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>
49 static void mlx4_en_vlan_rx_register(struct net_device
*dev
, struct vlan_group
*grp
)
51 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
52 struct mlx4_en_dev
*mdev
= priv
->mdev
;
55 en_dbg(HW
, priv
, "Registering VLAN group:%p\n", grp
);
58 mutex_lock(&mdev
->state_lock
);
59 if (mdev
->device_up
&& priv
->port_up
) {
60 err
= mlx4_SET_VLAN_FLTR(mdev
->dev
, priv
->port
, grp
);
62 en_err(priv
, "Failed configuring VLAN filter\n");
64 mutex_unlock(&mdev
->state_lock
);
67 static void mlx4_en_vlan_rx_add_vid(struct net_device
*dev
, unsigned short vid
)
69 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
70 struct mlx4_en_dev
*mdev
= priv
->mdev
;
76 en_dbg(HW
, priv
, "adding VLAN:%d (vlgrp entry:%p)\n",
77 vid
, vlan_group_get_device(priv
->vlgrp
, vid
));
79 /* Add VID to port VLAN filter */
80 mutex_lock(&mdev
->state_lock
);
81 if (mdev
->device_up
&& priv
->port_up
) {
82 err
= mlx4_SET_VLAN_FLTR(mdev
->dev
, priv
->port
, priv
->vlgrp
);
84 en_err(priv
, "Failed configuring VLAN filter\n");
86 mutex_unlock(&mdev
->state_lock
);
89 static void mlx4_en_vlan_rx_kill_vid(struct net_device
*dev
, unsigned short vid
)
91 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
92 struct mlx4_en_dev
*mdev
= priv
->mdev
;
98 en_dbg(HW
, priv
, "Killing VID:%d (vlgrp:%p vlgrp entry:%p)\n",
99 vid
, priv
->vlgrp
, vlan_group_get_device(priv
->vlgrp
, vid
));
100 vlan_group_set_device(priv
->vlgrp
, vid
, NULL
);
102 /* Remove VID from port VLAN filter */
103 mutex_lock(&mdev
->state_lock
);
104 if (mdev
->device_up
&& priv
->port_up
) {
105 err
= mlx4_SET_VLAN_FLTR(mdev
->dev
, priv
->port
, priv
->vlgrp
);
107 en_err(priv
, "Failed configuring VLAN filter\n");
109 mutex_unlock(&mdev
->state_lock
);
112 static u64
mlx4_en_mac_to_u64(u8
*addr
)
117 for (i
= 0; i
< ETH_ALEN
; i
++) {
124 static int mlx4_en_set_mac(struct net_device
*dev
, void *addr
)
126 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
127 struct mlx4_en_dev
*mdev
= priv
->mdev
;
128 struct sockaddr
*saddr
= addr
;
130 if (!is_valid_ether_addr(saddr
->sa_data
))
131 return -EADDRNOTAVAIL
;
133 memcpy(dev
->dev_addr
, saddr
->sa_data
, ETH_ALEN
);
134 priv
->mac
= mlx4_en_mac_to_u64(dev
->dev_addr
);
135 queue_work(mdev
->workqueue
, &priv
->mac_task
);
139 static void mlx4_en_do_set_mac(struct work_struct
*work
)
141 struct mlx4_en_priv
*priv
= container_of(work
, struct mlx4_en_priv
,
143 struct mlx4_en_dev
*mdev
= priv
->mdev
;
146 mutex_lock(&mdev
->state_lock
);
148 /* Remove old MAC and insert the new one */
149 mlx4_unregister_mac(mdev
->dev
, priv
->port
, priv
->mac_index
);
150 err
= mlx4_register_mac(mdev
->dev
, priv
->port
,
151 priv
->mac
, &priv
->mac_index
);
153 en_err(priv
, "Failed changing HW MAC address\n");
155 en_dbg(HW
, priv
, "Port is down while "
156 "registering mac, exiting...\n");
158 mutex_unlock(&mdev
->state_lock
);
161 static void mlx4_en_clear_list(struct net_device
*dev
)
163 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
165 kfree(priv
->mc_addrs
);
166 priv
->mc_addrs_cnt
= 0;
169 static void mlx4_en_cache_mclist(struct net_device
*dev
)
171 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
172 struct netdev_hw_addr
*ha
;
174 int mc_addrs_cnt
= netdev_mc_count(dev
);
177 mc_addrs
= kmalloc(mc_addrs_cnt
* ETH_ALEN
, GFP_ATOMIC
);
179 en_err(priv
, "failed to allocate multicast list\n");
183 netdev_for_each_mc_addr(ha
, dev
)
184 memcpy(mc_addrs
+ i
++ * ETH_ALEN
, ha
->addr
, ETH_ALEN
);
185 priv
->mc_addrs
= mc_addrs
;
186 priv
->mc_addrs_cnt
= mc_addrs_cnt
;
190 static void mlx4_en_set_multicast(struct net_device
