xref: /openbmc/linux/fs/btrfs/volumes.h (revision aac5987a)
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public
6  * License v2 as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public
14  * License along with this program; if not, write to the
15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16  * Boston, MA 021110-1307, USA.
17  */
18 
19 #ifndef __BTRFS_VOLUMES_
20 #define __BTRFS_VOLUMES_
21 
22 #include <linux/bio.h>
23 #include <linux/sort.h>
24 #include <linux/btrfs.h>
25 #include "async-thread.h"
26 
27 extern struct mutex uuid_mutex;
28 
29 #define BTRFS_STRIPE_LEN	SZ_64K
30 
31 struct buffer_head;
32 struct btrfs_pending_bios {
33 	struct bio *head;
34 	struct bio *tail;
35 };
36 
37 /*
38  * Use sequence counter to get consistent device stat data on
39  * 32-bit processors.
40  */
41 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
42 #include <linux/seqlock.h>
43 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
44 #define btrfs_device_data_ordered_init(device)	\
45 	seqcount_init(&device->data_seqcount)
46 #else
47 #define btrfs_device_data_ordered_init(device) do { } while (0)
48 #endif
49 
50 struct btrfs_device {
51 	struct list_head dev_list;
52 	struct list_head dev_alloc_list;
53 	struct btrfs_fs_devices *fs_devices;
54 	struct btrfs_fs_info *fs_info;
55 
56 	struct rcu_string *name;
57 
58 	u64 generation;
59 
60 	spinlock_t io_lock ____cacheline_aligned;
61 	int running_pending;
62 	/* regular prio bios */
63 	struct btrfs_pending_bios pending_bios;
64 	/* sync bios */
65 	struct btrfs_pending_bios pending_sync_bios;
66 
67 	struct block_device *bdev;
68 
69 	/* the mode sent to blkdev_get */
70 	fmode_t mode;
71 
72 	int writeable;
73 	int in_fs_metadata;
74 	int missing;
75 	int can_discard;
76 	int is_tgtdev_for_dev_replace;
77 
78 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
79 	seqcount_t data_seqcount;
80 #endif
81 
82 	/* the internal btrfs device id */
83 	u64 devid;
84 
85 	/* size of the device in memory */
86 	u64 total_bytes;
87 
88 	/* size of the device on disk */
89 	u64 disk_total_bytes;
90 
91 	/* bytes used */
92 	u64 bytes_used;
93 
94 	/* optimal io alignment for this device */
95 	u32 io_align;
96 
97 	/* optimal io width for this device */
98 	u32 io_width;
99 	/* type and info about this device */
100 	u64 type;
101 
102 	/* minimal io size for this device */
103 	u32 sector_size;
104 
105 	/* physical drive uuid (or lvm uuid) */
106 	u8 uuid[BTRFS_UUID_SIZE];
107 
108 	/*
109 	 * size of the device on the current transaction
110 	 *
111 	 * This variant is update when committing the transaction,
112 	 * and protected by device_list_mutex
113 	 */
114 	u64 commit_total_bytes;
115 
116 	/* bytes used on the current transaction */
117 	u64 commit_bytes_used;
118 	/*
119 	 * used to manage the device which is resized
120 	 *
121 	 * It is protected by chunk_lock.
122 	 */
123 	struct list_head resized_list;
124 
125 	/* for sending down flush barriers */
126 	int nobarriers;
127 	struct bio *flush_bio;
128 	struct completion flush_wait;
129 
130 	/* per-device scrub information */
131 	struct scrub_ctx *scrub_device;
132 
133 	struct btrfs_work work;
134 	struct rcu_head rcu;
135 	struct work_struct rcu_work;
136 
137 	/* readahead state */
138 	spinlock_t reada_lock;
139 	atomic_t reada_in_flight;
140 	u64 reada_next;
141 	struct reada_zone *reada_curr_zone;
142 	struct radix_tree_root reada_zones;
143 	struct radix_tree_root reada_extents;
144 
145 	/* disk I/O failure stats. For detailed description refer to
146 	 * enum btrfs_dev_stat_values in ioctl.h */
147 	int dev_stats_valid;
148 
149 	/* Counter to record the change of device stats */
150 	atomic_t dev_stats_ccnt;
151 	atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
152 };
153 
154 /*
155  * If we read those variants at the context of their own lock, we needn't
156  * use the following helpers, reading them directly is safe.
