2 * linux/fs/ext2/inode.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Goal-directed block allocation by Stephen Tweedie
16 * (sct@dcs.ed.ac.uk), 1993, 1998
17 * Big-endian to little-endian byte-swapping/bitmaps by
18 * David S. Miller (davem@caip.rutgers.edu), 1995
19 * 64-bit file support on 64-bit platforms by Jakub Jelinek
20 * (jj@sunsite.ms.mff.cuni.cz)
22 * Assorted race fixes, rewrite of ext2_get_block() by Al Viro, 2000
25 #include <linux/smp_lock.h>
26 #include <linux/time.h>
27 #include <linux/highuid.h>
28 #include <linux/pagemap.h>
29 #include <linux/quotaops.h>
30 #include <linux/module.h>
31 #include <linux/writeback.h>
32 #include <linux/buffer_head.h>
33 #include <linux/mpage.h>
38 MODULE_AUTHOR("Remy Card and others");
39 MODULE_DESCRIPTION("Second Extended Filesystem");
40 MODULE_LICENSE("GPL");
42 static int ext2_update_inode(struct inode
* inode
, int do_sync
);
45 * Test whether an inode is a fast symlink.
47 static inline int ext2_inode_is_fast_symlink(struct inode
*inode
)
49 int ea_blocks
= EXT2_I(inode
)->i_file_acl
?
50 (inode
->i_sb
->s_blocksize
>> 9) : 0;
52 return (S_ISLNK(inode
->i_mode
) &&
53 inode
->i_blocks
- ea_blocks
== 0);
57 * Called at each iput().
59 * The inode may be "bad" if ext2_read_inode() saw an error from
60 * ext2_get_inode(), so we need to check that to avoid freeing random disk
63 void ext2_put_inode(struct inode
*inode
)
65 if (!is_bad_inode(inode
))
66 ext2_discard_prealloc(inode
);
70 * Called at the last iput() if i_nlink is zero.
72 void ext2_delete_inode (struct inode
* inode
)
74 truncate_inode_pages(&inode
->i_data
, 0);
76 if (is_bad_inode(inode
))
78 EXT2_I(inode
)->i_dtime
= get_seconds();
79 mark_inode_dirty(inode
);
80 ext2_update_inode(inode
, inode_needs_sync(inode
));
84 ext2_truncate (inode
);
85 ext2_free_inode (inode
);
89 clear_inode(inode
); /* We must guarantee clearing of inode... */
92 void ext2_discard_prealloc (struct inode
* inode
)
94 #ifdef EXT2_PREALLOCATE
95 struct ext2_inode_info
*ei
= EXT2_I(inode
);
96 write_lock(&ei
->i_meta_lock
);
97 if (ei
->i_prealloc_count
) {
98 unsigned short total
= ei
->i_prealloc_count
;
99 unsigned long block
= ei
->i_prealloc_block
;
100 ei
->i_prealloc_count
= 0;
101 ei
->i_prealloc_block
= 0;
102 write_unlock(&ei
->i_meta_lock
);
103 ext2_free_blocks (inode
, block
, total
);
106 write_unlock(&ei
->i_meta_lock
);
110 static int ext2_alloc_block (struct inode
* inode
, unsigned long goal
, int *err
)
113 static unsigned long alloc_hits
, alloc_attempts
;
115 unsigned long result
;
118 #ifdef EXT2_PREALLOCATE
119 struct ext2_inode_info
*ei
= EXT2_I(inode
);
120 write_lock(&ei
->i_meta_lock
);
121 if (ei
->i_prealloc_count
&&
122 (goal
== ei
->i_prealloc_block
|| goal
+ 1 == ei
->i_prealloc_block
))
124 result
= ei
->i_prealloc_block
++;
125 ei
->i_prealloc_count
--;
126 write_unlock(&ei
->i_meta_lock
);
127 ext2_debug ("preallocation hit (%lu/%lu).\n",
128 ++alloc_hits
, ++alloc_attempts
);
130 write_unlock(&ei
->i_meta_lock
);
131 ext2_discard_prealloc (inode
);
132 ext2_debug ("preallocation miss (%lu/%lu).\n",
133 alloc_hits
, ++alloc_attempts
);
134 if (S_ISREG(inode
->i_mode
))
135 result
= ext2_new_block (inode
, goal
,
136 &ei
->i_prealloc_count
,
137 &ei
->i_prealloc_block
, err
);
139 result
= ext2_new_block(inode
, goal
, NULL
, NULL
, err
);
142 result
= ext2_new_block (inode
, goal
, 0, 0, err
);
150 struct buffer_head
*bh
;
153 static inline void add_chain(Indirect
*p
, struct buffer_head
*bh
, __le32
*v
)
155 p
->key
= *(p
->p
= v
);
159 static inline int verify_chain(Indirect
*from
, Indirect
*to
)
161 while (from
<= to
&& from
->key
== *from
->p
)
167 * ext2_block_to_path - parse the block number into array of offsets
168 * @inode: inode in question (we are only interested in its superblock)
169 * @i_block: block number to be parsed
170 * @offsets: array to store the offsets in
171 * @boundary: set this non-zero if the referred-to block is likely to be
172 * followed (on disk) by an indirect block.
173 * To store the locations of file's data ext2 uses a data structure common
174 * for UNIX filesystems - tree of pointers anchored in the inode, with
175 * data blocks at leaves and indirect blocks in intermediate nodes.
176 * This function translates the block number into path in that tree -
177 * return value is the path length and @offsets[n] is the offset of
178 * pointer to (n+1)th node in the nth one. If @block is out of range
179 * (negative or too large) warning is printed and zero returned.
