fs/erofs/internal.h

   1 /* SPDX-License-Identifier: GPL-2.0-only */
   2 /*
   3  * Copyright (C) 2017-2018 HUAWEI, Inc.
   4  *             https://www.huawei.com/
   5  * Copyright (C) 2021, Alibaba Cloud
   6  */
   7 #ifndef __EROFS_INTERNAL_H
   8 #define __EROFS_INTERNAL_H
   9
  10 #include <linux/fs.h>
  11 #include <linux/dax.h>
  12 #include <linux/dcache.h>
  13 #include <linux/mm.h>
  14 #include <linux/module.h>
  15 #include <linux/pagemap.h>
  16 #include <linux/bio.h>
  17 #include <linux/magic.h>
  18 #include <linux/slab.h>
  19 #include <linux/vmalloc.h>
  20 #include <linux/iomap.h>
  21 #include "erofs_fs.h"
  22
  23 __printf(2, 3) void _erofs_printk(struct super_block *sb, const char *fmt, ...);
  24 #define erofs_err(sb, fmt, ...) \
  25         _erofs_printk(sb, KERN_ERR fmt "\n", ##__VA_ARGS__)
  26 #define erofs_info(sb, fmt, ...) \
  27         _erofs_printk(sb, KERN_INFO fmt "\n", ##__VA_ARGS__)
  28
  29 #ifdef CONFIG_EROFS_FS_DEBUG
  30 #define DBG_BUGON               BUG_ON
  31 #else
  32 #define DBG_BUGON(x)            ((void)(x))
  33 #endif  /* !CONFIG_EROFS_FS_DEBUG */
  34
  35 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */
  36 #define EROFS_SUPER_MAGIC   EROFS_SUPER_MAGIC_V1
  37
  38 typedef u64 erofs_nid_t;
  39 typedef u64 erofs_off_t;
  40 /* data type for filesystem-wide blocks number */
  41 typedef u32 erofs_blk_t;
  42
  43 struct erofs_device_info {
  44         char *path;
  45         struct erofs_fscache *fscache;
  46         struct file *file;
  47         struct dax_device *dax_dev;
  48         u64 dax_part_off;
  49
  50         u32 blocks;
  51         u32 mapped_blkaddr;
  52 };
  53
  54 enum {
  55         EROFS_SYNC_DECOMPRESS_AUTO,
  56         EROFS_SYNC_DECOMPRESS_FORCE_ON,
  57         EROFS_SYNC_DECOMPRESS_FORCE_OFF
  58 };
  59
  60 struct erofs_mount_opts {
  61         /* current strategy of how to use managed cache */
  62         unsigned char cache_strategy;
  63         /* strategy of sync decompression (0 - auto, 1 - force on, 2 - force off) */
  64         unsigned int sync_decompress;
  65         /* threshold for decompression synchronously */
  66         unsigned int max_sync_decompress_pages;
  67         unsigned int mount_opt;
  68 };
  69
  70 struct erofs_dev_context {
  71         struct idr tree;
  72         struct rw_semaphore rwsem;
  73
  74         unsigned int extra_devices;
  75         bool flatdev;
  76 };
  77
  78 /* all filesystem-wide lz4 configurations */
  79 struct erofs_sb_lz4_info {
  80         /* # of pages needed for EROFS lz4 rolling decompression */
  81         u16 max_distance_pages;
  82         /* maximum possible blocks for pclusters in the filesystem */
  83         u16 max_pclusterblks;
  84 };
  85
  86 struct erofs_domain {
  87         refcount_t ref;
  88         struct list_head list;
  89         struct fscache_volume *volume;
  90         char *domain_id;
  91 };
  92
  93 struct erofs_fscache {
  94         struct fscache_cookie *cookie;
  95         struct inode *inode;    /* anonymous inode for the blob */
  96
  97         /* used for share domain mode */
  98         struct erofs_domain *domain;
  99         struct list_head node;
 100         refcount_t ref;
 101         char *name;
 102 };
 103
 104 struct erofs_xattr_prefix_item {
 105         struct erofs_xattr_long_prefix *prefix;
 106         u8 infix_len;
 107 };
 108
 109 struct erofs_sb_info {
 110         struct erofs_device_info dif0;
 111         struct erofs_mount_opts opt;    /* options */
 112 #ifdef CONFIG_EROFS_FS_ZIP
 113         /* list for all registered superblocks, mainly for shrinker */
 114         struct list_head list;
 115         struct mutex umount_mutex;
 116
 117         /* managed XArray arranged in physical block number */
 118         struct xarray managed_pslots;
 119
 120         unsigned int shrinker_run_no;
 121         u16 available_compr_algs;
 122
 123         /* pseudo inode to manage cached pages */
 124         struct inode *managed_cache;
 125
 126         struct erofs_sb_lz4_info lz4;
 127 #endif  /* CONFIG_EROFS_FS_ZIP */
 128         struct inode *packed_inode;
 129         struct erofs_dev_context *devs;
