| 1 | /* |
| 2 | * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved. |
| 3 | * |
| 4 | * This software is available to you under a choice of one of two |
| 5 | * licenses. You may choose to be licensed under the terms of the GNU |
| 6 | * General Public License (GPL) Version 2, available from the file |
| 7 | * COPYING in the main directory of this source tree, or the |
| 8 | * OpenIB.org BSD license below: |
| 9 | * |
| 10 | * Redistribution and use in source and binary forms, with or |
| 11 | * without modification, are permitted provided that the following |
| 12 | * conditions are met: |
| 13 | * |
| 14 | * - Redistributions of source code must retain the above |
| 15 | * copyright notice, this list of conditions and the following |
| 16 | * disclaimer. |
| 17 | * |
| 18 | * - Redistributions in binary form must reproduce the above |
| 19 | * copyright notice, this list of conditions and the following |
| 20 | * disclaimer in the documentation and/or other materials |
| 21 | * provided with the distribution. |
| 22 | * |
| 23 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, |
| 24 | * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
| 25 | * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND |
| 26 | * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS |
| 27 | * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN |
| 28 | * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN |
| 29 | * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 30 | * SOFTWARE. |
| 31 | */ |
| 32 | |
| 33 | #include <linux/kernel.h> |
| 34 | #include <linux/random.h> |
| 35 | #include <linux/vmalloc.h> |
| 36 | #include <linux/hardirq.h> |
| 37 | #include <linux/mlx5/driver.h> |
| 38 | #include <linux/kern_levels.h> |
| 39 | #include "mlx5_core.h" |
| 40 | #include "lib/eq.h" |
| 41 | #include "lib/mlx5.h" |
| 42 | #include "lib/events.h" |
| 43 | #include "lib/pci_vsc.h" |
| 44 | #include "lib/tout.h" |
| 45 | #include "diag/fw_tracer.h" |
| 46 | #include "diag/reporter_vnic.h" |
| 47 | |
| 48 | enum { |
| 49 | MAX_MISSES = 3, |
| 50 | }; |
| 51 | |
| 52 | enum { |
| 53 | MLX5_DROP_HEALTH_WORK, |
| 54 | }; |
| 55 | |
| 56 | enum { |
| 57 | MLX5_SENSOR_NO_ERR = 0, |
| 58 | MLX5_SENSOR_PCI_COMM_ERR = 1, |
| 59 | MLX5_SENSOR_PCI_ERR = 2, |
| 60 | MLX5_SENSOR_NIC_DISABLED = 3, |
| 61 | MLX5_SENSOR_NIC_SW_RESET = 4, |
| 62 | MLX5_SENSOR_FW_SYND_RFR = 5, |
| 63 | }; |
| 64 | |
| 65 | enum { |
| 66 | MLX5_SEVERITY_MASK = 0x7, |
| 67 | MLX5_SEVERITY_VALID_MASK = 0x8, |
| 68 | }; |
| 69 | |
| 70 | u8 mlx5_get_nic_state(struct mlx5_core_dev *dev) |
| 71 | { |
| 72 | return (ioread32be(&dev->iseg->cmdq_addr_l_sz) >> 8) & 7; |
| 73 | } |
| 74 | |
| 75 | void mlx5_set_nic_state(struct mlx5_core_dev *dev, u8 state) |
| 76 | { |
| 77 | u32 cur_cmdq_addr_l_sz; |
| 78 | |
| 79 | cur_cmdq_addr_l_sz = ioread32be(&dev->iseg->cmdq_addr_l_sz); |
| 80 | iowrite32be((cur_cmdq_addr_l_sz & 0xFFFFF000) | |
| 81 | state << MLX5_NIC_IFC_OFFSET, |
| 82 | &dev->iseg->cmdq_addr_l_sz); |
| 83 | } |
| 84 | |
| 85 | static bool sensor_pci_not_working(struct mlx5_core_dev *dev) |
| 86 | { |
| 87 | struct mlx5_core_health *health = &dev->priv.health; |
| 88 | struct health_buffer __iomem *h = health->health; |
| 89 | |
| 90 | /* Offline PCI reads return 0xffffffff */ |
| 91 | return (ioread32be(&h->fw_ver) == 0xffffffff); |
| 92 | } |
| 93 | |
| 94 | static int mlx5_health_get_rfr(u8 rfr_severity) |
| 95 | { |
| 96 | return rfr_severity >> MLX5_RFR_BIT_OFFSET; |
| 97 | } |
| 98 | |
| 99 | static int mlx5_health_get_crr(u8 rfr_severity) |
| 100 | { |
| 101 | return (rfr_severity >> MLX5_CRR_BIT_OFFSET) & 0x01; |
| 102 | } |
| 103 | |
| 104 | static bool sensor_fw_synd_rfr(struct mlx5_core_dev *dev) |
| 105 | { |
| 106 | struct mlx5_core_health *health = &dev->priv.health; |
| 107 | struct health_buffer __iomem *h = health->health; |
| 108 | u8 synd = ioread8(&h->synd); |
| 109 | u8 rfr; |
| 110 | |
| 111 | rfr = mlx5_health_get_rfr(rfr_severity: ioread8(&h->rfr_severity)); |
| 112 | |
| 113 | if (rfr && synd) |
| 114 | mlx5_core_dbg(dev, "FW requests reset, synd: %d\n" , synd); |
| 115 | return rfr && synd; |
| 116 | } |
| 117 | |
| 118 | u32 mlx5_health_check_fatal_sensors(struct mlx5_core_dev *dev) |
| 119 | { |
| 120 | if (sensor_pci_not_working(dev)) |
| 121 | return MLX5_SENSOR_PCI_COMM_ERR; |
| 122 | if (pci_channel_offline(pdev: dev->pdev)) |
| 123 | return MLX5_SENSOR_PCI_ERR; |
| 124 | if (mlx5_get_nic_state(dev) == MLX5_INITIAL_SEG_NIC_INTERFACE_DISABLED) |
| 125 | return MLX5_SENSOR_NIC_DISABLED; |
