2 * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
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:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
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.
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
34 #include <linux/etherdevice.h>
35 #include <linux/tcp.h>
36 #include <linux/if_vlan.h>
37 #include <linux/delay.h>
38 #include <linux/slab.h>
40 #include <linux/mlx4/driver.h>
41 #include <linux/mlx4/device.h>
42 #include <linux/mlx4/cmd.h>
43 #include <linux/mlx4/cq.h>
48 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev, unsigned short vid)
50 struct mlx4_en_priv *priv = netdev_priv(dev);
51 struct mlx4_en_dev *mdev = priv->mdev;
55 en_dbg(HW, priv, "adding VLAN:%d\n", vid);
57 set_bit(vid, priv->active_vlans);
59 /* Add VID to port VLAN filter */
60 mutex_lock(&mdev->state_lock);
61 if (mdev->device_up && priv->port_up) {
62 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
64 en_err(priv, "Failed configuring VLAN filter\n");
66 if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
67 en_err(priv, "failed adding vlan %d\n", vid);
68 mutex_unlock(&mdev->state_lock);
73 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev, unsigned short vid)
75 struct mlx4_en_priv *priv = netdev_priv(dev);
76 struct mlx4_en_dev *mdev = priv->mdev;
80 en_dbg(HW, priv, "Killing VID:%d\n", vid);
82 clear_bit(vid, priv->active_vlans);
84 /* Remove VID from port VLAN filter */
85 mutex_lock(&mdev->state_lock);
86 if (!mlx4_find_cached_vlan(mdev->dev, priv->port, vid, &idx))
87 mlx4_unregister_vlan(mdev->dev, priv->port, idx);
89 en_err(priv, "could not find vid %d in cache\n", vid);
91 if (mdev->device_up && priv->port_up) {
92 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
94 en_err(priv, "Failed configuring VLAN filter\n");
96 mutex_unlock(&mdev->state_lock);
101 u64 mlx4_en_mac_to_u64(u8 *addr)
106 for (i = 0; i < ETH_ALEN; i++) {
113 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
115 struct mlx4_en_priv *priv = netdev_priv(dev);
116 struct mlx4_en_dev *mdev = priv->mdev;
117 struct sockaddr *saddr = addr;
119 if (!is_valid_ether_addr(saddr->sa_data))
120 return -EADDRNOTAVAIL;
122 memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
123 priv->mac = mlx4_en_mac_to_u64(dev->dev_addr);
124 queue_work(mdev->workqueue, &priv->mac_task);
128 static void mlx4_en_do_set_mac(struct work_struct *work)
130 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
132 struct mlx4_en_dev *mdev = priv->mdev;
135 mutex_lock(&mdev->state_lock);
137 /* Remove old MAC and insert the new one */
138 err = mlx4_replace_mac(mdev->dev, priv->port,
139 priv->base_qpn, priv->mac);
141 en_err(priv, "Failed changing HW MAC address\n");
143 en_dbg(HW, priv, "Port is down while "
144 "registering mac, exiting...\n");
146 mutex_unlock(&mdev->state_lock);
149 static void mlx4_en_clear_list(struct net_device *dev)
151 struct mlx4_en_priv *priv = netdev_priv(dev);
153 kfree(priv->mc_addrs);
154 priv->mc_addrs_cnt = 0;
157 static void mlx4_en_cache_mclist(struct net_device *dev)
159 struct mlx4_en_priv *priv = netdev_priv(dev);
160 struct netdev_hw_addr *ha;
162 int mc_addrs_cnt = netdev_mc_count(dev);
165 mc_addrs = kmalloc(mc_addrs_cnt * ETH_ALEN, GFP_ATOMIC);
167 en_err(priv, "failed to allocate multicast list\n");
171 netdev_for_each_mc_addr(ha, dev)
172 memcpy(mc_addrs + i++ * ETH_ALEN, ha->addr, ETH_ALEN);
