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/buffer_head.h> 16 #include <linux/magic.h> 17 #include <linux/slab.h> 18 #include <linux/vmalloc.h> 19 #include <linux/iomap.h> 20 #include "erofs_fs.h" 21 22 /* redefine pr_fmt "erofs: " */ 23 #undef pr_fmt 24 #define pr_fmt(fmt) "erofs: " fmt 25 26 __printf(3, 4) void _erofs_err(struct super_block *sb, 27 const char *function, const char *fmt, ...); 28 #define erofs_err(sb, fmt, ...) \ 29 _erofs_err(sb, __func__, fmt "\n", ##__VA_ARGS__) 30 __printf(3, 4) void _erofs_info(struct super_block *sb, 31 const char *function, const char *fmt, ...); 32 #define erofs_info(sb, fmt, ...) \ 33 _erofs_info(sb, __func__, fmt "\n", ##__VA_ARGS__) 34 #ifdef CONFIG_EROFS_FS_DEBUG 35 #define erofs_dbg(x, ...) pr_debug(x "\n", ##__VA_ARGS__) 36 #define DBG_BUGON BUG_ON 37 #else 38 #define erofs_dbg(x, ...) ((void)0) 39 #define DBG_BUGON(x) ((void)(x)) 40 #endif /* !CONFIG_EROFS_FS_DEBUG */ 41 42 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */ 43 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1 44 45 typedef u64 erofs_nid_t; 46 typedef u64 erofs_off_t; 47 /* data type for filesystem-wide blocks number */ 48 typedef u32 erofs_blk_t; 49 50 struct erofs_device_info { 51 char *path; 52 struct erofs_fscache *fscache; 53 struct block_device *bdev; 54 struct dax_device *dax_dev; 55 u64 dax_part_off; 56 57 u32 blocks; 58 u32 mapped_blkaddr; 59 }; 60 61 enum { 62 EROFS_SYNC_DECOMPRESS_AUTO, 63 EROFS_SYNC_DECOMPRESS_FORCE_ON, 64 EROFS_SYNC_DECOMPRESS_FORCE_OFF 65 }; 66 67 struct erofs_mount_opts { 68 #ifdef CONFIG_EROFS_FS_ZIP 69 /* current strategy of how to use managed cache */ 70 unsigned char cache_strategy; 71 /* strategy of sync decompression (0 - auto, 1 - force on, 2 - force off) */ 72 unsigned int sync_decompress; 73 74 /* threshold for decompression synchronously */ 75 unsigned int max_sync_decompress_pages; 76 #endif 77 unsigned int mount_opt; 78 char *fsid; 79 }; 80 81 struct erofs_dev_context { 82 struct idr tree; 83 struct rw_semaphore rwsem; 84 85 unsigned int extra_devices; 86 }; 87 88 struct erofs_fs_context { 89 struct erofs_mount_opts opt; 90 struct erofs_dev_context *devs; 91 }; 92 93 /* all filesystem-wide lz4 configurations */ 94 struct erofs_sb_lz4_info { 95 /* # of pages needed for EROFS lz4 rolling decompression */ 96 u16 max_distance_pages; 97 /* maximum possible blocks for pclusters in the filesystem */ 98 u16 max_pclusterblks; 99 }; 100 101 struct erofs_fscache { 102 struct fscache_cookie *cookie; 103 struct inode *inode; 104 }; 105 106 struct erofs_sb_info { 107 struct erofs_mount_opts opt; /* options */ 108 #ifdef CONFIG_EROFS_FS_ZIP 109 /* list for all registered superblocks, mainly for shrinker */ 110 struct list_head list; 111 struct mutex umount_mutex; 112 113 /* managed XArray arranged in physical block number */ 114 struct xarray managed_pslots; 115 116 unsigned int shrinker_run_no; 117 u16 available_compr_algs; 118 119 /* pseudo inode to manage cached pages */ 120 struct inode *managed_cache; 121 122 struct erofs_sb_lz4_info lz4; 123 #endif /* CONFIG_EROFS_FS_ZIP */ 124 struct erofs_dev_context *devs; 125 struct dax_device *dax_dev; 126 u64 dax_part_off; 127 u64 total_blocks; 128 u32 primarydevice_blocks; 129 130 u32 meta_blkaddr; 131 #ifdef CONFIG_EROFS_FS_XATTR 132 u32 xattr_blkaddr; 133 #endif 134 u16 device_id_mask; /* valid bits of device id to be used */ 135 136 /* inode slot unit size in bit shift */ 137 unsigned char islotbits; 138 139 u32 sb_size; /* total superblock size */ 140 u32 build_time_nsec; 141 u64 build_time; 142 143 /* what we really care is nid, rather than ino.. */ 144 erofs_nid_t root_nid; 145 /* used for statfs, f_files - f_favail */ 146 u64 inos; 147 148 u8 uuid[16]; /* 128-bit uuid for volume */ 149 u8 volume_name[16]; /* volume name */ 150 u32 feature_compat; 151 u32 feature_incompat; 152 153 /* sysfs support */ 154 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */ 155 struct completion s_kobj_unregister; 156 157 /* fscache support */ 158 struct fscache_volume *volume; 159 struct erofs_fscache *s_fscache; 160 }; 161 162 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info) 163 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info) 164 165 /* Mount flags set via mount options or defaults */ 166 #define EROFS_MOUNT_XATTR_USER 0x00000010 167 #define EROFS_MOUNT_POSIX_ACL 0x00000020 168 #define EROFS_MOUNT_DAX_ALWAYS 0x00000040 169 #define EROFS_MOUNT_DAX_NEVER 0x00000080 170 171 #define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option) 172 #define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option) 173 #define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option) 174 175 static inline bool erofs_is_fscache_mode(struct super_block *sb) 176 { 177 return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && !sb->s_bdev; 178 } 179 180 enum { 181 EROFS_ZIP_CACHE_DISABLED, 182 EROFS_ZIP_CACHE_READAHEAD, 183 EROFS_ZIP_CACHE_READAROUND 184 }; 185 186 #ifdef CONFIG_EROFS_FS_ZIP 187 #define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL) 188 189 /* basic unit of the workstation of a super_block */ 190 struct erofs_workgroup { 191 /* the workgroup index in the workstation */ 192 pgoff_t index; 193 194 /* overall workgroup reference count */ 195 atomic_t refcount; 196 }; 197 198 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp, 199 int val) 200 { 201 preempt_disable(); 202 if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) { 203 preempt_enable(); 204 return false; 205 } 206 return true; 207 } 208 209 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp, 210 int orig_val) 211 { 212 /* 213 * other observers should notice all modifications 214 * in the freezing period. 215 */ 216 smp_mb(); 217 atomic_set(&grp->refcount, orig_val); 218 preempt_enable(); 219 } 220 221 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp) 222 { 223 return atomic_cond_read_relaxed(&grp->refcount, 224 VAL != EROFS_LOCKED_MAGIC); 225 } 226 #endif /* !CONFIG_EROFS_FS_ZIP */ 227 228 /* we strictly follow PAGE_SIZE and no buffer head yet */ 229 #define LOG_BLOCK_SIZE PAGE_SHIFT 230 231 #undef LOG_SECTORS_PER_BLOCK 232 #define LOG_SECTORS_PER_BLOCK (PAGE_SHIFT - 9) 233 234 #undef SECTORS_PER_BLOCK 235 #define SECTORS_PER_BLOCK (1 << SECTORS_PER_BLOCK) 236 237 #define EROFS_BLKSIZ (1 << LOG_BLOCK_SIZE) 238 239 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ) 240 #error erofs cannot be used in this platform 241 #endif 242 243 enum erofs_kmap_type { 244 EROFS_NO_KMAP, /* don't map the buffer */ 245 EROFS_KMAP, /* use kmap() to map the buffer */ 246 EROFS_KMAP_ATOMIC, /* use kmap_atomic() to map the buffer */ 247 }; 248 249 struct erofs_buf { 250 struct page *page; 251 void *base; 252 enum erofs_kmap_type kmap_type; 253 }; 254 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL }) 255 256 #define ROOT_NID(sb) ((sb)->root_nid) 257 258 #define erofs_blknr(addr) ((addr) / EROFS_BLKSIZ) 259 #define erofs_blkoff(addr) ((addr) % EROFS_BLKSIZ) 260 #define blknr_to_addr(nr) ((erofs_off_t)(nr) * EROFS_BLKSIZ) 261 262 static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid) 263 { 264 return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits); 265 } 266 267 #define EROFS_FEATURE_FUNCS(name, compat, feature) \ 268 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \ 269 { \ 270 return sbi->feature_##compat & EROFS_FEATURE_##feature; \ 271 } 272 273 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING) 274 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS) 275 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER) 276 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE) 277 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE) 278 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2) 279 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING) 280 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM) 281 282 /* atomic flag definitions */ 283 #define EROFS_I_EA_INITED_BIT 0 284 #define EROFS_I_Z_INITED_BIT 1 285 286 /* bitlock definitions (arranged in reverse order) */ 287 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1) 288 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2) 289 290 struct erofs_inode { 291 erofs_nid_t nid; 292 293 /* atomic flags (including bitlocks) */ 294 unsigned long flags; 295 296 unsigned char datalayout; 297 unsigned char inode_isize; 298 unsigned short xattr_isize; 299 300 unsigned int xattr_shared_count; 301 unsigned int *xattr_shared_xattrs; 302 303 union { 304 erofs_blk_t raw_blkaddr; 305 struct { 306 unsigned short chunkformat; 307 unsigned char chunkbits; 308 }; 309 #ifdef CONFIG_EROFS_FS_ZIP 310 struct { 311 unsigned short z_advise; 312 unsigned char z_algorithmtype[2]; 313 unsigned char z_logical_clusterbits; 314 unsigned long z_tailextent_headlcn; 315 erofs_off_t z_idataoff; 316 unsigned short z_idata_size; 317 }; 318 #endif /* CONFIG_EROFS_FS_ZIP */ 319 }; 320 /* the corresponding vfs inode */ 321 struct inode vfs_inode; 322 }; 323 324 #define EROFS_I(ptr) \ 325 container_of(ptr, struct erofs_inode, vfs_inode) 326 327 static inline unsigned long erofs_inode_datablocks(struct inode *inode) 328 { 329 /* since i_size cannot be changed */ 330 return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ); 331 } 332 333 static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit, 334 unsigned int bits) 335 { 336 337 return (value >> bit) & ((1 << bits) - 1); 338 } 339 340 341 static inline unsigned int erofs_inode_version(unsigned int value) 342 { 343 return erofs_bitrange(value, EROFS_I_VERSION_BIT, 344 EROFS_I_VERSION_BITS); 345 } 346 347 static inline unsigned int erofs_inode_datalayout(unsigned int value) 348 { 349 return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT, 350 EROFS_I_DATALAYOUT_BITS); 351 } 352 353 /* 354 * Different from grab_cache_page_nowait(), reclaiming is never triggered 355 * when allocating new pages. 356 */ 357 static inline 358 struct page *erofs_grab_cache_page_nowait(struct address_space *mapping, 359 pgoff_t index) 360 { 361 return pagecache_get_page(mapping, index, 362 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, 363 readahead_gfp_mask(mapping) & ~__GFP_RECLAIM); 364 } 365 366 extern const struct super_operations erofs_sops; 367 368 extern const struct address_space_operations erofs_raw_access_aops; 369 extern const struct address_space_operations z_erofs_aops; 370 371 enum { 372 BH_Encoded = BH_PrivateStart, 373 BH_FullMapped, 374 }; 375 376 /* Has a disk mapping */ 377 #define EROFS_MAP_MAPPED (1 << BH_Mapped) 378 /* Located in metadata (could be copied from bd_inode) */ 379 #define EROFS_MAP_META (1 << BH_Meta) 380 /* The extent is encoded */ 381 #define EROFS_MAP_ENCODED (1 << BH_Encoded) 382 /* The length of extent is full */ 383 #define EROFS_MAP_FULL_MAPPED (1 << BH_FullMapped) 384 385 struct erofs_map_blocks { 386 struct erofs_buf buf; 387 388 erofs_off_t m_pa, m_la; 389 u64 m_plen, m_llen; 390 391 unsigned short m_deviceid; 392 char m_algorithmformat; 393 unsigned int m_flags; 394 }; 395 396 /* Flags used by erofs_map_blocks_flatmode() */ 397 #define EROFS_GET_BLOCKS_RAW 0x0001 398 /* 399 * Used to get the exact decompressed length, e.g. fiemap (consider lookback 400 * approach instead if possible since it's more metadata lightweight.) 