| 126 | if (mlx5_get_nic_state(dev) == MLX5_INITIAL_SEG_NIC_INTERFACE_SW_RESET) |
| 127 | return MLX5_SENSOR_NIC_SW_RESET; |
| 128 | if (sensor_fw_synd_rfr(dev)) |
| 129 | return MLX5_SENSOR_FW_SYND_RFR; |
| 130 | |
| 131 | return MLX5_SENSOR_NO_ERR; |
| 132 | } |
| 133 | |
| 134 | static int lock_sem_sw_reset(struct mlx5_core_dev *dev, bool lock) |
| 135 | { |
| 136 | enum mlx5_vsc_state state; |
| 137 | int ret; |
| 138 | |
| 139 | if (!mlx5_core_is_pf(dev)) |
| 140 | return -EBUSY; |
| 141 | |
| 142 | /* Try to lock GW access, this stage doesn't return |
| 143 | * EBUSY because locked GW does not mean that other PF |
| 144 | * already started the reset. |
| 145 | */ |
| 146 | ret = mlx5_vsc_gw_lock(dev); |
| 147 | if (ret == -EBUSY) |
| 148 | return -EINVAL; |
| 149 | if (ret) |
| 150 | return ret; |
| 151 | |
| 152 | state = lock ? MLX5_VSC_LOCK : MLX5_VSC_UNLOCK; |
| 153 | /* At this stage, if the return status == EBUSY, then we know |
| 154 | * for sure that another PF started the reset, so don't allow |
| 155 | * another reset. |
| 156 | */ |
| 157 | ret = mlx5_vsc_sem_set_space(dev, space: MLX5_SEMAPHORE_SW_RESET, state); |
| 158 | if (ret) |
| 159 | mlx5_core_warn(dev, "Failed to lock SW reset semaphore\n" ); |
| 160 | |
| 161 | /* Unlock GW access */ |
| 162 | mlx5_vsc_gw_unlock(dev); |
| 163 | |
| 164 | return ret; |
| 165 | } |
| 166 | |
| 167 | static bool reset_fw_if_needed(struct mlx5_core_dev *dev) |
| 168 | { |
| 169 | bool supported = (ioread32be(&dev->iseg->initializing) >> |
| 170 | MLX5_FW_RESET_SUPPORTED_OFFSET) & 1; |
| 171 | u32 fatal_error; |
| 172 | |
| 173 | if (!supported) |
| 174 | return false; |
| 175 | |
| 176 | /* The reset only needs to be issued by one PF. The health buffer is |
| 177 | * shared between all functions, and will be cleared during a reset. |
| 178 | * Check again to avoid a redundant 2nd reset. If the fatal errors was |
| 179 | * PCI related a reset won't help. |
| 180 | */ |
| 181 | fatal_error = mlx5_health_check_fatal_sensors(dev); |
| 182 | if (fatal_error == MLX5_SENSOR_PCI_COMM_ERR || |
| 183 | fatal_error == MLX5_SENSOR_NIC_DISABLED || |
| 184 | fatal_error == MLX5_SENSOR_NIC_SW_RESET) { |
| 185 | mlx5_core_warn(dev, "Not issuing FW reset. Either it's already done or won't help." ); |
| 186 | return false; |
| 187 | } |
| 188 | |
| 189 | mlx5_core_warn(dev, "Issuing FW Reset\n" ); |
| 190 | /* Write the NIC interface field to initiate the reset, the command |
| 191 | * interface address also resides here, don't overwrite it. |
| 192 | */ |
| 193 | mlx5_set_nic_state(dev, state: MLX5_INITIAL_SEG_NIC_INTERFACE_SW_RESET); |
| 194 | |
| 195 | return true; |
| 196 | } |
| 197 | |
| 198 | static void enter_error_state(struct mlx5_core_dev *dev, bool force) |
| 199 | { |
| 200 | if (mlx5_health_check_fatal_sensors(dev) || force) { /* protected state setting */ |
| 201 | dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR; |
| 202 | mlx5_cmd_flush(dev); |
| 203 | } |
| 204 | |
| 205 | mlx5_notifier_call_chain(events: dev->priv.events, event: MLX5_DEV_EVENT_SYS_ERROR, data: (void *)1); |
| 206 | } |
| 207 | |
| 208 | void mlx5_enter_error_state(struct mlx5_core_dev *dev, bool force) |
| 209 | { |
| 210 | bool err_detected = false; |
| 211 | |
| 212 | /* Mark the device as fatal in order to abort FW commands */ |
| 213 | if ((mlx5_health_check_fatal_sensors(dev) || force) && |
| 214 | dev->state == MLX5_DEVICE_STATE_UP) { |
| 215 | dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR; |
| 216 | err_detected = true; |
| 217 | } |
| 218 | mutex_lock(&dev->intf_state_mutex); |
| 219 | if (!err_detected && dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) |
| 220 | goto unlock;/* a previous error is still being handled */ |
| 221 | |
| 222 | enter_error_state(dev, force); |
| 223 | unlock: |
| 224 | mutex_unlock(lock: &dev->intf_state_mutex); |
| 225 | } |
| 226 | |
| 227 | void mlx5_error_sw_reset(struct mlx5_core_dev *dev) |
| 228 | { |
| 229 | unsigned long end, delay_ms = mlx5_tout_ms(dev, PCI_TOGGLE); |
| 230 | int lock = -EBUSY; |
| 231 | |
| 232 | mutex_lock(&dev->intf_state_mutex); |
| 233 | if (dev->state != MLX5_DEVICE_STATE_INTERNAL_ERROR) |
| 234 | goto unlock; |
| 235 | |
| 236 | mlx5_core_err(dev, "start\n" ); |
| 237 | |
| 238 | if (mlx5_health_check_fatal_sensors(dev) == MLX5_SENSOR_FW_SYND_RFR) { |
| 239 | /* Get cr-dump and reset FW semaphore */ |
| 240 | lock = lock_sem_sw_reset(dev, lock: true); |
| 241 | |
| 242 | if (lock == -EBUSY) { |
| 243 | delay_ms = mlx5_tout_ms(dev, FULL_CRDUMP); |
| 244 | goto recover_from_sw_reset; |
| 245 | } |
