drivers/block/drbd/drbd_req.c

   1 /*
   2    drbd_req.c
   3
   4    This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
   5
   6    Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
   7    Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
   8    Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
   9
  10    drbd is free software; you can redistribute it and/or modify
  11    it under the terms of the GNU General Public License as published by
  12    the Free Software Foundation; either version 2, or (at your option)
  13    any later version.
  14
  15    drbd is distributed in the hope that it will be useful,
  16    but WITHOUT ANY WARRANTY; without even the implied warranty of
  17    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  18    GNU General Public License for more details.
  19
  20    You should have received a copy of the GNU General Public License
  21    along with drbd; see the file COPYING.  If not, write to
  22    the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
  23
  24  */
  25
  26 #include <linux/module.h>
  27
  28 #include <linux/slab.h>
  29 #include <linux/drbd.h>
  30 #include "drbd_int.h"
  31 #include "drbd_req.h"
  32
  33
  34 static bool drbd_may_do_local_read(struct drbd_conf *mdev, sector_t sector, int size);
  35
  36 /* Update disk stats at start of I/O request */
  37 static void _drbd_start_io_acct(struct drbd_conf *mdev, struct drbd_request *req, struct bio *bio)
  38 {
  39         const int rw = bio_data_dir(bio);
  40         int cpu;
  41         cpu = part_stat_lock();
  42         part_round_stats(cpu, &mdev->vdisk->part0);
  43         part_stat_inc(cpu, &mdev->vdisk->part0, ios[rw]);
  44         part_stat_add(cpu, &mdev->vdisk->part0, sectors[rw], bio_sectors(bio));
  45         (void) cpu; /* The macro invocations above want the cpu argument, I do not like
  46                        the compiler warning about cpu only assigned but never used... */
  47         part_inc_in_flight(&mdev->vdisk->part0, rw);
  48         part_stat_unlock();
  49 }
  50
  51 /* Update disk stats when completing request upwards */
  52 static void _drbd_end_io_acct(struct drbd_conf *mdev, struct drbd_request *req)
  53 {
  54         int rw = bio_data_dir(req->master_bio);
  55         unsigned long duration = jiffies - req->start_time;
  56         int cpu;
  57         cpu = part_stat_lock();
  58         part_stat_add(cpu, &mdev->vdisk->part0, ticks[rw], duration);
  59         part_round_stats(cpu, &mdev->vdisk->part0);
  60         part_dec_in_flight(&mdev->vdisk->part0, rw);
  61         part_stat_unlock();
  62 }
  63
  64 static struct drbd_request *drbd_req_new(struct drbd_conf *mdev,
  65                                                struct bio *bio_src)
  66 {
  67         struct drbd_request *req;
  68
  69         req = mempool_alloc(drbd_request_mempool, GFP_NOIO);
  70         if (!req)
  71                 return NULL;
  72
  73         drbd_req_make_private_bio(req, bio_src);
  74         req->rq_state    = bio_data_dir(bio_src) == WRITE ? RQ_WRITE : 0;
  75         req->w.mdev      = mdev;
  76         req->master_bio  = bio_src;
  77         req->epoch       = 0;
  78
  79         drbd_clear_interval(&req->i);
  80         req->i.sector     = bio_src->bi_sector;
  81         req->i.size      = bio_src->bi_size;
  82         req->i.local = true;
  83         req->i.waiting = false;
  84
  85         INIT_LIST_HEAD(&req->tl_requests);
  86         INIT_LIST_HEAD(&req->w.list);
  87
  88         /* one reference to be put by __drbd_make_request */
  89         atomic_set(&req->completion_ref, 1);
  90         /* one kref as long as completion_ref > 0 */
  91         kref_init(&req->kref);
  92         return req;
  93 }
  94
  95 void drbd_req_destroy(struct kref *kref)
  96 {
  97         struct drbd_request *req = container_of(kref, struct drbd_request, kref);
  98         struct drbd_conf *mdev = req->w.mdev;
  99         const unsigned s = req->rq_state;
 100
 101         if ((req->master_bio && !(s & RQ_POSTPONED)) ||
 102                 atomic_read(&req->completion_ref) ||
 103                 (s & RQ_LOCAL_PENDING) ||
 104                 ((s & RQ_NET_MASK) && !(s & RQ_NET_DONE))) {
 105                 dev_err(DEV, "drbd_req_destroy: Logic BUG rq_state = 0x%x, completion_ref = %d\n",
 106                                 s, atomic_read(&req->completion_ref));
 107                 return;
 108         }
 109
 110         /* remove it from the transfer log.
 111          * well, only if it had been there in the first
 112          * place... if it had not (local only or conflicting
 113          * and never sent), it should still be "empty" as
 114          * initialized in drbd_req_new(), so we can list_del() it
 115          * here unconditionally */
 116         list_del_init(&req->tl_requests);
 117
 118         /* if it was a write, we may have to set the corresponding
 119          * bit(s) out-of-sync first. If it had a local part, we need to
 120          * release the reference to the activity log. */
 121         if (s & RQ_WRITE) {
 122                 /* Set out-of-sync unless both OK flags are set
 123                  * (local only or remote failed).
 124                  * Other places where we set out-of-sync:
 125                  * READ with local io-error */
 126
 127                 /* There is a special case:
 128                  * we may notice late that IO was suspended,
 129                  * and postpone, or schedule for retry, a write,
 130                  * before it even was submitted or sent.
 131                  * In that case we do not want to touch the bitmap at all.
 132                  */
 133                 if ((s & (RQ_POSTPONED|RQ_LOCAL_MASK|RQ_NET_MASK)) != RQ_POSTPONED) {
 134                         if (!(s & RQ_NET_OK) || !(s & RQ_LOCAL_OK))
 135                                 drbd_set_out_of_sync(mdev, req->i.sector, req->i.size);
 136
 137                         if ((s & RQ_NET_OK) && (s & RQ_LOCAL_OK) && (s & RQ_NET_SIS))
 138                                 drbd_set_in_sync(mdev, req->i.sector, req->i.size);
 139                 }
 140
 141                 /* one might be tempted to move the drbd_al_complete_io
 142                  * to the local io completion callback drbd_request_endio.
 143                  * but, if this was a mirror write, we may only
 144                  * drbd_al_complete_io after this is RQ_NET_DONE,
 145                  * otherwise the extent could be dropped from the al
 146                  * before it has actually been written on the peer.
 147                  * if we crash before our peer knows about the request,
 148                  * but after the extent has been dropped from the al,
 149                  * we would forget to resync the corresponding extent.
