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>
37 MODULE_AUTHOR("Remy Card and others");
38 MODULE_DESCRIPTION("Second Extended Filesystem");
39 MODULE_LICENSE("GPL");
41 static int ext2_update_inode(struct inode * inode, int do_sync);
44 * Test whether an inode is a fast symlink.
46 static inline int ext2_inode_is_fast_symlink(struct inode *inode)
48 int ea_blocks = EXT2_I(inode)->i_file_acl ?
49 (inode->i_sb->s_blocksize >> 9) : 0;
51 return (S_ISLNK(inode->i_mode) &&
52 inode->i_blocks - ea_blocks == 0);
56 * Called at each iput().
58 * The inode may be "bad" if ext2_read_inode() saw an error from
59 * ext2_get_inode(), so we need to check that to avoid freeing random disk
62 void ext2_put_inode(struct inode *inode)
64 if (!is_bad_inode(inode))
65 ext2_discard_prealloc(inode);
69 * Called at the last iput() if i_nlink is zero.
71 void ext2_delete_inode (struct inode * inode)
73 if (is_bad_inode(inode))
75 EXT2_I(inode)->i_dtime = get_seconds();
76 mark_inode_dirty(inode);
77 ext2_update_inode(inode, inode_needs_sync(inode));
81 ext2_truncate (inode);
82 ext2_free_inode (inode);
86 clear_inode(inode); /* We must guarantee clearing of inode... */
89 void ext2_discard_prealloc (struct inode * inode)
91 #ifdef EXT2_PREALLOCATE
92 struct ext2_inode_info *ei = EXT2_I(inode);
93 write_lock(&ei->i_meta_lock);
94 if (ei->i_prealloc_count) {
95 unsigned short total = ei->i_prealloc_count;
96 unsigned long block = ei->i_prealloc_block;
97 ei->i_prealloc_count = 0;
98 ei->i_prealloc_block = 0;
99 write_unlock(&ei->i_meta_lock);
100 ext2_free_blocks (inode, block, total);
103 write_unlock(&ei->i_meta_lock);
107 static int ext2_alloc_block (struct inode * inode, unsigned long goal, int *err)
110 static unsigned long alloc_hits, alloc_attempts;
112 unsigned long result;
115 #ifdef EXT2_PREALLOCATE
116 struct ext2_inode_info *ei = EXT2_I(inode);
117 write_lock(&ei->i_meta_lock);
118 if (ei->i_prealloc_count &&
119 (goal == ei->i_prealloc_block || goal + 1 == ei->i_prealloc_block))
121 result = ei->i_prealloc_block++;
122 ei->i_prealloc_count--;
123 write_unlock(&ei->i_meta_lock);
124 ext2_debug ("preallocation hit (%lu/%lu).\n",
125 ++alloc_hits, ++alloc_attempts);
127 write_unlock(&ei->i_meta_lock);
128 ext2_discard_prealloc (inode);
129 ext2_debug ("preallocation miss (%lu/%lu).\n",
130 alloc_hits, ++alloc_attempts);
131 if (S_ISREG(inode->i_mode))
132 result = ext2_new_block (inode, goal,
133 &ei->i_prealloc_count,
134 &ei->i_prealloc_block, err);
136 result = ext2_new_block (inode, goal, 0, 0, err);
139 result = ext2_new_block (inode, goal, 0, 0, err);
147 struct buffer_head *bh;
150 static inline void add_chain(Indirect *p, struct buffer_head *bh, u32 *v)
152 p->key = *(p->p = v);
156 static inline int verify_chain(Indirect *from, Indirect *to)
158 while (from <= to && from->key == *from->p)
164 * ext2_block_to_path - parse the block number into array of offsets
165 * @inode: inode in question (we are only interested in its superblock)
166 * @i_block: block number to be parsed
167 * @offsets: array to store the offsets in
168 * @boundary: set this non-zero if the referred-to block is likely to be
169 * followed (on disk) by an indirect block.
170 * To store the locations of file's data ext2 uses a data structure common
171 * for UNIX filesystems - tree of pointers anchored in the inode, with
172 * data blocks at leaves and indirect blocks in intermediate nodes.
173 * This function translates the block number into path in that tree -
174 * return value is the path length and @offsets[n] is the offset of
175 * pointer to (n+1)th node in the nth one. If @block is out of range
176 * (negative or too large) warning is printed and zero returned.