*dev
)
192 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
197 queue_work(priv
->mdev
->workqueue
, &priv
->mcast_task
);
200 static void mlx4_en_do_set_multicast(struct work_struct
*work
)
202 struct mlx4_en_priv
*priv
= container_of(work
, struct mlx4_en_priv
,
204 struct mlx4_en_dev
*mdev
= priv
->mdev
;
205 struct net_device
*dev
= priv
->dev
;
209 mutex_lock(&mdev
->state_lock
);
210 if (!mdev
->device_up
) {
211 en_dbg(HW
, priv
, "Card is not up, "
212 "ignoring multicast change.\n");
215 if (!priv
->port_up
) {
216 en_dbg(HW
, priv
, "Port is down, "
217 "ignoring multicast change.\n");
222 * Promsicuous mode: disable all filters
225 if (dev
->flags
& IFF_PROMISC
) {
226 if (!(priv
->flags
& MLX4_EN_FLAG_PROMISC
)) {
227 if (netif_msg_rx_status(priv
))
228 en_warn(priv
, "Entering promiscuous mode\n");
229 priv
->flags
|= MLX4_EN_FLAG_PROMISC
;
231 /* Enable promiscouos mode */
232 err
= mlx4_SET_PORT_qpn_calc(mdev
->dev
, priv
->port
,
235 en_err(priv
, "Failed enabling "
236 "promiscous mode\n");
238 /* Disable port multicast filter (unconditionally) */
239 err
= mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
, 0,
240 0, MLX4_MCAST_DISABLE
);
242 en_err(priv
, "Failed disabling "
243 "multicast filter\n");
245 /* Disable port VLAN filter */
246 err
= mlx4_SET_VLAN_FLTR(mdev
->dev
, priv
->port
, NULL
);
248 en_err(priv
, "Failed disabling VLAN filter\n");
254 * Not in promiscous mode
257 if (priv
->flags
& MLX4_EN_FLAG_PROMISC
) {
258 if (netif_msg_rx_status(priv
))
259 en_warn(priv
, "Leaving promiscuous mode\n");
260 priv
->flags
&= ~MLX4_EN_FLAG_PROMISC
;
262 /* Disable promiscouos mode */
263 err
= mlx4_SET_PORT_qpn_calc(mdev
->dev
, priv
->port
,
266 en_err(priv
, "Failed disabling promiscous mode\n");
268 /* Enable port VLAN filter */
269 err
= mlx4_SET_VLAN_FLTR(mdev
->dev
, priv
->port
, priv
->vlgrp
);
271 en_err(priv
, "Failed enabling VLAN filter\n");
274 /* Enable/disable the multicast filter according to IFF_ALLMULTI */
275 if (dev
->flags
& IFF_ALLMULTI
) {
276 err
= mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
, 0,
277 0, MLX4_MCAST_DISABLE
);
279 en_err(priv
, "Failed disabling multicast filter\n");
283 err
= mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
, 0,
284 0, MLX4_MCAST_DISABLE
);
286 en_err(priv
, "Failed disabling multicast filter\n");
288 /* Flush mcast filter and init it with broadcast address */
289 mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
, ETH_BCAST
,
290 1, MLX4_MCAST_CONFIG
);
292 /* Update multicast list - we cache all addresses so they won't
293 * change while HW is updated holding the command semaphor */
294 netif_tx_lock_bh(dev
);
295 mlx4_en_cache_mclist(dev
);
296 netif_tx_unlock_bh(dev
);
297 for (i
= 0; i
< priv
->mc_addrs_cnt
; i
++) {
299 mlx4_en_mac_to_u64(priv
->mc_addrs
+ i
* ETH_ALEN
);
300 mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
,
301 mcast_addr
, 0, MLX4_MCAST_CONFIG
);
303 err
= mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
, 0,
304 0, MLX4_MCAST_ENABLE
);
306 en_err(priv
, "Failed enabling multicast filter\n");
308 mlx4_en_clear_list(dev
);
311 mutex_unlock(&mdev
->state_lock
);
314 #ifdef CONFIG_NET_POLL_CONTROLLER
315 static void mlx4_en_netpoll(struct net_device
*dev
)
317 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
318 struct mlx4_en_cq
*cq
;
322 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
323 cq
= &priv
->rx_cq
[i
];
324 spin_lock_irqsave(&cq
->lock
, flags
);
325 napi_synchronize(&cq
->napi
);
326 mlx4_en_process_rx_cq(dev
, cq
, 0);
327 spin_unlock_irqrestore(&cq
->lock
, flags
);
332 static void mlx4_en_tx_timeout(struct net_device
*dev
