2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
5 #include <linux/time.h>
10 #include <linux/exportfs.h>
11 #include <linux/pagemap.h>
12 #include <linux/highmem.h>
13 #include <linux/slab.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>
21 #include <linux/aio.h>
23 int reiserfs_commit_write(struct file *f, struct page *page,
24 unsigned from, unsigned to);
26 void reiserfs_evict_inode(struct inode *inode)
28 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
30 JOURNAL_PER_BALANCE_CNT * 2 +
31 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
32 struct reiserfs_transaction_handle th;
35 if (!inode->i_nlink && !is_bad_inode(inode))
36 dquot_initialize(inode);
38 truncate_inode_pages(&inode->i_data, 0);
42 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
43 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
45 reiserfs_delete_xattrs(inode);
47 reiserfs_write_lock(inode->i_sb);
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 reiserfs_write_unlock(inode->i_sb);
78 /* no object items are in the tree */
81 clear_inode(inode); /* note this must go after the journal_end to prevent deadlock */
91 static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
92 __u32 objectid, loff_t offset, int type, int length)
94 key->version = version;
96 key->on_disk_key.k_dir_id = dirid;
97 key->on_disk_key.k_objectid = objectid;
98 set_cpu_key_k_offset(key, offset);
99 set_cpu_key_k_type(key, type);
100 key->key_length = length;
103 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
104 offset and type of key */
105 void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
106 int type, int length)
108 _make_cpu_key(key, get_inode_item_key_version(inode),
109 le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
110 le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
115 // when key is 0, do not set version and short key
117 inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
119 loff_t offset, int type, int length,
120 int entry_count /*or ih_free_space */ )
123 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
124 ih->ih_key.k_objectid =
125 cpu_to_le32(key->on_disk_key.k_objectid);
127 put_ih_version(ih, version);
128 set_le_ih_k_offset(ih, offset);
129 set_le_ih_k_type(ih, type);
130 put_ih_item_len(ih, length);
131 /* set_ih_free_space (ih, 0); */
132 // for directory items it is entry count, for directs and stat
133 // datas - 0xffff, for indirects - 0
134 put_ih_entry_count(ih, entry_count);
138 // FIXME: we might cache recently accessed indirect item
140 // Ugh. Not too eager for that....
141 // I cut the code until such time as I see a convincing argument (benchmark).
142 // I don't want a bloated inode struct..., and I don't like code complexity....
144 /* cutting the code is fine, since it really isn't in use yet and is easy
145 ** to add back in. But, Vladimir has a really good idea here. Think
146 ** about what happens for reading a file. For each page,
147 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
148 ** an indirect item. This indirect item has X number of pointers, where
149 ** X is a big number if we've done the block allocation right. But,
150 ** we only use one or two of these pointers during each call to readpage,
151 ** needlessly researching again later on.
153 ** The size of the cache could be dynamic based on the size of the file.
155 ** I'd also like to see us cache the location the stat data item, since
156 ** we are needlessly researching for that frequently.
161 /* If this page has a file tail in it, and
162 ** it was read in by get_block_create_0, the page data is valid,
163 ** but tail is still sitting in a direct item, and we can't write to
164 ** it. So, look through this page, and check all the mapped buffers
165 ** to make sure they have valid block numbers. Any that don't need
166 ** to be unmapped, so that __block_write_begin will correctly call
167 ** reiserfs_get_block to convert the tail into an unformatted node
169 static inline void fix_tail_page_for_writing(struct page *page)
171 struct buffer_head *head, *next, *bh;
173 if (page && page_has_buffers(page)) {
174 head = page_buffers(page);
177 next = bh->b_this_page;
178 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
179 reiserfs_unmap_buffer(bh);
182 } while (bh != head);
186 /* reiserfs_get_block does not need to allocate a block only if it has been
187 done already or non-hole position has been found in the indirect item */
188 static inline int allocation_needed(int retval, b_blocknr_t allocated,
189 struct item_head *ih,
190 __le32 * item, int pos_in_item)
194 if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
195 get_block_num(item, pos_in_item))
200 static inline int indirect_item_found(int retval, struct item_head *ih)
202 return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
205 static inline void set_block_dev_mapped(struct buffer_head *bh,
206 b_blocknr_t block, struct inode *inode)
208 map_bh(bh, inode->i_sb, block);
212 // files which were created in the earlier version can not be longer,
215 static int file_capable(struct inode *inode, sector_t block)
217 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file.
218 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
224 static int restart_transaction(struct reiserfs_transaction_handle *th,
225 struct inode *inode, struct treepath *path)
227 struct super_block *s = th->t_super;
228 int len = th->t_blocks_allocated;
231 BUG_ON(!th->t_trans_id);
232 BUG_ON(!th->t_refcount);
236 /* we cannot restart while nested */
237 if (th->t_refcount > 1) {
240 reiserfs_update_sd(th, inode);
241 err = journal_end(th, s, len);
243 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
245 reiserfs_update_inode_transaction(inode);
250 // it is called by get_block when create == 0. Returns block number
251 // for 'block'-th logical block of file. When it hits direct item it
252 // returns 0 (being called from bmap) or read direct item into piece
253 // of page (bh_result)
255 // Please improve the english/clarity in the comment above, as it is
256 // hard to understand.
258 static int _get_block_create_0(struct inode *inode, sector_t block,
259 struct buffer_head *bh_result, int args)
261 INITIALIZE_PATH(path);
263 struct buffer_head *bh;
264 struct item_head *ih, tmp_ih;
271 unsigned long offset;
273 // prepare the key to look for the 'block'-th block of file
274 make_cpu_key(&key, inode,
275 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
278 result = search_for_position_by_key(inode->i_sb, &key, &path);
279 if (result != POSITION_FOUND) {
282 kunmap(bh_result->b_page);
283 if (result == IO_ERROR)
285 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
286 // That there is some MMAPED data associated with it that is yet to be written to disk.
287 if ((args & GET_BLOCK_NO_HOLE)
288 && !PageUptodate(bh_result->b_page)) {
294 bh = get_last_bh(&path);
296 if (is_indirect_le_ih(ih)) {
297 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih);
299 /* FIXME: here we could cache indirect item or part of it in
300 the inode to avoid search_by_key in case of subsequent
302 blocknr = get_block_num(ind_item, path.pos_in_item);
305 map_bh(bh_result, inode->i_sb, blocknr);
306 if (path.pos_in_item ==
307 ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
308 set_buffer_boundary(bh_result);
311 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
312 // That there is some MMAPED data associated with it that is yet to be written to disk.
313 if ((args & GET_BLOCK_NO_HOLE)
314 && !PageUptodate(bh_result->b_page)) {
320 kunmap(bh_result->b_page);
323 // requested data are in direct item(s)
324 if (!(args & GET_BLOCK_READ_DIRECT)) {
325 // we are called by bmap. FIXME: we can not map block of file
326 // when it is stored in direct item(s)
329 kunmap(bh_result->b_page);
333 /* if we've got a direct item, and the buffer or page was uptodate,
334 ** we don't want to pull data off disk again. skip to the
335 ** end, where we map the buffer and return
337 if (buffer_uptodate(bh_result)) {
341 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
342 ** pages without any buffers. If the page is up to date, we don't want
343 ** read old data off disk. Set the up to date bit on the buffer instead
344 ** and jump to the end
346 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
347 set_buffer_uptodate(bh_result);
350 // read file tail into part of page
351 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
352 copy_item_head(&tmp_ih, ih);
354 /* we only want to kmap if we are reading the tail into the page.
355 ** this is not the common case, so we don't kmap until we are
356 ** sure we need to. But, this means the item might move if
360 p = (char *)kmap(bh_result->b_page);
363 memset(p, 0, inode->i_sb->s_blocksize);
365 if (!is_direct_le_ih(ih)) {
368 /* make sure we don't read more bytes than actually exist in
369 ** the file. This can happen in odd cases where i_size isn't
370 ** correct, and when direct item padding results in a few
371 ** extra bytes at the end of the direct item
373 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
375 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
377 inode->i_size - (le_ih_k_offset(ih) - 1) -
381 chars = ih_item_len(ih) - path.pos_in_item;
383 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
390 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
391 // we done, if read direct item is not the last item of
392 // node FIXME: we could try to check right delimiting key
393 // to see whether direct item continues in the right
394 // neighbor or rely on i_size
397 // update key to look for the next piece
398 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
399 result = search_for_position_by_key(inode->i_sb, &key, &path);
400 if (result != POSITION_FOUND)
401 // i/o error most likely
403 bh = get_last_bh(&path);
407 flush_dcache_page(bh_result->b_page);
408 kunmap(bh_result->b_page);
413 if (result == IO_ERROR)
416 /* this buffer has valid data, but isn't valid for io. mapping it to
417 * block #0 tells the rest of reiserfs it just has a tail in it
419 map_bh(bh_result, inode->i_sb, 0);
420 set_buffer_uptodate(bh_result);
424 // this is called to create file map. So, _get_block_create_0 will not
426 static int reiserfs_bmap(struct inode *inode, sector_t block,
427 struct buffer_head *bh_result, int create)
429 if (!file_capable(inode, block))
432 reiserfs_write_lock(inode->i_sb);
433 /* do not read the direct item */
434 _get_block_create_0(inode, block, bh_result, 0);
435 reiserfs_write_unlock(inode->i_sb);
439 /* special version of get_block that is only used by grab_tail_page right
440 ** now. It is sent to __block_write_begin, and when you try to get a
441 ** block past the end of the file (or a block from a hole) it returns
442 ** -ENOENT instead of a valid buffer. __block_write_begin expects to
443 ** be able to do i/o on the buffers returned, unless an error value
446 ** So, this allows __block_write_begin to be used for reading a single block
447 ** in a page. Where it does not produce a valid page for holes, or past the
448 ** end of the file. This turns out to be exactly what we need for reading
449 ** tails for conversion.
451 ** The point of the wrapper is forcing a certain value for create, even
452 ** though the VFS layer is calling this function with create==1. If you
453 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
454 ** don't use this function.
