2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
5 #include <linux/config.h>
6 #include <linux/time.h>
8 #include <linux/reiserfs_fs.h>
9 #include <linux/reiserfs_acl.h>
10 #include <linux/reiserfs_xattr.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/vs_dlimit.h>
21 #include <linux/vserver/xid.h>
23 extern int reiserfs_default_io_size; /* default io size devuned in super.c */
25 static int reiserfs_commit_write(struct file *f, struct page *page,
26 unsigned from, unsigned to);
27 static int reiserfs_prepare_write(struct file *f, struct page *page,
28 unsigned from, unsigned to);
30 void reiserfs_delete_inode(struct inode *inode)
32 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
34 JOURNAL_PER_BALANCE_CNT * 2 +
35 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
36 struct reiserfs_transaction_handle th;
39 truncate_inode_pages(&inode->i_data, 0);
41 reiserfs_write_lock(inode->i_sb);
43 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
44 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
45 mutex_lock(&inode->i_mutex);
47 reiserfs_delete_xattrs(inode);
49 if (journal_begin(&th, inode->i_sb, jbegin_count)) {
50 mutex_unlock(&inode->i_mutex);
53 reiserfs_update_inode_transaction(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);
62 DLIMIT_FREE_INODE(inode);
64 if (journal_end(&th, inode->i_sb, jbegin_count)) {
65 mutex_unlock(&inode->i_mutex);
69 mutex_unlock(&inode->i_mutex);
71 /* check return value from reiserfs_delete_object after
72 * ending the transaction
77 /* all items of file are deleted, so we can remove "save" link */
78 remove_save_link(inode, 0 /* not truncate */ ); /* we can't do anything
79 * about an error here */
81 /* no object items are in the tree */
85 clear_inode(inode); /* note this must go after the journal_end to prevent deadlock */
87 reiserfs_write_unlock(inode->i_sb);
90 static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
91 __u32 objectid, loff_t offset, int type, int length)
93 key->version = version;
95 key->on_disk_key.k_dir_id = dirid;
96 key->on_disk_key.k_objectid = objectid;
97 set_cpu_key_k_offset(key, offset);
98 set_cpu_key_k_type(key, type);
99 key->key_length = length;
102 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
103 offset and type of key */
104 void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
105 int type, int length)
107 _make_cpu_key(key, get_inode_item_key_version(inode),
108 le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
109 le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
114 // when key is 0, do not set version and short key
116 inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
118 loff_t offset, int type, int length,
119 int entry_count /*or ih_free_space */ )
122 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
123 ih->ih_key.k_objectid =
124 cpu_to_le32(key->on_disk_key.k_objectid);
126 put_ih_version(ih, version);
127 set_le_ih_k_offset(ih, offset);
128 set_le_ih_k_type(ih, type);
129 put_ih_item_len(ih, length);
130 /* set_ih_free_space (ih, 0); */
131 // for directory items it is entry count, for directs and stat
132 // datas - 0xffff, for indirects - 0
133 put_ih_entry_count(ih, entry_count);
137 // FIXME: we might cache recently accessed indirect item
139 // Ugh. Not too eager for that....
140 // I cut the code until such time as I see a convincing argument (benchmark).
141 // I don't want a bloated inode struct..., and I don't like code complexity....
143 /* cutting the code is fine, since it really isn't in use yet and is easy
144 ** to add back in. But, Vladimir has a really good idea here. Think
145 ** about what happens for reading a file. For each page,
146 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
147 ** an indirect item. This indirect item has X number of pointers, where
148 ** X is a big number if we've done the block allocation right. But,
149 ** we only use one or two of these pointers during each call to readpage,
150 ** needlessly researching again later on.
152 ** The size of the cache could be dynamic based on the size of the file.
154 ** I'd also like to see us cache the location the stat data item, since
155 ** we are needlessly researching for that frequently.
160 /* If this page has a file tail in it, and
161 ** it was read in by get_block_create_0, the page data is valid,
162 ** but tail is still sitting in a direct item, and we can't write to
163 ** it. So, look through this page, and check all the mapped buffers
164 ** to make sure they have valid block numbers. Any that don't need
165 ** to be unmapped, so that block_prepare_write will correctly call
166 ** reiserfs_get_block to convert the tail into an unformatted node
168 static inline void fix_tail_page_for_writing(struct page *page)
170 struct buffer_head *head, *next, *bh;
172 if (page && page_has_buffers(page)) {
173 head = page_buffers(page);
176 next = bh->b_this_page;
177 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
178 reiserfs_unmap_buffer(bh);
181 } while (bh != head);
185 /* reiserfs_get_block does not need to allocate a block only if it has been
186 done already or non-hole position has been found in the indirect item */
187 static inline int allocation_needed(int retval, b_blocknr_t allocated,
188 struct item_head *ih,
189 __le32 * item, int pos_in_item)
193 if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
194 get_block_num(item, pos_in_item))
199 static inline int indirect_item_found(int retval, struct item_head *ih)
201 return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
204 static inline void set_block_dev_mapped(struct buffer_head *bh,
205 b_blocknr_t block, struct inode *inode)
207 map_bh(bh, inode->i_sb, block);
211 // files which were created in the earlier version can not be longer,
214 static int file_capable(struct inode *inode, long block)
216 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file.
217 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
223 /*static*/ int restart_transaction(struct reiserfs_transaction_handle *th,
224 struct inode *inode, struct path *path)
226 struct super_block *s = th->t_super;
227 int len = th->t_blocks_allocated;
230 BUG_ON(!th->t_trans_id);
231 BUG_ON(!th->t_refcount);
233 /* we cannot restart while nested */
234 if (th->t_refcount > 1) {
238 reiserfs_update_sd(th, inode);
239 err = journal_end(th, s, len);
241 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
243 reiserfs_update_inode_transaction(inode);
248 // it is called by get_block when create == 0. Returns block number
249 // for 'block'-th logical block of file. When it hits direct item it
250 // returns 0 (being called from bmap) or read direct item into piece
251 // of page (bh_result)
253 // Please improve the english/clarity in the comment above, as it is
254 // hard to understand.
256 static int _get_block_create_0(struct inode *inode, long block,
257 struct buffer_head *bh_result, int args)
259 INITIALIZE_PATH(path);
261 struct buffer_head *bh;
262 struct item_head *ih, tmp_ih;
270 unsigned long offset;
272 // prepare the key to look for the 'block'-th block of file
273 make_cpu_key(&key, inode,
274 (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 fs_gen = get_generation(inode->i_sb);
353 copy_item_head(&tmp_ih, ih);
355 /* we only want to kmap if we are reading the tail into the page.
356 ** this is not the common case, so we don't kmap until we are
357 ** sure we need to. But, this means the item might move if
361 p = (char *)kmap(bh_result->b_page);
362 if (fs_changed(fs_gen, inode->i_sb)
363 && item_moved(&tmp_ih, &path)) {
368 memset(p, 0, inode->i_sb->s_blocksize);
370 if (!is_direct_le_ih(ih)) {
373 /* make sure we don't read more bytes than actually exist in
374 ** the file. This can happen in odd cases where i_size isn't
375 ** correct, and when direct item padding results in a few
376 ** extra bytes at the end of the direct item
378 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
380 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
382 inode->i_size - (le_ih_k_offset(ih) - 1) -
386 chars = ih_item_len(ih) - path.pos_in_item;
388 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
395 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
396 // we done, if read direct item is not the last item of
397 // node FIXME: we could try to check right delimiting key
398 // to see whether direct item continues in the right
399 // neighbor or rely on i_size
402 // update key to look for the next piece
403 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
404 result = search_for_position_by_key(inode->i_sb, &key, &path);
405 if (result != POSITION_FOUND)
406 // i/o error most likely
408 bh = get_last_bh(&path);
412 flush_dcache_page(bh_result->b_page);
413 kunmap(bh_result->b_page);
418 if (result == IO_ERROR)
421 /* this buffer has valid data, but isn't valid for io. mapping it to
422 * block #0 tells the rest of reiserfs it just has a tail in it
424 map_bh(bh_result, inode->i_sb, 0);
425 set_buffer_uptodate(bh_result);
429 // this is called to create file map. So, _get_block_create_0 will not
431 static int reiserfs_bmap(struct inode *inode, sector_t block,
432 struct buffer_head *bh_result, int create)
434 if (!file_capable(inode, block))
437 reiserfs_write_lock(inode->i_sb);
438 /* do not read the direct item */
439 _get_block_create_0(inode, block, bh_result, 0);
440 reiserfs_write_unlock(inode->i_sb);
444 /* special version of get_block that is only used by grab_tail_page right
445 ** now. It is sent to block_prepare_write, and when you try to get a
446 ** block past the end of the file (or a block from a hole) it returns
447 ** -ENOENT instead of a valid buffer. block_prepare_write expects to
448 ** be able to do i/o on the buffers returned, unless an error value
451 ** So, this allows block_prepare_write to be used for reading a single block
452 ** in a page. Where it does not produce a valid page for holes, or past the
453 ** end of the file. This turns out to be exactly what we need for reading
454 ** tails for conversion.
456 ** The point of the wrapper is forcing a certain value for create, even
457 ** though the VFS layer is calling this function with create==1. If you
458 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
459 ** don't use this function.
