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 <asm/uaccess.h>
15 #include <asm/unaligned.h>
16 #include <linux/buffer_head.h>
17 #include <linux/mpage.h>
18 #include <linux/writeback.h>
19 #include <linux/quotaops.h>
20 #include <linux/swap.h>
22 int reiserfs_commit_write(struct file *f, struct page *page,
23 unsigned from, unsigned to);
24 int reiserfs_prepare_write(struct file *f, struct page *page,
25 unsigned from, unsigned to);
27 void reiserfs_delete_inode(struct inode *inode)
29 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
31 JOURNAL_PER_BALANCE_CNT * 2 +
32 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
33 struct reiserfs_transaction_handle th;
36 truncate_inode_pages(&inode->i_data, 0);
38 reiserfs_write_lock(inode->i_sb);
40 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
41 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
42 reiserfs_delete_xattrs(inode);
44 if (journal_begin(&th, inode->i_sb, jbegin_count))
46 reiserfs_update_inode_transaction(inode);
48 reiserfs_discard_prealloc(&th, inode);
50 err = reiserfs_delete_object(&th, inode);
52 /* Do quota update inside a transaction for journaled quotas. We must do that
53 * after delete_object so that quota updates go into the same transaction as
54 * stat data deletion */
56 vfs_dq_free_inode(inode);
58 if (journal_end(&th, inode->i_sb, jbegin_count))
61 /* check return value from reiserfs_delete_object after
62 * ending the transaction
67 /* all items of file are deleted, so we can remove "save" link */
68 remove_save_link(inode, 0 /* not truncate */ ); /* we can't do anything
69 * about an error here */
71 /* no object items are in the tree */
75 clear_inode(inode); /* note this must go after the journal_end to prevent deadlock */
77 reiserfs_write_unlock(inode->i_sb);
80 static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
81 __u32 objectid, loff_t offset, int type, int length)
83 key->version = version;
85 key->on_disk_key.k_dir_id = dirid;
86 key->on_disk_key.k_objectid = objectid;
87 set_cpu_key_k_offset(key, offset);
88 set_cpu_key_k_type(key, type);
89 key->key_length = length;
92 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
93 offset and type of key */
94 void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
97 _make_cpu_key(key, get_inode_item_key_version(inode),
98 le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
99 le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
104 // when key is 0, do not set version and short key
106 inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
108 loff_t offset, int type, int length,
109 int entry_count /*or ih_free_space */ )
112 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
113 ih->ih_key.k_objectid =
114 cpu_to_le32(key->on_disk_key.k_objectid);
116 put_ih_version(ih, version);
117 set_le_ih_k_offset(ih, offset);
118 set_le_ih_k_type(ih, type);
119 put_ih_item_len(ih, length);
120 /* set_ih_free_space (ih, 0); */
121 // for directory items it is entry count, for directs and stat
122 // datas - 0xffff, for indirects - 0
123 put_ih_entry_count(ih, entry_count);
127 // FIXME: we might cache recently accessed indirect item
129 // Ugh. Not too eager for that....
130 // I cut the code until such time as I see a convincing argument (benchmark).
131 // I don't want a bloated inode struct..., and I don't like code complexity....
133 /* cutting the code is fine, since it really isn't in use yet and is easy
134 ** to add back in. But, Vladimir has a really good idea here. Think
135 ** about what happens for reading a file. For each page,
136 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
137 ** an indirect item. This indirect item has X number of pointers, where
138 ** X is a big number if we've done the block allocation right. But,
139 ** we only use one or two of these pointers during each call to readpage,
140 ** needlessly researching again later on.
142 ** The size of the cache could be dynamic based on the size of the file.
144 ** I'd also like to see us cache the location the stat data item, since
145 ** we are needlessly researching for that frequently.
150 /* If this page has a file tail in it, and
151 ** it was read in by get_block_create_0, the page data is valid,
152 ** but tail is still sitting in a direct item, and we can't write to
153 ** it. So, look through this page, and check all the mapped buffers
154 ** to make sure they have valid block numbers. Any that don't need
155 ** to be unmapped, so that block_prepare_write will correctly call
156 ** reiserfs_get_block to convert the tail into an unformatted node
158 static inline void fix_tail_page_for_writing(struct page *page)
160 struct buffer_head *head, *next, *bh;
162 if (page && page_has_buffers(page)) {
163 head = page_buffers(page);
166 next = bh->b_this_page;
167 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
168 reiserfs_unmap_buffer(bh);
171 } while (bh != head);
175 /* reiserfs_get_block does not need to allocate a block only if it has been
176 done already or non-hole position has been found in the indirect item */
177 static inline int allocation_needed(int retval, b_blocknr_t allocated,
178 struct item_head *ih,
179 __le32 * item, int pos_in_item)
183 if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
184 get_block_num(item, pos_in_item))
189 static inline int indirect_item_found(int retval, struct item_head *ih)
191 return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
194 static inline void set_block_dev_mapped(struct buffer_head *bh,
195 b_blocknr_t block, struct inode *inode)
197 map_bh(bh, inode->i_sb, block);
201 // files which were created in the earlier version can not be longer,
204 static int file_capable(struct inode *inode, sector_t block)
206 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file.
207 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
213 static int restart_transaction(struct reiserfs_transaction_handle *th,
214 struct inode *inode, struct treepath *path)
216 struct super_block *s = th->t_super;
217 int len = th->t_blocks_allocated;
220 BUG_ON(!th->t_trans_id);
221 BUG_ON(!th->t_refcount);
225 /* we cannot restart while nested */
226 if (th->t_refcount > 1) {
229 reiserfs_update_sd(th, inode);
230 err = journal_end(th, s, len);
232 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
234 reiserfs_update_inode_transaction(inode);
239 // it is called by get_block when create == 0. Returns block number
240 // for 'block'-th logical block of file. When it hits direct item it
241 // returns 0 (being called from bmap) or read direct item into piece
242 // of page (bh_result)
244 // Please improve the english/clarity in the comment above, as it is
245 // hard to understand.
247 static int _get_block_create_0(struct inode *inode, sector_t block,
248 struct buffer_head *bh_result, int args)
250 INITIALIZE_PATH(path);
252 struct buffer_head *bh;
253 struct item_head *ih, tmp_ih;
261 unsigned long offset;
263 // prepare the key to look for the 'block'-th block of file
264 make_cpu_key(&key, inode,
265 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
269 result = search_for_position_by_key(inode->i_sb, &key, &path);
270 if (result != POSITION_FOUND) {
273 kunmap(bh_result->b_page);
274 if (result == IO_ERROR)
276 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
277 // That there is some MMAPED data associated with it that is yet to be written to disk.
278 if ((args & GET_BLOCK_NO_HOLE)
279 && !PageUptodate(bh_result->b_page)) {
285 bh = get_last_bh(&path);
287 if (is_indirect_le_ih(ih)) {
288 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih);
290 /* FIXME: here we could cache indirect item or part of it in
291 the inode to avoid search_by_key in case of subsequent
293 blocknr = get_block_num(ind_item, path.pos_in_item);
296 map_bh(bh_result, inode->i_sb, blocknr);
297 if (path.pos_in_item ==
298 ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
299 set_buffer_boundary(bh_result);
302 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
303 // That there is some MMAPED data associated with it that is yet to be written to disk.
304 if ((args & GET_BLOCK_NO_HOLE)
305 && !PageUptodate(bh_result->b_page)) {
311 kunmap(bh_result->b_page);
314 // requested data are in direct item(s)
315 if (!(args & GET_BLOCK_READ_DIRECT)) {
316 // we are called by bmap. FIXME: we can not map block of file
317 // when it is stored in direct item(s)
320 kunmap(bh_result->b_page);
324 /* if we've got a direct item, and the buffer or page was uptodate,
325 ** we don't want to pull data off disk again. skip to the
326 ** end, where we map the buffer and return
328 if (buffer_uptodate(bh_result)) {
332 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
333 ** pages without any buffers. If the page is up to date, we don't want
334 ** read old data off disk. Set the up to date bit on the buffer instead
335 ** and jump to the end
337 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
338 set_buffer_uptodate(bh_result);
341 // read file tail into part of page
342 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
343 fs_gen = get_generation(inode->i_sb);
344 copy_item_head(&tmp_ih, ih);
346 /* we only want to kmap if we are reading the tail into the page.
