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/vs_tag.h>
39 MODULE_AUTHOR("Remy Card and others");
40 MODULE_DESCRIPTION("Second Extended Filesystem");
41 MODULE_LICENSE("GPL");
43 static int ext2_update_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);
58 * Called at each iput().
60 * The inode may be "bad" if ext2_read_inode() saw an error from
61 * ext2_get_inode(), so we need to check that to avoid freeing random disk
64 void ext2_put_inode(struct inode *inode)
66 if (!is_bad_inode(inode))
67 ext2_discard_prealloc(inode);
71 * Called at the last iput() if i_nlink is zero.
73 void ext2_delete_inode (struct inode * inode)
75 truncate_inode_pages(&inode->i_data, 0);
77 if (is_bad_inode(inode))
79 EXT2_I(inode)->i_dtime = get_seconds();
80 mark_inode_dirty(inode);
81 ext2_update_inode(inode, inode_needs_sync(inode));
85 ext2_truncate (inode);
86 ext2_free_inode (inode);
90 clear_inode(inode); /* We must guarantee clearing of inode... */
93 void ext2_discard_prealloc (struct inode * inode)
95 #ifdef EXT2_PREALLOCATE
96 struct ext2_inode_info *ei = EXT2_I(inode);
97 write_lock(&ei->i_meta_lock);
98 if (ei->i_prealloc_count) {
99 unsigned short total = ei->i_prealloc_count;
100 unsigned long block = ei->i_prealloc_block;
101 ei->i_prealloc_count = 0;
102 ei->i_prealloc_block = 0;
103 write_unlock(&ei->i_meta_lock);
104 ext2_free_blocks (inode, block, total);
107 write_unlock(&ei->i_meta_lock);
111 static int ext2_alloc_block (struct inode * inode, unsigned long goal, int *err)
114 static unsigned long alloc_hits, alloc_attempts;
116 unsigned long result;
119 #ifdef EXT2_PREALLOCATE
120 struct ext2_inode_info *ei = EXT2_I(inode);
121 write_lock(&ei->i_meta_lock);
122 if (ei->i_prealloc_count &&
123 (goal == ei->i_prealloc_block || goal + 1 == ei->i_prealloc_block))
125 result = ei->i_prealloc_block++;
126 ei->i_prealloc_count--;
127 write_unlock(&ei->i_meta_lock);
128 ext2_debug ("preallocation hit (%lu/%lu).\n",
129 ++alloc_hits, ++alloc_attempts);
131 write_unlock(&ei->i_meta_lock);
132 ext2_discard_prealloc (inode);
133 ext2_debug ("preallocation miss (%lu/%lu).\n",
134 alloc_hits, ++alloc_attempts);
135 if (S_ISREG(inode->i_mode))
136 result = ext2_new_block (inode, goal,
137 &ei->i_prealloc_count,
138 &ei->i_prealloc_block, err);
140 result = ext2_new_block(inode, goal, NULL, NULL, err);
143 result = ext2_new_block (inode, goal, 0, 0, err);
151 struct buffer_head *bh;
154 static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v)
156 p->key = *(p->p = v);
160 static inline int verify_chain(Indirect *from, Indirect *to)
162 while (from <= to && from->key == *from->p)
168 * ext2_block_to_path - parse the block number into array of offsets
169 * @inode: inode in question (we are only interested in its superblock)
170 * @i_block: block number to be parsed
171 * @offsets: array to store the offsets in
172 * @boundary: set this non-zero if the referred-to block is likely to be
173 * followed (on disk) by an indirect block.
174 * To store the locations of file's data ext2 uses a data structure common
175 * for UNIX filesystems - tree of pointers anchored in the inode, with
176 * data blocks at leaves and indirect blocks in intermediate nodes.
177 * This function translates the block number into path in that tree -
178 * return value is the path length and @offsets[n] is the offset of
179 * pointer to (n+1)th node in the nth one. If @block is out of range
180 * (negative or too large) warning is printed and zero returned.
