xref: /linux/net/tls/tls_device.c (revision 81fa7a69c2174ed8de314b9c231ef30a8718e5e1)
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 /* device_offload_lock is used to synchronize tls_dev_add
42  * against NETDEV_DOWN notifications.
43  */
44 static DECLARE_RWSEM(device_offload_lock);
45 
46 static void tls_device_gc_task(struct work_struct *work);
47 
48 static DECLARE_WORK(tls_device_gc_work, tls_device_gc_task);
49 static LIST_HEAD(tls_device_gc_list);
50 static LIST_HEAD(tls_device_list);
51 static DEFINE_SPINLOCK(tls_device_lock);
52 
53 static void tls_device_free_ctx(struct tls_context *ctx)
54 {
55 	if (ctx->tx_conf == TLS_HW)
56 		kfree(tls_offload_ctx_tx(ctx));
57 
58 	if (ctx->rx_conf == TLS_HW)
59 		kfree(tls_offload_ctx_rx(ctx));
60 
61 	kfree(ctx);
62 }
63 
64 static void tls_device_gc_task(struct work_struct *work)
65 {
66 	struct tls_context *ctx, *tmp;
67 	unsigned long flags;
68 	LIST_HEAD(gc_list);
69 
70 	spin_lock_irqsave(&tls_device_lock, flags);
71 	list_splice_init(&tls_device_gc_list, &gc_list);
72 	spin_unlock_irqrestore(&tls_device_lock, flags);
73 
74 	list_for_each_entry_safe(ctx, tmp, &gc_list, list) {
75 		struct net_device *netdev = ctx->netdev;
76 
77 		if (netdev && ctx->tx_conf == TLS_HW) {
78 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
79 							TLS_OFFLOAD_CTX_DIR_TX);
80 			dev_put(netdev);
81 			ctx->netdev = NULL;
82 		}
83 
84 		list_del(&ctx->list);
85 		tls_device_free_ctx(ctx);
86 	}
87 }
88 
89 static void tls_device_attach(struct tls_context *ctx, struct sock *sk,
90 			      struct net_device *netdev)
91 {
92 	if (sk->sk_destruct != tls_device_sk_destruct) {
93 		refcount_set(&ctx->refcount, 1);
94 		dev_hold(netdev);
95 		ctx->netdev = netdev;
96 		spin_lock_irq(&tls_device_lock);
97 		list_add_tail(&ctx->list, &tls_device_list);
98 		spin_unlock_irq(&tls_device_lock);
99 
100 		ctx->sk_destruct = sk->sk_destruct;
101 		sk->sk_destruct = tls_device_sk_destruct;
102 	}
103 }
104 
105 static void tls_device_queue_ctx_destruction(struct tls_context *ctx)
106 {
107 	unsigned long flags;
108 
109 	spin_lock_irqsave(&tls_device_lock, flags);
110 	list_move_tail(&ctx->list, &tls_device_gc_list);
111 
112 	/* schedule_work inside the spinlock
113 	 * to make sure tls_device_down waits for that work.
114 	 */
115 	schedule_work(&tls_device_gc_work);
116 
117 	spin_unlock_irqrestore(&tls_device_lock, flags);
118 }
119 
120 /* We assume that the socket is already connected */
121 static struct net_device *get_netdev_for_sock(struct sock *sk)
122 {
123 	struct dst_entry *dst = sk_dst_get(sk);
124 	struct net_device *netdev = NULL;
125 
126 	if (likely(dst)) {
127 		netdev = dst->dev;
128 		dev_hold(netdev);
129 	}
130 
131 	dst_release(dst);
132 
133 	return netdev;
134 }
135 
136 static void destroy_record(struct tls_record_info *record)
137 {
138 	int nr_frags = record->num_frags;
139 	skb_frag_t *frag;
140 
141 	while (nr_frags-- > 0) {
142 		frag = &record->frags[nr_frags];
143 		__skb_frag_unref(frag);
144 	}
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 		list_del(&info->list);
178 		destroy_record(info);
179 		deleted_records++;
180 	}
181 
182 	list_for_each_entry_safe(info, temp, &ctx->records_list, list) {
183 		if (before(acked_seq, info->end_seq))
184 			break;
185 		list_del(&info->list);
186 
187 		destroy_record(info);
188 		deleted_records++;
189 	}
190 
191 	ctx->unacked_record_sn += deleted_records;
192 	spin_unlock_irqrestore(&ctx->lock, flags);
193 }
194 
195 /* At this point, there should be no references on this
196  * socket and no in-flight SKBs associated with this
197  * socket, so it is safe to free all the resources.
