1 /* A network driver using virtio.
3 * Copyright 2007 Rusty Russell <rusty@rustcorp.com.au> IBM Corporation
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 #include <linux/netdevice.h>
21 #include <linux/etherdevice.h>
22 #include <linux/ethtool.h>
23 #include <linux/module.h>
24 #include <linux/virtio.h>
25 #include <linux/virtio_net.h>
26 #include <linux/scatterlist.h>
27 #include <linux/if_vlan.h>
29 static int napi_weight
= 128;
30 module_param(napi_weight
, int, 0444);
32 static int csum
= 1, gso
= 1;
33 module_param(csum
, bool, 0444);
34 module_param(gso
, bool, 0444);
36 /* FIXME: MTU in config. */
37 #define MAX_PACKET_LEN (ETH_HLEN + VLAN_HLEN + ETH_DATA_LEN)
38 #define GOOD_COPY_LEN 128
40 #define VIRTNET_SEND_COMMAND_SG_MAX 2
44 struct virtio_device
*vdev
;
45 struct virtqueue
*rvq
, *svq
, *cvq
;
46 struct net_device
*dev
;
47 struct napi_struct napi
;
50 /* Number of input buffers, and max we've ever had. */
51 unsigned int num
, max
;
53 /* I like... big packets and I cannot lie! */
56 /* Host will merge rx buffers for big packets (shake it! shake it!) */
57 bool mergeable_rx_bufs
;
60 struct sk_buff_head send
;
62 /* Work struct for refilling if we run low on memory. */
63 struct delayed_work refill
;
65 /* Chain pages by the private ptr. */
71 struct virtio_net_hdr hdr
;
72 struct virtio_net_hdr_mrg_rxbuf mhdr
;
77 struct padded_vnet_hdr
{
78 struct virtio_net_hdr hdr
;
80 * virtio_net_hdr should be in a separated sg buffer because of a
81 * QEMU bug, and data sg buffer shares same page with this header sg.
82 * This padding makes next sg 16 byte aligned after virtio_net_hdr.
87 static inline struct skb_vnet_hdr
*skb_vnet_hdr(struct sk_buff
*skb
)
89 return (struct skb_vnet_hdr
*)skb
->cb
;
93 * private is used to chain pages for big packets, put the whole
94 * most recent used list in the beginning for reuse
96 static void give_pages(struct virtnet_info
*vi
, struct page
*page
)
100 /* Find end of list, sew whole thing into vi->pages. */
101 for (end
= page
; end
->private; end
= (struct page
*)end
->private);
102 end
->private = (unsigned long)vi
->pages
;
106 static struct page
*get_a_page(struct virtnet_info
*vi
, gfp_t gfp_mask
)
108 struct page
*p
= vi
->pages
;
111 vi
->pages
= (struct page
*)p
->private;
112 /* clear private here, it is used to chain pages */
115 p
= alloc_page(gfp_mask
);
119 static void skb_xmit_done(struct virtqueue
*svq
)
121 struct virtnet_info
*vi
= svq
->vdev
->priv
;
123 /* Suppress further interrupts. */
124 svq
->vq_ops
->disable_cb(svq
);
126 /* We were probably waiting for more output buffers. */
127 netif_wake_queue(vi
->dev
);
130 static void set_skb_frag(struct sk_buff
*skb
, struct page
*page
,
131 unsigned int offset
, unsigned int *len
)
133 int i
= skb_shinfo(skb
)->nr_frags
;
136 f
= &skb_shinfo(skb
)->frags
[i
];
137 f
->size
= min((unsigned)PAGE_SIZE
- offset
, *len
);
138 f
->page_offset
= offset
;
141 skb
->data_len
+= f
->size
;
143 skb_shinfo(skb
)->nr_frags
++;
147 static struct sk_buff
*page_to_skb(struct virtnet_info
*vi
,
148 struct page
*page
, unsigned int len
)
151 struct skb_vnet_hdr
*hdr
;
152 unsigned int copy
, hdr_len
, offset
;
155 p
= page_address(page
