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/bio.h>
10#include <linux/sort.h>
11#include <linux/btrfs.h>
12#include "async-thread.h"
13
14#define BTRFS_MAX_DATA_CHUNK_SIZE (10ULL * SZ_1G)
15
16extern struct mutex uuid_mutex;
17
18#define BTRFS_STRIPE_LEN SZ_64K
19
20struct buffer_head;
21struct btrfs_pending_bios {
22 struct bio *head;
23 struct bio *tail;
24};
25
26struct btrfs_io_geometry {
27 /* remaining bytes before crossing a stripe */
28 u64 len;
29 /* offset of logical address in chunk */
30 u64 offset;
31 /* length of single IO stripe */
32 u64 stripe_len;
33 /* number of stripe where address falls */
34 u64 stripe_nr;
35 /* offset of address in stripe */
36 u64 stripe_offset;
37 /* offset of raid56 stripe into the chunk */
38 u64 raid56_stripe_offset;
39};
40
41/*
42 * Use sequence counter to get consistent device stat data on
43 * 32-bit processors.
44 */
45#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
46#include <linux/seqlock.h>
47#define __BTRFS_NEED_DEVICE_DATA_ORDERED
48#define btrfs_device_data_ordered_init(device) \
49 seqcount_init(&device->data_seqcount)
50#else
51#define btrfs_device_data_ordered_init(device) do { } while (0)
52#endif
53
54#define BTRFS_DEV_STATE_WRITEABLE (0)
55#define BTRFS_DEV_STATE_IN_FS_METADATA (1)
56#define BTRFS_DEV_STATE_MISSING (2)
57#define BTRFS_DEV_STATE_REPLACE_TGT (3)
58#define BTRFS_DEV_STATE_FLUSH_SENT (4)
59
60struct btrfs_device {
61 struct list_head dev_list; /* device_list_mutex */
62 struct list_head dev_alloc_list; /* chunk mutex */
63 struct list_head post_commit_list; /* chunk mutex */
64 struct btrfs_fs_devices *fs_devices;
65 struct btrfs_fs_info *fs_info;
66
67 struct rcu_string *name;
68
69 u64 generation;
70
71 spinlock_t io_lock ____cacheline_aligned;
72 int running_pending;
73 /* regular prio bios */
74 struct btrfs_pending_bios pending_bios;
75 /* sync bios */
76 struct btrfs_pending_bios pending_sync_bios;
77
78 struct block_device *bdev;
79
80 /* the mode sent to blkdev_get */
81 fmode_t mode;
82
83 unsigned long dev_state;
84 blk_status_t last_flush_error;
85
86#ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
87 seqcount_t data_seqcount;
88#endif
89
90 /* the internal btrfs device id */
91 u64 devid;
92
93 /* size of the device in memory */
94 u64 total_bytes;
95
96 /* size of the device on disk */
97 u64 disk_total_bytes;
98
99 /* bytes used */
100 u64 bytes_used;
101
102 /* optimal io alignment for this device */
103 u32 io_align;
104
105 /* optimal io width for this device */
106 u32 io_width;
107 /* type and info about this device */
108 u64 type;
109
110 /* minimal io size for this device */
111 u32 sector_size;
112
113 /* physical drive uuid (or lvm uuid) */
114 u8 uuid[BTRFS_UUID_SIZE];
115
116 /*
117 * size of the device on the current transaction
118 *
119 * This variant is update when committing the transaction,
120 * and protected by chunk mutex
121 */
122 u64 commit_total_bytes;
123
124 /* bytes used on the current transaction */
125 u64 commit_bytes_used;
126
127 /* for sending down flush barriers */
128 struct bio *flush_bio;
129 struct completion flush_wait;
130
131 /* per-device scrub information */
132 struct scrub_ctx *scrub_ctx;
133
134 struct btrfs_work work;
135
136 /* readahead state */
137 atomic_t reada_in_flight;
138 u64 reada_next;
139 struct reada_zone *reada_curr_zone;
140 struct radix_tree_root reada_zones;
141 struct radix_tree_root reada_extents;
142
143 /* disk I/O failure stats. For detailed description refer to
144 * enum btrfs_dev_stat_values in ioctl.h */
145 int dev_stats_valid;
146
147 /* Counter to record the change of device stats */
148 atomic_t dev_stats_ccnt;
149 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
150
151 struct extent_io_tree alloc_state;
152};
153
154/*
155 * If we read those variants at the context of their own lock, we needn't
156 * use the following helpers, reading them directly is safe.