181 * Note: function doesn't find node addresses, so no IO is needed. All
182 * we need to know is the capacity of indirect blocks (taken from the
187 * Portability note: the last comparison (check that we fit into triple
188 * indirect block) is spelled differently, because otherwise on an
189 * architecture with 32-bit longs and 8Kb pages we might get into trouble
190 * if our filesystem had 8Kb blocks. We might use long long, but that would
191 * kill us on x86. Oh, well, at least the sign propagation does not matter -
192 * i_block would have to be negative in the very beginning, so we would not
196 static int ext2_block_to_path(struct inode
*inode
,
197 long i_block
, int offsets
[4], int *boundary
)
199 int ptrs
= EXT2_ADDR_PER_BLOCK(inode
->i_sb
);
200 int ptrs_bits
= EXT2_ADDR_PER_BLOCK_BITS(inode
->i_sb
);
201 const long direct_blocks
= EXT2_NDIR_BLOCKS
,
202 indirect_blocks
= ptrs
,
203 double_blocks
= (1 << (ptrs_bits
* 2));
208 ext2_warning (inode
->i_sb
, "ext2_block_to_path", "block < 0");
209 } else if (i_block
< direct_blocks
) {
210 offsets
[n
++] = i_block
;
211 final
= direct_blocks
;
212 } else if ( (i_block
-= direct_blocks
) < indirect_blocks
) {
213 offsets
[n
++] = EXT2_IND_BLOCK
;
214 offsets
[n
++] = i_block
;
216 } else if ((i_block
-= indirect_blocks
) < double_blocks
) {
217 offsets
[n
++] = EXT2_DIND_BLOCK
;
218 offsets
[n
++] = i_block
>> ptrs_bits
;
219 offsets
[n
++] = i_block
& (ptrs
- 1);
221 } else if (((i_block
-= double_blocks
) >> (ptrs_bits
* 2)) < ptrs
) {
222 offsets
[n
++] = EXT2_TIND_BLOCK
;
223 offsets
[n
++] = i_block
>> (ptrs_bits
* 2);
224 offsets
[n
++] = (i_block
>> ptrs_bits
) & (ptrs
- 1);
225 offsets
[n
++] = i_block
& (ptrs
- 1);
228 ext2_warning (inode
->i_sb
, "ext2_block_to_path", "block > big");
231 *boundary
= (i_block
& (ptrs
- 1)) == (final
- 1);
236 * ext2_get_branch - read the chain of indirect blocks leading to data
237 * @inode: inode in question
238 * @depth: depth of the chain (1 - direct pointer, etc.)
239 * @offsets: offsets of pointers in inode/indirect blocks
240 * @chain: place to store the result
241 * @err: here we store the error value
243 * Function fills the array of triples <key, p, bh> and returns %NULL
244 * if everything went OK or the pointer to the last filled triple
245 * (incomplete one) otherwise. Upon the return chain[i].key contains
246 * the number of (i+1)-th block in the chain (as it is stored in memory,
247 * i.e. little-endian 32-bit), chain[i].p contains the address of that
248 * number (it points into struct inode for i==0 and into the bh->b_data
249 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect
250 * block for i>0 and NULL for i==0. In other words, it holds the block
251 * numbers of the chain, addresses they were taken from (and where we can
252 * verify that chain did not change) and buffer_heads hosting these
255 * Function stops when it stumbles upon zero pointer (absent block)
256 * (pointer to last triple returned, *@err == 0)
257 * or when it gets an IO error reading an indirect block
258 * (ditto, *@err == -EIO)
259 * or when it notices that chain had been changed while it was reading
260 * (ditto, *@err == -EAGAIN)
261 * or when it reads all @depth-1 indirect blocks successfully and finds
262 * the whole chain, all way to the data (returns %NULL, *err == 0).
264 static Indirect
*ext2_get_branch(struct inode
*inode
,
270 struct super_block
*sb
= inode
->i_sb
;
272 struct buffer_head
*bh
;
275 /* i_data is not going away, no lock needed */
276 add_chain (chain
, NULL
, EXT2_I(inode
)->i_data
+ *offsets
);
280 bh
= sb_bread(sb
, le32_to_cpu(p
->key
));
283 read_lock(&EXT2_I(inode
)->i_meta_lock
);
284 if (!verify_chain(chain
, p
))
286 add_chain(++p
, bh
, (__le32
*)bh
->b_data
+ *++offsets
);
287 read_unlock(&EXT2_I(inode
)->i_meta_lock
);
294 read_unlock(&EXT2_I(inode
)->i_meta_lock
);
305 * ext2_find_near - find a place for allocation with sufficient locality
307 * @ind: descriptor of indirect block.
309 * This function returns the prefered place for block allocation.
310 * It is used when heuristic for sequential allocation fails.
312 * + if there is a block to the left of our position - allocate near it.
313 * + if pointer will live in indirect block - allocate near that block.
314 * + if pointer will live in inode - allocate in the same cylinder group.
316 * In the latter case we colour the starting block by the callers PID to
317 * prevent it from clashing with concurrent allocations for a different inode
318 * in the same block group. The PID is used here so that functionally related
319 * files will be close-by on-disk.
321 * Caller must make sure that @ind is valid and will stay that way.
324 static unsigned long ext2_find_near(struct inode
*inode
, Indirect
*ind
)
326 struct ext2_inode_info
*ei
= EXT2_I(inode
);
327 __le32
*start
= ind
->bh
? (__le32
*) ind
->bh
->b_data
: ei
->i_data
;
329 unsigned long bg_start
;
330 unsigned long colour
;
332 /* Try to find previous block */
333 for (p
= ind
->p
- 1; p
>= start
; p
--)
335 return le32_to_cpu(*p
);
337 /* No such thing, so let's try location of indirect block */
339 return ind
->bh
->b_blocknr
;
342 * It is going to be refered from inode itself? OK, just put it into
343 * the same cylinder group then.
