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 /* Used by sanity check for btrfs_raid_types. */
21 #define const_ffs(n) (__builtin_ctzll(n) + 1)
24 * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires
25 * RAID0 always to be the lowest profile bit.
26 * Although it's part of on-disk format and should never change, do extra
27 * compile-time sanity checks.
29 static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) <
30 const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0));
31 static_assert(const_ilog2(BTRFS_BLOCK_GROUP_RAID0) >
32 ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK));
34 /* ilog2() can handle both constants and variables */
35 #define BTRFS_BG_FLAG_TO_INDEX(profile) \
36 ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1))
38 enum btrfs_raid_types {
39 /* SINGLE is the special one as it doesn't have on-disk bit. */
40 BTRFS_RAID_SINGLE = 0,
42 BTRFS_RAID_RAID0 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0),
43 BTRFS_RAID_RAID1 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1),
44 BTRFS_RAID_DUP = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP),
45 BTRFS_RAID_RAID10 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10),
46 BTRFS_RAID_RAID5 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5),
47 BTRFS_RAID_RAID6 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6),
48 BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3),
49 BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4),
54 struct btrfs_io_geometry {
55 /* remaining bytes before crossing a stripe */
57 /* offset of logical address in chunk */
59 /* length of single IO stripe */
61 /* offset of address in stripe */
63 /* number of stripe where address falls */
65 /* offset of raid56 stripe into the chunk */
66 u64 raid56_stripe_offset;
70 * Use sequence counter to get consistent device stat data on
73 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
74 #include <linux/seqlock.h>
75 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
76 #define btrfs_device_data_ordered_init(device) \
77 seqcount_init(&device->data_seqcount)
79 #define btrfs_device_data_ordered_init(device) do { } while (0)
82 #define BTRFS_DEV_STATE_WRITEABLE (0)
83 #define BTRFS_DEV_STATE_IN_FS_METADATA (1)
84 #define BTRFS_DEV_STATE_MISSING (2)
85 #define BTRFS_DEV_STATE_REPLACE_TGT (3)
86 #define BTRFS_DEV_STATE_FLUSH_SENT (4)
87 #define BTRFS_DEV_STATE_NO_READA (5)
89 struct btrfs_zoned_device_info;
92 struct list_head dev_list; /* device_list_mutex */
93 struct list_head dev_alloc_list; /* chunk mutex */
94 struct list_head post_commit_list; /* chunk mutex */
95 struct btrfs_fs_devices *fs_devices;
96 struct btrfs_fs_info *fs_info;
98 struct rcu_string __rcu *name;
102 struct block_device *bdev;
104 struct btrfs_zoned_device_info *zone_info;
106 /* the mode sent to blkdev_get */
110 * Device's major-minor number. Must be set even if the device is not
111 * opened (bdev == NULL), unless the device is missing.
114 unsigned long dev_state;
115 blk_status_t last_flush_error;
117 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
118 seqcount_t data_seqcount;
121 /* the internal btrfs device id */
124 /* size of the device in memory */
127 /* size of the device on disk */
128 u64 disk_total_bytes;
133 /* optimal io alignment for this device */
136 /* optimal io width for this device */
138 /* type and info about this device */
141 /* minimal io size for this device */
144 /* physical drive uuid (or lvm uuid) */
145 u8 uuid[BTRFS_UUID_SIZE];
148 * size of the device on the current transaction
150 * This variant is update when committing the transaction,
151 * and protected by chunk mutex
153 u64 commit_total_bytes;
155 /* bytes used on the current transaction */
156 u64 commit_bytes_used;
158 /* Bio used for flushing device barriers */
159 struct bio flush_bio;
160 struct completion flush_wait;
162 /* per-device scrub information */
163 struct scrub_ctx *scrub_ctx;
165 /* disk I/O failure stats. For detailed description refer to
166 * enum btrfs_dev_stat_values in ioctl.h */
169 /* Counter to record the change of device stats */
170 atomic_t dev_stats_ccnt;
171 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
173 struct extent_io_tree alloc_state;
175 struct completion kobj_unregister;
176 /* For sysfs/FSID/devinfo/devid/ */
177 struct kobject devid_kobj;
179 /* Bandwidth limit for scrub, in bytes */
184 * If we read those variants at the context of their own lock, we needn't
185 * use the following helpers, reading them directly is safe.