173 priv->mc_addrs = mc_addrs;
174 priv->mc_addrs_cnt = mc_addrs_cnt;
178 static void mlx4_en_set_multicast(struct net_device *dev)
180 struct mlx4_en_priv *priv = netdev_priv(dev);
185 queue_work(priv->mdev->workqueue, &priv->mcast_task);
188 static void mlx4_en_do_set_multicast(struct work_struct *work)
190 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
192 struct mlx4_en_dev *mdev = priv->mdev;
193 struct net_device *dev = priv->dev;
195 u8 mc_list[16] = {0};
198 mutex_lock(&mdev->state_lock);
199 if (!mdev->device_up) {
200 en_dbg(HW, priv, "Card is not up, "
201 "ignoring multicast change.\n");
204 if (!priv->port_up) {
205 en_dbg(HW, priv, "Port is down, "
206 "ignoring multicast change.\n");
210 if (!netif_carrier_ok(dev)) {
211 if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
212 if (priv->port_state.link_state) {
213 priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
214 netif_carrier_on(dev);
215 en_dbg(LINK, priv, "Link Up\n");
221 * Promsicuous mode: disable all filters
224 if (dev->flags & IFF_PROMISC) {
225 if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
226 if (netif_msg_rx_status(priv))
227 en_warn(priv, "Entering promiscuous mode\n");
228 priv->flags |= MLX4_EN_FLAG_PROMISC;
230 /* Enable promiscouos mode */
231 if (!(mdev->dev->caps.flags &
232 MLX4_DEV_CAP_FLAG_VEP_UC_STEER))
233 err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port,
236 err = mlx4_unicast_promisc_add(mdev->dev, priv->base_qpn,
239 en_err(priv, "Failed enabling "
240 "promiscuous mode\n");
242 /* Disable port multicast filter (unconditionally) */
243 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
244 0, MLX4_MCAST_DISABLE);
246 en_err(priv, "Failed disabling "
247 "multicast filter\n");
249 /* Add the default qp number as multicast promisc */
250 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
251 err = mlx4_multicast_promisc_add(mdev->dev, priv->base_qpn,
254 en_err(priv, "Failed entering multicast promisc mode\n");
255 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
258 /* Disable port VLAN filter */
259 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
261 en_err(priv, "Failed disabling VLAN filter\n");
267 * Not in promiscuous mode
270 if (priv->flags & MLX4_EN_FLAG_PROMISC) {
271 if (netif_msg_rx_status(priv))
272 en_warn(priv, "Leaving promiscuous mode\n");
273 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
275 /* Disable promiscouos mode */
276 if (!(mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_VEP_UC_STEER))
277 err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port,
280 err = mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
283 en_err(priv, "Failed disabling promiscuous mode\n");
285 /* Disable Multicast promisc */
286 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
287 err = mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
290 en_err(priv, "Failed disabling multicast promiscuous mode\n");
291 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
294 /* Enable port VLAN filter */
295 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
297 en_err(priv, "Failed enabling VLAN filter\n");
300 /* Enable/disable the multicast filter according to IFF_ALLMULTI */
301 if (dev->flags & IFF_ALLMULTI) {
302 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
303 0, MLX4_MCAST_DISABLE);
305 en_err(priv, "Failed disabling multicast filter\n");
