fs/erofs/internal.h

   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_fs_context {
  51 #ifdef CONFIG_EROFS_FS_ZIP
  52         /* current strategy of how to use managed cache */
  53         unsigned char cache_strategy;
  54         /* strategy of sync decompression (false - auto, true - force on) */
  55         bool readahead_sync_decompress;
  56
  57         /* threshold for decompression synchronously */
  58         unsigned int max_sync_decompress_pages;
  59 #endif
  60         unsigned int mount_opt;
  61 };
  62
  63 /* all filesystem-wide lz4 configurations */
  64 struct erofs_sb_lz4_info {
  65         /* # of pages needed for EROFS lz4 rolling decompression */
  66         u16 max_distance_pages;
  67         /* maximum possible blocks for pclusters in the filesystem */
  68         u16 max_pclusterblks;
  69 };
  70
  71 struct erofs_sb_info {
  72 #ifdef CONFIG_EROFS_FS_ZIP
  73         /* list for all registered superblocks, mainly for shrinker */
  74         struct list_head list;
  75         struct mutex umount_mutex;
  76
  77         /* managed XArray arranged in physical block number */
  78         struct xarray managed_pslots;
  79
  80         unsigned int shrinker_run_no;
  81         u16 available_compr_algs;
  82
  83         /* pseudo inode to manage cached pages */
  84         struct inode *managed_cache;
  85
  86         struct erofs_sb_lz4_info lz4;
  87 #endif  /* CONFIG_EROFS_FS_ZIP */
  88         struct dax_device *dax_dev;
  89         u32 blocks;
  90         u32 meta_blkaddr;
  91 #ifdef CONFIG_EROFS_FS_XATTR
  92         u32 xattr_blkaddr;
  93 #endif
  94
  95         /* inode slot unit size in bit shift */
  96         unsigned char islotbits;
  97
  98         u32 sb_size;                    /* total superblock size */
  99         u32 build_time_nsec;
 100         u64 build_time;
 101
 102         /* what we really care is nid, rather than ino.. */
 103         erofs_nid_t root_nid;
 104         /* used for statfs, f_files - f_favail */
 105         u64 inos;
 106
 107         u8 uuid[16];                    /* 128-bit uuid for volume */
 108         u8 volume_name[16];             /* volume name */
 109         u32 feature_compat;
 110         u32 feature_incompat;
 111
 112         struct erofs_fs_context ctx;    /* options */
 113 };
 114
 115 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
 116 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
 117
 118 /* Mount flags set via mount options or defaults */
 119 #define EROFS_MOUNT_XATTR_USER          0x00000010
 120 #define EROFS_MOUNT_POSIX_ACL           0x00000020
 121 #define EROFS_MOUNT_DAX_ALWAYS          0x00000040
 122 #define EROFS_MOUNT_DAX_NEVER           0x00000080
 123
 124 #define clear_opt(ctx, option)  ((ctx)->mount_opt &= ~EROFS_MOUNT_##option)
 125 #define set_opt(ctx, option)    ((ctx)->mount_opt |= EROFS_MOUNT_##option)
 126 #define test_opt(ctx, option)   ((ctx)->mount_opt & EROFS_MOUNT_##option)
 127
 128 enum {
 129         EROFS_ZIP_CACHE_DISABLED,
 130         EROFS_ZIP_CACHE_READAHEAD,
 131         EROFS_ZIP_CACHE_READAROUND
 132 };
 133
 134 #ifdef CONFIG_EROFS_FS_ZIP
 135 #define EROFS_LOCKED_MAGIC     (INT_MIN | 0xE0F510CCL)
 136
 137 /* basic unit of the workstation of a super_block */
 138 struct erofs_workgroup {
 139         /* the workgroup index in the workstation */
 140         pgoff_t index;
 141
 142         /* overall workgroup reference count */
 143         atomic_t refcount;
 144 };
 145
 146 #if defined(CONFIG_SMP)
 147 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
 148                                                  int val)
 149 {
 150         preempt_disable();
 151         if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) {
 152                 preempt_enable();
 153                 return false;
 154         }
 155         return true;
 156 }
 157
 158 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
 159                                             int orig_val)
 160 {
 161         /*
 162          * other observers should notice all modifications
 163          * in the freezing period.
