Loading...
1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6#ifndef BTRFS_VOLUMES_H
7#define BTRFS_VOLUMES_H
8
9#include <linux/blk_types.h>
10#include <linux/sizes.h>
11#include <linux/atomic.h>
12#include <linux/sort.h>
13#include <linux/list.h>
14#include <linux/mutex.h>
15#include <linux/log2.h>
16#include <linux/kobject.h>
17#include <linux/refcount.h>
18#include <linux/completion.h>
19#include <linux/rbtree.h>
20#include <uapi/linux/btrfs.h>
21#include "messages.h"
22#include "rcu-string.h"
23
24struct block_device;
25struct bdev_handle;
26struct btrfs_fs_info;
27struct btrfs_block_group;
28struct btrfs_trans_handle;
29struct btrfs_zoned_device_info;
30
31#define BTRFS_MAX_DATA_CHUNK_SIZE (10ULL * SZ_1G)
32
33/*
34 * Arbitratry maximum size of one discard request to limit potentially long time
35 * spent in blkdev_issue_discard().
36 */
37#define BTRFS_MAX_DISCARD_CHUNK_SIZE (SZ_1G)
38
39extern struct mutex uuid_mutex;
40
41#define BTRFS_STRIPE_LEN SZ_64K
42#define BTRFS_STRIPE_LEN_SHIFT (16)
43#define BTRFS_STRIPE_LEN_MASK (BTRFS_STRIPE_LEN - 1)
44
45static_assert(const_ilog2(BTRFS_STRIPE_LEN) == BTRFS_STRIPE_LEN_SHIFT);
46
47/* Used by sanity check for btrfs_raid_types. */
48#define const_ffs(n) (__builtin_ctzll(n) + 1)
49
50/*
51 * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires
52 * RAID0 always to be the lowest profile bit.
53 * Although it's part of on-disk format and should never change, do extra
54 * compile-time sanity checks.
55 */
56static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) <
57 const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0));
58static_assert(const_ilog2(BTRFS_BLOCK_GROUP_RAID0) >
59 ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK));
60
61/* ilog2() can handle both constants and variables */
62#define BTRFS_BG_FLAG_TO_INDEX(profile) \
63 ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1))
64
65enum btrfs_raid_types {
66 /* SINGLE is the special one as it doesn't have on-disk bit. */
67 BTRFS_RAID_SINGLE = 0,
68
69 BTRFS_RAID_RAID0 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0),
70 BTRFS_RAID_RAID1 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1),
71 BTRFS_RAID_DUP = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP),
72 BTRFS_RAID_RAID10 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10),
73 BTRFS_RAID_RAID5 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5),
74 BTRFS_RAID_RAID6 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6),
75 BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3),
76 BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4),
77
78 BTRFS_NR_RAID_TYPES
79};
80
81/*
82 * Use sequence counter to get consistent device stat data on
83 * 32-bit processors.
84 */
85#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
86#include <linux/seqlock.h>
87#define __BTRFS_NEED_DEVICE_DATA_ORDERED
88#define btrfs_device_data_ordered_init(device) \
89 seqcount_init(&device->data_seqcount)
90#else
91#define btrfs_device_data_ordered_init(device) do { } while (0)
92#endif
93
94#define BTRFS_DEV_STATE_WRITEABLE (0)
95#define BTRFS_DEV_STATE_IN_FS_METADATA (1)
96#define BTRFS_DEV_STATE_MISSING (2)
97#define BTRFS_DEV_STATE_REPLACE_TGT (3)
98#define BTRFS_DEV_STATE_FLUSH_SENT (4)
99#define BTRFS_DEV_STATE_NO_READA (5)
100
101/* Special value encoding failure to write primary super block. */
102#define BTRFS_SUPER_PRIMARY_WRITE_ERROR (INT_MAX / 2)
103
104struct btrfs_fs_devices;
105
106struct btrfs_device {
107 struct list_head dev_list; /* device_list_mutex */
108 struct list_head dev_alloc_list; /* chunk mutex */
109 struct list_head post_commit_list; /* chunk mutex */
110 struct btrfs_fs_devices *fs_devices;
111 struct btrfs_fs_info *fs_info;
112
113 struct rcu_string __rcu *name;
114
115 u64 generation;
116
117 struct file *bdev_file;
118 struct block_device *bdev;
119
120 struct btrfs_zoned_device_info *zone_info;
121
122 /*
123 * Device's major-minor number. Must be set even if the device is not
124 * opened (bdev == NULL), unless the device is missing.