307 /* Add the default qp number as multicast promisc */
308 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
309 err = mlx4_multicast_promisc_add(mdev->dev, priv->base_qpn,
312 en_err(priv, "Failed entering multicast promisc mode\n");
313 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
317 /* Disable Multicast promisc */
318 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
319 err = mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
322 en_err(priv, "Failed disabling multicast promiscuous mode\n");
323 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
326 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
327 0, MLX4_MCAST_DISABLE);
329 en_err(priv, "Failed disabling multicast filter\n");
331 /* Detach our qp from all the multicast addresses */
332 for (i = 0; i < priv->mc_addrs_cnt; i++) {
333 memcpy(&mc_list[10], priv->mc_addrs + i * ETH_ALEN, ETH_ALEN);
334 mc_list[5] = priv->port;
335 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
336 mc_list, MLX4_PROT_ETH);
338 /* Flush mcast filter and init it with broadcast address */
339 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
340 1, MLX4_MCAST_CONFIG);
342 /* Update multicast list - we cache all addresses so they won't
343 * change while HW is updated holding the command semaphor */
344 netif_tx_lock_bh(dev);
345 mlx4_en_cache_mclist(dev);
346 netif_tx_unlock_bh(dev);
347 for (i = 0; i < priv->mc_addrs_cnt; i++) {
349 mlx4_en_mac_to_u64(priv->mc_addrs + i * ETH_ALEN);
350 memcpy(&mc_list[10], priv->mc_addrs + i * ETH_ALEN, ETH_ALEN);
351 mc_list[5] = priv->port;
352 mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp,
353 mc_list, 0, MLX4_PROT_ETH);
354 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
355 mcast_addr, 0, MLX4_MCAST_CONFIG);
357 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
358 0, MLX4_MCAST_ENABLE);
360 en_err(priv, "Failed enabling multicast filter\n");
363 mutex_unlock(&mdev->state_lock);
366 #ifdef CONFIG_NET_POLL_CONTROLLER
367 static void mlx4_en_netpoll(struct net_device *dev)
369 struct mlx4_en_priv *priv = netdev_priv(dev);
370 struct mlx4_en_cq *cq;
374 for (i = 0; i < priv->rx_ring_num; i++) {
375 cq = &priv->rx_cq[i];
376 spin_lock_irqsave(&cq->lock, flags);
377 napi_synchronize(&cq->napi);
378 mlx4_en_process_rx_cq(dev, cq, 0);
379 spin_unlock_irqrestore(&cq->lock, flags);
384 static void mlx4_en_tx_timeout(struct net_device *dev)
386 struct mlx4_en_priv *priv = netdev_priv(dev);
387 struct mlx4_en_dev *mdev = priv->mdev;
389 if (netif_msg_timer(priv))
390 en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
392 priv->port_stats.tx_timeout++;
393 en_dbg(DRV, priv, "Scheduling watchdog\n");
394 queue_work(mdev->workqueue, &priv->watchdog_task);
398 static struct net_device_stats *mlx4_en_get_stats(struct net_device *dev)
400 struct mlx4_en_priv *priv = netdev_priv(dev);
402 spin_lock_bh(&priv->stats_lock);
403 memcpy(&priv->ret_stats, &priv->stats, sizeof(priv->stats));
404 spin_unlock_bh(&priv->stats_lock);
406 return &priv->ret_stats;
409 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
411 struct mlx4_en_cq *cq;
414 /* If we haven't received a specific coalescing setting
415 * (module param), we set the moderation parameters as follows:
416 * - moder_cnt is set to the number of mtu sized packets to
417 * satisfy our coelsing target.
418 * - moder_time is set to a fixed value.