 130         u64 total_blocks;
 131
 132         u32 meta_blkaddr;
 133 #ifdef CONFIG_EROFS_FS_XATTR
 134         u32 xattr_blkaddr;
 135         u32 xattr_prefix_start;
 136         u8 xattr_prefix_count;
 137         struct erofs_xattr_prefix_item *xattr_prefixes;
 138         unsigned int xattr_filter_reserved;
 139 #endif
 140         u16 device_id_mask;     /* valid bits of device id to be used */
 141
 142         unsigned char islotbits;        /* inode slot unit size in bit shift */
 143         unsigned char blkszbits;        /* filesystem block size in bit shift */
 144
 145         u32 sb_size;                    /* total superblock size */
 146         u32 build_time_nsec;
 147         u64 build_time;
 148
 149         /* what we really care is nid, rather than ino.. */
 150         erofs_nid_t root_nid;
 151         erofs_nid_t packed_nid;
 152         /* used for statfs, f_files - f_favail */
 153         u64 inos;
 154
 155         u8 uuid[16];                    /* 128-bit uuid for volume */
 156         u8 volume_name[16];             /* volume name */
 157         u32 feature_compat;
 158         u32 feature_incompat;
 159
 160         /* sysfs support */
 161         struct kobject s_kobj;          /* /sys/fs/erofs/<devname> */
 162         struct completion s_kobj_unregister;
 163
 164         /* fscache support */
 165         struct fscache_volume *volume;
 166         struct erofs_domain *domain;
 167         char *fsid;
 168         char *domain_id;
 169 };
 170
 171 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
 172 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
 173
 174 /* Mount flags set via mount options or defaults */
 175 #define EROFS_MOUNT_XATTR_USER          0x00000010
 176 #define EROFS_MOUNT_POSIX_ACL           0x00000020
 177 #define EROFS_MOUNT_DAX_ALWAYS          0x00000040
 178 #define EROFS_MOUNT_DAX_NEVER           0x00000080
 179 #define EROFS_MOUNT_DIRECT_IO           0x00000100
 180
 181 #define clear_opt(opt, option)  ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
 182 #define set_opt(opt, option)    ((opt)->mount_opt |= EROFS_MOUNT_##option)
 183 #define test_opt(opt, option)   ((opt)->mount_opt & EROFS_MOUNT_##option)
 184
 185 static inline bool erofs_is_fileio_mode(struct erofs_sb_info *sbi)
 186 {
 187         return IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) && sbi->dif0.file;
 188 }
 189
 190 static inline bool erofs_is_fscache_mode(struct super_block *sb)
 191 {
 192         return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) &&
 193                         !erofs_is_fileio_mode(EROFS_SB(sb)) && !sb->s_bdev;
 194 }
 195
 196 enum {
 197         EROFS_ZIP_CACHE_DISABLED,
 198         EROFS_ZIP_CACHE_READAHEAD,
 199         EROFS_ZIP_CACHE_READAROUND
 200 };
 201
 202 enum erofs_kmap_type {
 203         EROFS_NO_KMAP,          /* don't map the buffer */
 204         EROFS_KMAP,             /* use kmap_local_page() to map the buffer */
 205 };
 206
 207 struct erofs_buf {
 208         struct address_space *mapping;
 209         struct file *file;
 210         struct page *page;
 211         void *base;
 212 };
 213 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL })
 214
 215 #define erofs_blknr(sb, addr)   ((erofs_blk_t)((addr) >> (sb)->s_blocksize_bits))
 216 #define erofs_blkoff(sb, addr)  ((addr) & ((sb)->s_blocksize - 1))
 217 #define erofs_pos(sb, blk)      ((erofs_off_t)(blk) << (sb)->s_blocksize_bits)
 218 #define erofs_iblks(i)  (round_up((i)->i_size, i_blocksize(i)) >> (i)->i_blkbits)
 219
 220 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
 221 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
 222 { \
 223         return sbi->feature_##compat & EROFS_FEATURE_##feature; \
 224 }
 225
 226 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
 227 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
 228 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
 229 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
 230 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
 231 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
 232 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
