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 struct buffer_head *bh_result,
477 bh_result->b_page = NULL;
479 /* We set the b_size before reiserfs_get_block call since it is
480 referenced in convert_tail_for_hole() that may be called from
481 reiserfs_get_block() */
482 bh_result->b_size = (1 << inode->i_blkbits);
484 ret = reiserfs_get_block(inode, iblock, bh_result,
485 create | GET_BLOCK_NO_DANGLE);
489 /* don't allow direct io onto tail pages */
490 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
491 /* make sure future calls to the direct io funcs for this offset
492 ** in the file fail by unmapping the buffer
494 clear_buffer_mapped(bh_result);
497 /* Possible unpacked tail. Flush the data before pages have
499 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
502 err = reiserfs_commit_for_inode(inode);
503 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
513 ** helper function for when reiserfs_get_block is called for a hole
514 ** but the file tail is still in a direct item
515 ** bh_result is the buffer head for the hole
516 ** tail_offset is the offset of the start of the tail in the file
518 ** This calls prepare_write, which will start a new transaction
519 ** you should not be in a transaction, or have any paths held when you
522 static int convert_tail_for_hole(struct inode *inode,
523 struct buffer_head *bh_result,
527 unsigned long tail_end;
528 unsigned long tail_start;
529 struct page *tail_page;
530 struct page *hole_page = bh_result->b_page;
533 if ((tail_offset & (bh_result->b_size - 1)) != 1)
536 /* always try to read until the end of the block */
537 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
538 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
540 index = tail_offset >> PAGE_CACHE_SHIFT;
541 /* hole_page can be zero in case of direct_io, we are sure
542 that we cannot get here if we write with O_DIRECT into
544 if (!hole_page || index != hole_page->index) {
545 tail_page = grab_cache_page(inode->i_mapping, index);
551 tail_page = hole_page;
554 /* we don't have to make sure the conversion did not happen while
555 ** we were locking the page because anyone that could convert
556 ** must first take i_mutex.
558 ** We must fix the tail page for writing because it might have buffers
559 ** that are mapped, but have a block number of 0. This indicates tail
560 ** data that has been read directly into the page, and block_prepare_write
561 ** won't trigger a get_block in this case.
563 fix_tail_page_for_writing(tail_page);
564 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
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 path *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;
629 reiserfs_write_lock(inode->i_sb);
630 version = get_inode_item_key_version(inode);
632 if (!file_capable(inode, block)) {
633 reiserfs_write_unlock(inode->i_sb);
637 /* if !create, we aren't changing the FS, so we don't need to
638 ** log anything, so we don't need to start a transaction
640 if (!(create & GET_BLOCK_CREATE)) {
642 /* find number of block-th logical block of the file */
643 ret = _get_block_create_0(inode, block, bh_result,
644 create | GET_BLOCK_READ_DIRECT);
645 reiserfs_write_unlock(inode->i_sb);
649 * if we're already in a transaction, make sure to close
650 * any new transactions we start in this func
652 if ((create & GET_BLOCK_NO_DANGLE) ||
653 reiserfs_transaction_running(inode->i_sb))
656 /* If file is of such a size, that it might have a tail and tails are enabled
657 ** we should mark it as possibly needing tail packing on close
659 if ((have_large_tails(inode->i_sb)
660 && inode->i_size < i_block_size(inode) * 4)
661 || (have_small_tails(inode->i_sb)
662 && inode->i_size < i_block_size(inode)))
663 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
665 /* set the key of the first byte in the 'block'-th block of file */
666 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
667 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
669 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
674 reiserfs_update_inode_transaction(inode);
678 retval = search_for_position_by_key(inode->i_sb, &key, &path);
679 if (retval == IO_ERROR) {
684 bh = get_last_bh(&path);
686 item = get_item(&path);
687 pos_in_item = path.pos_in_item;
689 fs_gen = get_generation(inode->i_sb);
690 copy_item_head(&tmp_ih, ih);
692 if (allocation_needed
693 (retval, allocated_block_nr, ih, item, pos_in_item)) {
694 /* we have to allocate block for the unformatted node */
701 _allocate_block(th, block, inode, &allocated_block_nr,
704 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
705 /* restart the transaction to give the journal a chance to free
706 ** some blocks. releases the path, so we have to go back to
707 ** research if we succeed on the second try
709 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
710 retval = restart_transaction(th, inode, &path);
714 _allocate_block(th, block, inode,
715 &allocated_block_nr, NULL, create);
717 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
720 if (repeat == QUOTA_EXCEEDED)
727 if (fs_changed(fs_gen, inode->i_sb)
728 && item_moved(&tmp_ih, &path)) {
733 if (indirect_item_found(retval, ih)) {
734 b_blocknr_t unfm_ptr;
735 /* 'block'-th block is in the file already (there is
736 corresponding cell in some indirect item). But it may be
737 zero unformatted node pointer (hole) */
738 unfm_ptr = get_block_num(item, pos_in_item);
740 /* use allocated block to plug the hole */
741 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
742 if (fs_changed(fs_gen, inode->i_sb)
743 && item_moved(&tmp_ih, &path)) {
744 reiserfs_restore_prepared_buffer(inode->i_sb,
748 set_buffer_new(bh_result);
749 if (buffer_dirty(bh_result)
750 && reiserfs_data_ordered(inode->i_sb))
751 reiserfs_add_ordered_list(inode, bh_result);
752 put_block_num(item, pos_in_item, allocated_block_nr);
753 unfm_ptr = allocated_block_nr;
754 journal_mark_dirty(th, inode->i_sb, bh);
755 reiserfs_update_sd(th, inode);
757 set_block_dev_mapped(bh_result, unfm_ptr, inode);
761 retval = reiserfs_end_persistent_transaction(th);
763 reiserfs_write_unlock(inode->i_sb);
765 /* the item was found, so new blocks were not added to the file
766 ** there is no need to make sure the inode is updated with this
777 /* desired position is not found or is in the direct item. We have
778 to append file with holes up to 'block'-th block converting
779 direct items to indirect one if necessary */
782 if (is_statdata_le_ih(ih)) {
784 struct cpu_key tmp_key;
786 /* indirect item has to be inserted */
787 make_le_item_head(&tmp_ih, &key, version, 1,
788 TYPE_INDIRECT, UNFM_P_SIZE,
789 0 /* free_space */ );
791 if (cpu_key_k_offset(&key) == 1) {
792 /* we are going to add 'block'-th block to the file. Use
793 allocated block for that */
794 unp = cpu_to_le32(allocated_block_nr);
795 set_block_dev_mapped(bh_result,
796 allocated_block_nr, inode);
797 set_buffer_new(bh_result);
801 set_cpu_key_k_offset(&tmp_key, 1);
802 PATH_LAST_POSITION(&path)++;
805 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
806 inode, (char *)&unp);
808 reiserfs_free_block(th, inode,
809 allocated_block_nr, 1);
810 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
812 //mark_tail_converted (inode);
813 } else if (is_direct_le_ih(ih)) {
814 /* direct item has to be converted */
818 ((le_ih_k_offset(ih) -
819 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
820 if (tail_offset == cpu_key_k_offset(&key)) {
821 /* direct item we just found fits into block we have
822 to map. Convert it into unformatted node: use
823 bh_result for the conversion */
824 set_block_dev_mapped(bh_result,
825 allocated_block_nr, inode);
829 /* we have to padd file tail stored in direct item(s)
830 up to block size and convert it to unformatted
831 node. FIXME: this should also get into page cache */
835 * ugly, but we can only end the transaction if
838 BUG_ON(!th->t_refcount);
839 if (th->t_refcount == 1) {
841 reiserfs_end_persistent_transaction
849 convert_tail_for_hole(inode, bh_result,
852 if (retval != -ENOSPC)
853 reiserfs_warning(inode->i_sb,
854 "clm-6004: convert tail failed inode %lu, error %d",
857 if (allocated_block_nr) {
858 /* the bitmap, the super, and the stat data == 3 */
860 th = reiserfs_persistent_transaction(inode->i_sb, 3);
862 reiserfs_free_block(th,
872 direct2indirect(th, inode, &path, unbh,
875 reiserfs_unmap_buffer(unbh);
876 reiserfs_free_block(th, inode,
877 allocated_block_nr, 1);
880 /* it is important the set_buffer_uptodate is done after
881 ** the direct2indirect. The buffer might contain valid
882 ** data newer than the data on disk (read by readpage, changed,
883 ** and then sent here by writepage). direct2indirect needs
884 ** to know if unbh was already up to date, so it can decide
885 ** if the data in unbh needs to be replaced with data from
888 set_buffer_uptodate(unbh);
890 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
891 buffer will disappear shortly, so it should not be added to
894 /* we've converted the tail, so we must
895 ** flush unbh before the transaction commits
897 reiserfs_add_tail_list(inode, unbh);
899 /* mark it dirty now to prevent commit_write from adding
900 ** this buffer to the inode's dirty buffer list
903 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
904 * It's still atomic, but it sets the page dirty too,
905 * which makes it eligible for writeback at any time by the
906 * VM (which was also the case with __mark_buffer_dirty())
908 mark_buffer_dirty(unbh);
911 /* append indirect item with holes if needed, when appending
912 pointer to 'block'-th block use block, which is already
914 struct cpu_key tmp_key;
915 unp_t unf_single = 0; // We use this in case we need to allocate only
916 // one block which is a fastpath
918 __u64 max_to_insert =
919 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
923 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
924 "vs-804: invalid position for append");
925 /* indirect item has to be appended, set up key of that position */
926 make_cpu_key(&tmp_key, inode,
927 le_key_k_offset(version,
930 inode->i_sb->s_blocksize),
931 //pos_in_item * inode->i_sb->s_blocksize,
932 TYPE_INDIRECT, 3); // key type is unimportant
934 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
935 "green-805: invalid offset");
938 ((cpu_key_k_offset(&key) -
939 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
942 if (blocks_needed == 1) {
945 un = kmalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_ATOMIC); // We need to avoid scheduling.
