Merge tag 'mlx5-updates-2019-06-13' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2019-06-13

Mlx5 devlink health fw reporters and sw reset support

This series provides mlx5 firmware reset support and firmware devlink health
reporters.

1) Add initial mlx5 kernel documentation and include devlink health reporters

2) Add CR-Space access and FW Crdump snapshot support via devlink region_snapshot

3) Issue software reset upon FW asserts

4) Add fw and fw_fatal devlink heath reporters to follow fw errors indication by
dump and recover procedures and enable trigger these functionality by user.

4.1) fw reporter:
The fw reporter implements diagnose and dump callbacks.
It follows symptoms of fw error such as fw syndrome by triggering
fw core dump and storing it and any other fw trace into the dump buffer.
The fw reporter diagnose command can be triggered any time by the user to check
current fw status.

4.2) fw_fatal repoter:
The fw_fatal reporter implements dump and recover callbacks.
It follows fatal errors indications by CR-space dump and recover flow.
The CR-space dump uses vsc interface which is valid even if the FW command
interface is not functional, which is the case in most FW fatal errors. The
CR-space dump is stored as a memory region snapshot to ease read by address.
The recover function runs recover flow which reloads the driver and triggers fw
reset if needed.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/networking/device_drivers/index.rst

@@ -21,6 +21,7 @@ Contents:
    intel/i40e
    intel/iavf
    intel/ice
+   mellanox/mlx5
 
 .. only::  subproject
 

Documentation/networking/device_drivers/mellanox/mlx5.rst

+.. SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+
+=================================================
+Mellanox ConnectX(R) mlx5 core VPI Network Driver
+=================================================
+
+Copyright (c) 2019, Mellanox Technologies LTD.
+
+Contents
+========
+
+- `Enabling the driver and kconfig options`_
+- `Devlink health reporters`_
+
+Enabling the driver and kconfig options
+================================================
+
+| mlx5 core is modular and most of the major mlx5 core driver features can be selected (compiled in/out)
+| at build time via kernel Kconfig flags.
+| Basic features, ethernet net device rx/tx offloads and XDP, are available with the most basic flags
+| CONFIG_MLX5_CORE=y/m and CONFIG_MLX5_CORE_EN=y.
+| For the list of advanced features please see below.
+
+**CONFIG_MLX5_CORE=(y/m/n)** (module mlx5_core.ko)
+
+|    The driver can be enabled by choosing CONFIG_MLX5_CORE=y/m in kernel config.
+|    This will provide mlx5 core driver for mlx5 ulps to interface with (mlx5e, mlx5_ib).
+
+
+**CONFIG_MLX5_CORE_EN=(y/n)**
+
+|    Choosing this option will allow basic ethernet netdevice support with all of the standard rx/tx offloads.
+|    mlx5e is the mlx5 ulp driver which provides netdevice kernel interface, when chosen, mlx5e will be
+|    built-in into mlx5_core.ko.
+
+
+**CONFIG_MLX5_EN_ARFS=(y/n)**
+
+|     Enables Hardware-accelerated receive flow steering (arfs) support, and ntuple filtering.
+|     https://community.mellanox.com/s/article/howto-configure-arfs-on-connectx-4
+
+
+**CONFIG_MLX5_EN_RXNFC=(y/n)**
+
+|    Enables ethtool receive network flow classification, which allows user defined
+|    flow rules to direct traffic into arbitrary rx queue via ethtool set/get_rxnfc API.
+
+
+**CONFIG_MLX5_CORE_EN_DCB=(y/n)**:
+
+|    Enables `Data Center Bridging (DCB) Support <https://community.mellanox.com/s/article/howto-auto-config-pfc-and-ets-on-connectx-4-via-lldp-dcbx>`_.
+
+
+**CONFIG_MLX5_MPFS=(y/n)**
+
+|    Ethernet Multi-Physical Function Switch (MPFS) support in ConnectX NIC.
+|    MPFs is required for when `Multi-Host <http://www.mellanox.com/page/multihost>`_ configuration is enabled to allow passing
+|    user configured unicast MAC addresses to the requesting PF.
+
+
+**CONFIG_MLX5_ESWITCH=(y/n)**
+
+|    Ethernet SRIOV E-Switch support in ConnectX NIC. E-Switch provides internal SRIOV packet steering
+|    and switching for the enabled VFs and PF in two available modes:
+|           1) `Legacy SRIOV mode (L2 mac vlan steering based) <https://community.mellanox.com/s/article/howto-configure-sr-iov-for-connectx-4-connectx-5-with-kvm--ethernet-x>`_.
+|           2) `Switchdev mode (eswitch offloads) <https://www.mellanox.com/related-docs/prod_software/ASAP2_Hardware_Offloading_for_vSwitches_User_Manual_v4.4.pdf>`_.
+
+
+**CONFIG_MLX5_CORE_IPOIB=(y/n)**
+
+|    IPoIB offloads & acceleration support.
+|    Requires CONFIG_MLX5_CORE_EN to provide an accelerated interface for the rdma
+|    IPoIB ulp netdevice.
+
+
+**CONFIG_MLX5_FPGA=(y/n)**
+
+|    Build support for the Innova family of network cards by Mellanox Technologies.
+|    Innova network cards are comprised of a ConnectX chip and an FPGA chip on one board.
+|    If you select this option, the mlx5_core driver will include the Innova FPGA core and allow
+|    building sandbox-specific client drivers.
+
+
+**CONFIG_MLX5_EN_IPSEC=(y/n)**
+
+|    Enables `IPSec XFRM cryptography-offload accelaration <http://www.mellanox.com/related-docs/prod_software/Mellanox_Innova_IPsec_Ethernet_Adapter_Card_User_Manual.pdf>`_.
+
+**CONFIG_MLX5_EN_TLS=(y/n)**
+
+|   TLS cryptography-offload accelaration.
+
+
+**CONFIG_MLX5_INFINIBAND=(y/n/m)** (module mlx5_ib.ko)
+
+|   Provides low-level InfiniBand/RDMA and `RoCE <https://community.mellanox.com/s/article/recommended-network-configuration-examples-for-roce-deployment>`_ support.
+
+
+**External options** ( Choose if the corresponding mlx5 feature is required )
+
+- CONFIG_PTP_1588_CLOCK: When chosen, mlx5 ptp support will be enabled
+- CONFIG_VXLAN: When chosen, mlx5 vxaln support will be enabled.
+- CONFIG_MLXFW: When chosen, mlx5 firmware flashing support will be enabled (via devlink and ethtool).
+
+
+Devlink health reporters
+========================
+
+tx reporter
+-----------
+The tx reporter is responsible of two error scenarios:
+
+- TX timeout
+    Report on kernel tx timeout detection.
+    Recover by searching lost interrupts.
+- TX error completion
+    Report on error tx completion.
+    Recover by flushing the TX queue and reset it.
+
+TX reporter also support Diagnose callback, on which it provides
+real time information of its send queues status.
+
+User commands examples:
+
+- Diagnose send queues status::
+
+    $ devlink health diagnose pci/0000:82:00.0 reporter tx
+
+- Show number of tx errors indicated, number of recover flows ended successfully,
+  is autorecover enabled and graceful period from last recover::
+
+    $ devlink health show pci/0000:82:00.0 reporter tx
+
+fw reporter
+-----------
+The fw reporter implements diagnose and dump callbacks.
+It follows symptoms of fw error such as fw syndrome by triggering
+fw core dump and storing it into the dump buffer.
+The fw reporter diagnose command can be triggered any time by the user to check
+current fw status.
+
+User commands examples:
+
+- Check fw heath status::
+
+    $ devlink health diagnose pci/0000:82:00.0 reporter fw
+
+- Read FW core dump if already stored or trigger new one::
+
+    $ devlink health dump show pci/0000:82:00.0 reporter fw
+
+NOTE: This command can run only on the PF which has fw tracer ownership,
+running it on other PF or any VF will return "Operation not permitted".
+
+fw fatal reporter
+-----------------
+The fw fatal reporter implements dump and recover callbacks.
+It follows fatal errors indications by CR-space dump and recover flow.
+The CR-space dump uses vsc interface which is valid even if the FW command
+interface is not functional, which is the case in most FW fatal errors.
+The recover function runs recover flow which reloads the driver and triggers fw
+reset if needed.
+
+User commands examples:
+
+- Run fw recover flow manually::
+
+    $ devlink health recover pci/0000:82:00.0 reporter fw_fatal
+
+- Read FW CR-space dump if already strored or trigger new one::
+
+    $ devlink health dump show pci/0000:82:00.1 reporter fw_fatal
+
+NOTE: This command can run only on PF.

MAINTAINERS

@@ -10108,6 +10108,7 @@ Q:	http://patchwork.ozlabs.org/project/netdev/list/
 S:	Supported
 F:	drivers/net/ethernet/mellanox/mlx5/core/
 F:	include/linux/mlx5/
+F:	Documentation/networking/device_drivers/mellanox/
 
 MELLANOX MLX5 IB driver
 M:	Leon Romanovsky <leonro@mellanox.com>

drivers/net/ethernet/mellanox/mlx5/core/Makefile

@@ -15,7 +15,8 @@ mlx5_core-y :=	main.o cmd.o debugfs.o fw.o eq.o uar.o pagealloc.o \
 		health.o mcg.o cq.o alloc.o qp.o port.o mr.o pd.o \
 		transobj.o vport.o sriov.o fs_cmd.o fs_core.o \
 		fs_counters.o rl.o lag.o dev.o events.o wq.o lib/gid.o \
-		lib/devcom.o diag/fs_tracepoint.o diag/fw_tracer.o
+		lib/devcom.o lib/pci_vsc.o diag/fs_tracepoint.o \
+		diag/fw_tracer.o diag/crdump.o devlink.o
 