 164          */
 165         smp_mb();
 166         atomic_set(&grp->refcount, orig_val);
 167         preempt_enable();
 168 }
 169
 170 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
 171 {
 172         return atomic_cond_read_relaxed(&grp->refcount,
 173                                         VAL != EROFS_LOCKED_MAGIC);
 174 }
 175 #else
 176 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
 177                                                  int val)
 178 {
 179         preempt_disable();
 180         /* no need to spin on UP platforms, let's just disable preemption. */
 181         if (val != atomic_read(&grp->refcount)) {
 182                 preempt_enable();
 183                 return false;
 184         }
 185         return true;
 186 }
 187
 188 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
 189                                             int orig_val)
 190 {
 191         preempt_enable();
 192 }
 193
 194 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
 195 {
 196         int v = atomic_read(&grp->refcount);
 197
 198         /* workgroup is never freezed on uniprocessor systems */
 199         DBG_BUGON(v == EROFS_LOCKED_MAGIC);
 200         return v;
 201 }
 202 #endif  /* !CONFIG_SMP */
 203 #endif  /* !CONFIG_EROFS_FS_ZIP */
 204
 205 /* we strictly follow PAGE_SIZE and no buffer head yet */
 206 #define LOG_BLOCK_SIZE          PAGE_SHIFT
 207
 208 #undef LOG_SECTORS_PER_BLOCK
 209 #define LOG_SECTORS_PER_BLOCK   (PAGE_SHIFT - 9)
 210
 211 #undef SECTORS_PER_BLOCK
 212 #define SECTORS_PER_BLOCK       (1 << SECTORS_PER_BLOCK)
 213
 214 #define EROFS_BLKSIZ            (1 << LOG_BLOCK_SIZE)
 215
 216 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ)
 217 #error erofs cannot be used in this platform
 218 #endif
 219
 220 #define ROOT_NID(sb)            ((sb)->root_nid)
 221
 222 #define erofs_blknr(addr)       ((addr) / EROFS_BLKSIZ)
 223 #define erofs_blkoff(addr)      ((addr) % EROFS_BLKSIZ)
 224 #define blknr_to_addr(nr)       ((erofs_off_t)(nr) * EROFS_BLKSIZ)
 225
 226 static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid)
 227 {
 228         return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits);
 229 }
 230
 231 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
 232 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
 233 { \
 234         return sbi->feature_##compat & EROFS_FEATURE_##feature; \
 235 }
 236
 237 EROFS_FEATURE_FUNCS(lz4_0padding, incompat, INCOMPAT_LZ4_0PADDING)
 238 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
 239 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
 240 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
 241
 242 /* atomic flag definitions */
 243 #define EROFS_I_EA_INITED_BIT   0
 244 #define EROFS_I_Z_INITED_BIT    1
 245
 246 /* bitlock definitions (arranged in reverse order) */
 247 #define EROFS_I_BL_XATTR_BIT    (BITS_PER_LONG - 1)
 248 #define EROFS_I_BL_Z_BIT        (BITS_PER_LONG - 2)
 249
 250 struct erofs_inode {
 251         erofs_nid_t nid;
 252
 253         /* atomic flags (including bitlocks) */
 254         unsigned long flags;
 255
 256         unsigned char datalayout;
 257         unsigned char inode_isize;
 258         unsigned short xattr_isize;
 259
 260         unsigned int xattr_shared_count;
 261         unsigned int *xattr_shared_xattrs;
 262
 263         union {
 264                 erofs_blk_t raw_blkaddr;