456 static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
457 struct buffer_head *bh_result,
460 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
463 /* This is special helper for reiserfs_get_block in case we are executing
464 direct_IO request. */
465 static int reiserfs_get_blocks_direct_io(struct inode *inode,
467 struct buffer_head *bh_result,
472 bh_result->b_page = NULL;
474 /* We set the b_size before reiserfs_get_block call since it is
475 referenced in convert_tail_for_hole() that may be called from
476 reiserfs_get_block() */
477 bh_result->b_size = (1 << inode->i_blkbits);
479 ret = reiserfs_get_block(inode, iblock, bh_result,
480 create | GET_BLOCK_NO_DANGLE);
484 /* don't allow direct io onto tail pages */
485 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
486 /* make sure future calls to the direct io funcs for this offset
487 ** in the file fail by unmapping the buffer
489 clear_buffer_mapped(bh_result);
492 /* Possible unpacked tail. Flush the data before pages have
494 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
497 reiserfs_write_lock(inode->i_sb);
499 err = reiserfs_commit_for_inode(inode);
500 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
502 reiserfs_write_unlock(inode->i_sb);
512 ** helper function for when reiserfs_get_block is called for a hole
513 ** but the file tail is still in a direct item
514 ** bh_result is the buffer head for the hole
515 ** tail_offset is the offset of the start of the tail in the file
517 ** This calls prepare_write, which will start a new transaction
518 ** you should not be in a transaction, or have any paths held when you
521 static int convert_tail_for_hole(struct inode *inode,
522 struct buffer_head *bh_result,
526 unsigned long tail_end;
527 unsigned long tail_start;
528 struct page *tail_page;
529 struct page *hole_page = bh_result->b_page;
532 if ((tail_offset & (bh_result->b_size - 1)) != 1)
535 /* always try to read until the end of the block */
536 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
537 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
539 index = tail_offset >> PAGE_CACHE_SHIFT;
540 /* hole_page can be zero in case of direct_io, we are sure
541 that we cannot get here if we write with O_DIRECT into
543 if (!hole_page || index != hole_page->index) {
544 tail_page = grab_cache_page(inode->i_mapping, index);
550 tail_page = hole_page;
553 /* we don't have to make sure the conversion did not happen while
554 ** we were locking the page because anyone that could convert
555 ** must first take i_mutex.
557 ** We must fix the tail page for writing because it might have buffers
558 ** that are mapped, but have a block number of 0. This indicates tail
559 ** data that has been read directly into the page, and
560 ** __block_write_begin won't trigger a get_block in this case.
562 fix_tail_page_for_writing(tail_page);
563 retval = __reiserfs_write_begin(tail_page, tail_start,
564 tail_end - tail_start);
568 /* tail conversion might change the data in the page */
569 flush_dcache_page(tail_page);
571 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
574 if (tail_page != hole_page) {
575 unlock_page(tail_page);
576 page_cache_release(tail_page);
582 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
585 b_blocknr_t * allocated_block_nr,
586 struct treepath *path, int flags)
588 BUG_ON(!th->t_trans_id);
590 #ifdef REISERFS_PREALLOCATE
591 if (!(flags & GET_BLOCK_NO_IMUX)) {
592 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
596 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
600 int reiserfs_get_block(struct inode *inode, sector_t block,
601 struct buffer_head *bh_result, int create)
603 int repeat, retval = 0;
604 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
605 INITIALIZE_PATH(path);
608 struct buffer_head *bh, *unbh = NULL;
609 struct item_head *ih, tmp_ih;
613 struct reiserfs_transaction_handle *th = NULL;
614 /* space reserved in transaction batch:
615 . 3 balancings in direct->indirect conversion
616 . 1 block involved into reiserfs_update_sd()
617 XXX in practically impossible worst case direct2indirect()
618 can incur (much) more than 3 balancings.
619 quota update for user, group */
621 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
622 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
626 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
628 reiserfs_write_lock(inode->i_sb);
629 version = get_inode_item_key_version(inode);
631 if (!file_capable(inode, block)) {
632 reiserfs_write_unlock(inode->i_sb);
636 /* if !create, we aren't changing the FS, so we don't need to
637 ** log anything, so we don't need to start a transaction
639 if (!(create & GET_BLOCK_CREATE)) {
641 /* find number of block-th logical block of the file */
642 ret = _get_block_create_0(inode, block, bh_result,
643 create | GET_BLOCK_READ_DIRECT);
644 reiserfs_write_unlock(inode->i_sb);
648 * if we're already in a transaction, make sure to close
649 * any new transactions we start in this func
651 if ((create & GET_BLOCK_NO_DANGLE) ||
652 reiserfs_transaction_running(inode->i_sb))
655 /* If file is of such a size, that it might have a tail and tails are enabled
656 ** we should mark it as possibly needing tail packing on close
658 if ((have_large_tails(inode->i_sb)
659 && inode->i_size < i_block_size(inode) * 4)
660 || (have_small_tails(inode->i_sb)
661 && inode->i_size < i_block_size(inode)))
662 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
664 /* set the key of the first byte in the 'block'-th block of file */
665 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
666 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
668 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
673 reiserfs_update_inode_transaction(inode);
677 retval = search_for_position_by_key(inode->i_sb, &key, &path);
678 if (retval == IO_ERROR) {
683 bh = get_last_bh(&path);
685 item = get_item(&path);
686 pos_in_item = path.pos_in_item;
688 fs_gen = get_generation(inode->i_sb);
689 copy_item_head(&tmp_ih, ih);
691 if (allocation_needed
692 (retval, allocated_block_nr, ih, item, pos_in_item)) {
693 /* we have to allocate block for the unformatted node */
700 _allocate_block(th, block, inode, &allocated_block_nr,
703 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
704 /* restart the transaction to give the journal a chance to free
705 ** some blocks. releases the path, so we have to go back to
706 ** research if we succeed on the second try
708 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
709 retval = restart_transaction(th, inode, &path);
713 _allocate_block(th, block, inode,
714 &allocated_block_nr, NULL, create);
716 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
719 if (repeat == QUOTA_EXCEEDED)
726 if (fs_changed(fs_gen, inode->i_sb)
727 && item_moved(&tmp_ih, &path)) {
732 if (indirect_item_found(retval, ih)) {
733 b_blocknr_t unfm_ptr;
734 /* 'block'-th block is in the file already (there is
735 corresponding cell in some indirect item). But it may be
736 zero unformatted node pointer (hole) */
737 unfm_ptr = get_block_num(item, pos_in_item);
739 /* use allocated block to plug the hole */
740 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
741 if (fs_changed(fs_gen, inode->i_sb)
742 && item_moved(&tmp_ih, &path)) {
743 reiserfs_restore_prepared_buffer(inode->i_sb,
747 set_buffer_new(bh_result);
748 if (buffer_dirty(bh_result)
749 && reiserfs_data_ordered(inode->i_sb))
750 reiserfs_add_ordered_list(inode, bh_result);
751 put_block_num(item, pos_in_item, allocated_block_nr);
752 unfm_ptr = allocated_block_nr;
753 journal_mark_dirty(th, inode->i_sb, bh);
754 reiserfs_update_sd(th, inode);
756 set_block_dev_mapped(bh_result, unfm_ptr, inode);
760 retval = reiserfs_end_persistent_transaction(th);
762 reiserfs_write_unlock(inode->i_sb);
764 /* the item was found, so new blocks were not added to the file
765 ** there is no need to make sure the inode is updated with this
776 /* desired position is not found or is in the direct item. We have
777 to append file with holes up to 'block'-th block converting
778 direct items to indirect one if necessary */
781 if (is_statdata_le_ih(ih)) {
783 struct cpu_key tmp_key;
785 /* indirect item has to be inserted */
786 make_le_item_head(&tmp_ih, &key, version, 1,
787 TYPE_INDIRECT, UNFM_P_SIZE,
788 0 /* free_space */ );
790 if (cpu_key_k_offset(&key) == 1) {
791 /* we are going to add 'block'-th block to the file. Use
792 allocated block for that */
793 unp = cpu_to_le32(allocated_block_nr);
794 set_block_dev_mapped(bh_result,
795 allocated_block_nr, inode);
796 set_buffer_new(bh_result);
800 set_cpu_key_k_offset(&tmp_key, 1);
801 PATH_LAST_POSITION(&path)++;
804 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
805 inode, (char *)&unp);
807 reiserfs_free_block(th, inode,
808 allocated_block_nr, 1);
809 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
811 //mark_tail_converted (inode);
812 } else if (is_direct_le_ih(ih)) {
813 /* direct item has to be converted */
817 ((le_ih_k_offset(ih) -
818 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
819 if (tail_offset == cpu_key_k_offset(&key)) {
820 /* direct item we just found fits into block we have
821 to map. Convert it into unformatted node: use
822 bh_result for the conversion */
823 set_block_dev_mapped(bh_result,
824 allocated_block_nr, inode);
828 /* we have to padd file tail stored in direct item(s)
829 up to block size and convert it to unformatted
830 node. FIXME: this should also get into page cache */
834 * ugly, but we can only end the transaction if
837 BUG_ON(!th->t_refcount);
838 if (th->t_refcount == 1) {
840 reiserfs_end_persistent_transaction
848 convert_tail_for_hole(inode, bh_result,
851 if (retval != -ENOSPC)
852 reiserfs_error(inode->i_sb,
854 "convert tail failed "
855 "inode %lu, error %d",
858 if (allocated_block_nr) {
859 /* the bitmap, the super, and the stat data == 3 */
861 th = reiserfs_persistent_transaction(inode->i_sb, 3);
863 reiserfs_free_block(th,
873 direct2indirect(th, inode, &path, unbh,
876 reiserfs_unmap_buffer(unbh);
877 reiserfs_free_block(th, inode,
878 allocated_block_nr, 1);
881 /* it is important the set_buffer_uptodate is done after
882 ** the direct2indirect. The buffer might contain valid
883 ** data newer than the data on disk (read by readpage, changed,
884 ** and then sent here by writepage). direct2indirect needs
885 ** to know if unbh was already up to date, so it can decide
886 ** if the data in unbh needs to be replaced with data from
889 set_buffer_uptodate(unbh);
891 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
892 buffer will disappear shortly, so it should not be added to
895 /* we've converted the tail, so we must
896 ** flush unbh before the transaction commits
898 reiserfs_add_tail_list(inode, unbh);
900 /* mark it dirty now to prevent commit_write from adding
901 ** this buffer to the inode's dirty buffer list
904 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
905 * It's still atomic, but it sets the page dirty too,
906 * which makes it eligible for writeback at any time by the
907 * VM (which was also the case with __mark_buffer_dirty())
909 mark_buffer_dirty(unbh);
912 /* append indirect item with holes if needed, when appending
913 pointer to 'block'-th block use block, which is already
915 struct cpu_key tmp_key;
916 unp_t unf_single = 0; // We use this in case we need to allocate only
917 // one block which is a fastpath
919 __u64 max_to_insert =
920 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
924 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
925 "vs-804: invalid position for append");
926 /* indirect item has to be appended, set up key of that position */
927 make_cpu_key(&tmp_key, inode,
928 le_key_k_offset(version,
931 inode->i_sb->s_blocksize),
932 //pos_in_item * inode->i_sb->s_blocksize,
933 TYPE_INDIRECT, 3); // key type is unimportant
935 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
936 "green-805: invalid offset");
939 ((cpu_key_k_offset(&key) -
940 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
943 if (blocks_needed == 1) {
946 un = kzalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_NOFS);
953 if (blocks_needed <= max_to_insert) {
954 /* we are going to add target block to the file. Use allocated
956 un[blocks_needed - 1] =
957 cpu_to_le32(allocated_block_nr);
958 set_block_dev_mapped(bh_result,
959 allocated_block_nr, inode);
960 set_buffer_new(bh_result);
963 /* paste hole to the indirect item */
964 /* If kmalloc failed, max_to_insert becomes zero and it means we
965 only have space for one block */
967 max_to_insert ? max_to_insert : 1;
970 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
975 if (blocks_needed != 1)
979 reiserfs_free_block(th, inode,
980 allocated_block_nr, 1);
984 /* We need to mark new file size in case this function will be
985 interrupted/aborted later on. And we may do this only for
988 inode->i_sb->s_blocksize * blocks_needed;
995 /* this loop could log more blocks than we had originally asked
996 ** for. So, we have to allow the transaction to end if it is
997 ** too big or too full. Update the inode so things are
998 ** consistent if we crash before the function returns
1000 ** release the path so that anybody waiting on the path before
1001 ** ending their transaction will be able to continue.
