xref: /openbmc/linux/fs/btrfs/volumes.h (revision 9fb29c73)
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/bio.h>
10 #include <linux/sort.h>
11 #include <linux/btrfs.h>
12 #include "async-thread.h"
13 
14 #define BTRFS_MAX_DATA_CHUNK_SIZE	(10ULL * SZ_1G)
15 
16 extern struct mutex uuid_mutex;
17 
18 #define BTRFS_STRIPE_LEN	SZ_64K
19 
20 struct buffer_head;
21 struct btrfs_pending_bios {
22 	struct bio *head;
23 	struct bio *tail;
24 };
25 
26 /*
27  * Use sequence counter to get consistent device stat data on
28  * 32-bit processors.
29  */
30 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
31 #include <linux/seqlock.h>
32 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
33 #define btrfs_device_data_ordered_init(device)	\
34 	seqcount_init(&device->data_seqcount)
35 #else
36 #define btrfs_device_data_ordered_init(device) do { } while (0)
37 #endif
38 
39 #define BTRFS_DEV_STATE_WRITEABLE	(0)
40 #define BTRFS_DEV_STATE_IN_FS_METADATA	(1)
41 #define BTRFS_DEV_STATE_MISSING		(2)
42 #define BTRFS_DEV_STATE_REPLACE_TGT	(3)
43 #define BTRFS_DEV_STATE_FLUSH_SENT	(4)
44 
45 struct btrfs_device {
46 	struct list_head dev_list;
47 	struct list_head dev_alloc_list;
48 	struct btrfs_fs_devices *fs_devices;
49 	struct btrfs_fs_info *fs_info;
50 
51 	struct rcu_string *name;
52 
53 	u64 generation;
54 
55 	spinlock_t io_lock ____cacheline_aligned;
56 	int running_pending;
57 	/* regular prio bios */
58 	struct btrfs_pending_bios pending_bios;
59 	/* sync bios */
60 	struct btrfs_pending_bios pending_sync_bios;
61 
62 	struct block_device *bdev;
63 
64 	/* the mode sent to blkdev_get */
65 	fmode_t mode;
66 
67 	unsigned long dev_state;
68 	blk_status_t last_flush_error;
69 	int flush_bio_sent;
70 
71 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
72 	seqcount_t data_seqcount;
73 #endif
74 
75 	/* the internal btrfs device id */
76 	u64 devid;
77 
78 	/* size of the device in memory */
79 	u64 total_bytes;
80 
81 	/* size of the device on disk */
82 	u64 disk_total_bytes;
83 
84 	/* bytes used */
85 	u64 bytes_used;
86 
87 	/* optimal io alignment for this device */
88 	u32 io_align;
89 
90 	/* optimal io width for this device */
91 	u32 io_width;
92 	/* type and info about this device */
93 	u64 type;
94 
95 	/* minimal io size for this device */
96 	u32 sector_size;
97 
98 	/* physical drive uuid (or lvm uuid) */
99 	u8 uuid[BTRFS_UUID_SIZE];
100 
101 	/*
102 	 * size of the device on the current transaction
103 	 *
104 	 * This variant is update when committing the transaction,
105 	 * and protected by device_list_mutex
106 	 */
107 	u64 commit_total_bytes;
108 
109 	/* bytes used on the current transaction */
110 	u64 commit_bytes_used;
111 	/*
112 	 * used to manage the device which is resized
113 	 *
114 	 * It is protected by chunk_lock.
115 	 */
116 	struct list_head resized_list;
117 
118 	/* for sending down flush barriers */
119 	struct bio *flush_bio;
120 	struct completion flush_wait;
121 
122 	/* per-device scrub information */
123 	struct scrub_ctx *scrub_ctx;
124 
125 	struct btrfs_work work;
126 	struct rcu_head rcu;
127 
128 	/* readahead state */
129 	atomic_t reada_in_flight;
130 	u64 reada_next;
131 	struct reada_zone *reada_curr_zone;
132 	struct radix_tree_root reada_zones;
133 	struct radix_tree_root reada_extents;
134 
135 	/* disk I/O failure stats. For detailed description refer to
136 	 * enum btrfs_dev_stat_values in ioctl.h */
137 	int dev_stats_valid;
138 
139 	/* Counter to record the change of device stats */
140 	atomic_t dev_stats_ccnt;
141 	atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
142 };
143 
144 /*
145  * If we read those variants at the context of their own lock, we needn't
146  * use the following helpers, reading them directly is safe.
