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_warn(sb, fmt, ...) \ 27 _erofs_printk(sb, KERN_WARNING fmt "\n", ##__VA_ARGS__) 28 #define erofs_info(sb, fmt, ...) \ 29 _erofs_printk(sb, KERN_INFO fmt "\n", ##__VA_ARGS__) 30 31 #ifdef CONFIG_EROFS_FS_DEBUG 32 #define DBG_BUGON BUG_ON 33 #else 34 #define DBG_BUGON(x) ((void)(x)) 35 #endif /* !CONFIG_EROFS_FS_DEBUG */ 36 37 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */ 38 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1 39 40 typedef u64 erofs_nid_t; 41 typedef u64 erofs_off_t; 42 typedef u64 erofs_blk_t; 43 44 struct erofs_device_info { 45 char *path; 46 struct file *file; 47 struct dax_device *dax_dev; 48 u64 fsoff, dax_part_off; 49 50 erofs_blk_t blocks; 51 erofs_blk_t uniaddr; 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 unsigned int mount_opt; 64 }; 65 66 struct erofs_dev_context { 67 struct idr tree; 68 struct rw_semaphore rwsem; 69 70 unsigned int extra_devices; 71 bool flatdev; 72 }; 73 74 struct erofs_sb_lz4_info { 75 u16 max_pclusterblks; /* maximum physical blocks for LZ4 pclusters */ 76 }; 77 78 struct erofs_xattr_prefix_item { 79 struct erofs_xattr_long_prefix *prefix; 80 u8 infix_len; 81 }; 82 83 struct erofs_sb_info { 84 struct erofs_device_info dif0; 85 struct erofs_mount_opts opt; /* options */ 86 #ifdef CONFIG_EROFS_FS_ZIP 87 /* list for all registered superblocks, mainly for shrinker */ 88 struct list_head list; 89 struct mutex umount_mutex; 90 91 /* managed XArray arranged in physical block number */ 92 struct xarray managed_pslots; 93 94 unsigned int sync_decompress; /* strategy for sync decompression */ 95 unsigned int shrinker_run_no; 96 97 struct erofs_sb_lz4_info lz4; 98 #endif /* CONFIG_EROFS_FS_ZIP */ 99 struct inode *managed_cache; /* pseudo inode to cache physical data */ 100 struct inode *packed_inode; 101 struct inode *metabox_inode; 102 struct erofs_dev_context *devs; 103 u64 total_blocks; 104 105 u32 meta_blkaddr; 106 #ifdef CONFIG_EROFS_FS_XATTR 107 u32 xattr_blkaddr; 108 u32 xattr_prefix_start; 109 u8 xattr_prefix_count; 110 u8 ishare_xattr_prefix_id; 111 struct erofs_xattr_prefix_item *xattr_prefixes; 112 unsigned int xattr_filter_reserved; 113 #endif 114 u16 device_id_mask; /* valid bits of device id to be used */ 115 116 unsigned char islotbits; /* inode slot unit size in bit shift */ 117 unsigned char blkszbits; /* filesystem block size in bit shift */ 118 119 u32 sb_size; /* total superblock size */ 120 u32 fixed_nsec; 121 s64 epoch; 122 123 /* what we really care is nid, rather than ino.. */ 124 erofs_nid_t root_nid; 125 erofs_nid_t packed_nid; 126 erofs_nid_t metabox_nid; 127 /* used for statfs, f_files - f_favail */ 128 u64 inos; 129 130 char *volume_name; 131 u32 feature_compat; 132 u32 feature_incompat; 133 u16 available_compr_algs; 134 135 /* sysfs support */ 136 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */ 137 struct completion s_kobj_unregister; 138 erofs_off_t dir_ra_bytes; 139 140 char *domain_id; 141 }; 142 143 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info) 144 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info) 145 146 /* Mount flags set via mount options or defaults */ 147 #define EROFS_MOUNT_XATTR_USER 0x00000010 148 #define EROFS_MOUNT_POSIX_ACL 0x00000020 149 #define EROFS_MOUNT_DAX_ALWAYS 0x00000040 150 #define EROFS_MOUNT_DAX_NEVER 0x00000080 151 #define EROFS_MOUNT_DIRECT_IO 0x00000100 152 #define EROFS_MOUNT_INODE_SHARE 0x00000200 153 154 #define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option) 