xref: /linux/net/tls/tls_device.c (revision a3a02a52bcfcbcc4a637d4b68bf1bc391c9fad02)
1 /* Copyright (c) 2018, Mellanox Technologies All rights reserved.
2  *
3  * This software is available to you under a choice of one of two
4  * licenses.  You may choose to be licensed under the terms of the GNU
5  * General Public License (GPL) Version 2, available from the file
6  * COPYING in the main directory of this source tree, or the
7  * OpenIB.org BSD license below:
8  *
9  *     Redistribution and use in source and binary forms, with or
10  *     without modification, are permitted provided that the following
11  *     conditions are met:
12  *
13  *      - Redistributions of source code must retain the above
14  *        copyright notice, this list of conditions and the following
15  *        disclaimer.
16  *
17  *      - Redistributions in binary form must reproduce the above
18  *        copyright notice, this list of conditions and the following
19  *        disclaimer in the documentation and/or other materials
20  *        provided with the distribution.
21  *
22  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
23  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
24  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
25  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
26  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
27  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
28  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
29  * SOFTWARE.
30  */
31 
32 #include <crypto/aead.h>
33 #include <linux/highmem.h>
34 #include <linux/module.h>
35 #include <linux/netdevice.h>
36 #include <net/dst.h>
37 #include <net/inet_connection_sock.h>
38 #include <net/tcp.h>
39 #include <net/tls.h>
40 #include <linux/skbuff_ref.h>
41 
42 #include "tls.h"
43 #include "trace.h"
44 
45 /* device_offload_lock is used to synchronize tls_dev_add
46  * against NETDEV_DOWN notifications.
47  */
48 static DECLARE_RWSEM(device_offload_lock);
49 
50 static struct workqueue_struct *destruct_wq __read_mostly;
51 
52 static LIST_HEAD(tls_device_list);
53 static LIST_HEAD(tls_device_down_list);
54 static DEFINE_SPINLOCK(tls_device_lock);
55 
56 static struct page *dummy_page;
57 
58 static void tls_device_free_ctx(struct tls_context *ctx)
59 {
60 	if (ctx->tx_conf == TLS_HW)
61 		kfree(tls_offload_ctx_tx(ctx));
62 
63 	if (ctx->rx_conf == TLS_HW)
64 		kfree(tls_offload_ctx_rx(ctx));
65 
66 	tls_ctx_free(NULL, ctx);
67 }
68 
69 static void tls_device_tx_del_task(struct work_struct *work)
70 {
71 	struct tls_offload_context_tx *offload_ctx =
72 		container_of(work, struct tls_offload_context_tx, destruct_work);
73 	struct tls_context *ctx = offload_ctx->ctx;
74 	struct net_device *netdev;
75 
76 	/* Safe, because this is the destroy flow, refcount is 0, so
77 	 * tls_device_down can't store this field in parallel.
78 	 */
79 	netdev = rcu_dereference_protected(ctx->netdev,
80 					   !refcount_read(&ctx->refcount));
81 
82 	netdev->tlsdev_ops->tls_dev_del(netdev, ctx, TLS_OFFLOAD_CTX_DIR_TX);
83 	dev_put(netdev);
84 	ctx->netdev = NULL;
85 	tls_device_free_ctx(ctx);
86 }
87 
88 static void tls_device_queue_ctx_destruction(struct tls_context *ctx)
89 {
90 	struct net_device *netdev;
91 	unsigned long flags;
92 	bool async_cleanup;
93 
94 	spin_lock_irqsave(&tls_device_lock, flags);
95 	if (unlikely(!refcount_dec_and_test(&ctx->refcount))) {
96 		spin_unlock_irqrestore(&tls_device_lock, flags);
97 		return;
98 	}
99 
100 	list_del(&ctx->list); /* Remove from tls_device_list / tls_device_down_list */
101 
102 	/* Safe, because this is the destroy flow, refcount is 0, so
103 	 * tls_device_down can't store this field in parallel.
104 	 */
105 	netdev = rcu_dereference_protected(ctx->netdev,
106 					   !refcount_read(&ctx->refcount));
107 
108 	async_cleanup = netdev && ctx->tx_conf == TLS_HW;
109 	if (async_cleanup) {
110 		struct tls_offload_context_tx *offload_ctx = tls_offload_ctx_tx(ctx);
111 
112 		/* queue_work inside the spinlock
113 		 * to make sure tls_device_down waits for that work.
114 		 */
115 		queue_work(destruct_wq, &offload_ctx->destruct_work);
116 	}
117 	spin_unlock_irqrestore(&tls_device_lock, flags);
118 
119 	if (!async_cleanup)
120 		tls_device_free_ctx(ctx);
121 }
122 
123 /* We assume that the socket is already connected */
124 static struct net_device *get_netdev_for_sock(struct sock *sk)
125 {
126 	struct dst_entry *dst = sk_dst_get(sk);
127 	struct net_device *netdev = NULL;
128 
129 	if (likely(dst)) {
130 		netdev = netdev_sk_get_lowest_dev(dst->dev, sk);
131 		dev_hold(netdev);
132 	}
133 
134 	dst_release(dst);
135 
136 	return netdev;
137 }
138 
139 static void destroy_record(struct tls_record_info *record)
140 {
141 	int i;
142 
143 	for (i = 0; i < record->num_frags; i++)
144 		__skb_frag_unref(&record->frags[i], false);
145 	kfree(record);
146 }
147 
148 static void delete_all_records(struct tls_offload_context_tx *offload_ctx)
149 {
150 	struct tls_record_info *info, *temp;
151 
152 	list_for_each_entry_safe(info, temp, &offload_ctx->records_list, list) {
153 		list_del(&info->list);
154 		destroy_record(info);
155 	}
156 
157 	offload_ctx->retransmit_hint = NULL;
158 }
159 
160 static void tls_icsk_clean_acked(struct sock *sk, u32 acked_seq)
161 {
162 	struct tls_context *tls_ctx = tls_get_ctx(sk);
163 	struct tls_record_info *info, *temp;
164 	struct tls_offload_context_tx *ctx;
165 	u64 deleted_records = 0;
166 	unsigned long flags;
167 
168 	if (!tls_ctx)
169 		return;
170 
171 	ctx = tls_offload_ctx_tx(tls_ctx);
172 
173 	spin_lock_irqsave(&ctx->lock, flags);
174 	info = ctx->retransmit_hint;
175 	if (info && !before(acked_seq, info->end_seq))
176 		ctx->retransmit_hint = NULL;
177 
178 	list_for_each_entry_safe(info, temp, &ctx->records_list, list) {
179 		if (before(acked_seq, info->end_seq))
180 			break;
181 		list_del(&info->list);
182 
183 		destroy_record(info);
184 		deleted_records++;
185 	}
186 
187 	ctx->unacked_record_sn += deleted_records;
188 	spin_unlock_irqrestore(&ctx->lock, flags);
189 }
190 
191 /* At this point, there should be no references on this
192  * socket and no in-flight SKBs associated with this
193  * socket, so it is safe to free all the resources.
