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 kfree(ctx->tx.rec_seq); 58 kfree(ctx->tx.iv); 59 } 60 61 if (ctx->rx_conf == TLS_HW) 62 kfree(tls_offload_ctx_rx(ctx)); 63 64 tls_ctx_free(ctx); 65 } 66 67 static void tls_device_gc_task(struct work_struct *work) 68 { 69 struct tls_context *ctx, *tmp; 70 unsigned long flags; 71 LIST_HEAD(gc_list); 72 73 spin_lock_irqsave(&tls_device_lock, flags); 74 list_splice_init(&tls_device_gc_list, &gc_list); 75 spin_unlock_irqrestore(&tls_device_lock, flags); 76 77 list_for_each_entry_safe(ctx, tmp, &gc_list, list) { 78 struct net_device *netdev = ctx->netdev; 79 80 if (netdev && ctx->tx_conf == TLS_HW) { 81 netdev->tlsdev_ops->tls_dev_del(netdev, ctx, 82 TLS_OFFLOAD_CTX_DIR_TX); 83 dev_put(netdev); 84 ctx->netdev = NULL; 85 } 86 87 list_del(&ctx->list); 88 tls_device_free_ctx(ctx); 89 } 90 } 91 92 static void tls_device_queue_ctx_destruction(struct tls_context *ctx) 93 { 94 unsigned long flags; 95 96 spin_lock_irqsave(&tls_device_lock, flags); 97 list_move_tail(&ctx->list, &tls_device_gc_list); 98 99 /* schedule_work inside the spinlock 100 * to make sure tls_device_down waits for that work. 101 */ 102 schedule_work(&tls_device_gc_work); 103 104 spin_unlock_irqrestore(&tls_device_lock, flags); 105 } 106 107 /* We assume that the socket is already connected */ 108 static struct net_device *get_netdev_for_sock(struct sock *sk) 109 { 110 struct dst_entry *dst = sk_dst_get(sk); 111 struct net_device *netdev = NULL; 112 113 if (likely(dst)) { 114 netdev = dst->dev; 115 dev_hold(netdev); 116 } 117 118 dst_release(dst); 119 120 return netdev; 121 } 122 123 static void destroy_record(struct tls_record_info *record) 124 { 125 int nr_frags = record->num_frags; 126 skb_frag_t *frag; 127 128 while (nr_frags-- > 0) { 129 frag = &record->frags[nr_frags]; 130 __skb_frag_unref(frag); 131 } 132 kfree(record); 133 } 134 135 static void delete_all_records(struct tls_offload_context_tx *offload_ctx) 136 { 137 struct tls_record_info *info, *temp; 138 139 list_for_each_entry_safe(info, temp, &offload_ctx->records_list, list) { 140 list_del(&info->list); 141 destroy_record(info); 142 } 143 144 offload_ctx->retransmit_hint = NULL; 145 } 146 147 static void tls_icsk_clean_acked(struct sock *sk, u32 acked_seq) 148 { 149 struct tls_context *tls_ctx = tls_get_ctx(sk); 150 struct tls_record_info *info, *temp; 151 struct tls_offload_context_tx *ctx; 152 u64 deleted_records = 0; 153 unsigned long flags; 154 155 if (!tls_ctx) 156 return; 157 158 ctx = tls_offload_ctx_tx(tls_ctx); 159 160 spin_lock_irqsave(&ctx->lock, flags); 161 info = ctx->retransmit_hint; 162 if (info && !before(acked_seq, info->end_seq)) { 163 ctx->retransmit_hint = NULL; 164 list_del(&info->list); 165 destroy_record(info); 166 deleted_records++; 167 } 168 169 list_for_each_entry_safe(info, temp, &ctx->records_list, list) { 170 if (before(acked_seq, info->end_seq)) 171 break; 172 list_del(&info->list); 173 174 destroy_record(info); 175 deleted_records++; 176 } 177 178 ctx->unacked_record_sn += deleted_records; 179 spin_unlock_irqrestore(&ctx->lock, flags); 180 } 181 182 /* At this point, there should be no references on this 183 * socket and no in-flight SKBs associated with this 184 * socket, so it is safe to free all the resources. 