1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * Stream Parser 4 * 5 * Copyright (c) 2016 Tom Herbert <tom@herbertland.com> 6 */ 7 8 #include <linux/bpf.h> 9 #include <linux/errno.h> 10 #include <linux/errqueue.h> 11 #include <linux/file.h> 12 #include <linux/in.h> 13 #include <linux/kernel.h> 14 #include <linux/export.h> 15 #include <linux/init.h> 16 #include <linux/net.h> 17 #include <linux/netdevice.h> 18 #include <linux/poll.h> 19 #include <linux/rculist.h> 20 #include <linux/skbuff.h> 21 #include <linux/socket.h> 22 #include <linux/uaccess.h> 23 #include <linux/workqueue.h> 24 #include <net/strparser.h> 25 #include <net/netns/generic.h> 26 #include <net/sock.h> 27 28 static struct workqueue_struct *strp_wq; 29 30 static inline struct _strp_msg *_strp_msg(struct sk_buff *skb) 31 { 32 return (struct _strp_msg *)((void *)skb->cb + 33 offsetof(struct sk_skb_cb, strp)); 34 } 35 36 /* Lower lock held */ 37 static void strp_abort_strp(struct strparser *strp, int err) 38 { 39 /* Unrecoverable error in receive */ 40 41 cancel_delayed_work(&strp->msg_timer_work); 42 43 if (strp->stopped) 44 return; 45 46 strp->stopped = 1; 47 48 if (strp->skb_head) { 49 kfree_skb(strp->skb_head); 50 strp->skb_head = NULL; 51 } 52 53 strp->skb_nextp = NULL; 54 strp->need_bytes = 0; 55 56 if (strp->sk) { 57 struct sock *sk = strp->sk; 58 59 /* Report an error on the lower socket */ 60 sk->sk_err = -err; 61 sk_error_report(sk); 62 } 63 } 64 65 static void strp_start_timer(struct strparser *strp, long timeo) 66 { 67 if (timeo && timeo != LONG_MAX) 68 mod_delayed_work(strp_wq, &strp->msg_timer_work, timeo); 69 } 70 71 /* Lower lock held */ 72 static void strp_parser_err(struct strparser *strp, int err, 73 read_descriptor_t *desc) 74 { 75 desc->error = err; 76 kfree_skb(strp->skb_head); 77 strp->skb_head = NULL; 78 strp->cb.abort_parser(strp, err); 79 } 80 81 static inline int strp_peek_len(struct strparser *strp) 82 { 83 if (strp->sk) { 84 struct socket *sock = strp->sk->sk_socket; 85 86 return sock->ops->peek_len(sock); 87 } 88 89 /* If we don't have an associated socket there's nothing to peek. 90 * Return int max to avoid stopping the strparser. 91 */ 92 93 return INT_MAX; 94 } 95 96 /* Lower socket lock held */ 97 static int __strp_recv(read_descriptor_t *desc, struct sk_buff *orig_skb, 98 unsigned int orig_offset, size_t orig_len, 99 size_t max_msg_size, long timeo) 100 { 101 struct strparser *strp = (struct strparser *)desc->arg.data; 102 struct _strp_msg *stm; 103 struct sk_buff *head, *skb; 104 size_t eaten = 0, cand_len; 105 ssize_t extra; 106 int err; 107 bool cloned_orig = false; 108 109 if (strp->paused) 110 return 0; 111 112 head = strp->skb_head; 113 if (head) { 114 /* Message already in progress */ 115 if (unlikely(orig_offset)) { 116 /* Getting data with a non-zero offset when a message is 117 * in progress is not expected. If it does happen, we 118 * need to clone and pull since we can't deal with 119 * offsets in the skbs for a message expect in the head. 120 */ 121 orig_skb = skb_clone(orig_skb, GFP_ATOMIC); 122 if (!orig_skb) { 123 STRP_STATS_INCR(strp->stats.mem_fail); 124 desc->error = -ENOMEM; 125 return 0; 126 } 127 if (!pskb_pull(orig_skb, orig_offset)) { 128 STRP_STATS_INCR(strp->stats.mem_fail); 129 kfree_skb(orig_skb); 130 desc->error = -ENOMEM; 131 return 0; 132 } 133 cloned_orig = true; 134 orig_offset = 0; 135 } 136 137 if (!strp->skb_nextp) { 138 /* We are going to append to the frag_list of head. 139 * Need to unshare the frag_list. 140 */ 141 err = skb_unclone(head, GFP_ATOMIC); 142 if (err) { 143 STRP_STATS_INCR(strp->stats.mem_fail); 144 desc->error = err; 145 return 0; 146 } 147 148 if (unlikely(skb_shinfo(head)->frag_list)) { 149 /* We can't append to an sk_buff that already 150 * has a frag_list. We create a new head, point 151 * the frag_list of that to the old head, and 152 * then are able to use the old head->next for 153 * appending to the message. 