1/* SPDX-License-Identifier: GPL-2.0 */
  2
  3#ifndef BTRFS_BLOCK_GROUP_H
  4#define BTRFS_BLOCK_GROUP_H
  5
  6#include "free-space-cache.h"
  7
  8enum btrfs_disk_cache_state {
  9	BTRFS_DC_WRITTEN,
 10	BTRFS_DC_ERROR,
 11	BTRFS_DC_CLEAR,
 12	BTRFS_DC_SETUP,
 13};
 14
 15/*
 16 * This describes the state of the block_group for async discard.  This is due
 17 * to the two pass nature of it where extent discarding is prioritized over
 18 * bitmap discarding.  BTRFS_DISCARD_RESET_CURSOR is set when we are resetting
 19 * between lists to prevent contention for discard state variables
 20 * (eg. discard_cursor).
 21 */
 22enum btrfs_discard_state {
 23	BTRFS_DISCARD_EXTENTS,
 24	BTRFS_DISCARD_BITMAPS,
 25	BTRFS_DISCARD_RESET_CURSOR,
 26};
 27
 28/*
 29 * Control flags for do_chunk_alloc's force field CHUNK_ALLOC_NO_FORCE means to
 30 * only allocate a chunk if we really need one.
 31 *
 32 * CHUNK_ALLOC_LIMITED means to only try and allocate one if we have very few
 33 * chunks already allocated.  This is used as part of the clustering code to
 34 * help make sure we have a good pool of storage to cluster in, without filling
 35 * the FS with empty chunks
 36 *
 37 * CHUNK_ALLOC_FORCE means it must try to allocate one
 38 */
 39enum btrfs_chunk_alloc_enum {
 40	CHUNK_ALLOC_NO_FORCE,
 41	CHUNK_ALLOC_LIMITED,
 42	CHUNK_ALLOC_FORCE,
 43};
 44
 45struct btrfs_caching_control {
 46	struct list_head list;
 47	struct mutex mutex;
 48	wait_queue_head_t wait;
 49	struct btrfs_work work;
 50	struct btrfs_block_group *block_group;
 51	u64 progress;
 52	refcount_t count;
 53};
 54
 55/* Once caching_thread() finds this much free space, it will wake up waiters. */
 56#define CACHING_CTL_WAKE_UP SZ_2M
 57
 58struct btrfs_block_group {
 59	struct btrfs_fs_info *fs_info;
 60	struct inode *inode;
 61	spinlock_t lock;
 62	u64 start;
 63	u64 length;
 64	u64 pinned;
 65	u64 reserved;
 66	u64 used;
 67	u64 delalloc_bytes;
 68	u64 bytes_super;
 69	u64 flags;
 70	u64 cache_generation;
 71
 72	/*
 73	 * If the free space extent count exceeds this number, convert the block
 74	 * group to bitmaps.
 75	 */
 76	u32 bitmap_high_thresh;
 77
 78	/*
 79	 * If the free space extent count drops below this number, convert the
 80	 * block group back to extents.
 81	 */
 82	u32 bitmap_low_thresh;
 83
 84	/*
 85	 * It is just used for the delayed data space allocation because
 86	 * only the data space allocation and the relative metadata update
 87	 * can be done cross the transaction.
 88	 */
 89	struct rw_semaphore data_rwsem;
 90
 91	/* For raid56, this is a full stripe, without parity */
 92	unsigned long full_stripe_len;
 93
 94	unsigned int ro;
 95	unsigned int iref:1;
 96	unsigned int has_caching_ctl:1;
 97	unsigned int removed:1;
 98	unsigned int to_copy:1;
 99	unsigned int relocating_repair:1;
100	unsigned int chunk_item_inserted:1;
101
102	int disk_cache_state;
103
104	/* Cache tracking stuff */
105	int cached;
106	struct btrfs_caching_control *caching_ctl;
107	u64 last_byte_to_unpin;
108
109	struct btrfs_space_info *space_info;
110
111	/* Free space cache stuff */
112	struct btrfs_free_space_ctl *free_space_ctl;
113
114	/* Block group cache stuff */
115	struct rb_node cache_node;
116
117	/* For block groups in the same raid type */
118	struct list_head list;
119
120	refcount_t refs;
121
122	/*
123	 * List of struct btrfs_free_clusters for this block group.
124	 * Today it will only have one thing on it, but that may change
125	 */
126	struct list_head cluster_list;
127
128	/* For delayed block group creation or deletion of empty block groups */
129	struct list_head bg_list;
130
131	/* For read-only block groups */
132	struct list_head ro_list;
133
134	/*
135	 * When non-zero it means the block group's logical address and its
136	 * device extents can not be reused for future block group allocations
137	 * until the counter goes down to 0. This is to prevent them from being
138	 * reused while some task is still using the block group after it was
139	 * deleted - we want to make sure they can only be reused for new block
140	 * groups after that task is done with the deleted block group.
141	 */
142	atomic_t frozen;
143
144	/* For discard operations */
145	struct list_head discard_list;
146	int discard_index;
147	u64 discard_eligible_time;
148	u64 discard_cursor;
149	enum btrfs_discard_state discard_state;
150
151	/* For dirty block groups */
152	struct list_head dirty_list;
153	struct list_head io_list;
154
155	struct btrfs_io_ctl io_ctl;
156
157	/*
158	 * Incremented when doing extent allocations and holding a read lock
159	 * on the space_info's groups_sem semaphore.