345 bg_start
= (ei
->i_block_group
* EXT2_BLOCKS_PER_GROUP(inode
->i_sb
)) +
346 le32_to_cpu(EXT2_SB(inode
->i_sb
)->s_es
->s_first_data_block
);
347 colour
= (current
->pid
% 16) *
348 (EXT2_BLOCKS_PER_GROUP(inode
->i_sb
) / 16);
349 return bg_start
+ colour
;
353 * ext2_find_goal - find a prefered place for allocation.
355 * @block: block we want
356 * @chain: chain of indirect blocks
357 * @partial: pointer to the last triple within a chain
358 * @goal: place to store the result.
360 * Normally this function find the prefered place for block allocation,
361 * stores it in *@goal and returns zero. If the branch had been changed
362 * under us we return -EAGAIN.
365 static inline int ext2_find_goal(struct inode
*inode
,
371 struct ext2_inode_info
*ei
= EXT2_I(inode
);
372 write_lock(&ei
->i_meta_lock
);
373 if ((block
== ei
->i_next_alloc_block
+ 1) && ei
->i_next_alloc_goal
) {
374 ei
->i_next_alloc_block
++;
375 ei
->i_next_alloc_goal
++;
377 if (verify_chain(chain
, partial
)) {
379 * try the heuristic for sequential allocation,
380 * failing that at least try to get decent locality.
382 if (block
== ei
->i_next_alloc_block
)
383 *goal
= ei
->i_next_alloc_goal
;
385 *goal
= ext2_find_near(inode
, partial
);
386 write_unlock(&ei
->i_meta_lock
);
389 write_unlock(&ei
->i_meta_lock
);
394 * ext2_alloc_branch - allocate and set up a chain of blocks.
396 * @num: depth of the chain (number of blocks to allocate)
397 * @offsets: offsets (in the blocks) to store the pointers to next.
398 * @branch: place to store the chain in.
400 * This function allocates @num blocks, zeroes out all but the last one,
401 * links them into chain and (if we are synchronous) writes them to disk.
402 * In other words, it prepares a branch that can be spliced onto the
403 * inode. It stores the information about that chain in the branch[], in
404 * the same format as ext2_get_branch() would do. We are calling it after
405 * we had read the existing part of chain and partial points to the last
406 * triple of that (one with zero ->key). Upon the exit we have the same
407 * picture as after the successful ext2_get_block(), excpet that in one
408 * place chain is disconnected - *branch->p is still zero (we did not
409 * set the last link), but branch->key contains the number that should
410 * be placed into *branch->p to fill that gap.
412 * If allocation fails we free all blocks we've allocated (and forget
413 * their buffer_heads) and return the error value the from failed
414 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
415 * as described above and return 0.
418 static int ext2_alloc_branch(struct inode
*inode
,
424 int blocksize
= inode
->i_sb
->s_blocksize
;
428 int parent
= ext2_alloc_block(inode
, goal
, &err
);
430 branch
[0].key
= cpu_to_le32(parent
);
431 if (parent
) for (n
= 1; n
< num
; n
++) {
432 struct buffer_head
*bh
;
433 /* Allocate the next block */
434 int nr
= ext2_alloc_block(inode
, parent
, &err
);
437 branch
[n
].key
= cpu_to_le32(nr
);
439 * Get buffer_head for parent block, zero it out and set
440 * the pointer to new one, then send parent to disk.
442 bh
= sb_getblk(inode
->i_sb
, parent
);
448 memset(bh
->b_data
, 0, blocksize
);
450 branch
[n
].p
= (__le32
*) bh
->b_data
+ offsets
[n
];
451 *branch
[n
].p
= branch
[n
].key
;
452 set_buffer_uptodate(bh
);
454 mark_buffer_dirty_inode(bh
, inode
);
455 /* We used to sync bh here if IS_SYNC(inode).
456 * But we now rely upon generic_osync_inode()
457 * and b_inode_buffers. But not for directories.
459 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
460 sync_dirty_buffer(bh
);
466 /* Allocation failed, free what we already allocated */
467 for (i
= 1; i
< n
; i
++)
468 bforget(branch
[i
].bh
);
469 for (i
= 0; i
< n
; i
++)
470 ext2_free_blocks(inode
, le32_to_cpu(branch
[i
].key
), 1);
475 * ext2_splice_branch - splice the allocated branch onto inode.
477 * @block: (logical) number of block we are adding
478 * @chain: chain of indirect blocks (with a missing link - see
480 * @where: location of missing link
481 * @num: number of blocks we are adding
483 * This function verifies that chain (up to the missing link) had not
484 * changed, fills the missing link and does all housekeeping needed in
485 * inode (->i_blocks, etc.). In case of success we end up with the full
486 * chain to new block and return 0. Otherwise (== chain had been changed)
487 * we free the new blocks (forgetting their buffer_heads, indeed) and
491 static inline int ext2_splice_branch(struct inode
*inode
,
497 struct ext2_inode_info
*ei
= EXT2_I(inode
);
500 /* Verify that place we are splicing to is still there and vacant */
502 write_lock(&ei
->i_meta_lock
);
503 if (!verify_chain(chain
, where
-1) || *where
->p
)
508 *where
->p
= where
->key
;
509 ei
->i_next_alloc_block
= block
;
510 ei
->i_next_alloc_goal
= le32_to_cpu(where
[num
-1].key
);
512 write_unlock(&ei
->i_meta_lock
);
514 /* We are done with atomic stuff, now do the rest of housekeeping */
516 inode
->i_ctime
= CURRENT_TIME_SEC
;
518 /* had we spliced it onto indirect block? */
520 mark_buffer_dirty_inode(where
->bh
, inode
);
522 mark_inode_dirty(inode
);
526 write_unlock(&ei
->i_meta_lock
);
527 for (i
= 1; i
< num
; i
++)
528 bforget(where
[i
].bh
);
529 for (i
= 0; i
< num
; i
++)
530 ext2_free_blocks(inode
, le32_to_cpu(where
[i
].key
), 1);
535 * Allocation strategy is simple: if we have to allocate something, we will
536 * have to go the whole way to leaf. So let's do it before attaching anything
537 * to tree, set linkage between the newborn blocks, write them if sync is
538 * required, recheck the path, free and repeat if check fails, otherwise
539 * set the last missing link (that will protect us from any truncate-generated
540 * removals - all blocks on the path are immune now) and possibly force the
541 * write on the parent block.
