1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 drbd_req.c 4 5 This file is part of DRBD by Philipp Reisner and Lars Ellenberg. 6 7 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. 8 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. 9 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. 10 11 12 */ 13 14 #include <linux/module.h> 15 16 #include <linux/slab.h> 17 #include <linux/drbd.h> 18 #include "drbd_int.h" 19 #include "drbd_req.h" 20 21 22 static bool drbd_may_do_local_read(struct drbd_device *device, sector_t sector, int size); 23 24 static struct drbd_request *drbd_req_new(struct drbd_device *device, struct bio *bio_src) 25 { 26 struct drbd_request *req; 27 28 req = mempool_alloc(&drbd_request_mempool, GFP_NOIO); 29 if (!req) 30 return NULL; 31 memset(req, 0, sizeof(*req)); 32 33 req->private_bio = bio_alloc_clone(device->ldev->backing_bdev, bio_src, 34 GFP_NOIO, &drbd_io_bio_set); 35 req->private_bio->bi_private = req; 36 req->private_bio->bi_end_io = drbd_request_endio; 37 38 req->rq_state = (bio_data_dir(bio_src) == WRITE ? RQ_WRITE : 0) 39 | (bio_op(bio_src) == REQ_OP_WRITE_ZEROES ? RQ_ZEROES : 0) 40 | (bio_op(bio_src) == REQ_OP_DISCARD ? RQ_UNMAP : 0); 41 req->device = device; 42 req->master_bio = bio_src; 43 req->epoch = 0; 44 45 drbd_clear_interval(&req->i); 46 req->i.sector = bio_src->bi_iter.bi_sector; 47 req->i.size = bio_src->bi_iter.bi_size; 48 req->i.local = true; 49 req->i.waiting = false; 50 51 INIT_LIST_HEAD(&req->tl_requests); 52 INIT_LIST_HEAD(&req->w.list); 53 INIT_LIST_HEAD(&req->req_pending_master_completion); 54 INIT_LIST_HEAD(&req->req_pending_local); 55 56 /* one reference to be put by __drbd_make_request */ 57 atomic_set(&req->completion_ref, 1); 58 /* one kref as long as completion_ref > 0 */ 59 kref_init(&req->kref); 60 return req; 61 } 62 63 static void drbd_remove_request_interval(struct rb_root *root, 64 struct drbd_request *req) 65 { 66 struct drbd_device *device = req->device; 67 struct drbd_interval *i = &req->i; 68 69 drbd_remove_interval(root, i); 70 71 /* Wake up any processes waiting for this request to complete. */ 72 if (i->waiting) 73 wake_up(&device->misc_wait); 74 } 75 76 void drbd_req_destroy(struct kref *kref) 77 { 78 struct drbd_request *req = container_of(kref, struct drbd_request, kref); 79 struct drbd_device *device = req->device; 80 const unsigned s = req->rq_state; 81 82 if ((req->master_bio && !(s & RQ_POSTPONED)) || 83 atomic_read(&req->completion_ref) || 84 (s & RQ_LOCAL_PENDING) || 85 ((s & RQ_NET_MASK) && !(s & RQ_NET_DONE))) { 86 drbd_err(device, "drbd_req_destroy: Logic BUG rq_state = 0x%x, completion_ref = %d\n", 87 s, atomic_read(&req->completion_ref)); 88 return; 89 } 90 91 /* If called from mod_rq_state (expected normal case) or 92 * drbd_send_and_submit (the less likely normal path), this holds the 93 * req_lock, and req->tl_requests will typicaly be on ->transfer_log, 94 * though it may be still empty (never added to the transfer log). 95 * 96 * If called from do_retry(), we do NOT hold the req_lock, but we are 97 * still allowed to unconditionally list_del(&req->tl_requests), 98 * because it will be on a local on-stack list only. */ 99 list_del_init(&req->tl_requests); 100 101 /* finally remove the request from the conflict detection 102 * respective block_id verification interval tree. */ 103 if (!drbd_interval_empty(&req->i)) { 104 struct rb_root *root; 105 106 if (s & RQ_WRITE) 107 root = &device->write_requests; 108 else 109 root = &device->read_requests; 110 drbd_remove_request_interval(root, req); 111 } else if (s & (RQ_NET_MASK & ~RQ_NET_DONE) && req->i.size != 0) 112 drbd_err(device, "drbd_req_destroy: Logic BUG: interval empty, but: rq_state=0x%x, sect=%llu, size=%u\n", 113 s, (unsigned long long)req->i.sector, req->i.size); 114 115 /* if it was a write, we may have to set the corresponding 116 * bit(s) out-of-sync first. If it had a local part, we need to 117 * release the reference to the activity log. */ 118 if (s & RQ_WRITE) { 119 /* Set out-of-sync unless both OK flags are set 120 * (local only or remote failed). 121 * Other places where we set out-of-sync: 122 * READ with local io-error */ 123 124 /* There is a special case: 125 * we may notice late that IO was suspended, 126 * and postpone, or schedule for retry, a write, 127 * before it even was submitted or sent. 128 * In that case we do not want to touch the bitmap at all. 129 */ 130 if ((s & (RQ_POSTPONED|RQ_LOCAL_MASK|RQ_NET_MASK)) != RQ_POSTPONED) { 131 if (!(s & RQ_NET_OK) || !(s & RQ_LOCAL_OK)) 132 drbd_set_out_of_sync(device, req->i.sector, req->i.size); 133 134 if ((s & RQ_NET_OK) && (s & RQ_LOCAL_OK) && (s & RQ_NET_SIS)) 135 drbd_set_in_sync(device, req->i.sector, req->i.size); 136 } 137 138 /* one might be tempted to move the drbd_al_complete_io 139 * to the local io completion callback drbd_request_endio. 140 * but, if this was a mirror write, we may only 141 * drbd_al_complete_io after this is RQ_NET_DONE, 142 * otherwise the extent could be dropped from the al 143 * before it has actually been written on the peer. 144 * if we crash before our peer knows about the request, 145 * but after the extent has been dropped from the al, 146 * we would forget to resync the corresponding extent. 147 */ 148 if (s & RQ_IN_ACT_LOG) { 149 if (get_ldev_if_state(device, D_FAILED)) { 150 drbd_al_complete_io(device, &req->i); 151 put_ldev(device); 152 } else if (__ratelimit(&drbd_ratelimit_state)) { 153 drbd_warn(device, "Should have called drbd_al_complete_io(, %llu, %u), " 154 "but my Disk seems to have failed :(\n", 155 (unsigned long long) req->i.sector, req->i.size); 156 } 157 } 158 } 159 160 mempool_free(req, &drbd_request_mempool); 161 } 162 163 static void wake_all_senders(struct drbd_connection *connection) 164 { 165 wake_up(&connection->sender_work.q_wait); 166 } 167 168 /* must hold resource->req_lock */ 169 void start_new_tl_epoch(struct drbd_connection *connection) 170 { 171 /* no point closing an epoch, if it is empty, anyways. */ 172 if (connection->current_tle_writes == 0) 173 return; 174 175 connection->current_tle_writes = 0; 176 atomic_inc(&connection->current_tle_nr); 177 wake_all_senders(connection); 178 } 179 180 void complete_master_bio(struct drbd_device *device, 181 struct bio_and_error *m) 182 { 183 if (unlikely(m->error)) 184 m->bio->bi_status = errno_to_blk_status(m->error); 185 bio_endio(m->bio); 186 dec_ap_bio(device); 187 } 188 189 190 /* Helper for __req_mod(). 191 * Set m->bio to the master bio, if it is fit to be completed, 192 * or leave it alone (it is initialized to NULL in __req_mod), 193 * if it has already been completed, or cannot be completed yet. 194 * If m->bio is set, the error status to be returned is placed in m->error. 195 */ 196 static 197 void drbd_req_complete(struct drbd_request *req, struct bio_and_error *m) 198 { 199 const unsigned s = req->rq_state; 200 struct drbd_device *device = req->device; 201 int error, ok; 202 203 /* we must not complete the master bio, while it is 204 * still being processed by _drbd_send_zc_bio (drbd_send_dblock) 205 * not yet acknowledged by the peer 206 * not yet completed by the local io subsystem 207 * these flags may get cleared in any order by 208 * the worker, 209 * the receiver, 210 * the bio_endio completion callbacks. 211 */ 212 if ((s & RQ_LOCAL_PENDING && !(s & RQ_LOCAL_ABORTED)) || 213 (s & RQ_NET_QUEUED) || (s & RQ_NET_PENDING) || 214 (s & RQ_COMPLETION_SUSP)) { 215 drbd_err(device, "drbd_req_complete: Logic BUG rq_state = 0x%x\n", s); 216 return; 217 } 218 219 if (!req->master_bio) { 220 drbd_err(device, "drbd_req_complete: Logic BUG, master_bio == NULL!\n"); 221 return; 222 } 223 224 /* 225 * figure out whether to report success or failure. 