182 * Note: function doesn't find node addresses, so no IO is needed. All
183 * we need to know is the capacity of indirect blocks (taken from the
188 * Portability note: the last comparison (check that we fit into triple
189 * indirect block) is spelled differently, because otherwise on an
190 * architecture with 32-bit longs and 8Kb pages we might get into trouble
191 * if our filesystem had 8Kb blocks. We might use long long, but that would
192 * kill us on x86. Oh, well, at least the sign propagation does not matter -
193 * i_block would have to be negative in the very beginning, so we would not
197 static int ext2_block_to_path(struct inode *inode,
198 long i_block, int offsets[4], int *boundary)
200 int ptrs = EXT2_ADDR_PER_BLOCK(inode->i_sb);
201 int ptrs_bits = EXT2_ADDR_PER_BLOCK_BITS(inode->i_sb);
202 const long direct_blocks = EXT2_NDIR_BLOCKS,
203 indirect_blocks = ptrs,
204 double_blocks = (1 << (ptrs_bits * 2));
209 ext2_warning (inode->i_sb, "ext2_block_to_path", "block < 0");
210 } else if (i_block < direct_blocks) {
211 offsets[n++] = i_block;
212 final = direct_blocks;
213 } else if ( (i_block -= direct_blocks) < indirect_blocks) {
214 offsets[n++] = EXT2_IND_BLOCK;
215 offsets[n++] = i_block;
217 } else if ((i_block -= indirect_blocks) < double_blocks) {
218 offsets[n++] = EXT2_DIND_BLOCK;
219 offsets[n++] = i_block >> ptrs_bits;
220 offsets[n++] = i_block & (ptrs - 1);
222 } else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
223 offsets[n++] = EXT2_TIND_BLOCK;
224 offsets[n++] = i_block >> (ptrs_bits * 2);
225 offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
226 offsets[n++] = i_block & (ptrs - 1);
229 ext2_warning (inode->i_sb, "ext2_block_to_path", "block > big");
232 *boundary = (i_block & (ptrs - 1)) == (final - 1);
237 * ext2_get_branch - read the chain of indirect blocks leading to data
238 * @inode: inode in question
239 * @depth: depth of the chain (1 - direct pointer, etc.)
240 * @offsets: offsets of pointers in inode/indirect blocks
241 * @chain: place to store the result
242 * @err: here we store the error value
244 * Function fills the array of triples <key, p, bh> and returns %NULL
245 * if everything went OK or the pointer to the last filled triple
246 * (incomplete one) otherwise. Upon the return chain[i].key contains
247 * the number of (i+1)-th block in the chain (as it is stored in memory,
248 * i.e. little-endian 32-bit), chain[i].p contains the address of that
249 * number (it points into struct inode for i==0 and into the bh->b_data
250 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect
251 * block for i>0 and NULL for i==0. In other words, it holds the block
252 * numbers of the chain, addresses they were taken from (and where we can
253 * verify that chain did not change) and buffer_heads hosting these
256 * Function stops when it stumbles upon zero pointer (absent block)
257 * (pointer to last triple returned, *@err == 0)
258 * or when it gets an IO error reading an indirect block
259 * (ditto, *@err == -EIO)
260 * or when it notices that chain had been changed while it was reading
261 * (ditto, *@err == -EAGAIN)
262 * or when it reads all @depth-1 indirect blocks successfully and finds
263 * the whole chain, all way to the data (returns %NULL, *err == 0).
265 static Indirect *ext2_get_branch(struct inode *inode,
271 struct super_block *sb = inode->i_sb;
273 struct buffer_head *bh;
276 /* i_data is not going away, no lock needed */
277 add_chain (chain, NULL, EXT2_I(inode)->i_data + *offsets);
281 bh = sb_bread(sb, le32_to_cpu(p->key));
284 read_lock(&EXT2_I(inode)->i_meta_lock);
285 if (!verify_chain(chain, p))
287 add_chain(++p, bh, (__le32*)bh->b_data + *++offsets);
288 read_unlock(&EXT2_I(inode)->i_meta_lock);
295 read_unlock(&EXT2_I(inode)->i_meta_lock);
306 * ext2_find_near - find a place for allocation with sufficient locality
308 * @ind: descriptor of indirect block.
310 * This function returns the prefered place for block allocation.
311 * It is used when heuristic for sequential allocation fails.
313 * + if there is a block to the left of our position - allocate near it.
314 * + if pointer will live in indirect block - allocate near that block.
315 * + if pointer will live in inode - allocate in the same cylinder group.
317 * In the latter case we colour the starting block by the callers PID to
318 * prevent it from clashing with concurrent allocations for a different inode
319 * in the same block group. The PID is used here so that functionally related
320 * files will be close-by on-disk.
322 * Caller must make sure that @ind is valid and will stay that way.
325 static unsigned long ext2_find_near(struct inode *inode, Indirect *ind)
327 struct ext2_inode_info *ei = EXT2_I(inode);
328 __le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data;
330 unsigned long bg_start;
331 unsigned long colour;
333 /* Try to find previous block */
334 for (p = ind->p - 1; p >= start; p--)
336 return le32_to_cpu(*p);
338 /* No such thing, so let's try location of indirect block */
340 return ind->bh->b_blocknr;
343 * It is going to be refered from inode itself? OK, just put it into
344 * the same cylinder group then.
346 bg_start = (ei->i_block_group * EXT2_BLOCKS_PER_GROUP(inode->i_sb)) +
347 le32_to_cpu(EXT2_SB(inode->i_sb)->s_es->s_first_data_block);
348 colour = (current->pid % 16) *
349 (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16);
350 return bg_start + colour;
354 * ext2_find_goal - find a prefered place for allocation.
