2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
5 #include <linux/config.h>
6 #include <linux/time.h>
8 #include <linux/reiserfs_fs.h>
9 #include <linux/reiserfs_acl.h>
10 #include <linux/reiserfs_xattr.h>
11 #include <linux/smp_lock.h>
12 #include <linux/pagemap.h>
13 #include <linux/highmem.h>
14 #include <asm/uaccess.h>
15 #include <asm/unaligned.h>
16 #include <linux/buffer_head.h>
17 #include <linux/mpage.h>
18 #include <linux/writeback.h>
19 #include <linux/quotaops.h>
20 #include <linux/vserver/xid.h>
22 extern int reiserfs_default_io_size; /* default io size devuned in super.c */
24 static int reiserfs_commit_write(struct file *f, struct page *page,
25 unsigned from, unsigned to);
26 static int reiserfs_prepare_write(struct file *f, struct page *page,
27 unsigned from, unsigned to);
29 void reiserfs_delete_inode (struct inode * inode)
31 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
32 int jbegin_count = JOURNAL_PER_BALANCE_CNT * 2 + 2 * REISERFS_QUOTA_INIT_BLOCKS;
33 struct reiserfs_transaction_handle th ;
35 reiserfs_write_lock(inode->i_sb);
37 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
38 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
41 reiserfs_delete_xattrs (inode);
43 if (journal_begin(&th, inode->i_sb, jbegin_count)) {
47 reiserfs_update_inode_transaction(inode) ;
49 if (reiserfs_delete_object (&th, inode)) {
54 /* Do quota update inside a transaction for journaled quotas. We must do that
55 * after delete_object so that quota updates go into the same transaction as
56 * stat data deletion */
57 DQUOT_FREE_INODE(inode);
59 if (journal_end(&th, inode->i_sb, jbegin_count)) {
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, __u32 objectid,
80 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;
92 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
93 offset and type of key */
94 void make_cpu_key (struct cpu_key * key, struct inode * inode, loff_t offset,
95 int type, int length )
97 _make_cpu_key (key, get_inode_item_key_version (inode), le32_to_cpu (INODE_PKEY (inode)->k_dir_id),
98 le32_to_cpu (INODE_PKEY (inode)->k_objectid),
99 offset, type, length);
104 // when key is 0, do not set version and short key
106 inline void make_le_item_head (struct item_head * ih, const struct cpu_key * key,
108 loff_t offset, int type, int length,
109 int entry_count/*or ih_free_space*/)
112 ih->ih_key.k_dir_id = cpu_to_le32 (key->on_disk_key.k_dir_id);
113 ih->ih_key.k_objectid = 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) {
158 struct buffer_head *head, *next, *bh ;
160 if (page && page_has_buffers(page)) {
161 head = page_buffers(page) ;
164 next = bh->b_this_page ;
165 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
166 reiserfs_unmap_buffer(bh) ;
169 } while (bh != head) ;
173 /* reiserfs_get_block does not need to allocate a block only if it has been
174 done already or non-hole position has been found in the indirect item */
175 static inline int allocation_needed (int retval, b_blocknr_t allocated,
176 struct item_head * ih,
177 __u32 * item, int pos_in_item)
181 if (retval == POSITION_FOUND && is_indirect_le_ih (ih) &&
182 get_block_num(item, pos_in_item))
187 static inline int indirect_item_found (int retval, struct item_head * ih)
189 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);
201 // files which were created in the earlier version can not be longer,
204 static int file_capable (struct inode * inode, long block)
206 if (get_inode_item_key_version (inode) != KEY_FORMAT_3_5 || // it is new file.
207 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
213 /*static*/ int restart_transaction(struct reiserfs_transaction_handle *th,
214 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,
249 INITIALIZE_PATH (path);
251 struct buffer_head * bh;
252 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, 3);
266 if (search_for_position_by_key (inode->i_sb, &key, &path) != POSITION_FOUND) {
269 kunmap(bh_result->b_page) ;
270 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
271 // That there is some MMAPED data associated with it that is yet to be written to disk.
272 if ((args & GET_BLOCK_NO_HOLE) && !PageUptodate(bh_result->b_page) ) {
279 bh = get_last_bh (&path);
281 if (is_indirect_le_ih (ih)) {
282 __u32 * ind_item = (__u32 *)B_I_PITEM (bh, ih);
284 /* FIXME: here we could cache indirect item or part of it in
285 the inode to avoid search_by_key in case of subsequent
287 blocknr = get_block_num(ind_item, path.pos_in_item) ;
290 map_bh(bh_result, inode->i_sb, blocknr);
291 if (path.pos_in_item == ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
292 set_buffer_boundary(bh_result);
295 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
296 // That there is some MMAPED data associated with it that is yet to be written to disk.
297 if ((args & GET_BLOCK_NO_HOLE) && !PageUptodate(bh_result->b_page) ) {
303 kunmap(bh_result->b_page) ;
307 // requested data are in direct item(s)
308 if (!(args & GET_BLOCK_READ_DIRECT)) {
309 // we are called by bmap. FIXME: we can not map block of file
310 // when it is stored in direct item(s)
313 kunmap(bh_result->b_page) ;
317 /* if we've got a direct item, and the buffer or page was uptodate,
318 ** we don't want to pull data off disk again. skip to the
319 ** end, where we map the buffer and return
321 if (buffer_uptodate(bh_result)) {
325 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
326 ** pages without any buffers. If the page is up to date, we don't want
327 ** read old data off disk. Set the up to date bit on the buffer instead
328 ** and jump to the end
330 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
331 set_buffer_uptodate(bh_result);
335 // read file tail into part of page
336 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1) ;
337 fs_gen = get_generation(inode->i_sb) ;
338 copy_item_head (&tmp_ih, ih);
340 /* we only want to kmap if we are reading the tail into the page.
341 ** this is not the common case, so we don't kmap until we are
342 ** sure we need to. But, this means the item might move if
346 p = (char *)kmap(bh_result->b_page) ;
347 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
352 memset (p, 0, inode->i_sb->s_blocksize);
354 if (!is_direct_le_ih (ih)) {
357 /* make sure we don't read more bytes than actually exist in
358 ** the file. This can happen in odd cases where i_size isn't
359 ** correct, and when direct item padding results in a few
360 ** extra bytes at the end of the direct item
362 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
364 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
365 chars = inode->i_size - (le_ih_k_offset(ih) - 1) - path.pos_in_item;
368 chars = ih_item_len(ih) - path.pos_in_item;
370 memcpy (p, B_I_PITEM (bh, ih) + path.pos_in_item, chars);
377 if (PATH_LAST_POSITION (&path) != (B_NR_ITEMS (bh) - 1))
378 // we done, if read direct item is not the last item of
379 // node FIXME: we could try to check right delimiting key
380 // to see whether direct item continues in the right
381 // neighbor or rely on i_size
384 // update key to look for the next piece
385 set_cpu_key_k_offset (&key, cpu_key_k_offset (&key) + chars);
386 if (search_for_position_by_key (inode->i_sb, &key, &path) != POSITION_FOUND)
387 // we read something from tail, even if now we got IO_ERROR
389 bh = get_last_bh (&path);
393 flush_dcache_page(bh_result->b_page) ;
394 kunmap(bh_result->b_page) ;
398 /* this buffer has valid data, but isn't valid for io. mapping it to
399 * block #0 tells the rest of reiserfs it just has a tail in it
401 map_bh(bh_result, inode->i_sb, 0);
402 set_buffer_uptodate (bh_result);
407 // this is called to create file map. So, _get_block_create_0 will not
409 static int reiserfs_bmap (struct inode * inode, sector_t block,
410 struct buffer_head * bh_result, int create)
412 if (!file_capable (inode, block))
415 reiserfs_write_lock(inode->i_sb);
416 /* do not read the direct item */
417 _get_block_create_0 (inode, block, bh_result, 0) ;
418 reiserfs_write_unlock(inode->i_sb);
422 /* special version of get_block that is only used by grab_tail_page right
423 ** now. It is sent to block_prepare_write, and when you try to get a
424 ** block past the end of the file (or a block from a hole) it returns
425 ** -ENOENT instead of a valid buffer. block_prepare_write expects to
426 ** be able to do i/o on the buffers returned, unless an error value
429 ** So, this allows block_prepare_write to be used for reading a single block
430 ** in a page. Where it does not produce a valid page for holes, or past the
431 ** end of the file. This turns out to be exactly what we need for reading
432 ** tails for conversion.
434 ** The point of the wrapper is forcing a certain value for create, even
435 ** though the VFS layer is calling this function with create==1. If you
436 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
437 ** don't use this function.
439 static int reiserfs_get_block_create_0 (struct inode * inode, sector_t block,
440 struct buffer_head * bh_result, int create) {
441 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE) ;
444 /* This is special helper for reiserfs_get_block in case we are executing
445 direct_IO request. */
446 static int reiserfs_get_blocks_direct_io(struct inode *inode,
448 unsigned long max_blocks,
449 struct buffer_head *bh_result,
454 bh_result->b_page = NULL;
456 /* We set the b_size before reiserfs_get_block call since it is
457 referenced in convert_tail_for_hole() that may be called from
458 reiserfs_get_block() */
459 bh_result->b_size = (1 << inode->i_blkbits);
461 ret = reiserfs_get_block(inode, iblock, bh_result,
462 create | GET_BLOCK_NO_DANGLE) ;
466 /* don't allow direct io onto tail pages */
467 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
468 /* make sure future calls to the direct io funcs for this offset
469 ** in the file fail by unmapping the buffer
471 clear_buffer_mapped(bh_result);
474 /* Possible unpacked tail. Flush the data before pages have
476 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
479 err = reiserfs_commit_for_inode(inode);
480 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
491 ** helper function for when reiserfs_get_block is called for a hole
492 ** but the file tail is still in a direct item
493 ** bh_result is the buffer head for the hole
494 ** tail_offset is the offset of the start of the tail in the file
496 ** This calls prepare_write, which will start a new transaction
497 ** you should not be in a transaction, or have any paths held when you
500 static int convert_tail_for_hole(struct inode *inode,
501 struct buffer_head *bh_result,
502 loff_t tail_offset) {
503 unsigned long index ;
504 unsigned long tail_end ;
505 unsigned long tail_start ;
506 struct page * tail_page ;
507 struct page * hole_page = bh_result->b_page ;
510 if ((tail_offset & (bh_result->b_size - 1)) != 1)
513 /* always try to read until the end of the block */
514 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1) ;
515 tail_end = (tail_start | (bh_result->b_size - 1)) + 1 ;
517 index = tail_offset >> PAGE_CACHE_SHIFT ;
518 /* hole_page can be zero in case of direct_io, we are sure
519 that we cannot get here if we write with O_DIRECT into
521 if (!hole_page || index != hole_page->index) {
522 tail_page = grab_cache_page(inode->i_mapping, index) ;
528 tail_page = hole_page ;
531 /* we don't have to make sure the conversion did not happen while
532 ** we were locking the page because anyone that could convert
533 ** must first take i_sem.
