1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext2/inode.c
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
12 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * Goal-directed block allocation by Stephen Tweedie
17 * (sct@dcs.ed.ac.uk), 1993, 1998
18 * Big-endian to little-endian byte-swapping/bitmaps by
19 * David S. Miller (davem@caip.rutgers.edu), 1995
20 * 64-bit file support on 64-bit platforms by Jakub Jelinek
21 * (jj@sunsite.ms.mff.cuni.cz)
23 * Assorted race fixes, rewrite of ext2_get_block() by Al Viro, 2000
26 #include <linux/time.h>
27 #include <linux/highuid.h>
28 #include <linux/pagemap.h>
29 #include <linux/dax.h>
30 #include <linux/blkdev.h>
31 #include <linux/quotaops.h>
32 #include <linux/writeback.h>
33 #include <linux/buffer_head.h>
34 #include <linux/mpage.h>
35 #include <linux/fiemap.h>
36 #include <linux/iomap.h>
37 #include <linux/namei.h>
38 #include <linux/uio.h>
43 static int __ext2_write_inode(struct inode
*inode
, int do_sync
);
46 * Test whether an inode is a fast symlink.
48 static inline int ext2_inode_is_fast_symlink(struct inode
*inode
)
50 int ea_blocks
= EXT2_I(inode
)->i_file_acl
?
51 (inode
->i_sb
->s_blocksize
>> 9) : 0;
53 return (S_ISLNK(inode
->i_mode
) &&
54 inode
->i_blocks
- ea_blocks
== 0);
57 static void ext2_truncate_blocks(struct inode
*inode
, loff_t offset
);
59 static void ext2_write_failed(struct address_space
*mapping
, loff_t to
)
61 struct inode
*inode
= mapping
->host
;
63 if (to
> inode
->i_size
) {
64 truncate_pagecache(inode
, inode
->i_size
);
65 ext2_truncate_blocks(inode
, inode
->i_size
);
70 * Called at the last iput() if i_nlink is zero.
72 void ext2_evict_inode(struct inode
* inode
)
74 struct ext2_block_alloc_info
*rsv
;
77 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
79 dquot_initialize(inode
);
84 truncate_inode_pages_final(&inode
->i_data
);
87 sb_start_intwrite(inode
->i_sb
);
89 EXT2_I(inode
)->i_dtime
= get_seconds();
90 mark_inode_dirty(inode
);
91 __ext2_write_inode(inode
, inode_needs_sync(inode
));
95 ext2_truncate_blocks(inode
, 0);
96 ext2_xattr_delete_inode(inode
);
99 invalidate_inode_buffers(inode
);
102 ext2_discard_reservation(inode
);
103 rsv
= EXT2_I(inode
)->i_block_alloc_info
;
104 EXT2_I(inode
)->i_block_alloc_info
= NULL
;
109 ext2_free_inode(inode
);
110 sb_end_intwrite(inode
->i_sb
);
117 struct buffer_head
*bh
;
120 static inline void add_chain(Indirect
*p
, struct buffer_head
*bh
, __le32
*v
)
122 p
->key
= *(p
->p
= v
);
126 static inline int verify_chain(Indirect
*from
, Indirect
*to
)
128 while (from
<= to
&& from
->key
== *from
->p
)
134 * ext2_block_to_path - parse the block number into array of offsets
135 * @inode: inode in question (we are only interested in its superblock)
136 * @i_block: block number to be parsed
137 * @offsets: array to store the offsets in
138 * @boundary: set this non-zero if the referred-to block is likely to be
139 * followed (on disk) by an indirect block.
140 * To store the locations of file's data ext2 uses a data structure common
141 * for UNIX filesystems - tree of pointers anchored in the inode, with
142 * data blocks at leaves and indirect blocks in intermediate nodes.
143 * This function translates the block number into path in that tree -
144 * return value is the path length and @offsets[n] is the offset of
145 * pointer to (n+1)th node in the nth one. If @block is out of range
146 * (negative or too large) warning is printed and zero returned.
148 * Note: function doesn't find node addresses, so no IO is needed. All
149 * we need to know is the capacity of indirect blocks (taken from the
154 * Portability note: the last comparison (check that we fit into triple
155 * indirect block) is spelled differently, because otherwise on an
156 * architecture with 32-bit longs and 8Kb pages we might get into trouble
157 * if our filesystem had 8Kb blocks. We might use long long, but that would
158 * kill us on x86. Oh, well, at least the sign propagation does not matter -
159 * i_block would have to be negative in the very beginning, so we would not
163 static int ext2_block_to_path(struct inode
*inode
,
164 long i_block
, int offsets
[4], int *boundary
)
166 int ptrs
= EXT2_ADDR_PER_BLOCK(inode
->i_sb
);
167 int ptrs_bits
= EXT2_ADDR_PER_BLOCK_BITS(inode
->i_sb
);
168 const long direct_blocks
= EXT2_NDIR_BLOCKS
,
169 indirect_blocks
= ptrs
,
170 double_blocks
= (1 << (ptrs_bits
* 2));
175 ext2_msg(inode
->i_sb
, KERN_WARNING
,
176 "warning: %s: block < 0", __func__
);
177 } else if (i_block
< direct_blocks
) {
178 offsets
[n
++] = i_block
;
179 final
= direct_blocks
;
180 } else if ( (i_block
-= direct_blocks
) < indirect_blocks
) {
181 offsets
[n
++] = EXT2_IND_BLOCK
;
182 offsets
[n
++] = i_block
;
184 } else if ((i_block
-= indirect_blocks
) < double_blocks
) {
185 offsets
[n
++] = EXT2_DIND_BLOCK
;
186 offsets
[n
++] = i_block
>> ptrs_bits
;
187 offsets
[n
++] = i_block
& (ptrs
- 1);
189 } else if (((i_block
-= double_blocks
) >> (ptrs_bits
* 2)) < ptrs
) {
190 offsets
[n
++] = EXT2_TIND_BLOCK
;
191 offsets
[n
++] = i_block
>> (ptrs_bits
* 2);
192 offsets
[n
++] = (i_block
>> ptrs_bits
) & (ptrs
- 1);
193 offsets
[n
++] = i_block
& (ptrs
- 1);
196 ext2_msg(inode
->i_sb
, KERN_WARNING
,
197 "warning: %s: block is too big", __func__
);
200 *boundary
= final
- 1 - (i_block
& (ptrs
- 1));
206 * ext2_get_branch - read the chain of indirect blocks leading to data
207 * @inode: inode in question
208 * @depth: depth of the chain (1 - direct pointer, etc.)