198  */
199 void tls_device_sk_destruct(struct sock *sk)
200 {
201 	struct tls_context *tls_ctx = tls_get_ctx(sk);
202 	struct tls_offload_context_tx *ctx = tls_offload_ctx_tx(tls_ctx);
203 
204 	tls_ctx->sk_destruct(sk);
205 
206 	if (tls_ctx->tx_conf == TLS_HW) {
207 		if (ctx->open_record)
208 			destroy_record(ctx->open_record);
209 		delete_all_records(ctx);
210 		crypto_free_aead(ctx->aead_send);
211 		clean_acked_data_disable(inet_csk(sk));
212 	}
213 
214 	if (refcount_dec_and_test(&tls_ctx->refcount))
215 		tls_device_queue_ctx_destruction(tls_ctx);
216 }
217 EXPORT_SYMBOL(tls_device_sk_destruct);
218 
219 static void tls_append_frag(struct tls_record_info *record,
220 			    struct page_frag *pfrag,
221 			    int size)
222 {
223 	skb_frag_t *frag;
224 
225 	frag = &record->frags[record->num_frags - 1];
226 	if (frag->page.p == pfrag->page &&
227 	    frag->page_offset + frag->size == pfrag->offset) {
228 		frag->size += size;
229 	} else {
230 		++frag;
231 		frag->page.p = pfrag->page;
232 		frag->page_offset = pfrag->offset;
233 		frag->size = size;
234 		++record->num_frags;
235 		get_page(pfrag->page);
236 	}
237 
238 	pfrag->offset += size;
239 	record->len += size;
240 }
241 
242 static int tls_push_record(struct sock *sk,
243 			   struct tls_context *ctx,
244 			   struct tls_offload_context_tx *offload_ctx,
245 			   struct tls_record_info *record,
246 			   struct page_frag *pfrag,
247 			   int flags,
248 			   unsigned char record_type)
249 {
250 	struct tcp_sock *tp = tcp_sk(sk);
251 	struct page_frag dummy_tag_frag;
252 	skb_frag_t *frag;
253 	int i;
254 
255 	/* fill prepend */
256 	frag = &record->frags[0];
257 	tls_fill_prepend(ctx,
258 			 skb_frag_address(frag),
259 			 record->len - ctx->tx.prepend_size,
260 			 record_type);
261 
262 	/* HW doesn't care about the data in the tag, because it fills it. */
263 	dummy_tag_frag.page = skb_frag_page(frag);
264 	dummy_tag_frag.offset = 0;
265 
266 	tls_append_frag(record, &dummy_tag_frag, ctx->tx.tag_size);
267 	record->end_seq = tp->write_seq + record->len;
268 	spin_lock_irq(&offload_ctx->lock);
269 	list_add_tail(&record->list, &offload_ctx->records_list);
270 	spin_unlock_irq(&offload_ctx->lock);
271 	offload_ctx->open_record = NULL;
272 	set_bit(TLS_PENDING_CLOSED_RECORD, &ctx->flags);
273 	tls_advance_record_sn(sk, &ctx->tx);
274 
275 	for (i = 0; i < record->num_frags; i++) {
276 		frag = &record->frags[i];
277 		sg_unmark_end(&offload_ctx->sg_tx_data[i]);
278 		sg_set_page(&offload_ctx->sg_tx_data[i], skb_frag_page(frag),
279 			    frag->size, frag->page_offset);
280 		sk_mem_charge(sk, frag->size);
281 		get_page(skb_frag_page(frag));
282 	}
283 	sg_mark_end(&offload_ctx->sg_tx_data[record->num_frags - 1]);
284 
285 	/* all ready, send */
286 	return tls_push_sg(sk, ctx, offload_ctx->sg_tx_data, 0, flags);
287 }
288 
289 static int tls_create_new_record(struct tls_offload_context_tx *offload_ctx,
290 				 struct page_frag *pfrag,
291 				 size_t prepend_size)
292 {
293 	struct tls_record_info *record;
294 	skb_frag_t *frag;
295 
296 	record = kmalloc(sizeof(*record), GFP_KERNEL);
297 	if (!record)
298 		return -ENOMEM;
299 
300 	frag = &record->frags[0];
301 	__skb_frag_set_page(frag, pfrag->page);
