aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/md
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/md')
-rw-r--r--drivers/md/Kconfig6
-rw-r--r--drivers/md/Makefile2
-rw-r--r--drivers/md/dm-bio-list.h4
-rw-r--r--drivers/md/dm-crypt.c33
-rw-r--r--drivers/md/dm-delay.c23
-rw-r--r--drivers/md/dm-exception-store.c76
-rw-r--r--drivers/md/dm-io.c5
-rw-r--r--drivers/md/dm-mpath-rdac.c700
-rw-r--r--drivers/md/dm-mpath.c34
-rw-r--r--drivers/md/dm-raid1.c75
-rw-r--r--drivers/md/dm-round-robin.c2
-rw-r--r--drivers/md/dm-snap.c116
-rw-r--r--drivers/md/dm-snap.h6
-rw-r--r--drivers/md/dm.c33
-rw-r--r--drivers/md/dm.h40
-rw-r--r--drivers/md/kcopyd.c11
16 files changed, 984 insertions, 182 deletions
diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig
index bfd9b9c6252c..64bf3a81db93 100644
--- a/drivers/md/Kconfig
+++ b/drivers/md/Kconfig
@@ -264,6 +264,12 @@ config DM_MULTIPATH_EMC
264 ---help--- 264 ---help---
265 Multipath support for EMC CX/AX series hardware. 265 Multipath support for EMC CX/AX series hardware.
266 266
267config DM_MULTIPATH_RDAC
268 tristate "LSI/Engenio RDAC multipath support (EXPERIMENTAL)"
269 depends on DM_MULTIPATH && BLK_DEV_DM && EXPERIMENTAL
270 ---help---
271 Multipath support for LSI/Engenio RDAC.
272
267config DM_DELAY 273config DM_DELAY
268 tristate "I/O delaying target (EXPERIMENTAL)" 274 tristate "I/O delaying target (EXPERIMENTAL)"
269 depends on BLK_DEV_DM && EXPERIMENTAL 275 depends on BLK_DEV_DM && EXPERIMENTAL
diff --git a/drivers/md/Makefile b/drivers/md/Makefile
index 71eb45f74171..c49366cdc05d 100644
--- a/drivers/md/Makefile
+++ b/drivers/md/Makefile
@@ -7,6 +7,7 @@ dm-mod-objs := dm.o dm-table.o dm-target.o dm-linear.o dm-stripe.o \
7dm-multipath-objs := dm-hw-handler.o dm-path-selector.o dm-mpath.o 7dm-multipath-objs := dm-hw-handler.o dm-path-selector.o dm-mpath.o
8dm-snapshot-objs := dm-snap.o dm-exception-store.o 8dm-snapshot-objs := dm-snap.o dm-exception-store.o
9dm-mirror-objs := dm-log.o dm-raid1.o 9dm-mirror-objs := dm-log.o dm-raid1.o
10dm-rdac-objs := dm-mpath-rdac.o
10md-mod-objs := md.o bitmap.o 11md-mod-objs := md.o bitmap.o
11raid456-objs := raid5.o raid6algos.o raid6recov.o raid6tables.o \ 12raid456-objs := raid5.o raid6algos.o raid6recov.o raid6tables.o \
12 raid6int1.o raid6int2.o raid6int4.o \ 13 raid6int1.o raid6int2.o raid6int4.o \
@@ -34,6 +35,7 @@ obj-$(CONFIG_DM_CRYPT) += dm-crypt.o
34obj-$(CONFIG_DM_DELAY) += dm-delay.o 35obj-$(CONFIG_DM_DELAY) += dm-delay.o
35obj-$(CONFIG_DM_MULTIPATH) += dm-multipath.o dm-round-robin.o 36obj-$(CONFIG_DM_MULTIPATH) += dm-multipath.o dm-round-robin.o
36obj-$(CONFIG_DM_MULTIPATH_EMC) += dm-emc.o 37obj-$(CONFIG_DM_MULTIPATH_EMC) += dm-emc.o
38obj-$(CONFIG_DM_MULTIPATH_RDAC) += dm-rdac.o
37obj-$(CONFIG_DM_SNAPSHOT) += dm-snapshot.o 39obj-$(CONFIG_DM_SNAPSHOT) += dm-snapshot.o
38obj-$(CONFIG_DM_MIRROR) += dm-mirror.o 40obj-$(CONFIG_DM_MIRROR) += dm-mirror.o
39obj-$(CONFIG_DM_ZERO) += dm-zero.o 41obj-$(CONFIG_DM_ZERO) += dm-zero.o
diff --git a/drivers/md/dm-bio-list.h b/drivers/md/dm-bio-list.h
index c6be88826fae..16ee3b018b3a 100644
--- a/drivers/md/dm-bio-list.h
+++ b/drivers/md/dm-bio-list.h
@@ -8,7 +8,6 @@
8#define DM_BIO_LIST_H 8#define DM_BIO_LIST_H
9 9
10#include <linux/bio.h> 10#include <linux/bio.h>
11#include <linux/prefetch.h>
12 11
13struct bio_list { 12struct bio_list {
14 struct bio *head; 13 struct bio *head;
@@ -31,8 +30,7 @@ static inline void bio_list_init(struct bio_list *bl)
31} 30}
32 31
33#define bio_list_for_each(bio, bl) \ 32#define bio_list_for_each(bio, bl) \
34 for (bio = (bl)->head; bio && ({ prefetch(bio->bi_next); 1; }); \ 33 for (bio = (bl)->head; bio; bio = bio->bi_next)
35 bio = bio->bi_next)
36 34
37static inline unsigned bio_list_size(const struct bio_list *bl) 35static inline unsigned bio_list_size(const struct bio_list *bl)
38{ 36{
diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 7b0fcfc9eaa5..ba952a032598 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -30,7 +30,7 @@
30/* 30/*
31 * per bio private data 31 * per bio private data
32 */ 32 */
33struct crypt_io { 33struct dm_crypt_io {
34 struct dm_target *target; 34 struct dm_target *target;
35 struct bio *base_bio; 35 struct bio *base_bio;
36 struct work_struct work; 36 struct work_struct work;
@@ -106,7 +106,7 @@ struct crypt_config {
106 106
107static struct kmem_cache *_crypt_io_pool; 107static struct kmem_cache *_crypt_io_pool;
108 108
109static void clone_init(struct crypt_io *, struct bio *); 109static void clone_init(struct dm_crypt_io *, struct bio *);
110 110
111/* 111/*
112 * Different IV generation algorithms: 112 * Different IV generation algorithms:
@@ -382,7 +382,7 @@ static int crypt_convert(struct crypt_config *cc,
382 382
383 static void dm_crypt_bio_destructor(struct bio *bio) 383 static void dm_crypt_bio_destructor(struct bio *bio)
384 { 384 {
385 struct crypt_io *io = bio->bi_private; 385 struct dm_crypt_io *io = bio->bi_private;
386 struct crypt_config *cc = io->target->private; 386 struct crypt_config *cc = io->target->private;
387 387
388 bio_free(bio, cc->bs); 388 bio_free(bio, cc->bs);
@@ -393,7 +393,7 @@ static int crypt_convert(struct crypt_config *cc,
393 * This should never violate the device limitations 393 * This should never violate the device limitations
394 * May return a smaller bio when running out of pages 394 * May return a smaller bio when running out of pages
395 */ 395 */
396static struct bio *crypt_alloc_buffer(struct crypt_io *io, unsigned int size) 396static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned size)
397{ 397{
398 struct crypt_config *cc = io->target->private; 398 struct crypt_config *cc = io->target->private;
399 struct bio *clone; 399 struct bio *clone;
@@ -479,7 +479,7 @@ static void crypt_free_buffer_pages(struct crypt_config *cc,
479 * One of the bios was finished. Check for completion of 479 * One of the bios was finished. Check for completion of
480 * the whole request and correctly clean up the buffer. 480 * the whole request and correctly clean up the buffer.
481 */ 481 */
482static void dec_pending(struct crypt_io *io, int error) 482static void dec_pending(struct dm_crypt_io *io, int error)
483{ 483{
484 struct crypt_config *cc = (struct crypt_config *) io->target->private; 484 struct crypt_config *cc = (struct crypt_config *) io->target->private;
485 485
@@ -503,7 +503,7 @@ static void dec_pending(struct crypt_io *io, int error)
503static struct workqueue_struct *_kcryptd_workqueue; 503static struct workqueue_struct *_kcryptd_workqueue;
504static void kcryptd_do_work(struct work_struct *work); 504static void kcryptd_do_work(struct work_struct *work);
505 505
506static void kcryptd_queue_io(struct crypt_io *io) 506static void kcryptd_queue_io(struct dm_crypt_io *io)
507{ 507{
508 INIT_WORK(&io->work, kcryptd_do_work); 508 INIT_WORK(&io->work, kcryptd_do_work);
509 queue_work(_kcryptd_workqueue, &io->work); 509 queue_work(_kcryptd_workqueue, &io->work);
@@ -511,7 +511,7 @@ static void kcryptd_queue_io(struct crypt_io *io)
511 511
512static int crypt_endio(struct bio *clone, unsigned int done, int error) 512static int crypt_endio(struct bio *clone, unsigned int done, int error)
513{ 513{
514 struct crypt_io *io = clone->bi_private; 514 struct dm_crypt_io *io = clone->bi_private;
515 struct crypt_config *cc = io->target->private; 515 struct crypt_config *cc = io->target->private;
516 unsigned read_io = bio_data_dir(clone) == READ; 516 unsigned read_io = bio_data_dir(clone) == READ;
517 517
@@ -545,7 +545,7 @@ out:
545 return error; 545 return error;
546} 546}
547 547
548static void clone_init(struct crypt_io *io, struct bio *clone) 548static void clone_init(struct dm_crypt_io *io, struct bio *clone)
549{ 549{
550 struct crypt_config *cc = io->target->private; 550 struct crypt_config *cc = io->target->private;
551 551
@@ -556,7 +556,7 @@ static void clone_init(struct crypt_io *io, struct bio *clone)
556 clone->bi_destructor = dm_crypt_bio_destructor; 556 clone->bi_destructor = dm_crypt_bio_destructor;
557} 557}
558 558
559static void process_read(struct crypt_io *io) 559static void process_read(struct dm_crypt_io *io)
560{ 560{
561 struct crypt_config *cc = io->target->private; 561 struct crypt_config *cc = io->target->private;
562 struct bio *base_bio = io->base_bio; 562 struct bio *base_bio = io->base_bio;
@@ -587,7 +587,7 @@ static void process_read(struct crypt_io *io)
587 generic_make_request(clone); 587 generic_make_request(clone);
588} 588}
589 589
590static void process_write(struct crypt_io *io) 590static void process_write(struct dm_crypt_io *io)
591{ 591{
592 struct crypt_config *cc = io->target->private; 592 struct crypt_config *cc = io->target->private;
593 struct bio *base_bio = io->base_bio; 593 struct bio *base_bio = io->base_bio;
@@ -644,7 +644,7 @@ static void process_write(struct crypt_io *io)
644 } 644 }
645} 645}
646 646
647static void process_read_endio(struct crypt_io *io) 647static void process_read_endio(struct dm_crypt_io *io)
648{ 648{
649 struct crypt_config *cc = io->target->private; 649 struct crypt_config *cc = io->target->private;
650 struct convert_context ctx; 650 struct convert_context ctx;
@@ -657,7 +657,7 @@ static void process_read_endio(struct crypt_io *io)
657 657
658static void kcryptd_do_work(struct work_struct *work) 658static void kcryptd_do_work(struct work_struct *work)
659{ 659{
660 struct crypt_io *io = container_of(work, struct crypt_io, work); 660 struct dm_crypt_io *io = container_of(work, struct dm_crypt_io, work);
661 661
662 if (io->post_process) 662 if (io->post_process)
663 process_read_endio(io); 663 process_read_endio(io);
@@ -939,10 +939,7 @@ static int crypt_map(struct dm_target *ti, struct bio *bio,
939 union map_info *map_context) 939 union map_info *map_context)
940{ 940{
941 struct crypt_config *cc = ti->private; 941 struct crypt_config *cc = ti->private;
942 struct crypt_io *io; 942 struct dm_crypt_io *io;
943
944 if (bio_barrier(bio))
945 return -EOPNOTSUPP;
946 943
947 io = mempool_alloc(cc->io_pool, GFP_NOIO); 944 io = mempool_alloc(cc->io_pool, GFP_NOIO);
948 io->target = ti; 945 io->target = ti;
@@ -1062,9 +1059,7 @@ static int __init dm_crypt_init(void)
1062{ 1059{
1063 int r; 1060 int r;
1064 1061
1065 _crypt_io_pool = kmem_cache_create("dm-crypt_io", 1062 _crypt_io_pool = KMEM_CACHE(dm_crypt_io, 0);
1066 sizeof(struct crypt_io),
1067 0, 0, NULL, NULL);
1068 if (!_crypt_io_pool) 1063 if (!_crypt_io_pool)
1069 return -ENOMEM; 1064 return -ENOMEM;
1070 1065
diff --git a/drivers/md/dm-delay.c b/drivers/md/dm-delay.c
index 52c7cf9e5803..6928c136d3c5 100644
--- a/drivers/md/dm-delay.c
+++ b/drivers/md/dm-delay.c
@@ -20,7 +20,7 @@
20 20
21struct delay_c { 21struct delay_c {
22 struct timer_list delay_timer; 22 struct timer_list delay_timer;
23 struct semaphore timer_lock; 23 struct mutex timer_lock;
24 struct work_struct flush_expired_bios; 24 struct work_struct flush_expired_bios;
25 struct list_head delayed_bios; 25 struct list_head delayed_bios;
26 atomic_t may_delay; 26 atomic_t may_delay;
@@ -37,7 +37,7 @@ struct delay_c {
37 unsigned writes; 37 unsigned writes;
