1 // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB 2 // Copyright (c) 2018 Mellanox Technologies 3 4 #include <linux/mlx5/driver.h> 5 6 #include "mlx5_core.h" 7 #include "lib/eq.h" 8 #include "lib/mlx5.h" 9 10 struct mlx5_event_nb { 11 struct mlx5_nb nb; 12 void *ctx; 13 }; 14 15 /* General events handlers for the low level mlx5_core driver 16 * 17 * Other Major feature specific events such as 18 * clock/eswitch/fpga/FW trace and many others, are handled elsewhere, with 19 * separate notifiers callbacks, specifically by those mlx5 components. 20 */ 21 static int any_notifier(struct notifier_block *, unsigned long, void *); 22 static int temp_warn(struct notifier_block *, unsigned long, void *); 23 static int port_module(struct notifier_block *, unsigned long, void *); 24 25 /* handler which forwards the event to events->nh, driver notifiers */ 26 static int forward_event(struct notifier_block *, unsigned long, void *); 27 28 static struct mlx5_nb events_nbs_ref[] = { 29 /* Events to be proccessed by mlx5_core */ 30 {.nb.notifier_call = any_notifier, .event_type = MLX5_EVENT_TYPE_NOTIFY_ANY }, 31 {.nb.notifier_call = temp_warn, .event_type = MLX5_EVENT_TYPE_TEMP_WARN_EVENT }, 32 {.nb.notifier_call = port_module, .event_type = MLX5_EVENT_TYPE_PORT_MODULE_EVENT }, 33 34 /* Events to be forwarded (as is) to mlx5 core interfaces (mlx5e/mlx5_ib) */ 35 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_PORT_CHANGE }, 36 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_GENERAL_EVENT }, 37 /* QP/WQ resource events to forward */ 38 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_DCT_DRAINED }, 39 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_PATH_MIG }, 40 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_COMM_EST }, 41 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_SQ_DRAINED }, 42 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_SRQ_LAST_WQE }, 43 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_WQ_CATAS_ERROR }, 44 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_PATH_MIG_FAILED }, 45 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR }, 46 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_WQ_ACCESS_ERROR }, 47 /* SRQ events */ 48 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_SRQ_CATAS_ERROR }, 49 {.nb.notifier_call = forward_event, .event_type = MLX5_EVENT_TYPE_SRQ_RQ_LIMIT }, 50 }; 51 52 struct mlx5_events { 53 struct mlx5_core_dev *dev; 54 struct mlx5_event_nb notifiers[ARRAY_SIZE(events_nbs_ref)]; 55 /* driver notifier chain */ 56 struct atomic_notifier_head nh; 57 /* port module events stats */ 58 struct mlx5_pme_stats pme_stats; 59 }; 60 61 static const char *eqe_type_str(u8 type) 62 { 63 switch (type) { 64 case MLX5_EVENT_TYPE_COMP: 65 return "MLX5_EVENT_TYPE_COMP"; 66 case MLX5_EVENT_TYPE_PATH_MIG: 67 return "MLX5_EVENT_TYPE_PATH_MIG"; 68 case MLX5_EVENT_TYPE_COMM_EST: 69 return "MLX5_EVENT_TYPE_COMM_EST"; 70 case MLX5_EVENT_TYPE_SQ_DRAINED: 71 