xref: /linux/fs/btrfs/volumes.h (revision 0eb4aaa230d725fa9b1cd758c0f17abca5597af6)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * Copyright (C) 2007 Oracle.  All rights reserved.
4  */
5 
6 #ifndef BTRFS_VOLUMES_H
7 #define BTRFS_VOLUMES_H
8 
9 #include <linux/blk_types.h>
10 #include <linux/sizes.h>
11 #include <linux/atomic.h>
12 #include <linux/sort.h>
13 #include <linux/list.h>
14 #include <linux/mutex.h>
15 #include <linux/log2.h>
16 #include <linux/kobject.h>
17 #include <linux/refcount.h>
18 #include <linux/completion.h>
19 #include <linux/rbtree.h>
20 #include <uapi/linux/btrfs.h>
21 #include "messages.h"
22 #include "rcu-string.h"
23 
24 struct block_device;
25 struct bdev_handle;
26 struct btrfs_fs_info;
27 struct btrfs_block_group;
28 struct btrfs_trans_handle;
29 struct btrfs_zoned_device_info;
30 
31 #define BTRFS_MAX_DATA_CHUNK_SIZE	(10ULL * SZ_1G)
32 
33 /*
34  * Arbitratry maximum size of one discard request to limit potentially long time
35  * spent in blkdev_issue_discard().
36  */
37 #define BTRFS_MAX_DISCARD_CHUNK_SIZE	(SZ_1G)
38 
39 extern struct mutex uuid_mutex;
40 
41 #define BTRFS_STRIPE_LEN		SZ_64K
42 #define BTRFS_STRIPE_LEN_SHIFT		(16)
43 #define BTRFS_STRIPE_LEN_MASK		(BTRFS_STRIPE_LEN - 1)
44 
45 static_assert(const_ilog2(BTRFS_STRIPE_LEN) == BTRFS_STRIPE_LEN_SHIFT);
46 
47 /* Used by sanity check for btrfs_raid_types. */
48 #define const_ffs(n) (__builtin_ctzll(n) + 1)
49 
50 /*
51  * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires
52  * RAID0 always to be the lowest profile bit.
53  * Although it's part of on-disk format and should never change, do extra
54  * compile-time sanity checks.
55  */
56 static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) <
57 	      const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0));
58 static_assert(const_ilog2(BTRFS_BLOCK_GROUP_RAID0) >
59 	      ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK));
60 
61 /* ilog2() can handle both constants and variables */
62 #define BTRFS_BG_FLAG_TO_INDEX(profile)					\
63 	ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1))
64 
65 enum btrfs_raid_types {
66 	/* SINGLE is the special one as it doesn't have on-disk bit. */
67 	BTRFS_RAID_SINGLE  = 0,
68 
69 	BTRFS_RAID_RAID0   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0),
70 	BTRFS_RAID_RAID1   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1),
71 	BTRFS_RAID_DUP	   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP),
72 	BTRFS_RAID_RAID10  = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10),
73 	BTRFS_RAID_RAID5   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5),
74 	BTRFS_RAID_RAID6   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6),
75 	BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3),
76 	BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4),
77 
78 	BTRFS_NR_RAID_TYPES
79 };
80 
81 /*
82  * Use sequence counter to get consistent device stat data on
83  * 32-bit processors.
