xref: /linux/fs/btrfs/volumes.h (revision 2697b79a469b68e3ad3640f55284359c1396278d)
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 extern struct mutex uuid_mutex;
34 
35 #define BTRFS_STRIPE_LEN		SZ_64K
36 #define BTRFS_STRIPE_LEN_SHIFT		(16)
37 #define BTRFS_STRIPE_LEN_MASK		(BTRFS_STRIPE_LEN - 1)
38 
39 static_assert(const_ilog2(BTRFS_STRIPE_LEN) == BTRFS_STRIPE_LEN_SHIFT);
40 
41 /* Used by sanity check for btrfs_raid_types. */
42 #define const_ffs(n) (__builtin_ctzll(n) + 1)
43 
44 /*
45  * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires
46  * RAID0 always to be the lowest profile bit.
47  * Although it's part of on-disk format and should never change, do extra
48  * compile-time sanity checks.
49  */
50 static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) <
51 	      const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0));
52 static_assert(const_ilog2(BTRFS_BLOCK_GROUP_RAID0) >
53 	      ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK));
54 
55 /* ilog2() can handle both constants and variables */
56 #define BTRFS_BG_FLAG_TO_INDEX(profile)					\
57 	ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1))
58 
59 enum btrfs_raid_types {
60 	/* SINGLE is the special one as it doesn't have on-disk bit. */
61 	BTRFS_RAID_SINGLE  = 0,
62 
63 	BTRFS_RAID_RAID0   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0),
64 	BTRFS_RAID_RAID1   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1),
65 	BTRFS_RAID_DUP	   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP),
66 	BTRFS_RAID_RAID10  = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10),
67 	BTRFS_RAID_RAID5   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5),
68 	BTRFS_RAID_RAID6   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6),
69 	BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3),
70 	BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4),
71 
72 	BTRFS_NR_RAID_TYPES
73 };
74 
75 /*
76  * Use sequence counter to get consistent device stat data on
77  * 32-bit processors.
78  */
79 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
80 #include <linux/seqlock.h>
81 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
82 #define btrfs_device_data_ordered_init(device)	\
83 	seqcount_init(&device->data_seqcount)
84 #else
85 #define btrfs_device_data_ordered_init(device) do { } while (0)
86 #endif
87 
88 #define BTRFS_DEV_STATE_WRITEABLE	(0)
89 #define BTRFS_DEV_STATE_IN_FS_METADATA	(1)
90 #define BTRFS_DEV_STATE_MISSING		(2)
91 #define BTRFS_DEV_STATE_REPLACE_TGT	(3)
92 #define BTRFS_DEV_STATE_FLUSH_SENT	(4)
93 #define BTRFS_DEV_STATE_NO_READA	(5)
94 
95 /* Special value encoding failure to write primary super block. */
96 #define BTRFS_SUPER_PRIMARY_WRITE_ERROR		(INT_MAX / 2)
97 
98 struct btrfs_fs_devices;
99 
100 struct btrfs_device {
101 	struct list_head dev_list; /* device_list_mutex */
102 	struct list_head dev_alloc_list; /* chunk mutex */
103 	struct list_head post_commit_list; /* chunk mutex */
104 	struct btrfs_fs_devices *fs_devices;
105 	struct btrfs_fs_info *fs_info;
106 
107 	struct rcu_string __rcu *name;
108 
109 	u64 generation;
110 
111 	struct file *bdev_file;
112 	struct block_device *bdev;
113 
114 	struct btrfs_zoned_device_info *zone_info;
115 
116 	/*
117 	 * Device's major-minor number. Must be set even if the device is not
118 	 * opened (bdev == NULL), unless the device is missing.