420 priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
421 priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
422 en_dbg(INTR, priv, "Default coalesing params for mtu:%d - "
423 "rx_frames:%d rx_usecs:%d\n",
424 priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
426 /* Setup cq moderation params */
427 for (i = 0; i < priv->rx_ring_num; i++) {
428 cq = &priv->rx_cq[i];
429 cq->moder_cnt = priv->rx_frames;
430 cq->moder_time = priv->rx_usecs;
431 priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
432 priv->last_moder_packets[i] = 0;
433 priv->last_moder_bytes[i] = 0;
436 for (i = 0; i < priv->tx_ring_num; i++) {
437 cq = &priv->tx_cq[i];
438 cq->moder_cnt = MLX4_EN_TX_COAL_PKTS;
439 cq->moder_time = MLX4_EN_TX_COAL_TIME;
442 /* Reset auto-moderation params */
443 priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
444 priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
445 priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
446 priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
447 priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
448 priv->adaptive_rx_coal = 1;
449 priv->last_moder_jiffies = 0;
450 priv->last_moder_tx_packets = 0;
453 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
455 unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
456 struct mlx4_en_cq *cq;
457 unsigned long packets;
459 unsigned long avg_pkt_size;
460 unsigned long rx_packets;
461 unsigned long rx_bytes;
462 unsigned long rx_pkt_diff;
466 if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
469 for (ring = 0; ring < priv->rx_ring_num; ring++) {
470 spin_lock_bh(&priv->stats_lock);
471 rx_packets = priv->rx_ring[ring].packets;
472 rx_bytes = priv->rx_ring[ring].bytes;
473 spin_unlock_bh(&priv->stats_lock);
475 rx_pkt_diff = ((unsigned long) (rx_packets -
476 priv->last_moder_packets[ring]));
477 packets = rx_pkt_diff;
478 rate = packets * HZ / period;
479 avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
480 priv->last_moder_bytes[ring])) / packets : 0;
482 /* Apply auto-moderation only when packet rate
483 * exceeds a rate that it matters */
484 if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
485 avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
486 if (rate < priv->pkt_rate_low)
487 moder_time = priv->rx_usecs_low;
488 else if (rate > priv->pkt_rate_high)
489 moder_time = priv->rx_usecs_high;
491 moder_time = (rate - priv->pkt_rate_low) *
492 (priv->rx_usecs_high - priv->rx_usecs_low) /
493 (priv->pkt_rate_high - priv->pkt_rate_low) +
496 moder_time = priv->rx_usecs_low;
499 if (moder_time != priv->last_moder_time[ring]) {
500 priv->last_moder_time[ring] = moder_time;
501 cq = &priv->rx_cq[ring];
502 cq->moder_time = moder_time;
503 err = mlx4_en_set_cq_moder(priv, cq);
505 en_err(priv, "Failed modifying moderation "
506 "for cq:%d\n", ring);
508 priv->last_moder_packets[ring] = rx_packets;
509 priv->last_moder_bytes[ring] = rx_bytes;
512 priv->last_moder_jiffies = jiffies;
515 static void mlx4_en_do_get_stats(struct work_struct *work)
517 struct delayed_work *delay = to_delayed_work(work);
518 struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
520 struct mlx4_en_dev *mdev = priv->mdev;
523 err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
525 en_dbg(HW, priv, "Could not update stats\n");
527 mutex_lock(&mdev->state_lock);
528 if (mdev->device_up) {
530 mlx4_en_auto_moderation(priv);
532 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
534 if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
535 queue_work(mdev->workqueue, &priv->mac_task);
536 mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
538 mutex_unlock(&mdev->state_lock);
541 static void mlx4_en_linkstate(struct work_struct *work)
543 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