542 * That has a nice additional property: no special recovery from the failed
543 * allocations is needed - we simply release blocks and do not touch anything
544 * reachable from inode.
547 int ext2_get_block(struct inode
*inode
, sector_t iblock
, struct buffer_head
*bh_result
, int create
)
556 int depth
= ext2_block_to_path(inode
, iblock
, offsets
, &boundary
);
562 partial
= ext2_get_branch(inode
, depth
, offsets
, chain
, &err
);
564 /* Simplest case - block found, no allocation needed */
567 map_bh(bh_result
, inode
->i_sb
, le32_to_cpu(chain
[depth
-1].key
));
569 set_buffer_boundary(bh_result
);
570 /* Clean up and exit */
571 partial
= chain
+depth
-1; /* the whole chain */
575 /* Next simple case - plain lookup or failed read of indirect block */
576 if (!create
|| err
== -EIO
) {
578 while (partial
> chain
) {
587 * Indirect block might be removed by truncate while we were
588 * reading it. Handling of that case (forget what we've got and
589 * reread) is taken out of the main path.
595 if (ext2_find_goal(inode
, iblock
, chain
, partial
, &goal
) < 0)
598 left
= (chain
+ depth
) - partial
;
599 err
= ext2_alloc_branch(inode
, left
, goal
,
600 offsets
+(partial
-chain
), partial
);
604 if (ext2_use_xip(inode
->i_sb
)) {
606 * we need to clear the block
608 err
= ext2_clear_xip_target (inode
,
609 le32_to_cpu(chain
[depth
-1].key
));
614 if (ext2_splice_branch(inode
, iblock
, chain
, partial
, left
) < 0)
617 set_buffer_new(bh_result
);
621 while (partial
> chain
) {
628 static int ext2_writepage(struct page
*page
, struct writeback_control
*wbc
)
630 return block_write_full_page(page
, ext2_get_block
, wbc
);
633 static int ext2_readpage(struct file
*file
, struct page
*page
)
635 return mpage_readpage(page
, ext2_get_block
);
639 ext2_readpages(struct file
*file
, struct address_space
*mapping
,
640 struct list_head
*pages
, unsigned nr_pages
)
642 return mpage_readpages(mapping
, pages
, nr_pages
, ext2_get_block
);
646 ext2_prepare_write(struct file
*file
, struct page
*page
,
647 unsigned from
, unsigned to
)
649 return block_prepare_write(page
,from
,to
,ext2_get_block
);
653 ext2_nobh_prepare_write(struct file
*file
, struct page
*page
,
654 unsigned from
, unsigned to
)
656 return nobh_prepare_write(page
,from
,to
,ext2_get_block
);
659 static int ext2_nobh_writepage(struct page
*page
,
660 struct writeback_control
*wbc
)
662 return nobh_writepage(page
, ext2_get_block
, wbc
);
665 static sector_t
ext2_bmap(struct address_space
*mapping
, sector_t block
)
667 return generic_block_bmap(mapping
,block
,ext2_get_block
);
671 ext2_direct_IO(int rw
, struct kiocb
*iocb
, const struct iovec
*iov
,
672 loff_t offset
, unsigned long nr_segs
)
674 struct file
*file
= iocb
->ki_filp
;
675 struct inode
*inode
= file
->f_mapping
->host
;
677 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
678 offset
, nr_segs
, ext2_get_block
, NULL
);
682 ext2_writepages(struct address_space
*mapping
, struct writeback_control
*wbc
)
684 return mpage_writepages(mapping
, wbc
, ext2_get_block
);
687 const struct address_space_operations ext2_aops
= {
688 .readpage
= ext2_readpage
,
689 .readpages
= ext2_readpages
,
690 .writepage
= ext2_writepage
,
691 .sync_page
= block_sync_page
,
692 .prepare_write
= ext2_prepare_write
,
693 .commit_write
= generic_commit_write
,
695 .direct_IO
= ext2_direct_IO
,
696 .writepages
= ext2_writepages
,
697 .migratepage
= buffer_migrate_page
,
700 const struct address_space_operations ext2_aops_xip
= {
702 .get_xip_page
= ext2_get_xip_page
,
705 const struct address_space_operations ext2_nobh_aops
= {
706 .readpage
= ext2_readpage
,
707 .readpages
= ext2_readpages
,
708 .writepage
= ext2_nobh_writepage
,
709 .sync_page
= block_sync_page
,
710 .prepare_write
= ext2_nobh_prepare_write
,
711 .commit_write
= nobh_commit_write
,
713 .direct_IO
= ext2_direct_IO
,
714 .writepages
= ext2_writepages
,
715 .migratepage
= buffer_migrate_page
,
719 * Probably it should be a library function... search for first non-zero word
720 * or memcmp with zero_page, whatever is better for particular architecture.
723 static inline int all_zeroes(__le32
*p
, __le32
*q
)
732 * ext2_find_shared - find the indirect blocks for partial truncation.
