drivers/block/drbd/drbd_worker.c

   1 /*
   2    drbd_worker.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 #include <linux/drbd.h>
  28 #include <linux/sched.h>
  29 #include <linux/smp_lock.h>
  30 #include <linux/wait.h>
  31 #include <linux/mm.h>
  32 #include <linux/memcontrol.h>
  33 #include <linux/mm_inline.h>
  34 #include <linux/slab.h>
  35 #include <linux/random.h>
  36 #include <linux/string.h>
  37 #include <linux/scatterlist.h>
  38
  39 #include "drbd_int.h"
  40 #include "drbd_req.h"
  41
  42 #define SLEEP_TIME (HZ/10)
  43
  44 static int w_make_ov_request(struct drbd_conf *mdev, struct drbd_work *w, int cancel);
  45
  46
  47
  48 /* defined here:
  49    drbd_md_io_complete
  50    drbd_endio_sec
  51    drbd_endio_pri
  52
  53  * more endio handlers:
  54    atodb_endio in drbd_actlog.c
  55    drbd_bm_async_io_complete in drbd_bitmap.c
  56
  57  * For all these callbacks, note the following:
  58  * The callbacks will be called in irq context by the IDE drivers,
  59  * and in Softirqs/Tasklets/BH context by the SCSI drivers.
  60  * Try to get the locking right :)
  61  *
  62  */
  63
  64
  65 /* About the global_state_lock
  66    Each state transition on an device holds a read lock. In case we have
  67    to evaluate the sync after dependencies, we grab a write lock, because
  68    we need stable states on all devices for that.  */
  69 rwlock_t global_state_lock;
  70
  71 /* used for synchronous meta data and bitmap IO
  72  * submitted by drbd_md_sync_page_io()
  73  */
  74 void drbd_md_io_complete(struct bio *bio, int error)
  75 {
  76         struct drbd_md_io *md_io;
  77
  78         md_io = (struct drbd_md_io *)bio->bi_private;
  79         md_io->error = error;
  80
  81         complete(&md_io->event);
  82 }
  83
  84 /* reads on behalf of the partner,
  85  * "submitted" by the receiver
  86  */
  87 void drbd_endio_read_sec_final(struct drbd_epoch_entry *e) __releases(local)
  88 {
  89         unsigned long flags = 0;
  90         struct drbd_conf *mdev = e->mdev;
  91
  92         D_ASSERT(e->block_id != ID_VACANT);
  93
  94         spin_lock_irqsave(&mdev->req_lock, flags);
  95         mdev->read_cnt += e->size >> 9;
  96         list_del(&e->w.list);
  97         if (list_empty(&mdev->read_ee))
  98                 wake_up(&mdev->ee_wait);
  99         if (test_bit(__EE_WAS_ERROR, &e->flags))
 100                 __drbd_chk_io_error(mdev, FALSE);
 101         spin_unlock_irqrestore(&mdev->req_lock, flags);
 102
 103         drbd_queue_work(&mdev->data.work, &e->w);
 104         put_ldev(mdev);
 105 }
 106
 107 static int is_failed_barrier(int ee_flags)
 108 {
 109         return (ee_flags & (EE_IS_BARRIER|EE_WAS_ERROR|EE_RESUBMITTED))
 110                         == (EE_IS_BARRIER|EE_WAS_ERROR);
 111 }
 112
 113 /* writes on behalf of the partner, or resync writes,
 114  * "submitted" by the receiver, final stage.  */
 115 static void drbd_endio_write_sec_final(struct drbd_epoch_entry *e) __releases(local)
 116 {
 117         unsigned long flags = 0;
 118         struct drbd_conf *mdev = e->mdev;
 119         sector_t e_sector;
 120         int do_wake;
 121         int is_syncer_req;
 122         int do_al_complete_io;
 123
 124         /* if this is a failed barrier request, disable use of barriers,
 125          * and schedule for resubmission */
 126         if (is_failed_barrier(e->flags)) {
 127                 drbd_bump_write_ordering(mdev, WO_bdev_flush);
 128                 spin_lock_irqsave(&mdev->req_lock, flags);
 129                 list_del(&e->w.list);
 130                 e->flags = (e->flags & ~EE_WAS_ERROR) | EE_RESUBMITTED;
 131                 e->w.cb = w_e_reissue;
 132                 /* put_ldev actually happens below, once we come here again. */
 133                 __release(local);
 134                 spin_unlock_irqrestore(&mdev->req_lock, flags);
 135                 drbd_queue_work(&mdev->data.work, &e->w);
 136                 return;
 137         }
 138
 139         D_ASSERT(e->block_id != ID_VACANT);
 140
 141         /* after we moved e to done_ee,
 142          * we may no longer access it,
 143          * it may be freed/reused already!
 144          * (as soon as we release the req_lock) */
 145         e_sector = e->sector;
 146         do_al_complete_io = e->flags & EE_CALL_AL_COMPLETE_IO;
 147         is_syncer_req = is_syncer_block_id(e->block_id);
 148
 149         spin_lock_irqsave(&mdev->req_lock, flags);
 150         mdev->writ_cnt += e->size >> 9;
 151         list_del(&e->w.list); /* has been on active_ee or sync_ee */
 152         list_add_tail(&e->w.list, &mdev->done_ee);
 153
 154         /* No hlist_del_init(&e->colision) here, we did not send the Ack yet,
 155          * neither did we wake possibly waiting conflicting requests.
 156          * done from "drbd_process_done_ee" within the appropriate w.cb
 157          * (e_end_block/e_end_resync_block) or from _drbd_clear_done_ee */
 158
 159         do_wake = is_syncer_req
 160                 ? list_empty(&mdev->sync_ee)
 161                 : list_empty(&mdev->active_ee);
 162
 163         if (test_bit(__EE_WAS_ERROR, &e->flags))
 164                 __drbd_chk_io_error(mdev, FALSE);
 165         spin_unlock_irqrestore(&mdev->req_lock, flags);
 166
 167         if (is_syncer_req)
 168                 drbd_rs_complete_io(mdev, e_sector);
 169
 170         if (do_wake)
 171                 wake_up(&mdev->ee_wait);
 172
 173         if (do_al_complete_io)
 174                 drbd_al_complete_io(mdev, e_sector);
 175
 176         wake_asender(mdev);
 177         put_ldev(mdev);
 178 }
 179
 180 /* writes on behalf of the partner, or resync writes,
 181  * "submitted" by the receiver.
 182  */
 183 void drbd_endio_sec(struct bio *bio, int error)
 184 {
 185         struct drbd_epoch_entry *e = bio->bi_private;
 186         struct drbd_conf *mdev = e->mdev;
 187         int uptodate = bio_flagged(bio, BIO_UPTODATE);
 188         int is_write = bio_data_dir(bio) == WRITE;
 189
 190         if (error)
 191                 dev_warn(DEV, "%s: error=%d s=%llus\n",
 192                                 is_write ? "write" : "read", error,
 193                                 (unsigned long long)e->sector);
 194         if (!error && !uptodate) {
 195                 dev_warn(DEV, "%s: setting error to -EIO s=%llus\n",
 196                                 is_write ? "write" : "read",
 197                                 (unsigned long long)e->sector);
 198                 /* strange behavior of some lower level drivers...
