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>
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);
57 * Called at the last iput() if i_nlink is zero.
59 void ext2_delete_inode (struct inode * inode)
61 truncate_inode_pages(&inode->i_data, 0);
63 if (is_bad_inode(inode))
65 EXT2_I(inode)->i_dtime = get_seconds();
66 mark_inode_dirty(inode);
67 ext2_update_inode(inode, inode_needs_sync(inode));
71 ext2_truncate (inode);
72 ext2_free_inode (inode);
76 clear_inode(inode); /* We must guarantee clearing of inode... */
82 struct buffer_head *bh;
85 static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v)
91 static inline int verify_chain(Indirect *from, Indirect *to)
93 while (from <= to && from->key == *from->p)
99 * ext2_block_to_path - parse the block number into array of offsets
100 * @inode: inode in question (we are only interested in its superblock)
101 * @i_block: block number to be parsed
102 * @offsets: array to store the offsets in
103 * @boundary: set this non-zero if the referred-to block is likely to be
104 * followed (on disk) by an indirect block.
105 * To store the locations of file's data ext2 uses a data structure common
106 * for UNIX filesystems - tree of pointers anchored in the inode, with
107 * data blocks at leaves and indirect blocks in intermediate nodes.
108 * This function translates the block number into path in that tree -
109 * return value is the path length and @offsets[n] is the offset of
110 * pointer to (n+1)th node in the nth one. If @block is out of range
111 * (negative or too large) warning is printed and zero returned.
113 * Note: function doesn't find node addresses, so no IO is needed. All
114 * we need to know is the capacity of indirect blocks (taken from the
119 * Portability note: the last comparison (check that we fit into triple
120 * indirect block) is spelled differently, because otherwise on an
121 * architecture with 32-bit longs and 8Kb pages we might get into trouble
122 * if our filesystem had 8Kb blocks. We might use long long, but that would
123 * kill us on x86. Oh, well, at least the sign propagation does not matter -
124 * i_block would have to be negative in the very beginning, so we would not
128 static int ext2_block_to_path(struct inode *inode,
129 long i_block, int offsets[4], int *boundary)
131 int ptrs = EXT2_ADDR_PER_BLOCK(inode->i_sb);
132 int ptrs_bits = EXT2_ADDR_PER_BLOCK_BITS(inode->i_sb);
133 const long direct_blocks = EXT2_NDIR_BLOCKS,
134 indirect_blocks = ptrs,
135 double_blocks = (1 << (ptrs_bits * 2));
140 ext2_warning (inode->i_sb, "ext2_block_to_path", "block < 0");
141 } else if (i_block < direct_blocks) {
142 offsets[n++] = i_block;
143 final = direct_blocks;
144 } else if ( (i_block -= direct_blocks) < indirect_blocks) {
145 offsets[n++] = EXT2_IND_BLOCK;
146 offsets[n++] = i_block;
148 } else if ((i_block -= indirect_blocks) < double_blocks) {
149 offsets[n++] = EXT2_DIND_BLOCK;
150 offsets[n++] = i_block >> ptrs_bits;
151 offsets[n++] = i_block & (ptrs - 1);
153 } else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
154 offsets[n++] = EXT2_TIND_BLOCK;
155 offsets[n++] = i_block >> (ptrs_bits * 2);
156 offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
157 offsets[n++] = i_block & (ptrs - 1);
160 ext2_warning (inode->i_sb, "ext2_block_to_path", "block > big");
163 *boundary = final - 1 - (i_block & (ptrs - 1));
169 * ext2_get_branch - read the chain of indirect blocks leading to data
170 * @inode: inode in question
171 * @depth: depth of the chain (1 - direct pointer, etc.)
172 * @offsets: offsets of pointers in inode/indirect blocks
173 * @chain: place to store the result
174 * @err: here we store the error value
176 * Function fills the array of triples <key, p, bh> and returns %NULL
177 * if everything went OK or the pointer to the last filled triple
178 * (incomplete one) otherwise. Upon the return chain[i].key contains
179 * the number of (i+1)-th block in the chain (as it is stored in memory,
180 * i.e. little-endian 32-bit), chain[i].p contains the address of that
181 * number (it points into struct inode for i==0 and into the bh->b_data
182 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect
183 * block for i>0 and NULL for i==0. In other words, it holds the block
184 * numbers of the chain, addresses they were taken from (and where we can
185 * verify that chain did not change) and buffer_heads hosting these
188 * Function stops when it stumbles upon zero pointer (absent block)
189 * (pointer to last triple returned, *@err == 0)
190 * or when it gets an IO error reading an indirect block
191 * (ditto, *@err == -EIO)
192 * or when it notices that chain had been changed while it was reading
193 * (ditto, *@err == -EAGAIN)
194 * or when it reads all @depth-1 indirect blocks successfully and finds
195 * the whole chain, all way to the data (returns %NULL, *err == 0).
