blob: fb3302fff62795b1c3723ed0b7982772d7f38d55 [file] [log] [blame]
Thomas Gleixner7a338472019-06-04 10:11:15 +02001// SPDX-License-Identifier: GPL-2.0-only
Asias He0ea9e1d2016-07-28 15:36:33 +01002/*
3 * virtio transport for vsock
4 *
5 * Copyright (C) 2013-2015 Red Hat, Inc.
6 * Author: Asias He <asias@redhat.com>
7 * Stefan Hajnoczi <stefanha@redhat.com>
8 *
9 * Some of the code is take from Gerd Hoffmann <kraxel@redhat.com>'s
10 * early virtio-vsock proof-of-concept bits.
Asias He0ea9e1d2016-07-28 15:36:33 +010011 */
12#include <linux/spinlock.h>
13#include <linux/module.h>
14#include <linux/list.h>
15#include <linux/atomic.h>
16#include <linux/virtio.h>
17#include <linux/virtio_ids.h>
18#include <linux/virtio_config.h>
19#include <linux/virtio_vsock.h>
20#include <net/sock.h>
21#include <linux/mutex.h>
22#include <net/af_vsock.h>
23
24static struct workqueue_struct *virtio_vsock_workqueue;
Stefano Garzarellaf9611342020-07-10 14:12:43 +020025static struct virtio_vsock __rcu *the_virtio_vsock;
Asias He0ea9e1d2016-07-28 15:36:33 +010026static DEFINE_MUTEX(the_virtio_vsock_mutex); /* protects the_virtio_vsock */
27
28struct virtio_vsock {
29 struct virtio_device *vdev;
30 struct virtqueue *vqs[VSOCK_VQ_MAX];
31
32 /* Virtqueue processing is deferred to a workqueue */
33 struct work_struct tx_work;
34 struct work_struct rx_work;
35 struct work_struct event_work;
36
37 /* The following fields are protected by tx_lock. vqs[VSOCK_VQ_TX]
38 * must be accessed with tx_lock held.
39 */
40 struct mutex tx_lock;
Stefano Garzarellab9175072019-07-05 13:04:53 +020041 bool tx_run;
Asias He0ea9e1d2016-07-28 15:36:33 +010042
43 struct work_struct send_pkt_work;
44 spinlock_t send_pkt_list_lock;
45 struct list_head send_pkt_list;
46
47 atomic_t queued_replies;
48
49 /* The following fields are protected by rx_lock. vqs[VSOCK_VQ_RX]
50 * must be accessed with rx_lock held.
51 */
52 struct mutex rx_lock;
Stefano Garzarellab9175072019-07-05 13:04:53 +020053 bool rx_run;
Asias He0ea9e1d2016-07-28 15:36:33 +010054 int rx_buf_nr;
55 int rx_buf_max_nr;
56
57 /* The following fields are protected by event_lock.
58 * vqs[VSOCK_VQ_EVENT] must be accessed with event_lock held.
59 */
60 struct mutex event_lock;
Stefano Garzarellab9175072019-07-05 13:04:53 +020061 bool event_run;
Asias He0ea9e1d2016-07-28 15:36:33 +010062 struct virtio_vsock_event event_list[8];
63
64 u32 guest_cid;
Arseny Krasnov53efbba2021-06-11 14:13:22 +030065 bool seqpacket_allow;
Asias He0ea9e1d2016-07-28 15:36:33 +010066};
67
Asias He0ea9e1d2016-07-28 15:36:33 +010068static u32 virtio_transport_get_local_cid(void)
69{
Stefano Garzarella0deab082019-07-05 13:04:52 +020070 struct virtio_vsock *vsock;
71 u32 ret;
Asias He0ea9e1d2016-07-28 15:36:33 +010072
Stefano Garzarella0deab082019-07-05 13:04:52 +020073 rcu_read_lock();
74 vsock = rcu_dereference(the_virtio_vsock);
75 if (!vsock) {
76 ret = VMADDR_CID_ANY;
77 goto out_rcu;
78 }
Stefano Garzarella22b5c0b2019-02-01 12:42:06 +010079
Stefano Garzarella0deab082019-07-05 13:04:52 +020080 ret = vsock->guest_cid;
81out_rcu:
82 rcu_read_unlock();
83 return ret;
Asias He0ea9e1d2016-07-28 15:36:33 +010084}
85
86static void
87virtio_transport_send_pkt_work(struct work_struct *work)
88{
89 struct virtio_vsock *vsock =
90 container_of(work, struct virtio_vsock, send_pkt_work);
