2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
5 #include <linux/time.h>
7 #include <linux/reiserfs_fs.h>
8 #include <linux/reiserfs_acl.h>
9 #include <linux/reiserfs_xattr.h>
10 #include <linux/exportfs.h>
11 #include <linux/smp_lock.h>
12 #include <linux/pagemap.h>
13 #include <linux/highmem.h>
14 #include <linux/slab.h>
15 #include <asm/uaccess.h>
16 #include <asm/unaligned.h>
17 #include <linux/buffer_head.h>
18 #include <linux/mpage.h>
19 #include <linux/writeback.h>
20 #include <linux/quotaops.h>
21 #include <linux/swap.h>
23 int reiserfs_commit_write(struct file *f, struct page *page,
24 unsigned from, unsigned to);
25 int reiserfs_prepare_write(struct file *f, struct page *page,
26 unsigned from, unsigned to);
28 void reiserfs_delete_inode(struct inode *inode)
30 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
32 JOURNAL_PER_BALANCE_CNT * 2 +
33 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
34 struct reiserfs_transaction_handle th;
38 if (!is_bad_inode(inode))
39 dquot_initialize(inode);
41 truncate_inode_pages(&inode->i_data, 0);
43 depth = reiserfs_write_lock_once(inode->i_sb);
45 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
46 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
47 reiserfs_delete_xattrs(inode);
49 if (journal_begin(&th, inode->i_sb, jbegin_count))
51 reiserfs_update_inode_transaction(inode);
53 reiserfs_discard_prealloc(&th, inode);
55 err = reiserfs_delete_object(&th, inode);
57 /* Do quota update inside a transaction for journaled quotas. We must do that
58 * after delete_object so that quota updates go into the same transaction as
59 * stat data deletion */
61 dquot_free_inode(inode);
63 if (journal_end(&th, inode->i_sb, jbegin_count))
66 /* check return value from reiserfs_delete_object after
67 * ending the transaction
72 /* all items of file are deleted, so we can remove "save" link */
73 remove_save_link(inode, 0 /* not truncate */ ); /* we can't do anything
74 * about an error here */
76 /* no object items are in the tree */
80 clear_inode(inode); /* note this must go after the journal_end to prevent deadlock */
82 reiserfs_write_unlock_once(inode->i_sb, depth);
85 static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
86 __u32 objectid, loff_t offset, int type, int length)
88 key->version = version;
90 key->on_disk_key.k_dir_id = dirid;
91 key->on_disk_key.k_objectid = objectid;
92 set_cpu_key_k_offset(key, offset);
93 set_cpu_key_k_type(key, type);
94 key->key_length = length;
97 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
98 offset and type of key */
99 void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
100 int type, int length)
102 _make_cpu_key(key, get_inode_item_key_version(inode),
103 le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
104 le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
109 // when key is 0, do not set version and short key
111 inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
113 loff_t offset, int type, int length,
114 int entry_count /*or ih_free_space */ )
117 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
118 ih->ih_key.k_objectid =
119 cpu_to_le32(key->on_disk_key.k_objectid);
121 put_ih_version(ih, version);
122 set_le_ih_k_offset(ih, offset);
123 set_le_ih_k_type(ih, type);
124 put_ih_item_len(ih, length);
125 /* set_ih_free_space (ih, 0); */
126 // for directory items it is entry count, for directs and stat
127 // datas - 0xffff, for indirects - 0
128 put_ih_entry_count(ih, entry_count);
132 // FIXME: we might cache recently accessed indirect item
134 // Ugh. Not too eager for that....
135 // I cut the code until such time as I see a convincing argument (benchmark).
136 // I don't want a bloated inode struct..., and I don't like code complexity....
138 /* cutting the code is fine, since it really isn't in use yet and is easy
139 ** to add back in. But, Vladimir has a really good idea here. Think
140 ** about what happens for reading a file. For each page,
141 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
142 ** an indirect item. This indirect item has X number of pointers, where
143 ** X is a big number if we've done the block allocation right. But,
144 ** we only use one or two of these pointers during each call to readpage,
145 ** needlessly researching again later on.
147 ** The size of the cache could be dynamic based on the size of the file.
149 ** I'd also like to see us cache the location the stat data item, since
150 ** we are needlessly researching for that frequently.
155 /* If this page has a file tail in it, and
156 ** it was read in by get_block_create_0, the page data is valid,
157 ** but tail is still sitting in a direct item, and we can't write to
158 ** it. So, look through this page, and check all the mapped buffers
159 ** to make sure they have valid block numbers. Any that don't need
160 ** to be unmapped, so that block_prepare_write will correctly call
161 ** reiserfs_get_block to convert the tail into an unformatted node
163 static inline void fix_tail_page_for_writing(struct page *page)
165 struct buffer_head *head, *next, *bh;
167 if (page && page_has_buffers(page)) {
168 head = page_buffers(page);
171 next = bh->b_this_page;
172 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
173 reiserfs_unmap_buffer(bh);
176 } while (bh != head);
180 /* reiserfs_get_block does not need to allocate a block only if it has been
181 done already or non-hole position has been found in the indirect item */
182 static inline int allocation_needed(int retval, b_blocknr_t allocated,
183 struct item_head *ih,
184 __le32 * item, int pos_in_item)
188 if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
189 get_block_num(item, pos_in_item))
194 static inline int indirect_item_found(int retval, struct item_head *ih)
196 return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
199 static inline void set_block_dev_mapped(struct buffer_head *bh,
200 b_blocknr_t block, struct inode *inode)
202 map_bh(bh, inode->i_sb, block);
206 // files which were created in the earlier version can not be longer,
209 static int file_capable(struct inode *inode, sector_t block)
211 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file.
212 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
218 static int restart_transaction(struct reiserfs_transaction_handle *th,
219 struct inode *inode, struct treepath *path)
221 struct super_block *s = th->t_super;
222 int len = th->t_blocks_allocated;
225 BUG_ON(!th->t_trans_id);
226 BUG_ON(!th->t_refcount);
230 /* we cannot restart while nested */
231 if (th->t_refcount > 1) {
234 reiserfs_update_sd(th, inode);
235 err = journal_end(th, s, len);
237 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
239 reiserfs_update_inode_transaction(inode);
244 // it is called by get_block when create == 0. Returns block number
245 // for 'block'-th logical block of file. When it hits direct item it
246 // returns 0 (being called from bmap) or read direct item into piece
247 // of page (bh_result)
249 // Please improve the english/clarity in the comment above, as it is
250 // hard to understand.
252 static int _get_block_create_0(struct inode *inode, sector_t block,
253 struct buffer_head *bh_result, int args)
255 INITIALIZE_PATH(path);
257 struct buffer_head *bh;
258 struct item_head *ih, tmp_ih;
265 unsigned long offset;
267 // prepare the key to look for the 'block'-th block of file
268 make_cpu_key(&key, inode,
269 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
272 result = search_for_position_by_key(inode->i_sb, &key, &path);
273 if (result != POSITION_FOUND) {
276 kunmap(bh_result->b_page);
277 if (result == IO_ERROR)
279 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
280 // That there is some MMAPED data associated with it that is yet to be written to disk.
281 if ((args & GET_BLOCK_NO_HOLE)
282 && !PageUptodate(bh_result->b_page)) {
288 bh = get_last_bh(&path);
290 if (is_indirect_le_ih(ih)) {
291 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih);
293 /* FIXME: here we could cache indirect item or part of it in
294 the inode to avoid search_by_key in case of subsequent
296 blocknr = get_block_num(ind_item, path.pos_in_item);
299 map_bh(bh_result, inode->i_sb, blocknr);
300 if (path.pos_in_item ==
301 ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
302 set_buffer_boundary(bh_result);
305 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
306 // That there is some MMAPED data associated with it that is yet to be written to disk.
307 if ((args & GET_BLOCK_NO_HOLE)
308 && !PageUptodate(bh_result->b_page)) {
314 kunmap(bh_result->b_page);
317 // requested data are in direct item(s)
318 if (!(args & GET_BLOCK_READ_DIRECT)) {
319 // we are called by bmap. FIXME: we can not map block of file
320 // when it is stored in direct item(s)
323 kunmap(bh_result->b_page);
327 /* if we've got a direct item, and the buffer or page was uptodate,
328 ** we don't want to pull data off disk again. skip to the
329 ** end, where we map the buffer and return
331 if (buffer_uptodate(bh_result)) {
335 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
336 ** pages without any buffers. If the page is up to date, we don't want
337 ** read old data off disk. Set the up to date bit on the buffer instead
338 ** and jump to the end
340 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
341 set_buffer_uptodate(bh_result);
344 // read file tail into part of page
345 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
346 copy_item_head(&tmp_ih, ih);
348 /* we only want to kmap if we are reading the tail into the page.
349 ** this is not the common case, so we don't kmap until we are
350 ** sure we need to. But, this means the item might move if
354 p = (char *)kmap(bh_result->b_page);
357 memset(p, 0, inode->i_sb->s_blocksize);
359 if (!is_direct_le_ih(ih)) {
362 /* make sure we don't read more bytes than actually exist in
363 ** the file. This can happen in odd cases where i_size isn't
364 ** correct, and when direct item padding results in a few
365 ** extra bytes at the end of the direct item
367 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
369 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
371 inode->i_size - (le_ih_k_offset(ih) - 1) -
375 chars = ih_item_len(ih) - path.pos_in_item;
377 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
384 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
385 // we done, if read direct item is not the last item of
386 // node FIXME: we could try to check right delimiting key
387 // to see whether direct item continues in the right
388 // neighbor or rely on i_size
391 // update key to look for the next piece
392 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
393 result = search_for_position_by_key(inode->i_sb, &key, &path);
394 if (result != POSITION_FOUND)
395 // i/o error most likely
397 bh = get_last_bh(&path);
401 flush_dcache_page(bh_result->b_page);
402 kunmap(bh_result->b_page);
407 if (result == IO_ERROR)
410 /* this buffer has valid data, but isn't valid for io. mapping it to
411 * block #0 tells the rest of reiserfs it just has a tail in it
413 map_bh(bh_result, inode->i_sb, 0);
414 set_buffer_uptodate(bh_result);
418 // this is called to create file map. So, _get_block_create_0 will not
420 static int reiserfs_bmap(struct inode *inode, sector_t block,
421 struct buffer_head *bh_result, int create)
423 if (!file_capable(inode, block))
426 reiserfs_write_lock(inode->i_sb);
427 /* do not read the direct item */
428 _get_block_create_0(inode, block, bh_result, 0);
429 reiserfs_write_unlock(inode->i_sb);
433 /* special version of get_block that is only used by grab_tail_page right
434 ** now. It is sent to block_prepare_write, and when you try to get a
435 ** block past the end of the file (or a block from a hole) it returns
436 ** -ENOENT instead of a valid buffer. block_prepare_write expects to
437 ** be able to do i/o on the buffers returned, unless an error value
440 ** So, this allows block_prepare_write to be used for reading a single block
441 ** in a page. Where it does not produce a valid page for holes, or past the
442 ** end of the file. This turns out to be exactly what we need for reading
443 ** tails for conversion.
