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
9 #include <linux/sort.h>
10 #include <linux/btrfs.h>
11 #include "async-thread.h"
13 #include "tree-checker.h"
14 #include "rcu-string.h"
16 #define BTRFS_MAX_DATA_CHUNK_SIZE (10ULL * SZ_1G)
18 extern struct mutex uuid_mutex;
20 #define BTRFS_STRIPE_LEN SZ_64K
22 /* Used by sanity check for btrfs_raid_types. */
23 #define const_ffs(n) (__builtin_ctzll(n) + 1)
26 * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires
27 * RAID0 always to be the lowest profile bit.
28 * Although it's part of on-disk format and should never change, do extra
29 * compile-time sanity checks.
31 static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) <
32 const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0));
33 static_assert(const_ilog2(BTRFS_BLOCK_GROUP_RAID0) >
34 ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK));
36 /* ilog2() can handle both constants and variables */
37 #define BTRFS_BG_FLAG_TO_INDEX(profile) \
38 ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1))
40 enum btrfs_raid_types {
41 /* SINGLE is the special one as it doesn't have on-disk bit. */
42 BTRFS_RAID_SINGLE = 0,
44 BTRFS_RAID_RAID0 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0),
45 BTRFS_RAID_RAID1 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1),
46 BTRFS_RAID_DUP = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP),
47 BTRFS_RAID_RAID10 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10),
48 BTRFS_RAID_RAID5 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5),
49 BTRFS_RAID_RAID6 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6),
50 BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3),
51 BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4),
56 struct btrfs_io_geometry {
57 /* remaining bytes before crossing a stripe */
59 /* offset of logical address in chunk */
61 /* length of single IO stripe */
63 /* offset of address in stripe */
65 /* number of stripe where address falls */
67 /* offset of raid56 stripe into the chunk */
68 u64 raid56_stripe_offset;
72 * Use sequence counter to get consistent device stat data on
75 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
76 #include <linux/seqlock.h>
77 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
78 #define btrfs_device_data_ordered_init(device) \
79 seqcount_init(&device->data_seqcount)
81 #define btrfs_device_data_ordered_init(device) do { } while (0)
84 #define BTRFS_DEV_STATE_WRITEABLE (0)
85 #define BTRFS_DEV_STATE_IN_FS_METADATA (1)
86 #define BTRFS_DEV_STATE_MISSING (2)
87 #define BTRFS_DEV_STATE_REPLACE_TGT (3)
88 #define BTRFS_DEV_STATE_FLUSH_SENT (4)
89 #define BTRFS_DEV_STATE_NO_READA (5)
91 struct btrfs_zoned_device_info;
94 struct list_head dev_list; /* device_list_mutex */
95 struct list_head dev_alloc_list; /* chunk mutex */
96 struct list_head post_commit_list; /* chunk mutex */
97 struct btrfs_fs_devices *fs_devices;
98 struct btrfs_fs_info *fs_info;
100 struct rcu_string __rcu *name;
104 struct block_device *bdev;
106 struct btrfs_zoned_device_info *zone_info;
108 /* the mode sent to blkdev_get */
112 * Device's major-minor number. Must be set even if the device is not
113 * opened (bdev == NULL), unless the device is missing.
116 unsigned long dev_state;
117 blk_status_t last_flush_error;
119 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
120 seqcount_t data_seqcount;
123 /* the internal btrfs device id */
126 /* size of the device in memory */
129 /* size of the device on disk */
130 u64 disk_total_bytes;
135 /* optimal io alignment for this device */
138 /* optimal io width for this device */
140 /* type and info about this device */
143 /* minimal io size for this device */
146 /* physical drive uuid (or lvm uuid) */
147 u8 uuid[BTRFS_UUID_SIZE];
150 * size of the device on the current transaction
152 * This variant is update when committing the transaction,
153 * and protected by chunk mutex
155 u64 commit_total_bytes;
157 /* bytes used on the current transaction */
158 u64 commit_bytes_used;
160 /* Bio used for flushing device barriers */
161 struct bio flush_bio;
162 struct completion flush_wait;
164 /* per-device scrub information */
165 struct scrub_ctx *scrub_ctx;
167 /* disk I/O failure stats. For detailed description refer to
168 * enum btrfs_dev_stat_values in ioctl.h */
171 /* Counter to record the change of device stats */
172 atomic_t dev_stats_ccnt;
173 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
175 struct extent_io_tree alloc_state;
177 struct completion kobj_unregister;
178 /* For sysfs/FSID/devinfo/devid/ */
179 struct kobject devid_kobj;
181 /* Bandwidth limit for scrub, in bytes */
186 * Block group or device which contains an active swapfile. Used for preventing
187 * unsafe operations while a swapfile is active.
189 * These are sorted on (ptr, inode) (note that a block group or device can
190 * contain more than one swapfile). We compare the pointer values because we
191 * don't actually care what the object is, we just need a quick check whether
192 * the object exists in the rbtree.
194 struct btrfs_swapfile_pin {
199 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr
200 * points to a struct btrfs_device.
204 * Only used when 'is_block_group' is true and it is the number of
205 * extents used by a swapfile for this block group ('ptr' field).