 233 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
 234 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
 235 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES)
 236 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
 237 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
 238
 239 /* atomic flag definitions */
 240 #define EROFS_I_EA_INITED_BIT   0
 241 #define EROFS_I_Z_INITED_BIT    1
 242
 243 /* bitlock definitions (arranged in reverse order) */
 244 #define EROFS_I_BL_XATTR_BIT    (BITS_PER_LONG - 1)
 245 #define EROFS_I_BL_Z_BIT        (BITS_PER_LONG - 2)
 246
 247 struct erofs_inode {
 248         erofs_nid_t nid;
 249
 250         /* atomic flags (including bitlocks) */
 251         unsigned long flags;
 252
 253         unsigned char datalayout;
 254         unsigned char inode_isize;
 255         unsigned int xattr_isize;
 256
 257         unsigned int xattr_name_filter;
 258         unsigned int xattr_shared_count;
 259         unsigned int *xattr_shared_xattrs;
 260
 261         union {
 262                 erofs_blk_t raw_blkaddr;
 263                 struct {
 264                         unsigned short  chunkformat;
 265                         unsigned char   chunkbits;
 266                 };
 267 #ifdef CONFIG_EROFS_FS_ZIP
 268                 struct {
 269                         unsigned short z_advise;
 270                         unsigned char  z_algorithmtype[2];
 271                         unsigned char  z_logical_clusterbits;
 272                         unsigned long  z_tailextent_headlcn;
 273                         union {
 274                                 struct {
 275                                         erofs_off_t    z_idataoff;
 276                                         unsigned short z_idata_size;
 277                                 };
 278                                 erofs_off_t z_fragmentoff;
 279                         };
 280                 };
 281 #endif  /* CONFIG_EROFS_FS_ZIP */
 282         };
 283         /* the corresponding vfs inode */
 284         struct inode vfs_inode;
 285 };
 286
 287 #define EROFS_I(ptr)    container_of(ptr, struct erofs_inode, vfs_inode)
 288
 289 static inline erofs_off_t erofs_iloc(struct inode *inode)
 290 {
 291         struct erofs_sb_info *sbi = EROFS_I_SB(inode);
 292
 293         return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
 294                 (EROFS_I(inode)->nid << sbi->islotbits);
 295 }
 296
 297 static inline unsigned int erofs_inode_version(unsigned int ifmt)
 298 {
 299         return (ifmt >> EROFS_I_VERSION_BIT) & EROFS_I_VERSION_MASK;
 300 }
 301
 302 static inline unsigned int erofs_inode_datalayout(unsigned int ifmt)
 303 {
 304         return (ifmt >> EROFS_I_DATALAYOUT_BIT) & EROFS_I_DATALAYOUT_MASK;
 305 }
 306
 307 /* reclaiming is never triggered when allocating new folios. */
 308 static inline struct folio *erofs_grab_folio_nowait(struct address_space *as,
 309                                                     pgoff_t index)
 310 {
 311         return __filemap_get_folio(as, index,
 312                         FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
 313                         readahead_gfp_mask(as) & ~__GFP_RECLAIM);
 314 }
 315
 316 /* Has a disk mapping */
 317 #define EROFS_MAP_MAPPED        0x0001
 318 /* Located in metadata (could be copied from bd_inode) */
 319 #define EROFS_MAP_META          0x0002
 320 /* The extent is encoded */
 321 #define EROFS_MAP_ENCODED       0x0004
 322 /* The length of extent is full */
 323 #define EROFS_MAP_FULL_MAPPED   0x0008
 324 /* Located in the special packed inode */
 325 #define EROFS_MAP_FRAGMENT      0x0010
 326 /* The extent refers to partial decompressed data */
 327 #define EROFS_MAP_PARTIAL_REF   0x0020
 328
 329 struct erofs_map_blocks {
 330         struct erofs_buf buf;
 331
 332         erofs_off_t m_pa, m_la;
 333         u64 m_plen, m_llen;
 334
 335         unsigned short m_deviceid;
 336         char m_algorithmformat;
 337         unsigned int m_flags;
 338 };
 339
 340 /*
 341  * Used to get the exact decompressed length, e.g. fiemap (consider lookback
 342  * approach instead if possible since it's more metadata lightweight.)
