1 /* SPDX-License-Identifier: GPL-2.0-only */ 2 /* 3 * Copyright (C) 2017-2018 HUAWEI, Inc. 4 * https://www.huawei.com/ 5 * Created by Gao Xiang <gaoxiang25@huawei.com> 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 "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 erofs_dbg(x, ...) pr_debug(x "\n", ##__VA_ARGS__) 35 #define DBG_BUGON BUG_ON 36 #else 37 #define erofs_dbg(x, ...) ((void)0) 38 #define DBG_BUGON(x) ((void)(x)) 39 #endif /* !CONFIG_EROFS_FS_DEBUG */ 40 41 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */ 42 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1 43 44 typedef u64 erofs_nid_t; 45 typedef u64 erofs_off_t; 46 /* data type for filesystem-wide blocks number */ 47 typedef u32 erofs_blk_t; 48 49 struct erofs_fs_context { 50 #ifdef CONFIG_EROFS_FS_ZIP 51 /* current strategy of how to use managed cache */ 52 unsigned char cache_strategy; 53 /* strategy of sync decompression (false - auto, true - force on) */ 54 bool readahead_sync_decompress; 55 56 /* threshold for decompression synchronously */ 57 unsigned int max_sync_decompress_pages; 58 #endif 59 unsigned int mount_opt; 60 }; 61 62 /* all filesystem-wide lz4 configurations */ 63 struct erofs_sb_lz4_info { 64 /* # of pages needed for EROFS lz4 rolling decompression */ 65 u16 max_distance_pages; 66 /* maximum possible blocks for pclusters in the filesystem */ 67 u16 max_pclusterblks; 68 }; 69 70 struct erofs_sb_info { 71 #ifdef CONFIG_EROFS_FS_ZIP 72 /* list for all registered superblocks, mainly for shrinker */ 73 struct list_head list; 74 struct mutex umount_mutex; 75 76 /* managed XArray arranged in physical block number */ 77 struct xarray managed_pslots; 78 79 unsigned int shrinker_run_no; 80 u16 available_compr_algs; 81 82 /* pseudo inode to manage cached pages */ 83 struct inode *managed_cache; 84 85 struct erofs_sb_lz4_info lz4; 86 #endif /* CONFIG_EROFS_FS_ZIP */ 87 u32 blocks; 88 u32 meta_blkaddr; 89 #ifdef CONFIG_EROFS_FS_XATTR 90 u32 xattr_blkaddr; 91 #endif 92 93 /* inode slot unit size in bit shift */ 94 unsigned char islotbits; 95 96 u32 sb_size; /* total superblock size */ 97 u32 build_time_nsec; 98 u64 build_time; 99 100 /* what we really care is nid, rather than ino.. */ 101 erofs_nid_t root_nid; 102 /* used for statfs, f_files - f_favail */ 103 u64 inos; 104 105 u8 uuid[16]; /* 128-bit uuid for volume */ 106 u8 volume_name[16]; /* volume name */ 107 u32 feature_compat; 108 u32 feature_incompat; 109 110 struct erofs_fs_context ctx; /* options */ 111 }; 112 113 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info) 114 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info) 115 116 /* Mount flags set via mount options or defaults */ 117 #define EROFS_MOUNT_XATTR_USER 0x00000010 118 #define EROFS_MOUNT_POSIX_ACL 0x00000020 119 120 #define clear_opt(ctx, option) ((ctx)->mount_opt &= ~EROFS_MOUNT_##option) 121 #define set_opt(ctx, option) ((ctx)->mount_opt |= EROFS_MOUNT_##option) 122 #define test_opt(ctx, option) ((ctx)->mount_opt & EROFS_MOUNT_##option) 123 124 enum { 125 EROFS_ZIP_CACHE_DISABLED, 126 EROFS_ZIP_CACHE_READAHEAD, 127 EROFS_ZIP_CACHE_READAROUND 128 }; 129 130 #ifdef CONFIG_EROFS_FS_ZIP 131 #define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL) 132 133 /* basic unit of the workstation of a super_block */ 134 struct erofs_workgroup { 135 /* the workgroup index in the workstation */ 136 pgoff_t index; 137 138 /* overall workgroup reference count */ 139 atomic_t refcount; 140 }; 141 142 #if defined(CONFIG_SMP) 143 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp, 144 int val) 145 { 146 preempt_disable(); 147 if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) { 148 preempt_enable(); 149 return false; 150 } 151 return true; 152 } 153 154 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp, 155 int orig_val) 156 { 157 /* 158 * other observers should notice all modifications 159 * in the freezing period. 