119 	 */
120 	dev_t devt;
121 	unsigned long dev_state;
122 	blk_status_t last_flush_error;
123 
124 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
125 	seqcount_t data_seqcount;
126 #endif
127 
128 	/* the internal btrfs device id */
129 	u64 devid;
130 
131 	/* size of the device in memory */
132 	u64 total_bytes;
133 
134 	/* size of the device on disk */
135 	u64 disk_total_bytes;
136 
137 	/* bytes used */
138 	u64 bytes_used;
139 
140 	/* optimal io alignment for this device */
141 	u32 io_align;
142 
143 	/* optimal io width for this device */
144 	u32 io_width;
145 	/* type and info about this device */
146 	u64 type;
147 
148 	/*
149 	 * Counter of super block write errors, values larger than
150 	 * BTRFS_SUPER_PRIMARY_WRITE_ERROR encode primary super block write failure.
151 	 */
152 	atomic_t sb_write_errors;
153 
154 	/* minimal io size for this device */
155 	u32 sector_size;
156 
157 	/* physical drive uuid (or lvm uuid) */
158 	u8 uuid[BTRFS_UUID_SIZE];
159 
160 	/*
161 	 * size of the device on the current transaction
162 	 *
163 	 * This variant is update when committing the transaction,
164 	 * and protected by chunk mutex
165 	 */
166 	u64 commit_total_bytes;
167 
168 	/* bytes used on the current transaction */
169 	u64 commit_bytes_used;
170 
171 	/* Bio used for flushing device barriers */
172 	struct bio flush_bio;
173 	struct completion flush_wait;
174 
175 	/* per-device scrub information */
176 	struct scrub_ctx *scrub_ctx;
177 
178 	/* disk I/O failure stats. For detailed description refer to
179 	 * enum btrfs_dev_stat_values in ioctl.h */
180 	int dev_stats_valid;
181 
182 	/* Counter to record the change of device stats */
183 	atomic_t dev_stats_ccnt;
184 	atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
185 
186 	struct extent_io_tree alloc_state;
187 
188 	struct completion kobj_unregister;
189 	/* For sysfs/FSID/devinfo/devid/ */
190 	struct kobject devid_kobj;
191 
192 	/* Bandwidth limit for scrub, in bytes */
193 	u64 scrub_speed_max;
194 };
195 
196 /*
197  * Block group or device which contains an active swapfile. Used for preventing
198  * unsafe operations while a swapfile is active.
199  *
200  * These are sorted on (ptr, inode) (note that a block group or device can
201  * contain more than one swapfile). We compare the pointer values because we
202  * don't actually care what the object is, we just need a quick check whether
203  * the object exists in the rbtree.
204  */
205 struct btrfs_swapfile_pin {
206 	struct rb_node node;
207 	void *ptr;
208 	struct inode *inode;
209 	/*
210 	 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr
211 	 * points to a struct btrfs_device.
212 	 */
213 	bool is_block_group;
214 	/*
215 	 * Only used when 'is_block_group' is true and it is the number of
216 	 * extents used by a swapfile for this block group ('ptr' field).
217 	 */
218 	int bg_extent_count;
219 };
220 
221 /*
222  * If we read those variants at the context of their own lock, we needn't
223  * use the following helpers, reading them directly is safe.
224  */
225 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
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 	unsigned int seq;						\
232 									\
233 	do {								\
234 		seq = read_seqcount_begin(&dev->data_seqcount);		\
235 		size = dev->name;					\
236 	} while (read_seqcount_retry(&dev->data_seqcount, seq));	\
237 	return size;							\
238 }									\
239 									\
240 static inline void							\
241 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
242 {									\
243 	preempt_disable();						\
244 	write_seqcount_begin(&dev->data_seqcount);			\
245 	dev->name = size;						\
246 	write_seqcount_end(&dev->data_seqcount);			\
247 	preempt_enable();						\
248 }
249 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
250 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
251 static inline u64							\
252 btrfs_device_get_##name(const struct btrfs_device *dev)			\
253 {									\
254 	u64 size;							\
255 									\
256 	preempt_disable();						\
257 	size = dev->name;						\
258 	preempt_enable();						\
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 	dev->name = size;						\
267 	preempt_enable();						\
268 }
269 #else
270 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
271 static inline u64							\
272 btrfs_device_get_##name(const struct btrfs_device *dev)			\
273 {									\
274 	return dev->name;						\
275 }									\
276 									\
277 static inline void							\
278 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
279 {									\
280 	dev->name = size;						\
281 }
282 #endif
283 
284 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
285 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
286 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
287 
288 enum btrfs_chunk_allocation_policy {
289 	BTRFS_CHUNK_ALLOC_REGULAR,
290 	BTRFS_CHUNK_ALLOC_ZONED,
291 };
292 
293 /*
294  * Read policies for mirrored block group profiles, read picks the stripe based
295  * on these policies.