154 */ 155 if (WARN_ON(head->next)) { 156 desc->error = -EINVAL; 157 return 0; 158 } 159 160 skb = alloc_skb_for_msg(head); 161 if (!skb) { 162 STRP_STATS_INCR(strp->stats.mem_fail); 163 desc->error = -ENOMEM; 164 return 0; 165 } 166 167 strp->skb_nextp = &head->next; 168 strp->skb_head = skb; 169 head = skb; 170 } else { 171 strp->skb_nextp = 172 &skb_shinfo(head)->frag_list; 173 } 174 } 175 } 176 177 while (eaten < orig_len) { 178 /* Always clone since we will consume something */ 179 skb = skb_clone(orig_skb, GFP_ATOMIC); 180 if (!skb) { 181 STRP_STATS_INCR(strp->stats.mem_fail); 182 desc->error = -ENOMEM; 183 break; 184 } 185 186 cand_len = orig_len - eaten; 187 188 head = strp->skb_head; 189 if (!head) { 190 head = skb; 191 strp->skb_head = head; 192 /* Will set skb_nextp on next packet if needed */ 193 strp->skb_nextp = NULL; 194 stm = _strp_msg(head); 195 memset(stm, 0, sizeof(*stm)); 196 stm->strp.offset = orig_offset + eaten; 197 } else { 198 /* Unclone if we are appending to an skb that we 199 * already share a frag_list with. 200 */ 201 if (skb_has_frag_list(skb)) { 202 err = skb_unclone(skb, GFP_ATOMIC); 203 if (err) { 204 STRP_STATS_INCR(strp->stats.mem_fail); 205 desc->error = err; 206 break; 207 } 208 } 209 210 stm = _strp_msg(head); 211 *strp->skb_nextp = skb; 212 strp->skb_nextp = &skb->next; 213 head->data_len += skb->len; 214 head->len += skb->len; 215 head->truesize += skb->truesize; 216 } 217 218 if (!stm->strp.full_len) { 219 ssize_t len; 220 221 len = (*strp->cb.parse_msg)(strp, head); 222 223 if (!len) { 224 /* Need more header to determine length */ 225 if (!stm->accum_len) { 226 /* Start RX timer for new message */ 227 strp_start_timer(strp, timeo); 228 } 229 stm->accum_len += cand_len; 230 eaten += cand_len; 231 STRP_STATS_INCR(strp->stats.need_more_hdr); 232 WARN_ON(eaten != orig_len); 233 break; 234 } else if (len < 0) { 235 if (len == -ESTRPIPE && stm->accum_len) { 236 len = -ENODATA; 237 strp->unrecov_intr = 1; 238 } else { 239 strp->interrupted = 1; 240 } 241 strp_parser_err(strp, len, desc); 242 break; 243 } else if (len > max_msg_size) { 244 /* Message length exceeds maximum allowed */ 245 STRP_STATS_INCR(strp->stats.msg_too_big); 246 strp_parser_err(strp, -EMSGSIZE, desc); 247 break; 248 } else if (len <= (ssize_t)head->len - 249 (ssize_t)skb->len - stm->strp.offset) { 250 /* Length must be into new skb (and also 251 * greater than zero) 252 */ 253 STRP_STATS_INCR(strp->stats.bad_hdr_len); 254 strp_parser_err(strp, -EPROTO, desc); 255 break; 256 } 257 258 stm->strp.full_len = len; 259 } 260 261 extra = (ssize_t)(stm->accum_len + cand_len) - 262 stm->strp.full_len; 263 264 if (extra < 0) { 265 /* Message not complete yet. */ 266 if (stm->strp.full_len - stm->accum_len > 267 strp_peek_len(strp)) { 268 /* Don't have the whole message in the socket 269 * buffer. Set strp->need_bytes to wait for 270 * the rest of the message. Also, set "early 271 * eaten" since we've already buffered the skb 272 * but don't consume yet per strp_read_sock. 273 */ 274 275 if (!stm->accum_len) { 276 /* Start RX timer for new message */ 277 strp_start_timer(strp, timeo); 278 } 