Loading...
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
45typedef u64 erofs_nid_t;
46typedef u64 erofs_off_t;
47/* data type for filesystem-wide blocks number */
48typedef u32 erofs_blk_t;
49
50struct 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
61enum {
62 EROFS_SYNC_DECOMPRESS_AUTO,
63 EROFS_SYNC_DECOMPRESS_FORCE_ON,
64 EROFS_SYNC_DECOMPRESS_FORCE_OFF
65};
66
67struct 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};
79
80struct erofs_dev_context {
81 struct idr tree;
82 struct rw_semaphore rwsem;
83
84 unsigned int extra_devices;
85};
86
87struct erofs_fs_context {
88 struct erofs_mount_opts opt;
89 struct erofs_dev_context *devs;
90 char *fsid;
91 char *domain_id;
92};
93
94/* all filesystem-wide lz4 configurations */
95struct erofs_sb_lz4_info {
96 /* # of pages needed for EROFS lz4 rolling decompression */
97 u16 max_distance_pages;
98 /* maximum possible blocks for pclusters in the filesystem */
99 u16 max_pclusterblks;
100};
101
102struct erofs_domain {
103 refcount_t ref;
104 struct list_head list;
105 struct fscache_volume *volume;
106 char *domain_id;
107};
108
109struct erofs_fscache {
110 struct fscache_cookie *cookie;
111 struct inode *inode;
112 struct inode *anon_inode;
113 struct erofs_domain *domain;
114 char *name;
115};
116
117struct erofs_sb_info {
118 struct erofs_mount_opts opt; /* options */
119#ifdef CONFIG_EROFS_FS_ZIP
120 /* list for all registered superblocks, mainly for shrinker */
121 struct list_head list;
122 struct mutex umount_mutex;
123
124 /* managed XArray arranged in physical block number */
125 struct xarray managed_pslots;
126
127 unsigned int shrinker_run_no;
128 u16 available_compr_algs;
129
130 /* pseudo inode to manage cached pages */
131 struct inode *managed_cache;
132
133 struct erofs_sb_lz4_info lz4;
134 struct inode *packed_inode;
135#endif /* CONFIG_EROFS_FS_ZIP */
136 struct erofs_dev_context *devs;
137 struct dax_device *dax_dev;
138 u64 dax_part_off;
139 u64 total_blocks;
140 u32 primarydevice_blocks;
141
142 u32 meta_blkaddr;
143#ifdef CONFIG_EROFS_FS_XATTR
144 u32 xattr_blkaddr;
145#endif
146 u16 device_id_mask; /* valid bits of device id to be used */
147
148 /* inode slot unit size in bit shift */
149 unsigned char islotbits;
150
151 u32 sb_size; /* total superblock size */
152 u32 build_time_nsec;
153 u64 build_time;
154
155 /* what we really care is nid, rather than ino.. */
156 erofs_nid_t root_nid;
157 /* used for statfs, f_files - f_favail */
158 u64 inos;
159
160 u8 uuid[16]; /* 128-bit uuid for volume */
161 u8 volume_name[16]; /* volume name */
162 u32 feature_compat;
163 u32 feature_incompat;
164
165 /* sysfs support */
166 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */
167 struct completion s_kobj_unregister;
168
169 /* fscache support */
170 struct fscache_volume *volume;
171 struct erofs_fscache *s_fscache;
172 struct erofs_domain *domain;
173 char *fsid;
174 char *domain_id;
175};
176
177#define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
178#define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
179
180/* Mount flags set via mount options or defaults */
181#define EROFS_MOUNT_XATTR_USER 0x00000010
182#define EROFS_MOUNT_POSIX_ACL 0x00000020
183#define EROFS_MOUNT_DAX_ALWAYS 0x00000040
184#define EROFS_MOUNT_DAX_NEVER 0x00000080
185
186#define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
187#define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option)
188#define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option)
189
190static inline bool erofs_is_fscache_mode(struct super_block *sb)
191{
192 return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && !sb->s_bdev;
193}
194
195enum {
196 EROFS_ZIP_CACHE_DISABLED,
197 EROFS_ZIP_CACHE_READAHEAD,
198 EROFS_ZIP_CACHE_READAROUND
199};
200
201#define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL)
202
203/* basic unit of the workstation of a super_block */
204struct erofs_workgroup {
205 /* the workgroup index in the workstation */
206 pgoff_t index;
207
208 /* overall workgroup reference count */
209 atomic_t refcount;
210};
211
212static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
213 int val)
214{
215 preempt_disable();
216 if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) {
217 preempt_enable();
218 return false;
219 }
220 return true;
221}
222
223static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
224 int orig_val)
225{
226 /*
227 * other observers should notice all modifications
228 * in the freezing period.
