xref: /linux/fs/btrfs/volumes.h (revision fd71def6d9abc5ae362fb9995d46049b7b0ed391)
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 "rcu-string.h"
25 #include "extent-io-tree.h"
26 
27 struct block_device;
28 struct bdev_handle;
29 struct btrfs_fs_info;
30 struct btrfs_block_group;
31 struct btrfs_trans_handle;
32 struct btrfs_transaction;
33 struct btrfs_zoned_device_info;
34 
35 #define BTRFS_MAX_DATA_CHUNK_SIZE	(10ULL * SZ_1G)
36 
37 /*
38  * Arbitratry 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(const_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 static_assert(const_ilog2(BTRFS_BLOCK_GROUP_RAID0) >
63 	      ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK));
64 
65 /* ilog2() can handle both constants and variables */
66 #define BTRFS_BG_FLAG_TO_INDEX(profile)					\
67 	ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1))
68 
69 enum btrfs_raid_types {
70 	/* SINGLE is the special one as it doesn't have on-disk bit. */
71 	BTRFS_RAID_SINGLE  = 0,
72 
73 	BTRFS_RAID_RAID0   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0),
74 	BTRFS_RAID_RAID1   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1),
75 	BTRFS_RAID_DUP	   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP),
76 	BTRFS_RAID_RAID10  = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10),
77 	BTRFS_RAID_RAID5   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5),
78 	BTRFS_RAID_RAID6   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6),
79 	BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3),
80 	BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4),
81 
82 	BTRFS_NR_RAID_TYPES
83 };
84 
85 /*
86  * Use sequence counter to get consistent device stat data on
87  * 32-bit processors.
88  */
89 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
90 #include <linux/seqlock.h>
91 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
92 #define btrfs_device_data_ordered_init(device)	\
93 	seqcount_init(&device->data_seqcount)
94 #else
95 #define btrfs_device_data_ordered_init(device) do { } while (0)
96 #endif
97 
98 #define BTRFS_DEV_STATE_WRITEABLE	(0)
99 #define BTRFS_DEV_STATE_IN_FS_METADATA	(1)
100 #define BTRFS_DEV_STATE_MISSING		(2)
101 #define BTRFS_DEV_STATE_REPLACE_TGT	(3)
102 #define BTRFS_DEV_STATE_FLUSH_SENT	(4)
103 #define BTRFS_DEV_STATE_NO_READA	(5)
104 
105 /* Special value encoding failure to write primary super block. */
106 #define BTRFS_SUPER_PRIMARY_WRITE_ERROR		(INT_MAX / 2)
107 
108 struct btrfs_fs_devices;
109 
110 struct btrfs_device {
111 	struct list_head dev_list; /* device_list_mutex */
112 	struct list_head dev_alloc_list; /* chunk mutex */
113 	struct list_head post_commit_list; /* chunk mutex */
114 	struct btrfs_fs_devices *fs_devices;
115 	struct btrfs_fs_info *fs_info;
116 
117 	struct rcu_string __rcu *name;
118 
119 	u64 generation;
120 
121 	struct file *bdev_file;
122 	struct block_device *bdev;
123 
124 	struct btrfs_zoned_device_info *zone_info;
125 
126 	/*
127 	 * Device's major-minor number. Must be set even if the device is not
128 	 * opened (bdev == NULL), unless the device is missing.
129 	 */
130 	dev_t devt;
131 	unsigned long dev_state;
132 	blk_status_t last_flush_error;
133 
134 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
135 	seqcount_t data_seqcount;
136 #endif
137 
138 	/* the internal btrfs device id */
139 	u64 devid;
140 
141 	/* size of the device in memory */
142 	u64 total_bytes;
143 
144 	/* size of the device on disk */
145 	u64 disk_total_bytes;
146 
147 	/* bytes used */
148 	u64 bytes_used;
149 
150 	/* optimal io alignment for this device */
151 	u32 io_align;
152 
153 	/* optimal io width for this device */
154 	u32 io_width;
155 	/* type and info about this device */
156 	u64 type;
157 
158 	/*
159 	 * Counter of super block write errors, values larger than
160 	 * BTRFS_SUPER_PRIMARY_WRITE_ERROR encode primary super block write failure.