535 ** We must fix the tail page for writing because it might have buffers
536 ** that are mapped, but have a block number of 0. This indicates tail
537 ** data that has been read directly into the page, and block_prepare_write
538 ** won't trigger a get_block in this case.
540 fix_tail_page_for_writing(tail_page) ;
541 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
545 /* tail conversion might change the data in the page */
546 flush_dcache_page(tail_page) ;
548 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end) ;
551 if (tail_page != hole_page) {
552 unlock_page(tail_page) ;
553 page_cache_release(tail_page) ;
559 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
562 b_blocknr_t *allocated_block_nr,
565 BUG_ON (!th->t_trans_id);
567 #ifdef REISERFS_PREALLOCATE
568 if (!(flags & GET_BLOCK_NO_ISEM)) {
569 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr, path, block);
572 return reiserfs_new_unf_blocknrs (th, inode, allocated_block_nr, path, block);
575 int reiserfs_get_block (struct inode * inode, sector_t block,
576 struct buffer_head * bh_result, int create)
578 int repeat, retval = 0;
579 b_blocknr_t allocated_block_nr = 0;// b_blocknr_t is (unsigned) 32 bit int
580 INITIALIZE_PATH(path);
583 struct buffer_head * bh, * unbh = NULL;
584 struct item_head * ih, tmp_ih;
588 struct reiserfs_transaction_handle *th = NULL;
589 /* space reserved in transaction batch:
590 . 3 balancings in direct->indirect conversion
591 . 1 block involved into reiserfs_update_sd()
592 XXX in practically impossible worst case direct2indirect()
593 can incur (much) more than 3 balancings.
594 quota update for user, group */
595 int jbegin_count = JOURNAL_PER_BALANCE_CNT * 3 + 1 + 2 * REISERFS_QUOTA_TRANS_BLOCKS;
598 loff_t new_offset = (((loff_t)block) << inode->i_sb->s_blocksize_bits) + 1 ;
601 reiserfs_write_lock(inode->i_sb);
602 version = get_inode_item_key_version (inode);
605 reiserfs_write_unlock(inode->i_sb);
609 if (!file_capable (inode, block)) {
610 reiserfs_write_unlock(inode->i_sb);
614 /* if !create, we aren't changing the FS, so we don't need to
615 ** log anything, so we don't need to start a transaction
617 if (!(create & GET_BLOCK_CREATE)) {
619 /* find number of block-th logical block of the file */
620 ret = _get_block_create_0 (inode, block, bh_result,
621 create | GET_BLOCK_READ_DIRECT) ;
622 reiserfs_write_unlock(inode->i_sb);
626 * if we're already in a transaction, make sure to close
627 * any new transactions we start in this func
629 if ((create & GET_BLOCK_NO_DANGLE) ||
630 reiserfs_transaction_running(inode->i_sb))
633 /* If file is of such a size, that it might have a tail and tails are enabled
634 ** we should mark it as possibly needing tail packing on close
636 if ( (have_large_tails (inode->i_sb) && inode->i_size < i_block_size (inode)*4) ||
637 (have_small_tails (inode->i_sb) && inode->i_size < i_block_size(inode)) )
638 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask ;
640 /* set the key of the first byte in the 'block'-th block of file */
641 make_cpu_key (&key, inode, new_offset,
642 TYPE_ANY, 3/*key length*/);
643 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
645 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
650 reiserfs_update_inode_transaction(inode) ;
654 retval = search_for_position_by_key (inode->i_sb, &key, &path);
655 if (retval == IO_ERROR) {
660 bh = get_last_bh (&path);
662 item = get_item (&path);
663 pos_in_item = path.pos_in_item;
665 fs_gen = get_generation (inode->i_sb);
666 copy_item_head (&tmp_ih, ih);
668 if (allocation_needed (retval, allocated_block_nr, ih, item, pos_in_item)) {
669 /* we have to allocate block for the unformatted node */
675 repeat = _allocate_block(th, block, inode, &allocated_block_nr, &path, create);
677 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
678 /* restart the transaction to give the journal a chance to free
679 ** some blocks. releases the path, so we have to go back to
680 ** research if we succeed on the second try
682 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
683 retval = restart_transaction(th, inode, &path) ;
686 repeat = _allocate_block(th, block, inode, &allocated_block_nr, NULL, create);
688 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
691 if (repeat == QUOTA_EXCEEDED)
698 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
703 if (indirect_item_found (retval, ih)) {
704 b_blocknr_t unfm_ptr;
705 /* 'block'-th block is in the file already (there is
706 corresponding cell in some indirect item). But it may be
707 zero unformatted node pointer (hole) */
708 unfm_ptr = get_block_num (item, pos_in_item);
710 /* use allocated block to plug the hole */
711 reiserfs_prepare_for_journal(inode->i_sb, bh, 1) ;
712 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
713 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
716 set_buffer_new(bh_result);
717 if (buffer_dirty(bh_result) && reiserfs_data_ordered(inode->i_sb))
718 reiserfs_add_ordered_list(inode, bh_result);
719 put_block_num(item, pos_in_item, allocated_block_nr) ;
720 unfm_ptr = allocated_block_nr;
721 journal_mark_dirty (th, inode->i_sb, bh);
722 reiserfs_update_sd(th, inode) ;
724 set_block_dev_mapped(bh_result, unfm_ptr, inode);
728 retval = reiserfs_end_persistent_transaction(th);
730 reiserfs_write_unlock(inode->i_sb);
732 /* the item was found, so new blocks were not added to the file
733 ** there is no need to make sure the inode is updated with this
744 /* desired position is not found or is in the direct item. We have
745 to append file with holes up to 'block'-th block converting
746 direct items to indirect one if necessary */
749 if (is_statdata_le_ih (ih)) {
751 struct cpu_key tmp_key;
753 /* indirect item has to be inserted */
754 make_le_item_head (&tmp_ih, &key, version, 1, TYPE_INDIRECT,
755 UNFM_P_SIZE, 0/* free_space */);
757 if (cpu_key_k_offset (&key) == 1) {
758 /* we are going to add 'block'-th block to the file. Use
759 allocated block for that */
760 unp = cpu_to_le32 (allocated_block_nr);
761 set_block_dev_mapped (bh_result, allocated_block_nr, inode);
762 set_buffer_new(bh_result);
766 set_cpu_key_k_offset (&tmp_key, 1);
767 PATH_LAST_POSITION(&path) ++;
769 retval = reiserfs_insert_item (th, &path, &tmp_key, &tmp_ih, inode, (char *)&unp);
771 reiserfs_free_block (th, inode, allocated_block_nr, 1);
772 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
774 //mark_tail_converted (inode);
775 } else if (is_direct_le_ih (ih)) {
776 /* direct item has to be converted */
779 tail_offset = ((le_ih_k_offset (ih) - 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
780 if (tail_offset == cpu_key_k_offset (&key)) {
781 /* direct item we just found fits into block we have
782 to map. Convert it into unformatted node: use
783 bh_result for the conversion */
784 set_block_dev_mapped (bh_result, allocated_block_nr, inode);
788 /* we have to padd file tail stored in direct item(s)
789 up to block size and convert it to unformatted
790 node. FIXME: this should also get into page cache */
794 * ugly, but we can only end the transaction if
797 BUG_ON (!th->t_refcount);
798 if (th->t_refcount == 1) {
799 retval = reiserfs_end_persistent_transaction(th);
805 retval = convert_tail_for_hole(inode, bh_result, tail_offset) ;
807 if ( retval != -ENOSPC )
808 reiserfs_warning (inode->i_sb, "clm-6004: convert tail failed inode %lu, error %d", inode->i_ino, retval) ;
809 if (allocated_block_nr) {
810 /* the bitmap, the super, and the stat data == 3 */
812 th = reiserfs_persistent_transaction(inode->i_sb,3);
814 reiserfs_free_block (th,inode,allocated_block_nr,1);
820 retval = direct2indirect (th, inode, &path, unbh, tail_offset);
822 reiserfs_unmap_buffer(unbh);
823 reiserfs_free_block (th, inode, allocated_block_nr, 1);
826 /* it is important the set_buffer_uptodate is done after
827 ** the direct2indirect. The buffer might contain valid
828 ** data newer than the data on disk (read by readpage, changed,
829 ** and then sent here by writepage). direct2indirect needs
830 ** to know if unbh was already up to date, so it can decide
831 ** if the data in unbh needs to be replaced with data from
834 set_buffer_uptodate (unbh);
836 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
837 buffer will disappear shortly, so it should not be added to
839 if ( unbh->b_page ) {
840 /* we've converted the tail, so we must
841 ** flush unbh before the transaction commits
843 reiserfs_add_tail_list(inode, unbh) ;
845 /* mark it dirty now to prevent commit_write from adding
846 ** this buffer to the inode's dirty buffer list
849 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
850 * It's still atomic, but it sets the page dirty too,
851 * which makes it eligible for writeback at any time by the
852 * VM (which was also the case with __mark_buffer_dirty())
854 mark_buffer_dirty(unbh) ;
857 /* append indirect item with holes if needed, when appending
858 pointer to 'block'-th block use block, which is already
860 struct cpu_key tmp_key;
861 unp_t unf_single=0; // We use this in case we need to allocate only
862 // one block which is a fastpath
864 __u64 max_to_insert=MAX_ITEM_LEN(inode->i_sb->s_blocksize)/UNFM_P_SIZE;
867 RFALSE( pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
868 "vs-804: invalid position for append");
869 /* indirect item has to be appended, set up key of that position */
870 make_cpu_key (&tmp_key, inode,
871 le_key_k_offset (version, &(ih->ih_key)) + op_bytes_number (ih, inode->i_sb->s_blocksize),
872 //pos_in_item * inode->i_sb->s_blocksize,
873 TYPE_INDIRECT, 3);// key type is unimportant
875 blocks_needed = 1 + ((cpu_key_k_offset (&key) - cpu_key_k_offset (&tmp_key)) >> inode->i_sb->s_blocksize_bits);
