• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 
41 #include "trace.h"
42 
43 /* device_offload_lock is used to synchronize tls_dev_add
44  * against NETDEV_DOWN notifications.
45  */
46 static DECLARE_RWSEM(device_offload_lock);
47 
48 static void tls_device_gc_task(struct work_struct *work);
49 
50 static DECLARE_WORK(tls_device_gc_work, tls_device_gc_task);
51 static LIST_HEAD(tls_device_gc_list);
52 static LIST_HEAD(tls_device_list);
53 static LIST_HEAD(tls_device_down_list);
54 static DEFINE_SPINLOCK(tls_device_lock);
55 
tls_device_free_ctx(struct tls_context * ctx)56 static void tls_device_free_ctx(struct tls_context *ctx)
57 {
58 	if (ctx->tx_conf == TLS_HW) {
59 		kfree(tls_offload_ctx_tx(ctx));
60 		kfree(ctx->tx.rec_seq);
61 		kfree(ctx->tx.iv);
62 	}
63 
64 	if (ctx->rx_conf == TLS_HW)
65 		kfree(tls_offload_ctx_rx(ctx));
66 
67 	tls_ctx_free(NULL, ctx);
68 }
69 
tls_device_gc_task(struct work_struct * work)70 static void tls_device_gc_task(struct work_struct *work)
71 {
72 	struct tls_context *ctx, *tmp;
73 	unsigned long flags;
74 	LIST_HEAD(gc_list);
75 
76 	spin_lock_irqsave(&tls_device_lock, flags);
77 	list_splice_init(&tls_device_gc_list, &gc_list);
78 	spin_unlock_irqrestore(&tls_device_lock, flags);
79 
80 	list_for_each_entry_safe(ctx, tmp, &gc_list, list) {
81 		struct net_device *netdev = ctx->netdev;
82 
83 		if (netdev && ctx->tx_conf == TLS_HW) {
84 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
85 							TLS_OFFLOAD_CTX_DIR_TX);
86 			dev_put(netdev);
87 			ctx->netdev = NULL;
88 		}
89 
90 		list_del(&ctx->list);
91 		tls_device_free_ctx(ctx);
92 	}
93 }
94 
tls_device_queue_ctx_destruction(struct tls_context * ctx)95 static void tls_device_queue_ctx_destruction(struct tls_context *ctx)
96 {
97 	unsigned long flags;
98 
99 	spin_lock_irqsave(&tls_device_lock, flags);
100 	if (unlikely(!refcount_dec_and_test(&ctx->refcount)))
101 		goto unlock;
102 
103 	list_move_tail(&ctx->list, &tls_device_gc_list);
104 
105 	/* schedule_work inside the spinlock
106 	 * to make sure tls_device_down waits for that work.
107 	 */
108 	schedule_work(&tls_device_gc_work);
109 unlock:
110 	spin_unlock_irqrestore(&tls_device_lock, flags);
111 }
112 
113 /* We assume that the socket is already connected */
get_netdev_for_sock(struct sock * sk)114 static struct net_device *get_netdev_for_sock(struct sock *sk)
115 {
116 	struct dst_entry *dst = sk_dst_get(sk);
117 	struct net_device *netdev = NULL;
118 
119 	if (likely(dst)) {
120 		netdev = dst->dev;
121 		dev_hold(netdev);
122 	}
123 
124 	dst_release(dst);
125 
126 	return netdev;
127 }
128 
destroy_record(struct tls_record_info * record)129 static void destroy_record(struct tls_record_info *record)
130 {
131 	int i;
132 
133 	for (i = 0; i < record->num_frags; i++)
134 		__skb_frag_unref(&record->frags[i]);
135 	kfree(record);
136 }
137 
delete_all_records(struct tls_offload_context_tx * offload_ctx)138 static void delete_all_records(struct tls_offload_context_tx *offload_ctx)
139 {
140 	struct tls_record_info *info, *temp;
141 
142 	list_for_each_entry_safe(info, temp, &offload_ctx->records_list, list) {
143 		list_del(&info->list);
144 		destroy_record(info);
145 	}
146 
147 	offload_ctx->retransmit_hint = NULL;
148 }
149 
tls_icsk_clean_acked(struct sock * sk,u32 acked_seq)150 static void tls_icsk_clean_acked(struct sock *sk, u32 acked_seq)
151 {
152 	struct tls_context *tls_ctx = tls_get_ctx(sk);
153 	struct tls_record_info *info, *temp;
154 	struct tls_offload_context_tx *ctx;
155 	u64 deleted_records = 0;
156 	unsigned long flags;
157 
158 	if (!tls_ctx)
159 		return;
160 
161 	ctx = tls_offload_ctx_tx(tls_ctx);
162 
163 	spin_lock_irqsave(&ctx->lock, flags);
164 	info = ctx->retransmit_hint;
165 	if (info && !before(acked_seq, info->end_seq))
166 		ctx->retransmit_hint = NULL;
167 
168 	list_for_each_entry_safe(info, temp, &ctx->records_list, list) {
169 		if (before(acked_seq, info->end_seq))
170 			break;
171 		list_del(&info->list);
172 
173 		destroy_record(info);
174 		deleted_records++;
175 	}
176 
177 	ctx->unacked_record_sn += deleted_records;
178 	spin_unlock_irqrestore(&ctx->lock, flags);
179 }
180 
181 /* At this point, there should be no references on this
182  * socket and no in-flight SKBs associated with this
183  * socket, so it is safe to free all the resources.