347 ** this is not the common case, so we don't kmap until we are
348 ** sure we need to. But, this means the item might move if
352 p = (char *)kmap(bh_result->b_page);
353 if (fs_changed(fs_gen, inode->i_sb)
354 && item_moved(&tmp_ih, &path)) {
359 memset(p, 0, inode->i_sb->s_blocksize);
361 if (!is_direct_le_ih(ih)) {
364 /* make sure we don't read more bytes than actually exist in
365 ** the file. This can happen in odd cases where i_size isn't
366 ** correct, and when direct item padding results in a few
367 ** extra bytes at the end of the direct item
369 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
371 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
373 inode->i_size - (le_ih_k_offset(ih) - 1) -
377 chars = ih_item_len(ih) - path.pos_in_item;
379 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
386 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
387 // we done, if read direct item is not the last item of
388 // node FIXME: we could try to check right delimiting key
389 // to see whether direct item continues in the right
390 // neighbor or rely on i_size
393 // update key to look for the next piece
394 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
395 result = search_for_position_by_key(inode->i_sb, &key, &path);
396 if (result != POSITION_FOUND)
397 // i/o error most likely
399 bh = get_last_bh(&path);
403 flush_dcache_page(bh_result->b_page);
404 kunmap(bh_result->b_page);
409 if (result == IO_ERROR)
412 /* this buffer has valid data, but isn't valid for io. mapping it to
413 * block #0 tells the rest of reiserfs it just has a tail in it
415 map_bh(bh_result, inode->i_sb, 0);
416 set_buffer_uptodate(bh_result);
420 // this is called to create file map. So, _get_block_create_0 will not
422 static int reiserfs_bmap(struct inode *inode, sector_t block,
423 struct buffer_head *bh_result, int create)
425 if (!file_capable(inode, block))
428 reiserfs_write_lock(inode->i_sb);
429 /* do not read the direct item */
430 _get_block_create_0(inode, block, bh_result, 0);
431 reiserfs_write_unlock(inode->i_sb);
435 /* special version of get_block that is only used by grab_tail_page right
436 ** now. It is sent to block_prepare_write, and when you try to get a
437 ** block past the end of the file (or a block from a hole) it returns
438 ** -ENOENT instead of a valid buffer. block_prepare_write expects to
439 ** be able to do i/o on the buffers returned, unless an error value
442 ** So, this allows block_prepare_write to be used for reading a single block
443 ** in a page. Where it does not produce a valid page for holes, or past the
444 ** end of the file. This turns out to be exactly what we need for reading
445 ** tails for conversion.
447 ** The point of the wrapper is forcing a certain value for create, even
448 ** though the VFS layer is calling this function with create==1. If you
449 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
450 ** don't use this function.
452 static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
453 struct buffer_head *bh_result,
456 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
459 /* This is special helper for reiserfs_get_block in case we are executing
460 direct_IO request. */
461 static int reiserfs_get_blocks_direct_io(struct inode *inode,
463 struct buffer_head *bh_result,
468 bh_result->b_page = NULL;
470 /* We set the b_size before reiserfs_get_block call since it is
471 referenced in convert_tail_for_hole() that may be called from
472 reiserfs_get_block() */
473 bh_result->b_size = (1 << inode->i_blkbits);
475 ret = reiserfs_get_block(inode, iblock, bh_result,
476 create | GET_BLOCK_NO_DANGLE);
480 /* don't allow direct io onto tail pages */
481 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
482 /* make sure future calls to the direct io funcs for this offset
483 ** in the file fail by unmapping the buffer
485 clear_buffer_mapped(bh_result);
488 /* Possible unpacked tail. Flush the data before pages have
490 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
493 reiserfs_write_lock(inode->i_sb);
495 err = reiserfs_commit_for_inode(inode);
496 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
498 reiserfs_write_unlock(inode->i_sb);
508 ** helper function for when reiserfs_get_block is called for a hole
509 ** but the file tail is still in a direct item
510 ** bh_result is the buffer head for the hole
511 ** tail_offset is the offset of the start of the tail in the file
513 ** This calls prepare_write, which will start a new transaction
514 ** you should not be in a transaction, or have any paths held when you
517 static int convert_tail_for_hole(struct inode *inode,
518 struct buffer_head *bh_result,
522 unsigned long tail_end;
523 unsigned long tail_start;
524 struct page *tail_page;
525 struct page *hole_page = bh_result->b_page;
528 if ((tail_offset & (bh_result->b_size - 1)) != 1)
531 /* always try to read until the end of the block */
532 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
533 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
535 index = tail_offset >> PAGE_CACHE_SHIFT;
536 /* hole_page can be zero in case of direct_io, we are sure
537 that we cannot get here if we write with O_DIRECT into
539 if (!hole_page || index != hole_page->index) {
540 tail_page = grab_cache_page(inode->i_mapping, index);
546 tail_page = hole_page;
549 /* we don't have to make sure the conversion did not happen while
550 ** we were locking the page because anyone that could convert
551 ** must first take i_mutex.
553 ** We must fix the tail page for writing because it might have buffers
554 ** that are mapped, but have a block number of 0. This indicates tail
555 ** data that has been read directly into the page, and block_prepare_write
556 ** won't trigger a get_block in this case.
558 fix_tail_page_for_writing(tail_page);
559 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
563 /* tail conversion might change the data in the page */
564 flush_dcache_page(tail_page);
566 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
569 if (tail_page != hole_page) {
570 unlock_page(tail_page);
571 page_cache_release(tail_page);
577 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
580 b_blocknr_t * allocated_block_nr,
581 struct treepath *path, int flags)
583 BUG_ON(!th->t_trans_id);
585 #ifdef REISERFS_PREALLOCATE
586 if (!(flags & GET_BLOCK_NO_IMUX)) {
587 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
591 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
595 int reiserfs_get_block(struct inode *inode, sector_t block,
596 struct buffer_head *bh_result, int create)
598 int repeat, retval = 0;
599 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
600 INITIALIZE_PATH(path);
603 struct buffer_head *bh, *unbh = NULL;
604 struct item_head *ih, tmp_ih;
608 struct reiserfs_transaction_handle *th = NULL;
609 /* space reserved in transaction batch:
610 . 3 balancings in direct->indirect conversion
611 . 1 block involved into reiserfs_update_sd()
612 XXX in practically impossible worst case direct2indirect()
613 can incur (much) more than 3 balancings.
614 quota update for user, group */
616 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
617 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
621 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
623 reiserfs_write_lock(inode->i_sb);
624 version = get_inode_item_key_version(inode);
626 if (!file_capable(inode, block)) {
627 reiserfs_write_unlock(inode->i_sb);
631 /* if !create, we aren't changing the FS, so we don't need to
632 ** log anything, so we don't need to start a transaction
634 if (!(create & GET_BLOCK_CREATE)) {
636 /* find number of block-th logical block of the file */
637 ret = _get_block_create_0(inode, block, bh_result,
638 create | GET_BLOCK_READ_DIRECT);
639 reiserfs_write_unlock(inode->i_sb);
643 * if we're already in a transaction, make sure to close
644 * any new transactions we start in this func
646 if ((create & GET_BLOCK_NO_DANGLE) ||
647 reiserfs_transaction_running(inode->i_sb))
650 /* If file is of such a size, that it might have a tail and tails are enabled
651 ** we should mark it as possibly needing tail packing on close
653 if ((have_large_tails(inode->i_sb)
654 && inode->i_size < i_block_size(inode) * 4)
655 || (have_small_tails(inode->i_sb)
656 && inode->i_size < i_block_size(inode)))
657 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
659 /* set the key of the first byte in the 'block'-th block of file */
660 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
661 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
663 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
668 reiserfs_update_inode_transaction(inode);
672 retval = search_for_position_by_key(inode->i_sb, &key, &path);
673 if (retval == IO_ERROR) {
678 bh = get_last_bh(&path);
680 item = get_item(&path);
681 pos_in_item = path.pos_in_item;
683 fs_gen = get_generation(inode->i_sb);