211 * If we read those variants at the context of their own lock, we needn't
212 * use the following helpers, reading them directly is safe.
214 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
215 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
217 btrfs_device_get_##name(const struct btrfs_device *dev) \
223 seq = read_seqcount_begin(&dev->data_seqcount); \
225 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
230 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
233 write_seqcount_begin(&dev->data_seqcount); \
235 write_seqcount_end(&dev->data_seqcount); \
238 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
239 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
241 btrfs_device_get_##name(const struct btrfs_device *dev) \
252 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
259 #define BTRFS_DEVICE_GETSET_FUNCS(name) \
261 btrfs_device_get_##name(const struct btrfs_device *dev) \
267 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
273 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
274 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
275 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
277 enum btrfs_chunk_allocation_policy {
278 BTRFS_CHUNK_ALLOC_REGULAR,
279 BTRFS_CHUNK_ALLOC_ZONED,
283 * Read policies for mirrored block group profiles, read picks the stripe based
286 enum btrfs_read_policy {
287 /* Use process PID to choose the stripe */
288 BTRFS_READ_POLICY_PID,
289 BTRFS_NR_READ_POLICY,
292 struct btrfs_fs_devices {
293 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
294 u8 metadata_uuid[BTRFS_FSID_SIZE];
296 struct list_head fs_list;
299 * Number of devices under this fsid including missing and
300 * replace-target device and excludes seed devices.
305 * The number of devices that successfully opened, including
306 * replace-target, excludes seed devices.
310 /* The number of devices that are under the chunk allocation list. */
313 /* Count of missing devices under this fsid excluding seed device. */
318 * Count of devices from btrfs_super_block::num_devices for this fsid,
319 * which includes the seed device, excludes the transient replace-target
324 /* Highest generation number of seen devices */
325 u64 latest_generation;
328 * The mount device or a device with highest generation after removal
331 struct btrfs_device *latest_dev;
333 /* all of the devices in the FS, protected by a mutex
334 * so we can safely walk it to write out the supers without
335 * worrying about add/remove by the multi-device code.
336 * Scrubbing super can kick off supers writing by holding
339 struct mutex device_list_mutex;
341 /* List of all devices, protected by device_list_mutex */
342 struct list_head devices;
345 * Devices which can satisfy space allocation. Protected by
348 struct list_head alloc_list;
350 struct list_head seed_list;
355 /* set when we find or add a device that doesn't have the
359 /* Devices support TRIM/discard commands */
362 struct btrfs_fs_info *fs_info;
364 struct kobject fsid_kobj;
365 struct kobject *devices_kobj;
366 struct kobject *devinfo_kobj;
367 struct completion kobj_unregister;
369 enum btrfs_chunk_allocation_policy chunk_alloc_policy;
371 /* Policy used to read the mirrored stripes */
372 enum btrfs_read_policy read_policy;
375 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \
376 - sizeof(struct btrfs_chunk)) \
377 / sizeof(struct btrfs_stripe) + 1)
379 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \
380 - 2 * sizeof(struct btrfs_disk_key) \
381 - 2 * sizeof(struct btrfs_chunk)) \
382 / sizeof(struct btrfs_stripe) + 1)
384 struct btrfs_io_stripe {
385 struct btrfs_device *dev;
389 /* For the endio handler */
390 struct btrfs_io_context *bioc;
394 struct btrfs_discard_stripe {
395 struct btrfs_device *dev;
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 struct btrfs_fs_info *fs_info;
419 u64 map_type; /* get from map_lookup->type */
420 struct bio *orig_bio;
428 * logical block numbers for the start of each stripe
429 * The last one or two are p/q. These are sorted,
430 * so raid_map[0] is the start of our full stripe
433 struct btrfs_io_stripe stripes[];
436 struct btrfs_device_info {
437 struct btrfs_device *dev;
443 struct btrfs_raid_attr {
444 u8 sub_stripes; /* sub_stripes info for map */
445 u8 dev_stripes; /* stripes per dev */
446 u8 devs_max; /* max devs to use */
447 u8 devs_min; /* min devs needed */
448 u8 tolerated_failures; /* max tolerated fail devs */
449 u8 devs_increment; /* ndevs has to be a multiple of this */
450 u8 ncopies; /* how many copies to data has */
451 u8 nparity; /* number of stripes worth of bytes to store
452 * parity information */
453 u8 mindev_error; /* error code if min devs requisite is unmet */
454 const char raid_name[8]; /* name of the raid */
455 u64 bg_flag; /* block group flag of the raid */
458 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
467 int verified_stripes; /* For mount time dev extent verification */
468 struct btrfs_io_stripe stripes[];
471 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
472 (sizeof(struct btrfs_io_stripe) * (n)))
474 struct btrfs_balance_args;
475 struct btrfs_balance_progress;
476 struct btrfs_balance_control {
477 struct btrfs_balance_args data;
478 struct btrfs_balance_args meta;
479 struct btrfs_balance_args sys;
483 struct btrfs_balance_progress stat;
487 * Search for a given device by the set parameters
489 struct btrfs_dev_lookup_args {
496 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
497 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
499 #define BTRFS_DEV_LOOKUP_ARGS(name) \
500 struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
506 BTRFS_MAP_GET_READ_MIRRORS,
509 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
511 switch (bio_op(bio)) {
513 return BTRFS_MAP_DISCARD;
515 case REQ_OP_ZONE_APPEND:
516 return BTRFS_MAP_WRITE;
521 return BTRFS_MAP_READ;
525 static inline unsigned long btrfs_chunk_item_size(int num_stripes)
528 return sizeof(struct btrfs_chunk) +
529 sizeof(struct btrfs_stripe) * (num_stripes - 1);
532 void btrfs_get_bioc(struct btrfs_io_context *bioc);
533 void btrfs_put_bioc(struct btrfs_io_context *bioc);
534 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