38}; 38};
39 39
40struct delay_info { 40struct dm_delay_info {
41 struct delay_c *context; 41 struct delay_c *context;
42 struct list_head list; 42 struct list_head list;
43 struct bio *bio; 43 struct bio *bio;
@@ -58,12 +58,12 @@ static void handle_delayed_timer(unsigned long data)
58 58
59static void queue_timeout(struct delay_c *dc, unsigned long expires) 59static void queue_timeout(struct delay_c *dc, unsigned long expires)
60{ 60{
61 down(&dc->timer_lock); 61 mutex_lock(&dc->timer_lock);
62 62
63 if (!timer_pending(&dc->delay_timer) || expires < dc->delay_timer.expires) 63 if (!timer_pending(&dc->delay_timer) || expires < dc->delay_timer.expires)
64 mod_timer(&dc->delay_timer, expires); 64 mod_timer(&dc->delay_timer, expires);
65 65
66 up(&dc->timer_lock); 66 mutex_unlock(&dc->timer_lock);
67} 67}
68 68
69static void flush_bios(struct bio *bio) 69static void flush_bios(struct bio *bio)
@@ -80,7 +80,7 @@ static void flush_bios(struct bio *bio)
80 80
81static struct bio *flush_delayed_bios(struct delay_c *dc, int flush_all) 81static struct bio *flush_delayed_bios(struct delay_c *dc, int flush_all)
82{ 82{
83 struct delay_info *delayed, *next; 83 struct dm_delay_info *delayed, *next;
84 unsigned long next_expires = 0; 84 unsigned long next_expires = 0;
85 int start_timer = 0; 85 int start_timer = 0;
86 BIO_LIST(flush_bios); 86 BIO_LIST(flush_bios);
@@ -193,13 +193,11 @@ out:
193 goto bad; 193 goto bad;
194 } 194 }
195 195
196 init_timer(&dc->delay_timer); 196 setup_timer(&dc->delay_timer, handle_delayed_timer, (unsigned long)dc);
197 dc->delay_timer.function = handle_delayed_timer;
198 dc->delay_timer.data = (unsigned long)dc;
199 197
200 INIT_WORK(&dc->flush_expired_bios, flush_expired_bios); 198 INIT_WORK(&dc->flush_expired_bios, flush_expired_bios);
201 INIT_LIST_HEAD(&dc->delayed_bios); 199 INIT_LIST_HEAD(&dc->delayed_bios);
202 init_MUTEX(&dc->timer_lock); 200 mutex_init(&dc->timer_lock);
203 atomic_set(&dc->may_delay, 1); 201 atomic_set(&dc->may_delay, 1);
204 202
205 ti->private = dc; 203 ti->private = dc;
@@ -227,7 +225,7 @@ static void delay_dtr(struct dm_target *ti)
227 225
228static int delay_bio(struct delay_c *dc, int delay, struct bio *bio) 226static int delay_bio(struct delay_c *dc, int delay, struct bio *bio)
229{ 227{
230 struct delay_info *delayed; 228 struct dm_delay_info *delayed;
231 unsigned long expires = 0; 229 unsigned long expires = 0;
232 230
233 if (!delay || !atomic_read(&dc->may_delay)) 231 if (!delay || !atomic_read(&dc->may_delay))
@@ -338,10 +336,7 @@ static int __init dm_delay_init(void)
338 goto bad_queue; 336 goto bad_queue;
339 } 337 }
340 338
341 delayed_cache = kmem_cache_create("dm-delay", 339 delayed_cache = KMEM_CACHE(dm_delay_info, 0);
342 sizeof(struct delay_info),
343 __alignof__(struct delay_info),
344 0, NULL, NULL);
345 if (!delayed_cache) { 340 if (!delayed_cache) {
346 DMERR("Couldn't create delayed bio cache."); 341 DMERR("Couldn't create delayed bio cache.");
347 goto bad_memcache; 342 goto bad_memcache;
diff --git a/drivers/md/dm-exception-store.c b/drivers/md/dm-exception-store.c
index 07e0a0c84f6e..3d65917a1bbb 100644
--- a/drivers/md/dm-exception-store.c
+++ b/drivers/md/dm-exception-store.c
@@ -125,9 +125,11 @@ struct pstore {
125 uint32_t callback_count; 125 uint32_t callback_count;
126 struct commit_callback *callbacks; 126 struct commit_callback *callbacks;
127 struct dm_io_client *io_client; 127 struct dm_io_client *io_client;
128
129 struct workqueue_struct *metadata_wq;
128}; 130};
129 131
130static inline unsigned int sectors_to_pages(unsigned int sectors) 132static unsigned sectors_to_pages(unsigned sectors)
131{ 133{
132 return sectors / (PAGE_SIZE >> 9); 134 return sectors / (PAGE_SIZE >> 9);
133} 135}
@@ -156,10 +158,24 @@ static void free_area(struct pstore *ps)
156 ps->area = NULL; 158 ps->area = NULL;
157} 159}
158 160
161struct mdata_req {
162 struct io_region *where;
163 struct dm_io_request *io_req;
164 struct work_struct work;
165 int result;
166};
167
168static void do_metadata(struct work_struct *work)
169{
170 struct mdata_req *req = container_of(work, struct mdata_req, work);
171
172 req->result = dm_io(req->io_req, 1, req->where, NULL);
173}
174
159/* 175/*
160 * 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.
161 */ 177 */
162static int chunk_io(struct pstore *ps, uint32_t chunk, int rw) 178static int chunk_io(struct pstore *ps, uint32_t chunk, int rw, int metadata)
163{ 179{
164 struct io_region where = { 180 struct io_region where = {
165 .bdev = ps->snap->cow->bdev, 181 .bdev = ps->snap->cow->bdev,
@@ -173,8 +189,23 @@ static int chunk_io(struct pstore *ps, uint32_t chunk, int rw)
173 .client = ps->io_client, 189 .client = ps->io_client,
174 .notify.fn = NULL, 190 .notify.fn = NULL,
175 }; 191 };
192 struct mdata_req req;
193
194 if (!metadata)
195 return dm_io(&io_req, 1, &where, NULL);
196
197 req.where = &where;
198 req.io_req = &io_req;
176 199
177 return dm_io(&io_req, 1, &where, NULL); 200 /*
201 * Issue the synchronous I/O from a different thread
202 * to avoid generic_make_request recursion.
203 */
204 INIT_WORK(&req.work, do_metadata);
205 queue_work(ps->metadata_wq, &req.work);
206 flush_workqueue(ps->metadata_wq);
207
208 return req.result;
178} 209}
179 210
180/* 211/*
@@ -189,7 +220,7 @@ static int area_io(struct pstore *ps, uint32_t area, int rw)
189 /* convert a metadata area index to a chunk index */ 220 /* convert a metadata area index to a chunk index */
190 chunk = 1 + ((ps->exceptions_per_area + 1) * area); 221 chunk = 1 + ((ps->exceptions_per_area + 1) * area);
191 222
192 r = chunk_io(ps, chunk, rw); 223 r = chunk_io(ps, chunk, rw, 0);
193 if (r) 224 if (r)
194 return r; 225 return r;
195 226
@@ -230,7 +261,7 @@ static int read_header(struct pstore *ps, int *new_snapshot)
230 if (r) 261 if (r)
231 return r; 262 return r;
232 263
233 r = chunk_io(ps, 0, READ); 264 r = chunk_io(ps, 0, READ, 1);
234 if (r) 265 if (r)
235 goto bad; 266 goto bad;
236 267
@@ -292,7 +323,7 @@ static int write_header(struct pstore *ps)
292 dh->version = cpu_to_le32(ps->version); 323 dh->version = cpu_to_le32(ps->version);
293 dh->chunk_size = cpu_to_le32(ps->snap->chunk_size); 324 dh->chunk_size = cpu_to_le32(ps->snap->chunk_size);
294 325
295 return chunk_io(ps, 0, WRITE); 326 return chunk_io(ps, 0, WRITE, 1);
296} 327}
297 328
298/* 329/*
@@ -393,7 +424,7 @@ static int read_exceptions(struct pstore *ps)
393 return 0; 424 return 0;
394} 425}
395 426
396static inline struct pstore *get_info(struct exception_store *store) 427static struct pstore *get_info(struct exception_store *store)
397{ 428{
398 return (struct pstore *) store->context; 429 return (struct pstore *) store->context;
399} 430}
@@ -409,6 +440,7 @@ static void persistent_destroy(struct exception_store *store)
409{ 440{
410 struct pstore *ps = get_info(store); 441 struct pstore *ps = get_info(store);
411 442
443 destroy_workqueue(ps->metadata_wq);
412 dm_io_client_destroy(ps->io_client); 444 dm_io_client_destroy(ps->io_client);
413 vfree(ps->callbacks); 445 vfree(ps->callbacks);
414 free_area(ps); 446 free_area(ps);
@@ -457,11 +489,6 @@ static int persistent_read_metadata(struct exception_store *store)
457 /* 489 /*
458 * Sanity checks. 490 * Sanity checks.
459 */ 491 */
460 if (!ps->valid) {
461 DMWARN("snapshot is marked invalid");
462 return -EINVAL;
463 }
464
465 if (ps->version != SNAPSHOT_DISK_VERSION) { 492 if (ps->version != SNAPSHOT_DISK_VERSION) {
466 DMWARN("unable to handle snapshot disk version %d", 493 DMWARN("unable to handle snapshot disk version %d",
467 ps->version); 494 ps->version);
@@ -469,6 +496,12 @@ static int persistent_read_metadata(struct exception_store *store)
469 } 496 }
470 497
471 /* 498 /*
499 * Metadata are valid, but snapshot is invalidated
500 */
501 if (!ps->valid)
502 return 1;
503
504 /*
472 * Read the metadata. 505 * Read the metadata.
473 */ 506 */
474 r = read_exceptions(ps); 507 r = read_exceptions(ps);
@@ -480,7 +513,7 @@ static int persistent_read_metadata(struct exception_store *store)
480} 513}
481 514
482static int persistent_prepare(struct exception_store *store, 515static int persistent_prepare(struct exception_store *store,
483 struct exception *e) 516 struct dm_snap_exception *e)
484{ 517{
485 struct pstore *ps = get_info(store); 518 struct pstore *ps = get_info(store);
486 uint32_t stride; 519 uint32_t stride;
@@ -505,7 +538,7 @@ static int persistent_prepare(struct exception_store *store,
505} 538}
506 539
507static void persistent_commit(struct exception_store *store, 540static void persistent_commit(struct exception_store *store,
508 struct exception *e, 541 struct dm_snap_exception *e,
509 void (*callback) (void *, int success), 542 void (*callback) (void *, int success),
510 void *callback_context) 543 void *callback_context)
511{ 544{
@@ -588,6 +621,12 @@ int dm_create_persistent(struct exception_store *store)
588 atomic_set(&ps->pending_count, 0); 621 atomic_set(&ps->pending_count, 0);
589 ps->callbacks = NULL; 622 ps->callbacks = NULL;
590 623
624 ps->metadata_wq = create_singlethread_workqueue("ksnaphd");
625 if (!ps->metadata_wq) {
626 DMERR("couldn't start header metadata update thread");
627 return -ENOMEM;
628 }
629
591 store->destroy = persistent_destroy; 630 store->destroy = persistent_destroy;
592 store->read_metadata = persistent_read_metadata; 631 store->read_metadata = persistent_read_metadata;
593 store->prepare_exception = persistent_prepare; 632 store->prepare_exception = persistent_prepare;
@@ -616,7 +655,8 @@ static int transient_read_metadata(struct exception_store *store)
616 return 0; 655 return 0;
617} 656}
618 657
619static int transient_prepare(struct exception_store *store, struct exception *e) 658static int transient_prepare(struct exception_store *store,
659 struct dm_snap_exception *e)
620{ 660{
621 struct transient_c *tc = (struct transient_c *) store->context; 661 struct transient_c *tc = (struct transient_c *) store->context;
622 sector_t size = get_dev_size(store->snap->cow->bdev); 662 sector_t size = get_dev_size(store->snap->cow->bdev);
@@ -631,9 +671,9 @@ static int transient_prepare(struct exception_store *store, struct exception *e)
631} 671}
632 672
633static void transient_commit(struct exception_store *store, 673static void transient_commit(struct exception_store *store,
634 struct exception *e, 674 struct dm_snap_exception *e,
635 void (*callback) (void *, int success), 675 void (*callback) (void *, int success),
636 void *callback_context) 676 void *callback_context)
637{ 677{
638 /* Just succeed */ 678 /* Just succeed */
639 callback(callback_context, 1); 679 callback(callback_context, 1);
diff --git a/drivers/md/dm-io.c b/drivers/md/dm-io.c
index 352c6fbeac53..f3a772486437 100644
--- a/drivers/md/dm-io.c
+++ b/drivers/md/dm-io.c
@@ -293,7 +293,10 @@ static void do_region(int rw, unsigned int region, struct io_region *where,
293 * bvec for bio_get/set_region() and decrement bi_max_vecs 293 * bvec for bio_get/set_region() and decrement bi_max_vecs
294 * to hide it from bio_add_page(). 294 * to hide it from bio_add_page().