1003 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
1004 retval = restart_transaction(th, inode, &path);
1009 * inserting indirect pointers for a hole can take a
1010 * long time. reschedule if needed and also release the write
1013 reiserfs_cond_resched(inode->i_sb);
1015 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1016 if (retval == IO_ERROR) {
1020 if (retval == POSITION_FOUND) {
1021 reiserfs_warning(inode->i_sb, "vs-825",
1022 "%K should not be found", &key);
1024 if (allocated_block_nr)
1025 reiserfs_free_block(th, inode,
1026 allocated_block_nr, 1);
1030 bh = get_last_bh(&path);
1032 item = get_item(&path);
1033 pos_in_item = path.pos_in_item;
1039 if (th && (!dangle || (retval && !th->t_trans_id))) {
1042 reiserfs_update_sd(th, inode);
1043 err = reiserfs_end_persistent_transaction(th);
1048 reiserfs_write_unlock(inode->i_sb);
1049 reiserfs_check_path(&path);
1054 reiserfs_readpages(struct file *file, struct address_space *mapping,
1055 struct list_head *pages, unsigned nr_pages)
1057 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
1060 /* Compute real number of used bytes by file
1061 * Following three functions can go away when we'll have enough space in stat item
1063 static int real_space_diff(struct inode *inode, int sd_size)
1066 loff_t blocksize = inode->i_sb->s_blocksize;
1068 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1071 /* End of file is also in full block with indirect reference, so round
1072 ** up to the next block.
1074 ** there is just no way to know if the tail is actually packed
1075 ** on the file, so we have to assume it isn't. When we pack the
1076 ** tail, we add 4 bytes to pretend there really is an unformatted
1081 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1086 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
1089 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1090 return inode->i_size +
1091 (loff_t) (real_space_diff(inode, sd_size));
1093 return ((loff_t) real_space_diff(inode, sd_size)) +
1094 (((loff_t) blocks) << 9);
1097 /* Compute number of blocks used by file in ReiserFS counting */
1098 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1100 loff_t bytes = inode_get_bytes(inode);
1101 loff_t real_space = real_space_diff(inode, sd_size);
1103 /* keeps fsck and non-quota versions of reiserfs happy */
1104 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1105 bytes += (loff_t) 511;
1108 /* files from before the quota patch might i_blocks such that
1109 ** bytes < real_space. Deal with that here to prevent it from
1112 if (bytes < real_space)
1114 return (bytes - real_space) >> 9;
1118 // BAD: new directories have stat data of new type and all other items
1119 // of old type. Version stored in the inode says about body items, so
1120 // in update_stat_data we can not rely on inode, but have to check
1121 // item version directly
1124 // called by read_locked_inode
1125 static void init_inode(struct inode *inode, struct treepath *path)
1127 struct buffer_head *bh;
1128 struct item_head *ih;
1130 //int version = ITEM_VERSION_1;
1132 bh = PATH_PLAST_BUFFER(path);
1133 ih = PATH_PITEM_HEAD(path);
1135 copy_key(INODE_PKEY(inode), &(ih->ih_key));
1137 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1138 REISERFS_I(inode)->i_flags = 0;
1139 REISERFS_I(inode)->i_prealloc_block = 0;
1140 REISERFS_I(inode)->i_prealloc_count = 0;
1141 REISERFS_I(inode)->i_trans_id = 0;
1142 REISERFS_I(inode)->i_jl = NULL;
1143 reiserfs_init_xattr_rwsem(inode);
1145 if (stat_data_v1(ih)) {
1146 struct stat_data_v1 *sd =
1147 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1148 unsigned long blocks;
1150 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1151 set_inode_sd_version(inode, STAT_DATA_V1);
1152 inode->i_mode = sd_v1_mode(sd);
1153 set_nlink(inode, sd_v1_nlink(sd));
1154 i_uid_write(inode, sd_v1_uid(sd));
1155 i_gid_write(inode, sd_v1_gid(sd));
1156 inode->i_size = sd_v1_size(sd);
1157 inode->i_atime.tv_sec = sd_v1_atime(sd);
1158 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1159 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1160 inode->i_atime.tv_nsec = 0;
1161 inode->i_ctime.tv_nsec = 0;
1162 inode->i_mtime.tv_nsec = 0;
1164 inode->i_blocks = sd_v1_blocks(sd);
1165 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1166 blocks = (inode->i_size + 511) >> 9;
1167 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1168 if (inode->i_blocks > blocks) {
1169 // there was a bug in <=3.5.23 when i_blocks could take negative
1170 // values. Starting from 3.5.17 this value could even be stored in
1171 // stat data. For such files we set i_blocks based on file
1172 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1173 // only updated if file's inode will ever change
1174 inode->i_blocks = blocks;
1177 rdev = sd_v1_rdev(sd);
1178 REISERFS_I(inode)->i_first_direct_byte =
1179 sd_v1_first_direct_byte(sd);
1180 /* an early bug in the quota code can give us an odd number for the
1181 ** block count. This is incorrect, fix it here.
1183 if (inode->i_blocks & 1) {
1186 inode_set_bytes(inode,
1187 to_real_used_space(inode, inode->i_blocks,
1189 /* nopack is initially zero for v1 objects. For v2 objects,
1190 nopack is initialised from sd_attrs */
1191 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1193 // new stat data found, but object may have old items
1194 // (directories and symlinks)
1195 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1197 inode->i_mode = sd_v2_mode(sd);
1198 set_nlink(inode, sd_v2_nlink(sd));
1199 i_uid_write(inode, sd_v2_uid(sd));
1200 inode->i_size = sd_v2_size(sd);
1201 i_gid_write(inode, sd_v2_gid(sd));
1202 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1203 inode->i_atime.tv_sec = sd_v2_atime(sd);
1204 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1205 inode->i_ctime.tv_nsec = 0;
1206 inode->i_mtime.tv_nsec = 0;
1207 inode->i_atime.tv_nsec = 0;
1208 inode->i_blocks = sd_v2_blocks(sd);
1209 rdev = sd_v2_rdev(sd);
1210 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1211 inode->i_generation =
1212 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1214 inode->i_generation = sd_v2_generation(sd);
1216 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1217 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1219 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1220 REISERFS_I(inode)->i_first_direct_byte = 0;
1221 set_inode_sd_version(inode, STAT_DATA_V2);
1222 inode_set_bytes(inode,
1223 to_real_used_space(inode, inode->i_blocks,
1225 /* read persistent inode attributes from sd and initialise
1226 generic inode flags from them */
1227 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1228 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
1232 if (S_ISREG(inode->i_mode)) {
1233 inode->i_op = &reiserfs_file_inode_operations;
1234 inode->i_fop = &reiserfs_file_operations;
1235 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1236 } else if (S_ISDIR(inode->i_mode)) {
1237 inode->i_op = &reiserfs_dir_inode_operations;
1238 inode->i_fop = &reiserfs_dir_operations;
1239 } else if (S_ISLNK(inode->i_mode)) {
1240 inode->i_op = &reiserfs_symlink_inode_operations;
1241 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1243 inode->i_blocks = 0;
1244 inode->i_op = &reiserfs_special_inode_operations;
1245 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1249 // update new stat data with inode fields
1250 static void inode2sd(void *sd, struct inode *inode, loff_t size)
1252 struct stat_data *sd_v2 = (struct stat_data *)sd;
1255 set_sd_v2_mode(sd_v2, inode->i_mode);
1256 set_sd_v2_nlink(sd_v2, inode->i_nlink);
1257 set_sd_v2_uid(sd_v2, i_uid_read(inode));
1258 set_sd_v2_size(sd_v2, size);
1259 set_sd_v2_gid(sd_v2, i_gid_read(inode));
1260 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1261 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1262 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1263 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1264 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1265 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1267 set_sd_v2_generation(sd_v2, inode->i_generation);
1268 flags = REISERFS_I(inode)->i_attrs;
1269 i_attrs_to_sd_attrs(inode, &flags);
1270 set_sd_v2_attrs(sd_v2, flags);
1273 // used to copy inode's fields to old stat data
1274 static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
1276 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
1278 set_sd_v1_mode(sd_v1, inode->i_mode);
1279 set_sd_v1_uid(sd_v1, i_uid_read(inode));
1280 set_sd_v1_gid(sd_v1, i_gid_read(inode));
1281 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1282 set_sd_v1_size(sd_v1, size);
1283 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1284 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1285 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
1287 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1288 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1290 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1292 // Sigh. i_first_direct_byte is back
1293 set_sd_v1_first_direct_byte(sd_v1,
1294 REISERFS_I(inode)->i_first_direct_byte);
1297 /* NOTE, you must prepare the buffer head before sending it here,
1298 ** and then log it after the call
1300 static void update_stat_data(struct treepath *path, struct inode *inode,
1303 struct buffer_head *bh;
