4 * Copyright 2017 6WIND S.A.
5 * Copyright 2017 Mellanox.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
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.
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.
36 #include <rte_debug.h>
37 #include <rte_atomic.h>
38 #include <rte_ethdev.h>
39 #include <rte_malloc.h>
41 #include <rte_cycles.h>
43 #include "failsafe_private.h"
45 static struct rte_eth_dev_info default_infos = {
46 /* Max possible number of elements */
47 .max_rx_pktlen = UINT32_MAX,
48 .max_rx_queues = RTE_MAX_QUEUES_PER_PORT,
49 .max_tx_queues = RTE_MAX_QUEUES_PER_PORT,
50 .max_mac_addrs = FAILSAFE_MAX_ETHADDR,
51 .max_hash_mac_addrs = UINT32_MAX,
52 .max_vfs = UINT16_MAX,
53 .max_vmdq_pools = UINT16_MAX,
58 .nb_seg_max = UINT16_MAX,
59 .nb_mtu_seg_max = UINT16_MAX,
65 .nb_seg_max = UINT16_MAX,
66 .nb_mtu_seg_max = UINT16_MAX,
69 * Set of capabilities that can be verified upon
70 * configuring a sub-device.
73 DEV_RX_OFFLOAD_VLAN_STRIP |
74 DEV_RX_OFFLOAD_QINQ_STRIP |
75 DEV_RX_OFFLOAD_IPV4_CKSUM |
76 DEV_RX_OFFLOAD_UDP_CKSUM |
77 DEV_RX_OFFLOAD_TCP_CKSUM |
78 DEV_RX_OFFLOAD_TCP_LRO,
79 .tx_offload_capa = 0x0,
80 .flow_type_rss_offloads = 0x0,
84 fs_dev_configure(struct rte_eth_dev *dev)
86 struct sub_device *sdev;
90 FOREACH_SUBDEV(sdev, i, dev) {
91 int rmv_interrupt = 0;
92 int lsc_interrupt = 0;
95 if (sdev->state != DEV_PROBED)
98 rmv_interrupt = ETH(sdev)->data->dev_flags &
101 DEBUG("Enabling RMV interrupts for sub_device %d", i);
102 dev->data->dev_conf.intr_conf.rmv = 1;
104 DEBUG("sub_device %d does not support RMV event", i);
106 lsc_enabled = dev->data->dev_conf.intr_conf.lsc;
107 lsc_interrupt = lsc_enabled &&
108 (ETH(sdev)->data->dev_flags &
109 RTE_ETH_DEV_INTR_LSC);
111 DEBUG("Enabling LSC interrupts for sub_device %d", i);
112 dev->data->dev_conf.intr_conf.lsc = 1;
113 } else if (lsc_enabled && !lsc_interrupt) {
114 DEBUG("Disabling LSC interrupts for sub_device %d", i);
115 dev->data->dev_conf.intr_conf.lsc = 0;
117 DEBUG("Configuring sub-device %d", i);
119 ret = rte_eth_dev_configure(PORT_ID(sdev),
120 dev->data->nb_rx_queues,
121 dev->data->nb_tx_queues,
122 &dev->data->dev_conf);
124 ERROR("Could not configure sub_device %d", i);
128 ret = rte_eth_dev_callback_register(PORT_ID(sdev),
129 RTE_ETH_EVENT_INTR_RMV,
130 failsafe_eth_rmv_event_callback,
133 WARN("Failed to register RMV callback for sub_device %d",
136 dev->data->dev_conf.intr_conf.rmv = 0;
138 ret = rte_eth_dev_callback_register(PORT_ID(sdev),
139 RTE_ETH_EVENT_INTR_LSC,
140 failsafe_eth_lsc_event_callback,
143 WARN("Failed to register LSC callback for sub_device %d",
146 dev->data->dev_conf.intr_conf.lsc = lsc_enabled;
147 sdev->state = DEV_ACTIVE;
149 if (PRIV(dev)->state < DEV_ACTIVE)
150 PRIV(dev)->state = DEV_ACTIVE;
155 fs_dev_start(struct rte_eth_dev *dev)
157 struct sub_device *sdev;
161 FOREACH_SUBDEV(sdev, i, dev) {
162 if (sdev->state != DEV_ACTIVE)
164 DEBUG("Starting sub_device %d", i);
165 ret = rte_eth_dev_start(PORT_ID(sdev));
168 sdev->state = DEV_STARTED;
170 if (PRIV(dev)->state < DEV_STARTED)
171 PRIV(dev)->state = DEV_STARTED;
172 fs_switch_dev(dev, NULL);
177 fs_dev_stop(struct rte_eth_dev *dev)
179 struct sub_device *sdev;
182 PRIV(dev)->state = DEV_STARTED - 1;