226 * 227 * report success when at least one of the operations succeeded. 228 * or, to put the other way, 229 * only report failure, when both operations failed. 230 * 231 * what to do about the failures is handled elsewhere. 232 * what we need to do here is just: complete the master_bio. 233 * 234 * local completion error, if any, has been stored as ERR_PTR 235 * in private_bio within drbd_request_endio. 236 */ 237 ok = (s & RQ_LOCAL_OK) || (s & RQ_NET_OK); 238 error = PTR_ERR(req->private_bio); 239 240 /* Before we can signal completion to the upper layers, 241 * we may need to close the current transfer log epoch. 242 * We are within the request lock, so we can simply compare 243 * the request epoch number with the current transfer log 244 * epoch number. If they match, increase the current_tle_nr, 245 * and reset the transfer log epoch write_cnt. 246 */ 247 if (op_is_write(bio_op(req->master_bio)) && 248 req->epoch == atomic_read(&first_peer_device(device)->connection->current_tle_nr)) 249 start_new_tl_epoch(first_peer_device(device)->connection); 250 251 /* Update disk stats */ 252 bio_end_io_acct(req->master_bio, req->start_jif); 253 254 /* If READ failed, 255 * have it be pushed back to the retry work queue, 256 * so it will re-enter __drbd_make_request(), 257 * and be re-assigned to a suitable local or remote path, 258 * or failed if we do not have access to good data anymore. 259 * 260 * Unless it was failed early by __drbd_make_request(), 261 * because no path was available, in which case 262 * it was not even added to the transfer_log. 263 * 264 * read-ahead may fail, and will not be retried. 265 * 266 * WRITE should have used all available paths already. 267 */ 268 if (!ok && 269 bio_op(req->master_bio) == REQ_OP_READ && 270 !(req->master_bio->bi_opf & REQ_RAHEAD) && 271 !list_empty(&req->tl_requests)) 272 req->rq_state |= RQ_POSTPONED; 273 274 if (!(req->rq_state & RQ_POSTPONED)) { 275 m->error = ok ? 0 : (error ?: -EIO); 276 m->bio = req->master_bio; 277 req->master_bio = NULL; 278 /* We leave it in the tree, to be able to verify later 279 * write-acks in protocol != C during resync. 280 * But we mark it as "complete", so it won't be counted as 281 * conflict in a multi-primary setup. */ 282 req->i.completed = true; 283 } 284 285 if (req->i.waiting) 286 wake_up(&device->misc_wait); 287 288 /* Either we are about to complete to upper layers, 289 * or we will restart this request. 290 * In either case, the request object will be destroyed soon, 291 * so better remove it from all lists. */ 292 list_del_init(&req->req_pending_master_completion); 293 } 294 295 /* still holds resource->req_lock */ 296 static void drbd_req_put_completion_ref(struct drbd_request *req, struct bio_and_error *m, int put) 297 { 298 struct drbd_device *device = req->device; 299 D_ASSERT(device, m || (req->rq_state & RQ_POSTPONED)); 300 301 if (!put) 302 return; 303 304 if (!atomic_sub_and_test(put, &req->completion_ref)) 305 return; 306 307 drbd_req_complete(req, m); 308 309 /* local completion may still come in later, 310 * we need to keep the req object around. */ 311 if (req->rq_state & RQ_LOCAL_ABORTED) 312 return; 313 314 if (req->rq_state & RQ_POSTPONED) { 315 /* don't destroy the req object just yet, 316 * but queue it for retry */ 317 drbd_restart_request(req); 318 return; 319 } 320 321 kref_put(&req->kref, drbd_req_destroy); 322 } 323 324 static void set_if_null_req_next(struct drbd_peer_device *peer_device, struct drbd_request *req) 325 { 326 struct drbd_connection *connection = peer_device ? peer_device->connection : NULL; 327 if (!connection) 328 return; 329 if (connection->req_next == NULL) 330 connection->req_next = req; 331 } 332 333 static void advance_conn_req_next(struct drbd_peer_device *peer_device, struct drbd_request *req) 334 { 335 struct drbd_connection *connection = peer_device ? peer_device->connection : NULL; 336 struct drbd_request *iter = req; 337 if (!connection) 338 return; 339 if (connection->req_next != req) 340 return; 341 342 req = NULL; 343 list_for_each_entry_continue(iter, &connection->transfer_log, tl_requests) { 344 const unsigned int s = iter->rq_state; 345 346 if (s & RQ_NET_QUEUED) { 347 req = iter; 348 break; 349 } 350 } 351 connection->req_next = req; 352 } 353 354 static void set_if_null_req_ack_pending(struct drbd_peer_device *peer_device, struct drbd_request *req) 355 { 356 struct drbd_connection *connection = peer_device ? peer_device->connection : NULL; 357 if (!connection) 358 return; 359 if (connection->req_ack_pending == NULL) 360 connection->req_ack_pending = req; 361 } 362 363 static void advance_conn_req_ack_pending(struct drbd_peer_device *peer_device, struct drbd_request *req) 364 { 365 struct drbd_connection *connection = peer_device ? peer_device->connection : NULL; 366 struct drbd_request *iter = req; 367 if (!connection) 368 return; 369 if (connection->req_ack_pending != req) 370 return; 371 372 req = NULL; 373 list_for_each_entry_continue(iter, &connection->transfer_log, tl_requests) { 374 const unsigned int s = iter->rq_state; 375 376 if ((s & RQ_NET_SENT) && (s & RQ_NET_PENDING)) { 377 req = iter; 378 break; 379 } 380 } 381 connection->req_ack_pending = req; 382 } 383 384 static void set_if_null_req_not_net_done(struct drbd_peer_device *peer_device, struct drbd_request *req) 385 { 386 struct drbd_connection *connection = peer_device ? peer_device->connection : NULL; 387 if (!connection) 388 return; 389 if (connection->req_not_net_done == NULL) 390 connection->req_not_net_done = req; 391 } 392 393 static void advance_conn_req_not_net_done(struct drbd_peer_device *peer_device, struct drbd_request *req) 394 { 395 struct drbd_connection *connection = peer_device ? peer_device->connection : NULL; 396 struct drbd_request *iter = req; 397 if (!connection) 398 return; 399 if (connection->req_not_net_done != req) 400 return; 401 402 req = NULL; 403 list_for_each_entry_continue(iter, &connection->transfer_log, tl_requests) { 404 const unsigned int s = iter->rq_state; 405 406 if ((s & RQ_NET_SENT) && !(s & RQ_NET_DONE)) { 407 req = iter; 408 break; 409 } 410 } 411 connection->req_not_net_done = req; 412 } 413 414 /* I'd like this to be the only place that manipulates 415 * req->completion_ref and req->kref. */ 416 static void mod_rq_state(struct drbd_request *req, struct bio_and_error *m, 417 int clear, int set) 418 { 419 struct drbd_device *device = req->device; 420 struct drbd_peer_device *peer_device = first_peer_device(device); 421 unsigned s = req->rq_state; 422 int c_put = 0; 423 424 if (drbd_suspended(device) && !((s | clear) & RQ_COMPLETION_SUSP)) 425 set |= RQ_COMPLETION_SUSP; 426 427 /* apply */ 428 429 req->rq_state &= ~clear; 430 req->rq_state |= set; 431 432 /* no change? */ 433 if (req->rq_state == s) 434 return; 435 436 /* intent: get references */ 437 438 kref_get(&req->kref); 439 440 if (!(s & RQ_LOCAL_PENDING) && (set & RQ_LOCAL_PENDING)) 441 atomic_inc(&req->completion_ref); 442 443 if (!(s & RQ_NET_PENDING) && (set & RQ_NET_PENDING)) { 444 inc_ap_pending(device); 445 atomic_inc(&req->completion_ref); 446 } 447 448 if (!(s & RQ_NET_QUEUED) && (set & RQ_NET_QUEUED)) { 449 atomic_inc(&req->completion_ref); 450 set_if_null_req_next(peer_device, req); 451 } 452 453 if (!(s & RQ_EXP_BARR_ACK) && (set & RQ_EXP_BARR_ACK)) 454 kref_get(&req->kref); /* wait for the DONE */ 455 456 if (!(s & RQ_NET_SENT) && (set & RQ_NET_SENT)) { 457 /* potentially already completed in the ack_receiver thread */ 458 if (!(s & RQ_NET_DONE)) { 459 atomic_add(req->i.size >> 9, &device->ap_in_flight); 460 set_if_null_req_not_net_done(peer_device, req); 461 } 462 if (req->rq_state & RQ_NET_PENDING) 463 set_if_null_req_ack_pending(peer_device, req); 464 } 465 466 if (!