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 8 enum 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 */ 22 enum 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 */ 39 enum btrfs_chunk_alloc_enum { 40 CHUNK_ALLOC_NO_FORCE, 41 CHUNK_ALLOC_LIMITED, 42 CHUNK_ALLOC_FORCE, 43 }; 44 45 struct 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 58 struct 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 u64 global_root_id; 72 73 /* 74 * If the free space extent count exceeds this number, convert the block 75 * group to bitmaps. 76 */ 77 u32 bitmap_high_thresh; 78 79 /* 80 * If the free space extent count drops below this number, convert the 81 * block group back to extents. 82 */ 83 u32 bitmap_low_thresh; 84 85 /* 86 * It is just used for the delayed data space allocation because 87 * only the data space allocation and the relative metadata update 88 * can be done cross the transaction. 89 */ 90 struct rw_semaphore data_rwsem; 91 92 /* For raid56, this is a full stripe, without parity */ 93 unsigned long full_stripe_len; 94 95 unsigned int ro; 96 unsigned int iref:1; 97 unsigned int has_caching_ctl:1; 98 unsigned int removed:1; 99 unsigned int to_copy:1; 100 unsigned int relocating_repair:1; 101 unsigned int chunk_item_inserted:1; 102 unsigned int zone_is_active:1; 103 104 int disk_cache_state; 105 106 /* Cache tracking stuff */ 107 int cached; 108 struct btrfs_caching_control *caching_ctl; 109 u64 last_byte_to_unpin; 110 111 struct btrfs_space_info *space_info; 112 113 /* Free space cache stuff */ 114 struct btrfs_free_space_ctl *free_space_ctl; 115 116 /* Block group cache stuff */ 117 struct rb_node cache_node; 118 119 /* For block groups in the same raid type */ 120 struct list_head list; 121 122 refcount_t refs; 123 124 /* 125 * List of struct btrfs_free_clusters for this block group. 126 * Today it will only have one thing on it, but that may change 127 */ 128 struct list_head cluster_list; 129 130 /* For delayed block group creation or deletion of empty block groups */ 131 struct list_head bg_list; 132 133 /* For read-only block groups */ 134 struct list_head ro_list; 135 136 /* 137 * When non-zero it means the block group's logical address and its 138 * device extents can not be reused for future block group allocations 139 * until the counter goes down to 0. This is to prevent them from being 140 * reused while some task is still using the block group after it was 141 * deleted - we want to make sure they can only be reused for new block 142 * groups after that task is done with the deleted block group. 143 */ 144 atomic_t frozen; 145 146 /* For discard operations */ 147 struct list_head discard_list; 148 int discard_index; 149 u64 discard_eligible_time; 150 u64 discard_cursor; 151 enum btrfs_discard_state discard_state; 152 153 /* For dirty block groups */ 154 struct list_head dirty_list; 155 struct list_head io_list; 156 157 struct btrfs_io_ctl io_ctl; 158 159 /* 160 * Incremented when doing extent allocations and holding a read lock 161 * on the space_info's groups_sem semaphore. 162 * Decremented when an ordered extent that represents an IO against this 163 * block group's range is created (after it's added to its inode's 164 * root's list of ordered extents) or immediately after the allocation 165 * if it's a metadata extent or fallocate extent (for these cases we 166 * don't create ordered extents). 167 */ 168 atomic_t reservations; 169 170 /* 171 * Incremented while holding the spinlock *lock* by a task checking if 172 * it can perform a nocow write (incremented if the value for the *ro* 173 * field is 0). Decremented by such tasks once they create an ordered 174 * extent or before that if some error happens before reaching that step. 175 * This is to prevent races between block group relocation and nocow 176 * writes through direct IO. 177 */ 178 atomic_t nocow_writers; 179 180 /* Lock for free space tree operations. */ 181 struct mutex free_space_lock; 182 183 /* 184 * Does the block group need to be added to the free space tree? 185 * Protected by free_space_lock. 186 */ 187 int needs_free_space; 188 189 /* Flag indicating this block group is placed on a sequential zone */ 190 bool seq_zone; 191 192 /* 193 * Number of extents in this block group used for swap files. 194 * All accesses protected by the spinlock 'lock'. 195 */ 196 int swap_extents; 197 198 /* Record locked full stripes for RAID5/6 block group */ 199 struct btrfs_full_stripe_locks_tree full_stripe_locks_root; 200 201 /* 202 * Allocation offset for the block group to implement sequential 203 * allocation. This is used only on a zoned filesystem. 204 */ 205 u64 alloc_offset; 206 u64 zone_unusable; 207 u64 zone_capacity; 208 u64 meta_write_pointer; 209 struct map_lookup *physical_map; 210 struct list_head active_bg_list; 211 }; 212 213 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group) 214 { 215 return (block_group->start + block_group->length); 216 } 217 218 static inline bool btrfs_is_block_group_data_only( 219 struct btrfs_block_group *block_group) 220 { 221 /* 222 * In mixed mode the fragmentation is expected to be high, lowering the 223 * efficiency, so only proper data block groups are considered. 