157  */
158 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
159 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
160 static inline u64							\
161 btrfs_device_get_##name(const struct btrfs_device *dev)			\
162 {									\
163 	u64 size;							\
164 	unsigned int seq;						\
165 									\
166 	do {								\
167 		seq = read_seqcount_begin(&dev->data_seqcount);		\
168 		size = dev->name;					\
169 	} while (read_seqcount_retry(&dev->data_seqcount, seq));	\
170 	return size;							\
171 }									\
172 									\
173 static inline void							\
174 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
175 {									\
176 	preempt_disable();						\
177 	write_seqcount_begin(&dev->data_seqcount);			\
178 	dev->name = size;						\
179 	write_seqcount_end(&dev->data_seqcount);			\
180 	preempt_enable();						\
181 }
182 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
183 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
184 static inline u64							\
185 btrfs_device_get_##name(const struct btrfs_device *dev)			\
186 {									\
187 	u64 size;							\
188 									\
189 	preempt_disable();						\
190 	size = dev->name;						\
191 	preempt_enable();						\
192 	return size;							\
193 }									\
194 									\
195 static inline void							\
196 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
197 {									\
198 	preempt_disable();						\
199 	dev->name = size;						\
200 	preempt_enable();						\
201 }
202 #else
203 #define BTRFS_DEVICE_GETSET_FUNCS(name)					\
204 static inline u64							\
205 btrfs_device_get_##name(const struct btrfs_device *dev)			\
206 {									\
207 	return dev->name;						\
208 }									\
209 									\
210 static inline void							\
211 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)		\
212 {									\
213 	dev->name = size;						\
214 }
215 #endif
216 
217 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
218 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
219 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
220 
221 struct btrfs_fs_devices {
222 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
223 
224 	u64 num_devices;
225 	u64 open_devices;
226 	u64 rw_devices;
227 	u64 missing_devices;
228 	u64 total_rw_bytes;
229 	u64 total_devices;
230 	struct block_device *latest_bdev;
231 
232 	/* all of the devices in the FS, protected by a mutex
233 	 * so we can safely walk it to write out the supers without
234 	 * worrying about add/remove by the multi-device code.
235 	 * Scrubbing super can kick off supers writing by holding
236 	 * this mutex lock.
237 	 */
238 	struct mutex device_list_mutex;
239 	struct list_head devices;
240 
241 	struct list_head resized_devices;
242 	/* devices not currently being allocated */
243 	struct list_head alloc_list;
244 	struct list_head list;
245 
246 	struct btrfs_fs_devices *seed;
247 	int seeding;
248 
249 	int opened;
250 
251 	/* set when we find or add a device that doesn't have the
252 	 * nonrot flag set
253 	 */
254 	int rotating;
255 
256 	struct btrfs_fs_info *fs_info;
257 	/* sysfs kobjects */
258 	struct kobject fsid_kobj;
259 	struct kobject *device_dir_kobj;
260 	struct completion kobj_unregister;
261 };
262 
263 #define BTRFS_BIO_INLINE_CSUM_SIZE	64
264 
265 /*
266  * we need the mirror number and stripe index to be passed around
267  * the call chain while we are processing end_io (especially errors).
268  * Really, what we need is a btrfs_bio structure that has this info
269  * and is properly sized with its stripe array, but we're not there
270  * quite yet.  We have our own btrfs bioset, and all of the bios
271  * we allocate are actually btrfs_io_bios.  We'll cram as much of
272  * struct btrfs_bio as we can into this over time.
