blob: 65b0c8c776a1db8144c74ce0d7074512da882b51 [file] [log] [blame]
Boaz Harroshe6af00f2008-10-28 15:38:12 +02001/*
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>
8 *
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)
14 * from
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
17 *
18 * This file is part of exofs.
19 *
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.
25 *
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.
30 *
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
34 */
35
36#include "exofs.h"
37
38static inline unsigned exofs_chunk_size(struct inode *inode)
39{
40 return inode->i_sb->s_blocksize;
41}
42
43static inline void exofs_put_page(struct page *page)
44{
45 kunmap(page);
46 page_cache_release(page);
47}
48
49/* Accesses dir's inode->i_size must be called under inode lock */
50static inline unsigned long dir_pages(struct inode *inode)
51{
52 return (inode->i_size + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
53}
54
55static unsigned exofs_last_byte(struct inode *inode, unsigned long page_nr)
56{
57 loff_t last_byte = inode->i_size;
58
59 last_byte -= page_nr << PAGE_CACHE_SHIFT;
60 if (last_byte > PAGE_CACHE_SIZE)
61 last_byte = PAGE_CACHE_SIZE;
62 return last_byte;
63}
64
65static int exofs_commit_chunk(struct page *page, loff_t pos, unsigned len)
66{
67 struct address_space *mapping = page->mapping;
68 struct inode *dir = mapping->host;
69 int err = 0;
70
71 dir->i_version++;
72
73 if (!PageUptodate(page))
74 SetPageUptodate(page);
75
76 if (pos+len > dir->i_size) {
77 i_size_write(dir, pos+len);
78 mark_inode_dirty(dir);
79 }
80 set_page_dirty(page);
81
82 if (IS_DIRSYNC(dir))
83 err = write_one_page(page, 1);
84 else
85 unlock_page(page);
86
87 return err;
88}
89
90static void exofs_check_page(struct page *page)
91{
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;
98 char *error;
99
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))
104 goto Ebadsize;
105 if (!limit)
106 goto out;
107 }
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);
111
112 if (rec_len < EXOFS_DIR_REC_LEN(1))
113 goto Eshort;
114 if (rec_len & 3)
115 goto Ealign;
116 if (rec_len < EXOFS_DIR_REC_LEN(p->name_len))
117 goto Enamelen;
118 if (((offs + rec_len - 1) ^ offs) & ~(chunk_size-1))
119 goto Espan;
120 }
121 if (offs != limit)
122 goto Eend;
123out:
124 SetPageChecked(page);
125 return;
126
127Ebadsize:
128 EXOFS_ERR("ERROR [exofs_check_page]: "
129 "size of directory #%lu is not a multiple of chunk size",
130 dir->i_ino
131 );
132 goto fail;
133Eshort:
134 error = "rec_len is smaller than minimal";
135 goto bad_entry;
136Ealign:
137 error = "unaligned directory entry";
138 goto bad_entry;
139Enamelen:
140 error = "rec_len is too small for name_len";
141 goto bad_entry;
142Espan:
143 error = "directory entry across blocks";
144 goto bad_entry;
145bad_entry:
146 EXOFS_ERR(
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);
152 goto fail;
153Eend:
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)));
160fail:
161 SetPageChecked(page);
162 SetPageError(page);
163}
164
165static struct page *exofs_get_page(struct inode *dir, unsigned long n)
166{
167 struct address_space *mapping = dir->i_mapping;
168 struct page *page = read_mapping_page(mapping, n, NULL);
169
170 if (!IS_ERR(page)) {
171 kmap(page);
172 if (!PageChecked(page))
173 exofs_check_page(page);
174 if (PageError(page))
175 goto fail;
176 }
177 return page;
178
179fail:
180 exofs_put_page(page);
181 return ERR_PTR(-EIO);
182}
183
184static inline int exofs_match(int len, const unsigned char *name,
185 struct exofs_dir_entry *de)
186{
187 if (len != de->name_len)
188 return 0;
189 if (!de->inode_no)
190 return 0;
191 return !memcmp(name, de->name, len);
192}
193
194static inline
195struct exofs_dir_entry *exofs_next_entry(struct exofs_dir_entry *p)
196{
197 return (struct exofs_dir_entry *)((char *)p + le16_to_cpu(p->rec_len));
198}
199
200static inline unsigned
201exofs_validate_entry(char *base, unsigned offset, unsigned mask)
202{
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) {
207 if (p->rec_len == 0)
208 break;
209 p = exofs_next_entry(p);
210 }
211 return (char *)p - base;
212}
213
214static 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,
223};
224
225#define S_SHIFT 12
226static 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,
234};
235
236static inline
237void exofs_set_de_type(struct exofs_dir_entry *de, struct inode *inode)
