1 /* SPDX-License-Identifier: GPL-2.0
2 *
3 * linux/drivers/staging/erofs/internal.h
4 *
5 * Copyright (C) 2017-2018 HUAWEI, Inc.
6 * http://www.huawei.com/
7 * Created by Gao Xiang <gaoxiang25@huawei.com>
8 *
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.
12 */
13 #ifndef __INTERNAL_H
14 #define __INTERNAL_H
15
16 #include <linux/fs.h>
17 #include <linux/dcache.h>
18 #include <linux/mm.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>
25 #include "erofs_fs.h"
26
27 /* redefine pr_fmt "erofs: " */
28 #undef pr_fmt
29 #define pr_fmt(fmt) "erofs: " fmt
30
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__)
35
36 #define dbg_might_sleep might_sleep
37 #define DBG_BUGON BUG_ON
38 #else
39 #define debugln(x, ...) ((void)0)
40
41 #define dbg_might_sleep() ((void)0)
42 #define DBG_BUGON(...) ((void)0)
43 #endif
44
45 #ifdef CONFIG_EROFS_FAULT_INJECTION
46 enum {
47 FAULT_KMALLOC,
48 FAULT_MAX,
49 };
50
51 extern char *erofs_fault_name[FAULT_MAX];
52 #define IS_FAULT_SET(fi, type) ((fi)->inject_type & (1 << (type)))
53
54 struct erofs_fault_info {
55 atomic_t inject_ops;
56 unsigned int inject_rate;
57 unsigned int inject_type;
58 };
59 #endif
60
61 #ifdef CONFIG_EROFS_FS_ZIP_CACHE_BIPOLAR
62 #define EROFS_FS_ZIP_CACHE_LVL (2)
63 #elif defined(EROFS_FS_ZIP_CACHE_UNIPOLAR)
64 #define EROFS_FS_ZIP_CACHE_LVL (1)
65 #else
66 #define EROFS_FS_ZIP_CACHE_LVL (0)
67 #endif
68
69 #if (!defined(EROFS_FS_HAS_MANAGED_CACHE) && (EROFS_FS_ZIP_CACHE_LVL > 0))
70 #define EROFS_FS_HAS_MANAGED_CACHE
71 #endif
72
73 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */
74 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1
75
76 typedef u64 erofs_nid_t;
77
78 struct erofs_sb_info {
79 /* list for all registered superblocks, mainly for shrinker */
80 struct list_head list;
81 struct mutex umount_mutex;
82
83 u32 blocks;
84 u32 meta_blkaddr;
85 #ifdef CONFIG_EROFS_FS_XATTR
86 u32 xattr_blkaddr;
87 #endif
88
89 /* inode slot unit size in bit shift */
90 unsigned char islotbits;
91 #ifdef CONFIG_EROFS_FS_ZIP
92 /* cluster size in bit shift */
93 unsigned char clusterbits;
94
95 /* the dedicated workstation for compression */
96 struct radix_tree_root workstn_tree;
97
98 #ifdef EROFS_FS_HAS_MANAGED_CACHE
99 struct inode *managed_cache;
100 #endif
101
102 #endif
103
104 u32 build_time_nsec;
105 u64 build_time;
106
107 /* what we really care is nid, rather than ino.. */
108 erofs_nid_t root_nid;
109 /* used for statfs, f_files - f_favail */
110 u64 inos;
111
112 u8 uuid[16]; /* 128-bit uuid for volume */
113 u8 volume_name[16]; /* volume name */
114 char *dev_name;
115
116 unsigned int mount_opt;
117 unsigned int shrinker_run_no;
118
119 #ifdef CONFIG_EROFS_FAULT_INJECTION
120 struct erofs_fault_info fault_info; /* For fault injection */
121 #endif
122 };
123
124 #ifdef CONFIG_EROFS_FAULT_INJECTION
125 #define erofs_show_injection_info(type) \
126 infoln("inject %s in %s of %pS", erofs_fault_name[type], \
127 __func__, __builtin_return_address(0))
128
time_to_inject(struct erofs_sb_info * sbi,int type)129 static inline bool time_to_inject(struct erofs_sb_info *sbi, int type)
130 {
131 struct erofs_fault_info *ffi = &sbi->fault_info;
132
133 if (!ffi->inject_rate)
134 return false;
135
136 if (!IS_FAULT_SET(ffi, type))
137 return false;
138
139 atomic_inc(&ffi->inject_ops);
