2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
5 #include <linux/time.h>
7 #include <linux/reiserfs_fs.h>
8 #include <linux/reiserfs_acl.h>
9 #include <linux/reiserfs_xattr.h>
10 #include <linux/smp_lock.h>
11 #include <linux/pagemap.h>
12 #include <linux/highmem.h>
13 #include <asm/uaccess.h>
14 #include <asm/unaligned.h>
15 #include <linux/buffer_head.h>
16 #include <linux/mpage.h>
17 #include <linux/writeback.h>
18 #include <linux/quotaops.h>
19 #include <linux/vs_dlimit.h>
20 #include <linux/vserver/xid.h>
22 static int reiserfs_commit_write(struct file *f, struct page *page,
23 unsigned from, unsigned to);
24 static int reiserfs_prepare_write(struct file *f, struct page *page,
25 unsigned from, unsigned to);
27 void reiserfs_delete_inode(struct inode *inode)
29 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
31 JOURNAL_PER_BALANCE_CNT * 2 +
32 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
33 struct reiserfs_transaction_handle th;
36 truncate_inode_pages(&inode->i_data, 0);
38 reiserfs_write_lock(inode->i_sb);
40 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
41 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
42 reiserfs_delete_xattrs(inode);
44 if (journal_begin(&th, inode->i_sb, jbegin_count))
46 reiserfs_update_inode_transaction(inode);
48 err = reiserfs_delete_object(&th, inode);
50 /* Do quota update inside a transaction for journaled quotas. We must do that
51 * after delete_object so that quota updates go into the same transaction as
52 * stat data deletion */
54 DQUOT_FREE_INODE(inode);
55 DLIMIT_FREE_INODE(inode);
57 if (journal_end(&th, inode->i_sb, jbegin_count))
60 /* check return value from reiserfs_delete_object after
61 * ending the transaction
66 /* all items of file are deleted, so we can remove "save" link */
67 remove_save_link(inode, 0 /* not truncate */ ); /* we can't do anything
68 * about an error here */
70 /* no object items are in the tree */
74 clear_inode(inode); /* note this must go after the journal_end to prevent deadlock */
76 reiserfs_write_unlock(inode->i_sb);
79 static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
80 __u32 objectid, loff_t offset, int type, int length)
82 key->version = version;
84 key->on_disk_key.k_dir_id = dirid;
85 key->on_disk_key.k_objectid = objectid;
86 set_cpu_key_k_offset(key, offset);
87 set_cpu_key_k_type(key, type);
88 key->key_length = length;
91 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
92 offset and type of key */
93 void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
96 _make_cpu_key(key, get_inode_item_key_version(inode),
97 le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
98 le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
103 // when key is 0, do not set version and short key
105 inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
107 loff_t offset, int type, int length,
108 int entry_count /*or ih_free_space */ )
111 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
112 ih->ih_key.k_objectid =
113 cpu_to_le32(key->on_disk_key.k_objectid);
115 put_ih_version(ih, version);
116 set_le_ih_k_offset(ih, offset);
117 set_le_ih_k_type(ih, type);
118 put_ih_item_len(ih, length);
119 /* set_ih_free_space (ih, 0); */
120 // for directory items it is entry count, for directs and stat
121 // datas - 0xffff, for indirects - 0
122 put_ih_entry_count(ih, entry_count);
126 // FIXME: we might cache recently accessed indirect item
128 // Ugh. Not too eager for that....
129 // I cut the code until such time as I see a convincing argument (benchmark).
130 // I don't want a bloated inode struct..., and I don't like code complexity....
132 /* cutting the code is fine, since it really isn't in use yet and is easy
133 ** to add back in. But, Vladimir has a really good idea here. Think
134 ** about what happens for reading a file. For each page,
135 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
136 ** an indirect item. This indirect item has X number of pointers, where
137 ** X is a big number if we've done the block allocation right. But,
138 ** we only use one or two of these pointers during each call to readpage,
139 ** needlessly researching again later on.
141 ** The size of the cache could be dynamic based on the size of the file.
143 ** I'd also like to see us cache the location the stat data item, since
144 ** we are needlessly researching for that frequently.
149 /* If this page has a file tail in it, and
150 ** it was read in by get_block_create_0, the page data is valid,
151 ** but tail is still sitting in a direct item, and we can't write to
152 ** it. So, look through this page, and check all the mapped buffers
153 ** to make sure they have valid block numbers. Any that don't need
154 ** to be unmapped, so that block_prepare_write will correctly call
155 ** reiserfs_get_block to convert the tail into an unformatted node
157 static inline void fix_tail_page_for_writing(struct page *page)
159 struct buffer_head *head, *next, *bh;
161 if (page && page_has_buffers(page)) {
162 head = page_buffers(page);
165 next = bh->b_this_page;
166 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
167 reiserfs_unmap_buffer(bh);
170 } while (bh != head);
174 /* reiserfs_get_block does not need to allocate a block only if it has been
175 done already or non-hole position has been found in the indirect item */
176 static inline int allocation_needed(int retval, b_blocknr_t allocated,
177 struct item_head *ih,
178 __le32 * item, int pos_in_item)
182 if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
183 get_block_num(item, pos_in_item))
188 static inline int indirect_item_found(int retval, struct item_head *ih)
190 return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
193 static inline void set_block_dev_mapped(struct buffer_head *bh,
194 b_blocknr_t block, struct inode *inode)
196 map_bh(bh, inode->i_sb, block);
200 // files which were created in the earlier version can not be longer,
203 static int file_capable(struct inode *inode, long block)
205 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file.
206 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
212 /*static*/ int restart_transaction(struct reiserfs_transaction_handle *th,
213 struct inode *inode, struct path *path)
215 struct super_block *s = th->t_super;
216 int len = th->t_blocks_allocated;
219 BUG_ON(!th->t_trans_id);
220 BUG_ON(!th->t_refcount);
222 /* we cannot restart while nested */
223 if (th->t_refcount > 1) {
227 reiserfs_update_sd(th, inode);
228 err = journal_end(th, s, len);
230 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
232 reiserfs_update_inode_transaction(inode);
237 // it is called by get_block when create == 0. Returns block number
238 // for 'block'-th logical block of file. When it hits direct item it
239 // returns 0 (being called from bmap) or read direct item into piece
240 // of page (bh_result)
242 // Please improve the english/clarity in the comment above, as it is
243 // hard to understand.
245 static int _get_block_create_0(struct inode *inode, long block,
246 struct buffer_head *bh_result, int args)
248 INITIALIZE_PATH(path);
250 struct buffer_head *bh;
251 struct item_head *ih, tmp_ih;
259 unsigned long offset;
261 // prepare the key to look for the 'block'-th block of file
262 make_cpu_key(&key, inode,
263 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
267 result = search_for_position_by_key(inode->i_sb, &key, &path);
268 if (result != POSITION_FOUND) {
271 kunmap(bh_result->b_page);
272 if (result == IO_ERROR)
274 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
275 // That there is some MMAPED data associated with it that is yet to be written to disk.
276 if ((args & GET_BLOCK_NO_HOLE)
277 && !PageUptodate(bh_result->b_page)) {
283 bh = get_last_bh(&path);
285 if (is_indirect_le_ih(ih)) {
286 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih);
288 /* FIXME: here we could cache indirect item or part of it in
289 the inode to avoid search_by_key in case of subsequent
291 blocknr = get_block_num(ind_item, path.pos_in_item);
294 map_bh(bh_result, inode->i_sb, blocknr);
295 if (path.pos_in_item ==
296 ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
297 set_buffer_boundary(bh_result);
300 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
301 // That there is some MMAPED data associated with it that is yet to be written to disk.
302 if ((args & GET_BLOCK_NO_HOLE)
303 && !PageUptodate(bh_result->b_page)) {
309 kunmap(bh_result->b_page);
312 // requested data are in direct item(s)
313 if (!(args & GET_BLOCK_READ_DIRECT)) {
314 // we are called by bmap. FIXME: we can not map block of file
315 // when it is stored in direct item(s)
318 kunmap(bh_result->b_page);
322 /* if we've got a direct item, and the buffer or page was uptodate,
323 ** we don't want to pull data off disk again. skip to the
324 ** end, where we map the buffer and return
326 if (buffer_uptodate(bh_result)) {
330 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
331 ** pages without any buffers. If the page is up to date, we don't want
332 ** read old data off disk. Set the up to date bit on the buffer instead
333 ** and jump to the end
335 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
336 set_buffer_uptodate(bh_result);
339 // read file tail into part of page
340 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
341 fs_gen = get_generation(inode->i_sb);
342 copy_item_head(&tmp_ih, ih);
344 /* we only want to kmap if we are reading the tail into the page.
