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/sort.h>
10#include <linux/btrfs.h>
11#include "async-thread.h"
12#include "messages.h"
13#include "tree-checker.h"
14#include "rcu-string.h"
15
16#define BTRFS_MAX_DATA_CHUNK_SIZE (10ULL * SZ_1G)
17
18extern struct mutex uuid_mutex;
19
20#define BTRFS_STRIPE_LEN SZ_64K
21
22/* Used by sanity check for btrfs_raid_types. */
23#define const_ffs(n) (__builtin_ctzll(n) + 1)
24
25/*
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.
30 */
31static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) <
32 const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0));
33static_assert(const_ilog2(BTRFS_BLOCK_GROUP_RAID0) >
34 ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK));
35
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))
39
40enum btrfs_raid_types {
41 /* SINGLE is the special one as it doesn't have on-disk bit. */
42 BTRFS_RAID_SINGLE = 0,
43
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),
52
53 BTRFS_NR_RAID_TYPES
54};
55
56struct btrfs_io_geometry {
57 /* remaining bytes before crossing a stripe */
58 u64 len;
59 /* offset of logical address in chunk */
60 u64 offset;
61 /* length of single IO stripe */
62 u32 stripe_len;
63 /* offset of address in stripe */
64 u32 stripe_offset;
65 /* number of stripe where address falls */
66 u64 stripe_nr;
67 /* offset of raid56 stripe into the chunk */
68 u64 raid56_stripe_offset;
69};
70
71/*
72 * Use sequence counter to get consistent device stat data on
73 * 32-bit processors.
74 */
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)
80#else
81#define btrfs_device_data_ordered_init(device) do { } while (0)
82#endif
83
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)
90
91struct btrfs_zoned_device_info;
92
93struct btrfs_device {
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;
99
100 struct rcu_string __rcu *name;
101
102 u64 generation;
103
104 struct block_device *bdev;
105
106 struct btrfs_zoned_device_info *zone_info;
107
108 /* the mode sent to blkdev_get */
109 fmode_t mode;
110
111 /*
112 * Device's major-minor number. Must be set even if the device is not
113 * opened (bdev == NULL), unless the device is missing.
114 */
115 dev_t devt;
116 unsigned long dev_state;
117 blk_status_t last_flush_error;
118
119#ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
120 seqcount_t data_seqcount;
121#endif
122
123 /* the internal btrfs device id */
124 u64 devid;
125
126 /* size of the device in memory */
127 u64 total_bytes;
128
129 /* size of the device on disk */
130 u64 disk_total_bytes;
131
132 /* bytes used */
133 u64 bytes_used;
134
135 /* optimal io alignment for this device */
136 u32 io_align;
137
138 /* optimal io width for this device */
139 u32 io_width;
140 /* type and info about this device */
141 u64 type;
142
143 /* minimal io size for this device */
144 u32 sector_size;
145
146 /* physical drive uuid (or lvm uuid) */
147 u8 uuid[BTRFS_UUID_SIZE];
148
149 /*
150 * size of the device on the current transaction
151 *
152 * This variant is update when committing the transaction,
153 * and protected by chunk mutex
154 */
155 u64 commit_total_bytes;
156
157 /* bytes used on the current transaction */
158 u64 commit_bytes_used;
159
160 /* Bio used for flushing device barriers */
161 struct bio flush_bio;
162 struct completion flush_wait;
163
164 /* per-device scrub information */
165 struct scrub_ctx *scrub_ctx;
166
167 /* disk I/O failure stats. For detailed description refer to
168 * enum btrfs_dev_stat_values in ioctl.h */
169 int dev_stats_valid;
170
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];
174
175 struct extent_io_tree alloc_state;
176
177 struct completion kobj_unregister;
178 /* For sysfs/FSID/devinfo/devid/ */
179 struct kobject devid_kobj;
180
181 /* Bandwidth limit for scrub, in bytes */
182 u64 scrub_speed_max;
183};
184
185/*
186 * Block group or device which contains an active swapfile. Used for preventing
187 * unsafe operations while a swapfile is active.
188 *
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.
193 */
194struct btrfs_swapfile_pin {
195 struct rb_node node;
196 void *ptr;
197 struct inode *inode;
198 /*
199 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr
200 * points to a struct btrfs_device.
201 */
202 bool is_block_group;
203 /*
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).