197 static Indirect *ext2_get_branch(struct inode *inode,
203 struct super_block *sb = inode->i_sb;
205 struct buffer_head *bh;
208 /* i_data is not going away, no lock needed */
209 add_chain (chain, NULL, EXT2_I(inode)->i_data + *offsets);
213 bh = sb_bread(sb, le32_to_cpu(p->key));
216 read_lock(&EXT2_I(inode)->i_meta_lock);
217 if (!verify_chain(chain, p))
219 add_chain(++p, bh, (__le32*)bh->b_data + *++offsets);
220 read_unlock(&EXT2_I(inode)->i_meta_lock);
227 read_unlock(&EXT2_I(inode)->i_meta_lock);
238 * ext2_find_near - find a place for allocation with sufficient locality
240 * @ind: descriptor of indirect block.
242 * This function returns the preferred place for block allocation.
243 * It is used when heuristic for sequential allocation fails.
245 * + if there is a block to the left of our position - allocate near it.
246 * + if pointer will live in indirect block - allocate near that block.
247 * + if pointer will live in inode - allocate in the same cylinder group.
249 * In the latter case we colour the starting block by the callers PID to
250 * prevent it from clashing with concurrent allocations for a different inode
251 * in the same block group. The PID is used here so that functionally related
252 * files will be close-by on-disk.
254 * Caller must make sure that @ind is valid and will stay that way.
257 static ext2_fsblk_t ext2_find_near(struct inode *inode, Indirect *ind)
259 struct ext2_inode_info *ei = EXT2_I(inode);
260 __le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data;
262 ext2_fsblk_t bg_start;
265 /* Try to find previous block */
266 for (p = ind->p - 1; p >= start; p--)
268 return le32_to_cpu(*p);
270 /* No such thing, so let's try location of indirect block */
272 return ind->bh->b_blocknr;
275 * It is going to be refered from inode itself? OK, just put it into
276 * the same cylinder group then.
278 bg_start = ext2_group_first_block_no(inode->i_sb, ei->i_block_group);
279 colour = (current->pid % 16) *
280 (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16);
281 return bg_start + colour;
285 * ext2_find_goal - find a preferred place for allocation.
287 * @block: block we want
288 * @partial: pointer to the last triple within a chain
290 * Returns preferred place for a block (the goal).
293 static inline ext2_fsblk_t ext2_find_goal(struct inode *inode, long block,
296 struct ext2_block_alloc_info *block_i;
298 block_i = EXT2_I(inode)->i_block_alloc_info;
301 * try the heuristic for sequential allocation,
302 * failing that at least try to get decent locality.
304 if (block_i && (block == block_i->last_alloc_logical_block + 1)
305 && (block_i->last_alloc_physical_block != 0)) {
306 return block_i->last_alloc_physical_block + 1;
309 return ext2_find_near(inode, partial);
313 * ext2_blks_to_allocate: Look up the block map and count the number
314 * of direct blocks need to be allocated for the given branch.
316 * @branch: chain of indirect blocks
317 * @k: number of blocks need for indirect blocks
318 * @blks: number of data blocks to be mapped.
319 * @blocks_to_boundary: the offset in the indirect block
321 * return the total number of blocks to be allocate, including the
322 * direct and indirect blocks.
325 ext2_blks_to_allocate(Indirect * branch, int k, unsigned long blks,
326 int blocks_to_boundary)
328 unsigned long count = 0;
331 * Simple case, [t,d]Indirect block(s) has not allocated yet
332 * then it's clear blocks on that path have not allocated
335 /* right now don't hanel cross boundary allocation */
336 if (blks < blocks_to_boundary + 1)
339 count += blocks_to_boundary + 1;
344 while (count < blks && count <= blocks_to_boundary
345 && le32_to_cpu(*(branch[0].p + count)) == 0) {
352 * ext2_alloc_blocks: multiple allocate blocks needed for a branch
353 * @indirect_blks: the number of blocks need to allocate for indirect
356 * @new_blocks: on return it will store the new block numbers for
357 * the indirect blocks(if needed) and the first direct block,
358 * @blks: on return it will store the total number of allocated
361 static int ext2_alloc_blocks(struct inode *inode,
362 ext2_fsblk_t goal, int indirect_blks, int blks,
363 ext2_fsblk_t new_blocks[4], int *err)
366 unsigned long count = 0;
368 ext2_fsblk_t current_block = 0;
372 * Here we try to allocate the requested multiple blocks at once,
373 * on a best-effort basis.
374 * To build a branch, we should allocate blocks for
375 * the indirect blocks(if not allocated yet), and at least
376 * the first direct block of this branch. That's the
377 * minimum number of blocks need to allocate(required)
379 target = blks + indirect_blks;
383 /* allocating blocks for indirect blocks and direct blocks */
384 current_block = ext2_new_blocks(inode,goal,&count,err);
389 /* allocate blocks for indirect blocks */
390 while (index < indirect_blks && count) {
391 new_blocks[index++] = current_block++;
399 /* save the new block number for the first direct block */
400 new_blocks[index] = current_block;
402 /* total number of blocks allocated for direct blocks */
407 for (i = 0; i <index; i++)
408 ext2_free_blocks(inode, new_blocks[i], 1);
413 * ext2_alloc_branch - allocate and set up a chain of blocks.