187 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
188 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
190 btrfs_device_get_##name(const struct btrfs_device *dev) \
196 seq = read_seqcount_begin(&dev->data_seqcount); \
198 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
203 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
206 write_seqcount_begin(&dev->data_seqcount); \
208 write_seqcount_end(&dev->data_seqcount); \
211 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
212 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
214 btrfs_device_get_##name(const struct btrfs_device *dev) \
225 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
232 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
234 btrfs_device_get_##name(const struct btrfs_device *dev) \
240 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
246 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
247 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
248 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
250 enum btrfs_chunk_allocation_policy {
251 BTRFS_CHUNK_ALLOC_REGULAR,
252 BTRFS_CHUNK_ALLOC_ZONED,
256 * Read policies for mirrored block group profiles, read picks the stripe based
259 enum btrfs_read_policy {
260 /* Use process PID to choose the stripe */
261 BTRFS_READ_POLICY_PID,
262 BTRFS_NR_READ_POLICY,
265 struct btrfs_fs_devices {
266 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
267 u8 metadata_uuid[BTRFS_FSID_SIZE];
269 struct list_head fs_list;
272 * Number of devices under this fsid including missing and
273 * replace-target device and excludes seed devices.
278 * The number of devices that successfully opened, including
279 * replace-target, excludes seed devices.
283 /* The number of devices that are under the chunk allocation list. */
286 /* Count of missing devices under this fsid excluding seed device. */
291 * Count of devices from btrfs_super_block::num_devices for this fsid,
292 * which includes the seed device, excludes the transient replace-target
297 /* Highest generation number of seen devices */
298 u64 latest_generation;
301 * The mount device or a device with highest generation after removal
304 struct btrfs_device *latest_dev;
306 /* all of the devices in the FS, protected by a mutex
307 * so we can safely walk it to write out the supers without
308 * worrying about add/remove by the multi-device code.
309 * Scrubbing super can kick off supers writing by holding
312 struct mutex device_list_mutex;
314 /* List of all devices, protected by device_list_mutex */
315 struct list_head devices;
318 * Devices which can satisfy space allocation. Protected by
321 struct list_head alloc_list;
323 struct list_head seed_list;
328 /* set when we find or add a device that doesn't have the
333 struct btrfs_fs_info *fs_info;
335 struct kobject fsid_kobj;
336 struct kobject *devices_kobj;
337 struct kobject *devinfo_kobj;
338 struct completion kobj_unregister;
340 enum btrfs_chunk_allocation_policy chunk_alloc_policy;
342 /* Policy used to read the mirrored stripes */
343 enum btrfs_read_policy read_policy;
346 #define BTRFS_BIO_INLINE_CSUM_SIZE 64
348 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \
349 - sizeof(struct btrfs_chunk)) \
350 / sizeof(struct btrfs_stripe) + 1)
352 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \
353 - 2 * sizeof(struct btrfs_disk_key) \
354 - 2 * sizeof(struct btrfs_chunk)) \
355 / sizeof(struct btrfs_stripe) + 1)
358 * Additional info to pass along bio.
360 * Mostly for btrfs specific features like csum and mirror_num.
363 unsigned int mirror_num;
368 /* @device is for stripe IO submission. */
369 struct btrfs_device *device;
371 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
372 struct bvec_iter iter;
375 * This member must come last, bio_alloc_bioset will allocate enough
376 * bytes for entire btrfs_bio but relies on bio being last.
381 static inline struct btrfs_bio *btrfs_bio(struct bio *bio)
383 return container_of(bio, struct btrfs_bio, bio);
386 static inline void btrfs_bio_free_csum(struct btrfs_bio *bbio)
388 if (bbio->csum != bbio->csum_inline) {
394 struct btrfs_io_stripe {
395 struct btrfs_device *dev;
397 u64 length; /* only used for discard mappings */
401 * Context for IO subsmission for device stripe.
403 * - Track the unfinished mirrors for mirror based profiles
404 * Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
406 * - Contain the logical -> physical mapping info
407 * Used by submit_stripe_bio() for mapping logical bio
408 * into physical device address.
410 * - Contain device replace info
411 * Used by handle_ops_on_dev_replace() to copy logical bios
412 * into the new device.