229 */
230 smp_mb();
231 atomic_set(&grp->refcount, orig_val);
232 preempt_enable();
233}
234
235static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
236{
237 return atomic_cond_read_relaxed(&grp->refcount,
238 VAL != EROFS_LOCKED_MAGIC);
239}
240
241/* we strictly follow PAGE_SIZE and no buffer head yet */
242#define LOG_BLOCK_SIZE PAGE_SHIFT
243
244#undef LOG_SECTORS_PER_BLOCK
245#define LOG_SECTORS_PER_BLOCK (PAGE_SHIFT - 9)
246
247#undef SECTORS_PER_BLOCK
248#define SECTORS_PER_BLOCK (1 << SECTORS_PER_BLOCK)
249
250#define EROFS_BLKSIZ (1 << LOG_BLOCK_SIZE)
251
252#if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ)
253#error erofs cannot be used in this platform
254#endif
255
256enum erofs_kmap_type {
257 EROFS_NO_KMAP, /* don't map the buffer */
258 EROFS_KMAP, /* use kmap_local_page() to map the buffer */
259};
260
261struct erofs_buf {
262 struct page *page;
263 void *base;
264 enum erofs_kmap_type kmap_type;
265};
266#define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL })
267
268#define ROOT_NID(sb) ((sb)->root_nid)
269
270#define erofs_blknr(addr) ((addr) / EROFS_BLKSIZ)
271#define erofs_blkoff(addr) ((addr) % EROFS_BLKSIZ)
272#define blknr_to_addr(nr) ((erofs_off_t)(nr) * EROFS_BLKSIZ)
273
274static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid)
275{
276 return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits);
277}
278
279#define EROFS_FEATURE_FUNCS(name, compat, feature) \
280static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
281{ \
282 return sbi->feature_##compat & EROFS_FEATURE_##feature; \
283}
284
285EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
286EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
287EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
288EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
289EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
290EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
291EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
292EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
293EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
294EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
295
296/* atomic flag definitions */
297#define EROFS_I_EA_INITED_BIT 0
298#define EROFS_I_Z_INITED_BIT 1
299
300/* bitlock definitions (arranged in reverse order) */
301#define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1)
302#define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2)
303
304struct erofs_inode {
305 erofs_nid_t nid;
306
307 /* atomic flags (including bitlocks) */
308 unsigned long flags;
309
310 unsigned char datalayout;
311 unsigned char inode_isize;
312 unsigned short xattr_isize;
313
314 unsigned int xattr_shared_count;
315 unsigned int *xattr_shared_xattrs;
316
317 union {
318 erofs_blk_t raw_blkaddr;
319 struct {
320 unsigned short chunkformat;
321 unsigned char chunkbits;
322 };
323#ifdef CONFIG_EROFS_FS_ZIP
324 struct {
325 unsigned short z_advise;
326 unsigned char z_algorithmtype[2];
327 unsigned char z_logical_clusterbits;
328 unsigned long z_tailextent_headlcn;
329 union {
330 struct {
331 erofs_off_t z_idataoff;
332 unsigned short z_idata_size;
333 };
334 erofs_off_t z_fragmentoff;
335 };
336 };
337#endif /* CONFIG_EROFS_FS_ZIP */
338 };
339 /* the corresponding vfs inode */
340 struct inode vfs_inode;
341};
342
343#define EROFS_I(ptr) \
344 container_of(ptr, struct erofs_inode, vfs_inode)
345
346static inline unsigned long erofs_inode_datablocks(struct inode *inode)
347{
348 /* since i_size cannot be changed */
349 return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
350}
351
352static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit,
353 unsigned int bits)
354{
355
356 return (value >> bit) & ((1 << bits) - 1);
357}
358
359
360static inline unsigned int erofs_inode_version(unsigned int value)
361{
362 return erofs_bitrange(value, EROFS_I_VERSION_BIT,
363 EROFS_I_VERSION_BITS);
364}
365
366static inline unsigned int erofs_inode_datalayout(unsigned int value)
367{
368 return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT,
369 EROFS_I_DATALAYOUT_BITS);
370}
371
372/*
373 * Different from grab_cache_page_nowait(), reclaiming is never triggered
374 * when allocating new pages.
