xref: /linux/fs/btrfs/block-group.h (revision 6f7e6393d1ce636bb7ec77a7fe7b77458fddf701)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 
3 #ifndef BTRFS_BLOCK_GROUP_H
4 #define BTRFS_BLOCK_GROUP_H
5 
6 #include <linux/atomic.h>
7 #include <linux/mutex.h>
8 #include <linux/list.h>
9 #include <linux/spinlock.h>
10 #include <linux/refcount.h>
11 #include <linux/wait.h>
12 #include <linux/sizes.h>
13 #include <linux/rwsem.h>
14 #include <linux/rbtree.h>
15 #include <uapi/linux/btrfs_tree.h>
16 #include "free-space-cache.h"
17 
18 struct btrfs_chunk_map;
19 struct btrfs_fs_info;
20 struct btrfs_inode;
21 struct btrfs_trans_handle;
22 
23 enum btrfs_disk_cache_state {
24 	BTRFS_DC_WRITTEN,
25 	BTRFS_DC_ERROR,
26 	BTRFS_DC_CLEAR,
27 	BTRFS_DC_SETUP,
28 };
29 
30 enum btrfs_block_group_size_class {
31 	/* Unset */
32 	BTRFS_BG_SZ_NONE,
33 	/* 0 < size <= 128K */
34 	BTRFS_BG_SZ_SMALL,
35 	/* 128K < size <= 8M */
36 	BTRFS_BG_SZ_MEDIUM,
37 	/* 8M < size < BG_LENGTH */
38 	BTRFS_BG_SZ_LARGE,
39 };
40 
41 /*
42  * This describes the state of the block_group for async discard.  This is due
43  * to the two pass nature of it where extent discarding is prioritized over
44  * bitmap discarding.  BTRFS_DISCARD_RESET_CURSOR is set when we are resetting
45  * between lists to prevent contention for discard state variables
46  * (eg. discard_cursor).
47  */
48 enum btrfs_discard_state {
49 	BTRFS_DISCARD_EXTENTS,
50 	BTRFS_DISCARD_BITMAPS,
51 	BTRFS_DISCARD_RESET_CURSOR,
52 	BTRFS_DISCARD_FULLY_REMAPPED,
53 };
54 
55 /*
56  * Control flags for do_chunk_alloc's force field CHUNK_ALLOC_NO_FORCE means to
57  * only allocate a chunk if we really need one.
58  *
59  * CHUNK_ALLOC_LIMITED means to only try and allocate one if we have very few
60  * chunks already allocated.  This is used as part of the clustering code to
61  * help make sure we have a good pool of storage to cluster in, without filling
62  * the FS with empty chunks
63  *
64  * CHUNK_ALLOC_FORCE means it must try to allocate one
65  *
66  * CHUNK_ALLOC_FORCE_FOR_EXTENT like CHUNK_ALLOC_FORCE but called from
67  * find_free_extent() that also activates the zone
68  */
69 enum btrfs_chunk_alloc_enum {
70 	CHUNK_ALLOC_NO_FORCE,
71 	CHUNK_ALLOC_LIMITED,
72 	CHUNK_ALLOC_FORCE,
73 	CHUNK_ALLOC_FORCE_FOR_EXTENT,
74 };
75 
76 /* Block group flags set at runtime */
77 enum btrfs_block_group_flags {
78 	BLOCK_GROUP_FLAG_IREF,
79 	BLOCK_GROUP_FLAG_REMOVED,
80 	BLOCK_GROUP_FLAG_TO_COPY,
81 	BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
82 	BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED,
83 	BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
84 	BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
85 	/* Does the block group need to be added to the free space tree? */
86 	BLOCK_GROUP_FLAG_NEEDS_FREE_SPACE,
87 	/* Set after we add a new block group to the free space tree. */
88 	BLOCK_GROUP_FLAG_FREE_SPACE_ADDED,
89 	/* Indicate that the block group is placed on a sequential zone */
90 	BLOCK_GROUP_FLAG_SEQUENTIAL_ZONE,
91 	/*
92 	 * Indicate that block group is in the list of new block groups of a
93 	 * transaction.