206 */
207 int bg_extent_count;
208};
209
210/*
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.
213 */
214#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
215#define BTRFS_DEVICE_GETSET_FUNCS(name) \
216static inline u64 \
217btrfs_device_get_##name(const struct btrfs_device *dev) \
218{ \
219 u64 size; \
220 unsigned int seq; \
221 \
222 do { \
223 seq = read_seqcount_begin(&dev->data_seqcount); \
224 size = dev->name; \
225 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
226 return size; \
227} \
228 \
229static inline void \
230btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
231{ \
232 preempt_disable(); \
233 write_seqcount_begin(&dev->data_seqcount); \
234 dev->name = size; \
235 write_seqcount_end(&dev->data_seqcount); \
236 preempt_enable(); \
237}
238#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
239#define BTRFS_DEVICE_GETSET_FUNCS(name) \
240static inline u64 \
241btrfs_device_get_##name(const struct btrfs_device *dev) \
242{ \
243 u64 size; \
244 \
245 preempt_disable(); \
246 size = dev->name; \
247 preempt_enable(); \
248 return size; \
249} \
250 \
251static inline void \
252btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
253{ \
254 preempt_disable(); \
255 dev->name = size; \
256 preempt_enable(); \
257}
258#else
259#define BTRFS_DEVICE_GETSET_FUNCS(name) \
260static inline u64 \
261btrfs_device_get_##name(const struct btrfs_device *dev) \
262{ \
263 return dev->name; \
264} \
265 \
266static inline void \
267btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
268{ \
269 dev->name = size; \
270}
271#endif
272
273BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
274BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
275BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
276
277enum btrfs_chunk_allocation_policy {
278 BTRFS_CHUNK_ALLOC_REGULAR,
279 BTRFS_CHUNK_ALLOC_ZONED,
280};
281
282/*
283 * Read policies for mirrored block group profiles, read picks the stripe based
284 * on these policies.
285 */
286enum btrfs_read_policy {
287 /* Use process PID to choose the stripe */
288 BTRFS_READ_POLICY_PID,
289 BTRFS_NR_READ_POLICY,
290};
291
292struct btrfs_fs_devices {
293 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
294 u8 metadata_uuid[BTRFS_FSID_SIZE];
295 bool fsid_change;
296 struct list_head fs_list;
297
298 /*
299 * Number of devices under this fsid including missing and
300 * replace-target device and excludes seed devices.
301 */
302 u64 num_devices;
303
304 /*
305 * The number of devices that successfully opened, including
306 * replace-target, excludes seed devices.
307 */
308 u64 open_devices;
309
310 /* The number of devices that are under the chunk allocation list. */
311 u64 rw_devices;
312
313 /* Count of missing devices under this fsid excluding seed device. */
314 u64 missing_devices;
315 u64 total_rw_bytes;
316
317 /*
318 * Count of devices from btrfs_super_block::num_devices for this fsid,
319 * which includes the seed device, excludes the transient replace-target
320 * device.
321 */
322 u64 total_devices;
323
324 /* Highest generation number of seen devices */
325 u64 latest_generation;
326
327 /*
328 * The mount device or a device with highest generation after removal
329 * or replace.
330 */
331 struct btrfs_device *latest_dev;
332
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
337 * this mutex lock.
338 */
339 struct mutex device_list_mutex;
340
341 /* List of all devices, protected by device_list_mutex */
342 struct list_head devices;
343
344 /*
345 * Devices which can satisfy space allocation. Protected by
346 * chunk_mutex
347 */
348 struct list_head alloc_list;
349
350 struct list_head seed_list;
351 bool seeding;
352
353 int opened;
354
355 /* set when we find or add a device that doesn't have the
356 * nonrot flag set
357 */
358 bool rotating;
359 /* Devices support TRIM/discard commands */
360 bool discardable;
361
362 struct btrfs_fs_info *fs_info;
363 /* sysfs kobjects */
364 struct kobject fsid_kobj;
365 struct kobject *devices_kobj;
366 struct kobject *devinfo_kobj;
367 struct completion kobj_unregister;
368
369 enum btrfs_chunk_allocation_policy chunk_alloc_policy;
370
371 /* Policy used to read the mirrored stripes */
372 enum btrfs_read_policy read_policy;
373};
374
375#define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \
376 - sizeof(struct btrfs_chunk)) \
377 / sizeof(struct btrfs_stripe) + 1)
378
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)
383
384struct btrfs_io_stripe {
385 struct btrfs_device *dev;
386 union {
387 /* Block mapping */
388 u64 physical;
389 /* For the endio handler */
390 struct btrfs_io_context *bioc;
391 };
392};
393
394struct btrfs_discard_stripe {
395 struct btrfs_device *dev;
396 u64 physical;
397 u64 length;
398};
399
400/*
401 * Context for IO subsmission for device stripe.