);
157 /* copy small packet so we can reuse these pages for small data */
158 skb
= netdev_alloc_skb_ip_align(vi
->dev
, GOOD_COPY_LEN
);
162 hdr
= skb_vnet_hdr(skb
);
164 if (vi
->mergeable_rx_bufs
) {
165 hdr_len
= sizeof hdr
->mhdr
;
168 hdr_len
= sizeof hdr
->hdr
;
169 offset
= sizeof(struct padded_vnet_hdr
);
172 memcpy(hdr
, p
, hdr_len
);
178 if (copy
> skb_tailroom(skb
))
179 copy
= skb_tailroom(skb
);
180 memcpy(skb_put(skb
, copy
), p
, copy
);
186 set_skb_frag(skb
, page
, offset
, &len
);
187 page
= (struct page
*)page
->private;
192 give_pages(vi
, page
);
197 static int receive_mergeable(struct virtnet_info
*vi
, struct sk_buff
*skb
)
199 struct skb_vnet_hdr
*hdr
= skb_vnet_hdr(skb
);
203 num_buf
= hdr
->mhdr
.num_buffers
;
205 i
= skb_shinfo(skb
)->nr_frags
;
206 if (i
>= MAX_SKB_FRAGS
) {
207 pr_debug("%s: packet too long\n", skb
->dev
->name
);
208 skb
->dev
->stats
.rx_length_errors
++;
212 page
= vi
->rvq
->vq_ops
->get_buf(vi
->rvq
, &len
);
214 pr_debug("%s: rx error: %d buffers missing\n",
215 skb
->dev
->name
, hdr
->mhdr
.num_buffers
);
216 skb
->dev
->stats
.rx_length_errors
++;
222 set_skb_frag(skb
, page
, 0, &len
);
229 static void receive_buf(struct net_device
*dev
, void *buf
, unsigned int len
)
231 struct virtnet_info
*vi
= netdev_priv(dev
);
234 struct skb_vnet_hdr
*hdr
;
236 if (unlikely(len
< sizeof(struct virtio_net_hdr
) + ETH_HLEN
)) {
237 pr_debug("%s: short packet %i\n", dev
->name
, len
);
238 dev
->stats
.rx_length_errors
++;
239 if (vi
->mergeable_rx_bufs
|| vi
->big_packets
)
246 if (!vi
->mergeable_rx_bufs
&& !vi
->big_packets
) {
248 len
-= sizeof(struct virtio_net_hdr
);
252 skb
= page_to_skb(vi
, page
, len
);
253 if (unlikely(!skb
)) {
254 dev
->stats
.rx_dropped
++;
255 give_pages(vi
, page
);
258 if (vi
->mergeable_rx_bufs
)
259 if (receive_mergeable(vi
, skb
)) {
265 hdr
= skb_vnet_hdr(skb
);
266 skb
->truesize
+= skb
->data_len
;
267 dev
->stats
.rx_bytes
+= skb
->len
;
268 dev
->stats
.rx_packets
++;
270 if (hdr
->hdr
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) {
271 pr_debug("Needs csum!\n");
272 if (!skb_partial_csum_set(skb
,
274 hdr
->hdr
.csum_offset
))
278 skb
->protocol
= eth_type_trans(skb
, dev
);
279 pr_debug("Receiving skb proto 0x%04x len %i type %i\n",
280 ntohs(skb
->protocol
), skb
->len
, skb
->pkt_type
);
282 if (hdr
->hdr
.gso_type
!= VIRTIO_NET_HDR_GSO_NONE
) {
284 switch (hdr
->hdr
.gso_type
& ~VIRTIO_NET_HDR_GSO_ECN
) {
285 case VIRTIO_NET_HDR_GSO_TCPV4
:
286 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
288 case VIRTIO_NET_HDR_GSO_UDP
:
289 skb_shinfo(skb
)->gso_type
= SKB_GSO_UDP
;
291 case VIRTIO_NET_HDR_GSO_TCPV6
:
292 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV6
;
296 printk(KERN_WARNING
"%s: bad gso type %u.\n",
297 dev
->name
, hdr
->hdr
.gso_type
);
301 if (hdr
->hdr
.gso_type
& VIRTIO_NET_HDR_GSO_ECN
)
302 skb_shinfo(skb
)->gso_type
|= SKB_GSO_TCP_ECN
;
304 skb_shinfo(skb
)->gso_size
= hdr
->hdr
.gso_size
;
305 if (skb_shinfo(skb
)->gso_size
== 0) {
307 printk(KERN_WARNING
"%s: zero gso size.\n",
312 /* Header must be checked, and gso_segs computed. */
313 skb_shinfo(skb
)->gso_type
|= SKB_GSO_DODGY
;
314 skb_shinfo(skb
