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>
34 #include <linux/vserver/xid.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);
56 static void ext2_truncate_nocheck (struct inode * inode);
59 * Called at each iput().
61 * The inode may be "bad" if ext2_read_inode() saw an error from
62 * ext2_get_inode(), so we need to check that to avoid freeing random disk
65 void ext2_put_inode(struct inode *inode)
67 if (!is_bad_inode(inode))
68 ext2_discard_prealloc(inode);
71 static void ext2_truncate_nocheck (struct inode * inode);
74 * Called at the last iput() if i_nlink is zero.
76 void ext2_delete_inode (struct inode * inode)
78 if (is_bad_inode(inode))
80 EXT2_I(inode)->i_dtime = get_seconds();
81 mark_inode_dirty(inode);
82 ext2_update_inode(inode, inode_needs_sync(inode));
86 ext2_truncate_nocheck(inode);
87 ext2_free_inode (inode);
91 clear_inode(inode); /* We must guarantee clearing of inode... */
94 void ext2_discard_prealloc (struct inode * inode)
96 #ifdef EXT2_PREALLOCATE
97 struct ext2_inode_info *ei = EXT2_I(inode);
98 write_lock(&ei->i_meta_lock);
99 if (ei->i_prealloc_count) {
100 unsigned short total = ei->i_prealloc_count;
101 unsigned long block = ei->i_prealloc_block;
102 ei->i_prealloc_count = 0;
103 ei->i_prealloc_block = 0;
104 write_unlock(&ei->i_meta_lock);
105 ext2_free_blocks (inode, block, total);
108 write_unlock(&ei->i_meta_lock);
112 static int ext2_alloc_block (struct inode * inode, unsigned long goal, int *err)
115 static unsigned long alloc_hits, alloc_attempts;
117 unsigned long result;
120 #ifdef EXT2_PREALLOCATE
121 struct ext2_inode_info *ei = EXT2_I(inode);
122 write_lock(&ei->i_meta_lock);
123 if (ei->i_prealloc_count &&
124 (goal == ei->i_prealloc_block || goal + 1 == ei->i_prealloc_block))
126 result = ei->i_prealloc_block++;
127 ei->i_prealloc_count--;
128 write_unlock(&ei->i_meta_lock);
129 ext2_debug ("preallocation hit (%lu/%lu).\n",
130 ++alloc_hits, ++alloc_attempts);
132 write_unlock(&ei->i_meta_lock);
133 ext2_discard_prealloc (inode);
134 ext2_debug ("preallocation miss (%lu/%lu).\n",
135 alloc_hits, ++alloc_attempts);
136 if (S_ISREG(inode->i_mode))
137 result = ext2_new_block (inode, goal,
138 &ei->i_prealloc_count,
139 &ei->i_prealloc_block, err);
141 result = ext2_new_block(inode, goal, NULL, NULL, err);
144 result = ext2_new_block (inode, goal, 0, 0, err);
152 struct buffer_head *bh;
155 static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v)
157 p->key = *(p->p = v);
161 static inline int verify_chain(Indirect *from, Indirect *to)
163 while (from <= to && from->key == *from->p)
169 * ext2_block_to_path - parse the block number into array of offsets
170 * @inode: inode in question (we are only interested in its superblock)
171 * @i_block: block number to be parsed
172 * @offsets: array to store the offsets in
173 * @boundary: set this non-zero if the referred-to block is likely to be
174 * followed (on disk) by an indirect block.
175 * To store the locations of file's data ext2 uses a data structure common
176 * for UNIX filesystems - tree of pointers anchored in the inode, with
177 * data blocks at leaves and indirect blocks in intermediate nodes.
178 * This function translates the block number into path in that tree -
179 * return value is the path length and @offsets[n] is the offset of
180 * pointer to (n+1)th node in the nth one. If @block is out of range
181 * (negative or too large) warning is printed and zero returned.
