aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/md
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2008-10-10 14:11:47 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2008-10-10 14:11:47 -0400
commitb0af205afb111e17ac8db64c3b9c4f2c332de92a (patch)
tree3999a2ffbd36e9d1cc6ca30e6b9d6280f4e50116 /drivers/md
parent73f6aa4d44ab6157badc456ddfa05b31e58de5f0 (diff)
parent0c2322e4ce144e130c03d813fe92de3798662c5e (diff)
Merge git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm
* git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm: dm: detect lost queue dm: publish dm_vcalloc dm: publish dm_table_unplug_all dm: publish dm_get_mapinfo dm: export struct dm_dev dm crypt: avoid unnecessary wait when splitting bio dm crypt: tidy ctx pending dm crypt: fix async inc_pending dm crypt: move dec_pending on error into write_io_submit dm crypt: remove inc_pending from write_io_submit dm crypt: tidy write loop pending dm crypt: tidy crypt alloc dm crypt: tidy inc pending dm exception store: use chunk_t for_areas dm exception store: introduce area_location function dm raid1: kcopyd should stop on error if errors handled dm mpath: remove is_active from struct dm_path dm mpath: use more error codes Fixed up trivial conflict in drivers/md/dm-mpath.c manually.
Diffstat (limited to 'drivers/md')
-rw-r--r--drivers/md/dm-crypt.c109
-rw-r--r--drivers/md/dm-exception-store.c29
-rw-r--r--drivers/md/dm-ioctl.c4
-rw-r--r--drivers/md/dm-mpath.c35
-rw-r--r--drivers/md/dm-mpath.h2
-rw-r--r--drivers/md/dm-raid1.c4
-rw-r--r--drivers/md/dm-table.c97
-rw-r--r--drivers/md/dm.h10
8 files changed, 173 insertions, 117 deletions
diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 13956437bc81..682ef9e6acd3 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -333,7 +333,6 @@ static void crypt_convert_init(struct crypt_config *cc,
333 ctx->idx_out = bio_out ? bio_out->bi_idx : 0; 333 ctx->idx_out = bio_out ? bio_out->bi_idx : 0;
334 ctx->sector = sector + cc->iv_offset; 334 ctx->sector = sector + cc->iv_offset;
335 init_completion(&ctx->restart); 335 init_completion(&ctx->restart);
336 atomic_set(&ctx->pending, 1);
337} 336}
338 337
339static int crypt_convert_block(struct crypt_config *cc, 338static int crypt_convert_block(struct crypt_config *cc,
@@ -408,6 +407,8 @@ static int crypt_convert(struct crypt_config *cc,
408{ 407{
409 int r; 408 int r;
410 409
410 atomic_set(&ctx->pending, 1);
411
411 while(ctx->idx_in < ctx->bio_in->bi_vcnt && 412 while(ctx->idx_in < ctx->bio_in->bi_vcnt &&
412 ctx->idx_out < ctx->bio_out->bi_vcnt) { 413 ctx->idx_out < ctx->bio_out->bi_vcnt) {
413 414
@@ -456,9 +457,11 @@ static void dm_crypt_bio_destructor(struct bio *bio)
456/* 457/*
457 * Generate a new unfragmented bio with the given size 458 * Generate a new unfragmented bio with the given size
458 * This should never violate the device limitations 459 * This should never violate the device limitations
459 * May return a smaller bio when running out of pages 460 * May return a smaller bio when running out of pages, indicated by
461 * *out_of_pages set to 1.
460 */ 462 */
461static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned size) 463static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned size,
464 unsigned *out_of_pages)
462{ 465{
463 struct crypt_config *cc = io->target->private; 466 struct crypt_config *cc = io->target->private;
464 struct bio *clone; 467 struct bio *clone;
@@ -472,11 +475,14 @@ static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned size)
472 return NULL; 475 return NULL;
473 476
474 clone_init(io, clone); 477 clone_init(io, clone);
478 *out_of_pages = 0;
475 479
476 for (i = 0; i < nr_iovecs; i++) { 480 for (i = 0; i < nr_iovecs; i++) {
477 page = mempool_alloc(cc->page_pool, gfp_mask); 481 page = mempool_alloc(cc->page_pool, gfp_mask);
478 if (!page) 482 if (!page) {
483 *out_of_pages = 1;
479 break; 484 break;
485 }
480 486
481 /* 487 /*
482 * if additional pages cannot be allocated without waiting, 488 * if additional pages cannot be allocated without waiting,
@@ -517,6 +523,27 @@ static void crypt_free_buffer_pages(struct crypt_config *cc, struct bio *clone)
517 } 523 }
518} 524}
519 525
526static struct dm_crypt_io *crypt_io_alloc(struct dm_target *ti,
527 struct bio *bio, sector_t sector)
528{
529 struct crypt_config *cc = ti->private;
530 struct dm_crypt_io *io;
531
532 io = mempool_alloc(cc->io_pool, GFP_NOIO);
533 io->target = ti;
534 io->base_bio = bio;
535 io->sector = sector;
536 io->error = 0;
537 atomic_set(&io->pending, 0);
538
539 return io;
540}
541
542static void crypt_inc_pending(struct dm_crypt_io *io)
543{
544 atomic_inc(&io->pending);
545}
546
520/* 547/*
521 * One of the bios was finished. Check for completion of 548 * One of the bios was finished. Check for completion of
522 * the whole request and correctly clean up the buffer. 549 * the whole request and correctly clean up the buffer.