185 */ 186 static void tls_device_sk_destruct(struct sock *sk) 187 { 188 struct tls_context *tls_ctx = tls_get_ctx(sk); 189 struct tls_offload_context_tx *ctx = tls_offload_ctx_tx(tls_ctx); 190 191 tls_ctx->sk_destruct(sk); 192 193 if (tls_ctx->tx_conf == TLS_HW) { 194 if (ctx->open_record) 195 destroy_record(ctx->open_record); 196 delete_all_records(ctx); 197 crypto_free_aead(ctx->aead_send); 198 clean_acked_data_disable(inet_csk(sk)); 199 } 200 201 if (refcount_dec_and_test(&tls_ctx->refcount)) 202 tls_device_queue_ctx_destruction(tls_ctx); 203 } 204 205 void tls_device_free_resources_tx(struct sock *sk) 206 { 207 struct tls_context *tls_ctx = tls_get_ctx(sk); 208 209 tls_free_partial_record(sk, tls_ctx); 210 } 211 212 static void tls_device_resync_tx(struct sock *sk, struct tls_context *tls_ctx, 213 u32 seq) 214 { 215 struct net_device *netdev; 216 struct sk_buff *skb; 217 u8 *rcd_sn; 218 219 skb = tcp_write_queue_tail(sk); 220 if (skb) 221 TCP_SKB_CB(skb)->eor = 1; 222 223 rcd_sn = tls_ctx->tx.rec_seq; 224 225 down_read(&device_offload_lock); 226 netdev = tls_ctx->netdev; 227 if (netdev) 228 netdev->tlsdev_ops->tls_dev_resync(netdev, sk, seq, rcd_sn, 229 TLS_OFFLOAD_CTX_DIR_TX); 230 up_read(&device_offload_lock); 231 232 clear_bit_unlock(TLS_TX_SYNC_SCHED, &tls_ctx->flags); 233 } 234 235 static void tls_append_frag(struct tls_record_info *record, 236 struct page_frag *pfrag, 237 int size) 238 { 239 skb_frag_t *frag; 240 241 frag = &record->frags[record->num_frags - 1]; 242 if (frag->page.p == pfrag->page && 243 frag->page_offset + frag->size == pfrag->offset) { 244 frag->size += size; 245 } else { 246 ++frag; 247 frag->page.p = pfrag->page; 248 frag->page_offset = pfrag->offset; 249 frag->size = size; 250 ++record->num_frags; 251 get_page(pfrag->page); 252 } 253 254 pfrag->offset += size; 255 record->len += size; 256 } 257 258 static int tls_push_record(struct sock *sk, 259 struct tls_context *ctx, 260 struct tls_offload_context_tx *offload_ctx, 261 struct tls_record_info *record, 262 struct page_frag *pfrag, 263 int flags, 264 unsigned char record_type) 265 { 266 struct tls_prot_info *prot = &ctx->prot_info; 267 struct tcp_sock *tp = tcp_sk(sk); 268 struct page_frag dummy_tag_frag; 269 skb_frag_t *frag; 270 int i; 271 272 /* fill prepend */ 273 frag = &record->frags[0]; 274 tls_fill_prepend(ctx, 275 skb_frag_address(frag), 276 record->len - prot->prepend_size, 277 record_type, 278 prot->version); 279 280 /* HW doesn't care about the data in the tag, because it fills it. */ 281 dummy_tag_frag.page = skb_frag_page(frag); 282 dummy_tag_frag.offset = 0; 283 284 tls_append_frag(record, &dummy_tag_frag, prot->tag_size); 285 record->end_seq = tp->write_seq + record->len; 286 spin_lock_irq(&offload_ctx->lock); 287 list_add_tail(&record->list, &offload_ctx->records_list); 288 spin_unlock_irq(&offload_ctx->lock); 289 offload_ctx->open_record = NULL; 290 291 if (test_bit(TLS_TX_SYNC_SCHED, &ctx->flags)) 292 tls_device_resync_tx(sk, ctx, tp->write_seq); 293 294 tls_advance_record_sn(sk, prot, &ctx->tx); 295 296 for (i = 0; i < record->num_frags; i++) { 297 frag = &record->frags[i]; 298 sg_unmark_end(&offload_ctx->sg_tx_data[i]); 299 sg_set_page(&offload_ctx->sg_tx_data[i], skb_frag_page(frag), 300 frag->size, frag->page_offset); 301 sk_mem_charge(sk, frag->size); 302 get_page(skb_frag_page(frag)); 303 } 304 sg_mark_end(&offload_ctx->sg_tx_data[record->num_frags - 1]); 305 306 /* all ready, send */ 307 return tls_push_sg(sk, ctx, offload_ctx->sg_tx_data, 0, flags); 308 } 309 310 static int tls_create_new_record(struct tls_offload_context_tx *offload_ctx, 311 struct page_frag *pfrag, 312 size_t prepend_size) 313 { 314 struct tls_record_info *record; 315 skb_frag_t *frag; 316 317 record = kmalloc(sizeof(*record), GFP_KERNEL); 318 if (!record) 319 return -ENOMEM; 320 321 frag = &record->frags[0]; 322 __skb_frag_set_page(frag, pfrag->page); 323 frag->page_offset = pfrag->offset; 324 skb_frag_size_set(frag, prepend_size); 325 326 get_page(pfrag->page); 327 pfrag->offset += prepend_size; 328 329 record->num_frags = 1; 330 record->len = prepend_size; 331 offload_ctx->open_record = record; 332 return 0; 333 } 334 335 static int tls_do_allocation(struct sock *sk, 336 struct tls_offload_context_tx *offload_ctx, 337 struct page_frag *pfrag, 338 size_t prepend_size) 339 { 340 