)
334 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
335 struct mlx4_en_dev
*mdev
= priv
->mdev
;
337 if (netif_msg_timer(priv
))
338 en_warn(priv
, "Tx timeout called on port:%d\n", priv
->port
);
340 priv
->port_stats
.tx_timeout
++;
341 en_dbg(DRV
, priv
, "Scheduling watchdog\n");
342 queue_work(mdev
->workqueue
, &priv
->watchdog_task
);
346 static struct net_device_stats
*mlx4_en_get_stats(struct net_device
*dev
)
348 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
350 spin_lock_bh(&priv
->stats_lock
);
351 memcpy(&priv
->ret_stats
, &priv
->stats
, sizeof(priv
->stats
));
352 spin_unlock_bh(&priv
->stats_lock
);
354 return &priv
->ret_stats
;
357 static void mlx4_en_set_default_moderation(struct mlx4_en_priv
*priv
)
359 struct mlx4_en_cq
*cq
;
362 /* If we haven't received a specific coalescing setting
363 * (module param), we set the moderation parameters as follows:
364 * - moder_cnt is set to the number of mtu sized packets to
365 * satisfy our coelsing target.
366 * - moder_time is set to a fixed value.
368 priv
->rx_frames
= MLX4_EN_RX_COAL_TARGET
;
369 priv
->rx_usecs
= MLX4_EN_RX_COAL_TIME
;
370 en_dbg(INTR
, priv
, "Default coalesing params for mtu:%d - "
371 "rx_frames:%d rx_usecs:%d\n",
372 priv
->dev
->mtu
, priv
->rx_frames
, priv
->rx_usecs
);
374 /* Setup cq moderation params */
375 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
376 cq
= &priv
->rx_cq
[i
];
377 cq
->moder_cnt
= priv
->rx_frames
;
378 cq
->moder_time
= priv
->rx_usecs
;
381 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
382 cq
= &priv
->tx_cq
[i
];
383 cq
->moder_cnt
= MLX4_EN_TX_COAL_PKTS
;
384 cq
->moder_time
= MLX4_EN_TX_COAL_TIME
;
387 /* Reset auto-moderation params */
388 priv
->pkt_rate_low
= MLX4_EN_RX_RATE_LOW
;
389 priv
->rx_usecs_low
= MLX4_EN_RX_COAL_TIME_LOW
;
390 priv
->pkt_rate_high
= MLX4_EN_RX_RATE_HIGH
;
391 priv
->rx_usecs_high
= MLX4_EN_RX_COAL_TIME_HIGH
;
392 priv
->sample_interval
= MLX4_EN_SAMPLE_INTERVAL
;
393 priv
->adaptive_rx_coal
= 1;
394 priv
->last_moder_time
= MLX4_EN_AUTO_CONF
;
395 priv
->last_moder_jiffies
= 0;
396 priv
->last_moder_packets
= 0;
397 priv
->last_moder_tx_packets
= 0;
398 priv
->last_moder_bytes
= 0;
401 static void mlx4_en_auto_moderation(struct mlx4_en_priv
*priv
)
403 unsigned long period
= (unsigned long) (jiffies
- priv
->last_moder_jiffies
);
404 struct mlx4_en_cq
*cq
;
405 unsigned long packets
;
407 unsigned long avg_pkt_size
;
408 unsigned long rx_packets
;
409 unsigned long rx_bytes
;
410 unsigned long rx_byte_diff
;
411 unsigned long tx_packets
;
412 unsigned long tx_pkt_diff
;
413 unsigned long rx_pkt_diff
;
417 if (!priv
->adaptive_rx_coal
|| period
< priv
->sample_interval
* HZ
)
420 spin_lock_bh(&priv
->stats_lock
);
421 rx_packets
= priv
->stats
.rx_packets
;
422 rx_bytes
= priv
->stats
.rx_bytes
;
423 tx_packets
= priv
->stats
.tx_packets
;
424 spin_unlock_bh(&priv
->stats_lock
);
426 if (!priv
->last_moder_jiffies
|| !period
)
429 tx_pkt_diff
= ((unsigned long) (tx_packets
-
430 priv
->last_moder_tx_packets
));
431 rx_pkt_diff
= ((unsigned long) (rx_packets
-
432 priv
->last_moder_packets
));
433 packets
= max(tx_pkt_diff
, rx_pkt_diff
);
434 rx_byte_diff
= rx_bytes
- priv
->last_moder_bytes
;
435 rx_byte_diff
= rx_byte_diff
? rx_byte_diff
: 1;
436 rate
= packets
* HZ
/ period
;
437 avg_pkt_size
= packets
? ((unsigned long) (rx_bytes
-
438 priv
->last_moder_bytes
)) / packets
: 0;
440 /* Apply auto-moderation only when packet rate exceeds a rate that
442 if (rate
> MLX4_EN_RX_RATE_THRESH
) {
443 /* If tx and rx packet rates are not balanced, assume that
444 * traffic is mainly BW bound and apply maximum moderation.
