Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (C) 2001, 2002 Sistina Software (UK) Limited. |
Milan Broz | 784aae7 | 2009-01-06 03:05:12 +0000 | [diff] [blame] | 3 | * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4 | * |
| 5 | * This file is released under the GPL. |
| 6 | */ |
| 7 | |
| 8 | #include "dm.h" |
Mike Anderson | 51e5b2b | 2007-10-19 22:48:00 +0100 | [diff] [blame] | 9 | #include "dm-uevent.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 10 | |
| 11 | #include <linux/init.h> |
| 12 | #include <linux/module.h> |
Arjan van de Ven | 48c9c27 | 2006-03-27 01:18:20 -0800 | [diff] [blame] | 13 | #include <linux/mutex.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 14 | #include <linux/moduleparam.h> |
| 15 | #include <linux/blkpg.h> |
| 16 | #include <linux/bio.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 17 | #include <linux/mempool.h> |
| 18 | #include <linux/slab.h> |
| 19 | #include <linux/idr.h> |
Darrick J. Wong | 3ac51e7 | 2006-03-27 01:17:54 -0800 | [diff] [blame] | 20 | #include <linux/hdreg.h> |
Kiyoshi Ueda | 3f77316d | 2010-08-12 04:13:56 +0100 | [diff] [blame] | 21 | #include <linux/delay.h> |
Li Zefan | 5578213 | 2009-06-09 13:43:05 +0800 | [diff] [blame] | 22 | |
| 23 | #include <trace/events/block.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | |
Alasdair G Kergon | 72d9486 | 2006-06-26 00:27:35 -0700 | [diff] [blame] | 25 | #define DM_MSG_PREFIX "core" |
| 26 | |
Namhyung Kim | 71a1673 | 2011-10-31 20:18:54 +0000 | [diff] [blame] | 27 | #ifdef CONFIG_PRINTK |
| 28 | /* |
| 29 | * ratelimit state to be used in DMXXX_LIMIT(). |
| 30 | */ |
| 31 | DEFINE_RATELIMIT_STATE(dm_ratelimit_state, |
| 32 | DEFAULT_RATELIMIT_INTERVAL, |
| 33 | DEFAULT_RATELIMIT_BURST); |
| 34 | EXPORT_SYMBOL(dm_ratelimit_state); |
| 35 | #endif |
| 36 | |
Milan Broz | 60935eb | 2009-06-22 10:12:30 +0100 | [diff] [blame] | 37 | /* |
| 38 | * Cookies are numeric values sent with CHANGE and REMOVE |
| 39 | * uevents while resuming, removing or renaming the device. |
| 40 | */ |
| 41 | #define DM_COOKIE_ENV_VAR_NAME "DM_COOKIE" |
| 42 | #define DM_COOKIE_LENGTH 24 |
| 43 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 44 | static const char *_name = DM_NAME; |
| 45 | |
| 46 | static unsigned int major = 0; |
| 47 | static unsigned int _major = 0; |
| 48 | |
Alasdair G Kergon | d15b774 | 2011-08-02 12:32:01 +0100 | [diff] [blame] | 49 | static DEFINE_IDR(_minor_idr); |
| 50 | |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 51 | static DEFINE_SPINLOCK(_minor_lock); |
Mikulas Patocka | 2c140a2 | 2013-11-01 18:27:41 -0400 | [diff] [blame] | 52 | |
| 53 | static void do_deferred_remove(struct work_struct *w); |
| 54 | |
| 55 | static DECLARE_WORK(deferred_remove_work, do_deferred_remove); |
| 56 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 57 | /* |
Kiyoshi Ueda | 8fbf26a | 2009-01-06 03:05:06 +0000 | [diff] [blame] | 58 | * For bio-based dm. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 59 | * One of these is allocated per bio. |
| 60 | */ |
| 61 | struct dm_io { |
| 62 | struct mapped_device *md; |
| 63 | int error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 64 | atomic_t io_count; |
Richard Kennedy | 6ae2fa6 | 2008-07-21 12:00:28 +0100 | [diff] [blame] | 65 | struct bio *bio; |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 66 | unsigned long start_time; |
Kiyoshi Ueda | f88fb98 | 2009-10-16 23:18:15 +0100 | [diff] [blame] | 67 | spinlock_t endio_lock; |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 68 | struct dm_stats_aux stats_aux; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 69 | }; |
| 70 | |
| 71 | /* |
Kiyoshi Ueda | 8fbf26a | 2009-01-06 03:05:06 +0000 | [diff] [blame] | 72 | * For request-based dm. |
| 73 | * One of these is allocated per request. |
| 74 | */ |
| 75 | struct dm_rq_target_io { |
| 76 | struct mapped_device *md; |
| 77 | struct dm_target *ti; |
| 78 | struct request *orig, clone; |
| 79 | int error; |
| 80 | union map_info info; |
| 81 | }; |
| 82 | |
| 83 | /* |
Kent Overstreet | 9481874 | 2012-09-07 13:44:01 -0700 | [diff] [blame] | 84 | * For request-based dm - the bio clones we allocate are embedded in these |
| 85 | * structs. |
| 86 | * |
| 87 | * We allocate these with bio_alloc_bioset, using the front_pad parameter when |
| 88 | * the bioset is created - this means the bio has to come at the end of the |
| 89 | * struct. |
Kiyoshi Ueda | 8fbf26a | 2009-01-06 03:05:06 +0000 | [diff] [blame] | 90 | */ |
| 91 | struct dm_rq_clone_bio_info { |
| 92 | struct bio *orig; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 93 | struct dm_rq_target_io *tio; |
Kent Overstreet | 9481874 | 2012-09-07 13:44:01 -0700 | [diff] [blame] | 94 | struct bio clone; |
Kiyoshi Ueda | 8fbf26a | 2009-01-06 03:05:06 +0000 | [diff] [blame] | 95 | }; |
| 96 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 97 | union map_info *dm_get_rq_mapinfo(struct request *rq) |
| 98 | { |
| 99 | if (rq && rq->end_io_data) |
| 100 | return &((struct dm_rq_target_io *)rq->end_io_data)->info; |
| 101 | return NULL; |
| 102 | } |
| 103 | EXPORT_SYMBOL_GPL(dm_get_rq_mapinfo); |
| 104 | |
Jeff Mahoney | ba61fdd | 2006-06-26 00:27:21 -0700 | [diff] [blame] | 105 | #define MINOR_ALLOCED ((void *)-1) |
| 106 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 107 | /* |
| 108 | * Bits for the md->flags field. |
| 109 | */ |
Alasdair G Kergon | 1eb787e | 2009-04-09 00:27:14 +0100 | [diff] [blame] | 110 | #define DMF_BLOCK_IO_FOR_SUSPEND 0 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 111 | #define DMF_SUSPENDED 1 |
Alasdair G Kergon | aa8d7c2 | 2006-01-06 00:20:06 -0800 | [diff] [blame] | 112 | #define DMF_FROZEN 2 |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 113 | #define DMF_FREEING 3 |
Alasdair G Kergon | 5c6bd75 | 2006-06-26 00:27:34 -0700 | [diff] [blame] | 114 | #define DMF_DELETING 4 |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 115 | #define DMF_NOFLUSH_SUSPENDING 5 |
Mikulas Patocka | d5b9dd0 | 2011-08-02 12:32:04 +0100 | [diff] [blame] | 116 | #define DMF_MERGE_IS_OPTIONAL 6 |
Mikulas Patocka | 2c140a2 | 2013-11-01 18:27:41 -0400 | [diff] [blame] | 117 | #define DMF_DEFERRED_REMOVE 7 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 118 | |
Milan Broz | 304f3f6 | 2008-02-08 02:11:17 +0000 | [diff] [blame] | 119 | /* |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 120 | * A dummy definition to make RCU happy. |
| 121 | * struct dm_table should never be dereferenced in this file. |
| 122 | */ |
| 123 | struct dm_table { |
| 124 | int undefined__; |
| 125 | }; |
| 126 | |
| 127 | /* |
Milan Broz | 304f3f6 | 2008-02-08 02:11:17 +0000 | [diff] [blame] | 128 | * Work processed by per-device workqueue. |
| 129 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 130 | struct mapped_device { |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 131 | struct srcu_struct io_barrier; |
Daniel Walker | e61290a | 2008-02-08 02:10:08 +0000 | [diff] [blame] | 132 | struct mutex suspend_lock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 133 | atomic_t holders; |
Alasdair G Kergon | 5c6bd75 | 2006-06-26 00:27:34 -0700 | [diff] [blame] | 134 | atomic_t open_count; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 135 | |
Mikulas Patocka | 2a7faeb | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 136 | /* |
| 137 | * The current mapping. |
| 138 | * Use dm_get_live_table{_fast} or take suspend_lock for |
| 139 | * dereference. |
| 140 | */ |
| 141 | struct dm_table *map; |
| 142 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 143 | unsigned long flags; |
| 144 | |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 145 | struct request_queue *queue; |
Mike Snitzer | a5664da | 2010-08-12 04:14:01 +0100 | [diff] [blame] | 146 | unsigned type; |
Mike Snitzer | 4a0b4dd | 2010-08-12 04:14:02 +0100 | [diff] [blame] | 147 | /* Protect queue and type against concurrent access. */ |
Mike Snitzer | a5664da | 2010-08-12 04:14:01 +0100 | [diff] [blame] | 148 | struct mutex type_lock; |
| 149 | |
Alasdair G Kergon | 36a0456 | 2011-10-31 20:19:04 +0000 | [diff] [blame] | 150 | struct target_type *immutable_target_type; |
| 151 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 152 | struct gendisk *disk; |
Mike Anderson | 7e51f25 | 2006-03-27 01:17:52 -0800 | [diff] [blame] | 153 | char name[16]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 154 | |
| 155 | void *interface_ptr; |
| 156 | |
| 157 | /* |
| 158 | * A list of ios that arrived while we were suspended. |
| 159 | */ |
Nikanth Karthikesan | 316d315 | 2009-10-06 20:16:55 +0200 | [diff] [blame] | 160 | atomic_t pending[2]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 161 | wait_queue_head_t wait; |
Mikulas Patocka | 53d5914 | 2009-04-02 19:55:37 +0100 | [diff] [blame] | 162 | struct work_struct work; |
Kiyoshi Ueda | 7485936 | 2006-12-08 02:41:02 -0800 | [diff] [blame] | 163 | struct bio_list deferred; |
Mikulas Patocka | 022c261 | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 164 | spinlock_t deferred_lock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 165 | |
| 166 | /* |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 167 | * Processing queue (flush) |
Milan Broz | 304f3f6 | 2008-02-08 02:11:17 +0000 | [diff] [blame] | 168 | */ |
| 169 | struct workqueue_struct *wq; |
| 170 | |
| 171 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 172 | * io objects are allocated from here. |
| 173 | */ |
| 174 | mempool_t *io_pool; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 175 | |
Stefan Bader | 9faf400 | 2006-10-03 01:15:41 -0700 | [diff] [blame] | 176 | struct bio_set *bs; |
| 177 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 178 | /* |
| 179 | * Event handling. |
| 180 | */ |
| 181 | atomic_t event_nr; |
| 182 | wait_queue_head_t eventq; |
Mike Anderson | 7a8c3d3 | 2007-10-19 22:48:01 +0100 | [diff] [blame] | 183 | atomic_t uevent_seq; |
| 184 | struct list_head uevent_list; |
| 185 | spinlock_t uevent_lock; /* Protect access to uevent_list */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 186 | |
| 187 | /* |
| 188 | * freeze/thaw support require holding onto a super block |
| 189 | */ |
| 190 | struct super_block *frozen_sb; |
Mikulas Patocka | db8fef4 | 2009-06-22 10:12:15 +0100 | [diff] [blame] | 191 | struct block_device *bdev; |
Darrick J. Wong | 3ac51e7 | 2006-03-27 01:17:54 -0800 | [diff] [blame] | 192 | |
| 193 | /* forced geometry settings */ |
| 194 | struct hd_geometry geometry; |
Milan Broz | 784aae7 | 2009-01-06 03:05:12 +0000 | [diff] [blame] | 195 | |
Mikulas Patocka | 2995fa7 | 2014-01-13 19:37:54 -0500 | [diff] [blame] | 196 | /* kobject and completion */ |
| 197 | struct dm_kobject_holder kobj_holder; |
Mikulas Patocka | be35f48 | 2014-01-06 23:01:22 -0500 | [diff] [blame] | 198 | |
Tejun Heo | d87f4c1 | 2010-09-03 11:56:19 +0200 | [diff] [blame] | 199 | /* zero-length flush that will be cloned and submitted to targets */ |
| 200 | struct bio flush_bio; |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 201 | |
| 202 | struct dm_stats stats; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 203 | }; |
| 204 | |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 205 | /* |
| 206 | * For mempools pre-allocation at the table loading time. |
| 207 | */ |
| 208 | struct dm_md_mempools { |
| 209 | mempool_t *io_pool; |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 210 | struct bio_set *bs; |
| 211 | }; |
| 212 | |
Mike Snitzer | 6cfa585 | 2013-09-12 18:06:11 -0400 | [diff] [blame] | 213 | #define RESERVED_BIO_BASED_IOS 16 |
| 214 | #define RESERVED_REQUEST_BASED_IOS 256 |
Mike Snitzer | f479082 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 215 | #define RESERVED_MAX_IOS 1024 |
Christoph Lameter | e18b890 | 2006-12-06 20:33:20 -0800 | [diff] [blame] | 216 | static struct kmem_cache *_io_cache; |
Kiyoshi Ueda | 8fbf26a | 2009-01-06 03:05:06 +0000 | [diff] [blame] | 217 | static struct kmem_cache *_rq_tio_cache; |
Kent Overstreet | 9481874 | 2012-09-07 13:44:01 -0700 | [diff] [blame] | 218 | |
Mike Snitzer | f479082 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 219 | /* |
Mike Snitzer | e860313 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 220 | * Bio-based DM's mempools' reserved IOs set by the user. |
| 221 | */ |
| 222 | static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS; |
| 223 | |
| 224 | /* |
Mike Snitzer | f479082 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 225 | * Request-based DM's mempools' reserved IOs set by the user. |
| 226 | */ |
| 227 | static unsigned reserved_rq_based_ios = RESERVED_REQUEST_BASED_IOS; |
| 228 | |
| 229 | static unsigned __dm_get_reserved_ios(unsigned *reserved_ios, |
| 230 | unsigned def, unsigned max) |
| 231 | { |
| 232 | unsigned ios = ACCESS_ONCE(*reserved_ios); |
| 233 | unsigned modified_ios = 0; |
| 234 | |
| 235 | if (!ios) |
| 236 | modified_ios = def; |
| 237 | else if (ios > max) |
| 238 | modified_ios = max; |
| 239 | |
| 240 | if (modified_ios) { |
| 241 | (void)cmpxchg(reserved_ios, ios, modified_ios); |
| 242 | ios = modified_ios; |
| 243 | } |
| 244 | |
| 245 | return ios; |
| 246 | } |
| 247 | |
Mike Snitzer | e860313 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 248 | unsigned dm_get_reserved_bio_based_ios(void) |
| 249 | { |
| 250 | return __dm_get_reserved_ios(&reserved_bio_based_ios, |
| 251 | RESERVED_BIO_BASED_IOS, RESERVED_MAX_IOS); |
| 252 | } |
| 253 | EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios); |
| 254 | |
Mike Snitzer | f479082 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 255 | unsigned dm_get_reserved_rq_based_ios(void) |
| 256 | { |
| 257 | return __dm_get_reserved_ios(&reserved_rq_based_ios, |
| 258 | RESERVED_REQUEST_BASED_IOS, RESERVED_MAX_IOS); |
| 259 | } |
| 260 | EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios); |
| 261 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 262 | static int __init local_init(void) |
| 263 | { |
Kiyoshi Ueda | 51157b4 | 2008-10-21 17:45:08 +0100 | [diff] [blame] | 264 | int r = -ENOMEM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 265 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 266 | /* allocate a slab for the dm_ios */ |
Alasdair G Kergon | 028867a | 2007-07-12 17:26:32 +0100 | [diff] [blame] | 267 | _io_cache = KMEM_CACHE(dm_io, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 268 | if (!_io_cache) |
Kiyoshi Ueda | 51157b4 | 2008-10-21 17:45:08 +0100 | [diff] [blame] | 269 | return r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 270 | |
Kiyoshi Ueda | 8fbf26a | 2009-01-06 03:05:06 +0000 | [diff] [blame] | 271 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
| 272 | if (!_rq_tio_cache) |
Mikulas Patocka | dba1416 | 2012-10-12 21:02:15 +0100 | [diff] [blame] | 273 | goto out_free_io_cache; |
Kiyoshi Ueda | 8fbf26a | 2009-01-06 03:05:06 +0000 | [diff] [blame] | 274 | |
Mike Anderson | 51e5b2b | 2007-10-19 22:48:00 +0100 | [diff] [blame] | 275 | r = dm_uevent_init(); |
Kiyoshi Ueda | 51157b4 | 2008-10-21 17:45:08 +0100 | [diff] [blame] | 276 | if (r) |
Jun'ichi Nomura | 23e5083 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 277 | goto out_free_rq_tio_cache; |
Mike Anderson | 51e5b2b | 2007-10-19 22:48:00 +0100 | [diff] [blame] | 278 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 279 | _major = major; |
| 280 | r = register_blkdev(_major, _name); |
Kiyoshi Ueda | 51157b4 | 2008-10-21 17:45:08 +0100 | [diff] [blame] | 281 | if (r < 0) |
| 282 | goto out_uevent_exit; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 283 | |
| 284 | if (!_major) |
| 285 | _major = r; |
| 286 | |
| 287 | return 0; |
Kiyoshi Ueda | 51157b4 | 2008-10-21 17:45:08 +0100 | [diff] [blame] | 288 | |
| 289 | out_uevent_exit: |
| 290 | dm_uevent_exit(); |
Kiyoshi Ueda | 8fbf26a | 2009-01-06 03:05:06 +0000 | [diff] [blame] | 291 | out_free_rq_tio_cache: |
| 292 | kmem_cache_destroy(_rq_tio_cache); |
Kiyoshi Ueda | 51157b4 | 2008-10-21 17:45:08 +0100 | [diff] [blame] | 293 | out_free_io_cache: |
| 294 | kmem_cache_destroy(_io_cache); |
| 295 | |
| 296 | return r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 297 | } |
| 298 | |
| 299 | static void local_exit(void) |
| 300 | { |
Mikulas Patocka | 2c140a2 | 2013-11-01 18:27:41 -0400 | [diff] [blame] | 301 | flush_scheduled_work(); |
| 302 | |
Kiyoshi Ueda | 8fbf26a | 2009-01-06 03:05:06 +0000 | [diff] [blame] | 303 | kmem_cache_destroy(_rq_tio_cache); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 304 | kmem_cache_destroy(_io_cache); |
Akinobu Mita | 00d5940 | 2007-07-17 04:03:46 -0700 | [diff] [blame] | 305 | unregister_blkdev(_major, _name); |
Mike Anderson | 51e5b2b | 2007-10-19 22:48:00 +0100 | [diff] [blame] | 306 | dm_uevent_exit(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 307 | |
| 308 | _major = 0; |
| 309 | |
| 310 | DMINFO("cleaned up"); |
| 311 | } |
| 312 | |
Alasdair G Kergon | b9249e5 | 2008-02-08 02:09:51 +0000 | [diff] [blame] | 313 | static int (*_inits[])(void) __initdata = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 314 | local_init, |
| 315 | dm_target_init, |
| 316 | dm_linear_init, |
| 317 | dm_stripe_init, |
Mikulas Patocka | 952b355 | 2009-12-10 23:51:57 +0000 | [diff] [blame] | 318 | dm_io_init, |
Mikulas Patocka | 945fa4d | 2008-04-24 21:43:49 +0100 | [diff] [blame] | 319 | dm_kcopyd_init, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 320 | dm_interface_init, |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 321 | dm_statistics_init, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 322 | }; |
| 323 | |
Alasdair G Kergon | b9249e5 | 2008-02-08 02:09:51 +0000 | [diff] [blame] | 324 | static void (*_exits[])(void) = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 325 | local_exit, |
| 326 | dm_target_exit, |
| 327 | dm_linear_exit, |
| 328 | dm_stripe_exit, |
Mikulas Patocka | 952b355 | 2009-12-10 23:51:57 +0000 | [diff] [blame] | 329 | dm_io_exit, |
Mikulas Patocka | 945fa4d | 2008-04-24 21:43:49 +0100 | [diff] [blame] | 330 | dm_kcopyd_exit, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 331 | dm_interface_exit, |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 332 | dm_statistics_exit, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 333 | }; |
| 334 | |
| 335 | static int __init dm_init(void) |
| 336 | { |
| 337 | const int count = ARRAY_SIZE(_inits); |
| 338 | |
| 339 | int r, i; |
| 340 | |
| 341 | for (i = 0; i < count; i++) { |
| 342 | r = _inits[i](); |
| 343 | if (r) |
| 344 | goto bad; |
| 345 | } |
| 346 | |
| 347 | return 0; |
| 348 | |
| 349 | bad: |
| 350 | while (i--) |
| 351 | _exits[i](); |
| 352 | |
| 353 | return r; |
| 354 | } |
| 355 | |
| 356 | static void __exit dm_exit(void) |
| 357 | { |
| 358 | int i = ARRAY_SIZE(_exits); |
| 359 | |
| 360 | while (i--) |
| 361 | _exits[i](); |
Alasdair G Kergon | d15b774 | 2011-08-02 12:32:01 +0100 | [diff] [blame] | 362 | |
| 363 | /* |
| 364 | * Should be empty by this point. |
| 365 | */ |
Alasdair G Kergon | d15b774 | 2011-08-02 12:32:01 +0100 | [diff] [blame] | 366 | idr_destroy(&_minor_idr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 367 | } |
| 368 | |
| 369 | /* |
| 370 | * Block device functions |
| 371 | */ |
Mike Anderson | 432a212 | 2009-12-10 23:52:20 +0000 | [diff] [blame] | 372 | int dm_deleting_md(struct mapped_device *md) |
| 373 | { |
| 374 | return test_bit(DMF_DELETING, &md->flags); |
| 375 | } |
| 376 | |
Al Viro | fe5f9f2 | 2008-03-02 10:29:31 -0500 | [diff] [blame] | 377 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 378 | { |
| 379 | struct mapped_device *md; |
| 380 | |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 381 | spin_lock(&_minor_lock); |
| 382 | |
Al Viro | fe5f9f2 | 2008-03-02 10:29:31 -0500 | [diff] [blame] | 383 | md = bdev->bd_disk->private_data; |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 384 | if (!md) |
| 385 | goto out; |
| 386 | |
Alasdair G Kergon | 5c6bd75 | 2006-06-26 00:27:34 -0700 | [diff] [blame] | 387 | if (test_bit(DMF_FREEING, &md->flags) || |
Mike Anderson | 432a212 | 2009-12-10 23:52:20 +0000 | [diff] [blame] | 388 | dm_deleting_md(md)) { |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 389 | md = NULL; |
| 390 | goto out; |
| 391 | } |
| 392 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 393 | dm_get(md); |
Alasdair G Kergon | 5c6bd75 | 2006-06-26 00:27:34 -0700 | [diff] [blame] | 394 | atomic_inc(&md->open_count); |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 395 | |