91 struct virtqueue *vq;
92 bool added = false;
93 bool restart_rx = false;
94
95 mutex_lock(&vsock->tx_lock);
96
Stefano Garzarellab9175072019-07-05 13:04:53 +020097 if (!vsock->tx_run)
98 goto out;
99
Asias He0ea9e1d2016-07-28 15:36:33 +0100100 vq = vsock->vqs[VSOCK_VQ_TX];
101
Asias He0ea9e1d2016-07-28 15:36:33 +0100102 for (;;) {
103 struct virtio_vsock_pkt *pkt;
104 struct scatterlist hdr, buf, *sgs[2];
105 int ret, in_sg = 0, out_sg = 0;
106 bool reply;
107
108 spin_lock_bh(&vsock->send_pkt_list_lock);
109 if (list_empty(&vsock->send_pkt_list)) {
110 spin_unlock_bh(&vsock->send_pkt_list_lock);
Asias He0ea9e1d2016-07-28 15:36:33 +0100111 break;
112 }
113
114 pkt = list_first_entry(&vsock->send_pkt_list,
115 struct virtio_vsock_pkt, list);
116 list_del_init(&pkt->list);
117 spin_unlock_bh(&vsock->send_pkt_list_lock);
118
Gerard Garcia82dfb5402017-04-21 10:10:46 +0100119 virtio_transport_deliver_tap_pkt(pkt);
120
Asias He0ea9e1d2016-07-28 15:36:33 +0100121 reply = pkt->reply;
122
123 sg_init_one(&hdr, &pkt->hdr, sizeof(pkt->hdr));
124 sgs[out_sg++] = &hdr;
125 if (pkt->buf) {
126 sg_init_one(&buf, pkt->buf, pkt->len);
127 sgs[out_sg++] = &buf;
128 }
129
130 ret = virtqueue_add_sgs(vq, sgs, out_sg, in_sg, pkt, GFP_KERNEL);
Gerard Garcia21bc54f2016-08-10 17:24:34 +0200131 /* Usually this means that there is no more space available in
132 * the vq
133 */
Asias He0ea9e1d2016-07-28 15:36:33 +0100134 if (ret < 0) {
135 spin_lock_bh(&vsock->send_pkt_list_lock);
136 list_add(&pkt->list, &vsock->send_pkt_list);
137 spin_unlock_bh(&vsock->send_pkt_list_lock);
Asias He0ea9e1d2016-07-28 15:36:33 +0100138 break;
139 }
140
141 if (reply) {
142 struct virtqueue *rx_vq = vsock->vqs[VSOCK_VQ_RX];
143 int val;
144
145 val = atomic_dec_return(&vsock->queued_replies);
146
147 /* Do we now have resources to resume rx processing? */
148 if (val + 1 == virtqueue_get_vring_size(rx_vq))
149 restart_rx = true;
150 }
151
152 added = true;
153 }
154
155 if (added)
156 virtqueue_kick(vq);
157
Stefano Garzarellab9175072019-07-05 13:04:53 +0200158out:
Asias He0ea9e1d2016-07-28 15:36:33 +0100159 mutex_unlock(&vsock->tx_lock);
160
161 if (restart_rx)
162 queue_work(virtio_vsock_workqueue, &vsock->rx_work);
163}
164
165static int
166virtio_transport_send_pkt(struct virtio_vsock_pkt *pkt)
167{
168 struct virtio_vsock *vsock;
169 int len = pkt->len;
170
Stefano Garzarella0deab082019-07-05 13:04:52 +0200171 rcu_read_lock();
172 vsock = rcu_dereference(the_virtio_vsock);
Asias He0ea9e1d2016-07-28 15:36:33 +0100173 if (!vsock) {
174 virtio_transport_free_pkt(pkt);
Stefano Garzarella0deab082019-07-05 13:04:52 +0200175 len = -ENODEV;
176 goto out_rcu;
Asias He0ea9e1d2016-07-28 15:36:33 +0100177 }
178
Stefano Garzarella0deab082019-07-05 13:04:52 +0200179 if (le64_to_cpu(pkt->hdr.dst_cid) == vsock->guest_cid) {
Stefano Garzarellabf5432b2019-12-10 11:43:07 +0100180 virtio_transport_free_pkt(pkt);
181 len = -ENODEV;
Stefano Garzarella0deab082019-07-05 13:04:52 +0200182 goto out_rcu;
183 }
Stefan Hajnoczib9116822016-11-21 13:56:31 +0000184
Asias He0ea9e1d2016-07-28 15:36:33 +0100185 if (pkt->reply)
186 atomic_inc(&vsock->queued_replies);
187
188 spin_lock_bh(&vsock->send_pkt_list_lock);
189 list_add_tail(&pkt->list, &vsock->send_pkt_list);