 199                  * fail the request by clearing the uptodate flag,
 200                  * but do not return any error?! */
 201                 error = -EIO;
 202         }
 203
 204         if (error)
 205                 set_bit(__EE_WAS_ERROR, &e->flags);
 206
 207         bio_put(bio); /* no need for the bio anymore */
 208         if (atomic_dec_and_test(&e->pending_bios)) {
 209                 if (is_write)
 210                         drbd_endio_write_sec_final(e);
 211                 else
 212                         drbd_endio_read_sec_final(e);
 213         }
 214 }
 215
 216 /* read, readA or write requests on R_PRIMARY coming from drbd_make_request
 217  */
 218 void drbd_endio_pri(struct bio *bio, int error)
 219 {
 220         unsigned long flags;
 221         struct drbd_request *req = bio->bi_private;
 222         struct drbd_conf *mdev = req->mdev;
 223         struct bio_and_error m;
 224         enum drbd_req_event what;
 225         int uptodate = bio_flagged(bio, BIO_UPTODATE);
 226
 227         if (!error && !uptodate) {
 228                 dev_warn(DEV, "p %s: setting error to -EIO\n",
 229                          bio_data_dir(bio) == WRITE ? "write" : "read");
 230                 /* strange behavior of some lower level drivers...
 231                  * fail the request by clearing the uptodate flag,
 232                  * but do not return any error?! */
 233                 error = -EIO;
 234         }
 235
 236         /* to avoid recursion in __req_mod */
 237         if (unlikely(error)) {
 238                 what = (bio_data_dir(bio) == WRITE)
 239                         ? write_completed_with_error
 240                         : (bio_rw(bio) == READ)
 241                           ? read_completed_with_error
 242                           : read_ahead_completed_with_error;
 243         } else
 244                 what = completed_ok;
 245
 246         bio_put(req->private_bio);
 247         req->private_bio = ERR_PTR(error);
 248
 249         spin_lock_irqsave(&mdev->req_lock, flags);
 250         __req_mod(req, what, &m);
 251         spin_unlock_irqrestore(&mdev->req_lock, flags);
 252
 253         if (m.bio)
 254                 complete_master_bio(mdev, &m);
 255 }
 256
 257 int w_read_retry_remote(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 258 {
 259         struct drbd_request *req = container_of(w, struct drbd_request, w);
 260
 261         /* We should not detach for read io-error,
 262          * but try to WRITE the P_DATA_REPLY to the failed location,
 263          * to give the disk the chance to relocate that block */
 264
 265         spin_lock_irq(&mdev->req_lock);
 266         if (cancel || mdev->state.pdsk != D_UP_TO_DATE) {
 267                 _req_mod(req, read_retry_remote_canceled);
 268                 spin_unlock_irq(&mdev->req_lock);
 269                 return 1;
 270         }
 271         spin_unlock_irq(&mdev->req_lock);
 272
 273         return w_send_read_req(mdev, w, 0);
 274 }
 275
 276 int w_resync_inactive(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 277 {
 278         ERR_IF(cancel) return 1;
 279         dev_err(DEV, "resync inactive, but callback triggered??\n");
 280         return 1; /* Simply ignore this! */
 281 }
 282
 283 void drbd_csum_ee(struct drbd_conf *mdev, struct crypto_hash *tfm, struct drbd_epoch_entry *e, void *digest)
 284 {
 285         struct hash_desc desc;
 286         struct scatterlist sg;
 287         struct page *page = e->pages;
 288         struct page *tmp;
 289         unsigned len;
 290
 291         desc.tfm = tfm;
 292         desc.flags = 0;
 293
 294         sg_init_table(&sg, 1);
 295         crypto_hash_init(&desc);
 296
 297         while ((tmp = page_chain_next(page))) {
 298                 /* all but the last page will be fully used */
 299                 sg_set_page(&sg, page, PAGE_SIZE, 0);
 300                 crypto_hash_update(&desc, &sg, sg.length);
 301                 page = tmp;
 302         }
 303         /* and now the last, possibly only partially used page */
 304         len = e->size & (PAGE_SIZE - 1);
 305         sg_set_page(&sg, page, len ?: PAGE_SIZE, 0);
 306         crypto_hash_update(&desc, &sg, sg.length);
 307         crypto_hash_final(&desc, digest);
 308 }
 309
 310 void drbd_csum_bio(struct drbd_conf *mdev, struct crypto_hash *tfm, struct bio *bio, void *digest)
 311 {
 312         struct hash_desc desc;
 313         struct scatterlist sg;
 314         struct bio_vec *bvec;
 315         int i;
 316
 317         desc.tfm = tfm;
 318         desc.flags = 0;
 319
 320         sg_init_table(&sg, 1);
 321         crypto_hash_init(&desc);
 322
 323         __bio_for_each_segment(bvec, bio, i, 0) {
 324                 sg_set_page(&sg, bvec->bv_page, bvec->bv_len, bvec->bv_offset);
 325                 crypto_hash_update(&desc, &sg, sg.length);
 326         }
 327         crypto_hash_final(&desc, digest);
 328 }
 329
 330 static int w_e_send_csum(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 331 {
 332         struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w);
 333         int digest_size;
 334         void *digest;
 335         int ok;
 336
 337         D_ASSERT(e->block_id == DRBD_MAGIC + 0xbeef);
 338
 339         if (unlikely(cancel)) {
 340                 drbd_free_ee(mdev, e);
 341                 return 1;
 342         }
 343
 344         if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 345                 digest_size = crypto_hash_digestsize(mdev->csums_tfm);
 346                 digest = kmalloc(digest_size, GFP_NOIO);
 347                 if (digest) {
 348                         drbd_csum_ee(mdev, mdev->csums_tfm, e, digest);
 349
 350                         inc_rs_pending(mdev);
 351                         ok = drbd_send_drequest_csum(mdev,
 352                                                      e->sector,
 353                                                      e->size,
 354                                                      digest,
 355                                                      digest_size,
 356                                                      P_CSUM_RS_REQUEST);
 357                         kfree(digest);
 358                 } else {
 359                         dev_err(DEV, "kmalloc() of digest failed.\n");
 360                         ok = 0;
 361                 }
 362         } else
 363                 ok = 1;
 364
 365         drbd_free_ee(mdev, e);
 366
 367         if (unlikely(!ok))
 368                 dev_err(DEV, "drbd_send_drequest(..., csum) failed\n");
 369         return ok;
 370 }
 371
 372 #define GFP_TRY (__GFP_HIGHMEM | __GFP_NOWARN)
 373
 374 static int read_for_csum(struct drbd_conf *mdev, sector_t sector, int size)
 375 {
 376         struct drbd_epoch_entry *e;
 377
 378         if (!get_ldev(mdev))
 379                 return 0;
 380
 381         /* GFP_TRY, because if there is no memory available right now, this may
 382          * be rescheduled for later. It is "only" background resync, after all. */
 383         e = drbd_alloc_ee(mdev, DRBD_MAGIC+0xbeef, sector, size, GFP_TRY);
 384         if (!e)
 385                 goto fail;
 386
 387         spin_lock_irq(&mdev->req_lock);
 388         list_add(&e->w.list, &mdev->read_ee);
 389         spin_unlock_irq(&mdev->req_lock);
 390
 391         e->w.cb = w_e_send_csum;
 392         if (drbd_submit_ee(mdev, e, READ, DRBD_FAULT_RS_RD) == 0)
 393                 return 1;
 394
 395         drbd_free_ee(mdev, e);
 396 fail:
 397         put_ldev(mdev);
 398         return 2;
 399 }
 400
 401 void resync_timer_fn(unsigned long data)
 402 {
 403         unsigned long flags;
 404         struct drbd_conf *mdev = (struct drbd_conf *) data;
 405         int queue;
 406
 407         spin_lock_irqsave(&mdev->req_lock, flags);
 408
 409         if (likely(!test_and_clear_bit(STOP_SYNC_TIMER, &mdev->flags))) {
 410                 queue = 1;
 411                 if (mdev->state.conn == C_VERIFY_S)
 412                         mdev->resync_work.cb = w_make_ov_request;
 413                 else
