xref: /openbmc/linux/fs/erofs/internal.h (revision 86db9f28)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Copyright (C) 2017-2018 HUAWEI, Inc.
4  *             http://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_sb_info {
50 #ifdef CONFIG_EROFS_FS_ZIP
51 	/* list for all registered superblocks, mainly for shrinker */
52 	struct list_head list;
53 	struct mutex umount_mutex;
54 
55 	/* the dedicated workstation for compression */
56 	struct radix_tree_root workstn_tree;
57 
58 	/* threshold for decompression synchronously */
59 	unsigned int max_sync_decompress_pages;
60 
61 	unsigned int shrinker_run_no;
62 
63 	/* current strategy of how to use managed cache */
64 	unsigned char cache_strategy;
65 
66 	/* pseudo inode to manage cached pages */
67 	struct inode *managed_cache;
68 #endif	/* CONFIG_EROFS_FS_ZIP */
69 	u32 blocks;
70 	u32 meta_blkaddr;
71 #ifdef CONFIG_EROFS_FS_XATTR
72 	u32 xattr_blkaddr;
73 #endif
74 
75 	/* inode slot unit size in bit shift */
76 	unsigned char islotbits;
77 
78 	u32 build_time_nsec;
79 	u64 build_time;
80 
81 	/* what we really care is nid, rather than ino.. */
82 	erofs_nid_t root_nid;
83 	/* used for statfs, f_files - f_favail */
84 	u64 inos;
85 
86 	u8 uuid[16];                    /* 128-bit uuid for volume */
87 	u8 volume_name[16];             /* volume name */
88 	u32 feature_incompat;
89 
90 	unsigned int mount_opt;
91 };
92 
93 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
94 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
95 
96 /* Mount flags set via mount options or defaults */
97 #define EROFS_MOUNT_XATTR_USER		0x00000010
98 #define EROFS_MOUNT_POSIX_ACL		0x00000020
99 
100 #define clear_opt(sbi, option)	((sbi)->mount_opt &= ~EROFS_MOUNT_##option)
101 #define set_opt(sbi, option)	((sbi)->mount_opt |= EROFS_MOUNT_##option)
102 #define test_opt(sbi, option)	((sbi)->mount_opt & EROFS_MOUNT_##option)
103 
104 #ifdef CONFIG_EROFS_FS_ZIP
105 enum {
106 	EROFS_ZIP_CACHE_DISABLED,
107 	EROFS_ZIP_CACHE_READAHEAD,
108 	EROFS_ZIP_CACHE_READAROUND
109 };
110 
111 #define EROFS_LOCKED_MAGIC     (INT_MIN | 0xE0F510CCL)
112 
113 /* basic unit of the workstation of a super_block */
114 struct erofs_workgroup {
115 	/* the workgroup index in the workstation */
116 	pgoff_t index;
117 
118 	/* overall workgroup reference count */
119 	atomic_t refcount;
120 };
121 
122 #if defined(CONFIG_SMP)
123 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
124 						 int val)
125 {
126 	preempt_disable();
127 	if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) {
128 		preempt_enable();
129 		return false;
130 	}
131 	return true;
132 }
133 
134 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
135 					    int orig_val)
136 {
137 	/*
138 	 * other observers should notice all modifications
139 	 * in the freezing period.
140 	 */
141 	smp_mb();
142 	atomic_set(&grp->refcount, orig_val);
143 	preempt_enable();
144 }
145 
146 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
147 {
148 	return atomic_cond_read_relaxed(&grp->refcount,
149 					VAL != EROFS_LOCKED_MAGIC);
150 }
151 #else
152 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
153 						 int val)
154 {
155 	preempt_disable();
156 	/* no need to spin on UP platforms, let's just disable preemption. */
157 	if (val != atomic_read(&grp->refcount)) {
158 		preempt_enable();
159 		return false;
160 	}
161 	return true;
162 }
163 
164 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
165 					    int orig_val)
166 {
167 	preempt_enable();
168 }
169 
170 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
171 {
172 	int v = atomic_read(&grp->refcount);
173 
174 	/* workgroup is never freezed on uniprocessor systems */
175 	DBG_BUGON(v == EROFS_LOCKED_MAGIC);
176 	return v;
177 }
178 #endif	/* !CONFIG_SMP */
179 
180 /* hard limit of pages per compressed cluster */
181 #define Z_EROFS_CLUSTER_MAX_PAGES       (CONFIG_EROFS_FS_CLUSTER_PAGE_LIMIT)
182 #define EROFS_PCPUBUF_NR_PAGES          Z_EROFS_CLUSTER_MAX_PAGES
183 #else
