2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
5 #include <linux/time.h>
10 #include <linux/exportfs.h>
11 #include <linux/pagemap.h>
12 #include <linux/highmem.h>
13 #include <linux/slab.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/swap.h>
22 int reiserfs_commit_write(struct file
*f
, struct page
*page
,
23 unsigned from
, unsigned to
);
25 void reiserfs_evict_inode(struct inode
*inode
)
27 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
29 JOURNAL_PER_BALANCE_CNT
* 2 +
30 2 * REISERFS_QUOTA_INIT_BLOCKS(inode
->i_sb
);
31 struct reiserfs_transaction_handle th
;
35 if (!inode
->i_nlink
&& !is_bad_inode(inode
))
36 dquot_initialize(inode
);
38 truncate_inode_pages(&inode
->i_data
, 0);
42 depth
= reiserfs_write_lock_once(inode
->i_sb
);
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 */
46 reiserfs_delete_xattrs(inode
);
48 if (journal_begin(&th
, inode
->i_sb
, jbegin_count
))
50 reiserfs_update_inode_transaction(inode
);
52 reiserfs_discard_prealloc(&th
, inode
);
54 err
= reiserfs_delete_object(&th
, inode
);
56 /* Do quota update inside a transaction for journaled quotas. We must do that
57 * after delete_object so that quota updates go into the same transaction as
58 * stat data deletion */
60 dquot_free_inode(inode
);
62 if (journal_end(&th
, inode
->i_sb
, jbegin_count
))
65 /* check return value from reiserfs_delete_object after
66 * ending the transaction
71 /* all items of file are deleted, so we can remove "save" link */
72 remove_save_link(inode
, 0 /* not truncate */ ); /* we can't do anything
73 * about an error here */
75 /* no object items are in the tree */
79 reiserfs_write_unlock_once(inode
->i_sb
, depth
);
80 clear_inode(inode
); /* note this must go after the journal_end to prevent deadlock */
90 static void _make_cpu_key(struct cpu_key
*key
, int version
, __u32 dirid
,
91 __u32 objectid
, loff_t offset
, int type
, int length
)
93 key
->version
= version
;
95 key
->on_disk_key
.k_dir_id
= dirid
;
96 key
->on_disk_key
.k_objectid
= objectid
;
97 set_cpu_key_k_offset(key
, offset
);
98 set_cpu_key_k_type(key
, type
);
99 key
->key_length
= length
;
102 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
103 offset and type of key */
104 void make_cpu_key(struct cpu_key
*key
, struct inode
*inode
, loff_t offset
,
105 int type
, int length
)
107 _make_cpu_key(key
, get_inode_item_key_version(inode
),
108 le32_to_cpu(INODE_PKEY(inode
)->k_dir_id
),
109 le32_to_cpu(INODE_PKEY(inode
)->k_objectid
), offset
, type
,
114 // when key is 0, do not set version and short key
116 inline void make_le_item_head(struct item_head
*ih
, const struct cpu_key
*key
,
118 loff_t offset
, int type
, int length
,
119 int entry_count
/*or ih_free_space */ )
122 ih
->ih_key
.k_dir_id
= cpu_to_le32(key
->on_disk_key
.k_dir_id
);
123 ih
->ih_key
.k_objectid
=
124 cpu_to_le32(key
->on_disk_key
.k_objectid
);
126 put_ih_version(ih
, version
);
127 set_le_ih_k_offset(ih
, offset
);
128 set_le_ih_k_type(ih
, type
);
129 put_ih_item_len(ih
, length
);
130 /* set_ih_free_space (ih, 0); */
131 // for directory items it is entry count, for directs and stat
132 // datas - 0xffff, for indirects - 0
133 put_ih_entry_count(ih
, entry_count
);
137 // FIXME: we might cache recently accessed indirect item
139 // Ugh. Not too eager for that....
140 // I cut the code until such time as I see a convincing argument (benchmark).
141 // I don't want a bloated inode struct..., and I don't like code complexity....
143 /* cutting the code is fine, since it really isn't in use yet and is easy
144 ** to add back in. But, Vladimir has a really good idea here. Think
145 ** about what happens for reading a file. For each page,
146 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
147 ** an indirect item. This indirect item has X number of pointers, where
148 ** X is a big number if we've done the block allocation right. But,
149 ** we only use one or two of these pointers during each call to readpage,
150 ** needlessly researching again later on.
152 ** The size of the cache could be dynamic based on the size of the file.
154 ** I'd also like to see us cache the location the stat data item, since
155 ** we are needlessly researching for that frequently.
160 /* If this page has a file tail in it, and
161 ** it was read in by get_block_create_0, the page data is valid,
162 ** but tail is still sitting in a direct item, and we can't write to
163 ** it. So, look through this page, and check all the mapped buffers
164 ** to make sure they have valid block numbers. Any that don't need
165 ** to be unmapped, so that __block_write_begin will correctly call
166 ** reiserfs_get_block to convert the tail into an unformatted node
168 static inline void fix_tail_page_for_writing(struct page
*page
)
170 struct buffer_head
*head
, *next
, *bh
;
172 if (page
&& page_has_buffers(page
)) {
173 head
= page_buffers(page
);
176 next
= bh
->b_this_page
;
177 if (buffer_mapped(bh
) && bh
->b_blocknr
== 0) {
178 reiserfs_unmap_buffer(bh
);
181 } while (bh
!= head
);
185 /* reiserfs_get_block does not need to allocate a block only if it has been
186 done already or non-hole position has been found in the indirect item */
187 static inline int allocation_needed(int retval
, b_blocknr_t allocated
,
188 struct item_head
*ih
,
189 __le32
* item
, int pos_in_item
)
193 if (retval
== POSITION_FOUND
&& is_indirect_le_ih(ih
) &&
194 get_block_num(item
, pos_in_item
))
199 static inline int indirect_item_found(int retval
, struct item_head
*ih
)
201 return (retval
== POSITION_FOUND
) && is_indirect_le_ih(ih
);
204 static inline void set_block_dev_mapped(struct buffer_head
*bh
,
205 b_blocknr_t block
, struct inode
*inode
)
207 map_bh(bh
, inode
->i_sb
, block
);
211 // files which were created in the earlier version can not be longer,
214 static int file_capable(struct inode
*inode
, sector_t block
)
216 if (get_inode_item_key_version(inode
) != KEY_FORMAT_3_5
|| // it is new file.
217 block
< (1 << (31 - inode
->i_sb
->s_blocksize_bits
))) // old file, but 'block' is inside of 2gb
223 static int restart_transaction(struct reiserfs_transaction_handle
*th
,
224 struct inode
*inode
, struct treepath
*path
)
226 struct super_block
*s
= th
->t_super
;
227 int len
= th
->t_blocks_allocated
;
230 BUG_ON(!th
->t_trans_id
);
231 BUG_ON(!th
->t_refcount
);
235 /* we cannot restart while nested */
236 if (th
->t_refcount
> 1) {
239 reiserfs_update_sd(th
, inode
);
240 err
= journal_end(th
, s
, len
);
242 err
= journal_begin(th
, s
, JOURNAL_PER_BALANCE_CNT
* 6);
244 reiserfs_update_inode_transaction(inode
);
249 // it is called by get_block when create == 0. Returns block number
250 // for 'block'-th logical block of file. When it hits direct item it
251 // returns 0 (being called from bmap) or read direct item into piece
252 // of page (bh_result)
254 // Please improve the english/clarity in the comment above, as it is
255 // hard to understand.
257 static int _get_block_create_0(struct inode
*inode
, sector_t block
,
258 struct buffer_head
*bh_result
, int args
)
260 INITIALIZE_PATH(path
);
262 struct buffer_head
*bh
;
263 struct item_head
*ih
, tmp_ih
;
270 unsigned long offset
;
272 // prepare the key to look for the 'block'-th block of file
273 make_cpu_key(&key
, inode
,
274 (loff_t
) block
* inode
->i_sb
->s_blocksize
+ 1, TYPE_ANY
,
277 result
= search_for_position_by_key(inode
->i_sb
, &key
, &path
);
278 if (result
!= POSITION_FOUND
) {
281 kunmap(bh_result
->b_page
);
282 if (result
== IO_ERROR
)
284 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
285 // That there is some MMAPED data associated with it that is yet to be written to disk.
286 if ((args
& GET_BLOCK_NO_HOLE
)
287 && !PageUptodate(bh_result
->b_page
)) {
293 bh
= get_last_bh(&path
);
295 if (is_indirect_le_ih(ih
)) {
296 __le32
*ind_item
= (__le32
*) B_I_PITEM(bh
, ih
);
298 /* FIXME: here we could cache indirect item or part of it in
299 the inode to avoid search_by_key in case of subsequent
301 blocknr
= get_block_num(ind_item
, path
.pos_in_item
);
304 map_bh(bh_result
, inode
->i_sb
, blocknr
);
305 if (path
.pos_in_item
==
306 ((ih_item_len(ih
) / UNFM_P_SIZE
) - 1)) {
307 set_buffer_boundary(bh_result
);
310 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
311 // That there is some MMAPED data associated with it that is yet to be written to disk.