209 * @offsets: offsets of pointers in inode/indirect blocks
210 * @chain: place to store the result
211 * @err: here we store the error value
213 * Function fills the array of triples <key, p, bh> and returns %NULL
214 * if everything went OK or the pointer to the last filled triple
215 * (incomplete one) otherwise. Upon the return chain[i].key contains
216 * the number of (i+1)-th block in the chain (as it is stored in memory,
217 * i.e. little-endian 32-bit), chain[i].p contains the address of that
218 * number (it points into struct inode for i==0 and into the bh->b_data
219 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect
220 * block for i>0 and NULL for i==0. In other words, it holds the block
221 * numbers of the chain, addresses they were taken from (and where we can
222 * verify that chain did not change) and buffer_heads hosting these
225 * Function stops when it stumbles upon zero pointer (absent block)
226 * (pointer to last triple returned, *@err == 0)
227 * or when it gets an IO error reading an indirect block
228 * (ditto, *@err == -EIO)
229 * or when it notices that chain had been changed while it was reading
230 * (ditto, *@err == -EAGAIN)
231 * or when it reads all @depth-1 indirect blocks successfully and finds
232 * the whole chain, all way to the data (returns %NULL, *err == 0).
234 static Indirect
*ext2_get_branch(struct inode
*inode
,
240 struct super_block
*sb
= inode
->i_sb
;
242 struct buffer_head
*bh
;
245 /* i_data is not going away, no lock needed */
246 add_chain (chain
, NULL
, EXT2_I(inode
)->i_data
+ *offsets
);
250 bh
= sb_bread(sb
, le32_to_cpu(p
->key
));
253 read_lock(&EXT2_I(inode
)->i_meta_lock
);
254 if (!verify_chain(chain
, p
))
256 add_chain(++p
, bh
, (__le32
*)bh
->b_data
+ *++offsets
);
257 read_unlock(&EXT2_I(inode
)->i_meta_lock
);
264 read_unlock(&EXT2_I(inode
)->i_meta_lock
);
275 * ext2_find_near - find a place for allocation with sufficient locality
277 * @ind: descriptor of indirect block.
279 * This function returns the preferred place for block allocation.
280 * It is used when heuristic for sequential allocation fails.
282 * + if there is a block to the left of our position - allocate near it.
283 * + if pointer will live in indirect block - allocate near that block.
284 * + if pointer will live in inode - allocate in the same cylinder group.
286 * In the latter case we colour the starting block by the callers PID to
287 * prevent it from clashing with concurrent allocations for a different inode
288 * in the same block group. The PID is used here so that functionally related
289 * files will be close-by on-disk.
291 * Caller must make sure that @ind is valid and will stay that way.
294 static ext2_fsblk_t
ext2_find_near(struct inode
*inode
, Indirect
*ind
)
296 struct ext2_inode_info
*ei
= EXT2_I(inode
);
297 __le32
*start
= ind
->bh
? (__le32
*) ind
->bh
->b_data
: ei
->i_data
;
299 ext2_fsblk_t bg_start
;
302 /* Try to find previous block */
303 for (p
= ind
->p
- 1; p
>= start
; p
--)
305 return le32_to_cpu(*p
);
307 /* No such thing, so let's try location of indirect block */
309 return ind
->bh
->b_blocknr
;
312 * It is going to be referred from inode itself? OK, just put it into
313 * the same cylinder group then.
315 bg_start
= ext2_group_first_block_no(inode
->i_sb
, ei
->i_block_group
);
316 colour
= (current
->pid
% 16) *
317 (EXT2_BLOCKS_PER_GROUP(inode
->i_sb
) / 16);
318 return bg_start
+ colour
;
322 * ext2_find_goal - find a preferred place for allocation.
324 * @block: block we want
325 * @partial: pointer to the last triple within a chain
327 * Returns preferred place for a block (the goal).
330 static inline ext2_fsblk_t
ext2_find_goal(struct inode
*inode
, long block
,
333 struct ext2_block_alloc_info
*block_i
;
335 block_i
= EXT2_I(inode
)->i_block_alloc_info
;
338 * try the heuristic for sequential allocation,
339 * failing that at least try to get decent locality.
341 if (block_i
&& (block
== block_i
->last_alloc_logical_block
+ 1)
342 && (block_i
->last_alloc_physical_block
!= 0)) {
343 return block_i
->last_alloc_physical_block
+ 1;
346 return ext2_find_near(inode
, partial
);
350 * ext2_blks_to_allocate: Look up the block map and count the number
351 * of direct blocks need to be allocated for the given branch.
353 * @branch: chain of indirect blocks
354 * @k: number of blocks need for indirect blocks
355 * @blks: number of data blocks to be mapped.
356 * @blocks_to_boundary: the offset in the indirect block
358 * return the total number of blocks to be allocate, including the
359 * direct and indirect blocks.
362 ext2_blks_to_allocate(Indirect
* branch
, int k
, unsigned long blks
,
363 int blocks_to_boundary
)
365 unsigned long count
= 0;
368 * Simple case, [t,d]Indirect block(s) has not allocated yet
369 * then it's clear blocks on that path have not allocated
372 /* right now don't hanel cross boundary allocation */
373 if (blks
< blocks_to_boundary
+ 1)
376 count
+= blocks_to_boundary
+ 1;
381 while (count
< blks
&& count
<= blocks_to_boundary
382 && le32_to_cpu(*(branch
[0].p
+ count
)) == 0) {
389 * ext2_alloc_blocks: multiple allocate blocks needed for a branch
390 * @indirect_blks: the number of blocks need to allocate for indirect
393 * @new_blocks: on return it will store the new block numbers for
394 * the indirect blocks(if needed) and the first direct block,
395 * @blks: on return it will store the total number of allocated
398 static int ext2_alloc_blocks(struct inode
*inode
,
399 ext2_fsblk_t goal
, int indirect_blks
, int blks
,
400 ext2_fsblk_t new_blocks
[4], int *err
)
403 unsigned long count
= 0;
405 ext2_fsblk_t current_block
= 0;
409 * Here we try to allocate the requested multiple blocks at once,
410 * on a best-effort basis.
411 * To build a branch, we should allocate blocks for
412 * the indirect blocks(if not allocated yet), and at least
413 * the first direct block of this branch. That's the
414 * minimum number of blocks need to allocate(required)
416 target
= blks
+ indirect_blks
;
420 /* allocating blocks for indirect blocks and direct blocks */
421 current_block
= ext2_new_blocks(inode
,goal
,&count
,err
);
426 /* allocate blocks for indirect blocks */
427 while (index
< indirect_blks
&& count
) {
428 new_blocks
[index
++] = current_block
++;
436 /* save the new block number for the first direct block */
437 new_blocks
[index
] = current_block
;
439 /* total number of blocks allocated for direct blocks */
444 for (i
= 0; i
<index
; i
++)
445 ext2_free_blocks(inode
, new_blocks
[i
], 1);
447 mark_inode_dirty(inode
);
452 * ext2_alloc_branch - allocate and set up a chain of blocks.
454 * @num: depth of the chain (number of blocks to allocate)
455 * @offsets: offsets (in the blocks) to store the pointers to next.
456 * @branch: place to store the chain in.
458 * This function allocates @num blocks, zeroes out all but the last one,
459 * links them into chain and (if we are synchronous) writes them to disk.
