Home Home > GIT Browse
summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorMike Marciniszyn <mike.marciniszyn@intel.com>2015-07-30 15:17:43 -0400
committerDoug Ledford <dledford@redhat.com>2015-08-28 22:59:36 -0400
commit7724105686e718ac476a6ad3304fea2fbcfcffde (patch)
tree743567a9714e95b0f803b74a7278c825e79d529f
parentd4ab347005fb26f414b98b2c8d5ef6de5778c3dc (diff)
IB/hfi1: add driver files
Signed-off-by: Andrew Friedley <andrew.friedley@intel.com> Signed-off-by: Arthur Kepner <arthur.kepner@intel.com> Signed-off-by: Brendan Cunningham <brendan.cunningham@intel.com> Signed-off-by: Brian Welty <brian.welty@intel.com> Signed-off-by: Caz Yokoyama <caz.yokoyama@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Harish Chegondi <harish.chegondi@intel.com> Signed-off-by: Ira Weiny <ira.weiny@intel.com> Signed-off-by: Jim Snow <jim.m.snow@intel.com> Signed-off-by: John Gregor <john.a.gregor@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Kaike Wan <kaike.wan@intel.com> Signed-off-by: Kevin Pine <kevin.pine@intel.com> Signed-off-by: Kyle Liddell <kyle.liddell@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Ravi Krishnaswamy <ravi.krishnaswamy@intel.com> Signed-off-by: Sadanand Warrier <sadanand.warrier@intel.com> Signed-off-by: Sanath Kumar <sanath.s.kumar@intel.com> Signed-off-by: Sudeep Dutt <sudeep.dutt@intel.com> Signed-off-by: Vlad Danushevsky <vladimir.danusevsky@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
-rw-r--r--Documentation/infiniband/sysfs.txt20
-rw-r--r--MAINTAINERS6
-rw-r--r--drivers/staging/rdma/Kconfig2
-rw-r--r--drivers/staging/rdma/Makefile1
-rw-r--r--drivers/staging/rdma/hfi1/Kconfig37
-rw-r--r--drivers/staging/rdma/hfi1/Makefile19
-rw-r--r--drivers/staging/rdma/hfi1/TODO6
-rw-r--r--drivers/staging/rdma/hfi1/chip.c10798
-rw-r--r--drivers/staging/rdma/hfi1/chip.h1035
-rw-r--r--drivers/staging/rdma/hfi1/chip_registers.h1289
-rw-r--r--drivers/staging/rdma/hfi1/common.h415
-rw-r--r--drivers/staging/rdma/hfi1/cq.c558
-rw-r--r--drivers/staging/rdma/hfi1/debugfs.c899
-rw-r--r--drivers/staging/rdma/hfi1/debugfs.h78
-rw-r--r--drivers/staging/rdma/hfi1/device.c142
-rw-r--r--drivers/staging/rdma/hfi1/device.h61
-rw-r--r--drivers/staging/rdma/hfi1/diag.c1873
-rw-r--r--drivers/staging/rdma/hfi1/dma.c186
-rw-r--r--drivers/staging/rdma/hfi1/driver.c1241
-rw-r--r--drivers/staging/rdma/hfi1/eprom.c475
-rw-r--r--drivers/staging/rdma/hfi1/eprom.h55
-rw-r--r--drivers/staging/rdma/hfi1/file_ops.c2140
-rw-r--r--drivers/staging/rdma/hfi1/firmware.c1620
-rw-r--r--drivers/staging/rdma/hfi1/hfi.h1821
-rw-r--r--drivers/staging/rdma/hfi1/init.c1722
-rw-r--r--drivers/staging/rdma/hfi1/intr.c207
-rw-r--r--drivers/staging/rdma/hfi1/iowait.h186
-rw-r--r--drivers/staging/rdma/hfi1/keys.c411
-rw-r--r--drivers/staging/rdma/hfi1/mad.c4257
-rw-r--r--drivers/staging/rdma/hfi1/mad.h325
-rw-r--r--drivers/staging/rdma/hfi1/mmap.c192
-rw-r--r--drivers/staging/rdma/hfi1/mr.c546
-rw-r--r--drivers/staging/rdma/hfi1/opa_compat.h129
-rw-r--r--drivers/staging/rdma/hfi1/pcie.c1253
-rw-r--r--drivers/staging/rdma/hfi1/pio.c1771
-rw-r--r--drivers/staging/rdma/hfi1/pio.h224
-rw-r--r--drivers/staging/rdma/hfi1/pio_copy.c858
-rw-r--r--drivers/staging/rdma/hfi1/platform_config.h286
-rw-r--r--drivers/staging/rdma/hfi1/qp.c1687
-rw-r--r--drivers/staging/rdma/hfi1/qp.h235
-rw-r--r--drivers/staging/rdma/hfi1/qsfp.c546
-rw-r--r--drivers/staging/rdma/hfi1/qsfp.h222
-rw-r--r--drivers/staging/rdma/hfi1/rc.c2426
-rw-r--r--drivers/staging/rdma/hfi1/ruc.c948
-rw-r--r--drivers/staging/rdma/hfi1/sdma.c2962
-rw-r--r--drivers/staging/rdma/hfi1/sdma.h1123
-rw-r--r--drivers/staging/rdma/hfi1/srq.c397
-rw-r--r--drivers/staging/rdma/hfi1/sysfs.c739
-rw-r--r--drivers/staging/rdma/hfi1/trace.c221
-rw-r--r--drivers/staging/rdma/hfi1/trace.h1409
-rw-r--r--drivers/staging/rdma/hfi1/twsi.c518
-rw-r--r--drivers/staging/rdma/hfi1/twsi.h68
-rw-r--r--drivers/staging/rdma/hfi1/uc.c585
-rw-r--r--drivers/staging/rdma/hfi1/ud.c885
-rw-r--r--drivers/staging/rdma/hfi1/user_pages.c156
-rw-r--r--drivers/staging/rdma/hfi1/user_sdma.c1444
-rw-r--r--drivers/staging/rdma/hfi1/user_sdma.h89
-rw-r--r--drivers/staging/rdma/hfi1/verbs.c2142
-rw-r--r--drivers/staging/rdma/hfi1/verbs.h1149
-rw-r--r--drivers/staging/rdma/hfi1/verbs_mcast.c385
60 files changed, 57480 insertions, 0 deletions
diff --git a/Documentation/infiniband/sysfs.txt b/Documentation/infiniband/sysfs.txt
index ddd519b72ee1..9028b025501a 100644
--- a/Documentation/infiniband/sysfs.txt
+++ b/Documentation/infiniband/sysfs.txt
@@ -64,3 +64,23 @@ MTHCA
fw_ver - Firmware version
hca_type - HCA type: "MT23108", "MT25208 (MT23108 compat mode)",
or "MT25208"
+
+HFI1
+
+ The hfi1 driver also creates these additional files:
+
+ hw_rev - hardware revision
+ board_id - manufacturing board id
+ tempsense - thermal sense information
+ serial - board serial number
+ nfreectxts - number of free user contexts
+ nctxts - number of allowed contexts (PSM2)
+ chip_reset - diagnostic (root only)
+ boardversion - board version
+ ports/1/
+ CMgtA/
+ cc_settings_bin - CCA tables used by PSM2
+ cc_table_bin
+ sc2v/ - 32 files (0 - 31) used to translate sl->vl
+ sl2sc/ - 32 files (0 - 31) used to translate sl->sc
+ vl2mtu/ - 16 (0 - 15) files used to determine MTU for vl
diff --git a/MAINTAINERS b/MAINTAINERS
index db1a523ed493..2249ac889076 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -9809,6 +9809,12 @@ M: Arnaud Patard <arnaud.patard@rtp-net.org>
S: Odd Fixes
F: drivers/staging/xgifb/
+HFI1 DRIVER
+M: Mike Marciniszyn <infinipath@intel.com>
+L: linux-rdma@vger.kernel.org
+S: Supported
+F: drivers/staging/rdma/hfi1
+
STARFIRE/DURALAN NETWORK DRIVER
M: Ion Badulescu <ionut@badula.org>
S: Odd Fixes
diff --git a/drivers/staging/rdma/Kconfig b/drivers/staging/rdma/Kconfig
index 5084088c8946..cf5fe9bb87a1 100644
--- a/drivers/staging/rdma/Kconfig
+++ b/drivers/staging/rdma/Kconfig
@@ -24,6 +24,8 @@ if STAGING_RDMA
source "drivers/staging/rdma/amso1100/Kconfig"
+source "drivers/staging/rdma/hfi1/Kconfig"
+
source "drivers/staging/rdma/ipath/Kconfig"
endif
diff --git a/drivers/staging/rdma/Makefile b/drivers/staging/rdma/Makefile
index a2a459ac8d67..cbd915ac7f20 100644
--- a/drivers/staging/rdma/Makefile
+++ b/drivers/staging/rdma/Makefile
@@ -1,3 +1,4 @@
# Entries for RDMA_STAGING tree
obj-$(CONFIG_INFINIBAND_AMSO1100) += amso1100/
+obj-$(CONFIG_INFINIBAND_HFI1) += hfi1/
obj-$(CONFIG_INFINIBAND_IPATH) += ipath/
diff --git a/drivers/staging/rdma/hfi1/Kconfig b/drivers/staging/rdma/hfi1/Kconfig
new file mode 100644
index 000000000000..fd25078ee923
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/Kconfig
@@ -0,0 +1,37 @@
+config INFINIBAND_HFI1
+ tristate "Intel OPA Gen1 support"
+ depends on X86_64
+ default m
+ ---help---
+ This is a low-level driver for Intel OPA Gen1 adapter.
+config HFI1_DEBUG_SDMA_ORDER
+ bool "HFI1 SDMA Order debug"
+ depends on INFINIBAND_HFI1
+ default n
+ ---help---
+ This is a debug flag to test for out of order
+ sdma completions for unit testing
+config HFI1_VERBS_31BIT_PSN
+ bool "HFI1 enable 31 bit PSN"
+ depends on INFINIBAND_HFI1
+ default y
+ ---help---
+ Setting this enables 31 BIT PSN
+ For verbs RC/UC
+config SDMA_VERBOSITY
+ bool "Config SDMA Verbosity"
+ depends on INFINIBAND_HFI1
+ default n
+ ---help---
+ This is a configuration flag to enable verbose
+ SDMA debug
+config PRESCAN_RXQ
+ bool "Enable prescanning of the RX queue for ECNs"
+ depends on INFINIBAND_HFI1
+ default n
+ ---help---
+ This option toggles the prescanning of the receive queue for
+ Explicit Congestion Notifications. If an ECN is detected, it
+ is processed as quickly as possible, the ECN is toggled off.
+ After the prescanning step, the receive queue is processed as
+ usual.
diff --git a/drivers/staging/rdma/hfi1/Makefile b/drivers/staging/rdma/hfi1/Makefile
new file mode 100644
index 000000000000..2e5daa6cdcc2
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/Makefile
@@ -0,0 +1,19 @@
+#
+# HFI driver
+#
+#
+#
+# Called from the kernel module build system.
+#
+obj-$(CONFIG_INFINIBAND_HFI1) += hfi1.o
+
+hfi1-y := chip.o cq.o device.o diag.o dma.o driver.o eprom.o file_ops.o firmware.o \
+ init.o intr.o keys.o mad.o mmap.o mr.o pcie.o pio.o pio_copy.o \
+ qp.o qsfp.o rc.o ruc.o sdma.o srq.o sysfs.o trace.o twsi.o \
+ uc.o ud.o user_pages.o user_sdma.o verbs_mcast.o verbs.o
+hfi1-$(CONFIG_DEBUG_FS) += debugfs.o
+
+CFLAGS_trace.o = -I$(src)
+ifdef MVERSION
+CFLAGS_driver.o = -DHFI_DRIVER_VERSION_BASE=\"$(MVERSION)\"
+endif
diff --git a/drivers/staging/rdma/hfi1/TODO b/drivers/staging/rdma/hfi1/TODO
new file mode 100644
index 000000000000..05de0dad8762
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/TODO
@@ -0,0 +1,6 @@
+July, 2015
+
+- Remove unneeded file entries in sysfs
+- Remove software processing of IB protocol and place in library for use
+ by qib, ipath (if still present), hfi1, and eventually soft-roce
+
diff --git a/drivers/staging/rdma/hfi1/chip.c b/drivers/staging/rdma/hfi1/chip.c
new file mode 100644
index 000000000000..654eafef1d30
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/chip.c
@@ -0,0 +1,10798 @@
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+/*
+ * This file contains all of the code that is specific to the HFI chip
+ */
+
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+
+#include "hfi.h"
+#include "trace.h"
+#include "mad.h"
+#include "pio.h"
+#include "sdma.h"
+#include "eprom.h"
+
+#define NUM_IB_PORTS 1
+
+uint kdeth_qp;
+module_param_named(kdeth_qp, kdeth_qp, uint, S_IRUGO);
+MODULE_PARM_DESC(kdeth_qp, "Set the KDETH queue pair prefix");
+
+uint num_vls = HFI1_MAX_VLS_SUPPORTED;
+module_param(num_vls, uint, S_IRUGO);
+MODULE_PARM_DESC(num_vls, "Set number of Virtual Lanes to use (1-8)");
+
+/*
+ * Default time to aggregate two 10K packets from the idle state
+ * (timer not running). The timer starts at the end of the first packet,
+ * so only the time for one 10K packet and header plus a bit extra is needed.
+ * 10 * 1024 + 64 header byte = 10304 byte
+ * 10304 byte / 12.5 GB/s = 824.32ns
+ */
+uint rcv_intr_timeout = (824 + 16); /* 16 is for coalescing interrupt */
+module_param(rcv_intr_timeout, uint, S_IRUGO);
+MODULE_PARM_DESC(rcv_intr_timeout, "Receive interrupt mitigation timeout in ns");
+
+uint rcv_intr_count = 16; /* same as qib */
+module_param(rcv_intr_count, uint, S_IRUGO);
+MODULE_PARM_DESC(rcv_intr_count, "Receive interrupt mitigation count");
+
+ushort link_crc_mask = SUPPORTED_CRCS;
+module_param(link_crc_mask, ushort, S_IRUGO);
+MODULE_PARM_DESC(link_crc_mask, "CRCs to use on the link");
+
+uint loopback;
+module_param_named(loopback, loopback, uint, S_IRUGO);
+MODULE_PARM_DESC(loopback, "Put into loopback mode (1 = serdes, 3 = external cable");
+
+/* Other driver tunables */
+uint rcv_intr_dynamic = 1; /* enable dynamic mode for rcv int mitigation*/
+static ushort crc_14b_sideband = 1;
+static uint use_flr = 1;
+uint quick_linkup; /* skip LNI */
+
+struct flag_table {
+ u64 flag; /* the flag */
+ char *str; /* description string */
+ u16 extra; /* extra information */
+ u16 unused0;
+ u32 unused1;
+};
+
+/* str must be a string constant */
+#define FLAG_ENTRY(str, extra, flag) {flag, str, extra}
+#define FLAG_ENTRY0(str, flag) {flag, str, 0}
+
+/* Send Error Consequences */
+#define SEC_WRITE_DROPPED 0x1
+#define SEC_PACKET_DROPPED 0x2
+#define SEC_SC_HALTED 0x4 /* per-context only */
+#define SEC_SPC_FREEZE 0x8 /* per-HFI only */
+
+#define VL15CTXT 1
+#define MIN_KERNEL_KCTXTS 2
+#define NUM_MAP_REGS 32
+
+/* Bit offset into the GUID which carries HFI id information */
+#define GUID_HFI_INDEX_SHIFT 39
+
+/* extract the emulation revision */
+#define emulator_rev(dd) ((dd)->irev >> 8)
+/* parallel and serial emulation versions are 3 and 4 respectively */
+#define is_emulator_p(dd) ((((dd)->irev) & 0xf) == 3)
+#define is_emulator_s(dd) ((((dd)->irev) & 0xf) == 4)
+
+/* RSM fields */
+
+/* packet type */
+#define IB_PACKET_TYPE 2ull
+#define QW_SHIFT 6ull
+/* QPN[7..1] */
+#define QPN_WIDTH 7ull
+
+/* LRH.BTH: QW 0, OFFSET 48 - for match */
+#define LRH_BTH_QW 0ull
+#define LRH_BTH_BIT_OFFSET 48ull
+#define LRH_BTH_OFFSET(off) ((LRH_BTH_QW << QW_SHIFT) | (off))
+#define LRH_BTH_MATCH_OFFSET LRH_BTH_OFFSET(LRH_BTH_BIT_OFFSET)
+#define LRH_BTH_SELECT
+#define LRH_BTH_MASK 3ull
+#define LRH_BTH_VALUE 2ull
+
+/* LRH.SC[3..0] QW 0, OFFSET 56 - for match */
+#define LRH_SC_QW 0ull
+#define LRH_SC_BIT_OFFSET 56ull
+#define LRH_SC_OFFSET(off) ((LRH_SC_QW << QW_SHIFT) | (off))
+#define LRH_SC_MATCH_OFFSET LRH_SC_OFFSET(LRH_SC_BIT_OFFSET)
+#define LRH_SC_MASK 128ull
+#define LRH_SC_VALUE 0ull
+
+/* SC[n..0] QW 0, OFFSET 60 - for select */
+#define LRH_SC_SELECT_OFFSET ((LRH_SC_QW << QW_SHIFT) | (60ull))
+
+/* QPN[m+n:1] QW 1, OFFSET 1 */
+#define QPN_SELECT_OFFSET ((1ull << QW_SHIFT) | (1ull))
+
+/* defines to build power on SC2VL table */
+#define SC2VL_VAL( \
+ num, \
+ sc0, sc0val, \
+ sc1, sc1val, \
+ sc2, sc2val, \
+ sc3, sc3val, \
+ sc4, sc4val, \
+ sc5, sc5val, \
+ sc6, sc6val, \
+ sc7, sc7val) \
+( \
+ ((u64)(sc0val) << SEND_SC2VLT##num##_SC##sc0##_SHIFT) | \
+ ((u64)(sc1val) << SEND_SC2VLT##num##_SC##sc1##_SHIFT) | \
+ ((u64)(sc2val) << SEND_SC2VLT##num##_SC##sc2##_SHIFT) | \
+ ((u64)(sc3val) << SEND_SC2VLT##num##_SC##sc3##_SHIFT) | \
+ ((u64)(sc4val) << SEND_SC2VLT##num##_SC##sc4##_SHIFT) | \
+ ((u64)(sc5val) << SEND_SC2VLT##num##_SC##sc5##_SHIFT) | \
+ ((u64)(sc6val) << SEND_SC2VLT##num##_SC##sc6##_SHIFT) | \
+ ((u64)(sc7val) << SEND_SC2VLT##num##_SC##sc7##_SHIFT) \
+)
+
+#define DC_SC_VL_VAL( \
+ range, \
+ e0, e0val, \
+ e1, e1val, \
+ e2, e2val, \
+ e3, e3val, \
+ e4, e4val, \
+ e5, e5val, \
+ e6, e6val, \
+ e7, e7val, \
+ e8, e8val, \
+ e9, e9val, \
+ e10, e10val, \
+ e11, e11val, \
+ e12, e12val, \
+ e13, e13val, \
+ e14, e14val, \
+ e15, e15val) \
+( \
+ ((u64)(e0val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e0##_SHIFT) | \
+ ((u64)(e1val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e1##_SHIFT) | \
+ ((u64)(e2val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e2##_SHIFT) | \
+ ((u64)(e3val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e3##_SHIFT) | \
+ ((u64)(e4val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e4##_SHIFT) | \
+ ((u64)(e5val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e5##_SHIFT) | \
+ ((u64)(e6val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e6##_SHIFT) | \
+ ((u64)(e7val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e7##_SHIFT) | \
+ ((u64)(e8val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e8##_SHIFT) | \
+ ((u64)(e9val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e9##_SHIFT) | \
+ ((u64)(e10val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e10##_SHIFT) | \
+ ((u64)(e11val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e11##_SHIFT) | \
+ ((u64)(e12val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e12##_SHIFT) | \
+ ((u64)(e13val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e13##_SHIFT) | \
+ ((u64)(e14val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e14##_SHIFT) | \
+ ((u64)(e15val) << DCC_CFG_SC_VL_TABLE_##range##_ENTRY##e15##_SHIFT) \
+)
+
+/* all CceStatus sub-block freeze bits */
+#define ALL_FROZE (CCE_STATUS_SDMA_FROZE_SMASK \
+ | CCE_STATUS_RXE_FROZE_SMASK \
+ | CCE_STATUS_TXE_FROZE_SMASK \
+ | CCE_STATUS_TXE_PIO_FROZE_SMASK)
+/* all CceStatus sub-block TXE pause bits */
+#define ALL_TXE_PAUSE (CCE_STATUS_TXE_PIO_PAUSED_SMASK \
+ | CCE_STATUS_TXE_PAUSED_SMASK \
+ | CCE_STATUS_SDMA_PAUSED_SMASK)
+/* all CceStatus sub-block RXE pause bits */
+#define ALL_RXE_PAUSE CCE_STATUS_RXE_PAUSED_SMASK
+
+/*
+ * CCE Error flags.
+ */
+static struct flag_table cce_err_status_flags[] = {
+/* 0*/ FLAG_ENTRY0("CceCsrParityErr",
+ CCE_ERR_STATUS_CCE_CSR_PARITY_ERR_SMASK),
+/* 1*/ FLAG_ENTRY0("CceCsrReadBadAddrErr",
+ CCE_ERR_STATUS_CCE_CSR_READ_BAD_ADDR_ERR_SMASK),
+/* 2*/ FLAG_ENTRY0("CceCsrWriteBadAddrErr",
+ CCE_ERR_STATUS_CCE_CSR_WRITE_BAD_ADDR_ERR_SMASK),
+/* 3*/ FLAG_ENTRY0("CceTrgtAsyncFifoParityErr",
+ CCE_ERR_STATUS_CCE_TRGT_ASYNC_FIFO_PARITY_ERR_SMASK),
+/* 4*/ FLAG_ENTRY0("CceTrgtAccessErr",
+ CCE_ERR_STATUS_CCE_TRGT_ACCESS_ERR_SMASK),
+/* 5*/ FLAG_ENTRY0("CceRspdDataParityErr",
+ CCE_ERR_STATUS_CCE_RSPD_DATA_PARITY_ERR_SMASK),
+/* 6*/ FLAG_ENTRY0("CceCli0AsyncFifoParityErr",
+ CCE_ERR_STATUS_CCE_CLI0_ASYNC_FIFO_PARITY_ERR_SMASK),
+/* 7*/ FLAG_ENTRY0("CceCsrCfgBusParityErr",
+ CCE_ERR_STATUS_CCE_CSR_CFG_BUS_PARITY_ERR_SMASK),
+/* 8*/ FLAG_ENTRY0("CceCli2AsyncFifoParityErr",
+ CCE_ERR_STATUS_CCE_CLI2_ASYNC_FIFO_PARITY_ERR_SMASK),
+/* 9*/ FLAG_ENTRY0("CceCli1AsyncFifoPioCrdtParityErr",
+ CCE_ERR_STATUS_CCE_CLI1_ASYNC_FIFO_PIO_CRDT_PARITY_ERR_SMASK),
+/*10*/ FLAG_ENTRY0("CceCli1AsyncFifoPioCrdtParityErr",
+ CCE_ERR_STATUS_CCE_CLI1_ASYNC_FIFO_SDMA_HD_PARITY_ERR_SMASK),
+/*11*/ FLAG_ENTRY0("CceCli1AsyncFifoRxdmaParityError",
+ CCE_ERR_STATUS_CCE_CLI1_ASYNC_FIFO_RXDMA_PARITY_ERROR_SMASK),
+/*12*/ FLAG_ENTRY0("CceCli1AsyncFifoDbgParityError",
+ CCE_ERR_STATUS_CCE_CLI1_ASYNC_FIFO_DBG_PARITY_ERROR_SMASK),
+/*13*/ FLAG_ENTRY0("PcicRetryMemCorErr",
+ CCE_ERR_STATUS_PCIC_RETRY_MEM_COR_ERR_SMASK),
+/*14*/ FLAG_ENTRY0("PcicRetryMemCorErr",
+ CCE_ERR_STATUS_PCIC_RETRY_SOT_MEM_COR_ERR_SMASK),
+/*15*/ FLAG_ENTRY0("PcicPostHdQCorErr",
+ CCE_ERR_STATUS_PCIC_POST_HD_QCOR_ERR_SMASK),
+/*16*/ FLAG_ENTRY0("PcicPostHdQCorErr",
+ CCE_ERR_STATUS_PCIC_POST_DAT_QCOR_ERR_SMASK),
+/*17*/ FLAG_ENTRY0("PcicPostHdQCorErr",
+ CCE_ERR_STATUS_PCIC_CPL_HD_QCOR_ERR_SMASK),
+/*18*/ FLAG_ENTRY0("PcicCplDatQCorErr",
+ CCE_ERR_STATUS_PCIC_CPL_DAT_QCOR_ERR_SMASK),
+/*19*/ FLAG_ENTRY0("PcicNPostHQParityErr",
+ CCE_ERR_STATUS_PCIC_NPOST_HQ_PARITY_ERR_SMASK),
+/*20*/ FLAG_ENTRY0("PcicNPostDatQParityErr",
+ CCE_ERR_STATUS_PCIC_NPOST_DAT_QPARITY_ERR_SMASK),
+/*21*/ FLAG_ENTRY0("PcicRetryMemUncErr",
+ CCE_ERR_STATUS_PCIC_RETRY_MEM_UNC_ERR_SMASK),
+/*22*/ FLAG_ENTRY0("PcicRetrySotMemUncErr",
+ CCE_ERR_STATUS_PCIC_RETRY_SOT_MEM_UNC_ERR_SMASK),
+/*23*/ FLAG_ENTRY0("PcicPostHdQUncErr",
+ CCE_ERR_STATUS_PCIC_POST_HD_QUNC_ERR_SMASK),
+/*24*/ FLAG_ENTRY0("PcicPostDatQUncErr",
+ CCE_ERR_STATUS_PCIC_POST_DAT_QUNC_ERR_SMASK),
+/*25*/ FLAG_ENTRY0("PcicCplHdQUncErr",
+ CCE_ERR_STATUS_PCIC_CPL_HD_QUNC_ERR_SMASK),
+/*26*/ FLAG_ENTRY0("PcicCplDatQUncErr",
+ CCE_ERR_STATUS_PCIC_CPL_DAT_QUNC_ERR_SMASK),
+/*27*/ FLAG_ENTRY0("PcicTransmitFrontParityErr",
+ CCE_ERR_STATUS_PCIC_TRANSMIT_FRONT_PARITY_ERR_SMASK),
+/*28*/ FLAG_ENTRY0("PcicTransmitBackParityErr",
+ CCE_ERR_STATUS_PCIC_TRANSMIT_BACK_PARITY_ERR_SMASK),
+/*29*/ FLAG_ENTRY0("PcicReceiveParityErr",
+ CCE_ERR_STATUS_PCIC_RECEIVE_PARITY_ERR_SMASK),
+/*30*/ FLAG_ENTRY0("CceTrgtCplTimeoutErr",
+ CCE_ERR_STATUS_CCE_TRGT_CPL_TIMEOUT_ERR_SMASK),
+/*31*/ FLAG_ENTRY0("LATriggered",
+ CCE_ERR_STATUS_LA_TRIGGERED_SMASK),
+/*32*/ FLAG_ENTRY0("CceSegReadBadAddrErr",
+ CCE_ERR_STATUS_CCE_SEG_READ_BAD_ADDR_ERR_SMASK),
+/*33*/ FLAG_ENTRY0("CceSegWriteBadAddrErr",
+ CCE_ERR_STATUS_CCE_SEG_WRITE_BAD_ADDR_ERR_SMASK),
+/*34*/ FLAG_ENTRY0("CceRcplAsyncFifoParityErr",
+ CCE_ERR_STATUS_CCE_RCPL_ASYNC_FIFO_PARITY_ERR_SMASK),
+/*35*/ FLAG_ENTRY0("CceRxdmaConvFifoParityErr",
+ CCE_ERR_STATUS_CCE_RXDMA_CONV_FIFO_PARITY_ERR_SMASK),
+/*36*/ FLAG_ENTRY0("CceMsixTableCorErr",
+ CCE_ERR_STATUS_CCE_MSIX_TABLE_COR_ERR_SMASK),
+/*37*/ FLAG_ENTRY0("CceMsixTableUncErr",
+ CCE_ERR_STATUS_CCE_MSIX_TABLE_UNC_ERR_SMASK),
+/*38*/ FLAG_ENTRY0("CceIntMapCorErr",
+ CCE_ERR_STATUS_CCE_INT_MAP_COR_ERR_SMASK),
+/*39*/ FLAG_ENTRY0("CceIntMapUncErr",
+ CCE_ERR_STATUS_CCE_INT_MAP_UNC_ERR_SMASK),
+/*40*/ FLAG_ENTRY0("CceMsixCsrParityErr",
+ CCE_ERR_STATUS_CCE_MSIX_CSR_PARITY_ERR_SMASK),
+/*41-63 reserved*/
+};
+
+/*
+ * Misc Error flags
+ */
+#define MES(text) MISC_ERR_STATUS_MISC_##text##_ERR_SMASK
+static struct flag_table misc_err_status_flags[] = {
+/* 0*/ FLAG_ENTRY0("CSR_PARITY", MES(CSR_PARITY)),
+/* 1*/ FLAG_ENTRY0("CSR_READ_BAD_ADDR", MES(CSR_READ_BAD_ADDR)),
+/* 2*/ FLAG_ENTRY0("CSR_WRITE_BAD_ADDR", MES(CSR_WRITE_BAD_ADDR)),
+/* 3*/ FLAG_ENTRY0("SBUS_WRITE_FAILED", MES(SBUS_WRITE_FAILED)),
+/* 4*/ FLAG_ENTRY0("KEY_MISMATCH", MES(KEY_MISMATCH)),
+/* 5*/ FLAG_ENTRY0("FW_AUTH_FAILED", MES(FW_AUTH_FAILED)),
+/* 6*/ FLAG_ENTRY0("EFUSE_CSR_PARITY", MES(EFUSE_CSR_PARITY)),
+/* 7*/ FLAG_ENTRY0("EFUSE_READ_BAD_ADDR", MES(EFUSE_READ_BAD_ADDR)),
+/* 8*/ FLAG_ENTRY0("EFUSE_WRITE", MES(EFUSE_WRITE)),
+/* 9*/ FLAG_ENTRY0("EFUSE_DONE_PARITY", MES(EFUSE_DONE_PARITY)),
+/*10*/ FLAG_ENTRY0("INVALID_EEP_CMD", MES(INVALID_EEP_CMD)),
+/*11*/ FLAG_ENTRY0("MBIST_FAIL", MES(MBIST_FAIL)),
+/*12*/ FLAG_ENTRY0("PLL_LOCK_FAIL", MES(PLL_LOCK_FAIL))
+};
+
+/*
+ * TXE PIO Error flags and consequences
+ */
+static struct flag_table pio_err_status_flags[] = {
+/* 0*/ FLAG_ENTRY("PioWriteBadCtxt",
+ SEC_WRITE_DROPPED,
+ SEND_PIO_ERR_STATUS_PIO_WRITE_BAD_CTXT_ERR_SMASK),
+/* 1*/ FLAG_ENTRY("PioWriteAddrParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_WRITE_ADDR_PARITY_ERR_SMASK),
+/* 2*/ FLAG_ENTRY("PioCsrParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_CSR_PARITY_ERR_SMASK),
+/* 3*/ FLAG_ENTRY("PioSbMemFifo0",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_SB_MEM_FIFO0_ERR_SMASK),
+/* 4*/ FLAG_ENTRY("PioSbMemFifo1",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_SB_MEM_FIFO1_ERR_SMASK),
+/* 5*/ FLAG_ENTRY("PioPccFifoParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_PCC_FIFO_PARITY_ERR_SMASK),
+/* 6*/ FLAG_ENTRY("PioPecFifoParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_PEC_FIFO_PARITY_ERR_SMASK),
+/* 7*/ FLAG_ENTRY("PioSbrdctlCrrelParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_SBRDCTL_CRREL_PARITY_ERR_SMASK),
+/* 8*/ FLAG_ENTRY("PioSbrdctrlCrrelFifoParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_SBRDCTRL_CRREL_FIFO_PARITY_ERR_SMASK),
+/* 9*/ FLAG_ENTRY("PioPktEvictFifoParityErr",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_PKT_EVICT_FIFO_PARITY_ERR_SMASK),
+/*10*/ FLAG_ENTRY("PioSmPktResetParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_SM_PKT_RESET_PARITY_ERR_SMASK),
+/*11*/ FLAG_ENTRY("PioVlLenMemBank0Unc",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK0_UNC_ERR_SMASK),
+/*12*/ FLAG_ENTRY("PioVlLenMemBank1Unc",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK1_UNC_ERR_SMASK),
+/*13*/ FLAG_ENTRY("PioVlLenMemBank0Cor",
+ 0,
+ SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK0_COR_ERR_SMASK),
+/*14*/ FLAG_ENTRY("PioVlLenMemBank1Cor",
+ 0,
+ SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK1_COR_ERR_SMASK),
+/*15*/ FLAG_ENTRY("PioCreditRetFifoParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_CREDIT_RET_FIFO_PARITY_ERR_SMASK),
+/*16*/ FLAG_ENTRY("PioPpmcPblFifo",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_PPMC_PBL_FIFO_ERR_SMASK),
+/*17*/ FLAG_ENTRY("PioInitSmIn",
+ 0,
+ SEND_PIO_ERR_STATUS_PIO_INIT_SM_IN_ERR_SMASK),
+/*18*/ FLAG_ENTRY("PioPktEvictSmOrArbSm",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_PKT_EVICT_SM_OR_ARB_SM_ERR_SMASK),
+/*19*/ FLAG_ENTRY("PioHostAddrMemUnc",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_HOST_ADDR_MEM_UNC_ERR_SMASK),
+/*20*/ FLAG_ENTRY("PioHostAddrMemCor",
+ 0,
+ SEND_PIO_ERR_STATUS_PIO_HOST_ADDR_MEM_COR_ERR_SMASK),
+/*21*/ FLAG_ENTRY("PioWriteDataParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_WRITE_DATA_PARITY_ERR_SMASK),
+/*22*/ FLAG_ENTRY("PioStateMachine",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_STATE_MACHINE_ERR_SMASK),
+/*23*/ FLAG_ENTRY("PioWriteQwValidParity",
+ SEC_WRITE_DROPPED|SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_WRITE_QW_VALID_PARITY_ERR_SMASK),
+/*24*/ FLAG_ENTRY("PioBlockQwCountParity",
+ SEC_WRITE_DROPPED|SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_BLOCK_QW_COUNT_PARITY_ERR_SMASK),
+/*25*/ FLAG_ENTRY("PioVlfVlLenParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_VLF_VL_LEN_PARITY_ERR_SMASK),
+/*26*/ FLAG_ENTRY("PioVlfSopParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_VLF_SOP_PARITY_ERR_SMASK),
+/*27*/ FLAG_ENTRY("PioVlFifoParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_VL_FIFO_PARITY_ERR_SMASK),
+/*28*/ FLAG_ENTRY("PioPpmcBqcMemParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_PPMC_BQC_MEM_PARITY_ERR_SMASK),
+/*29*/ FLAG_ENTRY("PioPpmcSopLen",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_PPMC_SOP_LEN_ERR_SMASK),
+/*30-31 reserved*/
+/*32*/ FLAG_ENTRY("PioCurrentFreeCntParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_CURRENT_FREE_CNT_PARITY_ERR_SMASK),
+/*33*/ FLAG_ENTRY("PioLastReturnedCntParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_LAST_RETURNED_CNT_PARITY_ERR_SMASK),
+/*34*/ FLAG_ENTRY("PioPccSopHeadParity",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_PCC_SOP_HEAD_PARITY_ERR_SMASK),
+/*35*/ FLAG_ENTRY("PioPecSopHeadParityErr",
+ SEC_SPC_FREEZE,
+ SEND_PIO_ERR_STATUS_PIO_PEC_SOP_HEAD_PARITY_ERR_SMASK),
+/*36-63 reserved*/
+};
+
+/* TXE PIO errors that cause an SPC freeze */
+#define ALL_PIO_FREEZE_ERR \
+ (SEND_PIO_ERR_STATUS_PIO_WRITE_ADDR_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_CSR_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_SB_MEM_FIFO0_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_SB_MEM_FIFO1_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_PCC_FIFO_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_PEC_FIFO_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_SBRDCTL_CRREL_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_SBRDCTRL_CRREL_FIFO_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_PKT_EVICT_FIFO_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_SM_PKT_RESET_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK0_UNC_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK1_UNC_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_CREDIT_RET_FIFO_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_PPMC_PBL_FIFO_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_PKT_EVICT_SM_OR_ARB_SM_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_HOST_ADDR_MEM_UNC_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_WRITE_DATA_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_STATE_MACHINE_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_WRITE_QW_VALID_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_BLOCK_QW_COUNT_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_VLF_VL_LEN_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_VLF_SOP_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_VL_FIFO_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_PPMC_BQC_MEM_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_PPMC_SOP_LEN_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_CURRENT_FREE_CNT_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_LAST_RETURNED_CNT_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_PCC_SOP_HEAD_PARITY_ERR_SMASK \
+ | SEND_PIO_ERR_STATUS_PIO_PEC_SOP_HEAD_PARITY_ERR_SMASK)
+
+/*
+ * TXE SDMA Error flags
+ */
+static struct flag_table sdma_err_status_flags[] = {
+/* 0*/ FLAG_ENTRY0("SDmaRpyTagErr",
+ SEND_DMA_ERR_STATUS_SDMA_RPY_TAG_ERR_SMASK),
+/* 1*/ FLAG_ENTRY0("SDmaCsrParityErr",
+ SEND_DMA_ERR_STATUS_SDMA_CSR_PARITY_ERR_SMASK),
+/* 2*/ FLAG_ENTRY0("SDmaPcieReqTrackingUncErr",
+ SEND_DMA_ERR_STATUS_SDMA_PCIE_REQ_TRACKING_UNC_ERR_SMASK),
+/* 3*/ FLAG_ENTRY0("SDmaPcieReqTrackingCorErr",
+ SEND_DMA_ERR_STATUS_SDMA_PCIE_REQ_TRACKING_COR_ERR_SMASK),
+/*04-63 reserved*/
+};
+
+/* TXE SDMA errors that cause an SPC freeze */
+#define ALL_SDMA_FREEZE_ERR \
+ (SEND_DMA_ERR_STATUS_SDMA_RPY_TAG_ERR_SMASK \
+ | SEND_DMA_ERR_STATUS_SDMA_CSR_PARITY_ERR_SMASK \
+ | SEND_DMA_ERR_STATUS_SDMA_PCIE_REQ_TRACKING_UNC_ERR_SMASK)
+
+/*
+ * TXE Egress Error flags
+ */
+#define SEES(text) SEND_EGRESS_ERR_STATUS_##text##_ERR_SMASK
+static struct flag_table egress_err_status_flags[] = {
+/* 0*/ FLAG_ENTRY0("TxPktIntegrityMemCorErr", SEES(TX_PKT_INTEGRITY_MEM_COR)),
+/* 1*/ FLAG_ENTRY0("TxPktIntegrityMemUncErr", SEES(TX_PKT_INTEGRITY_MEM_UNC)),
+/* 2 reserved */
+/* 3*/ FLAG_ENTRY0("TxEgressFifoUnderrunOrParityErr",
+ SEES(TX_EGRESS_FIFO_UNDERRUN_OR_PARITY)),
+/* 4*/ FLAG_ENTRY0("TxLinkdownErr", SEES(TX_LINKDOWN)),
+/* 5*/ FLAG_ENTRY0("TxIncorrectLinkStateErr", SEES(TX_INCORRECT_LINK_STATE)),
+/* 6 reserved */
+/* 7*/ FLAG_ENTRY0("TxPioLaunchIntfParityErr",
+ SEES(TX_PIO_LAUNCH_INTF_PARITY)),
+/* 8*/ FLAG_ENTRY0("TxSdmaLaunchIntfParityErr",
+ SEES(TX_SDMA_LAUNCH_INTF_PARITY)),
+/* 9-10 reserved */
+/*11*/ FLAG_ENTRY0("TxSbrdCtlStateMachineParityErr",
+ SEES(TX_SBRD_CTL_STATE_MACHINE_PARITY)),
+/*12*/ FLAG_ENTRY0("TxIllegalVLErr", SEES(TX_ILLEGAL_VL)),
+/*13*/ FLAG_ENTRY0("TxLaunchCsrParityErr", SEES(TX_LAUNCH_CSR_PARITY)),
+/*14*/ FLAG_ENTRY0("TxSbrdCtlCsrParityErr", SEES(TX_SBRD_CTL_CSR_PARITY)),
+/*15*/ FLAG_ENTRY0("TxConfigParityErr", SEES(TX_CONFIG_PARITY)),
+/*16*/ FLAG_ENTRY0("TxSdma0DisallowedPacketErr",
+ SEES(TX_SDMA0_DISALLOWED_PACKET)),
+/*17*/ FLAG_ENTRY0("TxSdma1DisallowedPacketErr",
+ SEES(TX_SDMA1_DISALLOWED_PACKET)),
+/*18*/ FLAG_ENTRY0("TxSdma2DisallowedPacketErr",
+ SEES(TX_SDMA2_DISALLOWED_PACKET)),
+/*19*/ FLAG_ENTRY0("TxSdma3DisallowedPacketErr",
+ SEES(TX_SDMA3_DISALLOWED_PACKET)),
+/*20*/ FLAG_ENTRY0("TxSdma4DisallowedPacketErr",
+ SEES(TX_SDMA4_DISALLOWED_PACKET)),
+/*21*/ FLAG_ENTRY0("TxSdma5DisallowedPacketErr",
+ SEES(TX_SDMA5_DISALLOWED_PACKET)),
+/*22*/ FLAG_ENTRY0("TxSdma6DisallowedPacketErr",
+ SEES(TX_SDMA6_DISALLOWED_PACKET)),
+/*23*/ FLAG_ENTRY0("TxSdma7DisallowedPacketErr",
+ SEES(TX_SDMA7_DISALLOWED_PACKET)),
+/*24*/ FLAG_ENTRY0("TxSdma8DisallowedPacketErr",
+ SEES(TX_SDMA8_DISALLOWED_PACKET)),
+/*25*/ FLAG_ENTRY0("TxSdma9DisallowedPacketErr",
+ SEES(TX_SDMA9_DISALLOWED_PACKET)),
+/*26*/ FLAG_ENTRY0("TxSdma10DisallowedPacketErr",
+ SEES(TX_SDMA10_DISALLOWED_PACKET)),
+/*27*/ FLAG_ENTRY0("TxSdma11DisallowedPacketErr",
+ SEES(TX_SDMA11_DISALLOWED_PACKET)),
+/*28*/ FLAG_ENTRY0("TxSdma12DisallowedPacketErr",
+ SEES(TX_SDMA12_DISALLOWED_PACKET)),
+/*29*/ FLAG_ENTRY0("TxSdma13DisallowedPacketErr",
+ SEES(TX_SDMA13_DISALLOWED_PACKET)),
+/*30*/ FLAG_ENTRY0("TxSdma14DisallowedPacketErr",
+ SEES(TX_SDMA14_DISALLOWED_PACKET)),
+/*31*/ FLAG_ENTRY0("TxSdma15DisallowedPacketErr",
+ SEES(TX_SDMA15_DISALLOWED_PACKET)),
+/*32*/ FLAG_ENTRY0("TxLaunchFifo0UncOrParityErr",
+ SEES(TX_LAUNCH_FIFO0_UNC_OR_PARITY)),
+/*33*/ FLAG_ENTRY0("TxLaunchFifo1UncOrParityErr",
+ SEES(TX_LAUNCH_FIFO1_UNC_OR_PARITY)),
+/*34*/ FLAG_ENTRY0("TxLaunchFifo2UncOrParityErr",
+ SEES(TX_LAUNCH_FIFO2_UNC_OR_PARITY)),
+/*35*/ FLAG_ENTRY0("TxLaunchFifo3UncOrParityErr",
+ SEES(TX_LAUNCH_FIFO3_UNC_OR_PARITY)),
+/*36*/ FLAG_ENTRY0("TxLaunchFifo4UncOrParityErr",
+ SEES(TX_LAUNCH_FIFO4_UNC_OR_PARITY)),
+/*37*/ FLAG_ENTRY0("TxLaunchFifo5UncOrParityErr",
+ SEES(TX_LAUNCH_FIFO5_UNC_OR_PARITY)),
+/*38*/ FLAG_ENTRY0("TxLaunchFifo6UncOrParityErr",
+ SEES(TX_LAUNCH_FIFO6_UNC_OR_PARITY)),
+/*39*/ FLAG_ENTRY0("TxLaunchFifo7UncOrParityErr",
+ SEES(TX_LAUNCH_FIFO7_UNC_OR_PARITY)),
+/*40*/ FLAG_ENTRY0("TxLaunchFifo8UncOrParityErr",
+ SEES(TX_LAUNCH_FIFO8_UNC_OR_PARITY)),
+/*41*/ FLAG_ENTRY0("TxCreditReturnParityErr", SEES(TX_CREDIT_RETURN_PARITY)),
+/*42*/ FLAG_ENTRY0("TxSbHdrUncErr", SEES(TX_SB_HDR_UNC)),
+/*43*/ FLAG_ENTRY0("TxReadSdmaMemoryUncErr", SEES(TX_READ_SDMA_MEMORY_UNC)),
+/*44*/ FLAG_ENTRY0("TxReadPioMemoryUncErr", SEES(TX_READ_PIO_MEMORY_UNC)),
+/*45*/ FLAG_ENTRY0("TxEgressFifoUncErr", SEES(TX_EGRESS_FIFO_UNC)),
+/*46*/ FLAG_ENTRY0("TxHcrcInsertionErr", SEES(TX_HCRC_INSERTION)),
+/*47*/ FLAG_ENTRY0("TxCreditReturnVLErr", SEES(TX_CREDIT_RETURN_VL)),
+/*48*/ FLAG_ENTRY0("TxLaunchFifo0CorErr", SEES(TX_LAUNCH_FIFO0_COR)),
+/*49*/ FLAG_ENTRY0("TxLaunchFifo1CorErr", SEES(TX_LAUNCH_FIFO1_COR)),
+/*50*/ FLAG_ENTRY0("TxLaunchFifo2CorErr", SEES(TX_LAUNCH_FIFO2_COR)),
+/*51*/ FLAG_ENTRY0("TxLaunchFifo3CorErr", SEES(TX_LAUNCH_FIFO3_COR)),
+/*52*/ FLAG_ENTRY0("TxLaunchFifo4CorErr", SEES(TX_LAUNCH_FIFO4_COR)),
+/*53*/ FLAG_ENTRY0("TxLaunchFifo5CorErr", SEES(TX_LAUNCH_FIFO5_COR)),
+/*54*/ FLAG_ENTRY0("TxLaunchFifo6CorErr", SEES(TX_LAUNCH_FIFO6_COR)),
+/*55*/ FLAG_ENTRY0("TxLaunchFifo7CorErr", SEES(TX_LAUNCH_FIFO7_COR)),
+/*56*/ FLAG_ENTRY0("TxLaunchFifo8CorErr", SEES(TX_LAUNCH_FIFO8_COR)),
+/*57*/ FLAG_ENTRY0("TxCreditOverrunErr", SEES(TX_CREDIT_OVERRUN)),
+/*58*/ FLAG_ENTRY0("TxSbHdrCorErr", SEES(TX_SB_HDR_COR)),
+/*59*/ FLAG_ENTRY0("TxReadSdmaMemoryCorErr", SEES(TX_READ_SDMA_MEMORY_COR)),
+/*60*/ FLAG_ENTRY0("TxReadPioMemoryCorErr", SEES(TX_READ_PIO_MEMORY_COR)),
+/*61*/ FLAG_ENTRY0("TxEgressFifoCorErr", SEES(TX_EGRESS_FIFO_COR)),
+/*62*/ FLAG_ENTRY0("TxReadSdmaMemoryCsrUncErr",
+ SEES(TX_READ_SDMA_MEMORY_CSR_UNC)),
+/*63*/ FLAG_ENTRY0("TxReadPioMemoryCsrUncErr",
+ SEES(TX_READ_PIO_MEMORY_CSR_UNC)),
+};
+
+/*
+ * TXE Egress Error Info flags
+ */
+#define SEEI(text) SEND_EGRESS_ERR_INFO_##text##_ERR_SMASK
+static struct flag_table egress_err_info_flags[] = {
+/* 0*/ FLAG_ENTRY0("Reserved", 0ull),
+/* 1*/ FLAG_ENTRY0("VLErr", SEEI(VL)),
+/* 2*/ FLAG_ENTRY0("JobKeyErr", SEEI(JOB_KEY)),
+/* 3*/ FLAG_ENTRY0("JobKeyErr", SEEI(JOB_KEY)),
+/* 4*/ FLAG_ENTRY0("PartitionKeyErr", SEEI(PARTITION_KEY)),
+/* 5*/ FLAG_ENTRY0("SLIDErr", SEEI(SLID)),
+/* 6*/ FLAG_ENTRY0("OpcodeErr", SEEI(OPCODE)),
+/* 7*/ FLAG_ENTRY0("VLMappingErr", SEEI(VL_MAPPING)),
+/* 8*/ FLAG_ENTRY0("RawErr", SEEI(RAW)),
+/* 9*/ FLAG_ENTRY0("RawIPv6Err", SEEI(RAW_IPV6)),
+/*10*/ FLAG_ENTRY0("GRHErr", SEEI(GRH)),
+/*11*/ FLAG_ENTRY0("BypassErr", SEEI(BYPASS)),
+/*12*/ FLAG_ENTRY0("KDETHPacketsErr", SEEI(KDETH_PACKETS)),
+/*13*/ FLAG_ENTRY0("NonKDETHPacketsErr", SEEI(NON_KDETH_PACKETS)),
+/*14*/ FLAG_ENTRY0("TooSmallIBPacketsErr", SEEI(TOO_SMALL_IB_PACKETS)),
+/*15*/ FLAG_ENTRY0("TooSmallBypassPacketsErr", SEEI(TOO_SMALL_BYPASS_PACKETS)),
+/*16*/ FLAG_ENTRY0("PbcTestErr", SEEI(PBC_TEST)),
+/*17*/ FLAG_ENTRY0("BadPktLenErr", SEEI(BAD_PKT_LEN)),
+/*18*/ FLAG_ENTRY0("TooLongIBPacketErr", SEEI(TOO_LONG_IB_PACKET)),
+/*19*/ FLAG_ENTRY0("TooLongBypassPacketsErr", SEEI(TOO_LONG_BYPASS_PACKETS)),
+/*20*/ FLAG_ENTRY0("PbcStaticRateControlErr", SEEI(PBC_STATIC_RATE_CONTROL)),
+/*21*/ FLAG_ENTRY0("BypassBadPktLenErr", SEEI(BAD_PKT_LEN)),
+};
+
+/* TXE Egress errors that cause an SPC freeze */
+#define ALL_TXE_EGRESS_FREEZE_ERR \
+ (SEES(TX_EGRESS_FIFO_UNDERRUN_OR_PARITY) \
+ | SEES(TX_PIO_LAUNCH_INTF_PARITY) \
+ | SEES(TX_SDMA_LAUNCH_INTF_PARITY) \
+ | SEES(TX_SBRD_CTL_STATE_MACHINE_PARITY) \
+ | SEES(TX_LAUNCH_CSR_PARITY) \
+ | SEES(TX_SBRD_CTL_CSR_PARITY) \
+ | SEES(TX_CONFIG_PARITY) \
+ | SEES(TX_LAUNCH_FIFO0_UNC_OR_PARITY) \
+ | SEES(TX_LAUNCH_FIFO1_UNC_OR_PARITY) \
+ | SEES(TX_LAUNCH_FIFO2_UNC_OR_PARITY) \
+ | SEES(TX_LAUNCH_FIFO3_UNC_OR_PARITY) \
+ | SEES(TX_LAUNCH_FIFO4_UNC_OR_PARITY) \
+ | SEES(TX_LAUNCH_FIFO5_UNC_OR_PARITY) \
+ | SEES(TX_LAUNCH_FIFO6_UNC_OR_PARITY) \
+ | SEES(TX_LAUNCH_FIFO7_UNC_OR_PARITY) \
+ | SEES(TX_LAUNCH_FIFO8_UNC_OR_PARITY) \
+ | SEES(TX_CREDIT_RETURN_PARITY))
+
+/*
+ * TXE Send error flags
+ */
+#define SES(name) SEND_ERR_STATUS_SEND_##name##_ERR_SMASK
+static struct flag_table send_err_status_flags[] = {
+/* 0*/ FLAG_ENTRY0("SDmaRpyTagErr", SES(CSR_PARITY)),
+/* 1*/ FLAG_ENTRY0("SendCsrReadBadAddrErr", SES(CSR_READ_BAD_ADDR)),
+/* 2*/ FLAG_ENTRY0("SendCsrWriteBadAddrErr", SES(CSR_WRITE_BAD_ADDR))
+};
+
+/*
+ * TXE Send Context Error flags and consequences
+ */
+static struct flag_table sc_err_status_flags[] = {
+/* 0*/ FLAG_ENTRY("InconsistentSop",
+ SEC_PACKET_DROPPED | SEC_SC_HALTED,
+ SEND_CTXT_ERR_STATUS_PIO_INCONSISTENT_SOP_ERR_SMASK),
+/* 1*/ FLAG_ENTRY("DisallowedPacket",
+ SEC_PACKET_DROPPED | SEC_SC_HALTED,
+ SEND_CTXT_ERR_STATUS_PIO_DISALLOWED_PACKET_ERR_SMASK),
+/* 2*/ FLAG_ENTRY("WriteCrossesBoundary",
+ SEC_WRITE_DROPPED | SEC_SC_HALTED,
+ SEND_CTXT_ERR_STATUS_PIO_WRITE_CROSSES_BOUNDARY_ERR_SMASK),
+/* 3*/ FLAG_ENTRY("WriteOverflow",
+ SEC_WRITE_DROPPED | SEC_SC_HALTED,
+ SEND_CTXT_ERR_STATUS_PIO_WRITE_OVERFLOW_ERR_SMASK),
+/* 4*/ FLAG_ENTRY("WriteOutOfBounds",
+ SEC_WRITE_DROPPED | SEC_SC_HALTED,
+ SEND_CTXT_ERR_STATUS_PIO_WRITE_OUT_OF_BOUNDS_ERR_SMASK),
+/* 5-63 reserved*/
+};
+
+/*
+ * RXE Receive Error flags
+ */
+#define RXES(name) RCV_ERR_STATUS_RX_##name##_ERR_SMASK
+static struct flag_table rxe_err_status_flags[] = {
+/* 0*/ FLAG_ENTRY0("RxDmaCsrCorErr", RXES(DMA_CSR_COR)),
+/* 1*/ FLAG_ENTRY0("RxDcIntfParityErr", RXES(DC_INTF_PARITY)),
+/* 2*/ FLAG_ENTRY0("RxRcvHdrUncErr", RXES(RCV_HDR_UNC)),
+/* 3*/ FLAG_ENTRY0("RxRcvHdrCorErr", RXES(RCV_HDR_COR)),
+/* 4*/ FLAG_ENTRY0("RxRcvDataUncErr", RXES(RCV_DATA_UNC)),
+/* 5*/ FLAG_ENTRY0("RxRcvDataCorErr", RXES(RCV_DATA_COR)),
+/* 6*/ FLAG_ENTRY0("RxRcvQpMapTableUncErr", RXES(RCV_QP_MAP_TABLE_UNC)),
+/* 7*/ FLAG_ENTRY0("RxRcvQpMapTableCorErr", RXES(RCV_QP_MAP_TABLE_COR)),
+/* 8*/ FLAG_ENTRY0("RxRcvCsrParityErr", RXES(RCV_CSR_PARITY)),
+/* 9*/ FLAG_ENTRY0("RxDcSopEopParityErr", RXES(DC_SOP_EOP_PARITY)),
+/*10*/ FLAG_ENTRY0("RxDmaFlagUncErr", RXES(DMA_FLAG_UNC)),
+/*11*/ FLAG_ENTRY0("RxDmaFlagCorErr", RXES(DMA_FLAG_COR)),
+/*12*/ FLAG_ENTRY0("RxRcvFsmEncodingErr", RXES(RCV_FSM_ENCODING)),
+/*13*/ FLAG_ENTRY0("RxRbufFreeListUncErr", RXES(RBUF_FREE_LIST_UNC)),
+/*14*/ FLAG_ENTRY0("RxRbufFreeListCorErr", RXES(RBUF_FREE_LIST_COR)),
+/*15*/ FLAG_ENTRY0("RxRbufLookupDesRegUncErr", RXES(RBUF_LOOKUP_DES_REG_UNC)),
+/*16*/ FLAG_ENTRY0("RxRbufLookupDesRegUncCorErr",
+ RXES(RBUF_LOOKUP_DES_REG_UNC_COR)),
+/*17*/ FLAG_ENTRY0("RxRbufLookupDesUncErr", RXES(RBUF_LOOKUP_DES_UNC)),
+/*18*/ FLAG_ENTRY0("RxRbufLookupDesCorErr", RXES(RBUF_LOOKUP_DES_COR)),
+/*19*/ FLAG_ENTRY0("RxRbufBlockListReadUncErr",
+ RXES(RBUF_BLOCK_LIST_READ_UNC)),
+/*20*/ FLAG_ENTRY0("RxRbufBlockListReadCorErr",
+ RXES(RBUF_BLOCK_LIST_READ_COR)),
+/*21*/ FLAG_ENTRY0("RxRbufCsrQHeadBufNumParityErr",
+ RXES(RBUF_CSR_QHEAD_BUF_NUM_PARITY)),
+/*22*/ FLAG_ENTRY0("RxRbufCsrQEntCntParityErr",
+ RXES(RBUF_CSR_QENT_CNT_PARITY)),
+/*23*/ FLAG_ENTRY0("RxRbufCsrQNextBufParityErr",
+ RXES(RBUF_CSR_QNEXT_BUF_PARITY)),
+/*24*/ FLAG_ENTRY0("RxRbufCsrQVldBitParityErr",
+ RXES(RBUF_CSR_QVLD_BIT_PARITY)),
+/*25*/ FLAG_ENTRY0("RxRbufCsrQHdPtrParityErr", RXES(RBUF_CSR_QHD_PTR_PARITY)),
+/*26*/ FLAG_ENTRY0("RxRbufCsrQTlPtrParityErr", RXES(RBUF_CSR_QTL_PTR_PARITY)),
+/*27*/ FLAG_ENTRY0("RxRbufCsrQNumOfPktParityErr",
+ RXES(RBUF_CSR_QNUM_OF_PKT_PARITY)),
+/*28*/ FLAG_ENTRY0("RxRbufCsrQEOPDWParityErr", RXES(RBUF_CSR_QEOPDW_PARITY)),
+/*29*/ FLAG_ENTRY0("RxRbufCtxIdParityErr", RXES(RBUF_CTX_ID_PARITY)),
+/*30*/ FLAG_ENTRY0("RxRBufBadLookupErr", RXES(RBUF_BAD_LOOKUP)),
+/*31*/ FLAG_ENTRY0("RxRbufFullErr", RXES(RBUF_FULL)),
+/*32*/ FLAG_ENTRY0("RxRbufEmptyErr", RXES(RBUF_EMPTY)),
+/*33*/ FLAG_ENTRY0("RxRbufFlRdAddrParityErr", RXES(RBUF_FL_RD_ADDR_PARITY)),
+/*34*/ FLAG_ENTRY0("RxRbufFlWrAddrParityErr", RXES(RBUF_FL_WR_ADDR_PARITY)),
+/*35*/ FLAG_ENTRY0("RxRbufFlInitdoneParityErr",
+ RXES(RBUF_FL_INITDONE_PARITY)),
+/*36*/ FLAG_ENTRY0("RxRbufFlInitWrAddrParityErr",
+ RXES(RBUF_FL_INIT_WR_ADDR_PARITY)),
+/*37*/ FLAG_ENTRY0("RxRbufNextFreeBufUncErr", RXES(RBUF_NEXT_FREE_BUF_UNC)),
+/*38*/ FLAG_ENTRY0("RxRbufNextFreeBufCorErr", RXES(RBUF_NEXT_FREE_BUF_COR)),
+/*39*/ FLAG_ENTRY0("RxLookupDesPart1UncErr", RXES(LOOKUP_DES_PART1_UNC)),
+/*40*/ FLAG_ENTRY0("RxLookupDesPart1UncCorErr",
+ RXES(LOOKUP_DES_PART1_UNC_COR)),
+/*41*/ FLAG_ENTRY0("RxLookupDesPart2ParityErr",
+ RXES(LOOKUP_DES_PART2_PARITY)),
+/*42*/ FLAG_ENTRY0("RxLookupRcvArrayUncErr", RXES(LOOKUP_RCV_ARRAY_UNC)),
+/*43*/ FLAG_ENTRY0("RxLookupRcvArrayCorErr", RXES(LOOKUP_RCV_ARRAY_COR)),
+/*44*/ FLAG_ENTRY0("RxLookupCsrParityErr", RXES(LOOKUP_CSR_PARITY)),
+/*45*/ FLAG_ENTRY0("RxHqIntrCsrParityErr", RXES(HQ_INTR_CSR_PARITY)),
+/*46*/ FLAG_ENTRY0("RxHqIntrFsmErr", RXES(HQ_INTR_FSM)),
+/*47*/ FLAG_ENTRY0("RxRbufDescPart1UncErr", RXES(RBUF_DESC_PART1_UNC)),
+/*48*/ FLAG_ENTRY0("RxRbufDescPart1CorErr", RXES(RBUF_DESC_PART1_COR)),
+/*49*/ FLAG_ENTRY0("RxRbufDescPart2UncErr", RXES(RBUF_DESC_PART2_UNC)),
+/*50*/ FLAG_ENTRY0("RxRbufDescPart2CorErr", RXES(RBUF_DESC_PART2_COR)),
+/*51*/ FLAG_ENTRY0("RxDmaHdrFifoRdUncErr", RXES(DMA_HDR_FIFO_RD_UNC)),
+/*52*/ FLAG_ENTRY0("RxDmaHdrFifoRdCorErr", RXES(DMA_HDR_FIFO_RD_COR)),
+/*53*/ FLAG_ENTRY0("RxDmaDataFifoRdUncErr", RXES(DMA_DATA_FIFO_RD_UNC)),
+/*54*/ FLAG_ENTRY0("RxDmaDataFifoRdCorErr", RXES(DMA_DATA_FIFO_RD_COR)),
+/*55*/ FLAG_ENTRY0("RxRbufDataUncErr", RXES(RBUF_DATA_UNC)),
+/*56*/ FLAG_ENTRY0("RxRbufDataCorErr", RXES(RBUF_DATA_COR)),
+/*57*/ FLAG_ENTRY0("RxDmaCsrParityErr", RXES(DMA_CSR_PARITY)),
+/*58*/ FLAG_ENTRY0("RxDmaEqFsmEncodingErr", RXES(DMA_EQ_FSM_ENCODING)),
+/*59*/ FLAG_ENTRY0("RxDmaDqFsmEncodingErr", RXES(DMA_DQ_FSM_ENCODING)),
+/*60*/ FLAG_ENTRY0("RxDmaCsrUncErr", RXES(DMA_CSR_UNC)),
+/*61*/ FLAG_ENTRY0("RxCsrReadBadAddrErr", RXES(CSR_READ_BAD_ADDR)),
+/*62*/ FLAG_ENTRY0("RxCsrWriteBadAddrErr", RXES(CSR_WRITE_BAD_ADDR)),
+/*63*/ FLAG_ENTRY0("RxCsrParityErr", RXES(CSR_PARITY))
+};
+
+/* RXE errors that will trigger an SPC freeze */
+#define ALL_RXE_FREEZE_ERR \
+ (RCV_ERR_STATUS_RX_RCV_QP_MAP_TABLE_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RCV_CSR_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_DMA_FLAG_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RCV_FSM_ENCODING_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_FREE_LIST_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_LOOKUP_DES_REG_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_LOOKUP_DES_REG_UNC_COR_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_LOOKUP_DES_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_BLOCK_LIST_READ_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_CSR_QHEAD_BUF_NUM_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_CSR_QENT_CNT_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_CSR_QNEXT_BUF_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_CSR_QVLD_BIT_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_CSR_QHD_PTR_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_CSR_QTL_PTR_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_CSR_QNUM_OF_PKT_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_CSR_QEOPDW_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_CTX_ID_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_BAD_LOOKUP_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_FULL_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_EMPTY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_FL_RD_ADDR_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_FL_WR_ADDR_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_FL_INITDONE_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_FL_INIT_WR_ADDR_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_NEXT_FREE_BUF_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_LOOKUP_DES_PART1_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_LOOKUP_DES_PART1_UNC_COR_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_LOOKUP_DES_PART2_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_LOOKUP_RCV_ARRAY_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_LOOKUP_CSR_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_HQ_INTR_CSR_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_HQ_INTR_FSM_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_DESC_PART1_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_DESC_PART1_COR_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_DESC_PART2_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_DMA_HDR_FIFO_RD_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_DMA_DATA_FIFO_RD_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_RBUF_DATA_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_DMA_CSR_PARITY_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_DMA_EQ_FSM_ENCODING_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_DMA_DQ_FSM_ENCODING_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_DMA_CSR_UNC_ERR_SMASK \
+ | RCV_ERR_STATUS_RX_CSR_PARITY_ERR_SMASK)
+
+#define RXE_FREEZE_ABORT_MASK \
+ (RCV_ERR_STATUS_RX_DMA_CSR_UNC_ERR_SMASK | \
+ RCV_ERR_STATUS_RX_DMA_HDR_FIFO_RD_UNC_ERR_SMASK | \
+ RCV_ERR_STATUS_RX_DMA_DATA_FIFO_RD_UNC_ERR_SMASK)
+
+/*
+ * DCC Error Flags
+ */
+#define DCCE(name) DCC_ERR_FLG_##name##_SMASK
+static struct flag_table dcc_err_flags[] = {
+ FLAG_ENTRY0("bad_l2_err", DCCE(BAD_L2_ERR)),
+ FLAG_ENTRY0("bad_sc_err", DCCE(BAD_SC_ERR)),
+ FLAG_ENTRY0("bad_mid_tail_err", DCCE(BAD_MID_TAIL_ERR)),
+ FLAG_ENTRY0("bad_preemption_err", DCCE(BAD_PREEMPTION_ERR)),
+ FLAG_ENTRY0("preemption_err", DCCE(PREEMPTION_ERR)),
+ FLAG_ENTRY0("preemptionvl15_err", DCCE(PREEMPTIONVL15_ERR)),
+ FLAG_ENTRY0("bad_vl_marker_err", DCCE(BAD_VL_MARKER_ERR)),
+ FLAG_ENTRY0("bad_dlid_target_err", DCCE(BAD_DLID_TARGET_ERR)),
+ FLAG_ENTRY0("bad_lver_err", DCCE(BAD_LVER_ERR)),
+ FLAG_ENTRY0("uncorrectable_err", DCCE(UNCORRECTABLE_ERR)),
+ FLAG_ENTRY0("bad_crdt_ack_err", DCCE(BAD_CRDT_ACK_ERR)),
+ FLAG_ENTRY0("unsup_pkt_type", DCCE(UNSUP_PKT_TYPE)),
+ FLAG_ENTRY0("bad_ctrl_flit_err", DCCE(BAD_CTRL_FLIT_ERR)),
+ FLAG_ENTRY0("event_cntr_parity_err", DCCE(EVENT_CNTR_PARITY_ERR)),
+ FLAG_ENTRY0("event_cntr_rollover_err", DCCE(EVENT_CNTR_ROLLOVER_ERR)),
+ FLAG_ENTRY0("link_err", DCCE(LINK_ERR)),
+ FLAG_ENTRY0("misc_cntr_rollover_err", DCCE(MISC_CNTR_ROLLOVER_ERR)),
+ FLAG_ENTRY0("bad_ctrl_dist_err", DCCE(BAD_CTRL_DIST_ERR)),
+ FLAG_ENTRY0("bad_tail_dist_err", DCCE(BAD_TAIL_DIST_ERR)),
+ FLAG_ENTRY0("bad_head_dist_err", DCCE(BAD_HEAD_DIST_ERR)),
+ FLAG_ENTRY0("nonvl15_state_err", DCCE(NONVL15_STATE_ERR)),
+ FLAG_ENTRY0("vl15_multi_err", DCCE(VL15_MULTI_ERR)),
+ FLAG_ENTRY0("bad_pkt_length_err", DCCE(BAD_PKT_LENGTH_ERR)),
+ FLAG_ENTRY0("unsup_vl_err", DCCE(UNSUP_VL_ERR)),
+ FLAG_ENTRY0("perm_nvl15_err", DCCE(PERM_NVL15_ERR)),
+ FLAG_ENTRY0("slid_zero_err", DCCE(SLID_ZERO_ERR)),
+ FLAG_ENTRY0("dlid_zero_err", DCCE(DLID_ZERO_ERR)),
+ FLAG_ENTRY0("length_mtu_err", DCCE(LENGTH_MTU_ERR)),
+ FLAG_ENTRY0("rx_early_drop_err", DCCE(RX_EARLY_DROP_ERR)),
+ FLAG_ENTRY0("late_short_err", DCCE(LATE_SHORT_ERR)),
+ FLAG_ENTRY0("late_long_err", DCCE(LATE_LONG_ERR)),
+ FLAG_ENTRY0("late_ebp_err", DCCE(LATE_EBP_ERR)),
+ FLAG_ENTRY0("fpe_tx_fifo_ovflw_err", DCCE(FPE_TX_FIFO_OVFLW_ERR)),
+ FLAG_ENTRY0("fpe_tx_fifo_unflw_err", DCCE(FPE_TX_FIFO_UNFLW_ERR)),
+ FLAG_ENTRY0("csr_access_blocked_host", DCCE(CSR_ACCESS_BLOCKED_HOST)),
+ FLAG_ENTRY0("csr_access_blocked_uc", DCCE(CSR_ACCESS_BLOCKED_UC)),
+ FLAG_ENTRY0("tx_ctrl_parity_err", DCCE(TX_CTRL_PARITY_ERR)),
+ FLAG_ENTRY0("tx_ctrl_parity_mbe_err", DCCE(TX_CTRL_PARITY_MBE_ERR)),
+ FLAG_ENTRY0("tx_sc_parity_err", DCCE(TX_SC_PARITY_ERR)),
+ FLAG_ENTRY0("rx_ctrl_parity_mbe_err", DCCE(RX_CTRL_PARITY_MBE_ERR)),
+ FLAG_ENTRY0("csr_parity_err", DCCE(CSR_PARITY_ERR)),
+ FLAG_ENTRY0("csr_inval_addr", DCCE(CSR_INVAL_ADDR)),
+ FLAG_ENTRY0("tx_byte_shft_parity_err", DCCE(TX_BYTE_SHFT_PARITY_ERR)),
+ FLAG_ENTRY0("rx_byte_shft_parity_err", DCCE(RX_BYTE_SHFT_PARITY_ERR)),
+ FLAG_ENTRY0("fmconfig_err", DCCE(FMCONFIG_ERR)),
+ FLAG_ENTRY0("rcvport_err", DCCE(RCVPORT_ERR)),
+};
+
+/*
+ * LCB error flags
+ */
+#define LCBE(name) DC_LCB_ERR_FLG_##name##_SMASK
+static struct flag_table lcb_err_flags[] = {
+/* 0*/ FLAG_ENTRY0("CSR_PARITY_ERR", LCBE(CSR_PARITY_ERR)),
+/* 1*/ FLAG_ENTRY0("INVALID_CSR_ADDR", LCBE(INVALID_CSR_ADDR)),
+/* 2*/ FLAG_ENTRY0("RST_FOR_FAILED_DESKEW", LCBE(RST_FOR_FAILED_DESKEW)),
+/* 3*/ FLAG_ENTRY0("ALL_LNS_FAILED_REINIT_TEST",
+ LCBE(ALL_LNS_FAILED_REINIT_TEST)),
+/* 4*/ FLAG_ENTRY0("LOST_REINIT_STALL_OR_TOS", LCBE(LOST_REINIT_STALL_OR_TOS)),
+/* 5*/ FLAG_ENTRY0("TX_LESS_THAN_FOUR_LNS", LCBE(TX_LESS_THAN_FOUR_LNS)),
+/* 6*/ FLAG_ENTRY0("RX_LESS_THAN_FOUR_LNS", LCBE(RX_LESS_THAN_FOUR_LNS)),
+/* 7*/ FLAG_ENTRY0("SEQ_CRC_ERR", LCBE(SEQ_CRC_ERR)),
+/* 8*/ FLAG_ENTRY0("REINIT_FROM_PEER", LCBE(REINIT_FROM_PEER)),
+/* 9*/ FLAG_ENTRY0("REINIT_FOR_LN_DEGRADE", LCBE(REINIT_FOR_LN_DEGRADE)),
+/*10*/ FLAG_ENTRY0("CRC_ERR_CNT_HIT_LIMIT", LCBE(CRC_ERR_CNT_HIT_LIMIT)),
+/*11*/ FLAG_ENTRY0("RCLK_STOPPED", LCBE(RCLK_STOPPED)),
+/*12*/ FLAG_ENTRY0("UNEXPECTED_REPLAY_MARKER", LCBE(UNEXPECTED_REPLAY_MARKER)),
+/*13*/ FLAG_ENTRY0("UNEXPECTED_ROUND_TRIP_MARKER",
+ LCBE(UNEXPECTED_ROUND_TRIP_MARKER)),
+/*14*/ FLAG_ENTRY0("ILLEGAL_NULL_LTP", LCBE(ILLEGAL_NULL_LTP)),
+/*15*/ FLAG_ENTRY0("ILLEGAL_FLIT_ENCODING", LCBE(ILLEGAL_FLIT_ENCODING)),
+/*16*/ FLAG_ENTRY0("FLIT_INPUT_BUF_OFLW", LCBE(FLIT_INPUT_BUF_OFLW)),
+/*17*/ FLAG_ENTRY0("VL_ACK_INPUT_BUF_OFLW", LCBE(VL_ACK_INPUT_BUF_OFLW)),
+/*18*/ FLAG_ENTRY0("VL_ACK_INPUT_PARITY_ERR", LCBE(VL_ACK_INPUT_PARITY_ERR)),
+/*19*/ FLAG_ENTRY0("VL_ACK_INPUT_WRONG_CRC_MODE",
+ LCBE(VL_ACK_INPUT_WRONG_CRC_MODE)),
+/*20*/ FLAG_ENTRY0("FLIT_INPUT_BUF_MBE", LCBE(FLIT_INPUT_BUF_MBE)),
+/*21*/ FLAG_ENTRY0("FLIT_INPUT_BUF_SBE", LCBE(FLIT_INPUT_BUF_SBE)),
+/*22*/ FLAG_ENTRY0("REPLAY_BUF_MBE", LCBE(REPLAY_BUF_MBE)),
+/*23*/ FLAG_ENTRY0("REPLAY_BUF_SBE", LCBE(REPLAY_BUF_SBE)),
+/*24*/ FLAG_ENTRY0("CREDIT_RETURN_FLIT_MBE", LCBE(CREDIT_RETURN_FLIT_MBE)),
+/*25*/ FLAG_ENTRY0("RST_FOR_LINK_TIMEOUT", LCBE(RST_FOR_LINK_TIMEOUT)),
+/*26*/ FLAG_ENTRY0("RST_FOR_INCOMPLT_RND_TRIP",
+ LCBE(RST_FOR_INCOMPLT_RND_TRIP)),
+/*27*/ FLAG_ENTRY0("HOLD_REINIT", LCBE(HOLD_REINIT)),
+/*28*/ FLAG_ENTRY0("NEG_EDGE_LINK_TRANSFER_ACTIVE",
+ LCBE(NEG_EDGE_LINK_TRANSFER_ACTIVE)),
+/*29*/ FLAG_ENTRY0("REDUNDANT_FLIT_PARITY_ERR",
+ LCBE(REDUNDANT_FLIT_PARITY_ERR))
+};
+
+/*
+ * DC8051 Error Flags
+ */
+#define D8E(name) DC_DC8051_ERR_FLG_##name##_SMASK
+static struct flag_table dc8051_err_flags[] = {
+ FLAG_ENTRY0("SET_BY_8051", D8E(SET_BY_8051)),
+ FLAG_ENTRY0("LOST_8051_HEART_BEAT", D8E(LOST_8051_HEART_BEAT)),
+ FLAG_ENTRY0("CRAM_MBE", D8E(CRAM_MBE)),
+ FLAG_ENTRY0("CRAM_SBE", D8E(CRAM_SBE)),
+ FLAG_ENTRY0("DRAM_MBE", D8E(DRAM_MBE)),
+ FLAG_ENTRY0("DRAM_SBE", D8E(DRAM_SBE)),
+ FLAG_ENTRY0("IRAM_MBE", D8E(IRAM_MBE)),
+ FLAG_ENTRY0("IRAM_SBE", D8E(IRAM_SBE)),
+ FLAG_ENTRY0("UNMATCHED_SECURE_MSG_ACROSS_BCC_LANES",
+ D8E(UNMATCHED_SECURE_MSG_ACROSS_BCC_LANES)),
+ FLAG_ENTRY0("INVALID_CSR_ADDR", D8E(INVALID_CSR_ADDR)),
+};
+
+/*
+ * DC8051 Information Error flags
+ *
+ * Flags in DC8051_DBG_ERR_INFO_SET_BY_8051.ERROR field.
+ */
+static struct flag_table dc8051_info_err_flags[] = {
+ FLAG_ENTRY0("Spico ROM check failed", SPICO_ROM_FAILED),
+ FLAG_ENTRY0("Unknown frame received", UNKNOWN_FRAME),
+ FLAG_ENTRY0("Target BER not met", TARGET_BER_NOT_MET),
+ FLAG_ENTRY0("Serdes internal loopback failure",
+ FAILED_SERDES_INTERNAL_LOOPBACK),
+ FLAG_ENTRY0("Failed SerDes init", FAILED_SERDES_INIT),
+ FLAG_ENTRY0("Failed LNI(Polling)", FAILED_LNI_POLLING),
+ FLAG_ENTRY0("Failed LNI(Debounce)", FAILED_LNI_DEBOUNCE),
+ FLAG_ENTRY0("Failed LNI(EstbComm)", FAILED_LNI_ESTBCOMM),
+ FLAG_ENTRY0("Failed LNI(OptEq)", FAILED_LNI_OPTEQ),
+ FLAG_ENTRY0("Failed LNI(VerifyCap_1)", FAILED_LNI_VERIFY_CAP1),
+ FLAG_ENTRY0("Failed LNI(VerifyCap_2)", FAILED_LNI_VERIFY_CAP2),
+ FLAG_ENTRY0("Failed LNI(ConfigLT)", FAILED_LNI_CONFIGLT)
+};
+
+/*
+ * DC8051 Information Host Information flags
+ *
+ * Flags in DC8051_DBG_ERR_INFO_SET_BY_8051.HOST_MSG field.
+ */
+static struct flag_table dc8051_info_host_msg_flags[] = {
+ FLAG_ENTRY0("Host request done", 0x0001),
+ FLAG_ENTRY0("BC SMA message", 0x0002),
+ FLAG_ENTRY0("BC PWR_MGM message", 0x0004),
+ FLAG_ENTRY0("BC Unknown message (BCC)", 0x0008),
+ FLAG_ENTRY0("BC Unknown message (LCB)", 0x0010),
+ FLAG_ENTRY0("External device config request", 0x0020),
+ FLAG_ENTRY0("VerifyCap all frames received", 0x0040),
+ FLAG_ENTRY0("LinkUp achieved", 0x0080),
+ FLAG_ENTRY0("Link going down", 0x0100),
+};
+
+
+static u32 encoded_size(u32 size);
+static u32 chip_to_opa_lstate(struct hfi1_devdata *dd, u32 chip_lstate);
+static int set_physical_link_state(struct hfi1_devdata *dd, u64 state);
+static void read_vc_remote_phy(struct hfi1_devdata *dd, u8 *power_management,
+ u8 *continuous);
+static void read_vc_remote_fabric(struct hfi1_devdata *dd, u8 *vau, u8 *z,
+ u8 *vcu, u16 *vl15buf, u8 *crc_sizes);
+static void read_vc_remote_link_width(struct hfi1_devdata *dd,
+ u8 *remote_tx_rate, u16 *link_widths);
+static void read_vc_local_link_width(struct hfi1_devdata *dd, u8 *misc_bits,
+ u8 *flag_bits, u16 *link_widths);
+static void read_remote_device_id(struct hfi1_devdata *dd, u16 *device_id,
+ u8 *device_rev);
+static void read_mgmt_allowed(struct hfi1_devdata *dd, u8 *mgmt_allowed);
+static void read_local_lni(struct hfi1_devdata *dd, u8 *enable_lane_rx);
+static int read_tx_settings(struct hfi1_devdata *dd, u8 *enable_lane_tx,
+ u8 *tx_polarity_inversion,
+ u8 *rx_polarity_inversion, u8 *max_rate);
+static void handle_sdma_eng_err(struct hfi1_devdata *dd,
+ unsigned int context, u64 err_status);
+static void handle_qsfp_int(struct hfi1_devdata *dd, u32 source, u64 reg);
+static void handle_dcc_err(struct hfi1_devdata *dd,
+ unsigned int context, u64 err_status);
+static void handle_lcb_err(struct hfi1_devdata *dd,
+ unsigned int context, u64 err_status);
+static void handle_8051_interrupt(struct hfi1_devdata *dd, u32 unused, u64 reg);
+static void handle_cce_err(struct hfi1_devdata *dd, u32 unused, u64 reg);
+static void handle_rxe_err(struct hfi1_devdata *dd, u32 unused, u64 reg);
+static void handle_misc_err(struct hfi1_devdata *dd, u32 unused, u64 reg);
+static void handle_pio_err(struct hfi1_devdata *dd, u32 unused, u64 reg);
+static void handle_sdma_err(struct hfi1_devdata *dd, u32 unused, u64 reg);
+static void handle_egress_err(struct hfi1_devdata *dd, u32 unused, u64 reg);
+static void handle_txe_err(struct hfi1_devdata *dd, u32 unused, u64 reg);
+static void set_partition_keys(struct hfi1_pportdata *);
+static const char *link_state_name(u32 state);
+static const char *link_state_reason_name(struct hfi1_pportdata *ppd,
+ u32 state);
+static int do_8051_command(struct hfi1_devdata *dd, u32 type, u64 in_data,
+ u64 *out_data);
+static int read_idle_sma(struct hfi1_devdata *dd, u64 *data);
+static int thermal_init(struct hfi1_devdata *dd);
+
+static int wait_logical_linkstate(struct hfi1_pportdata *ppd, u32 state,
+ int msecs);
+static void read_planned_down_reason_code(struct hfi1_devdata *dd, u8 *pdrrc);
+static void handle_temp_err(struct hfi1_devdata *);
+static void dc_shutdown(struct hfi1_devdata *);
+static void dc_start(struct hfi1_devdata *);
+
+/*
+ * Error interrupt table entry. This is used as input to the interrupt
+ * "clear down" routine used for all second tier error interrupt register.
+ * Second tier interrupt registers have a single bit representing them
+ * in the top-level CceIntStatus.
+ */
+struct err_reg_info {
+ u32 status; /* status CSR offset */
+ u32 clear; /* clear CSR offset */
+ u32 mask; /* mask CSR offset */
+ void (*handler)(struct hfi1_devdata *dd, u32 source, u64 reg);
+ const char *desc;
+};
+
+#define NUM_MISC_ERRS (IS_GENERAL_ERR_END - IS_GENERAL_ERR_START)
+#define NUM_DC_ERRS (IS_DC_END - IS_DC_START)
+#define NUM_VARIOUS (IS_VARIOUS_END - IS_VARIOUS_START)
+
+/*
+ * Helpers for building HFI and DC error interrupt table entries. Different
+ * helpers are needed because of inconsistent register names.
+ */
+#define EE(reg, handler, desc) \
+ { reg##_STATUS, reg##_CLEAR, reg##_MASK, \
+ handler, desc }
+#define DC_EE1(reg, handler, desc) \
+ { reg##_FLG, reg##_FLG_CLR, reg##_FLG_EN, handler, desc }
+#define DC_EE2(reg, handler, desc) \
+ { reg##_FLG, reg##_CLR, reg##_EN, handler, desc }
+
+/*
+ * Table of the "misc" grouping of error interrupts. Each entry refers to
+ * another register containing more information.
+ */
+static const struct err_reg_info misc_errs[NUM_MISC_ERRS] = {
+/* 0*/ EE(CCE_ERR, handle_cce_err, "CceErr"),
+/* 1*/ EE(RCV_ERR, handle_rxe_err, "RxeErr"),
+/* 2*/ EE(MISC_ERR, handle_misc_err, "MiscErr"),
+/* 3*/ { 0, 0, 0, NULL }, /* reserved */
+/* 4*/ EE(SEND_PIO_ERR, handle_pio_err, "PioErr"),
+/* 5*/ EE(SEND_DMA_ERR, handle_sdma_err, "SDmaErr"),
+/* 6*/ EE(SEND_EGRESS_ERR, handle_egress_err, "EgressErr"),
+/* 7*/ EE(SEND_ERR, handle_txe_err, "TxeErr")
+ /* the rest are reserved */
+};
+
+/*
+ * Index into the Various section of the interrupt sources
+ * corresponding to the Critical Temperature interrupt.
+ */
+#define TCRIT_INT_SOURCE 4
+
+/*
+ * SDMA error interrupt entry - refers to another register containing more
+ * information.
+ */
+static const struct err_reg_info sdma_eng_err =
+ EE(SEND_DMA_ENG_ERR, handle_sdma_eng_err, "SDmaEngErr");
+
+static const struct err_reg_info various_err[NUM_VARIOUS] = {
+/* 0*/ { 0, 0, 0, NULL }, /* PbcInt */
+/* 1*/ { 0, 0, 0, NULL }, /* GpioAssertInt */
+/* 2*/ EE(ASIC_QSFP1, handle_qsfp_int, "QSFP1"),
+/* 3*/ EE(ASIC_QSFP2, handle_qsfp_int, "QSFP2"),
+/* 4*/ { 0, 0, 0, NULL }, /* TCritInt */
+ /* rest are reserved */
+};
+
+/*
+ * The DC encoding of mtu_cap for 10K MTU in the DCC_CFG_PORT_CONFIG
+ * register can not be derived from the MTU value because 10K is not
+ * a power of 2. Therefore, we need a constant. Everything else can
+ * be calculated.
+ */
+#define DCC_CFG_PORT_MTU_CAP_10240 7
+
+/*
+ * Table of the DC grouping of error interrupts. Each entry refers to
+ * another register containing more information.