461 static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
462 struct buffer_head *bh_result,
465 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
468 /* This is special helper for reiserfs_get_block in case we are executing
469 direct_IO request. */
470 static int reiserfs_get_blocks_direct_io(struct inode *inode,
472 unsigned long max_blocks,
473 struct buffer_head *bh_result,
478 bh_result->b_page = NULL;
480 /* We set the b_size before reiserfs_get_block call since it is
481 referenced in convert_tail_for_hole() that may be called from
482 reiserfs_get_block() */
483 bh_result->b_size = (1 << inode->i_blkbits);
485 ret = reiserfs_get_block(inode, iblock, bh_result,
486 create | GET_BLOCK_NO_DANGLE);
490 /* don't allow direct io onto tail pages */
491 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
492 /* make sure future calls to the direct io funcs for this offset
493 ** in the file fail by unmapping the buffer
495 clear_buffer_mapped(bh_result);
498 /* Possible unpacked tail. Flush the data before pages have
500 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
503 err = reiserfs_commit_for_inode(inode);
504 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
514 ** helper function for when reiserfs_get_block is called for a hole
515 ** but the file tail is still in a direct item
516 ** bh_result is the buffer head for the hole
517 ** tail_offset is the offset of the start of the tail in the file
519 ** This calls prepare_write, which will start a new transaction
520 ** you should not be in a transaction, or have any paths held when you
523 static int convert_tail_for_hole(struct inode *inode,
524 struct buffer_head *bh_result,
528 unsigned long tail_end;
529 unsigned long tail_start;
530 struct page *tail_page;
531 struct page *hole_page = bh_result->b_page;
534 if ((tail_offset & (bh_result->b_size - 1)) != 1)
537 /* always try to read until the end of the block */
538 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
539 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
541 index = tail_offset >> PAGE_CACHE_SHIFT;
542 /* hole_page can be zero in case of direct_io, we are sure
543 that we cannot get here if we write with O_DIRECT into
545 if (!hole_page || index != hole_page->index) {
546 tail_page = grab_cache_page(inode->i_mapping, index);
552 tail_page = hole_page;
555 /* we don't have to make sure the conversion did not happen while
556 ** we were locking the page because anyone that could convert
557 ** must first take i_mutex.
559 ** We must fix the tail page for writing because it might have buffers
560 ** that are mapped, but have a block number of 0. This indicates tail
561 ** data that has been read directly into the page, and block_prepare_write
562 ** won't trigger a get_block in this case.
564 fix_tail_page_for_writing(tail_page);
565 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
569 /* tail conversion might change the data in the page */
570 flush_dcache_page(tail_page);
572 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
575 if (tail_page != hole_page) {
576 unlock_page(tail_page);
577 page_cache_release(tail_page);
583 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
586 b_blocknr_t * allocated_block_nr,
587 struct path *path, int flags)
589 BUG_ON(!th->t_trans_id);
591 #ifdef REISERFS_PREALLOCATE
592 if (!(flags & GET_BLOCK_NO_IMUX)) {
593 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
597 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
601 int reiserfs_get_block(struct inode *inode, sector_t block,
602 struct buffer_head *bh_result, int create)
604 int repeat, retval = 0;
605 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
606 INITIALIZE_PATH(path);
609 struct buffer_head *bh, *unbh = NULL;
610 struct item_head *ih, tmp_ih;
614 struct reiserfs_transaction_handle *th = NULL;
615 /* space reserved in transaction batch:
616 . 3 balancings in direct->indirect conversion
617 . 1 block involved into reiserfs_update_sd()
618 XXX in practically impossible worst case direct2indirect()
619 can incur (much) more than 3 balancings.
620 quota update for user, group */
622 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
623 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
627 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
630 reiserfs_write_lock(inode->i_sb);
631 version = get_inode_item_key_version(inode);
633 if (!file_capable(inode, block)) {
634 reiserfs_write_unlock(inode->i_sb);
638 /* if !create, we aren't changing the FS, so we don't need to
639 ** log anything, so we don't need to start a transaction
641 if (!(create & GET_BLOCK_CREATE)) {
643 /* find number of block-th logical block of the file */
644 ret = _get_block_create_0(inode, block, bh_result,
645 create | GET_BLOCK_READ_DIRECT);
646 reiserfs_write_unlock(inode->i_sb);
650 * if we're already in a transaction, make sure to close
651 * any new transactions we start in this func
653 if ((create & GET_BLOCK_NO_DANGLE) ||
654 reiserfs_transaction_running(inode->i_sb))
657 /* If file is of such a size, that it might have a tail and tails are enabled
658 ** we should mark it as possibly needing tail packing on close
660 if ((have_large_tails(inode->i_sb)
661 && inode->i_size < i_block_size(inode) * 4)
662 || (have_small_tails(inode->i_sb)
663 && inode->i_size < i_block_size(inode)))
664 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
666 /* set the key of the first byte in the 'block'-th block of file */
667 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
668 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
670 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
675 reiserfs_update_inode_transaction(inode);
679 retval = search_for_position_by_key(inode->i_sb, &key, &path);
680 if (retval == IO_ERROR) {
685 bh = get_last_bh(&path);
687 item = get_item(&path);
688 pos_in_item = path.pos_in_item;
690 fs_gen = get_generation(inode->i_sb);
691 copy_item_head(&tmp_ih, ih);
693 if (allocation_needed
694 (retval, allocated_block_nr, ih, item, pos_in_item)) {
695 /* we have to allocate block for the unformatted node */
702 _allocate_block(th, block, inode, &allocated_block_nr,
705 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
706 /* restart the transaction to give the journal a chance to free
707 ** some blocks. releases the path, so we have to go back to
708 ** research if we succeed on the second try
710 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
711 retval = restart_transaction(th, inode, &path);
715 _allocate_block(th, block, inode,
716 &allocated_block_nr, NULL, create);
718 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
721 if (repeat == QUOTA_EXCEEDED)
728 if (fs_changed(fs_gen, inode->i_sb)
729 && item_moved(&tmp_ih, &path)) {
734 if (indirect_item_found(retval, ih)) {
735 b_blocknr_t unfm_ptr;
736 /* 'block'-th block is in the file already (there is
737 corresponding cell in some indirect item). But it may be
738 zero unformatted node pointer (hole) */
739 unfm_ptr = get_block_num(item, pos_in_item);
741 /* use allocated block to plug the hole */
742 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
743 if (fs_changed(fs_gen, inode->i_sb)
744 && item_moved(&tmp_ih, &path)) {
745 reiserfs_restore_prepared_buffer(inode->i_sb,
749 set_buffer_new(bh_result);
750 if (buffer_dirty(bh_result)
751 && reiserfs_data_ordered(inode->i_sb))
752 reiserfs_add_ordered_list(inode, bh_result);
753 put_block_num(item, pos_in_item, allocated_block_nr);
754 unfm_ptr = allocated_block_nr;
755 journal_mark_dirty(th, inode->i_sb, bh);
756 reiserfs_update_sd(th, inode);
758 set_block_dev_mapped(bh_result, unfm_ptr, inode);
762 retval = reiserfs_end_persistent_transaction(th);
764 reiserfs_write_unlock(inode->i_sb);
766 /* the item was found, so new blocks were not added to the file
767 ** there is no need to make sure the inode is updated with this
778 /* desired position is not found or is in the direct item. We have
779 to append file with holes up to 'block'-th block converting
780 direct items to indirect one if necessary */
783 if (is_statdata_le_ih(ih)) {
785 struct cpu_key tmp_key;
787 /* indirect item has to be inserted */
788 make_le_item_head(&tmp_ih, &key, version, 1,
789 TYPE_INDIRECT, UNFM_P_SIZE,
790 0 /* free_space */ );
792 if (cpu_key_k_offset(&key) == 1) {
793 /* we are going to add 'block'-th block to the file. Use
794 allocated block for that */
795 unp = cpu_to_le32(allocated_block_nr);
796 set_block_dev_mapped(bh_result,
797 allocated_block_nr, inode);
798 set_buffer_new(bh_result);
802 set_cpu_key_k_offset(&tmp_key, 1);
803 PATH_LAST_POSITION(&path)++;
806 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
807 inode, (char *)&unp);
809 reiserfs_free_block(th, inode,
810 allocated_block_nr, 1);
811 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
813 //mark_tail_converted (inode);
814 } else if (is_direct_le_ih(ih)) {
815 /* direct item has to be converted */
819 ((le_ih_k_offset(ih) -
820 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
821 if (tail_offset == cpu_key_k_offset(&key)) {
822 /* direct item we just found fits into block we have
823 to map. Convert it into unformatted node: use
824 bh_result for the conversion */
825 set_block_dev_mapped(bh_result,
826 allocated_block_nr, inode);
830 /* we have to padd file tail stored in direct item(s)
831 up to block size and convert it to unformatted
832 node. FIXME: this should also get into page cache */
836 * ugly, but we can only end the transaction if
839 BUG_ON(!th->t_refcount);
840 if (th->t_refcount == 1) {
842 reiserfs_end_persistent_transaction
850 convert_tail_for_hole(inode, bh_result,
853 if (retval != -ENOSPC)
854 reiserfs_warning(inode->i_sb,
855 "clm-6004: convert tail failed 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 = kmalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_ATOMIC); // We need to avoid scheduling.