445 ** The point of the wrapper is forcing a certain value for create, even
446 ** though the VFS layer is calling this function with create==1. If you
447 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
448 ** don't use this function.
450 static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
451 struct buffer_head *bh_result,
454 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
457 /* This is special helper for reiserfs_get_block in case we are executing
458 direct_IO request. */
459 static int reiserfs_get_blocks_direct_io(struct inode *inode,
461 struct buffer_head *bh_result,
466 bh_result->b_page = NULL;
468 /* We set the b_size before reiserfs_get_block call since it is
469 referenced in convert_tail_for_hole() that may be called from
470 reiserfs_get_block() */
471 bh_result->b_size = (1 << inode->i_blkbits);
473 ret = reiserfs_get_block(inode, iblock, bh_result,
474 create | GET_BLOCK_NO_DANGLE);
478 /* don't allow direct io onto tail pages */
479 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
480 /* make sure future calls to the direct io funcs for this offset
481 ** in the file fail by unmapping the buffer
483 clear_buffer_mapped(bh_result);
486 /* Possible unpacked tail. Flush the data before pages have
488 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
491 reiserfs_write_lock(inode->i_sb);
493 err = reiserfs_commit_for_inode(inode);
494 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
496 reiserfs_write_unlock(inode->i_sb);
506 ** helper function for when reiserfs_get_block is called for a hole
507 ** but the file tail is still in a direct item
508 ** bh_result is the buffer head for the hole
509 ** tail_offset is the offset of the start of the tail in the file
511 ** This calls prepare_write, which will start a new transaction
512 ** you should not be in a transaction, or have any paths held when you
515 static int convert_tail_for_hole(struct inode *inode,
516 struct buffer_head *bh_result,
520 unsigned long tail_end;
521 unsigned long tail_start;
522 struct page *tail_page;
523 struct page *hole_page = bh_result->b_page;
526 if ((tail_offset & (bh_result->b_size - 1)) != 1)
529 /* always try to read until the end of the block */
530 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
531 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
533 index = tail_offset >> PAGE_CACHE_SHIFT;
534 /* hole_page can be zero in case of direct_io, we are sure
535 that we cannot get here if we write with O_DIRECT into
537 if (!hole_page || index != hole_page->index) {
538 tail_page = grab_cache_page(inode->i_mapping, index);
544 tail_page = hole_page;
547 /* we don't have to make sure the conversion did not happen while
548 ** we were locking the page because anyone that could convert
549 ** must first take i_mutex.
551 ** We must fix the tail page for writing because it might have buffers
552 ** that are mapped, but have a block number of 0. This indicates tail
553 ** data that has been read directly into the page, and block_prepare_write
554 ** won't trigger a get_block in this case.
556 fix_tail_page_for_writing(tail_page);
557 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
561 /* tail conversion might change the data in the page */
562 flush_dcache_page(tail_page);
564 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
567 if (tail_page != hole_page) {
568 unlock_page(tail_page);
569 page_cache_release(tail_page);
575 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
578 b_blocknr_t * allocated_block_nr,
579 struct treepath *path, int flags)
581 BUG_ON(!th->t_trans_id);
583 #ifdef REISERFS_PREALLOCATE
584 if (!(flags & GET_BLOCK_NO_IMUX)) {
585 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
589 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
593 int reiserfs_get_block(struct inode *inode, sector_t block,
594 struct buffer_head *bh_result, int create)
596 int repeat, retval = 0;
597 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
598 INITIALIZE_PATH(path);
601 struct buffer_head *bh, *unbh = NULL;
602 struct item_head *ih, tmp_ih;
607 struct reiserfs_transaction_handle *th = NULL;
608 /* space reserved in transaction batch:
609 . 3 balancings in direct->indirect conversion
610 . 1 block involved into reiserfs_update_sd()
611 XXX in practically impossible worst case direct2indirect()
612 can incur (much) more than 3 balancings.
613 quota update for user, group */
615 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
616 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
620 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
622 lock_depth = reiserfs_write_lock_once(inode->i_sb);
623 version = get_inode_item_key_version(inode);
625 if (!file_capable(inode, block)) {
626 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
630 /* if !create, we aren't changing the FS, so we don't need to
631 ** log anything, so we don't need to start a transaction
633 if (!(create & GET_BLOCK_CREATE)) {
635 /* find number of block-th logical block of the file */
636 ret = _get_block_create_0(inode, block, bh_result,
637 create | GET_BLOCK_READ_DIRECT);
638 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
642 * if we're already in a transaction, make sure to close
643 * any new transactions we start in this func
645 if ((create & GET_BLOCK_NO_DANGLE) ||
646 reiserfs_transaction_running(inode->i_sb))
649 /* If file is of such a size, that it might have a tail and tails are enabled
650 ** we should mark it as possibly needing tail packing on close
652 if ((have_large_tails(inode->i_sb)
653 && inode->i_size < i_block_size(inode) * 4)
654 || (have_small_tails(inode->i_sb)
655 && inode->i_size < i_block_size(inode)))
656 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
658 /* set the key of the first byte in the 'block'-th block of file */
659 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
660 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
662 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
667 reiserfs_update_inode_transaction(inode);
671 retval = search_for_position_by_key(inode->i_sb, &key, &path);
672 if (retval == IO_ERROR) {
677 bh = get_last_bh(&path);
679 item = get_item(&path);
680 pos_in_item = path.pos_in_item;
682 fs_gen = get_generation(inode->i_sb);
683 copy_item_head(&tmp_ih, ih);
685 if (allocation_needed
686 (retval, allocated_block_nr, ih, item, pos_in_item)) {
687 /* we have to allocate block for the unformatted node */
694 _allocate_block(th, block, inode, &allocated_block_nr,
697 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
698 /* restart the transaction to give the journal a chance to free
699 ** some blocks. releases the path, so we have to go back to
700 ** research if we succeed on the second try
702 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
703 retval = restart_transaction(th, inode, &path);
707 _allocate_block(th, block, inode,
708 &allocated_block_nr, NULL, create);
710 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
713 if (repeat == QUOTA_EXCEEDED)
720 if (fs_changed(fs_gen, inode->i_sb)
721 && item_moved(&tmp_ih, &path)) {
726 if (indirect_item_found(retval, ih)) {
727 b_blocknr_t unfm_ptr;
728 /* 'block'-th block is in the file already (there is
729 corresponding cell in some indirect item). But it may be
730 zero unformatted node pointer (hole) */
731 unfm_ptr = get_block_num(item, pos_in_item);
733 /* use allocated block to plug the hole */
734 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
735 if (fs_changed(fs_gen, inode->i_sb)
736 && item_moved(&tmp_ih, &path)) {
737 reiserfs_restore_prepared_buffer(inode->i_sb,
741 set_buffer_new(bh_result);
742 if (buffer_dirty(bh_result)
743 && reiserfs_data_ordered(inode->i_sb))
744 reiserfs_add_ordered_list(inode, bh_result);
745 put_block_num(item, pos_in_item, allocated_block_nr);
746 unfm_ptr = allocated_block_nr;
747 journal_mark_dirty(th, inode->i_sb, bh);
748 reiserfs_update_sd(th, inode);
750 set_block_dev_mapped(bh_result, unfm_ptr, inode);
754 retval = reiserfs_end_persistent_transaction(th);
756 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
758 /* the item was found, so new blocks were not added to the file
759 ** there is no need to make sure the inode is updated with this
770 /* desired position is not found or is in the direct item. We have
771 to append file with holes up to 'block'-th block converting
772 direct items to indirect one if necessary */
775 if (is_statdata_le_ih(ih)) {
777 struct cpu_key tmp_key;
779 /* indirect item has to be inserted */
780 make_le_item_head(&tmp_ih, &key, version, 1,
781 TYPE_INDIRECT, UNFM_P_SIZE,
782 0 /* free_space */ );
784 if (cpu_key_k_offset(&key) == 1) {
785 /* we are going to add 'block'-th block to the file. Use
786 allocated block for that */
787 unp = cpu_to_le32(allocated_block_nr);
788 set_block_dev_mapped(bh_result,
789 allocated_block_nr, inode);
790 set_buffer_new(bh_result);
794 set_cpu_key_k_offset(&tmp_key, 1);
795 PATH_LAST_POSITION(&path)++;
798 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
799 inode, (char *)&unp);
801 reiserfs_free_block(th, inode,
802 allocated_block_nr, 1);
803 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
805 //mark_tail_converted (inode);
806 } else if (is_direct_le_ih(ih)) {
807 /* direct item has to be converted */
811 ((le_ih_k_offset(ih) -
812 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