312 if ((args
& GET_BLOCK_NO_HOLE
)
313 && !PageUptodate(bh_result
->b_page
)) {
319 kunmap(bh_result
->b_page
);
322 // requested data are in direct item(s)
323 if (!(args
& GET_BLOCK_READ_DIRECT
)) {
324 // we are called by bmap. FIXME: we can not map block of file
325 // when it is stored in direct item(s)
328 kunmap(bh_result
->b_page
);
332 /* if we've got a direct item, and the buffer or page was uptodate,
333 ** we don't want to pull data off disk again. skip to the
334 ** end, where we map the buffer and return
336 if (buffer_uptodate(bh_result
)) {
340 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
341 ** pages without any buffers. If the page is up to date, we don't want
342 ** read old data off disk. Set the up to date bit on the buffer instead
343 ** and jump to the end
345 if (!bh_result
->b_page
|| PageUptodate(bh_result
->b_page
)) {
346 set_buffer_uptodate(bh_result
);
349 // read file tail into part of page
350 offset
= (cpu_key_k_offset(&key
) - 1) & (PAGE_CACHE_SIZE
- 1);
351 copy_item_head(&tmp_ih
, ih
);
353 /* we only want to kmap if we are reading the tail into the page.
354 ** this is not the common case, so we don't kmap until we are
355 ** sure we need to. But, this means the item might move if
359 p
= (char *)kmap(bh_result
->b_page
);
362 memset(p
, 0, inode
->i_sb
->s_blocksize
);
364 if (!is_direct_le_ih(ih
)) {
367 /* make sure we don't read more bytes than actually exist in
368 ** the file. This can happen in odd cases where i_size isn't
369 ** correct, and when direct item padding results in a few
370 ** extra bytes at the end of the direct item
372 if ((le_ih_k_offset(ih
) + path
.pos_in_item
) > inode
->i_size
)
374 if ((le_ih_k_offset(ih
) - 1 + ih_item_len(ih
)) > inode
->i_size
) {
376 inode
->i_size
- (le_ih_k_offset(ih
) - 1) -
380 chars
= ih_item_len(ih
) - path
.pos_in_item
;
382 memcpy(p
, B_I_PITEM(bh
, ih
) + path
.pos_in_item
, chars
);
389 if (PATH_LAST_POSITION(&path
) != (B_NR_ITEMS(bh
) - 1))
390 // we done, if read direct item is not the last item of
391 // node FIXME: we could try to check right delimiting key
392 // to see whether direct item continues in the right
393 // neighbor or rely on i_size
396 // update key to look for the next piece
397 set_cpu_key_k_offset(&key
, cpu_key_k_offset(&key
) + chars
);
398 result
= search_for_position_by_key(inode
->i_sb
, &key
, &path
);
399 if (result
!= POSITION_FOUND
)
400 // i/o error most likely
402 bh
= get_last_bh(&path
);
406 flush_dcache_page(bh_result
->b_page
);
407 kunmap(bh_result
->b_page
);
412 if (result
== IO_ERROR
)
415 /* this buffer has valid data, but isn't valid for io. mapping it to
416 * block #0 tells the rest of reiserfs it just has a tail in it
418 map_bh(bh_result
, inode
->i_sb
, 0);
419 set_buffer_uptodate(bh_result
);
423 // this is called to create file map. So, _get_block_create_0 will not
425 static int reiserfs_bmap(struct inode
*inode
, sector_t block
,
426 struct buffer_head
*bh_result
, int create
)
428 if (!file_capable(inode
, block
))
431 reiserfs_write_lock(inode
->i_sb
);
432 /* do not read the direct item */
433 _get_block_create_0(inode
, block
, bh_result
, 0);
434 reiserfs_write_unlock(inode
->i_sb
);
438 /* special version of get_block that is only used by grab_tail_page right
439 ** now. It is sent to __block_write_begin, and when you try to get a
440 ** block past the end of the file (or a block from a hole) it returns
441 ** -ENOENT instead of a valid buffer. __block_write_begin expects to
442 ** be able to do i/o on the buffers returned, unless an error value
445 ** So, this allows __block_write_begin to be used for reading a single block
446 ** in a page. Where it does not produce a valid page for holes, or past the
447 ** end of the file. This turns out to be exactly what we need for reading
448 ** tails for conversion.
450 ** The point of the wrapper is forcing a certain value for create, even
451 ** though the VFS layer is calling this function with create==1. If you
452 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
453 ** don't use this function.
455 static int reiserfs_get_block_create_0(struct inode
*inode
, sector_t block
,
456 struct buffer_head
*bh_result
,
459 return reiserfs_get_block(inode
, block
, bh_result
, GET_BLOCK_NO_HOLE
);
462 /* This is special helper for reiserfs_get_block in case we are executing
463 direct_IO request. */
464 static int reiserfs_get_blocks_direct_io(struct inode
*inode
,
466 struct buffer_head
*bh_result
,
471 bh_result
->b_page
= NULL
;
473 /* We set the b_size before reiserfs_get_block call since it is
474 referenced in convert_tail_for_hole() that may be called from
475 reiserfs_get_block() */
476 bh_result
->b_size
= (1 << inode
->i_blkbits
);
478 ret
= reiserfs_get_block(inode
, iblock
, bh_result
,
479 create
| GET_BLOCK_NO_DANGLE
);
483 /* don't allow direct io onto tail pages */
484 if (buffer_mapped(bh_result
) && bh_result
->b_blocknr
== 0) {
485 /* make sure future calls to the direct io funcs for this offset
486 ** in the file fail by unmapping the buffer
488 clear_buffer_mapped(bh_result
);
491 /* Possible unpacked tail. Flush the data before pages have
493 if (REISERFS_I(inode
)->i_flags
& i_pack_on_close_mask
) {
496 reiserfs_write_lock(inode
->i_sb
);
498 err
= reiserfs_commit_for_inode(inode
);
499 REISERFS_I(inode
)->i_flags
&= ~i_pack_on_close_mask
;
501 reiserfs_write_unlock(inode
->i_sb
);
511 ** helper function for when reiserfs_get_block is called for a hole
512 ** but the file tail is still in a direct item
513 ** bh_result is the buffer head for the hole
514 ** tail_offset is the offset of the start of the tail in the file
516 ** This calls prepare_write, which will start a new transaction
517 ** you should not be in a transaction, or have any paths held when you
520 static int convert_tail_for_hole(struct inode
*inode
,
521 struct buffer_head
*bh_result
,
525 unsigned long tail_end
;
526 unsigned long tail_start
;
527 struct page
*tail_page
;
528 struct page
*hole_page
= bh_result
->b_page
;
531 if ((tail_offset
& (bh_result
->b_size
- 1)) != 1)
534 /* always try to read until the end of the block */
535 tail_start
= tail_offset
& (PAGE_CACHE_SIZE
- 1);
536 tail_end
= (tail_start
| (bh_result
->b_size
- 1)) + 1;
538 index
= tail_offset
>> PAGE_CACHE_SHIFT
;
539 /* hole_page can be zero in case of direct_io, we are sure
540 that we cannot get here if we write with O_DIRECT into
542 if (!hole_page
|| index
!= hole_page
->index
) {
543 tail_page
= grab_cache_page(inode
->i_mapping
, index
);
549 tail_page
= hole_page
;
552 /* we don't have to make sure the conversion did not happen while
553 ** we were locking the page because anyone that could convert
554 ** must first take i_mutex.
556 ** We must fix the tail page for writing because it might have buffers
557 ** that are mapped, but have a block number of 0. This indicates tail
558 ** data that has been read directly into the page, and
559 ** __block_write_begin won't trigger a get_block in this case.
561 fix_tail_page_for_writing(tail_page
);
562 retval
= __reiserfs_write_begin(tail_page
, tail_start
,
563 tail_end
- tail_start
);
567 /* tail conversion might change the data in the page */
568 flush_dcache_page(tail_page
);
570 retval
= reiserfs_commit_write(NULL
, tail_page
, tail_start
, tail_end
);
573 if (tail_page
!= hole_page
) {
574 unlock_page(tail_page
);
575 page_cache_release(tail_page
);
581 static inline int _allocate_block(struct reiserfs_transaction_handle
*th
,
584 b_blocknr_t
* allocated_block_nr
,
585 struct treepath
*path
, int flags
)
587 BUG_ON(!th
->t_trans_id
);
589 #ifdef REISERFS_PREALLOCATE
590 if (!(flags
& GET_BLOCK_NO_IMUX
)) {
591 return reiserfs_new_unf_blocknrs2(th
, inode
, allocated_block_nr
,
595 return reiserfs_new_unf_blocknrs(th
, inode
, allocated_block_nr
, path
,
599 int reiserfs_get_block(struct inode
*inode
, sector_t block
,
600 struct buffer_head
*bh_result
, int create
)
602 int repeat
, retval
= 0;
603 b_blocknr_t allocated_block_nr
= 0; // b_blocknr_t is (unsigned) 32 bit int
604 INITIALIZE_PATH(path
);
607 struct buffer_head
*bh
, *unbh
= NULL
;
608 struct item_head
*ih
, tmp_ih
;
613 struct reiserfs_transaction_handle
*th
= NULL
;
614 /* space reserved in transaction batch:
615 . 3 balancings in direct->indirect conversion
616 . 1 block involved into reiserfs_update_sd()
617 XXX in practically impossible worst case direct2indirect()
618 can incur (much) more than 3 balancings.
