| 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
| 2 | |
| 3 | #ifndef BTRFS_EXTENT_IO_TREE_H |
| 4 | #define BTRFS_EXTENT_IO_TREE_H |
| 5 | |
| 6 | #include <linux/rbtree.h> |
| 7 | #include <linux/spinlock.h> |
| 8 | #include <linux/refcount.h> |
| 9 | #include <linux/list.h> |
| 10 | #include <linux/wait.h> |
| 11 | #include "misc.h" |
| 12 | |
| 13 | struct extent_changeset; |
| 14 | struct btrfs_fs_info; |
| 15 | struct btrfs_inode; |
| 16 | |
| 17 | /* Bits for the extent state */ |
| 18 | enum { |
| 19 | ENUM_BIT(EXTENT_DIRTY), |
| 20 | ENUM_BIT(EXTENT_LOCKED), |
| 21 | ENUM_BIT(EXTENT_DIO_LOCKED), |
| 22 | ENUM_BIT(EXTENT_DIRTY_LOG1), |
| 23 | ENUM_BIT(EXTENT_DIRTY_LOG2), |
| 24 | ENUM_BIT(EXTENT_DELALLOC), |
| 25 | ENUM_BIT(EXTENT_DEFRAG), |
| 26 | ENUM_BIT(EXTENT_BOUNDARY), |
| 27 | ENUM_BIT(EXTENT_NODATASUM), |
| 28 | ENUM_BIT(EXTENT_CLEAR_META_RESV), |
| 29 | ENUM_BIT(EXTENT_NEED_WAIT), |
| 30 | ENUM_BIT(EXTENT_NORESERVE), |
| 31 | ENUM_BIT(EXTENT_QGROUP_RESERVED), |
| 32 | ENUM_BIT(EXTENT_CLEAR_DATA_RESV), |
| 33 | /* |
| 34 | * Must be cleared only during ordered extent completion or on error |
| 35 | * paths if we did not manage to submit bios and create the ordered |
| 36 | * extents for the range. Should not be cleared during page release |
| 37 | * and page invalidation (if there is an ordered extent in flight), |
| 38 | * that is left for the ordered extent completion. |
| 39 | */ |
| 40 | ENUM_BIT(EXTENT_DELALLOC_NEW), |
| 41 | /* |
| 42 | * Mark that a range is being locked for finishing an ordered extent. |
| 43 | * Used together with EXTENT_LOCKED. |
| 44 | */ |
| 45 | ENUM_BIT(EXTENT_FINISHING_ORDERED), |
| 46 | /* |
| 47 | * When an ordered extent successfully completes for a region marked as |
| 48 | * a new delalloc range, use this flag when clearing a new delalloc |
| 49 | * range to indicate that the VFS' inode number of bytes should be |
| 50 | * incremented and the inode's new delalloc bytes decremented, in an |
| 51 | * atomic way to prevent races with stat(2). |
| 52 | */ |
| 53 | ENUM_BIT(EXTENT_ADD_INODE_BYTES), |
| 54 | /* |
| 55 | * Set during truncate when we're clearing an entire range and we just |
| 56 | * want the extent states to go away. |
| 57 | */ |
| 58 | ENUM_BIT(EXTENT_CLEAR_ALL_BITS), |
| 59 | |
| 60 | /* |
| 61 | * This must be last. |
| 62 | * |
| 63 | * Bit not representing a state but a request for NOWAIT semantics, |
| 64 | * e.g. when allocating memory, and must be masked out from the other |
| 65 | * bits. |
| 66 | */ |
| 67 | ENUM_BIT(EXTENT_NOWAIT) |
| 68 | }; |
| 69 | |
| 70 | #define EXTENT_DO_ACCOUNTING (EXTENT_CLEAR_META_RESV | \ |
| 71 | EXTENT_CLEAR_DATA_RESV) |
| 72 | #define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | \ |
| 73 | EXTENT_ADD_INODE_BYTES | \ |
| 74 | EXTENT_CLEAR_ALL_BITS) |
| 75 | |
| 76 | #define EXTENT_LOCK_BITS (EXTENT_LOCKED | EXTENT_DIO_LOCKED) |
| 77 | |
| 78 | /* |
| 79 | * Redefined bits above which are used only in the device allocation tree, |
| 80 | * shouldn't be using EXTENT_LOCKED / EXTENT_BOUNDARY / EXTENT_CLEAR_META_RESV |
| 81 | * / EXTENT_CLEAR_DATA_RESV because they have special meaning to the bit |
| 82 | * manipulation functions |
| 83 | */ |
| 84 | #define CHUNK_ALLOCATED EXTENT_DIRTY |
| 85 | #define CHUNK_TRIMMED EXTENT_DEFRAG |
| 86 | #define CHUNK_STATE_MASK (CHUNK_ALLOCATED | \ |