876 RFALSE( blocks_needed < 0, "green-805: invalid offset");
878 if ( blocks_needed == 1 ) {
881 un=kmalloc( min(blocks_needed,max_to_insert)*UNFM_P_SIZE,
882 GFP_ATOMIC); // We need to avoid scheduling.
888 memset(un, 0, UNFM_P_SIZE * min(blocks_needed,max_to_insert));
890 if ( blocks_needed <= max_to_insert) {
891 /* we are going to add target block to the file. Use allocated
893 un[blocks_needed-1] = cpu_to_le32 (allocated_block_nr);
894 set_block_dev_mapped (bh_result, allocated_block_nr, inode);
895 set_buffer_new(bh_result);
898 /* paste hole to the indirect item */
899 /* If kmalloc failed, max_to_insert becomes zero and it means we
900 only have space for one block */
901 blocks_needed=max_to_insert?max_to_insert:1;
903 retval = reiserfs_paste_into_item (th, &path, &tmp_key, inode, (char *)un, UNFM_P_SIZE * blocks_needed);
905 if (blocks_needed != 1)
909 reiserfs_free_block (th, inode, allocated_block_nr, 1);
913 /* We need to mark new file size in case this function will be
914 interrupted/aborted later on. And we may do this only for
916 inode->i_size += inode->i_sb->s_blocksize * blocks_needed;
923 /* this loop could log more blocks than we had originally asked
924 ** for. So, we have to allow the transaction to end if it is
925 ** too big or too full. Update the inode so things are
926 ** consistent if we crash before the function returns
928 ** release the path so that anybody waiting on the path before
929 ** ending their transaction will be able to continue.
931 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
932 retval = restart_transaction(th, inode, &path) ;
936 /* inserting indirect pointers for a hole can take a
937 ** long time. reschedule if needed
941 retval = search_for_position_by_key (inode->i_sb, &key, &path);
942 if (retval == IO_ERROR) {
946 if (retval == POSITION_FOUND) {
947 reiserfs_warning (inode->i_sb, "vs-825: reiserfs_get_block: "
948 "%K should not be found", &key);
950 if (allocated_block_nr)
951 reiserfs_free_block (th, inode, allocated_block_nr, 1);
955 bh = get_last_bh (&path);
957 item = get_item (&path);
958 pos_in_item = path.pos_in_item;
965 if (th && (!dangle || (retval && !th->t_trans_id))) {
968 reiserfs_update_sd(th, inode);
969 err = reiserfs_end_persistent_transaction(th);
974 reiserfs_write_unlock(inode->i_sb);
975 reiserfs_check_path(&path) ;
980 reiserfs_readpages(struct file *file, struct address_space *mapping,
981 struct list_head *pages, unsigned nr_pages)
983 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
986 /* Compute real number of used bytes by file
987 * Following three functions can go away when we'll have enough space in stat item
989 static int real_space_diff(struct inode *inode, int sd_size)
992 loff_t blocksize = inode->i_sb->s_blocksize ;
994 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
997 /* End of file is also in full block with indirect reference, so round
998 ** up to the next block.
1000 ** there is just no way to know if the tail is actually packed
1001 ** on the file, so we have to assume it isn't. When we pack the
1002 ** tail, we add 4 bytes to pretend there really is an unformatted
1005 bytes = ((inode->i_size + (blocksize-1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE + sd_size;
1009 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
1012 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1013 return inode->i_size + (loff_t)(real_space_diff(inode, sd_size)) ;
1015 return ((loff_t)real_space_diff(inode, sd_size)) + (((loff_t)blocks) << 9);
1018 /* Compute number of blocks used by file in ReiserFS counting */
1019 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1021 loff_t bytes = inode_get_bytes(inode) ;
1022 loff_t real_space = real_space_diff(inode, sd_size) ;
1024 /* keeps fsck and non-quota versions of reiserfs happy */
1025 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1026 bytes += (loff_t)511 ;
1029 /* files from before the quota patch might i_blocks such that
1030 ** bytes < real_space. Deal with that here to prevent it from
1033 if (bytes < real_space)
1035 return (bytes - real_space) >> 9;
1039 // BAD: new directories have stat data of new type and all other items
1040 // of old type. Version stored in the inode says about body items, so
1041 // in update_stat_data we can not rely on inode, but have to check
1042 // item version directly
1045 // called by read_locked_inode
1046 static void init_inode (struct inode * inode, struct path * path)
1048 struct buffer_head * bh;
1049 struct item_head * ih;
1053 //int version = ITEM_VERSION_1;
1055 bh = PATH_PLAST_BUFFER (path);
1056 ih = PATH_PITEM_HEAD (path);
1059 copy_key (INODE_PKEY (inode), &(ih->ih_key));
1060 inode->i_blksize = reiserfs_default_io_size;
1062 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list ));
1063 REISERFS_I(inode)->i_flags = 0;
1064 REISERFS_I(inode)->i_prealloc_block = 0;
1065 REISERFS_I(inode)->i_prealloc_count = 0;
1066 REISERFS_I(inode)->i_trans_id = 0;
1067 REISERFS_I(inode)->i_jl = NULL;
1068 REISERFS_I(inode)->i_acl_access = NULL;
1069 REISERFS_I(inode)->i_acl_default = NULL;
1070 init_rwsem (&REISERFS_I(inode)->xattr_sem);
1072 if (stat_data_v1 (ih)) {
1073 struct stat_data_v1 * sd = (struct stat_data_v1 *)B_I_PITEM (bh, ih);
1074 unsigned long blocks;
1076 uid = sd_v1_uid(sd);
1077 gid = sd_v1_gid(sd);
1079 set_inode_item_key_version (inode, KEY_FORMAT_3_5);
1080 set_inode_sd_version (inode, STAT_DATA_V1);
1081 inode->i_mode = sd_v1_mode(sd);
1082 inode->i_nlink = sd_v1_nlink(sd);
1083 inode->i_size = sd_v1_size(sd);
1084 inode->i_atime.tv_sec = sd_v1_atime(sd);
1085 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1086 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1087 inode->i_atime.tv_nsec = 0;
1088 inode->i_ctime.tv_nsec = 0;
1089 inode->i_mtime.tv_nsec = 0;
1091 inode->i_blocks = sd_v1_blocks(sd);
1092 inode->i_generation = le32_to_cpu (INODE_PKEY (inode)->k_dir_id);
1093 blocks = (inode->i_size + 511) >> 9;
1094 blocks = _ROUND_UP (blocks, inode->i_sb->s_blocksize >> 9);
1095 if (inode->i_blocks > blocks) {
1096 // there was a bug in <=3.5.23 when i_blocks could take negative
1097 // values. Starting from 3.5.17 this value could even be stored in
1098 // stat data. For such files we set i_blocks based on file
1099 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1100 // only updated if file's inode will ever change
1101 inode->i_blocks = blocks;
1104 rdev = sd_v1_rdev(sd);
1105 REISERFS_I(inode)->i_first_direct_byte = sd_v1_first_direct_byte(sd);
1106 /* an early bug in the quota code can give us an odd number for the
1107 ** block count. This is incorrect, fix it here.
1109 if (inode->i_blocks & 1) {
1112 inode_set_bytes(inode, to_real_used_space(inode, inode->i_blocks,
1114 /* nopack is initially zero for v1 objects. For v2 objects,
1115 nopack is initialised from sd_attrs */
1116 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1118 // new stat data found, but object may have old items
1119 // (directories and symlinks)
1120 struct stat_data * sd = (struct stat_data *)B_I_PITEM (bh, ih);
1122 uid = sd_v2_uid(sd);
1123 gid = sd_v2_gid(sd);
1125 inode->i_mode = sd_v2_mode(sd);
1126 inode->i_nlink = sd_v2_nlink(sd);
1127 inode->i_size = sd_v2_size(sd);
1128 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1129 inode->i_atime.tv_sec = sd_v2_atime(sd);
1130 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1131 inode->i_ctime.tv_nsec = 0;
1132 inode->i_mtime.tv_nsec = 0;
1133 inode->i_atime.tv_nsec = 0;
1134 inode->i_blocks = sd_v2_blocks(sd);
1135 rdev = sd_v2_rdev(sd);
1136 if( S_ISCHR( inode -> i_mode ) || S_ISBLK( inode -> i_mode ) )
1137 inode->i_generation = le32_to_cpu (INODE_PKEY (inode)->k_dir_id);
1139 inode->i_generation = sd_v2_generation(sd);
1141 if (S_ISDIR (inode->i_mode) || S_ISLNK (inode->i_mode))
1142 set_inode_item_key_version (inode, KEY_FORMAT_3_5);
1144 set_inode_item_key_version (inode, KEY_FORMAT_3_6);
1145 REISERFS_I(inode)->i_first_direct_byte = 0;
1146 set_inode_sd_version (inode, STAT_DATA_V2);
1147 inode_set_bytes(inode, to_real_used_space(inode, inode->i_blocks,
1149 /* read persistent inode attributes from sd and initalise
1150 generic inode flags from them */
1151 REISERFS_I(inode)->i_attrs = sd_v2_attrs( sd );
1152 sd_attrs_to_i_attrs( sd_v2_attrs( sd ), inode );
1154 inode->i_uid = INOXID_UID(XID_TAG(inode), uid, gid);
1155 inode->i_gid = INOXID_GID(XID_TAG(inode), uid, gid);
1156 inode->i_xid = INOXID_XID(XID_TAG(inode), uid, gid, 0);
1159 if (S_ISREG (inode->i_mode)) {
1160 inode->i_op = &reiserfs_file_inode_operations;
1161 inode->i_fop = &reiserfs_file_operations;
1162 inode->i_mapping->a_ops = &reiserfs_address_space_operations ;
1163 } else if (S_ISDIR (inode->i_mode)) {
1164 inode->i_op = &reiserfs_dir_inode_operations;
1165 inode->i_fop = &reiserfs_dir_operations;
1166 } else if (S_ISLNK (inode->i_mode)) {
1167 inode->i_op = &reiserfs_symlink_inode_operations;
1168 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1170 inode->i_blocks = 0;
1171 inode->i_op = &reiserfs_special_inode_operations;
1172 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1177 // update new stat data with inode fields
1178 static void inode2sd (void * sd, struct inode * inode, loff_t size)
1180 struct stat_data * sd_v2 = (struct stat_data *)sd;
1181 uid_t uid = XIDINO_UID(XID_TAG(inode), inode->i_uid, inode->i_xid);
1182 gid_t gid = XIDINO_GID(XID_TAG(inode), inode->i_gid, inode->i_xid);
