xref: /linux/fs/btrfs/volumes.h (revision 55d0969c451159cff86949b38c39171cab962069)
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 /*
300  * Read policies for mirrored block group profiles, read picks the stripe based
301  * on these policies.
302  */
303 enum btrfs_read_policy {
304 	/* Use process PID to choose the stripe */
305 	BTRFS_READ_POLICY_PID,
306 	BTRFS_NR_READ_POLICY,
307 };
308 
309 #ifdef CONFIG_BTRFS_DEBUG
310 /*
311  * Checksum mode - offload it to workqueues or do it synchronously in
312  * btrfs_submit_chunk().
313  */
314 enum btrfs_offload_csum_mode {
315 	/*
316 	 * Choose offloading checksum or do it synchronously automatically.
317 	 * Do it synchronously if the checksum is fast, or offload to workqueues
318 	 * otherwise.
319 	 */
320 	BTRFS_OFFLOAD_CSUM_AUTO,
321 	/* Always offload checksum to workqueues. */
322 	BTRFS_OFFLOAD_CSUM_FORCE_ON,
323 	/* Never offload checksum to workqueues. */
324 	BTRFS_OFFLOAD_CSUM_FORCE_OFF,
325 };
326 #endif
327 
328 struct btrfs_fs_devices {
329 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
330 
331 	/*
332 	 * UUID written into the btree blocks:
333 	 *
334 	 * - If metadata_uuid != fsid then super block must have
335 	 *   BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag set.
336 	 *
337 	 * - Following shall be true at all times:
338 	 *   - metadata_uuid == btrfs_header::fsid
339 	 *   - metadata_uuid == btrfs_dev_item::fsid
340 	 *
341 	 * - Relations between fsid and metadata_uuid in sb and fs_devices:
342 	 *   - Normal:
343 	 *       fs_devices->fsid == fs_devices->metadata_uuid == sb->fsid
344 	 *       sb->metadata_uuid == 0
345 	 *
346 	 *   - When the BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag is set:
347 	 *       fs_devices->fsid == sb->fsid
348 	 *       fs_devices->metadata_uuid == sb->metadata_uuid
349 	 *
350 	 *   - When in-memory fs_devices->temp_fsid is true
351 	 *	 fs_devices->fsid = random
352 	 *	 fs_devices->metadata_uuid == sb->fsid
353 	 */
354 	u8 metadata_uuid[BTRFS_FSID_SIZE];
355 
356 	struct list_head fs_list;
357 
358 	/*
359 	 * Number of devices under this fsid including missing and
360 	 * replace-target device and excludes seed devices.
361 	 */
362 	u64 num_devices;
363 
364 	/*
365 	 * The number of devices that successfully opened, including
366 	 * replace-target, excludes seed devices.
367 	 */
368 	u64 open_devices;
369 
370 	/* The number of devices that are under the chunk allocation list. */
371 	u64 rw_devices;
372 
373 	/* Count of missing devices under this fsid excluding seed device. */
374 	u64 missing_devices;
375 	u64 total_rw_bytes;
376 
377 	/*
378 	 * Count of devices from btrfs_super_block::num_devices for this fsid,
379 	 * which includes the seed device, excludes the transient replace-target
380 	 * device.
381 	 */
382 	u64 total_devices;
383 
384 	/* Highest generation number of seen devices */
385 	u64 latest_generation;
386 
387 	/*
388 	 * The mount device or a device with highest generation after removal
389 	 * or replace.
390 	 */
391 	struct btrfs_device *latest_dev;
392 
393 	/*
394 	 * All of the devices in the filesystem, protected by a mutex so we can
395 	 * safely walk it to write out the super blocks without worrying about
396 	 * adding/removing by the multi-device code. Scrubbing super block can
397 	 * kick off supers writing by holding this mutex lock.