415 * @num: depth of the chain (number of blocks to allocate)
416 * @offsets: offsets (in the blocks) to store the pointers to next.
417 * @branch: place to store the chain in.
419 * This function allocates @num blocks, zeroes out all but the last one,
420 * links them into chain and (if we are synchronous) writes them to disk.
421 * In other words, it prepares a branch that can be spliced onto the
422 * inode. It stores the information about that chain in the branch[], in
423 * the same format as ext2_get_branch() would do. We are calling it after
424 * we had read the existing part of chain and partial points to the last
425 * triple of that (one with zero ->key). Upon the exit we have the same
426 * picture as after the successful ext2_get_block(), excpet that in one
427 * place chain is disconnected - *branch->p is still zero (we did not
428 * set the last link), but branch->key contains the number that should
429 * be placed into *branch->p to fill that gap.
431 * If allocation fails we free all blocks we've allocated (and forget
432 * their buffer_heads) and return the error value the from failed
433 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
434 * as described above and return 0.
437 static int ext2_alloc_branch(struct inode *inode,
438 int indirect_blks, int *blks, ext2_fsblk_t goal,
439 int *offsets, Indirect *branch)
441 int blocksize = inode->i_sb->s_blocksize;
444 struct buffer_head *bh;
446 ext2_fsblk_t new_blocks[4];
447 ext2_fsblk_t current_block;
449 num = ext2_alloc_blocks(inode, goal, indirect_blks,
450 *blks, new_blocks, &err);
454 branch[0].key = cpu_to_le32(new_blocks[0]);
456 * metadata blocks and data blocks are allocated.
458 for (n = 1; n <= indirect_blks; n++) {
460 * Get buffer_head for parent block, zero it out
461 * and set the pointer to new one, then send
464 bh = sb_getblk(inode->i_sb, new_blocks[n-1]);
467 memset(bh->b_data, 0, blocksize);
468 branch[n].p = (__le32 *) bh->b_data + offsets[n];
469 branch[n].key = cpu_to_le32(new_blocks[n]);
470 *branch[n].p = branch[n].key;
471 if ( n == indirect_blks) {
472 current_block = new_blocks[n];
474 * End of chain, update the last new metablock of
475 * the chain to point to the new allocated
476 * data blocks numbers
478 for (i=1; i < num; i++)
479 *(branch[n].p + i) = cpu_to_le32(++current_block);
481 set_buffer_uptodate(bh);
483 mark_buffer_dirty_inode(bh, inode);
484 /* We used to sync bh here if IS_SYNC(inode).
485 * But we now rely upon generic_osync_inode()
486 * and b_inode_buffers. But not for directories.
488 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
489 sync_dirty_buffer(bh);
496 * ext2_splice_branch - splice the allocated branch onto inode.
498 * @block: (logical) number of block we are adding
499 * @chain: chain of indirect blocks (with a missing link - see
501 * @where: location of missing link
502 * @num: number of indirect blocks we are adding
503 * @blks: number of direct blocks we are adding
505 * This function fills the missing link and does all housekeeping needed in
506 * inode (->i_blocks, etc.). In case of success we end up with the full
507 * chain to new block and return 0.
509 static void ext2_splice_branch(struct inode *inode,
510 long block, Indirect *where, int num, int blks)
513 struct ext2_block_alloc_info *block_i;
514 ext2_fsblk_t current_block;
516 block_i = EXT2_I(inode)->i_block_alloc_info;
518 /* XXX LOCKING probably should have i_meta_lock ?*/
521 *where->p = where->key;
524 * Update the host buffer_head or inode to point to more just allocated
525 * direct blocks blocks
527 if (num == 0 && blks > 1) {
528 current_block = le32_to_cpu(where->key) + 1;
529 for (i = 1; i < blks; i++)
530 *(where->p + i ) = cpu_to_le32(current_block++);
534 * update the most recently allocated logical & physical block
535 * in i_block_alloc_info, to assist find the proper goal block for next
539 block_i->last_alloc_logical_block = block + blks - 1;
540 block_i->last_alloc_physical_block =
541 le32_to_cpu(where[num].key) + blks - 1;
544 /* We are done with atomic stuff, now do the rest of housekeeping */
546 /* had we spliced it onto indirect block? */
548 mark_buffer_dirty_inode(where->bh, inode);
550 inode->i_ctime = CURRENT_TIME_SEC;
551 mark_inode_dirty(inode);
555 * Allocation strategy is simple: if we have to allocate something, we will
556 * have to go the whole way to leaf. So let's do it before attaching anything
557 * to tree, set linkage between the newborn blocks, write them if sync is
558 * required, recheck the path, free and repeat if check fails, otherwise
559 * set the last missing link (that will protect us from any truncate-generated
560 * removals - all blocks on the path are immune now) and possibly force the
561 * write on the parent block.
562 * That has a nice additional property: no special recovery from the failed
563 * allocations is needed - we simply release blocks and do not touch anything
564 * reachable from inode.