140 if (atomic_read(&ffi->inject_ops) >= ffi->inject_rate) {
141 atomic_set(&ffi->inject_ops, 0);
142 return true;
143 }
144 return false;
145 }
146 #endif
147
erofs_kmalloc(struct erofs_sb_info * sbi,size_t size,gfp_t flags)148 static inline void *erofs_kmalloc(struct erofs_sb_info *sbi,
149 size_t size, gfp_t flags)
150 {
151 #ifdef CONFIG_EROFS_FAULT_INJECTION
152 if (time_to_inject(sbi, FAULT_KMALLOC)) {
153 erofs_show_injection_info(FAULT_KMALLOC);
154 return NULL;
155 }
156 #endif
157 return kmalloc(size, flags);
158 }
159
160 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
161 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
162
163 /* Mount flags set via mount options or defaults */
164 #define EROFS_MOUNT_XATTR_USER 0x00000010
165 #define EROFS_MOUNT_POSIX_ACL 0x00000020
166 #define EROFS_MOUNT_FAULT_INJECTION 0x00000040
167
168 #define clear_opt(sbi, option) ((sbi)->mount_opt &= ~EROFS_MOUNT_##option)
169 #define set_opt(sbi, option) ((sbi)->mount_opt |= EROFS_MOUNT_##option)
170 #define test_opt(sbi, option) ((sbi)->mount_opt & EROFS_MOUNT_##option)
171
172 #ifdef CONFIG_EROFS_FS_ZIP
173 #define erofs_workstn_lock(sbi) xa_lock(&(sbi)->workstn_tree)
174 #define erofs_workstn_unlock(sbi) xa_unlock(&(sbi)->workstn_tree)
175
176 /* basic unit of the workstation of a super_block */
177 struct erofs_workgroup {
178 /* the workgroup index in the workstation */
179 pgoff_t index;
180
181 /* overall workgroup reference count */
182 atomic_t refcount;
183 };
184
185 #define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL)
186
erofs_workgroup_try_to_freeze(struct erofs_workgroup * grp,int v)187 static inline bool erofs_workgroup_try_to_freeze(
188 struct erofs_workgroup *grp, int v)
189 {
190 #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK)
191 if (v != atomic_cmpxchg(&grp->refcount,
192 v, EROFS_LOCKED_MAGIC))
193 return false;
194 preempt_disable();
195 #else
196 preempt_disable();
197 if (atomic_read(&grp->refcount) != v) {
198 preempt_enable();
199 return false;
200 }
201 #endif
202 return true;
203 }
204
erofs_workgroup_unfreeze(struct erofs_workgroup * grp,int v)205 static inline void erofs_workgroup_unfreeze(
206 struct erofs_workgroup *grp, int v)
207 {
208 #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK)
209 atomic_set(&grp->refcount, v);
210 #endif
211 preempt_enable();
212 }
213
erofs_workgroup_get(struct erofs_workgroup * grp,int * ocnt)214 static inline bool erofs_workgroup_get(struct erofs_workgroup *grp, int *ocnt)
215 {
216 const int locked = (int)EROFS_LOCKED_MAGIC;
217 int o;
218
219 repeat:
220 o = atomic_read(&grp->refcount);
221
222 /* spin if it is temporarily locked at the reclaim path */
223 if (unlikely(o == locked)) {
224 #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK)
225 do
226 cpu_relax();
227 while (atomic_read(&grp->refcount) == locked);
228 #endif
229 goto repeat;
230 }
231
232 if (unlikely(o <= 0))
233 return -1;
234
235 if (unlikely(atomic_cmpxchg(&grp->refcount, o, o + 1) != o))
236 goto repeat;
237
238 *ocnt = o;
239 return 0;
240 }
241
242 #define __erofs_workgroup_get(grp) atomic_inc(&(grp)->refcount)
243
244 extern int erofs_workgroup_put(struct erofs_workgroup *grp);
245
246 extern struct erofs_workgroup *erofs_find_workgroup(
247 struct super_block *sb, pgoff_t index, bool *tag);
248
