xref: /openbmc/linux/net/core/skmsg.c (revision f125e2d4)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2017 - 2018 Covalent IO, Inc. http://covalent.io */
3 
4 #include <linux/skmsg.h>
5 #include <linux/skbuff.h>
6 #include <linux/scatterlist.h>
7 
8 #include <net/sock.h>
9 #include <net/tcp.h>
10 
11 static bool sk_msg_try_coalesce_ok(struct sk_msg *msg, int elem_first_coalesce)
12 {
13 	if (msg->sg.end > msg->sg.start &&
14 	    elem_first_coalesce < msg->sg.end)
15 		return true;
16 
17 	if (msg->sg.end < msg->sg.start &&
18 	    (elem_first_coalesce > msg->sg.start ||
19 	     elem_first_coalesce < msg->sg.end))
20 		return true;
21 
22 	return false;
23 }
24 
25 int sk_msg_alloc(struct sock *sk, struct sk_msg *msg, int len,
26 		 int elem_first_coalesce)
27 {
28 	struct page_frag *pfrag = sk_page_frag(sk);
29 	int ret = 0;
30 
31 	len -= msg->sg.size;
32 	while (len > 0) {
33 		struct scatterlist *sge;
34 		u32 orig_offset;
35 		int use, i;
36 
37 		if (!sk_page_frag_refill(sk, pfrag))
38 			return -ENOMEM;
39 
40 		orig_offset = pfrag->offset;
41 		use = min_t(int, len, pfrag->size - orig_offset);
42 		if (!sk_wmem_schedule(sk, use))
43 			return -ENOMEM;
44 
45 		i = msg->sg.end;
46 		sk_msg_iter_var_prev(i);
47 		sge = &msg->sg.data[i];
48 
49 		if (sk_msg_try_coalesce_ok(msg, elem_first_coalesce) &&
50 		    sg_page(sge) == pfrag->page &&
51 		    sge->offset + sge->length == orig_offset) {
52 			sge->length += use;
53 		} else {
54 			if (sk_msg_full(msg)) {
55 				ret = -ENOSPC;
56 				break;
57 			}
58 
59 			sge = &msg->sg.data[msg->sg.end];
60 			sg_unmark_end(sge);
61 			sg_set_page(sge, pfrag->page, use, orig_offset);
62 			get_page(pfrag->page);
63 			sk_msg_iter_next(msg, end);
64 		}
65 
66 		sk_mem_charge(sk, use);
67 		msg->sg.size += use;
68 		pfrag->offset += use;
69 		len -= use;
70 	}
71 
72 	return ret;
73 }
74 EXPORT_SYMBOL_GPL(sk_msg_alloc);
75 
76 int sk_msg_clone(struct sock *sk, struct sk_msg *dst, struct sk_msg *src,
77 		 u32 off, u32 len)
78 {
79 	int i = src->sg.start;
80 	struct scatterlist *sge = sk_msg_elem(src, i);
81 	struct scatterlist *sgd = NULL;
82 	u32 sge_len, sge_off;
83 
84 	while (off) {
85 		if (sge->length > off)
86 			break;
87 		off -= sge->length;
88 		sk_msg_iter_var_next(i);
89 		if (i == src->sg.end && off)
90 			return -ENOSPC;
91 		sge = sk_msg_elem(src, i);
92 	}
93 
94 	while (len) {
95 		sge_len = sge->length - off;
96 		if (sge_len > len)
97 			sge_len = len;
98 
99 		if (dst->sg.end)
100 			sgd = sk_msg_elem(dst, dst->sg.end - 1);
101 
102 		if (sgd &&
103 		    (sg_page(sge) == sg_page(sgd)) &&
104 		    (sg_virt(sge) + off == sg_virt(sgd) + sgd->length)) {
105 			sgd->length += sge_len;
106 			dst->sg.size += sge_len;
107 		} else if (!sk_msg_full(dst)) {
108 			sge_off = sge->offset + off;
109 			sk_msg_page_add(dst, sg_page(sge), sge_len, sge_off);
110 		} else {
111 			return -ENOSPC;
112 		}
113 
114 		off = 0;
115 		len -= sge_len;
116 		sk_mem_charge(sk, sge_len);
117 		sk_msg_iter_var_next(i);
118 		if (i == src->sg.end && len)
119 			return -ENOSPC;