84  */
85 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
86 #include <linux/seqlock.h>
87 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
88 #define btrfs_device_data_ordered_init(device)	\
89 	seqcount_init(&device->data_seqcount)
90 #else
91 #define btrfs_device_data_ordered_init(device) do { } while (0)
92 #endif
93 
94 #define BTRFS_DEV_STATE_WRITEABLE	(0)
95 #define BTRFS_DEV_STATE_IN_FS_METADATA	(1)
96 #define BTRFS_DEV_STATE_MISSING		(2)
97 #define BTRFS_DEV_STATE_REPLACE_TGT	(3)
98 #define BTRFS_DEV_STATE_FLUSH_SENT	(4)
99 #define BTRFS_DEV_STATE_NO_READA	(5)
100 
101 /* Special value encoding failure to write primary super block. */
102 #define BTRFS_SUPER_PRIMARY_WRITE_ERROR		(INT_MAX / 2)
103 
104 struct btrfs_fs_devices;
105 
106 struct btrfs_device {
107 	struct list_head dev_list; /* device_list_mutex */
108 	struct list_head dev_alloc_list; /* chunk mutex */
109 	struct list_head post_commit_list; /* chunk mutex */
110 	struct btrfs_fs_devices *fs_devices;
111 	struct btrfs_fs_info *fs_info;
112 
113 	struct rcu_string __rcu *name;
114 
115 	u64 generation;
116 
117 	struct file *bdev_file;
118 	struct block_device *bdev;
119 
120 	struct btrfs_zoned_device_info *zone_info;
121 
122 	/*
123 	 * Device's major-minor number. Must be set even if the device is not
124 	 * opened (bdev == NULL), unless the device is missing.
125 	 */
126 	dev_t devt;
127 	unsigned long dev_state;
128 	blk_status_t last_flush_error;
129 
130 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
131 	seqcount_t data_seqcount;
132 #endif
133 
134 	/* the internal btrfs device id */
135 	u64 devid;
136 
137 	/* size of the device in memory */
138 	u64 total_bytes;
139 
140 	/* size of the device on disk */
141 	u64 disk_total_bytes;
142 
143 	/* bytes used */
144 	u64 bytes_used;
145 
146 	/* optimal io alignment for this device */
147 	u32 io_align;
148 
149 	/* optimal io width for this device */
150 	u32 io_width;
151 	/* type and info about this device */
152 	u64 type;
153 
154 	/*
155 	 * Counter of super block write errors, values larger than
156 	 * BTRFS_SUPER_PRIMARY_WRITE_ERROR encode primary super block write failure.
157 	 */
158 	atomic_t sb_write_errors;
159 
160 	/* minimal io size for this device */
161 	u32 sector_size;
162 
163 	/* physical drive uuid (or lvm uuid) */
164 	u8 uuid[BTRFS_UUID_SIZE];
165 
166 	/*
167 	 * size of the device on the current transaction
168 	 *
169 	 * This variant is update when committing the transaction,
170 	 * and protected by chunk mutex
171 	 */
172 	u64 commit_total_bytes;
173 
174 	/* bytes used on the current transaction */
175 	u64 commit_bytes_used;
176 
177 	/* Bio used for flushing device barriers */
178 	struct bio flush_bio;
179 	struct completion flush_wait;
180 
181 	/* per-device scrub information */
182 	struct scrub_ctx *scrub_ctx;
183 
184 	/* disk I/O failure stats. For detailed description refer to
185 	 * enum btrfs_dev_stat_values in ioctl.h */
186 	int dev_stats_valid;
187 
188 	/* Counter to record the change of device stats */
189 	atomic_t dev_stats_ccnt;
190 	atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
191 
192 	struct extent_io_tree alloc_state;
193 
194 	struct completion kobj_unregister;
195 	/* For sysfs/FSID/devinfo/devid/ */
196 	struct kobject devid_kobj;
197 
198 	/* Bandwidth limit for scrub, in bytes */
199 	u64 scrub_speed_max;
200 };
201 
202 /*
203  * Block group or device which contains an active swapfile. Used for preventing
204  * unsafe operations while a swapfile is active.
205  *
206  * These are sorted on (ptr, inode) (note that a block group or device can
207  * contain more than one swapfile). We compare the pointer values because we
208  * don't actually care what the object is, we just need a quick check whether
209  * the object exists in the rbtree.
210  */
211 struct btrfs_swapfile_pin {
212 	struct rb_node node;
213 	void *ptr;
214 	struct inode *inode;
215 	/*
216 	 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr
217 	 * points to a struct btrfs_device.
