IPVS: fix netns if reading ip_vs_* procfs entries
[linux-2.6/linux-mips.git] / fs / nilfs2 / dir.c
blob3a1923943b14c9300bd8af952451efa86177c09c
1 /*
2 * dir.c - NILFS directory entry operations
4 * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 * Modified for NILFS by Amagai Yoshiji <amagai@osrg.net>
23 * linux/fs/ext2/dir.c
25 * Copyright (C) 1992, 1993, 1994, 1995
26 * Remy Card (card@masi.ibp.fr)
27 * Laboratoire MASI - Institut Blaise Pascal
28 * Universite Pierre et Marie Curie (Paris VI)
30 * from
32 * linux/fs/minix/dir.c
34 * Copyright (C) 1991, 1992 Linus Torvalds
36 * ext2 directory handling functions
38 * Big-endian to little-endian byte-swapping/bitmaps by
39 * David S. Miller (davem@caip.rutgers.edu), 1995
41 * All code that works with directory layout had been switched to pagecache
42 * and moved here. AV
45 #include <linux/pagemap.h>
46 #include "nilfs.h"
47 #include "page.h"
50 * nilfs uses block-sized chunks. Arguably, sector-sized ones would be
51 * more robust, but we have what we have
53 static inline unsigned nilfs_chunk_size(struct inode *inode)
55 return inode->i_sb->s_blocksize;
58 static inline void nilfs_put_page(struct page *page)
60 kunmap(page);
61 page_cache_release(page);
64 static inline unsigned long dir_pages(struct inode *inode)
66 return (inode->i_size+PAGE_CACHE_SIZE-1)>>PAGE_CACHE_SHIFT;
70 * Return the offset into page `page_nr' of the last valid
71 * byte in that page, plus one.
73 static unsigned nilfs_last_byte(struct inode *inode, unsigned long page_nr)
75 unsigned last_byte = inode->i_size;
77 last_byte -= page_nr << PAGE_CACHE_SHIFT;
78 if (last_byte > PAGE_CACHE_SIZE)
79 last_byte = PAGE_CACHE_SIZE;
80 return last_byte;
83 static int nilfs_prepare_chunk(struct page *page, unsigned from, unsigned to)
85 loff_t pos = page_offset(page) + from;
86 return __block_write_begin(page, pos, to - from, nilfs_get_block);
89 static void nilfs_commit_chunk(struct page *page,
90 struct address_space *mapping,
91 unsigned from, unsigned to)
93 struct inode *dir = mapping->host;
94 loff_t pos = page_offset(page) + from;
95 unsigned len = to - from;
96 unsigned nr_dirty, copied;
97 int err;
99 nr_dirty = nilfs_page_count_clean_buffers(page, from, to);
100 copied = block_write_end(NULL, mapping, pos, len, len, page, NULL);
101 if (pos + copied > dir->i_size)
102 i_size_write(dir, pos + copied);
103 if (IS_DIRSYNC(dir))
104 nilfs_set_transaction_flag(NILFS_TI_SYNC);
105 err = nilfs_set_file_dirty(dir, nr_dirty);
106 WARN_ON(err); /* do not happen */
107 unlock_page(page);
110 static void nilfs_check_page(struct page *page)