125 */
126 dev_t devt;
127 unsigned long dev_state;
128 blk_status_t last_flush_error;
129
130#ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
131 seqcount_t data_seqcount;
132#endif
133
134 /* the internal btrfs device id */
135 u64 devid;
136
137 /* size of the device in memory */
138 u64 total_bytes;
139
140 /* size of the device on disk */
141 u64 disk_total_bytes;
142
143 /* bytes used */
144 u64 bytes_used;
145
146 /* optimal io alignment for this device */
147 u32 io_align;
148
149 /* optimal io width for this device */
150 u32 io_width;
151 /* type and info about this device */
152 u64 type;
153
154 /*
155 * Counter of super block write errors, values larger than
156 * BTRFS_SUPER_PRIMARY_WRITE_ERROR encode primary super block write failure.
157 */
158 atomic_t sb_write_errors;
159
160 /* minimal io size for this device */
161 u32 sector_size;
162
163 /* physical drive uuid (or lvm uuid) */
164 u8 uuid[BTRFS_UUID_SIZE];
165
166 /*
167 * size of the device on the current transaction
168 *
169 * This variant is update when committing the transaction,
170 * and protected by chunk mutex
171 */
172 u64 commit_total_bytes;
173
174 /* bytes used on the current transaction */
175 u64 commit_bytes_used;
176
177 /* Bio used for flushing device barriers */
178 struct bio flush_bio;
179 struct completion flush_wait;
180
181 /* per-device scrub information */
182 struct scrub_ctx *scrub_ctx;
183
184 /* disk I/O failure stats. For detailed description refer to
185 * enum btrfs_dev_stat_values in ioctl.h */
186 int dev_stats_valid;
187
188 /* Counter to record the change of device stats */
189 atomic_t dev_stats_ccnt;
190 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
191
192 struct extent_io_tree alloc_state;
193
194 struct completion kobj_unregister;
195 /* For sysfs/FSID/devinfo/devid/ */
196 struct kobject devid_kobj;
197
198 /* Bandwidth limit for scrub, in bytes */
199 u64 scrub_speed_max;
200};
201
202/*
203 * Block group or device which contains an active swapfile. Used for preventing
204 * unsafe operations while a swapfile is active.
205 *
206 * These are sorted on (ptr, inode) (note that a block group or device can
207 * contain more than one swapfile). We compare the pointer values because we
208 * don't actually care what the object is, we just need a quick check whether
209 * the object exists in the rbtree.
210 */
211struct btrfs_swapfile_pin {
212 struct rb_node node;
213 void *ptr;
214 struct inode *inode;
215 /*
216 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr
217 * points to a struct btrfs_device.
218 */
219 bool is_block_group;
220 /*
221 * Only used when 'is_block_group' is true and it is the number of
222 * extents used by a swapfile for this block group ('ptr' field).
223 */
224 int bg_extent_count;
225};
226
227/*
228 * If we read those variants at the context of their own lock, we needn't
229 * use the following helpers, reading them directly is safe.
230 */
231#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
232#define BTRFS_DEVICE_GETSET_FUNCS(name) \
233static inline u64 \
234btrfs_device_get_##name(const struct btrfs_device *dev) \
235{ \
236 u64 size; \
237 unsigned int seq; \
238 \
239 do { \
240 seq = read_seqcount_begin(&dev->data_seqcount); \
241 size = dev->name; \
242 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
243 return size; \
244} \
245 \
246static inline void \
247btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
248{ \
249 preempt_disable(); \
250 write_seqcount_begin(&dev->data_seqcount); \
251 dev->name = size; \
252 write_seqcount_end(&dev->data_seqcount); \
253 preempt_enable(); \
254}
255#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
256#define BTRFS_DEVICE_GETSET_FUNCS(name) \
257static inline u64 \
258btrfs_device_get_##name(const struct btrfs_device *dev) \
259{ \
260 u64 size; \
261 \
262 preempt_disable(); \
263 size = dev->name; \
264 preempt_enable(); \
265 return size; \
266} \
267 \
268static inline void \
269btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
270{ \
271 preempt_disable(); \
272 dev->name = size; \
273 preempt_enable(); \
274}
275#else
276#define BTRFS_DEVICE_GETSET_FUNCS(name) \
277static inline u64 \
278btrfs_device_get_##name(const struct btrfs_device *dev) \
279{ \
280 return dev->name; \
281} \
282 \
283static inline void \
284btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
285{ \
286 dev->name = size; \
287}
288#endif
289
290BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
291BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
292BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
293
294enum btrfs_chunk_allocation_policy {
295 BTRFS_CHUNK_ALLOC_REGULAR,
296 BTRFS_CHUNK_ALLOC_ZONED,
297};
298
299/*
300 * Read policies for mirrored block group profiles, read picks the stripe based
301 * on these policies.
302 */
303enum btrfs_read_policy {
304 /* Use process PID to choose the stripe */
305 BTRFS_READ_POLICY_PID,
306 BTRFS_NR_READ_POLICY,
307};
308
309#ifdef CONFIG_BTRFS_EXPERIMENTAL
310/*
311 * Checksum mode - offload it to workqueues or do it synchronously in
312 * btrfs_submit_chunk().
313 */
314enum btrfs_offload_csum_mode {
315 /*
316 * Choose offloading checksum or do it synchronously automatically.