@@ -591,7 +618,7 @@ static void kcryptd_io_read(struct dm_crypt_io *io)
591 struct bio *base_bio = io->base_bio; 618 struct bio *base_bio = io->base_bio;
592 struct bio *clone; 619 struct bio *clone;
593 620
594 atomic_inc(&io->pending); 621 crypt_inc_pending(io);
595 622
596 /* 623 /*
597 * The block layer might modify the bvec array, so always 624 * The block layer might modify the bvec array, so always
@@ -653,6 +680,7 @@ static void kcryptd_crypt_write_io_submit(struct dm_crypt_io *io,
653 crypt_free_buffer_pages(cc, clone); 680 crypt_free_buffer_pages(cc, clone);
654 bio_put(clone); 681 bio_put(clone);
655 io->error = -EIO; 682 io->error = -EIO;
683 crypt_dec_pending(io);
656 return; 684 return;
657 } 685 }
658 686
@@ -664,28 +692,34 @@ static void kcryptd_crypt_write_io_submit(struct dm_crypt_io *io,
664 692
665 if (async) 693 if (async)
666 kcryptd_queue_io(io); 694 kcryptd_queue_io(io);
667 else { 695 else
668 atomic_inc(&io->pending);
669 generic_make_request(clone); 696 generic_make_request(clone);
670 }
671} 697}
672 698
673static void kcryptd_crypt_write_convert_loop(struct dm_crypt_io *io) 699static void kcryptd_crypt_write_convert(struct dm_crypt_io *io)
674{ 700{
675 struct crypt_config *cc = io->target->private; 701 struct crypt_config *cc = io->target->private;
676 struct bio *clone; 702 struct bio *clone;
703 int crypt_finished;
704 unsigned out_of_pages = 0;
677 unsigned remaining = io->base_bio->bi_size; 705 unsigned remaining = io->base_bio->bi_size;
678 int r; 706 int r;
679 707
680 /* 708 /*
709 * Prevent io from disappearing until this function completes.
710 */
711 crypt_inc_pending(io);
712 crypt_convert_init(cc, &io->ctx, NULL, io->base_bio, io->sector);
713
714 /*
681 * The allocated buffers can be smaller than the whole bio, 715 * The allocated buffers can be smaller than the whole bio,
682 * so repeat the whole process until all the data can be handled. 716 * so repeat the whole process until all the data can be handled.
683 */ 717 */
684 while (remaining) { 718 while (remaining) {
685 clone = crypt_alloc_buffer(io, remaining); 719 clone = crypt_alloc_buffer(io, remaining, &out_of_pages);
686 if (unlikely(!clone)) { 720 if (unlikely(!clone)) {
687 io->error = -ENOMEM; 721 io->error = -ENOMEM;
688 return; 722 break;
689 } 723 }
690 724
691 io->ctx.bio_out = clone; 725 io->ctx.bio_out = clone;
@@ -693,37 +727,32 @@ static void kcryptd_crypt_write_convert_loop(struct dm_crypt_io *io)
693 727
694 remaining -= clone->bi_size; 728 remaining -= clone->bi_size;
695 729
730 crypt_inc_pending(io);
696 r = crypt_convert(cc, &io->ctx); 731 r = crypt_convert(cc, &io->ctx);
732 crypt_finished = atomic_dec_and_test(&io->ctx.pending);
697 733
698 if (atomic_dec_and_test(&io->ctx.pending)) { 734 /* Encryption was already finished, submit io now */
699 /* processed, no running async crypto */ 735 if (crypt_finished) {
700 kcryptd_crypt_write_io_submit(io, r, 0); 736 kcryptd_crypt_write_io_submit(io, r, 0);
701 if (unlikely(r < 0))
702 return;
703 } else
704 atomic_inc(&io->pending);
705 737
706 /* out of memory -> run queues */ 738 /*
707 if (unlikely(remaining)) { 739 * If there was an error, do not try next fragments.
708 /* wait for async crypto then reinitialize pending */ 740 * For async, error is processed in async handler.