218 	 */
219 	bool is_block_group;
220 	/*
221 	 * Only used when 'is_block_group' is true and it is the number of
222 	 * extents used by a swapfile for this block group ('ptr' field).
223 	 */
224 	int bg_extent_count;
225 };
226 
227 /*
228  * If we read those variants at the context of their own lock, we needn't
229  * use the following helpers, reading them directly is safe.
230  */
231 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
232 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
233 static inline u64							\
234 btrfs_device_get_##name(const struct btrfs_device *dev)			\
235 {									\
236 	u64 size;							\
237 	unsigned int seq;						\
238 									\
239 	do {								\
240 		seq = read_seqcount_begin(&dev->data_seqcount);		\
241 		size = dev->name;					\
242 	} while (read_seqcount_retry(&dev->data_seqcount, seq));	\
243 	return size;							\
244 }									\
245 									\
246 static inline void							\
247 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
248 {									\
249 	preempt_disable();						\
250 	write_seqcount_begin(&dev->data_seqcount);			\
251 	dev->name = size;						\
252 	write_seqcount_end(&dev->data_seqcount);			\
253 	preempt_enable();						\
254 }
255 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
256 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
257 static inline u64							\
258 btrfs_device_get_##name(const struct btrfs_device *dev)			\
259 {									\
260 	u64 size;							\
261 									\
262 	preempt_disable();						\
263 	size = dev->name;						\
264 	preempt_enable();						\
265 	return size;							\
266 }									\
267 									\
268 static inline void							\
269 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
270 {									\
271 	preempt_disable();						\
272 	dev->name = size;						\
273 	preempt_enable();						\
274 }
275 #else
276 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
277 static inline u64							\
278 btrfs_device_get_##name(const struct btrfs_device *dev)			\
279 {									\
280 	return dev->name;						\
281 }									\
282 									\
283 static inline void							\
284 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
285 {									\
286 	dev->name = size;						\
287 }
288 #endif
289 
290 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
291 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
292 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
293 
294 enum btrfs_chunk_allocation_policy {
295 	BTRFS_CHUNK_ALLOC_REGULAR,
296 	BTRFS_CHUNK_ALLOC_ZONED,
297 };
298 
299 #define BTRFS_DEFAULT_RR_MIN_CONTIG_READ	(SZ_256K)
300 /* Keep in sync with raid_attr table, current maximum is RAID1C4. */
301 #define BTRFS_RAID1_MAX_MIRRORS			(4)
302 /*
303  * Read policies for mirrored block group profiles, read picks the stripe based
304  * on these policies.
305  */
306 enum btrfs_read_policy {
307 	/* Use process PID to choose the stripe */
308 	BTRFS_READ_POLICY_PID,
309 #ifdef CONFIG_BTRFS_EXPERIMENTAL
310 	/* Balancing RAID1 reads across all striped devices (round-robin). */
311 	BTRFS_READ_POLICY_RR,
312 	/* Read from a specific device. */
313 	BTRFS_READ_POLICY_DEVID,
314 #endif
315 	BTRFS_NR_READ_POLICY,
316 };
317 
318 #ifdef CONFIG_BTRFS_EXPERIMENTAL
319 /*
320  * Checksum mode - offload it to workqueues or do it synchronously in
321  * btrfs_submit_chunk().
322  */
323 enum btrfs_offload_csum_mode {
324 	/*
325 	 * Choose offloading checksum or do it synchronously automatically.
326 	 * Do it synchronously if the checksum is fast, or offload to workqueues
327 	 * otherwise.
