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 SZ_64K 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 struct btrfs_fs_info *fs_info; 55 56 struct rcu_string *name; 57 58 u64 generation; 59 60 spinlock_t io_lock ____cacheline_aligned; 61 int running_pending; 62 /* regular prio bios */ 63 struct btrfs_pending_bios pending_bios; 64 /* sync bios */ 65 struct btrfs_pending_bios pending_sync_bios; 66 67 struct block_device *bdev; 68 69 /* the mode sent to blkdev_get */ 70 fmode_t mode; 71 72 int writeable; 73 int in_fs_metadata; 74 int missing; 75 int can_discard; 76 int is_tgtdev_for_dev_replace; 77 78 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED 79 seqcount_t data_seqcount; 80 #endif 81 82 /* the internal btrfs device id */ 83 u64 devid; 84 85 /* size of the device in memory */ 86 u64 total_bytes; 87 88 /* size of the device on disk */ 89 u64 disk_total_bytes; 90 91 /* bytes used */ 92 u64 bytes_used; 93 94 /* optimal io alignment for this device */ 95 u32 io_align; 96 97 /* optimal io width for this device */ 98 u32 io_width; 99 /* type and info about this device */ 100 u64 type; 101 102 /* minimal io size for this device */ 103 u32 sector_size; 104 105 /* physical drive uuid (or lvm uuid) */ 106 u8 uuid[BTRFS_UUID_SIZE]; 107 108 /* 109 * size of the device on the current transaction 110 * 111 * This variant is update when committing the transaction, 112 * and protected by device_list_mutex 113 */ 114 u64 commit_total_bytes; 115 116 /* bytes used on the current transaction */ 117 u64 commit_bytes_used; 118 /* 119 * used to manage the device which is resized 120 * 121 * It is protected by chunk_lock. 122 */ 123 struct list_head resized_list; 124 125 /* for sending down flush barriers */ 126 struct bio *flush_bio; 127 struct completion flush_wait; 128 129 /* per-device scrub information */ 130 struct scrub_ctx *scrub_device; 131 132 struct btrfs_work work; 133 struct rcu_head rcu; 134 struct work_struct rcu_work; 135 136 /* readahead state */ 137 spinlock_t reada_lock; 138 atomic_t reada_in_flight; 139 u64 reada_next; 140 struct reada_zone *reada_curr_zone; 141 struct radix_tree_root reada_zones; 142 struct radix_tree_root reada_extents; 143 144 /* disk I/O failure stats. For detailed description refer to 145 * enum btrfs_dev_stat_values in ioctl.h */ 146 int dev_stats_valid; 147 148 /* Counter to record the change of device stats */ 149 atomic_t dev_stats_ccnt; 150 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX]; 151 }; 152 153 /* 154 * If we read those variants at the context of their own lock, we needn't 155 * use the following helpers, reading them directly is safe. 156 */ 157 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 158 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 159 static inline u64 \ 160 btrfs_device_get_##name(const struct btrfs_device *dev) \ 161 { \ 162 u64 size; \ 163 unsigned int seq; \ 164 \ 165 do { \ 166 seq = read_seqcount_begin(&dev->data_seqcount); \ 167 size = dev->name; \ 168 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \ 169 return size; \ 170 } \ 171 \ 172 static inline void \ 173 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 174 { \ 175 preempt_disable(); \ 176 write_seqcount_begin(&dev->data_seqcount); \ 177 dev->name = size; \ 178 write_seqcount_end(&dev->data_seqcount); \ 179 preempt_enable(); \ 180 } 181 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT) 182 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 183 static inline u64 \ 184 btrfs_device_get_##name(const struct btrfs_device *dev) \ 185 { \ 186 u64 size; \ 187 \ 188 preempt_disable(); \ 189 size = dev->name; \ 190 preempt_enable(); \ 191 return size; \ 192 } \ 193 \ 194 static inline void \ 195 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 196 { \ 197 preempt_disable(); \ 198 dev->name = size; \ 199 preempt_enable(); \ 200 } 201 #else 202 