619 quota update for user, group */
621 JOURNAL_PER_BALANCE_CNT
* 3 + 1 +
622 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode
->i_sb
);
626 (((loff_t
) block
) << inode
->i_sb
->s_blocksize_bits
) + 1;
628 lock_depth
= reiserfs_write_lock_once(inode
->i_sb
);
629 version
= get_inode_item_key_version(inode
);
631 if (!file_capable(inode
, block
)) {
632 reiserfs_write_unlock_once(inode
->i_sb
, lock_depth
);
636 /* if !create, we aren't changing the FS, so we don't need to
637 ** log anything, so we don't need to start a transaction
639 if (!(create
& GET_BLOCK_CREATE
)) {
641 /* find number of block-th logical block of the file */
642 ret
= _get_block_create_0(inode
, block
, bh_result
,
643 create
| GET_BLOCK_READ_DIRECT
);
644 reiserfs_write_unlock_once(inode
->i_sb
, lock_depth
);
648 * if we're already in a transaction, make sure to close
649 * any new transactions we start in this func
651 if ((create
& GET_BLOCK_NO_DANGLE
) ||
652 reiserfs_transaction_running(inode
->i_sb
))
655 /* If file is of such a size, that it might have a tail and tails are enabled
656 ** we should mark it as possibly needing tail packing on close
658 if ((have_large_tails(inode
->i_sb
)
659 && inode
->i_size
< i_block_size(inode
) * 4)
660 || (have_small_tails(inode
->i_sb
)
661 && inode
->i_size
< i_block_size(inode
)))
662 REISERFS_I(inode
)->i_flags
|= i_pack_on_close_mask
;
664 /* set the key of the first byte in the 'block'-th block of file */
665 make_cpu_key(&key
, inode
, new_offset
, TYPE_ANY
, 3 /*key length */ );
666 if ((new_offset
+ inode
->i_sb
->s_blocksize
- 1) > inode
->i_size
) {
668 th
= reiserfs_persistent_transaction(inode
->i_sb
, jbegin_count
);
673 reiserfs_update_inode_transaction(inode
);
677 retval
= search_for_position_by_key(inode
->i_sb
, &key
, &path
);
678 if (retval
== IO_ERROR
) {
683 bh
= get_last_bh(&path
);
685 item
= get_item(&path
);
686 pos_in_item
= path
.pos_in_item
;
688 fs_gen
= get_generation(inode
->i_sb
);
689 copy_item_head(&tmp_ih
, ih
);
691 if (allocation_needed
692 (retval
, allocated_block_nr
, ih
, item
, pos_in_item
)) {
693 /* we have to allocate block for the unformatted node */
700 _allocate_block(th
, block
, inode
, &allocated_block_nr
,
703 if (repeat
== NO_DISK_SPACE
|| repeat
== QUOTA_EXCEEDED
) {
704 /* restart the transaction to give the journal a chance to free
705 ** some blocks. releases the path, so we have to go back to
706 ** research if we succeed on the second try
708 SB_JOURNAL(inode
->i_sb
)->j_next_async_flush
= 1;
709 retval
= restart_transaction(th
, inode
, &path
);
713 _allocate_block(th
, block
, inode
,
714 &allocated_block_nr
, NULL
, create
);
716 if (repeat
!= NO_DISK_SPACE
&& repeat
!= QUOTA_EXCEEDED
) {
719 if (repeat
== QUOTA_EXCEEDED
)
726 if (fs_changed(fs_gen
, inode
->i_sb
)
727 && item_moved(&tmp_ih
, &path
)) {
732 if (indirect_item_found(retval
, ih
)) {
733 b_blocknr_t unfm_ptr
;
734 /* 'block'-th block is in the file already (there is
735 corresponding cell in some indirect item). But it may be
736 zero unformatted node pointer (hole) */
737 unfm_ptr
= get_block_num(item
, pos_in_item
);
739 /* use allocated block to plug the hole */
740 reiserfs_prepare_for_journal(inode
->i_sb
, bh
, 1);
741 if (fs_changed(fs_gen
, inode
->i_sb
)
742 && item_moved(&tmp_ih
, &path
)) {
743 reiserfs_restore_prepared_buffer(inode
->i_sb
,
747 set_buffer_new(bh_result
);
748 if (buffer_dirty(bh_result
)
749 && reiserfs_data_ordered(inode
->i_sb
))
750 reiserfs_add_ordered_list(inode
, bh_result
);
751 put_block_num(item
, pos_in_item
, allocated_block_nr
);
752 unfm_ptr
= allocated_block_nr
;
753 journal_mark_dirty(th
, inode
->i_sb
, bh
);
754 reiserfs_update_sd(th
, inode
);
756 set_block_dev_mapped(bh_result
, unfm_ptr
, inode
);
760 retval
= reiserfs_end_persistent_transaction(th
);
762 reiserfs_write_unlock_once(inode
->i_sb
, lock_depth
);
764 /* the item was found, so new blocks were not added to the file
765 ** there is no need to make sure the inode is updated with this
776 /* desired position is not found or is in the direct item. We have
777 to append file with holes up to 'block'-th block converting
778 direct items to indirect one if necessary */
781 if (is_statdata_le_ih(ih
)) {
783 struct cpu_key tmp_key
;
785 /* indirect item has to be inserted */
786 make_le_item_head(&tmp_ih
, &key
, version
, 1,
787 TYPE_INDIRECT
, UNFM_P_SIZE
,
788 0 /* free_space */ );
790 if (cpu_key_k_offset(&key
) == 1) {
791 /* we are going to add 'block'-th block to the file. Use
792 allocated block for that */
793 unp
= cpu_to_le32(allocated_block_nr
);
794 set_block_dev_mapped(bh_result
,
795 allocated_block_nr
, inode
);
796 set_buffer_new(bh_result
);
800 set_cpu_key_k_offset(&tmp_key
, 1);
801 PATH_LAST_POSITION(&path
)++;
804 reiserfs_insert_item(th
, &path
, &tmp_key
, &tmp_ih
,
805 inode
, (char *)&unp
);
807 reiserfs_free_block(th
, inode
,
808 allocated_block_nr
, 1);
809 goto failure
; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
811 //mark_tail_converted (inode);
812 } else if (is_direct_le_ih(ih
)) {
813 /* direct item has to be converted */
817 ((le_ih_k_offset(ih
) -
818 1) & ~(inode
->i_sb
->s_blocksize
- 1)) + 1;
819 if (tail_offset
== cpu_key_k_offset(&key
)) {
820 /* direct item we just found fits into block we have
821 to map. Convert it into unformatted node: use
822 bh_result for the conversion */
823 set_block_dev_mapped(bh_result
,
824 allocated_block_nr
, inode
);
828 /* we have to padd file tail stored in direct item(s)
829 up to block size and convert it to unformatted
830 node. FIXME: this should also get into page cache */
834 * ugly, but we can only end the transaction if
837 BUG_ON(!th
->t_refcount
);
838 if (th
->t_refcount
== 1) {
840 reiserfs_end_persistent_transaction
848 convert_tail_for_hole(inode
, bh_result
,
851 if (retval
!= -ENOSPC
)
852 reiserfs_error(inode
->i_sb
,
854 "convert tail failed "
855 "inode %lu, error %d",
858 if (allocated_block_nr
) {
859 /* the bitmap, the super, and the stat data == 3 */
861 th
= reiserfs_persistent_transaction(inode
->i_sb
, 3);
863 reiserfs_free_block(th
,
873 direct2indirect(th
, inode
, &path
, unbh
,
876 reiserfs_unmap_buffer(unbh
);
877 reiserfs_free_block(th
, inode
,
878 allocated_block_nr
, 1);
881 /* it is important the set_buffer_uptodate is done after
882 ** the direct2indirect. The buffer might contain valid
883 ** data newer than the data on disk (read by readpage, changed,
884 ** and then sent here by writepage). direct2indirect needs
885 ** to know if unbh was already up to date, so it can decide
886 ** if the data in unbh needs to be replaced with data from
889 set_buffer_uptodate(unbh
);
891 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
892 buffer will disappear shortly, so it should not be added to
895 /* we've converted the tail, so we must
896 ** flush unbh before the transaction commits
898 reiserfs_add_tail_list(inode
, unbh
);
900 /* mark it dirty now to prevent commit_write from adding
901 ** this buffer to the inode's dirty buffer list
904 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
905 * It's still atomic, but it sets the page dirty too,
906 * which makes it eligible for writeback at any time by the
907 * VM (which was also the case with __mark_buffer_dirty())
909 mark_buffer_dirty(unbh
);
912 /* append indirect item with holes if needed, when appending
913 pointer to 'block'-th block use block, which is already
