xref: /openbmc/linux/net/kcm/kcmsock.c (revision 86db9f28)
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);
2108