?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
convert rcv_conn_list to snd_conn_list
[cor.git] / net / kcm / kcmsock.c
blobea9e73428ed9c8b7bb3441947151c41f0c099185
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * Kernel Connection Multiplexor
4 *
5 * Copyright (c) 2016 Tom Herbert <tom@herbertland.com>
6 */
7
8 #include <linux/bpf.h>
9 #include <linux/errno.h>
10 #include <linux/errqueue.h>
11 #include <linux/file.h>
12 #include <linux/in.h>
13 #include <linux/kernel.h>
14 #include <linux/module.h>
15 #include <linux/net.h>
16 #include <linux/netdevice.h>
17 #include <linux/poll.h>
18 #include <linux/rculist.h>
19 #include <linux/skbuff.h>
20 #include <linux/socket.h>
21 #include <linux/uaccess.h>
22 #include <linux/workqueue.h>
23 #include <linux/syscalls.h>
24 #include <linux/sched/signal.h>
25
26 #include <net/kcm.h>
27 #include <net/netns/generic.h>
28 #include <net/sock.h>
29 #include <uapi/linux/kcm.h>
30
31 unsigned int kcm_net_id;
32
33 static struct kmem_cache *kcm_psockp __read_mostly;
34 static struct kmem_cache *kcm_muxp __read_mostly;
35 static struct workqueue_struct *kcm_wq;
36
37 static inline struct kcm_sock *kcm_sk(const struct sock *sk)
38 {
39 return (struct kcm_sock *)sk;
40 }
41
42 static inline struct kcm_tx_msg *kcm_tx_msg(struct sk_buff *skb)
43 {
44 return (struct kcm_tx_msg *)skb->cb;
45 }
46
47 static void report_csk_error(struct sock *csk, int err)
48 {
49 csk->sk_err = EPIPE;
50 csk->sk_error_report(csk);
51 }
52
53 static void kcm_abort_tx_psock(struct kcm_psock *psock, int err,
54 bool wakeup_kcm)
55 {
56 struct sock *csk = psock->sk;
57 struct kcm_mux *mux = psock->mux;
58
59 /* Unrecoverable error in transmit */
60
61 spin_lock_bh(&mux->lock);
62
63 if (psock->tx_stopped) {
64 spin_unlock_bh(&mux->lock);
65 return;
66 }
67
68 psock->tx_stopped = 1;
69 KCM_STATS_INCR(psock->stats.tx_aborts);
70
71 if (!psock->tx_kcm) {
72 /* Take off psocks_avail list */
73 list_del(&psock->psock_avail_list);
74 } else if (wakeup_kcm) {
75 /* In this case psock is being aborted while outside of
76 * write_msgs and psock is reserved. Schedule tx_work
77 * to handle the failure there. Need to commit tx_stopped
78 * before queuing work.
79 */
80 smp_mb();
81
82 queue_work(kcm_wq, &psock->tx_kcm->tx_work);
83 }
84
85 spin_unlock_bh(&mux->lock);
86
87 /* Report error on lower socket */
88 report_csk_error(csk, err);
89 }
90
91 /* RX mux lock held. */
92 static void kcm_update_rx_mux_stats(struct kcm_mux *mux,
93 struct kcm_psock *psock)
94 {
95 STRP_STATS_ADD(mux->stats.rx_bytes,
96 psock->strp.stats.bytes -
97 psock->saved_rx_bytes);
98 mux->stats.rx_msgs +=
99 psock->strp.stats.msgs - psock->saved_rx_msgs;
100 psock->saved_rx_msgs = psock->strp.stats.msgs;
101 psock->saved_rx_bytes = psock->strp.stats.bytes;
102 }
103
104 static void kcm_update_tx_mux_stats(struct kcm_mux *mux,
105 struct kcm_psock *psock)
106 {
107 KCM_STATS_ADD(mux->stats.tx_bytes,
108 psock->stats.tx_bytes - psock->saved_tx_bytes);
109 mux->stats.tx_msgs +=
110 psock->stats.tx_msgs - psock->saved_tx_msgs;
111 psock->saved_tx_msgs = psock->stats.tx_msgs;
112 psock->saved_tx_bytes = psock->stats.tx_bytes;
113 }
114
115 static int kcm_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
116
117 /* KCM is ready to receive messages on its queue-- either the KCM is new or
118 * has become unblocked after being blocked on full socket buffer. Queue any
119 * pending ready messages on a psock. RX mux lock held.
120 */
121 static void kcm_rcv_ready(struct kcm_sock *kcm)
122 {
123 struct kcm_mux *mux = kcm->mux;
124 struct kcm_psock *psock;
125 struct sk_buff *skb;
126
127 if (unlikely(kcm->rx_wait || kcm->rx_psock || kcm->rx_disabled))
128 return;
129
130 while (unlikely((skb = __skb_dequeue(&mux->rx_hold_queue)))) {
131 if (kcm_queue_rcv_skb(&kcm->sk, skb)) {
132 /* Assuming buffer limit has been reached */
133 skb_queue_head(&mux->rx_hold_queue, skb);
134 WARN_ON(!sk_rmem_alloc_get(&kcm->sk));
135 return;
136 }
137 }
138
139 while (!list_empty(&mux->psocks_ready)) {
140 psock = list_first_entry(&mux->psocks_ready, struct kcm_psock,
141 psock_ready_list);
142
143 if (kcm_queue_rcv_skb(&kcm->sk, psock->ready_rx_msg)) {
144 /* Assuming buffer limit has been reached */
145 WARN_ON(!sk_rmem_alloc_get(&kcm->sk));
146 return;
147 }
148
149 /* Consumed the ready message on the psock. Schedule rx_work to
150 * get more messages.
151 */
152 list_del(&psock->psock_ready_list);
153 psock->ready_rx_msg = NULL;
154 /* Commit clearing of ready_rx_msg for queuing work */
155 smp_mb();
156
157 strp_unpause(&psock->strp);
158 strp_check_rcv(&psock->strp);
159 }
160
161 /* Buffer limit is okay now, add to ready list */
162 list_add_tail(&kcm->wait_rx_list,
163 &kcm->mux->kcm_rx_waiters);
164 kcm->rx_wait = true;
165 }
166
167 static void kcm_rfree(struct sk_buff *skb)
168 {
169 struct sock *sk = skb->sk;
170 struct kcm_sock *kcm = kcm_sk(sk);
171 struct kcm_mux *mux = kcm->mux;
172 unsigned int len = skb->truesize;
173
174 sk_mem_uncharge(sk, len);
175 atomic_sub(len, &sk->sk_rmem_alloc);
176
177 /* For reading rx_wait and rx_psock without holding lock */
178 smp_mb__after_atomic();
179
180 if (!kcm->rx_wait && !kcm->rx_psock &&
181 sk_rmem_alloc_get(sk) < sk->sk_rcvlowat) {
182 spin_lock_bh(&mux->rx_lock);
183 kcm_rcv_ready(kcm);
184 spin_unlock_bh(&mux->rx_lock);
185 }
186 }
187
188 static int kcm_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
189 {
190 struct sk_buff_head *list = &sk->sk_receive_queue;
191
192 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
193 return -ENOMEM;
194
195 if (!sk_rmem_schedule(sk, skb, skb->truesize))
196 return -ENOBUFS;
197
198 skb->dev = NULL;
199
200 skb_orphan(skb);
201 skb->sk = sk;
202 skb->destructor = kcm_rfree;
203 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
204 sk_mem_charge(sk, skb->truesize);
205
206 skb_queue_tail(list, skb);
207
208 if (!sock_flag(sk, SOCK_DEAD))
209 sk->sk_data_ready(sk);
210
211 return 0;
212 }
213
214 /* Requeue received messages for a kcm socket to other kcm sockets. This is
215 * called with a kcm socket is receive disabled.
216 * RX mux lock held.