| 87 | CHUNK_TRIMMED) |
| 88 | |
| 89 | enum { |
| 90 | IO_TREE_FS_PINNED_EXTENTS, |
| 91 | IO_TREE_FS_EXCLUDED_EXTENTS, |
| 92 | IO_TREE_BTREE_INODE_IO, |
| 93 | IO_TREE_INODE_IO, |
| 94 | IO_TREE_RELOC_BLOCKS, |
| 95 | IO_TREE_TRANS_DIRTY_PAGES, |
| 96 | IO_TREE_ROOT_DIRTY_LOG_PAGES, |
| 97 | IO_TREE_INODE_FILE_EXTENT, |
| 98 | IO_TREE_LOG_CSUM_RANGE, |
| 99 | IO_TREE_SELFTEST, |
| 100 | IO_TREE_DEVICE_ALLOC_STATE, |
| 101 | }; |
| 102 | |
| 103 | struct extent_io_tree { |
| 104 | struct rb_root state; |
| 105 | /* |
| 106 | * The fs_info is needed for trace points, a tree attached to an inode |
| 107 | * needs the inode. |
| 108 | * |
| 109 | * owner == IO_TREE_INODE_IO - then inode is valid and fs_info can be |
| 110 | * accessed as inode->root->fs_info |
| 111 | */ |
| 112 | union { |
| 113 | struct btrfs_fs_info *fs_info; |
| 114 | struct btrfs_inode *inode; |
| 115 | }; |
| 116 | |
| 117 | /* Who owns this io tree, should be one of IO_TREE_* */ |
| 118 | u8 owner; |
| 119 | |
| 120 | spinlock_t lock; |
| 121 | }; |
| 122 | |
| 123 | struct extent_state { |
| 124 | u64 start; |
| 125 | u64 end; /* inclusive */ |
| 126 | struct rb_node rb_node; |
| 127 | |
| 128 | /* ADD NEW ELEMENTS AFTER THIS */ |
| 129 | wait_queue_head_t wq; |
| 130 | refcount_t refs; |
| 131 | u32 state; |
| 132 | |
| 133 | #ifdef CONFIG_BTRFS_DEBUG |
| 134 | struct list_head leak_list; |
| 135 | #endif |
| 136 | }; |
| 137 | |
| 138 | const struct btrfs_inode *btrfs_extent_io_tree_to_inode(const struct extent_io_tree *tree); |
| 139 | const struct btrfs_fs_info *btrfs_extent_io_tree_to_fs_info(const struct extent_io_tree *tree); |
| 140 | |
| 141 | void btrfs_extent_io_tree_init(struct btrfs_fs_info *fs_info, |
| 142 | struct extent_io_tree *tree, unsigned int owner); |
| 143 | void btrfs_extent_io_tree_release(struct extent_io_tree *tree); |
| 144 | int btrfs_lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, u32 bits, |
| 145 | struct extent_state **cached); |
| 146 | bool btrfs_try_lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, |
| 147 | u32 bits, struct extent_state **cached); |
| 148 | |
| 149 | static inline int btrfs_lock_extent(struct extent_io_tree *tree, u64 start, u64 end, |
| 150 | struct extent_state **cached) |
| 151 | { |
| 152 | return btrfs_lock_extent_bits(tree, start, end, bits: EXTENT_LOCKED, cached); |
| 153 | } |
| 154 | |
| 155 | static inline bool btrfs_try_lock_extent(struct extent_io_tree *tree, u64 start, |
| 156 | u64 end, struct extent_state **cached) |
| 157 | { |
| 158 | return btrfs_try_lock_extent_bits(tree, start, end, bits: EXTENT_LOCKED, cached); |
| 159 | } |
| 160 | |
| 161 | int __init btrfs_extent_state_init_cachep(void); |
| 162 | void __cold btrfs_extent_state_free_cachep(void); |
| 163 | |
| 164 | u64 btrfs_count_range_bits(struct extent_io_tree *tree, |
| 165 | u64 *start, u64 search_end, |
| 166 | u64 max_bytes, u32 bits, bool contig, |
| 167 | struct extent_state **cached_state); |
| 168 | |
| 169 | void btrfs_free_extent_state(struct extent_state *state); |
| 170 | bool btrfs_test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, u32 bit, |
| 171 | struct extent_state *cached_state); |
| 172 | bool btrfs_test_range_bit_exists(struct extent_io_tree *tree, u64 start, u64 end, u32 bit); |
| 173 | void btrfs_get_range_bits(struct extent_io_tree *tree, u64 start, u64 end, u32 *bits, |