375 */
376static inline
377struct page *erofs_grab_cache_page_nowait(struct address_space *mapping,
378 pgoff_t index)
379{
380 return pagecache_get_page(mapping, index,
381 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
382 readahead_gfp_mask(mapping) & ~__GFP_RECLAIM);
383}
384
385extern const struct super_operations erofs_sops;
386extern struct file_system_type erofs_fs_type;
387
388extern const struct address_space_operations erofs_raw_access_aops;
389extern const struct address_space_operations z_erofs_aops;
390
391enum {
392 BH_Encoded = BH_PrivateStart,
393 BH_FullMapped,
394 BH_Fragment,
395 BH_Partialref,
396};
397
398/* Has a disk mapping */
399#define EROFS_MAP_MAPPED (1 << BH_Mapped)
400/* Located in metadata (could be copied from bd_inode) */
401#define EROFS_MAP_META (1 << BH_Meta)
402/* The extent is encoded */
403#define EROFS_MAP_ENCODED (1 << BH_Encoded)
404/* The length of extent is full */
405#define EROFS_MAP_FULL_MAPPED (1 << BH_FullMapped)
406/* Located in the special packed inode */
407#define EROFS_MAP_FRAGMENT (1 << BH_Fragment)
408/* The extent refers to partial decompressed data */
409#define EROFS_MAP_PARTIAL_REF (1 << BH_Partialref)
410
411struct erofs_map_blocks {
412 struct erofs_buf buf;
413
414 erofs_off_t m_pa, m_la;
415 u64 m_plen, m_llen;
416
417 unsigned short m_deviceid;
418 char m_algorithmformat;
419 unsigned int m_flags;
420};
421
422/* Flags used by erofs_map_blocks_flatmode() */
423#define EROFS_GET_BLOCKS_RAW 0x0001
424/*
425 * Used to get the exact decompressed length, e.g. fiemap (consider lookback
426 * approach instead if possible since it's more metadata lightweight.)
427 */
428#define EROFS_GET_BLOCKS_FIEMAP 0x0002
429/* Used to map the whole extent if non-negligible data is requested for LZMA */
430#define EROFS_GET_BLOCKS_READMORE 0x0004
431/* Used to map tail extent for tailpacking inline or fragment pcluster */
432#define EROFS_GET_BLOCKS_FINDTAIL 0x0008
433
434enum {
435 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
436 Z_EROFS_COMPRESSION_INTERLACED,
437 Z_EROFS_COMPRESSION_RUNTIME_MAX
438};
439
440/* zmap.c */
441extern const struct iomap_ops z_erofs_iomap_report_ops;
442
443#ifdef CONFIG_EROFS_FS_ZIP
444int z_erofs_fill_inode(struct inode *inode);
445int z_erofs_map_blocks_iter(struct inode *inode,
446 struct erofs_map_blocks *map,
447 int flags);
448#else
449static inline int z_erofs_fill_inode(struct inode *inode) { return -EOPNOTSUPP; }
450static inline int z_erofs_map_blocks_iter(struct inode *inode,
451 struct erofs_map_blocks *map,
452 int flags)
453{
454 return -EOPNOTSUPP;
455}
456#endif /* !CONFIG_EROFS_FS_ZIP */
457
458struct erofs_map_dev {
459 struct erofs_fscache *m_fscache;
460 struct block_device *m_bdev;
461 struct dax_device *m_daxdev;
462 u64 m_dax_part_off;
463
464 erofs_off_t m_pa;
465 unsigned int m_deviceid;
466};
467
468/* data.c */
469extern const struct file_operations erofs_file_fops;
470void erofs_unmap_metabuf(struct erofs_buf *buf);
471void erofs_put_metabuf(struct erofs_buf *buf);
472void *erofs_bread(struct erofs_buf *buf, struct inode *inode,
473 erofs_blk_t blkaddr, enum erofs_kmap_type type);
474void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
475 erofs_blk_t blkaddr, enum erofs_kmap_type type);
476int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
477int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
478 u64 start, u64 len);
479int erofs_map_blocks(struct inode *inode,
480 struct erofs_map_blocks *map, int flags);
481
482/* inode.c */
483static inline unsigned long erofs_inode_hash(erofs_nid_t nid)
484{
485#if BITS_PER_LONG == 32
486 return (nid >> 32) ^ (nid & 0xffffffff);
487#else
488 return nid;
489#endif
490}
491
492extern const struct inode_operations erofs_generic_iops;
493extern const struct inode_operations erofs_symlink_iops;
494extern const struct inode_operations erofs_fast_symlink_iops;
495
496struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
497int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
498 struct kstat *stat, u32 request_mask,
499 unsigned int query_flags);
500
501/* namei.c */
502extern const struct inode_operations erofs_dir_iops;