+ */
+static const struct err_reg_info dc_errs[NUM_DC_ERRS] = {
+/* 0*/ DC_EE1(DCC_ERR, handle_dcc_err, "DCC Err"),
+/* 1*/ DC_EE2(DC_LCB_ERR, handle_lcb_err, "LCB Err"),
+/* 2*/ DC_EE2(DC_DC8051_ERR, handle_8051_interrupt, "DC8051 Interrupt"),
+/* 3*/ /* dc_lbm_int - special, see is_dc_int() */
+ /* the rest are reserved */
+};
+
+struct cntr_entry {
+ /*
+ * counter name
+ */
+ char *name;
+
+ /*
+ * csr to read for name (if applicable)
+ */
+ u64 csr;
+
+ /*
+ * offset into dd or ppd to store the counter's value
+ */
+ int offset;
+
+ /*
+ * flags
+ */
+ u8 flags;
+
+ /*
+ * accessor for stat element, context either dd or ppd
+ */
+ u64 (*rw_cntr)(const struct cntr_entry *,
+ void *context,
+ int vl,
+ int mode,
+ u64 data);
+};
+
+#define C_RCV_HDR_OVF_FIRST C_RCV_HDR_OVF_0
+#define C_RCV_HDR_OVF_LAST C_RCV_HDR_OVF_159
+
+#define CNTR_ELEM(name, csr, offset, flags, accessor) \
+{ \
+ name, \
+ csr, \
+ offset, \
+ flags, \
+ accessor \
+}
+
+/* 32bit RXE */
+#define RXE32_PORT_CNTR_ELEM(name, counter, flags) \
+CNTR_ELEM(#name, \
+ (counter * 8 + RCV_COUNTER_ARRAY32), \
+ 0, flags | CNTR_32BIT, \
+ port_access_u32_csr)
+
+#define RXE32_DEV_CNTR_ELEM(name, counter, flags) \
+CNTR_ELEM(#name, \
+ (counter * 8 + RCV_COUNTER_ARRAY32), \
+ 0, flags | CNTR_32BIT, \
+ dev_access_u32_csr)
+
+/* 64bit RXE */
+#define RXE64_PORT_CNTR_ELEM(name, counter, flags) \
+CNTR_ELEM(#name, \
+ (counter * 8 + RCV_COUNTER_ARRAY64), \
+ 0, flags, \
+ port_access_u64_csr)
+
+#define RXE64_DEV_CNTR_ELEM(name, counter, flags) \
+CNTR_ELEM(#name, \
+ (counter * 8 + RCV_COUNTER_ARRAY64), \
+ 0, flags, \
+ dev_access_u64_csr)
+
+#define OVR_LBL(ctx) C_RCV_HDR_OVF_ ## ctx
+#define OVR_ELM(ctx) \
+CNTR_ELEM("RcvHdrOvr" #ctx, \
+ (RCV_HDR_OVFL_CNT + ctx*0x100), \
+ 0, CNTR_NORMAL, port_access_u64_csr)
+
+/* 32bit TXE */
+#define TXE32_PORT_CNTR_ELEM(name, counter, flags) \
+CNTR_ELEM(#name, \
+ (counter * 8 + SEND_COUNTER_ARRAY32), \
+ 0, flags | CNTR_32BIT, \
+ port_access_u32_csr)
+
+/* 64bit TXE */
+#define TXE64_PORT_CNTR_ELEM(name, counter, flags) \
+CNTR_ELEM(#name, \
+ (counter * 8 + SEND_COUNTER_ARRAY64), \
+ 0, flags, \
+ port_access_u64_csr)
+
+# define TX64_DEV_CNTR_ELEM(name, counter, flags) \
+CNTR_ELEM(#name,\
+ counter * 8 + SEND_COUNTER_ARRAY64, \
+ 0, \
+ flags, \
+ dev_access_u64_csr)
+
+/* CCE */
+#define CCE_PERF_DEV_CNTR_ELEM(name, counter, flags) \
+CNTR_ELEM(#name, \
+ (counter * 8 + CCE_COUNTER_ARRAY32), \
+ 0, flags | CNTR_32BIT, \
+ dev_access_u32_csr)
+
+#define CCE_INT_DEV_CNTR_ELEM(name, counter, flags) \
+CNTR_ELEM(#name, \
+ (counter * 8 + CCE_INT_COUNTER_ARRAY32), \
+ 0, flags | CNTR_32BIT, \
+ dev_access_u32_csr)
+
+/* DC */
+#define DC_PERF_CNTR(name, counter, flags) \
+CNTR_ELEM(#name, \
+ counter, \
+ 0, \
+ flags, \
+ dev_access_u64_csr)
+
+#define DC_PERF_CNTR_LCB(name, counter, flags) \
+CNTR_ELEM(#name, \
+ counter, \
+ 0, \
+ flags, \
+ dc_access_lcb_cntr)
+
+/* ibp counters */
+#define SW_IBP_CNTR(name, cntr) \
+CNTR_ELEM(#name, \
+ 0, \
+ 0, \
+ CNTR_SYNTH, \
+ access_ibp_##cntr)
+
+u64 read_csr(const struct hfi1_devdata *dd, u32 offset)
+{
+ u64 val;
+
+ if (dd->flags & HFI1_PRESENT) {
+ val = readq((void __iomem *)dd->kregbase + offset);
+ return val;
+ }
+ return -1;
+}
+
+void write_csr(const struct hfi1_devdata *dd, u32 offset, u64 value)
+{
+ if (dd->flags & HFI1_PRESENT)
+ writeq(value, (void __iomem *)dd->kregbase + offset);
+}
+
+void __iomem *get_csr_addr(
+ struct hfi1_devdata *dd,
+ u32 offset)
+{
+ return (void __iomem *)dd->kregbase + offset;
+}
+
+static inline u64 read_write_csr(const struct hfi1_devdata *dd, u32 csr,
+ int mode, u64 value)
+{
+ u64 ret;
+
+
+ if (mode == CNTR_MODE_R) {
+ ret = read_csr(dd, csr);
+ } else if (mode == CNTR_MODE_W) {
+ write_csr(dd, csr, value);
+ ret = value;
+ } else {
+ dd_dev_err(dd, "Invalid cntr register access mode");
+ return 0;
+ }
+
+ hfi1_cdbg(CNTR, "csr 0x%x val 0x%llx mode %d", csr, ret, mode);
+ return ret;
+}
+
+/* Dev Access */
+static u64 dev_access_u32_csr(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+ return read_write_csr(dd, entry->csr, mode, data);
+}
+
+static u64 dev_access_u64_csr(const struct cntr_entry *entry, void *context,
+ int vl, int mode, u64 data)
+{
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+ u64 val = 0;
+ u64 csr = entry->csr;
+
+ if (entry->flags & CNTR_VL) {
+ if (vl == CNTR_INVALID_VL)
+ return 0;
+ csr += 8 * vl;
+ } else {
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+ }
+
+ val = read_write_csr(dd, csr, mode, data);
+ return val;
+}
+
+static u64 dc_access_lcb_cntr(const struct cntr_entry *entry, void *context,
+ int vl, int mode, u64 data)
+{
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+ u32 csr = entry->csr;
+ int ret = 0;
+
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+ if (mode == CNTR_MODE_R)
+ ret = read_lcb_csr(dd, csr, &data);
+ else if (mode == CNTR_MODE_W)
+ ret = write_lcb_csr(dd, csr, data);
+
+ if (ret) {
+ dd_dev_err(dd, "Could not acquire LCB for counter 0x%x", csr);
+ return 0;
+ }
+
+ hfi1_cdbg(CNTR, "csr 0x%x val 0x%llx mode %d", csr, data, mode);
+ return data;
+}
+
+/* Port Access */
+static u64 port_access_u32_csr(const struct cntr_entry *entry, void *context,
+ int vl, int mode, u64 data)
+{
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context;
+
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+ return read_write_csr(ppd->dd, entry->csr, mode, data);
+}
+
+static u64 port_access_u64_csr(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context;
+ u64 val;
+ u64 csr = entry->csr;
+
+ if (entry->flags & CNTR_VL) {
+ if (vl == CNTR_INVALID_VL)
+ return 0;
+ csr += 8 * vl;
+ } else {
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+ }
+ val = read_write_csr(ppd->dd, csr, mode, data);
+ return val;
+}
+
+/* Software defined */
+static inline u64 read_write_sw(struct hfi1_devdata *dd, u64 *cntr, int mode,
+ u64 data)
+{
+ u64 ret;
+
+ if (mode == CNTR_MODE_R) {
+ ret = *cntr;
+ } else if (mode == CNTR_MODE_W) {
+ *cntr = data;
+ ret = data;
+ } else {
+ dd_dev_err(dd, "Invalid cntr sw access mode");
+ return 0;
+ }
+
+ hfi1_cdbg(CNTR, "val 0x%llx mode %d", ret, mode);
+
+ return ret;
+}
+
+static u64 access_sw_link_dn_cnt(const struct cntr_entry *entry, void *context,
+ int vl, int mode, u64 data)
+{
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context;
+
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+ return read_write_sw(ppd->dd, &ppd->link_downed, mode, data);
+}
+
+static u64 access_sw_link_up_cnt(const struct cntr_entry *entry, void *context,
+ int vl, int mode, u64 data)
+{
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context;
+
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+ return read_write_sw(ppd->dd, &ppd->link_up, mode, data);
+}
+
+static u64 access_sw_xmit_discards(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context;
+
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+
+ return read_write_sw(ppd->dd, &ppd->port_xmit_discards, mode, data);
+}
+
+static u64 access_xmit_constraint_errs(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context;
+
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+
+ return read_write_sw(ppd->dd, &ppd->port_xmit_constraint_errors,
+ mode, data);
+}
+
+static u64 access_rcv_constraint_errs(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context;
+
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+
+ return read_write_sw(ppd->dd, &ppd->port_rcv_constraint_errors,
+ mode, data);
+}
+
+u64 get_all_cpu_total(u64 __percpu *cntr)
+{
+ int cpu;
+ u64 counter = 0;
+
+ for_each_possible_cpu(cpu)
+ counter += *per_cpu_ptr(cntr, cpu);
+ return counter;
+}
+
+static u64 read_write_cpu(struct hfi1_devdata *dd, u64 *z_val,
+ u64 __percpu *cntr,
+ int vl, int mode, u64 data)
+{
+
+ u64 ret = 0;
+
+ if (vl != CNTR_INVALID_VL)
+ return 0;
+
+ if (mode == CNTR_MODE_R) {
+ ret = get_all_cpu_total(cntr) - *z_val;
+ } else if (mode == CNTR_MODE_W) {
+ /* A write can only zero the counter */
+ if (data == 0)
+ *z_val = get_all_cpu_total(cntr);
+ else
+ dd_dev_err(dd, "Per CPU cntrs can only be zeroed");
+ } else {
+ dd_dev_err(dd, "Invalid cntr sw cpu access mode");
+ return 0;
+ }
+
+ return ret;
+}
+
+static u64 access_sw_cpu_intr(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+ return read_write_cpu(dd, &dd->z_int_counter, dd->int_counter, vl,
+ mode, data);
+}
+
+static u64 access_sw_cpu_rcv_limit(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+ return read_write_cpu(dd, &dd->z_rcv_limit, dd->rcv_limit, vl,
+ mode, data);
+}
+
+static u64 access_sw_pio_wait(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+ return dd->verbs_dev.n_piowait;
+}
+
+static u64 access_sw_vtx_wait(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+ return dd->verbs_dev.n_txwait;
+}
+
+static u64 access_sw_kmem_wait(const struct cntr_entry *entry,
+ void *context, int vl, int mode, u64 data)
+{
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+ return dd->verbs_dev.n_kmem_wait;
+}
+
+#define def_access_sw_cpu(cntr) \
+static u64 access_sw_cpu_##cntr(const struct cntr_entry *entry, \
+ void *context, int vl, int mode, u64 data) \
+{ \
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context; \
+ return read_write_cpu(ppd->dd, &ppd->ibport_data.z_ ##cntr, \
+ ppd->ibport_data.cntr, vl, \
+ mode, data); \
+}
+
+def_access_sw_cpu(rc_acks);
+def_access_sw_cpu(rc_qacks);
+def_access_sw_cpu(rc_delayed_comp);
+
+#define def_access_ibp_counter(cntr) \
+static u64 access_ibp_##cntr(const struct cntr_entry *entry, \
+ void *context, int vl, int mode, u64 data) \
+{ \
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context; \
+ \
+ if (vl != CNTR_INVALID_VL) \
+ return 0; \
+ \
+ return read_write_sw(ppd->dd, &ppd->ibport_data.n_ ##cntr, \
+ mode, data); \
+}
+
+def_access_ibp_counter(loop_pkts);
+def_access_ibp_counter(rc_resends);
+def_access_ibp_counter(rnr_naks);
+def_access_ibp_counter(other_naks);
+def_access_ibp_counter(rc_timeouts);
+def_access_ibp_counter(pkt_drops);
+def_access_ibp_counter(dmawait);
+def_access_ibp_counter(rc_seqnak);
+def_access_ibp_counter(rc_dupreq);
+def_access_ibp_counter(rdma_seq);
+def_access_ibp_counter(unaligned);
+def_access_ibp_counter(seq_naks);
+
+static struct cntr_entry dev_cntrs[DEV_CNTR_LAST] = {
+[C_RCV_OVF] = RXE32_DEV_CNTR_ELEM(RcvOverflow, RCV_BUF_OVFL_CNT, CNTR_SYNTH),
+[C_RX_TID_FULL] = RXE32_DEV_CNTR_ELEM(RxTIDFullEr, RCV_TID_FULL_ERR_CNT,
+ CNTR_NORMAL),
+[C_RX_TID_INVALID] = RXE32_DEV_CNTR_ELEM(RxTIDInvalid, RCV_TID_VALID_ERR_CNT,
+ CNTR_NORMAL),
+[C_RX_TID_FLGMS] = RXE32_DEV_CNTR_ELEM(RxTidFLGMs,
+ RCV_TID_FLOW_GEN_MISMATCH_CNT,
+ CNTR_NORMAL),
+[C_RX_CTX_RHQS] = RXE32_DEV_CNTR_ELEM(RxCtxRHQS, RCV_CONTEXT_RHQ_STALL,
+ CNTR_NORMAL),
+[C_RX_CTX_EGRS] = RXE32_DEV_CNTR_ELEM(RxCtxEgrS, RCV_CONTEXT_EGR_STALL,
+ CNTR_NORMAL),
+[C_RCV_TID_FLSMS] = RXE32_DEV_CNTR_ELEM(RxTidFLSMs,
+ RCV_TID_FLOW_SEQ_MISMATCH_CNT, CNTR_NORMAL),
+[C_CCE_PCI_CR_ST] = CCE_PERF_DEV_CNTR_ELEM(CcePciCrSt,
+ CCE_PCIE_POSTED_CRDT_STALL_CNT, CNTR_NORMAL),
+[C_CCE_PCI_TR_ST] = CCE_PERF_DEV_CNTR_ELEM(CcePciTrSt, CCE_PCIE_TRGT_STALL_CNT,
+ CNTR_NORMAL),
+[C_CCE_PIO_WR_ST] = CCE_PERF_DEV_CNTR_ELEM(CcePioWrSt, CCE_PIO_WR_STALL_CNT,
+ CNTR_NORMAL),
+[C_CCE_ERR_INT] = CCE_INT_DEV_CNTR_ELEM(CceErrInt, CCE_ERR_INT_CNT,
+ CNTR_NORMAL),
+[C_CCE_SDMA_INT] = CCE_INT_DEV_CNTR_ELEM(CceSdmaInt, CCE_SDMA_INT_CNT,
+ CNTR_NORMAL),
+[C_CCE_MISC_INT] = CCE_INT_DEV_CNTR_ELEM(CceMiscInt, CCE_MISC_INT_CNT,
+ CNTR_NORMAL),
+[C_CCE_RCV_AV_INT] = CCE_INT_DEV_CNTR_ELEM(CceRcvAvInt, CCE_RCV_AVAIL_INT_CNT,
+ CNTR_NORMAL),
+[C_CCE_RCV_URG_INT] = CCE_INT_DEV_CNTR_ELEM(CceRcvUrgInt,
+ CCE_RCV_URGENT_INT_CNT, CNTR_NORMAL),
+[C_CCE_SEND_CR_INT] = CCE_INT_DEV_CNTR_ELEM(CceSndCrInt,
+ CCE_SEND_CREDIT_INT_CNT, CNTR_NORMAL),
+[C_DC_UNC_ERR] = DC_PERF_CNTR(DcUnctblErr, DCC_ERR_UNCORRECTABLE_CNT,
+ CNTR_SYNTH),
+[C_DC_RCV_ERR] = DC_PERF_CNTR(DcRecvErr, DCC_ERR_PORTRCV_ERR_CNT, CNTR_SYNTH),
+[C_DC_FM_CFG_ERR] = DC_PERF_CNTR(DcFmCfgErr, DCC_ERR_FMCONFIG_ERR_CNT,
+ CNTR_SYNTH),
+[C_DC_RMT_PHY_ERR] = DC_PERF_CNTR(DcRmtPhyErr, DCC_ERR_RCVREMOTE_PHY_ERR_CNT,
+ CNTR_SYNTH),
+[C_DC_DROPPED_PKT] = DC_PERF_CNTR(DcDroppedPkt, DCC_ERR_DROPPED_PKT_CNT,
+ CNTR_SYNTH),
+[C_DC_MC_XMIT_PKTS] = DC_PERF_CNTR(DcMcXmitPkts,
+ DCC_PRF_PORT_XMIT_MULTICAST_CNT, CNTR_SYNTH),
+[C_DC_MC_RCV_PKTS] = DC_PERF_CNTR(DcMcRcvPkts,
+ DCC_PRF_PORT_RCV_MULTICAST_PKT_CNT,
+ CNTR_SYNTH),
+[C_DC_XMIT_CERR] = DC_PERF_CNTR(DcXmitCorr,
+ DCC_PRF_PORT_XMIT_CORRECTABLE_CNT, CNTR_SYNTH),
+[C_DC_RCV_CERR] = DC_PERF_CNTR(DcRcvCorrCnt, DCC_PRF_PORT_RCV_CORRECTABLE_CNT,
+ CNTR_SYNTH),
+[C_DC_RCV_FCC] = DC_PERF_CNTR(DcRxFCntl, DCC_PRF_RX_FLOW_CRTL_CNT,
+ CNTR_SYNTH),
+[C_DC_XMIT_FCC] = DC_PERF_CNTR(DcXmitFCntl, DCC_PRF_TX_FLOW_CRTL_CNT,
+ CNTR_SYNTH),
+[C_DC_XMIT_FLITS] = DC_PERF_CNTR(DcXmitFlits, DCC_PRF_PORT_XMIT_DATA_CNT,
+ CNTR_SYNTH),
+[C_DC_RCV_FLITS] = DC_PERF_CNTR(DcRcvFlits, DCC_PRF_PORT_RCV_DATA_CNT,
+ CNTR_SYNTH),
+[C_DC_XMIT_PKTS] = DC_PERF_CNTR(DcXmitPkts, DCC_PRF_PORT_XMIT_PKTS_CNT,
+ CNTR_SYNTH),
+[C_DC_RCV_PKTS] = DC_PERF_CNTR(DcRcvPkts, DCC_PRF_PORT_RCV_PKTS_CNT,
+ CNTR_SYNTH),
+[C_DC_RX_FLIT_VL] = DC_PERF_CNTR(DcRxFlitVl, DCC_PRF_PORT_VL_RCV_DATA_CNT,
+ CNTR_SYNTH | CNTR_VL),
+[C_DC_RX_PKT_VL] = DC_PERF_CNTR(DcRxPktVl, DCC_PRF_PORT_VL_RCV_PKTS_CNT,
+ CNTR_SYNTH | CNTR_VL),
+[C_DC_RCV_FCN] = DC_PERF_CNTR(DcRcvFcn, DCC_PRF_PORT_RCV_FECN_CNT, CNTR_SYNTH),
+[C_DC_RCV_FCN_VL] = DC_PERF_CNTR(DcRcvFcnVl, DCC_PRF_PORT_VL_RCV_FECN_CNT,
+ CNTR_SYNTH | CNTR_VL),
+[C_DC_RCV_BCN] = DC_PERF_CNTR(DcRcvBcn, DCC_PRF_PORT_RCV_BECN_CNT, CNTR_SYNTH),
+[C_DC_RCV_BCN_VL] = DC_PERF_CNTR(DcRcvBcnVl, DCC_PRF_PORT_VL_RCV_BECN_CNT,
+ CNTR_SYNTH | CNTR_VL),
+[C_DC_RCV_BBL] = DC_PERF_CNTR(DcRcvBbl, DCC_PRF_PORT_RCV_BUBBLE_CNT,
+ CNTR_SYNTH),
+[C_DC_RCV_BBL_VL] = DC_PERF_CNTR(DcRcvBblVl, DCC_PRF_PORT_VL_RCV_BUBBLE_CNT,
+ CNTR_SYNTH | CNTR_VL),
+[C_DC_MARK_FECN] = DC_PERF_CNTR(DcMarkFcn, DCC_PRF_PORT_MARK_FECN_CNT,
+ CNTR_SYNTH),
+[C_DC_MARK_FECN_VL] = DC_PERF_CNTR(DcMarkFcnVl, DCC_PRF_PORT_VL_MARK_FECN_CNT,
+ CNTR_SYNTH | CNTR_VL),
+[C_DC_TOTAL_CRC] =
+ DC_PERF_CNTR_LCB(DcTotCrc, DC_LCB_ERR_INFO_TOTAL_CRC_ERR,
+ CNTR_SYNTH),
+[C_DC_CRC_LN0] = DC_PERF_CNTR_LCB(DcCrcLn0, DC_LCB_ERR_INFO_CRC_ERR_LN0,
+ CNTR_SYNTH),
+[C_DC_CRC_LN1] = DC_PERF_CNTR_LCB(DcCrcLn1, DC_LCB_ERR_INFO_CRC_ERR_LN1,
+ CNTR_SYNTH),
+[C_DC_CRC_LN2] = DC_PERF_CNTR_LCB(DcCrcLn2, DC_LCB_ERR_INFO_CRC_ERR_LN2,
+ CNTR_SYNTH),
+[C_DC_CRC_LN3] = DC_PERF_CNTR_LCB(DcCrcLn3, DC_LCB_ERR_INFO_CRC_ERR_LN3,
+ CNTR_SYNTH),
+[C_DC_CRC_MULT_LN] =
+ DC_PERF_CNTR_LCB(DcMultLn, DC_LCB_ERR_INFO_CRC_ERR_MULTI_LN,
+ CNTR_SYNTH),
+[C_DC_TX_REPLAY] = DC_PERF_CNTR_LCB(DcTxReplay, DC_LCB_ERR_INFO_TX_REPLAY_CNT,
+ CNTR_SYNTH),
+[C_DC_RX_REPLAY] = DC_PERF_CNTR_LCB(DcRxReplay, DC_LCB_ERR_INFO_RX_REPLAY_CNT,
+ CNTR_SYNTH),
+[C_DC_SEQ_CRC_CNT] =
+ DC_PERF_CNTR_LCB(DcLinkSeqCrc, DC_LCB_ERR_INFO_SEQ_CRC_CNT,
+ CNTR_SYNTH),
+[C_DC_ESC0_ONLY_CNT] =
+ DC_PERF_CNTR_LCB(DcEsc0, DC_LCB_ERR_INFO_ESCAPE_0_ONLY_CNT,
+ CNTR_SYNTH),
+[C_DC_ESC0_PLUS1_CNT] =
+ DC_PERF_CNTR_LCB(DcEsc1, DC_LCB_ERR_INFO_ESCAPE_0_PLUS1_CNT,
+ CNTR_SYNTH),
+[C_DC_ESC0_PLUS2_CNT] =
+ DC_PERF_CNTR_LCB(DcEsc0Plus2, DC_LCB_ERR_INFO_ESCAPE_0_PLUS2_CNT,
+ CNTR_SYNTH),
+[C_DC_REINIT_FROM_PEER_CNT] =
+ DC_PERF_CNTR_LCB(DcReinitPeer, DC_LCB_ERR_INFO_REINIT_FROM_PEER_CNT,
+ CNTR_SYNTH),
+[C_DC_SBE_CNT] = DC_PERF_CNTR_LCB(DcSbe, DC_LCB_ERR_INFO_SBE_CNT,
+ CNTR_SYNTH),
+[C_DC_MISC_FLG_CNT] =
+ DC_PERF_CNTR_LCB(DcMiscFlg, DC_LCB_ERR_INFO_MISC_FLG_CNT,
+ CNTR_SYNTH),
+[C_DC_PRF_GOOD_LTP_CNT] =
+ DC_PERF_CNTR_LCB(DcGoodLTP, DC_LCB_PRF_GOOD_LTP_CNT, CNTR_SYNTH),
+[C_DC_PRF_ACCEPTED_LTP_CNT] =
+ DC_PERF_CNTR_LCB(DcAccLTP, DC_LCB_PRF_ACCEPTED_LTP_CNT,
+ CNTR_SYNTH),
+[C_DC_PRF_RX_FLIT_CNT] =
+ DC_PERF_CNTR_LCB(DcPrfRxFlit, DC_LCB_PRF_RX_FLIT_CNT, CNTR_SYNTH),
+[C_DC_PRF_TX_FLIT_CNT] =
+ DC_PERF_CNTR_LCB(DcPrfTxFlit, DC_LCB_PRF_TX_FLIT_CNT, CNTR_SYNTH),
+[C_DC_PRF_CLK_CNTR] =
+ DC_PERF_CNTR_LCB(DcPrfClk, DC_LCB_PRF_CLK_CNTR, CNTR_SYNTH),
+[C_DC_PG_DBG_FLIT_CRDTS_CNT] =
+ DC_PERF_CNTR_LCB(DcFltCrdts, DC_LCB_PG_DBG_FLIT_CRDTS_CNT, CNTR_SYNTH),
+[C_DC_PG_STS_PAUSE_COMPLETE_CNT] =
+ DC_PERF_CNTR_LCB(DcPauseComp, DC_LCB_PG_STS_PAUSE_COMPLETE_CNT,
+ CNTR_SYNTH),
+[C_DC_PG_STS_TX_SBE_CNT] =
+ DC_PERF_CNTR_LCB(DcStsTxSbe, DC_LCB_PG_STS_TX_SBE_CNT, CNTR_SYNTH),
+[C_DC_PG_STS_TX_MBE_CNT] =
+ DC_PERF_CNTR_LCB(DcStsTxMbe, DC_LCB_PG_STS_TX_MBE_CNT,
+ CNTR_SYNTH),
+[C_SW_CPU_INTR] = CNTR_ELEM("Intr", 0, 0, CNTR_NORMAL,
+ access_sw_cpu_intr),
+[C_SW_CPU_RCV_LIM] = CNTR_ELEM("RcvLimit", 0, 0, CNTR_NORMAL,
+ access_sw_cpu_rcv_limit),
+[C_SW_VTX_WAIT] = CNTR_ELEM("vTxWait", 0, 0, CNTR_NORMAL,
+ access_sw_vtx_wait),
+[C_SW_PIO_WAIT] = CNTR_ELEM("PioWait", 0, 0, CNTR_NORMAL,
+ access_sw_pio_wait),
+[C_SW_KMEM_WAIT] = CNTR_ELEM("KmemWait", 0, 0, CNTR_NORMAL,
+ access_sw_kmem_wait),
+};
+
+static struct cntr_entry port_cntrs[PORT_CNTR_LAST] = {
+[C_TX_UNSUP_VL] = TXE32_PORT_CNTR_ELEM(TxUnVLErr, SEND_UNSUP_VL_ERR_CNT,
+ CNTR_NORMAL),
+[C_TX_INVAL_LEN] = TXE32_PORT_CNTR_ELEM(TxInvalLen, SEND_LEN_ERR_CNT,
+ CNTR_NORMAL),
+[C_TX_MM_LEN_ERR] = TXE32_PORT_CNTR_ELEM(TxMMLenErr, SEND_MAX_MIN_LEN_ERR_CNT,
+ CNTR_NORMAL),
+[C_TX_UNDERRUN] = TXE32_PORT_CNTR_ELEM(TxUnderrun, SEND_UNDERRUN_CNT,
+ CNTR_NORMAL),
+[C_TX_FLOW_STALL] = TXE32_PORT_CNTR_ELEM(TxFlowStall, SEND_FLOW_STALL_CNT,
+ CNTR_NORMAL),
+[C_TX_DROPPED] = TXE32_PORT_CNTR_ELEM(TxDropped, SEND_DROPPED_PKT_CNT,
+ CNTR_NORMAL),
+[C_TX_HDR_ERR] = TXE32_PORT_CNTR_ELEM(TxHdrErr, SEND_HEADERS_ERR_CNT,
+ CNTR_NORMAL),
+[C_TX_PKT] = TXE64_PORT_CNTR_ELEM(TxPkt, SEND_DATA_PKT_CNT, CNTR_NORMAL),
+[C_TX_WORDS] = TXE64_PORT_CNTR_ELEM(TxWords, SEND_DWORD_CNT, CNTR_NORMAL),
+[C_TX_WAIT] = TXE64_PORT_CNTR_ELEM(TxWait, SEND_WAIT_CNT, CNTR_SYNTH),
+[C_TX_FLIT_VL] = TXE64_PORT_CNTR_ELEM(TxFlitVL, SEND_DATA_VL0_CNT,
+ CNTR_SYNTH | CNTR_VL),
+[C_TX_PKT_VL] = TXE64_PORT_CNTR_ELEM(TxPktVL, SEND_DATA_PKT_VL0_CNT,
+ CNTR_SYNTH | CNTR_VL),
+[C_TX_WAIT_VL] = TXE64_PORT_CNTR_ELEM(TxWaitVL, SEND_WAIT_VL0_CNT,
+ CNTR_SYNTH | CNTR_VL),
+[C_RX_PKT] = RXE64_PORT_CNTR_ELEM(RxPkt, RCV_DATA_PKT_CNT, CNTR_NORMAL),
+[C_RX_WORDS] = RXE64_PORT_CNTR_ELEM(RxWords, RCV_DWORD_CNT, CNTR_NORMAL),
+[C_SW_LINK_DOWN] = CNTR_ELEM("SwLinkDown", 0, 0, CNTR_SYNTH | CNTR_32BIT,
+ access_sw_link_dn_cnt),
+[C_SW_LINK_UP] = CNTR_ELEM("SwLinkUp", 0, 0, CNTR_SYNTH | CNTR_32BIT,
+ access_sw_link_up_cnt),
+[C_SW_XMIT_DSCD] = CNTR_ELEM("XmitDscd", 0, 0, CNTR_SYNTH | CNTR_32BIT,
+ access_sw_xmit_discards),
+[C_SW_XMIT_DSCD_VL] = CNTR_ELEM("XmitDscdVl", 0, 0,
+ CNTR_SYNTH | CNTR_32BIT | CNTR_VL,
+ access_sw_xmit_discards),
+[C_SW_XMIT_CSTR_ERR] = CNTR_ELEM("XmitCstrErr", 0, 0, CNTR_SYNTH,
+ access_xmit_constraint_errs),
+[C_SW_RCV_CSTR_ERR] = CNTR_ELEM("RcvCstrErr", 0, 0, CNTR_SYNTH,
+ access_rcv_constraint_errs),
+[C_SW_IBP_LOOP_PKTS] = SW_IBP_CNTR(LoopPkts, loop_pkts),
+[C_SW_IBP_RC_RESENDS] = SW_IBP_CNTR(RcResend, rc_resends),
+[C_SW_IBP_RNR_NAKS] = SW_IBP_CNTR(RnrNak, rnr_naks),
+[C_SW_IBP_OTHER_NAKS] = SW_IBP_CNTR(OtherNak, other_naks),
+[C_SW_IBP_RC_TIMEOUTS] = SW_IBP_CNTR(RcTimeOut, rc_timeouts),
+[C_SW_IBP_PKT_DROPS] = SW_IBP_CNTR(PktDrop, pkt_drops),
+[C_SW_IBP_DMA_WAIT] = SW_IBP_CNTR(DmaWait, dmawait),
+[C_SW_IBP_RC_SEQNAK] = SW_IBP_CNTR(RcSeqNak, rc_seqnak),
+[C_SW_IBP_RC_DUPREQ] = SW_IBP_CNTR(RcDupRew, rc_dupreq),
+[C_SW_IBP_RDMA_SEQ] = SW_IBP_CNTR(RdmaSeq, rdma_seq),
+[C_SW_IBP_UNALIGNED] = SW_IBP_CNTR(Unaligned, unaligned),
+[C_SW_IBP_SEQ_NAK] = SW_IBP_CNTR(SeqNak, seq_naks),
+[C_SW_CPU_RC_ACKS] = CNTR_ELEM("RcAcks", 0, 0, CNTR_NORMAL,
+ access_sw_cpu_rc_acks),
+[C_SW_CPU_RC_QACKS] = CNTR_ELEM("RcQacks", 0, 0, CNTR_NORMAL,
+ access_sw_cpu_rc_qacks),
+[C_SW_CPU_RC_DELAYED_COMP] = CNTR_ELEM("RcDelayComp", 0, 0, CNTR_NORMAL,
+ access_sw_cpu_rc_delayed_comp),
+[OVR_LBL(0)] = OVR_ELM(0), [OVR_LBL(1)] = OVR_ELM(1),
+[OVR_LBL(2)] = OVR_ELM(2), [OVR_LBL(3)] = OVR_ELM(3),
+[OVR_LBL(4)] = OVR_ELM(4), [OVR_LBL(5)] = OVR_ELM(5),
+[OVR_LBL(6)] = OVR_ELM(6), [OVR_LBL(7)] = OVR_ELM(7),
+[OVR_LBL(8)] = OVR_ELM(8), [OVR_LBL(9)] = OVR_ELM(9),
+[OVR_LBL(10)] = OVR_ELM(10), [OVR_LBL(11)] = OVR_ELM(11),
+[OVR_LBL(12)] = OVR_ELM(12), [OVR_LBL(13)] = OVR_ELM(13),
+[OVR_LBL(14)] = OVR_ELM(14), [OVR_LBL(15)] = OVR_ELM(15),
+[OVR_LBL(16)] = OVR_ELM(16), [OVR_LBL(17)] = OVR_ELM(17),
+[OVR_LBL(18)] = OVR_ELM(18), [OVR_LBL(19)] = OVR_ELM(19),
+[OVR_LBL(20)] = OVR_ELM(20), [OVR_LBL(21)] = OVR_ELM(21),
+[OVR_LBL(22)] = OVR_ELM(22), [OVR_LBL(23)] = OVR_ELM(23),
+[OVR_LBL(24)] = OVR_ELM(24), [OVR_LBL(25)] = OVR_ELM(25),
+[OVR_LBL(26)] = OVR_ELM(26), [OVR_LBL(27)] = OVR_ELM(27),
+[OVR_LBL(28)] = OVR_ELM(28), [OVR_LBL(29)] = OVR_ELM(29),
+[OVR_LBL(30)] = OVR_ELM(30), [OVR_LBL(31)] = OVR_ELM(31),
+[OVR_LBL(32)] = OVR_ELM(32), [OVR_LBL(33)] = OVR_ELM(33),
+[OVR_LBL(34)] = OVR_ELM(34), [OVR_LBL(35)] = OVR_ELM(35),
+[OVR_LBL(36)] = OVR_ELM(36), [OVR_LBL(37)] = OVR_ELM(37),
+[OVR_LBL(38)] = OVR_ELM(38), [OVR_LBL(39)] = OVR_ELM(39),
+[OVR_LBL(40)] = OVR_ELM(40), [OVR_LBL(41)] = OVR_ELM(41),
+[OVR_LBL(42)] = OVR_ELM(42), [OVR_LBL(43)] = OVR_ELM(43),
+[OVR_LBL(44)] = OVR_ELM(44), [OVR_LBL(45)] = OVR_ELM(45),
+[OVR_LBL(46)] = OVR_ELM(46), [OVR_LBL(47)] = OVR_ELM(47),
+[OVR_LBL(48)] = OVR_ELM(48), [OVR_LBL(49)] = OVR_ELM(49),
+[OVR_LBL(50)] = OVR_ELM(50), [OVR_LBL(51)] = OVR_ELM(51),
+[OVR_LBL(52)] = OVR_ELM(52), [OVR_LBL(53)] = OVR_ELM(53),
+[OVR_LBL(54)] = OVR_ELM(54), [OVR_LBL(55)] = OVR_ELM(55),
+[OVR_LBL(56)] = OVR_ELM(56), [OVR_LBL(57)] = OVR_ELM(57),
+[OVR_LBL(58)] = OVR_ELM(58), [OVR_LBL(59)] = OVR_ELM(59),
+[OVR_LBL(60)] = OVR_ELM(60), [OVR_LBL(61)] = OVR_ELM(61),
+[OVR_LBL(62)] = OVR_ELM(62), [OVR_LBL(63)] = OVR_ELM(63),
+[OVR_LBL(64)] = OVR_ELM(64), [OVR_LBL(65)] = OVR_ELM(65),
+[OVR_LBL(66)] = OVR_ELM(66), [OVR_LBL(67)] = OVR_ELM(67),
+[OVR_LBL(68)] = OVR_ELM(68), [OVR_LBL(69)] = OVR_ELM(69),
+[OVR_LBL(70)] = OVR_ELM(70), [OVR_LBL(71)] = OVR_ELM(71),
+[OVR_LBL(72)] = OVR_ELM(72), [OVR_LBL(73)] = OVR_ELM(73),
+[OVR_LBL(74)] = OVR_ELM(74), [OVR_LBL(75)] = OVR_ELM(75),
+[OVR_LBL(76)] = OVR_ELM(76), [OVR_LBL(77)] = OVR_ELM(77),
+[OVR_LBL(78)] = OVR_ELM(78), [OVR_LBL(79)] = OVR_ELM(79),
+[OVR_LBL(80)] = OVR_ELM(80), [OVR_LBL(81)] = OVR_ELM(81),
+[OVR_LBL(82)] = OVR_ELM(82), [OVR_LBL(83)] = OVR_ELM(83),
+[OVR_LBL(84)] = OVR_ELM(84), [OVR_LBL(85)] = OVR_ELM(85),
+[OVR_LBL(86)] = OVR_ELM(86), [OVR_LBL(87)] = OVR_ELM(87),
+[OVR_LBL(88)] = OVR_ELM(88), [OVR_LBL(89)] = OVR_ELM(89),
+[OVR_LBL(90)] = OVR_ELM(90), [OVR_LBL(91)] = OVR_ELM(91),
+[OVR_LBL(92)] = OVR_ELM(92), [OVR_LBL(93)] = OVR_ELM(93),
+[OVR_LBL(94)] = OVR_ELM(94), [OVR_LBL(95)] = OVR_ELM(95),
+[OVR_LBL(96)] = OVR_ELM(96), [OVR_LBL(97)] = OVR_ELM(97),
+[OVR_LBL(98)] = OVR_ELM(98), [OVR_LBL(99)] = OVR_ELM(99),
+[OVR_LBL(100)] = OVR_ELM(100), [OVR_LBL(101)] = OVR_ELM(101),
+[OVR_LBL(102)] = OVR_ELM(102), [OVR_LBL(103)] = OVR_ELM(103),
+[OVR_LBL(104)] = OVR_ELM(104), [OVR_LBL(105)] = OVR_ELM(105),
+[OVR_LBL(106)] = OVR_ELM(106), [OVR_LBL(107)] = OVR_ELM(107),
+[OVR_LBL(108)] = OVR_ELM(108), [OVR_LBL(109)] = OVR_ELM(109),
+[OVR_LBL(110)] = OVR_ELM(110), [OVR_LBL(111)] = OVR_ELM(111),
+[OVR_LBL(112)] = OVR_ELM(112), [OVR_LBL(113)] = OVR_ELM(113),
+[OVR_LBL(114)] = OVR_ELM(114), [OVR_LBL(115)] = OVR_ELM(115),
+[OVR_LBL(116)] = OVR_ELM(116), [OVR_LBL(117)] = OVR_ELM(117),
+[OVR_LBL(118)] = OVR_ELM(118), [OVR_LBL(119)] = OVR_ELM(119),
+[OVR_LBL(120)] = OVR_ELM(120), [OVR_LBL(121)] = OVR_ELM(121),
+[OVR_LBL(122)] = OVR_ELM(122), [OVR_LBL(123)] = OVR_ELM(123),
+[OVR_LBL(124)] = OVR_ELM(124), [OVR_LBL(125)] = OVR_ELM(125),
+[OVR_LBL(126)] = OVR_ELM(126), [OVR_LBL(127)] = OVR_ELM(127),
+[OVR_LBL(128)] = OVR_ELM(128), [OVR_LBL(129)] = OVR_ELM(129),
+[OVR_LBL(130)] = OVR_ELM(130), [OVR_LBL(131)] = OVR_ELM(131),
+[OVR_LBL(132)] = OVR_ELM(132), [OVR_LBL(133)] = OVR_ELM(133),
+[OVR_LBL(134)] = OVR_ELM(134), [OVR_LBL(135)] = OVR_ELM(135),
+[OVR_LBL(136)] = OVR_ELM(136), [OVR_LBL(137)] = OVR_ELM(137),
+[OVR_LBL(138)] = OVR_ELM(138), [OVR_LBL(139)] = OVR_ELM(139),
+[OVR_LBL(140)] = OVR_ELM(140), [OVR_LBL(141)] = OVR_ELM(141),
+[OVR_LBL(142)] = OVR_ELM(142), [OVR_LBL(143)] = OVR_ELM(143),
+[OVR_LBL(144)] = OVR_ELM(144), [OVR_LBL(145)] = OVR_ELM(145),
+[OVR_LBL(146)] = OVR_ELM(146), [OVR_LBL(147)] = OVR_ELM(147),
+[OVR_LBL(148)] = OVR_ELM(148), [OVR_LBL(149)] = OVR_ELM(149),
+[OVR_LBL(150)] = OVR_ELM(150), [OVR_LBL(151)] = OVR_ELM(151),
+[OVR_LBL(152)] = OVR_ELM(152), [OVR_LBL(153)] = OVR_ELM(153),
+[OVR_LBL(154)] = OVR_ELM(154), [OVR_LBL(155)] = OVR_ELM(155),
+[OVR_LBL(156)] = OVR_ELM(156), [OVR_LBL(157)] = OVR_ELM(157),
+[OVR_LBL(158)] = OVR_ELM(158), [OVR_LBL(159)] = OVR_ELM(159),
+};
+
+/* ======================================================================== */
+
+/* return true if this is chip revision revision a0 */
+int is_a0(struct hfi1_devdata *dd)
+{
+ return ((dd->revision >> CCE_REVISION_CHIP_REV_MINOR_SHIFT)
+ & CCE_REVISION_CHIP_REV_MINOR_MASK) == 0;
+}
+
+/* return true if this is chip revision revision a */
+int is_ax(struct hfi1_devdata *dd)
+{
+ u8 chip_rev_minor =
+ dd->revision >> CCE_REVISION_CHIP_REV_MINOR_SHIFT
+ & CCE_REVISION_CHIP_REV_MINOR_MASK;
+ return (chip_rev_minor & 0xf0) == 0;
+}
+
+/* return true if this is chip revision revision b */
+int is_bx(struct hfi1_devdata *dd)
+{
+ u8 chip_rev_minor =
+ dd->revision >> CCE_REVISION_CHIP_REV_MINOR_SHIFT
+ & CCE_REVISION_CHIP_REV_MINOR_MASK;
+ return !!(chip_rev_minor & 0x10);
+}
+
+/*
+ * Append string s to buffer buf. Arguments curp and len are the current
+ * position and remaining length, respectively.
+ *
+ * return 0 on success, 1 on out of room
+ */
+static int append_str(char *buf, char **curp, int *lenp, const char *s)
+{
+ char *p = *curp;
+ int len = *lenp;
+ int result = 0; /* success */
+ char c;
+
+ /* add a comma, if first in the buffer */
+ if (p != buf) {
+ if (len == 0) {
+ result = 1; /* out of room */
+ goto done;
+ }
+ *p++ = ',';
+ len--;
+ }
+
+ /* copy the string */
+ while ((c = *s++) != 0) {
+ if (len == 0) {
+ result = 1; /* out of room */
+ goto done;
+ }
+ *p++ = c;
+ len--;
+ }
+
+done:
+ /* write return values */
+ *curp = p;
+ *lenp = len;
+
+ return result;
+}
+
+/*
+ * Using the given flag table, print a comma separated string into
+ * the buffer. End in '*' if the buffer is too short.
+ */
+static char *flag_string(char *buf, int buf_len, u64 flags,
+ struct flag_table *table, int table_size)
+{
+ char extra[32];
+ char *p = buf;
+ int len = buf_len;
+ int no_room = 0;
+ int i;
+
+ /* make sure there is at least 2 so we can form "*" */
+ if (len < 2)
+ return "";
+
+ len--; /* leave room for a nul */
+ for (i = 0; i < table_size; i++) {
+ if (flags & table[i].flag) {
+ no_room = append_str(buf, &p, &len, table[i].str);
+ if (no_room)
+ break;
+ flags &= ~table[i].flag;
+ }
+ }
+
+ /* any undocumented bits left? */
+ if (!no_room && flags) {
+ snprintf(extra, sizeof(extra), "bits 0x%llx", flags);
+ no_room = append_str(buf, &p, &len, extra);
+ }
+
+ /* add * if ran out of room */
+ if (no_room) {
+ /* may need to back up to add space for a '*' */
+ if (len == 0)
+ --p;
+ *p++ = '*';
+ }
+
+ /* add final nul - space already allocated above */
+ *p = 0;
+ return buf;
+}
+
+/* first 8 CCE error interrupt source names */
+static const char * const cce_misc_names[] = {
+ "CceErrInt", /* 0 */
+ "RxeErrInt", /* 1 */
+ "MiscErrInt", /* 2 */
+ "Reserved3", /* 3 */
+ "PioErrInt", /* 4 */
+ "SDmaErrInt", /* 5 */
+ "EgressErrInt", /* 6 */
+ "TxeErrInt" /* 7 */
+};
+
+/*
+ * Return the miscellaneous error interrupt name.
+ */
+static char *is_misc_err_name(char *buf, size_t bsize, unsigned int source)
+{
+ if (source < ARRAY_SIZE(cce_misc_names))
+ strncpy(buf, cce_misc_names[source], bsize);
+ else
+ snprintf(buf,
+ bsize,
+ "Reserved%u",
+ source + IS_GENERAL_ERR_START);
+
+ return buf;
+}
+
+/*
+ * Return the SDMA engine error interrupt name.
+ */
+static char *is_sdma_eng_err_name(char *buf, size_t bsize, unsigned int source)
+{
+ snprintf(buf, bsize, "SDmaEngErrInt%u", source);
+ return buf;
+}
+
+/*
+ * Return the send context error interrupt name.
+ */
+static char *is_sendctxt_err_name(char *buf, size_t bsize, unsigned int source)
+{
+ snprintf(buf, bsize, "SendCtxtErrInt%u", source);
+ return buf;
+}
+
+static const char * const various_names[] = {
+ "PbcInt",
+ "GpioAssertInt",
+ "Qsfp1Int",
+ "Qsfp2Int",
+ "TCritInt"
+};
+
+/*
+ * Return the various interrupt name.
+ */
+static char *is_various_name(char *buf, size_t bsize, unsigned int source)
+{
+ if (source < ARRAY_SIZE(various_names))
+ strncpy(buf, various_names[source], bsize);
+ else
+ snprintf(buf, bsize, "Reserved%u", source+IS_VARIOUS_START);
+ return buf;
+}
+
+/*
+ * Return the DC interrupt name.
+ */
+static char *is_dc_name(char *buf, size_t bsize, unsigned int source)
+{
+ static const char * const dc_int_names[] = {
+ "common",
+ "lcb",
+ "8051",
+ "lbm" /* local block merge */
+ };
+
+ if (source < ARRAY_SIZE(dc_int_names))
+ snprintf(buf, bsize, "dc_%s_int", dc_int_names[source]);
+ else
+ snprintf(buf, bsize, "DCInt%u", source);
+ return buf;
+}
+
+static const char * const sdma_int_names[] = {
+ "SDmaInt",
+ "SdmaIdleInt",
+ "SdmaProgressInt",
+};
+
+/*
+ * Return the SDMA engine interrupt name.
+ */
+static char *is_sdma_eng_name(char *buf, size_t bsize, unsigned int source)
+{
+ /* what interrupt */
+ unsigned int what = source / TXE_NUM_SDMA_ENGINES;
+ /* which engine */
+ unsigned int which = source % TXE_NUM_SDMA_ENGINES;
+
+ if (likely(what < 3))
+ snprintf(buf, bsize, "%s%u", sdma_int_names[what], which);
+ else
+ snprintf(buf, bsize, "Invalid SDMA interrupt %u", source);
+ return buf;
+}
+
+/*
+ * Return the receive available interrupt name.
+ */
+static char *is_rcv_avail_name(char *buf, size_t bsize, unsigned int source)
+{
+ snprintf(buf, bsize, "RcvAvailInt%u", source);
+ return buf;
+}
+
+/*
+ * Return the receive urgent interrupt name.
+ */
+static char *is_rcv_urgent_name(char *buf, size_t bsize, unsigned int source)
+{
+ snprintf(buf, bsize, "RcvUrgentInt%u", source);
+ return buf;
+}
+
+/*
+ * Return the send credit interrupt name.
+ */
+static char *is_send_credit_name(char *buf, size_t bsize, unsigned int source)
+{
+ snprintf(buf, bsize, "SendCreditInt%u", source);
+ return buf;
+}
+
+/*
+ * Return the reserved interrupt name.
+ */
+static char *is_reserved_name(char *buf, size_t bsize, unsigned int source)
+{
+ snprintf(buf, bsize, "Reserved%u", source + IS_RESERVED_START);
+ return buf;
+}
+
+static char *cce_err_status_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags,
+ cce_err_status_flags, ARRAY_SIZE(cce_err_status_flags));
+}
+
+static char *rxe_err_status_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags,
+ rxe_err_status_flags, ARRAY_SIZE(rxe_err_status_flags));
+}
+
+static char *misc_err_status_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags, misc_err_status_flags,
+ ARRAY_SIZE(misc_err_status_flags));
+}
+
+static char *pio_err_status_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags,
+ pio_err_status_flags, ARRAY_SIZE(pio_err_status_flags));
+}
+
+static char *sdma_err_status_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags,
+ sdma_err_status_flags,
+ ARRAY_SIZE(sdma_err_status_flags));
+}
+
+static char *egress_err_status_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags,
+ egress_err_status_flags, ARRAY_SIZE(egress_err_status_flags));
+}
+
+static char *egress_err_info_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags,
+ egress_err_info_flags, ARRAY_SIZE(egress_err_info_flags));
+}
+
+static char *send_err_status_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags,
+ send_err_status_flags,
+ ARRAY_SIZE(send_err_status_flags));
+}
+
+static void handle_cce_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ char buf[96];
+
+ /*
+ * For most these errors, there is nothing that can be done except
+ * report or record it.
+ */
+ dd_dev_info(dd, "CCE Error: %s\n",
+ cce_err_status_string(buf, sizeof(buf), reg));
+
+ if ((reg & CCE_ERR_STATUS_CCE_CLI2_ASYNC_FIFO_PARITY_ERR_SMASK)
+ && is_a0(dd)
+ && (dd->icode != ICODE_FUNCTIONAL_SIMULATOR)) {
+ /* this error requires a manual drop into SPC freeze mode */
+ /* then a fix up */
+ start_freeze_handling(dd->pport, FREEZE_SELF);
+ }
+}
+
+/*
+ * Check counters for receive errors that do not have an interrupt
+ * associated with them.
+ */
+#define RCVERR_CHECK_TIME 10
+static void update_rcverr_timer(unsigned long opaque)
+{
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)opaque;
+ struct hfi1_pportdata *ppd = dd->pport;
+ u32 cur_ovfl_cnt = read_dev_cntr(dd, C_RCV_OVF, CNTR_INVALID_VL);
+
+ if (dd->rcv_ovfl_cnt < cur_ovfl_cnt &&
+ ppd->port_error_action & OPA_PI_MASK_EX_BUFFER_OVERRUN) {
+ dd_dev_info(dd, "%s: PortErrorAction bounce\n", __func__);
+ set_link_down_reason(ppd,
+ OPA_LINKDOWN_REASON_EXCESSIVE_BUFFER_OVERRUN, 0,
+ OPA_LINKDOWN_REASON_EXCESSIVE_BUFFER_OVERRUN);
+ queue_work(ppd->hfi1_wq, &ppd->link_bounce_work);
+ }
+ dd->rcv_ovfl_cnt = (u32) cur_ovfl_cnt;
+
+ mod_timer(&dd->rcverr_timer, jiffies + HZ * RCVERR_CHECK_TIME);
+}
+
+static int init_rcverr(struct hfi1_devdata *dd)
+{
+ init_timer(&dd->rcverr_timer);
+ dd->rcverr_timer.function = update_rcverr_timer;
+ dd->rcverr_timer.data = (unsigned long) dd;
+ /* Assume the hardware counter has been reset */
+ dd->rcv_ovfl_cnt = 0;
+ return mod_timer(&dd->rcverr_timer, jiffies + HZ * RCVERR_CHECK_TIME);
+}
+
+static void free_rcverr(struct hfi1_devdata *dd)
+{
+ if (dd->rcverr_timer.data)
+ del_timer_sync(&dd->rcverr_timer);
+ dd->rcverr_timer.data = 0;
+}
+
+static void handle_rxe_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ char buf[96];
+
+ dd_dev_info(dd, "Receive Error: %s\n",
+ rxe_err_status_string(buf, sizeof(buf), reg));
+
+ if (reg & ALL_RXE_FREEZE_ERR) {
+ int flags = 0;
+
+ /*
+ * Freeze mode recovery is disabled for the errors
+ * in RXE_FREEZE_ABORT_MASK
+ */
+ if (is_a0(dd) && (reg & RXE_FREEZE_ABORT_MASK))
+ flags = FREEZE_ABORT;
+
+ start_freeze_handling(dd->pport, flags);
+ }
+}
+
+static void handle_misc_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ char buf[96];
+
+ dd_dev_info(dd, "Misc Error: %s",
+ misc_err_status_string(buf, sizeof(buf), reg));
+}
+
+static void handle_pio_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ char buf[96];
+
+ dd_dev_info(dd, "PIO Error: %s\n",
+ pio_err_status_string(buf, sizeof(buf), reg));
+
+ if (reg & ALL_PIO_FREEZE_ERR)
+ start_freeze_handling(dd->pport, 0);
+}
+
+static void handle_sdma_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ char buf[96];
+
+ dd_dev_info(dd, "SDMA Error: %s\n",
+ sdma_err_status_string(buf, sizeof(buf), reg));
+
+ if (reg & ALL_SDMA_FREEZE_ERR)
+ start_freeze_handling(dd->pport, 0);
+}
+
+static void count_port_inactive(struct hfi1_devdata *dd)
+{
+ struct hfi1_pportdata *ppd = dd->pport;
+
+ if (ppd->port_xmit_discards < ~(u64)0)
+ ppd->port_xmit_discards++;
+}
+
+/*
+ * We have had a "disallowed packet" error during egress. Determine the
+ * integrity check which failed, and update relevant error counter, etc.
+ *
+ * Note that the SEND_EGRESS_ERR_INFO register has only a single
+ * bit of state per integrity check, and so we can miss the reason for an
+ * egress error if more than one packet fails the same integrity check
+ * since we cleared the corresponding bit in SEND_EGRESS_ERR_INFO.
+ */
+static void handle_send_egress_err_info(struct hfi1_devdata *dd)
+{
+ struct hfi1_pportdata *ppd = dd->pport;
+ u64 src = read_csr(dd, SEND_EGRESS_ERR_SOURCE); /* read first */
+ u64 info = read_csr(dd, SEND_EGRESS_ERR_INFO);
+ char buf[96];
+
+ /* clear down all observed info as quickly as possible after read */
+ write_csr(dd, SEND_EGRESS_ERR_INFO, info);
+
+ dd_dev_info(dd,
+ "Egress Error Info: 0x%llx, %s Egress Error Src 0x%llx\n",
+ info, egress_err_info_string(buf, sizeof(buf), info), src);
+
+ /* Eventually add other counters for each bit */
+
+ if (info & SEND_EGRESS_ERR_INFO_TOO_LONG_IB_PACKET_ERR_SMASK) {
+ if (ppd->port_xmit_discards < ~(u64)0)
+ ppd->port_xmit_discards++;
+ }
+}
+
+/*
+ * Input value is a bit position within the SEND_EGRESS_ERR_STATUS
+ * register. Does it represent a 'port inactive' error?
+ */
+static inline int port_inactive_err(u64 posn)
+{
+ return (posn >= SEES(TX_LINKDOWN) &&
+ posn <= SEES(TX_INCORRECT_LINK_STATE));
+}
+
+/*
+ * Input value is a bit position within the SEND_EGRESS_ERR_STATUS
+ * register. Does it represent a 'disallowed packet' error?
+ */
+static inline int disallowed_pkt_err(u64 posn)
+{
+ return (posn >= SEES(TX_SDMA0_DISALLOWED_PACKET) &&
+ posn <= SEES(TX_SDMA15_DISALLOWED_PACKET));
+}
+
+static void handle_egress_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ u64 reg_copy = reg, handled = 0;
+ char buf[96];
+
+ if (reg & ALL_TXE_EGRESS_FREEZE_ERR)
+ start_freeze_handling(dd->pport, 0);
+ if (is_a0(dd) && (reg &
+ SEND_EGRESS_ERR_STATUS_TX_CREDIT_RETURN_VL_ERR_SMASK)
+ && (dd->icode != ICODE_FUNCTIONAL_SIMULATOR))
+ start_freeze_handling(dd->pport, 0);
+
+ while (reg_copy) {
+ int posn = fls64(reg_copy);
+ /*
+ * fls64() returns a 1-based offset, but we generally
+ * want 0-based offsets.
+ */
+ int shift = posn - 1;
+
+ if (port_inactive_err(shift)) {
+ count_port_inactive(dd);
+ handled |= (1ULL << shift);
+ } else if (disallowed_pkt_err(shift)) {
+ handle_send_egress_err_info(dd);
+ handled |= (1ULL << shift);
+ }
+ clear_bit(shift, (unsigned long *)&reg_copy);
+ }
+
+ reg &= ~handled;
+
+ if (reg)
+ dd_dev_info(dd, "Egress Error: %s\n",
+ egress_err_status_string(buf, sizeof(buf), reg));
+}
+
+static void handle_txe_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ char buf[96];
+
+ dd_dev_info(dd, "Send Error: %s\n",
+ send_err_status_string(buf, sizeof(buf), reg));
+
+}
+
+/*
+ * The maximum number of times the error clear down will loop before
+ * blocking a repeating error. This value is arbitrary.
+ */
+#define MAX_CLEAR_COUNT 20
+
+/*
+ * Clear and handle an error register. All error interrupts are funneled
+ * through here to have a central location to correctly handle single-
+ * or multi-shot errors.
+ *
+ * For non per-context registers, call this routine with a context value
+ * of 0 so the per-context offset is zero.
+ *
+ * If the handler loops too many times, assume that something is wrong
+ * and can't be fixed, so mask the error bits.
+ */
+static void interrupt_clear_down(struct hfi1_devdata *dd,
+ u32 context,
+ const struct err_reg_info *eri)
+{
+ u64 reg;
+ u32 count;
+
+ /* read in a loop until no more errors are seen */
+ count = 0;
+ while (1) {
+ reg = read_kctxt_csr(dd, context, eri->status);
+ if (reg == 0)
+ break;
+ write_kctxt_csr(dd, context, eri->clear, reg);
+ if (likely(eri->handler))
+ eri->handler(dd, context, reg);
+ count++;
+ if (count > MAX_CLEAR_COUNT) {
+ u64 mask;
+
+ dd_dev_err(dd, "Repeating %s bits 0x%llx - masking\n",
+ eri->desc, reg);
+ /*
+ * Read-modify-write so any other masked bits
+ * remain masked.
+ */
+ mask = read_kctxt_csr(dd, context, eri->mask);
+ mask &= ~reg;
+ write_kctxt_csr(dd, context, eri->mask, mask);
+ break;
+ }
+ }
+}
+
+/*
+ * CCE block "misc" interrupt. Source is < 16.
+ */
+static void is_misc_err_int(struct hfi1_devdata *dd, unsigned int source)
+{
+ const struct err_reg_info *eri = &misc_errs[source];
+
+ if (eri->handler) {
+ interrupt_clear_down(dd, 0, eri);
+ } else {
+ dd_dev_err(dd, "Unexpected misc interrupt (%u) - reserved\n",
+ source);
+ }
+}
+
+static char *send_context_err_status_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags,
+ sc_err_status_flags, ARRAY_SIZE(sc_err_status_flags));
+}
+
+/*
+ * Send context error interrupt. Source (hw_context) is < 160.
+ *
+ * All send context errors cause the send context to halt. The normal
+ * clear-down mechanism cannot be used because we cannot clear the
+ * error bits until several other long-running items are done first.
+ * This is OK because with the context halted, nothing else is going
+ * to happen on it anyway.
+ */
+static void is_sendctxt_err_int(struct hfi1_devdata *dd,
+ unsigned int hw_context)
+{
+ struct send_context_info *sci;
+ struct send_context *sc;
+ char flags[96];
+ u64 status;
+ u32 sw_index;
+
+ sw_index = dd->hw_to_sw[hw_context];
+ if (sw_index >= dd->num_send_contexts) {
+ dd_dev_err(dd,
+ "out of range sw index %u for send context %u\n",
+ sw_index, hw_context);
+ return;
+ }
+ sci = &dd->send_contexts[sw_index];
+ sc = sci->sc;
+ if (!sc) {
+ dd_dev_err(dd, "%s: context %u(%u): no sc?\n", __func__,
+ sw_index, hw_context);
+ return;
+ }
+
+ /* tell the software that a halt has begun */
+ sc_stop(sc, SCF_HALTED);
+
+ status = read_kctxt_csr(dd, hw_context, SEND_CTXT_ERR_STATUS);
+
+ dd_dev_info(dd, "Send Context %u(%u) Error: %s\n", sw_index, hw_context,
+ send_context_err_status_string(flags, sizeof(flags), status));
+
+ if (status & SEND_CTXT_ERR_STATUS_PIO_DISALLOWED_PACKET_ERR_SMASK)
+ handle_send_egress_err_info(dd);
+
+ /*
+ * Automatically restart halted kernel contexts out of interrupt
+ * context. User contexts must ask the driver to restart the context.
+ */
+ if (sc->type != SC_USER)
+ queue_work(dd->pport->hfi1_wq, &sc->halt_work);
+}
+
+static void handle_sdma_eng_err(struct hfi1_devdata *dd,
+ unsigned int source, u64 status)
+{
+ struct sdma_engine *sde;
+
+ sde = &dd->per_sdma[source];
+#ifdef CONFIG_SDMA_VERBOSITY
+ dd_dev_err(sde->dd, "CONFIG SDMA(%u) %s:%d %s()\n", sde->this_idx,
+ slashstrip(__FILE__), __LINE__, __func__);
+ dd_dev_err(sde->dd, "CONFIG SDMA(%u) source: %u status 0x%llx\n",
+ sde->this_idx, source, (unsigned long long)status);
+#endif
+ sdma_engine_error(sde, status);
+}
+
+/*
+ * CCE block SDMA error interrupt. Source is < 16.
+ */
+static void is_sdma_eng_err_int(struct hfi1_devdata *dd, unsigned int source)
+{
+#ifdef CONFIG_SDMA_VERBOSITY
+ struct sdma_engine *sde = &dd->per_sdma[source];
+
+ dd_dev_err(dd, "CONFIG SDMA(%u) %s:%d %s()\n", sde->this_idx,
+ slashstrip(__FILE__), __LINE__, __func__);
+ dd_dev_err(dd, "CONFIG SDMA(%u) source: %u\n", sde->this_idx,
+ source);
+ sdma_dumpstate(sde);
+#endif
+ interrupt_clear_down(dd, source, &sdma_eng_err);
+}
+
+/*
+ * CCE block "various" interrupt. Source is < 8.
+ */
+static void is_various_int(struct hfi1_devdata *dd, unsigned int source)
+{
+ const struct err_reg_info *eri = &various_err[source];
+
+ /*
+ * TCritInt cannot go through interrupt_clear_down()
+ * because it is not a second tier interrupt. The handler
+ * should be called directly.
+ */
+ if (source == TCRIT_INT_SOURCE)
+ handle_temp_err(dd);
+ else if (eri->handler)
+ interrupt_clear_down(dd, 0, eri);
+ else
+ dd_dev_info(dd,
+ "%s: Unimplemented/reserved interrupt %d\n",
+ __func__, source);
+}
+
+static void handle_qsfp_int(struct hfi1_devdata *dd, u32 src_ctx, u64 reg)
+{
+ /* source is always zero */
+ struct hfi1_pportdata *ppd = dd->pport;
+ unsigned long flags;
+ u64 qsfp_int_mgmt = (u64)(QSFP_HFI0_INT_N | QSFP_HFI0_MODPRST_N);
+
+ if (reg & QSFP_HFI0_MODPRST_N) {
+
+ dd_dev_info(dd, "%s: ModPresent triggered QSFP interrupt\n",
+ __func__);
+
+ if (!qsfp_mod_present(ppd)) {
+ ppd->driver_link_ready = 0;
+ /*
+ * Cable removed, reset all our information about the
+ * cache and cable capabilities
+ */
+
+ spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
+ /*
+ * We don't set cache_refresh_required here as we expect
+ * an interrupt when a cable is inserted
+ */
+ ppd->qsfp_info.cache_valid = 0;
+ ppd->qsfp_info.qsfp_interrupt_functional = 0;
+ spin_unlock_irqrestore(&ppd->qsfp_info.qsfp_lock,
+ flags);
+ write_csr(dd,
+ dd->hfi1_id ?
+ ASIC_QSFP2_INVERT :
+ ASIC_QSFP1_INVERT,
+ qsfp_int_mgmt);
+ if (ppd->host_link_state == HLS_DN_POLL) {
+ /*
+ * The link is still in POLL. This means
+ * that the normal link down processing
+ * will not happen. We have to do it here
+ * before turning the DC off.
+ */
+ queue_work(ppd->hfi1_wq, &ppd->link_down_work);
+ }
+ } else {
+ spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
+ ppd->qsfp_info.cache_valid = 0;
+ ppd->qsfp_info.cache_refresh_required = 1;
+ spin_unlock_irqrestore(&ppd->qsfp_info.qsfp_lock,
+ flags);
+
+ qsfp_int_mgmt &= ~(u64)QSFP_HFI0_MODPRST_N;
+ write_csr(dd,
+ dd->hfi1_id ?
+ ASIC_QSFP2_INVERT :
+ ASIC_QSFP1_INVERT,
+ qsfp_int_mgmt);
+ }
+ }
+
+ if (reg & QSFP_HFI0_INT_N) {
+
+ dd_dev_info(dd, "%s: IntN triggered QSFP interrupt\n",
+ __func__);
+ spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
+ ppd->qsfp_info.check_interrupt_flags = 1;
+ ppd->qsfp_info.qsfp_interrupt_functional = 1;
+ spin_unlock_irqrestore(&ppd->qsfp_info.qsfp_lock, flags);
+ }
+
+ /* Schedule the QSFP work only if there is a cable attached. */
+ if (qsfp_mod_present(ppd))
+ queue_work(ppd->hfi1_wq, &ppd->qsfp_info.qsfp_work);
+}
+
+static int request_host_lcb_access(struct hfi1_devdata *dd)
+{
+ int ret;
+
+ ret = do_8051_command(dd, HCMD_MISC,
+ (u64)HCMD_MISC_REQUEST_LCB_ACCESS << LOAD_DATA_FIELD_ID_SHIFT,
+ NULL);
+ if (ret != HCMD_SUCCESS) {
+ dd_dev_err(dd, "%s: command failed with error %d\n",
+ __func__, ret);
+ }
+ return ret == HCMD_SUCCESS ? 0 : -EBUSY;
+}
+
+static int request_8051_lcb_access(struct hfi1_devdata *dd)
+{
+ int ret;
+
+ ret = do_8051_command(dd, HCMD_MISC,
+ (u64)HCMD_MISC_GRANT_LCB_ACCESS << LOAD_DATA_FIELD_ID_SHIFT,
+ NULL);
+ if (ret != HCMD_SUCCESS) {
+ dd_dev_err(dd, "%s: command failed with error %d\n",
+ __func__, ret);
+ }
+ return ret == HCMD_SUCCESS ? 0 : -EBUSY;
+}
+
+/*
+ * Set the LCB selector - allow host access. The DCC selector always
+ * points to the host.
+ */
+static inline void set_host_lcb_access(struct hfi1_devdata *dd)
+{
+ write_csr(dd, DC_DC8051_CFG_CSR_ACCESS_SEL,
+ DC_DC8051_CFG_CSR_ACCESS_SEL_DCC_SMASK
+ | DC_DC8051_CFG_CSR_ACCESS_SEL_LCB_SMASK);
+}
+
+/*
+ * Clear the LCB selector - allow 8051 access. The DCC selector always
+ * points to the host.
+ */
+static inline void set_8051_lcb_access(struct hfi1_devdata *dd)
+{
+ write_csr(dd, DC_DC8051_CFG_CSR_ACCESS_SEL,
+ DC_DC8051_CFG_CSR_ACCESS_SEL_DCC_SMASK);
+}
+
+/*
+ * Acquire LCB access from the 8051. If the host already has access,
+ * just increment a counter. Otherwise, inform the 8051 that the
+ * host is taking access.
+ *
+ * Returns:
+ * 0 on success
+ * -EBUSY if the 8051 has control and cannot be disturbed
+ * -errno if unable to acquire access from the 8051
+ */
+int acquire_lcb_access(struct hfi1_devdata *dd, int sleep_ok)
+{
+ struct hfi1_pportdata *ppd = dd->pport;
+ int ret = 0;
+
+ /*
+ * Use the host link state lock so the operation of this routine
+ * { link state check, selector change, count increment } can occur
+ * as a unit against a link state change. Otherwise there is a
+ * race between the state change and the count increment.
+ */
+ if (sleep_ok) {
+ mutex_lock(&ppd->hls_lock);
+ } else {
+ while (mutex_trylock(&ppd->hls_lock) == EBUSY)
+ udelay(1);
+ }
+
+ /* this access is valid only when the link is up */
+ if ((ppd->host_link_state & HLS_UP) == 0) {
+ dd_dev_info(dd, "%s: link state %s not up\n",
+ __func__, link_state_name(ppd->host_link_state));
+ ret = -EBUSY;
+ goto done;
+ }
+
+ if (dd->lcb_access_count == 0) {
+ ret = request_host_lcb_access(dd);
+ if (ret) {
+ dd_dev_err(dd,
+ "%s: unable to acquire LCB access, err %d\n",
+ __func__, ret);
+ goto done;
+ }
+ set_host_lcb_access(dd);
+ }
+ dd->lcb_access_count++;
+done:
+ mutex_unlock(&ppd->hls_lock);
+ return ret;
+}
+
+/*
+ * Release LCB access by decrementing the use count. If the count is moving
+ * from 1 to 0, inform 8051 that it has control back.
+ *
+ * Returns:
+ * 0 on success
+ * -errno if unable to release access to the 8051
+ */
+int release_lcb_access(struct hfi1_devdata *dd, int sleep_ok)
+{
+ int ret = 0;
+
+ /*
+ * Use the host link state lock because the acquire needed it.
+ * Here, we only need to keep { selector change, count decrement }
+ * as a unit.
+ */
+ if (sleep_ok) {
+ mutex_lock(&dd->pport->hls_lock);
+ } else {
+ while (mutex_trylock(&dd->pport->hls_lock) == EBUSY)
+ udelay(1);
+ }
+
+ if (dd->lcb_access_count == 0) {
+ dd_dev_err(dd, "%s: LCB access count is zero. Skipping.\n",
+ __func__);
+ goto done;
+ }
+
+ if (dd->lcb_access_count == 1) {
+ set_8051_lcb_access(dd);
+ ret = request_8051_lcb_access(dd);
+ if (ret) {
+ dd_dev_err(dd,
+ "%s: unable to release LCB access, err %d\n",
+ __func__, ret);
+ /* restore host access if the grant didn't work */
+ set_host_lcb_access(dd);
+ goto done;
+ }
+ }
+ dd->lcb_access_count--;
+done:
+ mutex_unlock(&dd->pport->hls_lock);
+ return ret;
+}
+
+/*
+ * Initialize LCB access variables and state. Called during driver load,
+ * after most of the initialization is finished.
+ *
+ * The DC default is LCB access on for the host. The driver defaults to
+ * leaving access to the 8051. Assign access now - this constrains the call
+ * to this routine to be after all LCB set-up is done. In particular, after
+ * hf1_init_dd() -> set_up_interrupts() -> clear_all_interrupts()
+ */
+static void init_lcb_access(struct hfi1_devdata *dd)
+{
+ dd->lcb_access_count = 0;
+}
+
+/*
+ * Write a response back to a 8051 request.
+ */
+static void hreq_response(struct hfi1_devdata *dd, u8 return_code, u16 rsp_data)
+{
+ write_csr(dd, DC_DC8051_CFG_EXT_DEV_0,
+ DC_DC8051_CFG_EXT_DEV_0_COMPLETED_SMASK
+ | (u64)return_code << DC_DC8051_CFG_EXT_DEV_0_RETURN_CODE_SHIFT
+ | (u64)rsp_data << DC_DC8051_CFG_EXT_DEV_0_RSP_DATA_SHIFT);
+}
+
+/*
+ * Handle requests from the 8051.
+ */
+static void handle_8051_request(struct hfi1_devdata *dd)
+{
+ u64 reg;
+ u16 data;
+ u8 type;
+
+ reg = read_csr(dd, DC_DC8051_CFG_EXT_DEV_1);
+ if ((reg & DC_DC8051_CFG_EXT_DEV_1_REQ_NEW_SMASK) == 0)
+ return; /* no request */
+
+ /* zero out COMPLETED so the response is seen */
+ write_csr(dd, DC_DC8051_CFG_EXT_DEV_0, 0);
+
+ /* extract request details */
+ type = (reg >> DC_DC8051_CFG_EXT_DEV_1_REQ_TYPE_SHIFT)
+ & DC_DC8051_CFG_EXT_DEV_1_REQ_TYPE_MASK;
+ data = (reg >> DC_DC8051_CFG_EXT_DEV_1_REQ_DATA_SHIFT)
+ & DC_DC8051_CFG_EXT_DEV_1_REQ_DATA_MASK;
+
+ switch (type) {
+ case HREQ_LOAD_CONFIG:
+ case HREQ_SAVE_CONFIG:
+ case HREQ_READ_CONFIG:
+ case HREQ_SET_TX_EQ_ABS:
+ case HREQ_SET_TX_EQ_REL:
+ case HREQ_ENABLE:
+ dd_dev_info(dd, "8051 request: request 0x%x not supported\n",
+ type);
+ hreq_response(dd, HREQ_NOT_SUPPORTED, 0);
+ break;
+
+ case HREQ_CONFIG_DONE:
+ hreq_response(dd, HREQ_SUCCESS, 0);
+ break;
+
+ case HREQ_INTERFACE_TEST:
+ hreq_response(dd, HREQ_SUCCESS, data);
+ break;
+
+ default:
+ dd_dev_err(dd, "8051 request: unknown request 0x%x\n", type);
+ hreq_response(dd, HREQ_NOT_SUPPORTED, 0);
+ break;
+ }
+}
+
+static void write_global_credit(struct hfi1_devdata *dd,
+ u8 vau, u16 total, u16 shared)
+{
+ write_csr(dd, SEND_CM_GLOBAL_CREDIT,
+ ((u64)total
+ << SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_SHIFT)
+ | ((u64)shared
+ << SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_SHIFT)
+ | ((u64)vau << SEND_CM_GLOBAL_CREDIT_AU_SHIFT));
+}
+
+/*
+ * Set up initial VL15 credits of the remote. Assumes the rest of
+ * the CM credit registers are zero from a previous global or credit reset .
+ */
+void set_up_vl15(struct hfi1_devdata *dd, u8 vau, u16 vl15buf)
+{
+ /* leave shared count at zero for both global and VL15 */
+ write_global_credit(dd, vau, vl15buf, 0);
+
+ /* We may need some credits for another VL when sending packets
+ * with the snoop interface. Dividing it down the middle for VL15
+ * and VL0 should suffice.
+ */
+ if (unlikely(dd->hfi1_snoop.mode_flag == HFI1_PORT_SNOOP_MODE)) {
+ write_csr(dd, SEND_CM_CREDIT_VL15, (u64)(vl15buf >> 1)
+ << SEND_CM_CREDIT_VL15_DEDICATED_LIMIT_VL_SHIFT);
+ write_csr(dd, SEND_CM_CREDIT_VL, (u64)(vl15buf >> 1)
+ << SEND_CM_CREDIT_VL_DEDICATED_LIMIT_VL_SHIFT);
+ } else {
+ write_csr(dd, SEND_CM_CREDIT_VL15, (u64)vl15buf
+ << SEND_CM_CREDIT_VL15_DEDICATED_LIMIT_VL_SHIFT);
+ }
+}
+
+/*
+ * Zero all credit details from the previous connection and
+ * reset the CM manager's internal counters.
+ */
+void reset_link_credits(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /* remove all previous VL credit limits */
+ for (i = 0; i < TXE_NUM_DATA_VL; i++)
+ write_csr(dd, SEND_CM_CREDIT_VL + (8*i), 0);
+ write_csr(dd, SEND_CM_CREDIT_VL15, 0);
+ write_global_credit(dd, 0, 0, 0);
+ /* reset the CM block */
+ pio_send_control(dd, PSC_CM_RESET);
+}
+
+/* convert a vCU to a CU */
+static u32 vcu_to_cu(u8 vcu)
+{
+ return 1 << vcu;
+}
+
+/* convert a CU to a vCU */
+static u8 cu_to_vcu(u32 cu)
+{
+ return ilog2(cu);
+}
+
+/* convert a vAU to an AU */
+static u32 vau_to_au(u8 vau)
+{
+ return 8 * (1 << vau);
+}
+
+static void set_linkup_defaults(struct hfi1_pportdata *ppd)
+{
+ ppd->sm_trap_qp = 0x0;
+ ppd->sa_qp = 0x1;
+}
+
+/*
+ * Graceful LCB shutdown. This leaves the LCB FIFOs in reset.
+ */
+static void lcb_shutdown(struct hfi1_devdata *dd, int abort)
+{
+ u64 reg;
+
+ /* clear lcb run: LCB_CFG_RUN.EN = 0 */
+ write_csr(dd, DC_LCB_CFG_RUN, 0);
+ /* set tx fifo reset: LCB_CFG_TX_FIFOS_RESET.VAL = 1 */
+ write_csr(dd, DC_LCB_CFG_TX_FIFOS_RESET,
+ 1ull << DC_LCB_CFG_TX_FIFOS_RESET_VAL_SHIFT);
+ /* set dcc reset csr: DCC_CFG_RESET.{reset_lcb,reset_rx_fpe} = 1 */
+ dd->lcb_err_en = read_csr(dd, DC_LCB_ERR_EN);
+ reg = read_csr(dd, DCC_CFG_RESET);
+ write_csr(dd, DCC_CFG_RESET,
+ reg
+ | (1ull << DCC_CFG_RESET_RESET_LCB_SHIFT)
+ | (1ull << DCC_CFG_RESET_RESET_RX_FPE_SHIFT));
+ (void) read_csr(dd, DCC_CFG_RESET); /* make sure the write completed */
+ if (!abort) {
+ udelay(1); /* must hold for the longer of 16cclks or 20ns */
+ write_csr(dd, DCC_CFG_RESET, reg);
+ write_csr(dd, DC_LCB_ERR_EN, dd->lcb_err_en);
+ }
+}
+
+/*
+ * This routine should be called after the link has been transitioned to
+ * OFFLINE (OFFLINE state has the side effect of putting the SerDes into
+ * reset).
+ *
+ * The expectation is that the caller of this routine would have taken
+ * care of properly transitioning the link into the correct state.
+ */
+static void dc_shutdown(struct hfi1_devdata *dd)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&dd->dc8051_lock, flags);
+ if (dd->dc_shutdown) {
+ spin_unlock_irqrestore(&dd->dc8051_lock, flags);
+ return;
+ }
+ dd->dc_shutdown = 1;
+ spin_unlock_irqrestore(&dd->dc8051_lock, flags);
+ /* Shutdown the LCB */
+ lcb_shutdown(dd, 1);
+ /* Going to OFFLINE would have causes the 8051 to put the
+ * SerDes into reset already. Just need to shut down the 8051,
+ * itself. */
+ write_csr(dd, DC_DC8051_CFG_RST, 0x1);
+}
+
+/* Calling this after the DC has been brought out of reset should not
+ * do any damage. */
+static void dc_start(struct hfi1_devdata *dd)
+{
+ unsigned long flags;
+ int ret;
+
+ spin_lock_irqsave(&dd->dc8051_lock, flags);
+ if (!dd->dc_shutdown)
+ goto done;
+ spin_unlock_irqrestore(&dd->dc8051_lock, flags);
+ /* Take the 8051 out of reset */
+ write_csr(dd, DC_DC8051_CFG_RST, 0ull);
+ /* Wait until 8051 is ready */
+ ret = wait_fm_ready(dd, TIMEOUT_8051_START);
+ if (ret) {
+ dd_dev_err(dd, "%s: timeout starting 8051 firmware\n",
+ __func__);
+ }
+ /* Take away reset for LCB and RX FPE (set in lcb_shutdown). */
+ write_csr(dd, DCC_CFG_RESET, 0x10);
+ /* lcb_shutdown() with abort=1 does not restore these */
+ write_csr(dd, DC_LCB_ERR_EN, dd->lcb_err_en);
+ spin_lock_irqsave(&dd->dc8051_lock, flags);
+ dd->dc_shutdown = 0;
+done:
+ spin_unlock_irqrestore(&dd->dc8051_lock, flags);
+}
+
+/*
+ * These LCB adjustments are for the Aurora SerDes core in the FPGA.
+ */
+static void adjust_lcb_for_fpga_serdes(struct hfi1_devdata *dd)
+{
+ u64 rx_radr, tx_radr;
+ u32 version;
+
+ if (dd->icode != ICODE_FPGA_EMULATION)
+ return;
+
+ /*
+ * These LCB defaults on emulator _s are good, nothing to do here:
+ * LCB_CFG_TX_FIFOS_RADR
+ * LCB_CFG_RX_FIFOS_RADR
+ * LCB_CFG_LN_DCLK
+ * LCB_CFG_IGNORE_LOST_RCLK
+ */
+ if (is_emulator_s(dd))
+ return;
+ /* else this is _p */
+
+ version = emulator_rev(dd);
+ if (!is_a0(dd))
+ version = 0x2d; /* all B0 use 0x2d or higher settings */
+
+ if (version <= 0x12) {
+ /* release 0x12 and below */
+
+ /*
+ * LCB_CFG_RX_FIFOS_RADR.RST_VAL = 0x9
+ * LCB_CFG_RX_FIFOS_RADR.OK_TO_JUMP_VAL = 0x9
+ * LCB_CFG_RX_FIFOS_RADR.DO_NOT_JUMP_VAL = 0xa
+ */
+ rx_radr =
+ 0xaull << DC_LCB_CFG_RX_FIFOS_RADR_DO_NOT_JUMP_VAL_SHIFT
+ | 0x9ull << DC_LCB_CFG_RX_FIFOS_RADR_OK_TO_JUMP_VAL_SHIFT
+ | 0x9ull << DC_LCB_CFG_RX_FIFOS_RADR_RST_VAL_SHIFT;
+ /*
+ * LCB_CFG_TX_FIFOS_RADR.ON_REINIT = 0 (default)
+ * LCB_CFG_TX_FIFOS_RADR.RST_VAL = 6
+ */
+ tx_radr = 6ull << DC_LCB_CFG_TX_FIFOS_RADR_RST_VAL_SHIFT;
+ } else if (version <= 0x18) {
+ /* release 0x13 up to 0x18 */
+ /* LCB_CFG_RX_FIFOS_RADR = 0x988 */
+ rx_radr =
+ 0x9ull << DC_LCB_CFG_RX_FIFOS_RADR_DO_NOT_JUMP_VAL_SHIFT
+ | 0x8ull << DC_LCB_CFG_RX_FIFOS_RADR_OK_TO_JUMP_VAL_SHIFT
+ | 0x8ull << DC_LCB_CFG_RX_FIFOS_RADR_RST_VAL_SHIFT;
+ tx_radr = 7ull << DC_LCB_CFG_TX_FIFOS_RADR_RST_VAL_SHIFT;
+ } else if (version == 0x19) {
+ /* release 0x19 */
+ /* LCB_CFG_RX_FIFOS_RADR = 0xa99 */
+ rx_radr =
+ 0xAull << DC_LCB_CFG_RX_FIFOS_RADR_DO_NOT_JUMP_VAL_SHIFT
+ | 0x9ull << DC_LCB_CFG_RX_FIFOS_RADR_OK_TO_JUMP_VAL_SHIFT
+ | 0x9ull << DC_LCB_CFG_RX_FIFOS_RADR_RST_VAL_SHIFT;
+ tx_radr = 3ull << DC_LCB_CFG_TX_FIFOS_RADR_RST_VAL_SHIFT;
+ } else if (version == 0x1a) {
+ /* release 0x1a */
+ /* LCB_CFG_RX_FIFOS_RADR = 0x988 */
+ rx_radr =
+ 0x9ull << DC_LCB_CFG_RX_FIFOS_RADR_DO_NOT_JUMP_VAL_SHIFT
+ | 0x8ull << DC_LCB_CFG_RX_FIFOS_RADR_OK_TO_JUMP_VAL_SHIFT
+ | 0x8ull << DC_LCB_CFG_RX_FIFOS_RADR_RST_VAL_SHIFT;
+ tx_radr = 7ull << DC_LCB_CFG_TX_FIFOS_RADR_RST_VAL_SHIFT;
+ write_csr(dd, DC_LCB_CFG_LN_DCLK, 1ull);
+ } else {
+ /* release 0x1b and higher */
+ /* LCB_CFG_RX_FIFOS_RADR = 0x877 */
+ rx_radr =
+ 0x8ull << DC_LCB_CFG_RX_FIFOS_RADR_DO_NOT_JUMP_VAL_SHIFT
+ | 0x7ull << DC_LCB_CFG_RX_FIFOS_RADR_OK_TO_JUMP_VAL_SHIFT
+ | 0x7ull << DC_LCB_CFG_RX_FIFOS_RADR_RST_VAL_SHIFT;
+ tx_radr = 3ull << DC_LCB_CFG_TX_FIFOS_RADR_RST_VAL_SHIFT;
+ }
+
+ write_csr(dd, DC_LCB_CFG_RX_FIFOS_RADR, rx_radr);
+ /* LCB_CFG_IGNORE_LOST_RCLK.EN = 1 */
+ write_csr(dd, DC_LCB_CFG_IGNORE_LOST_RCLK,
+ DC_LCB_CFG_IGNORE_LOST_RCLK_EN_SMASK);
+ write_csr(dd, DC_LCB_CFG_TX_FIFOS_RADR, tx_radr);
+}
+
+/*
+ * Handle a SMA idle message
+ *
+ * This is a work-queue function outside of the interrupt.
+ */
+void handle_sma_message(struct work_struct *work)
+{
+ struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+ sma_message_work);
+ struct hfi1_devdata *dd = ppd->dd;
+ u64 msg;
+ int ret;
+
+ /* msg is bytes 1-4 of the 40-bit idle message - the command code
+ is stripped off */
+ ret = read_idle_sma(dd, &msg);
+ if (ret)
+ return;
+ dd_dev_info(dd, "%s: SMA message 0x%llx\n", __func__, msg);
+ /*
+ * React to the SMA message. Byte[1] (0 for us) is the command.
+ */
+ switch (msg & 0xff) {
+ case SMA_IDLE_ARM:
+ /*
+ * See OPAv1 table 9-14 - HFI and External Switch Ports Key
+ * State Transitions
+ *
+ * Only expected in INIT or ARMED, discard otherwise.
+ */
+ if (ppd->host_link_state & (HLS_UP_INIT | HLS_UP_ARMED))
+ ppd->neighbor_normal = 1;
+ break;
+ case SMA_IDLE_ACTIVE:
+ /*
+ * See OPAv1 table 9-14 - HFI and External Switch Ports Key
+ * State Transitions
+ *
+ * Can activate the node. Discard otherwise.
+ */
+ if (ppd->host_link_state == HLS_UP_ARMED
+ && ppd->is_active_optimize_enabled) {
+ ppd->neighbor_normal = 1;
+ ret = set_link_state(ppd, HLS_UP_ACTIVE);
+ if (ret)
+ dd_dev_err(
+ dd,
+ "%s: received Active SMA idle message, couldn't set link to Active\n",
+ __func__);
+ }
+ break;
+ default:
+ dd_dev_err(dd,
+ "%s: received unexpected SMA idle message 0x%llx\n",
+ __func__, msg);
+ break;
+ }
+}
+
+static void adjust_rcvctrl(struct hfi1_devdata *dd, u64 add, u64 clear)
+{
+ u64 rcvctrl;
+ unsigned long flags;
+
+ spin_lock_irqsave(&dd->rcvctrl_lock, flags);
+ rcvctrl = read_csr(dd, RCV_CTRL);
+ rcvctrl |= add;
+ rcvctrl &= ~clear;
+ write_csr(dd, RCV_CTRL, rcvctrl);
+ spin_unlock_irqrestore(&dd->rcvctrl_lock, flags);
+}
+
+static inline void add_rcvctrl(struct hfi1_devdata *dd, u64 add)
+{
+ adjust_rcvctrl(dd, add, 0);
+}
+
+static inline void clear_rcvctrl(struct hfi1_devdata *dd, u64 clear)
+{
+ adjust_rcvctrl(dd, 0, clear);
+}
+
+/*
+ * Called from all interrupt handlers to start handling an SPC freeze.
+ */
+void start_freeze_handling(struct hfi1_pportdata *ppd, int flags)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ struct send_context *sc;
+ int i;
+
+ if (flags & FREEZE_SELF)
+ write_csr(dd, CCE_CTRL, CCE_CTRL_SPC_FREEZE_SMASK);
+
+ /* enter frozen mode */
+ dd->flags |= HFI1_FROZEN;
+
+ /* notify all SDMA engines that they are going into a freeze */
+ sdma_freeze_notify(dd, !!(flags & FREEZE_LINK_DOWN));
+
+ /* do halt pre-handling on all enabled send contexts */
+ for (i = 0; i < dd->num_send_contexts; i++) {
+ sc = dd->send_contexts[i].sc;
+ if (sc && (sc->flags & SCF_ENABLED))
+ sc_stop(sc, SCF_FROZEN | SCF_HALTED);
+ }
+
+ /* Send context are frozen. Notify user space */
+ hfi1_set_uevent_bits(ppd, _HFI1_EVENT_FROZEN_BIT);
+
+ if (flags & FREEZE_ABORT) {
+ dd_dev_err(dd,
+ "Aborted freeze recovery. Please REBOOT system\n");
+ return;
+ }
+ /* queue non-interrupt handler */
+ queue_work(ppd->hfi1_wq, &ppd->freeze_work);
+}
+
+/*
+ * Wait until all 4 sub-blocks indicate that they have frozen or unfrozen,
+ * depending on the "freeze" parameter.
+ *
+ * No need to return an error if it times out, our only option
+ * is to proceed anyway.
+ */
+static void wait_for_freeze_status(struct hfi1_devdata *dd, int freeze)
+{
+ unsigned long timeout;
+ u64 reg;
+
+ timeout = jiffies + msecs_to_jiffies(FREEZE_STATUS_TIMEOUT);
+ while (1) {
+ reg = read_csr(dd, CCE_STATUS);
+ if (freeze) {
+ /* waiting until all indicators are set */
+ if ((reg & ALL_FROZE) == ALL_FROZE)
+ return; /* all done */
+ } else {
+ /* waiting until all indicators are clear */
+ if ((reg & ALL_FROZE) == 0)
+ return; /* all done */
+ }
+
+ if (time_after(jiffies, timeout)) {
+ dd_dev_err(dd,
+ "Time out waiting for SPC %sfreeze, bits 0x%llx, expecting 0x%llx, continuing",
+ freeze ? "" : "un",
+ reg & ALL_FROZE,
+ freeze ? ALL_FROZE : 0ull);
+ return;
+ }
+ usleep_range(80, 120);
+ }
+}
+
+/*
+ * Do all freeze handling for the RXE block.
+ */
+static void rxe_freeze(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /* disable port */
+ clear_rcvctrl(dd, RCV_CTRL_RCV_PORT_ENABLE_SMASK);
+
+ /* disable all receive contexts */
+ for (i = 0; i < dd->num_rcv_contexts; i++)
+ hfi1_rcvctrl(dd, HFI1_RCVCTRL_CTXT_DIS, i);
+}
+
+/*
+ * Unfreeze handling for the RXE block - kernel contexts only.
+ * This will also enable the port. User contexts will do unfreeze
+ * handling on a per-context basis as they call into the driver.
+ *
+ */
+static void rxe_kernel_unfreeze(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /* enable all kernel contexts */
+ for (i = 0; i < dd->n_krcv_queues; i++)
+ hfi1_rcvctrl(dd, HFI1_RCVCTRL_CTXT_ENB, i);
+
+ /* enable port */
+ add_rcvctrl(dd, RCV_CTRL_RCV_PORT_ENABLE_SMASK);
+}
+
+/*
+ * Non-interrupt SPC freeze handling.
+ *
+ * This is a work-queue function outside of the triggering interrupt.
+ */
+void handle_freeze(struct work_struct *work)
+{
+ struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+ freeze_work);
+ struct hfi1_devdata *dd = ppd->dd;
+
+ /* wait for freeze indicators on all affected blocks */
+ dd_dev_info(dd, "Entering SPC freeze\n");
+ wait_for_freeze_status(dd, 1);
+
+ /* SPC is now frozen */
+
+ /* do send PIO freeze steps */
+ pio_freeze(dd);
+
+ /* do send DMA freeze steps */
+ sdma_freeze(dd);
+
+ /* do send egress freeze steps - nothing to do */
+
+ /* do receive freeze steps */
+ rxe_freeze(dd);
+
+ /*
+ * Unfreeze the hardware - clear the freeze, wait for each
+ * block's frozen bit to clear, then clear the frozen flag.
+ */
+ write_csr(dd, CCE_CTRL, CCE_CTRL_SPC_UNFREEZE_SMASK);
+ wait_for_freeze_status(dd, 0);
+
+ if (is_a0(dd)) {
+ write_csr(dd, CCE_CTRL, CCE_CTRL_SPC_FREEZE_SMASK);
+ wait_for_freeze_status(dd, 1);
+ write_csr(dd, CCE_CTRL, CCE_CTRL_SPC_UNFREEZE_SMASK);
+ wait_for_freeze_status(dd, 0);
+ }
+
+ /* do send PIO unfreeze steps for kernel contexts */
+ pio_kernel_unfreeze(dd);
+
+ /* do send DMA unfreeze steps */
+ sdma_unfreeze(dd);
+
+ /* do send egress unfreeze steps - nothing to do */
+
+ /* do receive unfreeze steps for kernel contexts */
+ rxe_kernel_unfreeze(dd);
+
+ /*
+ * The unfreeze procedure touches global device registers when
+ * it disables and re-enables RXE. Mark the device unfrozen
+ * after all that is done so other parts of the driver waiting
+ * for the device to unfreeze don't do things out of order.
+ *
+ * The above implies that the meaning of HFI1_FROZEN flag is
+ * "Device has gone into freeze mode and freeze mode handling
+ * is still in progress."
+ *
+ * The flag will be removed when freeze mode processing has
+ * completed.
+ */
+ dd->flags &= ~HFI1_FROZEN;
+ wake_up(&dd->event_queue);
+
+ /* no longer frozen */
+ dd_dev_err(dd, "Exiting SPC freeze\n");
+}
+
+/*
+ * Handle a link up interrupt from the 8051.
+ *
+ * This is a work-queue function outside of the interrupt.
+ */
+void handle_link_up(struct work_struct *work)
+{
+ struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+ link_up_work);
+ set_link_state(ppd, HLS_UP_INIT);
+
+ /* cache the read of DC_LCB_STS_ROUND_TRIP_LTP_CNT */
+ read_ltp_rtt(ppd->dd);
+ /*
+ * OPA specifies that certain counters are cleared on a transition
+ * to link up, so do that.
+ */
+ clear_linkup_counters(ppd->dd);
+ /*
+ * And (re)set link up default values.
+ */
+ set_linkup_defaults(ppd);
+
+ /* enforce link speed enabled */
+ if ((ppd->link_speed_active & ppd->link_speed_enabled) == 0) {
+ /* oops - current speed is not enabled, bounce */
+ dd_dev_err(ppd->dd,
+ "Link speed active 0x%x is outside enabled 0x%x, downing link\n",
+ ppd->link_speed_active, ppd->link_speed_enabled);
+ set_link_down_reason(ppd, OPA_LINKDOWN_REASON_SPEED_POLICY, 0,
+ OPA_LINKDOWN_REASON_SPEED_POLICY);
+ set_link_state(ppd, HLS_DN_OFFLINE);
+ start_link(ppd);
+ }
+}
+
+/* Several pieces of LNI information were cached for SMA in ppd.
+ * Reset these on link down */
+static void reset_neighbor_info(struct hfi1_pportdata *ppd)
+{
+ ppd->neighbor_guid = 0;
+ ppd->neighbor_port_number = 0;
+ ppd->neighbor_type = 0;
+ ppd->neighbor_fm_security = 0;
+}
+
+/*
+ * Handle a link down interrupt from the 8051.
+ *
+ * This is a work-queue function outside of the interrupt.
+ */
+void handle_link_down(struct work_struct *work)
+{
+ u8 lcl_reason, neigh_reason = 0;
+ struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+ link_down_work);
+
+ /* go offline first, then deal with reasons */
+ set_link_state(ppd, HLS_DN_OFFLINE);
+
+ lcl_reason = 0;
+ read_planned_down_reason_code(ppd->dd, &neigh_reason);
+
+ /*
+ * If no reason, assume peer-initiated but missed
+ * LinkGoingDown idle flits.
+ */
+ if (neigh_reason == 0)
+ lcl_reason = OPA_LINKDOWN_REASON_NEIGHBOR_UNKNOWN;
+
+ set_link_down_reason(ppd, lcl_reason, neigh_reason, 0);
+
+ reset_neighbor_info(ppd);
+
+ /* disable the port */
+ clear_rcvctrl(ppd->dd, RCV_CTRL_RCV_PORT_ENABLE_SMASK);
+
+ /* If there is no cable attached, turn the DC off. Otherwise,
+ * start the link bring up. */
+ if (!qsfp_mod_present(ppd))
+ dc_shutdown(ppd->dd);
+ else
+ start_link(ppd);
+}
+
+void handle_link_bounce(struct work_struct *work)
+{
+ struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+ link_bounce_work);
+
+ /*
+ * Only do something if the link is currently up.
+ */
+ if (ppd->host_link_state & HLS_UP) {
+ set_link_state(ppd, HLS_DN_OFFLINE);
+ start_link(ppd);
+ } else {
+ dd_dev_info(ppd->dd, "%s: link not up (%s), nothing to do\n",
+ __func__, link_state_name(ppd->host_link_state));
+ }
+}
+
+/*
+ * Mask conversion: Capability exchange to Port LTP. The capability
+ * exchange has an implicit 16b CRC that is mandatory.
+ */
+static int cap_to_port_ltp(int cap)
+{
+ int port_ltp = PORT_LTP_CRC_MODE_16; /* this mode is mandatory */
+
+ if (cap & CAP_CRC_14B)
+ port_ltp |= PORT_LTP_CRC_MODE_14;
+ if (cap & CAP_CRC_48B)
+ port_ltp |= PORT_LTP_CRC_MODE_48;
+ if (cap & CAP_CRC_12B_16B_PER_LANE)
+ port_ltp |= PORT_LTP_CRC_MODE_PER_LANE;
+
+ return port_ltp;
+}
+
+/*
+ * Convert an OPA Port LTP mask to capability mask
+ */
+int port_ltp_to_cap(int port_ltp)
+{
+ int cap_mask = 0;
+
+ if (port_ltp & PORT_LTP_CRC_MODE_14)
+ cap_mask |= CAP_CRC_14B;
+ if (port_ltp & PORT_LTP_CRC_MODE_48)
+ cap_mask |= CAP_CRC_48B;
+ if (port_ltp & PORT_LTP_CRC_MODE_PER_LANE)
+ cap_mask |= CAP_CRC_12B_16B_PER_LANE;
+
+ return cap_mask;
+}
+
+/*
+ * Convert a single DC LCB CRC mode to an OPA Port LTP mask.
+ */
+static int lcb_to_port_ltp(int lcb_crc)
+{
+ int port_ltp = 0;
+
+ if (lcb_crc == LCB_CRC_12B_16B_PER_LANE)
+ port_ltp = PORT_LTP_CRC_MODE_PER_LANE;
+ else if (lcb_crc == LCB_CRC_48B)
+ port_ltp = PORT_LTP_CRC_MODE_48;
+ else if (lcb_crc == LCB_CRC_14B)
+ port_ltp = PORT_LTP_CRC_MODE_14;
+ else
+ port_ltp = PORT_LTP_CRC_MODE_16;
+
+ return port_ltp;
+}
+
+/*
+ * Our neighbor has indicated that we are allowed to act as a fabric
+ * manager, so place the full management partition key in the second
+ * (0-based) pkey array position (see OPAv1, section 20.2.2.6.8). Note
+ * that we should already have the limited management partition key in
+ * array element 1, and also that the port is not yet up when
+ * add_full_mgmt_pkey() is invoked.
+ */
+static void add_full_mgmt_pkey(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+
+ /* Sanity check - ppd->pkeys[2] should be 0 */
+ if (ppd->pkeys[2] != 0)
+ dd_dev_err(dd, "%s pkey[2] already set to 0x%x, resetting it to 0x%x\n",
+ __func__, ppd->pkeys[2], FULL_MGMT_P_KEY);
+ ppd->pkeys[2] = FULL_MGMT_P_KEY;
+ (void)hfi1_set_ib_cfg(ppd, HFI1_IB_CFG_PKEYS, 0);
+}
+
+/*
+ * Convert the given link width to the OPA link width bitmask.
+ */
+static u16 link_width_to_bits(struct hfi1_devdata *dd, u16 width)
+{
+ switch (width) {
+ case 0:
+ /*
+ * Simulator and quick linkup do not set the width.
+ * Just set it to 4x without complaint.
+ */
+ if (dd->icode == ICODE_FUNCTIONAL_SIMULATOR || quick_linkup)
+ return OPA_LINK_WIDTH_4X;
+ return 0; /* no lanes up */
+ case 1: return OPA_LINK_WIDTH_1X;
+ case 2: return OPA_LINK_WIDTH_2X;
+ case 3: return OPA_LINK_WIDTH_3X;
+ default:
+ dd_dev_info(dd, "%s: invalid width %d, using 4\n",
+ __func__, width);
+ /* fall through */
+ case 4: return OPA_LINK_WIDTH_4X;
+ }
+}
+
+/*
+ * Do a population count on the bottom nibble.
+ */
+static const u8 bit_counts[16] = {
+ 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4
+};
+static inline u8 nibble_to_count(u8 nibble)
+{
+ return bit_counts[nibble & 0xf];
+}
+
+/*
+ * Read the active lane information from the 8051 registers and return
+ * their widths.
+ *
+ * Active lane information is found in these 8051 registers:
+ * enable_lane_tx
+ * enable_lane_rx
+ */
+static void get_link_widths(struct hfi1_devdata *dd, u16 *tx_width,
+ u16 *rx_width)
+{
+ u16 tx, rx;
+ u8 enable_lane_rx;
+ u8 enable_lane_tx;
+ u8 tx_polarity_inversion;
+ u8 rx_polarity_inversion;
+ u8 max_rate;
+
+ /* read the active lanes */
+ read_tx_settings(dd, &enable_lane_tx, &tx_polarity_inversion,
+ &rx_polarity_inversion, &max_rate);
+ read_local_lni(dd, &enable_lane_rx);
+
+ /* convert to counts */
+ tx = nibble_to_count(enable_lane_tx);
+ rx = nibble_to_count(enable_lane_rx);
+
+ /*
+ * Set link_speed_active here, overriding what was set in
+ * handle_verify_cap(). The ASIC 8051 firmware does not correctly
+ * set the max_rate field in handle_verify_cap until v0.19.
+ */
+ if ((dd->icode == ICODE_RTL_SILICON)
+ && (dd->dc8051_ver < dc8051_ver(0, 19))) {
+ /* max_rate: 0 = 12.5G, 1 = 25G */
+ switch (max_rate) {
+ case 0:
+ dd->pport[0].link_speed_active = OPA_LINK_SPEED_12_5G;
+ break;
+ default:
+ dd_dev_err(dd,
+ "%s: unexpected max rate %d, using 25Gb\n",
+ __func__, (int)max_rate);
+ /* fall through */
+ case 1:
+ dd->pport[0].link_speed_active = OPA_LINK_SPEED_25G;
+ break;
+ }
+ }
+
+ dd_dev_info(dd,
+ "Fabric active lanes (width): tx 0x%x (%d), rx 0x%x (%d)\n",
+ enable_lane_tx, tx, enable_lane_rx, rx);
+ *tx_width = link_width_to_bits(dd, tx);
+ *rx_width = link_width_to_bits(dd, rx);
+}
+
+/*
+ * Read verify_cap_local_fm_link_width[1] to obtain the link widths.
+ * Valid after the end of VerifyCap and during LinkUp. Does not change
+ * after link up. I.e. look elsewhere for downgrade information.
+ *
+ * Bits are:
+ * + bits [7:4] contain the number of active transmitters
+ * + bits [3:0] contain the number of active receivers
+ * These are numbers 1 through 4 and can be different values if the
+ * link is asymmetric.
+ *
+ * verify_cap_local_fm_link_width[0] retains its original value.
+ */
+static void get_linkup_widths(struct hfi1_devdata *dd, u16 *tx_width,
+ u16 *rx_width)
+{
+ u16 widths, tx, rx;
+ u8 misc_bits, local_flags;
+ u16 active_tx, active_rx;
+
+ read_vc_local_link_width(dd, &misc_bits, &local_flags, &widths);
+ tx = widths >> 12;
+ rx = (widths >> 8) & 0xf;
+
+ *tx_width = link_width_to_bits(dd, tx);
+ *rx_width = link_width_to_bits(dd, rx);
+
+ /* print the active widths */
+ get_link_widths(dd, &active_tx, &active_rx);
+}
+
+/*
+ * Set ppd->link_width_active and ppd->link_width_downgrade_active using
+ * hardware information when the link first comes up.
+ *
+ * The link width is not available until after VerifyCap.AllFramesReceived
+ * (the trigger for handle_verify_cap), so this is outside that routine
+ * and should be called when the 8051 signals linkup.
+ */
+void get_linkup_link_widths(struct hfi1_pportdata *ppd)
+{
+ u16 tx_width, rx_width;
+
+ /* get end-of-LNI link widths */
+ get_linkup_widths(ppd->dd, &tx_width, &rx_width);
+
+ /* use tx_width as the link is supposed to be symmetric on link up */
+ ppd->link_width_active = tx_width;
+ /* link width downgrade active (LWD.A) starts out matching LW.A */
+ ppd->link_width_downgrade_tx_active = ppd->link_width_active;
+ ppd->link_width_downgrade_rx_active = ppd->link_width_active;
+ /* per OPA spec, on link up LWD.E resets to LWD.S */
+ ppd->link_width_downgrade_enabled = ppd->link_width_downgrade_supported;
+ /* cache the active egress rate (units {10^6 bits/sec]) */
+ ppd->current_egress_rate = active_egress_rate(ppd);
+}
+
+/*
+ * Handle a verify capabilities interrupt from the 8051.
+ *
+ * This is a work-queue function outside of the interrupt.
+ */
+void handle_verify_cap(struct work_struct *work)
+{
+ struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+ link_vc_work);
+ struct hfi1_devdata *dd = ppd->dd;
+ u64 reg;
+ u8 power_management;
+ u8 continious;
+ u8 vcu;
+ u8 vau;
+ u8 z;
+ u16 vl15buf;
+ u16 link_widths;
+ u16 crc_mask;
+ u16 crc_val;
+ u16 device_id;
+ u16 active_tx, active_rx;
+ u8 partner_supported_crc;
+ u8 remote_tx_rate;
+ u8 device_rev;
+
+ set_link_state(ppd, HLS_VERIFY_CAP);
+
+ lcb_shutdown(dd, 0);
+ adjust_lcb_for_fpga_serdes(dd);
+
+ /*
+ * These are now valid:
+ * remote VerifyCap fields in the general LNI config
+ * CSR DC8051_STS_REMOTE_GUID
+ * CSR DC8051_STS_REMOTE_NODE_TYPE
+ * CSR DC8051_STS_REMOTE_FM_SECURITY
+ * CSR DC8051_STS_REMOTE_PORT_NO
+ */
+
+ read_vc_remote_phy(dd, &power_management, &continious);
+ read_vc_remote_fabric(
+ dd,
+ &vau,
+ &z,
+ &vcu,
+ &vl15buf,
+ &partner_supported_crc);
+ read_vc_remote_link_width(dd, &remote_tx_rate, &link_widths);
+ read_remote_device_id(dd, &device_id, &device_rev);
+ /*
+ * And the 'MgmtAllowed' information, which is exchanged during
+ * LNI, is also be available at this point.
+ */
+ read_mgmt_allowed(dd, &ppd->mgmt_allowed);
+ /* print the active widths */
+ get_link_widths(dd, &active_tx, &active_rx);
+ dd_dev_info(dd,
+ "Peer PHY: power management 0x%x, continuous updates 0x%x\n",
+ (int)power_management, (int)continious);
+ dd_dev_info(dd,
+ "Peer Fabric: vAU %d, Z %d, vCU %d, vl15 credits 0x%x, CRC sizes 0x%x\n",
+ (int)vau,
+ (int)z,
+ (int)vcu,
+ (int)vl15buf,
+ (int)partner_supported_crc);
+ dd_dev_info(dd, "Peer Link Width: tx rate 0x%x, widths 0x%x\n",
+ (u32)remote_tx_rate, (u32)link_widths);
+ dd_dev_info(dd, "Peer Device ID: 0x%04x, Revision 0x%02x\n",
+ (u32)device_id, (u32)device_rev);
+ /*
+ * The peer vAU value just read is the peer receiver value. HFI does
+ * not support a transmit vAU of 0 (AU == 8). We advertised that
+ * with Z=1 in the fabric capabilities sent to the peer. The peer
+ * will see our Z=1, and, if it advertised a vAU of 0, will move its
+ * receive to vAU of 1 (AU == 16). Do the same here. We do not care
+ * about the peer Z value - our sent vAU is 3 (hardwired) and is not
+ * subject to the Z value exception.
+ */
+ if (vau == 0)
+ vau = 1;
+ set_up_vl15(dd, vau, vl15buf);
+
+ /* set up the LCB CRC mode */
+ crc_mask = ppd->port_crc_mode_enabled & partner_supported_crc;
+
+ /* order is important: use the lowest bit in common */
+ if (crc_mask & CAP_CRC_14B)
+ crc_val = LCB_CRC_14B;
+ else if (crc_mask & CAP_CRC_48B)
+ crc_val = LCB_CRC_48B;
+ else if (crc_mask & CAP_CRC_12B_16B_PER_LANE)
+ crc_val = LCB_CRC_12B_16B_PER_LANE;
+ else
+ crc_val = LCB_CRC_16B;
+
+ dd_dev_info(dd, "Final LCB CRC mode: %d\n", (int)crc_val);
+ write_csr(dd, DC_LCB_CFG_CRC_MODE,
+ (u64)crc_val << DC_LCB_CFG_CRC_MODE_TX_VAL_SHIFT);
+
+ /* set (14b only) or clear sideband credit */
+ reg = read_csr(dd, SEND_CM_CTRL);
+ if (crc_val == LCB_CRC_14B && crc_14b_sideband) {
+ write_csr(dd, SEND_CM_CTRL,
+ reg | SEND_CM_CTRL_FORCE_CREDIT_MODE_SMASK);
+ } else {
+ write_csr(dd, SEND_CM_CTRL,
+ reg & ~SEND_CM_CTRL_FORCE_CREDIT_MODE_SMASK);
+ }
+
+ ppd->link_speed_active = 0; /* invalid value */
+ if (dd->dc8051_ver < dc8051_ver(0, 20)) {
+ /* remote_tx_rate: 0 = 12.5G, 1 = 25G */
+ switch (remote_tx_rate) {
+ case 0:
+ ppd->link_speed_active = OPA_LINK_SPEED_12_5G;
+ break;
+ case 1:
+ ppd->link_speed_active = OPA_LINK_SPEED_25G;
+ break;
+ }
+ } else {
+ /* actual rate is highest bit of the ANDed rates */
+ u8 rate = remote_tx_rate & ppd->local_tx_rate;
+
+ if (rate & 2)
+ ppd->link_speed_active = OPA_LINK_SPEED_25G;
+ else if (rate & 1)
+ ppd->link_speed_active = OPA_LINK_SPEED_12_5G;
+ }
+ if (ppd->link_speed_active == 0) {
+ dd_dev_err(dd, "%s: unexpected remote tx rate %d, using 25Gb\n",
+ __func__, (int)remote_tx_rate);
+ ppd->link_speed_active = OPA_LINK_SPEED_25G;
+ }
+
+ /*
+ * Cache the values of the supported, enabled, and active
+ * LTP CRC modes to return in 'portinfo' queries. But the bit
+ * flags that are returned in the portinfo query differ from
+ * what's in the link_crc_mask, crc_sizes, and crc_val
+ * variables. Convert these here.
+ */
+ ppd->port_ltp_crc_mode = cap_to_port_ltp(link_crc_mask) << 8;
+ /* supported crc modes */
+ ppd->port_ltp_crc_mode |=
+ cap_to_port_ltp(ppd->port_crc_mode_enabled) << 4;
+ /* enabled crc modes */
+ ppd->port_ltp_crc_mode |= lcb_to_port_ltp(crc_val);
+ /* active crc mode */
+
+ /* set up the remote credit return table */
+ assign_remote_cm_au_table(dd, vcu);
+
+ /*
+ * The LCB is reset on entry to handle_verify_cap(), so this must
+ * be applied on every link up.