)->gso_segs
= 0;
317 netif_receive_skb(skb
);
321 dev
->stats
.rx_frame_errors
++;
325 static int add_recvbuf_small(struct virtnet_info
*vi
, gfp_t gfp
)
328 struct skb_vnet_hdr
*hdr
;
329 struct scatterlist sg
[2];
332 skb
= netdev_alloc_skb_ip_align(vi
->dev
, MAX_PACKET_LEN
);
336 skb_put(skb
, MAX_PACKET_LEN
);
338 hdr
= skb_vnet_hdr(skb
);
339 sg_set_buf(sg
, &hdr
->hdr
, sizeof hdr
->hdr
);
341 skb_to_sgvec(skb
, sg
+ 1, 0, skb
->len
);
343 err
= vi
->rvq
->vq_ops
->add_buf(vi
->rvq
, sg
, 0, 2, skb
);
350 static int add_recvbuf_big(struct virtnet_info
*vi
, gfp_t gfp
)
352 struct scatterlist sg
[MAX_SKB_FRAGS
+ 2];
353 struct page
*first
, *list
= NULL
;
357 /* page in sg[MAX_SKB_FRAGS + 1] is list tail */
358 for (i
= MAX_SKB_FRAGS
+ 1; i
> 1; --i
) {
359 first
= get_a_page(vi
, gfp
);
362 give_pages(vi
, list
);
365 sg_set_buf(&sg
[i
], page_address(first
), PAGE_SIZE
);
367 /* chain new page in list head to match sg */
368 first
->private = (unsigned long)list
;
372 first
= get_a_page(vi
, gfp
);
374 give_pages(vi
, list
);
377 p
= page_address(first
);
379 /* sg[0], sg[1] share the same page */
380 /* a separated sg[0] for virtio_net_hdr only during to QEMU bug*/
381 sg_set_buf(&sg
[0], p
, sizeof(struct virtio_net_hdr
));
383 /* sg[1] for data packet, from offset */
384 offset
= sizeof(struct padded_vnet_hdr
);
385 sg_set_buf(&sg
[1], p
+ offset
, PAGE_SIZE
- offset
);
387 /* chain first in list head */
388 first
->private = (unsigned long)list
;
389 err
= vi
->rvq
->vq_ops
->add_buf(vi
->rvq
, sg
, 0, MAX_SKB_FRAGS
+ 2,
392 give_pages(vi
, first
);
397 static int add_recvbuf_mergeable(struct virtnet_info
*vi
, gfp_t gfp
)
400 struct scatterlist sg
;
403 page
= get_a_page(vi
, gfp
);
407 sg_init_one(&sg
, page_address(page
), PAGE_SIZE
);
409 err
= vi
->rvq
->vq_ops
->add_buf(vi
->rvq
, &sg
, 0, 1, page
);
411 give_pages(vi
, page
);
416 /* Returns false if we couldn't fill entirely (OOM). */
417 static bool try_fill_recv(struct virtnet_info
*vi
, gfp_t gfp
)
423 if (vi
->mergeable_rx_bufs
)
424 err
= add_recvbuf_mergeable(vi
, gfp
);
425 else if (vi
->big_packets
)
426 err
= add_recvbuf_big(vi
, gfp
);
428 err
= add_recvbuf_small(vi
, gfp
);
436 if (unlikely(vi
->num
> vi
->max
))
438 vi
->rvq
->vq_ops
->kick(vi
->rvq
);
442 static void skb_recv_done(struct virtqueue
*rvq
)
444 struct virtnet_info
*vi
= rvq
->vdev
->priv
;
445 /* Schedule NAPI, Suppress further interrupts if successful. */
446 if (napi_schedule_prep(&vi
->napi
)) {
447 rvq
->vq_ops
->disable_cb(rvq
);
448 __napi_schedule(&vi
->napi
);
452 static void refill_work(struct work_struct
*work
)
454 struct virtnet_info
*vi
;
457 vi
= container_of(work
, struct virtnet_info
, refill
.work
);
458 napi_disable(&vi
->napi
);
459 still_empty
= !try_fill_recv(vi
, GFP_KERNEL
);
460 napi_enable(&vi
->napi
);
462 /* In theory, this can happen: if we don't get any buffers in
463 * we will *never* try to fill again. */
465 schedule_delayed_work(&vi
->refill
, HZ
/2);
468 static int virtnet_poll(struct napi_struct
*napi
, int budget
)
470 struct virtnet_info
*vi
= container_of(napi
, struct virtnet_info
, napi
);
472 unsigned int len
, received
= 0;
475 while (received