 #
 # Netdev basic

drivers/net/ethernet/mellanox/mlx5/core/devlink.c

+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/* Copyright (c) 2019 Mellanox Technologies */
+
+#include <devlink.h>
+
+#include "mlx5_core.h"
+#include "eswitch.h"
+
+static int mlx5_devlink_flash_update(struct devlink *devlink,
+				     const char *file_name,
+				     const char *component,
+				     struct netlink_ext_ack *extack)
+{
+	struct mlx5_core_dev *dev = devlink_priv(devlink);
+	const struct firmware *fw;
+	int err;
+
+	if (component)
+		return -EOPNOTSUPP;
+
+	err = request_firmware_direct(&fw, file_name, &dev->pdev->dev);
+	if (err)
+		return err;
+
+	return mlx5_firmware_flash(dev, fw, extack);
+}
+
+static const struct devlink_ops mlx5_devlink_ops = {
+#ifdef CONFIG_MLX5_ESWITCH
+	.eswitch_mode_set = mlx5_devlink_eswitch_mode_set,
+	.eswitch_mode_get = mlx5_devlink_eswitch_mode_get,
+	.eswitch_inline_mode_set = mlx5_devlink_eswitch_inline_mode_set,
+	.eswitch_inline_mode_get = mlx5_devlink_eswitch_inline_mode_get,
+	.eswitch_encap_mode_set = mlx5_devlink_eswitch_encap_mode_set,
+	.eswitch_encap_mode_get = mlx5_devlink_eswitch_encap_mode_get,
+#endif
+	.flash_update = mlx5_devlink_flash_update,
+};
+
+struct devlink *mlx5_devlink_alloc()
+{
+	return devlink_alloc(&mlx5_devlink_ops, sizeof(struct mlx5_core_dev));
+}
+
+void mlx5_devlink_free(struct devlink *devlink)
+{
+	devlink_free(devlink);
+}
+
+int mlx5_devlink_register(struct devlink *devlink, struct device *dev)
+{
+	return devlink_register(devlink, dev);
+}
+
+void mlx5_devlink_unregister(struct devlink *devlink)
+{
+	devlink_unregister(devlink);
+}

drivers/net/ethernet/mellanox/mlx5/core/devlink.h

+/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
+/* Copyright (c) 2019, Mellanox Technologies */
+
+#ifndef __MLX5_DEVLINK_H__
+#define __MLX5_DEVLINK_H__
+
+#include <net/devlink.h>
+
+struct devlink *mlx5_devlink_alloc(void);
+void mlx5_devlink_free(struct devlink *devlink);
+int mlx5_devlink_register(struct devlink *devlink, struct device *dev);
+void mlx5_devlink_unregister(struct devlink *devlink);
+
+#endif /* __MLX5_DEVLINK_H__ */

drivers/net/ethernet/mellanox/mlx5/core/diag/crdump.c

+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/* Copyright (c) 2019 Mellanox Technologies */
+
+#include <linux/mlx5/driver.h>
+#include "mlx5_core.h"
+#include "lib/pci_vsc.h"
+#include "lib/mlx5.h"
+
+#define BAD_ACCESS			0xBADACCE5
+#define MLX5_PROTECTED_CR_SCAN_CRSPACE	0x7
+
+static bool mlx5_crdump_enabled(struct mlx5_core_dev *dev)
+{
+	return !!dev->priv.health.crdump_size;
+}
+
+static int mlx5_crdump_fill(struct mlx5_core_dev *dev, u32 *cr_data)
+{
+	u32 crdump_size = dev->priv.health.crdump_size;
+	int i, ret;
+
+	for (i = 0; i < (crdump_size / 4); i++)
+		cr_data[i] = BAD_ACCESS;
+
+	ret = mlx5_vsc_gw_read_block_fast(dev, cr_data, crdump_size);
+	if (ret <= 0) {
+		if (ret == 0)
+			return -EIO;
+		return ret;
+	}
+
+	if (crdump_size != ret) {
+		mlx5_core_warn(dev, "failed to read full dump, read %d out of %u\n",
+			       ret, crdump_size);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+int mlx5_crdump_collect(struct mlx5_core_dev *dev, u32 *cr_data)
+{
+	int ret;
+
+	if (!mlx5_crdump_enabled(dev))
+		return -ENODEV;
+
+	ret = mlx5_vsc_gw_lock(dev);
+	if (ret) {
+		mlx5_core_warn(dev, "crdump: failed to lock vsc gw err %d\n",
+			       ret);
+		return ret;
+	}
+	/* Verify no other PF is running cr-dump or sw reset */
+	ret = mlx5_vsc_sem_set_space(dev, MLX5_SEMAPHORE_SW_RESET,
+				     MLX5_VSC_LOCK);
+	if (ret) {
+		mlx5_core_warn(dev, "Failed to lock SW reset semaphore\n");
+		goto unlock_gw;
+	}
+
+	ret = mlx5_vsc_gw_set_space(dev, MLX5_VSC_SPACE_SCAN_CRSPACE, NULL);
+	if (ret)
+		goto unlock_sem;
+
+	ret = mlx5_crdump_fill(dev, cr_data);
+
+unlock_sem:
+	mlx5_vsc_sem_set_space(dev, MLX5_SEMAPHORE_SW_RESET, MLX5_VSC_UNLOCK);
+unlock_gw:
+	mlx5_vsc_gw_unlock(dev);
+	return ret;
+}
+
+int mlx5_crdump_enable(struct mlx5_core_dev *dev)
+{
+	struct mlx5_priv *priv = &dev->priv;
+	u32 space_size;
+	int ret;
+
+	if (!mlx5_core_is_pf(dev) || !mlx5_vsc_accessible(dev) ||
+	    mlx5_crdump_enabled(dev))
+		return 0;
+
+	ret = mlx5_vsc_gw_lock(dev);
+	if (ret)
+		return ret;
+
+	/* Check if space is supported and get space size */
+	ret = mlx5_vsc_gw_set_space(dev, MLX5_VSC_SPACE_SCAN_CRSPACE,
+				    &space_size);
+	if (ret) {
+		/* Unlock and mask error since space is not supported */
+		mlx5_vsc_gw_unlock(dev);
+		return 0;
+	}
+
+	if (!space_size) {
+		mlx5_core_warn(dev, "Invalid Crspace size, zero\n");
+		mlx5_vsc_gw_unlock(dev);
+		return -EINVAL;
+	}
+
+	ret = mlx5_vsc_gw_unlock(dev);
+	if (ret)
+		return ret;
+
+	priv->health.crdump_size = space_size;
+	return 0;
+}
+
+void mlx5_crdump_disable(struct mlx5_core_dev *dev)
+{
+	dev->priv.health.crdump_size = 0;
+}

drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.c

@@ -243,6 +243,19 @@ static int mlx5_fw_tracer_allocate_strings_db(struct mlx5_fw_tracer *tracer)
 	return -ENOMEM;
 }
 
+static void
+mlx5_fw_tracer_init_saved_traces_array(struct mlx5_fw_tracer *tracer)
+{
+	tracer->st_arr.saved_traces_index = 0;
+	mutex_init(&tracer->st_arr.lock);
+}
+
+static void
+mlx5_fw_tracer_clean_saved_traces_array(struct mlx5_fw_tracer *tracer)
+{
+	mutex_destroy(&tracer->st_arr.lock);
+}
+
 static void mlx5_tracer_read_strings_db(struct work_struct *work)
 {
 	struct mlx5_fw_tracer *tracer = container_of(work, struct mlx5_fw_tracer,
@@ -522,6 +535,24 @@ static void mlx5_fw_tracer_clean_ready_list(struct mlx5_fw_tracer *tracer)
 		list_del(&str_frmt->list);
 }
 
+static void mlx5_fw_tracer_save_trace(struct mlx5_fw_tracer *tracer,
+				      u64 timestamp, bool lost,
+				      u8 event_id, char *msg)
+{
+	struct mlx5_fw_trace_data *trace_data;
+
+	mutex_lock(&tracer->st_arr.lock);
+	trace_data = &tracer->st_arr.straces[tracer->st_arr.saved_traces_index];
+	trace_data->timestamp = timestamp;
+	trace_data->lost = lost;
+	trace_data->event_id = event_id;
+	strncpy(trace_data->msg, msg, TRACE_STR_MSG);
+
+	tracer->st_arr.saved_traces_index =
+		(tracer->st_arr.saved_traces_index + 1) & (SAVED_TRACES_NUM - 1);
+	mutex_unlock(&tracer->st_arr.lock);
+}
+
 static void mlx5_tracer_print_trace(struct tracer_string_format *str_frmt,
 				    struct mlx5_core_dev *dev,
 				    u64 trace_timestamp)
@@ -540,6 +571,9 @@ static void mlx5_tracer_print_trace(struct tracer_string_format *str_frmt,
 	trace_mlx5_fw(dev->tracer, trace_timestamp, str_frmt->lost,
 		      str_frmt->event_id, tmp);
 
+	mlx5_fw_tracer_save_trace(dev->tracer, trace_timestamp,
+				  str_frmt->lost, str_frmt->event_id, tmp);
+
 	/* remove it from hash */
 	mlx5_tracer_clean_message(str_frmt);
 }
@@ -786,6 +820,109 @@ static void mlx5_fw_tracer_ownership_change(struct work_struct *work)
 	mlx5_fw_tracer_start(tracer);
 }
 
+static int mlx5_fw_tracer_set_core_dump_reg(struct mlx5_core_dev *dev,
+					    u32 *in, int size_in)
+{
+	u32 out[MLX5_ST_SZ_DW(core_dump_reg)] = {};
+
+	if (!MLX5_CAP_DEBUG(dev, core_dump_general) &&
+	    !MLX5_CAP_DEBUG(dev, core_dump_qp))
+		return -EOPNOTSUPP;
+
+	return mlx5_core_access_reg(dev, in, size_in, out, sizeof(out),
+				    MLX5_REG_CORE_DUMP, 0, 1);
+}
+
+int mlx5_fw_tracer_trigger_core_dump_general(struct mlx5_core_dev *dev)
+{
+	struct mlx5_fw_tracer *tracer = dev->tracer;
+	u32 in[MLX5_ST_SZ_DW(core_dump_reg)] = {};
+	int err;
+
+	if (!MLX5_CAP_DEBUG(dev, core_dump_general) || !tracer)
+		return -EOPNOTSUPP;
+	if (!tracer->owner)
+		return -EPERM;
+
+	MLX5_SET(core_dump_reg, in, core_dump_type, 0x0);
+
+	err =  mlx5_fw_tracer_set_core_dump_reg(dev, in, sizeof(in));
+	if (err)
+		return err;
+	queue_work(tracer->work_queue, &tracer->handle_traces_work);
+	flush_workqueue(tracer->work_queue);
+	return 0;
+}
+
+static int
+mlx5_devlink_fmsg_fill_trace(struct devlink_fmsg *fmsg,
+			     struct mlx5_fw_trace_data *trace_data)
+{
+	int err;
+
+	err = devlink_fmsg_obj_nest_start(fmsg);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u64_pair_put(fmsg, "timestamp", trace_data->timestamp);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_bool_pair_put(fmsg, "lost", trace_data->lost);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u8_pair_put(fmsg, "event_id", trace_data->event_id);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_string_pair_put(fmsg, "msg", trace_data->msg);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_obj_nest_end(fmsg);
+	if (err)
+		return err;
+	return 0;
+}
+
+int mlx5_fw_tracer_get_saved_traces_objects(struct mlx5_fw_tracer *tracer,
+					    struct devlink_fmsg *fmsg)
+{
+	struct mlx5_fw_trace_data *straces = tracer->st_arr.straces;
+	u32 index, start_index, end_index;
+	u32 saved_traces_index;
+	int err;
+
+	if (!straces[0].timestamp)
+		return -ENOMSG;
+
+	mutex_lock(&tracer->st_arr.lock);
+	saved_traces_index = tracer->st_arr.saved_traces_index;
+	if (straces[saved_traces_index].timestamp)
+		start_index = saved_traces_index;
+	else
+		start_index = 0;
+	end_index = (saved_traces_index - 1) & (SAVED_TRACES_NUM - 1);
+
+	err = devlink_fmsg_arr_pair_nest_start(fmsg, "dump fw traces");
+	if (err)
+		goto unlock;
+	index = start_index;
+	while (index != end_index) {
+		err = mlx5_devlink_fmsg_fill_trace(fmsg, &straces[index]);
+		if (err)
+			goto unlock;
+
+		index = (index + 1) & (SAVED_TRACES_NUM - 1);
+	}
+
+	err = devlink_fmsg_arr_pair_nest_end(fmsg);
+unlock:
+	mutex_unlock(&tracer->st_arr.lock);
+	return err;
+}
+
 /* Create software resources (Buffers, etc ..) */
 struct mlx5_fw_tracer *mlx5_fw_tracer_create(struct mlx5_core_dev *dev)
 {
@@ -833,6 +970,7 @@ struct mlx5_fw_tracer *mlx5_fw_tracer_create(struct mlx5_core_dev *dev)
 		goto free_log_buf;
 	}
 