190 spin_unlock_bh(&vsock->send_pkt_list_lock);
191
192 queue_work(virtio_vsock_workqueue, &vsock->send_pkt_work);
Stefano Garzarella0deab082019-07-05 13:04:52 +0200193
194out_rcu:
195 rcu_read_unlock();
Asias He0ea9e1d2016-07-28 15:36:33 +0100196 return len;
197}
198
Peng Tao073b4f22017-03-15 09:32:16 +0800199static int
200virtio_transport_cancel_pkt(struct vsock_sock *vsk)
201{
202 struct virtio_vsock *vsock;
203 struct virtio_vsock_pkt *pkt, *n;
Stefano Garzarella0deab082019-07-05 13:04:52 +0200204 int cnt = 0, ret;
Peng Tao073b4f22017-03-15 09:32:16 +0800205 LIST_HEAD(freeme);
206
Stefano Garzarella0deab082019-07-05 13:04:52 +0200207 rcu_read_lock();
208 vsock = rcu_dereference(the_virtio_vsock);
Peng Tao073b4f22017-03-15 09:32:16 +0800209 if (!vsock) {
Stefano Garzarella0deab082019-07-05 13:04:52 +0200210 ret = -ENODEV;
211 goto out_rcu;
Peng Tao073b4f22017-03-15 09:32:16 +0800212 }
213
214 spin_lock_bh(&vsock->send_pkt_list_lock);
215 list_for_each_entry_safe(pkt, n, &vsock->send_pkt_list, list) {
216 if (pkt->vsk != vsk)
217 continue;
218 list_move(&pkt->list, &freeme);
219 }
220 spin_unlock_bh(&vsock->send_pkt_list_lock);
221
222 list_for_each_entry_safe(pkt, n, &freeme, list) {
223 if (pkt->reply)
224 cnt++;
225 list_del(&pkt->list);
226 virtio_transport_free_pkt(pkt);
227 }
228
229 if (cnt) {
230 struct virtqueue *rx_vq = vsock->vqs[VSOCK_VQ_RX];
231 int new_cnt;
232
233 new_cnt = atomic_sub_return(cnt, &vsock->queued_replies);
234 if (new_cnt + cnt >= virtqueue_get_vring_size(rx_vq) &&
235 new_cnt < virtqueue_get_vring_size(rx_vq))
236 queue_work(virtio_vsock_workqueue, &vsock->rx_work);
237 }
238
Stefano Garzarella0deab082019-07-05 13:04:52 +0200239 ret = 0;
240
241out_rcu:
242 rcu_read_unlock();
243 return ret;
Peng Tao073b4f22017-03-15 09:32:16 +0800244}
245
Asias He0ea9e1d2016-07-28 15:36:33 +0100246static void virtio_vsock_rx_fill(struct virtio_vsock *vsock)
247{
248 int buf_len = VIRTIO_VSOCK_DEFAULT_RX_BUF_SIZE;
249 struct virtio_vsock_pkt *pkt;
250 struct scatterlist hdr, buf, *sgs[2];
251 struct virtqueue *vq;
252 int ret;
253
254 vq = vsock->vqs[VSOCK_VQ_RX];
255
256 do {
257 pkt = kzalloc(sizeof(*pkt), GFP_KERNEL);
258 if (!pkt)
259 break;
260
261 pkt->buf = kmalloc(buf_len, GFP_KERNEL);
262 if (!pkt->buf) {
263 virtio_transport_free_pkt(pkt);
264 break;
265 }
266
Stefano Garzarella473c7392019-07-30 17:43:30 +0200267 pkt->buf_len = buf_len;
Asias He0ea9e1d2016-07-28 15:36:33 +0100268 pkt->len = buf_len;
269
270 sg_init_one(&hdr, &pkt->hdr, sizeof(pkt->hdr));
271 sgs[0] = &hdr;
272
273 sg_init_one(&buf, pkt->buf, buf_len);
274 sgs[1] = &buf;
275 ret = virtqueue_add_sgs(vq, sgs, 0, 2, pkt, GFP_KERNEL);
276 if (ret) {
277 virtio_transport_free_pkt(pkt);
278 break;
279 }
280 vsock->rx_buf_nr++;
281 } while (vq->num_free);
282 if (vsock->rx_buf_nr > vsock->rx_buf_max_nr)
283 vsock->rx_buf_max_nr = vsock->rx_buf_nr;
284 virtqueue_kick(vq);
285}
286
287static void virtio_transport_tx_work(struct work_struct *work)
288{
289 struct virtio_vsock *vsock =
290 container_of(work, struct virtio_vsock, tx_work);
291 struct virtqueue *vq;
292 bool added = false;
293
294 vq = vsock->vqs[VSOCK_VQ_TX];
295 mutex_lock(&vsock->tx_lock);
Stefano Garzarellab9175072019-07-05 13:04:53 +0200296
297 if (!vsock->tx_run)
298 goto out;
299
Asias He0ea9e1d2016-07-28 15:36:33 +0100300 do {