813 if (tail_offset == cpu_key_k_offset(&key)) {
814 /* direct item we just found fits into block we have
815 to map. Convert it into unformatted node: use
816 bh_result for the conversion */
817 set_block_dev_mapped(bh_result,
818 allocated_block_nr, inode);
822 /* we have to padd file tail stored in direct item(s)
823 up to block size and convert it to unformatted
824 node. FIXME: this should also get into page cache */
828 * ugly, but we can only end the transaction if
831 BUG_ON(!th->t_refcount);
832 if (th->t_refcount == 1) {
834 reiserfs_end_persistent_transaction
842 convert_tail_for_hole(inode, bh_result,
845 if (retval != -ENOSPC)
846 reiserfs_error(inode->i_sb,
848 "convert tail failed "
849 "inode %lu, error %d",
852 if (allocated_block_nr) {
853 /* the bitmap, the super, and the stat data == 3 */
855 th = reiserfs_persistent_transaction(inode->i_sb, 3);
857 reiserfs_free_block(th,
867 direct2indirect(th, inode, &path, unbh,
870 reiserfs_unmap_buffer(unbh);
871 reiserfs_free_block(th, inode,
872 allocated_block_nr, 1);
875 /* it is important the set_buffer_uptodate is done after
876 ** the direct2indirect. The buffer might contain valid
877 ** data newer than the data on disk (read by readpage, changed,
878 ** and then sent here by writepage). direct2indirect needs
879 ** to know if unbh was already up to date, so it can decide
880 ** if the data in unbh needs to be replaced with data from
883 set_buffer_uptodate(unbh);
885 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
886 buffer will disappear shortly, so it should not be added to
889 /* we've converted the tail, so we must
890 ** flush unbh before the transaction commits
892 reiserfs_add_tail_list(inode, unbh);
894 /* mark it dirty now to prevent commit_write from adding
895 ** this buffer to the inode's dirty buffer list
898 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
899 * It's still atomic, but it sets the page dirty too,
900 * which makes it eligible for writeback at any time by the
901 * VM (which was also the case with __mark_buffer_dirty())
903 mark_buffer_dirty(unbh);
906 /* append indirect item with holes if needed, when appending
907 pointer to 'block'-th block use block, which is already
909 struct cpu_key tmp_key;
910 unp_t unf_single = 0; // We use this in case we need to allocate only
911 // one block which is a fastpath
913 __u64 max_to_insert =
914 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
918 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
919 "vs-804: invalid position for append");
920 /* indirect item has to be appended, set up key of that position */
921 make_cpu_key(&tmp_key, inode,
922 le_key_k_offset(version,
925 inode->i_sb->s_blocksize),
926 //pos_in_item * inode->i_sb->s_blocksize,
927 TYPE_INDIRECT, 3); // key type is unimportant
929 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
930 "green-805: invalid offset");
933 ((cpu_key_k_offset(&key) -
934 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
937 if (blocks_needed == 1) {
940 un = kzalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_NOFS);
947 if (blocks_needed <= max_to_insert) {
948 /* we are going to add target block to the file. Use allocated
950 un[blocks_needed - 1] =
951 cpu_to_le32(allocated_block_nr);
952 set_block_dev_mapped(bh_result,
953 allocated_block_nr, inode);
954 set_buffer_new(bh_result);
957 /* paste hole to the indirect item */
958 /* If kmalloc failed, max_to_insert becomes zero and it means we
959 only have space for one block */
961 max_to_insert ? max_to_insert : 1;
964 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
969 if (blocks_needed != 1)
973 reiserfs_free_block(th, inode,
974 allocated_block_nr, 1);
978 /* We need to mark new file size in case this function will be
979 interrupted/aborted later on. And we may do this only for
982 inode->i_sb->s_blocksize * blocks_needed;
989 /* this loop could log more blocks than we had originally asked
990 ** for. So, we have to allow the transaction to end if it is
991 ** too big or too full. Update the inode so things are
992 ** consistent if we crash before the function returns
994 ** release the path so that anybody waiting on the path before
995 ** ending their transaction will be able to continue.
997 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
998 retval = restart_transaction(th, inode, &path);
1003 * inserting indirect pointers for a hole can take a
1004 * long time. reschedule if needed and also release the write
1007 if (need_resched()) {
1008 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
1010 lock_depth = reiserfs_write_lock_once(inode->i_sb);
1013 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1014 if (retval == IO_ERROR) {
1018 if (retval == POSITION_FOUND) {
1019 reiserfs_warning(inode->i_sb, "vs-825",
1020 "%K should not be found", &key);
1022 if (allocated_block_nr)
1023 reiserfs_free_block(th, inode,
1024 allocated_block_nr, 1);
1028 bh = get_last_bh(&path);
1030 item = get_item(&path);
1031 pos_in_item = path.pos_in_item;
1037 if (th && (!dangle || (retval && !th->t_trans_id))) {
1040 reiserfs_update_sd(th, inode);
1041 err = reiserfs_end_persistent_transaction(th);
1046 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
1047 reiserfs_check_path(&path);
1052 reiserfs_readpages(struct file *file, struct address_space *mapping,
1053 struct list_head *pages, unsigned nr_pages)
1055 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
1058 /* Compute real number of used bytes by file
1059 * Following three functions can go away when we'll have enough space in stat item
1061 static int real_space_diff(struct inode *inode, int sd_size)
1064 loff_t blocksize = inode->i_sb->s_blocksize;
1066 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1069 /* End of file is also in full block with indirect reference, so round
1070 ** up to the next block.
1072 ** there is just no way to know if the tail is actually packed
1073 ** on the file, so we have to assume it isn't. When we pack the
1074 ** tail, we add 4 bytes to pretend there really is an unformatted
1079 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1084 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
1087 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1088 return inode->i_size +
1089 (loff_t) (real_space_diff(inode, sd_size));
1091 return ((loff_t) real_space_diff(inode, sd_size)) +
1092 (((loff_t) blocks) << 9);
1095 /* Compute number of blocks used by file in ReiserFS counting */
1096 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1098 loff_t bytes = inode_get_bytes(inode);
1099 loff_t real_space = real_space_diff(inode, sd_size);
1101 /* keeps fsck and non-quota versions of reiserfs happy */
1102 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1103 bytes += (loff_t) 511;
1106 /* files from before the quota patch might i_blocks such that
1107 ** bytes < real_space. Deal with that here to prevent it from
1110 if (bytes < real_space)
1112 return (bytes - real_space) >> 9;
1116 // BAD: new directories have stat data of new type and all other items
1117 // of old type. Version stored in the inode says about body items, so
1118 // in update_stat_data we can not rely on inode, but have to check
1119 // item version directly
1122 // called by read_locked_inode
1123 static void init_inode(struct inode *inode, struct treepath *path)
1125 struct buffer_head *bh;
1126 struct item_head *ih;
1128 //int version = ITEM_VERSION_1;
1130 bh = PATH_PLAST_BUFFER(path);
1131 ih = PATH_PITEM_HEAD(path);
1133 copy_key(INODE_PKEY(inode), &(ih->ih_key));
1135 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1136 REISERFS_I(inode)->i_flags = 0;
1137 REISERFS_I(inode)->i_prealloc_block = 0;
1138 REISERFS_I(inode)->i_prealloc_count = 0;
1139 REISERFS_I(inode)->i_trans_id = 0;
1140 REISERFS_I(inode)->i_jl = NULL;
1141 reiserfs_init_xattr_rwsem(inode);
1143 if (stat_data_v1(ih)) {
1144 struct stat_data_v1 *sd =
1145 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1146 unsigned long blocks;
1148 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1149 set_inode_sd_version(inode, STAT_DATA_V1);
1150 inode->i_mode = sd_v1_mode(sd);
1151 inode->i_nlink = sd_v1_nlink(sd);
1152 inode->i_uid = sd_v1_uid(sd);
1153 inode->i_gid = sd_v1_gid(sd);
1154 inode->i_size = sd_v1_size(sd);
1155 inode->i_atime.tv_sec = sd_v1_atime(sd);
1156 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1157 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1158 inode->i_atime.tv_nsec = 0;
1159 inode->i_ctime.tv_nsec = 0;
1160 inode->i_mtime.tv_nsec = 0;
1162 inode->i_blocks = sd_v1_blocks(sd);
1163 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1164 blocks = (inode->i_size + 511) >> 9;
1165 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1166 if (inode->i_blocks > blocks) {
1167 // there was a bug in <=3.5.23 when i_blocks could take negative
1168 // values. Starting from 3.5.17 this value could even be stored in
1169 // stat data. For such files we set i_blocks based on file
1170 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1171 // only updated if file's inode will ever change
1172 inode->i_blocks = blocks;
1175 rdev = sd_v1_rdev(sd);
1176 REISERFS_I(inode)->i_first_direct_byte =
1177 sd_v1_first_direct_byte(sd);
1178 /* an early bug in the quota code can give us an odd number for the
1179 ** block count. This is incorrect, fix it here.