273  */
274 typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err);
275 struct btrfs_io_bio {
276 	unsigned int mirror_num;
277 	unsigned int stripe_index;
278 	u64 logical;
279 	u8 *csum;
280 	u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
281 	u8 *csum_allocated;
282 	btrfs_io_bio_end_io_t *end_io;
283 	struct bio bio;
284 };
285 
286 static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
287 {
288 	return container_of(bio, struct btrfs_io_bio, bio);
289 }
290 
291 struct btrfs_bio_stripe {
292 	struct btrfs_device *dev;
293 	u64 physical;
294 	u64 length; /* only used for discard mappings */
295 };
296 
297 struct btrfs_bio;
298 typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err);
299 
300 struct btrfs_bio {
301 	atomic_t refs;
302 	atomic_t stripes_pending;
303 	struct btrfs_fs_info *fs_info;
304 	u64 map_type; /* get from map_lookup->type */
305 	bio_end_io_t *end_io;
306 	struct bio *orig_bio;
307 	unsigned long flags;
308 	void *private;
309 	atomic_t error;
310 	int max_errors;
311 	int num_stripes;
312 	int mirror_num;
313 	int num_tgtdevs;
314 	int *tgtdev_map;
315 	/*
316 	 * logical block numbers for the start of each stripe
317 	 * The last one or two are p/q.  These are sorted,
318 	 * so raid_map[0] is the start of our full stripe
319 	 */
320 	u64 *raid_map;
321 	struct btrfs_bio_stripe stripes[];
322 };
323 
324 struct btrfs_device_info {
325 	struct btrfs_device *dev;
326 	u64 dev_offset;
327 	u64 max_avail;
328 	u64 total_avail;
329 };
330 
331 struct btrfs_raid_attr {
332 	int sub_stripes;	/* sub_stripes info for map */
333 	int dev_stripes;	/* stripes per dev */
334 	int devs_max;		/* max devs to use */
335 	int devs_min;		/* min devs needed */
336 	int tolerated_failures; /* max tolerated fail devs */
337 	int devs_increment;	/* ndevs has to be a multiple of this */
338 	int ncopies;		/* how many copies to data has */
339 };
340 
341 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
342 extern const int btrfs_raid_mindev_error[BTRFS_NR_RAID_TYPES];
343 extern const u64 btrfs_raid_group[BTRFS_NR_RAID_TYPES];
344 
345 struct map_lookup {
346 	u64 type;
347 	int io_align;
348 	int io_width;
349 	u64 stripe_len;
350 	int sector_size;
351 	int num_stripes;
352 	int sub_stripes;
353 	struct btrfs_bio_stripe stripes[];
354 };
355 
356 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
357 			    (sizeof(struct btrfs_bio_stripe) * (n)))
358 
359 struct btrfs_balance_args;
360 struct btrfs_balance_progress;
361 struct btrfs_balance_control {
362 	struct btrfs_fs_info *fs_info;
363 
364 	struct btrfs_balance_args data;
365 	struct btrfs_balance_args meta;
366 	struct btrfs_balance_args sys;
367 
368 	u64 flags;
369 
370 	struct btrfs_balance_progress stat;
371 };
372 
373 enum btrfs_map_op {
374 	BTRFS_MAP_READ,
375 	BTRFS_MAP_WRITE,
376 	BTRFS_MAP_DISCARD,
377 	BTRFS_MAP_GET_READ_MIRRORS,
378 };
379 
380 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
381 {
382 	switch (bio_op(bio)) {
383 	case REQ_OP_DISCARD:
384 		return BTRFS_MAP_DISCARD;
385 	case REQ_OP_WRITE:
386 		return BTRFS_MAP_WRITE;
387 	default:
388 		WARN_ON_ONCE(1);
389 	case REQ_OP_READ:
390 		return BTRFS_MAP_READ;
391 	}
392 }
393 
394 int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
395 				   u64 end, u64 *length);
396 void btrfs_get_bbio(struct btrfs_bio *bbio);
397 void btrfs_put_bbio(struct btrfs_bio *bbio);
398 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
399 		    u64 logical, u64 *length,
400 		    struct btrfs_bio **bbio_ret, int mirror_num);
401 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
402 		     u64 logical, u64 *length,
403 		     struct btrfs_bio **bbio_ret, int mirror_num,
404 		     int need_raid_map);
405 int btrfs_rmap_block(struct btrfs_fs_info *fs_info,
406 		     u64 chunk_start, u64 physical, u64 devid,
407 		     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,
411 		      struct btrfs_fs_info *fs_info, u64 type);
412 void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
413 void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
414 int btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
415 		  int mirror_num, int async_submit);
416 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
417 		       fmode_t flags, void *holder);
418 int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
419 			  struct btrfs_fs_devices **fs_devices_ret);
420 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
421 void btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices, int step);
422 void btrfs_assign_next_active_device(struct btrfs_fs_info *fs_info,
423 		struct btrfs_device *device, struct btrfs_device *this_dev);
424 int btrfs_find_device_missing_or_by_path(struct btrfs_fs_info *fs_info,
425 					 const char *device_path,
426 					 struct btrfs_device **device);