301 struct virtio_vsock_pkt *pkt;
302 unsigned int len;
303
304 virtqueue_disable_cb(vq);
305 while ((pkt = virtqueue_get_buf(vq, &len)) != NULL) {
306 virtio_transport_free_pkt(pkt);
307 added = true;
308 }
309 } while (!virtqueue_enable_cb(vq));
Stefano Garzarellab9175072019-07-05 13:04:53 +0200310
311out:
Asias He0ea9e1d2016-07-28 15:36:33 +0100312 mutex_unlock(&vsock->tx_lock);
313
314 if (added)
315 queue_work(virtio_vsock_workqueue, &vsock->send_pkt_work);
316}
317
318/* Is there space left for replies to rx packets? */
319static bool virtio_transport_more_replies(struct virtio_vsock *vsock)
320{
321 struct virtqueue *vq = vsock->vqs[VSOCK_VQ_RX];
322 int val;
323
324 smp_rmb(); /* paired with atomic_inc() and atomic_dec_return() */
325 val = atomic_read(&vsock->queued_replies);
326
327 return val < virtqueue_get_vring_size(vq);
328}
329
Asias He0ea9e1d2016-07-28 15:36:33 +0100330/* event_lock must be held */
331static int virtio_vsock_event_fill_one(struct virtio_vsock *vsock,
332 struct virtio_vsock_event *event)
333{
334 struct scatterlist sg;
335 struct virtqueue *vq;
336
337 vq = vsock->vqs[VSOCK_VQ_EVENT];
338
339 sg_init_one(&sg, event, sizeof(*event));
340
341 return virtqueue_add_inbuf(vq, &sg, 1, event, GFP_KERNEL);
342}
343
344/* event_lock must be held */
345static void virtio_vsock_event_fill(struct virtio_vsock *vsock)
346{
347 size_t i;
348
349 for (i = 0; i < ARRAY_SIZE(vsock->event_list); i++) {
350 struct virtio_vsock_event *event = &vsock->event_list[i];
351
352 virtio_vsock_event_fill_one(vsock, event);
353 }
354
355 virtqueue_kick(vsock->vqs[VSOCK_VQ_EVENT]);
356}
357
358static void virtio_vsock_reset_sock(struct sock *sk)
359{
Longpeng(Mike)49b0b6f2021-08-12 13:30:56 +0800360 /* vmci_transport.c doesn't take sk_lock here either. At least we're
361 * under vsock_table_lock so the sock cannot disappear while we're
362 * executing.
363 */
364
Stefan Hajnoczi3b4477d2017-10-05 16:46:52 -0400365 sk->sk_state = TCP_CLOSE;
Asias He0ea9e1d2016-07-28 15:36:33 +0100366 sk->sk_err = ECONNRESET;
Alexander Aringe3ae2362021-06-27 18:48:21 -0400367 sk_error_report(sk);
Asias He0ea9e1d2016-07-28 15:36:33 +0100368}
369
370static void virtio_vsock_update_guest_cid(struct virtio_vsock *vsock)
371{
372 struct virtio_device *vdev = vsock->vdev;
Michael S. Tsirkin6c7efaf2016-12-06 06:03:34 +0200373 __le64 guest_cid;
Asias He0ea9e1d2016-07-28 15:36:33 +0100374
375 vdev->config->get(vdev, offsetof(struct virtio_vsock_config, guest_cid),
376 &guest_cid, sizeof(guest_cid));
377 vsock->guest_cid = le64_to_cpu(guest_cid);
378}
379
380/* event_lock must be held */
381static void virtio_vsock_event_handle(struct virtio_vsock *vsock,
382 struct virtio_vsock_event *event)
383{
384 switch (le32_to_cpu(event->id)) {
385 case VIRTIO_VSOCK_EVENT_TRANSPORT_RESET:
386 virtio_vsock_update_guest_cid(vsock);
387 vsock_for_each_connected_socket(virtio_vsock_reset_sock);
388 break;
389 }
390}
391
392static void virtio_transport_event_work(struct work_struct *work)
393{
394 struct virtio_vsock *vsock =
395 container_of(work, struct virtio_vsock, event_work);
396 struct virtqueue *vq;
397
398 vq = vsock->vqs[VSOCK_VQ_EVENT];
399
400 mutex_lock(&vsock->event_lock);
401
Stefano Garzarellab9175072019-07-05 13:04:53 +0200402 if (!vsock->event_run)
