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
20 struct btrfs_io_geometry {
21 /* remaining bytes before crossing a stripe */
23 /* offset of logical address in chunk */
25 /* length of single IO stripe */
27 /* number of stripe where address falls */
29 /* offset of address in stripe */
31 /* offset of raid56 stripe into the chunk */
32 u64 raid56_stripe_offset;
36 * Use sequence counter to get consistent device stat data on
39 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
40 #include <linux/seqlock.h>
41 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
42 #define btrfs_device_data_ordered_init(device) \
43 seqcount_init(&device->data_seqcount)
45 #define btrfs_device_data_ordered_init(device) do { } while (0)
48 #define BTRFS_DEV_STATE_WRITEABLE (0)
49 #define BTRFS_DEV_STATE_IN_FS_METADATA (1)
50 #define BTRFS_DEV_STATE_MISSING (2)
51 #define BTRFS_DEV_STATE_REPLACE_TGT (3)
52 #define BTRFS_DEV_STATE_FLUSH_SENT (4)
53 #define BTRFS_DEV_STATE_NO_READA (5)
55 struct btrfs_zoned_device_info;
58 struct list_head dev_list; /* device_list_mutex */
59 struct list_head dev_alloc_list; /* chunk mutex */
60 struct list_head post_commit_list; /* chunk mutex */
61 struct btrfs_fs_devices *fs_devices;
62 struct btrfs_fs_info *fs_info;
64 struct rcu_string __rcu *name;
68 struct block_device *bdev;
70 struct btrfs_zoned_device_info *zone_info;
72 /* the mode sent to blkdev_get */
76 * Device's major-minor number. Must be set even if the device is not
77 * opened (bdev == NULL), unless the device is missing.
80 unsigned long dev_state;
81 blk_status_t last_flush_error;
83 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
84 seqcount_t data_seqcount;
87 /* the internal btrfs device id */
90 /* size of the device in memory */
93 /* size of the device on disk */
99 /* optimal io alignment for this device */
102 /* optimal io width for this device */
104 /* type and info about this device */
107 /* minimal io size for this device */
110 /* physical drive uuid (or lvm uuid) */
111 u8 uuid[BTRFS_UUID_SIZE];
114 * size of the device on the current transaction
116 * This variant is update when committing the transaction,
117 * and protected by chunk mutex
119 u64 commit_total_bytes;
121 /* bytes used on the current transaction */
122 u64 commit_bytes_used;
124 /* Bio used for flushing device barriers */
125 struct bio flush_bio;
126 struct completion flush_wait;
128 /* per-device scrub information */
129 struct scrub_ctx *scrub_ctx;
131 /* disk I/O failure stats. For detailed description refer to
132 * enum btrfs_dev_stat_values in ioctl.h */
135 /* Counter to record the change of device stats */
136 atomic_t dev_stats_ccnt;
137 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
139 struct extent_io_tree alloc_state;
141 struct completion kobj_unregister;
142 /* For sysfs/FSID/devinfo/devid/ */
143 struct kobject devid_kobj;
145 /* Bandwidth limit for scrub, in bytes */
150 * If we read those variants at the context of their own lock, we needn't
151 * use the following helpers, reading them directly is safe.
153 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
154 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
156 btrfs_device_get_##name(const struct btrfs_device *dev) \
162 seq = read_seqcount_begin(&dev->data_seqcount); \
164 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
169 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
172 write_seqcount_begin(&dev->data_seqcount); \
174 write_seqcount_end(&dev->data_seqcount); \
177 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
178 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
180 btrfs_device_get_##name(const struct btrfs_device *dev) \
191 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
198 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
200 btrfs_device_get_##name(const struct btrfs_device *dev) \
206 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
212 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
213 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
214 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
216 enum btrfs_chunk_allocation_policy {
217 BTRFS_CHUNK_ALLOC_REGULAR,
218 BTRFS_CHUNK_ALLOC_ZONED,
222 * Read policies for mirrored block group profiles, read picks the stripe based
225 enum btrfs_read_policy {
226 /* Use process PID to choose the stripe */
227 BTRFS_READ_POLICY_PID,
228 BTRFS_NR_READ_POLICY,
231 struct btrfs_fs_devices {
232 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
233 u8 metadata_uuid[BTRFS_FSID_SIZE];
235 struct list_head fs_list;
238 * Number of devices under this fsid including missing and
239 * replace-target device and excludes seed devices.
244 * The number of devices that successfully opened, including
245 * replace-target, excludes seed devices.