460 * In other words, it prepares a branch that can be spliced onto the
461 * inode. It stores the information about that chain in the branch[], in
462 * the same format as ext2_get_branch() would do. We are calling it after
463 * we had read the existing part of chain and partial points to the last
464 * triple of that (one with zero ->key). Upon the exit we have the same
465 * picture as after the successful ext2_get_block(), except that in one
466 * place chain is disconnected - *branch->p is still zero (we did not
467 * set the last link), but branch->key contains the number that should
468 * be placed into *branch->p to fill that gap.
470 * If allocation fails we free all blocks we've allocated (and forget
471 * their buffer_heads) and return the error value the from failed
472 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
473 * as described above and return 0.
476 static int ext2_alloc_branch(struct inode
*inode
,
477 int indirect_blks
, int *blks
, ext2_fsblk_t goal
,
478 int *offsets
, Indirect
*branch
)
480 int blocksize
= inode
->i_sb
->s_blocksize
;
483 struct buffer_head
*bh
;
485 ext2_fsblk_t new_blocks
[4];
486 ext2_fsblk_t current_block
;
488 num
= ext2_alloc_blocks(inode
, goal
, indirect_blks
,
489 *blks
, new_blocks
, &err
);
493 branch
[0].key
= cpu_to_le32(new_blocks
[0]);
495 * metadata blocks and data blocks are allocated.
497 for (n
= 1; n
<= indirect_blks
; n
++) {
499 * Get buffer_head for parent block, zero it out
500 * and set the pointer to new one, then send
503 bh
= sb_getblk(inode
->i_sb
, new_blocks
[n
-1]);
510 memset(bh
->b_data
, 0, blocksize
);
511 branch
[n
].p
= (__le32
*) bh
->b_data
+ offsets
[n
];
512 branch
[n
].key
= cpu_to_le32(new_blocks
[n
]);
513 *branch
[n
].p
= branch
[n
].key
;
514 if ( n
== indirect_blks
) {
515 current_block
= new_blocks
[n
];
517 * End of chain, update the last new metablock of
518 * the chain to point to the new allocated
519 * data blocks numbers
521 for (i
=1; i
< num
; i
++)
522 *(branch
[n
].p
+ i
) = cpu_to_le32(++current_block
);
524 set_buffer_uptodate(bh
);
526 mark_buffer_dirty_inode(bh
, inode
);
527 /* We used to sync bh here if IS_SYNC(inode).
528 * But we now rely upon generic_write_sync()
529 * and b_inode_buffers. But not for directories.
531 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
532 sync_dirty_buffer(bh
);
538 for (i
= 1; i
< n
; i
++)
539 bforget(branch
[i
].bh
);
540 for (i
= 0; i
< indirect_blks
; i
++)
541 ext2_free_blocks(inode
, new_blocks
[i
], 1);
542 ext2_free_blocks(inode
, new_blocks
[i
], num
);
547 * ext2_splice_branch - splice the allocated branch onto inode.
549 * @block: (logical) number of block we are adding
550 * @where: location of missing link
551 * @num: number of indirect blocks we are adding
552 * @blks: number of direct blocks we are adding
554 * This function fills the missing link and does all housekeeping needed in
555 * inode (->i_blocks, etc.). In case of success we end up with the full
556 * chain to new block and return 0.
558 static void ext2_splice_branch(struct inode
*inode
,
559 long block
, Indirect
*where
, int num
, int blks
)
562 struct ext2_block_alloc_info
*block_i
;
563 ext2_fsblk_t current_block
;
565 block_i
= EXT2_I(inode
)->i_block_alloc_info
;
567 /* XXX LOCKING probably should have i_meta_lock ?*/
570 *where
->p
= where
->key
;
573 * Update the host buffer_head or inode to point to more just allocated
574 * direct blocks blocks
576 if (num
== 0 && blks
> 1) {
577 current_block
= le32_to_cpu(where
->key
) + 1;
578 for (i
= 1; i
< blks
; i
++)
579 *(where
->p
+ i
) = cpu_to_le32(current_block
++);
583 * update the most recently allocated logical & physical block
584 * in i_block_alloc_info, to assist find the proper goal block for next
588 block_i
->last_alloc_logical_block
= block
+ blks
- 1;
589 block_i
->last_alloc_physical_block
=
590 le32_to_cpu(where
[num
].key
) + blks
- 1;
593 /* We are done with atomic stuff, now do the rest of housekeeping */
595 /* had we spliced it onto indirect block? */
597 mark_buffer_dirty_inode(where
->bh
, inode
);
599 inode
->i_ctime
= current_time(inode
);
600 mark_inode_dirty(inode
);
604 * Allocation strategy is simple: if we have to allocate something, we will
605 * have to go the whole way to leaf. So let's do it before attaching anything
606 * to tree, set linkage between the newborn blocks, write them if sync is
607 * required, recheck the path, free and repeat if check fails, otherwise
608 * set the last missing link (that will protect us from any truncate-generated
609 * removals - all blocks on the path are immune now) and possibly force the
610 * write on the parent block.
611 * That has a nice additional property: no special recovery from the failed
612 * allocations is needed - we simply release blocks and do not touch anything
613 * reachable from inode.
615 * `handle' can be NULL if create == 0.
617 * return > 0, # of blocks mapped or allocated.
618 * return = 0, if plain lookup failed.
619 * return < 0, error case.
621 static int ext2_get_blocks(struct inode
*inode
,
622 sector_t iblock
, unsigned long maxblocks
,
623 u32
*bno
, bool *new, bool *boundary
,
632 int blocks_to_boundary
= 0;
634 struct ext2_inode_info
*ei
= EXT2_I(inode
);
636 ext2_fsblk_t first_block
= 0;
638 BUG_ON(maxblocks
== 0);
640 depth
= ext2_block_to_path(inode
,iblock
,offsets
,&blocks_to_boundary
);
645 partial
= ext2_get_branch(inode
, depth
, offsets
, chain
, &err
);
646 /* Simplest case - block found, no allocation needed */
648 first_block
= le32_to_cpu(chain
[depth
- 1].key
);
651 while (count
< maxblocks
&& count
<= blocks_to_boundary
) {
654 if (!verify_chain(chain
, chain
+ depth
- 1)) {
656 * Indirect block might be removed by
657 * truncate while we were reading it.
658 * Handling of that case: forget what we've
659 * got now, go to reread.
663 partial
= chain
+ depth
- 1;
666 blk
= le32_to_cpu(*(chain
[depth
-1].p
+ count
));
667 if (blk
== first_block
+ count
)
676 /* Next simple case - plain lookup or failed read of indirect block */
677 if (!create
|| err
== -EIO
)
680 mutex_lock(&ei
->truncate_mutex
);
682 * If the indirect block is missing while we are reading
683 * the chain(ext2_get_branch() returns -EAGAIN err), or
684 * if the chain has been changed after we grab the semaphore,
685 * (either because another process truncated this branch, or
686 * another get_block allocated this branch) re-grab the chain to see if
687 * the request block has been allocated or not.
689 * Since we already block the truncate/other get_block
690 * at this point, we will have the current copy of the chain when we
691 * splice the branch into the tree.