317 * Do it synchronously if the checksum is fast, or offload to workqueues
318 * otherwise.
319 */
320 BTRFS_OFFLOAD_CSUM_AUTO,
321 /* Always offload checksum to workqueues. */
322 BTRFS_OFFLOAD_CSUM_FORCE_ON,
323 /* Never offload checksum to workqueues. */
324 BTRFS_OFFLOAD_CSUM_FORCE_OFF,
325};
326#endif
327
328struct btrfs_fs_devices {
329 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
330
331 /*
332 * UUID written into the btree blocks:
333 *
334 * - If metadata_uuid != fsid then super block must have
335 * BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag set.
336 *
337 * - Following shall be true at all times:
338 * - metadata_uuid == btrfs_header::fsid
339 * - metadata_uuid == btrfs_dev_item::fsid
340 *
341 * - Relations between fsid and metadata_uuid in sb and fs_devices:
342 * - Normal:
343 * fs_devices->fsid == fs_devices->metadata_uuid == sb->fsid
344 * sb->metadata_uuid == 0
345 *
346 * - When the BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag is set:
347 * fs_devices->fsid == sb->fsid
348 * fs_devices->metadata_uuid == sb->metadata_uuid
349 *
350 * - When in-memory fs_devices->temp_fsid is true
351 * fs_devices->fsid = random
352 * fs_devices->metadata_uuid == sb->fsid
353 */
354 u8 metadata_uuid[BTRFS_FSID_SIZE];
355
356 struct list_head fs_list;
357
358 /*
359 * Number of devices under this fsid including missing and
360 * replace-target device and excludes seed devices.
361 */
362 u64 num_devices;
363
364 /*
365 * The number of devices that successfully opened, including
366 * replace-target, excludes seed devices.
367 */
368 u64 open_devices;
369
370 /* The number of devices that are under the chunk allocation list. */
371 u64 rw_devices;
372
373 /* Count of missing devices under this fsid excluding seed device. */
374 u64 missing_devices;
375 u64 total_rw_bytes;
376
377 /*
378 * Count of devices from btrfs_super_block::num_devices for this fsid,
379 * which includes the seed device, excludes the transient replace-target
380 * device.
381 */
382 u64 total_devices;
383
384 /* Highest generation number of seen devices */
385 u64 latest_generation;
386
387 /*
388 * The mount device or a device with highest generation after removal
389 * or replace.
390 */
391 struct btrfs_device *latest_dev;
392
393 /*
394 * All of the devices in the filesystem, protected by a mutex so we can
395 * safely walk it to write out the super blocks without worrying about
396 * adding/removing by the multi-device code. Scrubbing super block can
397 * kick off supers writing by holding this mutex lock.
398 */
399 struct mutex device_list_mutex;
400
401 /* List of all devices, protected by device_list_mutex */
402 struct list_head devices;
403
404 /* Devices which can satisfy space allocation. Protected by * chunk_mutex. */
405 struct list_head alloc_list;
406
407 struct list_head seed_list;
408
409 /* Count fs-devices opened. */
410 int opened;
411
412 /* Set when we find or add a device that doesn't have the nonrot flag set. */
413 bool rotating;
414 /* Devices support TRIM/discard commands. */
415 bool discardable;
416 /* The filesystem is a seed filesystem. */
417 bool seeding;
418 /* The mount needs to use a randomly generated fsid. */
419 bool temp_fsid;
420
421 struct btrfs_fs_info *fs_info;
422 /* sysfs kobjects */
423 struct kobject fsid_kobj;
424 struct kobject *devices_kobj;
425 struct kobject *devinfo_kobj;
426 struct completion kobj_unregister;
427
428 enum btrfs_chunk_allocation_policy chunk_alloc_policy;
429
430 /* Policy used to read the mirrored stripes. */
431 enum btrfs_read_policy read_policy;
432
433#ifdef CONFIG_BTRFS_EXPERIMENTAL
434 /* Checksum mode - offload it or do it synchronously. */
435 enum btrfs_offload_csum_mode offload_csum_mode;
436#endif
437};
438
439#define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \
440 - sizeof(struct btrfs_chunk)) \
441 / sizeof(struct btrfs_stripe) + 1)
442
443#define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \
444 - 2 * sizeof(struct btrfs_disk_key) \
445 - 2 * sizeof(struct btrfs_chunk)) \
446 / sizeof(struct btrfs_stripe) + 1)
447
448struct btrfs_io_stripe {
449 struct btrfs_device *dev;
450 /* Block mapping. */
451 u64 physical;
452 u64 length;
453 bool rst_search_commit_root;
454 /* For the endio handler. */
455 struct btrfs_io_context *bioc;
456};
457
458struct btrfs_discard_stripe {
459 struct btrfs_device *dev;
460 u64 physical;
461 u64 length;
462};
463
464/*
465 * Context for IO subsmission for device stripe.