183 * Note: function doesn't find node addresses, so no IO is needed. All
184 * we need to know is the capacity of indirect blocks (taken from the
189 * Portability note: the last comparison (check that we fit into triple
190 * indirect block) is spelled differently, because otherwise on an
191 * architecture with 32-bit longs and 8Kb pages we might get into trouble
192 * if our filesystem had 8Kb blocks. We might use long long, but that would
193 * kill us on x86. Oh, well, at least the sign propagation does not matter -
194 * i_block would have to be negative in the very beginning, so we would not
198 static int ext2_block_to_path(struct inode *inode,
199 long i_block, int offsets[4], int *boundary)
201 int ptrs = EXT2_ADDR_PER_BLOCK(inode->i_sb);
202 int ptrs_bits = EXT2_ADDR_PER_BLOCK_BITS(inode->i_sb);
203 const long direct_blocks = EXT2_NDIR_BLOCKS,
204 indirect_blocks = ptrs,
205 double_blocks = (1 << (ptrs_bits * 2));
210 ext2_warning (inode->i_sb, "ext2_block_to_path", "block < 0");
211 } else if (i_block < direct_blocks) {
212 offsets[n++] = i_block;
213 final = direct_blocks;
214 } else if ( (i_block -= direct_blocks) < indirect_blocks) {
215 offsets[n++] = EXT2_IND_BLOCK;
216 offsets[n++] = i_block;
218 } else if ((i_block -= indirect_blocks) < double_blocks) {
219 offsets[n++] = EXT2_DIND_BLOCK;
220 offsets[n++] = i_block >> ptrs_bits;
221 offsets[n++] = i_block & (ptrs - 1);
223 } else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
224 offsets[n++] = EXT2_TIND_BLOCK;
225 offsets[n++] = i_block >> (ptrs_bits * 2);
226 offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
227 offsets[n++] = i_block & (ptrs - 1);
230 ext2_warning (inode->i_sb, "ext2_block_to_path", "block > big");
233 *boundary = (i_block & (ptrs - 1)) == (final - 1);
238 * ext2_get_branch - read the chain of indirect blocks leading to data
239 * @inode: inode in question
240 * @depth: depth of the chain (1 - direct pointer, etc.)
241 * @offsets: offsets of pointers in inode/indirect blocks
242 * @chain: place to store the result
243 * @err: here we store the error value
245 * Function fills the array of triples <key, p, bh> and returns %NULL
246 * if everything went OK or the pointer to the last filled triple
247 * (incomplete one) otherwise. Upon the return chain[i].key contains
248 * the number of (i+1)-th block in the chain (as it is stored in memory,
249 * i.e. little-endian 32-bit), chain[i].p contains the address of that
250 * number (it points into struct inode for i==0 and into the bh->b_data
251 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect
252 * block for i>0 and NULL for i==0. In other words, it holds the block
253 * numbers of the chain, addresses they were taken from (and where we can
254 * verify that chain did not change) and buffer_heads hosting these
257 * Function stops when it stumbles upon zero pointer (absent block)
258 * (pointer to last triple returned, *@err == 0)
259 * or when it gets an IO error reading an indirect block
260 * (ditto, *@err == -EIO)
261 * or when it notices that chain had been changed while it was reading
262 * (ditto, *@err == -EAGAIN)
263 * or when it reads all @depth-1 indirect blocks successfully and finds
264 * the whole chain, all way to the data (returns %NULL, *err == 0).
266 static Indirect *ext2_get_branch(struct inode *inode,
272 struct super_block *sb = inode->i_sb;
274 struct buffer_head *bh;
277 /* i_data is not going away, no lock needed */
278 add_chain (chain, NULL, EXT2_I(inode)->i_data + *offsets);
282 bh = sb_bread(sb, le32_to_cpu(p->key));
285 read_lock(&EXT2_I(inode)->i_meta_lock);
286 if (!verify_chain(chain, p))
288 add_chain(++p, bh, (__le32*)bh->b_data + *++offsets);
289 read_unlock(&EXT2_I(inode)->i_meta_lock);
296 read_unlock(&EXT2_I(inode)->i_meta_lock);
307 * ext2_find_near - find a place for allocation with sufficient locality
309 * @ind: descriptor of indirect block.
311 * This function returns the prefered place for block allocation.
312 * It is used when heuristic for sequential allocation fails.
314 * + if there is a block to the left of our position - allocate near it.
315 * + if pointer will live in indirect block - allocate near that block.
316 * + if pointer will live in inode - allocate in the same cylinder group.
318 * In the latter case we colour the starting block by the callers PID to
319 * prevent it from clashing with concurrent allocations for a different inode
320 * in the same block group. The PID is used here so that functionally related
321 * files will be close-by on-disk.
323 * Caller must make sure that @ind is valid and will stay that way.
326 static unsigned long ext2_find_near(struct inode *inode, Indirect *ind)
328 struct ext2_inode_info *ei = EXT2_I(inode);
329 __le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data;
331 unsigned long bg_start;
332 unsigned long colour;
334 /* Try to find previous block */
335 for (p = ind->p - 1; p >= start; p--)
337 return le32_to_cpu(*p);
339 /* No such thing, so let's try location of indirect block */
341 return ind->bh->b_blocknr;
344 * It is going to be refered from inode itself? OK, just put it into
345 * the same cylinder group then.