178 * Note: function doesn't find node addresses, so no IO is needed. All
179 * we need to know is the capacity of indirect blocks (taken from the
184 * Portability note: the last comparison (check that we fit into triple
185 * indirect block) is spelled differently, because otherwise on an
186 * architecture with 32-bit longs and 8Kb pages we might get into trouble
187 * if our filesystem had 8Kb blocks. We might use long long, but that would
188 * kill us on x86. Oh, well, at least the sign propagation does not matter -
189 * i_block would have to be negative in the very beginning, so we would not
193 static int ext2_block_to_path(struct inode *inode,
194 long i_block, int offsets[4], int *boundary)
196 int ptrs = EXT2_ADDR_PER_BLOCK(inode->i_sb);
197 int ptrs_bits = EXT2_ADDR_PER_BLOCK_BITS(inode->i_sb);
198 const long direct_blocks = EXT2_NDIR_BLOCKS,
199 indirect_blocks = ptrs,
200 double_blocks = (1 << (ptrs_bits * 2));
205 ext2_warning (inode->i_sb, "ext2_block_to_path", "block < 0");
206 } else if (i_block < direct_blocks) {
207 offsets[n++] = i_block;
208 final = direct_blocks;
209 } else if ( (i_block -= direct_blocks) < indirect_blocks) {
210 offsets[n++] = EXT2_IND_BLOCK;
211 offsets[n++] = i_block;
213 } else if ((i_block -= indirect_blocks) < double_blocks) {
214 offsets[n++] = EXT2_DIND_BLOCK;
215 offsets[n++] = i_block >> ptrs_bits;
216 offsets[n++] = i_block & (ptrs - 1);
218 } else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
219 offsets[n++] = EXT2_TIND_BLOCK;
220 offsets[n++] = i_block >> (ptrs_bits * 2);
221 offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
222 offsets[n++] = i_block & (ptrs - 1);
225 ext2_warning (inode->i_sb, "ext2_block_to_path", "block > big");
228 *boundary = (i_block & (ptrs - 1)) == (final - 1);
233 * ext2_get_branch - read the chain of indirect blocks leading to data
234 * @inode: inode in question
235 * @depth: depth of the chain (1 - direct pointer, etc.)
236 * @offsets: offsets of pointers in inode/indirect blocks
237 * @chain: place to store the result
238 * @err: here we store the error value
240 * Function fills the array of triples <key, p, bh> and returns %NULL
241 * if everything went OK or the pointer to the last filled triple
242 * (incomplete one) otherwise. Upon the return chain[i].key contains
243 * the number of (i+1)-th block in the chain (as it is stored in memory,
244 * i.e. little-endian 32-bit), chain[i].p contains the address of that
245 * number (it points into struct inode for i==0 and into the bh->b_data
246 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect
247 * block for i>0 and NULL for i==0. In other words, it holds the block
248 * numbers of the chain, addresses they were taken from (and where we can
249 * verify that chain did not change) and buffer_heads hosting these
252 * Function stops when it stumbles upon zero pointer (absent block)
253 * (pointer to last triple returned, *@err == 0)
254 * or when it gets an IO error reading an indirect block
255 * (ditto, *@err == -EIO)
256 * or when it notices that chain had been changed while it was reading
257 * (ditto, *@err == -EAGAIN)
258 * or when it reads all @depth-1 indirect blocks successfully and finds
259 * the whole chain, all way to the data (returns %NULL, *err == 0).
261 static Indirect *ext2_get_branch(struct inode *inode,
267 struct super_block *sb = inode->i_sb;
269 struct buffer_head *bh;
272 /* i_data is not going away, no lock needed */
273 add_chain (chain, NULL, EXT2_I(inode)->i_data + *offsets);
277 bh = sb_bread(sb, le32_to_cpu(p->key));
280 read_lock(&EXT2_I(inode)->i_meta_lock);
281 if (!verify_chain(chain, p))
283 add_chain(++p, bh, (u32*)bh->b_data + *++offsets);
284 read_unlock(&EXT2_I(inode)->i_meta_lock);
291 read_unlock(&EXT2_I(inode)->i_meta_lock);
302 * ext2_find_near - find a place for allocation with sufficient locality
304 * @ind: descriptor of indirect block.
306 * This function returns the prefered place for block allocation.
307 * It is used when heuristic for sequential allocation fails.
309 * + if there is a block to the left of our position - allocate near it.
310 * + if pointer will live in indirect block - allocate near that block.
311 * + if pointer will live in inode - allocate in the same cylinder group.
313 * In the latter case we colour the starting block by the callers PID to
314 * prevent it from clashing with concurrent allocations for a different inode
315 * in the same block group. The PID is used here so that functionally related
316 * files will be close-by on-disk.