952 UNFM_P_SIZE * min(blocks_needed,
955 if (blocks_needed <= max_to_insert) {
956 /* we are going to add target block to the file. Use allocated
958 un[blocks_needed - 1] =
959 cpu_to_le32(allocated_block_nr);
960 set_block_dev_mapped(bh_result,
961 allocated_block_nr, inode);
962 set_buffer_new(bh_result);
965 /* paste hole to the indirect item */
966 /* If kmalloc failed, max_to_insert becomes zero and it means we
967 only have space for one block */
969 max_to_insert ? max_to_insert : 1;
972 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
977 if (blocks_needed != 1)
981 reiserfs_free_block(th, inode,
982 allocated_block_nr, 1);
986 /* We need to mark new file size in case this function will be
987 interrupted/aborted later on. And we may do this only for
990 inode->i_sb->s_blocksize * blocks_needed;
997 /* this loop could log more blocks than we had originally asked
998 ** for. So, we have to allow the transaction to end if it is
999 ** too big or too full. Update the inode so things are
1000 ** consistent if we crash before the function returns
1002 ** release the path so that anybody waiting on the path before
1003 ** ending their transaction will be able to continue.
1005 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
1006 retval = restart_transaction(th, inode, &path);
1010 /* inserting indirect pointers for a hole can take a
1011 ** long time. reschedule if needed
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,
1022 "vs-825: reiserfs_get_block: "
1023 "%K should not be found", &key);
1025 if (allocated_block_nr)
1026 reiserfs_free_block(th, inode,
1027 allocated_block_nr, 1);
1031 bh = get_last_bh(&path);
1033 item = get_item(&path);
1034 pos_in_item = path.pos_in_item;
1040 if (th && (!dangle || (retval && !th->t_trans_id))) {
1043 reiserfs_update_sd(th, inode);
1044 err = reiserfs_end_persistent_transaction(th);
1049 reiserfs_write_unlock(inode->i_sb);
1050 reiserfs_check_path(&path);
1055 reiserfs_readpages(struct file *file, struct address_space *mapping,
1056 struct list_head *pages, unsigned nr_pages)
1058 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
1061 /* Compute real number of used bytes by file
1062 * Following three functions can go away when we'll have enough space in stat item
1064 static int real_space_diff(struct inode *inode, int sd_size)
1067 loff_t blocksize = inode->i_sb->s_blocksize;
1069 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1072 /* End of file is also in full block with indirect reference, so round
1073 ** up to the next block.
1075 ** there is just no way to know if the tail is actually packed
1076 ** on the file, so we have to assume it isn't. When we pack the
1077 ** tail, we add 4 bytes to pretend there really is an unformatted
1082 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1087 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
1090 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1091 return inode->i_size +
1092 (loff_t) (real_space_diff(inode, sd_size));
1094 return ((loff_t) real_space_diff(inode, sd_size)) +
1095 (((loff_t) blocks) << 9);
1098 /* Compute number of blocks used by file in ReiserFS counting */
1099 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1101 loff_t bytes = inode_get_bytes(inode);
1102 loff_t real_space = real_space_diff(inode, sd_size);
1104 /* keeps fsck and non-quota versions of reiserfs happy */
1105 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1106 bytes += (loff_t) 511;
1109 /* files from before the quota patch might i_blocks such that
1110 ** bytes < real_space. Deal with that here to prevent it from
1113 if (bytes < real_space)
1115 return (bytes - real_space) >> 9;
1119 // BAD: new directories have stat data of new type and all other items
1120 // of old type. Version stored in the inode says about body items, so
1121 // in update_stat_data we can not rely on inode, but have to check
1122 // item version directly
1125 // called by read_locked_inode
1126 static void init_inode(struct inode *inode, struct path *path)
1128 struct buffer_head *bh;
1129 struct item_head *ih;
1133 //int version = ITEM_VERSION_1;
1135 bh = PATH_PLAST_BUFFER(path);
1136 ih = PATH_PITEM_HEAD(path);
1138 copy_key(INODE_PKEY(inode), &(ih->ih_key));
1139 inode->i_blksize = reiserfs_default_io_size;
1141 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1142 REISERFS_I(inode)->i_flags = 0;
1143 REISERFS_I(inode)->i_prealloc_block = 0;
1144 REISERFS_I(inode)->i_prealloc_count = 0;
1145 REISERFS_I(inode)->i_trans_id = 0;
1146 REISERFS_I(inode)->i_jl = NULL;
1147 REISERFS_I(inode)->i_acl_access = NULL;
1148 REISERFS_I(inode)->i_acl_default = NULL;
1149 init_rwsem(&REISERFS_I(inode)->xattr_sem);
1151 if (stat_data_v1(ih)) {
1152 struct stat_data_v1 *sd =
1153 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1154 unsigned long blocks;
1156 uid = sd_v1_uid(sd);
1157 gid = sd_v1_gid(sd);
1159 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1160 set_inode_sd_version(inode, STAT_DATA_V1);
1161 inode->i_mode = sd_v1_mode(sd);
1162 inode->i_nlink = sd_v1_nlink(sd);
1163 inode->i_size = sd_v1_size(sd);
1164 inode->i_atime.tv_sec = sd_v1_atime(sd);
1165 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1166 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1167 inode->i_atime.tv_nsec = 0;
1168 inode->i_ctime.tv_nsec = 0;
1169 inode->i_mtime.tv_nsec = 0;
1171 inode->i_blocks = sd_v1_blocks(sd);
1172 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1173 blocks = (inode->i_size + 511) >> 9;
1174 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1175 if (inode->i_blocks > blocks) {
1176 // there was a bug in <=3.5.23 when i_blocks could take negative
1177 // values. Starting from 3.5.17 this value could even be stored in
1178 // stat data. For such files we set i_blocks based on file
1179 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1180 // only updated if file's inode will ever change
1181 inode->i_blocks = blocks;
1184 rdev = sd_v1_rdev(sd);
1185 REISERFS_I(inode)->i_first_direct_byte =
1186 sd_v1_first_direct_byte(sd);
1187 /* an early bug in the quota code can give us an odd number for the
1188 ** block count. This is incorrect, fix it here.