| 396 | out: |
| 397 | spin_unlock(&_minor_lock); |
| 398 | |
| 399 | return md ? 0 : -ENXIO; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 400 | } |
| 401 | |
Al Viro | db2a144 | 2013-05-05 21:52:57 -0400 | [diff] [blame] | 402 | static void dm_blk_close(struct gendisk *disk, fmode_t mode) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 403 | { |
Al Viro | fe5f9f2 | 2008-03-02 10:29:31 -0500 | [diff] [blame] | 404 | struct mapped_device *md = disk->private_data; |
Arnd Bergmann | 6e9624b | 2010-08-07 18:25:34 +0200 | [diff] [blame] | 405 | |
Milan Broz | 4a1aeb9 | 2011-01-13 19:59:48 +0000 | [diff] [blame] | 406 | spin_lock(&_minor_lock); |
| 407 | |
Mikulas Patocka | 2c140a2 | 2013-11-01 18:27:41 -0400 | [diff] [blame] | 408 | if (atomic_dec_and_test(&md->open_count) && |
| 409 | (test_bit(DMF_DEFERRED_REMOVE, &md->flags))) |
| 410 | schedule_work(&deferred_remove_work); |
| 411 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 412 | dm_put(md); |
Milan Broz | 4a1aeb9 | 2011-01-13 19:59:48 +0000 | [diff] [blame] | 413 | |
| 414 | spin_unlock(&_minor_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 415 | } |
| 416 | |
Alasdair G Kergon | 5c6bd75 | 2006-06-26 00:27:34 -0700 | [diff] [blame] | 417 | int dm_open_count(struct mapped_device *md) |
| 418 | { |
| 419 | return atomic_read(&md->open_count); |
| 420 | } |
| 421 | |
| 422 | /* |
| 423 | * Guarantees nothing is using the device before it's deleted. |
| 424 | */ |
Mikulas Patocka | 2c140a2 | 2013-11-01 18:27:41 -0400 | [diff] [blame] | 425 | int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred) |
Alasdair G Kergon | 5c6bd75 | 2006-06-26 00:27:34 -0700 | [diff] [blame] | 426 | { |
| 427 | int r = 0; |
| 428 | |
| 429 | spin_lock(&_minor_lock); |
| 430 | |
Mikulas Patocka | 2c140a2 | 2013-11-01 18:27:41 -0400 | [diff] [blame] | 431 | if (dm_open_count(md)) { |
Alasdair G Kergon | 5c6bd75 | 2006-06-26 00:27:34 -0700 | [diff] [blame] | 432 | r = -EBUSY; |
Mikulas Patocka | 2c140a2 | 2013-11-01 18:27:41 -0400 | [diff] [blame] | 433 | if (mark_deferred) |
| 434 | set_bit(DMF_DEFERRED_REMOVE, &md->flags); |
| 435 | } else if (only_deferred && !test_bit(DMF_DEFERRED_REMOVE, &md->flags)) |
| 436 | r = -EEXIST; |
Alasdair G Kergon | 5c6bd75 | 2006-06-26 00:27:34 -0700 | [diff] [blame] | 437 | else |
| 438 | set_bit(DMF_DELETING, &md->flags); |
| 439 | |
| 440 | spin_unlock(&_minor_lock); |
| 441 | |
| 442 | return r; |
| 443 | } |
| 444 | |
Mikulas Patocka | 2c140a2 | 2013-11-01 18:27:41 -0400 | [diff] [blame] | 445 | int dm_cancel_deferred_remove(struct mapped_device *md) |
| 446 | { |
| 447 | int r = 0; |
| 448 | |
| 449 | spin_lock(&_minor_lock); |
| 450 | |
| 451 | if (test_bit(DMF_DELETING, &md->flags)) |
| 452 | r = -EBUSY; |
| 453 | else |
| 454 | clear_bit(DMF_DEFERRED_REMOVE, &md->flags); |
| 455 | |
| 456 | spin_unlock(&_minor_lock); |
| 457 | |
| 458 | return r; |
| 459 | } |
| 460 | |
| 461 | static void do_deferred_remove(struct work_struct *w) |
| 462 | { |
| 463 | dm_deferred_remove(); |
| 464 | } |
| 465 | |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 466 | sector_t dm_get_size(struct mapped_device *md) |
| 467 | { |
| 468 | return get_capacity(md->disk); |
| 469 | } |
| 470 | |
Mike Snitzer | 9974fa2 | 2014-02-28 15:33:43 +0100 | [diff] [blame] | 471 | struct request_queue *dm_get_md_queue(struct mapped_device *md) |
| 472 | { |
| 473 | return md->queue; |
| 474 | } |
| 475 | |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 476 | struct dm_stats *dm_get_stats(struct mapped_device *md) |
| 477 | { |
| 478 | return &md->stats; |
| 479 | } |
| 480 | |
Darrick J. Wong | 3ac51e7 | 2006-03-27 01:17:54 -0800 | [diff] [blame] | 481 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
| 482 | { |
| 483 | struct mapped_device *md = bdev->bd_disk->private_data; |
| 484 | |
| 485 | return dm_get_geometry(md, geo); |
| 486 | } |
| 487 | |
Al Viro | fe5f9f2 | 2008-03-02 10:29:31 -0500 | [diff] [blame] | 488 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, |
Milan Broz | aa129a2 | 2006-10-03 01:15:15 -0700 | [diff] [blame] | 489 | unsigned int cmd, unsigned long arg) |
| 490 | { |
Al Viro | fe5f9f2 | 2008-03-02 10:29:31 -0500 | [diff] [blame] | 491 | struct mapped_device *md = bdev->bd_disk->private_data; |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 492 | int srcu_idx; |
Hannes Reinecke | 6c182cd | 2013-07-10 23:41:15 +0100 | [diff] [blame] | 493 | struct dm_table *map; |
Milan Broz | aa129a2 | 2006-10-03 01:15:15 -0700 | [diff] [blame] | 494 | struct dm_target *tgt; |
| 495 | int r = -ENOTTY; |
| 496 | |
Hannes Reinecke | 6c182cd | 2013-07-10 23:41:15 +0100 | [diff] [blame] | 497 | retry: |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 498 | map = dm_get_live_table(md, &srcu_idx); |
| 499 | |
Milan Broz | aa129a2 | 2006-10-03 01:15:15 -0700 | [diff] [blame] | 500 | if (!map || !dm_table_get_size(map)) |
| 501 | goto out; |
| 502 | |
| 503 | /* We only support devices that have a single target */ |
| 504 | if (dm_table_get_num_targets(map) != 1) |
| 505 | goto out; |
| 506 | |
| 507 | tgt = dm_table_get_target(map, 0); |
| 508 | |
Kiyoshi Ueda | 4f186f8 | 2009-12-10 23:52:26 +0000 | [diff] [blame] | 509 | if (dm_suspended_md(md)) { |
Milan Broz | aa129a2 | 2006-10-03 01:15:15 -0700 | [diff] [blame] | 510 | r = -EAGAIN; |
| 511 | goto out; |
| 512 | } |
| 513 | |
| 514 | if (tgt->type->ioctl) |
Al Viro | 647b3d0 | 2007-08-28 22:15:59 -0400 | [diff] [blame] | 515 | r = tgt->type->ioctl(tgt, cmd, arg); |
Milan Broz | aa129a2 | 2006-10-03 01:15:15 -0700 | [diff] [blame] | 516 | |
| 517 | out: |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 518 | dm_put_live_table(md, srcu_idx); |
Milan Broz | aa129a2 | 2006-10-03 01:15:15 -0700 | [diff] [blame] | 519 | |
Hannes Reinecke | 6c182cd | 2013-07-10 23:41:15 +0100 | [diff] [blame] | 520 | if (r == -ENOTCONN) { |
| 521 | msleep(10); |
| 522 | goto retry; |
| 523 | } |
| 524 | |
Milan Broz | aa129a2 | 2006-10-03 01:15:15 -0700 | [diff] [blame] | 525 | return r; |
| 526 | } |
| 527 | |
Alasdair G Kergon | 028867a | 2007-07-12 17:26:32 +0100 | [diff] [blame] | 528 | static struct dm_io *alloc_io(struct mapped_device *md) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 529 | { |
| 530 | return mempool_alloc(md->io_pool, GFP_NOIO); |
| 531 | } |
| 532 | |
Alasdair G Kergon | 028867a | 2007-07-12 17:26:32 +0100 | [diff] [blame] | 533 | static void free_io(struct mapped_device *md, struct dm_io *io) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 534 | { |
| 535 | mempool_free(io, md->io_pool); |
| 536 | } |
| 537 | |
Alasdair G Kergon | 028867a | 2007-07-12 17:26:32 +0100 | [diff] [blame] | 538 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 539 | { |
Mikulas Patocka | dba1416 | 2012-10-12 21:02:15 +0100 | [diff] [blame] | 540 | bio_put(&tio->clone); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 541 | } |
| 542 | |
Kiyoshi Ueda | 0888564 | 2009-12-10 23:52:15 +0000 | [diff] [blame] | 543 | static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md, |
| 544 | gfp_t gfp_mask) |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 545 | { |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 546 | return mempool_alloc(md->io_pool, gfp_mask); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 547 | } |
| 548 | |
| 549 | static void free_rq_tio(struct dm_rq_target_io *tio) |
| 550 | { |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 551 | mempool_free(tio, tio->md->io_pool); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 552 | } |
| 553 | |
Kiyoshi Ueda | 90abb8c | 2009-12-10 23:52:13 +0000 | [diff] [blame] | 554 | static int md_in_flight(struct mapped_device *md) |
| 555 | { |
| 556 | return atomic_read(&md->pending[READ]) + |
| 557 | atomic_read(&md->pending[WRITE]); |
| 558 | } |
| 559 | |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 560 | static void start_io_acct(struct dm_io *io) |
| 561 | { |
| 562 | struct mapped_device *md = io->md; |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 563 | struct bio *bio = io->bio; |
Tejun Heo | c995905 | 2008-08-25 19:47:21 +0900 | [diff] [blame] | 564 | int cpu; |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 565 | int rw = bio_data_dir(bio); |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 566 | |
| 567 | io->start_time = jiffies; |
| 568 | |
Tejun Heo | 074a7ac | 2008-08-25 19:56:14 +0900 | [diff] [blame] | 569 | cpu = part_stat_lock(); |
| 570 | part_round_stats(cpu, &dm_disk(md)->part0); |
| 571 | part_stat_unlock(); |
Shaohua Li | 1e9bb88 | 2011-03-22 08:35:35 +0100 | [diff] [blame] | 572 | atomic_set(&dm_disk(md)->part0.in_flight[rw], |
| 573 | atomic_inc_return(&md->pending[rw])); |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 574 | |
| 575 | if (unlikely(dm_stats_used(&md->stats))) |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 576 | dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_iter.bi_sector, |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 577 | bio_sectors(bio), false, 0, &io->stats_aux); |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 578 | } |
| 579 | |
Mikulas Patocka | d221d2e | 2008-11-13 23:39:10 +0000 | [diff] [blame] | 580 | static void end_io_acct(struct dm_io *io) |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 581 | { |
| 582 | struct mapped_device *md = io->md; |
| 583 | struct bio *bio = io->bio; |
| 584 | unsigned long duration = jiffies - io->start_time; |
Tejun Heo | c995905 | 2008-08-25 19:47:21 +0900 | [diff] [blame] | 585 | int pending, cpu; |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 586 | int rw = bio_data_dir(bio); |
| 587 | |
Tejun Heo | 074a7ac | 2008-08-25 19:56:14 +0900 | [diff] [blame] | 588 | cpu = part_stat_lock(); |
| 589 | part_round_stats(cpu, &dm_disk(md)->part0); |
| 590 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); |
| 591 | part_stat_unlock(); |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 592 | |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 593 | if (unlikely(dm_stats_used(&md->stats))) |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 594 | dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_iter.bi_sector, |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 595 | bio_sectors(bio), true, duration, &io->stats_aux); |
| 596 | |
Mikulas Patocka | af7e466 | 2009-04-09 00:27:16 +0100 | [diff] [blame] | 597 | /* |
| 598 | * After this is decremented the bio must not be touched if it is |
Tejun Heo | d87f4c1 | 2010-09-03 11:56:19 +0200 | [diff] [blame] | 599 | * a flush. |
Mikulas Patocka | af7e466 | 2009-04-09 00:27:16 +0100 | [diff] [blame] | 600 | */ |
Shaohua Li | 1e9bb88 | 2011-03-22 08:35:35 +0100 | [diff] [blame] | 601 | pending = atomic_dec_return(&md->pending[rw]); |
| 602 | atomic_set(&dm_disk(md)->part0.in_flight[rw], pending); |
Nikanth Karthikesan | 316d315 | 2009-10-06 20:16:55 +0200 | [diff] [blame] | 603 | pending += atomic_read(&md->pending[rw^0x1]); |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 604 | |
Mikulas Patocka | d221d2e | 2008-11-13 23:39:10 +0000 | [diff] [blame] | 605 | /* nudge anyone waiting on suspend queue */ |
| 606 | if (!pending) |
| 607 | wake_up(&md->wait); |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 608 | } |
| 609 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 610 | /* |
| 611 | * Add the bio to the list of deferred io. |
| 612 | */ |
Mikulas Patocka | 92c6390 | 2009-04-09 00:27:15 +0100 | [diff] [blame] | 613 | static void queue_io(struct mapped_device *md, struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 614 | { |
Kiyoshi Ueda | 05447420 | 2010-09-08 18:07:01 +0200 | [diff] [blame] | 615 | unsigned long flags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 616 | |
Kiyoshi Ueda | 05447420 | 2010-09-08 18:07:01 +0200 | [diff] [blame] | 617 | spin_lock_irqsave(&md->deferred_lock, flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 618 | bio_list_add(&md->deferred, bio); |
Kiyoshi Ueda | 05447420 | 2010-09-08 18:07:01 +0200 | [diff] [blame] | 619 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 620 | queue_work(md->wq, &md->work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 621 | } |
| 622 | |
| 623 | /* |
| 624 | * Everyone (including functions in this file), should use this |
| 625 | * function to access the md->map field, and make sure they call |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 626 | * dm_put_live_table() when finished. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 627 | */ |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 628 | struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 629 | { |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 630 | *srcu_idx = srcu_read_lock(&md->io_barrier); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 631 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 632 | return srcu_dereference(md->map, &md->io_barrier); |
| 633 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 634 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 635 | void dm_put_live_table(struct mapped_device *md, int srcu_idx) __releases(md->io_barrier) |
| 636 | { |
| 637 | srcu_read_unlock(&md->io_barrier, srcu_idx); |
| 638 | } |
| 639 | |
| 640 | void dm_sync_table(struct mapped_device *md) |
| 641 | { |
| 642 | synchronize_srcu(&md->io_barrier); |
| 643 | synchronize_rcu_expedited(); |
| 644 | } |
| 645 | |
| 646 | /* |
| 647 | * A fast alternative to dm_get_live_table/dm_put_live_table. |
| 648 | * The caller must not block between these two functions. |
| 649 | */ |
| 650 | static struct dm_table *dm_get_live_table_fast(struct mapped_device *md) __acquires(RCU) |
| 651 | { |
| 652 | rcu_read_lock(); |
| 653 | return rcu_dereference(md->map); |
| 654 | } |
| 655 | |
| 656 | static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU) |
| 657 | { |
| 658 | rcu_read_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 659 | } |
| 660 | |
Darrick J. Wong | 3ac51e7 | 2006-03-27 01:17:54 -0800 | [diff] [blame] | 661 | /* |
| 662 | * Get the geometry associated with a dm device |
| 663 | */ |
| 664 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) |
| 665 | { |
| 666 | *geo = md->geometry; |
| 667 | |
| 668 | return 0; |
| 669 | } |
| 670 | |
| 671 | /* |
| 672 | * Set the geometry of a device. |
| 673 | */ |
| 674 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) |
| 675 | { |
| 676 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; |
| 677 | |
| 678 | if (geo->start > sz) { |
| 679 | DMWARN("Start sector is beyond the geometry limits."); |
| 680 | return -EINVAL; |
| 681 | } |
| 682 | |
| 683 | md->geometry = *geo; |
| 684 | |
| 685 | return 0; |
| 686 | } |
| 687 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 688 | /*----------------------------------------------------------------- |
| 689 | * CRUD START: |
| 690 | * A more elegant soln is in the works that uses the queue |
| 691 | * merge fn, unfortunately there are a couple of changes to |
| 692 | * the block layer that I want to make for this. So in the |
| 693 | * interests of getting something for people to use I give |
| 694 | * you this clearly demarcated crap. |
| 695 | *---------------------------------------------------------------*/ |
| 696 | |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 697 | static int __noflush_suspending(struct mapped_device *md) |
| 698 | { |
| 699 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
| 700 | } |
| 701 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 702 | /* |
| 703 | * Decrements the number of outstanding ios that a bio has been |
| 704 | * cloned into, completing the original io if necc. |
| 705 | */ |
Arjan van de Ven | 858119e | 2006-01-14 13:20:43 -0800 | [diff] [blame] | 706 | static void dec_pending(struct dm_io *io, int error) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 707 | { |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 708 | unsigned long flags; |
Milan Broz | b35f8ca | 2009-03-16 17:44:36 +0000 | [diff] [blame] | 709 | int io_error; |
| 710 | struct bio *bio; |
| 711 | struct mapped_device *md = io->md; |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 712 | |
| 713 | /* Push-back supersedes any I/O errors */ |
Kiyoshi Ueda | f88fb98 | 2009-10-16 23:18:15 +0100 | [diff] [blame] | 714 | if (unlikely(error)) { |
| 715 | spin_lock_irqsave(&io->endio_lock, flags); |
| 716 | if (!(io->error > 0 && __noflush_suspending(md))) |
| 717 | io->error = error; |
| 718 | spin_unlock_irqrestore(&io->endio_lock, flags); |
| 719 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 720 | |
| 721 | if (atomic_dec_and_test(&io->io_count)) { |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 722 | if (io->error == DM_ENDIO_REQUEUE) { |
| 723 | /* |
| 724 | * Target requested pushing back the I/O. |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 725 | */ |
Mikulas Patocka | 022c261 | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 726 | spin_lock_irqsave(&md->deferred_lock, flags); |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 727 | if (__noflush_suspending(md)) |
| 728 | bio_list_add_head(&md->deferred, io->bio); |
| 729 | else |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 730 | /* noflush suspend was interrupted. */ |
| 731 | io->error = -EIO; |
Mikulas Patocka | 022c261 | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 732 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 733 | } |
| 734 | |
Milan Broz | b35f8ca | 2009-03-16 17:44:36 +0000 | [diff] [blame] | 735 | io_error = io->error; |
| 736 | bio = io->bio; |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 737 | end_io_acct(io); |
| 738 | free_io(md, io); |
Jens Axboe | 2056a78 | 2006-03-23 20:00:26 +0100 | [diff] [blame] | 739 | |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 740 | if (io_error == DM_ENDIO_REQUEUE) |
| 741 | return; |
| 742 | |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 743 | if ((bio->bi_rw & REQ_FLUSH) && bio->bi_iter.bi_size) { |
Mikulas Patocka | af7e466 | 2009-04-09 00:27:16 +0100 | [diff] [blame] | 744 | /* |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 745 | * Preflush done for flush with data, reissue |
| 746 | * without REQ_FLUSH. |
Mikulas Patocka | af7e466 | 2009-04-09 00:27:16 +0100 | [diff] [blame] | 747 | */ |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 748 | bio->bi_rw &= ~REQ_FLUSH; |
| 749 | queue_io(md, bio); |
Mikulas Patocka | af7e466 | 2009-04-09 00:27:16 +0100 | [diff] [blame] | 750 | } else { |
Mike Snitzer | b372d36 | 2010-09-08 18:07:01 +0200 | [diff] [blame] | 751 | /* done with normal IO or empty flush */ |
Linus Torvalds | 0a82a8d | 2013-04-18 09:00:26 -0700 | [diff] [blame] | 752 | trace_block_bio_complete(md->queue, bio, io_error); |
Mike Snitzer | b372d36 | 2010-09-08 18:07:01 +0200 | [diff] [blame] | 753 | bio_endio(bio, io_error); |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 754 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 755 | } |
| 756 | } |
| 757 | |
Mike Snitzer | 7eee4ae | 2014-06-02 15:50:06 -0400 | [diff] [blame] | 758 | static void disable_write_same(struct mapped_device *md) |
| 759 | { |
| 760 | struct queue_limits *limits = dm_get_queue_limits(md); |
| 761 | |
| 762 | /* device doesn't really support WRITE SAME, disable it */ |
| 763 | limits->max_write_same_sectors = 0; |
| 764 | } |
| 765 | |
NeilBrown | 6712ecf | 2007-09-27 12:47:43 +0200 | [diff] [blame] | 766 | static void clone_endio(struct bio *bio, int error) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 767 | { |
| 768 | int r = 0; |
Mikulas Patocka | bfc6d41 | 2014-03-04 18:24:49 -0500 | [diff] [blame] | 769 | struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone); |
Milan Broz | b35f8ca | 2009-03-16 17:44:36 +0000 | [diff] [blame] | 770 | struct dm_io *io = tio->io; |
Stefan Bader | 9faf400 | 2006-10-03 01:15:41 -0700 | [diff] [blame] | 771 | struct mapped_device *md = tio->io->md; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 772 | dm_endio_fn endio = tio->ti->type->end_io; |
| 773 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 774 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
| 775 | error = -EIO; |
| 776 | |
| 777 | if (endio) { |
Mikulas Patocka | 7de3ee5 | 2012-12-21 20:23:41 +0000 | [diff] [blame] | 778 | r = endio(tio->ti, bio, error); |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 779 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
| 780 | /* |
| 781 | * error and requeue request are handled |