 414                         mdev->resync_work.cb = w_make_resync_request;
 415         } else {
 416                 queue = 0;
 417                 mdev->resync_work.cb = w_resync_inactive;
 418         }
 419
 420         spin_unlock_irqrestore(&mdev->req_lock, flags);
 421
 422         /* harmless race: list_empty outside data.work.q_lock */
 423         if (list_empty(&mdev->resync_work.list) && queue)
 424                 drbd_queue_work(&mdev->data.work, &mdev->resync_work);
 425 }
 426
 427 int w_make_resync_request(struct drbd_conf *mdev,
 428                 struct drbd_work *w, int cancel)
 429 {
 430         unsigned long bit;
 431         sector_t sector;
 432         const sector_t capacity = drbd_get_capacity(mdev->this_bdev);
 433         int max_segment_size;
 434         int number, i, size, pe, mx;
 435         int align, queued, sndbuf;
 436
 437         if (unlikely(cancel))
 438                 return 1;
 439
 440         if (unlikely(mdev->state.conn < C_CONNECTED)) {
 441                 dev_err(DEV, "Confused in w_make_resync_request()! cstate < Connected");
 442                 return 0;
 443         }
 444
 445         if (mdev->state.conn != C_SYNC_TARGET)
 446                 dev_err(DEV, "%s in w_make_resync_request\n",
 447                         drbd_conn_str(mdev->state.conn));
 448
 449         if (!get_ldev(mdev)) {
 450                 /* Since we only need to access mdev->rsync a
 451                    get_ldev_if_state(mdev,D_FAILED) would be sufficient, but
 452                    to continue resync with a broken disk makes no sense at
 453                    all */
 454                 dev_err(DEV, "Disk broke down during resync!\n");
 455                 mdev->resync_work.cb = w_resync_inactive;
 456                 return 1;
 457         }
 458
 459         /* starting with drbd 8.3.8, we can handle multi-bio EEs,
 460          * if it should be necessary */
 461         max_segment_size = mdev->agreed_pro_version < 94 ?
 462                 queue_max_segment_size(mdev->rq_queue) : DRBD_MAX_SEGMENT_SIZE;
 463
 464         number = SLEEP_TIME * mdev->sync_conf.rate  / ((BM_BLOCK_SIZE / 1024) * HZ);
 465         pe = atomic_read(&mdev->rs_pending_cnt);
 466
 467         mutex_lock(&mdev->data.mutex);
 468         if (mdev->data.socket)
 469                 mx = mdev->data.socket->sk->sk_rcvbuf / sizeof(struct p_block_req);
 470         else
 471                 mx = 1;
 472         mutex_unlock(&mdev->data.mutex);
 473
 474         /* For resync rates >160MB/sec, allow more pending RS requests */
 475         if (number > mx)
 476                 mx = number;
 477
 478         /* Limit the number of pending RS requests to no more than the peer's receive buffer */
 479         if ((pe + number) > mx) {
 480                 number = mx - pe;
 481         }
 482
 483         for (i = 0; i < number; i++) {
 484                 /* Stop generating RS requests, when half of the send buffer is filled */
 485                 mutex_lock(&mdev->data.mutex);
 486                 if (mdev->data.socket) {
 487                         queued = mdev->data.socket->sk->sk_wmem_queued;
 488                         sndbuf = mdev->data.socket->sk->sk_sndbuf;
 489                 } else {
 490                         queued = 1;
 491                         sndbuf = 0;
 492                 }
 493                 mutex_unlock(&mdev->data.mutex);
 494                 if (queued > sndbuf / 2)
 495                         goto requeue;
 496
 497 next_sector:
 498                 size = BM_BLOCK_SIZE;
 499                 bit  = drbd_bm_find_next(mdev, mdev->bm_resync_fo);
 500
 501                 if (bit == -1UL) {
 502                         mdev->bm_resync_fo = drbd_bm_bits(mdev);
 503                         mdev->resync_work.cb = w_resync_inactive;
 504                         put_ldev(mdev);
 505                         return 1;
 506                 }
 507
 508                 sector = BM_BIT_TO_SECT(bit);
 509
 510                 if (drbd_try_rs_begin_io(mdev, sector)) {
 511                         mdev->bm_resync_fo = bit;
 512                         goto requeue;
 513                 }
 514                 mdev->bm_resync_fo = bit + 1;
 515
 516                 if (unlikely(drbd_bm_test_bit(mdev, bit) == 0)) {
 517                         drbd_rs_complete_io(mdev, sector);
 518                         goto next_sector;
 519                 }
 520
 521 #if DRBD_MAX_SEGMENT_SIZE > BM_BLOCK_SIZE
 522                 /* try to find some adjacent bits.
 523                  * we stop if we have already the maximum req size.
 524                  *
 525                  * Additionally always align bigger requests, in order to
 526                  * be prepared for all stripe sizes of software RAIDs.
 527                  */
 528                 align = 1;
 529                 for (;;) {
 530                         if (size + BM_BLOCK_SIZE > max_segment_size)
 531                                 break;
 532
 533                         /* Be always aligned */
 534                         if (sector & ((1<<(align+3))-1))
 535                                 break;
 536
 537                         /* do not cross extent boundaries */
 538                         if (((bit+1) & BM_BLOCKS_PER_BM_EXT_MASK) == 0)
 539                                 break;
 540                         /* now, is it actually dirty, after all?
 541                          * caution, drbd_bm_test_bit is tri-state for some
 542                          * obscure reason; ( b == 0 ) would get the out-of-band
 543                          * only accidentally right because of the "oddly sized"
 544                          * adjustment below */
 545                         if (drbd_bm_test_bit(mdev, bit+1) != 1)
 546                                 break;
 547                         bit++;
 548                         size += BM_BLOCK_SIZE;
 549                         if ((BM_BLOCK_SIZE << align) <= size)
 550                                 align++;
 551                         i++;
 552                 }
 553                 /* if we merged some,
 554                  * reset the offset to start the next drbd_bm_find_next from */
 555                 if (size > BM_BLOCK_SIZE)
 556                         mdev->bm_resync_fo = bit + 1;
 557 #endif
 558
 559                 /* adjust very last sectors, in case we are oddly sized */
 560                 if (sector + (size>>9) > capacity)
 561                         size = (capacity-sector)<<9;
 562                 if (mdev->agreed_pro_version >= 89 && mdev->csums_tfm) {
 563                         switch (read_for_csum(mdev, sector, size)) {
 564                         case 0: /* Disk failure*/
 565                                 put_ldev(mdev);
 566                                 return 0;
 567                         case 2: /* Allocation failed */
 568                                 drbd_rs_complete_io(mdev, sector);
 569                                 mdev->bm_resync_fo = BM_SECT_TO_BIT(sector);
 570                                 goto requeue;
 571                         /* case 1: everything ok */
 572                         }
 573                 } else {
 574                         inc_rs_pending(mdev);
 575                         if (!drbd_send_drequest(mdev, P_RS_DATA_REQUEST,
 576                                                sector, size, ID_SYNCER)) {
 577                                 dev_err(DEV, "drbd_send_drequest() failed, aborting...\n");
 578                                 dec_rs_pending(mdev);
 579                                 put_ldev(mdev);
 580                                 return 0;
 581                         }
 582                 }
 583         }
 584
 585         if (mdev->bm_resync_fo >= drbd_bm_bits(mdev)) {
 586                 /* last syncer _request_ was sent,
 587                  * but the P_RS_DATA_REPLY not yet received.  sync will end (and
 588                  * next sync group will resume), as soon as we receive the last
 589                  * resync data block, and the last bit is cleared.