| 246 | /* Execute SW reset */ |
| 247 | reset_fw_if_needed(dev); |
| 248 | } |
| 249 | |
| 250 | recover_from_sw_reset: |
| 251 | /* Recover from SW reset */ |
| 252 | end = jiffies + msecs_to_jiffies(m: delay_ms); |
| 253 | do { |
| 254 | if (mlx5_get_nic_state(dev) == MLX5_INITIAL_SEG_NIC_INTERFACE_DISABLED) |
| 255 | break; |
| 256 | if (pci_channel_offline(pdev: dev->pdev)) { |
| 257 | mlx5_core_err(dev, "PCI channel offline, stop waiting for NIC IFC\n" ); |
| 258 | goto unlock; |
| 259 | } |
| 260 | |
| 261 | msleep(msecs: 20); |
| 262 | } while (!time_after(jiffies, end)); |
| 263 | |
| 264 | if (mlx5_get_nic_state(dev) != MLX5_INITIAL_SEG_NIC_INTERFACE_DISABLED) { |
| 265 | dev_err(&dev->pdev->dev, "NIC IFC still %d after %lums.\n" , |
| 266 | mlx5_get_nic_state(dev), delay_ms); |
| 267 | } |
| 268 | |
| 269 | /* Release FW semaphore if you are the lock owner */ |
| 270 | if (!lock) |
| 271 | lock_sem_sw_reset(dev, lock: false); |
| 272 | |
| 273 | mlx5_core_err(dev, "end\n" ); |
| 274 | |
| 275 | unlock: |
| 276 | mutex_unlock(lock: &dev->intf_state_mutex); |
| 277 | } |
| 278 | |
| 279 | static void mlx5_handle_bad_state(struct mlx5_core_dev *dev) |
| 280 | { |
| 281 | u8 nic_interface = mlx5_get_nic_state(dev); |
| 282 | |
| 283 | switch (nic_interface) { |
| 284 | case MLX5_INITIAL_SEG_NIC_INTERFACE_FULL_DRIVER: |
| 285 | mlx5_core_warn(dev, "Expected to see disabled NIC but it is full driver\n" ); |
| 286 | break; |
| 287 | |
| 288 | case MLX5_INITIAL_SEG_NIC_INTERFACE_DISABLED: |
| 289 | mlx5_core_warn(dev, "starting teardown\n" ); |
| 290 | break; |
| 291 | |
| 292 | case MLX5_INITIAL_SEG_NIC_INTERFACE_NO_DRAM_NIC: |
| 293 | mlx5_core_warn(dev, "Expected to see disabled NIC but it is no dram nic\n" ); |
| 294 | break; |
| 295 | |
| 296 | case MLX5_INITIAL_SEG_NIC_INTERFACE_SW_RESET: |
| 297 | /* The IFC mode field is 3 bits, so it will read 0x7 in 2 cases: |
| 298 | * 1. PCI has been disabled (ie. PCI-AER, PF driver unloaded |
| 299 | * and this is a VF), this is not recoverable by SW reset. |
| 300 | * Logging of this is handled elsewhere. |
| 301 | * 2. FW reset has been issued by another function, driver can |
| 302 | * be reloaded to recover after the mode switches to |
| 303 | * MLX5_INITIAL_SEG_NIC_INTERFACE_DISABLED. |
| 304 | */ |
| 305 | if (dev->priv.health.fatal_error != MLX5_SENSOR_PCI_COMM_ERR) |
| 306 | mlx5_core_warn(dev, "NIC SW reset in progress\n" ); |
| 307 | break; |
| 308 | |
| 309 | default: |
| 310 | mlx5_core_warn(dev, "Expected to see disabled NIC but it is has invalid value %d\n" , |
| 311 | nic_interface); |
| 312 | } |
| 313 | |
| 314 | mlx5_disable_device(dev); |
| 315 | } |
| 316 | |
| 317 | int mlx5_health_wait_pci_up(struct mlx5_core_dev *dev) |
| 318 | { |
| 319 | unsigned long end; |
| 320 | |
| 321 | end = jiffies + msecs_to_jiffies(mlx5_tout_ms(dev, FW_RESET)); |
| 322 | while (sensor_pci_not_working(dev)) { |
| 323 | if (time_after(jiffies, end)) |
| 324 | return -ETIMEDOUT; |
| 325 | if (test_bit(MLX5_BREAK_FW_WAIT, &dev->intf_state)) { |
| 326 | mlx5_core_warn(dev, "device is being removed, stop waiting for PCI\n" ); |
| 327 | return -ENODEV; |
| 328 | } |
| 329 | if (pci_channel_offline(pdev: dev->pdev)) { |
| 330 | mlx5_core_err(dev, "PCI channel offline, stop waiting for PCI\n" ); |
| 331 | return -EACCES; |
| 332 | } |
| 333 | msleep(msecs: 100); |
| 334 | } |
| 335 | return 0; |
| 336 | } |
| 337 | |
| 338 | static int mlx5_health_try_recover(struct mlx5_core_dev *dev) |
| 339 | { |
| 340 | mlx5_core_warn(dev, "handling bad device here\n" ); |
| 341 | mlx5_handle_bad_state(dev); |
| 342 | if (mlx5_health_wait_pci_up(dev)) { |
| 343 | mlx5_core_err(dev, "health recovery flow aborted, PCI reads still not working\n" ); |
| 344 | return -EIO; |
| 345 | } |
| 346 | mlx5_core_err(dev, "starting health recovery flow\n" ); |
| 347 | if (mlx5_recover_device(dev) || mlx5_health_check_fatal_sensors(dev)) { |
| 348 | mlx5_core_err(dev, "health recovery failed\n" ); |
| 349 | return -EIO; |
| 350 | } |
| 351 | |
| 352 | mlx5_core_info(dev, "health recovery succeeded\n" ); |
| 353 | return 0; |
| 354 | } |
| 355 | |
| 356 | static const char *hsynd_str(u8 synd) |
| 357 | { |
| 358 | switch (synd) { |
| 359 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_FW_INTERNAL_ERR: |
| 360 | return "firmware internal error" ; |
| 361 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_DEAD_IRISC: |
| 362 | return "irisc not responding" ; |
| 363 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_HW_FATAL_ERR: |
| 364 | return "unrecoverable hardware error" ; |