953 UNFM_P_SIZE * min(blocks_needed,
956 if (blocks_needed <= max_to_insert) {
957 /* we are going to add target block to the file. Use allocated
959 un[blocks_needed - 1] =
960 cpu_to_le32(allocated_block_nr);
961 set_block_dev_mapped(bh_result,
962 allocated_block_nr, inode);
963 set_buffer_new(bh_result);
966 /* paste hole to the indirect item */
967 /* If kmalloc failed, max_to_insert becomes zero and it means we
968 only have space for one block */
970 max_to_insert ? max_to_insert : 1;
973 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
978 if (blocks_needed != 1)
982 reiserfs_free_block(th, inode,
983 allocated_block_nr, 1);
987 /* We need to mark new file size in case this function will be
988 interrupted/aborted later on. And we may do this only for
991 inode->i_sb->s_blocksize * blocks_needed;
998 /* this loop could log more blocks than we had originally asked
999 ** for. So, we have to allow the transaction to end if it is
1000 ** too big or too full. Update the inode so things are
1001 ** consistent if we crash before the function returns
1003 ** release the path so that anybody waiting on the path before
1004 ** ending their transaction will be able to continue.
1006 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
1007 retval = restart_transaction(th, inode, &path);
1011 /* inserting indirect pointers for a hole can take a
1012 ** long time. reschedule if needed
1016 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1017 if (retval == IO_ERROR) {
1021 if (retval == POSITION_FOUND) {
1022 reiserfs_warning(inode->i_sb,
1023 "vs-825: reiserfs_get_block: "
1024 "%K should not be found", &key);
1026 if (allocated_block_nr)
1027 reiserfs_free_block(th, inode,
1028 allocated_block_nr, 1);
1032 bh = get_last_bh(&path);
1034 item = get_item(&path);
1035 pos_in_item = path.pos_in_item;
1041 if (th && (!dangle || (retval && !th->t_trans_id))) {
1044 reiserfs_update_sd(th, inode);
1045 err = reiserfs_end_persistent_transaction(th);
1050 reiserfs_write_unlock(inode->i_sb);
1051 reiserfs_check_path(&path);
1056 reiserfs_readpages(struct file *file, struct address_space *mapping,
1057 struct list_head *pages, unsigned nr_pages)
1059 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
1062 /* Compute real number of used bytes by file
1063 * Following three functions can go away when we'll have enough space in stat item
1065 static int real_space_diff(struct inode *inode, int sd_size)
1068 loff_t blocksize = inode->i_sb->s_blocksize;
1070 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1073 /* End of file is also in full block with indirect reference, so round
1074 ** up to the next block.
1076 ** there is just no way to know if the tail is actually packed
1077 ** on the file, so we have to assume it isn't. When we pack the
1078 ** tail, we add 4 bytes to pretend there really is an unformatted
1083 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1088 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
1091 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1092 return inode->i_size +
1093 (loff_t) (real_space_diff(inode, sd_size));
1095 return ((loff_t) real_space_diff(inode, sd_size)) +
1096 (((loff_t) blocks) << 9);
1099 /* Compute number of blocks used by file in ReiserFS counting */
1100 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1102 loff_t bytes = inode_get_bytes(inode);
1103 loff_t real_space = real_space_diff(inode, sd_size);
1105 /* keeps fsck and non-quota versions of reiserfs happy */
1106 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1107 bytes += (loff_t) 511;
1110 /* files from before the quota patch might i_blocks such that
1111 ** bytes < real_space. Deal with that here to prevent it from
1114 if (bytes < real_space)
1116 return (bytes - real_space) >> 9;
1120 // BAD: new directories have stat data of new type and all other items
1121 // of old type. Version stored in the inode says about body items, so
1122 // in update_stat_data we can not rely on inode, but have to check
1123 // item version directly
1126 // called by read_locked_inode
1127 static void init_inode(struct inode *inode, struct path *path)
1129 struct buffer_head *bh;
1130 struct item_head *ih;
1134 //int version = ITEM_VERSION_1;
1136 bh = PATH_PLAST_BUFFER(path);
1137 ih = PATH_PITEM_HEAD(path);
1139 copy_key(INODE_PKEY(inode), &(ih->ih_key));
1140 inode->i_blksize = reiserfs_default_io_size;
1142 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1143 REISERFS_I(inode)->i_flags = 0;
1144 REISERFS_I(inode)->i_prealloc_block = 0;
1145 REISERFS_I(inode)->i_prealloc_count = 0;
1146 REISERFS_I(inode)->i_trans_id = 0;
1147 REISERFS_I(inode)->i_jl = NULL;
1148 REISERFS_I(inode)->i_acl_access = NULL;
1149 REISERFS_I(inode)->i_acl_default = NULL;
1150 init_rwsem(&REISERFS_I(inode)->xattr_sem);
1152 if (stat_data_v1(ih)) {
1153 struct stat_data_v1 *sd =
1154 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1155 unsigned long blocks;
1157 uid = sd_v1_uid(sd);
1158 gid = sd_v1_gid(sd);
1160 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1161 set_inode_sd_version(inode, STAT_DATA_V1);
1162 inode->i_mode = sd_v1_mode(sd);
1163 inode->i_nlink = sd_v1_nlink(sd);
1164 inode->i_size = sd_v1_size(sd);
1165 inode->i_atime.tv_sec = sd_v1_atime(sd);
1166 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1167 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1168 inode->i_atime.tv_nsec = 0;
1169 inode->i_ctime.tv_nsec = 0;
1170 inode->i_mtime.tv_nsec = 0;
1172 inode->i_blocks = sd_v1_blocks(sd);
1173 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1174 blocks = (inode->i_size + 511) >> 9;
1175 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1176 if (inode->i_blocks > blocks) {
1177 // there was a bug in <=3.5.23 when i_blocks could take negative
1178 // values. Starting from 3.5.17 this value could even be stored in
1179 // stat data. For such files we set i_blocks based on file
1180 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1181 // only updated if file's inode will ever change
1182 inode->i_blocks = blocks;
1185 rdev = sd_v1_rdev(sd);
1186 REISERFS_I(inode)->i_first_direct_byte =
1187 sd_v1_first_direct_byte(sd);
1188 /* an early bug in the quota code can give us an odd number for the
1189 ** block count. This is incorrect, fix it here.
1191 if (inode->i_blocks & 1) {
1194 inode_set_bytes(inode,
1195 to_real_used_space(inode, inode->i_blocks,
1197 /* nopack is initially zero for v1 objects. For v2 objects,
1198 nopack is initialised from sd_attrs */
1199 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1201 // new stat data found, but object may have old items
1202 // (directories and symlinks)
1203 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1205 uid = sd_v2_uid(sd);
1206 gid = sd_v2_gid(sd);
1208 inode->i_mode = sd_v2_mode(sd);
1209 inode->i_nlink = sd_v2_nlink(sd);
1210 inode->i_size = sd_v2_size(sd);
1211 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1212 inode->i_atime.tv_sec = sd_v2_atime(sd);
1213 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1214 inode->i_ctime.tv_nsec = 0;
1215 inode->i_mtime.tv_nsec = 0;
1216 inode->i_atime.tv_nsec = 0;
1217 inode->i_blocks = sd_v2_blocks(sd);
1218 rdev = sd_v2_rdev(sd);
1219 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1220 inode->i_generation =
1221 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1223 inode->i_generation = sd_v2_generation(sd);
1225 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1226 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1228 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1229 REISERFS_I(inode)->i_first_direct_byte = 0;
1230 set_inode_sd_version(inode, STAT_DATA_V2);
1231 inode_set_bytes(inode,
1232 to_real_used_space(inode, inode->i_blocks,
1234 /* read persistent inode attributes from sd and initalise
1235 generic inode flags from them */
1236 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1237 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
1240 inode->i_uid = INOXID_UID(XID_TAG(inode), uid, gid);
1241 inode->i_gid = INOXID_GID(XID_TAG(inode), uid, gid);
1242 inode->i_xid = INOXID_XID(XID_TAG(inode), uid, gid, 0);
1245 if (S_ISREG(inode->i_mode)) {
1246 inode->i_op = &reiserfs_file_inode_operations;
1247 inode->i_fop = &reiserfs_file_operations;
1248 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1249 } else if (S_ISDIR(inode->i_mode)) {
1250 inode->i_op = &reiserfs_dir_inode_operations;
1251 inode->i_fop = &reiserfs_dir_operations;
1252 } else if (S_ISLNK(inode->i_mode)) {
1253 inode->i_op = &reiserfs_symlink_inode_operations;
1254 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1256 inode->i_blocks = 0;
1257 inode->i_op = &reiserfs_special_inode_operations;
1258 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1262 // update new stat data with inode fields
1263 static void inode2sd(void *sd, struct inode *inode, loff_t size)
1265 struct stat_data *sd_v2 = (struct stat_data *)sd;
1266 uid_t uid = XIDINO_UID(XID_TAG(inode), inode->i_uid, inode->i_xid);
1267 gid_t gid = XIDINO_GID(XID_TAG(inode), inode->i_gid, inode->i_xid);
1270 set_sd_v2_uid(sd_v2, uid);
1271 set_sd_v2_gid(sd_v2, gid);
1272 set_sd_v2_mode(sd_v2, inode->i_mode);
1273 set_sd_v2_nlink(sd_v2, inode->i_nlink);
1274 set_sd_v2_size(sd_v2, size);
1275 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1276 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1277 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1278 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1279 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1280 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1282 set_sd_v2_generation(sd_v2, inode->i_generation);
1283 flags = REISERFS_I(inode)->i_attrs;
1284 i_attrs_to_sd_attrs(inode, &flags);
1285 set_sd_v2_attrs(sd_v2, flags);
1288 // used to copy inode's fields to old stat data
1289 static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