1185 set_sd_v2_uid(sd_v2, uid );
1186 set_sd_v2_gid(sd_v2, gid );
1187 set_sd_v2_mode(sd_v2, inode->i_mode );
1188 set_sd_v2_nlink(sd_v2, inode->i_nlink );
1189 set_sd_v2_size(sd_v2, size );
1190 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec );
1191 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec );
1192 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec );
1193 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1194 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1195 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1197 set_sd_v2_generation(sd_v2, inode->i_generation);
1198 flags = REISERFS_I(inode)->i_attrs;
1199 i_attrs_to_sd_attrs( inode, &flags );
1200 set_sd_v2_attrs( sd_v2, flags );
1204 // used to copy inode's fields to old stat data
1205 static void inode2sd_v1 (void * sd, struct inode * inode, loff_t size)
1207 struct stat_data_v1 * sd_v1 = (struct stat_data_v1 *)sd;
1209 set_sd_v1_mode(sd_v1, inode->i_mode );
1210 set_sd_v1_uid(sd_v1, inode->i_uid );
1211 set_sd_v1_gid(sd_v1, inode->i_gid );
1212 set_sd_v1_nlink(sd_v1, inode->i_nlink );
1213 set_sd_v1_size(sd_v1, size );
1214 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec );
1215 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec );
1216 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec );
1218 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1219 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1221 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1223 // Sigh. i_first_direct_byte is back
1224 set_sd_v1_first_direct_byte(sd_v1, REISERFS_I(inode)->i_first_direct_byte);
1228 /* NOTE, you must prepare the buffer head before sending it here,
1229 ** and then log it after the call
1231 static void update_stat_data (struct path * path, struct inode * inode,
1234 struct buffer_head * bh;
1235 struct item_head * ih;
1237 bh = PATH_PLAST_BUFFER (path);
1238 ih = PATH_PITEM_HEAD (path);
1240 if (!is_statdata_le_ih (ih))
1241 reiserfs_panic (inode->i_sb, "vs-13065: update_stat_data: key %k, found item %h",
1242 INODE_PKEY (inode), ih);
1244 if (stat_data_v1 (ih)) {
1245 // path points to old stat data
1246 inode2sd_v1 (B_I_PITEM (bh, ih), inode, size);
1248 inode2sd (B_I_PITEM (bh, ih), inode, size);
1255 void reiserfs_update_sd_size (struct reiserfs_transaction_handle *th,
1256 struct inode * inode, loff_t size)
1259 INITIALIZE_PATH(path);
1260 struct buffer_head *bh ;
1262 struct item_head *ih, tmp_ih ;
1265 BUG_ON (!th->t_trans_id);
1267 make_cpu_key (&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3);//key type is unimportant
1271 /* look for the object's stat data */
1272 retval = search_item (inode->i_sb, &key, &path);
1273 if (retval == IO_ERROR) {
1274 reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: "
1275 "i/o failure occurred trying to update %K stat data",
1279 if (retval == ITEM_NOT_FOUND) {
1280 pos = PATH_LAST_POSITION (&path);
1282 if (inode->i_nlink == 0) {
1283 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found");*/
1286 reiserfs_warning (inode->i_sb, "vs-13060: reiserfs_update_sd: "
1287 "stat data of object %k (nlink == %d) not found (pos %d)",
1288 INODE_PKEY (inode), inode->i_nlink, pos);
1289 reiserfs_check_path(&path) ;
1293 /* sigh, prepare_for_journal might schedule. When it schedules the
1294 ** FS might change. We have to detect that, and loop back to the
1295 ** search if the stat data item has moved
1297 bh = get_last_bh(&path) ;
1298 ih = get_ih(&path) ;
1299 copy_item_head (&tmp_ih, ih);
1300 fs_gen = get_generation (inode->i_sb);
1301 reiserfs_prepare_for_journal(inode->i_sb, bh, 1) ;
1302 if (fs_changed (fs_gen, inode->i_sb) && item_moved(&tmp_ih, &path)) {
1303 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
1304 continue ; /* Stat_data item has been moved after scheduling. */
1308 update_stat_data (&path, inode, size);
1309 journal_mark_dirty(th, th->t_super, bh) ;
1314 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1315 ** does a make_bad_inode when things go wrong. But, we need to make sure
1316 ** and clear the key in the private portion of the inode, otherwise a
1317 ** corresponding iput might try to delete whatever object the inode last
1320 static void reiserfs_make_bad_inode(struct inode *inode) {
1321 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1322 make_bad_inode(inode);
1326 // initially this function was derived from minix or ext2's analog and
1327 // evolved as the prototype did
1330 int reiserfs_init_locked_inode (struct inode * inode, void *p)
1332 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p ;
1333 inode->i_ino = args->objectid;
1334 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1338 /* looks for stat data in the tree, and fills up the fields of in-core
1339 inode stat data fields */
1340 void reiserfs_read_locked_inode (struct inode * inode, struct reiserfs_iget_args *args)
1342 INITIALIZE_PATH (path_to_sd);
1344 unsigned long dirino;
1347 dirino = args->dirid ;
1349 /* set version 1, version 2 could be used too, because stat data
1350 key is the same in both versions */
1351 key.version = KEY_FORMAT_3_5;
1352 key.on_disk_key.k_dir_id = dirino;
1353 key.on_disk_key.k_objectid = inode->i_ino;
1354 key.on_disk_key.u.k_offset_v1.k_offset = SD_OFFSET;
1355 key.on_disk_key.u.k_offset_v1.k_uniqueness = SD_UNIQUENESS;
1357 /* look for the object's stat data */
1358 retval = search_item (inode->i_sb, &key, &path_to_sd);
1359 if (retval == IO_ERROR) {
1360 reiserfs_warning (inode->i_sb, "vs-13070: reiserfs_read_locked_inode: "
1361 "i/o failure occurred trying to find stat data of %K",
1363 reiserfs_make_bad_inode(inode) ;
1366 if (retval != ITEM_FOUND) {
1367 /* a stale NFS handle can trigger this without it being an error */
1368 pathrelse (&path_to_sd);
1369 reiserfs_make_bad_inode(inode) ;
1374 init_inode (inode, &path_to_sd);
1376 /* It is possible that knfsd is trying to access inode of a file
1377 that is being removed from the disk by some other thread. As we
1378 update sd on unlink all that is required is to check for nlink
1379 here. This bug was first found by Sizif when debugging
1380 SquidNG/Butterfly, forgotten, and found again after Philippe
1381 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1383 More logical fix would require changes in fs/inode.c:iput() to
1384 remove inode from hash-table _after_ fs cleaned disk stuff up and
1385 in iget() to return NULL if I_FREEING inode is found in
1387 /* Currently there is one place where it's ok to meet inode with
1388 nlink==0: processing of open-unlinked and half-truncated files
1389 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1390 if( ( inode -> i_nlink == 0 ) &&
1391 ! REISERFS_SB(inode -> i_sb) -> s_is_unlinked_ok ) {
1392 reiserfs_warning (inode->i_sb,
1393 "vs-13075: reiserfs_read_locked_inode: "
1394 "dead inode read from disk %K. "
1395 "This is likely to be race with knfsd. Ignore",
1397 reiserfs_make_bad_inode( inode );
1400 reiserfs_check_path(&path_to_sd) ; /* init inode should be relsing */
1405 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1407 * @inode: inode from hash table to check
1408 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1410 * This function is called by iget5_locked() to distinguish reiserfs inodes
1411 * having the same inode numbers. Such inodes can only exist due to some
1412 * error condition. One of them should be bad. Inodes with identical
1413 * inode numbers (objectids) are distinguished by parent directory ids.
1416 int reiserfs_find_actor( struct inode *inode, void *opaque )
1418 struct reiserfs_iget_args *args;
1421 /* args is already in CPU order */
1422 return (inode->i_ino == args->objectid) &&
1423 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1426 struct inode * reiserfs_iget (struct super_block * s, const struct cpu_key * key)
1428 struct inode * inode;
1429 struct reiserfs_iget_args args ;
1431 args.objectid = key->on_disk_key.k_objectid ;
1432 args.dirid = key->on_disk_key.k_dir_id ;
1433 inode = iget5_locked (s, key->on_disk_key.k_objectid,
1434 reiserfs_find_actor, reiserfs_init_locked_inode, (void *)(&args));
1436 return ERR_PTR(-ENOMEM) ;
1438 if (inode->i_state & I_NEW) {
1439 reiserfs_read_locked_inode(inode, &args);
1440 unlock_new_inode(inode);
1443 if (comp_short_keys (INODE_PKEY (inode), key) || is_bad_inode (inode)) {
1444 /* either due to i/o error or a stale NFS handle */
1451 struct dentry *reiserfs_get_dentry(struct super_block *sb, void *vobjp)
1453 __u32 *data = vobjp;
1454 struct cpu_key key ;
1455 struct dentry *result;
1456 struct inode *inode;
1458 key.on_disk_key.k_objectid = data[0] ;
1459 key.on_disk_key.k_dir_id = data[1] ;
1460 reiserfs_write_lock(sb);
1461 inode = reiserfs_iget(sb, &key) ;
1462 if (inode && !IS_ERR(inode) && data[2] != 0 &&
1463 data[2] != inode->i_generation) {
1467 reiserfs_write_unlock(sb);
1469 inode = ERR_PTR(-ESTALE);
1471 return ERR_PTR(PTR_ERR(inode));
1472 result = d_alloc_anon(inode);
1475 return ERR_PTR(-ENOMEM);
1480 struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 *data,
1481 int len, int fhtype,
1482 int (*acceptable)(void *contect, struct dentry *de),
1484 __u32 obj[3], parent[3];
1486 /* fhtype happens to reflect the number of u32s encoded.
1487 * due to a bug in earlier code, fhtype might indicate there
1488 * are more u32s then actually fitted.
1489 * so if fhtype seems to be more than len, reduce fhtype.
