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>
21 extern int reiserfs_default_io_size; /* default io size devuned in super.c */
23 /* args for the create parameter of reiserfs_get_block */
24 #define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
25 #define GET_BLOCK_CREATE 1 /* add anything you need to find block */
26 #define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
27 #define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
28 #define GET_BLOCK_NO_ISEM 8 /* i_sem is not held, don't preallocate */
29 #define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
31 static int reiserfs_get_block (struct inode * inode, sector_t block,
32 struct buffer_head * bh_result, int create);
33 static int reiserfs_commit_write(struct file *f, struct page *page,
34 unsigned from, unsigned to);
36 void reiserfs_delete_inode (struct inode * inode)
38 int jbegin_count = JOURNAL_PER_BALANCE_CNT * 2;
39 struct reiserfs_transaction_handle th ;
41 reiserfs_write_lock(inode->i_sb);
43 DQUOT_FREE_INODE(inode);
44 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
45 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
48 reiserfs_delete_xattrs (inode);
50 journal_begin(&th, inode->i_sb, jbegin_count) ;
51 reiserfs_update_inode_transaction(inode) ;
53 reiserfs_delete_object (&th, inode);
55 journal_end(&th, inode->i_sb, jbegin_count) ;
59 /* all items of file are deleted, so we can remove "save" link */
60 remove_save_link (inode, 0/* not truncate */);
62 /* no object items are in the tree */
65 clear_inode (inode); /* note this must go after the journal_end to prevent deadlock */
67 reiserfs_write_unlock(inode->i_sb);
70 static void _make_cpu_key (struct cpu_key * key, int version, __u32 dirid, __u32 objectid,
71 loff_t offset, int type, int length )
73 key->version = version;
75 key->on_disk_key.k_dir_id = dirid;
76 key->on_disk_key.k_objectid = objectid;
77 set_cpu_key_k_offset (key, offset);
78 set_cpu_key_k_type (key, type);
79 key->key_length = length;
83 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
84 offset and type of key */
85 void make_cpu_key (struct cpu_key * key, struct inode * inode, loff_t offset,
86 int type, int length )
88 _make_cpu_key (key, get_inode_item_key_version (inode), le32_to_cpu (INODE_PKEY (inode)->k_dir_id),
89 le32_to_cpu (INODE_PKEY (inode)->k_objectid),
90 offset, type, length);
95 // when key is 0, do not set version and short key
97 inline void make_le_item_head (struct item_head * ih, const struct cpu_key * key,
99 loff_t offset, int type, int length,
100 int entry_count/*or ih_free_space*/)
103 ih->ih_key.k_dir_id = cpu_to_le32 (key->on_disk_key.k_dir_id);
104 ih->ih_key.k_objectid = cpu_to_le32 (key->on_disk_key.k_objectid);
106 put_ih_version( ih, version );
107 set_le_ih_k_offset (ih, offset);
108 set_le_ih_k_type (ih, type);
109 put_ih_item_len( ih, length );
110 /* set_ih_free_space (ih, 0);*/
111 // for directory items it is entry count, for directs and stat
112 // datas - 0xffff, for indirects - 0
113 put_ih_entry_count( ih, entry_count );
117 // FIXME: we might cache recently accessed indirect item
119 // Ugh. Not too eager for that....
120 // I cut the code until such time as I see a convincing argument (benchmark).
121 // I don't want a bloated inode struct..., and I don't like code complexity....
123 /* cutting the code is fine, since it really isn't in use yet and is easy
124 ** to add back in. But, Vladimir has a really good idea here. Think
125 ** about what happens for reading a file. For each page,
126 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
127 ** an indirect item. This indirect item has X number of pointers, where
128 ** X is a big number if we've done the block allocation right. But,
129 ** we only use one or two of these pointers during each call to readpage,
130 ** needlessly researching again later on.
132 ** The size of the cache could be dynamic based on the size of the file.
134 ** I'd also like to see us cache the location the stat data item, since
135 ** we are needlessly researching for that frequently.
140 /* If this page has a file tail in it, and
141 ** it was read in by get_block_create_0, the page data is valid,
142 ** but tail is still sitting in a direct item, and we can't write to
143 ** it. So, look through this page, and check all the mapped buffers
144 ** to make sure they have valid block numbers. Any that don't need
145 ** to be unmapped, so that block_prepare_write will correctly call
146 ** reiserfs_get_block to convert the tail into an unformatted node
148 static inline void fix_tail_page_for_writing(struct page *page) {
149 struct buffer_head *head, *next, *bh ;
151 if (page && page_has_buffers(page)) {
152 head = page_buffers(page) ;
155 next = bh->b_this_page ;
156 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
157 reiserfs_unmap_buffer(bh) ;
160 } while (bh != head) ;
164 /* reiserfs_get_block does not need to allocate a block only if it has been
165 done already or non-hole position has been found in the indirect item */
166 static inline int allocation_needed (int retval, b_blocknr_t allocated,
167 struct item_head * ih,
168 __u32 * item, int pos_in_item)
172 if (retval == POSITION_FOUND && is_indirect_le_ih (ih) &&
173 get_block_num(item, pos_in_item))
178 static inline int indirect_item_found (int retval, struct item_head * ih)
180 return (retval == POSITION_FOUND) && is_indirect_le_ih (ih);
184 static inline void set_block_dev_mapped (struct buffer_head * bh,
185 b_blocknr_t block, struct inode * inode)
187 map_bh(bh, inode->i_sb, block);
192 // files which were created in the earlier version can not be longer,
195 static int file_capable (struct inode * inode, long block)
197 if (get_inode_item_key_version (inode) != KEY_FORMAT_3_5 || // it is new file.
198 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
204 /*static*/ void restart_transaction(struct reiserfs_transaction_handle *th,
205 struct inode *inode, struct path *path) {
206 struct super_block *s = th->t_super ;
207 int len = th->t_blocks_allocated ;
209 /* we cannot restart while nested */
210 if (th->t_refcount > 1) {
214 reiserfs_update_sd(th, inode) ;
215 journal_end(th, s, len) ;
216 journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6) ;
217 reiserfs_update_inode_transaction(inode) ;
220 // it is called by get_block when create == 0. Returns block number
221 // for 'block'-th logical block of file. When it hits direct item it
222 // returns 0 (being called from bmap) or read direct item into piece
223 // of page (bh_result)
225 // Please improve the english/clarity in the comment above, as it is
226 // hard to understand.
228 static int _get_block_create_0 (struct inode * inode, long block,
229 struct buffer_head * bh_result,
232 INITIALIZE_PATH (path);
234 struct buffer_head * bh;
235 struct item_head * ih, tmp_ih;
242 unsigned long offset ;
244 // prepare the key to look for the 'block'-th block of file
245 make_cpu_key (&key, inode,
246 (loff_t)block * inode->i_sb->s_blocksize + 1, TYPE_ANY, 3);
249 if (search_for_position_by_key (inode->i_sb, &key, &path) != POSITION_FOUND) {
252 kunmap(bh_result->b_page) ;
253 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
254 // That there is some MMAPED data associated with it that is yet to be written to disk.
255 if ((args & GET_BLOCK_NO_HOLE) && !PageUptodate(bh_result->b_page) ) {
262 bh = get_last_bh (&path);
264 if (is_indirect_le_ih (ih)) {
265 __u32 * ind_item = (__u32 *)B_I_PITEM (bh, ih);
267 /* FIXME: here we could cache indirect item or part of it in
268 the inode to avoid search_by_key in case of subsequent
270 blocknr = get_block_num(ind_item, path.pos_in_item) ;
273 map_bh(bh_result, inode->i_sb, blocknr);
274 if (path.pos_in_item == ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
275 set_buffer_boundary(bh_result);
278 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
279 // That there is some MMAPED data associated with it that is yet to be written to disk.
280 if ((args & GET_BLOCK_NO_HOLE) && !PageUptodate(bh_result->b_page) ) {
286 kunmap(bh_result->b_page) ;
290 // requested data are in direct item(s)
291 if (!(args & GET_BLOCK_READ_DIRECT)) {
292 // we are called by bmap. FIXME: we can not map block of file
293 // when it is stored in direct item(s)
296 kunmap(bh_result->b_page) ;
300 /* if we've got a direct item, and the buffer or page was uptodate,
301 ** we don't want to pull data off disk again. skip to the
302 ** end, where we map the buffer and return
304 if (buffer_uptodate(bh_result)) {
308 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
309 ** pages without any buffers. If the page is up to date, we don't want
310 ** read old data off disk. Set the up to date bit on the buffer instead
311 ** and jump to the end
313 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
314 set_buffer_uptodate(bh_result);
318 // read file tail into part of page
319 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1) ;
320 fs_gen = get_generation(inode->i_sb) ;
321 copy_item_head (&tmp_ih, ih);
323 /* we only want to kmap if we are reading the tail into the page.
324 ** this is not the common case, so we don't kmap until we are
325 ** sure we need to. But, this means the item might move if
329 p = (char *)kmap(bh_result->b_page) ;
330 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
335 memset (p, 0, inode->i_sb->s_blocksize);
337 if (!is_direct_le_ih (ih)) {
340 /* make sure we don't read more bytes than actually exist in
341 ** the file. This can happen in odd cases where i_size isn't
342 ** correct, and when direct item padding results in a few
343 ** extra bytes at the end of the direct item
345 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
347 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
348 chars = inode->i_size - (le_ih_k_offset(ih) - 1) - path.pos_in_item;
351 chars = ih_item_len(ih) - path.pos_in_item;
353 memcpy (p, B_I_PITEM (bh, ih) + path.pos_in_item, chars);
360 if (PATH_LAST_POSITION (&path) != (B_NR_ITEMS (bh) - 1))
361 // we done, if read direct item is not the last item of
362 // node FIXME: we could try to check right delimiting key
363 // to see whether direct item continues in the right
364 // neighbor or rely on i_size
367 // update key to look for the next piece
368 set_cpu_key_k_offset (&key, cpu_key_k_offset (&key) + chars);
369 if (search_for_position_by_key (inode->i_sb, &key, &path) != POSITION_FOUND)
370 // we read something from tail, even if now we got IO_ERROR
372 bh = get_last_bh (&path);
376 flush_dcache_page(bh_result->b_page) ;
377 kunmap(bh_result->b_page) ;
381 /* this buffer has valid data, but isn't valid for io. mapping it to
382 * block #0 tells the rest of reiserfs it just has a tail in it
384 map_bh(bh_result, inode->i_sb, 0);
385 set_buffer_uptodate (bh_result);
390 // this is called to create file map. So, _get_block_create_0 will not
392 int reiserfs_bmap (struct inode * inode, sector_t block,
393 struct buffer_head * bh_result, int create)
395 if (!file_capable (inode, block))
398 reiserfs_write_lock(inode->i_sb);
399 /* do not read the direct item */
400 _get_block_create_0 (inode, block, bh_result, 0) ;
401 reiserfs_write_unlock(inode->i_sb);
405 /* special version of get_block that is only used by grab_tail_page right
406 ** now. It is sent to block_prepare_write, and when you try to get a
407 ** block past the end of the file (or a block from a hole) it returns
408 ** -ENOENT instead of a valid buffer. block_prepare_write expects to
409 ** be able to do i/o on the buffers returned, unless an error value
412 ** So, this allows block_prepare_write to be used for reading a single block
413 ** in a page. Where it does not produce a valid page for holes, or past the
414 ** end of the file. This turns out to be exactly what we need for reading
415 ** tails for conversion.