183 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_STARTED) {
184 rte_eth_dev_stop(PORT_ID(sdev));
185 sdev->state = DEV_STARTED - 1;
190 fs_dev_set_link_up(struct rte_eth_dev *dev)
192 struct sub_device *sdev;
196 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
197 DEBUG("Calling rte_eth_dev_set_link_up on sub_device %d", i);
198 ret = rte_eth_dev_set_link_up(PORT_ID(sdev));
200 ERROR("Operation rte_eth_dev_set_link_up failed for sub_device %d"
201 " with error %d", i, ret);
209 fs_dev_set_link_down(struct rte_eth_dev *dev)
211 struct sub_device *sdev;
215 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
216 DEBUG("Calling rte_eth_dev_set_link_down on sub_device %d", i);
217 ret = rte_eth_dev_set_link_down(PORT_ID(sdev));
219 ERROR("Operation rte_eth_dev_set_link_down failed for sub_device %d"
220 " with error %d", i, ret);
227 static void fs_dev_free_queues(struct rte_eth_dev *dev);
229 fs_dev_close(struct rte_eth_dev *dev)
231 struct sub_device *sdev;
234 failsafe_hotplug_alarm_cancel(dev);
235 if (PRIV(dev)->state == DEV_STARTED)
236 dev->dev_ops->dev_stop(dev);
237 PRIV(dev)->state = DEV_ACTIVE - 1;
238 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
239 DEBUG("Closing sub_device %d", i);
240 rte_eth_dev_close(PORT_ID(sdev));
241 sdev->state = DEV_ACTIVE - 1;
243 fs_dev_free_queues(dev);
247 fs_rx_queue_release(void *queue)
249 struct rte_eth_dev *dev;
250 struct sub_device *sdev;
257 dev = rxq->priv->dev;
258 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE)
259 SUBOPS(sdev, rx_queue_release)
260 (ETH(sdev)->data->rx_queues[rxq->qid]);
261 dev->data->rx_queues[rxq->qid] = NULL;
266 fs_rx_queue_setup(struct rte_eth_dev *dev,
267 uint16_t rx_queue_id,
269 unsigned int socket_id,
270 const struct rte_eth_rxconf *rx_conf,
271 struct rte_mempool *mb_pool)
273 struct sub_device *sdev;
278 rxq = dev->data->rx_queues[rx_queue_id];
280 fs_rx_queue_release(rxq);
281 dev->data->rx_queues[rx_queue_id] = NULL;
283 rxq = rte_zmalloc(NULL,
285 sizeof(rte_atomic64_t) * PRIV(dev)->subs_tail,
286 RTE_CACHE_LINE_SIZE);
289 FOREACH_SUBDEV(sdev, i, dev)
290 rte_atomic64_init(&rxq->refcnt[i]);
291 rxq->qid = rx_queue_id;
292 rxq->socket_id = socket_id;
293 rxq->info.mp = mb_pool;
294 rxq->info.conf = *rx_conf;
295 rxq->info.nb_desc = nb_rx_desc;
296 rxq->priv = PRIV(dev);
297 dev->data->rx_queues[rx_queue_id] = rxq;
298 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
299 ret = rte_eth_rx_queue_setup(PORT_ID(sdev),
301 nb_rx_desc, socket_id,
304 ERROR("RX queue setup failed for sub_device %d", i);
310 fs_rx_queue_release(rxq);
315 fs_tx_queue_release(void *queue)
317 struct rte_eth_dev *dev;
318 struct sub_device *sdev;
325 dev = txq->priv->dev;
326 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE)
327 SUBOPS(sdev, tx_queue_release)
328 (ETH(sdev)->data->tx_queues[txq->qid]);
329 dev->data->tx_queues[txq->qid] = NULL;
334 fs_tx_queue_setup(struct rte_eth_dev *dev,
335 uint16_t tx_queue_id,
337 unsigned int socket_id,
338 const struct rte_eth_txconf *tx_conf)
340 struct sub_device *sdev;
345 txq = dev->data->tx_queues[tx_queue_id];
347 fs_tx_queue_release(txq);
348 dev->data->tx_queues[tx_queue_id] = NULL;
350 txq = rte_zmalloc("ethdev TX queue",
352 sizeof(rte_atomic64_t) * PRIV(dev)->subs_tail,
