1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #ifndef BTRFS_VOLUMES_H
7 #define BTRFS_VOLUMES_H
10 #include <linux/sort.h>
11 #include <linux/btrfs.h>
12 #include "async-thread.h"
14 extern struct mutex uuid_mutex;
16 #define BTRFS_STRIPE_LEN SZ_64K
19 struct btrfs_pending_bios {
25 * Use sequence counter to get consistent device stat data on
28 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
29 #include <linux/seqlock.h>
30 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
31 #define btrfs_device_data_ordered_init(device) \
32 seqcount_init(&device->data_seqcount)
34 #define btrfs_device_data_ordered_init(device) do { } while (0)
37 #define BTRFS_DEV_STATE_WRITEABLE (0)
38 #define BTRFS_DEV_STATE_IN_FS_METADATA (1)
39 #define BTRFS_DEV_STATE_MISSING (2)
40 #define BTRFS_DEV_STATE_REPLACE_TGT (3)
41 #define BTRFS_DEV_STATE_FLUSH_SENT (4)
44 struct list_head dev_list;
45 struct list_head dev_alloc_list;
46 struct btrfs_fs_devices *fs_devices;
47 struct btrfs_fs_info *fs_info;
49 struct rcu_string *name;
53 spinlock_t io_lock ____cacheline_aligned;
55 /* regular prio bios */
56 struct btrfs_pending_bios pending_bios;
58 struct btrfs_pending_bios pending_sync_bios;
60 struct block_device *bdev;
62 /* the mode sent to blkdev_get */
65 unsigned long dev_state;
66 blk_status_t last_flush_error;
69 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
70 seqcount_t data_seqcount;
73 /* the internal btrfs device id */
76 /* size of the device in memory */
79 /* size of the device on disk */
85 /* optimal io alignment for this device */
88 /* optimal io width for this device */
90 /* type and info about this device */
93 /* minimal io size for this device */
96 /* physical drive uuid (or lvm uuid) */
97 u8 uuid[BTRFS_UUID_SIZE];
100 * size of the device on the current transaction
102 * This variant is update when committing the transaction,
103 * and protected by device_list_mutex
105 u64 commit_total_bytes;
107 /* bytes used on the current transaction */
108 u64 commit_bytes_used;
110 * used to manage the device which is resized
112 * It is protected by chunk_lock.
114 struct list_head resized_list;
116 /* for sending down flush barriers */
117 struct bio *flush_bio;
118 struct completion flush_wait;
120 /* per-device scrub information */
121 struct scrub_ctx *scrub_ctx;
123 struct btrfs_work work;
126 /* readahead state */
127 atomic_t reada_in_flight;
129 struct reada_zone *reada_curr_zone;
130 struct radix_tree_root reada_zones;
131 struct radix_tree_root reada_extents;
133 /* disk I/O failure stats. For detailed description refer to
134 * enum btrfs_dev_stat_values in ioctl.h */
137 /* Counter to record the change of device stats */
138 atomic_t dev_stats_ccnt;
139 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
143 * If we read those variants at the context of their own lock, we needn't
144 * use the following helpers, reading them directly is safe.
146 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
147 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
149 btrfs_device_get_##name(const struct btrfs_device *dev) \
155 seq = read_seqcount_begin(&dev->data_seqcount); \
157 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
162 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
165 write_seqcount_begin(&dev->data_seqcount); \
167 write_seqcount_end(&dev->data_seqcount); \
170 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
171 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
173 btrfs_device_get_##name(const struct btrfs_device *dev) \
184 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
191 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
193 btrfs_device_get_##name(const struct btrfs_device *dev) \
199 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
205 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
206 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
207 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
209 struct btrfs_fs_devices {
210 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
211 struct list_head fs_list;
219 struct block_device *latest_bdev;
221 /* all of the devices in the FS, protected by a mutex
222 * so we can safely walk it to write out the supers without
223 * worrying about add/remove by the multi-device code.
224 * Scrubbing super can kick off supers writing by holding
227 struct mutex device_list_mutex;
228 struct list_head devices;
230 struct list_head resized_devices;
231 /* devices not currently being allocated */
232 struct list_head alloc_list;
234 struct btrfs_fs_devices *seed;
239 /* set when we find or add a device that doesn't have the
244 struct btrfs_fs_info *fs_info;
246 struct kobject fsid_kobj;
247 struct kobject *device_dir_kobj;
248 struct completion kobj_unregister;
251 #define BTRFS_BIO_INLINE_CSUM_SIZE 64
254 * we need the mirror number and stripe index to be passed around
255 * the call chain while we are processing end_io (especially errors).
