projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'mac80211-for-net-2019-11-08' of git://git.kernel.org/pub/scm/linux/kernel...
[linux-2.6-block.git] / net / vmw_vsock / virtio_transport_common.c
CommitLineData
7a338472 1// SPDX-License-Identifier: GPL-2.0-only
06a8fc78
AH		 2/*
3 * common code for virtio vsock
4 *
5 * Copyright (C) 2013-2015 Red Hat, Inc.
6 * Author: Asias He <asias@redhat.com>
7 * Stefan Hajnoczi <stefanha@redhat.com>
06a8fc78
AH		 8 */
9#include <linux/spinlock.h>
10#include <linux/module.h>
174cd4b1 11#include <linux/sched/signal.h>
06a8fc78
AH		 12#include <linux/ctype.h>
13#include <linux/list.h>
14#include <linux/virtio.h>
15#include <linux/virtio_ids.h>
16#include <linux/virtio_config.h>
17#include <linux/virtio_vsock.h>
82dfb540 18#include <uapi/linux/vsockmon.h>
06a8fc78
AH		 19
20#include <net/sock.h>
21#include <net/af_vsock.h>
22
23#define CREATE_TRACE_POINTS
24#include <trace/events/vsock_virtio_transport_common.h>
25
26/* How long to wait for graceful shutdown of a connection */
27#define VSOCK_CLOSE_TIMEOUT (8 * HZ)
28
473c7391
SG		 29/* Threshold for detecting small packets to copy */
30#define GOOD_COPY_LEN 128
31
06a8fc78
AH		 32static const struct virtio_transport *virtio_transport_get_ops(void)
33{
34 const struct vsock_transport *t = vsock_core_get_transport();
35
36 return container_of(t, struct virtio_transport, transport);
37}
38
819483d8 39static struct virtio_vsock_pkt *
06a8fc78
AH		 40virtio_transport_alloc_pkt(struct virtio_vsock_pkt_info *info,
41 size_t len,
42 u32 src_cid,
43 u32 src_port,
44 u32 dst_cid,
45 u32 dst_port)
46{
47 struct virtio_vsock_pkt *pkt;
48 int err;
49
50 pkt = kzalloc(sizeof(*pkt), GFP_KERNEL);
51 if (!pkt)
52 return NULL;
53
54 pkt->hdr.type = cpu_to_le16(info->type);
55 pkt->hdr.op = cpu_to_le16(info->op);
56 pkt->hdr.src_cid = cpu_to_le64(src_cid);
57 pkt->hdr.dst_cid = cpu_to_le64(dst_cid);
58 pkt->hdr.src_port = cpu_to_le32(src_port);
59 pkt->hdr.dst_port = cpu_to_le32(dst_port);
60 pkt->hdr.flags = cpu_to_le32(info->flags);
61 pkt->len = len;
62 pkt->hdr.len = cpu_to_le32(len);
63 pkt->reply = info->reply;
36d277ba 64 pkt->vsk = info->vsk;
06a8fc78
AH		 65
66 if (info->msg && len > 0) {
67 pkt->buf = kmalloc(len, GFP_KERNEL);
68 if (!pkt->buf)
69 goto out_pkt;
473c7391
SG		 70
71 pkt->buf_len = len;
72
06a8fc78
AH		 73 err = memcpy_from_msg(pkt->buf, info->msg, len);
74 if (err)
75 goto out;
76 }
77
78 trace_virtio_transport_alloc_pkt(src_cid, src_port,
79 dst_cid, dst_port,
80 len,
81 info->type,
82 info->op,
83 info->flags);
84
85 return pkt;
86
87out:
88 kfree(pkt->buf);
89out_pkt:
90 kfree(pkt);
91 return NULL;
92}
06a8fc78 93
82dfb540
GG		 94/* Packet capture */
95static struct sk_buff *virtio_transport_build_skb(void *opaque)
96{
97 struct virtio_vsock_pkt *pkt = opaque;
82dfb540
GG		 98 struct af_vsockmon_hdr *hdr;
99 struct sk_buff *skb;
6dbd3e66
SG		 100 size_t payload_len;
101 void *payload_buf;
82dfb540 102
6dbd3e66
SG		 103 /* A packet could be split to fit the RX buffer, so we can retrieve
104 * the payload length from the header and the buffer pointer taking
105 * care of the offset in the original packet.