217 */
218 static void requeue_rx_msgs(struct kcm_mux *mux, struct sk_buff_head *head)
219 {
220 struct sk_buff *skb;
221 struct kcm_sock *kcm;
222
223 while ((skb = __skb_dequeue(head))) {
224 /* Reset destructor to avoid calling kcm_rcv_ready */
225 skb->destructor = sock_rfree;
226 skb_orphan(skb);
227 try_again:
228 if (list_empty(&mux->kcm_rx_waiters)) {
229 skb_queue_tail(&mux->rx_hold_queue, skb);
230 continue;
231 }
232
233 kcm = list_first_entry(&mux->kcm_rx_waiters,
234 struct kcm_sock, wait_rx_list);
235
236 if (kcm_queue_rcv_skb(&kcm->sk, skb)) {
237 /* Should mean socket buffer full */
238 list_del(&kcm->wait_rx_list);
239 kcm->rx_wait = false;
240
241 /* Commit rx_wait to read in kcm_free */
242 smp_wmb();
243
244 goto try_again;
245 }
246 }
247 }
248
249 /* Lower sock lock held */
250 static struct kcm_sock *reserve_rx_kcm(struct kcm_psock *psock,
251 struct sk_buff *head)
252 {
253 struct kcm_mux *mux = psock->mux;
254 struct kcm_sock *kcm;
255
256 WARN_ON(psock->ready_rx_msg);
257
258 if (psock->rx_kcm)
259 return psock->rx_kcm;
260
261 spin_lock_bh(&mux->rx_lock);
262
263 if (psock->rx_kcm) {
264 spin_unlock_bh(&mux->rx_lock);
265 return psock->rx_kcm;
266 }
267
268 kcm_update_rx_mux_stats(mux, psock);
269
270 if (list_empty(&mux->kcm_rx_waiters)) {
271 psock->ready_rx_msg = head;
272 strp_pause(&psock->strp);
273 list_add_tail(&psock->psock_ready_list,
274 &mux->psocks_ready);
275 spin_unlock_bh(&mux->rx_lock);
276 return NULL;
277 }
278
279 kcm = list_first_entry(&mux->kcm_rx_waiters,
280 struct kcm_sock, wait_rx_list);
281 list_del(&kcm->wait_rx_list);
282 kcm->rx_wait = false;
283
284 psock->rx_kcm = kcm;
285 kcm->rx_psock = psock;
286
287 spin_unlock_bh(&mux->rx_lock);
288
289 return kcm;
290 }
291
292 static void kcm_done(struct kcm_sock *kcm);
293
294 static void kcm_done_work(struct work_struct *w)
295 {
296 kcm_done(container_of(w, struct kcm_sock, done_work));
297 }
298
299 /* Lower sock held */
300 static void unreserve_rx_kcm(struct kcm_psock *psock,
301 bool rcv_ready)
302 {
303 struct kcm_sock *kcm = psock->rx_kcm;
304 struct kcm_mux *mux = psock->mux;
305
306 if (!kcm)
307 return;
308
309 spin_lock_bh(&mux->rx_lock);
310
311 psock->rx_kcm = NULL;
312 kcm->rx_psock = NULL;
313
314 /* Commit kcm->rx_psock before sk_rmem_alloc_get to sync with
315 * kcm_rfree
316 */
317 smp_mb();
318
319 if (unlikely(kcm->done)) {
320 spin_unlock_bh(&mux->rx_lock);
321
322 /* Need to run kcm_done in a task since we need to qcquire
323 * callback locks which may already be held here.
324 */
325 INIT_WORK(&kcm->done_work, kcm_done_work);
326 schedule_work(&kcm->done_work);
327 return;
328 }
329
330 if (unlikely(kcm->rx_disabled)) {
331 requeue_rx_msgs(mux, &kcm->sk.sk_receive_queue);
332 } else if (rcv_ready || unlikely(!sk_rmem_alloc_get(&kcm->sk))) {
333 /* Check for degenerative race with rx_wait that all
334 * data was dequeued (accounted for in kcm_rfree).
335 */
336 kcm_rcv_ready(kcm);
337 }
338 spin_unlock_bh(&mux->rx_lock);
339 }
340
341 /* Lower sock lock held */
342 static void psock_data_ready(struct sock *sk)
343 {
344 struct kcm_psock *psock;
345
346 read_lock_bh(&sk->sk_callback_lock);
347
348 psock = (struct kcm_psock *)sk->sk_user_data;
349 if (likely(psock))
350 strp_data_ready(&psock->strp);
351
352 read_unlock_bh(&sk->sk_callback_lock);
353 }
354
355 /* Called with lower sock held */
356 static void kcm_rcv_strparser(struct strparser *strp, struct sk_buff *skb)
357 {
358 struct kcm_psock *psock = container_of(strp, struct kcm_psock, strp);
359 struct kcm_sock *kcm;
360
361 try_queue:
362 kcm = reserve_rx_kcm(psock, skb);
363 if (!kcm) {
364 /* Unable to reserve a KCM, message is held in psock and strp
365 * is paused.
366 */
367 return;
368 }
369
370 if (kcm_queue_rcv_skb(&kcm->sk, skb)) {
371 /* Should mean socket buffer full */
372 unreserve_rx_kcm(psock, false);
373 goto try_queue;
374 }
375 }
376
377 static int kcm_parse_func_strparser(struct strparser *strp, struct sk_buff *skb)
378 {
379 struct kcm_psock *psock = container_of(strp, struct kcm_psock, strp);
380 struct bpf_prog *prog = psock->bpf_prog;
381 int res;
382
383 preempt_disable();
384 res = BPF_PROG_RUN(prog, skb);
385 preempt_enable();
386 return res;
387 }
388
389 static int kcm_read_sock_done(struct strparser *strp, int err)
390 {
391 struct kcm_psock *psock = container_of(strp, struct kcm_psock, strp);
392
393 unreserve_rx_kcm(psock, true);
394
395 return err;
396 }
397
398 static void psock_state_change(struct sock *sk)
399 {
400 /* TCP only does a EPOLLIN for a half close. Do a EPOLLHUP here
401 * since application will normally not poll with EPOLLIN
402 * on the TCP sockets.
403 */
404
405 report_csk_error(sk, EPIPE);
406 }
407
408 static void psock_write_space(struct sock *sk)
409 {
410 struct kcm_psock *psock;
411 struct kcm_mux *mux;
412 struct kcm_sock *kcm;
413
414 read_lock_bh(&sk->sk_callback_lock);
415
416 psock = (struct kcm_psock *)sk->sk_user_data;
417 if (unlikely(!psock))
418 goto out;
419 mux = psock->mux;
420
421 spin_lock_bh(&mux->lock);
422
423 /* Check if the socket is reserved so someone is waiting for sending. */
424 kcm = psock->tx_kcm;
425 if (kcm && !unlikely(kcm->tx_stopped))
426 queue_work(kcm_wq, &kcm->tx_work);
427
428 spin_unlock_bh(&mux->lock);
429 out:
430 read_unlock_bh(&sk->sk_callback_lock);
431 }
432
433 static void unreserve_psock(struct kcm_sock *kcm);
434
435 /* kcm sock is locked. */
436 static struct kcm_psock *reserve_psock(struct kcm_sock *kcm)
437 {
438 struct kcm_mux *mux = kcm->mux;
439 struct kcm_psock *psock;
440
441 psock = kcm->tx_psock;
442
443 smp_rmb(); /* Must read tx_psock before tx_wait */
444
445 if (psock) {
446 WARN_ON(kcm->tx_wait);
447 if (unlikely(psock->tx_stopped))
448 unreserve_psock(kcm);
449 else
450 return kcm->tx_psock;
451 }
452
453 spin_lock_bh(&mux->lock);
454
455 /* Check again under lock to see if psock was reserved for this
456 * psock via psock_unreserve.
457 */
458 psock = kcm->tx_psock;
459 if (unlikely(psock)) {
460 WARN_ON(kcm->tx_wait);
461 spin_unlock_bh(&mux->lock);
462 return kcm->tx_psock;
463 }
464
465 if (!list_empty(&mux->psocks_avail)) {
466 psock = list_first_entry(&mux->psocks_avail,
467 struct kcm_psock,
468 psock_avail_list);
469 list_del(&psock->psock_avail_list);
470 if (kcm->tx_wait) {
471 list_del(&kcm->wait_psock_list);
472 kcm->tx_wait = false;
473 }
474 kcm->tx_psock = psock;
475 psock->tx_kcm = kcm;
476 KCM_STATS_INCR(psock->stats.reserved);
477 } else if (!kcm->tx_wait) {
478 list_add_tail(&kcm->wait_psock_list,
479 &mux->kcm_tx_waiters);
480 kcm->tx_wait = true;
481 }
482
483 spin_unlock_bh(&mux->lock);
484
485 return psock;
486 }
487
488 /* mux lock held */
489 static void psock_now_avail(struct kcm_psock *psock)
490 {
491 struct kcm_mux *mux = psock->mux;
492 struct kcm_sock *kcm;
493
494 if (list_empty(&mux->kcm_tx_waiters)) {
495 list_add_tail(&psock->psock_avail_list,
496 &mux->psocks_avail);
497 } else {
498 kcm = list_first_entry(&mux->kcm_tx_waiters,
499 struct kcm_sock,
500 wait_psock_list);
501 list_del(&kcm->wait_psock_list);
502 kcm->tx_wait = false;
503 psock->tx_kcm = kcm;
504
505 /* Commit before changing tx_psock since that is read in
506 * reserve_psock before queuing work.
