1 /* SPDX-License-Identifier: GPL-2.0
3 * linux/drivers/staging/erofs/internal.h
5 * Copyright (C) 2017-2018 HUAWEI, Inc.
6 * http://www.huawei.com/
7 * Created by Gao Xiang <gaoxiang25@huawei.com>
9 * This file is subject to the terms and conditions of the GNU General Public
10 * License. See the file COPYING in the main directory of the Linux
11 * distribution for more details.
17 #include <linux/dcache.h>
19 #include <linux/pagemap.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/cleancache.h>
23 #include <linux/slab.h>
24 #include <linux/vmalloc.h>
27 /* redefine pr_fmt "erofs: " */
29 #define pr_fmt(fmt) "erofs: " fmt
31 #define errln(x, ...) pr_err(x "\n", ##__VA_ARGS__)
32 #define infoln(x, ...) pr_info(x "\n", ##__VA_ARGS__)
33 #ifdef CONFIG_EROFS_FS_DEBUG
34 #define debugln(x, ...) pr_debug(x "\n", ##__VA_ARGS__)
36 #define dbg_might_sleep might_sleep
37 #define DBG_BUGON BUG_ON
39 #define debugln(x, ...) ((void)0)
41 #define dbg_might_sleep() ((void)0)
42 #define DBG_BUGON(x) ((void)(x))
51 #ifdef CONFIG_EROFS_FAULT_INJECTION
52 extern const char *erofs_fault_name[FAULT_MAX];
53 #define IS_FAULT_SET(fi, type) ((fi)->inject_type & (1 << (type)))
55 struct erofs_fault_info {
57 unsigned int inject_rate;
58 unsigned int inject_type;
62 #ifdef CONFIG_EROFS_FS_ZIP_CACHE_BIPOLAR
63 #define EROFS_FS_ZIP_CACHE_LVL (2)
64 #elif defined(EROFS_FS_ZIP_CACHE_UNIPOLAR)
65 #define EROFS_FS_ZIP_CACHE_LVL (1)
67 #define EROFS_FS_ZIP_CACHE_LVL (0)
70 #if (!defined(EROFS_FS_HAS_MANAGED_CACHE) && (EROFS_FS_ZIP_CACHE_LVL > 0))
71 #define EROFS_FS_HAS_MANAGED_CACHE
74 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */
75 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1
77 typedef u64 erofs_nid_t;
79 struct erofs_sb_info {
80 /* list for all registered superblocks, mainly for shrinker */
81 struct list_head list;
82 struct mutex umount_mutex;
86 #ifdef CONFIG_EROFS_FS_XATTR
90 /* inode slot unit size in bit shift */
91 unsigned char islotbits;
92 #ifdef CONFIG_EROFS_FS_ZIP
93 /* cluster size in bit shift */
94 unsigned char clusterbits;
96 /* the dedicated workstation for compression */
97 struct radix_tree_root workstn_tree;
99 /* threshold for decompression synchronously */
100 unsigned int max_sync_decompress_pages;
102 #ifdef EROFS_FS_HAS_MANAGED_CACHE
103 struct inode *managed_cache;
111 /* what we really care is nid, rather than ino.. */
112 erofs_nid_t root_nid;
113 /* used for statfs, f_files - f_favail */
116 u8 uuid[16]; /* 128-bit uuid for volume */
117 u8 volume_name[16]; /* volume name */
122 unsigned int mount_opt;
123 unsigned int shrinker_run_no;
125 #ifdef CONFIG_EROFS_FAULT_INJECTION
126 struct erofs_fault_info fault_info; /* For fault injection */
130 #ifdef CONFIG_EROFS_FAULT_INJECTION
131 #define erofs_show_injection_info(type) \
132 infoln("inject %s in %s of %pS", erofs_fault_name[type], \
133 __func__, __builtin_return_address(0))
135 static inline bool time_to_inject(struct erofs_sb_info *sbi, int type)
137 struct erofs_fault_info *ffi = &sbi->fault_info;
139 if (!ffi->inject_rate)
142 if (!IS_FAULT_SET(ffi, type))
145 atomic_inc(&ffi->inject_ops);
146 if (atomic_read(&ffi->inject_ops) >= ffi->inject_rate) {
147 atomic_set(&ffi->inject_ops, 0);
153 static inline bool time_to_inject(struct erofs_sb_info *sbi, int type)
158 static inline void erofs_show_injection_info(int type)
163 static inline void *erofs_kmalloc(struct erofs_sb_info *sbi,
164 size_t size, gfp_t flags)
166 if (time_to_inject(sbi, FAULT_KMALLOC)) {
167 erofs_show_injection_info(FAULT_KMALLOC);
170 return kmalloc(size, flags);
173 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
174 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
176 /* Mount flags set via mount options or defaults */
177 #define EROFS_MOUNT_XATTR_USER 0x00000010
178 #define EROFS_MOUNT_POSIX_ACL 0x00000020
179 #define EROFS_MOUNT_FAULT_INJECTION 0x00000040
181 #define clear_opt(sbi, option) ((sbi)->mount_opt &= ~EROFS_MOUNT_##option)
182 #define set_opt(sbi, option) ((sbi)->mount_opt |= EROFS_MOUNT_##option)
183 #define test_opt(sbi, option) ((sbi)->mount_opt & EROFS_MOUNT_##option)
185 #ifdef CONFIG_EROFS_FS_ZIP
186 #define erofs_workstn_lock(sbi) xa_lock(&(sbi)->workstn_tree)
187 #define erofs_workstn_unlock(sbi) xa_unlock(&(sbi)->workstn_tree)
189 /* basic unit of the workstation of a super_block */
190 struct erofs_workgroup {
191 /* the workgroup index in the workstation */
194 /* overall workgroup reference count */
198 #define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL)
200 #if defined(CONFIG_SMP)
201 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
205 if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) {
212 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
216 * other observers should notice all modifications
217 * in the freezing period.