(s & RQ_COMPLETION_SUSP) && (set & RQ_COMPLETION_SUSP)) 467 atomic_inc(&req->completion_ref); 468 469 /* progress: put references */ 470 471 if ((s & RQ_COMPLETION_SUSP) && (clear & RQ_COMPLETION_SUSP)) 472 ++c_put; 473 474 if (!(s & RQ_LOCAL_ABORTED) && (set & RQ_LOCAL_ABORTED)) { 475 D_ASSERT(device, req->rq_state & RQ_LOCAL_PENDING); 476 ++c_put; 477 } 478 479 if ((s & RQ_LOCAL_PENDING) && (clear & RQ_LOCAL_PENDING)) { 480 if (req->rq_state & RQ_LOCAL_ABORTED) 481 kref_put(&req->kref, drbd_req_destroy); 482 else 483 ++c_put; 484 list_del_init(&req->req_pending_local); 485 } 486 487 if ((s & RQ_NET_PENDING) && (clear & RQ_NET_PENDING)) { 488 dec_ap_pending(device); 489 ++c_put; 490 req->acked_jif = jiffies; 491 advance_conn_req_ack_pending(peer_device, req); 492 } 493 494 if ((s & RQ_NET_QUEUED) && (clear & RQ_NET_QUEUED)) { 495 ++c_put; 496 advance_conn_req_next(peer_device, req); 497 } 498 499 if (!(s & RQ_NET_DONE) && (set & RQ_NET_DONE)) { 500 if (s & RQ_NET_SENT) 501 atomic_sub(req->i.size >> 9, &device->ap_in_flight); 502 if (s & RQ_EXP_BARR_ACK) 503 kref_put(&req->kref, drbd_req_destroy); 504 req->net_done_jif = jiffies; 505 506 /* in ahead/behind mode, or just in case, 507 * before we finally destroy this request, 508 * the caching pointers must not reference it anymore */ 509 advance_conn_req_next(peer_device, req); 510 advance_conn_req_ack_pending(peer_device, req); 511 advance_conn_req_not_net_done(peer_device, req); 512 } 513 514 /* potentially complete and destroy */ 515 516 /* If we made progress, retry conflicting peer requests, if any. */ 517 if (req->i.waiting) 518 wake_up(&device->misc_wait); 519 520 drbd_req_put_completion_ref(req, m, c_put); 521 kref_put(&req->kref, drbd_req_destroy); 522 } 523 524 static void drbd_report_io_error(struct drbd_device *device, struct drbd_request *req) 525 { 526 char b[BDEVNAME_SIZE]; 527 528 if (!__ratelimit(&drbd_ratelimit_state)) 529 return; 530 531 drbd_warn(device, "local %s IO error sector %llu+%u on %s\n", 532 (req->rq_state & RQ_WRITE) ? "WRITE" : "READ", 533 (unsigned long long)req->i.sector, 534 req->i.size >> 9, 535 bdevname(device->ldev->backing_bdev, b)); 536 } 537 538 /* Helper for HANDED_OVER_TO_NETWORK. 539 * Is this a protocol A write (neither WRITE_ACK nor RECEIVE_ACK expected)? 540 * Is it also still "PENDING"? 541 * --> If so, clear PENDING and set NET_OK below. 542 * If it is a protocol A write, but not RQ_PENDING anymore, neg-ack was faster 543 * (and we must not set RQ_NET_OK) */ 544 static inline bool is_pending_write_protocol_A(struct drbd_request *req) 545 { 546 return (req->rq_state & 547 (RQ_WRITE|RQ_NET_PENDING|RQ_EXP_WRITE_ACK|RQ_EXP_RECEIVE_ACK)) 548 == (RQ_WRITE|RQ_NET_PENDING); 549 } 550 551 /* obviously this could be coded as many single functions 552 * instead of one huge switch, 553 * or by putting the code directly in the respective locations 554 * (as it has been before). 555 * 556 * but having it this way 557 * enforces that it is all in this one place, where it is easier to audit, 558 * it makes it obvious that whatever "event" "happens" to a request should 559 * happen "atomically" within the req_lock, 560 * and it enforces that we have to think in a very structured manner 561 * about the "events" that may happen to a request during its life time ... 562 */ 563 int __req_mod(struct drbd_request *req, enum drbd_req_event what, 564 struct bio_and_error *m) 565 { 566 struct drbd_device *const device = req->device; 567 struct drbd_peer_device *const peer_device = first_peer_device(device); 568 struct drbd_connection *const connection = peer_device ? peer_device->connection : NULL; 569 struct net_conf *nc; 570 int p, rv = 0; 571 572 if (m) 573 m->bio = NULL; 574 575 switch (what) { 576 default: 577 drbd_err(device, "LOGIC BUG in %s:%u\n", __FILE__ , __LINE__); 578 break; 579 580 /* does not happen... 581 * initialization done in drbd_req_new 582 case CREATED: 583 break; 584 */ 585 586 case TO_BE_SENT: /* via network */ 587 /* reached via __drbd_make_request 588 * and from w_read_retry_remote */ 589 D_ASSERT(device, !(req->rq_state & RQ_NET_MASK)); 590 rcu_read_lock(); 591 nc = rcu_dereference(connection->net_conf); 592 p = nc->wire_protocol; 593 rcu_read_unlock(); 594 req->rq_state |= 595 p == DRBD_PROT_C ? RQ_EXP_WRITE_ACK : 596 p == DRBD_PROT_B ? RQ_EXP_RECEIVE_ACK : 0; 597 mod_rq_state(req, m, 0, RQ_NET_PENDING); 598 break; 599 600 case TO_BE_SUBMITTED: /* locally */ 601 /* reached via __drbd_make_request */ 602 D_ASSERT(device, !(req->rq_state & RQ_LOCAL_MASK)); 603 mod_rq_state(req, m, 0, RQ_LOCAL_PENDING); 604 break; 605 606 case COMPLETED_OK: 607 if (req->rq_state & RQ_WRITE) 608 device->writ_cnt += req->i.size >> 9; 609 else 610 device->read_cnt += req->i.size >> 9; 611 612 mod_rq_state(req, m, RQ_LOCAL_PENDING, 613 RQ_LOCAL_COMPLETED|RQ_LOCAL_OK); 614 break; 615 616 case ABORT_DISK_IO: 617 mod_rq_state(req, m, 0, RQ_LOCAL_ABORTED); 618 break; 619 620 case WRITE_COMPLETED_WITH_ERROR: 621 drbd_report_io_error(device, req); 622 __drbd_chk_io_error(device, DRBD_WRITE_ERROR); 623 mod_rq_state(req, m, RQ_LOCAL_PENDING, RQ_LOCAL_COMPLETED); 624 break; 625 626 case READ_COMPLETED_WITH_ERROR: 627 drbd_set_out_of_sync(device, req->i.sector, req->i.size); 628 drbd_report_io_error(device, req); 629 __drbd_chk_io_error(device, DRBD_READ_ERROR); 630 fallthrough; 631 case READ_AHEAD_COMPLETED_WITH_ERROR: 632 /* it is legal to fail read-ahead, no __drbd_chk_io_error in that case. */ 633 mod_rq_state(req, m, RQ_LOCAL_PENDING, RQ_LOCAL_COMPLETED); 634 break; 635 636 case DISCARD_COMPLETED_NOTSUPP: 637 case DISCARD_COMPLETED_WITH_ERROR: 638 /* I'd rather not detach from local disk just because it 639 * failed a REQ_OP_DISCARD. */ 640 mod_rq_state(req, m, RQ_LOCAL_PENDING, RQ_LOCAL_COMPLETED); 641 break; 642 643 case QUEUE_FOR_NET_READ: 644 /* READ, and 645 * no local disk, 646 * or target area marked as invalid, 647 * or just got an io-error. */ 648 /* from __drbd_make_request 649 * or from bio_endio during read io-error recovery */ 650 651 /* So we can verify the handle in the answer packet. 652 * Corresponding drbd_remove_request_interval is in 653 * drbd_req_complete() */ 654 D_ASSERT(device, drbd_interval_empty(&req->i)); 655 drbd_insert_interval(&device->read_requests, &req->i); 656 657 set_bit(UNPLUG_REMOTE, &device->flags); 658 659 D_ASSERT(device, req->rq_state & RQ_NET_PENDING); 660 D_ASSERT(device, (req->rq_state & RQ_LOCAL_MASK) == 0); 661 mod_rq_state(req, m, 0, RQ_NET_QUEUED); 662 req->w.cb = w_send_read_req; 663 drbd_queue_work(&connection->sender_work, 664 &req->w); 665 break; 666 667 case QUEUE_FOR_NET_WRITE: 668 /* assert something? */ 669 /* from __drbd_make_request only */ 670 671 /* Corresponding drbd_remove_request_interval is in 672 * drbd_req_complete() */ 673 D_ASSERT(device, drbd_interval_empty(&req->i)); 674 drbd_insert_interval(&device->write_requests, &req->i); 675 676 /* NOTE 677 * In case the req ended up on the transfer log before being 678 * queued on the worker, it could lead to this request being 679 * missed during cleanup after connection loss. 680 * So we have to do both operations here, 681 * within the same lock that protects the transfer log. 682 * 683 * _req_add_to_epoch(req); this has to be after the 684 * _maybe_start_new_epoch(req); which happened in 685 * __drbd_make_request, because we now may set the bit 686 * again ourselves to close the current epoch. 