417 ** The point of the wrapper is forcing a certain value for create, even
418 ** though the VFS layer is calling this function with create==1. If you
419 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
420 ** don't use this function.
422 static int reiserfs_get_block_create_0 (struct inode * inode, sector_t block,
423 struct buffer_head * bh_result, int create) {
424 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE) ;
427 /* This is special helper for reiserfs_get_block in case we are executing
428 direct_IO request. */
429 static int reiserfs_get_blocks_direct_io(struct inode *inode,
431 unsigned long max_blocks,
432 struct buffer_head *bh_result,
437 bh_result->b_page = NULL;
439 /* We set the b_size before reiserfs_get_block call since it is
440 referenced in convert_tail_for_hole() that may be called from
441 reiserfs_get_block() */
442 bh_result->b_size = (1 << inode->i_blkbits);
444 ret = reiserfs_get_block(inode, iblock, bh_result,
445 create | GET_BLOCK_NO_DANGLE) ;
447 /* don't allow direct io onto tail pages */
448 if (ret == 0 && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
449 /* make sure future calls to the direct io funcs for this offset
450 ** in the file fail by unmapping the buffer
452 clear_buffer_mapped(bh_result);
455 /* Possible unpacked tail. Flush the data before pages have
457 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
459 reiserfs_commit_for_inode(inode);
460 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
468 ** helper function for when reiserfs_get_block is called for a hole
469 ** but the file tail is still in a direct item
470 ** bh_result is the buffer head for the hole
471 ** tail_offset is the offset of the start of the tail in the file
473 ** This calls prepare_write, which will start a new transaction
474 ** you should not be in a transaction, or have any paths held when you
477 static int convert_tail_for_hole(struct inode *inode,
478 struct buffer_head *bh_result,
479 loff_t tail_offset) {
480 unsigned long index ;
481 unsigned long tail_end ;
482 unsigned long tail_start ;
483 struct page * tail_page ;
484 struct page * hole_page = bh_result->b_page ;
487 if ((tail_offset & (bh_result->b_size - 1)) != 1)
490 /* always try to read until the end of the block */
491 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1) ;
492 tail_end = (tail_start | (bh_result->b_size - 1)) + 1 ;
494 index = tail_offset >> PAGE_CACHE_SHIFT ;
495 /* hole_page can be zero in case of direct_io, we are sure
496 that we cannot get here if we write with O_DIRECT into
498 if (!hole_page || index != hole_page->index) {
499 tail_page = grab_cache_page(inode->i_mapping, index) ;
505 tail_page = hole_page ;
508 /* we don't have to make sure the conversion did not happen while
509 ** we were locking the page because anyone that could convert
510 ** must first take i_sem.
512 ** We must fix the tail page for writing because it might have buffers
513 ** that are mapped, but have a block number of 0. This indicates tail
514 ** data that has been read directly into the page, and block_prepare_write
515 ** won't trigger a get_block in this case.
517 fix_tail_page_for_writing(tail_page) ;
518 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
522 /* tail conversion might change the data in the page */
523 flush_dcache_page(tail_page) ;
525 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end) ;
528 if (tail_page != hole_page) {
529 unlock_page(tail_page) ;
530 page_cache_release(tail_page) ;
536 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
539 b_blocknr_t *allocated_block_nr,
543 #ifdef REISERFS_PREALLOCATE
544 if (!(flags & GET_BLOCK_NO_ISEM)) {
545 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr, path, block);
548 return reiserfs_new_unf_blocknrs (th, inode, allocated_block_nr, path, block);
551 int reiserfs_get_block (struct inode * inode, sector_t block,
552 struct buffer_head * bh_result, int create)
555 b_blocknr_t allocated_block_nr = 0;// b_blocknr_t is (unsigned) 32 bit int
556 INITIALIZE_PATH(path);
559 struct buffer_head * bh, * unbh = NULL;
560 struct item_head * ih, tmp_ih;
564 struct reiserfs_transaction_handle *th = NULL;
565 /* space reserved in transaction batch:
566 . 3 balancings in direct->indirect conversion
567 . 1 block involved into reiserfs_update_sd()
568 XXX in practically impossible worst case direct2indirect()
569 can incur (much) more that 3 balancings. */
570 int jbegin_count = JOURNAL_PER_BALANCE_CNT * 3 + 1;
573 loff_t new_offset = (((loff_t)block) << inode->i_sb->s_blocksize_bits) + 1 ;
576 reiserfs_write_lock(inode->i_sb);
577 version = get_inode_item_key_version (inode);
580 reiserfs_write_unlock(inode->i_sb);
584 if (!file_capable (inode, block)) {
585 reiserfs_write_unlock(inode->i_sb);
589 /* if !create, we aren't changing the FS, so we don't need to
590 ** log anything, so we don't need to start a transaction
592 if (!(create & GET_BLOCK_CREATE)) {
594 /* find number of block-th logical block of the file */
595 ret = _get_block_create_0 (inode, block, bh_result,
596 create | GET_BLOCK_READ_DIRECT) ;
597 reiserfs_write_unlock(inode->i_sb);
601 * if we're already in a transaction, make sure to close
602 * any new transactions we start in this func
604 if ((create & GET_BLOCK_NO_DANGLE) ||
605 reiserfs_transaction_running(inode->i_sb))
608 /* If file is of such a size, that it might have a tail and tails are enabled
609 ** we should mark it as possibly needing tail packing on close
611 if ( (have_large_tails (inode->i_sb) && inode->i_size < i_block_size (inode)*4) ||
612 (have_small_tails (inode->i_sb) && inode->i_size < i_block_size(inode)) )
613 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask ;
615 /* set the key of the first byte in the 'block'-th block of file */
616 make_cpu_key (&key, inode, new_offset,
617 TYPE_ANY, 3/*key length*/);
618 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
620 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
625 reiserfs_update_inode_transaction(inode) ;
629 retval = search_for_position_by_key (inode->i_sb, &key, &path);
630 if (retval == IO_ERROR) {
635 bh = get_last_bh (&path);
637 item = get_item (&path);
638 pos_in_item = path.pos_in_item;
640 fs_gen = get_generation (inode->i_sb);
641 copy_item_head (&tmp_ih, ih);
643 if (allocation_needed (retval, allocated_block_nr, ih, item, pos_in_item)) {
644 /* we have to allocate block for the unformatted node */
650 repeat = _allocate_block(th, block, inode, &allocated_block_nr, &path, create);
652 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
653 /* restart the transaction to give the journal a chance to free
654 ** some blocks. releases the path, so we have to go back to
655 ** research if we succeed on the second try
657 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
658 restart_transaction(th, inode, &path) ;
659 repeat = _allocate_block(th, block, inode, &allocated_block_nr, NULL, create);
661 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
664 if (repeat == QUOTA_EXCEEDED)
671 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
676 if (indirect_item_found (retval, ih)) {
677 b_blocknr_t unfm_ptr;
678 /* 'block'-th block is in the file already (there is
679 corresponding cell in some indirect item). But it may be
680 zero unformatted node pointer (hole) */
681 unfm_ptr = get_block_num (item, pos_in_item);
683 /* use allocated block to plug the hole */
684 reiserfs_prepare_for_journal(inode->i_sb, bh, 1) ;
685 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
686 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
689 set_buffer_new(bh_result);
690 if (buffer_dirty(bh_result) && reiserfs_data_ordered(inode->i_sb))
691 reiserfs_add_ordered_list(inode, bh_result);
692 put_block_num(item, pos_in_item, allocated_block_nr) ;
693 unfm_ptr = allocated_block_nr;
694 journal_mark_dirty (th, inode->i_sb, bh);
695 reiserfs_update_sd(th, inode) ;
697 set_block_dev_mapped(bh_result, unfm_ptr, inode);
700 reiserfs_end_persistent_transaction(th);
702 reiserfs_write_unlock(inode->i_sb);
704 /* the item was found, so new blocks were not added to the file
705 ** there is no need to make sure the inode is updated with this
716 /* desired position is not found or is in the direct item. We have
717 to append file with holes up to 'block'-th block converting
718 direct items to indirect one if necessary */
721 if (is_statdata_le_ih (ih)) {
723 struct cpu_key tmp_key;
725 /* indirect item has to be inserted */
726 make_le_item_head (&tmp_ih, &key, version, 1, TYPE_INDIRECT,
727 UNFM_P_SIZE, 0/* free_space */);
729 if (cpu_key_k_offset (&key) == 1) {
730 /* we are going to add 'block'-th block to the file. Use
731 allocated block for that */
732 unp = cpu_to_le32 (allocated_block_nr);
733 set_block_dev_mapped (bh_result, allocated_block_nr, inode);
734 set_buffer_new(bh_result);
738 set_cpu_key_k_offset (&tmp_key, 1);
739 PATH_LAST_POSITION(&path) ++;
741 retval = reiserfs_insert_item (th, &path, &tmp_key, &tmp_ih, inode, (char *)&unp);
743 reiserfs_free_block (th, inode, allocated_block_nr, 1);
744 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
746 //mark_tail_converted (inode);
747 } else if (is_direct_le_ih (ih)) {
748 /* direct item has to be converted */
751 tail_offset = ((le_ih_k_offset (ih) - 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
752 if (tail_offset == cpu_key_k_offset (&key)) {
753 /* direct item we just found fits into block we have
754 to map. Convert it into unformatted node: use