155 #define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option) 156 #define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option) 157 158 static inline bool erofs_is_fileio_mode(struct erofs_sb_info *sbi) 159 { 160 return IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) && sbi->dif0.file; 161 } 162 163 extern struct file_system_type erofs_anon_fs_type; 164 165 enum { 166 EROFS_ZIP_CACHE_DISABLED, 167 EROFS_ZIP_CACHE_READAHEAD, 168 EROFS_ZIP_CACHE_READAROUND 169 }; 170 171 struct erofs_buf { 172 struct address_space *mapping; 173 u64 off; 174 struct page *page; 175 void *base; 176 bool mc; 177 }; 178 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL }) 179 180 #define erofs_blknr(sb, pos) ((erofs_blk_t)((pos) >> (sb)->s_blocksize_bits)) 181 #define erofs_blkoff(sb, pos) ((pos) & ((sb)->s_blocksize - 1)) 182 #define erofs_pos(sb, blk) ((erofs_off_t)(blk) << (sb)->s_blocksize_bits) 183 #define erofs_iblks(i) (round_up((i)->i_size, i_blocksize(i)) >> (i)->i_blkbits) 184 185 #define EROFS_FEATURE_FUNCS(name, compat, feature) \ 186 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \ 187 { \ 188 return sbi->feature_##compat & EROFS_FEATURE_##feature; \ 189 } 190 191 EROFS_FEATURE_FUNCS(lz4_0padding, incompat, INCOMPAT_LZ4_0PADDING) 192 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS) 193 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER) 194 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE) 195 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE) 196 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2) 197 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING) 198 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS) 199 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE) 200 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES) 201 EROFS_FEATURE_FUNCS(48bit, incompat, INCOMPAT_48BIT) 202 EROFS_FEATURE_FUNCS(metabox, incompat, INCOMPAT_METABOX) 203 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM) 204 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER) 205 EROFS_FEATURE_FUNCS(shared_ea_in_metabox, compat, COMPAT_SHARED_EA_IN_METABOX) 206 EROFS_FEATURE_FUNCS(plain_xattr_pfx, compat, COMPAT_PLAIN_XATTR_PFX) 207 EROFS_FEATURE_FUNCS(ishare_xattrs, compat, COMPAT_ISHARE_XATTRS) 208 209 static inline u64 erofs_nid_to_ino64(struct erofs_sb_info *sbi, erofs_nid_t nid) 210 { 211 if (!erofs_sb_has_metabox(sbi)) 212 return nid; 213 214 /* 215 * When metadata compression is enabled, avoid generating excessively 216 * large inode numbers for metadata-compressed inodes. Shift NIDs in 217 * the 31-62 bit range left by one and move the metabox flag to bit 31. 218 * 219 * Note: on-disk NIDs remain unchanged as they are primarily used for 220 * compatibility with non-LFS 32-bit applications. 221 */ 222 return ((nid << 1) & GENMASK_ULL(63, 32)) | (nid & GENMASK(30, 0)) | 223 ((nid >> EROFS_DIRENT_NID_METABOX_BIT) << 31); 224 } 225 226 /* atomic flag definitions */ 227 #define EROFS_I_EA_INITED_BIT 0 228 #define EROFS_I_Z_INITED_BIT 1 229 230 /* bitlock definitions (arranged in reverse order) */ 231 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1) 232 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2) 233 234 /* default readahead size of directories */ 235 #define EROFS_DIR_RA_BYTES 16384 236 237 struct erofs_inode_fingerprint { 238 u8 *opaque; 239 int size; 240 }; 241 242 struct erofs_inode { 243 erofs_nid_t nid; 244 245 /* atomic flags (including bitlocks) */ 246 unsigned long flags; 247 248 unsigned char