220 atomic_set(&grp->refcount, orig_val);
224 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
226 return atomic_cond_read_relaxed(&grp->refcount,
227 VAL != EROFS_LOCKED_MAGIC);
230 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
234 /* no need to spin on UP platforms, let's just disable preemption. */
235 if (val != atomic_read(&grp->refcount)) {
242 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
248 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
250 int v = atomic_read(&grp->refcount);
252 /* workgroup is never freezed on uniprocessor systems */
253 DBG_BUGON(v == EROFS_LOCKED_MAGIC);
258 int erofs_workgroup_put(struct erofs_workgroup *grp);
259 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
260 pgoff_t index, bool *tag);
261 int erofs_register_workgroup(struct super_block *sb,
262 struct erofs_workgroup *grp, bool tag);
263 unsigned long erofs_shrink_workstation(struct erofs_sb_info *sbi,
264 unsigned long nr_shrink, bool cleanup);
265 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
267 #ifdef EROFS_FS_HAS_MANAGED_CACHE
268 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
269 struct erofs_workgroup *egrp);
270 int erofs_try_to_free_cached_page(struct address_space *mapping,
273 #define MNGD_MAPPING(sbi) ((sbi)->managed_cache->i_mapping)
274 static inline bool erofs_page_is_managed(const struct erofs_sb_info *sbi,
277 return page->mapping == MNGD_MAPPING(sbi);
280 #define MNGD_MAPPING(sbi) (NULL)
281 static inline bool erofs_page_is_managed(const struct erofs_sb_info *sbi,
282 struct page *page) { return false; }
285 #define DEFAULT_MAX_SYNC_DECOMPRESS_PAGES 3
287 static inline bool __should_decompress_synchronously(struct erofs_sb_info *sbi,
290 return nr <= sbi->max_sync_decompress_pages;
293 int __init z_erofs_init_zip_subsystem(void);
294 void z_erofs_exit_zip_subsystem(void);
296 /* dummy initializer/finalizer for the decompression subsystem */
297 static inline int z_erofs_init_zip_subsystem(void) { return 0; }
298 static inline void z_erofs_exit_zip_subsystem(void) {}
301 /* we strictly follow PAGE_SIZE and no buffer head yet */
302 #define LOG_BLOCK_SIZE PAGE_SHIFT
304 #undef LOG_SECTORS_PER_BLOCK
305 #define LOG_SECTORS_PER_BLOCK (PAGE_SHIFT - 9)
307 #undef SECTORS_PER_BLOCK
308 #define SECTORS_PER_BLOCK (1 << SECTORS_PER_BLOCK)
310 #define EROFS_BLKSIZ (1 << LOG_BLOCK_SIZE)
312 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ)
313 #error erofs cannot be used in this platform
316 #define ROOT_NID(sb) ((sb)->root_nid)
318 #ifdef CONFIG_EROFS_FS_ZIP
319 /* hard limit of pages per compressed cluster */
320 #define Z_EROFS_CLUSTER_MAX_PAGES (CONFIG_EROFS_FS_CLUSTER_PAGE_LIMIT)
322 /* page count of a compressed cluster */
323 #define erofs_clusterpages(sbi) ((1 << (sbi)->clusterbits) / PAGE_SIZE)
325 #define EROFS_PCPUBUF_NR_PAGES Z_EROFS_CLUSTER_MAX_PAGES
327 #define EROFS_PCPUBUF_NR_PAGES 0
330 typedef u64 erofs_off_t;
332 /* data type for filesystem-wide blocks number */
333 typedef u32 erofs_blk_t;
335 #define erofs_blknr(addr) ((addr) / EROFS_BLKSIZ)