157 */
158#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
159#define BTRFS_DEVICE_GETSET_FUNCS(name) \
160static inline u64 \
161btrfs_device_get_##name(const struct btrfs_device *dev) \
162{ \
163 u64 size; \
164 unsigned int seq; \
165 \
166 do { \
167 seq = read_seqcount_begin(&dev->data_seqcount); \
168 size = dev->name; \
169 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \
170 return size; \
171} \
172 \
173static inline void \
174btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
175{ \
176 preempt_disable(); \
177 write_seqcount_begin(&dev->data_seqcount); \
178 dev->name = size; \
179 write_seqcount_end(&dev->data_seqcount); \
180 preempt_enable(); \
181}
182#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
183#define BTRFS_DEVICE_GETSET_FUNCS(name) \
184static inline u64 \
185btrfs_device_get_##name(const struct btrfs_device *dev) \
186{ \
187 u64 size; \
188 \
189 preempt_disable(); \
190 size = dev->name; \
191 preempt_enable(); \
192 return size; \
193} \
194 \
195static inline void \
196btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
197{ \
198 preempt_disable(); \
199 dev->name = size; \
200 preempt_enable(); \
201}
202#else
203#define BTRFS_DEVICE_GETSET_FUNCS(name) \
204static inline u64 \
205btrfs_device_get_##name(const struct btrfs_device *dev) \
206{ \
207 return dev->name; \
208} \
209 \
210static inline void \
211btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \
212{ \
213 dev->name = size; \
214}
215#endif
216
217BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
218BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
219BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
220
221struct btrfs_fs_devices {
222 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
223 u8 metadata_uuid[BTRFS_FSID_SIZE];
224 bool fsid_change;
225 struct list_head fs_list;
226
227 u64 num_devices;
228 u64 open_devices;
229 u64 rw_devices;
230 u64 missing_devices;
231 u64 total_rw_bytes;
232 u64 total_devices;
233
234 /* Highest generation number of seen devices */
235 u64 latest_generation;
236
237 struct block_device *latest_bdev;
238
239 /* all of the devices in the FS, protected by a mutex
240 * so we can safely walk it to write out the supers without
241 * worrying about add/remove by the multi-device code.
242 * Scrubbing super can kick off supers writing by holding
243 * this mutex lock.
244 */
245 struct mutex device_list_mutex;
246
247 /* List of all devices, protected by device_list_mutex */
248 struct list_head devices;
249
250 /*
251 * Devices which can satisfy space allocation. Protected by
252 * chunk_mutex
253 */
254 struct list_head alloc_list;
255
256 struct btrfs_fs_devices *seed;
257 int seeding;
258
259 int opened;
260
261 /* set when we find or add a device that doesn't have the
262 * nonrot flag set
263 */
264 int rotating;
265
266 struct btrfs_fs_info *fs_info;
267 /* sysfs kobjects */
268 struct kobject fsid_kobj;
269 struct kobject *device_dir_kobj;
270 struct completion kobj_unregister;
271};
272
273#define BTRFS_BIO_INLINE_CSUM_SIZE 64
274
275#define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \
276 - sizeof(struct btrfs_chunk)) \
277 / sizeof(struct btrfs_stripe) + 1)
278
279#define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \
280 - 2 * sizeof(struct btrfs_disk_key) \
281 - 2 * sizeof(struct btrfs_chunk)) \
282 / sizeof(struct btrfs_stripe) + 1)
283
284/*
285 * we need the mirror number and stripe index to be passed around
286 * the call chain while we are processing end_io (especially errors).
287 * Really, what we need is a btrfs_bio structure that has this info
288 * and is properly sized with its stripe array, but we're not there
289 * quite yet. We have our own btrfs bioset, and all of the bios
290 * we allocate are actually btrfs_io_bios. We'll cram as much of
291 * struct btrfs_bio as we can into this over time.
292 */
293struct btrfs_io_bio {
294 unsigned int mirror_num;
295 unsigned int stripe_index;
296 u64 logical;
297 u8 *csum;
298 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
299 struct bvec_iter iter;
300 /*
301 * This member must come last, bio_alloc_bioset will allocate enough
302 * bytes for entire btrfs_io_bio but relies on bio being last.