295 */ 295 */
296 num_bvecs = (remaining / (PAGE_SIZE >> SECTOR_SHIFT)) + 2; 296 num_bvecs = dm_sector_div_up(remaining,
297 (PAGE_SIZE >> SECTOR_SHIFT));
298 num_bvecs = 1 + min_t(int, bio_get_nr_vecs(where->bdev),
299 num_bvecs);
297 bio = bio_alloc_bioset(GFP_NOIO, num_bvecs, io->client->bios); 300 bio = bio_alloc_bioset(GFP_NOIO, num_bvecs, io->client->bios);
298 bio->bi_sector = where->sector + (where->count - remaining); 301 bio->bi_sector = where->sector + (where->count - remaining);
299 bio->bi_bdev = where->bdev; 302 bio->bi_bdev = where->bdev;
diff --git a/drivers/md/dm-mpath-rdac.c b/drivers/md/dm-mpath-rdac.c
new file mode 100644
index 000000000000..8b776b8cb7f7
--- /dev/null
+++ b/drivers/md/dm-mpath-rdac.c
@@ -0,0 +1,700 @@
1/*
2 * Engenio/LSI RDAC DM HW handler
3 *
4 * Copyright (C) 2005 Mike Christie. All rights reserved.
5 * Copyright (C) Chandra Seetharaman, IBM Corp. 2007
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
20 *
21 */
22#include <scsi/scsi.h>
23#include <scsi/scsi_cmnd.h>
24#include <scsi/scsi_eh.h>
25
26#define DM_MSG_PREFIX "multipath rdac"
27
28#include "dm.h"
29#include "dm-hw-handler.h"
30
31#define RDAC_DM_HWH_NAME "rdac"
32#define RDAC_DM_HWH_VER "0.4"
33
34/*
35 * LSI mode page stuff
36 *
37 * These struct definitions and the forming of the
38 * mode page were taken from the LSI RDAC 2.4 GPL'd
39 * driver, and then converted to Linux conventions.
40 */
41#define RDAC_QUIESCENCE_TIME 20;
42/*
43 * Page Codes
44 */
45#define RDAC_PAGE_CODE_REDUNDANT_CONTROLLER 0x2c
46
47/*
48 * Controller modes definitions
49 */
50#define RDAC_MODE_TRANSFER_ALL_LUNS 0x01
51#define RDAC_MODE_TRANSFER_SPECIFIED_LUNS 0x02
52
53/*
54 * RDAC Options field
55 */
56#define RDAC_FORCED_QUIESENCE 0x02
57
58#define RDAC_FAILOVER_TIMEOUT (60 * HZ)
59
60struct rdac_mode_6_hdr {
61 u8 data_len;
62 u8 medium_type;
63 u8 device_params;
64 u8 block_desc_len;
65};
66
67struct rdac_mode_10_hdr {
68 u16 data_len;
69 u8 medium_type;
70 u8 device_params;
71 u16 reserved;
72 u16 block_desc_len;
73};
74
75struct rdac_mode_common {
76 u8 controller_serial[16];
77 u8 alt_controller_serial[16];
78 u8 rdac_mode[2];
79 u8 alt_rdac_mode[2];
80 u8 quiescence_timeout;
81 u8 rdac_options;
82};
83
84struct rdac_pg_legacy {
85 struct rdac_mode_6_hdr hdr;
86 u8 page_code;
87 u8 page_len;
88 struct rdac_mode_common common;
89#define MODE6_MAX_LUN 32
90 u8 lun_table[MODE6_MAX_LUN];
91 u8 reserved2[32];
92 u8 reserved3;
93 u8 reserved4;
94};
95
96struct rdac_pg_expanded {
97 struct rdac_mode_10_hdr hdr;
98 u8 page_code;
99 u8 subpage_code;
100 u8 page_len[2];
101 struct rdac_mode_common common;
102 u8 lun_table[256];
103 u8 reserved3;
104 u8 reserved4;
105};
106
107struct c9_inquiry {
108 u8 peripheral_info;
109 u8 page_code; /* 0xC9 */
110 u8 reserved1;
111 u8 page_len;
112 u8 page_id[4]; /* "vace" */
113 u8 avte_cvp;
114 u8 path_prio;
115 u8 reserved2[38];
116};
117
118#define SUBSYS_ID_LEN 16
119#define SLOT_ID_LEN 2
120
121struct c4_inquiry {
122 u8 peripheral_info;
123 u8 page_code; /* 0xC4 */
124 u8 reserved1;
125 u8 page_len;
126 u8 page_id[4]; /* "subs" */
127 u8 subsys_id[SUBSYS_ID_LEN];
128 u8 revision[4];
129 u8 slot_id[SLOT_ID_LEN];
130 u8 reserved[2];
131};
132
133struct rdac_controller {
134 u8 subsys_id[SUBSYS_ID_LEN];
135 u8 slot_id[SLOT_ID_LEN];
136 int use_10_ms;
137 struct kref kref;
138 struct list_head node; /* list of all controllers */
139 spinlock_t lock;
140 int submitted;
141 struct list_head cmd_list; /* list of commands to be submitted */
142 union {
143 struct rdac_pg_legacy legacy;
144 struct rdac_pg_expanded expanded;
145 } mode_select;
146};
147struct c8_inquiry {
148 u8 peripheral_info;
149 u8 page_code; /* 0xC8 */
150 u8 reserved1;
151 u8 page_len;
152 u8 page_id[4]; /* "edid" */
153 u8 reserved2[3];
154 u8 vol_uniq_id_len;
155 u8 vol_uniq_id[16];
156 u8 vol_user_label_len;
157 u8 vol_user_label[60];
158 u8 array_uniq_id_len;
159 u8 array_unique_id[16];
160 u8 array_user_label_len;
161 u8 array_user_label[60];
162 u8 lun[8];
163};
164
165struct c2_inquiry {
166 u8 peripheral_info;
167 u8 page_code; /* 0xC2 */
168 u8 reserved1;
169 u8 page_len;
170 u8 page_id[4]; /* "swr4" */
171 u8 sw_version[3];
172 u8 sw_date[3];
173 u8 features_enabled;
174 u8 max_lun_supported;
175 u8 partitions[239]; /* Total allocation length should be 0xFF */
176};
177
178struct rdac_handler {
179 struct list_head entry; /* list waiting to submit MODE SELECT */
180 unsigned timeout;
181 struct rdac_controller *ctlr;
182#define UNINITIALIZED_LUN (1 << 8)
183 unsigned lun;
184 unsigned char sense[SCSI_SENSE_BUFFERSIZE];
185 struct dm_path *path;
186 struct work_struct work;
187#define SEND_C2_INQUIRY 1
188#define SEND_C4_INQUIRY 2
189#define SEND_C8_INQUIRY 3
190#define SEND_C9_INQUIRY 4
191#define SEND_MODE_SELECT 5
192 int cmd_to_send;
193 union {
194 struct c2_inquiry c2;
195 struct c4_inquiry c4;
196 struct c8_inquiry c8;
197 struct c9_inquiry c9;
198 } inq;
199};
200
201static LIST_HEAD(ctlr_list);
202static DEFINE_SPINLOCK(list_lock);
203static struct workqueue_struct *rdac_wkqd;
204
205static inline int had_failures(struct request *req, int error)
206{
207 return (error || host_byte(req->errors) != DID_OK ||
208 msg_byte(req->errors) != COMMAND_COMPLETE);
209}
210
211static void rdac_resubmit_all(struct rdac_handler *h)
212{
213 struct rdac_controller *ctlr = h->ctlr;
214 struct rdac_handler *tmp, *h1;
215
216 spin_lock(&ctlr->lock);
217 list_for_each_entry_safe(h1, tmp, &ctlr->cmd_list, entry) {
218 h1->cmd_to_send = SEND_C9_INQUIRY;
219 queue_work(rdac_wkqd, &h1->work);
220 list_del(&h1->entry);
221 }
222 ctlr->submitted = 0;
223 spin_unlock(&ctlr->lock);
224}
225
226static void mode_select_endio(struct request *req, int error)
227{
228 struct rdac_handler *h = req->end_io_data;
229 struct scsi_sense_hdr sense_hdr;
230 int sense = 0, fail = 0;
231
232 if (had_failures(req, error)) {
233 fail = 1;
234 goto failed;
235 }
236
237 if (status_byte(req->errors) == CHECK_CONDITION) {
238 scsi_normalize_sense(req->sense, SCSI_SENSE_BUFFERSIZE,
239 &sense_hdr);
240 sense = (sense_hdr.sense_key << 16) | (sense_hdr.asc << 8) |
241 sense_hdr.ascq;
242 /* If it is retryable failure, submit the c9 inquiry again */
243 if (sense == 0x59136 || sense == 0x68b02 || sense == 0xb8b02 ||
244 sense == 0x62900) {
245 /* 0x59136 - Command lock contention
246 * 0x[6b]8b02 - Quiesense in progress or achieved
247 * 0x62900 - Power On, Reset, or Bus Device Reset
248 */
249 h->cmd_to_send = SEND_C9_INQUIRY;
250 queue_work(rdac_wkqd, &h->work);
251 goto done;
252 }
253 if (sense)
254 DMINFO("MODE_SELECT failed on %s with sense 0x%x",
255 h->path->dev->name, sense);
256 }
257failed:
258 if (fail || sense)
259 dm_pg_init_complete(h->path, MP_FAIL_PATH);
260 else
261 dm_pg_init_complete(h->path, 0);
262
263done:
264 rdac_resubmit_all(h);
265 __blk_put_request(req->q, req);
266}
267
268static struct request *get_rdac_req(struct rdac_handler *h,
269 void *buffer, unsigned buflen, int rw)
270{
271 struct request *rq;
272 struct request_queue *q = bdev_get_queue(h->path->dev->bdev);
273
274 rq = blk_get_request(q, rw, GFP_KERNEL);
275
276 if (!rq) {
277 DMINFO("get_rdac_req: blk_get_request failed");
278 return NULL;
279 }
280
281 if (buflen && blk_rq_map_kern(q, rq, buffer, buflen, GFP_KERNEL)) {
282 blk_put_request(rq);
283 DMINFO("get_rdac_req: blk_rq_map_kern failed");
284 return NULL;
285 }
286
287 memset(&rq->cmd, 0, BLK_MAX_CDB);
288 rq->sense = h->sense;
289 memset(rq->sense, 0, SCSI_SENSE_BUFFERSIZE);
290 rq->sense_len = 0;
291
292 rq->end_io_data = h;
293 rq->timeout = h->timeout;
294 rq->cmd_type = REQ_TYPE_BLOCK_PC;
295 rq->cmd_flags = REQ_FAILFAST | REQ_NOMERGE;
296 return rq;
297}
298
299static struct request *rdac_failover_get(struct rdac_handler *h)
300{
301 struct request *rq;
302 struct rdac_mode_common *common;
303 unsigned data_size;
304
305 if (h->ctlr->use_10_ms) {
306 struct rdac_pg_expanded *rdac_pg;
307
308 data_size = sizeof(struct rdac_pg_expanded);
309 rdac_pg = &h->ctlr->mode_select.expanded;
310 memset(rdac_pg, 0, data_size);
311 common = &rdac_pg->common;
312 rdac_pg->page_code = RDAC_PAGE_CODE_REDUNDANT_CONTROLLER + 0x40;
313 rdac_pg->subpage_code = 0x1;
314 rdac_pg->page_len[0] = 0x01;
315 rdac_pg->page_len[1] = 0x28;
316 rdac_pg->lun_table[h->lun] = 0x81;
317 } else {
318 struct rdac_pg_legacy *rdac_pg;
319
320 data_size = sizeof(struct rdac_pg_legacy);
321 rdac_pg = &h->ctlr->mode_select.legacy;
322 memset(rdac_pg, 0, data_size);
323 common = &rdac_pg->common;
324 rdac_pg->page_code = RDAC_PAGE_CODE_REDUNDANT_CONTROLLER;
325 rdac_pg->page_len = 0x68;
326 rdac_pg->lun_table[h->lun] = 0x81;
327 }
328 common->rdac_mode[1] = RDAC_MODE_TRANSFER_SPECIFIED_LUNS;
329 common->quiescence_timeout = RDAC_QUIESCENCE_TIME;
330 common->rdac_options = RDAC_FORCED_QUIESENCE;
331
332 /* get request for block layer packet command */
333 rq = get_rdac_req(h, &h->ctlr->mode_select, data_size, WRITE);
334 if (!rq) {
335 DMERR("rdac_failover_get: no rq");
336 return NULL;
337 }
338
339 /* Prepare the command. */
340 if (h->ctlr->use_10_ms) {
341 rq->cmd[0] = MODE_SELECT_10;
342 rq->cmd[7] = data_size >> 8;
343 rq->cmd[8] = data_size & 0xff;
344 } else {
345 rq->cmd[0] = MODE_SELECT;
346 rq->cmd[4] = data_size;
347 }
348 rq->cmd_len = COMMAND_SIZE(rq->cmd[0]);
349
350 return rq;
351}
352
353/* Acquires h->ctlr->lock */
354static void submit_mode_select(struct rdac_handler *h)