147  */
148 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
149 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
150 static inline u64							\
151 btrfs_device_get_##name(const struct btrfs_device *dev)			\
152 {									\
153 	u64 size;							\
154 	unsigned int seq;						\
155 									\
156 	do {								\
157 		seq = read_seqcount_begin(&dev->data_seqcount);		\
158 		size = dev->name;					\
159 	} while (read_seqcount_retry(&dev->data_seqcount, seq));	\
160 	return size;							\
161 }									\
162 									\
163 static inline void							\
164 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
165 {									\
166 	preempt_disable();						\
167 	write_seqcount_begin(&dev->data_seqcount);			\
168 	dev->name = size;						\
169 	write_seqcount_end(&dev->data_seqcount);			\
170 	preempt_enable();						\
171 }
172 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
173 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
174 static inline u64							\
175 btrfs_device_get_##name(const struct btrfs_device *dev)			\
176 {									\
177 	u64 size;							\
178 									\
179 	preempt_disable();						\
180 	size = dev->name;						\
181 	preempt_enable();						\
182 	return size;							\
183 }									\
184 									\
185 static inline void							\
186 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
187 {									\
188 	preempt_disable();						\
189 	dev->name = size;						\
190 	preempt_enable();						\
191 }
192 #else
193 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
194 static inline u64							\
195 btrfs_device_get_##name(const struct btrfs_device *dev)			\
196 {									\
197 	return dev->name;						\
198 }									\
199 									\
200 static inline void							\
201 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
202 {									\
203 	dev->name = size;						\
204 }
205 #endif
206 
207 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
208 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
209 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
210 
211 struct btrfs_fs_devices {
212 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
213 	u8 metadata_uuid[BTRFS_FSID_SIZE];
214 	bool fsid_change;
215 	struct list_head fs_list;
216 
217 	u64 num_devices;
218 	u64 open_devices;
219 	u64 rw_devices;
220 	u64 missing_devices;
221 	u64 total_rw_bytes;
222 	u64 total_devices;
223 
224 	/* Highest generation number of seen devices */
225 	u64 latest_generation;
226 
227 	struct block_device *latest_bdev;
228 
229 	/* all of the devices in the FS, protected by a mutex
230 	 * so we can safely walk it to write out the supers without
231 	 * worrying about add/remove by the multi-device code.
232 	 * Scrubbing super can kick off supers writing by holding
233 	 * this mutex lock.
234 	 */
235 	struct mutex device_list_mutex;
236 	struct list_head devices;
237 
238 	struct list_head resized_devices;
239 	/* devices not currently being allocated */
240 	struct list_head alloc_list;
241 
242 	struct btrfs_fs_devices *seed;
243 	int seeding;
244 
245 	int opened;
246 
247 	/* set when we find or add a device that doesn't have the
248 	 * nonrot flag set
249 	 */
250 	int rotating;
251 
252 	struct btrfs_fs_info *fs_info;
253 	/* sysfs kobjects */
254 	struct kobject fsid_kobj;
255 	struct kobject *device_dir_kobj;
256 	struct completion kobj_unregister;
257 };
258 
259 #define BTRFS_BIO_INLINE_CSUM_SIZE	64
260 
261 /*
262  * we need the mirror number and stripe index to be passed around
263  * the call chain while we are processing end_io (especially errors).
264  * Really, what we need is a btrfs_bio structure that has this info
265  * and is properly sized with its stripe array, but we're not there
266  * quite yet.  We have our own btrfs bioset, and all of the bios
267  * we allocate are actually btrfs_io_bios.  We'll cram as much of
268  * struct btrfs_bio as we can into this over time.
269  */
270 struct btrfs_io_bio {
271 	unsigned int mirror_num;
272 	unsigned int stripe_index;
273 	u64 logical;
274 	u8 *csum;
275 	u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
276 	struct bvec_iter iter;
277 	/*
278 	 * This member must come last, bio_alloc_bioset will allocate enough
279 	 * bytes for entire btrfs_io_bio but relies on bio being last.
