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/dcache.h> 12 #include <linux/mm.h> 13 #include <linux/pagemap.h> 14 #include <linux/bio.h> 15 #include <linux/magic.h> 16 #include <linux/slab.h> 17 #include <linux/vmalloc.h> 18 #include <linux/iomap.h> 19 #include "erofs_fs.h" 20 21 /* redefine pr_fmt "erofs: " */ 22 #undef pr_fmt 23 #define pr_fmt(fmt) "erofs: " fmt 24 25 __printf(3, 4) void _erofs_err(struct super_block *sb, 26 const char *function, const char *fmt, ...); 27 #define erofs_err(sb, fmt, ...) \ 28 _erofs_err(sb, __func__, fmt "\n", ##__VA_ARGS__) 29 __printf(3, 4) void _erofs_info(struct super_block *sb, 30 const char *function, const char *fmt, ...); 31 #define erofs_info(sb, fmt, ...) \ 32 _erofs_info(sb, __func__, fmt "\n", ##__VA_ARGS__) 33 #ifdef CONFIG_EROFS_FS_DEBUG 34 #define DBG_BUGON BUG_ON 35 #else 36 #define DBG_BUGON(x) ((void)(x)) 37 #endif /* !CONFIG_EROFS_FS_DEBUG */ 38 39 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */ 40 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1 41 42 typedef u64 erofs_nid_t; 43 typedef u64 erofs_off_t; 44 /* data type for filesystem-wide blocks number */ 45 typedef u32 erofs_blk_t; 46 47 struct erofs_device_info { 48 char *path; 49 struct erofs_fscache *fscache; 50 struct block_device *bdev; 51 struct dax_device *dax_dev; 52 u64 dax_part_off; 53 54 u32 blocks; 55 u32 mapped_blkaddr; 56 }; 57 58 enum { 59 EROFS_SYNC_DECOMPRESS_AUTO, 60 EROFS_SYNC_DECOMPRESS_FORCE_ON, 61 EROFS_SYNC_DECOMPRESS_FORCE_OFF 62 }; 63 64 struct erofs_mount_opts { 65 #ifdef CONFIG_EROFS_FS_ZIP 66 /* current strategy of how to use managed cache */ 67 unsigned char cache_strategy; 68 /* strategy of sync decompression (0 - auto, 1 - force on, 2 - force off) */ 69 unsigned int sync_decompress; 70 71 /* threshold for decompression synchronously */ 72 unsigned int max_sync_decompress_pages; 73 #endif 74 unsigned int mount_opt; 75 }; 76 77 struct erofs_dev_context { 78 struct idr tree; 79 struct rw_semaphore rwsem; 80 81 unsigned int extra_devices; 82 bool flatdev; 83 }; 84 85 struct erofs_fs_context { 86 struct erofs_mount_opts opt; 87 struct erofs_dev_context *devs; 88 char *fsid; 89 char *domain_id; 90 }; 91 92 /* all filesystem-wide lz4 configurations */ 93 struct erofs_sb_lz4_info { 94 /* # of pages needed for EROFS lz4 rolling decompression */ 95 u16 max_distance_pages; 96 /* maximum possible blocks for pclusters in the filesystem */ 97 u16 max_pclusterblks; 98 }; 99 100 struct erofs_domain { 101 refcount_t ref; 102 struct list_head list; 103 struct fscache_volume *volume; 104 char *domain_id; 105 }; 106 107 struct erofs_fscache { 108 struct fscache_cookie *cookie; 109 struct inode *inode; /* anonymous inode for the blob */ 110 111 /* used for share domain mode */ 112 struct erofs_domain *domain; 113 struct list_head node; 114 refcount_t ref; 115 char *name; 116 }; 117 118 struct erofs_xattr_prefix_item { 119 struct erofs_xattr_long_prefix *prefix; 120 u8 infix_len; 121 }; 122 123 struct erofs_sb_info { 124 struct erofs_mount_opts opt; /* options */ 125 #ifdef CONFIG_EROFS_FS_ZIP 126 /* list for all registered superblocks, mainly for shrinker */ 127 struct list_head list; 128 struct mutex umount_mutex; 129 130 /* managed XArray arranged in physical block number */ 131 struct xarray managed_pslots; 132 133 unsigned int shrinker_run_no; 134 u16 available_compr_algs; 135 136 /* pseudo inode to manage cached pages */ 137 struct inode *managed_cache; 138 139 struct erofs_sb_lz4_info lz4; 140 #endif /* CONFIG_EROFS_FS_ZIP */ 141 struct inode *packed_inode; 142 struct erofs_dev_context *devs; 143 struct dax_device *dax_dev; 144 u64 dax_part_off; 145 u64 total_blocks; 