 590                  * until then resync "work" is "inactive" ...
 591                  */
 592                 mdev->resync_work.cb = w_resync_inactive;
 593                 put_ldev(mdev);
 594                 return 1;
 595         }
 596
 597  requeue:
 598         mod_timer(&mdev->resync_timer, jiffies + SLEEP_TIME);
 599         put_ldev(mdev);
 600         return 1;
 601 }
 602
 603 static int w_make_ov_request(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 604 {
 605         int number, i, size;
 606         sector_t sector;
 607         const sector_t capacity = drbd_get_capacity(mdev->this_bdev);
 608
 609         if (unlikely(cancel))
 610                 return 1;
 611
 612         if (unlikely(mdev->state.conn < C_CONNECTED)) {
 613                 dev_err(DEV, "Confused in w_make_ov_request()! cstate < Connected");
 614                 return 0;
 615         }
 616
 617         number = SLEEP_TIME*mdev->sync_conf.rate / ((BM_BLOCK_SIZE/1024)*HZ);
 618         if (atomic_read(&mdev->rs_pending_cnt) > number)
 619                 goto requeue;
 620
 621         number -= atomic_read(&mdev->rs_pending_cnt);
 622
 623         sector = mdev->ov_position;
 624         for (i = 0; i < number; i++) {
 625                 if (sector >= capacity) {
 626                         mdev->resync_work.cb = w_resync_inactive;
 627                         return 1;
 628                 }
 629
 630                 size = BM_BLOCK_SIZE;
 631
 632                 if (drbd_try_rs_begin_io(mdev, sector)) {
 633                         mdev->ov_position = sector;
 634                         goto requeue;
 635                 }
 636
 637                 if (sector + (size>>9) > capacity)
 638                         size = (capacity-sector)<<9;
 639
 640                 inc_rs_pending(mdev);
 641                 if (!drbd_send_ov_request(mdev, sector, size)) {
 642                         dec_rs_pending(mdev);
 643                         return 0;
 644                 }
 645                 sector += BM_SECT_PER_BIT;
 646         }
 647         mdev->ov_position = sector;
 648
 649  requeue:
 650         mod_timer(&mdev->resync_timer, jiffies + SLEEP_TIME);
 651         return 1;
 652 }
 653
 654
 655 int w_ov_finished(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 656 {
 657         kfree(w);
 658         ov_oos_print(mdev);
 659         drbd_resync_finished(mdev);
 660
 661         return 1;
 662 }
 663
 664 static int w_resync_finished(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 665 {
 666         kfree(w);
 667
 668         drbd_resync_finished(mdev);
 669
 670         return 1;
 671 }
 672
 673 int drbd_resync_finished(struct drbd_conf *mdev)
 674 {
 675         unsigned long db, dt, dbdt;
 676         unsigned long n_oos;
 677         union drbd_state os, ns;
 678         struct drbd_work *w;
 679         char *khelper_cmd = NULL;
 680
 681         /* Remove all elements from the resync LRU. Since future actions
 682          * might set bits in the (main) bitmap, then the entries in the
 683          * resync LRU would be wrong. */
 684         if (drbd_rs_del_all(mdev)) {
 685                 /* In case this is not possible now, most probably because
 686                  * there are P_RS_DATA_REPLY Packets lingering on the worker's
 687                  * queue (or even the read operations for those packets
 688                  * is not finished by now).   Retry in 100ms. */
 689
 690                 drbd_kick_lo(mdev);
 691                 __set_current_state(TASK_INTERRUPTIBLE);
 692                 schedule_timeout(HZ / 10);
 693                 w = kmalloc(sizeof(struct drbd_work), GFP_ATOMIC);
 694                 if (w) {
 695                         w->cb = w_resync_finished;
 696                         drbd_queue_work(&mdev->data.work, w);
 697                         return 1;
 698                 }
 699                 dev_err(DEV, "Warn failed to drbd_rs_del_all() and to kmalloc(w).\n");
 700         }
 701
 702         dt = (jiffies - mdev->rs_start - mdev->rs_paused) / HZ;
 703         if (dt <= 0)
 704                 dt = 1;
 705         db = mdev->rs_total;
 706         dbdt = Bit2KB(db/dt);
 707         mdev->rs_paused /= HZ;
 708
 709         if (!get_ldev(mdev))
 710                 goto out;
 711
 712         spin_lock_irq(&mdev->req_lock);
 713         os = mdev->state;
 714
 715         /* This protects us against multiple calls (that can happen in the presence
 716            of application IO), and against connectivity loss just before we arrive here. */
 717         if (os.conn <= C_CONNECTED)
 718                 goto out_unlock;
 719
 720         ns = os;
 721         ns.conn = C_CONNECTED;
 722
 723         dev_info(DEV, "%s done (total %lu sec; paused %lu sec; %lu K/sec)\n",
 724              (os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) ?