507 */
508 smp_mb();
509
510 kcm->tx_psock = psock;
511 KCM_STATS_INCR(psock->stats.reserved);
512 queue_work(kcm_wq, &kcm->tx_work);
513 }
514 }
515
516 /* kcm sock is locked. */
517 static void unreserve_psock(struct kcm_sock *kcm)
518 {
519 struct kcm_psock *psock;
520 struct kcm_mux *mux = kcm->mux;
521
522 spin_lock_bh(&mux->lock);
523
524 psock = kcm->tx_psock;
525
526 if (WARN_ON(!psock)) {
527 spin_unlock_bh(&mux->lock);
528 return;
529 }
530
531 smp_rmb(); /* Read tx_psock before tx_wait */
532
533 kcm_update_tx_mux_stats(mux, psock);
534
535 WARN_ON(kcm->tx_wait);
536
537 kcm->tx_psock = NULL;
538 psock->tx_kcm = NULL;
539 KCM_STATS_INCR(psock->stats.unreserved);
540
541 if (unlikely(psock->tx_stopped)) {
542 if (psock->done) {
543 /* Deferred free */
544 list_del(&psock->psock_list);
545 mux->psocks_cnt--;
546 sock_put(psock->sk);
547 fput(psock->sk->sk_socket->file);
548 kmem_cache_free(kcm_psockp, psock);
549 }
550
551 /* Don't put back on available list */
552
553 spin_unlock_bh(&mux->lock);
554
555 return;
556 }
557
558 psock_now_avail(psock);
559
560 spin_unlock_bh(&mux->lock);
561 }
562
563 static void kcm_report_tx_retry(struct kcm_sock *kcm)
564 {
565 struct kcm_mux *mux = kcm->mux;
566
567 spin_lock_bh(&mux->lock);
568 KCM_STATS_INCR(mux->stats.tx_retries);
569 spin_unlock_bh(&mux->lock);
570 }
571
572 /* Write any messages ready on the kcm socket. Called with kcm sock lock
573 * held. Return bytes actually sent or error.
574 */
575 static int kcm_write_msgs(struct kcm_sock *kcm)
576 {
577 struct sock *sk = &kcm->sk;
578 struct kcm_psock *psock;
579 struct sk_buff *skb, *head;
580 struct kcm_tx_msg *txm;
581 unsigned short fragidx, frag_offset;
582 unsigned int sent, total_sent = 0;
583 int ret = 0;
584
585 kcm->tx_wait_more = false;
586 psock = kcm->tx_psock;
587 if (unlikely(psock && psock->tx_stopped)) {
588 /* A reserved psock was aborted asynchronously. Unreserve
589 * it and we'll retry the message.
590 */
591 unreserve_psock(kcm);
592 kcm_report_tx_retry(kcm);
593 if (skb_queue_empty(&sk->sk_write_queue))
594 return 0;
595
596 kcm_tx_msg(skb_peek(&sk->sk_write_queue))->sent = 0;
597
598 } else if (skb_queue_empty(&sk->sk_write_queue)) {
599 return 0;
600 }
601
602 head = skb_peek(&sk->sk_write_queue);
603 txm = kcm_tx_msg(head);
604
605 if (txm->sent) {
606 /* Send of first skbuff in queue already in progress */
607 if (WARN_ON(!psock)) {
608 ret = -EINVAL;
609 goto out;
610 }
611 sent = txm->sent;
612 frag_offset = txm->frag_offset;
613 fragidx = txm->fragidx;
614 skb = txm->frag_skb;
615
616 goto do_frag;
617 }
618
619 try_again:
620 psock = reserve_psock(kcm);
621 if (!psock)
622 goto out;
623
624 do {
625 skb = head;
626 txm = kcm_tx_msg(head);
627 sent = 0;
628
629 do_frag_list:
630 if (WARN_ON(!skb_shinfo(skb)->nr_frags)) {
631 ret = -EINVAL;
632 goto out;
633 }
634
635 for (fragidx = 0; fragidx < skb_shinfo(skb)->nr_frags;
636 fragidx++) {
637 skb_frag_t *frag;
638
639 frag_offset = 0;
640 do_frag:
641 frag = &skb_shinfo(skb)->frags[fragidx];
642 if (WARN_ON(!skb_frag_size(frag))) {
643 ret = -EINVAL;
644 goto out;
645 }
646
647 ret = kernel_sendpage(psock->sk->sk_socket,
648 skb_frag_page(frag),
649 skb_frag_off(frag) + frag_offset,
650 skb_frag_size(frag) - frag_offset,
651 MSG_DONTWAIT);
652 if (ret <= 0) {
653 if (ret == -EAGAIN) {
654 /* Save state to try again when there's
655 * write space on the socket
656 */
657 txm->sent = sent;
658 txm->frag_offset = frag_offset;
659 txm->fragidx = fragidx;
660 txm->frag_skb = skb;
661
662 ret = 0;
663 goto out;
664 }
665
666 /* Hard failure in sending message, abort this
667 * psock since it has lost framing
668 * synchonization and retry sending the
669 * message from the beginning.
670 */
671 kcm_abort_tx_psock(psock, ret ? -ret : EPIPE,
672 true);
673 unreserve_psock(kcm);
674
675 txm->sent = 0;
676 kcm_report_tx_retry(kcm);
677 ret = 0;
678
679 goto try_again;
680 }
681
682 sent += ret;
683 frag_offset += ret;
684 KCM_STATS_ADD(psock->stats.tx_bytes, ret);
685 if (frag_offset < skb_frag_size(frag)) {
686 /* Not finished with this frag */
687 goto do_frag;
688 }
689 }
690
691 if (skb == head) {
692 if (skb_has_frag_list(skb)) {
693 skb = skb_shinfo(skb)->frag_list;
694 goto do_frag_list;
695 }
696 } else if (skb->next) {
697 skb = skb->next;
698 goto do_frag_list;
699 }
700
701 /* Successfully sent the whole packet, account for it. */
702 skb_dequeue(&sk->sk_write_queue);
703 kfree_skb(head);
704 sk->sk_wmem_queued -= sent;
705 total_sent += sent;
706 KCM_STATS_INCR(psock->stats.tx_msgs);
707 } while ((head = skb_peek(&sk->sk_write_queue)));
708 out:
709 if (!head) {
710 /* Done with all queued messages. */
711 WARN_ON(!skb_queue_empty(&sk->sk_write_queue));
712 unreserve_psock(kcm);
713 }
714
715 /* Check if write space is available */
716 sk->sk_write_space(sk);
717
718 return total_sent ? : ret;
719 }
720
721 static void kcm_tx_work(struct work_struct *w)
722 {
723 struct kcm_sock *kcm = container_of(w, struct kcm_sock, tx_work);
724 struct sock *sk = &kcm->sk;
725 int err;
726
727 lock_sock(sk);
728
729 /* Primarily for SOCK_DGRAM sockets, also handle asynchronous tx
730 * aborts
731 */
732 err = kcm_write_msgs(kcm);
733 if (err < 0) {
734 /* Hard failure in write, report error on KCM socket */
735 pr_warn("KCM: Hard failure on kcm_write_msgs %d\n", err);
736 report_csk_error(&kcm->sk, -err);
737 goto out;
738 }
739
740 /* Primarily for SOCK_SEQPACKET sockets */
741 if (likely(sk->sk_socket) &&
742 test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
743 clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
744 sk->sk_write_space(sk);
745 }
746
747 out:
748 release_sock(sk);
749 }
750
751 static void kcm_push(struct kcm_sock *kcm)
752 {
753 if (kcm->tx_wait_more)
754 kcm_write_msgs(kcm);
755 }
756
757 static ssize_t kcm_sendpage(struct socket *sock, struct page *page,
758 int offset, size_t size, int flags)
759
760 {
761 struct sock *sk = sock->sk;
762 struct kcm_sock *kcm = kcm_sk(sk);
763 struct sk_buff *skb = NULL, *head = NULL;
764 long timeo = sock_sndtimeo(sk, flags & MSG_DONTWAIT);
765 bool eor;
766 int err = 0;
767 int i;
768
769 if (flags & MSG_SENDPAGE_NOTLAST)
770 flags |= MSG_MORE;
771
772 /* No MSG_EOR from splice, only look at MSG_MORE */
773 eor = !(flags & MSG_MORE);
774
775 lock_sock(sk);
776
777 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
778
779 err = -EPIPE;
780 if (sk->sk_err)
781 goto out_error;
782
783 if (kcm->seq_skb) {
784 /* Previously opened message */
785 head = kcm->seq_skb;
786 skb = kcm_tx_msg(head)->last_skb;
787 i = skb_shinfo(skb)->nr_frags;
788
789 if (skb_can_coalesce(skb, i, page, offset)) {
790 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], size);