687 * 688 * Add req to the (now) current epoch (barrier). */ 689 690 /* otherwise we may lose an unplug, which may cause some remote 691 * io-scheduler timeout to expire, increasing maximum latency, 692 * hurting performance. */ 693 set_bit(UNPLUG_REMOTE, &device->flags); 694 695 /* queue work item to send data */ 696 D_ASSERT(device, req->rq_state & RQ_NET_PENDING); 697 mod_rq_state(req, m, 0, RQ_NET_QUEUED|RQ_EXP_BARR_ACK); 698 req->w.cb = w_send_dblock; 699 drbd_queue_work(&connection->sender_work, 700 &req->w); 701 702 /* close the epoch, in case it outgrew the limit */ 703 rcu_read_lock(); 704 nc = rcu_dereference(connection->net_conf); 705 p = nc->max_epoch_size; 706 rcu_read_unlock(); 707 if (connection->current_tle_writes >= p) 708 start_new_tl_epoch(connection); 709 710 break; 711 712 case QUEUE_FOR_SEND_OOS: 713 mod_rq_state(req, m, 0, RQ_NET_QUEUED); 714 req->w.cb = w_send_out_of_sync; 715 drbd_queue_work(&connection->sender_work, 716 &req->w); 717 break; 718 719 case READ_RETRY_REMOTE_CANCELED: 720 case SEND_CANCELED: 721 case SEND_FAILED: 722 /* real cleanup will be done from tl_clear. just update flags 723 * so it is no longer marked as on the worker queue */ 724 mod_rq_state(req, m, RQ_NET_QUEUED, 0); 725 break; 726 727 case HANDED_OVER_TO_NETWORK: 728 /* assert something? */ 729 if (is_pending_write_protocol_A(req)) 730 /* this is what is dangerous about protocol A: 731 * pretend it was successfully written on the peer. */ 732 mod_rq_state(req, m, RQ_NET_QUEUED|RQ_NET_PENDING, 733 RQ_NET_SENT|RQ_NET_OK); 734 else 735 mod_rq_state(req, m, RQ_NET_QUEUED, RQ_NET_SENT); 736 /* It is still not yet RQ_NET_DONE until the 737 * corresponding epoch barrier got acked as well, 738 * so we know what to dirty on connection loss. */ 739 break; 740 741 case OOS_HANDED_TO_NETWORK: 742 /* Was not set PENDING, no longer QUEUED, so is now DONE 743 * as far as this connection is concerned. */ 744 mod_rq_state(req, m, RQ_NET_QUEUED, RQ_NET_DONE); 745 break; 746 747 case CONNECTION_LOST_WHILE_PENDING: 748 /* transfer log cleanup after connection loss */ 749 mod_rq_state(req, m, 750 RQ_NET_OK|RQ_NET_PENDING|RQ_COMPLETION_SUSP, 751 RQ_NET_DONE); 752 break; 753 754 case CONFLICT_RESOLVED: 755 /* for superseded conflicting writes of multiple primaries, 756 * there is no need to keep anything in the tl, potential 757 * node crashes are covered by the activity log. 758 * 759 * If this request had been marked as RQ_POSTPONED before, 760 * it will actually not be completed, but "restarted", 761 * resubmitted from the retry worker context. */ 762 D_ASSERT(device, req->rq_state & RQ_NET_PENDING); 763 D_ASSERT(device, req->rq_state & RQ_EXP_WRITE_ACK); 764 mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_DONE|RQ_NET_OK); 765 break; 766 767 case WRITE_ACKED_BY_PEER_AND_SIS: 768 req->rq_state |= RQ_NET_SIS; 769 fallthrough; 770 case WRITE_ACKED_BY_PEER: 771 /* Normal operation protocol C: successfully written on peer. 772 * During resync, even in protocol != C, 773 * we requested an explicit write ack anyways. 774 * Which means we cannot even assert anything here. 775 * Nothing more to do here. 776 * We want to keep the tl in place for all protocols, to cater 777 * for volatile write-back caches on lower level devices. */ 778 goto ack_common; 779 case RECV_ACKED_BY_PEER: 780 D_ASSERT(device, req->rq_state & RQ_EXP_RECEIVE_ACK); 781 /* protocol B; pretends to be successfully written on peer. 782 * see also notes above in HANDED_OVER_TO_NETWORK about 783 * protocol != C */ 784 ack_common: 785 mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_OK); 786 break; 787 788 case POSTPONE_WRITE: 789 D_ASSERT(device, req->rq_state & RQ_EXP_WRITE_ACK); 790 /* If this node has already detected the write conflict, the 791 * worker will be waiting on misc_wait. Wake it up once this 792 * request has completed locally. 793 */ 794 D_ASSERT(device, req->rq_state & RQ_NET_PENDING); 795 req->rq_state |= RQ_POSTPONED; 796 if (req->i.waiting) 797 wake_up(&device->misc_wait); 798 /* Do not clear RQ_NET_PENDING. This request will make further 799 * progress via restart_conflicting_writes() or 800 * fail_postponed_requests(). Hopefully. */ 801 break; 802 803 case NEG_ACKED: 804 mod_rq_state(req, m, RQ_NET_OK|RQ_NET_PENDING, 0); 805 break; 806 807 case FAIL_FROZEN_DISK_IO: 808 if (!(req->rq_state & RQ_LOCAL_COMPLETED)) 809 break; 810 mod_rq_state(req, m, RQ_COMPLETION_SUSP, 0); 811 break; 812 813 case RESTART_FROZEN_DISK_IO: 814 if (!(req->rq_state & RQ_LOCAL_COMPLETED)) 815 break; 816 817 mod_rq_state(req, m, 818 RQ_COMPLETION_SUSP|RQ_LOCAL_COMPLETED, 819 RQ_LOCAL_PENDING); 820 821 rv = MR_READ; 822 if (bio_data_dir(req->master_bio) == WRITE) 823 rv = MR_WRITE; 824 825 get_ldev(device); /* always succeeds in this call path */ 826 req->w.cb = w_restart_disk_io; 827 drbd_queue_work(&connection->sender_work, 828 &req->w); 829 break; 830 831 case RESEND: 832 /* Simply complete (local only) READs. */ 833 if (!(req->rq_state & RQ_WRITE) && !req->w.cb) { 834 mod_rq_state(req, m, RQ_COMPLETION_SUSP, 0); 835 break; 836 } 837 838 /* If RQ_NET_OK is already set, we got a P_WRITE_ACK or P_RECV_ACK 839 before the connection loss (B&C only); only P_BARRIER_ACK 840 (or the local completion?) was missing when we suspended. 841 Throwing them out of the TL here by pretending we got a BARRIER_ACK. 842 During connection handshake, we ensure that the peer was not rebooted. */ 843 if (!(req->rq_state & RQ_NET_OK)) { 844 /* FIXME could this possibly be a req->dw.cb == w_send_out_of_sync? 845 * in that case we must not set RQ_NET_PENDING. */ 846 847 mod_rq_state(req, m, RQ_COMPLETION_SUSP, RQ_NET_QUEUED|RQ_NET_PENDING); 848 if (req->w.cb) { 849 /* w.cb expected to be w_send_dblock, or w_send_read_req */ 850 drbd_queue_work(&connection->sender_work, 851 &req->w); 852 rv = req->rq_state & RQ_WRITE ? MR_WRITE : MR_READ; 853 } /* else: FIXME can this happen? */ 854 break; 855 } 856 fallthrough; /* to BARRIER_ACKED */ 857 858 case BARRIER_ACKED: 859 /* barrier ack for READ requests does not make sense */ 860 if (!(req->rq_state & RQ_WRITE)) 861 break; 862 863 if (req->rq_state & RQ_NET_PENDING) { 864 /* barrier came in before all requests were acked. 865 * this is bad, because if the connection is lost now, 866 * we won't be able to clean them up... */ 867 drbd_err(device, "FIXME (BARRIER_ACKED but pending)\n"); 868 } 869 /* Allowed to complete requests, even while suspended. 870 * As this is called for all requests within a matching epoch, 871 * we need to filter, and only set RQ_NET_DONE for those that 872 * have actually been on the wire. */ 873 mod_rq_state(req, m, RQ_COMPLETION_SUSP, 874 (req->rq_state & RQ_NET_MASK) ? RQ_NET_DONE : 0); 875 break; 876 877 case DATA_RECEIVED: 878 D_ASSERT(device, req->rq_state & RQ_NET_PENDING); 879 mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_OK|RQ_NET_DONE); 880 break; 881 882 case QUEUE_AS_DRBD_BARRIER: 883 start_new_tl_epoch(connection); 884 mod_rq_state(req, m, 0, RQ_NET_OK|RQ_NET_DONE); 885 break; 886 } 887 888 return rv; 889 } 890 891 /* we may do a local read if: 892 * - we are consistent (of course), 893 * - or we are generally inconsistent, 894 * BUT we are still/already IN SYNC for this area. 895 * since size may be bigger than BM_BLOCK_SIZE, 896 * we may need to check several bits. 897 */ 898 static bool drbd_may_do_local_read(struct drbd_device *device, sector_t sector, int size) 899 { 900 unsigned long sbnr, ebnr; 901 sector_t esector, nr_sectors; 902 903 if (device->state.disk == D_UP_TO_DATE) 904 return true; 905 if (device->state.disk != D_INCONSISTENT) 906 return false; 907 esector = sector + (size >> 9) - 1; 908 nr_sectors = get_capacity(device->vdisk); 909 D_ASSERT(device, sector < nr_sectors); 910 D_ASSERT(device, esector < nr_sectors); 911 912 sbnr = BM_SECT_TO_BIT(sector); 913 ebnr = BM_SECT_TO_BIT(esector); 914 915 return drbd_bm_count_bits(device, sbnr, ebnr) == 0; 916 } 917 