318 * Caller must make sure that @ind is valid and will stay that way.
321 static unsigned long ext2_find_near(struct inode *inode, Indirect *ind)
323 struct ext2_inode_info *ei = EXT2_I(inode);
324 u32 *start = ind->bh ? (u32*) ind->bh->b_data : ei->i_data;
326 unsigned long bg_start;
327 unsigned long colour;
329 /* Try to find previous block */
330 for (p = ind->p - 1; p >= start; p--)
332 return le32_to_cpu(*p);
334 /* No such thing, so let's try location of indirect block */
336 return ind->bh->b_blocknr;
339 * It is going to be refered from inode itself? OK, just put it into
340 * the same cylinder group then.
342 bg_start = (ei->i_block_group * EXT2_BLOCKS_PER_GROUP(inode->i_sb)) +
343 le32_to_cpu(EXT2_SB(inode->i_sb)->s_es->s_first_data_block);
344 colour = (current->pid % 16) *
345 (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16);
346 return bg_start + colour;
350 * ext2_find_goal - find a prefered place for allocation.
352 * @block: block we want
353 * @chain: chain of indirect blocks
354 * @partial: pointer to the last triple within a chain
355 * @goal: place to store the result.
357 * Normally this function find the prefered place for block allocation,
358 * stores it in *@goal and returns zero. If the branch had been changed
359 * under us we return -EAGAIN.
362 static inline int ext2_find_goal(struct inode *inode,
368 struct ext2_inode_info *ei = EXT2_I(inode);
369 write_lock(&ei->i_meta_lock);
370 if (block == ei->i_next_alloc_block + 1) {
371 ei->i_next_alloc_block++;
372 ei->i_next_alloc_goal++;
374 if (verify_chain(chain, partial)) {
376 * try the heuristic for sequential allocation,
377 * failing that at least try to get decent locality.
379 if (block == ei->i_next_alloc_block)
380 *goal = ei->i_next_alloc_goal;
382 *goal = ext2_find_near(inode, partial);
383 write_unlock(&ei->i_meta_lock);
386 write_unlock(&ei->i_meta_lock);
391 * ext2_alloc_branch - allocate and set up a chain of blocks.
393 * @num: depth of the chain (number of blocks to allocate)
394 * @offsets: offsets (in the blocks) to store the pointers to next.
395 * @branch: place to store the chain in.
397 * This function allocates @num blocks, zeroes out all but the last one,
398 * links them into chain and (if we are synchronous) writes them to disk.
399 * In other words, it prepares a branch that can be spliced onto the
400 * inode. It stores the information about that chain in the branch[], in
401 * the same format as ext2_get_branch() would do. We are calling it after
402 * we had read the existing part of chain and partial points to the last
403 * triple of that (one with zero ->key). Upon the exit we have the same
404 * picture as after the successful ext2_get_block(), excpet that in one
405 * place chain is disconnected - *branch->p is still zero (we did not
406 * set the last link), but branch->key contains the number that should
407 * be placed into *branch->p to fill that gap.
409 * If allocation fails we free all blocks we've allocated (and forget
410 * their buffer_heads) and return the error value the from failed
411 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
412 * as described above and return 0.
415 static int ext2_alloc_branch(struct inode *inode,
421 int blocksize = inode->i_sb->s_blocksize;
425 int parent = ext2_alloc_block(inode, goal, &err);
427 branch[0].key = cpu_to_le32(parent);
428 if (parent) for (n = 1; n < num; n++) {
429 struct buffer_head *bh;
430 /* Allocate the next block */
431 int nr = ext2_alloc_block(inode, parent, &err);
434 branch[n].key = cpu_to_le32(nr);
436 * Get buffer_head for parent block, zero it out and set
437 * the pointer to new one, then send parent to disk.
439 bh = sb_getblk(inode->i_sb, parent);
441 memset(bh->b_data, 0, blocksize);
443 branch[n].p = (u32*) bh->b_data + offsets[n];
444 *branch[n].p = branch[n].key;
445 set_buffer_uptodate(bh);
447 mark_buffer_dirty_inode(bh, inode);
448 /* We used to sync bh here if IS_SYNC(inode).
449 * But we now rely upon generic_osync_inode()
450 * and b_inode_buffers. But not for directories.
452 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
453 sync_dirty_buffer(bh);
459 /* Allocation failed, free what we already allocated */
460 for (i = 1; i < n; i++)
461 bforget(branch[i].bh);
462 for (i = 0; i < n; i++)
463 ext2_free_blocks(inode, le32_to_cpu(branch[i].key), 1);
468 * ext2_splice_branch - splice the allocated branch onto inode.