755 bh_result for the conversion */
756 set_block_dev_mapped (bh_result, allocated_block_nr, inode);
760 /* we have to padd file tail stored in direct item(s)
761 up to block size and convert it to unformatted
762 node. FIXME: this should also get into page cache */
766 * ugly, but we can only end the transaction if
769 if (th->t_refcount == 1) {
770 reiserfs_end_persistent_transaction(th);
774 retval = convert_tail_for_hole(inode, bh_result, tail_offset) ;
776 if ( retval != -ENOSPC )
777 reiserfs_warning (inode->i_sb, "clm-6004: convert tail failed inode %lu, error %d", inode->i_ino, retval) ;
778 if (allocated_block_nr) {
779 /* the bitmap, the super, and the stat data == 3 */
781 th = reiserfs_persistent_transaction(inode->i_sb,3);
783 reiserfs_free_block (th,inode,allocated_block_nr,1);
789 retval = direct2indirect (th, inode, &path, unbh, tail_offset);
791 reiserfs_unmap_buffer(unbh);
792 reiserfs_free_block (th, inode, allocated_block_nr, 1);
795 /* it is important the set_buffer_uptodate is done after
796 ** the direct2indirect. The buffer might contain valid
797 ** data newer than the data on disk (read by readpage, changed,
798 ** and then sent here by writepage). direct2indirect needs
799 ** to know if unbh was already up to date, so it can decide
800 ** if the data in unbh needs to be replaced with data from
803 set_buffer_uptodate (unbh);
805 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
806 buffer will disappear shortly, so it should not be added to
808 if ( unbh->b_page ) {
809 /* we've converted the tail, so we must
810 ** flush unbh before the transaction commits
812 reiserfs_add_tail_list(inode, unbh) ;
814 /* mark it dirty now to prevent commit_write from adding
815 ** this buffer to the inode's dirty buffer list
818 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
819 * It's still atomic, but it sets the page dirty too,
820 * which makes it eligible for writeback at any time by the
821 * VM (which was also the case with __mark_buffer_dirty())
823 mark_buffer_dirty(unbh) ;
826 /* append indirect item with holes if needed, when appending
827 pointer to 'block'-th block use block, which is already
829 struct cpu_key tmp_key;
830 unp_t unf_single=0; // We use this in case we need to allocate only
831 // one block which is a fastpath
833 __u64 max_to_insert=MAX_ITEM_LEN(inode->i_sb->s_blocksize)/UNFM_P_SIZE;
836 RFALSE( pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
837 "vs-804: invalid position for append");
838 /* indirect item has to be appended, set up key of that position */
839 make_cpu_key (&tmp_key, inode,
840 le_key_k_offset (version, &(ih->ih_key)) + op_bytes_number (ih, inode->i_sb->s_blocksize),
841 //pos_in_item * inode->i_sb->s_blocksize,
842 TYPE_INDIRECT, 3);// key type is unimportant
844 blocks_needed = 1 + ((cpu_key_k_offset (&key) - cpu_key_k_offset (&tmp_key)) >> inode->i_sb->s_blocksize_bits);
845 RFALSE( blocks_needed < 0, "green-805: invalid offset");
847 if ( blocks_needed == 1 ) {
850 un=kmalloc( min(blocks_needed,max_to_insert)*UNFM_P_SIZE,
851 GFP_ATOMIC); // We need to avoid scheduling.
857 memset(un, 0, UNFM_P_SIZE * min(blocks_needed,max_to_insert));
859 if ( blocks_needed <= max_to_insert) {
860 /* we are going to add target block to the file. Use allocated
862 un[blocks_needed-1] = cpu_to_le32 (allocated_block_nr);
863 set_block_dev_mapped (bh_result, allocated_block_nr, inode);
864 set_buffer_new(bh_result);
867 /* paste hole to the indirect item */
868 /* If kmalloc failed, max_to_insert becomes zero and it means we
869 only have space for one block */
870 blocks_needed=max_to_insert?max_to_insert:1;
872 retval = reiserfs_paste_into_item (th, &path, &tmp_key, inode, (char *)un, UNFM_P_SIZE * blocks_needed);
874 if (blocks_needed != 1)
878 reiserfs_free_block (th, inode, allocated_block_nr, 1);
882 /* We need to mark new file size in case this function will be
883 interrupted/aborted later on. And we may do this only for
885 inode->i_size += inode->i_sb->s_blocksize * blocks_needed;
892 /* this loop could log more blocks than we had originally asked
893 ** for. So, we have to allow the transaction to end if it is
894 ** too big or too full. Update the inode so things are
895 ** consistent if we crash before the function returns
897 ** release the path so that anybody waiting on the path before
898 ** ending their transaction will be able to continue.
900 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
901 restart_transaction(th, inode, &path) ;
903 /* inserting indirect pointers for a hole can take a
904 ** long time. reschedule if needed
908 retval = search_for_position_by_key (inode->i_sb, &key, &path);
909 if (retval == IO_ERROR) {
913 if (retval == POSITION_FOUND) {
914 reiserfs_warning (inode->i_sb, "vs-825: reiserfs_get_block: "
915 "%K should not be found", &key);
917 if (allocated_block_nr)
918 reiserfs_free_block (th, inode, allocated_block_nr, 1);
922 bh = get_last_bh (&path);
924 item = get_item (&path);
925 pos_in_item = path.pos_in_item;
933 reiserfs_update_sd(th, inode) ;
934 reiserfs_end_persistent_transaction(th);
936 reiserfs_write_unlock(inode->i_sb);
937 reiserfs_check_path(&path) ;
942 reiserfs_readpages(struct file *file, struct address_space *mapping,
943 struct list_head *pages, unsigned nr_pages)
945 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
948 /* Compute real number of used bytes by file
949 * Following three functions can go away when we'll have enough space in stat item
951 static int real_space_diff(struct inode *inode, int sd_size)
954 loff_t blocksize = inode->i_sb->s_blocksize ;
956 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
959 /* End of file is also in full block with indirect reference, so round
960 ** up to the next block.
962 ** there is just no way to know if the tail is actually packed
963 ** on the file, so we have to assume it isn't. When we pack the
964 ** tail, we add 4 bytes to pretend there really is an unformatted
967 bytes = ((inode->i_size + (blocksize-1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE + sd_size;
971 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
974 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
975 return inode->i_size + (loff_t)(real_space_diff(inode, sd_size)) ;
977 return ((loff_t)real_space_diff(inode, sd_size)) + (((loff_t)blocks) << 9);
980 /* Compute number of blocks used by file in ReiserFS counting */
981 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
983 loff_t bytes = inode_get_bytes(inode) ;
984 loff_t real_space = real_space_diff(inode, sd_size) ;
986 /* keeps fsck and non-quota versions of reiserfs happy */
987 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
988 bytes += (loff_t)511 ;
991 /* files from before the quota patch might i_blocks such that
992 ** bytes < real_space. Deal with that here to prevent it from
995 if (bytes < real_space)
997 return (bytes - real_space) >> 9;
1001 // BAD: new directories have stat data of new type and all other items
1002 // of old type. Version stored in the inode says about body items, so
1003 // in update_stat_data we can not rely on inode, but have to check
1004 // item version directly
1007 // called by read_locked_inode
1008 static void init_inode (struct inode * inode, struct path * path)
1010 struct buffer_head * bh;
1011 struct item_head * ih;
1013 //int version = ITEM_VERSION_1;
1015 bh = PATH_PLAST_BUFFER (path);
1016 ih = PATH_PITEM_HEAD (path);
1019 copy_key (INODE_PKEY (inode), &(ih->ih_key));
1020 inode->i_blksize = reiserfs_default_io_size;
1022 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list ));
1023 REISERFS_I(inode)->i_flags = 0;
1024 REISERFS_I(inode)->i_prealloc_block = 0;
1025 REISERFS_I(inode)->i_prealloc_count = 0;
1026 REISERFS_I(inode)->i_trans_id = 0;
1027 REISERFS_I(inode)->i_jl = NULL;
1028 REISERFS_I(inode)->i_acl_access = NULL;
1029 REISERFS_I(inode)->i_acl_default = NULL;
1030 init_rwsem (&REISERFS_I(inode)->xattr_sem);
1032 if (stat_data_v1 (ih)) {
1033 struct stat_data_v1 * sd = (struct stat_data_v1 *)B_I_PITEM (bh, ih);
1034 unsigned long blocks;
1036 set_inode_item_key_version (inode, KEY_FORMAT_3_5);
1037 set_inode_sd_version (inode, STAT_DATA_V1);
1038 inode->i_mode = sd_v1_mode(sd);
1039 inode->i_nlink = sd_v1_nlink(sd);
1040 inode->i_uid = sd_v1_uid(sd);
1041 inode->i_gid = sd_v1_gid(sd);
1042 inode->i_size = sd_v1_size(sd);
1043 inode->i_atime.tv_sec = sd_v1_atime(sd);
1044 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1045 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1046 inode->i_atime.tv_nsec = 0;
1047 inode->i_ctime.tv_nsec = 0;
1048 inode->i_mtime.tv_nsec = 0;
1050 inode->i_blocks = sd_v1_blocks(sd);
1051 inode->i_generation = le32_to_cpu (INODE_PKEY (inode)->k_dir_id);
1052 blocks = (inode->i_size + 511) >> 9;
1053 blocks = _ROUND_UP (blocks, inode->i_sb->s_blocksize >> 9);
1054 if (inode->i_blocks > blocks) {
1055 // there was a bug in <=3.5.23 when i_blocks could take negative
1056 // values. Starting from 3.5.17 this value could even be stored in
1057 // stat data. For such files we set i_blocks based on file
1058 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1059 // only updated if file's inode will ever change
1060 inode->i_blocks = blocks;
1063 rdev = sd_v1_rdev(sd);
1064 REISERFS_I(inode)->i_first_direct_byte = sd_v1_first_direct_byte(sd);
1065 /* an early bug in the quota code can give us an odd number for the
1066 ** block count. This is incorrect, fix it here.