414 * - Contain RAID56 full stripe logical bytenrs
416 struct btrfs_io_context {
418 atomic_t stripes_pending;
419 struct btrfs_fs_info *fs_info;
420 u64 map_type; /* get from map_lookup->type */
421 bio_end_io_t *end_io;
422 struct bio *orig_bio;
431 * logical block numbers for the start of each stripe
432 * The last one or two are p/q. These are sorted,
433 * so raid_map[0] is the start of our full stripe
436 struct btrfs_io_stripe stripes[];
439 struct btrfs_device_info {
440 struct btrfs_device *dev;
446 struct btrfs_raid_attr {
447 u8 sub_stripes; /* sub_stripes info for map */
448 u8 dev_stripes; /* stripes per dev */
449 u8 devs_max; /* max devs to use */
450 u8 devs_min; /* min devs needed */
451 u8 tolerated_failures; /* max tolerated fail devs */
452 u8 devs_increment; /* ndevs has to be a multiple of this */
453 u8 ncopies; /* how many copies to data has */
454 u8 nparity; /* number of stripes worth of bytes to store
455 * parity information */
456 u8 mindev_error; /* error code if min devs requisite is unmet */
457 const char raid_name[8]; /* name of the raid */
458 u64 bg_flag; /* block group flag of the raid */
461 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
470 int verified_stripes; /* For mount time dev extent verification */
471 struct btrfs_io_stripe stripes[];
474 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
475 (sizeof(struct btrfs_io_stripe) * (n)))
477 struct btrfs_balance_args;
478 struct btrfs_balance_progress;
479 struct btrfs_balance_control {
480 struct btrfs_balance_args data;
481 struct btrfs_balance_args meta;
482 struct btrfs_balance_args sys;
486 struct btrfs_balance_progress stat;
490 * Search for a given device by the set parameters
492 struct btrfs_dev_lookup_args {
499 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
500 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
502 #define BTRFS_DEV_LOOKUP_ARGS(name) \
503 struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
509 BTRFS_MAP_GET_READ_MIRRORS,
512 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
514 switch (bio_op(bio)) {
516 return BTRFS_MAP_DISCARD;
518 case REQ_OP_ZONE_APPEND:
519 return BTRFS_MAP_WRITE;
524 return BTRFS_MAP_READ;
528 void btrfs_get_bioc(struct btrfs_io_context *bioc);
529 void btrfs_put_bioc(struct btrfs_io_context *bioc);
530 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
531 u64 logical, u64 *length,
532 struct btrfs_io_context **bioc_ret, int mirror_num);
533 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
534 u64 logical, u64 *length,
535 struct btrfs_io_context **bioc_ret);
536 int btrfs_get_io_geometry(struct btrfs_fs_info *fs_info, struct extent_map *map,
537 enum btrfs_map_op op, u64 logical,
538 struct btrfs_io_geometry *io_geom);
539 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
540 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
541 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
543 void btrfs_mapping_tree_free(struct extent_map_tree *tree);
544 blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
546 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
547 fmode_t flags, void *holder);
548 struct btrfs_device *btrfs_scan_one_device(const char *path,
549 fmode_t flags, void *holder);
550 int btrfs_forget_devices(dev_t devt);
551 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
552 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
553 void btrfs_assign_next_active_device(struct btrfs_device *device,
554 struct btrfs_device *this_dev);
555 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
557 const char *devpath);
558 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
559 struct btrfs_dev_lookup_args *args,
561 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
564 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
565 void btrfs_free_device(struct btrfs_device *device);
566 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
567 struct btrfs_dev_lookup_args *args,
568 struct block_device **bdev, fmode_t *mode);
569 void __exit btrfs_cleanup_fs_uuids(void);
570 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
571 int btrfs_grow_device(struct btrfs_trans_handle *trans,
572 struct btrfs_device *device, u64 new_size);
573 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
574 const struct btrfs_dev_lookup_args *args);
575 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
576 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
577 int btrfs_balance(struct btrfs_fs_info *fs_info,
578 struct btrfs_balance_control *bctl,
579 struct btrfs_ioctl_balance_args *bargs);
580 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
581 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
582 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
583 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
584 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
585 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
586 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
587 int btrfs_uuid_scan_kthread(void *data);
588 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
589 int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes,
590 u64 *start, u64 *max_avail);
591 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
592 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
593 struct btrfs_ioctl_get_dev_stats *stats);
594 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
595 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
596 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
597 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
598 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
599 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
600 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
601 u64 logical, u64 len);
602 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
604 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
605 struct btrfs_block_group *bg);
606 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
607 struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
608 u64 logical, u64 length);
609 void btrfs_release_disk_super(struct btrfs_super_block *super);
611 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
614 atomic_inc(dev->dev_stat_values + index);
616 * This memory barrier orders stores updating statistics before stores
617 * updating dev_stats_ccnt.
619 * It pairs with smp_rmb() in btrfs_run_dev_stats().
621 smp_mb__before_atomic();
622 atomic_inc(&dev->dev_stats_ccnt);
625 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
628 return atomic_read(dev->dev_stat_values + index);
631 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
636 ret = atomic_xchg(dev->dev_stat_values + index, 0);
638 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
639 * - RMW operations that have a return value are fully ordered;
641 * This implicit memory barriers is paired with the smp_rmb in
642 * btrfs_run_dev_stats
644 atomic_inc(&dev->dev_stats_ccnt);
648 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
649 int index, unsigned long val)
651 atomic_set(dev->dev_stat_values + index, val);
653 * This memory barrier orders stores updating statistics before stores
654 * updating dev_stats_ccnt.
656 * It pairs with smp_rmb() in btrfs_run_dev_stats().
658 smp_mb__before_atomic();
659 atomic_inc(&dev->dev_stats_ccnt);
662 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
664 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
665 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
666 struct btrfs_device *failing_dev);
667 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info,
668 struct block_device *bdev,
669 const char *device_path);
671 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
672 int btrfs_bg_type_to_factor(u64 flags);
673 const char *btrfs_bg_type_to_raid_name(u64 flags);
674 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
675 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);