466 *
467 * - Track the unfinished mirrors for mirror based profiles
468 * Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
469 *
470 * - Contain the logical -> physical mapping info
471 * Used by submit_stripe_bio() for mapping logical bio
472 * into physical device address.
473 *
474 * - Contain device replace info
475 * Used by handle_ops_on_dev_replace() to copy logical bios
476 * into the new device.
477 *
478 * - Contain RAID56 full stripe logical bytenrs
479 */
480struct btrfs_io_context {
481 refcount_t refs;
482 struct btrfs_fs_info *fs_info;
483 /* Taken from struct btrfs_chunk_map::type. */
484 u64 map_type;
485 struct bio *orig_bio;
486 atomic_t error;
487 u16 max_errors;
488
489 u64 logical;
490 u64 size;
491 /* Raid stripe tree ordered entry. */
492 struct list_head rst_ordered_entry;
493
494 /*
495 * The total number of stripes, including the extra duplicated
496 * stripe for replace.
497 */
498 u16 num_stripes;
499
500 /*
501 * The mirror_num of this bioc.
502 *
503 * This is for reads which use 0 as mirror_num, thus we should return a
504 * valid mirror_num (>0) for the reader.
505 */
506 u16 mirror_num;
507
508 /*
509 * The following two members are for dev-replace case only.
510 *
511 * @replace_nr_stripes: Number of duplicated stripes which need to be
512 * written to replace target.
513 * Should be <= 2 (2 for DUP, otherwise <= 1).
514 * @replace_stripe_src: The array indicates where the duplicated stripes
515 * are from.
516 *
517 * The @replace_stripe_src[] array is mostly for RAID56 cases.
518 * As non-RAID56 stripes share the same contents of the mapped range,
519 * thus no need to bother where the duplicated ones are from.
520 *
521 * But for RAID56 case, all stripes contain different contents, thus
522 * we need a way to know the mapping.
523 *
524 * There is an example for the two members, using a RAID5 write:
525 *
526 * num_stripes: 4 (3 + 1 duplicated write)
527 * stripes[0]: dev = devid 1, physical = X
528 * stripes[1]: dev = devid 2, physical = Y
529 * stripes[2]: dev = devid 3, physical = Z
530 * stripes[3]: dev = devid 0, physical = Y
531 *
532 * replace_nr_stripes = 1
533 * replace_stripe_src = 1 <- Means stripes[1] is involved in replace.
534 * The duplicated stripe index would be
535 * (@num_stripes - 1).
536 *
537 * Note, that we can still have cases replace_nr_stripes = 2 for DUP.
538 * In that case, all stripes share the same content, thus we don't
539 * need to bother @replace_stripe_src value at all.
540 */
541 u16 replace_nr_stripes;
542 s16 replace_stripe_src;
543 /*
544 * Logical bytenr of the full stripe start, only for RAID56 cases.
545 *
546 * When this value is set to other than (u64)-1, the stripes[] should
547 * follow this pattern:
548 *
549 * (real_stripes = num_stripes - replace_nr_stripes)
550 * (data_stripes = (is_raid6) ? (real_stripes - 2) : (real_stripes - 1))
551 *
552 * stripes[0]: The first data stripe
553 * stripes[1]: The second data stripe
554 * ...
555 * stripes[data_stripes - 1]: The last data stripe
556 * stripes[data_stripes]: The P stripe
557 * stripes[data_stripes + 1]: The Q stripe (only for RAID6).