+	mlx5_fw_tracer_init_saved_traces_array(tracer);
 	mlx5_core_dbg(dev, "FWTracer: Tracer created\n");
 
 	return tracer;
@@ -917,6 +1055,7 @@ void mlx5_fw_tracer_destroy(struct mlx5_fw_tracer *tracer)
 	cancel_work_sync(&tracer->read_fw_strings_work);
 	mlx5_fw_tracer_clean_ready_list(tracer);
 	mlx5_fw_tracer_clean_print_hash(tracer);
+	mlx5_fw_tracer_clean_saved_traces_array(tracer);
 	mlx5_fw_tracer_free_strings_db(tracer);
 	mlx5_fw_tracer_destroy_log_buf(tracer);
 	flush_workqueue(tracer->work_queue);

drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.h

@@ -46,6 +46,9 @@
 #define TRACER_BLOCK_SIZE_BYTE 256
 #define TRACES_PER_BLOCK 32
 
+#define TRACE_STR_MSG 256
+#define SAVED_TRACES_NUM 8192
+
 #define TRACER_MAX_PARAMS 7
 #define MESSAGE_HASH_BITS 6
 #define MESSAGE_HASH_SIZE BIT(MESSAGE_HASH_BITS)
@@ -53,6 +56,13 @@
 #define MASK_52_7 (0x1FFFFFFFFFFF80)
 #define MASK_6_0  (0x7F)
 
+struct mlx5_fw_trace_data {
+	u64 timestamp;
+	bool lost;
+	u8 event_id;
+	char msg[TRACE_STR_MSG];
+};
+
 struct mlx5_fw_tracer {
 	struct mlx5_core_dev *dev;
 	struct mlx5_nb        nb;
@@ -83,6 +93,13 @@ struct mlx5_fw_tracer {
 		u32 consumer_index;
 	} buff;
 
+	/* Saved Traces Array */
+	struct {
+		struct mlx5_fw_trace_data straces[SAVED_TRACES_NUM];
+		u32 saved_traces_index;
+		struct mutex lock; /* Protect st_arr access */
+	} st_arr;
+
 	u64 last_timestamp;
 	struct work_struct handle_traces_work;
 	struct hlist_head hash[MESSAGE_HASH_SIZE];
@@ -171,5 +188,8 @@ struct mlx5_fw_tracer *mlx5_fw_tracer_create(struct mlx5_core_dev *dev);
 int mlx5_fw_tracer_init(struct mlx5_fw_tracer *tracer);
 void mlx5_fw_tracer_cleanup(struct mlx5_fw_tracer *tracer);
 void mlx5_fw_tracer_destroy(struct mlx5_fw_tracer *tracer);
+int mlx5_fw_tracer_trigger_core_dump_general(struct mlx5_core_dev *dev);
+int mlx5_fw_tracer_get_saved_traces_objects(struct mlx5_fw_tracer *tracer,
+					    struct devlink_fmsg *fmsg);
 
 #endif

drivers/net/ethernet/mellanox/mlx5/core/en_selftest.c

@@ -64,7 +64,7 @@ static int mlx5e_test_health_info(struct mlx5e_priv *priv)
 {
 	struct mlx5_core_health *health = &priv->mdev->priv.health;
 
-	return health->sick ? 1 : 0;
+	return health->fatal_error ? 1 : 0;
 }
 
 static int mlx5e_test_link_state(struct mlx5e_priv *priv)

drivers/net/ethernet/mellanox/mlx5/core/health.c

@@ -40,6 +40,8 @@
 #include "mlx5_core.h"
 #include "lib/eq.h"
 #include "lib/mlx5.h"
+#include "lib/pci_vsc.h"
+#include "diag/fw_tracer.h"
 
 enum {
 	MLX5_HEALTH_POLL_INTERVAL	= 2 * HZ,
@@ -62,12 +64,20 @@ enum {
 
 enum {
 	MLX5_DROP_NEW_HEALTH_WORK,
-	MLX5_DROP_NEW_RECOVERY_WORK,
+};
+
+enum  {
+	MLX5_SENSOR_NO_ERR		= 0,
+	MLX5_SENSOR_PCI_COMM_ERR	= 1,
+	MLX5_SENSOR_PCI_ERR		= 2,
+	MLX5_SENSOR_NIC_DISABLED	= 3,
+	MLX5_SENSOR_NIC_SW_RESET	= 4,
+	MLX5_SENSOR_FW_SYND_RFR		= 5,
 };
 
 u8 mlx5_get_nic_state(struct mlx5_core_dev *dev)
 {
-	return (ioread32be(&dev->iseg->cmdq_addr_l_sz) >> 8) & 3;
+	return (ioread32be(&dev->iseg->cmdq_addr_l_sz) >> 8) & 7;
 }
 
 void mlx5_set_nic_state(struct mlx5_core_dev *dev, u8 state)
@@ -80,18 +90,105 @@ void mlx5_set_nic_state(struct mlx5_core_dev *dev, u8 state)
 		    &dev->iseg->cmdq_addr_l_sz);
 }
 
-static int in_fatal(struct mlx5_core_dev *dev)
+static bool sensor_pci_not_working(struct mlx5_core_dev *dev)
 {
 	struct mlx5_core_health *health = &dev->priv.health;
 	struct health_buffer __iomem *h = health->health;
 
+	/* Offline PCI reads return 0xffffffff */
+	return (ioread32be(&h->fw_ver) == 0xffffffff);
+}
+
+static bool sensor_fw_synd_rfr(struct mlx5_core_dev *dev)
+{
+	struct mlx5_core_health *health = &dev->priv.health;
+	struct health_buffer __iomem *h = health->health;
+	u32 rfr = ioread32be(&h->rfr) >> MLX5_RFR_OFFSET;
+	u8 synd = ioread8(&h->synd);
+
+	if (rfr && synd)
+		mlx5_core_dbg(dev, "FW requests reset, synd: %d\n", synd);
+	return rfr && synd;
+}
+
+static u32 check_fatal_sensors(struct mlx5_core_dev *dev)
+{
+	if (sensor_pci_not_working(dev))
+		return MLX5_SENSOR_PCI_COMM_ERR;
+	if (pci_channel_offline(dev->pdev))
+		return MLX5_SENSOR_PCI_ERR;
 	if (mlx5_get_nic_state(dev) == MLX5_NIC_IFC_DISABLED)
-		return 1;
+		return MLX5_SENSOR_NIC_DISABLED;
+	if (mlx5_get_nic_state(dev) == MLX5_NIC_IFC_SW_RESET)
+		return MLX5_SENSOR_NIC_SW_RESET;
+	if (sensor_fw_synd_rfr(dev))
+		return MLX5_SENSOR_FW_SYND_RFR;
 
-	if (ioread32be(&h->fw_ver) == 0xffffffff)
-		return 1;
+	return MLX5_SENSOR_NO_ERR;
+}
 
-	return 0;
+static int lock_sem_sw_reset(struct mlx5_core_dev *dev, bool lock)
+{
+	enum mlx5_vsc_state state;
+	int ret;
+
+	if (!mlx5_core_is_pf(dev))
+		return -EBUSY;
+
+	/* Try to lock GW access, this stage doesn't return
+	 * EBUSY because locked GW does not mean that other PF
+	 * already started the reset.
+	 */
+	ret = mlx5_vsc_gw_lock(dev);
+	if (ret == -EBUSY)
+		return -EINVAL;
+	if (ret)
+		return ret;
+
+	state = lock ? MLX5_VSC_LOCK : MLX5_VSC_UNLOCK;
+	/* At this stage, if the return status == EBUSY, then we know
+	 * for sure that another PF started the reset, so don't allow
+	 * another reset.
+	 */
+	ret = mlx5_vsc_sem_set_space(dev, MLX5_SEMAPHORE_SW_RESET, state);
+	if (ret)
+		mlx5_core_warn(dev, "Failed to lock SW reset semaphore\n");
+
+	/* Unlock GW access */
+	mlx5_vsc_gw_unlock(dev);
+
+	return ret;
+}
+
+static bool reset_fw_if_needed(struct mlx5_core_dev *dev)
+{
+	bool supported = (ioread32be(&dev->iseg->initializing) >>
+			  MLX5_FW_RESET_SUPPORTED_OFFSET) & 1;
+	u32 fatal_error;
+
+	if (!supported)
+		return false;
+
+	/* The reset only needs to be issued by one PF. The health buffer is
+	 * shared between all functions, and will be cleared during a reset.
+	 * Check again to avoid a redundant 2nd reset. If the fatal erros was
+	 * PCI related a reset won't help.
+	 */
+	fatal_error = check_fatal_sensors(dev);
+	if (fatal_error == MLX5_SENSOR_PCI_COMM_ERR ||
+	    fatal_error == MLX5_SENSOR_NIC_DISABLED ||
+	    fatal_error == MLX5_SENSOR_NIC_SW_RESET) {
+		mlx5_core_warn(dev, "Not issuing FW reset. Either it's already done or won't help.");
+		return false;
+	}
+
+	mlx5_core_warn(dev, "Issuing FW Reset\n");
+	/* Write the NIC interface field to initiate the reset, the command
+	 * interface address also resides here, don't overwrite it.
+	 */
+	mlx5_set_nic_state(dev, MLX5_NIC_IFC_SW_RESET);
+
+	return true;
 }
 
 void mlx5_enter_error_state(struct mlx5_core_dev *dev, bool force)
@@ -99,14 +196,65 @@ void mlx5_enter_error_state(struct mlx5_core_dev *dev, bool force)
 	mutex_lock(&dev->intf_state_mutex);
 	if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR)
 		goto unlock;
+	if (dev->state == MLX5_DEVICE_STATE_UNINITIALIZED) {
+		dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR;
+		goto unlock;
+	}
 