280 	 */
281 	struct bio bio;
282 };
283 
284 static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
285 {
286 	return container_of(bio, struct btrfs_io_bio, bio);
287 }
288 
289 static inline void btrfs_io_bio_free_csum(struct btrfs_io_bio *io_bio)
290 {
291 	if (io_bio->csum != io_bio->csum_inline) {
292 		kfree(io_bio->csum);
293 		io_bio->csum = NULL;
294 	}
295 }
296 
297 struct btrfs_bio_stripe {
298 	struct btrfs_device *dev;
299 	u64 physical;
300 	u64 length; /* only used for discard mappings */
301 };
302 
303 struct btrfs_bio {
304 	refcount_t refs;
305 	atomic_t stripes_pending;
306 	struct btrfs_fs_info *fs_info;
307 	u64 map_type; /* get from map_lookup->type */
308 	bio_end_io_t *end_io;
309 	struct bio *orig_bio;
310 	unsigned long flags;
311 	void *private;
312 	atomic_t error;
313 	int max_errors;
314 	int num_stripes;
315 	int mirror_num;
316 	int num_tgtdevs;
317 	int *tgtdev_map;
318 	/*
319 	 * logical block numbers for the start of each stripe
320 	 * The last one or two are p/q.  These are sorted,
321 	 * so raid_map[0] is the start of our full stripe
322 	 */
323 	u64 *raid_map;
324 	struct btrfs_bio_stripe stripes[];
325 };
326 
327 struct btrfs_device_info {
328 	struct btrfs_device *dev;
329 	u64 dev_offset;
330 	u64 max_avail;
331 	u64 total_avail;
332 };
333 
334 struct btrfs_raid_attr {
335 	int sub_stripes;	/* sub_stripes info for map */
336 	int dev_stripes;	/* stripes per dev */
337 	int devs_max;		/* max devs to use */
338 	int devs_min;		/* min devs needed */
339 	int tolerated_failures; /* max tolerated fail devs */
340 	int devs_increment;	/* ndevs has to be a multiple of this */
341 	int ncopies;		/* how many copies to data has */
342 	int nparity;		/* number of stripes worth of bytes to store
343 				 * parity information */
344 	int mindev_error;	/* error code if min devs requisite is unmet */
345 	const char raid_name[8]; /* name of the raid */
346 	u64 bg_flag;		/* block group flag of the raid */
347 };
348 
349 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
350 
351 struct map_lookup {
352 	u64 type;
353 	int io_align;
354 	int io_width;
355 	u64 stripe_len;
356 	int num_stripes;
357 	int sub_stripes;
358 	int verified_stripes; /* For mount time dev extent verification */
359 	struct btrfs_bio_stripe stripes[];
360 };
361 
362 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
363 			    (sizeof(struct btrfs_bio_stripe) * (n)))
364 
365 struct btrfs_balance_args;
366 struct btrfs_balance_progress;
367 struct btrfs_balance_control {
368 	struct btrfs_balance_args data;
369 	struct btrfs_balance_args meta;
370 	struct btrfs_balance_args sys;
371 
372 	u64 flags;
373 
374 	struct btrfs_balance_progress stat;
375 };
376 
377 enum btrfs_map_op {
378 	BTRFS_MAP_READ,
379 	BTRFS_MAP_WRITE,
380 	BTRFS_MAP_DISCARD,
381 	BTRFS_MAP_GET_READ_MIRRORS,
382 };
383 
384 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
385 {
386 	switch (bio_op(bio)) {
387 	case REQ_OP_DISCARD:
388 		return BTRFS_MAP_DISCARD;
389 	case REQ_OP_WRITE:
390 		return BTRFS_MAP_WRITE;
391 	default:
392 		WARN_ON_ONCE(1);
393 	case REQ_OP_READ:
394 		return BTRFS_MAP_READ;
395 	}
396 }
397 
398 void btrfs_get_bbio(struct btrfs_bio *bbio);
399 void btrfs_put_bbio(struct btrfs_bio *bbio);
400 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
401 		    u64 logical, u64 *length,
402 		    struct btrfs_bio **bbio_ret, int mirror_num);
403 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
404 		     u64 logical, u64 *length,
405 		     struct btrfs_bio **bbio_ret);
406 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
407 		     u64 physical, u64 **logical, int *naddrs, int *stripe_len);
408 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
409 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
410 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 type);
411 void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
412 void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
413 blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
414 			   int mirror_num, int async_submit);
415 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
416 		       fmode_t flags, void *holder);
417 struct btrfs_device *btrfs_scan_one_device(const char *path,
418 					   fmode_t flags, void *holder);
419 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
420 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices, int step);
421 void btrfs_assign_next_active_device(struct btrfs_device *device,
422 				     struct btrfs_device *this_dev);
423 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
424 						  u64 devid,
425 						  const char *devpath);
426 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
427 					const u64 *devid,
428 					const u8 *uuid);
429 void btrfs_free_device(struct btrfs_device *device);
430 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
431 		    const char *device_path, u64 devid);
432 void __exit btrfs_cleanup_fs_uuids(void);
433 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
434 int btrfs_grow_device(struct btrfs_trans_handle *trans,
435 		      struct btrfs_device *device, u64 new_size);
436 struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
437 				       u8 *uuid, u8 *fsid);