356 * @block: block we want
357 * @chain: chain of indirect blocks
358 * @partial: pointer to the last triple within a chain
359 * @goal: place to store the result.
361 * Normally this function find the prefered place for block allocation,
362 * stores it in *@goal and returns zero. If the branch had been changed
363 * under us we return -EAGAIN.
366 static inline int ext2_find_goal(struct inode *inode,
372 struct ext2_inode_info *ei = EXT2_I(inode);
373 write_lock(&ei->i_meta_lock);
374 if ((block == ei->i_next_alloc_block + 1) && ei->i_next_alloc_goal) {
375 ei->i_next_alloc_block++;
376 ei->i_next_alloc_goal++;
378 if (verify_chain(chain, partial)) {
380 * try the heuristic for sequential allocation,
381 * failing that at least try to get decent locality.
383 if (block == ei->i_next_alloc_block)
384 *goal = ei->i_next_alloc_goal;
386 *goal = ext2_find_near(inode, partial);
387 write_unlock(&ei->i_meta_lock);
390 write_unlock(&ei->i_meta_lock);
395 * ext2_alloc_branch - allocate and set up a chain of blocks.
397 * @num: depth of the chain (number of blocks to allocate)
398 * @offsets: offsets (in the blocks) to store the pointers to next.
399 * @branch: place to store the chain in.
401 * This function allocates @num blocks, zeroes out all but the last one,
402 * links them into chain and (if we are synchronous) writes them to disk.
403 * In other words, it prepares a branch that can be spliced onto the
404 * inode. It stores the information about that chain in the branch[], in
405 * the same format as ext2_get_branch() would do. We are calling it after
406 * we had read the existing part of chain and partial points to the last
407 * triple of that (one with zero ->key). Upon the exit we have the same
408 * picture as after the successful ext2_get_block(), excpet that in one
409 * place chain is disconnected - *branch->p is still zero (we did not
410 * set the last link), but branch->key contains the number that should
411 * be placed into *branch->p to fill that gap.
413 * If allocation fails we free all blocks we've allocated (and forget
414 * their buffer_heads) and return the error value the from failed
415 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
416 * as described above and return 0.
419 static int ext2_alloc_branch(struct inode *inode,
425 int blocksize = inode->i_sb->s_blocksize;
429 int parent = ext2_alloc_block(inode, goal, &err);
431 branch[0].key = cpu_to_le32(parent);
432 if (parent) for (n = 1; n < num; n++) {
433 struct buffer_head *bh;
434 /* Allocate the next block */
435 int nr = ext2_alloc_block(inode, parent, &err);
438 branch[n].key = cpu_to_le32(nr);
440 * Get buffer_head for parent block, zero it out and set
441 * the pointer to new one, then send parent to disk.
443 bh = sb_getblk(inode->i_sb, parent);
449 memset(bh->b_data, 0, blocksize);
451 branch[n].p = (__le32 *) bh->b_data + offsets[n];
452 *branch[n].p = branch[n].key;
453 set_buffer_uptodate(bh);
455 mark_buffer_dirty_inode(bh, inode);
456 /* We used to sync bh here if IS_SYNC(inode).
457 * But we now rely upon generic_osync_inode()
458 * and b_inode_buffers. But not for directories.
460 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
461 sync_dirty_buffer(bh);
467 /* Allocation failed, free what we already allocated */
468 for (i = 1; i < n; i++)
469 bforget(branch[i].bh);
470 for (i = 0; i < n; i++)
471 ext2_free_blocks(inode, le32_to_cpu(branch[i].key), 1);
476 * ext2_splice_branch - splice the allocated branch onto inode.
478 * @block: (logical) number of block we are adding
479 * @chain: chain of indirect blocks (with a missing link - see
481 * @where: location of missing link
482 * @num: number of blocks we are adding
484 * This function verifies that chain (up to the missing link) had not
485 * changed, fills the missing link and does all housekeeping needed in
486 * inode (->i_blocks, etc.). In case of success we end up with the full
487 * chain to new block and return 0. Otherwise (== chain had been changed)
488 * we free the new blocks (forgetting their buffer_heads, indeed) and
492 static inline int ext2_splice_branch(struct inode *inode,
498 struct ext2_inode_info *ei = EXT2_I(inode);
501 /* Verify that place we are splicing to is still there and vacant */
503 write_lock(&ei->i_meta_lock);
504 if (!verify_chain(chain, where-1) || *where->p)
509 *where->p = where->key;
510 ei->i_next_alloc_block = block;
511 ei->i_next_alloc_goal = le32_to_cpu(where[num-1].key);
513 write_unlock(&ei->i_meta_lock);
515 /* We are done with atomic stuff, now do the rest of housekeeping */
517 inode->i_ctime = CURRENT_TIME_SEC;
519 /* had we spliced it onto indirect block? */
521 mark_buffer_dirty_inode(where->bh, inode);
523 mark_inode_dirty(inode);
527 write_unlock(&ei->i_meta_lock);
528 for (i = 1; i < num; i++)
529 bforget(where[i].bh);
530 for (i = 0; i < num; i++)
531 ext2_free_blocks(inode, le32_to_cpu(where[i].key), 1);
536 * Allocation strategy is simple: if we have to allocate something, we will
537 * have to go the whole way to leaf. So let's do it before attaching anything
538 * to tree, set linkage between the newborn blocks, write them if sync is
539 * required, recheck the path, free and repeat if check fails, otherwise
540 * set the last missing link (that will protect us from any truncate-generated
541 * removals - all blocks on the path are immune now) and possibly force the
542 * write on the parent block.