-	mlx5_core_err(dev, "start\n");
-	if (pci_channel_offline(dev->pdev) || in_fatal(dev) || force) {
+	if (check_fatal_sensors(dev) || force) {
 		dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR;
 		mlx5_cmd_flush(dev);
 	}
 
 	mlx5_notifier_call_chain(dev->priv.events, MLX5_DEV_EVENT_SYS_ERROR, (void *)1);
+unlock:
+	mutex_unlock(&dev->intf_state_mutex);
+}
+
+#define MLX5_CRDUMP_WAIT_MS	60000
+#define MLX5_FW_RESET_WAIT_MS	1000
+void mlx5_error_sw_reset(struct mlx5_core_dev *dev)
+{
+	unsigned long end, delay_ms = MLX5_FW_RESET_WAIT_MS;
+	int lock = -EBUSY;
+
+	mutex_lock(&dev->intf_state_mutex);
+	if (dev->state != MLX5_DEVICE_STATE_INTERNAL_ERROR)
+		goto unlock;
+
+	mlx5_core_err(dev, "start\n");
+
+	if (check_fatal_sensors(dev) == MLX5_SENSOR_FW_SYND_RFR) {
+		/* Get cr-dump and reset FW semaphore */
+		lock = lock_sem_sw_reset(dev, true);
+
+		if (lock == -EBUSY) {
+			delay_ms = MLX5_CRDUMP_WAIT_MS;
+			goto recover_from_sw_reset;
+		}
+		/* Execute SW reset */
+		reset_fw_if_needed(dev);
+	}
+
+recover_from_sw_reset:
+	/* Recover from SW reset */
+	end = jiffies + msecs_to_jiffies(delay_ms);
+	do {
+		if (mlx5_get_nic_state(dev) == MLX5_NIC_IFC_DISABLED)
+			break;
+
+		cond_resched();
+	} while (!time_after(jiffies, end));
+
+	if (mlx5_get_nic_state(dev) != MLX5_NIC_IFC_DISABLED) {
+		dev_err(&dev->pdev->dev, "NIC IFC still %d after %lums.\n",
+			mlx5_get_nic_state(dev), delay_ms);
+	}
+
+	/* Release FW semaphore if you are the lock owner */
+	if (!lock)
+		lock_sem_sw_reset(dev, false);
+
 	mlx5_core_err(dev, "end\n");
 
 unlock:
@@ -129,6 +277,20 @@ static void mlx5_handle_bad_state(struct mlx5_core_dev *dev)
 	case MLX5_NIC_IFC_NO_DRAM_NIC:
 		mlx5_core_warn(dev, "Expected to see disabled NIC but it is no dram nic\n");
 		break;
+
+	case MLX5_NIC_IFC_SW_RESET:
+		/* The IFC mode field is 3 bits, so it will read 0x7 in 2 cases:
+		 * 1. PCI has been disabled (ie. PCI-AER, PF driver unloaded
+		 *    and this is a VF), this is not recoverable by SW reset.
+		 *    Logging of this is handled elsewhere.
+		 * 2. FW reset has been issued by another function, driver can
+		 *    be reloaded to recover after the mode switches to
+		 *    MLX5_NIC_IFC_DISABLED.
+		 */
+		if (dev->priv.health.fatal_error != MLX5_SENSOR_PCI_COMM_ERR)
+			mlx5_core_warn(dev, "NIC SW reset in progress\n");
+		break;
+
 	default:
 		mlx5_core_warn(dev, "Expected to see disabled NIC but it is has invalid value %d\n",
 			       nic_interface);
@@ -137,52 +299,32 @@ static void mlx5_handle_bad_state(struct mlx5_core_dev *dev)
 	mlx5_disable_device(dev);
 }
 
-static void health_recover(struct work_struct *work)
-{
-	struct mlx5_core_health *health;
-	struct delayed_work *dwork;
-	struct mlx5_core_dev *dev;
-	struct mlx5_priv *priv;
-	u8 nic_state;
-
-	dwork = container_of(work, struct delayed_work, work);
-	health = container_of(dwork, struct mlx5_core_health, recover_work);
-	priv = container_of(health, struct mlx5_priv, health);
-	dev = container_of(priv, struct mlx5_core_dev, priv);
-
-	nic_state = mlx5_get_nic_state(dev);
-	if (nic_state == MLX5_NIC_IFC_INVALID) {
-		mlx5_core_err(dev, "health recovery flow aborted since the nic state is invalid\n");
-		return;
-	}
-
-	mlx5_core_err(dev, "starting health recovery flow\n");
-	mlx5_recover_device(dev);
-}
-
 /* How much time to wait until health resetting the driver (in msecs) */
-#define MLX5_RECOVERY_DELAY_MSECS 60000
-static void health_care(struct work_struct *work)
+#define MLX5_RECOVERY_WAIT_MSECS 60000
+static int mlx5_health_try_recover(struct mlx5_core_dev *dev)
 {
-	unsigned long recover_delay = msecs_to_jiffies(MLX5_RECOVERY_DELAY_MSECS);
-	struct mlx5_core_health *health;
-	struct mlx5_core_dev *dev;
-	struct mlx5_priv *priv;
-	unsigned long flags;
+	unsigned long end;
 
-	health = container_of(work, struct mlx5_core_health, work);
-	priv = container_of(health, struct mlx5_priv, health);
-	dev = container_of(priv, struct mlx5_core_dev, priv);
 	mlx5_core_warn(dev, "handling bad device here\n");
 	mlx5_handle_bad_state(dev);
+	end = jiffies + msecs_to_jiffies(MLX5_RECOVERY_WAIT_MSECS);
+	while (sensor_pci_not_working(dev)) {
+		if (time_after(jiffies, end)) {
+			mlx5_core_err(dev,
+				      "health recovery flow aborted, PCI reads still not working\n");
+			return -EIO;
+		}
+		msleep(100);
+	}
 
-	spin_lock_irqsave(&health->wq_lock, flags);
-	if (!test_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags))
-		schedule_delayed_work(&health->recover_work, recover_delay);
-	else
-		mlx5_core_err(dev,
-			      "new health works are not permitted at this stage\n");
-	spin_unlock_irqrestore(&health->wq_lock, flags);
+	mlx5_core_err(dev, "starting health recovery flow\n");
+	mlx5_recover_device(dev);
+	if (!test_bit(MLX5_INTERFACE_STATE_UP, &dev->intf_state) ||
+	    check_fatal_sensors(dev)) {
+		mlx5_core_err(dev, "health recovery failed\n");
+		return -EIO;
+	}
+	return 0;
 }
 
 static const char *hsynd_str(u8 synd)
@@ -246,6 +388,282 @@ static void print_health_info(struct mlx5_core_dev *dev)
 	mlx5_core_err(dev, "raw fw_ver 0x%08x\n", fw);
 }
 