| 782 | * in dec_pending(). |
| 783 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 784 | error = r; |
Kiyoshi Ueda | 45cbcd7 | 2006-12-08 02:41:05 -0800 | [diff] [blame] | 785 | else if (r == DM_ENDIO_INCOMPLETE) |
| 786 | /* The target will handle the io */ |
NeilBrown | 6712ecf | 2007-09-27 12:47:43 +0200 | [diff] [blame] | 787 | return; |
Kiyoshi Ueda | 45cbcd7 | 2006-12-08 02:41:05 -0800 | [diff] [blame] | 788 | else if (r) { |
| 789 | DMWARN("unimplemented target endio return value: %d", r); |
| 790 | BUG(); |
| 791 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 792 | } |
| 793 | |
Mike Snitzer | 7eee4ae | 2014-06-02 15:50:06 -0400 | [diff] [blame] | 794 | if (unlikely(r == -EREMOTEIO && (bio->bi_rw & REQ_WRITE_SAME) && |
| 795 | !bdev_get_queue(bio->bi_bdev)->limits.max_write_same_sectors)) |
| 796 | disable_write_same(md); |
| 797 | |
Stefan Bader | 9faf400 | 2006-10-03 01:15:41 -0700 | [diff] [blame] | 798 | free_tio(md, tio); |
Milan Broz | b35f8ca | 2009-03-16 17:44:36 +0000 | [diff] [blame] | 799 | dec_pending(io, error); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 800 | } |
| 801 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 802 | /* |
| 803 | * Partial completion handling for request-based dm |
| 804 | */ |
| 805 | static void end_clone_bio(struct bio *clone, int error) |
| 806 | { |
Mikulas Patocka | bfc6d41 | 2014-03-04 18:24:49 -0500 | [diff] [blame] | 807 | struct dm_rq_clone_bio_info *info = |
| 808 | container_of(clone, struct dm_rq_clone_bio_info, clone); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 809 | struct dm_rq_target_io *tio = info->tio; |
| 810 | struct bio *bio = info->orig; |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 811 | unsigned int nr_bytes = info->orig->bi_iter.bi_size; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 812 | |
| 813 | bio_put(clone); |
| 814 | |
| 815 | if (tio->error) |
| 816 | /* |
| 817 | * An error has already been detected on the request. |
| 818 | * Once error occurred, just let clone->end_io() handle |
| 819 | * the remainder. |
| 820 | */ |
| 821 | return; |
| 822 | else if (error) { |
| 823 | /* |
| 824 | * Don't notice the error to the upper layer yet. |
| 825 | * The error handling decision is made by the target driver, |
| 826 | * when the request is completed. |
| 827 | */ |
| 828 | tio->error = error; |
| 829 | return; |
| 830 | } |
| 831 | |
| 832 | /* |
| 833 | * I/O for the bio successfully completed. |
| 834 | * Notice the data completion to the upper layer. |
| 835 | */ |
| 836 | |
| 837 | /* |
| 838 | * bios are processed from the head of the list. |
| 839 | * So the completing bio should always be rq->bio. |
| 840 | * If it's not, something wrong is happening. |
| 841 | */ |
| 842 | if (tio->orig->bio != bio) |
| 843 | DMERR("bio completion is going in the middle of the request"); |
| 844 | |
| 845 | /* |
| 846 | * Update the original request. |
| 847 | * Do not use blk_end_request() here, because it may complete |
| 848 | * the original request before the clone, and break the ordering. |
| 849 | */ |
| 850 | blk_update_request(tio->orig, 0, nr_bytes); |
| 851 | } |
| 852 | |
| 853 | /* |
| 854 | * Don't touch any member of the md after calling this function because |
| 855 | * the md may be freed in dm_put() at the end of this function. |
| 856 | * Or do dm_get() before calling this function and dm_put() later. |
| 857 | */ |
Kiyoshi Ueda | b4324fe | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 858 | static void rq_completed(struct mapped_device *md, int rw, int run_queue) |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 859 | { |
Kiyoshi Ueda | b4324fe | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 860 | atomic_dec(&md->pending[rw]); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 861 | |
| 862 | /* nudge anyone waiting on suspend queue */ |
Kiyoshi Ueda | b4324fe | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 863 | if (!md_in_flight(md)) |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 864 | wake_up(&md->wait); |
| 865 | |
Jens Axboe | a8c32a5 | 2012-11-06 12:24:26 +0100 | [diff] [blame] | 866 | /* |
| 867 | * Run this off this callpath, as drivers could invoke end_io while |
| 868 | * inside their request_fn (and holding the queue lock). Calling |
| 869 | * back into ->request_fn() could deadlock attempting to grab the |
| 870 | * queue lock again. |
| 871 | */ |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 872 | if (run_queue) |
Jens Axboe | a8c32a5 | 2012-11-06 12:24:26 +0100 | [diff] [blame] | 873 | blk_run_queue_async(md->queue); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 874 | |
| 875 | /* |
| 876 | * dm_put() must be at the end of this function. See the comment above |
| 877 | */ |
| 878 | dm_put(md); |
| 879 | } |
| 880 | |
Kiyoshi Ueda | a77e28c | 2009-09-04 20:40:16 +0100 | [diff] [blame] | 881 | static void free_rq_clone(struct request *clone) |
| 882 | { |
| 883 | struct dm_rq_target_io *tio = clone->end_io_data; |
| 884 | |
| 885 | blk_rq_unprep_clone(clone); |
| 886 | free_rq_tio(tio); |
| 887 | } |
| 888 | |
Kiyoshi Ueda | 980691e | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 889 | /* |
| 890 | * Complete the clone and the original request. |
| 891 | * Must be called without queue lock. |
| 892 | */ |
| 893 | static void dm_end_request(struct request *clone, int error) |
| 894 | { |
| 895 | int rw = rq_data_dir(clone); |
| 896 | struct dm_rq_target_io *tio = clone->end_io_data; |
| 897 | struct mapped_device *md = tio->md; |
| 898 | struct request *rq = tio->orig; |
| 899 | |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 900 | if (rq->cmd_type == REQ_TYPE_BLOCK_PC) { |
Kiyoshi Ueda | 980691e | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 901 | rq->errors = clone->errors; |
| 902 | rq->resid_len = clone->resid_len; |
| 903 | |
| 904 | if (rq->sense) |
| 905 | /* |
| 906 | * We are using the sense buffer of the original |
| 907 | * request. |
| 908 | * So setting the length of the sense data is enough. |
| 909 | */ |
| 910 | rq->sense_len = clone->sense_len; |
| 911 | } |
| 912 | |
| 913 | free_rq_clone(clone); |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 914 | blk_end_request_all(rq, error); |
| 915 | rq_completed(md, rw, true); |
Kiyoshi Ueda | 980691e | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 916 | } |
| 917 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 918 | static void dm_unprep_request(struct request *rq) |
| 919 | { |
| 920 | struct request *clone = rq->special; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 921 | |
| 922 | rq->special = NULL; |
| 923 | rq->cmd_flags &= ~REQ_DONTPREP; |
| 924 | |
Kiyoshi Ueda | a77e28c | 2009-09-04 20:40:16 +0100 | [diff] [blame] | 925 | free_rq_clone(clone); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 926 | } |
| 927 | |
| 928 | /* |
| 929 | * Requeue the original request of a clone. |
| 930 | */ |
| 931 | void dm_requeue_unmapped_request(struct request *clone) |
| 932 | { |
Kiyoshi Ueda | b4324fe | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 933 | int rw = rq_data_dir(clone); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 934 | struct dm_rq_target_io *tio = clone->end_io_data; |
| 935 | struct mapped_device *md = tio->md; |
| 936 | struct request *rq = tio->orig; |
| 937 | struct request_queue *q = rq->q; |
| 938 | unsigned long flags; |
| 939 | |
| 940 | dm_unprep_request(rq); |
| 941 | |
| 942 | spin_lock_irqsave(q->queue_lock, flags); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 943 | blk_requeue_request(q, rq); |
| 944 | spin_unlock_irqrestore(q->queue_lock, flags); |
| 945 | |
Kiyoshi Ueda | b4324fe | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 946 | rq_completed(md, rw, 0); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 947 | } |
| 948 | EXPORT_SYMBOL_GPL(dm_requeue_unmapped_request); |
| 949 | |
| 950 | static void __stop_queue(struct request_queue *q) |
| 951 | { |
| 952 | blk_stop_queue(q); |
| 953 | } |
| 954 | |
| 955 | static void stop_queue(struct request_queue *q) |
| 956 | { |
| 957 | unsigned long flags; |
| 958 | |
| 959 | spin_lock_irqsave(q->queue_lock, flags); |
| 960 | __stop_queue(q); |
| 961 | spin_unlock_irqrestore(q->queue_lock, flags); |
| 962 | } |
| 963 | |
| 964 | static void __start_queue(struct request_queue *q) |
| 965 | { |
| 966 | if (blk_queue_stopped(q)) |
| 967 | blk_start_queue(q); |
| 968 | } |
| 969 | |
| 970 | static void start_queue(struct request_queue *q) |
| 971 | { |
| 972 | unsigned long flags; |
| 973 | |
| 974 | spin_lock_irqsave(q->queue_lock, flags); |
| 975 | __start_queue(q); |
| 976 | spin_unlock_irqrestore(q->queue_lock, flags); |
| 977 | } |
| 978 | |
Kiyoshi Ueda | 11a6824 | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 979 | static void dm_done(struct request *clone, int error, bool mapped) |
| 980 | { |
| 981 | int r = error; |
| 982 | struct dm_rq_target_io *tio = clone->end_io_data; |
Mike Snitzer | ba1cbad | 2012-09-26 23:45:42 +0100 | [diff] [blame] | 983 | dm_request_endio_fn rq_end_io = NULL; |
Kiyoshi Ueda | 11a6824 | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 984 | |
Mike Snitzer | ba1cbad | 2012-09-26 23:45:42 +0100 | [diff] [blame] | 985 | if (tio->ti) { |
| 986 | rq_end_io = tio->ti->type->rq_end_io; |
| 987 | |
| 988 | if (mapped && rq_end_io) |
| 989 | r = rq_end_io(tio->ti, clone, error, &tio->info); |
| 990 | } |
Kiyoshi Ueda | 11a6824 | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 991 | |
Mike Snitzer | 7eee4ae | 2014-06-02 15:50:06 -0400 | [diff] [blame] | 992 | if (unlikely(r == -EREMOTEIO && (clone->cmd_flags & REQ_WRITE_SAME) && |
| 993 | !clone->q->limits.max_write_same_sectors)) |
| 994 | disable_write_same(tio->md); |
| 995 | |
Kiyoshi Ueda | 11a6824 | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 996 | if (r <= 0) |
| 997 | /* The target wants to complete the I/O */ |
| 998 | dm_end_request(clone, r); |
| 999 | else if (r == DM_ENDIO_INCOMPLETE) |
| 1000 | /* The target will handle the I/O */ |
| 1001 | return; |
| 1002 | else if (r == DM_ENDIO_REQUEUE) |
| 1003 | /* The target wants to requeue the I/O */ |
| 1004 | dm_requeue_unmapped_request(clone); |
| 1005 | else { |
| 1006 | DMWARN("unimplemented target endio return value: %d", r); |
| 1007 | BUG(); |
| 1008 | } |
| 1009 | } |
| 1010 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1011 | /* |
| 1012 | * Request completion handler for request-based dm |
| 1013 | */ |
| 1014 | static void dm_softirq_done(struct request *rq) |
| 1015 | { |
Kiyoshi Ueda | 11a6824 | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 1016 | bool mapped = true; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1017 | struct request *clone = rq->completion_data; |
| 1018 | struct dm_rq_target_io *tio = clone->end_io_data; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1019 | |
Kiyoshi Ueda | 11a6824 | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 1020 | if (rq->cmd_flags & REQ_FAILED) |
| 1021 | mapped = false; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1022 | |
Kiyoshi Ueda | 11a6824 | 2009-12-10 23:52:17 +0000 | [diff] [blame] | 1023 | dm_done(clone, tio->error, mapped); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1024 | } |
| 1025 | |
| 1026 | /* |
| 1027 | * Complete the clone and the original request with the error status |
| 1028 | * through softirq context. |
| 1029 | */ |
| 1030 | static void dm_complete_request(struct request *clone, int error) |
| 1031 | { |
| 1032 | struct dm_rq_target_io *tio = clone->end_io_data; |
| 1033 | struct request *rq = tio->orig; |
| 1034 | |
| 1035 | tio->error = error; |
| 1036 | rq->completion_data = clone; |
| 1037 | blk_complete_request(rq); |
| 1038 | } |
| 1039 | |
| 1040 | /* |
| 1041 | * Complete the not-mapped clone and the original request with the error status |
| 1042 | * through softirq context. |
| 1043 | * Target's rq_end_io() function isn't called. |
| 1044 | * This may be used when the target's map_rq() function fails. |
| 1045 | */ |
| 1046 | void dm_kill_unmapped_request(struct request *clone, int error) |
| 1047 | { |
| 1048 | struct dm_rq_target_io *tio = clone->end_io_data; |
| 1049 | struct request *rq = tio->orig; |
| 1050 | |
| 1051 | rq->cmd_flags |= REQ_FAILED; |
| 1052 | dm_complete_request(clone, error); |
| 1053 | } |
| 1054 | EXPORT_SYMBOL_GPL(dm_kill_unmapped_request); |
| 1055 | |
| 1056 | /* |
| 1057 | * Called with the queue lock held |
| 1058 | */ |
| 1059 | static void end_clone_request(struct request *clone, int error) |
| 1060 | { |
| 1061 | /* |
| 1062 | * For just cleaning up the information of the queue in which |
| 1063 | * the clone was dispatched. |
| 1064 | * The clone is *NOT* freed actually here because it is alloced from |
| 1065 | * dm own mempool and REQ_ALLOCED isn't set in clone->cmd_flags. |
| 1066 | */ |
| 1067 | __blk_put_request(clone->q, clone); |
| 1068 | |
| 1069 | /* |
| 1070 | * Actual request completion is done in a softirq context which doesn't |
| 1071 | * hold the queue lock. Otherwise, deadlock could occur because: |
| 1072 | * - another request may be submitted by the upper level driver |
| 1073 | * of the stacking during the completion |
| 1074 | * - the submission which requires queue lock may be done |
| 1075 | * against this queue |
| 1076 | */ |
| 1077 | dm_complete_request(clone, error); |
| 1078 | } |
| 1079 | |
Mike Snitzer | 56a67df | 2010-08-12 04:14:10 +0100 | [diff] [blame] | 1080 | /* |
| 1081 | * Return maximum size of I/O possible at the supplied sector up to the current |
| 1082 | * target boundary. |
| 1083 | */ |
| 1084 | static sector_t max_io_len_target_boundary(sector_t sector, struct dm_target *ti) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1085 | { |
Mike Snitzer | 56a67df | 2010-08-12 04:14:10 +0100 | [diff] [blame] | 1086 | sector_t target_offset = dm_target_offset(ti, sector); |
| 1087 | |
| 1088 | return ti->len - target_offset; |
| 1089 | } |
| 1090 | |
| 1091 | static sector_t max_io_len(sector_t sector, struct dm_target *ti) |
| 1092 | { |
| 1093 | sector_t len = max_io_len_target_boundary(sector, ti); |
Mike Snitzer | 542f903 | 2012-07-27 15:08:00 +0100 | [diff] [blame] | 1094 | sector_t offset, max_len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1095 | |
| 1096 | /* |
Mike Snitzer | 542f903 | 2012-07-27 15:08:00 +0100 | [diff] [blame] | 1097 | * Does the target need to split even further? |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1098 | */ |
Mike Snitzer | 542f903 | 2012-07-27 15:08:00 +0100 | [diff] [blame] | 1099 | if (ti->max_io_len) { |
| 1100 | offset = dm_target_offset(ti, sector); |
| 1101 | if (unlikely(ti->max_io_len & (ti->max_io_len - 1))) |
| 1102 | max_len = sector_div(offset, ti->max_io_len); |
| 1103 | else |
| 1104 | max_len = offset & (ti->max_io_len - 1); |
| 1105 | max_len = ti->max_io_len - max_len; |
| 1106 | |
| 1107 | if (len > max_len) |
| 1108 | len = max_len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1109 | } |
| 1110 | |
| 1111 | return len; |
| 1112 | } |
| 1113 | |
Mike Snitzer | 542f903 | 2012-07-27 15:08:00 +0100 | [diff] [blame] | 1114 | int dm_set_target_max_io_len(struct dm_target *ti, sector_t len) |
| 1115 | { |
| 1116 | if (len > UINT_MAX) { |
| 1117 | DMERR("Specified maximum size of target IO (%llu) exceeds limit (%u)", |
| 1118 | (unsigned long long)len, UINT_MAX); |
| 1119 | ti->error = "Maximum size of target IO is too large"; |
| 1120 | return -EINVAL; |
| 1121 | } |
| 1122 | |
| 1123 | ti->max_io_len = (uint32_t) len; |
| 1124 | |
| 1125 | return 0; |
| 1126 | } |
| 1127 | EXPORT_SYMBOL_GPL(dm_set_target_max_io_len); |
| 1128 | |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1129 | /* |
| 1130 | * A target may call dm_accept_partial_bio only from the map routine. It is |
| 1131 | * allowed for all bio types except REQ_FLUSH. |
| 1132 | * |
| 1133 | * dm_accept_partial_bio informs the dm that the target only wants to process |
| 1134 | * additional n_sectors sectors of the bio and the rest of the data should be |
| 1135 | * sent in a next bio. |
| 1136 | * |
| 1137 | * A diagram that explains the arithmetics: |
| 1138 | * +--------------------+---------------+-------+ |
| 1139 | * | 1 | 2 | 3 | |
| 1140 | * +--------------------+---------------+-------+ |
| 1141 | * |
| 1142 | * <-------------- *tio->len_ptr ---------------> |
| 1143 | * <------- bi_size -------> |
| 1144 | * <-- n_sectors --> |
| 1145 | * |
| 1146 | * Region 1 was already iterated over with bio_advance or similar function. |
| 1147 | * (it may be empty if the target doesn't use bio_advance) |
| 1148 | * Region 2 is the remaining bio size that the target wants to process. |
| 1149 | * (it may be empty if region 1 is non-empty, although there is no reason |
| 1150 | * to make it empty) |
| 1151 | * The target requires that region 3 is to be sent in the next bio. |
| 1152 | * |
| 1153 | * If the target wants to receive multiple copies of the bio (via num_*bios, etc), |
| 1154 | * the partially processed part (the sum of regions 1+2) must be the same for all |
| 1155 | * copies of the bio. |
| 1156 | */ |
| 1157 | void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors) |
| 1158 | { |
| 1159 | struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone); |
| 1160 | unsigned bi_size = bio->bi_iter.bi_size >> SECTOR_SHIFT; |
| 1161 | BUG_ON(bio->bi_rw & REQ_FLUSH); |
| 1162 | BUG_ON(bi_size > *tio->len_ptr); |
| 1163 | BUG_ON(n_sectors > bi_size); |
| 1164 | *tio->len_ptr -= bi_size - n_sectors; |
| 1165 | bio->bi_iter.bi_size = n_sectors << SECTOR_SHIFT; |
| 1166 | } |
| 1167 | EXPORT_SYMBOL_GPL(dm_accept_partial_bio); |
| 1168 | |
Alasdair G Kergon | bd2a49b | 2013-03-01 22:45:46 +0000 | [diff] [blame] | 1169 | static void __map_bio(struct dm_target_io *tio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1170 | { |
| 1171 | int r; |
Jens Axboe | 2056a78 | 2006-03-23 20:00:26 +0100 | [diff] [blame] | 1172 | sector_t sector; |
Stefan Bader | 9faf400 | 2006-10-03 01:15:41 -0700 | [diff] [blame] | 1173 | struct mapped_device *md; |
Mikulas Patocka | dba1416 | 2012-10-12 21:02:15 +0100 | [diff] [blame] | 1174 | struct bio *clone = &tio->clone; |
Alasdair G Kergon | bd2a49b | 2013-03-01 22:45:46 +0000 | [diff] [blame] | 1175 | struct dm_target *ti = tio->ti; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1176 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1177 | clone->bi_end_io = clone_endio; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1178 | |
| 1179 | /* |
| 1180 | * Map the clone. If r == 0 we don't need to do |
| 1181 | * anything, the target has assumed ownership of |
| 1182 | * this io. |
| 1183 | */ |
| 1184 | atomic_inc(&tio->io->io_count); |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1185 | sector = clone->bi_iter.bi_sector; |
Mikulas Patocka | 7de3ee5 | 2012-12-21 20:23:41 +0000 | [diff] [blame] | 1186 | r = ti->type->map(ti, clone); |
Kiyoshi Ueda | 45cbcd7 | 2006-12-08 02:41:05 -0800 | [diff] [blame] | 1187 | if (r == DM_MAPIO_REMAPPED) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1188 | /* the bio has been remapped so dispatch it */ |
Jens Axboe | 2056a78 | 2006-03-23 20:00:26 +0100 | [diff] [blame] | 1189 | |
Mike Snitzer | d07335e | 2010-11-16 12:52:38 +0100 | [diff] [blame] | 1190 | trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone, |
| 1191 | tio->io->bio->bi_bdev->bd_dev, sector); |
Jens Axboe | 2056a78 | 2006-03-23 20:00:26 +0100 | [diff] [blame] | 1192 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1193 | generic_make_request(clone); |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 1194 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
| 1195 | /* error the io and bail out, or requeue it if needed */ |
Stefan Bader | 9faf400 | 2006-10-03 01:15:41 -0700 | [diff] [blame] | 1196 | md = tio->io->md; |
| 1197 | dec_pending(tio->io, r); |
Stefan Bader | 9faf400 | 2006-10-03 01:15:41 -0700 | [diff] [blame] | 1198 | free_tio(md, tio); |
Kiyoshi Ueda | 45cbcd7 | 2006-12-08 02:41:05 -0800 | [diff] [blame] | 1199 | } else if (r) { |
| 1200 | DMWARN("unimplemented target map return value: %d", r); |
| 1201 | BUG(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1202 | } |
| 1203 | } |
| 1204 | |
| 1205 | struct clone_info { |
| 1206 | struct mapped_device *md; |
| 1207 | struct dm_table *map; |
| 1208 | struct bio *bio; |
| 1209 | struct dm_io *io; |
| 1210 | sector_t sector; |