296  */
297 enum btrfs_read_policy {
298 	/* Use process PID to choose the stripe */
299 	BTRFS_READ_POLICY_PID,
300 	BTRFS_NR_READ_POLICY,
301 };
302 
303 #ifdef CONFIG_BTRFS_DEBUG
304 /*
305  * Checksum mode - offload it to workqueues or do it synchronously in
306  * btrfs_submit_chunk().
307  */
308 enum btrfs_offload_csum_mode {
309 	/*
310 	 * Choose offloading checksum or do it synchronously automatically.
311 	 * Do it synchronously if the checksum is fast, or offload to workqueues
312 	 * otherwise.
313 	 */
314 	BTRFS_OFFLOAD_CSUM_AUTO,
315 	/* Always offload checksum to workqueues. */
316 	BTRFS_OFFLOAD_CSUM_FORCE_ON,
317 	/* Never offload checksum to workqueues. */
318 	BTRFS_OFFLOAD_CSUM_FORCE_OFF,
319 };
320 #endif
321 
322 struct btrfs_fs_devices {
323 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
324 
325 	/*
326 	 * UUID written into the btree blocks:
327 	 *
328 	 * - If metadata_uuid != fsid then super block must have
329 	 *   BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag set.
330 	 *
331 	 * - Following shall be true at all times:
332 	 *   - metadata_uuid == btrfs_header::fsid
333 	 *   - metadata_uuid == btrfs_dev_item::fsid
334 	 *
335 	 * - Relations between fsid and metadata_uuid in sb and fs_devices:
336 	 *   - Normal:
337 	 *       fs_devices->fsid == fs_devices->metadata_uuid == sb->fsid
338 	 *       sb->metadata_uuid == 0
339 	 *
340 	 *   - When the BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag is set:
341 	 *       fs_devices->fsid == sb->fsid
342 	 *       fs_devices->metadata_uuid == sb->metadata_uuid
343 	 *
344 	 *   - When in-memory fs_devices->temp_fsid is true
345 	 *	 fs_devices->fsid = random
346 	 *	 fs_devices->metadata_uuid == sb->fsid
347 	 */
348 	u8 metadata_uuid[BTRFS_FSID_SIZE];
349 
350 	struct list_head fs_list;
351 
352 	/*
353 	 * Number of devices under this fsid including missing and
354 	 * replace-target device and excludes seed devices.
355 	 */
356 	u64 num_devices;
357 
358 	/*
359 	 * The number of devices that successfully opened, including
360 	 * replace-target, excludes seed devices.
361 	 */
362 	u64 open_devices;
363 
364 	/* The number of devices that are under the chunk allocation list. */
365 	u64 rw_devices;
366 
367 	/* Count of missing devices under this fsid excluding seed device. */
368 	u64 missing_devices;
369 	u64 total_rw_bytes;
370 
371 	/*
372 	 * Count of devices from btrfs_super_block::num_devices for this fsid,
373 	 * which includes the seed device, excludes the transient replace-target
374 	 * device.
375 	 */
376 	u64 total_devices;
377 
378 	/* Highest generation number of seen devices */
379 	u64 latest_generation;
380 
381 	/*
382 	 * The mount device or a device with highest generation after removal
383 	 * or replace.