684 copy_item_head(&tmp_ih, ih);
686 if (allocation_needed
687 (retval, allocated_block_nr, ih, item, pos_in_item)) {
688 /* we have to allocate block for the unformatted node */
695 _allocate_block(th, block, inode, &allocated_block_nr,
698 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
699 /* restart the transaction to give the journal a chance to free
700 ** some blocks. releases the path, so we have to go back to
701 ** research if we succeed on the second try
703 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
704 retval = restart_transaction(th, inode, &path);
708 _allocate_block(th, block, inode,
709 &allocated_block_nr, NULL, create);
711 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
714 if (repeat == QUOTA_EXCEEDED)
721 if (fs_changed(fs_gen, inode->i_sb)
722 && item_moved(&tmp_ih, &path)) {
727 if (indirect_item_found(retval, ih)) {
728 b_blocknr_t unfm_ptr;
729 /* 'block'-th block is in the file already (there is
730 corresponding cell in some indirect item). But it may be
731 zero unformatted node pointer (hole) */
732 unfm_ptr = get_block_num(item, pos_in_item);
734 /* use allocated block to plug the hole */
735 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
736 if (fs_changed(fs_gen, inode->i_sb)
737 && item_moved(&tmp_ih, &path)) {
738 reiserfs_restore_prepared_buffer(inode->i_sb,
742 set_buffer_new(bh_result);
743 if (buffer_dirty(bh_result)
744 && reiserfs_data_ordered(inode->i_sb))
745 reiserfs_add_ordered_list(inode, bh_result);
746 put_block_num(item, pos_in_item, allocated_block_nr);
747 unfm_ptr = allocated_block_nr;
748 journal_mark_dirty(th, inode->i_sb, bh);
749 reiserfs_update_sd(th, inode);
751 set_block_dev_mapped(bh_result, unfm_ptr, inode);
755 retval = reiserfs_end_persistent_transaction(th);
757 reiserfs_write_unlock(inode->i_sb);
759 /* the item was found, so new blocks were not added to the file
760 ** there is no need to make sure the inode is updated with this
771 /* desired position is not found or is in the direct item. We have
772 to append file with holes up to 'block'-th block converting
773 direct items to indirect one if necessary */
776 if (is_statdata_le_ih(ih)) {
778 struct cpu_key tmp_key;
780 /* indirect item has to be inserted */
781 make_le_item_head(&tmp_ih, &key, version, 1,
782 TYPE_INDIRECT, UNFM_P_SIZE,
783 0 /* free_space */ );
785 if (cpu_key_k_offset(&key) == 1) {
786 /* we are going to add 'block'-th block to the file. Use
787 allocated block for that */
788 unp = cpu_to_le32(allocated_block_nr);
789 set_block_dev_mapped(bh_result,
790 allocated_block_nr, inode);
791 set_buffer_new(bh_result);
795 set_cpu_key_k_offset(&tmp_key, 1);
796 PATH_LAST_POSITION(&path)++;
799 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
800 inode, (char *)&unp);
802 reiserfs_free_block(th, inode,
803 allocated_block_nr, 1);
804 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
806 //mark_tail_converted (inode);
807 } else if (is_direct_le_ih(ih)) {
808 /* direct item has to be converted */
812 ((le_ih_k_offset(ih) -
813 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
814 if (tail_offset == cpu_key_k_offset(&key)) {
815 /* direct item we just found fits into block we have
816 to map. Convert it into unformatted node: use
817 bh_result for the conversion */
818 set_block_dev_mapped(bh_result,
819 allocated_block_nr, inode);
823 /* we have to padd file tail stored in direct item(s)
824 up to block size and convert it to unformatted
825 node. FIXME: this should also get into page cache */
829 * ugly, but we can only end the transaction if
832 BUG_ON(!th->t_refcount);
833 if (th->t_refcount == 1) {
835 reiserfs_end_persistent_transaction
843 convert_tail_for_hole(inode, bh_result,
846 if (retval != -ENOSPC)
847 reiserfs_error(inode->i_sb,
849 "convert tail failed "
850 "inode %lu, error %d",
853 if (allocated_block_nr) {
854 /* the bitmap, the super, and the stat data == 3 */
856 th = reiserfs_persistent_transaction(inode->i_sb, 3);
858 reiserfs_free_block(th,
868 direct2indirect(th, inode, &path, unbh,
871 reiserfs_unmap_buffer(unbh);
872 reiserfs_free_block(th, inode,
873 allocated_block_nr, 1);
876 /* it is important the set_buffer_uptodate is done after
877 ** the direct2indirect. The buffer might contain valid
878 ** data newer than the data on disk (read by readpage, changed,
879 ** and then sent here by writepage). direct2indirect needs
880 ** to know if unbh was already up to date, so it can decide
881 ** if the data in unbh needs to be replaced with data from
884 set_buffer_uptodate(unbh);
886 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
887 buffer will disappear shortly, so it should not be added to
890 /* we've converted the tail, so we must
891 ** flush unbh before the transaction commits
893 reiserfs_add_tail_list(inode, unbh);
895 /* mark it dirty now to prevent commit_write from adding
896 ** this buffer to the inode's dirty buffer list
899 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
900 * It's still atomic, but it sets the page dirty too,
901 * which makes it eligible for writeback at any time by the
902 * VM (which was also the case with __mark_buffer_dirty())
904 mark_buffer_dirty(unbh);
907 /* append indirect item with holes if needed, when appending
908 pointer to 'block'-th block use block, which is already
910 struct cpu_key tmp_key;
911 unp_t unf_single = 0; // We use this in case we need to allocate only
912 // one block which is a fastpath
914 __u64 max_to_insert =
915 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
919 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
920 "vs-804: invalid position for append");
921 /* indirect item has to be appended, set up key of that position */
922 make_cpu_key(&tmp_key, inode,
923 le_key_k_offset(version,
926 inode->i_sb->s_blocksize),
927 //pos_in_item * inode->i_sb->s_blocksize,
928 TYPE_INDIRECT, 3); // key type is unimportant
930 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
931 "green-805: invalid offset");
934 ((cpu_key_k_offset(&key) -
935 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
938 if (blocks_needed == 1) {
941 un = kzalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_ATOMIC); // We need to avoid scheduling.
948 if (blocks_needed <= max_to_insert) {
949 /* we are going to add target block to the file. Use allocated
951 un[blocks_needed - 1] =
952 cpu_to_le32(allocated_block_nr);
953 set_block_dev_mapped(bh_result,
954 allocated_block_nr, inode);
955 set_buffer_new(bh_result);
958 /* paste hole to the indirect item */
959 /* If kmalloc failed, max_to_insert becomes zero and it means we
960 only have space for one block */
962 max_to_insert ? max_to_insert : 1;
965 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
970 if (blocks_needed != 1)
974 reiserfs_free_block(th, inode,
975 allocated_block_nr, 1);
979 /* We need to mark new file size in case this function will be
980 interrupted/aborted later on. And we may do this only for
983 inode->i_sb->s_blocksize * blocks_needed;
990 /* this loop could log more blocks than we had originally asked
991 ** for. So, we have to allow the transaction to end if it is
992 ** too big or too full. Update the inode so things are
993 ** consistent if we crash before the function returns
995 ** release the path so that anybody waiting on the path before
996 ** ending their transaction will be able to continue.
998 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
999 retval = restart_transaction(th, inode, &path);
1004 * inserting indirect pointers for a hole can take a
1005 * long time. reschedule if needed and also release the write
1008 reiserfs_write_unlock(inode->i_sb);
1010 reiserfs_write_lock(inode->i_sb);
1012 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1013 if (retval == IO_ERROR) {
1017 if (retval == POSITION_FOUND) {
1018 reiserfs_warning(inode->i_sb, "vs-825",
1019 "%K should not be found", &key);
1021 if (allocated_block_nr)
1022 reiserfs_free_block(th, inode,
1023 allocated_block_nr, 1);
1027 bh = get_last_bh(&path);
1029 item = get_item(&path);
1030 pos_in_item = path.pos_in_item;
1036 if (th && (!dangle || (retval && !th->t_trans_id))) {
1039 reiserfs_update_sd(th, inode);
1040 err = reiserfs_end_persistent_transaction(th);
1045 reiserfs_write_unlock(inode->i_sb);
1046 reiserfs_check_path(&path);
1051 reiserfs_readpages(struct file *file, struct address_space *mapping,
1052 struct list_head *pages, unsigned nr_pages)
1054 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
1057 /* Compute real number of used bytes by file
1058 * Following three functions can go away when we'll have enough space in stat item
1060 static int real_space_diff(struct inode *inode, int sd_size)
1063 loff_t blocksize = inode->i_sb->s_blocksize;
1065 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1068 /* End of file is also in full block with indirect reference, so round
1069 ** up to the next block.