1491 * 2 - objectid + dir_id - legacy support
1492 * 3 - objectid + dir_id + generation
1493 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1494 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1495 * 6 - as above plus generation of directory
1496 * 6 does not fit in NFSv2 handles
1499 if (fhtype != 6 || len != 5)
1500 reiserfs_warning (sb, "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1507 if (fhtype == 3 || fhtype >= 5)
1509 else obj[2] = 0; /* generation number */
1512 parent[0] = data[fhtype>=5?3:2] ;
1513 parent[1] = data[fhtype>=5?4:3] ;
1515 parent[2] = data[5];
1518 return sb->s_export_op->find_exported_dentry(sb, obj, fhtype < 4 ? NULL : parent,
1519 acceptable, context);
1522 int reiserfs_encode_fh(struct dentry *dentry, __u32 *data, int *lenp, int need_parent) {
1523 struct inode *inode = dentry->d_inode ;
1529 data[0] = inode->i_ino ;
1530 data[1] = le32_to_cpu(INODE_PKEY (inode)->k_dir_id) ;
1531 data[2] = inode->i_generation ;
1533 /* no room for directory info? return what we've stored so far */
1534 if (maxlen < 5 || ! need_parent)
1537 spin_lock(&dentry->d_lock);
1538 inode = dentry->d_parent->d_inode ;
1539 data[3] = inode->i_ino ;
1540 data[4] = le32_to_cpu(INODE_PKEY (inode)->k_dir_id) ;
1543 data[5] = inode->i_generation ;
1546 spin_unlock(&dentry->d_lock);
1551 /* looks for stat data, then copies fields to it, marks the buffer
1552 containing stat data as dirty */
1553 /* reiserfs inodes are never really dirty, since the dirty inode call
1554 ** always logs them. This call allows the VFS inode marking routines
1555 ** to properly mark inodes for datasync and such, but only actually
1556 ** does something when called for a synchronous update.
1558 int reiserfs_write_inode (struct inode * inode, int do_sync) {
1559 struct reiserfs_transaction_handle th ;
1560 int jbegin_count = 1 ;
1562 if (inode->i_sb->s_flags & MS_RDONLY)
1564 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1565 ** these cases are just when the system needs ram, not when the
1566 ** inode needs to reach disk for safety, and they can safely be
1567 ** ignored because the altered inode has already been logged.
1569 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1570 reiserfs_write_lock(inode->i_sb);
1571 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1572 reiserfs_update_sd (&th, inode);
1573 journal_end_sync(&th, inode->i_sb, jbegin_count) ;
1575 reiserfs_write_unlock(inode->i_sb);
1580 /* stat data of new object is inserted already, this inserts the item
1581 containing "." and ".." entries */
1582 static int reiserfs_new_directory (struct reiserfs_transaction_handle *th,
1583 struct inode *inode,
1584 struct item_head * ih, struct path * path,
1587 struct super_block * sb = th->t_super;
1588 char empty_dir [EMPTY_DIR_SIZE];
1589 char * body = empty_dir;
1593 BUG_ON (!th->t_trans_id);
1595 _make_cpu_key (&key, KEY_FORMAT_3_5, le32_to_cpu (ih->ih_key.k_dir_id),
1596 le32_to_cpu (ih->ih_key.k_objectid), DOT_OFFSET, TYPE_DIRENTRY, 3/*key length*/);
1598 /* compose item head for new item. Directories consist of items of
1599 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1600 is done by reiserfs_new_inode */
1601 if (old_format_only (sb)) {
1602 make_le_item_head (ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET, TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1604 make_empty_dir_item_v1 (body, ih->ih_key.k_dir_id, ih->ih_key.k_objectid,
1605 INODE_PKEY (dir)->k_dir_id,
1606 INODE_PKEY (dir)->k_objectid );
1608 make_le_item_head (ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET, TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1610 make_empty_dir_item (body, ih->ih_key.k_dir_id, ih->ih_key.k_objectid,
1611 INODE_PKEY (dir)->k_dir_id,
1612 INODE_PKEY (dir)->k_objectid );
1615 /* look for place in the tree for new item */
1616 retval = search_item (sb, &key, path);
1617 if (retval == IO_ERROR) {
1618 reiserfs_warning (sb, "vs-13080: reiserfs_new_directory: "
1619 "i/o failure occurred creating new directory");
1622 if (retval == ITEM_FOUND) {
1624 reiserfs_warning (sb, "vs-13070: reiserfs_new_directory: "
1625 "object with this key exists (%k)", &(ih->ih_key));
1629 /* insert item, that is empty directory item */
1630 return reiserfs_insert_item (th, path, &key, ih, inode, body);
1634 /* stat data of object has been inserted, this inserts the item
1635 containing the body of symlink */
1636 static int reiserfs_new_symlink (struct reiserfs_transaction_handle *th,
1637 struct inode *inode, /* Inode of symlink */
1638 struct item_head * ih,
1639 struct path * path, const char * symname, int item_len)
1641 struct super_block * sb = th->t_super;
1645 BUG_ON (!th->t_trans_id);
1647 _make_cpu_key (&key, KEY_FORMAT_3_5,
1648 le32_to_cpu (ih->ih_key.k_dir_id),
1649 le32_to_cpu (ih->ih_key.k_objectid),
1650 1, TYPE_DIRECT, 3/*key length*/);
1652 make_le_item_head (ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len, 0/*free_space*/);
1654 /* look for place in the tree for new item */
1655 retval = search_item (sb, &key, path);
1656 if (retval == IO_ERROR) {
1657 reiserfs_warning (sb, "vs-13080: reiserfs_new_symlinik: "
1658 "i/o failure occurred creating new symlink");
1661 if (retval == ITEM_FOUND) {
1663 reiserfs_warning (sb, "vs-13080: reiserfs_new_symlink: "
1664 "object with this key exists (%k)", &(ih->ih_key));
1668 /* insert item, that is body of symlink */
1669 return reiserfs_insert_item (th, path, &key, ih, inode, symname);
1673 /* inserts the stat data into the tree, and then calls
1674 reiserfs_new_directory (to insert ".", ".." item if new object is
1675 directory) or reiserfs_new_symlink (to insert symlink body if new
1676 object is symlink) or nothing (if new object is regular file)
1678 NOTE! uid and gid must already be set in the inode. If we return
1679 non-zero due to an error, we have to drop the quota previously allocated
1680 for the fresh inode. This can only be done outside a transaction, so
1681 if we return non-zero, we also end the transaction. */
1682 int reiserfs_new_inode (struct reiserfs_transaction_handle *th,
1683 struct inode * dir, int mode,
1684 const char * symname,
1685 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1686 strlen (symname) for symlinks)*/
1687 loff_t i_size, struct dentry *dentry,
1688 struct inode *inode)
1690 struct super_block * sb;
1691 INITIALIZE_PATH (path_to_key);
1693 struct item_head ih;
1694 struct stat_data sd;
1698 BUG_ON (!th->t_trans_id);
1700 if (DQUOT_ALLOC_INODE(inode)) {
1704 if (!dir || !dir->i_nlink) {
1711 /* item head of new item */
1712 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1713 ih.ih_key.k_objectid = cpu_to_le32 (reiserfs_get_unused_objectid (th));
1714 if (!ih.ih_key.k_objectid) {
1716 goto out_bad_inode ;
1718 if (old_format_only (sb))
1719 /* not a perfect generation count, as object ids can be reused, but
1720 ** this is as good as reiserfs can do right now.
1721 ** note that the private part of inode isn't filled in yet, we have
1722 ** to use the directory.
1724 inode->i_generation = le32_to_cpu (INODE_PKEY (dir)->k_objectid);
1726 #if defined( USE_INODE_GENERATION_COUNTER )
1727 inode->i_generation = le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1729 inode->i_generation = ++event;
1732 /* fill stat data */
1733 inode->i_nlink = (S_ISDIR (mode) ? 2 : 1);
1735 /* uid and gid must already be set by the caller for quota init */
1737 /* symlink cannot be immutable or append only, right? */
1738 if( S_ISLNK( inode -> i_mode ) )
1739 inode -> i_flags &= ~ ( S_IMMUTABLE | S_APPEND );
1741 inode->i_mtime = inode->i_atime = inode->i_ctime =
1743 inode->i_size = i_size;
1744 inode->i_blocks = 0;
1746 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1747 U32_MAX/*NO_BYTES_IN_DIRECT_ITEM*/;
1749 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list ));
1750 REISERFS_I(inode)->i_flags = 0;
1751 REISERFS_I(inode)->i_prealloc_block = 0;
1752 REISERFS_I(inode)->i_prealloc_count = 0;
1753 REISERFS_I(inode)->i_trans_id = 0;
1754 REISERFS_I(inode)->i_jl = NULL;
1755 REISERFS_I(inode)->i_attrs =
1756 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1757 sd_attrs_to_i_attrs( REISERFS_I(inode) -> i_attrs, inode );
1758 REISERFS_I(inode)->i_acl_access = NULL;
1759 REISERFS_I(inode)->i_acl_default = NULL;
1760 init_rwsem (&REISERFS_I(inode)->xattr_sem);
1762 if (old_format_only (sb))
1763 make_le_item_head (&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET, TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1765 make_le_item_head (&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET, TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1767 /* key to search for correct place for new stat data */
1768 _make_cpu_key (&key, KEY_FORMAT_3_6, le32_to_cpu (ih.ih_key.k_dir_id),
1769 le32_to_cpu (ih.ih_key.k_objectid), SD_OFFSET, TYPE_STAT_DATA, 3/*key length*/);
1771 /* find proper place for inserting of stat data */
1772 retval = search_item (sb, &key, &path_to_key);
1773 if (retval == IO_ERROR) {
1777 if (retval == ITEM_FOUND) {
1778 pathrelse (&path_to_key);
1782 if (old_format_only (sb)) {
1783 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1784 pathrelse (&path_to_key);
1785 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1789 inode2sd_v1 (&sd, inode, inode->i_size);
1791 inode2sd (&sd, inode, inode->i_size);
1793 // these do not go to on-disk stat data
1794 inode->i_ino = le32_to_cpu (ih.ih_key.k_objectid);
1795 inode->i_blksize = reiserfs_default_io_size;
1797 // store in in-core inode the key of stat data and version all
1798 // object items will have (directory items will have old offset
1799 // format, other new objects will consist of new items)
1800 memcpy (INODE_PKEY (inode), &(ih.ih_key), KEY_SIZE);
1801 if (old_format_only (sb) || S_ISDIR(mode) || S_ISLNK(mode))
1802 set_inode_item_key_version (inode, KEY_FORMAT_3_5);
1804 set_inode_item_key_version (inode, KEY_FORMAT_3_6);
1805 if (old_format_only (sb))
1806 set_inode_sd_version (inode, STAT_DATA_V1);
1808 set_inode_sd_version (inode, STAT_DATA_V2);
1810 /* insert the stat data into the tree */
1811 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1812 if (REISERFS_I(dir)->new_packing_locality)
1813 th->displace_new_blocks = 1;
1815 retval = reiserfs_insert_item (th, &path_to_key, &key, &ih, inode, (char *)(&sd));
1818 reiserfs_check_path(&path_to_key) ;
1822 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1823 if (!th->displace_new_blocks)
1824 REISERFS_I(dir)->new_packing_locality = 0;
1826 if (S_ISDIR(mode)) {
1827 /* insert item with "." and ".." */
1828 retval = reiserfs_new_directory (th, inode, &ih, &path_to_key, dir);
1831 if (S_ISLNK(mode)) {
1832 /* insert body of symlink */
1833 if (!old_format_only (sb))
1834 i_size = ROUND_UP(i_size);
1835 retval = reiserfs_new_symlink (th, inode, &ih, &path_to_key, symname, i_size);
1839 reiserfs_check_path(&path_to_key) ;
1840 journal_end(th, th->t_super, th->t_blocks_allocated);
1841 goto out_inserted_sd;
1844 /* XXX CHECK THIS */
1845 if (reiserfs_posixacl (inode->i_sb)) {
1846 retval = reiserfs_inherit_default_acl (dir, dentry, inode);
1849 reiserfs_check_path(&path_to_key) ;
1850 journal_end(th, th->t_super, th->t_blocks_allocated);
1851 goto out_inserted_sd;
1853 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1854 reiserfs_warning (inode->i_sb, "ACLs aren't enabled in the fs, "
1855 "but vfs thinks they are!");
1858 insert_inode_hash (inode);
1859 reiserfs_update_sd(th, inode);
1860 reiserfs_check_path(&path_to_key) ;
1864 /* it looks like you can easily compress these two goto targets into
1865 * one. Keeping it like this doesn't actually hurt anything, and they
1866 * are place holders for what the quota code actually needs.