302 	frag->page_offset = pfrag->offset;
303 	skb_frag_size_set(frag, prepend_size);
304 
305 	get_page(pfrag->page);
306 	pfrag->offset += prepend_size;
307 
308 	record->num_frags = 1;
309 	record->len = prepend_size;
310 	offload_ctx->open_record = record;
311 	return 0;
312 }
313 
314 static int tls_do_allocation(struct sock *sk,
315 			     struct tls_offload_context_tx *offload_ctx,
316 			     struct page_frag *pfrag,
317 			     size_t prepend_size)
318 {
319 	int ret;
320 
321 	if (!offload_ctx->open_record) {
322 		if (unlikely(!skb_page_frag_refill(prepend_size, pfrag,
323 						   sk->sk_allocation))) {
324 			sk->sk_prot->enter_memory_pressure(sk);
325 			sk_stream_moderate_sndbuf(sk);
326 			return -ENOMEM;
327 		}
328 
329 		ret = tls_create_new_record(offload_ctx, pfrag, prepend_size);
330 		if (ret)
331 			return ret;
332 
333 		if (pfrag->size > pfrag->offset)
334 			return 0;
335 	}
336 
337 	if (!sk_page_frag_refill(sk, pfrag))
338 		return -ENOMEM;
339 
340 	return 0;
341 }
342 
343 static int tls_push_data(struct sock *sk,
344 			 struct iov_iter *msg_iter,
345 			 size_t size, int flags,
346 			 unsigned char record_type)
347 {
348 	struct tls_context *tls_ctx = tls_get_ctx(sk);
349 	struct tls_offload_context_tx *ctx = tls_offload_ctx_tx(tls_ctx);
350 	int tls_push_record_flags = flags | MSG_SENDPAGE_NOTLAST;
351 	int more = flags & (MSG_SENDPAGE_NOTLAST | MSG_MORE);
352 	struct tls_record_info *record = ctx->open_record;
353 	struct page_frag *pfrag;
354 	size_t orig_size = size;
355 	u32 max_open_record_len;
356 	int copy, rc = 0;
357 	bool done = false;
358 	long timeo;
359 
360 	if (flags &
361 	    ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL | MSG_SENDPAGE_NOTLAST))
362 		return -ENOTSUPP;
363 
364 	if (sk->sk_err)
365 		return -sk->sk_err;
366 
367 	timeo = sock_sndtimeo(sk, flags & MSG_DONTWAIT);
368 	rc = tls_complete_pending_work(sk, tls_ctx, flags, &timeo);
369 	if (rc < 0)
370 		return rc;
371 
372 	pfrag = sk_page_frag(sk);
373 
374 	/* TLS_HEADER_SIZE is not counted as part of the TLS record, and
375 	 * we need to leave room for an authentication tag.
376 	 */
377 	max_open_record_len = TLS_MAX_PAYLOAD_SIZE +
378 			      tls_ctx->tx.prepend_size;
379 	do {
380 		rc = tls_do_allocation(sk, ctx, pfrag,
381 				       tls_ctx->tx.prepend_size);
382 		if (rc) {
383 			rc = sk_stream_wait_memory(sk, &timeo);
384 			if (!rc)
385 				continue;
386 
387 			record = ctx->open_record;
388 			if (!record)
389 				break;
390 handle_error:
391 			if (record_type != TLS_RECORD_TYPE_DATA) {
392 				/* avoid sending partial
393 				 * record with type !=
394 				 * application_data
395 				 */
396 				size = orig_size;
397 				destroy_record(record);
398 				ctx->open_record = NULL;
399 			} else if (record->len > tls_ctx->tx.prepend_size) {
400 				goto last_record;
401 			}
402 
403 			break;
404 		}
405 
406 		record = ctx->open_record;
407 		copy = min_t(size_t, size, (pfrag->size - pfrag->offset));
408 		copy = min_t(size_t, copy, (max_open_record_len - record->len));
409 
410 		if (copy_from_iter_nocache(page_address(pfrag->page) +
411 					       pfrag->offset,
412 					   copy, msg_iter) != copy) {
413 			rc = -EFAULT;
414 			goto handle_error;