94 	 */
95 	BLOCK_GROUP_FLAG_NEW,
96 	BLOCK_GROUP_FLAG_FULLY_REMAPPED,
97 	BLOCK_GROUP_FLAG_STRIPE_REMOVAL_PENDING,
98 };
99 
100 enum btrfs_caching_type {
101 	BTRFS_CACHE_NO,
102 	BTRFS_CACHE_STARTED,
103 	BTRFS_CACHE_FINISHED,
104 	BTRFS_CACHE_ERROR,
105 };
106 
107 struct btrfs_caching_control {
108 	struct list_head list;
109 	struct mutex mutex;
110 	wait_queue_head_t wait;
111 	struct btrfs_work work;
112 	struct btrfs_block_group *block_group;
113 	/* Track progress of caching during allocation. */
114 	atomic_t progress;
115 	refcount_t count;
116 };
117 
118 /* Once caching_thread() finds this much free space, it will wake up waiters. */
119 #define CACHING_CTL_WAKE_UP SZ_2M
120 
121 struct btrfs_block_group {
122 	struct btrfs_fs_info *fs_info;
123 	struct btrfs_inode *inode;
124 	spinlock_t lock;
125 	u64 start;
126 	u64 length;
127 	u64 pinned;
128 	u64 reserved;
129 	u64 used;
130 	u64 delalloc_bytes;
131 	u64 bytes_super;
132 	u64 flags;
133 	u64 cache_generation;
134 	u64 global_root_id;
135 	u64 remap_bytes;
136 	u32 identity_remap_count;
137 
138 	/*
139 	 * The last committed used bytes of this block group, if the above @used
140 	 * is still the same as @last_used, we don't need to update block
141 	 * group item of this block group.
142 	 */
143 	u64 last_used;
144 	/* The last committed remap_bytes value of this block group. */
145 	u64 last_remap_bytes;
146 	/* The last commited identity_remap_count value of this block group. */
147 	u32 last_identity_remap_count;
148 	/* The last committed flags value for this block group. */
149 	u64 last_flags;
150 
151 	/*
152 	 * If the free space extent count exceeds this number, convert the block
153 	 * group to bitmaps.
154 	 */
155 	u32 bitmap_high_thresh;
156 
157 	/*
158 	 * If the free space extent count drops below this number, convert the
159 	 * block group back to extents.
160 	 */
161 	u32 bitmap_low_thresh;
162 
163 	/*
164 	 * It is just used for the delayed data space allocation because
165 	 * only the data space allocation and the relative metadata update
166 	 * can be done cross the transaction.
167 	 */
168 	struct rw_semaphore data_rwsem;
169 
170 	/* For raid56, this is a full stripe, without parity */
171 	unsigned long full_stripe_len;
172 	unsigned long runtime_flags;
173 
174 	unsigned int ro;
175 
176 	int disk_cache_state;
177 
178 	/* Cache tracking stuff */
179 	int cached;
180 	struct btrfs_caching_control *caching_ctl;
181 
182 	struct btrfs_space_info *space_info;
183 
184 	/* Free space cache stuff */
185 	struct btrfs_free_space_ctl *free_space_ctl;
186 
187 	/* Block group cache stuff */
188 	struct rb_node cache_node;
189 
190 	/* For block groups in the same raid type */
191 	struct list_head list;
192 
193 	refcount_t refs;
194 
195 	/*
196 	 * List of struct btrfs_free_clusters for this block group.