345 ** this is not the common case, so we don't kmap until we are
346 ** sure we need to. But, this means the item might move if
350 p = (char *)kmap(bh_result->b_page);
351 if (fs_changed(fs_gen, inode->i_sb)
352 && item_moved(&tmp_ih, &path)) {
357 memset(p, 0, inode->i_sb->s_blocksize);
359 if (!is_direct_le_ih(ih)) {
362 /* make sure we don't read more bytes than actually exist in
363 ** the file. This can happen in odd cases where i_size isn't
364 ** correct, and when direct item padding results in a few
365 ** extra bytes at the end of the direct item
367 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
369 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
371 inode->i_size - (le_ih_k_offset(ih) - 1) -
375 chars = ih_item_len(ih) - path.pos_in_item;
377 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
384 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
385 // we done, if read direct item is not the last item of
386 // node FIXME: we could try to check right delimiting key
387 // to see whether direct item continues in the right
388 // neighbor or rely on i_size
391 // update key to look for the next piece
392 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
393 result = search_for_position_by_key(inode->i_sb, &key, &path);
394 if (result != POSITION_FOUND)
395 // i/o error most likely
397 bh = get_last_bh(&path);
401 flush_dcache_page(bh_result->b_page);
402 kunmap(bh_result->b_page);
407 if (result == IO_ERROR)
410 /* this buffer has valid data, but isn't valid for io. mapping it to
411 * block #0 tells the rest of reiserfs it just has a tail in it
413 map_bh(bh_result, inode->i_sb, 0);
414 set_buffer_uptodate(bh_result);
418 // this is called to create file map. So, _get_block_create_0 will not
420 static int reiserfs_bmap(struct inode *inode, sector_t block,
421 struct buffer_head *bh_result, int create)
423 if (!file_capable(inode, block))
426 reiserfs_write_lock(inode->i_sb);
427 /* do not read the direct item */
428 _get_block_create_0(inode, block, bh_result, 0);
429 reiserfs_write_unlock(inode->i_sb);
433 /* special version of get_block that is only used by grab_tail_page right
434 ** now. It is sent to block_prepare_write, and when you try to get a
435 ** block past the end of the file (or a block from a hole) it returns
436 ** -ENOENT instead of a valid buffer. block_prepare_write expects to
437 ** be able to do i/o on the buffers returned, unless an error value
440 ** So, this allows block_prepare_write to be used for reading a single block
441 ** in a page. Where it does not produce a valid page for holes, or past the
442 ** end of the file. This turns out to be exactly what we need for reading
443 ** tails for conversion.
445 ** The point of the wrapper is forcing a certain value for create, even
446 ** though the VFS layer is calling this function with create==1. If you
447 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
448 ** don't use this function.
450 static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
451 struct buffer_head *bh_result,
454 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
457 /* This is special helper for reiserfs_get_block in case we are executing
458 direct_IO request. */
459 static int reiserfs_get_blocks_direct_io(struct inode *inode,
461 struct buffer_head *bh_result,
466 bh_result->b_page = NULL;
468 /* We set the b_size before reiserfs_get_block call since it is
469 referenced in convert_tail_for_hole() that may be called from
470 reiserfs_get_block() */
471 bh_result->b_size = (1 << inode->i_blkbits);
473 ret = reiserfs_get_block(inode, iblock, bh_result,
474 create | GET_BLOCK_NO_DANGLE);
478 /* don't allow direct io onto tail pages */
479 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
480 /* make sure future calls to the direct io funcs for this offset
481 ** in the file fail by unmapping the buffer
483 clear_buffer_mapped(bh_result);
486 /* Possible unpacked tail. Flush the data before pages have
488 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
491 err = reiserfs_commit_for_inode(inode);
492 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
502 ** helper function for when reiserfs_get_block is called for a hole
503 ** but the file tail is still in a direct item
504 ** bh_result is the buffer head for the hole
505 ** tail_offset is the offset of the start of the tail in the file
507 ** This calls prepare_write, which will start a new transaction
508 ** you should not be in a transaction, or have any paths held when you
511 static int convert_tail_for_hole(struct inode *inode,
512 struct buffer_head *bh_result,
516 unsigned long tail_end;
517 unsigned long tail_start;
518 struct page *tail_page;
519 struct page *hole_page = bh_result->b_page;
522 if ((tail_offset & (bh_result->b_size - 1)) != 1)
525 /* always try to read until the end of the block */
526 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
527 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
529 index = tail_offset >> PAGE_CACHE_SHIFT;
530 /* hole_page can be zero in case of direct_io, we are sure
531 that we cannot get here if we write with O_DIRECT into
533 if (!hole_page || index != hole_page->index) {
534 tail_page = grab_cache_page(inode->i_mapping, index);
540 tail_page = hole_page;
543 /* we don't have to make sure the conversion did not happen while
544 ** we were locking the page because anyone that could convert
545 ** must first take i_mutex.
547 ** We must fix the tail page for writing because it might have buffers
548 ** that are mapped, but have a block number of 0. This indicates tail
549 ** data that has been read directly into the page, and block_prepare_write
550 ** won't trigger a get_block in this case.
552 fix_tail_page_for_writing(tail_page);
553 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
557 /* tail conversion might change the data in the page */
558 flush_dcache_page(tail_page);
560 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
563 if (tail_page != hole_page) {
564 unlock_page(tail_page);
565 page_cache_release(tail_page);
571 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
574 b_blocknr_t * allocated_block_nr,
575 struct path *path, int flags)
577 BUG_ON(!th->t_trans_id);
579 #ifdef REISERFS_PREALLOCATE
580 if (!(flags & GET_BLOCK_NO_IMUX)) {
581 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
585 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
589 int reiserfs_get_block(struct inode *inode, sector_t block,
590 struct buffer_head *bh_result, int create)
592 int repeat, retval = 0;
593 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
594 INITIALIZE_PATH(path);
597 struct buffer_head *bh, *unbh = NULL;
598 struct item_head *ih, tmp_ih;
602 struct reiserfs_transaction_handle *th = NULL;
603 /* space reserved in transaction batch:
604 . 3 balancings in direct->indirect conversion
605 . 1 block involved into reiserfs_update_sd()
606 XXX in practically impossible worst case direct2indirect()
607 can incur (much) more than 3 balancings.
608 quota update for user, group */
610 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
611 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
615 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
618 reiserfs_write_lock(inode->i_sb);
619 version = get_inode_item_key_version(inode);
621 if (!file_capable(inode, block)) {
622 reiserfs_write_unlock(inode->i_sb);
626 /* if !create, we aren't changing the FS, so we don't need to
627 ** log anything, so we don't need to start a transaction
629 if (!(create & GET_BLOCK_CREATE)) {
631 /* find number of block-th logical block of the file */
632 ret = _get_block_create_0(inode, block, bh_result,
633 create | GET_BLOCK_READ_DIRECT);
634 reiserfs_write_unlock(inode->i_sb);
638 * if we're already in a transaction, make sure to close
639 * any new transactions we start in this func
641 if ((create & GET_BLOCK_NO_DANGLE) ||
642 reiserfs_transaction_running(inode->i_sb))
645 /* If file is of such a size, that it might have a tail and tails are enabled
646 ** we should mark it as possibly needing tail packing on close
648 if ((have_large_tails(inode->i_sb)
649 && inode->i_size < i_block_size(inode) * 4)
650 || (have_small_tails(inode->i_sb)
651 && inode->i_size < i_block_size(inode)))
652 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
654 /* set the key of the first byte in the 'block'-th block of file */
655 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
656 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
658 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
663 reiserfs_update_inode_transaction(inode);
667 retval = search_for_position_by_key(inode->i_sb, &key, &path);
668 if (retval == IO_ERROR) {
673 bh = get_last_bh(&path);
675 item = get_item(&path);
676 pos_in_item = path.pos_in_item;
678 fs_gen = get_generation(inode->i_sb);
679 copy_item_head(&tmp_ih, ih);
681 if (allocation_needed
682 (retval, allocated_block_nr, ih, item, pos_in_item)) {
683 /* we have to allocate block for the unformatted node */
690 _allocate_block(th, block, inode, &allocated_block_nr,
693 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
694 /* restart the transaction to give the journal a chance to free
695 ** some blocks. releases the path, so we have to go back to
696 ** research if we succeed on the second try
698 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
699 retval = restart_transaction(th, inode, &path);
703 _allocate_block(th, block, inode,
704 &allocated_block_nr, NULL, create);
706 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
709 if (repeat == QUOTA_EXCEEDED)
716 if (fs_changed(fs_gen, inode->i_sb)
717 && item_moved(&tmp_ih, &path)) {
722 if (indirect_item_found(retval, ih)) {
723 b_blocknr_t unfm_ptr;
724 /* 'block'-th block is in the file already (there is
725 corresponding cell in some indirect item). But it may be
726 zero unformatted node pointer (hole) */
727 unfm_ptr = get_block_num(item, pos_in_item);
729 /* use allocated block to plug the hole */
730 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
731 if (fs_changed(fs_gen, inode->i_sb)
732 && item_moved(&tmp_ih, &path)) {
733 reiserfs_restore_prepared_buffer(inode->i_sb,
737 set_buffer_new(bh_result);
738 if (buffer_dirty(bh_result)
739 && reiserfs_data_ordered(inode->i_sb))
740 reiserfs_add_ordered_list(inode, bh_result);
741 put_block_num(item, pos_in_item, allocated_block_nr);
742 unfm_ptr = allocated_block_nr;
743 journal_mark_dirty(th, inode->i_sb, bh);
744 reiserfs_update_sd(th, inode);
746 set_block_dev_mapped(bh_result, unfm_ptr, inode);
750 retval = reiserfs_end_persistent_transaction(th);
752 reiserfs_write_unlock(inode->i_sb);
754 /* the item was found, so new blocks were not added to the file
755 ** there is no need to make sure the inode is updated with this
766 /* desired position is not found or is in the direct item. We have
767 to append file with holes up to 'block'-th block converting
768 direct items to indirect one if necessary */
771 if (is_statdata_le_ih(ih)) {
773 struct cpu_key tmp_key;
775 /* indirect item has to be inserted */
776 make_le_item_head(&tmp_ih, &key, version, 1,
777 TYPE_INDIRECT, UNFM_P_SIZE,
778 0 /* free_space */ );
780 if (cpu_key_k_offset(&key) == 1) {
781 /* we are going to add 'block'-th block to the file. Use
782 allocated block for that */
783 unp = cpu_to_le32(allocated_block_nr);
784 set_block_dev_mapped(bh_result,
785 allocated_block_nr, inode);
786 set_buffer_new(bh_result);
790 set_cpu_key_k_offset(&tmp_key, 1);
791 PATH_LAST_POSITION(&path)++;
794 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
795 inode, (char *)&unp);
797 reiserfs_free_block(th, inode,
798 allocated_block_nr, 1);
799 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
801 //mark_tail_converted (inode);
802 } else if (is_direct_le_ih(ih)) {
803 /* direct item has to be converted */
807 ((le_ih_k_offset(ih) -
808 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
809 if (tail_offset == cpu_key_k_offset(&key)) {
810 /* direct item we just found fits into block we have
811 to map. Convert it into unformatted node: use
812 bh_result for the conversion */
813 set_block_dev_mapped(bh_result,
814 allocated_block_nr, inode);
818 /* we have to padd file tail stored in direct item(s)
819 up to block size and convert it to unformatted
820 node. FIXME: this should also get into page cache */
824 * ugly, but we can only end the transaction if
827 BUG_ON(!th->t_refcount);
828 if (th->t_refcount == 1) {
830 reiserfs_end_persistent_transaction
838 convert_tail_for_hole(inode, bh_result,
841 if (retval != -ENOSPC)
842 reiserfs_warning(inode->i_sb,
843 "clm-6004: convert tail failed inode %lu, error %d",
846 if (allocated_block_nr) {
847 /* the bitmap, the super, and the stat data == 3 */
849 th = reiserfs_persistent_transaction(inode->i_sb, 3);
851 reiserfs_free_block(th,
861 direct2indirect(th, inode, &path, unbh,
864 reiserfs_unmap_buffer(unbh);
865 reiserfs_free_block(th, inode,
866 allocated_block_nr, 1);
869 /* it is important the set_buffer_uptodate is done after
870 ** the direct2indirect. The buffer might contain valid
871 ** data newer than the data on disk (read by readpage, changed,
872 ** and then sent here by writepage). direct2indirect needs
873 ** to know if unbh was already up to date, so it can decide
874 ** if the data in unbh needs to be replaced with data from
877 set_buffer_uptodate(unbh);
879 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
880 buffer will disappear shortly, so it should not be added to
883 /* we've converted the tail, so we must
884 ** flush unbh before the transaction commits
886 reiserfs_add_tail_list(inode, unbh);
888 /* mark it dirty now to prevent commit_write from adding
889 ** this buffer to the inode's dirty buffer list
892 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
893 * It's still atomic, but it sets the page dirty too,
894 * which makes it eligible for writeback at any time by the
895 * VM (which was also the case with __mark_buffer_dirty())
897 mark_buffer_dirty(unbh);
900 /* append indirect item with holes if needed, when appending
901 pointer to 'block'-th block use block, which is already
903 struct cpu_key tmp_key;
904 unp_t unf_single = 0; // We use this in case we need to allocate only
905 // one block which is a fastpath
907 __u64 max_to_insert =
908 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
912 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
913 "vs-804: invalid position for append");
914 /* indirect item has to be appended, set up key of that position */
915 make_cpu_key(&tmp_key, inode,
916 le_key_k_offset(version,
919 inode->i_sb->s_blocksize),
920 //pos_in_item * inode->i_sb->s_blocksize,
921 TYPE_INDIRECT, 3); // key type is unimportant
923 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
924 "green-805: invalid offset");
927 ((cpu_key_k_offset(&key) -
928 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
931 if (blocks_needed == 1) {
934 un = kmalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_ATOMIC); // We need to avoid scheduling.