1869 /* Invalidate the object, nothing was inserted yet */
1870 INODE_PKEY(inode)->k_objectid = 0;
1872 /* Quota change must be inside a transaction for journaling */
1873 DQUOT_FREE_INODE(inode);
1876 journal_end(th, th->t_super, th->t_blocks_allocated) ;
1877 /* Drop can be outside and it needs more credits so it's better to have it outside */
1879 inode->i_flags |= S_NOQUOTA;
1880 make_bad_inode(inode);
1884 th->t_trans_id = 0; /* so the caller can't use this handle later */
1890 ** finds the tail page in the page cache,
1891 ** reads the last block in.
1893 ** On success, page_result is set to a locked, pinned page, and bh_result
1894 ** is set to an up to date buffer for the last block in the file. returns 0.
1896 ** tail conversion is not done, so bh_result might not be valid for writing
1897 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1898 ** trying to write the block.
1900 ** on failure, nonzero is returned, page_result and bh_result are untouched.
1902 static int grab_tail_page(struct inode *p_s_inode,
1903 struct page **page_result,
1904 struct buffer_head **bh_result) {
1906 /* we want the page with the last byte in the file,
1907 ** not the page that will hold the next byte for appending
1909 unsigned long index = (p_s_inode->i_size-1) >> PAGE_CACHE_SHIFT ;
1910 unsigned long pos = 0 ;
1911 unsigned long start = 0 ;
1912 unsigned long blocksize = p_s_inode->i_sb->s_blocksize ;
1913 unsigned long offset = (p_s_inode->i_size) & (PAGE_CACHE_SIZE - 1) ;
1914 struct buffer_head *bh ;
1915 struct buffer_head *head ;
1916 struct page * page ;
1919 /* we know that we are only called with inode->i_size > 0.
1920 ** we also know that a file tail can never be as big as a block
1921 ** If i_size % blocksize == 0, our file is currently block aligned
1922 ** and it won't need converting or zeroing after a truncate.
1924 if ((offset & (blocksize - 1)) == 0) {
1927 page = grab_cache_page(p_s_inode->i_mapping, index) ;
1932 /* start within the page of the last block in the file */
1933 start = (offset / blocksize) * blocksize ;
1935 error = block_prepare_write(page, start, offset,
1936 reiserfs_get_block_create_0) ;
1940 head = page_buffers(page) ;
1946 bh = bh->b_this_page ;
1948 } while(bh != head) ;
1950 if (!buffer_uptodate(bh)) {
1951 /* note, this should never happen, prepare_write should
1952 ** be taking care of this for us. If the buffer isn't up to date,
1953 ** I've screwed up the code to find the buffer, or the code to
1954 ** call prepare_write
1956 reiserfs_warning (p_s_inode->i_sb,
1957 "clm-6000: error reading block %lu on dev %s",
1959 reiserfs_bdevname (p_s_inode->i_sb)) ;
1964 *page_result = page ;
1971 page_cache_release(page) ;
1976 ** vfs version of truncate file. Must NOT be called with
1977 ** a transaction already started.
1979 ** some code taken from block_truncate_page
1981 int reiserfs_truncate_file(struct inode *p_s_inode, int update_timestamps) {
1982 struct reiserfs_transaction_handle th ;
1983 /* we want the offset for the first byte after the end of the file */
1984 unsigned long offset = p_s_inode->i_size & (PAGE_CACHE_SIZE - 1) ;
1985 unsigned blocksize = p_s_inode->i_sb->s_blocksize ;
1987 struct page *page = NULL ;
1989 struct buffer_head *bh = NULL ;
1991 reiserfs_write_lock(p_s_inode->i_sb);
1993 if (p_s_inode->i_size > 0) {
1994 if ((error = grab_tail_page(p_s_inode, &page, &bh))) {
1995 // -ENOENT means we truncated past the end of the file,
1996 // and get_block_create_0 could not find a block to read in,
1998 if (error != -ENOENT)
1999 reiserfs_warning (p_s_inode->i_sb,
2000 "clm-6001: grab_tail_page failed %d",
2007 /* so, if page != NULL, we have a buffer head for the offset at
2008 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2009 ** then we have an unformatted node. Otherwise, we have a direct item,
2010 ** and no zeroing is required on disk. We zero after the truncate,
2011 ** because the truncate might pack the item anyway
2012 ** (it will unmap bh if it packs).
2014 /* it is enough to reserve space in transaction for 2 balancings:
2015 one for "save" link adding and another for the first
2016 cut_from_item. 1 is for update_sd */
2017 error = journal_begin (&th, p_s_inode->i_sb,
2018 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2021 reiserfs_update_inode_transaction(p_s_inode) ;
2022 if (update_timestamps)
2023 /* we are doing real truncate: if the system crashes before the last
2024 transaction of truncating gets committed - on reboot the file
2025 either appears truncated properly or not truncated at all */
2026 add_save_link (&th, p_s_inode, 1);
2027 error = reiserfs_do_truncate (&th, p_s_inode, page, update_timestamps) ;
2030 error = journal_end (&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2034 if (update_timestamps) {
2035 error = remove_save_link (p_s_inode, 1/* truncate */);
2041 length = offset & (blocksize - 1) ;
2042 /* if we are not on a block boundary */
2046 length = blocksize - length ;
2047 kaddr = kmap_atomic(page, KM_USER0) ;
2048 memset(kaddr + offset, 0, length) ;
2049 flush_dcache_page(page) ;
2050 kunmap_atomic(kaddr, KM_USER0) ;
2051 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2052 mark_buffer_dirty(bh) ;
2056 page_cache_release(page) ;
2059 reiserfs_write_unlock(p_s_inode->i_sb);
2064 page_cache_release (page);
2066 reiserfs_write_unlock(p_s_inode->i_sb);
2070 static int map_block_for_writepage(struct inode *inode,
2071 struct buffer_head *bh_result,
2072 unsigned long block) {
2073 struct reiserfs_transaction_handle th ;
2075 struct item_head tmp_ih ;
2076 struct item_head *ih ;
2077 struct buffer_head *bh ;
2079 struct cpu_key key ;
2080 INITIALIZE_PATH(path) ;
2082 int jbegin_count = JOURNAL_PER_BALANCE_CNT ;
2083 loff_t byte_offset = (block << inode->i_sb->s_blocksize_bits) + 1 ;
2085 int use_get_block = 0 ;
2086 int bytes_copied = 0 ;
2088 int trans_running = 0;
2090 /* catch places below that try to log something without starting a trans */
2093 if (!buffer_uptodate(bh_result)) {
2097 kmap(bh_result->b_page) ;
2099 reiserfs_write_lock(inode->i_sb);
2100 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3) ;
2103 retval = search_for_position_by_key(inode->i_sb, &key, &path) ;
2104 if (retval != POSITION_FOUND) {
2109 bh = get_last_bh(&path) ;
2110 ih = get_ih(&path) ;
2111 item = get_item(&path) ;
2112 pos_in_item = path.pos_in_item ;
2114 /* we've found an unformatted node */
2115 if (indirect_item_found(retval, ih)) {
2116 if (bytes_copied > 0) {
2117 reiserfs_warning (inode->i_sb, "clm-6002: bytes_copied %d",
2120 if (!get_block_num(item, pos_in_item)) {
2121 /* crap, we are writing to a hole */
2125 set_block_dev_mapped(bh_result, get_block_num(item,pos_in_item),inode);
2126 } else if (is_direct_le_ih(ih)) {
2128 p = page_address(bh_result->b_page) ;
2129 p += (byte_offset -1) & (PAGE_CACHE_SIZE - 1) ;
2130 copy_size = ih_item_len(ih) - pos_in_item;
2132 fs_gen = get_generation(inode->i_sb) ;
2133 copy_item_head(&tmp_ih, ih) ;
2135 if (!trans_running) {
2136 /* vs-3050 is gone, no need to drop the path */
2137 retval = journal_begin(&th, inode->i_sb, jbegin_count) ;
2140 reiserfs_update_inode_transaction(inode) ;
2142 if (fs_changed(fs_gen, inode->i_sb) && item_moved(&tmp_ih, &path)) {
2143 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
2148 reiserfs_prepare_for_journal(inode->i_sb, bh, 1) ;
2150 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
2151 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
2155 memcpy( B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied, copy_size) ;
2157 journal_mark_dirty(&th, inode->i_sb, bh) ;
2158 bytes_copied += copy_size ;
2159 set_block_dev_mapped(bh_result, 0, inode);
2161 /* are there still bytes left? */
2162 if (bytes_copied < bh_result->b_size &&
2163 (byte_offset + bytes_copied) < inode->i_size) {
2164 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + copy_size) ;
2168 reiserfs_warning (inode->i_sb,
2169 "clm-6003: bad item inode %lu, device %s",
2170 inode->i_ino, reiserfs_bdevname (inode->i_sb)) ;
2178 if (trans_running) {
2179 int err = journal_end(&th, inode->i_sb, jbegin_count) ;
2184 reiserfs_write_unlock(inode->i_sb);
2186 /* this is where we fill in holes in the file. */
2187 if (use_get_block) {
2188 retval = reiserfs_get_block(inode, block, bh_result,
2189 GET_BLOCK_CREATE | GET_BLOCK_NO_ISEM |
2190 GET_BLOCK_NO_DANGLE);
2192 if (!buffer_mapped(bh_result) || bh_result->b_blocknr == 0) {
2193 /* get_block failed to find a mapped unformatted node. */
2199 kunmap(bh_result->b_page) ;
2201 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2202 /* we've copied data from the page into the direct item, so the
2203 * buffer in the page is now clean, mark it to reflect that.