 725              "Online verify " : "Resync",
 726              dt + mdev->rs_paused, mdev->rs_paused, dbdt);
 727
 728         n_oos = drbd_bm_total_weight(mdev);
 729
 730         if (os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) {
 731                 if (n_oos) {
 732                         dev_alert(DEV, "Online verify found %lu %dk block out of sync!\n",
 733                               n_oos, Bit2KB(1));
 734                         khelper_cmd = "out-of-sync";
 735                 }
 736         } else {
 737                 D_ASSERT((n_oos - mdev->rs_failed) == 0);
 738
 739                 if (os.conn == C_SYNC_TARGET || os.conn == C_PAUSED_SYNC_T)
 740                         khelper_cmd = "after-resync-target";
 741
 742                 if (mdev->csums_tfm && mdev->rs_total) {
 743                         const unsigned long s = mdev->rs_same_csum;
 744                         const unsigned long t = mdev->rs_total;
 745                         const int ratio =
 746                                 (t == 0)     ? 0 :
 747                         (t < 100000) ? ((s*100)/t) : (s/(t/100));
 748                         dev_info(DEV, "%u %% had equal check sums, eliminated: %luK; "
 749                              "transferred %luK total %luK\n",
 750                              ratio,
 751                              Bit2KB(mdev->rs_same_csum),
 752                              Bit2KB(mdev->rs_total - mdev->rs_same_csum),
 753                              Bit2KB(mdev->rs_total));
 754                 }
 755         }
 756
 757         if (mdev->rs_failed) {
 758                 dev_info(DEV, "            %lu failed blocks\n", mdev->rs_failed);
 759
 760                 if (os.conn == C_SYNC_TARGET || os.conn == C_PAUSED_SYNC_T) {
 761                         ns.disk = D_INCONSISTENT;
 762                         ns.pdsk = D_UP_TO_DATE;
 763                 } else {
 764                         ns.disk = D_UP_TO_DATE;
 765                         ns.pdsk = D_INCONSISTENT;
 766                 }
 767         } else {
 768                 ns.disk = D_UP_TO_DATE;
 769                 ns.pdsk = D_UP_TO_DATE;
 770
 771                 if (os.conn == C_SYNC_TARGET || os.conn == C_PAUSED_SYNC_T) {
 772                         if (mdev->p_uuid) {
 773                                 int i;
 774                                 for (i = UI_BITMAP ; i <= UI_HISTORY_END ; i++)
 775                                         _drbd_uuid_set(mdev, i, mdev->p_uuid[i]);
 776                                 drbd_uuid_set(mdev, UI_BITMAP, mdev->ldev->md.uuid[UI_CURRENT]);
 777                                 _drbd_uuid_set(mdev, UI_CURRENT, mdev->p_uuid[UI_CURRENT]);
 778                         } else {
 779                                 dev_err(DEV, "mdev->p_uuid is NULL! BUG\n");
 780                         }
 781                 }
 782
 783                 drbd_uuid_set_bm(mdev, 0UL);
 784
 785                 if (mdev->p_uuid) {
 786                         /* Now the two UUID sets are equal, update what we
 787                          * know of the peer. */
 788                         int i;
 789                         for (i = UI_CURRENT ; i <= UI_HISTORY_END ; i++)
 790                                 mdev->p_uuid[i] = mdev->ldev->md.uuid[i];
 791                 }
 792         }
 793
 794         _drbd_set_state(mdev, ns, CS_VERBOSE, NULL);
 795 out_unlock:
 796         spin_unlock_irq(&mdev->req_lock);
 797         put_ldev(mdev);
 798 out:
 799         mdev->rs_total  = 0;
 800         mdev->rs_failed = 0;
 801         mdev->rs_paused = 0;
 802         mdev->ov_start_sector = 0;
 803
 804         if (test_and_clear_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags)) {
 805                 dev_warn(DEV, "Writing the whole bitmap, due to failed kmalloc\n");
 806                 drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL, "write from resync_finished");
 807         }
 808
 809         if (khelper_cmd)
 810                 drbd_khelper(mdev, khelper_cmd);
 811
 812         return 1;
 813 }
 814
 815 /* helper */
 816 static void move_to_net_ee_or_free(struct drbd_conf *mdev, struct drbd_epoch_entry *e)
 817 {
 818         if (drbd_ee_has_active_page(e)) {
 819                 /* This might happen if sendpage() has not finished */
 820                 spin_lock_irq(&mdev->req_lock);
 821                 list_add_tail(&e->w.list, &mdev->net_ee);
 822                 spin_unlock_irq(&mdev->req_lock);
 823         } else
 824                 drbd_free_ee(mdev, e);
 825 }
 826
 827 /**
 828  * w_e_end_data_req() - Worker callback, to send a P_DATA_REPLY packet in response to a P_DATA_REQUEST
 829  * @mdev:       DRBD device.
 830  * @w:          work object.
 831  * @cancel:     The connection will be closed anyways
 832  */
 833 int w_e_end_data_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 834 {
 835         struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w);
 836         int ok;
 837
 838         if (unlikely(cancel)) {
 839                 drbd_free_ee(mdev, e);
 840                 dec_unacked(mdev);
 841                 return 1;
 842         }
 843
 844         if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 845                 ok = drbd_send_block(mdev, P_DATA_REPLY, e);
 846         } else {
 847                 if (__ratelimit(&drbd_ratelimit_state))
 848                         dev_err(DEV, "Sending NegDReply. sector=%llus.\n",
 849                             (unsigned long long)e->sector);
 850
 851                 ok = drbd_send_ack(mdev, P_NEG_DREPLY, e);
 852         }
 853
 854         dec_unacked(mdev);
 855
 856         move_to_net_ee_or_free(mdev, e);
 857
 858         if (unlikely(!ok))
 859                 dev_err(DEV, "drbd_send_block() failed\n");
 860         return ok;
 861 }
 862
 863 /**
 864  * w_e_end_rsdata_req() - Worker callback to send a P_RS_DATA_REPLY packet in response to a P_RS_DATA_REQUESTRS
 865  * @mdev:       DRBD device.
 866  * @w:          work object.
 867  * @cancel:     The connection will be closed anyways
 868  */
 869 int w_e_end_rsdata_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 870 {
 871         struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w);
 872         int ok;
 873
 874         if (unlikely(cancel)) {
 875                 drbd_free_ee(mdev, e);
 876                 dec_unacked(mdev);
 877                 return 1;
 878         }
 879
 880         if (get_ldev_if_state(mdev, D_FAILED)) {
 881                 drbd_rs_complete_io(mdev, e->sector);
 882                 put_ldev(mdev);
 883         }
 884
 885         if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 886                 if (likely(mdev->state.pdsk >= D_INCONSISTENT)) {
 887                         inc_rs_pending(mdev);
 888                         ok = drbd_send_block(mdev, P_RS_DATA_REPLY, e);
 889                 } else {
 890                         if (__ratelimit(&drbd_ratelimit_state))
 891                                 dev_err(DEV, "Not sending RSDataReply, "
 892                                     "partner DISKLESS!\n");
 893                         ok = 1;
 894                 }
 895         } else {
 896                 if (__ratelimit(&drbd_ratelimit_state))
 897                         dev_err(DEV, "Sending NegRSDReply. sector %llus.\n",
 898                             (unsigned long long)e->sector);
 899
 900                 ok = drbd_send_ack(mdev, P_NEG_RS_DREPLY, e);
 901
 902                 /* update resync data with failure */
 903                 drbd_rs_failed_io(mdev, e->sector, e->size);
 904         }
 905
 906         dec_unacked(mdev);
 907
 908         move_to_net_ee_or_free(mdev, e);
 909
 910         if (unlikely(!ok))
 911                 dev_err(DEV, "drbd_send_block() failed\n");
 912         return ok;
 913 }
 914
 915 int w_e_end_csum_rs_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 916 {
 917         struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w);
 918         struct digest_info *di;
 919         int digest_size;
 920         void *digest = NULL;
 921         int ok, eq = 0;
 922
 923         if (unlikely(cancel)) {
 924                 drbd_free_ee(mdev, e);
 925                 dec_unacked(mdev);
 926                 return 1;
 927         }
 928
 929         drbd_rs_complete_io(mdev, e->sector);
 930
 931         di = (struct digest_info *)(unsigned long)e->block_id;
 932
 933         if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 934                 /* quick hack to try to avoid a race against reconfiguration.