 343  */
 344 #define EROFS_GET_BLOCKS_FIEMAP         0x0001
 345 /* Used to map the whole extent if non-negligible data is requested for LZMA */
 346 #define EROFS_GET_BLOCKS_READMORE       0x0002
 347 /* Used to map tail extent for tailpacking inline or fragment pcluster */
 348 #define EROFS_GET_BLOCKS_FINDTAIL       0x0004
 349
 350 enum {
 351         Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
 352         Z_EROFS_COMPRESSION_INTERLACED,
 353         Z_EROFS_COMPRESSION_RUNTIME_MAX
 354 };
 355
 356 struct erofs_map_dev {
 357         struct super_block *m_sb;
 358         struct erofs_device_info *m_dif;
 359         struct block_device *m_bdev;
 360
 361         erofs_off_t m_pa;
 362         unsigned int m_deviceid;
 363 };
 364
 365 extern const struct super_operations erofs_sops;
 366
 367 extern const struct address_space_operations erofs_aops;
 368 extern const struct address_space_operations erofs_fileio_aops;
 369 extern const struct address_space_operations z_erofs_aops;
 370 extern const struct address_space_operations erofs_fscache_access_aops;
 371
 372 extern const struct inode_operations erofs_generic_iops;
 373 extern const struct inode_operations erofs_symlink_iops;
 374 extern const struct inode_operations erofs_fast_symlink_iops;
 375 extern const struct inode_operations erofs_dir_iops;
 376
 377 extern const struct file_operations erofs_file_fops;
 378 extern const struct file_operations erofs_dir_fops;
 379
 380 extern const struct iomap_ops z_erofs_iomap_report_ops;
 381
 382 /* flags for erofs_fscache_register_cookie() */
 383 #define EROFS_REG_COOKIE_SHARE          0x0001
 384 #define EROFS_REG_COOKIE_NEED_NOEXIST   0x0002
 385
 386 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
 387                           erofs_off_t *offset, int *lengthp);
 388 void erofs_unmap_metabuf(struct erofs_buf *buf);
 389 void erofs_put_metabuf(struct erofs_buf *buf);
 390 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset,
 391                   enum erofs_kmap_type type);
 392 void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
 393 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
 394                          erofs_off_t offset, enum erofs_kmap_type type);
 395 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
 396 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 397                  u64 start, u64 len);
 398 int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map);
 399 void erofs_onlinefolio_init(struct folio *folio);
 400 void erofs_onlinefolio_split(struct folio *folio);
 401 void erofs_onlinefolio_end(struct folio *folio, int err);
 402 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
 403 int erofs_getattr(struct mnt_idmap *idmap, const struct path *path,
 404                   struct kstat *stat, u32 request_mask,
 405                   unsigned int query_flags);
 406 int erofs_namei(struct inode *dir, const struct qstr *name,
 407                 erofs_nid_t *nid, unsigned int *d_type);
 408
 409 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
 410 {
 411         int retried = 0;
 412
 413         while (1) {
 414                 void *p = vm_map_ram(pages, count, -1);
 415
 416                 /* retry two more times (totally 3 times) */
 417                 if (p || ++retried >= 3)
 418                         return p;
 419                 vm_unmap_aliases();
 420         }
 421         return NULL;
 422 }
 423
 424 int erofs_register_sysfs(struct super_block *sb);
 425 void erofs_unregister_sysfs(struct super_block *sb);
 426 int __init erofs_init_sysfs(void);
 427 void erofs_exit_sysfs(void);
