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