int ret; 341 342 if (!offload_ctx->open_record) { 343 if (unlikely(!skb_page_frag_refill(prepend_size, pfrag, 344 sk->sk_allocation))) { 345 sk->sk_prot->enter_memory_pressure(sk); 346 sk_stream_moderate_sndbuf(sk); 347 return -ENOMEM; 348 } 349 350 ret = tls_create_new_record(offload_ctx, pfrag, prepend_size); 351 if (ret) 352 return ret; 353 354 if (pfrag->size > pfrag->offset) 355 return 0; 356 } 357 358 if (!sk_page_frag_refill(sk, pfrag)) 359 return -ENOMEM; 360 361 return 0; 362 } 363 364 static int tls_push_data(struct sock *sk, 365 struct iov_iter *msg_iter, 366 size_t size, int flags, 367 unsigned char record_type) 368 { 369 struct tls_context *tls_ctx = tls_get_ctx(sk); 370 struct tls_prot_info *prot = &tls_ctx->prot_info; 371 struct tls_offload_context_tx *ctx = tls_offload_ctx_tx(tls_ctx); 372 int tls_push_record_flags = flags | MSG_SENDPAGE_NOTLAST; 373 int more = flags & (MSG_SENDPAGE_NOTLAST | MSG_MORE); 374 struct tls_record_info *record = ctx->open_record; 375 struct page_frag *pfrag; 376 size_t orig_size = size; 377 u32 max_open_record_len; 378 int copy, rc = 0; 379 bool done = false; 380 long timeo; 381 382 if (flags & 383 ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL | MSG_SENDPAGE_NOTLAST)) 384 return -ENOTSUPP; 385 386 if (sk->sk_err) 387 return -sk->sk_err; 388 389 timeo = sock_sndtimeo(sk, flags & MSG_DONTWAIT); 390 if (tls_is_partially_sent_record(tls_ctx)) { 391 rc = tls_push_partial_record(sk, tls_ctx, flags); 392 if (rc < 0) 393 return rc; 394 } 395 396 pfrag = sk_page_frag(sk); 397 398 /* TLS_HEADER_SIZE is not counted as part of the TLS record, and 399 * we need to leave room for an authentication tag. 400 */ 401 max_open_record_len = TLS_MAX_PAYLOAD_SIZE + 402 prot->prepend_size; 403 do { 404 rc = tls_do_allocation(sk, ctx, pfrag, 405 prot->prepend_size); 406 if (rc) { 407 rc = sk_stream_wait_memory(sk, &timeo); 408 if (!rc) 409 continue; 410 411 record = ctx->open_record; 412 if (!record) 413 break; 414 handle_error: 415 if (record_type != TLS_RECORD_TYPE_DATA) { 416 /* avoid sending partial 417 * record with type != 418 * application_data 419 */ 420 size = orig_size; 421 destroy_record(record); 422 ctx->open_record = NULL; 423 } else if (record->len > prot->prepend_size) { 424 goto last_record; 425 } 426 427 break; 428 } 429 430 record = ctx->open_record; 431 copy = min_t(size_t, size, (pfrag->size - pfrag->offset)); 432 copy = min_t(size_t, copy, (max_open_record_len - record->len)); 433 434 if (copy_from_iter_nocache(page_address(pfrag->page) + 435 pfrag->offset, 436 copy, msg_iter) != copy) { 437 rc = -EFAULT; 438 goto handle_error; 439 } 440 tls_append_frag(record, pfrag, copy); 441 442 size -= copy; 443 if (!size) { 444 last_record: 445 tls_push_record_flags = flags; 446 if (more) { 447 tls_ctx->pending_open_record_frags = 448 !!record->num_frags; 449 break; 450 } 451 452 done = true; 453 } 454 455 if (done || record->len >= max_open_record_len || 456 (record->num_frags >= MAX_SKB_FRAGS - 1)) { 457 rc = tls_push_record(sk, 458 tls_ctx, 459 ctx, 460 record, 461 pfrag, 462 tls_push_record_flags, 463 record_type); 464 if (rc < 0) 465 break; 466 } 467 } while (!done); 468 469 if (orig_size - size > 0) 470 rc = orig_size - size; 471 472 return rc; 473 } 474 475 int tls_device_sendmsg(struct sock *sk, struct msghdr *msg, size_t size) 476 { 477 unsigned char record_type = TLS_RECORD_TYPE_DATA; 478 int rc; 479 480 lock_sock(sk); 481 482 if (unlikely(msg->msg_controllen)) { 483 rc = tls_proccess_cmsg(sk, msg, &record_type); 484 if (rc) 485 goto out; 486 } 487 488 rc = tls_push_data(sk, &msg->msg_iter, size, 489 msg->msg_flags, record_type); 490 491 out: 492 release_sock(sk); 493 return rc; 494 } 495 496 int tls_device_sendpage(struct sock *sk, struct page *page, 497 int offset, size_t size, int