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_mount_opts opt;    /* options */
 111 #ifdef CONFIG_EROFS_FS_ZIP
 112         /* list for all registered superblocks, mainly for shrinker */
 113         struct list_head list;
 114         struct mutex umount_mutex;
 115
 116         /* managed XArray arranged in physical block number */
 117         struct xarray managed_pslots;
 118
 119         unsigned int shrinker_run_no;
 120         u16 available_compr_algs;
 121
 122         /* pseudo inode to manage cached pages */
 123         struct inode *managed_cache;
 124
 125         struct erofs_sb_lz4_info lz4;
 126 #endif  /* CONFIG_EROFS_FS_ZIP */
 127         struct file *fdev;
 128         struct inode *packed_inode;
 129         struct erofs_dev_context *devs;
 130         struct dax_device *dax_dev;
 131         u64 dax_part_off;
 132         u64 total_blocks;
 133         u32 primarydevice_blocks;
 134
 135         u32 meta_blkaddr;
 136 #ifdef CONFIG_EROFS_FS_XATTR
 137         u32 xattr_blkaddr;
 138         u32 xattr_prefix_start;
 139         u8 xattr_prefix_count;
 140         struct erofs_xattr_prefix_item *xattr_prefixes;
 141         unsigned int xattr_filter_reserved;
 142 #endif
 143         u16 device_id_mask;     /* valid bits of device id to be used */
 144
 145         unsigned char islotbits;        /* inode slot unit size in bit shift */
 146         unsigned char blkszbits;        /* filesystem block size in bit shift */
 147
 148         u32 sb_size;                    /* total superblock size */
 149         u32 build_time_nsec;
 150         u64 build_time;
 151
 152         /* what we really care is nid, rather than ino.. */
 153         erofs_nid_t root_nid;
 154         erofs_nid_t packed_nid;
 155         /* used for statfs, f_files - f_favail */
 156         u64 inos;
 157
 158         u8 uuid[16];                    /* 128-bit uuid for volume */
 159         u8 volume_name[16];             /* volume name */
 160         u32 feature_compat;
 161         u32 feature_incompat;
 162
 163         /* sysfs support */
 164         struct kobject s_kobj;          /* /sys/fs/erofs/<devname> */
 165         struct completion s_kobj_unregister;
 166
 167         /* fscache support */
 168         struct fscache_volume *volume;
 169         struct erofs_fscache *s_fscache;
 170         struct erofs_domain *domain;
 171         char *fsid;
 172         char *domain_id;
 173 };
 174
 175 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
 176 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
 177
 178 /* Mount flags set via mount options or defaults */
 179 #define EROFS_MOUNT_XATTR_USER          0x00000010
 180 #define EROFS_MOUNT_POSIX_ACL           0x00000020
 181 #define EROFS_MOUNT_DAX_ALWAYS          0x00000040
 182 #define EROFS_MOUNT_DAX_NEVER           0x00000080
 183
 184 #define clear_opt(opt, option)  ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
 185 #define set_opt(opt, option)    ((opt)->mount_opt |= EROFS_MOUNT_##option)
 186 #define test_opt(opt, option)   ((opt)->mount_opt & EROFS_MOUNT_##option)
 187
 188 static inline bool erofs_is_fileio_mode(struct erofs_sb_info *sbi)
 189 {
 190         return IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) && sbi->fdev;
 191 }
 192
 193 static inline bool erofs_is_fscache_mode(struct super_block *sb)
 194 {
 195         return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) &&
 196                         !erofs_is_fileio_mode(EROFS_SB(sb)) && !sb->s_bdev;
 197 }
 198
 199 enum {
 200         EROFS_ZIP_CACHE_DISABLED,
 201         EROFS_ZIP_CACHE_READAHEAD,
 202         EROFS_ZIP_CACHE_READAROUND
 203 };
 204
 205 enum erofs_kmap_type {
 206         EROFS_NO_KMAP,          /* don't map the buffer */
 207         EROFS_KMAP,             /* use kmap_local_page() to map the buffer */
 208 };
 209
 210 struct erofs_buf {
 211         struct address_space *mapping;
 212         struct file *file;
 213         struct page *page;
 214         void *base;
 215 };
 216 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL })
 217
 218 #define erofs_blknr(sb, addr)   ((erofs_blk_t)((addr) >> (sb)->s_blocksize_bits))
 219 #define erofs_blkoff(sb, addr)  ((addr) & ((sb)->s_blocksize - 1))
 220 #define erofs_pos(sb, blk)      ((erofs_off_t)(blk) << (sb)->s_blocksize_bits)
 221 #define erofs_iblks(i)  (round_up((i)->i_size, i_blocksize(i)) >> (i)->i_blkbits)
 222
 223 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
 224 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
 225 { \
 226         return sbi->feature_##compat & EROFS_FEATURE_##feature; \
 227 }
 228