918 static bool remote_due_to_read_balancing(struct drbd_device *device, sector_t sector, 919 enum drbd_read_balancing rbm) 920 { 921 int stripe_shift; 922 923 switch (rbm) { 924 case RB_CONGESTED_REMOTE: 925 return false; 926 case RB_LEAST_PENDING: 927 return atomic_read(&device->local_cnt) > 928 atomic_read(&device->ap_pending_cnt) + atomic_read(&device->rs_pending_cnt); 929 case RB_32K_STRIPING: /* stripe_shift = 15 */ 930 case RB_64K_STRIPING: 931 case RB_128K_STRIPING: 932 case RB_256K_STRIPING: 933 case RB_512K_STRIPING: 934 case RB_1M_STRIPING: /* stripe_shift = 20 */ 935 stripe_shift = (rbm - RB_32K_STRIPING + 15); 936 return (sector >> (stripe_shift - 9)) & 1; 937 case RB_ROUND_ROBIN: 938 return test_and_change_bit(READ_BALANCE_RR, &device->flags); 939 case RB_PREFER_REMOTE: 940 return true; 941 case RB_PREFER_LOCAL: 942 default: 943 return false; 944 } 945 } 946 947 /* 948 * complete_conflicting_writes - wait for any conflicting write requests 949 * 950 * The write_requests tree contains all active write requests which we 951 * currently know about. Wait for any requests to complete which conflict with 952 * the new one. 953 * 954 * Only way out: remove the conflicting intervals from the tree. 955 */ 956 static void complete_conflicting_writes(struct drbd_request *req) 957 { 958 DEFINE_WAIT(wait); 959 struct drbd_device *device = req->device; 960 struct drbd_interval *i; 961 sector_t sector = req->i.sector; 962 int size = req->i.size; 963 964 for (;;) { 965 drbd_for_each_overlap(i, &device->write_requests, sector, size) { 966 /* Ignore, if already completed to upper layers. */ 967 if (i->completed) 968 continue; 969 /* Handle the first found overlap. After the schedule 970 * we have to restart the tree walk. */ 971 break; 972 } 973 if (!i) /* if any */ 974 break; 975 976 /* Indicate to wake up device->misc_wait on progress. */ 977 prepare_to_wait(&device->misc_wait, &wait, TASK_UNINTERRUPTIBLE); 978 i->waiting = true; 979 spin_unlock_irq(&device->resource->req_lock); 980 schedule(); 981 spin_lock_irq(&device->resource->req_lock); 982 } 983 finish_wait(&device->misc_wait, &wait); 984 } 985 986 /* called within req_lock */ 987 static void maybe_pull_ahead(struct drbd_device *device) 988 { 989 struct drbd_connection *connection = first_peer_device(device)->connection; 990 struct net_conf *nc; 991 bool congested = false; 992 enum drbd_on_congestion on_congestion; 993 994 rcu_read_lock(); 995 nc = rcu_dereference(connection->net_conf); 996 on_congestion = nc ? nc->on_congestion : OC_BLOCK; 997 rcu_read_unlock(); 998 if (on_congestion == OC_BLOCK || 999 connection->agreed_pro_version < 96) 1000 return; 1001 1002 if (on_congestion == OC_PULL_AHEAD && device->state.conn == C_AHEAD) 1003 return; /* nothing to do ... */ 1004 1005 /* If I don't even have good local storage, we can not reasonably try 1006 * to pull ahead of the peer. We also need the local reference to make 1007 * sure device->act_log is there. 1008 */ 1009 if (!get_ldev_if_state(device, D_UP_TO_DATE)) 1010 return; 1011 1012 if (nc->cong_fill && 1013 atomic_read(&device->ap_in_flight) >= nc->cong_fill) { 1014 drbd_info(device, "Congestion-fill threshold reached\n"); 1015 congested = true; 1016 } 1017 1018 if (device->act_log->used >= nc->cong_extents) { 1019 drbd_info(device, "Congestion-extents threshold reached\n"); 1020 congested = true; 1021 } 1022 1023 if (congested) { 1024 /* start a new epoch for non-mirrored writes */ 1025 start_new_tl_epoch(first_peer_device(device)->connection); 1026 1027 if (on_congestion == OC_PULL_AHEAD) 1028 _drbd_set_state(_NS(device, conn, C_AHEAD), 0, NULL); 1029 else /*nc->on_congestion == OC_DISCONNECT */ 1030 _drbd_set_state(_NS(device, conn, C_DISCONNECTING), 0, NULL); 1031 } 1032 put_ldev(device); 1033 } 1034 1035 /* If this returns false, and req->private_bio is still set, 1036 * this should be submitted locally. 1037 * 1038 * If it returns false, but req->private_bio is not set, 1039 * we do not have access to good data :( 1040 * 1041 * Otherwise, this destroys req->private_bio, if any, 1042 * and returns true. 1043 */ 1044 static bool do_remote_read(struct drbd_request *req) 1045 { 1046 struct drbd_device *device = req->device; 1047 enum drbd_read_balancing rbm; 1048 1049 if (req->private_bio) { 1050 if (!drbd_may_do_local_read(device, 1051 req->i.sector, req->i.size)) { 1052 bio_put(req->private_bio); 1053 req->private_bio = NULL; 1054 put_ldev(device); 1055 } 1056 } 1057 1058 if (device->state.pdsk != D_UP_TO_DATE) 1059 return false; 1060 1061 if (req->private_bio == NULL) 1062 return true; 1063 1064 /* TODO: improve read balancing decisions, take into account drbd 1065 * protocol, pending requests etc. */ 1066 1067 rcu_read_lock(); 1068 rbm = rcu_dereference(device->ldev->disk_conf)->read_balancing; 1069 rcu_read_unlock(); 1070 1071 if (rbm == RB_PREFER_LOCAL && req->private_bio) 1072 return false; /* submit locally */ 1073 1074 if (remote_due_to_read_balancing(device, req->i.sector, rbm)) { 1075 if (req->private_bio) { 1076 bio_put(req->private_bio); 1077 req->private_bio = NULL; 1078 put_ldev(device); 1079 } 1080 return true; 1081 } 1082 1083 return false; 1084 } 1085 1086 bool drbd_should_do_remote(union drbd_dev_state s) 1087 { 1088 return s.pdsk == D_UP_TO_DATE || 1089 (s.pdsk >= D_INCONSISTENT && 1090 s.conn >= C_WF_BITMAP_T && 1091 s.conn < C_AHEAD); 1092 /* Before proto 96 that was >= CONNECTED instead of >= C_WF_BITMAP_T. 1093 That is equivalent since before 96 IO was frozen in the C_WF_BITMAP* 1094 states. */ 1095 } 1096 1097 static bool drbd_should_send_out_of_sync(union drbd_dev_state s) 1098 { 1099 return s.conn == C_AHEAD || s.conn == C_WF_BITMAP_S; 1100 /* pdsk = D_INCONSISTENT as a consequence. Protocol 96 check not necessary 1101 since we enter state C_AHEAD only if proto >= 96 */ 1102 } 1103 1104 /* returns number of connections (== 1, for drbd 8.4) 1105 * expected to actually write this data, 1106 * which does NOT include those that we are L_AHEAD for. */ 1107 static int drbd_process_write_request(struct drbd_request *req) 1108 { 1109 struct drbd_device *device = req->device; 1110 int remote, send_oos; 1111 1112 remote = drbd_should_do_remote(device->state); 1113 send_oos = drbd_should_send_out_of_sync(device->state); 1114 1115 /* Need to replicate writes. Unless it is an empty flush, 1116 * which is better mapped to a DRBD P_BARRIER packet, 1117 * also for drbd wire protocol compatibility reasons. 1118 * If this was a flush, just start a new epoch. 1119 * Unless the current epoch was empty anyways, or we are not currently 1120 * replicating, in which case there is no point. */ 1121 if (unlikely(req->i.size == 0)) { 1122 /* The only size==0 bios we expect are empty flushes. */ 1123 D_ASSERT(device, req->master_bio->bi_opf & REQ_PREFLUSH); 1124 if (remote) 1125 _req_mod(req, QUEUE_AS_DRBD_BARRIER); 1126 return remote; 1127 } 1128 1129 if (!remote && !send_oos) 1130 return 0; 1131 1132 D_ASSERT(device, !(remote && send_oos)); 1133 1134 if (remote) { 1135 _req_mod(req, TO_BE_SENT); 1136 _req_mod(req, QUEUE_FOR_NET_WRITE); 1137 } else if (drbd_set_out_of_sync(device, req->i.sector, req->i.size)) 1138 _req_mod(req, QUEUE_FOR_SEND_OOS); 1139 1140 return remote; 1141 } 1142 1143 static void drbd_process_discard_or_zeroes_req(struct drbd_request *req, int flags) 1144 { 1145 int err = drbd_issue_discard_or_zero_out(req->device, 1146 req->i.sector, req->i.size >> 9, flags); 1147 if (err) 1148 req->private_bio->bi_status = BLK_STS_IOERR; 1149 bio_endio(req->private_bio); 1150 } 1151 1152 static void 1153 drbd_submit_req_private_bio(struct drbd_request *req) 1154 { 1155 struct drbd_device *device = req->device; 1156 struct bio *bio = req->private_bio; 1157 unsigned int type; 1158 1159 if (bio_op(bio) != REQ_OP_READ) 1160 type = DRBD_FAULT_DT_WR; 1161 else if (bio->bi_opf & REQ_RAHEAD) 1162 type = DRBD_FAULT_DT_RA; 1163 else 1164 type = DRBD_FAULT_DT_RD; 1165 1166 /* State may have changed since we grabbed our reference on the 1167 * ->ldev member. Double check, and short-circuit to endio. 