566 * `handle' can be NULL if create == 0.
568 * return > 0, # of blocks mapped or allocated.
569 * return = 0, if plain lookup failed.
570 * return < 0, error case.
572 static int ext2_get_blocks(struct inode *inode,
573 sector_t iblock, unsigned long maxblocks,
574 struct buffer_head *bh_result,
583 int blocks_to_boundary = 0;
585 struct ext2_inode_info *ei = EXT2_I(inode);
587 ext2_fsblk_t first_block = 0;
589 depth = ext2_block_to_path(inode,iblock,offsets,&blocks_to_boundary);
594 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
596 /* Simplest case - block found, no allocation needed */
598 first_block = le32_to_cpu(chain[depth - 1].key);
599 clear_buffer_new(bh_result); /* What's this do? */
602 while (count < maxblocks && count <= blocks_to_boundary) {
605 if (!verify_chain(chain, partial)) {
607 * Indirect block might be removed by
608 * truncate while we were reading it.
609 * Handling of that case: forget what we've
610 * got now, go to reread.
615 blk = le32_to_cpu(*(chain[depth-1].p + count));
616 if (blk == first_block + count)
624 /* Next simple case - plain lookup or failed read of indirect block */
625 if (!create || err == -EIO)
628 mutex_lock(&ei->truncate_mutex);
631 * Okay, we need to do block allocation. Lazily initialize the block
632 * allocation info here if necessary
634 if (S_ISREG(inode->i_mode) && (!ei->i_block_alloc_info))
635 ext2_init_block_alloc_info(inode);
637 goal = ext2_find_goal(inode, iblock, partial);
639 /* the number of blocks need to allocate for [d,t]indirect blocks */
640 indirect_blks = (chain + depth) - partial - 1;
642 * Next look up the indirect map to count the totoal number of
643 * direct blocks to allocate for this branch.
645 count = ext2_blks_to_allocate(partial, indirect_blks,
646 maxblocks, blocks_to_boundary);
648 * XXX ???? Block out ext2_truncate while we alter the tree
650 err = ext2_alloc_branch(inode, indirect_blks, &count, goal,
651 offsets + (partial - chain), partial);
654 mutex_unlock(&ei->truncate_mutex);
658 if (ext2_use_xip(inode->i_sb)) {
660 * we need to clear the block
662 err = ext2_clear_xip_target (inode,
663 le32_to_cpu(chain[depth-1].key));
665 mutex_unlock(&ei->truncate_mutex);
670 ext2_splice_branch(inode, iblock, partial, indirect_blks, count);
671 mutex_unlock(&ei->truncate_mutex);
672 set_buffer_new(bh_result);
674 map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
675 if (count > blocks_to_boundary)
676 set_buffer_boundary(bh_result);
678 /* Clean up and exit */
679 partial = chain + depth - 1; /* the whole chain */
681 while (partial > chain) {
687 while (partial > chain) {
694 int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
696 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
697 int ret = ext2_get_blocks(inode, iblock, max_blocks,
700 bh_result->b_size = (ret << inode->i_blkbits);
707 static int ext2_writepage(struct page *page, struct writeback_control *wbc)
709 return block_write_full_page(page, ext2_get_block, wbc);
712 static int ext2_readpage(struct file *file, struct page *page)
714 return mpage_readpage(page, ext2_get_block);
718 ext2_readpages(struct file *file, struct address_space *mapping,
719 struct list_head *pages, unsigned nr_pages)
721 return mpage_readpages(mapping, pages, nr_pages, ext2_get_block);
724 int __ext2_write_begin(struct file *file, struct address_space *mapping,
725 loff_t pos, unsigned len, unsigned flags,
726 struct page **pagep, void **fsdata)
728 return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
733 ext2_write_begin(struct file *file, struct address_space *mapping,
734 loff_t pos, unsigned len, unsigned flags,
735 struct page **pagep, void **fsdata)
738 return __ext2_write_begin(file, mapping, pos, len, flags, pagep,fsdata);
742 ext2_nobh_write_begin(struct file *file, struct address_space *mapping,
743 loff_t pos, unsigned len, unsigned flags,
744 struct page **pagep, void **fsdata)
747 * Dir-in-pagecache still uses ext2_write_begin. Would have to rework
748 * directory handling code to pass around offsets rather than struct
749 * pages in order to make this work easily.