106 */
107 payload_len = le32_to_cpu(pkt->hdr.len);
108 payload_buf = pkt->buf + pkt->off;
109
110 skb = alloc_skb(sizeof(*hdr) + sizeof(pkt->hdr) + payload_len,
82dfb540
GG		 111 GFP_ATOMIC);
112 if (!skb)
113 return NULL;
114
4df864c1 115 hdr = skb_put(skb, sizeof(*hdr));
82dfb540
GG		 116
117 /* pkt->hdr is little-endian so no need to byteswap here */
118 hdr->src_cid = pkt->hdr.src_cid;
119 hdr->src_port = pkt->hdr.src_port;
120 hdr->dst_cid = pkt->hdr.dst_cid;
121 hdr->dst_port = pkt->hdr.dst_port;
122
123 hdr->transport = cpu_to_le16(AF_VSOCK_TRANSPORT_VIRTIO);
124 hdr->len = cpu_to_le16(sizeof(pkt->hdr));
125 memset(hdr->reserved, 0, sizeof(hdr->reserved));
126
127 switch (le16_to_cpu(pkt->hdr.op)) {
128 case VIRTIO_VSOCK_OP_REQUEST:
129 case VIRTIO_VSOCK_OP_RESPONSE:
130 hdr->op = cpu_to_le16(AF_VSOCK_OP_CONNECT);
131 break;
132 case VIRTIO_VSOCK_OP_RST:
133 case VIRTIO_VSOCK_OP_SHUTDOWN:
134 hdr->op = cpu_to_le16(AF_VSOCK_OP_DISCONNECT);
135 break;
136 case VIRTIO_VSOCK_OP_RW:
137 hdr->op = cpu_to_le16(AF_VSOCK_OP_PAYLOAD);
138 break;
139 case VIRTIO_VSOCK_OP_CREDIT_UPDATE:
140 case VIRTIO_VSOCK_OP_CREDIT_REQUEST:
141 hdr->op = cpu_to_le16(AF_VSOCK_OP_CONTROL);
142 break;
143 default:
144 hdr->op = cpu_to_le16(AF_VSOCK_OP_UNKNOWN);
145 break;
146 }
147
b952f4df 148 skb_put_data(skb, &pkt->hdr, sizeof(pkt->hdr));
82dfb540 149
6dbd3e66
SG		 150 if (payload_len) {
151 skb_put_data(skb, payload_buf, payload_len);
82dfb540
GG		 152 }
153
154 return skb;
155}
156
157void virtio_transport_deliver_tap_pkt(struct virtio_vsock_pkt *pkt)
158{
159 vsock_deliver_tap(virtio_transport_build_skb, pkt);
160}
161EXPORT_SYMBOL_GPL(virtio_transport_deliver_tap_pkt);
162
06a8fc78
AH		 163static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
164 struct virtio_vsock_pkt_info *info)
165{
166 u32 src_cid, src_port, dst_cid, dst_port;
167 struct virtio_vsock_sock *vvs;
168 struct virtio_vsock_pkt *pkt;
169 u32 pkt_len = info->pkt_len;
170
171 src_cid = vm_sockets_get_local_cid();
172 src_port = vsk->local_addr.svm_port;
173 if (!info->remote_cid) {
174 dst_cid = vsk->remote_addr.svm_cid;
175 dst_port = vsk->remote_addr.svm_port;
176 } else {
177 dst_cid = info->remote_cid;
178 dst_port = info->remote_port;
179 }
180
181 vvs = vsk->trans;
182
183 /* we can send less than pkt_len bytes */
0038ff35
SG		 184 if (pkt_len > VIRTIO_VSOCK_MAX_PKT_BUF_SIZE)
185 pkt_len = VIRTIO_VSOCK_MAX_PKT_BUF_SIZE;
06a8fc78
AH		 186
187 /* virtio_transport_get_credit might return less than pkt_len credit */
188 pkt_len = virtio_transport_get_credit(vvs, pkt_len);
189
190 /* Do not send zero length OP_RW pkt */
191 if (pkt_len == 0 && info->op == VIRTIO_VSOCK_OP_RW)
192 return pkt_len;
193
194 pkt = virtio_transport_alloc_pkt(info, pkt_len,
195 src_cid, src_port,
196 dst_cid, dst_port);
197 if (!pkt) {
198 virtio_transport_put_credit(vvs, pkt_len);
199 return -ENOMEM;
200 }
201
202 virtio_transport_inc_tx_pkt(vvs, pkt);
203
204 return virtio_transport_get_ops()->send_pkt(pkt);
205}
206
ae6fcfbf 207static bool virtio_transport_inc_rx_pkt(struct virtio_vsock_sock *vvs,
06a8fc78
AH		 208 struct virtio_vsock_pkt *pkt)
209{
ae6fcfbf
SG		 210 if (vvs->rx_bytes + pkt->len > vvs->buf_alloc)
211 return false;
212
06a8fc78 213 vvs->rx_bytes += pkt->len;
ae6fcfbf 214 return true;
06a8fc78
AH		 215}
216
217static void virtio_transport_dec_rx_pkt(struct virtio_vsock_sock *vvs,
218 struct virtio_vsock_pkt *pkt)
219{
220 vvs->rx_bytes -= pkt->len;
221 vvs->fwd_cnt += pkt->len;
222}
223
224void virtio_transport_inc_tx_pkt(struct virtio_vsock_sock *vvs, struct virtio_vsock_pkt *pkt)
225{
9632e9f6 226 spin_lock_bh(&vvs->rx_lock);
b89d882d 227 vvs->last_fwd_cnt = vvs->fwd_cnt;
06a8fc78
AH		 228 pkt->hdr.fwd_cnt = cpu_to_le32(vvs->fwd_cnt);
229 pkt->hdr.buf_alloc = cpu_to_le32(vvs->buf_alloc);
9632e9f6 230 spin_unlock_bh(&vvs->rx_lock);
06a8fc78
AH		 231}
232EXPORT_SYMBOL_GPL(virtio_transport_inc_tx_pkt);
233
234u32 virtio_transport_get_credit(struct virtio_vsock_sock *vvs, u32 credit)
235{
236 u32 ret;
237
238 spin_lock_bh(&vvs->tx_lock);
239 ret = vvs->peer_buf_alloc - (vvs->tx_cnt - vvs->peer_fwd_cnt);
240 if (ret > credit)
241 ret = credit;
242 vvs->tx_cnt += ret;
243 spin_unlock_bh(&vvs->tx_lock);
244
245 return ret;
246}
247EXPORT_SYMBOL_GPL(virtio_transport_get_credit);
248
249void virtio_transport_put_credit(struct virtio_vsock_sock *vvs, u32 credit)
250{
251 spin_lock_bh(&vvs->tx_lock);
252 vvs->tx_cnt -= credit;
253 spin_unlock_bh(&vvs->tx_lock);
254}
255EXPORT_SYMBOL_GPL(virtio_transport_put_credit);
256
257static int virtio_transport_send_credit_update(struct vsock_sock *vsk,
258 int type,
259 struct virtio_vsock_hdr *hdr)
260{
261 struct virtio_vsock_pkt_info info = {
262 .op = VIRTIO_VSOCK_OP_CREDIT_UPDATE,
263 .type = type,
36d277ba 264 .vsk = vsk,
06a8fc78
AH		 265 };
266
267 return virtio_transport_send_pkt_info(vsk, &info);
268}
269
270static ssize_t
271virtio_transport_stream_do_dequeue(struct vsock_sock *vsk,
272 struct msghdr *msg,
273 size_t len)
274{
275 struct virtio_vsock_sock *vvs = vsk->trans;
276 struct virtio_vsock_pkt *pkt;
277 size_t bytes, total = 0;
b89d882d 278 u32 free_space;
06a8fc78
AH		 279 int err = -EFAULT;
280
281 spin_lock_bh(&vvs->rx_lock);
282 while (total < len && !list_empty(&vvs->rx_queue)) {
283 pkt = list_first_entry(&vvs->rx_queue,
284 struct virtio_vsock_pkt, list);
285
286 bytes = len - total;
287 if (bytes > pkt->len - pkt->off)
288 bytes = pkt->len - pkt->off;
289
290 /* sk_lock is held by caller so no one else can dequeue.