1304 struct item_head *ih;
1306 bh = PATH_PLAST_BUFFER(path);
1307 ih = PATH_PITEM_HEAD(path);
1309 if (!is_statdata_le_ih(ih))
1310 reiserfs_panic(inode->i_sb, "vs-13065", "key %k, found item %h",
1311 INODE_PKEY(inode), ih);
1313 if (stat_data_v1(ih)) {
1314 // path points to old stat data
1315 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1317 inode2sd(B_I_PITEM(bh, ih), inode, size);
1323 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1324 struct inode *inode, loff_t size)
1327 INITIALIZE_PATH(path);
1328 struct buffer_head *bh;
1330 struct item_head *ih, tmp_ih;
1333 BUG_ON(!th->t_trans_id);
1335 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant
1339 /* look for the object's stat data */
1340 retval = search_item(inode->i_sb, &key, &path);
1341 if (retval == IO_ERROR) {
1342 reiserfs_error(inode->i_sb, "vs-13050",
1343 "i/o failure occurred trying to "
1344 "update %K stat data", &key);
1347 if (retval == ITEM_NOT_FOUND) {
1348 pos = PATH_LAST_POSITION(&path);
1350 if (inode->i_nlink == 0) {
1351 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1354 reiserfs_warning(inode->i_sb, "vs-13060",
1355 "stat data of object %k (nlink == %d) "
1356 "not found (pos %d)",
1357 INODE_PKEY(inode), inode->i_nlink,
1359 reiserfs_check_path(&path);
1363 /* sigh, prepare_for_journal might schedule. When it schedules the
1364 ** FS might change. We have to detect that, and loop back to the
1365 ** search if the stat data item has moved
1367 bh = get_last_bh(&path);
1369 copy_item_head(&tmp_ih, ih);
1370 fs_gen = get_generation(inode->i_sb);
1371 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1372 if (fs_changed(fs_gen, inode->i_sb)
1373 && item_moved(&tmp_ih, &path)) {
1374 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1375 continue; /* Stat_data item has been moved after scheduling. */
1379 update_stat_data(&path, inode, size);
1380 journal_mark_dirty(th, th->t_super, bh);
1385 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1386 ** does a make_bad_inode when things go wrong. But, we need to make sure
1387 ** and clear the key in the private portion of the inode, otherwise a
1388 ** corresponding iput might try to delete whatever object the inode last
1391 static void reiserfs_make_bad_inode(struct inode *inode)
1393 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1394 make_bad_inode(inode);
1398 // initially this function was derived from minix or ext2's analog and
1399 // evolved as the prototype did
1402 int reiserfs_init_locked_inode(struct inode *inode, void *p)
1404 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1405 inode->i_ino = args->objectid;
1406 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1410 /* looks for stat data in the tree, and fills up the fields of in-core
1411 inode stat data fields */
1412 void reiserfs_read_locked_inode(struct inode *inode,
1413 struct reiserfs_iget_args *args)
1415 INITIALIZE_PATH(path_to_sd);
1417 unsigned long dirino;
1420 dirino = args->dirid;
1422 /* set version 1, version 2 could be used too, because stat data
1423 key is the same in both versions */
1424 key.version = KEY_FORMAT_3_5;
1425 key.on_disk_key.k_dir_id = dirino;
1426 key.on_disk_key.k_objectid = inode->i_ino;
1427 key.on_disk_key.k_offset = 0;
1428 key.on_disk_key.k_type = 0;
1430 /* look for the object's stat data */
1431 retval = search_item(inode->i_sb, &key, &path_to_sd);
1432 if (retval == IO_ERROR) {
1433 reiserfs_error(inode->i_sb, "vs-13070",
1434 "i/o failure occurred trying to find "
1435 "stat data of %K", &key);
1436 reiserfs_make_bad_inode(inode);
1439 if (retval != ITEM_FOUND) {
1440 /* a stale NFS handle can trigger this without it being an error */
1441 pathrelse(&path_to_sd);
1442 reiserfs_make_bad_inode(inode);
1447 init_inode(inode, &path_to_sd);
1449 /* It is possible that knfsd is trying to access inode of a file
1450 that is being removed from the disk by some other thread. As we
1451 update sd on unlink all that is required is to check for nlink
1452 here. This bug was first found by Sizif when debugging
1453 SquidNG/Butterfly, forgotten, and found again after Philippe
1454 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1456 More logical fix would require changes in fs/inode.c:iput() to
1457 remove inode from hash-table _after_ fs cleaned disk stuff up and
1458 in iget() to return NULL if I_FREEING inode is found in
1460 /* Currently there is one place where it's ok to meet inode with
1461 nlink==0: processing of open-unlinked and half-truncated files
1462 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1463 if ((inode->i_nlink == 0) &&
1464 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
1465 reiserfs_warning(inode->i_sb, "vs-13075",
1466 "dead inode read from disk %K. "
1467 "This is likely to be race with knfsd. Ignore",
1469 reiserfs_make_bad_inode(inode);
1472 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
1475 * Stat data v1 doesn't support ACLs.
1477 if (get_inode_sd_version(inode) == STAT_DATA_V1)
1478 cache_no_acl(inode);
1482 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1484 * @inode: inode from hash table to check
1485 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1487 * This function is called by iget5_locked() to distinguish reiserfs inodes
1488 * having the same inode numbers. Such inodes can only exist due to some
1489 * error condition. One of them should be bad. Inodes with identical
1490 * inode numbers (objectids) are distinguished by parent directory ids.
1493 int reiserfs_find_actor(struct inode *inode, void *opaque)
1495 struct reiserfs_iget_args *args;
1498 /* args is already in CPU order */
1499 return (inode->i_ino == args->objectid) &&
1500 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1503 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
1505 struct inode *inode;
1506 struct reiserfs_iget_args args;
1509 args.objectid = key->on_disk_key.k_objectid;
1510 args.dirid = key->on_disk_key.k_dir_id;
1511 depth = reiserfs_write_unlock_nested(s);
1512 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1513 reiserfs_find_actor, reiserfs_init_locked_inode,
1515 reiserfs_write_lock_nested(s, depth);
1517 return ERR_PTR(-ENOMEM);
1519 if (inode->i_state & I_NEW) {
1520 reiserfs_read_locked_inode(inode, &args);
1521 unlock_new_inode(inode);
1524 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1525 /* either due to i/o error or a stale NFS handle */
1532 static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1533 u32 objectid, u32 dir_id, u32 generation)
1537 struct inode *inode;
1539 key.on_disk_key.k_objectid = objectid;
1540 key.on_disk_key.k_dir_id = dir_id;
1541 reiserfs_write_lock(sb);
1542 inode = reiserfs_iget(sb, &key);
1543 if (inode && !IS_ERR(inode) && generation != 0 &&
1544 generation != inode->i_generation) {
1548 reiserfs_write_unlock(sb);
1550 return d_obtain_alias(inode);
1553 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1554 int fh_len, int fh_type)
1556 /* fhtype happens to reflect the number of u32s encoded.
1557 * due to a bug in earlier code, fhtype might indicate there
1558 * are more u32s then actually fitted.
1559 * so if fhtype seems to be more than len, reduce fhtype.
1561 * 2 - objectid + dir_id - legacy support
1562 * 3 - objectid + dir_id + generation
1563 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1564 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1565 * 6 - as above plus generation of directory
1566 * 6 does not fit in NFSv2 handles
1568 if (fh_type > fh_len) {
1569 if (fh_type != 6 || fh_len != 5)
1570 reiserfs_warning(sb, "reiserfs-13077",
1571 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1578 return reiserfs_get_dentry(sb, fid->raw[0], fid->raw[1],
1579 (fh_type == 3 || fh_type >= 5) ? fid->raw[2] : 0);
1582 struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1583 int fh_len, int fh_type)
1585 if (fh_type > fh_len)
1590 return reiserfs_get_dentry(sb,
1591 (fh_type >= 5) ? fid->raw[3] : fid->raw[2],
1592 (fh_type >= 5) ? fid->raw[4] : fid->raw[3],
1593 (fh_type == 6) ? fid->raw[5] : 0);
1596 int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
1597 struct inode *parent)
1601 if (parent && (maxlen < 5)) {
1603 return FILEID_INVALID;
1604 } else if (maxlen < 3) {
1606 return FILEID_INVALID;
1609 data[0] = inode->i_ino;
1610 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1611 data[2] = inode->i_generation;
1614 data[3] = parent->i_ino;
1615 data[4] = le32_to_cpu(INODE_PKEY(parent)->k_dir_id);
1618 data[5] = parent->i_generation;
1625 /* looks for stat data, then copies fields to it, marks the buffer
1626 containing stat data as dirty */
1627 /* reiserfs inodes are never really dirty, since the dirty inode call
1628 ** always logs them. This call allows the VFS inode marking routines
1629 ** to properly mark inodes for datasync and such, but only actually
1630 ** does something when called for a synchronous update.
1632 int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc)
1634 struct reiserfs_transaction_handle th;
1635 int jbegin_count = 1;
1637 if (inode->i_sb->s_flags & MS_RDONLY)
1639 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1640 ** these cases are just when the system needs ram, not when the
1641 ** inode needs to reach disk for safety, and they can safely be
1642 ** ignored because the altered inode has already been logged.