1068 if (inode->i_blocks & 1) {
1071 inode_set_bytes(inode, to_real_used_space(inode, inode->i_blocks,
1073 /* nopack is initially zero for v1 objects. For v2 objects,
1074 nopack is initialised from sd_attrs */
1075 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1077 // new stat data found, but object may have old items
1078 // (directories and symlinks)
1079 struct stat_data * sd = (struct stat_data *)B_I_PITEM (bh, ih);
1081 inode->i_mode = sd_v2_mode(sd);
1082 inode->i_nlink = sd_v2_nlink(sd);
1083 inode->i_uid = sd_v2_uid(sd);
1084 inode->i_size = sd_v2_size(sd);
1085 inode->i_gid = sd_v2_gid(sd);
1086 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1087 inode->i_atime.tv_sec = sd_v2_atime(sd);
1088 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1089 inode->i_ctime.tv_nsec = 0;
1090 inode->i_mtime.tv_nsec = 0;
1091 inode->i_atime.tv_nsec = 0;
1092 inode->i_blocks = sd_v2_blocks(sd);
1093 rdev = sd_v2_rdev(sd);
1094 if( S_ISCHR( inode -> i_mode ) || S_ISBLK( inode -> i_mode ) )
1095 inode->i_generation = le32_to_cpu (INODE_PKEY (inode)->k_dir_id);
1097 inode->i_generation = sd_v2_generation(sd);
1099 if (S_ISDIR (inode->i_mode) || S_ISLNK (inode->i_mode))
1100 set_inode_item_key_version (inode, KEY_FORMAT_3_5);
1102 set_inode_item_key_version (inode, KEY_FORMAT_3_6);
1103 REISERFS_I(inode)->i_first_direct_byte = 0;
1104 set_inode_sd_version (inode, STAT_DATA_V2);
1105 inode_set_bytes(inode, to_real_used_space(inode, inode->i_blocks,
1107 /* read persistent inode attributes from sd and initalise
1108 generic inode flags from them */
1109 REISERFS_I(inode)->i_attrs = sd_v2_attrs( sd );
1110 sd_attrs_to_i_attrs( sd_v2_attrs( sd ), inode );
1114 if (S_ISREG (inode->i_mode)) {
1115 inode->i_op = &reiserfs_file_inode_operations;
1116 inode->i_fop = &reiserfs_file_operations;
1117 inode->i_mapping->a_ops = &reiserfs_address_space_operations ;
1118 } else if (S_ISDIR (inode->i_mode)) {
1119 inode->i_op = &reiserfs_dir_inode_operations;
1120 inode->i_fop = &reiserfs_dir_operations;
1121 } else if (S_ISLNK (inode->i_mode)) {
1122 inode->i_op = &reiserfs_symlink_inode_operations;
1123 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1125 inode->i_blocks = 0;
1126 inode->i_op = &reiserfs_special_inode_operations;
1127 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1132 // update new stat data with inode fields
1133 static void inode2sd (void * sd, struct inode * inode, loff_t size)
1135 struct stat_data * sd_v2 = (struct stat_data *)sd;
1138 set_sd_v2_mode(sd_v2, inode->i_mode );
1139 set_sd_v2_nlink(sd_v2, inode->i_nlink );
1140 set_sd_v2_uid(sd_v2, inode->i_uid );
1141 set_sd_v2_size(sd_v2, size );
1142 set_sd_v2_gid(sd_v2, inode->i_gid );
1143 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec );
1144 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec );
1145 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec );
1146 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1147 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1148 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1150 set_sd_v2_generation(sd_v2, inode->i_generation);
1151 flags = REISERFS_I(inode)->i_attrs;
1152 i_attrs_to_sd_attrs( inode, &flags );
1153 set_sd_v2_attrs( sd_v2, flags );
1157 // used to copy inode's fields to old stat data
1158 static void inode2sd_v1 (void * sd, struct inode * inode, loff_t size)
1160 struct stat_data_v1 * sd_v1 = (struct stat_data_v1 *)sd;
1162 set_sd_v1_mode(sd_v1, inode->i_mode );
1163 set_sd_v1_uid(sd_v1, inode->i_uid );
1164 set_sd_v1_gid(sd_v1, inode->i_gid );
1165 set_sd_v1_nlink(sd_v1, inode->i_nlink );
1166 set_sd_v1_size(sd_v1, size );
1167 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec );
1168 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec );
1169 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec );
1171 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1172 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1174 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1176 // Sigh. i_first_direct_byte is back
1177 set_sd_v1_first_direct_byte(sd_v1, REISERFS_I(inode)->i_first_direct_byte);
1181 /* NOTE, you must prepare the buffer head before sending it here,
1182 ** and then log it after the call
1184 static void update_stat_data (struct path * path, struct inode * inode,
1187 struct buffer_head * bh;
1188 struct item_head * ih;
1190 bh = PATH_PLAST_BUFFER (path);
1191 ih = PATH_PITEM_HEAD (path);
1193 if (!is_statdata_le_ih (ih))
1194 reiserfs_panic (inode->i_sb, "vs-13065: update_stat_data: key %k, found item %h",
1195 INODE_PKEY (inode), ih);
1197 if (stat_data_v1 (ih)) {
1198 // path points to old stat data
1199 inode2sd_v1 (B_I_PITEM (bh, ih), inode, size);
1201 inode2sd (B_I_PITEM (bh, ih), inode, size);
1208 void reiserfs_update_sd_size (struct reiserfs_transaction_handle *th,
1209 struct inode * inode, loff_t size)
1212 INITIALIZE_PATH(path);
1213 struct buffer_head *bh ;
1215 struct item_head *ih, tmp_ih ;
1218 make_cpu_key (&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3);//key type is unimportant
1222 /* look for the object's stat data */
1223 retval = search_item (inode->i_sb, &key, &path);
1224 if (retval == IO_ERROR) {
1225 reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: "
1226 "i/o failure occurred trying to update %K stat data",
1230 if (retval == ITEM_NOT_FOUND) {
1231 pos = PATH_LAST_POSITION (&path);
1233 if (inode->i_nlink == 0) {
1234 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found");*/
1237 reiserfs_warning (inode->i_sb, "vs-13060: reiserfs_update_sd: "
1238 "stat data of object %k (nlink == %d) not found (pos %d)",
1239 INODE_PKEY (inode), inode->i_nlink, pos);
1240 reiserfs_check_path(&path) ;
1244 /* sigh, prepare_for_journal might schedule. When it schedules the
1245 ** FS might change. We have to detect that, and loop back to the
1246 ** search if the stat data item has moved
1248 bh = get_last_bh(&path) ;
1249 ih = get_ih(&path) ;
1250 copy_item_head (&tmp_ih, ih);
1251 fs_gen = get_generation (inode->i_sb);
1252 reiserfs_prepare_for_journal(inode->i_sb, bh, 1) ;
1253 if (fs_changed (fs_gen, inode->i_sb) && item_moved(&tmp_ih, &path)) {
1254 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
1255 continue ; /* Stat_data item has been moved after scheduling. */
1259 update_stat_data (&path, inode, size);
1260 journal_mark_dirty(th, th->t_super, bh) ;
1265 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1266 ** does a make_bad_inode when things go wrong. But, we need to make sure
1267 ** and clear the key in the private portion of the inode, otherwise a
1268 ** corresponding iput might try to delete whatever object the inode last
1271 static void reiserfs_make_bad_inode(struct inode *inode) {
1272 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1273 make_bad_inode(inode);
1277 // initially this function was derived from minix or ext2's analog and
1278 // evolved as the prototype did
1281 int reiserfs_init_locked_inode (struct inode * inode, void *p)
1283 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p ;
1284 inode->i_ino = args->objectid;
1285 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1289 /* looks for stat data in the tree, and fills up the fields of in-core
1290 inode stat data fields */
1291 void reiserfs_read_locked_inode (struct inode * inode, struct reiserfs_iget_args *args)
1293 INITIALIZE_PATH (path_to_sd);
1295 unsigned long dirino;
1298 dirino = args->dirid ;
1300 /* set version 1, version 2 could be used too, because stat data
1301 key is the same in both versions */
1302 key.version = KEY_FORMAT_3_5;
1303 key.on_disk_key.k_dir_id = dirino;
1304 key.on_disk_key.k_objectid = inode->i_ino;
1305 key.on_disk_key.u.k_offset_v1.k_offset = SD_OFFSET;
1306 key.on_disk_key.u.k_offset_v1.k_uniqueness = SD_UNIQUENESS;
1308 /* look for the object's stat data */
1309 retval = search_item (inode->i_sb, &key, &path_to_sd);
1310 if (retval == IO_ERROR) {
1311 reiserfs_warning (inode->i_sb, "vs-13070: reiserfs_read_locked_inode: "
1312 "i/o failure occurred trying to find stat data of %K",
1314 reiserfs_make_bad_inode(inode) ;
1317 if (retval != ITEM_FOUND) {
1318 /* a stale NFS handle can trigger this without it being an error */
1319 pathrelse (&path_to_sd);
1320 reiserfs_make_bad_inode(inode) ;
1325 init_inode (inode, &path_to_sd);
1327 /* It is possible that knfsd is trying to access inode of a file
1328 that is being removed from the disk by some other thread. As we
1329 update sd on unlink all that is required is to check for nlink
1330 here. This bug was first found by Sizif when debugging
1331 SquidNG/Butterfly, forgotten, and found again after Philippe
1332 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1334 More logical fix would require changes in fs/inode.c:iput() to
1335 remove inode from hash-table _after_ fs cleaned disk stuff up and
1336 in iget() to return NULL if I_FREEING inode is found in
1338 /* Currently there is one place where it's ok to meet inode with
1339 nlink==0: processing of open-unlinked and half-truncated files
1340 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1341 if( ( inode -> i_nlink == 0 ) &&
1342 ! REISERFS_SB(inode -> i_sb) -> s_is_unlinked_ok ) {
1343 reiserfs_warning (inode->i_sb,
1344 "vs-13075: reiserfs_read_locked_inode: "
1345 "dead inode read from disk %K. "
1346 "This is likely to be race with knfsd. Ignore",
1348 reiserfs_make_bad_inode( inode );
1351 reiserfs_check_path(&path_to_sd) ; /* init inode should be relsing */
1356 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1358 * @inode: inode from hash table to check
1359 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1361 * This function is called by iget5_locked() to distinguish reiserfs inodes
1362 * having the same inode numbers. Such inodes can only exist due to some
1363 * error condition. One of them should be bad. Inodes with identical
1364 * inode numbers (objectids) are distinguished by parent directory ids.
