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 #define BTRFS_MAX_DATA_CHUNK_SIZE (10ULL * SZ_1G)
16 extern struct mutex uuid_mutex;
18 #define BTRFS_STRIPE_LEN SZ_64K
21 struct btrfs_pending_bios {
27 * Use sequence counter to get consistent device stat data on
30 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
31 #include <linux/seqlock.h>
32 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
33 #define btrfs_device_data_ordered_init(device) \
34 seqcount_init(&device->data_seqcount)
36 #define btrfs_device_data_ordered_init(device) do { } while (0)
39 #define BTRFS_DEV_STATE_WRITEABLE (0)
40 #define BTRFS_DEV_STATE_IN_FS_METADATA (1)
41 #define BTRFS_DEV_STATE_MISSING (2)
42 #define BTRFS_DEV_STATE_REPLACE_TGT (3)
43 #define BTRFS_DEV_STATE_FLUSH_SENT (4)
46 struct list_head dev_list;
47 struct list_head dev_alloc_list;
48 struct btrfs_fs_devices *fs_devices;
49 struct btrfs_fs_info *fs_info;
51 struct rcu_string *name;
55 spinlock_t io_lock ____cacheline_aligned;
57 /* regular prio bios */
58 struct btrfs_pending_bios pending_bios;
60 struct btrfs_pending_bios pending_sync_bios;
62 struct block_device *bdev;
64 /* the mode sent to blkdev_get */
67 unsigned long dev_state;
68 blk_status_t last_flush_error;
71 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
72 seqcount_t data_seqcount;
75 /* the internal btrfs device id */
78 /* size of the device in memory */
81 /* size of the device on disk */
87 /* optimal io alignment for this device */
90 /* optimal io width for this device */
92 /* type and info about this device */
95 /* minimal io size for this device */
98 /* physical drive uuid (or lvm uuid) */
99 u8 uuid[BTRFS_UUID_SIZE];
102 * size of the device on the current transaction
104 * This variant is update when committing the transaction,
105 * and protected by device_list_mutex
107 u64 commit_total_bytes;
109 /* bytes used on the current transaction */
110 u64 commit_bytes_used;
112 * used to manage the device which is resized
114 * It is protected by chunk_lock.
116 struct list_head resized_list;
118 /* for sending down flush barriers */
119 struct bio *flush_bio;
120 struct completion flush_wait;
122 /* per-device scrub information */
123 struct scrub_ctx *scrub_ctx;
125 struct btrfs_work work;
128 /* readahead state */
129 atomic_t reada_in_flight;
131 struct reada_zone *reada_curr_zone;
132 struct radix_tree_root reada_zones;
133 struct radix_tree_root reada_extents;
135 /* disk I/O failure stats. For detailed description refer to
136 * enum btrfs_dev_stat_values in ioctl.h */
139 /* Counter to record the change of device stats */
140 atomic_t dev_stats_ccnt;
141 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
145 * If we read those variants at the context of their own lock, we needn't
146 * use the following helpers, reading them directly is safe.
148 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
149 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
151 btrfs_device_get_##name(const struct btrfs_device *dev) \
157 seq = read_seqcount_begin(&dev->data_seqcount); \
159 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
164 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
167 write_seqcount_begin(&dev->data_seqcount); \
169 write_seqcount_end(&dev->data_seqcount); \
172 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
173 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
175 btrfs_device_get_##name(const struct btrfs_device *dev) \
186 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
193 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
195 btrfs_device_get_##name(const struct btrfs_device *dev) \
201 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
207 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
208 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
209 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
211 struct btrfs_fs_devices {
212 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
213 struct list_head fs_list;
221 struct block_device *latest_bdev;
223 /* all of the devices in the FS, protected by a mutex
224 * so we can safely walk it to write out the supers without
225 * worrying about add/remove by the multi-device code.
226 * Scrubbing super can kick off supers writing by holding
229 struct mutex device_list_mutex;
230 struct list_head devices;
232 struct list_head resized_devices;
233 /* devices not currently being allocated */
234 struct list_head alloc_list;
236 struct btrfs_fs_devices *seed;
241 /* set when we find or add a device that doesn't have the
246 struct btrfs_fs_info *fs_info;
248 struct kobject fsid_kobj;
249 struct kobject *device_dir_kobj;
250 struct completion kobj_unregister;
253 #define BTRFS_BIO_INLINE_CSUM_SIZE 64
256 * we need the mirror number and stripe index to be passed around
257 * the call chain while we are processing end_io (especially errors).