558 */
559 u64 full_stripe_logical;
560 struct btrfs_io_stripe stripes[];
561};
562
563struct btrfs_device_info {
564 struct btrfs_device *dev;
565 u64 dev_offset;
566 u64 max_avail;
567 u64 total_avail;
568};
569
570struct btrfs_raid_attr {
571 u8 sub_stripes; /* sub_stripes info for map */
572 u8 dev_stripes; /* stripes per dev */
573 u8 devs_max; /* max devs to use */
574 u8 devs_min; /* min devs needed */
575 u8 tolerated_failures; /* max tolerated fail devs */
576 u8 devs_increment; /* ndevs has to be a multiple of this */
577 u8 ncopies; /* how many copies to data has */
578 u8 nparity; /* number of stripes worth of bytes to store
579 * parity information */
580 u8 mindev_error; /* error code if min devs requisite is unmet */
581 const char raid_name[8]; /* name of the raid */
582 u64 bg_flag; /* block group flag of the raid */
583};
584
585extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
586
587struct btrfs_chunk_map {
588 struct rb_node rb_node;
589 /* For mount time dev extent verification. */
590 int verified_stripes;
591 refcount_t refs;
592 u64 start;
593 u64 chunk_len;
594 u64 stripe_size;
595 u64 type;
596 int io_align;
597 int io_width;
598 int num_stripes;
599 int sub_stripes;
600 struct btrfs_io_stripe stripes[];
601};
602
603#define btrfs_chunk_map_size(n) (sizeof(struct btrfs_chunk_map) + \
604 (sizeof(struct btrfs_io_stripe) * (n)))
605
606static inline void btrfs_free_chunk_map(struct btrfs_chunk_map *map)
607{
608 if (map && refcount_dec_and_test(&map->refs)) {
609 ASSERT(RB_EMPTY_NODE(&map->rb_node));
610 kfree(map);
611 }
612}
613
614struct btrfs_balance_control {
615 struct btrfs_balance_args data;
616 struct btrfs_balance_args meta;
617 struct btrfs_balance_args sys;
618
619 u64 flags;
620
621 struct btrfs_balance_progress stat;
622};
623
624/*
625 * Search for a given device by the set parameters
626 */
627struct btrfs_dev_lookup_args {
628 u64 devid;
629 u8 *uuid;
630 u8 *fsid;
631 bool missing;
632};
633
634/* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
635#define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
636
637#define BTRFS_DEV_LOOKUP_ARGS(name) \
638 struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
639
640enum btrfs_map_op {
641 BTRFS_MAP_READ,
642 BTRFS_MAP_WRITE,
643 BTRFS_MAP_GET_READ_MIRRORS,
644};
645
646static inline enum btrfs_map_op btrfs_op(struct bio *bio)
647{
648 switch (bio_op(bio)) {
649 case REQ_OP_WRITE:
650 case REQ_OP_ZONE_APPEND:
651 return BTRFS_MAP_WRITE;
652 default:
653 WARN_ON_ONCE(1);
654 fallthrough;
655 case REQ_OP_READ:
656 return BTRFS_MAP_READ;
657 }
658}
659
660static inline unsigned long btrfs_chunk_item_size(int num_stripes)
661{
662 ASSERT(num_stripes);
663 return sizeof(struct btrfs_chunk) +
664 sizeof(struct btrfs_stripe) * (num_stripes - 1);
665}
666
667/*
668 * Do the type safe conversion from stripe_nr to offset inside the chunk.
669 *
670 * @stripe_nr is u32, with left shift it can overflow u32 for chunks larger
671 * than 4G. This does the proper type cast to avoid overflow.
672 */
673static inline u64 btrfs_stripe_nr_to_offset(u32 stripe_nr)
674{
675 return (u64)stripe_nr << BTRFS_STRIPE_LEN_SHIFT;
676}
677
678void btrfs_get_bioc(struct btrfs_io_context *bioc);
679void btrfs_put_bioc(struct btrfs_io_context *bioc);
680int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
681 u64 logical, u64 *length,
682 struct btrfs_io_context **bioc_ret,
683 struct btrfs_io_stripe *smap, int *mirror_num_ret);
684int btrfs_map_repair_block(struct btrfs_fs_info *fs_info,
685 struct btrfs_io_stripe *smap, u64 logical,
686 u32 length, int mirror_num);
687struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,
688 u64 logical, u64 *length_ret,
689 u32 *num_stripes);
690int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
691int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
692struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
693 u64 type);
694void btrfs_mapping_tree_free(struct btrfs_fs_info *fs_info);
695int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
696 blk_mode_t flags, void *holder);
697struct btrfs_device *btrfs_scan_one_device(const char *path, blk_mode_t flags,
698 bool mount_arg_dev);
699int btrfs_forget_devices(dev_t devt);
700void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
701void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
702void btrfs_assign_next_active_device(struct btrfs_device *device,
703 struct btrfs_device *this_dev);
704struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
705 u64 devid,
706 const char *devpath);
707int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
708 struct btrfs_dev_lookup_args *args,
709 const char *path);
710struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
711 const u64 *devid, const u8 *uuid,
712 const char *path);
713void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
714int btrfs_rm_device(struct btrfs_fs_info *fs_info,
715 struct btrfs_dev_lookup_args *args,
716 struct file **bdev_file);
717void __exit btrfs_cleanup_fs_uuids(void);
718int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
719int btrfs_grow_device(struct btrfs_trans_handle *trans,
720 struct btrfs_device *device, u64 new_size);
721struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
722 const struct btrfs_dev_lookup_args *args);
723int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
724int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
725int btrfs_balance(struct btrfs_fs_info *fs_info,
726 struct btrfs_balance_control *bctl,
727 struct btrfs_ioctl_balance_args *bargs);
728void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
729int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
730int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
731int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
732int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
733int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
734bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
735void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
736int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
737 struct btrfs_ioctl_get_dev_stats *stats);
738int btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
739int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
740int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
741void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
742void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
743void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
744unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
745 u64 logical);
746u64 btrfs_calc_stripe_length(const struct btrfs_chunk_map *map);
747int btrfs_nr_parity_stripes(u64 type);
748int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
749 struct btrfs_block_group *bg);
750int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
751
752#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
753struct btrfs_chunk_map *btrfs_alloc_chunk_map(int num_stripes, gfp_t gfp);
754int btrfs_add_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
755#endif
756
757struct btrfs_chunk_map *btrfs_find_chunk_map(struct btrfs_fs_info *fs_info,
758 u64 logical, u64 length);
759struct btrfs_chunk_map *btrfs_find_chunk_map_nolock(struct btrfs_fs_info *fs_info,
760 u64 logical, u64 length);
761struct btrfs_chunk_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
762 u64 logical, u64 length);
763void btrfs_remove_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
764void btrfs_release_disk_super(struct btrfs_super_block *super);
765
766static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
767 int index)
768{
769 atomic_inc(dev->dev_stat_values + index);
770 /*
771 * This memory barrier orders stores updating statistics before stores
772 * updating dev_stats_ccnt.