flags) 498 { 499 struct iov_iter msg_iter; 500 char *kaddr = kmap(page); 501 struct kvec iov; 502 int rc; 503 504 if (flags & MSG_SENDPAGE_NOTLAST) 505 flags |= MSG_MORE; 506 507 lock_sock(sk); 508 509 if (flags & MSG_OOB) { 510 rc = -ENOTSUPP; 511 goto out; 512 } 513 514 iov.iov_base = kaddr + offset; 515 iov.iov_len = size; 516 iov_iter_kvec(&msg_iter, WRITE, &iov, 1, size); 517 rc = tls_push_data(sk, &msg_iter, size, 518 flags, TLS_RECORD_TYPE_DATA); 519 kunmap(page); 520 521 out: 522 release_sock(sk); 523 return rc; 524 } 525 526 struct tls_record_info *tls_get_record(struct tls_offload_context_tx *context, 527 u32 seq, u64 *p_record_sn) 528 { 529 u64 record_sn = context->hint_record_sn; 530 struct tls_record_info *info; 531 532 info = context->retransmit_hint; 533 if (!info || 534 before(seq, info->end_seq - info->len)) { 535 /* if retransmit_hint is irrelevant start 536 * from the beggining of the list 537 */ 538 info = list_first_entry(&context->records_list, 539 struct tls_record_info, list); 540 record_sn = context->unacked_record_sn; 541 } 542 543 list_for_each_entry_from(info, &context->records_list, list) { 544 if (before(seq, info->end_seq)) { 545 if (!context->retransmit_hint || 546 after(info->end_seq, 547 context->retransmit_hint->end_seq)) { 548 context->hint_record_sn = record_sn; 549 context->retransmit_hint = info; 550 } 551 *p_record_sn = record_sn; 552 return info; 553 } 554 record_sn++; 555 } 556 557 return NULL; 558 } 559 EXPORT_SYMBOL(tls_get_record); 560 561 static int tls_device_push_pending_record(struct sock *sk, int flags) 562 { 563 struct iov_iter msg_iter; 564 565 iov_iter_kvec(&msg_iter, WRITE, NULL, 0, 0); 566 return tls_push_data(sk, &msg_iter, 0, flags, TLS_RECORD_TYPE_DATA); 567 } 568 569 void tls_device_write_space(struct sock *sk, struct tls_context *ctx) 570 { 571 if (!sk->sk_write_pending && tls_is_partially_sent_record(ctx)) { 572 gfp_t sk_allocation = sk->sk_allocation; 573 574 sk->sk_allocation = GFP_ATOMIC; 575 tls_push_partial_record(sk, ctx, MSG_DONTWAIT | MSG_NOSIGNAL); 576 sk->sk_allocation = sk_allocation; 577 } 578 } 579 580 static void tls_device_resync_rx(struct tls_context *tls_ctx, 581 struct sock *sk, u32 seq, u8 *rcd_sn) 582 { 583 struct net_device *netdev; 584 585 if (WARN_ON(test_and_set_bit(TLS_RX_SYNC_RUNNING, &tls_ctx->flags))) 586 return; 587 netdev = READ_ONCE(tls_ctx->netdev); 588 if (netdev) 589 netdev->tlsdev_ops->tls_dev_resync(netdev, sk, seq, rcd_sn, 590 TLS_OFFLOAD_CTX_DIR_RX); 591 clear_bit_unlock(TLS_RX_SYNC_RUNNING, &tls_ctx->flags); 592 } 593 594 void tls_device_rx_resync_new_rec(struct sock *sk, u32 rcd_len, u32 seq) 595 { 596 struct tls_context *tls_ctx = tls_get_ctx(sk); 597 struct tls_offload_context_rx *rx_ctx; 598 u8 rcd_sn[TLS_MAX_REC_SEQ_SIZE]; 599 struct tls_prot_info *prot; 600 u32 is_req_pending; 601 s64 resync_req; 602 u32 req_seq; 603 604 if (tls_ctx->rx_conf != TLS_HW) 605 return; 606 607 prot = &tls_ctx->prot_info; 608 rx_ctx = tls_offload_ctx_rx(tls_ctx); 609 memcpy(rcd_sn, tls_ctx->rx.rec_seq, prot->rec_seq_size); 610 611 switch (rx_ctx->resync_type) { 612 case TLS_OFFLOAD_SYNC_TYPE_DRIVER_REQ: 613 resync_req = atomic64_read(&rx_ctx->resync_req); 614 req_seq = resync_req >> 32; 615 seq += TLS_HEADER_SIZE - 1; 616 is_req_pending = resync_req; 617 618 if (likely(!is_req_pending) || req_seq != seq || 619 !atomic64_try_cmpxchg(&rx_ctx->resync_req, &resync_req, 0)) 620 return; 621 break; 622 case TLS_OFFLOAD_SYNC_TYPE_CORE_NEXT_HINT: 623 if (likely(!rx_ctx->resync_nh_do_now)) 624 return; 625 626 /* head of next rec is already in, note that the sock_inq will 627 * include the currently parsed message when called from parser 628 */ 629 if (tcp_inq(sk) > rcd_len) 630 return; 631 632 rx_ctx->resync_nh_do_now = 0; 633 seq += rcd_len; 634 