 428
 429 struct page *__erofs_allocpage(struct page **pagepool, gfp_t gfp, bool tryrsv);
 430 static inline struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
 431 {
 432         return __erofs_allocpage(pagepool, gfp, false);
 433 }
 434 static inline void erofs_pagepool_add(struct page **pagepool, struct page *page)
 435 {
 436         set_page_private(page, (unsigned long)*pagepool);
 437         *pagepool = page;
 438 }
 439 void erofs_release_pages(struct page **pagepool);
 440
 441 #ifdef CONFIG_EROFS_FS_ZIP
 442 #define MNGD_MAPPING(sbi)       ((sbi)->managed_cache->i_mapping)
 443
 444 extern atomic_long_t erofs_global_shrink_cnt;
 445 void erofs_shrinker_register(struct super_block *sb);
 446 void erofs_shrinker_unregister(struct super_block *sb);
 447 int __init erofs_init_shrinker(void);
 448 void erofs_exit_shrinker(void);
 449 int __init z_erofs_init_subsystem(void);
 450 void z_erofs_exit_subsystem(void);
 451 unsigned long z_erofs_shrink_scan(struct erofs_sb_info *sbi,
 452                                   unsigned long nr_shrink);
 453 int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map,
 454                             int flags);
 455 void *z_erofs_get_gbuf(unsigned int requiredpages);
 456 void z_erofs_put_gbuf(void *ptr);
 457 int z_erofs_gbuf_growsize(unsigned int nrpages);
 458 int __init z_erofs_gbuf_init(void);
 459 void z_erofs_gbuf_exit(void);
 460 int erofs_init_managed_cache(struct super_block *sb);
 461 int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb);
 462 #else
 463 static inline void erofs_shrinker_register(struct super_block *sb) {}
 464 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
 465 static inline int erofs_init_shrinker(void) { return 0; }
 466 static inline void erofs_exit_shrinker(void) {}
 467 static inline int z_erofs_init_subsystem(void) { return 0; }
 468 static inline void z_erofs_exit_subsystem(void) {}
 469 static inline int erofs_init_managed_cache(struct super_block *sb) { return 0; }
 470 #endif  /* !CONFIG_EROFS_FS_ZIP */
 471
 472 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
 473 struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev);
 474 void erofs_fileio_submit_bio(struct bio *bio);
 475 #else
 476 static inline struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
 477 static inline void erofs_fileio_submit_bio(struct bio *bio) {}
 478 #endif
 479
 480 #ifdef CONFIG_EROFS_FS_ONDEMAND
 481 int erofs_fscache_register_fs(struct super_block *sb);
 482 void erofs_fscache_unregister_fs(struct super_block *sb);
 483
 484 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
 485                                         char *name, unsigned int flags);
 486 void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
 487 struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev);
 488 void erofs_fscache_submit_bio(struct bio *bio);
 489 #else
 490 static inline int erofs_fscache_register_fs(struct super_block *sb)
 491 {
 492         return -EOPNOTSUPP;
 493 }
 494 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
 495
 496 static inline
 497 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
 498                                         char *name, unsigned int flags)
 499 {
 500         return ERR_PTR(-EOPNOTSUPP);
 501 }
 502
 503 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
 504 {
 505 }
 506 static inline struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
 507 static inline void erofs_fscache_submit_bio(struct bio *bio) {}
 508 #endif
 509
 510 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
 511
 512 #endif  /* __EROFS_INTERNAL_H */