941 UNFM_P_SIZE * min(blocks_needed,
944 if (blocks_needed <= max_to_insert) {
945 /* we are going to add target block to the file. Use allocated
947 un[blocks_needed - 1] =
948 cpu_to_le32(allocated_block_nr);
949 set_block_dev_mapped(bh_result,
950 allocated_block_nr, inode);
951 set_buffer_new(bh_result);
954 /* paste hole to the indirect item */
955 /* If kmalloc failed, max_to_insert becomes zero and it means we
956 only have space for one block */
958 max_to_insert ? max_to_insert : 1;
961 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
966 if (blocks_needed != 1)
970 reiserfs_free_block(th, inode,
971 allocated_block_nr, 1);
975 /* We need to mark new file size in case this function will be
976 interrupted/aborted later on. And we may do this only for
979 inode->i_sb->s_blocksize * blocks_needed;
986 /* this loop could log more blocks than we had originally asked
987 ** for. So, we have to allow the transaction to end if it is
988 ** too big or too full. Update the inode so things are
989 ** consistent if we crash before the function returns
991 ** release the path so that anybody waiting on the path before
992 ** ending their transaction will be able to continue.
994 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
995 retval = restart_transaction(th, inode, &path);
999 /* inserting indirect pointers for a hole can take a
1000 ** long time. reschedule if needed
1004 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1005 if (retval == IO_ERROR) {
1009 if (retval == POSITION_FOUND) {
1010 reiserfs_warning(inode->i_sb,
1011 "vs-825: reiserfs_get_block: "
1012 "%K should not be found", &key);
1014 if (allocated_block_nr)
1015 reiserfs_free_block(th, inode,
1016 allocated_block_nr, 1);
1020 bh = get_last_bh(&path);
1022 item = get_item(&path);
1023 pos_in_item = path.pos_in_item;
1029 if (th && (!dangle || (retval && !th->t_trans_id))) {
1032 reiserfs_update_sd(th, inode);
1033 err = reiserfs_end_persistent_transaction(th);
1038 reiserfs_write_unlock(inode->i_sb);
1039 reiserfs_check_path(&path);
1044 reiserfs_readpages(struct file *file, struct address_space *mapping,
1045 struct list_head *pages, unsigned nr_pages)
1047 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
1050 /* Compute real number of used bytes by file
1051 * Following three functions can go away when we'll have enough space in stat item
1053 static int real_space_diff(struct inode *inode, int sd_size)
1056 loff_t blocksize = inode->i_sb->s_blocksize;
1058 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1061 /* End of file is also in full block with indirect reference, so round
1062 ** up to the next block.
1064 ** there is just no way to know if the tail is actually packed
1065 ** on the file, so we have to assume it isn't. When we pack the
1066 ** tail, we add 4 bytes to pretend there really is an unformatted
1071 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1076 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
1079 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1080 return inode->i_size +
1081 (loff_t) (real_space_diff(inode, sd_size));
1083 return ((loff_t) real_space_diff(inode, sd_size)) +
1084 (((loff_t) blocks) << 9);
1087 /* Compute number of blocks used by file in ReiserFS counting */
1088 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1090 loff_t bytes = inode_get_bytes(inode);
1091 loff_t real_space = real_space_diff(inode, sd_size);
1093 /* keeps fsck and non-quota versions of reiserfs happy */
1094 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1095 bytes += (loff_t) 511;
1098 /* files from before the quota patch might i_blocks such that
1099 ** bytes < real_space. Deal with that here to prevent it from
1102 if (bytes < real_space)
1104 return (bytes - real_space) >> 9;
1108 // BAD: new directories have stat data of new type and all other items
1109 // of old type. Version stored in the inode says about body items, so
1110 // in update_stat_data we can not rely on inode, but have to check
1111 // item version directly
1114 // called by read_locked_inode
1115 static void init_inode(struct inode *inode, struct path *path)
1117 struct buffer_head *bh;
1118 struct item_head *ih;
1122 //int version = ITEM_VERSION_1;
1124 bh = PATH_PLAST_BUFFER(path);
1125 ih = PATH_PITEM_HEAD(path);
1127 copy_key(INODE_PKEY(inode), &(ih->ih_key));
1129 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1130 REISERFS_I(inode)->i_flags = 0;
1131 REISERFS_I(inode)->i_prealloc_block = 0;
1132 REISERFS_I(inode)->i_prealloc_count = 0;
1133 REISERFS_I(inode)->i_trans_id = 0;
1134 REISERFS_I(inode)->i_jl = NULL;
1135 REISERFS_I(inode)->i_acl_access = NULL;
1136 REISERFS_I(inode)->i_acl_default = NULL;
1137 init_rwsem(&REISERFS_I(inode)->xattr_sem);
1139 if (stat_data_v1(ih)) {
1140 struct stat_data_v1 *sd =
1141 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1142 unsigned long blocks;
1144 uid = sd_v1_uid(sd);
1145 gid = sd_v1_gid(sd);
1147 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1148 set_inode_sd_version(inode, STAT_DATA_V1);
1149 inode->i_mode = sd_v1_mode(sd);
1150 inode->i_nlink = sd_v1_nlink(sd);
1151 inode->i_size = sd_v1_size(sd);
1152 inode->i_atime.tv_sec = sd_v1_atime(sd);
1153 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1154 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1155 inode->i_atime.tv_nsec = 0;
1156 inode->i_ctime.tv_nsec = 0;
1157 inode->i_mtime.tv_nsec = 0;
1159 inode->i_blocks = sd_v1_blocks(sd);
1160 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1161 blocks = (inode->i_size + 511) >> 9;
1162 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1163 if (inode->i_blocks > blocks) {
1164 // there was a bug in <=3.5.23 when i_blocks could take negative
1165 // values. Starting from 3.5.17 this value could even be stored in
1166 // stat data. For such files we set i_blocks based on file
1167 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1168 // only updated if file's inode will ever change
1169 inode->i_blocks = blocks;
1172 rdev = sd_v1_rdev(sd);
1173 REISERFS_I(inode)->i_first_direct_byte =
1174 sd_v1_first_direct_byte(sd);
1175 /* an early bug in the quota code can give us an odd number for the
1176 ** block count. This is incorrect, fix it here.
