1 /* SPDX-License-Identifier: GPL-2.0 */ 2 3 #ifndef BTRFS_EXTENT_IO_H 4 #define BTRFS_EXTENT_IO_H 5 6 #include <linux/rbtree.h> 7 #include <linux/refcount.h> 8 #include <linux/fiemap.h> 9 #include <linux/btrfs_tree.h> 10 #include "ulist.h" 11 12 /* 13 * flags for bio submission. The high bits indicate the compression 14 * type for this bio 15 */ 16 #define EXTENT_BIO_COMPRESSED 1 17 #define EXTENT_BIO_FLAG_SHIFT 16 18 19 enum { 20 EXTENT_BUFFER_UPTODATE, 21 EXTENT_BUFFER_DIRTY, 22 EXTENT_BUFFER_CORRUPT, 23 /* this got triggered by readahead */ 24 EXTENT_BUFFER_READAHEAD, 25 EXTENT_BUFFER_TREE_REF, 26 EXTENT_BUFFER_STALE, 27 EXTENT_BUFFER_WRITEBACK, 28 /* read IO error */ 29 EXTENT_BUFFER_READ_ERR, 30 EXTENT_BUFFER_UNMAPPED, 31 EXTENT_BUFFER_IN_TREE, 32 /* write IO error */ 33 EXTENT_BUFFER_WRITE_ERR, 34 EXTENT_BUFFER_NO_CHECK, 35 }; 36 37 /* these are flags for __process_pages_contig */ 38 #define PAGE_UNLOCK (1 << 0) 39 /* Page starts writeback, clear dirty bit and set writeback bit */ 40 #define PAGE_START_WRITEBACK (1 << 1) 41 #define PAGE_END_WRITEBACK (1 << 2) 42 #define PAGE_SET_PRIVATE2 (1 << 3) 43 #define PAGE_SET_ERROR (1 << 4) 44 #define PAGE_LOCK (1 << 5) 45 46 /* 47 * page->private values. Every page that is controlled by the extent 48 * map has page->private set to one. 49 */ 50 #define EXTENT_PAGE_PRIVATE 1 51 52 /* 53 * The extent buffer bitmap operations are done with byte granularity instead of 54 * word granularity for two reasons: 55 * 1. The bitmaps must be little-endian on disk. 56 * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a 57 * single word in a bitmap may straddle two pages in the extent buffer. 58 */ 59 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE) 60 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1) 61 #define BITMAP_FIRST_BYTE_MASK(start) \ 62 ((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK) 63 #define BITMAP_LAST_BYTE_MASK(nbits) \ 64 (BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1))) 65 66 struct btrfs_root; 67 struct btrfs_inode; 68 struct btrfs_io_bio; 69 struct io_failure_record; 70 struct extent_io_tree; 71 72 typedef blk_status_t (submit_bio_hook_t)(struct inode *inode, struct bio *bio, 73 int mirror_num, 74 unsigned long bio_flags); 75 76 typedef blk_status_t (extent_submit_bio_start_t)(struct inode *inode, 77 struct bio *bio, u64 dio_file_offset); 78 79 #define INLINE_EXTENT_BUFFER_PAGES (BTRFS_MAX_METADATA_BLOCKSIZE / PAGE_SIZE) 80 struct extent_buffer { 81 u64 start; 82 unsigned long len; 83 unsigned long bflags; 84 struct btrfs_fs_info *fs_info; 85 spinlock_t refs_lock; 86 atomic_t refs; 87 atomic_t io_pages; 88 int read_mirror; 89 struct rcu_head rcu_head; 90 pid_t lock_owner; 91 /* >= 0 if eb belongs to a log tree, -1 otherwise */ 92 s8 log_index; 93 94 struct rw_semaphore lock; 95 96 struct page *pages[INLINE_EXTENT_BUFFER_PAGES]; 97 struct list_head release_list; 98 #ifdef CONFIG_BTRFS_DEBUG 99 struct list_head leak_list; 100 #endif 101 }; 102 103 /* 104 * Structure to record how many bytes and which ranges are set/cleared 105 */ 106 struct extent_changeset { 107 /* How many bytes are set/cleared in this operation */ 108 unsigned int bytes_changed; 109 110 /* Changed ranges */ 111 struct ulist range_changed; 112 }; 113 114 static inline void extent_changeset_init(struct extent_changeset *changeset) 115 { 116 changeset->bytes_changed = 0; 117 ulist_init(&changeset->range_changed); 118 } 119 120 static inline struct extent_changeset *extent_changeset_alloc(void) 121 { 122 struct extent_changeset *ret; 123 124 ret = kmalloc(sizeof(*ret), GFP_KERNEL); 125 if (!ret) 126 return NULL; 127 128 extent_changeset_init(ret); 129 return ret; 130 } 131 132 static inline void extent_changeset_release(struct extent_changeset *changeset) 133 { 134 if (!changeset) 