xref: /linux/fs/btrfs/volumes.h (revision e564cd2511750a634f916ae406d1f6ff84e53d0d)
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/blkdev.h>
11 #include <linux/sizes.h>
12 #include <linux/atomic.h>
13 #include <linux/sort.h>
14 #include <linux/list.h>
15 #include <linux/mutex.h>
16 #include <linux/log2.h>
17 #include <linux/kobject.h>
18 #include <linux/refcount.h>
19 #include <linux/completion.h>
20 #include <linux/rbtree.h>
21 #include <uapi/linux/btrfs.h>
22 #include <uapi/linux/btrfs_tree.h>
23 #include "messages.h"
24 #include "extent-io-tree.h"
25 
26 struct block_device;
27 struct bdev_handle;
28 struct btrfs_fs_info;
29 struct btrfs_block_group;
30 struct btrfs_trans_handle;
31 struct btrfs_transaction;
32 struct btrfs_zoned_device_info;
33 struct btrfs_space_info;
34 
35 #define BTRFS_MAX_DATA_CHUNK_SIZE	(10ULL * SZ_1G)
36 
37 /*
38  * Arbitrary maximum size of one discard request to limit potentially long time
39  * spent in blkdev_issue_discard().
40  */
41 #define BTRFS_MAX_DISCARD_CHUNK_SIZE	(SZ_1G)
42 
43 extern struct mutex uuid_mutex;
44 
45 #define BTRFS_STRIPE_LEN		SZ_64K
46 #define BTRFS_STRIPE_LEN_SHIFT		(16)
47 #define BTRFS_STRIPE_LEN_MASK		(BTRFS_STRIPE_LEN - 1)
48 
49 static_assert(ilog2(BTRFS_STRIPE_LEN) == BTRFS_STRIPE_LEN_SHIFT);
50 
51 /* Used by sanity check for btrfs_raid_types. */
52 #define const_ffs(n) (__builtin_ctzll(n) + 1)
53 
54 /*
55  * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires
56  * RAID0 always to be the lowest profile bit.
57  * Although it's part of on-disk format and should never change, do extra
58  * compile-time sanity checks.
59  */
60 static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) <
61 	      const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0));
62 
63 /* ilog2() can handle both constants and variables */
64 #define BTRFS_BG_FLAG_TO_INDEX(profile)					\
65 	ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1))
66 
67 enum btrfs_raid_types {
68 	/* SINGLE is the special one as it doesn't have on-disk bit. */
69 	BTRFS_RAID_SINGLE  = 0,
70 
71 	BTRFS_RAID_RAID0   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0),
72 	BTRFS_RAID_RAID1   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1),
73 	BTRFS_RAID_DUP	   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP),
74 	BTRFS_RAID_RAID10  = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10),
75 	BTRFS_RAID_RAID5   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5),
76 	BTRFS_RAID_RAID6   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6),
77 	BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3),
78 	BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4),
79 
80 	BTRFS_NR_RAID_TYPES
81 };
82 
83 static_assert(BTRFS_RAID_RAID0 == 1);
84 static_assert(BTRFS_RAID_RAID1 == 2);
85 static_assert(BTRFS_RAID_DUP == 3);
86 static_assert(BTRFS_RAID_RAID10 == 4);
87 static_assert(BTRFS_RAID_RAID5 == 5);
88 static_assert(BTRFS_RAID_RAID6 == 6);
89 static_assert(BTRFS_RAID_RAID1C3 == 7);
90 static_assert(BTRFS_RAID_RAID1C4 == 8);
91 
92 /*
93  * Use sequence counter to get consistent device stat data on
94  * 32-bit processors.