709 wait_event(cc->writeq, !atomic_read(&io->ctx.pending)); 741 */
710 atomic_set(&io->ctx.pending, 1); 742 if (unlikely(r < 0))
711 congestion_wait(WRITE, HZ/100); 743 break;
712 } 744 }
713 }
714}
715 745
716static void kcryptd_crypt_write_convert(struct dm_crypt_io *io) 746 /*
717{ 747 * Out of memory -> run queues
718 struct crypt_config *cc = io->target->private; 748 * But don't wait if split was due to the io size restriction
719 749 */
720 /* 750 if (unlikely(out_of_pages))
721 * Prevent io from disappearing until this function completes. 751 congestion_wait(WRITE, HZ/100);
722 */
723 atomic_inc(&io->pending);
724 752
725 crypt_convert_init(cc, &io->ctx, NULL, io->base_bio, io->sector); 753 if (unlikely(remaining))
726 kcryptd_crypt_write_convert_loop(io); 754 wait_event(cc->writeq, !atomic_read(&io->ctx.pending));
755 }
727 756
728 crypt_dec_pending(io); 757 crypt_dec_pending(io);
729} 758}
@@ -741,7 +770,7 @@ static void kcryptd_crypt_read_convert(struct dm_crypt_io *io)
741 struct crypt_config *cc = io->target->private; 770 struct crypt_config *cc = io->target->private;
742 int r = 0; 771 int r = 0;
743 772
744 atomic_inc(&io->pending); 773 crypt_inc_pending(io);
745 774
746 crypt_convert_init(cc, &io->ctx, io->base_bio, io->base_bio, 775 crypt_convert_init(cc, &io->ctx, io->base_bio, io->base_bio,
747 io->sector); 776 io->sector);
@@ -1108,15 +1137,9 @@ static void crypt_dtr(struct dm_target *ti)
1108static int crypt_map(struct dm_target *ti, struct bio *bio, 1137static int crypt_map(struct dm_target *ti, struct bio *bio,
1109 union map_info *map_context) 1138 union map_info *map_context)
1110{ 1139{
1111 struct crypt_config *cc = ti->private;
1112 struct dm_crypt_io *io; 1140 struct dm_crypt_io *io;
1113 1141
1114 io = mempool_alloc(cc->io_pool, GFP_NOIO); 1142 io = crypt_io_alloc(ti, bio, bio->bi_sector - ti->begin);
1115 io->target = ti;
1116 io->base_bio = bio;
1117 io->sector = bio->bi_sector - ti->begin;
1118 io->error = 0;
1119 atomic_set(&io->pending, 0);
1120 1143
1121 if (bio_data_dir(io->base_bio) == READ) 1144 if (bio_data_dir(io->base_bio) == READ)
1122 kcryptd_queue_io(io); 1145 kcryptd_queue_io(io);
diff --git a/drivers/md/dm-exception-store.c b/drivers/md/dm-exception-store.c
index 41f408068a7c..769ab677f8e0 100644
--- a/drivers/md/dm-exception-store.c
+++ b/drivers/md/dm-exception-store.c
@@ -108,12 +108,12 @@ struct pstore {
108 * Used to keep track of which metadata area the data in 108 * Used to keep track of which metadata area the data in
109 * 'chunk' refers to. 109 * 'chunk' refers to.
110 */ 110 */
111 uint32_t current_area; 111 chunk_t current_area;
112 112
113 /* 113 /*
114 * The next free chunk for an exception. 114 * The next free chunk for an exception.
115 */ 115 */
116 uint32_t next_free; 116 chunk_t next_free;
117 117
118 /* 118 /*
119 * The index of next free exception in the current 119 * The index of next free exception in the current
@@ -175,7 +175,7 @@ static void do_metadata(struct work_struct *work)
175/* 175/*
176 * Read or write a chunk aligned and sized block of data from a device. 176 * Read or write a chunk aligned and sized block of data from a device.
177 */ 177 */
178static int chunk_io(struct pstore *ps, uint32_t chunk, int rw, int metadata) 178static int chunk_io(struct pstore *ps, chunk_t chunk, int rw, int metadata)
179{ 179{
180 struct dm_io_region where = { 180 struct dm_io_region where = {
181 .bdev = ps->snap->cow->bdev, 181 .bdev = ps->snap->cow->bdev,
@@ -209,16 +209,23 @@ static int chunk_io(struct pstore *ps, uint32_t chunk, int rw, int metadata)
209} 209}
210 210
211/* 211/*
212 * Convert a metadata area index to a chunk index.
213 */
214static chunk_t area_location(struct pstore *ps, chunk_t area)
215{
216 return 1 + ((ps->exceptions_per_area + 1) * area);
217}
218
219/*
212 * Read or write a metadata area. Remembering to skip the first 220 * Read or write a metadata area. Remembering to skip the first
213 * chunk which holds the header. 221 * chunk which holds the header.