| 365 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_FW_CRC_ERR: |
| 366 | return "firmware CRC error" ; |
| 367 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_ICM_FETCH_PCI_ERR: |
| 368 | return "ICM fetch PCI error" ; |
| 369 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_ICM_PAGE_ERR: |
| 370 | return "HW fatal error\n" ; |
| 371 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_ASYNCHRONOUS_EQ_BUF_OVERRUN: |
| 372 | return "async EQ buffer overrun" ; |
| 373 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_EQ_IN_ERR: |
| 374 | return "EQ error" ; |
| 375 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_EQ_INV: |
| 376 | return "Invalid EQ referenced" ; |
| 377 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_FFSER_ERR: |
| 378 | return "FFSER error" ; |
| 379 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_HIGH_TEMP_ERR: |
| 380 | return "High temperature" ; |
| 381 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_ICM_PCI_POISONED_ERR: |
| 382 | return "ICM fetch PCI data poisoned error" ; |
| 383 | case MLX5_INITIAL_SEG_HEALTH_SYNDROME_TRUST_LOCKDOWN_ERR: |
| 384 | return "Trust lockdown error" ; |
| 385 | default: |
| 386 | return "unrecognized error" ; |
| 387 | } |
| 388 | } |
| 389 | |
| 390 | static const char *mlx5_loglevel_str(int level) |
| 391 | { |
| 392 | switch (level) { |
| 393 | case LOGLEVEL_EMERG: |
| 394 | return "EMERGENCY" ; |
| 395 | case LOGLEVEL_ALERT: |
| 396 | return "ALERT" ; |
| 397 | case LOGLEVEL_CRIT: |
| 398 | return "CRITICAL" ; |
| 399 | case LOGLEVEL_ERR: |
| 400 | return "ERROR" ; |
| 401 | case LOGLEVEL_WARNING: |
| 402 | return "WARNING" ; |
| 403 | case LOGLEVEL_NOTICE: |
| 404 | return "NOTICE" ; |
| 405 | case LOGLEVEL_INFO: |
| 406 | return "INFO" ; |
| 407 | case LOGLEVEL_DEBUG: |
| 408 | return "DEBUG" ; |
| 409 | } |
| 410 | return "Unknown log level" ; |
| 411 | } |
| 412 | |
| 413 | static int mlx5_health_get_severity(u8 rfr_severity) |
| 414 | { |
| 415 | return rfr_severity & MLX5_SEVERITY_VALID_MASK ? |
| 416 | rfr_severity & MLX5_SEVERITY_MASK : LOGLEVEL_ERR; |
| 417 | } |
| 418 | |
| 419 | static void print_health_info(struct mlx5_core_dev *dev) |
| 420 | { |
| 421 | struct mlx5_core_health *health = &dev->priv.health; |
| 422 | struct health_buffer __iomem *h = health->health; |
| 423 | u8 rfr_severity; |
| 424 | int severity; |
| 425 | int i; |
| 426 | |
| 427 | /* If the syndrome is 0, the device is OK and no need to print buffer */ |
| 428 | if (!ioread8(&h->synd)) |
| 429 | return; |
| 430 | |
| 431 | if (ioread32be(&h->fw_ver) == 0xFFFFFFFF) { |
| 432 | mlx5_log(dev, LOGLEVEL_ERR, "PCI slot is unavailable\n" ); |
| 433 | return; |
| 434 | } |
| 435 | |
| 436 | rfr_severity = ioread8(&h->rfr_severity); |
| 437 | severity = mlx5_health_get_severity(rfr_severity); |
| 438 | mlx5_log(dev, severity, "Health issue observed, %s, severity(%d) %s:\n" , |
| 439 | hsynd_str(ioread8(&h->synd)), severity, mlx5_loglevel_str(severity)); |
| 440 | |
| 441 | for (i = 0; i < ARRAY_SIZE(h->assert_var); i++) |
| 442 | mlx5_log(dev, severity, "assert_var[%d] 0x%08x\n" , i, |
| 443 | ioread32be(h->assert_var + i)); |
| 444 | |
| 445 | mlx5_log(dev, severity, "assert_exit_ptr 0x%08x\n" , ioread32be(&h->assert_exit_ptr)); |
| 446 | mlx5_log(dev, severity, "assert_callra 0x%08x\n" , ioread32be(&h->assert_callra)); |
| 447 | mlx5_log(dev, severity, "fw_ver %d.%d.%d" , fw_rev_maj(dev), fw_rev_min(dev), |
| 448 | fw_rev_sub(dev)); |
| 449 | mlx5_log(dev, severity, "time %u\n" , ioread32be(&h->time)); |
| 450 | mlx5_log(dev, severity, "hw_id 0x%08x\n" , ioread32be(&h->hw_id)); |
| 451 | mlx5_log(dev, severity, "rfr %d\n" , mlx5_health_get_rfr(rfr_severity)); |
| 452 | mlx5_log(dev, severity, "crr %d\n" , mlx5_health_get_crr(rfr_severity)); |
| 453 | mlx5_log(dev, severity, "severity %d (%s)\n" , severity, mlx5_loglevel_str(severity)); |
| 454 | mlx5_log(dev, severity, "irisc_index %d\n" , ioread8(&h->irisc_index)); |
| 455 | mlx5_log(dev, severity, "synd 0x%x: %s\n" , ioread8(&h->synd), |
| 456 | hsynd_str(ioread8(&h->synd))); |
| 457 | mlx5_log(dev, severity, "ext_synd 0x%04x\n" , ioread16be(&h->ext_synd)); |
| 458 | mlx5_log(dev, severity, "raw fw_ver 0x%08x\n" , ioread32be(&h->fw_ver)); |
| 459 | if (mlx5_health_get_crr(rfr_severity)) |
| 460 | mlx5_core_warn(dev, "Cold reset is required\n" ); |
| 461 | } |
| 462 | |
| 463 | static int |
| 464 | mlx5_fw_reporter_diagnose(struct devlink_health_reporter *reporter, |
| 465 | struct devlink_fmsg *fmsg, |
| 466 | struct netlink_ext_ack *extack) |
| 467 | { |