95  */
96 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
97 #include <linux/seqlock.h>
98 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
99 #define btrfs_device_data_ordered_init(device)	\
100 	seqcount_init(&device->data_seqcount)
101 #else
102 #define btrfs_device_data_ordered_init(device) do { } while (0)
103 #endif
104 
105 #define BTRFS_DEV_STATE_WRITEABLE	(0)
106 #define BTRFS_DEV_STATE_IN_FS_METADATA	(1)
107 #define BTRFS_DEV_STATE_MISSING		(2)
108 #define BTRFS_DEV_STATE_REPLACE_TGT	(3)
109 #define BTRFS_DEV_STATE_FLUSH_SENT	(4)
110 #define BTRFS_DEV_STATE_NO_READA	(5)
111 #define BTRFS_DEV_STATE_FLUSH_FAILED	(6)
112 
113 /* Set when the device item is found in chunk tree, used to catch unexpected registered device. */
114 #define BTRFS_DEV_STATE_ITEM_FOUND	(7)
115 
116 /* Special value encoding failure to write primary super block. */
117 #define BTRFS_SUPER_PRIMARY_WRITE_ERROR		(INT_MAX / 2)
118 
119 struct btrfs_fs_devices;
120 
121 struct btrfs_device {
122 	struct list_head dev_list; /* device_list_mutex */
123 	struct list_head dev_alloc_list; /* chunk mutex */
124 	struct list_head post_commit_list; /* chunk mutex */
125 	struct btrfs_fs_devices *fs_devices;
126 	struct btrfs_fs_info *fs_info;
127 
128 	/* Device path or NULL if missing. */
129 	const char __rcu *name;
130 
131 	u64 generation;
132 
133 	struct file *bdev_file;
134 	struct block_device *bdev;
135 
136 	struct btrfs_zoned_device_info *zone_info;
137 
138 	unsigned long dev_state;
139 
140 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
141 	seqcount_t data_seqcount;
142 #endif
143 
144 	/* the internal btrfs device id */
145 	u64 devid;
146 
147 	/* size of the device in memory */
148 	u64 total_bytes;
149 
150 	/* size of the device on disk */
151 	u64 disk_total_bytes;
152 
153 	/* bytes used */
154 	u64 bytes_used;
155 
156 	/* optimal io alignment for this device */
157 	u32 io_align;
158 
159 	/* optimal io width for this device */
160 	u32 io_width;
161 	/* type and info about this device */
162 	u64 type;
163 
164 	/*
165 	 * Counter of super block write errors, values larger than
166 	 * BTRFS_SUPER_PRIMARY_WRITE_ERROR encode primary super block write failure.
167 	 */
168 	atomic_t sb_write_errors;
169 
170 	/* minimal io size for this device */
171 	u32 sector_size;
172 
173 	/* physical drive uuid (or lvm uuid) */
174 	u8 uuid[BTRFS_UUID_SIZE];
175 
176 	/*
177 	 * size of the device on the current transaction
178 	 *
179 	 * This variant is update when committing the transaction,
180 	 * and protected by chunk mutex
181 	 */
182 	u64 commit_total_bytes;
183 
184 	/* bytes used on the current transaction */
185 	u64 commit_bytes_used;
186 
187 	/* Bio used for flushing device barriers */
188 	struct bio flush_bio;
189 	struct completion flush_wait;
190 
191 	/* per-device scrub information */
192 	struct scrub_ctx *scrub_ctx;
193 
194 	/* disk I/O failure stats. For detailed description refer to
195 	 * enum btrfs_dev_stat_values in ioctl.h */
196 	int dev_stats_valid;
197 
198 	/* Counter to record the change of device stats */
199 	atomic_t dev_stats_ccnt;
200 	atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
201 
202 	/*
203 	 * Device's major-minor number. Must be set even if the device is not
204 	 * opened (bdev == NULL), unless the device is missing.
205 	 */
206 	dev_t devt;
207 
208 	struct extent_io_tree alloc_state;
209 
210 	struct completion kobj_unregister;
211 	/* For sysfs/FSID/devinfo/devid/ */
212 	struct kobject devid_kobj;
213 
214 	/* Bandwidth limit for scrub, in bytes */
215 	u64 scrub_speed_max;
216 };
217 
218 /*
219  * Block group or device which contains an active swapfile. Used for preventing
220  * unsafe operations while a swapfile is active.
221  *
222  * These are sorted on (ptr, inode) (note that a block group or device can
223  * contain more than one swapfile). We compare the pointer values because we
224  * don't actually care what the object is, we just need a quick check whether
225  * the object exists in the rbtree.