403 goto out;
404
Asias He0ea9e1d2016-07-28 15:36:33 +0100405 do {
406 struct virtio_vsock_event *event;
407 unsigned int len;
408
409 virtqueue_disable_cb(vq);
410 while ((event = virtqueue_get_buf(vq, &len)) != NULL) {
411 if (len == sizeof(*event))
412 virtio_vsock_event_handle(vsock, event);
413
414 virtio_vsock_event_fill_one(vsock, event);
415 }
416 } while (!virtqueue_enable_cb(vq));
417
418 virtqueue_kick(vsock->vqs[VSOCK_VQ_EVENT]);
Stefano Garzarellab9175072019-07-05 13:04:53 +0200419out:
Asias He0ea9e1d2016-07-28 15:36:33 +0100420 mutex_unlock(&vsock->event_lock);
421}
422
423static void virtio_vsock_event_done(struct virtqueue *vq)
424{
425 struct virtio_vsock *vsock = vq->vdev->priv;
426
427 if (!vsock)
428 return;
429 queue_work(virtio_vsock_workqueue, &vsock->event_work);
430}
431
432static void virtio_vsock_tx_done(struct virtqueue *vq)
433{
434 struct virtio_vsock *vsock = vq->vdev->priv;
435
436 if (!vsock)
437 return;
438 queue_work(virtio_vsock_workqueue, &vsock->tx_work);
439}
440
441static void virtio_vsock_rx_done(struct virtqueue *vq)
442{
443 struct virtio_vsock *vsock = vq->vdev->priv;
444
445 if (!vsock)
446 return;
447 queue_work(virtio_vsock_workqueue, &vsock->rx_work);
448}
449
Arseny Krasnov53efbba2021-06-11 14:13:22 +0300450static bool virtio_transport_seqpacket_allow(u32 remote_cid);
451
Asias He0ea9e1d2016-07-28 15:36:33 +0100452static struct virtio_transport virtio_transport = {
453 .transport = {
Stefano Garzarella6a2c09622019-11-14 10:57:48 +0100454 .module = THIS_MODULE,
455
Asias He0ea9e1d2016-07-28 15:36:33 +0100456 .get_local_cid = virtio_transport_get_local_cid,
457
458 .init = virtio_transport_do_socket_init,
459 .destruct = virtio_transport_destruct,
460 .release = virtio_transport_release,
461 .connect = virtio_transport_connect,
462 .shutdown = virtio_transport_shutdown,
Peng Tao073b4f22017-03-15 09:32:16 +0800463 .cancel_pkt = virtio_transport_cancel_pkt,
Asias He0ea9e1d2016-07-28 15:36:33 +0100464
465 .dgram_bind = virtio_transport_dgram_bind,
466 .dgram_dequeue = virtio_transport_dgram_dequeue,
467 .dgram_enqueue = virtio_transport_dgram_enqueue,
468 .dgram_allow = virtio_transport_dgram_allow,
469
470 .stream_dequeue = virtio_transport_stream_dequeue,
471 .stream_enqueue = virtio_transport_stream_enqueue,
472 .stream_has_data = virtio_transport_stream_has_data,
473 .stream_has_space = virtio_transport_stream_has_space,
474 .stream_rcvhiwat = virtio_transport_stream_rcvhiwat,
475 .stream_is_active = virtio_transport_stream_is_active,
476 .stream_allow = virtio_transport_stream_allow,
477
Arseny Krasnov53efbba2021-06-11 14:13:22 +0300478 .seqpacket_dequeue = virtio_transport_seqpacket_dequeue,
479 .seqpacket_enqueue = virtio_transport_seqpacket_enqueue,
480 .seqpacket_allow = virtio_transport_seqpacket_allow,
481 .seqpacket_has_data = virtio_transport_seqpacket_has_data,
482
Asias He0ea9e1d2016-07-28 15:36:33 +0100483 .notify_poll_in = virtio_transport_notify_poll_in,
484 .notify_poll_out = virtio_transport_notify_poll_out,
485 .notify_recv_init = virtio_transport_notify_recv_init,
486 .notify_recv_pre_block = virtio_transport_notify_recv_pre_block,
487 .notify_recv_pre_dequeue = virtio_transport_notify_recv_pre_dequeue,
488 .notify_recv_post_dequeue = virtio_transport_notify_recv_post_dequeue,
