2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
18 * This file is part of exofs.
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
38 static inline unsigned exofs_chunk_size(struct inode
*inode
)
40 return inode
->i_sb
->s_blocksize
;
43 static inline void exofs_put_page(struct page
*page
)
46 page_cache_release(page
);
49 /* Accesses dir's inode->i_size must be called under inode lock */
50 static inline unsigned long dir_pages(struct inode
*inode
)
52 return (inode
->i_size
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
55 static unsigned exofs_last_byte(struct inode
*inode
, unsigned long page_nr
)
57 loff_t last_byte
= inode
->i_size
;
59 last_byte
-= page_nr
<< PAGE_CACHE_SHIFT
;
60 if (last_byte
> PAGE_CACHE_SIZE
)
61 last_byte
= PAGE_CACHE_SIZE
;
65 static int exofs_commit_chunk(struct page
*page
, loff_t pos
, unsigned len
)
67 struct address_space
*mapping
= page
->mapping
;
68 struct inode
*dir
= mapping
->host
;
73 if (!PageUptodate(page
))
74 SetPageUptodate(page
);
76 if (pos
+len
> dir
->i_size
) {
77 i_size_write(dir
, pos
+len
);
78 mark_inode_dirty(dir
);
83 err
= write_one_page(page
, 1);
90 static void exofs_check_page(struct page
*page
)
92 struct inode
*dir
= page
->mapping
->host
;
93 unsigned chunk_size
= exofs_chunk_size(dir
);
94 char *kaddr
= page_address(page
);
95 unsigned offs
, rec_len
;
96 unsigned limit
= PAGE_CACHE_SIZE
;
97 struct exofs_dir_entry
*p
;
100 /* if the page is the last one in the directory */
101 if ((dir
->i_size
>> PAGE_CACHE_SHIFT
) == page
->index
) {
102 limit
= dir
->i_size
& ~PAGE_CACHE_MASK
;
103 if (limit
& (chunk_size
- 1))
108 for (offs
= 0; offs
<= limit
- EXOFS_DIR_REC_LEN(1); offs
+= rec_len
) {
109 p
= (struct exofs_dir_entry
*)(kaddr
+ offs
);
110 rec_len
= le16_to_cpu(p
->rec_len
);
112 if (rec_len
< EXOFS_DIR_REC_LEN(1))
116 if (rec_len
< EXOFS_DIR_REC_LEN(p
->name_len
))
118 if (((offs
+ rec_len
- 1) ^ offs
) & ~(chunk_size
-1))
124 SetPageChecked(page
);
128 EXOFS_ERR("ERROR [exofs_check_page]: "
129 "size of directory #%lu is not a multiple of chunk size",
134 error
= "rec_len is smaller than minimal";
137 error
= "unaligned directory entry";
140 error
= "rec_len is too small for name_len";
143 error
= "directory entry across blocks";
147 "ERROR [exofs_check_page]: bad entry in directory #%lu: %s - "
148 "offset=%lu, inode=%llu, rec_len=%d, name_len=%d",
149 dir
->i_ino
, error
, (page
->index
<<PAGE_CACHE_SHIFT
)+offs
,
150 _LLU(le64_to_cpu(p
->inode_no
)),
151 rec_len
, p
->name_len
);
154 p
= (struct exofs_dir_entry
*)(kaddr
+ offs
);
155 EXOFS_ERR("ERROR [exofs_check_page]: "
156 "entry in directory #%lu spans the page boundary"
157 "offset=%lu, inode=%llu",
158 dir
->i_ino
, (page
->index
<<PAGE_CACHE_SHIFT
)+offs
,
159 _LLU(le64_to_cpu(p
->inode_no
)));