347 bg_start = (ei->i_block_group * EXT2_BLOCKS_PER_GROUP(inode->i_sb)) +
348 le32_to_cpu(EXT2_SB(inode->i_sb)->s_es->s_first_data_block);
349 colour = (current->pid % 16) *
350 (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16);
351 return bg_start + colour;
355 * ext2_find_goal - find a prefered place for allocation.
357 * @block: block we want
358 * @chain: chain of indirect blocks
359 * @partial: pointer to the last triple within a chain
360 * @goal: place to store the result.
362 * Normally this function find the prefered place for block allocation,
363 * stores it in *@goal and returns zero. If the branch had been changed
364 * under us we return -EAGAIN.
367 static inline int ext2_find_goal(struct inode *inode,
373 struct ext2_inode_info *ei = EXT2_I(inode);
374 write_lock(&ei->i_meta_lock);
375 if ((block == ei->i_next_alloc_block + 1) && ei->i_next_alloc_goal) {
376 ei->i_next_alloc_block++;
377 ei->i_next_alloc_goal++;
379 if (verify_chain(chain, partial)) {
381 * try the heuristic for sequential allocation,
382 * failing that at least try to get decent locality.
384 if (block == ei->i_next_alloc_block)
385 *goal = ei->i_next_alloc_goal;
387 *goal = ext2_find_near(inode, partial);
388 write_unlock(&ei->i_meta_lock);
391 write_unlock(&ei->i_meta_lock);
396 * ext2_alloc_branch - allocate and set up a chain of blocks.
398 * @num: depth of the chain (number of blocks to allocate)
399 * @offsets: offsets (in the blocks) to store the pointers to next.
400 * @branch: place to store the chain in.
402 * This function allocates @num blocks, zeroes out all but the last one,
403 * links them into chain and (if we are synchronous) writes them to disk.
404 * In other words, it prepares a branch that can be spliced onto the
405 * inode. It stores the information about that chain in the branch[], in
406 * the same format as ext2_get_branch() would do. We are calling it after
407 * we had read the existing part of chain and partial points to the last
408 * triple of that (one with zero ->key). Upon the exit we have the same
409 * picture as after the successful ext2_get_block(), excpet that in one
410 * place chain is disconnected - *branch->p is still zero (we did not
411 * set the last link), but branch->key contains the number that should
412 * be placed into *branch->p to fill that gap.
414 * If allocation fails we free all blocks we've allocated (and forget
415 * their buffer_heads) and return the error value the from failed
416 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
417 * as described above and return 0.
420 static int ext2_alloc_branch(struct inode *inode,
426 int blocksize = inode->i_sb->s_blocksize;
430 int parent = ext2_alloc_block(inode, goal, &err);
432 branch[0].key = cpu_to_le32(parent);
433 if (parent) for (n = 1; n < num; n++) {
434 struct buffer_head *bh;
435 /* Allocate the next block */
436 int nr = ext2_alloc_block(inode, parent, &err);
439 branch[n].key = cpu_to_le32(nr);
441 * Get buffer_head for parent block, zero it out and set
442 * the pointer to new one, then send parent to disk.
444 bh = sb_getblk(inode->i_sb, parent);
446 memset(bh->b_data, 0, blocksize);
448 branch[n].p = (__le32 *) bh->b_data + offsets[n];
449 *branch[n].p = branch[n].key;
450 set_buffer_uptodate(bh);
452 mark_buffer_dirty_inode(bh, inode);
453 /* We used to sync bh here if IS_SYNC(inode).
454 * But we now rely upon generic_osync_inode()
455 * and b_inode_buffers. But not for directories.
457 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
458 sync_dirty_buffer(bh);
464 /* Allocation failed, free what we already allocated */
465 for (i = 1; i < n; i++)
466 bforget(branch[i].bh);
467 for (i = 0; i < n; i++)
468 ext2_free_blocks(inode, le32_to_cpu(branch[i].key), 1);
473 * ext2_splice_branch - splice the allocated branch onto inode.