384 	 */
385 	struct btrfs_device *latest_dev;
386 
387 	/*
388 	 * All of the devices in the filesystem, protected by a mutex so we can
389 	 * safely walk it to write out the super blocks without worrying about
390 	 * adding/removing by the multi-device code. Scrubbing super block can
391 	 * kick off supers writing by holding this mutex lock.
392 	 */
393 	struct mutex device_list_mutex;
394 
395 	/* List of all devices, protected by device_list_mutex */
396 	struct list_head devices;
397 
398 	/* Devices which can satisfy space allocation. Protected by * chunk_mutex. */
399 	struct list_head alloc_list;
400 
401 	struct list_head seed_list;
402 
403 	/* Count fs-devices opened. */
404 	int opened;
405 
406 	/* Set when we find or add a device that doesn't have the nonrot flag set. */
407 	bool rotating;
408 	/* Devices support TRIM/discard commands. */
409 	bool discardable;
410 	/* The filesystem is a seed filesystem. */
411 	bool seeding;
412 	/* The mount needs to use a randomly generated fsid. */
413 	bool temp_fsid;
414 
415 	struct btrfs_fs_info *fs_info;
416 	/* sysfs kobjects */
417 	struct kobject fsid_kobj;
418 	struct kobject *devices_kobj;
419 	struct kobject *devinfo_kobj;
420 	struct completion kobj_unregister;
421 
422 	enum btrfs_chunk_allocation_policy chunk_alloc_policy;
423 
424 	/* Policy used to read the mirrored stripes. */
425 	enum btrfs_read_policy read_policy;
426 
427 #ifdef CONFIG_BTRFS_DEBUG
428 	/* Checksum mode - offload it or do it synchronously. */
429 	enum btrfs_offload_csum_mode offload_csum_mode;
430 #endif
431 };
432 
433 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
434 			- sizeof(struct btrfs_chunk))		\
435 			/ sizeof(struct btrfs_stripe) + 1)
436 
437 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
438 				- 2 * sizeof(struct btrfs_disk_key)	\
439 				- 2 * sizeof(struct btrfs_chunk))	\
440 				/ sizeof(struct btrfs_stripe) + 1)
441 
442 struct btrfs_io_stripe {
443 	struct btrfs_device *dev;
444 	/* Block mapping. */
445 	u64 physical;
446 	u64 length;
447 	bool is_scrub;
448 	/* For the endio handler. */
449 	struct btrfs_io_context *bioc;
450 };
451 
452 struct btrfs_discard_stripe {
453 	struct btrfs_device *dev;
454 	u64 physical;
455 	u64 length;
456 };
457 
458 /*
459  * Context for IO subsmission for device stripe.
460  *
461  * - Track the unfinished mirrors for mirror based profiles
462  *   Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
463  *
464  * - Contain the logical -> physical mapping info
465  *   Used by submit_stripe_bio() for mapping logical bio
466  *   into physical device address.
467  *
468  * - Contain device replace info
469  *   Used by handle_ops_on_dev_replace() to copy logical bios
470  *   into the new device.
471  *
472  * - Contain RAID56 full stripe logical bytenrs
473  */
474 struct btrfs_io_context {
475 	refcount_t refs;
476 	struct btrfs_fs_info *fs_info;
477 	/* Taken from struct btrfs_chunk_map::type. */
478 	u64 map_type;
479 	struct bio *orig_bio;
480 	atomic_t error;
481 	u16 max_errors;
482 
483 	u64 logical;
484 	u64 size;
485 	/* Raid stripe tree ordered entry. */
486 	struct list_head rst_ordered_entry;
487 
488 	/*
489 	 * The total number of stripes, including the extra duplicated
490 	 * stripe for replace.
491 	 */
492 	u16 num_stripes;
493 
494 	/*
495 	 * The mirror_num of this bioc.
496 	 *
497 	 * This is for reads which use 0 as mirror_num, thus we should return a
498 	 * valid mirror_num (>0) for the reader.
499 	 */
500 	u16 mirror_num;
501 
502 	/*
503 	 * The following two members are for dev-replace case only.