791 skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
792 goto coalesced;
793 }
794
795 if (i >= MAX_SKB_FRAGS) {
796 struct sk_buff *tskb;
797
798 tskb = alloc_skb(0, sk->sk_allocation);
799 while (!tskb) {
800 kcm_push(kcm);
801 err = sk_stream_wait_memory(sk, &timeo);
802 if (err)
803 goto out_error;
804 }
805
806 if (head == skb)
807 skb_shinfo(head)->frag_list = tskb;
808 else
809 skb->next = tskb;
810
811 skb = tskb;
812 skb->ip_summed = CHECKSUM_UNNECESSARY;
813 i = 0;
814 }
815 } else {
816 /* Call the sk_stream functions to manage the sndbuf mem. */
817 if (!sk_stream_memory_free(sk)) {
818 kcm_push(kcm);
819 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
820 err = sk_stream_wait_memory(sk, &timeo);
821 if (err)
822 goto out_error;
823 }
824
825 head = alloc_skb(0, sk->sk_allocation);
826 while (!head) {
827 kcm_push(kcm);
828 err = sk_stream_wait_memory(sk, &timeo);
829 if (err)
830 goto out_error;
831 }
832
833 skb = head;
834 i = 0;
835 }
836
837 get_page(page);
838 skb_fill_page_desc(skb, i, page, offset, size);
839 skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
840
841 coalesced:
842 skb->len += size;
843 skb->data_len += size;
844 skb->truesize += size;
845 sk->sk_wmem_queued += size;
846 sk_mem_charge(sk, size);
847
848 if (head != skb) {
849 head->len += size;
850 head->data_len += size;
851 head->truesize += size;
852 }
853
854 if (eor) {
855 bool not_busy = skb_queue_empty(&sk->sk_write_queue);
856
857 /* Message complete, queue it on send buffer */
858 __skb_queue_tail(&sk->sk_write_queue, head);
859 kcm->seq_skb = NULL;
860 KCM_STATS_INCR(kcm->stats.tx_msgs);
861
862 if (flags & MSG_BATCH) {
863 kcm->tx_wait_more = true;
864 } else if (kcm->tx_wait_more || not_busy) {
865 err = kcm_write_msgs(kcm);
866 if (err < 0) {
867 /* We got a hard error in write_msgs but have
868 * already queued this message. Report an error
869 * in the socket, but don't affect return value
870 * from sendmsg
871 */
872 pr_warn("KCM: Hard failure on kcm_write_msgs\n");
873 report_csk_error(&kcm->sk, -err);
874 }
875 }
876 } else {
877 /* Message not complete, save state */
878 kcm->seq_skb = head;
879 kcm_tx_msg(head)->last_skb = skb;
880 }
881
882 KCM_STATS_ADD(kcm->stats.tx_bytes, size);
883
884 release_sock(sk);
885 return size;
886
887 out_error:
888 kcm_push(kcm);
889
890 err = sk_stream_error(sk, flags, err);
891
892 /* make sure we wake any epoll edge trigger waiter */
893 if (unlikely(skb_queue_len(&sk->sk_write_queue) == 0 && err == -EAGAIN))
894 sk->sk_write_space(sk);
895
896 release_sock(sk);
897 return err;
898 }
899
900 static int kcm_sendmsg(struct socket *sock, struct msghdr *msg, size_t len)
901 {
902 struct sock *sk = sock->sk;
903 struct kcm_sock *kcm = kcm_sk(sk);
904 struct sk_buff *skb = NULL, *head = NULL;
905 size_t copy, copied = 0;
906 long timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
907 int eor = (sock->type == SOCK_DGRAM) ?
908 !(msg->msg_flags & MSG_MORE) : !!(msg->msg_flags & MSG_EOR);
909 int err = -EPIPE;
910
911 lock_sock(sk);
912
913 /* Per tcp_sendmsg this should be in poll */
914 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
915
916 if (sk->sk_err)
917 goto out_error;
918
919 if (kcm->seq_skb) {
920 /* Previously opened message */
921 head = kcm->seq_skb;
922 skb = kcm_tx_msg(head)->last_skb;
923 goto start;
924 }
925
926 /* Call the sk_stream functions to manage the sndbuf mem. */
927 if (!sk_stream_memory_free(sk)) {
928 kcm_push(kcm);
929 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
930 err = sk_stream_wait_memory(sk, &timeo);
931 if (err)
932 goto out_error;
933 }
934
935 if (msg_data_left(msg)) {
936 /* New message, alloc head skb */
937 head = alloc_skb(0, sk->sk_allocation);
938 while (!head) {
939 kcm_push(kcm);
940 err = sk_stream_wait_memory(sk, &timeo);
941 if (err)
942 goto out_error;
943
944 head = alloc_skb(0, sk->sk_allocation);
945 }
946
947 skb = head;
948
949 /* Set ip_summed to CHECKSUM_UNNECESSARY to avoid calling
950 * csum_and_copy_from_iter from skb_do_copy_data_nocache.
951 */
952 skb->ip_summed = CHECKSUM_UNNECESSARY;
953 }
954
955 start:
956 while (msg_data_left(msg)) {
957 bool merge = true;
958 int i = skb_shinfo(skb)->nr_frags;
959 struct page_frag *pfrag = sk_page_frag(sk);
960
961 if (!sk_page_frag_refill(sk, pfrag))
962 goto wait_for_memory;
963
964 if (!skb_can_coalesce(skb, i, pfrag->page,
965 pfrag->offset)) {
966 if (i == MAX_SKB_FRAGS) {
967 struct sk_buff *tskb;
968
969 tskb = alloc_skb(0, sk->sk_allocation);
970 if (!tskb)
971 goto wait_for_memory;
972
973 if (head == skb)
974 skb_shinfo(head)->frag_list = tskb;
975 else
976 skb->next = tskb;
977
978 skb = tskb;
979 skb->ip_summed = CHECKSUM_UNNECESSARY;
980 continue;
981 }
982 merge = false;
983 }
984
985 copy = min_t(int, msg_data_left(msg),
986 pfrag->size - pfrag->offset);
987
988 if (!sk_wmem_schedule(sk, copy))
989 goto wait_for_memory;
990
991 err = skb_copy_to_page_nocache(sk, &msg->msg_iter, skb,
992 pfrag->page,
993 pfrag->offset,
994 copy);
995 if (err)
996 goto out_error;
997
998 /* Update the skb. */
999 if (merge) {
1000 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
1001 } else {
1002 skb_fill_page_desc(skb, i, pfrag->page,
1003 pfrag->offset, copy);
1004 get_page(pfrag->page);
1005 }
1006
1007 pfrag->offset += copy;
1008 copied += copy;
1009 if (head != skb) {
1010 head->len += copy;
1011 head->data_len += copy;
1012 }
1013
1014 continue;
1015
1016 wait_for_memory:
1017 kcm_push(kcm);
1018 err = sk_stream_wait_memory(sk, &timeo);
1019 if (err)
1020 goto out_error;
1021 }
1022
1023 if (eor) {
1024 bool not_busy = skb_queue_empty(&sk->sk_write_queue);
1025
1026 if (head) {
1027 /* Message complete, queue it on send buffer */
1028 __skb_queue_tail(&sk->sk_write_queue, head);
1029 kcm->seq_skb = NULL;
1030 KCM_STATS_INCR(kcm->stats.tx_msgs);
1031 }
1032
1033 if (msg->msg_flags & MSG_BATCH) {
1034 kcm->tx_wait_more = true;
1035 } else if (kcm->tx_wait_more || not_busy) {
1036 err = kcm_write_msgs(kcm);
1037 if (err < 0) {
1038 /* We got a hard error in write_msgs but have
1039 * already queued this message. Report an error
1040 * in the socket, but don't affect return value
1041 * from sendmsg
1042 */
1043 pr_warn("KCM: Hard failure on kcm_write_msgs\n");
1044 report_csk_error(&kcm->sk, -err);
1045 }
1046 }
1047 } else {
1048 /* Message not complete, save state */
1049 partial_message:
1050 if (head) {
1051 kcm->seq_skb = head;
1052 kcm_tx_msg(head)->last_skb = skb;
1053 }
1054 }
1055
1056 KCM_STATS_ADD(kcm->stats.tx_bytes, copied);
1057
1058 release_sock(sk);
1059 return copied;
1060
1061 out_error:
1062 kcm_push(kcm);
1063
1064 if (copied && sock->type == SOCK_SEQPACKET) {
1065 /* Wrote some bytes before encountering an
1066 * error, return partial success.