1168 * In case the last activity log transaction failed to get on 1169 * stable storage, and this is a WRITE, we may not even submit 1170 * this bio. */ 1171 if (get_ldev(device)) { 1172 if (drbd_insert_fault(device, type)) 1173 bio_io_error(bio); 1174 else if (bio_op(bio) == REQ_OP_WRITE_ZEROES) 1175 drbd_process_discard_or_zeroes_req(req, EE_ZEROOUT | 1176 ((bio->bi_opf & REQ_NOUNMAP) ? 0 : EE_TRIM)); 1177 else if (bio_op(bio) == REQ_OP_DISCARD) 1178 drbd_process_discard_or_zeroes_req(req, EE_TRIM); 1179 else 1180 submit_bio_noacct(bio); 1181 put_ldev(device); 1182 } else 1183 bio_io_error(bio); 1184 } 1185 1186 static void drbd_queue_write(struct drbd_device *device, struct drbd_request *req) 1187 { 1188 spin_lock_irq(&device->resource->req_lock); 1189 list_add_tail(&req->tl_requests, &device->submit.writes); 1190 list_add_tail(&req->req_pending_master_completion, 1191 &device->pending_master_completion[1 /* WRITE */]); 1192 spin_unlock_irq(&device->resource->req_lock); 1193 queue_work(device->submit.wq, &device->submit.worker); 1194 /* do_submit() may sleep internally on al_wait, too */ 1195 wake_up(&device->al_wait); 1196 } 1197 1198 /* returns the new drbd_request pointer, if the caller is expected to 1199 * drbd_send_and_submit() it (to save latency), or NULL if we queued the 1200 * request on the submitter thread. 1201 * Returns ERR_PTR(-ENOMEM) if we cannot allocate a drbd_request. 1202 */ 1203 static struct drbd_request * 1204 drbd_request_prepare(struct drbd_device *device, struct bio *bio) 1205 { 1206 const int rw = bio_data_dir(bio); 1207 struct drbd_request *req; 1208 1209 /* allocate outside of all locks; */ 1210 req = drbd_req_new(device, bio); 1211 if (!req) { 1212 dec_ap_bio(device); 1213 /* only pass the error to the upper layers. 1214 * if user cannot handle io errors, that's not our business. */ 1215 drbd_err(device, "could not kmalloc() req\n"); 1216 bio->bi_status = BLK_STS_RESOURCE; 1217 bio_endio(bio); 1218 return ERR_PTR(-ENOMEM); 1219 } 1220 1221 /* Update disk stats */ 1222 req->start_jif = bio_start_io_acct(req->master_bio); 1223 1224 if (!get_ldev(device)) { 1225 bio_put(req->private_bio); 1226 req->private_bio = NULL; 1227 } 1228 1229 /* process discards always from our submitter thread */ 1230 if (bio_op(bio) == REQ_OP_WRITE_ZEROES || 1231 bio_op(bio) == REQ_OP_DISCARD) 1232 goto queue_for_submitter_thread; 1233 1234 if (rw == WRITE && req->private_bio && req->i.size 1235 && !test_bit(AL_SUSPENDED, &device->flags)) { 1236 if (!drbd_al_begin_io_fastpath(device, &req->i)) 1237 goto queue_for_submitter_thread; 1238 req->rq_state |= RQ_IN_ACT_LOG; 1239 req->in_actlog_jif = jiffies; 1240 } 1241 return req; 1242 1243 queue_for_submitter_thread: 1244 atomic_inc(&device->ap_actlog_cnt); 1245 drbd_queue_write(device, req); 1246 return NULL; 1247 } 1248 1249 /* Require at least one path to current data. 1250 * We don't want to allow writes on C_STANDALONE D_INCONSISTENT: 1251 * We would not allow to read what was written, 1252 * we would not have bumped the data generation uuids, 1253 * we would cause data divergence for all the wrong reasons. 1254 * 1255 * If we don't see at least one D_UP_TO_DATE, we will fail this request, 1256 * which either returns EIO, or, if OND_SUSPEND_IO is set, suspends IO, 1257 * and queues for retry later. 1258 */ 1259 static bool may_do_writes(struct drbd_device *device) 1260 { 1261 const union drbd_dev_state s = device->state; 1262 return s.disk == D_UP_TO_DATE || s.pdsk == D_UP_TO_DATE; 1263 } 1264 1265 struct drbd_plug_cb { 1266 struct blk_plug_cb cb; 1267 struct drbd_request *most_recent_req; 1268 /* do we need more? */ 1269 }; 1270 1271 static void drbd_unplug(struct blk_plug_cb *cb, bool from_schedule) 1272 { 1273 struct drbd_plug_cb *plug = container_of(cb, struct drbd_plug_cb, cb); 1274 struct drbd_resource *resource = plug->cb.data; 1275 struct drbd_request *req = plug->most_recent_req; 1276 1277 kfree(cb); 1278 if (!req) 1279 return; 1280 1281 spin_lock_irq(&resource->req_lock); 1282 /* In case the sender did not process it yet, raise the flag to 1283 * have it followed with P_UNPLUG_REMOTE just after. */ 1284 req->rq_state |= RQ_UNPLUG; 1285 /* but also queue a generic unplug */ 1286 drbd_queue_unplug(req->device); 1287 kref_put(&req->kref, drbd_req_destroy); 1288 spin_unlock_irq(&resource->req_lock); 1289 } 1290 1291 static struct drbd_plug_cb* drbd_check_plugged(struct drbd_resource *resource) 1292 { 1293 /* A lot of text to say 1294 * return (struct drbd_plug_cb*)blk_check_plugged(); */ 1295 struct drbd_plug_cb *plug; 1296 struct blk_plug_cb *cb = blk_check_plugged(drbd_unplug, resource, sizeof(*plug)); 1297 1298 if (cb) 1299 plug = container_of(cb, struct drbd_plug_cb, cb); 1300 else 1301 plug = NULL; 1302 return plug; 1303 } 1304 1305 static void drbd_update_plug(struct drbd_plug_cb *plug, struct drbd_request *req) 1306 { 1307 struct drbd_request *tmp = plug->most_recent_req; 1308 /* Will be sent to some peer. 1309 * Remember to tag it with UNPLUG_REMOTE on unplug */ 1310 kref_get(&req->kref); 1311 plug->most_recent_req = req; 1312 if (tmp) 1313 kref_put(&tmp->kref, drbd_req_destroy); 1314 } 1315 1316 static void drbd_send_and_submit(struct drbd_device *device, struct drbd_request *req) 1317 { 1318 struct drbd_resource *resource = device->resource; 1319 const int rw = bio_data_dir(req->master_bio); 1320 struct bio_and_error m = { NULL, }; 1321 bool no_remote = false; 1322 bool submit_private_bio = false; 1323 1324 spin_lock_irq(&resource->req_lock); 1325 if (rw == WRITE) { 1326 /* This may temporarily give up the req_lock, 1327 * but will re-aquire it before it returns here. 1328 * Needs to be before the check on drbd_suspended() */ 1329 complete_conflicting_writes(req); 1330 /* no more giving up req_lock from now on! */ 1331 1332 /* check for congestion, and potentially stop sending 1333 * full data updates, but start sending "dirty bits" only. */ 1334 maybe_pull_ahead(device); 1335 } 1336 1337 1338 if (drbd_suspended(device)) { 1339 /* push back and retry: */ 1340 req->rq_state |= RQ_POSTPONED; 1341 if (req->private_bio) { 1342 bio_put(req->private_bio); 1343 req->private_bio = NULL; 1344 put_ldev(device); 1345 } 1346 goto out; 1347 } 1348 1349 /* We fail READ early, if we can not serve it. 1350 * We must do this before req is registered on any lists. 1351 * Otherwise, drbd_req_complete() will queue failed READ for retry. */ 1352 if (rw != WRITE) { 1353 if (!do_remote_read(req) && !req->private_bio) 1354 goto nodata; 1355 } 1356 1357 /* which transfer log epoch does this belong to? */ 1358 req->epoch = atomic_read(&first_peer_device(device)->connection->current_tle_nr); 1359 1360 /* no point in adding empty flushes to the transfer log, 1361 * they are mapped to drbd barriers already. */ 1362 if (likely(req->i.size!=0)) { 1363 if (rw == WRITE) 1364 first_peer_device(device)->connection->current_tle_writes++; 1365 1366 list_add_tail(&req->tl_requests, &first_peer_device(device)->connection->transfer_log); 1367 } 1368 1369 if (rw == WRITE) { 1370 if (req->private_bio && !may_do_writes(device)) { 1371 bio_put(req->private_bio); 1372 req->private_bio = NULL; 1373 put_ldev(device); 1374 goto nodata; 1375 } 1376 if (!drbd_process_write_request(req)) 1377 no_remote = true; 1378 } else { 1379 /* We either have a private_bio, or we can read from remote. 