249 /* The number of devices that are under the chunk allocation list. */
252 /* Count of missing devices under this fsid excluding seed device. */
257 * Count of devices from btrfs_super_block::num_devices for this fsid,
258 * which includes the seed device, excludes the transient replace-target
263 /* Highest generation number of seen devices */
264 u64 latest_generation;
267 * The mount device or a device with highest generation after removal
270 struct btrfs_device *latest_dev;
272 /* all of the devices in the FS, protected by a mutex
273 * so we can safely walk it to write out the supers without
274 * worrying about add/remove by the multi-device code.
275 * Scrubbing super can kick off supers writing by holding
278 struct mutex device_list_mutex;
280 /* List of all devices, protected by device_list_mutex */
281 struct list_head devices;
284 * Devices which can satisfy space allocation. Protected by
287 struct list_head alloc_list;
289 struct list_head seed_list;
294 /* set when we find or add a device that doesn't have the
299 struct btrfs_fs_info *fs_info;
301 struct kobject fsid_kobj;
302 struct kobject *devices_kobj;
303 struct kobject *devinfo_kobj;
304 struct completion kobj_unregister;
306 enum btrfs_chunk_allocation_policy chunk_alloc_policy;
308 /* Policy used to read the mirrored stripes */
309 enum btrfs_read_policy read_policy;
312 #define BTRFS_BIO_INLINE_CSUM_SIZE 64
314 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \
315 - sizeof(struct btrfs_chunk)) \
316 / sizeof(struct btrfs_stripe) + 1)
318 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \
319 - 2 * sizeof(struct btrfs_disk_key) \
320 - 2 * sizeof(struct btrfs_chunk)) \
321 / sizeof(struct btrfs_stripe) + 1)
324 * Additional info to pass along bio.
326 * Mostly for btrfs specific features like csum and mirror_num.
329 unsigned int mirror_num;
334 /* @device is for stripe IO submission. */
335 struct btrfs_device *device;
337 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
338 struct bvec_iter iter;
341 * This member must come last, bio_alloc_bioset will allocate enough
342 * bytes for entire btrfs_bio but relies on bio being last.
347 static inline struct btrfs_bio *btrfs_bio(struct bio *bio)
349 return container_of(bio, struct btrfs_bio, bio);
352 static inline void btrfs_bio_free_csum(struct btrfs_bio *bbio)
354 if (bbio->csum != bbio->csum_inline) {
360 struct btrfs_io_stripe {
361 struct btrfs_device *dev;
363 u64 length; /* only used for discard mappings */
367 * Context for IO subsmission for device stripe.
369 * - Track the unfinished mirrors for mirror based profiles
370 * Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
372 * - Contain the logical -> physical mapping info
373 * Used by submit_stripe_bio() for mapping logical bio
374 * into physical device address.
376 * - Contain device replace info
377 * Used by handle_ops_on_dev_replace() to copy logical bios
378 * into the new device.
380 * - Contain RAID56 full stripe logical bytenrs
382 struct btrfs_io_context {
384 atomic_t stripes_pending;
385 struct btrfs_fs_info *fs_info;
386 u64 map_type; /* get from map_lookup->type */
387 bio_end_io_t *end_io;
388 struct bio *orig_bio;
397 * logical block numbers for the start of each stripe
398 * The last one or two are p/q. These are sorted,
399 * so raid_map[0] is the start of our full stripe
402 struct btrfs_io_stripe stripes[];
405 struct btrfs_device_info {
406 struct btrfs_device *dev;
412 struct btrfs_raid_attr {
413 u8 sub_stripes; /* sub_stripes info for map */
414 u8 dev_stripes; /* stripes per dev */
415 u8 devs_max; /* max devs to use */
416 u8 devs_min; /* min devs needed */
417 u8 tolerated_failures; /* max tolerated fail devs */
418 u8 devs_increment; /* ndevs has to be a multiple of this */
419 u8 ncopies; /* how many copies to data has */
420 u8 nparity; /* number of stripes worth of bytes to store
421 * parity information */
422 u8 mindev_error; /* error code if min devs requisite is unmet */
423 const char raid_name[8]; /* name of the raid */
424 u64 bg_flag; /* block group flag of the raid */
427 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
436 int verified_stripes; /* For mount time dev extent verification */
437 struct btrfs_io_stripe stripes[];
440 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
441 (sizeof(struct btrfs_io_stripe) * (n)))
443 struct btrfs_balance_args;
444 struct btrfs_balance_progress;
445 struct btrfs_balance_control {
446 struct btrfs_balance_args data;
447 struct btrfs_balance_args meta;
448 struct btrfs_balance_args sys;
452 struct btrfs_balance_progress stat;
456 * Search for a given device by the set parameters
458 struct btrfs_dev_lookup_args {
465 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
466 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
468 #define BTRFS_DEV_LOOKUP_ARGS(name) \
469 struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
475 BTRFS_MAP_GET_READ_MIRRORS,
478 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
480 switch (bio_op(bio)) {
482 return BTRFS_MAP_DISCARD;
484 case REQ_OP_ZONE_APPEND:
485 return BTRFS_MAP_WRITE;
490 return BTRFS_MAP_READ;
494 void btrfs_get_bioc(struct btrfs_io_context *bioc);
495 void btrfs_put_bioc(struct btrfs_io_context *bioc);
496 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
497 u64 logical, u64 *length,
498 struct btrfs_io_context **bioc_ret, int mirror_num);
499 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