543 * That has a nice additional property: no special recovery from the failed
544 * allocations is needed - we simply release blocks and do not touch anything
545 * reachable from inode.
548 int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
557 int depth = ext2_block_to_path(inode, iblock, offsets, &boundary);
563 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
565 /* Simplest case - block found, no allocation needed */
568 map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
570 set_buffer_boundary(bh_result);
571 /* Clean up and exit */
572 partial = chain+depth-1; /* the whole chain */
576 /* Next simple case - plain lookup or failed read of indirect block */
577 if (!create || err == -EIO) {
579 while (partial > chain) {
588 * Indirect block might be removed by truncate while we were
589 * reading it. Handling of that case (forget what we've got and
590 * reread) is taken out of the main path.
596 if (ext2_find_goal(inode, iblock, chain, partial, &goal) < 0)
599 left = (chain + depth) - partial;
600 err = ext2_alloc_branch(inode, left, goal,
601 offsets+(partial-chain), partial);
605 if (ext2_use_xip(inode->i_sb)) {
607 * we need to clear the block
609 err = ext2_clear_xip_target (inode,
610 le32_to_cpu(chain[depth-1].key));
615 if (ext2_splice_branch(inode, iblock, chain, partial, left) < 0)
618 set_buffer_new(bh_result);
622 while (partial > chain) {
629 static int ext2_writepage(struct page *page, struct writeback_control *wbc)
631 return block_write_full_page(page, ext2_get_block, wbc);
634 static int ext2_readpage(struct file *file, struct page *page)
636 return mpage_readpage(page, ext2_get_block);
640 ext2_readpages(struct file *file, struct address_space *mapping,
641 struct list_head *pages, unsigned nr_pages)
643 return mpage_readpages(mapping, pages, nr_pages, ext2_get_block);
647 ext2_prepare_write(struct file *file, struct page *page,
648 unsigned from, unsigned to)
650 return block_prepare_write(page,from,to,ext2_get_block);
654 ext2_nobh_prepare_write(struct file *file, struct page *page,
655 unsigned from, unsigned to)
657 return nobh_prepare_write(page,from,to,ext2_get_block);
660 static int ext2_nobh_writepage(struct page *page,
661 struct writeback_control *wbc)
663 return nobh_writepage(page, ext2_get_block, wbc);
666 static sector_t ext2_bmap(struct address_space *mapping, sector_t block)
668 return generic_block_bmap(mapping,block,ext2_get_block);
672 ext2_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
673 loff_t offset, unsigned long nr_segs)
675 struct file *file = iocb->ki_filp;
676 struct inode *inode = file->f_mapping->host;
678 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
679 offset, nr_segs, ext2_get_block, NULL);
683 ext2_writepages(struct address_space *mapping, struct writeback_control *wbc)
685 return mpage_writepages(mapping, wbc, ext2_get_block);
688 const struct address_space_operations ext2_aops = {
689 .readpage = ext2_readpage,
690 .readpages = ext2_readpages,
691 .writepage = ext2_writepage,
692 .sync_page = block_sync_page,
693 .prepare_write = ext2_prepare_write,
694 .commit_write = generic_commit_write,
696 .direct_IO = ext2_direct_IO,
697 .writepages = ext2_writepages,
698 .migratepage = buffer_migrate_page,
701 const struct address_space_operations ext2_aops_xip = {
703 .get_xip_page = ext2_get_xip_page,
706 const struct address_space_operations ext2_nobh_aops = {
707 .readpage = ext2_readpage,
708 .readpages = ext2_readpages,
709 .writepage = ext2_nobh_writepage,
710 .sync_page = block_sync_page,
711 .prepare_write = ext2_nobh_prepare_write,
712 .commit_write = nobh_commit_write,
714 .direct_IO = ext2_direct_IO,
715 .writepages = ext2_writepages,
716 .migratepage = buffer_migrate_page,
720 * Probably it should be a library function... search for first non-zero word
721 * or memcmp with zero_page, whatever is better for particular architecture.