161 SetPageChecked(page
);
165 static struct page
*exofs_get_page(struct inode
*dir
, unsigned long n
)
167 struct address_space
*mapping
= dir
->i_mapping
;
168 struct page
*page
= read_mapping_page(mapping
, n
, NULL
);
172 if (!PageChecked(page
))
173 exofs_check_page(page
);
180 exofs_put_page(page
);
181 return ERR_PTR(-EIO
);
184 static inline int exofs_match(int len
, const unsigned char *name
,
185 struct exofs_dir_entry
*de
)
187 if (len
!= de
->name_len
)
191 return !memcmp(name
, de
->name
, len
);
195 struct exofs_dir_entry
*exofs_next_entry(struct exofs_dir_entry
*p
)
197 return (struct exofs_dir_entry
*)((char *)p
+ le16_to_cpu(p
->rec_len
));
200 static inline unsigned
201 exofs_validate_entry(char *base
, unsigned offset
, unsigned mask
)
203 struct exofs_dir_entry
*de
= (struct exofs_dir_entry
*)(base
+ offset
);
204 struct exofs_dir_entry
*p
=
205 (struct exofs_dir_entry
*)(base
+ (offset
&mask
));
206 while ((char *)p
< (char *)de
) {
209 p
= exofs_next_entry(p
);
211 return (char *)p
- base
;
214 static unsigned char exofs_filetype_table
[EXOFS_FT_MAX
] = {
215 [EXOFS_FT_UNKNOWN
] = DT_UNKNOWN
,
216 [EXOFS_FT_REG_FILE
] = DT_REG
,
217 [EXOFS_FT_DIR
] = DT_DIR
,
218 [EXOFS_FT_CHRDEV
] = DT_CHR
,
219 [EXOFS_FT_BLKDEV
] = DT_BLK
,
220 [EXOFS_FT_FIFO
] = DT_FIFO
,
221 [EXOFS_FT_SOCK
] = DT_SOCK
,
222 [EXOFS_FT_SYMLINK
] = DT_LNK
,
226 static unsigned char exofs_type_by_mode
[S_IFMT
>> S_SHIFT
] = {
227 [S_IFREG
>> S_SHIFT
] = EXOFS_FT_REG_FILE
,
228 [S_IFDIR
>> S_SHIFT
] = EXOFS_FT_DIR
,
229 [S_IFCHR
>> S_SHIFT
] = EXOFS_FT_CHRDEV
,
230 [S_IFBLK
>> S_SHIFT
] = EXOFS_FT_BLKDEV
,
231 [S_IFIFO
>> S_SHIFT
] = EXOFS_FT_FIFO
,
232 [S_IFSOCK
>> S_SHIFT
] = EXOFS_FT_SOCK
,
233 [S_IFLNK
>> S_SHIFT
] = EXOFS_FT_SYMLINK
,
237 void exofs_set_de_type(struct exofs_dir_entry
*de
, struct inode
*inode
)
239 mode_t mode
= inode
->i_mode
;
240 de
->file_type
= exofs_type_by_mode
[(mode
& S_IFMT
) >> S_SHIFT
];
244 exofs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
246 loff_t pos
= filp
->f_pos
;
247 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
248 unsigned int offset
= pos
& ~PAGE_CACHE_MASK
;
249 unsigned long n
= pos
>> PAGE_CACHE_SHIFT
;
250 unsigned long npages
= dir_pages(inode
);
251 unsigned chunk_mask
= ~(exofs_chunk_size(inode
)-1);
252 unsigned char *types
= NULL
;
253 int need_revalidate
= (filp
->f_version
!= inode
->i_version
);
255 if (pos
> inode
->i_size
- EXOFS_DIR_REC_LEN(1))
258 types
= exofs_filetype_table
;
260 for ( ; n
< npages
; n
++, offset
= 0) {
262 struct exofs_dir_entry
*de
;
263 struct page
*page
= exofs_get_page(inode
, n
);
269 filp
->f_pos
+= PAGE_CACHE_SIZE
- offset
;
270 return PTR_ERR(page
);
272 kaddr
= page_address(page
);
273 if (unlikely(need_revalidate
)) {
275 offset
= exofs_validate_entry(kaddr