120 		sge = sk_msg_elem(src, i);
121 	}
122 
123 	return 0;
124 }
125 EXPORT_SYMBOL_GPL(sk_msg_clone);
126 
127 void sk_msg_return_zero(struct sock *sk, struct sk_msg *msg, int bytes)
128 {
129 	int i = msg->sg.start;
130 
131 	do {
132 		struct scatterlist *sge = sk_msg_elem(msg, i);
133 
134 		if (bytes < sge->length) {
135 			sge->length -= bytes;
136 			sge->offset += bytes;
137 			sk_mem_uncharge(sk, bytes);
138 			break;
139 		}
140 
141 		sk_mem_uncharge(sk, sge->length);
142 		bytes -= sge->length;
143 		sge->length = 0;
144 		sge->offset = 0;
145 		sk_msg_iter_var_next(i);
146 	} while (bytes && i != msg->sg.end);
147 	msg->sg.start = i;
148 }
149 EXPORT_SYMBOL_GPL(sk_msg_return_zero);
150 
151 void sk_msg_return(struct sock *sk, struct sk_msg *msg, int bytes)
152 {
153 	int i = msg->sg.start;
154 
155 	do {
156 		struct scatterlist *sge = &msg->sg.data[i];
157 		int uncharge = (bytes < sge->length) ? bytes : sge->length;
158 
159 		sk_mem_uncharge(sk, uncharge);
160 		bytes -= uncharge;
161 		sk_msg_iter_var_next(i);
162 	} while (i != msg->sg.end);
163 }
164 EXPORT_SYMBOL_GPL(sk_msg_return);
165 
166 static int sk_msg_free_elem(struct sock *sk, struct sk_msg *msg, u32 i,
167 			    bool charge)
168 {
169 	struct scatterlist *sge = sk_msg_elem(msg, i);
170 	u32 len = sge->length;
171 
172 	if (charge)
173 		sk_mem_uncharge(sk, len);
174 	if (!msg->skb)
175 		put_page(sg_page(sge));
176 	memset(sge, 0, sizeof(*sge));
177 	return len;
178 }
179 
180 static int __sk_msg_free(struct sock *sk, struct sk_msg *msg, u32 i,
181 			 bool charge)
182 {
183 	struct scatterlist *sge = sk_msg_elem(msg, i);
184 	int freed = 0;
185 
186 	while (msg->sg.size) {
187 		msg->sg.size -= sge->length;
188 		freed += sk_msg_free_elem(sk, msg, i, charge);
189 		sk_msg_iter_var_next(i);
190 		sk_msg_check_to_free(msg, i, msg->sg.size);
191 		sge = sk_msg_elem(msg, i);
192 	}
193 	consume_skb(msg->skb);
194 	sk_msg_init(msg);
195 	return freed;
196 }
197 
198 int sk_msg_free_nocharge(struct sock *sk, struct sk_msg *msg)
199 {
200 	return __sk_msg_free(sk, msg, msg->sg.start, false);
201 }
202 EXPORT_SYMBOL_GPL(sk_msg_free_nocharge);
203 
204 int sk_msg_free(struct sock *sk, struct sk_msg *msg)
205 {
206 	return __sk_msg_free(sk, msg, msg->sg.start, true);
207 }
208 EXPORT_SYMBOL_GPL(sk_msg_free);
209 
210 static void __sk_msg_free_partial(struct sock *sk, struct sk_msg *msg,
211 				  u32 bytes, bool charge)
212 {
213 	struct scatterlist *sge;
214 	u32 i = msg->sg.start;
215 
216 	while (bytes) {
217 		sge = sk_msg_elem(msg, i);
218 		if (!sge->length)
219 			break;
220 		if (bytes < sge->length) {
221 			if (charge)
222 				sk_mem_uncharge(sk, bytes);
223 			sge->length -= bytes;
224 			sge->offset += bytes;
225 			msg->sg.size -= bytes;
226 			break;
227 		}
228 
229 		msg->sg.size -= sge->length;
230 		bytes -= sge->length;
231 		sk_msg_free_elem(sk, msg, i, charge);
232 		sk_msg_iter_var_next(i);
233 		sk_msg_check_to_free(msg, i, bytes);
234 	}
235 	msg->sg.start = i;
236 }
237 
238 void sk_msg_free_partial(struct sock *sk, struct sk_msg *msg, u32 bytes)