504 	 *
505 	 * @replace_nr_stripes:	Number of duplicated stripes which need to be
506 	 *			written to replace target.
507 	 *			Should be <= 2 (2 for DUP, otherwise <= 1).
508 	 * @replace_stripe_src:	The array indicates where the duplicated stripes
509 	 *			are from.
510 	 *
511 	 * The @replace_stripe_src[] array is mostly for RAID56 cases.
512 	 * As non-RAID56 stripes share the same contents of the mapped range,
513 	 * thus no need to bother where the duplicated ones are from.
514 	 *
515 	 * But for RAID56 case, all stripes contain different contents, thus
516 	 * we need a way to know the mapping.
517 	 *
518 	 * There is an example for the two members, using a RAID5 write:
519 	 *
520 	 *   num_stripes:	4 (3 + 1 duplicated write)
521 	 *   stripes[0]:	dev = devid 1, physical = X
522 	 *   stripes[1]:	dev = devid 2, physical = Y
523 	 *   stripes[2]:	dev = devid 3, physical = Z
524 	 *   stripes[3]:	dev = devid 0, physical = Y
525 	 *
526 	 * replace_nr_stripes = 1
527 	 * replace_stripe_src = 1	<- Means stripes[1] is involved in replace.
528 	 *				   The duplicated stripe index would be
529 	 *				   (@num_stripes - 1).
530 	 *
531 	 * Note, that we can still have cases replace_nr_stripes = 2 for DUP.
532 	 * In that case, all stripes share the same content, thus we don't
533 	 * need to bother @replace_stripe_src value at all.
534 	 */
535 	u16 replace_nr_stripes;
536 	s16 replace_stripe_src;
537 	/*
538 	 * Logical bytenr of the full stripe start, only for RAID56 cases.
539 	 *
540 	 * When this value is set to other than (u64)-1, the stripes[] should
541 	 * follow this pattern:
542 	 *
543 	 * (real_stripes = num_stripes - replace_nr_stripes)
544 	 * (data_stripes = (is_raid6) ? (real_stripes - 2) : (real_stripes - 1))
545 	 *
546 	 * stripes[0]:			The first data stripe
547 	 * stripes[1]:			The second data stripe
548 	 * ...
549 	 * stripes[data_stripes - 1]:	The last data stripe
550 	 * stripes[data_stripes]:	The P stripe
551 	 * stripes[data_stripes + 1]:	The Q stripe (only for RAID6).
552 	 */
553 	u64 full_stripe_logical;
554 	struct btrfs_io_stripe stripes[];
555 };
556 
557 struct btrfs_device_info {
558 	struct btrfs_device *dev;
559 	u64 dev_offset;
560 	u64 max_avail;
561 	u64 total_avail;
562 };
563 
564 struct btrfs_raid_attr {
565 	u8 sub_stripes;		/* sub_stripes info for map */
566 	u8 dev_stripes;		/* stripes per dev */
567 	u8 devs_max;		/* max devs to use */
568 	u8 devs_min;		/* min devs needed */
569 	u8 tolerated_failures;	/* max tolerated fail devs */
570 	u8 devs_increment;	/* ndevs has to be a multiple of this */
571 	u8 ncopies;		/* how many copies to data has */
572 	u8 nparity;		/* number of stripes worth of bytes to store
573 				 * parity information */
574 	u8 mindev_error;	/* error code if min devs requisite is unmet */
575 	const char raid_name[8]; /* name of the raid */
576 	u64 bg_flag;		/* block group flag of the raid */
577 };
578 
579 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
580 
581 struct btrfs_chunk_map {
582 	struct rb_node rb_node;
583 	/* For mount time dev extent verification. */
584 	int verified_stripes;
585 	refcount_t refs;
586 	u64 start;
587 	u64 chunk_len;
588 	u64 stripe_size;
589 	u64 type;
590 	int io_align;
591 	int io_width;
592 	int num_stripes;
593 	int sub_stripes;
594 	struct btrfs_io_stripe stripes[];
595 };
596 
597 #define btrfs_chunk_map_size(n) (sizeof(struct btrfs_chunk_map) + \
598 				 (sizeof(struct btrfs_io_stripe) * (n)))
599 
600 static inline void btrfs_free_chunk_map(struct btrfs_chunk_map *map)
601 {
602 	if (map && refcount_dec_and_test(&map->refs)) {
603 		ASSERT(RB_EMPTY_NODE(&map->rb_node));
604 		kfree(map);
605 	}
606 }
607 
608 struct btrfs_balance_control {
609 	struct btrfs_balance_args data;
610 	struct btrfs_balance_args meta;
611 	struct btrfs_balance_args sys;
612 
613 	u64 flags;
614 
615 	struct btrfs_balance_progress stat;
616 };
617 
618 /*
619  * Search for a given device by the set parameters
620  */
621 struct btrfs_dev_lookup_args {
622 	u64 devid;
623 	u8 *uuid;
624 	u8 *fsid;
625 	bool missing;
626 };
627 
628 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