146 u32 primarydevice_blocks; 147 148 u32 meta_blkaddr; 149 #ifdef CONFIG_EROFS_FS_XATTR 150 u32 xattr_blkaddr; 151 u32 xattr_prefix_start; 152 u8 xattr_prefix_count; 153 struct erofs_xattr_prefix_item *xattr_prefixes; 154 #endif 155 u16 device_id_mask; /* valid bits of device id to be used */ 156 157 unsigned char islotbits; /* inode slot unit size in bit shift */ 158 unsigned char blkszbits; /* filesystem block size in bit shift */ 159 160 u32 sb_size; /* total superblock size */ 161 u32 build_time_nsec; 162 u64 build_time; 163 164 /* what we really care is nid, rather than ino.. */ 165 erofs_nid_t root_nid; 166 erofs_nid_t packed_nid; 167 /* used for statfs, f_files - f_favail */ 168 u64 inos; 169 170 u8 uuid[16]; /* 128-bit uuid for volume */ 171 u8 volume_name[16]; /* volume name */ 172 u32 feature_compat; 173 u32 feature_incompat; 174 175 /* sysfs support */ 176 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */ 177 struct completion s_kobj_unregister; 178 179 /* fscache support */ 180 struct fscache_volume *volume; 181 struct erofs_fscache *s_fscache; 182 struct erofs_domain *domain; 183 char *fsid; 184 char *domain_id; 185 }; 186 187 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info) 188 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info) 189 190 /* Mount flags set via mount options or defaults */ 191 #define EROFS_MOUNT_XATTR_USER 0x00000010 192 #define EROFS_MOUNT_POSIX_ACL 0x00000020 193 #define EROFS_MOUNT_DAX_ALWAYS 0x00000040 194 #define EROFS_MOUNT_DAX_NEVER 0x00000080 195 196 #define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option) 197 #define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option) 198 #define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option) 199 200 static inline bool erofs_is_fscache_mode(struct super_block *sb) 201 { 202 return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && !sb->s_bdev; 203 } 204 205 enum { 206 EROFS_ZIP_CACHE_DISABLED, 207 EROFS_ZIP_CACHE_READAHEAD, 208 EROFS_ZIP_CACHE_READAROUND 209 }; 210 211 #define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL) 212 213 /* basic unit of the workstation of a super_block */ 214 struct erofs_workgroup { 215 /* the workgroup index in the workstation */ 216 pgoff_t index; 217 218 /* overall workgroup reference count */ 219 atomic_t refcount; 220 }; 221 222 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp, 223 int val) 224 { 225 preempt_disable(); 226 if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) { 227 preempt_enable(); 228 return false; 229 } 230 return true; 231 } 232 233 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp, 234 int orig_val) 235 { 236 /* 237 * other observers should notice all modifications 238 * in the freezing period. 239 */ 240 smp_mb(); 241 atomic_set(&grp->refcount, orig_val); 242 preempt_enable(); 243 } 244 245 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp) 246 { 247 return atomic_cond_read_relaxed(&grp->refcount, 248 VAL != EROFS_LOCKED_MAGIC); 249 } 250 251 enum erofs_kmap_type { 252 EROFS_NO_KMAP, /* don't map the buffer */ 253 EROFS_KMAP, /* use kmap_local_page() to map the buffer */ 254 }; 255 256 struct erofs_buf { 257 struct inode *inode; 258 struct page *page; 259 void *base; 260 enum erofs_kmap_type kmap_type; 261 }; 262 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL }) 263 264 #define ROOT_NID(sb) ((sb)->root_nid) 265 266 #define erofs_blknr(sb, addr) ((addr) >> (sb)->s_blocksize_bits) 267 #define erofs_blkoff(sb, addr) ((addr) & ((sb)->s_blocksize - 1)) 268 #define erofs_pos(sb, blk) ((erofs_off_t)(blk) << (sb)->s_blocksize_bits) 269 #define erofs_iblks(i) (round_up((i)->i_size, i_blocksize(i)) >> (i)->i_blkbits) 