470 * @block: (logical) number of block we are adding
471 * @chain: chain of indirect blocks (with a missing link - see
473 * @where: location of missing link
474 * @num: number of blocks we are adding
476 * This function verifies that chain (up to the missing link) had not
477 * changed, fills the missing link and does all housekeeping needed in
478 * inode (->i_blocks, etc.). In case of success we end up with the full
479 * chain to new block and return 0. Otherwise (== chain had been changed)
480 * we free the new blocks (forgetting their buffer_heads, indeed) and
484 static inline int ext2_splice_branch(struct inode *inode,
490 struct ext2_inode_info *ei = EXT2_I(inode);
493 /* Verify that place we are splicing to is still there and vacant */
495 write_lock(&ei->i_meta_lock);
496 if (!verify_chain(chain, where-1) || *where->p)
501 *where->p = where->key;
502 ei->i_next_alloc_block = block;
503 ei->i_next_alloc_goal = le32_to_cpu(where[num-1].key);
505 write_unlock(&ei->i_meta_lock);
507 /* We are done with atomic stuff, now do the rest of housekeeping */
509 inode->i_ctime = CURRENT_TIME;
511 /* had we spliced it onto indirect block? */
513 mark_buffer_dirty_inode(where->bh, inode);
515 mark_inode_dirty(inode);
519 write_unlock(&ei->i_meta_lock);
520 for (i = 1; i < num; i++)
521 bforget(where[i].bh);
522 for (i = 0; i < num; i++)
523 ext2_free_blocks(inode, le32_to_cpu(where[i].key), 1);
528 * Allocation strategy is simple: if we have to allocate something, we will
529 * have to go the whole way to leaf. So let's do it before attaching anything
530 * to tree, set linkage between the newborn blocks, write them if sync is
531 * required, recheck the path, free and repeat if check fails, otherwise
532 * set the last missing link (that will protect us from any truncate-generated
533 * removals - all blocks on the path are immune now) and possibly force the
534 * write on the parent block.
535 * That has a nice additional property: no special recovery from the failed
536 * allocations is needed - we simply release blocks and do not touch anything
537 * reachable from inode.
540 static int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
549 int depth = ext2_block_to_path(inode, iblock, offsets, &boundary);
555 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
557 /* Simplest case - block found, no allocation needed */
560 map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
562 set_buffer_boundary(bh_result);
563 /* Clean up and exit */
564 partial = chain+depth-1; /* the whole chain */
568 /* Next simple case - plain lookup or failed read of indirect block */
569 if (!create || err == -EIO) {
571 while (partial > chain) {
580 * Indirect block might be removed by truncate while we were
581 * reading it. Handling of that case (forget what we've got and
582 * reread) is taken out of the main path.
587 if (ext2_find_goal(inode, iblock, chain, partial, &goal) < 0)
590 left = (chain + depth) - partial;
591 err = ext2_alloc_branch(inode, left, goal,
592 offsets+(partial-chain), partial);
596 if (ext2_splice_branch(inode, iblock, chain, partial, left) < 0)
599 set_buffer_new(bh_result);
603 while (partial > chain) {
610 static int ext2_writepage(struct page *page, struct writeback_control *wbc)
612 return block_write_full_page(page, ext2_get_block, wbc);
615 static int ext2_readpage(struct file *file, struct page *page)
617 return mpage_readpage(page, ext2_get_block);
621 ext2_readpages(struct file *file, struct address_space *mapping,
622 struct list_head *pages, unsigned nr_pages)
624 return mpage_readpages(mapping, pages, nr_pages, ext2_get_block);
628 ext2_prepare_write(struct file *file, struct page *page,
629 unsigned from, unsigned to)
631 return block_prepare_write(page,from,to,ext2_get_block);
635 ext2_nobh_prepare_write(struct file *file, struct page *page,
636 unsigned from, unsigned to)
638 return nobh_prepare_write(page,from,to,ext2_get_block);
641 static sector_t ext2_bmap(struct address_space *mapping, sector_t block)
643 return generic_block_bmap(mapping,block,ext2_get_block);
647 ext2_get_blocks(struct inode *inode, sector_t iblock, unsigned long max_blocks,
648 struct buffer_head *bh_result, int create)
652 ret = ext2_get_block(inode, iblock, bh_result, create);
654 bh_result->b_size = (1 << inode->i_blkbits);
659 ext2_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
660 loff_t offset, unsigned long nr_segs)
662 struct file *file = iocb->ki_filp;
663 struct inode *inode = file->f_mapping->host;
665 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
666 offset, nr_segs, ext2_get_blocks, NULL);
670 ext2_writepages(struct address_space *mapping, struct writeback_control *wbc)
672 return mpage_writepages(mapping, wbc, ext2_get_block);
675 struct address_space_operations ext2_aops = {
676 .readpage = ext2_readpage,
677 .readpages = ext2_readpages,
678 .writepage = ext2_writepage,
679 .sync_page = block_sync_page,
680 .prepare_write = ext2_prepare_write,
681 .commit_write = generic_commit_write,
683 .direct_IO = ext2_direct_IO,
684 .writepages = ext2_writepages,
687 struct address_space_operations ext2_nobh_aops = {
688 .readpage = ext2_readpage,
689 .readpages = ext2_readpages,
690 .writepage = ext2_writepage,
691 .sync_page = block_sync_page,
692 .prepare_write = ext2_nobh_prepare_write,
693 .commit_write = nobh_commit_write,
695 .direct_IO = ext2_direct_IO,
696 .writepages = ext2_writepages,
700 * Probably it should be a library function... search for first non-zero word
701 * or memcmp with zero_page, whatever is better for particular architecture.