214 */ 222 */
215static int area_io(struct pstore *ps, uint32_t area, int rw) 223static int area_io(struct pstore *ps, chunk_t area, int rw)
216{ 224{
217 int r; 225 int r;
218 uint32_t chunk; 226 chunk_t chunk;
219 227
220 /* convert a metadata area index to a chunk index */ 228 chunk = area_location(ps, area);
221 chunk = 1 + ((ps->exceptions_per_area + 1) * area);
222 229
223 r = chunk_io(ps, chunk, rw, 0); 230 r = chunk_io(ps, chunk, rw, 0);
224 if (r) 231 if (r)
@@ -228,7 +235,7 @@ static int area_io(struct pstore *ps, uint32_t area, int rw)
228 return 0; 235 return 0;
229} 236}
230 237
231static int zero_area(struct pstore *ps, uint32_t area) 238static int zero_area(struct pstore *ps, chunk_t area)
232{ 239{
233 memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT); 240 memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
234 return area_io(ps, area, WRITE); 241 return area_io(ps, area, WRITE);
@@ -404,7 +411,7 @@ static int insert_exceptions(struct pstore *ps, int *full)
404 411
405static int read_exceptions(struct pstore *ps) 412static int read_exceptions(struct pstore *ps)
406{ 413{
407 uint32_t area; 414 chunk_t area;
408 int r, full = 1; 415 int r, full = 1;
409 416
410 /* 417 /*
@@ -517,6 +524,7 @@ static int persistent_prepare(struct exception_store *store,
517{ 524{
518 struct pstore *ps = get_info(store); 525 struct pstore *ps = get_info(store);
519 uint32_t stride; 526 uint32_t stride;
527 chunk_t next_free;
520 sector_t size = get_dev_size(store->snap->cow->bdev); 528 sector_t size = get_dev_size(store->snap->cow->bdev);
521 529
522 /* Is there enough room ? */ 530 /* Is there enough room ? */
@@ -530,7 +538,8 @@ static int persistent_prepare(struct exception_store *store,
530 * into account the location of the metadata chunks. 538 * into account the location of the metadata chunks.
531 */ 539 */
532 stride = (ps->exceptions_per_area + 1); 540 stride = (ps->exceptions_per_area + 1);
533 if ((++ps->next_free % stride) == 1) 541 next_free = ++ps->next_free;
542 if (sector_div(next_free, stride) == 1)
534 ps->next_free++; 543 ps->next_free++;
535 544
536 atomic_inc(&ps->pending_count); 545 atomic_inc(&ps->pending_count);
diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c
index 5b919159f084..dca401dc70a0 100644
--- a/drivers/md/dm-ioctl.c
+++ b/drivers/md/dm-ioctl.c
@@ -1131,7 +1131,7 @@ static void retrieve_deps(struct dm_table *table,
1131 unsigned int count = 0; 1131 unsigned int count = 0;
1132 struct list_head *tmp; 1132 struct list_head *tmp;
1133 size_t len, needed; 1133 size_t len, needed;
1134 struct dm_dev *dd; 1134 struct dm_dev_internal *dd;
1135 struct dm_target_deps *deps; 1135 struct dm_target_deps *deps;
1136 1136
1137 deps = get_result_buffer(param, param_size, &len); 1137 deps = get_result_buffer(param, param_size, &len);
@@ -1157,7 +1157,7 @@ static void retrieve_deps(struct dm_table *table,
1157 deps->count = count; 1157 deps->count = count;
1158 count = 0; 1158 count = 0;
1159 list_for_each_entry (dd, dm_table_get_devices(table), list) 1159 list_for_each_entry (dd, dm_table_get_devices(table), list)
1160 deps->dev[count++] = huge_encode_dev(dd->bdev->bd_dev); 1160 deps->dev[count++] = huge_encode_dev(dd->dm_dev.bdev->bd_dev);
1161 1161
1162 param->data_size = param->data_start + needed; 1162 param->data_size = param->data_start + needed;
1163} 1163}
diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index 3d3848132c69..103304c1e3b0 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -30,6 +30,7 @@ struct pgpath {
30 struct list_head list; 30 struct list_head list;
31 31
32 struct priority_group *pg; /* Owning PG */ 32 struct priority_group *pg; /* Owning PG */
33 unsigned is_active; /* Path status */
33 unsigned fail_count; /* Cumulative failure count */ 34 unsigned fail_count; /* Cumulative failure count */
34 35
35 struct dm_path path; 36 struct dm_path path;
@@ -125,7 +126,7 @@ static struct pgpath *alloc_pgpath(void)
125 struct pgpath *pgpath = kzalloc(sizeof(*pgpath), GFP_KERNEL); 126 struct pgpath *pgpath = kzalloc(sizeof(*pgpath), GFP_KERNEL);
126 127
127 if (pgpath) { 128 if (pgpath) {
128 pgpath->path.is_active = 1; 129 pgpath->is_active = 1;
129 INIT_WORK(&pgpath->deactivate_path, deactivate_path); 130 INIT_WORK(&pgpath->deactivate_path, deactivate_path);