733 * @inode: inode in question
734 * @depth: depth of the affected branch
735 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
736 * @chain: place to store the pointers to partial indirect blocks
737 * @top: place to the (detached) top of branch
739 * This is a helper function used by ext2_truncate().
741 * When we do truncate() we may have to clean the ends of several indirect
742 * blocks but leave the blocks themselves alive. Block is partially
743 * truncated if some data below the new i_size is refered from it (and
744 * it is on the path to the first completely truncated data block, indeed).
745 * We have to free the top of that path along with everything to the right
746 * of the path. Since no allocation past the truncation point is possible
747 * until ext2_truncate() finishes, we may safely do the latter, but top
748 * of branch may require special attention - pageout below the truncation
749 * point might try to populate it.
751 * We atomically detach the top of branch from the tree, store the block
752 * number of its root in *@top, pointers to buffer_heads of partially
753 * truncated blocks - in @chain[].bh and pointers to their last elements
754 * that should not be removed - in @chain[].p. Return value is the pointer
755 * to last filled element of @chain.
757 * The work left to caller to do the actual freeing of subtrees:
758 * a) free the subtree starting from *@top
759 * b) free the subtrees whose roots are stored in
760 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
761 * c) free the subtrees growing from the inode past the @chain[0].p
762 * (no partially truncated stuff there).
765 static Indirect
*ext2_find_shared(struct inode
*inode
,
771 Indirect
*partial
, *p
;
775 for (k
= depth
; k
> 1 && !offsets
[k
-1]; k
--)
777 partial
= ext2_get_branch(inode
, k
, offsets
, chain
, &err
);
779 partial
= chain
+ k
-1;
781 * If the branch acquired continuation since we've looked at it -
782 * fine, it should all survive and (new) top doesn't belong to us.
784 write_lock(&EXT2_I(inode
)->i_meta_lock
);
785 if (!partial
->key
&& *partial
->p
) {
786 write_unlock(&EXT2_I(inode
)->i_meta_lock
);
789 for (p
=partial
; p
>chain
&& all_zeroes((__le32
*)p
->bh
->b_data
,p
->p
); p
--)
792 * OK, we've found the last block that must survive. The rest of our
793 * branch should be detached before unlocking. However, if that rest
794 * of branch is all ours and does not grow immediately from the inode
795 * it's easier to cheat and just decrement partial->p.
797 if (p
== chain
+ k
- 1 && p
> chain
) {
803 write_unlock(&EXT2_I(inode
)->i_meta_lock
);
815 * ext2_free_data - free a list of data blocks
816 * @inode: inode we are dealing with
817 * @p: array of block numbers
818 * @q: points immediately past the end of array
820 * We are freeing all blocks refered from that array (numbers are
821 * stored as little-endian 32-bit) and updating @inode->i_blocks
824 static inline void ext2_free_data(struct inode
*inode
, __le32
*p
, __le32
*q
)
826 unsigned long block_to_free
= 0, count
= 0;
829 for ( ; p
< q
; p
++) {
830 nr
= le32_to_cpu(*p
);
833 /* accumulate blocks to free if they're contiguous */
836 else if (block_to_free
== nr
- count
)
839 mark_inode_dirty(inode
);
840 ext2_free_blocks (inode
, block_to_free
, count
);
848 mark_inode_dirty(inode
);
849 ext2_free_blocks (inode
, block_to_free
, count
);
854 * ext2_free_branches - free an array of branches
855 * @inode: inode we are dealing with
856 * @p: array of block numbers
857 * @q: pointer immediately past the end of array
858 * @depth: depth of the branches to free
860 * We are freeing all blocks refered from these branches (numbers are
861 * stored as little-endian 32-bit) and updating @inode->i_blocks
864 static void ext2_free_branches(struct inode
*inode
, __le32
*p
, __le32
*q
, int depth
)
866 struct buffer_head
* bh
;
870 int addr_per_block
= EXT2_ADDR_PER_BLOCK(inode
->i_sb
);
871 for ( ; p
< q
; p
++) {
872 nr
= le32_to_cpu(*p
);
876 bh
= sb_bread(inode
->i_sb
, nr
);
878 * A read failure? Report error and clear slot
882 ext2_error(inode
->i_sb
, "ext2_free_branches",
883 "Read failure, inode=%ld, block=%ld",
887 ext2_free_branches(inode
,
889 (__le32
*)bh
->b_data
+ addr_per_block
,
892 ext2_free_blocks(inode
, nr
, 1);
893 mark_inode_dirty(inode
);
896 ext2_free_data(inode
, p
, q
);
899 void ext2_truncate (struct inode
* inode
)
901 __le32
*i_data
= EXT2_I(inode
)->i_data
;
902 int addr_per_block
= EXT2_ADDR_PER_BLOCK(inode
->i_sb
);
911 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
912 S_ISLNK(inode
->i_mode
)))
914 if (ext2_inode_is_fast_symlink(inode
))
916 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
919 ext2_discard_prealloc(inode
);
921 blocksize
= inode
->i_sb
->s_blocksize
;
922 iblock
= (inode
->i_size
+ blocksize
-1)
923 >> EXT2_BLOCK_SIZE_BITS(inode
->i_sb
);