197 	 * Today it will only have one thing on it, but that may change
198 	 */
199 	struct list_head cluster_list;
200 
201 	/*
202 	 * Used for several lists:
203 	 *
204 	 * 1) struct btrfs_fs_info::unused_bgs
205 	 * 2) struct btrfs_fs_info::reclaim_bgs
206 	 * 3) struct btrfs_transaction::deleted_bgs
207 	 * 4) struct btrfs_trans_handle::new_bgs
208 	 */
209 	struct list_head bg_list;
210 
211 	/* For read-only block groups */
212 	struct list_head ro_list;
213 
214 	/*
215 	 * When non-zero it means the block group's logical address and its
216 	 * device extents can not be reused for future block group allocations
217 	 * until the counter goes down to 0. This is to prevent them from being
218 	 * reused while some task is still using the block group after it was
219 	 * deleted - we want to make sure they can only be reused for new block
220 	 * groups after that task is done with the deleted block group.
221 	 */
222 	atomic_t frozen;
223 
224 	/* For discard operations */
225 	struct list_head discard_list;
226 	int discard_index;
227 	u64 discard_eligible_time;
228 	u64 discard_cursor;
229 	enum btrfs_discard_state discard_state;
230 
231 	/* For dirty block groups */
232 	struct list_head dirty_list;
233 	struct list_head io_list;
234 
235 	struct btrfs_io_ctl io_ctl;
236 
237 	/*
238 	 * Incremented when doing extent allocations and holding a read lock
239 	 * on the space_info's groups_sem semaphore.
240 	 * Decremented when an ordered extent that represents an IO against this
241 	 * block group's range is created (after it's added to its inode's
242 	 * root's list of ordered extents) or immediately after the allocation
243 	 * if it's a metadata extent or fallocate extent (for these cases we
244 	 * don't create ordered extents).
245 	 */
246 	atomic_t reservations;
247 
248 	/*
249 	 * Incremented while holding the spinlock *lock* by a task checking if
250 	 * it can perform a nocow write (incremented if the value for the *ro*
251 	 * field is 0). Decremented by such tasks once they create an ordered
252 	 * extent or before that if some error happens before reaching that step.
253 	 * This is to prevent races between block group relocation and nocow
254 	 * writes through direct IO.
255 	 */
256 	atomic_t nocow_writers;
257 
258 	/* Lock for free space tree operations. */
259 	struct mutex free_space_lock;
260 
261 	/* Protected by @free_space_lock. */
262 	bool using_free_space_bitmaps;
263 	/* Protected by @free_space_lock. */
264 	bool using_free_space_bitmaps_cached;
265 
266 	/*
267 	 * Number of extents in this block group used for swap files.
268 	 * All accesses protected by the spinlock 'lock'.
269 	 */
270 	int swap_extents;
271 
272 	/*
273 	 * Allocation offset for the block group to implement sequential
274 	 * allocation. This is used only on a zoned filesystem.
275 	 */
276 	u64 alloc_offset;
277 	u64 zone_unusable;
278 	u64 zone_capacity;
279 	u64 meta_write_pointer;
280 	struct btrfs_chunk_map *physical_map;
281 	struct list_head active_bg_list;
282 	struct work_struct zone_finish_work;
283 	struct extent_buffer *last_eb;
284 	enum btrfs_block_group_size_class size_class;
285 	u64 reclaim_mark;
286 };
287 
288 static inline u64 btrfs_block_group_end(const struct btrfs_block_group *block_group)
289 {
290 	return (block_group->start + block_group->length);
291 }
292 
293 static inline bool btrfs_is_block_group_used(const struct btrfs_block_group *bg)
294 {
295 	lockdep_assert_held(&bg->lock);
296 
297 	return (bg->used > 0 || bg->reserved > 0 || bg->pinned > 0 ||
298 		bg->remap_bytes > 0);
299 }
300 
301 static inline bool btrfs_is_block_group_data_only(const struct btrfs_block_group *block_group)
302 {
303 	/*
304 	 * In mixed mode the fragmentation is expected to be high, lowering the
305 	 * efficiency, so only proper data block groups are considered.