398 	 */
399 	struct mutex device_list_mutex;
400 
401 	/* List of all devices, protected by device_list_mutex */
402 	struct list_head devices;
403 
404 	/* Devices which can satisfy space allocation. Protected by * chunk_mutex. */
405 	struct list_head alloc_list;
406 
407 	struct list_head seed_list;
408 
409 	/* Count fs-devices opened. */
410 	int opened;
411 
412 	/* Set when we find or add a device that doesn't have the nonrot flag set. */
413 	bool rotating;
414 	/* Devices support TRIM/discard commands. */
415 	bool discardable;
416 	/* The filesystem is a seed filesystem. */
417 	bool seeding;
418 	/* The mount needs to use a randomly generated fsid. */
419 	bool temp_fsid;
420 
421 	struct btrfs_fs_info *fs_info;
422 	/* sysfs kobjects */
423 	struct kobject fsid_kobj;
424 	struct kobject *devices_kobj;
425 	struct kobject *devinfo_kobj;
426 	struct completion kobj_unregister;
427 
428 	enum btrfs_chunk_allocation_policy chunk_alloc_policy;
429 
430 	/* Policy used to read the mirrored stripes. */
431 	enum btrfs_read_policy read_policy;
432 
433 #ifdef CONFIG_BTRFS_DEBUG
434 	/* Checksum mode - offload it or do it synchronously. */
435 	enum btrfs_offload_csum_mode offload_csum_mode;
436 #endif
437 };
438 
439 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
440 			- sizeof(struct btrfs_chunk))		\
441 			/ sizeof(struct btrfs_stripe) + 1)
442 
443 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
444 				- 2 * sizeof(struct btrfs_disk_key)	\
445 				- 2 * sizeof(struct btrfs_chunk))	\
446 				/ sizeof(struct btrfs_stripe) + 1)
447 
448 struct btrfs_io_stripe {
449 	struct btrfs_device *dev;
450 	/* Block mapping. */
451 	u64 physical;
452 	u64 length;
453 	bool rst_search_commit_root;
454 	/* For the endio handler. */
455 	struct btrfs_io_context *bioc;
456 };
457 
458 struct btrfs_discard_stripe {
459 	struct btrfs_device *dev;
460 	u64 physical;
461 	u64 length;
462 };
463 
464 /*
465  * Context for IO subsmission for device stripe.
466  *
467  * - Track the unfinished mirrors for mirror based profiles
468  *   Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
469  *
470  * - Contain the logical -> physical mapping info
471  *   Used by submit_stripe_bio() for mapping logical bio
472  *   into physical device address.
473  *
474  * - Contain device replace info
475  *   Used by handle_ops_on_dev_replace() to copy logical bios
476  *   into the new device.
477  *
478  * - Contain RAID56 full stripe logical bytenrs
479  */
480 struct btrfs_io_context {
481 	refcount_t refs;
482 	struct btrfs_fs_info *fs_info;
483 	/* Taken from struct btrfs_chunk_map::type. */
484 	u64 map_type;
485 	struct bio *orig_bio;
486 	atomic_t error;
487 	u16 max_errors;
488 
489 	u64 logical;
490 	u64 size;
491 	/* Raid stripe tree ordered entry. */
492 	struct list_head rst_ordered_entry;
493 
494 	/*
495 	 * The total number of stripes, including the extra duplicated
496 	 * stripe for replace.
497 	 */
498 	u16 num_stripes;
499 
500 	/*
501 	 * The mirror_num of this bioc.
502 	 *
503 	 * This is for reads which use 0 as mirror_num, thus we should return a
504 	 * valid mirror_num (>0) for the reader.
505 	 */
506 	u16 mirror_num;
507 
508 	/*
509 	 * The following two members are for dev-replace case only.
510 	 *
511 	 * @replace_nr_stripes:	Number of duplicated stripes which need to be
512 	 *			written to replace target.
513 	 *			Should be <= 2 (2 for DUP, otherwise <= 1).