2205 lock_buffer(bh_result);
2206 clear_buffer_dirty(bh_result);
2207 unlock_buffer(bh_result);
2213 * mason@suse.com: updated in 2.5.54 to follow the same general io
2214 * start/recovery path as __block_write_full_page, along with special
2215 * code to handle reiserfs tails.
2217 static int reiserfs_write_full_page(struct page *page, struct writeback_control *wbc) {
2218 struct inode *inode = page->mapping->host ;
2219 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT ;
2221 unsigned long block ;
2222 struct buffer_head *head, *bh;
2225 int checked = PageChecked(page);
2226 struct reiserfs_transaction_handle th;
2227 struct super_block *s = inode->i_sb;
2228 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2231 /* The page dirty bit is cleared before writepage is called, which
2232 * means we have to tell create_empty_buffers to make dirty buffers
2233 * The page really should be up to date at this point, so tossing
2234 * in the BH_Uptodate is just a sanity check.
2236 if (!page_has_buffers(page)) {
2237 create_empty_buffers(page, s->s_blocksize,
2238 (1 << BH_Dirty) | (1 << BH_Uptodate));
2240 head = page_buffers(page) ;
2242 /* last page in the file, zero out any contents past the
2243 ** last byte in the file
2245 if (page->index >= end_index) {
2247 unsigned last_offset;
2249 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1) ;
2250 /* no file contents in this page */
2251 if (page->index >= end_index + 1 || !last_offset) {
2255 kaddr = kmap_atomic(page, KM_USER0);
2256 memset(kaddr + last_offset, 0, PAGE_CACHE_SIZE-last_offset) ;
2257 flush_dcache_page(page) ;
2258 kunmap_atomic(kaddr, KM_USER0) ;
2261 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits) ;
2262 /* first map all the buffers, logging any direct items we find */
2264 if ((checked || buffer_dirty(bh)) && (!buffer_mapped(bh) ||
2265 (buffer_mapped(bh) && bh->b_blocknr == 0))) {
2266 /* not mapped yet, or it points to a direct item, search
2267 * the btree for the mapping info, and log any direct
2270 if ((error = map_block_for_writepage(inode, bh, block))) {
2274 bh = bh->b_this_page;
2276 } while(bh != head) ;
2279 * we start the transaction after map_block_for_writepage,
2280 * because it can create holes in the file (an unbounded operation).
2281 * starting it here, we can make a reliable estimate for how many
2282 * blocks we're going to log
2285 ClearPageChecked(page);
2286 reiserfs_write_lock(s);
2287 error = journal_begin(&th, s, bh_per_page + 1);
2289 reiserfs_write_unlock(s);
2292 reiserfs_update_inode_transaction(inode);
2294 /* now go through and lock any dirty buffers on the page */
2297 if (!buffer_mapped(bh))
2299 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2303 reiserfs_prepare_for_journal(s, bh, 1);
2304 journal_mark_dirty(&th, s, bh);
2307 /* from this point on, we know the buffer is mapped to a
2308 * real block and not a direct item
2310 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2313 if (test_set_buffer_locked(bh)) {
2314 redirty_page_for_writepage(wbc, page);
2318 if (test_clear_buffer_dirty(bh)) {
2319 mark_buffer_async_write(bh);
2323 } while((bh = bh->b_this_page) != head);
2326 error = journal_end(&th, s, bh_per_page + 1);
2327 reiserfs_write_unlock(s);
2331 BUG_ON(PageWriteback(page));
2332 set_page_writeback(page);
2336 * since any buffer might be the only dirty buffer on the page,
2337 * the first submit_bh can bring the page out of writeback.
2338 * be careful with the buffers.
2341 struct buffer_head *next = bh->b_this_page;
2342 if (buffer_async_write(bh)) {
2343 submit_bh(WRITE, bh);
2348 } while(bh != head);
2354 * if this page only had a direct item, it is very possible for
2355 * no io to be required without there being an error. Or,
2356 * someone else could have locked them and sent them down the
2357 * pipe without locking the page
2361 if (!buffer_uptodate(bh)) {
2365 bh = bh->b_this_page;
2366 } while(bh != head);
2368 SetPageUptodate(page);
2369 end_page_writeback(page);
2374 /* catches various errors, we need to make sure any valid dirty blocks
2375 * get to the media. The page is currently locked and not marked for
2378 ClearPageUptodate(page);
2382 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2384 mark_buffer_async_write(bh);
2387 * clear any dirty bits that might have come from getting
2388 * attached to a dirty page
2390 clear_buffer_dirty(bh);
2392 bh = bh->b_this_page;
2393 } while(bh != head);
2395 BUG_ON(PageWriteback(page));
2396 set_page_writeback(page);
2399 struct buffer_head *next = bh->b_this_page;
2400 if (buffer_async_write(bh)) {
2401 clear_buffer_dirty(bh);
2402 submit_bh(WRITE, bh);
2407 } while(bh != head);
2412 static int reiserfs_readpage (struct file *f, struct page * page)
2414 return block_read_full_page (page, reiserfs_get_block);
2418 static int reiserfs_writepage (struct page * page, struct writeback_control *wbc)
2420 struct inode *inode = page->mapping->host ;
2421 reiserfs_wait_on_write_block(inode->i_sb) ;
2422 return reiserfs_write_full_page(page, wbc) ;
2425 static int reiserfs_prepare_write(struct file *f, struct page *page,
2426 unsigned from, unsigned to) {
2427 struct inode *inode = page->mapping->host ;
2431 reiserfs_wait_on_write_block(inode->i_sb) ;
2432 fix_tail_page_for_writing(page) ;
2433 if (reiserfs_transaction_running(inode->i_sb)) {
2434 struct reiserfs_transaction_handle *th;
2435 th = (struct reiserfs_transaction_handle *)current->journal_info;
2436 BUG_ON (!th->t_refcount);
2437 BUG_ON (!th->t_trans_id);
2438 old_ref = th->t_refcount;
2442 ret = block_prepare_write(page, from, to, reiserfs_get_block) ;
2443 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2444 struct reiserfs_transaction_handle *th = current->journal_info;
2445 /* this gets a little ugly. If reiserfs_get_block returned an
2446 * error and left a transacstion running, we've got to close it,
2447 * and we've got to free handle if it was a persistent transaction.
2449 * But, if we had nested into an existing transaction, we need
2450 * to just drop the ref count on the handle.
2452 * If old_ref == 0, the transaction is from reiserfs_get_block,
2453 * and it was a persistent trans. Otherwise, it was nested above.
2455 if (th->t_refcount > old_ref) {
2460 reiserfs_write_lock(inode->i_sb);
2461 err = reiserfs_end_persistent_transaction(th);
2462 reiserfs_write_unlock(inode->i_sb);
2473 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block) {
2474 return generic_block_bmap(as, block, reiserfs_bmap) ;
2477 static int reiserfs_commit_write(struct file *f, struct page *page,
2478 unsigned from, unsigned to) {
2479 struct inode *inode = page->mapping->host ;
2480 loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
2483 struct reiserfs_transaction_handle *th = NULL;
2485 reiserfs_wait_on_write_block(inode->i_sb) ;
2486 if (reiserfs_transaction_running(inode->i_sb)) {
2487 th = current->journal_info;
2489 reiserfs_commit_page(inode, page, from, to);
2491 /* generic_commit_write does this for us, but does not update the
2492 ** transaction tracking stuff when the size changes. So, we have
2493 ** to do the i_size updates here.