239 {
240 	__sk_msg_free_partial(sk, msg, bytes, true);
241 }
242 EXPORT_SYMBOL_GPL(sk_msg_free_partial);
243 
244 void sk_msg_free_partial_nocharge(struct sock *sk, struct sk_msg *msg,
245 				  u32 bytes)
246 {
247 	__sk_msg_free_partial(sk, msg, bytes, false);
248 }
249 
250 void sk_msg_trim(struct sock *sk, struct sk_msg *msg, int len)
251 {
252 	int trim = msg->sg.size - len;
253 	u32 i = msg->sg.end;
254 
255 	if (trim <= 0) {
256 		WARN_ON(trim < 0);
257 		return;
258 	}
259 
260 	sk_msg_iter_var_prev(i);
261 	msg->sg.size = len;
262 	while (msg->sg.data[i].length &&
263 	       trim >= msg->sg.data[i].length) {
264 		trim -= msg->sg.data[i].length;
265 		sk_msg_free_elem(sk, msg, i, true);
266 		sk_msg_iter_var_prev(i);
267 		if (!trim)
268 			goto out;
269 	}
270 
271 	msg->sg.data[i].length -= trim;
272 	sk_mem_uncharge(sk, trim);
273 	/* Adjust copybreak if it falls into the trimmed part of last buf */
274 	if (msg->sg.curr == i && msg->sg.copybreak > msg->sg.data[i].length)
275 		msg->sg.copybreak = msg->sg.data[i].length;
276 out:
277 	sk_msg_iter_var_next(i);
278 	msg->sg.end = i;
279 
280 	/* If we trim data a full sg elem before curr pointer update
281 	 * copybreak and current so that any future copy operations
282 	 * start at new copy location.
283 	 * However trimed data that has not yet been used in a copy op
284 	 * does not require an update.
285 	 */
286 	if (!msg->sg.size) {
287 		msg->sg.curr = msg->sg.start;
288 		msg->sg.copybreak = 0;
289 	} else if (sk_msg_iter_dist(msg->sg.start, msg->sg.curr) >=
290 		   sk_msg_iter_dist(msg->sg.start, msg->sg.end)) {
291 		sk_msg_iter_var_prev(i);
292 		msg->sg.curr = i;
293 		msg->sg.copybreak = msg->sg.data[i].length;
294 	}
295 }
296 EXPORT_SYMBOL_GPL(sk_msg_trim);
297 
298 int sk_msg_zerocopy_from_iter(struct sock *sk, struct iov_iter *from,
299 			      struct sk_msg *msg, u32 bytes)
300 {
301 	int i, maxpages, ret = 0, num_elems = sk_msg_elem_used(msg);
302 	const int to_max_pages = MAX_MSG_FRAGS;
303 	struct page *pages[MAX_MSG_FRAGS];
304 	ssize_t orig, copied, use, offset;
305 
306 	orig = msg->sg.size;
307 	while (bytes > 0) {
308 		i = 0;
309 		maxpages = to_max_pages - num_elems;
310 		if (maxpages == 0) {
311 			ret = -EFAULT;
312 			goto out;
313 		}
314 
315 		copied = iov_iter_get_pages(from, pages, bytes, maxpages,
316 					    &offset);
317 		if (copied <= 0) {
318 			ret = -EFAULT;
319 			goto out;
320 		}
321 
322 		iov_iter_advance(from, copied);
323 		bytes -= copied;
324 		msg->sg.size += copied;
325 
326 		while (copied) {
327 			use = min_t(int, copied, PAGE_SIZE - offset);
328 			sg_set_page(&msg->sg.data[msg->sg.end],
329 				    pages[i], use, offset);
330 			sg_unmark_end(&msg->sg.data[msg->sg.end]);
331 			sk_mem_charge(sk, use);
332 
333 			offset = 0;
334 			copied -= use;
335 			sk_msg_iter_next(msg, end);
336 			num_elems++;
337 			i++;
338 		}
339 		/* When zerocopy is mixed with sk_msg_*copy* operations we
340 		 * may have a copybreak set in this case clear and prefer
341 		 * zerocopy remainder when possible.