1291 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
1293 set_sd_v1_mode(sd_v1, inode->i_mode);
1294 set_sd_v1_uid(sd_v1, inode->i_uid);
1295 set_sd_v1_gid(sd_v1, inode->i_gid);
1296 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1297 set_sd_v1_size(sd_v1, size);
1298 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1299 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1300 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
1302 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1303 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1305 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1307 // Sigh. i_first_direct_byte is back
1308 set_sd_v1_first_direct_byte(sd_v1,
1309 REISERFS_I(inode)->i_first_direct_byte);
1312 /* NOTE, you must prepare the buffer head before sending it here,
1313 ** and then log it after the call
1315 static void update_stat_data(struct path *path, struct inode *inode,
1318 struct buffer_head *bh;
1319 struct item_head *ih;
1321 bh = PATH_PLAST_BUFFER(path);
1322 ih = PATH_PITEM_HEAD(path);
1324 if (!is_statdata_le_ih(ih))
1325 reiserfs_panic(inode->i_sb,
1326 "vs-13065: update_stat_data: key %k, found item %h",
1327 INODE_PKEY(inode), ih);
1329 if (stat_data_v1(ih)) {
1330 // path points to old stat data
1331 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1333 inode2sd(B_I_PITEM(bh, ih), inode, size);
1339 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1340 struct inode *inode, loff_t size)
1343 INITIALIZE_PATH(path);
1344 struct buffer_head *bh;
1346 struct item_head *ih, tmp_ih;
1349 BUG_ON(!th->t_trans_id);
1351 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant
1355 /* look for the object's stat data */
1356 retval = search_item(inode->i_sb, &key, &path);
1357 if (retval == IO_ERROR) {
1358 reiserfs_warning(inode->i_sb,
1359 "vs-13050: reiserfs_update_sd: "
1360 "i/o failure occurred trying to update %K stat data",
1364 if (retval == ITEM_NOT_FOUND) {
1365 pos = PATH_LAST_POSITION(&path);
1367 if (inode->i_nlink == 0) {
1368 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1371 reiserfs_warning(inode->i_sb,
1372 "vs-13060: reiserfs_update_sd: "
1373 "stat data of object %k (nlink == %d) not found (pos %d)",
1374 INODE_PKEY(inode), inode->i_nlink,
1376 reiserfs_check_path(&path);
1380 /* sigh, prepare_for_journal might schedule. When it schedules the
1381 ** FS might change. We have to detect that, and loop back to the
1382 ** search if the stat data item has moved
1384 bh = get_last_bh(&path);
1386 copy_item_head(&tmp_ih, ih);
1387 fs_gen = get_generation(inode->i_sb);
1388 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1389 if (fs_changed(fs_gen, inode->i_sb)
1390 && item_moved(&tmp_ih, &path)) {
1391 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1392 continue; /* Stat_data item has been moved after scheduling. */
1396 update_stat_data(&path, inode, size);
1397 journal_mark_dirty(th, th->t_super, bh);
1402 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1403 ** does a make_bad_inode when things go wrong. But, we need to make sure
1404 ** and clear the key in the private portion of the inode, otherwise a
1405 ** corresponding iput might try to delete whatever object the inode last
1408 static void reiserfs_make_bad_inode(struct inode *inode)
1410 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1411 make_bad_inode(inode);
1415 // initially this function was derived from minix or ext2's analog and
1416 // evolved as the prototype did
1419 int reiserfs_init_locked_inode(struct inode *inode, void *p)
1421 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1422 inode->i_ino = args->objectid;
1423 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1427 /* looks for stat data in the tree, and fills up the fields of in-core
1428 inode stat data fields */
1429 void reiserfs_read_locked_inode(struct inode *inode,
1430 struct reiserfs_iget_args *args)
1432 INITIALIZE_PATH(path_to_sd);
1434 unsigned long dirino;
1437 dirino = args->dirid;
1439 /* set version 1, version 2 could be used too, because stat data
1440 key is the same in both versions */
1441 key.version = KEY_FORMAT_3_5;
1442 key.on_disk_key.k_dir_id = dirino;
1443 key.on_disk_key.k_objectid = inode->i_ino;
1444 key.on_disk_key.k_offset = 0;
1445 key.on_disk_key.k_type = 0;
1447 /* look for the object's stat data */
1448 retval = search_item(inode->i_sb, &key, &path_to_sd);
1449 if (retval == IO_ERROR) {
1450 reiserfs_warning(inode->i_sb,
1451 "vs-13070: reiserfs_read_locked_inode: "
1452 "i/o failure occurred trying to find stat data of %K",
1454 reiserfs_make_bad_inode(inode);
1457 if (retval != ITEM_FOUND) {
1458 /* a stale NFS handle can trigger this without it being an error */
1459 pathrelse(&path_to_sd);
1460 reiserfs_make_bad_inode(inode);
1465 init_inode(inode, &path_to_sd);
1467 /* It is possible that knfsd is trying to access inode of a file
1468 that is being removed from the disk by some other thread. As we
1469 update sd on unlink all that is required is to check for nlink
1470 here. This bug was first found by Sizif when debugging
1471 SquidNG/Butterfly, forgotten, and found again after Philippe
1472 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1474 More logical fix would require changes in fs/inode.c:iput() to
1475 remove inode from hash-table _after_ fs cleaned disk stuff up and
1476 in iget() to return NULL if I_FREEING inode is found in
1478 /* Currently there is one place where it's ok to meet inode with
1479 nlink==0: processing of open-unlinked and half-truncated files
1480 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1481 if ((inode->i_nlink == 0) &&
1482 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
1483 reiserfs_warning(inode->i_sb,
1484 "vs-13075: reiserfs_read_locked_inode: "
1485 "dead inode read from disk %K. "
1486 "This is likely to be race with knfsd. Ignore",
1488 reiserfs_make_bad_inode(inode);
1491 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
1496 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1498 * @inode: inode from hash table to check
1499 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1501 * This function is called by iget5_locked() to distinguish reiserfs inodes
1502 * having the same inode numbers. Such inodes can only exist due to some
1503 * error condition. One of them should be bad. Inodes with identical
1504 * inode numbers (objectids) are distinguished by parent directory ids.
1507 int reiserfs_find_actor(struct inode *inode, void *opaque)
1509 struct reiserfs_iget_args *args;
1512 /* args is already in CPU order */
1513 return (inode->i_ino == args->objectid) &&
1514 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1517 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
1519 struct inode *inode;
1520 struct reiserfs_iget_args args;
1522 args.objectid = key->on_disk_key.k_objectid;
1523 args.dirid = key->on_disk_key.k_dir_id;
1524 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1525 reiserfs_find_actor, reiserfs_init_locked_inode,
1528 return ERR_PTR(-ENOMEM);
1530 if (inode->i_state & I_NEW) {
1531 reiserfs_read_locked_inode(inode, &args);
1532 unlock_new_inode(inode);
1535 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1536 /* either due to i/o error or a stale NFS handle */
1543 struct dentry *reiserfs_get_dentry(struct super_block *sb, void *vobjp)
1545 __u32 *data = vobjp;
1547 struct dentry *result;
1548 struct inode *inode;
1550 key.on_disk_key.k_objectid = data[0];
1551 key.on_disk_key.k_dir_id = data[1];
1552 reiserfs_write_lock(sb);
1553 inode = reiserfs_iget(sb, &key);
1554 if (inode && !IS_ERR(inode) && data[2] != 0 &&
1555 data[2] != inode->i_generation) {
1559 reiserfs_write_unlock(sb);
1561 inode = ERR_PTR(-ESTALE);
1563 return ERR_PTR(PTR_ERR(inode));
1564 result = d_alloc_anon(inode);
1567 return ERR_PTR(-ENOMEM);
1572 struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 * data,
1573 int len, int fhtype,
1574 int (*acceptable) (void *contect,
1575 struct dentry * de),
1578 __u32 obj[3], parent[3];
1580 /* fhtype happens to reflect the number of u32s encoded.
1581 * due to a bug in earlier code, fhtype might indicate there
1582 * are more u32s then actually fitted.
1583 * so if fhtype seems to be more than len, reduce fhtype.
1585 * 2 - objectid + dir_id - legacy support
1586 * 3 - objectid + dir_id + generation
1587 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1588 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1589 * 6 - as above plus generation of directory
1590 * 6 does not fit in NFSv2 handles
1593 if (fhtype != 6 || len != 5)
1594 reiserfs_warning(sb,
1595 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1602 if (fhtype == 3 || fhtype >= 5)
1605 obj[2] = 0; /* generation number */
1608 parent[0] = data[fhtype >= 5 ? 3 : 2];
1609 parent[1] = data[fhtype >= 5 ? 4 : 3];
1611 parent[2] = data[5];
1615 return sb->s_export_op->find_exported_dentry(sb, obj,
1616 fhtype < 4 ? NULL : parent,
1617 acceptable, context);
1620 int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1623 struct inode *inode = dentry->d_inode;
1629 data[0] = inode->i_ino;
1630 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1631 data[2] = inode->i_generation;
1633 /* no room for directory info? return what we've stored so far */
1634 if (maxlen < 5 || !need_parent)
1637 spin_lock(&dentry->d_lock);
1638 inode = dentry->d_parent->d_inode;
1639 data[3] = inode->i_ino;
1640 data[4] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1643 data[5] = inode->i_generation;
1646 spin_unlock(&dentry->d_lock);
1650 /* looks for stat data, then copies fields to it, marks the buffer
1651 containing stat data as dirty */
1652 /* reiserfs inodes are never really dirty, since the dirty inode call
1653 ** always logs them. This call allows the VFS inode marking routines
1654 ** to properly mark inodes for datasync and such, but only actually
1655 ** does something when called for a synchronous update.
