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/vs_tag.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 treepath *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);
224 /* we cannot restart while nested */
225 if (th->t_refcount > 1) {
228 reiserfs_update_sd(th, inode);
229 err = journal_end(th, s, len);
231 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
233 reiserfs_update_inode_transaction(inode);
238 // it is called by get_block when create == 0. Returns block number
239 // for 'block'-th logical block of file. When it hits direct item it
240 // returns 0 (being called from bmap) or read direct item into piece
241 // of page (bh_result)
243 // Please improve the english/clarity in the comment above, as it is
244 // hard to understand.
246 static int _get_block_create_0(struct inode *inode, long block,
247 struct buffer_head *bh_result, int args)
249 INITIALIZE_PATH(path);
251 struct buffer_head *bh;
252 struct item_head *ih, tmp_ih;
260 unsigned long offset;
262 // prepare the key to look for the 'block'-th block of file
263 make_cpu_key(&key, inode,
264 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
268 result = search_for_position_by_key(inode->i_sb, &key, &path);
269 if (result != POSITION_FOUND) {
272 kunmap(bh_result->b_page);
273 if (result == IO_ERROR)
275 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
276 // That there is some MMAPED data associated with it that is yet to be written to disk.
277 if ((args & GET_BLOCK_NO_HOLE)
278 && !PageUptodate(bh_result->b_page)) {
284 bh = get_last_bh(&path);
286 if (is_indirect_le_ih(ih)) {
287 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih);
289 /* FIXME: here we could cache indirect item or part of it in
290 the inode to avoid search_by_key in case of subsequent
292 blocknr = get_block_num(ind_item, path.pos_in_item);
295 map_bh(bh_result, inode->i_sb, blocknr);
296 if (path.pos_in_item ==
297 ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
298 set_buffer_boundary(bh_result);
301 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
302 // That there is some MMAPED data associated with it that is yet to be written to disk.
303 if ((args & GET_BLOCK_NO_HOLE)
304 && !PageUptodate(bh_result->b_page)) {
310 kunmap(bh_result->b_page);
313 // requested data are in direct item(s)
314 if (!(args & GET_BLOCK_READ_DIRECT)) {
315 // we are called by bmap. FIXME: we can not map block of file
316 // when it is stored in direct item(s)
319 kunmap(bh_result->b_page);
323 /* if we've got a direct item, and the buffer or page was uptodate,
324 ** we don't want to pull data off disk again. skip to the
325 ** end, where we map the buffer and return
327 if (buffer_uptodate(bh_result)) {
331 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
332 ** pages without any buffers. If the page is up to date, we don't want
333 ** read old data off disk. Set the up to date bit on the buffer instead
334 ** and jump to the end
336 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
337 set_buffer_uptodate(bh_result);
340 // read file tail into part of page
341 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
342 fs_gen = get_generation(inode->i_sb);
343 copy_item_head(&tmp_ih, ih);
345 /* we only want to kmap if we are reading the tail into the page.
346 ** this is not the common case, so we don't kmap until we are
347 ** sure we need to. But, this means the item might move if
351 p = (char *)kmap(bh_result->b_page);
352 if (fs_changed(fs_gen, inode->i_sb)
353 && item_moved(&tmp_ih, &path)) {
358 memset(p, 0, inode->i_sb->s_blocksize);
360 if (!is_direct_le_ih(ih)) {
363 /* make sure we don't read more bytes than actually exist in
364 ** the file. This can happen in odd cases where i_size isn't
365 ** correct, and when direct item padding results in a few
366 ** extra bytes at the end of the direct item
368 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
370 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
372 inode->i_size - (le_ih_k_offset(ih) - 1) -
376 chars = ih_item_len(ih) - path.pos_in_item;
378 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
385 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
386 // we done, if read direct item is not the last item of
387 // node FIXME: we could try to check right delimiting key
388 // to see whether direct item continues in the right
389 // neighbor or rely on i_size
392 // update key to look for the next piece
393 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
394 result = search_for_position_by_key(inode->i_sb, &key, &path);
395 if (result != POSITION_FOUND)
396 // i/o error most likely
398 bh = get_last_bh(&path);
402 flush_dcache_page(bh_result->b_page);
403 kunmap(bh_result->b_page);
408 if (result == IO_ERROR)
411 /* this buffer has valid data, but isn't valid for io. mapping it to
412 * block #0 tells the rest of reiserfs it just has a tail in it
414 map_bh(bh_result, inode->i_sb, 0);
415 set_buffer_uptodate(bh_result);
419 // this is called to create file map. So, _get_block_create_0 will not
421 static int reiserfs_bmap(struct inode *inode, sector_t block,
422 struct buffer_head *bh_result, int create)
424 if (!file_capable(inode, block))
427 reiserfs_write_lock(inode->i_sb);
428 /* do not read the direct item */
429 _get_block_create_0(inode, block, bh_result, 0);
430 reiserfs_write_unlock(inode->i_sb);
434 /* special version of get_block that is only used by grab_tail_page right
435 ** now. It is sent to block_prepare_write, and when you try to get a
436 ** block past the end of the file (or a block from a hole) it returns
437 ** -ENOENT instead of a valid buffer. block_prepare_write expects to
438 ** be able to do i/o on the buffers returned, unless an error value
441 ** So, this allows block_prepare_write to be used for reading a single block
442 ** in a page. Where it does not produce a valid page for holes, or past the
443 ** end of the file. This turns out to be exactly what we need for reading
444 ** tails for conversion.
446 ** The point of the wrapper is forcing a certain value for create, even
447 ** though the VFS layer is calling this function with create==1. If you
448 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
449 ** don't use this function.
451 static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
452 struct buffer_head *bh_result,
455 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
458 /* This is special helper for reiserfs_get_block in case we are executing
459 direct_IO request. */
460 static int reiserfs_get_blocks_direct_io(struct inode *inode,
462 struct buffer_head *bh_result,
467 bh_result->b_page = NULL;
469 /* We set the b_size before reiserfs_get_block call since it is
470 referenced in convert_tail_for_hole() that may be called from
471 reiserfs_get_block() */
472 bh_result->b_size = (1 << inode->i_blkbits);
474 ret = reiserfs_get_block(inode, iblock, bh_result,
475 create | GET_BLOCK_NO_DANGLE);
479 /* don't allow direct io onto tail pages */
480 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
481 /* make sure future calls to the direct io funcs for this offset
482 ** in the file fail by unmapping the buffer
484 clear_buffer_mapped(bh_result);
487 /* Possible unpacked tail. Flush the data before pages have
489 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
492 err = reiserfs_commit_for_inode(inode);
493 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
503 ** helper function for when reiserfs_get_block is called for a hole
504 ** but the file tail is still in a direct item
505 ** bh_result is the buffer head for the hole
506 ** tail_offset is the offset of the start of the tail in the file
508 ** This calls prepare_write, which will start a new transaction
509 ** you should not be in a transaction, or have any paths held when you
512 static int convert_tail_for_hole(struct inode *inode,
513 struct buffer_head *bh_result,
517 unsigned long tail_end;
518 unsigned long tail_start;
519 struct page *tail_page;
520 struct page *hole_page = bh_result->b_page;
523 if ((tail_offset & (bh_result->b_size - 1)) != 1)
526 /* always try to read until the end of the block */
527 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
528 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
530 index = tail_offset >> PAGE_CACHE_SHIFT;
531 /* hole_page can be zero in case of direct_io, we are sure
532 that we cannot get here if we write with O_DIRECT into
534 if (!hole_page || index != hole_page->index) {
535 tail_page = grab_cache_page(inode->i_mapping, index);
541 tail_page = hole_page;
544 /* we don't have to make sure the conversion did not happen while
545 ** we were locking the page because anyone that could convert
546 ** must first take i_mutex.
548 ** We must fix the tail page for writing because it might have buffers
549 ** that are mapped, but have a block number of 0. This indicates tail
550 ** data that has been read directly into the page, and block_prepare_write
551 ** won't trigger a get_block in this case.
553 fix_tail_page_for_writing(tail_page);
554 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
558 /* tail conversion might change the data in the page */
559 flush_dcache_page(tail_page);
561 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
564 if (tail_page != hole_page) {
565 unlock_page(tail_page);
566 page_cache_release(tail_page);
572 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
575 b_blocknr_t * allocated_block_nr,
576 struct treepath *path, int flags)
578 BUG_ON(!th->t_trans_id);
580 #ifdef REISERFS_PREALLOCATE
581 if (!(flags & GET_BLOCK_NO_IMUX)) {
582 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
586 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
590 int reiserfs_get_block(struct inode *inode, sector_t block,
591 struct buffer_head *bh_result, int create)
593 int repeat, retval = 0;
594 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
595 INITIALIZE_PATH(path);
598 struct buffer_head *bh, *unbh = NULL;
599 struct item_head *ih, tmp_ih;
603 struct reiserfs_transaction_handle *th = NULL;
604 /* space reserved in transaction batch:
605 . 3 balancings in direct->indirect conversion
606 . 1 block involved into reiserfs_update_sd()
607 XXX in practically impossible worst case direct2indirect()
608 can incur (much) more than 3 balancings.