342 		 */
343 		msg->sg.copybreak = 0;
344 		msg->sg.curr = msg->sg.end;
345 	}
346 out:
347 	/* Revert iov_iter updates, msg will need to use 'trim' later if it
348 	 * also needs to be cleared.
349 	 */
350 	if (ret)
351 		iov_iter_revert(from, msg->sg.size - orig);
352 	return ret;
353 }
354 EXPORT_SYMBOL_GPL(sk_msg_zerocopy_from_iter);
355 
356 int sk_msg_memcopy_from_iter(struct sock *sk, struct iov_iter *from,
357 			     struct sk_msg *msg, u32 bytes)
358 {
359 	int ret = -ENOSPC, i = msg->sg.curr;
360 	struct scatterlist *sge;
361 	u32 copy, buf_size;
362 	void *to;
363 
364 	do {
365 		sge = sk_msg_elem(msg, i);
366 		/* This is possible if a trim operation shrunk the buffer */
367 		if (msg->sg.copybreak >= sge->length) {
368 			msg->sg.copybreak = 0;
369 			sk_msg_iter_var_next(i);
370 			if (i == msg->sg.end)
371 				break;
372 			sge = sk_msg_elem(msg, i);
373 		}
374 
375 		buf_size = sge->length - msg->sg.copybreak;
376 		copy = (buf_size > bytes) ? bytes : buf_size;
377 		to = sg_virt(sge) + msg->sg.copybreak;
378 		msg->sg.copybreak += copy;
379 		if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY)
380 			ret = copy_from_iter_nocache(to, copy, from);
381 		else
382 			ret = copy_from_iter(to, copy, from);
383 		if (ret != copy) {
384 			ret = -EFAULT;
385 			goto out;
386 		}
387 		bytes -= copy;
388 		if (!bytes)
389 			break;
390 		msg->sg.copybreak = 0;
391 		sk_msg_iter_var_next(i);
392 	} while (i != msg->sg.end);
393 out:
394 	msg->sg.curr = i;
395 	return ret;
396 }
397 EXPORT_SYMBOL_GPL(sk_msg_memcopy_from_iter);
398 
399 static int sk_psock_skb_ingress(struct sk_psock *psock, struct sk_buff *skb)
400 {
401 	struct sock *sk = psock->sk;
402 	int copied = 0, num_sge;
403 	struct sk_msg *msg;
404 
405 	msg = kzalloc(sizeof(*msg), __GFP_NOWARN | GFP_ATOMIC);
406 	if (unlikely(!msg))
407 		return -EAGAIN;
408 	if (!sk_rmem_schedule(sk, skb, skb->len)) {
409 		kfree(msg);
410 		return -EAGAIN;
411 	}
412 
413 	sk_msg_init(msg);
414 	num_sge = skb_to_sgvec(skb, msg->sg.data, 0, skb->len);
415 	if (unlikely(num_sge < 0)) {
416 		kfree(msg);
417 		return num_sge;
418 	}
419 
420 	sk_mem_charge(sk, skb->len);
421 	copied = skb->len;
422 	msg->sg.start = 0;
423 	msg->sg.size = copied;
424 	msg->sg.end = num_sge;
425 	msg->skb = skb;
426 
427 	sk_psock_queue_msg(psock, msg);
428 	sk_psock_data_ready(sk, psock);
429 	return copied;
430 }
431 
432 static int sk_psock_handle_skb(struct sk_psock *psock, struct sk_buff *skb,
433 			       u32 off, u32 len, bool ingress)
434 {
435 	if (ingress)
436 		return sk_psock_skb_ingress(psock, skb);
437 	else
438 		return skb_send_sock_locked(psock->sk, skb, off, len);
439 }
440 
441 static void sk_psock_backlog(struct work_struct *work)
442 {