415 		}
416 		tls_append_frag(record, pfrag, copy);
417 
418 		size -= copy;
419 		if (!size) {
420 last_record:
421 			tls_push_record_flags = flags;
422 			if (more) {
423 				tls_ctx->pending_open_record_frags =
424 						!!record->num_frags;
425 				break;
426 			}
427 
428 			done = true;
429 		}
430 
431 		if (done || record->len >= max_open_record_len ||
432 		    (record->num_frags >= MAX_SKB_FRAGS - 1)) {
433 			rc = tls_push_record(sk,
434 					     tls_ctx,
435 					     ctx,
436 					     record,
437 					     pfrag,
438 					     tls_push_record_flags,
439 					     record_type);
440 			if (rc < 0)
441 				break;
442 		}
443 	} while (!done);
444 
445 	if (orig_size - size > 0)
446 		rc = orig_size - size;
447 
448 	return rc;
449 }
450 
451 int tls_device_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
452 {
453 	unsigned char record_type = TLS_RECORD_TYPE_DATA;
454 	int rc;
455 
456 	lock_sock(sk);
457 
458 	if (unlikely(msg->msg_controllen)) {
459 		rc = tls_proccess_cmsg(sk, msg, &record_type);
460 		if (rc)
461 			goto out;
462 	}
463 
464 	rc = tls_push_data(sk, &msg->msg_iter, size,
465 			   msg->msg_flags, record_type);
466 
467 out:
468 	release_sock(sk);
469 	return rc;
470 }
471 
472 int tls_device_sendpage(struct sock *sk, struct page *page,
473 			int offset, size_t size, int flags)
474 {
475 	struct iov_iter	msg_iter;
476 	char *kaddr = kmap(page);
477 	struct kvec iov;
478 	int rc;
479 
480 	if (flags & MSG_SENDPAGE_NOTLAST)
481 		flags |= MSG_MORE;
482 
483 	lock_sock(sk);
484 
485 	if (flags & MSG_OOB) {
486 		rc = -ENOTSUPP;
487 		goto out;
488 	}
489 
490 	iov.iov_base = kaddr + offset;
491 	iov.iov_len = size;
492 	iov_iter_kvec(&msg_iter, WRITE | ITER_KVEC, &iov, 1, size);
493 	rc = tls_push_data(sk, &msg_iter, size,
494 			   flags, TLS_RECORD_TYPE_DATA);
495 	kunmap(page);
496 
497 out:
498 	release_sock(sk);
499 	return rc;
500 }
501 
502 struct tls_record_info *tls_get_record(struct tls_offload_context_tx *context,
503 				       u32 seq, u64 *p_record_sn)
504 {
505 	u64 record_sn = context->hint_record_sn;
506 	struct tls_record_info *info;
507 
508 	info = context->retransmit_hint;
509 	if (!info ||
510 	    before(seq, info->end_seq - info->len)) {
511 		/* if retransmit_hint is irrelevant start
512 		 * from the beggining of the list
513 		 */
514 		info = list_first_entry(&context->records_list,
515 					struct tls_record_info, list);
516 		record_sn = context->unacked_record_sn;
517 	}
518 
519 	list_for_each_entry_from(info, &context->records_list, list) {
520 		if (before(seq, info->end_seq)) {
521 			if (!context->retransmit_hint ||
522 			    after(info->end_seq,
523 				  context->retransmit_hint->end_seq)) {
524 				context->hint_record_sn = record_sn;
525 				context->retransmit_hint = info;
526 			}
527 			*p_record_sn = record_sn;
528 			return info;
529 		}
530 		record_sn++;
531 	}
532 
533 	return NULL;
534 }
535 EXPORT_SYMBOL(tls_get_record);
536 
537 static int tls_device_push_pending_record(struct sock *sk, int flags)
538 {
539 	struct iov_iter	msg_iter;
540 
541 	iov_iter_kvec(&msg_iter, WRITE | ITER_KVEC, NULL, 0, 0);
542 	return tls_push_data(sk, &msg_iter, 0, flags, TLS_RECORD_TYPE_DATA);
543 }
544 