160	 * Decremented when an ordered extent that represents an IO against this
161	 * block group's range is created (after it's added to its inode's
162	 * root's list of ordered extents) or immediately after the allocation
163	 * if it's a metadata extent or fallocate extent (for these cases we
164	 * don't create ordered extents).
165	 */
166	atomic_t reservations;
167
168	/*
169	 * Incremented while holding the spinlock *lock* by a task checking if
170	 * it can perform a nocow write (incremented if the value for the *ro*
171	 * field is 0). Decremented by such tasks once they create an ordered
172	 * extent or before that if some error happens before reaching that step.
173	 * This is to prevent races between block group relocation and nocow
174	 * writes through direct IO.
175	 */
176	atomic_t nocow_writers;
177
178	/* Lock for free space tree operations. */
179	struct mutex free_space_lock;
180
181	/*
182	 * Does the block group need to be added to the free space tree?
183	 * Protected by free_space_lock.
184	 */
185	int needs_free_space;
186
187	/* Flag indicating this block group is placed on a sequential zone */
188	bool seq_zone;
189
190	/*
191	 * Number of extents in this block group used for swap files.
192	 * All accesses protected by the spinlock 'lock'.
193	 */
194	int swap_extents;
195
196	/* Record locked full stripes for RAID5/6 block group */
197	struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
198
199	/*
200	 * Allocation offset for the block group to implement sequential
201	 * allocation. This is used only on a zoned filesystem.
202	 */
203	u64 alloc_offset;
204	u64 zone_unusable;
205	u64 meta_write_pointer;
206};
207
208static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
209{
210	return (block_group->start + block_group->length);
211}
212
213static inline bool btrfs_is_block_group_data_only(
214					struct btrfs_block_group *block_group)
215{
216	/*
217	 * In mixed mode the fragmentation is expected to be high, lowering the
218	 * efficiency, so only proper data block groups are considered.
219	 */
220	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
221	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
222}
223
224#ifdef CONFIG_BTRFS_DEBUG
225static inline int btrfs_should_fragment_free_space(
226		struct btrfs_block_group *block_group)
227{
228	struct btrfs_fs_info *fs_info = block_group->fs_info;
229
230	return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
231		block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
232	       (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
233		block_group->flags &  BTRFS_BLOCK_GROUP_DATA);
234}
235#endif
236
237struct btrfs_block_group *btrfs_lookup_first_block_group(
238		struct btrfs_fs_info *info, u64 bytenr);
239struct btrfs_block_group *btrfs_lookup_block_group(
240		struct btrfs_fs_info *info, u64 bytenr);
241struct btrfs_block_group *btrfs_next_block_group(
242		struct btrfs_block_group *cache);
243void btrfs_get_block_group(struct btrfs_block_group *cache);
244void btrfs_put_block_group(struct btrfs_block_group *cache);
245void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
246					const u64 start);
247void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
248bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
249void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
250void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
251void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
252				           u64 num_bytes);
253int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache);
254int btrfs_cache_block_group(struct btrfs_block_group *cache,
255			    int load_cache_only);
256void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
257struct btrfs_caching_control *btrfs_get_caching_control(
258		struct btrfs_block_group *cache);
259u64 add_new_free_space(struct btrfs_block_group *block_group,
260		       u64 start, u64 end);
261struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
262				struct btrfs_fs_info *fs_info,
263				const u64 chunk_offset);
264int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
265			     u64 group_start, struct extent_map *em);
266void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
267void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
268void btrfs_reclaim_bgs_work(struct work_struct *work);
269void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
270void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
271int btrfs_read_block_groups(struct btrfs_fs_info *info);
272struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
273						 u64 bytes_used, u64 type,
274						 u64 chunk_offset, u64 size);
275void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
276int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
277			     bool do_chunk_alloc);
278void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
279int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
280int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
281int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
282int btrfs_update_block_group(struct btrfs_trans_handle *trans,
283			     u64 bytenr, u64 num_bytes, int alloc);
284int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
285			     u64 ram_bytes, u64 num_bytes, int delalloc);
286void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
287			       u64 num_bytes, int delalloc);
288int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
289		      enum btrfs_chunk_alloc_enum force);
290int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
291void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
292u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
293void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
294int btrfs_free_block_groups(struct btrfs_fs_info *info);
295void btrfs_wait_space_cache_v1_finished(struct btrfs_block_group *cache,
296				struct btrfs_caching_control *caching_ctl);
297int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
298		       struct block_device *bdev, u64 physical, u64 **logical,
299		       int *naddrs, int *stripe_len);
300
301static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
302{
303	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
304}
305
306static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
307{
308	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
309}
310
311static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
312{
313	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
314}
315
316static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
317{
318	smp_mb();
319	return cache->cached == BTRFS_CACHE_FINISHED ||
320		cache->cached == BTRFS_CACHE_ERROR;
321}
322
323void btrfs_freeze_block_group(struct btrfs_block_group *cache);
324void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
325
326bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
327void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
328
329#endif /* BTRFS_BLOCK_GROUP_H */