751 return nobh_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
755 static int ext2_nobh_writepage(struct page *page,
756 struct writeback_control *wbc)
758 return nobh_writepage(page, ext2_get_block, wbc);
761 static sector_t ext2_bmap(struct address_space *mapping, sector_t block)
763 return generic_block_bmap(mapping,block,ext2_get_block);
767 ext2_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
768 loff_t offset, unsigned long nr_segs)
770 struct file *file = iocb->ki_filp;
771 struct inode *inode = file->f_mapping->host;
773 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
774 offset, nr_segs, ext2_get_block, NULL);
778 ext2_writepages(struct address_space *mapping, struct writeback_control *wbc)
780 return mpage_writepages(mapping, wbc, ext2_get_block);
783 const struct address_space_operations ext2_aops = {
784 .readpage = ext2_readpage,
785 .readpages = ext2_readpages,
786 .writepage = ext2_writepage,
787 .sync_page = block_sync_page,
788 .write_begin = ext2_write_begin,
789 .write_end = generic_write_end,
791 .direct_IO = ext2_direct_IO,
792 .writepages = ext2_writepages,
793 .migratepage = buffer_migrate_page,
794 .is_partially_uptodate = block_is_partially_uptodate,
797 const struct address_space_operations ext2_aops_xip = {
799 .get_xip_mem = ext2_get_xip_mem,
802 const struct address_space_operations ext2_nobh_aops = {
803 .readpage = ext2_readpage,
804 .readpages = ext2_readpages,
805 .writepage = ext2_nobh_writepage,
806 .sync_page = block_sync_page,
807 .write_begin = ext2_nobh_write_begin,
808 .write_end = nobh_write_end,
810 .direct_IO = ext2_direct_IO,
811 .writepages = ext2_writepages,
812 .migratepage = buffer_migrate_page,
816 * Probably it should be a library function... search for first non-zero word
817 * or memcmp with zero_page, whatever is better for particular architecture.
820 static inline int all_zeroes(__le32 *p, __le32 *q)
829 * ext2_find_shared - find the indirect blocks for partial truncation.
830 * @inode: inode in question
831 * @depth: depth of the affected branch
832 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
833 * @chain: place to store the pointers to partial indirect blocks
834 * @top: place to the (detached) top of branch
836 * This is a helper function used by ext2_truncate().
838 * When we do truncate() we may have to clean the ends of several indirect
839 * blocks but leave the blocks themselves alive. Block is partially
840 * truncated if some data below the new i_size is refered from it (and
841 * it is on the path to the first completely truncated data block, indeed).
842 * We have to free the top of that path along with everything to the right
843 * of the path. Since no allocation past the truncation point is possible
844 * until ext2_truncate() finishes, we may safely do the latter, but top
845 * of branch may require special attention - pageout below the truncation
846 * point might try to populate it.
848 * We atomically detach the top of branch from the tree, store the block
849 * number of its root in *@top, pointers to buffer_heads of partially
850 * truncated blocks - in @chain[].bh and pointers to their last elements
851 * that should not be removed - in @chain[].p. Return value is the pointer
852 * to last filled element of @chain.
854 * The work left to caller to do the actual freeing of subtrees:
855 * a) free the subtree starting from *@top
856 * b) free the subtrees whose roots are stored in
857 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
858 * c) free the subtrees growing from the inode past the @chain[0].p
859 * (no partially truncated stuff there).
862 static Indirect *ext2_find_shared(struct inode *inode,
868 Indirect *partial, *p;
872 for (k = depth; k > 1 && !offsets[k-1]; k--)
874 partial = ext2_get_branch(inode, k, offsets, chain, &err);
876 partial = chain + k-1;
878 * If the branch acquired continuation since we've looked at it -
879 * fine, it should all survive and (new) top doesn't belong to us.
881 write_lock(&EXT2_I(inode)->i_meta_lock);
882 if (!partial->key && *partial->p) {
883 write_unlock(&EXT2_I(inode)->i_meta_lock);
886 for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--)
889 * OK, we've found the last block that must survive. The rest of our
890 * branch should be detached before unlocking. However, if that rest
891 * of branch is all ours and does not grow immediately from the inode
892 * it's easier to cheat and just decrement partial->p.
894 if (p == chain + k - 1 && p > chain) {
900 write_unlock(&EXT2_I(inode)->i_meta_lock);
912 * ext2_free_data - free a list of data blocks
913 * @inode: inode we are dealing with
914 * @p: array of block numbers
915 * @q: points immediately past the end of array
917 * We are freeing all blocks refered from that array (numbers are
918 * stored as little-endian 32-bit) and updating @inode->i_blocks
921 static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q)
923 unsigned long block_to_free = 0, count = 0;
926 for ( ; p < q ; p++) {
927 nr = le32_to_cpu(*p);
930 /* accumulate blocks to free if they're contiguous */
933 else if (block_to_free == nr - count)
936 mark_inode_dirty(inode);
937 ext2_free_blocks (inode, block_to_free, count);
945 mark_inode_dirty(inode);
946 ext2_free_blocks (inode, block_to_free, count);
951 * ext2_free_branches - free an array of branches
952 * @inode: inode we are dealing with
953 * @p: array of block numbers
954 * @q: pointer immediately past the end of array
955 * @depth: depth of the branches to free
957 * We are freeing all blocks refered from these branches (numbers are
958 * stored as little-endian 32-bit) and updating @inode->i_blocks
961 static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth)
963 struct buffer_head * bh;
967 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
968 for ( ; p < q ; p++) {
969 nr = le32_to_cpu(*p);
973 bh = sb_bread(inode->i_sb, nr);
975 * A read failure? Report error and clear slot
979 ext2_error(inode->i_sb, "ext2_free_branches",
980 "Read failure, inode=%ld, block=%ld",
984 ext2_free_branches(inode,
986 (__le32*)bh->b_data + addr_per_block,
989 ext2_free_blocks(inode, nr, 1);
990 mark_inode_dirty(inode);
993 ext2_free_data(inode, p, q);
996 void ext2_truncate(struct inode *inode)
998 __le32 *i_data = EXT2_I(inode)->i_data;
999 struct ext2_inode_info *ei = EXT2_I(inode);
1000 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
1009 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1010 S_ISLNK(inode->i_mode)))
1012 if (ext2_inode_is_fast_symlink(inode))
1014 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1017 blocksize = inode->i_sb->s_blocksize;
1018 iblock = (inode->i_size + blocksize-1)
1019 >> EXT2_BLOCK_SIZE_BITS(inode->i_sb);
1021 if (mapping_is_xip(inode->i_mapping))
1022 xip_truncate_page(inode->i_mapping, inode->i_size);
1023 else if (test_opt(inode->i_sb, NOBH))
1024 nobh_truncate_page(inode->i_mapping,
1025 inode->i_size, ext2_get_block);
1027 block_truncate_page(inode->i_mapping,
1028 inode->i_size, ext2_get_block);
1030 n = ext2_block_to_path(inode, iblock, offsets, NULL);
1035 * From here we block out all ext2_get_block() callers who want to
1036 * modify the block allocation tree.