291 * Unlock rx_lock since memcpy_to_msg() may sleep.
292 */
293 spin_unlock_bh(&vvs->rx_lock);
294
295 err = memcpy_to_msg(msg, pkt->buf + pkt->off, bytes);
296 if (err)
297 goto out;
298
299 spin_lock_bh(&vvs->rx_lock);
300
301 total += bytes;
302 pkt->off += bytes;
303 if (pkt->off == pkt->len) {
304 virtio_transport_dec_rx_pkt(vvs, pkt);
305 list_del(&pkt->list);
306 virtio_transport_free_pkt(pkt);
307 }
308 }
b89d882d
SG		 309
310 free_space = vvs->buf_alloc - (vvs->fwd_cnt - vvs->last_fwd_cnt);
311
06a8fc78
AH		 312 spin_unlock_bh(&vvs->rx_lock);
313
f4d7c8e3
MT		 314 /* To reduce the number of credit update messages,
315 * don't update credits as long as lots of space is available.
316 * Note: the limit chosen here is arbitrary. Setting the limit
317 * too high causes extra messages. Too low causes transmitter
318 * stalls. As stalls are in theory more expensive than extra
319 * messages, we set the limit to a high value. TODO: experiment
320 * with different values.
b89d882d
SG		 321 */
322 if (free_space < VIRTIO_VSOCK_MAX_PKT_BUF_SIZE) {
323 virtio_transport_send_credit_update(vsk,
324 VIRTIO_VSOCK_TYPE_STREAM,
325 NULL);
326 }
06a8fc78
AH		 327
328 return total;
329
330out:
331 if (total)
332 err = total;
333 return err;
334}
335
336ssize_t
337virtio_transport_stream_dequeue(struct vsock_sock *vsk,
338 struct msghdr *msg,
339 size_t len, int flags)
340{
341 if (flags & MSG_PEEK)
342 return -EOPNOTSUPP;
343
344 return virtio_transport_stream_do_dequeue(vsk, msg, len);
345}
346EXPORT_SYMBOL_GPL(virtio_transport_stream_dequeue);
347
348int
349virtio_transport_dgram_dequeue(struct vsock_sock *vsk,
350 struct msghdr *msg,
351 size_t len, int flags)
352{
353 return -EOPNOTSUPP;
354}
355EXPORT_SYMBOL_GPL(virtio_transport_dgram_dequeue);
356
357s64 virtio_transport_stream_has_data(struct vsock_sock *vsk)
358{
359 struct virtio_vsock_sock *vvs = vsk->trans;
360 s64 bytes;
361
362 spin_lock_bh(&vvs->rx_lock);
363 bytes = vvs->rx_bytes;
364 spin_unlock_bh(&vvs->rx_lock);
365
366 return bytes;
367}
368EXPORT_SYMBOL_GPL(virtio_transport_stream_has_data);
369
370static s64 virtio_transport_has_space(struct vsock_sock *vsk)
371{
372 struct virtio_vsock_sock *vvs = vsk->trans;
373 s64 bytes;
374
375 bytes = vvs->peer_buf_alloc - (vvs->tx_cnt - vvs->peer_fwd_cnt);
376 if (bytes < 0)
377 bytes = 0;
378
379 return bytes;
380}
381
382s64 virtio_transport_stream_has_space(struct vsock_sock *vsk)
383{
384 struct virtio_vsock_sock *vvs = vsk->trans;
385 s64 bytes;
386
387 spin_lock_bh(&vvs->tx_lock);
388 bytes = virtio_transport_has_space(vsk);
389 spin_unlock_bh(&vvs->tx_lock);
390
391 return bytes;
392}
393EXPORT_SYMBOL_GPL(virtio_transport_stream_has_space);
394
395int virtio_transport_do_socket_init(struct vsock_sock *vsk,
396 struct vsock_sock *psk)
397{
398 struct virtio_vsock_sock *vvs;
399
400 vvs = kzalloc(sizeof(*vvs), GFP_KERNEL);
401 if (!vvs)
402 return -ENOMEM;
403
404 vsk->trans = vvs;
405 vvs->vsk = vsk;
406 if (psk) {
407 struct virtio_vsock_sock *ptrans = psk->trans;
408
409 vvs->buf_size = ptrans->buf_size;
410 vvs->buf_size_min = ptrans->buf_size_min;
411 vvs->buf_size_max = ptrans->buf_size_max;
412 vvs->peer_buf_alloc = ptrans->peer_buf_alloc;
413 } else {
414 vvs->buf_size = VIRTIO_VSOCK_DEFAULT_BUF_SIZE;
415 vvs->buf_size_min = VIRTIO_VSOCK_DEFAULT_MIN_BUF_SIZE;
416 vvs->buf_size_max = VIRTIO_VSOCK_DEFAULT_MAX_BUF_SIZE;
417 }
418
419 vvs->buf_alloc = vvs->buf_size;
420
421 spin_lock_init(&vvs->rx_lock);
422 spin_lock_init(&vvs->tx_lock);
423 INIT_LIST_HEAD(&vvs->rx_queue);
424
425 return 0;
426}
427EXPORT_SYMBOL_GPL(virtio_transport_do_socket_init);
428
429u64 virtio_transport_get_buffer_size(struct vsock_sock *vsk)
430{
431 struct virtio_vsock_sock *vvs = vsk->trans;