1657 int reiserfs_write_inode(struct inode *inode, int do_sync)
1659 struct reiserfs_transaction_handle th;
1660 int jbegin_count = 1;
1662 if (inode->i_sb->s_flags & MS_RDONLY)
1664 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1665 ** these cases are just when the system needs ram, not when the
1666 ** inode needs to reach disk for safety, and they can safely be
1667 ** ignored because the altered inode has already been logged.
1669 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1670 reiserfs_write_lock(inode->i_sb);
1671 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1672 reiserfs_update_sd(&th, inode);
1673 journal_end_sync(&th, inode->i_sb, jbegin_count);
1675 reiserfs_write_unlock(inode->i_sb);
1680 /* stat data of new object is inserted already, this inserts the item
1681 containing "." and ".." entries */
1682 static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1683 struct inode *inode,
1684 struct item_head *ih, struct path *path,
1687 struct super_block *sb = th->t_super;
1688 char empty_dir[EMPTY_DIR_SIZE];
1689 char *body = empty_dir;
1693 BUG_ON(!th->t_trans_id);
1695 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1696 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1697 TYPE_DIRENTRY, 3 /*key length */ );
1699 /* compose item head for new item. Directories consist of items of
1700 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1701 is done by reiserfs_new_inode */
1702 if (old_format_only(sb)) {
1703 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1704 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1706 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1707 ih->ih_key.k_objectid,
1708 INODE_PKEY(dir)->k_dir_id,
1709 INODE_PKEY(dir)->k_objectid);
1711 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1712 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1714 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1715 ih->ih_key.k_objectid,
1716 INODE_PKEY(dir)->k_dir_id,
1717 INODE_PKEY(dir)->k_objectid);
1720 /* look for place in the tree for new item */
1721 retval = search_item(sb, &key, path);
1722 if (retval == IO_ERROR) {
1723 reiserfs_warning(sb, "vs-13080: reiserfs_new_directory: "
1724 "i/o failure occurred creating new directory");
1727 if (retval == ITEM_FOUND) {
1729 reiserfs_warning(sb, "vs-13070: reiserfs_new_directory: "
1730 "object with this key exists (%k)",
1735 /* insert item, that is empty directory item */
1736 return reiserfs_insert_item(th, path, &key, ih, inode, body);
1739 /* stat data of object has been inserted, this inserts the item
1740 containing the body of symlink */
1741 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1742 struct item_head *ih,
1743 struct path *path, const char *symname,
1746 struct super_block *sb = th->t_super;
1750 BUG_ON(!th->t_trans_id);
1752 _make_cpu_key(&key, KEY_FORMAT_3_5,
1753 le32_to_cpu(ih->ih_key.k_dir_id),
1754 le32_to_cpu(ih->ih_key.k_objectid),
1755 1, TYPE_DIRECT, 3 /*key length */ );
1757 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1758 0 /*free_space */ );
1760 /* look for place in the tree for new item */
1761 retval = search_item(sb, &key, path);
1762 if (retval == IO_ERROR) {
1763 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlinik: "
1764 "i/o failure occurred creating new symlink");
1767 if (retval == ITEM_FOUND) {
1769 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlink: "
1770 "object with this key exists (%k)",
1775 /* insert item, that is body of symlink */
1776 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
1779 /* inserts the stat data into the tree, and then calls
1780 reiserfs_new_directory (to insert ".", ".." item if new object is
1781 directory) or reiserfs_new_symlink (to insert symlink body if new
1782 object is symlink) or nothing (if new object is regular file)
1784 NOTE! uid and gid must already be set in the inode. If we return
1785 non-zero due to an error, we have to drop the quota previously allocated
1786 for the fresh inode. This can only be done outside a transaction, so
1787 if we return non-zero, we also end the transaction. */
1788 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1789 struct inode *dir, int mode, const char *symname,
1790 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1791 strlen (symname) for symlinks) */
1792 loff_t i_size, struct dentry *dentry,
1793 struct inode *inode)
1795 struct super_block *sb;
1796 INITIALIZE_PATH(path_to_key);
1798 struct item_head ih;
1799 struct stat_data sd;
1803 BUG_ON(!th->t_trans_id);
1805 if (DLIMIT_ALLOC_INODE(inode)) {
1807 goto out_bad_dlimit;
1809 if (DQUOT_ALLOC_INODE(inode)) {
1813 if (!dir || !dir->i_nlink) {
1820 /* item head of new item */
1821 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1822 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1823 if (!ih.ih_key.k_objectid) {
1827 if (old_format_only(sb))
1828 /* not a perfect generation count, as object ids can be reused, but
1829 ** this is as good as reiserfs can do right now.
1830 ** note that the private part of inode isn't filled in yet, we have
1831 ** to use the directory.
1833 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1835 #if defined( USE_INODE_GENERATION_COUNTER )
1836 inode->i_generation =
1837 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1839 inode->i_generation = ++event;
1842 /* fill stat data */
1843 inode->i_nlink = (S_ISDIR(mode) ? 2 : 1);
1845 /* uid and gid must already be set by the caller for quota init */
1847 /* symlink cannot be immutable or append only, right? */
1848 if (S_ISLNK(inode->i_mode))
1849 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1851 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1852 inode->i_size = i_size;
1853 inode->i_blocks = 0;
1855 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1856 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1858 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1859 REISERFS_I(inode)->i_flags = 0;
1860 REISERFS_I(inode)->i_prealloc_block = 0;
1861 REISERFS_I(inode)->i_prealloc_count = 0;
1862 REISERFS_I(inode)->i_trans_id = 0;
1863 REISERFS_I(inode)->i_jl = NULL;
1864 REISERFS_I(inode)->i_attrs =
1865 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1866 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
1867 REISERFS_I(inode)->i_acl_access = NULL;
1868 REISERFS_I(inode)->i_acl_default = NULL;
1869 init_rwsem(&REISERFS_I(inode)->xattr_sem);
1871 if (old_format_only(sb))
1872 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1873 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1875 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1876 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1878 /* key to search for correct place for new stat data */
1879 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1880 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1881 TYPE_STAT_DATA, 3 /*key length */ );
1883 /* find proper place for inserting of stat data */
1884 retval = search_item(sb, &key, &path_to_key);
1885 if (retval == IO_ERROR) {
1889 if (retval == ITEM_FOUND) {
1890 pathrelse(&path_to_key);
1894 if (old_format_only(sb)) {
1895 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1896 pathrelse(&path_to_key);
1897 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1901 inode2sd_v1(&sd, inode, inode->i_size);
1903 inode2sd(&sd, inode, inode->i_size);
1905 // these do not go to on-disk stat data
1906 inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
1907 inode->i_blksize = reiserfs_default_io_size;
1909 // store in in-core inode the key of stat data and version all
1910 // object items will have (directory items will have old offset
1911 // format, other new objects will consist of new items)
1912 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1913 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1914 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1916 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1917 if (old_format_only(sb))
1918 set_inode_sd_version(inode, STAT_DATA_V1);
1920 set_inode_sd_version(inode, STAT_DATA_V2);
1922 /* insert the stat data into the tree */
1923 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1924 if (REISERFS_I(dir)->new_packing_locality)
1925 th->displace_new_blocks = 1;
1928 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1932 reiserfs_check_path(&path_to_key);
1935 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1936 if (!th->displace_new_blocks)
1937 REISERFS_I(dir)->new_packing_locality = 0;
1939 if (S_ISDIR(mode)) {
1940 /* insert item with "." and ".." */
1942 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1945 if (S_ISLNK(mode)) {
1946 /* insert body of symlink */
1947 if (!old_format_only(sb))
1948 i_size = ROUND_UP(i_size);
1950 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1955 reiserfs_check_path(&path_to_key);
1956 journal_end(th, th->t_super, th->t_blocks_allocated);
1957 goto out_inserted_sd;
1960 /* XXX CHECK THIS */
1961 if (reiserfs_posixacl(inode->i_sb)) {
1962 retval = reiserfs_inherit_default_acl(dir, dentry, inode);
1965 reiserfs_check_path(&path_to_key);
1966 journal_end(th, th->t_super, th->t_blocks_allocated);
1967 goto out_inserted_sd;
1969 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1970 reiserfs_warning(inode->i_sb, "ACLs aren't enabled in the fs, "
1971 "but vfs thinks they are!");
1972 } else if (is_reiserfs_priv_object(dir)) {
1973 reiserfs_mark_inode_private(inode);
1976 insert_inode_hash(inode);
1977 reiserfs_update_sd(th, inode);
1978 reiserfs_check_path(&path_to_key);
1982 /* it looks like you can easily compress these two goto targets into
1983 * one. Keeping it like this doesn't actually hurt anything, and they
1984 * are place holders for what the quota code actually needs.