1038 mutex_lock(&ei->truncate_mutex);
1041 ext2_free_data(inode, i_data+offsets[0],
1042 i_data + EXT2_NDIR_BLOCKS);
1046 partial = ext2_find_shared(inode, n, offsets, chain, &nr);
1047 /* Kill the top of shared branch (already detached) */
1049 if (partial == chain)
1050 mark_inode_dirty(inode);
1052 mark_buffer_dirty_inode(partial->bh, inode);
1053 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
1055 /* Clear the ends of indirect blocks on the shared branch */
1056 while (partial > chain) {
1057 ext2_free_branches(inode,
1059 (__le32*)partial->bh->b_data+addr_per_block,
1060 (chain+n-1) - partial);
1061 mark_buffer_dirty_inode(partial->bh, inode);
1062 brelse (partial->bh);
1066 /* Kill the remaining (whole) subtrees */
1067 switch (offsets[0]) {
1069 nr = i_data[EXT2_IND_BLOCK];
1071 i_data[EXT2_IND_BLOCK] = 0;
1072 mark_inode_dirty(inode);
1073 ext2_free_branches(inode, &nr, &nr+1, 1);
1075 case EXT2_IND_BLOCK:
1076 nr = i_data[EXT2_DIND_BLOCK];
1078 i_data[EXT2_DIND_BLOCK] = 0;
1079 mark_inode_dirty(inode);
1080 ext2_free_branches(inode, &nr, &nr+1, 2);
1082 case EXT2_DIND_BLOCK:
1083 nr = i_data[EXT2_TIND_BLOCK];
1085 i_data[EXT2_TIND_BLOCK] = 0;
1086 mark_inode_dirty(inode);
1087 ext2_free_branches(inode, &nr, &nr+1, 3);
1089 case EXT2_TIND_BLOCK:
1093 ext2_discard_reservation(inode);
1095 mutex_unlock(&ei->truncate_mutex);
1096 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
1097 if (inode_needs_sync(inode)) {
1098 sync_mapping_buffers(inode->i_mapping);
1099 ext2_sync_inode (inode);
1101 mark_inode_dirty(inode);
1105 static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
1106 struct buffer_head **p)
1108 struct buffer_head * bh;
1109 unsigned long block_group;
1110 unsigned long block;
1111 unsigned long offset;
1112 struct ext2_group_desc * gdp;
1115 if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) ||
1116 ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
1119 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
1120 gdp = ext2_get_group_desc(sb, block_group, NULL);
1124 * Figure out the offset within the block group inode table
1126 offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb);
1127 block = le32_to_cpu(gdp->bg_inode_table) +
1128 (offset >> EXT2_BLOCK_SIZE_BITS(sb));
1129 if (!(bh = sb_bread(sb, block)))
1133 offset &= (EXT2_BLOCK_SIZE(sb) - 1);
1134 return (struct ext2_inode *) (bh->b_data + offset);
1137 ext2_error(sb, "ext2_get_inode", "bad inode number: %lu",
1138 (unsigned long) ino);
1139 return ERR_PTR(-EINVAL);
1141 ext2_error(sb, "ext2_get_inode",
1142 "unable to read inode block - inode=%lu, block=%lu",
1143 (unsigned long) ino, block);
1145 return ERR_PTR(-EIO);
1148 void ext2_set_inode_flags(struct inode *inode)
1150 unsigned int flags = EXT2_I(inode)->i_flags;
1152 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
1153 if (flags & EXT2_SYNC_FL)
1154 inode->i_flags |= S_SYNC;
1155 if (flags & EXT2_APPEND_FL)
1156 inode->i_flags |= S_APPEND;
1157 if (flags & EXT2_IMMUTABLE_FL)
1158 inode->i_flags |= S_IMMUTABLE;
1159 if (flags & EXT2_NOATIME_FL)
1160 inode->i_flags |= S_NOATIME;