353 RTE_CACHE_LINE_SIZE);
356 FOREACH_SUBDEV(sdev, i, dev)
357 rte_atomic64_init(&txq->refcnt[i]);
358 txq->qid = tx_queue_id;
359 txq->socket_id = socket_id;
360 txq->info.conf = *tx_conf;
361 txq->info.nb_desc = nb_tx_desc;
362 txq->priv = PRIV(dev);
363 dev->data->tx_queues[tx_queue_id] = txq;
364 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
365 ret = rte_eth_tx_queue_setup(PORT_ID(sdev),
367 nb_tx_desc, socket_id,
370 ERROR("TX queue setup failed for sub_device %d", i);
376 fs_tx_queue_release(txq);
381 fs_dev_free_queues(struct rte_eth_dev *dev)
385 for (i = 0; i < dev->data->nb_rx_queues; i++) {
386 fs_rx_queue_release(dev->data->rx_queues[i]);
387 dev->data->rx_queues[i] = NULL;
389 dev->data->nb_rx_queues = 0;
390 for (i = 0; i < dev->data->nb_tx_queues; i++) {
391 fs_tx_queue_release(dev->data->tx_queues[i]);
392 dev->data->tx_queues[i] = NULL;
394 dev->data->nb_tx_queues = 0;
398 fs_promiscuous_enable(struct rte_eth_dev *dev)
400 struct sub_device *sdev;
403 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE)
404 rte_eth_promiscuous_enable(PORT_ID(sdev));
408 fs_promiscuous_disable(struct rte_eth_dev *dev)
410 struct sub_device *sdev;
413 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE)
414 rte_eth_promiscuous_disable(PORT_ID(sdev));
418 fs_allmulticast_enable(struct rte_eth_dev *dev)
420 struct sub_device *sdev;
423 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE)
424 rte_eth_allmulticast_enable(PORT_ID(sdev));
428 fs_allmulticast_disable(struct rte_eth_dev *dev)
430 struct sub_device *sdev;
433 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE)
434 rte_eth_allmulticast_disable(PORT_ID(sdev));
438 fs_link_update(struct rte_eth_dev *dev,
439 int wait_to_complete)
441 struct sub_device *sdev;
445 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
446 DEBUG("Calling link_update on sub_device %d", i);
447 ret = (SUBOPS(sdev, link_update))(ETH(sdev), wait_to_complete);
448 if (ret && ret != -1) {
449 ERROR("Link update failed for sub_device %d with error %d",
454 if (TX_SUBDEV(dev)) {
455 struct rte_eth_link *l1;
456 struct rte_eth_link *l2;
458 l1 = &dev->data->dev_link;
459 l2 = Ð(TX_SUBDEV(dev))->data->dev_link;
460 if (memcmp(l1, l2, sizeof(*l1))) {
469 fs_stats_get(struct rte_eth_dev *dev,
470 struct rte_eth_stats *stats)
472 struct sub_device *sdev;
476 rte_memcpy(stats, &PRIV(dev)->stats_accumulator, sizeof(*stats));
477 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
478 struct rte_eth_stats *snapshot = &sdev->stats_snapshot.stats;
479 uint64_t *timestamp = &sdev->stats_snapshot.timestamp;
481 ret = rte_eth_stats_get(PORT_ID(sdev), snapshot);
483 ERROR("Operation rte_eth_stats_get failed for sub_device %d with error %d",
488 *timestamp = rte_rdtsc();
489 failsafe_stats_increment(stats, snapshot);
495 fs_stats_reset(struct rte_eth_dev *dev)
497 struct sub_device *sdev;
500 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
501 rte_eth_stats_reset(PORT_ID(sdev));
502 memset(&sdev->stats_snapshot, 0, sizeof(struct rte_eth_stats));
504 memset(&PRIV(dev)->stats_accumulator, 0, sizeof(struct rte_eth_stats));
508 * Fail-safe dev_infos_get rules:
512 * Use the maximum possible values for any field, so as not
513 * to impede any further configuration effort.