475 * @block: (logical) number of block we are adding
476 * @chain: chain of indirect blocks (with a missing link - see
478 * @where: location of missing link
479 * @num: number of blocks we are adding
481 * This function verifies that chain (up to the missing link) had not
482 * changed, fills the missing link and does all housekeeping needed in
483 * inode (->i_blocks, etc.). In case of success we end up with the full
484 * chain to new block and return 0. Otherwise (== chain had been changed)
485 * we free the new blocks (forgetting their buffer_heads, indeed) and
489 static inline int ext2_splice_branch(struct inode *inode,
495 struct ext2_inode_info *ei = EXT2_I(inode);
498 /* Verify that place we are splicing to is still there and vacant */
500 write_lock(&ei->i_meta_lock);
501 if (!verify_chain(chain, where-1) || *where->p)
506 *where->p = where->key;
507 ei->i_next_alloc_block = block;
508 ei->i_next_alloc_goal = le32_to_cpu(where[num-1].key);
510 write_unlock(&ei->i_meta_lock);
512 /* We are done with atomic stuff, now do the rest of housekeeping */
514 inode->i_ctime = CURRENT_TIME_SEC;
516 /* had we spliced it onto indirect block? */
518 mark_buffer_dirty_inode(where->bh, inode);
520 mark_inode_dirty(inode);
524 write_unlock(&ei->i_meta_lock);
525 for (i = 1; i < num; i++)
526 bforget(where[i].bh);
527 for (i = 0; i < num; i++)
528 ext2_free_blocks(inode, le32_to_cpu(where[i].key), 1);
533 * Allocation strategy is simple: if we have to allocate something, we will
534 * have to go the whole way to leaf. So let's do it before attaching anything
535 * to tree, set linkage between the newborn blocks, write them if sync is
536 * required, recheck the path, free and repeat if check fails, otherwise
537 * set the last missing link (that will protect us from any truncate-generated
538 * removals - all blocks on the path are immune now) and possibly force the
539 * write on the parent block.
540 * That has a nice additional property: no special recovery from the failed
541 * allocations is needed - we simply release blocks and do not touch anything
542 * reachable from inode.
545 int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
554 int depth = ext2_block_to_path(inode, iblock, offsets, &boundary);
560 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
562 /* Simplest case - block found, no allocation needed */
565 map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
567 set_buffer_boundary(bh_result);
568 /* Clean up and exit */
569 partial = chain+depth-1; /* the whole chain */
573 /* Next simple case - plain lookup or failed read of indirect block */
574 if (!create || err == -EIO) {
576 while (partial > chain) {
585 * Indirect block might be removed by truncate while we were
586 * reading it. Handling of that case (forget what we've got and
587 * reread) is taken out of the main path.
593 if (ext2_find_goal(inode, iblock, chain, partial, &goal) < 0)
596 left = (chain + depth) - partial;
597 err = ext2_alloc_branch(inode, left, goal,
598 offsets+(partial-chain), partial);
602 if (ext2_splice_branch(inode, iblock, chain, partial, left) < 0)
605 set_buffer_new(bh_result);
609 while (partial > chain) {
616 static int ext2_writepage(struct page *page, struct writeback_control *wbc)
618 return block_write_full_page(page, ext2_get_block, wbc);
621 static int ext2_readpage(struct file *file, struct page *page)
623 return mpage_readpage(page, ext2_get_block);
627 ext2_readpages(struct file *file, struct address_space *mapping,
628 struct list_head *pages, unsigned nr_pages)
630 return mpage_readpages(mapping, pages, nr_pages, ext2_get_block);
634 ext2_prepare_write(struct file *file, struct page *page,
635 unsigned from, unsigned to)
637 return block_prepare_write(page,from,to,ext2_get_block);
641 ext2_nobh_prepare_write(struct file *file, struct page *page,
642 unsigned from, unsigned to)
644 return nobh_prepare_write(page,from,to,ext2_get_block);
647 static int ext2_nobh_writepage(struct page *page,
648 struct writeback_control *wbc)
650 return nobh_writepage(page, ext2_get_block, wbc);
653 static sector_t ext2_bmap(struct address_space *mapping, sector_t block)
655 return generic_block_bmap(mapping,block,ext2_get_block);
659 ext2_get_blocks(struct inode *inode, sector_t iblock, unsigned long max_blocks,
660 struct buffer_head *bh_result, int create)
664 ret = ext2_get_block(inode, iblock, bh_result, create);
666 bh_result->b_size = (1 << inode->i_blkbits);
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_blocks, NULL);
682 ext2_writepages(struct address_space *mapping, struct writeback_control *wbc)
684 return mpage_writepages(mapping, wbc, ext2_get_block);
687 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,
699 struct address_space_operations ext2_nobh_aops = {
700 .readpage = ext2_readpage,
701 .readpages = ext2_readpages,
702 .writepage = ext2_nobh_writepage,
703 .sync_page = block_sync_page,
704 .prepare_write = ext2_nobh_prepare_write,
705 .commit_write = nobh_commit_write,
707 .direct_IO = ext2_direct_IO,
708 .writepages = ext2_writepages,
712 * Probably it should be a library function... search for first non-zero word
713 * or memcmp with zero_page, whatever is better for particular architecture.
716 static inline int all_zeroes(__le32 *p, __le32 *q)
725 * ext2_find_shared - find the indirect blocks for partial truncation.