161 	 */
162 	atomic_t sb_write_errors;
163 
164 	/* minimal io size for this device */
165 	u32 sector_size;
166 
167 	/* physical drive uuid (or lvm uuid) */
168 	u8 uuid[BTRFS_UUID_SIZE];
169 
170 	/*
171 	 * size of the device on the current transaction
172 	 *
173 	 * This variant is update when committing the transaction,
174 	 * and protected by chunk mutex
175 	 */
176 	u64 commit_total_bytes;
177 
178 	/* bytes used on the current transaction */
179 	u64 commit_bytes_used;
180 
181 	/* Bio used for flushing device barriers */
182 	struct bio flush_bio;
183 	struct completion flush_wait;
184 
185 	/* per-device scrub information */
186 	struct scrub_ctx *scrub_ctx;
187 
188 	/* disk I/O failure stats. For detailed description refer to
189 	 * enum btrfs_dev_stat_values in ioctl.h */
190 	int dev_stats_valid;
191 
192 	/* Counter to record the change of device stats */
193 	atomic_t dev_stats_ccnt;
194 	atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
195 
196 	struct extent_io_tree alloc_state;
197 
198 	struct completion kobj_unregister;
199 	/* For sysfs/FSID/devinfo/devid/ */
200 	struct kobject devid_kobj;
201 
202 	/* Bandwidth limit for scrub, in bytes */
203 	u64 scrub_speed_max;
204 };
205 
206 /*
207  * Block group or device which contains an active swapfile. Used for preventing
208  * unsafe operations while a swapfile is active.
209  *
210  * These are sorted on (ptr, inode) (note that a block group or device can
211  * contain more than one swapfile). We compare the pointer values because we
212  * don't actually care what the object is, we just need a quick check whether
213  * the object exists in the rbtree.
214  */
215 struct btrfs_swapfile_pin {
216 	struct rb_node node;
217 	void *ptr;
218 	struct inode *inode;
219 	/*
220 	 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr
221 	 * points to a struct btrfs_device.
222 	 */
223 	bool is_block_group;
224 	/*
225 	 * Only used when 'is_block_group' is true and it is the number of
226 	 * extents used by a swapfile for this block group ('ptr' field).
227 	 */
228 	int bg_extent_count;
229 };
230 
231 /*
232  * If we read those variants at the context of their own lock, we needn't
233  * use the following helpers, reading them directly is safe.
234  */
235 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
236 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
237 static inline u64							\
238 btrfs_device_get_##name(const struct btrfs_device *dev)			\
239 {									\
240 	u64 size;							\
241 	unsigned int seq;						\
242 									\
243 	do {								\
244 		seq = read_seqcount_begin(&dev->data_seqcount);		\
245 		size = dev->name;					\
246 	} while (read_seqcount_retry(&dev->data_seqcount, seq));	\
247 	return size;							\
248 }									\
249 									\
250 static inline void							\
251 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
252 {									\
253 	preempt_disable();						\
254 	write_seqcount_begin(&dev->data_seqcount);			\
255 	dev->name = size;						\
256 	write_seqcount_end(&dev->data_seqcount);			\
257 	preempt_enable();						\
258 }
259 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
260 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
261 static inline u64							\
262 btrfs_device_get_##name(const struct btrfs_device *dev)			\
263 {									\
264 	u64 size;							\
265 									\
266 	preempt_disable();						\
267 	size = dev->name;						\
268 	preempt_enable();						\
269 	return size;							\
270 }									\
271 									\
272 static inline void							\
273 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
274 {									\
275 	preempt_disable();						\
276 	dev->name = size;						\
277 	preempt_enable();						\
278 }
279 #else
280 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
281 static inline u64							\
282 btrfs_device_get_##name(const struct btrfs_device *dev)			\
283 {									\
284 	return dev->name;						\
285 }									\
286 									\
287 static inline void							\
288 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
289 {									\
290 	dev->name = size;						\
291 }
292 #endif
293 
294 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
295 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
296 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
297 
298 enum btrfs_chunk_allocation_policy {
299 	BTRFS_CHUNK_ALLOC_REGULAR,
300 	BTRFS_CHUNK_ALLOC_ZONED,
301 };
302 
303 #define BTRFS_DEFAULT_RR_MIN_CONTIG_READ	(SZ_256K)
304 /* Keep in sync with raid_attr table, current maximum is RAID1C4. */
305 #define BTRFS_RAID1_MAX_MIRRORS			(4)
306 /*
307  * Read policies for mirrored block group profiles, read picks the stripe based
308  * on these policies.
309  */
310 enum btrfs_read_policy {
311 	/* Use process PID to choose the stripe */
312 	BTRFS_READ_POLICY_PID,
313 #ifdef CONFIG_BTRFS_EXPERIMENTAL
314 	/* Balancing RAID1 reads across all striped devices (round-robin). */
315 	BTRFS_READ_POLICY_RR,
316 	/* Read from a specific device. */
317 	BTRFS_READ_POLICY_DEVID,
318 #endif
319 	BTRFS_NR_READ_POLICY,
320 };
321 
322 #ifdef CONFIG_BTRFS_EXPERIMENTAL
323 /*
324  * Checksum mode - offload it to workqueues or do it synchronously in
325  * btrfs_submit_chunk().
326  */
327 enum btrfs_offload_csum_mode {
328 	/*
329 	 * Choose offloading checksum or do it synchronously automatically.
330 	 * Do it synchronously if the checksum is fast, or offload to workqueues
331 	 * otherwise.
332 	 */
333 	BTRFS_OFFLOAD_CSUM_AUTO,
334 	/* Always offload checksum to workqueues. */
335 	BTRFS_OFFLOAD_CSUM_FORCE_ON,
336 	/* Never offload checksum to workqueues. */
337 	BTRFS_OFFLOAD_CSUM_FORCE_OFF,
338 };
339 #endif
340 
341 struct btrfs_fs_devices {
342 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
343 
344 	/*
345 	 * UUID written into the btree blocks:
346 	 *
347 	 * - If metadata_uuid != fsid then super block must have
348 	 *   BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag set.
349 	 *
350 	 * - Following shall be true at all times:
351 	 *   - metadata_uuid == btrfs_header::fsid
352 	 *   - metadata_uuid == btrfs_dev_item::fsid
353 	 *
354 	 * - Relations between fsid and metadata_uuid in sb and fs_devices:
355 	 *   - Normal:
356 	 *       fs_devices->fsid == fs_devices->metadata_uuid == sb->fsid
357 	 *       sb->metadata_uuid == 0
358 	 *
359 	 *   - When the BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag is set:
360 	 *       fs_devices->fsid == sb->fsid
361 	 *       fs_devices->metadata_uuid == sb->metadata_uuid
362 	 *
363 	 *   - When in-memory fs_devices->temp_fsid is true
364 	 *	 fs_devices->fsid = random
365 	 *	 fs_devices->metadata_uuid == sb->fsid
366 	 */
367 	u8 metadata_uuid[BTRFS_FSID_SIZE];
368 
369 	struct list_head fs_list;
370 
371 	/*
372 	 * Number of devices under this fsid including missing and
373 	 * replace-target device and excludes seed devices.
374 	 */
375 	u64 num_devices;
376 
377 	/*
378 	 * The number of devices that successfully opened, including
379 	 * replace-target, excludes seed devices.
380 	 */
381 	u64 open_devices;
382 
383 	/* The number of devices that are under the chunk allocation list. */
384 	u64 rw_devices;
385 
386 	/* Count of missing devices under this fsid excluding seed device. */
387 	u64 missing_devices;
388 	u64 total_rw_bytes;
389 
390 	/*
391 	 * Count of devices from btrfs_super_block::num_devices for this fsid,
392 	 * which includes the seed device, excludes the transient replace-target
393 	 * device.
394 	 */
395 	u64 total_devices;
396 
397 	/* Highest generation number of seen devices */
398 	u64 latest_generation;
399 
400 	/*
401 	 * The mount device or a device with highest generation after removal
402 	 * or replace.
403 	 */
404 	struct btrfs_device *latest_dev;
405 
406 	/*
407 	 * All of the devices in the filesystem, protected by a mutex so we can
408 	 * safely walk it to write out the super blocks without worrying about
409 	 * adding/removing by the multi-device code. Scrubbing super block can
410 	 * kick off supers writing by holding this mutex lock.
411 	 */
412 	struct mutex device_list_mutex;
413 
414 	/* List of all devices, protected by device_list_mutex */
415 	struct list_head devices;
416 
417 	/* Devices which can satisfy space allocation. Protected by * chunk_mutex. */
418 	struct list_head alloc_list;
419 
420 	struct list_head seed_list;
421 
422 	/* Count fs-devices opened. */
423 	int opened;
424 
425 	/* Set when we find or add a device that doesn't have the nonrot flag set. */
426 	bool rotating;
427 	/* Devices support TRIM/discard commands. */
428 	bool discardable;
429 	/* The filesystem is a seed filesystem. */
430 	bool seeding;
431 	/* The mount needs to use a randomly generated fsid. */
432 	bool temp_fsid;
433 	/* Enable/disable the filesystem stats tracking. */
434 	bool collect_fs_stats;
435 
436 	struct btrfs_fs_info *fs_info;
437 	/* sysfs kobjects */
438 	struct kobject fsid_kobj;
439 	struct kobject *devices_kobj;
440 	struct kobject *devinfo_kobj;
441 	struct completion kobj_unregister;
442 
443 	enum btrfs_chunk_allocation_policy chunk_alloc_policy;
444 
445 	/* Policy used to read the mirrored stripes. */
446 	enum btrfs_read_policy read_policy;
447 
448 #ifdef CONFIG_BTRFS_EXPERIMENTAL
449 	/*
450 	 * Minimum contiguous reads before switching to next device, the unit
451 	 * is one block/sectorsize.
452 	 */
453 	u32 rr_min_contig_read;
454 
455 	/* Device to be used for reading in case of RAID1. */
456 	u64 read_devid;
457 
458 	/* Checksum mode - offload it or do it synchronously. */
459 	enum btrfs_offload_csum_mode offload_csum_mode;
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 	u64 length;
477 	bool rst_search_commit_root;
478 	/* For the endio handler. */
479 	struct btrfs_io_context *bioc;
480 };
481 
482 struct btrfs_discard_stripe {
483 	struct btrfs_device *dev;
484 	u64 physical;
485 	u64 length;
486 };
487 
488 /*
489  * Context for IO subsmission for device stripe.
490  *
491  * - Track the unfinished mirrors for mirror based profiles
492  *   Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
493  *
494  * - Contain the logical -> physical mapping info
495  *   Used by submit_stripe_bio() for mapping logical bio
496  *   into physical device address.
497  *
498  * - Contain device replace info
499  *   Used by handle_ops_on_dev_replace() to copy logical bios
500  *   into the new device.
501  *
502  * - Contain RAID56 full stripe logical bytenrs
503  */
504 struct btrfs_io_context {
505 	refcount_t refs;
506 	struct btrfs_fs_info *fs_info;
507 	/* Taken from struct btrfs_chunk_map::type. */
508 	u64 map_type;
509 	struct bio *orig_bio;
510 	atomic_t error;
511 	u16 max_errors;
512 	bool use_rst;
513 
514 	u64 logical;
515 	u64 size;
516 	/* Raid stripe tree ordered entry. */
517 	struct list_head rst_ordered_entry;
518 
519 	/*
520 	 * The total number of stripes, including the extra duplicated
521 	 * stripe for replace.
522 	 */
523 	u16 num_stripes;
524 
525 	/*
526 	 * The mirror_num of this bioc.
527 	 *
528 	 * This is for reads which use 0 as mirror_num, thus we should return a
529 	 * valid mirror_num (>0) for the reader.
530 	 */
531 	u16 mirror_num;
532 
533 	/*
534 	 * The following two members are for dev-replace case only.
535 	 *
536 	 * @replace_nr_stripes:	Number of duplicated stripes which need to be
537 	 *			written to replace target.
538 	 *			Should be <= 2 (2 for DUP, otherwise <= 1).
539 	 * @replace_stripe_src:	The array indicates where the duplicated stripes
540 	 *			are from.
541 	 *
542 	 * The @replace_stripe_src[] array is mostly for RAID56 cases.
543 	 * As non-RAID56 stripes share the same contents of the mapped range,
544 	 * thus no need to bother where the duplicated ones are from.
545 	 *
546 	 * But for RAID56 case, all stripes contain different contents, thus
547 	 * we need a way to know the mapping.
548 	 *
549 	 * There is an example for the two members, using a RAID5 write:
550 	 *
551 	 *   num_stripes:	4 (3 + 1 duplicated write)
552 	 *   stripes[0]:	dev = devid 1, physical = X
553 	 *   stripes[1]:	dev = devid 2, physical = Y
554 	 *   stripes[2]:	dev = devid 3, physical = Z
555 	 *   stripes[3]:	dev = devid 0, physical = Y
556 	 *
557 	 * replace_nr_stripes = 1
558 	 * replace_stripe_src = 1	<- Means stripes[1] is involved in replace.
559 	 *				   The duplicated stripe index would be
560 	 *				   (@num_stripes - 1).
561 	 *
562 	 * Note, that we can still have cases replace_nr_stripes = 2 for DUP.
563 	 * In that case, all stripes share the same content, thus we don't
564 	 * need to bother @replace_stripe_src value at all.
565 	 */
566 	u16 replace_nr_stripes;
567 	s16 replace_stripe_src;
568 	/*
569 	 * Logical bytenr of the full stripe start, only for RAID56 cases.
570 	 *
571 	 * When this value is set to other than (u64)-1, the stripes[] should
572 	 * follow this pattern:
573 	 *
574 	 * (real_stripes = num_stripes - replace_nr_stripes)
575 	 * (data_stripes = (is_raid6) ? (real_stripes - 2) : (real_stripes - 1))
576 	 *
577 	 * stripes[0]:			The first data stripe
578 	 * stripes[1]:			The second data stripe
579 	 * ...
580 	 * stripes[data_stripes - 1]:	The last data stripe
581 	 * stripes[data_stripes]:	The P stripe
582 	 * stripes[data_stripes + 1]:	The Q stripe (only for RAID6).
583 	 */
584 	u64 full_stripe_logical;
585 	struct btrfs_io_stripe stripes[];
586 };
587 
588 struct btrfs_device_info {
589 	struct btrfs_device *dev;
590 	u64 dev_offset;
591 	u64 max_avail;
592 	u64 total_avail;
593 };
594 
595 struct btrfs_raid_attr {
596 	u8 sub_stripes;		/* sub_stripes info for map */
597 	u8 dev_stripes;		/* stripes per dev */
598 	u8 devs_max;		/* max devs to use */
599 	u8 devs_min;		/* min devs needed */
600 	u8 tolerated_failures;	/* max tolerated fail devs */
601 	u8 devs_increment;	/* ndevs has to be a multiple of this */
602 	u8 ncopies;		/* how many copies to data has */
603 	u8 nparity;		/* number of stripes worth of bytes to store
604 				 * parity information */
605 	u8 mindev_error;	/* error code if min devs requisite is unmet */
606 	const char raid_name[8]; /* name of the raid */
607 	u64 bg_flag;		/* block group flag of the raid */
608 };
609 
610 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
611 
612 struct btrfs_chunk_map {
613 	struct rb_node rb_node;
614 	/* For mount time dev extent verification. */
615 	int verified_stripes;
616 	refcount_t refs;
617 	u64 start;
618 	u64 chunk_len;
619 	u64 stripe_size;
620 	u64 type;
621 	int io_align;
622 	int io_width;
623 	int num_stripes;
624 	int sub_stripes;
625 	struct btrfs_io_stripe stripes[];
626 };
627 
628 #define btrfs_chunk_map_size(n) (sizeof(struct btrfs_chunk_map) + \
629 				 (sizeof(struct btrfs_io_stripe) * (n)))
630 
btrfs_free_chunk_map(struct btrfs_chunk_map * map)631 static inline void btrfs_free_chunk_map(struct btrfs_chunk_map *map)
632 {
633 	if (map && refcount_dec_and_test(&map->refs)) {
634 		ASSERT(RB_EMPTY_NODE(&map->rb_node));
635 		kfree(map);
636 	}
637 }
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 	bool missing;
657 };
658 
659 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
660 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
661 
662 #define BTRFS_DEV_LOOKUP_ARGS(name) \
663 	struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
664 
665 enum btrfs_map_op {
666 	BTRFS_MAP_READ,
667 	BTRFS_MAP_WRITE,
668 	BTRFS_MAP_GET_READ_MIRRORS,
669 };
670 
btrfs_op(struct bio * bio)671 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
672 {
673 	switch (bio_op(bio)) {
674 	case REQ_OP_WRITE:
675 	case REQ_OP_ZONE_APPEND:
676 		return BTRFS_MAP_WRITE;
677 	default:
678 		WARN_ON_ONCE(1);
679 		fallthrough;
680 	case REQ_OP_READ:
681 		return BTRFS_MAP_READ;
682 	}
683 }
684 
btrfs_chunk_item_size(int num_stripes)685 static inline unsigned long btrfs_chunk_item_size(int num_stripes)
686 {
687 	ASSERT(num_stripes);
688 	return sizeof(struct btrfs_chunk) +
689 		sizeof(struct btrfs_stripe) * (num_stripes - 1);
690 }
691 
692 /*
693  * Do the type safe conversion from stripe_nr to offset inside the chunk.
694  *
695  * @stripe_nr is u32, with left shift it can overflow u32 for chunks larger
696  * than 4G.  This does the proper type cast to avoid overflow.
697  */
btrfs_stripe_nr_to_offset(u32 stripe_nr)698 static inline u64 btrfs_stripe_nr_to_offset(u32 stripe_nr)
699 {
700 	return (u64)stripe_nr << BTRFS_STRIPE_LEN_SHIFT;
701 }
702 
703 void btrfs_get_bioc(struct btrfs_io_context *bioc);
704 void btrfs_put_bioc(struct btrfs_io_context *bioc);
705 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
706 		    u64 logical, u64 *length,
707 		    struct btrfs_io_context **bioc_ret,
708 		    struct btrfs_io_stripe *smap, int *mirror_num_ret);
709 int btrfs_map_repair_block(struct btrfs_fs_info *fs_info,
710 			   struct btrfs_io_stripe *smap, u64 logical,
711 			   u32 length, int mirror_num);
712 struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,
713 					       u64 logical, u64 *length_ret,
714 					       u32 *num_stripes);
715 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
716 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
717 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
718 					    u64 type);
719 void btrfs_mapping_tree_free(struct btrfs_fs_info *fs_info);
720 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
721 		       blk_mode_t flags, void *holder);
722 struct btrfs_device *btrfs_scan_one_device(const char *path, blk_mode_t flags,
723 					   bool mount_arg_dev);
724 int btrfs_forget_devices(dev_t devt);
725 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
726 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
727 void btrfs_assign_next_active_device(struct btrfs_device *device,
728 				     struct btrfs_device *this_dev);
729 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
730 						  u64 devid,
731 						  const char *devpath);
732 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
733 				 struct btrfs_dev_lookup_args *args,
734 				 const char *path);
735 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
736 					const u64 *devid, const u8 *uuid,
737 					const char *path);
738 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
739 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
740 		    struct btrfs_dev_lookup_args *args,
741 		    struct file **bdev_file);
742 void __exit btrfs_cleanup_fs_uuids(void);
743 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
744 int btrfs_grow_device(struct btrfs_trans_handle *trans,
745 		      struct btrfs_device *device, u64 new_size);
746 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
747 				       const struct btrfs_dev_lookup_args *args);
748 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
749 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
750 int btrfs_balance(struct btrfs_fs_info *fs_info,
751 		  struct btrfs_balance_control *bctl,
752 		  struct btrfs_ioctl_balance_args *bargs);
753 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
754 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
755 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
756 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
757 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
758 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
759 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
760 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
761 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
762 			struct btrfs_ioctl_get_dev_stats *stats);
763 int btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
764 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
