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
26 #include <linux/locks.h>
27 #include <linux/smp_lock.h>
28 #include <linux/sched.h>
29 #include <linux/highuid.h>
31 static int ext2_update_inode(struct inode
* inode
, int do_sync
);
34 * Called at each iput()
36 void ext2_put_inode (struct inode
* inode
)
38 ext2_discard_prealloc (inode
);
42 * Called at the last iput() if i_nlink is zero.
44 void ext2_delete_inode (struct inode
* inode
)
48 if (is_bad_inode(inode
) ||
49 inode
->i_ino
== EXT2_ACL_IDX_INO
||
50 inode
->i_ino
== EXT2_ACL_DATA_INO
)
52 inode
->u
.ext2_i
.i_dtime
= CURRENT_TIME
;
53 mark_inode_dirty(inode
);
54 ext2_update_inode(inode
, IS_SYNC(inode
));
57 ext2_truncate (inode
);
58 ext2_free_inode (inode
);
64 clear_inode(inode
); /* We must guarantee clearing of inode... */
67 void ext2_discard_prealloc (struct inode
* inode
)
69 #ifdef EXT2_PREALLOCATE
71 /* Writer: ->i_prealloc* */
72 if (inode
->u
.ext2_i
.i_prealloc_count
) {
73 unsigned short total
= inode
->u
.ext2_i
.i_prealloc_count
;
74 unsigned long block
= inode
->u
.ext2_i
.i_prealloc_block
;
75 inode
->u
.ext2_i
.i_prealloc_count
= 0;
76 inode
->u
.ext2_i
.i_prealloc_block
= 0;
78 ext2_free_blocks (inode
, block
, total
);
84 static int ext2_alloc_block (struct inode
* inode
, unsigned long goal
, int *err
)
87 static unsigned long alloc_hits
= 0, alloc_attempts
= 0;
92 #ifdef EXT2_PREALLOCATE
93 /* Writer: ->i_prealloc* */
94 if (inode
->u
.ext2_i
.i_prealloc_count
&&
95 (goal
== inode
->u
.ext2_i
.i_prealloc_block
||
96 goal
+ 1 == inode
->u
.ext2_i
.i_prealloc_block
))
98 result
= inode
->u
.ext2_i
.i_prealloc_block
++;
99 inode
->u
.ext2_i
.i_prealloc_count
--;
102 ext2_debug ("preallocation hit (%lu/%lu).\n",
103 ++alloc_hits
, ++alloc_attempts
);
106 ext2_discard_prealloc (inode
);
108 ext2_debug ("preallocation miss (%lu/%lu).\n",
109 alloc_hits
, ++alloc_attempts
);
111 if (S_ISREG(inode
->i_mode
))
112 result
= ext2_new_block (inode
, goal
,
113 &inode
->u
.ext2_i
.i_prealloc_count
,
114 &inode
->u
.ext2_i
.i_prealloc_block
, err
);
116 result
= ext2_new_block (inode
, goal
, 0, 0, err
);
119 result
= ext2_new_block (inode
, goal
, 0, 0, err
);
127 struct buffer_head
*bh
;
130 static inline void add_chain(Indirect
*p
, struct buffer_head
*bh
, u32
*v
)
132 p
->key
= *(p
->p
= v
);
136 static inline int verify_chain(Indirect
*from
, Indirect
*to
)
138 while (from
<= to
&& from
->key
== *from
->p
)
144 * ext2_block_to_path - parse the block number into array of offsets
145 * @inode: inode in question (we are only interested in its superblock)
146 * @i_block: block number to be parsed
147 * @offsets: array to store the offsets in
149 * To store the locations of file's data ext2 uses a data structure common
150 * for UNIX filesystems - tree of pointers anchored in the inode, with
151 * data blocks at leaves and indirect blocks in intermediate nodes.
152 * This function translates the block number into path in that tree -
153 * return value is the path length and @offsets[n] is the offset of
154 * pointer to (n+1)th node in the nth one. If @block is out of range
155 * (negative or too large) warning is printed and zero returned.
157 * Note: function doesn't find node addresses, so no IO is needed. All
158 * we need to know is the capacity of indirect blocks (taken from the
163 * Portability note: the last comparison (check that we fit into triple
164 * indirect block) is spelled differently, because otherwise on an
165 * architecture with 32-bit longs and 8Kb pages we might get into trouble
166 * if our filesystem had 8Kb blocks. We might use long long, but that would
167 * kill us on x86. Oh, well, at least the sign propagation does not matter -
168 * i_block would have to be negative in the very beginning, so we would not
172 static int ext2_block_to_path(struct inode
*inode
, long i_block
, int offsets
[4])
174 int ptrs
= EXT2_ADDR_PER_BLOCK(inode
->i_sb
);
175 int ptrs_bits
= EXT2_ADDR_PER_BLOCK_BITS(inode
->i_sb
);
176 const long direct_blocks
= EXT2_NDIR_BLOCKS
,
177 indirect_blocks
= ptrs
,
178 double_blocks
= (1 << (ptrs_bits
* 2));
182 ext2_warning (inode
->i_sb
, "ext2_block_to_path", "block < 0");
183 } else if (i_block
< direct_blocks
) {
184 offsets
[n
++] = i_block
;
185 } else if ( (i_block
-= direct_blocks
) < indirect_blocks
) {
186 offsets
[n
++] = EXT2_IND_BLOCK
;
187 offsets
[n
++] = i_block
;
188 } else if ((i_block
-= indirect_blocks
) < double_blocks
) {
189 offsets
[n
++] = EXT2_DIND_BLOCK
;
190 offsets
[n
++] = i_block
>> ptrs_bits
;
191 offsets
[n
++] = i_block
& (ptrs
- 1);
192 } else if (((i_block
-= double_blocks
) >> (ptrs_bits
* 2)) < ptrs
) {
193 offsets
[n
++] = EXT2_TIND_BLOCK
;
194 offsets
[n
++] = i_block
>> (ptrs_bits
* 2);
195 offsets
[n
++] = (i_block
>> ptrs_bits
) & (ptrs
- 1);
196 offsets
[n
++] = i_block
& (ptrs
- 1);
198 ext2_warning (inode
->i_sb
, "ext2_block_to_path", "block > big");
204 * ext2_get_branch - read the chain of indirect blocks leading to data
205 * @inode: inode in question
206 * @depth: depth of the chain (1 - direct pointer, etc.)