336 #define erofs_blkoff(addr) ((addr) % EROFS_BLKSIZ)
337 #define blknr_to_addr(nr) ((erofs_off_t)(nr) * EROFS_BLKSIZ)
339 static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid)
341 return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits);
344 /* atomic flag definitions */
345 #define EROFS_V_EA_INITED_BIT 0
346 #define EROFS_V_Z_INITED_BIT 1
348 /* bitlock definitions (arranged in reverse order) */
349 #define EROFS_V_BL_XATTR_BIT (BITS_PER_LONG - 1)
350 #define EROFS_V_BL_Z_BIT (BITS_PER_LONG - 2)
355 /* atomic flags (including bitlocks) */
358 unsigned char datamode;
359 unsigned char inode_isize;
360 unsigned short xattr_isize;
362 unsigned xattr_shared_count;
363 unsigned *xattr_shared_xattrs;
366 erofs_blk_t raw_blkaddr;
367 #ifdef CONFIG_EROFS_FS_ZIP
369 unsigned short z_advise;
370 unsigned char z_algorithmtype[2];
371 unsigned char z_logical_clusterbits;
372 unsigned char z_physical_clusterbits[2];
376 /* the corresponding vfs inode */
377 struct inode vfs_inode;
380 #define EROFS_V(ptr) \
381 container_of(ptr, struct erofs_vnode, vfs_inode)
383 #define __inode_advise(x, bit, bits) \
384 (((x) >> (bit)) & ((1 << (bits)) - 1))
386 #define __inode_version(advise) \
387 __inode_advise(advise, EROFS_I_VERSION_BIT, \
388 EROFS_I_VERSION_BITS)
390 #define __inode_data_mapping(advise) \
391 __inode_advise(advise, EROFS_I_DATA_MAPPING_BIT,\
392 EROFS_I_DATA_MAPPING_BITS)
394 static inline unsigned long inode_datablocks(struct inode *inode)
396 /* since i_size cannot be changed */
397 return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
400 static inline bool is_inode_layout_compression(struct inode *inode)
402 return erofs_inode_is_data_compressed(EROFS_V(inode)->datamode);
405 static inline bool is_inode_flat_inline(struct inode *inode)
407 return EROFS_V(inode)->datamode == EROFS_INODE_FLAT_INLINE;
410 extern const struct super_operations erofs_sops;
412 extern const struct address_space_operations erofs_raw_access_aops;
413 #ifdef CONFIG_EROFS_FS_ZIP
414 extern const struct address_space_operations z_erofs_vle_normalaccess_aops;
418 * Logical to physical block mapping, used by erofs_map_blocks()
420 * Different with other file systems, it is used for 2 access modes:
422 * 1) RAW access mode:
424 * Users pass a valid (m_lblk, m_lofs -- usually 0) pair,
425 * and get the valid m_pblk, m_pofs and the longest m_len(in bytes).
427 * Note that m_lblk in the RAW access mode refers to the number of
428 * the compressed ondisk block rather than the uncompressed
429 * in-memory block for the compressed file.
431 * m_pofs equals to m_lofs except for the inline data page.
433 * 2) Normal access mode:
435 * If the inode is not compressed, it has no difference with
436 * the RAW access mode. However, if the inode is compressed,
437 * users should pass a valid (m_lblk, m_lofs) pair, and get
438 * the needed m_pblk, m_pofs, m_len to get the compressed data
439 * and the updated m_lblk, m_lofs which indicates the start
440 * of the corresponding uncompressed data in the file.