184  */
tls_device_sk_destruct(struct sock * sk)185 void tls_device_sk_destruct(struct sock *sk)
186 {
187 	struct tls_context *tls_ctx = tls_get_ctx(sk);
188 	struct tls_offload_context_tx *ctx = tls_offload_ctx_tx(tls_ctx);
189 
190 	tls_ctx->sk_destruct(sk);
191 
192 	if (tls_ctx->tx_conf == TLS_HW) {
193 		if (ctx->open_record)
194 			destroy_record(ctx->open_record);
195 		delete_all_records(ctx);
196 		crypto_free_aead(ctx->aead_send);
197 		clean_acked_data_disable(inet_csk(sk));
198 	}
199 
200 	tls_device_queue_ctx_destruction(tls_ctx);
201 }
202 EXPORT_SYMBOL_GPL(tls_device_sk_destruct);
203 
tls_device_free_resources_tx(struct sock * sk)204 void tls_device_free_resources_tx(struct sock *sk)
205 {
206 	struct tls_context *tls_ctx = tls_get_ctx(sk);
207 
208 	tls_free_partial_record(sk, tls_ctx);
209 }
210 
tls_offload_tx_resync_request(struct sock * sk,u32 got_seq,u32 exp_seq)211 void tls_offload_tx_resync_request(struct sock *sk, u32 got_seq, u32 exp_seq)
212 {
213 	struct tls_context *tls_ctx = tls_get_ctx(sk);
214 
215 	trace_tls_device_tx_resync_req(sk, got_seq, exp_seq);
216 	WARN_ON(test_and_set_bit(TLS_TX_SYNC_SCHED, &tls_ctx->flags));
217 }
218 EXPORT_SYMBOL_GPL(tls_offload_tx_resync_request);
219 
tls_device_resync_tx(struct sock * sk,struct tls_context * tls_ctx,u32 seq)220 static void tls_device_resync_tx(struct sock *sk, struct tls_context *tls_ctx,
221 				 u32 seq)
222 {
223 	struct net_device *netdev;
224 	struct sk_buff *skb;
225 	int err = 0;
226 	u8 *rcd_sn;
227 
228 	skb = tcp_write_queue_tail(sk);
229 	if (skb)
230 		TCP_SKB_CB(skb)->eor = 1;
231 
232 	rcd_sn = tls_ctx->tx.rec_seq;
233 
234 	trace_tls_device_tx_resync_send(sk, seq, rcd_sn);
235 	down_read(&device_offload_lock);
236 	netdev = tls_ctx->netdev;
237 	if (netdev)
238 		err = netdev->tlsdev_ops->tls_dev_resync(netdev, sk, seq,
239 							 rcd_sn,
240 							 TLS_OFFLOAD_CTX_DIR_TX);
241 	up_read(&device_offload_lock);
242 	if (err)
243 		return;
244 
245 	clear_bit_unlock(TLS_TX_SYNC_SCHED, &tls_ctx->flags);
246 }
247 
tls_append_frag(struct tls_record_info * record,struct page_frag * pfrag,int size)248 static void tls_append_frag(struct tls_record_info *record,
249 			    struct page_frag *pfrag,
250 			    int size)
251 {
252 	skb_frag_t *frag;
253 
254 	frag = &record->frags[record->num_frags - 1];
255 	if (skb_frag_page(frag) == pfrag->page &&
256 	    skb_frag_off(frag) + skb_frag_size(frag) == pfrag->offset) {
257 		skb_frag_size_add(frag, size);
258 	} else {
259 		++frag;
260 		__skb_frag_set_page(frag, pfrag->page);
261 		skb_frag_off_set(frag, pfrag->offset);
262 		skb_frag_size_set(frag, size);
263 		++record->num_frags;
264 		get_page(pfrag->page);
265 	}
266 
267 	pfrag->offset += size;
268 	record->len += size;
269 }
270 
tls_push_record(struct sock * sk,struct tls_context * ctx,struct tls_offload_context_tx * offload_ctx,struct tls_record_info * record,int flags)271 static int tls_push_record(struct sock *sk,
272 			   struct tls_context *ctx,
273 			   struct tls_offload_context_tx *offload_ctx,
274 			   struct tls_record_info *record,
275 			   int flags)
276 {
277 	struct tls_prot_info *prot = &ctx->prot_info;
278 	struct tcp_sock *tp = tcp_sk(sk);
279 	skb_frag_t *frag;
280 	int i;
281 
282 	record->end_seq = tp->write_seq + record->len;
283 	list_add_tail_rcu(&record->list, &offload_ctx->records_list);
284 	offload_ctx->open_record = NULL;
285 
286 	if (test_bit(TLS_TX_SYNC_SCHED, &ctx->flags))
287 		tls_device_resync_tx(sk, ctx, tp->write_seq);
288 
289 	tls_advance_record_sn(sk, prot, &ctx->tx);
290 
291 	for (i = 0; i < record->num_frags; i++) {
292 		frag = &record->frags[i];
293 		sg_unmark_end(&offload_ctx->sg_tx_data[i]);
294 		sg_set_page(&offload_ctx->sg_tx_data[i], skb_frag_page(frag),
295 			    skb_frag_size(frag), skb_frag_off(frag));
296 		sk_mem_charge(sk, skb_frag_size(frag));
297 		get_page(skb_frag_page(frag));
298 	}
299 	sg_mark_end(&offload_ctx->sg_tx_data[record->num_frags - 1]);
300 
301 	/* all ready, send */
302 	return tls_push_sg(sk, ctx, offload_ctx->sg_tx_data, 0, flags);
303 }
304 
tls_device_record_close(struct sock * sk,struct tls_context * ctx,struct tls_record_info * record,struct page_frag * pfrag,unsigned char record_type)305 static int tls_device_record_close(struct sock *sk,
306 				   struct tls_context *ctx,
307 				   struct tls_record_info *record,
308 				   struct page_frag *pfrag,
309 				   unsigned char record_type)
310 {
311 	struct tls_prot_info *prot = &ctx->prot_info;
312 	int ret;
313 
314 	/* append tag
315 	 * device will fill in the tag, we just need to append a placeholder
316 	 * use socket memory to improve coalescing (re-using a single buffer
317 	 * increases frag count)
318 	 * if we can't allocate memory now, steal some back from data
319 	 */
320 	if (likely(skb_page_frag_refill(prot->tag_size, pfrag,
321 					sk->sk_allocation))) {
322 		ret = 0;
323 		tls_append_frag(record, pfrag, prot->tag_size);
324 	} else {
325 		ret = prot->tag_size;
326 		if (record->len <= prot->overhead_size)
327 			return -ENOMEM;
328 	}
329 
330 	/* fill prepend */
331 	tls_fill_prepend(ctx, skb_frag_address(&record->frags[0]),
332 			 record->len - prot->overhead_size,
333 			 record_type, prot->version);
334 	return ret;
335 }
336 
tls_create_new_record(struct tls_offload_context_tx * offload_ctx,struct page_frag * pfrag,size_t prepend_size)337 static int tls_create_new_record(struct tls_offload_context_tx *offload_ctx,
338 				 struct page_frag *pfrag,
339 				 size_t prepend_size)
340 {
341 	struct tls_record_info *record;
342 	skb_frag_t *frag;
343 
344 	record = kmalloc(sizeof(*record), GFP_KERNEL);
345 	if (!record)
346 		return -ENOMEM;
347 
348 	frag = &record->frags[0];
349 	__skb_frag_set_page(frag, pfrag->page);
350 	skb_frag_off_set(frag, pfrag->offset);
351 	skb_frag_size_set(frag, prepend_size);
352 
353 	get_page(pfrag->page);
354 	pfrag->offset += prepend_size;
355 
356 	record->num_frags = 1;
357 	record->len = prepend_size;
358 	offload_ctx->open_record = record;
359 	return 0;
360 }
361 
tls_do_allocation(struct sock * sk,struct tls_offload_context_tx * offload_ctx,struct page_frag * pfrag,size_t prepend_size)362 static int tls_do_allocation(struct sock *sk,
363 			     struct tls_offload_context_tx *offload_ctx,
364 			     struct page_frag *pfrag,
365 			     size_t prepend_size)
366 {
367 	int ret;
368 
369 	if (!offload_ctx->open_record) {
370 		if (unlikely(!skb_page_frag_refill(prepend_size, pfrag,
371 						   sk->sk_allocation))) {
372 			READ_ONCE(sk->sk_prot)->enter_memory_pressure(sk);
373 			sk_stream_moderate_sndbuf(sk);
374 			return -ENOMEM;
375 		}
376 
377 		ret = tls_create_new_record(offload_ctx, pfrag, prepend_size);
378 		if (ret)
379 			return ret;
380 
381 		if (pfrag->size > pfrag->offset)