tls_bigint_increment(rcd_sn, prot->rec_seq_size); 635 break; 636 } 637 638 tls_device_resync_rx(tls_ctx, sk, seq, rcd_sn); 639 } 640 641 static void tls_device_core_ctrl_rx_resync(struct tls_context *tls_ctx, 642 struct tls_offload_context_rx *ctx, 643 struct sock *sk, struct sk_buff *skb) 644 { 645 struct strp_msg *rxm; 646 647 /* device will request resyncs by itself based on stream scan */ 648 if (ctx->resync_type != TLS_OFFLOAD_SYNC_TYPE_CORE_NEXT_HINT) 649 return; 650 /* already scheduled */ 651 if (ctx->resync_nh_do_now) 652 return; 653 /* seen decrypted fragments since last fully-failed record */ 654 if (ctx->resync_nh_reset) { 655 ctx->resync_nh_reset = 0; 656 ctx->resync_nh.decrypted_failed = 1; 657 ctx->resync_nh.decrypted_tgt = TLS_DEVICE_RESYNC_NH_START_IVAL; 658 return; 659 } 660 661 if (++ctx->resync_nh.decrypted_failed <= ctx->resync_nh.decrypted_tgt) 662 return; 663 664 /* doing resync, bump the next target in case it fails */ 665 if (ctx->resync_nh.decrypted_tgt < TLS_DEVICE_RESYNC_NH_MAX_IVAL) 666 ctx->resync_nh.decrypted_tgt *= 2; 667 else 668 ctx->resync_nh.decrypted_tgt += TLS_DEVICE_RESYNC_NH_MAX_IVAL; 669 670 rxm = strp_msg(skb); 671 672 /* head of next rec is already in, parser will sync for us */ 673 if (tcp_inq(sk) > rxm->full_len) { 674 ctx->resync_nh_do_now = 1; 675 } else { 676 struct tls_prot_info *prot = &tls_ctx->prot_info; 677 u8 rcd_sn[TLS_MAX_REC_SEQ_SIZE]; 678 679 memcpy(rcd_sn, tls_ctx->rx.rec_seq, prot->rec_seq_size); 680 tls_bigint_increment(rcd_sn, prot->rec_seq_size); 681 682 tls_device_resync_rx(tls_ctx, sk, tcp_sk(sk)->copied_seq, 683 rcd_sn); 684 } 685 } 686 687 static int tls_device_reencrypt(struct sock *sk, struct sk_buff *skb) 688 { 689 struct strp_msg *rxm = strp_msg(skb); 690 int err = 0, offset = rxm->offset, copy, nsg, data_len, pos; 691 struct sk_buff *skb_iter, *unused; 692 struct scatterlist sg[1]; 693 char *orig_buf, *buf; 694 695 orig_buf = kmalloc(rxm->full_len + TLS_HEADER_SIZE + 696 TLS_CIPHER_AES_GCM_128_IV_SIZE, sk->sk_allocation); 697 if (!orig_buf) 698 return -ENOMEM; 699 buf = orig_buf; 700 701 nsg = skb_cow_data(skb, 0, &unused); 702 if (unlikely(nsg < 0)) { 703 err = nsg; 704 goto free_buf; 705 } 706 707 sg_init_table(sg, 1); 708 sg_set_buf(&sg[0], buf, 709 rxm->full_len + TLS_HEADER_SIZE + 710 TLS_CIPHER_AES_GCM_128_IV_SIZE); 711 err = skb_copy_bits(skb, offset, buf, 712 TLS_HEADER_SIZE + TLS_CIPHER_AES_GCM_128_IV_SIZE); 713 if (err) 714 goto free_buf; 715 716 /* We are interested only in the decrypted data not the auth */ 717 err = decrypt_skb(sk, skb, sg); 718 if (err != -EBADMSG) 719 goto free_buf; 720 else 721 err = 0; 722 723 data_len = rxm->full_len - TLS_CIPHER_AES_GCM_128_TAG_SIZE; 724 725 if (skb_pagelen(skb) > offset) { 726 copy = min_t(int, skb_pagelen(skb) - offset, data_len); 727 728 if (skb->decrypted) { 729 err = skb_store_bits(skb, offset, buf, copy); 730 if (err) 731 goto free_buf; 732 } 733 734 offset += copy; 735 buf += copy; 736 } 737 738 pos = skb_pagelen(skb); 739 skb_walk_frags(skb, skb_iter) { 740 int frag_pos; 741 742 /* Practically all frags must belong to msg if reencrypt 743 * is needed with current strparser and coalescing logic, 744 * but strparser may "get optimized", so let's be safe. 745 */ 746 if (pos + skb_iter->len <= offset) 747 goto done_with_frag; 748 if (pos >= data_len + rxm->offset) 749 break; 750 751 frag_pos = offset - pos; 752 copy = min_t(int, skb_iter->len - frag_pos, 753 data_len + rxm->offset - offset); 754 755 if (skb_iter->decrypted) { 756 err = skb_store_bits(skb_iter, frag_pos, buf, copy); 757 if (err) 758 goto free_buf; 759 } 760 761 offset += copy; 762 buf += copy; 763 done_with_frag: 764 pos += skb_iter->len; 765 } 766 767 free_buf: 768 kfree(orig_buf); 769 return err; 770 } 771 772 int tls_device_decrypted(struct sock *sk, struct