Mikulas Patocka | e0d6609 | 2014-03-14 18:40:39 -0400 | [diff] [blame] | 1211 | unsigned sector_count; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1212 | }; |
| 1213 | |
Mikulas Patocka | e0d6609 | 2014-03-14 18:40:39 -0400 | [diff] [blame] | 1214 | static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len) |
Alasdair G Kergon | bd2a49b | 2013-03-01 22:45:46 +0000 | [diff] [blame] | 1215 | { |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1216 | bio->bi_iter.bi_sector = sector; |
| 1217 | bio->bi_iter.bi_size = to_bytes(len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1218 | } |
| 1219 | |
| 1220 | /* |
| 1221 | * Creates a bio that consists of range of complete bvecs. |
| 1222 | */ |
Mikulas Patocka | dba1416 | 2012-10-12 21:02:15 +0100 | [diff] [blame] | 1223 | static void clone_bio(struct dm_target_io *tio, struct bio *bio, |
Kent Overstreet | 1c3b13e | 2013-10-29 17:17:49 -0700 | [diff] [blame] | 1224 | sector_t sector, unsigned len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1225 | { |
Mikulas Patocka | dba1416 | 2012-10-12 21:02:15 +0100 | [diff] [blame] | 1226 | struct bio *clone = &tio->clone; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1227 | |
Kent Overstreet | 1c3b13e | 2013-10-29 17:17:49 -0700 | [diff] [blame] | 1228 | __bio_clone_fast(clone, bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1229 | |
Kent Overstreet | 1c3b13e | 2013-10-29 17:17:49 -0700 | [diff] [blame] | 1230 | if (bio_integrity(bio)) |
| 1231 | bio_integrity_clone(clone, bio, GFP_NOIO); |
| 1232 | |
| 1233 | bio_advance(clone, to_bytes(sector - clone->bi_iter.bi_sector)); |
| 1234 | clone->bi_iter.bi_size = to_bytes(len); |
| 1235 | |
| 1236 | if (bio_integrity(bio)) |
| 1237 | bio_integrity_trim(clone, 0, len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1238 | } |
| 1239 | |
Alasdair G Kergon | 9015df2 | 2009-06-22 10:12:21 +0100 | [diff] [blame] | 1240 | static struct dm_target_io *alloc_tio(struct clone_info *ci, |
Alasdair G Kergon | bd2a49b | 2013-03-01 22:45:46 +0000 | [diff] [blame] | 1241 | struct dm_target *ti, int nr_iovecs, |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1242 | unsigned target_bio_nr) |
Mikulas Patocka | f9ab94c | 2009-06-22 10:12:20 +0100 | [diff] [blame] | 1243 | { |
Mikulas Patocka | dba1416 | 2012-10-12 21:02:15 +0100 | [diff] [blame] | 1244 | struct dm_target_io *tio; |
| 1245 | struct bio *clone; |
| 1246 | |
| 1247 | clone = bio_alloc_bioset(GFP_NOIO, nr_iovecs, ci->md->bs); |
| 1248 | tio = container_of(clone, struct dm_target_io, clone); |
Mikulas Patocka | f9ab94c | 2009-06-22 10:12:20 +0100 | [diff] [blame] | 1249 | |
| 1250 | tio->io = ci->io; |
| 1251 | tio->ti = ti; |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1252 | tio->target_bio_nr = target_bio_nr; |
Alasdair G Kergon | 9015df2 | 2009-06-22 10:12:21 +0100 | [diff] [blame] | 1253 | |
| 1254 | return tio; |
| 1255 | } |
| 1256 | |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1257 | static void __clone_and_map_simple_bio(struct clone_info *ci, |
| 1258 | struct dm_target *ti, |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1259 | unsigned target_bio_nr, unsigned *len) |
Alasdair G Kergon | 9015df2 | 2009-06-22 10:12:21 +0100 | [diff] [blame] | 1260 | { |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1261 | struct dm_target_io *tio = alloc_tio(ci, ti, ci->bio->bi_max_vecs, target_bio_nr); |
Mikulas Patocka | dba1416 | 2012-10-12 21:02:15 +0100 | [diff] [blame] | 1262 | struct bio *clone = &tio->clone; |
Alasdair G Kergon | 9015df2 | 2009-06-22 10:12:21 +0100 | [diff] [blame] | 1263 | |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1264 | tio->len_ptr = len; |
| 1265 | |
Mike Snitzer | 06a426c | 2010-08-12 04:14:09 +0100 | [diff] [blame] | 1266 | /* |
| 1267 | * Discard requests require the bio's inline iovecs be initialized. |
| 1268 | * ci->bio->bi_max_vecs is BIO_INLINE_VECS anyway, for both flush |
| 1269 | * and discard, so no need for concern about wasted bvec allocations. |
| 1270 | */ |
Kent Overstreet | 1c3b13e | 2013-10-29 17:17:49 -0700 | [diff] [blame] | 1271 | __bio_clone_fast(clone, ci->bio); |
Alasdair G Kergon | bd2a49b | 2013-03-01 22:45:46 +0000 | [diff] [blame] | 1272 | if (len) |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1273 | bio_setup_sector(clone, ci->sector, *len); |
Mikulas Patocka | f9ab94c | 2009-06-22 10:12:20 +0100 | [diff] [blame] | 1274 | |
Alasdair G Kergon | bd2a49b | 2013-03-01 22:45:46 +0000 | [diff] [blame] | 1275 | __map_bio(tio); |
Mikulas Patocka | f9ab94c | 2009-06-22 10:12:20 +0100 | [diff] [blame] | 1276 | } |
| 1277 | |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1278 | static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti, |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1279 | unsigned num_bios, unsigned *len) |
Mike Snitzer | 06a426c | 2010-08-12 04:14:09 +0100 | [diff] [blame] | 1280 | { |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1281 | unsigned target_bio_nr; |
Mike Snitzer | 06a426c | 2010-08-12 04:14:09 +0100 | [diff] [blame] | 1282 | |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1283 | for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++) |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1284 | __clone_and_map_simple_bio(ci, ti, target_bio_nr, len); |
Mike Snitzer | 06a426c | 2010-08-12 04:14:09 +0100 | [diff] [blame] | 1285 | } |
| 1286 | |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1287 | static int __send_empty_flush(struct clone_info *ci) |
Mikulas Patocka | f9ab94c | 2009-06-22 10:12:20 +0100 | [diff] [blame] | 1288 | { |
Mike Snitzer | 06a426c | 2010-08-12 04:14:09 +0100 | [diff] [blame] | 1289 | unsigned target_nr = 0; |
Mikulas Patocka | f9ab94c | 2009-06-22 10:12:20 +0100 | [diff] [blame] | 1290 | struct dm_target *ti; |
| 1291 | |
Mike Snitzer | b372d36 | 2010-09-08 18:07:01 +0200 | [diff] [blame] | 1292 | BUG_ON(bio_has_data(ci->bio)); |
Mikulas Patocka | f9ab94c | 2009-06-22 10:12:20 +0100 | [diff] [blame] | 1293 | while ((ti = dm_table_get_target(ci->map, target_nr++))) |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1294 | __send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL); |
Mikulas Patocka | f9ab94c | 2009-06-22 10:12:20 +0100 | [diff] [blame] | 1295 | |
Mikulas Patocka | f9ab94c | 2009-06-22 10:12:20 +0100 | [diff] [blame] | 1296 | return 0; |
| 1297 | } |
| 1298 | |
Alasdair G Kergon | e4c9381 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1299 | static void __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti, |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1300 | sector_t sector, unsigned *len) |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1301 | { |
Mikulas Patocka | dba1416 | 2012-10-12 21:02:15 +0100 | [diff] [blame] | 1302 | struct bio *bio = ci->bio; |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1303 | struct dm_target_io *tio; |
Alasdair G Kergon | b0d8ed4 | 2013-03-01 22:45:49 +0000 | [diff] [blame] | 1304 | unsigned target_bio_nr; |
| 1305 | unsigned num_target_bios = 1; |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1306 | |
Alasdair G Kergon | b0d8ed4 | 2013-03-01 22:45:49 +0000 | [diff] [blame] | 1307 | /* |
| 1308 | * Does the target want to receive duplicate copies of the bio? |
| 1309 | */ |
| 1310 | if (bio_data_dir(bio) == WRITE && ti->num_write_bios) |
| 1311 | num_target_bios = ti->num_write_bios(ti, bio); |
Alasdair G Kergon | e4c9381 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1312 | |
Alasdair G Kergon | b0d8ed4 | 2013-03-01 22:45:49 +0000 | [diff] [blame] | 1313 | for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) { |
Kent Overstreet | 1c3b13e | 2013-10-29 17:17:49 -0700 | [diff] [blame] | 1314 | tio = alloc_tio(ci, ti, 0, target_bio_nr); |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1315 | tio->len_ptr = len; |
| 1316 | clone_bio(tio, bio, sector, *len); |
Alasdair G Kergon | b0d8ed4 | 2013-03-01 22:45:49 +0000 | [diff] [blame] | 1317 | __map_bio(tio); |
| 1318 | } |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1319 | } |
| 1320 | |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1321 | typedef unsigned (*get_num_bios_fn)(struct dm_target *ti); |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1322 | |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1323 | static unsigned get_num_discard_bios(struct dm_target *ti) |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1324 | { |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1325 | return ti->num_discard_bios; |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1326 | } |
| 1327 | |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1328 | static unsigned get_num_write_same_bios(struct dm_target *ti) |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1329 | { |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1330 | return ti->num_write_same_bios; |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1331 | } |
| 1332 | |
| 1333 | typedef bool (*is_split_required_fn)(struct dm_target *ti); |
| 1334 | |
| 1335 | static bool is_split_required_for_discard(struct dm_target *ti) |
| 1336 | { |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1337 | return ti->split_discard_bios; |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1338 | } |
| 1339 | |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1340 | static int __send_changing_extent_only(struct clone_info *ci, |
| 1341 | get_num_bios_fn get_num_bios, |
| 1342 | is_split_required_fn is_split_required) |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1343 | { |
| 1344 | struct dm_target *ti; |
Mikulas Patocka | e0d6609 | 2014-03-14 18:40:39 -0400 | [diff] [blame] | 1345 | unsigned len; |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1346 | unsigned num_bios; |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1347 | |
Mike Snitzer | a79245b | 2010-08-12 04:14:24 +0100 | [diff] [blame] | 1348 | do { |
| 1349 | ti = dm_table_find_target(ci->map, ci->sector); |
| 1350 | if (!dm_target_is_valid(ti)) |
| 1351 | return -EIO; |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1352 | |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1353 | /* |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1354 | * Even though the device advertised support for this type of |
| 1355 | * request, that does not mean every target supports it, and |
Mike Snitzer | 936688d | 2011-08-02 12:32:01 +0100 | [diff] [blame] | 1356 | * reconfiguration might also have changed that since the |
Mike Snitzer | a79245b | 2010-08-12 04:14:24 +0100 | [diff] [blame] | 1357 | * check was performed. |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1358 | */ |
Alasdair G Kergon | 55a62ee | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1359 | num_bios = get_num_bios ? get_num_bios(ti) : 0; |
| 1360 | if (!num_bios) |
Mike Snitzer | a79245b | 2010-08-12 04:14:24 +0100 | [diff] [blame] | 1361 | return -EOPNOTSUPP; |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1362 | |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1363 | if (is_split_required && !is_split_required(ti)) |
Mikulas Patocka | e0d6609 | 2014-03-14 18:40:39 -0400 | [diff] [blame] | 1364 | len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti)); |
Mikulas Patocka | 7acf027 | 2012-07-27 15:08:03 +0100 | [diff] [blame] | 1365 | else |
Mikulas Patocka | e0d6609 | 2014-03-14 18:40:39 -0400 | [diff] [blame] | 1366 | len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti)); |
Mike Snitzer | 06a426c | 2010-08-12 04:14:09 +0100 | [diff] [blame] | 1367 | |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1368 | __send_duplicate_bios(ci, ti, num_bios, &len); |
Mike Snitzer | a79245b | 2010-08-12 04:14:24 +0100 | [diff] [blame] | 1369 | |
| 1370 | ci->sector += len; |
| 1371 | } while (ci->sector_count -= len); |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1372 | |
| 1373 | return 0; |
| 1374 | } |
| 1375 | |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1376 | static int __send_discard(struct clone_info *ci) |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1377 | { |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1378 | return __send_changing_extent_only(ci, get_num_discard_bios, |
| 1379 | is_split_required_for_discard); |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1380 | } |
| 1381 | |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1382 | static int __send_write_same(struct clone_info *ci) |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1383 | { |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1384 | return __send_changing_extent_only(ci, get_num_write_same_bios, NULL); |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1385 | } |
| 1386 | |
Alasdair G Kergon | e4c9381 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1387 | /* |
Alasdair G Kergon | e4c9381 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1388 | * Select the correct strategy for processing a non-flush bio. |
| 1389 | */ |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1390 | static int __split_and_process_non_flush(struct clone_info *ci) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1391 | { |
Mikulas Patocka | dba1416 | 2012-10-12 21:02:15 +0100 | [diff] [blame] | 1392 | struct bio *bio = ci->bio; |
Jun'ichi Nomura | 512875b | 2007-12-13 14:15:25 +0000 | [diff] [blame] | 1393 | struct dm_target *ti; |
Kent Overstreet | 1c3b13e | 2013-10-29 17:17:49 -0700 | [diff] [blame] | 1394 | unsigned len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1395 | |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1396 | if (unlikely(bio->bi_rw & REQ_DISCARD)) |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1397 | return __send_discard(ci); |
Mike Snitzer | 23508a9 | 2012-12-21 20:23:37 +0000 | [diff] [blame] | 1398 | else if (unlikely(bio->bi_rw & REQ_WRITE_SAME)) |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1399 | return __send_write_same(ci); |
Mike Snitzer | 5ae89a8 | 2010-08-12 04:14:08 +0100 | [diff] [blame] | 1400 | |
Jun'ichi Nomura | 512875b | 2007-12-13 14:15:25 +0000 | [diff] [blame] | 1401 | ti = dm_table_find_target(ci->map, ci->sector); |
| 1402 | if (!dm_target_is_valid(ti)) |
| 1403 | return -EIO; |
| 1404 | |
Kent Overstreet | 1c3b13e | 2013-10-29 17:17:49 -0700 | [diff] [blame] | 1405 | len = min_t(sector_t, max_io_len(ci->sector, ti), ci->sector_count); |
Jun'ichi Nomura | 512875b | 2007-12-13 14:15:25 +0000 | [diff] [blame] | 1406 | |
Mikulas Patocka | 1dd40c3 | 2014-03-14 18:41:24 -0400 | [diff] [blame] | 1407 | __clone_and_map_data_bio(ci, ti, ci->sector, &len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1408 | |
Kent Overstreet | 1c3b13e | 2013-10-29 17:17:49 -0700 | [diff] [blame] | 1409 | ci->sector += len; |
| 1410 | ci->sector_count -= len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1411 | |
Kent Overstreet | 1c3b13e | 2013-10-29 17:17:49 -0700 | [diff] [blame] | 1412 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1413 | } |
| 1414 | |
| 1415 | /* |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1416 | * Entry point to split a bio into clones and submit them to the targets. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1417 | */ |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1418 | static void __split_and_process_bio(struct mapped_device *md, |
| 1419 | struct dm_table *map, struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1420 | { |
| 1421 | struct clone_info ci; |
Jun'ichi Nomura | 512875b | 2007-12-13 14:15:25 +0000 | [diff] [blame] | 1422 | int error = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1423 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1424 | if (unlikely(!map)) { |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1425 | bio_io_error(bio); |
Mikulas Patocka | f0b9a45 | 2009-04-02 19:55:38 +0100 | [diff] [blame] | 1426 | return; |
| 1427 | } |
Mikulas Patocka | 692d0eb | 2009-04-09 00:27:13 +0100 | [diff] [blame] | 1428 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1429 | ci.map = map; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1430 | ci.md = md; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1431 | ci.io = alloc_io(md); |
| 1432 | ci.io->error = 0; |
| 1433 | atomic_set(&ci.io->io_count, 1); |
| 1434 | ci.io->bio = bio; |
| 1435 | ci.io->md = md; |
Kiyoshi Ueda | f88fb98 | 2009-10-16 23:18:15 +0100 | [diff] [blame] | 1436 | spin_lock_init(&ci.io->endio_lock); |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1437 | ci.sector = bio->bi_iter.bi_sector; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1438 | |
Jun'ichi "Nick" Nomura | 3eaf840 | 2006-02-01 03:04:53 -0800 | [diff] [blame] | 1439 | start_io_acct(ci.io); |
Alasdair G Kergon | bd2a49b | 2013-03-01 22:45:46 +0000 | [diff] [blame] | 1440 | |
Mike Snitzer | b372d36 | 2010-09-08 18:07:01 +0200 | [diff] [blame] | 1441 | if (bio->bi_rw & REQ_FLUSH) { |
| 1442 | ci.bio = &ci.md->flush_bio; |
| 1443 | ci.sector_count = 0; |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1444 | error = __send_empty_flush(&ci); |
Mike Snitzer | b372d36 | 2010-09-08 18:07:01 +0200 | [diff] [blame] | 1445 | /* dec_pending submits any data associated with flush */ |
| 1446 | } else { |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1447 | ci.bio = bio; |
Tejun Heo | d87f4c1 | 2010-09-03 11:56:19 +0200 | [diff] [blame] | 1448 | ci.sector_count = bio_sectors(bio); |
Mike Snitzer | b372d36 | 2010-09-08 18:07:01 +0200 | [diff] [blame] | 1449 | while (ci.sector_count && !error) |
Alasdair G Kergon | 14fe594 | 2013-03-01 22:45:47 +0000 | [diff] [blame] | 1450 | error = __split_and_process_non_flush(&ci); |
Tejun Heo | d87f4c1 | 2010-09-03 11:56:19 +0200 | [diff] [blame] | 1451 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1452 | |
| 1453 | /* drop the extra reference count */ |
Jun'ichi Nomura | 512875b | 2007-12-13 14:15:25 +0000 | [diff] [blame] | 1454 | dec_pending(ci.io, error); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1455 | } |
| 1456 | /*----------------------------------------------------------------- |
| 1457 | * CRUD END |
| 1458 | *---------------------------------------------------------------*/ |
| 1459 | |
Milan Broz | f6fccb1 | 2008-07-21 12:00:37 +0100 | [diff] [blame] | 1460 | static int dm_merge_bvec(struct request_queue *q, |
| 1461 | struct bvec_merge_data *bvm, |
| 1462 | struct bio_vec *biovec) |
| 1463 | { |
| 1464 | struct mapped_device *md = q->queuedata; |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1465 | struct dm_table *map = dm_get_live_table_fast(md); |
Milan Broz | f6fccb1 | 2008-07-21 12:00:37 +0100 | [diff] [blame] | 1466 | struct dm_target *ti; |
| 1467 | sector_t max_sectors; |
Mikulas Patocka | 5037108 | 2008-10-01 14:39:17 +0100 | [diff] [blame] | 1468 | int max_size = 0; |
Milan Broz | f6fccb1 | 2008-07-21 12:00:37 +0100 | [diff] [blame] | 1469 | |
| 1470 | if (unlikely(!map)) |
Mikulas Patocka | 5037108 | 2008-10-01 14:39:17 +0100 | [diff] [blame] | 1471 | goto out; |
Milan Broz | f6fccb1 | 2008-07-21 12:00:37 +0100 | [diff] [blame] | 1472 | |
| 1473 | ti = dm_table_find_target(map, bvm->bi_sector); |
Mikulas Patocka | b01cd5a | 2008-10-01 14:39:24 +0100 | [diff] [blame] | 1474 | if (!dm_target_is_valid(ti)) |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1475 | goto out; |
Milan Broz | f6fccb1 | 2008-07-21 12:00:37 +0100 | [diff] [blame] | 1476 | |
| 1477 | /* |
| 1478 | * Find maximum amount of I/O that won't need splitting |
| 1479 | */ |
Mike Snitzer | 56a67df | 2010-08-12 04:14:10 +0100 | [diff] [blame] | 1480 | max_sectors = min(max_io_len(bvm->bi_sector, ti), |
Milan Broz | f6fccb1 | 2008-07-21 12:00:37 +0100 | [diff] [blame] | 1481 | (sector_t) BIO_MAX_SECTORS); |
| 1482 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; |
| 1483 | if (max_size < 0) |
| 1484 | max_size = 0; |
| 1485 | |
| 1486 | /* |
| 1487 | * merge_bvec_fn() returns number of bytes |
| 1488 | * it can accept at this offset |
| 1489 | * max is precomputed maximal io size |
| 1490 | */ |
| 1491 | if (max_size && ti->type->merge) |
| 1492 | max_size = ti->type->merge(ti, bvm, biovec, max_size); |
Mikulas Patocka | 8cbeb67 | 2009-06-22 10:12:14 +0100 | [diff] [blame] | 1493 | /* |
| 1494 | * If the target doesn't support merge method and some of the devices |
| 1495 | * provided their merge_bvec method (we know this by looking at |
| 1496 | * queue_max_hw_sectors), then we can't allow bios with multiple vector |
| 1497 | * entries. So always set max_size to 0, and the code below allows |
| 1498 | * just one page. |
| 1499 | */ |