401 */ 402 #define EROFS_GET_BLOCKS_FIEMAP 0x0002 403 /* Used to map the whole extent if non-negligible data is requested for LZMA */ 404 #define EROFS_GET_BLOCKS_READMORE 0x0004 405 /* Used to map tail extent for tailpacking inline pcluster */ 406 #define EROFS_GET_BLOCKS_FINDTAIL 0x0008 407 408 enum { 409 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX, 410 Z_EROFS_COMPRESSION_RUNTIME_MAX 411 }; 412 413 /* zmap.c */ 414 extern const struct iomap_ops z_erofs_iomap_report_ops; 415 416 #ifdef CONFIG_EROFS_FS_ZIP 417 int z_erofs_fill_inode(struct inode *inode); 418 int z_erofs_map_blocks_iter(struct inode *inode, 419 struct erofs_map_blocks *map, 420 int flags); 421 #else 422 static inline int z_erofs_fill_inode(struct inode *inode) { return -EOPNOTSUPP; } 423 static inline int z_erofs_map_blocks_iter(struct inode *inode, 424 struct erofs_map_blocks *map, 425 int flags) 426 { 427 return -EOPNOTSUPP; 428 } 429 #endif /* !CONFIG_EROFS_FS_ZIP */ 430 431 struct erofs_map_dev { 432 struct erofs_fscache *m_fscache; 433 struct block_device *m_bdev; 434 struct dax_device *m_daxdev; 435 u64 m_dax_part_off; 436 437 erofs_off_t m_pa; 438 unsigned int m_deviceid; 439 }; 440 441 /* data.c */ 442 extern const struct file_operations erofs_file_fops; 443 void erofs_unmap_metabuf(struct erofs_buf *buf); 444 void erofs_put_metabuf(struct erofs_buf *buf); 445 void *erofs_bread(struct erofs_buf *buf, struct inode *inode, 446 erofs_blk_t blkaddr, enum erofs_kmap_type type); 447 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb, 448 erofs_blk_t blkaddr, enum erofs_kmap_type type); 449 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev); 450 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 451 u64 start, u64 len); 452 int erofs_map_blocks(struct inode *inode, 453 struct erofs_map_blocks *map, int flags); 454 455 /* inode.c */ 456 static inline unsigned long erofs_inode_hash(erofs_nid_t nid) 457 { 458 #if BITS_PER_LONG == 32 459 return (nid >> 32) ^ (nid & 0xffffffff); 460 #else 461 return nid; 462 #endif 463 } 464 465 extern const struct inode_operations erofs_generic_iops; 466 extern const struct inode_operations erofs_symlink_iops; 467 extern const struct inode_operations erofs_fast_symlink_iops; 468 469 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid, bool dir); 470 int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path, 471 struct kstat *stat, u32 request_mask, 472 unsigned int query_flags); 473 474 /* namei.c */ 475 extern const struct inode_operations erofs_dir_iops; 476 477 int erofs_namei(struct inode *dir, const struct qstr *name, 478 erofs_nid_t *nid, unsigned int *d_type); 479 480 /* dir.c */ 481 extern const struct file_operations erofs_dir_fops; 482 483 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count) 484 { 485 int retried = 0; 486 487 while (1) { 488 void *p = vm_map_ram(pages, count, -1); 489 490 /* retry two more times (totally 3 times) */ 491 if (p || ++retried >= 3) 492 return p; 493 vm_unmap_aliases(); 494 } 495 return NULL; 496 } 497 498 /* pcpubuf.c */ 499 void *erofs_get_pcpubuf(unsigned int requiredpages); 500 void erofs_put_pcpubuf(void *ptr); 501 int erofs_pcpubuf_growsize(unsigned int nrpages); 502 void erofs_pcpubuf_init(void); 503 void erofs_pcpubuf_exit(void); 504 505 /* sysfs.c */ 506 int erofs_register_sysfs(struct super_block *sb); 507 void erofs_unregister_sysfs(struct super_block *sb); 508 int __init erofs_init_sysfs(void); 509 void