 229 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
 230 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
 231 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
 232 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
 233 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
 234 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
 235 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
 236 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
 237 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
 238 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES)
 239 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
 240 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
 241
 242 /* atomic flag definitions */
 243 #define EROFS_I_EA_INITED_BIT   0
 244 #define EROFS_I_Z_INITED_BIT    1
 245
 246 /* bitlock definitions (arranged in reverse order) */
 247 #define EROFS_I_BL_XATTR_BIT    (BITS_PER_LONG - 1)
 248 #define EROFS_I_BL_Z_BIT        (BITS_PER_LONG - 2)
 249
 250 struct erofs_inode {
 251         erofs_nid_t nid;
 252
 253         /* atomic flags (including bitlocks) */
 254         unsigned long flags;
 255
 256         unsigned char datalayout;
 257         unsigned char inode_isize;
 258         unsigned int xattr_isize;
 259
 260         unsigned int xattr_name_filter;
 261         unsigned int xattr_shared_count;
 262         unsigned int *xattr_shared_xattrs;
 263
 264         union {
 265                 erofs_blk_t raw_blkaddr;
 266                 struct {
 267                         unsigned short  chunkformat;
 268                         unsigned char   chunkbits;
 269                 };
 270 #ifdef CONFIG_EROFS_FS_ZIP
 271                 struct {
 272                         unsigned short z_advise;
 273                         unsigned char  z_algorithmtype[2];
 274                         unsigned char  z_logical_clusterbits;
 275                         unsigned long  z_tailextent_headlcn;
 276                         union {
 277                                 struct {
 278                                         erofs_off_t    z_idataoff;
 279                                         unsigned short z_idata_size;
 280                                 };
 281                                 erofs_off_t z_fragmentoff;
 282                         };
 283                 };
 284 #endif  /* CONFIG_EROFS_FS_ZIP */
 285         };
 286         /* the corresponding vfs inode */
 287         struct inode vfs_inode;
 288 };
 289
 290 #define EROFS_I(ptr)    container_of(ptr, struct erofs_inode, vfs_inode)
 291
 292 static inline erofs_off_t erofs_iloc(struct inode *inode)
 293 {
 294         struct erofs_sb_info *sbi = EROFS_I_SB(inode);
 295
 296         return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
 297                 (EROFS_I(inode)->nid << sbi->islotbits);
 298 }
 299
 300 static inline unsigned int erofs_inode_version(unsigned int ifmt)
 301 {
 302         return (ifmt >> EROFS_I_VERSION_BIT) & EROFS_I_VERSION_MASK;
 303 }
 304
 305 static inline unsigned int erofs_inode_datalayout(unsigned int ifmt)
 306 {
 307         return (ifmt >> EROFS_I_DATALAYOUT_BIT) & EROFS_I_DATALAYOUT_MASK;
 308 }
 309
 310 /* reclaiming is never triggered when allocating new folios. */
 311 static inline struct folio *erofs_grab_folio_nowait(struct address_space *as,
 312                                                     pgoff_t index)
 313 {
 314         return __filemap_get_folio(as, index,
 315                         FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
 316                         readahead_gfp_mask(as) & ~__GFP_RECLAIM);
 317 }
 318
 319 /* Has a disk mapping */
 320 #define EROFS_MAP_MAPPED        0x0001
 321 /* Located in metadata (could be copied from bd_inode) */
 322 #define EROFS_MAP_META          0x0002
 323 /* The extent is encoded */
 324 #define EROFS_MAP_ENCODED       0x0004
 325 /* The length of extent is full */
 326 #define EROFS_MAP_FULL_MAPPED   0x0008
 327 /* Located in the special packed inode */
 328 #define EROFS_MAP_FRAGMENT      0x0010
 329 /* The extent refers to partial decompressed data */
 330 #define EROFS_MAP_PARTIAL_REF   0x0020
 331
 332 struct erofs_map_blocks {
 333         struct erofs_buf buf;
 334
 335         erofs_off_t m_pa, m_la;
 336         u64 m_plen, m_llen;
 337
 338         unsigned short m_deviceid;
 339         char m_algorithmformat;
 340         unsigned int m_flags;
 341 };
 342
 343 /*
 344  * Used to get the exact decompressed length, e.g. fiemap (consider lookback
 345  * approach instead if possible since it's more metadata lightweight.)