724 static inline int all_zeroes(__le32 *p, __le32 *q)
733 * ext2_find_shared - find the indirect blocks for partial truncation.
734 * @inode: inode in question
735 * @depth: depth of the affected branch
736 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
737 * @chain: place to store the pointers to partial indirect blocks
738 * @top: place to the (detached) top of branch
740 * This is a helper function used by ext2_truncate().
742 * When we do truncate() we may have to clean the ends of several indirect
743 * blocks but leave the blocks themselves alive. Block is partially
744 * truncated if some data below the new i_size is refered from it (and
745 * it is on the path to the first completely truncated data block, indeed).
746 * We have to free the top of that path along with everything to the right
747 * of the path. Since no allocation past the truncation point is possible
748 * until ext2_truncate() finishes, we may safely do the latter, but top
749 * of branch may require special attention - pageout below the truncation
750 * point might try to populate it.
752 * We atomically detach the top of branch from the tree, store the block
753 * number of its root in *@top, pointers to buffer_heads of partially
754 * truncated blocks - in @chain[].bh and pointers to their last elements
755 * that should not be removed - in @chain[].p. Return value is the pointer
756 * to last filled element of @chain.
758 * The work left to caller to do the actual freeing of subtrees:
759 * a) free the subtree starting from *@top
760 * b) free the subtrees whose roots are stored in
761 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
762 * c) free the subtrees growing from the inode past the @chain[0].p
763 * (no partially truncated stuff there).
766 static Indirect *ext2_find_shared(struct inode *inode,
772 Indirect *partial, *p;
776 for (k = depth; k > 1 && !offsets[k-1]; k--)
778 partial = ext2_get_branch(inode, k, offsets, chain, &err);
780 partial = chain + k-1;
782 * If the branch acquired continuation since we've looked at it -
783 * fine, it should all survive and (new) top doesn't belong to us.
785 write_lock(&EXT2_I(inode)->i_meta_lock);
786 if (!partial->key && *partial->p) {
787 write_unlock(&EXT2_I(inode)->i_meta_lock);
790 for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--)
793 * OK, we've found the last block that must survive. The rest of our
794 * branch should be detached before unlocking. However, if that rest
795 * of branch is all ours and does not grow immediately from the inode
796 * it's easier to cheat and just decrement partial->p.
798 if (p == chain + k - 1 && p > chain) {
804 write_unlock(&EXT2_I(inode)->i_meta_lock);
816 * ext2_free_data - free a list of data blocks
817 * @inode: inode we are dealing with
818 * @p: array of block numbers
819 * @q: points immediately past the end of array
821 * We are freeing all blocks refered from that array (numbers are
822 * stored as little-endian 32-bit) and updating @inode->i_blocks
825 static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q)
827 unsigned long block_to_free = 0, count = 0;
830 for ( ; p < q ; p++) {
831 nr = le32_to_cpu(*p);
834 /* accumulate blocks to free if they're contiguous */
837 else if (block_to_free == nr - count)
840 mark_inode_dirty(inode);
841 ext2_free_blocks (inode, block_to_free, count);
849 mark_inode_dirty(inode);
850 ext2_free_blocks (inode, block_to_free, count);
855 * ext2_free_branches - free an array of branches
856 * @inode: inode we are dealing with
857 * @p: array of block numbers
858 * @q: pointer immediately past the end of array
859 * @depth: depth of the branches to free
861 * We are freeing all blocks refered from these branches (numbers are
862 * stored as little-endian 32-bit) and updating @inode->i_blocks
865 static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth)
867 struct buffer_head * bh;
871 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
872 for ( ; p < q ; p++) {
873 nr = le32_to_cpu(*p);
877 bh = sb_bread(inode->i_sb, nr);
879 * A read failure? Report error and clear slot
883 ext2_error(inode->i_sb, "ext2_free_branches",
884 "Read failure, inode=%ld, block=%ld",
888 ext2_free_branches(inode,
890 (__le32*)bh->b_data + addr_per_block,
893 ext2_free_blocks(inode, nr, 1);
894 mark_inode_dirty(inode);