1181 if (inode->i_blocks & 1) {
1184 inode_set_bytes(inode,
1185 to_real_used_space(inode, inode->i_blocks,
1187 /* nopack is initially zero for v1 objects. For v2 objects,
1188 nopack is initialised from sd_attrs */
1189 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1191 // new stat data found, but object may have old items
1192 // (directories and symlinks)
1193 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1195 inode->i_mode = sd_v2_mode(sd);
1196 inode->i_nlink = sd_v2_nlink(sd);
1197 inode->i_uid = sd_v2_uid(sd);
1198 inode->i_size = sd_v2_size(sd);
1199 inode->i_gid = sd_v2_gid(sd);
1200 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1201 inode->i_atime.tv_sec = sd_v2_atime(sd);
1202 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1203 inode->i_ctime.tv_nsec = 0;
1204 inode->i_mtime.tv_nsec = 0;
1205 inode->i_atime.tv_nsec = 0;
1206 inode->i_blocks = sd_v2_blocks(sd);
1207 rdev = sd_v2_rdev(sd);
1208 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1209 inode->i_generation =
1210 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1212 inode->i_generation = sd_v2_generation(sd);
1214 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1215 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1217 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1218 REISERFS_I(inode)->i_first_direct_byte = 0;
1219 set_inode_sd_version(inode, STAT_DATA_V2);
1220 inode_set_bytes(inode,
1221 to_real_used_space(inode, inode->i_blocks,
1223 /* read persistent inode attributes from sd and initalise
1224 generic inode flags from them */
1225 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1226 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
1230 if (S_ISREG(inode->i_mode)) {
1231 inode->i_op = &reiserfs_file_inode_operations;
1232 inode->i_fop = &reiserfs_file_operations;
1233 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1234 } else if (S_ISDIR(inode->i_mode)) {
1235 inode->i_op = &reiserfs_dir_inode_operations;
1236 inode->i_fop = &reiserfs_dir_operations;
1237 } else if (S_ISLNK(inode->i_mode)) {
1238 inode->i_op = &reiserfs_symlink_inode_operations;
1239 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1241 inode->i_blocks = 0;
1242 inode->i_op = &reiserfs_special_inode_operations;
1243 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1247 // update new stat data with inode fields
1248 static void inode2sd(void *sd, struct inode *inode, loff_t size)
1250 struct stat_data *sd_v2 = (struct stat_data *)sd;
1253 set_sd_v2_mode(sd_v2, inode->i_mode);
1254 set_sd_v2_nlink(sd_v2, inode->i_nlink);
1255 set_sd_v2_uid(sd_v2, inode->i_uid);
1256 set_sd_v2_size(sd_v2, size);
1257 set_sd_v2_gid(sd_v2, inode->i_gid);
1258 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1259 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1260 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1261 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1262 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1263 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1265 set_sd_v2_generation(sd_v2, inode->i_generation);
1266 flags = REISERFS_I(inode)->i_attrs;
1267 i_attrs_to_sd_attrs(inode, &flags);
1268 set_sd_v2_attrs(sd_v2, flags);
1271 // used to copy inode's fields to old stat data
1272 static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
1274 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
1276 set_sd_v1_mode(sd_v1, inode->i_mode);
1277 set_sd_v1_uid(sd_v1, inode->i_uid);
1278 set_sd_v1_gid(sd_v1, inode->i_gid);
1279 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1280 set_sd_v1_size(sd_v1, size);
1281 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1282 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1283 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
1285 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1286 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1288 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1290 // Sigh. i_first_direct_byte is back
1291 set_sd_v1_first_direct_byte(sd_v1,
1292 REISERFS_I(inode)->i_first_direct_byte);
1295 /* NOTE, you must prepare the buffer head before sending it here,
1296 ** and then log it after the call
1298 static void update_stat_data(struct treepath *path, struct inode *inode,
1301 struct buffer_head *bh;
1302 struct item_head *ih;
1304 bh = PATH_PLAST_BUFFER(path);
1305 ih = PATH_PITEM_HEAD(path);
1307 if (!is_statdata_le_ih(ih))
1308 reiserfs_panic(inode->i_sb, "vs-13065", "key %k, found item %h",
1309 INODE_PKEY(inode), ih);
1311 if (stat_data_v1(ih)) {
1312 // path points to old stat data
1313 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1315 inode2sd(B_I_PITEM(bh, ih), inode, size);
1321 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1322 struct inode *inode, loff_t size)
1325 INITIALIZE_PATH(path);
1326 struct buffer_head *bh;
1328 struct item_head *ih, tmp_ih;
1331 BUG_ON(!th->t_trans_id);
1333 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant
1337 /* look for the object's stat data */
1338 retval = search_item(inode->i_sb, &key, &path);
1339 if (retval == IO_ERROR) {
1340 reiserfs_error(inode->i_sb, "vs-13050",
1341 "i/o failure occurred trying to "
1342 "update %K stat data", &key);
1345 if (retval == ITEM_NOT_FOUND) {
1346 pos = PATH_LAST_POSITION(&path);
1348 if (inode->i_nlink == 0) {
1349 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1352 reiserfs_warning(inode->i_sb, "vs-13060",
1353 "stat data of object %k (nlink == %d) "
1354 "not found (pos %d)",
1355 INODE_PKEY(inode), inode->i_nlink,
1357 reiserfs_check_path(&path);
1361 /* sigh, prepare_for_journal might schedule. When it schedules the
1362 ** FS might change. We have to detect that, and loop back to the
1363 ** search if the stat data item has moved
1365 bh = get_last_bh(&path);
1367 copy_item_head(&tmp_ih, ih);
1368 fs_gen = get_generation(inode->i_sb);
1369 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1370 if (fs_changed(fs_gen, inode->i_sb)
1371 && item_moved(&tmp_ih, &path)) {
1372 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1373 continue; /* Stat_data item has been moved after scheduling. */
1377 update_stat_data(&path, inode, size);
1378 journal_mark_dirty(th, th->t_super, bh);
1383 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1384 ** does a make_bad_inode when things go wrong. But, we need to make sure
1385 ** and clear the key in the private portion of the inode, otherwise a
1386 ** corresponding iput might try to delete whatever object the inode last
1389 static void reiserfs_make_bad_inode(struct inode *inode)
1391 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1392 make_bad_inode(inode);
1396 // initially this function was derived from minix or ext2's analog and
1397 // evolved as the prototype did
1400 int reiserfs_init_locked_inode(struct inode *inode, void *p)
1402 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1403 inode->i_ino = args->objectid;
1404 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1408 /* looks for stat data in the tree, and fills up the fields of in-core
1409 inode stat data fields */
1410 void reiserfs_read_locked_inode(struct inode *inode,
1411 struct reiserfs_iget_args *args)
1413 INITIALIZE_PATH(path_to_sd);
1415 unsigned long dirino;
1418 dirino = args->dirid;
1420 /* set version 1, version 2 could be used too, because stat data
1421 key is the same in both versions */
1422 key.version = KEY_FORMAT_3_5;
1423 key.on_disk_key.k_dir_id = dirino;
1424 key.on_disk_key.k_objectid = inode->i_ino;
1425 key.on_disk_key.k_offset = 0;
1426 key.on_disk_key.k_type = 0;
1428 /* look for the object's stat data */
1429 retval = search_item(inode->i_sb, &key, &path_to_sd);
1430 if (retval == IO_ERROR) {
1431 reiserfs_error(inode->i_sb, "vs-13070",
1432 "i/o failure occurred trying to find "
1433 "stat data of %K", &key);
1434 reiserfs_make_bad_inode(inode);
1437 if (retval != ITEM_FOUND) {
1438 /* a stale NFS handle can trigger this without it being an error */
1439 pathrelse(&path_to_sd);
1440 reiserfs_make_bad_inode(inode);
1445 init_inode(inode, &path_to_sd);
1447 /* It is possible that knfsd is trying to access inode of a file
1448 that is being removed from the disk by some other thread. As we
1449 update sd on unlink all that is required is to check for nlink
1450 here. This bug was first found by Sizif when debugging
1451 SquidNG/Butterfly, forgotten, and found again after Philippe
1452 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1454 More logical fix would require changes in fs/inode.c:iput() to
1455 remove inode from hash-table _after_ fs cleaned disk stuff up and
1456 in iget() to return NULL if I_FREEING inode is found in
1458 /* Currently there is one place where it's ok to meet inode with
1459 nlink==0: processing of open-unlinked and half-truncated files
1460 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1461 if ((inode->i_nlink == 0) &&
1462 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
1463 reiserfs_warning(inode->i_sb, "vs-13075",
1464 "dead inode read from disk %K. "
1465 "This is likely to be race with knfsd. Ignore",
1467 reiserfs_make_bad_inode(inode);
1470 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
1475 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1477 * @inode: inode from hash table to check
1478 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1480 * This function is called by iget5_locked() to distinguish reiserfs inodes
1481 * having the same inode numbers. Such inodes can only exist due to some
1482 * error condition. One of them should be bad. Inodes with identical
1483 * inode numbers (objectids) are distinguished by parent directory ids.
1486 int reiserfs_find_actor(struct inode *inode, void *opaque)
1488 struct reiserfs_iget_args *args;
1491 /* args is already in CPU order */
1492 return (inode->i_ino == args->objectid) &&
1493 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1496 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
1498 struct inode *inode;
1499 struct reiserfs_iget_args args;
1501 args.objectid = key->on_disk_key.k_objectid;
1502 args.dirid = key->on_disk_key.k_dir_id;
1503 reiserfs_write_unlock(s);
1504 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1505 reiserfs_find_actor, reiserfs_init_locked_inode,
1507 reiserfs_write_lock(s);
1509 return ERR_PTR(-ENOMEM);
1511 if (inode->i_state & I_NEW) {
1512 reiserfs_read_locked_inode(inode, &args);
1513 unlock_new_inode(inode);
1516 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1517 /* either due to i/o error or a stale NFS handle */
1524 static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1525 u32 objectid, u32 dir_id, u32 generation)
1529 struct inode *inode;
1531 key.on_disk_key.k_objectid = objectid;
1532 key.on_disk_key.k_dir_id = dir_id;
1533 reiserfs_write_lock(sb);
1534 inode = reiserfs_iget(sb, &key);
1535 if (inode && !IS_ERR(inode) && generation != 0 &&
1536 generation != inode->i_generation) {
1540 reiserfs_write_unlock(sb);
1542 return d_obtain_alias(inode);
1545 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1546 int fh_len, int fh_type)
1548 /* fhtype happens to reflect the number of u32s encoded.
1549 * due to a bug in earlier code, fhtype might indicate there
1550 * are more u32s then actually fitted.
1551 * so if fhtype seems to be more than len, reduce fhtype.
1553 * 2 - objectid + dir_id - legacy support
1554 * 3 - objectid + dir_id + generation
1555 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1556 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1557 * 6 - as above plus generation of directory
1558 * 6 does not fit in NFSv2 handles
1560 if (fh_type > fh_len) {
1561 if (fh_type != 6 || fh_len != 5)
1562 reiserfs_warning(sb, "reiserfs-13077",
1563 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1568 return reiserfs_get_dentry(sb, fid->raw[0], fid->raw[1],
1569 (fh_type == 3 || fh_type >= 5) ? fid->raw[2] : 0);
1572 struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1573 int fh_len, int fh_type)
1578 return reiserfs_get_dentry(sb,
1579 (fh_type >= 5) ? fid->raw[3] : fid->raw[2],
1580 (fh_type >= 5) ? fid->raw[4] : fid->raw[3],
1581 (fh_type == 6) ? fid->raw[5] : 0);
1584 int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1587 struct inode *inode = dentry->d_inode;
1593 data[0] = inode->i_ino;
1594 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1595 data[2] = inode->i_generation;
1597 /* no room for directory info? return what we've stored so far */
1598 if (maxlen < 5 || !need_parent)
1601 spin_lock(&dentry->d_lock);
1602 inode = dentry->d_parent->d_inode;
1603 data[3] = inode->i_ino;
1604 data[4] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1607 data[5] = inode->i_generation;
1610 spin_unlock(&dentry->d_lock);
1614 /* looks for stat data, then copies fields to it, marks the buffer
1615 containing stat data as dirty */
1616 /* reiserfs inodes are never really dirty, since the dirty inode call
1617 ** always logs them. This call allows the VFS inode marking routines
1618 ** to properly mark inodes for datasync and such, but only actually
1619 ** does something when called for a synchronous update.
