net/mlx4_en: num cores tx rings for every UP
[linux-2.6.git] / drivers / net / ethernet / mellanox / mlx4 / en_netdev.c
blob926d8aac941c67c8e1160c120073b14aaddb778b
1 /*
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
12 * conditions are met:
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
16 * disclaimer.
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
30 * SOFTWARE.
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>
45 #include "mlx4_en.h"
46 #include "en_port.h"
48 static int mlx4_en_setup_tc(struct net_device *dev, u8 up)
50 struct mlx4_en_priv *priv = netdev_priv(dev);
51 int i;
52 unsigned int q, offset = 0;
54 if (up && up != MLX4_EN_NUM_UP)
55 return -EINVAL;
57 netdev_set_num_tc(dev, up);
59 /* Partition Tx queues evenly amongst UP's */
60 q = priv->tx_ring_num / up;
61 for (i = 0; i < up; i++) {
62 netdev_set_tc_queue(dev, i, q, offset);
63 offset += q;
66 return 0;
69 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev, unsigned short vid)
71 struct mlx4_en_priv *priv = netdev_priv(dev);
72 struct mlx4_en_dev *mdev = priv->mdev;
73 int err;
74 int idx;
76 en_dbg(HW, priv, "adding VLAN:%d\n", vid);
78 set_bit(vid, priv->active_vlans);
80 /* Add VID to port VLAN filter */
81 mutex_lock(&mdev->state_lock);
82 if (mdev->device_up && priv->port_up) {
83 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
84 if (err)
85 en_err(priv, "Failed configuring VLAN filter\n");
87 if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
88 en_err(priv, "failed adding vlan %d\n", vid);
89 mutex_unlock(&mdev->state_lock);
91 return 0;
94 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev, unsigned short vid)
96 struct mlx4_en_priv *priv = netdev_priv(dev);
97 struct mlx4_en_dev *mdev = priv->mdev;
98 int err;
99 int idx;
101 en_dbg(HW, priv, "Killing VID:%d\n", vid);
103 clear_bit(vid, priv->active_vlans);
105 /* Remove VID from port VLAN filter */
106 mutex_lock(&mdev->state_lock);
107 if (!mlx4_find_cached_vlan(mdev->dev, priv->port, vid, &idx))
108 mlx4_unregister_vlan(mdev->dev, priv->port, idx);
109 else
110 en_err(priv, "could not find vid %d in cache\n", vid);
112 if (mdev->device_up && priv->port_up) {
113 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
114 if (err)
115 en_err(priv, "Failed configuring VLAN filter\n");
117 mutex_unlock(&mdev->state_lock);
119 return 0;
122 u64 mlx4_en_mac_to_u64(u8 *addr)
124 u64 mac = 0;
125 int i;
127 for (i = 0; i < ETH_ALEN; i++) {
128 mac <<= 8;
129 mac |= addr[i];
131 return mac;
134 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
136 struct mlx4_en_priv *priv = netdev_priv(dev);
137 struct mlx4_en_dev *mdev = priv->mdev;
138 struct sockaddr *saddr = addr;
140 if (!is_valid_ether_addr(saddr->sa_data))
141 return -EADDRNOTAVAIL;
143 memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
144 priv->mac = mlx4_en_mac_to_u64(dev->dev_addr);
145 queue_work(mdev->workqueue, &priv->mac_task);
146 return 0;
149 static void mlx4_en_do_set_mac(struct work_struct *work)
151 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
152 mac_task);
153 struct mlx4_en_dev *mdev = priv->mdev;
154 int err = 0;
156 mutex_lock(&mdev->state_lock);
157 if (priv->port_up) {
158 /* Remove old MAC and insert the new one */
159 err = mlx4_replace_mac(mdev->dev, priv->port,
160 priv->base_qpn, priv->mac);
161 if (err)
162 en_err(priv, "Failed changing HW MAC address\n");
163 } else
164 en_dbg(HW, priv, "Port is down while "
165 "registering mac, exiting...\n");
167 mutex_unlock(&mdev->state_lock);
170 static void mlx4_en_clear_list(struct net_device *dev)
172 struct mlx4_en_priv *priv = netdev_priv(dev);
174 kfree(priv->mc_addrs);
175 priv->mc_addrs = NULL;
176 priv->mc_addrs_cnt = 0;
179 static void mlx4_en_cache_mclist(struct net_device *dev)
181 struct mlx4_en_priv *priv = netdev_priv(dev);
182 struct netdev_hw_addr *ha;
183 char *mc_addrs;
184 int mc_addrs_cnt = netdev_mc_count(dev);
185 int i;
187 mc_addrs = kmalloc(mc_addrs_cnt * ETH_ALEN, GFP_ATOMIC);
188 if (!mc_addrs) {
189 en_err(priv, "failed to allocate multicast list\n");
190 return;
192 i = 0;
193 netdev_for_each_mc_addr(ha, dev)
194 memcpy(mc_addrs + i++ * ETH_ALEN, ha->addr, ETH_ALEN);
195 mlx4_en_clear_list(dev);
196 priv->mc_addrs = mc_addrs;
197 priv->mc_addrs_cnt = mc_addrs_cnt;