609 quota update for user, group */
611 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
612 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
616 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
619 reiserfs_write_lock(inode->i_sb);
620 version = get_inode_item_key_version(inode);
622 if (!file_capable(inode, block)) {
623 reiserfs_write_unlock(inode->i_sb);
627 /* if !create, we aren't changing the FS, so we don't need to
628 ** log anything, so we don't need to start a transaction
630 if (!(create & GET_BLOCK_CREATE)) {
632 /* find number of block-th logical block of the file */
633 ret = _get_block_create_0(inode, block, bh_result,
634 create | GET_BLOCK_READ_DIRECT);
635 reiserfs_write_unlock(inode->i_sb);
639 * if we're already in a transaction, make sure to close
640 * any new transactions we start in this func
642 if ((create & GET_BLOCK_NO_DANGLE) ||
643 reiserfs_transaction_running(inode->i_sb))
646 /* If file is of such a size, that it might have a tail and tails are enabled
647 ** we should mark it as possibly needing tail packing on close
649 if ((have_large_tails(inode->i_sb)
650 && inode->i_size < i_block_size(inode) * 4)
651 || (have_small_tails(inode->i_sb)
652 && inode->i_size < i_block_size(inode)))
653 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
655 /* set the key of the first byte in the 'block'-th block of file */
656 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
657 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
659 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
664 reiserfs_update_inode_transaction(inode);
668 retval = search_for_position_by_key(inode->i_sb, &key, &path);
669 if (retval == IO_ERROR) {
674 bh = get_last_bh(&path);
676 item = get_item(&path);
677 pos_in_item = path.pos_in_item;
679 fs_gen = get_generation(inode->i_sb);
680 copy_item_head(&tmp_ih, ih);
682 if (allocation_needed
683 (retval, allocated_block_nr, ih, item, pos_in_item)) {
684 /* we have to allocate block for the unformatted node */
691 _allocate_block(th, block, inode, &allocated_block_nr,
694 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
695 /* restart the transaction to give the journal a chance to free
696 ** some blocks. releases the path, so we have to go back to
697 ** research if we succeed on the second try
699 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
700 retval = restart_transaction(th, inode, &path);
704 _allocate_block(th, block, inode,
705 &allocated_block_nr, NULL, create);
707 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
710 if (repeat == QUOTA_EXCEEDED)
717 if (fs_changed(fs_gen, inode->i_sb)
718 && item_moved(&tmp_ih, &path)) {
723 if (indirect_item_found(retval, ih)) {
724 b_blocknr_t unfm_ptr;
725 /* 'block'-th block is in the file already (there is
726 corresponding cell in some indirect item). But it may be
727 zero unformatted node pointer (hole) */
728 unfm_ptr = get_block_num(item, pos_in_item);
730 /* use allocated block to plug the hole */
731 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
732 if (fs_changed(fs_gen, inode->i_sb)
733 && item_moved(&tmp_ih, &path)) {
734 reiserfs_restore_prepared_buffer(inode->i_sb,
738 set_buffer_new(bh_result);
739 if (buffer_dirty(bh_result)
740 && reiserfs_data_ordered(inode->i_sb))
741 reiserfs_add_ordered_list(inode, bh_result);
742 put_block_num(item, pos_in_item, allocated_block_nr);
743 unfm_ptr = allocated_block_nr;
744 journal_mark_dirty(th, inode->i_sb, bh);
745 reiserfs_update_sd(th, inode);
747 set_block_dev_mapped(bh_result, unfm_ptr, inode);
751 retval = reiserfs_end_persistent_transaction(th);
753 reiserfs_write_unlock(inode->i_sb);
755 /* the item was found, so new blocks were not added to the file
756 ** there is no need to make sure the inode is updated with this
767 /* desired position is not found or is in the direct item. We have
768 to append file with holes up to 'block'-th block converting
769 direct items to indirect one if necessary */
772 if (is_statdata_le_ih(ih)) {
774 struct cpu_key tmp_key;
776 /* indirect item has to be inserted */
777 make_le_item_head(&tmp_ih, &key, version, 1,
778 TYPE_INDIRECT, UNFM_P_SIZE,
779 0 /* free_space */ );
781 if (cpu_key_k_offset(&key) == 1) {
782 /* we are going to add 'block'-th block to the file. Use
783 allocated block for that */
784 unp = cpu_to_le32(allocated_block_nr);
785 set_block_dev_mapped(bh_result,
786 allocated_block_nr, inode);
787 set_buffer_new(bh_result);
791 set_cpu_key_k_offset(&tmp_key, 1);
792 PATH_LAST_POSITION(&path)++;
795 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
796 inode, (char *)&unp);
798 reiserfs_free_block(th, inode,
799 allocated_block_nr, 1);
800 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
802 //mark_tail_converted (inode);
803 } else if (is_direct_le_ih(ih)) {
804 /* direct item has to be converted */
808 ((le_ih_k_offset(ih) -
809 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
810 if (tail_offset == cpu_key_k_offset(&key)) {
811 /* direct item we just found fits into block we have
812 to map. Convert it into unformatted node: use
813 bh_result for the conversion */
814 set_block_dev_mapped(bh_result,
815 allocated_block_nr, inode);
819 /* we have to padd file tail stored in direct item(s)
820 up to block size and convert it to unformatted
821 node. FIXME: this should also get into page cache */
825 * ugly, but we can only end the transaction if
828 BUG_ON(!th->t_refcount);
829 if (th->t_refcount == 1) {
831 reiserfs_end_persistent_transaction
839 convert_tail_for_hole(inode, bh_result,
842 if (retval != -ENOSPC)
843 reiserfs_warning(inode->i_sb,
844 "clm-6004: convert tail failed inode %lu, error %d",
847 if (allocated_block_nr) {
848 /* the bitmap, the super, and the stat data == 3 */
850 th = reiserfs_persistent_transaction(inode->i_sb, 3);
852 reiserfs_free_block(th,
862 direct2indirect(th, inode, &path, unbh,
865 reiserfs_unmap_buffer(unbh);
866 reiserfs_free_block(th, inode,
867 allocated_block_nr, 1);
870 /* it is important the set_buffer_uptodate is done after
871 ** the direct2indirect. The buffer might contain valid
872 ** data newer than the data on disk (read by readpage, changed,
873 ** and then sent here by writepage). direct2indirect needs
874 ** to know if unbh was already up to date, so it can decide
875 ** if the data in unbh needs to be replaced with data from
878 set_buffer_uptodate(unbh);
880 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
881 buffer will disappear shortly, so it should not be added to
884 /* we've converted the tail, so we must
885 ** flush unbh before the transaction commits
887 reiserfs_add_tail_list(inode, unbh);
889 /* mark it dirty now to prevent commit_write from adding
890 ** this buffer to the inode's dirty buffer list
893 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
894 * It's still atomic, but it sets the page dirty too,
895 * which makes it eligible for writeback at any time by the
896 * VM (which was also the case with __mark_buffer_dirty())
898 mark_buffer_dirty(unbh);
901 /* append indirect item with holes if needed, when appending
902 pointer to 'block'-th block use block, which is already
904 struct cpu_key tmp_key;
905 unp_t unf_single = 0; // We use this in case we need to allocate only
906 // one block which is a fastpath
908 __u64 max_to_insert =
909 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
913 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
914 "vs-804: invalid position for append");
915 /* indirect item has to be appended, set up key of that position */
916 make_cpu_key(&tmp_key, inode,
917 le_key_k_offset(version,
920 inode->i_sb->s_blocksize),
921 //pos_in_item * inode->i_sb->s_blocksize,
922 TYPE_INDIRECT, 3); // key type is unimportant
924 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
925 "green-805: invalid offset");
928 ((cpu_key_k_offset(&key) -
929 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
932 if (blocks_needed == 1) {
935 un = kzalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_ATOMIC); // We need to avoid scheduling.
