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