238{
239 mode_t mode = inode->i_mode;
240 de->file_type = exofs_type_by_mode[(mode & S_IFMT) >> S_SHIFT];
241}
242
243static int
244exofs_readdir(struct file *filp, void *dirent, filldir_t filldir)
245{
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);
254
255 if (pos > inode->i_size - EXOFS_DIR_REC_LEN(1))
256 return 0;
257
258 types = exofs_filetype_table;
259
260 for ( ; n < npages; n++, offset = 0) {
261 char *kaddr, *limit;
262 struct exofs_dir_entry *de;
263 struct page *page = exofs_get_page(inode, n);
264
265 if (IS_ERR(page)) {
266 EXOFS_ERR("ERROR: "
267 "bad page in #%lu",
268 inode->i_ino);
269 filp->f_pos += PAGE_CACHE_SIZE - offset;
270 return PTR_ERR(page);
271 }
272 kaddr = page_address(page);
273 if (unlikely(need_revalidate)) {
274 if (offset) {
275 offset = exofs_validate_entry(kaddr, offset,
276 chunk_mask);
277 filp->f_pos = (n<<PAGE_CACHE_SHIFT) + offset;
278 }
279 filp->f_version = inode->i_version;
280 need_revalidate = 0;
281 }
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) {
287 EXOFS_ERR("ERROR: "
288 "zero-length directory entry");
289 exofs_put_page(page);
290 return -EIO;
291 }
292 if (de->inode_no) {
293 int over;
294 unsigned char d_type = DT_UNKNOWN;
295
296 if (types && de->file_type < EXOFS_FT_MAX)
297 d_type = types[de->file_type];
298
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),
303 d_type);
304 if (over) {
305 exofs_put_page(page);
306 return 0;
307 }
308 }
309 filp->f_pos += le16_to_cpu(de->rec_len);
310 }
311 exofs_put_page(page);
312 }
313
314 return 0;
315}
316
317struct exofs_dir_entry *exofs_find_entry(struct inode *dir,
318 struct dentry *dentry, struct page **res_page)
319{
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;
328
329 if (npages == 0)
330 goto out;
331
332 *res_page = NULL;
333
334 start = oi->i_dir_start_lookup;
335 if (start >= npages)
336 start = 0;
337 n = start;
338 do {
339 char *kaddr;
340 page = exofs_get_page(dir, n);
341 if (!IS_ERR(page)) {
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) {
347 EXOFS_ERR(
348 "ERROR: exofs_find_entry: "
349 "zero-length directory entry");
350 exofs_put_page(page);
351 goto out;
352 }
353 if (exofs_match(namelen, name, de))
354 goto found;
355 de = exofs_next_entry(de);
356 }
357 exofs_put_page(page);
358 }
359 if (++n >= npages)
360 n = 0;
361 } while (n != start);
362out:
363 return NULL;
364
365found:
366 *res_page = page;
367 oi->i_dir_start_lookup = n;
368 return de;
369}
370
371struct exofs_dir_entry *exofs_dotdot(struct inode *dir, struct page **p)
372{
373 struct page *page = exofs_get_page(dir, 0);
374 struct exofs_dir_entry *de = NULL;
375
376 if (!IS_ERR(page)) {
377 de = exofs_next_entry(
378 (struct exofs_dir_entry *)page_address(page));
379 *p = page;
380 }
381 return de;
382}
383
Boaz Harrosh8cf74b32009-03-22 12:47:26 +0200384ino_t exofs_parent_ino(struct dentry *child)
385{
386 struct page *page;
387 struct exofs_dir_entry *de;
388 ino_t ino;
389
390 de = exofs_dotdot(child->d_inode, &page);
391 if (!de)
392 return 0;
393
394 ino = le64_to_cpu(de->inode_no);
395 exofs_put_page(page);
396 return ino;
397}
398
Boaz Harroshe6af00f2008-10-28 15:38:12 +0200399ino_t exofs_inode_by_name(struct inode *dir, struct dentry *dentry)
400{
401 ino_t res = 0;
402 struct exofs_dir_entry *de;
403 struct page *page;
404
405 de = exofs_find_entry(dir, dentry, &page);
406 if (de) {
407 res = le64_to_cpu(de->inode_no);
408 exofs_put_page(page);
409 }
410 return res;
411}
412
413int exofs_set_link(struct inode *dir, struct exofs_dir_entry *de,
414 struct page *page, struct inode *inode)
415{
416 loff_t pos = page_offset(page) +
417 (char *) de - (char *) page_address(page);
418 unsigned len = le16_to_cpu(de->rec_len);
419 int err;
420
421 lock_page(page);
422 err = exofs_write_begin(NULL, page->mapping, pos, len,
423 AOP_FLAG_UNINTERRUPTIBLE, &page, NULL);
424 if (err)
425 EXOFS_ERR("exofs_set_link: exofs_write_begin FAILD => %d\n",
426 err);
427
428 de->inode_no = cpu_to_le64(inode->i_ino);
429 exofs_set_de_type(de, inode);
430 if (likely(!err))
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);
435 return err;
436}
437
438int exofs_add_link(struct dentry *dentry, struct inode *inode)
439{
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);
450 unsigned long n;
451 char *kaddr;
452 loff_t pos;
453 int err;
454
455 for (n = 0; n <= npages; n++) {
456 char *dir_end;
457
458 page = exofs_get_page(dir, n);