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 203 static inline u64 \ 204 btrfs_device_get_##name(const struct btrfs_device *dev) \ 205 { \ 206 return dev->name; \ 207 } \ 208 \ 209 static inline void \ 210 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 211 { \ 212 dev->name = size; \ 213 } 214 #endif 215 216 BTRFS_DEVICE_GETSET_FUNCS(total_bytes); 217 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes); 218 BTRFS_DEVICE_GETSET_FUNCS(bytes_used); 219 220 struct btrfs_fs_devices { 221 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */ 222 223 u64 num_devices; 224 u64 open_devices; 225 u64 rw_devices; 226 u64 missing_devices; 227 u64 total_rw_bytes; 228 u64 total_devices; 229 struct block_device *latest_bdev; 230 231 /* all of the devices in the FS, protected by a mutex 232 * so we can safely walk it to write out the supers without 233 * worrying about add/remove by the multi-device code. 234 * Scrubbing super can kick off supers writing by holding 235 * this mutex lock. 236 */ 237 struct mutex device_list_mutex; 238 struct list_head devices; 239 240 struct list_head resized_devices; 241 /* devices not currently being allocated */ 242 struct list_head alloc_list; 243 struct list_head list; 244 245 struct btrfs_fs_devices *seed; 246 int seeding; 247 248 int opened; 249 250 /* set when we find or add a device that doesn't have the 251 * nonrot flag set 252 */ 253 int rotating; 254 255 struct btrfs_fs_info *fs_info; 256 /* sysfs kobjects */ 257 struct kobject fsid_kobj; 258 struct kobject *device_dir_kobj; 259 struct completion kobj_unregister; 260 }; 261 262 #define BTRFS_BIO_INLINE_CSUM_SIZE 64 263 264 /* 265 * we need the mirror number and stripe index to be passed around 266 * the call chain while we are processing end_io (especially errors). 267 * Really, what we need is a btrfs_bio structure that has this info 268 * and is properly sized with its stripe array, but we're not there 269 * quite yet. We have our own btrfs bioset, and all of the bios 270 * we allocate are actually btrfs_io_bios. We'll cram as much of 271 * struct btrfs_bio as we can into this over time. 272 */ 273 typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err); 274 struct btrfs_io_bio { 275 unsigned int mirror_num; 276 unsigned int stripe_index; 277 u64 logical; 278 u8 *csum; 279 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE]; 280 u8 *csum_allocated; 281 btrfs_io_bio_end_io_t *end_io; 282 struct bio bio; 283 }; 284 285 static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio) 286 { 287 return container_of(bio, struct btrfs_io_bio, bio); 288 } 289 290 struct btrfs_bio_stripe { 291 struct btrfs_device *dev; 292 u64 physical; 293 u64 length; /* only used for discard mappings */ 294 }; 295 296 struct btrfs_bio; 297 typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err); 298 299 struct btrfs_bio { 300 refcount_t refs; 301 atomic_t stripes_pending; 302 struct btrfs_fs_info *fs_info; 303 u64 map_type; /* get from map_lookup->type */ 304 bio_end_io_t *end_io; 305 struct bio *orig_bio; 306 unsigned long flags; 307 void *private; 308 atomic_t error; 309 int max_errors; 310 int num_stripes; 311 int mirror_num; 312 int num_tgtdevs; 313 int *tgtdev_map; 314 /* 315 * logical block numbers for the start of each stripe 316 * The last one or two are p/q. These are sorted, 317 * so raid_map[0] is the start of our full stripe 318 */ 319 u64 *raid_map; 320 struct btrfs_bio_stripe stripes[]; 321 }; 322 323 struct btrfs_device_info { 324 struct btrfs_device *dev; 325 u64 dev_offset; 326 u64 max_avail; 327 u64 total_avail; 328 }; 329 330 struct btrfs_raid_attr { 331 int sub_stripes; /* sub_stripes info for map */ 332 int dev_stripes; /* stripes per dev */ 333 int devs_max; /* max devs to use */ 334 int devs_min; /* min devs needed */ 335 int tolerated_failures; /* max tolerated fail devs */ 336 int devs_increment; /* ndevs