return "MLX5_EVENT_TYPE_SQ_DRAINED"; 72 case MLX5_EVENT_TYPE_SRQ_LAST_WQE: 73 return "MLX5_EVENT_TYPE_SRQ_LAST_WQE"; 74 case MLX5_EVENT_TYPE_SRQ_RQ_LIMIT: 75 return "MLX5_EVENT_TYPE_SRQ_RQ_LIMIT"; 76 case MLX5_EVENT_TYPE_CQ_ERROR: 77 return "MLX5_EVENT_TYPE_CQ_ERROR"; 78 case MLX5_EVENT_TYPE_WQ_CATAS_ERROR: 79 return "MLX5_EVENT_TYPE_WQ_CATAS_ERROR"; 80 case MLX5_EVENT_TYPE_PATH_MIG_FAILED: 81 return "MLX5_EVENT_TYPE_PATH_MIG_FAILED"; 82 case MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR: 83 return "MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR"; 84 case MLX5_EVENT_TYPE_WQ_ACCESS_ERROR: 85 return "MLX5_EVENT_TYPE_WQ_ACCESS_ERROR"; 86 case MLX5_EVENT_TYPE_SRQ_CATAS_ERROR: 87 return "MLX5_EVENT_TYPE_SRQ_CATAS_ERROR"; 88 case MLX5_EVENT_TYPE_INTERNAL_ERROR: 89 return "MLX5_EVENT_TYPE_INTERNAL_ERROR"; 90 case MLX5_EVENT_TYPE_PORT_CHANGE: 91 return "MLX5_EVENT_TYPE_PORT_CHANGE"; 92 case MLX5_EVENT_TYPE_GPIO_EVENT: 93 return "MLX5_EVENT_TYPE_GPIO_EVENT"; 94 case MLX5_EVENT_TYPE_PORT_MODULE_EVENT: 95 return "MLX5_EVENT_TYPE_PORT_MODULE_EVENT"; 96 case MLX5_EVENT_TYPE_TEMP_WARN_EVENT: 97 return "MLX5_EVENT_TYPE_TEMP_WARN_EVENT"; 98 case MLX5_EVENT_TYPE_REMOTE_CONFIG: 99 return "MLX5_EVENT_TYPE_REMOTE_CONFIG"; 100 case MLX5_EVENT_TYPE_DB_BF_CONGESTION: 101 return "MLX5_EVENT_TYPE_DB_BF_CONGESTION"; 102 case MLX5_EVENT_TYPE_STALL_EVENT: 103 return "MLX5_EVENT_TYPE_STALL_EVENT"; 104 case MLX5_EVENT_TYPE_CMD: 105 return "MLX5_EVENT_TYPE_CMD"; 106 case MLX5_EVENT_TYPE_PAGE_REQUEST: 107 return "MLX5_EVENT_TYPE_PAGE_REQUEST"; 108 case MLX5_EVENT_TYPE_PAGE_FAULT: 109 return "MLX5_EVENT_TYPE_PAGE_FAULT"; 110 case MLX5_EVENT_TYPE_PPS_EVENT: 111 return "MLX5_EVENT_TYPE_PPS_EVENT"; 112 case MLX5_EVENT_TYPE_NIC_VPORT_CHANGE: 113 return "MLX5_EVENT_TYPE_NIC_VPORT_CHANGE"; 114 case MLX5_EVENT_TYPE_FPGA_ERROR: 115 return "MLX5_EVENT_TYPE_FPGA_ERROR"; 116 case MLX5_EVENT_TYPE_FPGA_QP_ERROR: 117 return "MLX5_EVENT_TYPE_FPGA_QP_ERROR"; 118 case MLX5_EVENT_TYPE_GENERAL_EVENT: 119 return "MLX5_EVENT_TYPE_GENERAL_EVENT"; 120 case MLX5_EVENT_TYPE_MONITOR_COUNTER: 121 return "MLX5_EVENT_TYPE_MONITOR_COUNTER"; 122 case MLX5_EVENT_TYPE_DEVICE_TRACER: 123 return "MLX5_EVENT_TYPE_DEVICE_TRACER"; 124 default: 125 return "Unrecognized event"; 126 } 127 } 128 129 /* handles all FW events, type == eqe->type */ 130 static int any_notifier(struct notifier_block *nb, 131 unsigned long type, void *data) 132 { 133 struct mlx5_event_nb *event_nb = mlx5_nb_cof(nb, struct mlx5_event_nb, nb); 134 struct mlx5_events *events = event_nb->ctx; 135 struct mlx5_eqe *eqe = data; 136 137 mlx5_core_dbg(events->dev, "Async eqe type %s, subtype (%d)\n", 138 eqe_type_str(eqe->type), eqe->sub_type); 139 return NOTIFY_OK; 140 } 141 142 /* type == MLX5_EVENT_TYPE_TEMP_WARN_EVENT */ 143 static int temp_warn(struct notifier_block *nb, unsigned long type, void *data) 144 { 145 struct mlx5_event_nb *event_nb = mlx5_nb_cof(nb, struct mlx5_event_nb, nb); 146 struct