< budget
&&
476 (buf
= vi
->rvq
->vq_ops
->get_buf(vi
->rvq
, &len
)) != NULL
) {
477 receive_buf(vi
->dev
, buf
, len
);
482 if (vi
->num
< vi
->max
/ 2) {
483 if (!try_fill_recv(vi
, GFP_ATOMIC
))
484 schedule_delayed_work(&vi
->refill
, 0);
487 /* Out of packets? */
488 if (received
< budget
) {
490 if (unlikely(!vi
->rvq
->vq_ops
->enable_cb(vi
->rvq
)) &&
491 napi_schedule_prep(napi
)) {
492 vi
->rvq
->vq_ops
->disable_cb(vi
->rvq
);
493 __napi_schedule(napi
);
501 static unsigned int free_old_xmit_skbs(struct virtnet_info
*vi
)
504 unsigned int len
, tot_sgs
= 0;
506 while ((skb
= vi
->svq
->vq_ops
->get_buf(vi
->svq
, &len
)) != NULL
) {
507 pr_debug("Sent skb %p\n", skb
);
508 __skb_unlink(skb
, &vi
->send
);
509 vi
->dev
->stats
.tx_bytes
+= skb
->len
;
510 vi
->dev
->stats
.tx_packets
++;
511 tot_sgs
+= skb_vnet_hdr(skb
)->num_sg
;
512 dev_kfree_skb_any(skb
);
517 static int xmit_skb(struct virtnet_info
*vi
, struct sk_buff
*skb
)
519 struct scatterlist sg
[2+MAX_SKB_FRAGS
];
520 struct skb_vnet_hdr
*hdr
= skb_vnet_hdr(skb
);
521 const unsigned char *dest
= ((struct ethhdr
*)skb
->data
)->h_dest
;
523 sg_init_table(sg
, 2+MAX_SKB_FRAGS
);
525 pr_debug("%s: xmit %p %pM\n", vi
->dev
->name
, skb
, dest
);
527 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
528 hdr
->hdr
.flags
= VIRTIO_NET_HDR_F_NEEDS_CSUM
;
529 hdr
->hdr
.csum_start
= skb
->csum_start
- skb_headroom(skb
);
530 hdr
->hdr
.csum_offset
= skb
->csum_offset
;
533 hdr
->hdr
.csum_offset
= hdr
->hdr
.csum_start
= 0;
536 if (skb_is_gso(skb
)) {
537 hdr
->hdr
.hdr_len
= skb_headlen(skb
);
538 hdr
->hdr
.gso_size
= skb_shinfo(skb
)->gso_size
;
539 if (skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV4
)
540 hdr
->hdr
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV4
;
541 else if (skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
)
542 hdr
->hdr
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV6
;
543 else if (skb_shinfo(skb
)->gso_type
& SKB_GSO_UDP
)
544 hdr
->hdr
.gso_type
= VIRTIO_NET_HDR_GSO_UDP
;
547 if (skb_shinfo(skb
)->gso_type
& SKB_GSO_TCP_ECN
)
548 hdr
->hdr
.gso_type
|= VIRTIO_NET_HDR_GSO_ECN
;
550 hdr
->hdr
.gso_type
= VIRTIO_NET_HDR_GSO_NONE
;
551 hdr
->hdr
.gso_size
= hdr
->hdr
.hdr_len
= 0;
554 hdr
->mhdr
.num_buffers
= 0;
556 /* Encode metadata header at front. */
557 if (vi
->mergeable_rx_bufs
)
558 sg_set_buf(sg
, &hdr
->mhdr
, sizeof hdr
->mhdr
);
560 sg_set_buf(sg
, &hdr
->hdr
, sizeof hdr
->hdr
);
562 hdr
->num_sg
= skb_to_sgvec(skb
, sg
+1, 0, skb
->len
) + 1;
563 return vi
->svq
->vq_ops
->add_buf(vi
->svq
, sg
, hdr
->num_sg
, 0, skb
);
566 static netdev_tx_t
start_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
568 struct virtnet_info
*vi
= netdev_priv(dev
);
572 /* Free up any pending old buffers before queueing new ones. */
573 free_old_xmit_skbs(vi
);
575 /* Try to transmit */
576 capacity
= xmit_skb(vi
, skb
);
578 /* This can happen with OOM and indirect buffers. */
579 if (unlikely(capacity
< 0)) {
580 netif_stop_queue(dev
);
581 dev_warn(&dev
->dev
, "Unexpected full queue\n");
582 if (unlikely(!vi
->svq
->vq_ops
->enable_cb(vi
->svq
))) {
583 vi
->svq
->vq_ops
->disable_cb(vi
->svq
);
584 netif_start_queue(dev