datalayout; 249 unsigned char inode_isize; 250 bool dot_omitted; 251 unsigned int xattr_isize; 252 253 unsigned int xattr_name_filter; 254 unsigned int xattr_shared_count; 255 unsigned int *xattr_shared_xattrs; 256 257 union { 258 erofs_blk_t startblk; 259 struct { 260 unsigned short chunkformat; 261 unsigned char chunkbits; 262 }; 263 #ifdef CONFIG_EROFS_FS_ZIP 264 struct { 265 unsigned short z_advise; 266 unsigned char z_algofmt[2]; 267 unsigned char z_lclusterbits; 268 union { 269 u64 z_tailextent_headlcn; 270 u64 z_extents; 271 }; 272 erofs_off_t z_fragmentoff; 273 unsigned short z_idata_size; 274 }; 275 #endif /* CONFIG_EROFS_FS_ZIP */ 276 }; 277 #ifdef CONFIG_EROFS_FS_PAGE_CACHE_SHARE 278 struct list_head ishare_list; 279 union { 280 /* for each anon shared inode */ 281 struct { 282 struct erofs_inode_fingerprint fingerprint; 283 spinlock_t ishare_lock; 284 }; 285 /* for each real filesystem inode */ 286 struct dentry *sharedentry; 287 }; 288 #endif 289 /* the corresponding vfs inode */ 290 struct inode vfs_inode; 291 }; 292 293 #define EROFS_I(ptr) container_of(ptr, struct erofs_inode, vfs_inode) 294 295 static inline bool erofs_inode_in_metabox(struct inode *inode) 296 { 297 return EROFS_I(inode)->nid & BIT_ULL(EROFS_DIRENT_NID_METABOX_BIT); 298 } 299 300 static inline erofs_off_t erofs_iloc(struct inode *inode) 301 { 302 struct erofs_sb_info *sbi = EROFS_I_SB(inode); 303 erofs_nid_t nid_lo = EROFS_I(inode)->nid & EROFS_DIRENT_NID_MASK; 304 305 if (erofs_inode_in_metabox(inode)) 306 return nid_lo << sbi->islotbits; 307 return erofs_pos(inode->i_sb, sbi->meta_blkaddr) + 308 (nid_lo << sbi->islotbits); 309 } 310 311 static inline unsigned int erofs_inode_version(unsigned int ifmt) 312 { 313 return (ifmt >> EROFS_I_VERSION_BIT) & EROFS_I_VERSION_MASK; 314 } 315 316 static inline unsigned int erofs_inode_datalayout(unsigned int ifmt) 317 { 318 return (ifmt >> EROFS_I_DATALAYOUT_BIT) & EROFS_I_DATALAYOUT_MASK; 319 } 320 321 /* reclaiming is never triggered when allocating new folios. */ 322 static inline struct folio *erofs_grab_folio_nowait(struct address_space *as, 323 pgoff_t index) 324 { 325 return __filemap_get_folio(as, index, 326 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, 327 readahead_gfp_mask(as) & ~__GFP_RECLAIM); 328 } 329 330 /* Allocated on disk at @m_pa (e.g. NOT a fragment extent) */ 331 #define EROFS_MAP_MAPPED 0x0001 332 /* Located in metadata (could be copied from bd_inode) */ 333 #define EROFS_MAP_META 0x0002 334 /* @m_llen may be truncated by the runtime compared to the on-disk record */ 335 #define EROFS_MAP_PARTIAL_MAPPED 0x0004 336 /* The on-disk @m_llen may cover only part of the encoded data */ 337 #define EROFS_MAP_PARTIAL_REF 0x0008 338 /* Located in the special packed inode */ 339 #define EROFS_MAP_FRAGMENT 0x0010 340 /* The encoded on-disk data will be fully handled (decompressed) */ 341 #define EROFS_MAP_FULL(f) (!((f) & (EROFS_MAP_PARTIAL_MAPPED | \ 342 EROFS_MAP_PARTIAL_REF))) 343 344 struct erofs_map_blocks { 345 struct erofs_buf buf; 346 347 erofs_off_t m_pa, m_la; 348 u64 m_plen, m_llen; 349 350 unsigned short m_deviceid; 351 char m_algorithmformat; 352 unsigned int m_flags; 353 }; 354 355 /* 356 * Used to get the exact decompressed length, e.g. fiemap (consider lookback 357 * approach instead if possible since it's more metadata lightweight.) 