112 struct inode *dir = page->mapping->host;
113 struct super_block *sb = dir->i_sb;
114 unsigned chunk_size = nilfs_chunk_size(dir);
115 char *kaddr = page_address(page);
116 unsigned offs, rec_len;
117 unsigned limit = PAGE_CACHE_SIZE;
118 struct nilfs_dir_entry *p;
119 char *error;
121 if ((dir->i_size >> PAGE_CACHE_SHIFT) == page->index) {
122 limit = dir->i_size & ~PAGE_CACHE_MASK;
123 if (limit & (chunk_size - 1))
124 goto Ebadsize;
125 if (!limit)
126 goto out;
128 for (offs = 0; offs <= limit - NILFS_DIR_REC_LEN(1); offs += rec_len) {
129 p = (struct nilfs_dir_entry *)(kaddr + offs);
130 rec_len = nilfs_rec_len_from_disk(p->rec_len);
132 if (rec_len < NILFS_DIR_REC_LEN(1))
133 goto Eshort;
134 if (rec_len & 3)
135 goto Ealign;
136 if (rec_len < NILFS_DIR_REC_LEN(p->name_len))
137 goto Enamelen;
138 if (((offs + rec_len - 1) ^ offs) & ~(chunk_size-1))
139 goto Espan;
141 if (offs != limit)
142 goto Eend;
143 out:
144 SetPageChecked(page);
145 return;
147 /* Too bad, we had an error */
149 Ebadsize:
150 nilfs_error(sb, "nilfs_check_page",
151 "size of directory #%lu is not a multiple of chunk size",
152 dir->i_ino
154 goto fail;
155 Eshort:
156 error = "rec_len is smaller than minimal";
157 goto bad_entry;
158 Ealign:
159 error = "unaligned directory entry";
160 goto bad_entry;
161 Enamelen:
162 error = "rec_len is too small for name_len";
163 goto bad_entry;
164 Espan:
165 error = "directory entry across blocks";
166 bad_entry:
167 nilfs_error(sb, "nilfs_check_page", "bad entry in directory #%lu: %s - "
168 "offset=%lu, inode=%lu, rec_len=%d, name_len=%d",
169 dir->i_ino, error, (page->index<<PAGE_CACHE_SHIFT)+offs,
170 (unsigned long) le64_to_cpu(p->inode),
171 rec_len, p->name_len);
172 goto fail;
173 Eend:
174 p = (struct nilfs_dir_entry *)(kaddr + offs);
175 nilfs_error(sb, "nilfs_check_page",
176 "entry in directory #%lu spans the page boundary"
177 "offset=%lu, inode=%lu",
178 dir->i_ino, (page->index<<PAGE_CACHE_SHIFT)+offs,
179 (unsigned long) le64_to_cpu(p->inode));
180 fail:
181 SetPageChecked(page);
182 SetPageError(page);
185 static struct page *nilfs_get_page(struct inode *dir, unsigned long n)
187 struct address_space *mapping = dir->i_mapping;
188 struct page *page = read_mapping_page(mapping, n, NULL);
190 if (!IS_ERR(page)) {
191 kmap(page);
192 if (!PageChecked(page))
193 nilfs_check_page(page);
194 if (PageError(page))
195 goto fail;
197 return page;
199 fail:
200 nilfs_put_page(page);
201 return ERR_PTR(-EIO);
205 * NOTE! unlike strncmp, nilfs_match returns 1 for success, 0 for failure.
207 * len <= NILFS_NAME_LEN and de != NULL are guaranteed by caller.