382 			return 0;
383 	}
384 
385 	if (!sk_page_frag_refill(sk, pfrag))
386 		return -ENOMEM;
387 
388 	return 0;
389 }
390 
tls_device_copy_data(void * addr,size_t bytes,struct iov_iter * i)391 static int tls_device_copy_data(void *addr, size_t bytes, struct iov_iter *i)
392 {
393 	size_t pre_copy, nocache;
394 
395 	pre_copy = ~((unsigned long)addr - 1) & (SMP_CACHE_BYTES - 1);
396 	if (pre_copy) {
397 		pre_copy = min(pre_copy, bytes);
398 		if (copy_from_iter(addr, pre_copy, i) != pre_copy)
399 			return -EFAULT;
400 		bytes -= pre_copy;
401 		addr += pre_copy;
402 	}
403 
404 	nocache = round_down(bytes, SMP_CACHE_BYTES);
405 	if (copy_from_iter_nocache(addr, nocache, i) != nocache)
406 		return -EFAULT;
407 	bytes -= nocache;
408 	addr += nocache;
409 
410 	if (bytes && copy_from_iter(addr, bytes, i) != bytes)
411 		return -EFAULT;
412 
413 	return 0;
414 }
415 
tls_push_data(struct sock * sk,struct iov_iter * msg_iter,size_t size,int flags,unsigned char record_type)416 static int tls_push_data(struct sock *sk,
417 			 struct iov_iter *msg_iter,
418 			 size_t size, int flags,
419 			 unsigned char record_type)
420 {
421 	struct tls_context *tls_ctx = tls_get_ctx(sk);
422 	struct tls_prot_info *prot = &tls_ctx->prot_info;
423 	struct tls_offload_context_tx *ctx = tls_offload_ctx_tx(tls_ctx);
424 	struct tls_record_info *record = ctx->open_record;
425 	int tls_push_record_flags;
426 	struct page_frag *pfrag;
427 	size_t orig_size = size;
428 	u32 max_open_record_len;
429 	bool more = false;
430 	bool done = false;
431 	int copy, rc = 0;
432 	long timeo;
433 
434 	if (flags &
435 	    ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL | MSG_SENDPAGE_NOTLAST))
436 		return -EOPNOTSUPP;
437 
438 	if (unlikely(sk->sk_err))
439 		return -sk->sk_err;
440 
441 	flags |= MSG_SENDPAGE_DECRYPTED;
442 	tls_push_record_flags = flags | MSG_SENDPAGE_NOTLAST;
443 
444 	timeo = sock_sndtimeo(sk, flags & MSG_DONTWAIT);
445 	if (tls_is_partially_sent_record(tls_ctx)) {
446 		rc = tls_push_partial_record(sk, tls_ctx, flags);
447 		if (rc < 0)
448 			return rc;
449 	}
450 
451 	pfrag = sk_page_frag(sk);
452 
453 	/* TLS_HEADER_SIZE is not counted as part of the TLS record, and
454 	 * we need to leave room for an authentication tag.
455 	 */
456 	max_open_record_len = TLS_MAX_PAYLOAD_SIZE +
457 			      prot->prepend_size;
458 	do {
459 		rc = tls_do_allocation(sk, ctx, pfrag, prot->prepend_size);
460 		if (unlikely(rc)) {
461 			rc = sk_stream_wait_memory(sk, &timeo);
462 			if (!rc)
463 				continue;
464 
465 			record = ctx->open_record;
466 			if (!record)
467 				break;
468 handle_error:
469 			if (record_type != TLS_RECORD_TYPE_DATA) {
470 				/* avoid sending partial
471 				 * record with type !=
472 				 * application_data
473 				 */
474 				size = orig_size;
475 				destroy_record(record);
476 				ctx->open_record = NULL;
477 			} else if (record->len > prot->prepend_size) {
478 				goto last_record;
479 			}
480 
481 			break;
482 		}
483 
484 		record = ctx->open_record;
485 		copy = min_t(size_t, size, (pfrag->size - pfrag->offset));
486 		copy = min_t(size_t, copy, (max_open_record_len - record->len));
487 
488 		if (copy) {
489 			rc = tls_device_copy_data(page_address(pfrag->page) +
490 						  pfrag->offset, copy, msg_iter);
491 			if (rc)
492 				goto handle_error;
493 			tls_append_frag(record, pfrag, copy);
494 		}
495 
496 		size -= copy;
497 		if (!size) {
498 last_record:
499 			tls_push_record_flags = flags;
500 			if (flags & (MSG_SENDPAGE_NOTLAST | MSG_MORE)) {
501 				more = true;
502 				break;
503 			}
504 
505 			done = true;
506 		}
507 
508 		if (done || record->len >= max_open_record_len ||
509 		    (record->num_frags >= MAX_SKB_FRAGS - 1)) {
510 			rc = tls_device_record_close(sk, tls_ctx, record,
511 						     pfrag, record_type);
512 			if (rc) {
513 				if (rc > 0) {
514 					size += rc;
515 				} else {
516 					size = orig_size;
517 					destroy_record(record);
518 					ctx->open_record = NULL;
519 					break;
520 				}
521 			}
522 
523 			rc = tls_push_record(sk,
524 					     tls_ctx,
525 					     ctx,
526 					     record,
527 					     tls_push_record_flags);
528 			if (rc < 0)
529 				break;
530 		}
531 	} while (!done);
532 
533 	tls_ctx->pending_open_record_frags = more;
534 
535 	if (orig_size - size > 0)
536 		rc = orig_size - size;
537 
538 	return rc;
539 }
540 
tls_device_sendmsg(struct sock * sk,struct msghdr * msg,size_t size)541 int tls_device_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
542 {
543 	unsigned char record_type = TLS_RECORD_TYPE_DATA;
544 	struct tls_context *tls_ctx = tls_get_ctx(sk);
545 	int rc;
546 
547 	mutex_lock(&tls_ctx->tx_lock);
548 	lock_sock(sk);
549 
550 	if (unlikely(msg->msg_controllen)) {
551 		rc = tls_proccess_cmsg(sk, msg, &record_type);
552 		if (rc)
553 			goto out;
554 	}
555 
556 	rc = tls_push_data(sk, &msg->msg_iter, size,
557 			   msg->msg_flags, record_type);
558 
559 out:
560 	release_sock(sk);
561 	mutex_unlock(&tls_ctx->tx_lock);
562 	return rc;
563 }
564 
tls_device_sendpage(struct sock * sk,struct page * page,int offset,size_t size,int flags)565 int tls_device_sendpage(struct sock *sk, struct page *page,
566 			int offset, size_t size, int flags)
567 {
568 	struct tls_context *tls_ctx = tls_get_ctx(sk);
569 	struct iov_iter	msg_iter;
570 	char *kaddr;
571 	struct kvec iov;
572 	int rc;
573 
574 	if (flags & MSG_SENDPAGE_NOTLAST)
575 		flags |= MSG_MORE;
576 
577 	mutex_lock(&tls_ctx->tx_lock);
578 	lock_sock(sk);
579 
580 	if (flags & MSG_OOB) {
581 		rc = -EOPNOTSUPP;
582 		goto out;
583 	}
584 
585 	kaddr = kmap(page);
586 	iov.iov_base = kaddr + offset;
587 	iov.iov_len = size;
588 	iov_iter_kvec(&msg_iter, WRITE, &iov, 1, size);
589 	rc = tls_push_data(sk, &msg_iter, size,
590 			   flags, TLS_RECORD_TYPE_DATA);
591 	kunmap(page);
592 
593 out:
594 	release_sock(sk);
595 	mutex_unlock(&tls_ctx->tx_lock);
596 	return rc;
597 }
598 
tls_get_record(struct tls_offload_context_tx * context,u32 seq,u64 * p_record_sn)599 struct tls_record_info *tls_get_record(struct tls_offload_context_tx *context,
600 				       u32 seq, u64 *p_record_sn)
601 {
602 	u64 record_sn = context->hint_record_sn;
603 	struct tls_record_info *info, *last;
604 
605 	info = context->retransmit_hint;
606 	if (!info ||
607 	    before(seq, info->end_seq - info->len)) {
608 		/* if retransmit_hint is irrelevant start
609 		 * from the beggining of the list
610 		 */
611 		info = list_first_entry_or_null(&context->records_list,
612 						struct tls_record_info, list);
613 		if (!info)
614 			return NULL;
615 		/* send the start_marker record if seq number is before the
616 		 * tls offload start marker sequence number. This record is
617 		 * required to handle TCP packets which are before TLS offload
618 		 * started.