1621 int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc)
1623 struct reiserfs_transaction_handle th;
1624 int jbegin_count = 1;
1626 if (inode->i_sb->s_flags & MS_RDONLY)
1628 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1629 ** these cases are just when the system needs ram, not when the
1630 ** inode needs to reach disk for safety, and they can safely be
1631 ** ignored because the altered inode has already been logged.
1633 if (wbc->sync_mode == WB_SYNC_ALL && !(current->flags & PF_MEMALLOC)) {
1634 reiserfs_write_lock(inode->i_sb);
1635 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1636 reiserfs_update_sd(&th, inode);
1637 journal_end_sync(&th, inode->i_sb, jbegin_count);
1639 reiserfs_write_unlock(inode->i_sb);
1644 /* stat data of new object is inserted already, this inserts the item
1645 containing "." and ".." entries */
1646 static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1647 struct inode *inode,
1648 struct item_head *ih, struct treepath *path,
1651 struct super_block *sb = th->t_super;
1652 char empty_dir[EMPTY_DIR_SIZE];
1653 char *body = empty_dir;
1657 BUG_ON(!th->t_trans_id);
1659 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1660 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1661 TYPE_DIRENTRY, 3 /*key length */ );
1663 /* compose item head for new item. Directories consist of items of
1664 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1665 is done by reiserfs_new_inode */
1666 if (old_format_only(sb)) {
1667 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1668 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1670 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1671 ih->ih_key.k_objectid,
1672 INODE_PKEY(dir)->k_dir_id,
1673 INODE_PKEY(dir)->k_objectid);
1675 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1676 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1678 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1679 ih->ih_key.k_objectid,
1680 INODE_PKEY(dir)->k_dir_id,
1681 INODE_PKEY(dir)->k_objectid);
1684 /* look for place in the tree for new item */
1685 retval = search_item(sb, &key, path);
1686 if (retval == IO_ERROR) {
1687 reiserfs_error(sb, "vs-13080",
1688 "i/o failure occurred creating new directory");
1691 if (retval == ITEM_FOUND) {
1693 reiserfs_warning(sb, "vs-13070",
1694 "object with this key exists (%k)",
1699 /* insert item, that is empty directory item */
1700 return reiserfs_insert_item(th, path, &key, ih, inode, body);
1703 /* stat data of object has been inserted, this inserts the item
1704 containing the body of symlink */
1705 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1706 struct item_head *ih,
1707 struct treepath *path, const char *symname,
1710 struct super_block *sb = th->t_super;
1714 BUG_ON(!th->t_trans_id);
1716 _make_cpu_key(&key, KEY_FORMAT_3_5,
1717 le32_to_cpu(ih->ih_key.k_dir_id),
1718 le32_to_cpu(ih->ih_key.k_objectid),
1719 1, TYPE_DIRECT, 3 /*key length */ );
1721 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1722 0 /*free_space */ );
1724 /* look for place in the tree for new item */
1725 retval = search_item(sb, &key, path);
1726 if (retval == IO_ERROR) {
1727 reiserfs_error(sb, "vs-13080",
1728 "i/o failure occurred creating new symlink");
1731 if (retval == ITEM_FOUND) {
1733 reiserfs_warning(sb, "vs-13080",
1734 "object with this key exists (%k)",
1739 /* insert item, that is body of symlink */
1740 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
1743 /* inserts the stat data into the tree, and then calls
1744 reiserfs_new_directory (to insert ".", ".." item if new object is
1745 directory) or reiserfs_new_symlink (to insert symlink body if new
1746 object is symlink) or nothing (if new object is regular file)
1748 NOTE! uid and gid must already be set in the inode. If we return
1749 non-zero due to an error, we have to drop the quota previously allocated
1750 for the fresh inode. This can only be done outside a transaction, so
1751 if we return non-zero, we also end the transaction. */
1752 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1753 struct inode *dir, int mode, const char *symname,
1754 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1755 strlen (symname) for symlinks) */
1756 loff_t i_size, struct dentry *dentry,
1757 struct inode *inode,
1758 struct reiserfs_security_handle *security)
1760 struct super_block *sb;
1761 struct reiserfs_iget_args args;
1762 INITIALIZE_PATH(path_to_key);
1764 struct item_head ih;
1765 struct stat_data sd;
1769 BUG_ON(!th->t_trans_id);
1771 dquot_initialize(inode);
1772 err = dquot_alloc_inode(inode);
1775 if (!dir->i_nlink) {
1782 /* item head of new item */
1783 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1784 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1785 if (!ih.ih_key.k_objectid) {
1789 args.objectid = inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
1790 if (old_format_only(sb))
1791 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1792 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1794 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1795 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1796 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1797 args.dirid = le32_to_cpu(ih.ih_key.k_dir_id);
1798 if (insert_inode_locked4(inode, args.objectid,
1799 reiserfs_find_actor, &args) < 0) {
1803 if (old_format_only(sb))
1804 /* not a perfect generation count, as object ids can be reused, but
1805 ** this is as good as reiserfs can do right now.
1806 ** note that the private part of inode isn't filled in yet, we have
1807 ** to use the directory.
1809 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1811 #if defined( USE_INODE_GENERATION_COUNTER )
1812 inode->i_generation =
1813 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1815 inode->i_generation = ++event;
1818 /* fill stat data */
1819 inode->i_nlink = (S_ISDIR(mode) ? 2 : 1);
1821 /* uid and gid must already be set by the caller for quota init */
1823 /* symlink cannot be immutable or append only, right? */
1824 if (S_ISLNK(inode->i_mode))
1825 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1827 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1828 inode->i_size = i_size;
1829 inode->i_blocks = 0;
1831 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1832 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1834 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1835 REISERFS_I(inode)->i_flags = 0;
1836 REISERFS_I(inode)->i_prealloc_block = 0;
1837 REISERFS_I(inode)->i_prealloc_count = 0;
1838 REISERFS_I(inode)->i_trans_id = 0;
1839 REISERFS_I(inode)->i_jl = NULL;
1840 REISERFS_I(inode)->i_attrs =
1841 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1842 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
1843 reiserfs_init_xattr_rwsem(inode);
1845 /* key to search for correct place for new stat data */
1846 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1847 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1848 TYPE_STAT_DATA, 3 /*key length */ );
1850 /* find proper place for inserting of stat data */
1851 retval = search_item(sb, &key, &path_to_key);
1852 if (retval == IO_ERROR) {
1856 if (retval == ITEM_FOUND) {
1857 pathrelse(&path_to_key);
1861 if (old_format_only(sb)) {
1862 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1863 pathrelse(&path_to_key);
1864 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1868 inode2sd_v1(&sd, inode, inode->i_size);
1870 inode2sd(&sd, inode, inode->i_size);
1872 // store in in-core inode the key of stat data and version all
1873 // object items will have (directory items will have old offset
1874 // format, other new objects will consist of new items)
1875 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1876 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1878 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1879 if (old_format_only(sb))
1880 set_inode_sd_version(inode, STAT_DATA_V1);
1882 set_inode_sd_version(inode, STAT_DATA_V2);
1884 /* insert the stat data into the tree */
1885 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1886 if (REISERFS_I(dir)->new_packing_locality)
1887 th->displace_new_blocks = 1;
1890 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1894 reiserfs_check_path(&path_to_key);
1897 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1898 if (!th->displace_new_blocks)
1899 REISERFS_I(dir)->new_packing_locality = 0;
1901 if (S_ISDIR(mode)) {
1902 /* insert item with "." and ".." */
1904 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1907 if (S_ISLNK(mode)) {
1908 /* insert body of symlink */
1909 if (!old_format_only(sb))
1910 i_size = ROUND_UP(i_size);
1912 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1917 reiserfs_check_path(&path_to_key);
1918 journal_end(th, th->t_super, th->t_blocks_allocated);
1919 goto out_inserted_sd;
1922 if (reiserfs_posixacl(inode->i_sb)) {
1923 retval = reiserfs_inherit_default_acl(th, dir, dentry, inode);
1926 reiserfs_check_path(&path_to_key);
1927 journal_end(th, th->t_super, th->t_blocks_allocated);
1928 goto out_inserted_sd;
1930 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1931 reiserfs_warning(inode->i_sb, "jdm-13090",
1932 "ACLs aren't enabled in the fs, "
1933 "but vfs thinks they are!");
1934 } else if (IS_PRIVATE(dir))
1935 inode->i_flags |= S_PRIVATE;
1937 if (security->name) {
1938 retval = reiserfs_security_write(th, inode, security);
1941 reiserfs_check_path(&path_to_key);
1942 retval = journal_end(th, th->t_super,
1943 th->t_blocks_allocated);
1946 goto out_inserted_sd;
1950 reiserfs_update_sd(th, inode);
1951 reiserfs_check_path(&path_to_key);
1955 /* it looks like you can easily compress these two goto targets into
1956 * one. Keeping it like this doesn't actually hurt anything, and they
1957 * are place holders for what the quota code actually needs.
1960 /* Invalidate the object, nothing was inserted yet */
1961 INODE_PKEY(inode)->k_objectid = 0;
1963 /* Quota change must be inside a transaction for journaling */
1964 dquot_free_inode(inode);
1967 journal_end(th, th->t_super, th->t_blocks_allocated);
1968 /* Drop can be outside and it needs more credits so it's better to have it outside */
1970 inode->i_flags |= S_NOQUOTA;
1971 make_bad_inode(inode);
1975 th->t_trans_id = 0; /* so the caller can't use this handle later */
1976 unlock_new_inode(inode); /* OK to do even if we hadn't locked it */
1982 ** finds the tail page in the page cache,
1983 ** reads the last block in.