279 280 stm->accum_len += cand_len; 281 eaten += cand_len; 282 strp->need_bytes = stm->strp.full_len - 283 stm->accum_len; 284 STRP_STATS_ADD(strp->stats.bytes, cand_len); 285 desc->count = 0; /* Stop reading socket */ 286 break; 287 } 288 stm->accum_len += cand_len; 289 eaten += cand_len; 290 WARN_ON(eaten != orig_len); 291 break; 292 } 293 294 /* Positive extra indicates more bytes than needed for the 295 * message 296 */ 297 298 WARN_ON(extra > cand_len); 299 300 eaten += (cand_len - extra); 301 302 /* Hurray, we have a new message! */ 303 cancel_delayed_work(&strp->msg_timer_work); 304 strp->skb_head = NULL; 305 strp->need_bytes = 0; 306 STRP_STATS_INCR(strp->stats.msgs); 307 308 /* Give skb to upper layer */ 309 strp->cb.rcv_msg(strp, head); 310 311 if (unlikely(strp->paused)) { 312 /* Upper layer paused strp */ 313 break; 314 } 315 } 316 317 if (cloned_orig) 318 kfree_skb(orig_skb); 319 320 STRP_STATS_ADD(strp->stats.bytes, eaten); 321 322 return eaten; 323 } 324 325 int strp_process(struct strparser *strp, struct sk_buff *orig_skb, 326 unsigned int orig_offset, size_t orig_len, 327 size_t max_msg_size, long timeo) 328 { 329 read_descriptor_t desc; /* Dummy arg to strp_recv */ 330 331 desc.arg.data = strp; 332 333 return __strp_recv(&desc, orig_skb, orig_offset, orig_len, 334 max_msg_size, timeo); 335 } 336 EXPORT_SYMBOL_GPL(strp_process); 337 338 static int strp_recv(read_descriptor_t *desc, struct sk_buff *orig_skb, 339 unsigned int orig_offset, size_t orig_len) 340 { 341 struct strparser *strp = (struct strparser *)desc->arg.data; 342 343 return __strp_recv(desc, orig_skb, orig_offset, orig_len, 344 strp->sk->sk_rcvbuf, READ_ONCE(strp->sk->sk_rcvtimeo)); 345 } 346 347 static int default_read_sock_done(struct strparser *strp, int err) 348 { 349 return err; 350 } 351 352 /* Called with lock held on lower socket */ 353 static int strp_read_sock(struct strparser *strp) 354 { 355 struct socket *sock = strp->sk->sk_socket; 356 read_descriptor_t desc; 357 358 if (unlikely(!sock || !sock->ops)) 359 return -EBUSY; 360 361 if (unlikely(!strp->cb.read_sock && !sock->ops->read_sock)) 362 return -EBUSY; 363 364 desc.arg.data = strp; 365 desc.error = 0; 366 desc.count = 1; /* give more than one skb per call */ 367 368 /* sk should be locked here, so okay to do read_sock */ 369 if (strp->cb.read_sock) 370 strp->cb.read_sock(strp, &desc, strp_recv); 371 else 372 sock->ops->read_sock(strp->sk, &desc, strp_recv); 373 374 desc.error = strp->cb.read_sock_done(strp, desc.error); 375 376 return desc.error; 377 } 378 379 /* Lower sock lock held */ 380 void strp_data_ready(struct strparser *strp) 381 { 382 if (unlikely(strp->stopped) || strp->paused) 383 return; 384 385 /* This check is needed to synchronize with do_strp_work. 386 * do_strp_work acquires a process lock (lock_sock) whereas 387 * the lock held here is bh_lock_sock. The two locks can be 388 * held by different threads at the same time, but bh_lock_sock 389 * allows a thread in BH context to safely check if the process 390 * lock is held. In this case, if the lock is held, queue work. 391 */ 392 if (sock_owned_by_user_nocheck(strp->sk)) { 393 queue_work(strp_wq, &strp->work); 394 return; 395 } 396 397 if (strp->need_bytes) { 398 if (strp_peek_len(strp) < strp->need_bytes) 399 return; 400 } 401 402 if (strp_read_sock(strp) == -ENOMEM) 403 queue_work(strp_wq, &strp->work); 404 } 405 EXPORT_SYMBOL_GPL(strp_data_ready); 406 407 static void do_strp_work(struct strparser *strp) 408 { 409 /* We need the read lock to synchronize with