1071 ** there is just no way to know if the tail is actually packed
1072 ** on the file, so we have to assume it isn't. When we pack the
1073 ** tail, we add 4 bytes to pretend there really is an unformatted
1078 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1083 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
1086 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1087 return inode->i_size +
1088 (loff_t) (real_space_diff(inode, sd_size));
1090 return ((loff_t) real_space_diff(inode, sd_size)) +
1091 (((loff_t) blocks) << 9);
1094 /* Compute number of blocks used by file in ReiserFS counting */
1095 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1097 loff_t bytes = inode_get_bytes(inode);
1098 loff_t real_space = real_space_diff(inode, sd_size);
1100 /* keeps fsck and non-quota versions of reiserfs happy */
1101 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1102 bytes += (loff_t) 511;
1105 /* files from before the quota patch might i_blocks such that
1106 ** bytes < real_space. Deal with that here to prevent it from
1109 if (bytes < real_space)
1111 return (bytes - real_space) >> 9;
1115 // BAD: new directories have stat data of new type and all other items
1116 // of old type. Version stored in the inode says about body items, so
1117 // in update_stat_data we can not rely on inode, but have to check
1118 // item version directly
1121 // called by read_locked_inode
1122 static void init_inode(struct inode *inode, struct treepath *path)
1124 struct buffer_head *bh;
1125 struct item_head *ih;
1127 //int version = ITEM_VERSION_1;
1129 bh = PATH_PLAST_BUFFER(path);
1130 ih = PATH_PITEM_HEAD(path);
1132 copy_key(INODE_PKEY(inode), &(ih->ih_key));
1134 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1135 REISERFS_I(inode)->i_flags = 0;
1136 REISERFS_I(inode)->i_prealloc_block = 0;
1137 REISERFS_I(inode)->i_prealloc_count = 0;
1138 REISERFS_I(inode)->i_trans_id = 0;
1139 REISERFS_I(inode)->i_jl = NULL;
1140 mutex_init(&(REISERFS_I(inode)->i_mmap));
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 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1504 reiserfs_find_actor, reiserfs_init_locked_inode,
1507 return ERR_PTR(-ENOMEM);
1509 if (inode->i_state & I_NEW) {
1510 reiserfs_read_locked_inode(inode, &args);
1511 unlock_new_inode(inode);
1514 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1515 /* either due to i/o error or a stale NFS handle */
1522 static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1523 u32 objectid, u32 dir_id, u32 generation)
1527 struct inode *inode;
1529 key.on_disk_key.k_objectid = objectid;
1530 key.on_disk_key.k_dir_id = dir_id;
1531 reiserfs_write_lock(sb);
1532 inode = reiserfs_iget(sb, &key);
1533 if (inode && !IS_ERR(inode) && generation != 0 &&
1534 generation != inode->i_generation) {
1538 reiserfs_write_unlock(sb);
1540 return d_obtain_alias(inode);
1543 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1544 int fh_len, int fh_type)
1546 /* fhtype happens to reflect the number of u32s encoded.
1547 * due to a bug in earlier code, fhtype might indicate there
1548 * are more u32s then actually fitted.
1549 * so if fhtype seems to be more than len, reduce fhtype.
1551 * 2 - objectid + dir_id - legacy support
1552 * 3 - objectid + dir_id + generation
1553 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1554 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1555 * 6 - as above plus generation of directory
1556 * 6 does not fit in NFSv2 handles
1558 if (fh_type > fh_len) {
1559 if (fh_type != 6 || fh_len != 5)
1560 reiserfs_warning(sb, "reiserfs-13077",
1561 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1566 return reiserfs_get_dentry(sb, fid->raw[0], fid->raw[1],
1567 (fh_type == 3 || fh_type >= 5) ? fid->raw[2] : 0);
1570 struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1571 int fh_len, int fh_type)
1576 return reiserfs_get_dentry(sb,
1577 (fh_type >= 5) ? fid->raw[3] : fid->raw[2],
1578 (fh_type >= 5) ? fid->raw[4] : fid->raw[3],
1579 (fh_type == 6) ? fid->raw[5] : 0);
1582 int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1585 struct inode *inode = dentry->d_inode;
1591 data[0] = inode->i_ino;
1592 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1593 data[2] = inode->i_generation;
1595 /* no room for directory info? return what we've stored so far */
1596 if (maxlen < 5 || !need_parent)
1599 spin_lock(&dentry->d_lock);
1600 inode = dentry->d_parent->d_inode;
1601 data[3] = inode->i_ino;
1602 data[4] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1605 data[5] = inode->i_generation;
1608 spin_unlock(&dentry->d_lock);
1612 /* looks for stat data, then copies fields to it, marks the buffer
1613 containing stat data as dirty */
1614 /* reiserfs inodes are never really dirty, since the dirty inode call
1615 ** always logs them. This call allows the VFS inode marking routines
1616 ** to properly mark inodes for datasync and such, but only actually
1617 ** does something when called for a synchronous update.
1619 int reiserfs_write_inode(struct inode *inode, int do_sync)
1621 struct reiserfs_transaction_handle th;
1622 int jbegin_count = 1;
1624 if (inode->i_sb->s_flags & MS_RDONLY)
1626 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1627 ** these cases are just when the system needs ram, not when the
1628 ** inode needs to reach disk for safety, and they can safely be
1629 ** ignored because the altered inode has already been logged.
1631 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1632 reiserfs_write_lock(inode->i_sb);
1633 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1634 reiserfs_update_sd(&th, inode);
1635 journal_end_sync(&th, inode->i_sb, jbegin_count);
1637 reiserfs_write_unlock(inode->i_sb);
1642 /* stat data of new object is inserted already, this inserts the item
1643 containing "." and ".." entries */
1644 static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1645 struct inode *inode,
1646 struct item_head *ih, struct treepath *path,
1649 struct super_block *sb = th->t_super;
1650 char empty_dir[EMPTY_DIR_SIZE];
1651 char *body = empty_dir;
1655 BUG_ON(!th->t_trans_id);
1657 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1658 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1659 TYPE_DIRENTRY, 3 /*key length */ );
1661 /* compose item head for new item. Directories consist of items of
1662 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1663 is done by reiserfs_new_inode */
1664 if (old_format_only(sb)) {
1665 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1666 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1668 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1669 ih->ih_key.k_objectid,
1670 INODE_PKEY(dir)->k_dir_id,
1671 INODE_PKEY(dir)->k_objectid);
1673 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1674 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1676 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1677 ih->ih_key.k_objectid,
1678 INODE_PKEY(dir)->k_dir_id,
1679 INODE_PKEY(dir)->k_objectid);
1682 /* look for place in the tree for new item */
1683 retval = search_item(sb, &key, path);
1684 if (retval == IO_ERROR) {
1685 reiserfs_error(sb, "vs-13080",
1686 "i/o failure occurred creating new directory");
1689 if (retval == ITEM_FOUND) {
1691 reiserfs_warning(sb, "vs-13070",
1692 "object with this key exists (%k)",
1697 /* insert item, that is empty directory item */
1698 return reiserfs_insert_item(th, path, &key, ih, inode, body);
1701 /* stat data of object has been inserted, this inserts the item
1702 containing the body of symlink */
1703 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1704 struct item_head *ih,
1705 struct treepath *path, const char *symname,
1708 struct super_block *sb = th->t_super;
1712 BUG_ON(!th->t_trans_id);
1714 _make_cpu_key(&key, KEY_FORMAT_3_5,
1715 le32_to_cpu(ih->ih_key.k_dir_id),
1716 le32_to_cpu(ih->ih_key.k_objectid),
1717 1, TYPE_DIRECT, 3 /*key length */ );
1719 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1720 0 /*free_space */ );
1722 /* look for place in the tree for new item */
1723 retval = search_item(sb, &key, path);
1724 if (retval == IO_ERROR) {
1725 reiserfs_error(sb, "vs-13080",
1726 "i/o failure occurred creating new symlink");
1729 if (retval == ITEM_FOUND) {
1731 reiserfs_warning(sb, "vs-13080",
1732 "object with this key exists (%k)",
1737 /* insert item, that is body of symlink */
1738 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
1741 /* inserts the stat data into the tree, and then calls
1742 reiserfs_new_directory (to insert ".", ".." item if new object is
1743 directory) or reiserfs_new_symlink (to insert symlink body if new
1744 object is symlink) or nothing (if new object is regular file)
1746 NOTE! uid and gid must already be set in the inode. If we return
1747 non-zero due to an error, we have to drop the quota previously allocated
1748 for the fresh inode. This can only be done outside a transaction, so
1749 if we return non-zero, we also end the transaction. */
1750 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1751 struct inode *dir, int mode, const char *symname,
1752 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1753 strlen (symname) for symlinks) */
1754 loff_t i_size, struct dentry *dentry,
1755 struct inode *inode,
1756 struct reiserfs_security_handle *security)
1758 struct super_block *sb;
1759 struct reiserfs_iget_args args;
1760 INITIALIZE_PATH(path_to_key);
1762 struct item_head ih;
1763 struct stat_data sd;
1767 BUG_ON(!th->t_trans_id);
1769 if (vfs_dq_alloc_inode(inode)) {
1773 if (!dir->i_nlink) {
1780 /* item head of new item */
1781 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1782 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1783 if (!ih.ih_key.k_objectid) {
1787 args.objectid = inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
1788 if (old_format_only(sb))
1789 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1790 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1792 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1793 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1794 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1795 args.dirid = le32_to_cpu(ih.ih_key.k_dir_id);
1796 if (insert_inode_locked4(inode, args.objectid,
1797 reiserfs_find_actor, &args) < 0) {
1801 if (old_format_only(sb))
1802 /* not a perfect generation count, as object ids can be reused, but
1803 ** this is as good as reiserfs can do right now.