355{
356 struct request *rq;
357 struct request_queue *q = bdev_get_queue(h->path->dev->bdev);
358
359 spin_lock(&h->ctlr->lock);
360 if (h->ctlr->submitted) {
361 list_add(&h->entry, &h->ctlr->cmd_list);
362 goto drop_lock;
363 }
364
365 if (!q) {
366 DMINFO("submit_mode_select: no queue");
367 goto fail_path;
368 }
369
370 rq = rdac_failover_get(h);
371 if (!rq) {
372 DMERR("submit_mode_select: no rq");
373 goto fail_path;
374 }
375
376 DMINFO("queueing MODE_SELECT command on %s", h->path->dev->name);
377
378 blk_execute_rq_nowait(q, NULL, rq, 1, mode_select_endio);
379 h->ctlr->submitted = 1;
380 goto drop_lock;
381fail_path:
382 dm_pg_init_complete(h->path, MP_FAIL_PATH);
383drop_lock:
384 spin_unlock(&h->ctlr->lock);
385}
386
387static void release_ctlr(struct kref *kref)
388{
389 struct rdac_controller *ctlr;
390 ctlr = container_of(kref, struct rdac_controller, kref);
391
392 spin_lock(&list_lock);
393 list_del(&ctlr->node);
394 spin_unlock(&list_lock);
395 kfree(ctlr);
396}
397
398static struct rdac_controller *get_controller(u8 *subsys_id, u8 *slot_id)
399{
400 struct rdac_controller *ctlr, *tmp;
401
402 spin_lock(&list_lock);
403
404 list_for_each_entry(tmp, &ctlr_list, node) {
405 if ((memcmp(tmp->subsys_id, subsys_id, SUBSYS_ID_LEN) == 0) &&
406 (memcmp(tmp->slot_id, slot_id, SLOT_ID_LEN) == 0)) {
407 kref_get(&tmp->kref);
408 spin_unlock(&list_lock);
409 return tmp;
410 }
411 }
412 ctlr = kmalloc(sizeof(*ctlr), GFP_ATOMIC);
413 if (!ctlr)
414 goto done;
415
416 /* initialize fields of controller */
417 memcpy(ctlr->subsys_id, subsys_id, SUBSYS_ID_LEN);
418 memcpy(ctlr->slot_id, slot_id, SLOT_ID_LEN);
419 kref_init(&ctlr->kref);
420 spin_lock_init(&ctlr->lock);
421 ctlr->submitted = 0;
422 ctlr->use_10_ms = -1;
423 INIT_LIST_HEAD(&ctlr->cmd_list);
424 list_add(&ctlr->node, &ctlr_list);
425done:
426 spin_unlock(&list_lock);
427 return ctlr;
428}
429
430static void c4_endio(struct request *req, int error)
431{
432 struct rdac_handler *h = req->end_io_data;
433 struct c4_inquiry *sp;
434
435 if (had_failures(req, error)) {
436 dm_pg_init_complete(h->path, MP_FAIL_PATH);
437 goto done;
438 }
439
440 sp = &h->inq.c4;
441
442 h->ctlr = get_controller(sp->subsys_id, sp->slot_id);
443
444 if (h->ctlr) {
445 h->cmd_to_send = SEND_C9_INQUIRY;
446 queue_work(rdac_wkqd, &h->work);
447 } else
448 dm_pg_init_complete(h->path, MP_FAIL_PATH);
449done:
450 __blk_put_request(req->q, req);
451}
452
453static void c2_endio(struct request *req, int error)
454{
455 struct rdac_handler *h = req->end_io_data;
456 struct c2_inquiry *sp;
457
458 if (had_failures(req, error)) {
459 dm_pg_init_complete(h->path, MP_FAIL_PATH);
460 goto done;
461 }
462
463 sp = &h->inq.c2;
464
465 /* If more than MODE6_MAX_LUN luns are supported, use mode select 10 */
466 if (sp->max_lun_supported >= MODE6_MAX_LUN)
467 h->ctlr->use_10_ms = 1;
468 else
469 h->ctlr->use_10_ms = 0;
470
471 h->cmd_to_send = SEND_MODE_SELECT;
472 queue_work(rdac_wkqd, &h->work);
473done:
474 __blk_put_request(req->q, req);
475}
476
477static void c9_endio(struct request *req, int error)
478{
479 struct rdac_handler *h = req->end_io_data;
480 struct c9_inquiry *sp;
481
482 if (had_failures(req, error)) {
483 dm_pg_init_complete(h->path, MP_FAIL_PATH);
484 goto done;
485 }
486
487 /* We need to look at the sense keys here to take clear action.
488 * For now simple logic: If the host is in AVT mode or if controller
489 * owns the lun, return dm_pg_init_complete(), otherwise submit
490 * MODE SELECT.
491 */
492 sp = &h->inq.c9;
493
494 /* If in AVT mode, return success */
495 if ((sp->avte_cvp >> 7) == 0x1) {
496 dm_pg_init_complete(h->path, 0);
497 goto done;
498 }
499
500 /* If the controller on this path owns the LUN, return success */
501 if (sp->avte_cvp & 0x1) {
502 dm_pg_init_complete(h->path, 0);
503 goto done;
504 }
505
506 if (h->ctlr) {
507 if (h->ctlr->use_10_ms == -1)
508 h->cmd_to_send = SEND_C2_INQUIRY;
509 else
510 h->cmd_to_send = SEND_MODE_SELECT;
511 } else
512 h->cmd_to_send = SEND_C4_INQUIRY;
513 queue_work(rdac_wkqd, &h->work);
514done:
515 __blk_put_request(req->q, req);
516}
517
518static void c8_endio(struct request *req, int error)
519{
520 struct rdac_handler *h = req->end_io_data;
521 struct c8_inquiry *sp;
522
523 if (had_failures(req, error)) {
524 dm_pg_init_complete(h->path, MP_FAIL_PATH);
525 goto done;
526 }
527
528 /* We need to look at the sense keys here to take clear action.
529 * For now simple logic: Get the lun from the inquiry page.
530 */
531 sp = &h->inq.c8;
532 h->lun = sp->lun[7]; /* currently it uses only one byte */
533 h->cmd_to_send = SEND_C9_INQUIRY;
534 queue_work(rdac_wkqd, &h->work);
535done:
536 __blk_put_request(req->q, req);
537}
538
539static void submit_inquiry(struct rdac_handler *h, int page_code,
540 unsigned int len, rq_end_io_fn endio)
541{
542 struct request *rq;
543 struct request_queue *q = bdev_get_queue(h->path->dev->bdev);
544
545 if (!q)
546 goto fail_path;
547
548 rq = get_rdac_req(h, &h->inq, len, READ);
549 if (!rq)
550 goto fail_path;
551
552 /* Prepare the command. */
553 rq->cmd[0] = INQUIRY;
554 rq->cmd[1] = 1;
555 rq->cmd[2] = page_code;
556 rq->cmd[4] = len;
557 rq->cmd_len = COMMAND_SIZE(INQUIRY);
558 blk_execute_rq_nowait(q, NULL, rq, 1, endio);
559 return;
560
561fail_path:
562 dm_pg_init_complete(h->path, MP_FAIL_PATH);
563}
564
565static void service_wkq(struct work_struct *work)
566{
567 struct rdac_handler *h = container_of(work, struct rdac_handler, work);
568
569 switch (h->cmd_to_send) {
570 case SEND_C2_INQUIRY:
571 submit_inquiry(h, 0xC2, sizeof(struct c2_inquiry), c2_endio);
572 break;
573 case SEND_C4_INQUIRY:
574 submit_inquiry(h, 0xC4, sizeof(struct c4_inquiry), c4_endio);
575 break;
576 case SEND_C8_INQUIRY:
577 submit_inquiry(h, 0xC8, sizeof(struct c8_inquiry), c8_endio);
578 break;
579 case SEND_C9_INQUIRY:
580 submit_inquiry(h, 0xC9, sizeof(struct c9_inquiry), c9_endio);
581 break;
582 case SEND_MODE_SELECT:
583 submit_mode_select(h);
584 break;
585 default:
586 BUG();
587 }
588}
589/*
590 * only support subpage2c until we confirm that this is just a matter of
591 * of updating firmware or not, and RDAC (basic AVT works already) for now
592 * but we can add these in in when we get time and testers
593 */
594static int rdac_create(struct hw_handler *hwh, unsigned argc, char **argv)
595{
596 struct rdac_handler *h;
597 unsigned timeout;
598
599 if (argc == 0) {
600 /* No arguments: use defaults */
601 timeout = RDAC_FAILOVER_TIMEOUT;
602 } else if (argc != 1) {
603 DMWARN("incorrect number of arguments");
604 return -EINVAL;
605 } else {
606 if (sscanf(argv[1], "%u", &timeout) != 1) {
607 DMWARN("invalid timeout value");
608 return -EINVAL;
609 }
610 }
611
612 h = kzalloc(sizeof(*h), GFP_KERNEL);
613 if (!h)
614 return -ENOMEM;
615
616 hwh->context = h;
617 h->timeout = timeout;
618 h->lun = UNINITIALIZED_LUN;
619 INIT_WORK(&h->work, service_wkq);
620 DMWARN("using RDAC command with timeout %u", h->timeout);
621
622 return 0;
623}
624
625static void rdac_destroy(struct hw_handler *hwh)
626{
627 struct rdac_handler *h = hwh->context;
628
629 if (h->ctlr)
630 kref_put(&h->ctlr->kref, release_ctlr);
631 kfree(h);
632 hwh->context = NULL;
633}
634
635static unsigned rdac_error(struct hw_handler *hwh, struct bio *bio)
636{
637 /* Try default handler */
638 return dm_scsi_err_handler(hwh, bio);
639}
640
641static void rdac_pg_init(struct hw_handler *hwh, unsigned bypassed,
642 struct dm_path *path)
643{
644 struct rdac_handler *h = hwh->context;
645
646 h->path = path;
647 switch (h->lun) {
648 case UNINITIALIZED_LUN:
649 submit_inquiry(h, 0xC8, sizeof(struct c8_inquiry), c8_endio);
650 break;
651 default:
652 submit_inquiry(h, 0xC9, sizeof(struct c9_inquiry), c9_endio);
653 }
654}
655
656static struct hw_handler_type rdac_handler = {
657 .name = RDAC_DM_HWH_NAME,
658 .module = THIS_MODULE,
659 .create = rdac_create,
660 .destroy = rdac_destroy,
661 .pg_init = rdac_pg_init,
662 .error = rdac_error,
663};
664
665static int __init rdac_init(void)
666{
667 int r = dm_register_hw_handler(&rdac_handler);
668
669 if (r < 0) {
670 DMERR("%s: register failed %d", RDAC_DM_HWH_NAME, r);
671 return r;
672 }
673
674 rdac_wkqd = create_singlethread_workqueue("rdac_wkqd");
675 if (!rdac_wkqd) {
676 DMERR("Failed to create workqueue rdac_wkqd.");
677 dm_unregister_hw_handler(&rdac_handler);
678 return -ENOMEM;
679 }
680
681 DMINFO("%s: version %s loaded", RDAC_DM_HWH_NAME, RDAC_DM_HWH_VER);
682 return 0;
683}
684
685static void __exit rdac_exit(void)
686{
687 int r = dm_unregister_hw_handler(&rdac_handler);
688
689 destroy_workqueue(rdac_wkqd);
690 if (r < 0)
691 DMERR("%s: unregister failed %d", RDAC_DM_HWH_NAME, r);
692}
693
694module_init(rdac_init);
695module_exit(rdac_exit);
696
697MODULE_DESCRIPTION("DM Multipath LSI/Engenio RDAC support");
698MODULE_AUTHOR("Mike Christie, Chandra Seetharaman");
699MODULE_LICENSE("GPL");
700MODULE_VERSION(RDAC_DM_HWH_VER);
diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index de54b39e6ffe..d6ca9d0a6fd1 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -83,7 +83,7 @@ struct multipath {
83 struct work_struct trigger_event; 83 struct work_struct trigger_event;
84 84
85 /* 85 /*
86 * We must use a mempool of mpath_io structs so that we 86 * We must use a mempool of dm_mpath_io structs so that we
87 * can resubmit bios on error. 87 * can resubmit bios on error.