130 } 131 }
131 132
@@ -575,12 +576,12 @@ static struct pgpath *parse_path(struct arg_set *as, struct path_selector *ps,
575 /* we need at least a path arg */ 576 /* we need at least a path arg */
576 if (as->argc < 1) { 577 if (as->argc < 1) {
577 ti->error = "no device given"; 578 ti->error = "no device given";
578 return NULL; 579 return ERR_PTR(-EINVAL);
579 } 580 }
580 581
581 p = alloc_pgpath(); 582 p = alloc_pgpath();
582 if (!p) 583 if (!p)
583 return NULL; 584 return ERR_PTR(-ENOMEM);
584 585
585 r = dm_get_device(ti, shift(as), ti->begin, ti->len, 586 r = dm_get_device(ti, shift(as), ti->begin, ti->len,
586 dm_table_get_mode(ti->table), &p->path.dev); 587 dm_table_get_mode(ti->table), &p->path.dev);
@@ -608,7 +609,7 @@ static struct pgpath *parse_path(struct arg_set *as, struct path_selector *ps,
608 609
609 bad: 610 bad:
610 free_pgpath(p); 611 free_pgpath(p);
611 return NULL; 612 return ERR_PTR(r);
612} 613}
613 614
614static struct priority_group *parse_priority_group(struct arg_set *as, 615static struct priority_group *parse_priority_group(struct arg_set *as,
@@ -626,14 +627,14 @@ static struct priority_group *parse_priority_group(struct arg_set *as,
626 627
627 if (as->argc < 2) { 628 if (as->argc < 2) {
628 as->argc = 0; 629 as->argc = 0;
629 ti->error = "not enough priority group aruments"; 630 ti->error = "not enough priority group arguments";
630 return NULL; 631 return ERR_PTR(-EINVAL);
631 } 632 }
632 633
633 pg = alloc_priority_group(); 634 pg = alloc_priority_group();
634 if (!pg) { 635 if (!pg) {
635 ti->error = "couldn't allocate priority group"; 636 ti->error = "couldn't allocate priority group";
636 return NULL; 637 return ERR_PTR(-ENOMEM);
637 } 638 }
638 pg->m = m; 639 pg->m = m;
639 640
@@ -666,8 +667,10 @@ static struct priority_group *parse_priority_group(struct arg_set *as,
666 path_args.argv = as->argv; 667 path_args.argv = as->argv;
667 668
668 pgpath = parse_path(&path_args, &pg->ps, ti); 669 pgpath = parse_path(&path_args, &pg->ps, ti);
669 if (!pgpath) 670 if (IS_ERR(pgpath)) {
671 r = PTR_ERR(pgpath);
670 goto bad; 672 goto bad;
673 }
671 674
672 pgpath->pg = pg; 675 pgpath->pg = pg;
673 list_add_tail(&pgpath->list, &pg->pgpaths); 676 list_add_tail(&pgpath->list, &pg->pgpaths);
@@ -678,7 +681,7 @@ static struct priority_group *parse_priority_group(struct arg_set *as,
678 681
679 bad: 682 bad:
680 free_priority_group(pg, ti); 683 free_priority_group(pg, ti);
681 return NULL; 684 return ERR_PTR(r);
682} 685}
683 686
684static int parse_hw_handler(struct arg_set *as, struct multipath *m) 687static int parse_hw_handler(struct arg_set *as, struct multipath *m)
@@ -797,8 +800,8 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
797 struct priority_group *pg; 800 struct priority_group *pg;
798 801
799 pg = parse_priority_group(&as, m); 802 pg = parse_priority_group(&as, m);
800 if (!pg) { 803 if (IS_ERR(pg)) {
801 r = -EINVAL; 804 r = PTR_ERR(pg);
802 goto bad; 805 goto bad;
803 } 806 }
804 807
@@ -864,13 +867,13 @@ static int fail_path(struct pgpath *pgpath)
864 867
865 spin_lock_irqsave(&m->lock, flags); 868 spin_lock_irqsave(&m->lock, flags);
866 869
867 if (!pgpath->path.is_active) 870 if (!pgpath->is_active)
868 goto out; 871 goto out;
869 872
870 DMWARN("Failing path %s.", pgpath->path.dev->name); 873 DMWARN("Failing path %s.", pgpath->path.dev->name);
871 874
872 pgpath->pg->ps.type->fail_path(&pgpath->pg->ps, &pgpath->path); 875 pgpath->pg->ps.type->fail_path(&pgpath->pg->ps, &pgpath->path);
873 pgpath->path.is_active = 0; 876 pgpath->is_active = 0;
874 pgpath->fail_count++; 877 pgpath->fail_count++;
875 878
876 m->nr_valid_paths--; 879 m->nr_valid_paths--;
@@ -901,7 +904,7 @@ static int reinstate_path(struct pgpath *pgpath)
901 904
902 spin_lock_irqsave(&m->lock, flags); 905 spin_lock_irqsave(&m->lock, flags);
903 906
904 if (pgpath->path.is_active) 907 if (pgpath->is_active)
905 goto out; 908 goto out;
906 909
907 if (!pgpath->pg->ps.type->reinstate_path) { 910 if (!pgpath->pg->ps.type->reinstate_path) {
@@ -915,7 +918,7 @@ static int reinstate_path(struct pgpath *pgpath)
915 if (r) 918 if (r)
916 goto out; 919 goto out;
917 920
918 pgpath->path.is_active = 1; 921 pgpath->is_active = 1;
919 922