693 if (err
== -EAGAIN
|| !verify_chain(chain
, partial
)) {
694 while (partial
> chain
) {
698 partial
= ext2_get_branch(inode
, depth
, offsets
, chain
, &err
);
701 mutex_unlock(&ei
->truncate_mutex
);
706 mutex_unlock(&ei
->truncate_mutex
);
712 * Okay, we need to do block allocation. Lazily initialize the block
713 * allocation info here if necessary
715 if (S_ISREG(inode
->i_mode
) && (!ei
->i_block_alloc_info
))
716 ext2_init_block_alloc_info(inode
);
718 goal
= ext2_find_goal(inode
, iblock
, partial
);
720 /* the number of blocks need to allocate for [d,t]indirect blocks */
721 indirect_blks
= (chain
+ depth
) - partial
- 1;
723 * Next look up the indirect map to count the totoal number of
724 * direct blocks to allocate for this branch.
726 count
= ext2_blks_to_allocate(partial
, indirect_blks
,
727 maxblocks
, blocks_to_boundary
);
729 * XXX ???? Block out ext2_truncate while we alter the tree
731 err
= ext2_alloc_branch(inode
, indirect_blks
, &count
, goal
,
732 offsets
+ (partial
- chain
), partial
);
735 mutex_unlock(&ei
->truncate_mutex
);
741 * We must unmap blocks before zeroing so that writeback cannot
742 * overwrite zeros with stale data from block device page cache.
744 clean_bdev_aliases(inode
->i_sb
->s_bdev
,
745 le32_to_cpu(chain
[depth
-1].key
),
748 * block must be initialised before we put it in the tree
749 * so that it's not found by another thread before it's
752 err
= sb_issue_zeroout(inode
->i_sb
,
753 le32_to_cpu(chain
[depth
-1].key
), count
,
756 mutex_unlock(&ei
->truncate_mutex
);
762 ext2_splice_branch(inode
, iblock
, partial
, indirect_blks
, count
);
763 mutex_unlock(&ei
->truncate_mutex
);
765 if (count
> blocks_to_boundary
)
768 /* Clean up and exit */
769 partial
= chain
+ depth
- 1; /* the whole chain */
771 while (partial
> chain
) {
776 *bno
= le32_to_cpu(chain
[depth
-1].key
);
780 int ext2_get_block(struct inode
*inode
, sector_t iblock
,
781 struct buffer_head
*bh_result
, int create
)
783 unsigned max_blocks
= bh_result
->b_size
>> inode
->i_blkbits
;
784 bool new = false, boundary
= false;
788 ret
= ext2_get_blocks(inode
, iblock
, max_blocks
, &bno
, &new, &boundary
,
793 map_bh(bh_result
, inode
->i_sb
, bno
);
794 bh_result
->b_size
= (ret
<< inode
->i_blkbits
);
796 set_buffer_new(bh_result
);
798 set_buffer_boundary(bh_result
);
804 static int ext2_iomap_begin(struct inode
*inode
, loff_t offset
, loff_t length
,
805 unsigned flags
, struct iomap
*iomap
)
807 unsigned int blkbits
= inode
->i_blkbits
;
808 unsigned long first_block
= offset
>> blkbits
;
809 unsigned long max_blocks
= (length
+ (1 << blkbits
) - 1) >> blkbits
;
810 struct ext2_sb_info
*sbi
= EXT2_SB(inode
->i_sb
);
811 bool new = false, boundary
= false;
815 ret
= ext2_get_blocks(inode
, first_block
, max_blocks
,
816 &bno
, &new, &boundary
, flags
& IOMAP_WRITE
);
821 iomap
->bdev
= inode
->i_sb
->s_bdev
;
822 iomap
->offset
= (u64
)first_block
<< blkbits
;
823 iomap
->dax_dev
= sbi
->s_daxdev
;
826 iomap
->type
= IOMAP_HOLE
;
827 iomap
->blkno
= IOMAP_NULL_BLOCK
;
828 iomap
->length
= 1 << blkbits
;
830 iomap
->type
= IOMAP_MAPPED
;
831 iomap
->blkno
= (sector_t
)bno
<< (blkbits
- 9);
832 iomap
->length
= (u64
)ret
<< blkbits
;
833 iomap
->flags
|= IOMAP_F_MERGED
;
837 iomap
->flags
|= IOMAP_F_NEW
;
842 ext2_iomap_end(struct inode
*inode
, loff_t offset
, loff_t length
,
843 ssize_t written
, unsigned flags
, struct iomap
*iomap
)
845 if (iomap
->type
== IOMAP_MAPPED
&&
847 (flags
& IOMAP_WRITE
))
848 ext2_write_failed(inode
->i_mapping
, offset
+ length
);
852 const struct iomap_ops ext2_iomap_ops
= {
853 .iomap_begin
= ext2_iomap_begin
,
854 .iomap_end
= ext2_iomap_end
,
857 /* Define empty ops for !CONFIG_FS_DAX case to avoid ugly ifdefs */
858 const struct iomap_ops ext2_iomap_ops
;
859 #endif /* CONFIG_FS_DAX */
861 int ext2_fiemap(struct inode
*inode
, struct fiemap_extent_info
*fieinfo
,
864 return generic_block_fiemap(inode
, fieinfo
, start
, len
,
868 static int ext2_writepage(struct page
*page
, struct writeback_control
*wbc
)
870 return block_write_full_page(page
, ext2_get_block
, wbc
);
873 static int ext2_readpage(struct file
*file
, struct page
*page
)
875 return mpage_readpage(page
, ext2_get_block
);
879 ext2_readpages(struct file
*file
, struct address_space
*mapping
,
880 struct list_head
*pages
, unsigned nr_pages
)
882 return mpage_readpages(mapping
, pages
, nr_pages
, ext2_get_block
);
886 ext2_write_begin(struct file
*file
, struct address_space
*mapping
,
887 loff_t pos
, unsigned len
, unsigned flags
,
888 struct page
**pagep
, void **fsdata
)
892 ret
= block_write_begin(mapping
, pos
, len
, flags
, pagep
,
895 ext2_write_failed(mapping
, pos
+ len
);
899 static int ext2_write_end(struct file
*file
, struct address_space
*mapping
,
900 loff_t pos
, unsigned len
, unsigned copied
,
901 struct page
*page
, void *fsdata
)
905 ret
= generic_write_end(file
, mapping
, pos
, len
, copied
, page
, fsdata
);
907 ext2_write_failed(mapping
, pos
+ len
);
912 ext2_nobh_write_begin(struct file
*file
, struct address_space
*mapping
,
913 loff_t pos
, unsigned len
, unsigned flags
,
914 struct page
**pagep
, void **fsdata
)
918 ret
= nobh_write_begin(mapping
, pos
, len
, flags
, pagep
, fsdata
,
921 ext2_write_failed(mapping
, pos
+ len
);
925 static int ext2_nobh_writepage(struct page
*page
,
926 struct writeback_control
*wbc
)
928 return nobh_writepage(page
, ext2_get_block
, wbc
);
931 static sector_t
ext2_bmap(struct address_space
*mapping
, sector_t block
)
933 return generic_block_bmap(mapping
,block
,ext2_get_block
);
937 ext2_direct_IO(struct kiocb
*iocb
, struct iov_iter
*iter
)
939 struct file
*file
= iocb
->ki_filp
;
940 struct address_space
*mapping
= file
->f_mapping
;
941 struct inode
*inode
= mapping
->host
;
942 size_t count
= iov_iter_count(iter
);
943 loff_t offset
= iocb
->ki_pos
;
946 if (WARN_ON_ONCE(IS_DAX(inode
)))
949 ret
= blockdev_direct_IO(iocb
, inode
, iter
, ext2_get_block
);
950 if (ret
< 0 && iov_iter_rw(iter
) == WRITE
)
951 ext2_write_failed(mapping
, offset
+ count
);
956 ext2_writepages(struct address_space
*mapping
, struct writeback_control
*wbc
)
959 if (dax_mapping(mapping
)) {
960 return dax_writeback_mapping_range(mapping
,
961 mapping
->host
->i_sb
->s_bdev
,
966 return mpage_writepages(mapping
, wbc
, ext2_get_block