160 */ 161 smp_mb(); 162 atomic_set(&grp->refcount, orig_val); 163 preempt_enable(); 164 } 165 166 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp) 167 { 168 return atomic_cond_read_relaxed(&grp->refcount, 169 VAL != EROFS_LOCKED_MAGIC); 170 } 171 #else 172 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp, 173 int val) 174 { 175 preempt_disable(); 176 /* no need to spin on UP platforms, let's just disable preemption. */ 177 if (val != atomic_read(&grp->refcount)) { 178 preempt_enable(); 179 return false; 180 } 181 return true; 182 } 183 184 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp, 185 int orig_val) 186 { 187 preempt_enable(); 188 } 189 190 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp) 191 { 192 int v = atomic_read(&grp->refcount); 193 194 /* workgroup is never freezed on uniprocessor systems */ 195 DBG_BUGON(v == EROFS_LOCKED_MAGIC); 196 return v; 197 } 198 #endif /* !CONFIG_SMP */ 199 #endif /* !CONFIG_EROFS_FS_ZIP */ 200 201 /* we strictly follow PAGE_SIZE and no buffer head yet */ 202 #define LOG_BLOCK_SIZE PAGE_SHIFT 203 204 #undef LOG_SECTORS_PER_BLOCK 205 #define LOG_SECTORS_PER_BLOCK (PAGE_SHIFT - 9) 206 207 #undef SECTORS_PER_BLOCK 208 #define SECTORS_PER_BLOCK (1 << SECTORS_PER_BLOCK) 209 210 #define EROFS_BLKSIZ (1 << LOG_BLOCK_SIZE) 211 212 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ) 213 #error erofs cannot be used in this platform 214 #endif 215 216 #define ROOT_NID(sb) ((sb)->root_nid) 217 218 #define erofs_blknr(addr) ((addr) / EROFS_BLKSIZ) 219 #define erofs_blkoff(addr) ((addr) % EROFS_BLKSIZ) 220 #define blknr_to_addr(nr) ((erofs_off_t)(nr) * EROFS_BLKSIZ) 221 222 static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid) 223 { 224 return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits); 225 } 226 227 #define EROFS_FEATURE_FUNCS(name, compat, feature) \ 228 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \ 229 { \ 230 return sbi->feature_##compat & EROFS_FEATURE_##feature; \ 231 } 232 233 EROFS_FEATURE_FUNCS(lz4_0padding, incompat, INCOMPAT_LZ4_0PADDING) 234 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS) 235 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER) 236 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM) 237 238 /* atomic flag definitions */ 239 #define EROFS_I_EA_INITED_BIT 0 240 #define EROFS_I_Z_INITED_BIT 1 241 242 /* bitlock definitions (arranged in reverse order) */ 243 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1) 244 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2) 245 246 struct erofs_inode { 247 erofs_nid_t nid; 248 249 /* atomic flags (including bitlocks) */ 250 unsigned long flags; 251 252 unsigned char datalayout; 253 unsigned char inode_isize; 254 unsigned short xattr_isize; 255 256 unsigned int xattr_shared_count; 257 unsigned int *xattr_shared_xattrs; 258 259 union { 260 erofs_blk_t raw_blkaddr; 261 #ifdef