489 .notify_send_init = virtio_transport_notify_send_init,
490 .notify_send_pre_block = virtio_transport_notify_send_pre_block,
491 .notify_send_pre_enqueue = virtio_transport_notify_send_pre_enqueue,
492 .notify_send_post_enqueue = virtio_transport_notify_send_post_enqueue,
Stefano Garzarellab9f2b0f2019-11-14 10:57:42 +0100493 .notify_buffer_size = virtio_transport_notify_buffer_size,
Asias He0ea9e1d2016-07-28 15:36:33 +0100494 },
495
496 .send_pkt = virtio_transport_send_pkt,
497};
498
Arseny Krasnov53efbba2021-06-11 14:13:22 +0300499static bool virtio_transport_seqpacket_allow(u32 remote_cid)
500{
501 struct virtio_vsock *vsock;
502 bool seqpacket_allow;
503
Eric Dumazet64295f02021-06-21 07:53:48 -0700504 seqpacket_allow = false;
Arseny Krasnov53efbba2021-06-11 14:13:22 +0300505 rcu_read_lock();
506 vsock = rcu_dereference(the_virtio_vsock);
Eric Dumazet64295f02021-06-21 07:53:48 -0700507 if (vsock)
508 seqpacket_allow = vsock->seqpacket_allow;
Arseny Krasnov53efbba2021-06-11 14:13:22 +0300509 rcu_read_unlock();
510
511 return seqpacket_allow;
512}
513
Stefano Garzarella4c7246d2019-11-14 10:57:40 +0100514static void virtio_transport_rx_work(struct work_struct *work)
515{
516 struct virtio_vsock *vsock =
517 container_of(work, struct virtio_vsock, rx_work);
518 struct virtqueue *vq;
519
520 vq = vsock->vqs[VSOCK_VQ_RX];
521
522 mutex_lock(&vsock->rx_lock);
523
524 if (!vsock->rx_run)
525 goto out;
526
527 do {
528 virtqueue_disable_cb(vq);
529 for (;;) {
530 struct virtio_vsock_pkt *pkt;
531 unsigned int len;
532
533 if (!virtio_transport_more_replies(vsock)) {
534 /* Stop rx until the device processes already
535 * pending replies. Leave rx virtqueue
536 * callbacks disabled.
537 */
538 goto out;
539 }
540
541 pkt = virtqueue_get_buf(vq, &len);
542 if (!pkt) {
543 break;
544 }
545
546 vsock->rx_buf_nr--;
547
548 /* Drop short/long packets */
549 if (unlikely(len < sizeof(pkt->hdr) ||
550 len > sizeof(pkt->hdr) + pkt->len)) {
551 virtio_transport_free_pkt(pkt);
552 continue;
553 }
554
555 pkt->len = len - sizeof(pkt->hdr);
556 virtio_transport_deliver_tap_pkt(pkt);
557 virtio_transport_recv_pkt(&virtio_transport, pkt);
558 }
559 } while (!virtqueue_enable_cb(vq));
560
561out:
562 if (vsock->rx_buf_nr < vsock->rx_buf_max_nr / 2)
563 virtio_vsock_rx_fill(vsock);
564 mutex_unlock(&vsock->rx_lock);
565}
566
Asias He0ea9e1d2016-07-28 15:36:33 +0100567static int virtio_vsock_probe(struct virtio_device *vdev)
568{
569 vq_callback_t *callbacks[] = {
570 virtio_vsock_rx_done,
571 virtio_vsock_tx_done,
572 virtio_vsock_event_done,
573 };
574 static const char * const names[] = {
575 "rx",
576 "tx",
577 "event",
578 };
579 struct virtio_vsock *vsock = NULL;
580 int ret;
581
582 ret = mutex_lock_interruptible(&the_virtio_vsock_mutex);
583 if (ret)
584 return ret;
585
586 /* Only one virtio-vsock device per guest is supported */
Stefano Garzarella0deab082019-07-05 13:04:52 +0200587 if (rcu_dereference_protected(the_virtio_vsock,
588 lockdep_is_held(&the_virtio_vsock_mutex))) {
Asias He0ea9e1d2016-07-28 15:36:33 +0100589 ret = -EBUSY;
590 goto out;
591 }
592
593 vsock = kzalloc(sizeof(*vsock), GFP_KERNEL);
594 if (!vsock) {
595 ret = -ENOMEM;
596 goto out;
597 }
598
599 vsock->vdev = vdev;
600