+static int
+mlx5_fw_reporter_diagnose(struct devlink_health_reporter *reporter,
+			  struct devlink_fmsg *fmsg)
+{
+	struct mlx5_core_dev *dev = devlink_health_reporter_priv(reporter);
+	struct mlx5_core_health *health = &dev->priv.health;
+	struct health_buffer __iomem *h = health->health;
+	u8 synd;
+	int err;
+
+	synd = ioread8(&h->synd);
+	err = devlink_fmsg_u8_pair_put(fmsg, "Syndrome", synd);
+	if (err || !synd)
+		return err;
+	return devlink_fmsg_string_pair_put(fmsg, "Description", hsynd_str(synd));
+}
+
+struct mlx5_fw_reporter_ctx {
+	u8 err_synd;
+	int miss_counter;
+};
+
+static int
+mlx5_fw_reporter_ctx_pairs_put(struct devlink_fmsg *fmsg,
+			       struct mlx5_fw_reporter_ctx *fw_reporter_ctx)
+{
+	int err;
+
+	err = devlink_fmsg_u8_pair_put(fmsg, "syndrome",
+				       fw_reporter_ctx->err_synd);
+	if (err)
+		return err;
+	err = devlink_fmsg_u32_pair_put(fmsg, "fw_miss_counter",
+					fw_reporter_ctx->miss_counter);
+	if (err)
+		return err;
+	return 0;
+}
+
+static int
+mlx5_fw_reporter_heath_buffer_data_put(struct mlx5_core_dev *dev,
+				       struct devlink_fmsg *fmsg)
+{
+	struct mlx5_core_health *health = &dev->priv.health;
+	struct health_buffer __iomem *h = health->health;
+	int err;
+	int i;
+
+	if (!ioread8(&h->synd))
+		return 0;
+
+	err = devlink_fmsg_pair_nest_start(fmsg, "health buffer");
+	if (err)
+		return err;
+	err = devlink_fmsg_obj_nest_start(fmsg);
+	if (err)
+		return err;
+	err = devlink_fmsg_arr_pair_nest_start(fmsg, "assert_var");
+	if (err)
+		return err;
+
+	for (i = 0; i < ARRAY_SIZE(h->assert_var); i++) {
+		err = devlink_fmsg_u32_put(fmsg, ioread32be(h->assert_var + i));
+		if (err)
+			return err;
+	}
+	err = devlink_fmsg_arr_pair_nest_end(fmsg);
+	if (err)
+		return err;
+	err = devlink_fmsg_u32_pair_put(fmsg, "assert_exit_ptr",
+					ioread32be(&h->assert_exit_ptr));
+	if (err)
+		return err;
+	err = devlink_fmsg_u32_pair_put(fmsg, "assert_callra",
+					ioread32be(&h->assert_callra));
+	if (err)
+		return err;
+	err = devlink_fmsg_u32_pair_put(fmsg, "hw_id", ioread32be(&h->hw_id));
+	if (err)
+		return err;
+	err = devlink_fmsg_u8_pair_put(fmsg, "irisc_index",
+				       ioread8(&h->irisc_index));
+	if (err)
+		return err;
+	err = devlink_fmsg_u8_pair_put(fmsg, "synd", ioread8(&h->synd));
+	if (err)
+		return err;
+	err = devlink_fmsg_u32_pair_put(fmsg, "ext_synd",
+					ioread16be(&h->ext_synd));
+	if (err)
+		return err;
+	err = devlink_fmsg_u32_pair_put(fmsg, "raw_fw_ver",
+					ioread32be(&h->fw_ver));
+	if (err)
+		return err;
+	err = devlink_fmsg_obj_nest_end(fmsg);
+	if (err)
+		return err;
+	return devlink_fmsg_pair_nest_end(fmsg);
+}
+
+static int
+mlx5_fw_reporter_dump(struct devlink_health_reporter *reporter,
+		      struct devlink_fmsg *fmsg, void *priv_ctx)
+{
+	struct mlx5_core_dev *dev = devlink_health_reporter_priv(reporter);
+	int err;
+
+	err = mlx5_fw_tracer_trigger_core_dump_general(dev);
+	if (err)
+		return err;
+
+	if (priv_ctx) {
+		struct mlx5_fw_reporter_ctx *fw_reporter_ctx = priv_ctx;
+
+		err = mlx5_fw_reporter_ctx_pairs_put(fmsg, fw_reporter_ctx);
+		if (err)
+			return err;
+	}
+
+	err = mlx5_fw_reporter_heath_buffer_data_put(dev, fmsg);
+	if (err)
+		return err;
+	return mlx5_fw_tracer_get_saved_traces_objects(dev->tracer, fmsg);
+}
+
+static void mlx5_fw_reporter_err_work(struct work_struct *work)
+{
+	struct mlx5_fw_reporter_ctx fw_reporter_ctx;
+	struct mlx5_core_health *health;
+
+	health = container_of(work, struct mlx5_core_health, report_work);
+
+	if (IS_ERR_OR_NULL(health->fw_reporter))
+		return;
+
+	fw_reporter_ctx.err_synd = health->synd;
+	fw_reporter_ctx.miss_counter = health->miss_counter;
+	if (fw_reporter_ctx.err_synd) {
+		devlink_health_report(health->fw_reporter,
+				      "FW syndrom reported", &fw_reporter_ctx);
+		return;
+	}
+	if (fw_reporter_ctx.miss_counter)
+		devlink_health_report(health->fw_reporter,
+				      "FW miss counter reported",
+				      &fw_reporter_ctx);
+}
+
+static const struct devlink_health_reporter_ops mlx5_fw_reporter_ops = {
+		.name = "fw",
+		.diagnose = mlx5_fw_reporter_diagnose,
+		.dump = mlx5_fw_reporter_dump,
+};
+
+static int
+mlx5_fw_fatal_reporter_recover(struct devlink_health_reporter *reporter,
+			       void *priv_ctx)
+{
+	struct mlx5_core_dev *dev = devlink_health_reporter_priv(reporter);
+
+	return mlx5_health_try_recover(dev);
+}
+
+#define MLX5_CR_DUMP_CHUNK_SIZE 256
+static int
+mlx5_fw_fatal_reporter_dump(struct devlink_health_reporter *reporter,
+			    struct devlink_fmsg *fmsg, void *priv_ctx)
+{
+	struct mlx5_core_dev *dev = devlink_health_reporter_priv(reporter);
+	u32 crdump_size = dev->priv.health.crdump_size;
+	u32 *cr_data;
+	u32 data_size;
+	u32 offset;
+	int err;
+
+	if (!mlx5_core_is_pf(dev))
+		return -EPERM;
+
+	cr_data = kvmalloc(crdump_size, GFP_KERNEL);
+	if (!cr_data)
+		return -ENOMEM;
+	err = mlx5_crdump_collect(dev, cr_data);
+	if (err)
+		return err;
+
+	if (priv_ctx) {
+		struct mlx5_fw_reporter_ctx *fw_reporter_ctx = priv_ctx;
+
+		err = mlx5_fw_reporter_ctx_pairs_put(fmsg, fw_reporter_ctx);
+		if (err)
+			goto free_data;
+	}
+
+	err = devlink_fmsg_arr_pair_nest_start(fmsg, "crdump_data");
+	if (err)
+		goto free_data;
+	for (offset = 0; offset < crdump_size; offset += data_size) {
+		if (crdump_size - offset < MLX5_CR_DUMP_CHUNK_SIZE)
+			data_size = crdump_size - offset;
+		else
+			data_size = MLX5_CR_DUMP_CHUNK_SIZE;
+		err = devlink_fmsg_binary_put(fmsg, cr_data, data_size);
+		if (err)
+			goto free_data;
+	}
+	err = devlink_fmsg_arr_pair_nest_end(fmsg);
+
+free_data:
+	kfree(cr_data);
+	return err;
+}
+
+static void mlx5_fw_fatal_reporter_err_work(struct work_struct *work)
+{
+	struct mlx5_fw_reporter_ctx fw_reporter_ctx;
+	struct mlx5_core_health *health;
+	struct mlx5_core_dev *dev;
+	struct mlx5_priv *priv;
+
+	health = container_of(work, struct mlx5_core_health, fatal_report_work);
+	priv = container_of(health, struct mlx5_priv, health);
+	dev = container_of(priv, struct mlx5_core_dev, priv);
+
+	mlx5_enter_error_state(dev, false);
+	if (IS_ERR_OR_NULL(health->fw_fatal_reporter)) {
+		if (mlx5_health_try_recover(dev))
+			mlx5_core_err(dev, "health recovery failed\n");
+		return;
+	}
+	fw_reporter_ctx.err_synd = health->synd;
+	fw_reporter_ctx.miss_counter = health->miss_counter;
+	devlink_health_report(health->fw_fatal_reporter,
+			      "FW fatal error reported", &fw_reporter_ctx);
+}
+
+static const struct devlink_health_reporter_ops mlx5_fw_fatal_reporter_ops = {
+		.name = "fw_fatal",
+		.recover = mlx5_fw_fatal_reporter_recover,
+		.dump = mlx5_fw_fatal_reporter_dump,
+};
+
+#define MLX5_REPORTER_FW_GRACEFUL_PERIOD 1200000
+static void mlx5_fw_reporters_create(struct mlx5_core_dev *dev)
+{
+	struct mlx5_core_health *health = &dev->priv.health;
+	struct devlink *devlink = priv_to_devlink(dev);
+
+	health->fw_reporter =
+		devlink_health_reporter_create(devlink, &mlx5_fw_reporter_ops,
+					       0, false, dev);
+	if (IS_ERR(health->fw_reporter))
+		mlx5_core_warn(dev, "Failed to create fw reporter, err = %ld\n",
+			       PTR_ERR(health->fw_reporter));
+
+	health->fw_fatal_reporter =
+		devlink_health_reporter_create(devlink,
+					       &mlx5_fw_fatal_reporter_ops,
+					       MLX5_REPORTER_FW_GRACEFUL_PERIOD,
+					       true, dev);
+	if (IS_ERR(health->fw_fatal_reporter))
+		mlx5_core_warn(dev, "Failed to create fw fatal reporter, err = %ld\n",
+			       PTR_ERR(health->fw_fatal_reporter));
+}
+
+static void mlx5_fw_reporters_destroy(struct mlx5_core_dev *dev)
+{
+	struct mlx5_core_health *health = &dev->priv.health;
+
+	if (!IS_ERR_OR_NULL(health->fw_reporter))
+		devlink_health_reporter_destroy(health->fw_reporter);
+
+	if (!IS_ERR_OR_NULL(health->fw_fatal_reporter))
+		devlink_health_reporter_destroy(health->fw_fatal_reporter);
+}
+
 static unsigned long get_next_poll_jiffies(void)
 {
 	unsigned long next;
@@ -264,7 +682,7 @@ void mlx5_trigger_health_work(struct mlx5_core_dev *dev)
 
 	spin_lock_irqsave(&health->wq_lock, flags);
 	if (!test_bit(MLX5_DROP_NEW_HEALTH_WORK, &health->flags))
-		queue_work(health->wq, &health->work);
+		queue_work(health->wq, &health->fatal_report_work);
 	else
 		mlx5_core_err(dev, "new health works are not permitted at this stage\n");
 	spin_unlock_irqrestore(&health->wq_lock, flags);
@@ -274,6 +692,9 @@ static void poll_health(struct timer_list *t)
 {
 	struct mlx5_core_dev *dev = from_timer(dev, t, priv.health.timer);
 	struct mlx5_core_health *health = &dev->priv.health;
+	struct health_buffer __iomem *h = health->health;
+	u32 fatal_error;
+	u8 prev_synd;
 	u32 count;
 
 	if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR)
@@ -289,10 +710,19 @@ static void poll_health(struct timer_list *t)
 	if (health->miss_counter == MAX_MISSES) {
 		mlx5_core_err(dev, "device's health compromised - reached miss count\n");
 		print_health_info(dev);
+		queue_work(health->wq, &health->report_work);
 	}
 
-	if (in_fatal(dev) && !health->sick) {
-		health->sick = true;
+	prev_synd = health->synd;
+	health->synd = ioread8(&h->synd);
+	if (health->synd && health->synd != prev_synd)
+		queue_work(health->wq, &health->report_work);
+
+	fatal_error = check_fatal_sensors(dev);
+
+	if (fatal_error && !health->fatal_error) {
+		mlx5_core_err(dev, "Fatal error %u detected\n", fatal_error);
+		dev->priv.health.fatal_error = fatal_error;
 		print_health_info(dev);
 		mlx5_trigger_health_work(dev);
 	}
@@ -306,9 +736,8 @@ void mlx5_start_health_poll(struct mlx5_core_dev *dev)
 	struct mlx5_core_health *health = &dev->priv.health;
 
 	timer_setup(&health->timer, poll_health, 0);
-	health->sick = 0;
+	health->fatal_error = MLX5_SENSOR_NO_ERR;
 	clear_bit(MLX5_DROP_NEW_HEALTH_WORK, &health->flags);
-	clear_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags);
 	health->health = &dev->iseg->health;
 	health->health_counter = &dev->iseg->health_counter;
 
@@ -324,7 +753,6 @@ void mlx5_stop_health_poll(struct mlx5_core_dev *dev, bool disable_health)
 	if (disable_health) {
 		spin_lock_irqsave(&health->wq_lock, flags);
 		set_bit(MLX5_DROP_NEW_HEALTH_WORK, &health->flags);
-		set_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags);
 		spin_unlock_irqrestore(&health->wq_lock, flags);
 	}
 
@@ -338,21 +766,9 @@ void mlx5_drain_health_wq(struct mlx5_core_dev *dev)
 
 	spin_lock_irqsave(&health->wq_lock, flags);
 	set_bit(MLX5_DROP_NEW_HEALTH_WORK, &health->flags);
-	set_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags);
 	spin_unlock_irqrestore(&health->wq_lock, flags);
-	cancel_delayed_work_sync(&health->recover_work);
-	cancel_work_sync(&health->work);
-}
-
-void mlx5_drain_health_recovery(struct mlx5_core_dev *dev)
-{
-	struct mlx5_core_health *health = &dev->priv.health;
-	unsigned long flags;
-
-	spin_lock_irqsave(&health->wq_lock, flags);
-	set_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags);
-	spin_unlock_irqrestore(&health->wq_lock, flags);
-	cancel_delayed_work_sync(&dev->priv.health.recover_work);
+	cancel_work_sync(&health->report_work);
+	cancel_work_sync(&health->fatal_report_work);
 }
 
 void mlx5_health_flush(struct mlx5_core_dev *dev)
@@ -367,6 +783,7 @@ void mlx5_health_cleanup(struct mlx5_core_dev *dev)
 	struct mlx5_core_health *health = &dev->priv.health;
 
 	destroy_workqueue(health->wq);
+	mlx5_fw_reporters_destroy(dev);
 }
 
 int mlx5_health_init(struct mlx5_core_dev *dev)
@@ -374,20 +791,26 @@ int mlx5_health_init(struct mlx5_core_dev *dev)
 	struct mlx5_core_health *health;
 	char *name;
 