| 1500 | else if (queue_max_hw_sectors(q) <= PAGE_SIZE >> 9) |
Mikulas Patocka | 8cbeb67 | 2009-06-22 10:12:14 +0100 | [diff] [blame] | 1501 | max_size = 0; |
Milan Broz | f6fccb1 | 2008-07-21 12:00:37 +0100 | [diff] [blame] | 1502 | |
Mikulas Patocka | 5037108 | 2008-10-01 14:39:17 +0100 | [diff] [blame] | 1503 | out: |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1504 | dm_put_live_table_fast(md); |
Milan Broz | f6fccb1 | 2008-07-21 12:00:37 +0100 | [diff] [blame] | 1505 | /* |
| 1506 | * Always allow an entire first page |
| 1507 | */ |
| 1508 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) |
| 1509 | max_size = biovec->bv_len; |
| 1510 | |
Milan Broz | f6fccb1 | 2008-07-21 12:00:37 +0100 | [diff] [blame] | 1511 | return max_size; |
| 1512 | } |
| 1513 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1514 | /* |
| 1515 | * The request function that just remaps the bio built up by |
| 1516 | * dm_merge_bvec. |
| 1517 | */ |
Christoph Hellwig | 5a7bbad | 2011-09-12 12:12:01 +0200 | [diff] [blame] | 1518 | static void _dm_request(struct request_queue *q, struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1519 | { |
Kevin Corry | 12f03a4 | 2006-02-01 03:04:52 -0800 | [diff] [blame] | 1520 | int rw = bio_data_dir(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1521 | struct mapped_device *md = q->queuedata; |
Tejun Heo | c995905 | 2008-08-25 19:47:21 +0900 | [diff] [blame] | 1522 | int cpu; |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1523 | int srcu_idx; |
| 1524 | struct dm_table *map; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1525 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1526 | map = dm_get_live_table(md, &srcu_idx); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1527 | |
Tejun Heo | 074a7ac | 2008-08-25 19:56:14 +0900 | [diff] [blame] | 1528 | cpu = part_stat_lock(); |
| 1529 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); |
| 1530 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); |
| 1531 | part_stat_unlock(); |
Kevin Corry | 12f03a4 | 2006-02-01 03:04:52 -0800 | [diff] [blame] | 1532 | |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1533 | /* if we're suspended, we have to queue this io for later */ |
| 1534 | if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) { |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1535 | dm_put_live_table(md, srcu_idx); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1536 | |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1537 | if (bio_rw(bio) != READA) |
| 1538 | queue_io(md, bio); |
| 1539 | else |
Alasdair G Kergon | 54d9a1b | 2009-04-09 00:27:14 +0100 | [diff] [blame] | 1540 | bio_io_error(bio); |
Christoph Hellwig | 5a7bbad | 2011-09-12 12:12:01 +0200 | [diff] [blame] | 1541 | return; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1542 | } |
| 1543 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1544 | __split_and_process_bio(md, map, bio); |
| 1545 | dm_put_live_table(md, srcu_idx); |
Christoph Hellwig | 5a7bbad | 2011-09-12 12:12:01 +0200 | [diff] [blame] | 1546 | return; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1547 | } |
| 1548 | |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 1549 | int dm_request_based(struct mapped_device *md) |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1550 | { |
| 1551 | return blk_queue_stackable(md->queue); |
| 1552 | } |
| 1553 | |
Christoph Hellwig | 5a7bbad | 2011-09-12 12:12:01 +0200 | [diff] [blame] | 1554 | static void dm_request(struct request_queue *q, struct bio *bio) |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1555 | { |
| 1556 | struct mapped_device *md = q->queuedata; |
| 1557 | |
| 1558 | if (dm_request_based(md)) |
Christoph Hellwig | 5a7bbad | 2011-09-12 12:12:01 +0200 | [diff] [blame] | 1559 | blk_queue_bio(q, bio); |
| 1560 | else |
| 1561 | _dm_request(q, bio); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1562 | } |
| 1563 | |
| 1564 | void dm_dispatch_request(struct request *rq) |
| 1565 | { |
| 1566 | int r; |
| 1567 | |
| 1568 | if (blk_queue_io_stat(rq->q)) |
| 1569 | rq->cmd_flags |= REQ_IO_STAT; |
| 1570 | |
| 1571 | rq->start_time = jiffies; |
| 1572 | r = blk_insert_cloned_request(rq->q, rq); |
| 1573 | if (r) |
| 1574 | dm_complete_request(rq, r); |
| 1575 | } |
| 1576 | EXPORT_SYMBOL_GPL(dm_dispatch_request); |
| 1577 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1578 | static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig, |
| 1579 | void *data) |
| 1580 | { |
| 1581 | struct dm_rq_target_io *tio = data; |
Kent Overstreet | 9481874 | 2012-09-07 13:44:01 -0700 | [diff] [blame] | 1582 | struct dm_rq_clone_bio_info *info = |
| 1583 | container_of(bio, struct dm_rq_clone_bio_info, clone); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1584 | |
| 1585 | info->orig = bio_orig; |
| 1586 | info->tio = tio; |
| 1587 | bio->bi_end_io = end_clone_bio; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1588 | |
| 1589 | return 0; |
| 1590 | } |
| 1591 | |
| 1592 | static int setup_clone(struct request *clone, struct request *rq, |
| 1593 | struct dm_rq_target_io *tio) |
| 1594 | { |
Kiyoshi Ueda | d0bcb87 | 2009-12-10 23:52:18 +0000 | [diff] [blame] | 1595 | int r; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1596 | |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1597 | r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC, |
| 1598 | dm_rq_bio_constructor, tio); |
| 1599 | if (r) |
| 1600 | return r; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1601 | |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1602 | clone->cmd = rq->cmd; |
| 1603 | clone->cmd_len = rq->cmd_len; |
| 1604 | clone->sense = rq->sense; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1605 | clone->end_io = end_clone_request; |
| 1606 | clone->end_io_data = tio; |
| 1607 | |
| 1608 | return 0; |
| 1609 | } |
| 1610 | |
Kiyoshi Ueda | 6facdaf | 2009-12-10 23:52:15 +0000 | [diff] [blame] | 1611 | static struct request *clone_rq(struct request *rq, struct mapped_device *md, |
| 1612 | gfp_t gfp_mask) |
| 1613 | { |
| 1614 | struct request *clone; |
| 1615 | struct dm_rq_target_io *tio; |
| 1616 | |
| 1617 | tio = alloc_rq_tio(md, gfp_mask); |
| 1618 | if (!tio) |
| 1619 | return NULL; |
| 1620 | |
| 1621 | tio->md = md; |
| 1622 | tio->ti = NULL; |
| 1623 | tio->orig = rq; |
| 1624 | tio->error = 0; |
| 1625 | memset(&tio->info, 0, sizeof(tio->info)); |
| 1626 | |
| 1627 | clone = &tio->clone; |
| 1628 | if (setup_clone(clone, rq, tio)) { |
| 1629 | /* -ENOMEM */ |
| 1630 | free_rq_tio(tio); |
| 1631 | return NULL; |
| 1632 | } |
| 1633 | |
| 1634 | return clone; |
| 1635 | } |
| 1636 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1637 | /* |
| 1638 | * Called with the queue lock held. |
| 1639 | */ |
| 1640 | static int dm_prep_fn(struct request_queue *q, struct request *rq) |
| 1641 | { |
| 1642 | struct mapped_device *md = q->queuedata; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1643 | struct request *clone; |
| 1644 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1645 | if (unlikely(rq->special)) { |
| 1646 | DMWARN("Already has something in rq->special."); |
| 1647 | return BLKPREP_KILL; |
| 1648 | } |
| 1649 | |
Kiyoshi Ueda | 6facdaf | 2009-12-10 23:52:15 +0000 | [diff] [blame] | 1650 | clone = clone_rq(rq, md, GFP_ATOMIC); |
| 1651 | if (!clone) |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1652 | return BLKPREP_DEFER; |
| 1653 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1654 | rq->special = clone; |
| 1655 | rq->cmd_flags |= REQ_DONTPREP; |
| 1656 | |
| 1657 | return BLKPREP_OK; |
| 1658 | } |
| 1659 | |
Kiyoshi Ueda | 9eef87d | 2010-02-16 18:43:01 +0000 | [diff] [blame] | 1660 | /* |
| 1661 | * Returns: |
| 1662 | * 0 : the request has been processed (not requeued) |
| 1663 | * !0 : the request has been requeued |
| 1664 | */ |
| 1665 | static int map_request(struct dm_target *ti, struct request *clone, |
| 1666 | struct mapped_device *md) |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1667 | { |
Kiyoshi Ueda | 9eef87d | 2010-02-16 18:43:01 +0000 | [diff] [blame] | 1668 | int r, requeued = 0; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1669 | struct dm_rq_target_io *tio = clone->end_io_data; |
| 1670 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1671 | tio->ti = ti; |
| 1672 | r = ti->type->map_rq(ti, clone, &tio->info); |
| 1673 | switch (r) { |
| 1674 | case DM_MAPIO_SUBMITTED: |
| 1675 | /* The target has taken the I/O to submit by itself later */ |
| 1676 | break; |
| 1677 | case DM_MAPIO_REMAPPED: |
| 1678 | /* The target has remapped the I/O so dispatch it */ |
Jun'ichi Nomura | 6db4ccd | 2009-12-10 23:52:25 +0000 | [diff] [blame] | 1679 | trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)), |
| 1680 | blk_rq_pos(tio->orig)); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1681 | dm_dispatch_request(clone); |
| 1682 | break; |
| 1683 | case DM_MAPIO_REQUEUE: |
| 1684 | /* The target wants to requeue the I/O */ |
| 1685 | dm_requeue_unmapped_request(clone); |
Kiyoshi Ueda | 9eef87d | 2010-02-16 18:43:01 +0000 | [diff] [blame] | 1686 | requeued = 1; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1687 | break; |
| 1688 | default: |
| 1689 | if (r > 0) { |
| 1690 | DMWARN("unimplemented target map return value: %d", r); |
| 1691 | BUG(); |
| 1692 | } |
| 1693 | |
| 1694 | /* The target wants to complete the I/O */ |
| 1695 | dm_kill_unmapped_request(clone, r); |
| 1696 | break; |
| 1697 | } |
Kiyoshi Ueda | 9eef87d | 2010-02-16 18:43:01 +0000 | [diff] [blame] | 1698 | |
| 1699 | return requeued; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1700 | } |
| 1701 | |
Mike Snitzer | ba1cbad | 2012-09-26 23:45:42 +0100 | [diff] [blame] | 1702 | static struct request *dm_start_request(struct mapped_device *md, struct request *orig) |
| 1703 | { |
| 1704 | struct request *clone; |
| 1705 | |
| 1706 | blk_start_request(orig); |
| 1707 | clone = orig->special; |
| 1708 | atomic_inc(&md->pending[rq_data_dir(clone)]); |
| 1709 | |
| 1710 | /* |
| 1711 | * Hold the md reference here for the in-flight I/O. |
| 1712 | * We can't rely on the reference count by device opener, |
| 1713 | * because the device may be closed during the request completion |
| 1714 | * when all bios are completed. |
| 1715 | * See the comment in rq_completed() too. |
| 1716 | */ |
| 1717 | dm_get(md); |
| 1718 | |
| 1719 | return clone; |
| 1720 | } |
| 1721 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1722 | /* |
| 1723 | * q->request_fn for request-based dm. |
| 1724 | * Called with the queue lock held. |
| 1725 | */ |
| 1726 | static void dm_request_fn(struct request_queue *q) |
| 1727 | { |
| 1728 | struct mapped_device *md = q->queuedata; |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1729 | int srcu_idx; |
| 1730 | struct dm_table *map = dm_get_live_table(md, &srcu_idx); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1731 | struct dm_target *ti; |
Kiyoshi Ueda | b4324fe | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 1732 | struct request *rq, *clone; |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1733 | sector_t pos; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1734 | |
| 1735 | /* |
Kiyoshi Ueda | b4324fe | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 1736 | * For suspend, check blk_queue_stopped() and increment |
| 1737 | * ->pending within a single queue_lock not to increment the |
| 1738 | * number of in-flight I/Os after the queue is stopped in |
| 1739 | * dm_suspend(). |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1740 | */ |
Jens Axboe | 7eaceac | 2011-03-10 08:52:07 +0100 | [diff] [blame] | 1741 | while (!blk_queue_stopped(q)) { |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1742 | rq = blk_peek_request(q); |
| 1743 | if (!rq) |
Jens Axboe | 7eaceac | 2011-03-10 08:52:07 +0100 | [diff] [blame] | 1744 | goto delay_and_out; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1745 | |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1746 | /* always use block 0 to find the target for flushes for now */ |
| 1747 | pos = 0; |
| 1748 | if (!(rq->cmd_flags & REQ_FLUSH)) |
| 1749 | pos = blk_rq_pos(rq); |
Kiyoshi Ueda | d0bcb87 | 2009-12-10 23:52:18 +0000 | [diff] [blame] | 1750 | |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1751 | ti = dm_table_find_target(map, pos); |
Mike Snitzer | ba1cbad | 2012-09-26 23:45:42 +0100 | [diff] [blame] | 1752 | if (!dm_target_is_valid(ti)) { |
| 1753 | /* |
| 1754 | * Must perform setup, that dm_done() requires, |
| 1755 | * before calling dm_kill_unmapped_request |
| 1756 | */ |
| 1757 | DMERR_LIMIT("request attempted access beyond the end of device"); |
| 1758 | clone = dm_start_request(md, rq); |
| 1759 | dm_kill_unmapped_request(clone, -EIO); |
| 1760 | continue; |
| 1761 | } |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1762 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1763 | if (ti->type->busy && ti->type->busy(ti)) |
Jens Axboe | 7eaceac | 2011-03-10 08:52:07 +0100 | [diff] [blame] | 1764 | goto delay_and_out; |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1765 | |
Mike Snitzer | ba1cbad | 2012-09-26 23:45:42 +0100 | [diff] [blame] | 1766 | clone = dm_start_request(md, rq); |
Kiyoshi Ueda | b4324fe | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 1767 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1768 | spin_unlock(q->queue_lock); |
Kiyoshi Ueda | 9eef87d | 2010-02-16 18:43:01 +0000 | [diff] [blame] | 1769 | if (map_request(ti, clone, md)) |
| 1770 | goto requeued; |
| 1771 | |
Kiyoshi Ueda | 052189a | 2011-01-13 20:00:00 +0000 | [diff] [blame] | 1772 | BUG_ON(!irqs_disabled()); |
| 1773 | spin_lock(q->queue_lock); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1774 | } |
| 1775 | |
| 1776 | goto out; |
| 1777 | |
Kiyoshi Ueda | 9eef87d | 2010-02-16 18:43:01 +0000 | [diff] [blame] | 1778 | requeued: |
Kiyoshi Ueda | 052189a | 2011-01-13 20:00:00 +0000 | [diff] [blame] | 1779 | BUG_ON(!irqs_disabled()); |
| 1780 | spin_lock(q->queue_lock); |
Kiyoshi Ueda | 9eef87d | 2010-02-16 18:43:01 +0000 | [diff] [blame] | 1781 | |
Jens Axboe | 7eaceac | 2011-03-10 08:52:07 +0100 | [diff] [blame] | 1782 | delay_and_out: |
| 1783 | blk_delay_queue(q, HZ / 10); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1784 | out: |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1785 | dm_put_live_table(md, srcu_idx); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1786 | } |
| 1787 | |
| 1788 | int dm_underlying_device_busy(struct request_queue *q) |
| 1789 | { |
| 1790 | return blk_lld_busy(q); |
| 1791 | } |
| 1792 | EXPORT_SYMBOL_GPL(dm_underlying_device_busy); |
| 1793 | |
| 1794 | static int dm_lld_busy(struct request_queue *q) |
| 1795 | { |
| 1796 | int r; |
| 1797 | struct mapped_device *md = q->queuedata; |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1798 | struct dm_table *map = dm_get_live_table_fast(md); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1799 | |
| 1800 | if (!map || test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) |
| 1801 | r = 1; |
| 1802 | else |
| 1803 | r = dm_table_any_busy_target(map); |
| 1804 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1805 | dm_put_live_table_fast(md); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1806 | |
| 1807 | return r; |
| 1808 | } |
| 1809 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1810 | static int dm_any_congested(void *congested_data, int bdi_bits) |
| 1811 | { |
Chandra Seetharaman | 8a57dfc | 2008-11-13 23:39:14 +0000 | [diff] [blame] | 1812 | int r = bdi_bits; |
| 1813 | struct mapped_device *md = congested_data; |
| 1814 | struct dm_table *map; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1815 | |
Alasdair G Kergon | 1eb787e | 2009-04-09 00:27:14 +0100 | [diff] [blame] | 1816 | if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1817 | map = dm_get_live_table_fast(md); |
Chandra Seetharaman | 8a57dfc | 2008-11-13 23:39:14 +0000 | [diff] [blame] | 1818 | if (map) { |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 1819 | /* |
| 1820 | * Request-based dm cares about only own queue for |
| 1821 | * the query about congestion status of request_queue |
| 1822 | */ |
| 1823 | if (dm_request_based(md)) |
| 1824 | r = md->queue->backing_dev_info.state & |
| 1825 | bdi_bits; |
| 1826 | else |
| 1827 | r = dm_table_any_congested(map, bdi_bits); |
Chandra Seetharaman | 8a57dfc | 2008-11-13 23:39:14 +0000 | [diff] [blame] | 1828 | } |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1829 | dm_put_live_table_fast(md); |
Chandra Seetharaman | 8a57dfc | 2008-11-13 23:39:14 +0000 | [diff] [blame] | 1830 | } |
| 1831 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1832 | return r; |
| 1833 | } |
| 1834 | |
| 1835 | /*----------------------------------------------------------------- |
| 1836 | * An IDR is used to keep track of allocated minor numbers. |
| 1837 | *---------------------------------------------------------------*/ |
Alasdair G Kergon | 2b06cff | 2006-06-26 00:27:32 -0700 | [diff] [blame] | 1838 | static void free_minor(int minor) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1839 | { |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 1840 | spin_lock(&_minor_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1841 | idr_remove(&_minor_idr, minor); |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 1842 | spin_unlock(&_minor_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1843 | } |
| 1844 | |
| 1845 | /* |
| 1846 | * See if the device with a specific minor # is free. |
| 1847 | */ |
Frederik Deweerdt | cf13ab8 | 2008-04-24 22:10:59 +0100 | [diff] [blame] | 1848 | static int specific_minor(int minor) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1849 | { |
Tejun Heo | c9d76be | 2013-02-27 17:04:26 -0800 | [diff] [blame] | 1850 | int r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1851 | |
| 1852 | if (minor >= (1 << MINORBITS)) |
| 1853 | return -EINVAL; |
| 1854 | |
Tejun Heo | c9d76be | 2013-02-27 17:04:26 -0800 | [diff] [blame] | 1855 | idr_preload(GFP_KERNEL); |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 1856 | spin_lock(&_minor_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1857 | |
Tejun Heo | c9d76be | 2013-02-27 17:04:26 -0800 | [diff] [blame] | 1858 | r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1859 | |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 1860 | spin_unlock(&_minor_lock); |
Tejun Heo | c9d76be | 2013-02-27 17:04:26 -0800 | [diff] [blame] | 1861 | idr_preload_end(); |
| 1862 | if (r < 0) |
| 1863 | return r == -ENOSPC ? -EBUSY : r; |
| 1864 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1865 | } |
| 1866 | |
Frederik Deweerdt | cf13ab8 | 2008-04-24 22:10:59 +0100 | [diff] [blame] | 1867 | static int next_free_minor(int *minor) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1868 | { |
Tejun Heo | c9d76be | 2013-02-27 17:04:26 -0800 | [diff] [blame] | 1869 | int r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1870 | |
Tejun Heo | c9d76be | 2013-02-27 17:04:26 -0800 | [diff] [blame] | 1871 | idr_preload(GFP_KERNEL); |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 1872 | spin_lock(&_minor_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1873 | |
Tejun Heo | c9d76be | 2013-02-27 17:04:26 -0800 | [diff] [blame] | 1874 | r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1875 | |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 1876 | spin_unlock(&_minor_lock); |
Tejun Heo | c9d76be | 2013-02-27 17:04:26 -0800 | [diff] [blame] | 1877 | idr_preload_end(); |
| 1878 | if (r < 0) |
| 1879 | return r; |
| 1880 | *minor = r; |
| 1881 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1882 | } |
| 1883 | |
Alexey Dobriyan | 83d5cde | 2009-09-21 17:01:13 -0700 | [diff] [blame] | 1884 | static const struct block_device_operations dm_blk_dops; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1885 | |
Mikulas Patocka | 53d5914 | 2009-04-02 19:55:37 +0100 | [diff] [blame] | 1886 | static void dm_wq_work(struct work_struct *work); |
| 1887 | |
Mike Snitzer | 4a0b4dd | 2010-08-12 04:14:02 +0100 | [diff] [blame] | 1888 | static void dm_init_md_queue(struct mapped_device *md) |
| 1889 | { |
| 1890 | /* |
| 1891 | * Request-based dm devices cannot be stacked on top of bio-based dm |
| 1892 | * devices. The type of this dm device has not been decided yet. |
| 1893 | * The type is decided at the first table loading time. |