514 	 * @replace_stripe_src:	The array indicates where the duplicated stripes
515 	 *			are from.
516 	 *
517 	 * The @replace_stripe_src[] array is mostly for RAID56 cases.
518 	 * As non-RAID56 stripes share the same contents of the mapped range,
519 	 * thus no need to bother where the duplicated ones are from.
520 	 *
521 	 * But for RAID56 case, all stripes contain different contents, thus
522 	 * we need a way to know the mapping.
523 	 *
524 	 * There is an example for the two members, using a RAID5 write:
525 	 *
526 	 *   num_stripes:	4 (3 + 1 duplicated write)
527 	 *   stripes[0]:	dev = devid 1, physical = X
528 	 *   stripes[1]:	dev = devid 2, physical = Y
529 	 *   stripes[2]:	dev = devid 3, physical = Z
530 	 *   stripes[3]:	dev = devid 0, physical = Y
531 	 *
532 	 * replace_nr_stripes = 1
533 	 * replace_stripe_src = 1	<- Means stripes[1] is involved in replace.
534 	 *				   The duplicated stripe index would be
535 	 *				   (@num_stripes - 1).
536 	 *
537 	 * Note, that we can still have cases replace_nr_stripes = 2 for DUP.
538 	 * In that case, all stripes share the same content, thus we don't
539 	 * need to bother @replace_stripe_src value at all.
540 	 */
541 	u16 replace_nr_stripes;
542 	s16 replace_stripe_src;
543 	/*
544 	 * Logical bytenr of the full stripe start, only for RAID56 cases.
545 	 *
546 	 * When this value is set to other than (u64)-1, the stripes[] should
547 	 * follow this pattern:
548 	 *
549 	 * (real_stripes = num_stripes - replace_nr_stripes)
550 	 * (data_stripes = (is_raid6) ? (real_stripes - 2) : (real_stripes - 1))
551 	 *
552 	 * stripes[0]:			The first data stripe
553 	 * stripes[1]:			The second data stripe
554 	 * ...
555 	 * stripes[data_stripes - 1]:	The last data stripe
556 	 * stripes[data_stripes]:	The P stripe
557 	 * stripes[data_stripes + 1]:	The Q stripe (only for RAID6).
558 	 */
559 	u64 full_stripe_logical;
560 	struct btrfs_io_stripe stripes[];
561 };
562 
563 struct btrfs_device_info {
564 	struct btrfs_device *dev;
565 	u64 dev_offset;
566 	u64 max_avail;
567 	u64 total_avail;
568 };
569 
570 struct btrfs_raid_attr {
571 	u8 sub_stripes;		/* sub_stripes info for map */
572 	u8 dev_stripes;		/* stripes per dev */
573 	u8 devs_max;		/* max devs to use */
574 	u8 devs_min;		/* min devs needed */
575 	u8 tolerated_failures;	/* max tolerated fail devs */
576 	u8 devs_increment;	/* ndevs has to be a multiple of this */
577 	u8 ncopies;		/* how many copies to data has */
578 	u8 nparity;		/* number of stripes worth of bytes to store
579 				 * parity information */
580 	u8 mindev_error;	/* error code if min devs requisite is unmet */
581 	const char raid_name[8]; /* name of the raid */
582 	u64 bg_flag;		/* block group flag of the raid */
583 };
584 
585 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
586 
587 struct btrfs_chunk_map {
588 	struct rb_node rb_node;
589 	/* For mount time dev extent verification. */
590 	int verified_stripes;
591 	refcount_t refs;
592 	u64 start;
593 	u64 chunk_len;
594 	u64 stripe_size;
595 	u64 type;
596 	int io_align;
597 	int io_width;
598 	int num_stripes;
599 	int sub_stripes;
600 	struct btrfs_io_stripe stripes[];
601 };
602 
603 #define btrfs_chunk_map_size(n) (sizeof(struct btrfs_chunk_map) + \
604 				 (sizeof(struct btrfs_io_stripe) * (n)))
605 
606 static inline void btrfs_free_chunk_map(struct btrfs_chunk_map *map)
607 {
608 	if (map && refcount_dec_and_test(&map->refs)) {
609 		ASSERT(RB_EMPTY_NODE(&map->rb_node));
610 		kfree(map);
611 	}
612 }
613 
614 struct btrfs_balance_control {
615 	struct btrfs_balance_args data;
616 	struct btrfs_balance_args meta;
617 	struct btrfs_balance_args sys;
618 
619 	u64 flags;
620 
621 	struct btrfs_balance_progress stat;
622 };
623 
624 /*
625  * Search for a given device by the set parameters
626  */
627 struct btrfs_dev_lookup_args {
628 	u64 devid;
629 	u8 *uuid;
630 	u8 *fsid;
631 	bool missing;
632 };
633 
634 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
635 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
636 
637 #define BTRFS_DEV_LOOKUP_ARGS(name) \
