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