270 271 #define EROFS_FEATURE_FUNCS(name, compat, feature) \ 272 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \ 273 { \ 274 return sbi->feature_##compat & EROFS_FEATURE_##feature; \ 275 } 276 277 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING) 278 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS) 279 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER) 280 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE) 281 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE) 282 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2) 283 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING) 284 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS) 285 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE) 286 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES) 287 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM) 288 289 /* atomic flag definitions */ 290 #define EROFS_I_EA_INITED_BIT 0 291 #define EROFS_I_Z_INITED_BIT 1 292 293 /* bitlock definitions (arranged in reverse order) */ 294 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1) 295 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2) 296 297 struct erofs_inode { 298 erofs_nid_t nid; 299 300 /* atomic flags (including bitlocks) */ 301 unsigned long flags; 302 303 unsigned char datalayout; 304 unsigned char inode_isize; 305 unsigned int xattr_isize; 306 307 unsigned int xattr_shared_count; 308 unsigned int *xattr_shared_xattrs; 309 310 union { 311 erofs_blk_t raw_blkaddr; 312 struct { 313 unsigned short chunkformat; 314 unsigned char chunkbits; 315 }; 316 #ifdef CONFIG_EROFS_FS_ZIP 317 struct { 318 unsigned short z_advise; 319 unsigned char z_algorithmtype[2]; 320 unsigned char z_logical_clusterbits; 321 unsigned long z_tailextent_headlcn; 322 union { 323 struct { 324 erofs_off_t z_idataoff; 325 unsigned short z_idata_size; 326 }; 327 erofs_off_t z_fragmentoff; 328 }; 329 }; 330 #endif /* CONFIG_EROFS_FS_ZIP */ 331 }; 332 /* the corresponding vfs inode */ 333 struct inode vfs_inode; 334 }; 335 336 #define EROFS_I(ptr) container_of(ptr, struct erofs_inode, vfs_inode) 337 338 static inline erofs_off_t erofs_iloc(struct inode *inode) 339 { 340 struct erofs_sb_info *sbi = EROFS_I_SB(inode); 341 342 return erofs_pos(inode->i_sb, sbi->meta_blkaddr) + 343 (EROFS_I(inode)->nid << sbi->islotbits); 344 } 345 346 static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit, 347 unsigned int bits) 348 { 349 350 return (value >> bit) & ((1 << bits) - 1); 351 } 352 353 354 static inline unsigned int erofs_inode_version(unsigned int value) 355 { 356 return erofs_bitrange(value, EROFS_I_VERSION_BIT, 357 EROFS_I_VERSION_BITS); 358 } 359 360 static inline unsigned int erofs_inode_datalayout(unsigned int value) 361 { 362 return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT, 363 EROFS_I_DATALAYOUT_BITS); 364 } 365 366 /* 367 * Different from grab_cache_page_nowait(), reclaiming is never triggered 368 * when allocating new pages. 369 */ 370 static inline 371 struct page *erofs_grab_cache_page_nowait(struct address_space *mapping, 372 pgoff_t index) 373 { 374 return pagecache_get_page(mapping, index, 375 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, 376 readahead_gfp_mask(mapping) & ~__GFP_RECLAIM); 377 } 378 379 /* Has a disk mapping */ 380 #define EROFS_MAP_MAPPED 0x0001 381 /* Located in metadata (could be copied from bd_inode) */ 382 #define EROFS_MAP_META 0x0002 383 /* The extent is encoded */ 384 #define EROFS_MAP_ENCODED 0x0004 385 /* The length of extent is full */ 386 #define EROFS_MAP_FULL_MAPPED 0x0008 387 /* Located in the special packed inode */ 388 #define EROFS_MAP_FRAGMENT 0x0010 389 /* The extent refers to partial decompressed data */ 390 #define EROFS_MAP_PARTIAL_REF 0x0020 391 392 struct erofs_map_blocks { 393 struct erofs_buf buf; 394 395 erofs_off_t m_pa, m_la; 396 u64 m_plen, m_llen; 397 398 unsigned short m_deviceid; 399 char m_algorithmformat; 400 unsigned int m_flags; 401 }; 402 403 /* 404 * Used to get the exact decompressed length, e.g. fiemap (consider lookback 405 * approach instead if possible since it's more metadata lightweight.) 