503
504int erofs_namei(struct inode *dir, const struct qstr *name,
505 erofs_nid_t *nid, unsigned int *d_type);
506
507/* dir.c */
508extern const struct file_operations erofs_dir_fops;
509
510static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
511{
512 int retried = 0;
513
514 while (1) {
515 void *p = vm_map_ram(pages, count, -1);
516
517 /* retry two more times (totally 3 times) */
518 if (p || ++retried >= 3)
519 return p;
520 vm_unmap_aliases();
521 }
522 return NULL;
523}
524
525/* pcpubuf.c */
526void *erofs_get_pcpubuf(unsigned int requiredpages);
527void erofs_put_pcpubuf(void *ptr);
528int erofs_pcpubuf_growsize(unsigned int nrpages);
529void erofs_pcpubuf_init(void);
530void erofs_pcpubuf_exit(void);
531
532/* sysfs.c */
533int erofs_register_sysfs(struct super_block *sb);
534void erofs_unregister_sysfs(struct super_block *sb);
535int __init erofs_init_sysfs(void);
536void erofs_exit_sysfs(void);
537
538/* utils.c / zdata.c */
539struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp);
540static inline void erofs_pagepool_add(struct page **pagepool,
541 struct page *page)
542{
543 set_page_private(page, (unsigned long)*pagepool);
544 *pagepool = page;
545}
546void erofs_release_pages(struct page **pagepool);
547
548#ifdef CONFIG_EROFS_FS_ZIP
549int erofs_workgroup_put(struct erofs_workgroup *grp);
550struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
551 pgoff_t index);
552struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
553 struct erofs_workgroup *grp);
554void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
555void erofs_shrinker_register(struct super_block *sb);
556void erofs_shrinker_unregister(struct super_block *sb);
557int __init erofs_init_shrinker(void);
558void erofs_exit_shrinker(void);
559int __init z_erofs_init_zip_subsystem(void);
560void z_erofs_exit_zip_subsystem(void);
561int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
562 struct erofs_workgroup *egrp);
563int erofs_try_to_free_cached_page(struct page *page);
564int z_erofs_load_lz4_config(struct super_block *sb,
565 struct erofs_super_block *dsb,
566 struct z_erofs_lz4_cfgs *lz4, int len);
567#else
568static inline void erofs_shrinker_register(struct super_block *sb) {}
569static inline void erofs_shrinker_unregister(struct super_block *sb) {}
570static inline int erofs_init_shrinker(void) { return 0; }
571static inline void erofs_exit_shrinker(void) {}
572static inline int z_erofs_init_zip_subsystem(void) { return 0; }
573static inline void z_erofs_exit_zip_subsystem(void) {}
574static inline int z_erofs_load_lz4_config(struct super_block *sb,
575 struct erofs_super_block *dsb,
576 struct z_erofs_lz4_cfgs *lz4, int len)
577{
578 if (lz4 || dsb->u1.lz4_max_distance) {
579 erofs_err(sb, "lz4 algorithm isn't enabled");
580 return -EINVAL;
581 }
582 return 0;
583}
584#endif /* !CONFIG_EROFS_FS_ZIP */
585
586#ifdef CONFIG_EROFS_FS_ZIP_LZMA
587int z_erofs_lzma_init(void);
588void z_erofs_lzma_exit(void);
589int z_erofs_load_lzma_config(struct super_block *sb,
590 struct erofs_super_block *dsb,
591 struct z_erofs_lzma_cfgs *lzma, int size);
592#else
593static inline int z_erofs_lzma_init(void) { return 0; }
594static inline int z_erofs_lzma_exit(void) { return 0; }
595static inline int z_erofs_load_lzma_config(struct super_block *sb,
596 struct erofs_super_block *dsb,
597 struct z_erofs_lzma_cfgs *lzma, int size) {
598 if (lzma) {
599 erofs_err(sb, "lzma algorithm isn't enabled");
600 return -EINVAL;
601 }
602 return 0;
603}
604#endif /* !CONFIG_EROFS_FS_ZIP */
605
606/* flags for erofs_fscache_register_cookie() */
607#define EROFS_REG_COOKIE_NEED_INODE 1
608#define EROFS_REG_COOKIE_NEED_NOEXIST 2
609
610/* fscache.c */
611#ifdef CONFIG_EROFS_FS_ONDEMAND
612int erofs_fscache_register_fs(struct super_block *sb);
613void erofs_fscache_unregister_fs(struct super_block *sb);
614
615struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
616 char *name,
617 unsigned int flags);
618void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
619
620extern const struct address_space_operations erofs_fscache_access_aops;
621#else
622static inline int erofs_fscache_register_fs(struct super_block *sb)