1178 if (inode->i_blocks & 1) {
1181 inode_set_bytes(inode,
1182 to_real_used_space(inode, inode->i_blocks,
1184 /* nopack is initially zero for v1 objects. For v2 objects,
1185 nopack is initialised from sd_attrs */
1186 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1188 // new stat data found, but object may have old items
1189 // (directories and symlinks)
1190 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1192 uid = sd_v2_uid(sd);
1193 gid = sd_v2_gid(sd);
1195 inode->i_mode = sd_v2_mode(sd);
1196 inode->i_nlink = sd_v2_nlink(sd);
1197 inode->i_size = sd_v2_size(sd);
1198 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1199 inode->i_atime.tv_sec = sd_v2_atime(sd);
1200 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1201 inode->i_ctime.tv_nsec = 0;
1202 inode->i_mtime.tv_nsec = 0;
1203 inode->i_atime.tv_nsec = 0;
1204 inode->i_blocks = sd_v2_blocks(sd);
1205 rdev = sd_v2_rdev(sd);
1206 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1207 inode->i_generation =
1208 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1210 inode->i_generation = sd_v2_generation(sd);
1212 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1213 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1215 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1216 REISERFS_I(inode)->i_first_direct_byte = 0;
1217 set_inode_sd_version(inode, STAT_DATA_V2);
1218 inode_set_bytes(inode,
1219 to_real_used_space(inode, inode->i_blocks,
1221 /* read persistent inode attributes from sd and initalise
1222 generic inode flags from them */
1223 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1224 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
1227 inode->i_uid = INOXID_UID(XID_TAG(inode), uid, gid);
1228 inode->i_gid = INOXID_GID(XID_TAG(inode), uid, gid);
1229 inode->i_xid = INOXID_XID(XID_TAG(inode), uid, gid, 0);
1232 if (S_ISREG(inode->i_mode)) {
1233 inode->i_op = &reiserfs_file_inode_operations;
1234 inode->i_fop = &reiserfs_file_operations;
1235 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1236 } else if (S_ISDIR(inode->i_mode)) {
1237 inode->i_op = &reiserfs_dir_inode_operations;
1238 inode->i_fop = &reiserfs_dir_operations;
1239 } else if (S_ISLNK(inode->i_mode)) {
1240 inode->i_op = &reiserfs_symlink_inode_operations;
1241 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1243 inode->i_blocks = 0;
1244 inode->i_op = &reiserfs_special_inode_operations;
1245 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1249 // update new stat data with inode fields
1250 static void inode2sd(void *sd, struct inode *inode, loff_t size)
1252 struct stat_data *sd_v2 = (struct stat_data *)sd;
1253 uid_t uid = XIDINO_UID(XID_TAG(inode), inode->i_uid, inode->i_xid);
1254 gid_t gid = XIDINO_GID(XID_TAG(inode), inode->i_gid, inode->i_xid);
1257 set_sd_v2_uid(sd_v2, uid);
1258 set_sd_v2_gid(sd_v2, gid);
1259 set_sd_v2_mode(sd_v2, inode->i_mode);
1260 set_sd_v2_nlink(sd_v2, inode->i_nlink);
1261 set_sd_v2_size(sd_v2, size);
1262 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1263 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1264 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1265 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1266 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1267 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1269 set_sd_v2_generation(sd_v2, inode->i_generation);
1270 flags = REISERFS_I(inode)->i_attrs;
1271 i_attrs_to_sd_attrs(inode, &flags);
1272 set_sd_v2_attrs(sd_v2, flags);
1275 // used to copy inode's fields to old stat data
1276 static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
1278 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
1280 set_sd_v1_mode(sd_v1, inode->i_mode);
1281 set_sd_v1_uid(sd_v1, inode->i_uid);
1282 set_sd_v1_gid(sd_v1, inode->i_gid);
1283 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1284 set_sd_v1_size(sd_v1, size);
1285 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1286 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1287 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
1289 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1290 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1292 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1294 // Sigh. i_first_direct_byte is back
1295 set_sd_v1_first_direct_byte(sd_v1,
1296 REISERFS_I(inode)->i_first_direct_byte);
1299 /* NOTE, you must prepare the buffer head before sending it here,
1300 ** and then log it after the call
1302 static void update_stat_data(struct path *path, struct inode *inode,
1305 struct buffer_head *bh;
1306 struct item_head *ih;
1308 bh = PATH_PLAST_BUFFER(path);
1309 ih = PATH_PITEM_HEAD(path);
1311 if (!is_statdata_le_ih(ih))
1312 reiserfs_panic(inode->i_sb,
1313 "vs-13065: update_stat_data: key %k, found item %h",
1314 INODE_PKEY(inode), ih);
1316 if (stat_data_v1(ih)) {
1317 // path points to old stat data
1318 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1320 inode2sd(B_I_PITEM(bh, ih), inode, size);
1326 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1327 struct inode *inode, loff_t size)
1330 INITIALIZE_PATH(path);
1331 struct buffer_head *bh;
1333 struct item_head *ih, tmp_ih;
1336 BUG_ON(!th->t_trans_id);
1338 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant
1342 /* look for the object's stat data */
1343 retval = search_item(inode->i_sb, &key, &path);
1344 if (retval == IO_ERROR) {
1345 reiserfs_warning(inode->i_sb,
1346 "vs-13050: reiserfs_update_sd: "
1347 "i/o failure occurred trying to update %K stat data",
1351 if (retval == ITEM_NOT_FOUND) {
1352 pos = PATH_LAST_POSITION(&path);
1354 if (inode->i_nlink == 0) {
1355 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1358 reiserfs_warning(inode->i_sb,
1359 "vs-13060: reiserfs_update_sd: "
1360 "stat data of object %k (nlink == %d) not found (pos %d)",
1361 INODE_PKEY(inode), inode->i_nlink,
1363 reiserfs_check_path(&path);
1367 /* sigh, prepare_for_journal might schedule. When it schedules the
1368 ** FS might change. We have to detect that, and loop back to the
1369 ** search if the stat data item has moved
1371 bh = get_last_bh(&path);
1373 copy_item_head(&tmp_ih, ih);
1374 fs_gen = get_generation(inode->i_sb);
1375 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1376 if (fs_changed(fs_gen, inode->i_sb)
1377 && item_moved(&tmp_ih, &path)) {
1378 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1379 continue; /* Stat_data item has been moved after scheduling. */
1383 update_stat_data(&path, inode, size);
1384 journal_mark_dirty(th, th->t_super, bh);
1389 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1390 ** does a make_bad_inode when things go wrong. But, we need to make sure
1391 ** and clear the key in the private portion of the inode, otherwise a
1392 ** corresponding iput might try to delete whatever object the inode last
1395 static void reiserfs_make_bad_inode(struct inode *inode)
1397 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1398 make_bad_inode(inode);
1402 // initially this function was derived from minix or ext2's analog and
1403 // evolved as the prototype did
1406 int reiserfs_init_locked_inode(struct inode *inode, void *p)
1408 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1409 inode->i_ino = args->objectid;
1410 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1414 /* looks for stat data in the tree, and fills up the fields of in-core
1415 inode stat data fields */
1416 void reiserfs_read_locked_inode(struct inode *inode,
1417 struct reiserfs_iget_args *args)
1419 INITIALIZE_PATH(path_to_sd);
1421 unsigned long dirino;
1424 dirino = args->dirid;
1426 /* set version 1, version 2 could be used too, because stat data
1427 key is the same in both versions */
1428 key.version = KEY_FORMAT_3_5;
1429 key.on_disk_key.k_dir_id = dirino;
1430 key.on_disk_key.k_objectid = inode->i_ino;
1431 key.on_disk_key.k_offset = 0;
1432 key.on_disk_key.k_type = 0;
1434 /* look for the object's stat data */
1435 retval = search_item(inode->i_sb, &key, &path_to_sd);
1436 if (retval == IO_ERROR) {
1437 reiserfs_warning(inode->i_sb,
1438 "vs-13070: reiserfs_read_locked_inode: "
1439 "i/o failure occurred trying to find stat data of %K",
1441 reiserfs_make_bad_inode(inode);
1444 if (retval != ITEM_FOUND) {
1445 /* a stale NFS handle can trigger this without it being an error */
1446 pathrelse(&path_to_sd);
1447 reiserfs_make_bad_inode(inode);
1452 init_inode(inode, &path_to_sd);
1454 /* It is possible that knfsd is trying to access inode of a file
1455 that is being removed from the disk by some other thread. As we
1456 update sd on unlink all that is required is to check for nlink
1457 here. This bug was first found by Sizif when debugging
1458 SquidNG/Butterfly, forgotten, and found again after Philippe
1459 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1461 More logical fix would require changes in fs/inode.c:iput() to
1462 remove inode from hash-table _after_ fs cleaned disk stuff up and
1463 in iget() to return NULL if I_FREEING inode is found in
1465 /* Currently there is one place where it's ok to meet inode with
1466 nlink==0: processing of open-unlinked and half-truncated files
1467 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1468 if ((inode->i_nlink == 0) &&
1469 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
1470 reiserfs_warning(inode->i_sb,
1471 "vs-13075: reiserfs_read_locked_inode: "
1472 "dead inode read from disk %K. "
1473 "This is likely to be race with knfsd. Ignore",
1475 reiserfs_make_bad_inode(inode);
1478 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
1483 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1485 * @inode: inode from hash table to check
1486 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1488 * This function is called by iget5_locked() to distinguish reiserfs inodes
1489 * having the same inode numbers. Such inodes can only exist due to some
1490 * error condition. One of them should be bad. Inodes with identical
1491 * inode numbers (objectids) are distinguished by parent directory ids.
1494 int reiserfs_find_actor(struct inode *inode, void *opaque)
1496 struct reiserfs_iget_args *args;
1499 /* args is already in CPU order */
1500 return (inode->i_ino == args->objectid) &&
1501 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1504 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
1506 struct inode *inode;
1507 struct reiserfs_iget_args args;
1509 args.objectid = key->on_disk_key.k_objectid;
1510 args.dirid = key->on_disk_key.k_dir_id;
1511 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1512 reiserfs_find_actor, reiserfs_init_locked_inode,
1515 return ERR_PTR(-ENOMEM);
1517 if (inode->i_state & I_NEW) {
1518 reiserfs_read_locked_inode(inode, &args);
1519 unlock_new_inode(inode);
1522 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1523 /* either due to i/o error or a stale NFS handle */
1530 struct dentry *reiserfs_get_dentry(struct super_block *sb, void *vobjp)
1532 __u32 *data = vobjp;
1534 struct dentry *result;
1535 struct inode *inode;
1537 key.on_disk_key.k_objectid = data[0];
1538 key.on_disk_key.k_dir_id = data[1];
1539 reiserfs_write_lock(sb);
1540 inode = reiserfs_iget(sb, &key);
1541 if (inode && !IS_ERR(inode) && data[2] != 0 &&
1542 data[2] != inode->i_generation) {
1546 reiserfs_write_unlock(sb);
1548 inode = ERR_PTR(-ESTALE);
1550 return ERR_PTR(PTR_ERR(inode));
1551 result = d_alloc_anon(inode);
1554 return ERR_PTR(-ENOMEM);
1559 struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 * data,
1560 int len, int fhtype,
1561 int (*acceptable) (void *contect,
1562 struct dentry * de),
1565 __u32 obj[3], parent[3];
1567 /* fhtype happens to reflect the number of u32s encoded.