);
969 const struct address_space_operations ext2_aops
= {
970 .readpage
= ext2_readpage
,
971 .readpages
= ext2_readpages
,
972 .writepage
= ext2_writepage
,
973 .write_begin
= ext2_write_begin
,
974 .write_end
= ext2_write_end
,
976 .direct_IO
= ext2_direct_IO
,
977 .writepages
= ext2_writepages
,
978 .migratepage
= buffer_migrate_page
,
979 .is_partially_uptodate
= block_is_partially_uptodate
,
980 .error_remove_page
= generic_error_remove_page
,
983 const struct address_space_operations ext2_nobh_aops
= {
984 .readpage
= ext2_readpage
,
985 .readpages
= ext2_readpages
,
986 .writepage
= ext2_nobh_writepage
,
987 .write_begin
= ext2_nobh_write_begin
,
988 .write_end
= nobh_write_end
,
990 .direct_IO
= ext2_direct_IO
,
991 .writepages
= ext2_writepages
,
992 .migratepage
= buffer_migrate_page
,
993 .error_remove_page
= generic_error_remove_page
,
997 * Probably it should be a library function... search for first non-zero word
998 * or memcmp with zero_page, whatever is better for particular architecture.
1001 static inline int all_zeroes(__le32
*p
, __le32
*q
)
1010 * ext2_find_shared - find the indirect blocks for partial truncation.
1011 * @inode: inode in question
1012 * @depth: depth of the affected branch
1013 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
1014 * @chain: place to store the pointers to partial indirect blocks
1015 * @top: place to the (detached) top of branch
1017 * This is a helper function used by ext2_truncate().
1019 * When we do truncate() we may have to clean the ends of several indirect
1020 * blocks but leave the blocks themselves alive. Block is partially
1021 * truncated if some data below the new i_size is referred from it (and
1022 * it is on the path to the first completely truncated data block, indeed).
1023 * We have to free the top of that path along with everything to the right
1024 * of the path. Since no allocation past the truncation point is possible
1025 * until ext2_truncate() finishes, we may safely do the latter, but top
1026 * of branch may require special attention - pageout below the truncation
1027 * point might try to populate it.
1029 * We atomically detach the top of branch from the tree, store the block
1030 * number of its root in *@top, pointers to buffer_heads of partially
1031 * truncated blocks - in @chain[].bh and pointers to their last elements
1032 * that should not be removed - in @chain[].p. Return value is the pointer
1033 * to last filled element of @chain.
1035 * The work left to caller to do the actual freeing of subtrees:
1036 * a) free the subtree starting from *@top
1037 * b) free the subtrees whose roots are stored in
1038 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
1039 * c) free the subtrees growing from the inode past the @chain[0].p
1040 * (no partially truncated stuff there).
1043 static Indirect
*ext2_find_shared(struct inode
*inode
,
1049 Indirect
*partial
, *p
;
1053 for (k
= depth
; k
> 1 && !offsets
[k
-1]; k
--)
1055 partial
= ext2_get_branch(inode
, k
, offsets
, chain
, &err
);
1057 partial
= chain
+ k
-1;
1059 * If the branch acquired continuation since we've looked at it -
1060 * fine, it should all survive and (new) top doesn't belong to us.
1062 write_lock(&EXT2_I(inode
)->i_meta_lock
);
1063 if (!partial
->key
&& *partial
->p
) {
1064 write_unlock(&EXT2_I(inode
)->i_meta_lock
);
1067 for (p
=partial
; p
>chain
&& all_zeroes((__le32
*)p
->bh
->b_data
,p
->p
); p
--)
1070 * OK, we've found the last block that must survive. The rest of our
1071 * branch should be detached before unlocking. However, if that rest
1072 * of branch is all ours and does not grow immediately from the inode
1073 * it's easier to cheat and just decrement partial->p.
1075 if (p
== chain
+ k
- 1 && p
> chain
) {
1081 write_unlock(&EXT2_I(inode
)->i_meta_lock
);
1085 brelse(partial
->bh
);
1093 * ext2_free_data - free a list of data blocks
1094 * @inode: inode we are dealing with
1095 * @p: array of block numbers
1096 * @q: points immediately past the end of array
1098 * We are freeing all blocks referred from that array (numbers are
1099 * stored as little-endian 32-bit) and updating @inode->i_blocks
1102 static inline void ext2_free_data(struct inode
*inode
, __le32
*p
, __le32
*q
)
1104 unsigned long block_to_free
= 0, count
= 0;
1107 for ( ; p
< q
; p
++) {
1108 nr
= le32_to_cpu(*p
);
1111 /* accumulate blocks to free if they're contiguous */
1114 else if (block_to_free
== nr
- count
)
1117 ext2_free_blocks (inode
, block_to_free
, count
);
1118 mark_inode_dirty(inode
);
1126 ext2_free_blocks (inode
, block_to_free
, count
);
1127 mark_inode_dirty(inode
);
1132 * ext2_free_branches - free an array of branches
1133 * @inode: inode we are dealing with
1134 * @p: array of block numbers
1135 * @q: pointer immediately past the end of array
1136 * @depth: depth of the branches to free
1138 * We are freeing all blocks referred from these branches (numbers are
1139 * stored as little-endian 32-bit) and updating @inode->i_blocks
1142 static void ext2_free_branches(struct inode
*inode
, __le32
*p
, __le32
*q
, int depth
)
1144 struct buffer_head
* bh
;
1148 int addr_per_block
= EXT2_ADDR_PER_BLOCK(inode
->i_sb
);
1149 for ( ; p
< q
; p
++) {
1150 nr
= le32_to_cpu(*p
);
1154 bh
= sb_bread(inode
->i_sb
, nr
);
1156 * A read failure? Report error and clear slot
1160 ext2_error(inode
->i_sb
, "ext2_free_branches",
1161 "Read failure, inode=%ld, block=%ld",
1165 ext2_free_branches(inode
,
1166 (__le32
*)bh
->b_data
,
1167 (__le32
*)bh
->b_data
+ addr_per_block
,
1170 ext2_free_blocks(inode
, nr
, 1);
1171 mark_inode_dirty(inode
);
1174 ext2_free_data(inode
, p
, q
);
1177 /* dax_sem must be held when calling this function */
1178 static void __ext2_truncate_blocks(struct inode
*inode
, loff_t offset
)
1180 __le32
*i_data
= EXT2_I(inode
)->i_data
;
1181 struct ext2_inode_info
*ei
= EXT2_I(inode
);
1182 int addr_per_block
= EXT2_ADDR_PER_BLOCK(inode
->i_sb
);
1190 blocksize
= inode
->i_sb
->s_blocksize
;
1191 iblock
= (offset
+ blocksize
-1) >> EXT2_BLOCK_SIZE_BITS(inode
->i_sb
);
1193 #ifdef CONFIG_FS_DAX
1194 WARN_ON(!rwsem_is_locked(&ei
->dax_sem
));
1197 n
= ext2_block_to_path(inode
, iblock
, offsets
, NULL
);
1202 * From here we block out all ext2_get_block() callers who want to
1203 * modify the block allocation tree.