201 static void mlx4_en_set_multicast(struct net_device *dev)
203 struct mlx4_en_priv *priv = netdev_priv(dev);
205 if (!priv->port_up)
206 return;
208 queue_work(priv->mdev->workqueue, &priv->mcast_task);
211 static void mlx4_en_do_set_multicast(struct work_struct *work)
213 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
214 mcast_task);
215 struct mlx4_en_dev *mdev = priv->mdev;
216 struct net_device *dev = priv->dev;
217 u64 mcast_addr = 0;
218 u8 mc_list[16] = {0};
219 int err;
221 mutex_lock(&mdev->state_lock);
222 if (!mdev->device_up) {
223 en_dbg(HW, priv, "Card is not up, "
224 "ignoring multicast change.\n");
225 goto out;
227 if (!priv->port_up) {
228 en_dbg(HW, priv, "Port is down, "
229 "ignoring multicast change.\n");
230 goto out;
233 if (!netif_carrier_ok(dev)) {
234 if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
235 if (priv->port_state.link_state) {
236 priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
237 netif_carrier_on(dev);
238 en_dbg(LINK, priv, "Link Up\n");
244 * Promsicuous mode: disable all filters
247 if (dev->flags & IFF_PROMISC) {
248 if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
249 if (netif_msg_rx_status(priv))
250 en_warn(priv, "Entering promiscuous mode\n");
251 priv->flags |= MLX4_EN_FLAG_PROMISC;
253 /* Enable promiscouos mode */
254 if (!(mdev->dev->caps.flags &
255 MLX4_DEV_CAP_FLAG_VEP_UC_STEER))
256 err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port,
257 priv->base_qpn, 1);
258 else
259 err = mlx4_unicast_promisc_add(mdev->dev, priv->base_qpn,
260 priv->port);
261 if (err)
262 en_err(priv, "Failed enabling "
263 "promiscuous mode\n");
265 /* Disable port multicast filter (unconditionally) */
266 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
267 0, MLX4_MCAST_DISABLE);
268 if (err)
269 en_err(priv, "Failed disabling "
270 "multicast filter\n");
272 /* Add the default qp number as multicast promisc */
273 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
274 err = mlx4_multicast_promisc_add(mdev->dev, priv->base_qpn,
275 priv->port);
276 if (err)
277 en_err(priv, "Failed entering multicast promisc mode\n");
278 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
281 /* Disable port VLAN filter */
282 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
283 if (err)
284 en_err(priv, "Failed disabling VLAN filter\n");
286 goto out;
290 * Not in promiscuous mode
293 if (priv->flags & MLX4_EN_FLAG_PROMISC) {
294 if (netif_msg_rx_status(priv))
295 en_warn(priv, "Leaving promiscuous mode\n");
296 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
298 /* Disable promiscouos mode */
299 if (!(mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_VEP_UC_STEER))
300 err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port,
301 priv->base_qpn, 0);
302 else
303 err = mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
304 priv->port);
305 if (err)
306 en_err(priv, "Failed disabling promiscuous mode\n");
308 /* Disable Multicast promisc */
309 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
310 err = mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
311 priv->port);
312 if (err)
313 en_err(priv, "Failed disabling multicast promiscuous mode\n");
314 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
317 /* Enable port VLAN filter */
318 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
319 if (err)
320 en_err(priv, "Failed enabling VLAN filter\n");
323 /* Enable/disable the multicast filter according to IFF_ALLMULTI */
324 if (dev->flags & IFF_ALLMULTI) {
325 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
326 0, MLX4_MCAST_DISABLE);
327 if (err)
328 en_err(priv, "Failed disabling multicast filter\n");
330 /* Add the default qp number as multicast promisc */
331 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
332 err = mlx4_multicast_promisc_add(mdev->dev, priv->base_qpn,
333 priv->port);
334 if (err)
335 en_err(priv, "Failed entering multicast promisc mode\n");
336 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
338 } else {
339 int i;
340 /* Disable Multicast promisc */
341 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
342 err = mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
343 priv->port);
344 if (err)