1161 if (flags & EXT2_DIRSYNC_FL)
1162 inode->i_flags |= S_DIRSYNC;
1165 /* Propagate flags from i_flags to EXT2_I(inode)->i_flags */
1166 void ext2_get_inode_flags(struct ext2_inode_info *ei)
1168 unsigned int flags = ei->vfs_inode.i_flags;
1170 ei->i_flags &= ~(EXT2_SYNC_FL|EXT2_APPEND_FL|
1171 EXT2_IMMUTABLE_FL|EXT2_NOATIME_FL|EXT2_DIRSYNC_FL);
1173 ei->i_flags |= EXT2_SYNC_FL;
1174 if (flags & S_APPEND)
1175 ei->i_flags |= EXT2_APPEND_FL;
1176 if (flags & S_IMMUTABLE)
1177 ei->i_flags |= EXT2_IMMUTABLE_FL;
1178 if (flags & S_NOATIME)
1179 ei->i_flags |= EXT2_NOATIME_FL;
1180 if (flags & S_DIRSYNC)
1181 ei->i_flags |= EXT2_DIRSYNC_FL;
1184 struct inode *ext2_iget (struct super_block *sb, unsigned long ino)
1186 struct ext2_inode_info *ei;
1187 struct buffer_head * bh;
1188 struct ext2_inode *raw_inode;
1189 struct inode *inode;
1193 inode = iget_locked(sb, ino);
1195 return ERR_PTR(-ENOMEM);
1196 if (!(inode->i_state & I_NEW))
1200 #ifdef CONFIG_EXT2_FS_POSIX_ACL
1201 ei->i_acl = EXT2_ACL_NOT_CACHED;
1202 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
1204 ei->i_block_alloc_info = NULL;
1206 raw_inode = ext2_get_inode(inode->i_sb, ino, &bh);
1207 if (IS_ERR(raw_inode)) {
1208 ret = PTR_ERR(raw_inode);
1212 inode->i_mode = le16_to_cpu(raw_inode->i_mode);
1213 inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
1214 inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
1215 if (!(test_opt (inode->i_sb, NO_UID32))) {
1216 inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
1217 inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
1219 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
1220 inode->i_size = le32_to_cpu(raw_inode->i_size);
1221 inode->i_atime.tv_sec = (signed)le32_to_cpu(raw_inode->i_atime);
1222 inode->i_ctime.tv_sec = (signed)le32_to_cpu(raw_inode->i_ctime);
1223 inode->i_mtime.tv_sec = (signed)le32_to_cpu(raw_inode->i_mtime);
1224 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1225 ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
1226 /* We now have enough fields to check if the inode was active or not.
1227 * This is needed because nfsd might try to access dead inodes
1228 * the test is that same one that e2fsck uses
1229 * NeilBrown 1999oct15
1231 if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) {
1232 /* this inode is deleted */
1237 inode->i_blocks = le32_to_cpu(raw_inode->i_blocks);
1238 ei->i_flags = le32_to_cpu(raw_inode->i_flags);
1239 ei->i_faddr = le32_to_cpu(raw_inode->i_faddr);
1240 ei->i_frag_no = raw_inode->i_frag;
1241 ei->i_frag_size = raw_inode->i_fsize;
1242 ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl);
1244 if (S_ISREG(inode->i_mode))
1245 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32;
1247 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl);
1249 inode->i_generation = le32_to_cpu(raw_inode->i_generation);
1251 ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb);
1252 ei->i_dir_start_lookup = 0;
1255 * NOTE! The in-memory inode i_data array is in little-endian order
1256 * even on big-endian machines: we do NOT byteswap the block numbers!