209 static int
210 nilfs_match(int len, const unsigned char *name, struct nilfs_dir_entry *de)
212 if (len != de->name_len)
213 return 0;
214 if (!de->inode)
215 return 0;
216 return !memcmp(name, de->name, len);
220 * p is at least 6 bytes before the end of page
222 static struct nilfs_dir_entry *nilfs_next_entry(struct nilfs_dir_entry *p)
224 return (struct nilfs_dir_entry *)((char *)p +
225 nilfs_rec_len_from_disk(p->rec_len));
228 static unsigned char
229 nilfs_filetype_table[NILFS_FT_MAX] = {
230 [NILFS_FT_UNKNOWN] = DT_UNKNOWN,
231 [NILFS_FT_REG_FILE] = DT_REG,
232 [NILFS_FT_DIR] = DT_DIR,
233 [NILFS_FT_CHRDEV] = DT_CHR,
234 [NILFS_FT_BLKDEV] = DT_BLK,
235 [NILFS_FT_FIFO] = DT_FIFO,
236 [NILFS_FT_SOCK] = DT_SOCK,
237 [NILFS_FT_SYMLINK] = DT_LNK,
240 #define S_SHIFT 12
241 static unsigned char
242 nilfs_type_by_mode[S_IFMT >> S_SHIFT] = {
243 [S_IFREG >> S_SHIFT] = NILFS_FT_REG_FILE,
244 [S_IFDIR >> S_SHIFT] = NILFS_FT_DIR,
245 [S_IFCHR >> S_SHIFT] = NILFS_FT_CHRDEV,
246 [S_IFBLK >> S_SHIFT] = NILFS_FT_BLKDEV,
247 [S_IFIFO >> S_SHIFT] = NILFS_FT_FIFO,
248 [S_IFSOCK >> S_SHIFT] = NILFS_FT_SOCK,
249 [S_IFLNK >> S_SHIFT] = NILFS_FT_SYMLINK,
252 static void nilfs_set_de_type(struct nilfs_dir_entry *de, struct inode *inode)
254 mode_t mode = inode->i_mode;
256 de->file_type = nilfs_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
259 static int nilfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
261 loff_t pos = filp->f_pos;
262 struct inode *inode = filp->f_dentry->d_inode;
263 struct super_block *sb = inode->i_sb;
264 unsigned int offset = pos & ~PAGE_CACHE_MASK;
265 unsigned long n = pos >> PAGE_CACHE_SHIFT;
266 unsigned long npages = dir_pages(inode);
267 /* unsigned chunk_mask = ~(nilfs_chunk_size(inode)-1); */
268 unsigned char *types = NULL;
269 int ret;
271 if (pos > inode->i_size - NILFS_DIR_REC_LEN(1))
272 goto success;
274 types = nilfs_filetype_table;
276 for ( ; n < npages; n++, offset = 0) {
277 char *kaddr, *limit;
278 struct nilfs_dir_entry *de;
279 struct page *page = nilfs_get_page(inode, n);
281 if (IS_ERR(page)) {
282 nilfs_error(sb, __func__, "bad page in #%lu",
283 inode->i_ino);
284 filp->f_pos += PAGE_CACHE_SIZE - offset;
285 ret = -EIO;
286 goto done;
288 kaddr = page_address(page);
289 de = (struct nilfs_dir_entry *)(kaddr + offset);
290 limit = kaddr + nilfs_last_byte(inode, n) -
291 NILFS_DIR_REC_LEN(1);
292 for ( ; (char *)de <= limit; de = nilfs_next_entry(de)) {
293 if (de->rec_len == 0) {
294 nilfs_error(sb, __func__,
295 "zero-length directory entry");
296 ret = -EIO;
297 nilfs_put_page(page);
298 goto done;
300 if (de->inode) {
301 int over;
302 unsigned char d_type = DT_UNKNOWN;
304 if (types && de->file_type < NILFS_FT_MAX)
305 d_type = types[de->file_type];
307 offset = (char *)de - kaddr;
308 over = filldir(dirent, de->name, de->name_len,
309 (n<<PAGE_CACHE_SHIFT) | offset,
310 le64_to_cpu(de->inode), d_type);
311 if (over) {
312 nilfs_put_page(page);
313 goto success;
316 filp->f_pos += nilfs_rec_len_from_disk(de->rec_len);
318 nilfs_put_page(page);
321 success:
322 ret = 0;
323 done:
324 return ret;
328 * nilfs_find_entry()
330 * finds an entry in the specified directory with the wanted name. It
331 * returns the page in which the entry was found, and the entry itself
332 * (as a parameter - res_dir). Page is returned mapped and unlocked.
333 * Entry is guaranteed to be valid.