328 	 */
329 	BTRFS_OFFLOAD_CSUM_AUTO,
330 	/* Always offload checksum to workqueues. */
331 	BTRFS_OFFLOAD_CSUM_FORCE_ON,
332 	/* Never offload checksum to workqueues. */
333 	BTRFS_OFFLOAD_CSUM_FORCE_OFF,
334 };
335 #endif
336 
337 struct btrfs_fs_devices {
338 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
339 
340 	/*
341 	 * UUID written into the btree blocks:
342 	 *
343 	 * - If metadata_uuid != fsid then super block must have
344 	 *   BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag set.
345 	 *
346 	 * - Following shall be true at all times:
347 	 *   - metadata_uuid == btrfs_header::fsid
348 	 *   - metadata_uuid == btrfs_dev_item::fsid
349 	 *
350 	 * - Relations between fsid and metadata_uuid in sb and fs_devices:
351 	 *   - Normal:
352 	 *       fs_devices->fsid == fs_devices->metadata_uuid == sb->fsid
353 	 *       sb->metadata_uuid == 0
354 	 *
355 	 *   - When the BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag is set:
356 	 *       fs_devices->fsid == sb->fsid
357 	 *       fs_devices->metadata_uuid == sb->metadata_uuid
358 	 *
359 	 *   - When in-memory fs_devices->temp_fsid is true
360 	 *	 fs_devices->fsid = random
361 	 *	 fs_devices->metadata_uuid == sb->fsid
362 	 */
363 	u8 metadata_uuid[BTRFS_FSID_SIZE];
364 
365 	struct list_head fs_list;
366 
367 	/*
368 	 * Number of devices under this fsid including missing and
369 	 * replace-target device and excludes seed devices.
370 	 */
371 	u64 num_devices;
372 
373 	/*
374 	 * The number of devices that successfully opened, including
375 	 * replace-target, excludes seed devices.
376 	 */
377 	u64 open_devices;
378 
379 	/* The number of devices that are under the chunk allocation list. */
380 	u64 rw_devices;
381 
382 	/* Count of missing devices under this fsid excluding seed device. */
383 	u64 missing_devices;
384 	u64 total_rw_bytes;
385 
386 	/*
387 	 * Count of devices from btrfs_super_block::num_devices for this fsid,
388 	 * which includes the seed device, excludes the transient replace-target
389 	 * device.
390 	 */
391 	u64 total_devices;
392 
393 	/* Highest generation number of seen devices */
394 	u64 latest_generation;
395 
396 	/*
397 	 * The mount device or a device with highest generation after removal
398 	 * or replace.
399 	 */
400 	struct btrfs_device *latest_dev;
401 
402 	/*
403 	 * All of the devices in the filesystem, protected by a mutex so we can
404 	 * safely walk it to write out the super blocks without worrying about
405 	 * adding/removing by the multi-device code. Scrubbing super block can
406 	 * kick off supers writing by holding this mutex lock.
407 	 */
408 	struct mutex device_list_mutex;
409 
410 	/* List of all devices, protected by device_list_mutex */
411 	struct list_head devices;
412 
413 	/* Devices which can satisfy space allocation. Protected by * chunk_mutex. */
414 	struct list_head alloc_list;
415 
416 	struct list_head seed_list;
417 
418 	/* Count fs-devices opened. */
419 	int opened;
420 
421 	/* Set when we find or add a device that doesn't have the nonrot flag set. */
422 	bool rotating;
423 	/* Devices support TRIM/discard commands. */
424 	bool discardable;
425 	/* The filesystem is a seed filesystem. */
426 	bool seeding;
427 	/* The mount needs to use a randomly generated fsid. */
428 	bool temp_fsid;
429 	/* Enable/disable the filesystem stats tracking. */
430 	bool collect_fs_stats;
431 
432 	struct btrfs_fs_info *fs_info;
433 	/* sysfs kobjects */
434 	struct kobject fsid_kobj;
435 	struct kobject *devices_kobj;
436 	struct kobject *devinfo_kobj;
437 	struct completion kobj_unregister;
438 
439 	enum btrfs_chunk_allocation_policy chunk_alloc_policy;
440 
441 	/* Policy used to read the mirrored stripes. */
442 	enum btrfs_read_policy read_policy;
443 
444 #ifdef CONFIG_BTRFS_EXPERIMENTAL
445 	/*
446 	 * Minimum contiguous reads before switching to next device, the unit
447 	 * is one block/sectorsize.