1067 */
1068 goto partial_message;
1069 }
1070
1071 if (head != kcm->seq_skb)
1072 kfree_skb(head);
1073
1074 err = sk_stream_error(sk, msg->msg_flags, err);
1075
1076 /* make sure we wake any epoll edge trigger waiter */
1077 if (unlikely(skb_queue_len(&sk->sk_write_queue) == 0 && err == -EAGAIN))
1078 sk->sk_write_space(sk);
1079
1080 release_sock(sk);
1081 return err;
1082 }
1083
1084 static struct sk_buff *kcm_wait_data(struct sock *sk, int flags,
1085 long timeo, int *err)
1086 {
1087 struct sk_buff *skb;
1088
1089 while (!(skb = skb_peek(&sk->sk_receive_queue))) {
1090 if (sk->sk_err) {
1091 *err = sock_error(sk);
1092 return NULL;
1093 }
1094
1095 if (sock_flag(sk, SOCK_DONE))
1096 return NULL;
1097
1098 if ((flags & MSG_DONTWAIT) || !timeo) {
1099 *err = -EAGAIN;
1100 return NULL;
1101 }
1102
1103 sk_wait_data(sk, &timeo, NULL);
1104
1105 /* Handle signals */
1106 if (signal_pending(current)) {
1107 *err = sock_intr_errno(timeo);
1108 return NULL;
1109 }
1110 }
1111
1112 return skb;
1113 }
1114
1115 static int kcm_recvmsg(struct socket *sock, struct msghdr *msg,
1116 size_t len, int flags)
1117 {
1118 struct sock *sk = sock->sk;
1119 struct kcm_sock *kcm = kcm_sk(sk);
1120 int err = 0;
1121 long timeo;
1122 struct strp_msg *stm;
1123 int copied = 0;
1124 struct sk_buff *skb;
1125
1126 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1127
1128 lock_sock(sk);
1129
1130 skb = kcm_wait_data(sk, flags, timeo, &err);
1131 if (!skb)
1132 goto out;
1133
1134 /* Okay, have a message on the receive queue */
1135
1136 stm = strp_msg(skb);
1137
1138 if (len > stm->full_len)
1139 len = stm->full_len;
1140
1141 err = skb_copy_datagram_msg(skb, stm->offset, msg, len);
1142 if (err < 0)
1143 goto out;
1144
1145 copied = len;
1146 if (likely(!(flags & MSG_PEEK))) {
1147 KCM_STATS_ADD(kcm->stats.rx_bytes, copied);
1148 if (copied < stm->full_len) {
1149 if (sock->type == SOCK_DGRAM) {
1150 /* Truncated message */
1151 msg->msg_flags |= MSG_TRUNC;
1152 goto msg_finished;
1153 }
1154 stm->offset += copied;
1155 stm->full_len -= copied;
1156 } else {
1157 msg_finished:
1158 /* Finished with message */
1159 msg->msg_flags |= MSG_EOR;
1160 KCM_STATS_INCR(kcm->stats.rx_msgs);
1161 skb_unlink(skb, &sk->sk_receive_queue);
1162 kfree_skb(skb);
1163 }
1164 }
1165
1166 out:
1167 release_sock(sk);
1168
1169 return copied ? : err;
1170 }
1171
1172 static ssize_t kcm_splice_read(struct socket *sock, loff_t *ppos,
1173 struct pipe_inode_info *pipe, size_t len,
1174 unsigned int flags)
1175 {
1176 struct sock *sk = sock->sk;
1177 struct kcm_sock *kcm = kcm_sk(sk);
1178 long timeo;
1179 struct strp_msg *stm;
1180 int err = 0;
1181 ssize_t copied;
1182 struct sk_buff *skb;
1183
1184 /* Only support splice for SOCKSEQPACKET */
1185
1186 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1187
1188 lock_sock(sk);
1189
1190 skb = kcm_wait_data(sk, flags, timeo, &err);
1191 if (!skb)
1192 goto err_out;
1193
1194 /* Okay, have a message on the receive queue */
1195
1196 stm = strp_msg(skb);
1197
1198 if (len > stm->full_len)
1199 len = stm->full_len;
1200
1201 copied = skb_splice_bits(skb, sk, stm->offset, pipe, len, flags);
1202 if (copied < 0) {
1203 err = copied;
1204 goto err_out;
1205 }
1206
1207 KCM_STATS_ADD(kcm->stats.rx_bytes, copied);
1208
1209 stm->offset += copied;
1210 stm->full_len -= copied;
1211
1212 /* We have no way to return MSG_EOR. If all the bytes have been
1213 * read we still leave the message in the receive socket buffer.
1214 * A subsequent recvmsg needs to be done to return MSG_EOR and
1215 * finish reading the message.
1216 */
1217
1218 release_sock(sk);
1219
1220 return copied;
1221
1222 err_out:
1223 release_sock(sk);
1224
1225 return err;
1226 }
1227
1228 /* kcm sock lock held */
1229 static void kcm_recv_disable(struct kcm_sock *kcm)
1230 {
1231 struct kcm_mux *mux = kcm->mux;
1232
1233 if (kcm->rx_disabled)
1234 return;
1235
1236 spin_lock_bh(&mux->rx_lock);
1237
1238 kcm->rx_disabled = 1;
1239
1240 /* If a psock is reserved we'll do cleanup in unreserve */
1241 if (!kcm->rx_psock) {
1242 if (kcm->rx_wait) {
1243 list_del(&kcm->wait_rx_list);
1244 kcm->rx_wait = false;
1245 }
1246
1247 requeue_rx_msgs(mux, &kcm->sk.sk_receive_queue);
1248 }
1249
1250 spin_unlock_bh(&mux->rx_lock);
1251 }
1252
1253 /* kcm sock lock held */
1254 static void kcm_recv_enable(struct kcm_sock *kcm)
1255 {
1256 struct kcm_mux *mux = kcm->mux;
1257
1258 if (!kcm->rx_disabled)
1259 return;
1260
1261 spin_lock_bh(&mux->rx_lock);
1262
1263 kcm->rx_disabled = 0;
1264 kcm_rcv_ready(kcm);
1265
1266 spin_unlock_bh(&mux->rx_lock);
1267 }
1268
1269 static int kcm_setsockopt(struct socket *sock, int level, int optname,
1270 char __user *optval, unsigned int optlen)
1271 {
1272 struct kcm_sock *kcm = kcm_sk(sock->sk);
1273 int val, valbool;
1274 int err = 0;
1275
1276 if (level != SOL_KCM)
1277 return -ENOPROTOOPT;
1278
1279 if (optlen < sizeof(int))
1280 return -EINVAL;
1281
1282 if (get_user(val, (int __user *)optval))
1283 return -EINVAL;
1284
1285 valbool = val ? 1 : 0;
1286
1287 switch (optname) {
1288 case KCM_RECV_DISABLE:
1289 lock_sock(&kcm->sk);
1290 if (valbool)
1291 kcm_recv_disable(kcm);
1292 else
1293 kcm_recv_enable(kcm);
1294 release_sock(&kcm->sk);
1295 break;
1296 default:
1297 err = -ENOPROTOOPT;
1298 }
1299
1300 return err;
1301 }
1302
1303 static int kcm_getsockopt(struct socket *sock, int level, int optname,
1304 char __user *optval, int __user *optlen)
1305 {
1306 struct kcm_sock *kcm = kcm_sk(sock->sk);
1307 int val, len;
1308
1309 if (level != SOL_KCM)
1310 return -ENOPROTOOPT;
1311
1312 if (get_user(len, optlen))
1313 return -EFAULT;
1314
1315 len = min_t(unsigned int, len, sizeof(int));
1316 if (len < 0)
1317 return -EINVAL;
1318
1319 switch (optname) {
1320 case KCM_RECV_DISABLE:
1321 val = kcm->rx_disabled;
1322 break;
1323 default:
1324 return -ENOPROTOOPT;
1325 }
1326
1327 if (put_user(len, optlen))
1328 return -EFAULT;
1329 if (copy_to_user(optval, &val, len))
1330 return -EFAULT;
1331 return 0;
1332 }
1333
1334 static void init_kcm_sock(struct kcm_sock *kcm, struct kcm_mux *mux)
1335 {
1336 struct kcm_sock *tkcm;
1337 struct list_head *head;
1338 int index = 0;
1339
1340 /* For SOCK_SEQPACKET sock type, datagram_poll checks the sk_state, so