459 err = PTR_ERR(page);
460 if (IS_ERR(page))
461 goto out;
462 lock_page(page);
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) {
469 name_len = 0;
470 rec_len = chunk_size;
471 de->rec_len = cpu_to_le16(chunk_size);
472 de->inode_no = 0;
473 goto got_it;
474 }
475 if (de->rec_len == 0) {
476 EXOFS_ERR("ERROR: exofs_add_link: "
477 "zero-length directory entry");
478 err = -EIO;
479 goto out_unlock;
480 }
481 err = -EEXIST;
482 if (exofs_match(namelen, name, de))
483 goto out_unlock;
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)
487 goto got_it;
488 if (rec_len >= name_len + reclen)
489 goto got_it;
490 de = (struct exofs_dir_entry *) ((char *) de + rec_len);
491 }
492 unlock_page(page);
493 exofs_put_page(page);
494 }
495
496 EXOFS_ERR("exofs_add_link: BAD dentry=%p or inode=%p", dentry, inode);
497 return -EINVAL;
498
499got_it:
500 pos = page_offset(page) +
501 (char *)de - (char *)page_address(page);
502 err = exofs_write_begin(NULL, page->mapping, pos, rec_len, 0,
503 &page, NULL);
504 if (err)
505 goto out_unlock;
506 if (de->inode_no) {
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);
511 de = de1;
512 }
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);
520 sbi->s_numfiles++;
521
522out_put:
523 exofs_put_page(page);
524out:
525 return err;
526out_unlock:
527 unlock_page(page);
528 goto out_put;
529}
530
531int exofs_delete_entry(struct exofs_dir_entry *dir, struct page *page)
532{
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);
539 loff_t pos;
540 struct exofs_dir_entry *pde = NULL;
541 struct exofs_dir_entry *de = (struct exofs_dir_entry *) (kaddr + from);
542 int err;
543
544 while (de < dir) {
545 if (de->rec_len == 0) {
546 EXOFS_ERR("ERROR: exofs_delete_entry:"
547 "zero-length directory entry");
548 err = -EIO;
549 goto out;
550 }
551 pde = de;
552 de = exofs_next_entry(de);
553 }
554 if (pde)
555 from = (char *)pde - (char *)page_address(page);
556 pos = page_offset(page) + from;
557 lock_page(page);
558 err = exofs_write_begin(NULL, page->mapping, pos, to - from, 0,
559 &page, NULL);
560 if (err)
561 EXOFS_ERR("exofs_delete_entry: exofs_write_begin FAILD => %d\n",
562 err);
563 if (pde)
564 pde->rec_len = cpu_to_le16(to - from);
565 dir->inode_no = 0;
566 if (likely(!err))
567 err = exofs_commit_chunk(page, pos, to - from);
568 inode->i_ctime = inode->i_mtime = CURRENT_TIME;
569 mark_inode_dirty(inode);
570 sbi->s_numfiles--;
571out:
572 exofs_put_page(page);
573 return err;
574}
575
576/* kept aligned on 4 bytes */
577#define THIS_DIR ".\0\0"
578#define PARENT_DIR "..\0"
579
580int exofs_make_empty(struct inode *inode, struct inode *parent)
581{
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;
586 int err;
587 void *kaddr;
588
589 if (!page)
590 return -ENOMEM;
591
592 err = exofs_write_begin(NULL, page->mapping, 0, chunk_size, 0,
593 &page, NULL);
594 if (err) {
595 unlock_page(page);
596 goto fail;
597 }
598
599 kaddr = kmap_atomic(page, KM_USER0);
600 de = (struct exofs_dir_entry *)kaddr;
601 de->name_len = 1;
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);
606
607 de = (struct exofs_dir_entry *)(kaddr + EXOFS_DIR_REC_LEN(1));
608 de->name_len = 2;
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);
615fail:
616 page_cache_release(page);
617 return err;
618}
619
620int exofs_empty_dir(struct inode *inode)
621{
622 struct page *page = NULL;
623 unsigned long i, npages = dir_pages(inode);
624
625 for (i = 0; i < npages; i++) {
626 char *kaddr;
627 struct exofs_dir_entry *de;
628 page = exofs_get_page(inode, i);
629
630 if (IS_ERR(page))
631 continue;
632
633 kaddr = page_address(page);
634 de = (struct exofs_dir_entry *)kaddr;
635 kaddr += exofs_last_byte(inode, i) - EXOFS_DIR_REC_LEN(1);
636
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);
642 goto not_empty;
643 }
644 if (de->inode_no != 0) {
645 /* check for . and .. */
646 if (de->name[0] != '.')
647 goto not_empty;
648 if (de->name_len > 2)
649 goto not_empty;
650 if (de->name_len < 2) {
651 if (le64_to_cpu(de->inode_no) !=
652 inode->i_ino)
653 goto not_empty;
654 } else if (de->name[1] != '.')
655 goto not_empty;
656 }
657 de = exofs_next_entry(de);
658 }
659 exofs_put_page(page);
660 }
661 return 1;
662
663not_empty:
664 exofs_put_page(page);
665 return 0;
666}
667
668const struct file_operations exofs_dir_operations = {
669 .llseek = generic_file_llseek,
670 .read = generic_read_dir,
671 .readdir = exofs_readdir,
672};