+	mlx5_fw_reporters_create(dev);
+
 	health = &dev->priv.health;
 	name = kmalloc(64, GFP_KERNEL);
 	if (!name)
-		return -ENOMEM;
+		goto out_err;
 
 	strcpy(name, "mlx5_health");
 	strcat(name, dev_name(dev->device));
 	health->wq = create_singlethread_workqueue(name);
 	kfree(name);
 	if (!health->wq)
-		return -ENOMEM;
+		goto out_err;
 	spin_lock_init(&health->wq_lock);
-	INIT_WORK(&health->work, health_care);
-	INIT_DELAYED_WORK(&health->recover_work, health_recover);
+	INIT_WORK(&health->fatal_report_work, mlx5_fw_fatal_reporter_err_work);
+	INIT_WORK(&health->report_work, mlx5_fw_reporter_err_work);
 
 	return 0;
+
+out_err:
+	mlx5_fw_reporters_destroy(dev);
+	return -ENOMEM;
 }

drivers/net/ethernet/mellanox/mlx5/core/lib/mlx5.h

@@ -41,6 +41,9 @@ int  mlx5_core_reserve_gids(struct mlx5_core_dev *dev, unsigned int count);
 void mlx5_core_unreserve_gids(struct mlx5_core_dev *dev, unsigned int count);
 int  mlx5_core_reserved_gid_alloc(struct mlx5_core_dev *dev, int *gid_index);
 void mlx5_core_reserved_gid_free(struct mlx5_core_dev *dev, int gid_index);
+int mlx5_crdump_enable(struct mlx5_core_dev *dev);
+void mlx5_crdump_disable(struct mlx5_core_dev *dev);
+int mlx5_crdump_collect(struct mlx5_core_dev *dev, u32 *cr_data);
 
 /* TODO move to lib/events.h */
 

drivers/net/ethernet/mellanox/mlx5/core/lib/pci_vsc.c

+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/* Copyright (c) 2019 Mellanox Technologies */
+
+#include <linux/pci.h>
+#include "mlx5_core.h"
+#include "pci_vsc.h"
+
+#define MLX5_EXTRACT_C(source, offset, size)	\
+	((((u32)(source)) >> (offset)) & MLX5_ONES32(size))
+#define MLX5_EXTRACT(src, start, len)		\
+	(((len) == 32) ? (src) : MLX5_EXTRACT_C(src, start, len))
+#define MLX5_ONES32(size)			\
+	((size) ? (0xffffffff >> (32 - (size))) : 0)
+#define MLX5_MASK32(offset, size)		\
+	(MLX5_ONES32(size) << (offset))
+#define MLX5_MERGE_C(rsrc1, rsrc2, start, len)  \
+	((((rsrc2) << (start)) & (MLX5_MASK32((start), (len)))) | \
+	((rsrc1) & (~MLX5_MASK32((start), (len)))))
+#define MLX5_MERGE(rsrc1, rsrc2, start, len)	\
+	(((len) == 32) ? (rsrc2) : MLX5_MERGE_C(rsrc1, rsrc2, start, len))
+#define vsc_read(dev, offset, val) \
+	pci_read_config_dword((dev)->pdev, (dev)->vsc_addr + (offset), (val))
+#define vsc_write(dev, offset, val) \
+	pci_write_config_dword((dev)->pdev, (dev)->vsc_addr + (offset), (val))
+#define VSC_MAX_RETRIES 2048
+
+enum {
+	VSC_CTRL_OFFSET = 0x4,
+	VSC_COUNTER_OFFSET = 0x8,
+	VSC_SEMAPHORE_OFFSET = 0xc,
+	VSC_ADDR_OFFSET = 0x10,
+	VSC_DATA_OFFSET = 0x14,
+
+	VSC_FLAG_BIT_OFFS = 31,
+	VSC_FLAG_BIT_LEN = 1,
+
+	VSC_SYND_BIT_OFFS = 30,
+	VSC_SYND_BIT_LEN = 1,
+
+	VSC_ADDR_BIT_OFFS = 0,
+	VSC_ADDR_BIT_LEN = 30,
+
+	VSC_SPACE_BIT_OFFS = 0,
+	VSC_SPACE_BIT_LEN = 16,
+
+	VSC_SIZE_VLD_BIT_OFFS = 28,
+	VSC_SIZE_VLD_BIT_LEN = 1,
+
+	VSC_STATUS_BIT_OFFS = 29,
+	VSC_STATUS_BIT_LEN = 3,
+};
+
+void mlx5_pci_vsc_init(struct mlx5_core_dev *dev)
+{
+	if (!mlx5_core_is_pf(dev))
+		return;
+
+	dev->vsc_addr = pci_find_capability(dev->pdev,
+					    PCI_CAP_ID_VNDR);
+	if (!dev->vsc_addr)
+		mlx5_core_warn(dev, "Failed to get valid vendor specific ID\n");
+}
+
+int mlx5_vsc_gw_lock(struct mlx5_core_dev *dev)
+{
+	u32 counter = 0;
+	int retries = 0;
+	u32 lock_val;
+	int ret;
+
+	pci_cfg_access_lock(dev->pdev);
+	do {
+		if (retries > VSC_MAX_RETRIES) {
+			ret = -EBUSY;
+			goto pci_unlock;
+		}
+
+		/* Check if semaphore is already locked */
+		ret = vsc_read(dev, VSC_SEMAPHORE_OFFSET, &lock_val);
+		if (ret)
+			goto pci_unlock;
+
+		if (lock_val) {
+			retries++;
+			usleep_range(1000, 2000);
+			continue;
+		}
+
+		/* Read and write counter value, if written value is
+		 * the same, semaphore was acquired successfully.
+		 */
+		ret = vsc_read(dev, VSC_COUNTER_OFFSET, &counter);
+		if (ret)
+			goto pci_unlock;
+
+		ret = vsc_write(dev, VSC_SEMAPHORE_OFFSET, counter);
+		if (ret)
+			goto pci_unlock;
+
+		ret = vsc_read(dev, VSC_SEMAPHORE_OFFSET, &lock_val);
+		if (ret)
+			goto pci_unlock;
+
+		retries++;
+	} while (counter != lock_val);
+
+	return 0;
+
+pci_unlock:
+	pci_cfg_access_unlock(dev->pdev);
+	return ret;
+}
+
+int mlx5_vsc_gw_unlock(struct mlx5_core_dev *dev)
+{
+	int ret;
+
+	ret = vsc_write(dev, VSC_SEMAPHORE_OFFSET, MLX5_VSC_UNLOCK);
+	pci_cfg_access_unlock(dev->pdev);
+	return ret;
+}
+
+int mlx5_vsc_gw_set_space(struct mlx5_core_dev *dev, u16 space,
+			  u32 *ret_space_size)
+{
+	int ret;
+	u32 val = 0;
+
+	if (!mlx5_vsc_accessible(dev))
+		return -EINVAL;
+
+	if (ret_space_size)
+		*ret_space_size = 0;
+
+	/* Get a unique val */
+	ret = vsc_read(dev, VSC_CTRL_OFFSET, &val);
+	if (ret)
+		goto out;
+
+	/* Try to modify the lock */
+	val = MLX5_MERGE(val, space, VSC_SPACE_BIT_OFFS, VSC_SPACE_BIT_LEN);
+	ret = vsc_write(dev, VSC_CTRL_OFFSET, val);
+	if (ret)
+		goto out;
+
+	/* Verify lock was modified */
+	ret = vsc_read(dev, VSC_CTRL_OFFSET, &val);
+	if (ret)
+		goto out;
+
+	if (MLX5_EXTRACT(val, VSC_STATUS_BIT_OFFS, VSC_STATUS_BIT_LEN) == 0)
+		return -EINVAL;
+
+	/* Get space max address if indicated by size valid bit */
+	if (ret_space_size &&
+	    MLX5_EXTRACT(val, VSC_SIZE_VLD_BIT_OFFS, VSC_SIZE_VLD_BIT_LEN)) {
+		ret = vsc_read(dev, VSC_ADDR_OFFSET, &val);
+		if (ret) {
+			mlx5_core_warn(dev, "Failed to get max space size\n");
+			goto out;
+		}
+		*ret_space_size = MLX5_EXTRACT(val, VSC_ADDR_BIT_OFFS,
+					       VSC_ADDR_BIT_LEN);
+	}
+	return 0;
+
+out:
+	return ret;
+}
+
+static int mlx5_vsc_wait_on_flag(struct mlx5_core_dev *dev, u8 expected_val)
+{
+	int retries = 0;
+	u32 flag;
+	int ret;
+
+	do {
+		if (retries > VSC_MAX_RETRIES)
+			return -EBUSY;
+
+		ret = vsc_read(dev, VSC_ADDR_OFFSET, &flag);
+		if (ret)
+			return ret;
+		flag = MLX5_EXTRACT(flag, VSC_FLAG_BIT_OFFS, VSC_FLAG_BIT_LEN);
+		retries++;
+
+		if ((retries & 0xf) == 0)
+			usleep_range(1000, 2000);
+
+	} while (flag != expected_val);
+
+	return 0;
+}
+
+static int mlx5_vsc_gw_write(struct mlx5_core_dev *dev, unsigned int address,
+			     u32 data)
+{
+	int ret;
+
+	if (MLX5_EXTRACT(address, VSC_SYND_BIT_OFFS,
+			 VSC_FLAG_BIT_LEN + VSC_SYND_BIT_LEN))
+		return -EINVAL;
+
+	/* Set flag to 0x1 */
+	address = MLX5_MERGE(address, 1, VSC_FLAG_BIT_OFFS, 1);
+	ret = vsc_write(dev, VSC_DATA_OFFSET, data);
+	if (ret)
+		goto out;
+
+	ret = vsc_write(dev, VSC_ADDR_OFFSET, address);
+	if (ret)
+		goto out;
+
+	/* Wait for the flag to be cleared */
+	ret = mlx5_vsc_wait_on_flag(dev, 0);
+
+out:
+	return ret;
+}
+
+static int mlx5_vsc_gw_read(struct mlx5_core_dev *dev, unsigned int address,
+			    u32 *data)
+{
+	int ret;
+
+	if (MLX5_EXTRACT(address, VSC_SYND_BIT_OFFS,
+			 VSC_FLAG_BIT_LEN + VSC_SYND_BIT_LEN))
+		return -EINVAL;
+
+	ret = vsc_write(dev, VSC_ADDR_OFFSET, address);
+	if (ret)
+		goto out;
+
+	ret = mlx5_vsc_wait_on_flag(dev, 1);
+	if (ret)
+		goto out;
+
+	ret = vsc_read(dev, VSC_DATA_OFFSET, data);
+out:
+	return ret;
+}
+
+static int mlx5_vsc_gw_read_fast(struct mlx5_core_dev *dev,
+				 unsigned int read_addr,
+				 unsigned int *next_read_addr,
+				 u32 *data)
+{
+	int ret;
+
+	ret = mlx5_vsc_gw_read(dev, read_addr, data);
+	if (ret)
+		goto out;
+
+	ret = vsc_read(dev, VSC_ADDR_OFFSET, next_read_addr);
+	if (ret)
+		goto out;
+
+	*next_read_addr = MLX5_EXTRACT(*next_read_addr, VSC_ADDR_BIT_OFFS,
+				       VSC_ADDR_BIT_LEN);
+
+	if (*next_read_addr <= read_addr)
+		ret = -EINVAL;
+out:
+	return ret;
+}
+
+int mlx5_vsc_gw_read_block_fast(struct mlx5_core_dev *dev, u32 *data,
+				int length)
+{
+	unsigned int next_read_addr = 0;
+	unsigned int read_addr = 0;
+
+	while (read_addr < length) {
+		if (mlx5_vsc_gw_read_fast(dev, read_addr, &next_read_addr,
+					  &data[(read_addr >> 2)]))
+			return read_addr;
+
+		read_addr = next_read_addr;
+	}
+	return length;
+}
+
+int mlx5_vsc_sem_set_space(struct mlx5_core_dev *dev, u16 space,
+			   enum mlx5_vsc_state state)
+{
+	u32 data, id = 0;
+	int ret;
+
+	ret = mlx5_vsc_gw_set_space(dev, MLX5_SEMAPHORE_SPACE_DOMAIN, NULL);
+	if (ret) {
+		mlx5_core_warn(dev, "Failed to set gw space %d\n", ret);
+		return ret;
+	}
+
+	if (state == MLX5_VSC_LOCK) {
+		/* Get a unique ID based on the counter */
+		ret = vsc_read(dev, VSC_COUNTER_OFFSET, &id);
+		if (ret)
+			return ret;
+	}
+
+	/* Try to modify lock */
+	ret = mlx5_vsc_gw_write(dev, space, id);
+	if (ret)
+		return ret;
+
+	/* Verify lock was modified */
+	ret = mlx5_vsc_gw_read(dev, space, &data);
+	if (ret)
+		return -EINVAL;
+
+	if (data != id)
+		return -EBUSY;
+
+	return 0;
+}