249 extern int erofs_register_workgroup(struct super_block *sb,
250 struct erofs_workgroup *grp, bool tag);
251
252 extern unsigned long erofs_shrink_workstation(struct erofs_sb_info *sbi,
253 unsigned long nr_shrink, bool cleanup);
254
erofs_workstation_cleanup_all(struct super_block * sb)255 static inline void erofs_workstation_cleanup_all(struct super_block *sb)
256 {
257 erofs_shrink_workstation(EROFS_SB(sb), ~0UL, true);
258 }
259
260 #ifdef EROFS_FS_HAS_MANAGED_CACHE
261 #define EROFS_UNALLOCATED_CACHED_PAGE ((void *)0x5F0EF00D)
262
263 extern int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
264 struct erofs_workgroup *egrp);
265 extern int erofs_try_to_free_cached_page(struct address_space *mapping,
266 struct page *page);
267 #endif
268
269 #endif
270
271 /* we strictly follow PAGE_SIZE and no buffer head yet */
272 #define LOG_BLOCK_SIZE PAGE_SHIFT
273
274 #undef LOG_SECTORS_PER_BLOCK
275 #define LOG_SECTORS_PER_BLOCK (PAGE_SHIFT - 9)
276
277 #undef SECTORS_PER_BLOCK
278 #define SECTORS_PER_BLOCK (1 << SECTORS_PER_BLOCK)
279
280 #define EROFS_BLKSIZ (1 << LOG_BLOCK_SIZE)
281
282 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ)
283 #error erofs cannot be used in this platform
284 #endif
285
286 #define ROOT_NID(sb) ((sb)->root_nid)
287
288 #ifdef CONFIG_EROFS_FS_ZIP
289 /* hard limit of pages per compressed cluster */
290 #define Z_EROFS_CLUSTER_MAX_PAGES (CONFIG_EROFS_FS_CLUSTER_PAGE_LIMIT)
291
292 /* page count of a compressed cluster */
293 #define erofs_clusterpages(sbi) ((1 << (sbi)->clusterbits) / PAGE_SIZE)
294 #endif
295
296 typedef u64 erofs_off_t;
297
298 /* data type for filesystem-wide blocks number */
299 typedef u32 erofs_blk_t;
300
301 #define erofs_blknr(addr) ((addr) / EROFS_BLKSIZ)
302 #define erofs_blkoff(addr) ((addr) % EROFS_BLKSIZ)
303 #define blknr_to_addr(nr) ((erofs_off_t)(nr) * EROFS_BLKSIZ)
304
iloc(struct erofs_sb_info * sbi,erofs_nid_t nid)305 static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid)
306 {
307 return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits);
308 }
309
310 #define inode_set_inited_xattr(inode) (EROFS_V(inode)->flags |= 1)
311 #define inode_has_inited_xattr(inode) (EROFS_V(inode)->flags & 1)
312
313 struct erofs_vnode {
314 erofs_nid_t nid;
315 unsigned int flags;
316
317 unsigned char data_mapping_mode;
318 /* inline size in bytes */
319 unsigned char inode_isize;
320 unsigned short xattr_isize;
321
322 unsigned xattr_shared_count;
323 unsigned *xattr_shared_xattrs;
324
325 erofs_blk_t raw_blkaddr;
326
327 /* the corresponding vfs inode */
328 struct inode vfs_inode;
329 };
330
331 #define EROFS_V(ptr) \
332 container_of(ptr, struct erofs_vnode, vfs_inode)
333
334 #define __inode_advise(x, bit, bits) \
335 (((x) >> (bit)) & ((1 << (bits)) - 1))
336
337 #define __inode_version(advise) \
338 __inode_advise(advise, EROFS_I_VERSION_BIT, \
339 EROFS_I_VERSION_BITS)
340
341 #define __inode_data_mapping(advise) \
342 __inode_advise(advise, EROFS_I_DATA_MAPPING_BIT,\
343 EROFS_I_DATA_MAPPING_BITS)
344
inode_datablocks(struct inode * inode)345 static inline unsigned long inode_datablocks(struct inode *inode)
346 {
347 /* since i_size cannot be changed */
348 return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
349 }
350
is_inode_layout_plain(struct inode * inode)351 static inline bool is_inode_layout_plain(struct inode *inode)