432
433 return vvs->buf_size;
434}
435EXPORT_SYMBOL_GPL(virtio_transport_get_buffer_size);
436
437u64 virtio_transport_get_min_buffer_size(struct vsock_sock *vsk)
438{
439 struct virtio_vsock_sock *vvs = vsk->trans;
440
441 return vvs->buf_size_min;
442}
443EXPORT_SYMBOL_GPL(virtio_transport_get_min_buffer_size);
444
445u64 virtio_transport_get_max_buffer_size(struct vsock_sock *vsk)
446{
447 struct virtio_vsock_sock *vvs = vsk->trans;
448
449 return vvs->buf_size_max;
450}
451EXPORT_SYMBOL_GPL(virtio_transport_get_max_buffer_size);
452
453void virtio_transport_set_buffer_size(struct vsock_sock *vsk, u64 val)
454{
455 struct virtio_vsock_sock *vvs = vsk->trans;
456
457 if (val > VIRTIO_VSOCK_MAX_BUF_SIZE)
458 val = VIRTIO_VSOCK_MAX_BUF_SIZE;
459 if (val < vvs->buf_size_min)
460 vvs->buf_size_min = val;
461 if (val > vvs->buf_size_max)
462 vvs->buf_size_max = val;
463 vvs->buf_size = val;
464 vvs->buf_alloc = val;
ec3359b6
SG		 465
466 virtio_transport_send_credit_update(vsk, VIRTIO_VSOCK_TYPE_STREAM,
467 NULL);
06a8fc78
AH		 468}
469EXPORT_SYMBOL_GPL(virtio_transport_set_buffer_size);
470
471void virtio_transport_set_min_buffer_size(struct vsock_sock *vsk, u64 val)
472{
473 struct virtio_vsock_sock *vvs = vsk->trans;
474
475 if (val > VIRTIO_VSOCK_MAX_BUF_SIZE)
476 val = VIRTIO_VSOCK_MAX_BUF_SIZE;
477 if (val > vvs->buf_size)
478 vvs->buf_size = val;
479 vvs->buf_size_min = val;
480}
481EXPORT_SYMBOL_GPL(virtio_transport_set_min_buffer_size);
482
483void virtio_transport_set_max_buffer_size(struct vsock_sock *vsk, u64 val)
484{
485 struct virtio_vsock_sock *vvs = vsk->trans;
486
487 if (val > VIRTIO_VSOCK_MAX_BUF_SIZE)
488 val = VIRTIO_VSOCK_MAX_BUF_SIZE;
489 if (val < vvs->buf_size)
490 vvs->buf_size = val;
491 vvs->buf_size_max = val;
492}
493EXPORT_SYMBOL_GPL(virtio_transport_set_max_buffer_size);
494
495int
496virtio_transport_notify_poll_in(struct vsock_sock *vsk,
497 size_t target,
498 bool *data_ready_now)
499{
500 if (vsock_stream_has_data(vsk))
501 *data_ready_now = true;
502 else
503 *data_ready_now = false;
504
505 return 0;
506}
507EXPORT_SYMBOL_GPL(virtio_transport_notify_poll_in);
508
509int
510virtio_transport_notify_poll_out(struct vsock_sock *vsk,
511 size_t target,
512 bool *space_avail_now)
513{
514 s64 free_space;
515
516 free_space = vsock_stream_has_space(vsk);
517 if (free_space > 0)
518 *space_avail_now = true;
519 else if (free_space == 0)
520 *space_avail_now = false;
521
522 return 0;
523}
524EXPORT_SYMBOL_GPL(virtio_transport_notify_poll_out);
525
526int virtio_transport_notify_recv_init(struct vsock_sock *vsk,
527 size_t target, struct vsock_transport_recv_notify_data *data)
528{
529 return 0;
530}
531EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_init);
532
533int virtio_transport_notify_recv_pre_block(struct vsock_sock *vsk,
534 size_t target, struct vsock_transport_recv_notify_data *data)
535{
536 return 0;
537}
538EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_pre_block);
539
540int virtio_transport_notify_recv_pre_dequeue(struct vsock_sock *vsk,
541 size_t target, struct vsock_transport_recv_notify_data *data)
542{
543 return 0;
544}
545EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_pre_dequeue);
546
547int virtio_transport_notify_recv_post_dequeue(struct vsock_sock *vsk,
548 size_t target, ssize_t copied, bool data_read,
549 struct vsock_transport_recv_notify_data *data)
550{
551 return 0;
552}
553EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_post_dequeue);
554
555int virtio_transport_notify_send_init(struct vsock_sock *vsk,
556 struct vsock_transport_send_notify_data *data)
557{
558 return 0;
559}
560EXPORT_SYMBOL_GPL(virtio_transport_notify_send_init);
561
562int virtio_transport_notify_send_pre_block(struct vsock_sock *vsk,