545 struct mlx4_en_dev *mdev = priv->mdev;
546 int linkstate = priv->link_state;
548 mutex_lock(&mdev->state_lock);
549 /* If observable port state changed set carrier state and
550 * report to system log */
551 if (priv->last_link_state != linkstate) {
552 if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
553 en_info(priv, "Link Down\n");
554 netif_carrier_off(priv->dev);
556 en_info(priv, "Link Up\n");
557 netif_carrier_on(priv->dev);
560 priv->last_link_state = linkstate;
561 mutex_unlock(&mdev->state_lock);
565 int mlx4_en_start_port(struct net_device *dev)
567 struct mlx4_en_priv *priv = netdev_priv(dev);
568 struct mlx4_en_dev *mdev = priv->mdev;
569 struct mlx4_en_cq *cq;
570 struct mlx4_en_tx_ring *tx_ring;
576 u8 mc_list[16] = {0};
579 en_dbg(DRV, priv, "start port called while port already up\n");
583 /* Calculate Rx buf size */
584 dev->mtu = min(dev->mtu, priv->max_mtu);
585 mlx4_en_calc_rx_buf(dev);
586 en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
588 /* Configure rx cq's and rings */
589 err = mlx4_en_activate_rx_rings(priv);
591 en_err(priv, "Failed to activate RX rings\n");
594 for (i = 0; i < priv->rx_ring_num; i++) {
595 cq = &priv->rx_cq[i];
597 err = mlx4_en_activate_cq(priv, cq, i);
599 en_err(priv, "Failed activating Rx CQ\n");
602 for (j = 0; j < cq->size; j++)
603 cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
604 err = mlx4_en_set_cq_moder(priv, cq);
606 en_err(priv, "Failed setting cq moderation parameters");
607 mlx4_en_deactivate_cq(priv, cq);
610 mlx4_en_arm_cq(priv, cq);
611 priv->rx_ring[i].cqn = cq->mcq.cqn;
616 en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
617 err = mlx4_get_eth_qp(mdev->dev, priv->port,
618 priv->mac, &priv->base_qpn);
620 en_err(priv, "Failed getting eth qp\n");
623 mdev->mac_removed[priv->port] = 0;
625 err = mlx4_en_config_rss_steer(priv);
627 en_err(priv, "Failed configuring rss steering\n");
631 /* Configure tx cq's and rings */
632 for (i = 0; i < priv->tx_ring_num; i++) {
634 cq = &priv->tx_cq[i];
635 err = mlx4_en_activate_cq(priv, cq, i);
637 en_err(priv, "Failed allocating Tx CQ\n");
640 err = mlx4_en_set_cq_moder(priv, cq);
642 en_err(priv, "Failed setting cq moderation parameters");
643 mlx4_en_deactivate_cq(priv, cq);
646 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
647 cq->buf->wqe_index = cpu_to_be16(0xffff);
650 tx_ring = &priv->tx_ring[i];
651 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn);
653 en_err(priv, "Failed allocating Tx ring\n");
654 mlx4_en_deactivate_cq(priv, cq);
657 /* Set initial ownership of all Tx TXBBs to SW (1) */
658 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
659 *((u32 *) (tx_ring->buf + j)) = 0xffffffff;
664 err = mlx4_SET_PORT_general(mdev->dev, priv->port,
665 priv->rx_skb_size + ETH_FCS_LEN,
666 priv->prof->tx_pause,
668 priv->prof->rx_pause,
671 en_err(priv, "Failed setting port general configurations "
672 "for port %d, with error %d\n", priv->port, err);
675 /* Set default qp number */
676 err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
678 en_err(priv, "Failed setting default qp numbers\n");
683 en_dbg(HW, priv, "Initializing port\n");
684 err = mlx4_INIT_PORT(mdev->dev, priv->port);
686 en_err(priv, "Failed Initializing port\n");
690 /* Attach rx QP to bradcast address */
691 memset(&mc_list[10], 0xff, ETH_ALEN);
692 mc_list[5] = priv->port;
693 if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
695 mlx4_warn(mdev, "Failed Attaching Broadcast\n");
697 /* Must redo promiscuous mode setup. */
698 priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
700 /* Schedule multicast task to populate multicast list */
701 queue_work(mdev->workqueue, &priv->mcast_task);
703 priv->port_up = true;
704 netif_tx_start_all_queues(dev);