1205 mutex_lock(&ei
->truncate_mutex
);
1208 ext2_free_data(inode
, i_data
+offsets
[0],
1209 i_data
+ EXT2_NDIR_BLOCKS
);
1213 partial
= ext2_find_shared(inode
, n
, offsets
, chain
, &nr
);
1214 /* Kill the top of shared branch (already detached) */
1216 if (partial
== chain
)
1217 mark_inode_dirty(inode
);
1219 mark_buffer_dirty_inode(partial
->bh
, inode
);
1220 ext2_free_branches(inode
, &nr
, &nr
+1, (chain
+n
-1) - partial
);
1222 /* Clear the ends of indirect blocks on the shared branch */
1223 while (partial
> chain
) {
1224 ext2_free_branches(inode
,
1226 (__le32
*)partial
->bh
->b_data
+addr_per_block
,
1227 (chain
+n
-1) - partial
);
1228 mark_buffer_dirty_inode(partial
->bh
, inode
);
1229 brelse (partial
->bh
);
1233 /* Kill the remaining (whole) subtrees */
1234 switch (offsets
[0]) {
1236 nr
= i_data
[EXT2_IND_BLOCK
];
1238 i_data
[EXT2_IND_BLOCK
] = 0;
1239 mark_inode_dirty(inode
);
1240 ext2_free_branches(inode
, &nr
, &nr
+1, 1);
1242 case EXT2_IND_BLOCK
:
1243 nr
= i_data
[EXT2_DIND_BLOCK
];
1245 i_data
[EXT2_DIND_BLOCK
] = 0;
1246 mark_inode_dirty(inode
);
1247 ext2_free_branches(inode
, &nr
, &nr
+1, 2);
1249 case EXT2_DIND_BLOCK
:
1250 nr
= i_data
[EXT2_TIND_BLOCK
];
1252 i_data
[EXT2_TIND_BLOCK
] = 0;
1253 mark_inode_dirty(inode
);
1254 ext2_free_branches(inode
, &nr
, &nr
+1, 3);
1256 case EXT2_TIND_BLOCK
:
1260 ext2_discard_reservation(inode
);
1262 mutex_unlock(&ei
->truncate_mutex
);
1265 static void ext2_truncate_blocks(struct inode
*inode
, loff_t offset
)
1267 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
1268 S_ISLNK(inode
->i_mode
)))
1270 if (ext2_inode_is_fast_symlink(inode
))
1273 dax_sem_down_write(EXT2_I(inode
));
1274 __ext2_truncate_blocks(inode
, offset
);
1275 dax_sem_up_write(EXT2_I(inode
));
1278 static int ext2_setsize(struct inode
*inode
, loff_t newsize
)
1282 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
1283 S_ISLNK(inode
->i_mode
)))
1285 if (ext2_inode_is_fast_symlink(inode
))
1287 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
1290 inode_dio_wait(inode
);
1292 if (IS_DAX(inode
)) {
1293 error
= iomap_zero_range(inode
, newsize
,
1294 PAGE_ALIGN(newsize
) - newsize
, NULL
,
1296 } else if (test_opt(inode
->i_sb
, NOBH
))
1297 error
= nobh_truncate_page(inode
->i_mapping
,
1298 newsize
, ext2_get_block
);
1300 error
= block_truncate_page(inode
->i_mapping
,
1301 newsize
, ext2_get_block
);
1305 dax_sem_down_write(EXT2_I(inode
));
1306 truncate_setsize(inode
, newsize
);
1307 __ext2_truncate_blocks(inode
, newsize
);
1308 dax_sem_up_write(EXT2_I(inode
));
1310 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
1311 if (inode_needs_sync(inode
)) {
1312 sync_mapping_buffers(inode
->i_mapping
);
1313 sync_inode_metadata(inode
, 1);
1315 mark_inode_dirty(inode
);
1321 static struct ext2_inode
*ext2_get_inode(struct super_block
*sb
, ino_t ino
,
1322 struct buffer_head
**p
)
1324 struct buffer_head
* bh
;
1325 unsigned long block_group
;
1326 unsigned long block
;
1327 unsigned long offset
;
1328 struct ext2_group_desc
* gdp
;
1331 if ((ino
!= EXT2_ROOT_INO
&& ino
< EXT2_FIRST_INO(sb
)) ||
1332 ino
> le32_to_cpu(EXT2_SB(sb
)->s_es
->s_inodes_count
))
1335 block_group
= (ino
- 1) / EXT2_INODES_PER_GROUP(sb
);
1336 gdp
= ext2_get_group_desc(sb
, block_group
, NULL
);
1340 * Figure out the offset within the block group inode table
1342 offset
= ((ino
- 1) % EXT2_INODES_PER_GROUP(sb
)) * EXT2_INODE_SIZE(sb
);
1343 block
= le32_to_cpu(gdp
->bg_inode_table
) +
1344 (offset
>> EXT2_BLOCK_SIZE_BITS(sb
));
1345 if (!(bh
= sb_bread(sb
, block
)))
1349 offset
&= (EXT2_BLOCK_SIZE(sb
) - 1);
1350 return (struct ext2_inode
*) (bh
->b_data
+ offset
);