448 	 */
449 	u32 rr_min_contig_read;
450 
451 	/* Device to be used for reading in case of RAID1. */
452 	u64 read_devid;
453 
454 	/* Checksum mode - offload it or do it synchronously. */
455 	enum btrfs_offload_csum_mode offload_csum_mode;
456 #endif
457 };
458 
459 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
460 			- sizeof(struct btrfs_chunk))		\
461 			/ sizeof(struct btrfs_stripe) + 1)
462 
463 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
464 				- 2 * sizeof(struct btrfs_disk_key)	\
465 				- 2 * sizeof(struct btrfs_chunk))	\
466 				/ sizeof(struct btrfs_stripe) + 1)
467 
468 struct btrfs_io_stripe {
469 	struct btrfs_device *dev;
470 	/* Block mapping. */
471 	u64 physical;
472 	u64 length;
473 	bool rst_search_commit_root;
474 	/* For the endio handler. */
475 	struct btrfs_io_context *bioc;
476 };
477 
478 struct btrfs_discard_stripe {
479 	struct btrfs_device *dev;
480 	u64 physical;
481 	u64 length;
482 };
483 
484 /*
485  * Context for IO subsmission for device stripe.
486  *
487  * - Track the unfinished mirrors for mirror based profiles
488  *   Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
489  *
490  * - Contain the logical -> physical mapping info
491  *   Used by submit_stripe_bio() for mapping logical bio
492  *   into physical device address.
493  *
494  * - Contain device replace info
495  *   Used by handle_ops_on_dev_replace() to copy logical bios
496  *   into the new device.
497  *
498  * - Contain RAID56 full stripe logical bytenrs
499  */
500 struct btrfs_io_context {
501 	refcount_t refs;
502 	struct btrfs_fs_info *fs_info;
503 	/* Taken from struct btrfs_chunk_map::type. */
504 	u64 map_type;
505 	struct bio *orig_bio;
506 	atomic_t error;
507 	u16 max_errors;
508 	bool use_rst;
509 
510 	u64 logical;
511 	u64 size;
512 	/* Raid stripe tree ordered entry. */
513 	struct list_head rst_ordered_entry;
514 
515 	/*
516 	 * The total number of stripes, including the extra duplicated
517 	 * stripe for replace.
518 	 */
519 	u16 num_stripes;
520 
521 	/*
522 	 * The mirror_num of this bioc.
523 	 *
524 	 * This is for reads which use 0 as mirror_num, thus we should return a
525 	 * valid mirror_num (>0) for the reader.
526 	 */
527 	u16 mirror_num;
528 
529 	/*
530 	 * The following two members are for dev-replace case only.
531 	 *
532 	 * @replace_nr_stripes:	Number of duplicated stripes which need to be
533 	 *			written to replace target.
534 	 *			Should be <= 2 (2 for DUP, otherwise <= 1).
535 	 * @replace_stripe_src:	The array indicates where the duplicated stripes
536 	 *			are from.
537 	 *
538 	 * The @replace_stripe_src[] array is mostly for RAID56 cases.
539 	 * As non-RAID56 stripes share the same contents of the mapped range,
540 	 * thus no need to bother where the duplicated ones are from.
541 	 *
542 	 * But for RAID56 case, all stripes contain different contents, thus
543 	 * we need a way to know the mapping.