2495 if (pos > inode->i_size) {
2496 struct reiserfs_transaction_handle myth ;
2497 reiserfs_write_lock(inode->i_sb);
2498 /* If the file have grown beyond the border where it
2499 can have a tail, unmark it as needing a tail
2501 if ( (have_large_tails (inode->i_sb) && inode->i_size > i_block_size (inode)*4) ||
2502 (have_small_tails (inode->i_sb) && inode->i_size > i_block_size(inode)) )
2503 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask ;
2505 ret = journal_begin(&myth, inode->i_sb, 1) ;
2507 reiserfs_write_unlock(inode->i_sb);
2510 reiserfs_update_inode_transaction(inode) ;
2511 inode->i_size = pos ;
2512 reiserfs_update_sd(&myth, inode) ;
2514 ret = journal_end(&myth, inode->i_sb, 1) ;
2515 reiserfs_write_unlock(inode->i_sb);
2520 reiserfs_write_lock(inode->i_sb);
2522 reiserfs_update_sd(th, inode) ;
2523 ret = reiserfs_end_persistent_transaction(th);
2524 reiserfs_write_unlock(inode->i_sb);
2529 /* we test for O_SYNC here so we can commit the transaction
2530 ** for any packed tails the file might have had
2532 if (f && (f->f_flags & O_SYNC)) {
2533 reiserfs_write_lock(inode->i_sb);
2534 ret = reiserfs_commit_for_inode(inode) ;
2535 reiserfs_write_unlock(inode->i_sb);
2542 reiserfs_write_lock(inode->i_sb);
2544 reiserfs_update_sd(th, inode) ;
2545 ret = reiserfs_end_persistent_transaction(th);
2546 reiserfs_write_unlock(inode->i_sb);
2552 void sd_attrs_to_i_attrs( __u16 sd_attrs, struct inode *inode )
2554 if( reiserfs_attrs( inode -> i_sb ) ) {
2555 if( sd_attrs & REISERFS_SYNC_FL )
2556 inode -> i_flags |= S_SYNC;
2558 inode -> i_flags &= ~S_SYNC;
2559 if( sd_attrs & REISERFS_IMMUTABLE_FL )
2560 inode -> i_flags |= S_IMMUTABLE;
2562 inode -> i_flags &= ~S_IMMUTABLE;
2563 if( sd_attrs & REISERFS_IUNLINK_FL )
2564 inode -> i_flags |= S_IUNLINK;
2566 inode -> i_flags &= ~S_IUNLINK;
2567 if( sd_attrs & REISERFS_BARRIER_FL )
2568 inode -> i_flags |= S_BARRIER;
2570 inode -> i_flags &= ~S_BARRIER;
2571 if( sd_attrs & REISERFS_APPEND_FL )
2572 inode -> i_flags |= S_APPEND;
2574 inode -> i_flags &= ~S_APPEND;
2575 if( sd_attrs & REISERFS_NOATIME_FL )
2576 inode -> i_flags |= S_NOATIME;
2578 inode -> i_flags &= ~S_NOATIME;
2579 if( sd_attrs & REISERFS_NOTAIL_FL )
2580 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2582 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2586 void i_attrs_to_sd_attrs( struct inode *inode, __u16 *sd_attrs )
2588 if( reiserfs_attrs( inode -> i_sb ) ) {
2589 if( inode -> i_flags & S_IMMUTABLE )
2590 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2592 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2593 if( inode -> i_flags & S_IUNLINK )
2594 *sd_attrs |= REISERFS_IUNLINK_FL;
2596 *sd_attrs &= ~REISERFS_IUNLINK_FL;
2597 if( inode -> i_flags & S_BARRIER )
2598 *sd_attrs |= REISERFS_BARRIER_FL;
2600 *sd_attrs &= ~REISERFS_BARRIER_FL;
2601 if( inode -> i_flags & S_SYNC )
2602 *sd_attrs |= REISERFS_SYNC_FL;
2604 *sd_attrs &= ~REISERFS_SYNC_FL;
2605 if( inode -> i_flags & S_NOATIME )
2606 *sd_attrs |= REISERFS_NOATIME_FL;
2608 *sd_attrs &= ~REISERFS_NOATIME_FL;
2609 if( REISERFS_I(inode)->i_flags & i_nopack_mask )
2610 *sd_attrs |= REISERFS_NOTAIL_FL;
2612 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2616 /* decide if this buffer needs to stay around for data logging or ordered
2619 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2622 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb) ;
2624 spin_lock(&j->j_dirty_buffers_lock) ;
2625 if (!buffer_mapped(bh)) {
2628 /* the page is locked, and the only places that log a data buffer
2629 * also lock the page.
2631 if (reiserfs_file_data_log(inode)) {
2633 * very conservative, leave the buffer pinned if
2634 * anyone might need it.
2636 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2640 if (buffer_dirty(bh) || buffer_locked(bh)) {
2641 struct reiserfs_journal_list *jl;
2642 struct reiserfs_jh *jh = bh->b_private;
2644 /* why is this safe?
2645 * reiserfs_setattr updates i_size in the on disk
2646 * stat data before allowing vmtruncate to be called.
2648 * If buffer was put onto the ordered list for this
2649 * transaction, we know for sure either this transaction
2650 * or an older one already has updated i_size on disk,
2651 * and this ordered data won't be referenced in the file
2654 * if the buffer was put onto the ordered list for an older
2655 * transaction, we need to leave it around
2657 if (jh && (jl = jh->jl) && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2661 if (ret && bh->b_private) {
2662 reiserfs_free_jh(bh);
2664 spin_unlock(&j->j_dirty_buffers_lock) ;
2668 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2669 static int reiserfs_invalidatepage(struct page *page, unsigned long offset)
2671 struct buffer_head *head, *bh, *next;
2672 struct inode *inode = page->mapping->host;
2673 unsigned int curr_off = 0;
2676 BUG_ON(!PageLocked(page));
2679 ClearPageChecked(page);
2681 if (!page_has_buffers(page))
2684 head = page_buffers(page);
2687 unsigned int next_off = curr_off + bh->b_size;
2688 next = bh->b_this_page;
2691 * is this block fully invalidated?
2693 if (offset <= curr_off) {
2694 if (invalidatepage_can_drop(inode, bh))
2695 reiserfs_unmap_buffer(bh);
2699 curr_off = next_off;
2701 } while (bh != head);
2704 * We release buffers only if the entire page is being invalidated.
2705 * The get_block cached value has been unconditionally invalidated,
2706 * so real IO is not possible anymore.
2709 ret = try_to_release_page(page, 0);
2714 static int reiserfs_set_page_dirty(struct page *page) {
2715 struct inode *inode = page->mapping->host;
2716 if (reiserfs_file_data_log(inode)) {
2717 SetPageChecked(page);
2718 return __set_page_dirty_nobuffers(page);
2720 return __set_page_dirty_buffers(page);
2724 * Returns 1 if the page's buffers were dropped. The page is locked.
2726 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
2727 * in the buffers at page_buffers(page).
2729 * even in -o notail mode, we can't be sure an old mount without -o notail
2730 * didn't create files with tails.
2732 static int reiserfs_releasepage(struct page *page, int unused_gfp_flags)
2734 struct inode *inode = page->mapping->host ;
2735 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb) ;
2736 struct buffer_head *head ;
2737 struct buffer_head *bh ;
2740 WARN_ON(PageChecked(page));
2741 spin_lock(&j->j_dirty_buffers_lock) ;
2742 head = page_buffers(page) ;
2745 if (bh->b_private) {
2746 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
2747 reiserfs_free_jh(bh);
2753 bh = bh->b_this_page ;
2754 } while (bh != head) ;
2756 ret = try_to_free_buffers(page) ;
2757 spin_unlock(&j->j_dirty_buffers_lock) ;
2761 /* We thank Mingming Cao for helping us understand in great detail what
2762 to do in this section of the code. */
2763 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
2764 const struct iovec *iov, loff_t offset, unsigned long nr_segs)
2766 struct file *file = iocb->ki_filp;
2767 struct inode *inode = file->f_mapping->host;
2769 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2770 offset, nr_segs, reiserfs_get_blocks_direct_io, NULL);
2773 int reiserfs_setattr_flags(struct inode *inode, unsigned int flags)
2775 unsigned int oldflags, newflags;
2777 oldflags = REISERFS_I(inode)->i_flags;
2778 newflags = oldflags & ~(REISERFS_IMMUTABLE_FL |
2779 REISERFS_IUNLINK_FL | REISERFS_BARRIER_FL);
2780 if (flags & ATTR_FLAG_IMMUTABLE)
2781 newflags |= REISERFS_IMMUTABLE_FL;
2782 if (flags & ATTR_FLAG_IUNLINK)
2783 newflags |= REISERFS_IUNLINK_FL;
2784 if (flags & ATTR_FLAG_BARRIER)
2785 newflags |= REISERFS_BARRIER_FL;
2787 if (oldflags ^ newflags) {
2788 REISERFS_I(inode)->i_flags = newflags;
2789 inode->i_ctime = CURRENT_TIME;
2794 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr) {
2795 struct inode *inode = dentry->d_inode ;
2797 unsigned int ia_valid = attr->ia_valid;
2798 reiserfs_write_lock(inode->i_sb);
2799 if (attr->ia_valid & ATTR_SIZE) {
2800 /* version 2 items will be caught by the s_maxbytes check
2801 ** done for us in vmtruncate
2803 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
2804 attr->ia_size > MAX_NON_LFS) {
2808 /* fill in hole pointers in the expanding truncate case. */
2809 if (attr->ia_size > inode->i_size) {
2810 error = generic_cont_expand(inode, attr->ia_size) ;
2811 if (REISERFS_I(inode)->i_prealloc_count > 0) {
2813 struct reiserfs_transaction_handle th ;
2814 /* we're changing at most 2 bitmaps, inode + super */
2815 err = journal_begin(&th, inode->i_sb, 4) ;
2817 reiserfs_discard_prealloc (&th, inode);
2818 err = journal_end(&th, inode->i_sb, 4) ;
2828 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
2829 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
2830 (get_inode_sd_version (inode) == STAT_DATA_V1)) {
2831 /* stat data of format v3.5 has 16 bit uid and gid */
2836 error = inode_change_ok(inode, attr) ;
2838 if (!error && attr->ia_valid & ATTR_ATTR_FLAG)
2839 reiserfs_setattr_flags(inode, attr->ia_attr_flags);
2842 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
2843 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
2844 error = reiserfs_chown_xattrs (inode, attr);
2847 struct reiserfs_transaction_handle th;
2849 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
2850 journal_begin(&th, inode->i_sb, 4*REISERFS_QUOTA_INIT_BLOCKS+2);
2851 error = DQUOT_TRANSFER(inode, attr) ? -EDQUOT : 0;
2853 journal_end(&th, inode->i_sb, 4*REISERFS_QUOTA_INIT_BLOCKS+2);
2856 /* Update corresponding info in inode so that everything is in
2857 * one transaction */
2858 if (attr->ia_valid & ATTR_UID)
2859 inode->i_uid = attr->ia_uid;
2860 if (attr->ia_valid & ATTR_GID)
2861 inode->i_gid = attr->ia_gid;
2862 mark_inode_dirty(inode);
2863 journal_end(&th, inode->i_sb, 4*REISERFS_QUOTA_INIT_BLOCKS+2);
2867 error = inode_setattr(inode, attr) ;
2871 if (!error && reiserfs_posixacl (inode->i_sb)) {
2872 if (attr->ia_valid & ATTR_MODE)
2873 error = reiserfs_acl_chmod (inode);
2877 reiserfs_write_unlock(inode->i_sb);
2883 struct address_space_operations reiserfs_address_space_operations = {
2884 .writepage = reiserfs_writepage,
2885 .readpage = reiserfs_readpage,
2886 .readpages = reiserfs_readpages,
2887 .releasepage = reiserfs_releasepage,
2888 .invalidatepage = reiserfs_invalidatepage,
2889 .sync_page = block_sync_page,
2890 .prepare_write = reiserfs_prepare_write,
2891 .commit_write = reiserfs_commit_write,
2892 .bmap = reiserfs_aop_bmap,
2893 .direct_IO = reiserfs_direct_IO,
2894 .set_page_dirty = reiserfs_set_page_dirty,