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