638 	struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
639 
640 enum btrfs_map_op {
641 	BTRFS_MAP_READ,
642 	BTRFS_MAP_WRITE,
643 	BTRFS_MAP_GET_READ_MIRRORS,
644 };
645 
646 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
647 {
648 	switch (bio_op(bio)) {
649 	case REQ_OP_WRITE:
650 	case REQ_OP_ZONE_APPEND:
651 		return BTRFS_MAP_WRITE;
652 	default:
653 		WARN_ON_ONCE(1);
654 		fallthrough;
655 	case REQ_OP_READ:
656 		return BTRFS_MAP_READ;
657 	}
658 }
659 
660 static inline unsigned long btrfs_chunk_item_size(int num_stripes)
661 {
662 	ASSERT(num_stripes);
663 	return sizeof(struct btrfs_chunk) +
664 		sizeof(struct btrfs_stripe) * (num_stripes - 1);
665 }
666 
667 /*
668  * Do the type safe conversion from stripe_nr to offset inside the chunk.
669  *
670  * @stripe_nr is u32, with left shift it can overflow u32 for chunks larger
671  * than 4G.  This does the proper type cast to avoid overflow.
672  */
673 static inline u64 btrfs_stripe_nr_to_offset(u32 stripe_nr)
674 {
675 	return (u64)stripe_nr << BTRFS_STRIPE_LEN_SHIFT;
676 }
677 
678 void btrfs_get_bioc(struct btrfs_io_context *bioc);
679 void btrfs_put_bioc(struct btrfs_io_context *bioc);
680 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
681 		    u64 logical, u64 *length,
682 		    struct btrfs_io_context **bioc_ret,
683 		    struct btrfs_io_stripe *smap, int *mirror_num_ret);
684 int btrfs_map_repair_block(struct btrfs_fs_info *fs_info,
685 			   struct btrfs_io_stripe *smap, u64 logical,
686 			   u32 length, int mirror_num);
687 struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,
688 					       u64 logical, u64 *length_ret,
689 					       u32 *num_stripes);
690 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
691 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
692 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
693 					    u64 type);
694 void btrfs_mapping_tree_free(struct btrfs_fs_info *fs_info);
695 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
696 		       blk_mode_t flags, void *holder);
697 struct btrfs_device *btrfs_scan_one_device(const char *path, blk_mode_t flags,
698 					   bool mount_arg_dev);
699 int btrfs_forget_devices(dev_t devt);
700 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
701 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
702 void btrfs_assign_next_active_device(struct btrfs_device *device,
703 				     struct btrfs_device *this_dev);
704 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
705 						  u64 devid,
706 						  const char *devpath);
707 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
708 				 struct btrfs_dev_lookup_args *args,
709 				 const char *path);
710 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
711 					const u64 *devid, const u8 *uuid,
712 					const char *path);
713 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
714 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
715 		    struct btrfs_dev_lookup_args *args,
716 		    struct file **bdev_file);
717 void __exit btrfs_cleanup_fs_uuids(void);
718 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
719 int btrfs_grow_device(struct btrfs_trans_handle *trans,
720 		      struct btrfs_device *device, u64 new_size);
721 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
722 				       const struct btrfs_dev_lookup_args *args);
723 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
724 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
725 int btrfs_balance(struct btrfs_fs_info *fs_info,
726 		  struct btrfs_balance_control *bctl,
727 		  struct btrfs_ioctl_balance_args *bargs);
728 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
729 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
730 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
731 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
732 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
733 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
734 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
735 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