345 en_err(priv, "Failed disabling multicast promiscuous mode\n");
346 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
349 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
350 0, MLX4_MCAST_DISABLE);
351 if (err)
352 en_err(priv, "Failed disabling multicast filter\n");
354 /* Detach our qp from all the multicast addresses */
355 for (i = 0; i < priv->mc_addrs_cnt; i++) {
356 memcpy(&mc_list[10], priv->mc_addrs + i * ETH_ALEN, ETH_ALEN);
357 mc_list[5] = priv->port;
358 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
359 mc_list, MLX4_PROT_ETH);
361 /* Flush mcast filter and init it with broadcast address */
362 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
363 1, MLX4_MCAST_CONFIG);
365 /* Update multicast list - we cache all addresses so they won't
366 * change while HW is updated holding the command semaphor */
367 netif_tx_lock_bh(dev);
368 mlx4_en_cache_mclist(dev);
369 netif_tx_unlock_bh(dev);
370 for (i = 0; i < priv->mc_addrs_cnt; i++) {
371 mcast_addr =
372 mlx4_en_mac_to_u64(priv->mc_addrs + i * ETH_ALEN);
373 memcpy(&mc_list[10], priv->mc_addrs + i * ETH_ALEN, ETH_ALEN);
374 mc_list[5] = priv->port;
375 mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp,
376 mc_list, 0, MLX4_PROT_ETH);
377 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
378 mcast_addr, 0, MLX4_MCAST_CONFIG);
380 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
381 0, MLX4_MCAST_ENABLE);
382 if (err)
383 en_err(priv, "Failed enabling multicast filter\n");
385 out:
386 mutex_unlock(&mdev->state_lock);
389 #ifdef CONFIG_NET_POLL_CONTROLLER
390 static void mlx4_en_netpoll(struct net_device *dev)
392 struct mlx4_en_priv *priv = netdev_priv(dev);
393 struct mlx4_en_cq *cq;
394 unsigned long flags;
395 int i;
397 for (i = 0; i < priv->rx_ring_num; i++) {
398 cq = &priv->rx_cq[i];
399 spin_lock_irqsave(&cq->lock, flags);
400 napi_synchronize(&cq->napi);
401 mlx4_en_process_rx_cq(dev, cq, 0);
402 spin_unlock_irqrestore(&cq->lock, flags);
405 #endif
407 static void mlx4_en_tx_timeout(struct net_device *dev)
409 struct mlx4_en_priv *priv = netdev_priv(dev);
410 struct mlx4_en_dev *mdev = priv->mdev;
412 if (netif_msg_timer(priv))
413 en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
415 priv->port_stats.tx_timeout++;
416 en_dbg(DRV, priv, "Scheduling watchdog\n");
417 queue_work(mdev->workqueue, &priv->watchdog_task);
421 static struct net_device_stats *mlx4_en_get_stats(struct net_device *dev)
423 struct mlx4_en_priv *priv = netdev_priv(dev);
425 spin_lock_bh(&priv->stats_lock);
426 memcpy(&priv->ret_stats, &priv->stats, sizeof(priv->stats));
427 spin_unlock_bh(&priv->stats_lock);
429 return &priv->ret_stats;
432 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
434 struct mlx4_en_cq *cq;
435 int i;
437 /* If we haven't received a specific coalescing setting
438 * (module param), we set the moderation parameters as follows:
439 * - moder_cnt is set to the number of mtu sized packets to
440 * satisfy our coelsing target.
441 * - moder_time is set to a fixed value.
443 priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
444 priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
445 priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
446 priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
447 en_dbg(INTR, priv, "Default coalesing params for mtu:%d - "
448 "rx_frames:%d rx_usecs:%d\n",
449 priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
451 /* Setup cq moderation params */
452 for (i = 0; i < priv->rx_ring_num; i++) {
453 cq = &priv->rx_cq[i];
454 cq->moder_cnt = priv->rx_frames;
455 cq->moder_time = priv->rx_usecs;
456 priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
457 priv->last_moder_packets[i] = 0;
458 priv->last_moder_bytes[i] = 0;
461 for (i = 0; i < priv->tx_ring_num; i++) {
462 cq = &priv->tx_cq[i];
463 cq->moder_cnt = priv->tx_frames;
464 cq->moder_time = priv->tx_usecs;
467 /* Reset auto-moderation params */
468 priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
469 priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
470 priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
471 priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
472 priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
473 priv->adaptive_rx_coal = 1;
474 priv->last_moder_jiffies = 0;