184 #define EROFS_PCPUBUF_NR_PAGES          0
185 #endif	/* !CONFIG_EROFS_FS_ZIP */
186 
187 /* we strictly follow PAGE_SIZE and no buffer head yet */
188 #define LOG_BLOCK_SIZE		PAGE_SHIFT
189 
190 #undef LOG_SECTORS_PER_BLOCK
191 #define LOG_SECTORS_PER_BLOCK	(PAGE_SHIFT - 9)
192 
193 #undef SECTORS_PER_BLOCK
194 #define SECTORS_PER_BLOCK	(1 << SECTORS_PER_BLOCK)
195 
196 #define EROFS_BLKSIZ		(1 << LOG_BLOCK_SIZE)
197 
198 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ)
199 #error erofs cannot be used in this platform
200 #endif
201 
202 #define ROOT_NID(sb)		((sb)->root_nid)
203 
204 #define erofs_blknr(addr)       ((addr) / EROFS_BLKSIZ)
205 #define erofs_blkoff(addr)      ((addr) % EROFS_BLKSIZ)
206 #define blknr_to_addr(nr)       ((erofs_off_t)(nr) * EROFS_BLKSIZ)
207 
208 static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid)
209 {
210 	return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits);
211 }
212 
213 /* atomic flag definitions */
214 #define EROFS_I_EA_INITED_BIT	0
215 #define EROFS_I_Z_INITED_BIT	1
216 
217 /* bitlock definitions (arranged in reverse order) */
218 #define EROFS_I_BL_XATTR_BIT	(BITS_PER_LONG - 1)
219 #define EROFS_I_BL_Z_BIT	(BITS_PER_LONG - 2)
220 
221 struct erofs_inode {
222 	erofs_nid_t nid;
223 
224 	/* atomic flags (including bitlocks) */
225 	unsigned long flags;
226 
227 	unsigned char datalayout;
228 	unsigned char inode_isize;
229 	unsigned short xattr_isize;
230 
231 	unsigned int xattr_shared_count;
232 	unsigned int *xattr_shared_xattrs;
233 
234 	union {
235 		erofs_blk_t raw_blkaddr;
236 #ifdef CONFIG_EROFS_FS_ZIP
237 		struct {
238 			unsigned short z_advise;
239 			unsigned char  z_algorithmtype[2];
240 			unsigned char  z_logical_clusterbits;
241 			unsigned char  z_physical_clusterbits[2];
242 		};
243 #endif	/* CONFIG_EROFS_FS_ZIP */
244 	};
245 	/* the corresponding vfs inode */
246 	struct inode vfs_inode;
247 };
248 
249 #define EROFS_I(ptr)	\
250 	container_of(ptr, struct erofs_inode, vfs_inode)
251 
252 static inline unsigned long erofs_inode_datablocks(struct inode *inode)
253 {
254 	/* since i_size cannot be changed */
255 	return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
256 }
257 
258 static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit,
259 					  unsigned int bits)
260 {
261 
262 	return (value >> bit) & ((1 << bits) - 1);
263 }
264 
265 
266 static inline unsigned int erofs_inode_version(unsigned int value)
267 {
268 	return erofs_bitrange(value, EROFS_I_VERSION_BIT,
269 			      EROFS_I_VERSION_BITS);
270 }
271 
272 static inline unsigned int erofs_inode_datalayout(unsigned int value)
273 {
274 	return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT,
275 			      EROFS_I_DATALAYOUT_BITS);
276 }
277 
278 extern const struct super_operations erofs_sops;
279 
280 extern const struct address_space_operations erofs_raw_access_aops;
281 #ifdef CONFIG_EROFS_FS_ZIP
282 extern const struct address_space_operations z_erofs_vle_normalaccess_aops;
283 #endif
284 
285 /*
286  * Logical to physical block mapping, used by erofs_map_blocks()
287  *
288  * Different with other file systems, it is used for 2 access modes:
289  *
290  * 1) RAW access mode:
291  *
292  * Users pass a valid (m_lblk, m_lofs -- usually 0) pair,
293  * and get the valid m_pblk, m_pofs and the longest m_len(in bytes).
294  *
295  * Note that m_lblk in the RAW access mode refers to the number of
296  * the compressed ondisk block rather than the uncompressed
297  * in-memory block for the compressed file.
298  *
299  * m_pofs equals to m_lofs except for the inline data page.
300  *
301  * 2) Normal access mode:
302  *
303  * If the inode is not compressed, it has no difference with
304  * the RAW access mode. However, if the inode is compressed,
305  * users should pass a valid (m_lblk, m_lofs) pair, and get
306  * the needed m_pblk, m_pofs, m_len to get the compressed data
307  * and the updated m_lblk, m_lofs which indicates the start
308  * of the corresponding uncompressed data in the file.