mlx5_events *events = event_nb->ctx; 147 struct mlx5_eqe *eqe = data; 148 u64 value_lsb; 149 u64 value_msb; 150 151 value_lsb = be64_to_cpu(eqe->data.temp_warning.sensor_warning_lsb); 152 value_msb = be64_to_cpu(eqe->data.temp_warning.sensor_warning_msb); 153 154 mlx5_core_warn(events->dev, 155 "High temperature on sensors with bit set %llx %llx", 156 value_msb, value_lsb); 157 158 return NOTIFY_OK; 159 } 160 161 /* MLX5_EVENT_TYPE_PORT_MODULE_EVENT */ 162 static const char *mlx5_pme_status_to_string(enum port_module_event_status_type status) 163 { 164 switch (status) { 165 case MLX5_MODULE_STATUS_PLUGGED: 166 return "Cable plugged"; 167 case MLX5_MODULE_STATUS_UNPLUGGED: 168 return "Cable unplugged"; 169 case MLX5_MODULE_STATUS_ERROR: 170 return "Cable error"; 171 case MLX5_MODULE_STATUS_DISABLED: 172 return "Cable disabled"; 173 default: 174 return "Unknown status"; 175 } 176 } 177 178 static const char *mlx5_pme_error_to_string(enum port_module_event_error_type error) 179 { 180 switch (error) { 181 case MLX5_MODULE_EVENT_ERROR_POWER_BUDGET_EXCEEDED: 182 return "Power budget exceeded"; 183 case MLX5_MODULE_EVENT_ERROR_LONG_RANGE_FOR_NON_MLNX: 184 return "Long Range for non MLNX cable"; 185 case MLX5_MODULE_EVENT_ERROR_BUS_STUCK: 186 return "Bus stuck (I2C or data shorted)"; 187 case MLX5_MODULE_EVENT_ERROR_NO_EEPROM_RETRY_TIMEOUT: 188 return "No EEPROM/retry timeout"; 189 case MLX5_MODULE_EVENT_ERROR_ENFORCE_PART_NUMBER_LIST: 190 return "Enforce part number list"; 191 case MLX5_MODULE_EVENT_ERROR_UNKNOWN_IDENTIFIER: 192 return "Unknown identifier"; 193 case MLX5_MODULE_EVENT_ERROR_HIGH_TEMPERATURE: 194 return "High Temperature"; 195 case MLX5_MODULE_EVENT_ERROR_BAD_CABLE: 196 return "Bad or shorted cable/module"; 197 case MLX5_MODULE_EVENT_ERROR_PCIE_POWER_SLOT_EXCEEDED: 198 return "One or more network ports have been powered down due to insufficient/unadvertised power on the PCIe slot"; 199 default: 200 return "Unknown error"; 201 } 202 } 203 204 /* type == MLX5_EVENT_TYPE_PORT_MODULE_EVENT */ 205 static int port_module(struct notifier_block *nb, unsigned long type, void *data) 206 { 207 struct mlx5_event_nb *event_nb = mlx5_nb_cof(nb, struct mlx5_event_nb, nb); 208 struct mlx5_events *events = event_nb->ctx; 209 struct mlx5_eqe *eqe = data; 210 211 enum port_module_event_status_type module_status; 212 enum port_module_event_error_type error_type; 213 struct mlx5_eqe_port_module *module_event_eqe; 214 const char *status_str, *error_str; 215 u8 module_num; 216 217 module_event_eqe = &eqe->data.port_module; 218 module_num = module_event_eqe->module; 219 module_status = module_event_eqe->module_status & 220 PORT_MODULE_EVENT_MODULE_STATUS_MASK; 221 error_type = module_event_eqe->error_type & 222 PORT_MODULE_EVENT_ERROR_TYPE_MASK; 223 224 if (module_status < MLX5_MODULE_STATUS_NUM) 225 events->pme_stats.status_counters[module_status]++; 226 status_str = mlx5_pme_status_to_string(module_status); 227 228 if (module_status == MLX5_MODULE_STATUS_ERROR) { 229 if (error_type < MLX5_MODULE_EVENT_ERROR_NUM) 230 events->pme_stats.error_counters[error_type]++; 231 error_str = mlx5_pme_error_to_string(error_type); 232 } 233 234 if (!printk_ratelimit()) 235 return NOTIFY_OK; 236 237 if (module_status == MLX5_MODULE_STATUS_ERROR) 238 mlx5_core_err(events->dev, 239 "Port module event[error]: module %u, %s, %s\n", 240 module_num, status_str, error_str); 241 else 242 mlx5_core_info(events->dev, 243 "Port module event: module %u, %s\n", 244 module_num, status_str); 245 246 return NOTIFY_OK; 247 } 248 249 void mlx5_get_pme_stats(struct mlx5_core_dev *dev, struct mlx5_pme_stats *stats) 250 { 251 *stats = dev->priv.events->pme_stats; 252 } 253 254 /* forward event as is to registered interfaces (mlx5e/mlx5_ib) */ 255 static int forward_event(struct notifier_block *nb, unsigned long event, void *data) 256 { 257 struct mlx5_event_nb *event_nb = mlx5_nb_cof(nb, struct mlx5_event_nb, nb); 258 struct mlx5_events *events = event_nb->ctx; 259 struct mlx5_eqe *eqe = data; 260 261 mlx5_core_dbg(events->dev, "Async eqe type %s, subtype (%d) forward to interfaces\n", 262 eqe_type_str(eqe->type), eqe->sub_type); 263 atomic_notifier_call_chain(&events->nh, event, data); 264 return NOTIFY_OK; 265 } 266 267 int mlx5_events_init(struct mlx5_core_dev *dev) 268 { 269 struct mlx5_events *events = kzalloc(sizeof(*events), GFP_KERNEL); 270 271 if (!events) 272 return -ENOMEM; 273 274 ATOMIC_INIT_NOTIFIER_HEAD(&events->nh); 275 events->dev = dev; 276 dev->priv.events = events; 277 return 0; 278 } 279 280 void mlx5_events_cleanup(struct mlx5_core_dev *dev) 281 { 282 kvfree(dev->priv.events); 283 } 284 285 void mlx5_events_start(struct mlx5_core_dev *dev) 286 { 287 struct mlx5_events *events = dev->priv.events; 288 int i; 289 290 for (i = 0; i < ARRAY_SIZE(events_nbs_ref); i++) { 291 events->notifiers[i].nb = events_nbs_ref[i]; 292 events->notifiers[i].ctx = events; 293 mlx5_eq_notifier_register(dev, &events->notifiers[i].nb); 294 } 295 } 296 297 void mlx5_events_stop(struct mlx5_core_dev *dev) 298 { 299 struct mlx5_events *events = dev->priv.events; 300 int i; 301 302 for (i = ARRAY_SIZE(events_nbs_ref) - 1; i >= 0 ; i--) 303 mlx5_eq_notifier_unregister(dev, &events->notifiers[i].nb); 304 } 305 306 int mlx5_notifier_register(struct mlx5_core_dev *dev, struct notifier_block *nb) 307 { 308 struct mlx5_events *events = dev->priv.events; 309 310 return atomic_notifier_chain_register(&events->nh, nb); 311 } 312 EXPORT_SYMBOL(mlx5_notifier_register); 313 314 int mlx5_notifier_unregister(struct mlx5_core_dev *dev, struct notifier_block *nb) 315 { 316 struct mlx5_events *events = dev->priv.events; 317 318 return atomic_notifier_chain_unregister(&events->nh, nb); 319 } 320 EXPORT_SYMBOL(mlx5_notifier_unregister); 321 322 int mlx5_notifier_call_chain(struct mlx5_events *events, unsigned int event, void *data) 323 { 324 return atomic_notifier_call_chain(&events->nh, event, data); 325 } 326