 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_logical_clusterbits;
 274                 };
 275 #endif  /* CONFIG_EROFS_FS_ZIP */
 276         };
 277         /* the corresponding vfs inode */
 278         struct inode vfs_inode;
 279 };
 280
 281 #define EROFS_I(ptr)    \
 282         container_of(ptr, struct erofs_inode, vfs_inode)
 283
 284 static inline unsigned long erofs_inode_datablocks(struct inode *inode)
 285 {
 286         /* since i_size cannot be changed */
 287         return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
 288 }
 289
 290 static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit,
 291                                           unsigned int bits)
 292 {
 293
 294         return (value >> bit) & ((1 << bits) - 1);
 295 }
 296
 297
 298 static inline unsigned int erofs_inode_version(unsigned int value)
 299 {
 300         return erofs_bitrange(value, EROFS_I_VERSION_BIT,
 301                               EROFS_I_VERSION_BITS);
 302 }
 303
 304 static inline unsigned int erofs_inode_datalayout(unsigned int value)
 305 {
 306         return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT,
 307                               EROFS_I_DATALAYOUT_BITS);
 308 }
 309
 310 extern const struct super_operations erofs_sops;
 311
 312 extern const struct address_space_operations erofs_raw_access_aops;
 313 extern const struct address_space_operations z_erofs_aops;
 314
 315 /*
 316  * Logical to physical block mapping
 317  *
 318  * Different with other file systems, it is used for 2 access modes:
 319  *
 320  * 1) RAW access mode:
 321  *
 322  * Users pass a valid (m_lblk, m_lofs -- usually 0) pair,
 323  * and get the valid m_pblk, m_pofs and the longest m_len(in bytes).
 324  *
 325  * Note that m_lblk in the RAW access mode refers to the number of
 326  * the compressed ondisk block rather than the uncompressed
 327  * in-memory block for the compressed file.
 328  *
 329  * m_pofs equals to m_lofs except for the inline data page.
 330  *
 331  * 2) Normal access mode:
 332  *
 333  * If the inode is not compressed, it has no difference with
 334  * the RAW access mode. However, if the inode is compressed,
 335  * users should pass a valid (m_lblk, m_lofs) pair, and get
 336  * the needed m_pblk, m_pofs, m_len to get the compressed data
 337  * and the updated m_lblk, m_lofs which indicates the start
 338  * of the corresponding uncompressed data in the file.
 339  */
 340 enum {
 341         BH_Zipped = BH_PrivateStart,
 342         BH_FullMapped,
 343 };
 344
 345 /* Has a disk mapping */
 346 #define EROFS_MAP_MAPPED        (1 << BH_Mapped)
 347 /* Located in metadata (could be copied from bd_inode) */
 348 #define EROFS_MAP_META          (1 << BH_Meta)
 349 /* The extent has been compressed */
 350 #define EROFS_MAP_ZIPPED        (1 << BH_Zipped)
 351 /* The length of extent is full */
 352 #define EROFS_MAP_FULL_MAPPED   (1 << BH_FullMapped)
 353
 354 struct erofs_map_blocks {
 355         erofs_off_t m_pa, m_la;
 356         u64 m_plen, m_llen;
 357
 358         unsigned int m_flags;
 359
 360         struct page *mpage;
 361 };
 362
 363 /* Flags used by erofs_map_blocks_flatmode() */
 364 #define EROFS_GET_BLOCKS_RAW    0x0001
 365 /*
 366  * Used to get the exact decompressed length, e.g. fiemap (consider lookback
 367  * approach instead if possible since it's more metadata lightweight.)
