net/tls/tls_device.c

   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
  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
  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
  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 */
 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 = netdev_sk_get_lowest_dev(dst->dev, sk);
 121                 dev_hold(netdev);
 122         }
 123
 124         dst_release(dst);
 125
 126         return netdev;
 127 }
 128
 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], false);
 135         kfree(record);
 136 }
 137
 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
 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  */
 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
 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
 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
 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
 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
 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
 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);
 334         return ret;
 335 }
 336
 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
 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
 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
 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;
 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
 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
 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
 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 beginning 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
 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
 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
 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
 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
 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
 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
 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
 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
 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
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, salt_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                 salt_size = TLS_CIPHER_AES_GCM_128_SALT_SIZE;
1059                 rec_seq =
1060                  ((struct tls12_crypto_info_aes_gcm_128 *)crypto_info)->rec_seq;
1061                 break;
1062         default:
1063                 rc = -EINVAL;
1064                 goto free_offload_ctx;
1065         }
1066
1067         /* Sanity-check the rec_seq_size for stack allocations */
1068         if (rec_seq_size > TLS_MAX_REC_SEQ_SIZE) {
1069                 rc = -EINVAL;
1070                 goto free_offload_ctx;
1071         }
1072
1073         prot->version = crypto_info->version;
1074         prot->cipher_type = crypto_info->cipher_type;
1075         prot->prepend_size = TLS_HEADER_SIZE + nonce_size;
1076         prot->tag_size = tag_size;
1077         prot->overhead_size = prot->prepend_size + prot->tag_size;
1078         prot->iv_size = iv_size;
1079         prot->salt_size = salt_size;
1080         ctx->tx.iv = kmalloc(iv_size + TLS_CIPHER_AES_GCM_128_SALT_SIZE,
1081                              GFP_KERNEL);
1082         if (!ctx->tx.iv) {
1083                 rc = -ENOMEM;
1084                 goto free_offload_ctx;
1085         }
1086
1087         memcpy(ctx->tx.iv + TLS_CIPHER_AES_GCM_128_SALT_SIZE, iv, iv_size);
1088
1089         prot->rec_seq_size = rec_seq_size;
1090         ctx->tx.rec_seq = kmemdup(rec_seq, rec_seq_size, GFP_KERNEL);
1091         if (!ctx->tx.rec_seq) {
1092                 rc = -ENOMEM;
1093                 goto free_iv;
1094         }
1095
1096         rc = tls_sw_fallback_init(sk, offload_ctx, crypto_info);
1097         if (rc)
1098                 goto free_rec_seq;
1099
1100         /* start at rec_seq - 1 to account for the start marker record */
1101         memcpy(&rcd_sn, ctx->tx.rec_seq, sizeof(rcd_sn));
1102         offload_ctx->unacked_record_sn = be64_to_cpu(rcd_sn) - 1;
1103
1104         start_marker_record->end_seq = tcp_sk(sk)->write_seq;
1105         start_marker_record->len = 0;
1106         start_marker_record->num_frags = 0;
1107
1108         INIT_LIST_HEAD(&offload_ctx->records_list);
1109         list_add_tail(&start_marker_record->list, &offload_ctx->records_list);
1110         spin_lock_init(&offload_ctx->lock);
1111         sg_init_table(offload_ctx->sg_tx_data,
1112                       ARRAY_SIZE(offload_ctx->sg_tx_data));
1113
1114         clean_acked_data_enable(inet_csk(sk), &tls_icsk_clean_acked);
1115         ctx->push_pending_record = tls_device_push_pending_record;
1116
1117         /* TLS offload is greatly simplified if we don't send
1118          * SKBs where only part of the payload needs to be encrypted.
1119          * So mark the last skb in the write queue as end of record.
1120          */
1121         skb = tcp_write_queue_tail(sk);
1122         if (skb)
1123                 TCP_SKB_CB(skb)->eor = 1;
1124
1125         netdev = get_netdev_for_sock(sk);
1126         if (!netdev) {
1127                 pr_err_ratelimited("%s: netdev not found\n", __func__);
1128                 rc = -EINVAL;
1129                 goto disable_cad;
1130         }
1131
1132         if (!(netdev->features & NETIF_F_HW_TLS_TX)) {
1133                 rc = -EOPNOTSUPP;
1134                 goto release_netdev;
1135         }
1136
1137         /* Avoid offloading if the device is down
1138          * We don't want to offload new flows after
1139          * the NETDEV_DOWN event
1140          *
1141          * device_offload_lock is taken in tls_devices's NETDEV_DOWN
1142          * handler thus protecting from the device going down before
1143          * ctx was added to tls_device_list.
1144          */
1145         down_read(&device_offload_lock);
1146         if (!(netdev->flags & IFF_UP)) {
1147                 rc = -EINVAL;
1148                 goto release_lock;
1149         }
1150
1151         ctx->priv_ctx_tx = offload_ctx;
1152         rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_TX,
1153                                              &ctx->crypto_send.info,
1154                                              tcp_sk(sk)->write_seq);
1155         trace_tls_device_offload_set(sk, TLS_OFFLOAD_CTX_DIR_TX,
1156                                      tcp_sk(sk)->write_seq, rec_seq, rc);
1157         if (rc)
1158                 goto release_lock;
1159
1160         tls_device_attach(ctx, sk, netdev);
1161         up_read(&device_offload_lock);
1162
1163         /* following this assignment tls_is_sk_tx_device_offloaded
1164          * will return true and the context might be accessed
1165          * by the netdev's xmit function.
