fs/btrfs/extent_io.h

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