| 468 | struct mlx5_core_dev *dev = devlink_health_reporter_priv(reporter); |
| 469 | struct mlx5_core_health *health = &dev->priv.health; |
| 470 | struct health_buffer __iomem *h = health->health; |
| 471 | u8 synd = ioread8(&h->synd); |
| 472 | |
| 473 | devlink_fmsg_u8_pair_put(fmsg, name: "Syndrome" , value: synd); |
| 474 | if (!synd) |
| 475 | return 0; |
| 476 | |
| 477 | devlink_fmsg_string_pair_put(fmsg, name: "Description" , value: hsynd_str(synd)); |
| 478 | |
| 479 | return 0; |
| 480 | } |
| 481 | |
| 482 | struct mlx5_fw_reporter_ctx { |
| 483 | u8 err_synd; |
| 484 | int miss_counter; |
| 485 | }; |
| 486 | |
| 487 | static void |
| 488 | mlx5_fw_reporter_ctx_pairs_put(struct devlink_fmsg *fmsg, |
| 489 | struct mlx5_fw_reporter_ctx *fw_reporter_ctx) |
| 490 | { |
| 491 | devlink_fmsg_u8_pair_put(fmsg, name: "syndrome" , value: fw_reporter_ctx->err_synd); |
| 492 | devlink_fmsg_u32_pair_put(fmsg, name: "fw_miss_counter" , value: fw_reporter_ctx->miss_counter); |
| 493 | } |
| 494 | |
| 495 | static void |
| 496 | mlx5_fw_reporter_heath_buffer_data_put(struct mlx5_core_dev *dev, |
| 497 | struct devlink_fmsg *fmsg) |
| 498 | { |
| 499 | struct mlx5_core_health *health = &dev->priv.health; |
| 500 | struct health_buffer __iomem *h = health->health; |
| 501 | u8 rfr_severity; |
| 502 | int i; |
| 503 | |
| 504 | if (!ioread8(&h->synd)) |
| 505 | return; |
| 506 | |
| 507 | devlink_fmsg_pair_nest_start(fmsg, name: "health buffer" ); |
| 508 | devlink_fmsg_obj_nest_start(fmsg); |
| 509 | devlink_fmsg_arr_pair_nest_start(fmsg, name: "assert_var" ); |
| 510 | for (i = 0; i < ARRAY_SIZE(h->assert_var); i++) |
| 511 | devlink_fmsg_u32_put(fmsg, value: ioread32be(h->assert_var + i)); |
| 512 | devlink_fmsg_arr_pair_nest_end(fmsg); |
| 513 | devlink_fmsg_u32_pair_put(fmsg, name: "assert_exit_ptr" , |
| 514 | value: ioread32be(&h->assert_exit_ptr)); |
| 515 | devlink_fmsg_u32_pair_put(fmsg, name: "assert_callra" , |
| 516 | value: ioread32be(&h->assert_callra)); |
| 517 | devlink_fmsg_u32_pair_put(fmsg, name: "time" , value: ioread32be(&h->time)); |
| 518 | devlink_fmsg_u32_pair_put(fmsg, name: "hw_id" , value: ioread32be(&h->hw_id)); |
| 519 | rfr_severity = ioread8(&h->rfr_severity); |
| 520 | devlink_fmsg_u8_pair_put(fmsg, name: "rfr" , value: mlx5_health_get_rfr(rfr_severity)); |
| 521 | devlink_fmsg_u8_pair_put(fmsg, name: "severity" , value: mlx5_health_get_severity(rfr_severity)); |
| 522 | devlink_fmsg_u8_pair_put(fmsg, name: "irisc_index" , value: ioread8(&h->irisc_index)); |
| 523 | devlink_fmsg_u8_pair_put(fmsg, name: "synd" , value: ioread8(&h->synd)); |
| 524 | devlink_fmsg_u32_pair_put(fmsg, name: "ext_synd" , value: ioread16be(&h->ext_synd)); |
| 525 | devlink_fmsg_u32_pair_put(fmsg, name: "raw_fw_ver" , value: ioread32be(&h->fw_ver)); |
| 526 | devlink_fmsg_obj_nest_end(fmsg); |
| 527 | devlink_fmsg_pair_nest_end(fmsg); |
| 528 | } |
| 529 | |
| 530 | static int |
| 531 | mlx5_fw_reporter_dump(struct devlink_health_reporter *reporter, |
| 532 | struct devlink_fmsg *fmsg, void *priv_ctx, |
| 533 | struct netlink_ext_ack *extack) |
| 534 | { |
| 535 | struct mlx5_core_dev *dev = devlink_health_reporter_priv(reporter); |
| 536 | int err; |
| 537 | |
| 538 | err = mlx5_fw_tracer_trigger_core_dump_general(dev); |
| 539 | if (err) |
| 540 | return err; |
| 541 | |
| 542 | if (priv_ctx) { |
| 543 | struct mlx5_fw_reporter_ctx *fw_reporter_ctx = priv_ctx; |
| 544 | |
| 545 | mlx5_fw_reporter_ctx_pairs_put(fmsg, fw_reporter_ctx); |
| 546 | } |
| 547 | |
| 548 | mlx5_fw_reporter_heath_buffer_data_put(dev, fmsg); |
| 549 | |
| 550 | return mlx5_fw_tracer_get_saved_traces_objects(tracer: dev->tracer, fmsg); |
| 551 | } |
| 552 | |
| 553 | static void mlx5_fw_reporter_err_work(struct work_struct *work) |
| 554 | { |
| 555 | struct mlx5_fw_reporter_ctx fw_reporter_ctx; |
| 556 | struct mlx5_core_health *health; |
| 557 | |
| 558 | health = container_of(work, struct mlx5_core_health, report_work); |
| 559 | |
| 560 | if (IS_ERR_OR_NULL(ptr: health->fw_reporter)) |
| 561 | return; |
| 562 | |
| 563 | fw_reporter_ctx.err_synd = health->synd; |
| 564 | fw_reporter_ctx.miss_counter = health->miss_counter; |
| 565 | if (fw_reporter_ctx.err_synd) { |
| 566 | devlink_health_report(reporter: health->fw_reporter, |
| 567 | msg: "FW syndrome reported" , priv_ctx: &fw_reporter_ctx); |
| 568 | return; |
| 569 | } |
| 570 | if (fw_reporter_ctx.miss_counter) |
| 571 | devlink_health_report(reporter: health->fw_reporter, |
| 572 | msg: "FW miss counter reported" , |
| 573 | priv_ctx: &fw_reporter_ctx); |
| 574 | } |
| 575 | |