1985 ** On success, page_result is set to a locked, pinned page, and bh_result
1986 ** is set to an up to date buffer for the last block in the file. returns 0.
1988 ** tail conversion is not done, so bh_result might not be valid for writing
1989 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1990 ** trying to write the block.
1992 ** on failure, nonzero is returned, page_result and bh_result are untouched.
1994 static int grab_tail_page(struct inode *inode,
1995 struct page **page_result,
1996 struct buffer_head **bh_result)
1999 /* we want the page with the last byte in the file,
2000 ** not the page that will hold the next byte for appending
2002 unsigned long index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
2003 unsigned long pos = 0;
2004 unsigned long start = 0;
2005 unsigned long blocksize = inode->i_sb->s_blocksize;
2006 unsigned long offset = (inode->i_size) & (PAGE_CACHE_SIZE - 1);
2007 struct buffer_head *bh;
2008 struct buffer_head *head;
2012 /* we know that we are only called with inode->i_size > 0.
2013 ** we also know that a file tail can never be as big as a block
2014 ** If i_size % blocksize == 0, our file is currently block aligned
2015 ** and it won't need converting or zeroing after a truncate.
2017 if ((offset & (blocksize - 1)) == 0) {
2020 page = grab_cache_page(inode->i_mapping, index);
2025 /* start within the page of the last block in the file */
2026 start = (offset / blocksize) * blocksize;
2028 error = block_prepare_write(page, start, offset,
2029 reiserfs_get_block_create_0);
2033 head = page_buffers(page);
2039 bh = bh->b_this_page;
2041 } while (bh != head);
2043 if (!buffer_uptodate(bh)) {
2044 /* note, this should never happen, prepare_write should
2045 ** be taking care of this for us. If the buffer isn't up to date,
2046 ** I've screwed up the code to find the buffer, or the code to
2047 ** call prepare_write
2049 reiserfs_error(inode->i_sb, "clm-6000",
2050 "error reading block %lu", bh->b_blocknr);
2055 *page_result = page;
2062 page_cache_release(page);
2067 ** vfs version of truncate file. Must NOT be called with
2068 ** a transaction already started.
2070 ** some code taken from block_truncate_page
2072 int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
2074 struct reiserfs_transaction_handle th;
2075 /* we want the offset for the first byte after the end of the file */
2076 unsigned long offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2077 unsigned blocksize = inode->i_sb->s_blocksize;
2079 struct page *page = NULL;
2081 struct buffer_head *bh = NULL;
2085 lock_depth = reiserfs_write_lock_once(inode->i_sb);
2087 if (inode->i_size > 0) {
2088 error = grab_tail_page(inode, &page, &bh);
2090 // -ENOENT means we truncated past the end of the file,
2091 // and get_block_create_0 could not find a block to read in,
2093 if (error != -ENOENT)
2094 reiserfs_error(inode->i_sb, "clm-6001",
2095 "grab_tail_page failed %d",
2102 /* so, if page != NULL, we have a buffer head for the offset at
2103 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2104 ** then we have an unformatted node. Otherwise, we have a direct item,
2105 ** and no zeroing is required on disk. We zero after the truncate,
2106 ** because the truncate might pack the item anyway
2107 ** (it will unmap bh if it packs).
2109 /* it is enough to reserve space in transaction for 2 balancings:
2110 one for "save" link adding and another for the first
2111 cut_from_item. 1 is for update_sd */
2112 error = journal_begin(&th, inode->i_sb,
2113 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2116 reiserfs_update_inode_transaction(inode);
2117 if (update_timestamps)
2118 /* we are doing real truncate: if the system crashes before the last
2119 transaction of truncating gets committed - on reboot the file
2120 either appears truncated properly or not truncated at all */
2121 add_save_link(&th, inode, 1);
2122 err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps);
2124 journal_end(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2128 /* check reiserfs_do_truncate after ending the transaction */
2134 if (update_timestamps) {
2135 error = remove_save_link(inode, 1 /* truncate */);
2141 length = offset & (blocksize - 1);
2142 /* if we are not on a block boundary */
2144 length = blocksize - length;
2145 zero_user(page, offset, length);
2146 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2147 mark_buffer_dirty(bh);
2151 page_cache_release(page);
2154 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2160 page_cache_release(page);
2163 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2168 static int map_block_for_writepage(struct inode *inode,
2169 struct buffer_head *bh_result,
2170 unsigned long block)
2172 struct reiserfs_transaction_handle th;
2174 struct item_head tmp_ih;
2175 struct item_head *ih;
2176 struct buffer_head *bh;
2179 INITIALIZE_PATH(path);
2181 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
2182 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
2184 int use_get_block = 0;
2185 int bytes_copied = 0;
2187 int trans_running = 0;
2189 /* catch places below that try to log something without starting a trans */
2192 if (!buffer_uptodate(bh_result)) {
2196 kmap(bh_result->b_page);
2198 reiserfs_write_lock(inode->i_sb);
2199 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
2202 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2203 if (retval != POSITION_FOUND) {
2208 bh = get_last_bh(&path);
2210 item = get_item(&path);
2211 pos_in_item = path.pos_in_item;
2213 /* we've found an unformatted node */
2214 if (indirect_item_found(retval, ih)) {
2215 if (bytes_copied > 0) {
2216 reiserfs_warning(inode->i_sb, "clm-6002",
2217 "bytes_copied %d", bytes_copied);
2219 if (!get_block_num(item, pos_in_item)) {
2220 /* crap, we are writing to a hole */
2224 set_block_dev_mapped(bh_result,
2225 get_block_num(item, pos_in_item), inode);
2226 } else if (is_direct_le_ih(ih)) {
2228 p = page_address(bh_result->b_page);
2229 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2230 copy_size = ih_item_len(ih) - pos_in_item;
2232 fs_gen = get_generation(inode->i_sb);
2233 copy_item_head(&tmp_ih, ih);
2235 if (!trans_running) {
2236 /* vs-3050 is gone, no need to drop the path */
2237 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2240 reiserfs_update_inode_transaction(inode);
2242 if (fs_changed(fs_gen, inode->i_sb)
2243 && item_moved(&tmp_ih, &path)) {
2244 reiserfs_restore_prepared_buffer(inode->i_sb,
2250 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2252 if (fs_changed(fs_gen, inode->i_sb)
2253 && item_moved(&tmp_ih, &path)) {
2254 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2258 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2261 journal_mark_dirty(&th, inode->i_sb, bh);
2262 bytes_copied += copy_size;
2263 set_block_dev_mapped(bh_result, 0, inode);
2265 /* are there still bytes left? */
2266 if (bytes_copied < bh_result->b_size &&
2267 (byte_offset + bytes_copied) < inode->i_size) {
2268 set_cpu_key_k_offset(&key,
2269 cpu_key_k_offset(&key) +
2274 reiserfs_warning(inode->i_sb, "clm-6003",
2275 "bad item inode %lu", inode->i_ino);
2283 if (trans_running) {
2284 int err = journal_end(&th, inode->i_sb, jbegin_count);
2289 reiserfs_write_unlock(inode->i_sb);
2291 /* this is where we fill in holes in the file. */
2292 if (use_get_block) {
2293 retval = reiserfs_get_block(inode, block, bh_result,
2294 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2295 | GET_BLOCK_NO_DANGLE);
2297 if (!buffer_mapped(bh_result)
2298 || bh_result->b_blocknr == 0) {
2299 /* get_block failed to find a mapped unformatted node. */
2305 kunmap(bh_result->b_page);
2307 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2308 /* we've copied data from the page into the direct item, so the
2309 * buffer in the page is now clean, mark it to reflect that.
2311 lock_buffer(bh_result);
2312 clear_buffer_dirty(bh_result);
2313 unlock_buffer(bh_result);
2319 * mason@suse.com: updated in 2.5.54 to follow the same general io
2320 * start/recovery path as __block_write_full_page, along with special
2321 * code to handle reiserfs tails.
2323 static int reiserfs_write_full_page(struct page *page,
2324 struct writeback_control *wbc)
2326 struct inode *inode = page->mapping->host;
2327 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2329 unsigned long block;
2330 sector_t last_block;
2331 struct buffer_head *head, *bh;
2334 int checked = PageChecked(page);
2335 struct reiserfs_transaction_handle th;
2336 struct super_block *s = inode->i_sb;
2337 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2340 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2341 if (checked && (current->flags & PF_MEMALLOC)) {
2342 redirty_page_for_writepage(wbc, page);
2347 /* The page dirty bit is cleared before writepage is called, which
2348 * means we have to tell create_empty_buffers to make dirty buffers
2349 * The page really should be up to date at this point, so tossing
2350 * in the BH_Uptodate is just a sanity check.
2352 if (!page_has_buffers(page)) {
2353 create_empty_buffers(page, s->s_blocksize,
2354 (1 << BH_Dirty) | (1 << BH_Uptodate));
2356 head = page_buffers(page);
2358 /* last page in the file, zero out any contents past the
2359 ** last byte in the file
2361 if (page->index >= end_index) {
2362 unsigned last_offset;
2364 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2365 /* no file contents in this page */
2366 if (page->index >= end_index + 1 || !last_offset) {
2370 zero_user_segment(page, last_offset, PAGE_CACHE_SIZE);
2373 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
2374 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
2375 /* first map all the buffers, logging any direct items we find */
2377 if (block > last_block) {
2379 * This can happen when the block size is less than
2380 * the page size. The corresponding bytes in the page
2381 * were zero filled above
2383 clear_buffer_dirty(bh);
2384 set_buffer_uptodate(bh);
2385 } else if ((checked || buffer_dirty(bh)) &&
2386 (!buffer_mapped(bh) || (buffer_mapped(bh)
2389 /* not mapped yet, or it points to a direct item, search
2390 * the btree for the mapping info, and log any direct
2393 if ((error = map_block_for_writepage(inode, bh, block))) {
2397 bh = bh->b_this_page;
2399 } while (bh != head);
2402 * we start the transaction after map_block_for_writepage,
2403 * because it can create holes in the file (an unbounded operation).