, offset
,
277 filp
->f_pos
= (n
<<PAGE_CACHE_SHIFT
) + offset
;
279 filp
->f_version
= inode
->i_version
;
282 de
= (struct exofs_dir_entry
*)(kaddr
+ offset
);
283 limit
= kaddr
+ exofs_last_byte(inode
, n
) -
284 EXOFS_DIR_REC_LEN(1);
285 for (; (char *)de
<= limit
; de
= exofs_next_entry(de
)) {
286 if (de
->rec_len
== 0) {
288 "zero-length directory entry");
289 exofs_put_page(page
);
294 unsigned char d_type
= DT_UNKNOWN
;
296 if (types
&& de
->file_type
< EXOFS_FT_MAX
)
297 d_type
= types
[de
->file_type
];
299 offset
= (char *)de
- kaddr
;
300 over
= filldir(dirent
, de
->name
, de
->name_len
,
301 (n
<<PAGE_CACHE_SHIFT
) | offset
,
302 le64_to_cpu(de
->inode_no
),
305 exofs_put_page(page
);
309 filp
->f_pos
+= le16_to_cpu(de
->rec_len
);
311 exofs_put_page(page
);
317 struct exofs_dir_entry
*exofs_find_entry(struct inode
*dir
,
318 struct dentry
*dentry
, struct page
**res_page
)
320 const unsigned char *name
= dentry
->d_name
.name
;
321 int namelen
= dentry
->d_name
.len
;
322 unsigned reclen
= EXOFS_DIR_REC_LEN(namelen
);
323 unsigned long start
, n
;
324 unsigned long npages
= dir_pages(dir
);
325 struct page
*page
= NULL
;
326 struct exofs_i_info
*oi
= exofs_i(dir
);
327 struct exofs_dir_entry
*de
;
334 start
= oi
->i_dir_start_lookup
;
340 page
= exofs_get_page(dir
, n
);
342 kaddr
= page_address(page
);
343 de
= (struct exofs_dir_entry
*) kaddr
;
344 kaddr
+= exofs_last_byte(dir
, n
) - reclen
;
345 while ((char *) de
<= kaddr
) {
346 if (de
->rec_len
== 0) {
348 "ERROR: exofs_find_entry: "
349 "zero-length directory entry");
350 exofs_put_page(page
);
353 if (exofs_match(namelen
, name
, de
))
355 de
= exofs_next_entry(de
);
357 exofs_put_page(page
);
361 } while (n
!= start
);
367 oi
->i_dir_start_lookup
= n
;
371 struct exofs_dir_entry
*exofs_dotdot(struct inode
*dir
, struct page
**p
)
373 struct page
*page
= exofs_get_page(dir
, 0);
374 struct exofs_dir_entry
*de
= NULL
;
377 de
= exofs_next_entry(
378 (struct exofs_dir_entry
*)page_address(page
));
384 ino_t
exofs_parent_ino(struct dentry
*child
)
387 struct exofs_dir_entry
*de
;
390 de
= exofs_dotdot(child
->d_inode
, &page
);
394 ino
= le64_to_cpu(de
->inode_no
);
395 exofs_put_page(page
);
399 ino_t
exofs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
)
402 struct exofs_dir_entry
*de
;
405 de
= exofs_find_entry(dir
, dentry
, &page
);
407 res
= le64_to_cpu(de
->inode_no
);
408 exofs_put_page(page
);
413 int exofs_set_link(struct inode
*dir
, struct exofs_dir_entry
*de
,
414 struct page
*page
, struct inode
*inode
)
416 loff_t pos
= page_offset(page
) +
417 (char *) de
- (char *) page_address(page
);
418 unsigned len
= le16_to_cpu(de
->rec_len
);
422 err
= exofs_write_begin(NULL
, page
->mapping
, pos
, len
,
423 AOP_FLAG_UNINTERRUPTIBLE
, &page
, NULL
);