942 if (blocks_needed <= max_to_insert) {
943 /* we are going to add target block to the file. Use allocated
945 un[blocks_needed - 1] =
946 cpu_to_le32(allocated_block_nr);
947 set_block_dev_mapped(bh_result,
948 allocated_block_nr, inode);
949 set_buffer_new(bh_result);
952 /* paste hole to the indirect item */
953 /* If kmalloc failed, max_to_insert becomes zero and it means we
954 only have space for one block */
956 max_to_insert ? max_to_insert : 1;
959 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
964 if (blocks_needed != 1)
968 reiserfs_free_block(th, inode,
969 allocated_block_nr, 1);
973 /* We need to mark new file size in case this function will be
974 interrupted/aborted later on. And we may do this only for
977 inode->i_sb->s_blocksize * blocks_needed;
984 /* this loop could log more blocks than we had originally asked
985 ** for. So, we have to allow the transaction to end if it is
986 ** too big or too full. Update the inode so things are
987 ** consistent if we crash before the function returns
989 ** release the path so that anybody waiting on the path before
990 ** ending their transaction will be able to continue.
992 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
993 retval = restart_transaction(th, inode, &path);
997 /* inserting indirect pointers for a hole can take a
998 ** long time. reschedule if needed
1002 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1003 if (retval == IO_ERROR) {
1007 if (retval == POSITION_FOUND) {
1008 reiserfs_warning(inode->i_sb,
1009 "vs-825: reiserfs_get_block: "
1010 "%K should not be found", &key);
1012 if (allocated_block_nr)
1013 reiserfs_free_block(th, inode,
1014 allocated_block_nr, 1);
1018 bh = get_last_bh(&path);
1020 item = get_item(&path);
1021 pos_in_item = path.pos_in_item;
1027 if (th && (!dangle || (retval && !th->t_trans_id))) {
1030 reiserfs_update_sd(th, inode);
1031 err = reiserfs_end_persistent_transaction(th);
1036 reiserfs_write_unlock(inode->i_sb);
1037 reiserfs_check_path(&path);
1042 reiserfs_readpages(struct file *file, struct address_space *mapping,
1043 struct list_head *pages, unsigned nr_pages)
1045 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
1048 /* Compute real number of used bytes by file
1049 * Following three functions can go away when we'll have enough space in stat item
1051 static int real_space_diff(struct inode *inode, int sd_size)
1054 loff_t blocksize = inode->i_sb->s_blocksize;
1056 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1059 /* End of file is also in full block with indirect reference, so round
1060 ** up to the next block.
1062 ** there is just no way to know if the tail is actually packed
1063 ** on the file, so we have to assume it isn't. When we pack the
1064 ** tail, we add 4 bytes to pretend there really is an unformatted
1069 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1074 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
1077 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1078 return inode->i_size +
1079 (loff_t) (real_space_diff(inode, sd_size));
1081 return ((loff_t) real_space_diff(inode, sd_size)) +
1082 (((loff_t) blocks) << 9);
1085 /* Compute number of blocks used by file in ReiserFS counting */
1086 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1088 loff_t bytes = inode_get_bytes(inode);
1089 loff_t real_space = real_space_diff(inode, sd_size);
1091 /* keeps fsck and non-quota versions of reiserfs happy */
1092 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1093 bytes += (loff_t) 511;
1096 /* files from before the quota patch might i_blocks such that
1097 ** bytes < real_space. Deal with that here to prevent it from
1100 if (bytes < real_space)
1102 return (bytes - real_space) >> 9;
1106 // BAD: new directories have stat data of new type and all other items
1107 // of old type. Version stored in the inode says about body items, so
1108 // in update_stat_data we can not rely on inode, but have to check
1109 // item version directly
1112 // called by read_locked_inode
1113 static void init_inode(struct inode *inode, struct treepath *path)
1115 struct buffer_head *bh;
1116 struct item_head *ih;
1120 //int version = ITEM_VERSION_1;
1122 bh = PATH_PLAST_BUFFER(path);
1123 ih = PATH_PITEM_HEAD(path);
1125 copy_key(INODE_PKEY(inode), &(ih->ih_key));
1127 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1128 REISERFS_I(inode)->i_flags = 0;
1129 REISERFS_I(inode)->i_prealloc_block = 0;
1130 REISERFS_I(inode)->i_prealloc_count = 0;
1131 REISERFS_I(inode)->i_trans_id = 0;
1132 REISERFS_I(inode)->i_jl = NULL;
1133 mutex_init(&(REISERFS_I(inode)->i_mmap));
1134 reiserfs_init_acl_access(inode);
1135 reiserfs_init_acl_default(inode);
1136 reiserfs_init_xattr_rwsem(inode);
1138 if (stat_data_v1(ih)) {
1139 struct stat_data_v1 *sd =
1140 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1141 unsigned long blocks;
1143 uid = sd_v1_uid(sd);
1144 gid = sd_v1_gid(sd);
1146 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1147 set_inode_sd_version(inode, STAT_DATA_V1);
1148 inode->i_mode = sd_v1_mode(sd);
1149 inode->i_nlink = sd_v1_nlink(sd);
1150 inode->i_size = sd_v1_size(sd);
1151 inode->i_atime.tv_sec = sd_v1_atime(sd);
1152 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1153 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1154 inode->i_atime.tv_nsec = 0;
1155 inode->i_ctime.tv_nsec = 0;
1156 inode->i_mtime.tv_nsec = 0;
1158 inode->i_blocks = sd_v1_blocks(sd);
1159 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1160 blocks = (inode->i_size + 511) >> 9;
1161 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1162 if (inode->i_blocks > blocks) {
1163 // there was a bug in <=3.5.23 when i_blocks could take negative
1164 // values. Starting from 3.5.17 this value could even be stored in
1165 // stat data. For such files we set i_blocks based on file
1166 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1167 // only updated if file's inode will ever change
1168 inode->i_blocks = blocks;
1171 rdev = sd_v1_rdev(sd);
1172 REISERFS_I(inode)->i_first_direct_byte =
1173 sd_v1_first_direct_byte(sd);
1174 /* an early bug in the quota code can give us an odd number for the
1175 ** block count. This is incorrect, fix it here.
1177 if (inode->i_blocks & 1) {
1180 inode_set_bytes(inode,
1181 to_real_used_space(inode, inode->i_blocks,
1183 /* nopack is initially zero for v1 objects. For v2 objects,
1184 nopack is initialised from sd_attrs */
1185 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1187 // new stat data found, but object may have old items
1188 // (directories and symlinks)
1189 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1191 uid = sd_v2_uid(sd);
1192 gid = sd_v2_gid(sd);
1194 inode->i_mode = sd_v2_mode(sd);
1195 inode->i_nlink = sd_v2_nlink(sd);
1196 inode->i_size = sd_v2_size(sd);
1197 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1198 inode->i_atime.tv_sec = sd_v2_atime(sd);
1199 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1200 inode->i_ctime.tv_nsec = 0;
1201 inode->i_mtime.tv_nsec = 0;
1202 inode->i_atime.tv_nsec = 0;
1203 inode->i_blocks = sd_v2_blocks(sd);
1204 rdev = sd_v2_rdev(sd);
1205 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1206 inode->i_generation =
1207 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1209 inode->i_generation = sd_v2_generation(sd);
1211 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1212 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1214 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1215 REISERFS_I(inode)->i_first_direct_byte = 0;
1216 set_inode_sd_version(inode, STAT_DATA_V2);
1217 inode_set_bytes(inode,
1218 to_real_used_space(inode, inode->i_blocks,
1220 /* read persistent inode attributes from sd and initalise
1221 generic inode flags from them */
1222 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1223 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
1226 inode->i_uid = INOTAG_UID(DX_TAG(inode), uid, gid);
1227 inode->i_gid = INOTAG_GID(DX_TAG(inode), uid, gid);
1228 inode->i_tag = INOTAG_TAG(DX_TAG(inode), uid, gid, 0);
1231 if (S_ISREG(inode->i_mode)) {
1232 inode->i_op = &reiserfs_file_inode_operations;
1233 inode->i_fop = &reiserfs_file_operations;
1234 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1235 } else if (S_ISDIR(inode->i_mode)) {
1236 inode->i_op = &reiserfs_dir_inode_operations;
1237 inode->i_fop = &reiserfs_dir_operations;
1238 } else if (S_ISLNK(inode->i_mode)) {
1239 inode->i_op = &reiserfs_symlink_inode_operations;
1240 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1242 inode->i_blocks = 0;
1243 inode->i_op = &reiserfs_special_inode_operations;
1244 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1248 // update new stat data with inode fields
1249 static void inode2sd(void *sd, struct inode *inode, loff_t size)
1251 struct stat_data *sd_v2 = (struct stat_data *)sd;
1252 uid_t uid = TAGINO_UID(DX_TAG(inode), inode->i_uid, inode->i_tag);
1253 gid_t gid = TAGINO_GID(DX_TAG(inode), inode->i_gid, inode->i_tag);
1256 set_sd_v2_uid(sd_v2, uid);