773 *
774 * It pairs with smp_rmb() in btrfs_run_dev_stats().
775 */
776 smp_mb__before_atomic();
777 atomic_inc(&dev->dev_stats_ccnt);
778}
779
780static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
781 int index)
782{
783 return atomic_read(dev->dev_stat_values + index);
784}
785
786static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
787 int index)
788{
789 int ret;
790
791 ret = atomic_xchg(dev->dev_stat_values + index, 0);
792 /*
793 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
794 * - RMW operations that have a return value are fully ordered;
795 *
796 * This implicit memory barriers is paired with the smp_rmb in
797 * btrfs_run_dev_stats
798 */
799 atomic_inc(&dev->dev_stats_ccnt);
800 return ret;
801}
802
803static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
804 int index, unsigned long val)
805{
806 atomic_set(dev->dev_stat_values + index, val);
807 /*
808 * This memory barrier orders stores updating statistics before stores
809 * updating dev_stats_ccnt.
810 *
811 * It pairs with smp_rmb() in btrfs_run_dev_stats().
812 */
813 smp_mb__before_atomic();
814 atomic_inc(&dev->dev_stats_ccnt);
815}
816
817static inline const char *btrfs_dev_name(const struct btrfs_device *device)
818{
819 if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
820 return "<missing disk>";
821 else
822 return rcu_str_deref(device->name);
823}
824
825void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
826
827struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
828bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
829 struct btrfs_device *failing_dev);
830void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, struct btrfs_device *device);
831
832enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
833int btrfs_bg_type_to_factor(u64 flags);
834const char *btrfs_bg_type_to_raid_name(u64 flags);
835int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
836bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);
837
838bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr);
839const u8 *btrfs_sb_fsid_ptr(const struct btrfs_super_block *sb);
840
841#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
842struct btrfs_io_context *alloc_btrfs_io_context(struct btrfs_fs_info *fs_info,
843 u64 logical, u16 total_stripes);
844#endif
845
846#endif
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#ifndef __BTRFS_VOLUMES_
20#define __BTRFS_VOLUMES_
21
22#include <linux/bio.h>
23#include <linux/sort.h>
24#include <linux/btrfs.h>
25#include "async-thread.h"
26
27#define BTRFS_STRIPE_LEN (64 * 1024)
28
29struct buffer_head;
30struct btrfs_pending_bios {
31 struct bio *head;
32 struct bio *tail;
33};
34
35struct btrfs_device {
36 struct list_head dev_list;
37 struct list_head dev_alloc_list;
38 struct btrfs_fs_devices *fs_devices;
39 struct btrfs_root *dev_root;
40
41 /* regular prio bios */
42 struct btrfs_pending_bios pending_bios;
43 /* WRITE_SYNC bios */
44 struct btrfs_pending_bios pending_sync_bios;
45
46 u64 generation;
47 int running_pending;
48 int writeable;
49 int in_fs_metadata;
50 int missing;
51 int can_discard;
52 int is_tgtdev_for_dev_replace;
53
54 spinlock_t io_lock;
55 /* the mode sent to blkdev_get */
56 fmode_t mode;
57
58 struct block_device *bdev;
59
60
61 struct rcu_string *name;
62
63 /* the internal btrfs device id */
64 u64 devid;
65
66 /* size of the device */
67 u64 total_bytes;
68
69 /* size of the disk */
70 u64 disk_total_bytes;
71
72 /* bytes used */
73 u64 bytes_used;
74
75 /* optimal io alignment for this device */
76 u32 io_align;
77
78 /* optimal io width for this device */
79 u32 io_width;
80 /* type and info about this device */
81 u64 type;
82
83 /* minimal io size for this device */
84 u32 sector_size;
85
86
87 /* physical drive uuid (or lvm uuid) */
88 u8 uuid[BTRFS_UUID_SIZE];
89
90 /* for sending down flush barriers */
91 int nobarriers;
92 struct bio *flush_bio;
93 struct completion flush_wait;
94
95 /* per-device scrub information */
96 struct scrub_ctx *scrub_device;
97
98 struct btrfs_work work;
99 struct rcu_head rcu;
100 struct work_struct rcu_work;
101
102 /* readahead state */
103 spinlock_t reada_lock;
104 atomic_t reada_in_flight;
105 u64 reada_next;
106 struct reada_zone *reada_curr_zone;
107 struct radix_tree_root reada_zones;
108 struct radix_tree_root reada_extents;
109
110
111 /* disk I/O failure stats. For detailed description refer to
112 * enum btrfs_dev_stat_values in ioctl.h */
113 int dev_stats_valid;
114 int dev_stats_dirty; /* counters need to be written to disk */
115 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
116};
117
118struct btrfs_fs_devices {
119 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
120
121 /* the device with this id has the most recent copy of the super */
122 u64 latest_devid;
123 u64 latest_trans;
124 u64 num_devices;
125 u64 open_devices;
126 u64 rw_devices;
127 u64 missing_devices;
128 u64 total_rw_bytes;
129 u64 num_can_discard;
130 u64 total_devices;
131 struct block_device *latest_bdev;
132
133 /* all of the devices in the FS, protected by a mutex
134 * so we can safely walk it to write out the supers without
135 * worrying about add/remove by the multi-device code.