sk_buff *skb) 773 { 774 struct tls_context *tls_ctx = tls_get_ctx(sk); 775 struct tls_offload_context_rx *ctx = tls_offload_ctx_rx(tls_ctx); 776 int is_decrypted = skb->decrypted; 777 int is_encrypted = !is_decrypted; 778 struct sk_buff *skb_iter; 779 780 /* Check if all the data is decrypted already */ 781 skb_walk_frags(skb, skb_iter) { 782 is_decrypted &= skb_iter->decrypted; 783 is_encrypted &= !skb_iter->decrypted; 784 } 785 786 ctx->sw.decrypted |= is_decrypted; 787 788 /* Return immediately if the record is either entirely plaintext or 789 * entirely ciphertext. Otherwise handle reencrypt partially decrypted 790 * record. 791 */ 792 if (is_decrypted) { 793 ctx->resync_nh_reset = 1; 794 return 0; 795 } 796 if (is_encrypted) { 797 tls_device_core_ctrl_rx_resync(tls_ctx, ctx, sk, skb); 798 return 0; 799 } 800 801 ctx->resync_nh_reset = 1; 802 return tls_device_reencrypt(sk, skb); 803 } 804 805 static void tls_device_attach(struct tls_context *ctx, struct sock *sk, 806 struct net_device *netdev) 807 { 808 if (sk->sk_destruct != tls_device_sk_destruct) { 809 refcount_set(&ctx->refcount, 1); 810 dev_hold(netdev); 811 ctx->netdev = netdev; 812 spin_lock_irq(&tls_device_lock); 813 list_add_tail(&ctx->list, &tls_device_list); 814 spin_unlock_irq(&tls_device_lock); 815 816 ctx->sk_destruct = sk->sk_destruct; 817 sk->sk_destruct = tls_device_sk_destruct; 818 } 819 } 820 821 int tls_set_device_offload(struct sock *sk, struct tls_context *ctx) 822 { 823 u16 nonce_size, tag_size, iv_size, rec_seq_size; 824 struct tls_context *tls_ctx = tls_get_ctx(sk); 825 struct tls_prot_info *prot = &tls_ctx->prot_info; 826 struct tls_record_info *start_marker_record; 827 struct tls_offload_context_tx *offload_ctx; 828 struct tls_crypto_info *crypto_info; 829 struct net_device *netdev; 830 char *iv, *rec_seq; 831 struct sk_buff *skb; 832 int rc = -EINVAL; 833 __be64 rcd_sn; 834 835 if (!ctx) 836 goto out; 837 838 if (ctx->priv_ctx_tx) { 839 rc = -EEXIST; 840 goto out; 841 } 842 843 start_marker_record = kmalloc(sizeof(*start_marker_record), GFP_KERNEL); 844 if (!start_marker_record) { 845 rc = -ENOMEM; 846 goto out; 847 } 848 849 offload_ctx = kzalloc(TLS_OFFLOAD_CONTEXT_SIZE_TX, GFP_KERNEL); 850 if (!offload_ctx) { 851 rc = -ENOMEM; 852 goto free_marker_record; 853 } 854 855 crypto_info = &ctx->crypto_send.info; 856 if (crypto_info->version != TLS_1_2_VERSION) { 857 rc = -EOPNOTSUPP; 858 goto free_offload_ctx; 859 } 860 861 switch (crypto_info->cipher_type) { 862 case TLS_CIPHER_AES_GCM_128: 863 nonce_size = TLS_CIPHER_AES_GCM_128_IV_SIZE; 864 tag_size = TLS_CIPHER_AES_GCM_128_TAG_SIZE; 865 iv_size = TLS_CIPHER_AES_GCM_128_IV_SIZE; 866 iv = ((struct tls12_crypto_info_aes_gcm_128 *)crypto_info)->iv; 867 rec_seq_size = TLS_CIPHER_AES_GCM_128_REC_SEQ_SIZE; 868 rec_seq = 869 ((struct tls12_crypto_info_aes_gcm_128 *)crypto_info)->rec_seq; 870 break; 871 default: 872 rc = -EINVAL; 873 goto free_offload_ctx; 874 } 875 876 /* Sanity-check the rec_seq_size for stack allocations */ 877 if (rec_seq_size > TLS_MAX_REC_SEQ_SIZE) { 878 rc = -EINVAL; 879 goto free_offload_ctx; 880 } 881 882 prot->prepend_size = TLS_HEADER_SIZE + nonce_size; 883 prot->tag_size = tag_size; 884 prot->overhead_size = prot->prepend_size + prot->tag_size; 885 prot->iv_size = iv_size; 886 ctx->tx.iv = kmalloc(iv_size + TLS_CIPHER_AES_GCM_128_SALT_SIZE, 887 GFP_KERNEL); 888 if (!ctx->tx.iv) { 889 rc = -ENOMEM; 890 goto free_offload_ctx; 891 } 892 893 memcpy(ctx->tx.iv + TLS_CIPHER_AES_GCM_128_SALT_SIZE, iv, iv_size); 894 895 prot->rec_seq_size = rec_seq_size; 896 ctx->tx.rec_seq = kmemdup(rec_seq, rec_seq_size, GFP_KERNEL); 897 if (!ctx->tx.rec_seq) { 898 rc = -ENOMEM; 899 goto free_iv; 900 } 901 902 rc = tls_sw_fallback_init(sk, offload_ctx, crypto_info); 903 if (rc) 904 goto free_rec_seq; 905 906 /* start at rec_seq - 1 to account for the start marker record */ 907 memcpy(&rcd_sn, ctx->tx.rec_seq, sizeof(rcd_sn)); 908 offload_ctx->unacked_record_sn = be64_to_cpu(rcd_sn) - 1; 909 910 start_marker_record->end_seq = tcp_sk(sk)->write_seq; 911 start_marker_record->len = 0; 912 start_marker_record->num_frags = 0; 913 914 INIT_LIST_HEAD(&offload_ctx->records_list); 915 list_add_tail(&start_marker_record->list, &offload_ctx->records_list); 916 spin_lock_init(&offload_ctx->lock); 917 sg_init_table(offload_ctx->sg_tx_data, 918 ARRAY_SIZE(offload_ctx->sg_tx_data)); 919 920 clean_acked_data_enable(inet_csk(sk), &tls_icsk_clean_acked); 921 ctx->push_pending_record = tls_device_push_pending_record; 922 923 /* TLS offload is greatly simplified if we don't send 924 * SKBs where only part of the payload needs to be encrypted. 925 * So mark the last skb in the write queue as end of record. 926 */ 927 skb = tcp_write_queue_tail(sk); 928 if (skb) 929 TCP_SKB_CB(skb)->eor = 1; 930 931 /* We support starting offload on multiple sockets 932 * concurrently, so we only need a read lock here. 933 * This lock must precede get_netdev_for_sock to prevent races between 934 * NETDEV_DOWN and setsockopt. 935 */ 936 down_read(&device_offload_lock); 937 netdev = get_netdev_for_sock(sk); 938 if (!netdev) { 939 pr_err_ratelimited("%s: netdev not found\n", __func__); 940 rc = -EINVAL; 941 goto release_lock; 942 } 943 944 if (!(netdev->features & NETIF_F_HW_TLS_TX)) { 945 rc = -ENOTSUPP; 946 goto release_netdev; 947 } 948 949 /* Avoid offloading if the device is down 950 * We don't want to offload new flows after 951 * the NETDEV_DOWN event 952 */ 953 if (!(netdev->flags & IFF_UP)) { 954 rc = -EINVAL; 955 goto release_netdev; 956 } 957 958 ctx->priv_ctx_tx = offload_ctx; 959 rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_TX, 960 &ctx->crypto_send.info, 961 tcp_sk(sk)->write_seq); 962 if (rc) 963 goto release_netdev; 964 965 tls_device_attach(ctx, sk, netdev); 966 967 /* following this assignment tls_is_sk_tx_device_offloaded 968 * will return true and the context might be accessed 969 * by the netdev's xmit function. 970 */ 971 smp_store_release(&sk->sk_validate_xmit_skb, tls_validate_xmit_skb); 972 dev_put(netdev); 973 up_read(&device_offload_lock); 974 goto out; 975 976 release_netdev: 977 dev_put(netdev); 978 release_lock: 979 up_read(&device_offload_lock); 980 clean_acked_data_disable(inet_csk(sk)); 981 crypto_free_aead(offload_ctx->aead_send); 982 free_rec_seq: 983 kfree(ctx->tx.rec_seq); 984 free_iv: 985 kfree(ctx->tx.iv); 986 free_offload_ctx: 987 kfree(offload_ctx); 988 ctx->priv_ctx_tx = NULL; 989 free_marker_record: 990 kfree(start_marker_record); 991 out: 992 return rc; 993 } 994 995 int tls_set_device_offload_rx(struct sock *sk, struct tls_context *ctx) 996 { 997 struct tls_offload_context_rx *context; 998 struct net_device *netdev; 999 int rc = 0; 1000 1001 if (ctx->crypto_recv.info.version != TLS_1_2_VERSION) 1002 return -EOPNOTSUPP; 1003 1004 /* We support starting offload on multiple sockets 1005 * concurrently, so we only need a read lock here. 1006 * This lock must precede get_netdev_for_sock to prevent races between 1007 * NETDEV_DOWN and setsockopt. 1008 */ 1009 down_read(&device_offload_lock); 1010 netdev = get_netdev_for_sock(sk); 1011 if (!netdev) { 1012 pr_err_ratelimited("%s: netdev not found\n", __func__); 1013 rc = -EINVAL; 1014 goto release_lock; 1015 } 1016 1017 if (!(netdev->features & NETIF_F_HW_TLS_RX)) { 1018 rc = -ENOTSUPP; 1019 goto release_netdev; 1020 } 1021 1022 /* Avoid offloading if the device is down 1023 * We don't want to offload new flows after 1024 * the NETDEV_DOWN event 1025 */ 1026 if (!