1987 /* Invalidate the object, nothing was inserted yet */
1988 INODE_PKEY(inode)->k_objectid = 0;
1990 /* Quota change must be inside a transaction for journaling */
1991 DQUOT_FREE_INODE(inode);
1994 DLIMIT_FREE_INODE(inode);
1997 journal_end(th, th->t_super, th->t_blocks_allocated);
1998 /* Drop can be outside and it needs more credits so it's better to have it outside */
2000 inode->i_flags |= S_NOQUOTA;
2001 make_bad_inode(inode);
2005 th->t_trans_id = 0; /* so the caller can't use this handle later */
2007 /* If we were inheriting an ACL, we need to release the lock so that
2008 * iput doesn't deadlock in reiserfs_delete_xattrs. The locking
2009 * code really needs to be reworked, but this will take care of it
2010 * for now. -jeffm */
2011 if (REISERFS_I(dir)->i_acl_default && !IS_ERR(REISERFS_I(dir)->i_acl_default)) {
2012 reiserfs_write_unlock_xattrs(dir->i_sb);
2014 reiserfs_write_lock_xattrs(dir->i_sb);
2021 ** finds the tail page in the page cache,
2022 ** reads the last block in.
2024 ** On success, page_result is set to a locked, pinned page, and bh_result
2025 ** is set to an up to date buffer for the last block in the file. returns 0.
2027 ** tail conversion is not done, so bh_result might not be valid for writing
2028 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2029 ** trying to write the block.
2031 ** on failure, nonzero is returned, page_result and bh_result are untouched.
2033 static int grab_tail_page(struct inode *p_s_inode,
2034 struct page **page_result,
2035 struct buffer_head **bh_result)
2038 /* we want the page with the last byte in the file,
2039 ** not the page that will hold the next byte for appending
2041 unsigned long index = (p_s_inode->i_size - 1) >> PAGE_CACHE_SHIFT;
2042 unsigned long pos = 0;
2043 unsigned long start = 0;
2044 unsigned long blocksize = p_s_inode->i_sb->s_blocksize;
2045 unsigned long offset = (p_s_inode->i_size) & (PAGE_CACHE_SIZE - 1);
2046 struct buffer_head *bh;
2047 struct buffer_head *head;
2051 /* we know that we are only called with inode->i_size > 0.
2052 ** we also know that a file tail can never be as big as a block
2053 ** If i_size % blocksize == 0, our file is currently block aligned
2054 ** and it won't need converting or zeroing after a truncate.
2056 if ((offset & (blocksize - 1)) == 0) {
2059 page = grab_cache_page(p_s_inode->i_mapping, index);
2064 /* start within the page of the last block in the file */
2065 start = (offset / blocksize) * blocksize;
2067 error = block_prepare_write(page, start, offset,
2068 reiserfs_get_block_create_0);
2072 head = page_buffers(page);
2078 bh = bh->b_this_page;
2080 } while (bh != head);
2082 if (!buffer_uptodate(bh)) {
2083 /* note, this should never happen, prepare_write should
2084 ** be taking care of this for us. If the buffer isn't up to date,
2085 ** I've screwed up the code to find the buffer, or the code to
2086 ** call prepare_write
2088 reiserfs_warning(p_s_inode->i_sb,
2089 "clm-6000: error reading block %lu on dev %s",
2091 reiserfs_bdevname(p_s_inode->i_sb));
2096 *page_result = page;
2103 page_cache_release(page);
2108 ** vfs version of truncate file. Must NOT be called with
2109 ** a transaction already started.
2111 ** some code taken from block_truncate_page
2113 int reiserfs_truncate_file(struct inode *p_s_inode, int update_timestamps)
2115 struct reiserfs_transaction_handle th;
2116 /* we want the offset for the first byte after the end of the file */
2117 unsigned long offset = p_s_inode->i_size & (PAGE_CACHE_SIZE - 1);
2118 unsigned blocksize = p_s_inode->i_sb->s_blocksize;
2120 struct page *page = NULL;
2122 struct buffer_head *bh = NULL;
2125 reiserfs_write_lock(p_s_inode->i_sb);
2127 if (p_s_inode->i_size > 0) {
2128 if ((error = grab_tail_page(p_s_inode, &page, &bh))) {
2129 // -ENOENT means we truncated past the end of the file,
2130 // and get_block_create_0 could not find a block to read in,
2132 if (error != -ENOENT)
2133 reiserfs_warning(p_s_inode->i_sb,
2134 "clm-6001: grab_tail_page failed %d",
2141 /* so, if page != NULL, we have a buffer head for the offset at
2142 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2143 ** then we have an unformatted node. Otherwise, we have a direct item,
2144 ** and no zeroing is required on disk. We zero after the truncate,
2145 ** because the truncate might pack the item anyway
2146 ** (it will unmap bh if it packs).
2148 /* it is enough to reserve space in transaction for 2 balancings:
2149 one for "save" link adding and another for the first
2150 cut_from_item. 1 is for update_sd */
2151 error = journal_begin(&th, p_s_inode->i_sb,
2152 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2155 reiserfs_update_inode_transaction(p_s_inode);
2156 if (update_timestamps)
2157 /* we are doing real truncate: if the system crashes before the last
2158 transaction of truncating gets committed - on reboot the file
2159 either appears truncated properly or not truncated at all */
2160 add_save_link(&th, p_s_inode, 1);
2161 err2 = reiserfs_do_truncate(&th, p_s_inode, page, update_timestamps);
2163 journal_end(&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2167 /* check reiserfs_do_truncate after ending the transaction */
2173 if (update_timestamps) {
2174 error = remove_save_link(p_s_inode, 1 /* truncate */ );
2180 length = offset & (blocksize - 1);
2181 /* if we are not on a block boundary */
2185 length = blocksize - length;
2186 kaddr = kmap_atomic(page, KM_USER0);
2187 memset(kaddr + offset, 0, length);
2188 flush_dcache_page(page);
2189 kunmap_atomic(kaddr, KM_USER0);
2190 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2191 mark_buffer_dirty(bh);
2195 page_cache_release(page);
2198 reiserfs_write_unlock(p_s_inode->i_sb);
2203 page_cache_release(page);
2205 reiserfs_write_unlock(p_s_inode->i_sb);
2209 static int map_block_for_writepage(struct inode *inode,
2210 struct buffer_head *bh_result,
2211 unsigned long block)
2213 struct reiserfs_transaction_handle th;
2215 struct item_head tmp_ih;
2216 struct item_head *ih;
2217 struct buffer_head *bh;
2220 INITIALIZE_PATH(path);
2222 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
2223 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
2225 int use_get_block = 0;
2226 int bytes_copied = 0;
2228 int trans_running = 0;
2230 /* catch places below that try to log something without starting a trans */
2233 if (!buffer_uptodate(bh_result)) {
2237 kmap(bh_result->b_page);
2239 reiserfs_write_lock(inode->i_sb);
2240 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
2243 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2244 if (retval != POSITION_FOUND) {
2249 bh = get_last_bh(&path);
2251 item = get_item(&path);
2252 pos_in_item = path.pos_in_item;
2254 /* we've found an unformatted node */
2255 if (indirect_item_found(retval, ih)) {
2256 if (bytes_copied > 0) {
2257 reiserfs_warning(inode->i_sb,
2258 "clm-6002: bytes_copied %d",
2261 if (!get_block_num(item, pos_in_item)) {
2262 /* crap, we are writing to a hole */
2266 set_block_dev_mapped(bh_result,
2267 get_block_num(item, pos_in_item), inode);
2268 } else if (is_direct_le_ih(ih)) {
2270 p = page_address(bh_result->b_page);
2271 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2272 copy_size = ih_item_len(ih) - pos_in_item;
2274 fs_gen = get_generation(inode->i_sb);
2275 copy_item_head(&tmp_ih, ih);
2277 if (!trans_running) {
2278 /* vs-3050 is gone, no need to drop the path */
2279 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2282 reiserfs_update_inode_transaction(inode);
2284 if (fs_changed(fs_gen, inode->i_sb)
2285 && item_moved(&tmp_ih, &path)) {
2286 reiserfs_restore_prepared_buffer(inode->i_sb,
2292 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2294 if (fs_changed(fs_gen, inode->i_sb)
2295 && item_moved(&tmp_ih, &path)) {
2296 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2300 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2303 journal_mark_dirty(&th, inode->i_sb, bh);
2304 bytes_copied += copy_size;
2305 set_block_dev_mapped(bh_result, 0, inode);
2307 /* are there still bytes left? */
2308 if (bytes_copied < bh_result->b_size &&
2309 (byte_offset + bytes_copied) < inode->i_size) {
2310 set_cpu_key_k_offset(&key,
2311 cpu_key_k_offset(&key) +
2316 reiserfs_warning(inode->i_sb,
2317 "clm-6003: bad item inode %lu, device %s",
2318 inode->i_ino, reiserfs_bdevname(inode->i_sb));
2326 if (trans_running) {
2327 int err = journal_end(&th, inode->i_sb, jbegin_count);
2332 reiserfs_write_unlock(inode->i_sb);
2334 /* this is where we fill in holes in the file. */
2335 if (use_get_block) {
2336 retval = reiserfs_get_block(inode, block, bh_result,
2337 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2338 | GET_BLOCK_NO_DANGLE);
2340 if (!buffer_mapped(bh_result)
2341 || bh_result->b_blocknr == 0) {
2342 /* get_block failed to find a mapped unformatted node. */
2348 kunmap(bh_result->b_page);
2350 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2351 /* we've copied data from the page into the direct item, so the
2352 * buffer in the page is now clean, mark it to reflect that.