535 u64 logical, u64 *length,
536 struct btrfs_io_context **bioc_ret, int mirror_num);
537 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
538 u64 logical, u64 *length,
539 struct btrfs_io_context **bioc_ret);
540 int __btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
541 u64 logical, u64 *length,
542 struct btrfs_io_context **bioc_ret,
543 struct btrfs_io_stripe *smap, int *mirror_num_ret,
545 struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,
546 u64 logical, u64 *length_ret,
548 int btrfs_get_io_geometry(struct btrfs_fs_info *fs_info, struct extent_map *map,
549 enum btrfs_map_op op, u64 logical,
550 struct btrfs_io_geometry *io_geom);
551 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
552 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
553 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
555 void btrfs_mapping_tree_free(struct extent_map_tree *tree);
556 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
557 fmode_t flags, void *holder);
558 struct btrfs_device *btrfs_scan_one_device(const char *path,
559 fmode_t flags, void *holder);
560 int btrfs_forget_devices(dev_t devt);
561 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
562 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
563 void btrfs_assign_next_active_device(struct btrfs_device *device,
564 struct btrfs_device *this_dev);
565 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
567 const char *devpath);
568 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
569 struct btrfs_dev_lookup_args *args,
571 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
572 const u64 *devid, const u8 *uuid,
574 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
575 void btrfs_free_device(struct btrfs_device *device);
576 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
577 struct btrfs_dev_lookup_args *args,
578 struct block_device **bdev, fmode_t *mode);
579 void __exit btrfs_cleanup_fs_uuids(void);
580 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
581 int btrfs_grow_device(struct btrfs_trans_handle *trans,
582 struct btrfs_device *device, u64 new_size);
583 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
584 const struct btrfs_dev_lookup_args *args);
585 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
586 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
587 int btrfs_balance(struct btrfs_fs_info *fs_info,
588 struct btrfs_balance_control *bctl,
589 struct btrfs_ioctl_balance_args *bargs);
590 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
591 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
592 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
593 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
594 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
595 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
596 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
597 int btrfs_uuid_scan_kthread(void *data);
598 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
599 int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes,
600 u64 *start, u64 *max_avail);
601 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
602 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
603 struct btrfs_ioctl_get_dev_stats *stats);
604 int btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
605 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
606 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
607 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
608 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
609 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
610 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
611 u64 logical, u64 len);
612 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
614 u64 btrfs_calc_stripe_length(const struct extent_map *em);
615 int btrfs_nr_parity_stripes(u64 type);
616 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
617 struct btrfs_block_group *bg);
618 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
619 struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
620 u64 logical, u64 length);
621 void btrfs_release_disk_super(struct btrfs_super_block *super);
623 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
626 atomic_inc(dev->dev_stat_values + index);
628 * This memory barrier orders stores updating statistics before stores
629 * updating dev_stats_ccnt.
631 * It pairs with smp_rmb() in btrfs_run_dev_stats().
633 smp_mb__before_atomic();
634 atomic_inc(&dev->dev_stats_ccnt);
637 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
640 return atomic_read(dev->dev_stat_values + index);
643 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
648 ret = atomic_xchg(dev->dev_stat_values + index, 0);
650 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
651 * - RMW operations that have a return value are fully ordered;
653 * This implicit memory barriers is paired with the smp_rmb in
654 * btrfs_run_dev_stats
656 atomic_inc(&dev->dev_stats_ccnt);
660 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
661 int index, unsigned long val)
663 atomic_set(dev->dev_stat_values + index, val);
665 * This memory barrier orders stores updating statistics before stores
666 * updating dev_stats_ccnt.
668 * It pairs with smp_rmb() in btrfs_run_dev_stats().
670 smp_mb__before_atomic();
671 atomic_inc(&dev->dev_stats_ccnt);
674 static inline const char *btrfs_dev_name(const struct btrfs_device *device)
676 if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
677 return "<missing disk>";
679 return rcu_str_deref(device->name);
682 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
684 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
685 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
686 struct btrfs_device *failing_dev);
687 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info,
688 struct block_device *bdev,
689 const char *device_path);
691 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
692 int btrfs_bg_type_to_factor(u64 flags);
693 const char *btrfs_bg_type_to_raid_name(u64 flags);
694 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
695 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);
697 bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr);