xref: /openbmc/linux/fs/btrfs/volumes.h (revision 709f34b4)
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/sort.h>
10 #include <linux/btrfs.h>
11 #include "async-thread.h"
12 #include "messages.h"
13 #include "tree-checker.h"
14 #include "rcu-string.h"
15 
16 #define BTRFS_MAX_DATA_CHUNK_SIZE	(10ULL * SZ_1G)
17 
18 extern struct mutex uuid_mutex;
19 
20 #define BTRFS_STRIPE_LEN		SZ_64K
21 #define BTRFS_STRIPE_LEN_SHIFT		(16)
22 #define BTRFS_STRIPE_LEN_MASK		(BTRFS_STRIPE_LEN - 1)
23 
24 static_assert(const_ilog2(BTRFS_STRIPE_LEN) == BTRFS_STRIPE_LEN_SHIFT);
25 
26 /* Used by sanity check for btrfs_raid_types. */
27 #define const_ffs(n) (__builtin_ctzll(n) + 1)
28 
29 /*
30  * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires
31  * RAID0 always to be the lowest profile bit.
32  * Although it's part of on-disk format and should never change, do extra
33  * compile-time sanity checks.
34  */
35 static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) <
36 	      const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0));
37 static_assert(const_ilog2(BTRFS_BLOCK_GROUP_RAID0) >
38 	      ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK));
39 
40 /* ilog2() can handle both constants and variables */
41 #define BTRFS_BG_FLAG_TO_INDEX(profile)					\
42 	ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1))
43 
44 enum btrfs_raid_types {
45 	/* SINGLE is the special one as it doesn't have on-disk bit. */
46 	BTRFS_RAID_SINGLE  = 0,
47 
48 	BTRFS_RAID_RAID0   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0),
49 	BTRFS_RAID_RAID1   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1),
50 	BTRFS_RAID_DUP	   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP),
51 	BTRFS_RAID_RAID10  = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10),
52 	BTRFS_RAID_RAID5   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5),
53 	BTRFS_RAID_RAID6   = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6),
54 	BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3),
55 	BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4),
56 
57 	BTRFS_NR_RAID_TYPES
58 };
59 
60 /*
61  * Use sequence counter to get consistent device stat data on
62  * 32-bit processors.
63  */
64 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
65 #include <linux/seqlock.h>
66 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
67 #define btrfs_device_data_ordered_init(device)	\
68 	seqcount_init(&device->data_seqcount)
69 #else
70 #define btrfs_device_data_ordered_init(device) do { } while (0)
71 #endif
72 
73 #define BTRFS_DEV_STATE_WRITEABLE	(0)
74 #define BTRFS_DEV_STATE_IN_FS_METADATA	(1)
75 #define BTRFS_DEV_STATE_MISSING		(2)
76 #define BTRFS_DEV_STATE_REPLACE_TGT	(3)
77 #define BTRFS_DEV_STATE_FLUSH_SENT	(4)
78 #define BTRFS_DEV_STATE_NO_READA	(5)
79 
80 struct btrfs_zoned_device_info;
81 
82 struct btrfs_device {
83 	struct list_head dev_list; /* device_list_mutex */
84 	struct list_head dev_alloc_list; /* chunk mutex */
85 	struct list_head post_commit_list; /* chunk mutex */
86 	struct btrfs_fs_devices *fs_devices;
87 	struct btrfs_fs_info *fs_info;
88 
89 	struct rcu_string __rcu *name;
90 
91 	u64 generation;
92 
93 	struct block_device *bdev;
94 
95 	struct btrfs_zoned_device_info *zone_info;
96 
97 	/* the mode sent to blkdev_get */
98 	fmode_t mode;
99 
100 	/*
101 	 * Device's major-minor number. Must be set even if the device is not
102 	 * opened (bdev == NULL), unless the device is missing.
103 	 */
104 	dev_t devt;
105 	unsigned long dev_state;
106 	blk_status_t last_flush_error;
107 
108 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
109 	seqcount_t data_seqcount;
110 #endif
111 
112 	/* the internal btrfs device id */
113 	u64 devid;
114 
115 	/* size of the device in memory */
116 	u64 total_bytes;
117 
118 	/* size of the device on disk */
119 	u64 disk_total_bytes;
120 
121 	/* bytes used */
122 	u64 bytes_used;
123 
124 	/* optimal io alignment for this device */
125 	u32 io_align;
126 
127 	/* optimal io width for this device */
128 	u32 io_width;
129 	/* type and info about this device */
130 	u64 type;
131 
132 	/* minimal io size for this device */
133 	u32 sector_size;
134 
135 	/* physical drive uuid (or lvm uuid) */
136 	u8 uuid[BTRFS_UUID_SIZE];
137 
138 	/*
139 	 * size of the device on the current transaction
140 	 *
141 	 * This variant is update when committing the transaction,
142 	 * and protected by chunk mutex
143 	 */
144 	u64 commit_total_bytes;
145 
146 	/* bytes used on the current transaction */
147 	u64 commit_bytes_used;
148 
149 	/* Bio used for flushing device barriers */
150 	struct bio flush_bio;
151 	struct completion flush_wait;
152 
153 	/* per-device scrub information */
154 	struct scrub_ctx *scrub_ctx;
155 
156 	/* disk I/O failure stats. For detailed description refer to
157 	 * enum btrfs_dev_stat_values in ioctl.h */
158 	int dev_stats_valid;
159 
160 	/* Counter to record the change of device stats */
161 	atomic_t dev_stats_ccnt;
162 	atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
163 
164 	struct extent_io_tree alloc_state;
165 
166 	struct completion kobj_unregister;
167 	/* For sysfs/FSID/devinfo/devid/ */
168 	struct kobject devid_kobj;
169 
170 	/* Bandwidth limit for scrub, in bytes */
171 	u64 scrub_speed_max;
172 };
173 
174 /*
175  * Block group or device which contains an active swapfile. Used for preventing
176  * unsafe operations while a swapfile is active.
177  *
178  * These are sorted on (ptr, inode) (note that a block group or device can
179  * contain more than one swapfile). We compare the pointer values because we
180  * don't actually care what the object is, we just need a quick check whether
181  * the object exists in the rbtree.
182  */
183 struct btrfs_swapfile_pin {
184 	struct rb_node node;
185 	void *ptr;
186 	struct inode *inode;
187 	/*
188 	 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr
189 	 * points to a struct btrfs_device.
190 	 */
191 	bool is_block_group;
192 	/*
193 	 * Only used when 'is_block_group' is true and it is the number of
194 	 * extents used by a swapfile for this block group ('ptr' field).
195 	 */
196 	int bg_extent_count;
197 };
198 
199 /*
200  * If we read those variants at the context of their own lock, we needn't
201  * use the following helpers, reading them directly is safe.
202  */
203 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
204 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
205 static inline u64							\
206 btrfs_device_get_##name(const struct btrfs_device *dev)			\
207 {									\
208 	u64 size;							\
209 	unsigned int seq;						\
210 									\
211 	do {								\
212 		seq = read_seqcount_begin(&dev->data_seqcount);		\
213 		size = dev->name;					\
214 	} while (read_seqcount_retry(&dev->data_seqcount, seq));	\
215 	return size;							\
216 }									\
217 									\
218 static inline void							\
219 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
220 {									\
221 	preempt_disable();						\
222 	write_seqcount_begin(&dev->data_seqcount);			\
223 	dev->name = size;						\
224 	write_seqcount_end(&dev->data_seqcount);			\
225 	preempt_enable();						\
226 }
227 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
228 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
229 static inline u64							\
230 btrfs_device_get_##name(const struct btrfs_device *dev)			\
231 {									\
232 	u64 size;							\
233 									\
234 	preempt_disable();						\
235 	size = dev->name;						\
236 	preempt_enable();						\
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 	dev->name = size;						\
245 	preempt_enable();						\
246 }
247 #else
248 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
249 static inline u64							\
250 btrfs_device_get_##name(const struct btrfs_device *dev)			\
251 {									\
252 	return dev->name;						\
253 }									\
254 									\
255 static inline void							\
256 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
257 {									\
258 	dev->name = size;						\
259 }
260 #endif
261 
262 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
263 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
264 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
265 
266 enum btrfs_chunk_allocation_policy {
267 	BTRFS_CHUNK_ALLOC_REGULAR,
268 	BTRFS_CHUNK_ALLOC_ZONED,
269 };
270 
271 /*
272  * Read policies for mirrored block group profiles, read picks the stripe based
273  * on these policies.
274  */
275 enum btrfs_read_policy {
276 	/* Use process PID to choose the stripe */
277 	BTRFS_READ_POLICY_PID,
278 	BTRFS_NR_READ_POLICY,
279 };
280 
281 struct btrfs_fs_devices {
282 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
283 	u8 metadata_uuid[BTRFS_FSID_SIZE];
284 	bool fsid_change;
285 	struct list_head fs_list;
286 
287 	/*
288 	 * Number of devices under this fsid including missing and
289 	 * replace-target device and excludes seed devices.
290 	 */
291 	u64 num_devices;
292 
293 	/*
294 	 * The number of devices that successfully opened, including
295 	 * replace-target, excludes seed devices.
296 	 */
297 	u64 open_devices;
298 
299 	/* The number of devices that are under the chunk allocation list. */
300 	u64 rw_devices;
301 
302 	/* Count of missing devices under this fsid excluding seed device. */
303 	u64 missing_devices;
304 	u64 total_rw_bytes;
305 
306 	/*
307 	 * Count of devices from btrfs_super_block::num_devices for this fsid,
308 	 * which includes the seed device, excludes the transient replace-target
309 	 * device.
310 	 */
311 	u64 total_devices;
312 
313 	/* Highest generation number of seen devices */
314 	u64 latest_generation;
315 
316 	/*
317 	 * The mount device or a device with highest generation after removal
318 	 * or replace.
319 	 */
320 	struct btrfs_device *latest_dev;
321 
322 	/* all of the devices in the FS, protected by a mutex
323 	 * so we can safely walk it to write out the supers without
324 	 * worrying about add/remove by the multi-device code.
325 	 * Scrubbing super can kick off supers writing by holding
326 	 * this mutex lock.
327 	 */
328 	struct mutex device_list_mutex;
329 
330 	/* List of all devices, protected by device_list_mutex */
331 	struct list_head devices;
332 
333 	/*
334 	 * Devices which can satisfy space allocation. Protected by
335 	 * chunk_mutex
336 	 */
337 	struct list_head alloc_list;
338 
339 	struct list_head seed_list;
340 	bool seeding;
341 
342 	int opened;
343 
344 	/* set when we find or add a device that doesn't have the
345 	 * nonrot flag set
346 	 */
347 	bool rotating;
348 	/* Devices support TRIM/discard commands */
349 	bool discardable;
350 
351 	struct btrfs_fs_info *fs_info;
352 	/* sysfs kobjects */
353 	struct kobject fsid_kobj;
354 	struct kobject *devices_kobj;
355 	struct kobject *devinfo_kobj;
356 	struct completion kobj_unregister;
357 
358 	enum btrfs_chunk_allocation_policy chunk_alloc_policy;
359 
360 	/* Policy used to read the mirrored stripes */
361 	enum btrfs_read_policy read_policy;
362 };
363 
364 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
365 			- sizeof(struct btrfs_chunk))		\
366 			/ sizeof(struct btrfs_stripe) + 1)
367 
368 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
369 				- 2 * sizeof(struct btrfs_disk_key)	\
370 				- 2 * sizeof(struct btrfs_chunk))	\
371 				/ sizeof(struct btrfs_stripe) + 1)
372 
373 struct btrfs_io_stripe {
374 	struct btrfs_device *dev;
375 	union {
376 		/* Block mapping */
377 		u64 physical;
378 		/* For the endio handler */
379 		struct btrfs_io_context *bioc;
380 	};
381 };
382 
383 struct btrfs_discard_stripe {
384 	struct btrfs_device *dev;
385 	u64 physical;
386 	u64 length;
387 };
388 
389 /*
390  * Context for IO subsmission for device stripe.
391  *
392  * - Track the unfinished mirrors for mirror based profiles
393  *   Mirror based profiles are SINGLE/DUP/RAID1/RAID10.
394  *
395  * - Contain the logical -> physical mapping info
396  *   Used by submit_stripe_bio() for mapping logical bio
397  *   into physical device address.
398  *
399  * - Contain device replace info
400  *   Used by handle_ops_on_dev_replace() to copy logical bios
401  *   into the new device.
402  *
403  * - Contain RAID56 full stripe logical bytenrs
404  */
405 struct btrfs_io_context {
406 	refcount_t refs;
407 	struct btrfs_fs_info *fs_info;
408 	u64 map_type; /* get from map_lookup->type */
409 	struct bio *orig_bio;
410 	atomic_t error;
411 	u16 max_errors;
412 
413 	/*
414 	 * The total number of stripes, including the extra duplicated
415 	 * stripe for replace.
416 	 */
417 	u16 num_stripes;
418 
419 	/*
420 	 * The mirror_num of this bioc.
421 	 *
422 	 * This is for reads which use 0 as mirror_num, thus we should return a
423 	 * valid mirror_num (>0) for the reader.
424 	 */
425 	u16 mirror_num;
426 
427 	/*
428 	 * The following two members are for dev-replace case only.
429 	 *
430 	 * @replace_nr_stripes:	Number of duplicated stripes which need to be
431 	 *			written to replace target.
432 	 *			Should be <= 2 (2 for DUP, otherwise <= 1).
433 	 * @replace_stripe_src:	The array indicates where the duplicated stripes
434 	 *			are from.
435 	 *
436 	 * The @replace_stripe_src[] array is mostly for RAID56 cases.
437 	 * As non-RAID56 stripes share the same contents of the mapped range,
438 	 * thus no need to bother where the duplicated ones are from.
439 	 *
440 	 * But for RAID56 case, all stripes contain different contents, thus
441 	 * we need a way to know the mapping.
442 	 *
443 	 * There is an example for the two members, using a RAID5 write:
444 	 *
445 	 *   num_stripes:	4 (3 + 1 duplicated write)
446 	 *   stripes[0]:	dev = devid 1, physical = X
447 	 *   stripes[1]:	dev = devid 2, physical = Y
448 	 *   stripes[2]:	dev = devid 3, physical = Z
449 	 *   stripes[3]:	dev = devid 0, physical = Y
450 	 *
451 	 * replace_nr_stripes = 1
452 	 * replace_stripe_src = 1	<- Means stripes[1] is involved in replace.
453 	 *				   The duplicated stripe index would be
454 	 *				   (@num_stripes - 1).
455 	 *
456 	 * Note, that we can still have cases replace_nr_stripes = 2 for DUP.
457 	 * In that case, all stripes share the same content, thus we don't
458 	 * need to bother @replace_stripe_src value at all.
459 	 */
460 	u16 replace_nr_stripes;
461 	s16 replace_stripe_src;
462 	/*
463 	 * Logical bytenr of the full stripe start, only for RAID56 cases.
464 	 *
465 	 * When this value is set to other than (u64)-1, the stripes[] should
466 	 * follow this pattern:
467 	 *
468 	 * (real_stripes = num_stripes - replace_nr_stripes)
469 	 * (data_stripes = (is_raid6) ? (real_stripes - 2) : (real_stripes - 1))
470 	 *
471 	 * stripes[0]:			The first data stripe
472 	 * stripes[1]:			The second data stripe
473 	 * ...
474 	 * stripes[data_stripes - 1]:	The last data stripe
475 	 * stripes[data_stripes]:	The P stripe
476 	 * stripes[data_stripes + 1]:	The Q stripe (only for RAID6).
477 	 */
478 	u64 full_stripe_logical;
479 	struct btrfs_io_stripe stripes[];
480 };
481 
482 struct btrfs_device_info {
483 	struct btrfs_device *dev;
484 	u64 dev_offset;
485 	u64 max_avail;
486 	u64 total_avail;
487 };
488 
489 struct btrfs_raid_attr {
490 	u8 sub_stripes;		/* sub_stripes info for map */
491 	u8 dev_stripes;		/* stripes per dev */
492 	u8 devs_max;		/* max devs to use */
493 	u8 devs_min;		/* min devs needed */
494 	u8 tolerated_failures;	/* max tolerated fail devs */
495 	u8 devs_increment;	/* ndevs has to be a multiple of this */
496 	u8 ncopies;		/* how many copies to data has */
497 	u8 nparity;		/* number of stripes worth of bytes to store
498 				 * parity information */
499 	u8 mindev_error;	/* error code if min devs requisite is unmet */
500 	const char raid_name[8]; /* name of the raid */
501 	u64 bg_flag;		/* block group flag of the raid */
502 };
503 
504 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
505 
506 struct map_lookup {
507 	u64 type;
508 	int io_align;
509 	int io_width;
510 	int num_stripes;
511 	int sub_stripes;
512 	int verified_stripes; /* For mount time dev extent verification */
513 	struct btrfs_io_stripe stripes[];
514 };
515 
516 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
517 			    (sizeof(struct btrfs_io_stripe) * (n)))
518 
519 struct btrfs_balance_args;
520 struct btrfs_balance_progress;
521 struct btrfs_balance_control {
522 	struct btrfs_balance_args data;
523 	struct btrfs_balance_args meta;
524 	struct btrfs_balance_args sys;
525 
526 	u64 flags;
527 
528 	struct btrfs_balance_progress stat;
529 };
530 
531 /*
532  * Search for a given device by the set parameters
533  */
534 struct btrfs_dev_lookup_args {
535 	u64 devid;
536 	u8 *uuid;
537 	u8 *fsid;
538 	bool missing;
539 };
540 
541 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */
542 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 }
543 
544 #define BTRFS_DEV_LOOKUP_ARGS(name) \
545 	struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT
546 
547 enum btrfs_map_op {
548 	BTRFS_MAP_READ,
549 	BTRFS_MAP_WRITE,
550 	BTRFS_MAP_DISCARD,
551 	BTRFS_MAP_GET_READ_MIRRORS,
552 };
553 
554 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
555 {
556 	switch (bio_op(bio)) {
557 	case REQ_OP_DISCARD:
558 		return BTRFS_MAP_DISCARD;
559 	case REQ_OP_WRITE:
560 	case REQ_OP_ZONE_APPEND:
561 		return BTRFS_MAP_WRITE;
562 	default:
563 		WARN_ON_ONCE(1);
564 		fallthrough;
565 	case REQ_OP_READ:
566 		return BTRFS_MAP_READ;
567 	}
568 }
569 
570 static inline unsigned long btrfs_chunk_item_size(int num_stripes)
571 {
572 	ASSERT(num_stripes);
573 	return sizeof(struct btrfs_chunk) +
574 		sizeof(struct btrfs_stripe) * (num_stripes - 1);
575 }
576 
577 void btrfs_get_bioc(struct btrfs_io_context *bioc);
578 void btrfs_put_bioc(struct btrfs_io_context *bioc);
579 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
580 		    u64 logical, u64 *length,
581 		    struct btrfs_io_context **bioc_ret, int mirror_num);
582 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
583 		     u64 logical, u64 *length,
584 		     struct btrfs_io_context **bioc_ret);
585 int __btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
586 		      u64 logical, u64 *length,
587 		      struct btrfs_io_context **bioc_ret,
588 		      struct btrfs_io_stripe *smap, int *mirror_num_ret,
589 		      int need_raid_map);
590 int btrfs_map_repair_block(struct btrfs_fs_info *fs_info,
591 			   struct btrfs_io_stripe *smap, u64 logical,
592 			   u32 length, int mirror_num);
593 struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,
594 					       u64 logical, u64 *length_ret,
595 					       u32 *num_stripes);
596 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
597 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
598 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans,
599 					    u64 type);
600 void btrfs_mapping_tree_free(struct extent_map_tree *tree);
601 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
602 		       fmode_t flags, void *holder);
603 struct btrfs_device *btrfs_scan_one_device(const char *path,
604 					   fmode_t flags, void *holder);
605 int btrfs_forget_devices(dev_t devt);
606 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
607 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices);
608 void btrfs_assign_next_active_device(struct btrfs_device *device,
609 				     struct btrfs_device *this_dev);
610 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
611 						  u64 devid,
612 						  const char *devpath);
613 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info,
614 				 struct btrfs_dev_lookup_args *args,
615 				 const char *path);
616 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
617 					const u64 *devid, const u8 *uuid,
618 					const char *path);
619 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args);
620 void btrfs_free_device(struct btrfs_device *device);
621 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
622 		    struct btrfs_dev_lookup_args *args,
623 		    struct block_device **bdev, fmode_t *mode);
624 void __exit btrfs_cleanup_fs_uuids(void);
625 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
626 int btrfs_grow_device(struct btrfs_trans_handle *trans,
627 		      struct btrfs_device *device, u64 new_size);
628 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices,
629 				       const struct btrfs_dev_lookup_args *args);
630 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
631 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
632 int btrfs_balance(struct btrfs_fs_info *fs_info,
633 		  struct btrfs_balance_control *bctl,
634 		  struct btrfs_ioctl_balance_args *bargs);
635 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
636 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
637 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
638 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
639 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset);
640 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
641 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
642 int btrfs_uuid_scan_kthread(void *data);
643 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset);
644 int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes,
645 			 u64 *start, u64 *max_avail);
646 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
647 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
648 			struct btrfs_ioctl_get_dev_stats *stats);
649 int btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
650 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
651 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
652 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
653 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
654 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
655 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
656 			   u64 logical, u64 len);
657 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
658 				    u64 logical);
659 u64 btrfs_calc_stripe_length(const struct extent_map *em);
660 int btrfs_nr_parity_stripes(u64 type);
661 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
662 				     struct btrfs_block_group *bg);
663 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
664 struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
665 				       u64 logical, u64 length);
666 void btrfs_release_disk_super(struct btrfs_super_block *super);
667 
668 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
669 				      int index)
670 {
671 	atomic_inc(dev->dev_stat_values + index);
672 	/*
673 	 * This memory barrier orders stores updating statistics before stores
674 	 * updating dev_stats_ccnt.
675 	 *
676 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
677 	 */
678 	smp_mb__before_atomic();
679 	atomic_inc(&dev->dev_stats_ccnt);
680 }
681 
682 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
683 				      int index)
684 {
685 	return atomic_read(dev->dev_stat_values + index);
686 }
687 
688 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
689 						int index)
690 {
691 	int ret;
692 
693 	ret = atomic_xchg(dev->dev_stat_values + index, 0);
694 	/*
695 	 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
696 	 * - RMW operations that have a return value are fully ordered;
697 	 *
698 	 * This implicit memory barriers is paired with the smp_rmb in
699 	 * btrfs_run_dev_stats
700 	 */
701 	atomic_inc(&dev->dev_stats_ccnt);
702 	return ret;
703 }
704 
705 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
706 				      int index, unsigned long val)
707 {
708 	atomic_set(dev->dev_stat_values + index, val);
709 	/*
710 	 * This memory barrier orders stores updating statistics before stores
711 	 * updating dev_stats_ccnt.
712 	 *
713 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
714 	 */
715 	smp_mb__before_atomic();
716 	atomic_inc(&dev->dev_stats_ccnt);
717 }
718 
719 static inline const char *btrfs_dev_name(const struct btrfs_device *device)
720 {
721 	if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
722 		return "<missing disk>";
723 	else
724 		return rcu_str_deref(device->name);
725 }
726 
727 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
728 
729 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void);
730 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
731 					struct btrfs_device *failing_dev);
732 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info,
733 			       struct block_device *bdev,
734 			       const char *device_path);
735 
736 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags);
737 int btrfs_bg_type_to_factor(u64 flags);
738 const char *btrfs_bg_type_to_raid_name(u64 flags);
739 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
740 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical);
741 
742 bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr);
743 
744 #endif
745