915 struct cpu_key tmp_key
;
916 unp_t unf_single
= 0; // We use this in case we need to allocate only
917 // one block which is a fastpath
919 __u64 max_to_insert
=
920 MAX_ITEM_LEN(inode
->i_sb
->s_blocksize
) /
924 RFALSE(pos_in_item
!= ih_item_len(ih
) / UNFM_P_SIZE
,
925 "vs-804: invalid position for append");
926 /* indirect item has to be appended, set up key of that position */
927 make_cpu_key(&tmp_key
, inode
,
928 le_key_k_offset(version
,
931 inode
->i_sb
->s_blocksize
),
932 //pos_in_item * inode->i_sb->s_blocksize,
933 TYPE_INDIRECT
, 3); // key type is unimportant
935 RFALSE(cpu_key_k_offset(&tmp_key
) > cpu_key_k_offset(&key
),
936 "green-805: invalid offset");
939 ((cpu_key_k_offset(&key
) -
940 cpu_key_k_offset(&tmp_key
)) >> inode
->i_sb
->
943 if (blocks_needed
== 1) {
946 un
= kzalloc(min(blocks_needed
, max_to_insert
) * UNFM_P_SIZE
, GFP_NOFS
);
953 if (blocks_needed
<= max_to_insert
) {
954 /* we are going to add target block to the file. Use allocated
956 un
[blocks_needed
- 1] =
957 cpu_to_le32(allocated_block_nr
);
958 set_block_dev_mapped(bh_result
,
959 allocated_block_nr
, inode
);
960 set_buffer_new(bh_result
);
963 /* paste hole to the indirect item */
964 /* If kmalloc failed, max_to_insert becomes zero and it means we
965 only have space for one block */
967 max_to_insert
? max_to_insert
: 1;
970 reiserfs_paste_into_item(th
, &path
, &tmp_key
, inode
,
975 if (blocks_needed
!= 1)
979 reiserfs_free_block(th
, inode
,
980 allocated_block_nr
, 1);
984 /* We need to mark new file size in case this function will be
985 interrupted/aborted later on. And we may do this only for
988 inode
->i_sb
->s_blocksize
* blocks_needed
;
995 /* this loop could log more blocks than we had originally asked
996 ** for. So, we have to allow the transaction to end if it is
997 ** too big or too full. Update the inode so things are
998 ** consistent if we crash before the function returns
1000 ** release the path so that anybody waiting on the path before
1001 ** ending their transaction will be able to continue.
1003 if (journal_transaction_should_end(th
, th
->t_blocks_allocated
)) {
1004 retval
= restart_transaction(th
, inode
, &path
);
1009 * inserting indirect pointers for a hole can take a
1010 * long time. reschedule if needed and also release the write
1013 if (need_resched()) {
1014 reiserfs_write_unlock_once(inode
->i_sb
, lock_depth
);
1016 lock_depth
= reiserfs_write_lock_once(inode
->i_sb
);
1019 retval
= search_for_position_by_key(inode
->i_sb
, &key
, &path
);
1020 if (retval
== IO_ERROR
) {
1024 if (retval
== POSITION_FOUND
) {
1025 reiserfs_warning(inode
->i_sb
, "vs-825",
1026 "%K should not be found", &key
);
1028 if (allocated_block_nr
)
1029 reiserfs_free_block(th
, inode
,
1030 allocated_block_nr
, 1);
1034 bh
= get_last_bh(&path
);
1036 item
= get_item(&path
);
1037 pos_in_item
= path
.pos_in_item
;
1043 if (th
&& (!dangle
|| (retval
&& !th
->t_trans_id
))) {
1046 reiserfs_update_sd(th
, inode
);
1047 err
= reiserfs_end_persistent_transaction(th
);
1052 reiserfs_write_unlock_once(inode
->i_sb
, lock_depth
);
1053 reiserfs_check_path(&path
);
1058 reiserfs_readpages(struct file
*file
, struct address_space
*mapping
,
1059 struct list_head
*pages
, unsigned nr_pages
)
1061 return mpage_readpages(mapping
, pages
, nr_pages
, reiserfs_get_block
);
1064 /* Compute real number of used bytes by file
1065 * Following three functions can go away when we'll have enough space in stat item
1067 static int real_space_diff(struct inode
*inode
, int sd_size
)
1070 loff_t blocksize
= inode
->i_sb
->s_blocksize
;
1072 if (S_ISLNK(inode
->i_mode
) || S_ISDIR(inode
->i_mode
))
1075 /* End of file is also in full block with indirect reference, so round
1076 ** up to the next block.
1078 ** there is just no way to know if the tail is actually packed
1079 ** on the file, so we have to assume it isn't. When we pack the
1080 ** tail, we add 4 bytes to pretend there really is an unformatted
1085 (blocksize
- 1)) >> inode
->i_sb
->s_blocksize_bits
) * UNFM_P_SIZE
+
1090 static inline loff_t
to_real_used_space(struct inode
*inode
, ulong blocks
,
1093 if (S_ISLNK(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)) {
1094 return inode
->i_size
+
1095 (loff_t
) (real_space_diff(inode
, sd_size
));
1097 return ((loff_t
) real_space_diff(inode
, sd_size
)) +
1098 (((loff_t
) blocks
) << 9);
1101 /* Compute number of blocks used by file in ReiserFS counting */
1102 static inline ulong
to_fake_used_blocks(struct inode
*inode
, int sd_size
)
1104 loff_t bytes
= inode_get_bytes(inode
);
1105 loff_t real_space
= real_space_diff(inode
, sd_size
);
1107 /* keeps fsck and non-quota versions of reiserfs happy */
1108 if (S_ISLNK(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)) {
1109 bytes
+= (loff_t
) 511;
1112 /* files from before the quota patch might i_blocks such that
1113 ** bytes < real_space. Deal with that here to prevent it from
1116 if (bytes
< real_space
)
1118 return (bytes
- real_space
) >> 9;
1122 // BAD: new directories have stat data of new type and all other items
1123 // of old type. Version stored in the inode says about body items, so
1124 // in update_stat_data we can not rely on inode, but have to check
1125 // item version directly
1128 // called by read_locked_inode
1129 static void init_inode(struct inode
*inode
, struct treepath
*path
)
1131 struct buffer_head
*bh
;
1132 struct item_head
*ih
;
1134 //int version = ITEM_VERSION_1;
1136 bh
= PATH_PLAST_BUFFER(path
);
1137 ih
= PATH_PITEM_HEAD(path
);
1139 copy_key(INODE_PKEY(inode
), &(ih
->ih_key
));
1141 INIT_LIST_HEAD(&(REISERFS_I(inode
)->i_prealloc_list
));
1142 REISERFS_I(inode
)->i_flags
= 0;
1143 REISERFS_I(inode
)->i_prealloc_block
= 0;
1144 REISERFS_I(inode
)->i_prealloc_count
= 0;
1145 REISERFS_I(inode
)->i_trans_id
= 0;
1146 REISERFS_I(inode
)->i_jl
= NULL
;
1147 reiserfs_init_xattr_rwsem(inode
);
1149 if (stat_data_v1(ih
)) {
1150 struct stat_data_v1
*sd
=
1151 (struct stat_data_v1
*)B_I_PITEM(bh
, ih
);
1152 unsigned long blocks
;
1154 set_inode_item_key_version(inode
, KEY_FORMAT_3_5
);
1155 set_inode_sd_version(inode
, STAT_DATA_V1
);
1156 inode
->i_mode
= sd_v1_mode(sd
);
1157 set_nlink(inode
, sd_v1_nlink(sd
));
1158 i_uid_write(inode
, sd_v1_uid(sd
));
1159 i_gid_write(inode
, sd_v1_gid(sd
));
1160 inode
->i_size
= sd_v1_size(sd
);
1161 inode
->i_atime
.tv_sec
= sd_v1_atime(sd
);
1162 inode
->i_mtime
.tv_sec
= sd_v1_mtime(sd
);
1163 inode
->i_ctime
.tv_sec
= sd_v1_ctime(sd
);
1164 inode
->i_atime
.tv_nsec
= 0;
1165 inode
->i_ctime
.tv_nsec
= 0;
1166 inode
->i_mtime
.tv_nsec
= 0;
1168 inode
->i_blocks
= sd_v1_blocks(sd
);
1169 inode
->i_generation
= le32_to_cpu(INODE_PKEY(inode
)->k_dir_id
);
1170 blocks
= (inode
->i_size
+ 511) >> 9;
1171 blocks
= _ROUND_UP(blocks
, inode
->i_sb
->s_blocksize
>> 9);
1172 if (inode
->i_blocks
> blocks
) {
1173 // there was a bug in <=3.5.23 when i_blocks could take negative
1174 // values. Starting from 3.5.17 this value could even be stored in
1175 // stat data. For such files we set i_blocks based on file
1176 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1177 // only updated if file's inode will ever change
1178 inode
->i_blocks
= blocks
;
1181 rdev
= sd_v1_rdev(sd
);
1182 REISERFS_I(inode
)->i_first_direct_byte
=
1183 sd_v1_first_direct_byte(sd
);
1184 /* an early bug in the quota code can give us an odd number for the
1185 ** block count. This is incorrect, fix it here.