 346  */
 347 #define EROFS_GET_BLOCKS_FIEMAP         0x0001
 348 /* Used to map the whole extent if non-negligible data is requested for LZMA */
 349 #define EROFS_GET_BLOCKS_READMORE       0x0002
 350 /* Used to map tail extent for tailpacking inline or fragment pcluster */
 351 #define EROFS_GET_BLOCKS_FINDTAIL       0x0004
 352
 353 enum {
 354         Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
 355         Z_EROFS_COMPRESSION_INTERLACED,
 356         Z_EROFS_COMPRESSION_RUNTIME_MAX
 357 };
 358
 359 struct erofs_map_dev {
 360         struct erofs_fscache *m_fscache;
 361         struct block_device *m_bdev;
 362         struct dax_device *m_daxdev;
 363         struct file *m_fp;
 364         u64 m_dax_part_off;
 365
 366         erofs_off_t m_pa;
 367         unsigned int m_deviceid;
 368 };
 369
 370 extern const struct super_operations erofs_sops;
 371
 372 extern const struct address_space_operations erofs_aops;
 373 extern const struct address_space_operations erofs_fileio_aops;
 374 extern const struct address_space_operations z_erofs_aops;
 375 extern const struct address_space_operations erofs_fscache_access_aops;
 376
 377 extern const struct inode_operations erofs_generic_iops;
 378 extern const struct inode_operations erofs_symlink_iops;
 379 extern const struct inode_operations erofs_fast_symlink_iops;
 380 extern const struct inode_operations erofs_dir_iops;
 381
 382 extern const struct file_operations erofs_file_fops;
 383 extern const struct file_operations erofs_dir_fops;
 384
 385 extern const struct iomap_ops z_erofs_iomap_report_ops;
 386
 387 /* flags for erofs_fscache_register_cookie() */
 388 #define EROFS_REG_COOKIE_SHARE          0x0001
 389 #define EROFS_REG_COOKIE_NEED_NOEXIST   0x0002
 390
 391 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
 392                           erofs_off_t *offset, int *lengthp);
 393 void erofs_unmap_metabuf(struct erofs_buf *buf);
 394 void erofs_put_metabuf(struct erofs_buf *buf);
 395 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset,
 396                   enum erofs_kmap_type type);
 397 void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
 398 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
 399                          erofs_off_t offset, enum erofs_kmap_type type);
 400 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
 401 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 402                  u64 start, u64 len);
 403 int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map);
 404 void erofs_onlinefolio_init(struct folio *folio);
 405 void erofs_onlinefolio_split(struct folio *folio);
 406 void erofs_onlinefolio_end(struct folio *folio, int err);
 407 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
 408 int erofs_getattr(struct mnt_idmap *idmap, const struct path *path,
 409                   struct kstat *stat, u32 request_mask,
 410                   unsigned int query_flags);
 411 int erofs_namei(struct inode *dir, const struct qstr *name,
 412                 erofs_nid_t *nid, unsigned int *d_type);
 413
 414 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
 415 {
 416         int retried = 0;
 417
 418         while (1) {
 419                 void *p = vm_map_ram(pages, count, -1);
 420
 421                 /* retry two more times (totally 3 times) */
 422                 if (p || ++retried >= 3)
 423                         return p;
 424                 vm_unmap_aliases();
 425         }
 426         return NULL;
 427 }
 428
 429 int erofs_register_sysfs(struct super_block *sb);
 430 void erofs_unregister_sysfs(struct super_block *sb);
 431 int __init erofs_init_sysfs(void);