1353 ext2_error(sb
, "ext2_get_inode", "bad inode number: %lu",
1354 (unsigned long) ino
);
1355 return ERR_PTR(-EINVAL
);
1357 ext2_error(sb
, "ext2_get_inode",
1358 "unable to read inode block - inode=%lu, block=%lu",
1359 (unsigned long) ino
, block
);
1361 return ERR_PTR(-EIO
);
1364 void ext2_set_inode_flags(struct inode
*inode
)
1366 unsigned int flags
= EXT2_I(inode
)->i_flags
;
1368 inode
->i_flags
&= ~(S_SYNC
| S_APPEND
| S_IMMUTABLE
| S_NOATIME
|
1370 if (flags
& EXT2_SYNC_FL
)
1371 inode
->i_flags
|= S_SYNC
;
1372 if (flags
& EXT2_APPEND_FL
)
1373 inode
->i_flags
|= S_APPEND
;
1374 if (flags
& EXT2_IMMUTABLE_FL
)
1375 inode
->i_flags
|= S_IMMUTABLE
;
1376 if (flags
& EXT2_NOATIME_FL
)
1377 inode
->i_flags
|= S_NOATIME
;
1378 if (flags
& EXT2_DIRSYNC_FL
)
1379 inode
->i_flags
|= S_DIRSYNC
;
1380 if (test_opt(inode
->i_sb
, DAX
) && S_ISREG(inode
->i_mode
))
1381 inode
->i_flags
|= S_DAX
;
1384 struct inode
*ext2_iget (struct super_block
*sb
, unsigned long ino
)
1386 struct ext2_inode_info
*ei
;
1387 struct buffer_head
* bh
;
1388 struct ext2_inode
*raw_inode
;
1389 struct inode
*inode
;
1395 inode
= iget_locked(sb
, ino
);
1397 return ERR_PTR(-ENOMEM
);
1398 if (!(inode
->i_state
& I_NEW
))
1402 ei
->i_block_alloc_info
= NULL
;
1404 raw_inode
= ext2_get_inode(inode
->i_sb
, ino
, &bh
);
1405 if (IS_ERR(raw_inode
)) {
1406 ret
= PTR_ERR(raw_inode
);
1410 inode
->i_mode
= le16_to_cpu(raw_inode
->i_mode
);
1411 i_uid
= (uid_t
)le16_to_cpu(raw_inode
->i_uid_low
);
1412 i_gid
= (gid_t
)le16_to_cpu(raw_inode
->i_gid_low
);
1413 if (!(test_opt (inode
->i_sb
, NO_UID32
))) {
1414 i_uid
|= le16_to_cpu(raw_inode
->i_uid_high
) << 16;
1415 i_gid
|= le16_to_cpu(raw_inode
->i_gid_high
) << 16;
1417 i_uid_write(inode
, i_uid
);
1418 i_gid_write(inode
, i_gid
);
1419 set_nlink(inode
, le16_to_cpu(raw_inode
->i_links_count
));
1420 inode
->i_size
= le32_to_cpu(raw_inode
->i_size
);
1421 inode
->i_atime
.tv_sec
= (signed)le32_to_cpu(raw_inode
->i_atime
);
1422 inode
->i_ctime
.tv_sec
= (signed)le32_to_cpu(raw_inode
->i_ctime
);
1423 inode
->i_mtime
.tv_sec
= (signed)le32_to_cpu(raw_inode
->i_mtime
);
1424 inode
->i_atime
.tv_nsec
= inode
->i_mtime
.tv_nsec
= inode
->i_ctime
.tv_nsec
= 0;
1425 ei
->i_dtime
= le32_to_cpu(raw_inode
->i_dtime
);
1426 /* We now have enough fields to check if the inode was active or not.
1427 * This is needed because nfsd might try to access dead inodes
1428 * the test is that same one that e2fsck uses
1429 * NeilBrown 1999oct15
1431 if (inode
->i_nlink
== 0 && (inode
->i_mode
== 0 || ei
->i_dtime
)) {
1432 /* this inode is deleted */
1437 inode
->i_blocks
= le32_to_cpu(raw_inode
->i_blocks
);
1438 ei
->i_flags
= le32_to_cpu(raw_inode
->i_flags
);
1439 ei
->i_faddr
= le32_to_cpu(raw_inode
->i_faddr
);
1440 ei
->i_frag_no
= raw_inode
->i_frag
;
1441 ei
->i_frag_size
= raw_inode
->i_fsize
;
1442 ei
->i_file_acl
= le32_to_cpu(raw_inode
->i_file_acl
);
1445 if (ei
->i_file_acl
&&
1446 !ext2_data_block_valid(EXT2_SB(sb
), ei
->i_file_acl
, 1)) {
1447 ext2_error(sb
, "ext2_iget", "bad extended attribute block %u",
1450 ret
= -EFSCORRUPTED
;
1454 if (S_ISREG(inode
->i_mode
))
1455 inode
->i_size
|= ((__u64
)le32_to_cpu(raw_inode
->i_size_high
)) << 32;
1457 ei
->i_dir_acl
= le32_to_cpu(raw_inode
->i_dir_acl
);
1458 if (i_size_read(inode
) < 0) {
1459 ret
= -EFSCORRUPTED
;
1463 inode
->i_generation
= le32_to_cpu(raw_inode
->i_generation
);
1465 ei
->i_block_group
= (ino
- 1) / EXT2_INODES_PER_GROUP(inode
->i_sb
);
1466 ei
->i_dir_start_lookup
= 0;
1469 * NOTE! The in-memory inode i_data array is in little-endian order
1470 * even on big-endian machines: we do NOT byteswap the block numbers!