194  */
195 void tls_device_sk_destruct(struct sock *sk)
196 {
197 	struct tls_context *tls_ctx = tls_get_ctx(sk);
198 	struct tls_offload_context_tx *ctx = tls_offload_ctx_tx(tls_ctx);
199 
200 	tls_ctx->sk_destruct(sk);
201 
202 	if (tls_ctx->tx_conf == TLS_HW) {
203 		if (ctx->open_record)
204 			destroy_record(ctx->open_record);
205 		delete_all_records(ctx);
206 		crypto_free_aead(ctx->aead_send);
207 		clean_acked_data_disable(inet_csk(sk));
208 	}
209 
210 	tls_device_queue_ctx_destruction(tls_ctx);
211 }
212 EXPORT_SYMBOL_GPL(tls_device_sk_destruct);
213 
214 void tls_device_free_resources_tx(struct sock *sk)
215 {
216 	struct tls_context *tls_ctx = tls_get_ctx(sk);
217 
218 	tls_free_partial_record(sk, tls_ctx);
219 }
220 
221 void tls_offload_tx_resync_request(struct sock *sk, u32 got_seq, u32 exp_seq)
222 {
223 	struct tls_context *tls_ctx = tls_get_ctx(sk);
224 
225 	trace_tls_device_tx_resync_req(sk, got_seq, exp_seq);
226 	WARN_ON(test_and_set_bit(TLS_TX_SYNC_SCHED, &tls_ctx->flags));
227 }
228 EXPORT_SYMBOL_GPL(tls_offload_tx_resync_request);
229 
230 static void tls_device_resync_tx(struct sock *sk, struct tls_context *tls_ctx,
231 				 u32 seq)
232 {
233 	struct net_device *netdev;
234 	int err = 0;
235 	u8 *rcd_sn;
236 
237 	tcp_write_collapse_fence(sk);
238 	rcd_sn = tls_ctx->tx.rec_seq;
239 
240 	trace_tls_device_tx_resync_send(sk, seq, rcd_sn);
241 	down_read(&device_offload_lock);
242 	netdev = rcu_dereference_protected(tls_ctx->netdev,
243 					   lockdep_is_held(&device_offload_lock));
244 	if (netdev)
245 		err = netdev->tlsdev_ops->tls_dev_resync(netdev, sk, seq,
246 							 rcd_sn,
247 							 TLS_OFFLOAD_CTX_DIR_TX);
248 	up_read(&device_offload_lock);
249 	if (err)
250 		return;
251 
252 	clear_bit_unlock(TLS_TX_SYNC_SCHED, &tls_ctx->flags);
253 }
254 
255 static void tls_append_frag(struct tls_record_info *record,
256 			    struct page_frag *pfrag,
257 			    int size)
258 {
259 	skb_frag_t *frag;
260 
261 	frag = &record->frags[record->num_frags - 1];
262 	if (skb_frag_page(frag) == pfrag->page &&
263 	    skb_frag_off(frag) + skb_frag_size(frag) == pfrag->offset) {
264 		skb_frag_size_add(frag, size);
265 	} else {
266 		++frag;
267 		skb_frag_fill_page_desc(frag, pfrag->page, pfrag->offset,
268 					size);
269 		++record->num_frags;
270 		get_page(pfrag->page);
271 	}
272 
273 	pfrag->offset += size;
274 	record->len += size;
275 }
276 
277 static int tls_push_record(struct sock *sk,
278 			   struct tls_context *ctx,
279 			   struct tls_offload_context_tx *offload_ctx,
280 			   struct tls_record_info *record,
281 			   int flags)
282 {
283 	struct tls_prot_info *prot = &ctx->prot_info;
284 	struct tcp_sock *tp = tcp_sk(sk);
285 	skb_frag_t *frag;
286 	int i;
287 
288 	record->end_seq = tp->write_seq + record->len;
289 	list_add_tail_rcu(&record->list, &offload_ctx->records_list);
290 	offload_ctx->open_record = NULL;
291 
292 	if (test_bit(TLS_TX_SYNC_SCHED, &ctx->flags))
293 		tls_device_resync_tx(sk, ctx, tp->write_seq);
294 
295 	tls_advance_record_sn(sk, prot, &ctx->tx);
296 
297 	for (i = 0; i < record->num_frags; i++) {
298 		frag = &record->frags[i];
299 		sg_unmark_end(&offload_ctx->sg_tx_data[i]);
300 		sg_set_page(&offload_ctx->sg_tx_data[i], skb_frag_page(frag),
301 			    skb_frag_size(frag), skb_frag_off(frag));
302 		sk_mem_charge(sk, skb_frag_size(frag));
303 		get_page(skb_frag_page(frag));
304 	}
305 	sg_mark_end(&offload_ctx->sg_tx_data[record->num_frags - 1]);
306 
307 	/* all ready, send */
308 	return tls_push_sg(sk, ctx, offload_ctx->sg_tx_data, 0, flags);
309 }
310 
311 static void tls_device_record_close(struct sock *sk,
312 				    struct tls_context *ctx,
313 				    struct tls_record_info *record,
314 				    struct page_frag *pfrag,
315 				    unsigned char record_type)
316 {
317 	struct tls_prot_info *prot = &ctx->prot_info;
318 	struct page_frag dummy_tag_frag;
319 
320 	/* append tag
321 	 * device will fill in the tag, we just need to append a placeholder
322 	 * use socket memory to improve coalescing (re-using a single buffer
323 	 * increases frag count)
324 	 * if we can't allocate memory now use the dummy page
325 	 */
326 	if (unlikely(pfrag->size - pfrag->offset < prot->tag_size) &&
327 	    !skb_page_frag_refill(prot->tag_size, pfrag, sk->sk_allocation)) {
328 		dummy_tag_frag.page = dummy_page;
329 		dummy_tag_frag.offset = 0;
330 		pfrag = &dummy_tag_frag;
331 	}
332 	tls_append_frag(record, pfrag, prot->tag_size);
333 
334 	/* fill prepend */
335 	tls_fill_prepend(ctx, skb_frag_address(&record->frags[0]),
336 			 record->len - prot->overhead_size,
337 			 record_type);
338 }
339 
340 static int tls_create_new_record(struct tls_offload_context_tx *offload_ctx,
341 				 struct page_frag *pfrag,
342 				 size_t prepend_size)
343 {
344 	struct tls_record_info *record;
345 	skb_frag_t *frag;
346 
347 	record = kmalloc(sizeof(*record), GFP_KERNEL);
348 	if (!record)
349 		return -ENOMEM;
350 
351 	frag = &record->frags[0];
352 	skb_frag_fill_page_desc(frag, pfrag->page, pfrag->offset,
353 				prepend_size);
354 
355 	get_page(pfrag->page);
356 	pfrag->offset += prepend_size;
357 
358 	record->num_frags = 1;
359 	record->len = prepend_size;
360 	offload_ctx->open_record = record;
361 	return 0;
362 }
363 
364 static int tls_do_allocation(struct sock *sk,
365 			     struct tls_offload_context_tx *offload_ctx,
366 			     struct page_frag *pfrag,
367 			     size_t prepend_size)
368 {
369 	int ret;
370 
371 	if (!offload_ctx->open_record) {
372 		if (unlikely(!skb_page_frag_refill(prepend_size, pfrag,
373 						   sk->sk_allocation))) {
374 			READ_ONCE(sk->sk_prot)->enter_memory_pressure(sk);
375 			sk_stream_moderate_sndbuf(sk);
376 			return -ENOMEM;
377 		}
378 
379 		ret = tls_create_new_record(offload_ctx, pfrag, prepend_size);
380 		if (ret)
381 			return ret;
382 
383 		if (pfrag->size > pfrag->offset)
384 			return 0;
385 	}
386 
387 	if (!sk_page_frag_refill(sk, pfrag))
388 		return -ENOMEM;
389 
390 	return 0;
391 }
392 
393 static int tls_device_copy_data(void *addr, size_t bytes, struct iov_iter *i)
394 {
395 	size_t pre_copy, nocache;
396 
397 	pre_copy = ~((unsigned long)addr - 1) & (SMP_CACHE_BYTES - 1);
398 	if (pre_copy) {
399 		pre_copy = min(pre_copy, bytes);
400 		if (copy_from_iter(addr, pre_copy, i) != pre_copy)