709 mlx4_en_deactivate_tx_ring(priv, &priv->tx_ring[tx_index]);
710 mlx4_en_deactivate_cq(priv, &priv->tx_cq[tx_index]);
713 mlx4_en_release_rss_steer(priv);
715 mlx4_put_eth_qp(mdev->dev, priv->port, priv->mac, priv->base_qpn);
718 mlx4_en_deactivate_cq(priv, &priv->rx_cq[rx_index]);
719 for (i = 0; i < priv->rx_ring_num; i++)
720 mlx4_en_deactivate_rx_ring(priv, &priv->rx_ring[i]);
722 return err; /* need to close devices */
726 void mlx4_en_stop_port(struct net_device *dev)
728 struct mlx4_en_priv *priv = netdev_priv(dev);
729 struct mlx4_en_dev *mdev = priv->mdev;
731 u8 mc_list[16] = {0};
733 if (!priv->port_up) {
734 en_dbg(DRV, priv, "stop port called while port already down\n");
738 /* Synchronize with tx routine */
739 netif_tx_lock_bh(dev);
740 netif_tx_stop_all_queues(dev);
741 netif_tx_unlock_bh(dev);
743 /* Set port as not active */
744 priv->port_up = false;
746 /* Detach All multicasts */
747 memset(&mc_list[10], 0xff, ETH_ALEN);
748 mc_list[5] = priv->port;
749 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
751 for (i = 0; i < priv->mc_addrs_cnt; i++) {
752 memcpy(&mc_list[10], priv->mc_addrs + i * ETH_ALEN, ETH_ALEN);
753 mc_list[5] = priv->port;
754 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
755 mc_list, MLX4_PROT_ETH);
757 mlx4_en_clear_list(dev);
758 /* Flush multicast filter */
759 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
762 for (i = 0; i < priv->tx_ring_num; i++) {
763 mlx4_en_deactivate_tx_ring(priv, &priv->tx_ring[i]);
764 mlx4_en_deactivate_cq(priv, &priv->tx_cq[i]);
768 for (i = 0; i < priv->tx_ring_num; i++)
769 mlx4_en_free_tx_buf(dev, &priv->tx_ring[i]);
772 mlx4_en_release_rss_steer(priv);
774 /* Unregister Mac address for the port */
775 mlx4_put_eth_qp(mdev->dev, priv->port, priv->mac, priv->base_qpn);
776 mdev->mac_removed[priv->port] = 1;
779 for (i = 0; i < priv->rx_ring_num; i++) {
780 mlx4_en_deactivate_rx_ring(priv, &priv->rx_ring[i]);
781 while (test_bit(NAPI_STATE_SCHED, &priv->rx_cq[i].napi.state))
783 mlx4_en_deactivate_cq(priv, &priv->rx_cq[i]);
787 mlx4_CLOSE_PORT(mdev->dev, priv->port);
790 static void mlx4_en_restart(struct work_struct *work)
792 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
794 struct mlx4_en_dev *mdev = priv->mdev;
795 struct net_device *dev = priv->dev;
797 en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
799 mutex_lock(&mdev->state_lock);
801 mlx4_en_stop_port(dev);
802 if (mlx4_en_start_port(dev))
803 en_err(priv, "Failed restarting port %d\n", priv->port);
805 mutex_unlock(&mdev->state_lock);
809 static int mlx4_en_open(struct net_device *dev)
811 struct mlx4_en_priv *priv = netdev_priv(dev);
812 struct mlx4_en_dev *mdev = priv->mdev;
816 mutex_lock(&mdev->state_lock);
818 if (!mdev->device_up) {
819 en_err(priv, "Cannot open - device down/disabled\n");
824 /* Reset HW statistics and performance counters */
825 if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
826 en_dbg(HW, priv, "Failed dumping statistics\n");
828 memset(&priv->stats, 0, sizeof(priv->stats));
829 memset(&priv->pstats, 0, sizeof(priv->pstats));
831 for (i = 0; i < priv->tx_ring_num; i++) {
832 priv->tx_ring[i].bytes = 0;
833 priv->tx_ring[i].packets = 0;
835 for (i = 0; i < priv->rx_ring_num; i++) {
836 priv->rx_ring[i].bytes = 0;
837 priv->rx_ring[i].packets = 0;
840 err = mlx4_en_start_port(dev);
842 en_err(priv, "Failed starting port:%d\n", priv->port);
845 mutex_unlock(&mdev->state_lock);
850 static int mlx4_en_close(struct net_device *dev)
852 struct mlx4_en_priv *priv = netdev_priv(dev);
853 struct mlx4_en_dev *mdev = priv->mdev;
855 en_dbg(IFDOWN, priv, "Close port called\n");
857 mutex_lock(&mdev->state_lock);
859 mlx4_en_stop_port(dev);
860 netif_carrier_off(dev);
862 mutex_unlock(&mdev->state_lock);
866 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