897 ext2_free_data(inode, p, q);
900 void ext2_truncate (struct inode * inode)
902 __le32 *i_data = EXT2_I(inode)->i_data;
903 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
912 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
913 S_ISLNK(inode->i_mode)))
915 if (ext2_inode_is_fast_symlink(inode))
917 if (IS_APPEND(inode) || IS_IXORUNLINK(inode))
920 ext2_discard_prealloc(inode);
922 blocksize = inode->i_sb->s_blocksize;
923 iblock = (inode->i_size + blocksize-1)
924 >> EXT2_BLOCK_SIZE_BITS(inode->i_sb);
926 if (mapping_is_xip(inode->i_mapping))
927 xip_truncate_page(inode->i_mapping, inode->i_size);
928 else if (test_opt(inode->i_sb, NOBH))
929 nobh_truncate_page(inode->i_mapping, inode->i_size);
931 block_truncate_page(inode->i_mapping,
932 inode->i_size, ext2_get_block);
934 n = ext2_block_to_path(inode, iblock, offsets, NULL);
939 ext2_free_data(inode, i_data+offsets[0],
940 i_data + EXT2_NDIR_BLOCKS);
944 partial = ext2_find_shared(inode, n, offsets, chain, &nr);
945 /* Kill the top of shared branch (already detached) */
947 if (partial == chain)
948 mark_inode_dirty(inode);
950 mark_buffer_dirty_inode(partial->bh, inode);
951 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
953 /* Clear the ends of indirect blocks on the shared branch */
954 while (partial > chain) {
955 ext2_free_branches(inode,
957 (__le32*)partial->bh->b_data+addr_per_block,
958 (chain+n-1) - partial);
959 mark_buffer_dirty_inode(partial->bh, inode);
960 brelse (partial->bh);
964 /* Kill the remaining (whole) subtrees */
965 switch (offsets[0]) {
967 nr = i_data[EXT2_IND_BLOCK];
969 i_data[EXT2_IND_BLOCK] = 0;
970 mark_inode_dirty(inode);
971 ext2_free_branches(inode, &nr, &nr+1, 1);
974 nr = i_data[EXT2_DIND_BLOCK];
976 i_data[EXT2_DIND_BLOCK] = 0;
977 mark_inode_dirty(inode);
978 ext2_free_branches(inode, &nr, &nr+1, 2);
980 case EXT2_DIND_BLOCK:
981 nr = i_data[EXT2_TIND_BLOCK];
983 i_data[EXT2_TIND_BLOCK] = 0;
984 mark_inode_dirty(inode);
985 ext2_free_branches(inode, &nr, &nr+1, 3);
987 case EXT2_TIND_BLOCK:
990 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
991 if (inode_needs_sync(inode)) {
992 sync_mapping_buffers(inode->i_mapping);
993 ext2_sync_inode (inode);
995 mark_inode_dirty(inode);
999 static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
1000 struct buffer_head **p)
1002 struct buffer_head * bh;
1003 unsigned long block_group;
1004 unsigned long block;
1005 unsigned long offset;
1006 struct ext2_group_desc * gdp;
1009 if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) ||
1010 ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
1013 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
1014 gdp = ext2_get_group_desc(sb, block_group, &bh);
1018 * Figure out the offset within the block group inode table
1020 offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb);
1021 block = le32_to_cpu(gdp->bg_inode_table) +
1022 (offset >> EXT2_BLOCK_SIZE_BITS(sb));
1023 if (!(bh = sb_bread(sb, block)))
1027 offset &= (EXT2_BLOCK_SIZE(sb) - 1);
1028 return (struct ext2_inode *) (bh->b_data + offset);
1031 ext2_error(sb, "ext2_get_inode", "bad inode number: %lu",
1032 (unsigned long) ino);
1033 return ERR_PTR(-EINVAL);
1035 ext2_error(sb, "ext2_get_inode",
1036 "unable to read inode block - inode=%lu, block=%lu",
1037 (unsigned long) ino, block);
1039 return ERR_PTR(-EIO);
1042 void ext2_set_inode_flags(struct inode *inode)
1044 unsigned int flags = EXT2_I(inode)->i_flags;
1046 inode->i_flags &= ~(S_IMMUTABLE | S_IUNLINK | S_BARRIER |
1047 S_SYNC | S_APPEND | S_NOATIME | S_DIRSYNC);
1049 if (flags & EXT2_IMMUTABLE_FL)
1050 inode->i_flags |= S_IMMUTABLE;
1051 if (flags & EXT2_IUNLINK_FL)
1052 inode->i_flags |= S_IUNLINK;
1053 if (flags & EXT2_BARRIER_FL)
1054 inode->i_flags |= S_BARRIER;
1056 if (flags & EXT2_SYNC_FL)
1057 inode->i_flags |= S_SYNC;
1058 if (flags & EXT2_APPEND_FL)
1059 inode->i_flags |= S_APPEND;
1060 if (flags & EXT2_NOATIME_FL)
1061 inode->i_flags |= S_NOATIME;
1062 if (flags & EXT2_DIRSYNC_FL)
1063 inode->i_flags |= S_DIRSYNC;
1066 int ext2_sync_flags(struct inode *inode)
1068 unsigned int oldflags, newflags;
1070 oldflags = EXT2_I(inode)->i_flags;
1071 newflags = oldflags & ~(EXT2_IMMUTABLE_FL |
1072 EXT2_IUNLINK_FL | EXT2_BARRIER_FL);
1074 if (IS_IMMUTABLE(inode))
1075 newflags |= EXT2_IMMUTABLE_FL;