925 if (mapping_is_xip(inode
->i_mapping
))
926 xip_truncate_page(inode
->i_mapping
, inode
->i_size
);
927 else if (test_opt(inode
->i_sb
, NOBH
))
928 nobh_truncate_page(inode
->i_mapping
, inode
->i_size
);
930 block_truncate_page(inode
->i_mapping
,
931 inode
->i_size
, ext2_get_block
);
933 n
= ext2_block_to_path(inode
, iblock
, offsets
, NULL
);
938 ext2_free_data(inode
, i_data
+offsets
[0],
939 i_data
+ EXT2_NDIR_BLOCKS
);
943 partial
= ext2_find_shared(inode
, n
, offsets
, chain
, &nr
);
944 /* Kill the top of shared branch (already detached) */
946 if (partial
== chain
)
947 mark_inode_dirty(inode
);
949 mark_buffer_dirty_inode(partial
->bh
, inode
);
950 ext2_free_branches(inode
, &nr
, &nr
+1, (chain
+n
-1) - partial
);
952 /* Clear the ends of indirect blocks on the shared branch */
953 while (partial
> chain
) {
954 ext2_free_branches(inode
,
956 (__le32
*)partial
->bh
->b_data
+addr_per_block
,
957 (chain
+n
-1) - partial
);
958 mark_buffer_dirty_inode(partial
->bh
, inode
);
959 brelse (partial
->bh
);
963 /* Kill the remaining (whole) subtrees */
964 switch (offsets
[0]) {
966 nr
= i_data
[EXT2_IND_BLOCK
];
968 i_data
[EXT2_IND_BLOCK
] = 0;
969 mark_inode_dirty(inode
);
970 ext2_free_branches(inode
, &nr
, &nr
+1, 1);
973 nr
= i_data
[EXT2_DIND_BLOCK
];
975 i_data
[EXT2_DIND_BLOCK
] = 0;
976 mark_inode_dirty(inode
);
977 ext2_free_branches(inode
, &nr
, &nr
+1, 2);
979 case EXT2_DIND_BLOCK
:
980 nr
= i_data
[EXT2_TIND_BLOCK
];
982 i_data
[EXT2_TIND_BLOCK
] = 0;
983 mark_inode_dirty(inode
);
984 ext2_free_branches(inode
, &nr
, &nr
+1, 3);
986 case EXT2_TIND_BLOCK
:
989 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME_SEC
;
990 if (inode_needs_sync(inode
)) {
991 sync_mapping_buffers(inode
->i_mapping
);
992 ext2_sync_inode (inode
);
994 mark_inode_dirty(inode
);
998 static struct ext2_inode
*ext2_get_inode(struct super_block
*sb
, ino_t ino
,
999 struct buffer_head
**p
)
1001 struct buffer_head
* bh
;
1002 unsigned long block_group
;
1003 unsigned long block
;
1004 unsigned long offset
;
1005 struct ext2_group_desc
* gdp
;
1008 if ((ino
!= EXT2_ROOT_INO
&& ino
< EXT2_FIRST_INO(sb
)) ||
1009 ino
> le32_to_cpu(EXT2_SB(sb
)->s_es
->s_inodes_count
))
1012 block_group
= (ino
- 1) / EXT2_INODES_PER_GROUP(sb
);
1013 gdp
= ext2_get_group_desc(sb
, block_group
, &bh
);
1017 * Figure out the offset within the block group inode table
1019 offset
= ((ino
- 1) % EXT2_INODES_PER_GROUP(sb
)) * EXT2_INODE_SIZE(sb
);
1020 block
= le32_to_cpu(gdp
->bg_inode_table
) +
1021 (offset
>> EXT2_BLOCK_SIZE_BITS(sb
));
1022 if (!(bh
= sb_bread(sb
, block
)))
1026 offset
&= (EXT2_BLOCK_SIZE(sb
) - 1);
1027 return (struct ext2_inode
*) (bh
->b_data
+ offset
);
1030 ext2_error(sb
, "ext2_get_inode", "bad inode number: %lu",
1031 (unsigned long) ino
);
1032 return ERR_PTR(-EINVAL
);
1034 ext2_error(sb
, "ext2_get_inode",
1035 "unable to read inode block - inode=%lu, block=%lu",
1036 (unsigned long) ino
, block
);
1038 return ERR_PTR(-EIO
);
1041 void ext2_set_inode_flags(struct inode
*inode
)
1043 unsigned int flags
= EXT2_I(inode
)->i_flags
;
1045 inode
->i_flags
&= ~(S_SYNC
|S_APPEND
|S_IMMUTABLE
|S_NOATIME
|S_DIRSYNC
);
1046 if (flags
& EXT2_SYNC_FL
)
1047 inode
->i_flags
|= S_SYNC
;
1048 if (flags
& EXT2_APPEND_FL
)
1049 inode
->i_flags
|= S_APPEND
;
1050 if (flags
& EXT2_IMMUTABLE_FL
)
1051 inode
->i_flags
|= S_IMMUTABLE
;
1052 if (flags
& EXT2_NOATIME_FL
)
1053 inode
->i_flags
|= S_NOATIME
;
1054 if (flags
& EXT2_DIRSYNC_FL
)
1055 inode
->i_flags
|= S_DIRSYNC
;
1058 void ext2_read_inode (struct inode
* inode
)
1060 struct ext2_inode_info
*ei
= EXT2_I(inode
);
1061 ino_t ino
= inode
->i_ino
;
1062 struct buffer_head
* bh
;
1063 struct ext2_inode
* raw_inode
= ext2_get_inode(inode
->i_sb
, ino
, &bh
);
1066 #ifdef CONFIG_EXT2_FS_POSIX_ACL
1067 ei
->i_acl
= EXT2_ACL_NOT_CACHED
;
1068 ei
->i_default_acl
= EXT2_ACL_NOT_CACHED
;
1070 if (IS_ERR(raw_inode
))
1073 inode
->i_mode
= le16_to_cpu(raw_inode
->i_mode
);
1074 inode
->i_uid
= (uid_t
)le16_to_cpu(raw_inode
->i_uid_low
);
1075 inode
->i_gid
= (gid_t
)le16_to_cpu(raw_inode
->i_gid_low
);
1076 if (!(test_opt (inode
->i_sb
, NO_UID32
))) {
1077 inode
->i_uid
|= le16_to_cpu(raw_inode
->i_uid_high
) << 16;
1078 inode
->i_gid
|= le16_to_cpu(raw_inode
->i_gid_high
) << 16;
1080 inode
->i_nlink
= le16_to_cpu(raw_inode
->i_links_count
);
1081 inode
->i_size
= le32_to_cpu(raw_inode
->i_size
);
1082 inode
->i_atime
.tv_sec
= (signed)le32_to_cpu(raw_inode
->i_atime
);
1083 inode
->i_ctime
.tv_sec
= (signed)le32_to_cpu(raw_inode
->i_ctime
);
1084 inode
->i_mtime
.tv_sec
= (signed)le32_to_cpu(raw_inode
->i_mtime
);
1085 inode
->i_atime
.tv_nsec
= inode
->i_mtime
.tv_nsec
= inode
->i_ctime
.tv_nsec
= 0;
1086 ei
->i_dtime
= le32_to_cpu(raw_inode
->i_dtime
);
1087 /* We now have enough fields to check if the inode was active or not.