226  */
227 struct btrfs_swapfile_pin {
228 	struct rb_node node;
229 	void *ptr;
230 	struct inode *inode;
231 	/*
232 	 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr
233 	 * points to a struct btrfs_device.
234 	 */
235 	bool is_block_group;
236 	/*
237 	 * Only used when 'is_block_group' is true and it is the number of
238 	 * extents used by a swapfile for this block group ('ptr' field).
239 	 */
240 	int bg_extent_count;
241 };
242 
243 /*
244  * If we read those variants at the context of their own lock, we needn't
245  * use the following helpers, reading them directly is safe.
246  */
247 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
248 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
249 static inline u64							\
250 btrfs_device_get_##name(const struct btrfs_device *dev)			\
251 {									\
252 	u64 size;							\
253 	unsigned int seq;						\
254 									\
255 	do {								\
256 		seq = read_seqcount_begin(&dev->data_seqcount);		\
257 		size = dev->name;					\
258 	} while (read_seqcount_retry(&dev->data_seqcount, seq));	\
259 	return size;							\
260 }									\
261 									\
262 static inline void							\
263 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
264 {									\
265 	preempt_disable();						\
266 	write_seqcount_begin(&dev->data_seqcount);			\
267 	dev->name = size;						\
268 	write_seqcount_end(&dev->data_seqcount);			\
269 	preempt_enable();						\
270 }
271 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
272 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
273 static inline u64							\
274 btrfs_device_get_##name(const struct btrfs_device *dev)			\
275 {									\
276 	u64 size;							\
277 									\
278 	preempt_disable();						\
279 	size = dev->name;						\
280 	preempt_enable();						\
281 	return size;							\
282 }									\
283 									\
284 static inline void							\
285 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
286 {									\
287 	preempt_disable();						\
288 	dev->name = size;						\
289 	preempt_enable();						\
290 }
291 #else
292 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
293 static inline u64							\
294 btrfs_device_get_##name(const struct btrfs_device *dev)			\
295 {									\
296 	return dev->name;						\
297 }									\
298 									\
299 static inline void							\
300 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
301 {									\
302 	dev->name = size;						\
303 }
304 #endif
305 
306 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
307 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
308 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
309 
310 enum btrfs_chunk_allocation_policy {
311 	BTRFS_CHUNK_ALLOC_REGULAR,
312 	BTRFS_CHUNK_ALLOC_ZONED,
313 };
314 
315 #define BTRFS_DEFAULT_RR_MIN_CONTIG_READ	(SZ_256K)
316 /* Keep in sync with raid_attr table, current maximum is RAID1C4. */
317 #define BTRFS_RAID1_MAX_MIRRORS			(4)
318 /*
319  * Read policies for mirrored block group profiles, read picks the stripe based
320  * on these policies.
321  */
322 enum btrfs_read_policy {
323 	/* Use process PID to choose the stripe */
324 	BTRFS_READ_POLICY_PID,
325 #ifdef CONFIG_BTRFS_EXPERIMENTAL
326 	/* Balancing RAID1 reads across all striped devices (round-robin). */
327 	BTRFS_READ_POLICY_RR,
328 	/* Read from a specific device. */
329 	BTRFS_READ_POLICY_DEVID,
330 #endif
331 	BTRFS_NR_READ_POLICY,
332 };
333 
334 struct btrfs_fs_devices {
335 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
336 
337 	/*
338 	 * UUID written into the btree blocks:
339 	 *
340 	 * - If metadata_uuid != fsid then super block must have
341 	 *   BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag set.