88 */ 88 */
89 mempool_t *mpio_pool; 89 mempool_t *mpio_pool;
@@ -92,7 +92,7 @@ struct multipath {
92/* 92/*
93 * Context information attached to each bio we process. 93 * Context information attached to each bio we process.
94 */ 94 */
95struct mpath_io { 95struct dm_mpath_io {
96 struct pgpath *pgpath; 96 struct pgpath *pgpath;
97 struct dm_bio_details details; 97 struct dm_bio_details details;
98}; 98};
@@ -122,7 +122,7 @@ static struct pgpath *alloc_pgpath(void)
122 return pgpath; 122 return pgpath;
123} 123}
124 124
125static inline void free_pgpath(struct pgpath *pgpath) 125static void free_pgpath(struct pgpath *pgpath)
126{ 126{
127 kfree(pgpath); 127 kfree(pgpath);
128} 128}
@@ -299,8 +299,8 @@ static int __must_push_back(struct multipath *m)
299 dm_noflush_suspending(m->ti)); 299 dm_noflush_suspending(m->ti));
300} 300}
301 301
302static int map_io(struct multipath *m, struct bio *bio, struct mpath_io *mpio, 302static int map_io(struct multipath *m, struct bio *bio,
303 unsigned was_queued) 303 struct dm_mpath_io *mpio, unsigned was_queued)
304{ 304{
305 int r = DM_MAPIO_REMAPPED; 305 int r = DM_MAPIO_REMAPPED;
306 unsigned long flags; 306 unsigned long flags;
@@ -374,7 +374,7 @@ static void dispatch_queued_ios(struct multipath *m)
374 int r; 374 int r;
375 unsigned long flags; 375 unsigned long flags;
376 struct bio *bio = NULL, *next; 376 struct bio *bio = NULL, *next;
377 struct mpath_io *mpio; 377 struct dm_mpath_io *mpio;
378 union map_info *info; 378 union map_info *info;
379 379
380 spin_lock_irqsave(&m->lock, flags); 380 spin_lock_irqsave(&m->lock, flags);
@@ -795,12 +795,9 @@ static int multipath_map(struct dm_target *ti, struct bio *bio,
795 union map_info *map_context) 795 union map_info *map_context)
796{ 796{
797 int r; 797 int r;
798 struct mpath_io *mpio; 798 struct dm_mpath_io *mpio;
799 struct multipath *m = (struct multipath *) ti->private; 799 struct multipath *m = (struct multipath *) ti->private;
800 800
801 if (bio_barrier(bio))
802 return -EOPNOTSUPP;
803
804 mpio = mempool_alloc(m->mpio_pool, GFP_NOIO); 801 mpio = mempool_alloc(m->mpio_pool, GFP_NOIO);
805 dm_bio_record(&mpio->details, bio); 802 dm_bio_record(&mpio->details, bio);
806 803
@@ -1014,7 +1011,7 @@ void dm_pg_init_complete(struct dm_path *path, unsigned err_flags)
1014 * end_io handling 1011 * end_io handling
1015 */ 1012 */
1016static int do_end_io(struct multipath *m, struct bio *bio, 1013static int do_end_io(struct multipath *m, struct bio *bio,
1017 int error, struct mpath_io *mpio) 1014 int error, struct dm_mpath_io *mpio)
1018{ 1015{
1019 struct hw_handler *hwh = &m->hw_handler; 1016 struct hw_handler *hwh = &m->hw_handler;
1020 unsigned err_flags = MP_FAIL_PATH; /* Default behavior */ 1017 unsigned err_flags = MP_FAIL_PATH; /* Default behavior */
@@ -1075,8 +1072,8 @@ static int do_end_io(struct multipath *m, struct bio *bio,
1075static int multipath_end_io(struct dm_target *ti, struct bio *bio, 1072static int multipath_end_io(struct dm_target *ti, struct bio *bio,
1076 int error, union map_info *map_context) 1073 int error, union map_info *map_context)
1077{ 1074{
1078 struct multipath *m = (struct multipath *) ti->private; 1075 struct multipath *m = ti->private;
1079 struct mpath_io *mpio = (struct mpath_io *) map_context->ptr; 1076 struct dm_mpath_io *mpio = map_context->ptr;
1080 struct pgpath *pgpath = mpio->pgpath; 1077 struct pgpath *pgpath = mpio->pgpath;
1081 struct path_selector *ps; 1078 struct path_selector *ps;
1082 int r; 1079 int r;
@@ -1346,22 +1343,20 @@ static int __init dm_multipath_init(void)
1346 int r; 1343 int r;
1347 1344
1348 /* allocate a slab for the dm_ios */ 1345 /* allocate a slab for the dm_ios */
1349 _mpio_cache = kmem_cache_create("dm_mpath", sizeof(struct mpath_io), 1346 _mpio_cache = KMEM_CACHE(dm_mpath_io, 0);
1350 0, 0, NULL, NULL);
1351 if (!_mpio_cache) 1347 if (!_mpio_cache)
1352 return -ENOMEM; 1348 return -ENOMEM;
1353 1349
1354 r = dm_register_target(&multipath_target); 1350 r = dm_register_target(&multipath_target);
1355 if (r < 0) { 1351 if (r < 0) {
1356 DMERR("%s: register failed %d", multipath_target.name, r); 1352 DMERR("register failed %d", r);
1357 kmem_cache_destroy(_mpio_cache); 1353 kmem_cache_destroy(_mpio_cache);
1358 return -EINVAL; 1354 return -EINVAL;
1359 } 1355 }
1360 1356
1361 kmultipathd = create_workqueue("kmpathd"); 1357 kmultipathd = create_workqueue("kmpathd");
1362 if (!kmultipathd) { 1358 if (!kmultipathd) {
1363 DMERR("%s: failed to create workqueue kmpathd", 1359 DMERR("failed to create workqueue kmpathd");
1364 multipath_target.name);
1365 dm_unregister_target(&multipath_target); 1360 dm_unregister_target(&multipath_target);
1366 kmem_cache_destroy(_mpio_cache); 1361 kmem_cache_destroy(_mpio_cache);
1367 return -ENOMEM; 1362 return -ENOMEM;
@@ -1382,8 +1377,7 @@ static void __exit dm_multipath_exit(void)
1382 1377
1383 r = dm_unregister_target(&multipath_target); 1378 r = dm_unregister_target(&multipath_target);
1384 if (r < 0) 1379 if (r < 0)
1385 DMERR("%s: target unregister failed %d", 1380 DMERR("target unregister failed %d", r);
1386 multipath_target.name, r);
1387 kmem_cache_destroy(_mpio_cache); 1381 kmem_cache_destroy(_mpio_cache);
1388} 1382}
1389 1383
diff --git a/drivers/md/dm-raid1.c b/drivers/md/dm-raid1.c
index ef124b71ccc8..1a876f9965e0 100644
--- a/drivers/md/dm-raid1.c
+++ b/drivers/md/dm-raid1.c
@@ -24,6 +24,7 @@
24#define DM_IO_PAGES 64 24#define DM_IO_PAGES 64
25 25
26#define DM_RAID1_HANDLE_ERRORS 0x01 26#define DM_RAID1_HANDLE_ERRORS 0x01
27#define errors_handled(p) ((p)->features & DM_RAID1_HANDLE_ERRORS)
27 28
28static DECLARE_WAIT_QUEUE_HEAD(_kmirrord_recovery_stopped); 29static DECLARE_WAIT_QUEUE_HEAD(_kmirrord_recovery_stopped);
29 30
@@ -85,6 +86,7 @@ struct region_hash {
85 struct list_head clean_regions; 86 struct list_head clean_regions;
86 struct list_head quiesced_regions; 87 struct list_head quiesced_regions;
87 struct list_head recovered_regions; 88 struct list_head recovered_regions;
89 struct list_head failed_recovered_regions;
88}; 90};
89 91
90enum { 92enum {
@@ -132,6 +134,7 @@ struct mirror_set {
132 /* recovery */ 134 /* recovery */
133 region_t nr_regions; 135 region_t nr_regions;
134 int in_sync; 136 int in_sync;
137 int log_failure;
135 138
136 struct mirror *default_mirror; /* Default mirror */ 139 struct mirror *default_mirror; /* Default mirror */
137 140
@@ -204,6 +207,7 @@ static int rh_init(struct region_hash *rh, struct mirror_set *ms,
204 INIT_LIST_HEAD(&rh->clean_regions); 207 INIT_LIST_HEAD(&rh->clean_regions);
205 INIT_LIST_HEAD(&rh->quiesced_regions); 208 INIT_LIST_HEAD(&rh->quiesced_regions);
206 INIT_LIST_HEAD(&rh->recovered_regions); 209 INIT_LIST_HEAD(&rh->recovered_regions);
210 INIT_LIST_HEAD(&rh->failed_recovered_regions);
207 211
208 rh->region_pool = mempool_create_kmalloc_pool(MIN_REGIONS, 212 rh->region_pool = mempool_create_kmalloc_pool(MIN_REGIONS,
209 sizeof(struct region)); 213 sizeof(struct region));
@@ -368,6 +372,7 @@ static void rh_update_states(struct region_hash *rh)
368 372
369 LIST_HEAD(clean); 373 LIST_HEAD(clean);
370 LIST_HEAD(recovered); 374 LIST_HEAD(recovered);
375 LIST_HEAD(failed_recovered);
371 376
372 /* 377 /*
373 * Quickly grab the lists. 378 * Quickly grab the lists.
@@ -378,10 +383,8 @@ static void rh_update_states(struct region_hash *rh)
378 list_splice(&rh->clean_regions, &clean); 383 list_splice(&rh->clean_regions, &clean);
379 INIT_LIST_HEAD(&rh->clean_regions); 384 INIT_LIST_HEAD(&rh->clean_regions);
380 385
381 list_for_each_entry (reg, &clean, list) { 386 list_for_each_entry(reg, &clean, list)
382 rh->log->type->clear_region(rh->log, reg->key);
383 list_del(&reg->hash_list); 387 list_del(&reg->hash_list);
384 }
385 } 388 }
386 389
387 if (!list_empty(&rh->recovered_regions)) { 390 if (!list_empty(&rh->recovered_regions)) {
@@ -391,6 +394,15 @@ static void rh_update_states(struct region_hash *rh)
391 list_for_each_entry (reg, &recovered, list) 394 list_for_each_entry (reg, &recovered, list)
392 list_del(&reg->hash_list); 395 list_del(&reg->hash_list);
393 } 396 }
397
398 if (!list_empty(&rh->failed_recovered_regions)) {
399 list_splice(&rh->failed_recovered_regions, &failed_recovered);
400 INIT_LIST_HEAD(&rh->failed_recovered_regions);
401
402 list_for_each_entry(reg, &failed_recovered, list)
403 list_del(&reg->hash_list);
404 }
405
394 spin_unlock(&rh->region_lock); 406 spin_unlock(&rh->region_lock);
395 write_unlock_irq(&rh->hash_lock); 407 write_unlock_irq(&rh->hash_lock);
396 408
@@ -405,10 +417,17 @@ static void rh_update_states(struct region_hash *rh)
405 mempool_free(reg, rh->region_pool); 417 mempool_free(reg, rh->region_pool);
406 } 418 }
407 419
408 rh->log->type->flush(rh->log); 420 list_for_each_entry_safe(reg, next, &failed_recovered, list) {
421 complete_resync_work(reg, errors_handled(rh->ms) ? 0 : 1);
422 mempool_free(reg, rh->region_pool);
423 }
409 424
410 list_for_each_entry_safe (reg, next, &clean, list) 425 list_for_each_entry_safe(reg, next, &clean, list) {
426 rh->log->type->clear_region(rh->log, reg->key);
411 mempool_free(reg, rh->region_pool); 427 mempool_free(reg, rh->region_pool);
428 }
429
430 rh->log->type->flush(rh->log);
412} 431}
413 432
414static void rh_inc(struct region_hash *rh, region_t region) 433static void rh_inc(struct region_hash *rh, region_t region)
@@ -555,21 +574,25 @@ static struct region *rh_recovery_start(struct region_hash *rh)
555 return reg; 574 return reg;
556} 575}
557 576
558/* FIXME: success ignored for now */
559static void rh_recovery_end(struct region *reg, int success) 577static void rh_recovery_end(struct region *reg, int success)
560{ 578{
561 struct region_hash *rh = reg->rh; 579 struct region_hash *rh = reg->rh;
562 580
563 spin_lock_irq(&rh->region_lock); 581 spin_lock_irq(&rh->region_lock);
564 list_add(&reg->list, &reg->rh->recovered_regions); 582 if (success)
583 list_add(&reg->list, &reg->rh->recovered_regions);
584 else {
585 reg->state = RH_NOSYNC;
586 list_add(&reg->list, &reg->rh->failed_recovered_regions);
587 }
565 spin_unlock_irq(&rh->region_lock); 588 spin_unlock_irq(&rh->region_lock);
566 589
567 wake(rh->ms); 590 wake(rh->ms);
568} 591}
569 592
570static void rh_flush(struct region_hash *rh) 593static int rh_flush(struct region_hash *rh)
571{ 594{
572 rh->log->type->flush(rh->log); 595 return rh->log->type->flush(rh->log);
573} 596}
574 597
575static void rh_delay(struct region_hash *rh, struct bio *bio) 598static void rh_delay(struct region_hash *rh, struct bio *bio)
@@ -633,7 +656,14 @@ static void recovery_complete(int read_err, unsigned int write_err,
633{ 656{
634 struct region *reg = (struct region *) context; 657 struct region *reg = (struct region *) context;
635 658
636 /* FIXME: better error handling */ 659 if (read_err)
660 /* Read error means the failure of default mirror. */
661 DMERR_LIMIT("Unable to read primary mirror during recovery");
662
663 if (write_err)
664 DMERR_LIMIT("Write error during recovery (error = 0x%x)",
665 write_err);
666
637 rh_recovery_end(reg, !(read_err || write_err)); 667 rh_recovery_end(reg, !(read_err || write_err));
638} 668}
639 669
@@ -863,12 +893,15 @@ static void do_writes(struct mirror_set *ms, struct bio_list *writes)
863 */ 893 */
864 rh_inc_pending(&ms->rh, &sync); 894 rh_inc_pending(&ms->rh, &sync);
865 rh_inc_pending(&ms->rh, &nosync); 895 rh_inc_pending(&ms->rh, &nosync);
866 rh_flush(&ms->rh); 896 ms->log_failure = rh_flush(&ms->rh) ? 1 : 0;
867 897
868 /* 898 /*
869 * Dispatch io. 899 * Dispatch io.