1568 * due to a bug in earlier code, fhtype might indicate there
1569 * are more u32s then actually fitted.
1570 * so if fhtype seems to be more than len, reduce fhtype.
1572 * 2 - objectid + dir_id - legacy support
1573 * 3 - objectid + dir_id + generation
1574 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1575 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1576 * 6 - as above plus generation of directory
1577 * 6 does not fit in NFSv2 handles
1580 if (fhtype != 6 || len != 5)
1581 reiserfs_warning(sb,
1582 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1589 if (fhtype == 3 || fhtype >= 5)
1592 obj[2] = 0; /* generation number */
1595 parent[0] = data[fhtype >= 5 ? 3 : 2];
1596 parent[1] = data[fhtype >= 5 ? 4 : 3];
1598 parent[2] = data[5];
1602 return sb->s_export_op->find_exported_dentry(sb, obj,
1603 fhtype < 4 ? NULL : parent,
1604 acceptable, context);
1607 int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1610 struct inode *inode = dentry->d_inode;
1616 data[0] = inode->i_ino;
1617 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1618 data[2] = inode->i_generation;
1620 /* no room for directory info? return what we've stored so far */
1621 if (maxlen < 5 || !need_parent)
1624 spin_lock(&dentry->d_lock);
1625 inode = dentry->d_parent->d_inode;
1626 data[3] = inode->i_ino;
1627 data[4] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1630 data[5] = inode->i_generation;
1633 spin_unlock(&dentry->d_lock);
1637 /* looks for stat data, then copies fields to it, marks the buffer
1638 containing stat data as dirty */
1639 /* reiserfs inodes are never really dirty, since the dirty inode call
1640 ** always logs them. This call allows the VFS inode marking routines
1641 ** to properly mark inodes for datasync and such, but only actually
1642 ** does something when called for a synchronous update.
1644 int reiserfs_write_inode(struct inode *inode, int do_sync)
1646 struct reiserfs_transaction_handle th;
1647 int jbegin_count = 1;
1649 if (inode->i_sb->s_flags & MS_RDONLY)
1651 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1652 ** these cases are just when the system needs ram, not when the
1653 ** inode needs to reach disk for safety, and they can safely be
1654 ** ignored because the altered inode has already been logged.
1656 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1657 reiserfs_write_lock(inode->i_sb);
1658 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1659 reiserfs_update_sd(&th, inode);
1660 journal_end_sync(&th, inode->i_sb, jbegin_count);
1662 reiserfs_write_unlock(inode->i_sb);
1667 /* stat data of new object is inserted already, this inserts the item
1668 containing "." and ".." entries */
1669 static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1670 struct inode *inode,
1671 struct item_head *ih, struct path *path,
1674 struct super_block *sb = th->t_super;
1675 char empty_dir[EMPTY_DIR_SIZE];
1676 char *body = empty_dir;
1680 BUG_ON(!th->t_trans_id);
1682 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1683 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1684 TYPE_DIRENTRY, 3 /*key length */ );
1686 /* compose item head for new item. Directories consist of items of
1687 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1688 is done by reiserfs_new_inode */
1689 if (old_format_only(sb)) {
1690 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1691 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1693 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1694 ih->ih_key.k_objectid,
1695 INODE_PKEY(dir)->k_dir_id,
1696 INODE_PKEY(dir)->k_objectid);
1698 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1699 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1701 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1702 ih->ih_key.k_objectid,
1703 INODE_PKEY(dir)->k_dir_id,
1704 INODE_PKEY(dir)->k_objectid);
1707 /* look for place in the tree for new item */
1708 retval = search_item(sb, &key, path);
1709 if (retval == IO_ERROR) {
1710 reiserfs_warning(sb, "vs-13080: reiserfs_new_directory: "
1711 "i/o failure occurred creating new directory");
1714 if (retval == ITEM_FOUND) {
1716 reiserfs_warning(sb, "vs-13070: reiserfs_new_directory: "
1717 "object with this key exists (%k)",
1722 /* insert item, that is empty directory item */
1723 return reiserfs_insert_item(th, path, &key, ih, inode, body);
1726 /* stat data of object has been inserted, this inserts the item
1727 containing the body of symlink */
1728 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1729 struct item_head *ih,
1730 struct path *path, const char *symname,
1733 struct super_block *sb = th->t_super;
1737 BUG_ON(!th->t_trans_id);
1739 _make_cpu_key(&key, KEY_FORMAT_3_5,
1740 le32_to_cpu(ih->ih_key.k_dir_id),
1741 le32_to_cpu(ih->ih_key.k_objectid),
1742 1, TYPE_DIRECT, 3 /*key length */ );
1744 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1745 0 /*free_space */ );
1747 /* look for place in the tree for new item */
1748 retval = search_item(sb, &key, path);
1749 if (retval == IO_ERROR) {
1750 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlinik: "
1751 "i/o failure occurred creating new symlink");
1754 if (retval == ITEM_FOUND) {
1756 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlink: "
1757 "object with this key exists (%k)",
1762 /* insert item, that is body of symlink */
1763 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
1766 /* inserts the stat data into the tree, and then calls
1767 reiserfs_new_directory (to insert ".", ".." item if new object is
1768 directory) or reiserfs_new_symlink (to insert symlink body if new
1769 object is symlink) or nothing (if new object is regular file)
1771 NOTE! uid and gid must already be set in the inode. If we return
1772 non-zero due to an error, we have to drop the quota previously allocated
1773 for the fresh inode. This can only be done outside a transaction, so
1774 if we return non-zero, we also end the transaction. */
1775 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1776 struct inode *dir, int mode, const char *symname,
1777 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1778 strlen (symname) for symlinks) */
1779 loff_t i_size, struct dentry *dentry,
1780 struct inode *inode)
1782 struct super_block *sb;
1783 INITIALIZE_PATH(path_to_key);
1785 struct item_head ih;
1786 struct stat_data sd;
1790 BUG_ON(!th->t_trans_id);
1792 if (DLIMIT_ALLOC_INODE(inode)) {
1794 goto out_bad_dlimit;
1796 if (DQUOT_ALLOC_INODE(inode)) {
1800 if (!dir || !dir->i_nlink) {
1807 /* item head of new item */
1808 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1809 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1810 if (!ih.ih_key.k_objectid) {
1814 if (old_format_only(sb))
1815 /* not a perfect generation count, as object ids can be reused, but
1816 ** this is as good as reiserfs can do right now.
1817 ** note that the private part of inode isn't filled in yet, we have
1818 ** to use the directory.
1820 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1822 #if defined( USE_INODE_GENERATION_COUNTER )
1823 inode->i_generation =
1824 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1826 inode->i_generation = ++event;
1829 /* fill stat data */
1830 inode->i_nlink = (S_ISDIR(mode) ? 2 : 1);
1832 /* uid and gid must already be set by the caller for quota init */
1834 /* symlink cannot be immutable or append only, right? */
1835 if (S_ISLNK(inode->i_mode))
1836 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1838 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1839 inode->i_size = i_size;
1840 inode->i_blocks = 0;
1842 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1843 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1845 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1846 REISERFS_I(inode)->i_flags = 0;
1847 REISERFS_I(inode)->i_prealloc_block = 0;
1848 REISERFS_I(inode)->i_prealloc_count = 0;
1849 REISERFS_I(inode)->i_trans_id = 0;
1850 REISERFS_I(inode)->i_jl = NULL;
1851 REISERFS_I(inode)->i_attrs =
1852 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1853 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
1854 REISERFS_I(inode)->i_acl_access = NULL;
1855 REISERFS_I(inode)->i_acl_default = NULL;
1856 init_rwsem(&REISERFS_I(inode)->xattr_sem);
1858 if (old_format_only(sb))
1859 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1860 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1862 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1863 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1865 /* key to search for correct place for new stat data */
1866 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1867 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1868 TYPE_STAT_DATA, 3 /*key length */ );
1870 /* find proper place for inserting of stat data */
1871 retval = search_item(sb, &key, &path_to_key);
1872 if (retval == IO_ERROR) {
1876 if (retval == ITEM_FOUND) {
1877 pathrelse(&path_to_key);
1881 if (old_format_only(sb)) {
1882 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1883 pathrelse(&path_to_key);
1884 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1888 inode2sd_v1(&sd, inode, inode->i_size);
1890 inode2sd(&sd, inode, inode->i_size);
1892 // these do not go to on-disk stat data
1893 inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
1895 // store in in-core inode the key of stat data and version all
1896 // object items will have (directory items will have old offset
1897 // format, other new objects will consist of new items)
1898 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1899 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1900 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1902 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1903 if (old_format_only(sb))
1904 set_inode_sd_version(inode, STAT_DATA_V1);
1906 set_inode_sd_version(inode, STAT_DATA_V2);
1908 /* insert the stat data into the tree */
1909 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1910 if (REISERFS_I(dir)->new_packing_locality)
1911 th->displace_new_blocks = 1;
1914 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1918 reiserfs_check_path(&path_to_key);
1921 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1922 if (!th->displace_new_blocks)
1923 REISERFS_I(dir)->new_packing_locality = 0;
1925 if (S_ISDIR(mode)) {
1926 /* insert item with "." and ".." */
1928 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1931 if (S_ISLNK(mode)) {
1932 /* insert body of symlink */
1933 if (!old_format_only(sb))
1934 i_size = ROUND_UP(i_size);
1936 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1941 reiserfs_check_path(&path_to_key);
1942 journal_end(th, th->t_super, th->t_blocks_allocated);
1943 goto out_inserted_sd;
1946 /* XXX CHECK THIS */
1947 if (reiserfs_posixacl(inode->i_sb)) {
1948 retval = reiserfs_inherit_default_acl(dir, dentry, inode);
1951 reiserfs_check_path(&path_to_key);
1952 journal_end(th, th->t_super, th->t_blocks_allocated);
1953 goto out_inserted_sd;
1955 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1956 reiserfs_warning(inode->i_sb, "ACLs aren't enabled in the fs, "
1957 "but vfs thinks they are!");
1958 } else if (is_reiserfs_priv_object(dir)) {
1959 reiserfs_mark_inode_private(inode);
1962 insert_inode_hash(inode);
1963 reiserfs_update_sd(th, inode);
1964 reiserfs_check_path(&path_to_key);
1968 /* it looks like you can easily compress these two goto targets into
1969 * one. Keeping it like this doesn't actually hurt anything, and they
1970 * are place holders for what the quota code actually needs.