704 static inline int all_zeroes(u32 *p, u32 *q)
713 * ext2_find_shared - find the indirect blocks for partial truncation.
714 * @inode: inode in question
715 * @depth: depth of the affected branch
716 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
717 * @chain: place to store the pointers to partial indirect blocks
718 * @top: place to the (detached) top of branch
720 * This is a helper function used by ext2_truncate().
722 * When we do truncate() we may have to clean the ends of several indirect
723 * blocks but leave the blocks themselves alive. Block is partially
724 * truncated if some data below the new i_size is refered from it (and
725 * it is on the path to the first completely truncated data block, indeed).
726 * We have to free the top of that path along with everything to the right
727 * of the path. Since no allocation past the truncation point is possible
728 * until ext2_truncate() finishes, we may safely do the latter, but top
729 * of branch may require special attention - pageout below the truncation
730 * point might try to populate it.
732 * We atomically detach the top of branch from the tree, store the block
733 * number of its root in *@top, pointers to buffer_heads of partially
734 * truncated blocks - in @chain[].bh and pointers to their last elements
735 * that should not be removed - in @chain[].p. Return value is the pointer
736 * to last filled element of @chain.
738 * The work left to caller to do the actual freeing of subtrees:
739 * a) free the subtree starting from *@top
740 * b) free the subtrees whose roots are stored in
741 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
742 * c) free the subtrees growing from the inode past the @chain[0].p
743 * (no partially truncated stuff there).
746 static Indirect *ext2_find_shared(struct inode *inode,
752 Indirect *partial, *p;
756 for (k = depth; k > 1 && !offsets[k-1]; k--)
758 partial = ext2_get_branch(inode, k, offsets, chain, &err);
760 partial = chain + k-1;
762 * If the branch acquired continuation since we've looked at it -
763 * fine, it should all survive and (new) top doesn't belong to us.
765 write_lock(&EXT2_I(inode)->i_meta_lock);
766 if (!partial->key && *partial->p) {
767 write_unlock(&EXT2_I(inode)->i_meta_lock);
770 for (p=partial; p>chain && all_zeroes((u32*)p->bh->b_data,p->p); p--)
773 * OK, we've found the last block that must survive. The rest of our
774 * branch should be detached before unlocking. However, if that rest
775 * of branch is all ours and does not grow immediately from the inode
776 * it's easier to cheat and just decrement partial->p.