401 			return -EFAULT;
402 		bytes -= pre_copy;
403 		addr += pre_copy;
404 	}
405 
406 	nocache = round_down(bytes, SMP_CACHE_BYTES);
407 	if (copy_from_iter_nocache(addr, nocache, i) != nocache)
408 		return -EFAULT;
409 	bytes -= nocache;
410 	addr += nocache;
411 
412 	if (bytes && copy_from_iter(addr, bytes, i) != bytes)
413 		return -EFAULT;
414 
415 	return 0;
416 }
417 
418 static int tls_push_data(struct sock *sk,
419 			 struct iov_iter *iter,
420 			 size_t size, int flags,
421 			 unsigned char record_type)
422 {
423 	struct tls_context *tls_ctx = tls_get_ctx(sk);
424 	struct tls_prot_info *prot = &tls_ctx->prot_info;
425 	struct tls_offload_context_tx *ctx = tls_offload_ctx_tx(tls_ctx);
426 	struct tls_record_info *record;
427 	int tls_push_record_flags;
428 	struct page_frag *pfrag;
429 	size_t orig_size = size;
430 	u32 max_open_record_len;
431 	bool more = false;
432 	bool done = false;
433 	int copy, rc = 0;
434 	long timeo;
435 
436 	if (flags &
437 	    ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL |
438 	      MSG_SPLICE_PAGES | MSG_EOR))
439 		return -EOPNOTSUPP;
440 
441 	if ((flags & (MSG_MORE | MSG_EOR)) == (MSG_MORE | MSG_EOR))
442 		return -EINVAL;
443 
444 	if (unlikely(sk->sk_err))
445 		return -sk->sk_err;
446 
447 	flags |= MSG_SENDPAGE_DECRYPTED;
448 	tls_push_record_flags = flags | MSG_MORE;
449 
450 	timeo = sock_sndtimeo(sk, flags & MSG_DONTWAIT);
451 	if (tls_is_partially_sent_record(tls_ctx)) {
452 		rc = tls_push_partial_record(sk, tls_ctx, flags);
453 		if (rc < 0)
454 			return rc;
455 	}
456 
457 	pfrag = sk_page_frag(sk);
458 
459 	/* TLS_HEADER_SIZE is not counted as part of the TLS record, and
460 	 * we need to leave room for an authentication tag.
461 	 */
462 	max_open_record_len = TLS_MAX_PAYLOAD_SIZE +
463 			      prot->prepend_size;
464 	do {
465 		rc = tls_do_allocation(sk, ctx, pfrag, prot->prepend_size);
466 		if (unlikely(rc)) {
467 			rc = sk_stream_wait_memory(sk, &timeo);
468 			if (!rc)
469 				continue;
470 
471 			record = ctx->open_record;
472 			if (!record)
473 				break;
474 handle_error:
475 			if (record_type != TLS_RECORD_TYPE_DATA) {
476 				/* avoid sending partial
477 				 * record with type !=
478 				 * application_data
479 				 */
480 				size = orig_size;
481 				destroy_record(record);
482 				ctx->open_record = NULL;
483 			} else if (record->len > prot->prepend_size) {
484 				goto last_record;
485 			}
486 
487 			break;
488 		}
489 
490 		record = ctx->open_record;
491 
492 		copy = min_t(size_t, size, max_open_record_len - record->len);
493 		if (copy && (flags & MSG_SPLICE_PAGES)) {
494 			struct page_frag zc_pfrag;
495 			struct page **pages = &zc_pfrag.page;
496 			size_t off;
497 
498 			rc = iov_iter_extract_pages(iter, &pages,
499 						    copy, 1, 0, &off);
500 			if (rc <= 0) {
501 				if (rc == 0)
502 					rc = -EIO;
503 				goto handle_error;
504 			}
505 			copy = rc;
506 
507 			if (WARN_ON_ONCE(!sendpage_ok(zc_pfrag.page))) {
508 				iov_iter_revert(iter, copy);
509 				rc = -EIO;
510 				goto handle_error;
511 			}
512 
513 			zc_pfrag.offset = off;
514 			zc_pfrag.size = copy;
515 			tls_append_frag(record, &zc_pfrag, copy);
516 		} else if (copy) {
517 			copy = min_t(size_t, copy, pfrag->size - pfrag->offset);
518 
519 			rc = tls_device_copy_data(page_address(pfrag->page) +
520 						  pfrag->offset, copy,
521 						  iter);
522 			if (rc)
523 				goto handle_error;
524 			tls_append_frag(record, pfrag, copy);
525 		}
526 
527 		size -= copy;
528 		if (!size) {
529 last_record:
530 			tls_push_record_flags = flags;
531 			if (flags & MSG_MORE) {
532 				more = true;
533 				break;
534 			}
535 
536 			done = true;
537 		}
538 
539 		if (done || record->len >= max_open_record_len ||
540 		    (record->num_frags >= MAX_SKB_FRAGS - 1)) {
541 			tls_device_record_close(sk, tls_ctx, record,
542 						pfrag, record_type);
543 
544 			rc = tls_push_record(sk,
545 					     tls_ctx,
546 					     ctx,
547 					     record,
548 					     tls_push_record_flags);
549 			if (rc < 0)
550 				break;
551 		}
552 	} while (!done);
553 
554 	tls_ctx->pending_open_record_frags = more;
555 
556 	if (orig_size - size > 0)
557 		rc = orig_size - size;
558 
559 	return rc;
560 }
561 
562 int tls_device_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
563 {
564 	unsigned char record_type = TLS_RECORD_TYPE_DATA;
565 	struct tls_context *tls_ctx = tls_get_ctx(sk);
566 	int rc;
567 
568 	if (!tls_ctx->zerocopy_sendfile)
569 		msg->msg_flags &= ~MSG_SPLICE_PAGES;
570 
571 	mutex_lock(&tls_ctx->tx_lock);
572 	lock_sock(sk);
573 
574 	if (unlikely(msg->msg_controllen)) {
575 		rc = tls_process_cmsg(sk, msg, &record_type);
576 		if (rc)
577 			goto out;
578 	}
579 
580 	rc = tls_push_data(sk, &msg->msg_iter, size, msg->msg_flags,
581 			   record_type);
582 
583 out:
584 	release_sock(sk);
585 	mutex_unlock(&tls_ctx->tx_lock);
586 	return rc;
587 }
588 
589 void tls_device_splice_eof(struct socket *sock)
590 {
591 	struct sock *sk = sock->sk;
592 	struct tls_context *tls_ctx = tls_get_ctx(sk);
593 	struct iov_iter iter = {};
594 
595 	if (!tls_is_partially_sent_record(tls_ctx))
596 		return;
597 
598 	mutex_lock(&tls_ctx->tx_lock);
599 	lock_sock(sk);
600 
601 	if (tls_is_partially_sent_record(tls_ctx)) {
602 		iov_iter_bvec(&iter, ITER_SOURCE, NULL, 0, 0);
603 		tls_push_data(sk, &iter, 0, 0, TLS_RECORD_TYPE_DATA);
604 	}
605 
606 	release_sock(sk);
607 	mutex_unlock(&tls_ctx->tx_lock);
608 }
609 
610 struct tls_record_info *tls_get_record(struct tls_offload_context_tx *context,
611 				       u32 seq, u64 *p_record_sn)
612 {
613 	u64 record_sn = context->hint_record_sn;
614 	struct tls_record_info *info, *last;
615 
616 	info = context->retransmit_hint;
617 	if (!info ||
618 	    before(seq, info->end_seq - info->len)) {
619 		/* if retransmit_hint is irrelevant start
620 		 * from the beginning of the list
621 		 */
622 		info = list_first_entry_or_null(&context->records_list,
623 						struct tls_record_info, list);
624 		if (!info)
625 			return NULL;
626 		/* send the start_marker record if seq number is before the
627 		 * tls offload start marker sequence number. This record is
628 		 * required to handle TCP packets which are before TLS offload
629 		 * started.