342 	 *
343 	 * - Following shall be true at all times:
344 	 *   - metadata_uuid == btrfs_header::fsid
345 	 *   - metadata_uuid == btrfs_dev_item::fsid
346 	 *
347 	 * - Relations between fsid and metadata_uuid in sb and fs_devices:
348 	 *   - Normal:
349 	 *       fs_devices->fsid == fs_devices->metadata_uuid == sb->fsid
350 	 *       sb->metadata_uuid == 0
351 	 *
352 	 *   - When the BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag is set:
353 	 *       fs_devices->fsid == sb->fsid
354 	 *       fs_devices->metadata_uuid == sb->metadata_uuid
355 	 *
356 	 *   - When in-memory fs_devices->temp_fsid is true
357 	 *	 fs_devices->fsid = random
358 	 *	 fs_devices->metadata_uuid == sb->fsid
359 	 */
360 	u8 metadata_uuid[BTRFS_FSID_SIZE];
361 
362 	struct list_head fs_list;
363 
364 	/*
365 	 * Number of devices under this fsid including missing and
366 	 * replace-target device and excludes seed devices.
367 	 */
368 	u64 num_devices;
369 
370 	/*
371 	 * The number of devices that successfully opened, including
372 	 * replace-target, excludes seed devices.
373 	 */
374 	u64 open_devices;
375 
376 	/* The number of devices that are under the chunk allocation list. */
377 	u64 rw_devices;
378 
379 	/* Count of missing devices under this fsid excluding seed device. */
380 	u64 missing_devices;
381 	u64 total_rw_bytes;
382 
383 	/*
384 	 * Count of devices from btrfs_super_block::num_devices for this fsid,
385 	 * which includes the seed device, excludes the transient replace-target
386 	 * device.
387 	 */
388 	u64 total_devices;
389 
390 	/* Highest generation number of seen devices */
391 	u64 latest_generation;
392 
393 	/*
394 	 * The mount device or a device with highest generation after removal
395 	 * or replace.
396 	 */
397 	struct btrfs_device *latest_dev;
398 
399 	/*
400 	 * All of the devices in the filesystem, protected by a mutex so we can
401 	 * safely walk it to write out the super blocks without worrying about
402 	 * adding/removing by the multi-device code. Scrubbing super block can
403 	 * kick off supers writing by holding this mutex lock.
404 	 */
405 	struct mutex device_list_mutex;
406 
407 	/* List of all devices, protected by device_list_mutex */
408 	struct list_head devices;
409 
410 	/* Devices which can satisfy space allocation. Protected by * chunk_mutex. */
411 	struct list_head alloc_list;
412 
413 	struct list_head seed_list;
414 
415 	/* Count fs-devices opened. */
416 	int opened;
417 
418 	/*
419 	 * Counter of the processes that are holding this fs_devices but not
420 	 * yet opened.
421 	 * This is for mounting handling, as we can only open the fs_devices
422 	 * after a super block is created.  But we cannot take uuid_mutex
423 	 * during sget_fc(), thus we have to hold the fs_devices (meaning it
424 	 * cannot be released) until a super block is returned.
425 	 */
426 	int holding;
427 
428 	/* Set when we find or add a device that doesn't have the nonrot flag set. */
429 	bool rotating;
430 	/* Devices support TRIM/discard commands. */
431 	bool discardable;
432 	/* The filesystem is a seed filesystem. */
433 	bool seeding;
434 	/* The mount needs to use a randomly generated fsid. */
435 	bool temp_fsid;
436 	/* Enable/disable the filesystem stats tracking. */
437 	bool collect_fs_stats;
438 
439 	struct btrfs_fs_info *fs_info;
440 	/* sysfs kobjects */
441 	struct kobject fsid_kobj;
442 	struct kobject *devices_kobj;
443 	struct kobject *devinfo_kobj;
444 	struct completion kobj_unregister;
445 
446 	enum btrfs_chunk_allocation_policy chunk_alloc_policy;
447 
448 	/* Policy used to read the mirrored stripes. */
449 	enum btrfs_read_policy read_policy;
450 
451 #ifdef CONFIG_BTRFS_EXPERIMENTAL
452 	/*
453 	 * Minimum contiguous reads before switching to next device, the unit
454 	 * is one block/sectorsize.