425 EXOFS_ERR("exofs_set_link: exofs_write_begin FAILD => %d\n",
428 de
->inode_no
= cpu_to_le64(inode
->i_ino
);
429 exofs_set_de_type(de
, inode
);
431 err
= exofs_commit_chunk(page
, pos
, len
);
432 exofs_put_page(page
);
433 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
434 mark_inode_dirty(dir
);
438 int exofs_add_link(struct dentry
*dentry
, struct inode
*inode
)
440 struct inode
*dir
= dentry
->d_parent
->d_inode
;
441 const unsigned char *name
= dentry
->d_name
.name
;
442 int namelen
= dentry
->d_name
.len
;
443 unsigned chunk_size
= exofs_chunk_size(dir
);
444 unsigned reclen
= EXOFS_DIR_REC_LEN(namelen
);
445 unsigned short rec_len
, name_len
;
446 struct page
*page
= NULL
;
447 struct exofs_sb_info
*sbi
= inode
->i_sb
->s_fs_info
;
448 struct exofs_dir_entry
*de
;
449 unsigned long npages
= dir_pages(dir
);
455 for (n
= 0; n
<= npages
; n
++) {
458 page
= exofs_get_page(dir
, n
);
463 kaddr
= page_address(page
);
464 dir_end
= kaddr
+ exofs_last_byte(dir
, n
);
465 de
= (struct exofs_dir_entry
*)kaddr
;
466 kaddr
+= PAGE_CACHE_SIZE
- reclen
;
467 while ((char *)de
<= kaddr
) {
468 if ((char *)de
== dir_end
) {
470 rec_len
= chunk_size
;
471 de
->rec_len
= cpu_to_le16(chunk_size
);
475 if (de
->rec_len
== 0) {
476 EXOFS_ERR("ERROR: exofs_add_link: "
477 "zero-length directory entry");
482 if (exofs_match(namelen
, name
, de
))
484 name_len
= EXOFS_DIR_REC_LEN(de
->name_len
);
485 rec_len
= le16_to_cpu(de
->rec_len
);
486 if (!de
->inode_no
&& rec_len
>= reclen
)
488 if (rec_len
>= name_len
+ reclen
)
490 de
= (struct exofs_dir_entry
*) ((char *) de
+ rec_len
);
493 exofs_put_page(page
);
496 EXOFS_ERR("exofs_add_link: BAD dentry=%p or inode=%p", dentry
, inode
);
500 pos
= page_offset(page
) +
501 (char *)de
- (char *)page_address(page
);
502 err
= exofs_write_begin(NULL
, page
->mapping
, pos
, rec_len
, 0,
507 struct exofs_dir_entry
*de1
=
508 (struct exofs_dir_entry
*)((char *)de
+ name_len
);
509 de1
->rec_len
= cpu_to_le16(rec_len
- name_len
);
510 de
->rec_len
= cpu_to_le16(name_len
);
513 de
->name_len
= namelen
;
514 memcpy(de
->name
, name
, namelen
);
515 de
->inode_no
= cpu_to_le64(inode
->i_ino
);
516 exofs_set_de_type(de
, inode
);
517 err
= exofs_commit_chunk(page
, pos
, rec_len
);
518 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
519 mark_inode_dirty(dir
);
523 exofs_put_page(page
);
531 int exofs_delete_entry(struct exofs_dir_entry
*dir
, struct page
*page
)
533 struct address_space
*mapping
= page
->mapping
;
534 struct inode
*inode
= mapping
->host
;
535 struct exofs_sb_info
*sbi
= inode
->i_sb
->s_fs_info
;
536 char *kaddr
= page_address(page
);
537 unsigned from
= ((char *)dir
- kaddr
) & ~(exofs_chunk_size(inode
)-1);
538 unsigned to
= ((char *)dir
- kaddr
) + le16_to_cpu(dir
->rec_len
);
540 struct exofs_dir_entry