256 * Really, what we need is a btrfs_bio structure that has this info
257 * and is properly sized with its stripe array, but we're not there
258 * quite yet. We have our own btrfs bioset, and all of the bios
259 * we allocate are actually btrfs_io_bios. We'll cram as much of
260 * struct btrfs_bio as we can into this over time.
262 typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err);
263 struct btrfs_io_bio {
264 unsigned int mirror_num;
265 unsigned int stripe_index;
268 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
270 btrfs_io_bio_end_io_t *end_io;
271 struct bvec_iter iter;
273 * This member must come last, bio_alloc_bioset will allocate enough
274 * bytes for entire btrfs_io_bio but relies on bio being last.
279 static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
281 return container_of(bio, struct btrfs_io_bio, bio);
284 struct btrfs_bio_stripe {
285 struct btrfs_device *dev;
287 u64 length; /* only used for discard mappings */
291 typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err);
295 atomic_t stripes_pending;
296 struct btrfs_fs_info *fs_info;
297 u64 map_type; /* get from map_lookup->type */
298 bio_end_io_t *end_io;
299 struct bio *orig_bio;
309 * logical block numbers for the start of each stripe
310 * The last one or two are p/q. These are sorted,
311 * so raid_map[0] is the start of our full stripe
314 struct btrfs_bio_stripe stripes[];
317 struct btrfs_device_info {
318 struct btrfs_device *dev;
324 struct btrfs_raid_attr {
325 int sub_stripes; /* sub_stripes info for map */
326 int dev_stripes; /* stripes per dev */
327 int devs_max; /* max devs to use */
328 int devs_min; /* min devs needed */
329 int tolerated_failures; /* max tolerated fail devs */
330 int devs_increment; /* ndevs has to be a multiple of this */
331 int ncopies; /* how many copies to data has */
334 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
335 extern const int btrfs_raid_mindev_error[BTRFS_NR_RAID_TYPES];
336 extern const u64 btrfs_raid_group[BTRFS_NR_RAID_TYPES];
345 struct btrfs_bio_stripe stripes[];
348 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
349 (sizeof(struct btrfs_bio_stripe) * (n)))
351 struct btrfs_balance_args;
352 struct btrfs_balance_progress;
353 struct btrfs_balance_control {
354 struct btrfs_fs_info *fs_info;
356 struct btrfs_balance_args data;
357 struct btrfs_balance_args meta;
358 struct btrfs_balance_args sys;
362 struct btrfs_balance_progress stat;
369 BTRFS_MAP_GET_READ_MIRRORS,
372 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
374 switch (bio_op(bio)) {
376 return BTRFS_MAP_DISCARD;
378 return BTRFS_MAP_WRITE;
382 return BTRFS_MAP_READ;
386 int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
387 u64 end, u64 *length);
388 void btrfs_get_bbio(struct btrfs_bio *bbio);
389 void btrfs_put_bbio(struct btrfs_bio *bbio);
390 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
391 u64 logical, u64 *length,
392 struct btrfs_bio **bbio_ret, int mirror_num);
393 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
394 u64 logical, u64 *length,
395 struct btrfs_bio **bbio_ret);
396 int btrfs_rmap_block(struct btrfs_fs_info *fs_info,
397 u64 chunk_start, u64 physical, u64 devid,
398 u64 **logical, int *naddrs, int *stripe_len);
399 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
400 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
401 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
402 struct btrfs_fs_info *fs_info, u64 type);
403 void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
404 void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
405 blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
406 int mirror_num, int async_submit);
407 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
408 fmode_t flags, void *holder);
409 int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
410 struct btrfs_fs_devices **fs_devices_ret);
411 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
412 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices, int step);
413 void btrfs_assign_next_active_device(struct btrfs_fs_info *fs_info,
414 struct btrfs_device *device, struct btrfs_device *this_dev);
415 int btrfs_find_device_missing_or_by_path(struct btrfs_fs_info *fs_info,
416 const char *device_path,
417 struct btrfs_device **device);
418 int btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info, u64 devid,
420 struct btrfs_device **device);
421 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
424 void btrfs_free_device(struct btrfs_device *device);
425 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
426 const char *device_path, u64 devid);
427 void __exit btrfs_cleanup_fs_uuids(void);
428 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
429 int btrfs_grow_device(struct btrfs_trans_handle *trans,