726 * @inode: inode in question
727 * @depth: depth of the affected branch
728 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
729 * @chain: place to store the pointers to partial indirect blocks
730 * @top: place to the (detached) top of branch
732 * This is a helper function used by ext2_truncate().
734 * When we do truncate() we may have to clean the ends of several indirect
735 * blocks but leave the blocks themselves alive. Block is partially
736 * truncated if some data below the new i_size is refered from it (and
737 * it is on the path to the first completely truncated data block, indeed).
738 * We have to free the top of that path along with everything to the right
739 * of the path. Since no allocation past the truncation point is possible
740 * until ext2_truncate() finishes, we may safely do the latter, but top
741 * of branch may require special attention - pageout below the truncation
742 * point might try to populate it.
744 * We atomically detach the top of branch from the tree, store the block
745 * number of its root in *@top, pointers to buffer_heads of partially
746 * truncated blocks - in @chain[].bh and pointers to their last elements
747 * that should not be removed - in @chain[].p. Return value is the pointer
748 * to last filled element of @chain.
750 * The work left to caller to do the actual freeing of subtrees:
751 * a) free the subtree starting from *@top
752 * b) free the subtrees whose roots are stored in
753 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
754 * c) free the subtrees growing from the inode past the @chain[0].p
755 * (no partially truncated stuff there).
758 static Indirect *ext2_find_shared(struct inode *inode,
764 Indirect *partial, *p;
768 for (k = depth; k > 1 && !offsets[k-1]; k--)
770 partial = ext2_get_branch(inode, k, offsets, chain, &err);
772 partial = chain + k-1;
774 * If the branch acquired continuation since we've looked at it -
775 * fine, it should all survive and (new) top doesn't belong to us.
777 write_lock(&EXT2_I(inode)->i_meta_lock);
778 if (!partial->key && *partial->p) {
779 write_unlock(&EXT2_I(inode)->i_meta_lock);
782 for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--)
785 * OK, we've found the last block that must survive. The rest of our
786 * branch should be detached before unlocking. However, if that rest
787 * of branch is all ours and does not grow immediately from the inode
788 * it's easier to cheat and just decrement partial->p.
790 if (p == chain + k - 1 && p > chain) {
796 write_unlock(&EXT2_I(inode)->i_meta_lock);
808 * ext2_free_data - free a list of data blocks
809 * @inode: inode we are dealing with
810 * @p: array of block numbers
811 * @q: points immediately past the end of array
813 * We are freeing all blocks refered from that array (numbers are
814 * stored as little-endian 32-bit) and updating @inode->i_blocks
817 static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q)
819 unsigned long block_to_free = 0, count = 0;
822 for ( ; p < q ; p++) {
823 nr = le32_to_cpu(*p);
826 /* accumulate blocks to free if they're contiguous */
829 else if (block_to_free == nr - count)
832 mark_inode_dirty(inode);
833 ext2_free_blocks (inode, block_to_free, count);
841 mark_inode_dirty(inode);
842 ext2_free_blocks (inode, block_to_free, count);
847 * ext2_free_branches - free an array of branches
848 * @inode: inode we are dealing with
849 * @p: array of block numbers
850 * @q: pointer immediately past the end of array
851 * @depth: depth of the branches to free
853 * We are freeing all blocks refered from these branches (numbers are
854 * stored as little-endian 32-bit) and updating @inode->i_blocks
857 static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth)
859 struct buffer_head * bh;
863 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
864 for ( ; p < q ; p++) {
865 nr = le32_to_cpu(*p);
869 bh = sb_bread(inode->i_sb, nr);
871 * A read failure? Report error and clear slot
875 ext2_error(inode->i_sb, "ext2_free_branches",
876 "Read failure, inode=%ld, block=%ld",
880 ext2_free_branches(inode,
882 (__le32*)bh->b_data + addr_per_block,
885 ext2_free_blocks(inode, nr, 1);
886 mark_inode_dirty(inode);
889 ext2_free_data(inode, p, q);
892 static void ext2_truncate_nocheck(struct inode * inode)
894 __le32 *i_data = EXT2_I(inode)->i_data;
895 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
904 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
905 S_ISLNK(inode->i_mode)))
907 if (ext2_inode_is_fast_symlink(inode))
910 ext2_discard_prealloc(inode);
912 blocksize = inode->i_sb->s_blocksize;
913 iblock = (inode->i_size + blocksize-1)