1644 if (wbc->sync_mode == WB_SYNC_ALL && !(current->flags & PF_MEMALLOC)) {
1645 reiserfs_write_lock(inode->i_sb);
1646 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1647 reiserfs_update_sd(&th, inode);
1648 journal_end_sync(&th, inode->i_sb, jbegin_count);
1650 reiserfs_write_unlock(inode->i_sb);
1655 /* stat data of new object is inserted already, this inserts the item
1656 containing "." and ".." entries */
1657 static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1658 struct inode *inode,
1659 struct item_head *ih, struct treepath *path,
1662 struct super_block *sb = th->t_super;
1663 char empty_dir[EMPTY_DIR_SIZE];
1664 char *body = empty_dir;
1668 BUG_ON(!th->t_trans_id);
1670 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1671 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1672 TYPE_DIRENTRY, 3 /*key length */ );
1674 /* compose item head for new item. Directories consist of items of
1675 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1676 is done by reiserfs_new_inode */
1677 if (old_format_only(sb)) {
1678 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1679 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1681 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1682 ih->ih_key.k_objectid,
1683 INODE_PKEY(dir)->k_dir_id,
1684 INODE_PKEY(dir)->k_objectid);
1686 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1687 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1689 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1690 ih->ih_key.k_objectid,
1691 INODE_PKEY(dir)->k_dir_id,
1692 INODE_PKEY(dir)->k_objectid);
1695 /* look for place in the tree for new item */
1696 retval = search_item(sb, &key, path);
1697 if (retval == IO_ERROR) {
1698 reiserfs_error(sb, "vs-13080",
1699 "i/o failure occurred creating new directory");
1702 if (retval == ITEM_FOUND) {
1704 reiserfs_warning(sb, "vs-13070",
1705 "object with this key exists (%k)",
1710 /* insert item, that is empty directory item */
1711 return reiserfs_insert_item(th, path, &key, ih, inode, body);
1714 /* stat data of object has been inserted, this inserts the item
1715 containing the body of symlink */
1716 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1717 struct item_head *ih,
1718 struct treepath *path, const char *symname,
1721 struct super_block *sb = th->t_super;
1725 BUG_ON(!th->t_trans_id);
1727 _make_cpu_key(&key, KEY_FORMAT_3_5,
1728 le32_to_cpu(ih->ih_key.k_dir_id),
1729 le32_to_cpu(ih->ih_key.k_objectid),
1730 1, TYPE_DIRECT, 3 /*key length */ );
1732 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1733 0 /*free_space */ );
1735 /* look for place in the tree for new item */
1736 retval = search_item(sb, &key, path);
1737 if (retval == IO_ERROR) {
1738 reiserfs_error(sb, "vs-13080",
1739 "i/o failure occurred creating new symlink");
1742 if (retval == ITEM_FOUND) {
1744 reiserfs_warning(sb, "vs-13080",
1745 "object with this key exists (%k)",
1750 /* insert item, that is body of symlink */
1751 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
1754 /* inserts the stat data into the tree, and then calls
1755 reiserfs_new_directory (to insert ".", ".." item if new object is
1756 directory) or reiserfs_new_symlink (to insert symlink body if new
1757 object is symlink) or nothing (if new object is regular file)
1759 NOTE! uid and gid must already be set in the inode. If we return
1760 non-zero due to an error, we have to drop the quota previously allocated
1761 for the fresh inode. This can only be done outside a transaction, so
1762 if we return non-zero, we also end the transaction. */
1763 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1764 struct inode *dir, umode_t mode, const char *symname,
1765 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1766 strlen (symname) for symlinks) */
1767 loff_t i_size, struct dentry *dentry,
1768 struct inode *inode,
1769 struct reiserfs_security_handle *security)
1771 struct super_block *sb;
1772 struct reiserfs_iget_args args;
1773 INITIALIZE_PATH(path_to_key);
1775 struct item_head ih;
1776 struct stat_data sd;
1781 BUG_ON(!th->t_trans_id);
1783 reiserfs_write_unlock(inode->i_sb);
1784 err = dquot_alloc_inode(inode);
1785 reiserfs_write_lock(inode->i_sb);
1788 if (!dir->i_nlink) {
1795 /* item head of new item */
1796 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1797 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1798 if (!ih.ih_key.k_objectid) {
1802 args.objectid = inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
1803 if (old_format_only(sb))
1804 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1805 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1807 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1808 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1809 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1810 args.dirid = le32_to_cpu(ih.ih_key.k_dir_id);
1812 depth = reiserfs_write_unlock_nested(inode->i_sb);
1813 err = insert_inode_locked4(inode, args.objectid,
1814 reiserfs_find_actor, &args);
1815 reiserfs_write_lock_nested(inode->i_sb, depth);
1821 if (old_format_only(sb))
1822 /* not a perfect generation count, as object ids can be reused, but
1823 ** this is as good as reiserfs can do right now.
1824 ** note that the private part of inode isn't filled in yet, we have
1825 ** to use the directory.
1827 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1829 #if defined( USE_INODE_GENERATION_COUNTER )
1830 inode->i_generation =
1831 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1833 inode->i_generation = ++event;
1836 /* fill stat data */
1837 set_nlink(inode, (S_ISDIR(mode) ? 2 : 1));
1839 /* uid and gid must already be set by the caller for quota init */
1841 /* symlink cannot be immutable or append only, right? */
1842 if (S_ISLNK(inode->i_mode))
1843 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1845 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1846 inode->i_size = i_size;
1847 inode->i_blocks = 0;
1849 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1850 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1852 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1853 REISERFS_I(inode)->i_flags = 0;
1854 REISERFS_I(inode)->i_prealloc_block = 0;
1855 REISERFS_I(inode)->i_prealloc_count = 0;
1856 REISERFS_I(inode)->i_trans_id = 0;
1857 REISERFS_I(inode)->i_jl = NULL;
1858 REISERFS_I(inode)->i_attrs =
1859 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1860 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
1861 reiserfs_init_xattr_rwsem(inode);
1863 /* key to search for correct place for new stat data */
1864 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1865 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1866 TYPE_STAT_DATA, 3 /*key length */ );
1868 /* find proper place for inserting of stat data */
1869 retval = search_item(sb, &key, &path_to_key);
1870 if (retval == IO_ERROR) {
1874 if (retval == ITEM_FOUND) {
1875 pathrelse(&path_to_key);
1879 if (old_format_only(sb)) {
1880 if (i_uid_read(inode) & ~0xffff || i_gid_read(inode) & ~0xffff) {
1881 pathrelse(&path_to_key);
1882 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1886 inode2sd_v1(&sd, inode, inode->i_size);
1888 inode2sd(&sd, inode, inode->i_size);
1890 // store in in-core inode the key of stat data and version all
1891 // object items will have (directory items will have old offset
1892 // format, other new objects will consist of new items)
1893 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1894 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1896 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1897 if (old_format_only(sb))
1898 set_inode_sd_version(inode, STAT_DATA_V1);
1900 set_inode_sd_version(inode, STAT_DATA_V2);
1902 /* insert the stat data into the tree */
1903 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1904 if (REISERFS_I(dir)->new_packing_locality)
1905 th->displace_new_blocks = 1;
1908 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1912 reiserfs_check_path(&path_to_key);
1915 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1916 if (!th->displace_new_blocks)
1917 REISERFS_I(dir)->new_packing_locality = 0;
1919 if (S_ISDIR(mode)) {
1920 /* insert item with "." and ".." */
1922 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1925 if (S_ISLNK(mode)) {
1926 /* insert body of symlink */
1927 if (!old_format_only(sb))
1928 i_size = ROUND_UP(i_size);
1930 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1935 reiserfs_check_path(&path_to_key);
1936 journal_end(th, th->t_super, th->t_blocks_allocated);
1937 goto out_inserted_sd;
1940 if (reiserfs_posixacl(inode->i_sb)) {
1941 reiserfs_write_unlock(inode->i_sb);
1942 retval = reiserfs_inherit_default_acl(th, dir, dentry, inode);
1943 reiserfs_write_lock(inode->i_sb);
1946 reiserfs_check_path(&path_to_key);
1947 journal_end(th, th->t_super, th->t_blocks_allocated);
1948 goto out_inserted_sd;
1950 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1951 reiserfs_warning(inode->i_sb, "jdm-13090",
1952 "ACLs aren't enabled in the fs, "
1953 "but vfs thinks they are!");
1954 } else if (IS_PRIVATE(dir))
1955 inode->i_flags |= S_PRIVATE;
1957 if (security->name) {
1958 reiserfs_write_unlock(inode->i_sb);
1959 retval = reiserfs_security_write(th, inode, security);
1960 reiserfs_write_lock(inode->i_sb);
1963 reiserfs_check_path(&path_to_key);
1964 retval = journal_end(th, th->t_super,
1965 th->t_blocks_allocated);
1968 goto out_inserted_sd;
1972 reiserfs_update_sd(th, inode);
1973 reiserfs_check_path(&path_to_key);
1977 /* it looks like you can easily compress these two goto targets into
1978 * one. Keeping it like this doesn't actually hurt anything, and they
1979 * are place holders for what the quota code actually needs.
1982 /* Invalidate the object, nothing was inserted yet */
1983 INODE_PKEY(inode)->k_objectid = 0;
1985 /* Quota change must be inside a transaction for journaling */
1986 dquot_free_inode(inode);
1989 journal_end(th, th->t_super, th->t_blocks_allocated);
1990 reiserfs_write_unlock(inode->i_sb);
1991 /* Drop can be outside and it needs more credits so it's better to have it outside */
1993 reiserfs_write_lock(inode->i_sb);
1994 inode->i_flags |= S_NOQUOTA;
1995 make_bad_inode(inode);
1999 th->t_trans_id = 0; /* so the caller can't use this handle later */
2000 unlock_new_inode(inode); /* OK to do even if we hadn't locked it */
2006 ** finds the tail page in the page cache,
2007 ** reads the last block in.
2009 ** On success, page_result is set to a locked, pinned page, and bh_result
2010 ** is set to an up to date buffer for the last block in the file. returns 0.
2012 ** tail conversion is not done, so bh_result might not be valid for writing
2013 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2014 ** trying to write the block.
2016 ** on failure, nonzero is returned, page_result and bh_result are untouched.