358 */ 359 #define EROFS_GET_BLOCKS_FIEMAP 0x0001 360 /* Used to map the whole extent if non-negligible data is requested for LZMA */ 361 #define EROFS_GET_BLOCKS_READMORE 0x0002 362 /* Used to map tail extent for tailpacking inline or fragment pcluster */ 363 #define EROFS_GET_BLOCKS_FINDTAIL 0x0004 364 365 enum { 366 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX, 367 Z_EROFS_COMPRESSION_INTERLACED, 368 Z_EROFS_COMPRESSION_RUNTIME_MAX 369 }; 370 371 struct erofs_map_dev { 372 struct super_block *m_sb; 373 struct erofs_device_info *m_dif; 374 struct block_device *m_bdev; 375 376 erofs_off_t m_pa; 377 unsigned int m_deviceid; 378 }; 379 380 extern const struct super_operations erofs_sops; 381 extern const struct address_space_operations erofs_aops; 382 extern const struct address_space_operations erofs_fileio_aops; 383 extern const struct address_space_operations z_erofs_aops; 384 385 extern const struct inode_operations erofs_generic_iops; 386 extern const struct inode_operations erofs_symlink_iops; 387 extern const struct inode_operations erofs_fast_symlink_iops; 388 extern const struct inode_operations erofs_dir_iops; 389 390 extern const struct file_operations erofs_file_fops; 391 extern const struct file_operations erofs_dir_fops; 392 extern const struct file_operations erofs_ishare_fops; 393 394 extern const struct iomap_ops z_erofs_iomap_report_ops; 395 396 int erofs_setup_managed_cache(struct super_block *sb); 397 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE 398 int erofs_read_meta_folio(struct file *file, struct folio *folio); 399 #else 400 #define erofs_read_meta_folio NULL 401 #endif 402 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf, 403 erofs_off_t *offset, int *lengthp); 404 void erofs_unmap_metabuf(struct erofs_buf *buf); 405 void erofs_put_metabuf(struct erofs_buf *buf); 406 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap); 407 int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb, 408 bool in_metabox); 409 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb, 410 erofs_off_t offset, bool in_metabox); 411 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev); 412 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 413 u64 start, u64 len); 414 loff_t erofs_file_llseek(struct file *file, loff_t offset, int whence); 415 int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map); 416 void erofs_onlinefolio_init(struct folio *folio); 417 void erofs_onlinefolio_split(struct folio *folio); 418 void erofs_onlinefolio_end(struct folio *folio, int err, bool dirty); 419 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid); 420 int erofs_getattr(struct mnt_idmap *idmap, const struct path *path, 421 struct kstat *stat, u32 request_mask, 422 unsigned int query_flags); 423 int erofs_namei(struct inode *dir, const struct qstr *name, 424 erofs_nid_t *nid, unsigned int *d_type); 425 426 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count) 427 { 428 int retried = 0; 429 430 while (1) { 431 void *p = vm_map_ram(pages, count, -1); 432 433 /* retry two more times (totally 3 times) */ 434 if (p || ++retried >= 3) 435 return p; 436 vm_unmap_aliases(); 437 } 438 return NULL; 439 } 440 441 static inline const struct address_space_operations * 442 erofs_get_aops(struct inode *realinode) 443 { 444 if (erofs_inode_is_data_compressed(EROFS_I(realinode)->datalayout)) { 445 if (!IS_ENABLED(CONFIG_EROFS_FS_ZIP)) 446 return ERR_PTR(-EOPNOTSUPP); 447 DO_ONCE_LITE_IF(realinode->i_blkbits != PAGE_SHIFT, 448 erofs_info, realinode->i_sb, 449 "EXPERIMENTAL EROFS subpage compressed block support in use. Use at your own risk!"); 450 return &z_erofs_aops; 451 } 452 if (IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) && 