920 m->current_pgpath = NULL; 923 m->current_pgpath = NULL;
921 if (!m->nr_valid_paths++ && m->queue_size) 924 if (!m->nr_valid_paths++ && m->queue_size)
@@ -1303,7 +1306,7 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
1303 1306
1304 list_for_each_entry(p, &pg->pgpaths, list) { 1307 list_for_each_entry(p, &pg->pgpaths, list) {
1305 DMEMIT("%s %s %u ", p->path.dev->name, 1308 DMEMIT("%s %s %u ", p->path.dev->name,
1306 p->path.is_active ? "A" : "F", 1309 p->is_active ? "A" : "F",
1307 p->fail_count); 1310 p->fail_count);
1308 if (pg->ps.type->status) 1311 if (pg->ps.type->status)
1309 sz += pg->ps.type->status(&pg->ps, 1312 sz += pg->ps.type->status(&pg->ps,
diff --git a/drivers/md/dm-mpath.h b/drivers/md/dm-mpath.h
index c198b856a452..e230f7196259 100644
--- a/drivers/md/dm-mpath.h
+++ b/drivers/md/dm-mpath.h
@@ -13,8 +13,6 @@ struct dm_dev;
13 13
14struct dm_path { 14struct dm_path {
15 struct dm_dev *dev; /* Read-only */ 15 struct dm_dev *dev; /* Read-only */
16 unsigned is_active; /* Read-only */
17
18 void *pscontext; /* For path-selector use */ 16 void *pscontext; /* For path-selector use */
19}; 17};
20 18
diff --git a/drivers/md/dm-raid1.c b/drivers/md/dm-raid1.c
index ff05fe893083..29913e42c4ab 100644
--- a/drivers/md/dm-raid1.c
+++ b/drivers/md/dm-raid1.c
@@ -842,7 +842,9 @@ static int recover(struct mirror_set *ms, struct region *reg)
842 } 842 }
843 843
844 /* hand to kcopyd */ 844 /* hand to kcopyd */
845 set_bit(DM_KCOPYD_IGNORE_ERROR, &flags); 845 if (!errors_handled(ms))
846 set_bit(DM_KCOPYD_IGNORE_ERROR, &flags);
847
846 r = dm_kcopyd_copy(ms->kcopyd_client, &from, ms->nr_mirrors - 1, to, 848 r = dm_kcopyd_copy(ms->kcopyd_client, &from, ms->nr_mirrors - 1, to,
847 flags, recovery_complete, reg); 849 flags, recovery_complete, reg);
848 850
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
index 61f441409234..a740a6950f59 100644
--- a/drivers/md/dm-table.c
+++ b/drivers/md/dm-table.c
@@ -250,7 +250,8 @@ static void free_devices(struct list_head *devices)
250 struct list_head *tmp, *next; 250 struct list_head *tmp, *next;
251 251
252 list_for_each_safe(tmp, next, devices) { 252 list_for_each_safe(tmp, next, devices) {
253 struct dm_dev *dd = list_entry(tmp, struct dm_dev, list); 253 struct dm_dev_internal *dd =
254 list_entry(tmp, struct dm_dev_internal, list);
254 kfree(dd); 255 kfree(dd);
255 } 256 }
256} 257}
@@ -327,12 +328,12 @@ static int lookup_device(const char *path, dev_t *dev)
327/* 328/*
328 * See if we've already got a device in the list. 329 * See if we've already got a device in the list.
329 */ 330 */
330static struct dm_dev *find_device(struct list_head *l, dev_t dev) 331static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
331{ 332{
332 struct dm_dev *dd; 333 struct dm_dev_internal *dd;
333 334
334 list_for_each_entry (dd, l, list) 335 list_for_each_entry (dd, l, list)
335 if (dd->bdev->bd_dev == dev) 336 if (dd->dm_dev.bdev->bd_dev == dev)
336 return dd; 337 return dd;
337 338
338 return NULL; 339 return NULL;
@@ -341,45 +342,47 @@ static struct dm_dev *find_device(struct list_head *l, dev_t dev)
341/* 342/*
342 * Open a device so we can use it as a map destination. 343 * Open a device so we can use it as a map destination.
343 */ 344 */
344static int open_dev(struct dm_dev *d, dev_t dev, struct mapped_device *md) 345static int open_dev(struct dm_dev_internal *d, dev_t dev,
346 struct mapped_device *md)
345{ 347{
346 static char *_claim_ptr = "I belong to device-mapper"; 348 static char *_claim_ptr = "I belong to device-mapper";
347 struct block_device *bdev; 349 struct block_device *bdev;
348 350
349 int r; 351 int r;
350 352
351 BUG_ON(d->bdev); 353 BUG_ON(d->dm_dev.bdev);
352 354
353 bdev = open_by_devnum(dev, d->mode); 355 bdev = open_by_devnum(dev, d->dm_dev.mode);
354 if (IS_ERR(bdev)) 356 if (IS_ERR(bdev))
355 return PTR_ERR(bdev); 357 return PTR_ERR(bdev);
356 r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md)); 358 r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md));
357 if (r) 359 if (r)
358 blkdev_put(bdev); 360 blkdev_put(bdev);
359 else 361 else
360 d->bdev = bdev; 362 d->dm_dev.bdev = bdev;
361 return r; 363 return r;
362} 364}
363 365
364/* 366/*
365 * Close a device that we've been using. 367 * Close a device that we've been using.