455 	 */
456 	u32 rr_min_contig_read;
457 
458 	/* Device to be used for reading in case of RAID1. */
459 	u64 read_devid;
460 #endif
461 };
462 
463 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
464 			- sizeof(struct btrfs_chunk))		\
465 			/ sizeof(struct btrfs_stripe) + 1)
466 
467 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
468 				- 2 * sizeof(struct btrfs_disk_key)	\
469 				- 2 * sizeof(struct btrfs_chunk))	\
470 				/ sizeof(struct btrfs_stripe) + 1)
471 
472 struct btrfs_io_stripe {
473 	struct btrfs_device *dev;
474 	/* Block mapping. */
475 	u64 physical;
476 	bool rst_search_commit_root;
477 	/* For the endio handler. */
478 	struct btrfs_io_context *bioc;
479 };
480 
481 struct btrfs_discard_stripe {
482 	struct btrfs_device *dev;
483 	u64 physical;
484 	u64 length;
485 };
486 
487 /*
488  * Context for IO submission for device stripe.
489  *
490  * - Track the unfinished mirrors for mirror based profiles
491  *   Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
492  *
493  * - Contain the logical -> physical mapping info
494  *   Used by submit_stripe_bio() for mapping logical bio
495  *   into physical device address.
496  *
497  * - Contain device replace info
498  *   Used by handle_ops_on_dev_replace() to copy logical bios
499  *   into the new device.
500  *
501  * - Contain RAID56 full stripe logical bytenrs
502  */
503 struct btrfs_io_context {
504 	refcount_t refs;
505 	struct btrfs_fs_info *fs_info;
506 	/* Taken from struct btrfs_chunk_map::type. */
507 	u64 map_type;
508 	struct bio *orig_bio;
509 	atomic_t error;
510 	u16 max_errors;
511 	bool use_rst;
512 
513 	u64 logical;
514 	u64 size;
515 	/* Raid stripe tree ordered entry. */
516 	struct list_head rst_ordered_entry;
517 
518 	/*
519 	 * The total number of stripes, including the extra duplicated
520 	 * stripe for replace.
521 	 */
522 	u16 num_stripes;
523 
524 	/*
525 	 * The mirror_num of this bioc.
526 	 *
527 	 * This is for reads which use 0 as mirror_num, thus we should return a
528 	 * valid mirror_num (>0) for the reader.
529 	 */
530 	u16 mirror_num;
531 
532 	/*
533 	 * The following two members are for dev-replace case only.
534 	 *
535 	 * @replace_nr_stripes:	Number of duplicated stripes which need to be
536 	 *			written to replace target.
537 	 *			Should be <= 2 (2 for DUP, otherwise <= 1).
538 	 * @replace_stripe_src:	The array indicates where the duplicated stripes
539 	 *			are from.
540 	 *
541 	 * The @replace_stripe_src[] array is mostly for RAID56 cases.
542 	 * As non-RAID56 stripes share the same contents of the mapped range,
543 	 * thus no need to bother where the duplicated ones are from.
544 	 *
545 	 * But for RAID56 case, all stripes contain different contents, thus
546 	 * we need a way to know the mapping.
547 	 *
548 	 * There is an example for the two members, using a RAID5 write:
549 	 *
550 	 *   num_stripes:	4 (3 + 1 duplicated write)
551 	 *   stripes[0]:	dev = devid 1, physical = X
552 	 *   stripes[1]:	dev = devid 2, physical = Y
553 	 *   stripes[2]:	dev = devid 3, physical = Z
554 	 *   stripes[3]:	dev = devid 0, physical = Y
555 	 *
556 	 * replace_nr_stripes = 1
557 	 * replace_stripe_src = 1	<- Means stripes[1] is involved in replace.
558 	 *				   The duplicated stripe index would be
559 	 *				   (@num_stripes - 1).
560 	 *
561 	 * Note, that we can still have cases replace_nr_stripes = 2 for DUP.
562 	 * In that case, all stripes share the same content, thus we don't
563 	 * need to bother @replace_stripe_src value at all.
564 	 */
565 	u16 replace_nr_stripes;
566 	s16 replace_stripe_src;
567 	/*
568 	 * Logical bytenr of the full stripe start, only for RAID56 cases.