| 1894 | * To prevent problematic device stacking, clear the queue flag |
| 1895 | * for request stacking support until then. |
| 1896 | * |
| 1897 | * This queue is new, so no concurrency on the queue_flags. |
| 1898 | */ |
| 1899 | queue_flag_clear_unlocked(QUEUE_FLAG_STACKABLE, md->queue); |
| 1900 | |
| 1901 | md->queue->queuedata = md; |
| 1902 | md->queue->backing_dev_info.congested_fn = dm_any_congested; |
| 1903 | md->queue->backing_dev_info.congested_data = md; |
| 1904 | blk_queue_make_request(md->queue, dm_request); |
| 1905 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); |
Mike Snitzer | 4a0b4dd | 2010-08-12 04:14:02 +0100 | [diff] [blame] | 1906 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
| 1907 | } |
| 1908 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1909 | /* |
| 1910 | * Allocate and initialise a blank device with a given minor. |
| 1911 | */ |
Alasdair G Kergon | 2b06cff | 2006-06-26 00:27:32 -0700 | [diff] [blame] | 1912 | static struct mapped_device *alloc_dev(int minor) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1913 | { |
| 1914 | int r; |
Frederik Deweerdt | cf13ab8 | 2008-04-24 22:10:59 +0100 | [diff] [blame] | 1915 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
Jeff Mahoney | ba61fdd | 2006-06-26 00:27:21 -0700 | [diff] [blame] | 1916 | void *old_md; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1917 | |
| 1918 | if (!md) { |
| 1919 | DMWARN("unable to allocate device, out of memory."); |
| 1920 | return NULL; |
| 1921 | } |
| 1922 | |
Jeff Mahoney | 10da4f7 | 2006-06-26 00:27:25 -0700 | [diff] [blame] | 1923 | if (!try_module_get(THIS_MODULE)) |
Milan Broz | 6ed7ade | 2008-02-08 02:10:19 +0000 | [diff] [blame] | 1924 | goto bad_module_get; |
Jeff Mahoney | 10da4f7 | 2006-06-26 00:27:25 -0700 | [diff] [blame] | 1925 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1926 | /* get a minor number for the dev */ |
Alasdair G Kergon | 2b06cff | 2006-06-26 00:27:32 -0700 | [diff] [blame] | 1927 | if (minor == DM_ANY_MINOR) |
Frederik Deweerdt | cf13ab8 | 2008-04-24 22:10:59 +0100 | [diff] [blame] | 1928 | r = next_free_minor(&minor); |
Alasdair G Kergon | 2b06cff | 2006-06-26 00:27:32 -0700 | [diff] [blame] | 1929 | else |
Frederik Deweerdt | cf13ab8 | 2008-04-24 22:10:59 +0100 | [diff] [blame] | 1930 | r = specific_minor(minor); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1931 | if (r < 0) |
Milan Broz | 6ed7ade | 2008-02-08 02:10:19 +0000 | [diff] [blame] | 1932 | goto bad_minor; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1933 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 1934 | r = init_srcu_struct(&md->io_barrier); |
| 1935 | if (r < 0) |
| 1936 | goto bad_io_barrier; |
| 1937 | |
Mike Snitzer | a5664da | 2010-08-12 04:14:01 +0100 | [diff] [blame] | 1938 | md->type = DM_TYPE_NONE; |
Daniel Walker | e61290a | 2008-02-08 02:10:08 +0000 | [diff] [blame] | 1939 | mutex_init(&md->suspend_lock); |
Mike Snitzer | a5664da | 2010-08-12 04:14:01 +0100 | [diff] [blame] | 1940 | mutex_init(&md->type_lock); |
Mikulas Patocka | 022c261 | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 1941 | spin_lock_init(&md->deferred_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1942 | atomic_set(&md->holders, 1); |
Alasdair G Kergon | 5c6bd75 | 2006-06-26 00:27:34 -0700 | [diff] [blame] | 1943 | atomic_set(&md->open_count, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1944 | atomic_set(&md->event_nr, 0); |
Mike Anderson | 7a8c3d3 | 2007-10-19 22:48:01 +0100 | [diff] [blame] | 1945 | atomic_set(&md->uevent_seq, 0); |
| 1946 | INIT_LIST_HEAD(&md->uevent_list); |
| 1947 | spin_lock_init(&md->uevent_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1948 | |
Mike Snitzer | 4a0b4dd | 2010-08-12 04:14:02 +0100 | [diff] [blame] | 1949 | md->queue = blk_alloc_queue(GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1950 | if (!md->queue) |
Milan Broz | 6ed7ade | 2008-02-08 02:10:19 +0000 | [diff] [blame] | 1951 | goto bad_queue; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1952 | |
Mike Snitzer | 4a0b4dd | 2010-08-12 04:14:02 +0100 | [diff] [blame] | 1953 | dm_init_md_queue(md); |
Stefan Bader | 9faf400 | 2006-10-03 01:15:41 -0700 | [diff] [blame] | 1954 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1955 | md->disk = alloc_disk(1); |
| 1956 | if (!md->disk) |
Milan Broz | 6ed7ade | 2008-02-08 02:10:19 +0000 | [diff] [blame] | 1957 | goto bad_disk; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1958 | |
Nikanth Karthikesan | 316d315 | 2009-10-06 20:16:55 +0200 | [diff] [blame] | 1959 | atomic_set(&md->pending[0], 0); |
| 1960 | atomic_set(&md->pending[1], 0); |
Jeff Mahoney | f0b0411 | 2006-06-26 00:27:25 -0700 | [diff] [blame] | 1961 | init_waitqueue_head(&md->wait); |
Mikulas Patocka | 53d5914 | 2009-04-02 19:55:37 +0100 | [diff] [blame] | 1962 | INIT_WORK(&md->work, dm_wq_work); |
Jeff Mahoney | f0b0411 | 2006-06-26 00:27:25 -0700 | [diff] [blame] | 1963 | init_waitqueue_head(&md->eventq); |
Mikulas Patocka | 2995fa7 | 2014-01-13 19:37:54 -0500 | [diff] [blame] | 1964 | init_completion(&md->kobj_holder.completion); |
Jeff Mahoney | f0b0411 | 2006-06-26 00:27:25 -0700 | [diff] [blame] | 1965 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1966 | md->disk->major = _major; |
| 1967 | md->disk->first_minor = minor; |
| 1968 | md->disk->fops = &dm_blk_dops; |
| 1969 | md->disk->queue = md->queue; |
| 1970 | md->disk->private_data = md; |
| 1971 | sprintf(md->disk->disk_name, "dm-%d", minor); |
| 1972 | add_disk(md->disk); |
Mike Anderson | 7e51f25 | 2006-03-27 01:17:52 -0800 | [diff] [blame] | 1973 | format_dev_t(md->name, MKDEV(_major, minor)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1974 | |
Tejun Heo | 670368a | 2013-07-30 08:40:21 -0400 | [diff] [blame] | 1975 | md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0); |
Milan Broz | 304f3f6 | 2008-02-08 02:11:17 +0000 | [diff] [blame] | 1976 | if (!md->wq) |
| 1977 | goto bad_thread; |
| 1978 | |
Mikulas Patocka | 32a926d | 2009-06-22 10:12:17 +0100 | [diff] [blame] | 1979 | md->bdev = bdget_disk(md->disk, 0); |
| 1980 | if (!md->bdev) |
| 1981 | goto bad_bdev; |
| 1982 | |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 1983 | bio_init(&md->flush_bio); |
| 1984 | md->flush_bio.bi_bdev = md->bdev; |
| 1985 | md->flush_bio.bi_rw = WRITE_FLUSH; |
| 1986 | |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 1987 | dm_stats_init(&md->stats); |
| 1988 | |
Jeff Mahoney | ba61fdd | 2006-06-26 00:27:21 -0700 | [diff] [blame] | 1989 | /* Populate the mapping, nobody knows we exist yet */ |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 1990 | spin_lock(&_minor_lock); |
Jeff Mahoney | ba61fdd | 2006-06-26 00:27:21 -0700 | [diff] [blame] | 1991 | old_md = idr_replace(&_minor_idr, md, minor); |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 1992 | spin_unlock(&_minor_lock); |
Jeff Mahoney | ba61fdd | 2006-06-26 00:27:21 -0700 | [diff] [blame] | 1993 | |
| 1994 | BUG_ON(old_md != MINOR_ALLOCED); |
| 1995 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1996 | return md; |
| 1997 | |
Mikulas Patocka | 32a926d | 2009-06-22 10:12:17 +0100 | [diff] [blame] | 1998 | bad_bdev: |
| 1999 | destroy_workqueue(md->wq); |
Milan Broz | 304f3f6 | 2008-02-08 02:11:17 +0000 | [diff] [blame] | 2000 | bad_thread: |
Zdenek Kabelac | 03022c5 | 2009-10-16 23:18:15 +0100 | [diff] [blame] | 2001 | del_gendisk(md->disk); |
Milan Broz | 304f3f6 | 2008-02-08 02:11:17 +0000 | [diff] [blame] | 2002 | put_disk(md->disk); |
Milan Broz | 6ed7ade | 2008-02-08 02:10:19 +0000 | [diff] [blame] | 2003 | bad_disk: |
Al Viro | 1312f40 | 2006-03-12 11:02:03 -0500 | [diff] [blame] | 2004 | blk_cleanup_queue(md->queue); |
Milan Broz | 6ed7ade | 2008-02-08 02:10:19 +0000 | [diff] [blame] | 2005 | bad_queue: |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2006 | cleanup_srcu_struct(&md->io_barrier); |
| 2007 | bad_io_barrier: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2008 | free_minor(minor); |
Milan Broz | 6ed7ade | 2008-02-08 02:10:19 +0000 | [diff] [blame] | 2009 | bad_minor: |
Jeff Mahoney | 10da4f7 | 2006-06-26 00:27:25 -0700 | [diff] [blame] | 2010 | module_put(THIS_MODULE); |
Milan Broz | 6ed7ade | 2008-02-08 02:10:19 +0000 | [diff] [blame] | 2011 | bad_module_get: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2012 | kfree(md); |
| 2013 | return NULL; |
| 2014 | } |
| 2015 | |
Jun'ichi Nomura | ae9da83 | 2007-10-19 22:38:43 +0100 | [diff] [blame] | 2016 | static void unlock_fs(struct mapped_device *md); |
| 2017 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2018 | static void free_dev(struct mapped_device *md) |
| 2019 | { |
Tejun Heo | f331c02 | 2008-09-03 09:01:48 +0200 | [diff] [blame] | 2020 | int minor = MINOR(disk_devt(md->disk)); |
Jun'ichi Nomura | 63d94e4 | 2006-02-24 13:04:25 -0800 | [diff] [blame] | 2021 | |
Mikulas Patocka | 32a926d | 2009-06-22 10:12:17 +0100 | [diff] [blame] | 2022 | unlock_fs(md); |
| 2023 | bdput(md->bdev); |
Milan Broz | 304f3f6 | 2008-02-08 02:11:17 +0000 | [diff] [blame] | 2024 | destroy_workqueue(md->wq); |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2025 | if (md->io_pool) |
| 2026 | mempool_destroy(md->io_pool); |
| 2027 | if (md->bs) |
| 2028 | bioset_free(md->bs); |
Martin K. Petersen | 9c47008 | 2009-04-09 00:27:12 +0100 | [diff] [blame] | 2029 | blk_integrity_unregister(md->disk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2030 | del_gendisk(md->disk); |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2031 | cleanup_srcu_struct(&md->io_barrier); |
Jun'ichi Nomura | 63d94e4 | 2006-02-24 13:04:25 -0800 | [diff] [blame] | 2032 | free_minor(minor); |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 2033 | |
| 2034 | spin_lock(&_minor_lock); |
| 2035 | md->disk->private_data = NULL; |
| 2036 | spin_unlock(&_minor_lock); |
| 2037 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2038 | put_disk(md->disk); |
Al Viro | 1312f40 | 2006-03-12 11:02:03 -0500 | [diff] [blame] | 2039 | blk_cleanup_queue(md->queue); |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 2040 | dm_stats_cleanup(&md->stats); |
Jeff Mahoney | 10da4f7 | 2006-06-26 00:27:25 -0700 | [diff] [blame] | 2041 | module_put(THIS_MODULE); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2042 | kfree(md); |
| 2043 | } |
| 2044 | |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2045 | static void __bind_mempools(struct mapped_device *md, struct dm_table *t) |
| 2046 | { |
Mikulas Patocka | c0820cf | 2012-12-21 20:23:38 +0000 | [diff] [blame] | 2047 | struct dm_md_mempools *p = dm_table_get_md_mempools(t); |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2048 | |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2049 | if (md->io_pool && md->bs) { |
Jun'ichi Nomura | 16245bd | 2013-03-01 22:45:44 +0000 | [diff] [blame] | 2050 | /* The md already has necessary mempools. */ |
| 2051 | if (dm_table_get_type(t) == DM_TYPE_BIO_BASED) { |
| 2052 | /* |
| 2053 | * Reload bioset because front_pad may have changed |
| 2054 | * because a different table was loaded. |
| 2055 | */ |
| 2056 | bioset_free(md->bs); |
| 2057 | md->bs = p->bs; |
| 2058 | p->bs = NULL; |
| 2059 | } else if (dm_table_get_type(t) == DM_TYPE_REQUEST_BASED) { |
Jun'ichi Nomura | 16245bd | 2013-03-01 22:45:44 +0000 | [diff] [blame] | 2060 | /* |
| 2061 | * There's no need to reload with request-based dm |
| 2062 | * because the size of front_pad doesn't change. |
| 2063 | * Note for future: If you are to reload bioset, |
| 2064 | * prep-ed requests in the queue may refer |
| 2065 | * to bio from the old bioset, so you must walk |
| 2066 | * through the queue to unprep. |
| 2067 | */ |
| 2068 | } |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2069 | goto out; |
Mikulas Patocka | c0820cf | 2012-12-21 20:23:38 +0000 | [diff] [blame] | 2070 | } |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2071 | |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2072 | BUG_ON(!p || md->io_pool || md->bs); |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2073 | |
| 2074 | md->io_pool = p->io_pool; |
| 2075 | p->io_pool = NULL; |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2076 | md->bs = p->bs; |
| 2077 | p->bs = NULL; |
| 2078 | |
| 2079 | out: |
| 2080 | /* mempool bind completed, now no need any mempools in the table */ |
| 2081 | dm_table_free_md_mempools(t); |
| 2082 | } |
| 2083 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2084 | /* |
| 2085 | * Bind a table to the device. |
| 2086 | */ |
| 2087 | static void event_callback(void *context) |
| 2088 | { |
Mike Anderson | 7a8c3d3 | 2007-10-19 22:48:01 +0100 | [diff] [blame] | 2089 | unsigned long flags; |
| 2090 | LIST_HEAD(uevents); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2091 | struct mapped_device *md = (struct mapped_device *) context; |
| 2092 | |
Mike Anderson | 7a8c3d3 | 2007-10-19 22:48:01 +0100 | [diff] [blame] | 2093 | spin_lock_irqsave(&md->uevent_lock, flags); |
| 2094 | list_splice_init(&md->uevent_list, &uevents); |
| 2095 | spin_unlock_irqrestore(&md->uevent_lock, flags); |
| 2096 | |
Tejun Heo | ed9e198 | 2008-08-25 19:56:05 +0900 | [diff] [blame] | 2097 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
Mike Anderson | 7a8c3d3 | 2007-10-19 22:48:01 +0100 | [diff] [blame] | 2098 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2099 | atomic_inc(&md->event_nr); |
| 2100 | wake_up(&md->eventq); |
| 2101 | } |
| 2102 | |
Mike Snitzer | c217649 | 2011-01-13 19:53:46 +0000 | [diff] [blame] | 2103 | /* |
| 2104 | * Protected by md->suspend_lock obtained by dm_swap_table(). |
| 2105 | */ |
Alasdair G Kergon | 4e90188be | 2005-07-28 21:15:59 -0700 | [diff] [blame] | 2106 | static void __set_size(struct mapped_device *md, sector_t size) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2107 | { |
Alasdair G Kergon | 4e90188be | 2005-07-28 21:15:59 -0700 | [diff] [blame] | 2108 | set_capacity(md->disk, size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2109 | |
Mikulas Patocka | db8fef4 | 2009-06-22 10:12:15 +0100 | [diff] [blame] | 2110 | i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2111 | } |
| 2112 | |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2113 | /* |
Mikulas Patocka | d5b9dd0 | 2011-08-02 12:32:04 +0100 | [diff] [blame] | 2114 | * Return 1 if the queue has a compulsory merge_bvec_fn function. |
| 2115 | * |
| 2116 | * If this function returns 0, then the device is either a non-dm |
| 2117 | * device without a merge_bvec_fn, or it is a dm device that is |
| 2118 | * able to split any bios it receives that are too big. |
| 2119 | */ |
| 2120 | int dm_queue_merge_is_compulsory(struct request_queue *q) |
| 2121 | { |
| 2122 | struct mapped_device *dev_md; |
| 2123 | |
| 2124 | if (!q->merge_bvec_fn) |
| 2125 | return 0; |
| 2126 | |
| 2127 | if (q->make_request_fn == dm_request) { |
| 2128 | dev_md = q->queuedata; |
| 2129 | if (test_bit(DMF_MERGE_IS_OPTIONAL, &dev_md->flags)) |
| 2130 | return 0; |
| 2131 | } |
| 2132 | |
| 2133 | return 1; |
| 2134 | } |
| 2135 | |
| 2136 | static int dm_device_merge_is_compulsory(struct dm_target *ti, |
| 2137 | struct dm_dev *dev, sector_t start, |
| 2138 | sector_t len, void *data) |
| 2139 | { |
| 2140 | struct block_device *bdev = dev->bdev; |
| 2141 | struct request_queue *q = bdev_get_queue(bdev); |
| 2142 | |
| 2143 | return dm_queue_merge_is_compulsory(q); |
| 2144 | } |
| 2145 | |
| 2146 | /* |
| 2147 | * Return 1 if it is acceptable to ignore merge_bvec_fn based |
| 2148 | * on the properties of the underlying devices. |
| 2149 | */ |
| 2150 | static int dm_table_merge_is_optional(struct dm_table *table) |
| 2151 | { |
| 2152 | unsigned i = 0; |
| 2153 | struct dm_target *ti; |
| 2154 | |
| 2155 | while (i < dm_table_get_num_targets(table)) { |
| 2156 | ti = dm_table_get_target(table, i++); |
| 2157 | |
| 2158 | if (ti->type->iterate_devices && |
| 2159 | ti->type->iterate_devices(ti, dm_device_merge_is_compulsory, NULL)) |
| 2160 | return 0; |
| 2161 | } |
| 2162 | |
| 2163 | return 1; |
| 2164 | } |
| 2165 | |
| 2166 | /* |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2167 | * Returns old map, which caller must destroy. |
| 2168 | */ |
| 2169 | static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t, |
| 2170 | struct queue_limits *limits) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2171 | { |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2172 | struct dm_table *old_map; |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 2173 | struct request_queue *q = md->queue; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2174 | sector_t size; |
Mikulas Patocka | d5b9dd0 | 2011-08-02 12:32:04 +0100 | [diff] [blame] | 2175 | int merge_is_optional; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2176 | |
| 2177 | size = dm_table_get_size(t); |
Darrick J. Wong | 3ac51e7 | 2006-03-27 01:17:54 -0800 | [diff] [blame] | 2178 | |
| 2179 | /* |
| 2180 | * Wipe any geometry if the size of the table changed. |
| 2181 | */ |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 2182 | if (size != dm_get_size(md)) |
Darrick J. Wong | 3ac51e7 | 2006-03-27 01:17:54 -0800 | [diff] [blame] | 2183 | memset(&md->geometry, 0, sizeof(md->geometry)); |
| 2184 | |
Mikulas Patocka | 32a926d | 2009-06-22 10:12:17 +0100 | [diff] [blame] | 2185 | __set_size(md, size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2186 | |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2187 | dm_table_event_callback(t, event_callback, md); |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2188 | |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2189 | /* |
| 2190 | * The queue hasn't been stopped yet, if the old table type wasn't |
| 2191 | * for request-based during suspension. So stop it to prevent |
| 2192 | * I/O mapping before resume. |
| 2193 | * This must be done before setting the queue restrictions, |
| 2194 | * because request-based dm may be run just after the setting. |
| 2195 | */ |
| 2196 | if (dm_table_request_based(t) && !blk_queue_stopped(q)) |
| 2197 | stop_queue(q); |
| 2198 | |
| 2199 | __bind_mempools(md, t); |
| 2200 | |
Mikulas Patocka | d5b9dd0 | 2011-08-02 12:32:04 +0100 | [diff] [blame] | 2201 | merge_is_optional = dm_table_merge_is_optional(t); |
| 2202 | |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2203 | old_map = md->map; |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2204 | rcu_assign_pointer(md->map, t); |
Alasdair G Kergon | 36a0456 | 2011-10-31 20:19:04 +0000 | [diff] [blame] | 2205 | md->immutable_target_type = dm_table_get_immutable_target_type(t); |
| 2206 | |
Mike Snitzer | 754c5fc | 2009-06-22 10:12:34 +0100 | [diff] [blame] | 2207 | dm_table_set_restrictions(t, q, limits); |
Mikulas Patocka | d5b9dd0 | 2011-08-02 12:32:04 +0100 | [diff] [blame] | 2208 | if (merge_is_optional) |
| 2209 | set_bit(DMF_MERGE_IS_OPTIONAL, &md->flags); |
| 2210 | else |
| 2211 | clear_bit(DMF_MERGE_IS_OPTIONAL, &md->flags); |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2212 | dm_sync_table(md); |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2213 | |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2214 | return old_map; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2215 | } |
| 2216 | |
Alasdair G Kergon | a794015 | 2009-12-10 23:52:23 +0000 | [diff] [blame] | 2217 | /* |
| 2218 | * Returns unbound table for the caller to free. |
| 2219 | */ |
| 2220 | static struct dm_table *__unbind(struct mapped_device *md) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2221 | { |
| 2222 | struct dm_table *map = md->map; |
| 2223 | |
| 2224 | if (!map) |
Alasdair G Kergon | a794015 | 2009-12-10 23:52:23 +0000 | [diff] [blame] | 2225 | return NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2226 | |
| 2227 | dm_table_event_callback(map, NULL, NULL); |
Monam Agarwal | 9cdb852 | 2014-03-23 23:58:27 +0530 | [diff] [blame] | 2228 | RCU_INIT_POINTER(md->map, NULL); |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2229 | dm_sync_table(md); |
Alasdair G Kergon | a794015 | 2009-12-10 23:52:23 +0000 | [diff] [blame] | 2230 | |
| 2231 | return map; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2232 | } |
| 2233 | |
| 2234 | /* |
| 2235 | * Constructor for a new device. |
| 2236 | */ |
Alasdair G Kergon | 2b06cff | 2006-06-26 00:27:32 -0700 | [diff] [blame] | 2237 | int dm_create(int minor, struct mapped_device **result) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2238 | { |
| 2239 | struct mapped_device *md; |
| 2240 | |