352 {
353 return EROFS_V(inode)->data_mapping_mode == EROFS_INODE_LAYOUT_PLAIN;
354 }
355
is_inode_layout_compression(struct inode * inode)356 static inline bool is_inode_layout_compression(struct inode *inode)
357 {
358 return EROFS_V(inode)->data_mapping_mode ==
359 EROFS_INODE_LAYOUT_COMPRESSION;
360 }
361
is_inode_layout_inline(struct inode * inode)362 static inline bool is_inode_layout_inline(struct inode *inode)
363 {
364 return EROFS_V(inode)->data_mapping_mode == EROFS_INODE_LAYOUT_INLINE;
365 }
366
367 extern const struct super_operations erofs_sops;
368 extern const struct inode_operations erofs_dir_iops;
369 extern const struct file_operations erofs_dir_fops;
370
371 extern const struct address_space_operations erofs_raw_access_aops;
372 #ifdef CONFIG_EROFS_FS_ZIP
373 extern const struct address_space_operations z_erofs_vle_normalaccess_aops;
374 #endif
375
376 /*
377 * Logical to physical block mapping, used by erofs_map_blocks()
378 *
379 * Different with other file systems, it is used for 2 access modes:
380 *
381 * 1) RAW access mode:
382 *
383 * Users pass a valid (m_lblk, m_lofs -- usually 0) pair,
384 * and get the valid m_pblk, m_pofs and the longest m_len(in bytes).
385 *
386 * Note that m_lblk in the RAW access mode refers to the number of
387 * the compressed ondisk block rather than the uncompressed
388 * in-memory block for the compressed file.
389 *
390 * m_pofs equals to m_lofs except for the inline data page.
391 *
392 * 2) Normal access mode:
393 *
394 * If the inode is not compressed, it has no difference with
395 * the RAW access mode. However, if the inode is compressed,
396 * users should pass a valid (m_lblk, m_lofs) pair, and get
397 * the needed m_pblk, m_pofs, m_len to get the compressed data
398 * and the updated m_lblk, m_lofs which indicates the start
399 * of the corresponding uncompressed data in the file.
400 */
401 enum {
402 BH_Zipped = BH_PrivateStart,
403 };
404
405 /* Has a disk mapping */
406 #define EROFS_MAP_MAPPED (1 << BH_Mapped)
407 /* Located in metadata (could be copied from bd_inode) */
408 #define EROFS_MAP_META (1 << BH_Meta)
409 /* The extent has been compressed */
410 #define EROFS_MAP_ZIPPED (1 << BH_Zipped)
411
412 struct erofs_map_blocks {
413 erofs_off_t m_pa, m_la;
414 u64 m_plen, m_llen;
415
416 unsigned int m_flags;
417 };
418
419 /* Flags used by erofs_map_blocks() */
420 #define EROFS_GET_BLOCKS_RAW 0x0001
421
422 /* data.c */
prepare_bio(struct super_block * sb,erofs_blk_t blkaddr,unsigned nr_pages,bio_end_io_t endio)423 static inline struct bio *prepare_bio(
424 struct super_block *sb,
425 erofs_blk_t blkaddr, unsigned nr_pages,
426 bio_end_io_t endio)
427 {
428 gfp_t gfp = GFP_NOIO;
429 struct bio *bio = bio_alloc(gfp, nr_pages);
430
431 if (unlikely(bio == NULL) &&
432 (current->flags & PF_MEMALLOC)) {
433 do {
434 nr_pages /= 2;
435 if (unlikely(!nr_pages)) {
436 bio = bio_alloc(gfp | __GFP_NOFAIL, 1);
437 BUG_ON(bio == NULL);
438 break;
439 }
440 bio = bio_alloc(gfp, nr_pages);
441 } while (bio == NULL);
442 }
443
444 bio->bi_end_io = endio;
445 bio_set_dev(bio, sb->s_bdev);
446 bio->bi_iter.bi_sector = blkaddr << LOG_SECTORS_PER_BLOCK;
447 return bio;
448 }
449
__submit_bio(struct bio * bio,unsigned op,unsigned op_flags)450 static inline void __submit_bio(struct bio *bio, unsigned op, unsigned op_flags)
451 {