1166          */
1167         smp_store_release(&sk->sk_validate_xmit_skb, tls_validate_xmit_skb);
1168         dev_put(netdev);
1169
1170         return 0;
1171
1172 release_lock:
1173         up_read(&device_offload_lock);
1174 release_netdev:
1175         dev_put(netdev);
1176 disable_cad:
1177         clean_acked_data_disable(inet_csk(sk));
1178         crypto_free_aead(offload_ctx->aead_send);
1179 free_rec_seq:
1180         kfree(ctx->tx.rec_seq);
1181 free_iv:
1182         kfree(ctx->tx.iv);
1183 free_offload_ctx:
1184         kfree(offload_ctx);
1185         ctx->priv_ctx_tx = NULL;
1186 free_marker_record:
1187         kfree(start_marker_record);
1188         return rc;
1189 }
1190
1191 int tls_set_device_offload_rx(struct sock *sk, struct tls_context *ctx)
1192 {
1193         struct tls12_crypto_info_aes_gcm_128 *info;
1194         struct tls_offload_context_rx *context;
1195         struct net_device *netdev;
1196         int rc = 0;
1197
1198         if (ctx->crypto_recv.info.version != TLS_1_2_VERSION)
1199                 return -EOPNOTSUPP;
1200
1201         netdev = get_netdev_for_sock(sk);
1202         if (!netdev) {
1203                 pr_err_ratelimited("%s: netdev not found\n", __func__);
1204                 return -EINVAL;
1205         }
1206
1207         if (!(netdev->features & NETIF_F_HW_TLS_RX)) {
1208                 rc = -EOPNOTSUPP;
1209                 goto release_netdev;
1210         }
1211
1212         /* Avoid offloading if the device is down
1213          * We don't want to offload new flows after
1214          * the NETDEV_DOWN event
1215          *
1216          * device_offload_lock is taken in tls_devices's NETDEV_DOWN
1217          * handler thus protecting from the device going down before
1218          * ctx was added to tls_device_list.
1219          */
1220         down_read(&device_offload_lock);
1221         if (!(netdev->flags & IFF_UP)) {
1222                 rc = -EINVAL;
1223                 goto release_lock;
1224         }
1225
1226         context = kzalloc(TLS_OFFLOAD_CONTEXT_SIZE_RX, GFP_KERNEL);
1227         if (!context) {
1228                 rc = -ENOMEM;
1229                 goto release_lock;
1230         }
1231         context->resync_nh_reset = 1;
1232
1233         ctx->priv_ctx_rx = context;
1234         rc = tls_set_sw_offload(sk, ctx, 0);
1235         if (rc)
1236                 goto release_ctx;
1237
1238         rc = netdev->tlsdev_ops->tls_dev_add(netdev, sk, TLS_OFFLOAD_CTX_DIR_RX,
1239                                              &ctx->crypto_recv.info,
1240                                              tcp_sk(sk)->copied_seq);
1241         info = (void *)&ctx->crypto_recv.info;
1242         trace_tls_device_offload_set(sk, TLS_OFFLOAD_CTX_DIR_RX,
1243                                      tcp_sk(sk)->copied_seq, info->rec_seq, rc);
1244         if (rc)
1245                 goto free_sw_resources;
1246
1247         tls_device_attach(ctx, sk, netdev);
1248         up_read(&device_offload_lock);
1249
1250         dev_put(netdev);
1251
1252         return 0;
1253
1254 free_sw_resources:
1255         up_read(&device_offload_lock);
1256         tls_sw_free_resources_rx(sk);
1257         down_read(&device_offload_lock);
1258 release_ctx:
1259         ctx->priv_ctx_rx = NULL;
1260 release_lock:
1261         up_read(&device_offload_lock);
1262 release_netdev:
1263         dev_put(netdev);
1264         return rc;
1265 }
1266
1267 void tls_device_offload_cleanup_rx(struct sock *sk)
1268 {
1269         struct tls_context *tls_ctx = tls_get_ctx(sk);
1270         struct net_device *netdev;
1271
1272         down_read(&device_offload_lock);
1273         netdev = tls_ctx->netdev;
1274         if (!netdev)
1275                 goto out;
1276
1277         netdev->tlsdev_ops->tls_dev_del(netdev, tls_ctx,
1278                                         TLS_OFFLOAD_CTX_DIR_RX);
1279
1280         if (tls_ctx->tx_conf != TLS_HW) {
1281                 dev_put(netdev);
1282                 tls_ctx->netdev = NULL;
1283         } else {
1284                 set_bit(TLS_RX_DEV_CLOSED, &tls_ctx->flags);
1285         }
1286 out:
1287         up_read(&device_offload_lock);
1288         tls_sw_release_resources_rx(sk);
1289 }
1290
1291 static int tls_device_down(struct net_device *netdev)
1292 {
1293         struct tls_context *ctx, *tmp;
1294         unsigned long flags;
1295         LIST_HEAD(list);
1296
1297         /* Request a write lock to block new offload attempts */
1298         down_write(&device_offload_lock);
1299
1300         spin_lock_irqsave(&tls_device_lock, flags);
1301         list_for_each_entry_safe(ctx, tmp, &tls_device_list, list) {
1302                 if (ctx->netdev != netdev ||
1303                     !refcount_inc_not_zero(&ctx->refcount))
1304                         continue;
1305
1306                 list_move(&ctx->list, &list);
1307         }
1308         spin_unlock_irqrestore(&tls_device_lock, flags);
1309
1310         list_for_each_entry_safe(ctx, tmp, &list, list) {
1311                 /* Stop offloaded TX and switch to the fallback.