406 */ 407 #define EROFS_GET_BLOCKS_FIEMAP 0x0001 408 /* Used to map the whole extent if non-negligible data is requested for LZMA */ 409 #define EROFS_GET_BLOCKS_READMORE 0x0002 410 /* Used to map tail extent for tailpacking inline or fragment pcluster */ 411 #define EROFS_GET_BLOCKS_FINDTAIL 0x0004 412 413 enum { 414 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX, 415 Z_EROFS_COMPRESSION_INTERLACED, 416 Z_EROFS_COMPRESSION_RUNTIME_MAX 417 }; 418 419 struct erofs_map_dev { 420 struct erofs_fscache *m_fscache; 421 struct block_device *m_bdev; 422 struct dax_device *m_daxdev; 423 u64 m_dax_part_off; 424 425 erofs_off_t m_pa; 426 unsigned int m_deviceid; 427 }; 428 429 extern struct file_system_type erofs_fs_type; 430 extern const struct super_operations erofs_sops; 431 432 extern const struct address_space_operations erofs_raw_access_aops; 433 extern const struct address_space_operations z_erofs_aops; 434 extern const struct address_space_operations erofs_fscache_access_aops; 435 436 extern const struct inode_operations erofs_generic_iops; 437 extern const struct inode_operations erofs_symlink_iops; 438 extern const struct inode_operations erofs_fast_symlink_iops; 439 extern const struct inode_operations erofs_dir_iops; 440 441 extern const struct file_operations erofs_file_fops; 442 extern const struct file_operations erofs_dir_fops; 443 444 extern const struct iomap_ops z_erofs_iomap_report_ops; 445 446 /* flags for erofs_fscache_register_cookie() */ 447 #define EROFS_REG_COOKIE_SHARE 0x0001 448 #define EROFS_REG_COOKIE_NEED_NOEXIST 0x0002 449 450 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf, 451 erofs_off_t *offset, int *lengthp); 452 void erofs_unmap_metabuf(struct erofs_buf *buf); 453 void erofs_put_metabuf(struct erofs_buf *buf); 454 void *erofs_bread(struct erofs_buf *buf, erofs_blk_t blkaddr, 455 enum erofs_kmap_type type); 456 void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb); 457 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb, 458 erofs_blk_t blkaddr, enum erofs_kmap_type type); 459 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev); 460 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 461 u64 start, u64 len); 462 int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map); 463 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid); 464 int erofs_getattr(struct mnt_idmap *idmap, const struct path *path, 465 struct kstat *stat, u32 request_mask, 466 unsigned int query_flags); 467 int erofs_namei(struct inode *dir, const struct qstr *name, 468 erofs_nid_t *nid, unsigned int *d_type); 469 470 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count) 471 { 472 int retried = 0; 473 474 while (1) { 475 void *p = vm_map_ram(pages, count, -1); 476 477 /* retry two more times (totally 3 times) */ 478 if (p || ++retried >= 3) 479 return p; 480 vm_unmap_aliases(); 481 } 482 return NULL; 483 } 484 485 void *erofs_get_pcpubuf(unsigned int requiredpages); 486 void erofs_put_pcpubuf(void *ptr); 487 int erofs_pcpubuf_growsize(unsigned int nrpages); 488 void __init erofs_pcpubuf_init(void); 489 void erofs_pcpubuf_exit(void); 490 491 int erofs_register_sysfs(struct super_block *sb); 492 void