1257 set_sd_v2_gid(sd_v2, gid);
1258 set_sd_v2_mode(sd_v2, inode->i_mode);
1259 set_sd_v2_nlink(sd_v2, inode->i_nlink);
1260 set_sd_v2_size(sd_v2, size);
1261 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1262 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1263 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1264 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1265 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1266 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1268 set_sd_v2_generation(sd_v2, inode->i_generation);
1269 flags = REISERFS_I(inode)->i_attrs;
1270 i_attrs_to_sd_attrs(inode, &flags);
1271 set_sd_v2_attrs(sd_v2, flags);
1274 // used to copy inode's fields to old stat data
1275 static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
1277 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
1279 set_sd_v1_mode(sd_v1, inode->i_mode);
1280 set_sd_v1_uid(sd_v1, inode->i_uid);
1281 set_sd_v1_gid(sd_v1, inode->i_gid);
1282 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1283 set_sd_v1_size(sd_v1, size);
1284 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1285 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1286 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
1288 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1289 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1291 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1293 // Sigh. i_first_direct_byte is back
1294 set_sd_v1_first_direct_byte(sd_v1,
1295 REISERFS_I(inode)->i_first_direct_byte);
1298 /* NOTE, you must prepare the buffer head before sending it here,
1299 ** and then log it after the call
1301 static void update_stat_data(struct treepath *path, struct inode *inode,
1304 struct buffer_head *bh;
1305 struct item_head *ih;
1307 bh = PATH_PLAST_BUFFER(path);
1308 ih = PATH_PITEM_HEAD(path);
1310 if (!is_statdata_le_ih(ih))
1311 reiserfs_panic(inode->i_sb,
1312 "vs-13065: update_stat_data: key %k, found item %h",
1313 INODE_PKEY(inode), ih);
1315 if (stat_data_v1(ih)) {
1316 // path points to old stat data
1317 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1319 inode2sd(B_I_PITEM(bh, ih), inode, size);
1325 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1326 struct inode *inode, loff_t size)
1329 INITIALIZE_PATH(path);
1330 struct buffer_head *bh;
1332 struct item_head *ih, tmp_ih;
1335 BUG_ON(!th->t_trans_id);
1337 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant
1341 /* look for the object's stat data */
1342 retval = search_item(inode->i_sb, &key, &path);
1343 if (retval == IO_ERROR) {
1344 reiserfs_warning(inode->i_sb,
1345 "vs-13050: reiserfs_update_sd: "
1346 "i/o failure occurred trying to update %K stat data",
1350 if (retval == ITEM_NOT_FOUND) {
1351 pos = PATH_LAST_POSITION(&path);
1353 if (inode->i_nlink == 0) {
1354 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1357 reiserfs_warning(inode->i_sb,
1358 "vs-13060: reiserfs_update_sd: "
1359 "stat data of object %k (nlink == %d) not found (pos %d)",
1360 INODE_PKEY(inode), inode->i_nlink,
1362 reiserfs_check_path(&path);
1366 /* sigh, prepare_for_journal might schedule. When it schedules the
1367 ** FS might change. We have to detect that, and loop back to the
1368 ** search if the stat data item has moved
1370 bh = get_last_bh(&path);
1372 copy_item_head(&tmp_ih, ih);
1373 fs_gen = get_generation(inode->i_sb);
1374 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1375 if (fs_changed(fs_gen, inode->i_sb)
1376 && item_moved(&tmp_ih, &path)) {
1377 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1378 continue; /* Stat_data item has been moved after scheduling. */
1382 update_stat_data(&path, inode, size);
1383 journal_mark_dirty(th, th->t_super, bh);
1388 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1389 ** does a make_bad_inode when things go wrong. But, we need to make sure
1390 ** and clear the key in the private portion of the inode, otherwise a
1391 ** corresponding iput might try to delete whatever object the inode last
1394 static void reiserfs_make_bad_inode(struct inode *inode)
1396 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1397 make_bad_inode(inode);
1401 // initially this function was derived from minix or ext2's analog and
1402 // evolved as the prototype did
1405 int reiserfs_init_locked_inode(struct inode *inode, void *p)
1407 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1408 inode->i_ino = args->objectid;
1409 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1413 /* looks for stat data in the tree, and fills up the fields of in-core
1414 inode stat data fields */
1415 void reiserfs_read_locked_inode(struct inode *inode,
1416 struct reiserfs_iget_args *args)
1418 INITIALIZE_PATH(path_to_sd);
1420 unsigned long dirino;
1423 dirino = args->dirid;
1425 /* set version 1, version 2 could be used too, because stat data
1426 key is the same in both versions */
1427 key.version = KEY_FORMAT_3_5;
1428 key.on_disk_key.k_dir_id = dirino;
1429 key.on_disk_key.k_objectid = inode->i_ino;
1430 key.on_disk_key.k_offset = 0;
1431 key.on_disk_key.k_type = 0;
1433 /* look for the object's stat data */
1434 retval = search_item(inode->i_sb, &key, &path_to_sd);
1435 if (retval == IO_ERROR) {
1436 reiserfs_warning(inode->i_sb,
1437 "vs-13070: reiserfs_read_locked_inode: "
1438 "i/o failure occurred trying to find stat data of %K",
1440 reiserfs_make_bad_inode(inode);
1443 if (retval != ITEM_FOUND) {
1444 /* a stale NFS handle can trigger this without it being an error */
1445 pathrelse(&path_to_sd);
1446 reiserfs_make_bad_inode(inode);
1451 init_inode(inode, &path_to_sd);
1453 /* It is possible that knfsd is trying to access inode of a file
1454 that is being removed from the disk by some other thread. As we
1455 update sd on unlink all that is required is to check for nlink
1456 here. This bug was first found by Sizif when debugging
1457 SquidNG/Butterfly, forgotten, and found again after Philippe
1458 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1460 More logical fix would require changes in fs/inode.c:iput() to
1461 remove inode from hash-table _after_ fs cleaned disk stuff up and
1462 in iget() to return NULL if I_FREEING inode is found in
1464 /* Currently there is one place where it's ok to meet inode with
1465 nlink==0: processing of open-unlinked and half-truncated files
1466 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1467 if ((inode->i_nlink == 0) &&
1468 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
1469 reiserfs_warning(inode->i_sb,
1470 "vs-13075: reiserfs_read_locked_inode: "
1471 "dead inode read from disk %K. "
1472 "This is likely to be race with knfsd. Ignore",
1474 reiserfs_make_bad_inode(inode);
1477 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
1482 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1484 * @inode: inode from hash table to check
1485 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1487 * This function is called by iget5_locked() to distinguish reiserfs inodes
1488 * having the same inode numbers. Such inodes can only exist due to some
1489 * error condition. One of them should be bad. Inodes with identical
1490 * inode numbers (objectids) are distinguished by parent directory ids.
1493 int reiserfs_find_actor(struct inode *inode, void *opaque)
1495 struct reiserfs_iget_args *args;
1498 /* args is already in CPU order */
1499 return (inode->i_ino == args->objectid) &&
1500 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1503 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
1505 struct inode *inode;
1506 struct reiserfs_iget_args args;
1508 args.objectid = key->on_disk_key.k_objectid;
1509 args.dirid = key->on_disk_key.k_dir_id;
1510 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1511 reiserfs_find_actor, reiserfs_init_locked_inode,
1514 return ERR_PTR(-ENOMEM);
1516 if (inode->i_state & I_NEW) {
1517 reiserfs_read_locked_inode(inode, &args);
1518 unlock_new_inode(inode);
1521 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1522 /* either due to i/o error or a stale NFS handle */
1529 struct dentry *reiserfs_get_dentry(struct super_block *sb, void *vobjp)
1531 __u32 *data = vobjp;
1533 struct dentry *result;
1534 struct inode *inode;
1536 key.on_disk_key.k_objectid = data[0];
1537 key.on_disk_key.k_dir_id = data[1];
1538 reiserfs_write_lock(sb);
1539 inode = reiserfs_iget(sb, &key);
1540 if (inode && !IS_ERR(inode) && data[2] != 0 &&
1541 data[2] != inode->i_generation) {
1545 reiserfs_write_unlock(sb);
1547 inode = ERR_PTR(-ESTALE);
1549 return ERR_PTR(PTR_ERR(inode));
1550 result = d_alloc_anon(inode);
1553 return ERR_PTR(-ENOMEM);
1558 struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 * data,
1559 int len, int fhtype,
1560 int (*acceptable) (void *contect,
1561 struct dentry * de),
1564 __u32 obj[3], parent[3];
1566 /* fhtype happens to reflect the number of u32s encoded.
1567 * due to a bug in earlier code, fhtype might indicate there
1568 * are more u32s then actually fitted.
1569 * so if fhtype seems to be more than len, reduce fhtype.