544 	 *
545 	 * There is an example for the two members, using a RAID5 write:
546 	 *
547 	 *   num_stripes:	4 (3 + 1 duplicated write)
548 	 *   stripes[0]:	dev = devid 1, physical = X
549 	 *   stripes[1]:	dev = devid 2, physical = Y
550 	 *   stripes[2]:	dev = devid 3, physical = Z
551 	 *   stripes[3]:	dev = devid 0, physical = Y
552 	 *
553 	 * replace_nr_stripes = 1
554 	 * replace_stripe_src = 1	<- Means stripes[1] is involved in replace.
555 	 *				   The duplicated stripe index would be
556 	 *				   (@num_stripes - 1).
557 	 *
558 	 * Note, that we can still have cases replace_nr_stripes = 2 for DUP.
559 	 * In that case, all stripes share the same content, thus we don't
560 	 * need to bother @replace_stripe_src value at all.
561 	 */
562 	u16 replace_nr_stripes;
563 	s16 replace_stripe_src;
564 	/*
565 	 * Logical bytenr of the full stripe start, only for RAID56 cases.
566 	 *
567 	 * When this value is set to other than (u64)-1, the stripes[] should
568 	 * follow this pattern:
569 	 *
570 	 * (real_stripes = num_stripes - replace_nr_stripes)
571 	 * (data_stripes = (is_raid6) ? (real_stripes - 2) : (real_stripes - 1))
572 	 *
573 	 * stripes[0]:			The first data stripe
574 	 * stripes[1]:			The second data stripe
575 	 * ...
576 	 * stripes[data_stripes - 1]:	The last data stripe
577 	 * stripes[data_stripes]:	The P stripe
578 	 * stripes[data_stripes + 1]:	The Q stripe (only for RAID6).
579 	 */
580 	u64 full_stripe_logical;
581 	struct btrfs_io_stripe stripes[];
582 };
583 
584 struct btrfs_device_info {
585 	struct btrfs_device *dev;
586 	u64 dev_offset;
587 	u64 max_avail;
588 	u64 total_avail;
589 };
590 
591 struct btrfs_raid_attr {
592 	u8 sub_stripes;		/* sub_stripes info for map */
593 	u8 dev_stripes;		/* stripes per dev */
594 	u8 devs_max;		/* max devs to use */
595 	u8 devs_min;		/* min devs needed */
596 	u8 tolerated_failures;	/* max tolerated fail devs */
597 	u8 devs_increment;	/* ndevs has to be a multiple of this */
598 	u8 ncopies;		/* how many copies to data has */
599 	u8 nparity;		/* number of stripes worth of bytes to store
600 				 * parity information */
601 	u8 mindev_error;	/* error code if min devs requisite is unmet */
602 	const char raid_name[8]; /* name of the raid */
603 	u64 bg_flag;		/* block group flag of the raid */
604 };
605 
606 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
607 
608 struct btrfs_chunk_map {
609 	struct rb_node rb_node;
610 	/* For mount time dev extent verification. */
611 	int verified_stripes;
612 	refcount_t refs;
613 	u64 start;
614 	u64 chunk_len;
615 	u64 stripe_size;
616 	u64 type;
617 	int io_align;
618 	int io_width;
619 	int num_stripes;
620 	int sub_stripes;
621 	struct btrfs_io_stripe stripes[];
622 };
623 
624 #define btrfs_chunk_map_size(n) (sizeof(struct btrfs_chunk_map) + \
625 				 (sizeof(struct btrfs_io_stripe) * (n)))
626 
btrfs_free_chunk_map(struct btrfs_chunk_map * map)627 static inline void btrfs_free_chunk_map(struct btrfs_chunk_map *map)
628 {
629 	if (map && refcount_dec_and_test(&map->refs)) {
630 		ASSERT(RB_EMPTY_NODE(&map->rb_node));
631 		kfree(map);
632 	}
633 }
634 
635 struct btrfs_balance_control {
636 	struct btrfs_balance_args data;
637 	struct btrfs_balance_args meta;
638 	struct btrfs_balance_args sys;
639 
640 	u64 flags;
641 
642 	struct btrfs_balance_progress stat;
643 };
644 
645 /*
646  * Search for a given device by the set parameters
647  */
648 struct btrfs_dev_lookup_args {
649 	u64 devid;
650 	u8 *uuid;
651 	u8 *fsid;
652 	bool missing;
653 };
654 
655 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
656 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
657 
658 #define BTRFS_DEV_LOOKUP_ARGS(name) \
659 	struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
660 
661 enum btrfs_map_op {
662 	BTRFS_MAP_READ,
663 	BTRFS_MAP_WRITE,
664 	BTRFS_MAP_GET_READ_MIRRORS,
665 };
666 
btrfs_op(struct bio * bio)667 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
668 {
669 	switch (bio_op(bio)) {
670 	case REQ_OP_WRITE:
671 	case REQ_OP_ZONE_APPEND:
672 		return BTRFS_MAP_WRITE;
673 	default:
674 		WARN_ON_ONCE(1);
675 		fallthrough;