1187 if (inode
->i_blocks
& 1) {
1190 inode_set_bytes(inode
,
1191 to_real_used_space(inode
, inode
->i_blocks
,
1193 /* nopack is initially zero for v1 objects. For v2 objects,
1194 nopack is initialised from sd_attrs */
1195 REISERFS_I(inode
)->i_flags
&= ~i_nopack_mask
;
1197 // new stat data found, but object may have old items
1198 // (directories and symlinks)
1199 struct stat_data
*sd
= (struct stat_data
*)B_I_PITEM(bh
, ih
);
1201 inode
->i_mode
= sd_v2_mode(sd
);
1202 set_nlink(inode
, sd_v2_nlink(sd
));
1203 i_uid_write(inode
, sd_v2_uid(sd
));
1204 inode
->i_size
= sd_v2_size(sd
);
1205 i_gid_write(inode
, sd_v2_gid(sd
));
1206 inode
->i_mtime
.tv_sec
= sd_v2_mtime(sd
);
1207 inode
->i_atime
.tv_sec
= sd_v2_atime(sd
);
1208 inode
->i_ctime
.tv_sec
= sd_v2_ctime(sd
);
1209 inode
->i_ctime
.tv_nsec
= 0;
1210 inode
->i_mtime
.tv_nsec
= 0;
1211 inode
->i_atime
.tv_nsec
= 0;
1212 inode
->i_blocks
= sd_v2_blocks(sd
);
1213 rdev
= sd_v2_rdev(sd
);
1214 if (S_ISCHR(inode
->i_mode
) || S_ISBLK(inode
->i_mode
))
1215 inode
->i_generation
=
1216 le32_to_cpu(INODE_PKEY(inode
)->k_dir_id
);
1218 inode
->i_generation
= sd_v2_generation(sd
);
1220 if (S_ISDIR(inode
->i_mode
) || S_ISLNK(inode
->i_mode
))
1221 set_inode_item_key_version(inode
, KEY_FORMAT_3_5
);
1223 set_inode_item_key_version(inode
, KEY_FORMAT_3_6
);
1224 REISERFS_I(inode
)->i_first_direct_byte
= 0;
1225 set_inode_sd_version(inode
, STAT_DATA_V2
);
1226 inode_set_bytes(inode
,
1227 to_real_used_space(inode
, inode
->i_blocks
,
1229 /* read persistent inode attributes from sd and initialise
1230 generic inode flags from them */
1231 REISERFS_I(inode
)->i_attrs
= sd_v2_attrs(sd
);
1232 sd_attrs_to_i_attrs(sd_v2_attrs(sd
), inode
);
1236 if (S_ISREG(inode
->i_mode
)) {
1237 inode
->i_op
= &reiserfs_file_inode_operations
;
1238 inode
->i_fop
= &reiserfs_file_operations
;
1239 inode
->i_mapping
->a_ops
= &reiserfs_address_space_operations
;
1240 } else if (S_ISDIR(inode
->i_mode
)) {
1241 inode
->i_op
= &reiserfs_dir_inode_operations
;
1242 inode
->i_fop
= &reiserfs_dir_operations
;
1243 } else if (S_ISLNK(inode
->i_mode
)) {
1244 inode
->i_op
= &reiserfs_symlink_inode_operations
;
1245 inode
->i_mapping
->a_ops
= &reiserfs_address_space_operations
;
1247 inode
->i_blocks
= 0;
1248 inode
->i_op
= &reiserfs_special_inode_operations
;
1249 init_special_inode(inode
, inode
->i_mode
, new_decode_dev(rdev
));
1253 // update new stat data with inode fields
1254 static void inode2sd(void *sd
, struct inode
*inode
, loff_t size
)
1256 struct stat_data
*sd_v2
= (struct stat_data
*)sd
;
1259 set_sd_v2_mode(sd_v2
, inode
->i_mode
);
1260 set_sd_v2_nlink(sd_v2
, inode
->i_nlink
);
1261 set_sd_v2_uid(sd_v2
, i_uid_read(inode
));
1262 set_sd_v2_size(sd_v2
, size
);
1263 set_sd_v2_gid(sd_v2
, i_gid_read(inode
));
1264 set_sd_v2_mtime(sd_v2
, inode
->i_mtime
.tv_sec
);
1265 set_sd_v2_atime(sd_v2
, inode
->i_atime
.tv_sec
);
1266 set_sd_v2_ctime(sd_v2
, inode
->i_ctime
.tv_sec
);
1267 set_sd_v2_blocks(sd_v2
, to_fake_used_blocks(inode
, SD_V2_SIZE
));
1268 if (S_ISCHR(inode
->i_mode
) || S_ISBLK(inode
->i_mode
))
1269 set_sd_v2_rdev(sd_v2
, new_encode_dev(inode
->i_rdev
));
1271 set_sd_v2_generation(sd_v2
, inode
->i_generation
);
1272 flags
= REISERFS_I(inode
)->i_attrs
;
1273 i_attrs_to_sd_attrs(inode
, &flags
);
1274 set_sd_v2_attrs(sd_v2
, flags
);
1277 // used to copy inode's fields to old stat data
1278 static void inode2sd_v1(void *sd
, struct inode
*inode
, loff_t size
)
1280 struct stat_data_v1
*sd_v1
= (struct stat_data_v1
*)sd
;
1282 set_sd_v1_mode(sd_v1
, inode
->i_mode
);
1283 set_sd_v1_uid(sd_v1
, i_uid_read(inode
));
1284 set_sd_v1_gid(sd_v1
, i_gid_read(inode
));
1285 set_sd_v1_nlink(sd_v1
, inode
->i_nlink
);
1286 set_sd_v1_size(sd_v1
, size
);
1287 set_sd_v1_atime(sd_v1
, inode
->i_atime
.tv_sec
);
1288 set_sd_v1_ctime(sd_v1
, inode
->i_ctime
.tv_sec
);
1289 set_sd_v1_mtime(sd_v1
, inode
->i_mtime
.tv_sec
);
1291 if (S_ISCHR(inode
->i_mode
) || S_ISBLK(inode
->i_mode
))
1292 set_sd_v1_rdev(sd_v1
, new_encode_dev(inode
->i_rdev
));
1294 set_sd_v1_blocks(sd_v1
, to_fake_used_blocks(inode
, SD_V1_SIZE
));
1296 // Sigh. i_first_direct_byte is back
1297 set_sd_v1_first_direct_byte(sd_v1
,
1298 REISERFS_I(inode
)->i_first_direct_byte
);
1301 /* NOTE, you must prepare the buffer head before sending it here,
1302 ** and then log it after the call
1304 static void update_stat_data(struct treepath
*path
, struct inode
*inode
,
1307 struct buffer_head
*bh
;
1308 struct item_head
*ih
;
1310 bh
= PATH_PLAST_BUFFER(path
);
1311 ih
= PATH_PITEM_HEAD(path
);
1313 if (!is_statdata_le_ih(ih
))
1314 reiserfs_panic(inode
->i_sb
, "vs-13065", "key %k, found item %h",
1315 INODE_PKEY(inode
), ih
);
1317 if (stat_data_v1(ih
)) {
1318 // path points to old stat data
1319 inode2sd_v1(B_I_PITEM(bh
, ih
), inode
, size
);
1321 inode2sd(B_I_PITEM(bh
, ih
), inode
, size
);
1327 void reiserfs_update_sd_size(struct reiserfs_transaction_handle
*th
,
1328 struct inode
*inode
, loff_t size
)
1331 INITIALIZE_PATH(path
);
1332 struct buffer_head
*bh
;
1334 struct item_head
*ih
, tmp_ih
;
1337 BUG_ON(!th
->t_trans_id
);
1339 make_cpu_key(&key
, inode
, SD_OFFSET
, TYPE_STAT_DATA
, 3); //key type is unimportant
1343 /* look for the object's stat data */
1344 retval
= search_item(inode
->i_sb
, &key
, &path
);
1345 if (retval
== IO_ERROR
) {
1346 reiserfs_error(inode
->i_sb
, "vs-13050",
1347 "i/o failure occurred trying to "
1348 "update %K stat data", &key
);
1351 if (retval
== ITEM_NOT_FOUND
) {
1352 pos
= PATH_LAST_POSITION(&path
);
1354 if (inode
->i_nlink
== 0) {
1355 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1358 reiserfs_warning(inode
->i_sb
, "vs-13060",
1359 "stat data of object %k (nlink == %d) "
1360 "not found (pos %d)",
1361 INODE_PKEY(inode
), inode
->i_nlink
,
1363 reiserfs_check_path(&path
);
1367 /* sigh, prepare_for_journal might schedule. When it schedules the
1368 ** FS might change. We have to detect that, and loop back to the
1369 ** search if the stat data item has moved
1371 bh
= get_last_bh(&path
);
1373 copy_item_head(&tmp_ih
, ih
);
1374 fs_gen
= get_generation(inode
->i_sb
);
1375 reiserfs_prepare_for_journal(inode
->i_sb
, bh
, 1);
1376 if (fs_changed(fs_gen
, inode
->i_sb
)
1377 && item_moved(&tmp_ih
, &path
)) {
1378 reiserfs_restore_prepared_buffer(inode
->i_sb
, bh
);
1379 continue; /* Stat_data item has been moved after scheduling. */
1383 update_stat_data(&path
, inode
, size
);
1384 journal_mark_dirty(th
, th
->t_super
, bh
);
1389 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1390 ** does a make_bad_inode when things go wrong. But, we need to make sure
1391 ** and clear the key in the private portion of the inode, otherwise a
1392 ** corresponding iput might try to delete whatever object the inode last
1395 static void reiserfs_make_bad_inode(struct inode