207 * @offsets: offsets of pointers in inode/indirect blocks
208 * @chain: place to store the result
209 * @err: here we store the error value
211 * Function fills the array of triples <key, p, bh> and returns %NULL
212 * if everything went OK or the pointer to the last filled triple
213 * (incomplete one) otherwise. Upon the return chain[i].key contains
214 * the number of (i+1)-th block in the chain (as it is stored in memory,
215 * i.e. little-endian 32-bit), chain[i].p contains the address of that
216 * number (it points into struct inode for i==0 and into the bh->b_data
217 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect
218 * block for i>0 and NULL for i==0. In other words, it holds the block
219 * numbers of the chain, addresses they were taken from (and where we can
220 * verify that chain did not change) and buffer_heads hosting these
223 * Function stops when it stumbles upon zero pointer (absent block)
224 * (pointer to last triple returned, *@err == 0)
225 * or when it gets an IO error reading an indirect block
226 * (ditto, *@err == -EIO)
227 * or when it notices that chain had been changed while it was reading
228 * (ditto, *@err == -EAGAIN)
229 * or when it reads all @depth-1 indirect blocks successfully and finds
230 * the whole chain, all way to the data (returns %NULL, *err == 0).
232 static inline Indirect
*ext2_get_branch(struct inode
*inode
,
238 kdev_t dev
= inode
->i_dev
;
239 int size
= inode
->i_sb
->s_blocksize
;
241 struct buffer_head
*bh
;
244 /* i_data is not going away, no lock needed */
245 add_chain (chain
, NULL
, inode
->u
.ext2_i
.i_data
+ *offsets
);
249 bh
= bread(dev
, le32_to_cpu(p
->key
), size
);
252 /* Reader: pointers */
253 if (!verify_chain(chain
, p
))
255 add_chain(++p
, bh
, (u32
*)bh
->b_data
+ *++offsets
);
272 * ext2_find_near - find a place for allocation with sufficient locality
274 * @ind: descriptor of indirect block.
276 * This function returns the prefered place for block allocation.
277 * It is used when heuristic for sequential allocation fails.
279 * + if there is a block to the left of our position - allocate near it.
280 * + if pointer will live in indirect block - allocate near that block.
281 * + if pointer will live in inode - allocate in the same cylinder group.
282 * Caller must make sure that @ind is valid and will stay that way.
285 static inline unsigned long ext2_find_near(struct inode
*inode
, Indirect
*ind
)
287 u32
*start
= ind
->bh
? (u32
*) ind
->bh
->b_data
: inode
->u
.ext2_i
.i_data
;
290 /* Try to find previous block */
291 for (p
= ind
->p
- 1; p
>= start
; p
--)
293 return le32_to_cpu(*p
);
295 /* No such thing, so let's try location of indirect block */
297 return ind
->bh
->b_blocknr
;
300 * It is going to be refered from inode itself? OK, just put it into
301 * the same cylinder group then.
303 return (inode
->u
.ext2_i
.i_block_group
*
304 EXT2_BLOCKS_PER_GROUP(inode
->i_sb
)) +
305 le32_to_cpu(inode
->i_sb
->u
.ext2_sb
.s_es
->s_first_data_block
);
309 * ext2_find_goal - find a prefered place for allocation.
311 * @block: block we want
312 * @chain: chain of indirect blocks
313 * @partial: pointer to the last triple within a chain
314 * @goal: place to store the result.
316 * Normally this function find the prefered place for block allocation,
317 * stores it in *@goal and returns zero. If the branch had been changed
318 * under us we return -EAGAIN.
321 static inline int ext2_find_goal(struct inode
*inode
,
327 /* Writer: ->i_next_alloc* */
328 if (block
== inode
->u
.ext2_i
.i_next_alloc_block
+ 1) {
329 inode
->u
.ext2_i
.i_next_alloc_block
++;
330 inode
->u
.ext2_i
.i_next_alloc_goal
++;
333 /* Reader: pointers, ->i_next_alloc* */
334 if (verify_chain(chain
, partial
)) {
336 * try the heuristic for sequential allocation,
337 * failing that at least try to get decent locality.
339 if (block
== inode
->u
.ext2_i
.i_next_alloc_block
)
340 *goal
= inode
->u
.ext2_i
.i_next_alloc_goal
;
342 *goal
= ext2_find_near(inode
, partial
);
350 * ext2_alloc_branch - allocate and set up a chain of blocks.
352 * @num: depth of the chain (number of blocks to allocate)
353 * @offsets: offsets (in the blocks) to store the pointers to next.
354 * @branch: place to store the chain in.
356 * This function allocates @num blocks, zeroes out all but the last one,
357 * links them into chain and (if we are synchronous) writes them to disk.
358 * In other words, it prepares a branch that can be spliced onto the
359 * inode. It stores the information about that chain in the branch[], in
360 * the same format as ext2_get_branch() would do. We are calling it after
361 * we had read the existing part of chain and partial points to the last
362 * triple of that (one with zero ->key). Upon the exit we have the same
363 * picture as after the successful ext2_get_block(), excpet that in one
364 * place chain is disconnected - *branch->p is still zero (we did not
365 * set the last link), but branch->key contains the number that should
366 * be placed into *branch->p to fill that gap.