335 struct nilfs_dir_entry *
336 nilfs_find_entry(struct inode *dir, const struct qstr *qstr,
337 struct page **res_page)
339 const unsigned char *name = qstr->name;
340 int namelen = qstr->len;
341 unsigned reclen = NILFS_DIR_REC_LEN(namelen);
342 unsigned long start, n;
343 unsigned long npages = dir_pages(dir);
344 struct page *page = NULL;
345 struct nilfs_inode_info *ei = NILFS_I(dir);
346 struct nilfs_dir_entry *de;
348 if (npages == 0)
349 goto out;
351 /* OFFSET_CACHE */
352 *res_page = NULL;
354 start = ei->i_dir_start_lookup;
355 if (start >= npages)
356 start = 0;
357 n = start;
358 do {
359 char *kaddr;
360 page = nilfs_get_page(dir, n);
361 if (!IS_ERR(page)) {
362 kaddr = page_address(page);
363 de = (struct nilfs_dir_entry *)kaddr;
364 kaddr += nilfs_last_byte(dir, n) - reclen;
365 while ((char *) de <= kaddr) {
366 if (de->rec_len == 0) {
367 nilfs_error(dir->i_sb, __func__,
368 "zero-length directory entry");
369 nilfs_put_page(page);
370 goto out;
372 if (nilfs_match(namelen, name, de))
373 goto found;
374 de = nilfs_next_entry(de);
376 nilfs_put_page(page);
378 if (++n >= npages)
379 n = 0;
380 /* next page is past the blocks we've got */
381 if (unlikely(n > (dir->i_blocks >> (PAGE_CACHE_SHIFT - 9)))) {
382 nilfs_error(dir->i_sb, __func__,
383 "dir %lu size %lld exceeds block count %llu",
384 dir->i_ino, dir->i_size,
385 (unsigned long long)dir->i_blocks);
386 goto out;
388 } while (n != start);
389 out:
390 return NULL;
392 found:
393 *res_page = page;
394 ei->i_dir_start_lookup = n;
395 return de;
398 struct nilfs_dir_entry *nilfs_dotdot(struct inode *dir, struct page **p)
400 struct page *page = nilfs_get_page(dir, 0);
401 struct nilfs_dir_entry *de = NULL;
403 if (!IS_ERR(page)) {
404 de = nilfs_next_entry(
405 (struct nilfs_dir_entry *)page_address(page));
406 *p = page;
408 return de;
411 ino_t nilfs_inode_by_name(struct inode *dir, const struct qstr *qstr)
413 ino_t res = 0;
414 struct nilfs_dir_entry *de;
415 struct page *page;
417 de = nilfs_find_entry(dir, qstr, &page);
418 if (de) {
419 res = le64_to_cpu(de->inode);
420 kunmap(page);
421 page_cache_release(page);
423 return res;
426 /* Releases the page */
427 void nilfs_set_link(struct inode *dir, struct nilfs_dir_entry *de,
428 struct page *page, struct inode *inode)
430 unsigned from = (char *) de - (char *) page_address(page);
431 unsigned to = from + nilfs_rec_len_from_disk(de->rec_len);
432 struct address_space *mapping = page->mapping;
433 int err;
435 lock_page(page);
436 err = nilfs_prepare_chunk(page, from, to);
437 BUG_ON(err);
438 de->inode = cpu_to_le64(inode->i_ino);
439 nilfs_set_de_type(de, inode);
440 nilfs_commit_chunk(page, mapping, from, to);
441 nilfs_put_page(page);
442 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
446 * Parent is locked.
448 int nilfs_add_link(struct dentry *dentry, struct inode *inode)
450 struct inode *dir = dentry->d_parent->d_inode;
451 const unsigned char *name = dentry->d_name.name;
452 int namelen = dentry->d_name.len;
453 unsigned chunk_size = nilfs_chunk_size(dir);
454 unsigned reclen = NILFS_DIR_REC_LEN(namelen);
455 unsigned short rec_len, name_len;
456 struct page *page = NULL;
457 struct nilfs_dir_entry *de;
458 unsigned long npages = dir_pages(dir);
459 unsigned long n;
460 char *kaddr;
461 unsigned from, to;
462 int err;
465 * We take care of directory expansion in the same loop.