136 * Scrubbing super can kick off supers writing by holding
137 * this mutex lock.
138 */
139 struct mutex device_list_mutex;
140 struct list_head devices;
141
142 /* devices not currently being allocated */
143 struct list_head alloc_list;
144 struct list_head list;
145
146 struct btrfs_fs_devices *seed;
147 int seeding;
148
149 int opened;
150
151 /* set when we find or add a device that doesn't have the
152 * nonrot flag set
153 */
154 int rotating;
155};
156
157#define BTRFS_BIO_INLINE_CSUM_SIZE 64
158
159/*
160 * we need the mirror number and stripe index to be passed around
161 * the call chain while we are processing end_io (especially errors).
162 * Really, what we need is a btrfs_bio structure that has this info
163 * and is properly sized with its stripe array, but we're not there
164 * quite yet. We have our own btrfs bioset, and all of the bios
165 * we allocate are actually btrfs_io_bios. We'll cram as much of
166 * struct btrfs_bio as we can into this over time.
167 */
168typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err);
169struct btrfs_io_bio {
170 unsigned long mirror_num;
171 unsigned long stripe_index;
172 u8 *csum;
173 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
174 u8 *csum_allocated;
175 btrfs_io_bio_end_io_t *end_io;
176 struct bio bio;
177};
178
179static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
180{
181 return container_of(bio, struct btrfs_io_bio, bio);
182}
183
184struct btrfs_bio_stripe {
185 struct btrfs_device *dev;
186 u64 physical;
187 u64 length; /* only used for discard mappings */
188};
189
190struct btrfs_bio;
191typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err);
192
193struct btrfs_bio {
194 atomic_t stripes_pending;
195 struct btrfs_fs_info *fs_info;
196 bio_end_io_t *end_io;
197 struct bio *orig_bio;
198 void *private;
199 atomic_t error;
200 int max_errors;
201 int num_stripes;
202 int mirror_num;
203 struct btrfs_bio_stripe stripes[];
204};
205
206struct btrfs_device_info {
207 struct btrfs_device *dev;
208 u64 dev_offset;
209 u64 max_avail;
210 u64 total_avail;
211};
212
213struct btrfs_raid_attr {
214 int sub_stripes; /* sub_stripes info for map */
215 int dev_stripes; /* stripes per dev */
216 int devs_max; /* max devs to use */
217 int devs_min; /* min devs needed */
218 int devs_increment; /* ndevs has to be a multiple of this */
219 int ncopies; /* how many copies to data has */
220};
221
222struct map_lookup {
223 u64 type;
224 int io_align;
225 int io_width;
226 int stripe_len;
227 int sector_size;
228 int num_stripes;
229 int sub_stripes;
230 struct btrfs_bio_stripe stripes[];
231};
232
233#define map_lookup_size(n) (sizeof(struct map_lookup) + \
234 (sizeof(struct btrfs_bio_stripe) * (n)))
235
236/*
237 * Restriper's general type filter
238 */
239#define BTRFS_BALANCE_DATA (1ULL << 0)
240#define BTRFS_BALANCE_SYSTEM (1ULL << 1)
241#define BTRFS_BALANCE_METADATA (1ULL << 2)
242
243#define BTRFS_BALANCE_TYPE_MASK (BTRFS_BALANCE_DATA | \
244 BTRFS_BALANCE_SYSTEM | \
245 BTRFS_BALANCE_METADATA)
246
247#define BTRFS_BALANCE_FORCE (1ULL << 3)
248#define BTRFS_BALANCE_RESUME (1ULL << 4)
249
250/*
251 * Balance filters
252 */
253#define BTRFS_BALANCE_ARGS_PROFILES (1ULL << 0)
254#define BTRFS_BALANCE_ARGS_USAGE (1ULL << 1)
255#define BTRFS_BALANCE_ARGS_DEVID (1ULL << 2)
256#define BTRFS_BALANCE_ARGS_DRANGE (1ULL << 3)
257#define BTRFS_BALANCE_ARGS_VRANGE (1ULL << 4)
258
259/*
260 * Profile changing flags. When SOFT is set we won't relocate chunk if
261 * it already has the target profile (even though it may be
262 * half-filled).