430 struct btrfs_device *device, u64 new_size);
431 struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
433 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
434 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
435 int btrfs_init_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
436 const char *device_path,
437 struct btrfs_device *srcdev,
438 struct btrfs_device **device_out);
439 int btrfs_balance(struct btrfs_balance_control *bctl,
440 struct btrfs_ioctl_balance_args *bargs);
441 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
442 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
443 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
444 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
445 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
446 int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
447 int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset);
448 int find_free_dev_extent_start(struct btrfs_transaction *transaction,
449 struct btrfs_device *device, u64 num_bytes,
450 u64 search_start, u64 *start, u64 *max_avail);
451 int find_free_dev_extent(struct btrfs_trans_handle *trans,
452 struct btrfs_device *device, u64 num_bytes,
453 u64 *start, u64 *max_avail);
454 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
455 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
456 struct btrfs_ioctl_get_dev_stats *stats);
457 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
458 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
459 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
460 struct btrfs_fs_info *fs_info);
461 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_fs_info *fs_info,
462 struct btrfs_device *srcdev);
463 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_fs_info *fs_info,
464 struct btrfs_device *srcdev);
465 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
466 struct btrfs_device *tgtdev);
467 void btrfs_scratch_superblocks(struct block_device *bdev, const char *device_path);
468 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
469 u64 logical, u64 len);
470 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
472 int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
473 struct btrfs_fs_info *fs_info,
474 u64 chunk_offset, u64 chunk_size);
475 int btrfs_remove_chunk(struct btrfs_trans_handle *trans,
476 struct btrfs_fs_info *fs_info, u64 chunk_offset);
478 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
481 atomic_inc(dev->dev_stat_values + index);
483 * This memory barrier orders stores updating statistics before stores
484 * updating dev_stats_ccnt.
486 * It pairs with smp_rmb() in btrfs_run_dev_stats().
488 smp_mb__before_atomic();
489 atomic_inc(&dev->dev_stats_ccnt);
492 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
495 return atomic_read(dev->dev_stat_values + index);
498 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
503 ret = atomic_xchg(dev->dev_stat_values + index, 0);
505 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
506 * - RMW operations that have a return value are fully ordered;
508 * This implicit memory barriers is paired with the smp_rmb in
509 * btrfs_run_dev_stats
511 atomic_inc(&dev->dev_stats_ccnt);
515 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
516 int index, unsigned long val)
518 atomic_set(dev->dev_stat_values + index, val);
520 * This memory barrier orders stores updating statistics before stores
521 * updating dev_stats_ccnt.
523 * It pairs with smp_rmb() in btrfs_run_dev_stats().
525 smp_mb__before_atomic();
526 atomic_inc(&dev->dev_stats_ccnt);
529 static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
532 btrfs_dev_stat_set(dev, index, 0);
536 * Convert block group flags (BTRFS_BLOCK_GROUP_*) to btrfs_raid_types, which
537 * can be used as index to access btrfs_raid_array[].
539 static inline enum btrfs_raid_types btrfs_bg_flags_to_raid_index(u64 flags)
541 if (flags & BTRFS_BLOCK_GROUP_RAID10)
542 return BTRFS_RAID_RAID10;
543 else if (flags & BTRFS_BLOCK_GROUP_RAID1)
544 return BTRFS_RAID_RAID1;
545 else if (flags & BTRFS_BLOCK_GROUP_DUP)
546 return BTRFS_RAID_DUP;
547 else if (flags & BTRFS_BLOCK_GROUP_RAID0)
548 return BTRFS_RAID_RAID0;
549 else if (flags & BTRFS_BLOCK_GROUP_RAID5)
550 return BTRFS_RAID_RAID5;
551 else if (flags & BTRFS_BLOCK_GROUP_RAID6)
552 return BTRFS_RAID_RAID6;
554 return BTRFS_RAID_SINGLE; /* BTRFS_BLOCK_GROUP_SINGLE */
557 void btrfs_update_commit_device_size(struct btrfs_fs_info *fs_info);
558 void btrfs_update_commit_device_bytes_used(struct btrfs_transaction *trans);
560 struct list_head *btrfs_get_fs_uuids(void);
561 void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info);
562 void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info);
563 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
564 struct btrfs_device *failing_dev);