1 /* 2 * Copyright (C) 2007 Oracle. All rights reserved. 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public 6 * License v2 as published by the Free Software Foundation. 7 * 8 * This program is distributed in the hope that it will be useful, 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 11 * General Public License for more details. 12 * 13 * You should have received a copy of the GNU General Public 14 * License along with this program; if not, write to the 15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 16 * Boston, MA 021110-1307, USA. 17 */ 18 19 #ifndef __BTRFS_VOLUMES_ 20 #define __BTRFS_VOLUMES_ 21 22 #include <linux/bio.h> 23 #include <linux/sort.h> 24 #include <linux/btrfs.h> 25 #include "async-thread.h" 26 27 extern struct mutex uuid_mutex; 28 29 #define BTRFS_STRIPE_LEN (64 * 1024) 30 31 struct buffer_head; 32 struct btrfs_pending_bios { 33 struct bio *head; 34 struct bio *tail; 35 }; 36 37 /* 38 * Use sequence counter to get consistent device stat data on 39 * 32-bit processors. 40 */ 41 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 42 #include <linux/seqlock.h> 43 #define __BTRFS_NEED_DEVICE_DATA_ORDERED 44 #define btrfs_device_data_ordered_init(device) \ 45 seqcount_init(&device->data_seqcount) 46 #else 47 #define btrfs_device_data_ordered_init(device) do { } while (0) 48 #endif 49 50 struct btrfs_device { 51 struct list_head dev_list; 52 struct list_head dev_alloc_list; 53 struct btrfs_fs_devices *fs_devices; 54 55 struct btrfs_root *dev_root; 56 57 struct rcu_string *name; 58 59 u64 generation; 60 61 spinlock_t io_lock ____cacheline_aligned; 62 int running_pending; 63 /* regular prio bios */ 64 struct btrfs_pending_bios pending_bios; 65 /* WRITE_SYNC bios */ 66 struct btrfs_pending_bios pending_sync_bios; 67 68 struct block_device *bdev; 69 70 /* the mode sent to blkdev_get */ 71 fmode_t mode; 72 73 int writeable; 74 int in_fs_metadata; 75 int missing; 76 int can_discard; 77 int is_tgtdev_for_dev_replace; 78 79 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED 80 seqcount_t data_seqcount; 81 #endif 82 83 /* the internal btrfs device id */ 84 u64 devid; 85 86 /* size of the device in memory */ 87 u64 total_bytes; 88 89 /* size of the device on disk */ 90 u64 disk_total_bytes; 91 92 /* bytes used */ 93 u64 bytes_used; 94 95 /* optimal io alignment for this device */ 96 u32 io_align; 97 98 /* optimal io width for this device */ 99 u32 io_width; 100 /* type and info about this device */ 101 u64 type; 102 103 /* minimal io size for this device */ 104 u32 sector_size; 105 106 /* physical drive uuid (or lvm uuid) */ 107 u8 uuid[BTRFS_UUID_SIZE]; 108 109 /* 110 * size of the device on the current transaction 111 * 112 * This variant is update when committing the transaction, 113 * and protected by device_list_mutex 114 */ 115 u64 commit_total_bytes; 116 117 /* bytes used on the current transaction */ 118 u64 commit_bytes_used; 119 /* 120 * used to manage the device which is resized 121 * 122 * It is protected by chunk_lock. 123 */ 124 struct list_head resized_list; 125 126 /* for sending down flush barriers */ 127 int nobarriers; 128 struct bio *flush_bio; 129 struct completion flush_wait; 130 131 /* per-device scrub information */ 132 struct scrub_ctx *scrub_device; 133 134 struct btrfs_work work; 135 struct rcu_head rcu; 136 struct work_struct rcu_work; 137 138 /* readahead state */ 139 spinlock_t reada_lock; 140 atomic_t reada_in_flight; 141 u64 reada_next; 142 struct reada_zone *reada_curr_zone; 143 struct radix_tree_root reada_zones; 144 struct radix_tree_root reada_extents; 145 146 /* disk I/O failure stats. For detailed description refer to 147 * enum btrfs_dev_stat_values in ioctl.h */ 148 int dev_stats_valid; 149 150 /* Counter to record the change of device stats */ 151 atomic_t dev_stats_ccnt; 152 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX]; 153 }; 154 155 /* 156 * If we read those variants at the context of their own lock, we needn't 157 * use the following helpers, reading them directly is safe. 158 */ 159 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 160 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 161 static inline u64 \ 162 btrfs_device_get_##name(const struct btrfs_device *dev) \ 163 { \ 164 u64 size; \ 165 unsigned int seq; \ 166 \ 167 do { \ 168 seq = read_seqcount_begin(&dev->data_seqcount); \ 169 size = dev->name; \ 170 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \ 171 return size; \ 172 } \ 173 \ 174 static