443 	struct sk_psock *psock = container_of(work, struct sk_psock, work);
444 	struct sk_psock_work_state *state = &psock->work_state;
445 	struct sk_buff *skb;
446 	bool ingress;
447 	u32 len, off;
448 	int ret;
449 
450 	/* Lock sock to avoid losing sk_socket during loop. */
451 	lock_sock(psock->sk);
452 	if (state->skb) {
453 		skb = state->skb;
454 		len = state->len;
455 		off = state->off;
456 		state->skb = NULL;
457 		goto start;
458 	}
459 
460 	while ((skb = skb_dequeue(&psock->ingress_skb))) {
461 		len = skb->len;
462 		off = 0;
463 start:
464 		ingress = tcp_skb_bpf_ingress(skb);
465 		do {
466 			ret = -EIO;
467 			if (likely(psock->sk->sk_socket))
468 				ret = sk_psock_handle_skb(psock, skb, off,
469 							  len, ingress);
470 			if (ret <= 0) {
471 				if (ret == -EAGAIN) {
472 					state->skb = skb;
473 					state->len = len;
474 					state->off = off;
475 					goto end;
476 				}
477 				/* Hard errors break pipe and stop xmit. */
478 				sk_psock_report_error(psock, ret ? -ret : EPIPE);
479 				sk_psock_clear_state(psock, SK_PSOCK_TX_ENABLED);
480 				kfree_skb(skb);
481 				goto end;
482 			}
483 			off += ret;
484 			len -= ret;
485 		} while (len);
486 
487 		if (!ingress)
488 			kfree_skb(skb);
489 	}
490 end:
491 	release_sock(psock->sk);
492 }
493 
494 struct sk_psock *sk_psock_init(struct sock *sk, int node)
495 {
496 	struct sk_psock *psock = kzalloc_node(sizeof(*psock),
497 					      GFP_ATOMIC | __GFP_NOWARN,
498 					      node);
499 	if (!psock)
500 		return NULL;
501 
502 	psock->sk = sk;
503 	psock->eval =  __SK_NONE;
504 
505 	INIT_LIST_HEAD(&psock->link);
506 	spin_lock_init(&psock->link_lock);
507 
508 	INIT_WORK(&psock->work, sk_psock_backlog);
509 	INIT_LIST_HEAD(&psock->ingress_msg);
510 	skb_queue_head_init(&psock->ingress_skb);
511 
512 	sk_psock_set_state(psock, SK_PSOCK_TX_ENABLED);
513 	refcount_set(&psock->refcnt, 1);
514 
515 	rcu_assign_sk_user_data(sk, psock);
516 	sock_hold(sk);
517 
518 	return psock;
519 }
520 EXPORT_SYMBOL_GPL(sk_psock_init);
521 
522 struct sk_psock_link *sk_psock_link_pop(struct sk_psock *psock)
523 {
524 	struct sk_psock_link *link;
525 
526 	spin_lock_bh(&psock->link_lock);
527 	link = list_first_entry_or_null(&psock->link, struct sk_psock_link,
528 					list);
529 	if (link)
530 		list_del(&link->list);
531 	spin_unlock_bh(&psock->link_lock);
532 	return link;
533 }
534 
535 void __sk_psock_purge_ingress_msg(struct sk_psock *psock)
536 {
537 	struct sk_msg *msg, *tmp;
538 
539 	list_for_each_entry_safe(msg, tmp, &psock->ingress_msg, list) {
540 		list_del(&msg->list);
541 		sk_msg_free(psock->sk, msg);
542 		kfree(msg);
543 	}
544 }
545 
546 static void sk_psock_zap_ingress(struct sk_psock *psock)
547 {
548 	__skb_queue_purge(&psock->ingress_skb);