1088 * This is needed because nfsd might try to access dead inodes
1089 * the test is that same one that e2fsck uses
1090 * NeilBrown 1999oct15
1092 if (inode
->i_nlink
== 0 && (inode
->i_mode
== 0 || ei
->i_dtime
)) {
1093 /* this inode is deleted */
1097 inode
->i_blocks
= le32_to_cpu(raw_inode
->i_blocks
);
1098 ei
->i_flags
= le32_to_cpu(raw_inode
->i_flags
);
1099 ei
->i_faddr
= le32_to_cpu(raw_inode
->i_faddr
);
1100 ei
->i_frag_no
= raw_inode
->i_frag
;
1101 ei
->i_frag_size
= raw_inode
->i_fsize
;
1102 ei
->i_file_acl
= le32_to_cpu(raw_inode
->i_file_acl
);
1104 if (S_ISREG(inode
->i_mode
))
1105 inode
->i_size
|= ((__u64
)le32_to_cpu(raw_inode
->i_size_high
)) << 32;
1107 ei
->i_dir_acl
= le32_to_cpu(raw_inode
->i_dir_acl
);
1109 inode
->i_generation
= le32_to_cpu(raw_inode
->i_generation
);
1111 ei
->i_next_alloc_block
= 0;
1112 ei
->i_next_alloc_goal
= 0;
1113 ei
->i_prealloc_count
= 0;
1114 ei
->i_block_group
= (ino
- 1) / EXT2_INODES_PER_GROUP(inode
->i_sb
);
1115 ei
->i_dir_start_lookup
= 0;
1118 * NOTE! The in-memory inode i_data array is in little-endian order
1119 * even on big-endian machines: we do NOT byteswap the block numbers!
1121 for (n
= 0; n
< EXT2_N_BLOCKS
; n
++)
1122 ei
->i_data
[n
] = raw_inode
->i_block
[n
];
1124 if (S_ISREG(inode
->i_mode
)) {
1125 inode
->i_op
= &ext2_file_inode_operations
;
1126 if (ext2_use_xip(inode
->i_sb
)) {
1127 inode
->i_mapping
->a_ops
= &ext2_aops_xip
;
1128 inode
->i_fop
= &ext2_xip_file_operations
;
1129 } else if (test_opt(inode
->i_sb
, NOBH
)) {
1130 inode
->i_mapping
->a_ops
= &ext2_nobh_aops
;
1131 inode
->i_fop
= &ext2_file_operations
;
1133 inode
->i_mapping
->a_ops
= &ext2_aops
;
1134 inode
->i_fop
= &ext2_file_operations
;
1136 } else if (S_ISDIR(inode
->i_mode
)) {
1137 inode
->i_op
= &ext2_dir_inode_operations
;
1138 inode
->i_fop
= &ext2_dir_operations
;
1139 if (test_opt(inode
->i_sb
, NOBH
))
1140 inode
->i_mapping
->a_ops
= &ext2_nobh_aops
;
1142 inode
->i_mapping
->a_ops
= &ext2_aops
;
1143 } else if (S_ISLNK(inode
->i_mode
)) {
1144 if (ext2_inode_is_fast_symlink(inode
))
1145 inode
->i_op
= &ext2_fast_symlink_inode_operations
;
1147 inode
->i_op
= &ext2_symlink_inode_operations
;
1148 if (test_opt(inode
->i_sb
, NOBH
))
1149 inode
->i_mapping
->a_ops
= &ext2_nobh_aops
;
1151 inode
->i_mapping
->a_ops
= &ext2_aops
;
1154 inode
->i_op
= &ext2_special_inode_operations
;
1155 if (raw_inode
->i_block
[0])
1156 init_special_inode(inode
, inode
->i_mode
,
1157 old_decode_dev(le32_to_cpu(raw_inode
->i_block
[0])));
1159 init_special_inode(inode
, inode
->i_mode
,
1160 new_decode_dev(le32_to_cpu(raw_inode
->i_block
[1])));
1163 ext2_set_inode_flags(inode
);
1167 make_bad_inode(inode
);
1171 static int ext2_update_inode(struct inode
* inode
, int do_sync
)
1173 struct ext2_inode_info
*ei
= EXT2_I(inode
);
1174 struct super_block
*sb
= inode
->i_sb
;
1175 ino_t ino
= inode
->i_ino
;
1176 uid_t uid
= inode
->i_uid
;
1177 gid_t gid
= inode
->i_gid
;
1178 struct buffer_head
* bh
;
1179 struct ext2_inode
* raw_inode
= ext2_get_inode(sb
, ino
, &bh
);
1183 if (IS_ERR(raw_inode
))
1186 /* For fields not not tracking in the in-memory inode,
1187 * initialise them to zero for new inodes. */
1188 if (ei
->i_state
& EXT2_STATE_NEW
)
1189 memset(raw_inode
, 0, EXT2_SB(sb
)->s_inode_size
);
1191 raw_inode
->i_mode
= cpu_to_le16(inode
->i_mode
);
1192 if (!(test_opt(sb
, NO_UID32
))) {
1193 raw_inode
->i_uid_low
= cpu_to_le16(low_16_bits(uid
));
1194 raw_inode