*pde
= NULL
;
541 struct exofs_dir_entry
*de
= (struct exofs_dir_entry
*) (kaddr
+ from
);
545 if (de
->rec_len
== 0) {
546 EXOFS_ERR("ERROR: exofs_delete_entry:"
547 "zero-length directory entry");
552 de
= exofs_next_entry(de
);
555 from
= (char *)pde
- (char *)page_address(page
);
556 pos
= page_offset(page
) + from
;
558 err
= exofs_write_begin(NULL
, page
->mapping
, pos
, to
- from
, 0,
561 EXOFS_ERR("exofs_delete_entry: exofs_write_begin FAILD => %d\n",
564 pde
->rec_len
= cpu_to_le16(to
- from
);
567 err
= exofs_commit_chunk(page
, pos
, to
- from
);
568 inode
->i_ctime
= inode
->i_mtime
= CURRENT_TIME
;
569 mark_inode_dirty(inode
);
572 exofs_put_page(page
);
576 /* kept aligned on 4 bytes */
577 #define THIS_DIR ".\0\0"
578 #define PARENT_DIR "..\0"
580 int exofs_make_empty(struct inode
*inode
, struct inode
*parent
)
582 struct address_space
*mapping
= inode
->i_mapping
;
583 struct page
*page
= grab_cache_page(mapping
, 0);
584 unsigned chunk_size
= exofs_chunk_size(inode
);
585 struct exofs_dir_entry
*de
;
592 err
= exofs_write_begin(NULL
, page
->mapping
, 0, chunk_size
, 0,
599 kaddr
= kmap_atomic(page
, KM_USER0
);
600 de
= (struct exofs_dir_entry
*)kaddr
;
602 de
->rec_len
= cpu_to_le16(EXOFS_DIR_REC_LEN(1));
603 memcpy(de
->name
, THIS_DIR
, sizeof(THIS_DIR
));
604 de
->inode_no
= cpu_to_le64(inode
->i_ino
);
605 exofs_set_de_type(de
, inode
);
607 de
= (struct exofs_dir_entry
*)(kaddr
+ EXOFS_DIR_REC_LEN(1));
609 de
->rec_len
= cpu_to_le16(chunk_size
- EXOFS_DIR_REC_LEN(1));
610 de
->inode_no
= cpu_to_le64(parent
->i_ino
);
611 memcpy(de
->name
, PARENT_DIR
, sizeof(PARENT_DIR
));
612 exofs_set_de_type(de
, inode
);
613 kunmap_atomic(page
, KM_USER0
);
614 err
= exofs_commit_chunk(page
, 0, chunk_size
);
616 page_cache_release(page
);
620 int exofs_empty_dir(struct inode
*inode
)
622 struct page
*page
= NULL
;
623 unsigned long i
, npages
= dir_pages(inode
);
625 for (i
= 0; i
< npages
; i
++) {
627 struct exofs_dir_entry
*de
;
628 page
= exofs_get_page(inode
, i
);
633 kaddr
= page_address(page
);
634 de
= (struct exofs_dir_entry
*)kaddr
;
635 kaddr
+= exofs_last_byte(inode
, i
) - EXOFS_DIR_REC_LEN(1);
637 while ((char *)de
<= kaddr
) {
638 if (de
->rec_len
== 0) {
639 EXOFS_ERR("ERROR: exofs_empty_dir: "
640 "zero-length directory entry"
641 "kaddr=%p, de=%p\n", kaddr
, de
);
644 if (de
->inode_no
!= 0) {
645 /* check for . and .. */
646 if (de
->name
[0] != '.')
648 if (de
->name_len
> 2)
650 if (de
->name_len
< 2) {
651 if (le64_to_cpu(de
->inode_no
) !=
654 } else if (de
->name
[1] != '.')
657 de
= exofs_next_entry(de
);
659 exofs_put_page(page
);
664 exofs_put_page(page
);
668 const struct file_operations exofs_dir_operations
= {
669 .llseek
= generic_file_llseek
,
670 .read
= generic_read_dir
,
671 .readdir
= exofs_readdir
,