1973 /* Invalidate the object, nothing was inserted yet */
1974 INODE_PKEY(inode)->k_objectid = 0;
1976 /* Quota change must be inside a transaction for journaling */
1977 DQUOT_FREE_INODE(inode);
1980 DLIMIT_FREE_INODE(inode);
1983 journal_end(th, th->t_super, th->t_blocks_allocated);
1984 /* Drop can be outside and it needs more credits so it's better to have it outside */
1986 inode->i_flags |= S_NOQUOTA;
1987 make_bad_inode(inode);
1991 th->t_trans_id = 0; /* so the caller can't use this handle later */
1993 /* If we were inheriting an ACL, we need to release the lock so that
1994 * iput doesn't deadlock in reiserfs_delete_xattrs. The locking
1995 * code really needs to be reworked, but this will take care of it
1996 * for now. -jeffm */
1997 if (REISERFS_I(dir)->i_acl_default && !IS_ERR(REISERFS_I(dir)->i_acl_default)) {
1998 reiserfs_write_unlock_xattrs(dir->i_sb);
2000 reiserfs_write_lock_xattrs(dir->i_sb);
2007 ** finds the tail page in the page cache,
2008 ** reads the last block in.
2010 ** On success, page_result is set to a locked, pinned page, and bh_result
2011 ** is set to an up to date buffer for the last block in the file. returns 0.
2013 ** tail conversion is not done, so bh_result might not be valid for writing
2014 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2015 ** trying to write the block.
2017 ** on failure, nonzero is returned, page_result and bh_result are untouched.
2019 static int grab_tail_page(struct inode *p_s_inode,
2020 struct page **page_result,
2021 struct buffer_head **bh_result)
2024 /* we want the page with the last byte in the file,
2025 ** not the page that will hold the next byte for appending
2027 unsigned long index = (p_s_inode->i_size - 1) >> PAGE_CACHE_SHIFT;
2028 unsigned long pos = 0;
2029 unsigned long start = 0;
2030 unsigned long blocksize = p_s_inode->i_sb->s_blocksize;
2031 unsigned long offset = (p_s_inode->i_size) & (PAGE_CACHE_SIZE - 1);
2032 struct buffer_head *bh;
2033 struct buffer_head *head;
2037 /* we know that we are only called with inode->i_size > 0.
2038 ** we also know that a file tail can never be as big as a block
2039 ** If i_size % blocksize == 0, our file is currently block aligned
2040 ** and it won't need converting or zeroing after a truncate.
2042 if ((offset & (blocksize - 1)) == 0) {
2045 page = grab_cache_page(p_s_inode->i_mapping, index);
2050 /* start within the page of the last block in the file */
2051 start = (offset / blocksize) * blocksize;
2053 error = block_prepare_write(page, start, offset,
2054 reiserfs_get_block_create_0);
2058 head = page_buffers(page);
2064 bh = bh->b_this_page;
2066 } while (bh != head);
2068 if (!buffer_uptodate(bh)) {
2069 /* note, this should never happen, prepare_write should
2070 ** be taking care of this for us. If the buffer isn't up to date,
2071 ** I've screwed up the code to find the buffer, or the code to
2072 ** call prepare_write
2074 reiserfs_warning(p_s_inode->i_sb,
2075 "clm-6000: error reading block %lu on dev %s",
2077 reiserfs_bdevname(p_s_inode->i_sb));
2082 *page_result = page;
2089 page_cache_release(page);
2094 ** vfs version of truncate file. Must NOT be called with
2095 ** a transaction already started.
2097 ** some code taken from block_truncate_page
2099 int reiserfs_truncate_file(struct inode *p_s_inode, int update_timestamps)
2101 struct reiserfs_transaction_handle th;
2102 /* we want the offset for the first byte after the end of the file */
2103 unsigned long offset = p_s_inode->i_size & (PAGE_CACHE_SIZE - 1);
2104 unsigned blocksize = p_s_inode->i_sb->s_blocksize;
2106 struct page *page = NULL;
2108 struct buffer_head *bh = NULL;
2111 reiserfs_write_lock(p_s_inode->i_sb);
2113 if (p_s_inode->i_size > 0) {
2114 if ((error = grab_tail_page(p_s_inode, &page, &bh))) {
2115 // -ENOENT means we truncated past the end of the file,
2116 // and get_block_create_0 could not find a block to read in,
2118 if (error != -ENOENT)
2119 reiserfs_warning(p_s_inode->i_sb,
2120 "clm-6001: grab_tail_page failed %d",
2127 /* so, if page != NULL, we have a buffer head for the offset at
2128 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2129 ** then we have an unformatted node. Otherwise, we have a direct item,
2130 ** and no zeroing is required on disk. We zero after the truncate,
2131 ** because the truncate might pack the item anyway
2132 ** (it will unmap bh if it packs).
2134 /* it is enough to reserve space in transaction for 2 balancings:
2135 one for "save" link adding and another for the first
2136 cut_from_item. 1 is for update_sd */
2137 error = journal_begin(&th, p_s_inode->i_sb,
2138 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2141 reiserfs_update_inode_transaction(p_s_inode);
2142 if (update_timestamps)
2143 /* we are doing real truncate: if the system crashes before the last
2144 transaction of truncating gets committed - on reboot the file
2145 either appears truncated properly or not truncated at all */
2146 add_save_link(&th, p_s_inode, 1);
2147 err2 = reiserfs_do_truncate(&th, p_s_inode, page, update_timestamps);
2149 journal_end(&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2153 /* check reiserfs_do_truncate after ending the transaction */
2159 if (update_timestamps) {
2160 error = remove_save_link(p_s_inode, 1 /* truncate */ );
2166 length = offset & (blocksize - 1);
2167 /* if we are not on a block boundary */
2171 length = blocksize - length;
2172 kaddr = kmap_atomic(page, KM_USER0);
2173 memset(kaddr + offset, 0, length);
2174 flush_dcache_page(page);
2175 kunmap_atomic(kaddr, KM_USER0);
2176 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2177 mark_buffer_dirty(bh);
2181 page_cache_release(page);
2184 reiserfs_write_unlock(p_s_inode->i_sb);
2189 page_cache_release(page);
2191 reiserfs_write_unlock(p_s_inode->i_sb);
2195 static int map_block_for_writepage(struct inode *inode,
2196 struct buffer_head *bh_result,
2197 unsigned long block)
2199 struct reiserfs_transaction_handle th;
2201 struct item_head tmp_ih;
2202 struct item_head *ih;
2203 struct buffer_head *bh;
2206 INITIALIZE_PATH(path);
2208 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
2209 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
2211 int use_get_block = 0;
2212 int bytes_copied = 0;
2214 int trans_running = 0;
2216 /* catch places below that try to log something without starting a trans */
2219 if (!buffer_uptodate(bh_result)) {
2223 kmap(bh_result->b_page);
2225 reiserfs_write_lock(inode->i_sb);
2226 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
2229 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2230 if (retval != POSITION_FOUND) {
2235 bh = get_last_bh(&path);
2237 item = get_item(&path);
2238 pos_in_item = path.pos_in_item;
2240 /* we've found an unformatted node */
2241 if (indirect_item_found(retval, ih)) {
2242 if (bytes_copied > 0) {
2243 reiserfs_warning(inode->i_sb,
2244 "clm-6002: bytes_copied %d",
2247 if (!get_block_num(item, pos_in_item)) {
2248 /* crap, we are writing to a hole */
2252 set_block_dev_mapped(bh_result,
2253 get_block_num(item, pos_in_item), inode);
2254 } else if (is_direct_le_ih(ih)) {
2256 p = page_address(bh_result->b_page);
2257 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2258 copy_size = ih_item_len(ih) - pos_in_item;
2260 fs_gen = get_generation(inode->i_sb);
2261 copy_item_head(&tmp_ih, ih);
2263 if (!trans_running) {
2264 /* vs-3050 is gone, no need to drop the path */
2265 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2268 reiserfs_update_inode_transaction(inode);
2270 if (fs_changed(fs_gen, inode->i_sb)
2271 && item_moved(&tmp_ih, &path)) {
2272 reiserfs_restore_prepared_buffer(inode->i_sb,
2278 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2280 if (fs_changed(fs_gen, inode->i_sb)
2281 && item_moved(&tmp_ih, &path)) {
2282 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2286 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2289 journal_mark_dirty(&th, inode->i_sb, bh);
2290 bytes_copied += copy_size;
2291 set_block_dev_mapped(bh_result, 0, inode);
2293 /* are there still bytes left? */
2294 if (bytes_copied < bh_result->b_size &&
2295 (byte_offset + bytes_copied) < inode->i_size) {
2296 set_cpu_key_k_offset(&key,
2297 cpu_key_k_offset(&key) +
2302 reiserfs_warning(inode->i_sb,
2303 "clm-6003: bad item inode %lu, device %s",
2304 inode->i_ino, reiserfs_bdevname(inode->i_sb));
2312 if (trans_running) {
2313 int err = journal_end(&th, inode->i_sb, jbegin_count);
2318 reiserfs_write_unlock(inode->i_sb);
2320 /* this is where we fill in holes in the file. */
2321 if (use_get_block) {
2322 retval = reiserfs_get_block(inode, block, bh_result,
2323 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2324 | GET_BLOCK_NO_DANGLE);
2326 if (!buffer_mapped(bh_result)
2327 || bh_result->b_blocknr == 0) {
2328 /* get_block failed to find a mapped unformatted node. */
2334 kunmap(bh_result->b_page);
2336 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2337 /* we've copied data from the page into the direct item, so the
2338 * buffer in the page is now clean, mark it to reflect that.