);
587 return NETDEV_TX_BUSY
;
589 vi
->svq
->vq_ops
->kick(vi
->svq
);
592 * Put new one in send queue. You'd expect we'd need this before
593 * xmit_skb calls add_buf(), since the callback can be triggered
594 * immediately after that. But since the callback just triggers
595 * another call back here, normal network xmit locking prevents the
598 __skb_queue_head(&vi
->send
, skb
);
600 /* Don't wait up for transmitted skbs to be freed. */
604 /* Apparently nice girls don't return TX_BUSY; stop the queue
605 * before it gets out of hand. Naturally, this wastes entries. */
606 if (capacity
< 2+MAX_SKB_FRAGS
) {
607 netif_stop_queue(dev
);
608 if (unlikely(!vi
->svq
->vq_ops
->enable_cb(vi
->svq
))) {
609 /* More just got used, free them then recheck. */
610 capacity
+= free_old_xmit_skbs(vi
);
611 if (capacity
>= 2+MAX_SKB_FRAGS
) {
612 netif_start_queue(dev
);
613 vi
->svq
->vq_ops
->disable_cb(vi
->svq
);
621 static int virtnet_set_mac_address(struct net_device
*dev
, void *p
)
623 struct virtnet_info
*vi
= netdev_priv(dev
);
624 struct virtio_device
*vdev
= vi
->vdev
;
627 ret
= eth_mac_addr(dev
, p
);
631 if (virtio_has_feature(vdev
, VIRTIO_NET_F_MAC
))
632 vdev
->config
->set(vdev
, offsetof(struct virtio_net_config
, mac
),
633 dev
->dev_addr
, dev
->addr_len
);
638 #ifdef CONFIG_NET_POLL_CONTROLLER
639 static void virtnet_netpoll(struct net_device
*dev
)
641 struct virtnet_info
*vi
= netdev_priv(dev
);
643 napi_schedule(&vi
->napi
);
647 static int virtnet_open(struct net_device
*dev
)
649 struct virtnet_info
*vi
= netdev_priv(dev
);
651 napi_enable(&vi
->napi
);
653 /* If all buffers were filled by other side before we napi_enabled, we
654 * won't get another interrupt, so process any outstanding packets
655 * now. virtnet_poll wants re-enable the queue, so we disable here.
656 * We synchronize against interrupts via NAPI_STATE_SCHED */
657 if (napi_schedule_prep(&vi
->napi
)) {
658 vi
->rvq
->vq_ops
->disable_cb(vi
->rvq
);
659 __napi_schedule(&vi
->napi
);
665 * Send command via the control virtqueue and check status. Commands
666 * supported by the hypervisor, as indicated by feature bits, should
667 * never fail unless improperly formated.
669 static bool virtnet_send_command(struct virtnet_info
*vi
, u8
class, u8 cmd
,
670 struct scatterlist
*data
, int out
, int in
)
672 struct scatterlist
*s
, sg
[VIRTNET_SEND_COMMAND_SG_MAX
+ 2];
673 struct virtio_net_ctrl_hdr ctrl
;
674 virtio_net_ctrl_ack status
= ~0;
678 /* Caller should know better */
679 BUG_ON(!virtio_has_feature(vi
->vdev
, VIRTIO_NET_F_CTRL_VQ
) ||
680 (out
+ in
> VIRTNET_SEND_COMMAND_SG_MAX
));
682 out
++; /* Add header */
683 in
++; /* Add return status */
688 sg_init_table(sg
, out
+ in
);
690 sg_set_buf(&sg
[0], &ctrl
, sizeof(ctrl
));
691 for_each_sg(data
, s
, out
+ in
- 2, i
)
692 sg_set_buf(&sg
[i
+ 1], sg_virt(s
), s
->length
);
693 sg_set_buf(&sg
[out
+ in
- 1], &status
, sizeof(status
));
695 BUG_ON(vi
->cvq
->vq_ops
->add_buf(vi
->cvq
, sg
, out
, in
, vi
) < 0);
697 vi
->cvq
->vq_ops
->kick(vi
->cvq
);
700 * Spin for a response, the kick causes an ioport write, trapping
701 * into the hypervisor, so the request should be handled immediately.