676 	case REQ_OP_READ:
677 		return BTRFS_MAP_READ;
678 	}
679 }
680 
btrfs_chunk_item_size(int num_stripes)681 static inline unsigned long btrfs_chunk_item_size(int num_stripes)
682 {
683 	ASSERT(num_stripes);
684 	return sizeof(struct btrfs_chunk) +
685 		sizeof(struct btrfs_stripe) * (num_stripes - 1);
686 }
687 
688 /*
689  * Do the type safe conversion from stripe_nr to offset inside the chunk.
690  *
691  * @stripe_nr is u32, with left shift it can overflow u32 for chunks larger
692  * than 4G.  This does the proper type cast to avoid overflow.
693  */
btrfs_stripe_nr_to_offset(u32 stripe_nr)694 static inline u64 btrfs_stripe_nr_to_offset(u32 stripe_nr)
695 {
696 	return (u64)stripe_nr << BTRFS_STRIPE_LEN_SHIFT;
697 }
698 
699 void btrfs_get_bioc(struct btrfs_io_context *bioc);
700 void btrfs_put_bioc(struct btrfs_io_context *bioc);
701 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
702 		    u64 logical, u64 *length,
703 		    struct btrfs_io_context **bioc_ret,
704 		    struct btrfs_io_stripe *smap, int *mirror_num_ret);
705 int btrfs_map_repair_block(struct btrfs_fs_info *fs_info,
706 			   struct btrfs_io_stripe *smap, u64 logical,
707 			   u32 length, int mirror_num);
708 struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,
709 					       u64 logical, u64 *length_ret,
710 					       u32 *num_stripes);
711 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
712 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
713 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
714 					    u64 type);
715 void btrfs_mapping_tree_free(struct btrfs_fs_info *fs_info);
716 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
717 		       blk_mode_t flags, void *holder);
718 struct btrfs_device *btrfs_scan_one_device(const char *path, blk_mode_t flags,
719 					   bool mount_arg_dev);
720 int btrfs_forget_devices(dev_t devt);
721 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
722 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
723 void btrfs_assign_next_active_device(struct btrfs_device *device,
724 				     struct btrfs_device *this_dev);
725 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
726 						  u64 devid,
727 						  const char *devpath);
728 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
729 				 struct btrfs_dev_lookup_args *args,
730 				 const char *path);
731 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
732 					const u64 *devid, const u8 *uuid,
733 					const char *path);
734 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
735 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
736 		    struct btrfs_dev_lookup_args *args,
737 		    struct file **bdev_file);
738 void __exit btrfs_cleanup_fs_uuids(void);
739 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
740 int btrfs_grow_device(struct btrfs_trans_handle *trans,
741 		      struct btrfs_device *device, u64 new_size);
742 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
743 				       const struct btrfs_dev_lookup_args *args);
744 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
745 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
746 int btrfs_balance(struct btrfs_fs_info *fs_info,
747 		  struct btrfs_balance_control *bctl,
748 		  struct btrfs_ioctl_balance_args *bargs);
749 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
750 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
751 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
752 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
753 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
754 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
755 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
756 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
757 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
758 			struct btrfs_ioctl_get_dev_stats *stats);
759 int btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
760 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
761 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
762 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
763 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