135 return; 136 changeset->bytes_changed = 0; 137 ulist_release(&changeset->range_changed); 138 } 139 140 static inline void extent_changeset_free(struct extent_changeset *changeset) 141 { 142 if (!changeset) 143 return; 144 extent_changeset_release(changeset); 145 kfree(changeset); 146 } 147 148 static inline void extent_set_compress_type(unsigned long *bio_flags, 149 int compress_type) 150 { 151 *bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT; 152 } 153 154 static inline int extent_compress_type(unsigned long bio_flags) 155 { 156 return bio_flags >> EXTENT_BIO_FLAG_SHIFT; 157 } 158 159 struct extent_map_tree; 160 161 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode, 162 struct page *page, size_t pg_offset, 163 u64 start, u64 len); 164 165 int try_release_extent_mapping(struct page *page, gfp_t mask); 166 int try_release_extent_buffer(struct page *page); 167 168 int __must_check submit_one_bio(struct bio *bio, int mirror_num, 169 unsigned long bio_flags); 170 int btrfs_do_readpage(struct page *page, struct extent_map **em_cached, 171 struct bio **bio, unsigned long *bio_flags, 172 unsigned int read_flags, u64 *prev_em_start); 173 int extent_write_full_page(struct page *page, struct writeback_control *wbc); 174 int extent_write_locked_range(struct inode *inode, u64 start, u64 end, 175 int mode); 176 int extent_writepages(struct address_space *mapping, 177 struct writeback_control *wbc); 178 int btree_write_cache_pages(struct address_space *mapping, 179 struct writeback_control *wbc); 180 void extent_readahead(struct readahead_control *rac); 181 int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo, 182 u64 start, u64 len); 183 int set_page_extent_mapped(struct page *page); 184 void clear_page_extent_mapped(struct page *page); 185 186 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info, 187 u64 start, u64 owner_root, int level); 188 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info, 189 u64 start, unsigned long len); 190 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info, 191 u64 start); 192 struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src); 193 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info, 194 u64 start); 195 void free_extent_buffer(struct extent_buffer *eb); 196 void free_extent_buffer_stale(struct extent_buffer *eb); 197 #define WAIT_NONE 0 198 #define WAIT_COMPLETE 1 199 #define WAIT_PAGE_LOCK 2 200 int read_extent_buffer_pages(struct extent_buffer *eb, int wait, 201 int mirror_num); 202 void wait_on_extent_buffer_writeback(struct extent_buffer *eb); 203 void btrfs_readahead_tree_block(struct btrfs_fs_info *fs_info, 204 u64 bytenr, u64 owner_root, u64 gen, int level); 205 void btrfs_readahead_node_child(struct extent_buffer *node, int slot); 206 207 static inline int num_extent_pages(const struct extent_buffer *eb) 208 { 209 /* 210 * For sectorsize == PAGE_SIZE case, since nodesize is always aligned to 211 * sectorsize, it's just eb->len >> PAGE_SHIFT. 212 * 213 * For sectorsize < PAGE_SIZE case, we could have nodesize < PAGE_SIZE, 214 * thus have to ensure we get at least one page. 215 */ 216 return (eb->len >> PAGE_SHIFT) ?: 1; 217 } 218 219 static inline int extent_buffer_uptodate(const struct extent_buffer *eb) 220 { 221 return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); 222 } 223 224 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv, 225 unsigned long start, unsigned long len); 226 void read_extent_buffer(const struct extent_buffer *eb, void *dst, 227 unsigned long start, 228 unsigned long len); 229 int read_extent_buffer_to_user_nofault(const struct extent_buffer *eb, 230 void __user *dst, unsigned long start, 231 unsigned long len); 232 void write_extent_buffer_fsid(const struct extent_buffer *eb, const void *src); 233 void