619 		 *  And if it's not start marker, look if this seq number
620 		 * belongs to the list.
621 		 */
622 		if (likely(!tls_record_is_start_marker(info))) {
623 			/* we have the first record, get the last record to see
624 			 * if this seq number belongs to the list.
625 			 */
626 			last = list_last_entry(&context->records_list,
627 					       struct tls_record_info, list);
628 
629 			if (!between(seq, tls_record_start_seq(info),
630 				     last->end_seq))
631 				return NULL;
632 		}
633 		record_sn = context->unacked_record_sn;
634 	}
635 
636 	/* We just need the _rcu for the READ_ONCE() */
637 	rcu_read_lock();
638 	list_for_each_entry_from_rcu(info, &context->records_list, list) {
639 		if (before(seq, info->end_seq)) {
640 			if (!context->retransmit_hint ||
641 			    after(info->end_seq,
642 				  context->retransmit_hint->end_seq)) {
643 				context->hint_record_sn = record_sn;
644 				context->retransmit_hint = info;
645 			}
646 			*p_record_sn = record_sn;
647 			goto exit_rcu_unlock;
648 		}
649 		record_sn++;
650 	}
651 	info = NULL;
652 
653 exit_rcu_unlock:
654 	rcu_read_unlock();
655 	return info;
656 }
657 EXPORT_SYMBOL(tls_get_record);
658 
tls_device_push_pending_record(struct sock * sk,int flags)659 static int tls_device_push_pending_record(struct sock *sk, int flags)
660 {
661 	struct iov_iter	msg_iter;
662 
663 	iov_iter_kvec(&msg_iter, WRITE, NULL, 0, 0);
664 	return tls_push_data(sk, &msg_iter, 0, flags, TLS_RECORD_TYPE_DATA);
665 }
666 
tls_device_write_space(struct sock * sk,struct tls_context * ctx)667 void tls_device_write_space(struct sock *sk, struct tls_context *ctx)
668 {
669 	if (tls_is_partially_sent_record(ctx)) {
670 		gfp_t sk_allocation = sk->sk_allocation;
671 
672 		WARN_ON_ONCE(sk->sk_write_pending);
673 
674 		sk->sk_allocation = GFP_ATOMIC;
675 		tls_push_partial_record(sk, ctx,
676 					MSG_DONTWAIT | MSG_NOSIGNAL |
677 					MSG_SENDPAGE_DECRYPTED);
678 		sk->sk_allocation = sk_allocation;
679 	}
680 }
681 
tls_device_resync_rx(struct tls_context * tls_ctx,struct sock * sk,u32 seq,u8 * rcd_sn)682 static void tls_device_resync_rx(struct tls_context *tls_ctx,
683 				 struct sock *sk, u32 seq, u8 *rcd_sn)
684 {
685 	struct tls_offload_context_rx *rx_ctx = tls_offload_ctx_rx(tls_ctx);
686 	struct net_device *netdev;
687 
688 	trace_tls_device_rx_resync_send(sk, seq, rcd_sn, rx_ctx->resync_type);
689 	rcu_read_lock();
690 	netdev = READ_ONCE(tls_ctx->netdev);
691 	if (netdev)
692 		netdev->tlsdev_ops->tls_dev_resync(netdev, sk, seq, rcd_sn,
693 						   TLS_OFFLOAD_CTX_DIR_RX);
694 	rcu_read_unlock();
695 	TLS_INC_STATS(sock_net(sk), LINUX_MIB_TLSRXDEVICERESYNC);
696 }
697 
698 static bool
tls_device_rx_resync_async(struct tls_offload_resync_async * resync_async,s64 resync_req,u32 * seq,u16 * rcd_delta)699 tls_device_rx_resync_async(struct tls_offload_resync_async *resync_async,
700 			   s64 resync_req, u32 *seq, u16 *rcd_delta)
701 {
702 	u32 is_async = resync_req & RESYNC_REQ_ASYNC;
703 	u32 req_seq = resync_req >> 32;
704 	u32 req_end = req_seq + ((resync_req >> 16) & 0xffff);
705 	u16 i;
706 
707 	*rcd_delta = 0;
708 
709 	if (is_async) {
710 		/* shouldn't get to wraparound:
711 		 * too long in async stage, something bad happened
712 		 */
713 		if (WARN_ON_ONCE(resync_async->rcd_delta == USHRT_MAX))
714 			return false;
715 
716 		/* asynchronous stage: log all headers seq such that
717 		 * req_seq <= seq <= end_seq, and wait for real resync request
718 		 */
719 		if (before(*seq, req_seq))
720 			return false;
721 		if (!after(*seq, req_end) &&
722 		    resync_async->loglen < TLS_DEVICE_RESYNC_ASYNC_LOGMAX)
723 			resync_async->log[resync_async->loglen++] = *seq;
724 
725 		resync_async->rcd_delta++;
726 
727 		return false;
728 	}
729 
730 	/* synchronous stage: check against the logged entries and
731 	 * proceed to check the next entries if no match was found
732 	 */
733 	for (i = 0; i < resync_async->loglen; i++)
734 		if (req_seq == resync_async->log[i] &&
735 		    atomic64_try_cmpxchg(&resync_async->req, &resync_req, 0)) {
736 			*rcd_delta = resync_async->rcd_delta - i;
737 			*seq = req_seq;
738 			resync_async->loglen = 0;
739 			resync_async->rcd_delta = 0;
740 			return true;
741 		}
742 
743 	resync_async->loglen = 0;
744 	resync_async->rcd_delta = 0;
745 
746 	if (req_seq == *seq &&
747 	    atomic64_try_cmpxchg(&resync_async->req,
748 				 &resync_req, 0))
749 		return true;
750 
751 	return false;
752 }
753 
tls_device_rx_resync_new_rec(struct sock * sk,u32 rcd_len,u32 seq)754 void tls_device_rx_resync_new_rec(struct sock *sk, u32 rcd_len, u32 seq)
755 {
756 	struct tls_context *tls_ctx = tls_get_ctx(sk);
757 	struct tls_offload_context_rx *rx_ctx;
758 	u8 rcd_sn[TLS_MAX_REC_SEQ_SIZE];
759 	u32 sock_data, is_req_pending;
760 	struct tls_prot_info *prot;
761 	s64 resync_req;
762 	u16 rcd_delta;
763 	u32 req_seq;
764 
765 	if (tls_ctx->rx_conf != TLS_HW)
766 		return;
767 	if (unlikely(test_bit(TLS_RX_DEV_DEGRADED, &tls_ctx->flags)))
768 		return;
769 
770 	prot = &tls_ctx->prot_info;