563 struct vsock_transport_send_notify_data *data)
564{
565 return 0;
566}
567EXPORT_SYMBOL_GPL(virtio_transport_notify_send_pre_block);
568
569int virtio_transport_notify_send_pre_enqueue(struct vsock_sock *vsk,
570 struct vsock_transport_send_notify_data *data)
571{
572 return 0;
573}
574EXPORT_SYMBOL_GPL(virtio_transport_notify_send_pre_enqueue);
575
576int virtio_transport_notify_send_post_enqueue(struct vsock_sock *vsk,
577 ssize_t written, struct vsock_transport_send_notify_data *data)
578{
579 return 0;
580}
581EXPORT_SYMBOL_GPL(virtio_transport_notify_send_post_enqueue);
582
583u64 virtio_transport_stream_rcvhiwat(struct vsock_sock *vsk)
584{
585 struct virtio_vsock_sock *vvs = vsk->trans;
586
587 return vvs->buf_size;
588}
589EXPORT_SYMBOL_GPL(virtio_transport_stream_rcvhiwat);
590
591bool virtio_transport_stream_is_active(struct vsock_sock *vsk)
592{
593 return true;
594}
595EXPORT_SYMBOL_GPL(virtio_transport_stream_is_active);
596
597bool virtio_transport_stream_allow(u32 cid, u32 port)
598{
599 return true;
600}
601EXPORT_SYMBOL_GPL(virtio_transport_stream_allow);
602
603int virtio_transport_dgram_bind(struct vsock_sock *vsk,
604 struct sockaddr_vm *addr)
605{
606 return -EOPNOTSUPP;
607}
608EXPORT_SYMBOL_GPL(virtio_transport_dgram_bind);
609
610bool virtio_transport_dgram_allow(u32 cid, u32 port)
611{
612 return false;
613}
614EXPORT_SYMBOL_GPL(virtio_transport_dgram_allow);
615
616int virtio_transport_connect(struct vsock_sock *vsk)
617{
618 struct virtio_vsock_pkt_info info = {
619 .op = VIRTIO_VSOCK_OP_REQUEST,
620 .type = VIRTIO_VSOCK_TYPE_STREAM,
36d277ba 621 .vsk = vsk,
06a8fc78
AH		 622 };
623
624 return virtio_transport_send_pkt_info(vsk, &info);
625}
626EXPORT_SYMBOL_GPL(virtio_transport_connect);
627
628int virtio_transport_shutdown(struct vsock_sock *vsk, int mode)
629{
630 struct virtio_vsock_pkt_info info = {
631 .op = VIRTIO_VSOCK_OP_SHUTDOWN,
632 .type = VIRTIO_VSOCK_TYPE_STREAM,
633 .flags = (mode & RCV_SHUTDOWN ?
634 VIRTIO_VSOCK_SHUTDOWN_RCV : 0) |
635 (mode & SEND_SHUTDOWN ?
636 VIRTIO_VSOCK_SHUTDOWN_SEND : 0),
36d277ba 637 .vsk = vsk,
06a8fc78
AH		 638 };
639
640 return virtio_transport_send_pkt_info(vsk, &info);
641}
642EXPORT_SYMBOL_GPL(virtio_transport_shutdown);
643
644int
645virtio_transport_dgram_enqueue(struct vsock_sock *vsk,
646 struct sockaddr_vm *remote_addr,
647 struct msghdr *msg,
648 size_t dgram_len)
649{
650 return -EOPNOTSUPP;
651}
652EXPORT_SYMBOL_GPL(virtio_transport_dgram_enqueue);
653
654ssize_t
655virtio_transport_stream_enqueue(struct vsock_sock *vsk,
656 struct msghdr *msg,
657 size_t len)
658{
659 struct virtio_vsock_pkt_info info = {
660 .op = VIRTIO_VSOCK_OP_RW,
661 .type = VIRTIO_VSOCK_TYPE_STREAM,
662 .msg = msg,
663 .pkt_len = len,
36d277ba 664 .vsk = vsk,
06a8fc78
AH		 665 };
666
667 return virtio_transport_send_pkt_info(vsk, &info);
668}
669EXPORT_SYMBOL_GPL(virtio_transport_stream_enqueue);
670
671void virtio_transport_destruct(struct vsock_sock *vsk)
672{
673 struct virtio_vsock_sock *vvs = vsk->trans;
674
675 kfree(vvs);
676}
677EXPORT_SYMBOL_GPL(virtio_transport_destruct);
678
679static int virtio_transport_reset(struct vsock_sock *vsk,
680 struct virtio_vsock_pkt *pkt)
681{
682 struct virtio_vsock_pkt_info info = {
683 .op = VIRTIO_VSOCK_OP_RST,
684 .type = VIRTIO_VSOCK_TYPE_STREAM,
685 .reply = !!pkt,
36d277ba 686 .vsk = vsk,
06a8fc78
AH		 687 };
688
689 /* Send RST only if the original pkt is not a RST pkt */
690 if (pkt && le16_to_cpu(pkt->hdr.op) == VIRTIO_VSOCK_OP_RST)
691 return 0;
692
693 return virtio_transport_send_pkt_info(vsk, &info);
694}
695
696/* Normally packets are associated with a socket. There may be no socket if an
697 * attempt was made to connect to a socket that does not exist.