453 erofs_is_fileio_mode(EROFS_SB(realinode->i_sb))) 454 return &erofs_fileio_aops; 455 return &erofs_aops; 456 } 457 458 int erofs_register_sysfs(struct super_block *sb); 459 void erofs_unregister_sysfs(struct super_block *sb); 460 int __init erofs_init_sysfs(void); 461 void erofs_exit_sysfs(void); 462 463 struct page *__erofs_allocpage(struct page **pagepool, gfp_t gfp, bool tryrsv); 464 static inline struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp) 465 { 466 return __erofs_allocpage(pagepool, gfp, false); 467 } 468 static inline void erofs_pagepool_add(struct page **pagepool, struct page *page) 469 { 470 set_page_private(page, (unsigned long)*pagepool); 471 *pagepool = page; 472 } 473 void erofs_release_pages(struct page **pagepool); 474 475 #ifdef CONFIG_EROFS_FS_ZIP 476 #define MNGD_MAPPING(sbi) ((sbi)->managed_cache->i_mapping) 477 478 extern atomic_long_t erofs_global_shrink_cnt; 479 void erofs_shrinker_register(struct super_block *sb); 480 void erofs_shrinker_unregister(struct super_block *sb); 481 int __init erofs_init_shrinker(void); 482 void erofs_exit_shrinker(void); 483 int __init z_erofs_init_subsystem(void); 484 void z_erofs_exit_subsystem(void); 485 int z_erofs_init_super(struct super_block *sb); 486 unsigned long z_erofs_shrink_scan(struct erofs_sb_info *sbi, 487 unsigned long nr_shrink); 488 int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map, 489 int flags); 490 void *z_erofs_get_gbuf(unsigned int requiredpages); 491 void z_erofs_put_gbuf(void *ptr); 492 int z_erofs_gbuf_growsize(unsigned int nrpages); 493 int __init z_erofs_gbuf_init(void); 494 void z_erofs_gbuf_exit(void); 495 #else 496 static inline void erofs_shrinker_register(struct super_block *sb) {} 497 static inline void erofs_shrinker_unregister(struct super_block *sb) {} 498 static inline int erofs_init_shrinker(void) { return 0; } 499 static inline void erofs_exit_shrinker(void) {} 500 static inline int z_erofs_init_subsystem(void) { return 0; } 501 static inline void z_erofs_exit_subsystem(void) {} 502 static inline int z_erofs_init_super(struct super_block *sb) { return 0; } 503 #endif /* !CONFIG_EROFS_FS_ZIP */ 504 int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb); 505 506 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE 507 struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev); 508 void erofs_fileio_submit_bio(struct bio *bio); 509 #else 510 static inline struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev) { return NULL; } 511 static inline void erofs_fileio_submit_bio(struct bio *bio) {} 512 #endif 513 514 #ifdef CONFIG_EROFS_FS_PAGE_CACHE_SHARE 515 int __init erofs_init_ishare(void); 516 void erofs_exit_ishare(void); 517 bool erofs_ishare_fill_inode(struct inode *inode); 518 void erofs_ishare_free_inode(struct inode *inode); 519 struct inode *erofs_real_inode(struct inode *inode, bool *need_iput); 520 #else 521 static inline int erofs_init_ishare(void) { return 0; } 522 static inline void erofs_exit_ishare(void) {} 523 static inline bool erofs_ishare_fill_inode(struct inode *inode) { return false; } 524 static inline void erofs_ishare_free_inode(struct inode *inode) {} 525 static inline struct inode *erofs_real_inode(struct inode *inode, bool *need_iput) 526 { 527 *need_iput = false; 528 return inode; 529 } 530 #endif 531 532 long erofs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); 533 long erofs_compat_ioctl(struct file *filp, unsigned int cmd, 534 unsigned long arg); 535 536 #endif /* __EROFS_INTERNAL_H */ 537