870 */ 900 */
871 while ((bio = bio_list_pop(&sync))) 901 if (unlikely(ms->log_failure))
902 while ((bio = bio_list_pop(&sync)))
903 bio_endio(bio, bio->bi_size, -EIO);
904 else while ((bio = bio_list_pop(&sync)))
872 do_write(ms, bio); 905 do_write(ms, bio);
873 906
874 while ((bio = bio_list_pop(&recover))) 907 while ((bio = bio_list_pop(&recover)))
@@ -1145,6 +1178,15 @@ static int mirror_ctr(struct dm_target *ti, unsigned int argc, char **argv)
1145 argv += args_used; 1178 argv += args_used;
1146 argc -= args_used; 1179 argc -= args_used;
1147 1180
1181 /*
1182 * Any read-balancing addition depends on the
1183 * DM_RAID1_HANDLE_ERRORS flag being present.
1184 * This is because the decision to balance depends
1185 * on the sync state of a region. If the above
1186 * flag is not present, we ignore errors; and
1187 * the sync state may be inaccurate.
1188 */
1189
1148 if (argc) { 1190 if (argc) {
1149 ti->error = "Too many mirror arguments"; 1191 ti->error = "Too many mirror arguments";
1150 free_context(ms, ti, ms->nr_mirrors); 1192 free_context(ms, ti, ms->nr_mirrors);
@@ -1288,12 +1330,12 @@ static int mirror_status(struct dm_target *ti, status_type_t type,
1288 for (m = 0; m < ms->nr_mirrors; m++) 1330 for (m = 0; m < ms->nr_mirrors; m++)
1289 DMEMIT("%s ", ms->mirror[m].dev->name); 1331 DMEMIT("%s ", ms->mirror[m].dev->name);
1290 1332
1291 DMEMIT("%llu/%llu", 1333 DMEMIT("%llu/%llu 0 ",
1292 (unsigned long long)ms->rh.log->type-> 1334 (unsigned long long)ms->rh.log->type->
1293 get_sync_count(ms->rh.log), 1335 get_sync_count(ms->rh.log),
1294 (unsigned long long)ms->nr_regions); 1336 (unsigned long long)ms->nr_regions);
1295 1337
1296 sz = ms->rh.log->type->status(ms->rh.log, type, result, maxlen); 1338 sz += ms->rh.log->type->status(ms->rh.log, type, result+sz, maxlen-sz);
1297 1339
1298 break; 1340 break;
1299 1341
@@ -1335,8 +1377,7 @@ static int __init dm_mirror_init(void)
1335 1377
1336 r = dm_register_target(&mirror_target); 1378 r = dm_register_target(&mirror_target);
1337 if (r < 0) { 1379 if (r < 0) {
1338 DMERR("%s: Failed to register mirror target", 1380 DMERR("Failed to register mirror target");
1339 mirror_target.name);
1340 dm_dirty_log_exit(); 1381 dm_dirty_log_exit();
1341 } 1382 }
1342 1383
@@ -1349,7 +1390,7 @@ static void __exit dm_mirror_exit(void)
1349 1390
1350 r = dm_unregister_target(&mirror_target); 1391 r = dm_unregister_target(&mirror_target);
1351 if (r < 0) 1392 if (r < 0)
1352 DMERR("%s: unregister failed %d", mirror_target.name, r); 1393 DMERR("unregister failed %d", r);
1353 1394
1354 dm_dirty_log_exit(); 1395 dm_dirty_log_exit();
1355} 1396}
diff --git a/drivers/md/dm-round-robin.c b/drivers/md/dm-round-robin.c
index a348a97b65af..391dfa2ad434 100644
--- a/drivers/md/dm-round-robin.c
+++ b/drivers/md/dm-round-robin.c
@@ -205,7 +205,7 @@ static void __exit dm_rr_exit(void)
205 int r = dm_unregister_path_selector(&rr_ps); 205 int r = dm_unregister_path_selector(&rr_ps);
206 206
207 if (r < 0) 207 if (r < 0)
208 DMERR("round-robin: unregister failed %d", r); 208 DMERR("unregister failed %d", r);
209} 209}
210 210
211module_init(dm_rr_init); 211module_init(dm_rr_init);
diff --git a/drivers/md/dm-snap.c b/drivers/md/dm-snap.c
index 0821a2b68a73..83ddbfe6b8a4 100644
--- a/drivers/md/dm-snap.c
+++ b/drivers/md/dm-snap.c
@@ -42,8 +42,8 @@
42static struct workqueue_struct *ksnapd; 42static struct workqueue_struct *ksnapd;
43static void flush_queued_bios(struct work_struct *work); 43static void flush_queued_bios(struct work_struct *work);
44 44
45struct pending_exception { 45struct dm_snap_pending_exception {
46 struct exception e; 46 struct dm_snap_exception e;
47 47
48 /* 48 /*
49 * Origin buffers waiting for this to complete are held 49 * Origin buffers waiting for this to complete are held
@@ -63,7 +63,7 @@ struct pending_exception {
63 * group of pending_exceptions. It is always last to get freed. 63 * group of pending_exceptions. It is always last to get freed.
64 * These fields get set up when writing to the origin. 64 * These fields get set up when writing to the origin.
65 */ 65 */
66 struct pending_exception *primary_pe; 66 struct dm_snap_pending_exception *primary_pe;
67 67
68 /* 68 /*
69 * Number of pending_exceptions processing this chunk. 69 * Number of pending_exceptions processing this chunk.
@@ -137,7 +137,7 @@ static void exit_origin_hash(void)
137 kfree(_origins); 137 kfree(_origins);
138} 138}
139 139
140static inline unsigned int origin_hash(struct block_device *bdev) 140static unsigned origin_hash(struct block_device *bdev)
141{ 141{
142 return bdev->bd_dev & ORIGIN_MASK; 142 return bdev->bd_dev & ORIGIN_MASK;
143} 143}
@@ -231,7 +231,7 @@ static int init_exception_table(struct exception_table *et, uint32_t size)
231static void exit_exception_table(struct exception_table *et, struct kmem_cache *mem) 231static void exit_exception_table(struct exception_table *et, struct kmem_cache *mem)
232{ 232{
233 struct list_head *slot; 233 struct list_head *slot;
234 struct exception *ex, *next; 234 struct dm_snap_exception *ex, *next;
235 int i, size; 235 int i, size;
236 236
237 size = et->hash_mask + 1; 237 size = et->hash_mask + 1;
@@ -245,18 +245,19 @@ static void exit_exception_table(struct exception_table *et, struct kmem_cache *
245 vfree(et->table); 245 vfree(et->table);
246} 246}
247 247
248static inline uint32_t exception_hash(struct exception_table *et, chunk_t chunk) 248static uint32_t exception_hash(struct exception_table *et, chunk_t chunk)
249{ 249{
250 return chunk & et->hash_mask; 250 return chunk & et->hash_mask;
251} 251}
252 252
253static void insert_exception(struct exception_table *eh, struct exception *e) 253static void insert_exception(struct exception_table *eh,
254 struct dm_snap_exception *e)
254{ 255{
255 struct list_head *l = &eh->table[exception_hash(eh, e->old_chunk)]; 256 struct list_head *l = &eh->table[exception_hash(eh, e->old_chunk)];
256 list_add(&e->hash_list, l); 257 list_add(&e->hash_list, l);
257} 258}
258 259
259static inline void remove_exception(struct exception *e) 260static void remove_exception(struct dm_snap_exception *e)
260{ 261{
261 list_del(&e->hash_list); 262 list_del(&e->hash_list);
262} 263}
@@ -265,11 +266,11 @@ static inline void remove_exception(struct exception *e)
265 * Return the exception data for a sector, or NULL if not 266 * Return the exception data for a sector, or NULL if not
266 * remapped. 267 * remapped.
267 */ 268 */
268static struct exception *lookup_exception(struct exception_table *et, 269static struct dm_snap_exception *lookup_exception(struct exception_table *et,
269 chunk_t chunk) 270 chunk_t chunk)
270{ 271{
271 struct list_head *slot; 272 struct list_head *slot;
272 struct exception *e; 273 struct dm_snap_exception *e;
273 274
274 slot = &et->table[exception_hash(et, chunk)]; 275 slot = &et->table[exception_hash(et, chunk)];
275 list_for_each_entry (e, slot, hash_list) 276 list_for_each_entry (e, slot, hash_list)
@@ -279,9 +280,9 @@ static struct exception *lookup_exception(struct exception_table *et,
279 return NULL; 280 return NULL;
280} 281}
281 282
282static inline struct exception *alloc_exception(void) 283static struct dm_snap_exception *alloc_exception(void)
283{ 284{
284 struct exception *e; 285 struct dm_snap_exception *e;
285 286
286 e = kmem_cache_alloc(exception_cache, GFP_NOIO); 287 e = kmem_cache_alloc(exception_cache, GFP_NOIO);
287 if (!e) 288 if (!e)
@@ -290,24 +291,24 @@ static inline struct exception *alloc_exception(void)
290 return e; 291 return e;
291} 292}
292 293
293static inline void free_exception(struct exception *e) 294static void free_exception(struct dm_snap_exception *e)
294{ 295{
295 kmem_cache_free(exception_cache, e); 296 kmem_cache_free(exception_cache, e);
296} 297}
297 298
298static inline struct pending_exception *alloc_pending_exception(void) 299static struct dm_snap_pending_exception *alloc_pending_exception(void)
299{ 300{
300 return mempool_alloc(pending_pool, GFP_NOIO); 301 return mempool_alloc(pending_pool, GFP_NOIO);
301} 302}
302 303
303static inline void free_pending_exception(struct pending_exception *pe) 304static void free_pending_exception(struct dm_snap_pending_exception *pe)
304{ 305{
305 mempool_free(pe, pending_pool); 306 mempool_free(pe, pending_pool);
306} 307}
307 308
308int dm_add_exception(struct dm_snapshot *s, chunk_t old, chunk_t new) 309int dm_add_exception(struct dm_snapshot *s, chunk_t old, chunk_t new)
309{ 310{
310 struct exception *e; 311 struct dm_snap_exception *e;
311 312
312 e = alloc_exception(); 313 e = alloc_exception();
313 if (!e) 314 if (!e)
@@ -334,7 +335,7 @@ static int calc_max_buckets(void)
334/* 335/*
335 * Rounds a number down to a power of 2. 336 * Rounds a number down to a power of 2.
336 */ 337 */
337static inline uint32_t round_down(uint32_t n) 338static uint32_t round_down(uint32_t n)
338{ 339{
339 while (n & (n - 1)) 340 while (n & (n - 1))
340 n &= (n - 1); 341 n &= (n - 1);
@@ -384,7 +385,7 @@ static int init_hash_tables(struct dm_snapshot *s)
384 * Round a number up to the nearest 'size' boundary. size must 385 * Round a number up to the nearest 'size' boundary. size must
385 * be a power of 2. 386 * be a power of 2.
