1 /*- 2 * BSD LICENSE 3 * 4 * Copyright 2017 6WIND S.A. 5 * Copyright 2017 Mellanox. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 11 * * Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * * Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in 15 * the documentation and/or other materials provided with the 16 * distribution. 17 * * Neither the name of 6WIND S.A. nor the names of its 18 * contributors may be used to endorse or promote products derived 19 * from this software without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 #include <unistd.h> 35 36 #include <rte_flow.h> 37 #include <rte_flow_driver.h> 38 39 #include "failsafe_private.h" 40 41 /** Print a message out of a flow error. */ 42 static int 43 fs_flow_complain(struct rte_flow_error *error) 44 { 45 static const char *const errstrlist[] = { 46 [RTE_FLOW_ERROR_TYPE_NONE] = "no error", 47 [RTE_FLOW_ERROR_TYPE_UNSPECIFIED] = "cause unspecified", 48 [RTE_FLOW_ERROR_TYPE_HANDLE] = "flow rule (handle)", 49 [RTE_FLOW_ERROR_TYPE_ATTR_GROUP] = "group field", 50 [RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY] = "priority field", 51 [RTE_FLOW_ERROR_TYPE_ATTR_INGRESS] = "ingress field", 52 [RTE_FLOW_ERROR_TYPE_ATTR_EGRESS] = "egress field", 53 [RTE_FLOW_ERROR_TYPE_ATTR] = "attributes structure", 54 [RTE_FLOW_ERROR_TYPE_ITEM_NUM] = "pattern length", 55 [RTE_FLOW_ERROR_TYPE_ITEM] = "specific pattern item", 56 [RTE_FLOW_ERROR_TYPE_ACTION_NUM] = "number of actions", 57 [RTE_FLOW_ERROR_TYPE_ACTION] = "specific action", 58 }; 59 const char *errstr; 60 char buf[32]; 61 int err = rte_errno; 62 63 if ((unsigned int)error->type >= RTE_DIM(errstrlist) || 64 !errstrlist[error->type]) 65 errstr = "unknown type"; 66 else 67 errstr = errstrlist[error->type]; 68 ERROR("Caught error type %d (%s): %s%s\n", 69 error->type, errstr, 70 error->cause ? (snprintf(buf, sizeof(buf), "cause: %p, ", 71 error->cause), buf) : "", 72 error->message ? error->message : "(no stated reason)"); 73 return -err; 74 } 75 76 static int 77 eth_dev_flow_isolate_set(struct rte_eth_dev *dev, 78 struct sub_device *sdev) 79 { 80 struct rte_flow_error ferror; 81 int ret; 82 83 if (!PRIV(dev)->flow_isolated) { 84 DEBUG("Flow isolation already disabled"); 85 } else { 86 DEBUG("Enabling flow isolation"); 87 ret = rte_flow_isolate(PORT_ID(sdev), 88 PRIV(dev)->flow_isolated, 89 &ferror); 90 if (ret) { 91 fs_flow_complain(&ferror); 92 return ret; 93 } 94 } 95 return 0; 96 } 97 98 static int 99 fs_eth_dev_conf_apply(struct rte_eth_dev *dev, 100 struct sub_device *sdev) 101 { 102 struct rte_eth_dev *edev; 103 struct rte_vlan_filter_conf *vfc1; 104 struct rte_vlan_filter_conf *vfc2; 105 struct rte_flow *flow; 106 struct rte_flow_error ferror; 107 uint32_t i; 108 int ret; 109 110 edev = ETH(sdev); 111 /* RX queue setup */ 112 for (i = 0; i < dev->data->nb_rx_queues; i++) { 113 struct rxq *rxq; 114 115 rxq = dev->data->rx_queues[i]; 116 ret = rte_eth_rx_queue_setup(PORT_ID(sdev), i, 117 rxq->info.nb_desc, rxq->socket_id, 118 &rxq->info.conf, rxq->info.mp); 119 if (ret) { 120 ERROR("rx_queue_setup failed"); 121 return ret; 122 } 123 } 124 /* TX queue setup */ 125 for (i = 0; i < dev->data->nb_tx_queues; i++) { 126 struct txq *txq; 127 128 txq = dev->data->tx_queues[i]; 129 ret = rte_eth_tx_queue_setup(PORT_ID(sdev), i, 130 txq->info.nb_desc, txq->socket_id, 131 &txq->info.conf); 132 if (ret) { 133 ERROR("tx_queue_setup failed"); 134 return ret; 135 } 136 } 137 /* dev_link.link_status */ 138 if (dev->data->dev_link.link_status != 139 edev->data->dev_link.link_status) { 140 DEBUG("Configuring link_status"); 141 if (dev->data->dev_link.link_status) 142 ret = rte_eth_dev_set_link_up(PORT_ID(sdev)); 143 else 