1190 if (inode->i_blocks & 1) {
1193 inode_set_bytes(inode,
1194 to_real_used_space(inode, inode->i_blocks,
1196 /* nopack is initially zero for v1 objects. For v2 objects,
1197 nopack is initialised from sd_attrs */
1198 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1200 // new stat data found, but object may have old items
1201 // (directories and symlinks)
1202 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1204 uid = sd_v2_uid(sd);
1205 gid = sd_v2_gid(sd);
1207 inode->i_mode = sd_v2_mode(sd);
1208 inode->i_nlink = sd_v2_nlink(sd);
1209 inode->i_size = sd_v2_size(sd);
1210 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1211 inode->i_atime.tv_sec = sd_v2_atime(sd);
1212 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1213 inode->i_ctime.tv_nsec = 0;
1214 inode->i_mtime.tv_nsec = 0;
1215 inode->i_atime.tv_nsec = 0;
1216 inode->i_blocks = sd_v2_blocks(sd);
1217 rdev = sd_v2_rdev(sd);
1218 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1219 inode->i_generation =
1220 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1222 inode->i_generation = sd_v2_generation(sd);
1224 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1225 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1227 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1228 REISERFS_I(inode)->i_first_direct_byte = 0;
1229 set_inode_sd_version(inode, STAT_DATA_V2);
1230 inode_set_bytes(inode,
1231 to_real_used_space(inode, inode->i_blocks,
1233 /* read persistent inode attributes from sd and initalise
1234 generic inode flags from them */
1235 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1236 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
1239 inode->i_uid = INOXID_UID(XID_TAG(inode), uid, gid);
1240 inode->i_gid = INOXID_GID(XID_TAG(inode), uid, gid);
1241 inode->i_xid = INOXID_XID(XID_TAG(inode), uid, gid, 0);
1244 if (S_ISREG(inode->i_mode)) {
1245 inode->i_op = &reiserfs_file_inode_operations;
1246 inode->i_fop = &reiserfs_file_operations;
1247 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1248 } else if (S_ISDIR(inode->i_mode)) {
1249 inode->i_op = &reiserfs_dir_inode_operations;
1250 inode->i_fop = &reiserfs_dir_operations;
1251 } else if (S_ISLNK(inode->i_mode)) {
1252 inode->i_op = &reiserfs_symlink_inode_operations;
1253 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1255 inode->i_blocks = 0;
1256 inode->i_op = &reiserfs_special_inode_operations;
1257 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1261 // update new stat data with inode fields
1262 static void inode2sd(void *sd, struct inode *inode, loff_t size)
1264 struct stat_data *sd_v2 = (struct stat_data *)sd;
1265 uid_t uid = XIDINO_UID(XID_TAG(inode), inode->i_uid, inode->i_xid);
1266 gid_t gid = XIDINO_GID(XID_TAG(inode), inode->i_gid, inode->i_xid);
1269 set_sd_v2_uid(sd_v2, uid);
1270 set_sd_v2_gid(sd_v2, gid);
1271 set_sd_v2_mode(sd_v2, inode->i_mode);
1272 set_sd_v2_nlink(sd_v2, inode->i_nlink);
1273 set_sd_v2_size(sd_v2, size);
1274 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1275 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1276 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1277 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1278 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1279 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1281 set_sd_v2_generation(sd_v2, inode->i_generation);
1282 flags = REISERFS_I(inode)->i_attrs;
1283 i_attrs_to_sd_attrs(inode, &flags);
1284 set_sd_v2_attrs(sd_v2, flags);
1287 // used to copy inode's fields to old stat data
1288 static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
1290 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
1292 set_sd_v1_mode(sd_v1, inode->i_mode);
1293 set_sd_v1_uid(sd_v1, inode->i_uid);
1294 set_sd_v1_gid(sd_v1, inode->i_gid);
1295 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1296 set_sd_v1_size(sd_v1, size);
1297 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1298 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1299 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
1301 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1302 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1304 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1306 // Sigh. i_first_direct_byte is back
1307 set_sd_v1_first_direct_byte(sd_v1,
1308 REISERFS_I(inode)->i_first_direct_byte);
1311 /* NOTE, you must prepare the buffer head before sending it here,
1312 ** and then log it after the call
1314 static void update_stat_data(struct path *path, struct inode *inode,
1317 struct buffer_head *bh;
1318 struct item_head *ih;
1320 bh = PATH_PLAST_BUFFER(path);
1321 ih = PATH_PITEM_HEAD(path);
1323 if (!is_statdata_le_ih(ih))
1324 reiserfs_panic(inode->i_sb,
1325 "vs-13065: update_stat_data: key %k, found item %h",
1326 INODE_PKEY(inode), ih);
1328 if (stat_data_v1(ih)) {
1329 // path points to old stat data
1330 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1332 inode2sd(B_I_PITEM(bh, ih), inode, size);
1338 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1339 struct inode *inode, loff_t size)
1342 INITIALIZE_PATH(path);
1343 struct buffer_head *bh;
1345 struct item_head *ih, tmp_ih;
1348 BUG_ON(!th->t_trans_id);
1350 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant
1354 /* look for the object's stat data */
1355 retval = search_item(inode->i_sb, &key, &path);
1356 if (retval == IO_ERROR) {
1357 reiserfs_warning(inode->i_sb,
1358 "vs-13050: reiserfs_update_sd: "
1359 "i/o failure occurred trying to update %K stat data",
1363 if (retval == ITEM_NOT_FOUND) {
1364 pos = PATH_LAST_POSITION(&path);
1366 if (inode->i_nlink == 0) {
1367 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1370 reiserfs_warning(inode->i_sb,
1371 "vs-13060: reiserfs_update_sd: "
1372 "stat data of object %k (nlink == %d) not found (pos %d)",
1373 INODE_PKEY(inode), inode->i_nlink,
1375 reiserfs_check_path(&path);
1379 /* sigh, prepare_for_journal might schedule. When it schedules the
1380 ** FS might change. We have to detect that, and loop back to the
1381 ** search if the stat data item has moved
1383 bh = get_last_bh(&path);
1385 copy_item_head(&tmp_ih, ih);
1386 fs_gen = get_generation(inode->i_sb);
1387 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1388 if (fs_changed(fs_gen, inode->i_sb)
1389 && item_moved(&tmp_ih, &path)) {
1390 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1391 continue; /* Stat_data item has been moved after scheduling. */
1395 update_stat_data(&path, inode, size);
1396 journal_mark_dirty(th, th->t_super, bh);
1401 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1402 ** does a make_bad_inode when things go wrong. But, we need to make sure
1403 ** and clear the key in the private portion of the inode, otherwise a
1404 ** corresponding iput might try to delete whatever object the inode last
1407 static void reiserfs_make_bad_inode(struct inode *inode)
1409 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1410 make_bad_inode(inode);
1414 // initially this function was derived from minix or ext2's analog and
1415 // evolved as the prototype did
1418 int reiserfs_init_locked_inode(struct inode *inode, void *p)
1420 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1421 inode->i_ino = args->objectid;
1422 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1426 /* looks for stat data in the tree, and fills up the fields of in-core
1427 inode stat data fields */
1428 void reiserfs_read_locked_inode(struct inode *inode,
1429 struct reiserfs_iget_args *args)
1431 INITIALIZE_PATH(path_to_sd);
1433 unsigned long dirino;
1436 dirino = args->dirid;
1438 /* set version 1, version 2 could be used too, because stat data
1439 key is the same in both versions */
1440 key.version = KEY_FORMAT_3_5;
1441 key.on_disk_key.k_dir_id = dirino;
1442 key.on_disk_key.k_objectid = inode->i_ino;
1443 key.on_disk_key.k_offset = 0;
1444 key.on_disk_key.k_type = 0;
1446 /* look for the object's stat data */
1447 retval = search_item(inode->i_sb, &key, &path_to_sd);
1448 if (retval == IO_ERROR) {
1449 reiserfs_warning(inode->i_sb,
1450 "vs-13070: reiserfs_read_locked_inode: "
1451 "i/o failure occurred trying to find stat data of %K",
1453 reiserfs_make_bad_inode(inode);
1456 if (retval != ITEM_FOUND) {
1457 /* a stale NFS handle can trigger this without it being an error */
1458 pathrelse(&path_to_sd);
1459 reiserfs_make_bad_inode(inode);
1464 init_inode(inode, &path_to_sd);
1466 /* It is possible that knfsd is trying to access inode of a file
1467 that is being removed from the disk by some other thread. As we
1468 update sd on unlink all that is required is to check for nlink
1469 here. This bug was first found by Sizif when debugging
1470 SquidNG/Butterfly, forgotten, and found again after Philippe
1471 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1473 More logical fix would require changes in fs/inode.c:iput() to
1474 remove inode from hash-table _after_ fs cleaned disk stuff up and
1475 in iget() to return NULL if I_FREEING inode is found in
1477 /* Currently there is one place where it's ok to meet inode with
1478 nlink==0: processing of open-unlinked and half-truncated files
1479 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1480 if ((inode->i_nlink == 0) &&
1481 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
1482 reiserfs_warning(inode->i_sb,
1483 "vs-13075: reiserfs_read_locked_inode: "
1484 "dead inode read from disk %K. "
1485 "This is likely to be race with knfsd. Ignore",
1487 reiserfs_make_bad_inode(inode);
1490 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
1495 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1497 * @inode: inode from hash table to check
1498 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1500 * This function is called by iget5_locked() to distinguish reiserfs inodes
1501 * having the same inode numbers. Such inodes can only exist due to some
1502 * error condition. One of them should be bad. Inodes with identical
1503 * inode numbers (objectids) are distinguished by parent directory ids.