+ *
+ * Adjust LCB error kill enable to kill the link if
+ * these RBUF errors are seen:
+ * REPLAY_BUF_MBE_SMASK
+ * FLIT_INPUT_BUF_MBE_SMASK
+ */
+ if (is_a0(dd)) { /* fixed in B0 */
+ reg = read_csr(dd, DC_LCB_CFG_LINK_KILL_EN);
+ reg |= DC_LCB_CFG_LINK_KILL_EN_REPLAY_BUF_MBE_SMASK
+ | DC_LCB_CFG_LINK_KILL_EN_FLIT_INPUT_BUF_MBE_SMASK;
+ write_csr(dd, DC_LCB_CFG_LINK_KILL_EN, reg);
+ }
+
+ /* pull LCB fifos out of reset - all fifo clocks must be stable */
+ write_csr(dd, DC_LCB_CFG_TX_FIFOS_RESET, 0);
+
+ /* give 8051 access to the LCB CSRs */
+ write_csr(dd, DC_LCB_ERR_EN, 0); /* mask LCB errors */
+ set_8051_lcb_access(dd);
+
+ ppd->neighbor_guid =
+ read_csr(dd, DC_DC8051_STS_REMOTE_GUID);
+ ppd->neighbor_port_number = read_csr(dd, DC_DC8051_STS_REMOTE_PORT_NO) &
+ DC_DC8051_STS_REMOTE_PORT_NO_VAL_SMASK;
+ ppd->neighbor_type =
+ read_csr(dd, DC_DC8051_STS_REMOTE_NODE_TYPE) &
+ DC_DC8051_STS_REMOTE_NODE_TYPE_VAL_MASK;
+ ppd->neighbor_fm_security =
+ read_csr(dd, DC_DC8051_STS_REMOTE_FM_SECURITY) &
+ DC_DC8051_STS_LOCAL_FM_SECURITY_DISABLED_MASK;
+ dd_dev_info(dd,
+ "Neighbor Guid: %llx Neighbor type %d MgmtAllowed %d FM security bypass %d\n",
+ ppd->neighbor_guid, ppd->neighbor_type,
+ ppd->mgmt_allowed, ppd->neighbor_fm_security);
+ if (ppd->mgmt_allowed)
+ add_full_mgmt_pkey(ppd);
+
+ /* tell the 8051 to go to LinkUp */
+ set_link_state(ppd, HLS_GOING_UP);
+}
+
+/*
+ * Apply the link width downgrade enabled policy against the current active
+ * link widths.
+ *
+ * Called when the enabled policy changes or the active link widths change.
+ */
+void apply_link_downgrade_policy(struct hfi1_pportdata *ppd, int refresh_widths)
+{
+ int skip = 1;
+ int do_bounce = 0;
+ u16 lwde = ppd->link_width_downgrade_enabled;
+ u16 tx, rx;
+
+ mutex_lock(&ppd->hls_lock);
+ /* only apply if the link is up */
+ if (ppd->host_link_state & HLS_UP)
+ skip = 0;
+ mutex_unlock(&ppd->hls_lock);
+ if (skip)
+ return;
+
+ if (refresh_widths) {
+ get_link_widths(ppd->dd, &tx, &rx);
+ ppd->link_width_downgrade_tx_active = tx;
+ ppd->link_width_downgrade_rx_active = rx;
+ }
+
+ if (lwde == 0) {
+ /* downgrade is disabled */
+
+ /* bounce if not at starting active width */
+ if ((ppd->link_width_active !=
+ ppd->link_width_downgrade_tx_active)
+ || (ppd->link_width_active !=
+ ppd->link_width_downgrade_rx_active)) {
+ dd_dev_err(ppd->dd,
+ "Link downgrade is disabled and link has downgraded, downing link\n");
+ dd_dev_err(ppd->dd,
+ " original 0x%x, tx active 0x%x, rx active 0x%x\n",
+ ppd->link_width_active,
+ ppd->link_width_downgrade_tx_active,
+ ppd->link_width_downgrade_rx_active);
+ do_bounce = 1;
+ }
+ } else if ((lwde & ppd->link_width_downgrade_tx_active) == 0
+ || (lwde & ppd->link_width_downgrade_rx_active) == 0) {
+ /* Tx or Rx is outside the enabled policy */
+ dd_dev_err(ppd->dd,
+ "Link is outside of downgrade allowed, downing link\n");
+ dd_dev_err(ppd->dd,
+ " enabled 0x%x, tx active 0x%x, rx active 0x%x\n",
+ lwde,
+ ppd->link_width_downgrade_tx_active,
+ ppd->link_width_downgrade_rx_active);
+ do_bounce = 1;
+ }
+
+ if (do_bounce) {
+ set_link_down_reason(ppd, OPA_LINKDOWN_REASON_WIDTH_POLICY, 0,
+ OPA_LINKDOWN_REASON_WIDTH_POLICY);
+ set_link_state(ppd, HLS_DN_OFFLINE);
+ start_link(ppd);
+ }
+}
+
+/*
+ * Handle a link downgrade interrupt from the 8051.
+ *
+ * This is a work-queue function outside of the interrupt.
+ */
+void handle_link_downgrade(struct work_struct *work)
+{
+ struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+ link_downgrade_work);
+
+ dd_dev_info(ppd->dd, "8051: Link width downgrade\n");
+ apply_link_downgrade_policy(ppd, 1);
+}
+
+static char *dcc_err_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags, dcc_err_flags,
+ ARRAY_SIZE(dcc_err_flags));
+}
+
+static char *lcb_err_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags, lcb_err_flags,
+ ARRAY_SIZE(lcb_err_flags));
+}
+
+static char *dc8051_err_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags, dc8051_err_flags,
+ ARRAY_SIZE(dc8051_err_flags));
+}
+
+static char *dc8051_info_err_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags, dc8051_info_err_flags,
+ ARRAY_SIZE(dc8051_info_err_flags));
+}
+
+static char *dc8051_info_host_msg_string(char *buf, int buf_len, u64 flags)
+{
+ return flag_string(buf, buf_len, flags, dc8051_info_host_msg_flags,
+ ARRAY_SIZE(dc8051_info_host_msg_flags));
+}
+
+static void handle_8051_interrupt(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ struct hfi1_pportdata *ppd = dd->pport;
+ u64 info, err, host_msg;
+ int queue_link_down = 0;
+ char buf[96];
+
+ /* look at the flags */
+ if (reg & DC_DC8051_ERR_FLG_SET_BY_8051_SMASK) {
+ /* 8051 information set by firmware */
+ /* read DC8051_DBG_ERR_INFO_SET_BY_8051 for details */
+ info = read_csr(dd, DC_DC8051_DBG_ERR_INFO_SET_BY_8051);
+ err = (info >> DC_DC8051_DBG_ERR_INFO_SET_BY_8051_ERROR_SHIFT)
+ & DC_DC8051_DBG_ERR_INFO_SET_BY_8051_ERROR_MASK;
+ host_msg = (info >>
+ DC_DC8051_DBG_ERR_INFO_SET_BY_8051_HOST_MSG_SHIFT)
+ & DC_DC8051_DBG_ERR_INFO_SET_BY_8051_HOST_MSG_MASK;
+
+ /*
+ * Handle error flags.
+ */
+ if (err & FAILED_LNI) {
+ /*
+ * LNI error indications are cleared by the 8051
+ * only when starting polling. Only pay attention
+ * to them when in the states that occur during
+ * LNI.
+ */
+ if (ppd->host_link_state
+ & (HLS_DN_POLL | HLS_VERIFY_CAP | HLS_GOING_UP)) {
+ queue_link_down = 1;
+ dd_dev_info(dd, "Link error: %s\n",
+ dc8051_info_err_string(buf,
+ sizeof(buf),
+ err & FAILED_LNI));
+ }
+ err &= ~(u64)FAILED_LNI;
+ }
+ if (err) {
+ /* report remaining errors, but do not do anything */
+ dd_dev_err(dd, "8051 info error: %s\n",
+ dc8051_info_err_string(buf, sizeof(buf), err));
+ }
+
+ /*
+ * Handle host message flags.
+ */
+ if (host_msg & HOST_REQ_DONE) {
+ /*
+ * Presently, the driver does a busy wait for
+ * host requests to complete. This is only an
+ * informational message.
+ * NOTE: The 8051 clears the host message
+ * information *on the next 8051 command*.
+ * Therefore, when linkup is achieved,
+ * this flag will still be set.
+ */
+ host_msg &= ~(u64)HOST_REQ_DONE;
+ }
+ if (host_msg & BC_SMA_MSG) {
+ queue_work(ppd->hfi1_wq, &ppd->sma_message_work);
+ host_msg &= ~(u64)BC_SMA_MSG;
+ }
+ if (host_msg & LINKUP_ACHIEVED) {
+ dd_dev_info(dd, "8051: Link up\n");
+ queue_work(ppd->hfi1_wq, &ppd->link_up_work);
+ host_msg &= ~(u64)LINKUP_ACHIEVED;
+ }
+ if (host_msg & EXT_DEVICE_CFG_REQ) {
+ handle_8051_request(dd);
+ host_msg &= ~(u64)EXT_DEVICE_CFG_REQ;
+ }
+ if (host_msg & VERIFY_CAP_FRAME) {
+ queue_work(ppd->hfi1_wq, &ppd->link_vc_work);
+ host_msg &= ~(u64)VERIFY_CAP_FRAME;
+ }
+ if (host_msg & LINK_GOING_DOWN) {
+ const char *extra = "";
+ /* no downgrade action needed if going down */
+ if (host_msg & LINK_WIDTH_DOWNGRADED) {
+ host_msg &= ~(u64)LINK_WIDTH_DOWNGRADED;
+ extra = " (ignoring downgrade)";
+ }
+ dd_dev_info(dd, "8051: Link down%s\n", extra);
+ queue_link_down = 1;
+ host_msg &= ~(u64)LINK_GOING_DOWN;
+ }
+ if (host_msg & LINK_WIDTH_DOWNGRADED) {
+ queue_work(ppd->hfi1_wq, &ppd->link_downgrade_work);
+ host_msg &= ~(u64)LINK_WIDTH_DOWNGRADED;
+ }
+ if (host_msg) {
+ /* report remaining messages, but do not do anything */
+ dd_dev_info(dd, "8051 info host message: %s\n",
+ dc8051_info_host_msg_string(buf, sizeof(buf),
+ host_msg));
+ }
+
+ reg &= ~DC_DC8051_ERR_FLG_SET_BY_8051_SMASK;
+ }
+ if (reg & DC_DC8051_ERR_FLG_LOST_8051_HEART_BEAT_SMASK) {
+ /*
+ * Lost the 8051 heartbeat. If this happens, we
+ * receive constant interrupts about it. Disable
+ * the interrupt after the first.
+ */
+ dd_dev_err(dd, "Lost 8051 heartbeat\n");
+ write_csr(dd, DC_DC8051_ERR_EN,
+ read_csr(dd, DC_DC8051_ERR_EN)
+ & ~DC_DC8051_ERR_EN_LOST_8051_HEART_BEAT_SMASK);
+
+ reg &= ~DC_DC8051_ERR_FLG_LOST_8051_HEART_BEAT_SMASK;
+ }
+ if (reg) {
+ /* report the error, but do not do anything */
+ dd_dev_err(dd, "8051 error: %s\n",
+ dc8051_err_string(buf, sizeof(buf), reg));
+ }
+
+ if (queue_link_down) {
+ /* if the link is already going down or disabled, do not
+ * queue another */
+ if ((ppd->host_link_state
+ & (HLS_GOING_OFFLINE|HLS_LINK_COOLDOWN))
+ || ppd->link_enabled == 0) {
+ dd_dev_info(dd, "%s: not queuing link down\n",
+ __func__);
+ } else {
+ queue_work(ppd->hfi1_wq, &ppd->link_down_work);
+ }
+ }
+}
+
+static const char * const fm_config_txt[] = {
+[0] =
+ "BadHeadDist: Distance violation between two head flits",
+[1] =
+ "BadTailDist: Distance violation between two tail flits",
+[2] =
+ "BadCtrlDist: Distance violation between two credit control flits",
+[3] =
+ "BadCrdAck: Credits return for unsupported VL",
+[4] =
+ "UnsupportedVLMarker: Received VL Marker",
+[5] =
+ "BadPreempt: Exceeded the preemption nesting level",
+[6] =
+ "BadControlFlit: Received unsupported control flit",
+/* no 7 */
+[8] =
+ "UnsupportedVLMarker: Received VL Marker for unconfigured or disabled VL",
+};
+
+static const char * const port_rcv_txt[] = {
+[1] =
+ "BadPktLen: Illegal PktLen",
+[2] =
+ "PktLenTooLong: Packet longer than PktLen",
+[3] =
+ "PktLenTooShort: Packet shorter than PktLen",
+[4] =
+ "BadSLID: Illegal SLID (0, using multicast as SLID, does not include security validation of SLID)",
+[5] =
+ "BadDLID: Illegal DLID (0, doesn't match HFI)",
+[6] =
+ "BadL2: Illegal L2 opcode",
+[7] =
+ "BadSC: Unsupported SC",
+[9] =
+ "BadRC: Illegal RC",
+[11] =
+ "PreemptError: Preempting with same VL",
+[12] =
+ "PreemptVL15: Preempting a VL15 packet",
+};
+
+#define OPA_LDR_FMCONFIG_OFFSET 16
+#define OPA_LDR_PORTRCV_OFFSET 0
+static void handle_dcc_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ u64 info, hdr0, hdr1;
+ const char *extra;
+ char buf[96];
+ struct hfi1_pportdata *ppd = dd->pport;
+ u8 lcl_reason = 0;
+ int do_bounce = 0;
+
+ if (reg & DCC_ERR_FLG_UNCORRECTABLE_ERR_SMASK) {
+ if (!(dd->err_info_uncorrectable & OPA_EI_STATUS_SMASK)) {
+ info = read_csr(dd, DCC_ERR_INFO_UNCORRECTABLE);
+ dd->err_info_uncorrectable = info & OPA_EI_CODE_SMASK;
+ /* set status bit */
+ dd->err_info_uncorrectable |= OPA_EI_STATUS_SMASK;
+ }
+ reg &= ~DCC_ERR_FLG_UNCORRECTABLE_ERR_SMASK;
+ }
+
+ if (reg & DCC_ERR_FLG_LINK_ERR_SMASK) {
+ struct hfi1_pportdata *ppd = dd->pport;
+ /* this counter saturates at (2^32) - 1 */
+ if (ppd->link_downed < (u32)UINT_MAX)
+ ppd->link_downed++;
+ reg &= ~DCC_ERR_FLG_LINK_ERR_SMASK;
+ }
+
+ if (reg & DCC_ERR_FLG_FMCONFIG_ERR_SMASK) {
+ u8 reason_valid = 1;
+
+ info = read_csr(dd, DCC_ERR_INFO_FMCONFIG);
+ if (!(dd->err_info_fmconfig & OPA_EI_STATUS_SMASK)) {
+ dd->err_info_fmconfig = info & OPA_EI_CODE_SMASK;
+ /* set status bit */
+ dd->err_info_fmconfig |= OPA_EI_STATUS_SMASK;
+ }
+ switch (info) {
+ case 0:
+ case 1:
+ case 2:
+ case 3:
+ case 4:
+ case 5:
+ case 6:
+ extra = fm_config_txt[info];
+ break;
+ case 8:
+ extra = fm_config_txt[info];
+ if (ppd->port_error_action &
+ OPA_PI_MASK_FM_CFG_UNSUPPORTED_VL_MARKER) {
+ do_bounce = 1;
+ /*
+ * lcl_reason cannot be derived from info
+ * for this error
+ */
+ lcl_reason =
+ OPA_LINKDOWN_REASON_UNSUPPORTED_VL_MARKER;
+ }
+ break;
+ default:
+ reason_valid = 0;
+ snprintf(buf, sizeof(buf), "reserved%lld", info);
+ extra = buf;
+ break;
+ }
+
+ if (reason_valid && !do_bounce) {
+ do_bounce = ppd->port_error_action &
+ (1 << (OPA_LDR_FMCONFIG_OFFSET + info));
+ lcl_reason = info + OPA_LINKDOWN_REASON_BAD_HEAD_DIST;
+ }
+
+ /* just report this */
+ dd_dev_info(dd, "DCC Error: fmconfig error: %s\n", extra);
+ reg &= ~DCC_ERR_FLG_FMCONFIG_ERR_SMASK;
+ }
+
+ if (reg & DCC_ERR_FLG_RCVPORT_ERR_SMASK) {
+ u8 reason_valid = 1;
+
+ info = read_csr(dd, DCC_ERR_INFO_PORTRCV);
+ hdr0 = read_csr(dd, DCC_ERR_INFO_PORTRCV_HDR0);
+ hdr1 = read_csr(dd, DCC_ERR_INFO_PORTRCV_HDR1);
+ if (!(dd->err_info_rcvport.status_and_code &
+ OPA_EI_STATUS_SMASK)) {
+ dd->err_info_rcvport.status_and_code =
+ info & OPA_EI_CODE_SMASK;
+ /* set status bit */
+ dd->err_info_rcvport.status_and_code |=
+ OPA_EI_STATUS_SMASK;
+ /* save first 2 flits in the packet that caused
+ * the error */
+ dd->err_info_rcvport.packet_flit1 = hdr0;
+ dd->err_info_rcvport.packet_flit2 = hdr1;
+ }
+ switch (info) {
+ case 1:
+ case 2:
+ case 3:
+ case 4:
+ case 5:
+ case 6:
+ case 7:
+ case 9:
+ case 11:
+ case 12:
+ extra = port_rcv_txt[info];
+ break;
+ default:
+ reason_valid = 0;
+ snprintf(buf, sizeof(buf), "reserved%lld", info);
+ extra = buf;
+ break;
+ }
+
+ if (reason_valid && !do_bounce) {
+ do_bounce = ppd->port_error_action &
+ (1 << (OPA_LDR_PORTRCV_OFFSET + info));
+ lcl_reason = info + OPA_LINKDOWN_REASON_RCV_ERROR_0;
+ }
+
+ /* just report this */
+ dd_dev_info(dd, "DCC Error: PortRcv error: %s\n", extra);
+ dd_dev_info(dd, " hdr0 0x%llx, hdr1 0x%llx\n",
+ hdr0, hdr1);
+
+ reg &= ~DCC_ERR_FLG_RCVPORT_ERR_SMASK;
+ }
+
+ if (reg & DCC_ERR_FLG_EN_CSR_ACCESS_BLOCKED_UC_SMASK) {
+ /* informative only */
+ dd_dev_info(dd, "8051 access to LCB blocked\n");
+ reg &= ~DCC_ERR_FLG_EN_CSR_ACCESS_BLOCKED_UC_SMASK;
+ }
+ if (reg & DCC_ERR_FLG_EN_CSR_ACCESS_BLOCKED_HOST_SMASK) {
+ /* informative only */
+ dd_dev_info(dd, "host access to LCB blocked\n");
+ reg &= ~DCC_ERR_FLG_EN_CSR_ACCESS_BLOCKED_HOST_SMASK;
+ }
+
+ /* report any remaining errors */
+ if (reg)
+ dd_dev_info(dd, "DCC Error: %s\n",
+ dcc_err_string(buf, sizeof(buf), reg));
+
+ if (lcl_reason == 0)
+ lcl_reason = OPA_LINKDOWN_REASON_UNKNOWN;
+
+ if (do_bounce) {
+ dd_dev_info(dd, "%s: PortErrorAction bounce\n", __func__);
+ set_link_down_reason(ppd, lcl_reason, 0, lcl_reason);
+ queue_work(ppd->hfi1_wq, &ppd->link_bounce_work);
+ }
+}
+
+static void handle_lcb_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
+{
+ char buf[96];
+
+ dd_dev_info(dd, "LCB Error: %s\n",
+ lcb_err_string(buf, sizeof(buf), reg));
+}
+
+/*
+ * CCE block DC interrupt. Source is < 8.
+ */
+static void is_dc_int(struct hfi1_devdata *dd, unsigned int source)
+{
+ const struct err_reg_info *eri = &dc_errs[source];
+
+ if (eri->handler) {
+ interrupt_clear_down(dd, 0, eri);
+ } else if (source == 3 /* dc_lbm_int */) {
+ /*
+ * This indicates that a parity error has occurred on the
+ * address/control lines presented to the LBM. The error
+ * is a single pulse, there is no associated error flag,
+ * and it is non-maskable. This is because if a parity
+ * error occurs on the request the request is dropped.
+ * This should never occur, but it is nice to know if it
+ * ever does.
+ */
+ dd_dev_err(dd, "Parity error in DC LBM block\n");
+ } else {
+ dd_dev_err(dd, "Invalid DC interrupt %u\n", source);
+ }
+}
+
+/*
+ * TX block send credit interrupt. Source is < 160.
+ */
+static void is_send_credit_int(struct hfi1_devdata *dd, unsigned int source)
+{
+ sc_group_release_update(dd, source);
+}
+
+/*
+ * TX block SDMA interrupt. Source is < 48.
+ *
+ * SDMA interrupts are grouped by type:
+ *
+ * 0 - N-1 = SDma
+ * N - 2N-1 = SDmaProgress
+ * 2N - 3N-1 = SDmaIdle
+ */
+static void is_sdma_eng_int(struct hfi1_devdata *dd, unsigned int source)
+{
+ /* what interrupt */
+ unsigned int what = source / TXE_NUM_SDMA_ENGINES;
+ /* which engine */
+ unsigned int which = source % TXE_NUM_SDMA_ENGINES;
+
+#ifdef CONFIG_SDMA_VERBOSITY
+ dd_dev_err(dd, "CONFIG SDMA(%u) %s:%d %s()\n", which,
+ slashstrip(__FILE__), __LINE__, __func__);
+ sdma_dumpstate(&dd->per_sdma[which]);
+#endif
+
+ if (likely(what < 3 && which < dd->num_sdma)) {
+ sdma_engine_interrupt(&dd->per_sdma[which], 1ull << source);
+ } else {
+ /* should not happen */
+ dd_dev_err(dd, "Invalid SDMA interrupt 0x%x\n", source);
+ }
+}
+
+/*
+ * RX block receive available interrupt. Source is < 160.
+ */
+static void is_rcv_avail_int(struct hfi1_devdata *dd, unsigned int source)
+{
+ struct hfi1_ctxtdata *rcd;
+ char *err_detail;
+
+ if (likely(source < dd->num_rcv_contexts)) {
+ rcd = dd->rcd[source];
+ if (rcd) {
+ if (source < dd->first_user_ctxt)
+ rcd->do_interrupt(rcd);
+ else
+ handle_user_interrupt(rcd);
+ return; /* OK */
+ }
+ /* received an interrupt, but no rcd */
+ err_detail = "dataless";
+ } else {
+ /* received an interrupt, but are not using that context */
+ err_detail = "out of range";
+ }
+ dd_dev_err(dd, "unexpected %s receive available context interrupt %u\n",
+ err_detail, source);
+}
+
+/*
+ * RX block receive urgent interrupt. Source is < 160.
+ */
+static void is_rcv_urgent_int(struct hfi1_devdata *dd, unsigned int source)
+{
+ struct hfi1_ctxtdata *rcd;
+ char *err_detail;
+
+ if (likely(source < dd->num_rcv_contexts)) {
+ rcd = dd->rcd[source];
+ if (rcd) {
+ /* only pay attention to user urgent interrupts */
+ if (source >= dd->first_user_ctxt)
+ handle_user_interrupt(rcd);
+ return; /* OK */
+ }
+ /* received an interrupt, but no rcd */
+ err_detail = "dataless";
+ } else {
+ /* received an interrupt, but are not using that context */
+ err_detail = "out of range";
+ }
+ dd_dev_err(dd, "unexpected %s receive urgent context interrupt %u\n",
+ err_detail, source);
+}
+
+/*
+ * Reserved range interrupt. Should not be called in normal operation.
+ */
+static void is_reserved_int(struct hfi1_devdata *dd, unsigned int source)
+{
+ char name[64];
+
+ dd_dev_err(dd, "unexpected %s interrupt\n",
+ is_reserved_name(name, sizeof(name), source));
+}
+
+static const struct is_table is_table[] = {
+/* start end
+ name func interrupt func */
+{ IS_GENERAL_ERR_START, IS_GENERAL_ERR_END,
+ is_misc_err_name, is_misc_err_int },
+{ IS_SDMAENG_ERR_START, IS_SDMAENG_ERR_END,
+ is_sdma_eng_err_name, is_sdma_eng_err_int },
+{ IS_SENDCTXT_ERR_START, IS_SENDCTXT_ERR_END,
+ is_sendctxt_err_name, is_sendctxt_err_int },
+{ IS_SDMA_START, IS_SDMA_END,
+ is_sdma_eng_name, is_sdma_eng_int },
+{ IS_VARIOUS_START, IS_VARIOUS_END,
+ is_various_name, is_various_int },
+{ IS_DC_START, IS_DC_END,
+ is_dc_name, is_dc_int },
+{ IS_RCVAVAIL_START, IS_RCVAVAIL_END,
+ is_rcv_avail_name, is_rcv_avail_int },
+{ IS_RCVURGENT_START, IS_RCVURGENT_END,
+ is_rcv_urgent_name, is_rcv_urgent_int },
+{ IS_SENDCREDIT_START, IS_SENDCREDIT_END,
+ is_send_credit_name, is_send_credit_int},
+{ IS_RESERVED_START, IS_RESERVED_END,
+ is_reserved_name, is_reserved_int},
+};
+
+/*
+ * Interrupt source interrupt - called when the given source has an interrupt.
+ * Source is a bit index into an array of 64-bit integers.
+ */
+static void is_interrupt(struct hfi1_devdata *dd, unsigned int source)
+{
+ const struct is_table *entry;
+
+ /* avoids a double compare by walking the table in-order */
+ for (entry = &is_table[0]; entry->is_name; entry++) {
+ if (source < entry->end) {
+ trace_hfi1_interrupt(dd, entry, source);
+ entry->is_int(dd, source - entry->start);
+ return;
+ }
+ }
+ /* fell off the end */
+ dd_dev_err(dd, "invalid interrupt source %u\n", source);
+}
+
+/*
+ * General interrupt handler. This is able to correctly handle
+ * all interrupts in case INTx is used.
+ */
+static irqreturn_t general_interrupt(int irq, void *data)
+{
+ struct hfi1_devdata *dd = data;
+ u64 regs[CCE_NUM_INT_CSRS];
+ u32 bit;
+ int i;
+
+ this_cpu_inc(*dd->int_counter);
+
+ /* phase 1: scan and clear all handled interrupts */
+ for (i = 0; i < CCE_NUM_INT_CSRS; i++) {
+ if (dd->gi_mask[i] == 0) {
+ regs[i] = 0; /* used later */
+ continue;
+ }
+ regs[i] = read_csr(dd, CCE_INT_STATUS + (8 * i)) &
+ dd->gi_mask[i];
+ /* only clear if anything is set */
+ if (regs[i])
+ write_csr(dd, CCE_INT_CLEAR + (8 * i), regs[i]);
+ }
+
+ /* phase 2: call the appropriate handler */
+ for_each_set_bit(bit, (unsigned long *)&regs[0],
+ CCE_NUM_INT_CSRS*64) {
+ is_interrupt(dd, bit);
+ }
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t sdma_interrupt(int irq, void *data)
+{
+ struct sdma_engine *sde = data;
+ struct hfi1_devdata *dd = sde->dd;
+ u64 status;
+
+#ifdef CONFIG_SDMA_VERBOSITY
+ dd_dev_err(dd, "CONFIG SDMA(%u) %s:%d %s()\n", sde->this_idx,
+ slashstrip(__FILE__), __LINE__, __func__);
+ sdma_dumpstate(sde);
+#endif
+
+ this_cpu_inc(*dd->int_counter);
+
+ /* This read_csr is really bad in the hot path */
+ status = read_csr(dd,
+ CCE_INT_STATUS + (8*(IS_SDMA_START/64)))
+ & sde->imask;
+ if (likely(status)) {
+ /* clear the interrupt(s) */
+ write_csr(dd,
+ CCE_INT_CLEAR + (8*(IS_SDMA_START/64)),
+ status);
+
+ /* handle the interrupt(s) */
+ sdma_engine_interrupt(sde, status);
+ } else
+ dd_dev_err(dd, "SDMA engine %u interrupt, but no status bits set\n",
+ sde->this_idx);
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * NOTE: this routine expects to be on its own MSI-X interrupt. If
+ * multiple receive contexts share the same MSI-X interrupt, then this
+ * routine must check for who received it.
+ */
+static irqreturn_t receive_context_interrupt(int irq, void *data)
+{
+ struct hfi1_ctxtdata *rcd = data;
+ struct hfi1_devdata *dd = rcd->dd;
+
+ trace_hfi1_receive_interrupt(dd, rcd->ctxt);
+ this_cpu_inc(*dd->int_counter);
+
+ /* clear the interrupt */
+ write_csr(rcd->dd, CCE_INT_CLEAR + (8*rcd->ireg), rcd->imask);
+
+ /* handle the interrupt */
+ rcd->do_interrupt(rcd);
+
+ return IRQ_HANDLED;
+}
+
+/* ========================================================================= */
+
+u32 read_physical_state(struct hfi1_devdata *dd)
+{
+ u64 reg;
+
+ reg = read_csr(dd, DC_DC8051_STS_CUR_STATE);
+ return (reg >> DC_DC8051_STS_CUR_STATE_PORT_SHIFT)
+ & DC_DC8051_STS_CUR_STATE_PORT_MASK;
+}
+
+static u32 read_logical_state(struct hfi1_devdata *dd)
+{
+ u64 reg;
+
+ reg = read_csr(dd, DCC_CFG_PORT_CONFIG);
+ return (reg >> DCC_CFG_PORT_CONFIG_LINK_STATE_SHIFT)
+ & DCC_CFG_PORT_CONFIG_LINK_STATE_MASK;
+}
+
+static void set_logical_state(struct hfi1_devdata *dd, u32 chip_lstate)
+{
+ u64 reg;
+
+ reg = read_csr(dd, DCC_CFG_PORT_CONFIG);
+ /* clear current state, set new state */
+ reg &= ~DCC_CFG_PORT_CONFIG_LINK_STATE_SMASK;
+ reg |= (u64)chip_lstate << DCC_CFG_PORT_CONFIG_LINK_STATE_SHIFT;
+ write_csr(dd, DCC_CFG_PORT_CONFIG, reg);
+}
+
+/*
+ * Use the 8051 to read a LCB CSR.
+ */
+static int read_lcb_via_8051(struct hfi1_devdata *dd, u32 addr, u64 *data)
+{
+ u32 regno;
+ int ret;
+
+ if (dd->icode == ICODE_FUNCTIONAL_SIMULATOR) {
+ if (acquire_lcb_access(dd, 0) == 0) {
+ *data = read_csr(dd, addr);
+ release_lcb_access(dd, 0);
+ return 0;
+ }
+ return -EBUSY;
+ }
+
+ /* register is an index of LCB registers: (offset - base) / 8 */
+ regno = (addr - DC_LCB_CFG_RUN) >> 3;
+ ret = do_8051_command(dd, HCMD_READ_LCB_CSR, regno, data);
+ if (ret != HCMD_SUCCESS)
+ return -EBUSY;
+ return 0;
+}
+
+/*
+ * Read an LCB CSR. Access may not be in host control, so check.
+ * Return 0 on success, -EBUSY on failure.
+ */
+int read_lcb_csr(struct hfi1_devdata *dd, u32 addr, u64 *data)
+{
+ struct hfi1_pportdata *ppd = dd->pport;
+
+ /* if up, go through the 8051 for the value */
+ if (ppd->host_link_state & HLS_UP)
+ return read_lcb_via_8051(dd, addr, data);
+ /* if going up or down, no access */
+ if (ppd->host_link_state & (HLS_GOING_UP | HLS_GOING_OFFLINE))
+ return -EBUSY;
+ /* otherwise, host has access */
+ *data = read_csr(dd, addr);
+ return 0;
+}
+
+/*
+ * Use the 8051 to write a LCB CSR.
+ */
+static int write_lcb_via_8051(struct hfi1_devdata *dd, u32 addr, u64 data)
+{
+
+ if (acquire_lcb_access(dd, 0) == 0) {
+ write_csr(dd, addr, data);
+ release_lcb_access(dd, 0);
+ return 0;
+ }
+ return -EBUSY;
+}
+
+/*
+ * Write an LCB CSR. Access may not be in host control, so check.
+ * Return 0 on success, -EBUSY on failure.
+ */
+int write_lcb_csr(struct hfi1_devdata *dd, u32 addr, u64 data)
+{
+ struct hfi1_pportdata *ppd = dd->pport;
+
+ /* if up, go through the 8051 for the value */
+ if (ppd->host_link_state & HLS_UP)
+ return write_lcb_via_8051(dd, addr, data);
+ /* if going up or down, no access */
+ if (ppd->host_link_state & (HLS_GOING_UP | HLS_GOING_OFFLINE))
+ return -EBUSY;
+ /* otherwise, host has access */
+ write_csr(dd, addr, data);
+ return 0;
+}
+
+/*
+ * Returns:
+ * < 0 = Linux error, not able to get access
+ * > 0 = 8051 command RETURN_CODE
+ */
+static int do_8051_command(
+ struct hfi1_devdata *dd,
+ u32 type,
+ u64 in_data,
+ u64 *out_data)
+{
+ u64 reg, completed;
+ int return_code;
+ unsigned long flags;
+ unsigned long timeout;
+
+ hfi1_cdbg(DC8051, "type %d, data 0x%012llx", type, in_data);
+
+ /*
+ * Alternative to holding the lock for a long time:
+ * - keep busy wait - have other users bounce off
+ */
+ spin_lock_irqsave(&dd->dc8051_lock, flags);
+
+ /* We can't send any commands to the 8051 if it's in reset */
+ if (dd->dc_shutdown) {
+ return_code = -ENODEV;
+ goto fail;
+ }
+
+ /*
+ * If an 8051 host command timed out previously, then the 8051 is
+ * stuck.
+ *
+ * On first timeout, attempt to reset and restart the entire DC
+ * block (including 8051). (Is this too big of a hammer?)
+ *
+ * If the 8051 times out a second time, the reset did not bring it
+ * back to healthy life. In that case, fail any subsequent commands.
+ */
+ if (dd->dc8051_timed_out) {
+ if (dd->dc8051_timed_out > 1) {
+ dd_dev_err(dd,
+ "Previous 8051 host command timed out, skipping command %u\n",
+ type);
+ return_code = -ENXIO;
+ goto fail;
+ }
+ spin_unlock_irqrestore(&dd->dc8051_lock, flags);
+ dc_shutdown(dd);
+ dc_start(dd);
+ spin_lock_irqsave(&dd->dc8051_lock, flags);
+ }
+
+ /*
+ * If there is no timeout, then the 8051 command interface is
+ * waiting for a command.
+ */
+
+ /*
+ * Do two writes: the first to stabilize the type and req_data, the
+ * second to activate.
+ */
+ reg = ((u64)type & DC_DC8051_CFG_HOST_CMD_0_REQ_TYPE_MASK)
+ << DC_DC8051_CFG_HOST_CMD_0_REQ_TYPE_SHIFT
+ | (in_data & DC_DC8051_CFG_HOST_CMD_0_REQ_DATA_MASK)
+ << DC_DC8051_CFG_HOST_CMD_0_REQ_DATA_SHIFT;
+ write_csr(dd, DC_DC8051_CFG_HOST_CMD_0, reg);
+ reg |= DC_DC8051_CFG_HOST_CMD_0_REQ_NEW_SMASK;
+ write_csr(dd, DC_DC8051_CFG_HOST_CMD_0, reg);
+
+ /* wait for completion, alternate: interrupt */
+ timeout = jiffies + msecs_to_jiffies(DC8051_COMMAND_TIMEOUT);
+ while (1) {
+ reg = read_csr(dd, DC_DC8051_CFG_HOST_CMD_1);
+ completed = reg & DC_DC8051_CFG_HOST_CMD_1_COMPLETED_SMASK;
+ if (completed)
+ break;
+ if (time_after(jiffies, timeout)) {
+ dd->dc8051_timed_out++;
+ dd_dev_err(dd, "8051 host command %u timeout\n", type);
+ if (out_data)
+ *out_data = 0;
+ return_code = -ETIMEDOUT;
+ goto fail;
+ }
+ udelay(2);
+ }
+
+ if (out_data) {
+ *out_data = (reg >> DC_DC8051_CFG_HOST_CMD_1_RSP_DATA_SHIFT)
+ & DC_DC8051_CFG_HOST_CMD_1_RSP_DATA_MASK;
+ if (type == HCMD_READ_LCB_CSR) {
+ /* top 16 bits are in a different register */
+ *out_data |= (read_csr(dd, DC_DC8051_CFG_EXT_DEV_1)
+ & DC_DC8051_CFG_EXT_DEV_1_REQ_DATA_SMASK)
+ << (48
+ - DC_DC8051_CFG_EXT_DEV_1_REQ_DATA_SHIFT);
+ }
+ }
+ return_code = (reg >> DC_DC8051_CFG_HOST_CMD_1_RETURN_CODE_SHIFT)
+ & DC_DC8051_CFG_HOST_CMD_1_RETURN_CODE_MASK;
+ dd->dc8051_timed_out = 0;
+ /*
+ * Clear command for next user.
+ */
+ write_csr(dd, DC_DC8051_CFG_HOST_CMD_0, 0);
+
+fail:
+ spin_unlock_irqrestore(&dd->dc8051_lock, flags);
+
+ return return_code;
+}
+
+static int set_physical_link_state(struct hfi1_devdata *dd, u64 state)
+{
+ return do_8051_command(dd, HCMD_CHANGE_PHY_STATE, state, NULL);
+}
+
+static int load_8051_config(struct hfi1_devdata *dd, u8 field_id,
+ u8 lane_id, u32 config_data)
+{
+ u64 data;
+ int ret;
+
+ data = (u64)field_id << LOAD_DATA_FIELD_ID_SHIFT
+ | (u64)lane_id << LOAD_DATA_LANE_ID_SHIFT
+ | (u64)config_data << LOAD_DATA_DATA_SHIFT;
+ ret = do_8051_command(dd, HCMD_LOAD_CONFIG_DATA, data, NULL);
+ if (ret != HCMD_SUCCESS) {
+ dd_dev_err(dd,
+ "load 8051 config: field id %d, lane %d, err %d\n",
+ (int)field_id, (int)lane_id, ret);
+ }
+ return ret;
+}
+
+/*
+ * Read the 8051 firmware "registers". Use the RAM directly. Always
+ * set the result, even on error.
+ * Return 0 on success, -errno on failure
+ */
+static int read_8051_config(struct hfi1_devdata *dd, u8 field_id, u8 lane_id,
+ u32 *result)
+{
+ u64 big_data;
+ u32 addr;
+ int ret;
+
+ /* address start depends on the lane_id */
+ if (lane_id < 4)
+ addr = (4 * NUM_GENERAL_FIELDS)
+ + (lane_id * 4 * NUM_LANE_FIELDS);
+ else
+ addr = 0;
+ addr += field_id * 4;
+
+ /* read is in 8-byte chunks, hardware will truncate the address down */
+ ret = read_8051_data(dd, addr, 8, &big_data);
+
+ if (ret == 0) {
+ /* extract the 4 bytes we want */
+ if (addr & 0x4)
+ *result = (u32)(big_data >> 32);
+ else
+ *result = (u32)big_data;
+ } else {
+ *result = 0;
+ dd_dev_err(dd, "%s: direct read failed, lane %d, field %d!\n",
+ __func__, lane_id, field_id);
+ }
+
+ return ret;
+}
+
+static int write_vc_local_phy(struct hfi1_devdata *dd, u8 power_management,
+ u8 continuous)
+{
+ u32 frame;
+
+ frame = continuous << CONTINIOUS_REMOTE_UPDATE_SUPPORT_SHIFT
+ | power_management << POWER_MANAGEMENT_SHIFT;
+ return load_8051_config(dd, VERIFY_CAP_LOCAL_PHY,
+ GENERAL_CONFIG, frame);
+}
+
+static int write_vc_local_fabric(struct hfi1_devdata *dd, u8 vau, u8 z, u8 vcu,
+ u16 vl15buf, u8 crc_sizes)
+{
+ u32 frame;
+
+ frame = (u32)vau << VAU_SHIFT
+ | (u32)z << Z_SHIFT
+ | (u32)vcu << VCU_SHIFT
+ | (u32)vl15buf << VL15BUF_SHIFT
+ | (u32)crc_sizes << CRC_SIZES_SHIFT;
+ return load_8051_config(dd, VERIFY_CAP_LOCAL_FABRIC,
+ GENERAL_CONFIG, frame);
+}
+
+static void read_vc_local_link_width(struct hfi1_devdata *dd, u8 *misc_bits,
+ u8 *flag_bits, u16 *link_widths)
+{
+ u32 frame;
+
+ read_8051_config(dd, VERIFY_CAP_LOCAL_LINK_WIDTH, GENERAL_CONFIG,
+ &frame);
+ *misc_bits = (frame >> MISC_CONFIG_BITS_SHIFT) & MISC_CONFIG_BITS_MASK;
+ *flag_bits = (frame >> LOCAL_FLAG_BITS_SHIFT) & LOCAL_FLAG_BITS_MASK;
+ *link_widths = (frame >> LINK_WIDTH_SHIFT) & LINK_WIDTH_MASK;
+}
+
+static int write_vc_local_link_width(struct hfi1_devdata *dd,
+ u8 misc_bits,
+ u8 flag_bits,
+ u16 link_widths)
+{
+ u32 frame;
+
+ frame = (u32)misc_bits << MISC_CONFIG_BITS_SHIFT
+ | (u32)flag_bits << LOCAL_FLAG_BITS_SHIFT
+ | (u32)link_widths << LINK_WIDTH_SHIFT;
+ return load_8051_config(dd, VERIFY_CAP_LOCAL_LINK_WIDTH, GENERAL_CONFIG,
+ frame);
+}
+
+static int write_local_device_id(struct hfi1_devdata *dd, u16 device_id,
+ u8 device_rev)
+{
+ u32 frame;
+
+ frame = ((u32)device_id << LOCAL_DEVICE_ID_SHIFT)
+ | ((u32)device_rev << LOCAL_DEVICE_REV_SHIFT);
+ return load_8051_config(dd, LOCAL_DEVICE_ID, GENERAL_CONFIG, frame);
+}
+
+static void read_remote_device_id(struct hfi1_devdata *dd, u16 *device_id,
+ u8 *device_rev)
+{
+ u32 frame;
+
+ read_8051_config(dd, REMOTE_DEVICE_ID, GENERAL_CONFIG, &frame);
+ *device_id = (frame >> REMOTE_DEVICE_ID_SHIFT) & REMOTE_DEVICE_ID_MASK;
+ *device_rev = (frame >> REMOTE_DEVICE_REV_SHIFT)
+ & REMOTE_DEVICE_REV_MASK;
+}
+
+void read_misc_status(struct hfi1_devdata *dd, u8 *ver_a, u8 *ver_b)
+{
+ u32 frame;
+
+ read_8051_config(dd, MISC_STATUS, GENERAL_CONFIG, &frame);
+ *ver_a = (frame >> STS_FM_VERSION_A_SHIFT) & STS_FM_VERSION_A_MASK;
+ *ver_b = (frame >> STS_FM_VERSION_B_SHIFT) & STS_FM_VERSION_B_MASK;
+}
+
+static void read_vc_remote_phy(struct hfi1_devdata *dd, u8 *power_management,
+ u8 *continuous)
+{
+ u32 frame;
+
+ read_8051_config(dd, VERIFY_CAP_REMOTE_PHY, GENERAL_CONFIG, &frame);
+ *power_management = (frame >> POWER_MANAGEMENT_SHIFT)
+ & POWER_MANAGEMENT_MASK;
+ *continuous = (frame >> CONTINIOUS_REMOTE_UPDATE_SUPPORT_SHIFT)
+ & CONTINIOUS_REMOTE_UPDATE_SUPPORT_MASK;
+}
+
+static void read_vc_remote_fabric(struct hfi1_devdata *dd, u8 *vau, u8 *z,
+ u8 *vcu, u16 *vl15buf, u8 *crc_sizes)
+{
+ u32 frame;
+
+ read_8051_config(dd, VERIFY_CAP_REMOTE_FABRIC, GENERAL_CONFIG, &frame);
+ *vau = (frame >> VAU_SHIFT) & VAU_MASK;
+ *z = (frame >> Z_SHIFT) & Z_MASK;
+ *vcu = (frame >> VCU_SHIFT) & VCU_MASK;
+ *vl15buf = (frame >> VL15BUF_SHIFT) & VL15BUF_MASK;
+ *crc_sizes = (frame >> CRC_SIZES_SHIFT) & CRC_SIZES_MASK;
+}
+
+static void read_vc_remote_link_width(struct hfi1_devdata *dd,
+ u8 *remote_tx_rate,
+ u16 *link_widths)
+{
+ u32 frame;
+
+ read_8051_config(dd, VERIFY_CAP_REMOTE_LINK_WIDTH, GENERAL_CONFIG,
+ &frame);
+ *remote_tx_rate = (frame >> REMOTE_TX_RATE_SHIFT)
+ & REMOTE_TX_RATE_MASK;
+ *link_widths = (frame >> LINK_WIDTH_SHIFT) & LINK_WIDTH_MASK;
+}
+
+static void read_local_lni(struct hfi1_devdata *dd, u8 *enable_lane_rx)
+{
+ u32 frame;
+
+ read_8051_config(dd, LOCAL_LNI_INFO, GENERAL_CONFIG, &frame);
+ *enable_lane_rx = (frame >> ENABLE_LANE_RX_SHIFT) & ENABLE_LANE_RX_MASK;
+}
+
+static void read_mgmt_allowed(struct hfi1_devdata *dd, u8 *mgmt_allowed)
+{
+ u32 frame;
+
+ read_8051_config(dd, REMOTE_LNI_INFO, GENERAL_CONFIG, &frame);
+ *mgmt_allowed = (frame >> MGMT_ALLOWED_SHIFT) & MGMT_ALLOWED_MASK;
+}
+
+static void read_last_local_state(struct hfi1_devdata *dd, u32 *lls)
+{
+ read_8051_config(dd, LAST_LOCAL_STATE_COMPLETE, GENERAL_CONFIG, lls);
+}
+
+static void read_last_remote_state(struct hfi1_devdata *dd, u32 *lrs)
+{
+ read_8051_config(dd, LAST_REMOTE_STATE_COMPLETE, GENERAL_CONFIG, lrs);
+}
+
+void hfi1_read_link_quality(struct hfi1_devdata *dd, u8 *link_quality)
+{
+ u32 frame;
+ int ret;
+
+ *link_quality = 0;
+ if (dd->pport->host_link_state & HLS_UP) {
+ ret = read_8051_config(dd, LINK_QUALITY_INFO, GENERAL_CONFIG,
+ &frame);
+ if (ret == 0)
+ *link_quality = (frame >> LINK_QUALITY_SHIFT)
+ & LINK_QUALITY_MASK;
+ }
+}
+
+static void read_planned_down_reason_code(struct hfi1_devdata *dd, u8 *pdrrc)
+{
+ u32 frame;
+
+ read_8051_config(dd, LINK_QUALITY_INFO, GENERAL_CONFIG, &frame);
+ *pdrrc = (frame >> DOWN_REMOTE_REASON_SHIFT) & DOWN_REMOTE_REASON_MASK;
+}
+
+static int read_tx_settings(struct hfi1_devdata *dd,
+ u8 *enable_lane_tx,
+ u8 *tx_polarity_inversion,
+ u8 *rx_polarity_inversion,
+ u8 *max_rate)
+{
+ u32 frame;
+ int ret;
+
+ ret = read_8051_config(dd, TX_SETTINGS, GENERAL_CONFIG, &frame);
+ *enable_lane_tx = (frame >> ENABLE_LANE_TX_SHIFT)
+ & ENABLE_LANE_TX_MASK;
+ *tx_polarity_inversion = (frame >> TX_POLARITY_INVERSION_SHIFT)
+ & TX_POLARITY_INVERSION_MASK;
+ *rx_polarity_inversion = (frame >> RX_POLARITY_INVERSION_SHIFT)
+ & RX_POLARITY_INVERSION_MASK;
+ *max_rate = (frame >> MAX_RATE_SHIFT) & MAX_RATE_MASK;
+ return ret;
+}
+
+static int write_tx_settings(struct hfi1_devdata *dd,
+ u8 enable_lane_tx,
+ u8 tx_polarity_inversion,
+ u8 rx_polarity_inversion,
+ u8 max_rate)
+{
+ u32 frame;
+
+ /* no need to mask, all variable sizes match field widths */
+ frame = enable_lane_tx << ENABLE_LANE_TX_SHIFT
+ | tx_polarity_inversion << TX_POLARITY_INVERSION_SHIFT
+ | rx_polarity_inversion << RX_POLARITY_INVERSION_SHIFT
+ | max_rate << MAX_RATE_SHIFT;
+ return load_8051_config(dd, TX_SETTINGS, GENERAL_CONFIG, frame);
+}
+
+static void check_fabric_firmware_versions(struct hfi1_devdata *dd)
+{
+ u32 frame, version, prod_id;
+ int ret, lane;
+
+ /* 4 lanes */
+ for (lane = 0; lane < 4; lane++) {
+ ret = read_8051_config(dd, SPICO_FW_VERSION, lane, &frame);
+ if (ret) {
+ dd_dev_err(
+ dd,
+ "Unable to read lane %d firmware details\n",
+ lane);
+ continue;
+ }
+ version = (frame >> SPICO_ROM_VERSION_SHIFT)
+ & SPICO_ROM_VERSION_MASK;
+ prod_id = (frame >> SPICO_ROM_PROD_ID_SHIFT)
+ & SPICO_ROM_PROD_ID_MASK;
+ dd_dev_info(dd,
+ "Lane %d firmware: version 0x%04x, prod_id 0x%04x\n",
+ lane, version, prod_id);
+ }
+}
+
+/*
+ * Read an idle LCB message.
+ *
+ * Returns 0 on success, -EINVAL on error
+ */
+static int read_idle_message(struct hfi1_devdata *dd, u64 type, u64 *data_out)
+{
+ int ret;
+
+ ret = do_8051_command(dd, HCMD_READ_LCB_IDLE_MSG,
+ type, data_out);
+ if (ret != HCMD_SUCCESS) {
+ dd_dev_err(dd, "read idle message: type %d, err %d\n",
+ (u32)type, ret);
+ return -EINVAL;
+ }
+ dd_dev_info(dd, "%s: read idle message 0x%llx\n", __func__, *data_out);
+ /* return only the payload as we already know the type */
+ *data_out >>= IDLE_PAYLOAD_SHIFT;
+ return 0;
+}
+
+/*
+ * Read an idle SMA message. To be done in response to a notification from
+ * the 8051.
+ *
+ * Returns 0 on success, -EINVAL on error
+ */
+static int read_idle_sma(struct hfi1_devdata *dd, u64 *data)
+{
+ return read_idle_message(dd,
+ (u64)IDLE_SMA << IDLE_MSG_TYPE_SHIFT, data);
+}
+
+/*
+ * Send an idle LCB message.
+ *
+ * Returns 0 on success, -EINVAL on error
+ */
+static int send_idle_message(struct hfi1_devdata *dd, u64 data)
+{
+ int ret;
+
+ dd_dev_info(dd, "%s: sending idle message 0x%llx\n", __func__, data);
+ ret = do_8051_command(dd, HCMD_SEND_LCB_IDLE_MSG, data, NULL);
+ if (ret != HCMD_SUCCESS) {
+ dd_dev_err(dd, "send idle message: data 0x%llx, err %d\n",
+ data, ret);
+ return -EINVAL;
+ }
+ return 0;
+}
+
+/*
+ * Send an idle SMA message.
+ *
+ * Returns 0 on success, -EINVAL on error
+ */
+int send_idle_sma(struct hfi1_devdata *dd, u64 message)
+{
+ u64 data;
+
+ data = ((message & IDLE_PAYLOAD_MASK) << IDLE_PAYLOAD_SHIFT)
+ | ((u64)IDLE_SMA << IDLE_MSG_TYPE_SHIFT);
+ return send_idle_message(dd, data);
+}
+
+/*
+ * Initialize the LCB then do a quick link up. This may or may not be
+ * in loopback.
+ *
+ * return 0 on success, -errno on error
+ */
+static int do_quick_linkup(struct hfi1_devdata *dd)
+{
+ u64 reg;
+ unsigned long timeout;
+ int ret;
+
+ lcb_shutdown(dd, 0);
+
+ if (loopback) {
+ /* LCB_CFG_LOOPBACK.VAL = 2 */
+ /* LCB_CFG_LANE_WIDTH.VAL = 0 */
+ write_csr(dd, DC_LCB_CFG_LOOPBACK,
+ IB_PACKET_TYPE << DC_LCB_CFG_LOOPBACK_VAL_SHIFT);
+ write_csr(dd, DC_LCB_CFG_LANE_WIDTH, 0);
+ }
+
+ /* start the LCBs */
+ /* LCB_CFG_TX_FIFOS_RESET.VAL = 0 */
+ write_csr(dd, DC_LCB_CFG_TX_FIFOS_RESET, 0);
+
+ /* simulator only loopback steps */
+ if (loopback && dd->icode == ICODE_FUNCTIONAL_SIMULATOR) {
+ /* LCB_CFG_RUN.EN = 1 */
+ write_csr(dd, DC_LCB_CFG_RUN,
+ 1ull << DC_LCB_CFG_RUN_EN_SHIFT);
+
+ /* watch LCB_STS_LINK_TRANSFER_ACTIVE */
+ timeout = jiffies + msecs_to_jiffies(10);
+ while (1) {
+ reg = read_csr(dd,
+ DC_LCB_STS_LINK_TRANSFER_ACTIVE);
+ if (reg)
+ break;
+ if (time_after(jiffies, timeout)) {
+ dd_dev_err(dd,
+ "timeout waiting for LINK_TRANSFER_ACTIVE\n");
+ return -ETIMEDOUT;
+ }
+ udelay(2);
+ }
+
+ write_csr(dd, DC_LCB_CFG_ALLOW_LINK_UP,
+ 1ull << DC_LCB_CFG_ALLOW_LINK_UP_VAL_SHIFT);
+ }
+
+ if (!loopback) {
+ /*
+ * When doing quick linkup and not in loopback, both
+ * sides must be done with LCB set-up before either
+ * starts the quick linkup. Put a delay here so that
+ * both sides can be started and have a chance to be
+ * done with LCB set up before resuming.
+ */
+ dd_dev_err(dd,
+ "Pausing for peer to be finished with LCB set up\n");
+ msleep(5000);
+ dd_dev_err(dd,
+ "Continuing with quick linkup\n");
+ }
+
+ write_csr(dd, DC_LCB_ERR_EN, 0); /* mask LCB errors */
+ set_8051_lcb_access(dd);
+
+ /*
+ * State "quick" LinkUp request sets the physical link state to
+ * LinkUp without a verify capability sequence.
+ * This state is in simulator v37 and later.
+ */
+ ret = set_physical_link_state(dd, PLS_QUICK_LINKUP);
+ if (ret != HCMD_SUCCESS) {
+ dd_dev_err(dd,
+ "%s: set physical link state to quick LinkUp failed with return %d\n",
+ __func__, ret);
+
+ set_host_lcb_access(dd);
+ write_csr(dd, DC_LCB_ERR_EN, ~0ull); /* watch LCB errors */
+
+ if (ret >= 0)
+ ret = -EINVAL;
+ return ret;
+ }
+
+ return 0; /* success */
+}
+
+/*
+ * Set the SerDes to internal loopback mode.
+ * Returns 0 on success, -errno on error.
+ */
+static int set_serdes_loopback_mode(struct hfi1_devdata *dd)
+{
+ int ret;
+
+ ret = set_physical_link_state(dd, PLS_INTERNAL_SERDES_LOOPBACK);
+ if (ret == HCMD_SUCCESS)
+ return 0;
+ dd_dev_err(dd,
+ "Set physical link state to SerDes Loopback failed with return %d\n",
+ ret);
+ if (ret >= 0)
+ ret = -EINVAL;
+ return ret;
+}
+
+/*
+ * Do all special steps to set up loopback.
+ */
+static int init_loopback(struct hfi1_devdata *dd)
+{
+ dd_dev_info(dd, "Entering loopback mode\n");
+
+ /* all loopbacks should disable self GUID check */
+ write_csr(dd, DC_DC8051_CFG_MODE,
+ (read_csr(dd, DC_DC8051_CFG_MODE) | DISABLE_SELF_GUID_CHECK));
+
+ /*
+ * The simulator has only one loopback option - LCB. Switch
+ * to that option, which includes quick link up.
+ *
+ * Accept all valid loopback values.
+ */
+ if ((dd->icode == ICODE_FUNCTIONAL_SIMULATOR)
+ && (loopback == LOOPBACK_SERDES
+ || loopback == LOOPBACK_LCB
+ || loopback == LOOPBACK_CABLE)) {
+ loopback = LOOPBACK_LCB;
+ quick_linkup = 1;
+ return 0;
+ }
+
+ /* handle serdes loopback */
+ if (loopback == LOOPBACK_SERDES) {
+ /* internal serdes loopack needs quick linkup on RTL */
+ if (dd->icode == ICODE_RTL_SILICON)
+ quick_linkup = 1;
+ return set_serdes_loopback_mode(dd);
+ }
+
+ /* LCB loopback - handled at poll time */
+ if (loopback == LOOPBACK_LCB) {
+ quick_linkup = 1; /* LCB is always quick linkup */
+
+ /* not supported in emulation due to emulation RTL changes */
+ if (dd->icode == ICODE_FPGA_EMULATION) {
+ dd_dev_err(dd,
+ "LCB loopback not supported in emulation\n");
+ return -EINVAL;
+ }
+ return 0;
+ }
+
+ /* external cable loopback requires no extra steps */
+ if (loopback == LOOPBACK_CABLE)
+ return 0;
+
+ dd_dev_err(dd, "Invalid loopback mode %d\n", loopback);
+ return -EINVAL;
+}
+
+/*
+ * Translate from the OPA_LINK_WIDTH handed to us by the FM to bits
+ * used in the Verify Capability link width attribute.
+ */
+static u16 opa_to_vc_link_widths(u16 opa_widths)
+{
+ int i;
+ u16 result = 0;
+
+ static const struct link_bits {
+ u16 from;
+ u16 to;
+ } opa_link_xlate[] = {
+ { OPA_LINK_WIDTH_1X, 1 << (1-1) },
+ { OPA_LINK_WIDTH_2X, 1 << (2-1) },
+ { OPA_LINK_WIDTH_3X, 1 << (3-1) },
+ { OPA_LINK_WIDTH_4X, 1 << (4-1) },
+ };
+
+ for (i = 0; i < ARRAY_SIZE(opa_link_xlate); i++) {
+ if (opa_widths & opa_link_xlate[i].from)
+ result |= opa_link_xlate[i].to;
+ }
+ return result;
+}
+
+/*
+ * Set link attributes before moving to polling.
+ */
+static int set_local_link_attributes(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ u8 enable_lane_tx;
+ u8 tx_polarity_inversion;
+ u8 rx_polarity_inversion;
+ int ret;
+
+ /* reset our fabric serdes to clear any lingering problems */
+ fabric_serdes_reset(dd);
+
+ /* set the local tx rate - need to read-modify-write */
+ ret = read_tx_settings(dd, &enable_lane_tx, &tx_polarity_inversion,
+ &rx_polarity_inversion, &ppd->local_tx_rate);
+ if (ret)
+ goto set_local_link_attributes_fail;
+
+ if (dd->dc8051_ver < dc8051_ver(0, 20)) {
+ /* set the tx rate to the fastest enabled */
+ if (ppd->link_speed_enabled & OPA_LINK_SPEED_25G)
+ ppd->local_tx_rate = 1;
+ else
+ ppd->local_tx_rate = 0;
+ } else {
+ /* set the tx rate to all enabled */
+ ppd->local_tx_rate = 0;
+ if (ppd->link_speed_enabled & OPA_LINK_SPEED_25G)
+ ppd->local_tx_rate |= 2;
+ if (ppd->link_speed_enabled & OPA_LINK_SPEED_12_5G)
+ ppd->local_tx_rate |= 1;
+ }
+ ret = write_tx_settings(dd, enable_lane_tx, tx_polarity_inversion,
+ rx_polarity_inversion, ppd->local_tx_rate);
+ if (ret != HCMD_SUCCESS)
+ goto set_local_link_attributes_fail;
+
+ /*
+ * DC supports continuous updates.
+ */
+ ret = write_vc_local_phy(dd, 0 /* no power management */,
+ 1 /* continuous updates */);
+ if (ret != HCMD_SUCCESS)
+ goto set_local_link_attributes_fail;
+
+ /* z=1 in the next call: AU of 0 is not supported by the hardware */
+ ret = write_vc_local_fabric(dd, dd->vau, 1, dd->vcu, dd->vl15_init,
+ ppd->port_crc_mode_enabled);
+ if (ret != HCMD_SUCCESS)
+ goto set_local_link_attributes_fail;
+
+ ret = write_vc_local_link_width(dd, 0, 0,
+ opa_to_vc_link_widths(ppd->link_width_enabled));
+ if (ret != HCMD_SUCCESS)
+ goto set_local_link_attributes_fail;
+
+ /* let peer know who we are */
+ ret = write_local_device_id(dd, dd->pcidev->device, dd->minrev);
+ if (ret == HCMD_SUCCESS)
+ return 0;
+
+set_local_link_attributes_fail:
+ dd_dev_err(dd,
+ "Failed to set local link attributes, return 0x%x\n",
+ ret);
+ return ret;
+}
+
+/*
+ * Call this to start the link. Schedule a retry if the cable is not
+ * present or if unable to start polling. Do not do anything if the
+ * link is disabled. Returns 0 if link is disabled or moved to polling
+ */
+int start_link(struct hfi1_pportdata *ppd)
+{
+ if (!ppd->link_enabled) {
+ dd_dev_info(ppd->dd,
+ "%s: stopping link start because link is disabled\n",
+ __func__);
+ return 0;
+ }
+ if (!ppd->driver_link_ready) {
+ dd_dev_info(ppd->dd,
+ "%s: stopping link start because driver is not ready\n",
+ __func__);
+ return 0;
+ }
+
+ if (qsfp_mod_present(ppd) || loopback == LOOPBACK_SERDES ||
+ loopback == LOOPBACK_LCB ||
+ ppd->dd->icode == ICODE_FUNCTIONAL_SIMULATOR)
+ return set_link_state(ppd, HLS_DN_POLL);
+
+ dd_dev_info(ppd->dd,
+ "%s: stopping link start because no cable is present\n",
+ __func__);
+ return -EAGAIN;
+}
+
+static void reset_qsfp(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ u64 mask, qsfp_mask;
+
+ mask = (u64)QSFP_HFI0_RESET_N;
+ qsfp_mask = read_csr(dd,
+ dd->hfi1_id ? ASIC_QSFP2_OE : ASIC_QSFP1_OE);
+ qsfp_mask |= mask;
+ write_csr(dd,
+ dd->hfi1_id ? ASIC_QSFP2_OE : ASIC_QSFP1_OE,
+ qsfp_mask);
+
+ qsfp_mask = read_csr(dd,
+ dd->hfi1_id ? ASIC_QSFP2_OUT : ASIC_QSFP1_OUT);
+ qsfp_mask &= ~mask;
+ write_csr(dd,
+ dd->hfi1_id ? ASIC_QSFP2_OUT : ASIC_QSFP1_OUT,
+ qsfp_mask);
+
+ udelay(10);
+
+ qsfp_mask |= mask;
+ write_csr(dd,
+ dd->hfi1_id ? ASIC_QSFP2_OUT : ASIC_QSFP1_OUT,
+ qsfp_mask);
+}
+
+static int handle_qsfp_error_conditions(struct hfi1_pportdata *ppd,
+ u8 *qsfp_interrupt_status)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+
+ if ((qsfp_interrupt_status[0] & QSFP_HIGH_TEMP_ALARM) ||
+ (qsfp_interrupt_status[0] & QSFP_HIGH_TEMP_WARNING))
+ dd_dev_info(dd,
+ "%s: QSFP cable on fire\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[0] & QSFP_LOW_TEMP_ALARM) ||
+ (qsfp_interrupt_status[0] & QSFP_LOW_TEMP_WARNING))
+ dd_dev_info(dd,
+ "%s: QSFP cable temperature too low\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[1] & QSFP_HIGH_VCC_ALARM) ||
+ (qsfp_interrupt_status[1] & QSFP_HIGH_VCC_WARNING))
+ dd_dev_info(dd,
+ "%s: QSFP supply voltage too high\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[1] & QSFP_LOW_VCC_ALARM) ||
+ (qsfp_interrupt_status[1] & QSFP_LOW_VCC_WARNING))
+ dd_dev_info(dd,
+ "%s: QSFP supply voltage too low\n",
+ __func__);
+
+ /* Byte 2 is vendor specific */
+
+ if ((qsfp_interrupt_status[3] & QSFP_HIGH_POWER_ALARM) ||
+ (qsfp_interrupt_status[3] & QSFP_HIGH_POWER_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable RX channel 1/2 power too high\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[3] & QSFP_LOW_POWER_ALARM) ||
+ (qsfp_interrupt_status[3] & QSFP_LOW_POWER_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable RX channel 1/2 power too low\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[4] & QSFP_HIGH_POWER_ALARM) ||
+ (qsfp_interrupt_status[4] & QSFP_HIGH_POWER_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable RX channel 3/4 power too high\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[4] & QSFP_LOW_POWER_ALARM) ||
+ (qsfp_interrupt_status[4] & QSFP_LOW_POWER_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable RX channel 3/4 power too low\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[5] & QSFP_HIGH_BIAS_ALARM) ||
+ (qsfp_interrupt_status[5] & QSFP_HIGH_BIAS_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable TX channel 1/2 bias too high\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[5] & QSFP_LOW_BIAS_ALARM) ||
+ (qsfp_interrupt_status[5] & QSFP_LOW_BIAS_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable TX channel 1/2 bias too low\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[6] & QSFP_HIGH_BIAS_ALARM) ||
+ (qsfp_interrupt_status[6] & QSFP_HIGH_BIAS_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable TX channel 3/4 bias too high\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[6] & QSFP_LOW_BIAS_ALARM) ||
+ (qsfp_interrupt_status[6] & QSFP_LOW_BIAS_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable TX channel 3/4 bias too low\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[7] & QSFP_HIGH_POWER_ALARM) ||
+ (qsfp_interrupt_status[7] & QSFP_HIGH_POWER_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable TX channel 1/2 power too high\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[7] & QSFP_LOW_POWER_ALARM) ||
+ (qsfp_interrupt_status[7] & QSFP_LOW_POWER_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable TX channel 1/2 power too low\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[8] & QSFP_HIGH_POWER_ALARM) ||
+ (qsfp_interrupt_status[8] & QSFP_HIGH_POWER_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable TX channel 3/4 power too high\n",
+ __func__);
+
+ if ((qsfp_interrupt_status[8] & QSFP_LOW_POWER_ALARM) ||
+ (qsfp_interrupt_status[8] & QSFP_LOW_POWER_WARNING))
+ dd_dev_info(dd,
+ "%s: Cable TX channel 3/4 power too low\n",
+ __func__);
+
+ /* Bytes 9-10 and 11-12 are reserved */
+ /* Bytes 13-15 are vendor specific */
+
+ return 0;
+}
+
+static int do_pre_lni_host_behaviors(struct hfi1_pportdata *ppd)
+{
+ refresh_qsfp_cache(ppd, &ppd->qsfp_info);
+
+ return 0;
+}
+
+static int do_qsfp_intr_fallback(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ u8 qsfp_interrupt_status = 0;
+
+ if (qsfp_read(ppd, dd->hfi1_id, 2, &qsfp_interrupt_status, 1)
+ != 1) {
+ dd_dev_info(dd,
+ "%s: Failed to read status of QSFP module\n",
+ __func__);
+ return -EIO;
+ }
+
+ /* We don't care about alarms & warnings with a non-functional INT_N */
+ if (!(qsfp_interrupt_status & QSFP_DATA_NOT_READY))
+ do_pre_lni_host_behaviors(ppd);
+
+ return 0;
+}
+
+/* This routine will only be scheduled if the QSFP module is present */
+static void qsfp_event(struct work_struct *work)
+{
+ struct qsfp_data *qd;
+ struct hfi1_pportdata *ppd;
+ struct hfi1_devdata *dd;
+
+ qd = container_of(work, struct qsfp_data, qsfp_work);
+ ppd = qd->ppd;
+ dd = ppd->dd;
+
+ /* Sanity check */
+ if (!qsfp_mod_present(ppd))
+ return;
+
+ /*
+ * Turn DC back on after cables has been
+ * re-inserted. Up until now, the DC has been in
+ * reset to save power.
+ */
+ dc_start(dd);
+
+ if (qd->cache_refresh_required) {
+ msleep(3000);
+ reset_qsfp(ppd);
+
+ /* Check for QSFP interrupt after t_init (SFF 8679)
+ * + extra
+ */
+ msleep(3000);
+ if (!qd->qsfp_interrupt_functional) {
+ if (do_qsfp_intr_fallback(ppd) < 0)
+ dd_dev_info(dd, "%s: QSFP fallback failed\n",
+ __func__);
+ ppd->driver_link_ready = 1;
+ start_link(ppd);
+ }
+ }
+
+ if (qd->check_interrupt_flags) {
+ u8 qsfp_interrupt_status[16] = {0,};
+
+ if (qsfp_read(ppd, dd->hfi1_id, 6,
+ &qsfp_interrupt_status[0], 16) != 16) {
+ dd_dev_info(dd,
+ "%s: Failed to read status of QSFP module\n",
+ __func__);
+ } else {
+ unsigned long flags;
+ u8 data_status;
+
+ spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
+ ppd->qsfp_info.check_interrupt_flags = 0;
+ spin_unlock_irqrestore(&ppd->qsfp_info.qsfp_lock,
+ flags);
+
+ if (qsfp_read(ppd, dd->hfi1_id, 2, &data_status, 1)
+ != 1) {
+ dd_dev_info(dd,
+ "%s: Failed to read status of QSFP module\n",
+ __func__);
+ }
+ if (!(data_status & QSFP_DATA_NOT_READY)) {
+ do_pre_lni_host_behaviors(ppd);
+ start_link(ppd);
+ } else
+ handle_qsfp_error_conditions(ppd,
+ qsfp_interrupt_status);
+ }
+ }
+}
+
+void init_qsfp(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ u64 qsfp_mask;
+
+ if (loopback == LOOPBACK_SERDES || loopback == LOOPBACK_LCB ||
+ ppd->dd->icode == ICODE_FUNCTIONAL_SIMULATOR ||
+ !HFI1_CAP_IS_KSET(QSFP_ENABLED)) {
+ ppd->driver_link_ready = 1;
+ return;
+ }
+
+ ppd->qsfp_info.ppd = ppd;
+ INIT_WORK(&ppd->qsfp_info.qsfp_work, qsfp_event);
+
+ qsfp_mask = (u64)(QSFP_HFI0_INT_N | QSFP_HFI0_MODPRST_N);
+ /* Clear current status to avoid spurious interrupts */
+ write_csr(dd,
+ dd->hfi1_id ?
+ ASIC_QSFP2_CLEAR :
+ ASIC_QSFP1_CLEAR,
+ qsfp_mask);
+
+ /* Handle active low nature of INT_N and MODPRST_N pins */
+ if (qsfp_mod_present(ppd))
+ qsfp_mask &= ~(u64)QSFP_HFI0_MODPRST_N;
+ write_csr(dd,
+ dd->hfi1_id ? ASIC_QSFP2_INVERT : ASIC_QSFP1_INVERT,
+ qsfp_mask);
+
+ /* Allow only INT_N and MODPRST_N to trigger QSFP interrupts */
+ qsfp_mask |= (u64)QSFP_HFI0_MODPRST_N;
+ write_csr(dd,
+ dd->hfi1_id ? ASIC_QSFP2_MASK : ASIC_QSFP1_MASK,
+ qsfp_mask);
+
+ if (qsfp_mod_present(ppd)) {
+ msleep(3000);
+ reset_qsfp(ppd);
+
+ /* Check for QSFP interrupt after t_init (SFF 8679)
+ * + extra
+ */
+ msleep(3000);
+ if (!ppd->qsfp_info.qsfp_interrupt_functional) {
+ if (do_qsfp_intr_fallback(ppd) < 0)
+ dd_dev_info(dd,
+ "%s: QSFP fallback failed\n",
+ __func__);
+ ppd->driver_link_ready = 1;
+ }
+ }
+}
+
+int bringup_serdes(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ u64 guid;
+ int ret;
+
+ if (HFI1_CAP_IS_KSET(EXTENDED_PSN))
+ add_rcvctrl(dd, RCV_CTRL_RCV_EXTENDED_PSN_ENABLE_SMASK);
+
+ guid = ppd->guid;
+ if (!guid) {
+ if (dd->base_guid)
+ guid = dd->base_guid + ppd->port - 1;
+ ppd->guid = guid;
+ }
+
+ /* the link defaults to enabled */
+ ppd->link_enabled = 1;
+ /* Set linkinit_reason on power up per OPA spec */
+ ppd->linkinit_reason = OPA_LINKINIT_REASON_LINKUP;
+
+ if (loopback) {
+ ret = init_loopback(dd);
+ if (ret < 0)
+ return ret;
+ }
+
+ return start_link(ppd);
+}
+
+void hfi1_quiet_serdes(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+
+ /*
+ * Shut down the link and keep it down. First turn off that the
+ * driver wants to allow the link to be up (driver_link_ready).
+ * Then make sure the link is not automatically restarted
+ * (link_enabled). Cancel any pending restart. And finally
+ * go offline.
+ */
+ ppd->driver_link_ready = 0;
+ ppd->link_enabled = 0;
+
+ set_link_down_reason(ppd, OPA_LINKDOWN_REASON_SMA_DISABLED, 0,
+ OPA_LINKDOWN_REASON_SMA_DISABLED);
+ set_link_state(ppd, HLS_DN_OFFLINE);
+
+ /* disable the port */
+ clear_rcvctrl(dd, RCV_CTRL_RCV_PORT_ENABLE_SMASK);
+}
+
+static inline int init_cpu_counters(struct hfi1_devdata *dd)
+{
+ struct hfi1_pportdata *ppd;
+ int i;
+
+ ppd = (struct hfi1_pportdata *)(dd + 1);
+ for (i = 0; i < dd->num_pports; i++, ppd++) {
+ ppd->ibport_data.rc_acks = NULL;
+ ppd->ibport_data.rc_qacks = NULL;
+ ppd->ibport_data.rc_acks = alloc_percpu(u64);
+ ppd->ibport_data.rc_qacks = alloc_percpu(u64);
+ ppd->ibport_data.rc_delayed_comp = alloc_percpu(u64);
+ if ((ppd->ibport_data.rc_acks == NULL) ||
+ (ppd->ibport_data.rc_delayed_comp == NULL) ||
+ (ppd->ibport_data.rc_qacks == NULL))
+ return -ENOMEM;
+ }
+
+ return 0;
+}
+
+static const char * const pt_names[] = {
+ "expected",
+ "eager",
+ "invalid"
+};
+
+static const char *pt_name(u32 type)
+{
+ return type >= ARRAY_SIZE(pt_names) ? "unknown" : pt_names[type];
+}
+
+/*
+ * index is the index into the receive array
+ */
+void hfi1_put_tid(struct hfi1_devdata *dd, u32 index,
+ u32 type, unsigned long pa, u16 order)
+{
+ u64 reg;
+ void __iomem *base = (dd->rcvarray_wc ? dd->rcvarray_wc :
+ (dd->kregbase + RCV_ARRAY));
+
+ if (!(dd->flags & HFI1_PRESENT))
+ goto done;
+
+ if (type == PT_INVALID) {
+ pa = 0;
+ } else if (type > PT_INVALID) {
+ dd_dev_err(dd,
+ "unexpected receive array type %u for index %u, not handled\n",
+ type, index);
+ goto done;
+ }
+
+ hfi1_cdbg(TID, "type %s, index 0x%x, pa 0x%lx, bsize 0x%lx",
+ pt_name(type), index, pa, (unsigned long)order);
+
+#define RT_ADDR_SHIFT 12 /* 4KB kernel address boundary */
+ reg = RCV_ARRAY_RT_WRITE_ENABLE_SMASK
+ | (u64)order << RCV_ARRAY_RT_BUF_SIZE_SHIFT
+ | ((pa >> RT_ADDR_SHIFT) & RCV_ARRAY_RT_ADDR_MASK)
+ << RCV_ARRAY_RT_ADDR_SHIFT;
+ writeq(reg, base + (index * 8));
+
+ if (type == PT_EAGER)
+ /*
+ * Eager entries are written one-by-one so we have to push them
+ * after we write the entry.
+ */
+ flush_wc();
+done:
+ return;
+}
+
+void hfi1_clear_tids(struct hfi1_ctxtdata *rcd)
+{
+ struct hfi1_devdata *dd = rcd->dd;
+ u32 i;
+
+ /* this could be optimized */
+ for (i = rcd->eager_base; i < rcd->eager_base +
+ rcd->egrbufs.alloced; i++)
+ hfi1_put_tid(dd, i, PT_INVALID, 0, 0);
+
+ for (i = rcd->expected_base;
+ i < rcd->expected_base + rcd->expected_count; i++)
+ hfi1_put_tid(dd, i, PT_INVALID, 0, 0);
+}
+
+int hfi1_get_base_kinfo(struct hfi1_ctxtdata *rcd,
+ struct hfi1_ctxt_info *kinfo)
+{
+ kinfo->runtime_flags = (HFI1_MISC_GET() << HFI1_CAP_USER_SHIFT) |
+ HFI1_CAP_UGET(MASK) | HFI1_CAP_KGET(K2U);
+ return 0;
+}
+
+struct hfi1_message_header *hfi1_get_msgheader(
+ struct hfi1_devdata *dd, __le32 *rhf_addr)
+{
+ u32 offset = rhf_hdrq_offset(rhf_to_cpu(rhf_addr));
+
+ return (struct hfi1_message_header *)
+ (rhf_addr - dd->rhf_offset + offset);
+}
+
+static const char * const ib_cfg_name_strings[] = {
+ "HFI1_IB_CFG_LIDLMC",
+ "HFI1_IB_CFG_LWID_DG_ENB",
+ "HFI1_IB_CFG_LWID_ENB",
+ "HFI1_IB_CFG_LWID",
+ "HFI1_IB_CFG_SPD_ENB",
+ "HFI1_IB_CFG_SPD",
+ "HFI1_IB_CFG_RXPOL_ENB",
+ "HFI1_IB_CFG_LREV_ENB",
+ "HFI1_IB_CFG_LINKLATENCY",
+ "HFI1_IB_CFG_HRTBT",
+ "HFI1_IB_CFG_OP_VLS",
+ "HFI1_IB_CFG_VL_HIGH_CAP",
+ "HFI1_IB_CFG_VL_LOW_CAP",
+ "HFI1_IB_CFG_OVERRUN_THRESH",
+ "HFI1_IB_CFG_PHYERR_THRESH",
+ "HFI1_IB_CFG_LINKDEFAULT",
+ "HFI1_IB_CFG_PKEYS",
+ "HFI1_IB_CFG_MTU",
+ "HFI1_IB_CFG_LSTATE",
+ "HFI1_IB_CFG_VL_HIGH_LIMIT",
+ "HFI1_IB_CFG_PMA_TICKS",
+ "HFI1_IB_CFG_PORT"
+};
+
+static const char *ib_cfg_name(int which)
+{
+ if (which < 0 || which >= ARRAY_SIZE(ib_cfg_name_strings))
+ return "invalid";
+ return ib_cfg_name_strings[which];
+}
+
+int hfi1_get_ib_cfg(struct hfi1_pportdata *ppd, int which)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ int val = 0;
+
+ switch (which) {
+ case HFI1_IB_CFG_LWID_ENB: /* allowed Link-width */
+ val = ppd->link_width_enabled;
+ break;
+ case HFI1_IB_CFG_LWID: /* currently active Link-width */
+ val = ppd->link_width_active;
+ break;
+ case HFI1_IB_CFG_SPD_ENB: /* allowed Link speeds */
+ val = ppd->link_speed_enabled;
+ break;
+ case HFI1_IB_CFG_SPD: /* current Link speed */
+ val = ppd->link_speed_active;
+ break;
+
+ case HFI1_IB_CFG_RXPOL_ENB: /* Auto-RX-polarity enable */
+ case HFI1_IB_CFG_LREV_ENB: /* Auto-Lane-reversal enable */
+ case HFI1_IB_CFG_LINKLATENCY:
+ goto unimplemented;
+
+ case HFI1_IB_CFG_OP_VLS:
+ val = ppd->vls_operational;
+ break;
+ case HFI1_IB_CFG_VL_HIGH_CAP: /* VL arb high priority table size */
+ val = VL_ARB_HIGH_PRIO_TABLE_SIZE;
+ break;
+ case HFI1_IB_CFG_VL_LOW_CAP: /* VL arb low priority table size */
+ val = VL_ARB_LOW_PRIO_TABLE_SIZE;
+ break;
+ case HFI1_IB_CFG_OVERRUN_THRESH: /* IB overrun threshold */
+ val = ppd->overrun_threshold;
+ break;
+ case HFI1_IB_CFG_PHYERR_THRESH: /* IB PHY error threshold */
+ val = ppd->phy_error_threshold;
+ break;
+ case HFI1_IB_CFG_LINKDEFAULT: /* IB link default (sleep/poll) */
+ val = dd->link_default;
+ break;
+
+ case HFI1_IB_CFG_HRTBT: /* Heartbeat off/enable/auto */
+ case HFI1_IB_CFG_PMA_TICKS:
+ default:
+unimplemented:
+ if (HFI1_CAP_IS_KSET(PRINT_UNIMPL))
+ dd_dev_info(
+ dd,
+ "%s: which %s: not implemented\n",
+ __func__,
+ ib_cfg_name(which));
+ break;
+ }
+
+ return val;
+}
+
+/*
+ * The largest MAD packet size.
+ */
+#define MAX_MAD_PACKET 2048
+
+/*
+ * Return the maximum header bytes that can go on the _wire_
+ * for this device. This count includes the ICRC which is
+ * not part of the packet held in memory but it is appended
+ * by the HW.
+ * This is dependent on the device's receive header entry size.
+ * HFI allows this to be set per-receive context, but the
+ * driver presently enforces a global value.
+ */
+u32 lrh_max_header_bytes(struct hfi1_devdata *dd)
+{
+ /*
+ * The maximum non-payload (MTU) bytes in LRH.PktLen are
+ * the Receive Header Entry Size minus the PBC (or RHF) size
+ * plus one DW for the ICRC appended by HW.
+ *
+ * dd->rcd[0].rcvhdrqentsize is in DW.
+ * We use rcd[0] as all context will have the same value. Also,
+ * the first kernel context would have been allocated by now so
+ * we are guaranteed a valid value.
+ */
+ return (dd->rcd[0]->rcvhdrqentsize - 2/*PBC/RHF*/ + 1/*ICRC*/) << 2;
+}
+
+/*
+ * Set Send Length
+ * @ppd - per port data
+ *
+ * Set the MTU by limiting how many DWs may be sent. The SendLenCheck*
+ * registers compare against LRH.PktLen, so use the max bytes included
+ * in the LRH.
+ *
+ * This routine changes all VL values except VL15, which it maintains at
+ * the same value.
+ */
+static void set_send_length(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ u32 max_hb = lrh_max_header_bytes(dd), maxvlmtu = 0, dcmtu;
+ u64 len1 = 0, len2 = (((dd->vld[15].mtu + max_hb) >> 2)
+ & SEND_LEN_CHECK1_LEN_VL15_MASK) <<
+ SEND_LEN_CHECK1_LEN_VL15_SHIFT;
+ int i;
+
+ for (i = 0; i < ppd->vls_supported; i++) {
+ if (dd->vld[i].mtu > maxvlmtu)
+ maxvlmtu = dd->vld[i].mtu;
+ if (i <= 3)
+ len1 |= (((dd->vld[i].mtu + max_hb) >> 2)
+ & SEND_LEN_CHECK0_LEN_VL0_MASK) <<
+ ((i % 4) * SEND_LEN_CHECK0_LEN_VL1_SHIFT);
+ else
+ len2 |= (((dd->vld[i].mtu + max_hb) >> 2)
+ & SEND_LEN_CHECK1_LEN_VL4_MASK) <<
+ ((i % 4) * SEND_LEN_CHECK1_LEN_VL5_SHIFT);
+ }
+ write_csr(dd, SEND_LEN_CHECK0, len1);
+ write_csr(dd, SEND_LEN_CHECK1, len2);
+ /* adjust kernel credit return thresholds based on new MTUs */
+ /* all kernel receive contexts have the same hdrqentsize */
+ for (i = 0; i < ppd->vls_supported; i++) {
+ sc_set_cr_threshold(dd->vld[i].sc,
+ sc_mtu_to_threshold(dd->vld[i].sc, dd->vld[i].mtu,
+ dd->rcd[0]->rcvhdrqentsize));
+ }
+ sc_set_cr_threshold(dd->vld[15].sc,
+ sc_mtu_to_threshold(dd->vld[15].sc, dd->vld[15].mtu,
+ dd->rcd[0]->rcvhdrqentsize));
+
+ /* Adjust maximum MTU for the port in DC */
+ dcmtu = maxvlmtu == 10240 ? DCC_CFG_PORT_MTU_CAP_10240 :
+ (ilog2(maxvlmtu >> 8) + 1);
+ len1 = read_csr(ppd->dd, DCC_CFG_PORT_CONFIG);
+ len1 &= ~DCC_CFG_PORT_CONFIG_MTU_CAP_SMASK;
+ len1 |= ((u64)dcmtu & DCC_CFG_PORT_CONFIG_MTU_CAP_MASK) <<
+ DCC_CFG_PORT_CONFIG_MTU_CAP_SHIFT;
+ write_csr(ppd->dd, DCC_CFG_PORT_CONFIG, len1);
+}
+
+static void set_lidlmc(struct hfi1_pportdata *ppd)
+{
+ int i;
+ u64 sreg = 0;
+ struct hfi1_devdata *dd = ppd->dd;
+ u32 mask = ~((1U << ppd->lmc) - 1);
+ u64 c1 = read_csr(ppd->dd, DCC_CFG_PORT_CONFIG1);
+
+ if (dd->hfi1_snoop.mode_flag)
+ dd_dev_info(dd, "Set lid/lmc while snooping");
+
+ c1 &= ~(DCC_CFG_PORT_CONFIG1_TARGET_DLID_SMASK
+ | DCC_CFG_PORT_CONFIG1_DLID_MASK_SMASK);
+ c1 |= ((ppd->lid & DCC_CFG_PORT_CONFIG1_TARGET_DLID_MASK)
+ << DCC_CFG_PORT_CONFIG1_TARGET_DLID_SHIFT)|
+ ((mask & DCC_CFG_PORT_CONFIG1_DLID_MASK_MASK)
+ << DCC_CFG_PORT_CONFIG1_DLID_MASK_SHIFT);
+ write_csr(ppd->dd, DCC_CFG_PORT_CONFIG1, c1);
+
+ /*
+ * Iterate over all the send contexts and set their SLID check
+ */
+ sreg = ((mask & SEND_CTXT_CHECK_SLID_MASK_MASK) <<
+ SEND_CTXT_CHECK_SLID_MASK_SHIFT) |
+ (((ppd->lid & mask) & SEND_CTXT_CHECK_SLID_VALUE_MASK) <<
+ SEND_CTXT_CHECK_SLID_VALUE_SHIFT);
+
+ for (i = 0; i < dd->chip_send_contexts; i++) {
+ hfi1_cdbg(LINKVERB, "SendContext[%d].SLID_CHECK = 0x%x",
+ i, (u32)sreg);
+ write_kctxt_csr(dd, i, SEND_CTXT_CHECK_SLID, sreg);
+ }
+
+ /* Now we have to do the same thing for the sdma engines */
+ sdma_update_lmc(dd, mask, ppd->lid);
+}
+
+static int wait_phy_linkstate(struct hfi1_devdata *dd, u32 state, u32 msecs)
+{
+ unsigned long timeout;
+ u32 curr_state;
+
+ timeout = jiffies + msecs_to_jiffies(msecs);
+ while (1) {
+ curr_state = read_physical_state(dd);
+ if (curr_state == state)
+ break;
+ if (time_after(jiffies, timeout)) {
+ dd_dev_err(dd,
+ "timeout waiting for phy link state 0x%x, current state is 0x%x\n",
+ state, curr_state);
+ return -ETIMEDOUT;
+ }
+ usleep_range(1950, 2050); /* sleep 2ms-ish */
+ }
+
+ return 0;
+}
+
+/*
+ * Helper for set_link_state(). Do not call except from that routine.
+ * Expects ppd->hls_mutex to be held.
+ *
+ * @rem_reason value to be sent to the neighbor
+ *
+ * LinkDownReasons only set if transition succeeds.
+ */
+static int goto_offline(struct hfi1_pportdata *ppd, u8 rem_reason)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ u32 pstate, previous_state;
+ u32 last_local_state;
+ u32 last_remote_state;
+ int ret;
+ int do_transition;
+ int do_wait;
+
+ previous_state = ppd->host_link_state;
+ ppd->host_link_state = HLS_GOING_OFFLINE;
+ pstate = read_physical_state(dd);
+ if (pstate == PLS_OFFLINE) {
+ do_transition = 0; /* in right state */
+ do_wait = 0; /* ...no need to wait */
+ } else if ((pstate & 0xff) == PLS_OFFLINE) {
+ do_transition = 0; /* in an offline transient state */
+ do_wait = 1; /* ...wait for it to settle */
+ } else {
+ do_transition = 1; /* need to move to offline */
+ do_wait = 1; /* ...will need to wait */
+ }
+
+ if (do_transition) {
+ ret = set_physical_link_state(dd,
+ PLS_OFFLINE | (rem_reason << 8));
+
+ if (ret != HCMD_SUCCESS) {
+ dd_dev_err(dd,
+ "Failed to transition to Offline link state, return %d\n",
+ ret);
+ return -EINVAL;
+ }
+ if (ppd->offline_disabled_reason == OPA_LINKDOWN_REASON_NONE)
+ ppd->offline_disabled_reason =
+ OPA_LINKDOWN_REASON_TRANSIENT;
+ }
+
+ if (do_wait) {
+ /* it can take a while for the link to go down */
+ ret = wait_phy_linkstate(dd, PLS_OFFLINE, 5000);
+ if (ret < 0)
+ return ret;
+ }
+
+ /* make sure the logical state is also down */
+ wait_logical_linkstate(ppd, IB_PORT_DOWN, 1000);
+
+ /*
+ * Now in charge of LCB - must be after the physical state is
+ * offline.quiet and before host_link_state is changed.
+ */
+ set_host_lcb_access(dd);
+ write_csr(dd, DC_LCB_ERR_EN, ~0ull); /* watch LCB errors */
+ ppd->host_link_state = HLS_LINK_COOLDOWN; /* LCB access allowed */
+
+ /*
+ * The LNI has a mandatory wait time after the physical state
+ * moves to Offline.Quiet. The wait time may be different
+ * depending on how the link went down. The 8051 firmware
+ * will observe the needed wait time and only move to ready
+ * when that is completed. The largest of the quiet timeouts
+ * is 2.5s, so wait that long and then a bit more.
+ */
+ ret = wait_fm_ready(dd, 3000);
+ if (ret) {
+ dd_dev_err(dd,
+ "After going offline, timed out waiting for the 8051 to become ready to accept host requests\n");
+ /* state is really offline, so make it so */
+ ppd->host_link_state = HLS_DN_OFFLINE;
+ return ret;
+ }
+
+ /*
+ * The state is now offline and the 8051 is ready to accept host
+ * requests.
+ * - change our state
+ * - notify others if we were previously in a linkup state
+ */
+ ppd->host_link_state = HLS_DN_OFFLINE;
+ if (previous_state & HLS_UP) {
+ /* went down while link was up */
+ handle_linkup_change(dd, 0);
+ } else if (previous_state
+ & (HLS_DN_POLL | HLS_VERIFY_CAP | HLS_GOING_UP)) {
+ /* went down while attempting link up */
+ /* byte 1 of last_*_state is the failure reason */
+ read_last_local_state(dd, &last_local_state);
+ read_last_remote_state(dd, &last_remote_state);
+ dd_dev_err(dd,
+ "LNI failure last states: local 0x%08x, remote 0x%08x\n",
+ last_local_state, last_remote_state);
+ }
+
+ /* the active link width (downgrade) is 0 on link down */
+ ppd->link_width_active = 0;
+ ppd->link_width_downgrade_tx_active = 0;
+ ppd->link_width_downgrade_rx_active = 0;
+ ppd->current_egress_rate = 0;
+ return 0;
+}
+
+/* return the link state name */
+static const char *link_state_name(u32 state)
+{
+ const char *name;
+ int n = ilog2(state);
+ static const char * const names[] = {
+ [__HLS_UP_INIT_BP] = "INIT",
+ [__HLS_UP_ARMED_BP] = "ARMED",
+ [__HLS_UP_ACTIVE_BP] = "ACTIVE",
+ [__HLS_DN_DOWNDEF_BP] = "DOWNDEF",
+ [__HLS_DN_POLL_BP] = "POLL",
+ [__HLS_DN_DISABLE_BP] = "DISABLE",
+ [__HLS_DN_OFFLINE_BP] = "OFFLINE",
+ [__HLS_VERIFY_CAP_BP] = "VERIFY_CAP",
+ [__HLS_GOING_UP_BP] = "GOING_UP",
+ [__HLS_GOING_OFFLINE_BP] = "GOING_OFFLINE",
+ [__HLS_LINK_COOLDOWN_BP] = "LINK_COOLDOWN"
+ };
+
+ name = n < ARRAY_SIZE(names) ? names[n] : NULL;
+ return name ? name : "unknown";
+}
+
+/* return the link state reason name */
+static const char *link_state_reason_name(struct hfi1_pportdata *ppd, u32 state)
+{
+ if (state == HLS_UP_INIT) {
+ switch (ppd->linkinit_reason) {
+ case OPA_LINKINIT_REASON_LINKUP:
+ return "(LINKUP)";
+ case OPA_LINKINIT_REASON_FLAPPING:
+ return "(FLAPPING)";
+ case OPA_LINKINIT_OUTSIDE_POLICY:
+ return "(OUTSIDE_POLICY)";
+ case OPA_LINKINIT_QUARANTINED:
+ return "(QUARANTINED)";
+ case OPA_LINKINIT_INSUFIC_CAPABILITY:
+ return "(INSUFIC_CAPABILITY)";
+ default:
+ break;
+ }
+ }
+ return "";
+}
+
+/*
+ * driver_physical_state - convert the driver's notion of a port's
+ * state (an HLS_*) into a physical state (a {IB,OPA}_PORTPHYSSTATE_*).
+ * Return -1 (converted to a u32) to indicate error.
+ */
+u32 driver_physical_state(struct hfi1_pportdata *ppd)
+{
+ switch (ppd->host_link_state) {
+ case HLS_UP_INIT:
+ case HLS_UP_ARMED:
+ case HLS_UP_ACTIVE:
+ return IB_PORTPHYSSTATE_LINKUP;
+ case HLS_DN_POLL:
+ return IB_PORTPHYSSTATE_POLLING;
+ case HLS_DN_DISABLE:
+ return IB_PORTPHYSSTATE_DISABLED;
+ case HLS_DN_OFFLINE:
+ return OPA_PORTPHYSSTATE_OFFLINE;
+ case HLS_VERIFY_CAP:
+ return IB_PORTPHYSSTATE_POLLING;
+ case HLS_GOING_UP:
+ return IB_PORTPHYSSTATE_POLLING;
+ case HLS_GOING_OFFLINE:
+ return OPA_PORTPHYSSTATE_OFFLINE;
+ case HLS_LINK_COOLDOWN:
+ return OPA_PORTPHYSSTATE_OFFLINE;
+ case HLS_DN_DOWNDEF:
+ default:
+ dd_dev_err(ppd->dd, "invalid host_link_state 0x%x\n",
+ ppd->host_link_state);
+ return -1;
+ }
+}
+
+/*
+ * driver_logical_state - convert the driver's notion of a port's
+ * state (an HLS_*) into a logical state (a IB_PORT_*). Return -1
+ * (converted to a u32) to indicate error.
+ */
+u32 driver_logical_state(struct hfi1_pportdata *ppd)
+{
+ if (ppd->host_link_state && !(ppd->host_link_state & HLS_UP))
+ return IB_PORT_DOWN;
+
+ switch (ppd->host_link_state & HLS_UP) {
+ case HLS_UP_INIT:
+ return IB_PORT_INIT;
+ case HLS_UP_ARMED:
+ return IB_PORT_ARMED;
+ case HLS_UP_ACTIVE:
+ return IB_PORT_ACTIVE;
+ default:
+ dd_dev_err(ppd->dd, "invalid host_link_state 0x%x\n",
+ ppd->host_link_state);
+ return -1;
+ }
+}
+
+void set_link_down_reason(struct hfi1_pportdata *ppd, u8 lcl_reason,
+ u8 neigh_reason, u8 rem_reason)
+{
+ if (ppd->local_link_down_reason.latest == 0 &&
+ ppd->neigh_link_down_reason.latest == 0) {
+ ppd->local_link_down_reason.latest = lcl_reason;
+ ppd->neigh_link_down_reason.latest = neigh_reason;
+ ppd->remote_link_down_reason = rem_reason;
+ }
+}
+
+/*
+ * Change the physical and/or logical link state.
+ *
+ * Do not call this routine while inside an interrupt. It contains
+ * calls to routines that can take multiple seconds to finish.
+ *
+ * Returns 0 on success, -errno on failure.
+ */
+int set_link_state(struct hfi1_pportdata *ppd, u32 state)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ struct ib_event event = {.device = NULL};
+ int ret1, ret = 0;
+ int was_up, is_down;
+ int orig_new_state, poll_bounce;
+
+ mutex_lock(&ppd->hls_lock);
+
+ orig_new_state = state;
+ if (state == HLS_DN_DOWNDEF)
+ state = dd->link_default;
+
+ /* interpret poll -> poll as a link bounce */
+ poll_bounce = ppd->host_link_state == HLS_DN_POLL
+ && state == HLS_DN_POLL;
+
+ dd_dev_info(dd, "%s: current %s, new %s %s%s\n", __func__,
+ link_state_name(ppd->host_link_state),
+ link_state_name(orig_new_state),
+ poll_bounce ? "(bounce) " : "",
+ link_state_reason_name(ppd, state));
+
+ was_up = !!(ppd->host_link_state & HLS_UP);
+
+ /*
+ * If we're going to a (HLS_*) link state that implies the logical
+ * link state is neither of (IB_PORT_ARMED, IB_PORT_ACTIVE), then
+ * reset is_sm_config_started to 0.
+ */
+ if (!(state & (HLS_UP_ARMED | HLS_UP_ACTIVE)))
+ ppd->is_sm_config_started = 0;
+
+ /*
+ * Do nothing if the states match. Let a poll to poll link bounce
+ * go through.
+ */
+ if (ppd->host_link_state == state && !poll_bounce)
+ goto done;
+
+ switch (state) {
+ case HLS_UP_INIT:
+ if (ppd->host_link_state == HLS_DN_POLL && (quick_linkup
+ || dd->icode == ICODE_FUNCTIONAL_SIMULATOR)) {
+ /*
+ * Quick link up jumps from polling to here.
+ *
+ * Whether in normal or loopback mode, the
+ * simulator jumps from polling to link up.
+ * Accept that here.
+ */
+ /* OK */;
+ } else if (ppd->host_link_state != HLS_GOING_UP) {
+ goto unexpected;
+ }
+
+ ppd->host_link_state = HLS_UP_INIT;
+ ret = wait_logical_linkstate(ppd, IB_PORT_INIT, 1000);
+ if (ret) {
+ /* logical state didn't change, stay at going_up */
+ ppd->host_link_state = HLS_GOING_UP;
+ dd_dev_err(dd,
+ "%s: logical state did not change to INIT\n",
+ __func__);
+ } else {
+ /* clear old transient LINKINIT_REASON code */
+ if (ppd->linkinit_reason >= OPA_LINKINIT_REASON_CLEAR)
+ ppd->linkinit_reason =
+ OPA_LINKINIT_REASON_LINKUP;
+
+ /* enable the port */
+ add_rcvctrl(dd, RCV_CTRL_RCV_PORT_ENABLE_SMASK);
+
+ handle_linkup_change(dd, 1);
+ }
+ break;
+ case HLS_UP_ARMED:
+ if (ppd->host_link_state != HLS_UP_INIT)
+ goto unexpected;
+
+ ppd->host_link_state = HLS_UP_ARMED;
+ set_logical_state(dd, LSTATE_ARMED);
+ ret = wait_logical_linkstate(ppd, IB_PORT_ARMED, 1000);
+ if (ret) {
+ /* logical state didn't change, stay at init */
+ ppd->host_link_state = HLS_UP_INIT;
+ dd_dev_err(dd,
+ "%s: logical state did not change to ARMED\n",
+ __func__);
+ }
+ /*
+ * The simulator does not currently implement SMA messages,
+ * so neighbor_normal is not set. Set it here when we first
+ * move to Armed.