368 * If allocation fails we free all blocks we've allocated (and forget
369 * ther buffer_heads) and return the error value the from failed
370 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
371 * as described above and return 0.
374 static int ext2_alloc_branch(struct inode
*inode
,
380 int blocksize
= inode
->i_sb
->s_blocksize
;
384 int parent
= ext2_alloc_block(inode
, goal
, &err
);
386 branch
[0].key
= cpu_to_le32(parent
);
387 if (parent
) for (n
= 1; n
< num
; n
++) {
388 struct buffer_head
*bh
;
389 /* Allocate the next block */
390 int nr
= ext2_alloc_block(inode
, parent
, &err
);
393 branch
[n
].key
= cpu_to_le32(nr
);
395 * Get buffer_head for parent block, zero it out and set
396 * the pointer to new one, then send parent to disk.
398 bh
= getblk(inode
->i_dev
, parent
, blocksize
);
399 if (!buffer_uptodate(bh
))
401 memset(bh
->b_data
, 0, blocksize
);
403 branch
[n
].p
= (u32
*) bh
->b_data
+ offsets
[n
];
404 *branch
[n
].p
= branch
[n
].key
;
405 mark_buffer_uptodate(bh
, 1);
406 mark_buffer_dirty(bh
);
407 if (IS_SYNC(inode
) || inode
->u
.ext2_i
.i_osync
) {
408 ll_rw_block (WRITE
, 1, &bh
);
416 /* Allocation failed, free what we already allocated */
417 for (i
= 1; i
< n
; i
++)
418 bforget(branch
[i
].bh
);
419 for (i
= 0; i
< n
; i
++)
420 ext2_free_blocks(inode
, le32_to_cpu(branch
[i
].key
), 1);
425 * ext2_splice_branch - splice the allocated branch onto inode.
427 * @block: (logical) number of block we are adding
428 * @chain: chain of indirect blocks (with a missing link - see
430 * @where: location of missing link
431 * @num: number of blocks we are adding
433 * This function verifies that chain (up to the missing link) had not
434 * changed, fills the missing link and does all housekeeping needed in
435 * inode (->i_blocks, etc.). In case of success we end up with the full
436 * chain to new block and return 0. Otherwise (== chain had been changed)
437 * we free the new blocks (forgetting their buffer_heads, indeed) and
441 static inline int ext2_splice_branch(struct inode
*inode
,
449 /* Verify that place we are splicing to is still there and vacant */
451 /* Writer: pointers, ->i_next_alloc*, ->i_blocks */
452 if (!verify_chain(chain
, where
-1) || *where
->p
)
458 *where
->p
= where
->key
;
459 inode
->u
.ext2_i
.i_next_alloc_block
= block
;
460 inode
->u
.ext2_i
.i_next_alloc_goal
= le32_to_cpu(where
[num
-1].key
);
461 inode
->i_blocks
+= num
* inode
->i_sb
->s_blocksize
/512;
465 /* We are done with atomic stuff, now do the rest of housekeeping */
467 inode
->i_ctime
= CURRENT_TIME
;
469 /* had we spliced it onto indirect block? */
471 mark_buffer_dirty(where
->bh
);
472 if (IS_SYNC(inode
) || inode
->u
.ext2_i
.i_osync
) {
473 ll_rw_block (WRITE
, 1, &where
->bh
);
474 wait_on_buffer(where
->bh
);
478 if (IS_SYNC(inode
) || inode
->u
.ext2_i
.i_osync
)
479 ext2_sync_inode (inode
);
481 mark_inode_dirty(inode
);
485 for (i
= 1; i
< num
; i
++)
486 bforget(where
[i
].bh
);
487 for (i
= 0; i
< num
; i
++)
488 ext2_free_blocks(inode
, le32_to_cpu(where
[i
].key
), 1);
493 * Allocation strategy is simple: if we have to allocate something, we will
494 * have to go the whole way to leaf. So let's do it before attaching anything
495 * to tree, set linkage between the newborn blocks, write them if sync is
496 * required, recheck the path, free and repeat if check fails, otherwise
497 * set the last missing link (that will protect us from any truncate-generated
498 * removals - all blocks on the path are immune now) and possibly force the
499 * write on the parent block.
500 * That has a nice additional property: no special recovery from the failed
501 * allocations is needed - we simply release blocks and do not touch anything
502 * reachable from inode.
505 static int ext2_get_block(struct inode
*inode
, long iblock
, struct buffer_head
*bh_result
, int create
)
513 int depth
= ext2_block_to_path(inode
, iblock
, offsets
);
520 partial
= ext2_get_branch(inode
, depth
, offsets
, chain
, &err
);
522 /* Simplest case - block found, no allocation needed */
525 bh_result
->b_dev
= inode
->i_dev
;
526 bh_result
->b_blocknr
= le32_to_cpu(chain
[depth
-1].key
);
527 bh_result
->b_state
|= (1UL << BH_Mapped
);
528 /* Clean up and exit */
529 partial
= chain
+depth
-1; /* the whole chain */
533 /* Next simple case - plain lookup or failed read of indirect block */
534 if (!create
|| err
== -EIO
) {
536 while (partial
> chain
) {
546 * Indirect block might be removed by truncate while we were
547 * reading it. Handling of that case (forget what we've got and
548 * reread) is taken out of the main path.