has to be a multiple of this */ 337 int ncopies; /* how many copies to data has */ 338 }; 339 340 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES]; 341 extern const int btrfs_raid_mindev_error[BTRFS_NR_RAID_TYPES]; 342 extern const u64 btrfs_raid_group[BTRFS_NR_RAID_TYPES]; 343 344 struct map_lookup { 345 u64 type; 346 int io_align; 347 int io_width; 348 u64 stripe_len; 349 int sector_size; 350 int num_stripes; 351 int sub_stripes; 352 struct btrfs_bio_stripe stripes[]; 353 }; 354 355 #define map_lookup_size(n) (sizeof(struct map_lookup) + \ 356 (sizeof(struct btrfs_bio_stripe) * (n))) 357 358 struct btrfs_balance_args; 359 struct btrfs_balance_progress; 360 struct btrfs_balance_control { 361 struct btrfs_fs_info *fs_info; 362 363 struct btrfs_balance_args data; 364 struct btrfs_balance_args meta; 365 struct btrfs_balance_args sys; 366 367 u64 flags; 368 369 struct btrfs_balance_progress stat; 370 }; 371 372 enum btrfs_map_op { 373 BTRFS_MAP_READ, 374 BTRFS_MAP_WRITE, 375 BTRFS_MAP_DISCARD, 376 BTRFS_MAP_GET_READ_MIRRORS, 377 }; 378 379 static inline enum btrfs_map_op btrfs_op(struct bio *bio) 380 { 381 switch (bio_op(bio)) { 382 case REQ_OP_DISCARD: 383 return BTRFS_MAP_DISCARD; 384 case REQ_OP_WRITE: 385 return BTRFS_MAP_WRITE; 386 default: 387 WARN_ON_ONCE(1); 388 case REQ_OP_READ: 389 return BTRFS_MAP_READ; 390 } 391 } 392 393 int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start, 394 u64 end, u64 *length); 395 void btrfs_get_bbio(struct btrfs_bio *bbio); 396 void btrfs_put_bbio(struct btrfs_bio *bbio); 397 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, 398 u64 logical, u64 *length, 399 struct btrfs_bio **bbio_ret, int mirror_num); 400 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, 401 u64 logical, u64 *length, 402 struct btrfs_bio **bbio_ret); 403 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, 404 u64 chunk_start, u64 physical, u64 devid, 405 u64 **logical, int *naddrs, int *stripe_len); 406 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info); 407 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info); 408 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, 409 struct btrfs_fs_info *fs_info, u64 type); 410 void btrfs_mapping_init(struct btrfs_mapping_tree *tree); 411 void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree); 412 int btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio, 413 int mirror_num, int async_submit); 414 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, 415 fmode_t flags, void *holder); 416 int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder, 417 struct btrfs_fs_devices **fs_devices_ret); 418 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices); 419 void btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices, int step); 420 void btrfs_assign_next_active_device(struct btrfs_fs_info *fs_info, 421 struct btrfs_device *device, struct btrfs_device *this_dev); 422 int btrfs_find_device_missing_or_by_path(struct btrfs_fs_info *fs_info, 423 const char *device_path, 424 struct btrfs_device **device); 425 int btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info, u64 devid, 426 const char *devpath, 427 struct btrfs_device **device); 428 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info, 429 const u64 *devid, 430 const u8 *uuid); 431 int btrfs_rm_device(struct btrfs_fs_info *fs_info, 432 const char *device_path, u64 devid); 433 void btrfs_cleanup_fs_uuids(void); 434 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len); 435 int btrfs_grow_device(struct btrfs_trans_handle *trans, 436 struct btrfs_device *device, u64 new_size); 437 struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid, 438 u8 *uuid, u8 *fsid); 439 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); 440 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path); 441 int btrfs_init_dev_replace_tgtdev(struct btrfs_fs_info *fs_info, 442 const char *device_path, 443 struct btrfs_device *srcdev, 444 struct btrfs_device **device_out); 445 int btrfs_balance(struct btrfs_balance_control *bctl, 446 struct btrfs_ioctl_balance_args *bargs); 447 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info); 448 int btrfs_recover_balance(struct btrfs_fs_info *fs_info); 449 int btrfs_pause_balance(struct btrfs_fs_info *fs_info); 450 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info); 451 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info); 452 int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info); 453 int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset); 454 int find_free_dev_extent_start(struct btrfs_transaction *transaction, 455 struct btrfs_device *device, u64 num_bytes, 456 u64 search_start, u64 *start, u64 *max_avail); 457 int find_free_dev_extent(struct btrfs_trans_handle *trans, 458 struct btrfs_device *device, u64 num_bytes, 459 u64 *start, u64 *max_avail); 460 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index); 461 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info, 462 struct btrfs_ioctl_get_dev_stats *stats); 463 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info); 464 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info); 465 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans, 466 struct btrfs_fs_info *fs_info); 467 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_fs_info *fs_info, 468 struct btrfs_device *srcdev); 469 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_fs_info *fs_info, 470 struct btrfs_device *srcdev); 471 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info, 472 struct btrfs_device *tgtdev); 473 void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info, 474 struct btrfs_device *tgtdev); 475 void btrfs_scratch_superblocks(struct block_device *bdev, const char *device_path); 476 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info, 477 u64 logical, u64 len, int mirror_num); 478 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info, 479 struct btrfs_mapping_tree *map_tree, 480 u64 logical); 481 int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans, 482 struct btrfs_fs_info *fs_info, 483 u64 chunk_offset, u64 chunk_size); 484 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, 485 struct btrfs_fs_info *fs_info, u64 chunk_offset); 486 487 static inline int btrfs_dev_stats_dirty(struct btrfs_device *dev) 488 { 489 return atomic_read(&dev->dev_stats_ccnt); 490 } 491 492 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev, 493 int index) 494 { 495 atomic_inc(dev->dev_stat_values + index); 496 smp_mb__before_atomic(); 497 atomic_inc(&dev->dev_stats_ccnt); 498 } 499 500 static inline int btrfs_dev_stat_read(struct btrfs_device *dev, 501 int index) 502 { 503 return atomic_read(dev->dev_stat_values + index); 504 } 505 506 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev, 507 int index) 508 { 509 int ret; 510 511 ret = atomic_xchg(dev->dev_stat_values + index, 0); 512 smp_mb__before_atomic(); 513 atomic_inc(&dev->dev_stats_ccnt); 514 return ret; 515 } 516 517 static inline void btrfs_dev_stat_set(struct btrfs_device *dev, 518 int index, unsigned long val) 519 { 520 atomic_set(dev->dev_stat_values + index, val); 521 smp_mb__before_atomic(); 522 atomic_inc(&dev->dev_stats_ccnt); 523 } 524 525 static inline void btrfs_dev_stat_reset(struct btrfs_device *dev, 526 int index) 527 { 528 btrfs_dev_stat_set(dev, index, 0); 529 } 530 531 void btrfs_update_commit_device_size(struct btrfs_fs_info *fs_info); 532 void btrfs_update_commit_device_bytes_used(struct btrfs_fs_info *fs_info, 533 struct btrfs_transaction *transaction); 534 535 struct list_head *btrfs_get_fs_uuids(void); 536 void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info); 537 void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info); 538 539 #endif 540