466 * This code plays outside i_size, so it locks the page
467 * to protect that region.
469 for (n = 0; n <= npages; n++) {
470 char *dir_end;
472 page = nilfs_get_page(dir, n);
473 err = PTR_ERR(page);
474 if (IS_ERR(page))
475 goto out;
476 lock_page(page);
477 kaddr = page_address(page);
478 dir_end = kaddr + nilfs_last_byte(dir, n);
479 de = (struct nilfs_dir_entry *)kaddr;
480 kaddr += PAGE_CACHE_SIZE - reclen;
481 while ((char *)de <= kaddr) {
482 if ((char *)de == dir_end) {
483 /* We hit i_size */
484 name_len = 0;
485 rec_len = chunk_size;
486 de->rec_len = nilfs_rec_len_to_disk(chunk_size);
487 de->inode = 0;
488 goto got_it;
490 if (de->rec_len == 0) {
491 nilfs_error(dir->i_sb, __func__,
492 "zero-length directory entry");
493 err = -EIO;
494 goto out_unlock;
496 err = -EEXIST;
497 if (nilfs_match(namelen, name, de))
498 goto out_unlock;
499 name_len = NILFS_DIR_REC_LEN(de->name_len);
500 rec_len = nilfs_rec_len_from_disk(de->rec_len);
501 if (!de->inode && rec_len >= reclen)
502 goto got_it;
503 if (rec_len >= name_len + reclen)
504 goto got_it;
505 de = (struct nilfs_dir_entry *)((char *)de + rec_len);
507 unlock_page(page);
508 nilfs_put_page(page);
510 BUG();
511 return -EINVAL;
513 got_it:
514 from = (char *)de - (char *)page_address(page);
515 to = from + rec_len;
516 err = nilfs_prepare_chunk(page, from, to);
517 if (err)
518 goto out_unlock;
519 if (de->inode) {
520 struct nilfs_dir_entry *de1;
522 de1 = (struct nilfs_dir_entry *)((char *)de + name_len);
523 de1->rec_len = nilfs_rec_len_to_disk(rec_len - name_len);
524 de->rec_len = nilfs_rec_len_to_disk(name_len);
525 de = de1;
527 de->name_len = namelen;
528 memcpy(de->name, name, namelen);
529 de->inode = cpu_to_le64(inode->i_ino);
530 nilfs_set_de_type(de, inode);
531 nilfs_commit_chunk(page, page->mapping, from, to);
532 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
533 nilfs_mark_inode_dirty(dir);
534 /* OFFSET_CACHE */
535 out_put:
536 nilfs_put_page(page);
537 out:
538 return err;
539 out_unlock:
540 unlock_page(page);
541 goto out_put;
545 * nilfs_delete_entry deletes a directory entry by merging it with the
546 * previous entry. Page is up-to-date. Releases the page.
548 int nilfs_delete_entry(struct nilfs_dir_entry *dir, struct page *page)
550 struct address_space *mapping = page->mapping;
551 struct inode *inode = mapping->host;
552 char *kaddr = page_address(page);
553 unsigned from = ((char *)dir - kaddr) & ~(nilfs_chunk_size(inode) - 1);
554 unsigned to = ((char *)dir - kaddr) +
555 nilfs_rec_len_from_disk(dir->rec_len);
556 struct nilfs_dir_entry *pde = NULL;
557 struct nilfs_dir_entry *de = (struct nilfs_dir_entry *)(kaddr + from);
558 int err;
560 while ((char *)de < (char *)dir) {
561 if (de->rec_len == 0) {
562 nilfs_error(inode->i_sb, __func__,
563 "zero-length directory entry");
564 err = -EIO;
565 goto out;
567 pde = de;
568 de = nilfs_next_entry(de);
570 if (pde)
571 from = (char *)pde - (char *)page_address(page);
572 lock_page(page);
573 err = nilfs_prepare_chunk(page, from, to);
574 BUG_ON(err);
575 if (pde)
576 pde->rec_len = nilfs_rec_len_to_disk(to - from);
577 dir->inode = 0;
578 nilfs_commit_chunk(page, mapping, from, to);
579 inode->i_ctime = inode->i_mtime = CURRENT_TIME;
580 out:
581 nilfs_put_page(page);
582 return err;
586 * Set the first fragment of directory.