224 */ 225 return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) && 226 !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA); 227 } 228 229 #ifdef CONFIG_BTRFS_DEBUG 230 static inline int btrfs_should_fragment_free_space( 231 struct btrfs_block_group *block_group) 232 { 233 struct btrfs_fs_info *fs_info = block_group->fs_info; 234 235 return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) && 236 block_group->flags & BTRFS_BLOCK_GROUP_METADATA) || 237 (btrfs_test_opt(fs_info, FRAGMENT_DATA) && 238 block_group->flags & BTRFS_BLOCK_GROUP_DATA); 239 } 240 #endif 241 242 struct btrfs_block_group *btrfs_lookup_first_block_group( 243 struct btrfs_fs_info *info, u64 bytenr); 244 struct btrfs_block_group *btrfs_lookup_block_group( 245 struct btrfs_fs_info *info, u64 bytenr); 246 struct btrfs_block_group *btrfs_next_block_group( 247 struct btrfs_block_group *cache); 248 void btrfs_get_block_group(struct btrfs_block_group *cache); 249 void btrfs_put_block_group(struct btrfs_block_group *cache); 250 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info, 251 const u64 start); 252 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg); 253 bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr); 254 void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr); 255 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg); 256 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache, 257 u64 num_bytes); 258 int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache); 259 int btrfs_cache_block_group(struct btrfs_block_group *cache, 260 int load_cache_only); 261 void btrfs_put_caching_control(struct btrfs_caching_control *ctl); 262 struct btrfs_caching_control *btrfs_get_caching_control( 263 struct btrfs_block_group *cache); 264 u64 add_new_free_space(struct btrfs_block_group *block_group, 265 u64 start, u64 end); 266 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group( 267 struct btrfs_fs_info *fs_info, 268 const u64 chunk_offset); 269 int btrfs_remove_block_group(struct btrfs_trans_handle *trans, 270 u64 group_start, struct extent_map *em); 271 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info); 272 void btrfs_mark_bg_unused(struct btrfs_block_group *bg); 273 void btrfs_reclaim_bgs_work(struct work_struct *work); 274 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info); 275 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg); 276 int btrfs_read_block_groups(struct btrfs_fs_info *info); 277 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans, 278 u64 bytes_used, u64 type, 279 u64 chunk_offset, u64 size); 280 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans); 281 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache, 282 bool do_chunk_alloc); 283 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache); 284 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans); 285 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans); 286 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans); 287 int btrfs_update_block_group(struct btrfs_trans_handle *trans, 288 u64 bytenr, u64 num_bytes, bool alloc); 289 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache, 290 u64 ram_bytes, u64 num_bytes, int delalloc); 291 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache, 292 u64 num_bytes, int delalloc); 293 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags, 294 enum btrfs_chunk_alloc_enum force); 295 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type); 296 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type); 297 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans, 298 bool is_item_insertion); 299 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags); 300 void btrfs_put_block_group_cache(struct btrfs_fs_info *info); 301 int btrfs_free_block_groups(struct btrfs_fs_info *info); 302 void btrfs_wait_space_cache_v1_finished(struct btrfs_block_group *cache, 303 struct btrfs_caching_control *caching_ctl); 304 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start, 305 struct block_device *bdev, u64 physical, u64 **logical, 306 int *naddrs, int *stripe_len); 307 308 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info) 309 { 310 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA); 311 } 312 313 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info) 314 { 315 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA); 316 } 317 318 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info) 319 { 320 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM); 321 } 322 323 static inline int btrfs_block_group_done(struct btrfs_block_group *cache) 324 { 325 smp_mb(); 326 return cache->cached == BTRFS_CACHE_FINISHED || 327 cache->cached == BTRFS_CACHE_ERROR; 328 } 329 330 void btrfs_freeze_block_group(struct btrfs_block_group *cache); 331 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache); 332 333 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg); 334 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount); 335 336 #endif /* BTRFS_BLOCK_GROUP_H */ 337