1571 * 2 - objectid + dir_id - legacy support
1572 * 3 - objectid + dir_id + generation
1573 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1574 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1575 * 6 - as above plus generation of directory
1576 * 6 does not fit in NFSv2 handles
1579 if (fhtype != 6 || len != 5)
1580 reiserfs_warning(sb,
1581 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1588 if (fhtype == 3 || fhtype >= 5)
1591 obj[2] = 0; /* generation number */
1594 parent[0] = data[fhtype >= 5 ? 3 : 2];
1595 parent[1] = data[fhtype >= 5 ? 4 : 3];
1597 parent[2] = data[5];
1601 return sb->s_export_op->find_exported_dentry(sb, obj,
1602 fhtype < 4 ? NULL : parent,
1603 acceptable, context);
1606 int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1609 struct inode *inode = dentry->d_inode;
1615 data[0] = inode->i_ino;
1616 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1617 data[2] = inode->i_generation;
1619 /* no room for directory info? return what we've stored so far */
1620 if (maxlen < 5 || !need_parent)
1623 spin_lock(&dentry->d_lock);
1624 inode = dentry->d_parent->d_inode;
1625 data[3] = inode->i_ino;
1626 data[4] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1629 data[5] = inode->i_generation;
1632 spin_unlock(&dentry->d_lock);
1636 /* looks for stat data, then copies fields to it, marks the buffer
1637 containing stat data as dirty */
1638 /* reiserfs inodes are never really dirty, since the dirty inode call
1639 ** always logs them. This call allows the VFS inode marking routines
1640 ** to properly mark inodes for datasync and such, but only actually
1641 ** does something when called for a synchronous update.
1643 int reiserfs_write_inode(struct inode *inode, int do_sync)
1645 struct reiserfs_transaction_handle th;
1646 int jbegin_count = 1;
1648 if (inode->i_sb->s_flags & MS_RDONLY)
1650 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1651 ** these cases are just when the system needs ram, not when the
1652 ** inode needs to reach disk for safety, and they can safely be
1653 ** ignored because the altered inode has already been logged.
1655 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1656 reiserfs_write_lock(inode->i_sb);
1657 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1658 reiserfs_update_sd(&th, inode);
1659 journal_end_sync(&th, inode->i_sb, jbegin_count);
1661 reiserfs_write_unlock(inode->i_sb);
1666 /* stat data of new object is inserted already, this inserts the item
1667 containing "." and ".." entries */
1668 static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1669 struct inode *inode,
1670 struct item_head *ih, struct treepath *path,
1673 struct super_block *sb = th->t_super;
1674 char empty_dir[EMPTY_DIR_SIZE];
1675 char *body = empty_dir;
1679 BUG_ON(!th->t_trans_id);
1681 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1682 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1683 TYPE_DIRENTRY, 3 /*key length */ );
1685 /* compose item head for new item. Directories consist of items of
1686 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1687 is done by reiserfs_new_inode */
1688 if (old_format_only(sb)) {
1689 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1690 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1692 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1693 ih->ih_key.k_objectid,
1694 INODE_PKEY(dir)->k_dir_id,
1695 INODE_PKEY(dir)->k_objectid);
1697 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1698 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1700 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1701 ih->ih_key.k_objectid,
1702 INODE_PKEY(dir)->k_dir_id,
1703 INODE_PKEY(dir)->k_objectid);
1706 /* look for place in the tree for new item */
1707 retval = search_item(sb, &key, path);
1708 if (retval == IO_ERROR) {
1709 reiserfs_warning(sb, "vs-13080: reiserfs_new_directory: "
1710 "i/o failure occurred creating new directory");
1713 if (retval == ITEM_FOUND) {
1715 reiserfs_warning(sb, "vs-13070: reiserfs_new_directory: "
1716 "object with this key exists (%k)",
1721 /* insert item, that is empty directory item */
1722 return reiserfs_insert_item(th, path, &key, ih, inode, body);
1725 /* stat data of object has been inserted, this inserts the item
1726 containing the body of symlink */
1727 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1728 struct item_head *ih,
1729 struct treepath *path, const char *symname,
1732 struct super_block *sb = th->t_super;
1736 BUG_ON(!th->t_trans_id);
1738 _make_cpu_key(&key, KEY_FORMAT_3_5,
1739 le32_to_cpu(ih->ih_key.k_dir_id),
1740 le32_to_cpu(ih->ih_key.k_objectid),
1741 1, TYPE_DIRECT, 3 /*key length */ );
1743 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1744 0 /*free_space */ );
1746 /* look for place in the tree for new item */
1747 retval = search_item(sb, &key, path);
1748 if (retval == IO_ERROR) {
1749 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlinik: "
1750 "i/o failure occurred creating new symlink");
1753 if (retval == ITEM_FOUND) {
1755 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlink: "
1756 "object with this key exists (%k)",
1761 /* insert item, that is body of symlink */
1762 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
1765 /* inserts the stat data into the tree, and then calls
1766 reiserfs_new_directory (to insert ".", ".." item if new object is
1767 directory) or reiserfs_new_symlink (to insert symlink body if new
1768 object is symlink) or nothing (if new object is regular file)
1770 NOTE! uid and gid must already be set in the inode. If we return
1771 non-zero due to an error, we have to drop the quota previously allocated
1772 for the fresh inode. This can only be done outside a transaction, so
1773 if we return non-zero, we also end the transaction. */
1774 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1775 struct inode *dir, int mode, const char *symname,
1776 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1777 strlen (symname) for symlinks) */
1778 loff_t i_size, struct dentry *dentry,
1779 struct inode *inode)
1781 struct super_block *sb;
1782 INITIALIZE_PATH(path_to_key);
1784 struct item_head ih;
1785 struct stat_data sd;
1789 BUG_ON(!th->t_trans_id);
1791 if (DLIMIT_ALLOC_INODE(inode)) {
1793 goto out_bad_dlimit;
1795 if (DQUOT_ALLOC_INODE(inode)) {
1799 if (!dir->i_nlink) {
1806 /* item head of new item */
1807 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1808 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1809 if (!ih.ih_key.k_objectid) {
1813 if (old_format_only(sb))
1814 /* not a perfect generation count, as object ids can be reused, but
1815 ** this is as good as reiserfs can do right now.
1816 ** note that the private part of inode isn't filled in yet, we have
1817 ** to use the directory.
1819 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1821 #if defined( USE_INODE_GENERATION_COUNTER )
1822 inode->i_generation =
1823 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1825 inode->i_generation = ++event;
1828 /* fill stat data */
1829 inode->i_nlink = (S_ISDIR(mode) ? 2 : 1);
1831 /* uid and gid must already be set by the caller for quota init */
1833 /* symlink cannot be immutable or append only, right? */
1834 if (S_ISLNK(inode->i_mode))
1835 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1837 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1838 inode->i_size = i_size;
1839 inode->i_blocks = 0;
1841 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1842 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1844 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1845 REISERFS_I(inode)->i_flags = 0;
1846 REISERFS_I(inode)->i_prealloc_block = 0;
1847 REISERFS_I(inode)->i_prealloc_count = 0;
1848 REISERFS_I(inode)->i_trans_id = 0;
1849 REISERFS_I(inode)->i_jl = NULL;
1850 REISERFS_I(inode)->i_attrs =
1851 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1852 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
1853 mutex_init(&(REISERFS_I(inode)->i_mmap));
1854 reiserfs_init_acl_access(inode);
1855 reiserfs_init_acl_default(inode);
1856 reiserfs_init_xattr_rwsem(inode);
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 #ifdef CONFIG_REISERFS_FS_POSIX_ACL
1998 if (REISERFS_I(dir)->i_acl_default && !IS_ERR(REISERFS_I(dir)->i_acl_default)) {
1999 reiserfs_write_unlock_xattrs(dir->i_sb);
2001 reiserfs_write_lock_xattrs(dir->i_sb);
2009 ** finds the tail page in the page cache,
2010 ** reads the last block in.
2012 ** On success, page_result is set to a locked, pinned page, and bh_result
2013 ** is set to an up to date buffer for the last block in the file. returns 0.
2015 ** tail conversion is not done, so bh_result might not be valid for writing
2016 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2017 ** trying to write the block.
2019 ** on failure, nonzero is returned, page_result and bh_result are untouched.
2021 static int grab_tail_page(struct inode *p_s_inode,
2022 struct page **page_result,
2023 struct buffer_head **bh_result)
2026 /* we want the page with the last byte in the file,
2027 ** not the page that will hold the next byte for appending
2029 unsigned long index = (p_s_inode->i_size - 1) >> PAGE_CACHE_SHIFT;
2030 unsigned long pos = 0;
2031 unsigned long start = 0;
2032 unsigned long blocksize = p_s_inode->i_sb->s_blocksize;
2033 unsigned long offset = (p_s_inode->i_size) & (PAGE_CACHE_SIZE - 1);
2034 struct buffer_head *bh;
2035 struct buffer_head *head;
2039 /* we know that we are only called with inode->i_size > 0.