 935                  * a real fix would be much more involved,
 936                  * introducing more locking mechanisms */
 937                 if (mdev->csums_tfm) {
 938                         digest_size = crypto_hash_digestsize(mdev->csums_tfm);
 939                         D_ASSERT(digest_size == di->digest_size);
 940                         digest = kmalloc(digest_size, GFP_NOIO);
 941                 }
 942                 if (digest) {
 943                         drbd_csum_ee(mdev, mdev->csums_tfm, e, digest);
 944                         eq = !memcmp(digest, di->digest, digest_size);
 945                         kfree(digest);
 946                 }
 947
 948                 if (eq) {
 949                         drbd_set_in_sync(mdev, e->sector, e->size);
 950                         /* rs_same_csums unit is BM_BLOCK_SIZE */
 951                         mdev->rs_same_csum += e->size >> BM_BLOCK_SHIFT;
 952                         ok = drbd_send_ack(mdev, P_RS_IS_IN_SYNC, e);
 953                 } else {
 954                         inc_rs_pending(mdev);
 955                         e->block_id = ID_SYNCER;
 956                         ok = drbd_send_block(mdev, P_RS_DATA_REPLY, e);
 957                 }
 958         } else {
 959                 ok = drbd_send_ack(mdev, P_NEG_RS_DREPLY, e);
 960                 if (__ratelimit(&drbd_ratelimit_state))
 961                         dev_err(DEV, "Sending NegDReply. I guess it gets messy.\n");
 962         }
 963
 964         dec_unacked(mdev);
 965
 966         kfree(di);
 967
 968         move_to_net_ee_or_free(mdev, e);
 969
 970         if (unlikely(!ok))
 971                 dev_err(DEV, "drbd_send_block/ack() failed\n");
 972         return ok;
 973 }
 974
 975 int w_e_end_ov_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 976 {
 977         struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w);
 978         int digest_size;
 979         void *digest;
 980         int ok = 1;
 981
 982         if (unlikely(cancel))
 983                 goto out;
 984
 985         if (unlikely((e->flags & EE_WAS_ERROR) != 0))
 986                 goto out;
 987
 988         digest_size = crypto_hash_digestsize(mdev->verify_tfm);
 989         /* FIXME if this allocation fails, online verify will not terminate! */
 990         digest = kmalloc(digest_size, GFP_NOIO);
 991         if (digest) {
 992                 drbd_csum_ee(mdev, mdev->verify_tfm, e, digest);
 993                 inc_rs_pending(mdev);
 994                 ok = drbd_send_drequest_csum(mdev, e->sector, e->size,
 995                                              digest, digest_size, P_OV_REPLY);
 996                 if (!ok)
 997                         dec_rs_pending(mdev);
 998                 kfree(digest);
 999         }
1000
1001 out:
1002         drbd_free_ee(mdev, e);
1003
1004         dec_unacked(mdev);
1005
1006         return ok;
1007 }
1008
1009 void drbd_ov_oos_found(struct drbd_conf *mdev, sector_t sector, int size)
1010 {
1011         if (mdev->ov_last_oos_start + mdev->ov_last_oos_size == sector) {
1012                 mdev->ov_last_oos_size += size>>9;
1013         } else {
1014                 mdev->ov_last_oos_start = sector;
1015                 mdev->ov_last_oos_size = size>>9;
1016         }
1017         drbd_set_out_of_sync(mdev, sector, size);
1018         set_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags);
1019 }
1020
1021 int w_e_end_ov_reply(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
1022 {
1023         struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w);
1024         struct digest_info *di;
1025         int digest_size;
1026         void *digest;
1027         int ok, eq = 0;
1028
1029         if (unlikely(cancel)) {
1030                 drbd_free_ee(mdev, e);
1031                 dec_unacked(mdev);
1032                 return 1;
1033         }
1034
1035         /* after "cancel", because after drbd_disconnect/drbd_rs_cancel_all
1036          * the resync lru has been cleaned up already */
1037         drbd_rs_complete_io(mdev, e->sector);
1038
1039         di = (struct digest_info *)(unsigned long)e->block_id;
1040
1041         if (likely((e->flags & EE_WAS_ERROR) == 0)) {
1042                 digest_size = crypto_hash_digestsize(mdev->verify_tfm);
1043                 digest = kmalloc(digest_size, GFP_NOIO);
1044                 if (digest) {
1045                         drbd_csum_ee(mdev, mdev->verify_tfm, e, digest);
1046
1047                         D_ASSERT(digest_size == di->digest_size);
1048                         eq = !memcmp(digest, di->digest, digest_size);
1049                         kfree(digest);
1050                 }
1051         } else {
1052                 ok = drbd_send_ack(mdev, P_NEG_RS_DREPLY, e);
1053                 if (__ratelimit(&drbd_ratelimit_state))
1054                         dev_err(DEV, "Sending NegDReply. I guess it gets messy.\n");
1055         }
1056
1057         dec_unacked(mdev);
1058
1059         kfree(di);
1060
1061         if (!eq)
1062                 drbd_ov_oos_found(mdev, e->sector, e->size);
1063         else
1064                 ov_oos_print(mdev);
1065
1066         ok = drbd_send_ack_ex(mdev, P_OV_RESULT, e->sector, e->size,
1067                               eq ? ID_IN_SYNC : ID_OUT_OF_SYNC);
1068
1069         drbd_free_ee(mdev, e);
1070
1071         if (--mdev->ov_left == 0) {
1072                 ov_oos_print(mdev);
1073                 drbd_resync_finished(mdev);
1074         }
1075
1076         return ok;
1077 }
1078
1079 int w_prev_work_done(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
1080 {
1081         struct drbd_wq_barrier *b = container_of(w, struct drbd_wq_barrier, w);
1082         complete(&b->done);
1083         return 1;
1084 }
1085
1086 int w_send_barrier(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
1087 {
1088         struct drbd_tl_epoch *b = container_of(w, struct drbd_tl_epoch, w);
1089         struct p_barrier *p = &mdev->data.sbuf.barrier;
1090         int ok = 1;
1091
1092         /* really avoid racing with tl_clear.  w.cb may have been referenced
1093          * just before it was reassigned and re-queued, so double check that.
1094          * actually, this race was harmless, since we only try to send the
1095          * barrier packet here, and otherwise do nothing with the object.
1096          * but compare with the head of w_clear_epoch */
1097         spin_lock_irq(&mdev->req_lock);
1098         if (w->cb != w_send_barrier || mdev->state.conn < C_CONNECTED)
1099                 cancel = 1;
1100         spin_unlock_irq(&mdev->req_lock);
1101         if (cancel)
1102                 return 1;
1103
1104         if (!drbd_get_data_sock(mdev))
1105                 return 0;
1106         p->barrier = b->br_number;
1107         /* inc_ap_pending was done where this was queued.
1108          * dec_ap_pending will be done in got_BarrierAck
1109          * or (on connection loss) in w_clear_epoch.  */
1110         ok = _drbd_send_cmd(mdev, mdev->data.socket, P_BARRIER,
1111                                 (struct p_header *)p, sizeof(*p), 0);
1112         drbd_put_data_sock(mdev);
1113
1114         return ok;
1115 }
1116
1117 int w_send_write_hint(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
1118 {
1119         if (cancel)
1120                 return 1;
1121         return drbd_send_short_cmd(mdev, P_UNPLUG_REMOTE);
1122 }
1123
1124 /**
1125  * w_send_dblock() - Worker callback to send a P_DATA packet in order to mirror a write request
1126  * @mdev:       DRBD device.
1127  * @w:          work object.
1128  * @cancel:     The connection will be closed anyways
1129  */
1130 int w_send_dblock(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
1131 {
1132         struct drbd_request *req = container_of(w, struct drbd_request, w);
1133         int ok;
1134
1135         if (unlikely(cancel)) {
1136                 req_mod(req, send_canceled);
1137                 return 1;
1138         }
1139
1140         ok = drbd_send_dblock(mdev, req);
1141         req_mod(req, ok ? handed_over_to_network : send_failed);
1142
1143         return ok;
1144 }
1145
1146 /**
1147  * w_send_read_req() - Worker callback to send a read request (P_DATA_REQUEST) packet
1148  * @mdev:       DRBD device.
1149  * @w:          work object.