445 * Otherwise, moderate according to packet rate */
446 if (2 * tx_pkt_diff
> 3 * rx_pkt_diff
&&
447 rx_pkt_diff
/ rx_byte_diff
<
448 MLX4_EN_SMALL_PKT_SIZE
)
449 moder_time
= priv
->rx_usecs_low
;
450 else if (2 * rx_pkt_diff
> 3 * tx_pkt_diff
)
451 moder_time
= priv
->rx_usecs_high
;
453 if (rate
< priv
->pkt_rate_low
)
454 moder_time
= priv
->rx_usecs_low
;
455 else if (rate
> priv
->pkt_rate_high
)
456 moder_time
= priv
->rx_usecs_high
;
458 moder_time
= (rate
- priv
->pkt_rate_low
) *
459 (priv
->rx_usecs_high
- priv
->rx_usecs_low
) /
460 (priv
->pkt_rate_high
- priv
->pkt_rate_low
) +
464 /* When packet rate is low, use default moderation rather than
465 * 0 to prevent interrupt storms if traffic suddenly increases */
466 moder_time
= priv
->rx_usecs
;
469 en_dbg(INTR
, priv
, "tx rate:%lu rx_rate:%lu\n",
470 tx_pkt_diff
* HZ
/ period
, rx_pkt_diff
* HZ
/ period
);
472 en_dbg(INTR
, priv
, "Rx moder_time changed from:%d to %d period:%lu "
473 "[jiff] packets:%lu avg_pkt_size:%lu rate:%lu [p/s])\n",
474 priv
->last_moder_time
, moder_time
, period
, packets
,
477 if (moder_time
!= priv
->last_moder_time
) {
478 priv
->last_moder_time
= moder_time
;
479 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
480 cq
= &priv
->rx_cq
[i
];
481 cq
->moder_time
= moder_time
;
482 err
= mlx4_en_set_cq_moder(priv
, cq
);
484 en_err(priv
, "Failed modifying moderation for cq:%d\n", i
);
491 priv
->last_moder_packets
= rx_packets
;
492 priv
->last_moder_tx_packets
= tx_packets
;
493 priv
->last_moder_bytes
= rx_bytes
;
494 priv
->last_moder_jiffies
= jiffies
;
497 static void mlx4_en_do_get_stats(struct work_struct
*work
)
499 struct delayed_work
*delay
= to_delayed_work(work
);
500 struct mlx4_en_priv
*priv
= container_of(delay
, struct mlx4_en_priv
,
502 struct mlx4_en_dev
*mdev
= priv
->mdev
;
505 err
= mlx4_en_DUMP_ETH_STATS(mdev
, priv
->port
, 0);
507 en_dbg(HW
, priv
, "Could not update stats\n");
509 mutex_lock(&mdev
->state_lock
);
510 if (mdev
->device_up
) {
512 mlx4_en_auto_moderation(priv
);
514 queue_delayed_work(mdev
->workqueue
, &priv
->stats_task
, STATS_DELAY
);
516 mutex_unlock(&mdev
->state_lock
);
519 static void mlx4_en_linkstate(struct work_struct
*work
)
521 struct mlx4_en_priv
*priv
= container_of(work
, struct mlx4_en_priv
,
523 struct mlx4_en_dev
*mdev
= priv
->mdev
;
524 int linkstate
= priv
->link_state
;
526 mutex_lock(&mdev
->state_lock
);
527 /* If observable port state changed set carrier state and
528 * report to system log */
529 if (priv
->last_link_state
!= linkstate
) {
530 if (linkstate
== MLX4_DEV_EVENT_PORT_DOWN
) {
531 en_dbg(LINK
, priv
, "Link Down\n");
532 netif_carrier_off(priv
->dev
);
534 en_dbg(LINK
, priv
, "Link Up\n");
535 netif_carrier_on(priv
->dev
);
538 priv
->last_link_state
= linkstate
;
539 mutex_unlock(&mdev
->state_lock
);
543 int mlx4_en_start_port(struct net_device
*dev
)
545 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
546 struct mlx4_en_dev
*mdev
= priv
->mdev
;
547 struct mlx4_en_cq
*cq
;
548 struct mlx4_en_tx_ring
*tx_ring
;
556 en_dbg(DRV
, priv
, "start port called while port already up\n");
560 /* Calculate Rx buf size */
561 dev
->mtu
= min(dev
->mtu
, priv
->max_mtu
);
562 mlx4_en_calc_rx_buf(dev
);