Alasdair G Kergon | 2b06cff | 2006-06-26 00:27:32 -0700 | [diff] [blame] | 2241 | md = alloc_dev(minor); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2242 | if (!md) |
| 2243 | return -ENXIO; |
| 2244 | |
Milan Broz | 784aae7 | 2009-01-06 03:05:12 +0000 | [diff] [blame] | 2245 | dm_sysfs_init(md); |
| 2246 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2247 | *result = md; |
| 2248 | return 0; |
| 2249 | } |
| 2250 | |
Mike Snitzer | a5664da | 2010-08-12 04:14:01 +0100 | [diff] [blame] | 2251 | /* |
| 2252 | * Functions to manage md->type. |
| 2253 | * All are required to hold md->type_lock. |
| 2254 | */ |
| 2255 | void dm_lock_md_type(struct mapped_device *md) |
| 2256 | { |
| 2257 | mutex_lock(&md->type_lock); |
| 2258 | } |
| 2259 | |
| 2260 | void dm_unlock_md_type(struct mapped_device *md) |
| 2261 | { |
| 2262 | mutex_unlock(&md->type_lock); |
| 2263 | } |
| 2264 | |
| 2265 | void dm_set_md_type(struct mapped_device *md, unsigned type) |
| 2266 | { |
Mike Snitzer | 00c4fc3 | 2013-08-27 18:57:03 -0400 | [diff] [blame] | 2267 | BUG_ON(!mutex_is_locked(&md->type_lock)); |
Mike Snitzer | a5664da | 2010-08-12 04:14:01 +0100 | [diff] [blame] | 2268 | md->type = type; |
| 2269 | } |
| 2270 | |
| 2271 | unsigned dm_get_md_type(struct mapped_device *md) |
| 2272 | { |
Mike Snitzer | 00c4fc3 | 2013-08-27 18:57:03 -0400 | [diff] [blame] | 2273 | BUG_ON(!mutex_is_locked(&md->type_lock)); |
Mike Snitzer | a5664da | 2010-08-12 04:14:01 +0100 | [diff] [blame] | 2274 | return md->type; |
| 2275 | } |
| 2276 | |
Alasdair G Kergon | 36a0456 | 2011-10-31 20:19:04 +0000 | [diff] [blame] | 2277 | struct target_type *dm_get_immutable_target_type(struct mapped_device *md) |
| 2278 | { |
| 2279 | return md->immutable_target_type; |
| 2280 | } |
| 2281 | |
Mike Snitzer | 4a0b4dd | 2010-08-12 04:14:02 +0100 | [diff] [blame] | 2282 | /* |
Mike Snitzer | f84cb8a | 2013-09-19 12:13:58 -0400 | [diff] [blame] | 2283 | * The queue_limits are only valid as long as you have a reference |
| 2284 | * count on 'md'. |
| 2285 | */ |
| 2286 | struct queue_limits *dm_get_queue_limits(struct mapped_device *md) |
| 2287 | { |
| 2288 | BUG_ON(!atomic_read(&md->holders)); |
| 2289 | return &md->queue->limits; |
| 2290 | } |
| 2291 | EXPORT_SYMBOL_GPL(dm_get_queue_limits); |
| 2292 | |
| 2293 | /* |
Mike Snitzer | 4a0b4dd | 2010-08-12 04:14:02 +0100 | [diff] [blame] | 2294 | * Fully initialize a request-based queue (->elevator, ->request_fn, etc). |
| 2295 | */ |
| 2296 | static int dm_init_request_based_queue(struct mapped_device *md) |
| 2297 | { |
| 2298 | struct request_queue *q = NULL; |
| 2299 | |
| 2300 | if (md->queue->elevator) |
| 2301 | return 1; |
| 2302 | |
| 2303 | /* Fully initialize the queue */ |
| 2304 | q = blk_init_allocated_queue(md->queue, dm_request_fn, NULL); |
| 2305 | if (!q) |
| 2306 | return 0; |
| 2307 | |
| 2308 | md->queue = q; |
Mike Snitzer | 4a0b4dd | 2010-08-12 04:14:02 +0100 | [diff] [blame] | 2309 | dm_init_md_queue(md); |
| 2310 | blk_queue_softirq_done(md->queue, dm_softirq_done); |
| 2311 | blk_queue_prep_rq(md->queue, dm_prep_fn); |
| 2312 | blk_queue_lld_busy(md->queue, dm_lld_busy); |
Mike Snitzer | 4a0b4dd | 2010-08-12 04:14:02 +0100 | [diff] [blame] | 2313 | |
| 2314 | elv_register_queue(md->queue); |
| 2315 | |
| 2316 | return 1; |
| 2317 | } |
| 2318 | |
| 2319 | /* |
| 2320 | * Setup the DM device's queue based on md's type |
| 2321 | */ |
| 2322 | int dm_setup_md_queue(struct mapped_device *md) |
| 2323 | { |
| 2324 | if ((dm_get_md_type(md) == DM_TYPE_REQUEST_BASED) && |
| 2325 | !dm_init_request_based_queue(md)) { |
| 2326 | DMWARN("Cannot initialize queue for request-based mapped device"); |
| 2327 | return -EINVAL; |
| 2328 | } |
| 2329 | |
| 2330 | return 0; |
| 2331 | } |
| 2332 | |
David Teigland | 637842c | 2006-01-06 00:20:00 -0800 | [diff] [blame] | 2333 | static struct mapped_device *dm_find_md(dev_t dev) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2334 | { |
| 2335 | struct mapped_device *md; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2336 | unsigned minor = MINOR(dev); |
| 2337 | |
| 2338 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) |
| 2339 | return NULL; |
| 2340 | |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 2341 | spin_lock(&_minor_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2342 | |
| 2343 | md = idr_find(&_minor_idr, minor); |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 2344 | if (md && (md == MINOR_ALLOCED || |
Tejun Heo | f331c02 | 2008-09-03 09:01:48 +0200 | [diff] [blame] | 2345 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
Kiyoshi Ueda | abdc568 | 2010-08-12 04:13:54 +0100 | [diff] [blame] | 2346 | dm_deleting_md(md) || |
Alasdair G Kergon | 17b2f66 | 2006-06-26 00:27:33 -0700 | [diff] [blame] | 2347 | test_bit(DMF_FREEING, &md->flags))) { |
David Teigland | 637842c | 2006-01-06 00:20:00 -0800 | [diff] [blame] | 2348 | md = NULL; |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 2349 | goto out; |
| 2350 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2351 | |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 2352 | out: |
Jeff Mahoney | f32c10b | 2006-06-26 00:27:22 -0700 | [diff] [blame] | 2353 | spin_unlock(&_minor_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2354 | |
David Teigland | 637842c | 2006-01-06 00:20:00 -0800 | [diff] [blame] | 2355 | return md; |
| 2356 | } |
| 2357 | |
David Teigland | d229a95 | 2006-01-06 00:20:01 -0800 | [diff] [blame] | 2358 | struct mapped_device *dm_get_md(dev_t dev) |
| 2359 | { |
| 2360 | struct mapped_device *md = dm_find_md(dev); |
| 2361 | |
| 2362 | if (md) |
| 2363 | dm_get(md); |
| 2364 | |
| 2365 | return md; |
| 2366 | } |
Alasdair G Kergon | 3cf2e4b | 2011-10-31 20:19:06 +0000 | [diff] [blame] | 2367 | EXPORT_SYMBOL_GPL(dm_get_md); |
David Teigland | d229a95 | 2006-01-06 00:20:01 -0800 | [diff] [blame] | 2368 | |
Alasdair G Kergon | 9ade92a | 2006-03-27 01:17:53 -0800 | [diff] [blame] | 2369 | void *dm_get_mdptr(struct mapped_device *md) |
David Teigland | 637842c | 2006-01-06 00:20:00 -0800 | [diff] [blame] | 2370 | { |
Alasdair G Kergon | 9ade92a | 2006-03-27 01:17:53 -0800 | [diff] [blame] | 2371 | return md->interface_ptr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2372 | } |
| 2373 | |
| 2374 | void dm_set_mdptr(struct mapped_device *md, void *ptr) |
| 2375 | { |
| 2376 | md->interface_ptr = ptr; |
| 2377 | } |
| 2378 | |
| 2379 | void dm_get(struct mapped_device *md) |
| 2380 | { |
| 2381 | atomic_inc(&md->holders); |
Kiyoshi Ueda | 3f77316d | 2010-08-12 04:13:56 +0100 | [diff] [blame] | 2382 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2383 | } |
| 2384 | |
Alasdair G Kergon | 72d9486 | 2006-06-26 00:27:35 -0700 | [diff] [blame] | 2385 | const char *dm_device_name(struct mapped_device *md) |
| 2386 | { |
| 2387 | return md->name; |
| 2388 | } |
| 2389 | EXPORT_SYMBOL_GPL(dm_device_name); |
| 2390 | |
Kiyoshi Ueda | 3f77316d | 2010-08-12 04:13:56 +0100 | [diff] [blame] | 2391 | static void __dm_destroy(struct mapped_device *md, bool wait) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2392 | { |
Mike Anderson | 1134e5a | 2006-03-27 01:17:54 -0800 | [diff] [blame] | 2393 | struct dm_table *map; |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2394 | int srcu_idx; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2395 | |
Kiyoshi Ueda | 3f77316d | 2010-08-12 04:13:56 +0100 | [diff] [blame] | 2396 | might_sleep(); |
Jeff Mahoney | fba9f90 | 2006-06-26 00:27:23 -0700 | [diff] [blame] | 2397 | |
Kiyoshi Ueda | 3f77316d | 2010-08-12 04:13:56 +0100 | [diff] [blame] | 2398 | spin_lock(&_minor_lock); |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2399 | map = dm_get_live_table(md, &srcu_idx); |
Kiyoshi Ueda | 3f77316d | 2010-08-12 04:13:56 +0100 | [diff] [blame] | 2400 | idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); |
| 2401 | set_bit(DMF_FREEING, &md->flags); |
| 2402 | spin_unlock(&_minor_lock); |
| 2403 | |
| 2404 | if (!dm_suspended_md(md)) { |
| 2405 | dm_table_presuspend_targets(map); |
| 2406 | dm_table_postsuspend_targets(map); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2407 | } |
Kiyoshi Ueda | 3f77316d | 2010-08-12 04:13:56 +0100 | [diff] [blame] | 2408 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2409 | /* dm_put_live_table must be before msleep, otherwise deadlock is possible */ |
| 2410 | dm_put_live_table(md, srcu_idx); |
| 2411 | |
Kiyoshi Ueda | 3f77316d | 2010-08-12 04:13:56 +0100 | [diff] [blame] | 2412 | /* |
| 2413 | * Rare, but there may be I/O requests still going to complete, |
| 2414 | * for example. Wait for all references to disappear. |
| 2415 | * No one should increment the reference count of the mapped_device, |
| 2416 | * after the mapped_device state becomes DMF_FREEING. |
| 2417 | */ |
| 2418 | if (wait) |
| 2419 | while (atomic_read(&md->holders)) |
| 2420 | msleep(1); |
| 2421 | else if (atomic_read(&md->holders)) |
| 2422 | DMWARN("%s: Forcibly removing mapped_device still in use! (%d users)", |
| 2423 | dm_device_name(md), atomic_read(&md->holders)); |
| 2424 | |
| 2425 | dm_sysfs_exit(md); |
Kiyoshi Ueda | 3f77316d | 2010-08-12 04:13:56 +0100 | [diff] [blame] | 2426 | dm_table_destroy(__unbind(md)); |
| 2427 | free_dev(md); |
| 2428 | } |
| 2429 | |
| 2430 | void dm_destroy(struct mapped_device *md) |
| 2431 | { |
| 2432 | __dm_destroy(md, true); |
| 2433 | } |
| 2434 | |
| 2435 | void dm_destroy_immediate(struct mapped_device *md) |
| 2436 | { |
| 2437 | __dm_destroy(md, false); |
| 2438 | } |
| 2439 | |
| 2440 | void dm_put(struct mapped_device *md) |
| 2441 | { |
| 2442 | atomic_dec(&md->holders); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2443 | } |
Edward Goggin | 79eb885 | 2007-05-09 02:32:56 -0700 | [diff] [blame] | 2444 | EXPORT_SYMBOL_GPL(dm_put); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2445 | |
Mikulas Patocka | 401600d | 2009-04-02 19:55:38 +0100 | [diff] [blame] | 2446 | static int dm_wait_for_completion(struct mapped_device *md, int interruptible) |
Milan Broz | 46125c1 | 2008-02-08 02:10:30 +0000 | [diff] [blame] | 2447 | { |
| 2448 | int r = 0; |
Mikulas Patocka | b44ebeb | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 2449 | DECLARE_WAITQUEUE(wait, current); |
| 2450 | |
Mikulas Patocka | b44ebeb | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 2451 | add_wait_queue(&md->wait, &wait); |
Milan Broz | 46125c1 | 2008-02-08 02:10:30 +0000 | [diff] [blame] | 2452 | |
| 2453 | while (1) { |
Mikulas Patocka | 401600d | 2009-04-02 19:55:38 +0100 | [diff] [blame] | 2454 | set_current_state(interruptible); |
Milan Broz | 46125c1 | 2008-02-08 02:10:30 +0000 | [diff] [blame] | 2455 | |
Kiyoshi Ueda | b4324fe | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 2456 | if (!md_in_flight(md)) |
Milan Broz | 46125c1 | 2008-02-08 02:10:30 +0000 | [diff] [blame] | 2457 | break; |
| 2458 | |
Mikulas Patocka | 401600d | 2009-04-02 19:55:38 +0100 | [diff] [blame] | 2459 | if (interruptible == TASK_INTERRUPTIBLE && |
| 2460 | signal_pending(current)) { |
Milan Broz | 46125c1 | 2008-02-08 02:10:30 +0000 | [diff] [blame] | 2461 | r = -EINTR; |
| 2462 | break; |
| 2463 | } |
| 2464 | |
| 2465 | io_schedule(); |
| 2466 | } |
| 2467 | set_current_state(TASK_RUNNING); |
| 2468 | |
Mikulas Patocka | b44ebeb | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 2469 | remove_wait_queue(&md->wait, &wait); |
| 2470 | |
Milan Broz | 46125c1 | 2008-02-08 02:10:30 +0000 | [diff] [blame] | 2471 | return r; |
| 2472 | } |
| 2473 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2474 | /* |
| 2475 | * Process the deferred bios |
| 2476 | */ |
Mikulas Patocka | ef20858 | 2009-04-02 19:55:38 +0100 | [diff] [blame] | 2477 | static void dm_wq_work(struct work_struct *work) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2478 | { |
Mikulas Patocka | ef20858 | 2009-04-02 19:55:38 +0100 | [diff] [blame] | 2479 | struct mapped_device *md = container_of(work, struct mapped_device, |
| 2480 | work); |
Milan Broz | 6d6f10d | 2008-02-08 02:10:22 +0000 | [diff] [blame] | 2481 | struct bio *c; |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2482 | int srcu_idx; |
| 2483 | struct dm_table *map; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2484 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2485 | map = dm_get_live_table(md, &srcu_idx); |
Mikulas Patocka | ef20858 | 2009-04-02 19:55:38 +0100 | [diff] [blame] | 2486 | |
Mikulas Patocka | 3b00b20 | 2009-04-09 00:27:15 +0100 | [diff] [blame] | 2487 | while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
Alasdair G Kergon | df12ee9 | 2009-04-09 00:27:13 +0100 | [diff] [blame] | 2488 | spin_lock_irq(&md->deferred_lock); |
| 2489 | c = bio_list_pop(&md->deferred); |
| 2490 | spin_unlock_irq(&md->deferred_lock); |
Mikulas Patocka | 022c261 | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 2491 | |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 2492 | if (!c) |
Alasdair G Kergon | df12ee9 | 2009-04-09 00:27:13 +0100 | [diff] [blame] | 2493 | break; |
Alasdair G Kergon | df12ee9 | 2009-04-09 00:27:13 +0100 | [diff] [blame] | 2494 | |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2495 | if (dm_request_based(md)) |
| 2496 | generic_make_request(c); |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 2497 | else |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2498 | __split_and_process_bio(md, map, c); |
Mikulas Patocka | 022c261 | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 2499 | } |
Milan Broz | 73d410c | 2008-02-08 02:10:25 +0000 | [diff] [blame] | 2500 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2501 | dm_put_live_table(md, srcu_idx); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2502 | } |
| 2503 | |
Mikulas Patocka | 9a1fb46 | 2009-04-02 19:55:36 +0100 | [diff] [blame] | 2504 | static void dm_queue_flush(struct mapped_device *md) |
Milan Broz | 304f3f6 | 2008-02-08 02:11:17 +0000 | [diff] [blame] | 2505 | { |
Mikulas Patocka | 3b00b20 | 2009-04-09 00:27:15 +0100 | [diff] [blame] | 2506 | clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
Peter Zijlstra | 4e857c5 | 2014-03-17 18:06:10 +0100 | [diff] [blame] | 2507 | smp_mb__after_atomic(); |
Mikulas Patocka | 53d5914 | 2009-04-02 19:55:37 +0100 | [diff] [blame] | 2508 | queue_work(md->wq, &md->work); |
Milan Broz | 304f3f6 | 2008-02-08 02:11:17 +0000 | [diff] [blame] | 2509 | } |
| 2510 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2511 | /* |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2512 | * Swap in a new table, returning the old one for the caller to destroy. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2513 | */ |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2514 | struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2515 | { |
Mike Christie | 87eb5b2 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2516 | struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL); |
Mike Snitzer | 754c5fc | 2009-06-22 10:12:34 +0100 | [diff] [blame] | 2517 | struct queue_limits limits; |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2518 | int r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2519 | |
Daniel Walker | e61290a | 2008-02-08 02:10:08 +0000 | [diff] [blame] | 2520 | mutex_lock(&md->suspend_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2521 | |
| 2522 | /* device must be suspended */ |
Kiyoshi Ueda | 4f186f8 | 2009-12-10 23:52:26 +0000 | [diff] [blame] | 2523 | if (!dm_suspended_md(md)) |
Alasdair G Kergon | 93c534a | 2005-07-12 15:53:05 -0700 | [diff] [blame] | 2524 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2525 | |
Mike Snitzer | 3ae7065 | 2012-09-26 23:45:45 +0100 | [diff] [blame] | 2526 | /* |
| 2527 | * If the new table has no data devices, retain the existing limits. |
| 2528 | * This helps multipath with queue_if_no_path if all paths disappear, |
| 2529 | * then new I/O is queued based on these limits, and then some paths |
| 2530 | * reappear. |
| 2531 | */ |
| 2532 | if (dm_table_has_no_data_devices(table)) { |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2533 | live_map = dm_get_live_table_fast(md); |
Mike Snitzer | 3ae7065 | 2012-09-26 23:45:45 +0100 | [diff] [blame] | 2534 | if (live_map) |
| 2535 | limits = md->queue->limits; |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2536 | dm_put_live_table_fast(md); |
Mike Snitzer | 3ae7065 | 2012-09-26 23:45:45 +0100 | [diff] [blame] | 2537 | } |
| 2538 | |
Mike Christie | 87eb5b2 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2539 | if (!live_map) { |
| 2540 | r = dm_calculate_queue_limits(table, &limits); |
| 2541 | if (r) { |
| 2542 | map = ERR_PTR(r); |
| 2543 | goto out; |
| 2544 | } |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2545 | } |
Mike Snitzer | 754c5fc | 2009-06-22 10:12:34 +0100 | [diff] [blame] | 2546 | |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2547 | map = __bind(md, table, &limits); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2548 | |
Alasdair G Kergon | 93c534a | 2005-07-12 15:53:05 -0700 | [diff] [blame] | 2549 | out: |
Daniel Walker | e61290a | 2008-02-08 02:10:08 +0000 | [diff] [blame] | 2550 | mutex_unlock(&md->suspend_lock); |
Alasdair G Kergon | 042d2a9 | 2009-12-10 23:52:24 +0000 | [diff] [blame] | 2551 | return map; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2552 | } |
| 2553 | |
| 2554 | /* |
| 2555 | * Functions to lock and unlock any filesystem running on the |
| 2556 | * device. |
| 2557 | */ |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2558 | static int lock_fs(struct mapped_device *md) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2559 | { |
Alasdair G Kergon | e39e2e9 | 2006-01-06 00:20:05 -0800 | [diff] [blame] | 2560 | int r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2561 | |
| 2562 | WARN_ON(md->frozen_sb); |
Alasdair G Kergon | dfbe03f | 2005-05-05 16:16:04 -0700 | [diff] [blame] | 2563 | |
Mikulas Patocka | db8fef4 | 2009-06-22 10:12:15 +0100 | [diff] [blame] | 2564 | md->frozen_sb = freeze_bdev(md->bdev); |
Alasdair G Kergon | dfbe03f | 2005-05-05 16:16:04 -0700 | [diff] [blame] | 2565 | if (IS_ERR(md->frozen_sb)) { |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2566 | r = PTR_ERR(md->frozen_sb); |
Alasdair G Kergon | e39e2e9 | 2006-01-06 00:20:05 -0800 | [diff] [blame] | 2567 | md->frozen_sb = NULL; |
| 2568 | return r; |
Alasdair G Kergon | dfbe03f | 2005-05-05 16:16:04 -0700 | [diff] [blame] | 2569 | } |
| 2570 | |
Alasdair G Kergon | aa8d7c2 | 2006-01-06 00:20:06 -0800 | [diff] [blame] | 2571 | set_bit(DMF_FROZEN, &md->flags); |
| 2572 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2573 | return 0; |
| 2574 | } |
| 2575 | |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2576 | static void unlock_fs(struct mapped_device *md) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2577 | { |
Alasdair G Kergon | aa8d7c2 | 2006-01-06 00:20:06 -0800 | [diff] [blame] | 2578 | if (!test_bit(DMF_FROZEN, &md->flags)) |
| 2579 | return; |
| 2580 | |
Mikulas Patocka | db8fef4 | 2009-06-22 10:12:15 +0100 | [diff] [blame] | 2581 | thaw_bdev(md->bdev, md->frozen_sb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2582 | md->frozen_sb = NULL; |
Alasdair G Kergon | aa8d7c2 | 2006-01-06 00:20:06 -0800 | [diff] [blame] | 2583 | clear_bit(DMF_FROZEN, &md->flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2584 | } |
| 2585 | |
| 2586 | /* |
| 2587 | * We need to be able to change a mapping table under a mounted |
| 2588 | * filesystem. For example we might want to move some data in |
| 2589 | * the background. Before the table can be swapped with |
| 2590 | * dm_bind_table, dm_suspend must be called to flush any in |
| 2591 | * flight bios and ensure that any further io gets deferred. |
| 2592 | */ |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 2593 | /* |
| 2594 | * Suspend mechanism in request-based dm. |
| 2595 | * |
Kiyoshi Ueda | 9f518b2 | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 2596 | * 1. Flush all I/Os by lock_fs() if needed. |
| 2597 | * 2. Stop dispatching any I/O by stopping the request_queue. |
| 2598 | * 3. Wait for all in-flight I/Os to be completed or requeued. |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 2599 | * |
Kiyoshi Ueda | 9f518b2 | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 2600 | * To abort suspend, start the request_queue. |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 2601 | */ |