553 if (ext2_find_goal(inode
, iblock
, chain
, partial
, &goal
) < 0)
556 left
= (chain
+ depth
) - partial
;
557 err
= ext2_alloc_branch(inode
, left
, goal
,
558 offsets
+(partial
-chain
), partial
);
562 if (ext2_splice_branch(inode
, iblock
, chain
, partial
, left
) < 0)
565 bh_result
->b_state
|= (1UL << BH_New
);
569 while (partial
> chain
) {
570 bforget(partial
->bh
);
576 struct buffer_head
* ext2_getblk(struct inode
* inode
, long block
, int create
, int * err
)
578 struct buffer_head dummy
;
582 dummy
.b_blocknr
= -1000;
583 error
= ext2_get_block(inode
, block
, &dummy
, create
);
585 if (!error
&& buffer_mapped(&dummy
)) {
586 struct buffer_head
*bh
;
587 bh
= getblk(dummy
.b_dev
, dummy
.b_blocknr
, inode
->i_sb
->s_blocksize
);
588 if (buffer_new(&dummy
)) {
589 if (!buffer_uptodate(bh
))
591 memset(bh
->b_data
, 0, inode
->i_sb
->s_blocksize
);
592 mark_buffer_uptodate(bh
, 1);
593 mark_buffer_dirty(bh
);
600 struct buffer_head
* ext2_bread (struct inode
* inode
, int block
,
601 int create
, int *err
)
603 struct buffer_head
* bh
;
606 prev_blocks
= inode
->i_blocks
;
608 bh
= ext2_getblk (inode
, block
, create
, err
);
613 * If the inode has grown, and this is a directory, then perform
614 * preallocation of a few more blocks to try to keep directory
615 * fragmentation down.
618 S_ISDIR(inode
->i_mode
) &&
619 inode
->i_blocks
> prev_blocks
&&
620 EXT2_HAS_COMPAT_FEATURE(inode
->i_sb
,
621 EXT2_FEATURE_COMPAT_DIR_PREALLOC
)) {
623 struct buffer_head
*tmp_bh
;
626 i
< EXT2_SB(inode
->i_sb
)->s_es
->s_prealloc_dir_blocks
;
629 * ext2_getblk will zero out the contents of the
632 tmp_bh
= ext2_getblk(inode
, block
+i
, create
, err
);
641 if (buffer_uptodate(bh
))
643 ll_rw_block (READ
, 1, &bh
);
645 if (buffer_uptodate(bh
))
652 static int ext2_writepage(struct file
*file
, struct page
*page
)
654 return block_write_full_page(page
,ext2_get_block
);
656 static int ext2_readpage(struct file
*file
, struct page
*page
)
658 return block_read_full_page(page
,ext2_get_block
);
660 static int ext2_prepare_write(struct file
*file
, struct page
*page
, unsigned from
, unsigned to
)
662 return block_prepare_write(page
,from
,to
,ext2_get_block
);
664 static int ext2_bmap(struct address_space
*mapping
, long block
)
666 return generic_block_bmap(mapping
,block
,ext2_get_block
);
668 struct address_space_operations ext2_aops
= {
669 readpage
: ext2_readpage
,
670 writepage
: ext2_writepage
,
671 sync_page
: block_sync_page
,
672 prepare_write
: ext2_prepare_write
,
673 commit_write
: generic_commit_write
,
678 * Probably it should be a library function... search for first non-zero word
679 * or memcmp with zero_page, whatever is better for particular architecture.
682 static inline int all_zeroes(u32
*p
, u32
*q
)
691 * ext2_find_shared - find the indirect blocks for partial truncation.
692 * @inode: inode in question
693 * @depth: depth of the affected branch
694 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
695 * @chain: place to store the pointers to partial indirect blocks
696 * @top: place to the (detached) top of branch
698 * This is a helper function used by ext2_truncate().
700 * When we do truncate() we may have to clean the ends of several indirect
701 * blocks but leave the blocks themselves alive. Block is partially
702 * truncated if some data below the new i_size is refered from it (and
703 * it is on the path to the first completely truncated data block, indeed).
704 * We have to free the top of that path along with everything to the right
705 * of the path. Since no allocation past the truncation point is possible
706 * until ext2_truncate() finishes, we may safely do the latter, but top
707 * of branch may require special attention - pageout below the truncation
708 * point might try to populate it.
710 * We atomically detach the top of branch from the tree, store the block
711 * number of its root in *@top, pointers to buffer_heads of partially
712 * truncated blocks - in @chain[].bh and pointers to their last elements
713 * that should not be removed - in @chain[].p. Return value is the pointer
714 * to last filled element of @chain.
716 * The work left to caller to do the actual freeing of subtrees:
717 * a) free the subtree starting from *@top
718 * b) free the subtrees whose roots are stored in
719 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
720 * c) free the subtrees growing from the inode past the @chain[0].p
721 * (no partially truncated stuff there).
724 static Indirect
*ext2_find_shared(struct inode
*inode
,
730 Indirect
*partial
, *p
;
734 for (k
= depth
; k
> 1 && !offsets
[k
-1]; k
--)
736 partial
= ext2_get_branch(inode
, k
, offsets
, chain
, &err
);
737 /* Writer: pointers */
739 partial
= chain
+ k
-1;
741 * If the branch acquired continuation since we've looked at it -
742 * fine, it should all survive and (new) top doesn't belong to us.
744 if (!partial
->key
&& *partial
->p
)
747 for (p
=partial
; p
>chain
&& all_zeroes((u32
*)p
->bh
->b_data
,p
->p
); p
--)
750 * OK, we've found the last block that must survive. The rest of our
751 * branch should be detached before unlocking. However, if that rest
752 * of branch is all ours and does not grow immediately from the inode
753 * it's easier to cheat and just decrement partial->p.