strp_data_ready. We 410 * need the socket lock for calling strp_read_sock. 411 */ 412 strp->cb.lock(strp); 413 414 if (unlikely(strp->stopped)) 415 goto out; 416 417 if (strp->paused) 418 goto out; 419 420 if (strp_read_sock(strp) == -ENOMEM) 421 queue_work(strp_wq, &strp->work); 422 423 out: 424 strp->cb.unlock(strp); 425 } 426 427 static void strp_work(struct work_struct *w) 428 { 429 do_strp_work(container_of(w, struct strparser, work)); 430 } 431 432 static void strp_msg_timeout(struct work_struct *w) 433 { 434 struct strparser *strp = container_of(w, struct strparser, 435 msg_timer_work.work); 436 437 /* Message assembly timed out */ 438 STRP_STATS_INCR(strp->stats.msg_timeouts); 439 strp->cb.lock(strp); 440 strp->cb.abort_parser(strp, -ETIMEDOUT); 441 strp->cb.unlock(strp); 442 } 443 444 static void strp_sock_lock(struct strparser *strp) 445 { 446 lock_sock(strp->sk); 447 } 448 449 static void strp_sock_unlock(struct strparser *strp) 450 { 451 release_sock(strp->sk); 452 } 453 454 int strp_init(struct strparser *strp, struct sock *sk, 455 const struct strp_callbacks *cb) 456 { 457 458 if (!cb || !cb->rcv_msg || !cb->parse_msg) 459 return -EINVAL; 460 461 /* The sk (sock) arg determines the mode of the stream parser. 462 * 463 * If the sock is set then the strparser is in receive callback mode. 464 * The upper layer calls strp_data_ready to kick receive processing 465 * and strparser calls the read_sock function on the socket to 466 * get packets. 467 * 468 * If the sock is not set then the strparser is in general mode. 469 * The upper layer calls strp_process for each skb to be parsed. 470 */ 471 472 if (!sk) { 473 if (!cb->lock || !cb->unlock) 474 return -EINVAL; 475 } 476 477 memset(strp, 0, sizeof(*strp)); 478 479 strp->sk = sk; 480 481 strp->cb.lock = cb->lock ? : strp_sock_lock; 482 strp->cb.unlock = cb->unlock ? : strp_sock_unlock; 483 strp->cb.rcv_msg = cb->rcv_msg; 484 strp->cb.parse_msg = cb->parse_msg; 485 strp->cb.read_sock = cb->read_sock; 486 strp->cb.read_sock_done = cb->read_sock_done ? : default_read_sock_done; 487 strp->cb.abort_parser = cb->abort_parser ? : strp_abort_strp; 488 489 INIT_DELAYED_WORK(&strp->msg_timer_work, strp_msg_timeout); 490 INIT_WORK(&strp->work, strp_work); 491 492 return 0; 493 } 494 EXPORT_SYMBOL_GPL(strp_init); 495 496 void strp_unpause(struct strparser *strp) 497 { 498 strp->paused = 0; 499 500 /* Sync setting paused with RX work */ 501 smp_mb(); 502 503 queue_work(strp_wq, &strp->work); 504 } 505 EXPORT_SYMBOL_GPL(strp_unpause); 506 507 /* strp must already be stopped so that strp_recv will no longer be called. 508 * Note that strp_done is not called with the lower socket held. 509 */ 510 void strp_done(struct strparser *strp) 511 { 512 WARN_ON(!strp->stopped); 513 514 cancel_delayed_work_sync(&strp->msg_timer_work); 515 cancel_work_sync(&strp->work); 516 517 if (strp->skb_head) { 518 kfree_skb(strp->skb_head); 519 strp->skb_head = NULL; 520 } 521 } 522 EXPORT_SYMBOL_GPL(strp_done); 523 524 void strp_stop(struct strparser *strp) 525 { 526 strp->stopped = 1; 527 } 528 EXPORT_SYMBOL_GPL(strp_stop); 529 530 void strp_check_rcv(struct strparser *strp) 531 { 532 queue_work(strp_wq, &strp->work); 533 } 534 EXPORT_SYMBOL_GPL(strp_check_rcv); 535 536 static int __init strp_dev_init(void) 537 { 538 BUILD_BUG_ON(sizeof(struct sk_skb_cb) > 539 sizeof_field(struct sk_buff, cb)); 540 541 strp_wq = create_singlethread_workqueue("kstrp"); 542 if (unlikely(!strp_wq)) 543 return -ENOMEM; 544 545 return 0; 546 } 547 device_initcall(strp_dev_init); 548