623{
624 return -EOPNOTSUPP;
625}
626static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
627
628static inline
629struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
630 char *name,
631 unsigned int flags)
632{
633 return ERR_PTR(-EOPNOTSUPP);
634}
635
636static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
637{
638}
639#endif
640
641#define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
642
643#endif /* __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/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_info(sb, fmt, ...) \
27 _erofs_printk(sb, KERN_INFO fmt "\n", ##__VA_ARGS__)
28
29#ifdef CONFIG_EROFS_FS_DEBUG
30#define DBG_BUGON BUG_ON
31#else
32#define DBG_BUGON(x) ((void)(x))
33#endif /* !CONFIG_EROFS_FS_DEBUG */
34
35/* EROFS_SUPER_MAGIC_V1 to represent the whole file system */
36#define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1
37
38typedef u64 erofs_nid_t;
39typedef u64 erofs_off_t;
40/* data type for filesystem-wide blocks number */
41typedef u32 erofs_blk_t;
42
43struct erofs_device_info {
44 char *path;
45 struct erofs_fscache *fscache;
46 struct file *file;
47 struct dax_device *dax_dev;
48 u64 dax_part_off;
49
50 u32 blocks;
51 u32 mapped_blkaddr;
52};
53
54enum {
55 EROFS_SYNC_DECOMPRESS_AUTO,
56 EROFS_SYNC_DECOMPRESS_FORCE_ON,
57 EROFS_SYNC_DECOMPRESS_FORCE_OFF
58};
59
60struct erofs_mount_opts {
61 /* current strategy of how to use managed cache */
62 unsigned char cache_strategy;
63 /* strategy of sync decompression (0 - auto, 1 - force on, 2 - force off) */
64 unsigned int sync_decompress;
65 /* threshold for decompression synchronously */
66 unsigned int max_sync_decompress_pages;
67 unsigned int mount_opt;
68};
69
70struct erofs_dev_context {
71 struct idr tree;
72 struct rw_semaphore rwsem;
73
74 unsigned int extra_devices;
75 bool flatdev;
76};
77
78/* all filesystem-wide lz4 configurations */
79struct erofs_sb_lz4_info {
80 /* # of pages needed for EROFS lz4 rolling decompression */
81 u16 max_distance_pages;
82 /* maximum possible blocks for pclusters in the filesystem */
83 u16 max_pclusterblks;
84};
85
86struct erofs_domain {
87 refcount_t ref;
88 struct list_head list;
89 struct fscache_volume *volume;
90 char *domain_id;
91};
92
93struct erofs_fscache {
94 struct fscache_cookie *cookie;
95 struct inode *inode; /* anonymous inode for the blob */
96
97 /* used for share domain mode */
98 struct erofs_domain *domain;
99 struct list_head node;
100 refcount_t ref;
101 char *name;
102};
103
104struct erofs_xattr_prefix_item {
105 struct erofs_xattr_long_prefix *prefix;
106 u8 infix_len;
107};
108
109struct erofs_sb_info {
110 struct erofs_device_info dif0;
111 struct erofs_mount_opts opt; /* options */
112#ifdef CONFIG_EROFS_FS_ZIP
113 /* list for all registered superblocks, mainly for shrinker */
114 struct list_head list;
115 struct mutex umount_mutex;
116
117 /* managed XArray arranged in physical block number */
118 struct xarray managed_pslots;
119
120 unsigned int shrinker_run_no;
121 u16 available_compr_algs;
122
123 /* pseudo inode to manage cached pages */
124 struct inode *managed_cache;
125
126 struct erofs_sb_lz4_info lz4;
127#endif /* CONFIG_EROFS_FS_ZIP */
128 struct inode *packed_inode;
129 struct erofs_dev_context *devs;
130 u64 total_blocks;
131
132 u32 meta_blkaddr;
133#ifdef CONFIG_EROFS_FS_XATTR
134 u32 xattr_blkaddr;
135 u32 xattr_prefix_start;
136 u8 xattr_prefix_count;
137 struct erofs_xattr_prefix_item *xattr_prefixes;
138 unsigned int xattr_filter_reserved;
139#endif
140 u16 device_id_mask; /* valid bits of device id to be used */
141
142 unsigned char islotbits; /* inode slot unit size in bit shift */
143 unsigned char blkszbits; /* filesystem block size in bit shift */
144
145 u32 sb_size; /* total superblock size */
146 u32 build_time_nsec;
147 u64 build_time;
148
149 /* what we really care is nid, rather than ino.. */
150 erofs_nid_t root_nid;
151 erofs_nid_t packed_nid;
152 /* used for statfs, f_files - f_favail */
153 u64 inos;
154
155 u8 uuid[16]; /* 128-bit uuid for volume */
156 u8 volume_name[16]; /* volume name */
157 u32 feature_compat;