755 if (p
== chain
+ k
- 1 && p
> chain
) {
773 * ext2_free_data - free a list of data blocks
774 * @inode: inode we are dealing with
775 * @p: array of block numbers
776 * @q: points immediately past the end of array
778 * We are freeing all blocks refered from that array (numbers are
779 * stored as little-endian 32-bit) and updating @inode->i_blocks
782 static inline void ext2_free_data(struct inode
*inode
, u32
*p
, u32
*q
)
784 int blocks
= inode
->i_sb
->s_blocksize
/ 512;
785 unsigned long block_to_free
= 0, count
= 0;
788 for ( ; p
< q
; p
++) {
789 nr
= le32_to_cpu(*p
);
792 /* accumulate blocks to free if they're contiguous */
795 else if (block_to_free
== nr
- count
)
798 /* Writer: ->i_blocks */
799 inode
->i_blocks
-= blocks
* count
;
801 ext2_free_blocks (inode
, block_to_free
, count
);
802 mark_inode_dirty(inode
);
810 /* Writer: ->i_blocks */
811 inode
->i_blocks
-= blocks
* count
;
813 ext2_free_blocks (inode
, block_to_free
, count
);
814 mark_inode_dirty(inode
);
819 * ext2_free_branches - free an array of branches
820 * @inode: inode we are dealing with
821 * @p: array of block numbers
822 * @q: pointer immediately past the end of array
823 * @depth: depth of the branches to free
825 * We are freeing all blocks refered from these branches (numbers are
826 * stored as little-endian 32-bit) and updating @inode->i_blocks
829 static void ext2_free_branches(struct inode
*inode
, u32
*p
, u32
*q
, int depth
)
831 struct buffer_head
* bh
;
835 int addr_per_block
= EXT2_ADDR_PER_BLOCK(inode
->i_sb
);
836 for ( ; p
< q
; p
++) {
837 nr
= le32_to_cpu(*p
);
841 bh
= bread (inode
->i_dev
, nr
, inode
->i_sb
->s_blocksize
);
843 * A read failure? Report error and clear slot
847 ext2_error(inode
->i_sb
, "ext2_free_branches",
848 "Read failure, inode=%ld, block=%ld",
852 ext2_free_branches(inode
,
854 (u32
*)bh
->b_data
+ addr_per_block
,
857 /* Writer: ->i_blocks */
858 inode
->i_blocks
-= inode
->i_sb
->s_blocksize
/ 512;
860 ext2_free_blocks(inode
, nr
, 1);
861 mark_inode_dirty(inode
);
864 ext2_free_data(inode
, p
, q
);
867 void ext2_truncate (struct inode
* inode
)
869 u32
*i_data
= inode
->u
.ext2_i
.i_data
;
870 int addr_per_block
= EXT2_ADDR_PER_BLOCK(inode
->i_sb
);
877 unsigned blocksize
, tail
;
879 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
880 S_ISLNK(inode
->i_mode
)))
882 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
885 ext2_discard_prealloc(inode
);
887 blocksize
= inode
->i_sb
->s_blocksize
;
888 iblock
= (inode
->i_size
+ blocksize
-1)
889 >> EXT2_BLOCK_SIZE_BITS(inode
->i_sb
);
890 tail
= (iblock
<< EXT2_BLOCK_SIZE_BITS(inode
->i_sb
)) - inode
->i_size
;
892 block_zero_page(inode
->i_mapping
, inode
->i_size
, tail
);
894 n
= ext2_block_to_path(inode
, iblock
, offsets
);
899 ext2_free_data(inode
, i_data
+offsets
[0],
900 i_data
+ EXT2_NDIR_BLOCKS
);
904 partial
= ext2_find_shared(inode
, n
, offsets
, chain
, &nr
);
905 /* Kill the top of shared branch (already detached) */
907 if (partial
== chain
)
908 mark_inode_dirty(inode
);
910 mark_buffer_dirty(partial
->bh
);
911 ext2_free_branches(inode
, &nr
, &nr
+1, (chain
+n
-1) - partial
);
913 /* Clear the ends of indirect blocks on the shared branch */
914 while (partial
> chain
) {
915 ext2_free_branches(inode
,
917 (u32
*)partial
->bh
->b_data
+ addr_per_block
,
918 (chain
+n
-1) - partial
);
919 mark_buffer_dirty(partial
->bh
);
920 if (IS_SYNC(inode
)) {
921 ll_rw_block (WRITE
, 1, &partial
->bh
);
922 wait_on_buffer (partial
->bh
);
924 brelse (partial
->bh
);
928 /* Kill the remaining (whole) subtrees */
929 switch (offsets
[0]) {
931 nr
= i_data
[EXT2_IND_BLOCK
];
933 i_data
[EXT2_IND_BLOCK
] = 0;
934 mark_inode_dirty(inode
);
935 ext2_free_branches(inode
, &nr
, &nr
+1, 1);
938 nr
= i_data
[EXT2_DIND_BLOCK
];
940 i_data
[EXT2_DIND_BLOCK
] = 0;
941 mark_inode_dirty(inode
);
942 ext2_free_branches(inode
, &nr
, &nr
+1, 2);
944 case EXT2_DIND_BLOCK
:
945 nr
= i_data
[EXT2_TIND_BLOCK
];
947 i_data
[EXT2_TIND_BLOCK
] = 0;
948 mark_inode_dirty(inode
);
949 ext2_free_branches(inode
, &nr
, &nr