630 		 *  And if it's not start marker, look if this seq number
631 		 * belongs to the list.
632 		 */
633 		if (likely(!tls_record_is_start_marker(info))) {
634 			/* we have the first record, get the last record to see
635 			 * if this seq number belongs to the list.
636 			 */
637 			last = list_last_entry(&context->records_list,
638 					       struct tls_record_info, list);
639 
640 			if (!between(seq, tls_record_start_seq(info),
641 				     last->end_seq))
642 				return NULL;
643 		}
644 		record_sn = context->unacked_record_sn;
645 	}
646 
647 	/* We just need the _rcu for the READ_ONCE() */
648 	rcu_read_lock();
649 	list_for_each_entry_from_rcu(info, &context->records_list, list) {
650 		if (before(seq, info->end_seq)) {
651 			if (!context->retransmit_hint ||
652 			    after(info->end_seq,
653 				  context->retransmit_hint->end_seq)) {
654 				context->hint_record_sn = record_sn;
655 				context->retransmit_hint = info;
656 			}
657 			*p_record_sn = record_sn;
658 			goto exit_rcu_unlock;
659 		}
660 		record_sn++;
661 	}
662 	info = NULL;
663 
664 exit_rcu_unlock:
665 	rcu_read_unlock();
666 	return info;
667 }
668 EXPORT_SYMBOL(tls_get_record);
669 
670 static int tls_device_push_pending_record(struct sock *sk, int flags)
671 {
672 	struct iov_iter iter;
673 
674 	iov_iter_kvec(&iter, ITER_SOURCE, NULL, 0, 0);
675 	return tls_push_data(sk, &iter, 0, flags, TLS_RECORD_TYPE_DATA);
676 }
677 
678 void tls_device_write_space(struct sock *sk, struct tls_context *ctx)
679 {
680 	if (tls_is_partially_sent_record(ctx)) {
681 		gfp_t sk_allocation = sk->sk_allocation;
682 
683 		WARN_ON_ONCE(sk->sk_write_pending);
684 
685 		sk->sk_allocation = GFP_ATOMIC;
686 		tls_push_partial_record(sk, ctx,
687 					MSG_DONTWAIT | MSG_NOSIGNAL |
688 					MSG_SENDPAGE_DECRYPTED);
689 		sk->sk_allocation = sk_allocation;
690 	}
691 }
692 
693 static void tls_device_resync_rx(struct tls_context *tls_ctx,
694 				 struct sock *sk, u32 seq, u8 *rcd_sn)
695 {
696 	struct tls_offload_context_rx *rx_ctx = tls_offload_ctx_rx(tls_ctx);
697 	struct net_device *netdev;
698 
699 	trace_tls_device_rx_resync_send(sk, seq, rcd_sn, rx_ctx->resync_type);
700 	rcu_read_lock();
701 	netdev = rcu_dereference(tls_ctx->netdev);
702 	if (netdev)
703 		netdev->tlsdev_ops->tls_dev_resync(netdev, sk, seq, rcd_sn,
704 						   TLS_OFFLOAD_CTX_DIR_RX);
705 	rcu_read_unlock();
706 	TLS_INC_STATS(sock_net(sk), LINUX_MIB_TLSRXDEVICERESYNC);
707 }
708 
709 static bool
710 tls_device_rx_resync_async(struct tls_offload_resync_async *resync_async,
711 			   s64 resync_req, u32 *seq, u16 *rcd_delta)
712 {
713 	u32 is_async = resync_req & RESYNC_REQ_ASYNC;
714 	u32 req_seq = resync_req >> 32;
715 	u32 req_end = req_seq + ((resync_req >> 16) & 0xffff);
716 	u16 i;
717 
718 	*rcd_delta = 0;
719 
720 	if (is_async) {
721 		/* shouldn't get to wraparound:
722 		 * too long in async stage, something bad happened
723 		 */
724 		if (WARN_ON_ONCE(resync_async->rcd_delta == USHRT_MAX))
725 			return false;
726 
727 		/* asynchronous stage: log all headers seq such that
728 		 * req_seq <= seq <= end_seq, and wait for real resync request
729 		 */
730 		if (before(*seq, req_seq))
731 			return false;
732 		if (!after(*seq, req_end) &&
733 		    resync_async->loglen < TLS_DEVICE_RESYNC_ASYNC_LOGMAX)
734 			resync_async->log[resync_async->loglen++] = *seq;
735 
736 		resync_async->rcd_delta++;
737 
738 		return false;
739 	}
740 
741 	/* synchronous stage: check against the logged entries and
742 	 * proceed to check the next entries if no match was found
743 	 */
744 	for (i = 0; i < resync_async->loglen; i++)
745 		if (req_seq == resync_async->log[i] &&
746 		    atomic64_try_cmpxchg(&resync_async->req, &resync_req, 0)) {
747 			*rcd_delta = resync_async->rcd_delta - i;
748 			*seq = req_seq;
749 			resync_async->loglen = 0;
750 			resync_async->rcd_delta = 0;
751 			return true;
752 		}
753 
754 	resync_async->loglen = 0;
755 	resync_async->rcd_delta = 0;
756 
757 	if (req_seq == *seq &&
758 	    atomic64_try_cmpxchg(&resync_async->req,
759 				 &resync_req, 0))
760 		return true;
761 
762 	return false;
763 }
764 
765 void tls_device_rx_resync_new_rec(struct sock *sk, u32 rcd_len, u32 seq)
766 {
767 	struct tls_context *tls_ctx = tls_get_ctx(sk);
768 	struct tls_offload_context_rx *rx_ctx;
769 	u8 rcd_sn[TLS_MAX_REC_SEQ_SIZE];
770 	u32 sock_data, is_req_pending;
771 	struct tls_prot_info *prot;
772 	s64 resync_req;
773 	u16 rcd_delta;
774 	u32 req_seq;
775 
776 	if (tls_ctx->rx_conf != TLS_HW)
777 		return;
778 	if (unlikely(test_bit(TLS_RX_DEV_DEGRADED, &tls_ctx->flags)))
779 		return;
780 
781 	prot = &tls_ctx->prot_info;
782 	rx_ctx = tls_offload_ctx_rx(tls_ctx);
783 	memcpy(rcd_sn, tls_ctx->rx.rec_seq, prot->rec_seq_size);
784 
785 	switch (rx_ctx->resync_type) {
786 	case TLS_OFFLOAD_SYNC_TYPE_DRIVER_REQ:
787 		resync_req = atomic64_read(&rx_ctx->resync_req);