475 priv->last_moder_tx_packets = 0;
478 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
480 unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
481 struct mlx4_en_cq *cq;
482 unsigned long packets;
483 unsigned long rate;
484 unsigned long avg_pkt_size;
485 unsigned long rx_packets;
486 unsigned long rx_bytes;
487 unsigned long rx_pkt_diff;
488 int moder_time;
489 int ring, err;
491 if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
492 return;
494 for (ring = 0; ring < priv->rx_ring_num; ring++) {
495 spin_lock_bh(&priv->stats_lock);
496 rx_packets = priv->rx_ring[ring].packets;
497 rx_bytes = priv->rx_ring[ring].bytes;
498 spin_unlock_bh(&priv->stats_lock);
500 rx_pkt_diff = ((unsigned long) (rx_packets -
501 priv->last_moder_packets[ring]));
502 packets = rx_pkt_diff;
503 rate = packets * HZ / period;
504 avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
505 priv->last_moder_bytes[ring])) / packets : 0;
507 /* Apply auto-moderation only when packet rate
508 * exceeds a rate that it matters */
509 if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
510 avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
511 if (rate < priv->pkt_rate_low)
512 moder_time = priv->rx_usecs_low;
513 else if (rate > priv->pkt_rate_high)
514 moder_time = priv->rx_usecs_high;
515 else
516 moder_time = (rate - priv->pkt_rate_low) *
517 (priv->rx_usecs_high - priv->rx_usecs_low) /
518 (priv->pkt_rate_high - priv->pkt_rate_low) +
519 priv->rx_usecs_low;
520 } else {
521 moder_time = priv->rx_usecs_low;
524 if (moder_time != priv->last_moder_time[ring]) {
525 priv->last_moder_time[ring] = moder_time;
526 cq = &priv->rx_cq[ring];
527 cq->moder_time = moder_time;
528 err = mlx4_en_set_cq_moder(priv, cq);
529 if (err)
530 en_err(priv, "Failed modifying moderation "
531 "for cq:%d\n", ring);
533 priv->last_moder_packets[ring] = rx_packets;
534 priv->last_moder_bytes[ring] = rx_bytes;
537 priv->last_moder_jiffies = jiffies;
540 static void mlx4_en_do_get_stats(struct work_struct *work)
542 struct delayed_work *delay = to_delayed_work(work);
543 struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
544 stats_task);
545 struct mlx4_en_dev *mdev = priv->mdev;
546 int err;
548 err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
549 if (err)
550 en_dbg(HW, priv, "Could not update stats\n");
552 mutex_lock(&mdev->state_lock);
553 if (mdev->device_up) {
554 if (priv->port_up)
555 mlx4_en_auto_moderation(priv);
557 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
559 if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
560 queue_work(mdev->workqueue, &priv->mac_task);
561 mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
563 mutex_unlock(&mdev->state_lock);
566 static void mlx4_en_linkstate(struct work_struct *work)
568 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
569 linkstate_task);
570 struct mlx4_en_dev *mdev = priv->mdev;
571 int linkstate = priv->link_state;
573 mutex_lock(&mdev->state_lock);
574 /* If observable port state changed set carrier state and
575 * report to system log */
576 if (priv->last_link_state != linkstate) {
577 if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
578 en_info(priv, "Link Down\n");
579 netif_carrier_off(priv->dev);
580 } else {
581 en_info(priv, "Link Up\n");
582 netif_carrier_on(priv->dev);
585 priv->last_link_state = linkstate;
586 mutex_unlock(&mdev->state_lock);
590 int mlx4_en_start_port(struct net_device *dev)
592 struct mlx4_en_priv *priv = netdev_priv(dev);
593 struct mlx4_en_dev *mdev = priv->mdev;
594 struct mlx4_en_cq *cq;
595 struct mlx4_en_tx_ring *tx_ring;
596 int rx_index = 0;
597 int tx_index = 0;
598 int err = 0;
599 int i;
600 int j;
601 u8 mc_list[16] = {0};
603 if (priv->port_up) {
604 en_dbg(DRV, priv, "start port called while port already up\n");
605 return 0;
608 /* Calculate Rx buf size */
609 dev->mtu = min(dev->mtu, priv->max_mtu);
610 mlx4_en_calc_rx_buf(dev);
611 en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
613 /* Configure rx cq's and rings */
614 err = mlx4_en_activate_rx_rings(priv);
615 if (err) {
616 en_err(priv, "Failed to activate RX rings\n");
617 return err;
619 for (i = 0; i < priv->rx_ring_num; i++) {
620 cq = &priv->rx_cq[i];
622 err = mlx4_en_activate_cq(priv, cq, i);
