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 list_head post_commit_list; /* chunk mutex */
49 struct btrfs_fs_devices *fs_devices;
50 struct btrfs_fs_info *fs_info;
52 struct rcu_string *name;
56 spinlock_t io_lock ____cacheline_aligned;
58 /* regular prio bios */
59 struct btrfs_pending_bios pending_bios;
61 struct btrfs_pending_bios pending_sync_bios;
63 struct block_device *bdev;
65 /* the mode sent to blkdev_get */
68 unsigned long dev_state;
69 blk_status_t last_flush_error;
72 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
73 seqcount_t data_seqcount;
76 /* the internal btrfs device id */
79 /* size of the device in memory */
82 /* size of the device on disk */
88 /* optimal io alignment for this device */
91 /* optimal io width for this device */
93 /* type and info about this device */
96 /* minimal io size for this device */
99 /* physical drive uuid (or lvm uuid) */
100 u8 uuid[BTRFS_UUID_SIZE];
103 * size of the device on the current transaction
105 * This variant is update when committing the transaction,
106 * and protected by chunk mutex
108 u64 commit_total_bytes;
110 /* bytes used on the current transaction */
111 u64 commit_bytes_used;
113 /* for sending down flush barriers */
114 struct bio *flush_bio;
115 struct completion flush_wait;
117 /* per-device scrub information */
118 struct scrub_ctx *scrub_ctx;
120 struct btrfs_work work;
122 /* readahead state */
123 atomic_t reada_in_flight;
125 struct reada_zone *reada_curr_zone;
126 struct radix_tree_root reada_zones;
127 struct radix_tree_root reada_extents;
129 /* disk I/O failure stats. For detailed description refer to
130 * enum btrfs_dev_stat_values in ioctl.h */
133 /* Counter to record the change of device stats */
134 atomic_t dev_stats_ccnt;
135 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
137 struct extent_io_tree alloc_state;
141 * If we read those variants at the context of their own lock, we needn't
142 * use the following helpers, reading them directly is safe.
144 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
145 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
147 btrfs_device_get_##name(const struct btrfs_device *dev) \
153 seq = read_seqcount_begin(&dev->data_seqcount); \
155 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
160 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
163 write_seqcount_begin(&dev->data_seqcount); \
165 write_seqcount_end(&dev->data_seqcount); \
168 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
169 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
171 btrfs_device_get_##name(const struct btrfs_device *dev) \
182 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
189 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
191 btrfs_device_get_##name(const struct btrfs_device *dev) \
197 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
203 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
204 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
205 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
207 struct btrfs_fs_devices {
208 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
209 u8 metadata_uuid[BTRFS_FSID_SIZE];
211 struct list_head fs_list;
220 /* Highest generation number of seen devices */
221 u64 latest_generation;
223 struct block_device *latest_bdev;
225 /* all of the devices in the FS, protected by a mutex
226 * so we can safely walk it to write out the supers without
227 * worrying about add/remove by the multi-device code.
228 * Scrubbing super can kick off supers writing by holding
231 struct mutex device_list_mutex;
232 struct list_head devices;
234 /* devices not currently being allocated */
235 struct list_head alloc_list;
237 struct btrfs_fs_devices *seed;
242 /* set when we find or add a device that doesn't have the
247 struct btrfs_fs_info *fs_info;
249 struct kobject fsid_kobj;
250 struct kobject *device_dir_kobj;
251 struct completion kobj_unregister;
254 #define BTRFS_BIO_INLINE_CSUM_SIZE 64
256 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \
257 - sizeof(struct btrfs_chunk)) \
258 / sizeof(struct btrfs_stripe) + 1)
260 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \
261 - 2 * sizeof(struct btrfs_disk_key) \
262 - 2 * sizeof(struct btrfs_chunk)) \
263 / sizeof(struct btrfs_stripe) + 1)
266 * we need the mirror number and stripe index to be passed around
267 * the call chain while we are processing end_io (especially errors).