1506 int reiserfs_find_actor(struct inode *inode, void *opaque)
1508 struct reiserfs_iget_args *args;
1511 /* args is already in CPU order */
1512 return (inode->i_ino == args->objectid) &&
1513 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1516 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
1518 struct inode *inode;
1519 struct reiserfs_iget_args args;
1521 args.objectid = key->on_disk_key.k_objectid;
1522 args.dirid = key->on_disk_key.k_dir_id;
1523 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1524 reiserfs_find_actor, reiserfs_init_locked_inode,
1527 return ERR_PTR(-ENOMEM);
1529 if (inode->i_state & I_NEW) {
1530 reiserfs_read_locked_inode(inode, &args);
1531 unlock_new_inode(inode);
1534 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1535 /* either due to i/o error or a stale NFS handle */
1542 struct dentry *reiserfs_get_dentry(struct super_block *sb, void *vobjp)
1544 __u32 *data = vobjp;
1546 struct dentry *result;
1547 struct inode *inode;
1549 key.on_disk_key.k_objectid = data[0];
1550 key.on_disk_key.k_dir_id = data[1];
1551 reiserfs_write_lock(sb);
1552 inode = reiserfs_iget(sb, &key);
1553 if (inode && !IS_ERR(inode) && data[2] != 0 &&
1554 data[2] != inode->i_generation) {
1558 reiserfs_write_unlock(sb);
1560 inode = ERR_PTR(-ESTALE);
1562 return ERR_PTR(PTR_ERR(inode));
1563 result = d_alloc_anon(inode);
1566 return ERR_PTR(-ENOMEM);
1571 struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 * data,
1572 int len, int fhtype,
1573 int (*acceptable) (void *contect,
1574 struct dentry * de),
1577 __u32 obj[3], parent[3];
1579 /* fhtype happens to reflect the number of u32s encoded.
1580 * due to a bug in earlier code, fhtype might indicate there
1581 * are more u32s then actually fitted.
1582 * so if fhtype seems to be more than len, reduce fhtype.
1584 * 2 - objectid + dir_id - legacy support
1585 * 3 - objectid + dir_id + generation
1586 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1587 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1588 * 6 - as above plus generation of directory
1589 * 6 does not fit in NFSv2 handles
1592 if (fhtype != 6 || len != 5)
1593 reiserfs_warning(sb,
1594 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1601 if (fhtype == 3 || fhtype >= 5)
1604 obj[2] = 0; /* generation number */
1607 parent[0] = data[fhtype >= 5 ? 3 : 2];
1608 parent[1] = data[fhtype >= 5 ? 4 : 3];
1610 parent[2] = data[5];
1614 return sb->s_export_op->find_exported_dentry(sb, obj,
1615 fhtype < 4 ? NULL : parent,
1616 acceptable, context);
1619 int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1622 struct inode *inode = dentry->d_inode;
1628 data[0] = inode->i_ino;
1629 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1630 data[2] = inode->i_generation;
1632 /* no room for directory info? return what we've stored so far */
1633 if (maxlen < 5 || !need_parent)
1636 spin_lock(&dentry->d_lock);
1637 inode = dentry->d_parent->d_inode;
1638 data[3] = inode->i_ino;
1639 data[4] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1642 data[5] = inode->i_generation;
1645 spin_unlock(&dentry->d_lock);
1649 /* looks for stat data, then copies fields to it, marks the buffer
1650 containing stat data as dirty */
1651 /* reiserfs inodes are never really dirty, since the dirty inode call
1652 ** always logs them. This call allows the VFS inode marking routines
1653 ** to properly mark inodes for datasync and such, but only actually
1654 ** does something when called for a synchronous update.
1656 int reiserfs_write_inode(struct inode *inode, int do_sync)
1658 struct reiserfs_transaction_handle th;
1659 int jbegin_count = 1;
1661 if (inode->i_sb->s_flags & MS_RDONLY)
1663 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1664 ** these cases are just when the system needs ram, not when the
1665 ** inode needs to reach disk for safety, and they can safely be
1666 ** ignored because the altered inode has already been logged.
1668 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1669 reiserfs_write_lock(inode->i_sb);
1670 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1671 reiserfs_update_sd(&th, inode);
1672 journal_end_sync(&th, inode->i_sb, jbegin_count);
1674 reiserfs_write_unlock(inode->i_sb);
1679 /* stat data of new object is inserted already, this inserts the item
1680 containing "." and ".." entries */
1681 static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1682 struct inode *inode,
1683 struct item_head *ih, struct path *path,
1686 struct super_block *sb = th->t_super;
1687 char empty_dir[EMPTY_DIR_SIZE];
1688 char *body = empty_dir;
1692 BUG_ON(!th->t_trans_id);
1694 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1695 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1696 TYPE_DIRENTRY, 3 /*key length */ );
1698 /* compose item head for new item. Directories consist of items of
1699 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1700 is done by reiserfs_new_inode */
1701 if (old_format_only(sb)) {
1702 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1703 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1705 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1706 ih->ih_key.k_objectid,
1707 INODE_PKEY(dir)->k_dir_id,
1708 INODE_PKEY(dir)->k_objectid);
1710 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1711 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1713 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1714 ih->ih_key.k_objectid,
1715 INODE_PKEY(dir)->k_dir_id,
1716 INODE_PKEY(dir)->k_objectid);
1719 /* look for place in the tree for new item */
1720 retval = search_item(sb, &key, path);
1721 if (retval == IO_ERROR) {
1722 reiserfs_warning(sb, "vs-13080: reiserfs_new_directory: "
1723 "i/o failure occurred creating new directory");
1726 if (retval == ITEM_FOUND) {
1728 reiserfs_warning(sb, "vs-13070: reiserfs_new_directory: "
1729 "object with this key exists (%k)",
1734 /* insert item, that is empty directory item */
1735 return reiserfs_insert_item(th, path, &key, ih, inode, body);
1738 /* stat data of object has been inserted, this inserts the item
1739 containing the body of symlink */
1740 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1741 struct item_head *ih,
1742 struct path *path, const char *symname,
1745 struct super_block *sb = th->t_super;
1749 BUG_ON(!th->t_trans_id);
1751 _make_cpu_key(&key, KEY_FORMAT_3_5,
1752 le32_to_cpu(ih->ih_key.k_dir_id),
1753 le32_to_cpu(ih->ih_key.k_objectid),
1754 1, TYPE_DIRECT, 3 /*key length */ );
1756 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1757 0 /*free_space */ );
1759 /* look for place in the tree for new item */
1760 retval = search_item(sb, &key, path);
1761 if (retval == IO_ERROR) {
1762 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlinik: "
1763 "i/o failure occurred creating new symlink");
1766 if (retval == ITEM_FOUND) {
1768 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlink: "
1769 "object with this key exists (%k)",
1774 /* insert item, that is body of symlink */
1775 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
1778 /* inserts the stat data into the tree, and then calls
1779 reiserfs_new_directory (to insert ".", ".." item if new object is
1780 directory) or reiserfs_new_symlink (to insert symlink body if new
1781 object is symlink) or nothing (if new object is regular file)
1783 NOTE! uid and gid must already be set in the inode. If we return
1784 non-zero due to an error, we have to drop the quota previously allocated
1785 for the fresh inode. This can only be done outside a transaction, so
1786 if we return non-zero, we also end the transaction. */
1787 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1788 struct inode *dir, int mode, const char *symname,
1789 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1790 strlen (symname) for symlinks) */
1791 loff_t i_size, struct dentry *dentry,
1792 struct inode *inode)
1794 struct super_block *sb;
1795 INITIALIZE_PATH(path_to_key);
1797 struct item_head ih;
1798 struct stat_data sd;
1802 BUG_ON(!th->t_trans_id);
1804 if (DLIMIT_ALLOC_INODE(inode)) {
1806 goto out_bad_dlimit;
1808 if (DQUOT_ALLOC_INODE(inode)) {
1812 if (!dir || !dir->i_nlink) {
1819 /* item head of new item */
1820 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1821 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1822 if (!ih.ih_key.k_objectid) {
1826 if (old_format_only(sb))
1827 /* not a perfect generation count, as object ids can be reused, but
1828 ** this is as good as reiserfs can do right now.