438 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
439 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
440 int btrfs_balance(struct btrfs_fs_info *fs_info,
441 		  struct btrfs_balance_control *bctl,
442 		  struct btrfs_ioctl_balance_args *bargs);
443 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
444 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
445 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
446 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
447 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
448 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
449 int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
450 int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset);
451 int find_free_dev_extent_start(struct btrfs_transaction *transaction,
452 			 struct btrfs_device *device, u64 num_bytes,
453 			 u64 search_start, u64 *start, u64 *max_avail);
454 int find_free_dev_extent(struct btrfs_trans_handle *trans,
455 			 struct btrfs_device *device, u64 num_bytes,
456 			 u64 *start, u64 *max_avail);
457 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
458 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
459 			struct btrfs_ioctl_get_dev_stats *stats);
460 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
461 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
462 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
463 			struct btrfs_fs_info *fs_info);
464 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
465 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_fs_info *fs_info,
466 				      struct btrfs_device *srcdev);
467 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
468 void btrfs_scratch_superblocks(struct block_device *bdev, const char *device_path);
469 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
470 			   u64 logical, u64 len);
471 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
472 				    u64 logical);
473 int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
474 			     u64 chunk_offset, u64 chunk_size);
475 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
476 struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
477 				       u64 logical, u64 length);
478 
479 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
480 				      int index)
481 {
482 	atomic_inc(dev->dev_stat_values + index);
483 	/*
484 	 * This memory barrier orders stores updating statistics before stores
485 	 * updating dev_stats_ccnt.
486 	 *
487 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
488 	 */
489 	smp_mb__before_atomic();
490 	atomic_inc(&dev->dev_stats_ccnt);
491 }
492 
493 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
494 				      int index)
495 {
496 	return atomic_read(dev->dev_stat_values + index);
497 }
498 
499 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
500 						int index)
501 {
502 	int ret;
503 
504 	ret = atomic_xchg(dev->dev_stat_values + index, 0);
505 	/*
506 	 * atomic_xchg implies a full memory barriers as per atomic_t.txt:
507 	 * - RMW operations that have a return value are fully ordered;
508 	 *
509 	 * This implicit memory barriers is paired with the smp_rmb in
510 	 * btrfs_run_dev_stats
511 	 */
512 	atomic_inc(&dev->dev_stats_ccnt);
513 	return ret;
514 }
515 
516 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
517 				      int index, unsigned long val)
518 {
519 	atomic_set(dev->dev_stat_values + index, val);
520 	/*
521 	 * This memory barrier orders stores updating statistics before stores
522 	 * updating dev_stats_ccnt.
523 	 *
524 	 * It pairs with smp_rmb() in btrfs_run_dev_stats().
525 	 */
526 	smp_mb__before_atomic();
527 	atomic_inc(&dev->dev_stats_ccnt);
528 }
529 
530 static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
531 					int index)
532 {
533 	btrfs_dev_stat_set(dev, index, 0);
534 }
535 
536 /*
537  * Convert block group flags (BTRFS_BLOCK_GROUP_*) to btrfs_raid_types, which
538  * can be used as index to access btrfs_raid_array[].
539  */
540 static inline enum btrfs_raid_types btrfs_bg_flags_to_raid_index(u64 flags)
541 {
542 	if (flags & BTRFS_BLOCK_GROUP_RAID10)
543 		return BTRFS_RAID_RAID10;
544 	else if (flags & BTRFS_BLOCK_GROUP_RAID1)
545 		return BTRFS_RAID_RAID1;
546 	else if (flags & BTRFS_BLOCK_GROUP_DUP)
547 		return BTRFS_RAID_DUP;
548 	else if (flags & BTRFS_BLOCK_GROUP_RAID0)
549 		return BTRFS_RAID_RAID0;
550 	else if (flags & BTRFS_BLOCK_GROUP_RAID5)
551 		return BTRFS_RAID_RAID5;
552 	else if (flags & BTRFS_BLOCK_GROUP_RAID6)
553 		return BTRFS_RAID_RAID6;
554 
555 	return BTRFS_RAID_SINGLE; /* BTRFS_BLOCK_GROUP_SINGLE */
556 }
557 
558 const char *get_raid_name(enum btrfs_raid_types type);
559 
560 void btrfs_update_commit_device_size(struct btrfs_fs_info *fs_info);
561 void btrfs_update_commit_device_bytes_used(struct btrfs_transaction *trans);
562 
563 struct list_head *btrfs_get_fs_uuids(void);
564 void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info);
565 void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info);
566 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
567 					struct btrfs_device *failing_dev);
568 
569 int btrfs_bg_type_to_factor(u64 flags);
570 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
571 
572 #endif
573