788 		req_seq = resync_req >> 32;
789 		seq += TLS_HEADER_SIZE - 1;
790 		is_req_pending = resync_req;
791 
792 		if (likely(!is_req_pending) || req_seq != seq ||
793 		    !atomic64_try_cmpxchg(&rx_ctx->resync_req, &resync_req, 0))
794 			return;
795 		break;
796 	case TLS_OFFLOAD_SYNC_TYPE_CORE_NEXT_HINT:
797 		if (likely(!rx_ctx->resync_nh_do_now))
798 			return;
799 
800 		/* head of next rec is already in, note that the sock_inq will
801 		 * include the currently parsed message when called from parser
802 		 */
803 		sock_data = tcp_inq(sk);
804 		if (sock_data > rcd_len) {
805 			trace_tls_device_rx_resync_nh_delay(sk, sock_data,
806 							    rcd_len);
807 			return;
808 		}
809 
810 		rx_ctx->resync_nh_do_now = 0;
811 		seq += rcd_len;
812 		tls_bigint_increment(rcd_sn, prot->rec_seq_size);
813 		break;
814 	case TLS_OFFLOAD_SYNC_TYPE_DRIVER_REQ_ASYNC:
815 		resync_req = atomic64_read(&rx_ctx->resync_async->req);
816 		is_req_pending = resync_req;
817 		if (likely(!is_req_pending))
818 			return;
819 
820 		if (!tls_device_rx_resync_async(rx_ctx->resync_async,
821 						resync_req, &seq, &rcd_delta))
822 			return;
823 		tls_bigint_subtract(rcd_sn, rcd_delta);
824 		break;
825 	}
826 
827 	tls_device_resync_rx(tls_ctx, sk, seq, rcd_sn);
828 }
829 
830 static void tls_device_core_ctrl_rx_resync(struct tls_context *tls_ctx,
831 					   struct tls_offload_context_rx *ctx,
832 					   struct sock *sk, struct sk_buff *skb)
833 {
834 	struct strp_msg *rxm;
835 
836 	/* device will request resyncs by itself based on stream scan */
837 	if (ctx->resync_type != TLS_OFFLOAD_SYNC_TYPE_CORE_NEXT_HINT)
838 		return;
839 	/* already scheduled */
840 	if (ctx->resync_nh_do_now)
841 		return;
842 	/* seen decrypted fragments since last fully-failed record */
843 	if (ctx->resync_nh_reset) {
844 		ctx->resync_nh_reset = 0;
845 		ctx->resync_nh.decrypted_failed = 1;
846 		ctx->resync_nh.decrypted_tgt = TLS_DEVICE_RESYNC_NH_START_IVAL;
847 		return;
848 	}
849 
850 	if (++ctx->resync_nh.decrypted_failed <= ctx->resync_nh.decrypted_tgt)
851 		return;
852 
853 	/* doing resync, bump the next target in case it fails */
854 	if (ctx->resync_nh.decrypted_tgt < TLS_DEVICE_RESYNC_NH_MAX_IVAL)
855 		ctx->resync_nh.decrypted_tgt *= 2;
856 	else
857 		ctx->resync_nh.decrypted_tgt += TLS_DEVICE_RESYNC_NH_MAX_IVAL;
858 
859 	rxm = strp_msg(skb);
860 
861 	/* head of next rec is already in, parser will sync for us */
862 	if (tcp_inq(sk) > rxm->full_len) {
863 		trace_tls_device_rx_resync_nh_schedule(sk);
864 		ctx->resync_nh_do_now = 1;
865 	} else {
866 		struct tls_prot_info *prot = &tls_ctx->prot_info;
867 		u8 rcd_sn[TLS_MAX_REC_SEQ_SIZE];
868 
869 		memcpy(rcd_sn, tls_ctx->rx.rec_seq, prot->rec_seq_size);
870 		tls_bigint_increment(rcd_sn, prot->rec_seq_size);
871 
872 		tls_device_resync_rx(tls_ctx, sk, tcp_sk(sk)->copied_seq,
873 				     rcd_sn);
874 	}
875 }
876 
877 static int
878 tls_device_reencrypt(struct sock *sk, struct tls_context *tls_ctx)
879 {
880 	struct tls_sw_context_rx *sw_ctx = tls_sw_ctx_rx(tls_ctx);
881 	const struct tls_cipher_desc *cipher_desc;
882 	int err, offset, copy, data_len, pos;
883 	struct sk_buff *skb, *skb_iter;
884 	struct scatterlist sg[1];
885 	struct strp_msg *rxm;
886 	char *orig_buf, *buf;
887 
888 	cipher_desc = get_cipher_desc(tls_ctx->crypto_recv.info.cipher_type);
889 	DEBUG_NET_WARN_ON_ONCE(!cipher_desc || !cipher_desc->offloadable);
890 
891 	rxm = strp_msg(tls_strp_msg(sw_ctx));
892 	orig_buf = kmalloc(rxm->full_len + TLS_HEADER_SIZE + cipher_desc->iv,
893 			   sk->sk_allocation);
894 	if (!orig_buf)
895 		return -ENOMEM;
896 	buf = orig_buf;
897 
898 	err = tls_strp_msg_cow(sw_ctx);
899 	if (unlikely(err))
900 		goto free_buf;
901 
902 	skb = tls_strp_msg(sw_ctx);
903 	rxm = strp_msg(skb);
904 	offset = rxm->offset;
905 
906 	sg_init_table(sg, 1);
907 	sg_set_buf(&sg[0], buf,
908 		   rxm->full_len + TLS_HEADER_SIZE + cipher_desc->iv);
909 	err = skb_copy_bits(skb, offset, buf, TLS_HEADER_SIZE + cipher_desc->iv);
910 	if (err)
911 		goto free_buf;
912 
913 	/* We are interested only in the decrypted data not the auth */
914 	err = decrypt_skb(sk, sg);
915 	if (err != -EBADMSG)
916 		goto free_buf;
917 	else
918 		err = 0;
919 
920 	data_len = rxm->full_len - cipher_desc->tag;
921 
922 	if (skb_pagelen(skb) > offset) {
923 		copy = min_t(int, skb_pagelen(skb) - offset, data_len);
924 
925 		if (skb->decrypted) {
926 			err = skb_store_bits(skb, offset, buf, copy);
927 			if (err)
928 				goto free_buf;
929 		}
930 
931 		offset += copy;
932 		buf += copy;
933 	}
934 
935 	pos = skb_pagelen(skb);
936 	skb_walk_frags(skb, skb_iter) {
937 		int frag_pos;
938 
939 		/* Practically all frags must belong to msg if reencrypt
940 		 * is needed with current strparser and coalescing logic,
941 		 * but strparser may "get optimized", so let's be safe.