1829 ** note that the private part of inode isn't filled in yet, we have
1830 ** to use the directory.
1832 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1834 #if defined( USE_INODE_GENERATION_COUNTER )
1835 inode->i_generation =
1836 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1838 inode->i_generation = ++event;
1841 /* fill stat data */
1842 inode->i_nlink = (S_ISDIR(mode) ? 2 : 1);
1844 /* uid and gid must already be set by the caller for quota init */
1846 /* symlink cannot be immutable or append only, right? */
1847 if (S_ISLNK(inode->i_mode))
1848 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1850 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1851 inode->i_size = i_size;
1852 inode->i_blocks = 0;
1854 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1855 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1857 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1858 REISERFS_I(inode)->i_flags = 0;
1859 REISERFS_I(inode)->i_prealloc_block = 0;
1860 REISERFS_I(inode)->i_prealloc_count = 0;
1861 REISERFS_I(inode)->i_trans_id = 0;
1862 REISERFS_I(inode)->i_jl = NULL;
1863 REISERFS_I(inode)->i_attrs =
1864 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1865 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
1866 REISERFS_I(inode)->i_acl_access = NULL;
1867 REISERFS_I(inode)->i_acl_default = NULL;
1868 init_rwsem(&REISERFS_I(inode)->xattr_sem);
1870 if (old_format_only(sb))
1871 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1872 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1874 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1875 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1877 /* key to search for correct place for new stat data */
1878 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1879 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1880 TYPE_STAT_DATA, 3 /*key length */ );
1882 /* find proper place for inserting of stat data */
1883 retval = search_item(sb, &key, &path_to_key);
1884 if (retval == IO_ERROR) {
1888 if (retval == ITEM_FOUND) {
1889 pathrelse(&path_to_key);
1893 if (old_format_only(sb)) {
1894 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1895 pathrelse(&path_to_key);
1896 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1900 inode2sd_v1(&sd, inode, inode->i_size);
1902 inode2sd(&sd, inode, inode->i_size);
1904 // these do not go to on-disk stat data
1905 inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
1906 inode->i_blksize = reiserfs_default_io_size;
1908 // store in in-core inode the key of stat data and version all
1909 // object items will have (directory items will have old offset
1910 // format, other new objects will consist of new items)
1911 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1912 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1913 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1915 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1916 if (old_format_only(sb))
1917 set_inode_sd_version(inode, STAT_DATA_V1);
1919 set_inode_sd_version(inode, STAT_DATA_V2);
1921 /* insert the stat data into the tree */
1922 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1923 if (REISERFS_I(dir)->new_packing_locality)
1924 th->displace_new_blocks = 1;
1927 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1931 reiserfs_check_path(&path_to_key);
1934 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1935 if (!th->displace_new_blocks)
1936 REISERFS_I(dir)->new_packing_locality = 0;
1938 if (S_ISDIR(mode)) {
1939 /* insert item with "." and ".." */
1941 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1944 if (S_ISLNK(mode)) {
1945 /* insert body of symlink */
1946 if (!old_format_only(sb))
1947 i_size = ROUND_UP(i_size);
1949 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1954 reiserfs_check_path(&path_to_key);
1955 journal_end(th, th->t_super, th->t_blocks_allocated);
1956 goto out_inserted_sd;
1959 /* XXX CHECK THIS */
1960 if (reiserfs_posixacl(inode->i_sb)) {
1961 retval = reiserfs_inherit_default_acl(dir, dentry, inode);
1964 reiserfs_check_path(&path_to_key);
1965 journal_end(th, th->t_super, th->t_blocks_allocated);
1966 goto out_inserted_sd;
1968 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1969 reiserfs_warning(inode->i_sb, "ACLs aren't enabled in the fs, "
1970 "but vfs thinks they are!");
1971 } else if (is_reiserfs_priv_object(dir)) {
1972 reiserfs_mark_inode_private(inode);
1975 insert_inode_hash(inode);
1976 reiserfs_update_sd(th, inode);
1977 reiserfs_check_path(&path_to_key);
1981 /* it looks like you can easily compress these two goto targets into
1982 * one. Keeping it like this doesn't actually hurt anything, and they
1983 * are place holders for what the quota code actually needs.
1986 /* Invalidate the object, nothing was inserted yet */
1987 INODE_PKEY(inode)->k_objectid = 0;
1989 /* Quota change must be inside a transaction for journaling */
1990 DQUOT_FREE_INODE(inode);
1993 DLIMIT_FREE_INODE(inode);
1996 journal_end(th, th->t_super, th->t_blocks_allocated);
1997 /* Drop can be outside and it needs more credits so it's better to have it outside */
1999 inode->i_flags |= S_NOQUOTA;
2000 make_bad_inode(inode);
2004 th->t_trans_id = 0; /* so the caller can't use this handle later */
2006 /* If we were inheriting an ACL, we need to release the lock so that
2007 * iput doesn't deadlock in reiserfs_delete_xattrs. The locking
2008 * code really needs to be reworked, but this will take care of it
2009 * for now. -jeffm */
2010 if (REISERFS_I(dir)->i_acl_default && !IS_ERR(REISERFS_I(dir)->i_acl_default)) {
2011 reiserfs_write_unlock_xattrs(dir->i_sb);
2013 reiserfs_write_lock_xattrs(dir->i_sb);
2020 ** finds the tail page in the page cache,
2021 ** reads the last block in.
2023 ** On success, page_result is set to a locked, pinned page, and bh_result
2024 ** is set to an up to date buffer for the last block in the file. returns 0.
2026 ** tail conversion is not done, so bh_result might not be valid for writing
2027 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2028 ** trying to write the block.
2030 ** on failure, nonzero is returned, page_result and bh_result are untouched.
2032 static int grab_tail_page(struct inode *p_s_inode,
2033 struct page **page_result,
2034 struct buffer_head **bh_result)
2037 /* we want the page with the last byte in the file,
2038 ** not the page that will hold the next byte for appending
2040 unsigned long index = (p_s_inode->i_size - 1) >> PAGE_CACHE_SHIFT;
2041 unsigned long pos = 0;
2042 unsigned long start = 0;
2043 unsigned long blocksize = p_s_inode->i_sb->s_blocksize;
2044 unsigned long offset = (p_s_inode->i_size) & (PAGE_CACHE_SIZE - 1);
2045 struct buffer_head *bh;
2046 struct buffer_head *head;
2050 /* we know that we are only called with inode->i_size > 0.
2051 ** we also know that a file tail can never be as big as a block
2052 ** If i_size % blocksize == 0, our file is currently block aligned
2053 ** and it won't need converting or zeroing after a truncate.
2055 if ((offset & (blocksize - 1)) == 0) {
2058 page = grab_cache_page(p_s_inode->i_mapping, index);
2063 /* start within the page of the last block in the file */
2064 start = (offset / blocksize) * blocksize;
2066 error = block_prepare_write(page, start, offset,
2067 reiserfs_get_block_create_0);
2071 head = page_buffers(page);
2077 bh = bh->b_this_page;
2079 } while (bh != head);
2081 if (!buffer_uptodate(bh)) {
2082 /* note, this should never happen, prepare_write should
2083 ** be taking care of this for us. If the buffer isn't up to date,
2084 ** I've screwed up the code to find the buffer, or the code to
2085 ** call prepare_write
2087 reiserfs_warning(p_s_inode->i_sb,
2088 "clm-6000: error reading block %lu on dev %s",
2090 reiserfs_bdevname(p_s_inode->i_sb));
2095 *page_result = page;
2102 page_cache_release(page);
2107 ** vfs version of truncate file. Must NOT be called with
2108 ** a transaction already started.