1367 int reiserfs_find_actor( struct inode *inode, void *opaque )
1369 struct reiserfs_iget_args *args;
1372 /* args is already in CPU order */
1373 return (inode->i_ino == args->objectid) &&
1374 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1377 struct inode * reiserfs_iget (struct super_block * s, const struct cpu_key * key)
1379 struct inode * inode;
1380 struct reiserfs_iget_args args ;
1382 args.objectid = key->on_disk_key.k_objectid ;
1383 args.dirid = key->on_disk_key.k_dir_id ;
1384 inode = iget5_locked (s, key->on_disk_key.k_objectid,
1385 reiserfs_find_actor, reiserfs_init_locked_inode, (void *)(&args));
1387 return ERR_PTR(-ENOMEM) ;
1389 if (inode->i_state & I_NEW) {
1390 reiserfs_read_locked_inode(inode, &args);
1391 unlock_new_inode(inode);
1394 if (comp_short_keys (INODE_PKEY (inode), key) || is_bad_inode (inode)) {
1395 /* either due to i/o error or a stale NFS handle */
1402 struct dentry *reiserfs_get_dentry(struct super_block *sb, void *vobjp)
1404 __u32 *data = vobjp;
1405 struct cpu_key key ;
1406 struct dentry *result;
1407 struct inode *inode;
1409 key.on_disk_key.k_objectid = data[0] ;
1410 key.on_disk_key.k_dir_id = data[1] ;
1411 inode = reiserfs_iget(sb, &key) ;
1412 if (inode && !IS_ERR(inode) && data[2] != 0 &&
1413 data[2] != inode->i_generation) {
1418 inode = ERR_PTR(-ESTALE);
1420 return ERR_PTR(PTR_ERR(inode));
1421 result = d_alloc_anon(inode);
1424 return ERR_PTR(-ENOMEM);
1429 struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 *data,
1430 int len, int fhtype,
1431 int (*acceptable)(void *contect, struct dentry *de),
1433 __u32 obj[3], parent[3];
1435 /* fhtype happens to reflect the number of u32s encoded.
1436 * due to a bug in earlier code, fhtype might indicate there
1437 * are more u32s then actually fitted.
1438 * so if fhtype seems to be more than len, reduce fhtype.
1440 * 2 - objectid + dir_id - legacy support
1441 * 3 - objectid + dir_id + generation
1442 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1443 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1444 * 6 - as above plus generation of directory
1445 * 6 does not fit in NFSv2 handles
1448 if (fhtype != 6 || len != 5)
1449 reiserfs_warning (sb, "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1456 if (fhtype == 3 || fhtype >= 5)
1458 else obj[2] = 0; /* generation number */
1461 parent[0] = data[fhtype>=5?3:2] ;
1462 parent[1] = data[fhtype>=5?4:3] ;
1464 parent[2] = data[5];
1467 return sb->s_export_op->find_exported_dentry(sb, obj, fhtype < 4 ? NULL : parent,
1468 acceptable, context);
1471 int reiserfs_encode_fh(struct dentry *dentry, __u32 *data, int *lenp, int need_parent) {
1472 struct inode *inode = dentry->d_inode ;
1478 data[0] = inode->i_ino ;
1479 data[1] = le32_to_cpu(INODE_PKEY (inode)->k_dir_id) ;
1480 data[2] = inode->i_generation ;
1482 /* no room for directory info? return what we've stored so far */
1483 if (maxlen < 5 || ! need_parent)
1486 spin_lock(&dentry->d_lock);
1487 inode = dentry->d_parent->d_inode ;
1488 data[3] = inode->i_ino ;
1489 data[4] = le32_to_cpu(INODE_PKEY (inode)->k_dir_id) ;
1492 data[5] = inode->i_generation ;
1495 spin_unlock(&dentry->d_lock);
1500 /* looks for stat data, then copies fields to it, marks the buffer
1501 containing stat data as dirty */
1502 /* reiserfs inodes are never really dirty, since the dirty inode call
1503 ** always logs them. This call allows the VFS inode marking routines
1504 ** to properly mark inodes for datasync and such, but only actually
1505 ** does something when called for a synchronous update.
1507 void reiserfs_write_inode (struct inode * inode, int do_sync) {
1508 struct reiserfs_transaction_handle th ;
1509 int jbegin_count = 1 ;
1511 if (inode->i_sb->s_flags & MS_RDONLY) {
1512 reiserfs_warning (inode->i_sb,
1513 "clm-6005: writing inode %lu on readonly FS",
1517 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1518 ** these cases are just when the system needs ram, not when the
1519 ** inode needs to reach disk for safety, and they can safely be
1520 ** ignored because the altered inode has already been logged.
1522 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1523 reiserfs_write_lock(inode->i_sb);
1524 journal_begin(&th, inode->i_sb, jbegin_count) ;
1525 reiserfs_update_sd (&th, inode);
1526 journal_end_sync(&th, inode->i_sb, jbegin_count) ;
1527 reiserfs_write_unlock(inode->i_sb);
1531 /* FIXME: no need any more. right? */
1532 int reiserfs_sync_inode (struct reiserfs_transaction_handle *th, struct inode * inode)
1536 reiserfs_update_sd (th, inode);
1541 /* stat data of new object is inserted already, this inserts the item
1542 containing "." and ".." entries */
1543 static int reiserfs_new_directory (struct reiserfs_transaction_handle *th,
1544 struct inode *inode,
1545 struct item_head * ih, struct path * path,
1548 struct super_block * sb = th->t_super;
1549 char empty_dir [EMPTY_DIR_SIZE];
1550 char * body = empty_dir;
1554 _make_cpu_key (&key, KEY_FORMAT_3_5, le32_to_cpu (ih->ih_key.k_dir_id),
1555 le32_to_cpu (ih->ih_key.k_objectid), DOT_OFFSET, TYPE_DIRENTRY, 3/*key length*/);
1557 /* compose item head for new item. Directories consist of items of
1558 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1559 is done by reiserfs_new_inode */
1560 if (old_format_only (sb)) {
1561 make_le_item_head (ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET, TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1563 make_empty_dir_item_v1 (body, ih->ih_key.k_dir_id, ih->ih_key.k_objectid,
1564 INODE_PKEY (dir)->k_dir_id,
1565 INODE_PKEY (dir)->k_objectid );
1567 make_le_item_head (ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET, TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1569 make_empty_dir_item (body, ih->ih_key.k_dir_id, ih->ih_key.k_objectid,
1570 INODE_PKEY (dir)->k_dir_id,
1571 INODE_PKEY (dir)->k_objectid );
1574 /* look for place in the tree for new item */
1575 retval = search_item (sb, &key, path);
1576 if (retval == IO_ERROR) {
1577 reiserfs_warning (sb, "vs-13080: reiserfs_new_directory: "
1578 "i/o failure occurred creating new directory");
1581 if (retval == ITEM_FOUND) {
1583 reiserfs_warning (sb, "vs-13070: reiserfs_new_directory: "
1584 "object with this key exists (%k)", &(ih->ih_key));
1588 /* insert item, that is empty directory item */
1589 return reiserfs_insert_item (th, path, &key, ih, inode, body);
1593 /* stat data of object has been inserted, this inserts the item
1594 containing the body of symlink */
1595 static int reiserfs_new_symlink (struct reiserfs_transaction_handle *th,
1596 struct inode *inode, /* Inode of symlink */
1597 struct item_head * ih,
1598 struct path * path, const char * symname, int item_len)
1600 struct super_block * sb = th->t_super;
1604 _make_cpu_key (&key, KEY_FORMAT_3_5,
1605 le32_to_cpu (ih->ih_key.k_dir_id),
1606 le32_to_cpu (ih->ih_key.k_objectid),
1607 1, TYPE_DIRECT, 3/*key length*/);
1609 make_le_item_head (ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len, 0/*free_space*/);
1611 /* look for place in the tree for new item */
1612 retval = search_item (sb, &key, path);
1613 if (retval == IO_ERROR) {
1614 reiserfs_warning (sb, "vs-13080: reiserfs_new_symlinik: "
1615 "i/o failure occurred creating new symlink");
1618 if (retval == ITEM_FOUND) {
1620 reiserfs_warning (sb, "vs-13080: reiserfs_new_symlink: "
1621 "object with this key exists (%k)", &(ih->ih_key));
1625 /* insert item, that is body of symlink */
1626 return reiserfs_insert_item (th, path, &key, ih, inode, symname);
1630 /* inserts the stat data into the tree, and then calls
1631 reiserfs_new_directory (to insert ".", ".." item if new object is
1632 directory) or reiserfs_new_symlink (to insert symlink body if new
1633 object is symlink) or nothing (if new object is regular file)
1635 NOTE! uid and gid must already be set in the inode. If we return
1636 non-zero due to an error, we have to drop the quota previously allocated
1637 for the fresh inode. This can only be done outside a transaction, so
1638 if we return non-zero, we also end the transaction. */
1639 int reiserfs_new_inode (struct reiserfs_transaction_handle *th,
1640 struct inode * dir, int mode,
1641 const char * symname,
1642 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1643 strlen (symname) for symlinks)*/
1644 loff_t i_size, struct dentry *dentry,
1645 struct inode *inode)
1647 struct super_block * sb;
1648 INITIALIZE_PATH (path_to_key);
1650 struct item_head ih;
1651 struct stat_data sd;
1655 if (!dir || !dir->i_nlink) {
1662 /* item head of new item */
1663 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1664 ih.ih_key.k_objectid = cpu_to_le32 (reiserfs_get_unused_objectid (th));
1665 if (!ih.ih_key.k_objectid) {
1667 goto out_bad_inode ;
1669 if (old_format_only (sb))
1670 /* not a perfect generation count, as object ids can be reused, but
1671 ** this is as good as reiserfs can do right now.
1672 ** note that the private part of inode isn't filled in yet, we have
1673 ** to use the directory.