268 * Really, what we need is a btrfs_bio structure that has this info
269 * and is properly sized with its stripe array, but we're not there
270 * quite yet. We have our own btrfs bioset, and all of the bios
271 * we allocate are actually btrfs_io_bios. We'll cram as much of
272 * struct btrfs_bio as we can into this over time.
274 struct btrfs_io_bio {
275 unsigned int mirror_num;
276 unsigned int stripe_index;
279 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
280 struct bvec_iter iter;
282 * This member must come last, bio_alloc_bioset will allocate enough
283 * bytes for entire btrfs_io_bio but relies on bio being last.
288 static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
290 return container_of(bio, struct btrfs_io_bio, bio);
293 static inline void btrfs_io_bio_free_csum(struct btrfs_io_bio *io_bio)
295 if (io_bio->csum != io_bio->csum_inline) {
301 struct btrfs_bio_stripe {
302 struct btrfs_device *dev;
304 u64 length; /* only used for discard mappings */
309 atomic_t stripes_pending;
310 struct btrfs_fs_info *fs_info;
311 u64 map_type; /* get from map_lookup->type */
312 bio_end_io_t *end_io;
313 struct bio *orig_bio;
323 * logical block numbers for the start of each stripe
324 * The last one or two are p/q. These are sorted,
325 * so raid_map[0] is the start of our full stripe
328 struct btrfs_bio_stripe stripes[];
331 struct btrfs_device_info {
332 struct btrfs_device *dev;
338 struct btrfs_raid_attr {
339 int sub_stripes; /* sub_stripes info for map */
340 int dev_stripes; /* stripes per dev */
341 int devs_max; /* max devs to use */
342 int devs_min; /* min devs needed */
343 int tolerated_failures; /* max tolerated fail devs */
344 int devs_increment; /* ndevs has to be a multiple of this */
345 int ncopies; /* how many copies to data has */
346 int nparity; /* number of stripes worth of bytes to store
347 * parity information */
348 int mindev_error; /* error code if min devs requisite is unmet */
349 const char raid_name[8]; /* name of the raid */
350 u64 bg_flag; /* block group flag of the raid */
353 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
362 int verified_stripes; /* For mount time dev extent verification */
363 struct btrfs_bio_stripe stripes[];
366 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
367 (sizeof(struct btrfs_bio_stripe) * (n)))
369 struct btrfs_balance_args;
370 struct btrfs_balance_progress;
371 struct btrfs_balance_control {
372 struct btrfs_balance_args data;
373 struct btrfs_balance_args meta;
374 struct btrfs_balance_args sys;
378 struct btrfs_balance_progress stat;
385 BTRFS_MAP_GET_READ_MIRRORS,
388 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
390 switch (bio_op(bio)) {
392 return BTRFS_MAP_DISCARD;
394 return BTRFS_MAP_WRITE;
399 return BTRFS_MAP_READ;
403 void btrfs_get_bbio(struct btrfs_bio *bbio);
404 void btrfs_put_bbio(struct btrfs_bio *bbio);
405 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
406 u64 logical, u64 *length,
407 struct btrfs_bio **bbio_ret, int mirror_num);
408 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
409 u64 logical, u64 *length,
410 struct btrfs_bio **bbio_ret);
411 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
412 u64 physical, u64 **logical, int *naddrs, int *stripe_len);
413 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
414 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
415 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 type);
416 void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
417 void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
418 blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
419 int mirror_num, int async_submit);
420 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
421 fmode_t flags, void *holder);
422 struct btrfs_device *btrfs_scan_one_device(const char *path,
423 fmode_t flags, void *holder);
424 int btrfs_forget_devices(const char *path);
425 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
426 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices, int step);
427 void btrfs_assign_next_active_device(struct btrfs_device *device,
428 struct btrfs_device *this_dev);
429 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
431 const char *devpath);
432 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
435 void btrfs_free_device(struct btrfs_device *device);
436 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
437 const char *device_path, u64 devid);
438 void __exit btrfs_cleanup_fs_uuids(void);
439 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
440 int btrfs_grow_device(struct btrfs_trans_handle *trans,