569 	 *
570 	 * When this value is set to other than (u64)-1, the stripes[] should
571 	 * follow this pattern:
572 	 *
573 	 * (real_stripes = num_stripes - replace_nr_stripes)
574 	 * (data_stripes = (is_raid6) ? (real_stripes - 2) : (real_stripes - 1))
575 	 *
576 	 * stripes[0]:			The first data stripe
577 	 * stripes[1]:			The second data stripe
578 	 * ...
579 	 * stripes[data_stripes - 1]:	The last data stripe
580 	 * stripes[data_stripes]:	The P stripe
581 	 * stripes[data_stripes + 1]:	The Q stripe (only for RAID6).
582 	 */
583 	u64 full_stripe_logical;
584 	struct btrfs_io_stripe stripes[];
585 };
586 
587 struct btrfs_device_info {
588 	struct btrfs_device *dev;
589 	u64 dev_offset;
590 	u64 max_avail;
591 	u64 total_avail;
592 };
593 
594 struct btrfs_raid_attr {
595 	u8 sub_stripes;		/* sub_stripes info for map */
596 	u8 dev_stripes;		/* stripes per dev */
597 	u8 devs_max;		/* max devs to use */
598 	u8 devs_min;		/* min devs needed */
599 	u8 tolerated_failures;	/* max tolerated fail devs */
600 	u8 devs_increment;	/* ndevs has to be a multiple of this */
601 	u8 ncopies;		/* how many copies to data has */
602 	u8 nparity;		/* number of stripes worth of bytes to store
603 				 * parity information */
604 	u8 mindev_error;	/* error code if min devs requisite is unmet */
605 	const char raid_name[8]; /* name of the raid */
606 	u64 bg_flag;		/* block group flag of the raid */
607 };
608 
609 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
610 
611 struct btrfs_chunk_map {
612 	struct rb_node rb_node;
613 	/* For mount time dev extent verification. */
614 	int verified_stripes;
615 	refcount_t refs;
616 	u64 start;
617 	u64 chunk_len;
618 	u64 stripe_size;
619 	u64 type;
620 	int io_align;
621 	int io_width;
622 	int num_stripes;
623 	int sub_stripes;
624 	struct btrfs_io_stripe stripes[];
625 };
626 
627 #define btrfs_chunk_map_size(n) (sizeof(struct btrfs_chunk_map) + \
628 				 (sizeof(struct btrfs_io_stripe) * (n)))
629 
630 static inline void btrfs_free_chunk_map(struct btrfs_chunk_map *map)
631 {
632 	if (map && refcount_dec_and_test(&map->refs)) {
633 		ASSERT(RB_EMPTY_NODE(&map->rb_node));
634 		kfree(map);
635 	}
636 }
637 DEFINE_FREE(btrfs_free_chunk_map, struct btrfs_chunk_map *, btrfs_free_chunk_map(_T))
638 
639 struct btrfs_balance_control {
640 	struct btrfs_balance_args data;
641 	struct btrfs_balance_args meta;
642 	struct btrfs_balance_args sys;
643 
644 	u64 flags;
645 
646 	struct btrfs_balance_progress stat;
647 };
648 
649 /*
650  * Search for a given device by the set parameters
651  */
652 struct btrfs_dev_lookup_args {
653 	u64 devid;
654 	u8 *uuid;
655 	u8 *fsid;
656 	/*
657 	 * If devt is specified, all other members will be ignored as it is
658 	 * enough to uniquely locate a device.
659 	 */
660 	dev_t devt;
661 	bool missing;
662 };
663 
664 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
665 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
666 
667 #define BTRFS_DEV_LOOKUP_ARGS(name) \
668 	struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
669 
670 enum btrfs_map_op {
671 	BTRFS_MAP_READ,
672 	BTRFS_MAP_WRITE,
673 	BTRFS_MAP_GET_READ_MIRRORS,
674 };
675 
676 static inline enum btrfs_map_op btrfs_op(const struct bio *bio)
677 {
678 	switch (bio_op(bio)) {
679 	case REQ_OP_WRITE:
680 	case REQ_OP_ZONE_APPEND:
681 		return BTRFS_MAP_WRITE;
682 	default:
683 		WARN_ON_ONCE(1);
684 		fallthrough;
685 	case REQ_OP_READ:
686 		return BTRFS_MAP_READ;
687 	}
688 }
689 
690 static inline unsigned long btrfs_chunk_item_size(int num_stripes)
691 {
692 	ASSERT(num_stripes);
693 	return sizeof(struct btrfs_chunk) +
694 		sizeof(struct btrfs_stripe) * (num_stripes - 1);
695 }
696 
697 /*
698  * Do the type safe conversion from stripe_nr to offset inside the chunk.