736 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
737 			struct btrfs_ioctl_get_dev_stats *stats);
738 int btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
739 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
740 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
741 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
742 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
743 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
744 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
745 			   u64 logical, u64 len);
746 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
747 				    u64 logical);
748 u64 btrfs_calc_stripe_length(const struct btrfs_chunk_map *map);
749 int btrfs_nr_parity_stripes(u64 type);
750 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
751 				     struct btrfs_block_group *bg);
752 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
753 
754 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
755 struct btrfs_chunk_map *btrfs_alloc_chunk_map(int num_stripes, gfp_t gfp);
756 int btrfs_add_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
757 #endif
758 
759 struct btrfs_chunk_map *btrfs_find_chunk_map(struct btrfs_fs_info *fs_info,
760 					     u64 logical, u64 length);
761 struct btrfs_chunk_map *btrfs_find_chunk_map_nolock(struct btrfs_fs_info *fs_info,
762 						    u64 logical, u64 length);
763 struct btrfs_chunk_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
764 					    u64 logical, u64 length);
765 void btrfs_remove_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
766 void btrfs_release_disk_super(struct btrfs_super_block *super);
767 
768 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
769 				      int index)
770 {
771 	atomic_inc(dev->dev_stat_values + index);
772 	/*
773 	 * This memory barrier orders stores updating statistics before stores
774 	 * updating dev_stats_ccnt.
775 	 *
776 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
777 	 */
778 	smp_mb__before_atomic();
779 	atomic_inc(&dev->dev_stats_ccnt);
780 }
781 
782 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
783 				      int index)
784 {
785 	return atomic_read(dev->dev_stat_values + index);
786 }
787 
788 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
789 						int index)
790 {
791 	int ret;
792 
793 	ret = atomic_xchg(dev->dev_stat_values + index, 0);
794 	/*
795 	 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
796 	 * - RMW operations that have a return value are fully ordered;
797 	 *
798 	 * This implicit memory barriers is paired with the smp_rmb in
799 	 * btrfs_run_dev_stats
800 	 */
801 	atomic_inc(&dev->dev_stats_ccnt);
802 	return ret;
803 }
804 
805 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
806 				      int index, unsigned long val)
807 {
808 	atomic_set(dev->dev_stat_values + index, val);
809 	/*
810 	 * This memory barrier orders stores updating statistics before stores
811 	 * updating dev_stats_ccnt.
812 	 *
813 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
814 	 */
815 	smp_mb__before_atomic();
816 	atomic_inc(&dev->dev_stats_ccnt);
817 }
818 
819 static inline const char *btrfs_dev_name(const struct btrfs_device *device)
820 {
821 	if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
822 		return "<missing disk>";
823 	else
824 		return rcu_str_deref(device->name);
825 }
826 
827 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
828 
829 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
830 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
831 					struct btrfs_device *failing_dev);
832 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, struct btrfs_device *device);
833 
834 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
835 int btrfs_bg_type_to_factor(u64 flags);
836 const char *btrfs_bg_type_to_raid_name(u64 flags);
837 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
838 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);
839 
840 bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr);
841 const u8 *btrfs_sb_fsid_ptr(const struct btrfs_super_block *sb);
842 
843 #endif
844