| 576 | static const struct devlink_health_reporter_ops mlx5_fw_reporter_pf_ops = { |
| 577 | .name = "fw" , |
| 578 | .diagnose = mlx5_fw_reporter_diagnose, |
| 579 | .dump = mlx5_fw_reporter_dump, |
| 580 | }; |
| 581 | |
| 582 | static const struct devlink_health_reporter_ops mlx5_fw_reporter_ops = { |
| 583 | .name = "fw" , |
| 584 | .diagnose = mlx5_fw_reporter_diagnose, |
| 585 | }; |
| 586 | |
| 587 | static int |
| 588 | mlx5_fw_fatal_reporter_recover(struct devlink_health_reporter *reporter, |
| 589 | void *priv_ctx, |
| 590 | struct netlink_ext_ack *extack) |
| 591 | { |
| 592 | struct mlx5_core_dev *dev = devlink_health_reporter_priv(reporter); |
| 593 | |
| 594 | return mlx5_health_try_recover(dev); |
| 595 | } |
| 596 | |
| 597 | static int |
| 598 | mlx5_fw_fatal_reporter_dump(struct devlink_health_reporter *reporter, |
| 599 | struct devlink_fmsg *fmsg, void *priv_ctx, |
| 600 | struct netlink_ext_ack *extack) |
| 601 | { |
| 602 | struct mlx5_core_dev *dev = devlink_health_reporter_priv(reporter); |
| 603 | u32 crdump_size = dev->priv.health.crdump_size; |
| 604 | u32 *cr_data; |
| 605 | int err; |
| 606 | |
| 607 | if (!mlx5_core_is_pf(dev)) |
| 608 | return -EPERM; |
| 609 | |
| 610 | cr_data = kvmalloc(crdump_size, GFP_KERNEL); |
| 611 | if (!cr_data) |
| 612 | return -ENOMEM; |
| 613 | err = mlx5_crdump_collect(dev, cr_data); |
| 614 | if (err) |
| 615 | goto free_data; |
| 616 | |
| 617 | if (priv_ctx) { |
| 618 | struct mlx5_fw_reporter_ctx *fw_reporter_ctx = priv_ctx; |
| 619 | |
| 620 | mlx5_fw_reporter_ctx_pairs_put(fmsg, fw_reporter_ctx); |
| 621 | } |
| 622 | |
| 623 | devlink_fmsg_binary_pair_put(fmsg, name: "crdump_data" , value: cr_data, value_len: crdump_size); |
| 624 | |
| 625 | free_data: |
| 626 | kvfree(addr: cr_data); |
| 627 | return err; |
| 628 | } |
| 629 | |
| 630 | static void mlx5_fw_fatal_reporter_err_work(struct work_struct *work) |
| 631 | { |
| 632 | struct mlx5_fw_reporter_ctx fw_reporter_ctx; |
| 633 | struct mlx5_core_health *health; |
| 634 | struct mlx5_core_dev *dev; |
| 635 | struct devlink *devlink; |
| 636 | struct mlx5_priv *priv; |
| 637 | |
| 638 | health = container_of(work, struct mlx5_core_health, fatal_report_work); |
| 639 | priv = container_of(health, struct mlx5_priv, health); |
| 640 | dev = container_of(priv, struct mlx5_core_dev, priv); |
| 641 | devlink = priv_to_devlink(priv: dev); |
| 642 | |
| 643 | mutex_lock(&dev->intf_state_mutex); |
| 644 | if (test_bit(MLX5_DROP_HEALTH_WORK, &health->flags)) { |
| 645 | mlx5_core_err(dev, "health works are not permitted at this stage\n" ); |
| 646 | mutex_unlock(lock: &dev->intf_state_mutex); |
| 647 | return; |
| 648 | } |
| 649 | mutex_unlock(lock: &dev->intf_state_mutex); |
| 650 | enter_error_state(dev, force: false); |
| 651 | if (IS_ERR_OR_NULL(ptr: health->fw_fatal_reporter)) { |
| 652 | devl_lock(devlink); |
| 653 | if (mlx5_health_try_recover(dev)) |
| 654 | mlx5_core_err(dev, "health recovery failed\n" ); |
| 655 | devl_unlock(devlink); |
| 656 | return; |
| 657 | } |
| 658 | fw_reporter_ctx.err_synd = health->synd; |
| 659 | fw_reporter_ctx.miss_counter = health->miss_counter; |
| 660 | if (devlink_health_report(reporter: health->fw_fatal_reporter, |
| 661 | msg: "FW fatal error reported" , priv_ctx: &fw_reporter_ctx) == -ECANCELED) { |
| 662 | /* If recovery wasn't performed, due to grace period, |
| 663 | * unload the driver. This ensures that the driver |
| 664 | * closes all its resources and it is not subjected to |
| 665 | * requests from the kernel. |
| 666 | */ |
| 667 | mlx5_core_err(dev, "Driver is in error state. Unloading\n" ); |
| 668 | mlx5_unload_one(dev, suspend: false); |
| 669 | } |
| 670 | } |
| 671 | |
| 672 | #define MLX5_FW_REPORTER_ECPF_GRACEFUL_PERIOD 180000 |
| 673 | #define MLX5_FW_REPORTER_PF_GRACEFUL_PERIOD 60000 |
| 674 | #define MLX5_FW_REPORTER_VF_GRACEFUL_PERIOD 30000 |
| 675 | #define MLX5_FW_REPORTER_DEFAULT_GRACEFUL_PERIOD \ |
| 676 | MLX5_FW_REPORTER_VF_GRACEFUL_PERIOD |
| 677 | |
| 678 | static |
| 679 | const struct devlink_health_reporter_ops mlx5_fw_fatal_reporter_ecpf_ops = { |
| 680 | .name = "fw_fatal" , |
| 681 | .recover = mlx5_fw_fatal_reporter_recover, |
| 682 | .dump = mlx5_fw_fatal_reporter_dump, |
| 683 | .default_graceful_period = |
| 684 | MLX5_FW_REPORTER_ECPF_GRACEFUL_PERIOD, |
| 685 | }; |
| 686 | |
| 687 | static const struct devlink_health_reporter_ops mlx5_fw_fatal_reporter_pf_ops = { |
| 688 | .name = "fw_fatal" , |
| 689 | .recover = mlx5_fw_fatal_reporter_recover, |
| 690 | .dump = mlx5_fw_fatal_reporter_dump, |