402 *
403 * - Track the unfinished mirrors for mirror based profiles
404 * Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
405 *
406 * - Contain the logical -> physical mapping info
407 * Used by submit_stripe_bio() for mapping logical bio
408 * into physical device address.
409 *
410 * - Contain device replace info
411 * Used by handle_ops_on_dev_replace() to copy logical bios
412 * into the new device.
413 *
414 * - Contain RAID56 full stripe logical bytenrs
415 */
416struct btrfs_io_context {
417 refcount_t refs;
418 struct btrfs_fs_info *fs_info;
419 u64 map_type; /* get from map_lookup->type */
420 struct bio *orig_bio;
421 atomic_t error;
422 int max_errors;
423 int num_stripes;
424 int mirror_num;
425 int num_tgtdevs;
426 int *tgtdev_map;
427 /*
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
431 */
432 u64 *raid_map;
433 struct btrfs_io_stripe stripes[];
434};
435
436struct btrfs_device_info {
437 struct btrfs_device *dev;
438 u64 dev_offset;
439 u64 max_avail;
440 u64 total_avail;
441};
442
443struct 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 */
456};
457
458extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
459
460struct map_lookup {
461 u64 type;
462 int io_align;
463 int io_width;
464 u32 stripe_len;
465 int num_stripes;
466 int sub_stripes;
467 int verified_stripes; /* For mount time dev extent verification */
468 struct btrfs_io_stripe stripes[];
469};
470
471#define map_lookup_size(n) (sizeof(struct map_lookup) + \
472 (sizeof(struct btrfs_io_stripe) * (n)))
473
474struct btrfs_balance_args;
475struct btrfs_balance_progress;
476struct btrfs_balance_control {
477 struct btrfs_balance_args data;
478 struct btrfs_balance_args meta;
479 struct btrfs_balance_args sys;
480
481 u64 flags;
482
483 struct btrfs_balance_progress stat;
484};
485
486/*
487 * Search for a given device by the set parameters
488 */
489struct btrfs_dev_lookup_args {
490 u64 devid;
491 u8 *uuid;
492 u8 *fsid;
493 bool missing;
494};
495
496/* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
497#define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
498
499#define BTRFS_DEV_LOOKUP_ARGS(name) \
500 struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
501
502enum btrfs_map_op {
503 BTRFS_MAP_READ,
504 BTRFS_MAP_WRITE,
505 BTRFS_MAP_DISCARD,
506 BTRFS_MAP_GET_READ_MIRRORS,
507};
508
509static inline enum btrfs_map_op btrfs_op(struct bio *bio)
510{
511 switch (bio_op(bio)) {
512 case REQ_OP_DISCARD:
513 return BTRFS_MAP_DISCARD;
514 case REQ_OP_WRITE:
515 case REQ_OP_ZONE_APPEND:
516 return BTRFS_MAP_WRITE;
517 default:
518 WARN_ON_ONCE(1);
519 fallthrough;
520 case REQ_OP_READ:
521 return BTRFS_MAP_READ;
522 }
523}
524
525static inline unsigned long btrfs_chunk_item_size(int num_stripes)
526{
527 ASSERT(num_stripes);
528 return sizeof(struct btrfs_chunk) +
529 sizeof(struct btrfs_stripe) * (num_stripes - 1);
530}
531
532void btrfs_get_bioc(struct btrfs_io_context *bioc);
533void btrfs_put_bioc(struct btrfs_io_context *bioc);
534int 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);
537int 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);
540int __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,
544 int need_raid_map);
545struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,
546 u64 logical, u64 *length_ret,
547 u32 *num_stripes);
548int 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);
551int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
552int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
553struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
554 u64 type);
555void btrfs_mapping_tree_free(struct extent_map_tree *tree);
556int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
557 fmode_t flags, void *holder);
558struct btrfs_device *btrfs_scan_one_device(const char *path,
559 fmode_t flags, void *holder);
560int btrfs_forget_devices(dev_t devt);
561void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
562void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
563void btrfs_assign_next_active_device(struct btrfs_device *device,
564 struct btrfs_device *this_dev);
565struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
566 u64 devid,
567 const char *devpath);
568int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
569 struct btrfs_dev_lookup_args *args,
570 const char *path);
571struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
572 const u64 *devid, const u8 *uuid,
573 const char *path);
574void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
575void btrfs_free_device(struct btrfs_device *device);
576int btrfs_rm_device(struct btrfs_fs_info *fs_info,
577 struct btrfs_dev_lookup_args *args,
578 struct block_device **bdev, fmode_t *mode);
579void __exit btrfs_cleanup_fs_uuids(void);
580int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