549 	__sk_psock_purge_ingress_msg(psock);
550 }
551 
552 static void sk_psock_link_destroy(struct sk_psock *psock)
553 {
554 	struct sk_psock_link *link, *tmp;
555 
556 	list_for_each_entry_safe(link, tmp, &psock->link, list) {
557 		list_del(&link->list);
558 		sk_psock_free_link(link);
559 	}
560 }
561 
562 static void sk_psock_destroy_deferred(struct work_struct *gc)
563 {
564 	struct sk_psock *psock = container_of(gc, struct sk_psock, gc);
565 
566 	/* No sk_callback_lock since already detached. */
567 
568 	/* Parser has been stopped */
569 	if (psock->progs.skb_parser)
570 		strp_done(&psock->parser.strp);
571 
572 	cancel_work_sync(&psock->work);
573 
574 	psock_progs_drop(&psock->progs);
575 
576 	sk_psock_link_destroy(psock);
577 	sk_psock_cork_free(psock);
578 	sk_psock_zap_ingress(psock);
579 
580 	if (psock->sk_redir)
581 		sock_put(psock->sk_redir);
582 	sock_put(psock->sk);
583 	kfree(psock);
584 }
585 
586 void sk_psock_destroy(struct rcu_head *rcu)
587 {
588 	struct sk_psock *psock = container_of(rcu, struct sk_psock, rcu);
589 
590 	INIT_WORK(&psock->gc, sk_psock_destroy_deferred);
591 	schedule_work(&psock->gc);
592 }
593 EXPORT_SYMBOL_GPL(sk_psock_destroy);
594 
595 void sk_psock_drop(struct sock *sk, struct sk_psock *psock)
596 {
597 	sk_psock_cork_free(psock);
598 	sk_psock_zap_ingress(psock);
599 
600 	write_lock_bh(&sk->sk_callback_lock);
601 	sk_psock_restore_proto(sk, psock);
602 	rcu_assign_sk_user_data(sk, NULL);
603 	if (psock->progs.skb_parser)
604 		sk_psock_stop_strp(sk, psock);
605 	write_unlock_bh(&sk->sk_callback_lock);
606 	sk_psock_clear_state(psock, SK_PSOCK_TX_ENABLED);
607 
608 	call_rcu(&psock->rcu, sk_psock_destroy);
609 }
610 EXPORT_SYMBOL_GPL(sk_psock_drop);
611 
612 static int sk_psock_map_verd(int verdict, bool redir)
613 {
614 	switch (verdict) {
615 	case SK_PASS:
616 		return redir ? __SK_REDIRECT : __SK_PASS;
617 	case SK_DROP:
618 	default:
619 		break;
620 	}
621 
622 	return __SK_DROP;
623 }
624 
625 int sk_psock_msg_verdict(struct sock *sk, struct sk_psock *psock,
626 			 struct sk_msg *msg)
627 {
628 	struct bpf_prog *prog;
629 	int ret;
630 
631 	preempt_disable();
632 	rcu_read_lock();
633 	prog = READ_ONCE(psock->progs.msg_parser);
634 	if (unlikely(!prog)) {
635 		ret = __SK_PASS;
636 		goto out;
637 	}
638 
639 	sk_msg_compute_data_pointers(msg);
640 	msg->sk = sk;
641 	ret = BPF_PROG_RUN(prog, msg);
642 	ret = sk_psock_map_verd(ret, msg->sk_redir);
643 	psock->apply_bytes = msg->apply_bytes;
644 	if (ret == __SK_REDIRECT) {
645 		if (psock->sk_redir)
646 			sock_put(psock->sk_redir);
647 		psock->sk_redir = msg->sk_redir;
648 		if (!psock->sk_redir) {
649 			ret = __SK_DROP;
650 			goto out;
651 		}
652 		sock_hold(psock->sk_redir);
653 	}
654 out:
655 	rcu_read_unlock();