500 u64 logical, u64 *length,
501 struct btrfs_io_context **bioc_ret);
502 int btrfs_get_io_geometry(struct btrfs_fs_info *fs_info, struct extent_map *map,
503 enum btrfs_map_op op, u64 logical,
504 struct btrfs_io_geometry *io_geom);
505 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
506 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
507 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
509 void btrfs_mapping_tree_free(struct extent_map_tree *tree);
510 blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
512 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
513 fmode_t flags, void *holder);
514 struct btrfs_device *btrfs_scan_one_device(const char *path,
515 fmode_t flags, void *holder);
516 int btrfs_forget_devices(dev_t devt);
517 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
518 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
519 void btrfs_assign_next_active_device(struct btrfs_device *device,
520 struct btrfs_device *this_dev);
521 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
523 const char *devpath);
524 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
525 struct btrfs_dev_lookup_args *args,
527 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
530 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
531 void btrfs_free_device(struct btrfs_device *device);
532 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
533 struct btrfs_dev_lookup_args *args,
534 struct block_device **bdev, fmode_t *mode);
535 void __exit btrfs_cleanup_fs_uuids(void);
536 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
537 int btrfs_grow_device(struct btrfs_trans_handle *trans,
538 struct btrfs_device *device, u64 new_size);
539 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
540 const struct btrfs_dev_lookup_args *args);
541 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
542 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
543 int btrfs_balance(struct btrfs_fs_info *fs_info,
544 struct btrfs_balance_control *bctl,
545 struct btrfs_ioctl_balance_args *bargs);
546 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
547 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
548 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
549 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
550 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
551 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
552 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
553 int btrfs_uuid_scan_kthread(void *data);
554 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
555 int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes,
556 u64 *start, u64 *max_avail);
557 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
558 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
559 struct btrfs_ioctl_get_dev_stats *stats);
560 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
561 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
562 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
563 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
564 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
565 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
566 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
567 u64 logical, u64 len);
568 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
570 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
571 struct btrfs_block_group *bg);
572 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
573 struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
574 u64 logical, u64 length);
575 void btrfs_release_disk_super(struct btrfs_super_block *super);
577 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
580 atomic_inc(dev->dev_stat_values + index);
582 * This memory barrier orders stores updating statistics before stores
583 * updating dev_stats_ccnt.
585 * It pairs with smp_rmb() in btrfs_run_dev_stats().
587 smp_mb__before_atomic();
588 atomic_inc(&dev->dev_stats_ccnt);
591 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
594 return atomic_read(dev->dev_stat_values + index);
597 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
602 ret = atomic_xchg(dev->dev_stat_values + index, 0);
604 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
605 * - RMW operations that have a return value are fully ordered;
607 * This implicit memory barriers is paired with the smp_rmb in
608 * btrfs_run_dev_stats
610 atomic_inc(&dev->dev_stats_ccnt);
614 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
615 int index, unsigned long val)
617 atomic_set(dev->dev_stat_values + index, val);
619 * This memory barrier orders stores updating statistics before stores
620 * updating dev_stats_ccnt.
622 * It pairs with smp_rmb() in btrfs_run_dev_stats().
624 smp_mb__before_atomic();
625 atomic_inc(&dev->dev_stats_ccnt);
628 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
630 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
631 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
632 struct btrfs_device *failing_dev);
633 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info,
634 struct block_device *bdev,
635 const char *device_path);
637 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
638 int btrfs_bg_type_to_factor(u64 flags);
639 const char *btrfs_bg_type_to_raid_name(u64 flags);
640 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
641 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);