545 void handle_device_resync(struct sock *sk, u32 seq, u64 rcd_sn)
546 {
547 	struct tls_context *tls_ctx = tls_get_ctx(sk);
548 	struct net_device *netdev = tls_ctx->netdev;
549 	struct tls_offload_context_rx *rx_ctx;
550 	u32 is_req_pending;
551 	s64 resync_req;
552 	u32 req_seq;
553 
554 	if (tls_ctx->rx_conf != TLS_HW)
555 		return;
556 
557 	rx_ctx = tls_offload_ctx_rx(tls_ctx);
558 	resync_req = atomic64_read(&rx_ctx->resync_req);
559 	req_seq = ntohl(resync_req >> 32) - ((u32)TLS_HEADER_SIZE - 1);
560 	is_req_pending = resync_req;
561 
562 	if (unlikely(is_req_pending) && req_seq == seq &&
563 	    atomic64_try_cmpxchg(&rx_ctx->resync_req, &resync_req, 0))
564 		netdev->tlsdev_ops->tls_dev_resync_rx(netdev, sk,
565 						      seq + TLS_HEADER_SIZE - 1,
566 						      rcd_sn);
567 }
568 
569 static int tls_device_reencrypt(struct sock *sk, struct sk_buff *skb)
570 {
571 	struct strp_msg *rxm = strp_msg(skb);
572 	int err = 0, offset = rxm->offset, copy, nsg;
573 	struct sk_buff *skb_iter, *unused;
574 	struct scatterlist sg[1];
575 	char *orig_buf, *buf;
576 
577 	orig_buf = kmalloc(rxm->full_len + TLS_HEADER_SIZE +
578 			   TLS_CIPHER_AES_GCM_128_IV_SIZE, sk->sk_allocation);
579 	if (!orig_buf)
580 		return -ENOMEM;
581 	buf = orig_buf;
582 
583 	nsg = skb_cow_data(skb, 0, &unused);
584 	if (unlikely(nsg < 0)) {
585 		err = nsg;
586 		goto free_buf;
587 	}
588 
589 	sg_init_table(sg, 1);
590 	sg_set_buf(&sg[0], buf,
591 		   rxm->full_len + TLS_HEADER_SIZE +
592 		   TLS_CIPHER_AES_GCM_128_IV_SIZE);
593 	skb_copy_bits(skb, offset, buf,
594 		      TLS_HEADER_SIZE + TLS_CIPHER_AES_GCM_128_IV_SIZE);
595 
596 	/* We are interested only in the decrypted data not the auth */
597 	err = decrypt_skb(sk, skb, sg);
598 	if (err != -EBADMSG)
599 		goto free_buf;
600 	else
601 		err = 0;
602 
603 	copy = min_t(int, skb_pagelen(skb) - offset,
604 		     rxm->full_len - TLS_CIPHER_AES_GCM_128_TAG_SIZE);
605 
606 	if (skb->decrypted)
607 		skb_store_bits(skb, offset, buf, copy);
608 
609 	offset += copy;
610 	buf += copy;
611 
612 	skb_walk_frags(skb, skb_iter) {
613 		copy = min_t(int, skb_iter->len,
614 			     rxm->full_len - offset + rxm->offset -
615 			     TLS_CIPHER_AES_GCM_128_TAG_SIZE);
616 
617 		if (skb_iter->decrypted)
618 			skb_store_bits(skb_iter, offset, buf, copy);
619 
620 		offset += copy;
621 		buf += copy;
622 	}
623 
624 free_buf:
625 	kfree(orig_buf);
626 	return err;
627 }
628 
629 int tls_device_decrypted(struct sock *sk, struct sk_buff *skb)
630 {
631 	struct tls_context *tls_ctx = tls_get_ctx(sk);
632 	struct tls_offload_context_rx *ctx = tls_offload_ctx_rx(tls_ctx);
633 	int is_decrypted = skb->decrypted;
634 	int is_encrypted = !is_decrypted;
635 	struct sk_buff *skb_iter;
636 
637 	/* Skip if it is already decrypted */
638 	if (ctx->sw.decrypted)
639 		return 0;
640 
641 	/* Check if all the data is decrypted already */
642 	skb_walk_frags(skb, skb_iter) {
643 		is_decrypted &= skb_iter->decrypted;
644 		is_encrypted &= !skb_iter->decrypted;
645 	}
646 
647 	ctx->sw.decrypted |= is_decrypted;
648 
649 	/* Return immedeatly if the record is either entirely plaintext or
650 	 * entirely ciphertext. Otherwise handle reencrypt partially decrypted
651 	 * record.