2110 ** some code taken from block_truncate_page
2112 int reiserfs_truncate_file(struct inode *p_s_inode, int update_timestamps)
2114 struct reiserfs_transaction_handle th;
2115 /* we want the offset for the first byte after the end of the file */
2116 unsigned long offset = p_s_inode->i_size & (PAGE_CACHE_SIZE - 1);
2117 unsigned blocksize = p_s_inode->i_sb->s_blocksize;
2119 struct page *page = NULL;
2121 struct buffer_head *bh = NULL;
2124 reiserfs_write_lock(p_s_inode->i_sb);
2126 if (p_s_inode->i_size > 0) {
2127 if ((error = grab_tail_page(p_s_inode, &page, &bh))) {
2128 // -ENOENT means we truncated past the end of the file,
2129 // and get_block_create_0 could not find a block to read in,
2131 if (error != -ENOENT)
2132 reiserfs_warning(p_s_inode->i_sb,
2133 "clm-6001: grab_tail_page failed %d",
2140 /* so, if page != NULL, we have a buffer head for the offset at
2141 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2142 ** then we have an unformatted node. Otherwise, we have a direct item,
2143 ** and no zeroing is required on disk. We zero after the truncate,
2144 ** because the truncate might pack the item anyway
2145 ** (it will unmap bh if it packs).
2147 /* it is enough to reserve space in transaction for 2 balancings:
2148 one for "save" link adding and another for the first
2149 cut_from_item. 1 is for update_sd */
2150 error = journal_begin(&th, p_s_inode->i_sb,
2151 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2154 reiserfs_update_inode_transaction(p_s_inode);
2155 if (update_timestamps)
2156 /* we are doing real truncate: if the system crashes before the last
2157 transaction of truncating gets committed - on reboot the file
2158 either appears truncated properly or not truncated at all */
2159 add_save_link(&th, p_s_inode, 1);
2160 err2 = reiserfs_do_truncate(&th, p_s_inode, page, update_timestamps);
2162 journal_end(&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2166 /* check reiserfs_do_truncate after ending the transaction */
2172 if (update_timestamps) {
2173 error = remove_save_link(p_s_inode, 1 /* truncate */ );
2179 length = offset & (blocksize - 1);
2180 /* if we are not on a block boundary */
2184 length = blocksize - length;
2185 kaddr = kmap_atomic(page, KM_USER0);
2186 memset(kaddr + offset, 0, length);
2187 flush_dcache_page(page);
2188 kunmap_atomic(kaddr, KM_USER0);
2189 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2190 mark_buffer_dirty(bh);
2194 page_cache_release(page);
2197 reiserfs_write_unlock(p_s_inode->i_sb);
2202 page_cache_release(page);
2204 reiserfs_write_unlock(p_s_inode->i_sb);
2208 static int map_block_for_writepage(struct inode *inode,
2209 struct buffer_head *bh_result,
2210 unsigned long block)
2212 struct reiserfs_transaction_handle th;
2214 struct item_head tmp_ih;
2215 struct item_head *ih;
2216 struct buffer_head *bh;
2219 INITIALIZE_PATH(path);
2221 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
2222 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
2224 int use_get_block = 0;
2225 int bytes_copied = 0;
2227 int trans_running = 0;
2229 /* catch places below that try to log something without starting a trans */
2232 if (!buffer_uptodate(bh_result)) {
2236 kmap(bh_result->b_page);
2238 reiserfs_write_lock(inode->i_sb);
2239 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
2242 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2243 if (retval != POSITION_FOUND) {
2248 bh = get_last_bh(&path);
2250 item = get_item(&path);
2251 pos_in_item = path.pos_in_item;
2253 /* we've found an unformatted node */
2254 if (indirect_item_found(retval, ih)) {
2255 if (bytes_copied > 0) {
2256 reiserfs_warning(inode->i_sb,
2257 "clm-6002: bytes_copied %d",
2260 if (!get_block_num(item, pos_in_item)) {
2261 /* crap, we are writing to a hole */
2265 set_block_dev_mapped(bh_result,
2266 get_block_num(item, pos_in_item), inode);
2267 } else if (is_direct_le_ih(ih)) {
2269 p = page_address(bh_result->b_page);
2270 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2271 copy_size = ih_item_len(ih) - pos_in_item;
2273 fs_gen = get_generation(inode->i_sb);
2274 copy_item_head(&tmp_ih, ih);
2276 if (!trans_running) {
2277 /* vs-3050 is gone, no need to drop the path */
2278 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2281 reiserfs_update_inode_transaction(inode);
2283 if (fs_changed(fs_gen, inode->i_sb)
2284 && item_moved(&tmp_ih, &path)) {
2285 reiserfs_restore_prepared_buffer(inode->i_sb,
2291 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2293 if (fs_changed(fs_gen, inode->i_sb)
2294 && item_moved(&tmp_ih, &path)) {
2295 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2299 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2302 journal_mark_dirty(&th, inode->i_sb, bh);
2303 bytes_copied += copy_size;
2304 set_block_dev_mapped(bh_result, 0, inode);
2306 /* are there still bytes left? */
2307 if (bytes_copied < bh_result->b_size &&
2308 (byte_offset + bytes_copied) < inode->i_size) {
2309 set_cpu_key_k_offset(&key,
2310 cpu_key_k_offset(&key) +
2315 reiserfs_warning(inode->i_sb,
2316 "clm-6003: bad item inode %lu, device %s",
2317 inode->i_ino, reiserfs_bdevname(inode->i_sb));
2325 if (trans_running) {
2326 int err = journal_end(&th, inode->i_sb, jbegin_count);
2331 reiserfs_write_unlock(inode->i_sb);
2333 /* this is where we fill in holes in the file. */
2334 if (use_get_block) {
2335 retval = reiserfs_get_block(inode, block, bh_result,
2336 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2337 | GET_BLOCK_NO_DANGLE);
2339 if (!buffer_mapped(bh_result)
2340 || bh_result->b_blocknr == 0) {
2341 /* get_block failed to find a mapped unformatted node. */
2347 kunmap(bh_result->b_page);
2349 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2350 /* we've copied data from the page into the direct item, so the
2351 * buffer in the page is now clean, mark it to reflect that.
2353 lock_buffer(bh_result);
2354 clear_buffer_dirty(bh_result);
2355 unlock_buffer(bh_result);
2361 * mason@suse.com: updated in 2.5.54 to follow the same general io
2362 * start/recovery path as __block_write_full_page, along with special
2363 * code to handle reiserfs tails.
2365 static int reiserfs_write_full_page(struct page *page,
2366 struct writeback_control *wbc)
2368 struct inode *inode = page->mapping->host;
2369 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2371 unsigned long block;
2372 struct buffer_head *head, *bh;
2375 int checked = PageChecked(page);
2376 struct reiserfs_transaction_handle th;
2377 struct super_block *s = inode->i_sb;
2378 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2381 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2382 if (checked && (current->flags & PF_MEMALLOC)) {
2383 redirty_page_for_writepage(wbc, page);
2388 /* The page dirty bit is cleared before writepage is called, which
2389 * means we have to tell create_empty_buffers to make dirty buffers
2390 * The page really should be up to date at this point, so tossing
2391 * in the BH_Uptodate is just a sanity check.
2393 if (!page_has_buffers(page)) {
2394 create_empty_buffers(page, s->s_blocksize,
2395 (1 << BH_Dirty) | (1 << BH_Uptodate));
2397 head = page_buffers(page);
2399 /* last page in the file, zero out any contents past the
2400 ** last byte in the file
2402 if (page->index >= end_index) {
2404 unsigned last_offset;
2406 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2407 /* no file contents in this page */
2408 if (page->index >= end_index + 1 || !last_offset) {
2412 kaddr = kmap_atomic(page, KM_USER0);
2413 memset(kaddr + last_offset, 0, PAGE_CACHE_SIZE - last_offset);
2414 flush_dcache_page(page);
2415 kunmap_atomic(kaddr, KM_USER0);
2418 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
2419 /* first map all the buffers, logging any direct items we find */
2421 if ((checked || buffer_dirty(bh)) && (!buffer_mapped(bh) ||
2425 /* not mapped yet, or it points to a direct item, search
2426 * the btree for the mapping info, and log any direct
2429 if ((error = map_block_for_writepage(inode, bh, block))) {
2433 bh = bh->b_this_page;
2435 } while (bh != head);
2438 * we start the transaction after map_block_for_writepage,
2439 * because it can create holes in the file (an unbounded operation).