| 691 | .default_graceful_period = MLX5_FW_REPORTER_PF_GRACEFUL_PERIOD, |
| 692 | }; |
| 693 | |
| 694 | static const struct devlink_health_reporter_ops mlx5_fw_fatal_reporter_ops = { |
| 695 | .name = "fw_fatal" , |
| 696 | .recover = mlx5_fw_fatal_reporter_recover, |
| 697 | .default_graceful_period = |
| 698 | MLX5_FW_REPORTER_DEFAULT_GRACEFUL_PERIOD, |
| 699 | }; |
| 700 | |
| 701 | void mlx5_fw_reporters_create(struct mlx5_core_dev *dev) |
| 702 | { |
| 703 | const struct devlink_health_reporter_ops *fw_fatal_ops; |
| 704 | struct mlx5_core_health *health = &dev->priv.health; |
| 705 | const struct devlink_health_reporter_ops *fw_ops; |
| 706 | struct devlink *devlink = priv_to_devlink(priv: dev); |
| 707 | |
| 708 | fw_ops = &mlx5_fw_reporter_pf_ops; |
| 709 | if (mlx5_core_is_ecpf(dev)) { |
| 710 | fw_fatal_ops = &mlx5_fw_fatal_reporter_ecpf_ops; |
| 711 | } else if (mlx5_core_is_pf(dev)) { |
| 712 | fw_fatal_ops = &mlx5_fw_fatal_reporter_pf_ops; |
| 713 | } else { |
| 714 | /* VF or SF */ |
| 715 | fw_fatal_ops = &mlx5_fw_fatal_reporter_ops; |
| 716 | fw_ops = &mlx5_fw_reporter_ops; |
| 717 | } |
| 718 | |
| 719 | health->fw_reporter = devl_health_reporter_create(devlink, ops: fw_ops, priv: dev); |
| 720 | if (IS_ERR(ptr: health->fw_reporter)) |
| 721 | mlx5_core_warn(dev, "Failed to create fw reporter, err = %pe\n" , |
| 722 | health->fw_reporter); |
| 723 | |
| 724 | health->fw_fatal_reporter = devl_health_reporter_create(devlink, |
| 725 | ops: fw_fatal_ops, |
| 726 | priv: dev); |
| 727 | if (IS_ERR(ptr: health->fw_fatal_reporter)) |
| 728 | mlx5_core_warn(dev, "Failed to create fw fatal reporter, err = %pe\n" , |
| 729 | health->fw_fatal_reporter); |
| 730 | } |
| 731 | |
| 732 | static void mlx5_fw_reporters_destroy(struct mlx5_core_dev *dev) |
| 733 | { |
| 734 | struct mlx5_core_health *health = &dev->priv.health; |
| 735 | |
| 736 | if (!IS_ERR_OR_NULL(ptr: health->fw_reporter)) |
| 737 | devlink_health_reporter_destroy(reporter: health->fw_reporter); |
| 738 | |
| 739 | if (!IS_ERR_OR_NULL(ptr: health->fw_fatal_reporter)) |
| 740 | devlink_health_reporter_destroy(reporter: health->fw_fatal_reporter); |
| 741 | } |
| 742 | |
| 743 | static unsigned long get_next_poll_jiffies(struct mlx5_core_dev *dev) |
| 744 | { |
| 745 | unsigned long next; |
| 746 | |
| 747 | get_random_bytes(buf: &next, len: sizeof(next)); |
| 748 | next %= HZ; |
| 749 | next += jiffies + msecs_to_jiffies(mlx5_tout_ms(dev, HEALTH_POLL_INTERVAL)); |
| 750 | |
| 751 | return next; |
| 752 | } |
| 753 | |
| 754 | void mlx5_trigger_health_work(struct mlx5_core_dev *dev) |
| 755 | { |
| 756 | struct mlx5_core_health *health = &dev->priv.health; |
| 757 | |
| 758 | if (!mlx5_dev_is_lightweight(dev)) |
| 759 | queue_work(wq: health->wq, work: &health->fatal_report_work); |
| 760 | } |
| 761 | |
| 762 | #define MLX5_MSEC_PER_HOUR (MSEC_PER_SEC * 60 * 60) |
| 763 | static void mlx5_health_log_ts_update(struct work_struct *work) |
| 764 | { |
| 765 | struct delayed_work *dwork = to_delayed_work(work); |
| 766 | u32 out[MLX5_ST_SZ_DW(mrtc_reg)] = {}; |
| 767 | u32 in[MLX5_ST_SZ_DW(mrtc_reg)] = {}; |
| 768 | struct mlx5_core_health *health; |
| 769 | struct mlx5_core_dev *dev; |
| 770 | struct mlx5_priv *priv; |
| 771 | u64 now_us; |
| 772 | |
| 773 | health = container_of(dwork, struct mlx5_core_health, update_fw_log_ts_work); |
| 774 | priv = container_of(health, struct mlx5_priv, health); |
| 775 | dev = container_of(priv, struct mlx5_core_dev, priv); |
| 776 | |
| 777 | now_us = ktime_to_us(kt: ktime_get_real()); |
| 778 | |
| 779 | MLX5_SET(mrtc_reg, in, time_h, now_us >> 32); |
| 780 | MLX5_SET(mrtc_reg, in, time_l, now_us & 0xFFFFFFFF); |
| 781 | mlx5_core_access_reg(dev, data_in: in, size_in: sizeof(in), data_out: out, size_out: sizeof(out), reg_num: MLX5_REG_MRTC, arg: 0, write: 1); |
| 782 | |
| 783 | queue_delayed_work(wq: health->wq, dwork: &health->update_fw_log_ts_work, |
| 784 | delay: msecs_to_jiffies(MLX5_MSEC_PER_HOUR)); |
| 785 | } |
| 786 | |
| 787 | static void poll_health(struct timer_list *t) |
| 788 | { |
| 789 | struct mlx5_core_dev *dev = timer_container_of(dev, t, |
| 790 | priv.health.timer); |
| 791 | struct mlx5_core_health *health = &dev->priv.health; |
| 792 | struct health_buffer __iomem *h = health->health; |
| 793 | u32 fatal_error; |
| 794 | u8 prev_synd; |
| 795 | u32 count; |
| 796 | |
| 797 | if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) |
| 798 | goto out; |
| 799 | |
| 800 | fatal_error = mlx5_health_check_fatal_sensors(dev); |
| 801 | |
| 802 | if (fatal_error && !health->fatal_error) { |