 150                  */
 151                 if (s & RQ_IN_ACT_LOG) {
 152                         if (get_ldev_if_state(mdev, D_FAILED)) {
 153                                 drbd_al_complete_io(mdev, &req->i);
 154                                 put_ldev(mdev);
 155                         } else if (__ratelimit(&drbd_ratelimit_state)) {
 156                                 dev_warn(DEV, "Should have called drbd_al_complete_io(, %llu, %u), "
 157                                          "but my Disk seems to have failed :(\n",
 158                                          (unsigned long long) req->i.sector, req->i.size);
 159                         }
 160                 }
 161         }
 162
 163         mempool_free(req, drbd_request_mempool);
 164 }
 165
 166 static void wake_all_senders(struct drbd_tconn *tconn) {
 167         wake_up(&tconn->sender_work.q_wait);
 168 }
 169
 170 /* must hold resource->req_lock */
 171 static void start_new_tl_epoch(struct drbd_tconn *tconn)
 172 {
 173         /* no point closing an epoch, if it is empty, anyways. */
 174         if (tconn->current_tle_writes == 0)
 175                 return;
 176
 177         tconn->current_tle_writes = 0;
 178         atomic_inc(&tconn->current_tle_nr);
 179         wake_all_senders(tconn);
 180 }
 181
 182 void complete_master_bio(struct drbd_conf *mdev,
 183                 struct bio_and_error *m)
 184 {
 185         bio_endio(m->bio, m->error);
 186         dec_ap_bio(mdev);
 187 }
 188
 189
 190 static void drbd_remove_request_interval(struct rb_root *root,
 191                                          struct drbd_request *req)
 192 {
 193         struct drbd_conf *mdev = req->w.mdev;
 194         struct drbd_interval *i = &req->i;
 195
 196         drbd_remove_interval(root, i);
 197
 198         /* Wake up any processes waiting for this request to complete.  */
 199         if (i->waiting)
 200                 wake_up(&mdev->misc_wait);
 201 }
 202
 203 /* Helper for __req_mod().
 204  * Set m->bio to the master bio, if it is fit to be completed,
 205  * or leave it alone (it is initialized to NULL in __req_mod),
 206  * if it has already been completed, or cannot be completed yet.
 207  * If m->bio is set, the error status to be returned is placed in m->error.
 208  */
 209 static
 210 void drbd_req_complete(struct drbd_request *req, struct bio_and_error *m)
 211 {
 212         const unsigned s = req->rq_state;
 213         struct drbd_conf *mdev = req->w.mdev;
 214         int rw;
 215         int error, ok;
 216
 217         /* we must not complete the master bio, while it is
 218          *      still being processed by _drbd_send_zc_bio (drbd_send_dblock)
 219          *      not yet acknowledged by the peer
 220          *      not yet completed by the local io subsystem
 221          * these flags may get cleared in any order by
 222          *      the worker,
 223          *      the receiver,
 224          *      the bio_endio completion callbacks.
 225          */
 226         if ((s & RQ_LOCAL_PENDING && !(s & RQ_LOCAL_ABORTED)) ||
 227             (s & RQ_NET_QUEUED) || (s & RQ_NET_PENDING) ||
 228             (s & RQ_COMPLETION_SUSP)) {
 229                 dev_err(DEV, "drbd_req_complete: Logic BUG rq_state = 0x%x\n", s);
 230                 return;
 231         }
 232
 233         if (!req->master_bio) {
 234                 dev_err(DEV, "drbd_req_complete: Logic BUG, master_bio == NULL!\n");
 235                 return;
 236         }
 237
 238         rw = bio_rw(req->master_bio);
 239
 240         /*
 241          * figure out whether to report success or failure.
 242          *
 243          * report success when at least one of the operations succeeded.
 244          * or, to put the other way,
 245          * only report failure, when both operations failed.
 246          *
 247          * what to do about the failures is handled elsewhere.
 248          * what we need to do here is just: complete the master_bio.
 249          *
 250          * local completion error, if any, has been stored as ERR_PTR
 251          * in private_bio within drbd_request_endio.
 252          */
 253         ok = (s & RQ_LOCAL_OK) || (s & RQ_NET_OK);
 254         error = PTR_ERR(req->private_bio);
 255
 256         /* remove the request from the conflict detection
 257          * respective block_id verification hash */
 258         if (!drbd_interval_empty(&req->i)) {
 259                 struct rb_root *root;
 260
 261                 if (rw == WRITE)
 262                         root = &mdev->write_requests;
 263                 else
 264                         root = &mdev->read_requests;
 265                 drbd_remove_request_interval(root, req);
 266         } else if (!(s & RQ_POSTPONED))
 267                 D_ASSERT((s & (RQ_NET_MASK & ~RQ_NET_DONE)) == 0);
 268
 269         /* Before we can signal completion to the upper layers,
 270          * we may need to close the current transfer log epoch.
 271          * We are within the request lock, so we can simply compare
 272          * the request epoch number with the current transfer log
 273          * epoch number.  If they match, increase the current_tle_nr,
 274          * and reset the transfer log epoch write_cnt.
 275          */
 276         if (rw == WRITE &&
 277             req->epoch == atomic_read(&mdev->tconn->current_tle_nr))
 278                 start_new_tl_epoch(mdev->tconn);
 279
 280         /* Update disk stats */
 281         _drbd_end_io_acct(mdev, req);
 282
 283         /* If READ failed,
 284          * have it be pushed back to the retry work queue,
 285          * so it will re-enter __drbd_make_request(),
 286          * and be re-assigned to a suitable local or remote path,
 287          * or failed if we do not have access to good data anymore.
 288          *
 289          * Unless it was failed early by __drbd_make_request(),
 290          * because no path was available, in which case
 291          * it was not even added to the transfer_log.
 292          *
 293          * READA may fail, and will not be retried.
 294          *
 295          * WRITE should have used all available paths already.