441 struct btrfs_device *device, u64 new_size);
442 struct btrfs_device *btrfs_find_device(struct btrfs_fs_devices *fs_devices,
443 u64 devid, u8 *uuid, u8 *fsid, bool seed);
444 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
445 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
446 int btrfs_balance(struct btrfs_fs_info *fs_info,
447 struct btrfs_balance_control *bctl,
448 struct btrfs_ioctl_balance_args *bargs);
449 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
450 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
451 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
452 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
453 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
454 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
455 int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
456 int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset);
457 int find_free_dev_extent_start(struct btrfs_device *device, u64 num_bytes,
458 u64 search_start, u64 *start, u64 *max_avail);
459 int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes,
460 u64 *start, u64 *max_avail);
461 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
462 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
463 struct btrfs_ioctl_get_dev_stats *stats);
464 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
465 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
466 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
467 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
468 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
469 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
470 void btrfs_scratch_superblocks(struct block_device *bdev, const char *device_path);
471 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
472 u64 logical, u64 len);
473 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
475 int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
476 u64 chunk_offset, u64 chunk_size);
477 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
478 struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
479 u64 logical, u64 length);
481 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
484 atomic_inc(dev->dev_stat_values + index);
486 * This memory barrier orders stores updating statistics before stores
487 * updating dev_stats_ccnt.
489 * It pairs with smp_rmb() in btrfs_run_dev_stats().
491 smp_mb__before_atomic();
492 atomic_inc(&dev->dev_stats_ccnt);
495 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
498 return atomic_read(dev->dev_stat_values + index);
501 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
506 ret = atomic_xchg(dev->dev_stat_values + index, 0);
508 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
509 * - RMW operations that have a return value are fully ordered;
511 * This implicit memory barriers is paired with the smp_rmb in
512 * btrfs_run_dev_stats
514 atomic_inc(&dev->dev_stats_ccnt);
518 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
519 int index, unsigned long val)
521 atomic_set(dev->dev_stat_values + index, val);
523 * This memory barrier orders stores updating statistics before stores
524 * updating dev_stats_ccnt.
526 * It pairs with smp_rmb() in btrfs_run_dev_stats().
528 smp_mb__before_atomic();
529 atomic_inc(&dev->dev_stats_ccnt);
532 static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
535 btrfs_dev_stat_set(dev, index, 0);
539 * Convert block group flags (BTRFS_BLOCK_GROUP_*) to btrfs_raid_types, which
540 * can be used as index to access btrfs_raid_array[].
542 static inline enum btrfs_raid_types btrfs_bg_flags_to_raid_index(u64 flags)
544 if (flags & BTRFS_BLOCK_GROUP_RAID10)
545 return BTRFS_RAID_RAID10;
546 else if (flags & BTRFS_BLOCK_GROUP_RAID1)
547 return BTRFS_RAID_RAID1;
548 else if (flags & BTRFS_BLOCK_GROUP_DUP)
549 return BTRFS_RAID_DUP;
550 else if (flags & BTRFS_BLOCK_GROUP_RAID0)
551 return BTRFS_RAID_RAID0;
552 else if (flags & BTRFS_BLOCK_GROUP_RAID5)
553 return BTRFS_RAID_RAID5;
554 else if (flags & BTRFS_BLOCK_GROUP_RAID6)
555 return BTRFS_RAID_RAID6;
557 return BTRFS_RAID_SINGLE; /* BTRFS_BLOCK_GROUP_SINGLE */
560 const char *get_raid_name(enum btrfs_raid_types type);
562 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
564 struct list_head *btrfs_get_fs_uuids(void);
565 void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info);
566 void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info);
567 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
568 struct btrfs_device *failing_dev);
570 int btrfs_bg_type_to_factor(u64 flags);
571 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);