1312                  * tls_is_sk_tx_device_offloaded will return false.
1313                  */
1314                 WRITE_ONCE(ctx->sk->sk_validate_xmit_skb, tls_validate_xmit_skb_sw);
1315
1316                 /* Stop the RX and TX resync.
1317                  * tls_dev_resync must not be called after tls_dev_del.
1318                  */
1319                 WRITE_ONCE(ctx->netdev, NULL);
1320
1321                 /* Start skipping the RX resync logic completely. */
1322                 set_bit(TLS_RX_DEV_DEGRADED, &ctx->flags);
1323
1324                 /* Sync with inflight packets. After this point:
1325                  * TX: no non-encrypted packets will be passed to the driver.
1326                  * RX: resync requests from the driver will be ignored.
1327                  */
1328                 synchronize_net();
1329
1330                 /* Release the offload context on the driver side. */
1331                 if (ctx->tx_conf == TLS_HW)
1332                         netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
1333                                                         TLS_OFFLOAD_CTX_DIR_TX);
1334                 if (ctx->rx_conf == TLS_HW &&
1335                     !test_bit(TLS_RX_DEV_CLOSED, &ctx->flags))
1336                         netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
1337                                                         TLS_OFFLOAD_CTX_DIR_RX);
1338
1339                 dev_put(netdev);
1340
1341                 /* Move the context to a separate list for two reasons:
1342                  * 1. When the context is deallocated, list_del is called.
1343                  * 2. It's no longer an offloaded context, so we don't want to
1344                  *    run offload-specific code on this context.
1345                  */
1346                 spin_lock_irqsave(&tls_device_lock, flags);
1347                 list_move_tail(&ctx->list, &tls_device_down_list);
1348                 spin_unlock_irqrestore(&tls_device_lock, flags);
1349
1350                 /* Device contexts for RX and TX will be freed in on sk_destruct
1351                  * by tls_device_free_ctx. rx_conf and tx_conf stay in TLS_HW.
1352                  * Now release the ref taken above.
1353                  */
1354                 if (refcount_dec_and_test(&ctx->refcount)) {
1355                         /* sk_destruct ran after tls_device_down took a ref, and
1356                          * it returned early. Complete the destruction here.
1357                          */
1358                         list_del(&ctx->list);
1359                         tls_device_free_ctx(ctx);
1360                 }
1361         }
1362
1363         up_write(&device_offload_lock);
1364
1365         flush_work(&tls_device_gc_work);
1366
1367         return NOTIFY_DONE;
1368 }
1369
1370 static int tls_dev_event(struct notifier_block *this, unsigned long event,
1371                          void *ptr)
1372 {
1373         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1374
1375         if (!dev->tlsdev_ops &&
1376             !(dev->features & (NETIF_F_HW_TLS_RX | NETIF_F_HW_TLS_TX)))
1377                 return NOTIFY_DONE;
1378
1379         switch (event) {
1380         case NETDEV_REGISTER:
1381         case NETDEV_FEAT_CHANGE:
1382                 if (netif_is_bond_master(dev))
1383                         return NOTIFY_DONE;
1384                 if ((dev->features & NETIF_F_HW_TLS_RX) &&
1385                     !dev->tlsdev_ops->tls_dev_resync)
1386                         return NOTIFY_BAD;
1387
1388                 if  (dev->tlsdev_ops &&
1389                      dev->tlsdev_ops->tls_dev_add &&
1390                      dev->tlsdev_ops->tls_dev_del)
1391                         return NOTIFY_DONE;
1392                 else
1393                         return NOTIFY_BAD;
1394         case NETDEV_DOWN:
1395                 return tls_device_down(dev);
1396         }
1397         return NOTIFY_DONE;
1398 }
1399
1400 static struct notifier_block tls_dev_notifier = {
1401         .notifier_call  = tls_dev_event,
1402 };
1403
1404 int __init tls_device_init(void)
1405 {
1406         return register_netdevice_notifier(&tls_dev_notifier);
1407 }
1408
1409 void __exit tls_device_cleanup(void)
1410 {
1411         unregister_netdevice_notifier(&tls_dev_notifier);
1412         flush_work(&tls_device_gc_work);
1413         clean_acked_data_flush();
1414 }