CONFIG_EROFS_FS_ZIP 262 struct { 263 unsigned short z_advise; 264 unsigned char z_algorithmtype[2]; 265 unsigned char z_logical_clusterbits; 266 }; 267 #endif /* CONFIG_EROFS_FS_ZIP */ 268 }; 269 /* the corresponding vfs inode */ 270 struct inode vfs_inode; 271 }; 272 273 #define EROFS_I(ptr) \ 274 container_of(ptr, struct erofs_inode, vfs_inode) 275 276 static inline unsigned long erofs_inode_datablocks(struct inode *inode) 277 { 278 /* since i_size cannot be changed */ 279 return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ); 280 } 281 282 static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit, 283 unsigned int bits) 284 { 285 286 return (value >> bit) & ((1 << bits) - 1); 287 } 288 289 290 static inline unsigned int erofs_inode_version(unsigned int value) 291 { 292 return erofs_bitrange(value, EROFS_I_VERSION_BIT, 293 EROFS_I_VERSION_BITS); 294 } 295 296 static inline unsigned int erofs_inode_datalayout(unsigned int value) 297 { 298 return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT, 299 EROFS_I_DATALAYOUT_BITS); 300 } 301 302 extern const struct super_operations erofs_sops; 303 304 extern const struct address_space_operations erofs_raw_access_aops; 305 extern const struct address_space_operations z_erofs_aops; 306 307 /* 308 * Logical to physical block mapping 309 * 310 * Different with other file systems, it is used for 2 access modes: 311 * 312 * 1) RAW access mode: 313 * 314 * Users pass a valid (m_lblk, m_lofs -- usually 0) pair, 315 * and get the valid m_pblk, m_pofs and the longest m_len(in bytes). 316 * 317 * Note that m_lblk in the RAW access mode refers to the number of 318 * the compressed ondisk block rather than the uncompressed 319 * in-memory block for the compressed file. 320 * 321 * m_pofs equals to m_lofs except for the inline data page. 322 * 323 * 2) Normal access mode: 324 * 325 * If the inode is not compressed, it has no difference with 326 * the RAW access mode. However, if the inode is compressed, 327 * users should pass a valid (m_lblk, m_lofs) pair, and get 328 * the needed m_pblk, m_pofs, m_len to get the compressed data 329 * and the updated m_lblk, m_lofs which indicates the start 330 * of the corresponding uncompressed data in the file. 331 */ 332 enum { 333 BH_Zipped = BH_PrivateStart, 334 BH_FullMapped, 335 }; 336 337 /* Has a disk mapping */ 338 #define EROFS_MAP_MAPPED (1 << BH_Mapped) 339 /* Located in metadata (could be copied from bd_inode) */ 340 #define EROFS_MAP_META (1 << BH_Meta) 341 /* The extent has been compressed */ 342 #define EROFS_MAP_ZIPPED (1 << BH_Zipped) 343 /* The length of extent is full */ 344 #define EROFS_MAP_FULL_MAPPED (1 << BH_FullMapped) 345 346 struct erofs_map_blocks { 347 erofs_off_t m_pa, m_la; 348 u64 m_plen, m_llen; 349 350 unsigned int m_flags; 351 352 struct page *mpage; 353 }; 354 355 /* Flags used by erofs_map_blocks_flatmode() */ 356 #define EROFS_GET_BLOCKS_RAW 0x0001 357 358 /* zmap.c */ 359 #ifdef CONFIG_EROFS_FS_ZIP 360 int z_erofs_fill_inode(struct inode *inode); 361 int z_erofs_map_blocks_iter(struct inode *inode, 362 struct erofs_map_blocks *map, 363 int flags); 364 #else 365 static inline int z_erofs_fill_inode(struct inode *inode) { return -EOPNOTSUPP; } 366 static inline int z_erofs_map_blocks_iter(struct inode *inode, 367 struct erofs_map_blocks *map, 368 int flags) 369 { 370 return -EOPNOTSUPP; 371 } 372 #endif /* !CONFIG_EROFS_FS_ZIP */ 373 