2404 * starting it here, we can make a reliable estimate for how many
2405 * blocks we're going to log
2408 ClearPageChecked(page);
2409 reiserfs_write_lock(s);
2410 error = journal_begin(&th, s, bh_per_page + 1);
2412 reiserfs_write_unlock(s);
2415 reiserfs_update_inode_transaction(inode);
2417 /* now go through and lock any dirty buffers on the page */
2420 if (!buffer_mapped(bh))
2422 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2426 reiserfs_prepare_for_journal(s, bh, 1);
2427 journal_mark_dirty(&th, s, bh);
2430 /* from this point on, we know the buffer is mapped to a
2431 * real block and not a direct item
2433 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2436 if (!trylock_buffer(bh)) {
2437 redirty_page_for_writepage(wbc, page);
2441 if (test_clear_buffer_dirty(bh)) {
2442 mark_buffer_async_write(bh);
2446 } while ((bh = bh->b_this_page) != head);
2449 error = journal_end(&th, s, bh_per_page + 1);
2450 reiserfs_write_unlock(s);
2454 BUG_ON(PageWriteback(page));
2455 set_page_writeback(page);
2459 * since any buffer might be the only dirty buffer on the page,
2460 * the first submit_bh can bring the page out of writeback.
2461 * be careful with the buffers.
2464 struct buffer_head *next = bh->b_this_page;
2465 if (buffer_async_write(bh)) {
2466 submit_bh(WRITE, bh);
2471 } while (bh != head);
2477 * if this page only had a direct item, it is very possible for
2478 * no io to be required without there being an error. Or,
2479 * someone else could have locked them and sent them down the
2480 * pipe without locking the page
2484 if (!buffer_uptodate(bh)) {
2488 bh = bh->b_this_page;
2489 } while (bh != head);
2491 SetPageUptodate(page);
2492 end_page_writeback(page);
2497 /* catches various errors, we need to make sure any valid dirty blocks
2498 * get to the media. The page is currently locked and not marked for
2501 ClearPageUptodate(page);
2505 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2507 mark_buffer_async_write(bh);
2510 * clear any dirty bits that might have come from getting
2511 * attached to a dirty page
2513 clear_buffer_dirty(bh);
2515 bh = bh->b_this_page;
2516 } while (bh != head);
2518 BUG_ON(PageWriteback(page));
2519 set_page_writeback(page);
2522 struct buffer_head *next = bh->b_this_page;
2523 if (buffer_async_write(bh)) {
2524 clear_buffer_dirty(bh);
2525 submit_bh(WRITE, bh);
2530 } while (bh != head);
2534 static int reiserfs_readpage(struct file *f, struct page *page)
2536 return block_read_full_page(page, reiserfs_get_block);
2539 static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
2541 struct inode *inode = page->mapping->host;
2542 reiserfs_wait_on_write_block(inode->i_sb);
2543 return reiserfs_write_full_page(page, wbc);
2546 static void reiserfs_truncate_failed_write(struct inode *inode)
2548 truncate_inode_pages(inode->i_mapping, inode->i_size);
2549 reiserfs_truncate_file(inode, 0);
2552 static int reiserfs_write_begin(struct file *file,
2553 struct address_space *mapping,
2554 loff_t pos, unsigned len, unsigned flags,
2555 struct page **pagep, void **fsdata)
2557 struct inode *inode;
2563 inode = mapping->host;
2565 if (flags & AOP_FLAG_CONT_EXPAND &&
2566 (pos & (inode->i_sb->s_blocksize - 1)) == 0) {
2568 *fsdata = (void *)(unsigned long)flags;
2571 index = pos >> PAGE_CACHE_SHIFT;
2572 page = grab_cache_page_write_begin(mapping, index, flags);
2577 reiserfs_wait_on_write_block(inode->i_sb);
2578 fix_tail_page_for_writing(page);
2579 if (reiserfs_transaction_running(inode->i_sb)) {
2580 struct reiserfs_transaction_handle *th;
2581 th = (struct reiserfs_transaction_handle *)current->
2583 BUG_ON(!th->t_refcount);
2584 BUG_ON(!th->t_trans_id);
2585 old_ref = th->t_refcount;
2588 ret = __block_write_begin(page, pos, len, reiserfs_get_block);
2589 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2590 struct reiserfs_transaction_handle *th = current->journal_info;
2591 /* this gets a little ugly. If reiserfs_get_block returned an
2592 * error and left a transacstion running, we've got to close it,
2593 * and we've got to free handle if it was a persistent transaction.
2595 * But, if we had nested into an existing transaction, we need
2596 * to just drop the ref count on the handle.
2598 * If old_ref == 0, the transaction is from reiserfs_get_block,
2599 * and it was a persistent trans. Otherwise, it was nested above.
2601 if (th->t_refcount > old_ref) {
2606 reiserfs_write_lock(inode->i_sb);
2607 err = reiserfs_end_persistent_transaction(th);
2608 reiserfs_write_unlock(inode->i_sb);
2616 page_cache_release(page);
2617 /* Truncate allocated blocks */
2618 reiserfs_truncate_failed_write(inode);
2623 int reiserfs_prepare_write(struct file *f, struct page *page,
2624 unsigned from, unsigned to)
2626 struct inode *inode = page->mapping->host;
2630 reiserfs_write_unlock(inode->i_sb);
2631 reiserfs_wait_on_write_block(inode->i_sb);
2632 reiserfs_write_lock(inode->i_sb);
2634 fix_tail_page_for_writing(page);
2635 if (reiserfs_transaction_running(inode->i_sb)) {
2636 struct reiserfs_transaction_handle *th;
2637 th = (struct reiserfs_transaction_handle *)current->
2639 BUG_ON(!th->t_refcount);
2640 BUG_ON(!th->t_trans_id);
2641 old_ref = th->t_refcount;
2645 ret = block_prepare_write(page, from, to, reiserfs_get_block);
2646 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2647 struct reiserfs_transaction_handle *th = current->journal_info;
2648 /* this gets a little ugly. If reiserfs_get_block returned an
2649 * error and left a transacstion running, we've got to close it,
2650 * and we've got to free handle if it was a persistent transaction.
2652 * But, if we had nested into an existing transaction, we need
2653 * to just drop the ref count on the handle.
2655 * If old_ref == 0, the transaction is from reiserfs_get_block,
2656 * and it was a persistent trans. Otherwise, it was nested above.
2658 if (th->t_refcount > old_ref) {
2663 reiserfs_write_lock(inode->i_sb);
2664 err = reiserfs_end_persistent_transaction(th);
2665 reiserfs_write_unlock(inode->i_sb);
2675 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2677 return generic_block_bmap(as, block, reiserfs_bmap);
2680 static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2681 loff_t pos, unsigned len, unsigned copied,
2682 struct page *page, void *fsdata)
2684 struct inode *inode = page->mapping->host;
2687 struct reiserfs_transaction_handle *th;
2690 bool locked = false;
2692 if ((unsigned long)fsdata & AOP_FLAG_CONT_EXPAND)
2695 reiserfs_wait_on_write_block(inode->i_sb);
2696 if (reiserfs_transaction_running(inode->i_sb))
2697 th = current->journal_info;
2701 start = pos & (PAGE_CACHE_SIZE - 1);
2702 if (unlikely(copied < len)) {
2703 if (!PageUptodate(page))
2706 page_zero_new_buffers(page, start + copied, start + len);
2708 flush_dcache_page(page);
2710 reiserfs_commit_page(inode, page, start, start + copied);
2712 /* generic_commit_write does this for us, but does not update the
2713 ** transaction tracking stuff when the size changes. So, we have
2714 ** to do the i_size updates here.
2716 if (pos + copied > inode->i_size) {
2717 struct reiserfs_transaction_handle myth;
2718 lock_depth = reiserfs_write_lock_once(inode->i_sb);
2720 /* If the file have grown beyond the border where it
2721 can have a tail, unmark it as needing a tail
2723 if ((have_large_tails(inode->i_sb)
2724 && inode->i_size > i_block_size(inode) * 4)
2725 || (have_small_tails(inode->i_sb)
2726 && inode->i_size > i_block_size(inode)))
2727 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2729 ret = journal_begin(&myth, inode->i_sb, 1);
2733 reiserfs_update_inode_transaction(inode);
2734 inode->i_size = pos + copied;
2736 * this will just nest into our transaction. It's important
2737 * to use mark_inode_dirty so the inode gets pushed around on the
2738 * dirty lists, and so that O_SYNC works as expected
2740 mark_inode_dirty(inode);
2741 reiserfs_update_sd(&myth, inode);
2743 ret = journal_end(&myth, inode->i_sb, 1);
2749 lock_depth = reiserfs_write_lock_once(inode->i_sb);
2753 mark_inode_dirty(inode);
2754 ret = reiserfs_end_persistent_transaction(th);
2761 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2763 page_cache_release(page);
2765 if (pos + len > inode->i_size)
2766 reiserfs_truncate_failed_write(inode);
2768 return ret == 0 ? copied : ret;
2771 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2775 reiserfs_update_sd(th, inode);
2776 ret = reiserfs_end_persistent_transaction(th);
2781 int reiserfs_commit_write(struct file *f, struct page *page,
2782 unsigned from, unsigned to)
2784 struct inode *inode = page->mapping->host;
2785 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2788 struct reiserfs_transaction_handle *th = NULL;
2790 reiserfs_write_unlock(inode->i_sb);
2791 reiserfs_wait_on_write_block(inode->i_sb);
2792 reiserfs_write_lock(inode->i_sb);
2794 if (reiserfs_transaction_running(inode->i_sb)) {
2795 th = current->journal_info;
2797 reiserfs_commit_page(inode, page, from, to);
2799 /* generic_commit_write does this for us, but does not update the
2800 ** transaction tracking stuff when the size changes. So, we have
2801 ** to do the i_size updates here.