144 ret = rte_eth_dev_set_link_down(PORT_ID(sdev)); 145 if (ret) { 146 ERROR("Failed to apply link_status"); 147 return ret; 148 } 149 } else { 150 DEBUG("link_status already set"); 151 } 152 /* promiscuous */ 153 if (dev->data->promiscuous != edev->data->promiscuous) { 154 DEBUG("Configuring promiscuous"); 155 if (dev->data->promiscuous) 156 rte_eth_promiscuous_enable(PORT_ID(sdev)); 157 else 158 rte_eth_promiscuous_disable(PORT_ID(sdev)); 159 } else { 160 DEBUG("promiscuous already set"); 161 } 162 /* all_multicast */ 163 if (dev->data->all_multicast != edev->data->all_multicast) { 164 DEBUG("Configuring all_multicast"); 165 if (dev->data->all_multicast) 166 rte_eth_allmulticast_enable(PORT_ID(sdev)); 167 else 168 rte_eth_allmulticast_disable(PORT_ID(sdev)); 169 } else { 170 DEBUG("all_multicast already set"); 171 } 172 /* MTU */ 173 if (dev->data->mtu != edev->data->mtu) { 174 DEBUG("Configuring MTU"); 175 ret = rte_eth_dev_set_mtu(PORT_ID(sdev), dev->data->mtu); 176 if (ret) { 177 ERROR("Failed to apply MTU"); 178 return ret; 179 } 180 } else { 181 DEBUG("MTU already set"); 182 } 183 /* default MAC */ 184 DEBUG("Configuring default MAC address"); 185 ret = rte_eth_dev_default_mac_addr_set(PORT_ID(sdev), 186 &dev->data->mac_addrs[0]); 187 if (ret) { 188 ERROR("Setting default MAC address failed"); 189 return ret; 190 } 191 /* additional MAC */ 192 if (PRIV(dev)->nb_mac_addr > 1) 193 DEBUG("Configure additional MAC address%s", 194 (PRIV(dev)->nb_mac_addr > 2 ? "es" : "")); 195 for (i = 1; i < PRIV(dev)->nb_mac_addr; i++) { 196 struct ether_addr *ea; 197 198 ea = &dev->data->mac_addrs[i]; 199 ret = rte_eth_dev_mac_addr_add(PORT_ID(sdev), ea, 200 PRIV(dev)->mac_addr_pool[i]); 201 if (ret) { 202 char ea_fmt[ETHER_ADDR_FMT_SIZE]; 203 204 ether_format_addr(ea_fmt, ETHER_ADDR_FMT_SIZE, ea); 205 ERROR("Adding MAC address %s failed", ea_fmt); 206 } 207 } 208 /* VLAN filter */ 209 vfc1 = &dev->data->vlan_filter_conf; 210 vfc2 = &edev->data->vlan_filter_conf; 211 if (memcmp(vfc1, vfc2, sizeof(struct rte_vlan_filter_conf))) { 212 uint64_t vbit; 213 uint64_t ids; 214 size_t i; 215 uint16_t vlan_id; 216 217 DEBUG("Configuring VLAN filter"); 218 for (i = 0; i < RTE_DIM(vfc1->ids); i++) { 219 if (vfc1->ids[i] == 0) 220 continue; 221 ids = vfc1->ids[i]; 222 while (ids) { 223 vlan_id = 64 * i; 224 /* count trailing zeroes */ 225 vbit = ~ids & (ids - 1); 226 /* clear least significant bit set */ 227 ids ^= (ids ^ (ids - 1)) ^ vbit; 228 for (; vbit; vlan_id++) 229 vbit >>= 1; 230 ret = rte_eth_dev_vlan_filter( 231 PORT_ID(sdev), vlan_id, 1); 232 if (ret) { 233 ERROR("Failed to apply VLAN filter %hu", 234 vlan_id); 235 return ret; 236 } 237 } 238 } 239 } else { 240 DEBUG("VLAN filter already set"); 241 } 242 /* rte_flow */ 243 if (TAILQ_EMPTY(&PRIV(dev)->flow_list)) { 244 DEBUG("rte_flow already set"); 245 } else { 246 DEBUG("Resetting rte_flow configuration"); 247 ret = rte_flow_flush(PORT_ID(sdev), &ferror); 248 if (ret) { 249 fs_flow_complain(&ferror); 250 return ret; 251 } 252 i = 0; 253 rte_errno = 0; 254 DEBUG("Configuring rte_flow"); 255 TAILQ_FOREACH(flow, &PRIV(dev)->flow_list, next) { 256 DEBUG("Creating flow #%" PRIu32, i++); 257 flow->flows[SUB_ID(sdev)] = 258 rte_flow_create(PORT_ID(sdev), 259 &flow->fd->attr, 260 flow->fd->items, 261 flow->fd->actions, 262 &ferror); 263 ret = rte_errno; 264 if (ret) 265 break; 266 } 267 if (ret) { 268 fs_flow_complain(&ferror); 269 return ret; 270 } 271 } 272 return 0; 273 } 274 275 static void 276 fs_dev_remove(struct sub_device *sdev) 277 { 278 int ret; 279 280 if (sdev == NULL) 281 return; 282 switch (sdev->state) { 283 case DEV_STARTED: 284 rte_eth_dev_stop(PORT_ID(sdev)); 285 sdev->state = DEV_ACTIVE; 286 /* fallthrough */ 287 case DEV_ACTIVE: 288 rte_eth_dev_close(PORT_ID(sdev)); 289 sdev->state = DEV_PROBED; 290 /* fallthrough */ 291 case DEV_PROBED: 292 ret = rte_eal_hotplug_remove(sdev->bus->name, 293 sdev->dev->name); 294 if (ret) { 295 ERROR("Bus detach failed for sub_device %u", 296 SUB_ID(sdev)); 297 } else { 298 ETH(sdev)->state = RTE_ETH_DEV_UNUSED; 299 } 300 sdev->state = DEV_PARSED; 301 /* fallthrough */ 302 case DEV_PARSED: 303 case DEV_UNDEFINED: 304 sdev->state = DEV_UNDEFINED; 305 /* the end */ 306 break; 307 } 308 failsafe_hotplug_alarm_install(sdev->fs_dev); 309 } 310 311 static inline int 312 fs_rxtx_clean(struct sub_device *sdev) 313 { 314 uint16_t i; 315 316 for (i = 0; i < ETH(sdev)->data->nb_rx_queues; i++) 317 if (FS_ATOMIC_RX(sdev, i)) 318 return 0; 319 for (i = 0; i < ETH(sdev)->data->nb_tx_queues; i++) 320 if (FS_ATOMIC_TX(sdev, i)) 321 return 0; 322 return 1; 323 } 324 325 void 326 failsafe_dev_remove(struct rte_eth_dev *dev) 327 { 328 struct sub_device *sdev; 329 uint8_t i; 330 331 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) 332 if (sdev->remove && fs_rxtx_clean(sdev)) 333 fs_dev_remove(sdev); 334 } 335 336 int 337 failsafe_eth_dev_state_sync(struct rte_eth_dev *dev) 338 { 339 struct sub_device *sdev; 340 uint32_t inactive; 341 int ret; 342 uint8_t i; 343 344 if (PRIV(dev)->state < DEV_PARSED) 345 return 0; 346 347 ret = failsafe_args_parse_subs(dev); 348 if (ret) 349 goto err_remove; 350 351 if (PRIV(dev)->state < DEV_PROBED) 352 return 0; 353 ret = failsafe_eal_init(dev); 354 if (ret) 355 goto err_remove; 356 if (PRIV(dev)->state < DEV_ACTIVE) 357 return 0; 358 inactive = 0; 359 FOREACH_SUBDEV(sdev, i, dev) { 360 if (sdev->state == DEV_PROBED) { 361 inactive |= UINT32_C(1) << i; 362 ret = eth_dev_flow_isolate_set(dev, sdev); 363 if (ret) { 364 ERROR("Could not apply configuration to sub_device %d", 365 i); 366 goto err_remove; 367 } 368 } 369 } 370 ret = dev->dev_ops->dev_configure(dev); 371 if (ret) 372 goto err_remove; 373 FOREACH_SUBDEV(sdev, i, dev) { 374 if (inactive & (UINT32_C(1) << i)) { 375 ret = fs_eth_dev_conf_apply(dev, sdev); 376 if (ret) { 377 ERROR("Could not apply configuration to sub_device %d", 378 i); 379 goto err_remove; 380 } 381 } 382 } 383 /* 384 * If new devices have been configured, check if 385 * the link state has changed. 386 */ 387 if (inactive) 388 dev->dev_ops->link_update(dev, 1); 389 if (PRIV(dev)->state < DEV_STARTED) 390 return 0; 391 ret = dev->dev_ops->dev_start(dev); 392 if (ret) 393 goto err_remove; 394 return 0; 395 err_remove: 396 FOREACH_SUBDEV(sdev, i, dev) 397 if (sdev->state != PRIV(dev)->state) 398 sdev->remove = 1; 399 return ret; 400 } 401 402 int 403 failsafe_eth_rmv_event_callback(uint8_t port_id __rte_unused, 404 enum rte_eth_event_type event __rte_unused, 405 void *cb_arg, void *out __rte_unused) 406 { 407 struct sub_device *sdev = cb_arg; 408 409 /* Switch as soon as possible tx_dev. */ 410 fs_switch_dev(sdev->fs_dev, sdev); 411 /* Use safe bursts in any case. */ 412 set_burst_fn(sdev->fs_dev, 1); 413 /* 414 * Async removal, the sub-PMD will try to unregister 415 * the callback at the source of the current thread context. 416 */ 417 sdev->remove = 1; 418 return 0; 419 } 420 421 int 422 failsafe_eth_lsc_event_callback(uint8_t port_id __rte_unused, 423 enum rte_eth_event_type event __rte_unused, 424 void *cb_arg, void *out __rte_unused) 425 { 426 struct rte_eth_dev *dev = cb_arg; 427 int ret; 428 429 ret = dev->dev_ops->link_update(dev, 0); 430 /* We must pass on the LSC event */ 431 if (ret) 432 return _rte_eth_dev_callback_process(dev, 433 RTE_ETH_EVENT_INTR_LSC, 434 NULL, NULL); 435 else 436 return 0; 437 } 438