drivers/net/ethernet/mellanox/mlx5/core/lib/pci_vsc.h

+/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
+/* Copyright (c) 2019 Mellanox Technologies */
+
+#ifndef __MLX5_PCI_VSC_H__
+#define __MLX5_PCI_VSC_H__
+
+enum mlx5_vsc_state {
+	MLX5_VSC_UNLOCK,
+	MLX5_VSC_LOCK,
+};
+
+enum {
+	MLX5_VSC_SPACE_SCAN_CRSPACE = 0x7,
+};
+
+void mlx5_pci_vsc_init(struct mlx5_core_dev *dev);
+int mlx5_vsc_gw_lock(struct mlx5_core_dev *dev);
+int mlx5_vsc_gw_unlock(struct mlx5_core_dev *dev);
+int mlx5_vsc_gw_set_space(struct mlx5_core_dev *dev, u16 space,
+			  u32 *ret_space_size);
+int mlx5_vsc_gw_read_block_fast(struct mlx5_core_dev *dev, u32 *data,
+				int length);
+
+static inline bool mlx5_vsc_accessible(struct mlx5_core_dev *dev)
+{
+	return !!dev->vsc_addr;
+}
+
+int mlx5_vsc_sem_set_space(struct mlx5_core_dev *dev, u16 space,
+			   enum mlx5_vsc_state state);
+
+#endif /* __MLX5_PCI_VSC_H__ */

drivers/net/ethernet/mellanox/mlx5/core/main.c

@@ -56,6 +56,7 @@
 #include "fs_core.h"
 #include "lib/mpfs.h"
 #include "eswitch.h"
+#include "devlink.h"
 #include "lib/mlx5.h"
 #include "fpga/core.h"
 #include "fpga/ipsec.h"
@@ -65,6 +66,7 @@
 #include "lib/vxlan.h"
 #include "lib/geneve.h"
 #include "lib/devcom.h"
+#include "lib/pci_vsc.h"
 #include "diag/fw_tracer.h"
 #include "ecpf.h"
 
@@ -762,6 +764,8 @@ static int mlx5_pci_init(struct mlx5_core_dev *dev, struct pci_dev *pdev,
 		goto err_clr_master;
 	}
 
+	mlx5_pci_vsc_init(dev);
+
 	return 0;
 
 err_clr_master:
@@ -1187,7 +1191,7 @@ static int mlx5_unload_one(struct mlx5_core_dev *dev, bool cleanup)
 	int err = 0;
 
 	if (cleanup)
-		mlx5_drain_health_recovery(dev);
+		mlx5_drain_health_wq(dev);
 
 	mutex_lock(&dev->intf_state_mutex);
 	if (!test_bit(MLX5_INTERFACE_STATE_UP, &dev->intf_state)) {
@@ -1214,37 +1218,6 @@ static int mlx5_unload_one(struct mlx5_core_dev *dev, bool cleanup)
 	return err;
 }
 
-static int mlx5_devlink_flash_update(struct devlink *devlink,
-				     const char *file_name,
-				     const char *component,
-				     struct netlink_ext_ack *extack)
-{
-	struct mlx5_core_dev *dev = devlink_priv(devlink);
-	const struct firmware *fw;
-	int err;
-
-	if (component)
-		return -EOPNOTSUPP;
-
-	err = request_firmware_direct(&fw, file_name, &dev->pdev->dev);
-	if (err)
-		return err;
-
-	return mlx5_firmware_flash(dev, fw, extack);
-}
-
-static const struct devlink_ops mlx5_devlink_ops = {
-#ifdef CONFIG_MLX5_ESWITCH
-	.eswitch_mode_set = mlx5_devlink_eswitch_mode_set,
-	.eswitch_mode_get = mlx5_devlink_eswitch_mode_get,
-	.eswitch_inline_mode_set = mlx5_devlink_eswitch_inline_mode_set,
-	.eswitch_inline_mode_get = mlx5_devlink_eswitch_inline_mode_get,
-	.eswitch_encap_mode_set = mlx5_devlink_eswitch_encap_mode_set,
-	.eswitch_encap_mode_get = mlx5_devlink_eswitch_encap_mode_get,
-#endif
-	.flash_update = mlx5_devlink_flash_update,
-};
-
 static int mlx5_mdev_init(struct mlx5_core_dev *dev, int profile_idx)
 {
 	struct mlx5_priv *priv = &dev->priv;
@@ -1306,9 +1279,9 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 	struct devlink *devlink;
 	int err;
 
-	devlink = devlink_alloc(&mlx5_devlink_ops, sizeof(*dev));
+	devlink = mlx5_devlink_alloc();
 	if (!devlink) {
-		dev_err(&pdev->dev, "kzalloc failed\n");
+		dev_err(&pdev->dev, "devlink alloc failed\n");
 		return -ENOMEM;
 	}
 
@@ -1336,10 +1309,14 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 
 	request_module_nowait(MLX5_IB_MOD);
 
-	err = devlink_register(devlink, &pdev->dev);
+	err = mlx5_devlink_register(devlink, &pdev->dev);
 	if (err)
 		goto clean_load;
 
+	err = mlx5_crdump_enable(dev);
+	if (err)
+		dev_err(&pdev->dev, "mlx5_crdump_enable failed with error code %d\n", err);
+
 	pci_save_state(pdev);
 	return 0;
 
@@ -1351,7 +1328,7 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 pci_init_err:
 	mlx5_mdev_uninit(dev);
 mdev_init_err:
-	devlink_free(devlink);
+	mlx5_devlink_free(devlink);
 
 	return err;
 }
@@ -1361,7 +1338,8 @@ static void remove_one(struct pci_dev *pdev)
 	struct mlx5_core_dev *dev  = pci_get_drvdata(pdev);
 	struct devlink *devlink = priv_to_devlink(dev);
 
-	devlink_unregister(devlink);
+	mlx5_crdump_disable(dev);
+	mlx5_devlink_unregister(devlink);
 	mlx5_unregister_device(dev);
 
 	if (mlx5_unload_one(dev, true)) {
@@ -1372,7 +1350,7 @@ static void remove_one(struct pci_dev *pdev)
 
 	mlx5_pci_close(dev);
 	mlx5_mdev_uninit(dev);
-	devlink_free(devlink);
+	mlx5_devlink_free(devlink);
 }
 
 static pci_ers_result_t mlx5_pci_err_detected(struct pci_dev *pdev,
@@ -1383,12 +1361,10 @@ static pci_ers_result_t mlx5_pci_err_detected(struct pci_dev *pdev,
 	mlx5_core_info(dev, "%s was called\n", __func__);
 
 	mlx5_enter_error_state(dev, false);
+	mlx5_error_sw_reset(dev);
 	mlx5_unload_one(dev, false);
-	/* In case of kernel call drain the health wq */
-	if (state) {
-		mlx5_drain_health_wq(dev);
-		mlx5_pci_disable_device(dev);
-	}
+	mlx5_drain_health_wq(dev);
+	mlx5_pci_disable_device(dev);
 
 	return state == pci_channel_io_perm_failure ?
 		PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_NEED_RESET;
@@ -1556,7 +1532,8 @@ MODULE_DEVICE_TABLE(pci, mlx5_core_pci_table);
 
 void mlx5_disable_device(struct mlx5_core_dev *dev)
 {
-	mlx5_pci_err_detected(dev->pdev, 0);
+	mlx5_error_sw_reset(dev);
+	mlx5_unload_one(dev, false);
 }
 
 void mlx5_recover_device(struct mlx5_core_dev *dev)

drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h

@@ -111,6 +111,11 @@ enum {
 	MLX5_DRIVER_SYND = 0xbadd00de,
 };
 