771 	rx_ctx = tls_offload_ctx_rx(tls_ctx);
772 	memcpy(rcd_sn, tls_ctx->rx.rec_seq, prot->rec_seq_size);
773 
774 	switch (rx_ctx->resync_type) {
775 	case TLS_OFFLOAD_SYNC_TYPE_DRIVER_REQ:
776 		resync_req = atomic64_read(&rx_ctx->resync_req);
777 		req_seq = resync_req >> 32;
778 		seq += TLS_HEADER_SIZE - 1;
779 		is_req_pending = resync_req;
780 
781 		if (likely(!is_req_pending) || req_seq != seq ||
782 		    !atomic64_try_cmpxchg(&rx_ctx->resync_req, &resync_req, 0))
783 			return;
784 		break;
785 	case TLS_OFFLOAD_SYNC_TYPE_CORE_NEXT_HINT:
786 		if (likely(!rx_ctx->resync_nh_do_now))
787 			return;
788 
789 		/* head of next rec is already in, note that the sock_inq will
790 		 * include the currently parsed message when called from parser
791 		 */
792 		sock_data = tcp_inq(sk);
793 		if (sock_data > rcd_len) {
794 			trace_tls_device_rx_resync_nh_delay(sk, sock_data,
795 							    rcd_len);
796 			return;
797 		}
798 
799 		rx_ctx->resync_nh_do_now = 0;
800 		seq += rcd_len;
801 		tls_bigint_increment(rcd_sn, prot->rec_seq_size);
802 		break;
803 	case TLS_OFFLOAD_SYNC_TYPE_DRIVER_REQ_ASYNC:
804 		resync_req = atomic64_read(&rx_ctx->resync_async->req);
805 		is_req_pending = resync_req;
806 		if (likely(!is_req_pending))
807 			return;
808 
809 		if (!tls_device_rx_resync_async(rx_ctx->resync_async,
810 						resync_req, &seq, &rcd_delta))
811 			return;
812 		tls_bigint_subtract(rcd_sn, rcd_delta);
813 		break;
814 	}
815 
816 	tls_device_resync_rx(tls_ctx, sk, seq, rcd_sn);
817 }
818 
tls_device_core_ctrl_rx_resync(struct tls_context * tls_ctx,struct tls_offload_context_rx * ctx,struct sock * sk,struct sk_buff * skb)819 static void tls_device_core_ctrl_rx_resync(struct tls_context *tls_ctx,
820 					   struct tls_offload_context_rx *ctx,
821 					   struct sock *sk, struct sk_buff *skb)
822 {
823 	struct strp_msg *rxm;
824 
825 	/* device will request resyncs by itself based on stream scan */
826 	if (ctx->resync_type != TLS_OFFLOAD_SYNC_TYPE_CORE_NEXT_HINT)
827 		return;
828 	/* already scheduled */
829 	if (ctx->resync_nh_do_now)
830 		return;
831 	/* seen decrypted fragments since last fully-failed record */
832 	if (ctx->resync_nh_reset) {
833 		ctx->resync_nh_reset = 0;
834 		ctx->resync_nh.decrypted_failed = 1;
835 		ctx->resync_nh.decrypted_tgt = TLS_DEVICE_RESYNC_NH_START_IVAL;
836 		return;
837 	}
838 
839 	if (++ctx->resync_nh.decrypted_failed <= ctx->resync_nh.decrypted_tgt)
840 		return;
841 
842 	/* doing resync, bump the next target in case it fails */
843 	if (ctx->resync_nh.decrypted_tgt < TLS_DEVICE_RESYNC_NH_MAX_IVAL)
844 		ctx->resync_nh.decrypted_tgt *= 2;
845 	else
846 		ctx->resync_nh.decrypted_tgt += TLS_DEVICE_RESYNC_NH_MAX_IVAL;
847 
848 	rxm = strp_msg(skb);
849 
850 	/* head of next rec is already in, parser will sync for us */
851 	if (tcp_inq(sk) > rxm->full_len) {
852 		trace_tls_device_rx_resync_nh_schedule(sk);
853 		ctx->resync_nh_do_now = 1;
854 	} else {
855 		struct tls_prot_info *prot = &tls_ctx->prot_info;
856 		u8 rcd_sn[TLS_MAX_REC_SEQ_SIZE];
857 
858 		memcpy(rcd_sn, tls_ctx->rx.rec_seq, prot->rec_seq_size);
859 		tls_bigint_increment(rcd_sn, prot->rec_seq_size);
860 
861 		tls_device_resync_rx(tls_ctx, sk, tcp_sk(sk)->copied_seq,
862 				     rcd_sn);
863 	}
864 }
865 
tls_device_reencrypt(struct sock * sk,struct sk_buff * skb)866 static int tls_device_reencrypt(struct sock *sk, struct sk_buff *skb)
867 {
868 	struct strp_msg *rxm = strp_msg(skb);
869 	int err = 0, offset = rxm->offset, copy, nsg, data_len, pos;
870 	struct sk_buff *skb_iter, *unused;
871 	struct scatterlist sg[1];
872 	char *orig_buf, *buf;
873 
874 	orig_buf = kmalloc(rxm->full_len + TLS_HEADER_SIZE +
875 			   TLS_CIPHER_AES_GCM_128_IV_SIZE, sk->sk_allocation);
876 	if (!orig_buf)
877 		return -ENOMEM;
878 	buf = orig_buf;
879 
880 	nsg = skb_cow_data(skb, 0, &unused);
881 	if (unlikely(nsg < 0)) {
882 		err = nsg;
883 		goto free_buf;
884 	}
885 
886 	sg_init_table(sg, 1);
887 	sg_set_buf(&sg[0], buf,
888 		   rxm->full_len + TLS_HEADER_SIZE +
889 		   TLS_CIPHER_AES_GCM_128_IV_SIZE);
890 	err = skb_copy_bits(skb, offset, buf,
891 			    TLS_HEADER_SIZE + TLS_CIPHER_AES_GCM_128_IV_SIZE);
892 	if (err)
893 		goto free_buf;
894 
895 	/* We are interested only in the decrypted data not the auth */
896 	err = decrypt_skb(sk, skb, sg);
897 	if (err != -EBADMSG)
898 		goto free_buf;
899 	else
900 		err = 0;
901 
902 	data_len = rxm->full_len - TLS_CIPHER_AES_GCM_128_TAG_SIZE;
903 
904 	if (skb_pagelen(skb) > offset) {
905 		copy = min_t(int, skb_pagelen(skb) - offset, data_len);
906 
907 		if (skb->decrypted) {
908 			err = skb_store_bits(skb, offset, buf, copy);
909 			if (err)
910 				goto free_buf;
911 		}
912 
913 		offset += copy;
914 		buf += copy;
915 	}
916 
917 	pos = skb_pagelen(skb);
918 	skb_walk_frags(skb, skb_iter) {
919 		int frag_pos;
920 
921 		/* Practically all frags must belong to msg if reencrypt
922 		 * is needed with current strparser and coalescing logic,
923 		 * but strparser may "get optimized", so let's be safe.