652 	 */
653 	return (is_encrypted || is_decrypted) ? 0 :
654 		tls_device_reencrypt(sk, skb);
655 }
656 
657 int tls_set_device_offload(struct sock *sk, struct tls_context *ctx)
658 {
659 	u16 nonce_size, tag_size, iv_size, rec_seq_size;
660 	struct tls_record_info *start_marker_record;
661 	struct tls_offload_context_tx *offload_ctx;
662 	struct tls_crypto_info *crypto_info;
663 	struct net_device *netdev;
664 	char *iv, *rec_seq;
665 	struct sk_buff *skb;
666 	int rc = -EINVAL;
667 	__be64 rcd_sn;
668 
669 	if (!ctx)
670 		goto out;
671 
672 	if (ctx->priv_ctx_tx) {
673 		rc = -EEXIST;
674 		goto out;
675 	}
676 
677 	start_marker_record = kmalloc(sizeof(*start_marker_record), GFP_KERNEL);
678 	if (!start_marker_record) {
679 		rc = -ENOMEM;
680 		goto out;
681 	}
682 
683 	offload_ctx = kzalloc(TLS_OFFLOAD_CONTEXT_SIZE_TX, GFP_KERNEL);
684 	if (!offload_ctx) {
685 		rc = -ENOMEM;
686 		goto free_marker_record;
687 	}
688 
689 	crypto_info = &ctx->crypto_send.info;
690 	switch (crypto_info->cipher_type) {
691 	case TLS_CIPHER_AES_GCM_128:
692 		nonce_size = TLS_CIPHER_AES_GCM_128_IV_SIZE;
693 		tag_size = TLS_CIPHER_AES_GCM_128_TAG_SIZE;
694 		iv_size = TLS_CIPHER_AES_GCM_128_IV_SIZE;
695 		iv = ((struct tls12_crypto_info_aes_gcm_128 *)crypto_info)->iv;
696 		rec_seq_size = TLS_CIPHER_AES_GCM_128_REC_SEQ_SIZE;
697 		rec_seq =
698 		 ((struct tls12_crypto_info_aes_gcm_128 *)crypto_info)->rec_seq;
699 		break;
700 	default:
701 		rc = -EINVAL;
702 		goto free_offload_ctx;
703 	}
704 
705 	ctx->tx.prepend_size = TLS_HEADER_SIZE + nonce_size;
706 	ctx->tx.tag_size = tag_size;
707 	ctx->tx.overhead_size = ctx->tx.prepend_size + ctx->tx.tag_size;
708 	ctx->tx.iv_size = iv_size;
709 	ctx->tx.iv = kmalloc(iv_size + TLS_CIPHER_AES_GCM_128_SALT_SIZE,
710 			     GFP_KERNEL);
711 	if (!ctx->tx.iv) {
712 		rc = -ENOMEM;
713 		goto free_offload_ctx;
714 	}
715 
716 	memcpy(ctx->tx.iv + TLS_CIPHER_AES_GCM_128_SALT_SIZE, iv, iv_size);
717 
718 	ctx->tx.rec_seq_size = rec_seq_size;
719 	ctx->tx.rec_seq = kmemdup(rec_seq, rec_seq_size, GFP_KERNEL);
720 	if (!ctx->tx.rec_seq) {
721 		rc = -ENOMEM;
722 		goto free_iv;
723 	}
724 
725 	rc = tls_sw_fallback_init(sk, offload_ctx, crypto_info);
726 	if (rc)
727 		goto free_rec_seq;
728 
729 	/* start at rec_seq - 1 to account for the start marker record */
730 	memcpy(&rcd_sn, ctx->tx.rec_seq, sizeof(rcd_sn));
731 	offload_ctx->unacked_record_sn = be64_to_cpu(rcd_sn) - 1;
732 
733 	start_marker_record->end_seq = tcp_sk(sk)->write_seq;
734 	start_marker_record->len = 0;
735 	start_marker_record->num_frags = 0;
736 
737 	INIT_LIST_HEAD(&offload_ctx->records_list);
738 	list_add_tail(&start_marker_record->list, &offload_ctx->records_list);
739 	spin_lock_init(&offload_ctx->lock);
740 	sg_init_table(offload_ctx->sg_tx_data,
741 		      ARRAY_SIZE(offload_ctx->sg_tx_data));
742 
743 	clean_acked_data_enable(inet_csk(sk), &tls_icsk_clean_acked);
744 	ctx->push_pending_record = tls_device_push_pending_record;
745 
746 	/* TLS offload is greatly simplified if we don't send
747 	 * SKBs where only part of the payload needs to be encrypted.
748 	 * So mark the last skb in the write queue as end of record.
749 	 */
750 	skb = tcp_write_queue_tail(sk);
751 	if (skb)
752 		TCP_SKB_CB(skb)->eor = 1;
753 
754 	/* We support starting offload on multiple sockets
755 	 * concurrently, so we only need a read lock here.