2040 ** we also know that a file tail can never be as big as a block
2041 ** If i_size % blocksize == 0, our file is currently block aligned
2042 ** and it won't need converting or zeroing after a truncate.
2044 if ((offset & (blocksize - 1)) == 0) {
2047 page = grab_cache_page(p_s_inode->i_mapping, index);
2052 /* start within the page of the last block in the file */
2053 start = (offset / blocksize) * blocksize;
2055 error = block_prepare_write(page, start, offset,
2056 reiserfs_get_block_create_0);
2060 head = page_buffers(page);
2066 bh = bh->b_this_page;
2068 } while (bh != head);
2070 if (!buffer_uptodate(bh)) {
2071 /* note, this should never happen, prepare_write should
2072 ** be taking care of this for us. If the buffer isn't up to date,
2073 ** I've screwed up the code to find the buffer, or the code to
2074 ** call prepare_write
2076 reiserfs_warning(p_s_inode->i_sb,
2077 "clm-6000: error reading block %lu on dev %s",
2079 reiserfs_bdevname(p_s_inode->i_sb));
2084 *page_result = page;
2091 page_cache_release(page);
2096 ** vfs version of truncate file. Must NOT be called with
2097 ** a transaction already started.
2099 ** some code taken from block_truncate_page
2101 int reiserfs_truncate_file(struct inode *p_s_inode, int update_timestamps)
2103 struct reiserfs_transaction_handle th;
2104 /* we want the offset for the first byte after the end of the file */
2105 unsigned long offset = p_s_inode->i_size & (PAGE_CACHE_SIZE - 1);
2106 unsigned blocksize = p_s_inode->i_sb->s_blocksize;
2108 struct page *page = NULL;
2110 struct buffer_head *bh = NULL;
2113 reiserfs_write_lock(p_s_inode->i_sb);
2115 if (p_s_inode->i_size > 0) {
2116 if ((error = grab_tail_page(p_s_inode, &page, &bh))) {
2117 // -ENOENT means we truncated past the end of the file,
2118 // and get_block_create_0 could not find a block to read in,
2120 if (error != -ENOENT)
2121 reiserfs_warning(p_s_inode->i_sb,
2122 "clm-6001: grab_tail_page failed %d",
2129 /* so, if page != NULL, we have a buffer head for the offset at
2130 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2131 ** then we have an unformatted node. Otherwise, we have a direct item,
2132 ** and no zeroing is required on disk. We zero after the truncate,
2133 ** because the truncate might pack the item anyway
2134 ** (it will unmap bh if it packs).
2136 /* it is enough to reserve space in transaction for 2 balancings:
2137 one for "save" link adding and another for the first
2138 cut_from_item. 1 is for update_sd */
2139 error = journal_begin(&th, p_s_inode->i_sb,
2140 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2143 reiserfs_update_inode_transaction(p_s_inode);
2144 if (update_timestamps)
2145 /* we are doing real truncate: if the system crashes before the last
2146 transaction of truncating gets committed - on reboot the file
2147 either appears truncated properly or not truncated at all */
2148 add_save_link(&th, p_s_inode, 1);
2149 err2 = reiserfs_do_truncate(&th, p_s_inode, page, update_timestamps);
2151 journal_end(&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2155 /* check reiserfs_do_truncate after ending the transaction */
2161 if (update_timestamps) {
2162 error = remove_save_link(p_s_inode, 1 /* truncate */ );
2168 length = offset & (blocksize - 1);
2169 /* if we are not on a block boundary */
2173 length = blocksize - length;
2174 kaddr = kmap_atomic(page, KM_USER0);
2175 memset(kaddr + offset, 0, length);
2176 flush_dcache_page(page);
2177 kunmap_atomic(kaddr, KM_USER0);
2178 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2179 mark_buffer_dirty(bh);
2183 page_cache_release(page);
2186 reiserfs_write_unlock(p_s_inode->i_sb);
2191 page_cache_release(page);
2193 reiserfs_write_unlock(p_s_inode->i_sb);
2197 static int map_block_for_writepage(struct inode *inode,
2198 struct buffer_head *bh_result,
2199 unsigned long block)
2201 struct reiserfs_transaction_handle th;
2203 struct item_head tmp_ih;
2204 struct item_head *ih;
2205 struct buffer_head *bh;
2208 INITIALIZE_PATH(path);
2210 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
2211 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
2213 int use_get_block = 0;
2214 int bytes_copied = 0;
2216 int trans_running = 0;
2218 /* catch places below that try to log something without starting a trans */
2221 if (!buffer_uptodate(bh_result)) {
2225 kmap(bh_result->b_page);
2227 reiserfs_write_lock(inode->i_sb);
2228 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
2231 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2232 if (retval != POSITION_FOUND) {
2237 bh = get_last_bh(&path);
2239 item = get_item(&path);
2240 pos_in_item = path.pos_in_item;
2242 /* we've found an unformatted node */
2243 if (indirect_item_found(retval, ih)) {
2244 if (bytes_copied > 0) {
2245 reiserfs_warning(inode->i_sb,
2246 "clm-6002: bytes_copied %d",
2249 if (!get_block_num(item, pos_in_item)) {
2250 /* crap, we are writing to a hole */
2254 set_block_dev_mapped(bh_result,
2255 get_block_num(item, pos_in_item), inode);
2256 } else if (is_direct_le_ih(ih)) {
2258 p = page_address(bh_result->b_page);
2259 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2260 copy_size = ih_item_len(ih) - pos_in_item;
2262 fs_gen = get_generation(inode->i_sb);
2263 copy_item_head(&tmp_ih, ih);
2265 if (!trans_running) {
2266 /* vs-3050 is gone, no need to drop the path */
2267 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2270 reiserfs_update_inode_transaction(inode);
2272 if (fs_changed(fs_gen, inode->i_sb)
2273 && item_moved(&tmp_ih, &path)) {
2274 reiserfs_restore_prepared_buffer(inode->i_sb,
2280 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2282 if (fs_changed(fs_gen, inode->i_sb)
2283 && item_moved(&tmp_ih, &path)) {
2284 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2288 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2291 journal_mark_dirty(&th, inode->i_sb, bh);
2292 bytes_copied += copy_size;
2293 set_block_dev_mapped(bh_result, 0, inode);
2295 /* are there still bytes left? */
2296 if (bytes_copied < bh_result->b_size &&
2297 (byte_offset + bytes_copied) < inode->i_size) {
2298 set_cpu_key_k_offset(&key,
2299 cpu_key_k_offset(&key) +
2304 reiserfs_warning(inode->i_sb,
2305 "clm-6003: bad item inode %lu, device %s",
2306 inode->i_ino, reiserfs_bdevname(inode->i_sb));
2314 if (trans_running) {
2315 int err = journal_end(&th, inode->i_sb, jbegin_count);
2320 reiserfs_write_unlock(inode->i_sb);
2322 /* this is where we fill in holes in the file. */
2323 if (use_get_block) {
2324 retval = reiserfs_get_block(inode, block, bh_result,
2325 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2326 | GET_BLOCK_NO_DANGLE);
2328 if (!buffer_mapped(bh_result)
2329 || bh_result->b_blocknr == 0) {
2330 /* get_block failed to find a mapped unformatted node. */
2336 kunmap(bh_result->b_page);
2338 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2339 /* we've copied data from the page into the direct item, so the
2340 * buffer in the page is now clean, mark it to reflect that.
2342 lock_buffer(bh_result);
2343 clear_buffer_dirty(bh_result);
2344 unlock_buffer(bh_result);
2350 * mason@suse.com: updated in 2.5.54 to follow the same general io
2351 * start/recovery path as __block_write_full_page, along with special
2352 * code to handle reiserfs tails.
2354 static int reiserfs_write_full_page(struct page *page,
2355 struct writeback_control *wbc)
2357 struct inode *inode = page->mapping->host;
2358 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2360 unsigned long block;
2361 sector_t last_block;
2362 struct buffer_head *head, *bh;
2365 int checked = PageChecked(page);
2366 struct reiserfs_transaction_handle th;
2367 struct super_block *s = inode->i_sb;
2368 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2371 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2372 if (checked && (current->flags & PF_MEMALLOC)) {
2373 redirty_page_for_writepage(wbc, page);
2378 /* The page dirty bit is cleared before writepage is called, which
2379 * means we have to tell create_empty_buffers to make dirty buffers
2380 * The page really should be up to date at this point, so tossing
2381 * in the BH_Uptodate is just a sanity check.