| 803 | mlx5_core_err(dev, "Fatal error %u detected\n" , fatal_error); |
| 804 | dev->priv.health.fatal_error = fatal_error; |
| 805 | print_health_info(dev); |
| 806 | dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR; |
| 807 | mlx5_trigger_health_work(dev); |
| 808 | return; |
| 809 | } |
| 810 | |
| 811 | count = ioread32be(health->health_counter); |
| 812 | if (count == health->prev) |
| 813 | ++health->miss_counter; |
| 814 | else |
| 815 | health->miss_counter = 0; |
| 816 | |
| 817 | health->prev = count; |
| 818 | if (health->miss_counter == MAX_MISSES) { |
| 819 | mlx5_core_err(dev, "device's health compromised - reached miss count\n" ); |
| 820 | health->synd = ioread8(&h->synd); |
| 821 | print_health_info(dev); |
| 822 | queue_work(wq: health->wq, work: &health->report_work); |
| 823 | } |
| 824 | |
| 825 | prev_synd = health->synd; |
| 826 | health->synd = ioread8(&h->synd); |
| 827 | if (health->synd && health->synd != prev_synd) { |
| 828 | print_health_info(dev); |
| 829 | queue_work(wq: health->wq, work: &health->report_work); |
| 830 | } |
| 831 | |
| 832 | out: |
| 833 | mod_timer(timer: &health->timer, expires: get_next_poll_jiffies(dev)); |
| 834 | } |
| 835 | |
| 836 | void mlx5_start_health_poll(struct mlx5_core_dev *dev) |
| 837 | { |
| 838 | u64 poll_interval_ms = mlx5_tout_ms(dev, HEALTH_POLL_INTERVAL); |
| 839 | struct mlx5_core_health *health = &dev->priv.health; |
| 840 | |
| 841 | timer_setup(&health->timer, poll_health, 0); |
| 842 | health->fatal_error = MLX5_SENSOR_NO_ERR; |
| 843 | clear_bit(nr: MLX5_DROP_HEALTH_WORK, addr: &health->flags); |
| 844 | health->health = &dev->iseg->health; |
| 845 | health->health_counter = &dev->iseg->health_counter; |
| 846 | |
| 847 | health->timer.expires = jiffies + msecs_to_jiffies(m: poll_interval_ms); |
| 848 | add_timer(timer: &health->timer); |
| 849 | } |
| 850 | |
| 851 | void mlx5_stop_health_poll(struct mlx5_core_dev *dev, bool disable_health) |
| 852 | { |
| 853 | struct mlx5_core_health *health = &dev->priv.health; |
| 854 | |
| 855 | if (disable_health) |
| 856 | set_bit(nr: MLX5_DROP_HEALTH_WORK, addr: &health->flags); |
| 857 | |
| 858 | timer_delete_sync(timer: &health->timer); |
| 859 | } |
| 860 | |
| 861 | void mlx5_start_health_fw_log_up(struct mlx5_core_dev *dev) |
| 862 | { |
| 863 | struct mlx5_core_health *health = &dev->priv.health; |
| 864 | |
| 865 | if (mlx5_core_is_pf(dev) && MLX5_CAP_MCAM_REG(dev, mrtc)) |
| 866 | queue_delayed_work(wq: health->wq, dwork: &health->update_fw_log_ts_work, delay: 0); |
| 867 | } |
| 868 | |
| 869 | void mlx5_drain_health_wq(struct mlx5_core_dev *dev) |
| 870 | { |
| 871 | struct mlx5_core_health *health = &dev->priv.health; |
| 872 | |
| 873 | set_bit(nr: MLX5_DROP_HEALTH_WORK, addr: &health->flags); |
| 874 | cancel_delayed_work_sync(dwork: &health->update_fw_log_ts_work); |
| 875 | cancel_work_sync(work: &health->report_work); |
| 876 | cancel_work_sync(work: &health->fatal_report_work); |
| 877 | } |
| 878 | |
| 879 | void mlx5_health_cleanup(struct mlx5_core_dev *dev) |
| 880 | { |
| 881 | struct mlx5_core_health *health = &dev->priv.health; |
| 882 | |
| 883 | cancel_delayed_work_sync(dwork: &health->update_fw_log_ts_work); |
| 884 | destroy_workqueue(wq: health->wq); |
| 885 | mlx5_reporter_vnic_destroy(dev); |
| 886 | mlx5_fw_reporters_destroy(dev); |
| 887 | } |
| 888 | |
| 889 | int mlx5_health_init(struct mlx5_core_dev *dev) |
| 890 | { |
| 891 | struct devlink *devlink = priv_to_devlink(priv: dev); |
| 892 | struct mlx5_core_health *health; |
| 893 | char *name; |
| 894 | |
| 895 | if (!mlx5_dev_is_lightweight(dev)) { |
| 896 | devl_lock(devlink); |
| 897 | mlx5_fw_reporters_create(dev); |
| 898 | devl_unlock(devlink); |
| 899 | } |
| 900 | mlx5_reporter_vnic_create(dev); |
| 901 | |
| 902 | health = &dev->priv.health; |
| 903 | name = kmalloc(64, GFP_KERNEL); |
| 904 | if (!name) |
| 905 | goto out_err; |
| 906 | |
| 907 | strcpy(p: name, q: "mlx5_health" ); |
| 908 | strcat(p: name, q: dev_name(dev: dev->device)); |
| 909 | health->wq = create_singlethread_workqueue(name); |
| 910 | kfree(objp: name); |
| 911 | if (!health->wq) |
| 912 | goto out_err; |
| 913 | INIT_WORK(&health->fatal_report_work, mlx5_fw_fatal_reporter_err_work); |
| 914 | INIT_WORK(&health->report_work, mlx5_fw_reporter_err_work); |
| 915 | INIT_DELAYED_WORK(&health->update_fw_log_ts_work, mlx5_health_log_ts_update); |
| 916 | |
| 917 | return 0; |
| 918 | |
| 919 | out_err: |
| 920 | mlx5_reporter_vnic_destroy(dev); |
| 921 | mlx5_fw_reporters_destroy(dev); |
| 922 | return -ENOMEM; |
| 923 | } |
| 924 | |