1341 * we set sk_state, otherwise epoll_wait always returns right away with
1342 * EPOLLHUP
1343 */
1344 kcm->sk.sk_state = TCP_ESTABLISHED;
1345
1346 /* Add to mux's kcm sockets list */
1347 kcm->mux = mux;
1348 spin_lock_bh(&mux->lock);
1349
1350 head = &mux->kcm_socks;
1351 list_for_each_entry(tkcm, &mux->kcm_socks, kcm_sock_list) {
1352 if (tkcm->index != index)
1353 break;
1354 head = &tkcm->kcm_sock_list;
1355 index++;
1356 }
1357
1358 list_add(&kcm->kcm_sock_list, head);
1359 kcm->index = index;
1360
1361 mux->kcm_socks_cnt++;
1362 spin_unlock_bh(&mux->lock);
1363
1364 INIT_WORK(&kcm->tx_work, kcm_tx_work);
1365
1366 spin_lock_bh(&mux->rx_lock);
1367 kcm_rcv_ready(kcm);
1368 spin_unlock_bh(&mux->rx_lock);
1369 }
1370
1371 static int kcm_attach(struct socket *sock, struct socket *csock,
1372 struct bpf_prog *prog)
1373 {
1374 struct kcm_sock *kcm = kcm_sk(sock->sk);
1375 struct kcm_mux *mux = kcm->mux;
1376 struct sock *csk;
1377 struct kcm_psock *psock = NULL, *tpsock;
1378 struct list_head *head;
1379 int index = 0;
1380 static const struct strp_callbacks cb = {
1381 .rcv_msg = kcm_rcv_strparser,
1382 .parse_msg = kcm_parse_func_strparser,
1383 .read_sock_done = kcm_read_sock_done,
1384 };
1385 int err = 0;
1386
1387 csk = csock->sk;
1388 if (!csk)
1389 return -EINVAL;
1390
1391 lock_sock(csk);
1392
1393 /* Only allow TCP sockets to be attached for now */
1394 if ((csk->sk_family != AF_INET && csk->sk_family != AF_INET6) ||
1395 csk->sk_protocol != IPPROTO_TCP) {
1396 err = -EOPNOTSUPP;
1397 goto out;
1398 }
1399
1400 /* Don't allow listeners or closed sockets */
1401 if (csk->sk_state == TCP_LISTEN || csk->sk_state == TCP_CLOSE) {
1402 err = -EOPNOTSUPP;
1403 goto out;
1404 }
1405
1406 psock = kmem_cache_zalloc(kcm_psockp, GFP_KERNEL);
1407 if (!psock) {
1408 err = -ENOMEM;
1409 goto out;
1410 }
1411
1412 psock->mux = mux;
1413 psock->sk = csk;
1414 psock->bpf_prog = prog;
1415
1416 err = strp_init(&psock->strp, csk, &cb);
1417 if (err) {
1418 kmem_cache_free(kcm_psockp, psock);
1419 goto out;
1420 }
1421
1422 write_lock_bh(&csk->sk_callback_lock);
1423
1424 /* Check if sk_user_data is aready by KCM or someone else.
1425 * Must be done under lock to prevent race conditions.
1426 */
1427 if (csk->sk_user_data) {
1428 write_unlock_bh(&csk->sk_callback_lock);
1429 strp_stop(&psock->strp);
1430 strp_done(&psock->strp);
1431 kmem_cache_free(kcm_psockp, psock);
1432 err = -EALREADY;
1433 goto out;
1434 }
1435
1436 psock->save_data_ready = csk->sk_data_ready;
1437 psock->save_write_space = csk->sk_write_space;
1438 psock->save_state_change = csk->sk_state_change;
1439 csk->sk_user_data = psock;
1440 csk->sk_data_ready = psock_data_ready;
1441 csk->sk_write_space = psock_write_space;
1442 csk->sk_state_change = psock_state_change;
1443
1444 write_unlock_bh(&csk->sk_callback_lock);
1445
1446 sock_hold(csk);
1447
1448 /* Finished initialization, now add the psock to the MUX. */
1449 spin_lock_bh(&mux->lock);
1450 head = &mux->psocks;
1451 list_for_each_entry(tpsock, &mux->psocks, psock_list) {
1452 if (tpsock->index != index)
1453 break;
1454 head = &tpsock->psock_list;
1455 index++;
1456 }
1457
1458 list_add(&psock->psock_list, head);
1459 psock->index = index;
1460
1461 KCM_STATS_INCR(mux->stats.psock_attach);
1462 mux->psocks_cnt++;
1463 psock_now_avail(psock);
1464 spin_unlock_bh(&mux->lock);
1465
1466 /* Schedule RX work in case there are already bytes queued */
1467 strp_check_rcv(&psock->strp);
1468
1469 out:
1470 release_sock(csk);
1471
1472 return err;
1473 }
1474
1475 static int kcm_attach_ioctl(struct socket *sock, struct kcm_attach *info)
1476 {
1477 struct socket *csock;
1478 struct bpf_prog *prog;
1479 int err;
1480
1481 csock = sockfd_lookup(info->fd, &err);
1482 if (!csock)
1483 return -ENOENT;
1484
1485 prog = bpf_prog_get_type(info->bpf_fd, BPF_PROG_TYPE_SOCKET_FILTER);
1486 if (IS_ERR(prog)) {
1487 err = PTR_ERR(prog);
1488 goto out;
1489 }
1490
1491 err = kcm_attach(sock, csock, prog);
1492 if (err) {
1493 bpf_prog_put(prog);
1494 goto out;
1495 }
1496
1497 /* Keep reference on file also */
1498
1499 return 0;
1500 out:
1501 fput(csock->file);
1502 return err;
1503 }
1504
1505 static void kcm_unattach(struct kcm_psock *psock)
1506 {
1507 struct sock *csk = psock->sk;
1508 struct kcm_mux *mux = psock->mux;
1509
1510 lock_sock(csk);
1511
1512 /* Stop getting callbacks from TCP socket. After this there should
1513 * be no way to reserve a kcm for this psock.
1514 */
1515 write_lock_bh(&csk->sk_callback_lock);
1516 csk->sk_user_data = NULL;
1517 csk->sk_data_ready = psock->save_data_ready;
1518 csk->sk_write_space = psock->save_write_space;
1519 csk->sk_state_change = psock->save_state_change;
1520 strp_stop(&psock->strp);
1521
1522 if (WARN_ON(psock->rx_kcm)) {
1523 write_unlock_bh(&csk->sk_callback_lock);
1524 release_sock(csk);
1525 return;
1526 }
1527
1528 spin_lock_bh(&mux->rx_lock);
1529
1530 /* Stop receiver activities. After this point psock should not be
1531 * able to get onto ready list either through callbacks or work.
1532 */
1533 if (psock->ready_rx_msg) {
1534 list_del(&psock->psock_ready_list);
1535 kfree_skb(psock->ready_rx_msg);
1536 psock->ready_rx_msg = NULL;
1537 KCM_STATS_INCR(mux->stats.rx_ready_drops);
1538 }
1539
1540 spin_unlock_bh(&mux->rx_lock);
1541
1542 write_unlock_bh(&csk->sk_callback_lock);
1543
1544 /* Call strp_done without sock lock */
1545 release_sock(csk);
1546 strp_done(&psock->strp);
1547 lock_sock(csk);
1548
1549 bpf_prog_put(psock->bpf_prog);
1550
1551 spin_lock_bh(&mux->lock);
1552
1553 aggregate_psock_stats(&psock->stats, &mux->aggregate_psock_stats);
1554 save_strp_stats(&psock->strp, &mux->aggregate_strp_stats);
1555
1556 KCM_STATS_INCR(mux->stats.psock_unattach);
1557
1558 if (psock->tx_kcm) {
1559 /* psock was reserved. Just mark it finished and we will clean
1560 * up in the kcm paths, we need kcm lock which can not be
1561 * acquired here.
1562 */
1563 KCM_STATS_INCR(mux->stats.psock_unattach_rsvd);
1564 spin_unlock_bh(&mux->lock);
1565
1566 /* We are unattaching a socket that is reserved. Abort the
1567 * socket since we may be out of sync in sending on it. We need
1568 * to do this without the mux lock.