765 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
766 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
767 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
768 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
769 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
770 				    u64 logical);
771 u64 btrfs_calc_stripe_length(const struct btrfs_chunk_map *map);
772 int btrfs_nr_parity_stripes(u64 type);
773 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
774 				     struct btrfs_block_group *bg);
775 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
776 
777 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
778 struct btrfs_chunk_map *btrfs_alloc_chunk_map(int num_stripes, gfp_t gfp);
779 int btrfs_add_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
780 #endif
781 
782 struct btrfs_chunk_map *btrfs_find_chunk_map(struct btrfs_fs_info *fs_info,
783 					     u64 logical, u64 length);
784 struct btrfs_chunk_map *btrfs_find_chunk_map_nolock(struct btrfs_fs_info *fs_info,
785 						    u64 logical, u64 length);
786 struct btrfs_chunk_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
787 					    u64 logical, u64 length);
788 void btrfs_remove_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map);
789 void btrfs_release_disk_super(struct btrfs_super_block *super);
790 
btrfs_dev_stat_inc(struct btrfs_device * dev,int index)791 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
792 				      int index)
793 {
794 	atomic_inc(dev->dev_stat_values + index);
795 	/*
796 	 * This memory barrier orders stores updating statistics before stores
797 	 * updating dev_stats_ccnt.
798 	 *
799 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
800 	 */
801 	smp_mb__before_atomic();
802 	atomic_inc(&dev->dev_stats_ccnt);
803 }
804 
btrfs_dev_stat_read(struct btrfs_device * dev,int index)805 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
806 				      int index)
807 {
808 	return atomic_read(dev->dev_stat_values + index);
809 }
810 
btrfs_dev_stat_read_and_reset(struct btrfs_device * dev,int index)811 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
812 						int index)
813 {
814 	int ret;
815 
816 	ret = atomic_xchg(dev->dev_stat_values + index, 0);
817 	/*
818 	 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
819 	 * - RMW operations that have a return value are fully ordered;
820 	 *
821 	 * This implicit memory barriers is paired with the smp_rmb in
822 	 * btrfs_run_dev_stats
823 	 */
824 	atomic_inc(&dev->dev_stats_ccnt);
825 	return ret;
826 }
827 
btrfs_dev_stat_set(struct btrfs_device * dev,int index,unsigned long val)828 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
829 				      int index, unsigned long val)
830 {
831 	atomic_set(dev->dev_stat_values + index, val);
832 	/*
833 	 * This memory barrier orders stores updating statistics before stores
834 	 * updating dev_stats_ccnt.
835 	 *
836 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
837 	 */
838 	smp_mb__before_atomic();
839 	atomic_inc(&dev->dev_stats_ccnt);
840 }
841 
btrfs_dev_name(const struct btrfs_device * device)842 static inline const char *btrfs_dev_name(const struct btrfs_device *device)
843 {
844 	if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
845 		return "<missing disk>";
846 	else
847 		return rcu_str_deref(device->name);
848 }
849 
850 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
851 
852 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
853 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
854 					struct btrfs_device *failing_dev);
855 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, struct btrfs_device *device);
856 
857 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
858 int btrfs_bg_type_to_factor(u64 flags);
859 const char *btrfs_bg_type_to_raid_name(u64 flags);
860 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
861 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);
862 
863 bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr);
864 const u8 *btrfs_sb_fsid_ptr(const struct btrfs_super_block *sb);
865 
866 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
867 struct btrfs_io_context *alloc_btrfs_io_context(struct btrfs_fs_info *fs_info,
868 						u64 logical, u16 total_stripes);
869 #endif
870 
871 #endif
872