443 BH_Zipped = BH_PrivateStart,
447 /* Has a disk mapping */
448 #define EROFS_MAP_MAPPED (1 << BH_Mapped)
449 /* Located in metadata (could be copied from bd_inode) */
450 #define EROFS_MAP_META (1 << BH_Meta)
451 /* The extent has been compressed */
452 #define EROFS_MAP_ZIPPED (1 << BH_Zipped)
453 /* The length of extent is full */
454 #define EROFS_MAP_FULL_MAPPED (1 << BH_FullMapped)
456 struct erofs_map_blocks {
457 erofs_off_t m_pa, m_la;
460 unsigned int m_flags;
465 /* Flags used by erofs_map_blocks() */
466 #define EROFS_GET_BLOCKS_RAW 0x0001
469 #ifdef CONFIG_EROFS_FS_ZIP
470 int z_erofs_fill_inode(struct inode *inode);
471 int z_erofs_map_blocks_iter(struct inode *inode,
472 struct erofs_map_blocks *map,
475 static inline int z_erofs_fill_inode(struct inode *inode) { return -ENOTSUPP; }
476 static inline int z_erofs_map_blocks_iter(struct inode *inode,
477 struct erofs_map_blocks *map,
485 static inline struct bio *
486 erofs_grab_bio(struct super_block *sb,
487 erofs_blk_t blkaddr, unsigned int nr_pages, void *bi_private,
488 bio_end_io_t endio, bool nofail)
490 const gfp_t gfp = GFP_NOIO;
495 bio = bio_alloc(gfp | (nofail ? __GFP_NOFAIL : 0), 1);
496 if (unlikely(!bio)) {
498 return ERR_PTR(-ENOMEM);
502 bio = bio_alloc(gfp, nr_pages);
504 } while (unlikely(!bio));
506 bio->bi_end_io = endio;
507 bio_set_dev(bio, sb->s_bdev);
508 bio->bi_iter.bi_sector = (sector_t)blkaddr << LOG_SECTORS_PER_BLOCK;
509 bio->bi_private = bi_private;
513 static inline void __submit_bio(struct bio *bio, unsigned op, unsigned op_flags)
515 bio_set_op_attrs(bio, op, op_flags);
519 #ifndef CONFIG_EROFS_FS_IO_MAX_RETRIES
520 #define EROFS_IO_MAX_RETRIES_NOFAIL 0
522 #define EROFS_IO_MAX_RETRIES_NOFAIL CONFIG_EROFS_FS_IO_MAX_RETRIES
525 struct page *__erofs_get_meta_page(struct super_block *sb, erofs_blk_t blkaddr,
526 bool prio, bool nofail);
528 static inline struct page *erofs_get_meta_page(struct super_block *sb,
529 erofs_blk_t blkaddr, bool prio)
531 return __erofs_get_meta_page(sb, blkaddr, prio, false);
534 static inline struct page *erofs_get_meta_page_nofail(struct super_block *sb,
535 erofs_blk_t blkaddr, bool prio)
537 return __erofs_get_meta_page(sb, blkaddr, prio, true);
540 int erofs_map_blocks(struct inode *, struct erofs_map_blocks *, int);
542 static inline struct page *
543 erofs_get_inline_page(struct inode *inode,
546 return erofs_get_meta_page(inode->i_sb,
547 blkaddr, S_ISDIR(inode->i_mode));
551 static inline unsigned long erofs_inode_hash(erofs_nid_t nid)
553 #if BITS_PER_LONG == 32
554 return (nid >> 32) ^ (nid & 0xffffffff);
560 extern const struct inode_operations erofs_generic_iops;
561 extern const struct inode_operations erofs_symlink_iops;
562 extern const struct inode_operations erofs_fast_symlink_iops;
564 static inline void set_inode_fast_symlink(struct inode *inode)
566 inode->i_op = &erofs_fast_symlink_iops;
569 static inline bool is_inode_fast_symlink(struct inode *inode)
571 return inode->i_op == &erofs_fast_symlink_iops;
574 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid, bool dir);
575 int erofs_getattr(const struct path *path, struct kstat *stat,
576 u32 request_mask, unsigned int query_flags);
579 extern const struct inode_operations erofs_dir_iops;
581 int erofs_namei(struct inode *dir, struct qstr *name,
582 erofs_nid_t *nid, unsigned int *d_type);
585 extern const struct file_operations erofs_dir_fops;
587 static inline void *erofs_vmap(struct page **pages, unsigned int count)
589 #ifdef CONFIG_EROFS_FS_USE_VM_MAP_RAM
593 void *addr = vm_map_ram(pages, count, -1, PAGE_KERNEL);
594 /* retry two more times (totally 3 times) */
595 if (addr || ++i >= 3)
601 return vmap(pages, count, VM_MAP, PAGE_KERNEL);
605 static inline void erofs_vunmap(const void *mem, unsigned int count)
607 #ifdef CONFIG_EROFS_FS_USE_VM_MAP_RAM
608 vm_unmap_ram(mem, count);
615 extern struct shrinker erofs_shrinker_info;
617 struct page *erofs_allocpage(struct list_head *pool, gfp_t gfp);
619 #if (EROFS_PCPUBUF_NR_PAGES > 0)
620 void *erofs_get_pcpubuf(unsigned int pagenr);
621 #define erofs_put_pcpubuf(buf) do { \
626 static inline void *erofs_get_pcpubuf(unsigned int pagenr)
628 return ERR_PTR(-ENOTSUPP);
631 #define erofs_put_pcpubuf(buf) do {} while (0)
634 void erofs_register_super(struct super_block *sb);
635 void erofs_unregister_super(struct super_block *sb);
638 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))