263 */
264#define BTRFS_BALANCE_ARGS_CONVERT (1ULL << 8)
265#define BTRFS_BALANCE_ARGS_SOFT (1ULL << 9)
266
267struct btrfs_balance_args;
268struct btrfs_balance_progress;
269struct btrfs_balance_control {
270 struct btrfs_fs_info *fs_info;
271
272 struct btrfs_balance_args data;
273 struct btrfs_balance_args meta;
274 struct btrfs_balance_args sys;
275
276 u64 flags;
277
278 struct btrfs_balance_progress stat;
279};
280
281int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
282 u64 end, u64 *length);
283
284#define btrfs_bio_size(n) (sizeof(struct btrfs_bio) + \
285 (sizeof(struct btrfs_bio_stripe) * (n)))
286
287int btrfs_map_block(struct btrfs_fs_info *fs_info, int rw,
288 u64 logical, u64 *length,
289 struct btrfs_bio **bbio_ret, int mirror_num);
290int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
291 u64 chunk_start, u64 physical, u64 devid,
292 u64 **logical, int *naddrs, int *stripe_len);
293int btrfs_read_sys_array(struct btrfs_root *root);
294int btrfs_read_chunk_tree(struct btrfs_root *root);
295int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
296 struct btrfs_root *extent_root, u64 type);
297void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
298void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
299int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
300 int mirror_num, int async_submit);
301int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
302 fmode_t flags, void *holder);
303int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
304 struct btrfs_fs_devices **fs_devices_ret);
305int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
306void btrfs_close_extra_devices(struct btrfs_fs_info *fs_info,
307 struct btrfs_fs_devices *fs_devices, int step);
308int btrfs_find_device_missing_or_by_path(struct btrfs_root *root,
309 char *device_path,
310 struct btrfs_device **device);
311struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
312 const u64 *devid,
313 const u8 *uuid);
314int btrfs_rm_device(struct btrfs_root *root, char *device_path);
315void btrfs_cleanup_fs_uuids(void);
316int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
317int btrfs_grow_device(struct btrfs_trans_handle *trans,
318 struct btrfs_device *device, u64 new_size);
319struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
320 u8 *uuid, u8 *fsid);
321int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
322int btrfs_init_new_device(struct btrfs_root *root, char *path);
323int btrfs_init_dev_replace_tgtdev(struct btrfs_root *root, char *device_path,
324 struct btrfs_device **device_out);
325int btrfs_balance(struct btrfs_balance_control *bctl,
326 struct btrfs_ioctl_balance_args *bargs);
327int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
328int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
329int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
330int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
331int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
332int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
333int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset);
334int find_free_dev_extent(struct btrfs_trans_handle *trans,
335 struct btrfs_device *device, u64 num_bytes,
336 u64 *start, u64 *max_avail);
337void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
338int btrfs_get_dev_stats(struct btrfs_root *root,
339 struct btrfs_ioctl_get_dev_stats *stats);
340void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
341int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
342int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
343 struct btrfs_fs_info *fs_info);
344void btrfs_rm_dev_replace_srcdev(struct btrfs_fs_info *fs_info,
345 struct btrfs_device *srcdev);
346void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
347 struct btrfs_device *tgtdev);
348void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info,
349 struct btrfs_device *tgtdev);
350int btrfs_scratch_superblock(struct btrfs_device *device);
351int btrfs_is_parity_mirror(struct btrfs_mapping_tree *map_tree,
352 u64 logical, u64 len, int mirror_num);
353unsigned long btrfs_full_stripe_len(struct btrfs_root *root,
354 struct btrfs_mapping_tree *map_tree,
355 u64 logical);
356int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
357 struct btrfs_root *extent_root,
358 u64 chunk_offset, u64 chunk_size);
359static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
360 int index)
361{
362 atomic_inc(dev->dev_stat_values + index);
363 dev->dev_stats_dirty = 1;
364}
365
366static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
367 int index)
368{
369 return atomic_read(dev->dev_stat_values + index);
370}
371
372static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
373 int index)
374{
375 int ret;
376
377 ret = atomic_xchg(dev->dev_stat_values + index, 0);
378 dev->dev_stats_dirty = 1;
379 return ret;
380}
381
382static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
383 int index, unsigned long val)
384{
385 atomic_set(dev->dev_stat_values + index, val);
386 dev->dev_stats_dirty = 1;
387}
388
389static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
390 int index)
391{
392 btrfs_dev_stat_set(dev, index, 0);
393}
394#endif