563 en_dbg(DRV
, priv
, "Rx buf size:%d\n", priv
->rx_skb_size
);
565 /* Configure rx cq's and rings */
566 err
= mlx4_en_activate_rx_rings(priv
);
568 en_err(priv
, "Failed to activate RX rings\n");
571 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
572 cq
= &priv
->rx_cq
[i
];
574 err
= mlx4_en_activate_cq(priv
, cq
);
576 en_err(priv
, "Failed activating Rx CQ\n");
579 for (j
= 0; j
< cq
->size
; j
++)
580 cq
->buf
[j
].owner_sr_opcode
= MLX4_CQE_OWNER_MASK
;
581 err
= mlx4_en_set_cq_moder(priv
, cq
);
583 en_err(priv
, "Failed setting cq moderation parameters");
584 mlx4_en_deactivate_cq(priv
, cq
);
587 mlx4_en_arm_cq(priv
, cq
);
588 priv
->rx_ring
[i
].cqn
= cq
->mcq
.cqn
;
592 err
= mlx4_en_config_rss_steer(priv
);
594 en_err(priv
, "Failed configuring rss steering\n");
598 /* Configure tx cq's and rings */
599 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
601 cq
= &priv
->tx_cq
[i
];
602 err
= mlx4_en_activate_cq(priv
, cq
);
604 en_err(priv
, "Failed allocating Tx CQ\n");
607 err
= mlx4_en_set_cq_moder(priv
, cq
);
609 en_err(priv
, "Failed setting cq moderation parameters");
610 mlx4_en_deactivate_cq(priv
, cq
);
613 en_dbg(DRV
, priv
, "Resetting index of collapsed CQ:%d to -1\n", i
);
614 cq
->buf
->wqe_index
= cpu_to_be16(0xffff);
617 tx_ring
= &priv
->tx_ring
[i
];
618 err
= mlx4_en_activate_tx_ring(priv
, tx_ring
, cq
->mcq
.cqn
);
620 en_err(priv
, "Failed allocating Tx ring\n");
621 mlx4_en_deactivate_cq(priv
, cq
);
624 /* Set initial ownership of all Tx TXBBs to SW (1) */
625 for (j
= 0; j
< tx_ring
->buf_size
; j
+= STAMP_STRIDE
)
626 *((u32
*) (tx_ring
->buf
+ j
)) = 0xffffffff;
631 err
= mlx4_SET_PORT_general(mdev
->dev
, priv
->port
,
632 priv
->rx_skb_size
+ ETH_FCS_LEN
,
633 priv
->prof
->tx_pause
,
635 priv
->prof
->rx_pause
,
638 en_err(priv
, "Failed setting port general configurations "
639 "for port %d, with error %d\n", priv
->port
, err
);
642 /* Set default qp number */
643 err
= mlx4_SET_PORT_qpn_calc(mdev
->dev
, priv
->port
, priv
->base_qpn
, 0);
645 en_err(priv
, "Failed setting default qp numbers\n");
648 /* Set port mac number */
649 en_dbg(DRV
, priv
, "Setting mac for port %d\n", priv
->port
);
650 err
= mlx4_register_mac(mdev
->dev
, priv
->port
,
651 priv
->mac
, &priv
->mac_index
);
653 en_err(priv
, "Failed setting port mac\n");
658 en_dbg(HW
, priv
, "Initializing port\n");
659 err
= mlx4_INIT_PORT(mdev
->dev
, priv
->port
);
661 en_err(priv
, "Failed Initializing port\n");
665 /* Schedule multicast task to populate multicast list */
666 queue_work(mdev
->workqueue
, &priv
->mcast_task
);
668 priv
->port_up
= true;
669 netif_tx_start_all_queues(dev
);
673 mlx4_unregister_mac(mdev
->dev
, priv
->port
, priv
->mac_index
);
676 mlx4_en_deactivate_tx_ring(priv
, &priv
->tx_ring
[tx_index
]);
677 mlx4_en_deactivate_cq(priv
, &priv
->tx_cq
[tx_index
]);
680 mlx4_en_release_rss_steer(priv
);
683 mlx4_en_deactivate_cq(priv
, &priv
->rx_cq
[rx_index
]);
684 for (i
= 0; i
< priv
->rx_ring_num
; i
++)
685 mlx4_en_deactivate_rx_ring(priv
, &priv
->rx_ring
[i
]);
687 return err
; /* need to close devices */
691 void mlx4_en_stop_port(struct net_device
*dev
)
693 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
694 struct mlx4_en_dev
*mdev
= priv
->mdev
;
697 if (!priv
->port_up
) {
698 en_dbg(DRV
, priv
, "stop port called while port already down\n");
702 /* Synchronize with tx routine */
703 netif_tx_lock_bh(dev
);
704 netif_tx_stop_all_queues(dev
);
705 netif_tx_unlock_bh(dev
);
708 priv
->port_up
= false;
709 mlx4_CLOSE_PORT(mdev
->dev
, priv
->port
);
711 /* Unregister Mac address for the port */
712 mlx4_unregister_mac(mdev
->dev
, priv
->port
, priv
->mac_index
);
715 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
716 mlx4_en_deactivate_tx_ring(priv
, &priv
->tx_ring
[i
]);
717 mlx4_en_deactivate_cq(priv
, &priv
->tx_cq
[i
]);
721 for (i
= 0; i
< priv
->tx_ring_num
; i
++)
722 mlx4_en_free_tx_buf(dev
, &priv
->tx_ring
[i
]);
725 mlx4_en_release_rss_steer(priv
);
728 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
729 mlx4_en_deactivate_rx_ring(priv
, &priv
->rx_ring
[i
]);
730 while (test_bit(NAPI_STATE_SCHED
, &priv
->rx_cq
[i
].napi
.state
))
732 mlx4_en_deactivate_cq(priv
, &priv
->rx_cq
[i
]);
736 static void mlx4_en_restart(struct work_struct
*work
)
738 struct mlx4_en_priv
*priv
= container_of(work
, struct mlx4_en_priv
,
740 struct mlx4_en_dev
*mdev
= priv
->mdev
;
741 struct net_device
*dev
= priv
->dev
;
743 en_dbg(DRV
, priv
, "Watchdog task called for port %d\n", priv
->port
);
745 mutex_lock(&mdev
->state_lock
);
747 mlx4_en_stop_port(dev
);
748 if (mlx4_en_start_port(dev
))
749 en_err(priv
, "Failed restarting port %d\n", priv
->port
);
751 mutex_unlock(&mdev
->state_lock
);
755 static int mlx4_en_open(struct net_device
*dev
)
757 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
758 struct mlx4_en_dev
*mdev
= priv
->mdev
;
762 mutex_lock(&mdev
->state_lock
);
764 if (!mdev
->device_up
) {
765 en_err(priv
, "Cannot open - device down/disabled\n");
770 /* Reset HW statistics and performance counters */
771 if (mlx4_en_DUMP_ETH_STATS(mdev
, priv
->port
, 1))
772 en_dbg(HW
, priv
, "Failed dumping statistics\n");
774 memset(&priv
->stats
, 0, sizeof(priv
->stats
));
775 memset(&priv
->pstats
, 0, sizeof(priv
->pstats
));
777 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
778 priv
->tx_ring
[i
].bytes
= 0;
779 priv
->tx_ring
[i
].packets
= 0;
781 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
782 priv
->rx_ring
[i
].bytes
= 0;
783 priv
->rx_ring
[i
].packets
= 0;
786 mlx4_en_set_default_moderation(priv
);
787 err
= mlx4_en_start_port(dev
);
789 en_err(priv
, "Failed starting port:%d\n", priv
->port
);
792 mutex_unlock(&mdev
->state_lock
);
797 static int mlx4_en_close(struct net_device
*dev
)
799 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
800 struct mlx4_en_dev
*mdev
= priv
->mdev
;
802 en_dbg(IFDOWN
, priv
, "Close port called\n");
804 mutex_lock(&mdev
->state_lock
);
806 mlx4_en_stop_port(dev
);
807 netif_carrier_off(dev
);
809 mutex_unlock(&mdev
->state_lock
);
813 void mlx4_en_free_resources(struct mlx4_en_priv
*priv
)
817 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
818 if (priv
->tx_ring
[i
].tx_info
)
819 mlx4_en_destroy_tx_ring(priv
, &priv
->tx_ring
[i
]);
820 if (priv
->tx_cq
[i
].buf
)
821 mlx4_en_destroy_cq(priv
, &priv
->tx_cq
[i
]);
824 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
825 if (priv
->rx_ring
[i
].rx_info
)