703 while (!vi
->cvq
->vq_ops
->get_buf(vi
->cvq
, &tmp
))
706 return status
== VIRTIO_NET_OK
;
709 static int virtnet_close(struct net_device
*dev
)
711 struct virtnet_info
*vi
= netdev_priv(dev
);
713 napi_disable(&vi
->napi
);
718 static int virtnet_set_tx_csum(struct net_device
*dev
, u32 data
)
720 struct virtnet_info
*vi
= netdev_priv(dev
);
721 struct virtio_device
*vdev
= vi
->vdev
;
723 if (data
&& !virtio_has_feature(vdev
, VIRTIO_NET_F_CSUM
))
726 return ethtool_op_set_tx_hw_csum(dev
, data
);
729 static void virtnet_set_rx_mode(struct net_device
*dev
)
731 struct virtnet_info
*vi
= netdev_priv(dev
);
732 struct scatterlist sg
[2];
733 u8 promisc
, allmulti
;
734 struct virtio_net_ctrl_mac
*mac_data
;
735 struct dev_addr_list
*addr
;
736 struct netdev_hw_addr
*ha
;
741 /* We can't dynamicaly set ndo_set_rx_mode, so return gracefully */
742 if (!virtio_has_feature(vi
->vdev
, VIRTIO_NET_F_CTRL_RX
))
745 promisc
= ((dev
->flags
& IFF_PROMISC
) != 0);
746 allmulti
= ((dev
->flags
& IFF_ALLMULTI
) != 0);
748 sg_init_one(sg
, &promisc
, sizeof(promisc
));
750 if (!virtnet_send_command(vi
, VIRTIO_NET_CTRL_RX
,
751 VIRTIO_NET_CTRL_RX_PROMISC
,
753 dev_warn(&dev
->dev
, "Failed to %sable promisc mode.\n",
754 promisc
? "en" : "dis");
756 sg_init_one(sg
, &allmulti
, sizeof(allmulti
));
758 if (!virtnet_send_command(vi
, VIRTIO_NET_CTRL_RX
,
759 VIRTIO_NET_CTRL_RX_ALLMULTI
,
761 dev_warn(&dev
->dev
, "Failed to %sable allmulti mode.\n",
762 allmulti
? "en" : "dis");
764 uc_count
= netdev_uc_count(dev
);
765 /* MAC filter - use one buffer for both lists */
766 mac_data
= buf
= kzalloc(((uc_count
+ dev
->mc_count
) * ETH_ALEN
) +
767 (2 * sizeof(mac_data
->entries
)), GFP_ATOMIC
);
769 dev_warn(&dev
->dev
, "No memory for MAC address buffer\n");
773 sg_init_table(sg
, 2);
775 /* Store the unicast list and count in the front of the buffer */
776 mac_data
->entries
= uc_count
;
778 netdev_for_each_uc_addr(ha
, dev
)
779 memcpy(&mac_data
->macs
[i
++][0], ha
->addr
, ETH_ALEN
);
781 sg_set_buf(&sg
[0], mac_data
,
782 sizeof(mac_data
->entries
) + (uc_count
* ETH_ALEN
));
784 /* multicast list and count fill the end */
785 mac_data
= (void *)&mac_data
->macs
[uc_count
][0];
787 mac_data
->entries
= dev
->mc_count
;
789 for (i
= 0; i
< dev
->mc_count
; i
++, addr
= addr
->next
)
790 memcpy(&mac_data
->macs
[i
][0], addr
->da_addr
, ETH_ALEN
);
792 sg_set_buf(&sg
[1], mac_data
,
793 sizeof(mac_data
->entries
) + (dev
->mc_count
* ETH_ALEN
));
795 if (!virtnet_send_command(vi
, VIRTIO_NET_CTRL_MAC
,
796 VIRTIO_NET_CTRL_MAC_TABLE_SET
,
798 dev_warn(&dev
->dev
, "Failed to set MAC fitler table.\n");
803 static void virtnet_vlan_rx_add_vid(struct net_device
*dev
, u16 vid
)
805 struct virtnet_info
*vi
= netdev_priv(dev
);
806 struct scatterlist sg
;
808 sg_init_one(&sg
, &vid
, sizeof(vid
));
810 if (!virtnet_send_command(vi
, VIRTIO_NET_CTRL_VLAN
,
811 VIRTIO_NET_CTRL_VLAN_ADD
, &sg
, 1, 0))
812 dev_warn(&dev
->dev
, "Failed to add VLAN ID %d.\n", vid
);
815 static void virtnet_vlan_rx_kill_vid(struct net_device
*dev
, u16 vid
)
817 struct virtnet_info
*vi
= netdev_priv(dev
);
818 struct scatterlist sg
;
820 sg_init_one(&sg
, &vid
, sizeof(vid
));
822 if (!virtnet_send_command(vi
, VIRTIO_NET_CTRL_VLAN
,
823 VIRTIO_NET_CTRL_VLAN_DEL
, &sg
, 1, 0))
824 dev_warn(&dev
->dev
, "Failed to kill VLAN ID %d.\n", vid
);
827 static const struct ethtool_ops virtnet_ethtool_ops