870 for (i = 0; i < priv->tx_ring_num; i++) {
871 if (priv->tx_ring[i].tx_info)
872 mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
873 if (priv->tx_cq[i].buf)
874 mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
877 for (i = 0; i < priv->rx_ring_num; i++) {
878 if (priv->rx_ring[i].rx_info)
879 mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i]);
880 if (priv->rx_cq[i].buf)
881 mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
885 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
887 struct mlx4_en_port_profile *prof = priv->prof;
889 int base_tx_qpn, err;
891 err = mlx4_qp_reserve_range(priv->mdev->dev, priv->tx_ring_num, 256, &base_tx_qpn);
893 en_err(priv, "failed reserving range for TX rings\n");
897 /* Create tx Rings */
898 for (i = 0; i < priv->tx_ring_num; i++) {
899 if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
900 prof->tx_ring_size, i, TX))
903 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i], base_tx_qpn + i,
904 prof->tx_ring_size, TXBB_SIZE))
908 /* Create rx Rings */
909 for (i = 0; i < priv->rx_ring_num; i++) {
910 if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
911 prof->rx_ring_size, i, RX))
914 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
915 prof->rx_ring_size, priv->stride))
922 en_err(priv, "Failed to allocate NIC resources\n");
923 mlx4_qp_release_range(priv->mdev->dev, base_tx_qpn, priv->tx_ring_num);
928 void mlx4_en_destroy_netdev(struct net_device *dev)
930 struct mlx4_en_priv *priv = netdev_priv(dev);
931 struct mlx4_en_dev *mdev = priv->mdev;
933 en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
935 /* Unregister device - this will close the port if it was up */
936 if (priv->registered)
937 unregister_netdev(dev);
940 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
942 cancel_delayed_work(&priv->stats_task);
943 /* flush any pending task for this netdev */
944 flush_workqueue(mdev->workqueue);
946 /* Detach the netdev so tasks would not attempt to access it */
947 mutex_lock(&mdev->state_lock);
948 mdev->pndev[priv->port] = NULL;
949 mutex_unlock(&mdev->state_lock);
951 mlx4_en_free_resources(priv);
955 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
957 struct mlx4_en_priv *priv = netdev_priv(dev);
958 struct mlx4_en_dev *mdev = priv->mdev;
961 en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
964 if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
965 en_err(priv, "Bad MTU size:%d.\n", new_mtu);
970 if (netif_running(dev)) {
971 mutex_lock(&mdev->state_lock);
972 if (!mdev->device_up) {
973 /* NIC is probably restarting - let watchdog task reset
975 en_dbg(DRV, priv, "Change MTU called with card down!?\n");
977 mlx4_en_stop_port(dev);
978 err = mlx4_en_start_port(dev);
980 en_err(priv, "Failed restarting port:%d\n",
982 queue_work(mdev->workqueue, &priv->watchdog_task);
985 mutex_unlock(&mdev->state_lock);
990 static int mlx4_en_set_features(struct net_device *netdev,
991 netdev_features_t features)
993 struct mlx4_en_priv *priv = netdev_priv(netdev);
995 if (features & NETIF_F_LOOPBACK)
996 priv->ctrl_flags |= cpu_to_be32(MLX4_WQE_CTRL_FORCE_LOOPBACK);
999 cpu_to_be32(~MLX4_WQE_CTRL_FORCE_LOOPBACK);
1005 static const struct net_device_ops mlx4_netdev_ops = {
1006 .ndo_open = mlx4_en_open,
1007 .ndo_stop = mlx4_en_close,
1008 .ndo_start_xmit = mlx4_en_xmit,
1009 .ndo_select_queue = mlx4_en_select_queue,
1010 .ndo_get_stats = mlx4_en_get_stats,
1011 .ndo_set_rx_mode = mlx4_en_set_multicast,
1012 .ndo_set_mac_address = mlx4_en_set_mac,
1013 .ndo_validate_addr = eth_validate_addr,
1014 .ndo_change_mtu = mlx4_en_change_mtu,
1015 .ndo_tx_timeout = mlx4_en_tx_timeout,
1016 .ndo_vlan_rx_add_vid = mlx4_en_vlan_rx_add_vid,
1017 .ndo_vlan_rx_kill_vid = mlx4_en_vlan_rx_kill_vid,