1675 inode->i_generation = le32_to_cpu (INODE_PKEY (dir)->k_objectid);
1677 #if defined( USE_INODE_GENERATION_COUNTER )
1678 inode->i_generation = le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1680 inode->i_generation = ++event;
1683 /* fill stat data */
1684 inode->i_nlink = (S_ISDIR (mode) ? 2 : 1);
1686 /* uid and gid must already be set by the caller for quota init */
1688 /* symlink cannot be immutable or append only, right? */
1689 if( S_ISLNK( inode -> i_mode ) )
1690 inode -> i_flags &= ~ ( S_IMMUTABLE | S_APPEND );
1692 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1693 inode->i_size = i_size;
1694 inode->i_blocks = 0;
1696 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1697 U32_MAX/*NO_BYTES_IN_DIRECT_ITEM*/;
1699 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list ));
1700 REISERFS_I(inode)->i_flags = 0;
1701 REISERFS_I(inode)->i_prealloc_block = 0;
1702 REISERFS_I(inode)->i_prealloc_count = 0;
1703 REISERFS_I(inode)->i_trans_id = 0;
1704 REISERFS_I(inode)->i_jl = NULL;
1705 REISERFS_I(inode)->i_attrs =
1706 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1707 sd_attrs_to_i_attrs( REISERFS_I(inode) -> i_attrs, inode );
1708 REISERFS_I(inode)->i_acl_access = NULL;
1709 REISERFS_I(inode)->i_acl_default = NULL;
1710 init_rwsem (&REISERFS_I(inode)->xattr_sem);
1712 if (old_format_only (sb))
1713 make_le_item_head (&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET, TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1715 make_le_item_head (&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET, TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1717 /* key to search for correct place for new stat data */
1718 _make_cpu_key (&key, KEY_FORMAT_3_6, le32_to_cpu (ih.ih_key.k_dir_id),
1719 le32_to_cpu (ih.ih_key.k_objectid), SD_OFFSET, TYPE_STAT_DATA, 3/*key length*/);
1721 /* find proper place for inserting of stat data */
1722 retval = search_item (sb, &key, &path_to_key);
1723 if (retval == IO_ERROR) {
1727 if (retval == ITEM_FOUND) {
1728 pathrelse (&path_to_key);
1732 if (old_format_only (sb)) {
1733 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1734 pathrelse (&path_to_key);
1735 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1739 inode2sd_v1 (&sd, inode, inode->i_size);
1741 inode2sd (&sd, inode, inode->i_size);
1743 // these do not go to on-disk stat data
1744 inode->i_ino = le32_to_cpu (ih.ih_key.k_objectid);
1745 inode->i_blksize = reiserfs_default_io_size;
1747 // store in in-core inode the key of stat data and version all
1748 // object items will have (directory items will have old offset
1749 // format, other new objects will consist of new items)
1750 memcpy (INODE_PKEY (inode), &(ih.ih_key), KEY_SIZE);
1751 if (old_format_only (sb) || S_ISDIR(mode) || S_ISLNK(mode))
1752 set_inode_item_key_version (inode, KEY_FORMAT_3_5);
1754 set_inode_item_key_version (inode, KEY_FORMAT_3_6);
1755 if (old_format_only (sb))
1756 set_inode_sd_version (inode, STAT_DATA_V1);
1758 set_inode_sd_version (inode, STAT_DATA_V2);
1760 /* insert the stat data into the tree */
1761 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1762 if (REISERFS_I(dir)->new_packing_locality)
1763 th->displace_new_blocks = 1;
1765 retval = reiserfs_insert_item (th, &path_to_key, &key, &ih, inode, (char *)(&sd));
1768 reiserfs_check_path(&path_to_key) ;
1772 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1773 if (!th->displace_new_blocks)
1774 REISERFS_I(dir)->new_packing_locality = 0;
1776 if (S_ISDIR(mode)) {
1777 /* insert item with "." and ".." */
1778 retval = reiserfs_new_directory (th, inode, &ih, &path_to_key, dir);
1781 if (S_ISLNK(mode)) {
1782 /* insert body of symlink */
1783 if (!old_format_only (sb))
1784 i_size = ROUND_UP(i_size);
1785 retval = reiserfs_new_symlink (th, inode, &ih, &path_to_key, symname, i_size);
1789 reiserfs_check_path(&path_to_key) ;
1790 journal_end(th, th->t_super, th->t_blocks_allocated);
1791 goto out_inserted_sd;
1794 /* XXX CHECK THIS */
1795 if (reiserfs_posixacl (inode->i_sb)) {
1796 retval = reiserfs_inherit_default_acl (dir, dentry, inode);
1799 reiserfs_check_path(&path_to_key) ;
1800 journal_end(th, th->t_super, th->t_blocks_allocated);
1801 goto out_inserted_sd;
1803 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1804 reiserfs_warning (inode->i_sb, "ACLs aren't enabled in the fs, "
1805 "but vfs thinks they are!");
1808 insert_inode_hash (inode);
1809 reiserfs_update_sd(th, inode);
1810 reiserfs_check_path(&path_to_key) ;
1814 /* it looks like you can easily compress these two goto targets into
1815 * one. Keeping it like this doesn't actually hurt anything, and they
1816 * are place holders for what the quota code actually needs.
1819 /* Invalidate the object, nothing was inserted yet */
1820 INODE_PKEY(inode)->k_objectid = 0;
1822 /* dquot_drop must be done outside a transaction */
1823 journal_end(th, th->t_super, th->t_blocks_allocated) ;
1824 DQUOT_FREE_INODE(inode);
1826 inode->i_flags |= S_NOQUOTA;
1827 make_bad_inode(inode);
1831 th->t_trans_id = 0; /* so the caller can't use this handle later */
1837 ** finds the tail page in the page cache,
1838 ** reads the last block in.
1840 ** On success, page_result is set to a locked, pinned page, and bh_result
1841 ** is set to an up to date buffer for the last block in the file. returns 0.
1843 ** tail conversion is not done, so bh_result might not be valid for writing
1844 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1845 ** trying to write the block.
1847 ** on failure, nonzero is returned, page_result and bh_result are untouched.
1849 static int grab_tail_page(struct inode *p_s_inode,
1850 struct page **page_result,
1851 struct buffer_head **bh_result) {
1853 /* we want the page with the last byte in the file,
1854 ** not the page that will hold the next byte for appending
1856 unsigned long index = (p_s_inode->i_size-1) >> PAGE_CACHE_SHIFT ;
1857 unsigned long pos = 0 ;
1858 unsigned long start = 0 ;
1859 unsigned long blocksize = p_s_inode->i_sb->s_blocksize ;
1860 unsigned long offset = (p_s_inode->i_size) & (PAGE_CACHE_SIZE - 1) ;
1861 struct buffer_head *bh ;
1862 struct buffer_head *head ;
1863 struct page * page ;
1866 /* we know that we are only called with inode->i_size > 0.
1867 ** we also know that a file tail can never be as big as a block
1868 ** If i_size % blocksize == 0, our file is currently block aligned
1869 ** and it won't need converting or zeroing after a truncate.
1871 if ((offset & (blocksize - 1)) == 0) {
1874 page = grab_cache_page(p_s_inode->i_mapping, index) ;
1879 /* start within the page of the last block in the file */
1880 start = (offset / blocksize) * blocksize ;
1882 error = block_prepare_write(page, start, offset,
1883 reiserfs_get_block_create_0) ;
1887 head = page_buffers(page) ;
1893 bh = bh->b_this_page ;
1895 } while(bh != head) ;
1897 if (!buffer_uptodate(bh)) {
1898 /* note, this should never happen, prepare_write should
1899 ** be taking care of this for us. If the buffer isn't up to date,
1900 ** I've screwed up the code to find the buffer, or the code to
1901 ** call prepare_write
1903 reiserfs_warning (p_s_inode->i_sb,
1904 "clm-6000: error reading block %lu on dev %s",
1906 reiserfs_bdevname (p_s_inode->i_sb)) ;
1911 *page_result = page ;
1918 page_cache_release(page) ;
1923 ** vfs version of truncate file. Must NOT be called with
1924 ** a transaction already started.
1926 ** some code taken from block_truncate_page
1928 void reiserfs_truncate_file(struct inode *p_s_inode, int update_timestamps) {
1929 struct reiserfs_transaction_handle th ;
1930 /* we want the offset for the first byte after the end of the file */
1931 unsigned long offset = p_s_inode->i_size & (PAGE_CACHE_SIZE - 1) ;
1932 unsigned blocksize = p_s_inode->i_sb->s_blocksize ;
1934 struct page *page = NULL ;
1936 struct buffer_head *bh = NULL ;
1938 reiserfs_write_lock(p_s_inode->i_sb);
1940 if (p_s_inode->i_size > 0) {
1941 if ((error = grab_tail_page(p_s_inode, &page, &bh))) {
1942 // -ENOENT means we truncated past the end of the file,
1943 // and get_block_create_0 could not find a block to read in,
1945 if (error != -ENOENT)
1946 reiserfs_warning (p_s_inode->i_sb,
1947 "clm-6001: grab_tail_page failed %d",
1954 /* so, if page != NULL, we have a buffer head for the offset at
1955 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
1956 ** then we have an unformatted node. Otherwise, we have a direct item,
1957 ** and no zeroing is required on disk. We zero after the truncate,
1958 ** because the truncate might pack the item anyway
1959 ** (it will unmap bh if it packs).
