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
sk_msg_try_coalesce_ok(struct sk_msg * msg,int elem_first_coalesce)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
sk_msg_alloc(struct sock * sk,struct sk_msg * msg,int len,int elem_first_coalesce)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
sk_msg_clone(struct sock * sk,struct sk_msg * dst,struct sk_msg * src,u32 off,u32 len)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
sk_msg_return_zero(struct sock * sk,struct sk_msg * msg,int bytes)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
sk_msg_return(struct sock * sk,struct sk_msg * msg,int bytes)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
sk_msg_free_elem(struct sock * sk,struct sk_msg * msg,u32 i,bool charge)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
__sk_msg_free(struct sock * sk,struct sk_msg * msg,u32 i,bool charge)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
sk_msg_free_nocharge(struct sock * sk,struct sk_msg * msg)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
sk_msg_free(struct sock * sk,struct sk_msg * msg)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
__sk_msg_free_partial(struct sock * sk,struct sk_msg * msg,u32 bytes,bool charge)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
sk_msg_free_partial(struct sock * sk,struct sk_msg * msg,u32 bytes)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
sk_msg_free_partial_nocharge(struct sock * sk,struct sk_msg * msg,u32 bytes)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
sk_msg_trim(struct sock * sk,struct sk_msg * msg,int len)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
sk_msg_zerocopy_from_iter(struct sock * sk,struct iov_iter * from,struct sk_msg * msg,u32 bytes)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
sk_msg_memcopy_from_iter(struct sock * sk,struct iov_iter * from,struct sk_msg * msg,u32 bytes)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
sk_psock_skb_ingress(struct sk_psock * psock,struct sk_buff * skb)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
sk_psock_handle_skb(struct sk_psock * psock,struct sk_buff * skb,u32 off,u32 len,bool ingress)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
sk_psock_backlog(struct work_struct * work)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
sk_psock_init(struct sock * sk,int node)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
sk_psock_link_pop(struct sk_psock * psock)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
__sk_psock_purge_ingress_msg(struct sk_psock * psock)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
sk_psock_zap_ingress(struct sk_psock * psock)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
sk_psock_link_destroy(struct sk_psock * psock)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
sk_psock_destroy_deferred(struct work_struct * gc)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
sk_psock_destroy(struct rcu_head * rcu)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
sk_psock_drop(struct sock * sk,struct sk_psock * psock)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
sk_psock_map_verd(int verdict,bool redir)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
sk_psock_msg_verdict(struct sock * sk,struct sk_psock * psock,struct sk_msg * msg)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
sk_psock_bpf_run(struct sk_psock * psock,struct bpf_prog * prog,struct sk_buff * skb)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
sk_psock_from_strp(struct strparser * strp)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
sk_psock_verdict_apply(struct sk_psock * psock,struct sk_buff * skb,int verdict)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
sk_psock_strp_read(struct strparser * strp,struct sk_buff * skb)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
sk_psock_strp_read_done(struct strparser * strp,int err)759 static int sk_psock_strp_read_done(struct strparser *strp, int err)
760 {
761 return err;
762 }
763
sk_psock_strp_parse(struct strparser * strp,struct sk_buff * skb)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. */
sk_psock_strp_data_ready(struct sock * sk)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
sk_psock_write_space(struct sock * sk)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
sk_psock_init_strp(struct sock * sk,struct sk_psock * psock)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
sk_psock_start_strp(struct sock * sk,struct sk_psock * psock)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
sk_psock_stop_strp(struct sock * sk,struct sk_psock * psock)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