588 int nilfs_make_empty(struct inode *inode, struct inode *parent)
590 struct address_space *mapping = inode->i_mapping;
591 struct page *page = grab_cache_page(mapping, 0);
592 unsigned chunk_size = nilfs_chunk_size(inode);
593 struct nilfs_dir_entry *de;
594 int err;
595 void *kaddr;
597 if (!page)
598 return -ENOMEM;
600 err = nilfs_prepare_chunk(page, 0, chunk_size);
601 if (unlikely(err)) {
602 unlock_page(page);
603 goto fail;
605 kaddr = kmap_atomic(page, KM_USER0);
606 memset(kaddr, 0, chunk_size);
607 de = (struct nilfs_dir_entry *)kaddr;
608 de->name_len = 1;
609 de->rec_len = nilfs_rec_len_to_disk(NILFS_DIR_REC_LEN(1));
610 memcpy(de->name, ".\0\0", 4);
611 de->inode = cpu_to_le64(inode->i_ino);
612 nilfs_set_de_type(de, inode);
614 de = (struct nilfs_dir_entry *)(kaddr + NILFS_DIR_REC_LEN(1));
615 de->name_len = 2;
616 de->rec_len = nilfs_rec_len_to_disk(chunk_size - NILFS_DIR_REC_LEN(1));
617 de->inode = cpu_to_le64(parent->i_ino);
618 memcpy(de->name, "..\0", 4);
619 nilfs_set_de_type(de, inode);
620 kunmap_atomic(kaddr, KM_USER0);
621 nilfs_commit_chunk(page, mapping, 0, chunk_size);
622 fail:
623 page_cache_release(page);
624 return err;
628 * routine to check that the specified directory is empty (for rmdir)
630 int nilfs_empty_dir(struct inode *inode)
632 struct page *page = NULL;
633 unsigned long i, npages = dir_pages(inode);
635 for (i = 0; i < npages; i++) {
636 char *kaddr;
637 struct nilfs_dir_entry *de;
639 page = nilfs_get_page(inode, i);
640 if (IS_ERR(page))
641 continue;
643 kaddr = page_address(page);
644 de = (struct nilfs_dir_entry *)kaddr;
645 kaddr += nilfs_last_byte(inode, i) - NILFS_DIR_REC_LEN(1);
647 while ((char *)de <= kaddr) {
648 if (de->rec_len == 0) {
649 nilfs_error(inode->i_sb, __func__,
650 "zero-length directory entry "
651 "(kaddr=%p, de=%p)\n", kaddr, de);
652 goto not_empty;
654 if (de->inode != 0) {
655 /* check for . and .. */
656 if (de->name[0] != '.')
657 goto not_empty;
658 if (de->name_len > 2)
659 goto not_empty;
660 if (de->name_len < 2) {
661 if (de->inode !=
662 cpu_to_le64(inode->i_ino))
663 goto not_empty;
664 } else if (de->name[1] != '.')
665 goto not_empty;
667 de = nilfs_next_entry(de);
669 nilfs_put_page(page);
671 return 1;
673 not_empty:
674 nilfs_put_page(page);
675 return 0;
678 const struct file_operations nilfs_dir_operations = {
679 .llseek = generic_file_llseek,
680 .read = generic_read_dir,
681 .readdir = nilfs_readdir,
682 .unlocked_ioctl = nilfs_ioctl,
683 #ifdef CONFIG_COMPAT
684 .compat_ioctl = nilfs_compat_ioctl,
685 #endif /* CONFIG_COMPAT */
686 .fsync = nilfs_sync_file,