= {
828 .set_tx_csum
= virtnet_set_tx_csum
,
829 .set_sg
= ethtool_op_set_sg
,
830 .set_tso
= ethtool_op_set_tso
,
831 .set_ufo
= ethtool_op_set_ufo
,
832 .get_link
= ethtool_op_get_link
,
836 #define MAX_MTU 65535
838 static int virtnet_change_mtu(struct net_device
*dev
, int new_mtu
)
840 if (new_mtu
< MIN_MTU
|| new_mtu
> MAX_MTU
)
846 static const struct net_device_ops virtnet_netdev
= {
847 .ndo_open
= virtnet_open
,
848 .ndo_stop
= virtnet_close
,
849 .ndo_start_xmit
= start_xmit
,
850 .ndo_validate_addr
= eth_validate_addr
,
851 .ndo_set_mac_address
= virtnet_set_mac_address
,
852 .ndo_set_rx_mode
= virtnet_set_rx_mode
,
853 .ndo_change_mtu
= virtnet_change_mtu
,
854 .ndo_vlan_rx_add_vid
= virtnet_vlan_rx_add_vid
,
855 .ndo_vlan_rx_kill_vid
= virtnet_vlan_rx_kill_vid
,
856 #ifdef CONFIG_NET_POLL_CONTROLLER
857 .ndo_poll_controller
= virtnet_netpoll
,
861 static void virtnet_update_status(struct virtnet_info
*vi
)
865 if (!virtio_has_feature(vi
->vdev
, VIRTIO_NET_F_STATUS
))
868 vi
->vdev
->config
->get(vi
->vdev
,
869 offsetof(struct virtio_net_config
, status
),
872 /* Ignore unknown (future) status bits */
873 v
&= VIRTIO_NET_S_LINK_UP
;
880 if (vi
->status
& VIRTIO_NET_S_LINK_UP
) {
881 netif_carrier_on(vi
->dev
);
882 netif_wake_queue(vi
->dev
);
884 netif_carrier_off(vi
->dev
);
885 netif_stop_queue(vi
->dev
);
889 static void virtnet_config_changed(struct virtio_device
*vdev
)
891 struct virtnet_info
*vi
= vdev
->priv
;
893 virtnet_update_status(vi
);
896 static int virtnet_probe(struct virtio_device
*vdev
)
899 struct net_device
*dev
;
900 struct virtnet_info
*vi
;
901 struct virtqueue
*vqs
[3];
902 vq_callback_t
*callbacks
[] = { skb_recv_done
, skb_xmit_done
, NULL
};
903 const char *names
[] = { "input", "output", "control" };
906 /* Allocate ourselves a network device with room for our info */
907 dev
= alloc_etherdev(sizeof(struct virtnet_info
));
911 /* Set up network device as normal. */
912 dev
->netdev_ops
= &virtnet_netdev
;
913 dev
->features
= NETIF_F_HIGHDMA
;
914 SET_ETHTOOL_OPS(dev
, &virtnet_ethtool_ops
);
915 SET_NETDEV_DEV(dev
, &vdev
->dev
);
917 /* Do we support "hardware" checksums? */
918 if (csum
&& virtio_has_feature(vdev
, VIRTIO_NET_F_CSUM
)) {
919 /* This opens up the world of extra features. */
920 dev
->features
|= NETIF_F_HW_CSUM
|NETIF_F_SG
|NETIF_F_FRAGLIST
;
921 if (gso
&& virtio_has_feature(vdev
, VIRTIO_NET_F_GSO
)) {
922 dev
->features
|= NETIF_F_TSO
| NETIF_F_UFO
923 | NETIF_F_TSO_ECN
| NETIF_F_TSO6
;
925 /* Individual feature bits: what can host handle? */
926 if (gso
&& virtio_has_feature(vdev
, VIRTIO_NET_F_HOST_TSO4
))
927 dev
->features
|= NETIF_F_TSO
;
928 if (gso
&& virtio_has_feature(vdev
, VIRTIO_NET_F_HOST_TSO6
))
929 dev
->features
|= NETIF_F_TSO6
;
930 if (gso
&& virtio_has_feature(vdev
, VIRTIO_NET_F_HOST_ECN
))
931 dev
->features
|= NETIF_F_TSO_ECN
;
932 if (gso
&& virtio_has_feature(vdev
, VIRTIO_NET_F_HOST_UFO
))
933 dev
->features
|= NETIF_F_UFO
;
936 /* Configuration may specify what MAC to use. Otherwise random. */
937 if (virtio_has_feature(vdev
, VIRTIO_NET_F_MAC
)) {
938 vdev
->config
->get(vdev
,
939 offsetof(struct virtio_net_config
, mac
),
940 dev
->dev_addr
, dev
->addr_len
);
942 random_ether_addr(dev
->dev_addr
);
944 /* Set up our device-specific information */
945 vi
= netdev_priv(dev
);
946 netif_napi_add(dev
, &vi
->napi
, virtnet_poll
, napi_weight
);