 296          */
 297         if (!ok && rw == READ && !list_empty(&req->tl_requests))
 298                 req->rq_state |= RQ_POSTPONED;
 299
 300         if (!(req->rq_state & RQ_POSTPONED)) {
 301                 m->error = ok ? 0 : (error ?: -EIO);
 302                 m->bio = req->master_bio;
 303                 req->master_bio = NULL;
 304         }
 305 }
 306
 307 static int drbd_req_put_completion_ref(struct drbd_request *req, struct bio_and_error *m, int put)
 308 {
 309         struct drbd_conf *mdev = req->w.mdev;
 310         D_ASSERT(m || (req->rq_state & RQ_POSTPONED));
 311
 312         if (!atomic_sub_and_test(put, &req->completion_ref))
 313                 return 0;
 314
 315         drbd_req_complete(req, m);
 316
 317         if (req->rq_state & RQ_POSTPONED) {
 318                 /* don't destroy the req object just yet,
 319                  * but queue it for retry */
 320                 drbd_restart_request(req);
 321                 return 0;
 322         }
 323
 324         return 1;
 325 }
 326
 327 /* I'd like this to be the only place that manipulates
 328  * req->completion_ref and req->kref. */
 329 static void mod_rq_state(struct drbd_request *req, struct bio_and_error *m,
 330                 int clear, int set)
 331 {
 332         struct drbd_conf *mdev = req->w.mdev;
 333         unsigned s = req->rq_state;
 334         int c_put = 0;
 335         int k_put = 0;
 336
 337         if (drbd_suspended(mdev) && !((s | clear) & RQ_COMPLETION_SUSP))
 338                 set |= RQ_COMPLETION_SUSP;
 339
 340         /* apply */
 341
 342         req->rq_state &= ~clear;
 343         req->rq_state |= set;
 344
 345         /* no change? */
 346         if (req->rq_state == s)
 347                 return;
 348
 349         /* intent: get references */
 350
 351         if (!(s & RQ_LOCAL_PENDING) && (set & RQ_LOCAL_PENDING))
 352                 atomic_inc(&req->completion_ref);
 353
 354         if (!(s & RQ_NET_PENDING) && (set & RQ_NET_PENDING)) {
 355                 inc_ap_pending(mdev);
 356                 atomic_inc(&req->completion_ref);
 357         }
 358
 359         if (!(s & RQ_NET_QUEUED) && (set & RQ_NET_QUEUED))
 360                 atomic_inc(&req->completion_ref);
 361
 362         if (!(s & RQ_EXP_BARR_ACK) && (set & RQ_EXP_BARR_ACK))
 363                 kref_get(&req->kref); /* wait for the DONE */
 364
 365         if (!(s & RQ_NET_SENT) && (set & RQ_NET_SENT))
 366                 atomic_add(req->i.size >> 9, &mdev->ap_in_flight);
 367
 368         if (!(s & RQ_COMPLETION_SUSP) && (set & RQ_COMPLETION_SUSP))
 369                 atomic_inc(&req->completion_ref);
 370
 371         /* progress: put references */
 372
 373         if ((s & RQ_COMPLETION_SUSP) && (clear & RQ_COMPLETION_SUSP))
 374                 ++c_put;
 375
 376         if (!(s & RQ_LOCAL_ABORTED) && (set & RQ_LOCAL_ABORTED)) {
 377                 D_ASSERT(req->rq_state & RQ_LOCAL_PENDING);
 378                 /* local completion may still come in later,
 379                  * we need to keep the req object around. */
 380                 kref_get(&req->kref);
 381                 ++c_put;
 382         }
 383
 384         if ((s & RQ_LOCAL_PENDING) && (clear & RQ_LOCAL_PENDING)) {
 385                 if (req->rq_state & RQ_LOCAL_ABORTED)
 386                         ++k_put;
 387                 else
 388                         ++c_put;
 389         }
 390
 391         if ((s & RQ_NET_PENDING) && (clear & RQ_NET_PENDING)) {
 392                 dec_ap_pending(mdev);
 393                 ++c_put;
 394         }
 395
 396         if ((s & RQ_NET_QUEUED) && (clear & RQ_NET_QUEUED))
 397                 ++c_put;
 398
 399         if ((s & RQ_EXP_BARR_ACK) && !(s & RQ_NET_DONE) && (set & RQ_NET_DONE)) {
 400                 if (req->rq_state & RQ_NET_SENT)
 401                         atomic_sub(req->i.size >> 9, &mdev->ap_in_flight);
 402                 ++k_put;
 403         }
 404
 405         /* potentially complete and destroy */
 406
 407         if (k_put || c_put) {
 408                 /* Completion does it's own kref_put.  If we are going to
 409                  * kref_sub below, we need req to be still around then. */
 410                 int at_least = k_put + !!c_put;
 411                 int refcount = atomic_read(&req->kref.refcount);
 412                 if (refcount < at_least)
 413                         dev_err(DEV,
 414                                 "mod_rq_state: Logic BUG: %x -> %x: refcount = %d, should be >= %d\n",
 415                                 s, req->rq_state, refcount, at_least);
 416         }
 417
 418         /* If we made progress, retry conflicting peer requests, if any. */
 419         if (req->i.waiting)
 420                 wake_up(&mdev->misc_wait);
 421
 422         if (c_put)
 423                 k_put += drbd_req_put_completion_ref(req, m, c_put);
 424         if (k_put)
 425                 kref_sub(&req->kref, k_put, drbd_req_destroy);
 426 }
 427
 428 static void drbd_report_io_error(struct drbd_conf *mdev, struct drbd_request *req)
 429 {
 430         char b[BDEVNAME_SIZE];
 431
 432         if (!__ratelimit(&drbd_ratelimit_state))
 433                 return;
 434
 435         dev_warn(DEV, "local %s IO error sector %llu+%u on %s\n",
 436                         (req->rq_state & RQ_WRITE) ? "WRITE" : "READ",
 437                         (unsigned long long)req->i.sector,
 438                         req->i.size >> 9,
 439                         bdevname(mdev->ldev->backing_bdev, b));
 440 }
 441
 442 /* obviously this could be coded as many single functions
 443  * instead of one huge switch,
 444  * or by putting the code directly in the respective locations
 445  * (as it has been before).
 446  *
 447  * but having it this way
 448  *  enforces that it is all in this one place, where it is easier to audit,
 449  *  it makes it obvious that whatever "event" "happens" to a request should
 450  *  happen "atomically" within the req_lock,
 451  *  and it enforces that we have to think in a very structured manner
 452  *  about the "events" that may happen to a request during its life time ...
 453  */
 454 int __req_mod(struct drbd_request *req, enum drbd_req_event what,
 455                 struct bio_and_error *m)
 456 {
 457         struct drbd_conf *mdev = req->w.mdev;
 458         struct net_conf *nc;
 459         int p, rv = 0;
 460
 461         if (m)
 462                 m->bio = NULL;
 463
 464         switch (what) {
 465         default:
 466                 dev_err(DEV, "LOGIC BUG in %s:%u\n", __FILE__ , __LINE__);
 467                 break;
 468
 469         /* does not happen...