581int btrfs_grow_device(struct btrfs_trans_handle *trans,
582 struct btrfs_device *device, u64 new_size);
583struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
584 const struct btrfs_dev_lookup_args *args);
585int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
586int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
587int btrfs_balance(struct btrfs_fs_info *fs_info,
588 struct btrfs_balance_control *bctl,
589 struct btrfs_ioctl_balance_args *bargs);
590void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
591int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
592int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
593int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
594int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
595int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
596int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
597int btrfs_uuid_scan_kthread(void *data);
598bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
599int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes,
600 u64 *start, u64 *max_avail);
601void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
602int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
603 struct btrfs_ioctl_get_dev_stats *stats);
604int btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
605int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
606int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
607void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
608void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
609void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
610int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
611 u64 logical, u64 len);
612unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
613 u64 logical);
614u64 btrfs_calc_stripe_length(const struct extent_map *em);
615int btrfs_nr_parity_stripes(u64 type);
616int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
617 struct btrfs_block_group *bg);
618int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
619struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
620 u64 logical, u64 length);
621void btrfs_release_disk_super(struct btrfs_super_block *super);
622
623static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
624 int index)
625{
626 atomic_inc(dev->dev_stat_values + index);
627 /*
628 * This memory barrier orders stores updating statistics before stores
629 * updating dev_stats_ccnt.
630 *
631 * It pairs with smp_rmb() in btrfs_run_dev_stats().
632 */
633 smp_mb__before_atomic();
634 atomic_inc(&dev->dev_stats_ccnt);
635}
636
637static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
638 int index)
639{
640 return atomic_read(dev->dev_stat_values + index);
641}
642
643static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
644 int index)
645{
646 int ret;
647
648 ret = atomic_xchg(dev->dev_stat_values + index, 0);
649 /*
650 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
651 * - RMW operations that have a return value are fully ordered;
652 *
653 * This implicit memory barriers is paired with the smp_rmb in
654 * btrfs_run_dev_stats
655 */
656 atomic_inc(&dev->dev_stats_ccnt);
657 return ret;
658}
659
660static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
661 int index, unsigned long val)
662{
663 atomic_set(dev->dev_stat_values + index, val);
664 /*
665 * This memory barrier orders stores updating statistics before stores
666 * updating dev_stats_ccnt.
667 *
668 * It pairs with smp_rmb() in btrfs_run_dev_stats().
669 */
670 smp_mb__before_atomic();
671 atomic_inc(&dev->dev_stats_ccnt);
672}
673
674static inline const char *btrfs_dev_name(const struct btrfs_device *device)
675{
676 if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
677 return "<missing disk>";
678 else
679 return rcu_str_deref(device->name);
680}
681
682void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
683
684struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
685bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
686 struct btrfs_device *failing_dev);
687void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info,
688 struct block_device *bdev,
689 const char *device_path);
690
691enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
692int btrfs_bg_type_to_factor(u64 flags);
693const char *btrfs_bg_type_to_raid_name(u64 flags);
694int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
695bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);
696
697bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr);
698
699#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
27extern struct mutex uuid_mutex;
28
29#define BTRFS_STRIPE_LEN SZ_64K
30
31struct buffer_head;
32struct btrfs_pending_bios {
33 struct bio *head;
34 struct bio *tail;
35};
36
37/*
38 * Use sequence counter to get consistent device stat data on
39 * 32-bit processors.