 432 void erofs_exit_sysfs(void);
 433
 434 struct page *__erofs_allocpage(struct page **pagepool, gfp_t gfp, bool tryrsv);
 435 static inline struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
 436 {
 437         return __erofs_allocpage(pagepool, gfp, false);
 438 }
 439 static inline void erofs_pagepool_add(struct page **pagepool, struct page *page)
 440 {
 441         set_page_private(page, (unsigned long)*pagepool);
 442         *pagepool = page;
 443 }
 444 void erofs_release_pages(struct page **pagepool);
 445
 446 #ifdef CONFIG_EROFS_FS_ZIP
 447 #define MNGD_MAPPING(sbi)       ((sbi)->managed_cache->i_mapping)
 448
 449 extern atomic_long_t erofs_global_shrink_cnt;
 450 void erofs_shrinker_register(struct super_block *sb);
 451 void erofs_shrinker_unregister(struct super_block *sb);
 452 int __init erofs_init_shrinker(void);
 453 void erofs_exit_shrinker(void);
 454 int __init z_erofs_init_subsystem(void);
 455 void z_erofs_exit_subsystem(void);
 456 unsigned long z_erofs_shrink_scan(struct erofs_sb_info *sbi,
 457                                   unsigned long nr_shrink);
 458 int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map,
 459                             int flags);
 460 void *z_erofs_get_gbuf(unsigned int requiredpages);
 461 void z_erofs_put_gbuf(void *ptr);
 462 int z_erofs_gbuf_growsize(unsigned int nrpages);
 463 int __init z_erofs_gbuf_init(void);
 464 void z_erofs_gbuf_exit(void);
 465 int erofs_init_managed_cache(struct super_block *sb);
 466 int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb);
 467 #else
 468 static inline void erofs_shrinker_register(struct super_block *sb) {}
 469 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
 470 static inline int erofs_init_shrinker(void) { return 0; }
 471 static inline void erofs_exit_shrinker(void) {}
 472 static inline int z_erofs_init_subsystem(void) { return 0; }
 473 static inline void z_erofs_exit_subsystem(void) {}
 474 static inline int erofs_init_managed_cache(struct super_block *sb) { return 0; }
 475 #endif  /* !CONFIG_EROFS_FS_ZIP */
 476
 477 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
 478 struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev);
 479 void erofs_fileio_submit_bio(struct bio *bio);
 480 #else
 481 static inline struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
 482 static inline void erofs_fileio_submit_bio(struct bio *bio) {}
 483 #endif
 484
 485 #ifdef CONFIG_EROFS_FS_ONDEMAND
 486 int erofs_fscache_register_fs(struct super_block *sb);
 487 void erofs_fscache_unregister_fs(struct super_block *sb);
 488
 489 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
 490                                         char *name, unsigned int flags);
 491 void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
 492 struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev);
 493 void erofs_fscache_submit_bio(struct bio *bio);
 494 #else
 495 static inline int erofs_fscache_register_fs(struct super_block *sb)
 496 {
 497         return -EOPNOTSUPP;
 498 }
 499 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
 500
 501 static inline
 502 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
 503                                         char *name, unsigned int flags)
 504 {
 505         return ERR_PTR(-EOPNOTSUPP);
 506 }
 507
 508 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
 509 {
 510 }
 511 static inline struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
 512 static inline void erofs_fscache_submit_bio(struct bio *bio) {}
 513 #endif
 514
 515 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
 516
 517 #endif  /* __EROFS_INTERNAL_H */