 470          * initialization done in drbd_req_new
 471         case CREATED:
 472                 break;
 473                 */
 474
 475         case TO_BE_SENT: /* via network */
 476                 /* reached via __drbd_make_request
 477                  * and from w_read_retry_remote */
 478                 D_ASSERT(!(req->rq_state & RQ_NET_MASK));
 479                 rcu_read_lock();
 480                 nc = rcu_dereference(mdev->tconn->net_conf);
 481                 p = nc->wire_protocol;
 482                 rcu_read_unlock();
 483                 req->rq_state |=
 484                         p == DRBD_PROT_C ? RQ_EXP_WRITE_ACK :
 485                         p == DRBD_PROT_B ? RQ_EXP_RECEIVE_ACK : 0;
 486                 mod_rq_state(req, m, 0, RQ_NET_PENDING);
 487                 break;
 488
 489         case TO_BE_SUBMITTED: /* locally */
 490                 /* reached via __drbd_make_request */
 491                 D_ASSERT(!(req->rq_state & RQ_LOCAL_MASK));
 492                 mod_rq_state(req, m, 0, RQ_LOCAL_PENDING);
 493                 break;
 494
 495         case COMPLETED_OK:
 496                 if (req->rq_state & RQ_WRITE)
 497                         mdev->writ_cnt += req->i.size >> 9;
 498                 else
 499                         mdev->read_cnt += req->i.size >> 9;
 500
 501                 mod_rq_state(req, m, RQ_LOCAL_PENDING,
 502                                 RQ_LOCAL_COMPLETED|RQ_LOCAL_OK);
 503                 break;
 504
 505         case ABORT_DISK_IO:
 506                 mod_rq_state(req, m, 0, RQ_LOCAL_ABORTED);
 507                 break;
 508
 509         case WRITE_COMPLETED_WITH_ERROR:
 510                 drbd_report_io_error(mdev, req);
 511                 __drbd_chk_io_error(mdev, DRBD_WRITE_ERROR);
 512                 mod_rq_state(req, m, RQ_LOCAL_PENDING, RQ_LOCAL_COMPLETED);
 513                 break;
 514
 515         case READ_COMPLETED_WITH_ERROR:
 516                 drbd_set_out_of_sync(mdev, req->i.sector, req->i.size);
 517                 drbd_report_io_error(mdev, req);
 518                 __drbd_chk_io_error(mdev, DRBD_READ_ERROR);
 519                 /* fall through. */
 520         case READ_AHEAD_COMPLETED_WITH_ERROR:
 521                 /* it is legal to fail READA, no __drbd_chk_io_error in that case. */
 522                 mod_rq_state(req, m, RQ_LOCAL_PENDING, RQ_LOCAL_COMPLETED);
 523                 break;
 524
 525         case QUEUE_FOR_NET_READ:
 526                 /* READ or READA, and
 527                  * no local disk,
 528                  * or target area marked as invalid,
 529                  * or just got an io-error. */
 530                 /* from __drbd_make_request
 531                  * or from bio_endio during read io-error recovery */
 532
 533                 /* So we can verify the handle in the answer packet.
 534                  * Corresponding drbd_remove_request_interval is in
 535                  * drbd_req_complete() */
 536                 D_ASSERT(drbd_interval_empty(&req->i));
 537                 drbd_insert_interval(&mdev->read_requests, &req->i);
 538
 539                 set_bit(UNPLUG_REMOTE, &mdev->flags);
 540
 541                 D_ASSERT(req->rq_state & RQ_NET_PENDING);
 542                 D_ASSERT((req->rq_state & RQ_LOCAL_MASK) == 0);
 543                 mod_rq_state(req, m, 0, RQ_NET_QUEUED);
 544                 req->w.cb = w_send_read_req;
 545                 drbd_queue_work(&mdev->tconn->sender_work, &req->w);
 546                 break;
 547
 548         case QUEUE_FOR_NET_WRITE:
 549                 /* assert something? */
 550                 /* from __drbd_make_request only */
 551
 552                 /* Corresponding drbd_remove_request_interval is in
 553                  * drbd_req_complete() */
 554                 D_ASSERT(drbd_interval_empty(&req->i));
 555                 drbd_insert_interval(&mdev->write_requests, &req->i);
 556
 557                 /* NOTE
 558                  * In case the req ended up on the transfer log before being
 559                  * queued on the worker, it could lead to this request being
 560                  * missed during cleanup after connection loss.
 561                  * So we have to do both operations here,
 562                  * within the same lock that protects the transfer log.
 563                  *
 564                  * _req_add_to_epoch(req); this has to be after the
 565                  * _maybe_start_new_epoch(req); which happened in
 566                  * __drbd_make_request, because we now may set the bit
 567                  * again ourselves to close the current epoch.
 568                  *
 569                  * Add req to the (now) current epoch (barrier). */
 570
 571                 /* otherwise we may lose an unplug, which may cause some remote
 572                  * io-scheduler timeout to expire, increasing maximum latency,
 573                  * hurting performance. */
 574                 set_bit(UNPLUG_REMOTE, &mdev->flags);
 575
 576                 /* queue work item to send data */
 577                 D_ASSERT(req->rq_state & RQ_NET_PENDING);
 578                 mod_rq_state(req, m, 0, RQ_NET_QUEUED|RQ_EXP_BARR_ACK);
 579                 req->w.cb =  w_send_dblock;
 580                 drbd_queue_work(&mdev->tconn->sender_work, &req->w);
 581
 582                 /* close the epoch, in case it outgrew the limit */
 583                 rcu_read_lock();
 584                 nc = rcu_dereference(mdev->tconn->net_conf);
 585                 p = nc->max_epoch_size;
 586                 rcu_read_unlock();
 587                 if (mdev->tconn->current_tle_writes >= p)
 588                         start_new_tl_epoch(mdev->tconn);
 589
 590                 break;
 591
 592         case QUEUE_FOR_SEND_OOS:
 593                 mod_rq_state(req, m, 0, RQ_NET_QUEUED);
 594                 req->w.cb =  w_send_out_of_sync;
 595                 drbd_queue_work(&mdev->tconn->sender_work, &req->w);
 596                 break;
 597
 598         case READ_RETRY_REMOTE_CANCELED:
 599         case SEND_CANCELED:
 600         case SEND_FAILED:
 601                 /* real cleanup will be done from tl_clear.  just update flags
 602                  * so it is no longer marked as on the worker queue */
 603                 mod_rq_state(req, m, RQ_NET_QUEUED, 0);
 604                 break;
 605
 606         case HANDED_OVER_TO_NETWORK:
 607                 /* assert something? */
 608                 if (bio_data_dir(req->master_bio) == WRITE &&
 609                     !(req->rq_state & (RQ_EXP_RECEIVE_ACK | RQ_EXP_WRITE_ACK))) {
 610                         /* this is what is dangerous about protocol A:
 611                          * pretend it was successfully written on the peer. */
 612                         if (req->rq_state & RQ_NET_PENDING)
 613                                 mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_OK);
 614                         /* else: neg-ack was faster... */
 615                         /* it is still not yet RQ_NET_DONE until the
 616                          * corresponding epoch barrier got acked as well,
 617                          * so we know what to dirty on connection loss */
 618                 }
 619                 mod_rq_state(req, m, RQ_NET_QUEUED, RQ_NET_SENT);
 620                 break;
 621
 622         case OOS_HANDED_TO_NETWORK:
 623                 /* Was not set PENDING, no longer QUEUED, so is now DONE
 624                  * as far as this connection is concerned. */
 625                 mod_rq_state(req, m, RQ_NET_QUEUED, RQ_NET_DONE);
 626                 break;
 627
 628         case CONNECTION_LOST_WHILE_PENDING:
 629                 /* transfer log cleanup after connection loss */
 630                 mod_rq_state(req, m,
 631                                 RQ_NET_OK|RQ_NET_PENDING|RQ_COMPLETION_SUSP,
 632                                 RQ_NET_DONE);
 633                 break;
 634
 635         case CONFLICT_RESOLVED:
 636                 /* for superseded conflicting writes of multiple primaries,
 637                  * there is no need to keep anything in the tl, potential
 638                  * node crashes are covered by the activity log.