write_extent_buffer_chunk_tree_uuid(const struct extent_buffer *eb, 234 const void *src); 235 void write_extent_buffer(const struct extent_buffer *eb, const void *src, 236 unsigned long start, unsigned long len); 237 void copy_extent_buffer_full(const struct extent_buffer *dst, 238 const struct extent_buffer *src); 239 void copy_extent_buffer(const struct extent_buffer *dst, 240 const struct extent_buffer *src, 241 unsigned long dst_offset, unsigned long src_offset, 242 unsigned long len); 243 void memcpy_extent_buffer(const struct extent_buffer *dst, 244 unsigned long dst_offset, unsigned long src_offset, 245 unsigned long len); 246 void memmove_extent_buffer(const struct extent_buffer *dst, 247 unsigned long dst_offset, unsigned long src_offset, 248 unsigned long len); 249 void memzero_extent_buffer(const struct extent_buffer *eb, unsigned long start, 250 unsigned long len); 251 int extent_buffer_test_bit(const struct extent_buffer *eb, unsigned long start, 252 unsigned long pos); 253 void extent_buffer_bitmap_set(const struct extent_buffer *eb, unsigned long start, 254 unsigned long pos, unsigned long len); 255 void extent_buffer_bitmap_clear(const struct extent_buffer *eb, 256 unsigned long start, unsigned long pos, 257 unsigned long len); 258 void clear_extent_buffer_dirty(const struct extent_buffer *eb); 259 bool set_extent_buffer_dirty(struct extent_buffer *eb); 260 void set_extent_buffer_uptodate(struct extent_buffer *eb); 261 void clear_extent_buffer_uptodate(struct extent_buffer *eb); 262 int extent_buffer_under_io(const struct extent_buffer *eb); 263 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end); 264 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end); 265 void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end, 266 struct page *locked_page, 267 u32 bits_to_clear, unsigned long page_ops); 268 struct bio *btrfs_bio_alloc(u64 first_byte); 269 struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs); 270 struct bio *btrfs_bio_clone(struct bio *bio); 271 struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size); 272 273 struct btrfs_fs_info; 274 struct btrfs_inode; 275 276 int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start, 277 u64 length, u64 logical, struct page *page, 278 unsigned int pg_offset, int mirror_num); 279 void end_extent_writepage(struct page *page, int err, u64 start, u64 end); 280 int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, int mirror_num); 281 282 /* 283 * When IO fails, either with EIO or csum verification fails, we 284 * try other mirrors that might have a good copy of the data. This 285 * io_failure_record is used to record state as we go through all the 286 * mirrors. If another mirror has good data, the page is set up to date 287 * and things continue. If a good mirror can't be found, the original 288 * bio end_io callback is called to indicate things have failed. 289 */ 290 struct io_failure_record { 291 struct page *page; 292 u64 start; 293 u64 len; 294 u64 logical; 295 unsigned long bio_flags; 296 int this_mirror; 297 int failed_mirror; 298 int in_validation; 299 }; 300 301 302 blk_status_t btrfs_submit_read_repair(struct inode *inode, 303 struct bio *failed_bio, u32 bio_offset, 304 struct page *page, unsigned int pgoff, 305 u64 start, u64 end, int failed_mirror, 306 submit_bio_hook_t *submit_bio_hook); 307 308 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 309 bool find_lock_delalloc_range(struct inode *inode, 310 struct page *locked_page, u64 *start, 311 u64 *end); 312 #endif 313 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info, 314 u64 start); 315 316 #ifdef CONFIG_BTRFS_DEBUG 317 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info); 318 #else 319 #define btrfs_extent_buffer_leak_debug_check(fs_info) do {} while (0) 320 #endif 321 322 #endif 323