452 bio_set_op_attrs(bio, op, op_flags);
453 submit_bio(bio);
454 }
455
456 extern struct page *erofs_get_meta_page(struct super_block *sb,
457 erofs_blk_t blkaddr, bool prio);
458 extern int erofs_map_blocks(struct inode *, struct erofs_map_blocks *, int);
459 extern int erofs_map_blocks_iter(struct inode *, struct erofs_map_blocks *,
460 struct page **, int);
461
462 struct erofs_map_blocks_iter {
463 struct erofs_map_blocks map;
464 struct page *mpage;
465 };
466
467
erofs_get_inline_page(struct inode * inode,erofs_blk_t blkaddr)468 static inline struct page *erofs_get_inline_page(struct inode *inode,
469 erofs_blk_t blkaddr)
470 {
471 return erofs_get_meta_page(inode->i_sb,
472 blkaddr, S_ISDIR(inode->i_mode));
473 }
474
475 /* inode.c */
476 extern struct inode *erofs_iget(struct super_block *sb,
477 erofs_nid_t nid, bool dir);
478
479 /* dir.c */
480 int erofs_namei(struct inode *dir, struct qstr *name,
481 erofs_nid_t *nid, unsigned *d_type);
482
483 /* xattr.c */
484 #ifdef CONFIG_EROFS_FS_XATTR
485 extern const struct xattr_handler *erofs_xattr_handlers[];
486 #endif
487
488 /* symlink */
489 #ifdef CONFIG_EROFS_FS_XATTR
490 extern const struct inode_operations erofs_symlink_xattr_iops;
491 extern const struct inode_operations erofs_fast_symlink_xattr_iops;
492 extern const struct inode_operations erofs_special_inode_operations;
493 #endif
494
set_inode_fast_symlink(struct inode * inode)495 static inline void set_inode_fast_symlink(struct inode *inode)
496 {
497 #ifdef CONFIG_EROFS_FS_XATTR
498 inode->i_op = &erofs_fast_symlink_xattr_iops;
499 #else
500 inode->i_op = &simple_symlink_inode_operations;
501 #endif
502 }
503
is_inode_fast_symlink(struct inode * inode)504 static inline bool is_inode_fast_symlink(struct inode *inode)
505 {
506 #ifdef CONFIG_EROFS_FS_XATTR
507 return inode->i_op == &erofs_fast_symlink_xattr_iops;
508 #else
509 return inode->i_op == &simple_symlink_inode_operations;
510 #endif
511 }
512
erofs_vmap(struct page ** pages,unsigned int count)513 static inline void *erofs_vmap(struct page **pages, unsigned int count)
514 {
515 #ifdef CONFIG_EROFS_FS_USE_VM_MAP_RAM
516 int i = 0;
517
518 while (1) {
519 void *addr = vm_map_ram(pages, count, -1, PAGE_KERNEL);
520 /* retry two more times (totally 3 times) */
521 if (addr != NULL || ++i >= 3)
522 return addr;
523 vm_unmap_aliases();
524 }
525 return NULL;
526 #else
527 return vmap(pages, count, VM_MAP, PAGE_KERNEL);
528 #endif
529 }
530
erofs_vunmap(const void * mem,unsigned int count)531 static inline void erofs_vunmap(const void *mem, unsigned int count)
532 {
533 #ifdef CONFIG_EROFS_FS_USE_VM_MAP_RAM
534 vm_unmap_ram(mem, count);
535 #else
536 vunmap(mem);
537 #endif
538 }
539
540 /* utils.c */
541 extern struct page *erofs_allocpage(struct list_head *pool, gfp_t gfp);
542
543 extern void erofs_register_super(struct super_block *sb);
544 extern void erofs_unregister_super(struct super_block *sb);
545
546 extern unsigned long erofs_shrink_count(struct shrinker *shrink,
547 struct shrink_control *sc);
548 extern unsigned long erofs_shrink_scan(struct shrinker *shrink,
549 struct shrink_control *sc);
550
551 #ifndef lru_to_page
552 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
553 #endif
554
555 #endif
556
557