623 if (err) {
624 en_err(priv, "Failed activating Rx CQ\n");
625 goto cq_err;
627 for (j = 0; j < cq->size; j++)
628 cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
629 err = mlx4_en_set_cq_moder(priv, cq);
630 if (err) {
631 en_err(priv, "Failed setting cq moderation parameters");
632 mlx4_en_deactivate_cq(priv, cq);
633 goto cq_err;
635 mlx4_en_arm_cq(priv, cq);
636 priv->rx_ring[i].cqn = cq->mcq.cqn;
637 ++rx_index;
640 /* Set qp number */
641 en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
642 err = mlx4_get_eth_qp(mdev->dev, priv->port,
643 priv->mac, &priv->base_qpn);
644 if (err) {
645 en_err(priv, "Failed getting eth qp\n");
646 goto cq_err;
648 mdev->mac_removed[priv->port] = 0;
650 err = mlx4_en_config_rss_steer(priv);
651 if (err) {
652 en_err(priv, "Failed configuring rss steering\n");
653 goto mac_err;
656 /* Configure tx cq's and rings */
657 for (i = 0; i < priv->tx_ring_num; i++) {
658 /* Configure cq */
659 cq = &priv->tx_cq[i];
660 err = mlx4_en_activate_cq(priv, cq, i);
661 if (err) {
662 en_err(priv, "Failed allocating Tx CQ\n");
663 goto tx_err;
665 err = mlx4_en_set_cq_moder(priv, cq);
666 if (err) {
667 en_err(priv, "Failed setting cq moderation parameters");
668 mlx4_en_deactivate_cq(priv, cq);
669 goto tx_err;
671 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
672 cq->buf->wqe_index = cpu_to_be16(0xffff);
674 /* Configure ring */
675 tx_ring = &priv->tx_ring[i];
676 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
677 i / priv->mdev->profile.num_tx_rings_p_up);
678 if (err) {
679 en_err(priv, "Failed allocating Tx ring\n");
680 mlx4_en_deactivate_cq(priv, cq);
681 goto tx_err;
683 tx_ring->tx_queue = netdev_get_tx_queue(dev, i);
685 /* Arm CQ for TX completions */
686 mlx4_en_arm_cq(priv, cq);
688 /* Set initial ownership of all Tx TXBBs to SW (1) */
689 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
690 *((u32 *) (tx_ring->buf + j)) = 0xffffffff;
691 ++tx_index;
694 /* Configure port */
695 err = mlx4_SET_PORT_general(mdev->dev, priv->port,
696 priv->rx_skb_size + ETH_FCS_LEN,
697 priv->prof->tx_pause,
698 priv->prof->tx_ppp,
699 priv->prof->rx_pause,
700 priv->prof->rx_ppp);
701 if (err) {
702 en_err(priv, "Failed setting port general configurations "
703 "for port %d, with error %d\n", priv->port, err);
704 goto tx_err;
706 /* Set default qp number */
707 err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
708 if (err) {
709 en_err(priv, "Failed setting default qp numbers\n");
710 goto tx_err;
713 /* Init port */
714 en_dbg(HW, priv, "Initializing port\n");
715 err = mlx4_INIT_PORT(mdev->dev, priv->port);
716 if (err) {
717 en_err(priv, "Failed Initializing port\n");
718 goto tx_err;
721 /* Attach rx QP to bradcast address */
722 memset(&mc_list[10], 0xff, ETH_ALEN);
723 mc_list[5] = priv->port;
724 if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
725 0, MLX4_PROT_ETH))
726 mlx4_warn(mdev, "Failed Attaching Broadcast\n");
728 /* Must redo promiscuous mode setup. */
729 priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
731 /* Schedule multicast task to populate multicast list */
732 queue_work(mdev->workqueue, &priv->mcast_task);
734 mlx4_set_stats_bitmap(mdev->dev, &priv->stats_bitmap);
736 priv->port_up = true;
737 netif_tx_start_all_queues(dev);
738 return 0;
740 tx_err:
741 while (tx_index--) {
742 mlx4_en_deactivate_tx_ring(priv, &priv->tx_ring[tx_index]);
743 mlx4_en_deactivate_cq(priv, &priv->tx_cq[tx_index]);
746 mlx4_en_release_rss_steer(priv);
747 mac_err:
748 mlx4_put_eth_qp(mdev->dev, priv->port, priv->mac, priv->base_qpn);
749 cq_err:
750 while (rx_index--)
751 mlx4_en_deactivate_cq(priv, &priv->rx_cq[rx_index]);
752 for (i = 0; i < priv->rx_ring_num; i++)
753 mlx4_en_deactivate_rx_ring(priv, &priv->rx_ring[i]);
755 return err; /* need to close devices */
759 void mlx4_en_stop_port(struct net_device *dev)
761 struct mlx4_en_priv *priv = netdev_priv(dev);
762 struct mlx4_en_dev *mdev = priv->mdev;
763 int i;
764 u8 mc_list[16] = {0};
766 if (!priv->port_up) {
767 en_dbg(DRV, priv, "stop port called while port already down\n");
768 return;
771 /* Synchronize with tx routine */
772 netif_tx_lock_bh(dev);
773 netif_tx_stop_all_queues(dev);
774 netif_tx_unlock_bh(dev);
776 /* Set port as not active */