+1, 3);
951 case EXT2_TIND_BLOCK
:
954 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
956 ext2_sync_inode (inode
);
958 mark_inode_dirty(inode
);
961 void ext2_read_inode (struct inode
* inode
)
963 struct buffer_head
* bh
;
964 struct ext2_inode
* raw_inode
;
965 unsigned long block_group
;
966 unsigned long group_desc
;
969 unsigned long offset
;
970 struct ext2_group_desc
* gdp
;
972 if ((inode
->i_ino
!= EXT2_ROOT_INO
&& inode
->i_ino
!= EXT2_ACL_IDX_INO
&&
973 inode
->i_ino
!= EXT2_ACL_DATA_INO
&&
974 inode
->i_ino
< EXT2_FIRST_INO(inode
->i_sb
)) ||
975 inode
->i_ino
> le32_to_cpu(inode
->i_sb
->u
.ext2_sb
.s_es
->s_inodes_count
)) {
976 ext2_error (inode
->i_sb
, "ext2_read_inode",
977 "bad inode number: %lu", inode
->i_ino
);
980 block_group
= (inode
->i_ino
- 1) / EXT2_INODES_PER_GROUP(inode
->i_sb
);
981 if (block_group
>= inode
->i_sb
->u
.ext2_sb
.s_groups_count
) {
982 ext2_error (inode
->i_sb
, "ext2_read_inode",
983 "group >= groups count");
986 group_desc
= block_group
>> EXT2_DESC_PER_BLOCK_BITS(inode
->i_sb
);
987 desc
= block_group
& (EXT2_DESC_PER_BLOCK(inode
->i_sb
) - 1);
988 bh
= inode
->i_sb
->u
.ext2_sb
.s_group_desc
[group_desc
];
990 ext2_error (inode
->i_sb
, "ext2_read_inode",
991 "Descriptor not loaded");
995 gdp
= (struct ext2_group_desc
*) bh
->b_data
;
997 * Figure out the offset within the block group inode table
999 offset
= ((inode
->i_ino
- 1) % EXT2_INODES_PER_GROUP(inode
->i_sb
)) *
1000 EXT2_INODE_SIZE(inode
->i_sb
);
1001 block
= le32_to_cpu(gdp
[desc
].bg_inode_table
) +
1002 (offset
>> EXT2_BLOCK_SIZE_BITS(inode
->i_sb
));
1003 if (!(bh
= bread (inode
->i_dev
, block
, inode
->i_sb
->s_blocksize
))) {
1004 ext2_error (inode
->i_sb
, "ext2_read_inode",
1005 "unable to read inode block - "
1006 "inode=%lu, block=%lu", inode
->i_ino
, block
);
1009 offset
&= (EXT2_BLOCK_SIZE(inode
->i_sb
) - 1);
1010 raw_inode
= (struct ext2_inode
*) (bh
->b_data
+ offset
);
1012 inode
->i_mode
= le16_to_cpu(raw_inode
->i_mode
);
1013 inode
->i_uid
= (uid_t
)le16_to_cpu(raw_inode
->i_uid_low
);
1014 inode
->i_gid
= (gid_t
)le16_to_cpu(raw_inode
->i_gid_low
);
1015 if(!(test_opt (inode
->i_sb
, NO_UID32
))) {
1016 inode
->i_uid
|= le16_to_cpu(raw_inode
->i_uid_high
) << 16;
1017 inode
->i_gid
|= le16_to_cpu(raw_inode
->i_gid_high
) << 16;
1019 inode
->i_nlink
= le16_to_cpu(raw_inode
->i_links_count
);
1020 inode
->i_size
= le32_to_cpu(raw_inode
->i_size
);
1021 inode
->i_atime
= le32_to_cpu(raw_inode
->i_atime
);
1022 inode
->i_ctime
= le32_to_cpu(raw_inode
->i_ctime
);
1023 inode
->i_mtime
= le32_to_cpu(raw_inode
->i_mtime
);
1024 inode
->u
.ext2_i
.i_dtime
= le32_to_cpu(raw_inode
->i_dtime
);
1025 /* We now have enough fields to check if the inode was active or not.
1026 * This is needed because nfsd might try to access dead inodes
1027 * the test is that same one that e2fsck uses
1028 * NeilBrown 1999oct15
1030 if (inode
->i_nlink
== 0 && (inode
->i_mode
== 0 || inode
->u
.ext2_i
.i_dtime
)) {
1031 /* this inode is deleted */
1035 inode
->i_blksize
= PAGE_SIZE
; /* This is the optimal IO size (for stat), not the fs block size */
1036 inode
->i_blocks
= le32_to_cpu(raw_inode
->i_blocks
);
1037 inode
->i_version
= ++event
;
1038 inode
->u
.ext2_i
.i_flags
= le32_to_cpu(raw_inode
->i_flags
);
1039 inode
->u
.ext2_i
.i_faddr
= le32_to_cpu(raw_inode
->i_faddr
);
1040 inode
->u
.ext2_i
.i_frag_no
= raw_inode
->i_frag
;
1041 inode
->u
.ext2_i
.i_frag_size
= raw_inode
->i_fsize
;
1042 inode
->u
.ext2_i
.i_file_acl
= le32_to_cpu(raw_inode
->i_file_acl
);
1043 if (S_ISDIR(inode
->i_mode
))
1044 inode
->u
.ext2_i
.i_dir_acl
= le32_to_cpu(raw_inode
->i_dir_acl
);
1046 inode
->u
.ext2_i
.i_high_size
= le32_to_cpu(raw_inode
->i_size_high
);
1047 inode
->i_size
|= ((__u64
)le32_to_cpu(raw_inode
->i_size_high
)) << 32;
1049 inode
->i_generation
= le32_to_cpu(raw_inode
->i_generation
);
1050 inode
->u
.ext2_i
.i_block_group
= block_group
;
1053 * NOTE! The in-memory inode i_data array is in little-endian order
1054 * even on big-endian machines: we do NOT byteswap the block numbers!