| 174 | struct extent_state **cached_state); |
| 175 | int btrfs_clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, |
| 176 | u32 bits, struct extent_changeset *changeset); |
| 177 | int btrfs_clear_extent_bit_changeset(struct extent_io_tree *tree, u64 start, u64 end, |
| 178 | u32 bits, struct extent_state **cached, |
| 179 | struct extent_changeset *changeset); |
| 180 | |
| 181 | static inline int btrfs_clear_extent_bit(struct extent_io_tree *tree, u64 start, |
| 182 | u64 end, u32 bits, |
| 183 | struct extent_state **cached) |
| 184 | { |
| 185 | return btrfs_clear_extent_bit_changeset(tree, start, end, bits, cached, NULL); |
| 186 | } |
| 187 | |
| 188 | static inline int btrfs_unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, |
| 189 | struct extent_state **cached) |
| 190 | { |
| 191 | return btrfs_clear_extent_bit_changeset(tree, start, end, bits: EXTENT_LOCKED, |
| 192 | cached, NULL); |
| 193 | } |
| 194 | |
| 195 | int btrfs_set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, |
| 196 | u32 bits, struct extent_changeset *changeset); |
| 197 | int btrfs_set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, |
| 198 | u32 bits, struct extent_state **cached_state); |
| 199 | |
| 200 | static inline int btrfs_clear_extent_dirty(struct extent_io_tree *tree, u64 start, |
| 201 | u64 end, struct extent_state **cached) |
| 202 | { |
| 203 | return btrfs_clear_extent_bit(tree, start, end, |
| 204 | bits: EXTENT_DIRTY | EXTENT_DELALLOC | |
| 205 | EXTENT_DO_ACCOUNTING, cached); |
| 206 | } |
| 207 | |
| 208 | int btrfs_convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, |
| 209 | u32 bits, u32 clear_bits, |
| 210 | struct extent_state **cached_state); |
| 211 | |
| 212 | bool btrfs_find_first_extent_bit(struct extent_io_tree *tree, u64 start, |
| 213 | u64 *start_ret, u64 *end_ret, u32 bits, |
| 214 | struct extent_state **cached_state); |
| 215 | void btrfs_find_first_clear_extent_bit(struct extent_io_tree *tree, u64 start, |
| 216 | u64 *start_ret, u64 *end_ret, u32 bits); |
| 217 | bool btrfs_find_contiguous_extent_bit(struct extent_io_tree *tree, u64 start, |
| 218 | u64 *start_ret, u64 *end_ret, u32 bits); |
| 219 | bool btrfs_find_delalloc_range(struct extent_io_tree *tree, u64 *start, |
| 220 | u64 *end, u64 max_bytes, |
| 221 | struct extent_state **cached_state); |
| 222 | static inline int btrfs_lock_dio_extent(struct extent_io_tree *tree, u64 start, |
| 223 | u64 end, struct extent_state **cached) |
| 224 | { |
| 225 | return btrfs_lock_extent_bits(tree, start, end, bits: EXTENT_DIO_LOCKED, cached); |
| 226 | } |
| 227 | |
| 228 | static inline bool btrfs_try_lock_dio_extent(struct extent_io_tree *tree, u64 start, |
| 229 | u64 end, struct extent_state **cached) |
| 230 | { |
| 231 | return btrfs_try_lock_extent_bits(tree, start, end, bits: EXTENT_DIO_LOCKED, cached); |
| 232 | } |
| 233 | |
| 234 | static inline int btrfs_unlock_dio_extent(struct extent_io_tree *tree, u64 start, |
| 235 | u64 end, struct extent_state **cached) |
| 236 | { |
| 237 | return btrfs_clear_extent_bit_changeset(tree, start, end, bits: EXTENT_DIO_LOCKED, |
| 238 | cached, NULL); |
| 239 | } |
| 240 | |
| 241 | struct extent_state *btrfs_next_extent_state(struct extent_io_tree *tree, |
| 242 | struct extent_state *state); |
| 243 | |
| 244 | #endif /* BTRFS_EXTENT_IO_TREE_H */ |
| 245 | |