914 >> EXT2_BLOCK_SIZE_BITS(inode->i_sb);
916 if (test_opt(inode->i_sb, NOBH))
917 nobh_truncate_page(inode->i_mapping, inode->i_size);
919 block_truncate_page(inode->i_mapping,
920 inode->i_size, ext2_get_block);
922 n = ext2_block_to_path(inode, iblock, offsets, NULL);
927 ext2_free_data(inode, i_data+offsets[0],
928 i_data + EXT2_NDIR_BLOCKS);
932 partial = ext2_find_shared(inode, n, offsets, chain, &nr);
933 /* Kill the top of shared branch (already detached) */
935 if (partial == chain)
936 mark_inode_dirty(inode);
938 mark_buffer_dirty_inode(partial->bh, inode);
939 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
941 /* Clear the ends of indirect blocks on the shared branch */
942 while (partial > chain) {
943 ext2_free_branches(inode,
945 (__le32*)partial->bh->b_data+addr_per_block,
946 (chain+n-1) - partial);
947 mark_buffer_dirty_inode(partial->bh, inode);
948 brelse (partial->bh);
952 /* Kill the remaining (whole) subtrees */
953 switch (offsets[0]) {
955 nr = i_data[EXT2_IND_BLOCK];
957 i_data[EXT2_IND_BLOCK] = 0;
958 mark_inode_dirty(inode);
959 ext2_free_branches(inode, &nr, &nr+1, 1);
962 nr = i_data[EXT2_DIND_BLOCK];
964 i_data[EXT2_DIND_BLOCK] = 0;
965 mark_inode_dirty(inode);
966 ext2_free_branches(inode, &nr, &nr+1, 2);
968 case EXT2_DIND_BLOCK:
969 nr = i_data[EXT2_TIND_BLOCK];
971 i_data[EXT2_TIND_BLOCK] = 0;
972 mark_inode_dirty(inode);
973 ext2_free_branches(inode, &nr, &nr+1, 3);
975 case EXT2_TIND_BLOCK:
978 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
979 if (inode_needs_sync(inode)) {
980 sync_mapping_buffers(inode->i_mapping);
981 ext2_sync_inode (inode);
983 mark_inode_dirty(inode);
987 static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
988 struct buffer_head **p)
990 struct buffer_head * bh;
991 unsigned long block_group;
993 unsigned long offset;
994 struct ext2_group_desc * gdp;
997 if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) ||
998 ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
1001 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
1002 gdp = ext2_get_group_desc(sb, block_group, &bh);
1006 * Figure out the offset within the block group inode table
1008 offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb);
1009 block = le32_to_cpu(gdp->bg_inode_table) +
1010 (offset >> EXT2_BLOCK_SIZE_BITS(sb));
1011 if (!(bh = sb_bread(sb, block)))
1015 offset &= (EXT2_BLOCK_SIZE(sb) - 1);
1016 return (struct ext2_inode *) (bh->b_data + offset);
1019 ext2_error(sb, "ext2_get_inode", "bad inode number: %lu",
1020 (unsigned long) ino);
1021 return ERR_PTR(-EINVAL);
1023 ext2_error(sb, "ext2_get_inode",
1024 "unable to read inode block - inode=%lu, block=%lu",
1025 (unsigned long) ino, block);
1027 return ERR_PTR(-EIO);
1030 void ext2_truncate (struct inode * inode)
1032 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1034 ext2_truncate_nocheck(inode);
1037 void ext2_set_inode_flags(struct inode *inode)
1039 unsigned int flags = EXT2_I(inode)->i_flags;
1041 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_IUNLINK|S_BARRIER|S_NOATIME|S_DIRSYNC);
1042 if (flags & EXT2_SYNC_FL)
1043 inode->i_flags |= S_SYNC;
1044 if (flags & EXT2_APPEND_FL)
1045 inode->i_flags |= S_APPEND;
1046 if (flags & EXT2_IMMUTABLE_FL)
1047 inode->i_flags |= S_IMMUTABLE;
1048 if (flags & EXT2_IUNLINK_FL)
1049 inode->i_flags |= S_IUNLINK;
1050 if (flags & EXT2_BARRIER_FL)
1051 inode->i_flags |= S_BARRIER;
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);
1068 #ifdef CONFIG_EXT2_FS_POSIX_ACL
1069 ei->i_acl = EXT2_ACL_NOT_CACHED;
1070 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
1072 if (IS_ERR(raw_inode))
1075 inode->i_mode = le16_to_cpu(raw_inode->i_mode);
1076 uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
1077 gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
1078 if (!(test_opt (inode->i_sb, NO_UID32))) {
1079 uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
1080 gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
1082 inode->i_uid = INOXID_UID(XID_TAG(inode), uid, gid);
1083 inode->i_gid = INOXID_GID(XID_TAG(inode), uid, gid);
1084 inode->i_xid = INOXID_XID(XID_TAG(inode), uid, gid,
1085 le16_to_cpu(raw_inode->i_raw_xid));
1087 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
1088 inode->i_size = le32_to_cpu(raw_inode->i_size);
1089 inode->i_atime.tv_sec = le32_to_cpu(raw_inode->i_atime);
1090 inode->i_ctime.tv_sec = le32_to_cpu(raw_inode->i_ctime);
1091 inode->i_mtime.tv_sec = le32_to_cpu(raw_inode->i_mtime);
1092 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1093 ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
1094 /* We now have enough fields to check if the inode was active or not.