1076 if (IS_IUNLINK(inode))
1077 newflags |= EXT2_IUNLINK_FL;
1078 if (IS_BARRIER(inode))
1079 newflags |= EXT2_BARRIER_FL;
1081 if (oldflags ^ newflags) {
1082 EXT2_I(inode)->i_flags = newflags;
1083 inode->i_ctime = CURRENT_TIME;
1084 mark_inode_dirty(inode);
1089 void ext2_read_inode (struct inode * inode)
1091 struct ext2_inode_info *ei = EXT2_I(inode);
1092 ino_t ino = inode->i_ino;
1093 struct buffer_head * bh;
1094 struct ext2_inode * raw_inode = ext2_get_inode(inode->i_sb, ino, &bh);
1099 #ifdef CONFIG_EXT2_FS_POSIX_ACL
1100 ei->i_acl = EXT2_ACL_NOT_CACHED;
1101 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
1103 if (IS_ERR(raw_inode))
1106 inode->i_mode = le16_to_cpu(raw_inode->i_mode);
1107 uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
1108 gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
1109 if (!(test_opt (inode->i_sb, NO_UID32))) {
1110 uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
1111 gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
1113 inode->i_uid = INOTAG_UID(DX_TAG(inode), uid, gid);
1114 inode->i_gid = INOTAG_GID(DX_TAG(inode), uid, gid);
1115 inode->i_tag = INOTAG_TAG(DX_TAG(inode), uid, gid,
1116 le16_to_cpu(raw_inode->i_raw_tag));
1118 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
1119 inode->i_size = le32_to_cpu(raw_inode->i_size);
1120 inode->i_atime.tv_sec = le32_to_cpu(raw_inode->i_atime);
1121 inode->i_ctime.tv_sec = le32_to_cpu(raw_inode->i_ctime);
1122 inode->i_mtime.tv_sec = le32_to_cpu(raw_inode->i_mtime);
1123 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1124 ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
1125 /* We now have enough fields to check if the inode was active or not.
1126 * This is needed because nfsd might try to access dead inodes
1127 * the test is that same one that e2fsck uses
1128 * NeilBrown 1999oct15
1130 if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) {
1131 /* this inode is deleted */
1135 inode->i_blocks = le32_to_cpu(raw_inode->i_blocks);
1136 ei->i_flags = le32_to_cpu(raw_inode->i_flags);
1137 ei->i_faddr = le32_to_cpu(raw_inode->i_faddr);
1138 ei->i_frag_no = raw_inode->i_frag;
1139 ei->i_frag_size = raw_inode->i_fsize;
1140 ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl);
1142 if (S_ISREG(inode->i_mode))
1143 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32;
1145 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl);
1147 inode->i_generation = le32_to_cpu(raw_inode->i_generation);
1149 ei->i_next_alloc_block = 0;
1150 ei->i_next_alloc_goal = 0;
1151 ei->i_prealloc_count = 0;
1152 ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb);
1153 ei->i_dir_start_lookup = 0;
1156 * NOTE! The in-memory inode i_data array is in little-endian order
1157 * even on big-endian machines: we do NOT byteswap the block numbers!
1159 for (n = 0; n < EXT2_N_BLOCKS; n++)
1160 ei->i_data[n] = raw_inode->i_block[n];
1162 if (S_ISREG(inode->i_mode)) {
1163 inode->i_op = &ext2_file_inode_operations;
1164 if (ext2_use_xip(inode->i_sb)) {
1165 inode->i_mapping->a_ops = &ext2_aops_xip;
1166 inode->i_fop = &ext2_xip_file_operations;
1167 } else if (test_opt(inode->i_sb, NOBH)) {
1168 inode->i_mapping->a_ops = &ext2_nobh_aops;
1169 inode->i_fop = &ext2_file_operations;
1171 inode->i_mapping->a_ops = &ext2_aops;
1172 inode->i_fop = &ext2_file_operations;
1174 } else if (S_ISDIR(inode->i_mode)) {
1175 inode->i_op = &ext2_dir_inode_operations;
1176 inode->i_fop = &ext2_dir_operations;
1177 if (test_opt(inode->i_sb, NOBH))
1178 inode->i_mapping->a_ops = &ext2_nobh_aops;
1180 inode->i_mapping->a_ops = &ext2_aops;
1181 } else if (S_ISLNK(inode->i_mode)) {
1182 if (ext2_inode_is_fast_symlink(inode))
1183 inode->i_op = &ext2_fast_symlink_inode_operations;
1185 inode->i_op = &ext2_symlink_inode_operations;
1186 if (test_opt(inode->i_sb, NOBH))
1187 inode->i_mapping->a_ops = &ext2_nobh_aops;
1189 inode->i_mapping->a_ops = &ext2_aops;
1192 inode->i_op = &ext2_special_inode_operations;
1193 if (raw_inode->i_block[0])
1194 init_special_inode(inode, inode->i_mode,
1195 old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
1197 init_special_inode(inode, inode->i_mode,
1198 new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
1201 ext2_set_inode_flags(inode);
1205 make_bad_inode(inode);
1209 static int ext2_update_inode(struct inode * inode, int do_sync)