1961 /* it is enough to reserve space in transaction for 2 balancings:
1962 one for "save" link adding and another for the first
1963 cut_from_item. 1 is for update_sd */
1964 journal_begin(&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1 ) ;
1965 reiserfs_update_inode_transaction(p_s_inode) ;
1966 if (update_timestamps)
1967 /* we are doing real truncate: if the system crashes before the last
1968 transaction of truncating gets committed - on reboot the file
1969 either appears truncated properly or not truncated at all */
1970 add_save_link (&th, p_s_inode, 1);
1971 reiserfs_do_truncate (&th, p_s_inode, page, update_timestamps) ;
1972 journal_end(&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1 ) ;
1974 if (update_timestamps)
1975 remove_save_link (p_s_inode, 1/* truncate */);
1978 length = offset & (blocksize - 1) ;
1979 /* if we are not on a block boundary */
1983 length = blocksize - length ;
1984 kaddr = kmap_atomic(page, KM_USER0) ;
1985 memset(kaddr + offset, 0, length) ;
1986 flush_dcache_page(page) ;
1987 kunmap_atomic(kaddr, KM_USER0) ;
1988 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
1989 mark_buffer_dirty(bh) ;
1993 page_cache_release(page) ;
1996 reiserfs_write_unlock(p_s_inode->i_sb);
1999 static int map_block_for_writepage(struct inode *inode,
2000 struct buffer_head *bh_result,
2001 unsigned long block) {
2002 struct reiserfs_transaction_handle th ;
2004 struct item_head tmp_ih ;
2005 struct item_head *ih ;
2006 struct buffer_head *bh ;
2008 struct cpu_key key ;
2009 INITIALIZE_PATH(path) ;
2011 int jbegin_count = JOURNAL_PER_BALANCE_CNT ;
2012 loff_t byte_offset = (block << inode->i_sb->s_blocksize_bits) + 1 ;
2014 int use_get_block = 0 ;
2015 int bytes_copied = 0 ;
2017 int trans_running = 0;
2019 /* catch places below that try to log something without starting a trans */
2022 if (!buffer_uptodate(bh_result)) {
2026 kmap(bh_result->b_page) ;
2028 reiserfs_write_lock(inode->i_sb);
2029 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3) ;
2032 retval = search_for_position_by_key(inode->i_sb, &key, &path) ;
2033 if (retval != POSITION_FOUND) {
2038 bh = get_last_bh(&path) ;
2039 ih = get_ih(&path) ;
2040 item = get_item(&path) ;
2041 pos_in_item = path.pos_in_item ;
2043 /* we've found an unformatted node */
2044 if (indirect_item_found(retval, ih)) {
2045 if (bytes_copied > 0) {
2046 reiserfs_warning (inode->i_sb, "clm-6002: bytes_copied %d",
2049 if (!get_block_num(item, pos_in_item)) {
2050 /* crap, we are writing to a hole */
2054 set_block_dev_mapped(bh_result, get_block_num(item,pos_in_item),inode);
2055 } else if (is_direct_le_ih(ih)) {
2057 p = page_address(bh_result->b_page) ;
2058 p += (byte_offset -1) & (PAGE_CACHE_SIZE - 1) ;
2059 copy_size = ih_item_len(ih) - pos_in_item;
2061 fs_gen = get_generation(inode->i_sb) ;
2062 copy_item_head(&tmp_ih, ih) ;
2064 if (!trans_running) {
2065 /* vs-3050 is gone, no need to drop the path */
2066 journal_begin(&th, inode->i_sb, jbegin_count) ;
2067 reiserfs_update_inode_transaction(inode) ;
2069 if (fs_changed(fs_gen, inode->i_sb) && item_moved(&tmp_ih, &path)) {
2070 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
2075 reiserfs_prepare_for_journal(inode->i_sb, bh, 1) ;
2077 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
2078 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
2082 memcpy( B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied, copy_size) ;
2084 journal_mark_dirty(&th, inode->i_sb, bh) ;
2085 bytes_copied += copy_size ;
2086 set_block_dev_mapped(bh_result, 0, inode);
2088 /* are there still bytes left? */
2089 if (bytes_copied < bh_result->b_size &&
2090 (byte_offset + bytes_copied) < inode->i_size) {
2091 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + copy_size) ;
2095 reiserfs_warning (inode->i_sb,
2096 "clm-6003: bad item inode %lu, device %s",
2097 inode->i_ino, reiserfs_bdevname (inode->i_sb)) ;
2105 if (trans_running) {
2106 journal_end(&th, inode->i_sb, jbegin_count) ;
2109 reiserfs_write_unlock(inode->i_sb);
2111 /* this is where we fill in holes in the file. */
2112 if (use_get_block) {
2113 retval = reiserfs_get_block(inode, block, bh_result,
2114 GET_BLOCK_CREATE | GET_BLOCK_NO_ISEM |
2115 GET_BLOCK_NO_DANGLE);
2117 if (!buffer_mapped(bh_result) || bh_result->b_blocknr == 0) {
2118 /* get_block failed to find a mapped unformatted node. */
2124 kunmap(bh_result->b_page) ;
2126 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2127 /* we've copied data from the page into the direct item, so the
2128 * buffer in the page is now clean, mark it to reflect that.
2130 lock_buffer(bh_result);
2131 clear_buffer_dirty(bh_result);
2132 unlock_buffer(bh_result);
2138 * mason@suse.com: updated in 2.5.54 to follow the same general io
2139 * start/recovery path as __block_write_full_page, along with special
2140 * code to handle reiserfs tails.
2142 static int reiserfs_write_full_page(struct page *page, struct writeback_control *wbc) {
2143 struct inode *inode = page->mapping->host ;
2144 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT ;
2146 unsigned long block ;
2147 struct buffer_head *head, *bh;
2150 int checked = PageChecked(page);
2151 struct reiserfs_transaction_handle th;
2152 struct super_block *s = inode->i_sb;
2153 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2156 /* The page dirty bit is cleared before writepage is called, which
2157 * means we have to tell create_empty_buffers to make dirty buffers
2158 * The page really should be up to date at this point, so tossing
2159 * in the BH_Uptodate is just a sanity check.
2161 if (!page_has_buffers(page)) {
2162 create_empty_buffers(page, s->s_blocksize,
2163 (1 << BH_Dirty) | (1 << BH_Uptodate));
2165 head = page_buffers(page) ;
2167 /* last page in the file, zero out any contents past the
2168 ** last byte in the file
2170 if (page->index >= end_index) {
2172 unsigned last_offset;
2174 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1) ;
2175 /* no file contents in this page */
2176 if (page->index >= end_index + 1 || !last_offset) {
2180 kaddr = kmap_atomic(page, KM_USER0);
2181 memset(kaddr + last_offset, 0, PAGE_CACHE_SIZE-last_offset) ;
2182 flush_dcache_page(page) ;
2183 kunmap_atomic(kaddr, KM_USER0) ;
2186 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits) ;
2187 /* first map all the buffers, logging any direct items we find */
2189 if ((checked || buffer_dirty(bh)) && (!buffer_mapped(bh) ||
2190 (buffer_mapped(bh) && bh->b_blocknr == 0))) {
2191 /* not mapped yet, or it points to a direct item, search
2192 * the btree for the mapping info, and log any direct
2195 if ((error = map_block_for_writepage(inode, bh, block))) {
2199 bh = bh->b_this_page;
2201 } while(bh != head) ;
2204 * we start the transaction after map_block_for_writepage,
2205 * because it can create holes in the file (an unbounded operation).
2206 * starting it here, we can make a reliable estimate for how many
2207 * blocks we're going to log
2210 ClearPageChecked(page);
2211 reiserfs_write_lock(s);
2212 journal_begin(&th, s, bh_per_page + 1);
2213 reiserfs_update_inode_transaction(inode);
2215 /* now go through and lock any dirty buffers on the page */
2218 if (!buffer_mapped(bh))
2220 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2224 reiserfs_prepare_for_journal(s, bh, 1);
2225 journal_mark_dirty(&th, s, bh);
2228 /* from this point on, we know the buffer is mapped to a
2229 * real block and not a direct item
2231 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2234 if (test_set_buffer_locked(bh)) {
2235 redirty_page_for_writepage(wbc, page);
2239 if (test_clear_buffer_dirty(bh)) {
2240 mark_buffer_async_write(bh);
2244 } while((bh = bh->b_this_page) != head);
2247 journal_end(&th, s, bh_per_page + 1);
2248 reiserfs_write_unlock(s);
2250 BUG_ON(PageWriteback(page));
2251 set_page_writeback(page);
2255 * since any buffer might be the only dirty buffer on the page,
2256 * the first submit_bh can bring the page out of writeback.
2257 * be careful with the buffers.
2260 struct buffer_head *next = bh->b_this_page;
2261 if (buffer_async_write(bh)) {
2262 submit_bh(WRITE, bh);
2267 } while(bh != head);
2273 * if this page only had a direct item, it is very possible for
2274 * no io to be required without there being an error. Or,
2275 * someone else could have locked them and sent them down the
2276 * pipe without locking the page
2280 if (!buffer_uptodate(bh)) {
2284 bh = bh->b_this_page;
2285 } while(bh != head);
2287 SetPageUptodate(page);
2288 end_page_writeback(page);
2293 /* catches various errors, we need to make sure any valid dirty blocks
2294 * get to the media. The page is currently locked and not marked for
2297 ClearPageUptodate(page);
2301 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2303 mark_buffer_async_write(bh);
2306 * clear any dirty bits that might have come from getting
2307 * attached to a dirty page
2309 clear_buffer_dirty(bh);
2311 bh = bh->b_this_page;
2312 } while(bh != head);
2314 BUG_ON(PageWriteback(page));
2315 set_page_writeback(page);
2318 struct buffer_head *next = bh->b_this_page;
2319 if (buffer_async_write(bh)) {
2320 clear_buffer_dirty(bh);
2321 submit_bh(WRITE, bh);
2326 } while(bh != head);
2331 static int reiserfs_readpage (struct file *f, struct page * page)
2333 return block_read_full_page (page, reiserfs_get_block);
2337 static int reiserfs_writepage (struct page * page, struct writeback_control *wbc)
2339 struct inode *inode = page->mapping->host ;
2340 reiserfs_wait_on_write_block(inode->i_sb) ;
2341 return reiserfs_write_full_page(page, wbc) ;
2344 int reiserfs_prepare_write(struct file *f, struct page *page,
2345 unsigned from, unsigned to) {
2346 struct inode *inode = page->mapping->host ;
2350 reiserfs_wait_on_write_block(inode->i_sb) ;
2351 fix_tail_page_for_writing(page) ;
2352 if (reiserfs_transaction_running(inode->i_sb)) {
2353 struct reiserfs_transaction_handle *th;
2354 th = (struct reiserfs_transaction_handle *)current->journal_info;
2355 old_ref = th->t_refcount;
2359 ret = block_prepare_write(page, from, to, reiserfs_get_block) ;
2360 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2361 struct reiserfs_transaction_handle *th = current->journal_info;
2362 /* this gets a little ugly. If reiserfs_get_block returned an
2363 * error and left a transacstion running, we've got to close it,
2364 * and we've got to free handle if it was a persistent transaction.
2366 * But, if we had nested into an existing transaction, we need
2367 * to just drop the ref count on the handle.
2369 * If old_ref == 0, the transaction is from reiserfs_get_block,
2370 * and it was a persistent trans. Otherwise, it was nested above.