+ */
+ if (dd->icode == ICODE_FUNCTIONAL_SIMULATOR)
+ ppd->neighbor_normal = 1;
+ break;
+ case HLS_UP_ACTIVE:
+ if (ppd->host_link_state != HLS_UP_ARMED)
+ goto unexpected;
+
+ ppd->host_link_state = HLS_UP_ACTIVE;
+ set_logical_state(dd, LSTATE_ACTIVE);
+ ret = wait_logical_linkstate(ppd, IB_PORT_ACTIVE, 1000);
+ if (ret) {
+ /* logical state didn't change, stay at armed */
+ ppd->host_link_state = HLS_UP_ARMED;
+ dd_dev_err(dd,
+ "%s: logical state did not change to ACTIVE\n",
+ __func__);
+ } else {
+
+ /* tell all engines to go running */
+ sdma_all_running(dd);
+
+ /* Signal the IB layer that the port has went active */
+ event.device = &dd->verbs_dev.ibdev;
+ event.element.port_num = ppd->port;
+ event.event = IB_EVENT_PORT_ACTIVE;
+ }
+ break;
+ case HLS_DN_POLL:
+ if ((ppd->host_link_state == HLS_DN_DISABLE ||
+ ppd->host_link_state == HLS_DN_OFFLINE) &&
+ dd->dc_shutdown)
+ dc_start(dd);
+ /* Hand LED control to the DC */
+ write_csr(dd, DCC_CFG_LED_CNTRL, 0);
+
+ if (ppd->host_link_state != HLS_DN_OFFLINE) {
+ u8 tmp = ppd->link_enabled;
+
+ ret = goto_offline(ppd, ppd->remote_link_down_reason);
+ if (ret) {
+ ppd->link_enabled = tmp;
+ break;
+ }
+ ppd->remote_link_down_reason = 0;
+
+ if (ppd->driver_link_ready)
+ ppd->link_enabled = 1;
+ }
+
+ ret = set_local_link_attributes(ppd);
+ if (ret)
+ break;
+
+ ppd->port_error_action = 0;
+ ppd->host_link_state = HLS_DN_POLL;
+
+ if (quick_linkup) {
+ /* quick linkup does not go into polling */
+ ret = do_quick_linkup(dd);
+ } else {
+ ret1 = set_physical_link_state(dd, PLS_POLLING);
+ if (ret1 != HCMD_SUCCESS) {
+ dd_dev_err(dd,
+ "Failed to transition to Polling link state, return 0x%x\n",
+ ret1);
+ ret = -EINVAL;
+ }
+ }
+ ppd->offline_disabled_reason = OPA_LINKDOWN_REASON_NONE;
+ /*
+ * If an error occurred above, go back to offline. The
+ * caller may reschedule another attempt.
+ */
+ if (ret)
+ goto_offline(ppd, 0);
+ break;
+ case HLS_DN_DISABLE:
+ /* link is disabled */
+ ppd->link_enabled = 0;
+
+ /* allow any state to transition to disabled */
+
+ /* must transition to offline first */
+ if (ppd->host_link_state != HLS_DN_OFFLINE) {
+ ret = goto_offline(ppd, ppd->remote_link_down_reason);
+ if (ret)
+ break;
+ ppd->remote_link_down_reason = 0;
+ }
+
+ ret1 = set_physical_link_state(dd, PLS_DISABLED);
+ if (ret1 != HCMD_SUCCESS) {
+ dd_dev_err(dd,
+ "Failed to transition to Disabled link state, return 0x%x\n",
+ ret1);
+ ret = -EINVAL;
+ break;
+ }
+ ppd->host_link_state = HLS_DN_DISABLE;
+ dc_shutdown(dd);
+ break;
+ case HLS_DN_OFFLINE:
+ if (ppd->host_link_state == HLS_DN_DISABLE)
+ dc_start(dd);
+
+ /* allow any state to transition to offline */
+ ret = goto_offline(ppd, ppd->remote_link_down_reason);
+ if (!ret)
+ ppd->remote_link_down_reason = 0;
+ break;
+ case HLS_VERIFY_CAP:
+ if (ppd->host_link_state != HLS_DN_POLL)
+ goto unexpected;
+ ppd->host_link_state = HLS_VERIFY_CAP;
+ break;
+ case HLS_GOING_UP:
+ if (ppd->host_link_state != HLS_VERIFY_CAP)
+ goto unexpected;
+
+ ret1 = set_physical_link_state(dd, PLS_LINKUP);
+ if (ret1 != HCMD_SUCCESS) {
+ dd_dev_err(dd,
+ "Failed to transition to link up state, return 0x%x\n",
+ ret1);
+ ret = -EINVAL;
+ break;
+ }
+ ppd->host_link_state = HLS_GOING_UP;
+ break;
+
+ case HLS_GOING_OFFLINE: /* transient within goto_offline() */
+ case HLS_LINK_COOLDOWN: /* transient within goto_offline() */
+ default:
+ dd_dev_info(dd, "%s: state 0x%x: not supported\n",
+ __func__, state);
+ ret = -EINVAL;
+ break;
+ }
+
+ is_down = !!(ppd->host_link_state & (HLS_DN_POLL |
+ HLS_DN_DISABLE | HLS_DN_OFFLINE));
+
+ if (was_up && is_down && ppd->local_link_down_reason.sma == 0 &&
+ ppd->neigh_link_down_reason.sma == 0) {
+ ppd->local_link_down_reason.sma =
+ ppd->local_link_down_reason.latest;
+ ppd->neigh_link_down_reason.sma =
+ ppd->neigh_link_down_reason.latest;
+ }
+
+ goto done;
+
+unexpected:
+ dd_dev_err(dd, "%s: unexpected state transition from %s to %s\n",
+ __func__, link_state_name(ppd->host_link_state),
+ link_state_name(state));
+ ret = -EINVAL;
+
+done:
+ mutex_unlock(&ppd->hls_lock);
+
+ if (event.device)
+ ib_dispatch_event(&event);
+
+ return ret;
+}
+
+int hfi1_set_ib_cfg(struct hfi1_pportdata *ppd, int which, u32 val)
+{
+ u64 reg;
+ int ret = 0;
+
+ switch (which) {
+ case HFI1_IB_CFG_LIDLMC:
+ set_lidlmc(ppd);
+ break;
+ case HFI1_IB_CFG_VL_HIGH_LIMIT:
+ /*
+ * The VL Arbitrator high limit is sent in units of 4k
+ * bytes, while HFI stores it in units of 64 bytes.
+ */
+ val *= 4096/64;
+ reg = ((u64)val & SEND_HIGH_PRIORITY_LIMIT_LIMIT_MASK)
+ << SEND_HIGH_PRIORITY_LIMIT_LIMIT_SHIFT;
+ write_csr(ppd->dd, SEND_HIGH_PRIORITY_LIMIT, reg);
+ break;
+ case HFI1_IB_CFG_LINKDEFAULT: /* IB link default (sleep/poll) */
+ /* HFI only supports POLL as the default link down state */
+ if (val != HLS_DN_POLL)
+ ret = -EINVAL;
+ break;
+ case HFI1_IB_CFG_OP_VLS:
+ if (ppd->vls_operational != val) {
+ ppd->vls_operational = val;
+ if (!ppd->port)
+ ret = -EINVAL;
+ else
+ ret = sdma_map_init(
+ ppd->dd,
+ ppd->port - 1,
+ val,
+ NULL);
+ }
+ break;
+ /*
+ * For link width, link width downgrade, and speed enable, always AND
+ * the setting with what is actually supported. This has two benefits.
+ * First, enabled can't have unsupported values, no matter what the
+ * SM or FM might want. Second, the ALL_SUPPORTED wildcards that mean
+ * "fill in with your supported value" have all the bits in the
+ * field set, so simply ANDing with supported has the desired result.
+ */
+ case HFI1_IB_CFG_LWID_ENB: /* set allowed Link-width */
+ ppd->link_width_enabled = val & ppd->link_width_supported;
+ break;
+ case HFI1_IB_CFG_LWID_DG_ENB: /* set allowed link width downgrade */
+ ppd->link_width_downgrade_enabled =
+ val & ppd->link_width_downgrade_supported;
+ break;
+ case HFI1_IB_CFG_SPD_ENB: /* allowed Link speeds */
+ ppd->link_speed_enabled = val & ppd->link_speed_supported;
+ break;
+ case HFI1_IB_CFG_OVERRUN_THRESH: /* IB overrun threshold */
+ /*
+ * HFI does not follow IB specs, save this value
+ * so we can report it, if asked.
+ */
+ ppd->overrun_threshold = val;
+ break;
+ case HFI1_IB_CFG_PHYERR_THRESH: /* IB PHY error threshold */
+ /*
+ * HFI does not follow IB specs, save this value
+ * so we can report it, if asked.
+ */
+ ppd->phy_error_threshold = val;
+ break;
+
+ case HFI1_IB_CFG_MTU:
+ set_send_length(ppd);
+ break;
+
+ case HFI1_IB_CFG_PKEYS:
+ if (HFI1_CAP_IS_KSET(PKEY_CHECK))
+ set_partition_keys(ppd);
+ break;
+
+ default:
+ if (HFI1_CAP_IS_KSET(PRINT_UNIMPL))
+ dd_dev_info(ppd->dd,
+ "%s: which %s, val 0x%x: not implemented\n",
+ __func__, ib_cfg_name(which), val);
+ break;
+ }
+ return ret;
+}
+
+/* begin functions related to vl arbitration table caching */
+static void init_vl_arb_caches(struct hfi1_pportdata *ppd)
+{
+ int i;
+
+ BUILD_BUG_ON(VL_ARB_TABLE_SIZE !=
+ VL_ARB_LOW_PRIO_TABLE_SIZE);
+ BUILD_BUG_ON(VL_ARB_TABLE_SIZE !=
+ VL_ARB_HIGH_PRIO_TABLE_SIZE);
+
+ /*
+ * Note that we always return values directly from the
+ * 'vl_arb_cache' (and do no CSR reads) in response to a
+ * 'Get(VLArbTable)'. This is obviously correct after a
+ * 'Set(VLArbTable)', since the cache will then be up to
+ * date. But it's also correct prior to any 'Set(VLArbTable)'
+ * since then both the cache, and the relevant h/w registers
+ * will be zeroed.
+ */
+
+ for (i = 0; i < MAX_PRIO_TABLE; i++)
+ spin_lock_init(&ppd->vl_arb_cache[i].lock);
+}
+
+/*
+ * vl_arb_lock_cache
+ *
+ * All other vl_arb_* functions should be called only after locking
+ * the cache.
+ */
+static inline struct vl_arb_cache *
+vl_arb_lock_cache(struct hfi1_pportdata *ppd, int idx)
+{
+ if (idx != LO_PRIO_TABLE && idx != HI_PRIO_TABLE)
+ return NULL;
+ spin_lock(&ppd->vl_arb_cache[idx].lock);
+ return &ppd->vl_arb_cache[idx];
+}
+
+static inline void vl_arb_unlock_cache(struct hfi1_pportdata *ppd, int idx)
+{
+ spin_unlock(&ppd->vl_arb_cache[idx].lock);
+}
+
+static void vl_arb_get_cache(struct vl_arb_cache *cache,
+ struct ib_vl_weight_elem *vl)
+{
+ memcpy(vl, cache->table, VL_ARB_TABLE_SIZE * sizeof(*vl));
+}
+
+static void vl_arb_set_cache(struct vl_arb_cache *cache,
+ struct ib_vl_weight_elem *vl)
+{
+ memcpy(cache->table, vl, VL_ARB_TABLE_SIZE * sizeof(*vl));
+}
+
+static int vl_arb_match_cache(struct vl_arb_cache *cache,
+ struct ib_vl_weight_elem *vl)
+{
+ return !memcmp(cache->table, vl, VL_ARB_TABLE_SIZE * sizeof(*vl));
+}
+/* end functions related to vl arbitration table caching */
+
+static int set_vl_weights(struct hfi1_pportdata *ppd, u32 target,
+ u32 size, struct ib_vl_weight_elem *vl)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ u64 reg;
+ unsigned int i, is_up = 0;
+ int drain, ret = 0;
+
+ mutex_lock(&ppd->hls_lock);
+
+ if (ppd->host_link_state & HLS_UP)
+ is_up = 1;
+
+ drain = !is_ax(dd) && is_up;
+
+ if (drain)
+ /*
+ * Before adjusting VL arbitration weights, empty per-VL
+ * FIFOs, otherwise a packet whose VL weight is being
+ * set to 0 could get stuck in a FIFO with no chance to
+ * egress.
+ */
+ ret = stop_drain_data_vls(dd);
+
+ if (ret) {
+ dd_dev_err(
+ dd,
+ "%s: cannot stop/drain VLs - refusing to change VL arbitration weights\n",
+ __func__);
+ goto err;
+ }
+
+ for (i = 0; i < size; i++, vl++) {
+ /*
+ * NOTE: The low priority shift and mask are used here, but
+ * they are the same for both the low and high registers.
+ */
+ reg = (((u64)vl->vl & SEND_LOW_PRIORITY_LIST_VL_MASK)
+ << SEND_LOW_PRIORITY_LIST_VL_SHIFT)
+ | (((u64)vl->weight
+ & SEND_LOW_PRIORITY_LIST_WEIGHT_MASK)
+ << SEND_LOW_PRIORITY_LIST_WEIGHT_SHIFT);
+ write_csr(dd, target + (i * 8), reg);
+ }
+ pio_send_control(dd, PSC_GLOBAL_VLARB_ENABLE);
+
+ if (drain)
+ open_fill_data_vls(dd); /* reopen all VLs */
+
+err:
+ mutex_unlock(&ppd->hls_lock);
+
+ return ret;
+}
+
+/*
+ * Read one credit merge VL register.
+ */
+static void read_one_cm_vl(struct hfi1_devdata *dd, u32 csr,
+ struct vl_limit *vll)
+{
+ u64 reg = read_csr(dd, csr);
+
+ vll->dedicated = cpu_to_be16(
+ (reg >> SEND_CM_CREDIT_VL_DEDICATED_LIMIT_VL_SHIFT)
+ & SEND_CM_CREDIT_VL_DEDICATED_LIMIT_VL_MASK);
+ vll->shared = cpu_to_be16(
+ (reg >> SEND_CM_CREDIT_VL_SHARED_LIMIT_VL_SHIFT)
+ & SEND_CM_CREDIT_VL_SHARED_LIMIT_VL_MASK);
+}
+
+/*
+ * Read the current credit merge limits.
+ */
+static int get_buffer_control(struct hfi1_devdata *dd,
+ struct buffer_control *bc, u16 *overall_limit)
+{
+ u64 reg;
+ int i;
+
+ /* not all entries are filled in */
+ memset(bc, 0, sizeof(*bc));
+
+ /* OPA and HFI have a 1-1 mapping */
+ for (i = 0; i < TXE_NUM_DATA_VL; i++)
+ read_one_cm_vl(dd, SEND_CM_CREDIT_VL + (8*i), &bc->vl[i]);
+
+ /* NOTE: assumes that VL* and VL15 CSRs are bit-wise identical */
+ read_one_cm_vl(dd, SEND_CM_CREDIT_VL15, &bc->vl[15]);
+
+ reg = read_csr(dd, SEND_CM_GLOBAL_CREDIT);
+ bc->overall_shared_limit = cpu_to_be16(
+ (reg >> SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_SHIFT)
+ & SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_MASK);
+ if (overall_limit)
+ *overall_limit = (reg
+ >> SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_SHIFT)
+ & SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_MASK;
+ return sizeof(struct buffer_control);
+}
+
+static int get_sc2vlnt(struct hfi1_devdata *dd, struct sc2vlnt *dp)
+{
+ u64 reg;
+ int i;
+
+ /* each register contains 16 SC->VLnt mappings, 4 bits each */
+ reg = read_csr(dd, DCC_CFG_SC_VL_TABLE_15_0);
+ for (i = 0; i < sizeof(u64); i++) {
+ u8 byte = *(((u8 *)&reg) + i);
+
+ dp->vlnt[2 * i] = byte & 0xf;
+ dp->vlnt[(2 * i) + 1] = (byte & 0xf0) >> 4;
+ }
+
+ reg = read_csr(dd, DCC_CFG_SC_VL_TABLE_31_16);
+ for (i = 0; i < sizeof(u64); i++) {
+ u8 byte = *(((u8 *)&reg) + i);
+
+ dp->vlnt[16 + (2 * i)] = byte & 0xf;
+ dp->vlnt[16 + (2 * i) + 1] = (byte & 0xf0) >> 4;
+ }
+ return sizeof(struct sc2vlnt);
+}
+
+static void get_vlarb_preempt(struct hfi1_devdata *dd, u32 nelems,
+ struct ib_vl_weight_elem *vl)
+{
+ unsigned int i;
+
+ for (i = 0; i < nelems; i++, vl++) {
+ vl->vl = 0xf;
+ vl->weight = 0;
+ }
+}
+
+static void set_sc2vlnt(struct hfi1_devdata *dd, struct sc2vlnt *dp)
+{
+ write_csr(dd, DCC_CFG_SC_VL_TABLE_15_0,
+ DC_SC_VL_VAL(15_0,
+ 0, dp->vlnt[0] & 0xf,
+ 1, dp->vlnt[1] & 0xf,
+ 2, dp->vlnt[2] & 0xf,
+ 3, dp->vlnt[3] & 0xf,
+ 4, dp->vlnt[4] & 0xf,
+ 5, dp->vlnt[5] & 0xf,
+ 6, dp->vlnt[6] & 0xf,
+ 7, dp->vlnt[7] & 0xf,
+ 8, dp->vlnt[8] & 0xf,
+ 9, dp->vlnt[9] & 0xf,
+ 10, dp->vlnt[10] & 0xf,
+ 11, dp->vlnt[11] & 0xf,
+ 12, dp->vlnt[12] & 0xf,
+ 13, dp->vlnt[13] & 0xf,
+ 14, dp->vlnt[14] & 0xf,
+ 15, dp->vlnt[15] & 0xf));
+ write_csr(dd, DCC_CFG_SC_VL_TABLE_31_16,
+ DC_SC_VL_VAL(31_16,
+ 16, dp->vlnt[16] & 0xf,
+ 17, dp->vlnt[17] & 0xf,
+ 18, dp->vlnt[18] & 0xf,
+ 19, dp->vlnt[19] & 0xf,
+ 20, dp->vlnt[20] & 0xf,
+ 21, dp->vlnt[21] & 0xf,
+ 22, dp->vlnt[22] & 0xf,
+ 23, dp->vlnt[23] & 0xf,
+ 24, dp->vlnt[24] & 0xf,
+ 25, dp->vlnt[25] & 0xf,
+ 26, dp->vlnt[26] & 0xf,
+ 27, dp->vlnt[27] & 0xf,
+ 28, dp->vlnt[28] & 0xf,
+ 29, dp->vlnt[29] & 0xf,
+ 30, dp->vlnt[30] & 0xf,
+ 31, dp->vlnt[31] & 0xf));
+}
+
+static void nonzero_msg(struct hfi1_devdata *dd, int idx, const char *what,
+ u16 limit)
+{
+ if (limit != 0)
+ dd_dev_info(dd, "Invalid %s limit %d on VL %d, ignoring\n",
+ what, (int)limit, idx);
+}
+
+/* change only the shared limit portion of SendCmGLobalCredit */
+static void set_global_shared(struct hfi1_devdata *dd, u16 limit)
+{
+ u64 reg;
+
+ reg = read_csr(dd, SEND_CM_GLOBAL_CREDIT);
+ reg &= ~SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_SMASK;
+ reg |= (u64)limit << SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_SHIFT;
+ write_csr(dd, SEND_CM_GLOBAL_CREDIT, reg);
+}
+
+/* change only the total credit limit portion of SendCmGLobalCredit */
+static void set_global_limit(struct hfi1_devdata *dd, u16 limit)
+{
+ u64 reg;
+
+ reg = read_csr(dd, SEND_CM_GLOBAL_CREDIT);
+ reg &= ~SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_SMASK;
+ reg |= (u64)limit << SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_SHIFT;
+ write_csr(dd, SEND_CM_GLOBAL_CREDIT, reg);
+}
+
+/* set the given per-VL shared limit */
+static void set_vl_shared(struct hfi1_devdata *dd, int vl, u16 limit)
+{
+ u64 reg;
+ u32 addr;
+
+ if (vl < TXE_NUM_DATA_VL)
+ addr = SEND_CM_CREDIT_VL + (8 * vl);
+ else
+ addr = SEND_CM_CREDIT_VL15;
+
+ reg = read_csr(dd, addr);
+ reg &= ~SEND_CM_CREDIT_VL_SHARED_LIMIT_VL_SMASK;
+ reg |= (u64)limit << SEND_CM_CREDIT_VL_SHARED_LIMIT_VL_SHIFT;
+ write_csr(dd, addr, reg);
+}
+
+/* set the given per-VL dedicated limit */
+static void set_vl_dedicated(struct hfi1_devdata *dd, int vl, u16 limit)
+{
+ u64 reg;
+ u32 addr;
+
+ if (vl < TXE_NUM_DATA_VL)
+ addr = SEND_CM_CREDIT_VL + (8 * vl);
+ else
+ addr = SEND_CM_CREDIT_VL15;
+
+ reg = read_csr(dd, addr);
+ reg &= ~SEND_CM_CREDIT_VL_DEDICATED_LIMIT_VL_SMASK;
+ reg |= (u64)limit << SEND_CM_CREDIT_VL_DEDICATED_LIMIT_VL_SHIFT;
+ write_csr(dd, addr, reg);
+}
+
+/* spin until the given per-VL status mask bits clear */
+static void wait_for_vl_status_clear(struct hfi1_devdata *dd, u64 mask,
+ const char *which)
+{
+ unsigned long timeout;
+ u64 reg;
+
+ timeout = jiffies + msecs_to_jiffies(VL_STATUS_CLEAR_TIMEOUT);
+ while (1) {
+ reg = read_csr(dd, SEND_CM_CREDIT_USED_STATUS) & mask;
+
+ if (reg == 0)
+ return; /* success */
+ if (time_after(jiffies, timeout))
+ break; /* timed out */
+ udelay(1);
+ }
+
+ dd_dev_err(dd,
+ "%s credit change status not clearing after %dms, mask 0x%llx, not clear 0x%llx\n",
+ which, VL_STATUS_CLEAR_TIMEOUT, mask, reg);
+ /*
+ * If this occurs, it is likely there was a credit loss on the link.
+ * The only recovery from that is a link bounce.
+ */
+ dd_dev_err(dd,
+ "Continuing anyway. A credit loss may occur. Suggest a link bounce\n");
+}
+
+/*
+ * The number of credits on the VLs may be changed while everything
+ * is "live", but the following algorithm must be followed due to
+ * how the hardware is actually implemented. In particular,
+ * Return_Credit_Status[] is the only correct status check.
+ *
+ * if (reducing Global_Shared_Credit_Limit or any shared limit changing)
+ * set Global_Shared_Credit_Limit = 0
+ * use_all_vl = 1
+ * mask0 = all VLs that are changing either dedicated or shared limits
+ * set Shared_Limit[mask0] = 0
+ * spin until Return_Credit_Status[use_all_vl ? all VL : mask0] == 0
+ * if (changing any dedicated limit)
+ * mask1 = all VLs that are lowering dedicated limits
+ * lower Dedicated_Limit[mask1]
+ * spin until Return_Credit_Status[mask1] == 0
+ * raise Dedicated_Limits
+ * raise Shared_Limits
+ * raise Global_Shared_Credit_Limit
+ *
+ * lower = if the new limit is lower, set the limit to the new value
+ * raise = if the new limit is higher than the current value (may be changed
+ * earlier in the algorithm), set the new limit to the new value
+ */
+static int set_buffer_control(struct hfi1_devdata *dd,
+ struct buffer_control *new_bc)
+{
+ u64 changing_mask, ld_mask, stat_mask;
+ int change_count;
+ int i, use_all_mask;
+ int this_shared_changing;
+ /*
+ * A0: add the variable any_shared_limit_changing below and in the
+ * algorithm above. If removing A0 support, it can be removed.
+ */
+ int any_shared_limit_changing;
+ struct buffer_control cur_bc;
+ u8 changing[OPA_MAX_VLS];
+ u8 lowering_dedicated[OPA_MAX_VLS];
+ u16 cur_total;
+ u32 new_total = 0;
+ const u64 all_mask =
+ SEND_CM_CREDIT_USED_STATUS_VL0_RETURN_CREDIT_STATUS_SMASK
+ | SEND_CM_CREDIT_USED_STATUS_VL1_RETURN_CREDIT_STATUS_SMASK
+ | SEND_CM_CREDIT_USED_STATUS_VL2_RETURN_CREDIT_STATUS_SMASK
+ | SEND_CM_CREDIT_USED_STATUS_VL3_RETURN_CREDIT_STATUS_SMASK
+ | SEND_CM_CREDIT_USED_STATUS_VL4_RETURN_CREDIT_STATUS_SMASK
+ | SEND_CM_CREDIT_USED_STATUS_VL5_RETURN_CREDIT_STATUS_SMASK
+ | SEND_CM_CREDIT_USED_STATUS_VL6_RETURN_CREDIT_STATUS_SMASK
+ | SEND_CM_CREDIT_USED_STATUS_VL7_RETURN_CREDIT_STATUS_SMASK
+ | SEND_CM_CREDIT_USED_STATUS_VL15_RETURN_CREDIT_STATUS_SMASK;
+
+#define valid_vl(idx) ((idx) < TXE_NUM_DATA_VL || (idx) == 15)
+#define NUM_USABLE_VLS 16 /* look at VL15 and less */
+
+
+ /* find the new total credits, do sanity check on unused VLs */
+ for (i = 0; i < OPA_MAX_VLS; i++) {
+ if (valid_vl(i)) {
+ new_total += be16_to_cpu(new_bc->vl[i].dedicated);
+ continue;
+ }
+ nonzero_msg(dd, i, "dedicated",
+ be16_to_cpu(new_bc->vl[i].dedicated));
+ nonzero_msg(dd, i, "shared",
+ be16_to_cpu(new_bc->vl[i].shared));
+ new_bc->vl[i].dedicated = 0;
+ new_bc->vl[i].shared = 0;
+ }
+ new_total += be16_to_cpu(new_bc->overall_shared_limit);
+ if (new_total > (u32)dd->link_credits)
+ return -EINVAL;
+ /* fetch the current values */
+ get_buffer_control(dd, &cur_bc, &cur_total);
+
+ /*
+ * Create the masks we will use.
+ */
+ memset(changing, 0, sizeof(changing));
+ memset(lowering_dedicated, 0, sizeof(lowering_dedicated));
+ /* NOTE: Assumes that the individual VL bits are adjacent and in
+ increasing order */
+ stat_mask =
+ SEND_CM_CREDIT_USED_STATUS_VL0_RETURN_CREDIT_STATUS_SMASK;
+ changing_mask = 0;
+ ld_mask = 0;
+ change_count = 0;
+ any_shared_limit_changing = 0;
+ for (i = 0; i < NUM_USABLE_VLS; i++, stat_mask <<= 1) {
+ if (!valid_vl(i))
+ continue;
+ this_shared_changing = new_bc->vl[i].shared
+ != cur_bc.vl[i].shared;
+ if (this_shared_changing)
+ any_shared_limit_changing = 1;
+ if (new_bc->vl[i].dedicated != cur_bc.vl[i].dedicated
+ || this_shared_changing) {
+ changing[i] = 1;
+ changing_mask |= stat_mask;
+ change_count++;
+ }
+ if (be16_to_cpu(new_bc->vl[i].dedicated) <
+ be16_to_cpu(cur_bc.vl[i].dedicated)) {
+ lowering_dedicated[i] = 1;
+ ld_mask |= stat_mask;
+ }
+ }
+
+ /* bracket the credit change with a total adjustment */
+ if (new_total > cur_total)
+ set_global_limit(dd, new_total);
+
+ /*
+ * Start the credit change algorithm.
+ */
+ use_all_mask = 0;
+ if ((be16_to_cpu(new_bc->overall_shared_limit) <
+ be16_to_cpu(cur_bc.overall_shared_limit))
+ || (is_a0(dd) && any_shared_limit_changing)) {
+ set_global_shared(dd, 0);
+ cur_bc.overall_shared_limit = 0;
+ use_all_mask = 1;
+ }
+
+ for (i = 0; i < NUM_USABLE_VLS; i++) {
+ if (!valid_vl(i))
+ continue;
+
+ if (changing[i]) {
+ set_vl_shared(dd, i, 0);
+ cur_bc.vl[i].shared = 0;
+ }
+ }
+
+ wait_for_vl_status_clear(dd, use_all_mask ? all_mask : changing_mask,
+ "shared");
+
+ if (change_count > 0) {
+ for (i = 0; i < NUM_USABLE_VLS; i++) {
+ if (!valid_vl(i))
+ continue;
+
+ if (lowering_dedicated[i]) {
+ set_vl_dedicated(dd, i,
+ be16_to_cpu(new_bc->vl[i].dedicated));
+ cur_bc.vl[i].dedicated =
+ new_bc->vl[i].dedicated;
+ }
+ }
+
+ wait_for_vl_status_clear(dd, ld_mask, "dedicated");
+
+ /* now raise all dedicated that are going up */
+ for (i = 0; i < NUM_USABLE_VLS; i++) {
+ if (!valid_vl(i))
+ continue;
+
+ if (be16_to_cpu(new_bc->vl[i].dedicated) >
+ be16_to_cpu(cur_bc.vl[i].dedicated))
+ set_vl_dedicated(dd, i,
+ be16_to_cpu(new_bc->vl[i].dedicated));
+ }
+ }
+
+ /* next raise all shared that are going up */
+ for (i = 0; i < NUM_USABLE_VLS; i++) {
+ if (!valid_vl(i))
+ continue;
+
+ if (be16_to_cpu(new_bc->vl[i].shared) >
+ be16_to_cpu(cur_bc.vl[i].shared))
+ set_vl_shared(dd, i, be16_to_cpu(new_bc->vl[i].shared));
+ }
+
+ /* finally raise the global shared */
+ if (be16_to_cpu(new_bc->overall_shared_limit) >
+ be16_to_cpu(cur_bc.overall_shared_limit))
+ set_global_shared(dd,
+ be16_to_cpu(new_bc->overall_shared_limit));
+
+ /* bracket the credit change with a total adjustment */
+ if (new_total < cur_total)
+ set_global_limit(dd, new_total);
+ return 0;
+}
+
+/*
+ * Read the given fabric manager table. Return the size of the
+ * table (in bytes) on success, and a negative error code on
+ * failure.
+ */
+int fm_get_table(struct hfi1_pportdata *ppd, int which, void *t)
+
+{
+ int size;
+ struct vl_arb_cache *vlc;
+
+ switch (which) {
+ case FM_TBL_VL_HIGH_ARB:
+ size = 256;
+ /*
+ * OPA specifies 128 elements (of 2 bytes each), though
+ * HFI supports only 16 elements in h/w.
+ */
+ vlc = vl_arb_lock_cache(ppd, HI_PRIO_TABLE);
+ vl_arb_get_cache(vlc, t);
+ vl_arb_unlock_cache(ppd, HI_PRIO_TABLE);
+ break;
+ case FM_TBL_VL_LOW_ARB:
+ size = 256;
+ /*
+ * OPA specifies 128 elements (of 2 bytes each), though
+ * HFI supports only 16 elements in h/w.
+ */
+ vlc = vl_arb_lock_cache(ppd, LO_PRIO_TABLE);
+ vl_arb_get_cache(vlc, t);
+ vl_arb_unlock_cache(ppd, LO_PRIO_TABLE);
+ break;
+ case FM_TBL_BUFFER_CONTROL:
+ size = get_buffer_control(ppd->dd, t, NULL);
+ break;
+ case FM_TBL_SC2VLNT:
+ size = get_sc2vlnt(ppd->dd, t);
+ break;
+ case FM_TBL_VL_PREEMPT_ELEMS:
+ size = 256;
+ /* OPA specifies 128 elements, of 2 bytes each */
+ get_vlarb_preempt(ppd->dd, OPA_MAX_VLS, t);
+ break;
+ case FM_TBL_VL_PREEMPT_MATRIX:
+ size = 256;
+ /*
+ * OPA specifies that this is the same size as the VL
+ * arbitration tables (i.e., 256 bytes).
+ */
+ break;
+ default:
+ return -EINVAL;
+ }
+ return size;
+}
+
+/*
+ * Write the given fabric manager table.
+ */
+int fm_set_table(struct hfi1_pportdata *ppd, int which, void *t)
+{
+ int ret = 0;
+ struct vl_arb_cache *vlc;
+
+ switch (which) {
+ case FM_TBL_VL_HIGH_ARB:
+ vlc = vl_arb_lock_cache(ppd, HI_PRIO_TABLE);
+ if (vl_arb_match_cache(vlc, t)) {
+ vl_arb_unlock_cache(ppd, HI_PRIO_TABLE);
+ break;
+ }
+ vl_arb_set_cache(vlc, t);
+ vl_arb_unlock_cache(ppd, HI_PRIO_TABLE);
+ ret = set_vl_weights(ppd, SEND_HIGH_PRIORITY_LIST,
+ VL_ARB_HIGH_PRIO_TABLE_SIZE, t);
+ break;
+ case FM_TBL_VL_LOW_ARB:
+ vlc = vl_arb_lock_cache(ppd, LO_PRIO_TABLE);
+ if (vl_arb_match_cache(vlc, t)) {
+ vl_arb_unlock_cache(ppd, LO_PRIO_TABLE);
+ break;
+ }
+ vl_arb_set_cache(vlc, t);
+ vl_arb_unlock_cache(ppd, LO_PRIO_TABLE);
+ ret = set_vl_weights(ppd, SEND_LOW_PRIORITY_LIST,
+ VL_ARB_LOW_PRIO_TABLE_SIZE, t);
+ break;
+ case FM_TBL_BUFFER_CONTROL:
+ ret = set_buffer_control(ppd->dd, t);
+ break;
+ case FM_TBL_SC2VLNT:
+ set_sc2vlnt(ppd->dd, t);
+ break;
+ default:
+ ret = -EINVAL;
+ }
+ return ret;
+}
+
+/*
+ * Disable all data VLs.
+ *
+ * Return 0 if disabled, non-zero if the VLs cannot be disabled.
+ */
+static int disable_data_vls(struct hfi1_devdata *dd)
+{
+ if (is_a0(dd))
+ return 1;
+
+ pio_send_control(dd, PSC_DATA_VL_DISABLE);
+
+ return 0;
+}
+
+/*
+ * open_fill_data_vls() - the counterpart to stop_drain_data_vls().
+ * Just re-enables all data VLs (the "fill" part happens
+ * automatically - the name was chosen for symmetry with
+ * stop_drain_data_vls()).
+ *
+ * Return 0 if successful, non-zero if the VLs cannot be enabled.
+ */
+int open_fill_data_vls(struct hfi1_devdata *dd)
+{
+ if (is_a0(dd))
+ return 1;
+
+ pio_send_control(dd, PSC_DATA_VL_ENABLE);
+
+ return 0;
+}
+
+/*
+ * drain_data_vls() - assumes that disable_data_vls() has been called,
+ * wait for occupancy (of per-VL FIFOs) for all contexts, and SDMA
+ * engines to drop to 0.
+ */
+static void drain_data_vls(struct hfi1_devdata *dd)
+{
+ sc_wait(dd);
+ sdma_wait(dd);
+ pause_for_credit_return(dd);
+}
+
+/*
+ * stop_drain_data_vls() - disable, then drain all per-VL fifos.
+ *
+ * Use open_fill_data_vls() to resume using data VLs. This pair is
+ * meant to be used like this:
+ *
+ * stop_drain_data_vls(dd);
+ * // do things with per-VL resources
+ * open_fill_data_vls(dd);
+ */
+int stop_drain_data_vls(struct hfi1_devdata *dd)
+{
+ int ret;
+
+ ret = disable_data_vls(dd);
+ if (ret == 0)
+ drain_data_vls(dd);
+
+ return ret;
+}
+
+/*
+ * Convert a nanosecond time to a cclock count. No matter how slow
+ * the cclock, a non-zero ns will always have a non-zero result.
+ */
+u32 ns_to_cclock(struct hfi1_devdata *dd, u32 ns)
+{
+ u32 cclocks;
+
+ if (dd->icode == ICODE_FPGA_EMULATION)
+ cclocks = (ns * 1000) / FPGA_CCLOCK_PS;
+ else /* simulation pretends to be ASIC */
+ cclocks = (ns * 1000) / ASIC_CCLOCK_PS;
+ if (ns && !cclocks) /* if ns nonzero, must be at least 1 */
+ cclocks = 1;
+ return cclocks;
+}
+
+/*
+ * Convert a cclock count to nanoseconds. Not matter how slow
+ * the cclock, a non-zero cclocks will always have a non-zero result.
+ */
+u32 cclock_to_ns(struct hfi1_devdata *dd, u32 cclocks)
+{
+ u32 ns;
+
+ if (dd->icode == ICODE_FPGA_EMULATION)
+ ns = (cclocks * FPGA_CCLOCK_PS) / 1000;
+ else /* simulation pretends to be ASIC */
+ ns = (cclocks * ASIC_CCLOCK_PS) / 1000;
+ if (cclocks && !ns)
+ ns = 1;
+ return ns;
+}
+
+/*
+ * Dynamically adjust the receive interrupt timeout for a context based on
+ * incoming packet rate.
+ *
+ * NOTE: Dynamic adjustment does not allow rcv_intr_count to be zero.
+ */
+static void adjust_rcv_timeout(struct hfi1_ctxtdata *rcd, u32 npkts)
+{
+ struct hfi1_devdata *dd = rcd->dd;
+ u32 timeout = rcd->rcvavail_timeout;
+
+ /*
+ * This algorithm doubles or halves the timeout depending on whether
+ * the number of packets received in this interrupt were less than or
+ * greater equal the interrupt count.
+ *
+ * The calculations below do not allow a steady state to be achieved.
+ * Only at the endpoints it is possible to have an unchanging
+ * timeout.
+ */
+ if (npkts < rcv_intr_count) {
+ /*
+ * Not enough packets arrived before the timeout, adjust
+ * timeout downward.
+ */
+ if (timeout < 2) /* already at minimum? */
+ return;
+ timeout >>= 1;
+ } else {
+ /*
+ * More than enough packets arrived before the timeout, adjust
+ * timeout upward.
+ */
+ if (timeout >= dd->rcv_intr_timeout_csr) /* already at max? */
+ return;
+ timeout = min(timeout << 1, dd->rcv_intr_timeout_csr);
+ }
+
+ rcd->rcvavail_timeout = timeout;
+ /* timeout cannot be larger than rcv_intr_timeout_csr which has already
+ been verified to be in range */
+ write_kctxt_csr(dd, rcd->ctxt, RCV_AVAIL_TIME_OUT,
+ (u64)timeout << RCV_AVAIL_TIME_OUT_TIME_OUT_RELOAD_SHIFT);
+}
+
+void update_usrhead(struct hfi1_ctxtdata *rcd, u32 hd, u32 updegr, u32 egrhd,
+ u32 intr_adjust, u32 npkts)
+{
+ struct hfi1_devdata *dd = rcd->dd;
+ u64 reg;
+ u32 ctxt = rcd->ctxt;
+
+ /*
+ * Need to write timeout register before updating RcvHdrHead to ensure
+ * that a new value is used when the HW decides to restart counting.
+ */
+ if (intr_adjust)
+ adjust_rcv_timeout(rcd, npkts);
+ if (updegr) {
+ reg = (egrhd & RCV_EGR_INDEX_HEAD_HEAD_MASK)
+ << RCV_EGR_INDEX_HEAD_HEAD_SHIFT;
+ write_uctxt_csr(dd, ctxt, RCV_EGR_INDEX_HEAD, reg);
+ }
+ mmiowb();
+ reg = ((u64)rcv_intr_count << RCV_HDR_HEAD_COUNTER_SHIFT) |
+ (((u64)hd & RCV_HDR_HEAD_HEAD_MASK)
+ << RCV_HDR_HEAD_HEAD_SHIFT);
+ write_uctxt_csr(dd, ctxt, RCV_HDR_HEAD, reg);
+ mmiowb();
+}
+
+u32 hdrqempty(struct hfi1_ctxtdata *rcd)
+{
+ u32 head, tail;
+
+ head = (read_uctxt_csr(rcd->dd, rcd->ctxt, RCV_HDR_HEAD)
+ & RCV_HDR_HEAD_HEAD_SMASK) >> RCV_HDR_HEAD_HEAD_SHIFT;
+
+ if (rcd->rcvhdrtail_kvaddr)
+ tail = get_rcvhdrtail(rcd);
+ else
+ tail = read_uctxt_csr(rcd->dd, rcd->ctxt, RCV_HDR_TAIL);
+
+ return head == tail;
+}
+
+/*
+ * Context Control and Receive Array encoding for buffer size:
+ * 0x0 invalid
+ * 0x1 4 KB
+ * 0x2 8 KB
+ * 0x3 16 KB
+ * 0x4 32 KB
+ * 0x5 64 KB
+ * 0x6 128 KB
+ * 0x7 256 KB
+ * 0x8 512 KB (Receive Array only)
+ * 0x9 1 MB (Receive Array only)
+ * 0xa 2 MB (Receive Array only)
+ *
+ * 0xB-0xF - reserved (Receive Array only)
+ *
+ *
+ * This routine assumes that the value has already been sanity checked.
+ */
+static u32 encoded_size(u32 size)
+{
+ switch (size) {
+ case 4*1024: return 0x1;
+ case 8*1024: return 0x2;
+ case 16*1024: return 0x3;
+ case 32*1024: return 0x4;
+ case 64*1024: return 0x5;
+ case 128*1024: return 0x6;
+ case 256*1024: return 0x7;
+ case 512*1024: return 0x8;
+ case 1*1024*1024: return 0x9;
+ case 2*1024*1024: return 0xa;
+ }
+ return 0x1; /* if invalid, go with the minimum size */
+}
+
+void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op, int ctxt)
+{
+ struct hfi1_ctxtdata *rcd;
+ u64 rcvctrl, reg;
+ int did_enable = 0;
+
+ rcd = dd->rcd[ctxt];
+ if (!rcd)
+ return;
+
+ hfi1_cdbg(RCVCTRL, "ctxt %d op 0x%x", ctxt, op);
+
+ rcvctrl = read_kctxt_csr(dd, ctxt, RCV_CTXT_CTRL);
+ /* if the context already enabled, don't do the extra steps */
+ if ((op & HFI1_RCVCTRL_CTXT_ENB)
+ && !(rcvctrl & RCV_CTXT_CTRL_ENABLE_SMASK)) {
+ /* reset the tail and hdr addresses, and sequence count */
+ write_kctxt_csr(dd, ctxt, RCV_HDR_ADDR,
+ rcd->rcvhdrq_phys);
+ if (HFI1_CAP_KGET_MASK(rcd->flags, DMA_RTAIL))
+ write_kctxt_csr(dd, ctxt, RCV_HDR_TAIL_ADDR,
+ rcd->rcvhdrqtailaddr_phys);
+ rcd->seq_cnt = 1;
+
+ /* reset the cached receive header queue head value */
+ rcd->head = 0;
+
+ /*
+ * Zero the receive header queue so we don't get false
+ * positives when checking the sequence number. The
+ * sequence numbers could land exactly on the same spot.
+ * E.g. a rcd restart before the receive header wrapped.
+ */
+ memset(rcd->rcvhdrq, 0, rcd->rcvhdrq_size);
+
+ /* starting timeout */
+ rcd->rcvavail_timeout = dd->rcv_intr_timeout_csr;
+
+ /* enable the context */
+ rcvctrl |= RCV_CTXT_CTRL_ENABLE_SMASK;
+
+ /* clean the egr buffer size first */
+ rcvctrl &= ~RCV_CTXT_CTRL_EGR_BUF_SIZE_SMASK;
+ rcvctrl |= ((u64)encoded_size(rcd->egrbufs.rcvtid_size)
+ & RCV_CTXT_CTRL_EGR_BUF_SIZE_MASK)
+ << RCV_CTXT_CTRL_EGR_BUF_SIZE_SHIFT;
+
+ /* zero RcvHdrHead - set RcvHdrHead.Counter after enable */
+ write_uctxt_csr(dd, ctxt, RCV_HDR_HEAD, 0);
+ did_enable = 1;
+
+ /* zero RcvEgrIndexHead */
+ write_uctxt_csr(dd, ctxt, RCV_EGR_INDEX_HEAD, 0);
+
+ /* set eager count and base index */
+ reg = (((u64)(rcd->egrbufs.alloced >> RCV_SHIFT)
+ & RCV_EGR_CTRL_EGR_CNT_MASK)
+ << RCV_EGR_CTRL_EGR_CNT_SHIFT) |
+ (((rcd->eager_base >> RCV_SHIFT)
+ & RCV_EGR_CTRL_EGR_BASE_INDEX_MASK)
+ << RCV_EGR_CTRL_EGR_BASE_INDEX_SHIFT);
+ write_kctxt_csr(dd, ctxt, RCV_EGR_CTRL, reg);
+
+ /*
+ * Set TID (expected) count and base index.
+ * rcd->expected_count is set to individual RcvArray entries,
+ * not pairs, and the CSR takes a pair-count in groups of
+ * four, so divide by 8.
+ */
+ reg = (((rcd->expected_count >> RCV_SHIFT)
+ & RCV_TID_CTRL_TID_PAIR_CNT_MASK)
+ << RCV_TID_CTRL_TID_PAIR_CNT_SHIFT) |
+ (((rcd->expected_base >> RCV_SHIFT)
+ & RCV_TID_CTRL_TID_BASE_INDEX_MASK)
+ << RCV_TID_CTRL_TID_BASE_INDEX_SHIFT);
+ write_kctxt_csr(dd, ctxt, RCV_TID_CTRL, reg);
+ if (ctxt == VL15CTXT)
+ write_csr(dd, RCV_VL15, VL15CTXT);
+ }
+ if (op & HFI1_RCVCTRL_CTXT_DIS) {
+ write_csr(dd, RCV_VL15, 0);
+ rcvctrl &= ~RCV_CTXT_CTRL_ENABLE_SMASK;
+ }
+ if (op & HFI1_RCVCTRL_INTRAVAIL_ENB)
+ rcvctrl |= RCV_CTXT_CTRL_INTR_AVAIL_SMASK;
+ if (op & HFI1_RCVCTRL_INTRAVAIL_DIS)
+ rcvctrl &= ~RCV_CTXT_CTRL_INTR_AVAIL_SMASK;
+ if (op & HFI1_RCVCTRL_TAILUPD_ENB && rcd->rcvhdrqtailaddr_phys)
+ rcvctrl |= RCV_CTXT_CTRL_TAIL_UPD_SMASK;
+ if (op & HFI1_RCVCTRL_TAILUPD_DIS)
+ rcvctrl &= ~RCV_CTXT_CTRL_TAIL_UPD_SMASK;
+ if (op & HFI1_RCVCTRL_TIDFLOW_ENB)
+ rcvctrl |= RCV_CTXT_CTRL_TID_FLOW_ENABLE_SMASK;
+ if (op & HFI1_RCVCTRL_TIDFLOW_DIS)
+ rcvctrl &= ~RCV_CTXT_CTRL_TID_FLOW_ENABLE_SMASK;
+ if (op & HFI1_RCVCTRL_ONE_PKT_EGR_ENB) {
+ /* In one-packet-per-eager mode, the size comes from
+ the RcvArray entry. */
+ rcvctrl &= ~RCV_CTXT_CTRL_EGR_BUF_SIZE_SMASK;
+ rcvctrl |= RCV_CTXT_CTRL_ONE_PACKET_PER_EGR_BUFFER_SMASK;
+ }
+ if (op & HFI1_RCVCTRL_ONE_PKT_EGR_DIS)
+ rcvctrl &= ~RCV_CTXT_CTRL_ONE_PACKET_PER_EGR_BUFFER_SMASK;
+ if (op & HFI1_RCVCTRL_NO_RHQ_DROP_ENB)
+ rcvctrl |= RCV_CTXT_CTRL_DONT_DROP_RHQ_FULL_SMASK;
+ if (op & HFI1_RCVCTRL_NO_RHQ_DROP_DIS)
+ rcvctrl &= ~RCV_CTXT_CTRL_DONT_DROP_RHQ_FULL_SMASK;
+ if (op & HFI1_RCVCTRL_NO_EGR_DROP_ENB)
+ rcvctrl |= RCV_CTXT_CTRL_DONT_DROP_EGR_FULL_SMASK;
+ if (op & HFI1_RCVCTRL_NO_EGR_DROP_DIS)
+ rcvctrl &= ~RCV_CTXT_CTRL_DONT_DROP_EGR_FULL_SMASK;
+ rcd->rcvctrl = rcvctrl;
+ hfi1_cdbg(RCVCTRL, "ctxt %d rcvctrl 0x%llx\n", ctxt, rcvctrl);
+ write_kctxt_csr(dd, ctxt, RCV_CTXT_CTRL, rcd->rcvctrl);
+
+ /* work around sticky RcvCtxtStatus.BlockedRHQFull */
+ if (did_enable
+ && (rcvctrl & RCV_CTXT_CTRL_DONT_DROP_RHQ_FULL_SMASK)) {
+ reg = read_kctxt_csr(dd, ctxt, RCV_CTXT_STATUS);
+ if (reg != 0) {
+ dd_dev_info(dd, "ctxt %d status %lld (blocked)\n",
+ ctxt, reg);
+ read_uctxt_csr(dd, ctxt, RCV_HDR_HEAD);
+ write_uctxt_csr(dd, ctxt, RCV_HDR_HEAD, 0x10);
+ write_uctxt_csr(dd, ctxt, RCV_HDR_HEAD, 0x00);
+ read_uctxt_csr(dd, ctxt, RCV_HDR_HEAD);
+ reg = read_kctxt_csr(dd, ctxt, RCV_CTXT_STATUS);
+ dd_dev_info(dd, "ctxt %d status %lld (%s blocked)\n",
+ ctxt, reg, reg == 0 ? "not" : "still");
+ }
+ }
+
+ if (did_enable) {
+ /*
+ * The interrupt timeout and count must be set after
+ * the context is enabled to take effect.
+ */
+ /* set interrupt timeout */
+ write_kctxt_csr(dd, ctxt, RCV_AVAIL_TIME_OUT,
+ (u64)rcd->rcvavail_timeout <<
+ RCV_AVAIL_TIME_OUT_TIME_OUT_RELOAD_SHIFT);
+
+ /* set RcvHdrHead.Counter, zero RcvHdrHead.Head (again) */
+ reg = (u64)rcv_intr_count << RCV_HDR_HEAD_COUNTER_SHIFT;
+ write_uctxt_csr(dd, ctxt, RCV_HDR_HEAD, reg);
+ }
+
+ if (op & (HFI1_RCVCTRL_TAILUPD_DIS | HFI1_RCVCTRL_CTXT_DIS))
+ /*
+ * If the context has been disabled and the Tail Update has
+ * been cleared, clear the RCV_HDR_TAIL_ADDR CSR so
+ * it doesn't contain an address that is invalid.
+ */
+ write_kctxt_csr(dd, ctxt, RCV_HDR_TAIL_ADDR, 0);
+}
+
+u32 hfi1_read_cntrs(struct hfi1_devdata *dd, loff_t pos, char **namep,
+ u64 **cntrp)
+{
+ int ret;
+ u64 val = 0;
+
+ if (namep) {
+ ret = dd->cntrnameslen;
+ if (pos != 0) {
+ dd_dev_err(dd, "read_cntrs does not support indexing");
+ return 0;
+ }
+ *namep = dd->cntrnames;
+ } else {
+ const struct cntr_entry *entry;
+ int i, j;
+
+ ret = (dd->ndevcntrs) * sizeof(u64);
+ if (pos != 0) {
+ dd_dev_err(dd, "read_cntrs does not support indexing");
+ return 0;
+ }
+
+ /* Get the start of the block of counters */
+ *cntrp = dd->cntrs;
+
+ /*
+ * Now go and fill in each counter in the block.
+ */
+ for (i = 0; i < DEV_CNTR_LAST; i++) {
+ entry = &dev_cntrs[i];
+ hfi1_cdbg(CNTR, "reading %s", entry->name);
+ if (entry->flags & CNTR_DISABLED) {
+ /* Nothing */
+ hfi1_cdbg(CNTR, "\tDisabled\n");
+ } else {
+ if (entry->flags & CNTR_VL) {
+ hfi1_cdbg(CNTR, "\tPer VL\n");
+ for (j = 0; j < C_VL_COUNT; j++) {
+ val = entry->rw_cntr(entry,
+ dd, j,
+ CNTR_MODE_R,
+ 0);
+ hfi1_cdbg(
+ CNTR,
+ "\t\tRead 0x%llx for %d\n",
+ val, j);
+ dd->cntrs[entry->offset + j] =
+ val;
+ }
+ } else {
+ val = entry->rw_cntr(entry, dd,
+ CNTR_INVALID_VL,
+ CNTR_MODE_R, 0);
+ dd->cntrs[entry->offset] = val;
+ hfi1_cdbg(CNTR, "\tRead 0x%llx", val);
+ }
+ }
+ }
+ }
+ return ret;
+}
+
+/*
+ * Used by sysfs to create files for hfi stats to read
+ */
+u32 hfi1_read_portcntrs(struct hfi1_devdata *dd, loff_t pos, u32 port,
+ char **namep, u64 **cntrp)
+{
+ int ret;
+ u64 val = 0;
+
+ if (namep) {
+ ret = dd->portcntrnameslen;
+ if (pos != 0) {
+ dd_dev_err(dd, "index not supported");
+ return 0;
+ }
+ *namep = dd->portcntrnames;
+ } else {
+ const struct cntr_entry *entry;
+ struct hfi1_pportdata *ppd;
+ int i, j;
+
+ ret = (dd->nportcntrs) * sizeof(u64);
+ if (pos != 0) {
+ dd_dev_err(dd, "indexing not supported");
+ return 0;
+ }
+ ppd = (struct hfi1_pportdata *)(dd + 1 + port);
+ *cntrp = ppd->cntrs;
+
+ for (i = 0; i < PORT_CNTR_LAST; i++) {
+ entry = &port_cntrs[i];
+ hfi1_cdbg(CNTR, "reading %s", entry->name);
+ if (entry->flags & CNTR_DISABLED) {
+ /* Nothing */
+ hfi1_cdbg(CNTR, "\tDisabled\n");
+ continue;
+ }
+
+ if (entry->flags & CNTR_VL) {
+ hfi1_cdbg(CNTR, "\tPer VL");
+ for (j = 0; j < C_VL_COUNT; j++) {
+ val = entry->rw_cntr(entry, ppd, j,
+ CNTR_MODE_R,
+ 0);
+ hfi1_cdbg(
+ CNTR,
+ "\t\tRead 0x%llx for %d",
+ val, j);
+ ppd->cntrs[entry->offset + j] = val;
+ }
+ } else {
+ val = entry->rw_cntr(entry, ppd,
+ CNTR_INVALID_VL,
+ CNTR_MODE_R,
+ 0);
+ ppd->cntrs[entry->offset] = val;
+ hfi1_cdbg(CNTR, "\tRead 0x%llx", val);
+ }
+ }
+ }
+ return ret;
+}
+
+static void free_cntrs(struct hfi1_devdata *dd)
+{
+ struct hfi1_pportdata *ppd;
+ int i;
+
+ if (dd->synth_stats_timer.data)
+ del_timer_sync(&dd->synth_stats_timer);
+ dd->synth_stats_timer.data = 0;
+ ppd = (struct hfi1_pportdata *)(dd + 1);
+ for (i = 0; i < dd->num_pports; i++, ppd++) {
+ kfree(ppd->cntrs);
+ kfree(ppd->scntrs);
+ free_percpu(ppd->ibport_data.rc_acks);
+ free_percpu(ppd->ibport_data.rc_qacks);
+ free_percpu(ppd->ibport_data.rc_delayed_comp);
+ ppd->cntrs = NULL;
+ ppd->scntrs = NULL;
+ ppd->ibport_data.rc_acks = NULL;
+ ppd->ibport_data.rc_qacks = NULL;
+ ppd->ibport_data.rc_delayed_comp = NULL;
+ }
+ kfree(dd->portcntrnames);
+ dd->portcntrnames = NULL;
+ kfree(dd->cntrs);
+ dd->cntrs = NULL;
+ kfree(dd->scntrs);
+ dd->scntrs = NULL;
+ kfree(dd->cntrnames);
+ dd->cntrnames = NULL;
+}
+
+#define CNTR_MAX 0xFFFFFFFFFFFFFFFFULL
+#define CNTR_32BIT_MAX 0x00000000FFFFFFFF
+
+static u64 read_dev_port_cntr(struct hfi1_devdata *dd, struct cntr_entry *entry,
+ u64 *psval, void *context, int vl)
+{
+ u64 val;
+ u64 sval = *psval;
+
+ if (entry->flags & CNTR_DISABLED) {
+ dd_dev_err(dd, "Counter %s not enabled", entry->name);
+ return 0;
+ }
+
+ hfi1_cdbg(CNTR, "cntr: %s vl %d psval 0x%llx", entry->name, vl, *psval);
+
+ val = entry->rw_cntr(entry, context, vl, CNTR_MODE_R, 0);
+
+ /* If its a synthetic counter there is more work we need to do */
+ if (entry->flags & CNTR_SYNTH) {
+ if (sval == CNTR_MAX) {
+ /* No need to read already saturated */
+ return CNTR_MAX;
+ }
+
+ if (entry->flags & CNTR_32BIT) {
+ /* 32bit counters can wrap multiple times */
+ u64 upper = sval >> 32;
+ u64 lower = (sval << 32) >> 32;
+
+ if (lower > val) { /* hw wrapped */
+ if (upper == CNTR_32BIT_MAX)
+ val = CNTR_MAX;
+ else
+ upper++;
+ }
+
+ if (val != CNTR_MAX)
+ val = (upper << 32) | val;
+
+ } else {
+ /* If we rolled we are saturated */
+ if ((val < sval) || (val > CNTR_MAX))
+ val = CNTR_MAX;
+ }
+ }
+
+ *psval = val;
+
+ hfi1_cdbg(CNTR, "\tNew val=0x%llx", val);
+
+ return val;
+}
+
+static u64 write_dev_port_cntr(struct hfi1_devdata *dd,
+ struct cntr_entry *entry,
+ u64 *psval, void *context, int vl, u64 data)
+{
+ u64 val;
+
+ if (entry->flags & CNTR_DISABLED) {
+ dd_dev_err(dd, "Counter %s not enabled", entry->name);
+ return 0;
+ }
+
+ hfi1_cdbg(CNTR, "cntr: %s vl %d psval 0x%llx", entry->name, vl, *psval);
+
+ if (entry->flags & CNTR_SYNTH) {
+ *psval = data;
+ if (entry->flags & CNTR_32BIT) {
+ val = entry->rw_cntr(entry, context, vl, CNTR_MODE_W,
+ (data << 32) >> 32);
+ val = data; /* return the full 64bit value */
+ } else {
+ val = entry->rw_cntr(entry, context, vl, CNTR_MODE_W,
+ data);
+ }
+ } else {
+ val = entry->rw_cntr(entry, context, vl, CNTR_MODE_W, data);
+ }
+
+ *psval = val;
+
+ hfi1_cdbg(CNTR, "\tNew val=0x%llx", val);
+
+ return val;
+}
+
+u64 read_dev_cntr(struct hfi1_devdata *dd, int index, int vl)
+{
+ struct cntr_entry *entry;
+ u64 *sval;
+
+ entry = &dev_cntrs[index];
+ sval = dd->scntrs + entry->offset;
+
+ if (vl != CNTR_INVALID_VL)
+ sval += vl;
+
+ return read_dev_port_cntr(dd, entry, sval, dd, vl);
+}
+
+u64 write_dev_cntr(struct hfi1_devdata *dd, int index, int vl, u64 data)
+{
+ struct cntr_entry *entry;
+ u64 *sval;
+
+ entry = &dev_cntrs[index];
+ sval = dd->scntrs + entry->offset;
+
+ if (vl != CNTR_INVALID_VL)
+ sval += vl;
+
+ return write_dev_port_cntr(dd, entry, sval, dd, vl, data);
+}
+
+u64 read_port_cntr(struct hfi1_pportdata *ppd, int index, int vl)
+{
+ struct cntr_entry *entry;
+ u64 *sval;
+
+ entry = &port_cntrs[index];
+ sval = ppd->scntrs + entry->offset;
+
+ if (vl != CNTR_INVALID_VL)
+ sval += vl;
+
+ if ((index >= C_RCV_HDR_OVF_FIRST + ppd->dd->num_rcv_contexts) &&
+ (index <= C_RCV_HDR_OVF_LAST)) {
+ /* We do not want to bother for disabled contexts */
+ return 0;
+ }
+
+ return read_dev_port_cntr(ppd->dd, entry, sval, ppd, vl);
+}
+
+u64 write_port_cntr(struct hfi1_pportdata *ppd, int index, int vl, u64 data)
+{
+ struct cntr_entry *entry;
+ u64 *sval;
+
+ entry = &port_cntrs[index];
+ sval = ppd->scntrs + entry->offset;
+
+ if (vl != CNTR_INVALID_VL)
+ sval += vl;
+
+ if ((index >= C_RCV_HDR_OVF_FIRST + ppd->dd->num_rcv_contexts) &&
+ (index <= C_RCV_HDR_OVF_LAST)) {
+ /* We do not want to bother for disabled contexts */
+ return 0;
+ }
+
+ return write_dev_port_cntr(ppd->dd, entry, sval, ppd, vl, data);
+}
+
+static void update_synth_timer(unsigned long opaque)
+{
+ u64 cur_tx;
+ u64 cur_rx;
+ u64 total_flits;
+ u8 update = 0;
+ int i, j, vl;
+ struct hfi1_pportdata *ppd;
+ struct cntr_entry *entry;
+
+ struct hfi1_devdata *dd = (struct hfi1_devdata *)opaque;
+
+ /*
+ * Rather than keep beating on the CSRs pick a minimal set that we can
+ * check to watch for potential roll over. We can do this by looking at
+ * the number of flits sent/recv. If the total flits exceeds 32bits then
+ * we have to iterate all the counters and update.
+ */
+ entry = &dev_cntrs[C_DC_RCV_FLITS];
+ cur_rx = entry->rw_cntr(entry, dd, CNTR_INVALID_VL, CNTR_MODE_R, 0);
+
+ entry = &dev_cntrs[C_DC_XMIT_FLITS];
+ cur_tx = entry->rw_cntr(entry, dd, CNTR_INVALID_VL, CNTR_MODE_R, 0);
+
+ hfi1_cdbg(
+ CNTR,
+ "[%d] curr tx=0x%llx rx=0x%llx :: last tx=0x%llx rx=0x%llx\n",
+ dd->unit, cur_tx, cur_rx, dd->last_tx, dd->last_rx);
+
+ if ((cur_tx < dd->last_tx) || (cur_rx < dd->last_rx)) {
+ /*
+ * May not be strictly necessary to update but it won't hurt and
+ * simplifies the logic here.
+ */
+ update = 1;
+ hfi1_cdbg(CNTR, "[%d] Tripwire counter rolled, updating",
+ dd->unit);
+ } else {
+ total_flits = (cur_tx - dd->last_tx) + (cur_rx - dd->last_rx);
+ hfi1_cdbg(CNTR,
+ "[%d] total flits 0x%llx limit 0x%llx\n", dd->unit,
+ total_flits, (u64)CNTR_32BIT_MAX);
+ if (total_flits >= CNTR_32BIT_MAX) {
+ hfi1_cdbg(CNTR, "[%d] 32bit limit hit, updating",
+ dd->unit);
+ update = 1;
+ }
+ }
+
+ if (update) {
+ hfi1_cdbg(CNTR, "[%d] Updating dd and ppd counters", dd->unit);
+ for (i = 0; i < DEV_CNTR_LAST; i++) {
+ entry = &dev_cntrs[i];
+ if (entry->flags & CNTR_VL) {
+ for (vl = 0; vl < C_VL_COUNT; vl++)
+ read_dev_cntr(dd, i, vl);
+ } else {
+ read_dev_cntr(dd, i, CNTR_INVALID_VL);
+ }
+ }
+ ppd = (struct hfi1_pportdata *)(dd + 1);
+ for (i = 0; i < dd->num_pports; i++, ppd++) {
+ for (j = 0; j < PORT_CNTR_LAST; j++) {
+ entry = &port_cntrs[j];
+ if (entry->flags & CNTR_VL) {
+ for (vl = 0; vl < C_VL_COUNT; vl++)
+ read_port_cntr(ppd, j, vl);
+ } else {
+ read_port_cntr(ppd, j, CNTR_INVALID_VL);
+ }
+ }
+ }
+
+ /*
+ * We want the value in the register. The goal is to keep track
+ * of the number of "ticks" not the counter value. In other
+ * words if the register rolls we want to notice it and go ahead
+ * and force an update.
+ */
+ entry = &dev_cntrs[C_DC_XMIT_FLITS];
+ dd->last_tx = entry->rw_cntr(entry, dd, CNTR_INVALID_VL,
+ CNTR_MODE_R, 0);
+
+ entry = &dev_cntrs[C_DC_RCV_FLITS];
+ dd->last_rx = entry->rw_cntr(entry, dd, CNTR_INVALID_VL,
+ CNTR_MODE_R, 0);
+
+ hfi1_cdbg(CNTR, "[%d] setting last tx/rx to 0x%llx 0x%llx",
+ dd->unit, dd->last_tx, dd->last_rx);
+
+ } else {
+ hfi1_cdbg(CNTR, "[%d] No update necessary", dd->unit);
+ }
+
+mod_timer(&dd->synth_stats_timer, jiffies + HZ * SYNTH_CNT_TIME);
+}
+
+#define C_MAX_NAME 13 /* 12 chars + one for /0 */
+static int init_cntrs(struct hfi1_devdata *dd)
+{
+ int i, rcv_ctxts, index, j;
+ size_t sz;
+ char *p;
+ char name[C_MAX_NAME];
+ struct hfi1_pportdata *ppd;
+
+ /* set up the stats timer; the add_timer is done at the end */
+ init_timer(&dd->synth_stats_timer);
+ dd->synth_stats_timer.function = update_synth_timer;
+ dd->synth_stats_timer.data = (unsigned long) dd;
+
+ /***********************/
+ /* per device counters */
+ /***********************/
+
+ /* size names and determine how many we have*/
+ dd->ndevcntrs = 0;
+ sz = 0;
+ index = 0;
+
+ for (i = 0; i < DEV_CNTR_LAST; i++) {
+ hfi1_dbg_early("Init cntr %s\n", dev_cntrs[i].name);
+ if (dev_cntrs[i].flags & CNTR_DISABLED) {
+ hfi1_dbg_early("\tSkipping %s\n", dev_cntrs[i].name);
+ continue;
+ }
+
+ if (dev_cntrs[i].flags & CNTR_VL) {
+ hfi1_dbg_early("\tProcessing VL cntr\n");
+ dev_cntrs[i].offset = index;
+ for (j = 0; j < C_VL_COUNT; j++) {
+ memset(name, '\0', C_MAX_NAME);
+ snprintf(name, C_MAX_NAME, "%s%d",
+ dev_cntrs[i].name,
+ vl_from_idx(j));
+ sz += strlen(name);
+ sz++;
+ hfi1_dbg_early("\t\t%s\n", name);
+ dd->ndevcntrs++;
+ index++;
+ }
+ } else {
+ /* +1 for newline */
+ sz += strlen(dev_cntrs[i].name) + 1;
+ dd->ndevcntrs++;
+ dev_cntrs[i].offset = index;
+ index++;
+ hfi1_dbg_early("\tAdding %s\n", dev_cntrs[i].name);
+ }
+ }
+
+ /* allocate space for the counter values */
+ dd->cntrs = kcalloc(index, sizeof(u64), GFP_KERNEL);
+ if (!dd->cntrs)
+ goto bail;
+
+ dd->scntrs = kcalloc(index, sizeof(u64), GFP_KERNEL);
+ if (!dd->scntrs)
+ goto bail;
+
+
+ /* allocate space for the counter names */
+ dd->cntrnameslen = sz;
+ dd->cntrnames = kmalloc(sz, GFP_KERNEL);
+ if (!dd->cntrnames)
+ goto bail;
+
+ /* fill in the names */
+ for (p = dd->cntrnames, i = 0, index = 0; i < DEV_CNTR_LAST; i++) {
+ if (dev_cntrs[i].flags & CNTR_DISABLED) {
+ /* Nothing */
+ } else {
+ if (dev_cntrs[i].flags & CNTR_VL) {
+ for (j = 0; j < C_VL_COUNT; j++) {
+ memset(name, '\0', C_MAX_NAME);
+ snprintf(name, C_MAX_NAME, "%s%d",
+ dev_cntrs[i].name,
+ vl_from_idx(j));
+ memcpy(p, name, strlen(name));
+ p += strlen(name);
+ *p++ = '\n';
+ }
+ } else {
+ memcpy(p, dev_cntrs[i].name,
+ strlen(dev_cntrs[i].name));
+ p += strlen(dev_cntrs[i].name);
+ *p++ = '\n';
+ }
+ index++;
+ }
+ }
+
+ /*********************/
+ /* per port counters */
+ /*********************/
+
+ /*
+ * Go through the counters for the overflows and disable the ones we
+ * don't need. This varies based on platform so we need to do it
+ * dynamically here.
+ */
+ rcv_ctxts = dd->num_rcv_contexts;
+ for (i = C_RCV_HDR_OVF_FIRST + rcv_ctxts;
+ i <= C_RCV_HDR_OVF_LAST; i++) {
+ port_cntrs[i].flags |= CNTR_DISABLED;
+ }
+
+ /* size port counter names and determine how many we have*/
+ sz = 0;
+ dd->nportcntrs = 0;
+ for (i = 0; i < PORT_CNTR_LAST; i++) {
+ hfi1_dbg_early("Init pcntr %s\n", port_cntrs[i].name);
+ if (port_cntrs[i].flags & CNTR_DISABLED) {
+ hfi1_dbg_early("\tSkipping %s\n", port_cntrs[i].name);
+ continue;
+ }
+
+ if (port_cntrs[i].flags & CNTR_VL) {
+ hfi1_dbg_early("\tProcessing VL cntr\n");
+ port_cntrs[i].offset = dd->nportcntrs;
+ for (j = 0; j < C_VL_COUNT; j++) {
+ memset(name, '\0', C_MAX_NAME);
+ snprintf(name, C_MAX_NAME, "%s%d",
+ port_cntrs[i].name,
+ vl_from_idx(j));
+ sz += strlen(name);
+ sz++;
+ hfi1_dbg_early("\t\t%s\n", name);
+ dd->nportcntrs++;
+ }
+ } else {
+ /* +1 for newline */
+ sz += strlen(port_cntrs[i].name) + 1;
+ port_cntrs[i].offset = dd->nportcntrs;
+ dd->nportcntrs++;
+ hfi1_dbg_early("\tAdding %s\n", port_cntrs[i].name);
+ }
+ }
+
+ /* allocate space for the counter names */
+ dd->portcntrnameslen = sz;
+ dd->portcntrnames = kmalloc(sz, GFP_KERNEL);
+ if (!dd->portcntrnames)
+ goto bail;
+
+ /* fill in port cntr names */
+ for (p = dd->portcntrnames, i = 0; i < PORT_CNTR_LAST; i++) {
+ if (port_cntrs[i].flags & CNTR_DISABLED)
+ continue;
+
+ if (port_cntrs[i].flags & CNTR_VL) {
+ for (j = 0; j < C_VL_COUNT; j++) {
+ memset(name, '\0', C_MAX_NAME);
+ snprintf(name, C_MAX_NAME, "%s%d",
+ port_cntrs[i].name,
+ vl_from_idx(j));
+ memcpy(p, name, strlen(name));
+ p += strlen(name);
+ *p++ = '\n';
+ }
+ } else {
+ memcpy(p, port_cntrs[i].name,
+ strlen(port_cntrs[i].name));
+ p += strlen(port_cntrs[i].name);
+ *p++ = '\n';
+ }
+ }
+
+ /* allocate per port storage for counter values */
+ ppd = (struct hfi1_pportdata *)(dd + 1);
+ for (i = 0; i < dd->num_pports; i++, ppd++) {
+ ppd->cntrs = kcalloc(dd->nportcntrs, sizeof(u64), GFP_KERNEL);
+ if (!ppd->cntrs)
+ goto bail;
+
+ ppd->scntrs = kcalloc(dd->nportcntrs, sizeof(u64), GFP_KERNEL);
+ if (!ppd->scntrs)
+ goto bail;
+ }
+
+ /* CPU counters need to be allocated and zeroed */
+ if (init_cpu_counters(dd))
+ goto bail;
+
+ mod_timer(&dd->synth_stats_timer, jiffies + HZ * SYNTH_CNT_TIME);
+ return 0;
+bail:
+ free_cntrs(dd);
+ return -ENOMEM;
+}
+
+
+static u32 chip_to_opa_lstate(struct hfi1_devdata *dd, u32 chip_lstate)
+{
+ switch (chip_lstate) {
+ default:
+ dd_dev_err(dd,
+ "Unknown logical state 0x%x, reporting IB_PORT_DOWN\n",
+ chip_lstate);
+ /* fall through */
+ case LSTATE_DOWN:
+ return IB_PORT_DOWN;
+ case LSTATE_INIT:
+ return IB_PORT_INIT;
+ case LSTATE_ARMED:
+ return IB_PORT_ARMED;
+ case LSTATE_ACTIVE:
+ return IB_PORT_ACTIVE;
+ }
+}
+
+u32 chip_to_opa_pstate(struct hfi1_devdata *dd, u32 chip_pstate)
+{
+ /* look at the HFI meta-states only */
+ switch (chip_pstate & 0xf0) {
+ default:
+ dd_dev_err(dd, "Unexpected chip physical state of 0x%x\n",
+ chip_pstate);
+ /* fall through */
+ case PLS_DISABLED:
+ return IB_PORTPHYSSTATE_DISABLED;
+ case PLS_OFFLINE:
+ return OPA_PORTPHYSSTATE_OFFLINE;
+ case PLS_POLLING:
+ return IB_PORTPHYSSTATE_POLLING;
+ case PLS_CONFIGPHY:
+ return IB_PORTPHYSSTATE_TRAINING;
+ case PLS_LINKUP:
+ return IB_PORTPHYSSTATE_LINKUP;
+ case PLS_PHYTEST:
+ return IB_PORTPHYSSTATE_PHY_TEST;
+ }
+}
+
+/* return the OPA port logical state name */
+const char *opa_lstate_name(u32 lstate)
+{
+ static const char * const port_logical_names[] = {
+ "PORT_NOP",
+ "PORT_DOWN",
+ "PORT_INIT",
+ "PORT_ARMED",
+ "PORT_ACTIVE",
+ "PORT_ACTIVE_DEFER",
+ };
+ if (lstate < ARRAY_SIZE(port_logical_names))
+ return port_logical_names[lstate];
+ return "unknown";
+}
+
+/* return the OPA port physical state name */
+const char *opa_pstate_name(u32 pstate)
+{
+ static const char * const port_physical_names[] = {
+ "PHYS_NOP",
+ "reserved1",
+ "PHYS_POLL",
+ "PHYS_DISABLED",
+ "PHYS_TRAINING",
+ "PHYS_LINKUP",
+ "PHYS_LINK_ERR_RECOVER",
+ "PHYS_PHY_TEST",
+ "reserved8",
+ "PHYS_OFFLINE",
+ "PHYS_GANGED",
+ "PHYS_TEST",
+ };
+ if (pstate < ARRAY_SIZE(port_physical_names))
+ return port_physical_names[pstate];
+ return "unknown";
+}
+
+/*
+ * Read the hardware link state and set the driver's cached value of it.
+ * Return the (new) current value.
+ */
+u32 get_logical_state(struct hfi1_pportdata *ppd)
+{
+ u32 new_state;
+
+ new_state = chip_to_opa_lstate(ppd->dd, read_logical_state(ppd->dd));
+ if (new_state != ppd->lstate) {
+ dd_dev_info(ppd->dd, "logical state changed to %s (0x%x)\n",
+ opa_lstate_name(new_state), new_state);
+ ppd->lstate = new_state;
+ }
+ /*
+ * Set port status flags in the page mapped into userspace
+ * memory. Do it here to ensure a reliable state - this is
+ * the only function called by all state handling code.
+ * Always set the flags due to the fact that the cache value
+ * might have been changed explicitly outside of this
+ * function.
+ */
+ if (ppd->statusp) {
+ switch (ppd->lstate) {
+ case IB_PORT_DOWN:
+ case IB_PORT_INIT:
+ *ppd->statusp &= ~(HFI1_STATUS_IB_CONF |
+ HFI1_STATUS_IB_READY);
+ break;
+ case IB_PORT_ARMED:
+ *ppd->statusp |= HFI1_STATUS_IB_CONF;
+ break;
+ case IB_PORT_ACTIVE:
+ *ppd->statusp |= HFI1_STATUS_IB_READY;
+ break;
+ }
+ }
+ return ppd->lstate;
+}
+
+/**
+ * wait_logical_linkstate - wait for an IB link state change to occur
+ * @ppd: port device
+ * @state: the state to wait for
+ * @msecs: the number of milliseconds to wait
+ *
+ * Wait up to msecs milliseconds for IB link state change to occur.
+ * For now, take the easy polling route.
+ * Returns 0 if state reached, otherwise -ETIMEDOUT.
+ */
+static int wait_logical_linkstate(struct hfi1_pportdata *ppd, u32 state,
+ int msecs)
+{
+ unsigned long timeout;
+
+ timeout = jiffies + msecs_to_jiffies(msecs);
+ while (1) {
+ if (get_logical_state(ppd) == state)
+ return 0;
+ if (time_after(jiffies, timeout))
+ break;
+ msleep(20);
+ }
+ dd_dev_err(ppd->dd, "timeout waiting for link state 0x%x\n", state);
+
+ return -ETIMEDOUT;
+}
+
+u8 hfi1_ibphys_portstate(struct hfi1_pportdata *ppd)
+{
+ static u32 remembered_state = 0xff;
+ u32 pstate;
+ u32 ib_pstate;
+
+ pstate = read_physical_state(ppd->dd);
+ ib_pstate = chip_to_opa_pstate(ppd->dd, pstate);
+ if (remembered_state != ib_pstate) {
+ dd_dev_info(ppd->dd,
+ "%s: physical state changed to %s (0x%x), phy 0x%x\n",
+ __func__, opa_pstate_name(ib_pstate), ib_pstate,
+ pstate);
+ remembered_state = ib_pstate;
+ }
+ return ib_pstate;
+}
+
+/*
+ * Read/modify/write ASIC_QSFP register bits as selected by mask
+ * data: 0 or 1 in the positions depending on what needs to be written
+ * dir: 0 for read, 1 for write
+ * mask: select by setting
+ * I2CCLK (bit 0)
+ * I2CDATA (bit 1)
+ */
+u64 hfi1_gpio_mod(struct hfi1_devdata *dd, u32 target, u32 data, u32 dir,
+ u32 mask)
+{
+ u64 qsfp_oe, target_oe;
+
+ target_oe = target ? ASIC_QSFP2_OE : ASIC_QSFP1_OE;
+ if (mask) {
+ /* We are writing register bits, so lock access */
+ dir &= mask;
+ data &= mask;
+
+ qsfp_oe = read_csr(dd, target_oe);
+ qsfp_oe = (qsfp_oe & ~(u64)mask) | (u64)dir;
+ write_csr(dd, target_oe, qsfp_oe);
+ }
+ /* We are exclusively reading bits here, but it is unlikely
+ * we'll get valid data when we set the direction of the pin
+ * in the same call, so read should call this function again
+ * to get valid data
+ */
+ return read_csr(dd, target ? ASIC_QSFP2_IN : ASIC_QSFP1_IN);
+}
+
+#define CLEAR_STATIC_RATE_CONTROL_SMASK(r) \
+(r &= ~SEND_CTXT_CHECK_ENABLE_DISALLOW_PBC_STATIC_RATE_CONTROL_SMASK)
+
+#define SET_STATIC_RATE_CONTROL_SMASK(r) \
+(r |= SEND_CTXT_CHECK_ENABLE_DISALLOW_PBC_STATIC_RATE_CONTROL_SMASK)
+
+int hfi1_init_ctxt(struct send_context *sc)
+{
+ if (sc != NULL) {
+ struct hfi1_devdata *dd = sc->dd;
+ u64 reg;
+ u8 set = (sc->type == SC_USER ?
+ HFI1_CAP_IS_USET(STATIC_RATE_CTRL) :
+ HFI1_CAP_IS_KSET(STATIC_RATE_CTRL));
+ reg = read_kctxt_csr(dd, sc->hw_context,
+ SEND_CTXT_CHECK_ENABLE);
+ if (set)
+ CLEAR_STATIC_RATE_CONTROL_SMASK(reg);
+ else
+ SET_STATIC_RATE_CONTROL_SMASK(reg);
+ write_kctxt_csr(dd, sc->hw_context,
+ SEND_CTXT_CHECK_ENABLE, reg);
+ }
+ return 0;
+}
+
+int hfi1_tempsense_rd(struct hfi1_devdata *dd, struct hfi1_temp *temp)
+{
+ int ret = 0;
+ u64 reg;
+
+ if (dd->icode != ICODE_RTL_SILICON) {
+ if (HFI1_CAP_IS_KSET(PRINT_UNIMPL))
+ dd_dev_info(dd, "%s: tempsense not supported by HW\n",
+ __func__);
+ return -EINVAL;
+ }
+ reg = read_csr(dd, ASIC_STS_THERM);
+ temp->curr = ((reg >> ASIC_STS_THERM_CURR_TEMP_SHIFT) &
+ ASIC_STS_THERM_CURR_TEMP_MASK);
+ temp->lo_lim = ((reg >> ASIC_STS_THERM_LO_TEMP_SHIFT) &
+ ASIC_STS_THERM_LO_TEMP_MASK);
+ temp->hi_lim = ((reg >> ASIC_STS_THERM_HI_TEMP_SHIFT) &
+ ASIC_STS_THERM_HI_TEMP_MASK);
+ temp->crit_lim = ((reg >> ASIC_STS_THERM_CRIT_TEMP_SHIFT) &
+ ASIC_STS_THERM_CRIT_TEMP_MASK);
+ /* triggers is a 3-bit value - 1 bit per trigger. */
+ temp->triggers = (u8)((reg >> ASIC_STS_THERM_LOW_SHIFT) & 0x7);
+
+ return ret;
+}
+
+/* ========================================================================= */
+
+/*
+ * Enable/disable chip from delivering interrupts.
+ */
+void set_intr_state(struct hfi1_devdata *dd, u32 enable)
+{
+ int i;
+
+ /*
+ * In HFI, the mask needs to be 1 to allow interrupts.
+ */
+ if (enable) {
+ u64 cce_int_mask;
+ const int qsfp1_int_smask = QSFP1_INT % 64;
+ const int qsfp2_int_smask = QSFP2_INT % 64;
+
+ /* enable all interrupts */
+ for (i = 0; i < CCE_NUM_INT_CSRS; i++)
+ write_csr(dd, CCE_INT_MASK + (8*i), ~(u64)0);
+
+ /*
+ * disable QSFP1 interrupts for HFI1, QSFP2 interrupts for HFI0
+ * Qsfp1Int and Qsfp2Int are adjacent bits in the same CSR,
+ * therefore just one of QSFP1_INT/QSFP2_INT can be used to find
+ * the index of the appropriate CSR in the CCEIntMask CSR array
+ */
+ cce_int_mask = read_csr(dd, CCE_INT_MASK +
+ (8*(QSFP1_INT/64)));
+ if (dd->hfi1_id) {
+ cce_int_mask &= ~((u64)1 << qsfp1_int_smask);
+ write_csr(dd, CCE_INT_MASK + (8*(QSFP1_INT/64)),
+ cce_int_mask);
+ } else {
+ cce_int_mask &= ~((u64)1 << qsfp2_int_smask);
+ write_csr(dd, CCE_INT_MASK + (8*(QSFP2_INT/64)),
+ cce_int_mask);
+ }
+ } else {
+ for (i = 0; i < CCE_NUM_INT_CSRS; i++)
+ write_csr(dd, CCE_INT_MASK + (8*i), 0ull);
+ }
+}
+
+/*
+ * Clear all interrupt sources on the chip.
+ */
+static void clear_all_interrupts(struct hfi1_devdata *dd)
+{
+ int i;
+
+ for (i = 0; i < CCE_NUM_INT_CSRS; i++)
+ write_csr(dd, CCE_INT_CLEAR + (8*i), ~(u64)0);
+
+ write_csr(dd, CCE_ERR_CLEAR, ~(u64)0);
+ write_csr(dd, MISC_ERR_CLEAR, ~(u64)0);
+ write_csr(dd, RCV_ERR_CLEAR, ~(u64)0);
+ write_csr(dd, SEND_ERR_CLEAR, ~(u64)0);
+ write_csr(dd, SEND_PIO_ERR_CLEAR, ~(u64)0);
+ write_csr(dd, SEND_DMA_ERR_CLEAR, ~(u64)0);
+ write_csr(dd, SEND_EGRESS_ERR_CLEAR, ~(u64)0);
+ for (i = 0; i < dd->chip_send_contexts; i++)
+ write_kctxt_csr(dd, i, SEND_CTXT_ERR_CLEAR, ~(u64)0);
+ for (i = 0; i < dd->chip_sdma_engines; i++)
+ write_kctxt_csr(dd, i, SEND_DMA_ENG_ERR_CLEAR, ~(u64)0);
+
+ write_csr(dd, DCC_ERR_FLG_CLR, ~(u64)0);
+ write_csr(dd, DC_LCB_ERR_CLR, ~(u64)0);
+ write_csr(dd, DC_DC8051_ERR_CLR, ~(u64)0);
+}
+
+/* Move to pcie.c? */
+static void disable_intx(struct pci_dev *pdev)
+{
+ pci_intx(pdev, 0);
+}
+
+static void clean_up_interrupts(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /* remove irqs - must happen before disabling/turning off */
+ if (dd->num_msix_entries) {
+ /* MSI-X */
+ struct hfi1_msix_entry *me = dd->msix_entries;
+
+ for (i = 0; i < dd->num_msix_entries; i++, me++) {
+ if (me->arg == NULL) /* => no irq, no affinity */
+ break;
+ irq_set_affinity_hint(dd->msix_entries[i].msix.vector,
+ NULL);
+ free_irq(me->msix.vector, me->arg);
+ }
+ } else {
+ /* INTx */
+ if (dd->requested_intx_irq) {
+ free_irq(dd->pcidev->irq, dd);
+ dd->requested_intx_irq = 0;
+ }
+ }
+
+ /* turn off interrupts */
+ if (dd->num_msix_entries) {
+ /* MSI-X */
+ hfi1_nomsix(dd);
+ } else {
+ /* INTx */
+ disable_intx(dd->pcidev);
+ }
+
+ /* clean structures */
+ for (i = 0; i < dd->num_msix_entries; i++)
+ free_cpumask_var(dd->msix_entries[i].mask);
+ kfree(dd->msix_entries);
+ dd->msix_entries = NULL;
+ dd->num_msix_entries = 0;
+}
+
+/*
+ * Remap the interrupt source from the general handler to the given MSI-X
+ * interrupt.
+ */
+static void remap_intr(struct hfi1_devdata *dd, int isrc, int msix_intr)
+{
+ u64 reg;
+ int m, n;
+
+ /* clear from the handled mask of the general interrupt */
+ m = isrc / 64;
+ n = isrc % 64;
+ dd->gi_mask[m] &= ~((u64)1 << n);
+
+ /* direct the chip source to the given MSI-X interrupt */
+ m = isrc / 8;
+ n = isrc % 8;
+ reg = read_csr(dd, CCE_INT_MAP + (8*m));
+ reg &= ~((u64)0xff << (8*n));
+ reg |= ((u64)msix_intr & 0xff) << (8*n);
+ write_csr(dd, CCE_INT_MAP + (8*m), reg);
+}
+
+static void remap_sdma_interrupts(struct hfi1_devdata *dd,
+ int engine, int msix_intr)
+{
+ /*
+ * SDMA engine interrupt sources grouped by type, rather than
+ * engine. Per-engine interrupts are as follows:
+ * SDMA
+ * SDMAProgress
+ * SDMAIdle
+ */
+ remap_intr(dd, IS_SDMA_START + 0*TXE_NUM_SDMA_ENGINES + engine,
+ msix_intr);
+ remap_intr(dd, IS_SDMA_START + 1*TXE_NUM_SDMA_ENGINES + engine,
+ msix_intr);
+ remap_intr(dd, IS_SDMA_START + 2*TXE_NUM_SDMA_ENGINES + engine,
+ msix_intr);
+}
+
+static void remap_receive_available_interrupt(struct hfi1_devdata *dd,
+ int rx, int msix_intr)
+{
+ remap_intr(dd, IS_RCVAVAIL_START + rx, msix_intr);
+}
+
+static int request_intx_irq(struct hfi1_devdata *dd)
+{
+ int ret;
+
+ snprintf(dd->intx_name, sizeof(dd->intx_name), DRIVER_NAME"_%d",
+ dd->unit);
+ ret = request_irq(dd->pcidev->irq, general_interrupt,
+ IRQF_SHARED, dd->intx_name, dd);
+ if (ret)
+ dd_dev_err(dd, "unable to request INTx interrupt, err %d\n",
+ ret);
+ else
+ dd->requested_intx_irq = 1;
+ return ret;
+}
+
+static int request_msix_irqs(struct hfi1_devdata *dd)
+{
+ const struct cpumask *local_mask;
+ cpumask_var_t def, rcv;
+ bool def_ret, rcv_ret;
+ int first_general, last_general;
+ int first_sdma, last_sdma;
+ int first_rx, last_rx;
+ int first_cpu, restart_cpu, curr_cpu;
+ int rcv_cpu, sdma_cpu;
+ int i, ret = 0, possible;
+ int ht;
+
+ /* calculate the ranges we are going to use */
+ first_general = 0;
+ first_sdma = last_general = first_general + 1;
+ first_rx = last_sdma = first_sdma + dd->num_sdma;
+ last_rx = first_rx + dd->n_krcv_queues;
+
+ /*
+ * Interrupt affinity.
+ *
+ * non-rcv avail gets a default mask that
+ * starts as possible cpus with threads reset
+ * and each rcv avail reset.
+ *
+ * rcv avail gets node relative 1 wrapping back
+ * to the node relative 1 as necessary.
+ *
+ */
+ local_mask = cpumask_of_pcibus(dd->pcidev->bus);
+ /* if first cpu is invalid, use NUMA 0 */
+ if (cpumask_first(local_mask) >= nr_cpu_ids)
+ local_mask = topology_core_cpumask(0);
+
+ def_ret = zalloc_cpumask_var(&def, GFP_KERNEL);
+ rcv_ret = zalloc_cpumask_var(&rcv, GFP_KERNEL);
+ if (!def_ret || !rcv_ret)
+ goto bail;
+ /* use local mask as default */
+ cpumask_copy(def, local_mask);
+ possible = cpumask_weight(def);
+ /* disarm threads from default */
+ ht = cpumask_weight(
+ topology_sibling_cpumask(cpumask_first(local_mask)));
+ for (i = possible/ht; i < possible; i++)
+ cpumask_clear_cpu(i, def);
+ /* reset possible */
+ possible = cpumask_weight(def);
+ /* def now has full cores on chosen node*/
+ first_cpu = cpumask_first(def);
+ if (nr_cpu_ids >= first_cpu)
+ first_cpu++;
+ restart_cpu = first_cpu;
+ curr_cpu = restart_cpu;
+
+ for (i = first_cpu; i < dd->n_krcv_queues + first_cpu; i++) {
+ cpumask_clear_cpu(curr_cpu, def);
+ cpumask_set_cpu(curr_cpu, rcv);
+ if (curr_cpu >= possible)
+ curr_cpu = restart_cpu;
+ else
+ curr_cpu++;
+ }
+ /* def mask has non-rcv, rcv has recv mask */
+ rcv_cpu = cpumask_first(rcv);
+ sdma_cpu = cpumask_first(def);
+
+ /*
+ * Sanity check - the code expects all SDMA chip source
+ * interrupts to be in the same CSR, starting at bit 0. Verify
+ * that this is true by checking the bit location of the start.
+ */
+ BUILD_BUG_ON(IS_SDMA_START % 64);
+
+ for (i = 0; i < dd->num_msix_entries; i++) {
+ struct hfi1_msix_entry *me = &dd->msix_entries[i];
+ const char *err_info;
+ irq_handler_t handler;
+ void *arg;
+ int idx;
+ struct hfi1_ctxtdata *rcd = NULL;
+ struct sdma_engine *sde = NULL;
+
+ /* obtain the arguments to request_irq */
+ if (first_general <= i && i < last_general) {
+ idx = i - first_general;
+ handler = general_interrupt;
+ arg = dd;
+ snprintf(me->name, sizeof(me->name),
+ DRIVER_NAME"_%d", dd->unit);
+ err_info = "general";
+ } else if (first_sdma <= i && i < last_sdma) {
+ idx = i - first_sdma;
+ sde = &dd->per_sdma[idx];
+ handler = sdma_interrupt;
+ arg = sde;
+ snprintf(me->name, sizeof(me->name),
+ DRIVER_NAME"_%d sdma%d", dd->unit, idx);
+ err_info = "sdma";
+ remap_sdma_interrupts(dd, idx, i);
+ } else if (first_rx <= i && i < last_rx) {
+ idx = i - first_rx;
+ rcd = dd->rcd[idx];
+ /* no interrupt if no rcd */
+ if (!rcd)
+ continue;
+ /*
+ * Set the interrupt register and mask for this
+ * context's interrupt.
+ */
+ rcd->ireg = (IS_RCVAVAIL_START+idx) / 64;
+ rcd->imask = ((u64)1) <<
+ ((IS_RCVAVAIL_START+idx) % 64);
+ handler = receive_context_interrupt;
+ arg = rcd;
+ snprintf(me->name, sizeof(me->name),
+ DRIVER_NAME"_%d kctxt%d", dd->unit, idx);
+ err_info = "receive context";
+ remap_receive_available_interrupt(dd, idx, i);
+ } else {
+ /* not in our expected range - complain, then
+ ignore it */
+ dd_dev_err(dd,
+ "Unexpected extra MSI-X interrupt %d\n", i);
+ continue;
+ }
+ /* no argument, no interrupt */
+ if (arg == NULL)
+ continue;
+ /* make sure the name is terminated */
+ me->name[sizeof(me->name)-1] = 0;
+
+ ret = request_irq(me->msix.vector, handler, 0, me->name, arg);
+ if (ret) {
+ dd_dev_err(dd,
+ "unable to allocate %s interrupt, vector %d, index %d, err %d\n",
+ err_info, me->msix.vector, idx, ret);
+ return ret;
+ }
+ /*
+ * assign arg after request_irq call, so it will be
+ * cleaned up
+ */
+ me->arg = arg;
+
+ if (!zalloc_cpumask_var(
+ &dd->msix_entries[i].mask,
+ GFP_KERNEL))
+ goto bail;
+ if (handler == sdma_interrupt) {
+ dd_dev_info(dd, "sdma engine %d cpu %d\n",
+ sde->this_idx, sdma_cpu);
+ cpumask_set_cpu(sdma_cpu, dd->msix_entries[i].mask);
+ sdma_cpu = cpumask_next(sdma_cpu, def);
+ if (sdma_cpu >= nr_cpu_ids)
+ sdma_cpu = cpumask_first(def);
+ } else if (handler == receive_context_interrupt) {
+ dd_dev_info(dd, "rcv ctxt %d cpu %d\n",
+ rcd->ctxt, rcv_cpu);
+ cpumask_set_cpu(rcv_cpu, dd->msix_entries[i].mask);
+ rcv_cpu = cpumask_next(rcv_cpu, rcv);
+ if (rcv_cpu >= nr_cpu_ids)
+ rcv_cpu = cpumask_first(rcv);
+ } else {
+ /* otherwise first def */
+ dd_dev_info(dd, "%s cpu %d\n",
+ err_info, cpumask_first(def));
+ cpumask_set_cpu(
+ cpumask_first(def), dd->msix_entries[i].mask);
+ }
+ irq_set_affinity_hint(
+ dd->msix_entries[i].msix.vector,
+ dd->msix_entries[i].mask);
+ }
+
+out:
+ free_cpumask_var(def);
+ free_cpumask_var(rcv);
+ return ret;
+bail:
+ ret = -ENOMEM;
+ goto out;
+}
+
+/*
+ * Set the general handler to accept all interrupts, remap all
+ * chip interrupts back to MSI-X 0.
+ */
+static void reset_interrupts(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /* all interrupts handled by the general handler */
+ for (i = 0; i < CCE_NUM_INT_CSRS; i++)
+ dd->gi_mask[i] = ~(u64)0;
+
+ /* all chip interrupts map to MSI-X 0 */
+ for (i = 0; i < CCE_NUM_INT_MAP_CSRS; i++)
+ write_csr(dd, CCE_INT_MAP + (8*i), 0);
+}
+
+static int set_up_interrupts(struct hfi1_devdata *dd)
+{
+ struct hfi1_msix_entry *entries;
+ u32 total, request;
+ int i, ret;
+ int single_interrupt = 0; /* we expect to have all the interrupts */
+
+ /*
+ * Interrupt count:
+ * 1 general, "slow path" interrupt (includes the SDMA engines
+ * slow source, SDMACleanupDone)
+ * N interrupts - one per used SDMA engine
+ * M interrupt - one per kernel receive context
+ */
+ total = 1 + dd->num_sdma + dd->n_krcv_queues;
+
+ entries = kcalloc(total, sizeof(*entries), GFP_KERNEL);
+ if (!entries) {
+ dd_dev_err(dd, "cannot allocate msix table\n");
+ ret = -ENOMEM;
+ goto fail;
+ }
+ /* 1-1 MSI-X entry assignment */
+ for (i = 0; i < total; i++)
+ entries[i].msix.entry = i;
+
+ /* ask for MSI-X interrupts */
+ request = total;
+ request_msix(dd, &request, entries);
+
+ if (request == 0) {
+ /* using INTx */
+ /* dd->num_msix_entries already zero */
+ kfree(entries);
+ single_interrupt = 1;
+ dd_dev_err(dd, "MSI-X failed, using INTx interrupts\n");
+ } else {
+ /* using MSI-X */
+ dd->num_msix_entries = request;
+ dd->msix_entries = entries;
+
+ if (request != total) {
+ /* using MSI-X, with reduced interrupts */
+ dd_dev_err(
+ dd,
+ "cannot handle reduced interrupt case, want %u, got %u\n",
+ total, request);
+ ret = -EINVAL;
+ goto fail;
+ }
+ dd_dev_info(dd, "%u MSI-X interrupts allocated\n", total);
+ }
+
+ /* mask all interrupts */
+ set_intr_state(dd, 0);
+ /* clear all pending interrupts */
+ clear_all_interrupts(dd);
+
+ /* reset general handler mask, chip MSI-X mappings */
+ reset_interrupts(dd);
+
+ if (single_interrupt)
+ ret = request_intx_irq(dd);
+ else
+ ret = request_msix_irqs(dd);
+ if (ret)
+ goto fail;
+
+ return 0;
+
+fail:
+ clean_up_interrupts(dd);
+ return ret;
+}
+
+/*
+ * Set up context values in dd. Sets:
+ *
+ * num_rcv_contexts - number of contexts being used
+ * n_krcv_queues - number of kernel contexts
+ * first_user_ctxt - first non-kernel context in array of contexts
+ * freectxts - number of free user contexts
+ * num_send_contexts - number of PIO send contexts being used
+ */
+static int set_up_context_variables(struct hfi1_devdata *dd)
+{
+ int num_kernel_contexts;
+ int num_user_contexts;
+ int total_contexts;
+ int ret;
+ unsigned ngroups;
+
+ /*
+ * Kernel contexts: (to be fixed later):
+ * - min or 2 or 1 context/numa
+ * - Context 0 - default/errors
+ * - Context 1 - VL15
+ */
+ if (n_krcvqs)
+ num_kernel_contexts = n_krcvqs + MIN_KERNEL_KCTXTS;
+ else
+ num_kernel_contexts = num_online_nodes();
+ num_kernel_contexts =
+ max_t(int, MIN_KERNEL_KCTXTS, num_kernel_contexts);
+ /*
+ * Every kernel receive context needs an ACK send context.
+ * one send context is allocated for each VL{0-7} and VL15
+ */
+ if (num_kernel_contexts > (dd->chip_send_contexts - num_vls - 1)) {
+ dd_dev_err(dd,
+ "Reducing # kernel rcv contexts to: %d, from %d\n",
+ (int)(dd->chip_send_contexts - num_vls - 1),
+ (int)num_kernel_contexts);
+ num_kernel_contexts = dd->chip_send_contexts - num_vls - 1;
+ }
+ /*
+ * User contexts: (to be fixed later)
+ * - set to num_rcv_contexts if non-zero
+ * - default to 1 user context per CPU
+ */
+ if (num_rcv_contexts)
+ num_user_contexts = num_rcv_contexts;
+ else
+ num_user_contexts = num_online_cpus();
+
+ total_contexts = num_kernel_contexts + num_user_contexts;
+
+ /*
+ * Adjust the counts given a global max.
+ */
+ if (total_contexts > dd->chip_rcv_contexts) {
+ dd_dev_err(dd,
+ "Reducing # user receive contexts to: %d, from %d\n",
+ (int)(dd->chip_rcv_contexts - num_kernel_contexts),
+ (int)num_user_contexts);
+ num_user_contexts = dd->chip_rcv_contexts - num_kernel_contexts;
+ /* recalculate */
+ total_contexts = num_kernel_contexts + num_user_contexts;
+ }
+
+ /* the first N are kernel contexts, the rest are user contexts */
+ dd->num_rcv_contexts = total_contexts;
+ dd->n_krcv_queues = num_kernel_contexts;
+ dd->first_user_ctxt = num_kernel_contexts;
+ dd->freectxts = num_user_contexts;
+ dd_dev_info(dd,
+ "rcv contexts: chip %d, used %d (kernel %d, user %d)\n",
+ (int)dd->chip_rcv_contexts,
+ (int)dd->num_rcv_contexts,
+ (int)dd->n_krcv_queues,
+ (int)dd->num_rcv_contexts - dd->n_krcv_queues);
+
+ /*
+ * Receive array allocation:
+ * All RcvArray entries are divided into groups of 8. This
+ * is required by the hardware and will speed up writes to
+ * consecutive entries by using write-combining of the entire
+ * cacheline.