374 /* data.c */ 375 struct page *erofs_get_meta_page(struct super_block *sb, erofs_blk_t blkaddr); 376 377 /* inode.c */ 378 static inline unsigned long erofs_inode_hash(erofs_nid_t nid) 379 { 380 #if BITS_PER_LONG == 32 381 return (nid >> 32) ^ (nid & 0xffffffff); 382 #else 383 return nid; 384 #endif 385 } 386 387 extern const struct inode_operations erofs_generic_iops; 388 extern const struct inode_operations erofs_symlink_iops; 389 extern const struct inode_operations erofs_fast_symlink_iops; 390 391 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid, bool dir); 392 int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path, 393 struct kstat *stat, u32 request_mask, 394 unsigned int query_flags); 395 396 /* namei.c */ 397 extern const struct inode_operations erofs_dir_iops; 398 399 int erofs_namei(struct inode *dir, struct qstr *name, 400 erofs_nid_t *nid, unsigned int *d_type); 401 402 /* dir.c */ 403 extern const struct file_operations erofs_dir_fops; 404 405 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count) 406 { 407 int retried = 0; 408 409 while (1) { 410 void *p = vm_map_ram(pages, count, -1); 411 412 /* retry two more times (totally 3 times) */ 413 if (p || ++retried >= 3) 414 return p; 415 vm_unmap_aliases(); 416 } 417 return NULL; 418 } 419 420 /* pcpubuf.c */ 421 void *erofs_get_pcpubuf(unsigned int requiredpages); 422 void erofs_put_pcpubuf(void *ptr); 423 int erofs_pcpubuf_growsize(unsigned int nrpages); 424 void erofs_pcpubuf_init(void); 425 void erofs_pcpubuf_exit(void); 426 427 /* utils.c / zdata.c */ 428 struct page *erofs_allocpage(struct list_head *pool, gfp_t gfp); 429 430 #ifdef CONFIG_EROFS_FS_ZIP 431 int erofs_workgroup_put(struct erofs_workgroup *grp); 432 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb, 433 pgoff_t index); 434 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb, 435 struct erofs_workgroup *grp); 436 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp); 437 void erofs_shrinker_register(struct super_block *sb); 438 void erofs_shrinker_unregister(struct super_block *sb); 439 int __init erofs_init_shrinker(void); 440 void erofs_exit_shrinker(void); 441 int __init z_erofs_init_zip_subsystem(void); 442 void z_erofs_exit_zip_subsystem(void); 443 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi, 444 struct erofs_workgroup *egrp); 445 int erofs_try_to_free_cached_page(struct address_space *mapping, 446 struct page *page); 447 int z_erofs_load_lz4_config(struct super_block *sb, 448 struct erofs_super_block *dsb, 449 struct z_erofs_lz4_cfgs *lz4, int len); 450 #else 451 static inline void erofs_shrinker_register(struct super_block *sb) {} 452 static inline void erofs_shrinker_unregister(struct super_block *sb) {} 453 static inline int erofs_init_shrinker(void) { return 0; } 454 static inline void erofs_exit_shrinker(void) {} 455 static inline int z_erofs_init_zip_subsystem(void) { return 0; } 456 static inline void z_erofs_exit_zip_subsystem(void) {} 457 static inline int z_erofs_load_lz4_config(struct super_block *sb, 458 struct erofs_super_block *dsb, 459 struct z_erofs_lz4_cfgs *lz4, int len) 460 { 461 if (lz4 || dsb->u1.lz4_max_distance) { 462 erofs_err(sb, "lz4 algorithm isn't enabled"); 463 return -EINVAL; 464 } 465 return 0; 466 } 467 #endif /* !CONFIG_EROFS_FS_ZIP */ 468 469 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 470 471 #endif /* __EROFS_INTERNAL_H */ 472 473