764 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
765 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
766 				    u64 logical);
767 u64 btrfs_calc_stripe_length(const struct btrfs_chunk_map *map);
768 int btrfs_nr_parity_stripes(u64 type);
769 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
770 				     struct btrfs_block_group *bg);
771 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
772 
773 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
774 struct btrfs_chunk_map *btrfs_alloc_chunk_map(int num_stripes, gfp_t gfp);
775 int btrfs_add_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
776 #endif
777 
778 struct btrfs_chunk_map *btrfs_find_chunk_map(struct btrfs_fs_info *fs_info,
779 					     u64 logical, u64 length);
780 struct btrfs_chunk_map *btrfs_find_chunk_map_nolock(struct btrfs_fs_info *fs_info,
781 						    u64 logical, u64 length);
782 struct btrfs_chunk_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
783 					    u64 logical, u64 length);
784 void btrfs_remove_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
785 void btrfs_release_disk_super(struct btrfs_super_block *super);
786 
btrfs_dev_stat_inc(struct btrfs_device * dev,int index)787 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
788 				      int index)
789 {
790 	atomic_inc(dev->dev_stat_values + index);
791 	/*
792 	 * This memory barrier orders stores updating statistics before stores
793 	 * updating dev_stats_ccnt.
794 	 *
795 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
796 	 */
797 	smp_mb__before_atomic();
798 	atomic_inc(&dev->dev_stats_ccnt);
799 }
800 
btrfs_dev_stat_read(struct btrfs_device * dev,int index)801 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
802 				      int index)
803 {
804 	return atomic_read(dev->dev_stat_values + index);
805 }
806 
btrfs_dev_stat_read_and_reset(struct btrfs_device * dev,int index)807 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
808 						int index)
809 {
810 	int ret;
811 
812 	ret = atomic_xchg(dev->dev_stat_values + index, 0);
813 	/*
814 	 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
815 	 * - RMW operations that have a return value are fully ordered;
816 	 *
817 	 * This implicit memory barriers is paired with the smp_rmb in
818 	 * btrfs_run_dev_stats
819 	 */
820 	atomic_inc(&dev->dev_stats_ccnt);
821 	return ret;
822 }
823 
btrfs_dev_stat_set(struct btrfs_device * dev,int index,unsigned long val)824 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
825 				      int index, unsigned long val)
826 {
827 	atomic_set(dev->dev_stat_values + index, val);
828 	/*
829 	 * This memory barrier orders stores updating statistics before stores
830 	 * updating dev_stats_ccnt.
831 	 *
832 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
833 	 */
834 	smp_mb__before_atomic();
835 	atomic_inc(&dev->dev_stats_ccnt);
836 }
837 
btrfs_dev_name(const struct btrfs_device * device)838 static inline const char *btrfs_dev_name(const struct btrfs_device *device)
839 {
840 	if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
841 		return "<missing disk>";
842 	else
843 		return rcu_str_deref(device->name);
844 }
845 
846 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
847 
848 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
849 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
850 					struct btrfs_device *failing_dev);
851 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, struct btrfs_device *device);
852 
853 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
854 int btrfs_bg_type_to_factor(u64 flags);
855 const char *btrfs_bg_type_to_raid_name(u64 flags);
856 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
857 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);
858 
859 bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr);
860 const u8 *btrfs_sb_fsid_ptr(const struct btrfs_super_block *sb);
861 
862 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
863 struct btrfs_io_context *alloc_btrfs_io_context(struct btrfs_fs_info *fs_info,
864 						u64 logical, u16 total_stripes);
865 #endif
866 
867 #endif
868