2372 if (th->t_refcount > old_ref) {
2376 reiserfs_write_lock(inode->i_sb);
2377 reiserfs_end_persistent_transaction(th);
2378 reiserfs_write_unlock(inode->i_sb);
2387 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block) {
2388 return generic_block_bmap(as, block, reiserfs_bmap) ;
2391 static int reiserfs_commit_write(struct file *f, struct page *page,
2392 unsigned from, unsigned to) {
2393 struct inode *inode = page->mapping->host ;
2394 loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
2397 struct reiserfs_transaction_handle *th = NULL;
2399 reiserfs_wait_on_write_block(inode->i_sb) ;
2400 if (reiserfs_transaction_running(inode->i_sb)) {
2401 th = current->journal_info;
2403 reiserfs_commit_page(inode, page, from, to);
2405 /* generic_commit_write does this for us, but does not update the
2406 ** transaction tracking stuff when the size changes. So, we have
2407 ** to do the i_size updates here.
2409 if (pos > inode->i_size) {
2410 struct reiserfs_transaction_handle myth ;
2411 reiserfs_write_lock(inode->i_sb);
2412 /* If the file have grown beyond the border where it
2413 can have a tail, unmark it as needing a tail
2415 if ( (have_large_tails (inode->i_sb) && inode->i_size > i_block_size (inode)*4) ||
2416 (have_small_tails (inode->i_sb) && inode->i_size > i_block_size(inode)) )
2417 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask ;
2419 journal_begin(&myth, inode->i_sb, 1) ;
2420 reiserfs_update_inode_transaction(inode) ;
2421 inode->i_size = pos ;
2422 reiserfs_update_sd(&myth, inode) ;
2424 journal_end(&myth, inode->i_sb, 1) ;
2425 reiserfs_write_unlock(inode->i_sb);
2428 reiserfs_write_lock(inode->i_sb);
2430 reiserfs_update_sd(th, inode) ;
2431 reiserfs_end_persistent_transaction(th);
2432 reiserfs_write_unlock(inode->i_sb);
2435 /* we test for O_SYNC here so we can commit the transaction
2436 ** for any packed tails the file might have had
2438 if (f && (f->f_flags & O_SYNC)) {
2439 reiserfs_write_lock(inode->i_sb);
2440 reiserfs_commit_for_inode(inode) ;
2441 reiserfs_write_unlock(inode->i_sb);
2446 void sd_attrs_to_i_attrs( __u16 sd_attrs, struct inode *inode )
2448 if( reiserfs_attrs( inode -> i_sb ) ) {
2449 if( sd_attrs & REISERFS_SYNC_FL )
2450 inode -> i_flags |= S_SYNC;
2452 inode -> i_flags &= ~S_SYNC;
2453 if( sd_attrs & REISERFS_IMMUTABLE_FL )
2454 inode -> i_flags |= S_IMMUTABLE;
2456 inode -> i_flags &= ~S_IMMUTABLE;
2457 if( sd_attrs & REISERFS_APPEND_FL )
2458 inode -> i_flags |= S_APPEND;
2460 inode -> i_flags &= ~S_APPEND;
2461 if( sd_attrs & REISERFS_NOATIME_FL )
2462 inode -> i_flags |= S_NOATIME;
2464 inode -> i_flags &= ~S_NOATIME;
2465 if( sd_attrs & REISERFS_NOTAIL_FL )
2466 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2468 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2472 void i_attrs_to_sd_attrs( struct inode *inode, __u16 *sd_attrs )
2474 if( reiserfs_attrs( inode -> i_sb ) ) {
2475 if( inode -> i_flags & S_IMMUTABLE )
2476 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2478 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2479 if( inode -> i_flags & S_SYNC )
2480 *sd_attrs |= REISERFS_SYNC_FL;
2482 *sd_attrs &= ~REISERFS_SYNC_FL;
2483 if( inode -> i_flags & S_NOATIME )
2484 *sd_attrs |= REISERFS_NOATIME_FL;
2486 *sd_attrs &= ~REISERFS_NOATIME_FL;
2487 if( REISERFS_I(inode)->i_flags & i_nopack_mask )
2488 *sd_attrs |= REISERFS_NOTAIL_FL;
2490 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2494 /* decide if this buffer needs to stay around for data logging or ordered
2497 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2500 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb) ;
2502 spin_lock(&j->j_dirty_buffers_lock) ;
2503 if (!buffer_mapped(bh)) {
2506 /* the page is locked, and the only places that log a data buffer
2507 * also lock the page.
2509 if (reiserfs_file_data_log(inode)) {
2511 * very conservative, leave the buffer pinned if
2512 * anyone might need it.
2514 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2518 if (buffer_dirty(bh) || buffer_locked(bh)) {
2519 struct reiserfs_journal_list *jl;
2520 struct reiserfs_jh *jh = bh->b_private;
2522 /* why is this safe?
2523 * reiserfs_setattr updates i_size in the on disk
2524 * stat data before allowing vmtruncate to be called.
2526 * If buffer was put onto the ordered list for this
2527 * transaction, we know for sure either this transaction
2528 * or an older one already has updated i_size on disk,
2529 * and this ordered data won't be referenced in the file
2532 * if the buffer was put onto the ordered list for an older
2533 * transaction, we need to leave it around
2535 if (jh && (jl = jh->jl) && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2539 if (ret && bh->b_private) {
2540 reiserfs_free_jh(bh);
2542 spin_unlock(&j->j_dirty_buffers_lock) ;
2546 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2547 static int reiserfs_invalidatepage(struct page *page, unsigned long offset)
2549 struct buffer_head *head, *bh, *next;
2550 struct inode *inode = page->mapping->host;
2551 unsigned int curr_off = 0;
2554 BUG_ON(!PageLocked(page));
2557 ClearPageChecked(page);
2559 if (!page_has_buffers(page))
2562 head = page_buffers(page);
2565 unsigned int next_off = curr_off + bh->b_size;
2566 next = bh->b_this_page;
2569 * is this block fully invalidated?
2571 if (offset <= curr_off) {
2572 if (invalidatepage_can_drop(inode, bh))
2573 reiserfs_unmap_buffer(bh);
2577 curr_off = next_off;
2579 } while (bh != head);
2582 * We release buffers only if the entire page is being invalidated.
2583 * The get_block cached value has been unconditionally invalidated,
2584 * so real IO is not possible anymore.
2587 ret = try_to_release_page(page, 0);
2592 static int reiserfs_set_page_dirty(struct page *page) {
2593 struct inode *inode = page->mapping->host;
2594 if (reiserfs_file_data_log(inode)) {
2595 SetPageChecked(page);
2596 return __set_page_dirty_nobuffers(page);
2598 return __set_page_dirty_buffers(page);
2602 * Returns 1 if the page's buffers were dropped. The page is locked.
2604 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
2605 * in the buffers at page_buffers(page).
2607 * even in -o notail mode, we can't be sure an old mount without -o notail
2608 * didn't create files with tails.
2610 static int reiserfs_releasepage(struct page *page, int unused_gfp_flags)
2612 struct inode *inode = page->mapping->host ;
2613 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb) ;
2614 struct buffer_head *head ;
2615 struct buffer_head *bh ;
2618 WARN_ON(PageChecked(page));
2619 spin_lock(&j->j_dirty_buffers_lock) ;
2620 head = page_buffers(page) ;
2623 if (bh->b_private) {
2624 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
2625 reiserfs_free_jh(bh);
2631 bh = bh->b_this_page ;
2632 } while (bh != head) ;
2634 ret = try_to_free_buffers(page) ;
2635 spin_unlock(&j->j_dirty_buffers_lock) ;
2639 /* We thank Mingming Cao for helping us understand in great detail what
2640 to do in this section of the code. */
2641 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
2642 const struct iovec *iov, loff_t offset, unsigned long nr_segs)
2644 struct file *file = iocb->ki_filp;
2645 struct inode *inode = file->f_mapping->host;
2647 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2648 offset, nr_segs, reiserfs_get_blocks_direct_io, NULL);
2651 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr) {
2652 struct inode *inode = dentry->d_inode ;
2654 unsigned int ia_valid = attr->ia_valid;
2655 reiserfs_write_lock(inode->i_sb);
2656 if (attr->ia_valid & ATTR_SIZE) {
2657 /* version 2 items will be caught by the s_maxbytes check
2658 ** done for us in vmtruncate
2660 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
2661 attr->ia_size > MAX_NON_LFS) {
2665 /* fill in hole pointers in the expanding truncate case. */
2666 if (attr->ia_size > inode->i_size) {
2667 error = generic_cont_expand(inode, attr->ia_size) ;
2668 if (REISERFS_I(inode)->i_prealloc_count > 0) {
2669 struct reiserfs_transaction_handle th ;
2670 /* we're changing at most 2 bitmaps, inode + super */
2671 journal_begin(&th, inode->i_sb, 4) ;
2672 reiserfs_discard_prealloc (&th, inode);
2673 journal_end(&th, inode->i_sb, 4) ;
2680 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
2681 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
2682 (get_inode_sd_version (inode) == STAT_DATA_V1)) {
2683 /* stat data of format v3.5 has 16 bit uid and gid */
2688 error = inode_change_ok(inode, attr) ;
2690 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
2691 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
2692 error = reiserfs_chown_xattrs (inode, attr);
2695 error = DQUOT_TRANSFER(inode, attr) ? -EDQUOT : 0;
2698 error = inode_setattr(inode, attr) ;
2702 if (!error && reiserfs_posixacl (inode->i_sb)) {
2703 if (attr->ia_valid & ATTR_MODE)
2704 error = reiserfs_acl_chmod (inode);
2708 reiserfs_write_unlock(inode->i_sb);
2714 struct address_space_operations reiserfs_address_space_operations = {
2715 .writepage = reiserfs_writepage,
2716 .readpage = reiserfs_readpage,
2717 .readpages = reiserfs_readpages,
2718 .releasepage = reiserfs_releasepage,
2719 .invalidatepage = reiserfs_invalidatepage,
2720 .sync_page = block_sync_page,
2721 .prepare_write = reiserfs_prepare_write,
2722 .commit_write = reiserfs_commit_write,
2723 .bmap = reiserfs_aop_bmap,
2724 .direct_IO = reiserfs_direct_IO,
2725 .set_page_dirty = reiserfs_set_page_dirty,