40 */
41#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
42#include <linux/seqlock.h>
43#define __BTRFS_NEED_DEVICE_DATA_ORDERED
44#define btrfs_device_data_ordered_init(device) \
45 seqcount_init(&device->data_seqcount)
46#else
47#define btrfs_device_data_ordered_init(device) do { } while (0)
48#endif
49
50struct btrfs_device {
51 struct list_head dev_list;
52 struct list_head dev_alloc_list;
53 struct btrfs_fs_devices *fs_devices;
54
55 struct btrfs_root *dev_root;
56
57 struct rcu_string *name;
58
59 u64 generation;
60
61 spinlock_t io_lock ____cacheline_aligned;
62 int running_pending;
63 /* regular prio bios */
64 struct btrfs_pending_bios pending_bios;
65 /* WRITE_SYNC bios */
66 struct btrfs_pending_bios pending_sync_bios;
67
68 struct block_device *bdev;
69
70 /* the mode sent to blkdev_get */
71 fmode_t mode;
72
73 int writeable;
74 int in_fs_metadata;
75 int missing;
76 int can_discard;
77 int is_tgtdev_for_dev_replace;
78
79#ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
80 seqcount_t data_seqcount;
81#endif
82
83 /* the internal btrfs device id */
84 u64 devid;
85
86 /* size of the device in memory */
87 u64 total_bytes;
88
89 /* size of the device on disk */
90 u64 disk_total_bytes;
91
92 /* bytes used */
93 u64 bytes_used;
94
95 /* optimal io alignment for this device */
96 u32 io_align;
97
98 /* optimal io width for this device */
99 u32 io_width;
100 /* type and info about this device */
101 u64 type;
102
103 /* minimal io size for this device */
104 u32 sector_size;
105
106 /* physical drive uuid (or lvm uuid) */
107 u8 uuid[BTRFS_UUID_SIZE];
108
109 /*
110 * size of the device on the current transaction
111 *
112 * This variant is update when committing the transaction,
113 * and protected by device_list_mutex
114 */
115 u64 commit_total_bytes;
116
117 /* bytes used on the current transaction */
118 u64 commit_bytes_used;
119 /*
120 * used to manage the device which is resized
121 *
122 * It is protected by chunk_lock.
123 */
124 struct list_head resized_list;
125
126 /* for sending down flush barriers */
127 int nobarriers;
128 struct bio *flush_bio;
129 struct completion flush_wait;
130
131 /* per-device scrub information */
132 struct scrub_ctx *scrub_device;
133
134 struct btrfs_work work;
135 struct rcu_head rcu;
136 struct work_struct rcu_work;
137
138 /* readahead state */
139 spinlock_t reada_lock;
140 atomic_t reada_in_flight;
141 u64 reada_next;
142 struct reada_zone *reada_curr_zone;
143 struct radix_tree_root reada_zones;
144 struct radix_tree_root reada_extents;
145
146 /* disk I/O failure stats. For detailed description refer to
147 * enum btrfs_dev_stat_values in ioctl.h */
148 int dev_stats_valid;
149
150 /* Counter to record the change of device stats */
151 atomic_t dev_stats_ccnt;
152 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
153};
154
155/*
156 * If we read those variants at the context of their own lock, we needn't
157 * use the following helpers, reading them directly is safe.
158 */
159#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
160#define BTRFS_DEVICE_GETSET_FUNCS(name) \
161static inline u64 \
162btrfs_device_get_##name(const struct btrfs_device *dev) \
163{ \
164 u64 size; \
165 unsigned int seq; \
166 \
167 do { \
168 seq = read_seqcount_begin(&dev->data_seqcount); \
169 size = dev->name; \
170 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
171 return size; \
172} \
173 \
174static inline void \
175btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
176{ \
177 preempt_disable(); \
178 write_seqcount_begin(&dev->data_seqcount); \
179 dev->name = size; \
180 write_seqcount_end(&dev->data_seqcount); \
181 preempt_enable(); \
182}
183#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
184#define BTRFS_DEVICE_GETSET_FUNCS(name) \
185static inline u64 \
186btrfs_device_get_##name(const struct btrfs_device *dev) \
187{ \
188 u64 size; \
189 \
190 preempt_disable(); \
191 size = dev->name; \
192 preempt_enable(); \
193 return size; \
194} \
195 \
196static inline void \
197btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
198{ \
199 preempt_disable(); \
200 dev->name = size; \
201 preempt_enable(); \
202}
203#else
204#define BTRFS_DEVICE_GETSET_FUNCS(name) \
205static inline u64 \
206btrfs_device_get_##name(const struct btrfs_device *dev) \
207{ \
208 return dev->name; \
209} \
210 \
211static inline void \
212btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
213{ \
214 dev->name = size; \
215}
216#endif
217
218BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
219BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
220BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
221
222struct btrfs_fs_devices {
223 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
224
225 u64 num_devices;
226 u64 open_devices;
227 u64 rw_devices;
228 u64 missing_devices;
229 u64 total_rw_bytes;
230 u64 total_devices;
231 struct block_device *latest_bdev;
232
233 /* all of the devices in the FS, protected by a mutex
234 * so we can safely walk it to write out the supers without
235 * worrying about add/remove by the multi-device code.