1804 ** note that the private part of inode isn't filled in yet, we have
1805 ** to use the directory.
1807 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1809 #if defined( USE_INODE_GENERATION_COUNTER )
1810 inode->i_generation =
1811 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1813 inode->i_generation = ++event;
1816 /* fill stat data */
1817 inode->i_nlink = (S_ISDIR(mode) ? 2 : 1);
1819 /* uid and gid must already be set by the caller for quota init */
1821 /* symlink cannot be immutable or append only, right? */
1822 if (S_ISLNK(inode->i_mode))
1823 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1825 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1826 inode->i_size = i_size;
1827 inode->i_blocks = 0;
1829 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1830 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1832 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1833 REISERFS_I(inode)->i_flags = 0;
1834 REISERFS_I(inode)->i_prealloc_block = 0;
1835 REISERFS_I(inode)->i_prealloc_count = 0;
1836 REISERFS_I(inode)->i_trans_id = 0;
1837 REISERFS_I(inode)->i_jl = NULL;
1838 REISERFS_I(inode)->i_attrs =
1839 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1840 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
1841 mutex_init(&(REISERFS_I(inode)->i_mmap));
1842 reiserfs_init_xattr_rwsem(inode);
1844 /* key to search for correct place for new stat data */
1845 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1846 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1847 TYPE_STAT_DATA, 3 /*key length */ );
1849 /* find proper place for inserting of stat data */
1850 retval = search_item(sb, &key, &path_to_key);
1851 if (retval == IO_ERROR) {
1855 if (retval == ITEM_FOUND) {
1856 pathrelse(&path_to_key);
1860 if (old_format_only(sb)) {
1861 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1862 pathrelse(&path_to_key);
1863 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1867 inode2sd_v1(&sd, inode, inode->i_size);
1869 inode2sd(&sd, inode, inode->i_size);
1871 // store in in-core inode the key of stat data and version all
1872 // object items will have (directory items will have old offset
1873 // format, other new objects will consist of new items)
1874 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1875 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1877 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1878 if (old_format_only(sb))
1879 set_inode_sd_version(inode, STAT_DATA_V1);
1881 set_inode_sd_version(inode, STAT_DATA_V2);
1883 /* insert the stat data into the tree */
1884 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1885 if (REISERFS_I(dir)->new_packing_locality)
1886 th->displace_new_blocks = 1;
1889 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1893 reiserfs_check_path(&path_to_key);
1896 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1897 if (!th->displace_new_blocks)
1898 REISERFS_I(dir)->new_packing_locality = 0;
1900 if (S_ISDIR(mode)) {
1901 /* insert item with "." and ".." */
1903 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1906 if (S_ISLNK(mode)) {
1907 /* insert body of symlink */
1908 if (!old_format_only(sb))
1909 i_size = ROUND_UP(i_size);
1911 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1916 reiserfs_check_path(&path_to_key);
1917 journal_end(th, th->t_super, th->t_blocks_allocated);
1918 goto out_inserted_sd;
1921 if (reiserfs_posixacl(inode->i_sb)) {
1922 retval = reiserfs_inherit_default_acl(th, dir, dentry, inode);
1925 reiserfs_check_path(&path_to_key);
1926 journal_end(th, th->t_super, th->t_blocks_allocated);
1927 goto out_inserted_sd;
1929 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1930 reiserfs_warning(inode->i_sb, "jdm-13090",
1931 "ACLs aren't enabled in the fs, "
1932 "but vfs thinks they are!");
1933 } else if (IS_PRIVATE(dir))
1934 inode->i_flags |= S_PRIVATE;
1936 if (security->name) {
1937 retval = reiserfs_security_write(th, inode, security);
1940 reiserfs_check_path(&path_to_key);
1941 retval = journal_end(th, th->t_super,
1942 th->t_blocks_allocated);
1945 goto out_inserted_sd;
1949 reiserfs_update_sd(th, inode);
1950 reiserfs_check_path(&path_to_key);
1954 /* it looks like you can easily compress these two goto targets into
1955 * one. Keeping it like this doesn't actually hurt anything, and they
1956 * are place holders for what the quota code actually needs.
1959 /* Invalidate the object, nothing was inserted yet */
1960 INODE_PKEY(inode)->k_objectid = 0;
1962 /* Quota change must be inside a transaction for journaling */
1963 vfs_dq_free_inode(inode);
1966 journal_end(th, th->t_super, th->t_blocks_allocated);
1967 /* Drop can be outside and it needs more credits so it's better to have it outside */
1969 inode->i_flags |= S_NOQUOTA;
1970 make_bad_inode(inode);
1974 th->t_trans_id = 0; /* so the caller can't use this handle later */
1975 unlock_new_inode(inode); /* OK to do even if we hadn't locked it */
1981 ** finds the tail page in the page cache,
1982 ** reads the last block in.
1984 ** On success, page_result is set to a locked, pinned page, and bh_result
1985 ** is set to an up to date buffer for the last block in the file. returns 0.
1987 ** tail conversion is not done, so bh_result might not be valid for writing
1988 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1989 ** trying to write the block.
1991 ** on failure, nonzero is returned, page_result and bh_result are untouched.
1993 static int grab_tail_page(struct inode *inode,
1994 struct page **page_result,
1995 struct buffer_head **bh_result)
1998 /* we want the page with the last byte in the file,
1999 ** not the page that will hold the next byte for appending
2001 unsigned long index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
2002 unsigned long pos = 0;
2003 unsigned long start = 0;
2004 unsigned long blocksize = inode->i_sb->s_blocksize;
2005 unsigned long offset = (inode->i_size) & (PAGE_CACHE_SIZE - 1);
2006 struct buffer_head *bh;
2007 struct buffer_head *head;
2011 /* we know that we are only called with inode->i_size > 0.
2012 ** we also know that a file tail can never be as big as a block
2013 ** If i_size % blocksize == 0, our file is currently block aligned
2014 ** and it won't need converting or zeroing after a truncate.
2016 if ((offset & (blocksize - 1)) == 0) {
2019 page = grab_cache_page(inode->i_mapping, index);
2024 /* start within the page of the last block in the file */
2025 start = (offset / blocksize) * blocksize;
2027 error = block_prepare_write(page, start, offset,
2028 reiserfs_get_block_create_0);
2032 head = page_buffers(page);
2038 bh = bh->b_this_page;
2040 } while (bh != head);
2042 if (!buffer_uptodate(bh)) {
2043 /* note, this should never happen, prepare_write should
2044 ** be taking care of this for us. If the buffer isn't up to date,
2045 ** I've screwed up the code to find the buffer, or the code to
2046 ** call prepare_write
2048 reiserfs_error(inode->i_sb, "clm-6000",
2049 "error reading block %lu", bh->b_blocknr);
2054 *page_result = page;
2061 page_cache_release(page);
2066 ** vfs version of truncate file. Must NOT be called with
2067 ** a transaction already started.
2069 ** some code taken from block_truncate_page
2071 int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
2073 struct reiserfs_transaction_handle th;
2074 /* we want the offset for the first byte after the end of the file */
2075 unsigned long offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2076 unsigned blocksize = inode->i_sb->s_blocksize;
2078 struct page *page = NULL;
2080 struct buffer_head *bh = NULL;
2084 lock_depth = reiserfs_write_lock_once(inode->i_sb);
2086 if (inode->i_size > 0) {
2087 error = grab_tail_page(inode, &page, &bh);
2089 // -ENOENT means we truncated past the end of the file,
2090 // and get_block_create_0 could not find a block to read in,
2092 if (error != -ENOENT)
2093 reiserfs_error(inode->i_sb, "clm-6001",
2094 "grab_tail_page failed %d",
2101 /* so, if page != NULL, we have a buffer head for the offset at
2102 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2103 ** then we have an unformatted node. Otherwise, we have a direct item,
2104 ** and no zeroing is required on disk. We zero after the truncate,
2105 ** because the truncate might pack the item anyway
2106 ** (it will unmap bh if it packs).