 368  */
 369 #define EROFS_GET_BLOCKS_FIEMAP 0x0002
 370
 371 /* zmap.c */
 372 extern const struct iomap_ops z_erofs_iomap_report_ops;
 373
 374 #ifdef CONFIG_EROFS_FS_ZIP
 375 int z_erofs_fill_inode(struct inode *inode);
 376 int z_erofs_map_blocks_iter(struct inode *inode,
 377                             struct erofs_map_blocks *map,
 378                             int flags);
 379 #else
 380 static inline int z_erofs_fill_inode(struct inode *inode) { return -EOPNOTSUPP; }
 381 static inline int z_erofs_map_blocks_iter(struct inode *inode,
 382                                           struct erofs_map_blocks *map,
 383                                           int flags)
 384 {
 385         return -EOPNOTSUPP;
 386 }
 387 #endif  /* !CONFIG_EROFS_FS_ZIP */
 388
 389 /* data.c */
 390 extern const struct file_operations erofs_file_fops;
 391 struct page *erofs_get_meta_page(struct super_block *sb, erofs_blk_t blkaddr);
 392 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 393                  u64 start, u64 len);
 394
 395 /* inode.c */
 396 static inline unsigned long erofs_inode_hash(erofs_nid_t nid)
 397 {
 398 #if BITS_PER_LONG == 32
 399         return (nid >> 32) ^ (nid & 0xffffffff);
 400 #else
 401         return nid;
 402 #endif
 403 }
 404
 405 extern const struct inode_operations erofs_generic_iops;
 406 extern const struct inode_operations erofs_symlink_iops;
 407 extern const struct inode_operations erofs_fast_symlink_iops;
 408
 409 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid, bool dir);
 410 int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 411                   struct kstat *stat, u32 request_mask,
 412                   unsigned int query_flags);
 413
 414 /* namei.c */
 415 extern const struct inode_operations erofs_dir_iops;
 416
 417 int erofs_namei(struct inode *dir, struct qstr *name,
 418                 erofs_nid_t *nid, unsigned int *d_type);
 419
 420 /* dir.c */
 421 extern const struct file_operations erofs_dir_fops;
 422
 423 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
 424 {
 425         int retried = 0;
 426
 427         while (1) {
 428                 void *p = vm_map_ram(pages, count, -1);
 429
 430                 /* retry two more times (totally 3 times) */
 431                 if (p || ++retried >= 3)
 432                         return p;
 433                 vm_unmap_aliases();
 434         }
 435         return NULL;
 436 }
 437
 438 /* pcpubuf.c */
 439 void *erofs_get_pcpubuf(unsigned int requiredpages);
 440 void erofs_put_pcpubuf(void *ptr);
 441 int erofs_pcpubuf_growsize(unsigned int nrpages);
 442 void erofs_pcpubuf_init(void);
 443 void erofs_pcpubuf_exit(void);
 444
 445 /* utils.c / zdata.c */
 446 struct page *erofs_allocpage(struct list_head *pool, gfp_t gfp);
 447
 448 #ifdef CONFIG_EROFS_FS_ZIP
 449 int erofs_workgroup_put(struct erofs_workgroup *grp);
 450 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
 451                                              pgoff_t index);
 452 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
 453                                                struct erofs_workgroup *grp);
 454 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
 455 void erofs_shrinker_register(struct super_block *sb);
 456 void erofs_shrinker_unregister(struct super_block *sb);
 457 int __init erofs_init_shrinker(void);
 458 void erofs_exit_shrinker(void);
 459 int __init z_erofs_init_zip_subsystem(void);
 460 void z_erofs_exit_zip_subsystem(void);
 461 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
 462                                        struct erofs_workgroup *egrp);
 463 int erofs_try_to_free_cached_page(struct page *page);
 464 int z_erofs_load_lz4_config(struct super_block *sb,
 465                             struct erofs_super_block *dsb,
 466                             struct z_erofs_lz4_cfgs *lz4, int len);
 467 #else
 468 static inline void erofs_shrinker_register(struct super_block *sb) {}
 469 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
 470 static inline int erofs_init_shrinker(void) { return 0; }
 471 static inline void erofs_exit_shrinker(void) {}
 472 static inline int z_erofs_init_zip_subsystem(void) { return 0; }
 473 static inline void z_erofs_exit_zip_subsystem(void) {}
 474 static inline int z_erofs_load_lz4_config(struct super_block *sb,
 475                                   struct erofs_super_block *dsb,
 476                                   struct z_erofs_lz4_cfgs *lz4, int len)
 477 {
 478         if (lz4 || dsb->u1.lz4_max_distance) {
 479                 erofs_err(sb, "lz4 algorithm isn't enabled");
 480                 return -EINVAL;
 481         }
 482         return 0;
 483 }
 484 #endif  /* !CONFIG_EROFS_FS_ZIP */
 485
 486 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
 487
 488 #endif  /* __EROFS_INTERNAL_H */