236 * Scrubbing super can kick off supers writing by holding
237 * this mutex lock.
238 */
239 struct mutex device_list_mutex;
240 struct list_head devices;
241
242 struct list_head resized_devices;
243 /* devices not currently being allocated */
244 struct list_head alloc_list;
245 struct list_head list;
246
247 struct btrfs_fs_devices *seed;
248 int seeding;
249
250 int opened;
251
252 /* set when we find or add a device that doesn't have the
253 * nonrot flag set
254 */
255 int rotating;
256
257 struct btrfs_fs_info *fs_info;
258 /* sysfs kobjects */
259 struct kobject fsid_kobj;
260 struct kobject *device_dir_kobj;
261 struct completion kobj_unregister;
262};
263
264#define BTRFS_BIO_INLINE_CSUM_SIZE 64
265
266/*
267 * we need the mirror number and stripe index to be passed around
268 * the call chain while we are processing end_io (especially errors).
269 * Really, what we need is a btrfs_bio structure that has this info
270 * and is properly sized with its stripe array, but we're not there
271 * quite yet. We have our own btrfs bioset, and all of the bios
272 * we allocate are actually btrfs_io_bios. We'll cram as much of
273 * struct btrfs_bio as we can into this over time.
274 */
275typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err);
276struct btrfs_io_bio {
277 unsigned int mirror_num;
278 unsigned int stripe_index;
279 u64 logical;
280 u8 *csum;
281 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
282 u8 *csum_allocated;
283 btrfs_io_bio_end_io_t *end_io;
284 struct bio bio;
285};
286
287static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
288{
289 return container_of(bio, struct btrfs_io_bio, bio);
290}
291
292struct btrfs_bio_stripe {
293 struct btrfs_device *dev;
294 u64 physical;
295 u64 length; /* only used for discard mappings */
296};
297
298struct btrfs_bio;
299typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err);
300
301struct btrfs_bio {
302 atomic_t refs;
303 atomic_t stripes_pending;
304 struct btrfs_fs_info *fs_info;
305 u64 map_type; /* get from map_lookup->type */
306 bio_end_io_t *end_io;
307 struct bio *orig_bio;
308 unsigned long flags;
309 void *private;
310 atomic_t error;
311 int max_errors;
312 int num_stripes;
313 int mirror_num;
314 int num_tgtdevs;
315 int *tgtdev_map;
316 /*
317 * logical block numbers for the start of each stripe
318 * The last one or two are p/q. These are sorted,
319 * so raid_map[0] is the start of our full stripe
320 */
321 u64 *raid_map;
322 struct btrfs_bio_stripe stripes[];
323};
324
325struct btrfs_device_info {
326 struct btrfs_device *dev;
327 u64 dev_offset;
328 u64 max_avail;
329 u64 total_avail;
330};
331
332struct btrfs_raid_attr {
333 int sub_stripes; /* sub_stripes info for map */
334 int dev_stripes; /* stripes per dev */
335 int devs_max; /* max devs to use */
336 int devs_min; /* min devs needed */
337 int tolerated_failures; /* max tolerated fail devs */
338 int devs_increment; /* ndevs has to be a multiple of this */
339 int ncopies; /* how many copies to data has */
340};
341
342extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
343
344extern const u64 btrfs_raid_group[BTRFS_NR_RAID_TYPES];
345
346struct map_lookup {
347 u64 type;
348 int io_align;
349 int io_width;
350 int stripe_len;
351 int sector_size;
352 int num_stripes;
353 int sub_stripes;
354 struct btrfs_bio_stripe stripes[];
355};
356
357#define map_lookup_size(n) (sizeof(struct map_lookup) + \
358 (sizeof(struct btrfs_bio_stripe) * (n)))
359
360/*
361 * Restriper's general type filter
362 */
363#define BTRFS_BALANCE_DATA (1ULL << 0)
364#define BTRFS_BALANCE_SYSTEM (1ULL << 1)
365#define BTRFS_BALANCE_METADATA (1ULL << 2)
366
367#define BTRFS_BALANCE_TYPE_MASK (BTRFS_BALANCE_DATA | \
368 BTRFS_BALANCE_SYSTEM | \
369 BTRFS_BALANCE_METADATA)
370
371#define BTRFS_BALANCE_FORCE (1ULL << 3)