1018 #ifdef CONFIG_NET_POLL_CONTROLLER
1019 .ndo_poll_controller = mlx4_en_netpoll,
1021 .ndo_set_features = mlx4_en_set_features,
1024 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
1025 struct mlx4_en_port_profile *prof)
1027 struct net_device *dev;
1028 struct mlx4_en_priv *priv;
1032 dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
1033 prof->tx_ring_num, prof->rx_ring_num);
1035 mlx4_err(mdev, "Net device allocation failed\n");
1039 SET_NETDEV_DEV(dev, &mdev->dev->pdev->dev);
1040 dev->dev_id = port - 1;
1043 * Initialize driver private data
1046 priv = netdev_priv(dev);
1047 memset(priv, 0, sizeof(struct mlx4_en_priv));
1052 priv->port_up = false;
1053 priv->flags = prof->flags;
1054 priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
1055 MLX4_WQE_CTRL_SOLICITED);
1056 priv->tx_ring_num = prof->tx_ring_num;
1057 priv->rx_ring_num = prof->rx_ring_num;
1058 priv->mac_index = -1;
1059 priv->msg_enable = MLX4_EN_MSG_LEVEL;
1060 spin_lock_init(&priv->stats_lock);
1061 INIT_WORK(&priv->mcast_task, mlx4_en_do_set_multicast);
1062 INIT_WORK(&priv->mac_task, mlx4_en_do_set_mac);
1063 INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
1064 INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
1065 INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
1067 /* Query for default mac and max mtu */
1068 priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
1069 priv->mac = mdev->dev->caps.def_mac[priv->port];
1070 if (ILLEGAL_MAC(priv->mac)) {
1071 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
1072 priv->port, priv->mac);
1077 priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
1078 DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
1079 err = mlx4_en_alloc_resources(priv);
1083 /* Allocate page for receive rings */
1084 err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
1085 MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
1087 en_err(priv, "Failed to allocate page for rx qps\n");
1090 priv->allocated = 1;
1093 * Initialize netdev entry points
1095 dev->netdev_ops = &mlx4_netdev_ops;
1096 dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
1097 netif_set_real_num_tx_queues(dev, priv->tx_ring_num);
1098 netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
1100 SET_ETHTOOL_OPS(dev, &mlx4_en_ethtool_ops);
1102 /* Set defualt MAC */
1103 dev->addr_len = ETH_ALEN;
1104 for (i = 0; i < ETH_ALEN; i++) {
1105 dev->dev_addr[ETH_ALEN - 1 - i] = (u8) (priv->mac >> (8 * i));
1106 dev->perm_addr[ETH_ALEN - 1 - i] = (u8) (priv->mac >> (8 * i));
1110 * Set driver features
1112 dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
1113 if (mdev->LSO_support)
1114 dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
1116 dev->vlan_features = dev->hw_features;
1118 dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
1119 dev->features = dev->hw_features | NETIF_F_HIGHDMA |
1120 NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX |
1121 NETIF_F_HW_VLAN_FILTER;
1122 dev->hw_features |= NETIF_F_LOOPBACK;
1124 mdev->pndev[port] = dev;
1126 netif_carrier_off(dev);
1127 err = register_netdev(dev);
1129 en_err(priv, "Netdev registration failed for port %d\n", port);
1132 priv->registered = 1;
1134 en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
1135 en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
1137 /* Configure port */
1138 err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1142 en_err(priv, "Failed setting port general configurations "
1143 "for port %d, with error %d\n", priv->port, err);
1148 en_warn(priv, "Initializing port\n");
1149 err = mlx4_INIT_PORT(mdev->dev, priv->port);
1151 en_err(priv, "Failed Initializing port\n");
1154 mlx4_en_set_default_moderation(priv);
1155 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1159 mlx4_en_destroy_netdev(dev);