+ *
+ * The number of groups are evenly divided among all contexts.
+ * any left over groups will be given to the first N user
+ * contexts.
+ */
+ dd->rcv_entries.group_size = RCV_INCREMENT;
+ ngroups = dd->chip_rcv_array_count / dd->rcv_entries.group_size;
+ dd->rcv_entries.ngroups = ngroups / dd->num_rcv_contexts;
+ dd->rcv_entries.nctxt_extra = ngroups -
+ (dd->num_rcv_contexts * dd->rcv_entries.ngroups);
+ dd_dev_info(dd, "RcvArray groups %u, ctxts extra %u\n",
+ dd->rcv_entries.ngroups,
+ dd->rcv_entries.nctxt_extra);
+ if (dd->rcv_entries.ngroups * dd->rcv_entries.group_size >
+ MAX_EAGER_ENTRIES * 2) {
+ dd->rcv_entries.ngroups = (MAX_EAGER_ENTRIES * 2) /
+ dd->rcv_entries.group_size;
+ dd_dev_info(dd,
+ "RcvArray group count too high, change to %u\n",
+ dd->rcv_entries.ngroups);
+ dd->rcv_entries.nctxt_extra = 0;
+ }
+ /*
+ * PIO send contexts
+ */
+ ret = init_sc_pools_and_sizes(dd);
+ if (ret >= 0) { /* success */
+ dd->num_send_contexts = ret;
+ dd_dev_info(
+ dd,
+ "send contexts: chip %d, used %d (kernel %d, ack %d, user %d)\n",
+ dd->chip_send_contexts,
+ dd->num_send_contexts,
+ dd->sc_sizes[SC_KERNEL].count,
+ dd->sc_sizes[SC_ACK].count,
+ dd->sc_sizes[SC_USER].count);
+ ret = 0; /* success */
+ }
+
+ return ret;
+}
+
+/*
+ * Set the device/port partition key table. The MAD code
+ * will ensure that, at least, the partial management
+ * partition key is present in the table.
+ */
+static void set_partition_keys(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ u64 reg = 0;
+ int i;
+
+ dd_dev_info(dd, "Setting partition keys\n");
+ for (i = 0; i < hfi1_get_npkeys(dd); i++) {
+ reg |= (ppd->pkeys[i] &
+ RCV_PARTITION_KEY_PARTITION_KEY_A_MASK) <<
+ ((i % 4) *
+ RCV_PARTITION_KEY_PARTITION_KEY_B_SHIFT);
+ /* Each register holds 4 PKey values. */
+ if ((i % 4) == 3) {
+ write_csr(dd, RCV_PARTITION_KEY +
+ ((i - 3) * 2), reg);
+ reg = 0;
+ }
+ }
+
+ /* Always enable HW pkeys check when pkeys table is set */
+ add_rcvctrl(dd, RCV_CTRL_RCV_PARTITION_KEY_ENABLE_SMASK);
+}
+
+/*
+ * These CSRs and memories are uninitialized on reset and must be
+ * written before reading to set the ECC/parity bits.
+ *
+ * NOTE: All user context CSRs that are not mmaped write-only
+ * (e.g. the TID flows) must be initialized even if the driver never
+ * reads them.
+ */
+static void write_uninitialized_csrs_and_memories(struct hfi1_devdata *dd)
+{
+ int i, j;
+
+ /* CceIntMap */
+ for (i = 0; i < CCE_NUM_INT_MAP_CSRS; i++)
+ write_csr(dd, CCE_INT_MAP+(8*i), 0);
+
+ /* SendCtxtCreditReturnAddr */
+ for (i = 0; i < dd->chip_send_contexts; i++)
+ write_kctxt_csr(dd, i, SEND_CTXT_CREDIT_RETURN_ADDR, 0);
+
+ /* PIO Send buffers */
+ /* SDMA Send buffers */
+ /* These are not normally read, and (presently) have no method
+ to be read, so are not pre-initialized */
+
+ /* RcvHdrAddr */
+ /* RcvHdrTailAddr */
+ /* RcvTidFlowTable */
+ for (i = 0; i < dd->chip_rcv_contexts; i++) {
+ write_kctxt_csr(dd, i, RCV_HDR_ADDR, 0);
+ write_kctxt_csr(dd, i, RCV_HDR_TAIL_ADDR, 0);
+ for (j = 0; j < RXE_NUM_TID_FLOWS; j++)
+ write_uctxt_csr(dd, i, RCV_TID_FLOW_TABLE+(8*j), 0);
+ }
+
+ /* RcvArray */
+ for (i = 0; i < dd->chip_rcv_array_count; i++)
+ write_csr(dd, RCV_ARRAY + (8*i),
+ RCV_ARRAY_RT_WRITE_ENABLE_SMASK);
+
+ /* RcvQPMapTable */
+ for (i = 0; i < 32; i++)
+ write_csr(dd, RCV_QP_MAP_TABLE + (8 * i), 0);
+}
+
+/*
+ * Use the ctrl_bits in CceCtrl to clear the status_bits in CceStatus.
+ */
+static void clear_cce_status(struct hfi1_devdata *dd, u64 status_bits,
+ u64 ctrl_bits)
+{
+ unsigned long timeout;
+ u64 reg;
+
+ /* is the condition present? */
+ reg = read_csr(dd, CCE_STATUS);
+ if ((reg & status_bits) == 0)
+ return;
+
+ /* clear the condition */
+ write_csr(dd, CCE_CTRL, ctrl_bits);
+
+ /* wait for the condition to clear */
+ timeout = jiffies + msecs_to_jiffies(CCE_STATUS_TIMEOUT);
+ while (1) {
+ reg = read_csr(dd, CCE_STATUS);
+ if ((reg & status_bits) == 0)
+ return;
+ if (time_after(jiffies, timeout)) {
+ dd_dev_err(dd,
+ "Timeout waiting for CceStatus to clear bits 0x%llx, remaining 0x%llx\n",
+ status_bits, reg & status_bits);
+ return;
+ }
+ udelay(1);
+ }
+}
+
+/* set CCE CSRs to chip reset defaults */
+static void reset_cce_csrs(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /* CCE_REVISION read-only */
+ /* CCE_REVISION2 read-only */
+ /* CCE_CTRL - bits clear automatically */
+ /* CCE_STATUS read-only, use CceCtrl to clear */
+ clear_cce_status(dd, ALL_FROZE, CCE_CTRL_SPC_UNFREEZE_SMASK);
+ clear_cce_status(dd, ALL_TXE_PAUSE, CCE_CTRL_TXE_RESUME_SMASK);
+ clear_cce_status(dd, ALL_RXE_PAUSE, CCE_CTRL_RXE_RESUME_SMASK);
+ for (i = 0; i < CCE_NUM_SCRATCH; i++)
+ write_csr(dd, CCE_SCRATCH + (8 * i), 0);
+ /* CCE_ERR_STATUS read-only */
+ write_csr(dd, CCE_ERR_MASK, 0);
+ write_csr(dd, CCE_ERR_CLEAR, ~0ull);
+ /* CCE_ERR_FORCE leave alone */
+ for (i = 0; i < CCE_NUM_32_BIT_COUNTERS; i++)
+ write_csr(dd, CCE_COUNTER_ARRAY32 + (8 * i), 0);
+ write_csr(dd, CCE_DC_CTRL, CCE_DC_CTRL_RESETCSR);
+ /* CCE_PCIE_CTRL leave alone */
+ for (i = 0; i < CCE_NUM_MSIX_VECTORS; i++) {
+ write_csr(dd, CCE_MSIX_TABLE_LOWER + (8 * i), 0);
+ write_csr(dd, CCE_MSIX_TABLE_UPPER + (8 * i),
+ CCE_MSIX_TABLE_UPPER_RESETCSR);
+ }
+ for (i = 0; i < CCE_NUM_MSIX_PBAS; i++) {
+ /* CCE_MSIX_PBA read-only */
+ write_csr(dd, CCE_MSIX_INT_GRANTED, ~0ull);
+ write_csr(dd, CCE_MSIX_VEC_CLR_WITHOUT_INT, ~0ull);
+ }
+ for (i = 0; i < CCE_NUM_INT_MAP_CSRS; i++)
+ write_csr(dd, CCE_INT_MAP, 0);
+ for (i = 0; i < CCE_NUM_INT_CSRS; i++) {
+ /* CCE_INT_STATUS read-only */
+ write_csr(dd, CCE_INT_MASK + (8 * i), 0);
+ write_csr(dd, CCE_INT_CLEAR + (8 * i), ~0ull);
+ /* CCE_INT_FORCE leave alone */
+ /* CCE_INT_BLOCKED read-only */
+ }
+ for (i = 0; i < CCE_NUM_32_BIT_INT_COUNTERS; i++)
+ write_csr(dd, CCE_INT_COUNTER_ARRAY32 + (8 * i), 0);
+}
+
+/* set ASIC CSRs to chip reset defaults */
+static void reset_asic_csrs(struct hfi1_devdata *dd)
+{
+ static DEFINE_MUTEX(asic_mutex);
+ static int called;
+ int i;
+
+ /*
+ * If the HFIs are shared between separate nodes or VMs,
+ * then more will need to be done here. One idea is a module
+ * parameter that returns early, letting the first power-on or
+ * a known first load do the reset and blocking all others.
+ */
+
+ /*
+ * These CSRs should only be reset once - the first one here will
+ * do the work. Use a mutex so that a non-first caller waits until
+ * the first is finished before it can proceed.
+ */
+ mutex_lock(&asic_mutex);
+ if (called)
+ goto done;
+ called = 1;
+
+ if (dd->icode != ICODE_FPGA_EMULATION) {
+ /* emulation does not have an SBus - leave these alone */
+ /*
+ * All writes to ASIC_CFG_SBUS_REQUEST do something.
+ * Notes:
+ * o The reset is not zero if aimed at the core. See the
+ * SBus documentation for details.
+ * o If the SBus firmware has been updated (e.g. by the BIOS),
+ * will the reset revert that?
+ */
+ /* ASIC_CFG_SBUS_REQUEST leave alone */
+ write_csr(dd, ASIC_CFG_SBUS_EXECUTE, 0);
+ }
+ /* ASIC_SBUS_RESULT read-only */
+ write_csr(dd, ASIC_STS_SBUS_COUNTERS, 0);
+ for (i = 0; i < ASIC_NUM_SCRATCH; i++)
+ write_csr(dd, ASIC_CFG_SCRATCH + (8 * i), 0);
+ write_csr(dd, ASIC_CFG_MUTEX, 0); /* this will clear it */
+ write_csr(dd, ASIC_CFG_DRV_STR, 0);
+ write_csr(dd, ASIC_CFG_THERM_POLL_EN, 0);
+ /* ASIC_STS_THERM read-only */
+ /* ASIC_CFG_RESET leave alone */
+
+ write_csr(dd, ASIC_PCIE_SD_HOST_CMD, 0);
+ /* ASIC_PCIE_SD_HOST_STATUS read-only */
+ write_csr(dd, ASIC_PCIE_SD_INTRPT_DATA_CODE, 0);
+ write_csr(dd, ASIC_PCIE_SD_INTRPT_ENABLE, 0);
+ /* ASIC_PCIE_SD_INTRPT_PROGRESS read-only */
+ write_csr(dd, ASIC_PCIE_SD_INTRPT_STATUS, ~0ull); /* clear */
+ /* ASIC_HFI0_PCIE_SD_INTRPT_RSPD_DATA read-only */
+ /* ASIC_HFI1_PCIE_SD_INTRPT_RSPD_DATA read-only */
+ for (i = 0; i < 16; i++)
+ write_csr(dd, ASIC_PCIE_SD_INTRPT_LIST + (8 * i), 0);
+
+ /* ASIC_GPIO_IN read-only */
+ write_csr(dd, ASIC_GPIO_OE, 0);
+ write_csr(dd, ASIC_GPIO_INVERT, 0);
+ write_csr(dd, ASIC_GPIO_OUT, 0);
+ write_csr(dd, ASIC_GPIO_MASK, 0);
+ /* ASIC_GPIO_STATUS read-only */
+ write_csr(dd, ASIC_GPIO_CLEAR, ~0ull);
+ /* ASIC_GPIO_FORCE leave alone */
+
+ /* ASIC_QSFP1_IN read-only */
+ write_csr(dd, ASIC_QSFP1_OE, 0);
+ write_csr(dd, ASIC_QSFP1_INVERT, 0);
+ write_csr(dd, ASIC_QSFP1_OUT, 0);
+ write_csr(dd, ASIC_QSFP1_MASK, 0);
+ /* ASIC_QSFP1_STATUS read-only */
+ write_csr(dd, ASIC_QSFP1_CLEAR, ~0ull);
+ /* ASIC_QSFP1_FORCE leave alone */
+
+ /* ASIC_QSFP2_IN read-only */
+ write_csr(dd, ASIC_QSFP2_OE, 0);
+ write_csr(dd, ASIC_QSFP2_INVERT, 0);
+ write_csr(dd, ASIC_QSFP2_OUT, 0);
+ write_csr(dd, ASIC_QSFP2_MASK, 0);
+ /* ASIC_QSFP2_STATUS read-only */
+ write_csr(dd, ASIC_QSFP2_CLEAR, ~0ull);
+ /* ASIC_QSFP2_FORCE leave alone */
+
+ write_csr(dd, ASIC_EEP_CTL_STAT, ASIC_EEP_CTL_STAT_RESETCSR);
+ /* this also writes a NOP command, clearing paging mode */
+ write_csr(dd, ASIC_EEP_ADDR_CMD, 0);
+ write_csr(dd, ASIC_EEP_DATA, 0);
+
+done:
+ mutex_unlock(&asic_mutex);
+}
+
+/* set MISC CSRs to chip reset defaults */
+static void reset_misc_csrs(struct hfi1_devdata *dd)
+{
+ int i;
+
+ for (i = 0; i < 32; i++) {
+ write_csr(dd, MISC_CFG_RSA_R2 + (8 * i), 0);
+ write_csr(dd, MISC_CFG_RSA_SIGNATURE + (8 * i), 0);
+ write_csr(dd, MISC_CFG_RSA_MODULUS + (8 * i), 0);
+ }
+ /* MISC_CFG_SHA_PRELOAD leave alone - always reads 0 and can
+ only be written 128-byte chunks */
+ /* init RSA engine to clear lingering errors */
+ write_csr(dd, MISC_CFG_RSA_CMD, 1);
+ write_csr(dd, MISC_CFG_RSA_MU, 0);
+ write_csr(dd, MISC_CFG_FW_CTRL, 0);
+ /* MISC_STS_8051_DIGEST read-only */
+ /* MISC_STS_SBM_DIGEST read-only */
+ /* MISC_STS_PCIE_DIGEST read-only */
+ /* MISC_STS_FAB_DIGEST read-only */
+ /* MISC_ERR_STATUS read-only */
+ write_csr(dd, MISC_ERR_MASK, 0);
+ write_csr(dd, MISC_ERR_CLEAR, ~0ull);
+ /* MISC_ERR_FORCE leave alone */
+}
+
+/* set TXE CSRs to chip reset defaults */
+static void reset_txe_csrs(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /*
+ * TXE Kernel CSRs
+ */
+ write_csr(dd, SEND_CTRL, 0);
+ __cm_reset(dd, 0); /* reset CM internal state */
+ /* SEND_CONTEXTS read-only */
+ /* SEND_DMA_ENGINES read-only */
+ /* SEND_PIO_MEM_SIZE read-only */
+ /* SEND_DMA_MEM_SIZE read-only */
+ write_csr(dd, SEND_HIGH_PRIORITY_LIMIT, 0);
+ pio_reset_all(dd); /* SEND_PIO_INIT_CTXT */
+ /* SEND_PIO_ERR_STATUS read-only */
+ write_csr(dd, SEND_PIO_ERR_MASK, 0);
+ write_csr(dd, SEND_PIO_ERR_CLEAR, ~0ull);
+ /* SEND_PIO_ERR_FORCE leave alone */
+ /* SEND_DMA_ERR_STATUS read-only */
+ write_csr(dd, SEND_DMA_ERR_MASK, 0);
+ write_csr(dd, SEND_DMA_ERR_CLEAR, ~0ull);
+ /* SEND_DMA_ERR_FORCE leave alone */
+ /* SEND_EGRESS_ERR_STATUS read-only */
+ write_csr(dd, SEND_EGRESS_ERR_MASK, 0);
+ write_csr(dd, SEND_EGRESS_ERR_CLEAR, ~0ull);
+ /* SEND_EGRESS_ERR_FORCE leave alone */
+ write_csr(dd, SEND_BTH_QP, 0);
+ write_csr(dd, SEND_STATIC_RATE_CONTROL, 0);
+ write_csr(dd, SEND_SC2VLT0, 0);
+ write_csr(dd, SEND_SC2VLT1, 0);
+ write_csr(dd, SEND_SC2VLT2, 0);
+ write_csr(dd, SEND_SC2VLT3, 0);
+ write_csr(dd, SEND_LEN_CHECK0, 0);
+ write_csr(dd, SEND_LEN_CHECK1, 0);
+ /* SEND_ERR_STATUS read-only */
+ write_csr(dd, SEND_ERR_MASK, 0);
+ write_csr(dd, SEND_ERR_CLEAR, ~0ull);
+ /* SEND_ERR_FORCE read-only */
+ for (i = 0; i < VL_ARB_LOW_PRIO_TABLE_SIZE; i++)
+ write_csr(dd, SEND_LOW_PRIORITY_LIST + (8*i), 0);
+ for (i = 0; i < VL_ARB_HIGH_PRIO_TABLE_SIZE; i++)
+ write_csr(dd, SEND_HIGH_PRIORITY_LIST + (8*i), 0);
+ for (i = 0; i < dd->chip_send_contexts/NUM_CONTEXTS_PER_SET; i++)
+ write_csr(dd, SEND_CONTEXT_SET_CTRL + (8*i), 0);
+ for (i = 0; i < TXE_NUM_32_BIT_COUNTER; i++)
+ write_csr(dd, SEND_COUNTER_ARRAY32 + (8*i), 0);
+ for (i = 0; i < TXE_NUM_64_BIT_COUNTER; i++)
+ write_csr(dd, SEND_COUNTER_ARRAY64 + (8*i), 0);
+ write_csr(dd, SEND_CM_CTRL, SEND_CM_CTRL_RESETCSR);
+ write_csr(dd, SEND_CM_GLOBAL_CREDIT,
+ SEND_CM_GLOBAL_CREDIT_RESETCSR);
+ /* SEND_CM_CREDIT_USED_STATUS read-only */
+ write_csr(dd, SEND_CM_TIMER_CTRL, 0);
+ write_csr(dd, SEND_CM_LOCAL_AU_TABLE0_TO3, 0);
+ write_csr(dd, SEND_CM_LOCAL_AU_TABLE4_TO7, 0);
+ write_csr(dd, SEND_CM_REMOTE_AU_TABLE0_TO3, 0);
+ write_csr(dd, SEND_CM_REMOTE_AU_TABLE4_TO7, 0);
+ for (i = 0; i < TXE_NUM_DATA_VL; i++)
+ write_csr(dd, SEND_CM_CREDIT_VL + (8*i), 0);
+ write_csr(dd, SEND_CM_CREDIT_VL15, 0);
+ /* SEND_CM_CREDIT_USED_VL read-only */
+ /* SEND_CM_CREDIT_USED_VL15 read-only */
+ /* SEND_EGRESS_CTXT_STATUS read-only */
+ /* SEND_EGRESS_SEND_DMA_STATUS read-only */
+ write_csr(dd, SEND_EGRESS_ERR_INFO, ~0ull);
+ /* SEND_EGRESS_ERR_INFO read-only */
+ /* SEND_EGRESS_ERR_SOURCE read-only */
+
+ /*
+ * TXE Per-Context CSRs
+ */
+ for (i = 0; i < dd->chip_send_contexts; i++) {
+ write_kctxt_csr(dd, i, SEND_CTXT_CTRL, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_CREDIT_CTRL, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_CREDIT_RETURN_ADDR, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_CREDIT_FORCE, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_ERR_MASK, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_ERR_CLEAR, ~0ull);
+ write_kctxt_csr(dd, i, SEND_CTXT_CHECK_ENABLE, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_CHECK_VL, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_CHECK_JOB_KEY, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_CHECK_PARTITION_KEY, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_CHECK_SLID, 0);
+ write_kctxt_csr(dd, i, SEND_CTXT_CHECK_OPCODE, 0);
+ }
+
+ /*
+ * TXE Per-SDMA CSRs
+ */
+ for (i = 0; i < dd->chip_sdma_engines; i++) {
+ write_kctxt_csr(dd, i, SEND_DMA_CTRL, 0);
+ /* SEND_DMA_STATUS read-only */
+ write_kctxt_csr(dd, i, SEND_DMA_BASE_ADDR, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_LEN_GEN, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_TAIL, 0);
+ /* SEND_DMA_HEAD read-only */
+ write_kctxt_csr(dd, i, SEND_DMA_HEAD_ADDR, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_PRIORITY_THLD, 0);
+ /* SEND_DMA_IDLE_CNT read-only */
+ write_kctxt_csr(dd, i, SEND_DMA_RELOAD_CNT, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_DESC_CNT, 0);
+ /* SEND_DMA_DESC_FETCHED_CNT read-only */
+ /* SEND_DMA_ENG_ERR_STATUS read-only */
+ write_kctxt_csr(dd, i, SEND_DMA_ENG_ERR_MASK, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_ENG_ERR_CLEAR, ~0ull);
+ /* SEND_DMA_ENG_ERR_FORCE leave alone */
+ write_kctxt_csr(dd, i, SEND_DMA_CHECK_ENABLE, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_CHECK_VL, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_CHECK_JOB_KEY, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_CHECK_PARTITION_KEY, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_CHECK_SLID, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_CHECK_OPCODE, 0);
+ write_kctxt_csr(dd, i, SEND_DMA_MEMORY, 0);
+ }
+}
+
+/*
+ * Expect on entry:
+ * o Packet ingress is disabled, i.e. RcvCtrl.RcvPortEnable == 0
+ */
+static void init_rbufs(struct hfi1_devdata *dd)
+{
+ u64 reg;
+ int count;
+
+ /*
+ * Wait for DMA to stop: RxRbufPktPending and RxPktInProgress are
+ * clear.
+ */
+ count = 0;
+ while (1) {
+ reg = read_csr(dd, RCV_STATUS);
+ if ((reg & (RCV_STATUS_RX_RBUF_PKT_PENDING_SMASK
+ | RCV_STATUS_RX_PKT_IN_PROGRESS_SMASK)) == 0)
+ break;
+ /*
+ * Give up after 1ms - maximum wait time.
+ *
+ * RBuf size is 148KiB. Slowest possible is PCIe Gen1 x1 at
+ * 250MB/s bandwidth. Lower rate to 66% for overhead to get:
+ * 148 KB / (66% * 250MB/s) = 920us
+ */
+ if (count++ > 500) {
+ dd_dev_err(dd,
+ "%s: in-progress DMA not clearing: RcvStatus 0x%llx, continuing\n",
+ __func__, reg);
+ break;
+ }
+ udelay(2); /* do not busy-wait the CSR */
+ }
+
+ /* start the init - expect RcvCtrl to be 0 */
+ write_csr(dd, RCV_CTRL, RCV_CTRL_RX_RBUF_INIT_SMASK);
+
+ /*
+ * Read to force the write of Rcvtrl.RxRbufInit. There is a brief
+ * period after the write before RcvStatus.RxRbufInitDone is valid.
+ * The delay in the first run through the loop below is sufficient and
+ * required before the first read of RcvStatus.RxRbufInintDone.
+ */
+ read_csr(dd, RCV_CTRL);
+
+ /* wait for the init to finish */
+ count = 0;
+ while (1) {
+ /* delay is required first time through - see above */
+ udelay(2); /* do not busy-wait the CSR */
+ reg = read_csr(dd, RCV_STATUS);
+ if (reg & (RCV_STATUS_RX_RBUF_INIT_DONE_SMASK))
+ break;
+
+ /* give up after 100us - slowest possible at 33MHz is 73us */
+ if (count++ > 50) {
+ dd_dev_err(dd,
+ "%s: RcvStatus.RxRbufInit not set, continuing\n",
+ __func__);
+ break;
+ }
+ }
+}
+
+/* set RXE CSRs to chip reset defaults */
+static void reset_rxe_csrs(struct hfi1_devdata *dd)
+{
+ int i, j;
+
+ /*
+ * RXE Kernel CSRs
+ */
+ write_csr(dd, RCV_CTRL, 0);
+ init_rbufs(dd);
+ /* RCV_STATUS read-only */
+ /* RCV_CONTEXTS read-only */
+ /* RCV_ARRAY_CNT read-only */
+ /* RCV_BUF_SIZE read-only */
+ write_csr(dd, RCV_BTH_QP, 0);
+ write_csr(dd, RCV_MULTICAST, 0);
+ write_csr(dd, RCV_BYPASS, 0);
+ write_csr(dd, RCV_VL15, 0);
+ /* this is a clear-down */
+ write_csr(dd, RCV_ERR_INFO,
+ RCV_ERR_INFO_RCV_EXCESS_BUFFER_OVERRUN_SMASK);
+ /* RCV_ERR_STATUS read-only */
+ write_csr(dd, RCV_ERR_MASK, 0);
+ write_csr(dd, RCV_ERR_CLEAR, ~0ull);
+ /* RCV_ERR_FORCE leave alone */
+ for (i = 0; i < 32; i++)
+ write_csr(dd, RCV_QP_MAP_TABLE + (8 * i), 0);
+ for (i = 0; i < 4; i++)
+ write_csr(dd, RCV_PARTITION_KEY + (8 * i), 0);
+ for (i = 0; i < RXE_NUM_32_BIT_COUNTERS; i++)
+ write_csr(dd, RCV_COUNTER_ARRAY32 + (8 * i), 0);
+ for (i = 0; i < RXE_NUM_64_BIT_COUNTERS; i++)
+ write_csr(dd, RCV_COUNTER_ARRAY64 + (8 * i), 0);
+ for (i = 0; i < RXE_NUM_RSM_INSTANCES; i++) {
+ write_csr(dd, RCV_RSM_CFG + (8 * i), 0);
+ write_csr(dd, RCV_RSM_SELECT + (8 * i), 0);
+ write_csr(dd, RCV_RSM_MATCH + (8 * i), 0);
+ }
+ for (i = 0; i < 32; i++)
+ write_csr(dd, RCV_RSM_MAP_TABLE + (8 * i), 0);
+
+ /*
+ * RXE Kernel and User Per-Context CSRs
+ */
+ for (i = 0; i < dd->chip_rcv_contexts; i++) {
+ /* kernel */
+ write_kctxt_csr(dd, i, RCV_CTXT_CTRL, 0);
+ /* RCV_CTXT_STATUS read-only */
+ write_kctxt_csr(dd, i, RCV_EGR_CTRL, 0);
+ write_kctxt_csr(dd, i, RCV_TID_CTRL, 0);
+ write_kctxt_csr(dd, i, RCV_KEY_CTRL, 0);
+ write_kctxt_csr(dd, i, RCV_HDR_ADDR, 0);
+ write_kctxt_csr(dd, i, RCV_HDR_CNT, 0);
+ write_kctxt_csr(dd, i, RCV_HDR_ENT_SIZE, 0);
+ write_kctxt_csr(dd, i, RCV_HDR_SIZE, 0);
+ write_kctxt_csr(dd, i, RCV_HDR_TAIL_ADDR, 0);
+ write_kctxt_csr(dd, i, RCV_AVAIL_TIME_OUT, 0);
+ write_kctxt_csr(dd, i, RCV_HDR_OVFL_CNT, 0);
+
+ /* user */
+ /* RCV_HDR_TAIL read-only */
+ write_uctxt_csr(dd, i, RCV_HDR_HEAD, 0);
+ /* RCV_EGR_INDEX_TAIL read-only */
+ write_uctxt_csr(dd, i, RCV_EGR_INDEX_HEAD, 0);
+ /* RCV_EGR_OFFSET_TAIL read-only */
+ for (j = 0; j < RXE_NUM_TID_FLOWS; j++) {
+ write_uctxt_csr(dd, i, RCV_TID_FLOW_TABLE + (8 * j),
+ 0);
+ }
+ }
+}
+
+/*
+ * Set sc2vl tables.
+ *
+ * They power on to zeros, so to avoid send context errors
+ * they need to be set:
+ *
+ * SC 0-7 -> VL 0-7 (respectively)
+ * SC 15 -> VL 15
+ * otherwise
+ * -> VL 0
+ */
+static void init_sc2vl_tables(struct hfi1_devdata *dd)
+{
+ int i;
+ /* init per architecture spec, constrained by hardware capability */
+
+ /* HFI maps sent packets */
+ write_csr(dd, SEND_SC2VLT0, SC2VL_VAL(
+ 0,
+ 0, 0, 1, 1,
+ 2, 2, 3, 3,
+ 4, 4, 5, 5,
+ 6, 6, 7, 7));
+ write_csr(dd, SEND_SC2VLT1, SC2VL_VAL(
+ 1,
+ 8, 0, 9, 0,
+ 10, 0, 11, 0,
+ 12, 0, 13, 0,
+ 14, 0, 15, 15));
+ write_csr(dd, SEND_SC2VLT2, SC2VL_VAL(
+ 2,
+ 16, 0, 17, 0,
+ 18, 0, 19, 0,
+ 20, 0, 21, 0,
+ 22, 0, 23, 0));
+ write_csr(dd, SEND_SC2VLT3, SC2VL_VAL(
+ 3,
+ 24, 0, 25, 0,
+ 26, 0, 27, 0,
+ 28, 0, 29, 0,
+ 30, 0, 31, 0));
+
+ /* DC maps received packets */
+ write_csr(dd, DCC_CFG_SC_VL_TABLE_15_0, DC_SC_VL_VAL(
+ 15_0,
+ 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7,
+ 8, 0, 9, 0, 10, 0, 11, 0, 12, 0, 13, 0, 14, 0, 15, 15));
+ write_csr(dd, DCC_CFG_SC_VL_TABLE_31_16, DC_SC_VL_VAL(
+ 31_16,
+ 16, 0, 17, 0, 18, 0, 19, 0, 20, 0, 21, 0, 22, 0, 23, 0,
+ 24, 0, 25, 0, 26, 0, 27, 0, 28, 0, 29, 0, 30, 0, 31, 0));
+
+ /* initialize the cached sc2vl values consistently with h/w */
+ for (i = 0; i < 32; i++) {
+ if (i < 8 || i == 15)
+ *((u8 *)(dd->sc2vl) + i) = (u8)i;
+ else
+ *((u8 *)(dd->sc2vl) + i) = 0;
+ }
+}
+
+/*
+ * Read chip sizes and then reset parts to sane, disabled, values. We cannot
+ * depend on the chip going through a power-on reset - a driver may be loaded
+ * and unloaded many times.
+ *
+ * Do not write any CSR values to the chip in this routine - there may be
+ * a reset following the (possible) FLR in this routine.
+ *
+ */
+static void init_chip(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /*
+ * Put the HFI CSRs in a known state.
+ * Combine this with a DC reset.
+ *
+ * Stop the device from doing anything while we do a
+ * reset. We know there are no other active users of
+ * the device since we are now in charge. Turn off
+ * off all outbound and inbound traffic and make sure
+ * the device does not generate any interrupts.
+ */
+
+ /* disable send contexts and SDMA engines */
+ write_csr(dd, SEND_CTRL, 0);
+ for (i = 0; i < dd->chip_send_contexts; i++)
+ write_kctxt_csr(dd, i, SEND_CTXT_CTRL, 0);
+ for (i = 0; i < dd->chip_sdma_engines; i++)
+ write_kctxt_csr(dd, i, SEND_DMA_CTRL, 0);
+ /* disable port (turn off RXE inbound traffic) and contexts */
+ write_csr(dd, RCV_CTRL, 0);
+ for (i = 0; i < dd->chip_rcv_contexts; i++)
+ write_csr(dd, RCV_CTXT_CTRL, 0);
+ /* mask all interrupt sources */
+ for (i = 0; i < CCE_NUM_INT_CSRS; i++)
+ write_csr(dd, CCE_INT_MASK + (8*i), 0ull);
+
+ /*
+ * DC Reset: do a full DC reset before the register clear.
+ * A recommended length of time to hold is one CSR read,
+ * so reread the CceDcCtrl. Then, hold the DC in reset
+ * across the clear.
+ */
+ write_csr(dd, CCE_DC_CTRL, CCE_DC_CTRL_DC_RESET_SMASK);
+ (void) read_csr(dd, CCE_DC_CTRL);
+
+ if (use_flr) {
+ /*
+ * A FLR will reset the SPC core and part of the PCIe.
+ * The parts that need to be restored have already been
+ * saved.
+ */
+ dd_dev_info(dd, "Resetting CSRs with FLR\n");
+
+ /* do the FLR, the DC reset will remain */
+ hfi1_pcie_flr(dd);
+
+ /* restore command and BARs */
+ restore_pci_variables(dd);
+
+ if (is_a0(dd)) {
+ dd_dev_info(dd, "Resetting CSRs with FLR\n");
+ hfi1_pcie_flr(dd);
+ restore_pci_variables(dd);
+ }
+
+ } else {
+ dd_dev_info(dd, "Resetting CSRs with writes\n");
+ reset_cce_csrs(dd);
+ reset_txe_csrs(dd);
+ reset_rxe_csrs(dd);
+ reset_asic_csrs(dd);
+ reset_misc_csrs(dd);
+ }
+ /* clear the DC reset */
+ write_csr(dd, CCE_DC_CTRL, 0);
+ /* Set the LED off */
+ if (is_a0(dd))
+ setextled(dd, 0);
+ /*
+ * Clear the QSFP reset.
+ * A0 leaves the out lines floating on power on, then on an FLR
+ * enforces a 0 on all out pins. The driver does not touch
+ * ASIC_QSFPn_OUT otherwise. This leaves RESET_N low and
+ * anything plugged constantly in reset, if it pays attention
+ * to RESET_N.
+ * A prime example of this is SiPh. For now, set all pins high.
+ * I2CCLK and I2CDAT will change per direction, and INT_N and
+ * MODPRS_N are input only and their value is ignored.
+ */
+ if (is_a0(dd)) {
+ write_csr(dd, ASIC_QSFP1_OUT, 0x1f);
+ write_csr(dd, ASIC_QSFP2_OUT, 0x1f);
+ }
+}
+
+static void init_early_variables(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /* assign link credit variables */
+ dd->vau = CM_VAU;
+ dd->link_credits = CM_GLOBAL_CREDITS;
+ if (is_a0(dd))
+ dd->link_credits--;
+ dd->vcu = cu_to_vcu(hfi1_cu);
+ /* enough room for 8 MAD packets plus header - 17K */
+ dd->vl15_init = (8 * (2048 + 128)) / vau_to_au(dd->vau);
+ if (dd->vl15_init > dd->link_credits)
+ dd->vl15_init = dd->link_credits;
+
+ write_uninitialized_csrs_and_memories(dd);
+
+ if (HFI1_CAP_IS_KSET(PKEY_CHECK))
+ for (i = 0; i < dd->num_pports; i++) {
+ struct hfi1_pportdata *ppd = &dd->pport[i];
+
+ set_partition_keys(ppd);
+ }
+ init_sc2vl_tables(dd);
+}
+
+static void init_kdeth_qp(struct hfi1_devdata *dd)
+{
+ /* user changed the KDETH_QP */
+ if (kdeth_qp != 0 && kdeth_qp >= 0xff) {
+ /* out of range or illegal value */
+ dd_dev_err(dd, "Invalid KDETH queue pair prefix, ignoring");
+ kdeth_qp = 0;
+ }
+ if (kdeth_qp == 0) /* not set, or failed range check */
+ kdeth_qp = DEFAULT_KDETH_QP;
+
+ write_csr(dd, SEND_BTH_QP,
+ (kdeth_qp & SEND_BTH_QP_KDETH_QP_MASK)
+ << SEND_BTH_QP_KDETH_QP_SHIFT);
+
+ write_csr(dd, RCV_BTH_QP,
+ (kdeth_qp & RCV_BTH_QP_KDETH_QP_MASK)
+ << RCV_BTH_QP_KDETH_QP_SHIFT);
+}
+
+/**
+ * init_qpmap_table
+ * @dd - device data
+ * @first_ctxt - first context
+ * @last_ctxt - first context
+ *
+ * This return sets the qpn mapping table that
+ * is indexed by qpn[8:1].
+ *
+ * The routine will round robin the 256 settings
+ * from first_ctxt to last_ctxt.
+ *
+ * The first/last looks ahead to having specialized
+ * receive contexts for mgmt and bypass. Normal
+ * verbs traffic will assumed to be on a range
+ * of receive contexts.
+ */
+static void init_qpmap_table(struct hfi1_devdata *dd,
+ u32 first_ctxt,
+ u32 last_ctxt)
+{
+ u64 reg = 0;
+ u64 regno = RCV_QP_MAP_TABLE;
+ int i;
+ u64 ctxt = first_ctxt;
+
+ for (i = 0; i < 256;) {
+ if (ctxt == VL15CTXT) {
+ ctxt++;
+ if (ctxt > last_ctxt)
+ ctxt = first_ctxt;
+ continue;
+ }
+ reg |= ctxt << (8 * (i % 8));
+ i++;
+ ctxt++;
+ if (ctxt > last_ctxt)
+ ctxt = first_ctxt;
+ if (i % 8 == 0) {
+ write_csr(dd, regno, reg);
+ reg = 0;
+ regno += 8;
+ }
+ }
+ if (i % 8)
+ write_csr(dd, regno, reg);
+
+ add_rcvctrl(dd, RCV_CTRL_RCV_QP_MAP_ENABLE_SMASK
+ | RCV_CTRL_RCV_BYPASS_ENABLE_SMASK);
+}
+
+/**
+ * init_qos - init RX qos
+ * @dd - device data
+ * @first_context
+ *
+ * This routine initializes Rule 0 and the
+ * RSM map table to implement qos.
+ *
+ * If all of the limit tests succeed,
+ * qos is applied based on the array
+ * interpretation of krcvqs where
+ * entry 0 is VL0.
+ *
+ * The number of vl bits (n) and the number of qpn
+ * bits (m) are computed to feed both the RSM map table
+ * and the single rule.
+ *
+ */
+static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
+{
+ u8 max_by_vl = 0;
+ unsigned qpns_per_vl, ctxt, i, qpn, n = 1, m;
+ u64 *rsmmap;
+ u64 reg;
+ u8 rxcontext = is_a0(dd) ? 0 : 0xff; /* 0 is default if a0 ver. */
+
+ /* validate */
+ if (dd->n_krcv_queues <= MIN_KERNEL_KCTXTS ||
+ num_vls == 1 ||
+ krcvqsset <= 1)
+ goto bail;
+ for (i = 0; i < min_t(unsigned, num_vls, krcvqsset); i++)
+ if (krcvqs[i] > max_by_vl)
+ max_by_vl = krcvqs[i];
+ if (max_by_vl > 32)
+ goto bail;
+ qpns_per_vl = __roundup_pow_of_two(max_by_vl);
+ /* determine bits vl */
+ n = ilog2(num_vls);
+ /* determine bits for qpn */
+ m = ilog2(qpns_per_vl);
+ if ((m + n) > 7)
+ goto bail;
+ if (num_vls * qpns_per_vl > dd->chip_rcv_contexts)
+ goto bail;
+ rsmmap = kmalloc_array(NUM_MAP_REGS, sizeof(u64), GFP_KERNEL);
+ memset(rsmmap, rxcontext, NUM_MAP_REGS * sizeof(u64));
+ /* init the local copy of the table */
+ for (i = 0, ctxt = first_ctxt; i < num_vls; i++) {
+ unsigned tctxt;
+
+ for (qpn = 0, tctxt = ctxt;
+ krcvqs[i] && qpn < qpns_per_vl; qpn++) {
+ unsigned idx, regoff, regidx;
+
+ /* generate index <= 128 */
+ idx = (qpn << n) ^ i;
+ regoff = (idx % 8) * 8;
+ regidx = idx / 8;
+ reg = rsmmap[regidx];
+ /* replace 0xff with context number */
+ reg &= ~(RCV_RSM_MAP_TABLE_RCV_CONTEXT_A_MASK
+ << regoff);
+ reg |= (u64)(tctxt++) << regoff;
+ rsmmap[regidx] = reg;
+ if (tctxt == ctxt + krcvqs[i])
+ tctxt = ctxt;
+ }
+ ctxt += krcvqs[i];
+ }
+ /* flush cached copies to chip */
+ for (i = 0; i < NUM_MAP_REGS; i++)
+ write_csr(dd, RCV_RSM_MAP_TABLE + (8 * i), rsmmap[i]);
+ /* add rule0 */
+ write_csr(dd, RCV_RSM_CFG /* + (8 * 0) */,
+ RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_MASK
+ << RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_SHIFT |
+ 2ull << RCV_RSM_CFG_PACKET_TYPE_SHIFT);
+ write_csr(dd, RCV_RSM_SELECT /* + (8 * 0) */,
+ LRH_BTH_MATCH_OFFSET
+ << RCV_RSM_SELECT_FIELD1_OFFSET_SHIFT |
+ LRH_SC_MATCH_OFFSET << RCV_RSM_SELECT_FIELD2_OFFSET_SHIFT |
+ LRH_SC_SELECT_OFFSET << RCV_RSM_SELECT_INDEX1_OFFSET_SHIFT |
+ ((u64)n) << RCV_RSM_SELECT_INDEX1_WIDTH_SHIFT |
+ QPN_SELECT_OFFSET << RCV_RSM_SELECT_INDEX2_OFFSET_SHIFT |
+ ((u64)m + (u64)n) << RCV_RSM_SELECT_INDEX2_WIDTH_SHIFT);
+ write_csr(dd, RCV_RSM_MATCH /* + (8 * 0) */,
+ LRH_BTH_MASK << RCV_RSM_MATCH_MASK1_SHIFT |
+ LRH_BTH_VALUE << RCV_RSM_MATCH_VALUE1_SHIFT |
+ LRH_SC_MASK << RCV_RSM_MATCH_MASK2_SHIFT |
+ LRH_SC_VALUE << RCV_RSM_MATCH_VALUE2_SHIFT);
+ /* Enable RSM */
+ add_rcvctrl(dd, RCV_CTRL_RCV_RSM_ENABLE_SMASK);
+ kfree(rsmmap);
+ /* map everything else (non-VL15) to context 0 */
+ init_qpmap_table(
+ dd,
+ 0,
+ 0);
+ dd->qos_shift = n + 1;
+ return;
+bail:
+ dd->qos_shift = 1;
+ init_qpmap_table(
+ dd,
+ dd->n_krcv_queues > MIN_KERNEL_KCTXTS ? MIN_KERNEL_KCTXTS : 0,
+ dd->n_krcv_queues - 1);
+}
+
+static void init_rxe(struct hfi1_devdata *dd)
+{
+ /* enable all receive errors */
+ write_csr(dd, RCV_ERR_MASK, ~0ull);
+ /* setup QPN map table - start where VL15 context leaves off */
+ init_qos(
+ dd,
+ dd->n_krcv_queues > MIN_KERNEL_KCTXTS ? MIN_KERNEL_KCTXTS : 0);
+ /*
+ * make sure RcvCtrl.RcvWcb <= PCIe Device Control
+ * Register Max_Payload_Size (PCI_EXP_DEVCTL in Linux PCIe config
+ * space, PciCfgCap2.MaxPayloadSize in HFI). There is only one
+ * invalid configuration: RcvCtrl.RcvWcb set to its max of 256 and
+ * Max_PayLoad_Size set to its minimum of 128.
+ *
+ * Presently, RcvCtrl.RcvWcb is not modified from its default of 0
+ * (64 bytes). Max_Payload_Size is possibly modified upward in
+ * tune_pcie_caps() which is called after this routine.
+ */
+}
+
+static void init_other(struct hfi1_devdata *dd)
+{
+ /* enable all CCE errors */
+ write_csr(dd, CCE_ERR_MASK, ~0ull);
+ /* enable *some* Misc errors */
+ write_csr(dd, MISC_ERR_MASK, DRIVER_MISC_MASK);
+ /* enable all DC errors, except LCB */
+ write_csr(dd, DCC_ERR_FLG_EN, ~0ull);
+ write_csr(dd, DC_DC8051_ERR_EN, ~0ull);
+}
+
+/*
+ * Fill out the given AU table using the given CU. A CU is defined in terms
+ * AUs. The table is a an encoding: given the index, how many AUs does that
+ * represent?
+ *
+ * NOTE: Assumes that the register layout is the same for the
+ * local and remote tables.
+ */
+static void assign_cm_au_table(struct hfi1_devdata *dd, u32 cu,
+ u32 csr0to3, u32 csr4to7)
+{
+ write_csr(dd, csr0to3,
+ 0ull <<
+ SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE0_SHIFT
+ | 1ull <<
+ SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE1_SHIFT
+ | 2ull * cu <<
+ SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE2_SHIFT
+ | 4ull * cu <<
+ SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE3_SHIFT);
+ write_csr(dd, csr4to7,
+ 8ull * cu <<
+ SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE4_SHIFT
+ | 16ull * cu <<
+ SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE5_SHIFT
+ | 32ull * cu <<
+ SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE6_SHIFT
+ | 64ull * cu <<
+ SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE7_SHIFT);
+
+}
+
+static void assign_local_cm_au_table(struct hfi1_devdata *dd, u8 vcu)
+{
+ assign_cm_au_table(dd, vcu_to_cu(vcu), SEND_CM_LOCAL_AU_TABLE0_TO3,
+ SEND_CM_LOCAL_AU_TABLE4_TO7);
+}
+
+void assign_remote_cm_au_table(struct hfi1_devdata *dd, u8 vcu)
+{
+ assign_cm_au_table(dd, vcu_to_cu(vcu), SEND_CM_REMOTE_AU_TABLE0_TO3,
+ SEND_CM_REMOTE_AU_TABLE4_TO7);
+}
+
+static void init_txe(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /* enable all PIO, SDMA, general, and Egress errors */
+ write_csr(dd, SEND_PIO_ERR_MASK, ~0ull);
+ write_csr(dd, SEND_DMA_ERR_MASK, ~0ull);
+ write_csr(dd, SEND_ERR_MASK, ~0ull);
+ write_csr(dd, SEND_EGRESS_ERR_MASK, ~0ull);
+
+ /* enable all per-context and per-SDMA engine errors */
+ for (i = 0; i < dd->chip_send_contexts; i++)
+ write_kctxt_csr(dd, i, SEND_CTXT_ERR_MASK, ~0ull);
+ for (i = 0; i < dd->chip_sdma_engines; i++)
+ write_kctxt_csr(dd, i, SEND_DMA_ENG_ERR_MASK, ~0ull);
+
+ /* set the local CU to AU mapping */
+ assign_local_cm_au_table(dd, dd->vcu);
+
+ /*
+ * Set reasonable default for Credit Return Timer
+ * Don't set on Simulator - causes it to choke.
+ */
+ if (dd->icode != ICODE_FUNCTIONAL_SIMULATOR)
+ write_csr(dd, SEND_CM_TIMER_CTRL, HFI1_CREDIT_RETURN_RATE);
+}
+
+int hfi1_set_ctxt_jkey(struct hfi1_devdata *dd, unsigned ctxt, u16 jkey)
+{
+ struct hfi1_ctxtdata *rcd = dd->rcd[ctxt];
+ unsigned sctxt;
+ int ret = 0;
+ u64 reg;
+
+ if (!rcd || !rcd->sc) {
+ ret = -EINVAL;
+ goto done;
+ }
+ sctxt = rcd->sc->hw_context;
+ reg = SEND_CTXT_CHECK_JOB_KEY_MASK_SMASK | /* mask is always 1's */
+ ((jkey & SEND_CTXT_CHECK_JOB_KEY_VALUE_MASK) <<
+ SEND_CTXT_CHECK_JOB_KEY_VALUE_SHIFT);
+ /* JOB_KEY_ALLOW_PERMISSIVE is not allowed by default */
+ if (HFI1_CAP_KGET_MASK(rcd->flags, ALLOW_PERM_JKEY))
+ reg |= SEND_CTXT_CHECK_JOB_KEY_ALLOW_PERMISSIVE_SMASK;
+ write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_JOB_KEY, reg);
+ /*
+ * Enable send-side J_KEY integrity check, unless this is A0 h/w
+ * (due to A0 erratum).
+ */
+ if (!is_a0(dd)) {
+ reg = read_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE);
+ reg |= SEND_CTXT_CHECK_ENABLE_CHECK_JOB_KEY_SMASK;
+ write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE, reg);
+ }
+
+ /* Enable J_KEY check on receive context. */
+ reg = RCV_KEY_CTRL_JOB_KEY_ENABLE_SMASK |
+ ((jkey & RCV_KEY_CTRL_JOB_KEY_VALUE_MASK) <<
+ RCV_KEY_CTRL_JOB_KEY_VALUE_SHIFT);
+ write_kctxt_csr(dd, ctxt, RCV_KEY_CTRL, reg);
+done:
+ return ret;
+}
+
+int hfi1_clear_ctxt_jkey(struct hfi1_devdata *dd, unsigned ctxt)
+{
+ struct hfi1_ctxtdata *rcd = dd->rcd[ctxt];
+ unsigned sctxt;
+ int ret = 0;
+ u64 reg;
+
+ if (!rcd || !rcd->sc) {
+ ret = -EINVAL;
+ goto done;
+ }
+ sctxt = rcd->sc->hw_context;
+ write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_JOB_KEY, 0);
+ /*
+ * Disable send-side J_KEY integrity check, unless this is A0 h/w.
+ * This check would not have been enabled for A0 h/w, see
+ * set_ctxt_jkey().
+ */
+ if (!is_a0(dd)) {
+ reg = read_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE);
+ reg &= ~SEND_CTXT_CHECK_ENABLE_CHECK_JOB_KEY_SMASK;
+ write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE, reg);
+ }
+ /* Turn off the J_KEY on the receive side */
+ write_kctxt_csr(dd, ctxt, RCV_KEY_CTRL, 0);
+done:
+ return ret;
+}
+
+int hfi1_set_ctxt_pkey(struct hfi1_devdata *dd, unsigned ctxt, u16 pkey)
+{
+ struct hfi1_ctxtdata *rcd;
+ unsigned sctxt;
+ int ret = 0;
+ u64 reg;
+
+ if (ctxt < dd->num_rcv_contexts)
+ rcd = dd->rcd[ctxt];
+ else {
+ ret = -EINVAL;
+ goto done;
+ }
+ if (!rcd || !rcd->sc) {
+ ret = -EINVAL;
+ goto done;
+ }
+ sctxt = rcd->sc->hw_context;
+ reg = ((u64)pkey & SEND_CTXT_CHECK_PARTITION_KEY_VALUE_MASK) <<
+ SEND_CTXT_CHECK_PARTITION_KEY_VALUE_SHIFT;
+ write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_PARTITION_KEY, reg);
+ reg = read_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE);
+ reg |= SEND_CTXT_CHECK_ENABLE_CHECK_PARTITION_KEY_SMASK;
+ write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE, reg);
+done:
+ return ret;
+}
+
+int hfi1_clear_ctxt_pkey(struct hfi1_devdata *dd, unsigned ctxt)
+{
+ struct hfi1_ctxtdata *rcd;
+ unsigned sctxt;
+ int ret = 0;
+ u64 reg;
+
+ if (ctxt < dd->num_rcv_contexts)
+ rcd = dd->rcd[ctxt];
+ else {
+ ret = -EINVAL;
+ goto done;
+ }
+ if (!rcd || !rcd->sc) {
+ ret = -EINVAL;
+ goto done;
+ }
+ sctxt = rcd->sc->hw_context;
+ reg = read_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE);
+ reg &= ~SEND_CTXT_CHECK_ENABLE_CHECK_PARTITION_KEY_SMASK;
+ write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_ENABLE, reg);
+ write_kctxt_csr(dd, sctxt, SEND_CTXT_CHECK_PARTITION_KEY, 0);
+done:
+ return ret;
+}
+
+/*
+ * Start doing the clean up the the chip. Our clean up happens in multiple
+ * stages and this is just the first.
+ */
+void hfi1_start_cleanup(struct hfi1_devdata *dd)
+{
+ free_cntrs(dd);
+ free_rcverr(dd);
+ clean_up_interrupts(dd);
+}
+
+#define HFI_BASE_GUID(dev) \
+ ((dev)->base_guid & ~(1ULL << GUID_HFI_INDEX_SHIFT))
+
+/*
+ * Certain chip functions need to be initialized only once per asic
+ * instead of per-device. This function finds the peer device and
+ * checks whether that chip initialization needs to be done by this
+ * device.
+ */
+static void asic_should_init(struct hfi1_devdata *dd)
+{
+ unsigned long flags;
+ struct hfi1_devdata *tmp, *peer = NULL;
+
+ spin_lock_irqsave(&hfi1_devs_lock, flags);
+ /* Find our peer device */
+ list_for_each_entry(tmp, &hfi1_dev_list, list) {
+ if ((HFI_BASE_GUID(dd) == HFI_BASE_GUID(tmp)) &&
+ dd->unit != tmp->unit) {
+ peer = tmp;
+ break;
+ }
+ }
+
+ /*
+ * "Claim" the ASIC for initialization if it hasn't been
+ " "claimed" yet.
+ */
+ if (!peer || !(peer->flags & HFI1_DO_INIT_ASIC))
+ dd->flags |= HFI1_DO_INIT_ASIC;
+ spin_unlock_irqrestore(&hfi1_devs_lock, flags);
+}
+
+/**
+ * Allocate an initialize the device structure for the hfi.
+ * @dev: the pci_dev for hfi1_ib device
+ * @ent: pci_device_id struct for this dev
+ *
+ * Also allocates, initializes, and returns the devdata struct for this
+ * device instance
+ *
+ * This is global, and is called directly at init to set up the
+ * chip-specific function pointers for later use.
+ */
+struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ struct hfi1_devdata *dd;
+ struct hfi1_pportdata *ppd;
+ u64 reg;
+ int i, ret;
+ static const char * const inames[] = { /* implementation names */
+ "RTL silicon",
+ "RTL VCS simulation",
+ "RTL FPGA emulation",
+ "Functional simulator"
+ };
+
+ dd = hfi1_alloc_devdata(pdev,
+ NUM_IB_PORTS * sizeof(struct hfi1_pportdata));
+ if (IS_ERR(dd))
+ goto bail;
+ ppd = dd->pport;
+ for (i = 0; i < dd->num_pports; i++, ppd++) {
+ int vl;
+ /* init common fields */
+ hfi1_init_pportdata(pdev, ppd, dd, 0, 1);
+ /* DC supports 4 link widths */
+ ppd->link_width_supported =
+ OPA_LINK_WIDTH_1X | OPA_LINK_WIDTH_2X |
+ OPA_LINK_WIDTH_3X | OPA_LINK_WIDTH_4X;
+ ppd->link_width_downgrade_supported =
+ ppd->link_width_supported;
+ /* start out enabling only 4X */
+ ppd->link_width_enabled = OPA_LINK_WIDTH_4X;
+ ppd->link_width_downgrade_enabled =
+ ppd->link_width_downgrade_supported;
+ /* link width active is 0 when link is down */
+ /* link width downgrade active is 0 when link is down */
+
+ if (num_vls < HFI1_MIN_VLS_SUPPORTED
+ || num_vls > HFI1_MAX_VLS_SUPPORTED) {
+ hfi1_early_err(&pdev->dev,
+ "Invalid num_vls %u, using %u VLs\n",
+ num_vls, HFI1_MAX_VLS_SUPPORTED);
+ num_vls = HFI1_MAX_VLS_SUPPORTED;
+ }
+ ppd->vls_supported = num_vls;
+ ppd->vls_operational = ppd->vls_supported;
+ /* Set the default MTU. */
+ for (vl = 0; vl < num_vls; vl++)
+ dd->vld[vl].mtu = hfi1_max_mtu;
+ dd->vld[15].mtu = MAX_MAD_PACKET;
+ /*
+ * Set the initial values to reasonable default, will be set
+ * for real when link is up.
+ */
+ ppd->lstate = IB_PORT_DOWN;
+ ppd->overrun_threshold = 0x4;
+ ppd->phy_error_threshold = 0xf;
+ ppd->port_crc_mode_enabled = link_crc_mask;
+ /* initialize supported LTP CRC mode */
+ ppd->port_ltp_crc_mode = cap_to_port_ltp(link_crc_mask) << 8;
+ /* initialize enabled LTP CRC mode */
+ ppd->port_ltp_crc_mode |= cap_to_port_ltp(link_crc_mask) << 4;
+ /* start in offline */
+ ppd->host_link_state = HLS_DN_OFFLINE;
+ init_vl_arb_caches(ppd);
+ }
+
+ dd->link_default = HLS_DN_POLL;
+
+ /*
+ * Do remaining PCIe setup and save PCIe values in dd.
+ * Any error printing is already done by the init code.
+ * On return, we have the chip mapped.
+ */
+ ret = hfi1_pcie_ddinit(dd, pdev, ent);
+ if (ret < 0)
+ goto bail_free;
+
+ /* verify that reads actually work, save revision for reset check */
+ dd->revision = read_csr(dd, CCE_REVISION);
+ if (dd->revision == ~(u64)0) {
+ dd_dev_err(dd, "cannot read chip CSRs\n");
+ ret = -EINVAL;
+ goto bail_cleanup;
+ }
+ dd->majrev = (dd->revision >> CCE_REVISION_CHIP_REV_MAJOR_SHIFT)
+ & CCE_REVISION_CHIP_REV_MAJOR_MASK;
+ dd->minrev = (dd->revision >> CCE_REVISION_CHIP_REV_MINOR_SHIFT)
+ & CCE_REVISION_CHIP_REV_MINOR_MASK;
+
+ /* obtain the hardware ID - NOT related to unit, which is a
+ software enumeration */
+ reg = read_csr(dd, CCE_REVISION2);
+ dd->hfi1_id = (reg >> CCE_REVISION2_HFI_ID_SHIFT)
+ & CCE_REVISION2_HFI_ID_MASK;
+ /* the variable size will remove unwanted bits */
+ dd->icode = reg >> CCE_REVISION2_IMPL_CODE_SHIFT;
+ dd->irev = reg >> CCE_REVISION2_IMPL_REVISION_SHIFT;
+ dd_dev_info(dd, "Implementation: %s, revision 0x%x\n",
+ dd->icode < ARRAY_SIZE(inames) ? inames[dd->icode] : "unknown",
+ (int)dd->irev);
+
+ /* speeds the hardware can support */
+ dd->pport->link_speed_supported = OPA_LINK_SPEED_25G;
+ /* speeds allowed to run at */
+ dd->pport->link_speed_enabled = dd->pport->link_speed_supported;
+ /* give a reasonable active value, will be set on link up */
+ dd->pport->link_speed_active = OPA_LINK_SPEED_25G;
+
+ dd->chip_rcv_contexts = read_csr(dd, RCV_CONTEXTS);
+ dd->chip_send_contexts = read_csr(dd, SEND_CONTEXTS);
+ dd->chip_sdma_engines = read_csr(dd, SEND_DMA_ENGINES);
+ dd->chip_pio_mem_size = read_csr(dd, SEND_PIO_MEM_SIZE);
+ dd->chip_sdma_mem_size = read_csr(dd, SEND_DMA_MEM_SIZE);
+ /* fix up link widths for emulation _p */
+ ppd = dd->pport;
+ if (dd->icode == ICODE_FPGA_EMULATION && is_emulator_p(dd)) {
+ ppd->link_width_supported =
+ ppd->link_width_enabled =
+ ppd->link_width_downgrade_supported =
+ ppd->link_width_downgrade_enabled =
+ OPA_LINK_WIDTH_1X;
+ }
+ /* insure num_vls isn't larger than number of sdma engines */
+ if (HFI1_CAP_IS_KSET(SDMA) && num_vls > dd->chip_sdma_engines) {
+ dd_dev_err(dd, "num_vls %u too large, using %u VLs\n",
+ num_vls, HFI1_MAX_VLS_SUPPORTED);
+ ppd->vls_supported = num_vls = HFI1_MAX_VLS_SUPPORTED;
+ ppd->vls_operational = ppd->vls_supported;
+ }
+
+ /*
+ * Convert the ns parameter to the 64 * cclocks used in the CSR.
+ * Limit the max if larger than the field holds. If timeout is
+ * non-zero, then the calculated field will be at least 1.
+ *
+ * Must be after icode is set up - the cclock rate depends
+ * on knowing the hardware being used.
+ */
+ dd->rcv_intr_timeout_csr = ns_to_cclock(dd, rcv_intr_timeout) / 64;
+ if (dd->rcv_intr_timeout_csr >
+ RCV_AVAIL_TIME_OUT_TIME_OUT_RELOAD_MASK)
+ dd->rcv_intr_timeout_csr =
+ RCV_AVAIL_TIME_OUT_TIME_OUT_RELOAD_MASK;
+ else if (dd->rcv_intr_timeout_csr == 0 && rcv_intr_timeout)
+ dd->rcv_intr_timeout_csr = 1;
+
+ /* obtain chip sizes, reset chip CSRs */
+ init_chip(dd);
+
+ /* read in the PCIe link speed information */
+ ret = pcie_speeds(dd);
+ if (ret)
+ goto bail_cleanup;
+
+ /* needs to be done before we look for the peer device */
+ read_guid(dd);
+
+ asic_should_init(dd);
+
+ /* read in firmware */
+ ret = hfi1_firmware_init(dd);
+ if (ret)
+ goto bail_cleanup;
+
+ /*
+ * In general, the PCIe Gen3 transition must occur after the
+ * chip has been idled (so it won't initiate any PCIe transactions
+ * e.g. an interrupt) and before the driver changes any registers
+ * (the transition will reset the registers).
+ *
+ * In particular, place this call after:
+ * - init_chip() - the chip will not initiate any PCIe transactions
+ * - pcie_speeds() - reads the current link speed
+ * - hfi1_firmware_init() - the needed firmware is ready to be
+ * downloaded
+ */
+ ret = do_pcie_gen3_transition(dd);
+ if (ret)
+ goto bail_cleanup;
+
+ /* start setting dd values and adjusting CSRs */
+ init_early_variables(dd);
+
+ parse_platform_config(dd);
+
+ /* add board names as they are defined */
+ dd->boardname = kmalloc(64, GFP_KERNEL);
+ if (!dd->boardname)
+ goto bail_cleanup;
+ snprintf(dd->boardname, 64, "Board ID 0x%llx",
+ dd->revision >> CCE_REVISION_BOARD_ID_LOWER_NIBBLE_SHIFT
+ & CCE_REVISION_BOARD_ID_LOWER_NIBBLE_MASK);
+
+ snprintf(dd->boardversion, BOARD_VERS_MAX,
+ "ChipABI %u.%u, %s, ChipRev %u.%u, SW Compat %llu\n",
+ HFI1_CHIP_VERS_MAJ, HFI1_CHIP_VERS_MIN,
+ dd->boardname,
+ (u32)dd->majrev,
+ (u32)dd->minrev,
+ (dd->revision >> CCE_REVISION_SW_SHIFT)
+ & CCE_REVISION_SW_MASK);
+
+ ret = set_up_context_variables(dd);
+ if (ret)
+ goto bail_cleanup;
+
+ /* set initial RXE CSRs */
+ init_rxe(dd);
+ /* set initial TXE CSRs */
+ init_txe(dd);
+ /* set initial non-RXE, non-TXE CSRs */
+ init_other(dd);
+ /* set up KDETH QP prefix in both RX and TX CSRs */
+ init_kdeth_qp(dd);
+
+ /* send contexts must be set up before receive contexts */
+ ret = init_send_contexts(dd);
+ if (ret)
+ goto bail_cleanup;
+
+ ret = hfi1_create_ctxts(dd);
+ if (ret)
+ goto bail_cleanup;
+
+ dd->rcvhdrsize = DEFAULT_RCVHDRSIZE;
+ /*
+ * rcd[0] is guaranteed to be valid by this point. Also, all
+ * context are using the same value, as per the module parameter.
+ */
+ dd->rhf_offset = dd->rcd[0]->rcvhdrqentsize - sizeof(u64) / sizeof(u32);
+
+ ret = init_pervl_scs(dd);
+ if (ret)
+ goto bail_cleanup;
+
+ /* sdma init */
+ for (i = 0; i < dd->num_pports; ++i) {
+ ret = sdma_init(dd, i);
+ if (ret)
+ goto bail_cleanup;
+ }
+
+ /* use contexts created by hfi1_create_ctxts */
+ ret = set_up_interrupts(dd);
+ if (ret)
+ goto bail_cleanup;
+
+ /* set up LCB access - must be after set_up_interrupts() */
+ init_lcb_access(dd);
+
+ snprintf(dd->serial, SERIAL_MAX, "0x%08llx\n",
+ dd->base_guid & 0xFFFFFF);
+
+ dd->oui1 = dd->base_guid >> 56 & 0xFF;
+ dd->oui2 = dd->base_guid >> 48 & 0xFF;
+ dd->oui3 = dd->base_guid >> 40 & 0xFF;
+
+ ret = load_firmware(dd); /* asymmetric with dispose_firmware() */
+ if (ret)
+ goto bail_clear_intr;
+ check_fabric_firmware_versions(dd);
+
+ thermal_init(dd);
+
+ ret = init_cntrs(dd);
+ if (ret)
+ goto bail_clear_intr;
+
+ ret = init_rcverr(dd);
+ if (ret)
+ goto bail_free_cntrs;
+
+ ret = eprom_init(dd);
+ if (ret)
+ goto bail_free_rcverr;
+
+ goto bail;
+
+bail_free_rcverr:
+ free_rcverr(dd);
+bail_free_cntrs:
+ free_cntrs(dd);
+bail_clear_intr:
+ clean_up_interrupts(dd);
+bail_cleanup:
+ hfi1_pcie_ddcleanup(dd);
+bail_free:
+ hfi1_free_devdata(dd);
+ dd = ERR_PTR(ret);
+bail:
+ return dd;
+}
+
+static u16 delay_cycles(struct hfi1_pportdata *ppd, u32 desired_egress_rate,
+ u32 dw_len)
+{
+ u32 delta_cycles;
+ u32 current_egress_rate = ppd->current_egress_rate;
+ /* rates here are in units of 10^6 bits/sec */
+
+ if (desired_egress_rate == -1)
+ return 0; /* shouldn't happen */
+
+ if (desired_egress_rate >= current_egress_rate)
+ return 0; /* we can't help go faster, only slower */
+
+ delta_cycles = egress_cycles(dw_len * 4, desired_egress_rate) -
+ egress_cycles(dw_len * 4, current_egress_rate);
+
+ return (u16)delta_cycles;
+}
+
+
+/**
+ * create_pbc - build a pbc for transmission
+ * @flags: special case flags or-ed in built pbc
+ * @srate: static rate
+ * @vl: vl
+ * @dwlen: dword length (header words + data words + pbc words)
+ *
+ * Create a PBC with the given flags, rate, VL, and length.
+ *
+ * NOTE: The PBC created will not insert any HCRC - all callers but one are
+ * for verbs, which does not use this PSM feature. The lone other caller
+ * is for the diagnostic interface which calls this if the user does not
+ * supply their own PBC.
+ */
+u64 create_pbc(struct hfi1_pportdata *ppd, u64 flags, int srate_mbs, u32 vl,
+ u32 dw_len)
+{
+ u64 pbc, delay = 0;
+
+ if (unlikely(srate_mbs))
+ delay = delay_cycles(ppd, srate_mbs, dw_len);
+
+ pbc = flags
+ | (delay << PBC_STATIC_RATE_CONTROL_COUNT_SHIFT)
+ | ((u64)PBC_IHCRC_NONE << PBC_INSERT_HCRC_SHIFT)
+ | (vl & PBC_VL_MASK) << PBC_VL_SHIFT
+ | (dw_len & PBC_LENGTH_DWS_MASK)
+ << PBC_LENGTH_DWS_SHIFT;
+
+ return pbc;
+}
+
+#define SBUS_THERMAL 0x4f
+#define SBUS_THERM_MONITOR_MODE 0x1
+
+#define THERM_FAILURE(dev, ret, reason) \
+ dd_dev_err((dd), \
+ "Thermal sensor initialization failed: %s (%d)\n", \
+ (reason), (ret))
+
+/*
+ * Initialize the Avago Thermal sensor.
+ *
+ * After initialization, enable polling of thermal sensor through
+ * SBus interface. In order for this to work, the SBus Master
+ * firmware has to be loaded due to the fact that the HW polling
+ * logic uses SBus interrupts, which are not supported with
+ * default firmware. Otherwise, no data will be returned through
+ * the ASIC_STS_THERM CSR.
+ */
+static int thermal_init(struct hfi1_devdata *dd)
+{
+ int ret = 0;
+
+ if (dd->icode != ICODE_RTL_SILICON ||
+ !(dd->flags & HFI1_DO_INIT_ASIC))
+ return ret;
+
+ acquire_hw_mutex(dd);
+ dd_dev_info(dd, "Initializing thermal sensor\n");
+ /* Thermal Sensor Initialization */
+ /* Step 1: Reset the Thermal SBus Receiver */
+ ret = sbus_request_slow(dd, SBUS_THERMAL, 0x0,
+ RESET_SBUS_RECEIVER, 0);
+ if (ret) {
+ THERM_FAILURE(dd, ret, "Bus Reset");
+ goto done;
+ }
+ /* Step 2: Set Reset bit in Thermal block */
+ ret = sbus_request_slow(dd, SBUS_THERMAL, 0x0,
+ WRITE_SBUS_RECEIVER, 0x1);
+ if (ret) {
+ THERM_FAILURE(dd, ret, "Therm Block Reset");
+ goto done;
+ }
+ /* Step 3: Write clock divider value (100MHz -> 2MHz) */
+ ret = sbus_request_slow(dd, SBUS_THERMAL, 0x1,
+ WRITE_SBUS_RECEIVER, 0x32);
+ if (ret) {
+ THERM_FAILURE(dd, ret, "Write Clock Div");
+ goto done;
+ }
+ /* Step 4: Select temperature mode */
+ ret = sbus_request_slow(dd, SBUS_THERMAL, 0x3,
+ WRITE_SBUS_RECEIVER,
+ SBUS_THERM_MONITOR_MODE);
+ if (ret) {
+ THERM_FAILURE(dd, ret, "Write Mode Sel");
+ goto done;
+ }
+ /* Step 5: De-assert block reset and start conversion */
+ ret = sbus_request_slow(dd, SBUS_THERMAL, 0x0,
+ WRITE_SBUS_RECEIVER, 0x2);
+ if (ret) {
+ THERM_FAILURE(dd, ret, "Write Reset Deassert");
+ goto done;
+ }
+ /* Step 5.1: Wait for first conversion (21.5ms per spec) */
+ msleep(22);
+
+ /* Enable polling of thermal readings */
+ write_csr(dd, ASIC_CFG_THERM_POLL_EN, 0x1);
+done:
+ release_hw_mutex(dd);
+ return ret;
+}
+
+static void handle_temp_err(struct hfi1_devdata *dd)
+{
+ struct hfi1_pportdata *ppd = &dd->pport[0];
+ /*
+ * Thermal Critical Interrupt
+ * Put the device into forced freeze mode, take link down to
+ * offline, and put DC into reset.
+ */
+ dd_dev_emerg(dd,
+ "Critical temperature reached! Forcing device into freeze mode!\n");
+ dd->flags |= HFI1_FORCED_FREEZE;
+ start_freeze_handling(ppd, FREEZE_SELF|FREEZE_ABORT);
+ /*
+ * Shut DC down as much and as quickly as possible.
+ *
+ * Step 1: Take the link down to OFFLINE. This will cause the
+ * 8051 to put the Serdes in reset. However, we don't want to
+ * go through the entire link state machine since we want to
+ * shutdown ASAP. Furthermore, this is not a graceful shutdown
+ * but rather an attempt to save the chip.
+ * Code below is almost the same as quiet_serdes() but avoids
+ * all the extra work and the sleeps.
+ */
+ ppd->driver_link_ready = 0;
+ ppd->link_enabled = 0;
+ set_physical_link_state(dd, PLS_OFFLINE |
+ (OPA_LINKDOWN_REASON_SMA_DISABLED << 8));
+ /*
+ * Step 2: Shutdown LCB and 8051
+ * After shutdown, do not restore DC_CFG_RESET value.
+ */
+ dc_shutdown(dd);
+}
diff --git a/drivers/staging/rdma/hfi1/chip.h b/drivers/staging/rdma/hfi1/chip.h
new file mode 100644
index 000000000000..f89a432c7334
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/chip.h
@@ -0,0 +1,1035 @@
+#ifndef _CHIP_H
+#define _CHIP_H
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+/*
+ * This file contains all of the defines that is specific to the HFI chip
+ */
+
+/* sizes */
+#define CCE_NUM_MSIX_VECTORS 256
+#define CCE_NUM_INT_CSRS 12
+#define CCE_NUM_INT_MAP_CSRS 96
+#define NUM_INTERRUPT_SOURCES 768
+#define RXE_NUM_CONTEXTS 160
+#define RXE_PER_CONTEXT_SIZE 0x1000 /* 4k */
+#define RXE_NUM_TID_FLOWS 32
+#define RXE_NUM_DATA_VL 8
+#define TXE_NUM_CONTEXTS 160
+#define TXE_NUM_SDMA_ENGINES 16
+#define NUM_CONTEXTS_PER_SET 8
+#define VL_ARB_HIGH_PRIO_TABLE_SIZE 16
+#define VL_ARB_LOW_PRIO_TABLE_SIZE 16
+#define VL_ARB_TABLE_SIZE 16
+#define TXE_NUM_32_BIT_COUNTER 7
+#define TXE_NUM_64_BIT_COUNTER 30
+#define TXE_NUM_DATA_VL 8
+#define TXE_PIO_SIZE (32 * 0x100000) /* 32 MB */
+#define PIO_BLOCK_SIZE 64 /* bytes */
+#define SDMA_BLOCK_SIZE 64 /* bytes */
+#define RCV_BUF_BLOCK_SIZE 64 /* bytes */
+#define PIO_CMASK 0x7ff /* counter mask for free and fill counters */
+#define MAX_EAGER_ENTRIES 2048 /* max receive eager entries */
+#define MAX_TID_PAIR_ENTRIES 1024 /* max receive expected pairs */
+/* Virtual? Allocation Unit, defined as AU = 8*2^vAU, 64 bytes, AU is fixed
+ at 64 bytes for all generation one devices */
+#define CM_VAU 3
+/* HFI link credit count, AKA receive buffer depth (RBUF_DEPTH) */
+#define CM_GLOBAL_CREDITS 0x940
+/* Number of PKey entries in the HW */
+#define MAX_PKEY_VALUES 16
+
+#include "chip_registers.h"
+
+#define RXE_PER_CONTEXT_USER (RXE + RXE_PER_CONTEXT_OFFSET)
+#define TXE_PIO_SEND (TXE + TXE_PIO_SEND_OFFSET)
+
+/* PBC flags */
+#define PBC_INTR (1ull << 31)
+#define PBC_DC_INFO_SHIFT (30)
+#define PBC_DC_INFO (1ull << PBC_DC_INFO_SHIFT)
+#define PBC_TEST_EBP (1ull << 29)
+#define PBC_PACKET_BYPASS (1ull << 28)
+#define PBC_CREDIT_RETURN (1ull << 25)
+#define PBC_INSERT_BYPASS_ICRC (1ull << 24)
+#define PBC_TEST_BAD_ICRC (1ull << 23)
+#define PBC_FECN (1ull << 22)
+
+/* PbcInsertHcrc field settings */
+#define PBC_IHCRC_LKDETH 0x0 /* insert @ local KDETH offset */
+#define PBC_IHCRC_GKDETH 0x1 /* insert @ global KDETH offset */
+#define PBC_IHCRC_NONE 0x2 /* no HCRC inserted */
+
+/* PBC fields */
+#define PBC_STATIC_RATE_CONTROL_COUNT_SHIFT 32
+#define PBC_STATIC_RATE_CONTROL_COUNT_MASK 0xffffull
+#define PBC_STATIC_RATE_CONTROL_COUNT_SMASK \
+ (PBC_STATIC_RATE_CONTROL_COUNT_MASK << \
+ PBC_STATIC_RATE_CONTROL_COUNT_SHIFT)
+
+#define PBC_INSERT_HCRC_SHIFT 26
+#define PBC_INSERT_HCRC_MASK 0x3ull
+#define PBC_INSERT_HCRC_SMASK \
+ (PBC_INSERT_HCRC_MASK << PBC_INSERT_HCRC_SHIFT)
+
+#define PBC_VL_SHIFT 12
+#define PBC_VL_MASK 0xfull
+#define PBC_VL_SMASK (PBC_VL_MASK << PBC_VL_SHIFT)
+
+#define PBC_LENGTH_DWS_SHIFT 0
+#define PBC_LENGTH_DWS_MASK 0xfffull
+#define PBC_LENGTH_DWS_SMASK \
+ (PBC_LENGTH_DWS_MASK << PBC_LENGTH_DWS_SHIFT)
+
+/* Credit Return Fields */
+#define CR_COUNTER_SHIFT 0
+#define CR_COUNTER_MASK 0x7ffull
+#define CR_COUNTER_SMASK (CR_COUNTER_MASK << CR_COUNTER_SHIFT)
+
+#define CR_STATUS_SHIFT 11
+#define CR_STATUS_MASK 0x1ull
+#define CR_STATUS_SMASK (CR_STATUS_MASK << CR_STATUS_SHIFT)
+
+#define CR_CREDIT_RETURN_DUE_TO_PBC_SHIFT 12
+#define CR_CREDIT_RETURN_DUE_TO_PBC_MASK 0x1ull
+#define CR_CREDIT_RETURN_DUE_TO_PBC_SMASK \
+ (CR_CREDIT_RETURN_DUE_TO_PBC_MASK << \
+ CR_CREDIT_RETURN_DUE_TO_PBC_SHIFT)
+
+#define CR_CREDIT_RETURN_DUE_TO_THRESHOLD_SHIFT 13
+#define CR_CREDIT_RETURN_DUE_TO_THRESHOLD_MASK 0x1ull
+#define CR_CREDIT_RETURN_DUE_TO_THRESHOLD_SMASK \
+ (CR_CREDIT_RETURN_DUE_TO_THRESHOLD_MASK << \
+ CR_CREDIT_RETURN_DUE_TO_THRESHOLD_SHIFT)
+
+#define CR_CREDIT_RETURN_DUE_TO_ERR_SHIFT 14
+#define CR_CREDIT_RETURN_DUE_TO_ERR_MASK 0x1ull
+#define CR_CREDIT_RETURN_DUE_TO_ERR_SMASK \
+ (CR_CREDIT_RETURN_DUE_TO_ERR_MASK << \
+ CR_CREDIT_RETURN_DUE_TO_ERR_SHIFT)
+
+#define CR_CREDIT_RETURN_DUE_TO_FORCE_SHIFT 15
+#define CR_CREDIT_RETURN_DUE_TO_FORCE_MASK 0x1ull
+#define CR_CREDIT_RETURN_DUE_TO_FORCE_SMASK \
+ (CR_CREDIT_RETURN_DUE_TO_FORCE_MASK << \
+ CR_CREDIT_RETURN_DUE_TO_FORCE_SHIFT)
+
+/* interrupt source numbers */
+#define IS_GENERAL_ERR_START 0
+#define IS_SDMAENG_ERR_START 16
+#define IS_SENDCTXT_ERR_START 32
+#define IS_SDMA_START 192 /* includes SDmaProgress,SDmaIdle */
+#define IS_VARIOUS_START 240
+#define IS_DC_START 248
+#define IS_RCVAVAIL_START 256
+#define IS_RCVURGENT_START 416
+#define IS_SENDCREDIT_START 576
+#define IS_RESERVED_START 736
+#define IS_MAX_SOURCES 768
+
+/* derived interrupt source values */
+#define IS_GENERAL_ERR_END IS_SDMAENG_ERR_START
+#define IS_SDMAENG_ERR_END IS_SENDCTXT_ERR_START
+#define IS_SENDCTXT_ERR_END IS_SDMA_START
+#define IS_SDMA_END IS_VARIOUS_START
+#define IS_VARIOUS_END IS_DC_START
+#define IS_DC_END IS_RCVAVAIL_START
+#define IS_RCVAVAIL_END IS_RCVURGENT_START
+#define IS_RCVURGENT_END IS_SENDCREDIT_START
+#define IS_SENDCREDIT_END IS_RESERVED_START
+#define IS_RESERVED_END IS_MAX_SOURCES
+
+/* absolute interrupt numbers for QSFP1Int and QSFP2Int */
+#define QSFP1_INT 242
+#define QSFP2_INT 243
+
+/* DCC_CFG_PORT_CONFIG logical link states */
+#define LSTATE_DOWN 0x1
+#define LSTATE_INIT 0x2
+#define LSTATE_ARMED 0x3
+#define LSTATE_ACTIVE 0x4
+
+/* DC8051_STS_CUR_STATE port values (physical link states) */
+#define PLS_DISABLED 0x30
+#define PLS_OFFLINE 0x90
+#define PLS_OFFLINE_QUIET 0x90
+#define PLS_OFFLINE_PLANNED_DOWN_INFORM 0x91
+#define PLS_OFFLINE_READY_TO_QUIET_LT 0x92
+#define PLS_OFFLINE_REPORT_FAILURE 0x93
+#define PLS_OFFLINE_READY_TO_QUIET_BCC 0x94
+#define PLS_POLLING 0x20
+#define PLS_POLLING_QUIET 0x20
+#define PLS_POLLING_ACTIVE 0x21
+#define PLS_CONFIGPHY 0x40
+#define PLS_CONFIGPHY_DEBOUCE 0x40
+#define PLS_CONFIGPHY_ESTCOMM 0x41
+#define PLS_CONFIGPHY_ESTCOMM_TXRX_HUNT 0x42
+#define PLS_CONFIGPHY_ESTcOMM_LOCAL_COMPLETE 0x43
+#define PLS_CONFIGPHY_OPTEQ 0x44
+#define PLS_CONFIGPHY_OPTEQ_OPTIMIZING 0x44
+#define PLS_CONFIGPHY_OPTEQ_LOCAL_COMPLETE 0x45
+#define PLS_CONFIGPHY_VERIFYCAP 0x46
+#define PLS_CONFIGPHY_VERIFYCAP_EXCHANGE 0x46
+#define PLS_CONFIGPHY_VERIFYCAP_LOCAL_COMPLETE 0x47
+#define PLS_CONFIGLT 0x48
+#define PLS_CONFIGLT_CONFIGURE 0x48
+#define PLS_CONFIGLT_LINK_TRANSFER_ACTIVE 0x49
+#define PLS_LINKUP 0x50
+#define PLS_PHYTEST 0xB0
+#define PLS_INTERNAL_SERDES_LOOPBACK 0xe1
+#define PLS_QUICK_LINKUP 0xe2
+
+/* DC_DC8051_CFG_HOST_CMD_0.REQ_TYPE - 8051 host commands */
+#define HCMD_LOAD_CONFIG_DATA 0x01
+#define HCMD_READ_CONFIG_DATA 0x02
+#define HCMD_CHANGE_PHY_STATE 0x03
+#define HCMD_SEND_LCB_IDLE_MSG 0x04
+#define HCMD_MISC 0x05
+#define HCMD_READ_LCB_IDLE_MSG 0x06
+#define HCMD_READ_LCB_CSR 0x07
+#define HCMD_INTERFACE_TEST 0xff
+
+/* DC_DC8051_CFG_HOST_CMD_1.RETURN_CODE - 8051 host command return */
+#define HCMD_SUCCESS 2
+
+/* DC_DC8051_DBG_ERR_INFO_SET_BY_8051.ERROR - error flags */
+#define SPICO_ROM_FAILED (1 << 0)
+#define UNKNOWN_FRAME (1 << 1)
+#define TARGET_BER_NOT_MET (1 << 2)
+#define FAILED_SERDES_INTERNAL_LOOPBACK (1 << 3)
+#define FAILED_SERDES_INIT (1 << 4)
+#define FAILED_LNI_POLLING (1 << 5)
+#define FAILED_LNI_DEBOUNCE (1 << 6)
+#define FAILED_LNI_ESTBCOMM (1 << 7)
+#define FAILED_LNI_OPTEQ (1 << 8)
+#define FAILED_LNI_VERIFY_CAP1 (1 << 9)
+#define FAILED_LNI_VERIFY_CAP2 (1 << 10)
+#define FAILED_LNI_CONFIGLT (1 << 11)
+
+#define FAILED_LNI (FAILED_LNI_POLLING | FAILED_LNI_DEBOUNCE \
+ | FAILED_LNI_ESTBCOMM | FAILED_LNI_OPTEQ \
+ | FAILED_LNI_VERIFY_CAP1 \
+ | FAILED_LNI_VERIFY_CAP2 \
+ | FAILED_LNI_CONFIGLT)
+
+/* DC_DC8051_DBG_ERR_INFO_SET_BY_8051.HOST_MSG - host message flags */
+#define HOST_REQ_DONE (1 << 0)
+#define BC_PWR_MGM_MSG (1 << 1)
+#define BC_SMA_MSG (1 << 2)
+#define BC_BCC_UNKOWN_MSG (1 << 3)
+#define BC_IDLE_UNKNOWN_MSG (1 << 4)
+#define EXT_DEVICE_CFG_REQ (1 << 5)
+#define VERIFY_CAP_FRAME (1 << 6)
+#define LINKUP_ACHIEVED (1 << 7)
+#define LINK_GOING_DOWN (1 << 8)
+#define LINK_WIDTH_DOWNGRADED (1 << 9)
+
+/* DC_DC8051_CFG_EXT_DEV_1.REQ_TYPE - 8051 host requests */
+#define HREQ_LOAD_CONFIG 0x01
+#define HREQ_SAVE_CONFIG 0x02
+#define HREQ_READ_CONFIG 0x03
+#define HREQ_SET_TX_EQ_ABS 0x04
+#define HREQ_SET_TX_EQ_REL 0x05
+#define HREQ_ENABLE 0x06
+#define HREQ_CONFIG_DONE 0xfe
+#define HREQ_INTERFACE_TEST 0xff
+
+/* DC_DC8051_CFG_EXT_DEV_0.RETURN_CODE - 8051 host request return codes */
+#define HREQ_INVALID 0x01
+#define HREQ_SUCCESS 0x02
+#define HREQ_NOT_SUPPORTED 0x03
+#define HREQ_FEATURE_NOT_SUPPORTED 0x04 /* request specific feature */
+#define HREQ_REQUEST_REJECTED 0xfe
+#define HREQ_EXECUTION_ONGOING 0xff
+
+/* MISC host command functions */
+#define HCMD_MISC_REQUEST_LCB_ACCESS 0x1
+#define HCMD_MISC_GRANT_LCB_ACCESS 0x2
+
+/* idle flit message types */
+#define IDLE_PHYSICAL_LINK_MGMT 0x1
+#define IDLE_CRU 0x2
+#define IDLE_SMA 0x3
+#define IDLE_POWER_MGMT 0x4
+
+/* idle flit message send fields (both send and read) */
+#define IDLE_PAYLOAD_MASK 0xffffffffffull /* 40 bits */
+#define IDLE_PAYLOAD_SHIFT 8
+#define IDLE_MSG_TYPE_MASK 0xf
+#define IDLE_MSG_TYPE_SHIFT 0
+
+/* idle flit message read fields */
+#define READ_IDLE_MSG_TYPE_MASK 0xf
+#define READ_IDLE_MSG_TYPE_SHIFT 0
+
+/* SMA idle flit payload commands */
+#define SMA_IDLE_ARM 1
+#define SMA_IDLE_ACTIVE 2
+
+/* DC_DC8051_CFG_MODE.GENERAL bits */
+#define DISABLE_SELF_GUID_CHECK 0x2
+
+/*
+ * Eager buffer minimum and maximum sizes supported by the hardware.
+ * All power-of-two sizes in between are supported as well.
+ * MAX_EAGER_BUFFER_TOTAL is the maximum size of memory
+ * allocatable for Eager buffer to a single context. All others
+ * are limits for the RcvArray entries.
+ */
+#define MIN_EAGER_BUFFER (4 * 1024)
+#define MAX_EAGER_BUFFER (256 * 1024)
+#define MAX_EAGER_BUFFER_TOTAL (64 * (1 << 20)) /* max per ctxt 64MB */
+#define MAX_EXPECTED_BUFFER (2048 * 1024)
+
+/*
+ * Receive expected base and count and eager base and count increment -
+ * the CSR fields hold multiples of this value.
+ */
+#define RCV_SHIFT 3
+#define RCV_INCREMENT (1 << RCV_SHIFT)
+
+/*
+ * Receive header queue entry increment - the CSR holds multiples of
+ * this value.
+ */
+#define HDRQ_SIZE_SHIFT 5
+#define HDRQ_INCREMENT (1 << HDRQ_SIZE_SHIFT)
+
+/*
+ * Freeze handling flags
+ */
+#define FREEZE_ABORT 0x01 /* do not do recovery */
+#define FREEZE_SELF 0x02 /* initiate the freeze */
+#define FREEZE_LINK_DOWN 0x04 /* link is down */
+
+/*
+ * Chip implementation codes.
+ */
+#define ICODE_RTL_SILICON 0x00
+#define ICODE_RTL_VCS_SIMULATION 0x01
+#define ICODE_FPGA_EMULATION 0x02
+#define ICODE_FUNCTIONAL_SIMULATOR 0x03
+
+/*
+ * 8051 data memory size.
+ */
+#define DC8051_DATA_MEM_SIZE 0x1000
+
+/*
+ * 8051 firmware registers
+ */
+#define NUM_GENERAL_FIELDS 0x17
+#define NUM_LANE_FIELDS 0x8
+
+/* 8051 general register Field IDs */
+#define TX_SETTINGS 0x06
+#define VERIFY_CAP_LOCAL_PHY 0x07
+#define VERIFY_CAP_LOCAL_FABRIC 0x08
+#define VERIFY_CAP_LOCAL_LINK_WIDTH 0x09
+#define LOCAL_DEVICE_ID 0x0a
+#define LOCAL_LNI_INFO 0x0c
+#define REMOTE_LNI_INFO 0x0d
+#define MISC_STATUS 0x0e
+#define VERIFY_CAP_REMOTE_PHY 0x0f
+#define VERIFY_CAP_REMOTE_FABRIC 0x10
+#define VERIFY_CAP_REMOTE_LINK_WIDTH 0x11
+#define LAST_LOCAL_STATE_COMPLETE 0x12
+#define LAST_REMOTE_STATE_COMPLETE 0x13
+#define LINK_QUALITY_INFO 0x14
+#define REMOTE_DEVICE_ID 0x15
+
+/* Lane ID for general configuration registers */
+#define GENERAL_CONFIG 4
+
+/* LOAD_DATA 8051 command shifts and fields */
+#define LOAD_DATA_FIELD_ID_SHIFT 40
+#define LOAD_DATA_FIELD_ID_MASK 0xfull
+#define LOAD_DATA_LANE_ID_SHIFT 32
+#define LOAD_DATA_LANE_ID_MASK 0xfull
+#define LOAD_DATA_DATA_SHIFT 0x0
+#define LOAD_DATA_DATA_MASK 0xffffffffull
+
+/* READ_DATA 8051 command shifts and fields */
+#define READ_DATA_FIELD_ID_SHIFT 40
+#define READ_DATA_FIELD_ID_MASK 0xffull
+#define READ_DATA_LANE_ID_SHIFT 32
+#define READ_DATA_LANE_ID_MASK 0xffull
+#define READ_DATA_DATA_SHIFT 0x0
+#define READ_DATA_DATA_MASK 0xffffffffull
+
+/* TX settings fields */
+#define ENABLE_LANE_TX_SHIFT 0
+#define ENABLE_LANE_TX_MASK 0xff
+#define TX_POLARITY_INVERSION_SHIFT 8
+#define TX_POLARITY_INVERSION_MASK 0xff
+#define RX_POLARITY_INVERSION_SHIFT 16
+#define RX_POLARITY_INVERSION_MASK 0xff
+#define MAX_RATE_SHIFT 24
+#define MAX_RATE_MASK 0xff
+
+/* verify capability PHY fields */
+#define CONTINIOUS_REMOTE_UPDATE_SUPPORT_SHIFT 0x4
+#define CONTINIOUS_REMOTE_UPDATE_SUPPORT_MASK 0x1
+#define POWER_MANAGEMENT_SHIFT 0x0
+#define POWER_MANAGEMENT_MASK 0xf
+
+/* 8051 lane register Field IDs */
+#define SPICO_FW_VERSION 0x7 /* SPICO firmware version */
+
+/* SPICO firmware version fields */
+#define SPICO_ROM_VERSION_SHIFT 0
+#define SPICO_ROM_VERSION_MASK 0xffff
+#define SPICO_ROM_PROD_ID_SHIFT 16
+#define SPICO_ROM_PROD_ID_MASK 0xffff
+
+/* verify capability fabric fields */
+#define VAU_SHIFT 0
+#define VAU_MASK 0x0007
+#define Z_SHIFT 3
+#define Z_MASK 0x0001
+#define VCU_SHIFT 4
+#define VCU_MASK 0x0007
+#define VL15BUF_SHIFT 8
+#define VL15BUF_MASK 0x0fff
+#define CRC_SIZES_SHIFT 20
+#define CRC_SIZES_MASK 0x7
+
+/* verify capability local link width fields */
+#define LINK_WIDTH_SHIFT 0 /* also for remote link width */
+#define LINK_WIDTH_MASK 0xffff /* also for remote link width */
+#define LOCAL_FLAG_BITS_SHIFT 16
+#define LOCAL_FLAG_BITS_MASK 0xff
+#define MISC_CONFIG_BITS_SHIFT 24
+#define MISC_CONFIG_BITS_MASK 0xff
+
+/* verify capability remote link width fields */
+#define REMOTE_TX_RATE_SHIFT 16
+#define REMOTE_TX_RATE_MASK 0xff
+
+/* LOCAL_DEVICE_ID fields */
+#define LOCAL_DEVICE_REV_SHIFT 0
+#define LOCAL_DEVICE_REV_MASK 0xff
+#define LOCAL_DEVICE_ID_SHIFT 8
+#define LOCAL_DEVICE_ID_MASK 0xffff
+
+/* REMOTE_DEVICE_ID fields */
+#define REMOTE_DEVICE_REV_SHIFT 0
+#define REMOTE_DEVICE_REV_MASK 0xff
+#define REMOTE_DEVICE_ID_SHIFT 8
+#define REMOTE_DEVICE_ID_MASK 0xffff
+
+/* local LNI link width fields */
+#define ENABLE_LANE_RX_SHIFT 16
+#define ENABLE_LANE_RX_MASK 0xff
+
+/* mask, shift for reading 'mgmt_enabled' value from REMOTE_LNI_INFO field */
+#define MGMT_ALLOWED_SHIFT 23
+#define MGMT_ALLOWED_MASK 0x1
+
+/* mask, shift for 'link_quality' within LINK_QUALITY_INFO field */
+#define LINK_QUALITY_SHIFT 24
+#define LINK_QUALITY_MASK 0x7
+
+/*
+ * mask, shift for reading 'planned_down_remote_reason_code'
+ * from LINK_QUALITY_INFO field
+ */
+#define DOWN_REMOTE_REASON_SHIFT 16
+#define DOWN_REMOTE_REASON_MASK 0xff
+
+/* verify capability PHY power management bits */
+#define PWRM_BER_CONTROL 0x1
+#define PWRM_BANDWIDTH_CONTROL 0x2
+
+/* verify capability fabric CRC size bits */
+enum {
+ CAP_CRC_14B = (1 << 0), /* 14b CRC */
+ CAP_CRC_48B = (1 << 1), /* 48b CRC */
+ CAP_CRC_12B_16B_PER_LANE = (1 << 2) /* 12b-16b per lane CRC */
+};
+
+#define SUPPORTED_CRCS (CAP_CRC_14B | CAP_CRC_48B)
+
+/* misc status version fields */
+#define STS_FM_VERSION_A_SHIFT 16
+#define STS_FM_VERSION_A_MASK 0xff
+#define STS_FM_VERSION_B_SHIFT 24
+#define STS_FM_VERSION_B_MASK 0xff
+
+/* LCB_CFG_CRC_MODE TX_VAL and RX_VAL CRC mode values */
+#define LCB_CRC_16B 0x0 /* 16b CRC */
+#define LCB_CRC_14B 0x1 /* 14b CRC */
+#define LCB_CRC_48B 0x2 /* 48b CRC */
+#define LCB_CRC_12B_16B_PER_LANE 0x3 /* 12b-16b per lane CRC */
+
+/* the following enum is (almost) a copy/paste of the definition
+ * in the OPA spec, section 20.2.2.6.8 (PortInfo) */
+enum {
+ PORT_LTP_CRC_MODE_NONE = 0,
+ PORT_LTP_CRC_MODE_14 = 1, /* 14-bit LTP CRC mode (optional) */
+ PORT_LTP_CRC_MODE_16 = 2, /* 16-bit LTP CRC mode */
+ PORT_LTP_CRC_MODE_48 = 4,
+ /* 48-bit overlapping LTP CRC mode (optional) */
+ PORT_LTP_CRC_MODE_PER_LANE = 8
+ /* 12 to 16 bit per lane LTP CRC mode (optional) */
+};
+
+/* timeouts */
+#define LINK_RESTART_DELAY 1000 /* link restart delay, in ms */
+#define TIMEOUT_8051_START 5000 /* 8051 start timeout, in ms */
+#define DC8051_COMMAND_TIMEOUT 20000 /* DC8051 command timeout, in ms */
+#define FREEZE_STATUS_TIMEOUT 20 /* wait for freeze indicators, in ms */
+#define VL_STATUS_CLEAR_TIMEOUT 5000 /* per-VL status clear, in ms */
+#define CCE_STATUS_TIMEOUT 10 /* time to clear CCE Status, in ms */
+
+/* cclock tick time, in picoseconds per tick: 1/speed * 10^12 */
+#define ASIC_CCLOCK_PS 1242 /* 805 MHz */
+#define FPGA_CCLOCK_PS 30300 /* 33 MHz */
+
+/*
+ * Mask of enabled MISC errors. Do not enable the two RSA engine errors -
+ * see firmware.c:run_rsa() for details.
+ */
+#define DRIVER_MISC_MASK \
+ (~(MISC_ERR_STATUS_MISC_FW_AUTH_FAILED_ERR_SMASK \
+ | MISC_ERR_STATUS_MISC_KEY_MISMATCH_ERR_SMASK))
+
+/* valid values for the loopback module parameter */
+#define LOOPBACK_NONE 0 /* no loopback - default */
+#define LOOPBACK_SERDES 1
+#define LOOPBACK_LCB 2
+#define LOOPBACK_CABLE 3 /* external cable */
+
+/* read and write hardware registers */
+u64 read_csr(const struct hfi1_devdata *dd, u32 offset);
+void write_csr(const struct hfi1_devdata *dd, u32 offset, u64 value);
+
+/*
+ * The *_kctxt_* flavor of the CSR read/write functions are for
+ * per-context or per-SDMA CSRs that are not mappable to user-space.
+ * Their spacing is not a PAGE_SIZE multiple.