+enum mlx5_semaphore_space_address {
+	MLX5_SEMAPHORE_SPACE_DOMAIN     = 0xA,
+	MLX5_SEMAPHORE_SW_RESET         = 0x20,
+};
+
 int mlx5_query_hca_caps(struct mlx5_core_dev *dev);
 int mlx5_query_board_id(struct mlx5_core_dev *dev);
 int mlx5_cmd_init_hca(struct mlx5_core_dev *dev, uint32_t *sw_owner_id);
@@ -118,6 +123,7 @@ int mlx5_cmd_teardown_hca(struct mlx5_core_dev *dev);
 int mlx5_cmd_force_teardown_hca(struct mlx5_core_dev *dev);
 int mlx5_cmd_fast_teardown_hca(struct mlx5_core_dev *dev);
 void mlx5_enter_error_state(struct mlx5_core_dev *dev, bool force);
+void mlx5_error_sw_reset(struct mlx5_core_dev *dev);
 void mlx5_disable_device(struct mlx5_core_dev *dev);
 void mlx5_recover_device(struct mlx5_core_dev *dev);
 int mlx5_sriov_init(struct mlx5_core_dev *dev);
@@ -214,7 +220,7 @@ enum {
 	MLX5_NIC_IFC_FULL		= 0,
 	MLX5_NIC_IFC_DISABLED		= 1,
 	MLX5_NIC_IFC_NO_DRAM_NIC	= 2,
-	MLX5_NIC_IFC_INVALID		= 3
+	MLX5_NIC_IFC_SW_RESET		= 7
 };
 
 u8 mlx5_get_nic_state(struct mlx5_core_dev *dev);

include/linux/mlx5/device.h

@@ -510,6 +510,10 @@ struct mlx5_cmd_layout {
 	u8		status_own;
 };
 
+enum mlx5_fatal_assert_bit_offsets {
+	MLX5_RFR_OFFSET = 31,
+};
+
 struct health_buffer {
 	__be32		assert_var[5];
 	__be32		rsvd0[3];
@@ -518,12 +522,16 @@ struct health_buffer {
 	__be32		rsvd1[2];
 	__be32		fw_ver;
 	__be32		hw_id;
-	__be32		rsvd2;
+	__be32		rfr;
 	u8		irisc_index;
 	u8		synd;
 	__be16		ext_synd;
 };
 
+enum mlx5_initializing_bit_offsets {
+	MLX5_FW_RESET_SUPPORTED_OFFSET = 30,
+};
+
 enum mlx5_cmd_addr_l_sz_offset {
 	MLX5_NIC_IFC_OFFSET = 8,
 };

include/linux/mlx5/driver.h

@@ -53,6 +53,7 @@
 #include <linux/mlx5/eq.h>
 #include <linux/timecounter.h>
 #include <linux/ptp_clock_kernel.h>
+#include <net/devlink.h>
 
 enum {
 	MLX5_BOARD_ID_LEN = 64,
@@ -434,13 +435,18 @@ struct mlx5_core_health {
 	struct timer_list		timer;
 	u32				prev;
 	int				miss_counter;
-	bool				sick;
+	u8				synd;
+	u32				fatal_error;
+	u32				crdump_size;
 	/* wq spinlock to synchronize draining */
 	spinlock_t			wq_lock;
 	struct workqueue_struct	       *wq;
 	unsigned long			flags;
-	struct work_struct		work;
+	struct work_struct		fatal_report_work;
+	struct work_struct		report_work;
 	struct delayed_work		recover_work;
+	struct devlink_health_reporter *fw_reporter;
+	struct devlink_health_reporter *fw_fatal_reporter;
 };
 
 struct mlx5_qp_table {
@@ -581,6 +587,7 @@ struct mlx5_priv {
 };
 
 enum mlx5_device_state {
+	MLX5_DEVICE_STATE_UNINITIALIZED,
 	MLX5_DEVICE_STATE_UP,
 	MLX5_DEVICE_STATE_INTERNAL_ERROR,
 };
@@ -693,6 +700,7 @@ struct mlx5_core_dev {
 	struct mlx5_clock        clock;
 	struct mlx5_ib_clock_info  *clock_info;
 	struct mlx5_fw_tracer   *tracer;
+	u32                      vsc_addr;
 };
 
 struct mlx5_db {
@@ -904,7 +912,6 @@ void mlx5_start_health_poll(struct mlx5_core_dev *dev);
 void mlx5_stop_health_poll(struct mlx5_core_dev *dev, bool disable_health);
 void mlx5_drain_health_wq(struct mlx5_core_dev *dev);
 void mlx5_trigger_health_work(struct mlx5_core_dev *dev);
-void mlx5_drain_health_recovery(struct mlx5_core_dev *dev);
 int mlx5_buf_alloc_node(struct mlx5_core_dev *dev, int size,
 			struct mlx5_frag_buf *buf, int node);
 int mlx5_buf_alloc(struct mlx5_core_dev *dev,

net/core/devlink.c

@@ -4518,6 +4518,35 @@ static int devlink_fmsg_snd(struct devlink_fmsg *fmsg,
 	return err;
 }
 
+static int devlink_fmsg_dumpit(struct devlink_fmsg *fmsg, struct sk_buff *skb,
+			       struct netlink_callback *cb,
+			       enum devlink_command cmd)
+{
+	int index = cb->args[0];
+	int tmp_index = index;
+	void *hdr;
+	int err;
+
+	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
+			  &devlink_nl_family, NLM_F_ACK | NLM_F_MULTI, cmd);
+	if (!hdr) {
+		err = -EMSGSIZE;
+		goto nla_put_failure;
+	}
+
+	err = devlink_fmsg_prepare_skb(fmsg, skb, &index);
+	if ((err && err != -EMSGSIZE) || tmp_index == index)
+		goto nla_put_failure;
+
+	cb->args[0] = index;
+	genlmsg_end(skb, hdr);
+	return skb->len;
+
+nla_put_failure:
+	genlmsg_cancel(skb, hdr);
+	return err;
+}
+
 struct devlink_health_reporter {
 	struct list_head list;
 	void *priv;
@@ -4750,17 +4779,16 @@ int devlink_health_report(struct devlink_health_reporter *reporter,
 EXPORT_SYMBOL_GPL(devlink_health_report);
 
 static struct devlink_health_reporter *
-devlink_health_reporter_get_from_info(struct devlink *devlink,
-				      struct genl_info *info)
+devlink_health_reporter_get_from_attrs(struct devlink *devlink,
+				       struct nlattr **attrs)
 {
 	struct devlink_health_reporter *reporter;
 	char *reporter_name;
 
-	if (!info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_NAME])
+	if (!attrs[DEVLINK_ATTR_HEALTH_REPORTER_NAME])
 		return NULL;
 
-	reporter_name =
-		nla_data(info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_NAME]);
+	reporter_name = nla_data(attrs[DEVLINK_ATTR_HEALTH_REPORTER_NAME]);
 	mutex_lock(&devlink->reporters_lock);
 	reporter = devlink_health_reporter_find_by_name(devlink, reporter_name);
 	if (reporter)
@@ -4769,6 +4797,48 @@ devlink_health_reporter_get_from_info(struct devlink *devlink,
 	return reporter;
 }
 
+static struct devlink_health_reporter *
+devlink_health_reporter_get_from_info(struct devlink *devlink,
+				      struct genl_info *info)
+{
+	return devlink_health_reporter_get_from_attrs(devlink, info->attrs);
+}
+
+static struct devlink_health_reporter *
+devlink_health_reporter_get_from_cb(struct netlink_callback *cb)
+{
+	struct devlink_health_reporter *reporter;
+	struct devlink *devlink;
+	struct nlattr **attrs;
+	int err;
+
+	attrs = kmalloc_array(DEVLINK_ATTR_MAX + 1, sizeof(*attrs), GFP_KERNEL);
+	if (!attrs)
+		return NULL;
+
+	err = nlmsg_parse_deprecated(cb->nlh,
+				     GENL_HDRLEN + devlink_nl_family.hdrsize,
+				     attrs, DEVLINK_ATTR_MAX,
+				     devlink_nl_family.policy, cb->extack);
+	if (err)
+		goto free;
+
+	mutex_lock(&devlink_mutex);
+	devlink = devlink_get_from_attrs(sock_net(cb->skb->sk), attrs);
+	if (IS_ERR(devlink))
+		goto unlock;
+
+	reporter = devlink_health_reporter_get_from_attrs(devlink, attrs);
+	mutex_unlock(&devlink_mutex);
+	kfree(attrs);
+	return reporter;
+unlock:
+	mutex_unlock(&devlink_mutex);
+free:
+	kfree(attrs);
+	return NULL;
+}
+
 static void
 devlink_health_reporter_put(struct devlink_health_reporter *reporter)
 {
@@ -5004,32 +5074,40 @@ static int devlink_nl_cmd_health_reporter_diagnose_doit(struct sk_buff *skb,
 	return err;
 }
 
-static int devlink_nl_cmd_health_reporter_dump_get_doit(struct sk_buff *skb,
-							struct genl_info *info)
+static int
+devlink_nl_cmd_health_reporter_dump_get_dumpit(struct sk_buff *skb,
+					       struct netlink_callback *cb)
 {
-	struct devlink *devlink = info->user_ptr[0];
 	struct devlink_health_reporter *reporter;
+	u64 start = cb->args[0];
 	int err;
 
-	reporter = devlink_health_reporter_get_from_info(devlink, info);
+	reporter = devlink_health_reporter_get_from_cb(cb);
 	if (!reporter)
 		return -EINVAL;
 
 	if (!reporter->ops->dump) {
-		devlink_health_reporter_put(reporter);
-		return -EOPNOTSUPP;
+		err = -EOPNOTSUPP;
+		goto out;
 	}
-
 	mutex_lock(&reporter->dump_lock);
-	err = devlink_health_do_dump(reporter, NULL);
-	if (err)
-		goto out;
-
-	err = devlink_fmsg_snd(reporter->dump_fmsg, info,
-			       DEVLINK_CMD_HEALTH_REPORTER_DUMP_GET, 0);
+	if (!start) {
+		err = devlink_health_do_dump(reporter, NULL);
+		if (err)
+			goto unlock;
+		cb->args[1] = reporter->dump_ts;
+	}
+	if (!reporter->dump_fmsg || cb->args[1] != reporter->dump_ts) {
+		NL_SET_ERR_MSG_MOD(cb->extack, "Dump trampled, please retry");
+		err = -EAGAIN;
+		goto unlock;
+	}
 
-out:
+	err = devlink_fmsg_dumpit(reporter->dump_fmsg, skb, cb,
+				  DEVLINK_CMD_HEALTH_REPORTER_DUMP_GET);
+unlock:
 	mutex_unlock(&reporter->dump_lock);
+out:
 	devlink_health_reporter_put(reporter);
 	return err;
 }
@@ -5366,7 +5444,7 @@ static const struct genl_ops devlink_nl_ops[] = {
 	{
 		.cmd = DEVLINK_CMD_HEALTH_REPORTER_DUMP_GET,
 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
-		.doit = devlink_nl_cmd_health_reporter_dump_get_doit,
+		.dumpit = devlink_nl_cmd_health_reporter_dump_get_dumpit,
 		.flags = GENL_ADMIN_PERM,
 		.internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK |
 				  DEVLINK_NL_FLAG_NO_LOCK,