2340 lock_buffer(bh_result);
2341 clear_buffer_dirty(bh_result);
2342 unlock_buffer(bh_result);
2348 * mason@suse.com: updated in 2.5.54 to follow the same general io
2349 * start/recovery path as __block_write_full_page, along with special
2350 * code to handle reiserfs tails.
2352 static int reiserfs_write_full_page(struct page *page,
2353 struct writeback_control *wbc)
2355 struct inode *inode = page->mapping->host;
2356 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2358 unsigned long block;
2359 sector_t last_block;
2360 struct buffer_head *head, *bh;
2363 int checked = PageFsMisc(page);
2364 struct reiserfs_transaction_handle th;
2365 struct super_block *s = inode->i_sb;
2366 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2369 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2370 if (checked && (current->flags & PF_MEMALLOC)) {
2371 redirty_page_for_writepage(wbc, page);
2376 /* The page dirty bit is cleared before writepage is called, which
2377 * means we have to tell create_empty_buffers to make dirty buffers
2378 * The page really should be up to date at this point, so tossing
2379 * in the BH_Uptodate is just a sanity check.
2381 if (!page_has_buffers(page)) {
2382 create_empty_buffers(page, s->s_blocksize,
2383 (1 << BH_Dirty) | (1 << BH_Uptodate));
2385 head = page_buffers(page);
2387 /* last page in the file, zero out any contents past the
2388 ** last byte in the file
2390 if (page->index >= end_index) {
2392 unsigned last_offset;
2394 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2395 /* no file contents in this page */
2396 if (page->index >= end_index + 1 || !last_offset) {
2400 kaddr = kmap_atomic(page, KM_USER0);
2401 memset(kaddr + last_offset, 0, PAGE_CACHE_SIZE - last_offset);
2402 flush_dcache_page(page);
2403 kunmap_atomic(kaddr, KM_USER0);
2406 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
2407 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
2408 /* first map all the buffers, logging any direct items we find */
2410 if (block > last_block) {
2412 * This can happen when the block size is less than
2413 * the page size. The corresponding bytes in the page
2414 * were zero filled above
2416 clear_buffer_dirty(bh);
2417 set_buffer_uptodate(bh);
2418 } else if ((checked || buffer_dirty(bh)) &&
2419 (!buffer_mapped(bh) || (buffer_mapped(bh)
2422 /* not mapped yet, or it points to a direct item, search
2423 * the btree for the mapping info, and log any direct
2426 if ((error = map_block_for_writepage(inode, bh, block))) {
2430 bh = bh->b_this_page;
2432 } while (bh != head);
2435 * we start the transaction after map_block_for_writepage,
2436 * because it can create holes in the file (an unbounded operation).
2437 * starting it here, we can make a reliable estimate for how many
2438 * blocks we're going to log
2441 ClearPageFsMisc(page);
2442 reiserfs_write_lock(s);
2443 error = journal_begin(&th, s, bh_per_page + 1);
2445 reiserfs_write_unlock(s);
2448 reiserfs_update_inode_transaction(inode);
2450 /* now go through and lock any dirty buffers on the page */
2453 if (!buffer_mapped(bh))
2455 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2459 reiserfs_prepare_for_journal(s, bh, 1);
2460 journal_mark_dirty(&th, s, bh);
2463 /* from this point on, we know the buffer is mapped to a
2464 * real block and not a direct item
2466 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2469 if (test_set_buffer_locked(bh)) {
2470 redirty_page_for_writepage(wbc, page);
2474 if (test_clear_buffer_dirty(bh)) {
2475 mark_buffer_async_write(bh);
2479 } while ((bh = bh->b_this_page) != head);
2482 error = journal_end(&th, s, bh_per_page + 1);
2483 reiserfs_write_unlock(s);
2487 BUG_ON(PageWriteback(page));
2488 set_page_writeback(page);
2492 * since any buffer might be the only dirty buffer on the page,
2493 * the first submit_bh can bring the page out of writeback.
2494 * be careful with the buffers.
2497 struct buffer_head *next = bh->b_this_page;
2498 if (buffer_async_write(bh)) {
2499 submit_bh(WRITE, bh);
2504 } while (bh != head);
2510 * if this page only had a direct item, it is very possible for
2511 * no io to be required without there being an error. Or,
2512 * someone else could have locked them and sent them down the
2513 * pipe without locking the page
2517 if (!buffer_uptodate(bh)) {
2521 bh = bh->b_this_page;
2522 } while (bh != head);
2524 SetPageUptodate(page);
2525 end_page_writeback(page);
2530 /* catches various errors, we need to make sure any valid dirty blocks
2531 * get to the media. The page is currently locked and not marked for
2534 ClearPageUptodate(page);
2538 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2540 mark_buffer_async_write(bh);
2543 * clear any dirty bits that might have come from getting
2544 * attached to a dirty page
2546 clear_buffer_dirty(bh);
2548 bh = bh->b_this_page;
2549 } while (bh != head);
2551 BUG_ON(PageWriteback(page));
2552 set_page_writeback(page);
2555 struct buffer_head *next = bh->b_this_page;
2556 if (buffer_async_write(bh)) {
2557 clear_buffer_dirty(bh);
2558 submit_bh(WRITE, bh);
2563 } while (bh != head);
2567 static int reiserfs_readpage(struct file *f, struct page *page)
2569 return block_read_full_page(page, reiserfs_get_block);
2572 static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
2574 struct inode *inode = page->mapping->host;
2575 reiserfs_wait_on_write_block(inode->i_sb);
2576 return reiserfs_write_full_page(page, wbc);
2579 static int reiserfs_prepare_write(struct file *f, struct page *page,
2580 unsigned from, unsigned to)
2582 struct inode *inode = page->mapping->host;
2586 reiserfs_wait_on_write_block(inode->i_sb);
2587 fix_tail_page_for_writing(page);
2588 if (reiserfs_transaction_running(inode->i_sb)) {
2589 struct reiserfs_transaction_handle *th;
2590 th = (struct reiserfs_transaction_handle *)current->
2592 BUG_ON(!th->t_refcount);
2593 BUG_ON(!th->t_trans_id);
2594 old_ref = th->t_refcount;
2598 ret = block_prepare_write(page, from, to, reiserfs_get_block);
2599 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2600 struct reiserfs_transaction_handle *th = current->journal_info;
2601 /* this gets a little ugly. If reiserfs_get_block returned an
2602 * error and left a transacstion running, we've got to close it,
2603 * and we've got to free handle if it was a persistent transaction.
2605 * But, if we had nested into an existing transaction, we need
2606 * to just drop the ref count on the handle.
2608 * If old_ref == 0, the transaction is from reiserfs_get_block,
2609 * and it was a persistent trans. Otherwise, it was nested above.
2611 if (th->t_refcount > old_ref) {
2616 reiserfs_write_lock(inode->i_sb);
2617 err = reiserfs_end_persistent_transaction(th);
2618 reiserfs_write_unlock(inode->i_sb);
2628 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2630 return generic_block_bmap(as, block, reiserfs_bmap);
2633 static int reiserfs_commit_write(struct file *f, struct page *page,
2634 unsigned from, unsigned to)
2636 struct inode *inode = page->mapping->host;
2637 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2640 struct reiserfs_transaction_handle *th = NULL;
2642 reiserfs_wait_on_write_block(inode->i_sb);
2643 if (reiserfs_transaction_running(inode->i_sb)) {
2644 th = current->journal_info;
2646 reiserfs_commit_page(inode, page, from, to);
2648 /* generic_commit_write does this for us, but does not update the
2649 ** transaction tracking stuff when the size changes. So, we have
2650 ** to do the i_size updates here.