258 * Really, what we need is a btrfs_bio structure that has this info
259 * and is properly sized with its stripe array, but we're not there
260 * quite yet. We have our own btrfs bioset, and all of the bios
261 * we allocate are actually btrfs_io_bios. We'll cram as much of
262 * struct btrfs_bio as we can into this over time.
264 typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err);
265 struct btrfs_io_bio {
266 unsigned int mirror_num;
267 unsigned int stripe_index;
270 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
272 btrfs_io_bio_end_io_t *end_io;
273 struct bvec_iter iter;
275 * This member must come last, bio_alloc_bioset will allocate enough
276 * bytes for entire btrfs_io_bio but relies on bio being last.
281 static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
283 return container_of(bio, struct btrfs_io_bio, bio);
286 struct btrfs_bio_stripe {
287 struct btrfs_device *dev;
289 u64 length; /* only used for discard mappings */
293 typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err);
297 atomic_t stripes_pending;
298 struct btrfs_fs_info *fs_info;
299 u64 map_type; /* get from map_lookup->type */
300 bio_end_io_t *end_io;
301 struct bio *orig_bio;
311 * logical block numbers for the start of each stripe
312 * The last one or two are p/q. These are sorted,
313 * so raid_map[0] is the start of our full stripe
316 struct btrfs_bio_stripe stripes[];
319 struct btrfs_device_info {
320 struct btrfs_device *dev;
326 struct btrfs_raid_attr {
327 int sub_stripes; /* sub_stripes info for map */
328 int dev_stripes; /* stripes per dev */
329 int devs_max; /* max devs to use */
330 int devs_min; /* min devs needed */
331 int tolerated_failures; /* max tolerated fail devs */
332 int devs_increment; /* ndevs has to be a multiple of this */
333 int ncopies; /* how many copies to data has */
334 int mindev_error; /* error code if min devs requisite is unmet */
335 const char raid_name[8]; /* name of the raid */
336 u64 bg_flag; /* block group flag of the raid */
339 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
348 struct btrfs_bio_stripe stripes[];
351 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
352 (sizeof(struct btrfs_bio_stripe) * (n)))
354 struct btrfs_balance_args;
355 struct btrfs_balance_progress;
356 struct btrfs_balance_control {
357 struct btrfs_balance_args data;
358 struct btrfs_balance_args meta;
359 struct btrfs_balance_args sys;
363 struct btrfs_balance_progress stat;
370 BTRFS_MAP_GET_READ_MIRRORS,
373 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
375 switch (bio_op(bio)) {
377 return BTRFS_MAP_DISCARD;
379 return BTRFS_MAP_WRITE;
383 return BTRFS_MAP_READ;
387 void btrfs_get_bbio(struct btrfs_bio *bbio);
388 void btrfs_put_bbio(struct btrfs_bio *bbio);
389 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
390 u64 logical, u64 *length,
391 struct btrfs_bio **bbio_ret, int mirror_num);
392 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
393 u64 logical, u64 *length,
394 struct btrfs_bio **bbio_ret);
395 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
396 u64 physical, u64 **logical, int *naddrs, int *stripe_len);
397 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
398 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
399 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 type);
400 void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
401 void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
402 blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
403 int mirror_num, int async_submit);
404 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
405 fmode_t flags, void *holder);
406 struct btrfs_device *btrfs_scan_one_device(const char *path,
407 fmode_t flags, void *holder);
408 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
409 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices, int step);
410 void btrfs_assign_next_active_device(struct btrfs_fs_info *fs_info,
411 struct btrfs_device *device, struct btrfs_device *this_dev);
412 int btrfs_find_device_missing_or_by_path(struct btrfs_fs_info *fs_info,
413 const char *device_path,
414 struct btrfs_device **device);
415 int btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info, u64 devid,
417 struct btrfs_device **device);
418 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
421 void btrfs_free_device(struct btrfs_device *device);
422 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
423 const char *device_path, u64 devid);
424 void __exit btrfs_cleanup_fs_uuids(void);
425 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
426 int btrfs_grow_device(struct btrfs_trans_handle *trans,
427 struct btrfs_device *device, u64 new_size);