->i_gid_low
= cpu_to_le16(low_16_bits(gid
));
1196 * Fix up interoperability with old kernels. Otherwise, old inodes get
1197 * re-used with the upper 16 bits of the uid/gid intact
1200 raw_inode
->i_uid_high
= cpu_to_le16(high_16_bits(uid
));
1201 raw_inode
->i_gid_high
= cpu_to_le16(high_16_bits(gid
));
1203 raw_inode
->i_uid_high
= 0;
1204 raw_inode
->i_gid_high
= 0;
1207 raw_inode
->i_uid_low
= cpu_to_le16(fs_high2lowuid(uid
));
1208 raw_inode
->i_gid_low
= cpu_to_le16(fs_high2lowgid(gid
));
1209 raw_inode
->i_uid_high
= 0;
1210 raw_inode
->i_gid_high
= 0;
1212 raw_inode
->i_links_count
= cpu_to_le16(inode
->i_nlink
);
1213 raw_inode
->i_size
= cpu_to_le32(inode
->i_size
);
1214 raw_inode
->i_atime
= cpu_to_le32(inode
->i_atime
.tv_sec
);
1215 raw_inode
->i_ctime
= cpu_to_le32(inode
->i_ctime
.tv_sec
);
1216 raw_inode
->i_mtime
= cpu_to_le32(inode
->i_mtime
.tv_sec
);
1218 raw_inode
->i_blocks
= cpu_to_le32(inode
->i_blocks
);
1219 raw_inode
->i_dtime
= cpu_to_le32(ei
->i_dtime
);
1220 raw_inode
->i_flags
= cpu_to_le32(ei
->i_flags
);
1221 raw_inode
->i_faddr
= cpu_to_le32(ei
->i_faddr
);
1222 raw_inode
->i_frag
= ei
->i_frag_no
;
1223 raw_inode
->i_fsize
= ei
->i_frag_size
;
1224 raw_inode
->i_file_acl
= cpu_to_le32(ei
->i_file_acl
);
1225 if (!S_ISREG(inode
->i_mode
))
1226 raw_inode
->i_dir_acl
= cpu_to_le32(ei
->i_dir_acl
);
1228 raw_inode
->i_size_high
= cpu_to_le32(inode
->i_size
>> 32);
1229 if (inode
->i_size
> 0x7fffffffULL
) {
1230 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb
,
1231 EXT2_FEATURE_RO_COMPAT_LARGE_FILE
) ||
1232 EXT2_SB(sb
)->s_es
->s_rev_level
==
1233 cpu_to_le32(EXT2_GOOD_OLD_REV
)) {
1234 /* If this is the first large file
1235 * created, add a flag to the superblock.
1238 ext2_update_dynamic_rev(sb
);
1239 EXT2_SET_RO_COMPAT_FEATURE(sb
,
1240 EXT2_FEATURE_RO_COMPAT_LARGE_FILE
);
1242 ext2_write_super(sb
);
1247 raw_inode
->i_generation
= cpu_to_le32(inode
->i_generation
);
1248 if (S_ISCHR(inode
->i_mode
) || S_ISBLK(inode
->i_mode
)) {
1249 if (old_valid_dev(inode
->i_rdev
)) {
1250 raw_inode
->i_block
[0] =
1251 cpu_to_le32(old_encode_dev(inode
->i_rdev
));
1252 raw_inode
->i_block
[1] = 0;
1254 raw_inode
->i_block
[0] = 0;
1255 raw_inode
->i_block
[1] =
1256 cpu_to_le32(new_encode_dev(inode
->i_rdev
));
1257 raw_inode
->i_block
[2] = 0;
1259 } else for (n
= 0; n
< EXT2_N_BLOCKS
; n
++)
1260 raw_inode
->i_block
[n
] = ei
->i_data
[n
];
1261 mark_buffer_dirty(bh
);
1263 sync_dirty_buffer(bh
);
1264 if (buffer_req(bh
) && !buffer_uptodate(bh
)) {
1265 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1266 sb
->s_id
, (unsigned long) ino
);
1270 ei
->i_state
&= ~EXT2_STATE_NEW
;
1275 int ext2_write_inode(struct inode
*inode
, int wait
)
1277 return ext2_update_inode(inode
, wait
);
1280 int ext2_sync_inode(struct inode
*inode
)
1282 struct writeback_control wbc
= {
1283 .sync_mode
= WB_SYNC_ALL
,
1284 .nr_to_write
= 0, /* sys_fsync did this */
1286 return sync_inode(inode
, &wbc
);
1289 int ext2_setattr(struct dentry
*dentry
, struct iattr
*iattr
)
1291 struct inode
*inode
= dentry
->d_inode
;
1294 error
= inode_change_ok(inode
, iattr
);
1297 if ((iattr
->ia_valid
& ATTR_UID
&& iattr
->ia_uid
!= inode
->i_uid
) ||
1298 (iattr
->ia_valid
& ATTR_GID
&& iattr
->ia_gid
!= inode
->i_gid
)) {
1299 error
= DQUOT_TRANSFER(inode
, iattr
) ? -EDQUOT
: 0;
1303 error
= inode_setattr(inode
, iattr
);
1304 if (!error
&& (iattr
->ia_valid
& ATTR_MODE
))
1305 error
= ext2_acl_chmod(inode
);