2354 lock_buffer(bh_result);
2355 clear_buffer_dirty(bh_result);
2356 unlock_buffer(bh_result);
2362 * mason@suse.com: updated in 2.5.54 to follow the same general io
2363 * start/recovery path as __block_write_full_page, along with special
2364 * code to handle reiserfs tails.
2366 static int reiserfs_write_full_page(struct page *page,
2367 struct writeback_control *wbc)
2369 struct inode *inode = page->mapping->host;
2370 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2372 unsigned long block;
2373 struct buffer_head *head, *bh;
2376 int checked = PageChecked(page);
2377 struct reiserfs_transaction_handle th;
2378 struct super_block *s = inode->i_sb;
2379 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2382 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2383 if (checked && (current->flags & PF_MEMALLOC)) {
2384 redirty_page_for_writepage(wbc, page);
2389 /* The page dirty bit is cleared before writepage is called, which
2390 * means we have to tell create_empty_buffers to make dirty buffers
2391 * The page really should be up to date at this point, so tossing
2392 * in the BH_Uptodate is just a sanity check.
2394 if (!page_has_buffers(page)) {
2395 create_empty_buffers(page, s->s_blocksize,
2396 (1 << BH_Dirty) | (1 << BH_Uptodate));
2398 head = page_buffers(page);
2400 /* last page in the file, zero out any contents past the
2401 ** last byte in the file
2403 if (page->index >= end_index) {
2405 unsigned last_offset;
2407 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2408 /* no file contents in this page */
2409 if (page->index >= end_index + 1 || !last_offset) {
2413 kaddr = kmap_atomic(page, KM_USER0);
2414 memset(kaddr + last_offset, 0, PAGE_CACHE_SIZE - last_offset);
2415 flush_dcache_page(page);
2416 kunmap_atomic(kaddr, KM_USER0);
2419 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
2420 /* first map all the buffers, logging any direct items we find */
2422 if ((checked || buffer_dirty(bh)) && (!buffer_mapped(bh) ||
2426 /* not mapped yet, or it points to a direct item, search
2427 * the btree for the mapping info, and log any direct
2430 if ((error = map_block_for_writepage(inode, bh, block))) {
2434 bh = bh->b_this_page;
2436 } while (bh != head);
2439 * we start the transaction after map_block_for_writepage,
2440 * because it can create holes in the file (an unbounded operation).
2441 * starting it here, we can make a reliable estimate for how many
2442 * blocks we're going to log
2445 ClearPageChecked(page);
2446 reiserfs_write_lock(s);
2447 error = journal_begin(&th, s, bh_per_page + 1);
2449 reiserfs_write_unlock(s);
2452 reiserfs_update_inode_transaction(inode);
2454 /* now go through and lock any dirty buffers on the page */
2457 if (!buffer_mapped(bh))
2459 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2463 reiserfs_prepare_for_journal(s, bh, 1);
2464 journal_mark_dirty(&th, s, bh);
2467 /* from this point on, we know the buffer is mapped to a
2468 * real block and not a direct item
2470 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2473 if (test_set_buffer_locked(bh)) {
2474 redirty_page_for_writepage(wbc, page);
2478 if (test_clear_buffer_dirty(bh)) {
2479 mark_buffer_async_write(bh);
2483 } while ((bh = bh->b_this_page) != head);
2486 error = journal_end(&th, s, bh_per_page + 1);
2487 reiserfs_write_unlock(s);
2491 BUG_ON(PageWriteback(page));
2492 set_page_writeback(page);
2496 * since any buffer might be the only dirty buffer on the page,
2497 * the first submit_bh can bring the page out of writeback.
2498 * be careful with the buffers.
2501 struct buffer_head *next = bh->b_this_page;
2502 if (buffer_async_write(bh)) {
2503 submit_bh(WRITE, bh);
2508 } while (bh != head);
2514 * if this page only had a direct item, it is very possible for
2515 * no io to be required without there being an error. Or,
2516 * someone else could have locked them and sent them down the
2517 * pipe without locking the page
2521 if (!buffer_uptodate(bh)) {
2525 bh = bh->b_this_page;
2526 } while (bh != head);
2528 SetPageUptodate(page);
2529 end_page_writeback(page);
2534 /* catches various errors, we need to make sure any valid dirty blocks
2535 * get to the media. The page is currently locked and not marked for
2538 ClearPageUptodate(page);
2542 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2544 mark_buffer_async_write(bh);
2547 * clear any dirty bits that might have come from getting
2548 * attached to a dirty page
2550 clear_buffer_dirty(bh);
2552 bh = bh->b_this_page;
2553 } while (bh != head);
2555 BUG_ON(PageWriteback(page));
2556 set_page_writeback(page);
2559 struct buffer_head *next = bh->b_this_page;
2560 if (buffer_async_write(bh)) {
2561 clear_buffer_dirty(bh);
2562 submit_bh(WRITE, bh);
2567 } while (bh != head);
2571 static int reiserfs_readpage(struct file *f, struct page *page)
2573 return block_read_full_page(page, reiserfs_get_block);
2576 static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
2578 struct inode *inode = page->mapping->host;
2579 reiserfs_wait_on_write_block(inode->i_sb);
2580 return reiserfs_write_full_page(page, wbc);
2583 static int reiserfs_prepare_write(struct file *f, struct page *page,
2584 unsigned from, unsigned to)
2586 struct inode *inode = page->mapping->host;
2590 reiserfs_wait_on_write_block(inode->i_sb);
2591 fix_tail_page_for_writing(page);
2592 if (reiserfs_transaction_running(inode->i_sb)) {
2593 struct reiserfs_transaction_handle *th;
2594 th = (struct reiserfs_transaction_handle *)current->
2596 BUG_ON(!th->t_refcount);
2597 BUG_ON(!th->t_trans_id);
2598 old_ref = th->t_refcount;
2602 ret = block_prepare_write(page, from, to, reiserfs_get_block);
2603 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2604 struct reiserfs_transaction_handle *th = current->journal_info;
2605 /* this gets a little ugly. If reiserfs_get_block returned an
2606 * error and left a transacstion running, we've got to close it,
2607 * and we've got to free handle if it was a persistent transaction.
2609 * But, if we had nested into an existing transaction, we need
2610 * to just drop the ref count on the handle.
2612 * If old_ref == 0, the transaction is from reiserfs_get_block,
2613 * and it was a persistent trans. Otherwise, it was nested above.
2615 if (th->t_refcount > old_ref) {
2620 reiserfs_write_lock(inode->i_sb);
2621 err = reiserfs_end_persistent_transaction(th);
2622 reiserfs_write_unlock(inode->i_sb);
2632 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2634 return generic_block_bmap(as, block, reiserfs_bmap);
2637 static int reiserfs_commit_write(struct file *f, struct page *page,
2638 unsigned from, unsigned to)
2640 struct inode *inode = page->mapping->host;
2641 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2644 struct reiserfs_transaction_handle *th = NULL;
2646 reiserfs_wait_on_write_block(inode->i_sb);
2647 if (reiserfs_transaction_running(inode->i_sb)) {
2648 th = current->journal_info;
2650 reiserfs_commit_page(inode, page, from, to);
2652 /* generic_commit_write does this for us, but does not update the
2653 ** transaction tracking stuff when the size changes. So, we have
2654 ** to do the i_size updates here.