1380 * Otherwise we had done the goto nodata above. */ 1381 if (req->private_bio == NULL) { 1382 _req_mod(req, TO_BE_SENT); 1383 _req_mod(req, QUEUE_FOR_NET_READ); 1384 } else 1385 no_remote = true; 1386 } 1387 1388 if (no_remote == false) { 1389 struct drbd_plug_cb *plug = drbd_check_plugged(resource); 1390 if (plug) 1391 drbd_update_plug(plug, req); 1392 } 1393 1394 /* If it took the fast path in drbd_request_prepare, add it here. 1395 * The slow path has added it already. */ 1396 if (list_empty(&req->req_pending_master_completion)) 1397 list_add_tail(&req->req_pending_master_completion, 1398 &device->pending_master_completion[rw == WRITE]); 1399 if (req->private_bio) { 1400 /* needs to be marked within the same spinlock */ 1401 req->pre_submit_jif = jiffies; 1402 list_add_tail(&req->req_pending_local, 1403 &device->pending_completion[rw == WRITE]); 1404 _req_mod(req, TO_BE_SUBMITTED); 1405 /* but we need to give up the spinlock to submit */ 1406 submit_private_bio = true; 1407 } else if (no_remote) { 1408 nodata: 1409 if (__ratelimit(&drbd_ratelimit_state)) 1410 drbd_err(device, "IO ERROR: neither local nor remote data, sector %llu+%u\n", 1411 (unsigned long long)req->i.sector, req->i.size >> 9); 1412 /* A write may have been queued for send_oos, however. 1413 * So we can not simply free it, we must go through drbd_req_put_completion_ref() */ 1414 } 1415 1416 out: 1417 drbd_req_put_completion_ref(req, &m, 1); 1418 spin_unlock_irq(&resource->req_lock); 1419 1420 /* Even though above is a kref_put(), this is safe. 1421 * As long as we still need to submit our private bio, 1422 * we hold a completion ref, and the request cannot disappear. 1423 * If however this request did not even have a private bio to submit 1424 * (e.g. remote read), req may already be invalid now. 1425 * That's why we cannot check on req->private_bio. */ 1426 if (submit_private_bio) 1427 drbd_submit_req_private_bio(req); 1428 if (m.bio) 1429 complete_master_bio(device, &m); 1430 } 1431 1432 void __drbd_make_request(struct drbd_device *device, struct bio *bio) 1433 { 1434 struct drbd_request *req = drbd_request_prepare(device, bio); 1435 if (IS_ERR_OR_NULL(req)) 1436 return; 1437 drbd_send_and_submit(device, req); 1438 } 1439 1440 static void submit_fast_path(struct drbd_device *device, struct list_head *incoming) 1441 { 1442 struct blk_plug plug; 1443 struct drbd_request *req, *tmp; 1444 1445 blk_start_plug(&plug); 1446 list_for_each_entry_safe(req, tmp, incoming, tl_requests) { 1447 const int rw = bio_data_dir(req->master_bio); 1448 1449 if (rw == WRITE /* rw != WRITE should not even end up here! */ 1450 && req->private_bio && req->i.size 1451 && !test_bit(AL_SUSPENDED, &device->flags)) { 1452 if (!drbd_al_begin_io_fastpath(device, &req->i)) 1453 continue; 1454 1455 req->rq_state |= RQ_IN_ACT_LOG; 1456 req->in_actlog_jif = jiffies; 1457 atomic_dec(&device->ap_actlog_cnt); 1458 } 1459 1460 list_del_init(&req->tl_requests); 1461 drbd_send_and_submit(device, req); 1462 } 1463 blk_finish_plug(&plug); 1464 } 1465 1466 static bool prepare_al_transaction_nonblock(struct drbd_device *device, 1467 struct list_head *incoming, 1468 struct list_head *pending, 1469 struct list_head *later) 1470 { 1471 struct drbd_request *req; 1472 int wake = 0; 1473 int err; 1474 1475 spin_lock_irq(&device->al_lock); 1476 while ((req = list_first_entry_or_null(incoming, struct drbd_request, tl_requests))) { 1477 err = drbd_al_begin_io_nonblock(device, &req->i); 1478 if (err == -ENOBUFS) 1479 break; 1480 if (err == -EBUSY) 1481 wake = 1; 1482 if (err) 1483 list_move_tail(&req->tl_requests, later); 1484 else 1485 list_move_tail(&req->tl_requests, pending); 1486 } 1487 spin_unlock_irq(&device->al_lock); 1488 if (wake) 1489 wake_up(&device->al_wait); 1490 return !list_empty(pending); 1491 } 1492 1493 static void send_and_submit_pending(struct drbd_device *device, struct list_head *pending) 1494 { 1495 struct blk_plug plug; 1496 struct drbd_request *req; 1497 1498 blk_start_plug(&plug); 1499 while ((req = list_first_entry_or_null(pending, struct drbd_request, tl_requests))) { 1500 req->rq_state |= RQ_IN_ACT_LOG; 1501 req->in_actlog_jif = jiffies; 1502 atomic_dec(&device->ap_actlog_cnt); 1503 list_del_init(&req->tl_requests); 1504 drbd_send_and_submit(device, req); 1505 } 1506 blk_finish_plug(&plug); 1507 } 1508 1509 void do_submit(struct work_struct *ws) 1510 { 1511 struct drbd_device *device = container_of(ws, struct drbd_device, submit.worker); 1512 LIST_HEAD(incoming); /* from drbd_make_request() */ 1513 LIST_HEAD(pending); /* to be submitted after next AL-transaction commit */ 1514 LIST_HEAD(busy); /* blocked by resync requests */ 1515 1516 /* grab new incoming requests */ 1517 spin_lock_irq(&device->resource->req_lock); 1518 list_splice_tail_init(&device->submit.writes, &incoming); 1519 spin_unlock_irq(&device->resource->req_lock); 1520 1521 for (;;) { 1522 DEFINE_WAIT(wait); 1523 1524 /* move used-to-be-busy back to front of incoming */ 1525 list_splice_init(&busy, &incoming); 1526 submit_fast_path(device, &incoming); 1527 if (list_empty(&incoming)) 1528 break; 1529 1530 for (;;) { 1531 prepare_to_wait(&device->al_wait, &wait, TASK_UNINTERRUPTIBLE); 1532 1533 list_splice_init(&busy, &incoming); 1534 prepare_al_transaction_nonblock(device, &incoming, &pending, &busy); 1535 if (!list_empty(&pending)) 1536 break; 1537 1538 schedule(); 1539 1540 /* If all currently "hot" activity log extents are kept busy by 1541 * incoming requests, we still must not totally starve new 1542 * requests to "cold" extents. 1543 * Something left on &incoming means there had not been 1544 * enough update slots available, and the activity log 1545 * has been marked as "starving". 1546 * 1547 * Try again now, without looking for new requests, 1548 * effectively blocking all new requests until we made 1549 * at least _some_ progress with what we currently have. 1550 */ 1551 if (!list_empty(&incoming)) 1552 continue; 1553 1554 /* Nothing moved to pending, but nothing left 1555 * on incoming: all moved to busy! 1556 * Grab new and iterate. */ 1557 spin_lock_irq(&device->resource->req_lock); 1558 list_splice_tail_init(&device->submit.writes, &incoming); 1559 spin_unlock_irq(&device->resource->req_lock); 1560 } 1561 finish_wait(&device->al_wait, &wait); 1562 1563 /* If the transaction was full, before all incoming requests 1564 * had been processed, skip ahead to commit, and iterate 1565 * without splicing in more incoming requests from upper layers. 1566 * 1567 * Else, if all incoming have been processed, 1568 * they have become either "pending" (to be submitted after 1569 * next transaction commit) or "busy" (blocked by resync). 1570 * 1571 * Maybe more was queued, while we prepared the transaction? 1572 * Try to stuff those into this transaction as well. 1573 * Be strictly non-blocking here, 1574 * we already have something to commit. 1575 * 1576 * Commit if we don't make any more progres. 