 639                  *
 640                  * If this request had been marked as RQ_POSTPONED before,
 641                  * it will actually not be completed, but "restarted",
 642                  * resubmitted from the retry worker context. */
 643                 D_ASSERT(req->rq_state & RQ_NET_PENDING);
 644                 D_ASSERT(req->rq_state & RQ_EXP_WRITE_ACK);
 645                 mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_DONE|RQ_NET_OK);
 646                 break;
 647
 648         case WRITE_ACKED_BY_PEER_AND_SIS:
 649                 req->rq_state |= RQ_NET_SIS;
 650         case WRITE_ACKED_BY_PEER:
 651                 D_ASSERT(req->rq_state & RQ_EXP_WRITE_ACK);
 652                 /* protocol C; successfully written on peer.
 653                  * Nothing more to do here.
 654                  * We want to keep the tl in place for all protocols, to cater
 655                  * for volatile write-back caches on lower level devices. */
 656
 657                 goto ack_common;
 658         case RECV_ACKED_BY_PEER:
 659                 D_ASSERT(req->rq_state & RQ_EXP_RECEIVE_ACK);
 660                 /* protocol B; pretends to be successfully written on peer.
 661                  * see also notes above in HANDED_OVER_TO_NETWORK about
 662                  * protocol != C */
 663         ack_common:
 664                 D_ASSERT(req->rq_state & RQ_NET_PENDING);
 665                 mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_OK);
 666                 break;
 667
 668         case POSTPONE_WRITE:
 669                 D_ASSERT(req->rq_state & RQ_EXP_WRITE_ACK);
 670                 /* If this node has already detected the write conflict, the
 671                  * worker will be waiting on misc_wait.  Wake it up once this
 672                  * request has completed locally.
 673                  */
 674                 D_ASSERT(req->rq_state & RQ_NET_PENDING);
 675                 req->rq_state |= RQ_POSTPONED;
 676                 if (req->i.waiting)
 677                         wake_up(&mdev->misc_wait);
 678                 /* Do not clear RQ_NET_PENDING. This request will make further
 679                  * progress via restart_conflicting_writes() or
 680                  * fail_postponed_requests(). Hopefully. */
 681                 break;
 682
 683         case NEG_ACKED:
 684                 mod_rq_state(req, m, RQ_NET_OK|RQ_NET_PENDING, 0);
 685                 break;
 686
 687         case FAIL_FROZEN_DISK_IO:
 688                 if (!(req->rq_state & RQ_LOCAL_COMPLETED))
 689                         break;
 690                 mod_rq_state(req, m, RQ_COMPLETION_SUSP, 0);
 691                 break;
 692
 693         case RESTART_FROZEN_DISK_IO:
 694                 if (!(req->rq_state & RQ_LOCAL_COMPLETED))
 695                         break;
 696
 697                 mod_rq_state(req, m,
 698                                 RQ_COMPLETION_SUSP|RQ_LOCAL_COMPLETED,
 699                                 RQ_LOCAL_PENDING);
 700
 701                 rv = MR_READ;
 702                 if (bio_data_dir(req->master_bio) == WRITE)
 703                         rv = MR_WRITE;
 704
 705                 get_ldev(mdev); /* always succeeds in this call path */
 706                 req->w.cb = w_restart_disk_io;
 707                 drbd_queue_work(&mdev->tconn->sender_work, &req->w);
 708                 break;
 709
 710         case RESEND:
 711                 /* Simply complete (local only) READs. */
 712                 if (!(req->rq_state & RQ_WRITE) && !req->w.cb) {
 713                         mod_rq_state(req, m, RQ_COMPLETION_SUSP, 0);
 714                         break;
 715                 }
 716
 717                 /* If RQ_NET_OK is already set, we got a P_WRITE_ACK or P_RECV_ACK
 718                    before the connection loss (B&C only); only P_BARRIER_ACK
 719                    (or the local completion?) was missing when we suspended.
 720                    Throwing them out of the TL here by pretending we got a BARRIER_ACK.
 721                    During connection handshake, we ensure that the peer was not rebooted. */
 722                 if (!(req->rq_state & RQ_NET_OK)) {
 723                         /* FIXME could this possibly be a req->w.cb == w_send_out_of_sync?
 724                          * in that case we must not set RQ_NET_PENDING. */
 725
 726                         mod_rq_state(req, m, RQ_COMPLETION_SUSP, RQ_NET_QUEUED|RQ_NET_PENDING);
 727                         if (req->w.cb) {
 728                                 drbd_queue_work(&mdev->tconn->sender_work, &req->w);
 729                                 rv = req->rq_state & RQ_WRITE ? MR_WRITE : MR_READ;
 730                         } /* else: FIXME can this happen? */
 731                         break;
 732                 }
 733                 /* else, fall through to BARRIER_ACKED */
 734
 735         case BARRIER_ACKED:
 736                 /* barrier ack for READ requests does not make sense */
 737                 if (!(req->rq_state & RQ_WRITE))
 738                         break;
 739
 740                 if (req->rq_state & RQ_NET_PENDING) {
 741                         /* barrier came in before all requests were acked.
 742                          * this is bad, because if the connection is lost now,
 743                          * we won't be able to clean them up... */
 744                         dev_err(DEV, "FIXME (BARRIER_ACKED but pending)\n");
 745                 }
 746                 /* Allowed to complete requests, even while suspended.
 747                  * As this is called for all requests within a matching epoch,
 748                  * we need to filter, and only set RQ_NET_DONE for those that
 749                  * have actually been on the wire. */
 750                 mod_rq_state(req, m, RQ_COMPLETION_SUSP,
 751                                 (req->rq_state & RQ_NET_MASK) ? RQ_NET_DONE : 0);
 752                 break;
 753
 754         case DATA_RECEIVED:
 755                 D_ASSERT(req->rq_state & RQ_NET_PENDING);
 756                 mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_OK|RQ_NET_DONE);
 757                 break;
 758         };
 759
 760         return rv;
 761 }
 762
 763 /* we may do a local read if:
 764  * - we are consistent (of course),
 765  * - or we are generally inconsistent,
 766  *   BUT we are still/already IN SYNC for this area.
 767  *   since size may be bigger than BM_BLOCK_SIZE,
 768  *   we may need to check several bits.