2018 static int grab_tail_page(struct inode *inode,
2019 struct page **page_result,
2020 struct buffer_head **bh_result)
2023 /* we want the page with the last byte in the file,
2024 ** not the page that will hold the next byte for appending
2026 unsigned long index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
2027 unsigned long pos = 0;
2028 unsigned long start = 0;
2029 unsigned long blocksize = inode->i_sb->s_blocksize;
2030 unsigned long offset = (inode->i_size) & (PAGE_CACHE_SIZE - 1);
2031 struct buffer_head *bh;
2032 struct buffer_head *head;
2036 /* we know that we are only called with inode->i_size > 0.
2037 ** we also know that a file tail can never be as big as a block
2038 ** If i_size % blocksize == 0, our file is currently block aligned
2039 ** and it won't need converting or zeroing after a truncate.
2041 if ((offset & (blocksize - 1)) == 0) {
2044 page = grab_cache_page(inode->i_mapping, index);
2049 /* start within the page of the last block in the file */
2050 start = (offset / blocksize) * blocksize;
2052 error = __block_write_begin(page, start, offset - start,
2053 reiserfs_get_block_create_0);
2057 head = page_buffers(page);
2063 bh = bh->b_this_page;
2065 } while (bh != head);
2067 if (!buffer_uptodate(bh)) {
2068 /* note, this should never happen, prepare_write should
2069 ** be taking care of this for us. If the buffer isn't up to date,
2070 ** I've screwed up the code to find the buffer, or the code to
2071 ** call prepare_write
2073 reiserfs_error(inode->i_sb, "clm-6000",
2074 "error reading block %lu", bh->b_blocknr);
2079 *page_result = page;
2086 page_cache_release(page);
2091 ** vfs version of truncate file. Must NOT be called with
2092 ** a transaction already started.
2094 ** some code taken from block_truncate_page
2096 int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
2098 struct reiserfs_transaction_handle th;
2099 /* we want the offset for the first byte after the end of the file */
2100 unsigned long offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2101 unsigned blocksize = inode->i_sb->s_blocksize;
2103 struct page *page = NULL;
2105 struct buffer_head *bh = NULL;
2108 reiserfs_write_lock(inode->i_sb);
2110 if (inode->i_size > 0) {
2111 error = grab_tail_page(inode, &page, &bh);
2113 // -ENOENT means we truncated past the end of the file,
2114 // and get_block_create_0 could not find a block to read in,
2116 if (error != -ENOENT)
2117 reiserfs_error(inode->i_sb, "clm-6001",
2118 "grab_tail_page failed %d",
2125 /* so, if page != NULL, we have a buffer head for the offset at
2126 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2127 ** then we have an unformatted node. Otherwise, we have a direct item,
2128 ** and no zeroing is required on disk. We zero after the truncate,
2129 ** because the truncate might pack the item anyway
2130 ** (it will unmap bh if it packs).
2132 /* it is enough to reserve space in transaction for 2 balancings:
2133 one for "save" link adding and another for the first
2134 cut_from_item. 1 is for update_sd */
2135 error = journal_begin(&th, inode->i_sb,
2136 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2139 reiserfs_update_inode_transaction(inode);
2140 if (update_timestamps)
2141 /* we are doing real truncate: if the system crashes before the last
2142 transaction of truncating gets committed - on reboot the file
2143 either appears truncated properly or not truncated at all */
2144 add_save_link(&th, inode, 1);
2145 err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps);
2147 journal_end(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2151 /* check reiserfs_do_truncate after ending the transaction */
2157 if (update_timestamps) {
2158 error = remove_save_link(inode, 1 /* truncate */);
2164 length = offset & (blocksize - 1);
2165 /* if we are not on a block boundary */
2167 length = blocksize - length;
2168 zero_user(page, offset, length);
2169 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2170 mark_buffer_dirty(bh);
2174 page_cache_release(page);
2177 reiserfs_write_unlock(inode->i_sb);
2183 page_cache_release(page);
2186 reiserfs_write_unlock(inode->i_sb);
2191 static int map_block_for_writepage(struct inode *inode,
2192 struct buffer_head *bh_result,
2193 unsigned long block)
2195 struct reiserfs_transaction_handle th;
2197 struct item_head tmp_ih;
2198 struct item_head *ih;
2199 struct buffer_head *bh;
2202 INITIALIZE_PATH(path);
2204 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
2205 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
2207 int use_get_block = 0;
2208 int bytes_copied = 0;
2210 int trans_running = 0;
2212 /* catch places below that try to log something without starting a trans */
2215 if (!buffer_uptodate(bh_result)) {
2219 kmap(bh_result->b_page);
2221 reiserfs_write_lock(inode->i_sb);
2222 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
2225 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2226 if (retval != POSITION_FOUND) {
2231 bh = get_last_bh(&path);
2233 item = get_item(&path);
2234 pos_in_item = path.pos_in_item;
2236 /* we've found an unformatted node */
2237 if (indirect_item_found(retval, ih)) {
2238 if (bytes_copied > 0) {
2239 reiserfs_warning(inode->i_sb, "clm-6002",
2240 "bytes_copied %d", bytes_copied);
2242 if (!get_block_num(item, pos_in_item)) {
2243 /* crap, we are writing to a hole */
2247 set_block_dev_mapped(bh_result,
2248 get_block_num(item, pos_in_item), inode);
2249 } else if (is_direct_le_ih(ih)) {
2251 p = page_address(bh_result->b_page);
2252 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2253 copy_size = ih_item_len(ih) - pos_in_item;
2255 fs_gen = get_generation(inode->i_sb);
2256 copy_item_head(&tmp_ih, ih);
2258 if (!trans_running) {
2259 /* vs-3050 is gone, no need to drop the path */
2260 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2263 reiserfs_update_inode_transaction(inode);
2265 if (fs_changed(fs_gen, inode->i_sb)
2266 && item_moved(&tmp_ih, &path)) {
2267 reiserfs_restore_prepared_buffer(inode->i_sb,
2273 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2275 if (fs_changed(fs_gen, inode->i_sb)
2276 && item_moved(&tmp_ih, &path)) {
2277 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2281 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2284 journal_mark_dirty(&th, inode->i_sb, bh);
2285 bytes_copied += copy_size;
2286 set_block_dev_mapped(bh_result, 0, inode);
2288 /* are there still bytes left? */
2289 if (bytes_copied < bh_result->b_size &&
2290 (byte_offset + bytes_copied) < inode->i_size) {
2291 set_cpu_key_k_offset(&key,
2292 cpu_key_k_offset(&key) +
2297 reiserfs_warning(inode->i_sb, "clm-6003",
2298 "bad item inode %lu", inode->i_ino);
2306 if (trans_running) {
2307 int err = journal_end(&th, inode->i_sb, jbegin_count);
2312 reiserfs_write_unlock(inode->i_sb);
2314 /* this is where we fill in holes in the file. */
2315 if (use_get_block) {
2316 retval = reiserfs_get_block(inode, block, bh_result,
2317 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2318 | GET_BLOCK_NO_DANGLE);
2320 if (!buffer_mapped(bh_result)
2321 || bh_result->b_blocknr == 0) {
2322 /* get_block failed to find a mapped unformatted node. */
2328 kunmap(bh_result->b_page);
2330 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2331 /* we've copied data from the page into the direct item, so the
2332 * buffer in the page is now clean, mark it to reflect that.
2334 lock_buffer(bh_result);
2335 clear_buffer_dirty(bh_result);
2336 unlock_buffer(bh_result);
2342 * mason@suse.com: updated in 2.5.54 to follow the same general io
2343 * start/recovery path as __block_write_full_page, along with special
2344 * code to handle reiserfs tails.
2346 static int reiserfs_write_full_page(struct page *page,
2347 struct writeback_control *wbc)
2349 struct inode *inode = page->mapping->host;
2350 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2352 unsigned long block;
2353 sector_t last_block;
2354 struct buffer_head *head, *bh;
2357 int checked = PageChecked(page);
2358 struct reiserfs_transaction_handle th;
2359 struct super_block *s = inode->i_sb;
2360 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2363 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2364 if (checked && (current->flags & PF_MEMALLOC)) {
2365 redirty_page_for_writepage(wbc, page);
2370 /* The page dirty bit is cleared before writepage is called, which
2371 * means we have to tell create_empty_buffers to make dirty buffers
2372 * The page really should be up to date at this point, so tossing
2373 * in the BH_Uptodate is just a sanity check.
2375 if (!page_has_buffers(page)) {
2376 create_empty_buffers(page, s->s_blocksize,
2377 (1 << BH_Dirty) | (1 << BH_Uptodate));
2379 head = page_buffers(page);
2381 /* last page in the file, zero out any contents past the
2382 ** last byte in the file
2384 if (page->index >= end_index) {
2385 unsigned last_offset;
2387 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2388 /* no file contents in this page */
2389 if (page->index >= end_index + 1 || !last_offset) {
2393 zero_user_segment(page, last_offset, PAGE_CACHE_SIZE);
2396 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
2397 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
2398 /* first map all the buffers, logging any direct items we find */
2400 if (block > last_block) {
2402 * This can happen when the block size is less than
2403 * the page size. The corresponding bytes in the page
2404 * were zero filled above
2406 clear_buffer_dirty(bh);
2407 set_buffer_uptodate(bh);
2408 } else if ((checked || buffer_dirty(bh)) &&
2409 (!buffer_mapped(bh) || (buffer_mapped(bh)
2412 /* not mapped yet, or it points to a direct item, search
2413 * the btree for the mapping info, and log any direct
2416 if ((error = map_block_for_writepage(inode, bh, block))) {
2420 bh = bh->b_this_page;
2422 } while (bh != head);
2425 * we start the transaction after map_block_for_writepage,
2426 * because it can create holes in the file (an unbounded operation).
2427 * starting it here, we can make a reliable estimate for how many
2428 * blocks we're going to log
2431 ClearPageChecked(page);
2432 reiserfs_write_lock(s);
2433 error = journal_begin(&th, s, bh_per_page + 1);
2435 reiserfs_write_unlock(s);
2438 reiserfs_update_inode_transaction(inode);
2440 /* now go through and lock any dirty buffers on the page */
2443 if (!buffer_mapped(bh))
2445 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2449 reiserfs_prepare_for_journal(s, bh, 1);
2450 journal_mark_dirty(&th, s, bh);
2453 /* from this point on, we know the buffer is mapped to a
2454 * real block and not a direct item
2456 if (wbc->sync_mode != WB_SYNC_NONE) {
2459 if (!trylock_buffer(bh)) {
2460 redirty_page_for_writepage(wbc, page);
2464 if (test_clear_buffer_dirty(bh)) {
2465 mark_buffer_async_write(bh);
2469 } while ((bh = bh->b_this_page) != head);
2472 error = journal_end(&th, s, bh_per_page + 1);
2473 reiserfs_write_unlock(s);
2477 BUG_ON(PageWriteback(page));
2478 set_page_writeback(page);
2482 * since any buffer might be the only dirty buffer on the page,
2483 * the first submit_bh can bring the page out of writeback.