2440 * starting it here, we can make a reliable estimate for how many
2441 * blocks we're going to log
2444 ClearPageChecked(page);
2445 reiserfs_write_lock(s);
2446 error = journal_begin(&th, s, bh_per_page + 1);
2448 reiserfs_write_unlock(s);
2451 reiserfs_update_inode_transaction(inode);
2453 /* now go through and lock any dirty buffers on the page */
2456 if (!buffer_mapped(bh))
2458 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2462 reiserfs_prepare_for_journal(s, bh, 1);
2463 journal_mark_dirty(&th, s, bh);
2466 /* from this point on, we know the buffer is mapped to a
2467 * real block and not a direct item
2469 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2472 if (test_set_buffer_locked(bh)) {
2473 redirty_page_for_writepage(wbc, page);
2477 if (test_clear_buffer_dirty(bh)) {
2478 mark_buffer_async_write(bh);
2482 } while ((bh = bh->b_this_page) != head);
2485 error = journal_end(&th, s, bh_per_page + 1);
2486 reiserfs_write_unlock(s);
2490 BUG_ON(PageWriteback(page));
2491 set_page_writeback(page);
2495 * since any buffer might be the only dirty buffer on the page,
2496 * the first submit_bh can bring the page out of writeback.
2497 * be careful with the buffers.
2500 struct buffer_head *next = bh->b_this_page;
2501 if (buffer_async_write(bh)) {
2502 submit_bh(WRITE, bh);
2507 } while (bh != head);
2513 * if this page only had a direct item, it is very possible for
2514 * no io to be required without there being an error. Or,
2515 * someone else could have locked them and sent them down the
2516 * pipe without locking the page
2520 if (!buffer_uptodate(bh)) {
2524 bh = bh->b_this_page;
2525 } while (bh != head);
2527 SetPageUptodate(page);
2528 end_page_writeback(page);
2533 /* catches various errors, we need to make sure any valid dirty blocks
2534 * get to the media. The page is currently locked and not marked for
2537 ClearPageUptodate(page);
2541 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2543 mark_buffer_async_write(bh);
2546 * clear any dirty bits that might have come from getting
2547 * attached to a dirty page
2549 clear_buffer_dirty(bh);
2551 bh = bh->b_this_page;
2552 } while (bh != head);
2554 BUG_ON(PageWriteback(page));
2555 set_page_writeback(page);
2558 struct buffer_head *next = bh->b_this_page;
2559 if (buffer_async_write(bh)) {
2560 clear_buffer_dirty(bh);
2561 submit_bh(WRITE, bh);
2566 } while (bh != head);
2570 static int reiserfs_readpage(struct file *f, struct page *page)
2572 return block_read_full_page(page, reiserfs_get_block);
2575 static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
2577 struct inode *inode = page->mapping->host;
2578 reiserfs_wait_on_write_block(inode->i_sb);
2579 return reiserfs_write_full_page(page, wbc);
2582 static int reiserfs_prepare_write(struct file *f, struct page *page,
2583 unsigned from, unsigned to)
2585 struct inode *inode = page->mapping->host;
2589 reiserfs_wait_on_write_block(inode->i_sb);
2590 fix_tail_page_for_writing(page);
2591 if (reiserfs_transaction_running(inode->i_sb)) {
2592 struct reiserfs_transaction_handle *th;
2593 th = (struct reiserfs_transaction_handle *)current->
2595 BUG_ON(!th->t_refcount);
2596 BUG_ON(!th->t_trans_id);
2597 old_ref = th->t_refcount;
2601 ret = block_prepare_write(page, from, to, reiserfs_get_block);
2602 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2603 struct reiserfs_transaction_handle *th = current->journal_info;
2604 /* this gets a little ugly. If reiserfs_get_block returned an
2605 * error and left a transacstion running, we've got to close it,
2606 * and we've got to free handle if it was a persistent transaction.
2608 * But, if we had nested into an existing transaction, we need
2609 * to just drop the ref count on the handle.
2611 * If old_ref == 0, the transaction is from reiserfs_get_block,
2612 * and it was a persistent trans. Otherwise, it was nested above.
2614 if (th->t_refcount > old_ref) {
2619 reiserfs_write_lock(inode->i_sb);
2620 err = reiserfs_end_persistent_transaction(th);
2621 reiserfs_write_unlock(inode->i_sb);
2631 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2633 return generic_block_bmap(as, block, reiserfs_bmap);
2636 static int reiserfs_commit_write(struct file *f, struct page *page,
2637 unsigned from, unsigned to)
2639 struct inode *inode = page->mapping->host;
2640 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2643 struct reiserfs_transaction_handle *th = NULL;
2645 reiserfs_wait_on_write_block(inode->i_sb);
2646 if (reiserfs_transaction_running(inode->i_sb)) {
2647 th = current->journal_info;
2649 reiserfs_commit_page(inode, page, from, to);
2651 /* generic_commit_write does this for us, but does not update the
2652 ** transaction tracking stuff when the size changes. So, we have
2653 ** to do the i_size updates here.
2655 if (pos > inode->i_size) {
2656 struct reiserfs_transaction_handle myth;
2657 reiserfs_write_lock(inode->i_sb);
2658 /* If the file have grown beyond the border where it
2659 can have a tail, unmark it as needing a tail
2661 if ((have_large_tails(inode->i_sb)
2662 && inode->i_size > i_block_size(inode) * 4)
2663 || (have_small_tails(inode->i_sb)
2664 && inode->i_size > i_block_size(inode)))
2665 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2667 ret = journal_begin(&myth, inode->i_sb, 1);
2669 reiserfs_write_unlock(inode->i_sb);
2672 reiserfs_update_inode_transaction(inode);
2673 inode->i_size = pos;
2675 * this will just nest into our transaction. It's important
2676 * to use mark_inode_dirty so the inode gets pushed around on the
2677 * dirty lists, and so that O_SYNC works as expected
2679 mark_inode_dirty(inode);
2680 reiserfs_update_sd(&myth, inode);
2682 ret = journal_end(&myth, inode->i_sb, 1);
2683 reiserfs_write_unlock(inode->i_sb);
2688 reiserfs_write_lock(inode->i_sb);
2690 mark_inode_dirty(inode);
2691 ret = reiserfs_end_persistent_transaction(th);
2692 reiserfs_write_unlock(inode->i_sb);
2702 reiserfs_write_lock(inode->i_sb);
2704 reiserfs_update_sd(th, inode);
2705 ret = reiserfs_end_persistent_transaction(th);
2706 reiserfs_write_unlock(inode->i_sb);
2712 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2714 if (reiserfs_attrs(inode->i_sb)) {
2715 if (sd_attrs & REISERFS_SYNC_FL)
2716 inode->i_flags |= S_SYNC;
2718 inode->i_flags &= ~S_SYNC;
2719 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2720 inode->i_flags |= S_IMMUTABLE;
2722 inode->i_flags &= ~S_IMMUTABLE;
2723 if (sd_attrs & REISERFS_IUNLINK_FL)
2724 inode->i_flags |= S_IUNLINK;
2726 inode->i_flags &= ~S_IUNLINK;
2727 if (sd_attrs & REISERFS_BARRIER_FL)
2728 inode->i_flags |= S_BARRIER;
2730 inode->i_flags &= ~S_BARRIER;
2731 if (sd_attrs & REISERFS_APPEND_FL)
2732 inode->i_flags |= S_APPEND;
2734 inode->i_flags &= ~S_APPEND;
2735 if (sd_attrs & REISERFS_NOATIME_FL)
2736 inode->i_flags |= S_NOATIME;
2738 inode->i_flags &= ~S_NOATIME;
2739 if (sd_attrs & REISERFS_NOTAIL_FL)
2740 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2742 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2746 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
2748 if (reiserfs_attrs(inode->i_sb)) {
2749 if (inode->i_flags & S_IMMUTABLE)
2750 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2752 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2753 if (inode->i_flags & S_IUNLINK)
2754 *sd_attrs |= REISERFS_IUNLINK_FL;
2756 *sd_attrs &= ~REISERFS_IUNLINK_FL;
2757 if (inode->i_flags & S_BARRIER)
2758 *sd_attrs |= REISERFS_BARRIER_FL;
2760 *sd_attrs &= ~REISERFS_BARRIER_FL;
2761 if (inode->i_flags & S_SYNC)
2762 *sd_attrs |= REISERFS_SYNC_FL;
2764 *sd_attrs &= ~REISERFS_SYNC_FL;
2765 if (inode->i_flags & S_NOATIME)
2766 *sd_attrs |= REISERFS_NOATIME_FL;
2768 *sd_attrs &= ~REISERFS_NOATIME_FL;
2769 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
2770 *sd_attrs |= REISERFS_NOTAIL_FL;
2772 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2776 /* decide if this buffer needs to stay around for data logging or ordered
2779 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2782 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2785 spin_lock(&j->j_dirty_buffers_lock);
2786 if (!buffer_mapped(bh)) {
2789 /* the page is locked, and the only places that log a data buffer
2790 * also lock the page.