inline void \ 175 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 176 { \ 177 preempt_disable(); \ 178 write_seqcount_begin(&dev->data_seqcount); \ 179 dev->name = size; \ 180 write_seqcount_end(&dev->data_seqcount); \ 181 preempt_enable(); \ 182 } 183 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT) 184 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 185 static inline u64 \ 186 btrfs_device_get_##name(const struct btrfs_device *dev) \ 187 { \ 188 u64 size; \ 189 \ 190 preempt_disable(); \ 191 size = dev->name; \ 192 preempt_enable(); \ 193 return size; \ 194 } \ 195 \ 196 static inline void \ 197 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 198 { \ 199 preempt_disable(); \ 200 dev->name = size; \ 201 preempt_enable(); \ 202 } 203 #else 204 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 205 static inline u64 \ 206 btrfs_device_get_##name(const struct btrfs_device *dev) \ 207 { \ 208 return dev->name; \ 209 } \ 210 \ 211 static inline void \ 212 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 213 { \ 214 dev->name = size; \ 215 } 216 #endif 217 218 BTRFS_DEVICE_GETSET_FUNCS(total_bytes); 219 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes); 220 BTRFS_DEVICE_GETSET_FUNCS(bytes_used); 221 222 struct btrfs_fs_devices { 223 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */ 224 225 u64 num_devices; 226 u64 open_devices; 227 u64 rw_devices; 228 u64 missing_devices; 229 u64 total_rw_bytes; 230 u64 total_devices; 231 struct block_device *latest_bdev; 232 233 /* all of the devices in the FS, protected by a mutex 234 * so we can safely walk it to write out the supers without 235 * worrying about add/remove by the multi-device code. 236 * Scrubbing super can kick off supers writing by holding 237 * this mutex lock. 238 */ 239 struct mutex device_list_mutex; 240 struct list_head devices; 241 242 struct list_head resized_devices; 243 /* devices not currently being allocated */ 244 struct list_head alloc_list; 245 struct list_head list; 246 247 struct btrfs_fs_devices *seed; 248 int seeding; 249 250 int opened; 251 252 /* set when we find or add a device that doesn't have the 253 * nonrot flag set 254 */ 255 int rotating; 256 257 struct btrfs_fs_info *fs_info; 258 /* sysfs kobjects */ 259 struct kobject fsid_kobj; 260 struct kobject *device_dir_kobj; 261 struct completion kobj_unregister; 262 }; 263 264 #define BTRFS_BIO_INLINE_CSUM_SIZE 64 265 266 /* 267 * we need the mirror number and stripe index to be passed around 268 * the call chain while we are processing end_io (especially errors). 269 * Really, what we need is a btrfs_bio structure that has this info 270 * and is properly sized with its stripe array, but we're not there 271 * quite yet. We have our own btrfs bioset, and all of the bios 272 * we allocate are actually btrfs_io_bios. We'll cram as much of 273 * struct btrfs_bio as we can into this over time. 274 */ 275 typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err); 276 struct btrfs_io_bio { 277 unsigned int mirror_num; 278 unsigned int stripe_index; 279 u64 logical; 280 u8 *csum; 281 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE]; 282 u8 *csum_allocated; 283 btrfs_io_bio_end_io_t *end_io; 284 struct bio bio; 285 }; 286 287 static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio) 288 { 289 return container_of(bio, struct btrfs_io_bio, bio); 290 } 291 292 struct btrfs_bio_stripe { 293 struct btrfs_device *dev; 294 u64 physical; 295 u64 length; /* only used for discard mappings */ 296 }; 297 298 struct btrfs_bio; 299 typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err); 300 301 struct btrfs_bio { 302 atomic_t refs; 303 atomic_t stripes_pending; 304 struct btrfs_fs_info *fs_info; 305 u64 map_type; /* get from map_lookup->type */ 306 bio_end_io_t *end_io; 307 struct bio *orig_bio; 308 unsigned long flags; 309 void *private; 310 atomic_t error; 311 int max_errors; 312 int num_stripes; 313 int mirror_num; 314 int num_tgtdevs; 315 int *tgtdev_map; 316 /* 317 * logical block numbers for the start of each stripe 318 * The last one or two are p/q. These are sorted, 319 * so raid_map[0] is the start of our full stripe 320 */ 321 u64 *raid_map; 322 struct btrfs_bio_stripe stripes[]; 323 }; 324 325 struct btrfs_device_info { 326 struct btrfs_device *dev; 327 u64 dev_offset; 328 u64 max_avail; 329 u64 total_avail; 330 }; 331 332 struct btrfs_raid_attr { 333 int sub_stripes; /* sub_stripes info for map */ 334 int dev_stripes; /* stripes per dev */ 