158 u32 feature_incompat;
159
160 /* sysfs support */
161 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */
162 struct completion s_kobj_unregister;
163
164 /* fscache support */
165 struct fscache_volume *volume;
166 struct erofs_domain *domain;
167 char *fsid;
168 char *domain_id;
169};
170
171#define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
172#define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
173
174/* Mount flags set via mount options or defaults */
175#define EROFS_MOUNT_XATTR_USER 0x00000010
176#define EROFS_MOUNT_POSIX_ACL 0x00000020
177#define EROFS_MOUNT_DAX_ALWAYS 0x00000040
178#define EROFS_MOUNT_DAX_NEVER 0x00000080
179#define EROFS_MOUNT_DIRECT_IO 0x00000100
180
181#define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
182#define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option)
183#define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option)
184
185static inline bool erofs_is_fileio_mode(struct erofs_sb_info *sbi)
186{
187 return IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) && sbi->dif0.file;
188}
189
190static inline bool erofs_is_fscache_mode(struct super_block *sb)
191{
192 return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) &&
193 !erofs_is_fileio_mode(EROFS_SB(sb)) && !sb->s_bdev;
194}
195
196enum {
197 EROFS_ZIP_CACHE_DISABLED,
198 EROFS_ZIP_CACHE_READAHEAD,
199 EROFS_ZIP_CACHE_READAROUND
200};
201
202enum erofs_kmap_type {
203 EROFS_NO_KMAP, /* don't map the buffer */
204 EROFS_KMAP, /* use kmap_local_page() to map the buffer */
205};
206
207struct erofs_buf {
208 struct address_space *mapping;
209 struct file *file;
210 struct page *page;
211 void *base;
212};
213#define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL })
214
215#define erofs_blknr(sb, addr) ((erofs_blk_t)((addr) >> (sb)->s_blocksize_bits))
216#define erofs_blkoff(sb, addr) ((addr) & ((sb)->s_blocksize - 1))
217#define erofs_pos(sb, blk) ((erofs_off_t)(blk) << (sb)->s_blocksize_bits)
218#define erofs_iblks(i) (round_up((i)->i_size, i_blocksize(i)) >> (i)->i_blkbits)
219
220#define EROFS_FEATURE_FUNCS(name, compat, feature) \
221static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
222{ \
223 return sbi->feature_##compat & EROFS_FEATURE_##feature; \
224}
225
226EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
227EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
228EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
229EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
230EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
231EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
232EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
233EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
234EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
235EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES)
236EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
237EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
238
239/* atomic flag definitions */
240#define EROFS_I_EA_INITED_BIT 0
241#define EROFS_I_Z_INITED_BIT 1
242
243/* bitlock definitions (arranged in reverse order) */
244#define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1)
245#define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2)
246
247struct erofs_inode {
248 erofs_nid_t nid;
249
250 /* atomic flags (including bitlocks) */
251 unsigned long flags;
252
253 unsigned char datalayout;
254 unsigned char inode_isize;
255 unsigned int xattr_isize;
256
257 unsigned int xattr_name_filter;
258 unsigned int xattr_shared_count;
259 unsigned int *xattr_shared_xattrs;
260
261 union {
262 erofs_blk_t raw_blkaddr;
263 struct {
264 unsigned short chunkformat;
265 unsigned char chunkbits;
266 };
267#ifdef CONFIG_EROFS_FS_ZIP
268 struct {
269 unsigned short z_advise;
270 unsigned char z_algorithmtype[2];
271 unsigned char z_logical_clusterbits;
272 unsigned long z_tailextent_headlcn;
273 union {
274 struct {
275 erofs_off_t z_idataoff;
276 unsigned short z_idata_size;
277 };
278 erofs_off_t z_fragmentoff;
279 };
280 };
281#endif /* CONFIG_EROFS_FS_ZIP */
282 };
283 /* the corresponding vfs inode */
284 struct inode vfs_inode;
285};
286