2108 /* it is enough to reserve space in transaction for 2 balancings:
2109 one for "save" link adding and another for the first
2110 cut_from_item. 1 is for update_sd */
2111 error = journal_begin(&th, inode->i_sb,
2112 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2115 reiserfs_update_inode_transaction(inode);
2116 if (update_timestamps)
2117 /* we are doing real truncate: if the system crashes before the last
2118 transaction of truncating gets committed - on reboot the file
2119 either appears truncated properly or not truncated at all */
2120 add_save_link(&th, inode, 1);
2121 err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps);
2123 journal_end(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2127 /* check reiserfs_do_truncate after ending the transaction */
2133 if (update_timestamps) {
2134 error = remove_save_link(inode, 1 /* truncate */);
2140 length = offset & (blocksize - 1);
2141 /* if we are not on a block boundary */
2143 length = blocksize - length;
2144 zero_user(page, offset, length);
2145 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2146 mark_buffer_dirty(bh);
2150 page_cache_release(page);
2153 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2159 page_cache_release(page);
2162 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2167 static int map_block_for_writepage(struct inode *inode,
2168 struct buffer_head *bh_result,
2169 unsigned long block)
2171 struct reiserfs_transaction_handle th;
2173 struct item_head tmp_ih;
2174 struct item_head *ih;
2175 struct buffer_head *bh;
2178 INITIALIZE_PATH(path);
2180 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
2181 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
2183 int use_get_block = 0;
2184 int bytes_copied = 0;
2186 int trans_running = 0;
2188 /* catch places below that try to log something without starting a trans */
2191 if (!buffer_uptodate(bh_result)) {
2195 kmap(bh_result->b_page);
2197 reiserfs_write_lock(inode->i_sb);
2198 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
2201 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2202 if (retval != POSITION_FOUND) {
2207 bh = get_last_bh(&path);
2209 item = get_item(&path);
2210 pos_in_item = path.pos_in_item;
2212 /* we've found an unformatted node */
2213 if (indirect_item_found(retval, ih)) {
2214 if (bytes_copied > 0) {
2215 reiserfs_warning(inode->i_sb, "clm-6002",
2216 "bytes_copied %d", bytes_copied);
2218 if (!get_block_num(item, pos_in_item)) {
2219 /* crap, we are writing to a hole */
2223 set_block_dev_mapped(bh_result,
2224 get_block_num(item, pos_in_item), inode);
2225 } else if (is_direct_le_ih(ih)) {
2227 p = page_address(bh_result->b_page);
2228 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2229 copy_size = ih_item_len(ih) - pos_in_item;
2231 fs_gen = get_generation(inode->i_sb);
2232 copy_item_head(&tmp_ih, ih);
2234 if (!trans_running) {
2235 /* vs-3050 is gone, no need to drop the path */
2236 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2239 reiserfs_update_inode_transaction(inode);
2241 if (fs_changed(fs_gen, inode->i_sb)
2242 && item_moved(&tmp_ih, &path)) {
2243 reiserfs_restore_prepared_buffer(inode->i_sb,
2249 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2251 if (fs_changed(fs_gen, inode->i_sb)
2252 && item_moved(&tmp_ih, &path)) {
2253 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2257 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2260 journal_mark_dirty(&th, inode->i_sb, bh);
2261 bytes_copied += copy_size;
2262 set_block_dev_mapped(bh_result, 0, inode);
2264 /* are there still bytes left? */
2265 if (bytes_copied < bh_result->b_size &&
2266 (byte_offset + bytes_copied) < inode->i_size) {
2267 set_cpu_key_k_offset(&key,
2268 cpu_key_k_offset(&key) +
2273 reiserfs_warning(inode->i_sb, "clm-6003",
2274 "bad item inode %lu", inode->i_ino);
2282 if (trans_running) {
2283 int err = journal_end(&th, inode->i_sb, jbegin_count);
2288 reiserfs_write_unlock(inode->i_sb);
2290 /* this is where we fill in holes in the file. */
2291 if (use_get_block) {
2292 retval = reiserfs_get_block(inode, block, bh_result,
2293 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2294 | GET_BLOCK_NO_DANGLE);
2296 if (!buffer_mapped(bh_result)
2297 || bh_result->b_blocknr == 0) {
2298 /* get_block failed to find a mapped unformatted node. */
2304 kunmap(bh_result->b_page);
2306 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2307 /* we've copied data from the page into the direct item, so the
2308 * buffer in the page is now clean, mark it to reflect that.
2310 lock_buffer(bh_result);
2311 clear_buffer_dirty(bh_result);
2312 unlock_buffer(bh_result);
2318 * mason@suse.com: updated in 2.5.54 to follow the same general io
2319 * start/recovery path as __block_write_full_page, along with special
2320 * code to handle reiserfs tails.
2322 static int reiserfs_write_full_page(struct page *page,
2323 struct writeback_control *wbc)
2325 struct inode *inode = page->mapping->host;
2326 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2328 unsigned long block;
2329 sector_t last_block;
2330 struct buffer_head *head, *bh;
2333 int checked = PageChecked(page);
2334 struct reiserfs_transaction_handle th;
2335 struct super_block *s = inode->i_sb;
2336 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2339 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2340 if (checked && (current->flags & PF_MEMALLOC)) {
2341 redirty_page_for_writepage(wbc, page);
2346 /* The page dirty bit is cleared before writepage is called, which
2347 * means we have to tell create_empty_buffers to make dirty buffers
2348 * The page really should be up to date at this point, so tossing
2349 * in the BH_Uptodate is just a sanity check.
2351 if (!page_has_buffers(page)) {
2352 create_empty_buffers(page, s->s_blocksize,
2353 (1 << BH_Dirty) | (1 << BH_Uptodate));
2355 head = page_buffers(page);
2357 /* last page in the file, zero out any contents past the
2358 ** last byte in the file
2360 if (page->index >= end_index) {
2361 unsigned last_offset;
2363 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2364 /* no file contents in this page */
2365 if (page->index >= end_index + 1 || !last_offset) {
2369 zero_user_segment(page, last_offset, PAGE_CACHE_SIZE);
2372 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
2373 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
2374 /* first map all the buffers, logging any direct items we find */
2376 if (block > last_block) {
2378 * This can happen when the block size is less than
2379 * the page size. The corresponding bytes in the page
2380 * were zero filled above
2382 clear_buffer_dirty(bh);
2383 set_buffer_uptodate(bh);
2384 } else if ((checked || buffer_dirty(bh)) &&
2385 (!buffer_mapped(bh) || (buffer_mapped(bh)
2388 /* not mapped yet, or it points to a direct item, search
2389 * the btree for the mapping info, and log any direct
2392 if ((error = map_block_for_writepage(inode, bh, block))) {
2396 bh = bh->b_this_page;
2398 } while (bh != head);
2401 * we start the transaction after map_block_for_writepage,
2402 * because it can create holes in the file (an unbounded operation).
2403 * starting it here, we can make a reliable estimate for how many
2404 * blocks we're going to log
2407 ClearPageChecked(page);
2408 reiserfs_write_lock(s);
2409 error = journal_begin(&th, s, bh_per_page + 1);
2411 reiserfs_write_unlock(s);
2414 reiserfs_update_inode_transaction(inode);
2416 /* now go through and lock any dirty buffers on the page */
2419 if (!buffer_mapped(bh))
2421 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2425 reiserfs_prepare_for_journal(s, bh, 1);
2426 journal_mark_dirty(&th, s, bh);
2429 /* from this point on, we know the buffer is mapped to a
2430 * real block and not a direct item
2432 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2435 if (!trylock_buffer(bh)) {
2436 redirty_page_for_writepage(wbc, page);
2440 if (test_clear_buffer_dirty(bh)) {
2441 mark_buffer_async_write(bh);
2445 } while ((bh = bh->b_this_page) != head);
2448 error = journal_end(&th, s, bh_per_page + 1);
2449 reiserfs_write_unlock(s);
2453 BUG_ON(PageWriteback(page));
2454 set_page_writeback(page);
2458 * since any buffer might be the only dirty buffer on the page,
2459 * the first submit_bh can bring the page out of writeback.
2460 * be careful with the buffers.