2383 if (!page_has_buffers(page)) {
2384 create_empty_buffers(page, s->s_blocksize,
2385 (1 << BH_Dirty) | (1 << BH_Uptodate));
2387 head = page_buffers(page);
2389 /* last page in the file, zero out any contents past the
2390 ** last byte in the file
2392 if (page->index >= end_index) {
2394 unsigned last_offset;
2396 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2397 /* no file contents in this page */
2398 if (page->index >= end_index + 1 || !last_offset) {
2402 kaddr = kmap_atomic(page, KM_USER0);
2403 memset(kaddr + last_offset, 0, PAGE_CACHE_SIZE - last_offset);
2404 flush_dcache_page(page);
2405 kunmap_atomic(kaddr, KM_USER0);
2408 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
2409 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
2410 /* first map all the buffers, logging any direct items we find */
2412 if (block > last_block) {
2414 * This can happen when the block size is less than
2415 * the page size. The corresponding bytes in the page
2416 * were zero filled above
2418 clear_buffer_dirty(bh);
2419 set_buffer_uptodate(bh);
2420 } else if ((checked || buffer_dirty(bh)) &&
2421 (!buffer_mapped(bh) || (buffer_mapped(bh)
2424 /* not mapped yet, or it points to a direct item, search
2425 * the btree for the mapping info, and log any direct
2428 if ((error = map_block_for_writepage(inode, bh, block))) {
2432 bh = bh->b_this_page;
2434 } while (bh != head);
2437 * we start the transaction after map_block_for_writepage,
2438 * because it can create holes in the file (an unbounded operation).
2439 * starting it here, we can make a reliable estimate for how many
2440 * blocks we're going to log
2443 ClearPageChecked(page);
2444 reiserfs_write_lock(s);
2445 error = journal_begin(&th, s, bh_per_page + 1);
2447 reiserfs_write_unlock(s);
2450 reiserfs_update_inode_transaction(inode);
2452 /* now go through and lock any dirty buffers on the page */
2455 if (!buffer_mapped(bh))
2457 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2461 reiserfs_prepare_for_journal(s, bh, 1);
2462 journal_mark_dirty(&th, s, bh);
2465 /* from this point on, we know the buffer is mapped to a
2466 * real block and not a direct item
2468 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2471 if (test_set_buffer_locked(bh)) {
2472 redirty_page_for_writepage(wbc, page);
2476 if (test_clear_buffer_dirty(bh)) {
2477 mark_buffer_async_write(bh);
2481 } while ((bh = bh->b_this_page) != head);
2484 error = journal_end(&th, s, bh_per_page + 1);
2485 reiserfs_write_unlock(s);
2489 BUG_ON(PageWriteback(page));
2490 set_page_writeback(page);
2494 * since any buffer might be the only dirty buffer on the page,
2495 * the first submit_bh can bring the page out of writeback.
2496 * be careful with the buffers.
2499 struct buffer_head *next = bh->b_this_page;
2500 if (buffer_async_write(bh)) {
2501 submit_bh(WRITE, bh);
2506 } while (bh != head);
2512 * if this page only had a direct item, it is very possible for
2513 * no io to be required without there being an error. Or,
2514 * someone else could have locked them and sent them down the
2515 * pipe without locking the page
2519 if (!buffer_uptodate(bh)) {
2523 bh = bh->b_this_page;
2524 } while (bh != head);
2526 SetPageUptodate(page);
2527 end_page_writeback(page);
2532 /* catches various errors, we need to make sure any valid dirty blocks
2533 * get to the media. The page is currently locked and not marked for
2536 ClearPageUptodate(page);
2540 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2542 mark_buffer_async_write(bh);
2545 * clear any dirty bits that might have come from getting
2546 * attached to a dirty page
2548 clear_buffer_dirty(bh);
2550 bh = bh->b_this_page;
2551 } while (bh != head);
2553 BUG_ON(PageWriteback(page));
2554 set_page_writeback(page);
2557 struct buffer_head *next = bh->b_this_page;
2558 if (buffer_async_write(bh)) {
2559 clear_buffer_dirty(bh);
2560 submit_bh(WRITE, bh);
2565 } while (bh != head);
2569 static int reiserfs_readpage(struct file *f, struct page *page)
2571 return block_read_full_page(page, reiserfs_get_block);
2574 static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
2576 struct inode *inode = page->mapping->host;
2577 reiserfs_wait_on_write_block(inode->i_sb);
2578 return reiserfs_write_full_page(page, wbc);
2581 static int reiserfs_prepare_write(struct file *f, struct page *page,
2582 unsigned from, unsigned to)
2584 struct inode *inode = page->mapping->host;
2588 reiserfs_wait_on_write_block(inode->i_sb);
2589 fix_tail_page_for_writing(page);
2590 if (reiserfs_transaction_running(inode->i_sb)) {
2591 struct reiserfs_transaction_handle *th;
2592 th = (struct reiserfs_transaction_handle *)current->
2594 BUG_ON(!th->t_refcount);
2595 BUG_ON(!th->t_trans_id);
2596 old_ref = th->t_refcount;
2600 ret = block_prepare_write(page, from, to, reiserfs_get_block);
2601 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2602 struct reiserfs_transaction_handle *th = current->journal_info;
2603 /* this gets a little ugly. If reiserfs_get_block returned an
2604 * error and left a transacstion running, we've got to close it,
2605 * and we've got to free handle if it was a persistent transaction.
2607 * But, if we had nested into an existing transaction, we need
2608 * to just drop the ref count on the handle.
2610 * If old_ref == 0, the transaction is from reiserfs_get_block,
2611 * and it was a persistent trans. Otherwise, it was nested above.
2613 if (th->t_refcount > old_ref) {
2618 reiserfs_write_lock(inode->i_sb);
2619 err = reiserfs_end_persistent_transaction(th);
2620 reiserfs_write_unlock(inode->i_sb);
2630 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2632 return generic_block_bmap(as, block, reiserfs_bmap);
2635 static int reiserfs_commit_write(struct file *f, struct page *page,
2636 unsigned from, unsigned to)
2638 struct inode *inode = page->mapping->host;
2639 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2642 struct reiserfs_transaction_handle *th = NULL;
2644 reiserfs_wait_on_write_block(inode->i_sb);
2645 if (reiserfs_transaction_running(inode->i_sb)) {
2646 th = current->journal_info;
2648 reiserfs_commit_page(inode, page, from, to);
2650 /* generic_commit_write does this for us, but does not update the
2651 ** transaction tracking stuff when the size changes. So, we have
2652 ** to do the i_size updates here.
2654 if (pos > inode->i_size) {
2655 struct reiserfs_transaction_handle myth;
2656 reiserfs_write_lock(inode->i_sb);
2657 /* If the file have grown beyond the border where it
2658 can have a tail, unmark it as needing a tail
2660 if ((have_large_tails(inode->i_sb)
2661 && inode->i_size > i_block_size(inode) * 4)
2662 || (have_small_tails(inode->i_sb)
2663 && inode->i_size > i_block_size(inode)))
2664 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2666 ret = journal_begin(&myth, inode->i_sb, 1);
2668 reiserfs_write_unlock(inode->i_sb);
2671 reiserfs_update_inode_transaction(inode);
2672 inode->i_size = pos;
2674 * this will just nest into our transaction. It's important
2675 * to use mark_inode_dirty so the inode gets pushed around on the
2676 * dirty lists, and so that O_SYNC works as expected
2678 mark_inode_dirty(inode);
2679 reiserfs_update_sd(&myth, inode);
2681 ret = journal_end(&myth, inode->i_sb, 1);
2682 reiserfs_write_unlock(inode->i_sb);
2687 reiserfs_write_lock(inode->i_sb);
2689 mark_inode_dirty(inode);
2690 ret = reiserfs_end_persistent_transaction(th);
2691 reiserfs_write_unlock(inode->i_sb);
2701 reiserfs_write_lock(inode->i_sb);
2703 reiserfs_update_sd(th, inode);
2704 ret = reiserfs_end_persistent_transaction(th);
2705 reiserfs_write_unlock(inode->i_sb);
2711 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2713 if (reiserfs_attrs(inode->i_sb)) {
2714 if (sd_attrs & REISERFS_SYNC_FL)
2715 inode->i_flags |= S_SYNC;
2717 inode->i_flags &= ~S_SYNC;
2718 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2719 inode->i_flags |= S_IMMUTABLE;
2721 inode->i_flags &= ~S_IMMUTABLE;
2722 if (sd_attrs & REISERFS_IUNLINK_FL)
2723 inode->i_flags |= S_IUNLINK;
2725 inode->i_flags &= ~S_IUNLINK;
2726 if (sd_attrs & REISERFS_BARRIER_FL)
2727 inode->i_flags |= S_BARRIER;
2729 inode->i_flags &= ~S_BARRIER;
2730 if (sd_attrs & REISERFS_APPEND_FL)
2731 inode->i_flags |= S_APPEND;
2733 inode->i_flags &= ~S_APPEND;
2734 if (sd_attrs & REISERFS_NOATIME_FL)
2735 inode->i_flags |= S_NOATIME;
2737 inode->i_flags &= ~S_NOATIME;
2738 if (sd_attrs & REISERFS_NOTAIL_FL)
2739 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2741 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2745 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
2747 if (reiserfs_attrs(inode->i_sb)) {
2748 if (inode->i_flags & S_IMMUTABLE)
2749 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2751 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2752 if (inode->i_flags & S_IUNLINK)
2753 *sd_attrs |= REISERFS_IUNLINK_FL;
2755 *sd_attrs &= ~REISERFS_IUNLINK_FL;
2756 if (inode->i_flags & S_BARRIER)
2757 *sd_attrs |= REISERFS_BARRIER_FL;
2759 *sd_attrs &= ~REISERFS_BARRIER_FL;
2760 if (inode->i_flags & S_SYNC)
2761 *sd_attrs |= REISERFS_SYNC_FL;
2763 *sd_attrs &= ~REISERFS_SYNC_FL;
2764 if (inode->i_flags & S_NOATIME)
2765 *sd_attrs |= REISERFS_NOATIME_FL;
2767 *sd_attrs &= ~REISERFS_NOATIME_FL;
2768 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
2769 *sd_attrs |= REISERFS_NOTAIL_FL;
2771 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2775 /* decide if this buffer needs to stay around for data logging or ordered
2778 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2781 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2784 spin_lock(&j->j_dirty_buffers_lock);
2785 if (!buffer_mapped(bh)) {
2788 /* the page is locked, and the only places that log a data buffer
2789 * also lock the page.