366 */ 368 */
367static void close_dev(struct dm_dev *d, struct mapped_device *md) 369static void close_dev(struct dm_dev_internal *d, struct mapped_device *md)
368{ 370{
369 if (!d->bdev) 371 if (!d->dm_dev.bdev)
370 return; 372 return;
371 373
372 bd_release_from_disk(d->bdev, dm_disk(md)); 374 bd_release_from_disk(d->dm_dev.bdev, dm_disk(md));
373 blkdev_put(d->bdev); 375 blkdev_put(d->dm_dev.bdev);
374 d->bdev = NULL; 376 d->dm_dev.bdev = NULL;
375} 377}
376 378
377/* 379/*
378 * If possible, this checks an area of a destination device is valid. 380 * If possible, this checks an area of a destination device is valid.
379 */ 381 */
380static int check_device_area(struct dm_dev *dd, sector_t start, sector_t len) 382static int check_device_area(struct dm_dev_internal *dd, sector_t start,
383 sector_t len)
381{ 384{
382 sector_t dev_size = dd->bdev->bd_inode->i_size >> SECTOR_SHIFT; 385 sector_t dev_size = dd->dm_dev.bdev->bd_inode->i_size >> SECTOR_SHIFT;
383 386
384 if (!dev_size) 387 if (!dev_size)
385 return 1; 388 return 1;
@@ -392,16 +395,17 @@ static int check_device_area(struct dm_dev *dd, sector_t start, sector_t len)
392 * careful to leave things as they were if we fail to reopen the 395 * careful to leave things as they were if we fail to reopen the
393 * device. 396 * device.
394 */ 397 */
395static int upgrade_mode(struct dm_dev *dd, int new_mode, struct mapped_device *md) 398static int upgrade_mode(struct dm_dev_internal *dd, int new_mode,
399 struct mapped_device *md)
396{ 400{
397 int r; 401 int r;
398 struct dm_dev dd_copy; 402 struct dm_dev_internal dd_copy;
399 dev_t dev = dd->bdev->bd_dev; 403 dev_t dev = dd->dm_dev.bdev->bd_dev;
400 404
401 dd_copy = *dd; 405 dd_copy = *dd;
402 406
403 dd->mode |= new_mode; 407 dd->dm_dev.mode |= new_mode;
404 dd->bdev = NULL; 408 dd->dm_dev.bdev = NULL;
405 r = open_dev(dd, dev, md); 409 r = open_dev(dd, dev, md);
406 if (!r) 410 if (!r)
407 close_dev(&dd_copy, md); 411 close_dev(&dd_copy, md);
@@ -421,7 +425,7 @@ static int __table_get_device(struct dm_table *t, struct dm_target *ti,
421{ 425{
422 int r; 426 int r;
423 dev_t uninitialized_var(dev); 427 dev_t uninitialized_var(dev);
424 struct dm_dev *dd; 428 struct dm_dev_internal *dd;
425 unsigned int major, minor; 429 unsigned int major, minor;
426 430
427 BUG_ON(!t); 431 BUG_ON(!t);
@@ -443,20 +447,20 @@ static int __table_get_device(struct dm_table *t, struct dm_target *ti,
443 if (!dd) 447 if (!dd)
444 return -ENOMEM; 448 return -ENOMEM;
445 449
446 dd->mode = mode; 450 dd->dm_dev.mode = mode;
447 dd->bdev = NULL; 451 dd->dm_dev.bdev = NULL;
448 452
449 if ((r = open_dev(dd, dev, t->md))) { 453 if ((r = open_dev(dd, dev, t->md))) {
450 kfree(dd); 454 kfree(dd);
451 return r; 455 return r;
452 } 456 }
453 457
454 format_dev_t(dd->name, dev); 458 format_dev_t(dd->dm_dev.name, dev);
455 459
456 atomic_set(&dd->count, 0); 460 atomic_set(&dd->count, 0);
457 list_add(&dd->list, &t->devices); 461 list_add(&dd->list, &t->devices);
458 462
459 } else if (dd->mode != (mode | dd->mode)) { 463 } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) {
460 r = upgrade_mode(dd, mode, t->md); 464 r = upgrade_mode(dd, mode, t->md);
461 if (r) 465 if (r)
462 return r; 466 return r;
@@ -465,11 +469,11 @@ static int __table_get_device(struct dm_table *t, struct dm_target *ti,
465 469
466 if (!check_device_area(dd, start, len)) { 470 if (!check_device_area(dd, start, len)) {
467 DMWARN("device %s too small for target", path); 471 DMWARN("device %s too small for target", path);
468 dm_put_device(ti, dd); 472 dm_put_device(ti, &dd->dm_dev);
469 return -EINVAL; 473 return -EINVAL;
470 } 474 }
471 475
472 *result = dd; 476 *result = &dd->dm_dev;
473 477
474 return 0; 478 return 0;
475} 479}
@@ -478,6 +482,13 @@ void dm_set_device_limits(struct dm_target *ti, struct block_device *bdev)
478{ 482{
479 struct request_queue *q = bdev_get_queue(bdev); 483 struct request_queue *q = bdev_get_queue(bdev);
480 struct io_restrictions *rs = &ti->limits; 484 struct io_restrictions *rs = &ti->limits;
485 char b[BDEVNAME_SIZE];
486
487 if (unlikely(!q)) {
488 DMWARN("%s: Cannot set limits for nonexistent device %s",
489 dm_device_name(ti->table->md), bdevname(bdev, b));
490 return;
491 }
481 492
482 /* 493 /*
483 * Combine the device limits low. 494 * Combine the device limits low.