2484 * be careful with the buffers.
2487 struct buffer_head *next = bh->b_this_page;
2488 if (buffer_async_write(bh)) {
2489 submit_bh(WRITE, bh);
2494 } while (bh != head);
2500 * if this page only had a direct item, it is very possible for
2501 * no io to be required without there being an error. Or,
2502 * someone else could have locked them and sent them down the
2503 * pipe without locking the page
2507 if (!buffer_uptodate(bh)) {
2511 bh = bh->b_this_page;
2512 } while (bh != head);
2514 SetPageUptodate(page);
2515 end_page_writeback(page);
2520 /* catches various errors, we need to make sure any valid dirty blocks
2521 * get to the media. The page is currently locked and not marked for
2524 ClearPageUptodate(page);
2528 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2530 mark_buffer_async_write(bh);
2533 * clear any dirty bits that might have come from getting
2534 * attached to a dirty page
2536 clear_buffer_dirty(bh);
2538 bh = bh->b_this_page;
2539 } while (bh != head);
2541 BUG_ON(PageWriteback(page));
2542 set_page_writeback(page);
2545 struct buffer_head *next = bh->b_this_page;
2546 if (buffer_async_write(bh)) {
2547 clear_buffer_dirty(bh);
2548 submit_bh(WRITE, bh);
2553 } while (bh != head);
2557 static int reiserfs_readpage(struct file *f, struct page *page)
2559 return block_read_full_page(page, reiserfs_get_block);
2562 static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
2564 struct inode *inode = page->mapping->host;
2565 reiserfs_wait_on_write_block(inode->i_sb);
2566 return reiserfs_write_full_page(page, wbc);
2569 static void reiserfs_truncate_failed_write(struct inode *inode)
2571 truncate_inode_pages(inode->i_mapping, inode->i_size);
2572 reiserfs_truncate_file(inode, 0);
2575 static int reiserfs_write_begin(struct file *file,
2576 struct address_space *mapping,
2577 loff_t pos, unsigned len, unsigned flags,
2578 struct page **pagep, void **fsdata)
2580 struct inode *inode;
2586 inode = mapping->host;
2588 if (flags & AOP_FLAG_CONT_EXPAND &&
2589 (pos & (inode->i_sb->s_blocksize - 1)) == 0) {
2591 *fsdata = (void *)(unsigned long)flags;
2594 index = pos >> PAGE_CACHE_SHIFT;
2595 page = grab_cache_page_write_begin(mapping, index, flags);
2600 reiserfs_wait_on_write_block(inode->i_sb);
2601 fix_tail_page_for_writing(page);
2602 if (reiserfs_transaction_running(inode->i_sb)) {
2603 struct reiserfs_transaction_handle *th;
2604 th = (struct reiserfs_transaction_handle *)current->
2606 BUG_ON(!th->t_refcount);
2607 BUG_ON(!th->t_trans_id);
2608 old_ref = th->t_refcount;
2611 ret = __block_write_begin(page, pos, len, reiserfs_get_block);
2612 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2613 struct reiserfs_transaction_handle *th = current->journal_info;
2614 /* this gets a little ugly. If reiserfs_get_block returned an
2615 * error and left a transacstion running, we've got to close it,
2616 * and we've got to free handle if it was a persistent transaction.
2618 * But, if we had nested into an existing transaction, we need
2619 * to just drop the ref count on the handle.
2621 * If old_ref == 0, the transaction is from reiserfs_get_block,
2622 * and it was a persistent trans. Otherwise, it was nested above.
2624 if (th->t_refcount > old_ref) {
2629 reiserfs_write_lock(inode->i_sb);
2630 err = reiserfs_end_persistent_transaction(th);
2631 reiserfs_write_unlock(inode->i_sb);
2639 page_cache_release(page);
2640 /* Truncate allocated blocks */
2641 reiserfs_truncate_failed_write(inode);
2646 int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len)
2648 struct inode *inode = page->mapping->host;
2653 depth = reiserfs_write_unlock_nested(inode->i_sb);
2654 reiserfs_wait_on_write_block(inode->i_sb);
2655 reiserfs_write_lock_nested(inode->i_sb, depth);
2657 fix_tail_page_for_writing(page);
2658 if (reiserfs_transaction_running(inode->i_sb)) {
2659 struct reiserfs_transaction_handle *th;
2660 th = (struct reiserfs_transaction_handle *)current->
2662 BUG_ON(!th->t_refcount);
2663 BUG_ON(!th->t_trans_id);
2664 old_ref = th->t_refcount;
2668 ret = __block_write_begin(page, from, len, reiserfs_get_block);
2669 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2670 struct reiserfs_transaction_handle *th = current->journal_info;
2671 /* this gets a little ugly. If reiserfs_get_block returned an
2672 * error and left a transacstion running, we've got to close it,
2673 * and we've got to free handle if it was a persistent transaction.
2675 * But, if we had nested into an existing transaction, we need
2676 * to just drop the ref count on the handle.
2678 * If old_ref == 0, the transaction is from reiserfs_get_block,
2679 * and it was a persistent trans. Otherwise, it was nested above.
2681 if (th->t_refcount > old_ref) {
2686 reiserfs_write_lock(inode->i_sb);
2687 err = reiserfs_end_persistent_transaction(th);
2688 reiserfs_write_unlock(inode->i_sb);
2698 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2700 return generic_block_bmap(as, block, reiserfs_bmap);
2703 static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2704 loff_t pos, unsigned len, unsigned copied,
2705 struct page *page, void *fsdata)
2707 struct inode *inode = page->mapping->host;
2710 struct reiserfs_transaction_handle *th;
2712 bool locked = false;
2714 if ((unsigned long)fsdata & AOP_FLAG_CONT_EXPAND)
2717 reiserfs_wait_on_write_block(inode->i_sb);
2718 if (reiserfs_transaction_running(inode->i_sb))
2719 th = current->journal_info;
2723 start = pos & (PAGE_CACHE_SIZE - 1);
2724 if (unlikely(copied < len)) {
2725 if (!PageUptodate(page))
2728 page_zero_new_buffers(page, start + copied, start + len);
2730 flush_dcache_page(page);
2732 reiserfs_commit_page(inode, page, start, start + copied);
2734 /* generic_commit_write does this for us, but does not update the
2735 ** transaction tracking stuff when the size changes. So, we have
2736 ** to do the i_size updates here.
2738 if (pos + copied > inode->i_size) {
2739 struct reiserfs_transaction_handle myth;
2740 reiserfs_write_lock(inode->i_sb);
2742 /* If the file have grown beyond the border where it
2743 can have a tail, unmark it as needing a tail
2745 if ((have_large_tails(inode->i_sb)
2746 && inode->i_size > i_block_size(inode) * 4)
2747 || (have_small_tails(inode->i_sb)
2748 && inode->i_size > i_block_size(inode)))
2749 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2751 ret = journal_begin(&myth, inode->i_sb, 1);
2755 reiserfs_update_inode_transaction(inode);
2756 inode->i_size = pos + copied;
2758 * this will just nest into our transaction. It's important
2759 * to use mark_inode_dirty so the inode gets pushed around on the
2760 * dirty lists, and so that O_SYNC works as expected
2762 mark_inode_dirty(inode);
2763 reiserfs_update_sd(&myth, inode);
2765 ret = journal_end(&myth, inode->i_sb, 1);
2771 reiserfs_write_lock(inode->i_sb);
2775 mark_inode_dirty(inode);
2776 ret = reiserfs_end_persistent_transaction(th);
2783 reiserfs_write_unlock(inode->i_sb);
2785 page_cache_release(page);
2787 if (pos + len > inode->i_size)
2788 reiserfs_truncate_failed_write(inode);
2790 return ret == 0 ? copied : ret;
2793 reiserfs_write_unlock(inode->i_sb);
2797 reiserfs_update_sd(th, inode);
2798 ret = reiserfs_end_persistent_transaction(th);
2803 int reiserfs_commit_write(struct file *f, struct page *page,
2804 unsigned from, unsigned to)
2806 struct inode *inode = page->mapping->host;
2807 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2810 struct reiserfs_transaction_handle *th = NULL;
2813 depth = reiserfs_write_unlock_nested(inode->i_sb);
2814 reiserfs_wait_on_write_block(inode->i_sb);
2815 reiserfs_write_lock_nested(inode->i_sb, depth);
2817 if (reiserfs_transaction_running(inode->i_sb)) {
2818 th = current->journal_info;
2820 reiserfs_commit_page(inode, page, from, to);
2822 /* generic_commit_write does this for us, but does not update the
2823 ** transaction tracking stuff when the size changes. So, we have
2824 ** to do the i_size updates here.