306 	 */
307 	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
308 	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
309 }
310 
311 static inline u64 btrfs_block_group_available_space(const struct btrfs_block_group *bg)
312 {
313 	lockdep_assert_held(&bg->lock);
314 
315 	return (bg->length - bg->used - bg->pinned - bg->reserved -
316 		bg->bytes_super - bg->zone_unusable);
317 }
318 
319 #ifdef CONFIG_BTRFS_DEBUG
320 int btrfs_should_fragment_free_space(const struct btrfs_block_group *block_group);
321 #endif
322 
323 struct btrfs_block_group *btrfs_lookup_first_block_group(
324 		struct btrfs_fs_info *info, u64 bytenr);
325 struct btrfs_block_group *btrfs_lookup_block_group(
326 		struct btrfs_fs_info *info, u64 bytenr);
327 struct btrfs_block_group *btrfs_next_block_group(
328 		struct btrfs_block_group *cache);
329 void btrfs_get_block_group(struct btrfs_block_group *cache);
330 void btrfs_put_block_group(struct btrfs_block_group *cache);
331 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
332 					const u64 start);
333 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
334 struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
335 						  u64 bytenr);
336 void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
337 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
338 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
339 				           u64 num_bytes);
340 int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
341 struct btrfs_caching_control *btrfs_get_caching_control(
342 		struct btrfs_block_group *cache);
343 int btrfs_add_new_free_space(struct btrfs_block_group *block_group,
344 			     u64 start, u64 end, u64 *total_added_ret);
345 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
346 				struct btrfs_fs_info *fs_info,
347 				const u64 chunk_offset);
348 void btrfs_remove_bg_from_sinfo(struct btrfs_block_group *bg);
349 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
350 			     struct btrfs_chunk_map *map);
351 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
352 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
353 void btrfs_reclaim_bgs_work(struct work_struct *work);
354 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
355 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
356 int btrfs_read_block_groups(struct btrfs_fs_info *info);
357 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
358 						 struct btrfs_space_info *space_info,
359 						 u64 type, u64 chunk_offset, u64 size);
360 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
361 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
362 			     bool do_chunk_alloc);
363 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
364 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
365 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
366 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
367 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
368 			     u64 bytenr, u64 num_bytes, bool alloc);
369 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
370 			     u64 ram_bytes, u64 num_bytes, bool delalloc,
371 			     bool force_wrong_size_class);
372 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache, u64 num_bytes,
373 			       bool is_delalloc);
374 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans,
375 		      struct btrfs_space_info *space_info, u64 flags,
376 		      enum btrfs_chunk_alloc_enum force);
377 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
378 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
379 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
380 				  bool is_item_insertion);
381 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
382 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
383 int btrfs_free_block_groups(struct btrfs_fs_info *info);
384 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
385 		     u64 physical, u64 **logical, int *naddrs, int *stripe_len);
386 
387 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
388 {
389 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
390 }
391 
392 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
393 {
394 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
395 }
396 
397 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
398 {
399 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
400 }
401 
402 static inline int btrfs_block_group_done(const struct btrfs_block_group *cache)
403 {
404 	smp_mb();
405 	return cache->cached == BTRFS_CACHE_FINISHED ||
406 		cache->cached == BTRFS_CACHE_ERROR;
407 }
408 
409 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
410 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
411 
412 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
413 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
414 
415 enum btrfs_block_group_size_class btrfs_calc_block_group_size_class(u64 size);
416 int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
417 				     enum btrfs_block_group_size_class size_class,
418 				     bool force_wrong_size_class);
419 bool btrfs_block_group_should_use_size_class(const struct btrfs_block_group *bg);
420 void btrfs_mark_bg_fully_remapped(struct btrfs_block_group *bg,
421 				  struct btrfs_trans_handle *trans);
422 int btrfs_populate_fully_remapped_bgs_list(struct btrfs_fs_info *fs_info);
423 
424 #endif /* BTRFS_BLOCK_GROUP_H */
425