1569 */
1570 kcm_abort_tx_psock(psock, EPIPE, false);
1571
1572 spin_lock_bh(&mux->lock);
1573 if (!psock->tx_kcm) {
1574 /* psock now unreserved in window mux was unlocked */
1575 goto no_reserved;
1576 }
1577 psock->done = 1;
1578
1579 /* Commit done before queuing work to process it */
1580 smp_mb();
1581
1582 /* Queue tx work to make sure psock->done is handled */
1583 queue_work(kcm_wq, &psock->tx_kcm->tx_work);
1584 spin_unlock_bh(&mux->lock);
1585 } else {
1586 no_reserved:
1587 if (!psock->tx_stopped)
1588 list_del(&psock->psock_avail_list);
1589 list_del(&psock->psock_list);
1590 mux->psocks_cnt--;
1591 spin_unlock_bh(&mux->lock);
1592
1593 sock_put(csk);
1594 fput(csk->sk_socket->file);
1595 kmem_cache_free(kcm_psockp, psock);
1596 }
1597
1598 release_sock(csk);
1599 }
1600
1601 static int kcm_unattach_ioctl(struct socket *sock, struct kcm_unattach *info)
1602 {
1603 struct kcm_sock *kcm = kcm_sk(sock->sk);
1604 struct kcm_mux *mux = kcm->mux;
1605 struct kcm_psock *psock;
1606 struct socket *csock;
1607 struct sock *csk;
1608 int err;
1609
1610 csock = sockfd_lookup(info->fd, &err);
1611 if (!csock)
1612 return -ENOENT;
1613
1614 csk = csock->sk;
1615 if (!csk) {
1616 err = -EINVAL;
1617 goto out;
1618 }
1619
1620 err = -ENOENT;
1621
1622 spin_lock_bh(&mux->lock);
1623
1624 list_for_each_entry(psock, &mux->psocks, psock_list) {
1625 if (psock->sk != csk)
1626 continue;
1627
1628 /* Found the matching psock */
1629
1630 if (psock->unattaching || WARN_ON(psock->done)) {
1631 err = -EALREADY;
1632 break;
1633 }
1634
1635 psock->unattaching = 1;
1636
1637 spin_unlock_bh(&mux->lock);
1638
1639 /* Lower socket lock should already be held */
1640 kcm_unattach(psock);
1641
1642 err = 0;
1643 goto out;
1644 }
1645
1646 spin_unlock_bh(&mux->lock);
1647
1648 out:
1649 fput(csock->file);
1650 return err;
1651 }
1652
1653 static struct proto kcm_proto = {
1654 .name = "KCM",
1655 .owner = THIS_MODULE,
1656 .obj_size = sizeof(struct kcm_sock),
1657 };
1658
1659 /* Clone a kcm socket. */
1660 static struct file *kcm_clone(struct socket *osock)
1661 {
1662 struct socket *newsock;
1663 struct sock *newsk;
1664
1665 newsock = sock_alloc();
1666 if (!newsock)
1667 return ERR_PTR(-ENFILE);
1668
1669 newsock->type = osock->type;
1670 newsock->ops = osock->ops;
1671
1672 __module_get(newsock->ops->owner);
1673
1674 newsk = sk_alloc(sock_net(osock->sk), PF_KCM, GFP_KERNEL,
1675 &kcm_proto, false);
1676 if (!newsk) {
1677 sock_release(newsock);
1678 return ERR_PTR(-ENOMEM);
1679 }
1680 sock_init_data(newsock, newsk);
1681 init_kcm_sock(kcm_sk(newsk), kcm_sk(osock->sk)->mux);
1682
1683 return sock_alloc_file(newsock, 0, osock->sk->sk_prot_creator->name);
1684 }
1685
1686 static int kcm_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1687 {
1688 int err;
1689
1690 switch (cmd) {
1691 case SIOCKCMATTACH: {
1692 struct kcm_attach info;
1693
1694 if (copy_from_user(&info, (void __user *)arg, sizeof(info)))
1695 return -EFAULT;
1696
1697 err = kcm_attach_ioctl(sock, &info);
1698
1699 break;
1700 }
1701 case SIOCKCMUNATTACH: {
1702 struct kcm_unattach info;
1703
1704 if (copy_from_user(&info, (void __user *)arg, sizeof(info)))
1705 return -EFAULT;
1706
1707 err = kcm_unattach_ioctl(sock, &info);
1708
1709 break;
1710 }
1711 case SIOCKCMCLONE: {
1712 struct kcm_clone info;
1713 struct file *file;
1714
1715 info.fd = get_unused_fd_flags(0);
1716 if (unlikely(info.fd < 0))
1717 return info.fd;
1718
1719 file = kcm_clone(sock);
1720 if (IS_ERR(file)) {
1721 put_unused_fd(info.fd);
1722 return PTR_ERR(file);
1723 }
1724 if (copy_to_user((void __user *)arg, &info,
1725 sizeof(info))) {
1726 put_unused_fd(info.fd);
1727 fput(file);
1728 return -EFAULT;
1729 }
1730 fd_install(info.fd, file);
1731 err = 0;
1732 break;
1733 }
1734 default:
1735 err = -ENOIOCTLCMD;
1736 break;
1737 }
1738
1739 return err;
1740 }
1741
1742 static void free_mux(struct rcu_head *rcu)
1743 {
1744 struct kcm_mux *mux = container_of(rcu,
1745 struct kcm_mux, rcu);
1746
1747 kmem_cache_free(kcm_muxp, mux);
1748 }
1749
1750 static void release_mux(struct kcm_mux *mux)
1751 {
1752 struct kcm_net *knet = mux->knet;
1753 struct kcm_psock *psock, *tmp_psock;
1754
1755 /* Release psocks */
1756 list_for_each_entry_safe(psock, tmp_psock,
1757 &mux->psocks, psock_list) {
1758 if (!WARN_ON(psock->unattaching))
1759 kcm_unattach(psock);
1760 }
1761
1762 if (WARN_ON(mux->psocks_cnt))
1763 return;
1764
1765 __skb_queue_purge(&mux->rx_hold_queue);
1766
1767 mutex_lock(&knet->mutex);
1768 aggregate_mux_stats(&mux->stats, &knet->aggregate_mux_stats);
1769 aggregate_psock_stats(&mux->aggregate_psock_stats,
1770 &knet->aggregate_psock_stats);
1771 aggregate_strp_stats(&mux->aggregate_strp_stats,
1772 &knet->aggregate_strp_stats);
1773 list_del_rcu(&mux->kcm_mux_list);
1774 knet->count--;
1775 mutex_unlock(&knet->mutex);
1776
1777 call_rcu(&mux->rcu, free_mux);
1778 }
1779
1780 static void kcm_done(struct kcm_sock *kcm)
1781 {
1782 struct kcm_mux *mux = kcm->mux;
1783 struct sock *sk = &kcm->sk;
1784 int socks_cnt;
1785
1786 spin_lock_bh(&mux->rx_lock);
1787 if (kcm->rx_psock) {
1788 /* Cleanup in unreserve_rx_kcm */
1789 WARN_ON(kcm->done);
1790 kcm->rx_disabled = 1;
1791 kcm->done = 1;
1792 spin_unlock_bh(&mux->rx_lock);
1793 return;
1794 }
1795
1796 if (kcm->rx_wait) {
1797 list_del(&kcm->wait_rx_list);
1798 kcm->rx_wait = false;
1799 }
1800 /* Move any pending receive messages to other kcm sockets */
1801 requeue_rx_msgs(mux, &sk->sk_receive_queue);
1802
1803 spin_unlock_bh(&mux->rx_lock);
1804
1805 if (WARN_ON(sk_rmem_alloc_get(sk)))
1806 return;
1807
1808 /* Detach from MUX */
1809 spin_lock_bh(&mux->lock);
1810
1811 list_del(&kcm->kcm_sock_list);
1812 mux->kcm_socks_cnt--;
1813 socks_cnt = mux->kcm_socks_cnt;
1814
1815 spin_unlock_bh(&mux->lock);
1816
1817 if (!socks_cnt) {
1818 /* We are done with the mux now. */
1819 release_mux(mux);
1820 }
1821
1822 WARN_ON(kcm->rx_wait);
1823
1824 sock_put(&kcm->sk);
1825 }
1826
1827 /* Called by kcm_release to close a KCM socket.
1828 * If this is the last KCM socket on the MUX, destroy the MUX.
1829 */
1830 static int kcm_release(struct socket *sock)
1831 {
1832 struct sock *sk = sock->sk;
1833 struct kcm_sock *kcm;
1834 struct kcm_mux *mux;
1835 struct kcm_psock *psock;
1836
1837 if (!sk)
1838 return 0;
1839
1840 kcm = kcm_sk(sk);
1841 mux = kcm->mux;
1842
1843 sock_orphan(sk);
1844 kfree_skb(kcm->seq_skb);
1845
1846 lock_sock(sk);
1847 /* Purge queue under lock to avoid race condition with tx_work trying
1848 * to act when queue is nonempty. If tx_work runs after this point
1849 * it will just return.
1850 */
1851 __skb_queue_purge(&sk->sk_write_queue);
1852
1853 /* Set tx_stopped. This is checked when psock is bound to a kcm and we
1854 * get a writespace callback. This prevents further work being queued
1855 * from the callback (unbinding the psock occurs after canceling work.
1856 */
1857 kcm->tx_stopped = 1;
1858
1859 release_sock(sk);
1860
1861 spin_lock_bh(&mux->lock);
1862 if (kcm->tx_wait) {
1863 /* Take of tx_wait list, after this point there should be no way
1864 * that a psock will be assigned to this kcm.