1472 for (n
= 0; n
< EXT2_N_BLOCKS
; n
++)
1473 ei
->i_data
[n
] = raw_inode
->i_block
[n
];
1475 if (S_ISREG(inode
->i_mode
)) {
1476 inode
->i_op
= &ext2_file_inode_operations
;
1477 if (test_opt(inode
->i_sb
, NOBH
)) {
1478 inode
->i_mapping
->a_ops
= &ext2_nobh_aops
;
1479 inode
->i_fop
= &ext2_file_operations
;
1481 inode
->i_mapping
->a_ops
= &ext2_aops
;
1482 inode
->i_fop
= &ext2_file_operations
;
1484 } else if (S_ISDIR(inode
->i_mode
)) {
1485 inode
->i_op
= &ext2_dir_inode_operations
;
1486 inode
->i_fop
= &ext2_dir_operations
;
1487 if (test_opt(inode
->i_sb
, NOBH
))
1488 inode
->i_mapping
->a_ops
= &ext2_nobh_aops
;
1490 inode
->i_mapping
->a_ops
= &ext2_aops
;
1491 } else if (S_ISLNK(inode
->i_mode
)) {
1492 if (ext2_inode_is_fast_symlink(inode
)) {
1493 inode
->i_link
= (char *)ei
->i_data
;
1494 inode
->i_op
= &ext2_fast_symlink_inode_operations
;
1495 nd_terminate_link(ei
->i_data
, inode
->i_size
,
1496 sizeof(ei
->i_data
) - 1);
1498 inode
->i_op
= &ext2_symlink_inode_operations
;
1499 inode_nohighmem(inode
);
1500 if (test_opt(inode
->i_sb
, NOBH
))
1501 inode
->i_mapping
->a_ops
= &ext2_nobh_aops
;
1503 inode
->i_mapping
->a_ops
= &ext2_aops
;
1506 inode
->i_op
= &ext2_special_inode_operations
;
1507 if (raw_inode
->i_block
[0])
1508 init_special_inode(inode
, inode
->i_mode
,
1509 old_decode_dev(le32_to_cpu(raw_inode
->i_block
[0])));
1511 init_special_inode(inode
, inode
->i_mode
,
1512 new_decode_dev(le32_to_cpu(raw_inode
->i_block
[1])));
1515 ext2_set_inode_flags(inode
);
1516 unlock_new_inode(inode
);
1521 return ERR_PTR(ret
);
1524 static int __ext2_write_inode(struct inode
*inode
, int do_sync
)
1526 struct ext2_inode_info
*ei
= EXT2_I(inode
);
1527 struct super_block
*sb
= inode
->i_sb
;
1528 ino_t ino
= inode
->i_ino
;
1529 uid_t uid
= i_uid_read(inode
);
1530 gid_t gid
= i_gid_read(inode
);
1531 struct buffer_head
* bh
;
1532 struct ext2_inode
* raw_inode
= ext2_get_inode(sb
, ino
, &bh
);
1536 if (IS_ERR(raw_inode
))
1539 /* For fields not not tracking in the in-memory inode,
1540 * initialise them to zero for new inodes. */
1541 if (ei
->i_state
& EXT2_STATE_NEW
)
1542 memset(raw_inode
, 0, EXT2_SB(sb
)->s_inode_size
);
1544 raw_inode
->i_mode
= cpu_to_le16(inode
->i_mode
);
1545 if (!(test_opt(sb
, NO_UID32
))) {
1546 raw_inode
->i_uid_low
= cpu_to_le16(low_16_bits(uid
));
1547 raw_inode
->i_gid_low
= cpu_to_le16(low_16_bits(gid
));
1549 * Fix up interoperability with old kernels. Otherwise, old inodes get
1550 * re-used with the upper 16 bits of the uid/gid intact
1553 raw_inode
->i_uid_high
= cpu_to_le16(high_16_bits(uid
));
1554 raw_inode
->i_gid_high
= cpu_to_le16(high_16_bits(gid
));
1556 raw_inode
->i_uid_high
= 0;
1557 raw_inode
->i_gid_high
= 0;
1560 raw_inode
->i_uid_low
= cpu_to_le16(fs_high2lowuid(uid
));
1561 raw_inode
->i_gid_low
= cpu_to_le16(fs_high2lowgid(gid
));
1562 raw_inode
->i_uid_high
= 0;
1563 raw_inode
->i_gid_high
= 0;
1565 raw_inode
->i_links_count
= cpu_to_le16(inode
->i_nlink
);
1566 raw_inode
->i_size
= cpu_to_le32(inode
->i_size
);
1567 raw_inode
->i_atime
= cpu_to_le32(inode
->i_atime
.tv_sec
);
1568 raw_inode
->i_ctime
= cpu_to_le32(inode
->i_ctime
.tv_sec
);
1569 raw_inode
->i_mtime
= cpu_to_le32(inode
->i_mtime
.tv_sec
);
1571 raw_inode
->i_blocks
= cpu_to_le32(inode
->i_blocks
);
1572 raw_inode
->i_dtime
= cpu_to_le32(ei
->i_dtime
);
1573 raw_inode
->i_flags
= cpu_to_le32(ei
->i_flags
);
1574 raw_inode
->i_faddr
= cpu_to_le32(ei
->i_faddr
);
1575 raw_inode
->i_frag
= ei
->i_frag_no
;
1576 raw_inode
->i_fsize
= ei
->i_frag_size
;
1577 raw_inode
->i_file_acl
= cpu_to_le32(ei
->i_file_acl
);
1578 if (!S_ISREG(inode
->i_mode
))
1579 raw_inode
->i_dir_acl
= cpu_to_le32(ei
->i_dir_acl
);
1581 raw_inode
->i_size_high
= cpu_to_le32(inode
->i_size
>> 32);
1582 if (inode
->i_size
> 0x7fffffffULL
) {
1583 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb
,
1584 EXT2_FEATURE_RO_COMPAT_LARGE_FILE
) ||
1585 EXT2_SB(sb
)->s_es
->s_rev_level
==
1586 cpu_to_le32(EXT2_GOOD_OLD_REV
)) {
1587 /* If this is the first large file
1588 * created, add a flag to the superblock.
1590 spin_lock(&EXT2_SB(sb
)->s_lock
);
1591 ext2_update_dynamic_rev(sb
);
1592 EXT2_SET_RO_COMPAT_FEATURE(sb
,
1593 EXT2_FEATURE_RO_COMPAT_LARGE_FILE
);
1594 spin_unlock(&EXT2_SB(sb
)->s_lock
);
1595 ext2_sync_super(sb
, EXT2_SB(sb
)->s_es
, 1);
1600 raw_inode
->i_generation
= cpu_to_le32(inode
->i_generation
);
1601 if (S_ISCHR(inode
->i_mode
) || S_ISBLK(inode
->i_mode
)) {
1602 if (old_valid_dev(inode
->i_rdev
)) {
1603 raw_inode
->i_block
[0] =
1604 cpu_to_le32(old_encode_dev(inode
->i_rdev
));
1605 raw_inode
->i_block
[1] = 0;
1607 raw_inode
->i_block
[0] = 0;
1608 raw_inode
->i_block
[1] =
1609 cpu_to_le32(new_encode_dev(inode
->i_rdev
));
1610 raw_inode
->i_block
[2] = 0;
1612 } else for (n
= 0; n
< EXT2_N_BLOCKS
; n
++)
1613 raw_inode
->i_block
[n
] = ei
->i_data
[n
];
1614 mark_buffer_dirty(bh
);
1616 sync_dirty_buffer(bh
);
1617 if (buffer_req(bh
) && !buffer_uptodate(bh
)) {
1618 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1619 sb
->s_id
, (unsigned long) ino
);
1623 ei
->i_state
&= ~EXT2_STATE_NEW
;
1628 int ext2_write_inode(struct inode
*inode
, struct writeback_control
*wbc
)
1630 return __ext2_write_inode(inode
, wbc
->sync_mode
== WB_SYNC_ALL
);
1633 int ext2_setattr(struct dentry
*dentry
, struct iattr
*iattr
)
1635 struct inode
*inode
= d_inode(dentry
);
1638 error
= setattr_prepare(dentry
, iattr
);
1642 if (is_quota_modification(inode
, iattr
)) {
1643 error
= dquot_initialize(inode
);
1647 if ((iattr
->ia_valid
& ATTR_UID
&& !uid_eq(iattr
->ia_uid
, inode
->i_uid
)) ||
1648 (iattr
->ia_valid
& ATTR_GID
&& !gid_eq(iattr
->ia_gid
, inode
->i_gid
))) {
1649 error
= dquot_transfer(inode
, iattr
);
1653 if (iattr
->ia_valid
& ATTR_SIZE
&& iattr
->ia_size
!= inode
->i_size
) {
1654 error
= ext2_setsize(inode
, iattr
->ia_size
);
1658 setattr_copy(inode
, iattr
);
1659 if (iattr
->ia_valid
& ATTR_MODE
)
1660 error
= posix_acl_chmod(inode
, inode
->i_mode
);
1661 mark_inode_dirty(inode
);