+ */
+static inline u64 read_kctxt_csr(const struct hfi1_devdata *dd, int ctxt,
+ u32 offset0)
+{
+ /* kernel per-context CSRs are separated by 0x100 */
+ return read_csr(dd, offset0 + (0x100 * ctxt));
+}
+
+static inline void write_kctxt_csr(struct hfi1_devdata *dd, int ctxt,
+ u32 offset0, u64 value)
+{
+ /* kernel per-context CSRs are separated by 0x100 */
+ write_csr(dd, offset0 + (0x100 * ctxt), value);
+}
+
+int read_lcb_csr(struct hfi1_devdata *dd, u32 offset, u64 *data);
+int write_lcb_csr(struct hfi1_devdata *dd, u32 offset, u64 data);
+
+void __iomem *get_csr_addr(
+ struct hfi1_devdata *dd,
+ u32 offset);
+
+static inline void __iomem *get_kctxt_csr_addr(
+ struct hfi1_devdata *dd,
+ int ctxt,
+ u32 offset0)
+{
+ return get_csr_addr(dd, offset0 + (0x100 * ctxt));
+}
+
+/*
+ * The *_uctxt_* flavor of the CSR read/write functions are for
+ * per-context CSRs that are mappable to user space. All these CSRs
+ * are spaced by a PAGE_SIZE multiple in order to be mappable to
+ * different processes without exposing other contexts' CSRs
+ */
+static inline u64 read_uctxt_csr(const struct hfi1_devdata *dd, int ctxt,
+ u32 offset0)
+{
+ /* user per-context CSRs are separated by 0x1000 */
+ return read_csr(dd, offset0 + (0x1000 * ctxt));
+}
+
+static inline void write_uctxt_csr(struct hfi1_devdata *dd, int ctxt,
+ u32 offset0, u64 value)
+{
+ /* user per-context CSRs are separated by 0x1000 */
+ write_csr(dd, offset0 + (0x1000 * ctxt), value);
+}
+
+u64 create_pbc(struct hfi1_pportdata *ppd, u64, int, u32, u32);
+
+/* firmware.c */
+#define NUM_PCIE_SERDES 16 /* number of PCIe serdes on the SBus */
+extern const u8 pcie_serdes_broadcast[];
+extern const u8 pcie_pcs_addrs[2][NUM_PCIE_SERDES];
+/* SBus commands */
+#define RESET_SBUS_RECEIVER 0x20
+#define WRITE_SBUS_RECEIVER 0x21
+void sbus_request(struct hfi1_devdata *dd,
+ u8 receiver_addr, u8 data_addr, u8 command, u32 data_in);
+int sbus_request_slow(struct hfi1_devdata *dd,
+ u8 receiver_addr, u8 data_addr, u8 command, u32 data_in);
+void set_sbus_fast_mode(struct hfi1_devdata *dd);
+void clear_sbus_fast_mode(struct hfi1_devdata *dd);
+int hfi1_firmware_init(struct hfi1_devdata *dd);
+int load_pcie_firmware(struct hfi1_devdata *dd);
+int load_firmware(struct hfi1_devdata *dd);
+void dispose_firmware(void);
+int acquire_hw_mutex(struct hfi1_devdata *dd);
+void release_hw_mutex(struct hfi1_devdata *dd);
+void fabric_serdes_reset(struct hfi1_devdata *dd);
+int read_8051_data(struct hfi1_devdata *dd, u32 addr, u32 len, u64 *result);
+
+/* chip.c */
+void read_misc_status(struct hfi1_devdata *dd, u8 *ver_a, u8 *ver_b);
+void read_guid(struct hfi1_devdata *dd);
+int wait_fm_ready(struct hfi1_devdata *dd, u32 mstimeout);
+void set_link_down_reason(struct hfi1_pportdata *ppd, u8 lcl_reason,
+ u8 neigh_reason, u8 rem_reason);
+int set_link_state(struct hfi1_pportdata *, u32 state);
+int port_ltp_to_cap(int port_ltp);
+void handle_verify_cap(struct work_struct *work);
+void handle_freeze(struct work_struct *work);
+void handle_link_up(struct work_struct *work);
+void handle_link_down(struct work_struct *work);
+void handle_link_downgrade(struct work_struct *work);
+void handle_link_bounce(struct work_struct *work);
+void handle_sma_message(struct work_struct *work);
+void start_freeze_handling(struct hfi1_pportdata *ppd, int flags);
+int send_idle_sma(struct hfi1_devdata *dd, u64 message);
+int start_link(struct hfi1_pportdata *ppd);
+void init_qsfp(struct hfi1_pportdata *ppd);
+int bringup_serdes(struct hfi1_pportdata *ppd);
+void set_intr_state(struct hfi1_devdata *dd, u32 enable);
+void apply_link_downgrade_policy(struct hfi1_pportdata *ppd,
+ int refresh_widths);
+void update_usrhead(struct hfi1_ctxtdata *, u32, u32, u32, u32, u32);
+int stop_drain_data_vls(struct hfi1_devdata *dd);
+int open_fill_data_vls(struct hfi1_devdata *dd);
+u32 ns_to_cclock(struct hfi1_devdata *dd, u32 ns);
+u32 cclock_to_ns(struct hfi1_devdata *dd, u32 cclock);
+void get_linkup_link_widths(struct hfi1_pportdata *ppd);
+void read_ltp_rtt(struct hfi1_devdata *dd);
+void clear_linkup_counters(struct hfi1_devdata *dd);
+u32 hdrqempty(struct hfi1_ctxtdata *rcd);
+int is_a0(struct hfi1_devdata *dd);
+int is_ax(struct hfi1_devdata *dd);
+int is_bx(struct hfi1_devdata *dd);
+u32 read_physical_state(struct hfi1_devdata *dd);
+u32 chip_to_opa_pstate(struct hfi1_devdata *dd, u32 chip_pstate);
+u32 get_logical_state(struct hfi1_pportdata *ppd);
+const char *opa_lstate_name(u32 lstate);
+const char *opa_pstate_name(u32 pstate);
+u32 driver_physical_state(struct hfi1_pportdata *ppd);
+u32 driver_logical_state(struct hfi1_pportdata *ppd);
+
+int acquire_lcb_access(struct hfi1_devdata *dd, int sleep_ok);
+int release_lcb_access(struct hfi1_devdata *dd, int sleep_ok);
+#define LCB_START DC_LCB_CSRS
+#define LCB_END DC_8051_CSRS /* next block is 8051 */
+static inline int is_lcb_offset(u32 offset)
+{
+ return (offset >= LCB_START && offset < LCB_END);
+}
+
+extern uint num_vls;
+
+extern uint disable_integrity;
+u64 read_dev_cntr(struct hfi1_devdata *dd, int index, int vl);
+u64 write_dev_cntr(struct hfi1_devdata *dd, int index, int vl, u64 data);
+u64 read_port_cntr(struct hfi1_pportdata *ppd, int index, int vl);
+u64 write_port_cntr(struct hfi1_pportdata *ppd, int index, int vl, u64 data);
+
+/* Per VL indexes */
+enum {
+ C_VL_0 = 0,
+ C_VL_1,
+ C_VL_2,
+ C_VL_3,
+ C_VL_4,
+ C_VL_5,
+ C_VL_6,
+ C_VL_7,
+ C_VL_15,
+ C_VL_COUNT
+};
+
+static inline int vl_from_idx(int idx)
+{
+ return (idx == C_VL_15 ? 15 : idx);
+}
+
+static inline int idx_from_vl(int vl)
+{
+ return (vl == 15 ? C_VL_15 : vl);
+}
+
+/* Per device counter indexes */
+enum {
+ C_RCV_OVF = 0,
+ C_RX_TID_FULL,
+ C_RX_TID_INVALID,
+ C_RX_TID_FLGMS,
+ C_RX_CTX_RHQS,
+ C_RX_CTX_EGRS,
+ C_RCV_TID_FLSMS,
+ C_CCE_PCI_CR_ST,
+ C_CCE_PCI_TR_ST,
+ C_CCE_PIO_WR_ST,
+ C_CCE_ERR_INT,
+ C_CCE_SDMA_INT,
+ C_CCE_MISC_INT,
+ C_CCE_RCV_AV_INT,
+ C_CCE_RCV_URG_INT,
+ C_CCE_SEND_CR_INT,
+ C_DC_UNC_ERR,
+ C_DC_RCV_ERR,
+ C_DC_FM_CFG_ERR,
+ C_DC_RMT_PHY_ERR,
+ C_DC_DROPPED_PKT,
+ C_DC_MC_XMIT_PKTS,
+ C_DC_MC_RCV_PKTS,
+ C_DC_XMIT_CERR,
+ C_DC_RCV_CERR,
+ C_DC_RCV_FCC,
+ C_DC_XMIT_FCC,
+ C_DC_XMIT_FLITS,
+ C_DC_RCV_FLITS,
+ C_DC_XMIT_PKTS,
+ C_DC_RCV_PKTS,
+ C_DC_RX_FLIT_VL,
+ C_DC_RX_PKT_VL,
+ C_DC_RCV_FCN,
+ C_DC_RCV_FCN_VL,
+ C_DC_RCV_BCN,
+ C_DC_RCV_BCN_VL,
+ C_DC_RCV_BBL,
+ C_DC_RCV_BBL_VL,
+ C_DC_MARK_FECN,
+ C_DC_MARK_FECN_VL,
+ C_DC_TOTAL_CRC,
+ C_DC_CRC_LN0,
+ C_DC_CRC_LN1,
+ C_DC_CRC_LN2,
+ C_DC_CRC_LN3,
+ C_DC_CRC_MULT_LN,
+ C_DC_TX_REPLAY,
+ C_DC_RX_REPLAY,
+ C_DC_SEQ_CRC_CNT,
+ C_DC_ESC0_ONLY_CNT,
+ C_DC_ESC0_PLUS1_CNT,
+ C_DC_ESC0_PLUS2_CNT,
+ C_DC_REINIT_FROM_PEER_CNT,
+ C_DC_SBE_CNT,
+ C_DC_MISC_FLG_CNT,
+ C_DC_PRF_GOOD_LTP_CNT,
+ C_DC_PRF_ACCEPTED_LTP_CNT,
+ C_DC_PRF_RX_FLIT_CNT,
+ C_DC_PRF_TX_FLIT_CNT,
+ C_DC_PRF_CLK_CNTR,
+ C_DC_PG_DBG_FLIT_CRDTS_CNT,
+ C_DC_PG_STS_PAUSE_COMPLETE_CNT,
+ C_DC_PG_STS_TX_SBE_CNT,
+ C_DC_PG_STS_TX_MBE_CNT,
+ C_SW_CPU_INTR,
+ C_SW_CPU_RCV_LIM,
+ C_SW_VTX_WAIT,
+ C_SW_PIO_WAIT,
+ C_SW_KMEM_WAIT,
+ DEV_CNTR_LAST /* Must be kept last */
+};
+
+/* Per port counter indexes */
+enum {
+ C_TX_UNSUP_VL = 0,
+ C_TX_INVAL_LEN,
+ C_TX_MM_LEN_ERR,
+ C_TX_UNDERRUN,
+ C_TX_FLOW_STALL,
+ C_TX_DROPPED,
+ C_TX_HDR_ERR,
+ C_TX_PKT,
+ C_TX_WORDS,
+ C_TX_WAIT,
+ C_TX_FLIT_VL,
+ C_TX_PKT_VL,
+ C_TX_WAIT_VL,
+ C_RX_PKT,
+ C_RX_WORDS,
+ C_SW_LINK_DOWN,
+ C_SW_LINK_UP,
+ C_SW_XMIT_DSCD,
+ C_SW_XMIT_DSCD_VL,
+ C_SW_XMIT_CSTR_ERR,
+ C_SW_RCV_CSTR_ERR,
+ C_SW_IBP_LOOP_PKTS,
+ C_SW_IBP_RC_RESENDS,
+ C_SW_IBP_RNR_NAKS,
+ C_SW_IBP_OTHER_NAKS,
+ C_SW_IBP_RC_TIMEOUTS,
+ C_SW_IBP_PKT_DROPS,
+ C_SW_IBP_DMA_WAIT,
+ C_SW_IBP_RC_SEQNAK,
+ C_SW_IBP_RC_DUPREQ,
+ C_SW_IBP_RDMA_SEQ,
+ C_SW_IBP_UNALIGNED,
+ C_SW_IBP_SEQ_NAK,
+ C_SW_CPU_RC_ACKS,
+ C_SW_CPU_RC_QACKS,
+ C_SW_CPU_RC_DELAYED_COMP,
+ C_RCV_HDR_OVF_0,
+ C_RCV_HDR_OVF_1,
+ C_RCV_HDR_OVF_2,
+ C_RCV_HDR_OVF_3,
+ C_RCV_HDR_OVF_4,
+ C_RCV_HDR_OVF_5,
+ C_RCV_HDR_OVF_6,
+ C_RCV_HDR_OVF_7,
+ C_RCV_HDR_OVF_8,
+ C_RCV_HDR_OVF_9,
+ C_RCV_HDR_OVF_10,
+ C_RCV_HDR_OVF_11,
+ C_RCV_HDR_OVF_12,
+ C_RCV_HDR_OVF_13,
+ C_RCV_HDR_OVF_14,
+ C_RCV_HDR_OVF_15,
+ C_RCV_HDR_OVF_16,
+ C_RCV_HDR_OVF_17,
+ C_RCV_HDR_OVF_18,
+ C_RCV_HDR_OVF_19,
+ C_RCV_HDR_OVF_20,
+ C_RCV_HDR_OVF_21,
+ C_RCV_HDR_OVF_22,
+ C_RCV_HDR_OVF_23,
+ C_RCV_HDR_OVF_24,
+ C_RCV_HDR_OVF_25,
+ C_RCV_HDR_OVF_26,
+ C_RCV_HDR_OVF_27,
+ C_RCV_HDR_OVF_28,
+ C_RCV_HDR_OVF_29,
+ C_RCV_HDR_OVF_30,
+ C_RCV_HDR_OVF_31,
+ C_RCV_HDR_OVF_32,
+ C_RCV_HDR_OVF_33,
+ C_RCV_HDR_OVF_34,
+ C_RCV_HDR_OVF_35,
+ C_RCV_HDR_OVF_36,
+ C_RCV_HDR_OVF_37,
+ C_RCV_HDR_OVF_38,
+ C_RCV_HDR_OVF_39,
+ C_RCV_HDR_OVF_40,
+ C_RCV_HDR_OVF_41,
+ C_RCV_HDR_OVF_42,
+ C_RCV_HDR_OVF_43,
+ C_RCV_HDR_OVF_44,
+ C_RCV_HDR_OVF_45,
+ C_RCV_HDR_OVF_46,
+ C_RCV_HDR_OVF_47,
+ C_RCV_HDR_OVF_48,
+ C_RCV_HDR_OVF_49,
+ C_RCV_HDR_OVF_50,
+ C_RCV_HDR_OVF_51,
+ C_RCV_HDR_OVF_52,
+ C_RCV_HDR_OVF_53,
+ C_RCV_HDR_OVF_54,
+ C_RCV_HDR_OVF_55,
+ C_RCV_HDR_OVF_56,
+ C_RCV_HDR_OVF_57,
+ C_RCV_HDR_OVF_58,
+ C_RCV_HDR_OVF_59,
+ C_RCV_HDR_OVF_60,
+ C_RCV_HDR_OVF_61,
+ C_RCV_HDR_OVF_62,
+ C_RCV_HDR_OVF_63,
+ C_RCV_HDR_OVF_64,
+ C_RCV_HDR_OVF_65,
+ C_RCV_HDR_OVF_66,
+ C_RCV_HDR_OVF_67,
+ C_RCV_HDR_OVF_68,
+ C_RCV_HDR_OVF_69,
+ C_RCV_HDR_OVF_70,
+ C_RCV_HDR_OVF_71,
+ C_RCV_HDR_OVF_72,
+ C_RCV_HDR_OVF_73,
+ C_RCV_HDR_OVF_74,
+ C_RCV_HDR_OVF_75,
+ C_RCV_HDR_OVF_76,
+ C_RCV_HDR_OVF_77,
+ C_RCV_HDR_OVF_78,
+ C_RCV_HDR_OVF_79,
+ C_RCV_HDR_OVF_80,
+ C_RCV_HDR_OVF_81,
+ C_RCV_HDR_OVF_82,
+ C_RCV_HDR_OVF_83,
+ C_RCV_HDR_OVF_84,
+ C_RCV_HDR_OVF_85,
+ C_RCV_HDR_OVF_86,
+ C_RCV_HDR_OVF_87,
+ C_RCV_HDR_OVF_88,
+ C_RCV_HDR_OVF_89,
+ C_RCV_HDR_OVF_90,
+ C_RCV_HDR_OVF_91,
+ C_RCV_HDR_OVF_92,
+ C_RCV_HDR_OVF_93,
+ C_RCV_HDR_OVF_94,
+ C_RCV_HDR_OVF_95,
+ C_RCV_HDR_OVF_96,
+ C_RCV_HDR_OVF_97,
+ C_RCV_HDR_OVF_98,
+ C_RCV_HDR_OVF_99,
+ C_RCV_HDR_OVF_100,
+ C_RCV_HDR_OVF_101,
+ C_RCV_HDR_OVF_102,
+ C_RCV_HDR_OVF_103,
+ C_RCV_HDR_OVF_104,
+ C_RCV_HDR_OVF_105,
+ C_RCV_HDR_OVF_106,
+ C_RCV_HDR_OVF_107,
+ C_RCV_HDR_OVF_108,
+ C_RCV_HDR_OVF_109,
+ C_RCV_HDR_OVF_110,
+ C_RCV_HDR_OVF_111,
+ C_RCV_HDR_OVF_112,
+ C_RCV_HDR_OVF_113,
+ C_RCV_HDR_OVF_114,
+ C_RCV_HDR_OVF_115,
+ C_RCV_HDR_OVF_116,
+ C_RCV_HDR_OVF_117,
+ C_RCV_HDR_OVF_118,
+ C_RCV_HDR_OVF_119,
+ C_RCV_HDR_OVF_120,
+ C_RCV_HDR_OVF_121,
+ C_RCV_HDR_OVF_122,
+ C_RCV_HDR_OVF_123,
+ C_RCV_HDR_OVF_124,
+ C_RCV_HDR_OVF_125,
+ C_RCV_HDR_OVF_126,
+ C_RCV_HDR_OVF_127,
+ C_RCV_HDR_OVF_128,
+ C_RCV_HDR_OVF_129,
+ C_RCV_HDR_OVF_130,
+ C_RCV_HDR_OVF_131,
+ C_RCV_HDR_OVF_132,
+ C_RCV_HDR_OVF_133,
+ C_RCV_HDR_OVF_134,
+ C_RCV_HDR_OVF_135,
+ C_RCV_HDR_OVF_136,
+ C_RCV_HDR_OVF_137,
+ C_RCV_HDR_OVF_138,
+ C_RCV_HDR_OVF_139,
+ C_RCV_HDR_OVF_140,
+ C_RCV_HDR_OVF_141,
+ C_RCV_HDR_OVF_142,
+ C_RCV_HDR_OVF_143,
+ C_RCV_HDR_OVF_144,
+ C_RCV_HDR_OVF_145,
+ C_RCV_HDR_OVF_146,
+ C_RCV_HDR_OVF_147,
+ C_RCV_HDR_OVF_148,
+ C_RCV_HDR_OVF_149,
+ C_RCV_HDR_OVF_150,
+ C_RCV_HDR_OVF_151,
+ C_RCV_HDR_OVF_152,
+ C_RCV_HDR_OVF_153,
+ C_RCV_HDR_OVF_154,
+ C_RCV_HDR_OVF_155,
+ C_RCV_HDR_OVF_156,
+ C_RCV_HDR_OVF_157,
+ C_RCV_HDR_OVF_158,
+ C_RCV_HDR_OVF_159,
+ PORT_CNTR_LAST /* Must be kept last */
+};
+
+u64 get_all_cpu_total(u64 __percpu *cntr);
+void hfi1_start_cleanup(struct hfi1_devdata *dd);
+void hfi1_clear_tids(struct hfi1_ctxtdata *rcd);
+struct hfi1_message_header *hfi1_get_msgheader(
+ struct hfi1_devdata *dd, __le32 *rhf_addr);
+int hfi1_get_base_kinfo(struct hfi1_ctxtdata *rcd,
+ struct hfi1_ctxt_info *kinfo);
+u64 hfi1_gpio_mod(struct hfi1_devdata *dd, u32 target, u32 data, u32 dir,
+ u32 mask);
+int hfi1_init_ctxt(struct send_context *sc);
+void hfi1_put_tid(struct hfi1_devdata *dd, u32 index,
+ u32 type, unsigned long pa, u16 order);
+void hfi1_quiet_serdes(struct hfi1_pportdata *ppd);
+void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op, int ctxt);
+u32 hfi1_read_cntrs(struct hfi1_devdata *dd, loff_t pos, char **namep,
+ u64 **cntrp);
+u32 hfi1_read_portcntrs(struct hfi1_devdata *dd, loff_t pos, u32 port,
+ char **namep, u64 **cntrp);
+u8 hfi1_ibphys_portstate(struct hfi1_pportdata *ppd);
+int hfi1_get_ib_cfg(struct hfi1_pportdata *ppd, int which);
+int hfi1_set_ib_cfg(struct hfi1_pportdata *ppd, int which, u32 val);
+int hfi1_set_ctxt_jkey(struct hfi1_devdata *dd, unsigned ctxt, u16 jkey);
+int hfi1_clear_ctxt_jkey(struct hfi1_devdata *dd, unsigned ctxt);
+int hfi1_set_ctxt_pkey(struct hfi1_devdata *dd, unsigned ctxt, u16 pkey);
+int hfi1_clear_ctxt_pkey(struct hfi1_devdata *dd, unsigned ctxt);
+void hfi1_read_link_quality(struct hfi1_devdata *dd, u8 *link_quality);
+
+/*
+ * Interrupt source table.
+ *
+ * Each entry is an interrupt source "type". It is ordered by increasing
+ * number.
+ */
+struct is_table {
+ int start; /* interrupt source type start */
+ int end; /* interrupt source type end */
+ /* routine that returns the name of the interrupt source */
+ char *(*is_name)(char *name, size_t size, unsigned int source);
+ /* routine to call when receiving an interrupt */
+ void (*is_int)(struct hfi1_devdata *dd, unsigned int source);
+};
+
+#endif /* _CHIP_H */
diff --git a/drivers/staging/rdma/hfi1/chip_registers.h b/drivers/staging/rdma/hfi1/chip_registers.h
new file mode 100644
index 000000000000..6521030018d8
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/chip_registers.h
@@ -0,0 +1,1289 @@
+#ifndef DEF_CHIP_REG
+#define DEF_CHIP_REG
+
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#define CORE 0x000000000000
+#define CCE (CORE + 0x000000000000)
+#define ASIC (CORE + 0x000000400000)
+#define MISC (CORE + 0x000000500000)
+#define DC_TOP_CSRS (CORE + 0x000000600000)
+#define CHIP_DEBUG (CORE + 0x000000700000)
+#define RXE (CORE + 0x000001000000)
+#define TXE (CORE + 0x000001800000)
+#define DCC_CSRS (DC_TOP_CSRS + 0x000000000000)
+#define DC_LCB_CSRS (DC_TOP_CSRS + 0x000000001000)
+#define DC_8051_CSRS (DC_TOP_CSRS + 0x000000002000)
+#define PCIE 0
+
+#define ASIC_NUM_SCRATCH 4
+#define CCE_ERR_INT_CNT 0
+#define CCE_MISC_INT_CNT 2
+#define CCE_NUM_32_BIT_COUNTERS 3
+#define CCE_NUM_32_BIT_INT_COUNTERS 6
+#define CCE_NUM_INT_CSRS 12
+#define CCE_NUM_INT_MAP_CSRS 96
+#define CCE_NUM_MSIX_PBAS 4
+#define CCE_NUM_MSIX_VECTORS 256
+#define CCE_NUM_SCRATCH 4
+#define CCE_PCIE_POSTED_CRDT_STALL_CNT 2
+#define CCE_PCIE_TRGT_STALL_CNT 0
+#define CCE_PIO_WR_STALL_CNT 1
+#define CCE_RCV_AVAIL_INT_CNT 3
+#define CCE_RCV_URGENT_INT_CNT 4
+#define CCE_SDMA_INT_CNT 1
+#define CCE_SEND_CREDIT_INT_CNT 5
+#define DCC_CFG_LED_CNTRL (DCC_CSRS + 0x000000000040)
+#define DCC_CFG_LED_CNTRL_LED_CNTRL_SMASK 0x10ull
+#define DCC_CFG_LED_CNTRL_LED_SW_BLINK_RATE_SHIFT 0
+#define DCC_CFG_LED_CNTRL_LED_SW_BLINK_RATE_SMASK 0xFull
+#define DCC_CFG_PORT_CONFIG (DCC_CSRS + 0x000000000008)
+#define DCC_CFG_PORT_CONFIG1 (DCC_CSRS + 0x000000000010)
+#define DCC_CFG_PORT_CONFIG1_DLID_MASK_MASK 0xFFFFull
+#define DCC_CFG_PORT_CONFIG1_DLID_MASK_SHIFT 16
+#define DCC_CFG_PORT_CONFIG1_DLID_MASK_SMASK 0xFFFF0000ull
+#define DCC_CFG_PORT_CONFIG1_TARGET_DLID_MASK 0xFFFFull
+#define DCC_CFG_PORT_CONFIG1_TARGET_DLID_SHIFT 0
+#define DCC_CFG_PORT_CONFIG1_TARGET_DLID_SMASK 0xFFFFull
+#define DCC_CFG_PORT_CONFIG_LINK_STATE_MASK 0x7ull
+#define DCC_CFG_PORT_CONFIG_LINK_STATE_SHIFT 48
+#define DCC_CFG_PORT_CONFIG_LINK_STATE_SMASK 0x7000000000000ull
+#define DCC_CFG_PORT_CONFIG_MTU_CAP_MASK 0x7ull
+#define DCC_CFG_PORT_CONFIG_MTU_CAP_SHIFT 32
+#define DCC_CFG_PORT_CONFIG_MTU_CAP_SMASK 0x700000000ull
+#define DCC_CFG_RESET (DCC_CSRS + 0x000000000000)
+#define DCC_CFG_RESET_RESET_LCB_SHIFT 0
+#define DCC_CFG_RESET_RESET_RX_FPE_SHIFT 2
+#define DCC_CFG_SC_VL_TABLE_15_0 (DCC_CSRS + 0x000000000028)
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY0_SHIFT 0
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY10_SHIFT 40
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY11_SHIFT 44
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY12_SHIFT 48
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY13_SHIFT 52
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY14_SHIFT 56
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY15_SHIFT 60
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY1_SHIFT 4
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY2_SHIFT 8
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY3_SHIFT 12
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY4_SHIFT 16
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY5_SHIFT 20
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY6_SHIFT 24
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY7_SHIFT 28
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY8_SHIFT 32
+#define DCC_CFG_SC_VL_TABLE_15_0_ENTRY9_SHIFT 36
+#define DCC_CFG_SC_VL_TABLE_31_16 (DCC_CSRS + 0x000000000030)
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY16_SHIFT 0
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY17_SHIFT 4
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY18_SHIFT 8
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY19_SHIFT 12
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY20_SHIFT 16
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY21_SHIFT 20
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY22_SHIFT 24
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY23_SHIFT 28
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY24_SHIFT 32
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY25_SHIFT 36
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY26_SHIFT 40
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY27_SHIFT 44
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY28_SHIFT 48
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY29_SHIFT 52
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY30_SHIFT 56
+#define DCC_CFG_SC_VL_TABLE_31_16_ENTRY31_SHIFT 60
+#define DCC_ERR_DROPPED_PKT_CNT (DCC_CSRS + 0x000000000120)
+#define DCC_ERR_FLG (DCC_CSRS + 0x000000000050)
+#define DCC_ERR_FLG_BAD_CRDT_ACK_ERR_SMASK 0x4000ull
+#define DCC_ERR_FLG_BAD_CTRL_DIST_ERR_SMASK 0x200000ull
+#define DCC_ERR_FLG_BAD_CTRL_FLIT_ERR_SMASK 0x10000ull
+#define DCC_ERR_FLG_BAD_DLID_TARGET_ERR_SMASK 0x200ull
+#define DCC_ERR_FLG_BAD_HEAD_DIST_ERR_SMASK 0x800000ull
+#define DCC_ERR_FLG_BAD_L2_ERR_SMASK 0x2ull
+#define DCC_ERR_FLG_BAD_LVER_ERR_SMASK 0x400ull
+#define DCC_ERR_FLG_BAD_MID_TAIL_ERR_SMASK 0x8ull
+#define DCC_ERR_FLG_BAD_PKT_LENGTH_ERR_SMASK 0x4000000ull
+#define DCC_ERR_FLG_BAD_PREEMPTION_ERR_SMASK 0x10ull
+#define DCC_ERR_FLG_BAD_SC_ERR_SMASK 0x4ull
+#define DCC_ERR_FLG_BAD_TAIL_DIST_ERR_SMASK 0x400000ull
+#define DCC_ERR_FLG_BAD_VL_MARKER_ERR_SMASK 0x80ull
+#define DCC_ERR_FLG_CLR (DCC_CSRS + 0x000000000060)
+#define DCC_ERR_FLG_CSR_ACCESS_BLOCKED_HOST_SMASK 0x8000000000ull
+#define DCC_ERR_FLG_CSR_ACCESS_BLOCKED_UC_SMASK 0x10000000000ull
+#define DCC_ERR_FLG_CSR_INVAL_ADDR_SMASK 0x400000000000ull
+#define DCC_ERR_FLG_CSR_PARITY_ERR_SMASK 0x200000000000ull
+#define DCC_ERR_FLG_DLID_ZERO_ERR_SMASK 0x40000000ull
+#define DCC_ERR_FLG_EN (DCC_CSRS + 0x000000000058)
+#define DCC_ERR_FLG_EN_CSR_ACCESS_BLOCKED_HOST_SMASK 0x8000000000ull
+#define DCC_ERR_FLG_EN_CSR_ACCESS_BLOCKED_UC_SMASK 0x10000000000ull
+#define DCC_ERR_FLG_EVENT_CNTR_PARITY_ERR_SMASK 0x20000ull
+#define DCC_ERR_FLG_EVENT_CNTR_ROLLOVER_ERR_SMASK 0x40000ull
+#define DCC_ERR_FLG_FMCONFIG_ERR_SMASK 0x40000000000000ull
+#define DCC_ERR_FLG_FPE_TX_FIFO_OVFLW_ERR_SMASK 0x2000000000ull
+#define DCC_ERR_FLG_FPE_TX_FIFO_UNFLW_ERR_SMASK 0x4000000000ull
+#define DCC_ERR_FLG_LATE_EBP_ERR_SMASK 0x1000000000ull
+#define DCC_ERR_FLG_LATE_LONG_ERR_SMASK 0x800000000ull
+#define DCC_ERR_FLG_LATE_SHORT_ERR_SMASK 0x400000000ull
+#define DCC_ERR_FLG_LENGTH_MTU_ERR_SMASK 0x80000000ull
+#define DCC_ERR_FLG_LINK_ERR_SMASK 0x80000ull
+#define DCC_ERR_FLG_MISC_CNTR_ROLLOVER_ERR_SMASK 0x100000ull
+#define DCC_ERR_FLG_NONVL15_STATE_ERR_SMASK 0x1000000ull
+#define DCC_ERR_FLG_PERM_NVL15_ERR_SMASK 0x10000000ull
+#define DCC_ERR_FLG_PREEMPTION_ERR_SMASK 0x20ull
+#define DCC_ERR_FLG_PREEMPTIONVL15_ERR_SMASK 0x40ull
+#define DCC_ERR_FLG_RCVPORT_ERR_SMASK 0x80000000000000ull
+#define DCC_ERR_FLG_RX_BYTE_SHFT_PARITY_ERR_SMASK 0x1000000000000ull
+#define DCC_ERR_FLG_RX_CTRL_PARITY_MBE_ERR_SMASK 0x100000000000ull
+#define DCC_ERR_FLG_RX_EARLY_DROP_ERR_SMASK 0x200000000ull
+#define DCC_ERR_FLG_SLID_ZERO_ERR_SMASK 0x20000000ull
+#define DCC_ERR_FLG_TX_BYTE_SHFT_PARITY_ERR_SMASK 0x800000000000ull
+#define DCC_ERR_FLG_TX_CTRL_PARITY_ERR_SMASK 0x20000000000ull
+#define DCC_ERR_FLG_TX_CTRL_PARITY_MBE_ERR_SMASK 0x40000000000ull
+#define DCC_ERR_FLG_TX_SC_PARITY_ERR_SMASK 0x80000000000ull
+#define DCC_ERR_FLG_UNCORRECTABLE_ERR_SMASK 0x2000ull
+#define DCC_ERR_FLG_UNSUP_PKT_TYPE_SMASK 0x8000ull
+#define DCC_ERR_FLG_UNSUP_VL_ERR_SMASK 0x8000000ull
+#define DCC_ERR_FLG_VL15_MULTI_ERR_SMASK 0x2000000ull
+#define DCC_ERR_FMCONFIG_ERR_CNT (DCC_CSRS + 0x000000000110)
+#define DCC_ERR_INFO_FMCONFIG (DCC_CSRS + 0x000000000090)
+#define DCC_ERR_INFO_PORTRCV (DCC_CSRS + 0x000000000078)
+#define DCC_ERR_INFO_PORTRCV_HDR0 (DCC_CSRS + 0x000000000080)
+#define DCC_ERR_INFO_PORTRCV_HDR1 (DCC_CSRS + 0x000000000088)
+#define DCC_ERR_INFO_UNCORRECTABLE (DCC_CSRS + 0x000000000098)
+#define DCC_ERR_PORTRCV_ERR_CNT (DCC_CSRS + 0x000000000108)
+#define DCC_ERR_RCVREMOTE_PHY_ERR_CNT (DCC_CSRS + 0x000000000118)
+#define DCC_ERR_UNCORRECTABLE_CNT (DCC_CSRS + 0x000000000100)
+#define DCC_PRF_PORT_MARK_FECN_CNT (DCC_CSRS + 0x000000000330)
+#define DCC_PRF_PORT_RCV_BECN_CNT (DCC_CSRS + 0x000000000290)
+#define DCC_PRF_PORT_RCV_BUBBLE_CNT (DCC_CSRS + 0x0000000002E0)
+#define DCC_PRF_PORT_RCV_CORRECTABLE_CNT (DCC_CSRS + 0x000000000140)
+#define DCC_PRF_PORT_RCV_DATA_CNT (DCC_CSRS + 0x000000000198)
+#define DCC_PRF_PORT_RCV_FECN_CNT (DCC_CSRS + 0x000000000240)
+#define DCC_PRF_PORT_RCV_MULTICAST_PKT_CNT (DCC_CSRS + 0x000000000130)
+#define DCC_PRF_PORT_RCV_PKTS_CNT (DCC_CSRS + 0x0000000001A8)
+#define DCC_PRF_PORT_VL_MARK_FECN_CNT (DCC_CSRS + 0x000000000338)
+#define DCC_PRF_PORT_VL_RCV_BECN_CNT (DCC_CSRS + 0x000000000298)
+#define DCC_PRF_PORT_VL_RCV_BUBBLE_CNT (DCC_CSRS + 0x0000000002E8)
+#define DCC_PRF_PORT_VL_RCV_DATA_CNT (DCC_CSRS + 0x0000000001B0)
+#define DCC_PRF_PORT_VL_RCV_FECN_CNT (DCC_CSRS + 0x000000000248)
+#define DCC_PRF_PORT_VL_RCV_PKTS_CNT (DCC_CSRS + 0x0000000001F8)
+#define DCC_PRF_PORT_XMIT_CORRECTABLE_CNT (DCC_CSRS + 0x000000000138)
+#define DCC_PRF_PORT_XMIT_DATA_CNT (DCC_CSRS + 0x000000000190)
+#define DCC_PRF_PORT_XMIT_MULTICAST_CNT (DCC_CSRS + 0x000000000128)
+#define DCC_PRF_PORT_XMIT_PKTS_CNT (DCC_CSRS + 0x0000000001A0)
+#define DCC_PRF_RX_FLOW_CRTL_CNT (DCC_CSRS + 0x000000000180)
+#define DCC_PRF_TX_FLOW_CRTL_CNT (DCC_CSRS + 0x000000000188)
+#define DC_DC8051_CFG_CSR_ACCESS_SEL (DC_8051_CSRS + 0x000000000110)
+#define DC_DC8051_CFG_CSR_ACCESS_SEL_DCC_SMASK 0x2ull
+#define DC_DC8051_CFG_CSR_ACCESS_SEL_LCB_SMASK 0x1ull
+#define DC_DC8051_CFG_EXT_DEV_0 (DC_8051_CSRS + 0x000000000118)
+#define DC_DC8051_CFG_EXT_DEV_0_COMPLETED_SMASK 0x1ull
+#define DC_DC8051_CFG_EXT_DEV_0_RETURN_CODE_SHIFT 8
+#define DC_DC8051_CFG_EXT_DEV_0_RSP_DATA_SHIFT 16
+#define DC_DC8051_CFG_EXT_DEV_1 (DC_8051_CSRS + 0x000000000120)
+#define DC_DC8051_CFG_EXT_DEV_1_REQ_DATA_MASK 0xFFFFull
+#define DC_DC8051_CFG_EXT_DEV_1_REQ_DATA_SHIFT 16
+#define DC_DC8051_CFG_EXT_DEV_1_REQ_DATA_SMASK 0xFFFF0000ull
+#define DC_DC8051_CFG_EXT_DEV_1_REQ_NEW_SMASK 0x1ull
+#define DC_DC8051_CFG_EXT_DEV_1_REQ_TYPE_MASK 0xFFull
+#define DC_DC8051_CFG_EXT_DEV_1_REQ_TYPE_SHIFT 8
+#define DC_DC8051_CFG_HOST_CMD_0 (DC_8051_CSRS + 0x000000000028)
+#define DC_DC8051_CFG_HOST_CMD_0_REQ_DATA_MASK 0xFFFFFFFFFFFFull
+#define DC_DC8051_CFG_HOST_CMD_0_REQ_DATA_SHIFT 16
+#define DC_DC8051_CFG_HOST_CMD_0_REQ_NEW_SMASK 0x1ull
+#define DC_DC8051_CFG_HOST_CMD_0_REQ_TYPE_MASK 0xFFull
+#define DC_DC8051_CFG_HOST_CMD_0_REQ_TYPE_SHIFT 8
+#define DC_DC8051_CFG_HOST_CMD_1 (DC_8051_CSRS + 0x000000000030)
+#define DC_DC8051_CFG_HOST_CMD_1_COMPLETED_SMASK 0x1ull
+#define DC_DC8051_CFG_HOST_CMD_1_RETURN_CODE_MASK 0xFFull
+#define DC_DC8051_CFG_HOST_CMD_1_RETURN_CODE_SHIFT 8
+#define DC_DC8051_CFG_HOST_CMD_1_RSP_DATA_MASK 0xFFFFFFFFFFFFull
+#define DC_DC8051_CFG_HOST_CMD_1_RSP_DATA_SHIFT 16
+#define DC_DC8051_CFG_LOCAL_GUID (DC_8051_CSRS + 0x000000000038)
+#define DC_DC8051_CFG_MODE (DC_8051_CSRS + 0x000000000070)
+#define DC_DC8051_CFG_RAM_ACCESS_CTRL (DC_8051_CSRS + 0x000000000008)
+#define DC_DC8051_CFG_RAM_ACCESS_CTRL_ADDRESS_MASK 0x7FFFull
+#define DC_DC8051_CFG_RAM_ACCESS_CTRL_ADDRESS_SHIFT 0
+#define DC_DC8051_CFG_RAM_ACCESS_CTRL_WRITE_ENA_SMASK 0x1000000ull
+#define DC_DC8051_CFG_RAM_ACCESS_CTRL_READ_ENA_SMASK 0x10000ull
+#define DC_DC8051_CFG_RAM_ACCESS_SETUP (DC_8051_CSRS + 0x000000000000)
+#define DC_DC8051_CFG_RAM_ACCESS_SETUP_AUTO_INCR_ADDR_SMASK 0x100ull
+#define DC_DC8051_CFG_RAM_ACCESS_SETUP_RAM_SEL_SMASK 0x1ull
+#define DC_DC8051_CFG_RAM_ACCESS_STATUS (DC_8051_CSRS + 0x000000000018)
+#define DC_DC8051_CFG_RAM_ACCESS_STATUS_ACCESS_COMPLETED_SMASK 0x10000ull
+#define DC_DC8051_CFG_RAM_ACCESS_WR_DATA (DC_8051_CSRS + 0x000000000010)
+#define DC_DC8051_CFG_RAM_ACCESS_RD_DATA (DC_8051_CSRS + 0x000000000020)
+#define DC_DC8051_CFG_RST (DC_8051_CSRS + 0x000000000068)
+#define DC_DC8051_CFG_RST_CRAM_SMASK 0x2ull
+#define DC_DC8051_CFG_RST_DRAM_SMASK 0x4ull
+#define DC_DC8051_CFG_RST_IRAM_SMASK 0x8ull
+#define DC_DC8051_CFG_RST_M8051W_SMASK 0x1ull
+#define DC_DC8051_CFG_RST_SFR_SMASK 0x10ull
+#define DC_DC8051_DBG_ERR_INFO_SET_BY_8051 (DC_8051_CSRS + 0x0000000000D8)
+#define DC_DC8051_DBG_ERR_INFO_SET_BY_8051_ERROR_MASK 0xFFFFFFFFull
+#define DC_DC8051_DBG_ERR_INFO_SET_BY_8051_ERROR_SHIFT 16
+#define DC_DC8051_DBG_ERR_INFO_SET_BY_8051_HOST_MSG_MASK 0xFFFFull
+#define DC_DC8051_DBG_ERR_INFO_SET_BY_8051_HOST_MSG_SHIFT 0
+#define DC_DC8051_ERR_CLR (DC_8051_CSRS + 0x0000000000E8)
+#define DC_DC8051_ERR_EN (DC_8051_CSRS + 0x0000000000F0)
+#define DC_DC8051_ERR_EN_LOST_8051_HEART_BEAT_SMASK 0x2ull
+#define DC_DC8051_ERR_FLG (DC_8051_CSRS + 0x0000000000E0)
+#define DC_DC8051_ERR_FLG_CRAM_MBE_SMASK 0x4ull
+#define DC_DC8051_ERR_FLG_CRAM_SBE_SMASK 0x8ull
+#define DC_DC8051_ERR_FLG_DRAM_MBE_SMASK 0x10ull
+#define DC_DC8051_ERR_FLG_DRAM_SBE_SMASK 0x20ull
+#define DC_DC8051_ERR_FLG_INVALID_CSR_ADDR_SMASK 0x400ull
+#define DC_DC8051_ERR_FLG_IRAM_MBE_SMASK 0x40ull
+#define DC_DC8051_ERR_FLG_IRAM_SBE_SMASK 0x80ull
+#define DC_DC8051_ERR_FLG_LOST_8051_HEART_BEAT_SMASK 0x2ull
+#define DC_DC8051_ERR_FLG_SET_BY_8051_SMASK 0x1ull
+#define DC_DC8051_ERR_FLG_UNMATCHED_SECURE_MSG_ACROSS_BCC_LANES_SMASK 0x100ull
+#define DC_DC8051_STS_CUR_STATE (DC_8051_CSRS + 0x000000000060)
+#define DC_DC8051_STS_CUR_STATE_FIRMWARE_MASK 0xFFull
+#define DC_DC8051_STS_CUR_STATE_FIRMWARE_SHIFT 16
+#define DC_DC8051_STS_CUR_STATE_PORT_MASK 0xFFull
+#define DC_DC8051_STS_CUR_STATE_PORT_SHIFT 0
+#define DC_DC8051_STS_LOCAL_FM_SECURITY (DC_8051_CSRS + 0x000000000050)
+#define DC_DC8051_STS_LOCAL_FM_SECURITY_DISABLED_MASK 0x1ull
+#define DC_DC8051_STS_REMOTE_FM_SECURITY (DC_8051_CSRS + 0x000000000058)
+#define DC_DC8051_STS_REMOTE_GUID (DC_8051_CSRS + 0x000000000040)
+#define DC_DC8051_STS_REMOTE_NODE_TYPE (DC_8051_CSRS + 0x000000000048)
+#define DC_DC8051_STS_REMOTE_NODE_TYPE_VAL_MASK 0x3ull
+#define DC_DC8051_STS_REMOTE_PORT_NO (DC_8051_CSRS + 0x000000000130)
+#define DC_DC8051_STS_REMOTE_PORT_NO_VAL_SMASK 0xFFull
+#define DC_LCB_CFG_ALLOW_LINK_UP (DC_LCB_CSRS + 0x000000000128)
+#define DC_LCB_CFG_ALLOW_LINK_UP_VAL_SHIFT 0
+#define DC_LCB_CFG_CRC_MODE (DC_LCB_CSRS + 0x000000000058)
+#define DC_LCB_CFG_CRC_MODE_TX_VAL_SHIFT 0
+#define DC_LCB_CFG_IGNORE_LOST_RCLK (DC_LCB_CSRS + 0x000000000020)
+#define DC_LCB_CFG_IGNORE_LOST_RCLK_EN_SMASK 0x1ull
+#define DC_LCB_CFG_LANE_WIDTH (DC_LCB_CSRS + 0x000000000100)
+#define DC_LCB_CFG_LINK_KILL_EN (DC_LCB_CSRS + 0x000000000120)
+#define DC_LCB_CFG_LINK_KILL_EN_FLIT_INPUT_BUF_MBE_SMASK 0x100000ull
+#define DC_LCB_CFG_LINK_KILL_EN_REPLAY_BUF_MBE_SMASK 0x400000ull
+#define DC_LCB_CFG_LN_DCLK (DC_LCB_CSRS + 0x000000000060)
+#define DC_LCB_CFG_LOOPBACK (DC_LCB_CSRS + 0x0000000000F8)
+#define DC_LCB_CFG_LOOPBACK_VAL_SHIFT 0
+#define DC_LCB_CFG_RUN (DC_LCB_CSRS + 0x000000000000)
+#define DC_LCB_CFG_RUN_EN_SHIFT 0
+#define DC_LCB_CFG_RX_FIFOS_RADR (DC_LCB_CSRS + 0x000000000018)
+#define DC_LCB_CFG_RX_FIFOS_RADR_DO_NOT_JUMP_VAL_SHIFT 8
+#define DC_LCB_CFG_RX_FIFOS_RADR_OK_TO_JUMP_VAL_SHIFT 4
+#define DC_LCB_CFG_RX_FIFOS_RADR_RST_VAL_SHIFT 0
+#define DC_LCB_CFG_TX_FIFOS_RADR (DC_LCB_CSRS + 0x000000000010)
+#define DC_LCB_CFG_TX_FIFOS_RADR_RST_VAL_SHIFT 0
+#define DC_LCB_CFG_TX_FIFOS_RESET (DC_LCB_CSRS + 0x000000000008)
+#define DC_LCB_CFG_TX_FIFOS_RESET_VAL_SHIFT 0
+#define DC_LCB_ERR_CLR (DC_LCB_CSRS + 0x000000000308)
+#define DC_LCB_ERR_EN (DC_LCB_CSRS + 0x000000000310)
+#define DC_LCB_ERR_FLG (DC_LCB_CSRS + 0x000000000300)
+#define DC_LCB_ERR_FLG_REDUNDANT_FLIT_PARITY_ERR_SMASK 0x20000000ull
+#define DC_LCB_ERR_FLG_NEG_EDGE_LINK_TRANSFER_ACTIVE_SMASK 0x10000000ull
+#define DC_LCB_ERR_FLG_HOLD_REINIT_SMASK 0x8000000ull
+#define DC_LCB_ERR_FLG_RST_FOR_INCOMPLT_RND_TRIP_SMASK 0x4000000ull
+#define DC_LCB_ERR_FLG_RST_FOR_LINK_TIMEOUT_SMASK 0x2000000ull
+#define DC_LCB_ERR_FLG_CREDIT_RETURN_FLIT_MBE_SMASK 0x1000000ull
+#define DC_LCB_ERR_FLG_REPLAY_BUF_SBE_SMASK 0x800000ull
+#define DC_LCB_ERR_FLG_REPLAY_BUF_MBE_SMASK 0x400000ull
+#define DC_LCB_ERR_FLG_FLIT_INPUT_BUF_SBE_SMASK 0x200000ull
+#define DC_LCB_ERR_FLG_FLIT_INPUT_BUF_MBE_SMASK 0x100000ull
+#define DC_LCB_ERR_FLG_VL_ACK_INPUT_WRONG_CRC_MODE_SMASK 0x80000ull
+#define DC_LCB_ERR_FLG_VL_ACK_INPUT_PARITY_ERR_SMASK 0x40000ull
+#define DC_LCB_ERR_FLG_VL_ACK_INPUT_BUF_OFLW_SMASK 0x20000ull
+#define DC_LCB_ERR_FLG_FLIT_INPUT_BUF_OFLW_SMASK 0x10000ull
+#define DC_LCB_ERR_FLG_ILLEGAL_FLIT_ENCODING_SMASK 0x8000ull
+#define DC_LCB_ERR_FLG_ILLEGAL_NULL_LTP_SMASK 0x4000ull
+#define DC_LCB_ERR_FLG_UNEXPECTED_ROUND_TRIP_MARKER_SMASK 0x2000ull
+#define DC_LCB_ERR_FLG_UNEXPECTED_REPLAY_MARKER_SMASK 0x1000ull
+#define DC_LCB_ERR_FLG_RCLK_STOPPED_SMASK 0x800ull
+#define DC_LCB_ERR_FLG_CRC_ERR_CNT_HIT_LIMIT_SMASK 0x400ull
+#define DC_LCB_ERR_FLG_REINIT_FOR_LN_DEGRADE_SMASK 0x200ull
+#define DC_LCB_ERR_FLG_REINIT_FROM_PEER_SMASK 0x100ull
+#define DC_LCB_ERR_FLG_SEQ_CRC_ERR_SMASK 0x80ull
+#define DC_LCB_ERR_FLG_RX_LESS_THAN_FOUR_LNS_SMASK 0x40ull
+#define DC_LCB_ERR_FLG_TX_LESS_THAN_FOUR_LNS_SMASK 0x20ull
+#define DC_LCB_ERR_FLG_LOST_REINIT_STALL_OR_TOS_SMASK 0x10ull
+#define DC_LCB_ERR_FLG_ALL_LNS_FAILED_REINIT_TEST_SMASK 0x8ull
+#define DC_LCB_ERR_FLG_RST_FOR_FAILED_DESKEW_SMASK 0x4ull
+#define DC_LCB_ERR_FLG_INVALID_CSR_ADDR_SMASK 0x2ull
+#define DC_LCB_ERR_FLG_CSR_PARITY_ERR_SMASK 0x1ull
+#define DC_LCB_ERR_INFO_CRC_ERR_LN0 (DC_LCB_CSRS + 0x000000000328)
+#define DC_LCB_ERR_INFO_CRC_ERR_LN1 (DC_LCB_CSRS + 0x000000000330)
+#define DC_LCB_ERR_INFO_CRC_ERR_LN2 (DC_LCB_CSRS + 0x000000000338)
+#define DC_LCB_ERR_INFO_CRC_ERR_LN3 (DC_LCB_CSRS + 0x000000000340)
+#define DC_LCB_ERR_INFO_CRC_ERR_MULTI_LN (DC_LCB_CSRS + 0x000000000348)
+#define DC_LCB_ERR_INFO_ESCAPE_0_ONLY_CNT (DC_LCB_CSRS + 0x000000000368)
+#define DC_LCB_ERR_INFO_ESCAPE_0_PLUS1_CNT (DC_LCB_CSRS + 0x000000000370)
+#define DC_LCB_ERR_INFO_ESCAPE_0_PLUS2_CNT (DC_LCB_CSRS + 0x000000000378)
+#define DC_LCB_ERR_INFO_MISC_FLG_CNT (DC_LCB_CSRS + 0x000000000390)
+#define DC_LCB_ERR_INFO_REINIT_FROM_PEER_CNT (DC_LCB_CSRS + 0x000000000380)
+#define DC_LCB_ERR_INFO_RX_REPLAY_CNT (DC_LCB_CSRS + 0x000000000358)
+#define DC_LCB_ERR_INFO_SBE_CNT (DC_LCB_CSRS + 0x000000000388)
+#define DC_LCB_ERR_INFO_SEQ_CRC_CNT (DC_LCB_CSRS + 0x000000000360)
+#define DC_LCB_ERR_INFO_TOTAL_CRC_ERR (DC_LCB_CSRS + 0x000000000320)
+#define DC_LCB_ERR_INFO_TX_REPLAY_CNT (DC_LCB_CSRS + 0x000000000350)
+#define DC_LCB_PG_DBG_FLIT_CRDTS_CNT (DC_LCB_CSRS + 0x000000000580)
+#define DC_LCB_PG_STS_PAUSE_COMPLETE_CNT (DC_LCB_CSRS + 0x0000000005F8)
+#define DC_LCB_PG_STS_TX_MBE_CNT (DC_LCB_CSRS + 0x000000000608)
+#define DC_LCB_PG_STS_TX_SBE_CNT (DC_LCB_CSRS + 0x000000000600)
+#define DC_LCB_PRF_ACCEPTED_LTP_CNT (DC_LCB_CSRS + 0x000000000408)
+#define DC_LCB_PRF_CLK_CNTR (DC_LCB_CSRS + 0x000000000420)
+#define DC_LCB_PRF_GOOD_LTP_CNT (DC_LCB_CSRS + 0x000000000400)
+#define DC_LCB_PRF_RX_FLIT_CNT (DC_LCB_CSRS + 0x000000000410)
+#define DC_LCB_PRF_TX_FLIT_CNT (DC_LCB_CSRS + 0x000000000418)
+#define DC_LCB_STS_LINK_TRANSFER_ACTIVE (DC_LCB_CSRS + 0x000000000468)
+#define DC_LCB_STS_ROUND_TRIP_LTP_CNT (DC_LCB_CSRS + 0x0000000004B0)
+#define RCV_BUF_OVFL_CNT 10
+#define RCV_CONTEXT_EGR_STALL 22
+#define RCV_CONTEXT_RHQ_STALL 21
+#define RCV_DATA_PKT_CNT 0
+#define RCV_DWORD_CNT 1
+#define RCV_TID_FLOW_GEN_MISMATCH_CNT 20
+#define RCV_TID_FLOW_SEQ_MISMATCH_CNT 23
+#define RCV_TID_FULL_ERR_CNT 18
+#define RCV_TID_VALID_ERR_CNT 19
+#define RXE_NUM_32_BIT_COUNTERS 24
+#define RXE_NUM_64_BIT_COUNTERS 2
+#define RXE_NUM_RSM_INSTANCES 4
+#define RXE_NUM_TID_FLOWS 32
+#define RXE_PER_CONTEXT_OFFSET 0x0300000
+#define SEND_DATA_PKT_CNT 0
+#define SEND_DATA_PKT_VL0_CNT 12
+#define SEND_DATA_VL0_CNT 3
+#define SEND_DROPPED_PKT_CNT 5
+#define SEND_DWORD_CNT 1
+#define SEND_FLOW_STALL_CNT 4
+#define SEND_HEADERS_ERR_CNT 6
+#define SEND_LEN_ERR_CNT 1
+#define SEND_MAX_MIN_LEN_ERR_CNT 2
+#define SEND_UNDERRUN_CNT 3
+#define SEND_UNSUP_VL_ERR_CNT 0
+#define SEND_WAIT_CNT 2
+#define SEND_WAIT_VL0_CNT 21
+#define TXE_PIO_SEND_OFFSET 0x0800000
+#define ASIC_CFG_DRV_STR (ASIC + 0x000000000048)
+#define ASIC_CFG_MUTEX (ASIC + 0x000000000040)
+#define ASIC_CFG_SBUS_EXECUTE (ASIC + 0x000000000008)
+#define ASIC_CFG_SBUS_EXECUTE_EXECUTE_SMASK 0x1ull
+#define ASIC_CFG_SBUS_EXECUTE_FAST_MODE_SMASK 0x2ull
+#define ASIC_CFG_SBUS_REQUEST (ASIC + 0x000000000000)
+#define ASIC_CFG_SBUS_REQUEST_COMMAND_SHIFT 16
+#define ASIC_CFG_SBUS_REQUEST_DATA_ADDR_SHIFT 8
+#define ASIC_CFG_SBUS_REQUEST_DATA_IN_SHIFT 32
+#define ASIC_CFG_SBUS_REQUEST_RECEIVER_ADDR_SHIFT 0
+#define ASIC_CFG_SCRATCH (ASIC + 0x000000000020)
+#define ASIC_CFG_THERM_POLL_EN (ASIC + 0x000000000050)
+#define ASIC_EEP_ADDR_CMD (ASIC + 0x000000000308)
+#define ASIC_EEP_ADDR_CMD_EP_ADDR_MASK 0xFFFFFFull
+#define ASIC_EEP_CTL_STAT (ASIC + 0x000000000300)
+#define ASIC_EEP_CTL_STAT_EP_RESET_SMASK 0x4ull
+#define ASIC_EEP_CTL_STAT_RATE_SPI_SHIFT 8
+#define ASIC_EEP_CTL_STAT_RESETCSR 0x0000000083818000ull
+#define ASIC_EEP_DATA (ASIC + 0x000000000310)
+#define ASIC_GPIO_CLEAR (ASIC + 0x000000000230)
+#define ASIC_GPIO_FORCE (ASIC + 0x000000000238)
+#define ASIC_GPIO_IN (ASIC + 0x000000000200)
+#define ASIC_GPIO_INVERT (ASIC + 0x000000000210)
+#define ASIC_GPIO_MASK (ASIC + 0x000000000220)
+#define ASIC_GPIO_OE (ASIC + 0x000000000208)
+#define ASIC_GPIO_OUT (ASIC + 0x000000000218)
+#define ASIC_PCIE_SD_HOST_CMD (ASIC + 0x000000000100)
+#define ASIC_PCIE_SD_HOST_CMD_INTRPT_CMD_SHIFT 0
+#define ASIC_PCIE_SD_HOST_CMD_SBR_MODE_SMASK 0x400ull
+#define ASIC_PCIE_SD_HOST_CMD_SBUS_RCVR_ADDR_SHIFT 2
+#define ASIC_PCIE_SD_HOST_CMD_TIMER_MASK 0xFFFFFull
+#define ASIC_PCIE_SD_HOST_CMD_TIMER_SHIFT 12
+#define ASIC_PCIE_SD_HOST_STATUS (ASIC + 0x000000000108)
+#define ASIC_PCIE_SD_HOST_STATUS_FW_DNLD_ERR_MASK 0x7ull
+#define ASIC_PCIE_SD_HOST_STATUS_FW_DNLD_ERR_SHIFT 2
+#define ASIC_PCIE_SD_HOST_STATUS_FW_DNLD_STS_MASK 0x3ull
+#define ASIC_PCIE_SD_HOST_STATUS_FW_DNLD_STS_SHIFT 0
+#define ASIC_PCIE_SD_INTRPT_DATA_CODE (ASIC + 0x000000000110)
+#define ASIC_PCIE_SD_INTRPT_ENABLE (ASIC + 0x000000000118)
+#define ASIC_PCIE_SD_INTRPT_LIST (ASIC + 0x000000000180)
+#define ASIC_PCIE_SD_INTRPT_LIST_INTRPT_CODE_SHIFT 16
+#define ASIC_PCIE_SD_INTRPT_LIST_INTRPT_DATA_SHIFT 0
+#define ASIC_PCIE_SD_INTRPT_STATUS (ASIC + 0x000000000128)
+#define ASIC_QSFP1_CLEAR (ASIC + 0x000000000270)
+#define ASIC_QSFP1_FORCE (ASIC + 0x000000000278)
+#define ASIC_QSFP1_IN (ASIC + 0x000000000240)
+#define ASIC_QSFP1_INVERT (ASIC + 0x000000000250)
+#define ASIC_QSFP1_MASK (ASIC + 0x000000000260)
+#define ASIC_QSFP1_OE (ASIC + 0x000000000248)
+#define ASIC_QSFP1_OUT (ASIC + 0x000000000258)
+#define ASIC_QSFP1_STATUS (ASIC + 0x000000000268)
+#define ASIC_QSFP2_CLEAR (ASIC + 0x0000000002B0)
+#define ASIC_QSFP2_FORCE (ASIC + 0x0000000002B8)
+#define ASIC_QSFP2_IN (ASIC + 0x000000000280)
+#define ASIC_QSFP2_INVERT (ASIC + 0x000000000290)
+#define ASIC_QSFP2_MASK (ASIC + 0x0000000002A0)
+#define ASIC_QSFP2_OE (ASIC + 0x000000000288)
+#define ASIC_QSFP2_OUT (ASIC + 0x000000000298)
+#define ASIC_QSFP2_STATUS (ASIC + 0x0000000002A8)
+#define ASIC_STS_SBUS_COUNTERS (ASIC + 0x000000000018)
+#define ASIC_STS_SBUS_COUNTERS_EXECUTE_CNT_MASK 0xFFFFull
+#define ASIC_STS_SBUS_COUNTERS_EXECUTE_CNT_SHIFT 0
+#define ASIC_STS_SBUS_COUNTERS_RCV_DATA_VALID_CNT_MASK 0xFFFFull
+#define ASIC_STS_SBUS_COUNTERS_RCV_DATA_VALID_CNT_SHIFT 16
+#define ASIC_STS_SBUS_RESULT (ASIC + 0x000000000010)
+#define ASIC_STS_SBUS_RESULT_DONE_SMASK 0x1ull
+#define ASIC_STS_SBUS_RESULT_RCV_DATA_VALID_SMASK 0x2ull
+#define ASIC_STS_THERM (ASIC + 0x000000000058)
+#define ASIC_STS_THERM_CRIT_TEMP_MASK 0x7FFull
+#define ASIC_STS_THERM_CRIT_TEMP_SHIFT 18
+#define ASIC_STS_THERM_CURR_TEMP_MASK 0x7FFull
+#define ASIC_STS_THERM_CURR_TEMP_SHIFT 2
+#define ASIC_STS_THERM_HI_TEMP_MASK 0x7FFull
+#define ASIC_STS_THERM_HI_TEMP_SHIFT 50
+#define ASIC_STS_THERM_LO_TEMP_MASK 0x7FFull
+#define ASIC_STS_THERM_LO_TEMP_SHIFT 34
+#define ASIC_STS_THERM_LOW_SHIFT 13
+#define CCE_COUNTER_ARRAY32 (CCE + 0x000000000060)
+#define CCE_CTRL (CCE + 0x000000000010)
+#define CCE_CTRL_RXE_RESUME_SMASK 0x800ull
+#define CCE_CTRL_SPC_FREEZE_SMASK 0x100ull
+#define CCE_CTRL_SPC_UNFREEZE_SMASK 0x200ull
+#define CCE_CTRL_TXE_RESUME_SMASK 0x2000ull
+#define CCE_DC_CTRL (CCE + 0x0000000000B8)
+#define CCE_DC_CTRL_DC_RESET_SMASK 0x1ull
+#define CCE_DC_CTRL_RESETCSR 0x0000000000000001ull
+#define CCE_ERR_CLEAR (CCE + 0x000000000050)
+#define CCE_ERR_MASK (CCE + 0x000000000048)
+#define CCE_ERR_STATUS (CCE + 0x000000000040)
+#define CCE_ERR_STATUS_CCE_CLI0_ASYNC_FIFO_PARITY_ERR_SMASK 0x40ull
+#define CCE_ERR_STATUS_CCE_CLI1_ASYNC_FIFO_DBG_PARITY_ERROR_SMASK 0x1000ull
+#define CCE_ERR_STATUS_CCE_CLI1_ASYNC_FIFO_PIO_CRDT_PARITY_ERR_SMASK \
+ 0x200ull
+#define CCE_ERR_STATUS_CCE_CLI1_ASYNC_FIFO_RXDMA_PARITY_ERROR_SMASK \
+ 0x800ull
+#define CCE_ERR_STATUS_CCE_CLI1_ASYNC_FIFO_SDMA_HD_PARITY_ERR_SMASK \
+ 0x400ull
+#define CCE_ERR_STATUS_CCE_CLI2_ASYNC_FIFO_PARITY_ERR_SMASK 0x100ull
+#define CCE_ERR_STATUS_CCE_CSR_CFG_BUS_PARITY_ERR_SMASK 0x80ull
+#define CCE_ERR_STATUS_CCE_CSR_PARITY_ERR_SMASK 0x1ull
+#define CCE_ERR_STATUS_CCE_CSR_READ_BAD_ADDR_ERR_SMASK 0x2ull
+#define CCE_ERR_STATUS_CCE_CSR_WRITE_BAD_ADDR_ERR_SMASK 0x4ull
+#define CCE_ERR_STATUS_CCE_INT_MAP_COR_ERR_SMASK 0x4000000000ull
+#define CCE_ERR_STATUS_CCE_INT_MAP_UNC_ERR_SMASK 0x8000000000ull
+#define CCE_ERR_STATUS_CCE_MSIX_CSR_PARITY_ERR_SMASK 0x10000000000ull
+#define CCE_ERR_STATUS_CCE_MSIX_TABLE_COR_ERR_SMASK 0x1000000000ull
+#define CCE_ERR_STATUS_CCE_MSIX_TABLE_UNC_ERR_SMASK 0x2000000000ull
+#define CCE_ERR_STATUS_CCE_RCPL_ASYNC_FIFO_PARITY_ERR_SMASK 0x400000000ull
+#define CCE_ERR_STATUS_CCE_RSPD_DATA_PARITY_ERR_SMASK 0x20ull
+#define CCE_ERR_STATUS_CCE_RXDMA_CONV_FIFO_PARITY_ERR_SMASK 0x800000000ull
+#define CCE_ERR_STATUS_CCE_SEG_READ_BAD_ADDR_ERR_SMASK 0x100000000ull
+#define CCE_ERR_STATUS_CCE_SEG_WRITE_BAD_ADDR_ERR_SMASK 0x200000000ull
+#define CCE_ERR_STATUS_CCE_TRGT_ACCESS_ERR_SMASK 0x10ull
+#define CCE_ERR_STATUS_CCE_TRGT_ASYNC_FIFO_PARITY_ERR_SMASK 0x8ull
+#define CCE_ERR_STATUS_CCE_TRGT_CPL_TIMEOUT_ERR_SMASK 0x40000000ull
+#define CCE_ERR_STATUS_LA_TRIGGERED_SMASK 0x80000000ull
+#define CCE_ERR_STATUS_PCIC_CPL_DAT_QCOR_ERR_SMASK 0x40000ull
+#define CCE_ERR_STATUS_PCIC_CPL_DAT_QUNC_ERR_SMASK 0x4000000ull
+#define CCE_ERR_STATUS_PCIC_CPL_HD_QCOR_ERR_SMASK 0x20000ull
+#define CCE_ERR_STATUS_PCIC_CPL_HD_QUNC_ERR_SMASK 0x2000000ull
+#define CCE_ERR_STATUS_PCIC_NPOST_DAT_QPARITY_ERR_SMASK 0x100000ull
+#define CCE_ERR_STATUS_PCIC_NPOST_HQ_PARITY_ERR_SMASK 0x80000ull
+#define CCE_ERR_STATUS_PCIC_POST_DAT_QCOR_ERR_SMASK 0x10000ull
+#define CCE_ERR_STATUS_PCIC_POST_DAT_QUNC_ERR_SMASK 0x1000000ull
+#define CCE_ERR_STATUS_PCIC_POST_HD_QCOR_ERR_SMASK 0x8000ull
+#define CCE_ERR_STATUS_PCIC_POST_HD_QUNC_ERR_SMASK 0x800000ull
+#define CCE_ERR_STATUS_PCIC_RECEIVE_PARITY_ERR_SMASK 0x20000000ull
+#define CCE_ERR_STATUS_PCIC_RETRY_MEM_COR_ERR_SMASK 0x2000ull
+#define CCE_ERR_STATUS_PCIC_RETRY_MEM_UNC_ERR_SMASK 0x200000ull
+#define CCE_ERR_STATUS_PCIC_RETRY_SOT_MEM_COR_ERR_SMASK 0x4000ull
+#define CCE_ERR_STATUS_PCIC_RETRY_SOT_MEM_UNC_ERR_SMASK 0x400000ull
+#define CCE_ERR_STATUS_PCIC_TRANSMIT_BACK_PARITY_ERR_SMASK 0x10000000ull
+#define CCE_ERR_STATUS_PCIC_TRANSMIT_FRONT_PARITY_ERR_SMASK 0x8000000ull
+#define CCE_INT_CLEAR (CCE + 0x000000110A00)
+#define CCE_INT_COUNTER_ARRAY32 (CCE + 0x000000110D00)
+#define CCE_INT_FORCE (CCE + 0x000000110B00)
+#define CCE_INT_MAP (CCE + 0x000000110500)
+#define CCE_INT_MASK (CCE + 0x000000110900)
+#define CCE_INT_STATUS (CCE + 0x000000110800)
+#define CCE_MSIX_INT_GRANTED (CCE + 0x000000110200)
+#define CCE_MSIX_TABLE_LOWER (CCE + 0x000000100000)
+#define CCE_MSIX_TABLE_UPPER (CCE + 0x000000100008)
+#define CCE_MSIX_TABLE_UPPER_RESETCSR 0x0000000100000000ull
+#define CCE_MSIX_VEC_CLR_WITHOUT_INT (CCE + 0x000000110400)
+#define CCE_REVISION (CCE + 0x000000000000)
+#define CCE_REVISION2 (CCE + 0x000000000008)
+#define CCE_REVISION2_HFI_ID_MASK 0x1ull
+#define CCE_REVISION2_HFI_ID_SHIFT 0
+#define CCE_REVISION2_IMPL_CODE_SHIFT 8
+#define CCE_REVISION2_IMPL_REVISION_SHIFT 16
+#define CCE_REVISION_BOARD_ID_LOWER_NIBBLE_MASK 0xFull
+#define CCE_REVISION_BOARD_ID_LOWER_NIBBLE_SHIFT 32
+#define CCE_REVISION_CHIP_REV_MAJOR_MASK 0xFFull
+#define CCE_REVISION_CHIP_REV_MAJOR_SHIFT 8
+#define CCE_REVISION_CHIP_REV_MINOR_MASK 0xFFull
+#define CCE_REVISION_CHIP_REV_MINOR_SHIFT 0
+#define CCE_REVISION_SW_MASK 0xFFull
+#define CCE_REVISION_SW_SHIFT 24
+#define CCE_SCRATCH (CCE + 0x000000000020)
+#define CCE_STATUS (CCE + 0x000000000018)
+#define CCE_STATUS_RXE_FROZE_SMASK 0x2ull
+#define CCE_STATUS_RXE_PAUSED_SMASK 0x20ull
+#define CCE_STATUS_SDMA_FROZE_SMASK 0x1ull
+#define CCE_STATUS_SDMA_PAUSED_SMASK 0x10ull
+#define CCE_STATUS_TXE_FROZE_SMASK 0x4ull
+#define CCE_STATUS_TXE_PAUSED_SMASK 0x40ull
+#define CCE_STATUS_TXE_PIO_FROZE_SMASK 0x8ull
+#define CCE_STATUS_TXE_PIO_PAUSED_SMASK 0x80ull
+#define MISC_CFG_FW_CTRL (MISC + 0x000000001000)
+#define MISC_CFG_FW_CTRL_FW_8051_LOADED_SMASK 0x2ull
+#define MISC_CFG_FW_CTRL_RSA_STATUS_SHIFT 2
+#define MISC_CFG_FW_CTRL_RSA_STATUS_SMASK 0xCull
+#define MISC_CFG_RSA_CMD (MISC + 0x000000000A08)
+#define MISC_CFG_RSA_MODULUS (MISC + 0x000000000400)
+#define MISC_CFG_RSA_MU (MISC + 0x000000000A10)
+#define MISC_CFG_RSA_R2 (MISC + 0x000000000000)
+#define MISC_CFG_RSA_SIGNATURE (MISC + 0x000000000200)
+#define MISC_CFG_SHA_PRELOAD (MISC + 0x000000000A00)
+#define MISC_ERR_CLEAR (MISC + 0x000000002010)
+#define MISC_ERR_MASK (MISC + 0x000000002008)
+#define MISC_ERR_STATUS (MISC + 0x000000002000)
+#define MISC_ERR_STATUS_MISC_PLL_LOCK_FAIL_ERR_SMASK 0x1000ull
+#define MISC_ERR_STATUS_MISC_MBIST_FAIL_ERR_SMASK 0x800ull
+#define MISC_ERR_STATUS_MISC_INVALID_EEP_CMD_ERR_SMASK 0x400ull
+#define MISC_ERR_STATUS_MISC_EFUSE_DONE_PARITY_ERR_SMASK 0x200ull
+#define MISC_ERR_STATUS_MISC_EFUSE_WRITE_ERR_SMASK 0x100ull
+#define MISC_ERR_STATUS_MISC_EFUSE_READ_BAD_ADDR_ERR_SMASK 0x80ull
+#define MISC_ERR_STATUS_MISC_EFUSE_CSR_PARITY_ERR_SMASK 0x40ull
+#define MISC_ERR_STATUS_MISC_FW_AUTH_FAILED_ERR_SMASK 0x20ull
+#define MISC_ERR_STATUS_MISC_KEY_MISMATCH_ERR_SMASK 0x10ull
+#define MISC_ERR_STATUS_MISC_SBUS_WRITE_FAILED_ERR_SMASK 0x8ull
+#define MISC_ERR_STATUS_MISC_CSR_WRITE_BAD_ADDR_ERR_SMASK 0x4ull
+#define MISC_ERR_STATUS_MISC_CSR_READ_BAD_ADDR_ERR_SMASK 0x2ull
+#define MISC_ERR_STATUS_MISC_CSR_PARITY_ERR_SMASK 0x1ull
+#define PCI_CFG_MSIX0 (PCIE + 0x0000000000B0)
+#define PCI_CFG_REG1 (PCIE + 0x000000000004)
+#define PCI_CFG_REG11 (PCIE + 0x00000000002C)
+#define PCIE_CFG_SPCIE1 (PCIE + 0x00000000014C)
+#define PCIE_CFG_SPCIE2 (PCIE + 0x000000000150)
+#define PCIE_CFG_TPH2 (PCIE + 0x000000000180)
+#define RCV_ARRAY (RXE + 0x000000200000)
+#define RCV_ARRAY_CNT (RXE + 0x000000000018)
+#define RCV_ARRAY_RT_ADDR_MASK 0xFFFFFFFFFull
+#define RCV_ARRAY_RT_ADDR_SHIFT 0
+#define RCV_ARRAY_RT_BUF_SIZE_SHIFT 36
+#define RCV_ARRAY_RT_WRITE_ENABLE_SMASK 0x8000000000000000ull
+#define RCV_AVAIL_TIME_OUT (RXE + 0x000000100050)
+#define RCV_AVAIL_TIME_OUT_TIME_OUT_RELOAD_MASK 0xFFull
+#define RCV_AVAIL_TIME_OUT_TIME_OUT_RELOAD_SHIFT 0
+#define RCV_BTH_QP (RXE + 0x000000000028)
+#define RCV_BTH_QP_KDETH_QP_MASK 0xFFull
+#define RCV_BTH_QP_KDETH_QP_SHIFT 16
+#define RCV_BYPASS (RXE + 0x000000000038)
+#define RCV_CONTEXTS (RXE + 0x000000000010)
+#define RCV_COUNTER_ARRAY32 (RXE + 0x000000000400)
+#define RCV_COUNTER_ARRAY64 (RXE + 0x000000000500)
+#define RCV_CTRL (RXE + 0x000000000000)
+#define RCV_CTRL_RCV_BYPASS_ENABLE_SMASK 0x10ull
+#define RCV_CTRL_RCV_EXTENDED_PSN_ENABLE_SMASK 0x40ull
+#define RCV_CTRL_RCV_PARTITION_KEY_ENABLE_SMASK 0x4ull
+#define RCV_CTRL_RCV_PORT_ENABLE_SMASK 0x1ull
+#define RCV_CTRL_RCV_QP_MAP_ENABLE_SMASK 0x2ull
+#define RCV_CTRL_RCV_RSM_ENABLE_SMASK 0x20ull
+#define RCV_CTRL_RX_RBUF_INIT_SMASK 0x200ull
+#define RCV_CTXT_CTRL (RXE + 0x000000100000)
+#define RCV_CTXT_CTRL_DONT_DROP_EGR_FULL_SMASK 0x4ull
+#define RCV_CTXT_CTRL_DONT_DROP_RHQ_FULL_SMASK 0x8ull
+#define RCV_CTXT_CTRL_EGR_BUF_SIZE_MASK 0x7ull
+#define RCV_CTXT_CTRL_EGR_BUF_SIZE_SHIFT 8
+#define RCV_CTXT_CTRL_EGR_BUF_SIZE_SMASK 0x700ull
+#define RCV_CTXT_CTRL_ENABLE_SMASK 0x1ull
+#define RCV_CTXT_CTRL_INTR_AVAIL_SMASK 0x20ull
+#define RCV_CTXT_CTRL_ONE_PACKET_PER_EGR_BUFFER_SMASK 0x2ull
+#define RCV_CTXT_CTRL_TAIL_UPD_SMASK 0x40ull
+#define RCV_CTXT_CTRL_TID_FLOW_ENABLE_SMASK 0x10ull
+#define RCV_CTXT_STATUS (RXE + 0x000000100008)
+#define RCV_EGR_CTRL (RXE + 0x000000100010)
+#define RCV_EGR_CTRL_EGR_BASE_INDEX_MASK 0x1FFFull
+#define RCV_EGR_CTRL_EGR_BASE_INDEX_SHIFT 0
+#define RCV_EGR_CTRL_EGR_CNT_MASK 0x1FFull
+#define RCV_EGR_CTRL_EGR_CNT_SHIFT 32
+#define RCV_EGR_INDEX_HEAD (RXE + 0x000000300018)
+#define RCV_EGR_INDEX_HEAD_HEAD_MASK 0x7FFull
+#define RCV_EGR_INDEX_HEAD_HEAD_SHIFT 0
+#define RCV_ERR_CLEAR (RXE + 0x000000000070)
+#define RCV_ERR_INFO (RXE + 0x000000000050)
+#define RCV_ERR_INFO_RCV_EXCESS_BUFFER_OVERRUN_SC_SMASK 0x1Full
+#define RCV_ERR_INFO_RCV_EXCESS_BUFFER_OVERRUN_SMASK 0x20ull
+#define RCV_ERR_MASK (RXE + 0x000000000068)
+#define RCV_ERR_STATUS (RXE + 0x000000000060)
+#define RCV_ERR_STATUS_RX_CSR_PARITY_ERR_SMASK 0x8000000000000000ull
+#define RCV_ERR_STATUS_RX_CSR_READ_BAD_ADDR_ERR_SMASK 0x2000000000000000ull
+#define RCV_ERR_STATUS_RX_CSR_WRITE_BAD_ADDR_ERR_SMASK \
+ 0x4000000000000000ull
+#define RCV_ERR_STATUS_RX_DC_INTF_PARITY_ERR_SMASK 0x2ull
+#define RCV_ERR_STATUS_RX_DC_SOP_EOP_PARITY_ERR_SMASK 0x200ull
+#define RCV_ERR_STATUS_RX_DMA_CSR_COR_ERR_SMASK 0x1ull
+#define RCV_ERR_STATUS_RX_DMA_CSR_PARITY_ERR_SMASK 0x200000000000000ull
+#define RCV_ERR_STATUS_RX_DMA_CSR_UNC_ERR_SMASK 0x1000000000000000ull
+#define RCV_ERR_STATUS_RX_DMA_DATA_FIFO_RD_COR_ERR_SMASK \
+ 0x40000000000000ull
+#define RCV_ERR_STATUS_RX_DMA_DATA_FIFO_RD_UNC_ERR_SMASK \
+ 0x20000000000000ull
+#define RCV_ERR_STATUS_RX_DMA_DQ_FSM_ENCODING_ERR_SMASK \
+ 0x800000000000000ull
+#define RCV_ERR_STATUS_RX_DMA_EQ_FSM_ENCODING_ERR_SMASK \
+ 0x400000000000000ull
+#define RCV_ERR_STATUS_RX_DMA_FLAG_COR_ERR_SMASK 0x800ull
+#define RCV_ERR_STATUS_RX_DMA_FLAG_UNC_ERR_SMASK 0x400ull
+#define RCV_ERR_STATUS_RX_DMA_HDR_FIFO_RD_COR_ERR_SMASK 0x10000000000000ull
+#define RCV_ERR_STATUS_RX_DMA_HDR_FIFO_RD_UNC_ERR_SMASK 0x8000000000000ull
+#define RCV_ERR_STATUS_RX_HQ_INTR_CSR_PARITY_ERR_SMASK 0x200000000000ull
+#define RCV_ERR_STATUS_RX_HQ_INTR_FSM_ERR_SMASK 0x400000000000ull
+#define RCV_ERR_STATUS_RX_LOOKUP_CSR_PARITY_ERR_SMASK 0x100000000000ull
+#define RCV_ERR_STATUS_RX_LOOKUP_DES_PART1_UNC_COR_ERR_SMASK \
+ 0x10000000000ull
+#define RCV_ERR_STATUS_RX_LOOKUP_DES_PART1_UNC_ERR_SMASK 0x8000000000ull
+#define RCV_ERR_STATUS_RX_LOOKUP_DES_PART2_PARITY_ERR_SMASK \
+ 0x20000000000ull
+#define RCV_ERR_STATUS_RX_LOOKUP_RCV_ARRAY_COR_ERR_SMASK 0x80000000000ull
+#define RCV_ERR_STATUS_RX_LOOKUP_RCV_ARRAY_UNC_ERR_SMASK 0x40000000000ull
+#define RCV_ERR_STATUS_RX_RBUF_BAD_LOOKUP_ERR_SMASK 0x40000000ull
+#define RCV_ERR_STATUS_RX_RBUF_BLOCK_LIST_READ_COR_ERR_SMASK 0x100000ull
+#define RCV_ERR_STATUS_RX_RBUF_BLOCK_LIST_READ_UNC_ERR_SMASK 0x80000ull
+#define RCV_ERR_STATUS_RX_RBUF_CSR_QENT_CNT_PARITY_ERR_SMASK 0x400000ull
+#define RCV_ERR_STATUS_RX_RBUF_CSR_QEOPDW_PARITY_ERR_SMASK 0x10000000ull
+#define RCV_ERR_STATUS_RX_RBUF_CSR_QHD_PTR_PARITY_ERR_SMASK 0x2000000ull
+#define RCV_ERR_STATUS_RX_RBUF_CSR_QHEAD_BUF_NUM_PARITY_ERR_SMASK \
+ 0x200000ull
+#define RCV_ERR_STATUS_RX_RBUF_CSR_QNEXT_BUF_PARITY_ERR_SMASK 0x800000ull
+#define RCV_ERR_STATUS_RX_RBUF_CSR_QNUM_OF_PKT_PARITY_ERR_SMASK \
+ 0x8000000ull
+#define RCV_ERR_STATUS_RX_RBUF_CSR_QTL_PTR_PARITY_ERR_SMASK 0x4000000ull
+#define RCV_ERR_STATUS_RX_RBUF_CSR_QVLD_BIT_PARITY_ERR_SMASK 0x1000000ull
+#define RCV_ERR_STATUS_RX_RBUF_CTX_ID_PARITY_ERR_SMASK 0x20000000ull
+#define RCV_ERR_STATUS_RX_RBUF_DATA_COR_ERR_SMASK 0x100000000000000ull
+#define RCV_ERR_STATUS_RX_RBUF_DATA_UNC_ERR_SMASK 0x80000000000000ull
+#define RCV_ERR_STATUS_RX_RBUF_DESC_PART1_COR_ERR_SMASK 0x1000000000000ull
+#define RCV_ERR_STATUS_RX_RBUF_DESC_PART1_UNC_ERR_SMASK 0x800000000000ull
+#define RCV_ERR_STATUS_RX_RBUF_DESC_PART2_COR_ERR_SMASK 0x4000000000000ull
+#define RCV_ERR_STATUS_RX_RBUF_DESC_PART2_UNC_ERR_SMASK 0x2000000000000ull
+#define RCV_ERR_STATUS_RX_RBUF_EMPTY_ERR_SMASK 0x100000000ull
+#define RCV_ERR_STATUS_RX_RBUF_FL_INITDONE_PARITY_ERR_SMASK 0x800000000ull
+#define RCV_ERR_STATUS_RX_RBUF_FL_INIT_WR_ADDR_PARITY_ERR_SMASK \
+ 0x1000000000ull
+#define RCV_ERR_STATUS_RX_RBUF_FL_RD_ADDR_PARITY_ERR_SMASK 0x200000000ull
+#define RCV_ERR_STATUS_RX_RBUF_FL_WR_ADDR_PARITY_ERR_SMASK 0x400000000ull
+#define RCV_ERR_STATUS_RX_RBUF_FREE_LIST_COR_ERR_SMASK 0x4000ull
+#define RCV_ERR_STATUS_RX_RBUF_FREE_LIST_UNC_ERR_SMASK 0x2000ull
+#define RCV_ERR_STATUS_RX_RBUF_FULL_ERR_SMASK 0x80000000ull
+#define RCV_ERR_STATUS_RX_RBUF_LOOKUP_DES_COR_ERR_SMASK 0x40000ull
+#define RCV_ERR_STATUS_RX_RBUF_LOOKUP_DES_REG_UNC_COR_ERR_SMASK 0x10000ull
+#define RCV_ERR_STATUS_RX_RBUF_LOOKUP_DES_REG_UNC_ERR_SMASK 0x8000ull
+#define RCV_ERR_STATUS_RX_RBUF_LOOKUP_DES_UNC_ERR_SMASK 0x20000ull
+#define RCV_ERR_STATUS_RX_RBUF_NEXT_FREE_BUF_COR_ERR_SMASK 0x4000000000ull
+#define RCV_ERR_STATUS_RX_RBUF_NEXT_FREE_BUF_UNC_ERR_SMASK 0x2000000000ull
+#define RCV_ERR_STATUS_RX_RCV_CSR_PARITY_ERR_SMASK 0x100ull
+#define RCV_ERR_STATUS_RX_RCV_DATA_COR_ERR_SMASK 0x20ull
+#define RCV_ERR_STATUS_RX_RCV_DATA_UNC_ERR_SMASK 0x10ull
+#define RCV_ERR_STATUS_RX_RCV_FSM_ENCODING_ERR_SMASK 0x1000ull
+#define RCV_ERR_STATUS_RX_RCV_HDR_COR_ERR_SMASK 0x8ull
+#define RCV_ERR_STATUS_RX_RCV_HDR_UNC_ERR_SMASK 0x4ull
+#define RCV_ERR_STATUS_RX_RCV_QP_MAP_TABLE_COR_ERR_SMASK 0x80ull
+#define RCV_ERR_STATUS_RX_RCV_QP_MAP_TABLE_UNC_ERR_SMASK 0x40ull
+#define RCV_HDR_ADDR (RXE + 0x000000100028)
+#define RCV_HDR_CNT (RXE + 0x000000100030)
+#define RCV_HDR_CNT_CNT_MASK 0x1FFull
+#define RCV_HDR_CNT_CNT_SHIFT 0
+#define RCV_HDR_ENT_SIZE (RXE + 0x000000100038)
+#define RCV_HDR_ENT_SIZE_ENT_SIZE_MASK 0x7ull
+#define RCV_HDR_ENT_SIZE_ENT_SIZE_SHIFT 0
+#define RCV_HDR_HEAD (RXE + 0x000000300008)
+#define RCV_HDR_HEAD_COUNTER_MASK 0xFFull
+#define RCV_HDR_HEAD_COUNTER_SHIFT 32
+#define RCV_HDR_HEAD_HEAD_MASK 0x7FFFFull
+#define RCV_HDR_HEAD_HEAD_SHIFT 0
+#define RCV_HDR_HEAD_HEAD_SMASK 0x7FFFFull
+#define RCV_HDR_OVFL_CNT (RXE + 0x000000100058)
+#define RCV_HDR_SIZE (RXE + 0x000000100040)
+#define RCV_HDR_SIZE_HDR_SIZE_MASK 0x1Full
+#define RCV_HDR_SIZE_HDR_SIZE_SHIFT 0
+#define RCV_HDR_TAIL (RXE + 0x000000300000)
+#define RCV_HDR_TAIL_ADDR (RXE + 0x000000100048)
+#define RCV_KEY_CTRL (RXE + 0x000000100020)
+#define RCV_KEY_CTRL_JOB_KEY_ENABLE_SMASK 0x200000000ull
+#define RCV_KEY_CTRL_JOB_KEY_VALUE_MASK 0xFFFFull
+#define RCV_KEY_CTRL_JOB_KEY_VALUE_SHIFT 0
+#define RCV_MULTICAST (RXE + 0x000000000030)
+#define RCV_PARTITION_KEY (RXE + 0x000000000200)
+#define RCV_PARTITION_KEY_PARTITION_KEY_A_MASK 0xFFFFull
+#define RCV_PARTITION_KEY_PARTITION_KEY_B_SHIFT 16
+#define RCV_QP_MAP_TABLE (RXE + 0x000000000100)
+#define RCV_RSM_CFG (RXE + 0x000000000600)
+#define RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_MASK 0x1ull
+#define RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_SHIFT 0
+#define RCV_RSM_CFG_PACKET_TYPE_SHIFT 60
+#define RCV_RSM_MAP_TABLE (RXE + 0x000000000900)
+#define RCV_RSM_MAP_TABLE_RCV_CONTEXT_A_MASK 0xFFull
+#define RCV_RSM_MATCH (RXE + 0x000000000800)
+#define RCV_RSM_MATCH_MASK1_SHIFT 0
+#define RCV_RSM_MATCH_MASK2_SHIFT 16
+#define RCV_RSM_MATCH_VALUE1_SHIFT 8
+#define RCV_RSM_MATCH_VALUE2_SHIFT 24
+#define RCV_RSM_SELECT (RXE + 0x000000000700)
+#define RCV_RSM_SELECT_FIELD1_OFFSET_SHIFT 0
+#define RCV_RSM_SELECT_FIELD2_OFFSET_SHIFT 16
+#define RCV_RSM_SELECT_INDEX1_OFFSET_SHIFT 32
+#define RCV_RSM_SELECT_INDEX1_WIDTH_SHIFT 44
+#define RCV_RSM_SELECT_INDEX2_OFFSET_SHIFT 48
+#define RCV_RSM_SELECT_INDEX2_WIDTH_SHIFT 60
+#define RCV_STATUS (RXE + 0x000000000008)
+#define RCV_STATUS_RX_PKT_IN_PROGRESS_SMASK 0x1ull
+#define RCV_STATUS_RX_RBUF_INIT_DONE_SMASK 0x200ull
+#define RCV_STATUS_RX_RBUF_PKT_PENDING_SMASK 0x40ull
+#define RCV_TID_CTRL (RXE + 0x000000100018)
+#define RCV_TID_CTRL_TID_BASE_INDEX_MASK 0x1FFFull
+#define RCV_TID_CTRL_TID_BASE_INDEX_SHIFT 0
+#define RCV_TID_CTRL_TID_PAIR_CNT_MASK 0x1FFull
+#define RCV_TID_CTRL_TID_PAIR_CNT_SHIFT 32
+#define RCV_TID_FLOW_TABLE (RXE + 0x000000300800)
+#define RCV_VL15 (RXE + 0x000000000048)
+#define SEND_BTH_QP (TXE + 0x0000000000A0)
+#define SEND_BTH_QP_KDETH_QP_MASK 0xFFull
+#define SEND_BTH_QP_KDETH_QP_SHIFT 16
+#define SEND_CM_CREDIT_USED_STATUS (TXE + 0x000000000510)
+#define SEND_CM_CREDIT_USED_STATUS_VL0_RETURN_CREDIT_STATUS_SMASK \
+ 0x1000000000000ull
+#define SEND_CM_CREDIT_USED_STATUS_VL15_RETURN_CREDIT_STATUS_SMASK \
+ 0x8000000000000000ull
+#define SEND_CM_CREDIT_USED_STATUS_VL1_RETURN_CREDIT_STATUS_SMASK \
+ 0x2000000000000ull
+#define SEND_CM_CREDIT_USED_STATUS_VL2_RETURN_CREDIT_STATUS_SMASK \
+ 0x4000000000000ull
+#define SEND_CM_CREDIT_USED_STATUS_VL3_RETURN_CREDIT_STATUS_SMASK \
+ 0x8000000000000ull
+#define SEND_CM_CREDIT_USED_STATUS_VL4_RETURN_CREDIT_STATUS_SMASK \
+ 0x10000000000000ull
+#define SEND_CM_CREDIT_USED_STATUS_VL5_RETURN_CREDIT_STATUS_SMASK \
+ 0x20000000000000ull
+#define SEND_CM_CREDIT_USED_STATUS_VL6_RETURN_CREDIT_STATUS_SMASK \
+ 0x40000000000000ull
+#define SEND_CM_CREDIT_USED_STATUS_VL7_RETURN_CREDIT_STATUS_SMASK \
+ 0x80000000000000ull
+#define SEND_CM_CREDIT_VL (TXE + 0x000000000600)
+#define SEND_CM_CREDIT_VL15 (TXE + 0x000000000678)
+#define SEND_CM_CREDIT_VL15_DEDICATED_LIMIT_VL_SHIFT 0
+#define SEND_CM_CREDIT_VL_DEDICATED_LIMIT_VL_MASK 0xFFFFull
+#define SEND_CM_CREDIT_VL_DEDICATED_LIMIT_VL_SHIFT 0
+#define SEND_CM_CREDIT_VL_DEDICATED_LIMIT_VL_SMASK 0xFFFFull
+#define SEND_CM_CREDIT_VL_SHARED_LIMIT_VL_MASK 0xFFFFull
+#define SEND_CM_CREDIT_VL_SHARED_LIMIT_VL_SHIFT 16
+#define SEND_CM_CREDIT_VL_SHARED_LIMIT_VL_SMASK 0xFFFF0000ull
+#define SEND_CM_CTRL (TXE + 0x000000000500)
+#define SEND_CM_CTRL_FORCE_CREDIT_MODE_SMASK 0x8ull
+#define SEND_CM_CTRL_RESETCSR 0x0000000000000020ull
+#define SEND_CM_GLOBAL_CREDIT (TXE + 0x000000000508)
+#define SEND_CM_GLOBAL_CREDIT_AU_SHIFT 16
+#define SEND_CM_GLOBAL_CREDIT_RESETCSR 0x0000094000030000ull
+#define SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_MASK 0xFFFFull
+#define SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_SHIFT 0
+#define SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_SMASK 0xFFFFull
+#define SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_MASK 0xFFFFull
+#define SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_SHIFT 32
+#define SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_SMASK 0xFFFF00000000ull
+#define SEND_CM_LOCAL_AU_TABLE0_TO3 (TXE + 0x000000000520)
+#define SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE0_SHIFT 0
+#define SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE1_SHIFT 16
+#define SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE2_SHIFT 32
+#define SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE3_SHIFT 48
+#define SEND_CM_LOCAL_AU_TABLE4_TO7 (TXE + 0x000000000528)
+#define SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE4_SHIFT 0
+#define SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE5_SHIFT 16
+#define SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE6_SHIFT 32
+#define SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE7_SHIFT 48
+#define SEND_CM_REMOTE_AU_TABLE0_TO3 (TXE + 0x000000000530)
+#define SEND_CM_REMOTE_AU_TABLE4_TO7 (TXE + 0x000000000538)
+#define SEND_CM_TIMER_CTRL (TXE + 0x000000000518)
+#define SEND_CONTEXTS (TXE + 0x000000000010)
+#define SEND_CONTEXT_SET_CTRL (TXE + 0x000000000200)
+#define SEND_COUNTER_ARRAY32 (TXE + 0x000000000300)
+#define SEND_COUNTER_ARRAY64 (TXE + 0x000000000400)
+#define SEND_CTRL (TXE + 0x000000000000)
+#define SEND_CTRL_CM_RESET_SMASK 0x4ull
+#define SEND_CTRL_SEND_ENABLE_SMASK 0x1ull
+#define SEND_CTRL_VL_ARBITER_ENABLE_SMASK 0x2ull
+#define SEND_CTXT_CHECK_ENABLE (TXE + 0x000000100080)
+#define SEND_CTXT_CHECK_ENABLE_CHECK_BYPASS_VL_MAPPING_SMASK 0x80ull
+#define SEND_CTXT_CHECK_ENABLE_CHECK_ENABLE_SMASK 0x1ull
+#define SEND_CTXT_CHECK_ENABLE_CHECK_JOB_KEY_SMASK 0x4ull
+#define SEND_CTXT_CHECK_ENABLE_CHECK_OPCODE_SMASK 0x20ull
+#define SEND_CTXT_CHECK_ENABLE_CHECK_PARTITION_KEY_SMASK 0x8ull
+#define SEND_CTXT_CHECK_ENABLE_CHECK_SLID_SMASK 0x10ull
+#define SEND_CTXT_CHECK_ENABLE_CHECK_VL_MAPPING_SMASK 0x40ull
+#define SEND_CTXT_CHECK_ENABLE_CHECK_VL_SMASK 0x2ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_BAD_PKT_LEN_SMASK 0x20000ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_BYPASS_BAD_PKT_LEN_SMASK \
+ 0x200000ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_BYPASS_SMASK 0x800ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_GRH_SMASK 0x400ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_KDETH_PACKETS_SMASK 0x1000ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_NON_KDETH_PACKETS_SMASK 0x2000ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_PBC_STATIC_RATE_CONTROL_SMASK \
+ 0x100000ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_PBC_TEST_SMASK 0x10000ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_RAW_IPV6_SMASK 0x200ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_RAW_SMASK 0x100ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_TOO_LONG_BYPASS_PACKETS_SMASK \
+ 0x80000ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_TOO_LONG_IB_PACKETS_SMASK \
+ 0x40000ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_TOO_SMALL_BYPASS_PACKETS_SMASK \
+ 0x8000ull
+#define SEND_CTXT_CHECK_ENABLE_DISALLOW_TOO_SMALL_IB_PACKETS_SMASK \
+ 0x4000ull
+#define SEND_CTXT_CHECK_JOB_KEY (TXE + 0x000000100090)
+#define SEND_CTXT_CHECK_JOB_KEY_ALLOW_PERMISSIVE_SMASK 0x100000000ull
+#define SEND_CTXT_CHECK_JOB_KEY_MASK_SMASK 0xFFFF0000ull
+#define SEND_CTXT_CHECK_JOB_KEY_VALUE_MASK 0xFFFFull
+#define SEND_CTXT_CHECK_JOB_KEY_VALUE_SHIFT 0
+#define SEND_CTXT_CHECK_OPCODE (TXE + 0x0000001000A8)
+#define SEND_CTXT_CHECK_OPCODE_MASK_SHIFT 8
+#define SEND_CTXT_CHECK_OPCODE_VALUE_SHIFT 0
+#define SEND_CTXT_CHECK_PARTITION_KEY (TXE + 0x000000100098)
+#define SEND_CTXT_CHECK_PARTITION_KEY_VALUE_MASK 0xFFFFull
+#define SEND_CTXT_CHECK_PARTITION_KEY_VALUE_SHIFT 0
+#define SEND_CTXT_CHECK_SLID (TXE + 0x0000001000A0)
+#define SEND_CTXT_CHECK_SLID_MASK_MASK 0xFFFFull
+#define SEND_CTXT_CHECK_SLID_MASK_SHIFT 16
+#define SEND_CTXT_CHECK_SLID_VALUE_MASK 0xFFFFull
+#define SEND_CTXT_CHECK_SLID_VALUE_SHIFT 0
+#define SEND_CTXT_CHECK_VL (TXE + 0x000000100088)
+#define SEND_CTXT_CREDIT_CTRL (TXE + 0x000000100010)
+#define SEND_CTXT_CREDIT_CTRL_CREDIT_INTR_SMASK 0x20000ull
+#define SEND_CTXT_CREDIT_CTRL_EARLY_RETURN_SMASK 0x10000ull
+#define SEND_CTXT_CREDIT_CTRL_THRESHOLD_MASK 0x7FFull
+#define SEND_CTXT_CREDIT_CTRL_THRESHOLD_SHIFT 0
+#define SEND_CTXT_CREDIT_CTRL_THRESHOLD_SMASK 0x7FFull
+#define SEND_CTXT_CREDIT_FORCE (TXE + 0x000000100028)
+#define SEND_CTXT_CREDIT_FORCE_FORCE_RETURN_SMASK 0x1ull
+#define SEND_CTXT_CREDIT_RETURN_ADDR (TXE + 0x000000100020)
+#define SEND_CTXT_CREDIT_RETURN_ADDR_ADDRESS_SMASK 0xFFFFFFFFFFC0ull
+#define SEND_CTXT_CTRL (TXE + 0x000000100000)
+#define SEND_CTXT_CTRL_CTXT_BASE_MASK 0x3FFFull
+#define SEND_CTXT_CTRL_CTXT_BASE_SHIFT 32
+#define SEND_CTXT_CTRL_CTXT_DEPTH_MASK 0x7FFull
+#define SEND_CTXT_CTRL_CTXT_DEPTH_SHIFT 48
+#define SEND_CTXT_CTRL_CTXT_ENABLE_SMASK 0x1ull
+#define SEND_CTXT_ERR_CLEAR (TXE + 0x000000100050)
+#define SEND_CTXT_ERR_MASK (TXE + 0x000000100048)
+#define SEND_CTXT_ERR_STATUS (TXE + 0x000000100040)
+#define SEND_CTXT_ERR_STATUS_PIO_DISALLOWED_PACKET_ERR_SMASK 0x2ull
+#define SEND_CTXT_ERR_STATUS_PIO_INCONSISTENT_SOP_ERR_SMASK 0x1ull
+#define SEND_CTXT_ERR_STATUS_PIO_WRITE_CROSSES_BOUNDARY_ERR_SMASK 0x4ull
+#define SEND_CTXT_ERR_STATUS_PIO_WRITE_OUT_OF_BOUNDS_ERR_SMASK 0x10ull
+#define SEND_CTXT_ERR_STATUS_PIO_WRITE_OVERFLOW_ERR_SMASK 0x8ull
+#define SEND_CTXT_STATUS (TXE + 0x000000100008)
+#define SEND_CTXT_STATUS_CTXT_HALTED_SMASK 0x1ull
+#define SEND_DMA_BASE_ADDR (TXE + 0x000000200010)
+#define SEND_DMA_CHECK_ENABLE (TXE + 0x000000200080)
+#define SEND_DMA_CHECK_ENABLE_CHECK_BYPASS_VL_MAPPING_SMASK 0x80ull
+#define SEND_DMA_CHECK_ENABLE_CHECK_ENABLE_SMASK 0x1ull
+#define SEND_DMA_CHECK_ENABLE_CHECK_JOB_KEY_SMASK 0x4ull
+#define SEND_DMA_CHECK_ENABLE_CHECK_OPCODE_SMASK 0x20ull
+#define SEND_DMA_CHECK_ENABLE_CHECK_PARTITION_KEY_SMASK 0x8ull
+#define SEND_DMA_CHECK_ENABLE_CHECK_SLID_SMASK 0x10ull
+#define SEND_DMA_CHECK_ENABLE_CHECK_VL_MAPPING_SMASK 0x40ull
+#define SEND_DMA_CHECK_ENABLE_CHECK_VL_SMASK 0x2ull
+#define SEND_DMA_CHECK_ENABLE_DISALLOW_BAD_PKT_LEN_SMASK 0x20000ull
+#define SEND_DMA_CHECK_ENABLE_DISALLOW_BYPASS_BAD_PKT_LEN_SMASK 0x200000ull
+#define SEND_DMA_CHECK_ENABLE_DISALLOW_PBC_STATIC_RATE_CONTROL_SMASK \
+ 0x100000ull
+#define SEND_DMA_CHECK_ENABLE_DISALLOW_RAW_IPV6_SMASK 0x200ull
+#define SEND_DMA_CHECK_ENABLE_DISALLOW_RAW_SMASK 0x100ull
+#define SEND_DMA_CHECK_ENABLE_DISALLOW_TOO_LONG_BYPASS_PACKETS_SMASK \
+ 0x80000ull
+#define SEND_DMA_CHECK_ENABLE_DISALLOW_TOO_LONG_IB_PACKETS_SMASK 0x40000ull
+#define SEND_DMA_CHECK_ENABLE_DISALLOW_TOO_SMALL_BYPASS_PACKETS_SMASK \
+ 0x8000ull
+#define SEND_DMA_CHECK_ENABLE_DISALLOW_TOO_SMALL_IB_PACKETS_SMASK 0x4000ull
+#define SEND_DMA_CHECK_JOB_KEY (TXE + 0x000000200090)
+#define SEND_DMA_CHECK_OPCODE (TXE + 0x0000002000A8)
+#define SEND_DMA_CHECK_PARTITION_KEY (TXE + 0x000000200098)
+#define SEND_DMA_CHECK_SLID (TXE + 0x0000002000A0)
+#define SEND_DMA_CHECK_SLID_MASK_MASK 0xFFFFull
+#define SEND_DMA_CHECK_SLID_MASK_SHIFT 16
+#define SEND_DMA_CHECK_SLID_VALUE_MASK 0xFFFFull
+#define SEND_DMA_CHECK_SLID_VALUE_SHIFT 0
+#define SEND_DMA_CHECK_VL (TXE + 0x000000200088)
+#define SEND_DMA_CTRL (TXE + 0x000000200000)
+#define SEND_DMA_CTRL_SDMA_CLEANUP_SMASK 0x4ull
+#define SEND_DMA_CTRL_SDMA_ENABLE_SMASK 0x1ull
+#define SEND_DMA_CTRL_SDMA_HALT_SMASK 0x2ull
+#define SEND_DMA_CTRL_SDMA_INT_ENABLE_SMASK 0x8ull
+#define SEND_DMA_DESC_CNT (TXE + 0x000000200050)
+#define SEND_DMA_DESC_CNT_CNT_MASK 0xFFFFull
+#define SEND_DMA_DESC_CNT_CNT_SHIFT 0
+#define SEND_DMA_ENG_ERR_CLEAR (TXE + 0x000000200070)
+#define SEND_DMA_ENG_ERR_CLEAR_SDMA_HEADER_REQUEST_FIFO_UNC_ERR_MASK 0x1ull
+#define SEND_DMA_ENG_ERR_CLEAR_SDMA_HEADER_REQUEST_FIFO_UNC_ERR_SHIFT 18
+#define SEND_DMA_ENG_ERR_MASK (TXE + 0x000000200068)
+#define SEND_DMA_ENG_ERR_STATUS (TXE + 0x000000200060)
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_ASSEMBLY_UNC_ERR_SMASK 0x8000ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_DESC_TABLE_UNC_ERR_SMASK 0x4000ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_FIRST_DESC_ERR_SMASK 0x10ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_GEN_MISMATCH_ERR_SMASK 0x2ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_HALT_ERR_SMASK 0x40ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_HEADER_ADDRESS_ERR_SMASK 0x800ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_HEADER_LENGTH_ERR_SMASK 0x1000ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_HEADER_REQUEST_FIFO_UNC_ERR_SMASK \
+ 0x40000ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_HEADER_SELECT_ERR_SMASK 0x400ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_HEADER_STORAGE_UNC_ERR_SMASK \
+ 0x20000ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_LENGTH_MISMATCH_ERR_SMASK 0x80ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_MEM_READ_ERR_SMASK 0x20ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_PACKET_DESC_OVERFLOW_ERR_SMASK \
+ 0x100ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_PACKET_TRACKING_UNC_ERR_SMASK \
+ 0x10000ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_TAIL_OUT_OF_BOUNDS_ERR_SMASK 0x8ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_TIMEOUT_ERR_SMASK 0x2000ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_TOO_LONG_ERR_SMASK 0x4ull
+#define SEND_DMA_ENG_ERR_STATUS_SDMA_WRONG_DW_ERR_SMASK 0x1ull
+#define SEND_DMA_ENGINES (TXE + 0x000000000018)
+#define SEND_DMA_ERR_CLEAR (TXE + 0x000000000070)
+#define SEND_DMA_ERR_MASK (TXE + 0x000000000068)
+#define SEND_DMA_ERR_STATUS (TXE + 0x000000000060)
+#define SEND_DMA_ERR_STATUS_SDMA_CSR_PARITY_ERR_SMASK 0x2ull
+#define SEND_DMA_ERR_STATUS_SDMA_PCIE_REQ_TRACKING_COR_ERR_SMASK 0x8ull
+#define SEND_DMA_ERR_STATUS_SDMA_PCIE_REQ_TRACKING_UNC_ERR_SMASK 0x4ull
+#define SEND_DMA_ERR_STATUS_SDMA_RPY_TAG_ERR_SMASK 0x1ull
+#define SEND_DMA_HEAD (TXE + 0x000000200028)
+#define SEND_DMA_HEAD_ADDR (TXE + 0x000000200030)
+#define SEND_DMA_LEN_GEN (TXE + 0x000000200018)
+#define SEND_DMA_LEN_GEN_GENERATION_SHIFT 16
+#define SEND_DMA_LEN_GEN_LENGTH_SHIFT 6
+#define SEND_DMA_MEMORY (TXE + 0x0000002000B0)
+#define SEND_DMA_MEMORY_SDMA_MEMORY_CNT_SHIFT 16
+#define SEND_DMA_MEMORY_SDMA_MEMORY_INDEX_SHIFT 0
+#define SEND_DMA_MEM_SIZE (TXE + 0x000000000028)
+#define SEND_DMA_PRIORITY_THLD (TXE + 0x000000200038)
+#define SEND_DMA_RELOAD_CNT (TXE + 0x000000200048)
+#define SEND_DMA_STATUS (TXE + 0x000000200008)
+#define SEND_DMA_STATUS_ENG_CLEANED_UP_SMASK 0x200000000000000ull
+#define SEND_DMA_STATUS_ENG_HALTED_SMASK 0x100000000000000ull
+#define SEND_DMA_TAIL (TXE + 0x000000200020)
+#define SEND_EGRESS_CTXT_STATUS (TXE + 0x000000000800)
+#define SEND_EGRESS_CTXT_STATUS_CTXT_EGRESS_HALT_STATUS_SMASK 0x10000ull
+#define SEND_EGRESS_CTXT_STATUS_CTXT_EGRESS_PACKET_OCCUPANCY_SHIFT 0
+#define SEND_EGRESS_CTXT_STATUS_CTXT_EGRESS_PACKET_OCCUPANCY_SMASK \
+ 0x3FFFull
+#define SEND_EGRESS_ERR_CLEAR (TXE + 0x000000000090)
+#define SEND_EGRESS_ERR_INFO (TXE + 0x000000000F00)
+#define SEND_EGRESS_ERR_INFO_BAD_PKT_LEN_ERR_SMASK 0x20000ull
+#define SEND_EGRESS_ERR_INFO_BYPASS_ERR_SMASK 0x800ull
+#define SEND_EGRESS_ERR_INFO_GRH_ERR_SMASK 0x400ull
+#define SEND_EGRESS_ERR_INFO_JOB_KEY_ERR_SMASK 0x4ull
+#define SEND_EGRESS_ERR_INFO_KDETH_PACKETS_ERR_SMASK 0x1000ull
+#define SEND_EGRESS_ERR_INFO_NON_KDETH_PACKETS_ERR_SMASK 0x2000ull
+#define SEND_EGRESS_ERR_INFO_OPCODE_ERR_SMASK 0x20ull
+#define SEND_EGRESS_ERR_INFO_PARTITION_KEY_ERR_SMASK 0x8ull
+#define SEND_EGRESS_ERR_INFO_PBC_STATIC_RATE_CONTROL_ERR_SMASK 0x100000ull
+#define SEND_EGRESS_ERR_INFO_PBC_TEST_ERR_SMASK 0x10000ull
+#define SEND_EGRESS_ERR_INFO_RAW_ERR_SMASK 0x100ull
+#define SEND_EGRESS_ERR_INFO_RAW_IPV6_ERR_SMASK 0x200ull
+#define SEND_EGRESS_ERR_INFO_SLID_ERR_SMASK 0x10ull
+#define SEND_EGRESS_ERR_INFO_TOO_LONG_BYPASS_PACKETS_ERR_SMASK 0x80000ull
+#define SEND_EGRESS_ERR_INFO_TOO_LONG_IB_PACKET_ERR_SMASK 0x40000ull
+#define SEND_EGRESS_ERR_INFO_TOO_SMALL_BYPASS_PACKETS_ERR_SMASK 0x8000ull
+#define SEND_EGRESS_ERR_INFO_TOO_SMALL_IB_PACKETS_ERR_SMASK 0x4000ull
+#define SEND_EGRESS_ERR_INFO_VL_ERR_SMASK 0x2ull
+#define SEND_EGRESS_ERR_INFO_VL_MAPPING_ERR_SMASK 0x40ull
+#define SEND_EGRESS_ERR_MASK (TXE + 0x000000000088)
+#define SEND_EGRESS_ERR_SOURCE (TXE + 0x000000000F08)
+#define SEND_EGRESS_ERR_STATUS (TXE + 0x000000000080)
+#define SEND_EGRESS_ERR_STATUS_TX_CONFIG_PARITY_ERR_SMASK 0x8000ull
+#define SEND_EGRESS_ERR_STATUS_TX_CREDIT_OVERRUN_ERR_SMASK \
+ 0x200000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_CREDIT_RETURN_PARITY_ERR_SMASK \
+ 0x20000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_CREDIT_RETURN_VL_ERR_SMASK \
+ 0x800000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_EGRESS_FIFO_COR_ERR_SMASK \
+ 0x2000000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_EGRESS_FIFO_UNC_ERR_SMASK \
+ 0x200000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_EGRESS_FIFO_UNDERRUN_OR_PARITY_ERR_SMASK \
+ 0x8ull
+#define SEND_EGRESS_ERR_STATUS_TX_HCRC_INSERTION_ERR_SMASK \
+ 0x400000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_ILLEGAL_VL_ERR_SMASK 0x1000ull
+#define SEND_EGRESS_ERR_STATUS_TX_INCORRECT_LINK_STATE_ERR_SMASK 0x20ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_CSR_PARITY_ERR_SMASK 0x2000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO0_COR_ERR_SMASK \
+ 0x1000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO0_UNC_OR_PARITY_ERR_SMASK \
+ 0x100000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO1_COR_ERR_SMASK \
+ 0x2000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO1_UNC_OR_PARITY_ERR_SMASK \
+ 0x200000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO2_COR_ERR_SMASK \
+ 0x4000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO2_UNC_OR_PARITY_ERR_SMASK \
+ 0x400000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO3_COR_ERR_SMASK \
+ 0x8000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO3_UNC_OR_PARITY_ERR_SMASK \
+ 0x800000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO4_COR_ERR_SMASK \
+ 0x10000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO4_UNC_OR_PARITY_ERR_SMASK \
+ 0x1000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO5_COR_ERR_SMASK \
+ 0x20000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO5_UNC_OR_PARITY_ERR_SMASK \
+ 0x2000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO6_COR_ERR_SMASK \
+ 0x40000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO6_UNC_OR_PARITY_ERR_SMASK \
+ 0x4000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO7_COR_ERR_SMASK \
+ 0x80000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO7_UNC_OR_PARITY_ERR_SMASK \
+ 0x8000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO8_COR_ERR_SMASK \
+ 0x100000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LAUNCH_FIFO8_UNC_OR_PARITY_ERR_SMASK \
+ 0x10000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_LINKDOWN_ERR_SMASK 0x10ull
+#define SEND_EGRESS_ERR_STATUS_TX_PIO_LAUNCH_INTF_PARITY_ERR_SMASK 0x80ull
+#define SEND_EGRESS_ERR_STATUS_TX_PKT_INTEGRITY_MEM_COR_ERR_SMASK 0x1ull
+#define SEND_EGRESS_ERR_STATUS_TX_PKT_INTEGRITY_MEM_UNC_ERR_SMASK 0x2ull
+#define SEND_EGRESS_ERR_STATUS_TX_READ_PIO_MEMORY_COR_ERR_SMASK \
+ 0x1000000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_READ_PIO_MEMORY_CSR_UNC_ERR_SMASK \
+ 0x8000000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_READ_PIO_MEMORY_UNC_ERR_SMASK \
+ 0x100000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_READ_SDMA_MEMORY_COR_ERR_SMASK \
+ 0x800000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_READ_SDMA_MEMORY_CSR_UNC_ERR_SMASK \
+ 0x4000000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_READ_SDMA_MEMORY_UNC_ERR_SMASK \
+ 0x80000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SB_HDR_COR_ERR_SMASK 0x400000000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SB_HDR_UNC_ERR_SMASK 0x40000000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SBRD_CTL_CSR_PARITY_ERR_SMASK 0x4000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SBRD_CTL_STATE_MACHINE_PARITY_ERR_SMASK \
+ 0x800ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA0_DISALLOWED_PACKET_ERR_SMASK \
+ 0x10000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA10_DISALLOWED_PACKET_ERR_SMASK \
+ 0x4000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA11_DISALLOWED_PACKET_ERR_SMASK \
+ 0x8000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA12_DISALLOWED_PACKET_ERR_SMASK \
+ 0x10000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA13_DISALLOWED_PACKET_ERR_SMASK \
+ 0x20000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA14_DISALLOWED_PACKET_ERR_SMASK \
+ 0x40000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA15_DISALLOWED_PACKET_ERR_SMASK \
+ 0x80000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA1_DISALLOWED_PACKET_ERR_SMASK \
+ 0x20000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA2_DISALLOWED_PACKET_ERR_SMASK \
+ 0x40000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA3_DISALLOWED_PACKET_ERR_SMASK \
+ 0x80000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA4_DISALLOWED_PACKET_ERR_SMASK \
+ 0x100000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA5_DISALLOWED_PACKET_ERR_SMASK \
+ 0x200000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA6_DISALLOWED_PACKET_ERR_SMASK \
+ 0x400000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA7_DISALLOWED_PACKET_ERR_SMASK \
+ 0x800000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA8_DISALLOWED_PACKET_ERR_SMASK \
+ 0x1000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA9_DISALLOWED_PACKET_ERR_SMASK \
+ 0x2000000ull
+#define SEND_EGRESS_ERR_STATUS_TX_SDMA_LAUNCH_INTF_PARITY_ERR_SMASK \
+ 0x100ull
+#define SEND_EGRESS_SEND_DMA_STATUS (TXE + 0x000000000E00)
+#define SEND_EGRESS_SEND_DMA_STATUS_SDMA_EGRESS_PACKET_OCCUPANCY_SHIFT 0
+#define SEND_EGRESS_SEND_DMA_STATUS_SDMA_EGRESS_PACKET_OCCUPANCY_SMASK \
+ 0x3FFFull
+#define SEND_ERR_CLEAR (TXE + 0x0000000000F0)
+#define SEND_ERR_MASK (TXE + 0x0000000000E8)
+#define SEND_ERR_STATUS (TXE + 0x0000000000E0)
+#define SEND_ERR_STATUS_SEND_CSR_PARITY_ERR_SMASK 0x1ull
+#define SEND_ERR_STATUS_SEND_CSR_READ_BAD_ADDR_ERR_SMASK 0x2ull
+#define SEND_ERR_STATUS_SEND_CSR_WRITE_BAD_ADDR_ERR_SMASK 0x4ull
+#define SEND_HIGH_PRIORITY_LIMIT (TXE + 0x000000000030)
+#define SEND_HIGH_PRIORITY_LIMIT_LIMIT_MASK 0x3FFFull
+#define SEND_HIGH_PRIORITY_LIMIT_LIMIT_SHIFT 0
+#define SEND_HIGH_PRIORITY_LIST (TXE + 0x000000000180)
+#define SEND_LEN_CHECK0 (TXE + 0x0000000000D0)
+#define SEND_LEN_CHECK0_LEN_VL0_MASK 0xFFFull
+#define SEND_LEN_CHECK0_LEN_VL1_SHIFT 12
+#define SEND_LEN_CHECK1 (TXE + 0x0000000000D8)
+#define SEND_LEN_CHECK1_LEN_VL15_MASK 0xFFFull
+#define SEND_LEN_CHECK1_LEN_VL15_SHIFT 48
+#define SEND_LEN_CHECK1_LEN_VL4_MASK 0xFFFull
+#define SEND_LEN_CHECK1_LEN_VL5_SHIFT 12
+#define SEND_LOW_PRIORITY_LIST (TXE + 0x000000000100)
+#define SEND_LOW_PRIORITY_LIST_VL_MASK 0x7ull
+#define SEND_LOW_PRIORITY_LIST_VL_SHIFT 16
+#define SEND_LOW_PRIORITY_LIST_WEIGHT_MASK 0xFFull
+#define SEND_LOW_PRIORITY_LIST_WEIGHT_SHIFT 0
+#define SEND_PIO_ERR_CLEAR (TXE + 0x000000000050)
+#define SEND_PIO_ERR_CLEAR_PIO_INIT_SM_IN_ERR_SMASK 0x20000ull
+#define SEND_PIO_ERR_MASK (TXE + 0x000000000048)
+#define SEND_PIO_ERR_STATUS (TXE + 0x000000000040)
+#define SEND_PIO_ERR_STATUS_PIO_BLOCK_QW_COUNT_PARITY_ERR_SMASK \
+ 0x1000000ull
+#define SEND_PIO_ERR_STATUS_PIO_CREDIT_RET_FIFO_PARITY_ERR_SMASK 0x8000ull
+#define SEND_PIO_ERR_STATUS_PIO_CSR_PARITY_ERR_SMASK 0x4ull
+#define SEND_PIO_ERR_STATUS_PIO_CURRENT_FREE_CNT_PARITY_ERR_SMASK \
+ 0x100000000ull
+#define SEND_PIO_ERR_STATUS_PIO_HOST_ADDR_MEM_COR_ERR_SMASK 0x100000ull
+#define SEND_PIO_ERR_STATUS_PIO_HOST_ADDR_MEM_UNC_ERR_SMASK 0x80000ull
+#define SEND_PIO_ERR_STATUS_PIO_INIT_SM_IN_ERR_SMASK 0x20000ull
+#define SEND_PIO_ERR_STATUS_PIO_LAST_RETURNED_CNT_PARITY_ERR_SMASK \
+ 0x200000000ull
+#define SEND_PIO_ERR_STATUS_PIO_PCC_FIFO_PARITY_ERR_SMASK 0x20ull
+#define SEND_PIO_ERR_STATUS_PIO_PCC_SOP_HEAD_PARITY_ERR_SMASK \
+ 0x400000000ull
+#define SEND_PIO_ERR_STATUS_PIO_PEC_FIFO_PARITY_ERR_SMASK 0x40ull
+#define SEND_PIO_ERR_STATUS_PIO_PEC_SOP_HEAD_PARITY_ERR_SMASK \
+ 0x800000000ull
+#define SEND_PIO_ERR_STATUS_PIO_PKT_EVICT_FIFO_PARITY_ERR_SMASK 0x200ull
+#define SEND_PIO_ERR_STATUS_PIO_PKT_EVICT_SM_OR_ARB_SM_ERR_SMASK 0x40000ull
+#define SEND_PIO_ERR_STATUS_PIO_PPMC_BQC_MEM_PARITY_ERR_SMASK 0x10000000ull
+#define SEND_PIO_ERR_STATUS_PIO_PPMC_PBL_FIFO_ERR_SMASK 0x10000ull
+#define SEND_PIO_ERR_STATUS_PIO_PPMC_SOP_LEN_ERR_SMASK 0x20000000ull
+#define SEND_PIO_ERR_STATUS_PIO_SB_MEM_FIFO0_ERR_SMASK 0x8ull
+#define SEND_PIO_ERR_STATUS_PIO_SB_MEM_FIFO1_ERR_SMASK 0x10ull
+#define SEND_PIO_ERR_STATUS_PIO_SBRDCTL_CRREL_PARITY_ERR_SMASK 0x80ull
+#define SEND_PIO_ERR_STATUS_PIO_SBRDCTRL_CRREL_FIFO_PARITY_ERR_SMASK \
+ 0x100ull
+#define SEND_PIO_ERR_STATUS_PIO_SM_PKT_RESET_PARITY_ERR_SMASK 0x400ull
+#define SEND_PIO_ERR_STATUS_PIO_STATE_MACHINE_ERR_SMASK 0x400000ull
+#define SEND_PIO_ERR_STATUS_PIO_VL_FIFO_PARITY_ERR_SMASK 0x8000000ull
+#define SEND_PIO_ERR_STATUS_PIO_VLF_SOP_PARITY_ERR_SMASK 0x4000000ull
+#define SEND_PIO_ERR_STATUS_PIO_VLF_VL_LEN_PARITY_ERR_SMASK 0x2000000ull
+#define SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK0_COR_ERR_SMASK 0x2000ull
+#define SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK0_UNC_ERR_SMASK 0x800ull
+#define SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK1_COR_ERR_SMASK 0x4000ull
+#define SEND_PIO_ERR_STATUS_PIO_VL_LEN_MEM_BANK1_UNC_ERR_SMASK 0x1000ull
+#define SEND_PIO_ERR_STATUS_PIO_WRITE_ADDR_PARITY_ERR_SMASK 0x2ull
+#define SEND_PIO_ERR_STATUS_PIO_WRITE_BAD_CTXT_ERR_SMASK 0x1ull
+#define SEND_PIO_ERR_STATUS_PIO_WRITE_DATA_PARITY_ERR_SMASK 0x200000ull
+#define SEND_PIO_ERR_STATUS_PIO_WRITE_QW_VALID_PARITY_ERR_SMASK 0x800000ull
+#define SEND_PIO_INIT_CTXT (TXE + 0x000000000038)
+#define SEND_PIO_INIT_CTXT_PIO_ALL_CTXT_INIT_SMASK 0x1ull
+#define SEND_PIO_INIT_CTXT_PIO_CTXT_NUM_MASK 0xFFull
+#define SEND_PIO_INIT_CTXT_PIO_CTXT_NUM_SHIFT 8
+#define SEND_PIO_INIT_CTXT_PIO_INIT_ERR_SMASK 0x8ull
+#define SEND_PIO_INIT_CTXT_PIO_INIT_IN_PROGRESS_SMASK 0x4ull
+#define SEND_PIO_INIT_CTXT_PIO_SINGLE_CTXT_INIT_SMASK 0x2ull
+#define SEND_PIO_MEM_SIZE (TXE + 0x000000000020)
+#define SEND_SC2VLT0 (TXE + 0x0000000000B0)
+#define SEND_SC2VLT0_SC0_SHIFT 0
+#define SEND_SC2VLT0_SC1_SHIFT 8
+#define SEND_SC2VLT0_SC2_SHIFT 16
+#define SEND_SC2VLT0_SC3_SHIFT 24
+#define SEND_SC2VLT0_SC4_SHIFT 32
+#define SEND_SC2VLT0_SC5_SHIFT 40
+#define SEND_SC2VLT0_SC6_SHIFT 48
+#define SEND_SC2VLT0_SC7_SHIFT 56
+#define SEND_SC2VLT1 (TXE + 0x0000000000B8)
+#define SEND_SC2VLT1_SC10_SHIFT 16
+#define SEND_SC2VLT1_SC11_SHIFT 24
+#define SEND_SC2VLT1_SC12_SHIFT 32
+#define SEND_SC2VLT1_SC13_SHIFT 40
+#define SEND_SC2VLT1_SC14_SHIFT 48
+#define SEND_SC2VLT1_SC15_SHIFT 56
+#define SEND_SC2VLT1_SC8_SHIFT 0
+#define SEND_SC2VLT1_SC9_SHIFT 8
+#define SEND_SC2VLT2 (TXE + 0x0000000000C0)
+#define SEND_SC2VLT2_SC16_SHIFT 0
+#define SEND_SC2VLT2_SC17_SHIFT 8
+#define SEND_SC2VLT2_SC18_SHIFT 16
+#define SEND_SC2VLT2_SC19_SHIFT 24
+#define SEND_SC2VLT2_SC20_SHIFT 32
+#define SEND_SC2VLT2_SC21_SHIFT 40
+#define SEND_SC2VLT2_SC22_SHIFT 48
+#define SEND_SC2VLT2_SC23_SHIFT 56
+#define SEND_SC2VLT3 (TXE + 0x0000000000C8)
+#define SEND_SC2VLT3_SC24_SHIFT 0
+#define SEND_SC2VLT3_SC25_SHIFT 8
+#define SEND_SC2VLT3_SC26_SHIFT 16
+#define SEND_SC2VLT3_SC27_SHIFT 24
+#define SEND_SC2VLT3_SC28_SHIFT 32
+#define SEND_SC2VLT3_SC29_SHIFT 40
+#define SEND_SC2VLT3_SC30_SHIFT 48
+#define SEND_SC2VLT3_SC31_SHIFT 56
+#define SEND_STATIC_RATE_CONTROL (TXE + 0x0000000000A8)
+#define SEND_STATIC_RATE_CONTROL_CSR_SRC_RELOAD_SHIFT 0
+#define SEND_STATIC_RATE_CONTROL_CSR_SRC_RELOAD_SMASK 0xFFFFull
+#define PCIE_CFG_REG_PL2 (PCIE + 0x000000000708)
+#define PCIE_CFG_REG_PL102 (PCIE + 0x000000000898)
+#define PCIE_CFG_REG_PL102_GEN3_EQ_POST_CURSOR_PSET_SHIFT 12
+#define PCIE_CFG_REG_PL102_GEN3_EQ_CURSOR_PSET_SHIFT 6
+#define PCIE_CFG_REG_PL102_GEN3_EQ_PRE_CURSOR_PSET_SHIFT 0
+#define PCIE_CFG_REG_PL103 (PCIE + 0x00000000089C)
+#define PCIE_CFG_REG_PL105 (PCIE + 0x0000000008A4)
+#define PCIE_CFG_REG_PL105_GEN3_EQ_VIOLATE_COEF_RULES_SMASK 0x1ull
+#define PCIE_CFG_REG_PL2_LOW_PWR_ENT_CNT_SHIFT 24
+#define PCIE_CFG_REG_PL100 (PCIE + 0x000000000890)
+#define PCIE_CFG_REG_PL100_EQ_EIEOS_CNT_SMASK 0x400ull
+#define PCIE_CFG_REG_PL101 (PCIE + 0x000000000894)
+#define PCIE_CFG_REG_PL101_GEN3_EQ_LOCAL_FS_SHIFT 6
+#define PCIE_CFG_REG_PL101_GEN3_EQ_LOCAL_LF_SHIFT 0
+#define PCIE_CFG_REG_PL106 (PCIE + 0x0000000008A8)
+#define PCIE_CFG_REG_PL106_GEN3_EQ_PSET_REQ_VEC_SHIFT 8
+#define PCIE_CFG_REG_PL106_GEN3_EQ_EVAL2MS_DISABLE_SMASK 0x20ull
+#define PCIE_CFG_REG_PL106_GEN3_EQ_PHASE23_EXIT_MODE_SMASK 0x10ull
+
+#endif /* DEF_CHIP_REG */
diff --git a/drivers/staging/rdma/hfi1/common.h b/drivers/staging/rdma/hfi1/common.h
new file mode 100644
index 000000000000..5f2293729cf9
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/common.h
@@ -0,0 +1,415 @@
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#ifndef _COMMON_H
+#define _COMMON_H
+
+#include <rdma/hfi/hfi1_user.h>
+
+/*
+ * This file contains defines, structures, etc. that are used
+ * to communicate between kernel and user code.