386 */ 387 */
387static inline ulong round_up(ulong n, ulong size) 388static ulong round_up(ulong n, ulong size)
388{ 389{
389 size--; 390 size--;
390 return (n + size) & ~size; 391 return (n + size) & ~size;
@@ -522,9 +523,12 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
522 523
523 /* Metadata must only be loaded into one table at once */ 524 /* Metadata must only be loaded into one table at once */
524 r = s->store.read_metadata(&s->store); 525 r = s->store.read_metadata(&s->store);
525 if (r) { 526 if (r < 0) {
526 ti->error = "Failed to read snapshot metadata"; 527 ti->error = "Failed to read snapshot metadata";
527 goto bad6; 528 goto bad6;
529 } else if (r > 0) {
530 s->valid = 0;
531 DMWARN("Snapshot is marked invalid.");
528 } 532 }
529 533
530 bio_list_init(&s->queued_bios); 534 bio_list_init(&s->queued_bios);
@@ -577,7 +581,7 @@ static void __free_exceptions(struct dm_snapshot *s)
577 581
578static void snapshot_dtr(struct dm_target *ti) 582static void snapshot_dtr(struct dm_target *ti)
579{ 583{
580 struct dm_snapshot *s = (struct dm_snapshot *) ti->private; 584 struct dm_snapshot *s = ti->private;
581 585
582 flush_workqueue(ksnapd); 586 flush_workqueue(ksnapd);
583 587
@@ -655,14 +659,14 @@ static void __invalidate_snapshot(struct dm_snapshot *s, int err)
655 dm_table_event(s->table); 659 dm_table_event(s->table);
656} 660}
657 661
658static void get_pending_exception(struct pending_exception *pe) 662static void get_pending_exception(struct dm_snap_pending_exception *pe)
659{ 663{
660 atomic_inc(&pe->ref_count); 664 atomic_inc(&pe->ref_count);
661} 665}
662 666
663static struct bio *put_pending_exception(struct pending_exception *pe) 667static struct bio *put_pending_exception(struct dm_snap_pending_exception *pe)
664{ 668{
665 struct pending_exception *primary_pe; 669 struct dm_snap_pending_exception *primary_pe;
666 struct bio *origin_bios = NULL; 670 struct bio *origin_bios = NULL;
667 671
668 primary_pe = pe->primary_pe; 672 primary_pe = pe->primary_pe;
@@ -692,9 +696,9 @@ static struct bio *put_pending_exception(struct pending_exception *pe)
692 return origin_bios; 696 return origin_bios;
693} 697}
694 698
695static void pending_complete(struct pending_exception *pe, int success) 699static void pending_complete(struct dm_snap_pending_exception *pe, int success)
696{ 700{
697 struct exception *e; 701 struct dm_snap_exception *e;
698 struct dm_snapshot *s = pe->snap; 702 struct dm_snapshot *s = pe->snap;
699 struct bio *origin_bios = NULL; 703 struct bio *origin_bios = NULL;
700 struct bio *snapshot_bios = NULL; 704 struct bio *snapshot_bios = NULL;
@@ -748,7 +752,8 @@ static void pending_complete(struct pending_exception *pe, int success)
748 752
749static void commit_callback(void *context, int success) 753static void commit_callback(void *context, int success)
750{ 754{
751 struct pending_exception *pe = (struct pending_exception *) context; 755 struct dm_snap_pending_exception *pe = context;
756
752 pending_complete(pe, success); 757 pending_complete(pe, success);
753} 758}
754 759
@@ -758,7 +763,7 @@ static void commit_callback(void *context, int success)
758 */ 763 */
759static void copy_callback(int read_err, unsigned int write_err, void *context) 764static void copy_callback(int read_err, unsigned int write_err, void *context)
760{ 765{
761 struct pending_exception *pe = (struct pending_exception *) context; 766 struct dm_snap_pending_exception *pe = context;
762 struct dm_snapshot *s = pe->snap; 767 struct dm_snapshot *s = pe->snap;
763 768
764 if (read_err || write_err) 769 if (read_err || write_err)
@@ -773,7 +778,7 @@ static void copy_callback(int read_err, unsigned int write_err, void *context)
773/* 778/*
774 * Dispatches the copy operation to kcopyd. 779 * Dispatches the copy operation to kcopyd.
775 */ 780 */
776static void start_copy(struct pending_exception *pe) 781static void start_copy(struct dm_snap_pending_exception *pe)
777{ 782{
778 struct dm_snapshot *s = pe->snap; 783 struct dm_snapshot *s = pe->snap;
779 struct io_region src, dest; 784 struct io_region src, dest;
@@ -803,11 +808,11 @@ static void start_copy(struct pending_exception *pe)
803 * NOTE: a write lock must be held on snap->lock before calling 808 * NOTE: a write lock must be held on snap->lock before calling
804 * this. 809 * this.
805 */ 810 */
806static struct pending_exception * 811static struct dm_snap_pending_exception *
807__find_pending_exception(struct dm_snapshot *s, struct bio *bio) 812__find_pending_exception(struct dm_snapshot *s, struct bio *bio)
808{ 813{
809 struct exception *e; 814 struct dm_snap_exception *e;
810 struct pending_exception *pe; 815 struct dm_snap_pending_exception *pe;
811 chunk_t chunk = sector_to_chunk(s, bio->bi_sector); 816 chunk_t chunk = sector_to_chunk(s, bio->bi_sector);
812 817
813 /* 818 /*
@@ -816,7 +821,7 @@ __find_pending_exception(struct dm_snapshot *s, struct bio *bio)
816 e = lookup_exception(&s->pending, chunk); 821 e = lookup_exception(&s->pending, chunk);
817 if (e) { 822 if (e) {
818 /* cast the exception to a pending exception */ 823 /* cast the exception to a pending exception */
819 pe = container_of(e, struct pending_exception, e); 824 pe = container_of(e, struct dm_snap_pending_exception, e);
820 goto out; 825 goto out;
821 } 826 }
822 827
@@ -836,7 +841,7 @@ __find_pending_exception(struct dm_snapshot *s, struct bio *bio)
836 e = lookup_exception(&s->pending, chunk); 841 e = lookup_exception(&s->pending, chunk);
837 if (e) { 842 if (e) {
838 free_pending_exception(pe); 843 free_pending_exception(pe);
839 pe = container_of(e, struct pending_exception, e); 844 pe = container_of(e, struct dm_snap_pending_exception, e);
840 goto out; 845 goto out;
841 } 846 }
842 847
@@ -860,8 +865,8 @@ __find_pending_exception(struct dm_snapshot *s, struct bio *bio)
860 return pe; 865 return pe;
861} 866}
862 867
863static inline void remap_exception(struct dm_snapshot *s, struct exception *e, 868static void remap_exception(struct dm_snapshot *s, struct dm_snap_exception *e,
864 struct bio *bio) 869 struct bio *bio)
865{ 870{
866 bio->bi_bdev = s->cow->bdev; 871 bio->bi_bdev = s->cow->bdev;
867 bio->bi_sector = chunk_to_sector(s, e->new_chunk) + 872 bio->bi_sector = chunk_to_sector(s, e->new_chunk) +
@@ -871,11 +876,11 @@ static inline void remap_exception(struct dm_snapshot *s, struct exception *e,
871static int snapshot_map(struct dm_target *ti, struct bio *bio, 876static int snapshot_map(struct dm_target *ti, struct bio *bio,
872 union map_info *map_context) 877 union map_info *map_context)
873{ 878{
874 struct exception *e; 879 struct dm_snap_exception *e;
875 struct dm_snapshot *s = (struct dm_snapshot *) ti->private; 880 struct dm_snapshot *s = ti->private;
876 int r = DM_MAPIO_REMAPPED; 881 int r = DM_MAPIO_REMAPPED;
877 chunk_t chunk; 882 chunk_t chunk;
878 struct pending_exception *pe = NULL; 883 struct dm_snap_pending_exception *pe = NULL;
879 884
880 chunk = sector_to_chunk(s, bio->bi_sector); 885 chunk = sector_to_chunk(s, bio->bi_sector);
881 886
@@ -884,9 +889,6 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
884 if (!s->valid) 889 if (!s->valid)
885 return -EIO; 890 return -EIO;
886 891
887 if (unlikely(bio_barrier(bio)))
888 return -EOPNOTSUPP;
889
890 /* FIXME: should only take write lock if we need 892 /* FIXME: should only take write lock if we need
891 * to copy an exception */ 893 * to copy an exception */
892 down_write(&s->lock); 894 down_write(&s->lock);
@@ -945,7 +947,7 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
945 947
946static void snapshot_resume(struct dm_target *ti) 948static void snapshot_resume(struct dm_target *ti)
947{ 949{
948 struct dm_snapshot *s = (struct dm_snapshot *) ti->private; 950 struct dm_snapshot *s = ti->private;
949 951
950 down_write(&s->lock); 952 down_write(&s->lock);
951 s->active = 1; 953 s->active = 1;
@@ -955,7 +957,7 @@ static void snapshot_resume(struct dm_target *ti)
955static int snapshot_status(struct dm_target *ti, status_type_t type, 957static int snapshot_status(struct dm_target *ti, status_type_t type,
956 char *result, unsigned int maxlen) 958 char *result, unsigned int maxlen)
957{ 959{
958 struct dm_snapshot *snap = (struct dm_snapshot *) ti->private; 960 struct dm_snapshot *snap = ti->private;
959 961
960 switch (type) { 962 switch (type) {
961 case STATUSTYPE_INFO: 963 case STATUSTYPE_INFO:
@@ -999,8 +1001,8 @@ static int __origin_write(struct list_head *snapshots, struct bio *bio)
999{ 1001{
1000 int r = DM_MAPIO_REMAPPED, first = 0; 1002 int r = DM_MAPIO_REMAPPED, first = 0;
1001 struct dm_snapshot *snap; 1003 struct dm_snapshot *snap;
1002 struct exception *e; 1004 struct dm_snap_exception *e;
1003 struct pending_exception *pe, *next_pe, *primary_pe = NULL; 1005 struct dm_snap_pending_exception *pe, *next_pe, *primary_pe = NULL;
1004 chunk_t chunk; 1006 chunk_t chunk;
1005 LIST_HEAD(pe_queue); 1007 LIST_HEAD(pe_queue);
1006 1008
@@ -1147,19 +1149,16 @@ static int origin_ctr(struct dm_target *ti, unsigned int argc, char **argv)
1147 1149
1148static void origin_dtr(struct dm_target *ti) 1150static void origin_dtr(struct dm_target *ti)
1149{ 1151{
1150 struct dm_dev *dev = (struct dm_dev *) ti->private; 1152 struct dm_dev *dev = ti->private;
1151 dm_put_device(ti, dev); 1153 dm_put_device(ti, dev);
1152} 1154}
1153 1155
1154static int origin_map(struct dm_target *ti, struct bio *bio, 1156static int origin_map(struct dm_target *ti, struct bio *bio,
1155 union map_info *map_context) 1157 union map_info *map_context)
1156{ 1158{
1157 struct dm_dev *dev = (struct dm_dev *) ti->private; 1159 struct dm_dev *dev = ti->private;
1158 bio->bi_bdev = dev->bdev; 1160 bio->bi_bdev = dev->bdev;
1159 1161
1160 if (unlikely(bio_barrier(bio)))
1161 return -EOPNOTSUPP;
1162
1163 /* Only tell snapshots if this is a write */ 1162 /* Only tell snapshots if this is a write */
1164 return (bio_rw(bio) == WRITE) ? do_origin(dev, bio) : DM_MAPIO_REMAPPED; 1163 return (bio_rw(bio) == WRITE) ? do_origin(dev, bio) : DM_MAPIO_REMAPPED;
1165} 1164}
@@ -1172,7 +1171,7 @@ static int origin_map(struct dm_target *ti, struct bio *bio,
1172 */ 1171 */
1173static void origin_resume(struct dm_target *ti) 1172static void origin_resume(struct dm_target *ti)
1174{ 1173{
1175 struct dm_dev *dev = (struct dm_dev *) ti->private; 1174 struct dm_dev *dev = ti->private;
1176 struct dm_snapshot *snap; 1175 struct dm_snapshot *snap;
1177 struct origin *o; 1176 struct origin *o;
1178 chunk_t chunk_size = 0; 1177 chunk_t chunk_size = 0;
@@ -1190,7 +1189,7 @@ static void origin_resume(struct dm_target *ti)
1190static int origin_status(struct dm_target *ti, status_type_t type, char *result, 1189static int origin_status(struct dm_target *ti, status_type_t type, char *result,
1191 unsigned int maxlen) 1190 unsigned int maxlen)
1192{ 1191{
1193 struct dm_dev *dev = (struct dm_dev *) ti->private; 1192 struct dm_dev *dev = ti->private;
1194 1193
1195 switch (type) { 1194 switch (type) {
1196 case STATUSTYPE_INFO: 1195 case STATUSTYPE_INFO:
@@ -1249,21 +1248,14 @@ static int __init dm_snapshot_init(void)
1249 goto bad2; 1248 goto bad2;
1250 } 1249 }
1251 1250
1252 exception_cache = kmem_cache_create("dm-snapshot-ex", 1251 exception_cache = KMEM_CACHE(dm_snap_exception, 0);
1253 sizeof(struct exception),
1254 __alignof__(struct exception),
1255 0, NULL, NULL);
1256 if (!exception_cache) { 1252 if (!exception_cache) {
1257 DMERR("Couldn't create exception cache."); 1253 DMERR("Couldn't create exception cache.");
1258 r = -ENOMEM; 1254 r = -ENOMEM;
1259 goto bad3; 1255 goto bad3;
1260 } 1256 }
1261 1257
1262 pending_cache = 1258 pending_cache = KMEM_CACHE(dm_snap_pending_exception, 0);
1263 kmem_cache_create("dm-snapshot-in",
1264 sizeof(struct pending_exception),
1265 __alignof__(struct pending_exception),
1266 0, NULL, NULL);
1267 if (!pending_cache) { 1259 if (!pending_cache) {
1268 DMERR("Couldn't create pending cache."); 1260 DMERR("Couldn't create pending cache.");
1269 r = -ENOMEM; 1261 r = -ENOMEM;
diff --git a/drivers/md/dm-snap.h b/drivers/md/dm-snap.h
index 15fa2ae6cdc2..650e0f1f51d8 100644
--- a/drivers/md/dm-snap.h
+++ b/drivers/md/dm-snap.h
@@ -30,7 +30,7 @@ typedef sector_t chunk_t;
30 * An exception is used where an old chunk of data has been 30 * An exception is used where an old chunk of data has been
31 * replaced by a new one. 31 * replaced by a new one.