777 priv->port_up = false;
779 /* Detach All multicasts */
780 memset(&mc_list[10], 0xff, ETH_ALEN);
781 mc_list[5] = priv->port;
782 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
783 MLX4_PROT_ETH);
784 for (i = 0; i < priv->mc_addrs_cnt; i++) {
785 memcpy(&mc_list[10], priv->mc_addrs + i * ETH_ALEN, ETH_ALEN);
786 mc_list[5] = priv->port;
787 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
788 mc_list, MLX4_PROT_ETH);
790 mlx4_en_clear_list(dev);
791 /* Flush multicast filter */
792 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
794 /* Free TX Rings */
795 for (i = 0; i < priv->tx_ring_num; i++) {
796 mlx4_en_deactivate_tx_ring(priv, &priv->tx_ring[i]);
797 mlx4_en_deactivate_cq(priv, &priv->tx_cq[i]);
799 msleep(10);
801 for (i = 0; i < priv->tx_ring_num; i++)
802 mlx4_en_free_tx_buf(dev, &priv->tx_ring[i]);
804 /* Free RSS qps */
805 mlx4_en_release_rss_steer(priv);
807 /* Unregister Mac address for the port */
808 mlx4_put_eth_qp(mdev->dev, priv->port, priv->mac, priv->base_qpn);
809 mdev->mac_removed[priv->port] = 1;
811 /* Free RX Rings */
812 for (i = 0; i < priv->rx_ring_num; i++) {
813 mlx4_en_deactivate_rx_ring(priv, &priv->rx_ring[i]);
814 while (test_bit(NAPI_STATE_SCHED, &priv->rx_cq[i].napi.state))
815 msleep(1);
816 mlx4_en_deactivate_cq(priv, &priv->rx_cq[i]);
819 /* close port*/
820 mlx4_CLOSE_PORT(mdev->dev, priv->port);
823 static void mlx4_en_restart(struct work_struct *work)
825 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
826 watchdog_task);
827 struct mlx4_en_dev *mdev = priv->mdev;
828 struct net_device *dev = priv->dev;
829 int i;
831 en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
833 mutex_lock(&mdev->state_lock);
834 if (priv->port_up) {
835 mlx4_en_stop_port(dev);
836 for (i = 0; i < priv->tx_ring_num; i++)
837 netdev_tx_reset_queue(priv->tx_ring[i].tx_queue);
838 if (mlx4_en_start_port(dev))
839 en_err(priv, "Failed restarting port %d\n", priv->port);
841 mutex_unlock(&mdev->state_lock);
844 static void mlx4_en_clear_stats(struct net_device *dev)
846 struct mlx4_en_priv *priv = netdev_priv(dev);
847 struct mlx4_en_dev *mdev = priv->mdev;
848 int i;
850 if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
851 en_dbg(HW, priv, "Failed dumping statistics\n");
853 memset(&priv->stats, 0, sizeof(priv->stats));
854 memset(&priv->pstats, 0, sizeof(priv->pstats));
855 memset(&priv->pkstats, 0, sizeof(priv->pkstats));
856 memset(&priv->port_stats, 0, sizeof(priv->port_stats));
858 for (i = 0; i < priv->tx_ring_num; i++) {
859 priv->tx_ring[i].bytes = 0;
860 priv->tx_ring[i].packets = 0;
861 priv->tx_ring[i].tx_csum = 0;
863 for (i = 0; i < priv->rx_ring_num; i++) {
864 priv->rx_ring[i].bytes = 0;
865 priv->rx_ring[i].packets = 0;
866 priv->rx_ring[i].csum_ok = 0;
867 priv->rx_ring[i].csum_none = 0;
871 static int mlx4_en_open(struct net_device *dev)
873 struct mlx4_en_priv *priv = netdev_priv(dev);
874 struct mlx4_en_dev *mdev = priv->mdev;
875 int err = 0;
877 mutex_lock(&mdev->state_lock);
879 if (!mdev->device_up) {
880 en_err(priv, "Cannot open - device down/disabled\n");
881 err = -EBUSY;
882 goto out;
885 /* Reset HW statistics and SW counters */
886 mlx4_en_clear_stats(dev);
888 err = mlx4_en_start_port(dev);
889 if (err)
890 en_err(priv, "Failed starting port:%d\n", priv->port);
892 out:
893 mutex_unlock(&mdev->state_lock);
894 return err;
898 static int mlx4_en_close(struct net_device *dev)
900 struct mlx4_en_priv *priv = netdev_priv(dev);
901 struct mlx4_en_dev *mdev = priv->mdev;
903 en_dbg(IFDOWN, priv, "Close port called\n");
905 mutex_lock(&mdev->state_lock);
907 mlx4_en_stop_port(dev);
908 netif_carrier_off(dev);
910 mutex_unlock(&mdev->state_lock);
911 return 0;
914 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
916 int i;
918 for (i = 0; i < priv->tx_ring_num; i++) {
919 if (priv->tx_ring[i].tx_info)
920 mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
921 if (priv->tx_cq[i].buf)
922 mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
925 for (i = 0; i < priv->rx_ring_num; i++) {
926 if (priv->rx_ring[i].rx_info)
927 mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
928 priv->prof->rx_ring_size, priv->stride);
929 if (priv->rx_cq[i].buf)
930 mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