309  */
310 enum {
311 	BH_Zipped = BH_PrivateStart,
312 	BH_FullMapped,
313 };
314 
315 /* Has a disk mapping */
316 #define EROFS_MAP_MAPPED	(1 << BH_Mapped)
317 /* Located in metadata (could be copied from bd_inode) */
318 #define EROFS_MAP_META		(1 << BH_Meta)
319 /* The extent has been compressed */
320 #define EROFS_MAP_ZIPPED	(1 << BH_Zipped)
321 /* The length of extent is full */
322 #define EROFS_MAP_FULL_MAPPED	(1 << BH_FullMapped)
323 
324 struct erofs_map_blocks {
325 	erofs_off_t m_pa, m_la;
326 	u64 m_plen, m_llen;
327 
328 	unsigned int m_flags;
329 
330 	struct page *mpage;
331 };
332 
333 /* Flags used by erofs_map_blocks() */
334 #define EROFS_GET_BLOCKS_RAW    0x0001
335 
336 /* zmap.c */
337 #ifdef CONFIG_EROFS_FS_ZIP
338 int z_erofs_fill_inode(struct inode *inode);
339 int z_erofs_map_blocks_iter(struct inode *inode,
340 			    struct erofs_map_blocks *map,
341 			    int flags);
342 #else
343 static inline int z_erofs_fill_inode(struct inode *inode) { return -EOPNOTSUPP; }
344 static inline int z_erofs_map_blocks_iter(struct inode *inode,
345 					  struct erofs_map_blocks *map,
346 					  int flags)
347 {
348 	return -EOPNOTSUPP;
349 }
350 #endif	/* !CONFIG_EROFS_FS_ZIP */
351 
352 /* data.c */
353 struct page *erofs_get_meta_page(struct super_block *sb, erofs_blk_t blkaddr);
354 
355 int erofs_map_blocks(struct inode *, struct erofs_map_blocks *, int);
356 
357 /* inode.c */
358 static inline unsigned long erofs_inode_hash(erofs_nid_t nid)
359 {
360 #if BITS_PER_LONG == 32
361 	return (nid >> 32) ^ (nid & 0xffffffff);
362 #else
363 	return nid;
364 #endif
365 }
366 
367 extern const struct inode_operations erofs_generic_iops;
368 extern const struct inode_operations erofs_symlink_iops;
369 extern const struct inode_operations erofs_fast_symlink_iops;
370 
371 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid, bool dir);
372 int erofs_getattr(const struct path *path, struct kstat *stat,
373 		  u32 request_mask, unsigned int query_flags);
374 
375 /* namei.c */
376 extern const struct inode_operations erofs_dir_iops;
377 
378 int erofs_namei(struct inode *dir, struct qstr *name,
379 		erofs_nid_t *nid, unsigned int *d_type);
380 
381 /* dir.c */
382 extern const struct file_operations erofs_dir_fops;
383 
384 /* utils.c / zdata.c */
385 struct page *erofs_allocpage(struct list_head *pool, gfp_t gfp, bool nofail);
386 
387 #if (EROFS_PCPUBUF_NR_PAGES > 0)
388 void *erofs_get_pcpubuf(unsigned int pagenr);
389 #define erofs_put_pcpubuf(buf) do { \
390 	(void)&(buf);	\
391 	preempt_enable();	\
392 } while (0)
393 #else
394 static inline void *erofs_get_pcpubuf(unsigned int pagenr)
395 {
396 	return ERR_PTR(-EOPNOTSUPP);
397 }
398 
399 #define erofs_put_pcpubuf(buf) do {} while (0)
400 #endif
401 
402 #ifdef CONFIG_EROFS_FS_ZIP
403 int erofs_workgroup_put(struct erofs_workgroup *grp);
404 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
405 					     pgoff_t index, bool *tag);
406 int erofs_register_workgroup(struct super_block *sb,
407 			     struct erofs_workgroup *grp, bool tag);
408 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
409 void erofs_shrinker_register(struct super_block *sb);
410 void erofs_shrinker_unregister(struct super_block *sb);
411 int __init erofs_init_shrinker(void);
412 void erofs_exit_shrinker(void);
413 int __init z_erofs_init_zip_subsystem(void);
414 void z_erofs_exit_zip_subsystem(void);
415 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
416 				       struct erofs_workgroup *egrp);
417 int erofs_try_to_free_cached_page(struct address_space *mapping,
418 				  struct page *page);
419 #else
420 static inline void erofs_shrinker_register(struct super_block *sb) {}
421 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
422 static inline int erofs_init_shrinker(void) { return 0; }
423 static inline void erofs_exit_shrinker(void) {}
424 static inline int z_erofs_init_zip_subsystem(void) { return 0; }
425 static inline void z_erofs_exit_zip_subsystem(void) {}
426 #endif	/* !CONFIG_EROFS_FS_ZIP */
427 
428 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
429 
430 #endif	/* __EROFS_INTERNAL_H */
431 
432