2826 if (pos > inode->i_size) {
2827 struct reiserfs_transaction_handle myth;
2828 /* If the file have grown beyond the border where it
2829 can have a tail, unmark it as needing a tail
2831 if ((have_large_tails(inode->i_sb)
2832 && inode->i_size > i_block_size(inode) * 4)
2833 || (have_small_tails(inode->i_sb)
2834 && inode->i_size > i_block_size(inode)))
2835 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2837 ret = journal_begin(&myth, inode->i_sb, 1);
2841 reiserfs_update_inode_transaction(inode);
2842 inode->i_size = pos;
2844 * this will just nest into our transaction. It's important
2845 * to use mark_inode_dirty so the inode gets pushed around on the
2846 * dirty lists, and so that O_SYNC works as expected
2848 mark_inode_dirty(inode);
2849 reiserfs_update_sd(&myth, inode);
2851 ret = journal_end(&myth, inode->i_sb, 1);
2857 mark_inode_dirty(inode);
2858 ret = reiserfs_end_persistent_transaction(th);
2869 reiserfs_update_sd(th, inode);
2870 ret = reiserfs_end_persistent_transaction(th);
2876 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2878 if (reiserfs_attrs(inode->i_sb)) {
2879 if (sd_attrs & REISERFS_SYNC_FL)
2880 inode->i_flags |= S_SYNC;
2882 inode->i_flags &= ~S_SYNC;
2883 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2884 inode->i_flags |= S_IMMUTABLE;
2886 inode->i_flags &= ~S_IMMUTABLE;
2887 if (sd_attrs & REISERFS_APPEND_FL)
2888 inode->i_flags |= S_APPEND;
2890 inode->i_flags &= ~S_APPEND;
2891 if (sd_attrs & REISERFS_NOATIME_FL)
2892 inode->i_flags |= S_NOATIME;
2894 inode->i_flags &= ~S_NOATIME;
2895 if (sd_attrs & REISERFS_NOTAIL_FL)
2896 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2898 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2902 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
2904 if (reiserfs_attrs(inode->i_sb)) {
2905 if (inode->i_flags & S_IMMUTABLE)
2906 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2908 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2909 if (inode->i_flags & S_SYNC)
2910 *sd_attrs |= REISERFS_SYNC_FL;
2912 *sd_attrs &= ~REISERFS_SYNC_FL;
2913 if (inode->i_flags & S_NOATIME)
2914 *sd_attrs |= REISERFS_NOATIME_FL;
2916 *sd_attrs &= ~REISERFS_NOATIME_FL;
2917 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
2918 *sd_attrs |= REISERFS_NOTAIL_FL;
2920 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2924 /* decide if this buffer needs to stay around for data logging or ordered
2927 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2930 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2933 spin_lock(&j->j_dirty_buffers_lock);
2934 if (!buffer_mapped(bh)) {
2937 /* the page is locked, and the only places that log a data buffer
2938 * also lock the page.
2940 if (reiserfs_file_data_log(inode)) {
2942 * very conservative, leave the buffer pinned if
2943 * anyone might need it.
2945 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2948 } else if (buffer_dirty(bh)) {
2949 struct reiserfs_journal_list *jl;
2950 struct reiserfs_jh *jh = bh->b_private;
2952 /* why is this safe?
2953 * reiserfs_setattr updates i_size in the on disk
2954 * stat data before allowing vmtruncate to be called.
2956 * If buffer was put onto the ordered list for this
2957 * transaction, we know for sure either this transaction
2958 * or an older one already has updated i_size on disk,
2959 * and this ordered data won't be referenced in the file
2962 * if the buffer was put onto the ordered list for an older
2963 * transaction, we need to leave it around
2965 if (jh && (jl = jh->jl)
2966 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2970 if (ret && bh->b_private) {
2971 reiserfs_free_jh(bh);
2973 spin_unlock(&j->j_dirty_buffers_lock);
2978 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2979 static void reiserfs_invalidatepage(struct page *page, unsigned int offset,
2980 unsigned int length)
2982 struct buffer_head *head, *bh, *next;
2983 struct inode *inode = page->mapping->host;
2984 unsigned int curr_off = 0;
2985 unsigned int stop = offset + length;
2986 int partial_page = (offset || length < PAGE_CACHE_SIZE);
2989 BUG_ON(!PageLocked(page));
2992 ClearPageChecked(page);
2994 if (!page_has_buffers(page))
2997 head = page_buffers(page);
3000 unsigned int next_off = curr_off + bh->b_size;
3001 next = bh->b_this_page;
3003 if (next_off > stop)
3007 * is this block fully invalidated?
3009 if (offset <= curr_off) {
3010 if (invalidatepage_can_drop(inode, bh))
3011 reiserfs_unmap_buffer(bh);
3015 curr_off = next_off;
3017 } while (bh != head);
3020 * We release buffers only if the entire page is being invalidated.
3021 * The get_block cached value has been unconditionally invalidated,
3022 * so real IO is not possible anymore.
3024 if (!partial_page && ret) {
3025 ret = try_to_release_page(page, 0);
3026 /* maybe should BUG_ON(!ret); - neilb */
3032 static int reiserfs_set_page_dirty(struct page *page)
3034 struct inode *inode = page->mapping->host;
3035 if (reiserfs_file_data_log(inode)) {
3036 SetPageChecked(page);
3037 return __set_page_dirty_nobuffers(page);
3039 return __set_page_dirty_buffers(page);
3043 * Returns 1 if the page's buffers were dropped. The page is locked.
3045 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3046 * in the buffers at page_buffers(page).
3048 * even in -o notail mode, we can't be sure an old mount without -o notail
3049 * didn't create files with tails.
3051 static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
3053 struct inode *inode = page->mapping->host;
3054 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
3055 struct buffer_head *head;
3056 struct buffer_head *bh;
3059 WARN_ON(PageChecked(page));
3060 spin_lock(&j->j_dirty_buffers_lock);
3061 head = page_buffers(page);
3064 if (bh->b_private) {
3065 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
3066 reiserfs_free_jh(bh);
3072 bh = bh->b_this_page;
3073 } while (bh != head);
3075 ret = try_to_free_buffers(page);
3076 spin_unlock(&j->j_dirty_buffers_lock);
3080 /* We thank Mingming Cao for helping us understand in great detail what
3081 to do in this section of the code. */
3082 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
3083 const struct iovec *iov, loff_t offset,
3084 unsigned long nr_segs)
3086 struct file *file = iocb->ki_filp;
3087 struct inode *inode = file->f_mapping->host;
3090 ret = blockdev_direct_IO(rw, iocb, inode, iov, offset, nr_segs,
3091 reiserfs_get_blocks_direct_io);
3094 * In case of error extending write may have instantiated a few
3095 * blocks outside i_size. Trim these off again.
3097 if (unlikely((rw & WRITE) && ret < 0)) {
3098 loff_t isize = i_size_read(inode);
3099 loff_t end = offset + iov_length(iov, nr_segs);
3101 if ((end > isize) && inode_newsize_ok(inode, isize) == 0) {
3102 truncate_setsize(inode, isize);
3103 reiserfs_vfs_truncate_file(inode);
3110 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3112 struct inode *inode = dentry->d_inode;
3113 unsigned int ia_valid;
3116 error = inode_change_ok(inode, attr);
3120 /* must be turned off for recursive notify_change calls */
3121 ia_valid = attr->ia_valid &= ~(ATTR_KILL_SUID|ATTR_KILL_SGID);
3123 if (is_quota_modification(inode, attr))
3124 dquot_initialize(inode);
3125 reiserfs_write_lock(inode->i_sb);
3126 if (attr->ia_valid & ATTR_SIZE) {
3127 /* version 2 items will be caught by the s_maxbytes check
3128 ** done for us in vmtruncate
3130 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
3131 attr->ia_size > MAX_NON_LFS) {
3132 reiserfs_write_unlock(inode->i_sb);
3137 inode_dio_wait(inode);
3139 /* fill in hole pointers in the expanding truncate case. */
3140 if (attr->ia_size > inode->i_size) {
3141 error = generic_cont_expand_simple(inode, attr->ia_size);
3142 if (REISERFS_I(inode)->i_prealloc_count > 0) {
3144 struct reiserfs_transaction_handle th;
3145 /* we're changing at most 2 bitmaps, inode + super */
3146 err = journal_begin(&th, inode->i_sb, 4);
3148 reiserfs_discard_prealloc(&th, inode);
3149 err = journal_end(&th, inode->i_sb, 4);
3155 reiserfs_write_unlock(inode->i_sb);
3159 * file size is changed, ctime and mtime are
3162 attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
3165 reiserfs_write_unlock(inode->i_sb);
3167 if ((((attr->ia_valid & ATTR_UID) && (from_kuid(&init_user_ns, attr->ia_uid) & ~0xffff)) ||
3168 ((attr->ia_valid & ATTR_GID) && (from_kgid(&init_user_ns, attr->ia_gid) & ~0xffff))) &&
3169 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
3170 /* stat data of format v3.5 has 16 bit uid and gid */
3175 if ((ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
3176 (ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
3177 struct reiserfs_transaction_handle th;
3180 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3181 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3184 error = reiserfs_chown_xattrs(inode, attr);
3189 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3190 reiserfs_write_lock(inode->i_sb);
3191 error = journal_begin(&th, inode->i_sb, jbegin_count);
3192 reiserfs_write_unlock(inode->i_sb);
3195 error = dquot_transfer(inode, attr);
3196 reiserfs_write_lock(inode->i_sb);
3198 journal_end(&th, inode->i_sb, jbegin_count);
3199 reiserfs_write_unlock(inode->i_sb);
3203 /* Update corresponding info in inode so that everything is in
3204 * one transaction */
3205 if (attr->ia_valid & ATTR_UID)
3206 inode->i_uid = attr->ia_uid;
3207 if (attr->ia_valid & ATTR_GID)
3208 inode->i_gid = attr->ia_gid;
3209 mark_inode_dirty(inode);
3210 error = journal_end(&th, inode->i_sb, jbegin_count);
3211 reiserfs_write_unlock(inode->i_sb);
3216 if ((attr->ia_valid & ATTR_SIZE) &&
3217 attr->ia_size != i_size_read(inode)) {
3218 error = inode_newsize_ok(inode, attr->ia_size);
3220 truncate_setsize(inode, attr->ia_size);
3221 reiserfs_vfs_truncate_file(inode);
3226 setattr_copy(inode, attr);
3227 mark_inode_dirty(inode);
3230 if (!error && reiserfs_posixacl(inode->i_sb)) {
3231 if (attr->ia_valid & ATTR_MODE)
3232 error = reiserfs_acl_chmod(inode);
3239 const struct address_space_operations reiserfs_address_space_operations = {
3240 .writepage = reiserfs_writepage,
3241 .readpage = reiserfs_readpage,
3242 .readpages = reiserfs_readpages,
3243 .releasepage = reiserfs_releasepage,
3244 .invalidatepage = reiserfs_invalidatepage,
3245 .write_begin = reiserfs_write_begin,
3246 .write_end = reiserfs_write_end,
3247 .bmap = reiserfs_aop_bmap,
3248 .direct_IO = reiserfs_direct_IO,
3249 .set_page_dirty = reiserfs_set_page_dirty,