2463 struct buffer_head *next = bh->b_this_page;
2464 if (buffer_async_write(bh)) {
2465 submit_bh(WRITE, bh);
2470 } while (bh != head);
2476 * if this page only had a direct item, it is very possible for
2477 * no io to be required without there being an error. Or,
2478 * someone else could have locked them and sent them down the
2479 * pipe without locking the page
2483 if (!buffer_uptodate(bh)) {
2487 bh = bh->b_this_page;
2488 } while (bh != head);
2490 SetPageUptodate(page);
2491 end_page_writeback(page);
2496 /* catches various errors, we need to make sure any valid dirty blocks
2497 * get to the media. The page is currently locked and not marked for
2500 ClearPageUptodate(page);
2504 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2506 mark_buffer_async_write(bh);
2509 * clear any dirty bits that might have come from getting
2510 * attached to a dirty page
2512 clear_buffer_dirty(bh);
2514 bh = bh->b_this_page;
2515 } while (bh != head);
2517 BUG_ON(PageWriteback(page));
2518 set_page_writeback(page);
2521 struct buffer_head *next = bh->b_this_page;
2522 if (buffer_async_write(bh)) {
2523 clear_buffer_dirty(bh);
2524 submit_bh(WRITE, bh);
2529 } while (bh != head);
2533 static int reiserfs_readpage(struct file *f, struct page *page)
2535 return block_read_full_page(page, reiserfs_get_block);
2538 static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
2540 struct inode *inode = page->mapping->host;
2541 reiserfs_wait_on_write_block(inode->i_sb);
2542 return reiserfs_write_full_page(page, wbc);
2545 static int reiserfs_write_begin(struct file *file,
2546 struct address_space *mapping,
2547 loff_t pos, unsigned len, unsigned flags,
2548 struct page **pagep, void **fsdata)
2550 struct inode *inode;
2556 inode = mapping->host;
2558 if (flags & AOP_FLAG_CONT_EXPAND &&
2559 (pos & (inode->i_sb->s_blocksize - 1)) == 0) {
2561 *fsdata = (void *)(unsigned long)flags;
2564 index = pos >> PAGE_CACHE_SHIFT;
2565 page = grab_cache_page_write_begin(mapping, index, flags);
2570 reiserfs_wait_on_write_block(inode->i_sb);
2571 fix_tail_page_for_writing(page);
2572 if (reiserfs_transaction_running(inode->i_sb)) {
2573 struct reiserfs_transaction_handle *th;
2574 th = (struct reiserfs_transaction_handle *)current->
2576 BUG_ON(!th->t_refcount);
2577 BUG_ON(!th->t_trans_id);
2578 old_ref = th->t_refcount;
2581 ret = block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
2582 reiserfs_get_block);
2583 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2584 struct reiserfs_transaction_handle *th = current->journal_info;
2585 /* this gets a little ugly. If reiserfs_get_block returned an
2586 * error and left a transacstion running, we've got to close it,
2587 * and we've got to free handle if it was a persistent transaction.
2589 * But, if we had nested into an existing transaction, we need
2590 * to just drop the ref count on the handle.
2592 * If old_ref == 0, the transaction is from reiserfs_get_block,
2593 * and it was a persistent trans. Otherwise, it was nested above.
2595 if (th->t_refcount > old_ref) {
2600 reiserfs_write_lock(inode->i_sb);
2601 err = reiserfs_end_persistent_transaction(th);
2602 reiserfs_write_unlock(inode->i_sb);
2610 page_cache_release(page);
2615 int reiserfs_prepare_write(struct file *f, struct page *page,
2616 unsigned from, unsigned to)
2618 struct inode *inode = page->mapping->host;
2622 reiserfs_write_unlock(inode->i_sb);
2623 reiserfs_wait_on_write_block(inode->i_sb);
2624 reiserfs_write_lock(inode->i_sb);
2626 fix_tail_page_for_writing(page);
2627 if (reiserfs_transaction_running(inode->i_sb)) {
2628 struct reiserfs_transaction_handle *th;
2629 th = (struct reiserfs_transaction_handle *)current->
2631 BUG_ON(!th->t_refcount);
2632 BUG_ON(!th->t_trans_id);
2633 old_ref = th->t_refcount;
2637 ret = block_prepare_write(page, from, to, reiserfs_get_block);
2638 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2639 struct reiserfs_transaction_handle *th = current->journal_info;
2640 /* this gets a little ugly. If reiserfs_get_block returned an
2641 * error and left a transacstion running, we've got to close it,
2642 * and we've got to free handle if it was a persistent transaction.
2644 * But, if we had nested into an existing transaction, we need
2645 * to just drop the ref count on the handle.
2647 * If old_ref == 0, the transaction is from reiserfs_get_block,
2648 * and it was a persistent trans. Otherwise, it was nested above.
2650 if (th->t_refcount > old_ref) {
2655 reiserfs_write_lock(inode->i_sb);
2656 err = reiserfs_end_persistent_transaction(th);
2657 reiserfs_write_unlock(inode->i_sb);
2667 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2669 return generic_block_bmap(as, block, reiserfs_bmap);
2672 static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2673 loff_t pos, unsigned len, unsigned copied,
2674 struct page *page, void *fsdata)
2676 struct inode *inode = page->mapping->host;
2679 struct reiserfs_transaction_handle *th;
2682 if ((unsigned long)fsdata & AOP_FLAG_CONT_EXPAND)
2685 reiserfs_wait_on_write_block(inode->i_sb);
2686 if (reiserfs_transaction_running(inode->i_sb))
2687 th = current->journal_info;
2691 start = pos & (PAGE_CACHE_SIZE - 1);
2692 if (unlikely(copied < len)) {
2693 if (!PageUptodate(page))
2696 page_zero_new_buffers(page, start + copied, start + len);
2698 flush_dcache_page(page);
2700 reiserfs_commit_page(inode, page, start, start + copied);
2702 /* generic_commit_write does this for us, but does not update the
2703 ** transaction tracking stuff when the size changes. So, we have
2704 ** to do the i_size updates here.
2707 if (pos > inode->i_size) {
2708 struct reiserfs_transaction_handle myth;
2709 reiserfs_write_lock(inode->i_sb);
2710 /* If the file have grown beyond the border where it
2711 can have a tail, unmark it as needing a tail
2713 if ((have_large_tails(inode->i_sb)
2714 && inode->i_size > i_block_size(inode) * 4)
2715 || (have_small_tails(inode->i_sb)
2716 && inode->i_size > i_block_size(inode)))
2717 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2719 ret = journal_begin(&myth, inode->i_sb, 1);
2721 reiserfs_write_unlock(inode->i_sb);
2724 reiserfs_update_inode_transaction(inode);
2725 inode->i_size = pos;
2727 * this will just nest into our transaction. It's important
2728 * to use mark_inode_dirty so the inode gets pushed around on the
2729 * dirty lists, and so that O_SYNC works as expected
2731 mark_inode_dirty(inode);
2732 reiserfs_update_sd(&myth, inode);
2734 ret = journal_end(&myth, inode->i_sb, 1);
2735 reiserfs_write_unlock(inode->i_sb);
2740 reiserfs_write_lock(inode->i_sb);
2742 mark_inode_dirty(inode);
2743 ret = reiserfs_end_persistent_transaction(th);
2744 reiserfs_write_unlock(inode->i_sb);
2751 page_cache_release(page);
2752 return ret == 0 ? copied : ret;
2756 reiserfs_write_lock(inode->i_sb);
2758 reiserfs_update_sd(th, inode);
2759 ret = reiserfs_end_persistent_transaction(th);
2760 reiserfs_write_unlock(inode->i_sb);
2766 int reiserfs_commit_write(struct file *f, struct page *page,
2767 unsigned from, unsigned to)
2769 struct inode *inode = page->mapping->host;
2770 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2773 struct reiserfs_transaction_handle *th = NULL;
2775 reiserfs_write_unlock(inode->i_sb);
2776 reiserfs_wait_on_write_block(inode->i_sb);
2777 reiserfs_write_lock(inode->i_sb);
2779 if (reiserfs_transaction_running(inode->i_sb)) {
2780 th = current->journal_info;
2782 reiserfs_commit_page(inode, page, from, to);
2784 /* generic_commit_write does this for us, but does not update the
2785 ** transaction tracking stuff when the size changes. So, we have
2786 ** to do the i_size updates here.