335 int devs_max; /* max devs to use */ 336 int devs_min; /* min devs needed */ 337 int tolerated_failures; /* max tolerated fail devs */ 338 int devs_increment; /* ndevs has to be a multiple of this */ 339 int ncopies; /* how many copies to data has */ 340 }; 341 342 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES]; 343 344 extern const u64 btrfs_raid_group[BTRFS_NR_RAID_TYPES]; 345 346 struct map_lookup { 347 u64 type; 348 int io_align; 349 int io_width; 350 int stripe_len; 351 int sector_size; 352 int num_stripes; 353 int sub_stripes; 354 struct btrfs_bio_stripe stripes[]; 355 }; 356 357 #define map_lookup_size(n) (sizeof(struct map_lookup) + \ 358 (sizeof(struct btrfs_bio_stripe) * (n))) 359 360 /* 361 * Restriper's general type filter 362 */ 363 #define BTRFS_BALANCE_DATA (1ULL << 0) 364 #define BTRFS_BALANCE_SYSTEM (1ULL << 1) 365 #define BTRFS_BALANCE_METADATA (1ULL << 2) 366 367 #define BTRFS_BALANCE_TYPE_MASK (BTRFS_BALANCE_DATA | \ 368 BTRFS_BALANCE_SYSTEM | \ 369 BTRFS_BALANCE_METADATA) 370 371 #define BTRFS_BALANCE_FORCE (1ULL << 3) 372 #define BTRFS_BALANCE_RESUME (1ULL << 4) 373 374 /* 375 * Balance filters 376 */ 377 #define BTRFS_BALANCE_ARGS_PROFILES (1ULL << 0) 378 #define BTRFS_BALANCE_ARGS_USAGE (1ULL << 1) 379 #define BTRFS_BALANCE_ARGS_DEVID (1ULL << 2) 380 #define BTRFS_BALANCE_ARGS_DRANGE (1ULL << 3) 381 #define BTRFS_BALANCE_ARGS_VRANGE (1ULL << 4) 382 #define BTRFS_BALANCE_ARGS_LIMIT (1ULL << 5) 383 #define BTRFS_BALANCE_ARGS_LIMIT_RANGE (1ULL << 6) 384 #define BTRFS_BALANCE_ARGS_STRIPES_RANGE (1ULL << 7) 385 #define BTRFS_BALANCE_ARGS_USAGE_RANGE (1ULL << 10) 386 387 #define BTRFS_BALANCE_ARGS_MASK \ 388 (BTRFS_BALANCE_ARGS_PROFILES | \ 389 BTRFS_BALANCE_ARGS_USAGE | \ 390 BTRFS_BALANCE_ARGS_DEVID | \ 391 BTRFS_BALANCE_ARGS_DRANGE | \ 392 BTRFS_BALANCE_ARGS_VRANGE | \ 393 BTRFS_BALANCE_ARGS_LIMIT | \ 394 BTRFS_BALANCE_ARGS_LIMIT_RANGE | \ 395 BTRFS_BALANCE_ARGS_STRIPES_RANGE | \ 396 BTRFS_BALANCE_ARGS_USAGE_RANGE) 397 398 /* 399 * Profile changing flags. When SOFT is set we won't relocate chunk if 400 * it already has the target profile (even though it may be 401 * half-filled). 402 */ 403 #define BTRFS_BALANCE_ARGS_CONVERT (1ULL << 8) 404 #define BTRFS_BALANCE_ARGS_SOFT (1ULL << 9) 405 406 struct btrfs_balance_args; 407 struct btrfs_balance_progress; 408 struct btrfs_balance_control { 409 struct btrfs_fs_info *fs_info; 410 411 struct btrfs_balance_args data; 412 struct btrfs_balance_args meta; 413 struct btrfs_balance_args sys; 414 415 u64 flags; 416 417 struct btrfs_balance_progress stat; 418 }; 419 420 int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start, 421 u64 end, u64 *length); 422 void btrfs_get_bbio(struct btrfs_bio *bbio); 423 void btrfs_put_bbio(struct btrfs_bio *bbio); 424 int btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, 425 u64 logical, u64 *length, 426 struct btrfs_bio **bbio_ret, int mirror_num); 427 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, int rw, 428 u64 logical, u64 *length, 429 struct btrfs_bio **bbio_ret, int mirror_num, 430 int need_raid_map); 431 int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree, 432 u64 chunk_start, u64 physical, u64 devid, 433 u64 **logical, int *naddrs, int *stripe_len); 434 int btrfs_read_sys_array(struct btrfs_root *root); 435 int btrfs_read_chunk_tree(struct btrfs_root *root); 436 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, 437 struct btrfs_root *extent_root, u64 type); 438 void btrfs_mapping_init(struct btrfs_mapping_tree *tree); 439 void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree); 440 int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, 441 int mirror_num, int async_submit); 442 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, 443 fmode_t flags, void *holder); 444 int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder, 445 struct btrfs_fs_devices **fs_devices_ret); 446 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices); 447 void btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices, int step); 448 int btrfs_find_device_missing_or_by_path(struct btrfs_root *root, 449 char *device_path, 450 struct btrfs_device **device); 451 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info, 452 const u64 *devid, 453 const u8 *uuid); 454 int btrfs_rm_device(struct btrfs_root *root, char *device_path); 455 void btrfs_cleanup_fs_uuids(void); 456 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len); 457 int btrfs_grow_device(struct btrfs_trans_handle *trans, 458 struct btrfs_device *device, u64 new_size); 459 struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid, 460 u8 *uuid, u8 *fsid); 461 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); 462 int btrfs_init_new_device(struct btrfs_root *root, char *path); 463 int btrfs_init_dev_replace_tgtdev(struct btrfs_root *root, char *device_path, 464 struct btrfs_device *srcdev, 465 struct btrfs_device **device_out); 466 int btrfs_balance(struct btrfs_balance_control *bctl, 467 struct btrfs_ioctl_balance_args *bargs); 468 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info); 469 int btrfs_recover_balance(struct btrfs_fs_info *fs_info); 470 int btrfs_pause_balance(struct btrfs_fs_info *fs_info); 471 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info); 472 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info); 473 int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info); 474 int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset); 475 int find_free_dev_extent_start(struct btrfs_transaction *transaction, 476 struct btrfs_device *device, u64 num_bytes, 477 u64 search_start, u64 *start, u64 *max_avail); 478 int find_free_dev_extent(struct btrfs_trans_handle *trans, 479 struct btrfs_device *device, u64 num_bytes, 480 u64 *start, u64 *max_avail); 481 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index); 482 int btrfs_get_dev_stats(struct btrfs_root *root, 483 struct btrfs_ioctl_get_dev_stats *stats); 484 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info); 485 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info); 486 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans, 487 struct btrfs_fs_info *fs_info); 488 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_fs_info *fs_info, 489 struct btrfs_device *srcdev); 490 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_fs_info *fs_info, 491 struct btrfs_device *srcdev); 492 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info, 493 struct btrfs_device *tgtdev); 494 void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info, 495 struct btrfs_device *tgtdev); 496 void btrfs_scratch_superblocks(struct block_device *bdev, char *device_path); 497 int btrfs_is_parity_mirror(struct btrfs_mapping_tree *map_tree, 498 u64 logical, u64 len, int mirror_num); 499 unsigned long btrfs_full_stripe_len(struct btrfs_root *root, 500 struct btrfs_mapping_tree *map_tree, 501 u64 logical); 502 int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans, 503 struct btrfs_root *extent_root, 504 u64 chunk_offset, u64 chunk_size); 505 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, 506 struct btrfs_root *root, u64 chunk_offset); 507 508 static inline int btrfs_dev_stats_dirty(struct btrfs_device *dev) 509 { 510 return atomic_read(&dev->dev_stats_ccnt); 511 } 512 513 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev, 514 int index) 515 { 516 atomic_inc(dev->dev_stat_values + index); 517 smp_mb__before_atomic(); 518 atomic_inc(&dev->dev_stats_ccnt); 519 } 520 521 static inline int btrfs_dev_stat_read(struct btrfs_device *dev, 522 int index) 523 { 524 return atomic_read(dev->dev_stat_values + index); 525 } 526 527 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev, 528 int index) 529 { 530 int ret; 531 532 ret = atomic_xchg(dev->dev_stat_values + index, 0); 533 smp_mb__before_atomic(); 534 atomic_inc(&dev->dev_stats_ccnt); 535 return ret; 536 } 537 538 static inline void btrfs_dev_stat_set(struct btrfs_device *dev, 539 int index, unsigned long val) 540 { 541 atomic_set(dev->dev_stat_values + index, val); 542 smp_mb__before_atomic(); 543 atomic_inc(&dev->dev_stats_ccnt); 544 } 545 546 static inline void btrfs_dev_stat_reset(struct btrfs_device *dev, 547 int index) 548 { 549 btrfs_dev_stat_set(dev, index, 0); 550 } 551 552 void btrfs_update_commit_device_size(struct btrfs_fs_info *fs_info); 553 void btrfs_update_commit_device_bytes_used(struct btrfs_root *root, 554 struct btrfs_transaction *transaction); 555 556 static inline void lock_chunks(struct btrfs_root *root) 557 { 558 mutex_lock(&root->fs_info->chunk_mutex); 559 } 560 561 static inline void unlock_chunks(struct btrfs_root *root) 562 { 563 mutex_unlock(&root->fs_info->chunk_mutex); 564 } 565 566 struct list_head *btrfs_get_fs_uuids(void); 567 void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info); 568 void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info); 569 void btrfs_close_one_device(struct btrfs_device *device); 570 571 #endif 572