erofs_unregister_sysfs(struct super_block *sb); 493 int __init erofs_init_sysfs(void); 494 void erofs_exit_sysfs(void); 495 496 struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp); 497 static inline void erofs_pagepool_add(struct page **pagepool, struct page *page) 498 { 499 set_page_private(page, (unsigned long)*pagepool); 500 *pagepool = page; 501 } 502 void erofs_release_pages(struct page **pagepool); 503 504 #ifdef CONFIG_EROFS_FS_ZIP 505 int erofs_workgroup_put(struct erofs_workgroup *grp); 506 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb, 507 pgoff_t index); 508 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb, 509 struct erofs_workgroup *grp); 510 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp); 511 void erofs_shrinker_register(struct super_block *sb); 512 void erofs_shrinker_unregister(struct super_block *sb); 513 int __init erofs_init_shrinker(void); 514 void erofs_exit_shrinker(void); 515 int __init z_erofs_init_zip_subsystem(void); 516 void z_erofs_exit_zip_subsystem(void); 517 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi, 518 struct erofs_workgroup *egrp); 519 int erofs_try_to_free_cached_page(struct page *page); 520 int z_erofs_load_lz4_config(struct super_block *sb, 521 struct erofs_super_block *dsb, 522 struct z_erofs_lz4_cfgs *lz4, int len); 523 int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map, 524 int flags); 525 #else 526 static inline void erofs_shrinker_register(struct super_block *sb) {} 527 static inline void erofs_shrinker_unregister(struct super_block *sb) {} 528 static inline int erofs_init_shrinker(void) { return 0; } 529 static inline void erofs_exit_shrinker(void) {} 530 static inline int z_erofs_init_zip_subsystem(void) { return 0; } 531 static inline void z_erofs_exit_zip_subsystem(void) {} 532 static inline int z_erofs_load_lz4_config(struct super_block *sb, 533 struct erofs_super_block *dsb, 534 struct z_erofs_lz4_cfgs *lz4, int len) 535 { 536 if (lz4 || dsb->u1.lz4_max_distance) { 537 erofs_err(sb, "lz4 algorithm isn't enabled"); 538 return -EINVAL; 539 } 540 return 0; 541 } 542 #endif /* !CONFIG_EROFS_FS_ZIP */ 543 544 #ifdef CONFIG_EROFS_FS_ZIP_LZMA 545 int __init z_erofs_lzma_init(void); 546 void z_erofs_lzma_exit(void); 547 int z_erofs_load_lzma_config(struct super_block *sb, 548 struct erofs_super_block *dsb, 549 struct z_erofs_lzma_cfgs *lzma, int size); 550 #else 551 static inline int z_erofs_lzma_init(void) { return 0; } 552 static inline int z_erofs_lzma_exit(void) { return 0; } 553 static inline int z_erofs_load_lzma_config(struct super_block *sb, 554 struct erofs_super_block *dsb, 555 struct z_erofs_lzma_cfgs *lzma, int size) { 556 if (lzma) { 557 erofs_err(sb, "lzma algorithm isn't enabled"); 558 return -EINVAL; 559 } 560 return 0; 561 } 562 #endif /* !CONFIG_EROFS_FS_ZIP_LZMA */ 563 564 #ifdef CONFIG_EROFS_FS_ONDEMAND 565 int erofs_fscache_register_fs(struct super_block *sb); 566 void erofs_fscache_unregister_fs(struct super_block *sb); 567 568 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb, 569 char *name, unsigned int flags); 570 void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache); 571 #else 572 static inline int erofs_fscache_register_fs(struct super_block *sb) 573 { 574 return -EOPNOTSUPP; 575 } 576 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {} 577 578 static inline 579 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb, 580 char *name, unsigned int flags) 581 { 582 return ERR_PTR(-EOPNOTSUPP); 583 } 584 585 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache) 586 { 587 } 588 #endif 589 590 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 591 592 #endif /* __EROFS_INTERNAL_H */ 593