2791 if (reiserfs_file_data_log(inode)) {
2793 * very conservative, leave the buffer pinned if
2794 * anyone might need it.
2796 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2799 } else if (buffer_dirty(bh)) {
2800 struct reiserfs_journal_list *jl;
2801 struct reiserfs_jh *jh = bh->b_private;
2803 /* why is this safe?
2804 * reiserfs_setattr updates i_size in the on disk
2805 * stat data before allowing vmtruncate to be called.
2807 * If buffer was put onto the ordered list for this
2808 * transaction, we know for sure either this transaction
2809 * or an older one already has updated i_size on disk,
2810 * and this ordered data won't be referenced in the file
2813 * if the buffer was put onto the ordered list for an older
2814 * transaction, we need to leave it around
2816 if (jh && (jl = jh->jl)
2817 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2821 if (ret && bh->b_private) {
2822 reiserfs_free_jh(bh);
2824 spin_unlock(&j->j_dirty_buffers_lock);
2829 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2830 static void reiserfs_invalidatepage(struct page *page, unsigned long offset)
2832 struct buffer_head *head, *bh, *next;
2833 struct inode *inode = page->mapping->host;
2834 unsigned int curr_off = 0;
2837 BUG_ON(!PageLocked(page));
2840 ClearPageChecked(page);
2842 if (!page_has_buffers(page))
2845 head = page_buffers(page);
2848 unsigned int next_off = curr_off + bh->b_size;
2849 next = bh->b_this_page;
2852 * is this block fully invalidated?
2854 if (offset <= curr_off) {
2855 if (invalidatepage_can_drop(inode, bh))
2856 reiserfs_unmap_buffer(bh);
2860 curr_off = next_off;
2862 } while (bh != head);
2865 * We release buffers only if the entire page is being invalidated.
2866 * The get_block cached value has been unconditionally invalidated,
2867 * so real IO is not possible anymore.
2869 if (!offset && ret) {
2870 ret = try_to_release_page(page, 0);
2871 /* maybe should BUG_ON(!ret); - neilb */
2877 static int reiserfs_set_page_dirty(struct page *page)
2879 struct inode *inode = page->mapping->host;
2880 if (reiserfs_file_data_log(inode)) {
2881 SetPageChecked(page);
2882 return __set_page_dirty_nobuffers(page);
2884 return __set_page_dirty_buffers(page);
2888 * Returns 1 if the page's buffers were dropped. The page is locked.
2890 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
2891 * in the buffers at page_buffers(page).
2893 * even in -o notail mode, we can't be sure an old mount without -o notail
2894 * didn't create files with tails.
2896 static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
2898 struct inode *inode = page->mapping->host;
2899 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
2900 struct buffer_head *head;
2901 struct buffer_head *bh;
2904 WARN_ON(PageChecked(page));
2905 spin_lock(&j->j_dirty_buffers_lock);
2906 head = page_buffers(page);
2909 if (bh->b_private) {
2910 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
2911 reiserfs_free_jh(bh);
2917 bh = bh->b_this_page;
2918 } while (bh != head);
2920 ret = try_to_free_buffers(page);
2921 spin_unlock(&j->j_dirty_buffers_lock);
2925 /* We thank Mingming Cao for helping us understand in great detail what
2926 to do in this section of the code. */
2927 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
2928 const struct iovec *iov, loff_t offset,
2929 unsigned long nr_segs)
2931 struct file *file = iocb->ki_filp;
2932 struct inode *inode = file->f_mapping->host;
2934 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2936 reiserfs_get_blocks_direct_io, NULL);
2939 int reiserfs_sync_flags(struct inode *inode)
2941 u16 oldflags, newflags;
2943 oldflags = REISERFS_I(inode)->i_attrs;
2944 newflags = oldflags;
2945 i_attrs_to_sd_attrs(inode, &newflags);
2947 if (oldflags ^ newflags) {
2948 REISERFS_I(inode)->i_attrs = newflags;
2949 inode->i_ctime = CURRENT_TIME_SEC;
2950 mark_inode_dirty(inode);
2955 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
2957 struct inode *inode = dentry->d_inode;
2959 unsigned int ia_valid = attr->ia_valid;
2960 reiserfs_write_lock(inode->i_sb);
2961 if (attr->ia_valid & ATTR_SIZE) {
2962 /* version 2 items will be caught by the s_maxbytes check
2963 ** done for us in vmtruncate
2965 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
2966 attr->ia_size > MAX_NON_LFS) {
2970 /* fill in hole pointers in the expanding truncate case. */
2971 if (attr->ia_size > inode->i_size) {
2972 error = generic_cont_expand(inode, attr->ia_size);
2973 if (REISERFS_I(inode)->i_prealloc_count > 0) {
2975 struct reiserfs_transaction_handle th;
2976 /* we're changing at most 2 bitmaps, inode + super */
2977 err = journal_begin(&th, inode->i_sb, 4);
2979 reiserfs_discard_prealloc(&th, inode);
2980 err = journal_end(&th, inode->i_sb, 4);
2988 * file size is changed, ctime and mtime are
2991 attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
2995 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
2996 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
2997 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
2998 /* stat data of format v3.5 has 16 bit uid and gid */
3003 error = inode_change_ok(inode, attr);
3006 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
3007 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid) ||
3008 (ia_valid & ATTR_TAG && attr->ia_tag != inode->i_tag)) {
3009 error = reiserfs_chown_xattrs(inode, attr);
3012 struct reiserfs_transaction_handle th;
3015 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3016 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3019 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3021 journal_begin(&th, inode->i_sb,
3026 DQUOT_TRANSFER(inode, attr) ? -EDQUOT : 0;
3028 journal_end(&th, inode->i_sb,
3032 /* Update corresponding info in inode so that everything is in
3033 * one transaction */
3034 if (attr->ia_valid & ATTR_UID)
3035 inode->i_uid = attr->ia_uid;
3036 if (attr->ia_valid & ATTR_GID)
3037 inode->i_gid = attr->ia_gid;
3038 if ((attr->ia_valid & ATTR_TAG) &&
3040 inode->i_tag = attr->ia_tag;
3041 mark_inode_dirty(inode);
3043 journal_end(&th, inode->i_sb, jbegin_count);
3047 error = inode_setattr(inode, attr);
3050 if (!error && reiserfs_posixacl(inode->i_sb)) {
3051 if (attr->ia_valid & ATTR_MODE)
3052 error = reiserfs_acl_chmod(inode);
3056 reiserfs_write_unlock(inode->i_sb);
3060 const struct address_space_operations reiserfs_address_space_operations = {
3061 .writepage = reiserfs_writepage,
3062 .readpage = reiserfs_readpage,
3063 .readpages = reiserfs_readpages,
3064 .releasepage = reiserfs_releasepage,
3065 .invalidatepage = reiserfs_invalidatepage,
3066 .sync_page = block_sync_page,
3067 .prepare_write = reiserfs_prepare_write,
3068 .commit_write = reiserfs_commit_write,
3069 .bmap = reiserfs_aop_bmap,
3070 .direct_IO = reiserfs_direct_IO,
3071 .set_page_dirty = reiserfs_set_page_dirty,