698 */
699static int virtio_transport_reset_no_sock(struct virtio_vsock_pkt *pkt)
700{
4c404ce2
AL		 701 const struct virtio_transport *t;
702 struct virtio_vsock_pkt *reply;
06a8fc78
AH		 703 struct virtio_vsock_pkt_info info = {
704 .op = VIRTIO_VSOCK_OP_RST,
705 .type = le16_to_cpu(pkt->hdr.type),
706 .reply = true,
707 };
708
709 /* Send RST only if the original pkt is not a RST pkt */
710 if (le16_to_cpu(pkt->hdr.op) == VIRTIO_VSOCK_OP_RST)
711 return 0;
712
4c404ce2
AL		 713 reply = virtio_transport_alloc_pkt(&info, 0,
714 le64_to_cpu(pkt->hdr.dst_cid),
715 le32_to_cpu(pkt->hdr.dst_port),
716 le64_to_cpu(pkt->hdr.src_cid),
717 le32_to_cpu(pkt->hdr.src_port));
718 if (!reply)
06a8fc78
AH		 719 return -ENOMEM;
720
4c404ce2
AL		 721 t = virtio_transport_get_ops();
722 if (!t) {
723 virtio_transport_free_pkt(reply);
724 return -ENOTCONN;
725 }
726
727 return t->send_pkt(reply);
06a8fc78
AH		 728}
729
730static void virtio_transport_wait_close(struct sock *sk, long timeout)
731{
732 if (timeout) {
d9dc8b0f
WC		 733 DEFINE_WAIT_FUNC(wait, woken_wake_function);
734
735 add_wait_queue(sk_sleep(sk), &wait);
06a8fc78
AH		 736
737 do {
06a8fc78 738 if (sk_wait_event(sk, &timeout,
d9dc8b0f 739 sock_flag(sk, SOCK_DONE), &wait))
06a8fc78
AH		 740 break;
741 } while (!signal_pending(current) && timeout);
742
d9dc8b0f 743 remove_wait_queue(sk_sleep(sk), &wait);
06a8fc78
AH		 744 }
745}
746
747static void virtio_transport_do_close(struct vsock_sock *vsk,
748 bool cancel_timeout)
749{
750 struct sock *sk = sk_vsock(vsk);
751
752 sock_set_flag(sk, SOCK_DONE);
753 vsk->peer_shutdown = SHUTDOWN_MASK;
754 if (vsock_stream_has_data(vsk) <= 0)
3b4477d2 755 sk->sk_state = TCP_CLOSING;
06a8fc78
AH		 756 sk->sk_state_change(sk);
757
758 if (vsk->close_work_scheduled &&
759 (!cancel_timeout || cancel_delayed_work(&vsk->close_work))) {
760 vsk->close_work_scheduled = false;
761
762 vsock_remove_sock(vsk);
763
764 /* Release refcnt obtained when we scheduled the timeout */
765 sock_put(sk);
766 }
767}
768
769static void virtio_transport_close_timeout(struct work_struct *work)
770{
771 struct vsock_sock *vsk =
772 container_of(work, struct vsock_sock, close_work.work);
773 struct sock *sk = sk_vsock(vsk);
774
775 sock_hold(sk);
776 lock_sock(sk);
777
778 if (!sock_flag(sk, SOCK_DONE)) {
779 (void)virtio_transport_reset(vsk, NULL);
780
781 virtio_transport_do_close(vsk, false);
782 }
783
784 vsk->close_work_scheduled = false;
785
786 release_sock(sk);
787 sock_put(sk);
788}
789
790/* User context, vsk->sk is locked */
791static bool virtio_transport_close(struct vsock_sock *vsk)
792{
793 struct sock *sk = &vsk->sk;
794
3b4477d2
SH		 795 if (!(sk->sk_state == TCP_ESTABLISHED ||
796 sk->sk_state == TCP_CLOSING))
06a8fc78
AH		 797 return true;
798
799 /* Already received SHUTDOWN from peer, reply with RST */
800 if ((vsk->peer_shutdown & SHUTDOWN_MASK) == SHUTDOWN_MASK) {
801 (void)virtio_transport_reset(vsk, NULL);
802 return true;
803 }
804
805 if ((sk->sk_shutdown & SHUTDOWN_MASK) != SHUTDOWN_MASK)
806 (void)virtio_transport_shutdown(vsk, SHUTDOWN_MASK);
807
808 if (sock_flag(sk, SOCK_LINGER) && !(current->flags & PF_EXITING))
809 virtio_transport_wait_close(sk, sk->sk_lingertime);
810
811 if (sock_flag(sk, SOCK_DONE)) {
812 return true;
813 }
814
815 sock_hold(sk);
816 INIT_DELAYED_WORK(&vsk->close_work,
817 virtio_transport_close_timeout);
818 vsk->close_work_scheduled = true;
819 schedule_delayed_work(&vsk->close_work, VSOCK_CLOSE_TIMEOUT);
820 return false;
821}
822
823void virtio_transport_release(struct vsock_sock *vsk)
824{
ac03046e
SG		 825 struct virtio_vsock_sock *vvs = vsk->trans;
826 struct virtio_vsock_pkt *pkt, *tmp;
06a8fc78
AH		 827 struct sock *sk = &vsk->sk;
828 bool remove_sock = true;
829
0d9138ff 830 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
06a8fc78
AH		 831 if (sk->sk_type == SOCK_STREAM)
832 remove_sock = virtio_transport_close(vsk);
ac03046e
SG		 833
834 list_for_each_entry_safe(pkt, tmp, &vvs->rx_queue, list) {
835 list_del(&pkt->list);
836 virtio_transport_free_pkt(pkt);
837 }
06a8fc78
AH		 838 release_sock(sk);
839
840 if (remove_sock)
841 vsock_remove_sock(vsk);
842}
843EXPORT_SYMBOL_GPL(virtio_transport_release);
844
845static int
846virtio_transport_recv_connecting(struct sock *sk,
847 struct virtio_vsock_pkt *pkt)
848{
849 struct vsock_sock *vsk = vsock_sk(sk);
850 int err;
851 int skerr;
852
853 switch (le16_to_cpu(pkt->hdr.op)) {
854 case VIRTIO_VSOCK_OP_RESPONSE:
3b4477d2 855 sk->sk_state = TCP_ESTABLISHED;
06a8fc78
AH		 856 sk->sk_socket->state = SS_CONNECTED;
857 vsock_insert_connected(vsk);
858 sk->sk_state_change(sk);
859 break;
860 case VIRTIO_VSOCK_OP_INVALID:
861 break;
862 case VIRTIO_VSOCK_OP_RST:
863 skerr = ECONNRESET;
864 err = 0;
865 goto destroy;
866 default:
867 skerr = EPROTO;
868 err = -EINVAL;
869 goto destroy;
870 }
871 return 0;
872
873destroy:
874 virtio_transport_reset(vsk, pkt);
3b4477d2 875 sk->sk_state = TCP_CLOSE;
06a8fc78
AH		 876 sk->sk_err = skerr;
877 sk->sk_error_report(sk);
878 return err;
879}
880
473c7391
SG		 881static void
882virtio_transport_recv_enqueue(struct vsock_sock *vsk,
883 struct virtio_vsock_pkt *pkt)
884{
885 struct virtio_vsock_sock *vvs = vsk->trans;
ae6fcfbf 886 bool can_enqueue, free_pkt = false;
473c7391
SG		 887
888 pkt->len = le32_to_cpu(pkt->hdr.len);
889 pkt->off = 0;
890
891 spin_lock_bh(&vvs->rx_lock);
892
ae6fcfbf
SG		 893 can_enqueue = virtio_transport_inc_rx_pkt(vvs, pkt);
894 if (!can_enqueue) {
895 free_pkt = true;
896 goto out;
897 }
473c7391
SG		 898
899 /* Try to copy small packets into the buffer of last packet queued,
900 * to avoid wasting memory queueing the entire buffer with a small
901 * payload.