1056 for (block
= 0; block
< EXT2_N_BLOCKS
; block
++)
1057 inode
->u
.ext2_i
.i_data
[block
] = raw_inode
->i_block
[block
];
1059 if (inode
->i_ino
== EXT2_ACL_IDX_INO
||
1060 inode
->i_ino
== EXT2_ACL_DATA_INO
)
1061 /* Nothing to do */ ;
1062 else if (S_ISREG(inode
->i_mode
)) {
1063 inode
->i_op
= &ext2_file_inode_operations
;
1064 inode
->i_fop
= &ext2_file_operations
;
1065 inode
->i_mapping
->a_ops
= &ext2_aops
;
1066 } else if (S_ISDIR(inode
->i_mode
)) {
1067 inode
->i_op
= &ext2_dir_inode_operations
;
1068 inode
->i_fop
= &ext2_dir_operations
;
1069 } else if (S_ISLNK(inode
->i_mode
)) {
1070 if (!inode
->i_blocks
)
1071 inode
->i_op
= &ext2_fast_symlink_inode_operations
;
1073 inode
->i_op
= &page_symlink_inode_operations
;
1074 inode
->i_mapping
->a_ops
= &ext2_aops
;
1077 init_special_inode(inode
, inode
->i_mode
,
1078 le32_to_cpu(raw_inode
->i_block
[0]));
1080 inode
->i_attr_flags
= 0;
1081 if (inode
->u
.ext2_i
.i_flags
& EXT2_SYNC_FL
) {
1082 inode
->i_attr_flags
|= ATTR_FLAG_SYNCRONOUS
;
1083 inode
->i_flags
|= S_SYNC
;
1085 if (inode
->u
.ext2_i
.i_flags
& EXT2_APPEND_FL
) {
1086 inode
->i_attr_flags
|= ATTR_FLAG_APPEND
;
1087 inode
->i_flags
|= S_APPEND
;
1089 if (inode
->u
.ext2_i
.i_flags
& EXT2_IMMUTABLE_FL
) {
1090 inode
->i_attr_flags
|= ATTR_FLAG_IMMUTABLE
;
1091 inode
->i_flags
|= S_IMMUTABLE
;
1093 if (inode
->u
.ext2_i
.i_flags
& EXT2_NOATIME_FL
) {
1094 inode
->i_attr_flags
|= ATTR_FLAG_NOATIME
;
1095 inode
->i_flags
|= S_NOATIME
;
1100 make_bad_inode(inode
);
1104 static int ext2_update_inode(struct inode
* inode
, int do_sync
)
1106 struct buffer_head
* bh
;
1107 struct ext2_inode
* raw_inode
;
1108 unsigned long block_group
;
1109 unsigned long group_desc
;
1111 unsigned long block
;
1112 unsigned long offset
;
1114 struct ext2_group_desc
* gdp
;
1116 if ((inode
->i_ino
!= EXT2_ROOT_INO
&&
1117 inode
->i_ino
< EXT2_FIRST_INO(inode
->i_sb
)) ||
1118 inode
->i_ino
> le32_to_cpu(inode
->i_sb
->u
.ext2_sb
.s_es
->s_inodes_count
)) {
1119 ext2_error (inode
->i_sb
, "ext2_write_inode",
1120 "bad inode number: %lu", inode
->i_ino
);
1123 block_group
= (inode
->i_ino
- 1) / EXT2_INODES_PER_GROUP(inode
->i_sb
);
1124 if (block_group
>= inode
->i_sb
->u
.ext2_sb
.s_groups_count
) {
1125 ext2_error (inode
->i_sb
, "ext2_write_inode",
1126 "group >= groups count");
1129 group_desc
= block_group
>> EXT2_DESC_PER_BLOCK_BITS(inode
->i_sb
);
1130 desc
= block_group
& (EXT2_DESC_PER_BLOCK(inode
->i_sb
) - 1);
1131 bh
= inode
->i_sb
->u
.ext2_sb
.s_group_desc
[group_desc
];
1133 ext2_error (inode
->i_sb
, "ext2_write_inode",
1134 "Descriptor not loaded");
1137 gdp
= (struct ext2_group_desc
*) bh
->b_data
;
1139 * Figure out the offset within the block group inode table
1141 offset
= ((inode
->i_ino
- 1) % EXT2_INODES_PER_GROUP(inode
->i_sb
)) *
1142 EXT2_INODE_SIZE(inode
->i_sb
);
1143 block
= le32_to_cpu(gdp
[desc
].bg_inode_table
) +
1144 (offset
>> EXT2_BLOCK_SIZE_BITS(inode
->i_sb
));
1145 if (!(bh
= bread (inode
->i_dev
, block
, inode
->i_sb
->s_blocksize
))) {
1146 ext2_error (inode
->i_sb
, "ext2_write_inode",
1147 "unable to read inode block - "
1148 "inode=%lu, block=%lu", inode
->i_ino
, block
);
1151 offset
&= EXT2_BLOCK_SIZE(inode
->i_sb
) - 1;
1152 raw_inode
= (struct ext2_inode
*) (bh
->b_data
+ offset
);
1154 raw_inode
->i_mode
= cpu_to_le16(inode
->i_mode
);
1155 if(!(test_opt(inode
->i_sb
, NO_UID32
))) {
1156 raw_inode
->i_uid_low
= cpu_to_le16(low_16_bits(inode
->i_uid
));
1157 raw_inode
->i_gid_low
= cpu_to_le16(low_16_bits(inode
->i_gid
));
1159 * Fix up interoperability with old kernels. Otherwise, old inodes get
1160 * re-used with the upper 16 bits of the uid/gid intact
1162 if(!inode
->u
.ext2_i
.i_dtime
) {
1163 raw_inode
->i_uid_high
= cpu_to_le16(high_16_bits(inode
->i_uid
));
1164 raw_inode
->i_gid_high
= cpu_to_le16(high_16_bits(inode
->i_gid
));
1166 raw_inode
->i_uid_high
= 0;