428 struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
430 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
431 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
432 int btrfs_balance(struct btrfs_fs_info *fs_info,
433 struct btrfs_balance_control *bctl,
434 struct btrfs_ioctl_balance_args *bargs);
435 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
436 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
437 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
438 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
439 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
440 int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
441 int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset);
442 int find_free_dev_extent_start(struct btrfs_transaction *transaction,
443 struct btrfs_device *device, u64 num_bytes,
444 u64 search_start, u64 *start, u64 *max_avail);
445 int find_free_dev_extent(struct btrfs_trans_handle *trans,
446 struct btrfs_device *device, u64 num_bytes,
447 u64 *start, u64 *max_avail);
448 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
449 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
450 struct btrfs_ioctl_get_dev_stats *stats);
451 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
452 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
453 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
454 struct btrfs_fs_info *fs_info);
455 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
456 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_fs_info *fs_info,
457 struct btrfs_device *srcdev);
458 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
459 struct btrfs_device *tgtdev);
460 void btrfs_scratch_superblocks(struct block_device *bdev, const char *device_path);
461 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
462 u64 logical, u64 len);
463 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
465 int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
466 struct btrfs_fs_info *fs_info,
467 u64 chunk_offset, u64 chunk_size);
468 int btrfs_remove_chunk(struct btrfs_trans_handle *trans,
469 struct btrfs_fs_info *fs_info, u64 chunk_offset);
471 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
474 atomic_inc(dev->dev_stat_values + index);
476 * This memory barrier orders stores updating statistics before stores
477 * updating dev_stats_ccnt.
479 * It pairs with smp_rmb() in btrfs_run_dev_stats().
481 smp_mb__before_atomic();
482 atomic_inc(&dev->dev_stats_ccnt);
485 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
488 return atomic_read(dev->dev_stat_values + index);
491 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
496 ret = atomic_xchg(dev->dev_stat_values + index, 0);
498 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
499 * - RMW operations that have a return value are fully ordered;
501 * This implicit memory barriers is paired with the smp_rmb in
502 * btrfs_run_dev_stats
504 atomic_inc(&dev->dev_stats_ccnt);
508 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
509 int index, unsigned long val)
511 atomic_set(dev->dev_stat_values + index, val);
513 * This memory barrier orders stores updating statistics before stores
514 * updating dev_stats_ccnt.
516 * It pairs with smp_rmb() in btrfs_run_dev_stats().
518 smp_mb__before_atomic();
519 atomic_inc(&dev->dev_stats_ccnt);
522 static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
525 btrfs_dev_stat_set(dev, index, 0);
529 * Convert block group flags (BTRFS_BLOCK_GROUP_*) to btrfs_raid_types, which
530 * can be used as index to access btrfs_raid_array[].
532 static inline enum btrfs_raid_types btrfs_bg_flags_to_raid_index(u64 flags)
534 if (flags & BTRFS_BLOCK_GROUP_RAID10)
535 return BTRFS_RAID_RAID10;
536 else if (flags & BTRFS_BLOCK_GROUP_RAID1)
537 return BTRFS_RAID_RAID1;
538 else if (flags & BTRFS_BLOCK_GROUP_DUP)
539 return BTRFS_RAID_DUP;
540 else if (flags & BTRFS_BLOCK_GROUP_RAID0)
541 return BTRFS_RAID_RAID0;
542 else if (flags & BTRFS_BLOCK_GROUP_RAID5)
543 return BTRFS_RAID_RAID5;
544 else if (flags & BTRFS_BLOCK_GROUP_RAID6)
545 return BTRFS_RAID_RAID6;
547 return BTRFS_RAID_SINGLE; /* BTRFS_BLOCK_GROUP_SINGLE */
550 const char *get_raid_name(enum btrfs_raid_types type);
552 void btrfs_update_commit_device_size(struct btrfs_fs_info *fs_info);
553 void btrfs_update_commit_device_bytes_used(struct btrfs_transaction *trans);
555 struct list_head *btrfs_get_fs_uuids(void);
556 void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info);
557 void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info);
558 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
559 struct btrfs_device *failing_dev);
561 int btrfs_bg_type_to_factor(u64 flags);