287#define EROFS_I(ptr) container_of(ptr, struct erofs_inode, vfs_inode)
288
289static inline erofs_off_t erofs_iloc(struct inode *inode)
290{
291 struct erofs_sb_info *sbi = EROFS_I_SB(inode);
292
293 return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
294 (EROFS_I(inode)->nid << sbi->islotbits);
295}
296
297static inline unsigned int erofs_inode_version(unsigned int ifmt)
298{
299 return (ifmt >> EROFS_I_VERSION_BIT) & EROFS_I_VERSION_MASK;
300}
301
302static inline unsigned int erofs_inode_datalayout(unsigned int ifmt)
303{
304 return (ifmt >> EROFS_I_DATALAYOUT_BIT) & EROFS_I_DATALAYOUT_MASK;
305}
306
307/* reclaiming is never triggered when allocating new folios. */
308static inline struct folio *erofs_grab_folio_nowait(struct address_space *as,
309 pgoff_t index)
310{
311 return __filemap_get_folio(as, index,
312 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
313 readahead_gfp_mask(as) & ~__GFP_RECLAIM);
314}
315
316/* Has a disk mapping */
317#define EROFS_MAP_MAPPED 0x0001
318/* Located in metadata (could be copied from bd_inode) */
319#define EROFS_MAP_META 0x0002
320/* The extent is encoded */
321#define EROFS_MAP_ENCODED 0x0004
322/* The length of extent is full */
323#define EROFS_MAP_FULL_MAPPED 0x0008
324/* Located in the special packed inode */
325#define EROFS_MAP_FRAGMENT 0x0010
326/* The extent refers to partial decompressed data */
327#define EROFS_MAP_PARTIAL_REF 0x0020
328
329struct erofs_map_blocks {
330 struct erofs_buf buf;
331
332 erofs_off_t m_pa, m_la;
333 u64 m_plen, m_llen;
334
335 unsigned short m_deviceid;
336 char m_algorithmformat;
337 unsigned int m_flags;
338};
339
340/*
341 * Used to get the exact decompressed length, e.g. fiemap (consider lookback
342 * approach instead if possible since it's more metadata lightweight.)
343 */
344#define EROFS_GET_BLOCKS_FIEMAP 0x0001
345/* Used to map the whole extent if non-negligible data is requested for LZMA */
346#define EROFS_GET_BLOCKS_READMORE 0x0002
347/* Used to map tail extent for tailpacking inline or fragment pcluster */
348#define EROFS_GET_BLOCKS_FINDTAIL 0x0004
349
350enum {
351 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
352 Z_EROFS_COMPRESSION_INTERLACED,
353 Z_EROFS_COMPRESSION_RUNTIME_MAX
354};
355
356struct erofs_map_dev {
357 struct super_block *m_sb;
358 struct erofs_device_info *m_dif;
359 struct block_device *m_bdev;
360
361 erofs_off_t m_pa;
362 unsigned int m_deviceid;
363};
364
365extern const struct super_operations erofs_sops;
366
367extern const struct address_space_operations erofs_aops;
368extern const struct address_space_operations erofs_fileio_aops;
369extern const struct address_space_operations z_erofs_aops;
370extern const struct address_space_operations erofs_fscache_access_aops;
371
372extern const struct inode_operations erofs_generic_iops;
373extern const struct inode_operations erofs_symlink_iops;
374extern const struct inode_operations erofs_fast_symlink_iops;
375extern const struct inode_operations erofs_dir_iops;
376
377extern const struct file_operations erofs_file_fops;
378extern const struct file_operations erofs_dir_fops;
379
380extern const struct iomap_ops z_erofs_iomap_report_ops;
381
382/* flags for erofs_fscache_register_cookie() */
383#define EROFS_REG_COOKIE_SHARE 0x0001
384#define EROFS_REG_COOKIE_NEED_NOEXIST 0x0002
385
386void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
387 erofs_off_t *offset, int *lengthp);
388void erofs_unmap_metabuf(struct erofs_buf *buf);
389void erofs_put_metabuf(struct erofs_buf *buf);
390void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset,
391 enum erofs_kmap_type type);
392void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
393void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
394 erofs_off_t offset, enum erofs_kmap_type type);
395int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
396int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
397 u64 start, u64 len);
398int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map);
399void erofs_onlinefolio_init(struct folio *folio);
400void erofs_onlinefolio_split(struct folio *folio);
401void erofs_onlinefolio_end(struct folio *folio, int err);
402struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
403int erofs_getattr(struct mnt_idmap *idmap, const struct path *path,
404 struct kstat *stat, u32 request_mask,
405 unsigned int query_flags);
406int erofs_namei(struct inode *dir, const struct qstr *name,
407 erofs_nid_t *nid, unsigned int *d_type);
408
409static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
410{
411 int retried = 0;
412
413 while (1) {
414 void *p = vm_map_ram(pages, count, -1);
415
416 /* retry two more times (totally 3 times) */
417 if (p || ++retried >= 3)
418 return p;
419 vm_unmap_aliases();
420 }
421 return NULL;
422}
423
424int erofs_register_sysfs(struct super_block *sb);
425void erofs_unregister_sysfs(struct super_block *sb);
426int __init erofs_init_sysfs(void);
427void erofs_exit_sysfs(void);
428
429struct page *__erofs_allocpage(struct page **pagepool, gfp_t gfp, bool tryrsv);
430static inline struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
431{
432 return __erofs_allocpage(pagepool, gfp, false);
433}
434static inline void erofs_pagepool_add(struct page **pagepool, struct page *page)
435{
436 set_page_private(page, (unsigned long)*pagepool);
437 *pagepool = page;
438}
439void erofs_release_pages(struct page **pagepool);
440
441#ifdef CONFIG_EROFS_FS_ZIP
442#define MNGD_MAPPING(sbi) ((sbi)->managed_cache->i_mapping)
443
444extern atomic_long_t erofs_global_shrink_cnt;
445void erofs_shrinker_register(struct super_block *sb);
446void erofs_shrinker_unregister(struct super_block *sb);
447int __init erofs_init_shrinker(void);
448void erofs_exit_shrinker(void);
449int __init z_erofs_init_subsystem(void);
450void z_erofs_exit_subsystem(void);
451unsigned long z_erofs_shrink_scan(struct erofs_sb_info *sbi,
452 unsigned long nr_shrink);
453int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map,
454 int flags);
455void *z_erofs_get_gbuf(unsigned int requiredpages);
456void z_erofs_put_gbuf(void *ptr);
457int z_erofs_gbuf_growsize(unsigned int nrpages);
458int __init z_erofs_gbuf_init(void);
459void z_erofs_gbuf_exit(void);
460int erofs_init_managed_cache(struct super_block *sb);
461int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb);
462#else
463static inline void erofs_shrinker_register(struct super_block *sb) {}
464static inline void erofs_shrinker_unregister(struct super_block *sb) {}
465static inline int erofs_init_shrinker(void) { return 0; }
466static inline void erofs_exit_shrinker(void) {}
467static inline int z_erofs_init_subsystem(void) { return 0; }
468static inline void z_erofs_exit_subsystem(void) {}
469static inline int erofs_init_managed_cache(struct super_block *sb) { return 0; }
470#endif /* !CONFIG_EROFS_FS_ZIP */
471
472#ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
473struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev);
474void erofs_fileio_submit_bio(struct bio *bio);
475#else
476static inline struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
477static inline void erofs_fileio_submit_bio(struct bio *bio) {}
478#endif
479
480#ifdef CONFIG_EROFS_FS_ONDEMAND
481int erofs_fscache_register_fs(struct super_block *sb);
482void erofs_fscache_unregister_fs(struct super_block *sb);
483
484struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
485 char *name, unsigned int flags);
486void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
487struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev);
488void erofs_fscache_submit_bio(struct bio *bio);
489#else
490static inline int erofs_fscache_register_fs(struct super_block *sb)
491{
492 return -EOPNOTSUPP;
493}
494static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
495
496static inline
497struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
498 char *name, unsigned int flags)
499{
500 return ERR_PTR(-EOPNOTSUPP);
501}
502
503static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
504{
505}
506static inline struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
507static inline void erofs_fscache_submit_bio(struct bio *bio) {}
508#endif
509
510#define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
511
512#endif /* __EROFS_INTERNAL_H */