778 if (p == chain + k - 1 && p > chain) {
784 write_unlock(&EXT2_I(inode)->i_meta_lock);
796 * ext2_free_data - free a list of data blocks
797 * @inode: inode we are dealing with
798 * @p: array of block numbers
799 * @q: points immediately past the end of array
801 * We are freeing all blocks refered from that array (numbers are
802 * stored as little-endian 32-bit) and updating @inode->i_blocks
805 static inline void ext2_free_data(struct inode *inode, u32 *p, u32 *q)
807 unsigned long block_to_free = 0, count = 0;
810 for ( ; p < q ; p++) {
811 nr = le32_to_cpu(*p);
814 /* accumulate blocks to free if they're contiguous */
817 else if (block_to_free == nr - count)
820 mark_inode_dirty(inode);
821 ext2_free_blocks (inode, block_to_free, count);
829 mark_inode_dirty(inode);
830 ext2_free_blocks (inode, block_to_free, count);
835 * ext2_free_branches - free an array of branches
836 * @inode: inode we are dealing with
837 * @p: array of block numbers
838 * @q: pointer immediately past the end of array
839 * @depth: depth of the branches to free
841 * We are freeing all blocks refered from these branches (numbers are
842 * stored as little-endian 32-bit) and updating @inode->i_blocks
845 static void ext2_free_branches(struct inode *inode, u32 *p, u32 *q, int depth)
847 struct buffer_head * bh;
851 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
852 for ( ; p < q ; p++) {
853 nr = le32_to_cpu(*p);
857 bh = sb_bread(inode->i_sb, nr);
859 * A read failure? Report error and clear slot
863 ext2_error(inode->i_sb, "ext2_free_branches",
864 "Read failure, inode=%ld, block=%ld",
868 ext2_free_branches(inode,
870 (u32*)bh->b_data + addr_per_block,
873 ext2_free_blocks(inode, nr, 1);
874 mark_inode_dirty(inode);
877 ext2_free_data(inode, p, q);
880 void ext2_truncate (struct inode * inode)
882 u32 *i_data = EXT2_I(inode)->i_data;
883 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
892 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
893 S_ISLNK(inode->i_mode)))
895 if (ext2_inode_is_fast_symlink(inode))
897 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
900 ext2_discard_prealloc(inode);
902 blocksize = inode->i_sb->s_blocksize;
903 iblock = (inode->i_size + blocksize-1)
904 >> EXT2_BLOCK_SIZE_BITS(inode->i_sb);
906 if (test_opt(inode->i_sb, NOBH))
907 nobh_truncate_page(inode->i_mapping, inode->i_size);
909 block_truncate_page(inode->i_mapping,
910 inode->i_size, ext2_get_block);
912 n = ext2_block_to_path(inode, iblock, offsets, NULL);
917 ext2_free_data(inode, i_data+offsets[0],
918 i_data + EXT2_NDIR_BLOCKS);
922 partial = ext2_find_shared(inode, n, offsets, chain, &nr);
923 /* Kill the top of shared branch (already detached) */
925 if (partial == chain)
926 mark_inode_dirty(inode);
928 mark_buffer_dirty_inode(partial->bh, inode);
929 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
931 /* Clear the ends of indirect blocks on the shared branch */
932 while (partial > chain) {
933 ext2_free_branches(inode,
935 (u32*)partial->bh->b_data + addr_per_block,
936 (chain+n-1) - partial);
937 mark_buffer_dirty_inode(partial->bh, inode);
938 brelse (partial->bh);
942 /* Kill the remaining (whole) subtrees */
943 switch (offsets[0]) {
945 nr = i_data[EXT2_IND_BLOCK];
947 i_data[EXT2_IND_BLOCK] = 0;
948 mark_inode_dirty(inode);
949 ext2_free_branches(inode, &nr, &nr+1, 1);
952 nr = i_data[EXT2_DIND_BLOCK];
954 i_data[EXT2_DIND_BLOCK] = 0;
955 mark_inode_dirty(inode);
956 ext2_free_branches(inode, &nr, &nr+1, 2);
958 case EXT2_DIND_BLOCK:
959 nr = i_data[EXT2_TIND_BLOCK];
961 i_data[EXT2_TIND_BLOCK] = 0;
962 mark_inode_dirty(inode);
963 ext2_free_branches(inode, &nr, &nr+1, 3);
965 case EXT2_TIND_BLOCK:
968 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
969 if (inode_needs_sync(inode)) {
970 sync_mapping_buffers(inode->i_mapping);
971 ext2_sync_inode (inode);
973 mark_inode_dirty(inode);
977 static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