+ */
+
+/* version of protocol header (known to chip also). In the long run,
+ * we should be able to generate and accept a range of version numbers;
+ * for now we only accept one, and it's compiled in.
+ */
+#define IPS_PROTO_VERSION 2
+
+/*
+ * These are compile time constants that you may want to enable or disable
+ * if you are trying to debug problems with code or performance.
+ * HFI1_VERBOSE_TRACING define as 1 if you want additional tracing in
+ * fast path code
+ * HFI1_TRACE_REGWRITES define as 1 if you want register writes to be
+ * traced in fast path code
+ * _HFI1_TRACING define as 0 if you want to remove all tracing in a
+ * compilation unit
+ */
+
+/*
+ * If a packet's QP[23:16] bits match this value, then it is
+ * a PSM packet and the hardware will expect a KDETH header
+ * following the BTH.
+ */
+#define DEFAULT_KDETH_QP 0x80
+
+/* driver/hw feature set bitmask */
+#define HFI1_CAP_USER_SHIFT 24
+#define HFI1_CAP_MASK ((1UL << HFI1_CAP_USER_SHIFT) - 1)
+/* locked flag - if set, only HFI1_CAP_WRITABLE_MASK bits can be set */
+#define HFI1_CAP_LOCKED_SHIFT 63
+#define HFI1_CAP_LOCKED_MASK 0x1ULL
+#define HFI1_CAP_LOCKED_SMASK (HFI1_CAP_LOCKED_MASK << HFI1_CAP_LOCKED_SHIFT)
+/* extra bits used between kernel and user processes */
+#define HFI1_CAP_MISC_SHIFT (HFI1_CAP_USER_SHIFT * 2)
+#define HFI1_CAP_MISC_MASK ((1ULL << (HFI1_CAP_LOCKED_SHIFT - \
+ HFI1_CAP_MISC_SHIFT)) - 1)
+
+#define HFI1_CAP_KSET(cap) ({ hfi1_cap_mask |= HFI1_CAP_##cap; hfi1_cap_mask; })
+#define HFI1_CAP_KCLEAR(cap) \
+ ({ \
+ hfi1_cap_mask &= ~HFI1_CAP_##cap; \
+ hfi1_cap_mask; \
+ })
+#define HFI1_CAP_USET(cap) \
+ ({ \
+ hfi1_cap_mask |= (HFI1_CAP_##cap << HFI1_CAP_USER_SHIFT); \
+ hfi1_cap_mask; \
+ })
+#define HFI1_CAP_UCLEAR(cap) \
+ ({ \
+ hfi1_cap_mask &= ~(HFI1_CAP_##cap << HFI1_CAP_USER_SHIFT); \
+ hfi1_cap_mask; \
+ })
+#define HFI1_CAP_SET(cap) \
+ ({ \
+ hfi1_cap_mask |= (HFI1_CAP_##cap | (HFI1_CAP_##cap << \
+ HFI1_CAP_USER_SHIFT)); \
+ hfi1_cap_mask; \
+ })
+#define HFI1_CAP_CLEAR(cap) \
+ ({ \
+ hfi1_cap_mask &= ~(HFI1_CAP_##cap | \
+ (HFI1_CAP_##cap << HFI1_CAP_USER_SHIFT)); \
+ hfi1_cap_mask; \
+ })
+#define HFI1_CAP_LOCK() \
+ ({ hfi1_cap_mask |= HFI1_CAP_LOCKED_SMASK; hfi1_cap_mask; })
+#define HFI1_CAP_LOCKED() (!!(hfi1_cap_mask & HFI1_CAP_LOCKED_SMASK))
+/*
+ * The set of capability bits that can be changed after initial load
+ * This set is the same for kernel and user contexts. However, for
+ * user contexts, the set can be further filtered by using the
+ * HFI1_CAP_RESERVED_MASK bits.
+ */
+#define HFI1_CAP_WRITABLE_MASK (HFI1_CAP_SDMA_AHG | \
+ HFI1_CAP_HDRSUPP | \
+ HFI1_CAP_MULTI_PKT_EGR | \
+ HFI1_CAP_NODROP_RHQ_FULL | \
+ HFI1_CAP_NODROP_EGR_FULL | \
+ HFI1_CAP_ALLOW_PERM_JKEY | \
+ HFI1_CAP_STATIC_RATE_CTRL | \
+ HFI1_CAP_PRINT_UNIMPL)
+/*
+ * A set of capability bits that are "global" and are not allowed to be
+ * set in the user bitmask.
+ */
+#define HFI1_CAP_RESERVED_MASK ((HFI1_CAP_SDMA | \
+ HFI1_CAP_USE_SDMA_HEAD | \
+ HFI1_CAP_EXTENDED_PSN | \
+ HFI1_CAP_PRINT_UNIMPL | \
+ HFI1_CAP_QSFP_ENABLED | \
+ HFI1_CAP_NO_INTEGRITY | \
+ HFI1_CAP_PKEY_CHECK) << \
+ HFI1_CAP_USER_SHIFT)
+/*
+ * Set of capabilities that need to be enabled for kernel context in
+ * order to be allowed for user contexts, as well.
+ */
+#define HFI1_CAP_MUST_HAVE_KERN (HFI1_CAP_STATIC_RATE_CTRL)
+/* Default enabled capabilities (both kernel and user) */
+#define HFI1_CAP_MASK_DEFAULT (HFI1_CAP_HDRSUPP | \
+ HFI1_CAP_NODROP_RHQ_FULL | \
+ HFI1_CAP_NODROP_EGR_FULL | \
+ HFI1_CAP_SDMA | \
+ HFI1_CAP_PRINT_UNIMPL | \
+ HFI1_CAP_STATIC_RATE_CTRL | \
+ HFI1_CAP_QSFP_ENABLED | \
+ HFI1_CAP_PKEY_CHECK | \
+ HFI1_CAP_MULTI_PKT_EGR | \
+ HFI1_CAP_EXTENDED_PSN | \
+ ((HFI1_CAP_HDRSUPP | \
+ HFI1_CAP_MULTI_PKT_EGR | \
+ HFI1_CAP_STATIC_RATE_CTRL | \
+ HFI1_CAP_PKEY_CHECK | \
+ HFI1_CAP_EARLY_CREDIT_RETURN) << \
+ HFI1_CAP_USER_SHIFT))
+/*
+ * A bitmask of kernel/global capabilities that should be communicated
+ * to user level processes.
+ */
+#define HFI1_CAP_K2U (HFI1_CAP_SDMA | \
+ HFI1_CAP_EXTENDED_PSN | \
+ HFI1_CAP_PKEY_CHECK | \
+ HFI1_CAP_NO_INTEGRITY)
+
+#define HFI1_USER_SWVERSION ((HFI1_USER_SWMAJOR << 16) | HFI1_USER_SWMINOR)
+
+#ifndef HFI1_KERN_TYPE
+#define HFI1_KERN_TYPE 0
+#endif
+
+/*
+ * Similarly, this is the kernel version going back to the user. It's
+ * slightly different, in that we want to tell if the driver was built as
+ * part of a Intel release, or from the driver from openfabrics.org,
+ * kernel.org, or a standard distribution, for support reasons.
+ * The high bit is 0 for non-Intel and 1 for Intel-built/supplied.
+ *
+ * It's returned by the driver to the user code during initialization in the
+ * spi_sw_version field of hfi1_base_info, so the user code can in turn
+ * check for compatibility with the kernel.
+*/
+#define HFI1_KERN_SWVERSION ((HFI1_KERN_TYPE << 31) | HFI1_USER_SWVERSION)
+
+/*
+ * Define the driver version number. This is something that refers only
+ * to the driver itself, not the software interfaces it supports.
+ */
+#ifndef HFI1_DRIVER_VERSION_BASE
+#define HFI1_DRIVER_VERSION_BASE "0.9-248"
+#endif
+
+/* create the final driver version string */
+#ifdef HFI1_IDSTR
+#define HFI1_DRIVER_VERSION HFI1_DRIVER_VERSION_BASE " " HFI1_IDSTR
+#else
+#define HFI1_DRIVER_VERSION HFI1_DRIVER_VERSION_BASE
+#endif
+
+/*
+ * Diagnostics can send a packet by writing the following
+ * struct to the diag packet special file.
+ *
+ * This allows a custom PBC qword, so that special modes and deliberate
+ * changes to CRCs can be used.
+ */
+#define _DIAG_PKT_VERS 1
+struct diag_pkt {
+ __u16 version; /* structure version */
+ __u16 unit; /* which device */
+ __u16 sw_index; /* send sw index to use */
+ __u16 len; /* data length, in bytes */
+ __u16 port; /* port number */
+ __u16 unused;
+ __u32 flags; /* call flags */
+ __u64 data; /* user data pointer */
+ __u64 pbc; /* PBC for the packet */
+};
+
+/* diag_pkt flags */
+#define F_DIAGPKT_WAIT 0x1 /* wait until packet is sent */
+
+/*
+ * The next set of defines are for packet headers, and chip register
+ * and memory bits that are visible to and/or used by user-mode software.
+ */
+
+/*
+ * Receive Header Flags
+ */
+#define RHF_PKT_LEN_SHIFT 0
+#define RHF_PKT_LEN_MASK 0xfffull
+#define RHF_PKT_LEN_SMASK (RHF_PKT_LEN_MASK << RHF_PKT_LEN_SHIFT)
+
+#define RHF_RCV_TYPE_SHIFT 12
+#define RHF_RCV_TYPE_MASK 0x7ull
+#define RHF_RCV_TYPE_SMASK (RHF_RCV_TYPE_MASK << RHF_RCV_TYPE_SHIFT)
+
+#define RHF_USE_EGR_BFR_SHIFT 15
+#define RHF_USE_EGR_BFR_MASK 0x1ull
+#define RHF_USE_EGR_BFR_SMASK (RHF_USE_EGR_BFR_MASK << RHF_USE_EGR_BFR_SHIFT)
+
+#define RHF_EGR_INDEX_SHIFT 16
+#define RHF_EGR_INDEX_MASK 0x7ffull
+#define RHF_EGR_INDEX_SMASK (RHF_EGR_INDEX_MASK << RHF_EGR_INDEX_SHIFT)
+
+#define RHF_DC_INFO_SHIFT 27
+#define RHF_DC_INFO_MASK 0x1ull
+#define RHF_DC_INFO_SMASK (RHF_DC_INFO_MASK << RHF_DC_INFO_SHIFT)
+
+#define RHF_RCV_SEQ_SHIFT 28
+#define RHF_RCV_SEQ_MASK 0xfull
+#define RHF_RCV_SEQ_SMASK (RHF_RCV_SEQ_MASK << RHF_RCV_SEQ_SHIFT)
+
+#define RHF_EGR_OFFSET_SHIFT 32
+#define RHF_EGR_OFFSET_MASK 0xfffull
+#define RHF_EGR_OFFSET_SMASK (RHF_EGR_OFFSET_MASK << RHF_EGR_OFFSET_SHIFT)
+#define RHF_HDRQ_OFFSET_SHIFT 44
+#define RHF_HDRQ_OFFSET_MASK 0x1ffull
+#define RHF_HDRQ_OFFSET_SMASK (RHF_HDRQ_OFFSET_MASK << RHF_HDRQ_OFFSET_SHIFT)
+#define RHF_K_HDR_LEN_ERR (0x1ull << 53)
+#define RHF_DC_UNC_ERR (0x1ull << 54)
+#define RHF_DC_ERR (0x1ull << 55)
+#define RHF_RCV_TYPE_ERR_SHIFT 56
+#define RHF_RCV_TYPE_ERR_MASK 0x7ul
+#define RHF_RCV_TYPE_ERR_SMASK (RHF_RCV_TYPE_ERR_MASK << RHF_RCV_TYPE_ERR_SHIFT)
+#define RHF_TID_ERR (0x1ull << 59)
+#define RHF_LEN_ERR (0x1ull << 60)
+#define RHF_ECC_ERR (0x1ull << 61)
+#define RHF_VCRC_ERR (0x1ull << 62)
+#define RHF_ICRC_ERR (0x1ull << 63)
+
+#define RHF_ERROR_SMASK 0xffe0000000000000ull /* bits 63:53 */
+
+/* RHF receive types */
+#define RHF_RCV_TYPE_EXPECTED 0
+#define RHF_RCV_TYPE_EAGER 1
+#define RHF_RCV_TYPE_IB 2 /* normal IB, IB Raw, or IPv6 */
+#define RHF_RCV_TYPE_ERROR 3
+#define RHF_RCV_TYPE_BYPASS 4
+#define RHF_RCV_TYPE_INVALID5 5
+#define RHF_RCV_TYPE_INVALID6 6
+#define RHF_RCV_TYPE_INVALID7 7
+
+/* RHF receive type error - expected packet errors */
+#define RHF_RTE_EXPECTED_FLOW_SEQ_ERR 0x2
+#define RHF_RTE_EXPECTED_FLOW_GEN_ERR 0x4
+
+/* RHF receive type error - eager packet errors */
+#define RHF_RTE_EAGER_NO_ERR 0x0
+
+/* RHF receive type error - IB packet errors */
+#define RHF_RTE_IB_NO_ERR 0x0
+
+/* RHF receive type error - error packet errors */
+#define RHF_RTE_ERROR_NO_ERR 0x0
+#define RHF_RTE_ERROR_OP_CODE_ERR 0x1
+#define RHF_RTE_ERROR_KHDR_MIN_LEN_ERR 0x2
+#define RHF_RTE_ERROR_KHDR_HCRC_ERR 0x3
+#define RHF_RTE_ERROR_KHDR_KVER_ERR 0x4
+#define RHF_RTE_ERROR_CONTEXT_ERR 0x5
+#define RHF_RTE_ERROR_KHDR_TID_ERR 0x6
+
+/* RHF receive type error - bypass packet errors */
+#define RHF_RTE_BYPASS_NO_ERR 0x0
+
+/*
+ * This structure contains the first field common to all protocols
+ * that employ this chip.
+ */
+struct hfi1_message_header {
+ __be16 lrh[4];
+};
+
+/* IB - LRH header constants */
+#define HFI1_LRH_GRH 0x0003 /* 1. word of IB LRH - next header: GRH */
+#define HFI1_LRH_BTH 0x0002 /* 1. word of IB LRH - next header: BTH */
+
+/* misc. */
+#define SIZE_OF_CRC 1
+
+#define LIM_MGMT_P_KEY 0x7FFF
+#define FULL_MGMT_P_KEY 0xFFFF
+
+#define DEFAULT_P_KEY LIM_MGMT_P_KEY
+#define HFI1_PERMISSIVE_LID 0xFFFF
+#define HFI1_AETH_CREDIT_SHIFT 24
+#define HFI1_AETH_CREDIT_MASK 0x1F
+#define HFI1_AETH_CREDIT_INVAL 0x1F
+#define HFI1_MSN_MASK 0xFFFFFF
+#define HFI1_QPN_MASK 0xFFFFFF
+#define HFI1_FECN_SHIFT 31
+#define HFI1_FECN_MASK 1
+#define HFI1_FECN_SMASK (1 << HFI1_FECN_SHIFT)
+#define HFI1_BECN_SHIFT 30
+#define HFI1_BECN_MASK 1
+#define HFI1_BECN_SMASK (1 << HFI1_BECN_SHIFT)
+#define HFI1_MULTICAST_LID_BASE 0xC000
+
+static inline __u64 rhf_to_cpu(const __le32 *rbuf)
+{
+ return __le64_to_cpu(*((__le64 *)rbuf));
+}
+
+static inline u64 rhf_err_flags(u64 rhf)
+{
+ return rhf & RHF_ERROR_SMASK;
+}
+
+static inline u32 rhf_rcv_type(u64 rhf)
+{
+ return (rhf >> RHF_RCV_TYPE_SHIFT) & RHF_RCV_TYPE_MASK;
+}
+
+static inline u32 rhf_rcv_type_err(u64 rhf)
+{
+ return (rhf >> RHF_RCV_TYPE_ERR_SHIFT) & RHF_RCV_TYPE_ERR_MASK;
+}
+
+/* return size is in bytes, not DWORDs */
+static inline u32 rhf_pkt_len(u64 rhf)
+{
+ return ((rhf & RHF_PKT_LEN_SMASK) >> RHF_PKT_LEN_SHIFT) << 2;
+}
+
+static inline u32 rhf_egr_index(u64 rhf)
+{
+ return (rhf >> RHF_EGR_INDEX_SHIFT) & RHF_EGR_INDEX_MASK;
+}
+
+static inline u32 rhf_rcv_seq(u64 rhf)
+{
+ return (rhf >> RHF_RCV_SEQ_SHIFT) & RHF_RCV_SEQ_MASK;
+}
+
+/* returned offset is in DWORDS */
+static inline u32 rhf_hdrq_offset(u64 rhf)
+{
+ return (rhf >> RHF_HDRQ_OFFSET_SHIFT) & RHF_HDRQ_OFFSET_MASK;
+}
+
+static inline u64 rhf_use_egr_bfr(u64 rhf)
+{
+ return rhf & RHF_USE_EGR_BFR_SMASK;
+}
+
+static inline u64 rhf_dc_info(u64 rhf)
+{
+ return rhf & RHF_DC_INFO_SMASK;
+}
+
+static inline u32 rhf_egr_buf_offset(u64 rhf)
+{
+ return (rhf >> RHF_EGR_OFFSET_SHIFT) & RHF_EGR_OFFSET_MASK;
+}
+#endif /* _COMMON_H */
diff --git a/drivers/staging/rdma/hfi1/cq.c b/drivers/staging/rdma/hfi1/cq.c
new file mode 100644
index 000000000000..4f046ffe7e60
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/cq.c
@@ -0,0 +1,558 @@
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/kthread.h>
+
+#include "verbs.h"
+#include "hfi.h"
+
+/**
+ * hfi1_cq_enter - add a new entry to the completion queue
+ * @cq: completion queue
+ * @entry: work completion entry to add
+ * @sig: true if @entry is a solicited entry
+ *
+ * This may be called with qp->s_lock held.
+ */
+void hfi1_cq_enter(struct hfi1_cq *cq, struct ib_wc *entry, int solicited)
+{
+ struct hfi1_cq_wc *wc;
+ unsigned long flags;
+ u32 head;
+ u32 next;
+
+ spin_lock_irqsave(&cq->lock, flags);
+
+ /*
+ * Note that the head pointer might be writable by user processes.
+ * Take care to verify it is a sane value.
+ */
+ wc = cq->queue;
+ head = wc->head;
+ if (head >= (unsigned) cq->ibcq.cqe) {
+ head = cq->ibcq.cqe;
+ next = 0;
+ } else
+ next = head + 1;
+ if (unlikely(next == wc->tail)) {
+ spin_unlock_irqrestore(&cq->lock, flags);
+ if (cq->ibcq.event_handler) {
+ struct ib_event ev;
+
+ ev.device = cq->ibcq.device;
+ ev.element.cq = &cq->ibcq;
+ ev.event = IB_EVENT_CQ_ERR;
+ cq->ibcq.event_handler(&ev, cq->ibcq.cq_context);
+ }
+ return;
+ }
+ if (cq->ip) {
+ wc->uqueue[head].wr_id = entry->wr_id;
+ wc->uqueue[head].status = entry->status;
+ wc->uqueue[head].opcode = entry->opcode;
+ wc->uqueue[head].vendor_err = entry->vendor_err;
+ wc->uqueue[head].byte_len = entry->byte_len;
+ wc->uqueue[head].ex.imm_data =
+ (__u32 __force)entry->ex.imm_data;
+ wc->uqueue[head].qp_num = entry->qp->qp_num;
+ wc->uqueue[head].src_qp = entry->src_qp;
+ wc->uqueue[head].wc_flags = entry->wc_flags;
+ wc->uqueue[head].pkey_index = entry->pkey_index;
+ wc->uqueue[head].slid = entry->slid;
+ wc->uqueue[head].sl = entry->sl;
+ wc->uqueue[head].dlid_path_bits = entry->dlid_path_bits;
+ wc->uqueue[head].port_num = entry->port_num;
+ /* Make sure entry is written before the head index. */
+ smp_wmb();
+ } else
+ wc->kqueue[head] = *entry;
+ wc->head = next;
+
+ if (cq->notify == IB_CQ_NEXT_COMP ||
+ (cq->notify == IB_CQ_SOLICITED &&
+ (solicited || entry->status != IB_WC_SUCCESS))) {
+ struct kthread_worker *worker;
+ /*
+ * This will cause send_complete() to be called in
+ * another thread.
+ */
+ smp_read_barrier_depends(); /* see hfi1_cq_exit */
+ worker = cq->dd->worker;
+ if (likely(worker)) {
+ cq->notify = IB_CQ_NONE;
+ cq->triggered++;
+ queue_kthread_work(worker, &cq->comptask);
+ }
+ }
+
+ spin_unlock_irqrestore(&cq->lock, flags);
+}
+
+/**
+ * hfi1_poll_cq - poll for work completion entries
+ * @ibcq: the completion queue to poll
+ * @num_entries: the maximum number of entries to return
+ * @entry: pointer to array where work completions are placed
+ *
+ * Returns the number of completion entries polled.
+ *
+ * This may be called from interrupt context. Also called by ib_poll_cq()
+ * in the generic verbs code.
+ */
+int hfi1_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
+{
+ struct hfi1_cq *cq = to_icq(ibcq);
+ struct hfi1_cq_wc *wc;
+ unsigned long flags;
+ int npolled;
+ u32 tail;
+
+ /* The kernel can only poll a kernel completion queue */
+ if (cq->ip) {
+ npolled = -EINVAL;
+ goto bail;
+ }
+
+ spin_lock_irqsave(&cq->lock, flags);
+
+ wc = cq->queue;
+ tail = wc->tail;
+ if (tail > (u32) cq->ibcq.cqe)
+ tail = (u32) cq->ibcq.cqe;
+ for (npolled = 0; npolled < num_entries; ++npolled, ++entry) {
+ if (tail == wc->head)
+ break;
+ /* The kernel doesn't need a RMB since it has the lock. */
+ *entry = wc->kqueue[tail];
+ if (tail >= cq->ibcq.cqe)
+ tail = 0;
+ else
+ tail++;
+ }
+ wc->tail = tail;
+
+ spin_unlock_irqrestore(&cq->lock, flags);
+
+bail:
+ return npolled;
+}
+
+static void send_complete(struct kthread_work *work)
+{
+ struct hfi1_cq *cq = container_of(work, struct hfi1_cq, comptask);
+
+ /*
+ * The completion handler will most likely rearm the notification
+ * and poll for all pending entries. If a new completion entry
+ * is added while we are in this routine, queue_work()
+ * won't call us again until we return so we check triggered to
+ * see if we need to call the handler again.
+ */
+ for (;;) {
+ u8 triggered = cq->triggered;
+
+ /*
+ * IPoIB connected mode assumes the callback is from a
+ * soft IRQ. We simulate this by blocking "bottom halves".
+ * See the implementation for ipoib_cm_handle_tx_wc(),
+ * netif_tx_lock_bh() and netif_tx_lock().
+ */
+ local_bh_disable();
+ cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
+ local_bh_enable();
+
+ if (cq->triggered == triggered)
+ return;
+ }
+}
+
+/**
+ * hfi1_create_cq - create a completion queue
+ * @ibdev: the device this completion queue is attached to
+ * @attr: creation attributes
+ * @context: unused by the driver
+ * @udata: user data for libibverbs.so
+ *
+ * Returns a pointer to the completion queue or negative errno values
+ * for failure.
+ *
+ * Called by ib_create_cq() in the generic verbs code.
+ */
+struct ib_cq *hfi1_create_cq(
+ struct ib_device *ibdev,
+ const struct ib_cq_init_attr *attr,
+ struct ib_ucontext *context,
+ struct ib_udata *udata)
+{
+ struct hfi1_ibdev *dev = to_idev(ibdev);
+ struct hfi1_cq *cq;
+ struct hfi1_cq_wc *wc;
+ struct ib_cq *ret;
+ u32 sz;
+ unsigned int entries = attr->cqe;
+
+ if (attr->flags)
+ return ERR_PTR(-EINVAL);
+
+ if (entries < 1 || entries > hfi1_max_cqes)
+ return ERR_PTR(-EINVAL);
+
+ /* Allocate the completion queue structure. */
+ cq = kmalloc(sizeof(*cq), GFP_KERNEL);
+ if (!cq)
+ return ERR_PTR(-ENOMEM);
+
+ /*
+ * Allocate the completion queue entries and head/tail pointers.
+ * This is allocated separately so that it can be resized and
+ * also mapped into user space.
+ * We need to use vmalloc() in order to support mmap and large
+ * numbers of entries.
+ */
+ sz = sizeof(*wc);
+ if (udata && udata->outlen >= sizeof(__u64))
+ sz += sizeof(struct ib_uverbs_wc) * (entries + 1);
+ else
+ sz += sizeof(struct ib_wc) * (entries + 1);
+ wc = vmalloc_user(sz);
+ if (!wc) {
+ ret = ERR_PTR(-ENOMEM);
+ goto bail_cq;
+ }
+
+ /*
+ * Return the address of the WC as the offset to mmap.
+ * See hfi1_mmap() for details.
+ */
+ if (udata && udata->outlen >= sizeof(__u64)) {
+ int err;
+
+ cq->ip = hfi1_create_mmap_info(dev, sz, context, wc);
+ if (!cq->ip) {
+ ret = ERR_PTR(-ENOMEM);
+ goto bail_wc;
+ }
+
+ err = ib_copy_to_udata(udata, &cq->ip->offset,
+ sizeof(cq->ip->offset));
+ if (err) {
+ ret = ERR_PTR(err);
+ goto bail_ip;
+ }
+ } else
+ cq->ip = NULL;
+
+ spin_lock(&dev->n_cqs_lock);
+ if (dev->n_cqs_allocated == hfi1_max_cqs) {
+ spin_unlock(&dev->n_cqs_lock);
+ ret = ERR_PTR(-ENOMEM);
+ goto bail_ip;
+ }
+
+ dev->n_cqs_allocated++;
+ spin_unlock(&dev->n_cqs_lock);
+
+ if (cq->ip) {
+ spin_lock_irq(&dev->pending_lock);
+ list_add(&cq->ip->pending_mmaps, &dev->pending_mmaps);
+ spin_unlock_irq(&dev->pending_lock);
+ }
+
+ /*
+ * ib_create_cq() will initialize cq->ibcq except for cq->ibcq.cqe.
+ * The number of entries should be >= the number requested or return
+ * an error.
+ */
+ cq->dd = dd_from_dev(dev);
+ cq->ibcq.cqe = entries;
+ cq->notify = IB_CQ_NONE;
+ cq->triggered = 0;
+ spin_lock_init(&cq->lock);
+ init_kthread_work(&cq->comptask, send_complete);
+ wc->head = 0;
+ wc->tail = 0;
+ cq->queue = wc;
+
+ ret = &cq->ibcq;
+
+ goto done;
+
+bail_ip:
+ kfree(cq->ip);
+bail_wc:
+ vfree(wc);
+bail_cq:
+ kfree(cq);
+done:
+ return ret;
+}
+
+/**
+ * hfi1_destroy_cq - destroy a completion queue
+ * @ibcq: the completion queue to destroy.
+ *
+ * Returns 0 for success.
+ *
+ * Called by ib_destroy_cq() in the generic verbs code.
+ */
+int hfi1_destroy_cq(struct ib_cq *ibcq)
+{
+ struct hfi1_ibdev *dev = to_idev(ibcq->device);
+ struct hfi1_cq *cq = to_icq(ibcq);
+
+ flush_kthread_work(&cq->comptask);
+ spin_lock(&dev->n_cqs_lock);
+ dev->n_cqs_allocated--;
+ spin_unlock(&dev->n_cqs_lock);
+ if (cq->ip)
+ kref_put(&cq->ip->ref, hfi1_release_mmap_info);
+ else
+ vfree(cq->queue);
+ kfree(cq);
+
+ return 0;
+}
+
+/**
+ * hfi1_req_notify_cq - change the notification type for a completion queue
+ * @ibcq: the completion queue
+ * @notify_flags: the type of notification to request
+ *
+ * Returns 0 for success.
+ *
+ * This may be called from interrupt context. Also called by
+ * ib_req_notify_cq() in the generic verbs code.
+ */
+int hfi1_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
+{
+ struct hfi1_cq *cq = to_icq(ibcq);
+ unsigned long flags;
+ int ret = 0;
+
+ spin_lock_irqsave(&cq->lock, flags);
+ /*
+ * Don't change IB_CQ_NEXT_COMP to IB_CQ_SOLICITED but allow
+ * any other transitions (see C11-31 and C11-32 in ch. 11.4.2.2).
+ */
+ if (cq->notify != IB_CQ_NEXT_COMP)
+ cq->notify = notify_flags & IB_CQ_SOLICITED_MASK;
+
+ if ((notify_flags & IB_CQ_REPORT_MISSED_EVENTS) &&
+ cq->queue->head != cq->queue->tail)
+ ret = 1;
+
+ spin_unlock_irqrestore(&cq->lock, flags);
+
+ return ret;
+}
+
+/**
+ * hfi1_resize_cq - change the size of the CQ
+ * @ibcq: the completion queue
+ *
+ * Returns 0 for success.
+ */
+int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
+{
+ struct hfi1_cq *cq = to_icq(ibcq);
+ struct hfi1_cq_wc *old_wc;
+ struct hfi1_cq_wc *wc;
+ u32 head, tail, n;
+ int ret;
+ u32 sz;
+
+ if (cqe < 1 || cqe > hfi1_max_cqes) {
+ ret = -EINVAL;
+ goto bail;
+ }
+
+ /*
+ * Need to use vmalloc() if we want to support large #s of entries.
+ */
+ sz = sizeof(*wc);
+ if (udata && udata->outlen >= sizeof(__u64))
+ sz += sizeof(struct ib_uverbs_wc) * (cqe + 1);
+ else
+ sz += sizeof(struct ib_wc) * (cqe + 1);
+ wc = vmalloc_user(sz);
+ if (!wc) {
+ ret = -ENOMEM;
+ goto bail;
+ }
+
+ /* Check that we can write the offset to mmap. */
+ if (udata && udata->outlen >= sizeof(__u64)) {
+ __u64 offset = 0;
+
+ ret = ib_copy_to_udata(udata, &offset, sizeof(offset));
+ if (ret)
+ goto bail_free;
+ }
+
+ spin_lock_irq(&cq->lock);
+ /*
+ * Make sure head and tail are sane since they
+ * might be user writable.
+ */
+ old_wc = cq->queue;
+ head = old_wc->head;
+ if (head > (u32) cq->ibcq.cqe)
+ head = (u32) cq->ibcq.cqe;
+ tail = old_wc->tail;
+ if (tail > (u32) cq->ibcq.cqe)
+ tail = (u32) cq->ibcq.cqe;
+ if (head < tail)
+ n = cq->ibcq.cqe + 1 + head - tail;
+ else
+ n = head - tail;
+ if (unlikely((u32)cqe < n)) {
+ ret = -EINVAL;
+ goto bail_unlock;
+ }
+ for (n = 0; tail != head; n++) {
+ if (cq->ip)
+ wc->uqueue[n] = old_wc->uqueue[tail];
+ else
+ wc->kqueue[n] = old_wc->kqueue[tail];
+ if (tail == (u32) cq->ibcq.cqe)
+ tail = 0;
+ else
+ tail++;
+ }
+ cq->ibcq.cqe = cqe;
+ wc->head = n;
+ wc->tail = 0;
+ cq->queue = wc;
+ spin_unlock_irq(&cq->lock);
+
+ vfree(old_wc);
+
+ if (cq->ip) {
+ struct hfi1_ibdev *dev = to_idev(ibcq->device);
+ struct hfi1_mmap_info *ip = cq->ip;
+
+ hfi1_update_mmap_info(dev, ip, sz, wc);
+
+ /*
+ * Return the offset to mmap.
+ * See hfi1_mmap() for details.
+ */
+ if (udata && udata->outlen >= sizeof(__u64)) {
+ ret = ib_copy_to_udata(udata, &ip->offset,
+ sizeof(ip->offset));
+ if (ret)
+ goto bail;
+ }
+
+ spin_lock_irq(&dev->pending_lock);
+ if (list_empty(&ip->pending_mmaps))
+ list_add(&ip->pending_mmaps, &dev->pending_mmaps);
+ spin_unlock_irq(&dev->pending_lock);
+ }
+
+ ret = 0;
+ goto bail;
+
+bail_unlock:
+ spin_unlock_irq(&cq->lock);
+bail_free:
+ vfree(wc);
+bail:
+ return ret;
+}
+
+int hfi1_cq_init(struct hfi1_devdata *dd)
+{
+ int ret = 0;
+ int cpu;
+ struct task_struct *task;
+
+ if (dd->worker)
+ return 0;
+ dd->worker = kzalloc(sizeof(*dd->worker), GFP_KERNEL);
+ if (!dd->worker)
+ return -ENOMEM;
+ init_kthread_worker(dd->worker);
+ task = kthread_create_on_node(
+ kthread_worker_fn,
+ dd->worker,
+ dd->assigned_node_id,
+ "hfi1_cq%d", dd->unit);
+ if (IS_ERR(task))
+ goto task_fail;
+ cpu = cpumask_first(cpumask_of_node(dd->assigned_node_id));
+ kthread_bind(task, cpu);
+ wake_up_process(task);
+out:
+ return ret;
+task_fail:
+ ret = PTR_ERR(task);
+ kfree(dd->worker);
+ dd->worker = NULL;
+ goto out;
+}
+
+void hfi1_cq_exit(struct hfi1_devdata *dd)
+{
+ struct kthread_worker *worker;
+
+ worker = dd->worker;
+ if (!worker)
+ return;
+ /* blocks future queuing from send_complete() */
+ dd->worker = NULL;
+ smp_wmb(); /* See hfi1_cq_enter */
+ flush_kthread_worker(worker);
+ kthread_stop(worker->task);
+ kfree(worker);
+}
diff --git a/drivers/staging/rdma/hfi1/debugfs.c b/drivers/staging/rdma/hfi1/debugfs.c
new file mode 100644
index 000000000000..acd2269e9f14
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/debugfs.c
@@ -0,0 +1,899 @@
+#ifdef CONFIG_DEBUG_FS
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <linux/kernel.h>
+#include <linux/export.h>
+
+#include "hfi.h"
+#include "debugfs.h"
+#include "device.h"
+#include "qp.h"
+#include "sdma.h"
+
+static struct dentry *hfi1_dbg_root;
+
+#define private2dd(file) (file_inode(file)->i_private)
+#define private2ppd(file) (file_inode(file)->i_private)
+
+#define DEBUGFS_SEQ_FILE_OPS(name) \
+static const struct seq_operations _##name##_seq_ops = { \
+ .start = _##name##_seq_start, \
+ .next = _##name##_seq_next, \
+ .stop = _##name##_seq_stop, \
+ .show = _##name##_seq_show \
+}
+#define DEBUGFS_SEQ_FILE_OPEN(name) \
+static int _##name##_open(struct inode *inode, struct file *s) \
+{ \
+ struct seq_file *seq; \
+ int ret; \
+ ret = seq_open(s, &_##name##_seq_ops); \
+ if (ret) \
+ return ret; \
+ seq = s->private_data; \
+ seq->private = inode->i_private; \
+ return 0; \
+}
+
+#define DEBUGFS_FILE_OPS(name) \
+static const struct file_operations _##name##_file_ops = { \
+ .owner = THIS_MODULE, \
+ .open = _##name##_open, \
+ .read = seq_read, \
+ .llseek = seq_lseek, \
+ .release = seq_release \
+}
+
+#define DEBUGFS_FILE_CREATE(name, parent, data, ops, mode) \
+do { \
+ struct dentry *ent; \
+ ent = debugfs_create_file(name, mode, parent, \
+ data, ops); \
+ if (!ent) \
+ pr_warn("create of %s failed\n", name); \
+} while (0)
+
+
+#define DEBUGFS_SEQ_FILE_CREATE(name, parent, data) \
+ DEBUGFS_FILE_CREATE(#name, parent, data, &_##name##_file_ops, S_IRUGO)
+
+static void *_opcode_stats_seq_start(struct seq_file *s, loff_t *pos)
+__acquires(RCU)
+{
+ struct hfi1_opcode_stats_perctx *opstats;
+
+ rcu_read_lock();
+ if (*pos >= ARRAY_SIZE(opstats->stats))
+ return NULL;
+ return pos;
+}
+
+static void *_opcode_stats_seq_next(struct seq_file *s, void *v, loff_t *pos)
+{
+ struct hfi1_opcode_stats_perctx *opstats;
+
+ ++*pos;
+ if (*pos >= ARRAY_SIZE(opstats->stats))
+ return NULL;
+ return pos;
+}
+
+
+static void _opcode_stats_seq_stop(struct seq_file *s, void *v)
+__releases(RCU)
+{
+ rcu_read_unlock();
+}
+
+static int _opcode_stats_seq_show(struct seq_file *s, void *v)
+{
+ loff_t *spos = v;
+ loff_t i = *spos, j;
+ u64 n_packets = 0, n_bytes = 0;
+ struct hfi1_ibdev *ibd = (struct hfi1_ibdev *)s->private;
+ struct hfi1_devdata *dd = dd_from_dev(ibd);
+
+ for (j = 0; j < dd->first_user_ctxt; j++) {
+ if (!dd->rcd[j])
+ continue;
+ n_packets += dd->rcd[j]->opstats->stats[i].n_packets;
+ n_bytes += dd->rcd[j]->opstats->stats[i].n_bytes;
+ }
+ if (!n_packets && !n_bytes)
+ return SEQ_SKIP;
+ seq_printf(s, "%02llx %llu/%llu\n", i,
+ (unsigned long long) n_packets,
+ (unsigned long long) n_bytes);
+
+ return 0;
+}
+
+DEBUGFS_SEQ_FILE_OPS(opcode_stats);
+DEBUGFS_SEQ_FILE_OPEN(opcode_stats)
+DEBUGFS_FILE_OPS(opcode_stats);
+
+static void *_ctx_stats_seq_start(struct seq_file *s, loff_t *pos)
+{
+ struct hfi1_ibdev *ibd = (struct hfi1_ibdev *)s->private;
+ struct hfi1_devdata *dd = dd_from_dev(ibd);
+
+ if (!*pos)
+ return SEQ_START_TOKEN;
+ if (*pos >= dd->first_user_ctxt)
+ return NULL;
+ return pos;
+}
+
+static void *_ctx_stats_seq_next(struct seq_file *s, void *v, loff_t *pos)
+{
+ struct hfi1_ibdev *ibd = (struct hfi1_ibdev *)s->private;
+ struct hfi1_devdata *dd = dd_from_dev(ibd);
+
+ if (v == SEQ_START_TOKEN)
+ return pos;
+
+ ++*pos;
+ if (*pos >= dd->first_user_ctxt)
+ return NULL;
+ return pos;
+}
+
+static void _ctx_stats_seq_stop(struct seq_file *s, void *v)
+{
+ /* nothing allocated */
+}
+
+static int _ctx_stats_seq_show(struct seq_file *s, void *v)
+{
+ loff_t *spos;
+ loff_t i, j;
+ u64 n_packets = 0;
+ struct hfi1_ibdev *ibd = (struct hfi1_ibdev *)s->private;
+ struct hfi1_devdata *dd = dd_from_dev(ibd);
+
+ if (v == SEQ_START_TOKEN) {
+ seq_puts(s, "Ctx:npkts\n");
+ return 0;
+ }
+
+ spos = v;
+ i = *spos;
+
+ if (!dd->rcd[i])
+ return SEQ_SKIP;
+
+ for (j = 0; j < ARRAY_SIZE(dd->rcd[i]->opstats->stats); j++)
+ n_packets += dd->rcd[i]->opstats->stats[j].n_packets;
+
+ if (!n_packets)
+ return SEQ_SKIP;
+
+ seq_printf(s, " %llu:%llu\n", i, n_packets);
+ return 0;
+}
+
+DEBUGFS_SEQ_FILE_OPS(ctx_stats);
+DEBUGFS_SEQ_FILE_OPEN(ctx_stats)
+DEBUGFS_FILE_OPS(ctx_stats);
+
+static void *_qp_stats_seq_start(struct seq_file *s, loff_t *pos)
+__acquires(RCU)
+{
+ struct qp_iter *iter;
+ loff_t n = *pos;
+
+ rcu_read_lock();
+ iter = qp_iter_init(s->private);
+ if (!iter)
+ return NULL;
+
+ while (n--) {
+ if (qp_iter_next(iter)) {
+ kfree(iter);
+ return NULL;
+ }
+ }
+
+ return iter;
+}
+
+static void *_qp_stats_seq_next(struct seq_file *s, void *iter_ptr,
+ loff_t *pos)
+{
+ struct qp_iter *iter = iter_ptr;
+
+ (*pos)++;
+
+ if (qp_iter_next(iter)) {
+ kfree(iter);
+ return NULL;
+ }
+
+ return iter;
+}
+
+static void _qp_stats_seq_stop(struct seq_file *s, void *iter_ptr)
+__releases(RCU)
+{
+ rcu_read_unlock();
+}
+
+static int _qp_stats_seq_show(struct seq_file *s, void *iter_ptr)
+{
+ struct qp_iter *iter = iter_ptr;
+
+ if (!iter)
+ return 0;
+
+ qp_iter_print(s, iter);
+
+ return 0;
+}
+
+DEBUGFS_SEQ_FILE_OPS(qp_stats);
+DEBUGFS_SEQ_FILE_OPEN(qp_stats)
+DEBUGFS_FILE_OPS(qp_stats);
+
+static void *_sdes_seq_start(struct seq_file *s, loff_t *pos)
+__acquires(RCU)
+{
+ struct hfi1_ibdev *ibd;
+ struct hfi1_devdata *dd;
+
+ rcu_read_lock();
+ ibd = (struct hfi1_ibdev *)s->private;
+ dd = dd_from_dev(ibd);
+ if (!dd->per_sdma || *pos >= dd->num_sdma)
+ return NULL;
+ return pos;
+}
+
+static void *_sdes_seq_next(struct seq_file *s, void *v, loff_t *pos)
+{
+ struct hfi1_ibdev *ibd = (struct hfi1_ibdev *)s->private;
+ struct hfi1_devdata *dd = dd_from_dev(ibd);
+
+ ++*pos;
+ if (!dd->per_sdma || *pos >= dd->num_sdma)
+ return NULL;
+ return pos;
+}
+
+
+static void _sdes_seq_stop(struct seq_file *s, void *v)
+__releases(RCU)
+{
+ rcu_read_unlock();
+}
+
+static int _sdes_seq_show(struct seq_file *s, void *v)
+{
+ struct hfi1_ibdev *ibd = (struct hfi1_ibdev *)s->private;
+ struct hfi1_devdata *dd = dd_from_dev(ibd);
+ loff_t *spos = v;
+ loff_t i = *spos;
+
+ sdma_seqfile_dump_sde(s, &dd->per_sdma[i]);
+ return 0;
+}
+
+DEBUGFS_SEQ_FILE_OPS(sdes);
+DEBUGFS_SEQ_FILE_OPEN(sdes)
+DEBUGFS_FILE_OPS(sdes);
+
+/* read the per-device counters */
+static ssize_t dev_counters_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ u64 *counters;
+ size_t avail;
+ struct hfi1_devdata *dd;
+ ssize_t rval;
+
+ rcu_read_lock();
+ dd = private2dd(file);
+ avail = hfi1_read_cntrs(dd, *ppos, NULL, &counters);
+ rval = simple_read_from_buffer(buf, count, ppos, counters, avail);
+ rcu_read_unlock();
+ return rval;
+}
+
+/* read the per-device counters */
+static ssize_t dev_names_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ char *names;
+ size_t avail;
+ struct hfi1_devdata *dd;
+ ssize_t rval;
+
+ rcu_read_lock();
+ dd = private2dd(file);
+ avail = hfi1_read_cntrs(dd, *ppos, &names, NULL);
+ rval = simple_read_from_buffer(buf, count, ppos, names, avail);
+ rcu_read_unlock();
+ return rval;
+}
+
+struct counter_info {
+ char *name;
+ const struct file_operations ops;
+};
+
+/*
+ * Could use file_inode(file)->i_ino to figure out which file,
+ * instead of separate routine for each, but for now, this works...
+ */
+
+/* read the per-port names (same for each port) */
+static ssize_t portnames_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ char *names;
+ size_t avail;
+ struct hfi1_devdata *dd;
+ ssize_t rval;
+
+ rcu_read_lock();
+ dd = private2dd(file);
+ /* port number n/a here since names are constant */
+ avail = hfi1_read_portcntrs(dd, *ppos, 0, &names, NULL);
+ rval = simple_read_from_buffer(buf, count, ppos, names, avail);
+ rcu_read_unlock();
+ return rval;
+}
+
+/* read the per-port counters */
+static ssize_t portcntrs_debugfs_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ u64 *counters;
+ size_t avail;
+ struct hfi1_devdata *dd;
+ struct hfi1_pportdata *ppd;
+ ssize_t rval;
+
+ rcu_read_lock();
+ ppd = private2ppd(file);
+ dd = ppd->dd;
+ avail = hfi1_read_portcntrs(dd, *ppos, ppd->port - 1, NULL, &counters);
+ rval = simple_read_from_buffer(buf, count, ppos, counters, avail);
+ rcu_read_unlock();
+ return rval;
+}
+
+/*
+ * read the per-port QSFP data for ppd
+ */
+static ssize_t qsfp_debugfs_dump(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ struct hfi1_pportdata *ppd;
+ char *tmp;
+ int ret;
+
+ rcu_read_lock();
+ ppd = private2ppd(file);
+ tmp = kmalloc(PAGE_SIZE, GFP_KERNEL);
+ if (!tmp) {
+ rcu_read_unlock();
+ return -ENOMEM;
+ }
+
+ ret = qsfp_dump(ppd, tmp, PAGE_SIZE);
+ if (ret > 0)
+ ret = simple_read_from_buffer(buf, count, ppos, tmp, ret);
+ rcu_read_unlock();
+ kfree(tmp);
+ return ret;
+}
+
+/* Do an i2c write operation on the chain for the given HFI. */
+static ssize_t __i2c_debugfs_write(struct file *file, const char __user *buf,
+ size_t count, loff_t *ppos, u32 target)
+{
+ struct hfi1_pportdata *ppd;
+ char *buff;
+ int ret;
+ int i2c_addr;
+ int offset;
+ int total_written;
+
+ rcu_read_lock();
+ ppd = private2ppd(file);
+
+ buff = kmalloc(count, GFP_KERNEL);
+ if (!buff) {
+ ret = -ENOMEM;
+ goto _return;
+ }
+
+ ret = copy_from_user(buff, buf, count);
+ if (ret > 0) {
+ ret = -EFAULT;
+ goto _free;
+ }
+
+ i2c_addr = (*ppos >> 16) & 0xff;
+ offset = *ppos & 0xffff;
+
+ total_written = i2c_write(ppd, target, i2c_addr, offset, buff, count);
+ if (total_written < 0) {
+ ret = total_written;
+ goto _free;
+ }
+
+ *ppos += total_written;
+
+ ret = total_written;
+
+ _free:
+ kfree(buff);
+ _return:
+ rcu_read_unlock();
+ return ret;
+}
+
+/* Do an i2c write operation on chain for HFI 0. */
+static ssize_t i2c1_debugfs_write(struct file *file, const char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ return __i2c_debugfs_write(file, buf, count, ppos, 0);
+}
+
+/* Do an i2c write operation on chain for HFI 1. */
+static ssize_t i2c2_debugfs_write(struct file *file, const char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ return __i2c_debugfs_write(file, buf, count, ppos, 1);
+}
+
+/* Do an i2c read operation on the chain for the given HFI. */
+static ssize_t __i2c_debugfs_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos, u32 target)
+{
+ struct hfi1_pportdata *ppd;
+ char *buff;
+ int ret;
+ int i2c_addr;
+ int offset;
+ int total_read;
+
+ rcu_read_lock();
+ ppd = private2ppd(file);
+
+ buff = kmalloc(count, GFP_KERNEL);
+ if (!buff) {
+ ret = -ENOMEM;
+ goto _return;
+ }
+
+ i2c_addr = (*ppos >> 16) & 0xff;
+ offset = *ppos & 0xffff;
+
+ total_read = i2c_read(ppd, target, i2c_addr, offset, buff, count);
+ if (total_read < 0) {
+ ret = total_read;
+ goto _free;
+ }
+
+ *ppos += total_read;
+
+ ret = copy_to_user(buf, buff, total_read);
+ if (ret > 0) {
+ ret = -EFAULT;
+ goto _free;
+ }
+
+ ret = total_read;
+
+ _free:
+ kfree(buff);
+ _return:
+ rcu_read_unlock();
+ return ret;
+}
+
+/* Do an i2c read operation on chain for HFI 0. */
+static ssize_t i2c1_debugfs_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ return __i2c_debugfs_read(file, buf, count, ppos, 0);
+}
+
+/* Do an i2c read operation on chain for HFI 1. */
+static ssize_t i2c2_debugfs_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ return __i2c_debugfs_read(file, buf, count, ppos, 1);
+}
+
+/* Do a QSFP write operation on the i2c chain for the given HFI. */
+static ssize_t __qsfp_debugfs_write(struct file *file, const char __user *buf,
+ size_t count, loff_t *ppos, u32 target)
+{
+ struct hfi1_pportdata *ppd;
+ char *buff;
+ int ret;
+ int total_written;
+
+ rcu_read_lock();
+ if (*ppos + count > QSFP_PAGESIZE * 4) { /* base page + page00-page03 */
+ ret = -EINVAL;
+ goto _return;
+ }
+
+ ppd = private2ppd(file);
+
+ buff = kmalloc(count, GFP_KERNEL);
+ if (!buff) {
+ ret = -ENOMEM;
+ goto _return;
+ }
+
+ ret = copy_from_user(buff, buf, count);
+ if (ret > 0) {
+ ret = -EFAULT;
+ goto _free;
+ }
+
+ total_written = qsfp_write(ppd, target, *ppos, buff, count);
+ if (total_written < 0) {
+ ret = total_written;
+ goto _free;
+ }
+
+ *ppos += total_written;
+
+ ret = total_written;
+
+ _free:
+ kfree(buff);
+ _return:
+ rcu_read_unlock();
+ return ret;
+}
+
+/* Do a QSFP write operation on i2c chain for HFI 0. */
+static ssize_t qsfp1_debugfs_write(struct file *file, const char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ return __qsfp_debugfs_write(file, buf, count, ppos, 0);
+}
+
+/* Do a QSFP write operation on i2c chain for HFI 1. */
+static ssize_t qsfp2_debugfs_write(struct file *file, const char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ return __qsfp_debugfs_write(file, buf, count, ppos, 1);
+}
+
+/* Do a QSFP read operation on the i2c chain for the given HFI. */
+static ssize_t __qsfp_debugfs_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos, u32 target)
+{
+ struct hfi1_pportdata *ppd;
+ char *buff;
+ int ret;
+ int total_read;
+
+ rcu_read_lock();
+ if (*ppos + count > QSFP_PAGESIZE * 4) { /* base page + page00-page03 */
+ ret = -EINVAL;
+ goto _return;
+ }
+
+ ppd = private2ppd(file);
+
+ buff = kmalloc(count, GFP_KERNEL);
+ if (!buff) {
+ ret = -ENOMEM;
+ goto _return;
+ }
+
+ total_read = qsfp_read(ppd, target, *ppos, buff, count);
+ if (total_read < 0) {
+ ret = total_read;
+ goto _free;
+ }
+
+ *ppos += total_read;
+
+ ret = copy_to_user(buf, buff, total_read);
+ if (ret > 0) {
+ ret = -EFAULT;
+ goto _free;
+ }
+
+ ret = total_read;
+
+ _free:
+ kfree(buff);
+ _return:
+ rcu_read_unlock();
+ return ret;
+}
+
+/* Do a QSFP read operation on i2c chain for HFI 0. */
+static ssize_t qsfp1_debugfs_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ return __qsfp_debugfs_read(file, buf, count, ppos, 0);
+}
+
+/* Do a QSFP read operation on i2c chain for HFI 1. */
+static ssize_t qsfp2_debugfs_read(struct file *file, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ return __qsfp_debugfs_read(file, buf, count, ppos, 1);
+}
+
+#define DEBUGFS_OPS(nm, readroutine, writeroutine) \
+{ \
+ .name = nm, \
+ .ops = { \
+ .read = readroutine, \
+ .write = writeroutine, \
+ .llseek = generic_file_llseek, \
+ }, \
+}
+
+static const struct counter_info cntr_ops[] = {
+ DEBUGFS_OPS("counter_names", dev_names_read, NULL),
+ DEBUGFS_OPS("counters", dev_counters_read, NULL),
+ DEBUGFS_OPS("portcounter_names", portnames_read, NULL),
+};
+
+static const struct counter_info port_cntr_ops[] = {
+ DEBUGFS_OPS("port%dcounters", portcntrs_debugfs_read, NULL),
+ DEBUGFS_OPS("i2c1", i2c1_debugfs_read, i2c1_debugfs_write),
+ DEBUGFS_OPS("i2c2", i2c2_debugfs_read, i2c2_debugfs_write),
+ DEBUGFS_OPS("qsfp_dump%d", qsfp_debugfs_dump, NULL),
+ DEBUGFS_OPS("qsfp1", qsfp1_debugfs_read, qsfp1_debugfs_write),
+ DEBUGFS_OPS("qsfp2", qsfp2_debugfs_read, qsfp2_debugfs_write),
+};
+
+void hfi1_dbg_ibdev_init(struct hfi1_ibdev *ibd)
+{
+ char name[sizeof("port0counters") + 1];
+ char link[10];
+ struct hfi1_devdata *dd = dd_from_dev(ibd);
+ struct hfi1_pportdata *ppd;
+ int unit = dd->unit;
+ int i, j;
+
+ if (!hfi1_dbg_root)
+ return;
+ snprintf(name, sizeof(name), "%s_%d", class_name(), unit);
+ snprintf(link, sizeof(link), "%d", unit);
+ ibd->hfi1_ibdev_dbg = debugfs_create_dir(name, hfi1_dbg_root);
+ if (!ibd->hfi1_ibdev_dbg) {
+ pr_warn("create of %s failed\n", name);
+ return;
+ }
+ ibd->hfi1_ibdev_link =
+ debugfs_create_symlink(link, hfi1_dbg_root, name);
+ if (!ibd->hfi1_ibdev_link) {
+ pr_warn("create of %s symlink failed\n", name);
+ return;
+ }
+ DEBUGFS_SEQ_FILE_CREATE(opcode_stats, ibd->hfi1_ibdev_dbg, ibd);
+ DEBUGFS_SEQ_FILE_CREATE(ctx_stats, ibd->hfi1_ibdev_dbg, ibd);
+ DEBUGFS_SEQ_FILE_CREATE(qp_stats, ibd->hfi1_ibdev_dbg, ibd);
+ DEBUGFS_SEQ_FILE_CREATE(sdes, ibd->hfi1_ibdev_dbg, ibd);
+ /* dev counter files */
+ for (i = 0; i < ARRAY_SIZE(cntr_ops); i++)
+ DEBUGFS_FILE_CREATE(cntr_ops[i].name,
+ ibd->hfi1_ibdev_dbg,
+ dd,
+ &cntr_ops[i].ops, S_IRUGO);
+ /* per port files */
+ for (ppd = dd->pport, j = 0; j < dd->num_pports; j++, ppd++)
+ for (i = 0; i < ARRAY_SIZE(port_cntr_ops); i++) {
+ snprintf(name,
+ sizeof(name),
+ port_cntr_ops[i].name,
+ j + 1);
+ DEBUGFS_FILE_CREATE(name,
+ ibd->hfi1_ibdev_dbg,
+ ppd,
+ &port_cntr_ops[i].ops,
+ port_cntr_ops[i].ops.write == NULL ?
+ S_IRUGO : S_IRUGO|S_IWUSR);
+ }
+}
+
+void hfi1_dbg_ibdev_exit(struct hfi1_ibdev *ibd)
+{
+ if (!hfi1_dbg_root)
+ goto out;
+ debugfs_remove(ibd->hfi1_ibdev_link);
+ debugfs_remove_recursive(ibd->hfi1_ibdev_dbg);
+out:
+ ibd->hfi1_ibdev_dbg = NULL;
+ synchronize_rcu();
+}
+
+/*
+ * driver stats field names, one line per stat, single string. Used by
+ * programs like hfistats to print the stats in a way which works for
+ * different versions of drivers, without changing program source.
+ * if hfi1_ib_stats changes, this needs to change. Names need to be
+ * 12 chars or less (w/o newline), for proper display by hfistats utility.
+ */
+static const char * const hfi1_statnames[] = {
+ /* must be element 0*/
+ "KernIntr",
+ "ErrorIntr",
+ "Tx_Errs",
+ "Rcv_Errs",
+ "H/W_Errs",
+ "NoPIOBufs",
+ "CtxtsOpen",
+ "RcvLen_Errs",
+ "EgrBufFull",
+ "EgrHdrFull"
+};
+
+static void *_driver_stats_names_seq_start(struct seq_file *s, loff_t *pos)
+__acquires(RCU)
+{
+ rcu_read_lock();
+ if (*pos >= ARRAY_SIZE(hfi1_statnames))
+ return NULL;
+ return pos;
+}
+
+static void *_driver_stats_names_seq_next(
+ struct seq_file *s,
+ void *v,
+ loff_t *pos)
+{
+ ++*pos;
+ if (*pos >= ARRAY_SIZE(hfi1_statnames))
+ return NULL;
+ return pos;
+}
+
+static void _driver_stats_names_seq_stop(struct seq_file *s, void *v)
+__releases(RCU)
+{
+ rcu_read_unlock();
+}
+
+static int _driver_stats_names_seq_show(struct seq_file *s, void *v)
+{
+ loff_t *spos = v;
+
+ seq_printf(s, "%s\n", hfi1_statnames[*spos]);
+ return 0;
+}
+
+DEBUGFS_SEQ_FILE_OPS(driver_stats_names);
+DEBUGFS_SEQ_FILE_OPEN(driver_stats_names)
+DEBUGFS_FILE_OPS(driver_stats_names);
+
+static void *_driver_stats_seq_start(struct seq_file *s, loff_t *pos)
+__acquires(RCU)
+{
+ rcu_read_lock();
+ if (*pos >= ARRAY_SIZE(hfi1_statnames))
+ return NULL;
+ return pos;
+}
+
+static void *_driver_stats_seq_next(struct seq_file *s, void *v, loff_t *pos)
+{
+ ++*pos;
+ if (*pos >= ARRAY_SIZE(hfi1_statnames))
+ return NULL;
+ return pos;
+}
+
+static void _driver_stats_seq_stop(struct seq_file *s, void *v)
+__releases(RCU)
+{
+ rcu_read_unlock();
+}
+
+static u64 hfi1_sps_ints(void)
+{
+ unsigned long flags;
+ struct hfi1_devdata *dd;
+ u64 sps_ints = 0;
+
+ spin_lock_irqsave(&hfi1_devs_lock, flags);
+ list_for_each_entry(dd, &hfi1_dev_list, list) {
+ sps_ints += get_all_cpu_total(dd->int_counter);
+ }
+ spin_unlock_irqrestore(&hfi1_devs_lock, flags);
+ return sps_ints;
+}
+
+static int _driver_stats_seq_show(struct seq_file *s, void *v)
+{
+ loff_t *spos = v;
+ char *buffer;
+ u64 *stats = (u64 *)&hfi1_stats;
+ size_t sz = seq_get_buf(s, &buffer);
+
+ if (sz < sizeof(u64))
+ return SEQ_SKIP;
+ /* special case for interrupts */
+ if (*spos == 0)
+ *(u64 *)buffer = hfi1_sps_ints();
+ else
+ *(u64 *)buffer = stats[*spos];
+ seq_commit(s, sizeof(u64));
+ return 0;
+}
+
+DEBUGFS_SEQ_FILE_OPS(driver_stats);
+DEBUGFS_SEQ_FILE_OPEN(driver_stats)
+DEBUGFS_FILE_OPS(driver_stats);
+
+void hfi1_dbg_init(void)
+{
+ hfi1_dbg_root = debugfs_create_dir(DRIVER_NAME, NULL);
+ if (!hfi1_dbg_root)
+ pr_warn("init of debugfs failed\n");
+ DEBUGFS_SEQ_FILE_CREATE(driver_stats_names, hfi1_dbg_root, NULL);
+ DEBUGFS_SEQ_FILE_CREATE(driver_stats, hfi1_dbg_root, NULL);
+}
+
+void hfi1_dbg_exit(void)
+{
+ debugfs_remove_recursive(hfi1_dbg_root);
+ hfi1_dbg_root = NULL;
+}
+
+#endif
diff --git a/drivers/staging/rdma/hfi1/debugfs.h b/drivers/staging/rdma/hfi1/debugfs.h
new file mode 100644
index 000000000000..92d6fe146714
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/debugfs.h
@@ -0,0 +1,78 @@
+#ifndef _HFI1_DEBUGFS_H
+#define _HFI1_DEBUGFS_H
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+struct hfi1_ibdev;
+#ifdef CONFIG_DEBUG_FS
+void hfi1_dbg_ibdev_init(struct hfi1_ibdev *ibd);
+void hfi1_dbg_ibdev_exit(struct hfi1_ibdev *ibd);
+void hfi1_dbg_init(void);
+void hfi1_dbg_exit(void);
+#else
+static inline void hfi1_dbg_ibdev_init(struct hfi1_ibdev *ibd)
+{
+}
+
+void hfi1_dbg_ibdev_exit(struct hfi1_ibdev *ibd)
+{
+}
+
+void hfi1_dbg_init(void)
+{
+}
+
+void hfi1_dbg_exit(void)
+{
+}
+
+#endif
+
+#endif /* _HFI1_DEBUGFS_H */
diff --git a/drivers/staging/rdma/hfi1/device.c b/drivers/staging/rdma/hfi1/device.c
new file mode 100644
index 000000000000..07c87a87775f
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/device.c
@@ -0,0 +1,142 @@
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <linux/cdev.h>
+#include <linux/module.h>
+#include <linux/device.h>
+#include <linux/fs.h>
+
+#include "hfi.h"
+#include "device.h"
+
+static struct class *class;
+static dev_t hfi1_dev;
+
+int hfi1_cdev_init(int minor, const char *name,
+ const struct file_operations *fops,
+ struct cdev *cdev, struct device **devp)
+{
+ const dev_t dev = MKDEV(MAJOR(hfi1_dev), minor);
+ struct device *device = NULL;
+ int ret;
+
+ cdev_init(cdev, fops);
+ cdev->owner = THIS_MODULE;
+ kobject_set_name(&cdev->kobj, name);
+
+ ret = cdev_add(cdev, dev, 1);
+ if (ret < 0) {
+ pr_err("Could not add cdev for minor %d, %s (err %d)\n",
+ minor, name, -ret);
+ goto done;
+ }
+
+ device = device_create(class, NULL, dev, NULL, "%s", name);
+ if (!IS_ERR(device))
+ goto done;
+ ret = PTR_ERR(device);
+ device = NULL;
+ pr_err("Could not create device for minor %d, %s (err %d)\n",
+ minor, name, -ret);
+ cdev_del(cdev);
+done:
+ *devp = device;
+ return ret;
+}
+
+void hfi1_cdev_cleanup(struct cdev *cdev, struct device **devp)
+{
+ struct device *device = *devp;
+
+ if (device) {
+ device_unregister(device);
+ *devp = NULL;
+
+ cdev_del(cdev);
+ }
+}
+
+static const char *hfi1_class_name = "hfi1";
+
+const char *class_name(void)
+{
+ return hfi1_class_name;
+}
+
+int __init dev_init(void)
+{
+ int ret;
+
+ ret = alloc_chrdev_region(&hfi1_dev, 0, HFI1_NMINORS, DRIVER_NAME);
+ if (ret < 0) {
+ pr_err("Could not allocate chrdev region (err %d)\n", -ret);
+ goto done;
+ }
+
+ class = class_create(THIS_MODULE, class_name());
+ if (IS_ERR(class)) {
+ ret = PTR_ERR(class);
+ pr_err("Could not create device class (err %d)\n", -ret);
+ unregister_chrdev_region(hfi1_dev, HFI1_NMINORS);
+ }
+
+done:
+ return ret;
+}
+
+void dev_cleanup(void)
+{
+ if (class) {
+ class_destroy(class);
+ class = NULL;
+ }
+
+ unregister_chrdev_region(hfi1_dev, HFI1_NMINORS);
+}
diff --git a/drivers/staging/rdma/hfi1/device.h b/drivers/staging/rdma/hfi1/device.h
new file mode 100644
index 000000000000..98caecd3d807
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/device.h
@@ -0,0 +1,61 @@
+#ifndef _HFI1_DEVICE_H
+#define _HFI1_DEVICE_H
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+int hfi1_cdev_init(int minor, const char *name,
+ const struct file_operations *fops,
+ struct cdev *cdev, struct device **devp);
+void hfi1_cdev_cleanup(struct cdev *cdev, struct device **devp);
+const char *class_name(void);
+int __init dev_init(void);
+void dev_cleanup(void);
+
+#endif /* _HFI1_DEVICE_H */
diff --git a/drivers/staging/rdma/hfi1/diag.c b/drivers/staging/rdma/hfi1/diag.c
new file mode 100644
index 000000000000..6777d6b659cf
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/diag.c
@@ -0,0 +1,1873 @@
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+/*
+ * This file contains support for diagnostic functions. It is accessed by
+ * opening the hfi1_diag device, normally minor number 129. Diagnostic use
+ * of the chip may render the chip or board unusable until the driver
+ * is unloaded, or in some cases, until the system is rebooted.
+ *
+ * Accesses to the chip through this interface are not similar to going
+ * through the /sys/bus/pci resource mmap interface.
+ */
+
+#include <linux/io.h>
+#include <linux/pci.h>
+#include <linux/poll.h>
+#include <linux/vmalloc.h>
+#include <linux/export.h>
+#include <linux/fs.h>
+#include <linux/uaccess.h>
+#include <linux/module.h>
+#include <rdma/ib_smi.h>
+#include "hfi.h"
+#include "device.h"
+#include "common.h"
+#include "trace.h"
+
+#undef pr_fmt
+#define pr_fmt(fmt) DRIVER_NAME ": " fmt
+#define snoop_dbg(fmt, ...) \
+ hfi1_cdbg(SNOOP, fmt, ##__VA_ARGS__)
+
+/* Snoop option mask */
+#define SNOOP_DROP_SEND (1 << 0)
+#define SNOOP_USE_METADATA (1 << 1)
+
+static u8 snoop_flags;
+
+/*
+ * Extract packet length from LRH header.
+ * Why & 0x7FF? Because len is only 11 bits in case it wasn't 0'd we throw the
+ * bogus bits away. This is in Dwords so multiply by 4 to get size in bytes
+ */
+#define HFI1_GET_PKT_LEN(x) (((be16_to_cpu((x)->lrh[2]) & 0x7FF)) << 2)
+
+enum hfi1_filter_status {
+ HFI1_FILTER_HIT,
+ HFI1_FILTER_ERR,
+ HFI1_FILTER_MISS
+};
+
+/* snoop processing functions */
+rhf_rcv_function_ptr snoop_rhf_rcv_functions[8] = {
+ [RHF_RCV_TYPE_EXPECTED] = snoop_recv_handler,
+ [RHF_RCV_TYPE_EAGER] = snoop_recv_handler,
+ [RHF_RCV_TYPE_IB] = snoop_recv_handler,
+ [RHF_RCV_TYPE_ERROR] = snoop_recv_handler,
+ [RHF_RCV_TYPE_BYPASS] = snoop_recv_handler,
+ [RHF_RCV_TYPE_INVALID5] = process_receive_invalid,
+ [RHF_RCV_TYPE_INVALID6] = process_receive_invalid,
+ [RHF_RCV_TYPE_INVALID7] = process_receive_invalid
+};
+
+/* Snoop packet structure */
+struct snoop_packet {
+ struct list_head list;
+ u32 total_len;
+ u8 data[];
+};
+
+/* Do not make these an enum or it will blow up the capture_md */
+#define PKT_DIR_EGRESS 0x0
+#define PKT_DIR_INGRESS 0x1
+
+/* Packet capture metadata returned to the user with the packet. */
+struct capture_md {
+ u8 port;
+ u8 dir;
+ u8 reserved[6];
+ union {
+ u64 pbc;
+ u64 rhf;
+ } u;
+};
+
+static atomic_t diagpkt_count = ATOMIC_INIT(0);
+static struct cdev diagpkt_cdev;
+static struct device *diagpkt_device;
+
+static ssize_t diagpkt_write(struct file *fp, const char __user *data,
+ size_t count, loff_t *off);
+
+static const struct file_operations diagpkt_file_ops = {
+ .owner = THIS_MODULE,
+ .write = diagpkt_write,
+ .llseek = noop_llseek,
+};
+
+/*
+ * This is used for communication with user space for snoop extended IOCTLs
+ */
+struct hfi1_link_info {
+ __be64 node_guid;
+ u8 port_mode;
+ u8 port_state;
+ u16 link_speed_active;
+ u16 link_width_active;
+ u16 vl15_init;
+ u8 port_number;
+ /*
+ * Add padding to make this a full IB SMP payload. Note: changing the
+ * size of this structure will make the IOCTLs created with _IOWR
+ * change.
+ * Be sure to run tests on all IOCTLs when making changes to this
+ * structure.
+ */
+ u8 res[47];
+};
+
+/*
+ * This starts our ioctl sequence numbers *way* off from the ones
+ * defined in ib_core.
+ */
+#define SNOOP_CAPTURE_VERSION 0x1
+
+#define IB_IOCTL_MAGIC 0x1b /* See Documentation/ioctl-number.txt */
+#define HFI1_SNOOP_IOC_MAGIC IB_IOCTL_MAGIC
+#define HFI1_SNOOP_IOC_BASE_SEQ 0x80
+
+#define HFI1_SNOOP_IOCGETLINKSTATE \
+ _IO(HFI1_SNOOP_IOC_MAGIC, HFI1_SNOOP_IOC_BASE_SEQ)
+#define HFI1_SNOOP_IOCSETLINKSTATE \
+ _IO(HFI1_SNOOP_IOC_MAGIC, HFI1_SNOOP_IOC_BASE_SEQ+1)
+#define HFI1_SNOOP_IOCCLEARQUEUE \
+ _IO(HFI1_SNOOP_IOC_MAGIC, HFI1_SNOOP_IOC_BASE_SEQ+2)
+#define HFI1_SNOOP_IOCCLEARFILTER \
+ _IO(HFI1_SNOOP_IOC_MAGIC, HFI1_SNOOP_IOC_BASE_SEQ+3)
+#define HFI1_SNOOP_IOCSETFILTER \
+ _IO(HFI1_SNOOP_IOC_MAGIC, HFI1_SNOOP_IOC_BASE_SEQ+4)
+#define HFI1_SNOOP_IOCGETVERSION \
+ _IO(HFI1_SNOOP_IOC_MAGIC, HFI1_SNOOP_IOC_BASE_SEQ+5)
+#define HFI1_SNOOP_IOCSET_OPTS \
+ _IO(HFI1_SNOOP_IOC_MAGIC, HFI1_SNOOP_IOC_BASE_SEQ+6)
+
+/*
+ * These offsets +6/+7 could change, but these are already known and used
+ * IOCTL numbers so don't change them without a good reason.