756 	 * This lock must precede get_netdev_for_sock to prevent races between
757 	 * NETDEV_DOWN and setsockopt.
758 	 */
759 	down_read(&device_offload_lock);
760 	netdev = get_netdev_for_sock(sk);
761 	if (!netdev) {
762 		pr_err_ratelimited("%s: netdev not found\n", __func__);
763 		rc = -EINVAL;
764 		goto release_lock;
765 	}
766 
767 	if (!(netdev->features & NETIF_F_HW_TLS_TX)) {
768 		rc = -ENOTSUPP;
769 		goto release_netdev;
770 	}
771 
772 	/* Avoid offloading if the device is down
773 	 * We don't want to offload new flows after
774 	 * the NETDEV_DOWN event
775 	 */
776 	if (!(netdev->flags & IFF_UP)) {
777 		rc = -EINVAL;
778 		goto release_netdev;
779 	}
780 
781 	ctx->priv_ctx_tx = offload_ctx;
782 	rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_TX,
783 					     &ctx->crypto_send.info,
784 					     tcp_sk(sk)->write_seq);
785 	if (rc)
786 		goto release_netdev;
787 
788 	tls_device_attach(ctx, sk, netdev);
789 
790 	/* following this assignment tls_is_sk_tx_device_offloaded
791 	 * will return true and the context might be accessed
792 	 * by the netdev's xmit function.
793 	 */
794 	smp_store_release(&sk->sk_validate_xmit_skb, tls_validate_xmit_skb);
795 	dev_put(netdev);
796 	up_read(&device_offload_lock);
797 	goto out;
798 
799 release_netdev:
800 	dev_put(netdev);
801 release_lock:
802 	up_read(&device_offload_lock);
803 	clean_acked_data_disable(inet_csk(sk));
804 	crypto_free_aead(offload_ctx->aead_send);
805 free_rec_seq:
806 	kfree(ctx->tx.rec_seq);
807 free_iv:
808 	kfree(ctx->tx.iv);
809 free_offload_ctx:
810 	kfree(offload_ctx);
811 	ctx->priv_ctx_tx = NULL;
812 free_marker_record:
813 	kfree(start_marker_record);
814 out:
815 	return rc;
816 }
817 
818 int tls_set_device_offload_rx(struct sock *sk, struct tls_context *ctx)
819 {
820 	struct tls_offload_context_rx *context;
821 	struct net_device *netdev;
822 	int rc = 0;
823 
824 	/* We support starting offload on multiple sockets
825 	 * concurrently, so we only need a read lock here.
826 	 * This lock must precede get_netdev_for_sock to prevent races between
827 	 * NETDEV_DOWN and setsockopt.
828 	 */
829 	down_read(&device_offload_lock);
830 	netdev = get_netdev_for_sock(sk);
831 	if (!netdev) {
832 		pr_err_ratelimited("%s: netdev not found\n", __func__);
833 		rc = -EINVAL;
834 		goto release_lock;
835 	}
836 
837 	if (!(netdev->features & NETIF_F_HW_TLS_RX)) {
838 		pr_err_ratelimited("%s: netdev %s with no TLS offload\n",
839 				   __func__, netdev->name);
840 		rc = -ENOTSUPP;
841 		goto release_netdev;
842 	}
843 
844 	/* Avoid offloading if the device is down
845 	 * We don't want to offload new flows after
846 	 * the NETDEV_DOWN event
847 	 */
848 	if (!(netdev->flags & IFF_UP)) {
849 		rc = -EINVAL;
850 		goto release_netdev;
851 	}
852 
853 	context = kzalloc(TLS_OFFLOAD_CONTEXT_SIZE_RX, GFP_KERNEL);
854 	if (!context) {
855 		rc = -ENOMEM;
856 		goto release_netdev;
857 	}
858 
859 	ctx->priv_ctx_rx = context;
860 	rc = tls_set_sw_offload(sk, ctx, 0);
861 	if (rc)
862 		goto release_ctx;
863 
864 	rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_RX,
865 					     &ctx->crypto_recv.info,
866 					     tcp_sk(sk)->copied_seq);
867 	if (rc) {
868 		pr_err_ratelimited("%s: The netdev has refused to offload this socket\n",
869 				   __func__);
870 		goto free_sw_resources;
871 	}
872 