978 struct buffer_head **p)
980 struct buffer_head * bh;
981 unsigned long block_group;
983 unsigned long offset;
984 struct ext2_group_desc * gdp;
987 if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) ||
988 ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
991 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
992 gdp = ext2_get_group_desc(sb, block_group, &bh);
996 * Figure out the offset within the block group inode table
998 offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb);
999 block = le32_to_cpu(gdp->bg_inode_table) +
1000 (offset >> EXT2_BLOCK_SIZE_BITS(sb));
1001 if (!(bh = sb_bread(sb, block)))
1005 offset &= (EXT2_BLOCK_SIZE(sb) - 1);
1006 return (struct ext2_inode *) (bh->b_data + offset);
1009 ext2_error(sb, "ext2_get_inode", "bad inode number: %lu",
1010 (unsigned long) ino);
1011 return ERR_PTR(-EINVAL);
1013 ext2_error(sb, "ext2_get_inode",
1014 "unable to read inode block - inode=%lu, block=%lu",
1015 (unsigned long) ino, block);
1017 return ERR_PTR(-EIO);
1020 void ext2_set_inode_flags(struct inode *inode)
1022 unsigned int flags = EXT2_I(inode)->i_flags;
1024 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
1025 if (flags & EXT2_SYNC_FL)
1026 inode->i_flags |= S_SYNC;
1027 if (flags & EXT2_APPEND_FL)
1028 inode->i_flags |= S_APPEND;
1029 if (flags & EXT2_IMMUTABLE_FL)
1030 inode->i_flags |= S_IMMUTABLE;
1031 if (flags & EXT2_NOATIME_FL)
1032 inode->i_flags |= S_NOATIME;
1033 if (flags & EXT2_DIRSYNC_FL)
1034 inode->i_flags |= S_DIRSYNC;
1037 void ext2_read_inode (struct inode * inode)
1039 struct ext2_inode_info *ei = EXT2_I(inode);
1040 ino_t ino = inode->i_ino;
1041 struct buffer_head * bh;
1042 struct ext2_inode * raw_inode = ext2_get_inode(inode->i_sb, ino, &bh);
1045 #ifdef CONFIG_EXT2_FS_POSIX_ACL
1046 ei->i_acl = EXT2_ACL_NOT_CACHED;
1047 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
1049 if (IS_ERR(raw_inode))
1052 inode->i_mode = le16_to_cpu(raw_inode->i_mode);
1053 inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
1054 inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
1055 if (!(test_opt (inode->i_sb, NO_UID32))) {
1056 inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
1057 inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
1059 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
1060 inode->i_size = le32_to_cpu(raw_inode->i_size);
1061 inode->i_atime.tv_sec = le32_to_cpu(raw_inode->i_atime);
1062 inode->i_ctime.tv_sec = le32_to_cpu(raw_inode->i_ctime);
1063 inode->i_mtime.tv_sec = le32_to_cpu(raw_inode->i_mtime);
1064 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1065 ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
1066 /* We now have enough fields to check if the inode was active or not.
1067 * This is needed because nfsd might try to access dead inodes
1068 * the test is that same one that e2fsck uses
1069 * NeilBrown 1999oct15
1071 if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) {
1072 /* this inode is deleted */
1076 inode->i_blksize = PAGE_SIZE; /* This is the optimal IO size (for stat), not the fs block size */
1077 inode->i_blocks = le32_to_cpu(raw_inode->i_blocks);
1078 ei->i_flags = le32_to_cpu(raw_inode->i_flags);
1079 ei->i_faddr = le32_to_cpu(raw_inode->i_faddr);
1080 ei->i_frag_no = raw_inode->i_frag;
1081 ei->i_frag_size = raw_inode->i_fsize;
1082 ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl);
1084 if (S_ISREG(inode->i_mode))
1085 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32;
1087 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl);
1089 inode->i_generation = le32_to_cpu(raw_inode->i_generation);
1091 ei->i_next_alloc_block = 0;
1092 ei->i_next_alloc_goal = 0;
1093 ei->i_prealloc_count = 0;
1094 ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb);
1095 ei->i_dir_start_lookup = 0;
1098 * NOTE! The in-memory inode i_data array is in little-endian order
1099 * even on big-endian machines: we do NOT byteswap the block numbers!