Kiyoshi Ueda | a3d77d3 | 2006-12-08 02:41:04 -0800 | [diff] [blame] | 2602 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2603 | { |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2604 | struct dm_table *map = NULL; |
Milan Broz | 46125c1 | 2008-02-08 02:10:30 +0000 | [diff] [blame] | 2605 | int r = 0; |
Kiyoshi Ueda | a3d77d3 | 2006-12-08 02:41:04 -0800 | [diff] [blame] | 2606 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 2607 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2608 | |
Daniel Walker | e61290a | 2008-02-08 02:10:08 +0000 | [diff] [blame] | 2609 | mutex_lock(&md->suspend_lock); |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2610 | |
Kiyoshi Ueda | 4f186f8 | 2009-12-10 23:52:26 +0000 | [diff] [blame] | 2611 | if (dm_suspended_md(md)) { |
Milan Broz | 73d410c | 2008-02-08 02:10:25 +0000 | [diff] [blame] | 2612 | r = -EINVAL; |
Alasdair G Kergon | d287483 | 2006-11-08 17:44:43 -0800 | [diff] [blame] | 2613 | goto out_unlock; |
Milan Broz | 73d410c | 2008-02-08 02:10:25 +0000 | [diff] [blame] | 2614 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2615 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2616 | map = md->map; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2617 | |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 2618 | /* |
| 2619 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. |
| 2620 | * This flag is cleared before dm_suspend returns. |
| 2621 | */ |
| 2622 | if (noflush) |
| 2623 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
| 2624 | |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2625 | /* This does not get reverted if there's an error later. */ |
| 2626 | dm_table_presuspend_targets(map); |
| 2627 | |
Mikulas Patocka | 32a926d | 2009-06-22 10:12:17 +0100 | [diff] [blame] | 2628 | /* |
Kiyoshi Ueda | 9f518b2 | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 2629 | * Flush I/O to the device. |
| 2630 | * Any I/O submitted after lock_fs() may not be flushed. |
| 2631 | * noflush takes precedence over do_lockfs. |
| 2632 | * (lock_fs() flushes I/Os and waits for them to complete.) |
Mikulas Patocka | 32a926d | 2009-06-22 10:12:17 +0100 | [diff] [blame] | 2633 | */ |
| 2634 | if (!noflush && do_lockfs) { |
| 2635 | r = lock_fs(md); |
| 2636 | if (r) |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2637 | goto out_unlock; |
Alasdair G Kergon | aa8d7c2 | 2006-01-06 00:20:06 -0800 | [diff] [blame] | 2638 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2639 | |
| 2640 | /* |
Mikulas Patocka | 3b00b20 | 2009-04-09 00:27:15 +0100 | [diff] [blame] | 2641 | * Here we must make sure that no processes are submitting requests |
| 2642 | * to target drivers i.e. no one may be executing |
| 2643 | * __split_and_process_bio. This is called from dm_request and |
| 2644 | * dm_wq_work. |
| 2645 | * |
| 2646 | * To get all processes out of __split_and_process_bio in dm_request, |
| 2647 | * we take the write lock. To prevent any process from reentering |
Tejun Heo | 6a8736d | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 2648 | * __split_and_process_bio from dm_request and quiesce the thread |
| 2649 | * (dm_wq_work), we set BMF_BLOCK_IO_FOR_SUSPEND and call |
| 2650 | * flush_workqueue(md->wq). |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2651 | */ |
Alasdair G Kergon | 1eb787e | 2009-04-09 00:27:14 +0100 | [diff] [blame] | 2652 | set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2653 | synchronize_srcu(&md->io_barrier); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2654 | |
Kiyoshi Ueda | d0bcb87 | 2009-12-10 23:52:18 +0000 | [diff] [blame] | 2655 | /* |
Tejun Heo | 29e4013 | 2010-09-08 18:07:00 +0200 | [diff] [blame] | 2656 | * Stop md->queue before flushing md->wq in case request-based |
| 2657 | * dm defers requests to md->wq from md->queue. |
Kiyoshi Ueda | d0bcb87 | 2009-12-10 23:52:18 +0000 | [diff] [blame] | 2658 | */ |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 2659 | if (dm_request_based(md)) |
Kiyoshi Ueda | 9f518b2 | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 2660 | stop_queue(md->queue); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 2661 | |
Kiyoshi Ueda | d0bcb87 | 2009-12-10 23:52:18 +0000 | [diff] [blame] | 2662 | flush_workqueue(md->wq); |
| 2663 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2664 | /* |
Mikulas Patocka | 3b00b20 | 2009-04-09 00:27:15 +0100 | [diff] [blame] | 2665 | * At this point no more requests are entering target request routines. |
| 2666 | * We call dm_wait_for_completion to wait for all existing requests |
| 2667 | * to finish. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2668 | */ |
Mikulas Patocka | 401600d | 2009-04-02 19:55:38 +0100 | [diff] [blame] | 2669 | r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2670 | |
Milan Broz | 6d6f10d | 2008-02-08 02:10:22 +0000 | [diff] [blame] | 2671 | if (noflush) |
Mikulas Patocka | 022c261 | 2009-04-02 19:55:39 +0100 | [diff] [blame] | 2672 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2673 | synchronize_srcu(&md->io_barrier); |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 2674 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2675 | /* were we interrupted ? */ |
Milan Broz | 46125c1 | 2008-02-08 02:10:30 +0000 | [diff] [blame] | 2676 | if (r < 0) { |
Mikulas Patocka | 9a1fb46 | 2009-04-02 19:55:36 +0100 | [diff] [blame] | 2677 | dm_queue_flush(md); |
Milan Broz | 73d410c | 2008-02-08 02:10:25 +0000 | [diff] [blame] | 2678 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 2679 | if (dm_request_based(md)) |
Kiyoshi Ueda | 9f518b2 | 2009-12-10 23:52:16 +0000 | [diff] [blame] | 2680 | start_queue(md->queue); |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 2681 | |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2682 | unlock_fs(md); |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2683 | goto out_unlock; /* pushback list is already flushed, so skip flush */ |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2684 | } |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2685 | |
Mikulas Patocka | 3b00b20 | 2009-04-09 00:27:15 +0100 | [diff] [blame] | 2686 | /* |
| 2687 | * If dm_wait_for_completion returned 0, the device is completely |
| 2688 | * quiescent now. There is no request-processing activity. All new |
| 2689 | * requests are being added to md->deferred list. |
| 2690 | */ |
| 2691 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2692 | set_bit(DMF_SUSPENDED, &md->flags); |
| 2693 | |
Kiyoshi Ueda | 4d4471c | 2009-12-10 23:52:26 +0000 | [diff] [blame] | 2694 | dm_table_postsuspend_targets(map); |
| 2695 | |
Alasdair G Kergon | d287483 | 2006-11-08 17:44:43 -0800 | [diff] [blame] | 2696 | out_unlock: |
Daniel Walker | e61290a | 2008-02-08 02:10:08 +0000 | [diff] [blame] | 2697 | mutex_unlock(&md->suspend_lock); |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2698 | return r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2699 | } |
| 2700 | |
| 2701 | int dm_resume(struct mapped_device *md) |
| 2702 | { |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2703 | int r = -EINVAL; |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2704 | struct dm_table *map = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2705 | |
Daniel Walker | e61290a | 2008-02-08 02:10:08 +0000 | [diff] [blame] | 2706 | mutex_lock(&md->suspend_lock); |
Kiyoshi Ueda | 4f186f8 | 2009-12-10 23:52:26 +0000 | [diff] [blame] | 2707 | if (!dm_suspended_md(md)) |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2708 | goto out; |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2709 | |
Mikulas Patocka | 83d5e5b | 2013-07-10 23:41:18 +0100 | [diff] [blame] | 2710 | map = md->map; |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2711 | if (!map || !dm_table_get_size(map)) |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2712 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2713 | |
Milan Broz | 8757b77 | 2006-10-03 01:15:36 -0700 | [diff] [blame] | 2714 | r = dm_table_resume_targets(map); |
| 2715 | if (r) |
| 2716 | goto out; |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2717 | |
Mikulas Patocka | 9a1fb46 | 2009-04-02 19:55:36 +0100 | [diff] [blame] | 2718 | dm_queue_flush(md); |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2719 | |
Kiyoshi Ueda | cec47e3 | 2009-06-22 10:12:35 +0100 | [diff] [blame] | 2720 | /* |
| 2721 | * Flushing deferred I/Os must be done after targets are resumed |
| 2722 | * so that mapping of targets can work correctly. |
| 2723 | * Request-based dm is queueing the deferred I/Os in its request_queue. |
| 2724 | */ |
| 2725 | if (dm_request_based(md)) |
| 2726 | start_queue(md->queue); |
| 2727 | |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2728 | unlock_fs(md); |
| 2729 | |
| 2730 | clear_bit(DMF_SUSPENDED, &md->flags); |
| 2731 | |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2732 | r = 0; |
| 2733 | out: |
Daniel Walker | e61290a | 2008-02-08 02:10:08 +0000 | [diff] [blame] | 2734 | mutex_unlock(&md->suspend_lock); |
Alasdair G Kergon | 2ca3310 | 2005-07-28 21:16:00 -0700 | [diff] [blame] | 2735 | |
Alasdair G Kergon | cf222b3 | 2005-07-28 21:15:57 -0700 | [diff] [blame] | 2736 | return r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2737 | } |
| 2738 | |
Mikulas Patocka | fd2ed4d | 2013-08-16 10:54:23 -0400 | [diff] [blame] | 2739 | /* |
| 2740 | * Internal suspend/resume works like userspace-driven suspend. It waits |
| 2741 | * until all bios finish and prevents issuing new bios to the target drivers. |
| 2742 | * It may be used only from the kernel. |
| 2743 | * |
| 2744 | * Internal suspend holds md->suspend_lock, which prevents interaction with |
| 2745 | * userspace-driven suspend. |
| 2746 | */ |
| 2747 | |
| 2748 | void dm_internal_suspend(struct mapped_device *md) |
| 2749 | { |
| 2750 | mutex_lock(&md->suspend_lock); |
| 2751 | if (dm_suspended_md(md)) |
| 2752 | return; |
| 2753 | |
| 2754 | set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
| 2755 | synchronize_srcu(&md->io_barrier); |
| 2756 | flush_workqueue(md->wq); |
| 2757 | dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE); |
| 2758 | } |
| 2759 | |
| 2760 | void dm_internal_resume(struct mapped_device *md) |
| 2761 | { |
| 2762 | if (dm_suspended_md(md)) |
| 2763 | goto done; |
| 2764 | |
| 2765 | dm_queue_flush(md); |
| 2766 | |
| 2767 | done: |
| 2768 | mutex_unlock(&md->suspend_lock); |
| 2769 | } |
| 2770 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2771 | /*----------------------------------------------------------------- |
| 2772 | * Event notification. |
| 2773 | *---------------------------------------------------------------*/ |
Peter Rajnoha | 3abf85b | 2010-03-06 02:32:31 +0000 | [diff] [blame] | 2774 | int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action, |
Milan Broz | 60935eb | 2009-06-22 10:12:30 +0100 | [diff] [blame] | 2775 | unsigned cookie) |
Alasdair G Kergon | 69267a3 | 2007-12-13 14:15:57 +0000 | [diff] [blame] | 2776 | { |
Milan Broz | 60935eb | 2009-06-22 10:12:30 +0100 | [diff] [blame] | 2777 | char udev_cookie[DM_COOKIE_LENGTH]; |
| 2778 | char *envp[] = { udev_cookie, NULL }; |
| 2779 | |
| 2780 | if (!cookie) |
Peter Rajnoha | 3abf85b | 2010-03-06 02:32:31 +0000 | [diff] [blame] | 2781 | return kobject_uevent(&disk_to_dev(md->disk)->kobj, action); |
Milan Broz | 60935eb | 2009-06-22 10:12:30 +0100 | [diff] [blame] | 2782 | else { |
| 2783 | snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u", |
| 2784 | DM_COOKIE_ENV_VAR_NAME, cookie); |
Peter Rajnoha | 3abf85b | 2010-03-06 02:32:31 +0000 | [diff] [blame] | 2785 | return kobject_uevent_env(&disk_to_dev(md->disk)->kobj, |
| 2786 | action, envp); |
Milan Broz | 60935eb | 2009-06-22 10:12:30 +0100 | [diff] [blame] | 2787 | } |
Alasdair G Kergon | 69267a3 | 2007-12-13 14:15:57 +0000 | [diff] [blame] | 2788 | } |
| 2789 | |
Mike Anderson | 7a8c3d3 | 2007-10-19 22:48:01 +0100 | [diff] [blame] | 2790 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
| 2791 | { |
| 2792 | return atomic_add_return(1, &md->uevent_seq); |
| 2793 | } |
| 2794 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2795 | uint32_t dm_get_event_nr(struct mapped_device *md) |
| 2796 | { |
| 2797 | return atomic_read(&md->event_nr); |
| 2798 | } |
| 2799 | |
| 2800 | int dm_wait_event(struct mapped_device *md, int event_nr) |
| 2801 | { |
| 2802 | return wait_event_interruptible(md->eventq, |
| 2803 | (event_nr != atomic_read(&md->event_nr))); |
| 2804 | } |
| 2805 | |
Mike Anderson | 7a8c3d3 | 2007-10-19 22:48:01 +0100 | [diff] [blame] | 2806 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
| 2807 | { |
| 2808 | unsigned long flags; |
| 2809 | |
| 2810 | spin_lock_irqsave(&md->uevent_lock, flags); |
| 2811 | list_add(elist, &md->uevent_list); |
| 2812 | spin_unlock_irqrestore(&md->uevent_lock, flags); |
| 2813 | } |
| 2814 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2815 | /* |
| 2816 | * The gendisk is only valid as long as you have a reference |
| 2817 | * count on 'md'. |
| 2818 | */ |
| 2819 | struct gendisk *dm_disk(struct mapped_device *md) |
| 2820 | { |
| 2821 | return md->disk; |
| 2822 | } |
| 2823 | |
Milan Broz | 784aae7 | 2009-01-06 03:05:12 +0000 | [diff] [blame] | 2824 | struct kobject *dm_kobject(struct mapped_device *md) |
| 2825 | { |
Mikulas Patocka | 2995fa7 | 2014-01-13 19:37:54 -0500 | [diff] [blame] | 2826 | return &md->kobj_holder.kobj; |
Milan Broz | 784aae7 | 2009-01-06 03:05:12 +0000 | [diff] [blame] | 2827 | } |
| 2828 | |
Milan Broz | 784aae7 | 2009-01-06 03:05:12 +0000 | [diff] [blame] | 2829 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) |
| 2830 | { |
| 2831 | struct mapped_device *md; |
| 2832 | |
Mikulas Patocka | 2995fa7 | 2014-01-13 19:37:54 -0500 | [diff] [blame] | 2833 | md = container_of(kobj, struct mapped_device, kobj_holder.kobj); |
Milan Broz | 784aae7 | 2009-01-06 03:05:12 +0000 | [diff] [blame] | 2834 | |
Milan Broz | 4d89b7b | 2009-06-22 10:12:11 +0100 | [diff] [blame] | 2835 | if (test_bit(DMF_FREEING, &md->flags) || |
Mike Anderson | 432a212 | 2009-12-10 23:52:20 +0000 | [diff] [blame] | 2836 | dm_deleting_md(md)) |
Milan Broz | 4d89b7b | 2009-06-22 10:12:11 +0100 | [diff] [blame] | 2837 | return NULL; |
| 2838 | |
Milan Broz | 784aae7 | 2009-01-06 03:05:12 +0000 | [diff] [blame] | 2839 | dm_get(md); |
| 2840 | return md; |
| 2841 | } |
| 2842 | |
Kiyoshi Ueda | 4f186f8 | 2009-12-10 23:52:26 +0000 | [diff] [blame] | 2843 | int dm_suspended_md(struct mapped_device *md) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2844 | { |
| 2845 | return test_bit(DMF_SUSPENDED, &md->flags); |
| 2846 | } |
| 2847 | |
Mikulas Patocka | 2c140a2 | 2013-11-01 18:27:41 -0400 | [diff] [blame] | 2848 | int dm_test_deferred_remove_flag(struct mapped_device *md) |
| 2849 | { |
| 2850 | return test_bit(DMF_DEFERRED_REMOVE, &md->flags); |
| 2851 | } |
| 2852 | |
Kiyoshi Ueda | 64dbce5 | 2009-12-10 23:52:27 +0000 | [diff] [blame] | 2853 | int dm_suspended(struct dm_target *ti) |
| 2854 | { |
Kiyoshi Ueda | ecdb2e2 | 2010-03-06 02:29:52 +0000 | [diff] [blame] | 2855 | return dm_suspended_md(dm_table_get_md(ti->table)); |
Kiyoshi Ueda | 64dbce5 | 2009-12-10 23:52:27 +0000 | [diff] [blame] | 2856 | } |
| 2857 | EXPORT_SYMBOL_GPL(dm_suspended); |
| 2858 | |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 2859 | int dm_noflush_suspending(struct dm_target *ti) |
| 2860 | { |
Kiyoshi Ueda | ecdb2e2 | 2010-03-06 02:29:52 +0000 | [diff] [blame] | 2861 | return __noflush_suspending(dm_table_get_md(ti->table)); |
Kiyoshi Ueda | 2e93ccc | 2006-12-08 02:41:09 -0800 | [diff] [blame] | 2862 | } |
| 2863 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); |
| 2864 | |
Mikulas Patocka | c0820cf | 2012-12-21 20:23:38 +0000 | [diff] [blame] | 2865 | struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity, unsigned per_bio_data_size) |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2866 | { |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2867 | struct dm_md_mempools *pools = kzalloc(sizeof(*pools), GFP_KERNEL); |
| 2868 | struct kmem_cache *cachep; |
| 2869 | unsigned int pool_size; |
| 2870 | unsigned int front_pad; |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2871 | |
| 2872 | if (!pools) |
| 2873 | return NULL; |
| 2874 | |
Jun'ichi Nomura | 23e5083 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2875 | if (type == DM_TYPE_BIO_BASED) { |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2876 | cachep = _io_cache; |
Mike Snitzer | e860313 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 2877 | pool_size = dm_get_reserved_bio_based_ios(); |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2878 | front_pad = roundup(per_bio_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone); |
| 2879 | } else if (type == DM_TYPE_REQUEST_BASED) { |
| 2880 | cachep = _rq_tio_cache; |
Mike Snitzer | f479082 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 2881 | pool_size = dm_get_reserved_rq_based_ios(); |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2882 | front_pad = offsetof(struct dm_rq_clone_bio_info, clone); |
| 2883 | /* per_bio_data_size is not used. See __bind_mempools(). */ |
| 2884 | WARN_ON(per_bio_data_size != 0); |
| 2885 | } else |
| 2886 | goto out; |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2887 | |
Mike Snitzer | 6cfa585 | 2013-09-12 18:06:11 -0400 | [diff] [blame] | 2888 | pools->io_pool = mempool_create_slab_pool(pool_size, cachep); |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2889 | if (!pools->io_pool) |
| 2890 | goto out; |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2891 | |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2892 | pools->bs = bioset_create(pool_size, front_pad); |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2893 | if (!pools->bs) |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2894 | goto out; |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2895 | |
Martin K. Petersen | a91a278 | 2011-03-17 11:11:05 +0100 | [diff] [blame] | 2896 | if (integrity && bioset_integrity_create(pools->bs, pool_size)) |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2897 | goto out; |
Martin K. Petersen | a91a278 | 2011-03-17 11:11:05 +0100 | [diff] [blame] | 2898 | |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2899 | return pools; |
| 2900 | |
Jun'ichi Nomura | 5f01520 | 2013-03-01 22:45:48 +0000 | [diff] [blame] | 2901 | out: |
| 2902 | dm_free_md_mempools(pools); |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2903 | |
| 2904 | return NULL; |
| 2905 | } |
| 2906 | |
| 2907 | void dm_free_md_mempools(struct dm_md_mempools *pools) |
| 2908 | { |
| 2909 | if (!pools) |
| 2910 | return; |
| 2911 | |
| 2912 | if (pools->io_pool) |
| 2913 | mempool_destroy(pools->io_pool); |
| 2914 | |
Kiyoshi Ueda | e6ee8c0 | 2009-06-22 10:12:36 +0100 | [diff] [blame] | 2915 | if (pools->bs) |
| 2916 | bioset_free(pools->bs); |
| 2917 | |
| 2918 | kfree(pools); |
| 2919 | } |
| 2920 | |
Alexey Dobriyan | 83d5cde | 2009-09-21 17:01:13 -0700 | [diff] [blame] | 2921 | static const struct block_device_operations dm_blk_dops = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2922 | .open = dm_blk_open, |
| 2923 | .release = dm_blk_close, |
Milan Broz | aa129a2 | 2006-10-03 01:15:15 -0700 | [diff] [blame] | 2924 | .ioctl = dm_blk_ioctl, |
Darrick J. Wong | 3ac51e7 | 2006-03-27 01:17:54 -0800 | [diff] [blame] | 2925 | .getgeo = dm_blk_getgeo, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2926 | .owner = THIS_MODULE |
| 2927 | }; |
| 2928 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2929 | /* |
| 2930 | * module hooks |
| 2931 | */ |
| 2932 | module_init(dm_init); |
| 2933 | module_exit(dm_exit); |
| 2934 | |
| 2935 | module_param(major, uint, 0); |
| 2936 | MODULE_PARM_DESC(major, "The major number of the device mapper"); |
Mike Snitzer | f479082 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 2937 | |
Mike Snitzer | e860313 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 2938 | module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR); |
| 2939 | MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools"); |
| 2940 | |
Mike Snitzer | f479082 | 2013-09-12 18:06:12 -0400 | [diff] [blame] | 2941 | module_param(reserved_rq_based_ios, uint, S_IRUGO | S_IWUSR); |
| 2942 | MODULE_PARM_DESC(reserved_rq_based_ios, "Reserved IOs in request-based mempools"); |
| 2943 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2944 | MODULE_DESCRIPTION(DM_NAME " driver"); |
| 2945 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); |
| 2946 | MODULE_LICENSE("GPL"); |