*inode
)
1397 memset(INODE_PKEY(inode
), 0, KEY_SIZE
);
1398 make_bad_inode(inode
);
1402 // initially this function was derived from minix or ext2's analog and
1403 // evolved as the prototype did
1406 int reiserfs_init_locked_inode(struct inode
*inode
, void *p
)
1408 struct reiserfs_iget_args
*args
= (struct reiserfs_iget_args
*)p
;
1409 inode
->i_ino
= args
->objectid
;
1410 INODE_PKEY(inode
)->k_dir_id
= cpu_to_le32(args
->dirid
);
1414 /* looks for stat data in the tree, and fills up the fields of in-core
1415 inode stat data fields */
1416 void reiserfs_read_locked_inode(struct inode
*inode
,
1417 struct reiserfs_iget_args
*args
)
1419 INITIALIZE_PATH(path_to_sd
);
1421 unsigned long dirino
;
1424 dirino
= args
->dirid
;
1426 /* set version 1, version 2 could be used too, because stat data
1427 key is the same in both versions */
1428 key
.version
= KEY_FORMAT_3_5
;
1429 key
.on_disk_key
.k_dir_id
= dirino
;
1430 key
.on_disk_key
.k_objectid
= inode
->i_ino
;
1431 key
.on_disk_key
.k_offset
= 0;
1432 key
.on_disk_key
.k_type
= 0;
1434 /* look for the object's stat data */
1435 retval
= search_item(inode
->i_sb
, &key
, &path_to_sd
);
1436 if (retval
== IO_ERROR
) {
1437 reiserfs_error(inode
->i_sb
, "vs-13070",
1438 "i/o failure occurred trying to find "
1439 "stat data of %K", &key
);
1440 reiserfs_make_bad_inode(inode
);
1443 if (retval
!= ITEM_FOUND
) {
1444 /* a stale NFS handle can trigger this without it being an error */
1445 pathrelse(&path_to_sd
);
1446 reiserfs_make_bad_inode(inode
);
1451 init_inode(inode
, &path_to_sd
);
1453 /* It is possible that knfsd is trying to access inode of a file
1454 that is being removed from the disk by some other thread. As we
1455 update sd on unlink all that is required is to check for nlink
1456 here. This bug was first found by Sizif when debugging
1457 SquidNG/Butterfly, forgotten, and found again after Philippe
1458 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1460 More logical fix would require changes in fs/inode.c:iput() to
1461 remove inode from hash-table _after_ fs cleaned disk stuff up and
1462 in iget() to return NULL if I_FREEING inode is found in
1464 /* Currently there is one place where it's ok to meet inode with
1465 nlink==0: processing of open-unlinked and half-truncated files
1466 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1467 if ((inode
->i_nlink
== 0) &&
1468 !REISERFS_SB(inode
->i_sb
)->s_is_unlinked_ok
) {
1469 reiserfs_warning(inode
->i_sb
, "vs-13075",
1470 "dead inode read from disk %K. "
1471 "This is likely to be race with knfsd. Ignore",
1473 reiserfs_make_bad_inode(inode
);
1476 reiserfs_check_path(&path_to_sd
); /* init inode should be relsing */
1479 * Stat data v1 doesn't support ACLs.
1481 if (get_inode_sd_version(inode
) == STAT_DATA_V1
)
1482 cache_no_acl(inode
);
1486 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1488 * @inode: inode from hash table to check
1489 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1491 * This function is called by iget5_locked() to distinguish reiserfs inodes
1492 * having the same inode numbers. Such inodes can only exist due to some
1493 * error condition. One of them should be bad. Inodes with identical
1494 * inode numbers (objectids) are distinguished by parent directory ids.
1497 int reiserfs_find_actor(struct inode
*inode
, void *opaque
)
1499 struct reiserfs_iget_args
*args
;
1502 /* args is already in CPU order */
1503 return (inode
->i_ino
== args
->objectid
) &&
1504 (le32_to_cpu(INODE_PKEY(inode
)->k_dir_id
) == args
->dirid
);
1507 struct inode
*reiserfs_iget(struct super_block
*s
, const struct cpu_key
*key
)
1509 struct inode
*inode
;
1510 struct reiserfs_iget_args args
;
1512 args
.objectid
= key
->on_disk_key
.k_objectid
;
1513 args
.dirid
= key
->on_disk_key
.k_dir_id
;
1514 reiserfs_write_unlock(s
);
1515 inode
= iget5_locked(s
, key
->on_disk_key
.k_objectid
,
1516 reiserfs_find_actor
, reiserfs_init_locked_inode
,
1518 reiserfs_write_lock(s
);
1520 return ERR_PTR(-ENOMEM
);
1522 if (inode
->i_state
& I_NEW
) {
1523 reiserfs_read_locked_inode(inode
, &args
);
1524 unlock_new_inode(inode
);
1527 if (comp_short_keys(INODE_PKEY(inode
), key
) || is_bad_inode(inode
)) {
1528 /* either due to i/o error or a stale NFS handle */
1535 static struct dentry
*reiserfs_get_dentry(struct super_block
*sb
,
1536 u32 objectid
, u32 dir_id
, u32 generation
)
1540 struct inode
*inode
;
1542 key
.on_disk_key
.k_objectid
= objectid
;
1543 key
.on_disk_key
.k_dir_id
= dir_id
;
1544 reiserfs_write_lock(sb
);
1545 inode
= reiserfs_iget(sb
, &key
);
1546 if (inode
&& !IS_ERR(inode
) && generation
!= 0 &&
1547 generation
!= inode
->i_generation
) {
1551 reiserfs_write_unlock(sb
);
1553 return d_obtain_alias(inode
);
1556 struct dentry
*reiserfs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1557 int fh_len
, int fh_type
)
1559 /* fhtype happens to reflect the number of u32s encoded.
1560 * due to a bug in earlier code, fhtype might indicate there
1561 * are more u32s then actually fitted.
1562 * so if fhtype seems to be more than len, reduce fhtype.
1564 * 2 - objectid + dir_id - legacy support
1565 * 3 - objectid + dir_id + generation
1566 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1567 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1568 * 6 - as above plus generation of directory
1569 * 6 does not fit in NFSv2 handles
1571 if (fh_type
> fh_len
) {
1572 if (fh_type
!= 6 || fh_len
!= 5)
1573 reiserfs_warning(sb
, "reiserfs-13077",
1574 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1581 return reiserfs_get_dentry(sb
, fid
->raw
[0], fid
->raw
[1],
1582 (fh_type
== 3 || fh_type
>= 5) ? fid
->raw
[2] : 0);
1585 struct dentry
*reiserfs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1586 int fh_len
, int fh_type
)
1588 if (fh_type
> fh_len
)
1593 return reiserfs_get_dentry(sb
,
1594 (fh_type
>= 5) ? fid
->raw
[3] : fid
->raw
[2],
1595 (fh_type
>= 5) ? fid
->raw
[4] : fid
->raw
[3],
1596 (fh_type
== 6) ? fid
->raw
[5] : 0);
1599 int reiserfs_encode_fh(struct inode
*inode
, __u32
* data
, int *lenp
,
1600 struct inode
*parent
)
1604 if (parent
&& (maxlen
< 5)) {
1607 } else if (maxlen
< 3) {
1612 data
[0] = inode
->i_ino
;
1613 data
[1] = le32_to_cpu(INODE_PKEY(inode
)->k_dir_id
);
1614 data
[2] = inode
->i_generation
;
1617 data
[3] = parent
->i_ino
;
1618 data
[4] = le32_to_cpu(INODE_PKEY(parent
)->k_dir_id
);
1621 data
[5] = parent
->i_generation
;
1628 /* looks for stat data, then copies fields to it, marks the buffer
1629 containing stat data as dirty */
1630 /* reiserfs inodes are never really dirty, since the dirty inode call
1631 ** always logs them. This call allows the VFS inode marking routines
1632 ** to properly mark inodes for datasync and such, but only actually
1633 ** does something when called for a synchronous update.