1865 */
1866 list_del(&kcm->wait_psock_list);
1867 kcm->tx_wait = false;
1868 }
1869 spin_unlock_bh(&mux->lock);
1870
1871 /* Cancel work. After this point there should be no outside references
1872 * to the kcm socket.
1873 */
1874 cancel_work_sync(&kcm->tx_work);
1875
1876 lock_sock(sk);
1877 psock = kcm->tx_psock;
1878 if (psock) {
1879 /* A psock was reserved, so we need to kill it since it
1880 * may already have some bytes queued from a message. We
1881 * need to do this after removing kcm from tx_wait list.
1882 */
1883 kcm_abort_tx_psock(psock, EPIPE, false);
1884 unreserve_psock(kcm);
1885 }
1886 release_sock(sk);
1887
1888 WARN_ON(kcm->tx_wait);
1889 WARN_ON(kcm->tx_psock);
1890
1891 sock->sk = NULL;
1892
1893 kcm_done(kcm);
1894
1895 return 0;
1896 }
1897
1898 static const struct proto_ops kcm_dgram_ops = {
1899 .family = PF_KCM,
1900 .owner = THIS_MODULE,
1901 .release = kcm_release,
1902 .bind = sock_no_bind,
1903 .connect = sock_no_connect,
1904 .socketpair = sock_no_socketpair,
1905 .accept = sock_no_accept,
1906 .getname = sock_no_getname,
1907 .poll = datagram_poll,
1908 .ioctl = kcm_ioctl,
1909 .listen = sock_no_listen,
1910 .shutdown = sock_no_shutdown,
1911 .setsockopt = kcm_setsockopt,
1912 .getsockopt = kcm_getsockopt,
1913 .sendmsg = kcm_sendmsg,
1914 .recvmsg = kcm_recvmsg,
1915 .mmap = sock_no_mmap,
1916 .sendpage = kcm_sendpage,
1917 };
1918
1919 static const struct proto_ops kcm_seqpacket_ops = {
1920 .family = PF_KCM,
1921 .owner = THIS_MODULE,
1922 .release = kcm_release,
1923 .bind = sock_no_bind,
1924 .connect = sock_no_connect,
1925 .socketpair = sock_no_socketpair,
1926 .accept = sock_no_accept,
1927 .getname = sock_no_getname,
1928 .poll = datagram_poll,
1929 .ioctl = kcm_ioctl,
1930 .listen = sock_no_listen,
1931 .shutdown = sock_no_shutdown,
1932 .setsockopt = kcm_setsockopt,
1933 .getsockopt = kcm_getsockopt,
1934 .sendmsg = kcm_sendmsg,
1935 .recvmsg = kcm_recvmsg,
1936 .mmap = sock_no_mmap,
1937 .sendpage = kcm_sendpage,
1938 .splice_read = kcm_splice_read,
1939 };
1940
1941 /* Create proto operation for kcm sockets */
1942 static int kcm_create(struct net *net, struct socket *sock,
1943 int protocol, int kern)
1944 {
1945 struct kcm_net *knet = net_generic(net, kcm_net_id);
1946 struct sock *sk;
1947 struct kcm_mux *mux;
1948
1949 switch (sock->type) {
1950 case SOCK_DGRAM:
1951 sock->ops = &kcm_dgram_ops;
1952 break;
1953 case SOCK_SEQPACKET:
1954 sock->ops = &kcm_seqpacket_ops;
1955 break;
1956 default:
1957 return -ESOCKTNOSUPPORT;
1958 }
1959
1960 if (protocol != KCMPROTO_CONNECTED)
1961 return -EPROTONOSUPPORT;
1962
1963 sk = sk_alloc(net, PF_KCM, GFP_KERNEL, &kcm_proto, kern);
1964 if (!sk)
1965 return -ENOMEM;
1966
1967 /* Allocate a kcm mux, shared between KCM sockets */
1968 mux = kmem_cache_zalloc(kcm_muxp, GFP_KERNEL);
1969 if (!mux) {
1970 sk_free(sk);
1971 return -ENOMEM;
1972 }
1973
1974 spin_lock_init(&mux->lock);
1975 spin_lock_init(&mux->rx_lock);
1976 INIT_LIST_HEAD(&mux->kcm_socks);
1977 INIT_LIST_HEAD(&mux->kcm_rx_waiters);
1978 INIT_LIST_HEAD(&mux->kcm_tx_waiters);
1979
1980 INIT_LIST_HEAD(&mux->psocks);
1981 INIT_LIST_HEAD(&mux->psocks_ready);
1982 INIT_LIST_HEAD(&mux->psocks_avail);
1983
1984 mux->knet = knet;
1985
1986 /* Add new MUX to list */
1987 mutex_lock(&knet->mutex);
1988 list_add_rcu(&mux->kcm_mux_list, &knet->mux_list);
1989 knet->count++;
1990 mutex_unlock(&knet->mutex);
1991
1992 skb_queue_head_init(&mux->rx_hold_queue);
1993
1994 /* Init KCM socket */
1995 sock_init_data(sock, sk);
1996 init_kcm_sock(kcm_sk(sk), mux);
1997
1998 return 0;
1999 }
2000
2001 static const struct net_proto_family kcm_family_ops = {
2002 .family = PF_KCM,
2003 .create = kcm_create,
2004 .owner = THIS_MODULE,
2005 };
2006
2007 static __net_init int kcm_init_net(struct net *net)
2008 {
2009 struct kcm_net *knet = net_generic(net, kcm_net_id);
2010
2011 INIT_LIST_HEAD_RCU(&knet->mux_list);
2012 mutex_init(&knet->mutex);
2013
2014 return 0;
2015 }
2016
2017 static __net_exit void kcm_exit_net(struct net *net)
2018 {
2019 struct kcm_net *knet = net_generic(net, kcm_net_id);
2020
2021 /* All KCM sockets should be closed at this point, which should mean
2022 * that all multiplexors and psocks have been destroyed.
2023 */
2024 WARN_ON(!list_empty(&knet->mux_list));
2025 }
2026
2027 static struct pernet_operations kcm_net_ops = {
2028 .init = kcm_init_net,
2029 .exit = kcm_exit_net,
2030 .id = &kcm_net_id,
2031 .size = sizeof(struct kcm_net),
2032 };
2033
2034 static int __init kcm_init(void)
2035 {
2036 int err = -ENOMEM;
2037
2038 kcm_muxp = kmem_cache_create("kcm_mux_cache",
2039 sizeof(struct kcm_mux), 0,
2040 SLAB_HWCACHE_ALIGN, NULL);
2041 if (!kcm_muxp)
2042 goto fail;
2043
2044 kcm_psockp = kmem_cache_create("kcm_psock_cache",
2045 sizeof(struct kcm_psock), 0,
2046 SLAB_HWCACHE_ALIGN, NULL);
2047 if (!kcm_psockp)
2048 goto fail;
2049
2050 kcm_wq = create_singlethread_workqueue("kkcmd");
2051 if (!kcm_wq)
2052 goto fail;
2053
2054 err = proto_register(&kcm_proto, 1);
2055 if (err)
2056 goto fail;
2057
2058 err = register_pernet_device(&kcm_net_ops);
2059 if (err)
2060 goto net_ops_fail;
2061
2062 err = sock_register(&kcm_family_ops);
2063 if (err)
2064 goto sock_register_fail;
2065
2066 err = kcm_proc_init();
2067 if (err)
2068 goto proc_init_fail;
2069
2070 return 0;
2071
2072 proc_init_fail:
2073 sock_unregister(PF_KCM);
2074
2075 sock_register_fail:
2076 unregister_pernet_device(&kcm_net_ops);
2077
2078 net_ops_fail:
2079 proto_unregister(&kcm_proto);
2080
2081 fail:
2082 kmem_cache_destroy(kcm_muxp);
2083 kmem_cache_destroy(kcm_psockp);
2084
2085 if (kcm_wq)
2086 destroy_workqueue(kcm_wq);
2087
2088 return err;
2089 }
2090
2091 static void __exit kcm_exit(void)
2092 {
2093 kcm_proc_exit();
2094 sock_unregister(PF_KCM);
2095 unregister_pernet_device(&kcm_net_ops);
2096 proto_unregister(&kcm_proto);
2097 destroy_workqueue(kcm_wq);
2098
2099 kmem_cache_destroy(kcm_muxp);
2100 kmem_cache_destroy(kcm_psockp);
2101 }
2102
2103 module_init(kcm_init);
2104 module_exit(kcm_exit);
2105
2106 MODULE_LICENSE("GPL");
2107 MODULE_ALIAS_NETPROTO(PF_KCM);
connection oriented routing