1258 for (n = 0; n < EXT2_N_BLOCKS; n++)
1259 ei->i_data[n] = raw_inode->i_block[n];
1261 if (S_ISREG(inode->i_mode)) {
1262 inode->i_op = &ext2_file_inode_operations;
1263 if (ext2_use_xip(inode->i_sb)) {
1264 inode->i_mapping->a_ops = &ext2_aops_xip;
1265 inode->i_fop = &ext2_xip_file_operations;
1266 } else if (test_opt(inode->i_sb, NOBH)) {
1267 inode->i_mapping->a_ops = &ext2_nobh_aops;
1268 inode->i_fop = &ext2_file_operations;
1270 inode->i_mapping->a_ops = &ext2_aops;
1271 inode->i_fop = &ext2_file_operations;
1273 } else if (S_ISDIR(inode->i_mode)) {
1274 inode->i_op = &ext2_dir_inode_operations;
1275 inode->i_fop = &ext2_dir_operations;
1276 if (test_opt(inode->i_sb, NOBH))
1277 inode->i_mapping->a_ops = &ext2_nobh_aops;
1279 inode->i_mapping->a_ops = &ext2_aops;
1280 } else if (S_ISLNK(inode->i_mode)) {
1281 if (ext2_inode_is_fast_symlink(inode))
1282 inode->i_op = &ext2_fast_symlink_inode_operations;
1284 inode->i_op = &ext2_symlink_inode_operations;
1285 if (test_opt(inode->i_sb, NOBH))
1286 inode->i_mapping->a_ops = &ext2_nobh_aops;
1288 inode->i_mapping->a_ops = &ext2_aops;
1291 inode->i_op = &ext2_special_inode_operations;
1292 if (raw_inode->i_block[0])
1293 init_special_inode(inode, inode->i_mode,
1294 old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
1296 init_special_inode(inode, inode->i_mode,
1297 new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
1300 ext2_set_inode_flags(inode);
1301 unlock_new_inode(inode);
1306 return ERR_PTR(ret);
1309 static int ext2_update_inode(struct inode * inode, int do_sync)
1311 struct ext2_inode_info *ei = EXT2_I(inode);
1312 struct super_block *sb = inode->i_sb;
1313 ino_t ino = inode->i_ino;
1314 uid_t uid = inode->i_uid;
1315 gid_t gid = inode->i_gid;
1316 struct buffer_head * bh;
1317 struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh);
1321 if (IS_ERR(raw_inode))
1324 /* For fields not not tracking in the in-memory inode,
1325 * initialise them to zero for new inodes. */
1326 if (ei->i_state & EXT2_STATE_NEW)
1327 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
1329 ext2_get_inode_flags(ei);
1330 raw_inode->i_mode = cpu_to_le16(inode->i_mode);
1331 if (!(test_opt(sb, NO_UID32))) {
1332 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid));
1333 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid));
1335 * Fix up interoperability with old kernels. Otherwise, old inodes get
1336 * re-used with the upper 16 bits of the uid/gid intact
1339 raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid));
1340 raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid));
1342 raw_inode->i_uid_high = 0;
1343 raw_inode->i_gid_high = 0;
1346 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid));
1347 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid));
1348 raw_inode->i_uid_high = 0;
1349 raw_inode->i_gid_high = 0;
1351 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
1352 raw_inode->i_size = cpu_to_le32(inode->i_size);
1353 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1354 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1355 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1357 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
1358 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
1359 raw_inode->i_flags = cpu_to_le32(ei->i_flags);
1360 raw_inode->i_faddr = cpu_to_le32(ei->i_faddr);
1361 raw_inode->i_frag = ei->i_frag_no;
1362 raw_inode->i_fsize = ei->i_frag_size;
1363 raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl);
1364 if (!S_ISREG(inode->i_mode))
1365 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl);
1367 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32);
1368 if (inode->i_size > 0x7fffffffULL) {
1369 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb,
1370 EXT2_FEATURE_RO_COMPAT_LARGE_FILE) ||
1371 EXT2_SB(sb)->s_es->s_rev_level ==
1372 cpu_to_le32(EXT2_GOOD_OLD_REV)) {
1373 /* If this is the first large file
1374 * created, add a flag to the superblock.
1377 ext2_update_dynamic_rev(sb);
1378 EXT2_SET_RO_COMPAT_FEATURE(sb,
1379 EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
1381 ext2_write_super(sb);
1386 raw_inode->i_generation = cpu_to_le32(inode->i_generation);
1387 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1388 if (old_valid_dev(inode->i_rdev)) {
1389 raw_inode->i_block[0] =
1390 cpu_to_le32(old_encode_dev(inode->i_rdev));
1391 raw_inode->i_block[1] = 0;
1393 raw_inode->i_block[0] = 0;
1394 raw_inode->i_block[1] =
1395 cpu_to_le32(new_encode_dev(inode->i_rdev));
1396 raw_inode->i_block[2] = 0;
1398 } else for (n = 0; n < EXT2_N_BLOCKS; n++)
1399 raw_inode->i_block[n] = ei->i_data[n];
1400 mark_buffer_dirty(bh);
1402 sync_dirty_buffer(bh);
1403 if (buffer_req(bh) && !buffer_uptodate(bh)) {
1404 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1405 sb->s_id, (unsigned long) ino);
1409 ei->i_state &= ~EXT2_STATE_NEW;
1414 int ext2_write_inode(struct inode *inode, int wait)
1416 return ext2_update_inode(inode, wait);
1419 int ext2_sync_inode(struct inode *inode)
1421 struct writeback_control wbc = {
1422 .sync_mode = WB_SYNC_ALL,
1423 .nr_to_write = 0, /* sys_fsync did this */
1425 return sync_inode(inode, &wbc);
1428 int ext2_setattr(struct dentry *dentry, struct iattr *iattr)
1430 struct inode *inode = dentry->d_inode;
1433 error = inode_change_ok(inode, iattr);
1436 if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) ||
1437 (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid)) {
1438 error = DQUOT_TRANSFER(inode, iattr) ? -EDQUOT : 0;
1442 error = inode_setattr(inode, iattr);
1443 if (!error && (iattr->ia_valid & ATTR_MODE))
1444 error = ext2_acl_chmod(inode);