934 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
936 struct mlx4_en_port_profile *prof = priv->prof;
937 int i;
938 int base_tx_qpn, err;
940 err = mlx4_qp_reserve_range(priv->mdev->dev, priv->tx_ring_num, 256, &base_tx_qpn);
941 if (err) {
942 en_err(priv, "failed reserving range for TX rings\n");
943 return err;
946 /* Create tx Rings */
947 for (i = 0; i < priv->tx_ring_num; i++) {
948 if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
949 prof->tx_ring_size, i, TX))
950 goto err;
952 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i], base_tx_qpn + i,
953 prof->tx_ring_size, TXBB_SIZE))
954 goto err;
957 /* Create rx Rings */
958 for (i = 0; i < priv->rx_ring_num; i++) {
959 if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
960 prof->rx_ring_size, i, RX))
961 goto err;
963 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
964 prof->rx_ring_size, priv->stride))
965 goto err;
968 return 0;
970 err:
971 en_err(priv, "Failed to allocate NIC resources\n");
972 mlx4_qp_release_range(priv->mdev->dev, base_tx_qpn, priv->tx_ring_num);
973 return -ENOMEM;
977 void mlx4_en_destroy_netdev(struct net_device *dev)
979 struct mlx4_en_priv *priv = netdev_priv(dev);
980 struct mlx4_en_dev *mdev = priv->mdev;
982 en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
984 /* Unregister device - this will close the port if it was up */
985 if (priv->registered)
986 unregister_netdev(dev);
988 if (priv->allocated)
989 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
991 cancel_delayed_work(&priv->stats_task);
992 /* flush any pending task for this netdev */
993 flush_workqueue(mdev->workqueue);
995 /* Detach the netdev so tasks would not attempt to access it */
996 mutex_lock(&mdev->state_lock);
997 mdev->pndev[priv->port] = NULL;
998 mutex_unlock(&mdev->state_lock);
1000 mlx4_en_free_resources(priv);
1002 kfree(priv->tx_ring);
1003 kfree(priv->tx_cq);
1005 free_netdev(dev);
1008 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
1010 struct mlx4_en_priv *priv = netdev_priv(dev);
1011 struct mlx4_en_dev *mdev = priv->mdev;
1012 int err = 0;
1014 en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
1015 dev->mtu, new_mtu);
1017 if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1018 en_err(priv, "Bad MTU size:%d.\n", new_mtu);
1019 return -EPERM;
1021 dev->mtu = new_mtu;
1023 if (netif_running(dev)) {
1024 mutex_lock(&mdev->state_lock);
1025 if (!mdev->device_up) {
1026 /* NIC is probably restarting - let watchdog task reset
1027 * the port */
1028 en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1029 } else {
1030 mlx4_en_stop_port(dev);
1031 err = mlx4_en_start_port(dev);
1032 if (err) {
1033 en_err(priv, "Failed restarting port:%d\n",
1034 priv->port);
1035 queue_work(mdev->workqueue, &priv->watchdog_task);
1038 mutex_unlock(&mdev->state_lock);
1040 return 0;
1043 static int mlx4_en_set_features(struct net_device *netdev,
1044 netdev_features_t features)
1046 struct mlx4_en_priv *priv = netdev_priv(netdev);
1048 if (features & NETIF_F_LOOPBACK)
1049 priv->ctrl_flags |= cpu_to_be32(MLX4_WQE_CTRL_FORCE_LOOPBACK);
1050 else
1051 priv->ctrl_flags &=
1052 cpu_to_be32(~MLX4_WQE_CTRL_FORCE_LOOPBACK);
1054 return 0;
1058 static const struct net_device_ops mlx4_netdev_ops = {
1059 .ndo_open = mlx4_en_open,
1060 .ndo_stop = mlx4_en_close,
1061 .ndo_start_xmit = mlx4_en_xmit,
1062 .ndo_select_queue = mlx4_en_select_queue,
1063 .ndo_get_stats = mlx4_en_get_stats,
1064 .ndo_set_rx_mode = mlx4_en_set_multicast,
1065 .ndo_set_mac_address = mlx4_en_set_mac,
1066 .ndo_validate_addr = eth_validate_addr,
1067 .ndo_change_mtu = mlx4_en_change_mtu,
1068 .ndo_tx_timeout = mlx4_en_tx_timeout,
1069 .ndo_vlan_rx_add_vid = mlx4_en_vlan_rx_add_vid,
1070 .ndo_vlan_rx_kill_vid = mlx4_en_vlan_rx_kill_vid,
1071 #ifdef CONFIG_NET_POLL_CONTROLLER
1072 .ndo_poll_controller = mlx4_en_netpoll,
1073 #endif
1074 .ndo_set_features = mlx4_en_set_features,
1075 .ndo_setup_tc = mlx4_en_setup_tc,
1078 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
1079 struct mlx4_en_port_profile *prof)
1081 struct net_device *dev;
1082 struct mlx4_en_priv *priv;
1083 int i;
1084 int err;
1086 dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
1087 prof->tx_ring_num, prof->rx_ring_num);