2803 if (pos > inode->i_size) {
2804 struct reiserfs_transaction_handle myth;
2805 /* If the file have grown beyond the border where it
2806 can have a tail, unmark it as needing a tail
2808 if ((have_large_tails(inode->i_sb)
2809 && inode->i_size > i_block_size(inode) * 4)
2810 || (have_small_tails(inode->i_sb)
2811 && inode->i_size > i_block_size(inode)))
2812 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2814 ret = journal_begin(&myth, inode->i_sb, 1);
2818 reiserfs_update_inode_transaction(inode);
2819 inode->i_size = pos;
2821 * this will just nest into our transaction. It's important
2822 * to use mark_inode_dirty so the inode gets pushed around on the
2823 * dirty lists, and so that O_SYNC works as expected
2825 mark_inode_dirty(inode);
2826 reiserfs_update_sd(&myth, inode);
2828 ret = journal_end(&myth, inode->i_sb, 1);
2834 mark_inode_dirty(inode);
2835 ret = reiserfs_end_persistent_transaction(th);
2846 reiserfs_update_sd(th, inode);
2847 ret = reiserfs_end_persistent_transaction(th);
2853 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2855 if (reiserfs_attrs(inode->i_sb)) {
2856 if (sd_attrs & REISERFS_SYNC_FL)
2857 inode->i_flags |= S_SYNC;
2859 inode->i_flags &= ~S_SYNC;
2860 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2861 inode->i_flags |= S_IMMUTABLE;
2863 inode->i_flags &= ~S_IMMUTABLE;
2864 if (sd_attrs & REISERFS_APPEND_FL)
2865 inode->i_flags |= S_APPEND;
2867 inode->i_flags &= ~S_APPEND;
2868 if (sd_attrs & REISERFS_NOATIME_FL)
2869 inode->i_flags |= S_NOATIME;
2871 inode->i_flags &= ~S_NOATIME;
2872 if (sd_attrs & REISERFS_NOTAIL_FL)
2873 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2875 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2879 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
2881 if (reiserfs_attrs(inode->i_sb)) {
2882 if (inode->i_flags & S_IMMUTABLE)
2883 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2885 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2886 if (inode->i_flags & S_SYNC)
2887 *sd_attrs |= REISERFS_SYNC_FL;
2889 *sd_attrs &= ~REISERFS_SYNC_FL;
2890 if (inode->i_flags & S_NOATIME)
2891 *sd_attrs |= REISERFS_NOATIME_FL;
2893 *sd_attrs &= ~REISERFS_NOATIME_FL;
2894 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
2895 *sd_attrs |= REISERFS_NOTAIL_FL;
2897 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2901 /* decide if this buffer needs to stay around for data logging or ordered
2904 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2907 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2910 spin_lock(&j->j_dirty_buffers_lock);
2911 if (!buffer_mapped(bh)) {
2914 /* the page is locked, and the only places that log a data buffer
2915 * also lock the page.
2917 if (reiserfs_file_data_log(inode)) {
2919 * very conservative, leave the buffer pinned if
2920 * anyone might need it.
2922 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2925 } else if (buffer_dirty(bh)) {
2926 struct reiserfs_journal_list *jl;
2927 struct reiserfs_jh *jh = bh->b_private;
2929 /* why is this safe?
2930 * reiserfs_setattr updates i_size in the on disk
2931 * stat data before allowing vmtruncate to be called.
2933 * If buffer was put onto the ordered list for this
2934 * transaction, we know for sure either this transaction
2935 * or an older one already has updated i_size on disk,
2936 * and this ordered data won't be referenced in the file
2939 * if the buffer was put onto the ordered list for an older
2940 * transaction, we need to leave it around
2942 if (jh && (jl = jh->jl)
2943 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2947 if (ret && bh->b_private) {
2948 reiserfs_free_jh(bh);
2950 spin_unlock(&j->j_dirty_buffers_lock);
2955 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2956 static void reiserfs_invalidatepage(struct page *page, unsigned long offset)
2958 struct buffer_head *head, *bh, *next;
2959 struct inode *inode = page->mapping->host;
2960 unsigned int curr_off = 0;
2963 BUG_ON(!PageLocked(page));
2966 ClearPageChecked(page);
2968 if (!page_has_buffers(page))
2971 head = page_buffers(page);
2974 unsigned int next_off = curr_off + bh->b_size;
2975 next = bh->b_this_page;
2978 * is this block fully invalidated?
2980 if (offset <= curr_off) {
2981 if (invalidatepage_can_drop(inode, bh))
2982 reiserfs_unmap_buffer(bh);
2986 curr_off = next_off;
2988 } while (bh != head);
2991 * We release buffers only if the entire page is being invalidated.
2992 * The get_block cached value has been unconditionally invalidated,
2993 * so real IO is not possible anymore.
2995 if (!offset && ret) {
2996 ret = try_to_release_page(page, 0);
2997 /* maybe should BUG_ON(!ret); - neilb */
3003 static int reiserfs_set_page_dirty(struct page *page)
3005 struct inode *inode = page->mapping->host;
3006 if (reiserfs_file_data_log(inode)) {
3007 SetPageChecked(page);
3008 return __set_page_dirty_nobuffers(page);
3010 return __set_page_dirty_buffers(page);
3014 * Returns 1 if the page's buffers were dropped. The page is locked.
3016 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3017 * in the buffers at page_buffers(page).
3019 * even in -o notail mode, we can't be sure an old mount without -o notail
3020 * didn't create files with tails.
3022 static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
3024 struct inode *inode = page->mapping->host;
3025 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
3026 struct buffer_head *head;
3027 struct buffer_head *bh;
3030 WARN_ON(PageChecked(page));
3031 spin_lock(&j->j_dirty_buffers_lock);
3032 head = page_buffers(page);
3035 if (bh->b_private) {
3036 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
3037 reiserfs_free_jh(bh);
3043 bh = bh->b_this_page;
3044 } while (bh != head);
3046 ret = try_to_free_buffers(page);
3047 spin_unlock(&j->j_dirty_buffers_lock);
3051 /* We thank Mingming Cao for helping us understand in great detail what
3052 to do in this section of the code. */
3053 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
3054 const struct iovec *iov, loff_t offset,
3055 unsigned long nr_segs)
3057 struct file *file = iocb->ki_filp;
3058 struct inode *inode = file->f_mapping->host;
3061 ret = blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
3063 reiserfs_get_blocks_direct_io, NULL);
3066 * In case of error extending write may have instantiated a few
3067 * blocks outside i_size. Trim these off again.
3069 if (unlikely((rw & WRITE) && ret < 0)) {
3070 loff_t isize = i_size_read(inode);
3071 loff_t end = offset + iov_length(iov, nr_segs);
3074 vmtruncate(inode, isize);
3080 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3082 struct inode *inode = dentry->d_inode;
3083 unsigned int ia_valid;
3087 /* must be turned off for recursive notify_change calls */
3088 ia_valid = attr->ia_valid &= ~(ATTR_KILL_SUID|ATTR_KILL_SGID);
3090 depth = reiserfs_write_lock_once(inode->i_sb);
3091 if (is_quota_modification(inode, attr))
3092 dquot_initialize(inode);
3094 if (attr->ia_valid & ATTR_SIZE) {
3095 /* version 2 items will be caught by the s_maxbytes check
3096 ** done for us in vmtruncate
3098 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
3099 attr->ia_size > MAX_NON_LFS) {
3103 /* fill in hole pointers in the expanding truncate case. */
3104 if (attr->ia_size > inode->i_size) {
3105 error = generic_cont_expand_simple(inode, attr->ia_size);
3106 if (REISERFS_I(inode)->i_prealloc_count > 0) {
3108 struct reiserfs_transaction_handle th;
3109 /* we're changing at most 2 bitmaps, inode + super */
3110 err = journal_begin(&th, inode->i_sb, 4);
3112 reiserfs_discard_prealloc(&th, inode);
3113 err = journal_end(&th, inode->i_sb, 4);
3121 * file size is changed, ctime and mtime are
3124 attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
3128 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
3129 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
3130 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
3131 /* stat data of format v3.5 has 16 bit uid and gid */
3136 error = inode_change_ok(inode, attr);
3138 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
3139 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
3140 error = reiserfs_chown_xattrs(inode, attr);
3143 struct reiserfs_transaction_handle th;
3146 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3147 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3150 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3152 journal_begin(&th, inode->i_sb,
3156 error = dquot_transfer(inode, attr);
3158 journal_end(&th, inode->i_sb,
3162 /* Update corresponding info in inode so that everything is in
3163 * one transaction */
3164 if (attr->ia_valid & ATTR_UID)
3165 inode->i_uid = attr->ia_uid;
3166 if (attr->ia_valid & ATTR_GID)
3167 inode->i_gid = attr->ia_gid;
3168 mark_inode_dirty(inode);
3170 journal_end(&th, inode->i_sb, jbegin_count);
3175 * Relax the lock here, as it might truncate the
3176 * inode pages and wait for inode pages locks.
3177 * To release such page lock, the owner needs the
3180 reiserfs_write_unlock_once(inode->i_sb, depth);
3181 error = inode_setattr(inode, attr);
3182 depth = reiserfs_write_lock_once(inode->i_sb);
3186 if (!error && reiserfs_posixacl(inode->i_sb)) {
3187 if (attr->ia_valid & ATTR_MODE)
3188 error = reiserfs_acl_chmod(inode);
3192 reiserfs_write_unlock_once(inode->i_sb, depth);
3197 const struct address_space_operations reiserfs_address_space_operations = {
3198 .writepage = reiserfs_writepage,
3199 .readpage = reiserfs_readpage,
3200 .readpages = reiserfs_readpages,
3201 .releasepage = reiserfs_releasepage,
3202 .invalidatepage = reiserfs_invalidatepage,
3203 .sync_page = block_sync_page,
3204 .write_begin = reiserfs_write_begin,
3205 .write_end = reiserfs_write_end,
3206 .bmap = reiserfs_aop_bmap,
3207 .direct_IO = reiserfs_direct_IO,
3208 .set_page_dirty = reiserfs_set_page_dirty,