1635 int reiserfs_write_inode(struct inode
*inode
, struct writeback_control
*wbc
)
1637 struct reiserfs_transaction_handle th
;
1638 int jbegin_count
= 1;
1640 if (inode
->i_sb
->s_flags
& MS_RDONLY
)
1642 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1643 ** these cases are just when the system needs ram, not when the
1644 ** inode needs to reach disk for safety, and they can safely be
1645 ** ignored because the altered inode has already been logged.
1647 if (wbc
->sync_mode
== WB_SYNC_ALL
&& !(current
->flags
& PF_MEMALLOC
)) {
1648 reiserfs_write_lock(inode
->i_sb
);
1649 if (!journal_begin(&th
, inode
->i_sb
, jbegin_count
)) {
1650 reiserfs_update_sd(&th
, inode
);
1651 journal_end_sync(&th
, inode
->i_sb
, jbegin_count
);
1653 reiserfs_write_unlock(inode
->i_sb
);
1658 /* stat data of new object is inserted already, this inserts the item
1659 containing "." and ".." entries */
1660 static int reiserfs_new_directory(struct reiserfs_transaction_handle
*th
,
1661 struct inode
*inode
,
1662 struct item_head
*ih
, struct treepath
*path
,
1665 struct super_block
*sb
= th
->t_super
;
1666 char empty_dir
[EMPTY_DIR_SIZE
];
1667 char *body
= empty_dir
;
1671 BUG_ON(!th
->t_trans_id
);
1673 _make_cpu_key(&key
, KEY_FORMAT_3_5
, le32_to_cpu(ih
->ih_key
.k_dir_id
),
1674 le32_to_cpu(ih
->ih_key
.k_objectid
), DOT_OFFSET
,
1675 TYPE_DIRENTRY
, 3 /*key length */ );
1677 /* compose item head for new item. Directories consist of items of
1678 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1679 is done by reiserfs_new_inode */
1680 if (old_format_only(sb
)) {
1681 make_le_item_head(ih
, NULL
, KEY_FORMAT_3_5
, DOT_OFFSET
,
1682 TYPE_DIRENTRY
, EMPTY_DIR_SIZE_V1
, 2);
1684 make_empty_dir_item_v1(body
, ih
->ih_key
.k_dir_id
,
1685 ih
->ih_key
.k_objectid
,
1686 INODE_PKEY(dir
)->k_dir_id
,
1687 INODE_PKEY(dir
)->k_objectid
);
1689 make_le_item_head(ih
, NULL
, KEY_FORMAT_3_5
, DOT_OFFSET
,
1690 TYPE_DIRENTRY
, EMPTY_DIR_SIZE
, 2);
1692 make_empty_dir_item(body
, ih
->ih_key
.k_dir_id
,
1693 ih
->ih_key
.k_objectid
,
1694 INODE_PKEY(dir
)->k_dir_id
,
1695 INODE_PKEY(dir
)->k_objectid
);
1698 /* look for place in the tree for new item */
1699 retval
= search_item(sb
, &key
, path
);
1700 if (retval
== IO_ERROR
) {
1701 reiserfs_error(sb
, "vs-13080",
1702 "i/o failure occurred creating new directory");
1705 if (retval
== ITEM_FOUND
) {
1707 reiserfs_warning(sb
, "vs-13070",
1708 "object with this key exists (%k)",
1713 /* insert item, that is empty directory item */
1714 return reiserfs_insert_item(th
, path
, &key
, ih
, inode
, body
);
1717 /* stat data of object has been inserted, this inserts the item
1718 containing the body of symlink */
1719 static int reiserfs_new_symlink(struct reiserfs_transaction_handle
*th
, struct inode
*inode
, /* Inode of symlink */
1720 struct item_head
*ih
,
1721 struct treepath
*path
, const char *symname
,
1724 struct super_block
*sb
= th
->t_super
;
1728 BUG_ON(!th
->t_trans_id
);
1730 _make_cpu_key(&key
, KEY_FORMAT_3_5
,
1731 le32_to_cpu(ih
->ih_key
.k_dir_id
),
1732 le32_to_cpu(ih
->ih_key
.k_objectid
),
1733 1, TYPE_DIRECT
, 3 /*key length */ );
1735 make_le_item_head(ih
, NULL
, KEY_FORMAT_3_5
, 1, TYPE_DIRECT
, item_len
,
1736 0 /*free_space */ );
1738 /* look for place in the tree for new item */
1739 retval
= search_item(sb
, &key
, path
);
1740 if (retval
== IO_ERROR
) {
1741 reiserfs_error(sb
, "vs-13080",
1742 "i/o failure occurred creating new symlink");
1745 if (retval
== ITEM_FOUND
) {
1747 reiserfs_warning(sb
, "vs-13080",
1748 "object with this key exists (%k)",
1753 /* insert item, that is body of symlink */
1754 return reiserfs_insert_item(th
, path
, &key
, ih
, inode
, symname
);
1757 /* inserts the stat data into the tree, and then calls
1758 reiserfs_new_directory (to insert ".", ".." item if new object is
1759 directory) or reiserfs_new_symlink (to insert symlink body if new
1760 object is symlink) or nothing (if new object is regular file)
1762 NOTE! uid and gid must already be set in the inode. If we return
1763 non-zero due to an error, we have to drop the quota previously allocated
1764 for the fresh inode. This can only be done outside a transaction, so
1765 if we return non-zero, we also end the transaction. */
1766 int reiserfs_new_inode(struct reiserfs_transaction_handle
*th
,
1767 struct inode
*dir
, umode_t mode
, const char *symname
,
1768 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1769 strlen (symname) for symlinks) */
1770 loff_t i_size
, struct dentry
*dentry
,
1771 struct inode
*inode
,
1772 struct reiserfs_security_handle
*security
)
1774 struct super_block
*sb
;
1775 struct reiserfs_iget_args args
;
1776 INITIALIZE_PATH(path_to_key
);
1778 struct item_head ih
;
1779 struct stat_data sd
;
1783 BUG_ON(!th
->t_trans_id
);
1785 dquot_initialize(inode
);
1786 err
= dquot_alloc_inode(inode
);
1789 if (!dir
->i_nlink
) {
1796 /* item head of new item */
1797 ih
.ih_key
.k_dir_id
= reiserfs_choose_packing(dir
);
1798 ih
.ih_key
.k_objectid
= cpu_to_le32(reiserfs_get_unused_objectid(th
));
1799 if (!ih
.ih_key
.k_objectid
) {
1803 args
.objectid
= inode
->i_ino
= le32_to_cpu(ih
.ih_key
.k_objectid
);
1804 if (old_format_only(sb
))
1805 make_le_item_head(&ih
, NULL
, KEY_FORMAT_3_5
, SD_OFFSET
,
1806 TYPE_STAT_DATA
, SD_V1_SIZE
, MAX_US_INT
);
1808 make_le_item_head(&ih
, NULL
, KEY_FORMAT_3_6
, SD_OFFSET
,
1809 TYPE_STAT_DATA
, SD_SIZE
, MAX_US_INT
);
1810 memcpy(INODE_PKEY(inode
), &(ih
.ih_key
), KEY_SIZE
);
1811 args
.dirid
= le32_to_cpu(ih
.ih_key
.k_dir_id
);
1812 if (insert_inode_locked4(inode
, args
.objectid
,
1813 reiserfs_find_actor
, &args
) < 0) {
1817 if (old_format_only(sb
))
1818 /* not a perfect generation count, as object ids can be reused, but
1819 ** this is as good as reiserfs can do right now.
1820 ** note that the private part of inode isn't filled in yet, we have
1821 ** to use the directory.