372#define BTRFS_BALANCE_RESUME (1ULL << 4)
373
374/*
375 * Balance filters
376 */
377#define BTRFS_BALANCE_ARGS_PROFILES (1ULL << 0)
378#define BTRFS_BALANCE_ARGS_USAGE (1ULL << 1)
379#define BTRFS_BALANCE_ARGS_DEVID (1ULL << 2)
380#define BTRFS_BALANCE_ARGS_DRANGE (1ULL << 3)
381#define BTRFS_BALANCE_ARGS_VRANGE (1ULL << 4)
382#define BTRFS_BALANCE_ARGS_LIMIT (1ULL << 5)
383#define BTRFS_BALANCE_ARGS_LIMIT_RANGE (1ULL << 6)
384#define BTRFS_BALANCE_ARGS_STRIPES_RANGE (1ULL << 7)
385#define BTRFS_BALANCE_ARGS_USAGE_RANGE (1ULL << 10)
386
387#define BTRFS_BALANCE_ARGS_MASK \
388 (BTRFS_BALANCE_ARGS_PROFILES | \
389 BTRFS_BALANCE_ARGS_USAGE | \
390 BTRFS_BALANCE_ARGS_DEVID | \
391 BTRFS_BALANCE_ARGS_DRANGE | \
392 BTRFS_BALANCE_ARGS_VRANGE | \
393 BTRFS_BALANCE_ARGS_LIMIT | \
394 BTRFS_BALANCE_ARGS_LIMIT_RANGE | \
395 BTRFS_BALANCE_ARGS_STRIPES_RANGE | \
396 BTRFS_BALANCE_ARGS_USAGE_RANGE)
397
398/*
399 * Profile changing flags. When SOFT is set we won't relocate chunk if
400 * it already has the target profile (even though it may be
401 * half-filled).
402 */
403#define BTRFS_BALANCE_ARGS_CONVERT (1ULL << 8)
404#define BTRFS_BALANCE_ARGS_SOFT (1ULL << 9)
405
406struct btrfs_balance_args;
407struct btrfs_balance_progress;
408struct btrfs_balance_control {
409 struct btrfs_fs_info *fs_info;
410
411 struct btrfs_balance_args data;
412 struct btrfs_balance_args meta;
413 struct btrfs_balance_args sys;
414
415 u64 flags;
416
417 struct btrfs_balance_progress stat;
418};
419
420int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
421 u64 end, u64 *length);
422void btrfs_get_bbio(struct btrfs_bio *bbio);
423void btrfs_put_bbio(struct btrfs_bio *bbio);
424int btrfs_map_block(struct btrfs_fs_info *fs_info, int rw,
425 u64 logical, u64 *length,
426 struct btrfs_bio **bbio_ret, int mirror_num);
427int btrfs_map_sblock(struct btrfs_fs_info *fs_info, int rw,
428 u64 logical, u64 *length,
429 struct btrfs_bio **bbio_ret, int mirror_num,
430 int need_raid_map);
431int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
432 u64 chunk_start, u64 physical, u64 devid,
433 u64 **logical, int *naddrs, int *stripe_len);
434int btrfs_read_sys_array(struct btrfs_root *root);
435int btrfs_read_chunk_tree(struct btrfs_root *root);
436int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
437 struct btrfs_root *extent_root, u64 type);
438void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
439void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
440int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
441 int mirror_num, int async_submit);
442int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
443 fmode_t flags, void *holder);
444int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
445 struct btrfs_fs_devices **fs_devices_ret);
446int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
447void btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices, int step);
448int btrfs_find_device_missing_or_by_path(struct btrfs_root *root,
449 char *device_path,
450 struct btrfs_device **device);
451struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
452 const u64 *devid,
453 const u8 *uuid);
454int btrfs_rm_device(struct btrfs_root *root, char *device_path);
455void btrfs_cleanup_fs_uuids(void);
456int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
457int btrfs_grow_device(struct btrfs_trans_handle *trans,
458 struct btrfs_device *device, u64 new_size);
459struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
460 u8 *uuid, u8 *fsid);
461int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
462int btrfs_init_new_device(struct btrfs_root *root, char *path);
463int btrfs_init_dev_replace_tgtdev(struct btrfs_root *root, char *device_path,
464 struct btrfs_device *srcdev,
465 struct btrfs_device **device_out);
466int btrfs_balance(struct btrfs_balance_control *bctl,