924 		 */
925 		if (pos + skb_iter->len <= offset)
926 			goto done_with_frag;
927 		if (pos >= data_len + rxm->offset)
928 			break;
929 
930 		frag_pos = offset - pos;
931 		copy = min_t(int, skb_iter->len - frag_pos,
932 			     data_len + rxm->offset - offset);
933 
934 		if (skb_iter->decrypted) {
935 			err = skb_store_bits(skb_iter, frag_pos, buf, copy);
936 			if (err)
937 				goto free_buf;
938 		}
939 
940 		offset += copy;
941 		buf += copy;
942 done_with_frag:
943 		pos += skb_iter->len;
944 	}
945 
946 free_buf:
947 	kfree(orig_buf);
948 	return err;
949 }
950 
tls_device_decrypted(struct sock * sk,struct tls_context * tls_ctx,struct sk_buff * skb,struct strp_msg * rxm)951 int tls_device_decrypted(struct sock *sk, struct tls_context *tls_ctx,
952 			 struct sk_buff *skb, struct strp_msg *rxm)
953 {
954 	struct tls_offload_context_rx *ctx = tls_offload_ctx_rx(tls_ctx);
955 	int is_decrypted = skb->decrypted;
956 	int is_encrypted = !is_decrypted;
957 	struct sk_buff *skb_iter;
958 
959 	/* Check if all the data is decrypted already */
960 	skb_walk_frags(skb, skb_iter) {
961 		is_decrypted &= skb_iter->decrypted;
962 		is_encrypted &= !skb_iter->decrypted;
963 	}
964 
965 	trace_tls_device_decrypted(sk, tcp_sk(sk)->copied_seq - rxm->full_len,
966 				   tls_ctx->rx.rec_seq, rxm->full_len,
967 				   is_encrypted, is_decrypted);
968 
969 	ctx->sw.decrypted |= is_decrypted;
970 
971 	if (unlikely(test_bit(TLS_RX_DEV_DEGRADED, &tls_ctx->flags))) {
972 		if (likely(is_encrypted || is_decrypted))
973 			return 0;
974 
975 		/* After tls_device_down disables the offload, the next SKB will
976 		 * likely have initial fragments decrypted, and final ones not
977 		 * decrypted. We need to reencrypt that single SKB.
978 		 */
979 		return tls_device_reencrypt(sk, skb);
980 	}
981 
982 	/* Return immediately if the record is either entirely plaintext or
983 	 * entirely ciphertext. Otherwise handle reencrypt partially decrypted
984 	 * record.
985 	 */
986 	if (is_decrypted) {
987 		ctx->resync_nh_reset = 1;
988 		return 0;
989 	}
990 	if (is_encrypted) {
991 		tls_device_core_ctrl_rx_resync(tls_ctx, ctx, sk, skb);
992 		return 0;
993 	}
994 
995 	ctx->resync_nh_reset = 1;
996 	return tls_device_reencrypt(sk, skb);
997 }
998 
tls_device_attach(struct tls_context * ctx,struct sock * sk,struct net_device * netdev)999 static void tls_device_attach(struct tls_context *ctx, struct sock *sk,
1000 			      struct net_device *netdev)
1001 {
1002 	if (sk->sk_destruct != tls_device_sk_destruct) {
1003 		refcount_set(&ctx->refcount, 1);
1004 		dev_hold(netdev);
1005 		ctx->netdev = netdev;
1006 		spin_lock_irq(&tls_device_lock);
1007 		list_add_tail(&ctx->list, &tls_device_list);
1008 		spin_unlock_irq(&tls_device_lock);
1009 
1010 		ctx->sk_destruct = sk->sk_destruct;
1011 		smp_store_release(&sk->sk_destruct, tls_device_sk_destruct);
1012 	}
1013 }
1014 
tls_set_device_offload(struct sock * sk,struct tls_context * ctx)1015 int tls_set_device_offload(struct sock *sk, struct tls_context *ctx)
1016 {
1017 	u16 nonce_size, tag_size, iv_size, rec_seq_size;
1018 	struct tls_context *tls_ctx = tls_get_ctx(sk);
1019 	struct tls_prot_info *prot = &tls_ctx->prot_info;
1020 	struct tls_record_info *start_marker_record;
1021 	struct tls_offload_context_tx *offload_ctx;
1022 	struct tls_crypto_info *crypto_info;
1023 	struct net_device *netdev;
1024 	char *iv, *rec_seq;
1025 	struct sk_buff *skb;
1026 	__be64 rcd_sn;
1027 	int rc;
1028 
1029 	if (!ctx)
1030 		return -EINVAL;
1031 
1032 	if (ctx->priv_ctx_tx)
1033 		return -EEXIST;
1034 
1035 	start_marker_record = kmalloc(sizeof(*start_marker_record), GFP_KERNEL);
1036 	if (!start_marker_record)
1037 		return -ENOMEM;
1038 
1039 	offload_ctx = kzalloc(TLS_OFFLOAD_CONTEXT_SIZE_TX, GFP_KERNEL);
1040 	if (!offload_ctx) {
1041 		rc = -ENOMEM;
1042 		goto free_marker_record;
1043 	}
1044 
1045 	crypto_info = &ctx->crypto_send.info;
1046 	if (crypto_info->version != TLS_1_2_VERSION) {
1047 		rc = -EOPNOTSUPP;
1048 		goto free_offload_ctx;
1049 	}
1050 
1051 	switch (crypto_info->cipher_type) {
1052 	case TLS_CIPHER_AES_GCM_128:
1053 		nonce_size = TLS_CIPHER_AES_GCM_128_IV_SIZE;
1054 		tag_size = TLS_CIPHER_AES_GCM_128_TAG_SIZE;
1055 		iv_size = TLS_CIPHER_AES_GCM_128_IV_SIZE;
1056 		iv = ((struct tls12_crypto_info_aes_gcm_128 *)crypto_info)->iv;
1057 		rec_seq_size = TLS_CIPHER_AES_GCM_128_REC_SEQ_SIZE;
1058 		rec_seq =
1059 		 ((struct tls12_crypto_info_aes_gcm_128 *)crypto_info)->rec_seq;
1060 		break;
1061 	default:
1062 		rc = -EINVAL;
1063 		goto free_offload_ctx;
1064 	}
1065 
1066 	/* Sanity-check the rec_seq_size for stack allocations */
1067 	if (rec_seq_size > TLS_MAX_REC_SEQ_SIZE) {
1068 		rc = -EINVAL;
1069 		goto free_offload_ctx;
1070 	}
1071 
1072 	prot->version = crypto_info->version;
