block-group.h

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