+ */
+#define HFI1_SNOOP_IOCGETLINKSTATE_EXTRA \
+ _IOWR(HFI1_SNOOP_IOC_MAGIC, HFI1_SNOOP_IOC_BASE_SEQ+6, \
+ struct hfi1_link_info)
+#define HFI1_SNOOP_IOCSETLINKSTATE_EXTRA \
+ _IOWR(HFI1_SNOOP_IOC_MAGIC, HFI1_SNOOP_IOC_BASE_SEQ+7, \
+ struct hfi1_link_info)
+
+static int hfi1_snoop_open(struct inode *in, struct file *fp);
+static ssize_t hfi1_snoop_read(struct file *fp, char __user *data,
+ size_t pkt_len, loff_t *off);
+static ssize_t hfi1_snoop_write(struct file *fp, const char __user *data,
+ size_t count, loff_t *off);
+static long hfi1_ioctl(struct file *fp, unsigned int cmd, unsigned long arg);
+static unsigned int hfi1_snoop_poll(struct file *fp,
+ struct poll_table_struct *wait);
+static int hfi1_snoop_release(struct inode *in, struct file *fp);
+
+struct hfi1_packet_filter_command {
+ int opcode;
+ int length;
+ void *value_ptr;
+};
+
+/* Can't re-use PKT_DIR_*GRESS here because 0 means no packets for this */
+#define HFI1_SNOOP_INGRESS 0x1
+#define HFI1_SNOOP_EGRESS 0x2
+
+enum hfi1_packet_filter_opcodes {
+ FILTER_BY_LID,
+ FILTER_BY_DLID,
+ FILTER_BY_MAD_MGMT_CLASS,
+ FILTER_BY_QP_NUMBER,
+ FILTER_BY_PKT_TYPE,
+ FILTER_BY_SERVICE_LEVEL,
+ FILTER_BY_PKEY,
+ FILTER_BY_DIRECTION,
+};
+
+static const struct file_operations snoop_file_ops = {
+ .owner = THIS_MODULE,
+ .open = hfi1_snoop_open,
+ .read = hfi1_snoop_read,
+ .unlocked_ioctl = hfi1_ioctl,
+ .poll = hfi1_snoop_poll,
+ .write = hfi1_snoop_write,
+ .release = hfi1_snoop_release
+};
+
+struct hfi1_filter_array {
+ int (*filter)(void *, void *, void *);
+};
+
+static int hfi1_filter_lid(void *ibhdr, void *packet_data, void *value);
+static int hfi1_filter_dlid(void *ibhdr, void *packet_data, void *value);
+static int hfi1_filter_mad_mgmt_class(void *ibhdr, void *packet_data,
+ void *value);
+static int hfi1_filter_qp_number(void *ibhdr, void *packet_data, void *value);
+static int hfi1_filter_ibpacket_type(void *ibhdr, void *packet_data,
+ void *value);
+static int hfi1_filter_ib_service_level(void *ibhdr, void *packet_data,
+ void *value);
+static int hfi1_filter_ib_pkey(void *ibhdr, void *packet_data, void *value);
+static int hfi1_filter_direction(void *ibhdr, void *packet_data, void *value);
+
+static struct hfi1_filter_array hfi1_filters[] = {
+ { hfi1_filter_lid },
+ { hfi1_filter_dlid },
+ { hfi1_filter_mad_mgmt_class },
+ { hfi1_filter_qp_number },
+ { hfi1_filter_ibpacket_type },
+ { hfi1_filter_ib_service_level },
+ { hfi1_filter_ib_pkey },
+ { hfi1_filter_direction },
+};
+
+#define HFI1_MAX_FILTERS ARRAY_SIZE(hfi1_filters)
+#define HFI1_DIAG_MINOR_BASE 129
+
+static int hfi1_snoop_add(struct hfi1_devdata *dd, const char *name);
+
+int hfi1_diag_add(struct hfi1_devdata *dd)
+{
+ char name[16];
+ int ret = 0;
+
+ snprintf(name, sizeof(name), "%s_diagpkt%d", class_name(),
+ dd->unit);
+ /*
+ * Do this for each device as opposed to the normal diagpkt
+ * interface which is one per host
+ */
+ ret = hfi1_snoop_add(dd, name);
+ if (ret)
+ dd_dev_err(dd, "Unable to init snoop/capture device");
+
+ snprintf(name, sizeof(name), "%s_diagpkt", class_name());
+ if (atomic_inc_return(&diagpkt_count) == 1) {
+ ret = hfi1_cdev_init(HFI1_DIAGPKT_MINOR, name,
+ &diagpkt_file_ops, &diagpkt_cdev,
+ &diagpkt_device);
+ }
+
+ return ret;
+}
+
+/* this must be called w/ dd->snoop_in_lock held */
+static void drain_snoop_list(struct list_head *queue)
+{
+ struct list_head *pos, *q;
+ struct snoop_packet *packet;
+
+ list_for_each_safe(pos, q, queue) {
+ packet = list_entry(pos, struct snoop_packet, list);
+ list_del(pos);
+ kfree(packet);
+ }
+}
+
+static void hfi1_snoop_remove(struct hfi1_devdata *dd)
+{
+ unsigned long flags = 0;
+
+ spin_lock_irqsave(&dd->hfi1_snoop.snoop_lock, flags);
+ drain_snoop_list(&dd->hfi1_snoop.queue);
+ hfi1_cdev_cleanup(&dd->hfi1_snoop.cdev, &dd->hfi1_snoop.class_dev);
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+}
+
+void hfi1_diag_remove(struct hfi1_devdata *dd)
+{
+
+ hfi1_snoop_remove(dd);
+ if (atomic_dec_and_test(&diagpkt_count))
+ hfi1_cdev_cleanup(&diagpkt_cdev, &diagpkt_device);
+ hfi1_cdev_cleanup(&dd->diag_cdev, &dd->diag_device);
+}
+
+
+/*
+ * Allocated structure shared between the credit return mechanism and
+ * diagpkt_send().
+ */
+struct diagpkt_wait {
+ struct completion credits_returned;
+ int code;
+ atomic_t count;
+};
+
+/*
+ * When each side is finished with the structure, they call this.
+ * The last user frees the structure.
+ */
+static void put_diagpkt_wait(struct diagpkt_wait *wait)
+{
+ if (atomic_dec_and_test(&wait->count))
+ kfree(wait);
+}
+
+/*
+ * Callback from the credit return code. Set the complete, which
+ * will let diapkt_send() continue.
+ */
+static void diagpkt_complete(void *arg, int code)
+{
+ struct diagpkt_wait *wait = (struct diagpkt_wait *)arg;
+
+ wait->code = code;
+ complete(&wait->credits_returned);
+ put_diagpkt_wait(wait); /* finished with the structure */
+}
+
+/**
+ * diagpkt_send - send a packet
+ * @dp: diag packet descriptor
+ */
+static ssize_t diagpkt_send(struct diag_pkt *dp)
+{
+ struct hfi1_devdata *dd;
+ struct send_context *sc;
+ struct pio_buf *pbuf;
+ u32 *tmpbuf = NULL;
+ ssize_t ret = 0;
+ u32 pkt_len, total_len;
+ pio_release_cb credit_cb = NULL;
+ void *credit_arg = NULL;
+ struct diagpkt_wait *wait = NULL;
+
+ dd = hfi1_lookup(dp->unit);
+ if (!dd || !(dd->flags & HFI1_PRESENT) || !dd->kregbase) {
+ ret = -ENODEV;
+ goto bail;
+ }
+ if (!(dd->flags & HFI1_INITTED)) {
+ /* no hardware, freeze, etc. */
+ ret = -ENODEV;
+ goto bail;
+ }
+
+ if (dp->version != _DIAG_PKT_VERS) {
+ dd_dev_err(dd, "Invalid version %u for diagpkt_write\n",
+ dp->version);
+ ret = -EINVAL;
+ goto bail;
+ }
+
+ /* send count must be an exact number of dwords */
+ if (dp->len & 3) {
+ ret = -EINVAL;
+ goto bail;
+ }
+
+ /* there is only port 1 */
+ if (dp->port != 1) {
+ ret = -EINVAL;
+ goto bail;
+ }
+
+ /* need a valid context */
+ if (dp->sw_index >= dd->num_send_contexts) {
+ ret = -EINVAL;
+ goto bail;
+ }
+ /* can only use kernel contexts */
+ if (dd->send_contexts[dp->sw_index].type != SC_KERNEL) {
+ ret = -EINVAL;
+ goto bail;
+ }
+ /* must be allocated */
+ sc = dd->send_contexts[dp->sw_index].sc;
+ if (!sc) {
+ ret = -EINVAL;
+ goto bail;
+ }
+ /* must be enabled */
+ if (!(sc->flags & SCF_ENABLED)) {
+ ret = -EINVAL;
+ goto bail;
+ }
+
+ /* allocate a buffer and copy the data in */
+ tmpbuf = vmalloc(dp->len);
+ if (!tmpbuf) {
+ ret = -ENOMEM;
+ goto bail;
+ }
+
+ if (copy_from_user(tmpbuf,
+ (const void __user *) (unsigned long) dp->data,
+ dp->len)) {
+ ret = -EFAULT;
+ goto bail;
+ }
+
+ /*
+ * pkt_len is how much data we have to write, includes header and data.
+ * total_len is length of the packet in Dwords plus the PBC should not
+ * include the CRC.
+ */
+ pkt_len = dp->len >> 2;
+ total_len = pkt_len + 2; /* PBC + packet */
+
+ /* if 0, fill in a default */
+ if (dp->pbc == 0) {
+ struct hfi1_pportdata *ppd = dd->pport;
+
+ hfi1_cdbg(PKT, "Generating PBC");
+ dp->pbc = create_pbc(ppd, 0, 0, 0, total_len);
+ } else {
+ hfi1_cdbg(PKT, "Using passed in PBC");
+ }
+
+ hfi1_cdbg(PKT, "Egress PBC content is 0x%llx", dp->pbc);
+
+ /*
+ * The caller wants to wait until the packet is sent and to
+ * check for errors. The best we can do is wait until
+ * the buffer credits are returned and check if any packet
+ * error has occurred. If there are any late errors, this
+ * could miss it. If there are other senders who generate
+ * an error, this may find it. However, in general, it
+ * should catch most.
+ */
+ if (dp->flags & F_DIAGPKT_WAIT) {
+ /* always force a credit return */
+ dp->pbc |= PBC_CREDIT_RETURN;
+ /* turn on credit return interrupts */
+ sc_add_credit_return_intr(sc);
+ wait = kmalloc(sizeof(*wait), GFP_KERNEL);
+ if (!wait) {
+ ret = -ENOMEM;
+ goto bail;
+ }
+ init_completion(&wait->credits_returned);
+ atomic_set(&wait->count, 2);
+ wait->code = PRC_OK;
+
+ credit_cb = diagpkt_complete;
+ credit_arg = wait;
+ }
+
+ pbuf = sc_buffer_alloc(sc, total_len, credit_cb, credit_arg);
+ if (!pbuf) {
+ /*
+ * No send buffer means no credit callback. Undo
+ * the wait set-up that was done above. We free wait
+ * because the callback will never be called.
+ */
+ if (dp->flags & F_DIAGPKT_WAIT) {
+ sc_del_credit_return_intr(sc);
+ kfree(wait);
+ wait = NULL;
+ }
+ ret = -ENOSPC;
+ goto bail;
+ }
+
+ pio_copy(dd, pbuf, dp->pbc, tmpbuf, pkt_len);
+ /* no flush needed as the HW knows the packet size */
+
+ ret = sizeof(*dp);
+
+ if (dp->flags & F_DIAGPKT_WAIT) {
+ /* wait for credit return */
+ ret = wait_for_completion_interruptible(
+ &wait->credits_returned);
+ /*
+ * If the wait returns an error, the wait was interrupted,
+ * e.g. with a ^C in the user program. The callback is
+ * still pending. This is OK as the wait structure is
+ * kmalloc'ed and the structure will free itself when
+ * all users are done with it.
+ *
+ * A context disable occurs on a send context restart, so
+ * include that in the list of errors below to check for.
+ * NOTE: PRC_FILL_ERR is at best informational and cannot
+ * be depended on.
+ */
+ if (!ret && (((wait->code & PRC_STATUS_ERR)
+ || (wait->code & PRC_FILL_ERR)
+ || (wait->code & PRC_SC_DISABLE))))
+ ret = -EIO;
+
+ put_diagpkt_wait(wait); /* finished with the structure */
+ sc_del_credit_return_intr(sc);
+ }
+
+bail:
+ vfree(tmpbuf);
+ return ret;
+}
+
+static ssize_t diagpkt_write(struct file *fp, const char __user *data,
+ size_t count, loff_t *off)
+{
+ struct hfi1_devdata *dd;
+ struct send_context *sc;
+ u8 vl;
+
+ struct diag_pkt dp;
+
+ if (count != sizeof(dp))
+ return -EINVAL;
+
+ if (copy_from_user(&dp, data, sizeof(dp)))
+ return -EFAULT;
+
+ /*
+ * The Send Context is derived from the PbcVL value
+ * if PBC is populated
+ */
+ if (dp.pbc) {
+ dd = hfi1_lookup(dp.unit);
+ if (dd == NULL)
+ return -ENODEV;
+ vl = (dp.pbc >> PBC_VL_SHIFT) & PBC_VL_MASK;
+ sc = dd->vld[vl].sc;
+ if (sc) {
+ dp.sw_index = sc->sw_index;
+ hfi1_cdbg(
+ PKT,
+ "Packet sent over VL %d via Send Context %u(%u)",
+ vl, sc->sw_index, sc->hw_context);
+ }
+ }
+
+ return diagpkt_send(&dp);
+}
+
+static int hfi1_snoop_add(struct hfi1_devdata *dd, const char *name)
+{
+ int ret = 0;
+
+ dd->hfi1_snoop.mode_flag = 0;
+ spin_lock_init(&dd->hfi1_snoop.snoop_lock);
+ INIT_LIST_HEAD(&dd->hfi1_snoop.queue);
+ init_waitqueue_head(&dd->hfi1_snoop.waitq);
+
+ ret = hfi1_cdev_init(HFI1_SNOOP_CAPTURE_BASE + dd->unit, name,
+ &snoop_file_ops,
+ &dd->hfi1_snoop.cdev, &dd->hfi1_snoop.class_dev);
+
+ if (ret) {
+ dd_dev_err(dd, "Couldn't create %s device: %d", name, ret);
+ hfi1_cdev_cleanup(&dd->hfi1_snoop.cdev,
+ &dd->hfi1_snoop.class_dev);
+ }
+
+ return ret;
+}
+
+static struct hfi1_devdata *hfi1_dd_from_sc_inode(struct inode *in)
+{
+ int unit = iminor(in) - HFI1_SNOOP_CAPTURE_BASE;
+ struct hfi1_devdata *dd = NULL;
+
+ dd = hfi1_lookup(unit);
+ return dd;
+
+}
+
+/* clear or restore send context integrity checks */
+static void adjust_integrity_checks(struct hfi1_devdata *dd)
+{
+ struct send_context *sc;
+ unsigned long sc_flags;
+ int i;
+
+ spin_lock_irqsave(&dd->sc_lock, sc_flags);
+ for (i = 0; i < dd->num_send_contexts; i++) {
+ int enable;
+
+ sc = dd->send_contexts[i].sc;
+
+ if (!sc)
+ continue; /* not allocated */
+
+ enable = likely(!HFI1_CAP_IS_KSET(NO_INTEGRITY)) &&
+ dd->hfi1_snoop.mode_flag != HFI1_PORT_SNOOP_MODE;
+
+ set_pio_integrity(sc);
+
+ if (enable) /* take HFI_CAP_* flags into account */
+ hfi1_init_ctxt(sc);
+ }
+ spin_unlock_irqrestore(&dd->sc_lock, sc_flags);
+}
+
+static int hfi1_snoop_open(struct inode *in, struct file *fp)
+{
+ int ret;
+ int mode_flag = 0;
+ unsigned long flags = 0;
+ struct hfi1_devdata *dd;
+ struct list_head *queue;
+
+ mutex_lock(&hfi1_mutex);
+
+ dd = hfi1_dd_from_sc_inode(in);
+ if (dd == NULL) {
+ ret = -ENODEV;
+ goto bail;
+ }
+
+ /*
+ * File mode determines snoop or capture. Some existing user
+ * applications expect the capture device to be able to be opened RDWR
+ * because they expect a dedicated capture device. For this reason we
+ * support a module param to force capture mode even if the file open
+ * mode matches snoop.
+ */
+ if ((fp->f_flags & O_ACCMODE) == O_RDONLY) {
+ snoop_dbg("Capture Enabled");
+ mode_flag = HFI1_PORT_CAPTURE_MODE;
+ } else if ((fp->f_flags & O_ACCMODE) == O_RDWR) {
+ snoop_dbg("Snoop Enabled");
+ mode_flag = HFI1_PORT_SNOOP_MODE;
+ } else {
+ snoop_dbg("Invalid");
+ ret = -EINVAL;
+ goto bail;
+ }
+ queue = &dd->hfi1_snoop.queue;
+
+ /*
+ * We are not supporting snoop and capture at the same time.
+ */
+ spin_lock_irqsave(&dd->hfi1_snoop.snoop_lock, flags);
+ if (dd->hfi1_snoop.mode_flag) {
+ ret = -EBUSY;
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+ goto bail;
+ }
+
+ dd->hfi1_snoop.mode_flag = mode_flag;
+ drain_snoop_list(queue);
+
+ dd->hfi1_snoop.filter_callback = NULL;
+ dd->hfi1_snoop.filter_value = NULL;
+
+ /*
+ * Send side packet integrity checks are not helpful when snooping so
+ * disable and re-enable when we stop snooping.
+ */
+ if (mode_flag == HFI1_PORT_SNOOP_MODE) {
+ /* clear after snoop mode is on */
+ adjust_integrity_checks(dd); /* clear */
+
+ /*
+ * We also do not want to be doing the DLID LMC check for
+ * ingressed packets.
+ */
+ dd->hfi1_snoop.dcc_cfg = read_csr(dd, DCC_CFG_PORT_CONFIG1);
+ write_csr(dd, DCC_CFG_PORT_CONFIG1,
+ (dd->hfi1_snoop.dcc_cfg >> 32) << 32);
+ }
+
+ /*
+ * As soon as we set these function pointers the recv and send handlers
+ * are active. This is a race condition so we must make sure to drain
+ * the queue and init filter values above. Technically we should add
+ * locking here but all that will happen is on recv a packet will get
+ * allocated and get stuck on the snoop_lock before getting added to the
+ * queue. Same goes for send.
+ */
+ dd->rhf_rcv_function_map = snoop_rhf_rcv_functions;
+ dd->process_pio_send = snoop_send_pio_handler;
+ dd->process_dma_send = snoop_send_pio_handler;
+ dd->pio_inline_send = snoop_inline_pio_send;
+
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+ ret = 0;
+
+bail:
+ mutex_unlock(&hfi1_mutex);
+
+ return ret;
+}
+
+static int hfi1_snoop_release(struct inode *in, struct file *fp)
+{
+ unsigned long flags = 0;
+ struct hfi1_devdata *dd;
+ int mode_flag;
+
+ dd = hfi1_dd_from_sc_inode(in);
+ if (dd == NULL)
+ return -ENODEV;
+
+ spin_lock_irqsave(&dd->hfi1_snoop.snoop_lock, flags);
+
+ /* clear the snoop mode before re-adjusting send context CSRs */
+ mode_flag = dd->hfi1_snoop.mode_flag;
+ dd->hfi1_snoop.mode_flag = 0;
+
+ /*
+ * Drain the queue and clear the filters we are done with it. Don't
+ * forget to restore the packet integrity checks
+ */
+ drain_snoop_list(&dd->hfi1_snoop.queue);
+ if (mode_flag == HFI1_PORT_SNOOP_MODE) {
+ /* restore after snoop mode is clear */
+ adjust_integrity_checks(dd); /* restore */
+
+ /*
+ * Also should probably reset the DCC_CONFIG1 register for DLID
+ * checking on incoming packets again. Use the value saved when
+ * opening the snoop device.
+ */
+ write_csr(dd, DCC_CFG_PORT_CONFIG1, dd->hfi1_snoop.dcc_cfg);
+ }
+
+ dd->hfi1_snoop.filter_callback = NULL;
+ kfree(dd->hfi1_snoop.filter_value);
+ dd->hfi1_snoop.filter_value = NULL;
+
+ /*
+ * User is done snooping and capturing, return control to the normal
+ * handler. Re-enable SDMA handling.
+ */
+ dd->rhf_rcv_function_map = dd->normal_rhf_rcv_functions;
+ dd->process_pio_send = hfi1_verbs_send_pio;
+ dd->process_dma_send = hfi1_verbs_send_dma;
+ dd->pio_inline_send = pio_copy;
+
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+
+ snoop_dbg("snoop/capture device released");
+
+ return 0;
+}
+
+static unsigned int hfi1_snoop_poll(struct file *fp,
+ struct poll_table_struct *wait)
+{
+ int ret = 0;
+ unsigned long flags = 0;
+
+ struct hfi1_devdata *dd;
+
+ dd = hfi1_dd_from_sc_inode(fp->f_inode);
+ if (dd == NULL)
+ return -ENODEV;
+
+ spin_lock_irqsave(&dd->hfi1_snoop.snoop_lock, flags);
+
+ poll_wait(fp, &dd->hfi1_snoop.waitq, wait);
+ if (!list_empty(&dd->hfi1_snoop.queue))
+ ret |= POLLIN | POLLRDNORM;
+
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+ return ret;
+
+}
+
+static ssize_t hfi1_snoop_write(struct file *fp, const char __user *data,
+ size_t count, loff_t *off)
+{
+ struct diag_pkt dpkt;
+ struct hfi1_devdata *dd;
+ size_t ret;
+ u8 byte_two, sl, sc5, sc4, vl, byte_one;
+ struct send_context *sc;
+ u32 len;
+ u64 pbc;
+ struct hfi1_ibport *ibp;
+ struct hfi1_pportdata *ppd;
+
+ dd = hfi1_dd_from_sc_inode(fp->f_inode);
+ if (dd == NULL)
+ return -ENODEV;
+
+ ppd = dd->pport;
+ snoop_dbg("received %lu bytes from user", count);
+
+ memset(&dpkt, 0, sizeof(struct diag_pkt));
+ dpkt.version = _DIAG_PKT_VERS;
+ dpkt.unit = dd->unit;
+ dpkt.port = 1;
+
+ if (likely(!(snoop_flags & SNOOP_USE_METADATA))) {
+ /*
+ * We need to generate the PBC and not let diagpkt_send do it,
+ * to do this we need the VL and the length in dwords.
+ * The VL can be determined by using the SL and looking up the
+ * SC. Then the SC can be converted into VL. The exception to
+ * this is those packets which are from an SMI queue pair.
+ * Since we can't detect anything about the QP here we have to
+ * rely on the SC. If its 0xF then we assume its SMI and
+ * do not look at the SL.
+ */
+ if (copy_from_user(&byte_one, data, 1))
+ return -EINVAL;
+
+ if (copy_from_user(&byte_two, data+1, 1))
+ return -EINVAL;
+
+ sc4 = (byte_one >> 4) & 0xf;
+ if (sc4 == 0xF) {
+ snoop_dbg("Detected VL15 packet ignoring SL in packet");
+ vl = sc4;
+ } else {
+ sl = (byte_two >> 4) & 0xf;
+ ibp = to_iport(&dd->verbs_dev.ibdev, 1);
+ sc5 = ibp->sl_to_sc[sl];
+ vl = sc_to_vlt(dd, sc5);
+ if (vl != sc4) {
+ snoop_dbg("VL %d does not match SC %d of packet",
+ vl, sc4);
+ return -EINVAL;
+ }
+ }
+
+ sc = dd->vld[vl].sc; /* Look up the context based on VL */
+ if (sc) {
+ dpkt.sw_index = sc->sw_index;
+ snoop_dbg("Sending on context %u(%u)", sc->sw_index,
+ sc->hw_context);
+ } else {
+ snoop_dbg("Could not find context for vl %d", vl);
+ return -EINVAL;
+ }
+
+ len = (count >> 2) + 2; /* Add in PBC */
+ pbc = create_pbc(ppd, 0, 0, vl, len);
+ } else {
+ if (copy_from_user(&pbc, data, sizeof(pbc)))
+ return -EINVAL;
+ vl = (pbc >> PBC_VL_SHIFT) & PBC_VL_MASK;
+ sc = dd->vld[vl].sc; /* Look up the context based on VL */
+ if (sc) {
+ dpkt.sw_index = sc->sw_index;
+ } else {
+ snoop_dbg("Could not find context for vl %d", vl);
+ return -EINVAL;
+ }
+ data += sizeof(pbc);
+ count -= sizeof(pbc);
+ }
+ dpkt.len = count;
+ dpkt.data = (unsigned long)data;
+
+ snoop_dbg("PBC: vl=0x%llx Length=0x%llx",
+ (pbc >> 12) & 0xf,
+ (pbc & 0xfff));
+
+ dpkt.pbc = pbc;
+ ret = diagpkt_send(&dpkt);
+ /*
+ * diagpkt_send only returns number of bytes in the diagpkt so patch
+ * that up here before returning.
+ */
+ if (ret == sizeof(dpkt))
+ return count;
+
+ return ret;
+}
+
+static ssize_t hfi1_snoop_read(struct file *fp, char __user *data,
+ size_t pkt_len, loff_t *off)
+{
+ ssize_t ret = 0;
+ unsigned long flags = 0;
+ struct snoop_packet *packet = NULL;
+ struct hfi1_devdata *dd;
+
+ dd = hfi1_dd_from_sc_inode(fp->f_inode);
+ if (dd == NULL)
+ return -ENODEV;
+
+ spin_lock_irqsave(&dd->hfi1_snoop.snoop_lock, flags);
+
+ while (list_empty(&dd->hfi1_snoop.queue)) {
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+
+ if (fp->f_flags & O_NONBLOCK)
+ return -EAGAIN;
+
+ if (wait_event_interruptible(
+ dd->hfi1_snoop.waitq,
+ !list_empty(&dd->hfi1_snoop.queue)))
+ return -EINTR;
+
+ spin_lock_irqsave(&dd->hfi1_snoop.snoop_lock, flags);
+ }
+
+ if (!list_empty(&dd->hfi1_snoop.queue)) {
+ packet = list_entry(dd->hfi1_snoop.queue.next,
+ struct snoop_packet, list);
+ list_del(&packet->list);
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+ if (pkt_len >= packet->total_len) {
+ if (copy_to_user(data, packet->data,
+ packet->total_len))
+ ret = -EFAULT;
+ else
+ ret = packet->total_len;
+ } else
+ ret = -EINVAL;
+
+ kfree(packet);
+ } else
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+
+ return ret;
+}
+
+static long hfi1_ioctl(struct file *fp, unsigned int cmd, unsigned long arg)
+{
+ struct hfi1_devdata *dd;
+ void *filter_value = NULL;
+ long ret = 0;
+ int value = 0;
+ u8 physState = 0;
+ u8 linkState = 0;
+ u16 devState = 0;
+ unsigned long flags = 0;
+ unsigned long *argp = NULL;
+ struct hfi1_packet_filter_command filter_cmd = {0};
+ int mode_flag = 0;
+ struct hfi1_pportdata *ppd = NULL;
+ unsigned int index;
+ struct hfi1_link_info link_info;
+
+ dd = hfi1_dd_from_sc_inode(fp->f_inode);
+ if (dd == NULL)
+ return -ENODEV;
+
+ spin_lock_irqsave(&dd->hfi1_snoop.snoop_lock, flags);
+
+ mode_flag = dd->hfi1_snoop.mode_flag;
+
+ if (((_IOC_DIR(cmd) & _IOC_READ)
+ && !access_ok(VERIFY_WRITE, (void __user *)arg, _IOC_SIZE(cmd)))
+ || ((_IOC_DIR(cmd) & _IOC_WRITE)
+ && !access_ok(VERIFY_READ, (void __user *)arg, _IOC_SIZE(cmd)))) {
+ ret = -EFAULT;
+ } else if (!capable(CAP_SYS_ADMIN)) {
+ ret = -EPERM;
+ } else if ((mode_flag & HFI1_PORT_CAPTURE_MODE) &&
+ (cmd != HFI1_SNOOP_IOCCLEARQUEUE) &&
+ (cmd != HFI1_SNOOP_IOCCLEARFILTER) &&
+ (cmd != HFI1_SNOOP_IOCSETFILTER)) {
+ /* Capture devices are allowed only 3 operations
+ * 1.Clear capture queue
+ * 2.Clear capture filter
+ * 3.Set capture filter
+ * Other are invalid.
+ */
+ ret = -EINVAL;
+ } else {
+ switch (cmd) {
+ case HFI1_SNOOP_IOCSETLINKSTATE:
+ snoop_dbg("HFI1_SNOOP_IOCSETLINKSTATE is not valid");
+ ret = -EINVAL;
+ break;
+
+ case HFI1_SNOOP_IOCSETLINKSTATE_EXTRA:
+ memset(&link_info, 0, sizeof(link_info));
+
+ ret = copy_from_user(&link_info,
+ (struct hfi1_link_info __user *)arg,
+ sizeof(link_info));
+ if (ret)
+ break;
+
+ value = link_info.port_state;
+ index = link_info.port_number;
+ if (index > dd->num_pports - 1) {
+ ret = -EINVAL;
+ break;
+ }
+
+ ppd = &dd->pport[index];
+ if (!ppd) {
+ ret = -EINVAL;
+ break;
+ }
+
+ /* What we want to transition to */
+ physState = (value >> 4) & 0xF;
+ linkState = value & 0xF;
+ snoop_dbg("Setting link state 0x%x", value);
+
+ switch (linkState) {
+ case IB_PORT_NOP:
+ if (physState == 0)
+ break;
+ /* fall through */
+ case IB_PORT_DOWN:
+ switch (physState) {
+ case 0:
+ devState = HLS_DN_DOWNDEF;
+ break;
+ case 2:
+ devState = HLS_DN_POLL;
+ break;
+ case 3:
+ devState = HLS_DN_DISABLE;
+ break;
+ default:
+ ret = -EINVAL;
+ goto done;
+ }
+ ret = set_link_state(ppd, devState);
+ break;
+ case IB_PORT_ARMED:
+ ret = set_link_state(ppd, HLS_UP_ARMED);
+ if (!ret)
+ send_idle_sma(dd, SMA_IDLE_ARM);
+ break;
+ case IB_PORT_ACTIVE:
+ ret = set_link_state(ppd, HLS_UP_ACTIVE);
+ if (!ret)
+ send_idle_sma(dd, SMA_IDLE_ACTIVE);
+ break;
+ default:
+ ret = -EINVAL;
+ break;
+ }
+
+ if (ret)
+ break;
+ /* fall through */
+ case HFI1_SNOOP_IOCGETLINKSTATE:
+ case HFI1_SNOOP_IOCGETLINKSTATE_EXTRA:
+ if (cmd == HFI1_SNOOP_IOCGETLINKSTATE_EXTRA) {
+ memset(&link_info, 0, sizeof(link_info));
+ ret = copy_from_user(&link_info,
+ (struct hfi1_link_info __user *)arg,
+ sizeof(link_info));
+ index = link_info.port_number;
+ } else {
+ ret = __get_user(index, (int __user *) arg);
+ if (ret != 0)
+ break;
+ }
+
+ if (index > dd->num_pports - 1) {
+ ret = -EINVAL;
+ break;
+ }
+
+ ppd = &dd->pport[index];
+ if (!ppd) {
+ ret = -EINVAL;
+ break;
+ }
+ value = hfi1_ibphys_portstate(ppd);
+ value <<= 4;
+ value |= driver_lstate(ppd);
+
+ snoop_dbg("Link port | Link State: %d", value);
+
+ if ((cmd == HFI1_SNOOP_IOCGETLINKSTATE_EXTRA) ||
+ (cmd == HFI1_SNOOP_IOCSETLINKSTATE_EXTRA)) {
+ link_info.port_state = value;
+ link_info.node_guid = cpu_to_be64(ppd->guid);
+ link_info.link_speed_active =
+ ppd->link_speed_active;
+ link_info.link_width_active =
+ ppd->link_width_active;
+ ret = copy_to_user(
+ (struct hfi1_link_info __user *)arg,
+ &link_info, sizeof(link_info));
+ } else {
+ ret = __put_user(value, (int __user *)arg);
+ }
+ break;
+
+ case HFI1_SNOOP_IOCCLEARQUEUE:
+ snoop_dbg("Clearing snoop queue");
+ drain_snoop_list(&dd->hfi1_snoop.queue);
+ break;
+
+ case HFI1_SNOOP_IOCCLEARFILTER:
+ snoop_dbg("Clearing filter");
+ if (dd->hfi1_snoop.filter_callback) {
+ /* Drain packets first */
+ drain_snoop_list(&dd->hfi1_snoop.queue);
+ dd->hfi1_snoop.filter_callback = NULL;
+ }
+ kfree(dd->hfi1_snoop.filter_value);
+ dd->hfi1_snoop.filter_value = NULL;
+ break;
+
+ case HFI1_SNOOP_IOCSETFILTER:
+ snoop_dbg("Setting filter");
+ /* just copy command structure */
+ argp = (unsigned long *)arg;
+ ret = copy_from_user(&filter_cmd, (void __user *)argp,
+ sizeof(filter_cmd));
+ if (ret < 0) {
+ pr_alert("Error copying filter command\n");
+ break;
+ }
+ if (filter_cmd.opcode >= HFI1_MAX_FILTERS) {
+ pr_alert("Invalid opcode in request\n");
+ ret = -EINVAL;
+ break;
+ }
+
+ snoop_dbg("Opcode %d Len %d Ptr %p",
+ filter_cmd.opcode, filter_cmd.length,
+ filter_cmd.value_ptr);
+
+ filter_value = kzalloc(
+ filter_cmd.length * sizeof(u8),
+ GFP_KERNEL);
+ if (!filter_value) {
+ pr_alert("Not enough memory\n");
+ ret = -ENOMEM;
+ break;
+ }
+ /* copy remaining data from userspace */
+ ret = copy_from_user((u8 *)filter_value,
+ (void __user *)filter_cmd.value_ptr,
+ filter_cmd.length);
+ if (ret < 0) {
+ kfree(filter_value);
+ pr_alert("Error copying filter data\n");
+ break;
+ }
+ /* Drain packets first */
+ drain_snoop_list(&dd->hfi1_snoop.queue);
+ dd->hfi1_snoop.filter_callback =
+ hfi1_filters[filter_cmd.opcode].filter;
+ /* just in case we see back to back sets */
+ kfree(dd->hfi1_snoop.filter_value);
+ dd->hfi1_snoop.filter_value = filter_value;
+
+ break;
+ case HFI1_SNOOP_IOCGETVERSION:
+ value = SNOOP_CAPTURE_VERSION;
+ snoop_dbg("Getting version: %d", value);
+ ret = __put_user(value, (int __user *)arg);
+ break;
+ case HFI1_SNOOP_IOCSET_OPTS:
+ snoop_flags = 0;
+ ret = __get_user(value, (int __user *) arg);
+ if (ret != 0)
+ break;
+
+ snoop_dbg("Setting snoop option %d", value);
+ if (value & SNOOP_DROP_SEND)
+ snoop_flags |= SNOOP_DROP_SEND;
+ if (value & SNOOP_USE_METADATA)
+ snoop_flags |= SNOOP_USE_METADATA;
+ break;
+ default:
+ ret = -ENOTTY;
+ break;
+ }
+ }
+done:
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+ return ret;
+}
+
+static void snoop_list_add_tail(struct snoop_packet *packet,
+ struct hfi1_devdata *dd)
+{
+ unsigned long flags = 0;
+
+ spin_lock_irqsave(&dd->hfi1_snoop.snoop_lock, flags);
+ if (likely((dd->hfi1_snoop.mode_flag & HFI1_PORT_SNOOP_MODE) ||
+ (dd->hfi1_snoop.mode_flag & HFI1_PORT_CAPTURE_MODE))) {
+ list_add_tail(&packet->list, &dd->hfi1_snoop.queue);
+ snoop_dbg("Added packet to list");
+ }
+
+ /*
+ * Technically we can could have closed the snoop device while waiting
+ * on the above lock and it is gone now. The snoop mode_flag will
+ * prevent us from adding the packet to the queue though.
+ */
+
+ spin_unlock_irqrestore(&dd->hfi1_snoop.snoop_lock, flags);
+ wake_up_interruptible(&dd->hfi1_snoop.waitq);
+}
+
+static inline int hfi1_filter_check(void *val, const char *msg)
+{
+ if (!val) {
+ snoop_dbg("Error invalid %s value for filter", msg);
+ return HFI1_FILTER_ERR;
+ }
+ return 0;
+}
+
+static int hfi1_filter_lid(void *ibhdr, void *packet_data, void *value)
+{
+ struct hfi1_ib_header *hdr;
+ int ret;
+
+ ret = hfi1_filter_check(ibhdr, "header");
+ if (ret)
+ return ret;
+ ret = hfi1_filter_check(value, "user");
+ if (ret)
+ return ret;
+ hdr = (struct hfi1_ib_header *)ibhdr;
+
+ if (*((u16 *)value) == be16_to_cpu(hdr->lrh[3])) /* matches slid */
+ return HFI1_FILTER_HIT; /* matched */
+
+ return HFI1_FILTER_MISS; /* Not matched */
+}
+
+static int hfi1_filter_dlid(void *ibhdr, void *packet_data, void *value)
+{
+ struct hfi1_ib_header *hdr;
+ int ret;
+
+ ret = hfi1_filter_check(ibhdr, "header");
+ if (ret)
+ return ret;
+ ret = hfi1_filter_check(value, "user");
+ if (ret)
+ return ret;
+
+ hdr = (struct hfi1_ib_header *)ibhdr;
+
+ if (*((u16 *)value) == be16_to_cpu(hdr->lrh[1]))
+ return HFI1_FILTER_HIT;
+
+ return HFI1_FILTER_MISS;
+}
+
+/* Not valid for outgoing packets, send handler passes null for data*/
+static int hfi1_filter_mad_mgmt_class(void *ibhdr, void *packet_data,
+ void *value)
+{
+ struct hfi1_ib_header *hdr;
+ struct hfi1_other_headers *ohdr = NULL;
+ struct ib_smp *smp = NULL;
+ u32 qpn = 0;
+ int ret;
+
+ ret = hfi1_filter_check(ibhdr, "header");
+ if (ret)
+ return ret;
+ ret = hfi1_filter_check(packet_data, "packet_data");
+ if (ret)
+ return ret;
+ ret = hfi1_filter_check(value, "user");
+ if (ret)
+ return ret;
+
+ hdr = (struct hfi1_ib_header *)ibhdr;
+
+ /* Check for GRH */
+ if ((be16_to_cpu(hdr->lrh[0]) & 3) == HFI1_LRH_BTH)
+ ohdr = &hdr->u.oth; /* LRH + BTH + DETH */
+ else
+ ohdr = &hdr->u.l.oth; /* LRH + GRH + BTH + DETH */
+
+ qpn = be32_to_cpu(ohdr->bth[1]) & 0x00FFFFFF;
+ if (qpn <= 1) {
+ smp = (struct ib_smp *)packet_data;
+ if (*((u8 *)value) == smp->mgmt_class)
+ return HFI1_FILTER_HIT;
+ else
+ return HFI1_FILTER_MISS;
+ }
+ return HFI1_FILTER_ERR;
+}
+
+static int hfi1_filter_qp_number(void *ibhdr, void *packet_data, void *value)
+{
+
+ struct hfi1_ib_header *hdr;
+ struct hfi1_other_headers *ohdr = NULL;
+ int ret;
+
+ ret = hfi1_filter_check(ibhdr, "header");
+ if (ret)
+ return ret;
+ ret = hfi1_filter_check(value, "user");
+ if (ret)
+ return ret;
+
+ hdr = (struct hfi1_ib_header *)ibhdr;
+
+ /* Check for GRH */
+ if ((be16_to_cpu(hdr->lrh[0]) & 3) == HFI1_LRH_BTH)
+ ohdr = &hdr->u.oth; /* LRH + BTH + DETH */
+ else
+ ohdr = &hdr->u.l.oth; /* LRH + GRH + BTH + DETH */
+ if (*((u32 *)value) == (be32_to_cpu(ohdr->bth[1]) & 0x00FFFFFF))
+ return HFI1_FILTER_HIT;
+
+ return HFI1_FILTER_MISS;
+}
+
+static int hfi1_filter_ibpacket_type(void *ibhdr, void *packet_data,
+ void *value)
+{
+ u32 lnh = 0;
+ u8 opcode = 0;
+ struct hfi1_ib_header *hdr;
+ struct hfi1_other_headers *ohdr = NULL;
+ int ret;
+
+ ret = hfi1_filter_check(ibhdr, "header");
+ if (ret)
+ return ret;
+ ret = hfi1_filter_check(value, "user");
+ if (ret)
+ return ret;
+
+ hdr = (struct hfi1_ib_header *)ibhdr;
+
+ lnh = (be16_to_cpu(hdr->lrh[0]) & 3);
+
+ if (lnh == HFI1_LRH_BTH)
+ ohdr = &hdr->u.oth;
+ else if (lnh == HFI1_LRH_GRH)
+ ohdr = &hdr->u.l.oth;
+ else
+ return HFI1_FILTER_ERR;
+
+ opcode = be32_to_cpu(ohdr->bth[0]) >> 24;
+
+ if (*((u8 *)value) == ((opcode >> 5) & 0x7))
+ return HFI1_FILTER_HIT;
+
+ return HFI1_FILTER_MISS;
+}
+
+static int hfi1_filter_ib_service_level(void *ibhdr, void *packet_data,
+ void *value)
+{
+ struct hfi1_ib_header *hdr;
+ int ret;
+
+ ret = hfi1_filter_check(ibhdr, "header");
+ if (ret)
+ return ret;
+ ret = hfi1_filter_check(value, "user");
+ if (ret)
+ return ret;
+
+ hdr = (struct hfi1_ib_header *)ibhdr;
+
+ if ((*((u8 *)value)) == ((be16_to_cpu(hdr->lrh[0]) >> 4) & 0xF))
+ return HFI1_FILTER_HIT;
+
+ return HFI1_FILTER_MISS;
+}
+
+static int hfi1_filter_ib_pkey(void *ibhdr, void *packet_data, void *value)
+{
+
+ u32 lnh = 0;
+ struct hfi1_ib_header *hdr;
+ struct hfi1_other_headers *ohdr = NULL;
+ int ret;
+
+ ret = hfi1_filter_check(ibhdr, "header");
+ if (ret)
+ return ret;
+ ret = hfi1_filter_check(value, "user");
+ if (ret)
+ return ret;
+
+ hdr = (struct hfi1_ib_header *)ibhdr;
+
+ lnh = (be16_to_cpu(hdr->lrh[0]) & 3);
+ if (lnh == HFI1_LRH_BTH)
+ ohdr = &hdr->u.oth;
+ else if (lnh == HFI1_LRH_GRH)
+ ohdr = &hdr->u.l.oth;
+ else
+ return HFI1_FILTER_ERR;
+
+ /* P_key is 16-bit entity, however top most bit indicates
+ * type of membership. 0 for limited and 1 for Full.
+ * Limited members cannot accept information from other
+ * Limited members, but communication is allowed between
+ * every other combination of membership.
+ * Hence we'll omit comparing top-most bit while filtering
+ */
+
+ if ((*(u16 *)value & 0x7FFF) ==
+ ((be32_to_cpu(ohdr->bth[0])) & 0x7FFF))
+ return HFI1_FILTER_HIT;
+
+ return HFI1_FILTER_MISS;
+}
+
+/*
+ * If packet_data is NULL then this is coming from one of the send functions.
+ * Thus we know if its an ingressed or egressed packet.
+ */
+static int hfi1_filter_direction(void *ibhdr, void *packet_data, void *value)
+{
+ u8 user_dir = *(u8 *)value;
+ int ret;
+
+ ret = hfi1_filter_check(value, "user");
+ if (ret)
+ return ret;
+
+ if (packet_data) {
+ /* Incoming packet */
+ if (user_dir & HFI1_SNOOP_INGRESS)
+ return HFI1_FILTER_HIT;
+ } else {
+ /* Outgoing packet */
+ if (user_dir & HFI1_SNOOP_EGRESS)
+ return HFI1_FILTER_HIT;
+ }
+
+ return HFI1_FILTER_MISS;
+}
+
+/*
+ * Allocate a snoop packet. The structure that is stored in the ring buffer, not
+ * to be confused with an hfi packet type.
+ */
+static struct snoop_packet *allocate_snoop_packet(u32 hdr_len,
+ u32 data_len,
+ u32 md_len)
+{
+
+ struct snoop_packet *packet = NULL;
+
+ packet = kzalloc(sizeof(struct snoop_packet) + hdr_len + data_len
+ + md_len,
+ GFP_ATOMIC | __GFP_NOWARN);
+ if (likely(packet))
+ INIT_LIST_HEAD(&packet->list);
+
+
+ return packet;
+}
+
+/*
+ * Instead of having snoop and capture code intermixed with the recv functions,
+ * both the interrupt handler and hfi1_ib_rcv() we are going to hijack the call
+ * and land in here for snoop/capture but if not enabled the call will go
+ * through as before. This gives us a single point to constrain all of the snoop
+ * snoop recv logic. There is nothing special that needs to happen for bypass
+ * packets. This routine should not try to look into the packet. It just copied
+ * it. There is no guarantee for filters when it comes to bypass packets as
+ * there is no specific support. Bottom line is this routine does now even know
+ * what a bypass packet is.
+ */
+int snoop_recv_handler(struct hfi1_packet *packet)
+{
+ struct hfi1_pportdata *ppd = packet->rcd->ppd;
+ struct hfi1_ib_header *hdr = packet->hdr;
+ int header_size = packet->hlen;
+ void *data = packet->ebuf;
+ u32 tlen = packet->tlen;
+ struct snoop_packet *s_packet = NULL;
+ int ret;
+ int snoop_mode = 0;
+ u32 md_len = 0;
+ struct capture_md md;
+
+ snoop_dbg("PACKET IN: hdr size %d tlen %d data %p", header_size, tlen,
+ data);
+
+ trace_snoop_capture(ppd->dd, header_size, hdr, tlen - header_size,
+ data);
+
+ if (!ppd->dd->hfi1_snoop.filter_callback) {
+ snoop_dbg("filter not set");
+ ret = HFI1_FILTER_HIT;
+ } else {
+ ret = ppd->dd->hfi1_snoop.filter_callback(hdr, data,
+ ppd->dd->hfi1_snoop.filter_value);
+ }
+
+ switch (ret) {
+ case HFI1_FILTER_ERR:
+ snoop_dbg("Error in filter call");
+ break;
+ case HFI1_FILTER_MISS:
+ snoop_dbg("Filter Miss");
+ break;
+ case HFI1_FILTER_HIT:
+
+ if (ppd->dd->hfi1_snoop.mode_flag & HFI1_PORT_SNOOP_MODE)
+ snoop_mode = 1;
+ if ((snoop_mode == 0) ||
+ unlikely(snoop_flags & SNOOP_USE_METADATA))
+ md_len = sizeof(struct capture_md);
+
+
+ s_packet = allocate_snoop_packet(header_size,
+ tlen - header_size,
+ md_len);
+
+ if (unlikely(s_packet == NULL)) {
+ dd_dev_warn_ratelimited(ppd->dd, "Unable to allocate snoop/capture packet\n");
+ break;
+ }
+
+ if (md_len > 0) {
+ memset(&md, 0, sizeof(struct capture_md));
+ md.port = 1;
+ md.dir = PKT_DIR_INGRESS;
+ md.u.rhf = packet->rhf;
+ memcpy(s_packet->data, &md, md_len);
+ }
+
+ /* We should always have a header */
+ if (hdr) {
+ memcpy(s_packet->data + md_len, hdr, header_size);
+ } else {
+ dd_dev_err(ppd->dd, "Unable to copy header to snoop/capture packet\n");
+ kfree(s_packet);
+ break;
+ }
+
+ /*
+ * Packets with no data are possible. If there is no data needed
+ * to take care of the last 4 bytes which are normally included
+ * with data buffers and are included in tlen. Since we kzalloc
+ * the buffer we do not need to set any values but if we decide
+ * not to use kzalloc we should zero them.
+ */
+ if (data)
+ memcpy(s_packet->data + header_size + md_len, data,
+ tlen - header_size);
+
+ s_packet->total_len = tlen + md_len;
+ snoop_list_add_tail(s_packet, ppd->dd);
+
+ /*
+ * If we are snooping the packet not capturing then throw away
+ * after adding to the list.
+ */
+ snoop_dbg("Capturing packet");
+ if (ppd->dd->hfi1_snoop.mode_flag & HFI1_PORT_SNOOP_MODE) {
+ snoop_dbg("Throwing packet away");
+ /*
+ * If we are dropping the packet we still may need to
+ * handle the case where error flags are set, this is
+ * normally done by the type specific handler but that
+ * won't be called in this case.
+ */
+ if (unlikely(rhf_err_flags(packet->rhf)))
+ handle_eflags(packet);
+
+ /* throw the packet on the floor */
+ return RHF_RCV_CONTINUE;
+ }
+ break;
+ default:
+ break;
+ }
+
+ /*
+ * We do not care what type of packet came in here - just pass it off
+ * to the normal handler.
+ */
+ return ppd->dd->normal_rhf_rcv_functions[rhf_rcv_type(packet->rhf)]
+ (packet);
+}
+
+/*
+ * Handle snooping and capturing packets when sdma is being used.
+ */
+int snoop_send_dma_handler(struct hfi1_qp *qp, struct ahg_ib_header *ibhdr,
+ u32 hdrwords, struct hfi1_sge_state *ss, u32 len,
+ u32 plen, u32 dwords, u64 pbc)
+{
+ pr_alert("Snooping/Capture of Send DMA Packets Is Not Supported!\n");
+ snoop_dbg("Unsupported Operation");
+ return hfi1_verbs_send_dma(qp, ibhdr, hdrwords, ss, len, plen, dwords,
+ 0);
+}
+
+/*
+ * Handle snooping and capturing packets when pio is being used. Does not handle
+ * bypass packets. The only way to send a bypass packet currently is to use the
+ * diagpkt interface. When that interface is enable snoop/capture is not.
+ */
+int snoop_send_pio_handler(struct hfi1_qp *qp, struct ahg_ib_header *ahdr,
+ u32 hdrwords, struct hfi1_sge_state *ss, u32 len,
+ u32 plen, u32 dwords, u64 pbc)
+{
+ struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+ struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
+ struct snoop_packet *s_packet = NULL;
+ u32 *hdr = (u32 *)&ahdr->ibh;
+ u32 length = 0;
+ struct hfi1_sge_state temp_ss;
+ void *data = NULL;
+ void *data_start = NULL;
+ int ret;
+ int snoop_mode = 0;
+ int md_len = 0;
+ struct capture_md md;
+ u32 vl;
+ u32 hdr_len = hdrwords << 2;
+ u32 tlen = HFI1_GET_PKT_LEN(&ahdr->ibh);
+
+ md.u.pbc = 0;
+
+ snoop_dbg("PACKET OUT: hdrword %u len %u plen %u dwords %u tlen %u",
+ hdrwords, len, plen, dwords, tlen);
+ if (ppd->dd->hfi1_snoop.mode_flag & HFI1_PORT_SNOOP_MODE)
+ snoop_mode = 1;
+ if ((snoop_mode == 0) ||
+ unlikely(snoop_flags & SNOOP_USE_METADATA))
+ md_len = sizeof(struct capture_md);
+
+ /* not using ss->total_len as arg 2 b/c that does not count CRC */
+ s_packet = allocate_snoop_packet(hdr_len, tlen - hdr_len, md_len);
+
+ if (unlikely(s_packet == NULL)) {
+ dd_dev_warn_ratelimited(ppd->dd, "Unable to allocate snoop/capture packet\n");
+ goto out;
+ }
+
+ s_packet->total_len = tlen + md_len;
+
+ if (md_len > 0) {
+ memset(&md, 0, sizeof(struct capture_md));
+ md.port = 1;
+ md.dir = PKT_DIR_EGRESS;
+ if (likely(pbc == 0)) {
+ vl = be16_to_cpu(ahdr->ibh.lrh[0]) >> 12;
+ md.u.pbc = create_pbc(ppd, 0, qp->s_srate, vl, plen);
+ } else {
+ md.u.pbc = 0;
+ }
+ memcpy(s_packet->data, &md, md_len);
+ } else {
+ md.u.pbc = pbc;
+ }
+
+ /* Copy header */
+ if (likely(hdr)) {
+ memcpy(s_packet->data + md_len, hdr, hdr_len);
+ } else {
+ dd_dev_err(ppd->dd,
+ "Unable to copy header to snoop/capture packet\n");
+ kfree(s_packet);
+ goto out;
+ }
+
+ if (ss) {
+ data = s_packet->data + hdr_len + md_len;
+ data_start = data;
+
+ /*
+ * Copy SGE State
+ * The update_sge() function below will not modify the
+ * individual SGEs in the array. It will make a copy each time
+ * and operate on that. So we only need to copy this instance
+ * and it won't impact PIO.
+ */
+ temp_ss = *ss;
+ length = len;
+
+ snoop_dbg("Need to copy %d bytes", length);
+ while (length) {
+ void *addr = temp_ss.sge.vaddr;
+ u32 slen = temp_ss.sge.length;
+
+ if (slen > length) {
+ slen = length;
+ snoop_dbg("slen %d > len %d", slen, length);
+ }
+ snoop_dbg("copy %d to %p", slen, addr);
+ memcpy(data, addr, slen);
+ update_sge(&temp_ss, slen);
+ length -= slen;
+ data += slen;
+ snoop_dbg("data is now %p bytes left %d", data, length);
+ }
+ snoop_dbg("Completed SGE copy");
+ }
+
+ /*
+ * Why do the filter check down here? Because the event tracing has its
+ * own filtering and we need to have the walked the SGE list.
+ */
+ if (!ppd->dd->hfi1_snoop.filter_callback) {
+ snoop_dbg("filter not set\n");
+ ret = HFI1_FILTER_HIT;
+ } else {
+ ret = ppd->dd->hfi1_snoop.filter_callback(
+ &ahdr->ibh,
+ NULL,
+ ppd->dd->hfi1_snoop.filter_value);
+ }
+
+ switch (ret) {
+ case HFI1_FILTER_ERR:
+ snoop_dbg("Error in filter call");
+ /* fall through */
+ case HFI1_FILTER_MISS:
+ snoop_dbg("Filter Miss");
+ kfree(s_packet);
+ break;
+ case HFI1_FILTER_HIT:
+ snoop_dbg("Capturing packet");
+ snoop_list_add_tail(s_packet, ppd->dd);
+
+ if (unlikely((snoop_flags & SNOOP_DROP_SEND) &&
+ (ppd->dd->hfi1_snoop.mode_flag &
+ HFI1_PORT_SNOOP_MODE))) {
+ unsigned long flags;
+
+ snoop_dbg("Dropping packet");
+ if (qp->s_wqe) {
+ spin_lock_irqsave(&qp->s_lock, flags);
+ hfi1_send_complete(
+ qp,
+ qp->s_wqe,
+ IB_WC_SUCCESS);
+ spin_unlock_irqrestore(&qp->s_lock, flags);
+ } else if (qp->ibqp.qp_type == IB_QPT_RC) {
+ spin_lock_irqsave(&qp->s_lock, flags);
+ hfi1_rc_send_complete(qp, &ahdr->ibh);
+ spin_unlock_irqrestore(&qp->s_lock, flags);
+ }
+ return 0;
+ }
+ break;
+ default:
+ kfree(s_packet);
+ break;
+ }
+out:
+ return hfi1_verbs_send_pio(qp, ahdr, hdrwords, ss, len, plen, dwords,
+ md.u.pbc);
+}
+
+/*
+ * Callers of this must pass a hfi1_ib_header type for the from ptr. Currently
+ * this can be used anywhere, but the intention is for inline ACKs for RC and
+ * CCA packets. We don't restrict this usage though.
+ */
+void snoop_inline_pio_send(struct hfi1_devdata *dd, struct pio_buf *pbuf,
+ u64 pbc, const void *from, size_t count)
+{
+ int snoop_mode = 0;
+ int md_len = 0;
+ struct capture_md md;
+ struct snoop_packet *s_packet = NULL;
+
+ /*
+ * count is in dwords so we need to convert to bytes.
+ * We also need to account for CRC which would be tacked on by hardware.
+ */
+ int packet_len = (count << 2) + 4;
+ int ret;
+
+ snoop_dbg("ACK OUT: len %d", packet_len);
+
+ if (!dd->hfi1_snoop.filter_callback) {
+ snoop_dbg("filter not set");
+ ret = HFI1_FILTER_HIT;
+ } else {
+ ret = dd->hfi1_snoop.filter_callback(
+ (struct hfi1_ib_header *)from,
+ NULL,
+ dd->hfi1_snoop.filter_value);
+ }
+
+ switch (ret) {
+ case HFI1_FILTER_ERR:
+ snoop_dbg("Error in filter call");
+ /* fall through */
+ case HFI1_FILTER_MISS:
+ snoop_dbg("Filter Miss");
+ break;
+ case HFI1_FILTER_HIT:
+ snoop_dbg("Capturing packet");
+ if (dd->hfi1_snoop.mode_flag & HFI1_PORT_SNOOP_MODE)
+ snoop_mode = 1;
+ if ((snoop_mode == 0) ||
+ unlikely(snoop_flags & SNOOP_USE_METADATA))
+ md_len = sizeof(struct capture_md);
+
+ s_packet = allocate_snoop_packet(packet_len, 0, md_len);
+
+ if (unlikely(s_packet == NULL)) {
+ dd_dev_warn_ratelimited(dd, "Unable to allocate snoop/capture packet\n");
+ goto inline_pio_out;
+ }
+
+ s_packet->total_len = packet_len + md_len;
+
+ /* Fill in the metadata for the packet */
+ if (md_len > 0) {
+ memset(&md, 0, sizeof(struct capture_md));
+ md.port = 1;
+ md.dir = PKT_DIR_EGRESS;
+ md.u.pbc = pbc;
+ memcpy(s_packet->data, &md, md_len);
+ }
+
+ /* Add the packet data which is a single buffer */
+ memcpy(s_packet->data + md_len, from, packet_len);
+
+ snoop_list_add_tail(s_packet, dd);
+
+ if (unlikely((snoop_flags & SNOOP_DROP_SEND) && snoop_mode)) {
+ snoop_dbg("Dropping packet");
+ return;
+ }
+ break;
+ default:
+ break;
+ }
+
+inline_pio_out:
+ pio_copy(dd, pbuf, pbc, from, count);
+
+}
diff --git a/drivers/staging/rdma/hfi1/dma.c b/drivers/staging/rdma/hfi1/dma.c
new file mode 100644
index 000000000000..e03bd735173c
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/dma.c
@@ -0,0 +1,186 @@
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+#include <linux/types.h>
+#include <linux/scatterlist.h>
+
+#include "verbs.h"
+
+#define BAD_DMA_ADDRESS ((u64) 0)
+
+/*
+ * The following functions implement driver specific replacements
+ * for the ib_dma_*() functions.
+ *
+ * These functions return kernel virtual addresses instead of
+ * device bus addresses since the driver uses the CPU to copy
+ * data instead of using hardware DMA.
+ */
+
+static int hfi1_mapping_error(struct ib_device *dev, u64 dma_addr)
+{
+ return dma_addr == BAD_DMA_ADDRESS;
+}
+
+static u64 hfi1_dma_map_single(struct ib_device *dev, void *cpu_addr,
+ size_t size, enum dma_data_direction direction)
+{
+ if (WARN_ON(!valid_dma_direction(direction)))
+ return BAD_DMA_ADDRESS;
+
+ return (u64) cpu_addr;
+}
+
+static void hfi1_dma_unmap_single(struct ib_device *dev, u64 addr, size_t size,
+ enum dma_data_direction direction)
+{
+ /* This is a stub, nothing to be done here */
+}
+
+static u64 hfi1_dma_map_page(struct ib_device *dev, struct page *page,
+ unsigned long offset, size_t size,
+ enum dma_data_direction direction)
+{
+ u64 addr;
+
+ if (WARN_ON(!valid_dma_direction(direction)))
+ return BAD_DMA_ADDRESS;
+
+ if (offset + size > PAGE_SIZE)
+ return BAD_DMA_ADDRESS;
+
+ addr = (u64) page_address(page);
+ if (addr)
+ addr += offset;
+
+ return addr;
+}
+
+static void hfi1_dma_unmap_page(struct ib_device *dev, u64 addr, size_t size,
+ enum dma_data_direction direction)
+{
+ /* This is a stub, nothing to be done here */
+}
+
+static int hfi1_map_sg(struct ib_device *dev, struct scatterlist *sgl,
+ int nents, enum dma_data_direction direction)
+{
+ struct scatterlist *sg;
+ u64 addr;
+ int i;
+ int ret = nents;
+
+ if (WARN_ON(!valid_dma_direction(direction)))
+ return BAD_DMA_ADDRESS;
+
+ for_each_sg(sgl, sg, nents, i) {
+ addr = (u64) page_address(sg_page(sg));
+ if (!addr) {
+ ret = 0;
+ break;
+ }
+ sg->dma_address = addr + sg->offset;
+#ifdef CONFIG_NEED_SG_DMA_LENGTH
+ sg->dma_length = sg->length;
+#endif
+ }
+ return ret;
+}
+
+static void hfi1_unmap_sg(struct ib_device *dev,
+ struct scatterlist *sg, int nents,
+ enum dma_data_direction direction)
+{
+ /* This is a stub, nothing to be done here */
+}
+
+static void hfi1_sync_single_for_cpu(struct ib_device *dev, u64 addr,
+ size_t size, enum dma_data_direction dir)
+{
+}
+
+static void hfi1_sync_single_for_device(struct ib_device *dev, u64 addr,
+ size_t size,
+ enum dma_data_direction dir)
+{
+}
+
+static void *hfi1_dma_alloc_coherent(struct ib_device *dev, size_t size,
+ u64 *dma_handle, gfp_t flag)
+{
+ struct page *p;
+ void *addr = NULL;
+
+ p = alloc_pages(flag, get_order(size));
+ if (p)
+ addr = page_address(p);
+ if (dma_handle)
+ *dma_handle = (u64) addr;
+ return addr;
+}
+
+static void hfi1_dma_free_coherent(struct ib_device *dev, size_t size,
+ void *cpu_addr, u64 dma_handle)
+{
+ free_pages((unsigned long) cpu_addr, get_order(size));
+}
+
+struct ib_dma_mapping_ops hfi1_dma_mapping_ops = {
+ .mapping_error = hfi1_mapping_error,
+ .map_single = hfi1_dma_map_single,
+ .unmap_single = hfi1_dma_unmap_single,
+ .map_page = hfi1_dma_map_page,
+ .unmap_page = hfi1_dma_unmap_page,
+ .map_sg = hfi1_map_sg,
+ .unmap_sg = hfi1_unmap_sg,
+ .sync_single_for_cpu = hfi1_sync_single_for_cpu,
+ .sync_single_for_device = hfi1_sync_single_for_device,
+ .alloc_coherent = hfi1_dma_alloc_coherent,
+ .free_coherent = hfi1_dma_free_coherent
+};
diff --git a/drivers/staging/rdma/hfi1/driver.c b/drivers/staging/rdma/hfi1/driver.c
new file mode 100644
index 000000000000..c0a59001e5cd
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/driver.c
@@ -0,0 +1,1241 @@
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <linux/spinlock.h>
+#include <linux/pci.h>
+#include <linux/io.h>
+#include <linux/delay.h>
+#include <linux/netdevice.h>
+#include <linux/vmalloc.h>
+#include <linux/module.h>
+#include <linux/prefetch.h>
+
+#include "hfi.h"
+#include "trace.h"
+#include "qp.h"
+#include "sdma.h"
+
+#undef pr_fmt
+#define pr_fmt(fmt) DRIVER_NAME ": " fmt
+
+/*
+ * The size has to be longer than this string, so we can append
+ * board/chip information to it in the initialization code.
+ */
+const char ib_hfi1_version[] = HFI1_DRIVER_VERSION "\n";
+
+DEFINE_SPINLOCK(hfi1_devs_lock);
+LIST_HEAD(hfi1_dev_list);
+DEFINE_MUTEX(hfi1_mutex); /* general driver use */
+
+unsigned int hfi1_max_mtu = HFI1_DEFAULT_MAX_MTU;
+module_param_named(max_mtu, hfi1_max_mtu, uint, S_IRUGO);
+MODULE_PARM_DESC(max_mtu, "Set max MTU bytes, default is 8192");
+
+unsigned int hfi1_cu = 1;
+module_param_named(cu, hfi1_cu, uint, S_IRUGO);
+MODULE_PARM_DESC(cu, "Credit return units");
+
+unsigned long hfi1_cap_mask = HFI1_CAP_MASK_DEFAULT;
+static int hfi1_caps_set(const char *, const struct kernel_param *);
+static int hfi1_caps_get(char *, const struct kernel_param *);
+static const struct kernel_param_ops cap_ops = {
+ .set = hfi1_caps_set,
+ .get = hfi1_caps_get
+};
+module_param_cb(cap_mask, &cap_ops, &hfi1_cap_mask, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(cap_mask, "Bit mask of enabled/disabled HW features");
+
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_DESCRIPTION("Intel Omni-Path Architecture driver");
+MODULE_VERSION(HFI1_DRIVER_VERSION);
+
+/*
+ * MAX_PKT_RCV is the max # if packets processed per receive interrupt.
+ */
+#define MAX_PKT_RECV 64
+#define EGR_HEAD_UPDATE_THRESHOLD 16
+
+struct hfi1_ib_stats hfi1_stats;
+
+static int hfi1_caps_set(const char *val, const struct kernel_param *kp)
+{
+ int ret = 0;
+ unsigned long *cap_mask_ptr = (unsigned long *)kp->arg,
+ cap_mask = *cap_mask_ptr, value, diff,
+ write_mask = ((HFI1_CAP_WRITABLE_MASK << HFI1_CAP_USER_SHIFT) |
+ HFI1_CAP_WRITABLE_MASK);
+
+ ret = kstrtoul(val, 0, &value);
+ if (ret) {
+ pr_warn("Invalid module parameter value for 'cap_mask'\n");
+ goto done;
+ }
+ /* Get the changed bits (except the locked bit) */
+ diff = value ^ (cap_mask & ~HFI1_CAP_LOCKED_SMASK);
+
+ /* Remove any bits that are not allowed to change after driver load */
+ if (HFI1_CAP_LOCKED() && (diff & ~write_mask)) {
+ pr_warn("Ignoring non-writable capability bits %#lx\n",
+ diff & ~write_mask);
+ diff &= write_mask;
+ }
+
+ /* Mask off any reserved bits */
+ diff &= ~HFI1_CAP_RESERVED_MASK;
+ /* Clear any previously set and changing bits */
+ cap_mask &= ~diff;
+ /* Update the bits with the new capability */
+ cap_mask |= (value & diff);
+ /* Check for any kernel/user restrictions */
+ diff = (cap_mask & (HFI1_CAP_MUST_HAVE_KERN << HFI1_CAP_USER_SHIFT)) ^
+ ((cap_mask & HFI1_CAP_MUST_HAVE_KERN) << HFI1_CAP_USER_SHIFT);
+ cap_mask &= ~diff;
+ /* Set the bitmask to the final set */
+ *cap_mask_ptr = cap_mask;
+done:
+ return ret;
+}
+
+static int hfi1_caps_get(char *buffer, const struct kernel_param *kp)
+{
+ unsigned long cap_mask = *(unsigned long *)kp->arg;
+
+ cap_mask &= ~HFI1_CAP_LOCKED_SMASK;
+ cap_mask |= ((cap_mask & HFI1_CAP_K2U) << HFI1_CAP_USER_SHIFT);
+
+ return scnprintf(buffer, PAGE_SIZE, "0x%lx", cap_mask);
+}
+
+const char *get_unit_name(int unit)
+{
+ static char iname[16];
+
+ snprintf(iname, sizeof(iname), DRIVER_NAME"_%u", unit);
+ return iname;
+}
+
+/*
+ * Return count of units with at least one port ACTIVE.
+ */
+int hfi1_count_active_units(void)
+{
+ struct hfi1_devdata *dd;
+ struct hfi1_pportdata *ppd;
+ unsigned long flags;
+ int pidx, nunits_active = 0;
+
+ spin_lock_irqsave(&hfi1_devs_lock, flags);
+ list_for_each_entry(dd, &hfi1_dev_list, list) {
+ if (!(dd->flags & HFI1_PRESENT) || !dd->kregbase)
+ continue;
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+ if (ppd->lid && ppd->linkup) {
+ nunits_active++;
+ break;
+ }
+ }
+ }
+ spin_unlock_irqrestore(&hfi1_devs_lock, flags);
+ return nunits_active;
+}
+
+/*
+ * Return count of all units, optionally return in arguments
+ * the number of usable (present) units, and the number of
+ * ports that are up.
+ */
+int hfi1_count_units(int *npresentp, int *nupp)
+{
+ int nunits = 0, npresent = 0, nup = 0;
+ struct hfi1_devdata *dd;
+ unsigned long flags;
+ int pidx;
+ struct hfi1_pportdata *ppd;
+
+ spin_lock_irqsave(&hfi1_devs_lock, flags);
+
+ list_for_each_entry(dd, &hfi1_dev_list, list) {
+ nunits++;
+ if ((dd->flags & HFI1_PRESENT) && dd->kregbase)
+ npresent++;
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+ if (ppd->lid && ppd->linkup)
+ nup++;
+ }
+ }
+
+ spin_unlock_irqrestore(&hfi1_devs_lock, flags);
+
+ if (npresentp)
+ *npresentp = npresent;
+ if (nupp)
+ *nupp = nup;
+
+ return nunits;
+}
+
+/*
+ * Get address of eager buffer from it's index (allocated in chunks, not
+ * contiguous).
+ */
+static inline void *get_egrbuf(const struct hfi1_ctxtdata *rcd, u64 rhf,
+ u8 *update)
+{
+ u32 idx = rhf_egr_index(rhf), offset = rhf_egr_buf_offset(rhf);
+
+ *update |= !(idx & (rcd->egrbufs.threshold - 1)) && !offset;
+ return (void *)(((u64)(rcd->egrbufs.rcvtids[idx].addr)) +
+ (offset * RCV_BUF_BLOCK_SIZE));
+}
+
+/*
+ * Validate and encode the a given RcvArray Buffer size.
+ * The function will check whether the given size falls within
+ * allowed size ranges for the respective type and, optionally,
+ * return the proper encoding.
+ */
+inline int hfi1_rcvbuf_validate(u32 size, u8 type, u16 *encoded)
+{
+ if (unlikely(!IS_ALIGNED(size, PAGE_SIZE)))
+ return 0;
+ if (unlikely(size < MIN_EAGER_BUFFER))
+ return 0;
+ if (size >
+ (type == PT_EAGER ? MAX_EAGER_BUFFER : MAX_EXPECTED_BUFFER))
+ return 0;
+ if (encoded)
+ *encoded = ilog2(size / PAGE_SIZE) + 1;
+ return 1;
+}
+
+static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
+ struct hfi1_packet *packet)
+{
+ struct hfi1_message_header *rhdr = packet->hdr;
+ u32 rte = rhf_rcv_type_err(packet->rhf);
+ int lnh = be16_to_cpu(rhdr->lrh[0]) & 3;
+ struct hfi1_ibport *ibp = &ppd->ibport_data;
+
+ if (packet->rhf & (RHF_VCRC_ERR | RHF_ICRC_ERR))
+ return;
+
+ if (packet->rhf & RHF_TID_ERR) {
+ /* For TIDERR and RC QPs preemptively schedule a NAK */
+ struct hfi1_ib_header *hdr = (struct hfi1_ib_header *)rhdr;
+ struct hfi1_other_headers *ohdr = NULL;
+ u32 tlen = rhf_pkt_len(packet->rhf); /* in bytes */
+ u16 lid = be16_to_cpu(hdr->lrh[1]);
+ u32 qp_num;
+ u32 rcv_flags = 0;
+
+ /* Sanity check packet */
+ if (tlen < 24)
+ goto drop;
+
+ /* Check for GRH */
+ if (lnh == HFI1_LRH_BTH)
+ ohdr = &hdr->u.oth;
+ else if (lnh == HFI1_LRH_GRH) {
+ u32 vtf;
+
+ ohdr = &hdr->u.l.oth;
+ if (hdr->u.l.grh.next_hdr != IB_GRH_NEXT_HDR)
+ goto drop;
+ vtf = be32_to_cpu(hdr->u.l.grh.version_tclass_flow);
+ if ((vtf >> IB_GRH_VERSION_SHIFT) != IB_GRH_VERSION)
+ goto drop;
+ rcv_flags |= HFI1_HAS_GRH;
+ } else
+ goto drop;
+
+ /* Get the destination QP number. */
+ qp_num = be32_to_cpu(ohdr->bth[1]) & HFI1_QPN_MASK;
+ if (lid < HFI1_MULTICAST_LID_BASE) {
+ struct hfi1_qp *qp;
+
+ rcu_read_lock();
+ qp = hfi1_lookup_qpn(ibp, qp_num);
+ if (!qp) {
+ rcu_read_unlock();
+ goto drop;
+ }
+
+ /*
+ * Handle only RC QPs - for other QP types drop error
+ * packet.
+ */
+ spin_lock(&qp->r_lock);
+
+ /* Check for valid receive state. */
+ if (!(ib_hfi1_state_ops[qp->state] &
+ HFI1_PROCESS_RECV_OK)) {
+ ibp->n_pkt_drops++;
+ }
+
+ switch (qp->ibqp.qp_type) {
+ case IB_QPT_RC:
+ hfi1_rc_hdrerr(
+ rcd,
+ hdr,
+ rcv_flags,
+ qp);
+ break;
+ default:
+ /* For now don't handle any other QP types */
+ break;
+ }
+
+ spin_unlock(&qp->r_lock);
+ rcu_read_unlock();
+ } /* Unicast QP */
+ } /* Valid packet with TIDErr */
+
+ /* handle "RcvTypeErr" flags */
+ switch (rte) {
+ case RHF_RTE_ERROR_OP_CODE_ERR:
+ {
+ u32 opcode;
+ void *ebuf = NULL;
+ __be32 *bth = NULL;
+
+ if (rhf_use_egr_bfr(packet->rhf))
+ ebuf = packet->ebuf;
+
+ if (ebuf == NULL)
+ goto drop; /* this should never happen */
+
+ if (lnh == HFI1_LRH_BTH)
+ bth = (__be32 *)ebuf;
+ else if (lnh == HFI1_LRH_GRH)
+ bth = (__be32 *)((char *)ebuf + sizeof(struct ib_grh));
+ else
+ goto drop;
+
+ opcode = be32_to_cpu(bth[0]) >> 24;
+ opcode &= 0xff;
+
+ if (opcode == IB_OPCODE_CNP) {
+ /*
+ * Only in pre-B0 h/w is the CNP_OPCODE handled
+ * via this code path (errata 291394).
+ */
+ struct hfi1_qp *qp = NULL;
+ u32 lqpn, rqpn;
+ u16 rlid;
+ u8 svc_type, sl, sc5;
+
+ sc5 = (be16_to_cpu(rhdr->lrh[0]) >> 12) & 0xf;
+ if (rhf_dc_info(packet->rhf))
+ sc5 |= 0x10;
+ sl = ibp->sc_to_sl[sc5];
+
+ lqpn = be32_to_cpu(bth[1]) & HFI1_QPN_MASK;
+ rcu_read_lock();
+ qp = hfi1_lookup_qpn(ibp, lqpn);
+ if (qp == NULL) {
+ rcu_read_unlock();
+ goto drop;
+ }
+
+ switch (qp->ibqp.qp_type) {
+ case IB_QPT_UD:
+ rlid = 0;
+ rqpn = 0;
+ svc_type = IB_CC_SVCTYPE_UD;
+ break;
+ case IB_QPT_UC:
+ rlid = be16_to_cpu(rhdr->lrh[3]);
+ rqpn = qp->remote_qpn;
+ svc_type = IB_CC_SVCTYPE_UC;
+ break;
+ default:
+ goto drop;
+ }
+
+ process_becn(ppd, sl, rlid, lqpn, rqpn, svc_type);
+ rcu_read_unlock();
+ }
+
+ packet->rhf &= ~RHF_RCV_TYPE_ERR_SMASK;
+ break;
+ }
+ default:
+ break;
+ }
+
+drop:
+ return;
+}
+
+static inline void init_packet(struct hfi1_ctxtdata *rcd,
+ struct hfi1_packet *packet)
+{
+
+ packet->rsize = rcd->rcvhdrqentsize; /* words */
+ packet->maxcnt = rcd->rcvhdrq_cnt * packet->rsize; /* words */
+ packet->rcd = rcd;
+ packet->updegr = 0;
+ packet->etail = -1;
+ packet->rhf_addr = (__le32 *) rcd->rcvhdrq + rcd->head +
+ rcd->dd->rhf_offset;
+ packet->rhf = rhf_to_cpu(packet->rhf_addr);
+ packet->rhqoff = rcd->head;
+ packet->numpkt = 0;
+ packet->rcv_flags = 0;
+}
+
+#ifndef CONFIG_PRESCAN_RXQ
+static void prescan_rxq(struct hfi1_packet *packet) {}
+#else /* CONFIG_PRESCAN_RXQ */
+static int prescan_receive_queue;
+
+static void process_ecn(struct hfi1_qp *qp, struct hfi1_ib_header *hdr,
+ struct hfi1_other_headers *ohdr,
+ u64 rhf, struct ib_grh *grh)
+{
+ struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+ u32 bth1;
+ u8 sc5, svc_type;
+ int is_fecn, is_becn;
+
+ switch (qp->ibqp.qp_type) {
+ case IB_QPT_UD:
+ svc_type = IB_CC_SVCTYPE_UD;
+ break;
+ case IB_QPT_UC: /* LATER */
+ case IB_QPT_RC: /* LATER */
+ default:
+ return;
+ }
+
+ is_fecn = (be32_to_cpu(ohdr->bth[1]) >> HFI1_FECN_SHIFT) &
+ HFI1_FECN_MASK;
+ is_becn = (be32_to_cpu(ohdr->bth[1]) >> HFI1_BECN_SHIFT) &
+ HFI1_BECN_MASK;
+
+ sc5 = (be16_to_cpu(hdr->lrh[0]) >> 12) & 0xf;
+ if (rhf_dc_info(rhf))
+ sc5 |= 0x10;
+
+ if (is_fecn) {
+ u32 src_qpn = be32_to_cpu(ohdr->u.ud.deth[1]) & HFI1_QPN_MASK;
+ u16 pkey = (u16)be32_to_cpu(ohdr->bth[0]);
+ u16 dlid = be16_to_cpu(hdr->lrh[1]);
+ u16 slid = be16_to_cpu(hdr->lrh[3]);
+
+ return_cnp(ibp, qp, src_qpn, pkey, dlid, slid, sc5, grh);
+ }
+
+ if (is_becn) {
+ struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
+ u32 lqpn = be32_to_cpu(ohdr->bth[1]) & HFI1_QPN_MASK;
+ u8 sl = ibp->sc_to_sl[sc5];
+
+ process_becn(ppd, sl, 0, lqpn, 0, svc_type);
+ }
+
+ /* turn off BECN, or FECN */
+ bth1 = be32_to_cpu(ohdr->bth[1]);
+ bth1 &= ~(HFI1_FECN_MASK << HFI1_FECN_SHIFT);
+ bth1 &= ~(HFI1_BECN_MASK << HFI1_BECN_SHIFT);
+ ohdr->bth[1] = cpu_to_be32(bth1);
+}
+
+struct ps_mdata {
+ struct hfi1_ctxtdata *rcd;
+ u32 rsize;
+ u32 maxcnt;
+ u32 ps_head;
+ u32 ps_tail;
+ u32 ps_seq;
+};
+
+static inline void init_ps_mdata(struct ps_mdata *mdata,
+ struct hfi1_packet *packet)
+{
+ struct hfi1_ctxtdata *rcd = packet->rcd;
+
+ mdata->rcd = rcd;
+ mdata->rsize = packet->rsize;
+ mdata->maxcnt = packet->maxcnt;
+
+ if (rcd->ps_state.initialized == 0) {
+ mdata->ps_head = packet->rhqoff;
+ rcd->ps_state.initialized++;
+ } else
+ mdata->ps_head = rcd->ps_state.ps_head;
+
+ if (HFI1_CAP_IS_KSET(DMA_RTAIL)) {
+ mdata->ps_tail = packet->hdrqtail;
+ mdata->ps_seq = 0; /* not used with DMA_RTAIL */
+ } else {
+ mdata->ps_tail = 0; /* used only with DMA_RTAIL*/
+ mdata->ps_seq = rcd->seq_cnt;
+ }
+}
+
+static inline int ps_done(struct ps_mdata *mdata, u64 rhf)
+{
+ if (HFI1_CAP_IS_KSET(DMA_RTAIL))
+ return mdata->ps_head == mdata->ps_tail;
+ return mdata->ps_seq != rhf_rcv_seq(rhf);
+}
+
+static inline void update_ps_mdata(struct ps_mdata *mdata)
+{
+ struct hfi1_ctxtdata *rcd = mdata->rcd;
+
+ mdata->ps_head += mdata->rsize;
+ if (mdata->ps_head > mdata->maxcnt)
+ mdata->ps_head = 0;
+ rcd->ps_state.ps_head = mdata->ps_head;
+ if (!HFI1_CAP_IS_KSET(DMA_RTAIL)) {
+ if (++mdata->ps_seq > 13)
+ mdata->ps_seq = 1;
+ }
+}
+
+/*
+ * prescan_rxq - search through the receive queue looking for packets
+ * containing Excplicit Congestion Notifications (FECNs, or BECNs).
+ * When an ECN is found, process the Congestion Notification, and toggle
+ * it off.
+ */
+static void prescan_rxq(struct hfi1_packet *packet)
+{
+ struct hfi1_ctxtdata *rcd = packet->rcd;
+ struct ps_mdata mdata;
+
+ if (!prescan_receive_queue)
+ return;
+
+ init_ps_mdata(&mdata, packet);
+
+ while (1) {
+ struct hfi1_devdata *dd = rcd->dd;
+ struct hfi1_ibport *ibp = &rcd->ppd->ibport_data;
+ __le32 *rhf_addr = (__le32 *) rcd->rcvhdrq + mdata.ps_head +
+ dd->rhf_offset;
+ struct hfi1_qp *qp;
+ struct hfi1_ib_header *hdr;
+ struct hfi1_other_headers *ohdr;
+ struct ib_grh *grh = NULL;
+ u64 rhf = rhf_to_cpu(rhf_addr);
+ u32 etype = rhf_rcv_type(rhf), qpn;
+ int is_ecn = 0;
+ u8 lnh;
+
+ if (ps_done(&mdata, rhf))
+ break;
+
+ if (etype != RHF_RCV_TYPE_IB)
+ goto next;
+
+ hdr = (struct hfi1_ib_header *)
+ hfi1_get_msgheader(dd, rhf_addr);
+ lnh = be16_to_cpu(hdr->lrh[0]) & 3;
+
+ if (lnh == HFI1_LRH_BTH)
+ ohdr = &hdr->u.oth;
+ else if (lnh == HFI1_LRH_GRH) {
+ ohdr = &hdr->u.l.oth;
+ grh = &hdr->u.l.grh;
+ } else
+ goto next; /* just in case */
+
+ is_ecn |= be32_to_cpu(ohdr->bth[1]) &
+ (HFI1_FECN_MASK << HFI1_FECN_SHIFT);
+ is_ecn |= be32_to_cpu(ohdr->bth[1]) &
+ (HFI1_BECN_MASK << HFI1_BECN_SHIFT);
+
+ if (!is_ecn)
+ goto next;
+
+ qpn = be32_to_cpu(ohdr->bth[1]) & HFI1_QPN_MASK;
+ rcu_read_lock();
+ qp = hfi1_lookup_qpn(ibp, qpn);
+
+ if (qp == NULL) {
+ rcu_read_unlock();
+ goto next;
+ }
+
+ process_ecn(qp, hdr, ohdr, rhf, grh);
+ rcu_read_unlock();
+next:
+ update_ps_mdata(&mdata);
+ }
+}
+#endif /* CONFIG_PRESCAN_RXQ */
+
+#define RCV_PKT_OK 0x0
+#define RCV_PKT_MAX 0x1
+
+static inline int process_rcv_packet(struct hfi1_packet *packet)
+{
+ int ret = RCV_PKT_OK;
+
+ packet->hdr = hfi1_get_msgheader(packet->rcd->dd,
+ packet->rhf_addr);
+ packet->hlen = (u8 *)packet->rhf_addr - (u8 *)packet->hdr;
+ packet->etype = rhf_rcv_type(packet->rhf);
+ /* total length */
+ packet->tlen = rhf_pkt_len(packet->rhf); /* in bytes */
+ /* retrieve eager buffer details */
+ packet->ebuf = NULL;
+ if (rhf_use_egr_bfr(packet->rhf)) {
+ packet->etail = rhf_egr_index(packet->rhf);
+ packet->ebuf = get_egrbuf(packet->rcd, packet->rhf,
+ &packet->updegr);
+ /*
+ * Prefetch the contents of the eager buffer. It is
+ * OK to send a negative length to prefetch_range().
+ * The +2 is the size of the RHF.
+ */
+ prefetch_range(packet->ebuf,
+ packet->tlen - ((packet->rcd->rcvhdrqentsize -
+ (rhf_hdrq_offset(packet->rhf)+2)) * 4));
+ }
+
+ /*
+ * Call a type specific handler for the packet. We
+ * should be able to trust that etype won't be beyond
+ * the range of valid indexes. If so something is really
+ * wrong and we can probably just let things come
+ * crashing down. There is no need to eat another
+ * comparison in this performance critical code.
+ */
+ packet->rcd->dd->rhf_rcv_function_map[packet->etype](packet);
+ packet->numpkt++;
+
+ /* Set up for the next packet */
+ packet->rhqoff += packet->rsize;
+ if (packet->rhqoff >= packet->maxcnt)
+ packet->rhqoff = 0;
+
+ if (packet->numpkt == MAX_PKT_RECV) {
+ ret = RCV_PKT_MAX;
+ this_cpu_inc(*packet->rcd->dd->rcv_limit);
+ }
+
+ packet->rhf_addr = (__le32 *) packet->rcd->rcvhdrq + packet->rhqoff +
+ packet->rcd->dd->rhf_offset;
+ packet->rhf = rhf_to_cpu(packet->rhf_addr);
+
+ return ret;
+}
+
+static inline void process_rcv_update(int last, struct hfi1_packet *packet)
+{
+ /*
+ * Update head regs etc., every 16 packets, if not last pkt,
+ * to help prevent rcvhdrq overflows, when many packets
+ * are processed and queue is nearly full.
+ * Don't request an interrupt for intermediate updates.
+ */
+ if (!last && !(packet->numpkt & 0xf)) {
+ update_usrhead(packet->rcd, packet->rhqoff, packet->updegr,
+ packet->etail, 0, 0);
+ packet->updegr = 0;
+ }
+ packet->rcv_flags = 0;
+}
+
+static inline void finish_packet(struct hfi1_packet *packet)
+{
+
+ /*
+ * Nothing we need to free for the packet.
+ *
+ * The only thing we need to do is a final update and call for an
+ * interrupt
+ */
+ update_usrhead(packet->rcd, packet->rcd->head, packet->updegr,
+ packet->etail, rcv_intr_dynamic, packet->numpkt);
+
+}
+
+static inline void process_rcv_qp_work(struct hfi1_packet *packet)
+{
+
+ struct hfi1_ctxtdata *rcd;
+ struct hfi1_qp *qp, *nqp;
+
+ rcd = packet->rcd;
+ rcd->head = packet->rhqoff;
+
+ /*
+ * Iterate over all QPs waiting to respond.
+ * The list won't change since the IRQ is only run on one CPU.
+ */
+ list_for_each_entry_safe(qp, nqp, &rcd->qp_wait_list, rspwait) {
+ list_del_init(&qp->rspwait);
+ if (qp->r_flags & HFI1_R_RSP_NAK) {
+ qp->r_flags &= ~HFI1_R_RSP_NAK;
+ hfi1_send_rc_ack(rcd, qp, 0);
+ }
+ if (qp->r_flags & HFI1_R_RSP_SEND) {
+ unsigned long flags;
+
+ qp->r_flags &= ~HFI1_R_RSP_SEND;
+ spin_lock_irqsave(&qp->s_lock, flags);
+ if (ib_hfi1_state_ops[qp->state] &
+ HFI1_PROCESS_OR_FLUSH_SEND)
+ hfi1_schedule_send(qp);
+ spin_unlock_irqrestore(&qp->s_lock, flags);
+ }
+ if (atomic_dec_and_test(&qp->refcount))
+ wake_up(&qp->wait);
+ }
+}
+
+/*
+ * Handle receive interrupts when using the no dma rtail option.
+ */
+void handle_receive_interrupt_nodma_rtail(struct hfi1_ctxtdata *rcd)
+{
+ u32 seq;
+ int last = 0;
+ struct hfi1_packet packet;
+
+ init_packet(rcd, &packet);
+ seq = rhf_rcv_seq(packet.rhf);
+ if (seq != rcd->seq_cnt)
+ goto bail;
+
+ prescan_rxq(&packet);
+
+ while (!last) {
+ last = process_rcv_packet(&packet);
+ seq = rhf_rcv_seq(packet.rhf);
+ if (++rcd->seq_cnt > 13)
+ rcd->seq_cnt = 1;
+ if (seq != rcd->seq_cnt)
+ last = 1;
+ process_rcv_update(last, &packet);
+ }
+ process_rcv_qp_work(&packet);
+bail:
+ finish_packet(&packet);
+}
+
+void handle_receive_interrupt_dma_rtail(struct hfi1_ctxtdata *rcd)
+{
+ u32 hdrqtail;
+ int last = 0;
+ struct hfi1_packet packet;
+
+ init_packet(rcd, &packet);
+ hdrqtail = get_rcvhdrtail(rcd);
+ if (packet.rhqoff == hdrqtail)
+ goto bail;
+ smp_rmb(); /* prevent speculative reads of dma'ed hdrq */
+
+ prescan_rxq(&packet);
+
+ while (!last) {
+ last = process_rcv_packet(&packet);
+ if (packet.rhqoff == hdrqtail)
+ last = 1;
+ process_rcv_update(last, &packet);
+ }
+ process_rcv_qp_work(&packet);
+bail:
+ finish_packet(&packet);
+
+}
+
+static inline void set_all_nodma_rtail(struct hfi1_devdata *dd)
+{
+ int i;
+
+ for (i = 0; i < dd->first_user_ctxt; i++)
+ dd->rcd[i]->do_interrupt =
+ &handle_receive_interrupt_nodma_rtail;
+}
+
+static inline void set_all_dma_rtail(struct hfi1_devdata *dd)
+{
+ int i;
+
+ for (i = 0; i < dd->first_user_ctxt; i++)
+ dd->rcd[i]->do_interrupt =
+ &handle_receive_interrupt_dma_rtail;
+}
+
+/*
+ * handle_receive_interrupt - receive a packet
+ * @rcd: the context
+ *
+ * Called from interrupt handler for errors or receive interrupt.
+ * This is the slow path interrupt handler.
+ */
+void handle_receive_interrupt(struct hfi1_ctxtdata *rcd)
+{
+
+ struct hfi1_devdata *dd = rcd->dd;
+ u32 hdrqtail;
+ int last = 0, needset = 1;
+ struct hfi1_packet packet;
+
+ init_packet(rcd, &packet);
+
+ if (!HFI1_CAP_IS_KSET(DMA_RTAIL)) {
+ u32 seq = rhf_rcv_seq(packet.rhf);
+
+ if (seq != rcd->seq_cnt)
+ goto bail;
+ hdrqtail = 0;
+ } else {
+ hdrqtail = get_rcvhdrtail(rcd);
+ if (packet.rhqoff == hdrqtail)
+ goto bail;
+ smp_rmb(); /* prevent speculative reads of dma'ed hdrq */
+ }
+
+ prescan_rxq(&packet);
+
+ while (!last) {
+
+ if (unlikely(dd->do_drop && atomic_xchg(&dd->drop_packet,
+ DROP_PACKET_OFF) == DROP_PACKET_ON)) {
+ dd->do_drop = 0;
+
+ /* On to the next packet */
+ packet.rhqoff += packet.rsize;
+ packet.rhf_addr = (__le32 *) rcd->rcvhdrq +
+ packet.rhqoff +
+ dd->rhf_offset;
+ packet.rhf = rhf_to_cpu(packet.rhf_addr);
+
+ } else {
+ last = process_rcv_packet(&packet);
+ }
+
+ if (!HFI1_CAP_IS_KSET(DMA_RTAIL)) {
+ u32 seq = rhf_rcv_seq(packet.rhf);
+
+ if (++rcd->seq_cnt > 13)
+ rcd->seq_cnt = 1;
+ if (seq != rcd->seq_cnt)
+ last = 1;
+ if (needset) {
+ dd_dev_info(dd,
+ "Switching to NO_DMA_RTAIL\n");
+ set_all_nodma_rtail(dd);
+ needset = 0;
+ }
+ } else {
+ if (packet.rhqoff == hdrqtail)
+ last = 1;
+ if (needset) {
+ dd_dev_info(dd,
+ "Switching to DMA_RTAIL\n");
+ set_all_dma_rtail(dd);
+ needset = 0;
+ }
+ }
+
+ process_rcv_update(last, &packet);
+ }
+
+ process_rcv_qp_work(&packet);
+
+bail:
+ /*
+ * Always write head at end, and setup rcv interrupt, even
+ * if no packets were processed.
+ */
+ finish_packet(&packet);
+}
+
+/*
+ * Convert a given MTU size to the on-wire MAD packet enumeration.
+ * Return -1 if the size is invalid.
+ */
+int mtu_to_enum(u32 mtu, int default_if_bad)
+{
+ switch (mtu) {
+ case 0: return OPA_MTU_0;
+ case 256: return OPA_MTU_256;
+ case 512: return OPA_MTU_512;
+ case 1024: return OPA_MTU_1024;
+ case 2048: return OPA_MTU_2048;
+ case 4096: return OPA_MTU_4096;
+ case 8192: return OPA_MTU_8192;
+ case 10240: return OPA_MTU_10240;
+ }
+ return default_if_bad;
+}
+
+u16 enum_to_mtu(int mtu)
+{
+ switch (mtu) {
+ case OPA_MTU_0: return 0;
+ case OPA_MTU_256: return 256;
+ case OPA_MTU_512: return 512;
+ case OPA_MTU_1024: return 1024;
+ case OPA_MTU_2048: return 2048;
+ case OPA_MTU_4096: return 4096;
+ case OPA_MTU_8192: return 8192;
+ case OPA_MTU_10240: return 10240;
+ default: return 0xffff;
+ }
+}
+
+/*
+ * set_mtu - set the MTU
+ * @ppd: the per port data
+ *
+ * We can handle "any" incoming size, the issue here is whether we
+ * need to restrict our outgoing size. We do not deal with what happens
+ * to programs that are already running when the size changes.
+ */
+int set_mtu(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ int i, drain, ret = 0, is_up = 0;
+
+ ppd->ibmtu = 0;
+ for (i = 0; i < ppd->vls_supported; i++)
+ if (ppd->ibmtu < dd->vld[i].mtu)
+ ppd->ibmtu = dd->vld[i].mtu;
+ ppd->ibmaxlen = ppd->ibmtu + lrh_max_header_bytes(ppd->dd);
+
+ mutex_lock(&ppd->hls_lock);
+ if (ppd->host_link_state == HLS_UP_INIT
+ || ppd->host_link_state == HLS_UP_ARMED
+ || ppd->host_link_state == HLS_UP_ACTIVE)
+ is_up = 1;
+
+ drain = !is_ax(dd) && is_up;
+
+ if (drain)
+ /*
+ * MTU is specified per-VL. To ensure that no packet gets
+ * stuck (due, e.g., to the MTU for the packet's VL being
+ * reduced), empty the per-VL FIFOs before adjusting MTU.
+ */
+ ret = stop_drain_data_vls(dd);
+
+ if (ret) {
+ dd_dev_err(dd, "%s: cannot stop/drain VLs - refusing to change per-VL MTUs\n",
+ __func__);
+ goto err;
+ }
+
+ hfi1_set_ib_cfg(ppd, HFI1_IB_CFG_MTU, 0);
+
+ if (drain)
+ open_fill_data_vls(dd); /* reopen all VLs */
+
+err:
+ mutex_unlock(&ppd->hls_lock);
+
+ return ret;
+}
+
+int hfi1_set_lid(struct hfi1_pportdata *ppd, u32 lid, u8 lmc)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+
+ ppd->lid = lid;
+ ppd->lmc = lmc;
+ hfi1_set_ib_cfg(ppd, HFI1_IB_CFG_LIDLMC, 0);
+
+ dd_dev_info(dd, "IB%u:%u got a lid: 0x%x\n", dd->unit, ppd->port, lid);
+
+ return 0;
+}
+
+/*
+ * Following deal with the "obviously simple" task of overriding the state
+ * of the LEDs, which normally indicate link physical and logical status.
+ * The complications arise in dealing with different hardware mappings
+ * and the board-dependent routine being called from interrupts.
+ * and then there's the requirement to _flash_ them.
+ */
+#define LED_OVER_FREQ_SHIFT 8
+#define LED_OVER_FREQ_MASK (0xFF<<LED_OVER_FREQ_SHIFT)
+/* Below is "non-zero" to force override, but both actual LEDs are off */
+#define LED_OVER_BOTH_OFF (8)
+
+static void run_led_override(unsigned long opaque)
+{
+ struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)opaque;
+ struct hfi1_devdata *dd = ppd->dd;
+ int timeoff;
+ int ph_idx;
+
+ if (!(dd->flags & HFI1_INITTED))
+ return;
+
+ ph_idx = ppd->led_override_phase++ & 1;
+ ppd->led_override = ppd->led_override_vals[ph_idx];
+ timeoff = ppd->led_override_timeoff;
+
+ /*
+ * don't re-fire the timer if user asked for it to be off; we let
+ * it fire one more time after they turn it off to simplify
+ */
+ if (ppd->led_override_vals[0] || ppd->led_override_vals[1])
+ mod_timer(&ppd->led_override_timer, jiffies + timeoff);
+}
+
+void hfi1_set_led_override(struct hfi1_pportdata *ppd, unsigned int val)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ int timeoff, freq;
+
+ if (!(dd->flags & HFI1_INITTED))
+ return;
+
+ /* First check if we are blinking. If not, use 1HZ polling */
+ timeoff = HZ;
+ freq = (val & LED_OVER_FREQ_MASK) >> LED_OVER_FREQ_SHIFT;
+
+ if (freq) {
+ /* For blink, set each phase from one nybble of val */
+ ppd->led_override_vals[0] = val & 0xF;
+ ppd->led_override_vals[1] = (val >> 4) & 0xF;
+ timeoff = (HZ << 4)/freq;
+ } else {
+ /* Non-blink set both phases the same. */
+ ppd->led_override_vals[0] = val & 0xF;
+ ppd->led_override_vals[1] = val & 0xF;
+ }
+ ppd->led_override_timeoff = timeoff;
+
+ /*
+ * If the timer has not already been started, do so. Use a "quick"
+ * timeout so the function will be called soon, to look at our request.
+ */
+ if (atomic_inc_return(&ppd->led_override_timer_active) == 1) {
+ /* Need to start timer */
+ init_timer(&ppd->led_override_timer);
+ ppd->led_override_timer.function = run_led_override;
+ ppd->led_override_timer.data = (unsigned long) ppd;
+ ppd->led_override_timer.expires = jiffies + 1;
+ add_timer(&ppd->led_override_timer);
+ } else {
+ if (ppd->led_override_vals[0] || ppd->led_override_vals[1])
+ mod_timer(&ppd->led_override_timer, jiffies + 1);
+ atomic_dec(&ppd->led_override_timer_active);
+ }
+}
+
+/**
+ * hfi1_reset_device - reset the chip if possible
+ * @unit: the device to reset
+ *
+ * Whether or not reset is successful, we attempt to re-initialize the chip
+ * (that is, much like a driver unload/reload). We clear the INITTED flag
+ * so that the various entry points will fail until we reinitialize. For
+ * now, we only allow this if no user contexts are open that use chip resources
+ */
+int hfi1_reset_device(int unit)
+{
+ int ret, i;
+ struct hfi1_devdata *dd = hfi1_lookup(unit);
+ struct hfi1_pportdata *ppd;
+ unsigned long flags;
+ int pidx;
+
+ if (!dd) {
+ ret = -ENODEV;
+ goto bail;
+ }
+
+ dd_dev_info(dd, "Reset on unit %u requested\n", unit);
+
+ if (!dd->kregbase || !(dd->flags & HFI1_PRESENT)) {
+ dd_dev_info(dd,
+ "Invalid unit number %u or not initialized or not present\n",
+ unit);
+ ret = -ENXIO;
+ goto bail;
+ }
+
+ spin_lock_irqsave(&dd->uctxt_lock, flags);
+ if (dd->rcd)
+ for (i = dd->first_user_ctxt; i < dd->num_rcv_contexts; i++) {
+ if (!dd->rcd[i] || !dd->rcd[i]->cnt)
+ continue;
+ spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+ ret = -EBUSY;
+ goto bail;
+ }
+ spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+ if (atomic_read(&ppd->led_override_timer_active)) {
+ /* Need to stop LED timer, _then_ shut off LEDs */
+ del_timer_sync(&ppd->led_override_timer);
+ atomic_set(&ppd->led_override_timer_active, 0);
+ }
+
+ /* Shut off LEDs after we are sure timer is not running */
+ ppd->led_override = LED_OVER_BOTH_OFF;
+ }
+ if (dd->flags & HFI1_HAS_SEND_DMA)
+ sdma_exit(dd);
+
+ hfi1_reset_cpu_counters(dd);
+
+ ret = hfi1_init(dd, 1);
+
+ if (ret)
+ dd_dev_err(dd,
+ "Reinitialize unit %u after reset failed with %d\n",
+ unit, ret);
+ else
+ dd_dev_info(dd, "Reinitialized unit %u after resetting\n",
+ unit);
+
+bail:
+ return ret;
+}
+
+void handle_eflags(struct hfi1_packet *packet)
+{
+ struct hfi1_ctxtdata *rcd = packet->rcd;
+ u32 rte = rhf_rcv_type_err(packet-