873 	tls_device_attach(ctx, sk, netdev);
874 	goto release_netdev;
875 
876 free_sw_resources:
877 	tls_sw_free_resources_rx(sk);
878 release_ctx:
879 	ctx->priv_ctx_rx = NULL;
880 release_netdev:
881 	dev_put(netdev);
882 release_lock:
883 	up_read(&device_offload_lock);
884 	return rc;
885 }
886 
887 void tls_device_offload_cleanup_rx(struct sock *sk)
888 {
889 	struct tls_context *tls_ctx = tls_get_ctx(sk);
890 	struct net_device *netdev;
891 
892 	down_read(&device_offload_lock);
893 	netdev = tls_ctx->netdev;
894 	if (!netdev)
895 		goto out;
896 
897 	if (!(netdev->features & NETIF_F_HW_TLS_RX)) {
898 		pr_err_ratelimited("%s: device is missing NETIF_F_HW_TLS_RX cap\n",
899 				   __func__);
900 		goto out;
901 	}
902 
903 	netdev->tlsdev_ops->tls_dev_del(netdev, tls_ctx,
904 					TLS_OFFLOAD_CTX_DIR_RX);
905 
906 	if (tls_ctx->tx_conf != TLS_HW) {
907 		dev_put(netdev);
908 		tls_ctx->netdev = NULL;
909 	}
910 out:
911 	up_read(&device_offload_lock);
912 	kfree(tls_ctx->rx.rec_seq);
913 	kfree(tls_ctx->rx.iv);
914 	tls_sw_release_resources_rx(sk);
915 }
916 
917 static int tls_device_down(struct net_device *netdev)
918 {
919 	struct tls_context *ctx, *tmp;
920 	unsigned long flags;
921 	LIST_HEAD(list);
922 
923 	/* Request a write lock to block new offload attempts */
924 	down_write(&device_offload_lock);
925 
926 	spin_lock_irqsave(&tls_device_lock, flags);
927 	list_for_each_entry_safe(ctx, tmp, &tls_device_list, list) {
928 		if (ctx->netdev != netdev ||
929 		    !refcount_inc_not_zero(&ctx->refcount))
930 			continue;
931 
932 		list_move(&ctx->list, &list);
933 	}
934 	spin_unlock_irqrestore(&tls_device_lock, flags);
935 
936 	list_for_each_entry_safe(ctx, tmp, &list, list)	{
937 		if (ctx->tx_conf == TLS_HW)
938 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
939 							TLS_OFFLOAD_CTX_DIR_TX);
940 		if (ctx->rx_conf == TLS_HW)
941 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
942 							TLS_OFFLOAD_CTX_DIR_RX);
943 		ctx->netdev = NULL;
944 		dev_put(netdev);
945 		list_del_init(&ctx->list);
946 
947 		if (refcount_dec_and_test(&ctx->refcount))
948 			tls_device_free_ctx(ctx);
949 	}
950 
951 	up_write(&device_offload_lock);
952 
953 	flush_work(&tls_device_gc_work);
954 
955 	return NOTIFY_DONE;
956 }
957 
958 static int tls_dev_event(struct notifier_block *this, unsigned long event,
959 			 void *ptr)
960 {
961 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
962 
963 	if (!(dev->features & (NETIF_F_HW_TLS_RX | NETIF_F_HW_TLS_TX)))
964 		return NOTIFY_DONE;
965 
966 	switch (event) {
967 	case NETDEV_REGISTER:
968 	case NETDEV_FEAT_CHANGE:
969 		if ((dev->features & NETIF_F_HW_TLS_RX) &&
970 		    !dev->tlsdev_ops->tls_dev_resync_rx)
971 			return NOTIFY_BAD;
972 
973 		if  (dev->tlsdev_ops &&
974 		     dev->tlsdev_ops->tls_dev_add &&
975 		     dev->tlsdev_ops->tls_dev_del)
976 			return NOTIFY_DONE;
977 		else
978 			return NOTIFY_BAD;
979 	case NETDEV_DOWN:
980 		return tls_device_down(dev);
981 	}
982 	return NOTIFY_DONE;
983 }
984 
985 static struct notifier_block tls_dev_notifier = {
986 	.notifier_call	= tls_dev_event,
987 };
988 
989 void __init tls_device_init(void)
990 {
991 	register_netdevice_notifier(&tls_dev_notifier);
992 }
993 
994 void __exit tls_device_cleanup(void)
995 {
996 	unregister_netdevice_notifier(&tls_dev_notifier);
997 	flush_work(&tls_device_gc_work);
998 }
999