1150  * @cancel:     The connection will be closed anyways
1151  */
1152 int w_send_read_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
1153 {
1154         struct drbd_request *req = container_of(w, struct drbd_request, w);
1155         int ok;
1156
1157         if (unlikely(cancel)) {
1158                 req_mod(req, send_canceled);
1159                 return 1;
1160         }
1161
1162         ok = drbd_send_drequest(mdev, P_DATA_REQUEST, req->sector, req->size,
1163                                 (unsigned long)req);
1164
1165         if (!ok) {
1166                 /* ?? we set C_TIMEOUT or C_BROKEN_PIPE in drbd_send();
1167                  * so this is probably redundant */
1168                 if (mdev->state.conn >= C_CONNECTED)
1169                         drbd_force_state(mdev, NS(conn, C_NETWORK_FAILURE));
1170         }
1171         req_mod(req, ok ? handed_over_to_network : send_failed);
1172
1173         return ok;
1174 }
1175
1176 static int _drbd_may_sync_now(struct drbd_conf *mdev)
1177 {
1178         struct drbd_conf *odev = mdev;
1179
1180         while (1) {
1181                 if (odev->sync_conf.after == -1)
1182                         return 1;
1183                 odev = minor_to_mdev(odev->sync_conf.after);
1184                 ERR_IF(!odev) return 1;
1185                 if ((odev->state.conn >= C_SYNC_SOURCE &&
1186                      odev->state.conn <= C_PAUSED_SYNC_T) ||
1187                     odev->state.aftr_isp || odev->state.peer_isp ||
1188                     odev->state.user_isp)
1189                         return 0;
1190         }
1191 }
1192
1193 /**
1194  * _drbd_pause_after() - Pause resync on all devices that may not resync now
1195  * @mdev:       DRBD device.
1196  *
1197  * Called from process context only (admin command and after_state_ch).
1198  */
1199 static int _drbd_pause_after(struct drbd_conf *mdev)
1200 {
1201         struct drbd_conf *odev;
1202         int i, rv = 0;
1203
1204         for (i = 0; i < minor_count; i++) {
1205                 odev = minor_to_mdev(i);
1206                 if (!odev)
1207                         continue;
1208                 if (odev->state.conn == C_STANDALONE && odev->state.disk == D_DISKLESS)
1209                         continue;
1210                 if (!_drbd_may_sync_now(odev))
1211                         rv |= (__drbd_set_state(_NS(odev, aftr_isp, 1), CS_HARD, NULL)
1212                                != SS_NOTHING_TO_DO);
1213         }
1214
1215         return rv;
1216 }
1217
1218 /**
1219  * _drbd_resume_next() - Resume resync on all devices that may resync now
1220  * @mdev:       DRBD device.
1221  *
1222  * Called from process context only (admin command and worker).
1223  */
1224 static int _drbd_resume_next(struct drbd_conf *mdev)
1225 {
1226         struct drbd_conf *odev;
1227         int i, rv = 0;
1228
1229         for (i = 0; i < minor_count; i++) {
1230                 odev = minor_to_mdev(i);
1231                 if (!odev)
1232                         continue;
1233                 if (odev->state.conn == C_STANDALONE && odev->state.disk == D_DISKLESS)
1234                         continue;
1235                 if (odev->state.aftr_isp) {
1236                         if (_drbd_may_sync_now(odev))
1237                                 rv |= (__drbd_set_state(_NS(odev, aftr_isp, 0),
1238                                                         CS_HARD, NULL)
1239                                        != SS_NOTHING_TO_DO) ;
1240                 }
1241         }
1242         return rv;
1243 }
1244
1245 void resume_next_sg(struct drbd_conf *mdev)
1246 {
1247         write_lock_irq(&global_state_lock);
1248         _drbd_resume_next(mdev);
1249         write_unlock_irq(&global_state_lock);
1250 }
1251
1252 void suspend_other_sg(struct drbd_conf *mdev)
1253 {
1254         write_lock_irq(&global_state_lock);
1255         _drbd_pause_after(mdev);
1256         write_unlock_irq(&global_state_lock);
1257 }
1258
1259 static int sync_after_error(struct drbd_conf *mdev, int o_minor)
1260 {
1261         struct drbd_conf *odev;
1262
1263         if (o_minor == -1)
1264                 return NO_ERROR;
1265         if (o_minor < -1 || minor_to_mdev(o_minor) == NULL)
1266                 return ERR_SYNC_AFTER;
1267
1268         /* check for loops */
1269         odev = minor_to_mdev(o_minor);
1270         while (1) {
1271                 if (odev == mdev)
1272                         return ERR_SYNC_AFTER_CYCLE;
1273
1274                 /* dependency chain ends here, no cycles. */
1275                 if (odev->sync_conf.after == -1)
1276                         return NO_ERROR;
1277
1278                 /* follow the dependency chain */
1279                 odev = minor_to_mdev(odev->sync_conf.after);
1280         }
1281 }
1282
1283 int drbd_alter_sa(struct drbd_conf *mdev, int na)
1284 {
1285         int changes;
1286         int retcode;
1287
1288         write_lock_irq(&global_state_lock);
1289         retcode = sync_after_error(mdev, na);
1290         if (retcode == NO_ERROR) {
1291                 mdev->sync_conf.after = na;
1292                 do {
1293                         changes  = _drbd_pause_after(mdev);
1294                         changes |= _drbd_resume_next(mdev);
1295                 } while (changes);
1296         }
1297         write_unlock_irq(&global_state_lock);
1298         return retcode;
1299 }
1300
1301 static void ping_peer(struct drbd_conf *mdev)
1302 {
1303         clear_bit(GOT_PING_ACK, &mdev->flags);
1304         request_ping(mdev);
1305         wait_event(mdev->misc_wait,
1306                    test_bit(GOT_PING_ACK, &mdev->flags) || mdev->state.conn < C_CONNECTED);
1307 }
1308
1309 /**
1310  * drbd_start_resync() - Start the resync process
1311  * @mdev:       DRBD device.
1312  * @side:       Either C_SYNC_SOURCE or C_SYNC_TARGET
1313  *
1314  * This function might bring you directly into one of the
1315  * C_PAUSED_SYNC_* states.