902 */
903 if (pkt->len <= GOOD_COPY_LEN && !list_empty(&vvs->rx_queue)) {
904 struct virtio_vsock_pkt *last_pkt;
905
906 last_pkt = list_last_entry(&vvs->rx_queue,
907 struct virtio_vsock_pkt, list);
908
909 /* If there is space in the last packet queued, we copy the
910 * new packet in its buffer.
911 */
912 if (pkt->len <= last_pkt->buf_len - last_pkt->len) {
913 memcpy(last_pkt->buf + last_pkt->len, pkt->buf,
914 pkt->len);
915 last_pkt->len += pkt->len;
916 free_pkt = true;
917 goto out;
918 }
919 }
920
921 list_add_tail(&pkt->list, &vvs->rx_queue);
922
923out:
924 spin_unlock_bh(&vvs->rx_lock);
925 if (free_pkt)
926 virtio_transport_free_pkt(pkt);
927}
928
06a8fc78
AH		 929static int
930virtio_transport_recv_connected(struct sock *sk,
931 struct virtio_vsock_pkt *pkt)
932{
933 struct vsock_sock *vsk = vsock_sk(sk);
06a8fc78
AH		 934 int err = 0;
935
936 switch (le16_to_cpu(pkt->hdr.op)) {
937 case VIRTIO_VSOCK_OP_RW:
473c7391 938 virtio_transport_recv_enqueue(vsk, pkt);
06a8fc78
AH		 939 sk->sk_data_ready(sk);
940 return err;
941 case VIRTIO_VSOCK_OP_CREDIT_UPDATE:
942 sk->sk_write_space(sk);
943 break;
944 case VIRTIO_VSOCK_OP_SHUTDOWN:
945 if (le32_to_cpu(pkt->hdr.flags) & VIRTIO_VSOCK_SHUTDOWN_RCV)
946 vsk->peer_shutdown |= RCV_SHUTDOWN;
947 if (le32_to_cpu(pkt->hdr.flags) & VIRTIO_VSOCK_SHUTDOWN_SEND)
948 vsk->peer_shutdown |= SEND_SHUTDOWN;
949 if (vsk->peer_shutdown == SHUTDOWN_MASK &&
42f5cda5
SB		 950 vsock_stream_has_data(vsk) <= 0) {
951 sock_set_flag(sk, SOCK_DONE);
3b4477d2 952 sk->sk_state = TCP_CLOSING;
42f5cda5 953 }
06a8fc78
AH		 954 if (le32_to_cpu(pkt->hdr.flags))
955 sk->sk_state_change(sk);
956 break;
957 case VIRTIO_VSOCK_OP_RST:
958 virtio_transport_do_close(vsk, true);
959 break;
960 default:
961 err = -EINVAL;
962 break;
963 }
964
965 virtio_transport_free_pkt(pkt);
966 return err;
967}
968
969static void
970virtio_transport_recv_disconnecting(struct sock *sk,
971 struct virtio_vsock_pkt *pkt)
972{
973 struct vsock_sock *vsk = vsock_sk(sk);
974
975 if (le16_to_cpu(pkt->hdr.op) == VIRTIO_VSOCK_OP_RST)
976 virtio_transport_do_close(vsk, true);
977}
978
979static int
980virtio_transport_send_response(struct vsock_sock *vsk,
981 struct virtio_vsock_pkt *pkt)
982{
983 struct virtio_vsock_pkt_info info = {
984 .op = VIRTIO_VSOCK_OP_RESPONSE,
985 .type = VIRTIO_VSOCK_TYPE_STREAM,
f83f12d6 986 .remote_cid = le64_to_cpu(pkt->hdr.src_cid),
06a8fc78
AH		 987 .remote_port = le32_to_cpu(pkt->hdr.src_port),
988 .reply = true,
36d277ba 989 .vsk = vsk,
06a8fc78
AH		 990 };
991
992 return virtio_transport_send_pkt_info(vsk, &info);
993}
994
995/* Handle server socket */
996static int
997virtio_transport_recv_listen(struct sock *sk, struct virtio_vsock_pkt *pkt)
998{
999 struct vsock_sock *vsk = vsock_sk(sk);
1000 struct vsock_sock *vchild;
1001 struct sock *child;
1002
1003 if (le16_to_cpu(pkt->hdr.op) != VIRTIO_VSOCK_OP_REQUEST) {
1004 virtio_transport_reset(vsk, pkt);
1005 return -EINVAL;
1006 }
1007
1008 if (sk_acceptq_is_full(sk)) {
1009 virtio_transport_reset(vsk, pkt);
1010 return -ENOMEM;
1011 }
1012
1013 child = __vsock_create(sock_net(sk), NULL, sk, GFP_KERNEL,
1014 sk->sk_type, 0);
1015 if (!child) {
1016 virtio_transport_reset(vsk, pkt);
1017 return -ENOMEM;
1018 }
1019
1020 sk->sk_ack_backlog++;
1021
1022 lock_sock_nested(child, SINGLE_DEPTH_NESTING);
1023
3b4477d2 1024 child->sk_state = TCP_ESTABLISHED;
06a8fc78
AH		 1025
1026 vchild = vsock_sk(child);
f83f12d6 1027 vsock_addr_init(&vchild->local_addr, le64_to_cpu(pkt->hdr.dst_cid),
06a8fc78 1028 le32_to_cpu(pkt->hdr.dst_port));
f83f12d6 1029 vsock_addr_init(&vchild->remote_addr, le64_to_cpu(pkt->hdr.src_cid),
06a8fc78