1088 if (dev == NULL)
1089 return -ENOMEM;
1091 SET_NETDEV_DEV(dev, &mdev->dev->pdev->dev);
1092 dev->dev_id = port - 1;
1095 * Initialize driver private data
1098 priv = netdev_priv(dev);
1099 memset(priv, 0, sizeof(struct mlx4_en_priv));
1100 priv->dev = dev;
1101 priv->mdev = mdev;
1102 priv->ddev = &mdev->pdev->dev;
1103 priv->prof = prof;
1104 priv->port = port;
1105 priv->port_up = false;
1106 priv->flags = prof->flags;
1107 priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
1108 MLX4_WQE_CTRL_SOLICITED);
1109 priv->tx_ring_num = prof->tx_ring_num;
1110 priv->tx_ring = kzalloc(sizeof(struct mlx4_en_tx_ring) *
1111 priv->tx_ring_num, GFP_KERNEL);
1112 if (!priv->tx_ring) {
1113 err = -ENOMEM;
1114 goto out;
1116 priv->tx_cq = kzalloc(sizeof(struct mlx4_en_cq) * priv->tx_ring_num,
1117 GFP_KERNEL);
1118 if (!priv->tx_cq) {
1119 err = -ENOMEM;
1120 goto out;
1122 priv->rx_ring_num = prof->rx_ring_num;
1123 priv->mac_index = -1;
1124 priv->msg_enable = MLX4_EN_MSG_LEVEL;
1125 spin_lock_init(&priv->stats_lock);
1126 INIT_WORK(&priv->mcast_task, mlx4_en_do_set_multicast);
1127 INIT_WORK(&priv->mac_task, mlx4_en_do_set_mac);
1128 INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
1129 INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
1130 INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
1131 #ifdef CONFIG_MLX4_EN_DCB
1132 if (!mlx4_is_slave(priv->mdev->dev))
1133 dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
1134 #endif
1136 /* Query for default mac and max mtu */
1137 priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
1138 priv->mac = mdev->dev->caps.def_mac[priv->port];
1139 if (ILLEGAL_MAC(priv->mac)) {
1140 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
1141 priv->port, priv->mac);
1142 err = -EINVAL;
1143 goto out;
1146 priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
1147 DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
1148 err = mlx4_en_alloc_resources(priv);
1149 if (err)
1150 goto out;
1152 /* Allocate page for receive rings */
1153 err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
1154 MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
1155 if (err) {
1156 en_err(priv, "Failed to allocate page for rx qps\n");
1157 goto out;
1159 priv->allocated = 1;
1162 * Initialize netdev entry points
1164 dev->netdev_ops = &mlx4_netdev_ops;
1165 dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
1166 netif_set_real_num_tx_queues(dev, priv->tx_ring_num);
1167 netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
1169 SET_ETHTOOL_OPS(dev, &mlx4_en_ethtool_ops);
1171 /* Set defualt MAC */
1172 dev->addr_len = ETH_ALEN;
1173 for (i = 0; i < ETH_ALEN; i++) {
1174 dev->dev_addr[ETH_ALEN - 1 - i] = (u8) (priv->mac >> (8 * i));
1175 dev->perm_addr[ETH_ALEN - 1 - i] = (u8) (priv->mac >> (8 * i));
1179 * Set driver features
1181 dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
1182 if (mdev->LSO_support)
1183 dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
1185 dev->vlan_features = dev->hw_features;
1187 dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
1188 dev->features = dev->hw_features | NETIF_F_HIGHDMA |
1189 NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX |
1190 NETIF_F_HW_VLAN_FILTER;
1191 dev->hw_features |= NETIF_F_LOOPBACK;
1193 mdev->pndev[port] = dev;
1195 netif_carrier_off(dev);
1196 err = register_netdev(dev);
1197 if (err) {
1198 en_err(priv, "Netdev registration failed for port %d\n", port);
1199 goto out;
1201 priv->registered = 1;
1203 en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
1204 en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
1206 /* Configure port */
1207 err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1208 MLX4_EN_MIN_MTU,
1209 0, 0, 0, 0);
1210 if (err) {
1211 en_err(priv, "Failed setting port general configurations "
1212 "for port %d, with error %d\n", priv->port, err);
1213 goto out;
1216 /* Init port */
1217 en_warn(priv, "Initializing port\n");
1218 err = mlx4_INIT_PORT(mdev->dev, priv->port);
1219 if (err) {
1220 en_err(priv, "Failed Initializing port\n");
1221 goto out;
1223 mlx4_en_set_default_moderation(priv);
1224 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1225 return 0;
1227 out:
1228 mlx4_en_destroy_netdev(dev);
1229 return err;