 769  */
 770 static bool drbd_may_do_local_read(struct drbd_conf *mdev, sector_t sector, int size)
 771 {
 772         unsigned long sbnr, ebnr;
 773         sector_t esector, nr_sectors;
 774
 775         if (mdev->state.disk == D_UP_TO_DATE)
 776                 return true;
 777         if (mdev->state.disk != D_INCONSISTENT)
 778                 return false;
 779         esector = sector + (size >> 9) - 1;
 780         nr_sectors = drbd_get_capacity(mdev->this_bdev);
 781         D_ASSERT(sector  < nr_sectors);
 782         D_ASSERT(esector < nr_sectors);
 783
 784         sbnr = BM_SECT_TO_BIT(sector);
 785         ebnr = BM_SECT_TO_BIT(esector);
 786
 787         return drbd_bm_count_bits(mdev, sbnr, ebnr) == 0;
 788 }
 789
 790 static bool remote_due_to_read_balancing(struct drbd_conf *mdev, sector_t sector,
 791                 enum drbd_read_balancing rbm)
 792 {
 793         struct backing_dev_info *bdi;
 794         int stripe_shift;
 795
 796         switch (rbm) {
 797         case RB_CONGESTED_REMOTE:
 798                 bdi = &mdev->ldev->backing_bdev->bd_disk->queue->backing_dev_info;
 799                 return bdi_read_congested(bdi);
 800         case RB_LEAST_PENDING:
 801                 return atomic_read(&mdev->local_cnt) >
 802                         atomic_read(&mdev->ap_pending_cnt) + atomic_read(&mdev->rs_pending_cnt);
 803         case RB_32K_STRIPING:  /* stripe_shift = 15 */
 804         case RB_64K_STRIPING:
 805         case RB_128K_STRIPING:
 806         case RB_256K_STRIPING:
 807         case RB_512K_STRIPING:
 808         case RB_1M_STRIPING:   /* stripe_shift = 20 */
 809                 stripe_shift = (rbm - RB_32K_STRIPING + 15);
 810                 return (sector >> (stripe_shift - 9)) & 1;
 811         case RB_ROUND_ROBIN:
 812                 return test_and_change_bit(READ_BALANCE_RR, &mdev->flags);
 813         case RB_PREFER_REMOTE:
 814                 return true;
 815         case RB_PREFER_LOCAL:
 816         default:
 817                 return false;
 818         }
 819 }
 820
 821 /*
 822  * complete_conflicting_writes  -  wait for any conflicting write requests
 823  *
 824  * The write_requests tree contains all active write requests which we
 825  * currently know about.  Wait for any requests to complete which conflict with
 826  * the new one.
 827  *
 828  * Only way out: remove the conflicting intervals from the tree.
 829  */
 830 static void complete_conflicting_writes(struct drbd_request *req)
 831 {
 832         DEFINE_WAIT(wait);
 833         struct drbd_conf *mdev = req->w.mdev;
 834         struct drbd_interval *i;
 835         sector_t sector = req->i.sector;
 836         int size = req->i.size;
 837
 838         i = drbd_find_overlap(&mdev->write_requests, sector, size);
 839         if (!i)
 840                 return;
 841
 842         for (;;) {
 843                 prepare_to_wait(&mdev->misc_wait, &wait, TASK_UNINTERRUPTIBLE);
 844                 i = drbd_find_overlap(&mdev->write_requests, sector, size);
 845                 if (!i)
 846                         break;
 847                 /* Indicate to wake up device->misc_wait on progress.  */
 848                 i->waiting = true;
 849                 spin_unlock_irq(&mdev->tconn->req_lock);
 850                 schedule();
 851                 spin_lock_irq(&mdev->tconn->req_lock);
 852         }
 853         finish_wait(&mdev->misc_wait, &wait);
 854 }
 855
 856 /* called within req_lock and rcu_read_lock() */
 857 static void maybe_pull_ahead(struct drbd_conf *mdev)
 858 {
 859         struct drbd_tconn *tconn = mdev->tconn;
 860         struct net_conf *nc;
 861         bool congested = false;
 862         enum drbd_on_congestion on_congestion;
 863
 864         nc = rcu_dereference(tconn->net_conf);
 865         on_congestion = nc ? nc->on_congestion : OC_BLOCK;
 866         if (on_congestion == OC_BLOCK ||
 867             tconn->agreed_pro_version < 96)
 868                 return;
 869
 870         /* If I don't even have good local storage, we can not reasonably try
 871          * to pull ahead of the peer. We also need the local reference to make
 872          * sure mdev->act_log is there.
 873          */
 874         if (!get_ldev_if_state(mdev, D_UP_TO_DATE))
 875                 return;
 876
 877         if (nc->cong_fill &&
 878             atomic_read(&mdev->ap_in_flight) >= nc->cong_fill) {
 879                 dev_info(DEV, "Congestion-fill threshold reached\n");
 880                 congested = true;
 881         }
 882
 883         if (mdev->act_log->used >= nc->cong_extents) {
 884                 dev_info(DEV, "Congestion-extents threshold reached\n");
 885                 congested = true;
 886         }
 887
 888         if (congested) {
 889                 /* start a new epoch for non-mirrored writes */
 890                 start_new_tl_epoch(mdev->tconn);
 891
 892                 if (on_congestion == OC_PULL_AHEAD)
 893                         _drbd_set_state(_NS(mdev, conn, C_AHEAD), 0, NULL);
 894                 else  /*nc->on_congestion == OC_DISCONNECT */
 895                         _drbd_set_state(_NS(mdev, conn, C_DISCONNECTING), 0, NULL);
 896         }
 897         put_ldev(mdev);
 898 }
 899
 900 /* If this returns false, and req->private_bio is still set,
 901  * this should be submitted locally.
 902  *
 903  * If it returns false, but req->private_bio is not set,
 904  * we do not have access to good data :(
 905  *
 906  * Otherwise, this destroys req->private_bio, if any,
 907  * and returns true.
 908  */
 909 static bool do_remote_read(struct drbd_request *req)
 910 {
 911         struct drbd_conf *mdev = req->w.mdev;
 912         enum drbd_read_balancing rbm;
 913
 914         if (req->private_bio) {
 915                 if (!drbd_may_do_local_read(mdev,
 916                                         req->i.sector, req->i.size)) {
 917                         bio_put(req->private_bio);
 918                         req->private_bio = NULL;
 919                         put_ldev(mdev);
 920                 }
 921         }
 922
 923         if (mdev->state.pdsk != D_UP_TO_DATE)
 924                 return false;
 925
 926         if (req->private_bio == NULL)
 927                 return true;
 928
 929         /* TODO: improve read balancing decisions, take into account drbd
 930          * protocol, pending requests etc. */
 931
 932         rcu_read_lock();
 933         rbm = rcu_dereference(mdev->ldev->disk_conf)->read_balancing;
 934         rcu_read_unlock();
 935
 936         if (rbm == RB_PREFER_LOCAL && req->private_bio)
 937                 return false; /* submit locally */
 938
 939         if (remote_due_to_read_balancing(mdev, req->i.sector, rbm)) {
 940                 if (req->private_bio) {
 941                         bio_put(req->private_bio);
 942                         req->private_bio = NULL;
 943                         put_ldev(mdev);
 944                 }
 945                 return true;
 946         }
 947
 948         return false;
 949 }
 950
 951 /* returns number of connections (== 1, for drbd 8.4)
 952  * expected to actually write this data,
 953  * which does NOT include those that we are L_AHEAD for. */
 954 static int drbd_process_write_request(struct drbd_request *req)
 955 {
 956         struct drbd_conf *mdev = req->w.mdev;
 957         int remote, send_oos;
 958
 959         rcu_read_lock();
 960         remote = drbd_should_do_remote(mdev->state);
 961         if (remote) {
 962                 maybe_pull_ahead(mdev);
 963                 remote = drbd_should_do_remote(mdev->state);
 964         }
 965         send_oos = drbd_should_send_out_of_sync(mdev->state);
 966         rcu_read_unlock();
 967
 968         /* Need to replicate writes.  Unless it is an empty flush,
 969          * which is better mapped to a DRBD P_BARRIER packet,
 970          * also for drbd wire protocol compatibility reasons.