2656 if (pos > inode->i_size) {
2657 struct reiserfs_transaction_handle myth;
2658 reiserfs_write_lock(inode->i_sb);
2659 /* If the file have grown beyond the border where it
2660 can have a tail, unmark it as needing a tail
2662 if ((have_large_tails(inode->i_sb)
2663 && inode->i_size > i_block_size(inode) * 4)
2664 || (have_small_tails(inode->i_sb)
2665 && inode->i_size > i_block_size(inode)))
2666 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2668 ret = journal_begin(&myth, inode->i_sb, 1);
2670 reiserfs_write_unlock(inode->i_sb);
2673 reiserfs_update_inode_transaction(inode);
2674 inode->i_size = pos;
2676 * this will just nest into our transaction. It's important
2677 * to use mark_inode_dirty so the inode gets pushed around on the
2678 * dirty lists, and so that O_SYNC works as expected
2680 mark_inode_dirty(inode);
2681 reiserfs_update_sd(&myth, inode);
2683 ret = journal_end(&myth, inode->i_sb, 1);
2684 reiserfs_write_unlock(inode->i_sb);
2689 reiserfs_write_lock(inode->i_sb);
2691 mark_inode_dirty(inode);
2692 ret = reiserfs_end_persistent_transaction(th);
2693 reiserfs_write_unlock(inode->i_sb);
2703 reiserfs_write_lock(inode->i_sb);
2705 reiserfs_update_sd(th, inode);
2706 ret = reiserfs_end_persistent_transaction(th);
2707 reiserfs_write_unlock(inode->i_sb);
2713 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2715 if (reiserfs_attrs(inode->i_sb)) {
2716 if (sd_attrs & REISERFS_SYNC_FL)
2717 inode->i_flags |= S_SYNC;
2719 inode->i_flags &= ~S_SYNC;
2720 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2721 inode->i_flags |= S_IMMUTABLE;
2723 inode->i_flags &= ~S_IMMUTABLE;
2724 if (sd_attrs & REISERFS_IUNLINK_FL)
2725 inode->i_flags |= S_IUNLINK;
2727 inode->i_flags &= ~S_IUNLINK;
2728 if (sd_attrs & REISERFS_BARRIER_FL)
2729 inode->i_flags |= S_BARRIER;
2731 inode->i_flags &= ~S_BARRIER;
2732 if (sd_attrs & REISERFS_APPEND_FL)
2733 inode->i_flags |= S_APPEND;
2735 inode->i_flags &= ~S_APPEND;
2736 if (sd_attrs & REISERFS_NOATIME_FL)
2737 inode->i_flags |= S_NOATIME;
2739 inode->i_flags &= ~S_NOATIME;
2740 if (sd_attrs & REISERFS_NOTAIL_FL)
2741 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2743 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2747 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
2749 if (reiserfs_attrs(inode->i_sb)) {
2750 if (inode->i_flags & S_IMMUTABLE)
2751 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2753 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2754 if (inode->i_flags & S_IUNLINK)
2755 *sd_attrs |= REISERFS_IUNLINK_FL;
2757 *sd_attrs &= ~REISERFS_IUNLINK_FL;
2758 if (inode->i_flags & S_BARRIER)
2759 *sd_attrs |= REISERFS_BARRIER_FL;
2761 *sd_attrs &= ~REISERFS_BARRIER_FL;
2762 if (inode->i_flags & S_SYNC)
2763 *sd_attrs |= REISERFS_SYNC_FL;
2765 *sd_attrs &= ~REISERFS_SYNC_FL;
2766 if (inode->i_flags & S_NOATIME)
2767 *sd_attrs |= REISERFS_NOATIME_FL;
2769 *sd_attrs &= ~REISERFS_NOATIME_FL;
2770 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
2771 *sd_attrs |= REISERFS_NOTAIL_FL;
2773 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2777 /* decide if this buffer needs to stay around for data logging or ordered
2780 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2783 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2786 spin_lock(&j->j_dirty_buffers_lock);
2787 if (!buffer_mapped(bh)) {
2790 /* the page is locked, and the only places that log a data buffer
2791 * also lock the page.
2793 if (reiserfs_file_data_log(inode)) {
2795 * very conservative, leave the buffer pinned if
2796 * anyone might need it.
2798 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2801 } else if (buffer_dirty(bh)) {
2802 struct reiserfs_journal_list *jl;
2803 struct reiserfs_jh *jh = bh->b_private;
2805 /* why is this safe?
2806 * reiserfs_setattr updates i_size in the on disk
2807 * stat data before allowing vmtruncate to be called.
2809 * If buffer was put onto the ordered list for this
2810 * transaction, we know for sure either this transaction
2811 * or an older one already has updated i_size on disk,
2812 * and this ordered data won't be referenced in the file
2815 * if the buffer was put onto the ordered list for an older
2816 * transaction, we need to leave it around
2818 if (jh && (jl = jh->jl)
2819 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2823 if (ret && bh->b_private) {
2824 reiserfs_free_jh(bh);
2826 spin_unlock(&j->j_dirty_buffers_lock);
2831 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2832 static int reiserfs_invalidatepage(struct page *page, unsigned long offset)
2834 struct buffer_head *head, *bh, *next;
2835 struct inode *inode = page->mapping->host;
2836 unsigned int curr_off = 0;
2839 BUG_ON(!PageLocked(page));
2842 ClearPageChecked(page);
2844 if (!page_has_buffers(page))
2847 head = page_buffers(page);
2850 unsigned int next_off = curr_off + bh->b_size;
2851 next = bh->b_this_page;
2854 * is this block fully invalidated?
2856 if (offset <= curr_off) {
2857 if (invalidatepage_can_drop(inode, bh))
2858 reiserfs_unmap_buffer(bh);
2862 curr_off = next_off;
2864 } while (bh != head);
2867 * We release buffers only if the entire page is being invalidated.
2868 * The get_block cached value has been unconditionally invalidated,
2869 * so real IO is not possible anymore.
2872 ret = try_to_release_page(page, 0);
2877 static int reiserfs_set_page_dirty(struct page *page)
2879 struct inode *inode = page->mapping->host;
2880 if (reiserfs_file_data_log(inode)) {
2881 SetPageChecked(page);
2882 return __set_page_dirty_nobuffers(page);
2884 return __set_page_dirty_buffers(page);
2888 * Returns 1 if the page's buffers were dropped. The page is locked.
2890 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
2891 * in the buffers at page_buffers(page).
2893 * even in -o notail mode, we can't be sure an old mount without -o notail
2894 * didn't create files with tails.
2896 static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
2898 struct inode *inode = page->mapping->host;
2899 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2900 struct buffer_head *head;
2901 struct buffer_head *bh;
2904 WARN_ON(PageChecked(page));
2905 spin_lock(&j->j_dirty_buffers_lock);
2906 head = page_buffers(page);
2909 if (bh->b_private) {
2910 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
2911 reiserfs_free_jh(bh);
2917 bh = bh->b_this_page;
2918 } while (bh != head);
2920 ret = try_to_free_buffers(page);
2921 spin_unlock(&j->j_dirty_buffers_lock);
2925 /* We thank Mingming Cao for helping us understand in great detail what
2926 to do in this section of the code. */
2927 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
2928 const struct iovec *iov, loff_t offset,
2929 unsigned long nr_segs)
2931 struct file *file = iocb->ki_filp;
2932 struct inode *inode = file->f_mapping->host;
2934 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2936 reiserfs_get_blocks_direct_io, NULL);
2939 int reiserfs_sync_flags(struct inode *inode)
2941 u16 oldflags, newflags;
2943 oldflags = REISERFS_I(inode)->i_attrs;
2944 newflags = oldflags;
2945 i_attrs_to_sd_attrs(inode, &newflags);
2947 if (oldflags ^ newflags) {
2948 REISERFS_I(inode)->i_attrs = newflags;
2949 inode->i_ctime = CURRENT_TIME_SEC;
2950 mark_inode_dirty(inode);
2955 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
2957 struct inode *inode = dentry->d_inode;
2959 unsigned int ia_valid = attr->ia_valid;
2960 reiserfs_write_lock(inode->i_sb);
2961 if (attr->ia_valid & ATTR_SIZE) {
2962 /* version 2 items will be caught by the s_maxbytes check
2963 ** done for us in vmtruncate
2965 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
2966 attr->ia_size > MAX_NON_LFS) {
2970 /* fill in hole pointers in the expanding truncate case. */
2971 if (attr->ia_size > inode->i_size) {
2972 error = generic_cont_expand(inode, attr->ia_size);
2973 if (REISERFS_I(inode)->i_prealloc_count > 0) {
2975 struct reiserfs_transaction_handle th;
2976 /* we're changing at most 2 bitmaps, inode + super */
2977 err = journal_begin(&th, inode->i_sb, 4);
2979 reiserfs_discard_prealloc(&th, inode);
2980 err = journal_end(&th, inode->i_sb, 4);
2990 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
2991 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
2992 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
2993 /* stat data of format v3.5 has 16 bit uid and gid */
2998 error = inode_change_ok(inode, attr);
3001 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
3002 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid) ||
3003 (ia_valid & ATTR_XID && attr->ia_xid != inode->i_xid)) {
3004 error = reiserfs_chown_xattrs(inode, attr);
3007 struct reiserfs_transaction_handle th;
3010 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3011 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3014 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3016 journal_begin(&th, inode->i_sb,
3021 DQUOT_TRANSFER(inode, attr) ? -EDQUOT : 0;
3023 journal_end(&th, inode->i_sb,
3027 /* Update corresponding info in inode so that everything is in
3028 * one transaction */
3029 if (attr->ia_valid & ATTR_UID)
3030 inode->i_uid = attr->ia_uid;
3031 if (attr->ia_valid & ATTR_GID)
3032 inode->i_gid = attr->ia_gid;
3033 if ((attr->ia_valid & ATTR_XID) &&
3035 inode->i_xid = attr->ia_xid;
3036 mark_inode_dirty(inode);
3038 journal_end(&th, inode->i_sb, jbegin_count);
3042 error = inode_setattr(inode, attr);
3045 if (!error && reiserfs_posixacl(inode->i_sb)) {
3046 if (attr->ia_valid & ATTR_MODE)
3047 error = reiserfs_acl_chmod(inode);
3051 reiserfs_write_unlock(inode->i_sb);
3055 struct address_space_operations reiserfs_address_space_operations = {
3056 .writepage = reiserfs_writepage,
3057 .readpage = reiserfs_readpage,
3058 .readpages = reiserfs_readpages,
3059 .releasepage = reiserfs_releasepage,
3060 .invalidatepage = reiserfs_invalidatepage,
3061 .sync_page = block_sync_page,
3062 .prepare_write = reiserfs_prepare_write,
3063 .commit_write = reiserfs_commit_write,
3064 .bmap = reiserfs_aop_bmap,
3065 .direct_IO = reiserfs_direct_IO,
3066 .set_page_dirty = reiserfs_set_page_dirty,