2652 if (pos > inode->i_size) {
2653 struct reiserfs_transaction_handle myth;
2654 reiserfs_write_lock(inode->i_sb);
2655 /* If the file have grown beyond the border where it
2656 can have a tail, unmark it as needing a tail
2658 if ((have_large_tails(inode->i_sb)
2659 && inode->i_size > i_block_size(inode) * 4)
2660 || (have_small_tails(inode->i_sb)
2661 && inode->i_size > i_block_size(inode)))
2662 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2664 ret = journal_begin(&myth, inode->i_sb, 1);
2666 reiserfs_write_unlock(inode->i_sb);
2669 reiserfs_update_inode_transaction(inode);
2670 inode->i_size = pos;
2672 * this will just nest into our transaction. It's important
2673 * to use mark_inode_dirty so the inode gets pushed around on the
2674 * dirty lists, and so that O_SYNC works as expected
2676 mark_inode_dirty(inode);
2677 reiserfs_update_sd(&myth, inode);
2679 ret = journal_end(&myth, inode->i_sb, 1);
2680 reiserfs_write_unlock(inode->i_sb);
2685 reiserfs_write_lock(inode->i_sb);
2687 mark_inode_dirty(inode);
2688 ret = reiserfs_end_persistent_transaction(th);
2689 reiserfs_write_unlock(inode->i_sb);
2699 reiserfs_write_lock(inode->i_sb);
2701 reiserfs_update_sd(th, inode);
2702 ret = reiserfs_end_persistent_transaction(th);
2703 reiserfs_write_unlock(inode->i_sb);
2709 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2711 if (reiserfs_attrs(inode->i_sb)) {
2712 if (sd_attrs & REISERFS_SYNC_FL)
2713 inode->i_flags |= S_SYNC;
2715 inode->i_flags &= ~S_SYNC;
2716 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2717 inode->i_flags |= S_IMMUTABLE;
2719 inode->i_flags &= ~S_IMMUTABLE;
2720 if (sd_attrs & REISERFS_IUNLINK_FL)
2721 inode->i_flags |= S_IUNLINK;
2723 inode->i_flags &= ~S_IUNLINK;
2724 if (sd_attrs & REISERFS_BARRIER_FL)
2725 inode->i_flags |= S_BARRIER;
2727 inode->i_flags &= ~S_BARRIER;
2728 if (sd_attrs & REISERFS_APPEND_FL)
2729 inode->i_flags |= S_APPEND;
2731 inode->i_flags &= ~S_APPEND;
2732 if (sd_attrs & REISERFS_NOATIME_FL)
2733 inode->i_flags |= S_NOATIME;
2735 inode->i_flags &= ~S_NOATIME;
2736 if (sd_attrs & REISERFS_NOTAIL_FL)
2737 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2739 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2743 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
2745 if (reiserfs_attrs(inode->i_sb)) {
2746 if (inode->i_flags & S_IMMUTABLE)
2747 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2749 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2750 if (inode->i_flags & S_IUNLINK)
2751 *sd_attrs |= REISERFS_IUNLINK_FL;
2753 *sd_attrs &= ~REISERFS_IUNLINK_FL;
2754 if (inode->i_flags & S_BARRIER)
2755 *sd_attrs |= REISERFS_BARRIER_FL;
2757 *sd_attrs &= ~REISERFS_BARRIER_FL;
2758 if (inode->i_flags & S_SYNC)
2759 *sd_attrs |= REISERFS_SYNC_FL;
2761 *sd_attrs &= ~REISERFS_SYNC_FL;
2762 if (inode->i_flags & S_NOATIME)
2763 *sd_attrs |= REISERFS_NOATIME_FL;
2765 *sd_attrs &= ~REISERFS_NOATIME_FL;
2766 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
2767 *sd_attrs |= REISERFS_NOTAIL_FL;
2769 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2773 /* decide if this buffer needs to stay around for data logging or ordered
2776 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2779 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2782 spin_lock(&j->j_dirty_buffers_lock);
2783 if (!buffer_mapped(bh)) {
2786 /* the page is locked, and the only places that log a data buffer
2787 * also lock the page.
2789 if (reiserfs_file_data_log(inode)) {
2791 * very conservative, leave the buffer pinned if
2792 * anyone might need it.
2794 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2797 } else if (buffer_dirty(bh)) {
2798 struct reiserfs_journal_list *jl;
2799 struct reiserfs_jh *jh = bh->b_private;
2801 /* why is this safe?
2802 * reiserfs_setattr updates i_size in the on disk
2803 * stat data before allowing vmtruncate to be called.
2805 * If buffer was put onto the ordered list for this
2806 * transaction, we know for sure either this transaction
2807 * or an older one already has updated i_size on disk,
2808 * and this ordered data won't be referenced in the file
2811 * if the buffer was put onto the ordered list for an older
2812 * transaction, we need to leave it around
2814 if (jh && (jl = jh->jl)
2815 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2819 if (ret && bh->b_private) {
2820 reiserfs_free_jh(bh);
2822 spin_unlock(&j->j_dirty_buffers_lock);
2827 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2828 static void reiserfs_invalidatepage(struct page *page, unsigned long offset)
2830 struct buffer_head *head, *bh, *next;
2831 struct inode *inode = page->mapping->host;
2832 unsigned int curr_off = 0;
2835 BUG_ON(!PageLocked(page));
2838 ClearPageFsMisc(page);
2840 if (!page_has_buffers(page))
2843 head = page_buffers(page);
2846 unsigned int next_off = curr_off + bh->b_size;
2847 next = bh->b_this_page;
2850 * is this block fully invalidated?
2852 if (offset <= curr_off) {
2853 if (invalidatepage_can_drop(inode, bh))
2854 reiserfs_unmap_buffer(bh);
2858 curr_off = next_off;
2860 } while (bh != head);
2863 * We release buffers only if the entire page is being invalidated.
2864 * The get_block cached value has been unconditionally invalidated,
2865 * so real IO is not possible anymore.
2867 if (!offset && ret) {
2868 ret = try_to_release_page(page, 0);
2869 /* maybe should BUG_ON(!ret); - neilb */
2875 static int reiserfs_set_page_dirty(struct page *page)
2877 struct inode *inode = page->mapping->host;
2878 if (reiserfs_file_data_log(inode)) {
2879 SetPageFsMisc(page);
2880 return __set_page_dirty_nobuffers(page);
2882 return __set_page_dirty_buffers(page);
2886 * Returns 1 if the page's buffers were dropped. The page is locked.
2888 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
2889 * in the buffers at page_buffers(page).
2891 * even in -o notail mode, we can't be sure an old mount without -o notail
2892 * didn't create files with tails.
2894 static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
2896 struct inode *inode = page->mapping->host;
2897 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2898 struct buffer_head *head;
2899 struct buffer_head *bh;
2902 WARN_ON(PageFsMisc(page));
2903 spin_lock(&j->j_dirty_buffers_lock);
2904 head = page_buffers(page);
2907 if (bh->b_private) {
2908 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
2909 reiserfs_free_jh(bh);
2915 bh = bh->b_this_page;
2916 } while (bh != head);
2918 ret = try_to_free_buffers(page);
2919 spin_unlock(&j->j_dirty_buffers_lock);
2923 /* We thank Mingming Cao for helping us understand in great detail what
2924 to do in this section of the code. */
2925 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
2926 const struct iovec *iov, loff_t offset,
2927 unsigned long nr_segs)
2929 struct file *file = iocb->ki_filp;
2930 struct inode *inode = file->f_mapping->host;
2932 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2934 reiserfs_get_blocks_direct_io, NULL);
2937 int reiserfs_sync_flags(struct inode *inode)
2939 u16 oldflags, newflags;
2941 oldflags = REISERFS_I(inode)->i_attrs;
2942 newflags = oldflags;
2943 i_attrs_to_sd_attrs(inode, &newflags);
2945 if (oldflags ^ newflags) {
2946 REISERFS_I(inode)->i_attrs = newflags;
2947 inode->i_ctime = CURRENT_TIME_SEC;
2948 mark_inode_dirty(inode);
2953 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
2955 struct inode *inode = dentry->d_inode;
2957 unsigned int ia_valid = attr->ia_valid;
2958 reiserfs_write_lock(inode->i_sb);
2959 if (attr->ia_valid & ATTR_SIZE) {
2960 /* version 2 items will be caught by the s_maxbytes check
2961 ** done for us in vmtruncate
2963 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
2964 attr->ia_size > MAX_NON_LFS) {
2968 /* fill in hole pointers in the expanding truncate case. */
2969 if (attr->ia_size > inode->i_size) {
2970 error = generic_cont_expand(inode, attr->ia_size);
2971 if (REISERFS_I(inode)->i_prealloc_count > 0) {
2973 struct reiserfs_transaction_handle th;
2974 /* we're changing at most 2 bitmaps, inode + super */
2975 err = journal_begin(&th, inode->i_sb, 4);
2977 reiserfs_discard_prealloc(&th, inode);
2978 err = journal_end(&th, inode->i_sb, 4);
2986 * file size is changed, ctime and mtime are
2989 attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
2993 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
2994 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
2995 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
2996 /* stat data of format v3.5 has 16 bit uid and gid */
3001 error = inode_change_ok(inode, attr);
3004 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
3005 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid) ||
3006 (ia_valid & ATTR_XID && attr->ia_xid != inode->i_xid)) {
3007 error = reiserfs_chown_xattrs(inode, attr);
3010 struct reiserfs_transaction_handle th;
3013 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3014 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3017 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3019 journal_begin(&th, inode->i_sb,
3024 DQUOT_TRANSFER(inode, attr) ? -EDQUOT : 0;
3026 journal_end(&th, inode->i_sb,
3030 /* Update corresponding info in inode so that everything is in
3031 * one transaction */
3032 if (attr->ia_valid & ATTR_UID)
3033 inode->i_uid = attr->ia_uid;
3034 if (attr->ia_valid & ATTR_GID)
3035 inode->i_gid = attr->ia_gid;
3036 if ((attr->ia_valid & ATTR_XID) &&
3038 inode->i_xid = attr->ia_xid;
3039 mark_inode_dirty(inode);
3041 journal_end(&th, inode->i_sb, jbegin_count);
3045 error = inode_setattr(inode, attr);
3048 if (!error && reiserfs_posixacl(inode->i_sb)) {
3049 if (attr->ia_valid & ATTR_MODE)
3050 error = reiserfs_acl_chmod(inode);
3054 reiserfs_write_unlock(inode->i_sb);
3058 const struct address_space_operations reiserfs_address_space_operations = {
3059 .writepage = reiserfs_writepage,
3060 .readpage = reiserfs_readpage,
3061 .readpages = reiserfs_readpages,
3062 .releasepage = reiserfs_releasepage,
3063 .invalidatepage = reiserfs_invalidatepage,
3064 .sync_page = block_sync_page,
3065 .prepare_write = reiserfs_prepare_write,
3066 .commit_write = reiserfs_commit_write,
3067 .bmap = reiserfs_aop_bmap,
3068 .direct_IO = reiserfs_direct_IO,
3069 .set_page_dirty = reiserfs_set_page_dirty,