(netdev->flags & IFF_UP)) { 1027 rc = -EINVAL; 1028 goto release_netdev; 1029 } 1030 1031 context = kzalloc(TLS_OFFLOAD_CONTEXT_SIZE_RX, GFP_KERNEL); 1032 if (!context) { 1033 rc = -ENOMEM; 1034 goto release_netdev; 1035 } 1036 context->resync_nh_reset = 1; 1037 1038 ctx->priv_ctx_rx = context; 1039 rc = tls_set_sw_offload(sk, ctx, 0); 1040 if (rc) 1041 goto release_ctx; 1042 1043 rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_RX, 1044 &ctx->crypto_recv.info, 1045 tcp_sk(sk)->copied_seq); 1046 if (rc) 1047 goto free_sw_resources; 1048 1049 tls_device_attach(ctx, sk, netdev); 1050 goto release_netdev; 1051 1052 free_sw_resources: 1053 up_read(&device_offload_lock); 1054 tls_sw_free_resources_rx(sk); 1055 down_read(&device_offload_lock); 1056 release_ctx: 1057 ctx->priv_ctx_rx = NULL; 1058 release_netdev: 1059 dev_put(netdev); 1060 release_lock: 1061 up_read(&device_offload_lock); 1062 return rc; 1063 } 1064 1065 void tls_device_offload_cleanup_rx(struct sock *sk) 1066 { 1067 struct tls_context *tls_ctx = tls_get_ctx(sk); 1068 struct net_device *netdev; 1069 1070 down_read(&device_offload_lock); 1071 netdev = tls_ctx->netdev; 1072 if (!netdev) 1073 goto out; 1074 1075 netdev->tlsdev_ops->tls_dev_del(netdev, tls_ctx, 1076 TLS_OFFLOAD_CTX_DIR_RX); 1077 1078 if (tls_ctx->tx_conf != TLS_HW) { 1079 dev_put(netdev); 1080 tls_ctx->netdev = NULL; 1081 } 1082 out: 1083 up_read(&device_offload_lock); 1084 tls_sw_release_resources_rx(sk); 1085 } 1086 1087 static int tls_device_down(struct net_device *netdev) 1088 { 1089 struct tls_context *ctx, *tmp; 1090 unsigned long flags; 1091 LIST_HEAD(list); 1092 1093 /* Request a write lock to block new offload attempts */ 1094 down_write(&device_offload_lock); 1095 1096 spin_lock_irqsave(&tls_device_lock, flags); 1097 list_for_each_entry_safe(ctx, tmp, &tls_device_list, list) { 1098 if (ctx->netdev != netdev || 1099 !refcount_inc_not_zero(&ctx->refcount)) 1100 continue; 1101 1102 list_move(&ctx->list, &list); 1103 } 1104 spin_unlock_irqrestore(&tls_device_lock, flags); 1105 1106 list_for_each_entry_safe(ctx, tmp, &list, list) { 1107 if (ctx->tx_conf == TLS_HW) 1108 netdev->tlsdev_ops->tls_dev_del(netdev, ctx, 1109 TLS_OFFLOAD_CTX_DIR_TX); 1110 if (ctx->rx_conf == TLS_HW) 1111 netdev->tlsdev_ops->tls_dev_del(netdev, ctx, 1112 TLS_OFFLOAD_CTX_DIR_RX); 1113 WRITE_ONCE(ctx->netdev, NULL); 1114 smp_mb__before_atomic(); /* pairs with test_and_set_bit() */ 1115 while (test_bit(TLS_RX_SYNC_RUNNING, &ctx->flags)) 1116 usleep_range(10, 200); 1117 dev_put(netdev); 1118 list_del_init(&ctx->list); 1119 1120 if (refcount_dec_and_test(&ctx->refcount)) 1121 tls_device_free_ctx(ctx); 1122 } 1123 1124 up_write(&device_offload_lock); 1125 1126 flush_work(&tls_device_gc_work); 1127 1128 return NOTIFY_DONE; 1129 } 1130 1131 static int tls_dev_event(struct notifier_block *this, unsigned long event, 1132 void *ptr) 1133 { 1134 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1135 1136 if (!dev->tlsdev_ops && 1137 !(dev->features & (NETIF_F_HW_TLS_RX | NETIF_F_HW_TLS_TX))) 1138 return NOTIFY_DONE; 1139 1140 switch (event) { 1141 case NETDEV_REGISTER: 1142 case NETDEV_FEAT_CHANGE: 1143 if ((dev->features & NETIF_F_HW_TLS_RX) && 1144 !dev->tlsdev_ops->tls_dev_resync) 1145 return NOTIFY_BAD; 1146 1147 if (dev->tlsdev_ops && 1148 dev->tlsdev_ops->tls_dev_add && 1149 dev->tlsdev_ops->tls_dev_del) 1150 return NOTIFY_DONE; 1151 else 1152 return NOTIFY_BAD; 1153 case NETDEV_DOWN: 1154 return tls_device_down(dev); 1155 } 1156 return NOTIFY_DONE; 1157 } 1158 1159 static struct notifier_block tls_dev_notifier = { 1160 .notifier_call = tls_dev_event, 1161 }; 1162 1163 void __init tls_device_init(void) 1164 { 1165 register_netdevice_notifier(&tls_dev_notifier); 1166 } 1167 1168 void __exit tls_device_cleanup(void) 1169 { 1170 unregister_netdevice_notifier(&tls_dev_notifier); 1171 flush_work(&tls_device_gc_work); 1172 clean_acked_data_flush(); 1173 } 1174