942 		 */
943 		if (pos + skb_iter->len <= offset)
944 			goto done_with_frag;
945 		if (pos >= data_len + rxm->offset)
946 			break;
947 
948 		frag_pos = offset - pos;
949 		copy = min_t(int, skb_iter->len - frag_pos,
950 			     data_len + rxm->offset - offset);
951 
952 		if (skb_iter->decrypted) {
953 			err = skb_store_bits(skb_iter, frag_pos, buf, copy);
954 			if (err)
955 				goto free_buf;
956 		}
957 
958 		offset += copy;
959 		buf += copy;
960 done_with_frag:
961 		pos += skb_iter->len;
962 	}
963 
964 free_buf:
965 	kfree(orig_buf);
966 	return err;
967 }
968 
969 int tls_device_decrypted(struct sock *sk, struct tls_context *tls_ctx)
970 {
971 	struct tls_offload_context_rx *ctx = tls_offload_ctx_rx(tls_ctx);
972 	struct tls_sw_context_rx *sw_ctx = tls_sw_ctx_rx(tls_ctx);
973 	struct sk_buff *skb = tls_strp_msg(sw_ctx);
974 	struct strp_msg *rxm = strp_msg(skb);
975 	int is_decrypted, is_encrypted;
976 
977 	if (!tls_strp_msg_mixed_decrypted(sw_ctx)) {
978 		is_decrypted = skb->decrypted;
979 		is_encrypted = !is_decrypted;
980 	} else {
981 		is_decrypted = 0;
982 		is_encrypted = 0;
983 	}
984 
985 	trace_tls_device_decrypted(sk, tcp_sk(sk)->copied_seq - rxm->full_len,
986 				   tls_ctx->rx.rec_seq, rxm->full_len,
987 				   is_encrypted, is_decrypted);
988 
989 	if (unlikely(test_bit(TLS_RX_DEV_DEGRADED, &tls_ctx->flags))) {
990 		if (likely(is_encrypted || is_decrypted))
991 			return is_decrypted;
992 
993 		/* After tls_device_down disables the offload, the next SKB will
994 		 * likely have initial fragments decrypted, and final ones not
995 		 * decrypted. We need to reencrypt that single SKB.
996 		 */
997 		return tls_device_reencrypt(sk, tls_ctx);
998 	}
999 
1000 	/* Return immediately if the record is either entirely plaintext or
1001 	 * entirely ciphertext. Otherwise handle reencrypt partially decrypted
1002 	 * record.
1003 	 */
1004 	if (is_decrypted) {
1005 		ctx->resync_nh_reset = 1;
1006 		return is_decrypted;
1007 	}
1008 	if (is_encrypted) {
1009 		tls_device_core_ctrl_rx_resync(tls_ctx, ctx, sk, skb);
1010 		return 0;
1011 	}
1012 
1013 	ctx->resync_nh_reset = 1;
1014 	return tls_device_reencrypt(sk, tls_ctx);
1015 }
1016 
1017 static void tls_device_attach(struct tls_context *ctx, struct sock *sk,
1018 			      struct net_device *netdev)
1019 {
1020 	if (sk->sk_destruct != tls_device_sk_destruct) {
1021 		refcount_set(&ctx->refcount, 1);
1022 		dev_hold(netdev);
1023 		RCU_INIT_POINTER(ctx->netdev, netdev);
1024 		spin_lock_irq(&tls_device_lock);
1025 		list_add_tail(&ctx->list, &tls_device_list);
1026 		spin_unlock_irq(&tls_device_lock);
1027 
1028 		ctx->sk_destruct = sk->sk_destruct;
1029 		smp_store_release(&sk->sk_destruct, tls_device_sk_destruct);
1030 	}
1031 }
1032 
1033 static struct tls_offload_context_tx *alloc_offload_ctx_tx(struct tls_context *ctx)
1034 {
1035 	struct tls_offload_context_tx *offload_ctx;
1036 	__be64 rcd_sn;
1037 
1038 	offload_ctx = kzalloc(sizeof(*offload_ctx), GFP_KERNEL);
1039 	if (!offload_ctx)
1040 		return NULL;
1041 
1042 	INIT_WORK(&offload_ctx->destruct_work, tls_device_tx_del_task);
1043 	INIT_LIST_HEAD(&offload_ctx->records_list);
1044 	spin_lock_init(&offload_ctx->lock);
1045 	sg_init_table(offload_ctx->sg_tx_data,
1046 		      ARRAY_SIZE(offload_ctx->sg_tx_data));
1047 
1048 	/* start at rec_seq - 1 to account for the start marker record */
1049 	memcpy(&rcd_sn, ctx->tx.rec_seq, sizeof(rcd_sn));
1050 	offload_ctx->unacked_record_sn = be64_to_cpu(rcd_sn) - 1;
1051 
1052 	offload_ctx->ctx = ctx;
1053 
1054 	return offload_ctx;
1055 }
1056 
1057 int tls_set_device_offload(struct sock *sk)
1058 {
1059 	struct tls_record_info *start_marker_record;
1060 	struct tls_offload_context_tx *offload_ctx;
1061 	const struct tls_cipher_desc *cipher_desc;
1062 	struct tls_crypto_info *crypto_info;
1063 	struct tls_prot_info *prot;
1064 	struct net_device *netdev;
1065 	struct tls_context *ctx;
1066 	char *iv, *rec_seq;
1067 	int rc;
1068 
1069 	ctx = tls_get_ctx(sk);
1070 	prot = &ctx->prot_info;
1071 
1072 	if (ctx->priv_ctx_tx)
1073 		return -EEXIST;
1074 
1075 	netdev = get_netdev_for_sock(sk);
1076 	if (!netdev) {
1077 		pr_err_ratelimited("%s: netdev not found\n", __func__);
1078 		return -EINVAL;
1079 	}
1080 
1081 	if (!(netdev->features & NETIF_F_HW_TLS_TX)) {
1082 		rc = -EOPNOTSUPP;
1083 		goto release_netdev;
1084 	}
1085 
1086 	crypto_info = &ctx->crypto_send.info;
1087 	if (crypto_info->version != TLS_1_2_VERSION) {
1088 		rc = -EOPNOTSUPP;
1089 		goto release_netdev;
1090 	}
1091 
1092 	cipher_desc = get_cipher_desc(crypto_info->cipher_type);
1093 	if (!cipher_desc || !cipher_desc->offloadable) {
1094 		rc = -EINVAL;
1095 		goto release_netdev;
1096 	}
1097 
1098 	rc = init_prot_info(prot, crypto_info, cipher_desc);
1099 	if (rc)
1100 		goto release_netdev;
1101 
1102 	iv = crypto_info_iv(crypto_info, cipher_desc);
1103 	rec_seq = crypto_info_rec_seq(crypto_info, cipher_desc);
1104 
1105 	memcpy(ctx->tx.iv + cipher_desc->salt, iv, cipher_desc->iv);
1106 	memcpy(ctx->tx.rec_seq, rec_seq, cipher_desc->rec_seq);
1107 
1108 	start_marker_record = kmalloc(sizeof(*start_marker_record), GFP_KERNEL);
1109 	if (!start_marker_record) {
1110 		rc = -ENOMEM;
1111 		goto release_netdev;
1112 	}
1113 
1114 	offload_ctx = alloc_offload_ctx_tx(ctx);
1115 	if (!offload_ctx) {
1116 		rc = -ENOMEM;
1117 		goto free_marker_record;
1118 	}
1119 
1120 	rc = tls_sw_fallback_init(sk, offload_ctx, crypto_info);
1121 	if (rc)
1122 		goto free_offload_ctx;
1123 
1124 	start_marker_record->end_seq = tcp_sk(sk)->write_seq;
1125 	start_marker_record->len = 0;
1126 	start_marker_record->num_frags = 0;
1127 	list_add_tail(&start_marker_record->list, &offload_ctx->records_list);
1128 
1129 	clean_acked_data_enable(inet_csk(sk), &tls_icsk_clean_acked);
1130 	ctx->push_pending_record = tls_device_push_pending_record;
1131 
1132 	/* TLS offload is greatly simplified if we don't send
1133 	 * SKBs where only part of the payload needs to be encrypted.