 971          * If this was a flush, just start a new epoch.
 972          * Unless the current epoch was empty anyways, or we are not currently
 973          * replicating, in which case there is no point. */
 974         if (unlikely(req->i.size == 0)) {
 975                 /* The only size==0 bios we expect are empty flushes. */
 976                 D_ASSERT(req->master_bio->bi_rw & REQ_FLUSH);
 977                 if (remote)
 978                         start_new_tl_epoch(mdev->tconn);
 979                 return 0;
 980         }
 981
 982         if (!remote && !send_oos)
 983                 return 0;
 984
 985         D_ASSERT(!(remote && send_oos));
 986
 987         if (remote) {
 988                 _req_mod(req, TO_BE_SENT);
 989                 _req_mod(req, QUEUE_FOR_NET_WRITE);
 990         } else if (drbd_set_out_of_sync(mdev, req->i.sector, req->i.size))
 991                 _req_mod(req, QUEUE_FOR_SEND_OOS);
 992
 993         return remote;
 994 }
 995
 996 static void
 997 drbd_submit_req_private_bio(struct drbd_request *req)
 998 {
 999         struct drbd_conf *mdev = req->w.mdev;
1000         struct bio *bio = req->private_bio;
1001         const int rw = bio_rw(bio);
1002
1003         bio->bi_bdev = mdev->ldev->backing_bdev;
1004
1005         /* State may have changed since we grabbed our reference on the
1006          * ->ldev member. Double check, and short-circuit to endio.
1007          * In case the last activity log transaction failed to get on
1008          * stable storage, and this is a WRITE, we may not even submit
1009          * this bio. */
1010         if (get_ldev(mdev)) {
1011                 if (drbd_insert_fault(mdev,
1012                                       rw == WRITE ? DRBD_FAULT_DT_WR
1013                                     : rw == READ  ? DRBD_FAULT_DT_RD
1014                                     :               DRBD_FAULT_DT_RA))
1015                         bio_endio(bio, -EIO);
1016                 else
1017                         generic_make_request(bio);
1018                 put_ldev(mdev);
1019         } else
1020                 bio_endio(bio, -EIO);
1021 }
1022
1023 void __drbd_make_request(struct drbd_conf *mdev, struct bio *bio, unsigned long start_time)
1024 {
1025         const int rw = bio_rw(bio);
1026         struct bio_and_error m = { NULL, };
1027         struct drbd_request *req;
1028         bool no_remote = false;
1029
1030         /* allocate outside of all locks; */
1031         req = drbd_req_new(mdev, bio);
1032         if (!req) {
1033                 dec_ap_bio(mdev);
1034                 /* only pass the error to the upper layers.
1035                  * if user cannot handle io errors, that's not our business. */
1036                 dev_err(DEV, "could not kmalloc() req\n");
1037                 bio_endio(bio, -ENOMEM);
1038                 return;
1039         }
1040         req->start_time = start_time;
1041
1042         if (!get_ldev(mdev)) {
1043                 bio_put(req->private_bio);
1044                 req->private_bio = NULL;
1045         }
1046
1047         /* For WRITES going to the local disk, grab a reference on the target
1048          * extent.  This waits for any resync activity in the corresponding
1049          * resync extent to finish, and, if necessary, pulls in the target
1050          * extent into the activity log, which involves further disk io because
1051          * of transactional on-disk meta data updates.
1052          * Empty flushes don't need to go into the activity log, they can only
1053          * flush data for pending writes which are already in there. */
1054         if (rw == WRITE && req->private_bio && req->i.size
1055         && !test_bit(AL_SUSPENDED, &mdev->flags)) {
1056                 req->rq_state |= RQ_IN_ACT_LOG;
1057                 drbd_al_begin_io(mdev, &req->i);
1058         }
1059
1060         spin_lock_irq(&mdev->tconn->req_lock);
1061         if (rw == WRITE) {
1062                 /* This may temporarily give up the req_lock,
1063                  * but will re-aquire it before it returns here.
1064                  * Needs to be before the check on drbd_suspended() */
1065                 complete_conflicting_writes(req);
1066         }
1067
1068         /* no more giving up req_lock from now on! */
1069
1070         if (drbd_suspended(mdev)) {
1071                 /* push back and retry: */
1072                 req->rq_state |= RQ_POSTPONED;
1073                 if (req->private_bio) {
1074                         bio_put(req->private_bio);
1075                         req->private_bio = NULL;
1076                         put_ldev(mdev);
1077                 }
1078                 goto out;
1079         }
1080
1081         /* Update disk stats */
1082         _drbd_start_io_acct(mdev, req, bio);
1083
1084         /* We fail READ/READA early, if we can not serve it.
1085          * We must do this before req is registered on any lists.
1086          * Otherwise, drbd_req_complete() will queue failed READ for retry. */
1087         if (rw != WRITE) {
1088                 if (!do_remote_read(req) && !req->private_bio)
1089                         goto nodata;
1090         }
1091
1092         /* which transfer log epoch does this belong to? */
1093         req->epoch = atomic_read(&mdev->tconn->current_tle_nr);
1094
1095         /* no point in adding empty flushes to the transfer log,
1096          * they are mapped to drbd barriers already. */
1097         if (likely(req->i.size!=0)) {
1098                 if (rw == WRITE)
1099                         mdev->tconn->current_tle_writes++;
1100
1101                 list_add_tail(&req->tl_requests, &mdev->tconn->transfer_log);
1102         }
1103
1104         if (rw == WRITE) {
1105                 if (!drbd_process_write_request(req))
1106                         no_remote = true;
1107         } else {
1108                 /* We either have a private_bio, or we can read from remote.