629 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
630 
631 #define BTRFS_DEV_LOOKUP_ARGS(name) \
632 	struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
633 
634 enum btrfs_map_op {
635 	BTRFS_MAP_READ,
636 	BTRFS_MAP_WRITE,
637 	BTRFS_MAP_GET_READ_MIRRORS,
638 };
639 
640 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
641 {
642 	switch (bio_op(bio)) {
643 	case REQ_OP_WRITE:
644 	case REQ_OP_ZONE_APPEND:
645 		return BTRFS_MAP_WRITE;
646 	default:
647 		WARN_ON_ONCE(1);
648 		fallthrough;
649 	case REQ_OP_READ:
650 		return BTRFS_MAP_READ;
651 	}
652 }
653 
654 static inline unsigned long btrfs_chunk_item_size(int num_stripes)
655 {
656 	ASSERT(num_stripes);
657 	return sizeof(struct btrfs_chunk) +
658 		sizeof(struct btrfs_stripe) * (num_stripes - 1);
659 }
660 
661 /*
662  * Do the type safe conversion from stripe_nr to offset inside the chunk.
663  *
664  * @stripe_nr is u32, with left shift it can overflow u32 for chunks larger
665  * than 4G.  This does the proper type cast to avoid overflow.
666  */
667 static inline u64 btrfs_stripe_nr_to_offset(u32 stripe_nr)
668 {
669 	return (u64)stripe_nr << BTRFS_STRIPE_LEN_SHIFT;
670 }
671 
672 void btrfs_get_bioc(struct btrfs_io_context *bioc);
673 void btrfs_put_bioc(struct btrfs_io_context *bioc);
674 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
675 		    u64 logical, u64 *length,
676 		    struct btrfs_io_context **bioc_ret,
677 		    struct btrfs_io_stripe *smap, int *mirror_num_ret);
678 int btrfs_map_repair_block(struct btrfs_fs_info *fs_info,
679 			   struct btrfs_io_stripe *smap, u64 logical,
680 			   u32 length, int mirror_num);
681 struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,
682 					       u64 logical, u64 *length_ret,
683 					       u32 *num_stripes);
684 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
685 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
686 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
687 					    u64 type);
688 void btrfs_mapping_tree_free(struct btrfs_fs_info *fs_info);
689 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
690 		       blk_mode_t flags, void *holder);
691 struct btrfs_device *btrfs_scan_one_device(const char *path, blk_mode_t flags,
692 					   bool mount_arg_dev);
693 int btrfs_forget_devices(dev_t devt);
694 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
695 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
696 void btrfs_assign_next_active_device(struct btrfs_device *device,
697 				     struct btrfs_device *this_dev);
698 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
699 						  u64 devid,
700 						  const char *devpath);
701 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
702 				 struct btrfs_dev_lookup_args *args,
703 				 const char *path);
704 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
705 					const u64 *devid, const u8 *uuid,
706 					const char *path);
707 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
708 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
709 		    struct btrfs_dev_lookup_args *args,
710 		    struct file **bdev_file);
711 void __exit btrfs_cleanup_fs_uuids(void);
712 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
713 int btrfs_grow_device(struct btrfs_trans_handle *trans,
714 		      struct btrfs_device *device, u64 new_size);
715 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
716 				       const struct btrfs_dev_lookup_args *args);
717 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
718 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
719 int btrfs_balance(struct btrfs_fs_info *fs_info,
720 		  struct btrfs_balance_control *bctl,
721 		  struct btrfs_ioctl_balance_args *bargs);
722 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
723 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
724 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
725 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
726 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
727 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
728 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
729 int btrfs_uuid_scan_kthread(void *data);
730 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