Michael S. Tsirkin9b2bbdb2017-03-06 18:19:39 +0200601 ret = virtio_find_vqs(vsock->vdev, VSOCK_VQ_MAX,
602 vsock->vqs, callbacks, names,
603 NULL);
Asias He0ea9e1d2016-07-28 15:36:33 +0100604 if (ret < 0)
605 goto out;
606
607 virtio_vsock_update_guest_cid(vsock);
608
Asias He0ea9e1d2016-07-28 15:36:33 +0100609 vsock->rx_buf_nr = 0;
610 vsock->rx_buf_max_nr = 0;
611 atomic_set(&vsock->queued_replies, 0);
612
Asias He0ea9e1d2016-07-28 15:36:33 +0100613 mutex_init(&vsock->tx_lock);
614 mutex_init(&vsock->rx_lock);
615 mutex_init(&vsock->event_lock);
616 spin_lock_init(&vsock->send_pkt_list_lock);
617 INIT_LIST_HEAD(&vsock->send_pkt_list);
618 INIT_WORK(&vsock->rx_work, virtio_transport_rx_work);
619 INIT_WORK(&vsock->tx_work, virtio_transport_tx_work);
620 INIT_WORK(&vsock->event_work, virtio_transport_event_work);
621 INIT_WORK(&vsock->send_pkt_work, virtio_transport_send_pkt_work);
622
Stefano Garzarellab9175072019-07-05 13:04:53 +0200623 mutex_lock(&vsock->tx_lock);
624 vsock->tx_run = true;
625 mutex_unlock(&vsock->tx_lock);
626
Asias He0ea9e1d2016-07-28 15:36:33 +0100627 mutex_lock(&vsock->rx_lock);
628 virtio_vsock_rx_fill(vsock);
Stefano Garzarellab9175072019-07-05 13:04:53 +0200629 vsock->rx_run = true;
Asias He0ea9e1d2016-07-28 15:36:33 +0100630 mutex_unlock(&vsock->rx_lock);
631
632 mutex_lock(&vsock->event_lock);
633 virtio_vsock_event_fill(vsock);
Stefano Garzarellab9175072019-07-05 13:04:53 +0200634 vsock->event_run = true;
Asias He0ea9e1d2016-07-28 15:36:33 +0100635 mutex_unlock(&vsock->event_lock);
636
Arseny Krasnov53efbba2021-06-11 14:13:22 +0300637 if (virtio_has_feature(vdev, VIRTIO_VSOCK_F_SEQPACKET))
638 vsock->seqpacket_allow = true;
639
Stefano Garzarella0deab082019-07-05 13:04:52 +0200640 vdev->priv = vsock;
641 rcu_assign_pointer(the_virtio_vsock, vsock);
642
Asias He0ea9e1d2016-07-28 15:36:33 +0100643 mutex_unlock(&the_virtio_vsock_mutex);
Arseny Krasnov53efbba2021-06-11 14:13:22 +0300644
Asias He0ea9e1d2016-07-28 15:36:33 +0100645 return 0;
646
Asias He0ea9e1d2016-07-28 15:36:33 +0100647out:
648 kfree(vsock);
649 mutex_unlock(&the_virtio_vsock_mutex);
650 return ret;
651}
652
653static void virtio_vsock_remove(struct virtio_device *vdev)
654{
655 struct virtio_vsock *vsock = vdev->priv;
656 struct virtio_vsock_pkt *pkt;
657
Stefano Garzarella0deab082019-07-05 13:04:52 +0200658 mutex_lock(&the_virtio_vsock_mutex);
659
660 vdev->priv = NULL;
661 rcu_assign_pointer(the_virtio_vsock, NULL);
662 synchronize_rcu();
663
Stefano Garzarella85965482019-02-01 12:42:07 +0100664 /* Reset all connected sockets when the device disappear */
665 vsock_for_each_connected_socket(virtio_vsock_reset_sock);
666
Stefano Garzarellab9175072019-07-05 13:04:53 +0200667 /* Stop all work handlers to make sure no one is accessing the device,
Michael S. Tsirkind9679d02021-10-13 06:55:44 -0400668 * so we can safely call virtio_reset_device().
Stefano Garzarellab9175072019-07-05 13:04:53 +0200669 */
670 mutex_lock(&vsock->rx_lock);
671 vsock->rx_run = false;
672 mutex_unlock(&vsock->rx_lock);
673
674 mutex_lock(&vsock->tx_lock);
675 vsock->tx_run = false;
676 mutex_unlock(&vsock->tx_lock);
677
678 mutex_lock(&vsock->event_lock);
679 vsock->event_run = false;
680 mutex_unlock(&vsock->event_lock);
681
682 /* Flush all device writes and interrupts, device will not use any
683 * more buffers.