1577 */ 1578 1579 while (list_empty(&incoming)) { 1580 LIST_HEAD(more_pending); 1581 LIST_HEAD(more_incoming); 1582 bool made_progress; 1583 1584 /* It is ok to look outside the lock, 1585 * it's only an optimization anyways */ 1586 if (list_empty(&device->submit.writes)) 1587 break; 1588 1589 spin_lock_irq(&device->resource->req_lock); 1590 list_splice_tail_init(&device->submit.writes, &more_incoming); 1591 spin_unlock_irq(&device->resource->req_lock); 1592 1593 if (list_empty(&more_incoming)) 1594 break; 1595 1596 made_progress = prepare_al_transaction_nonblock(device, &more_incoming, &more_pending, &busy); 1597 1598 list_splice_tail_init(&more_pending, &pending); 1599 list_splice_tail_init(&more_incoming, &incoming); 1600 if (!made_progress) 1601 break; 1602 } 1603 1604 drbd_al_begin_io_commit(device); 1605 send_and_submit_pending(device, &pending); 1606 } 1607 } 1608 1609 void drbd_submit_bio(struct bio *bio) 1610 { 1611 struct drbd_device *device = bio->bi_bdev->bd_disk->private_data; 1612 1613 blk_queue_split(&bio); 1614 1615 /* 1616 * what we "blindly" assume: 1617 */ 1618 D_ASSERT(device, IS_ALIGNED(bio->bi_iter.bi_size, 512)); 1619 1620 inc_ap_bio(device); 1621 __drbd_make_request(device, bio); 1622 } 1623 1624 static bool net_timeout_reached(struct drbd_request *net_req, 1625 struct drbd_connection *connection, 1626 unsigned long now, unsigned long ent, 1627 unsigned int ko_count, unsigned int timeout) 1628 { 1629 struct drbd_device *device = net_req->device; 1630 1631 if (!time_after(now, net_req->pre_send_jif + ent)) 1632 return false; 1633 1634 if (time_in_range(now, connection->last_reconnect_jif, connection->last_reconnect_jif + ent)) 1635 return false; 1636 1637 if (net_req->rq_state & RQ_NET_PENDING) { 1638 drbd_warn(device, "Remote failed to finish a request within %ums > ko-count (%u) * timeout (%u * 0.1s)\n", 1639 jiffies_to_msecs(now - net_req->pre_send_jif), ko_count, timeout); 1640 return true; 1641 } 1642 1643 /* We received an ACK already (or are using protocol A), 1644 * but are waiting for the epoch closing barrier ack. 1645 * Check if we sent the barrier already. We should not blame the peer 1646 * for being unresponsive, if we did not even ask it yet. */ 1647 if (net_req->epoch == connection->send.current_epoch_nr) { 1648 drbd_warn(device, 1649 "We did not send a P_BARRIER for %ums > ko-count (%u) * timeout (%u * 0.1s); drbd kernel thread blocked?\n", 1650 jiffies_to_msecs(now - net_req->pre_send_jif), ko_count, timeout); 1651 return false; 1652 } 1653 1654 /* Worst case: we may have been blocked for whatever reason, then 1655 * suddenly are able to send a lot of requests (and epoch separating 1656 * barriers) in quick succession. 1657 * The timestamp of the net_req may be much too old and not correspond 1658 * to the sending time of the relevant unack'ed barrier packet, so 1659 * would trigger a spurious timeout. The latest barrier packet may 1660 * have a too recent timestamp to trigger the timeout, potentially miss 1661 * a timeout. Right now we don't have a place to conveniently store 1662 * these timestamps. 1663 * But in this particular situation, the application requests are still 1664 * completed to upper layers, DRBD should still "feel" responsive. 1665 * No need yet to kill this connection, it may still recover. 1666 * If not, eventually we will have queued enough into the network for 1667 * us to block. From that point of view, the timestamp of the last sent 1668 * barrier packet is relevant enough. 1669 */ 1670 if (time_after(now, connection->send.last_sent_barrier_jif + ent)) { 1671 drbd_warn(device, "Remote failed to answer a P_BARRIER (sent at %lu jif; now=%lu jif) within %ums > ko-count (%u) * timeout (%u * 0.1s)\n", 1672 connection->send.last_sent_barrier_jif, now, 1673 jiffies_to_msecs(now - connection->send.last_sent_barrier_jif), ko_count, timeout); 1674 return true; 1675 } 1676 return false; 1677 } 1678 1679 /* A request is considered timed out, if 1680 * - we have some effective timeout from the configuration, 1681 * with some state restrictions applied, 1682 * - the oldest request is waiting for a response from the network 1683 * resp. the local disk, 1684 * - the oldest request is in fact older than the effective timeout, 1685 * - the connection was established (resp. disk was attached) 1686 * for longer than the timeout already. 1687 * Note that for 32bit jiffies and very stable connections/disks, 1688 * we may have a wrap around, which is catched by 1689 * !time_in_range(now, last_..._jif, last_..._jif + timeout). 1690 * 1691 * Side effect: once per 32bit wrap-around interval, which means every 1692 * ~198 days with 250 HZ, we have a window where the timeout would need 1693 * to expire twice (worst case) to become effective. Good enough. 1694 */ 1695 1696 void request_timer_fn(struct timer_list *t) 1697 { 1698 struct drbd_device *device = from_timer(device, t, request_timer); 1699 struct drbd_connection *connection = first_peer_device(device)->connection; 1700 struct drbd_request *req_read, *req_write, *req_peer; /* oldest request */ 1701 struct net_conf *nc; 1702 unsigned long oldest_submit_jif; 1703 unsigned long ent = 0, dt = 0, et, nt; /* effective timeout = ko_count * timeout */ 1704 unsigned long now; 1705 unsigned int ko_count = 0, timeout = 0; 1706 1707 rcu_read_lock(); 1708 nc = rcu_dereference(connection->net_conf); 1709 if (nc && device->state.conn >= C_WF_REPORT_PARAMS) { 1710 ko_count = nc->ko_count; 1711 timeout = nc->timeout; 1712 } 1713 1714 if (get_ldev(device)) { /* implicit state.disk >= D_INCONSISTENT */ 1715 dt = rcu_dereference(device->ldev->disk_conf)->disk_timeout * HZ / 10; 1716 put_ldev(device); 1717 } 1718 rcu_read_unlock(); 1719 1720 1721 ent = timeout * HZ/10 * ko_count; 1722 et = min_not_zero(dt, ent); 1723 1724 if (!et) 1725 return; /* Recurring timer stopped */ 1726 1727 now = jiffies; 1728 nt = now + et; 1729 1730 spin_lock_irq(&device->resource->req_lock); 1731 req_read = list_first_entry_or_null(&device->pending_completion[0], struct drbd_request, req_pending_local); 1732 req_write = list_first_entry_or_null(&device->pending_completion[1], struct drbd_request, req_pending_local); 1733 1734 /* maybe the oldest request waiting for the peer is in fact still 1735 * blocking in tcp sendmsg. That's ok, though, that's handled via the 1736 * socket send timeout, requesting a ping, and bumping ko-count in 1737 * we_should_drop_the_connection(). 1738 */ 1739 1740 /* check the oldest request we did successfully sent, 1741 * but which is still waiting for an ACK. */ 1742 req_peer = connection->req_ack_pending; 1743 1744 /* if we don't have such request (e.g. protocoll A) 1745 * check the oldest requests which is still waiting on its epoch 1746 * closing barrier ack. */ 1747 if (!req_peer) 1748 req_peer = connection->req_not_net_done; 1749 1750 /* evaluate the oldest peer request only in one timer! */ 1751 if (req_peer && req_peer->device != device) 1752 req_peer = NULL; 1753 1754 /* do we have something to evaluate? */ 1755 if (req_peer == NULL && req_write == NULL && req_read == NULL) 1756 goto out; 1757 1758 oldest_submit_jif = 1759 (req_write && req_read) 1760 ? ( time_before(req_write->pre_submit_jif, req_read->pre_submit_jif) 1761 ? req_write->pre_submit_jif : req_read->pre_submit_jif ) 1762 : req_write ? req_write->pre_submit_jif 1763 : req_read ? req_read->pre_submit_jif : now; 1764 1765 if (ent && req_peer && net_timeout_reached(req_peer, connection, now, ent, ko_count, timeout)) 1766 _conn_request_state(connection, NS(conn, C_TIMEOUT), CS_VERBOSE | CS_HARD); 1767 1768 if (dt && oldest_submit_jif != now && 1769 time_after(now, oldest_submit_jif + dt) && 1770 !time_in_range(now, device->last_reattach_jif, device->last_reattach_jif + dt)) { 1771 drbd_warn(device, "Local backing device failed to meet the disk-timeout\n"); 1772 __drbd_chk_io_error(device, DRBD_FORCE_DETACH); 1773 } 1774 1775 /* Reschedule timer for the nearest not already expired timeout. 1776 * Fallback to now + min(effective network timeout, disk timeout). */ 1777 ent = (ent && req_peer && time_before(now, req_peer->pre_send_jif + ent)) 1778 ? req_peer->pre_send_jif + ent : now + et; 1779 dt = (dt && oldest_submit_jif != now && time_before(now, oldest_submit_jif + dt)) 1780 ? oldest_submit_jif + dt : now + et; 1781 nt = time_before(ent, dt) ? ent : dt; 1782 out: 1783 spin_unlock_irq(&device->resource->req_lock); 1784 mod_timer(&device->request_timer, nt); 1785 } 1786