731 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
732 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
733 			struct btrfs_ioctl_get_dev_stats *stats);
734 int btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
735 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
736 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
737 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
738 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
739 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
740 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
741 			   u64 logical, u64 len);
742 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
743 				    u64 logical);
744 u64 btrfs_calc_stripe_length(const struct btrfs_chunk_map *map);
745 int btrfs_nr_parity_stripes(u64 type);
746 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
747 				     struct btrfs_block_group *bg);
748 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
749 
750 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
751 struct btrfs_chunk_map *btrfs_alloc_chunk_map(int num_stripes, gfp_t gfp);
752 int btrfs_add_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
753 #endif
754 
755 struct btrfs_chunk_map *btrfs_find_chunk_map(struct btrfs_fs_info *fs_info,
756 					     u64 logical, u64 length);
757 struct btrfs_chunk_map *btrfs_find_chunk_map_nolock(struct btrfs_fs_info *fs_info,
758 						    u64 logical, u64 length);
759 struct btrfs_chunk_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
760 					    u64 logical, u64 length);
761 void btrfs_remove_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
762 void btrfs_release_disk_super(struct btrfs_super_block *super);
763 
764 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
765 				      int index)
766 {
767 	atomic_inc(dev->dev_stat_values + index);
768 	/*
769 	 * This memory barrier orders stores updating statistics before stores
770 	 * updating dev_stats_ccnt.
771 	 *
772 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
773 	 */
774 	smp_mb__before_atomic();
775 	atomic_inc(&dev->dev_stats_ccnt);
776 }
777 
778 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
779 				      int index)
780 {
781 	return atomic_read(dev->dev_stat_values + index);
782 }
783 
784 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
785 						int index)
786 {
787 	int ret;
788 
789 	ret = atomic_xchg(dev->dev_stat_values + index, 0);
790 	/*
791 	 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
792 	 * - RMW operations that have a return value are fully ordered;
793 	 *
794 	 * This implicit memory barriers is paired with the smp_rmb in
795 	 * btrfs_run_dev_stats
796 	 */
797 	atomic_inc(&dev->dev_stats_ccnt);
798 	return ret;
799 }
800 
801 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
802 				      int index, unsigned long val)
803 {
804 	atomic_set(dev->dev_stat_values + index, val);
805 	/*
806 	 * This memory barrier orders stores updating statistics before stores
807 	 * updating dev_stats_ccnt.
808 	 *
809 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
810 	 */
811 	smp_mb__before_atomic();
812 	atomic_inc(&dev->dev_stats_ccnt);
813 }
814 
815 static inline const char *btrfs_dev_name(const struct btrfs_device *device)
816 {
817 	if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
818 		return "<missing disk>";
819 	else
820 		return rcu_str_deref(device->name);
821 }
822 
823 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
824 
825 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
826 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
827 					struct btrfs_device *failing_dev);
828 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, struct btrfs_device *device);
829 
830 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
831 int btrfs_bg_type_to_factor(u64 flags);
832 const char *btrfs_bg_type_to_raid_name(u64 flags);
833 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
834 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);
835 
836 bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr);
837 u8 *btrfs_sb_fsid_ptr(struct btrfs_super_block *sb);
838 
839 #endif
840