2792 if (reiserfs_file_data_log(inode)) {
2794 * very conservative, leave the buffer pinned if
2795 * anyone might need it.
2797 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2800 } else if (buffer_dirty(bh)) {
2801 struct reiserfs_journal_list *jl;
2802 struct reiserfs_jh *jh = bh->b_private;
2804 /* why is this safe?
2805 * reiserfs_setattr updates i_size in the on disk
2806 * stat data before allowing vmtruncate to be called.
2808 * If buffer was put onto the ordered list for this
2809 * transaction, we know for sure either this transaction
2810 * or an older one already has updated i_size on disk,
2811 * and this ordered data won't be referenced in the file
2814 * if the buffer was put onto the ordered list for an older
2815 * transaction, we need to leave it around
2817 if (jh && (jl = jh->jl)
2818 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2822 if (ret && bh->b_private) {
2823 reiserfs_free_jh(bh);
2825 spin_unlock(&j->j_dirty_buffers_lock);
2830 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2831 static void reiserfs_invalidatepage(struct page *page, unsigned long offset)
2833 struct buffer_head *head, *bh, *next;
2834 struct inode *inode = page->mapping->host;
2835 unsigned int curr_off = 0;
2838 BUG_ON(!PageLocked(page));
2841 ClearPageChecked(page);
2843 if (!page_has_buffers(page))
2846 head = page_buffers(page);
2849 unsigned int next_off = curr_off + bh->b_size;
2850 next = bh->b_this_page;
2853 * is this block fully invalidated?
2855 if (offset <= curr_off) {
2856 if (invalidatepage_can_drop(inode, bh))
2857 reiserfs_unmap_buffer(bh);
2861 curr_off = next_off;
2863 } while (bh != head);
2866 * We release buffers only if the entire page is being invalidated.
2867 * The get_block cached value has been unconditionally invalidated,
2868 * so real IO is not possible anymore.
2870 if (!offset && ret) {
2871 ret = try_to_release_page(page, 0);
2872 /* maybe should BUG_ON(!ret); - neilb */
2878 static int reiserfs_set_page_dirty(struct page *page)
2880 struct inode *inode = page->mapping->host;
2881 if (reiserfs_file_data_log(inode)) {
2882 SetPageChecked(page);
2883 return __set_page_dirty_nobuffers(page);
2885 return __set_page_dirty_buffers(page);
2889 * Returns 1 if the page's buffers were dropped. The page is locked.
2891 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
2892 * in the buffers at page_buffers(page).
2894 * even in -o notail mode, we can't be sure an old mount without -o notail
2895 * didn't create files with tails.
2897 static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
2899 struct inode *inode = page->mapping->host;
2900 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2901 struct buffer_head *head;
2902 struct buffer_head *bh;
2905 WARN_ON(PageChecked(page));
2906 spin_lock(&j->j_dirty_buffers_lock);
2907 head = page_buffers(page);
2910 if (bh->b_private) {
2911 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
2912 reiserfs_free_jh(bh);
2918 bh = bh->b_this_page;
2919 } while (bh != head);
2921 ret = try_to_free_buffers(page);
2922 spin_unlock(&j->j_dirty_buffers_lock);
2926 /* We thank Mingming Cao for helping us understand in great detail what
2927 to do in this section of the code. */
2928 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
2929 const struct iovec *iov, loff_t offset,
2930 unsigned long nr_segs)
2932 struct file *file = iocb->ki_filp;
2933 struct inode *inode = file->f_mapping->host;
2935 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2937 reiserfs_get_blocks_direct_io, NULL);
2940 int reiserfs_sync_flags(struct inode *inode)
2942 u16 oldflags, newflags;
2944 oldflags = REISERFS_I(inode)->i_attrs;
2945 newflags = oldflags;
2946 i_attrs_to_sd_attrs(inode, &newflags);
2948 if (oldflags ^ newflags) {
2949 REISERFS_I(inode)->i_attrs = newflags;
2950 inode->i_ctime = CURRENT_TIME_SEC;
2951 mark_inode_dirty(inode);
2956 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
2958 struct inode *inode = dentry->d_inode;
2960 unsigned int ia_valid = attr->ia_valid;
2961 reiserfs_write_lock(inode->i_sb);
2962 if (attr->ia_valid & ATTR_SIZE) {
2963 /* version 2 items will be caught by the s_maxbytes check
2964 ** done for us in vmtruncate
2966 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
2967 attr->ia_size > MAX_NON_LFS) {
2971 /* fill in hole pointers in the expanding truncate case. */
2972 if (attr->ia_size > inode->i_size) {
2973 error = generic_cont_expand(inode, attr->ia_size);
2974 if (REISERFS_I(inode)->i_prealloc_count > 0) {
2976 struct reiserfs_transaction_handle th;
2977 /* we're changing at most 2 bitmaps, inode + super */
2978 err = journal_begin(&th, inode->i_sb, 4);
2980 reiserfs_discard_prealloc(&th, inode);
2981 err = journal_end(&th, inode->i_sb, 4);
2991 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
2992 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
2993 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
2994 /* stat data of format v3.5 has 16 bit uid and gid */
2999 error = inode_change_ok(inode, attr);
3002 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
3003 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid) ||
3004 (ia_valid & ATTR_XID && attr->ia_xid != inode->i_xid)) {
3005 error = reiserfs_chown_xattrs(inode, attr);
3008 struct reiserfs_transaction_handle th;
3011 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3012 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3015 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3017 journal_begin(&th, inode->i_sb,
3022 DQUOT_TRANSFER(inode, attr) ? -EDQUOT : 0;
3024 journal_end(&th, inode->i_sb,
3028 /* Update corresponding info in inode so that everything is in
3029 * one transaction */
3030 if (attr->ia_valid & ATTR_UID)
3031 inode->i_uid = attr->ia_uid;
3032 if (attr->ia_valid & ATTR_GID)
3033 inode->i_gid = attr->ia_gid;
3034 if ((attr->ia_valid & ATTR_XID) &&
3036 inode->i_xid = attr->ia_xid;
3037 mark_inode_dirty(inode);
3039 journal_end(&th, inode->i_sb, jbegin_count);
3043 error = inode_setattr(inode, attr);
3046 if (!error && reiserfs_posixacl(inode->i_sb)) {
3047 if (attr->ia_valid & ATTR_MODE)
3048 error = reiserfs_acl_chmod(inode);
3052 reiserfs_write_unlock(inode->i_sb);
3056 struct address_space_operations reiserfs_address_space_operations = {
3057 .writepage = reiserfs_writepage,
3058 .readpage = reiserfs_readpage,
3059 .readpages = reiserfs_readpages,
3060 .releasepage = reiserfs_releasepage,
3061 .invalidatepage = reiserfs_invalidatepage,
3062 .sync_page = block_sync_page,
3063 .prepare_write = reiserfs_prepare_write,
3064 .commit_write = reiserfs_commit_write,
3065 .bmap = reiserfs_aop_bmap,
3066 .direct_IO = reiserfs_direct_IO,
3067 .set_page_dirty = reiserfs_set_page_dirty,