1109                  * Otherwise we had done the goto nodata above. */
1110                 if (req->private_bio == NULL) {
1111                         _req_mod(req, TO_BE_SENT);
1112                         _req_mod(req, QUEUE_FOR_NET_READ);
1113                 } else
1114                         no_remote = true;
1115         }
1116
1117         if (req->private_bio) {
1118                 /* needs to be marked within the same spinlock */
1119                 _req_mod(req, TO_BE_SUBMITTED);
1120                 /* but we need to give up the spinlock to submit */
1121                 spin_unlock_irq(&mdev->tconn->req_lock);
1122                 drbd_submit_req_private_bio(req);
1123                 spin_lock_irq(&mdev->tconn->req_lock);
1124         } else if (no_remote) {
1125 nodata:
1126                 if (__ratelimit(&drbd_ratelimit_state))
1127                         dev_err(DEV, "IO ERROR: neither local nor remote data, sector %llu+%u\n",
1128                                         (unsigned long long)req->i.sector, req->i.size >> 9);
1129                 /* A write may have been queued for send_oos, however.
1130                  * So we can not simply free it, we must go through drbd_req_put_completion_ref() */
1131         }
1132
1133 out:
1134         if (drbd_req_put_completion_ref(req, &m, 1))
1135                 kref_put(&req->kref, drbd_req_destroy);
1136         spin_unlock_irq(&mdev->tconn->req_lock);
1137
1138         if (m.bio)
1139                 complete_master_bio(mdev, &m);
1140         return;
1141 }
1142
1143 int drbd_make_request(struct request_queue *q, struct bio *bio)
1144 {
1145         struct drbd_conf *mdev = (struct drbd_conf *) q->queuedata;
1146         unsigned long start_time;
1147
1148         start_time = jiffies;
1149
1150         /*
1151          * what we "blindly" assume:
1152          */
1153         D_ASSERT(IS_ALIGNED(bio->bi_size, 512));
1154
1155         inc_ap_bio(mdev);
1156         __drbd_make_request(mdev, bio, start_time);
1157
1158         return 0;
1159 }
1160
1161 /* This is called by bio_add_page().
1162  *
1163  * q->max_hw_sectors and other global limits are already enforced there.
1164  *
1165  * We need to call down to our lower level device,
1166  * in case it has special restrictions.
1167  *
1168  * We also may need to enforce configured max-bio-bvecs limits.
1169  *
1170  * As long as the BIO is empty we have to allow at least one bvec,
1171  * regardless of size and offset, so no need to ask lower levels.
1172  */
1173 int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec)
1174 {
1175         struct drbd_conf *mdev = (struct drbd_conf *) q->queuedata;
1176         unsigned int bio_size = bvm->bi_size;
1177         int limit = DRBD_MAX_BIO_SIZE;
1178         int backing_limit;
1179
1180         if (bio_size && get_ldev(mdev)) {
1181                 struct request_queue * const b =
1182                         mdev->ldev->backing_bdev->bd_disk->queue;
1183                 if (b->merge_bvec_fn) {
1184                         backing_limit = b->merge_bvec_fn(b, bvm, bvec);
1185                         limit = min(limit, backing_limit);
1186                 }
1187                 put_ldev(mdev);
1188         }
1189         return limit;
1190 }
1191
1192 struct drbd_request *find_oldest_request(struct drbd_tconn *tconn)
1193 {
1194         /* Walk the transfer log,
1195          * and find the oldest not yet completed request */
1196         struct drbd_request *r;
1197         list_for_each_entry(r, &tconn->transfer_log, tl_requests) {
1198                 if (atomic_read(&r->completion_ref))
1199                         return r;
1200         }
1201         return NULL;
1202 }
1203
1204 void request_timer_fn(unsigned long data)
1205 {
1206         struct drbd_conf *mdev = (struct drbd_conf *) data;
1207         struct drbd_tconn *tconn = mdev->tconn;
1208         struct drbd_request *req; /* oldest request */
1209         struct net_conf *nc;
1210         unsigned long ent = 0, dt = 0, et, nt; /* effective timeout = ko_count * timeout */
1211         unsigned long now;
1212
1213         rcu_read_lock();
1214         nc = rcu_dereference(tconn->net_conf);
1215         if (nc && mdev->state.conn >= C_WF_REPORT_PARAMS)
1216                 ent = nc->timeout * HZ/10 * nc->ko_count;
1217
1218         if (get_ldev(mdev)) { /* implicit state.disk >= D_INCONSISTENT */
1219                 dt = rcu_dereference(mdev->ldev->disk_conf)->disk_timeout * HZ / 10;
1220                 put_ldev(mdev);
1221         }
1222         rcu_read_unlock();
1223
1224         et = min_not_zero(dt, ent);
1225
1226         if (!et)
1227                 return; /* Recurring timer stopped */
1228
1229         now = jiffies;
1230
1231         spin_lock_irq(&tconn->req_lock);
1232         req = find_oldest_request(tconn);
1233         if (!req) {
1234                 spin_unlock_irq(&tconn->req_lock);
1235                 mod_timer(&mdev->request_timer, now + et);
1236                 return;
1237         }
1238
1239         /* The request is considered timed out, if
1240          * - we have some effective timeout from the configuration,
1241          *   with above state restrictions applied,
1242          * - the oldest request is waiting for a response from the network
1243          *   resp. the local disk,
1244          * - the oldest request is in fact older than the effective timeout,
1245          * - the connection was established (resp. disk was attached)
1246          *   for longer than the timeout already.
1247          * Note that for 32bit jiffies and very stable connections/disks,
1248          * we may have a wrap around, which is catched by
1249          *   !time_in_range(now, last_..._jif, last_..._jif + timeout).
1250          *
1251          * Side effect: once per 32bit wrap-around interval, which means every
1252          * ~198 days with 250 HZ, we have a window where the timeout would need
1253          * to expire twice (worst case) to become effective. Good enough.
1254          */
1255         if (ent && req->rq_state & RQ_NET_PENDING &&
1256                  time_after(now, req->start_time + ent) &&
1257                 !time_in_range(now, tconn->last_reconnect_jif, tconn->last_reconnect_jif + ent)) {
1258                 dev_warn(DEV, "Remote failed to finish a request within ko-count * timeout\n");
1259                 _drbd_set_state(_NS(mdev, conn, C_TIMEOUT), CS_VERBOSE | CS_HARD, NULL);
1260         }
1261         if (dt && req->rq_state & RQ_LOCAL_PENDING && req->w.mdev == mdev &&
1262                  time_after(now, req->start_time + dt) &&
1263                 !time_in_range(now, mdev->last_reattach_jif, mdev->last_reattach_jif + dt)) {
1264                 dev_warn(DEV, "Local backing device failed to meet the disk-timeout\n");
1265                 __drbd_chk_io_error(mdev, DRBD_FORCE_DETACH);
1266         }
1267         nt = (time_after(now, req->start_time + et) ? now : req->start_time) + et;
1268         spin_unlock_irq(&tconn->req_lock);
1269         mod_timer(&mdev->request_timer, nt);
1270 }