1095 * This is needed because nfsd might try to access dead inodes
1096 * the test is that same one that e2fsck uses
1097 * NeilBrown 1999oct15
1099 if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) {
1100 /* this inode is deleted */
1104 inode->i_blksize = PAGE_SIZE; /* This is the optimal IO size (for stat), not the fs block size */
1105 inode->i_blocks = le32_to_cpu(raw_inode->i_blocks);
1106 ei->i_flags = le32_to_cpu(raw_inode->i_flags);
1107 ei->i_faddr = le32_to_cpu(raw_inode->i_faddr);
1108 ei->i_frag_no = raw_inode->i_frag;
1109 ei->i_frag_size = raw_inode->i_fsize;
1110 ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl);
1112 if (S_ISREG(inode->i_mode))
1113 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32;
1115 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl);
1117 inode->i_generation = le32_to_cpu(raw_inode->i_generation);
1119 ei->i_next_alloc_block = 0;
1120 ei->i_next_alloc_goal = 0;
1121 ei->i_prealloc_count = 0;
1122 ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb);
1123 ei->i_dir_start_lookup = 0;
1126 * NOTE! The in-memory inode i_data array is in little-endian order
1127 * even on big-endian machines: we do NOT byteswap the block numbers!
1129 for (n = 0; n < EXT2_N_BLOCKS; n++)
1130 ei->i_data[n] = raw_inode->i_block[n];
1132 if (S_ISREG(inode->i_mode)) {
1133 inode->i_op = &ext2_file_inode_operations;
1134 inode->i_fop = &ext2_file_operations;
1135 if (test_opt(inode->i_sb, NOBH))
1136 inode->i_mapping->a_ops = &ext2_nobh_aops;
1138 inode->i_mapping->a_ops = &ext2_aops;
1139 } else if (S_ISDIR(inode->i_mode)) {
1140 inode->i_op = &ext2_dir_inode_operations;
1141 inode->i_fop = &ext2_dir_operations;
1142 if (test_opt(inode->i_sb, NOBH))
1143 inode->i_mapping->a_ops = &ext2_nobh_aops;
1145 inode->i_mapping->a_ops = &ext2_aops;
1146 } else if (S_ISLNK(inode->i_mode)) {
1147 if (ext2_inode_is_fast_symlink(inode))
1148 inode->i_op = &ext2_fast_symlink_inode_operations;
1150 inode->i_op = &ext2_symlink_inode_operations;
1151 if (test_opt(inode->i_sb, NOBH))
1152 inode->i_mapping->a_ops = &ext2_nobh_aops;
1154 inode->i_mapping->a_ops = &ext2_aops;
1157 inode->i_op = &ext2_special_inode_operations;
1158 if (raw_inode->i_block[0])
1159 init_special_inode(inode, inode->i_mode,
1160 old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
1162 init_special_inode(inode, inode->i_mode,
1163 new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
1166 ext2_set_inode_flags(inode);
1170 make_bad_inode(inode);
1174 static int ext2_update_inode(struct inode * inode, int do_sync)
1176 struct ext2_inode_info *ei = EXT2_I(inode);
1177 struct super_block *sb = inode->i_sb;
1178 ino_t ino = inode->i_ino;
1179 uid_t uid = XIDINO_UID(XID_TAG(inode), inode->i_uid, inode->i_xid);
1180 gid_t gid = XIDINO_GID(XID_TAG(inode), inode->i_gid, inode->i_xid);
1181 struct buffer_head * bh;
1182 struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh);
1186 if (IS_ERR(raw_inode))
1189 /* For fields not not tracking in the in-memory inode,
1190 * initialise them to zero for new inodes. */
1191 if (ei->i_state & EXT2_STATE_NEW)
1192 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
1194 raw_inode->i_mode = cpu_to_le16(inode->i_mode);
1195 if (!(test_opt(sb, NO_UID32))) {
1196 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid));
1197 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid));
1199 * Fix up interoperability with old kernels. Otherwise, old inodes get
1200 * re-used with the upper 16 bits of the uid/gid intact
1203 raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid));
1204 raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid));
1206 raw_inode->i_uid_high = 0;
1207 raw_inode->i_gid_high = 0;
1210 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid));