951 INIT_DELAYED_WORK(&vi
->refill
, refill_work
);
953 /* If we can receive ANY GSO packets, we must allocate large ones. */
954 if (virtio_has_feature(vdev
, VIRTIO_NET_F_GUEST_TSO4
) ||
955 virtio_has_feature(vdev
, VIRTIO_NET_F_GUEST_TSO6
) ||
956 virtio_has_feature(vdev
, VIRTIO_NET_F_GUEST_ECN
))
957 vi
->big_packets
= true;
959 if (virtio_has_feature(vdev
, VIRTIO_NET_F_MRG_RXBUF
))
960 vi
->mergeable_rx_bufs
= true;
962 /* We expect two virtqueues, receive then send,
963 * and optionally control. */
964 nvqs
= virtio_has_feature(vi
->vdev
, VIRTIO_NET_F_CTRL_VQ
) ? 3 : 2;
966 err
= vdev
->config
->find_vqs(vdev
, nvqs
, vqs
, callbacks
, names
);
973 if (virtio_has_feature(vi
->vdev
, VIRTIO_NET_F_CTRL_VQ
)) {
976 if (virtio_has_feature(vi
->vdev
, VIRTIO_NET_F_CTRL_VLAN
))
977 dev
->features
|= NETIF_F_HW_VLAN_FILTER
;
980 /* Initialize our empty send queue. */
981 skb_queue_head_init(&vi
->send
);
983 err
= register_netdev(dev
);
985 pr_debug("virtio_net: registering device failed\n");
989 /* Last of all, set up some receive buffers. */
990 try_fill_recv(vi
, GFP_KERNEL
);
992 /* If we didn't even get one input buffer, we're useless. */
998 vi
->status
= VIRTIO_NET_S_LINK_UP
;
999 virtnet_update_status(vi
);
1000 netif_carrier_on(dev
);
1002 pr_debug("virtnet: registered device %s\n", dev
->name
);
1006 unregister_netdev(dev
);
1007 cancel_delayed_work_sync(&vi
->refill
);
1009 vdev
->config
->del_vqs(vdev
);
1015 static void free_unused_bufs(struct virtnet_info
*vi
)
1019 buf
= vi
->rvq
->vq_ops
->detach_unused_buf(vi
->rvq
);
1022 if (vi
->mergeable_rx_bufs
|| vi
->big_packets
)
1023 give_pages(vi
, buf
);
1028 BUG_ON(vi
->num
!= 0);
1031 static void __devexit
virtnet_remove(struct virtio_device
*vdev
)
1033 struct virtnet_info
*vi
= vdev
->priv
;
1035 /* Stop all the virtqueues. */
1036 vdev
->config
->reset(vdev
);
1038 /* Free our skbs in send queue, if any. */
1039 __skb_queue_purge(&vi
->send
);
1041 unregister_netdev(vi
->dev
);
1042 cancel_delayed_work_sync(&vi
->refill
);
1043 free_unused_bufs(vi
);
1045 vdev
->config
->del_vqs(vi
->vdev
);
1048 __free_pages(get_a_page(vi
, GFP_KERNEL
), 0);
1050 free_netdev(vi
->dev
);
1053 static struct virtio_device_id id_table
[] = {
1054 { VIRTIO_ID_NET
, VIRTIO_DEV_ANY_ID
},
1058 static unsigned int features
[] = {
1059 VIRTIO_NET_F_CSUM
, VIRTIO_NET_F_GUEST_CSUM
,
1060 VIRTIO_NET_F_GSO
, VIRTIO_NET_F_MAC
,
1061 VIRTIO_NET_F_HOST_TSO4
, VIRTIO_NET_F_HOST_UFO
, VIRTIO_NET_F_HOST_TSO6
,
1062 VIRTIO_NET_F_HOST_ECN
, VIRTIO_NET_F_GUEST_TSO4
, VIRTIO_NET_F_GUEST_TSO6
,
1063 VIRTIO_NET_F_GUEST_ECN
, VIRTIO_NET_F_GUEST_UFO
,
1064 VIRTIO_NET_F_MRG_RXBUF
, VIRTIO_NET_F_STATUS
, VIRTIO_NET_F_CTRL_VQ
,
1065 VIRTIO_NET_F_CTRL_RX
, VIRTIO_NET_F_CTRL_VLAN
,
1068 static struct virtio_driver virtio_net_driver
= {
1069 .feature_table
= features
,
1070 .feature_table_size
= ARRAY_SIZE(features
),
1071 .driver
.name
= KBUILD_MODNAME
,
1072 .driver
.owner
= THIS_MODULE
,
1073 .id_table
= id_table
,
1074 .probe
= virtnet_probe
,
1075 .remove
= __devexit_p(virtnet_remove
),
1076 .config_changed
= virtnet_config_changed
,
1079 static int __init
init(void)
1081 return register_virtio_driver(&virtio_net_driver
);
1084 static void __exit
fini(void)
1086 unregister_virtio_driver(&virtio_net_driver
);
1091 MODULE_DEVICE_TABLE(virtio
, id_table
);
1092 MODULE_DESCRIPTION("Virtio network driver");
1093 MODULE_LICENSE("GPL");