656 	preempt_enable();
657 	return ret;
658 }
659 EXPORT_SYMBOL_GPL(sk_psock_msg_verdict);
660 
661 static int sk_psock_bpf_run(struct sk_psock *psock, struct bpf_prog *prog,
662 			    struct sk_buff *skb)
663 {
664 	int ret;
665 
666 	skb->sk = psock->sk;
667 	bpf_compute_data_end_sk_skb(skb);
668 	preempt_disable();
669 	ret = BPF_PROG_RUN(prog, skb);
670 	preempt_enable();
671 	/* strparser clones the skb before handing it to a upper layer,
672 	 * meaning skb_orphan has been called. We NULL sk on the way out
673 	 * to ensure we don't trigger a BUG_ON() in skb/sk operations
674 	 * later and because we are not charging the memory of this skb
675 	 * to any socket yet.
676 	 */
677 	skb->sk = NULL;
678 	return ret;
679 }
680 
681 static struct sk_psock *sk_psock_from_strp(struct strparser *strp)
682 {
683 	struct sk_psock_parser *parser;
684 
685 	parser = container_of(strp, struct sk_psock_parser, strp);
686 	return container_of(parser, struct sk_psock, parser);
687 }
688 
689 static void sk_psock_verdict_apply(struct sk_psock *psock,
690 				   struct sk_buff *skb, int verdict)
691 {
692 	struct sk_psock *psock_other;
693 	struct sock *sk_other;
694 	bool ingress;
695 
696 	switch (verdict) {
697 	case __SK_PASS:
698 		sk_other = psock->sk;
699 		if (sock_flag(sk_other, SOCK_DEAD) ||
700 		    !sk_psock_test_state(psock, SK_PSOCK_TX_ENABLED)) {
701 			goto out_free;
702 		}
703 		if (atomic_read(&sk_other->sk_rmem_alloc) <=
704 		    sk_other->sk_rcvbuf) {
705 			struct tcp_skb_cb *tcp = TCP_SKB_CB(skb);
706 
707 			tcp->bpf.flags |= BPF_F_INGRESS;
708 			skb_queue_tail(&psock->ingress_skb, skb);
709 			schedule_work(&psock->work);
710 			break;
711 		}
712 		goto out_free;
713 	case __SK_REDIRECT:
714 		sk_other = tcp_skb_bpf_redirect_fetch(skb);
715 		if (unlikely(!sk_other))
716 			goto out_free;
717 		psock_other = sk_psock(sk_other);
718 		if (!psock_other || sock_flag(sk_other, SOCK_DEAD) ||
719 		    !sk_psock_test_state(psock_other, SK_PSOCK_TX_ENABLED))
720 			goto out_free;
721 		ingress = tcp_skb_bpf_ingress(skb);
722 		if ((!ingress && sock_writeable(sk_other)) ||
723 		    (ingress &&
724 		     atomic_read(&sk_other->sk_rmem_alloc) <=
725 		     sk_other->sk_rcvbuf)) {
726 			if (!ingress)
727 				skb_set_owner_w(skb, sk_other);
728 			skb_queue_tail(&psock_other->ingress_skb, skb);
729 			schedule_work(&psock_other->work);
730 			break;
731 		}
732 		/* fall-through */
733 	case __SK_DROP:
734 		/* fall-through */
735 	default:
736 out_free:
737 		kfree_skb(skb);
738 	}
739 }
740 
741 static void sk_psock_strp_read(struct strparser *strp, struct sk_buff *skb)
742 {
743 	struct sk_psock *psock = sk_psock_from_strp(strp);
744 	struct bpf_prog *prog;
745 	int ret = __SK_DROP;
746 