826 mlx4_en_destroy_rx_ring(priv
, &priv
->rx_ring
[i
]);
827 if (priv
->rx_cq
[i
].buf
)
828 mlx4_en_destroy_cq(priv
, &priv
->rx_cq
[i
]);
832 int mlx4_en_alloc_resources(struct mlx4_en_priv
*priv
)
834 struct mlx4_en_port_profile
*prof
= priv
->prof
;
837 /* Create tx Rings */
838 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
839 if (mlx4_en_create_cq(priv
, &priv
->tx_cq
[i
],
840 prof
->tx_ring_size
, i
, TX
))
843 if (mlx4_en_create_tx_ring(priv
, &priv
->tx_ring
[i
],
844 prof
->tx_ring_size
, TXBB_SIZE
))
848 /* Create rx Rings */
849 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
850 if (mlx4_en_create_cq(priv
, &priv
->rx_cq
[i
],
851 prof
->rx_ring_size
, i
, RX
))
854 if (mlx4_en_create_rx_ring(priv
, &priv
->rx_ring
[i
],
855 prof
->rx_ring_size
, priv
->stride
))
862 en_err(priv
, "Failed to allocate NIC resources\n");
867 void mlx4_en_destroy_netdev(struct net_device
*dev
)
869 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
870 struct mlx4_en_dev
*mdev
= priv
->mdev
;
872 en_dbg(DRV
, priv
, "Destroying netdev on port:%d\n", priv
->port
);
874 /* Unregister device - this will close the port if it was up */
875 if (priv
->registered
)
876 unregister_netdev(dev
);
879 mlx4_free_hwq_res(mdev
->dev
, &priv
->res
, MLX4_EN_PAGE_SIZE
);
881 cancel_delayed_work(&priv
->stats_task
);
882 /* flush any pending task for this netdev */
883 flush_workqueue(mdev
->workqueue
);
885 /* Detach the netdev so tasks would not attempt to access it */
886 mutex_lock(&mdev
->state_lock
);
887 mdev
->pndev
[priv
->port
] = NULL
;
888 mutex_unlock(&mdev
->state_lock
);
890 mlx4_en_free_resources(priv
);
894 static int mlx4_en_change_mtu(struct net_device
*dev
, int new_mtu
)
896 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
897 struct mlx4_en_dev
*mdev
= priv
->mdev
;
900 en_dbg(DRV
, priv
, "Change MTU called - current:%d new:%d\n",
903 if ((new_mtu
< MLX4_EN_MIN_MTU
) || (new_mtu
> priv
->max_mtu
)) {
904 en_err(priv
, "Bad MTU size:%d.\n", new_mtu
);
909 if (netif_running(dev
)) {
910 mutex_lock(&mdev
->state_lock
);
911 if (!mdev
->device_up
) {
912 /* NIC is probably restarting - let watchdog task reset
914 en_dbg(DRV
, priv
, "Change MTU called with card down!?\n");
916 mlx4_en_stop_port(dev
);
917 mlx4_en_set_default_moderation(priv
);
918 err
= mlx4_en_start_port(dev
);
920 en_err(priv
, "Failed restarting port:%d\n",
922 queue_work(mdev
->workqueue
, &priv
->watchdog_task
);
925 mutex_unlock(&mdev
->state_lock
);
930 static const struct net_device_ops mlx4_netdev_ops
= {
931 .ndo_open
= mlx4_en_open
,
932 .ndo_stop
= mlx4_en_close
,
933 .ndo_start_xmit
= mlx4_en_xmit
,
934 .ndo_select_queue
= mlx4_en_select_queue
,
935 .ndo_get_stats
= mlx4_en_get_stats
,
936 .ndo_set_multicast_list
= mlx4_en_set_multicast
,
937 .ndo_set_mac_address
= mlx4_en_set_mac
,
938 .ndo_validate_addr
= eth_validate_addr
,
939 .ndo_change_mtu
= mlx4_en_change_mtu
,
940 .ndo_tx_timeout
= mlx4_en_tx_timeout
,
941 .ndo_vlan_rx_register
= mlx4_en_vlan_rx_register
,
942 .ndo_vlan_rx_add_vid
= mlx4_en_vlan_rx_add_vid
,
943 .ndo_vlan_rx_kill_vid
= mlx4_en_vlan_rx_kill_vid
,
944 #ifdef CONFIG_NET_POLL_CONTROLLER
945 .ndo_poll_controller
= mlx4_en_netpoll
,
949 int mlx4_en_init_netdev(struct mlx4_en_dev