303 */
304 struct bio bio;
305};
306
307static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
308{
309 return container_of(bio, struct btrfs_io_bio, bio);
310}
311
312static inline void btrfs_io_bio_free_csum(struct btrfs_io_bio *io_bio)
313{
314 if (io_bio->csum != io_bio->csum_inline) {
315 kfree(io_bio->csum);
316 io_bio->csum = NULL;
317 }
318}
319
320struct btrfs_bio_stripe {
321 struct btrfs_device *dev;
322 u64 physical;
323 u64 length; /* only used for discard mappings */
324};
325
326struct btrfs_bio {
327 refcount_t refs;
328 atomic_t stripes_pending;
329 struct btrfs_fs_info *fs_info;
330 u64 map_type; /* get from map_lookup->type */
331 bio_end_io_t *end_io;
332 struct bio *orig_bio;
333 unsigned long flags;
334 void *private;
335 atomic_t error;
336 int max_errors;
337 int num_stripes;
338 int mirror_num;
339 int num_tgtdevs;
340 int *tgtdev_map;
341 /*
342 * logical block numbers for the start of each stripe
343 * The last one or two are p/q. These are sorted,
344 * so raid_map[0] is the start of our full stripe
345 */
346 u64 *raid_map;
347 struct btrfs_bio_stripe stripes[];
348};
349
350struct btrfs_device_info {
351 struct btrfs_device *dev;
352 u64 dev_offset;
353 u64 max_avail;
354 u64 total_avail;
355};
356
357struct btrfs_raid_attr {
358 u8 sub_stripes; /* sub_stripes info for map */
359 u8 dev_stripes; /* stripes per dev */
360 u8 devs_max; /* max devs to use */
361 u8 devs_min; /* min devs needed */
362 u8 tolerated_failures; /* max tolerated fail devs */
363 u8 devs_increment; /* ndevs has to be a multiple of this */
364 u8 ncopies; /* how many copies to data has */
365 u8 nparity; /* number of stripes worth of bytes to store
366 * parity information */
367 u8 mindev_error; /* error code if min devs requisite is unmet */
368 const char raid_name[8]; /* name of the raid */
369 u64 bg_flag; /* block group flag of the raid */
370};
371
372extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
373
374struct map_lookup {
375 u64 type;
376 int io_align;
377 int io_width;
378 u64 stripe_len;
379 int num_stripes;
380 int sub_stripes;
381 int verified_stripes; /* For mount time dev extent verification */
382 struct btrfs_bio_stripe stripes[];
383};
384
385#define map_lookup_size(n) (sizeof(struct map_lookup) + \
386 (sizeof(struct btrfs_bio_stripe) * (n)))
387
388struct btrfs_balance_args;
389struct btrfs_balance_progress;
390struct btrfs_balance_control {
391 struct btrfs_balance_args data;
392 struct btrfs_balance_args meta;
393 struct btrfs_balance_args sys;
394
395 u64 flags;
396
397 struct btrfs_balance_progress stat;
398};
399
400enum btrfs_map_op {
401 BTRFS_MAP_READ,
402 BTRFS_MAP_WRITE,
403 BTRFS_MAP_DISCARD,
404 BTRFS_MAP_GET_READ_MIRRORS,
405};
406
407static inline enum btrfs_map_op btrfs_op(struct bio *bio)
408{
409 switch (bio_op(bio)) {
410 case REQ_OP_DISCARD:
411 return BTRFS_MAP_DISCARD;
412 case REQ_OP_WRITE:
413 return BTRFS_MAP_WRITE;
414 default:
415 WARN_ON_ONCE(1);
416 /* fall through */
417 case REQ_OP_READ:
418 return BTRFS_MAP_READ;
419 }
420}
421
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, enum btrfs_map_op op,
425 u64 logical, u64 *length,
426 struct btrfs_bio **bbio_ret, int mirror_num);
427int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
428 u64 logical, u64 *length,
429 struct btrfs_bio **bbio_ret);
430int btrfs_get_io_geometry(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
431 u64 logical, u64 len, struct btrfs_io_geometry *io_geom);
432int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
433 u64 physical, u64 **logical, int *naddrs, int *stripe_len);
434int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
435int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
436int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 type);
437void btrfs_mapping_tree_free(struct extent_map_tree *tree);
438blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
439 int mirror_num, int async_submit);
440int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
441 fmode_t flags, void *holder);
442struct btrfs_device *btrfs_scan_one_device(const char *path,
443 fmode_t flags, void *holder);
444int btrfs_forget_devices(const char *path);
445int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
446void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices, int step);
447void btrfs_assign_next_active_device(struct btrfs_device *device,
448 struct btrfs_device *this_dev);
449struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
450 u64 devid,
451 const char *devpath);
452struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
453 const u64 *devid,
454 const u8 *uuid);
455void btrfs_free_device(struct btrfs_device *device);
456int btrfs_rm_device(struct btrfs_fs_info *fs_info,
457 const char *device_path, u64 devid);
458void __exit btrfs_cleanup_fs_uuids(void);