747 	rcu_read_lock();
748 	prog = READ_ONCE(psock->progs.skb_verdict);
749 	if (likely(prog)) {
750 		skb_orphan(skb);
751 		tcp_skb_bpf_redirect_clear(skb);
752 		ret = sk_psock_bpf_run(psock, prog, skb);
753 		ret = sk_psock_map_verd(ret, tcp_skb_bpf_redirect_fetch(skb));
754 	}
755 	rcu_read_unlock();
756 	sk_psock_verdict_apply(psock, skb, ret);
757 }
758 
759 static int sk_psock_strp_read_done(struct strparser *strp, int err)
760 {
761 	return err;
762 }
763 
764 static int sk_psock_strp_parse(struct strparser *strp, struct sk_buff *skb)
765 {
766 	struct sk_psock *psock = sk_psock_from_strp(strp);
767 	struct bpf_prog *prog;
768 	int ret = skb->len;
769 
770 	rcu_read_lock();
771 	prog = READ_ONCE(psock->progs.skb_parser);
772 	if (likely(prog))
773 		ret = sk_psock_bpf_run(psock, prog, skb);
774 	rcu_read_unlock();
775 	return ret;
776 }
777 
778 /* Called with socket lock held. */
779 static void sk_psock_strp_data_ready(struct sock *sk)
780 {
781 	struct sk_psock *psock;
782 
783 	rcu_read_lock();
784 	psock = sk_psock(sk);
785 	if (likely(psock)) {
786 		write_lock_bh(&sk->sk_callback_lock);
787 		strp_data_ready(&psock->parser.strp);
788 		write_unlock_bh(&sk->sk_callback_lock);
789 	}
790 	rcu_read_unlock();
791 }
792 
793 static void sk_psock_write_space(struct sock *sk)
794 {
795 	struct sk_psock *psock;
796 	void (*write_space)(struct sock *sk) = NULL;
797 
798 	rcu_read_lock();
799 	psock = sk_psock(sk);
800 	if (likely(psock)) {
801 		if (sk_psock_test_state(psock, SK_PSOCK_TX_ENABLED))
802 			schedule_work(&psock->work);
803 		write_space = psock->saved_write_space;
804 	}
805 	rcu_read_unlock();
806 	if (write_space)
807 		write_space(sk);
808 }
809 
810 int sk_psock_init_strp(struct sock *sk, struct sk_psock *psock)
811 {
812 	static const struct strp_callbacks cb = {
813 		.rcv_msg	= sk_psock_strp_read,
814 		.read_sock_done	= sk_psock_strp_read_done,
815 		.parse_msg	= sk_psock_strp_parse,
816 	};
817 
818 	psock->parser.enabled = false;
819 	return strp_init(&psock->parser.strp, sk, &cb);
820 }
821 
822 void sk_psock_start_strp(struct sock *sk, struct sk_psock *psock)
823 {
824 	struct sk_psock_parser *parser = &psock->parser;
825 
826 	if (parser->enabled)
827 		return;
828 
829 	parser->saved_data_ready = sk->sk_data_ready;
830 	sk->sk_data_ready = sk_psock_strp_data_ready;
831 	sk->sk_write_space = sk_psock_write_space;
832 	parser->enabled = true;
833 }
834 
835 void sk_psock_stop_strp(struct sock *sk, struct sk_psock *psock)
836 {
837 	struct sk_psock_parser *parser = &psock->parser;
838 
839 	if (!parser->enabled)
840 		return;
841 
842 	sk->sk_data_ready = parser->saved_data_ready;
843 	parser->saved_data_ready = NULL;
844 	strp_stop(&parser->strp);
845 	parser->enabled = false;
846 }
847