684 */
Michael S. Tsirkind9679d02021-10-13 06:55:44 -0400685 virtio_reset_device(vdev);
Asias He0ea9e1d2016-07-28 15:36:33 +0100686
687 mutex_lock(&vsock->rx_lock);
688 while ((pkt = virtqueue_detach_unused_buf(vsock->vqs[VSOCK_VQ_RX])))
689 virtio_transport_free_pkt(pkt);
690 mutex_unlock(&vsock->rx_lock);
691
692 mutex_lock(&vsock->tx_lock);
693 while ((pkt = virtqueue_detach_unused_buf(vsock->vqs[VSOCK_VQ_TX])))
694 virtio_transport_free_pkt(pkt);
695 mutex_unlock(&vsock->tx_lock);
696
697 spin_lock_bh(&vsock->send_pkt_list_lock);
698 while (!list_empty(&vsock->send_pkt_list)) {
699 pkt = list_first_entry(&vsock->send_pkt_list,
700 struct virtio_vsock_pkt, list);
701 list_del(&pkt->list);
702 virtio_transport_free_pkt(pkt);
703 }
704 spin_unlock_bh(&vsock->send_pkt_list_lock);
705
Stefano Garzarellab9175072019-07-05 13:04:53 +0200706 /* Delete virtqueues and flush outstanding callbacks if any */
Asias He0ea9e1d2016-07-28 15:36:33 +0100707 vdev->config->del_vqs(vdev);
708
Stefano Garzarellae2261212019-07-05 13:04:54 +0200709 /* Other works can be queued before 'config->del_vqs()', so we flush
710 * all works before to free the vsock object to avoid use after free.
711 */
Stefano Garzarellae2261212019-07-05 13:04:54 +0200712 flush_work(&vsock->rx_work);
713 flush_work(&vsock->tx_work);
714 flush_work(&vsock->event_work);
715 flush_work(&vsock->send_pkt_work);
716
Stefano Garzarella0deab082019-07-05 13:04:52 +0200717 mutex_unlock(&the_virtio_vsock_mutex);
718
Asias He0ea9e1d2016-07-28 15:36:33 +0100719 kfree(vsock);
720}
721
722static struct virtio_device_id id_table[] = {
723 { VIRTIO_ID_VSOCK, VIRTIO_DEV_ANY_ID },
724 { 0 },
725};
726
727static unsigned int features[] = {
Arseny Krasnov53efbba2021-06-11 14:13:22 +0300728 VIRTIO_VSOCK_F_SEQPACKET
Asias He0ea9e1d2016-07-28 15:36:33 +0100729};
730
731static struct virtio_driver virtio_vsock_driver = {
732 .feature_table = features,
733 .feature_table_size = ARRAY_SIZE(features),
734 .driver.name = KBUILD_MODNAME,
735 .driver.owner = THIS_MODULE,
736 .id_table = id_table,
737 .probe = virtio_vsock_probe,
738 .remove = virtio_vsock_remove,
739};
740
741static int __init virtio_vsock_init(void)
742{
743 int ret;
744
745 virtio_vsock_workqueue = alloc_workqueue("virtio_vsock", 0, 0);
746 if (!virtio_vsock_workqueue)
747 return -ENOMEM;
Stefano Garzarella22b5c0b2019-02-01 12:42:06 +0100748
Stefano Garzarellac0cfa2d2019-11-14 10:57:46 +0100749 ret = vsock_core_register(&virtio_transport.transport,
750 VSOCK_TRANSPORT_F_G2H);
Asias He0ea9e1d2016-07-28 15:36:33 +0100751 if (ret)
Stefano Garzarella22b5c0b2019-02-01 12:42:06 +0100752 goto out_wq;
753
Jorge E. Moreiraba95e5df2019-05-16 13:51:07 -0700754 ret = register_virtio_driver(&virtio_vsock_driver);
Stefano Garzarella22b5c0b2019-02-01 12:42:06 +0100755 if (ret)
Jorge E. Moreiraba95e5df2019-05-16 13:51:07 -0700756 goto out_vci;
Stefano Garzarella22b5c0b2019-02-01 12:42:06 +0100757
758 return 0;
759
Jorge E. Moreiraba95e5df2019-05-16 13:51:07 -0700760out_vci:
Stefano Garzarellac0cfa2d2019-11-14 10:57:46 +0100761 vsock_core_unregister(&virtio_transport.transport);
Stefano Garzarella22b5c0b2019-02-01 12:42:06 +0100762out_wq:
763 destroy_workqueue(virtio_vsock_workqueue);
Asias He0ea9e1d2016-07-28 15:36:33 +0100764 return ret;
765}
766
767static void __exit virtio_vsock_exit(void)
768{
Asias He0ea9e1d2016-07-28 15:36:33 +0100769 unregister_virtio_driver(&virtio_vsock_driver);
Stefano Garzarellac0cfa2d2019-11-14 10:57:46 +0100770 vsock_core_unregister(&virtio_transport.transport);
Asias He0ea9e1d2016-07-28 15:36:33 +0100771 destroy_workqueue(virtio_vsock_workqueue);
772}
773
774module_init(virtio_vsock_init);
775module_exit(virtio_vsock_exit);
776MODULE_LICENSE("GPL v2");
777MODULE_AUTHOR("Asias He");
778MODULE_DESCRIPTION("virtio transport for vsock");
779MODULE_DEVICE_TABLE(virtio, id_table);