459int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
460int btrfs_grow_device(struct btrfs_trans_handle *trans,
461 struct btrfs_device *device, u64 new_size);
462struct btrfs_device *btrfs_find_device(struct btrfs_fs_devices *fs_devices,
463 u64 devid, u8 *uuid, u8 *fsid, bool seed);
464int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
465int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
466int btrfs_balance(struct btrfs_fs_info *fs_info,
467 struct btrfs_balance_control *bctl,
468 struct btrfs_ioctl_balance_args *bargs);
469void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
470int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
471int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
472int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
473int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
474int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
475int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
476int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset);
477int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes,
478 u64 *start, u64 *max_avail);
479void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
480int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
481 struct btrfs_ioctl_get_dev_stats *stats);
482void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
483int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
484int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
485void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
486void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
487void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
488void btrfs_scratch_superblocks(struct block_device *bdev, const char *device_path);
489int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
490 u64 logical, u64 len);
491unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
492 u64 logical);
493int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
494 u64 chunk_offset, u64 chunk_size);
495int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
496struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
497 u64 logical, u64 length);
498
499static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
500 int index)
501{
502 atomic_inc(dev->dev_stat_values + index);
503 /*
504 * This memory barrier orders stores updating statistics before stores
505 * updating dev_stats_ccnt.
506 *
507 * It pairs with smp_rmb() in btrfs_run_dev_stats().
508 */
509 smp_mb__before_atomic();
510 atomic_inc(&dev->dev_stats_ccnt);
511}
512
513static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
514 int index)
515{
516 return atomic_read(dev->dev_stat_values + index);
517}
518
519static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
520 int index)
521{
522 int ret;
523
524 ret = atomic_xchg(dev->dev_stat_values + index, 0);
525 /*
526 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
527 * - RMW operations that have a return value are fully ordered;
528 *
529 * This implicit memory barriers is paired with the smp_rmb in
530 * btrfs_run_dev_stats
531 */
532 atomic_inc(&dev->dev_stats_ccnt);
533 return ret;
534}
535
536static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
537 int index, unsigned long val)
538{
539 atomic_set(dev->dev_stat_values + index, val);
540 /*
541 * This memory barrier orders stores updating statistics before stores
542 * updating dev_stats_ccnt.
543 *
544 * It pairs with smp_rmb() in btrfs_run_dev_stats().
545 */
546 smp_mb__before_atomic();
547 atomic_inc(&dev->dev_stats_ccnt);
548}
549
550/*
551 * Convert block group flags (BTRFS_BLOCK_GROUP_*) to btrfs_raid_types, which
552 * can be used as index to access btrfs_raid_array[].
553 */
554static inline enum btrfs_raid_types btrfs_bg_flags_to_raid_index(u64 flags)
555{
556 if (flags & BTRFS_BLOCK_GROUP_RAID10)
557 return BTRFS_RAID_RAID10;
558 else if (flags & BTRFS_BLOCK_GROUP_RAID1)
559 return BTRFS_RAID_RAID1;
560 else if (flags & BTRFS_BLOCK_GROUP_DUP)
561 return BTRFS_RAID_DUP;
562 else if (flags & BTRFS_BLOCK_GROUP_RAID0)
563 return BTRFS_RAID_RAID0;
564 else if (flags & BTRFS_BLOCK_GROUP_RAID5)
565 return BTRFS_RAID_RAID5;
566 else if (flags & BTRFS_BLOCK_GROUP_RAID6)
567 return BTRFS_RAID_RAID6;
568
569 return BTRFS_RAID_SINGLE; /* BTRFS_BLOCK_GROUP_SINGLE */
570}
571
572void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
573
574struct list_head *btrfs_get_fs_uuids(void);
575void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info);
576void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info);
577bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
578 struct btrfs_device *failing_dev);
579
580int btrfs_bg_type_to_factor(u64 flags);
581const char *btrfs_bg_type_to_raid_name(u64 flags);
582int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
583
584#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