erofs_exit_sysfs(void); 510 511 /* utils.c / zdata.c */ 512 struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp); 513 static inline void erofs_pagepool_add(struct page **pagepool, 514 struct page *page) 515 { 516 set_page_private(page, (unsigned long)*pagepool); 517 *pagepool = page; 518 } 519 void erofs_release_pages(struct page **pagepool); 520 521 #ifdef CONFIG_EROFS_FS_ZIP 522 int erofs_workgroup_put(struct erofs_workgroup *grp); 523 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb, 524 pgoff_t index); 525 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb, 526 struct erofs_workgroup *grp); 527 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp); 528 void erofs_shrinker_register(struct super_block *sb); 529 void erofs_shrinker_unregister(struct super_block *sb); 530 int __init erofs_init_shrinker(void); 531 void erofs_exit_shrinker(void); 532 int __init z_erofs_init_zip_subsystem(void); 533 void z_erofs_exit_zip_subsystem(void); 534 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi, 535 struct erofs_workgroup *egrp); 536 int erofs_try_to_free_cached_page(struct page *page); 537 int z_erofs_load_lz4_config(struct super_block *sb, 538 struct erofs_super_block *dsb, 539 struct z_erofs_lz4_cfgs *lz4, int len); 540 #else 541 static inline void erofs_shrinker_register(struct super_block *sb) {} 542 static inline void erofs_shrinker_unregister(struct super_block *sb) {} 543 static inline int erofs_init_shrinker(void) { return 0; } 544 static inline void erofs_exit_shrinker(void) {} 545 static inline int z_erofs_init_zip_subsystem(void) { return 0; } 546 static inline void z_erofs_exit_zip_subsystem(void) {} 547 static inline int z_erofs_load_lz4_config(struct super_block *sb, 548 struct erofs_super_block *dsb, 549 struct z_erofs_lz4_cfgs *lz4, int len) 550 { 551 if (lz4 || dsb->u1.lz4_max_distance) { 552 erofs_err(sb, "lz4 algorithm isn't enabled"); 553 return -EINVAL; 554 } 555 return 0; 556 } 557 #endif /* !CONFIG_EROFS_FS_ZIP */ 558 559 #ifdef CONFIG_EROFS_FS_ZIP_LZMA 560 int z_erofs_lzma_init(void); 561 void z_erofs_lzma_exit(void); 562 int z_erofs_load_lzma_config(struct super_block *sb, 563 struct erofs_super_block *dsb, 564 struct z_erofs_lzma_cfgs *lzma, int size); 565 #else 566 static inline int z_erofs_lzma_init(void) { return 0; } 567 static inline int z_erofs_lzma_exit(void) { return 0; } 568 static inline int z_erofs_load_lzma_config(struct super_block *sb, 569 struct erofs_super_block *dsb, 570 struct z_erofs_lzma_cfgs *lzma, int size) { 571 if (lzma) { 572 erofs_err(sb, "lzma algorithm isn't enabled"); 573 return -EINVAL; 574 } 575 return 0; 576 } 577 #endif /* !CONFIG_EROFS_FS_ZIP */ 578 579 /* fscache.c */ 580 #ifdef CONFIG_EROFS_FS_ONDEMAND 581 int erofs_fscache_register_fs(struct super_block *sb); 582 void erofs_fscache_unregister_fs(struct super_block *sb); 583 584 int erofs_fscache_register_cookie(struct super_block *sb, 585 struct erofs_fscache **fscache, 586 char *name, bool need_inode); 587 void erofs_fscache_unregister_cookie(struct erofs_fscache **fscache); 588 589 extern const struct address_space_operations erofs_fscache_access_aops; 590 #else 591 static inline int erofs_fscache_register_fs(struct super_block *sb) 592 { 593 return 0; 594 } 595 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {} 596 597 static inline int erofs_fscache_register_cookie(struct super_block *sb, 598 struct erofs_fscache **fscache, 599 char *name, bool need_inode) 600 { 601 return -EOPNOTSUPP; 602 } 603 604 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache **fscache) 605 { 606 } 607 #endif 608 609 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 610 611 #endif /* __EROFS_INTERNAL_H */ 612