1211 struct ext2_inode_info *ei = EXT2_I(inode);
1212 struct super_block *sb = inode->i_sb;
1213 ino_t ino = inode->i_ino;
1214 uid_t uid = TAGINO_UID(DX_TAG(inode), inode->i_uid, inode->i_tag);
1215 gid_t gid = TAGINO_GID(DX_TAG(inode), inode->i_gid, inode->i_tag);
1216 struct buffer_head * bh;
1217 struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh);
1221 if (IS_ERR(raw_inode))
1224 /* For fields not not tracking in the in-memory inode,
1225 * initialise them to zero for new inodes. */
1226 if (ei->i_state & EXT2_STATE_NEW)
1227 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
1229 raw_inode->i_mode = cpu_to_le16(inode->i_mode);
1230 if (!(test_opt(sb, NO_UID32))) {
1231 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid));
1232 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid));
1234 * Fix up interoperability with old kernels. Otherwise, old inodes get
1235 * re-used with the upper 16 bits of the uid/gid intact
1238 raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid));
1239 raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid));
1241 raw_inode->i_uid_high = 0;
1242 raw_inode->i_gid_high = 0;
1245 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid));
1246 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid));
1247 raw_inode->i_uid_high = 0;
1248 raw_inode->i_gid_high = 0;
1250 #ifdef CONFIG_TAGGING_INTERN
1251 raw_inode->i_raw_tag = cpu_to_le16(inode->i_tag);
1253 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
1254 raw_inode->i_size = cpu_to_le32(inode->i_size);
1255 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1256 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1257 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1259 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
1260 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
1261 raw_inode->i_flags = cpu_to_le32(ei->i_flags);
1262 raw_inode->i_faddr = cpu_to_le32(ei->i_faddr);
1263 raw_inode->i_frag = ei->i_frag_no;
1264 raw_inode->i_fsize = ei->i_frag_size;
1265 raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl);
1266 if (!S_ISREG(inode->i_mode))
1267 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl);
1269 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32);
1270 if (inode->i_size > 0x7fffffffULL) {
1271 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb,
1272 EXT2_FEATURE_RO_COMPAT_LARGE_FILE) ||
1273 EXT2_SB(sb)->s_es->s_rev_level ==
1274 cpu_to_le32(EXT2_GOOD_OLD_REV)) {
1275 /* If this is the first large file
1276 * created, add a flag to the superblock.
1279 ext2_update_dynamic_rev(sb);
1280 EXT2_SET_RO_COMPAT_FEATURE(sb,
1281 EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
1283 ext2_write_super(sb);
1288 raw_inode->i_generation = cpu_to_le32(inode->i_generation);
1289 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1290 if (old_valid_dev(inode->i_rdev)) {
1291 raw_inode->i_block[0] =
1292 cpu_to_le32(old_encode_dev(inode->i_rdev));
1293 raw_inode->i_block[1] = 0;
1295 raw_inode->i_block[0] = 0;
1296 raw_inode->i_block[1] =
1297 cpu_to_le32(new_encode_dev(inode->i_rdev));
1298 raw_inode->i_block[2] = 0;
1300 } else for (n = 0; n < EXT2_N_BLOCKS; n++)
1301 raw_inode->i_block[n] = ei->i_data[n];
1302 mark_buffer_dirty(bh);
1304 sync_dirty_buffer(bh);
1305 if (buffer_req(bh) && !buffer_uptodate(bh)) {
1306 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1307 sb->s_id, (unsigned long) ino);
1311 ei->i_state &= ~EXT2_STATE_NEW;
1316 int ext2_write_inode(struct inode *inode, int wait)
1318 return ext2_update_inode(inode, wait);
1321 int ext2_sync_inode(struct inode *inode)
1323 struct writeback_control wbc = {
1324 .sync_mode = WB_SYNC_ALL,
1325 .nr_to_write = 0, /* sys_fsync did this */
1327 return sync_inode(inode, &wbc);
1330 int ext2_setattr(struct dentry *dentry, struct iattr *iattr)
1332 struct inode *inode = dentry->d_inode;
1335 error = inode_change_ok(inode, iattr);
1338 if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) ||
1339 (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid) ||
1340 (iattr->ia_valid & ATTR_TAG && iattr->ia_tag != inode->i_tag)) {
1341 error = DQUOT_TRANSFER(inode, iattr) ? -EDQUOT : 0;
1345 error = inode_setattr(inode, iattr);
1346 if (!error && (iattr->ia_valid & ATTR_MODE))
1347 error = ext2_acl_chmod(inode);