1167 raw_inode
->i_gid_high
= 0;
1170 raw_inode
->i_uid_low
= cpu_to_le16(fs_high2lowuid(inode
->i_uid
));
1171 raw_inode
->i_gid_low
= cpu_to_le16(fs_high2lowgid(inode
->i_gid
));
1172 raw_inode
->i_uid_high
= 0;
1173 raw_inode
->i_gid_high
= 0;
1175 raw_inode
->i_links_count
= cpu_to_le16(inode
->i_nlink
);
1176 raw_inode
->i_size
= cpu_to_le32(inode
->i_size
);
1177 raw_inode
->i_atime
= cpu_to_le32(inode
->i_atime
);
1178 raw_inode
->i_ctime
= cpu_to_le32(inode
->i_ctime
);
1179 raw_inode
->i_mtime
= cpu_to_le32(inode
->i_mtime
);
1180 raw_inode
->i_blocks
= cpu_to_le32(inode
->i_blocks
);
1181 raw_inode
->i_dtime
= cpu_to_le32(inode
->u
.ext2_i
.i_dtime
);
1182 raw_inode
->i_flags
= cpu_to_le32(inode
->u
.ext2_i
.i_flags
);
1183 raw_inode
->i_faddr
= cpu_to_le32(inode
->u
.ext2_i
.i_faddr
);
1184 raw_inode
->i_frag
= inode
->u
.ext2_i
.i_frag_no
;
1185 raw_inode
->i_fsize
= inode
->u
.ext2_i
.i_frag_size
;
1186 raw_inode
->i_file_acl
= cpu_to_le32(inode
->u
.ext2_i
.i_file_acl
);
1187 if (S_ISDIR(inode
->i_mode
))
1188 raw_inode
->i_dir_acl
= cpu_to_le32(inode
->u
.ext2_i
.i_dir_acl
);
1190 raw_inode
->i_size_high
= cpu_to_le32(inode
->i_size
>> 32);
1191 if (raw_inode
->i_size_high
) {
1192 struct super_block
*sb
= inode
->i_sb
;
1193 struct ext2_super_block
*es
= sb
->u
.ext2_sb
.s_es
;
1194 if (!(es
->s_feature_ro_compat
& cpu_to_le32(EXT2_FEATURE_RO_COMPAT_LARGE_FILE
))) {
1195 /* If this is the first large file
1196 * created, add a flag to the superblock.
1199 es
->s_feature_ro_compat
|= cpu_to_le32(EXT2_FEATURE_RO_COMPAT_LARGE_FILE
);
1201 ext2_write_super(sb
);
1206 raw_inode
->i_generation
= cpu_to_le32(inode
->i_generation
);
1207 if (S_ISCHR(inode
->i_mode
) || S_ISBLK(inode
->i_mode
))
1208 raw_inode
->i_block
[0] = cpu_to_le32(kdev_t_to_nr(inode
->i_rdev
));
1209 else for (block
= 0; block
< EXT2_N_BLOCKS
; block
++)
1210 raw_inode
->i_block
[block
] = inode
->u
.ext2_i
.i_data
[block
];
1211 mark_buffer_dirty(bh
);
1213 ll_rw_block (WRITE
, 1, &bh
);
1214 wait_on_buffer (bh
);
1215 if (buffer_req(bh
) && !buffer_uptodate(bh
)) {
1216 printk ("IO error syncing ext2 inode ["
1218 bdevname(inode
->i_dev
), inode
->i_ino
);
1226 void ext2_write_inode (struct inode
* inode
, int wait
)
1229 ext2_update_inode (inode
, wait
);
1233 int ext2_sync_inode (struct inode
*inode
)
1235 return ext2_update_inode (inode
, 1);
1238 int ext2_notify_change(struct dentry
*dentry
, struct iattr
*iattr
)
1240 struct inode
*inode
= dentry
->d_inode
;
1245 if (iattr
->ia_valid
& ATTR_ATTR_FLAG
&&
1246 ((!(iattr
->ia_attr_flags
& ATTR_FLAG_APPEND
) !=
1247 !(inode
->u
.ext2_i
.i_flags
& EXT2_APPEND_FL
)) ||
1248 (!(iattr
->ia_attr_flags
& ATTR_FLAG_IMMUTABLE
) !=
1249 !(inode
->u
.ext2_i
.i_flags
& EXT2_IMMUTABLE_FL
)))) {
1250 if (!capable(CAP_LINUX_IMMUTABLE
))
1252 } else if ((current
->fsuid
!= inode
->i_uid
) && !capable(CAP_FOWNER
))
1255 retval
= inode_change_ok(inode
, iattr
);
1259 inode_setattr(inode
, iattr
);
1261 flags
= iattr
->ia_attr_flags
;
1262 if (flags
& ATTR_FLAG_SYNCRONOUS
) {
1263 inode
->i_flags
|= S_SYNC
;
1264 inode
->u
.ext2_i
.i_flags
|= EXT2_SYNC_FL
;
1266 inode
->i_flags
&= ~S_SYNC
;
1267 inode
->u
.ext2_i
.i_flags
&= ~EXT2_SYNC_FL
;
1269 if (flags
& ATTR_FLAG_NOATIME
) {
1270 inode
->i_flags
|= S_NOATIME
;
1271 inode
->u
.ext2_i
.i_flags
|= EXT2_NOATIME_FL
;
1273 inode
->i_flags
&= ~S_NOATIME
;
1274 inode
->u
.ext2_i
.i_flags
&= ~EXT2_NOATIME_FL
;
1276 if (flags
& ATTR_FLAG_APPEND
) {
1277 inode
->i_flags
|= S_APPEND
;
1278 inode
->u
.ext2_i
.i_flags
|= EXT2_APPEND_FL
;
1280 inode
->i_flags
&= ~S_APPEND
;
1281 inode
->u
.ext2_i
.i_flags
&= ~EXT2_APPEND_FL
;
1283 if (flags
& ATTR_FLAG_IMMUTABLE
) {
1284 inode
->i_flags
|= S_IMMUTABLE
;
1285 inode
->u
.ext2_i
.i_flags
|= EXT2_IMMUTABLE_FL
;
1287 inode
->i_flags
&= ~S_IMMUTABLE
;
1288 inode
->u
.ext2_i
.i_flags
&= ~EXT2_IMMUTABLE_FL
;
1290 mark_inode_dirty(inode
);