1134 	 * So mark the last skb in the write queue as end of record.
1135 	 */
1136 	tcp_write_collapse_fence(sk);
1137 
1138 	/* Avoid offloading if the device is down
1139 	 * We don't want to offload new flows after
1140 	 * the NETDEV_DOWN event
1141 	 *
1142 	 * device_offload_lock is taken in tls_devices's NETDEV_DOWN
1143 	 * handler thus protecting from the device going down before
1144 	 * ctx was added to tls_device_list.
1145 	 */
1146 	down_read(&device_offload_lock);
1147 	if (!(netdev->flags & IFF_UP)) {
1148 		rc = -EINVAL;
1149 		goto release_lock;
1150 	}
1151 
1152 	ctx->priv_ctx_tx = offload_ctx;
1153 	rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_TX,
1154 					     &ctx->crypto_send.info,
1155 					     tcp_sk(sk)->write_seq);
1156 	trace_tls_device_offload_set(sk, TLS_OFFLOAD_CTX_DIR_TX,
1157 				     tcp_sk(sk)->write_seq, rec_seq, rc);
1158 	if (rc)
1159 		goto release_lock;
1160 
1161 	tls_device_attach(ctx, sk, netdev);
1162 	up_read(&device_offload_lock);
1163 
1164 	/* following this assignment tls_is_skb_tx_device_offloaded
1165 	 * will return true and the context might be accessed
1166 	 * by the netdev's xmit function.
1167 	 */
1168 	smp_store_release(&sk->sk_validate_xmit_skb, tls_validate_xmit_skb);
1169 	dev_put(netdev);
1170 
1171 	return 0;
1172 
1173 release_lock:
1174 	up_read(&device_offload_lock);
1175 	clean_acked_data_disable(inet_csk(sk));
1176 	crypto_free_aead(offload_ctx->aead_send);
1177 free_offload_ctx:
1178 	kfree(offload_ctx);
1179 	ctx->priv_ctx_tx = NULL;
1180 free_marker_record:
1181 	kfree(start_marker_record);
1182 release_netdev:
1183 	dev_put(netdev);
1184 	return rc;
1185 }
1186 
1187 int tls_set_device_offload_rx(struct sock *sk, struct tls_context *ctx)
1188 {
1189 	struct tls12_crypto_info_aes_gcm_128 *info;
1190 	struct tls_offload_context_rx *context;
1191 	struct net_device *netdev;
1192 	int rc = 0;
1193 
1194 	if (ctx->crypto_recv.info.version != TLS_1_2_VERSION)
1195 		return -EOPNOTSUPP;
1196 
1197 	netdev = get_netdev_for_sock(sk);
1198 	if (!netdev) {
1199 		pr_err_ratelimited("%s: netdev not found\n", __func__);
1200 		return -EINVAL;
1201 	}
1202 
1203 	if (!(netdev->features & NETIF_F_HW_TLS_RX)) {
1204 		rc = -EOPNOTSUPP;
1205 		goto release_netdev;
1206 	}
1207 
1208 	/* Avoid offloading if the device is down
1209 	 * We don't want to offload new flows after
1210 	 * the NETDEV_DOWN event
1211 	 *
1212 	 * device_offload_lock is taken in tls_devices's NETDEV_DOWN
1213 	 * handler thus protecting from the device going down before
1214 	 * ctx was added to tls_device_list.