427 int btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info, u64 devid,
428 					 const char *devpath,
429 					 struct btrfs_device **device);
430 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
431 					const u64 *devid,
432 					const u8 *uuid);
433 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
434 		    const char *device_path, u64 devid);
435 void btrfs_cleanup_fs_uuids(void);
436 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
437 int btrfs_grow_device(struct btrfs_trans_handle *trans,
438 		      struct btrfs_device *device, u64 new_size);
439 struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
440 				       u8 *uuid, u8 *fsid);
441 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
442 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
443 int btrfs_init_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
444 				  const char *device_path,
445 				  struct btrfs_device *srcdev,
446 				  struct btrfs_device **device_out);
447 int btrfs_balance(struct btrfs_balance_control *bctl,
448 		  struct btrfs_ioctl_balance_args *bargs);
449 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
450 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
451 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
452 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
453 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
454 int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
455 int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset);
456 int find_free_dev_extent_start(struct btrfs_transaction *transaction,
457 			 struct btrfs_device *device, u64 num_bytes,
458 			 u64 search_start, u64 *start, u64 *max_avail);
459 int find_free_dev_extent(struct btrfs_trans_handle *trans,
460 			 struct btrfs_device *device, u64 num_bytes,
461 			 u64 *start, u64 *max_avail);
462 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
463 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
464 			struct btrfs_ioctl_get_dev_stats *stats);
465 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
466 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
467 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
468 			struct btrfs_fs_info *fs_info);
469 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_fs_info *fs_info,
470 					struct btrfs_device *srcdev);
471 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_fs_info *fs_info,
472 				      struct btrfs_device *srcdev);
473 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
474 				      struct btrfs_device *tgtdev);
475 void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info,
476 					      struct btrfs_device *tgtdev);
477 void btrfs_scratch_superblocks(struct block_device *bdev, const char *device_path);
478 int btrfs_is_parity_mirror(struct btrfs_mapping_tree *map_tree,
479 			   u64 logical, u64 len, int mirror_num);
480 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
481 				    struct btrfs_mapping_tree *map_tree,
482 				    u64 logical);
483 int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
484 				struct btrfs_fs_info *fs_info,
485 				u64 chunk_offset, u64 chunk_size);
486 int btrfs_remove_chunk(struct btrfs_trans_handle *trans,
487 		       struct btrfs_fs_info *fs_info, u64 chunk_offset);
488 
489 static inline int btrfs_dev_stats_dirty(struct btrfs_device *dev)
490 {
491 	return atomic_read(&dev->dev_stats_ccnt);
492 }
493 
494 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
495 				      int index)
496 {
497 	atomic_inc(dev->dev_stat_values + index);
498 	smp_mb__before_atomic();
499 	atomic_inc(&dev->dev_stats_ccnt);
500 }
501 
502 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
503 				      int index)
504 {
505 	return atomic_read(dev->dev_stat_values + index);
506 }
507 
508 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
509 						int index)
510 {
511 	int ret;
512 
513 	ret = atomic_xchg(dev->dev_stat_values + index, 0);
514 	smp_mb__before_atomic();
515 	atomic_inc(&dev->dev_stats_ccnt);
516 	return ret;
517 }
518 
519 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
520 				      int index, unsigned long val)
521 {
522 	atomic_set(dev->dev_stat_values + index, val);
523 	smp_mb__before_atomic();
524 	atomic_inc(&dev->dev_stats_ccnt);
525 }
526 
527 static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
528 					int index)
529 {
530 	btrfs_dev_stat_set(dev, index, 0);
531 }
532 
533 void btrfs_update_commit_device_size(struct btrfs_fs_info *fs_info);
534 void btrfs_update_commit_device_bytes_used(struct btrfs_fs_info *fs_info,
535 					struct btrfs_transaction *transaction);
536 
537 struct list_head *btrfs_get_fs_uuids(void);
538 void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info);
539 void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info);
540 
541 #endif
542