1211 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid));
1212 raw_inode->i_uid_high = 0;
1213 raw_inode->i_gid_high = 0;
1215 #ifdef CONFIG_INOXID_INTERN
1216 raw_inode->i_raw_xid = cpu_to_le16(inode->i_xid);
1218 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
1219 raw_inode->i_size = cpu_to_le32(inode->i_size);
1220 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1221 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1222 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1224 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
1225 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
1226 raw_inode->i_flags = cpu_to_le32(ei->i_flags);
1227 raw_inode->i_faddr = cpu_to_le32(ei->i_faddr);
1228 raw_inode->i_frag = ei->i_frag_no;
1229 raw_inode->i_fsize = ei->i_frag_size;
1230 raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl);
1231 if (!S_ISREG(inode->i_mode))
1232 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl);
1234 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32);
1235 if (inode->i_size > 0x7fffffffULL) {
1236 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb,
1237 EXT2_FEATURE_RO_COMPAT_LARGE_FILE) ||
1238 EXT2_SB(sb)->s_es->s_rev_level ==
1239 cpu_to_le32(EXT2_GOOD_OLD_REV)) {
1240 /* If this is the first large file
1241 * created, add a flag to the superblock.
1244 ext2_update_dynamic_rev(sb);
1245 EXT2_SET_RO_COMPAT_FEATURE(sb,
1246 EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
1248 ext2_write_super(sb);
1253 raw_inode->i_generation = cpu_to_le32(inode->i_generation);
1254 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1255 if (old_valid_dev(inode->i_rdev)) {
1256 raw_inode->i_block[0] =
1257 cpu_to_le32(old_encode_dev(inode->i_rdev));
1258 raw_inode->i_block[1] = 0;
1260 raw_inode->i_block[0] = 0;
1261 raw_inode->i_block[1] =
1262 cpu_to_le32(new_encode_dev(inode->i_rdev));
1263 raw_inode->i_block[2] = 0;
1265 } else for (n = 0; n < EXT2_N_BLOCKS; n++)
1266 raw_inode->i_block[n] = ei->i_data[n];
1267 mark_buffer_dirty(bh);
1269 sync_dirty_buffer(bh);
1270 if (buffer_req(bh) && !buffer_uptodate(bh)) {
1271 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1272 sb->s_id, (unsigned long) ino);
1276 ei->i_state &= ~EXT2_STATE_NEW;
1281 int ext2_write_inode(struct inode *inode, int wait)
1283 return ext2_update_inode(inode, wait);
1286 int ext2_sync_inode(struct inode *inode)
1288 struct writeback_control wbc = {
1289 .sync_mode = WB_SYNC_ALL,
1290 .nr_to_write = 0, /* sys_fsync did this */
1292 return sync_inode(inode, &wbc);
1295 int ext2_setattr_flags(struct inode *inode, unsigned int flags)
1297 unsigned int oldflags, newflags;
1299 oldflags = EXT2_I(inode)->i_flags;
1300 newflags = oldflags &
1301 ~(EXT2_IMMUTABLE_FL | EXT2_IUNLINK_FL | EXT2_BARRIER_FL);
1302 if (flags & ATTR_FLAG_IMMUTABLE)
1303 newflags |= EXT2_IMMUTABLE_FL;
1304 if (flags & ATTR_FLAG_IUNLINK)
1305 newflags |= EXT2_IUNLINK_FL;
1306 if (flags & ATTR_FLAG_BARRIER)
1307 newflags |= EXT2_BARRIER_FL;
1309 if (oldflags ^ newflags) {
1310 EXT2_I(inode)->i_flags = newflags;
1311 inode->i_ctime = CURRENT_TIME;
1316 int ext2_setattr(struct dentry *dentry, struct iattr *iattr)
1318 struct inode *inode = dentry->d_inode;
1321 error = inode_change_ok(inode, iattr);
1324 if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) ||
1325 (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid) ||
1326 (iattr->ia_valid & ATTR_XID && iattr->ia_xid != inode->i_xid)) {
1327 error = DQUOT_TRANSFER(inode, iattr) ? -EDQUOT : 0;
1331 if (iattr->ia_valid & ATTR_ATTR_FLAG)
1332 ext2_setattr_flags(inode, iattr->ia_attr_flags);
1334 error = inode_setattr(inode, iattr);
1335 if (!error && (iattr->ia_valid & ATTR_MODE))
1336 error = ext2_acl_chmod(inode);