*mdev
, int port
,
950 struct mlx4_en_port_profile
*prof
)
952 struct net_device
*dev
;
953 struct mlx4_en_priv
*priv
;
957 dev
= alloc_etherdev_mq(sizeof(struct mlx4_en_priv
), prof
->tx_ring_num
);
959 mlx4_err(mdev
, "Net device allocation failed\n");
963 SET_NETDEV_DEV(dev
, &mdev
->dev
->pdev
->dev
);
964 dev
->dev_id
= port
- 1;
967 * Initialize driver private data
970 priv
= netdev_priv(dev
);
971 memset(priv
, 0, sizeof(struct mlx4_en_priv
));
976 priv
->port_up
= false;
978 priv
->flags
= prof
->flags
;
979 priv
->tx_ring_num
= prof
->tx_ring_num
;
980 priv
->rx_ring_num
= prof
->rx_ring_num
;
981 priv
->mac_index
= -1;
982 priv
->msg_enable
= MLX4_EN_MSG_LEVEL
;
983 spin_lock_init(&priv
->stats_lock
);
984 INIT_WORK(&priv
->mcast_task
, mlx4_en_do_set_multicast
);
985 INIT_WORK(&priv
->mac_task
, mlx4_en_do_set_mac
);
986 INIT_WORK(&priv
->watchdog_task
, mlx4_en_restart
);
987 INIT_WORK(&priv
->linkstate_task
, mlx4_en_linkstate
);
988 INIT_DELAYED_WORK(&priv
->stats_task
, mlx4_en_do_get_stats
);
990 /* Query for default mac and max mtu */
991 priv
->max_mtu
= mdev
->dev
->caps
.eth_mtu_cap
[priv
->port
];
992 priv
->mac
= mdev
->dev
->caps
.def_mac
[priv
->port
];
993 if (ILLEGAL_MAC(priv
->mac
)) {
994 en_err(priv
, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
995 priv
->port
, priv
->mac
);
1000 priv
->stride
= roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc
) +
1001 DS_SIZE
* MLX4_EN_MAX_RX_FRAGS
);
1002 err
= mlx4_en_alloc_resources(priv
);
1006 /* Allocate page for receive rings */
1007 err
= mlx4_alloc_hwq_res(mdev
->dev
, &priv
->res
,
1008 MLX4_EN_PAGE_SIZE
, MLX4_EN_PAGE_SIZE
);
1010 en_err(priv
, "Failed to allocate page for rx qps\n");
1013 priv
->allocated
= 1;
1016 * Initialize netdev entry points
1018 dev
->netdev_ops
= &mlx4_netdev_ops
;
1019 dev
->watchdog_timeo
= MLX4_EN_WATCHDOG_TIMEOUT
;
1020 dev
->real_num_tx_queues
= MLX4_EN_NUM_TX_RINGS
;
1022 SET_ETHTOOL_OPS(dev
, &mlx4_en_ethtool_ops
);
1024 /* Set defualt MAC */
1025 dev
->addr_len
= ETH_ALEN
;
1026 for (i
= 0; i
< ETH_ALEN
; i
++)
1027 dev
->dev_addr
[ETH_ALEN
- 1 - i
] =
1028 (u8
) (priv
->mac
>> (8 * i
));
1031 * Set driver features
1033 dev
->features
|= NETIF_F_SG
;
1034 dev
->vlan_features
|= NETIF_F_SG
;
1035 dev
->features
|= NETIF_F_IP_CSUM
| NETIF_F_IPV6_CSUM
;
1036 dev
->vlan_features
|= NETIF_F_IP_CSUM
| NETIF_F_IPV6_CSUM
;
1037 dev
->features
|= NETIF_F_HIGHDMA
;
1038 dev
->features
|= NETIF_F_HW_VLAN_TX
|
1039 NETIF_F_HW_VLAN_RX
|
1040 NETIF_F_HW_VLAN_FILTER
;
1041 if (mdev
->profile
.num_lro
)
1042 dev
->features
|= NETIF_F_LRO
;
1043 if (mdev
->LSO_support
) {
1044 dev
->features
|= NETIF_F_TSO
;
1045 dev
->features
|= NETIF_F_TSO6
;
1046 dev
->vlan_features
|= NETIF_F_TSO
;
1047 dev
->vlan_features
|= NETIF_F_TSO6
;
1050 mdev
->pndev
[port
] = dev
;
1052 netif_carrier_off(dev
);
1053 err
= register_netdev(dev
);
1055 en_err(priv
, "Netdev registration failed for port %d\n", port
);
1059 en_warn(priv
, "Using %d TX rings\n", prof
->tx_ring_num
);
1060 en_warn(priv
, "Using %d RX rings\n", prof
->rx_ring_num
);
1062 priv
->registered
= 1;
1063 queue_delayed_work(mdev
->workqueue
, &priv
->stats_task
, STATS_DELAY
);
1067 mlx4_en_destroy_netdev(dev
);