2788 if (pos > inode->i_size) {
2789 struct reiserfs_transaction_handle myth;
2790 reiserfs_write_lock(inode->i_sb);
2791 /* If the file have grown beyond the border where it
2792 can have a tail, unmark it as needing a tail
2794 if ((have_large_tails(inode->i_sb)
2795 && inode->i_size > i_block_size(inode) * 4)
2796 || (have_small_tails(inode->i_sb)
2797 && inode->i_size > i_block_size(inode)))
2798 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2800 ret = journal_begin(&myth, inode->i_sb, 1);
2802 reiserfs_write_unlock(inode->i_sb);
2805 reiserfs_update_inode_transaction(inode);
2806 inode->i_size = pos;
2808 * this will just nest into our transaction. It's important
2809 * to use mark_inode_dirty so the inode gets pushed around on the
2810 * dirty lists, and so that O_SYNC works as expected
2812 mark_inode_dirty(inode);
2813 reiserfs_update_sd(&myth, inode);
2815 ret = journal_end(&myth, inode->i_sb, 1);
2816 reiserfs_write_unlock(inode->i_sb);
2821 reiserfs_write_lock(inode->i_sb);
2823 mark_inode_dirty(inode);
2824 ret = reiserfs_end_persistent_transaction(th);
2825 reiserfs_write_unlock(inode->i_sb);
2835 reiserfs_write_lock(inode->i_sb);
2837 reiserfs_update_sd(th, inode);
2838 ret = reiserfs_end_persistent_transaction(th);
2839 reiserfs_write_unlock(inode->i_sb);
2845 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2847 if (reiserfs_attrs(inode->i_sb)) {
2848 if (sd_attrs & REISERFS_SYNC_FL)
2849 inode->i_flags |= S_SYNC;
2851 inode->i_flags &= ~S_SYNC;
2852 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2853 inode->i_flags |= S_IMMUTABLE;
2855 inode->i_flags &= ~S_IMMUTABLE;
2856 if (sd_attrs & REISERFS_APPEND_FL)
2857 inode->i_flags |= S_APPEND;
2859 inode->i_flags &= ~S_APPEND;
2860 if (sd_attrs & REISERFS_NOATIME_FL)
2861 inode->i_flags |= S_NOATIME;
2863 inode->i_flags &= ~S_NOATIME;
2864 if (sd_attrs & REISERFS_NOTAIL_FL)
2865 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2867 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2871 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
2873 if (reiserfs_attrs(inode->i_sb)) {
2874 if (inode->i_flags & S_IMMUTABLE)
2875 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2877 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2878 if (inode->i_flags & S_SYNC)
2879 *sd_attrs |= REISERFS_SYNC_FL;
2881 *sd_attrs &= ~REISERFS_SYNC_FL;
2882 if (inode->i_flags & S_NOATIME)
2883 *sd_attrs |= REISERFS_NOATIME_FL;
2885 *sd_attrs &= ~REISERFS_NOATIME_FL;
2886 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
2887 *sd_attrs |= REISERFS_NOTAIL_FL;
2889 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2893 /* decide if this buffer needs to stay around for data logging or ordered
2896 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2899 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2902 spin_lock(&j->j_dirty_buffers_lock);
2903 if (!buffer_mapped(bh)) {
2906 /* the page is locked, and the only places that log a data buffer
2907 * also lock the page.
2909 if (reiserfs_file_data_log(inode)) {
2911 * very conservative, leave the buffer pinned if
2912 * anyone might need it.
2914 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2917 } else if (buffer_dirty(bh)) {
2918 struct reiserfs_journal_list *jl;
2919 struct reiserfs_jh *jh = bh->b_private;
2921 /* why is this safe?
2922 * reiserfs_setattr updates i_size in the on disk
2923 * stat data before allowing vmtruncate to be called.
2925 * If buffer was put onto the ordered list for this
2926 * transaction, we know for sure either this transaction
2927 * or an older one already has updated i_size on disk,
2928 * and this ordered data won't be referenced in the file
2931 * if the buffer was put onto the ordered list for an older
2932 * transaction, we need to leave it around
2934 if (jh && (jl = jh->jl)
2935 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2939 if (ret && bh->b_private) {
2940 reiserfs_free_jh(bh);
2942 spin_unlock(&j->j_dirty_buffers_lock);
2947 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2948 static void reiserfs_invalidatepage(struct page *page, unsigned long offset)
2950 struct buffer_head *head, *bh, *next;
2951 struct inode *inode = page->mapping->host;
2952 unsigned int curr_off = 0;
2955 BUG_ON(!PageLocked(page));
2958 ClearPageChecked(page);
2960 if (!page_has_buffers(page))
2963 head = page_buffers(page);
2966 unsigned int next_off = curr_off + bh->b_size;
2967 next = bh->b_this_page;
2970 * is this block fully invalidated?
2972 if (offset <= curr_off) {
2973 if (invalidatepage_can_drop(inode, bh))
2974 reiserfs_unmap_buffer(bh);
2978 curr_off = next_off;
2980 } while (bh != head);
2983 * We release buffers only if the entire page is being invalidated.
2984 * The get_block cached value has been unconditionally invalidated,
2985 * so real IO is not possible anymore.
2987 if (!offset && ret) {
2988 ret = try_to_release_page(page, 0);
2989 /* maybe should BUG_ON(!ret); - neilb */
2995 static int reiserfs_set_page_dirty(struct page *page)
2997 struct inode *inode = page->mapping->host;
2998 if (reiserfs_file_data_log(inode)) {
2999 SetPageChecked(page);
3000 return __set_page_dirty_nobuffers(page);
3002 return __set_page_dirty_buffers(page);
3006 * Returns 1 if the page's buffers were dropped. The page is locked.
3008 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3009 * in the buffers at page_buffers(page).
3011 * even in -o notail mode, we can't be sure an old mount without -o notail
3012 * didn't create files with tails.
3014 static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
3016 struct inode *inode = page->mapping->host;
3017 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
3018 struct buffer_head *head;
3019 struct buffer_head *bh;
3022 WARN_ON(PageChecked(page));
3023 spin_lock(&j->j_dirty_buffers_lock);
3024 head = page_buffers(page);
3027 if (bh->b_private) {
3028 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
3029 reiserfs_free_jh(bh);
3035 bh = bh->b_this_page;
3036 } while (bh != head);
3038 ret = try_to_free_buffers(page);
3039 spin_unlock(&j->j_dirty_buffers_lock);
3043 /* We thank Mingming Cao for helping us understand in great detail what
3044 to do in this section of the code. */
3045 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
3046 const struct iovec *iov, loff_t offset,
3047 unsigned long nr_segs)
3049 struct file *file = iocb->ki_filp;
3050 struct inode *inode = file->f_mapping->host;
3052 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
3054 reiserfs_get_blocks_direct_io, NULL);
3057 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3059 struct inode *inode = dentry->d_inode;
3061 unsigned int ia_valid;
3063 /* must be turned off for recursive notify_change calls */
3064 ia_valid = attr->ia_valid &= ~(ATTR_KILL_SUID|ATTR_KILL_SGID);
3066 reiserfs_write_lock(inode->i_sb);
3067 if (attr->ia_valid & ATTR_SIZE) {
3068 /* version 2 items will be caught by the s_maxbytes check
3069 ** done for us in vmtruncate
3071 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
3072 attr->ia_size > MAX_NON_LFS) {
3076 /* fill in hole pointers in the expanding truncate case. */
3077 if (attr->ia_size > inode->i_size) {
3078 error = generic_cont_expand_simple(inode, attr->ia_size);
3079 if (REISERFS_I(inode)->i_prealloc_count > 0) {
3081 struct reiserfs_transaction_handle th;
3082 /* we're changing at most 2 bitmaps, inode + super */
3083 err = journal_begin(&th, inode->i_sb, 4);
3085 reiserfs_discard_prealloc(&th, inode);
3086 err = journal_end(&th, inode->i_sb, 4);
3094 * file size is changed, ctime and mtime are
3097 attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
3101 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
3102 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
3103 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
3104 /* stat data of format v3.5 has 16 bit uid and gid */
3109 error = inode_change_ok(inode, attr);
3111 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
3112 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
3113 error = reiserfs_chown_xattrs(inode, attr);
3116 struct reiserfs_transaction_handle th;
3119 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3120 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3123 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3125 journal_begin(&th, inode->i_sb,
3130 vfs_dq_transfer(inode, attr) ? -EDQUOT : 0;
3132 journal_end(&th, inode->i_sb,
3136 /* Update corresponding info in inode so that everything is in
3137 * one transaction */
3138 if (attr->ia_valid & ATTR_UID)
3139 inode->i_uid = attr->ia_uid;
3140 if (attr->ia_valid & ATTR_GID)
3141 inode->i_gid = attr->ia_gid;
3142 mark_inode_dirty(inode);
3144 journal_end(&th, inode->i_sb, jbegin_count);
3148 error = inode_setattr(inode, attr);
3151 if (!error && reiserfs_posixacl(inode->i_sb)) {
3152 if (attr->ia_valid & ATTR_MODE)
3153 error = reiserfs_acl_chmod(inode);
3157 reiserfs_write_unlock(inode->i_sb);
3161 const struct address_space_operations reiserfs_address_space_operations = {
3162 .writepage = reiserfs_writepage,
3163 .readpage = reiserfs_readpage,
3164 .readpages = reiserfs_readpages,
3165 .releasepage = reiserfs_releasepage,
3166 .invalidatepage = reiserfs_invalidatepage,
3167 .sync_page = block_sync_page,
3168 .write_begin = reiserfs_write_begin,
3169 .write_end = reiserfs_write_end,
3170 .bmap = reiserfs_aop_bmap,
3171 .direct_IO = reiserfs_direct_IO,
3172 .set_page_dirty = reiserfs_set_page_dirty,