699  *
700  * @stripe_nr is u32, with left shift it can overflow u32 for chunks larger
701  * than 4G.  This does the proper type cast to avoid overflow.
702  */
703 static inline u64 btrfs_stripe_nr_to_offset(u32 stripe_nr)
704 {
705 	return (u64)stripe_nr << BTRFS_STRIPE_LEN_SHIFT;
706 }
707 
708 void btrfs_get_bioc(struct btrfs_io_context *bioc);
709 void btrfs_put_bioc(struct btrfs_io_context *bioc);
710 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
711 		    u64 logical, u64 *length,
712 		    struct btrfs_io_context **bioc_ret,
713 		    struct btrfs_io_stripe *smap, int *mirror_num_ret);
714 int btrfs_map_repair_block(struct btrfs_fs_info *fs_info,
715 			   struct btrfs_io_stripe *smap, u64 logical,
716 			   u32 length, int mirror_num);
717 struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,
718 					       u64 logical, u64 *length_ret,
719 					       u32 *num_stripes, bool do_remap);
720 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
721 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
722 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
723 					     struct btrfs_space_info *space_info,
724 					     u64 type);
725 void btrfs_mapping_tree_free(struct btrfs_fs_info *fs_info);
726 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
727 		       blk_mode_t flags, void *holder);
728 struct btrfs_device *btrfs_scan_one_device(const char *path, bool mount_arg_dev);
729 int btrfs_forget_devices(dev_t devt);
730 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
731 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
732 void btrfs_assign_next_active_device(struct btrfs_device *device,
733 				     struct btrfs_device *this_dev);
734 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
735 						  u64 devid,
736 						  const char *devpath);
737 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
738 				 struct btrfs_dev_lookup_args *args,
739 				 const char *path);
740 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
741 					const u64 *devid, const u8 *uuid,
742 					const char *path);
743 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
744 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
745 		    struct btrfs_dev_lookup_args *args,
746 		    struct file **bdev_file);
747 void __exit btrfs_cleanup_fs_uuids(void);
748 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
749 int btrfs_grow_device(struct btrfs_trans_handle *trans,
750 		      struct btrfs_device *device, u64 new_size);
751 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
752 				       const struct btrfs_dev_lookup_args *args);
753 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
754 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
755 int btrfs_balance(struct btrfs_fs_info *fs_info,
756 		  struct btrfs_balance_control *bctl,
757 		  struct btrfs_ioctl_balance_args *bargs);
758 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
759 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
760 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
761 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
762 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset,
763 			 bool verbose);
764 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
765 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
766 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
767 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
768 			struct btrfs_ioctl_get_dev_stats *stats);
769 int btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
770 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
771 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
772 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
773 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
774 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
775 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
776 				    u64 logical);
777 u64 btrfs_calc_stripe_length(const struct btrfs_chunk_map *map);
778 int btrfs_nr_parity_stripes(u64 type);
779 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
780 				     struct btrfs_block_group *bg);
781 int btrfs_remove_dev_extents(struct btrfs_trans_handle *trans, struct btrfs_chunk_map *map);
782 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
783 
784 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
785 struct btrfs_chunk_map *btrfs_alloc_chunk_map(int num_stripes, gfp_t gfp);
786 int btrfs_add_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
787 #endif
788 
789 struct btrfs_chunk_map *btrfs_find_chunk_map(struct btrfs_fs_info *fs_info,
790 					     u64 logical, u64 length);
791 struct btrfs_chunk_map *btrfs_find_chunk_map_nolock(struct btrfs_fs_info *fs_info,
792 						    u64 logical, u64 length);
793 struct btrfs_chunk_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
794 					    u64 logical, u64 length);
795 void btrfs_remove_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
796 struct btrfs_super_block *btrfs_read_disk_super(struct block_device *bdev,
797 						int copy_num, bool drop_cache);
798 void btrfs_release_disk_super(struct btrfs_super_block *super);
799 
800 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
801 				      int index)
802 {
803 	atomic_inc(dev->dev_stat_values + index);
804 	/*
805 	 * This memory barrier orders stores updating statistics before stores
806 	 * updating dev_stats_ccnt.