32 */ 32 */
33struct exception { 33struct dm_snap_exception {
34 struct list_head hash_list; 34 struct list_head hash_list;
35 35
36 chunk_t old_chunk; 36 chunk_t old_chunk;
@@ -58,13 +58,13 @@ struct exception_store {
58 * Find somewhere to store the next exception. 58 * Find somewhere to store the next exception.
59 */ 59 */
60 int (*prepare_exception) (struct exception_store *store, 60 int (*prepare_exception) (struct exception_store *store,
61 struct exception *e); 61 struct dm_snap_exception *e);
62 62
63 /* 63 /*
64 * Update the metadata with this exception. 64 * Update the metadata with this exception.
65 */ 65 */
66 void (*commit_exception) (struct exception_store *store, 66 void (*commit_exception) (struct exception_store *store,
67 struct exception *e, 67 struct dm_snap_exception *e,
68 void (*callback) (void *, int success), 68 void (*callback) (void *, int success),
69 void *callback_context); 69 void *callback_context);
70 70
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 2717a355dc5b..f4f7d35561ab 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -45,7 +45,7 @@ struct dm_io {
45 * One of these is allocated per target within a bio. Hopefully 45 * One of these is allocated per target within a bio. Hopefully
46 * this will be simplified out one day. 46 * this will be simplified out one day.
47 */ 47 */
48struct target_io { 48struct dm_target_io {
49 struct dm_io *io; 49 struct dm_io *io;
50 struct dm_target *ti; 50 struct dm_target *ti;
51 union map_info info; 51 union map_info info;
@@ -54,7 +54,7 @@ struct target_io {
54union map_info *dm_get_mapinfo(struct bio *bio) 54union map_info *dm_get_mapinfo(struct bio *bio)
55{ 55{
56 if (bio && bio->bi_private) 56 if (bio && bio->bi_private)
57 return &((struct target_io *)bio->bi_private)->info; 57 return &((struct dm_target_io *)bio->bi_private)->info;
58 return NULL; 58 return NULL;
59} 59}
60 60
@@ -132,14 +132,12 @@ static int __init local_init(void)
132 int r; 132 int r;
133 133
134 /* allocate a slab for the dm_ios */ 134 /* allocate a slab for the dm_ios */
135 _io_cache = kmem_cache_create("dm_io", 135 _io_cache = KMEM_CACHE(dm_io, 0);
136 sizeof(struct dm_io), 0, 0, NULL, NULL);
137 if (!_io_cache) 136 if (!_io_cache)
138 return -ENOMEM; 137 return -ENOMEM;
139 138
140 /* allocate a slab for the target ios */ 139 /* allocate a slab for the target ios */
141 _tio_cache = kmem_cache_create("dm_tio", sizeof(struct target_io), 140 _tio_cache = KMEM_CACHE(dm_target_io, 0);
142 0, 0, NULL, NULL);
143 if (!_tio_cache) { 141 if (!_tio_cache) {
144 kmem_cache_destroy(_io_cache); 142 kmem_cache_destroy(_io_cache);
145 return -ENOMEM; 143 return -ENOMEM;
@@ -325,22 +323,22 @@ out:
325 return r; 323 return r;
326} 324}
327 325
328static inline struct dm_io *alloc_io(struct mapped_device *md) 326static struct dm_io *alloc_io(struct mapped_device *md)
329{ 327{
330 return mempool_alloc(md->io_pool, GFP_NOIO); 328 return mempool_alloc(md->io_pool, GFP_NOIO);
331} 329}
332 330
333static inline void free_io(struct mapped_device *md, struct dm_io *io) 331static void free_io(struct mapped_device *md, struct dm_io *io)
334{ 332{
335 mempool_free(io, md->io_pool); 333 mempool_free(io, md->io_pool);
336} 334}
337 335
338static inline struct target_io *alloc_tio(struct mapped_device *md) 336static struct dm_target_io *alloc_tio(struct mapped_device *md)
339{ 337{
340 return mempool_alloc(md->tio_pool, GFP_NOIO); 338 return mempool_alloc(md->tio_pool, GFP_NOIO);
341} 339}
342 340
343static inline void free_tio(struct mapped_device *md, struct target_io *tio) 341static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
344{ 342{
345 mempool_free(tio, md->tio_pool); 343 mempool_free(tio, md->tio_pool);
346} 344}
@@ -498,7 +496,7 @@ static void dec_pending(struct dm_io *io, int error)
498static int clone_endio(struct bio *bio, unsigned int done, int error) 496static int clone_endio(struct bio *bio, unsigned int done, int error)
499{ 497{
500 int r = 0; 498 int r = 0;
501 struct target_io *tio = bio->bi_private; 499 struct dm_target_io *tio = bio->bi_private;
502 struct mapped_device *md = tio->io->md; 500 struct mapped_device *md = tio->io->md;
503 dm_endio_fn endio = tio->ti->type->end_io; 501 dm_endio_fn endio = tio->ti->type->end_io;
504 502
@@ -558,7 +556,7 @@ static sector_t max_io_len(struct mapped_device *md,
558} 556}
559 557
560static void __map_bio(struct dm_target *ti, struct bio *clone, 558static void __map_bio(struct dm_target *ti, struct bio *clone,
561 struct target_io *tio) 559 struct dm_target_io *tio)
562{ 560{
563 int r; 561 int r;
564 sector_t sector; 562 sector_t sector;
@@ -672,7 +670,7 @@ static void __clone_and_map(struct clone_info *ci)
672 struct bio *clone, *bio = ci->bio; 670 struct bio *clone, *bio = ci->bio;
673 struct dm_target *ti = dm_table_find_target(ci->map, ci->sector); 671 struct dm_target *ti = dm_table_find_target(ci->map, ci->sector);
674 sector_t len = 0, max = max_io_len(ci->md, ci->sector, ti); 672 sector_t len = 0, max = max_io_len(ci->md, ci->sector, ti);
675 struct target_io *tio; 673 struct dm_target_io *tio;
676 674
677 /* 675 /*
678 * Allocate a target io object. 676 * Allocate a target io object.
@@ -802,6 +800,15 @@ static int dm_request(request_queue_t *q, struct bio *bio)
802 int rw = bio_data_dir(bio); 800 int rw = bio_data_dir(bio);
803 struct mapped_device *md = q->queuedata; 801 struct mapped_device *md = q->queuedata;
804 802
803 /*
804 * There is no use in forwarding any barrier request since we can't
805 * guarantee it is (or can be) handled by the targets correctly.
806 */
807 if (unlikely(bio_barrier(bio))) {
808 bio_endio(bio, bio->bi_size, -EOPNOTSUPP);
809 return 0;
810 }
811
805 down_read(&md->io_lock); 812 down_read(&md->io_lock);
806 813
807 disk_stat_inc(dm_disk(md), ios[rw]); 814 disk_stat_inc(dm_disk(md), ios[rw]);
diff --git a/drivers/md/dm.h b/drivers/md/dm.h
index 2f796b1436b2..462ee652a890 100644
--- a/drivers/md/dm.h
+++ b/drivers/md/dm.h
@@ -18,13 +18,45 @@
18 18
19#define DM_NAME "device-mapper" 19#define DM_NAME "device-mapper"
20 20
21#define DMERR(f, arg...) printk(KERN_ERR DM_NAME ": " DM_MSG_PREFIX ": " f "\n", ## arg) 21#define DMERR(f, arg...) \
22#define DMWARN(f, arg...) printk(KERN_WARNING DM_NAME ": " DM_MSG_PREFIX ": " f "\n", ## arg) 22 printk(KERN_ERR DM_NAME ": " DM_MSG_PREFIX ": " f "\n", ## arg)
23#define DMINFO(f, arg...) printk(KERN_INFO DM_NAME ": " DM_MSG_PREFIX ": " f "\n", ## arg) 23#define DMERR_LIMIT(f, arg...) \
24 do { \
25 if (printk_ratelimit()) \
26 printk(KERN_ERR DM_NAME ": " DM_MSG_PREFIX ": " \
27 f "\n", ## arg); \
28 } while (0)
29
30#define DMWARN(f, arg...) \
31 printk(KERN_WARNING DM_NAME ": " DM_MSG_PREFIX ": " f "\n", ## arg)
32#define DMWARN_LIMIT(f, arg...) \
33 do { \
34 if (printk_ratelimit()) \
35 printk(KERN_WARNING DM_NAME ": " DM_MSG_PREFIX ": " \
36 f "\n", ## arg); \
37 } while (0)
38
39#define DMINFO(f, arg...) \
40 printk(KERN_INFO DM_NAME ": " DM_MSG_PREFIX ": " f "\n", ## arg)
41#define DMINFO_LIMIT(f, arg...) \
42 do { \
43 if (printk_ratelimit()) \
44 printk(KERN_INFO DM_NAME ": " DM_MSG_PREFIX ": " f \
45 "\n", ## arg); \
46 } while (0)
47
24#ifdef CONFIG_DM_DEBUG 48#ifdef CONFIG_DM_DEBUG
25# define DMDEBUG(f, arg...) printk(KERN_DEBUG DM_NAME ": " DM_MSG_PREFIX " DEBUG: " f "\n", ## arg) 49# define DMDEBUG(f, arg...) \
50 printk(KERN_DEBUG DM_NAME ": " DM_MSG_PREFIX " DEBUG: " f "\n", ## arg)
51# define DMDEBUG_LIMIT(f, arg...) \
52 do { \
53 if (printk_ratelimit()) \
54 printk(KERN_DEBUG DM_NAME ": " DM_MSG_PREFIX ": " f \
55 "\n", ## arg); \
56 } while (0)
26#else 57#else
27# define DMDEBUG(f, arg...) do {} while (0) 58# define DMDEBUG(f, arg...) do {} while (0)
59# define DMDEBUG_LIMIT(f, arg...) do {} while (0)
28#endif 60#endif
29 61
30#define DMEMIT(x...) sz += ((sz >= maxlen) ? \ 62#define DMEMIT(x...) sz += ((sz >= maxlen) ? \
diff --git a/drivers/md/kcopyd.c b/drivers/md/kcopyd.c
index dbc234e3c69f..7e052378c47e 100644
--- a/drivers/md/kcopyd.c
+++ b/drivers/md/kcopyd.c
@@ -29,7 +29,7 @@
29static struct workqueue_struct *_kcopyd_wq; 29static struct workqueue_struct *_kcopyd_wq;
30static struct work_struct _kcopyd_work; 30static struct work_struct _kcopyd_work;
31 31
32static inline void wake(void) 32static void wake(void)
33{ 33{
34 queue_work(_kcopyd_wq, &_kcopyd_work); 34 queue_work(_kcopyd_wq, &_kcopyd_work);
35} 35}
@@ -226,10 +226,7 @@ static LIST_HEAD(_pages_jobs);
226 226
227static int jobs_init(void) 227static int jobs_init(void)
228{ 228{
229 _job_cache = kmem_cache_create("kcopyd-jobs", 229 _job_cache = KMEM_CACHE(kcopyd_job, 0);
230 sizeof(struct kcopyd_job),
231 __alignof__(struct kcopyd_job),
232 0, NULL, NULL);
233 if (!_job_cache) 230 if (!_job_cache)
234 return -ENOMEM; 231 return -ENOMEM;
235 232
@@ -258,7 +255,7 @@ static void jobs_exit(void)
258 * Functions to push and pop a job onto the head of a given job 255 * Functions to push and pop a job onto the head of a given job
259 * list. 256 * list.
260 */ 257 */
261static inline struct kcopyd_job *pop(struct list_head *jobs) 258static struct kcopyd_job *pop(struct list_head *jobs)
262{ 259{
263 struct kcopyd_job *job = NULL; 260 struct kcopyd_job *job = NULL;
264 unsigned long flags; 261 unsigned long flags;
@@ -274,7 +271,7 @@ static inline struct kcopyd_job *pop(struct list_head *jobs)
274 return job; 271 return job;
275} 272}
276 273
277static inline void push(struct list_head *jobs, struct kcopyd_job *job) 274static void push(struct list_head *jobs, struct kcopyd_job *job)
278{ 275{
279 unsigned long flags; 276 unsigned long flags;
280 277
generated by cgit v1.2.2 (git 2.25.0) at 2025-01-18 10:39:46 -0500