1215 	 */
1216 	down_read(&device_offload_lock);
1217 	if (!(netdev->flags & IFF_UP)) {
1218 		rc = -EINVAL;
1219 		goto release_lock;
1220 	}
1221 
1222 	context = kzalloc(sizeof(*context), GFP_KERNEL);
1223 	if (!context) {
1224 		rc = -ENOMEM;
1225 		goto release_lock;
1226 	}
1227 	context->resync_nh_reset = 1;
1228 
1229 	ctx->priv_ctx_rx = context;
1230 	rc = tls_set_sw_offload(sk, 0);
1231 	if (rc)
1232 		goto release_ctx;
1233 
1234 	rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_RX,
1235 					     &ctx->crypto_recv.info,
1236 					     tcp_sk(sk)->copied_seq);
1237 	info = (void *)&ctx->crypto_recv.info;
1238 	trace_tls_device_offload_set(sk, TLS_OFFLOAD_CTX_DIR_RX,
1239 				     tcp_sk(sk)->copied_seq, info->rec_seq, rc);
1240 	if (rc)
1241 		goto free_sw_resources;
1242 
1243 	tls_device_attach(ctx, sk, netdev);
1244 	up_read(&device_offload_lock);
1245 
1246 	dev_put(netdev);
1247 
1248 	return 0;
1249 
1250 free_sw_resources:
1251 	up_read(&device_offload_lock);
1252 	tls_sw_free_resources_rx(sk);
1253 	down_read(&device_offload_lock);
1254 release_ctx:
1255 	ctx->priv_ctx_rx = NULL;
1256 release_lock:
1257 	up_read(&device_offload_lock);
1258 release_netdev:
1259 	dev_put(netdev);
1260 	return rc;
1261 }
1262 
1263 void tls_device_offload_cleanup_rx(struct sock *sk)
1264 {
1265 	struct tls_context *tls_ctx = tls_get_ctx(sk);
1266 	struct net_device *netdev;
1267 
1268 	down_read(&device_offload_lock);
1269 	netdev = rcu_dereference_protected(tls_ctx->netdev,
1270 					   lockdep_is_held(&device_offload_lock));
1271 	if (!netdev)
1272 		goto out;
1273 
1274 	netdev->tlsdev_ops->tls_dev_del(netdev, tls_ctx,
1275 					TLS_OFFLOAD_CTX_DIR_RX);
1276 
1277 	if (tls_ctx->tx_conf != TLS_HW) {
1278 		dev_put(netdev);
1279 		rcu_assign_pointer(tls_ctx->netdev, NULL);
1280 	} else {
1281 		set_bit(TLS_RX_DEV_CLOSED, &tls_ctx->flags);
1282 	}
1283 out:
1284 	up_read(&device_offload_lock);
1285 	tls_sw_release_resources_rx(sk);
1286 }
1287 
1288 static int tls_device_down(struct net_device *netdev)
1289 {
1290 	struct tls_context *ctx, *tmp;
1291 	unsigned long flags;
1292 	LIST_HEAD(list);
1293 
1294 	/* Request a write lock to block new offload attempts */
1295 	down_write(&device_offload_lock);
1296 
1297 	spin_lock_irqsave(&tls_device_lock, flags);
1298 	list_for_each_entry_safe(ctx, tmp, &tls_device_list, list) {
1299 		struct net_device *ctx_netdev =
1300 			rcu_dereference_protected(ctx->netdev,
1301 						  lockdep_is_held(&device_offload_lock));
1302 
1303 		if (ctx_netdev != netdev ||
1304 		    !refcount_inc_not_zero(&ctx->refcount))
1305 			continue;
1306 
1307 		list_move(&ctx->list, &list);
1308 	}
1309 	spin_unlock_irqrestore(&tls_device_lock, flags);
1310 
1311 	list_for_each_entry_safe(ctx, tmp, &list, list)	{
1312 		/* Stop offloaded TX and switch to the fallback.
1313 		 * tls_is_skb_tx_device_offloaded will return false.
1314 		 */
1315 		WRITE_ONCE(ctx->sk->sk_validate_xmit_skb, tls_validate_xmit_skb_sw);
1316 
1317 		/* Stop the RX and TX resync.
1318 		 * tls_dev_resync must not be called after tls_dev_del.
1319 		 */
1320 		rcu_assign_pointer(ctx->netdev, NULL);
1321 
1322 		/* Start skipping the RX resync logic completely. */
1323 		set_bit(TLS_RX_DEV_DEGRADED, &ctx->flags);
1324 
1325 		/* Sync with inflight packets. After this point:
1326 		 * TX: no non-encrypted packets will be passed to the driver.
1327 		 * RX: resync requests from the driver will be ignored.
1328 		 */
1329 		synchronize_net();
1330 
1331 		/* Release the offload context on the driver side. */
1332 		if (ctx->tx_conf == TLS_HW)
1333 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
1334 							TLS_OFFLOAD_CTX_DIR_TX);
1335 		if (ctx->rx_conf == TLS_HW &&
1336 		    !test_bit(TLS_RX_DEV_CLOSED, &ctx->flags))
1337 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
1338 							TLS_OFFLOAD_CTX_DIR_RX);
1339 
1340 		dev_put(netdev);
1341 
1342 		/* Move the context to a separate list for two reasons:
1343 		 * 1. When the context is deallocated, list_del is called.
1344 		 * 2. It's no longer an offloaded context, so we don't want to
1345 		 *    run offload-specific code on this context.
1346 		 */
1347 		spin_lock_irqsave(&tls_device_lock, flags);
1348 		list_move_tail(&ctx->list, &tls_device_down_list);
1349 		spin_unlock_irqrestore(&tls_device_lock, flags);
1350 
1351 		/* Device contexts for RX and TX will be freed in on sk_destruct
1352 		 * by tls_device_free_ctx. rx_conf and tx_conf stay in TLS_HW.
1353 		 * Now release the ref taken above.
1354 		 */
1355 		if (refcount_dec_and_test(&ctx->refcount)) {
1356 			/* sk_destruct ran after tls_device_down took a ref, and
1357 			 * it returned early. Complete the destruction here.
1358 			 */
1359 			list_del(&ctx->list);
1360 			tls_device_free_ctx(ctx);
1361 		}
1362 	}
1363 
1364 	up_write(&device_offload_lock);
1365 
1366 	flush_workqueue(destruct_wq);
1367 
1368 	return NOTIFY_DONE;
1369 }
1370 
1371 static int tls_dev_event(struct notifier_block *this, unsigned long event,
1372 			 void *ptr)
1373 {
1374 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1375 
1376 	if (!dev->tlsdev_ops &&
1377 	    !(dev->features & (NETIF_F_HW_TLS_RX | NETIF_F_HW_TLS_TX)))
1378 		return NOTIFY_DONE;
1379 
1380 	switch (event) {
1381 	case NETDEV_REGISTER:
1382 	case NETDEV_FEAT_CHANGE:
1383 		if (netif_is_bond_master(dev))
1384 			return NOTIFY_DONE;
1385 		if ((dev->features & NETIF_F_HW_TLS_RX) &&
1386 		    !dev->tlsdev_ops->tls_dev_resync)
1387 			return NOTIFY_BAD;
1388 
1389 		if  (dev->tlsdev_ops &&
1390 		     dev->tlsdev_ops->tls_dev_add &&
1391 		     dev->tlsdev_ops->tls_dev_del)
1392 			return NOTIFY_DONE;
1393 		else
1394 			return NOTIFY_BAD;
1395 	case NETDEV_DOWN:
1396 		return tls_device_down(dev);
1397 	}
1398 	return NOTIFY_DONE;
1399 }
1400 
1401 static struct notifier_block tls_dev_notifier = {
1402 	.notifier_call	= tls_dev_event,
1403 };
1404 
1405 int __init tls_device_init(void)
1406 {
1407 	int err;
1408 
1409 	dummy_page = alloc_page(GFP_KERNEL);
1410 	if (!dummy_page)
1411 		return -ENOMEM;
1412 
1413 	destruct_wq = alloc_workqueue("ktls_device_destruct", 0, 0);
1414 	if (!destruct_wq) {
1415 		err = -ENOMEM;
1416 		goto err_free_dummy;
1417 	}
1418 
1419 	err = register_netdevice_notifier(&tls_dev_notifier);
1420 	if (err)
1421 		goto err_destroy_wq;
1422 
1423 	return 0;
1424 
1425 err_destroy_wq:
1426 	destroy_workqueue(destruct_wq);
1427 err_free_dummy:
1428 	put_page(dummy_page);
1429 	return err;
1430 }
1431 
1432 void __exit tls_device_cleanup(void)
1433 {
1434 	unregister_netdevice_notifier(&tls_dev_notifier);
1435 	destroy_workqueue(destruct_wq);
1436 	clean_acked_data_flush();
1437 	put_page(dummy_page);
1438 }
1439