515 * Limits capabilities to those that are understood by the
516 * fail-safe PMD. This understanding stems from the fail-safe
517 * being capable of verifying that the related capability is
518 * expressed within the device configuration (struct rte_eth_conf).
520 * At least one probed sub_device:
522 * Uses values from the active probed sub_device
523 * The rationale here is that if any sub_device is less capable
524 * (for example concerning the number of queues) than the active
525 * sub_device, then its subsequent configuration will fail.
526 * It is impossible to foresee this failure when the failing sub_device
527 * is supposed to be plugged-in later on, so the configuration process
528 * is the single point of failure and error reporting.
530 * Uses a logical AND of RX capabilities among
531 * all sub_devices and the default capabilities.
532 * Uses a logical AND of TX capabilities among
533 * the active probed sub_device and the default capabilities.
537 fs_dev_infos_get(struct rte_eth_dev *dev,
538 struct rte_eth_dev_info *infos)
540 struct sub_device *sdev;
543 sdev = TX_SUBDEV(dev);
545 DEBUG("No probed device, using default infos");
546 rte_memcpy(&PRIV(dev)->infos, &default_infos,
547 sizeof(default_infos));
549 uint32_t rx_offload_capa;
551 rx_offload_capa = default_infos.rx_offload_capa;
552 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_PROBED) {
553 rte_eth_dev_info_get(PORT_ID(sdev),
555 rx_offload_capa &= PRIV(dev)->infos.rx_offload_capa;
557 sdev = TX_SUBDEV(dev);
558 rte_eth_dev_info_get(PORT_ID(sdev), &PRIV(dev)->infos);
559 PRIV(dev)->infos.rx_offload_capa = rx_offload_capa;
560 PRIV(dev)->infos.tx_offload_capa &=
561 default_infos.tx_offload_capa;
562 PRIV(dev)->infos.flow_type_rss_offloads &=
563 default_infos.flow_type_rss_offloads;
565 rte_memcpy(infos, &PRIV(dev)->infos, sizeof(*infos));
568 static const uint32_t *
569 fs_dev_supported_ptypes_get(struct rte_eth_dev *dev)
571 struct sub_device *sdev;
572 struct rte_eth_dev *edev;
574 sdev = TX_SUBDEV(dev);
578 /* ENOTSUP: counts as no supported ptypes */
579 if (SUBOPS(sdev, dev_supported_ptypes_get) == NULL)
582 * The API does not permit to do a clean AND of all ptypes,
583 * It is also incomplete by design and we do not really care
584 * to have a best possible value in this context.
585 * We just return the ptypes of the device of highest
586 * priority, usually the PREFERRED device.
588 return SUBOPS(sdev, dev_supported_ptypes_get)(edev);
592 fs_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
594 struct sub_device *sdev;
598 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
599 DEBUG("Calling rte_eth_dev_set_mtu on sub_device %d", i);
600 ret = rte_eth_dev_set_mtu(PORT_ID(sdev), mtu);
602 ERROR("Operation rte_eth_dev_set_mtu failed for sub_device %d with error %d",
611 fs_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
613 struct sub_device *sdev;
617 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
618 DEBUG("Calling rte_eth_dev_vlan_filter on sub_device %d", i);
619 ret = rte_eth_dev_vlan_filter(PORT_ID(sdev), vlan_id, on);
621 ERROR("Operation rte_eth_dev_vlan_filter failed for sub_device %d"
622 " with error %d", i, ret);
630 fs_flow_ctrl_get(struct rte_eth_dev *dev,
631 struct rte_eth_fc_conf *fc_conf)
633 struct sub_device *sdev;
635 sdev = TX_SUBDEV(dev);
638 if (SUBOPS(sdev, flow_ctrl_get) == NULL)
640 return SUBOPS(sdev, flow_ctrl_get)(ETH(sdev), fc_conf);
644 fs_flow_ctrl_set(struct rte_eth_dev *dev,
645 struct rte_eth_fc_conf *fc_conf)
647 struct sub_device *sdev;
651 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
652 DEBUG("Calling rte_eth_dev_flow_ctrl_set on sub_device %d", i);
653 ret = rte_eth_dev_flow_ctrl_set(PORT_ID(sdev), fc_conf);
655 ERROR("Operation rte_eth_dev_flow_ctrl_set failed for sub_device %d"
656 " with error %d", i, ret);
664 fs_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index)
666 struct sub_device *sdev;
669 /* No check: already done within the rte_eth_dev_mac_addr_remove
670 * call for the fail-safe device.