1073 	prot->cipher_type = crypto_info->cipher_type;
1074 	prot->prepend_size = TLS_HEADER_SIZE + nonce_size;
1075 	prot->tag_size = tag_size;
1076 	prot->overhead_size = prot->prepend_size + prot->tag_size;
1077 	prot->iv_size = iv_size;
1078 	ctx->tx.iv = kmalloc(iv_size + TLS_CIPHER_AES_GCM_128_SALT_SIZE,
1079 			     GFP_KERNEL);
1080 	if (!ctx->tx.iv) {
1081 		rc = -ENOMEM;
1082 		goto free_offload_ctx;
1083 	}
1084 
1085 	memcpy(ctx->tx.iv + TLS_CIPHER_AES_GCM_128_SALT_SIZE, iv, iv_size);
1086 
1087 	prot->rec_seq_size = rec_seq_size;
1088 	ctx->tx.rec_seq = kmemdup(rec_seq, rec_seq_size, GFP_KERNEL);
1089 	if (!ctx->tx.rec_seq) {
1090 		rc = -ENOMEM;
1091 		goto free_iv;
1092 	}
1093 
1094 	rc = tls_sw_fallback_init(sk, offload_ctx, crypto_info);
1095 	if (rc)
1096 		goto free_rec_seq;
1097 
1098 	/* start at rec_seq - 1 to account for the start marker record */
1099 	memcpy(&rcd_sn, ctx->tx.rec_seq, sizeof(rcd_sn));
1100 	offload_ctx->unacked_record_sn = be64_to_cpu(rcd_sn) - 1;
1101 
1102 	start_marker_record->end_seq = tcp_sk(sk)->write_seq;
1103 	start_marker_record->len = 0;
1104 	start_marker_record->num_frags = 0;
1105 
1106 	INIT_LIST_HEAD(&offload_ctx->records_list);
1107 	list_add_tail(&start_marker_record->list, &offload_ctx->records_list);
1108 	spin_lock_init(&offload_ctx->lock);
1109 	sg_init_table(offload_ctx->sg_tx_data,
1110 		      ARRAY_SIZE(offload_ctx->sg_tx_data));
1111 
1112 	clean_acked_data_enable(inet_csk(sk), &tls_icsk_clean_acked);
1113 	ctx->push_pending_record = tls_device_push_pending_record;
1114 
1115 	/* TLS offload is greatly simplified if we don't send
1116 	 * SKBs where only part of the payload needs to be encrypted.
1117 	 * So mark the last skb in the write queue as end of record.
1118 	 */
1119 	skb = tcp_write_queue_tail(sk);
1120 	if (skb)
1121 		TCP_SKB_CB(skb)->eor = 1;
1122 
1123 	netdev = get_netdev_for_sock(sk);
1124 	if (!netdev) {
1125 		pr_err_ratelimited("%s: netdev not found\n", __func__);
1126 		rc = -EINVAL;
1127 		goto disable_cad;
1128 	}
1129 
1130 	if (!(netdev->features & NETIF_F_HW_TLS_TX)) {
1131 		rc = -EOPNOTSUPP;
1132 		goto release_netdev;
1133 	}
1134 
1135 	/* Avoid offloading if the device is down
1136 	 * We don't want to offload new flows after
1137 	 * the NETDEV_DOWN event
1138 	 *
1139 	 * device_offload_lock is taken in tls_devices's NETDEV_DOWN
1140 	 * handler thus protecting from the device going down before
1141 	 * ctx was added to tls_device_list.
1142 	 */
1143 	down_read(&device_offload_lock);
1144 	if (!(netdev->flags & IFF_UP)) {
1145 		rc = -EINVAL;
1146 		goto release_lock;
1147 	}
1148 
1149 	ctx->priv_ctx_tx = offload_ctx;
1150 	rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_TX,
1151 					     &ctx->crypto_send.info,
1152 					     tcp_sk(sk)->write_seq);
1153 	trace_tls_device_offload_set(sk, TLS_OFFLOAD_CTX_DIR_TX,
1154 				     tcp_sk(sk)->write_seq, rec_seq, rc);
1155 	if (rc)
1156 		goto release_lock;
1157 
1158 	tls_device_attach(ctx, sk, netdev);
1159 	up_read(&device_offload_lock);
1160 
1161 	/* following this assignment tls_is_sk_tx_device_offloaded
1162 	 * will return true and the context might be accessed
1163 	 * by the netdev's xmit function.
1164 	 */
1165 	smp_store_release(&sk->sk_validate_xmit_skb, tls_validate_xmit_skb);
1166 	dev_put(netdev);
1167 
1168 	return 0;
1169 
1170 release_lock:
1171 	up_read(&device_offload_lock);
1172 release_netdev:
1173 	dev_put(netdev);
1174 disable_cad:
1175 	clean_acked_data_disable(inet_csk(sk));
1176 	crypto_free_aead(offload_ctx->aead_send);
1177 free_rec_seq:
1178 	kfree(ctx->tx.rec_seq);
1179 free_iv:
1180 	kfree(ctx->tx.iv);
1181 free_offload_ctx:
1182 	kfree(offload_ctx);
1183 	ctx->priv_ctx_tx = NULL;
1184 free_marker_record:
1185 	kfree(start_marker_record);
1186 	return rc;
1187 }
1188 
tls_set_device_offload_rx(struct sock * sk,struct tls_context * ctx)1189 int tls_set_device_offload_rx(struct sock *sk, struct tls_context *ctx)
1190 {
1191 	struct tls12_crypto_info_aes_gcm_128 *info;
1192 	struct tls_offload_context_rx *context;
1193 	struct net_device *netdev;
1194 	int rc = 0;
1195 
1196 	if (ctx->crypto_recv.info.version != TLS_1_2_VERSION)
1197 		return -EOPNOTSUPP;
1198 
1199 	netdev = get_netdev_for_sock(sk);
1200 	if (!netdev) {
1201 		pr_err_ratelimited("%s: netdev not found\n", __func__);
1202 		return -EINVAL;
1203 	}
1204 
1205 	if (!(netdev->features & NETIF_F_HW_TLS_RX)) {
1206 		rc = -EOPNOTSUPP;
1207 		goto release_netdev;
1208 	}
1209 
1210 	/* Avoid offloading if the device is down
1211 	 * We don't want to offload new flows after
1212 	 * the NETDEV_DOWN event
1213 	 *
1214 	 * device_offload_lock is taken in tls_devices's NETDEV_DOWN
1215 	 * handler thus protecting from the device going down before
1216 	 * ctx was added to tls_device_list.