467 struct btrfs_ioctl_balance_args *bargs);
468int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
469int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
470int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
471int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
472int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
473int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
474int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset);
475int find_free_dev_extent_start(struct btrfs_transaction *transaction,
476 struct btrfs_device *device, u64 num_bytes,
477 u64 search_start, u64 *start, u64 *max_avail);
478int find_free_dev_extent(struct btrfs_trans_handle *trans,
479 struct btrfs_device *device, u64 num_bytes,
480 u64 *start, u64 *max_avail);
481void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
482int btrfs_get_dev_stats(struct btrfs_root *root,
483 struct btrfs_ioctl_get_dev_stats *stats);
484void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
485int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
486int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
487 struct btrfs_fs_info *fs_info);
488void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_fs_info *fs_info,
489 struct btrfs_device *srcdev);
490void btrfs_rm_dev_replace_free_srcdev(struct btrfs_fs_info *fs_info,
491 struct btrfs_device *srcdev);
492void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
493 struct btrfs_device *tgtdev);
494void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info,
495 struct btrfs_device *tgtdev);
496void btrfs_scratch_superblocks(struct block_device *bdev, char *device_path);
497int btrfs_is_parity_mirror(struct btrfs_mapping_tree *map_tree,
498 u64 logical, u64 len, int mirror_num);
499unsigned long btrfs_full_stripe_len(struct btrfs_root *root,
500 struct btrfs_mapping_tree *map_tree,
501 u64 logical);
502int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
503 struct btrfs_root *extent_root,
504 u64 chunk_offset, u64 chunk_size);
505int btrfs_remove_chunk(struct btrfs_trans_handle *trans,
506 struct btrfs_root *root, u64 chunk_offset);
507
508static inline int btrfs_dev_stats_dirty(struct btrfs_device *dev)
509{
510 return atomic_read(&dev->dev_stats_ccnt);
511}
512
513static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
514 int index)
515{
516 atomic_inc(dev->dev_stat_values + index);
517 smp_mb__before_atomic();
518 atomic_inc(&dev->dev_stats_ccnt);
519}
520
521static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
522 int index)
523{
524 return atomic_read(dev->dev_stat_values + index);
525}
526
527static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
528 int index)
529{
530 int ret;
531
532 ret = atomic_xchg(dev->dev_stat_values + index, 0);
533 smp_mb__before_atomic();
534 atomic_inc(&dev->dev_stats_ccnt);
535 return ret;
536}
537
538static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
539 int index, unsigned long val)
540{
541 atomic_set(dev->dev_stat_values + index, val);
542 smp_mb__before_atomic();
543 atomic_inc(&dev->dev_stats_ccnt);
544}
545
546static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
547 int index)
548{
549 btrfs_dev_stat_set(dev, index, 0);
550}
551
552void btrfs_update_commit_device_size(struct btrfs_fs_info *fs_info);
553void btrfs_update_commit_device_bytes_used(struct btrfs_root *root,
554 struct btrfs_transaction *transaction);
555
556static inline void lock_chunks(struct btrfs_root *root)
557{
558 mutex_lock(&root->fs_info->chunk_mutex);
559}
560
561static inline void unlock_chunks(struct btrfs_root *root)
562{
563 mutex_unlock(&root->fs_info->chunk_mutex);
564}
565
566struct list_head *btrfs_get_fs_uuids(void);
567void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info);
568void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info);
569
570#endif