672 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE)
673 rte_eth_dev_mac_addr_remove(PORT_ID(sdev),
674 &dev->data->mac_addrs[index]);
675 PRIV(dev)->mac_addr_pool[index] = 0;
679 fs_mac_addr_add(struct rte_eth_dev *dev,
680 struct ether_addr *mac_addr,
684 struct sub_device *sdev;
688 RTE_ASSERT(index < FAILSAFE_MAX_ETHADDR);
689 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
690 ret = rte_eth_dev_mac_addr_add(PORT_ID(sdev), mac_addr, vmdq);
692 ERROR("Operation rte_eth_dev_mac_addr_add failed for sub_device %"
693 PRIu8 " with error %d", i, ret);
697 if (index >= PRIV(dev)->nb_mac_addr) {
698 DEBUG("Growing mac_addrs array");
699 PRIV(dev)->nb_mac_addr = index;
701 PRIV(dev)->mac_addr_pool[index] = vmdq;
706 fs_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
708 struct sub_device *sdev;
711 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE)
712 rte_eth_dev_default_mac_addr_set(PORT_ID(sdev), mac_addr);
716 fs_filter_ctrl(struct rte_eth_dev *dev,
717 enum rte_filter_type type,
718 enum rte_filter_op op,
721 struct sub_device *sdev;
725 if (type == RTE_ETH_FILTER_GENERIC &&
726 op == RTE_ETH_FILTER_GET) {
727 *(const void **)arg = &fs_flow_ops;
730 FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
731 DEBUG("Calling rte_eth_dev_filter_ctrl on sub_device %d", i);
732 ret = rte_eth_dev_filter_ctrl(PORT_ID(sdev), type, op, arg);
734 ERROR("Operation rte_eth_dev_filter_ctrl failed for sub_device %d"
735 " with error %d", i, ret);
742 const struct eth_dev_ops failsafe_ops = {
743 .dev_configure = fs_dev_configure,
744 .dev_start = fs_dev_start,
745 .dev_stop = fs_dev_stop,
746 .dev_set_link_down = fs_dev_set_link_down,
747 .dev_set_link_up = fs_dev_set_link_up,
748 .dev_close = fs_dev_close,
749 .promiscuous_enable = fs_promiscuous_enable,
750 .promiscuous_disable = fs_promiscuous_disable,
751 .allmulticast_enable = fs_allmulticast_enable,
752 .allmulticast_disable = fs_allmulticast_disable,
753 .link_update = fs_link_update,
754 .stats_get = fs_stats_get,
755 .stats_reset = fs_stats_reset,
756 .dev_infos_get = fs_dev_infos_get,
757 .dev_supported_ptypes_get = fs_dev_supported_ptypes_get,
758 .mtu_set = fs_mtu_set,
759 .vlan_filter_set = fs_vlan_filter_set,
760 .rx_queue_setup = fs_rx_queue_setup,
761 .tx_queue_setup = fs_tx_queue_setup,
762 .rx_queue_release = fs_rx_queue_release,
763 .tx_queue_release = fs_tx_queue_release,
764 .flow_ctrl_get = fs_flow_ctrl_get,
765 .flow_ctrl_set = fs_flow_ctrl_set,
766 .mac_addr_remove = fs_mac_addr_remove,
767 .mac_addr_add = fs_mac_addr_add,
768 .mac_addr_set = fs_mac_addr_set,
769 .filter_ctrl = fs_filter_ctrl,