1316  */
1317 void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side)
1318 {
1319         union drbd_state ns;
1320         int r;
1321
1322         if (mdev->state.conn >= C_SYNC_SOURCE) {
1323                 dev_err(DEV, "Resync already running!\n");
1324                 return;
1325         }
1326
1327         /* In case a previous resync run was aborted by an IO error/detach on the peer. */
1328         drbd_rs_cancel_all(mdev);
1329
1330         if (side == C_SYNC_TARGET) {
1331                 /* Since application IO was locked out during C_WF_BITMAP_T and
1332                    C_WF_SYNC_UUID we are still unmodified. Before going to C_SYNC_TARGET
1333                    we check that we might make the data inconsistent. */
1334                 r = drbd_khelper(mdev, "before-resync-target");
1335                 r = (r >> 8) & 0xff;
1336                 if (r > 0) {
1337                         dev_info(DEV, "before-resync-target handler returned %d, "
1338                              "dropping connection.\n", r);
1339                         drbd_force_state(mdev, NS(conn, C_DISCONNECTING));
1340                         return;
1341                 }
1342         }
1343
1344         drbd_state_lock(mdev);
1345
1346         if (!get_ldev_if_state(mdev, D_NEGOTIATING)) {
1347                 drbd_state_unlock(mdev);
1348                 return;
1349         }
1350
1351         if (side == C_SYNC_TARGET) {
1352                 mdev->bm_resync_fo = 0;
1353         } else /* side == C_SYNC_SOURCE */ {
1354                 u64 uuid;
1355
1356                 get_random_bytes(&uuid, sizeof(u64));
1357                 drbd_uuid_set(mdev, UI_BITMAP, uuid);
1358                 drbd_send_sync_uuid(mdev, uuid);
1359
1360                 D_ASSERT(mdev->state.disk == D_UP_TO_DATE);
1361         }
1362
1363         write_lock_irq(&global_state_lock);
1364         ns = mdev->state;
1365
1366         ns.aftr_isp = !_drbd_may_sync_now(mdev);
1367
1368         ns.conn = side;
1369
1370         if (side == C_SYNC_TARGET)
1371                 ns.disk = D_INCONSISTENT;
1372         else /* side == C_SYNC_SOURCE */
1373                 ns.pdsk = D_INCONSISTENT;
1374
1375         r = __drbd_set_state(mdev, ns, CS_VERBOSE, NULL);
1376         ns = mdev->state;
1377
1378         if (ns.conn < C_CONNECTED)
1379                 r = SS_UNKNOWN_ERROR;
1380
1381         if (r == SS_SUCCESS) {
1382                 mdev->rs_total     =
1383                 mdev->rs_mark_left = drbd_bm_total_weight(mdev);
1384                 mdev->rs_failed    = 0;
1385                 mdev->rs_paused    = 0;
1386                 mdev->rs_start     =
1387                 mdev->rs_mark_time = jiffies;
1388                 mdev->rs_same_csum = 0;
1389                 _drbd_pause_after(mdev);
1390         }
1391         write_unlock_irq(&global_state_lock);
1392         put_ldev(mdev);
1393
1394         if (r == SS_SUCCESS) {
1395                 dev_info(DEV, "Began resync as %s (will sync %lu KB [%lu bits set]).\n",
1396                      drbd_conn_str(ns.conn),
1397                      (unsigned long) mdev->rs_total << (BM_BLOCK_SHIFT-10),
1398                      (unsigned long) mdev->rs_total);
1399
1400                 if (mdev->rs_total == 0) {
1401                         /* Peer still reachable? Beware of failing before-resync-target handlers! */
1402                         ping_peer(mdev);
1403                         drbd_resync_finished(mdev);
1404                 }
1405
1406                 /* ns.conn may already be != mdev->state.conn,
1407                  * we may have been paused in between, or become paused until
1408                  * the timer triggers.
1409                  * No matter, that is handled in resync_timer_fn() */
1410                 if (ns.conn == C_SYNC_TARGET)
1411                         mod_timer(&mdev->resync_timer, jiffies);
1412
1413                 drbd_md_sync(mdev);
1414         }
1415         drbd_state_unlock(mdev);
1416 }
1417
1418 int drbd_worker(struct drbd_thread *thi)
1419 {
1420         struct drbd_conf *mdev = thi->mdev;
1421         struct drbd_work *w = NULL;
1422         LIST_HEAD(work_list);
1423         int intr = 0, i;
1424
1425         sprintf(current->comm, "drbd%d_worker", mdev_to_minor(mdev));
1426
1427         while (get_t_state(thi) == Running) {
1428                 drbd_thread_current_set_cpu(mdev);
1429
1430                 if (down_trylock(&mdev->data.work.s)) {
1431                         mutex_lock(&mdev->data.mutex);
1432                         if (mdev->data.socket && !mdev->net_conf->no_cork)
1433                                 drbd_tcp_uncork(mdev->data.socket);
1434                         mutex_unlock(&mdev->data.mutex);
1435
1436                         intr = down_interruptible(&mdev->data.work.s);
1437
1438                         mutex_lock(&mdev->data.mutex);
1439                         if (mdev->data.socket  && !mdev->net_conf->no_cork)
1440                                 drbd_tcp_cork(mdev->data.socket);
1441                         mutex_unlock(&mdev->data.mutex);
1442                 }
1443
1444                 if (intr) {
1445                         D_ASSERT(intr == -EINTR);
1446                         flush_signals(current);
1447                         ERR_IF (get_t_state(thi) == Running)
1448                                 continue;
1449                         break;
1450                 }
1451
1452                 if (get_t_state(thi) != Running)
1453                         break;
1454                 /* With this break, we have done a down() but not consumed
1455                    the entry from the list. The cleanup code takes care of
1456                    this...   */
1457
1458                 w = NULL;
1459                 spin_lock_irq(&mdev->data.work.q_lock);
1460                 ERR_IF(list_empty(&mdev->data.work.q)) {
1461                         /* something terribly wrong in our logic.
1462                          * we were able to down() the semaphore,
1463                          * but the list is empty... doh.
1464                          *
1465                          * what is the best thing to do now?
1466                          * try again from scratch, restarting the receiver,
1467                          * asender, whatnot? could break even more ugly,
1468                          * e.g. when we are primary, but no good local data.
1469                          *
1470                          * I'll try to get away just starting over this loop.
1471                          */
1472                         spin_unlock_irq(&mdev->data.work.q_lock);
1473                         continue;
1474                 }
1475                 w = list_entry(mdev->data.work.q.next, struct drbd_work, list);
1476                 list_del_init(&w->list);
1477                 spin_unlock_irq(&mdev->data.work.q_lock);
1478
1479                 if (!w->cb(mdev, w, mdev->state.conn < C_CONNECTED)) {
1480                         /* dev_warn(DEV, "worker: a callback failed! \n"); */
1481                         if (mdev->state.conn >= C_CONNECTED)
1482                                 drbd_force_state(mdev,
1483                                                 NS(conn, C_NETWORK_FAILURE));
1484                 }
1485         }
1486         D_ASSERT(test_bit(DEVICE_DYING, &mdev->flags));
1487         D_ASSERT(test_bit(CONFIG_PENDING, &mdev->flags));
1488
1489         spin_lock_irq(&mdev->data.work.q_lock);
1490         i = 0;
1491         while (!list_empty(&mdev->data.work.q)) {
1492                 list_splice_init(&mdev->data.work.q, &work_list);
1493                 spin_unlock_irq(&mdev->data.work.q_lock);
1494
1495                 while (!list_empty(&work_list)) {
1496                         w = list_entry(work_list.next, struct drbd_work, list);
1497                         list_del_init(&w->list);
1498                         w->cb(mdev, w, 1);
1499                         i++; /* dead debugging code */
1500                 }
1501
1502                 spin_lock_irq(&mdev->data.work.q_lock);
1503         }
1504         sema_init(&mdev->data.work.s, 0);
1505         /* DANGEROUS race: if someone did queue his work within the spinlock,
1506          * but up() ed outside the spinlock, we could get an up() on the
1507          * semaphore without corresponding list entry.
1508          * So don't do that.
1509          */
1510         spin_unlock_irq(&mdev->data.work.q_lock);
1511
1512         D_ASSERT(mdev->state.disk == D_DISKLESS && mdev->state.conn == C_STANDALONE);
1513         /* _drbd_set_state only uses stop_nowait.
1514          * wait here for the Exiting receiver. */
1515         drbd_thread_stop(&mdev->receiver);
1516         drbd_mdev_cleanup(mdev);
1517
1518         dev_info(DEV, "worker terminated\n");
1519
1520         clear_bit(DEVICE_DYING, &mdev->flags);
1521         clear_bit(CONFIG_PENDING, &mdev->flags);
1522         wake_up(&mdev->state_wait);
1523
1524         return 0;
1525 }