@@ -540,8 +551,11 @@ int dm_get_device(struct dm_target *ti, const char *path, sector_t start,
540/* 551/*
541 * Decrement a devices use count and remove it if necessary. 552 * Decrement a devices use count and remove it if necessary.
542 */ 553 */
543void dm_put_device(struct dm_target *ti, struct dm_dev *dd) 554void dm_put_device(struct dm_target *ti, struct dm_dev *d)
544{ 555{
556 struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal,
557 dm_dev);
558
545 if (atomic_dec_and_test(&dd->count)) { 559 if (atomic_dec_and_test(&dd->count)) {
546 close_dev(dd, ti->table->md); 560 close_dev(dd, ti->table->md);
547 list_del(&dd->list); 561 list_del(&dd->list);
@@ -937,13 +951,20 @@ int dm_table_resume_targets(struct dm_table *t)
937 951
938int dm_table_any_congested(struct dm_table *t, int bdi_bits) 952int dm_table_any_congested(struct dm_table *t, int bdi_bits)
939{ 953{
940 struct dm_dev *dd; 954 struct dm_dev_internal *dd;
941 struct list_head *devices = dm_table_get_devices(t); 955 struct list_head *devices = dm_table_get_devices(t);
942 int r = 0; 956 int r = 0;
943 957
944 list_for_each_entry(dd, devices, list) { 958 list_for_each_entry(dd, devices, list) {
945 struct request_queue *q = bdev_get_queue(dd->bdev); 959 struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
946 r |= bdi_congested(&q->backing_dev_info, bdi_bits); 960 char b[BDEVNAME_SIZE];
961
962 if (likely(q))
963 r |= bdi_congested(&q->backing_dev_info, bdi_bits);
964 else
965 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
966 dm_device_name(t->md),
967 bdevname(dd->dm_dev.bdev, b));
947 } 968 }
948 969
949 return r; 970 return r;
@@ -951,13 +972,19 @@ int dm_table_any_congested(struct dm_table *t, int bdi_bits)
951 972
952void dm_table_unplug_all(struct dm_table *t) 973void dm_table_unplug_all(struct dm_table *t)
953{ 974{
954 struct dm_dev *dd; 975 struct dm_dev_internal *dd;
955 struct list_head *devices = dm_table_get_devices(t); 976 struct list_head *devices = dm_table_get_devices(t);
956 977
957 list_for_each_entry(dd, devices, list) { 978 list_for_each_entry(dd, devices, list) {
958 struct request_queue *q = bdev_get_queue(dd->bdev); 979 struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
959 980 char b[BDEVNAME_SIZE];
960 blk_unplug(q); 981
982 if (likely(q))
983 blk_unplug(q);
984 else
985 DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s",
986 dm_device_name(t->md),
987 bdevname(dd->dm_dev.bdev, b));
961 } 988 }
962} 989}
963 990
diff --git a/drivers/md/dm.h b/drivers/md/dm.h
index 1e59a0b0a78a..cd189da2b2fa 100644
--- a/drivers/md/dm.h
+++ b/drivers/md/dm.h
@@ -25,13 +25,10 @@
25/* 25/*
26 * List of devices that a metadevice uses and should open/close. 26 * List of devices that a metadevice uses and should open/close.
27 */ 27 */
28struct dm_dev { 28struct dm_dev_internal {
29 struct list_head list; 29 struct list_head list;
30
31 atomic_t count; 30 atomic_t count;
32 int mode; 31 struct dm_dev dm_dev;
33 struct block_device *bdev;
34 char name[16];
35}; 32};
36 33
37struct dm_table; 34struct dm_table;
@@ -49,7 +46,6 @@ void dm_table_presuspend_targets(struct dm_table *t);
49void dm_table_postsuspend_targets(struct dm_table *t); 46void dm_table_postsuspend_targets(struct dm_table *t);
50int dm_table_resume_targets(struct dm_table *t); 47int dm_table_resume_targets(struct dm_table *t);
51int dm_table_any_congested(struct dm_table *t, int bdi_bits); 48int dm_table_any_congested(struct dm_table *t, int bdi_bits);
52void dm_table_unplug_all(struct dm_table *t);
53 49
54/* 50/*
55 * To check the return value from dm_table_find_target(). 51 * To check the return value from dm_table_find_target().
@@ -93,8 +89,6 @@ void dm_linear_exit(void);
93int dm_stripe_init(void); 89int dm_stripe_init(void);
94void dm_stripe_exit(void); 90void dm_stripe_exit(void);
95 91
96void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size);
97union map_info *dm_get_mapinfo(struct bio *bio);
98int dm_open_count(struct mapped_device *md); 92int dm_open_count(struct mapped_device *md);
99int dm_lock_for_deletion(struct mapped_device *md); 93int dm_lock_for_deletion(struct mapped_device *md);
100 94