807 	 *
808 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
809 	 */
810 	smp_mb__before_atomic();
811 	atomic_inc(&dev->dev_stats_ccnt);
812 }
813 
814 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
815 				      int index)
816 {
817 	return atomic_read(dev->dev_stat_values + index);
818 }
819 
820 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
821 						int index)
822 {
823 	int ret;
824 
825 	ret = atomic_xchg(dev->dev_stat_values + index, 0);
826 	/*
827 	 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
828 	 * - RMW operations that have a return value are fully ordered;
829 	 *
830 	 * This implicit memory barriers is paired with the smp_rmb in
831 	 * btrfs_run_dev_stats
832 	 */
833 	atomic_inc(&dev->dev_stats_ccnt);
834 	return ret;
835 }
836 
837 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
838 				      int index, unsigned long val)
839 {
840 	atomic_set(dev->dev_stat_values + index, val);
841 	/*
842 	 * This memory barrier orders stores updating statistics before stores
843 	 * updating dev_stats_ccnt.
844 	 *
845 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
846 	 */
847 	smp_mb__before_atomic();
848 	atomic_inc(&dev->dev_stats_ccnt);
849 }
850 
851 static inline const char *btrfs_dev_name(const struct btrfs_device *device)
852 {
853 	if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
854 		return "<missing disk>";
855 	else
856 		return rcu_dereference(device->name);
857 }
858 
859 static inline void btrfs_warn_unknown_chunk_allocation(enum btrfs_chunk_allocation_policy pol)
860 {
861 	WARN_ONCE(1, "unknown allocation policy %d, fallback to regular", pol);
862 }
863 
864 static inline void btrfs_fs_devices_inc_holding(struct btrfs_fs_devices *fs_devices)
865 {
866 	lockdep_assert_held(&uuid_mutex);
867 	ASSERT(fs_devices->holding >= 0);
868 	fs_devices->holding++;
869 }
870 
871 static inline void btrfs_fs_devices_dec_holding(struct btrfs_fs_devices *fs_devices)
872 {
873 	lockdep_assert_held(&uuid_mutex);
874 	ASSERT(fs_devices->holding > 0);
875 	fs_devices->holding--;
876 }
877 
878 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
879 
880 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
881 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
882 					struct btrfs_device *failing_dev);
883 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, struct btrfs_device *device);
884 
885 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
886 int btrfs_bg_type_to_factor(u64 flags);
887 const char *btrfs_bg_type_to_raid_name(u64 flags);
888 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
889 bool btrfs_verify_dev_items(const struct btrfs_fs_info *fs_info);
890 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);
891 
892 bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr);
893 const u8 *btrfs_sb_fsid_ptr(const struct btrfs_super_block *sb);
894 int btrfs_update_device(struct btrfs_trans_handle *trans, struct btrfs_device *device);
895 void btrfs_chunk_map_device_clear_bits(struct btrfs_chunk_map *map, unsigned int bits);
896 
897 bool btrfs_first_pending_extent(struct btrfs_device *device, u64 start, u64 len,
898 				u64 *pending_start, u64 *pending_end);
899 bool btrfs_find_hole_in_pending_extents(struct btrfs_device *device,
900 					u64 *start, u64 *len, u64 min_hole_size);
901 
902 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
903 struct btrfs_io_context *alloc_btrfs_io_context(struct btrfs_fs_info *fs_info,
904 						u64 logical, u16 total_stripes);
905 #endif
906 
907 #endif
908