AH		 1030 le32_to_cpu(pkt->hdr.src_port));
1031
1032 vsock_insert_connected(vchild);
1033 vsock_enqueue_accept(sk, child);
1034 virtio_transport_send_response(vchild, pkt);
1035
1036 release_sock(child);
1037
1038 sk->sk_data_ready(sk);
1039 return 0;
1040}
1041
1042static bool virtio_transport_space_update(struct sock *sk,
1043 struct virtio_vsock_pkt *pkt)
1044{
1045 struct vsock_sock *vsk = vsock_sk(sk);
1046 struct virtio_vsock_sock *vvs = vsk->trans;
1047 bool space_available;
1048
1049 /* buf_alloc and fwd_cnt is always included in the hdr */
1050 spin_lock_bh(&vvs->tx_lock);
1051 vvs->peer_buf_alloc = le32_to_cpu(pkt->hdr.buf_alloc);
1052 vvs->peer_fwd_cnt = le32_to_cpu(pkt->hdr.fwd_cnt);
1053 space_available = virtio_transport_has_space(vsk);
1054 spin_unlock_bh(&vvs->tx_lock);
1055 return space_available;
1056}
1057
1058/* We are under the virtio-vsock's vsock->rx_lock or vhost-vsock's vq->mutex
1059 * lock.
1060 */
1061void virtio_transport_recv_pkt(struct virtio_vsock_pkt *pkt)
1062{
1063 struct sockaddr_vm src, dst;
1064 struct vsock_sock *vsk;
1065 struct sock *sk;
1066 bool space_available;
1067
f83f12d6 1068 vsock_addr_init(&src, le64_to_cpu(pkt->hdr.src_cid),
06a8fc78 1069 le32_to_cpu(pkt->hdr.src_port));
f83f12d6 1070 vsock_addr_init(&dst, le64_to_cpu(pkt->hdr.dst_cid),
06a8fc78
AH		 1071 le32_to_cpu(pkt->hdr.dst_port));
1072
1073 trace_virtio_transport_recv_pkt(src.svm_cid, src.svm_port,
1074 dst.svm_cid, dst.svm_port,
1075 le32_to_cpu(pkt->hdr.len),
1076 le16_to_cpu(pkt->hdr.type),
1077 le16_to_cpu(pkt->hdr.op),
1078 le32_to_cpu(pkt->hdr.flags),
1079 le32_to_cpu(pkt->hdr.buf_alloc),
1080 le32_to_cpu(pkt->hdr.fwd_cnt));
1081
1082 if (le16_to_cpu(pkt->hdr.type) != VIRTIO_VSOCK_TYPE_STREAM) {
1083 (void)virtio_transport_reset_no_sock(pkt);
1084 goto free_pkt;
1085 }
1086
1087 /* The socket must be in connected or bound table
1088 * otherwise send reset back
1089 */
1090 sk = vsock_find_connected_socket(&src, &dst);
1091 if (!sk) {
1092 sk = vsock_find_bound_socket(&dst);
1093 if (!sk) {
1094 (void)virtio_transport_reset_no_sock(pkt);
1095 goto free_pkt;
1096 }
1097 }
1098
1099 vsk = vsock_sk(sk);
1100
1101 space_available = virtio_transport_space_update(sk, pkt);
1102
1103 lock_sock(sk);
1104
1105 /* Update CID in case it has changed after a transport reset event */
1106 vsk->local_addr.svm_cid = dst.svm_cid;
1107
1108 if (space_available)
1109 sk->sk_write_space(sk);
1110
1111 switch (sk->sk_state) {
3b4477d2 1112 case TCP_LISTEN:
06a8fc78
AH		 1113 virtio_transport_recv_listen(sk, pkt);
1114 virtio_transport_free_pkt(pkt);
1115 break;
3b4477d2 1116 case TCP_SYN_SENT:
06a8fc78
AH		 1117 virtio_transport_recv_connecting(sk, pkt);
1118 virtio_transport_free_pkt(pkt);
1119 break;
3b4477d2 1120 case TCP_ESTABLISHED:
06a8fc78
AH		 1121 virtio_transport_recv_connected(sk, pkt);
1122 break;
3b4477d2 1123 case TCP_CLOSING:
06a8fc78
AH		 1124 virtio_transport_recv_disconnecting(sk, pkt);
1125 virtio_transport_free_pkt(pkt);
1126 break;
1127 default:
1128 virtio_transport_free_pkt(pkt);
1129 break;
1130 }
1131 release_sock(sk);
1132
1133 /* Release refcnt obtained when we fetched this socket out of the
1134 * bound or connected list.
1135 */
1136 sock_put(sk);
1137 return;
1138
1139free_pkt:
1140 virtio_transport_free_pkt(pkt);
1141}
1142EXPORT_SYMBOL_GPL(virtio_transport_recv_pkt);
1143
1144void virtio_transport_free_pkt(struct virtio_vsock_pkt *pkt)
1145{
1146 kfree(pkt->buf);
1147 kfree(pkt);
1148}
1149EXPORT_SYMBOL_GPL(virtio_transport_free_pkt);
1150
1151MODULE_LICENSE("GPL v2");
1152MODULE_AUTHOR("Asias He");
1153MODULE_DESCRIPTION("common code for virtio vsock");
Linux 6.x block layer and io_uring tree(s)