1217 	 */
1218 	down_read(&device_offload_lock);
1219 	if (!(netdev->flags & IFF_UP)) {
1220 		rc = -EINVAL;
1221 		goto release_lock;
1222 	}
1223 
1224 	context = kzalloc(TLS_OFFLOAD_CONTEXT_SIZE_RX, GFP_KERNEL);
1225 	if (!context) {
1226 		rc = -ENOMEM;
1227 		goto release_lock;
1228 	}
1229 	context->resync_nh_reset = 1;
1230 
1231 	ctx->priv_ctx_rx = context;
1232 	rc = tls_set_sw_offload(sk, ctx, 0);
1233 	if (rc)
1234 		goto release_ctx;
1235 
1236 	rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_RX,
1237 					     &ctx->crypto_recv.info,
1238 					     tcp_sk(sk)->copied_seq);
1239 	info = (void *)&ctx->crypto_recv.info;
1240 	trace_tls_device_offload_set(sk, TLS_OFFLOAD_CTX_DIR_RX,
1241 				     tcp_sk(sk)->copied_seq, info->rec_seq, rc);
1242 	if (rc)
1243 		goto free_sw_resources;
1244 
1245 	tls_device_attach(ctx, sk, netdev);
1246 	up_read(&device_offload_lock);
1247 
1248 	dev_put(netdev);
1249 
1250 	return 0;
1251 
1252 free_sw_resources:
1253 	up_read(&device_offload_lock);
1254 	tls_sw_free_resources_rx(sk);
1255 	down_read(&device_offload_lock);
1256 release_ctx:
1257 	ctx->priv_ctx_rx = NULL;
1258 release_lock:
1259 	up_read(&device_offload_lock);
1260 release_netdev:
1261 	dev_put(netdev);
1262 	return rc;
1263 }
1264 
tls_device_offload_cleanup_rx(struct sock * sk)1265 void tls_device_offload_cleanup_rx(struct sock *sk)
1266 {
1267 	struct tls_context *tls_ctx = tls_get_ctx(sk);
1268 	struct net_device *netdev;
1269 
1270 	down_read(&device_offload_lock);
1271 	netdev = tls_ctx->netdev;
1272 	if (!netdev)
1273 		goto out;
1274 
1275 	netdev->tlsdev_ops->tls_dev_del(netdev, tls_ctx,
1276 					TLS_OFFLOAD_CTX_DIR_RX);
1277 
1278 	if (tls_ctx->tx_conf != TLS_HW) {
1279 		dev_put(netdev);
1280 		tls_ctx->netdev = NULL;
1281 	} else {
1282 		set_bit(TLS_RX_DEV_CLOSED, &tls_ctx->flags);
1283 	}
1284 out:
1285 	up_read(&device_offload_lock);
1286 	tls_sw_release_resources_rx(sk);
1287 }
1288 
tls_device_down(struct net_device * netdev)1289 static int tls_device_down(struct net_device *netdev)
1290 {
1291 	struct tls_context *ctx, *tmp;
1292 	unsigned long flags;
1293 	LIST_HEAD(list);
1294 
1295 	/* Request a write lock to block new offload attempts */
1296 	down_write(&device_offload_lock);
1297 
1298 	spin_lock_irqsave(&tls_device_lock, flags);
1299 	list_for_each_entry_safe(ctx, tmp, &tls_device_list, list) {
1300 		if (ctx->netdev != netdev ||
1301 		    !refcount_inc_not_zero(&ctx->refcount))
1302 			continue;
1303 
1304 		list_move(&ctx->list, &list);
1305 	}
1306 	spin_unlock_irqrestore(&tls_device_lock, flags);
1307 
1308 	list_for_each_entry_safe(ctx, tmp, &list, list)	{
1309 		/* Stop offloaded TX and switch to the fallback.
1310 		 * tls_is_sk_tx_device_offloaded will return false.
1311 		 */
1312 		WRITE_ONCE(ctx->sk->sk_validate_xmit_skb, tls_validate_xmit_skb_sw);
1313 
1314 		/* Stop the RX and TX resync.
1315 		 * tls_dev_resync must not be called after tls_dev_del.
1316 		 */
1317 		WRITE_ONCE(ctx->netdev, NULL);
1318 
1319 		/* Start skipping the RX resync logic completely. */
1320 		set_bit(TLS_RX_DEV_DEGRADED, &ctx->flags);
1321 
1322 		/* Sync with inflight packets. After this point:
1323 		 * TX: no non-encrypted packets will be passed to the driver.
1324 		 * RX: resync requests from the driver will be ignored.
1325 		 */
1326 		synchronize_net();
1327 
1328 		/* Release the offload context on the driver side. */
1329 		if (ctx->tx_conf == TLS_HW)
1330 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
1331 							TLS_OFFLOAD_CTX_DIR_TX);
1332 		if (ctx->rx_conf == TLS_HW &&
1333 		    !test_bit(TLS_RX_DEV_CLOSED, &ctx->flags))
1334 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
1335 							TLS_OFFLOAD_CTX_DIR_RX);
1336 
1337 		dev_put(netdev);
1338 
1339 		/* Move the context to a separate list for two reasons:
1340 		 * 1. When the context is deallocated, list_del is called.
1341 		 * 2. It's no longer an offloaded context, so we don't want to
1342 		 *    run offload-specific code on this context.
1343 		 */
1344 		spin_lock_irqsave(&tls_device_lock, flags);
1345 		list_move_tail(&ctx->list, &tls_device_down_list);
1346 		spin_unlock_irqrestore(&tls_device_lock, flags);
1347 
1348 		/* Device contexts for RX and TX will be freed in on sk_destruct
1349 		 * by tls_device_free_ctx. rx_conf and tx_conf stay in TLS_HW.
1350 		 * Now release the ref taken above.
1351 		 */
1352 		if (refcount_dec_and_test(&ctx->refcount)) {
1353 			/* sk_destruct ran after tls_device_down took a ref, and
1354 			 * it returned early. Complete the destruction here.
1355 			 */
1356 			list_del(&ctx->list);
1357 			tls_device_free_ctx(ctx);
1358 		}
1359 	}
1360 
1361 	up_write(&device_offload_lock);
1362 
1363 	flush_work(&tls_device_gc_work);
1364 
1365 	return NOTIFY_DONE;
1366 }
1367 
tls_dev_event(struct notifier_block * this,unsigned long event,void * ptr)1368 static int tls_dev_event(struct notifier_block *this, unsigned long event,
1369 			 void *ptr)
1370 {
1371 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1372 
1373 	if (!dev->tlsdev_ops &&
1374 	    !(dev->features & (NETIF_F_HW_TLS_RX | NETIF_F_HW_TLS_TX)))
1375 		return NOTIFY_DONE;
1376 
1377 	switch (event) {
1378 	case NETDEV_REGISTER:
1379 	case NETDEV_FEAT_CHANGE:
1380 		if ((dev->features & NETIF_F_HW_TLS_RX) &&
1381 		    !dev->tlsdev_ops->tls_dev_resync)
1382 			return NOTIFY_BAD;
1383 
1384 		if  (dev->tlsdev_ops &&
1385 		     dev->tlsdev_ops->tls_dev_add &&
1386 		     dev->tlsdev_ops->tls_dev_del)
1387 			return NOTIFY_DONE;
1388 		else
1389 			return NOTIFY_BAD;
1390 	case NETDEV_DOWN:
1391 		return tls_device_down(dev);
1392 	}
1393 	return NOTIFY_DONE;
1394 }
1395 
1396 static struct notifier_block tls_dev_notifier = {
1397 	.notifier_call	= tls_dev_event,
1398 };
1399 
tls_device_init(void)1400 int __init tls_device_init(void)
1401 {
1402 	return register_netdevice_notifier(&tls_dev_notifier);
1403 }
1404 
tls_device_cleanup(void)1405 void __exit tls_device_cleanup(void)
1406 {
1407 	unregister_netdevice_notifier(&tls_dev_notifier);
1408 	flush_work(&tls_device_gc_work);
1409 	clean_acked_data_flush();
1410 }
1411