1101 for (n = 0; n < EXT2_N_BLOCKS; n++)
1102 ei->i_data[n] = raw_inode->i_block[n];
1104 if (S_ISREG(inode->i_mode)) {
1105 inode->i_op = &ext2_file_inode_operations;
1106 inode->i_fop = &ext2_file_operations;
1107 if (test_opt(inode->i_sb, NOBH))
1108 inode->i_mapping->a_ops = &ext2_nobh_aops;
1110 inode->i_mapping->a_ops = &ext2_aops;
1111 } else if (S_ISDIR(inode->i_mode)) {
1112 inode->i_op = &ext2_dir_inode_operations;
1113 inode->i_fop = &ext2_dir_operations;
1114 if (test_opt(inode->i_sb, NOBH))
1115 inode->i_mapping->a_ops = &ext2_nobh_aops;
1117 inode->i_mapping->a_ops = &ext2_aops;
1118 } else if (S_ISLNK(inode->i_mode)) {
1119 if (ext2_inode_is_fast_symlink(inode))
1120 inode->i_op = &ext2_fast_symlink_inode_operations;
1122 inode->i_op = &ext2_symlink_inode_operations;
1123 if (test_opt(inode->i_sb, NOBH))
1124 inode->i_mapping->a_ops = &ext2_nobh_aops;
1126 inode->i_mapping->a_ops = &ext2_aops;
1129 inode->i_op = &ext2_special_inode_operations;
1130 if (raw_inode->i_block[0])
1131 init_special_inode(inode, inode->i_mode,
1132 old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
1134 init_special_inode(inode, inode->i_mode,
1135 new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
1138 ext2_set_inode_flags(inode);
1142 make_bad_inode(inode);
1146 static int ext2_update_inode(struct inode * inode, int do_sync)
1148 struct ext2_inode_info *ei = EXT2_I(inode);
1149 struct super_block *sb = inode->i_sb;
1150 ino_t ino = inode->i_ino;
1151 uid_t uid = inode->i_uid;
1152 gid_t gid = inode->i_gid;
1153 struct buffer_head * bh;
1154 struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh);
1158 if (IS_ERR(raw_inode))
1161 /* For fields not not tracking in the in-memory inode,
1162 * initialise them to zero for new inodes. */
1163 if (ei->i_state & EXT2_STATE_NEW)
1164 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
1166 raw_inode->i_mode = cpu_to_le16(inode->i_mode);
1167 if (!(test_opt(sb, NO_UID32))) {
1168 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid));
1169 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid));
1171 * Fix up interoperability with old kernels. Otherwise, old inodes get
1172 * re-used with the upper 16 bits of the uid/gid intact
1175 raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid));
1176 raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid));
1178 raw_inode->i_uid_high = 0;
1179 raw_inode->i_gid_high = 0;
1182 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid));
1183 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid));
1184 raw_inode->i_uid_high = 0;
1185 raw_inode->i_gid_high = 0;
1187 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
1188 raw_inode->i_size = cpu_to_le32(inode->i_size);
1189 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1190 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1191 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1193 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
1194 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
1195 raw_inode->i_flags = cpu_to_le32(ei->i_flags);
1196 raw_inode->i_faddr = cpu_to_le32(ei->i_faddr);
1197 raw_inode->i_frag = ei->i_frag_no;
1198 raw_inode->i_fsize = ei->i_frag_size;
1199 raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl);
1200 if (!S_ISREG(inode->i_mode))
1201 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl);
1203 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32);
1204 if (inode->i_size > 0x7fffffffULL) {
1205 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb,
1206 EXT2_FEATURE_RO_COMPAT_LARGE_FILE) ||
1207 EXT2_SB(sb)->s_es->s_rev_level ==
1208 cpu_to_le32(EXT2_GOOD_OLD_REV)) {
1209 /* If this is the first large file
1210 * created, add a flag to the superblock.
1213 ext2_update_dynamic_rev(sb);
1214 EXT2_SET_RO_COMPAT_FEATURE(sb,
1215 EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
1217 ext2_write_super(sb);
1222 raw_inode->i_generation = cpu_to_le32(inode->i_generation);
1223 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1224 if (old_valid_dev(inode->i_rdev)) {
1225 raw_inode->i_block[0] =
1226 cpu_to_le32(old_encode_dev(inode->i_rdev));
1227 raw_inode->i_block[1] = 0;
1229 raw_inode->i_block[0] = 0;
1230 raw_inode->i_block[1] =
1231 cpu_to_le32(new_encode_dev(inode->i_rdev));
1232 raw_inode->i_block[2] = 0;
1234 } else for (n = 0; n < EXT2_N_BLOCKS; n++)
1235 raw_inode->i_block[n] = ei->i_data[n];
1236 mark_buffer_dirty(bh);
1238 sync_dirty_buffer(bh);
1239 if (buffer_req(bh) && !buffer_uptodate(bh)) {
1240 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1241 sb->s_id, (unsigned long) ino);
1245 ei->i_state &= ~EXT2_STATE_NEW;
1250 void ext2_write_inode(struct inode *inode, int wait)
1252 ext2_update_inode(inode, wait);
1255 int ext2_sync_inode(struct inode *inode)
1257 struct writeback_control wbc = {
1258 .sync_mode = WB_SYNC_ALL,
1259 .nr_to_write = 0, /* sys_fsync did this */
1261 return sync_inode(inode, &wbc);
1264 int ext2_setattr(struct dentry *dentry, struct iattr *iattr)
1266 struct inode *inode = dentry->d_inode;
1269 error = inode_change_ok(inode, iattr);
1272 if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) ||
1273 (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid)) {
1274 error = DQUOT_TRANSFER(inode, iattr) ? -EDQUOT : 0;
1278 inode_setattr(inode, iattr);
1279 if (iattr->ia_valid & ATTR_MODE)
1280 error = ext2_acl_chmod(inode);