1823 inode
->i_generation
= le32_to_cpu(INODE_PKEY(dir
)->k_objectid
);
1825 #if defined( USE_INODE_GENERATION_COUNTER )
1826 inode
->i_generation
=
1827 le32_to_cpu(REISERFS_SB(sb
)->s_rs
->s_inode_generation
);
1829 inode
->i_generation
= ++event
;
1832 /* fill stat data */
1833 set_nlink(inode
, (S_ISDIR(mode
) ? 2 : 1));
1835 /* uid and gid must already be set by the caller for quota init */
1837 /* symlink cannot be immutable or append only, right? */
1838 if (S_ISLNK(inode
->i_mode
))
1839 inode
->i_flags
&= ~(S_IMMUTABLE
| S_APPEND
);
1841 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME_SEC
;
1842 inode
->i_size
= i_size
;
1843 inode
->i_blocks
= 0;
1845 REISERFS_I(inode
)->i_first_direct_byte
= S_ISLNK(mode
) ? 1 :
1846 U32_MAX
/*NO_BYTES_IN_DIRECT_ITEM */ ;
1848 INIT_LIST_HEAD(&(REISERFS_I(inode
)->i_prealloc_list
));
1849 REISERFS_I(inode
)->i_flags
= 0;
1850 REISERFS_I(inode
)->i_prealloc_block
= 0;
1851 REISERFS_I(inode
)->i_prealloc_count
= 0;
1852 REISERFS_I(inode
)->i_trans_id
= 0;
1853 REISERFS_I(inode
)->i_jl
= NULL
;
1854 REISERFS_I(inode
)->i_attrs
=
1855 REISERFS_I(dir
)->i_attrs
& REISERFS_INHERIT_MASK
;
1856 sd_attrs_to_i_attrs(REISERFS_I(inode
)->i_attrs
, inode
);
1857 reiserfs_init_xattr_rwsem(inode
);
1859 /* key to search for correct place for new stat data */
1860 _make_cpu_key(&key
, KEY_FORMAT_3_6
, le32_to_cpu(ih
.ih_key
.k_dir_id
),
1861 le32_to_cpu(ih
.ih_key
.k_objectid
), SD_OFFSET
,
1862 TYPE_STAT_DATA
, 3 /*key length */ );
1864 /* find proper place for inserting of stat data */
1865 retval
= search_item(sb
, &key
, &path_to_key
);
1866 if (retval
== IO_ERROR
) {
1870 if (retval
== ITEM_FOUND
) {
1871 pathrelse(&path_to_key
);
1875 if (old_format_only(sb
)) {
1876 if (i_uid_read(inode
) & ~0xffff || i_gid_read(inode
) & ~0xffff) {
1877 pathrelse(&path_to_key
);
1878 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1882 inode2sd_v1(&sd
, inode
, inode
->i_size
);
1884 inode2sd(&sd
, inode
, inode
->i_size
);
1886 // store in in-core inode the key of stat data and version all
1887 // object items will have (directory items will have old offset
1888 // format, other new objects will consist of new items)
1889 if (old_format_only(sb
) || S_ISDIR(mode
) || S_ISLNK(mode
))
1890 set_inode_item_key_version(inode
, KEY_FORMAT_3_5
);
1892 set_inode_item_key_version(inode
, KEY_FORMAT_3_6
);
1893 if (old_format_only(sb
))
1894 set_inode_sd_version(inode
, STAT_DATA_V1
);
1896 set_inode_sd_version(inode
, STAT_DATA_V2
);
1898 /* insert the stat data into the tree */
1899 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1900 if (REISERFS_I(dir
)->new_packing_locality
)
1901 th
->displace_new_blocks
= 1;
1904 reiserfs_insert_item(th
, &path_to_key
, &key
, &ih
, inode
,
1908 reiserfs_check_path(&path_to_key
);
1911 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1912 if (!th
->displace_new_blocks
)
1913 REISERFS_I(dir
)->new_packing_locality
= 0;
1915 if (S_ISDIR(mode
)) {
1916 /* insert item with "." and ".." */
1918 reiserfs_new_directory(th
, inode
, &ih
, &path_to_key
, dir
);
1921 if (S_ISLNK(mode
)) {
1922 /* insert body of symlink */
1923 if (!old_format_only(sb
))
1924 i_size
= ROUND_UP(i_size
);
1926 reiserfs_new_symlink(th
, inode
, &ih
, &path_to_key
, symname
,
1931 reiserfs_check_path(&path_to_key
);
1932 journal_end(th
, th
->t_super
, th
->t_blocks_allocated
);
1933 goto out_inserted_sd
;
1936 if (reiserfs_posixacl(inode
->i_sb
)) {
1937 retval
= reiserfs_inherit_default_acl(th
, dir
, dentry
, inode
);
1940 reiserfs_check_path(&path_to_key
);
1941 journal_end(th
, th
->t_super
, th
->t_blocks_allocated
);
1942 goto out_inserted_sd
;
1944 } else if (inode
->i_sb
->s_flags
& MS_POSIXACL
) {
1945 reiserfs_warning(inode
->i_sb
, "jdm-13090",
1946 "ACLs aren't enabled in the fs, "
1947 "but vfs thinks they are!");
1948 } else if (IS_PRIVATE(dir
))
1949 inode
->i_flags
|= S_PRIVATE
;
1951 if (security
->name
) {
1952 retval
= reiserfs_security_write(th
, inode
, security
);
1955 reiserfs_check_path(&path_to_key
);
1956 retval
= journal_end(th
, th
->t_super
,
1957 th
->t_blocks_allocated
);
1960 goto out_inserted_sd
;
1964 reiserfs_update_sd(th
, inode
);
1965 reiserfs_check_path(&path_to_key
);
1969 /* it looks like you can easily compress these two goto targets into
1970 * one. Keeping it like this doesn't actually hurt anything, and they
1971 * are place holders for what the quota code actually needs.
1974 /* Invalidate the object, nothing was inserted yet */
1975 INODE_PKEY(inode
)->k_objectid
= 0;
1977 /* Quota change must be inside a transaction for journaling */
1978 dquot_free_inode(inode
);
1981 journal_end(th
, th
->t_super
, th
->t_blocks_allocated
);
1982 /* Drop can be outside and it needs more credits so it's better to have it outside */
1984 inode
->i_flags
|= S_NOQUOTA
;
1985 make_bad_inode(inode
);
1989 th
->t_trans_id
= 0; /* so the caller can't use this handle later */
1990 unlock_new_inode(inode
); /* OK to do even if we hadn't locked it */
1996 ** finds the tail page in the page cache,
1997 ** reads the last block in.
1999 ** On success, page_result is set to a locked, pinned page, and bh_result
2000 ** is set to an up to date buffer for the last block in the file. returns 0.
2002 ** tail conversion is not done, so bh_result might not be valid for writing
2003 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2004 ** trying to write the block.
2006 ** on failure, nonzero is returned, page_result and bh_result are untouched.
2008 static int grab_tail_page(struct inode
*inode
,
2009 struct page
**page_result
,
2010 struct buffer_head
**bh_result
)
2013 /* we want the page with the last byte in the file,
2014 ** not the page that will hold the next byte for appending
2016 unsigned long index
= (inode
->i_size
- 1) >> PAGE_CACHE_SHIFT
;
2017 unsigned long pos
= 0;
2018 unsigned long start
= 0;
2019 unsigned long blocksize
= inode
->i_sb
->s_blocksize
;
2020 unsigned long offset
= (inode
->i_size
) & (PAGE_CACHE_SIZE
- 1);
2021 struct buffer_head
*bh
;
2022 struct buffer_head
*head
;
2026 /* we know that we are only called with inode->i_size > 0.
2027 ** we also know that a file tail can never be as big as a block
2028 ** If i_size % blocksize == 0, our file is currently block aligned
2029 ** and it won't need converting or zeroing after a truncate.
2031 if ((offset
& (blocksize
- 1)) == 0) {
2034 page
= grab_cache_page(inode
->i_mapping
, index
);
2039 /* start within the page of the last block in the file */
2040 start
= (offset
/ blocksize
) * blocksize
;
2042 error
= __block_write_begin(page
, start
, offset
- start
,
2043 reiserfs_get_block_create_0
);
2047 head
= page_buffers(page
);
2053 bh
= bh
->b_this_page
;
2055 } while (bh
!= head
);
2057 if (!buffer_uptodate(bh
)) {
2058 /* note, this should never happen, prepare_write should
2059 ** be taking care of this for us. If the buffer isn't up to date,
2060 ** I've screwed up the code to find the buffer, or the code to
2061 ** call prepare_write
2063 reiserfs_error(inode
->i_sb
, "clm-6000",
2064 "error reading block %lu", bh
->b_blocknr
);
2069 *page_result
= page
;
2076 page_cache_release(page
);
2081 ** vfs version of truncate file. Must NOT be called with
2082 ** a transaction already started.
2084 ** some code taken from block_truncate_page
2086 int reiserfs_truncate_file(struct inode
*inode
, int update_timestamps
)
2088 struct reiserfs_transaction_handle th
;
2089 /* we want the offset for the first byte after the end of the file */
2090 unsigned long offset
= inode
->i_size
& (PAGE_CACHE_SIZE
- 1);
2091 unsigned blocksize
= inode
->i_sb
->s_blocksize
;
2093 struct page
*page
= NULL
;
2095 struct buffer_head
*bh
= NULL
;
2099 lock_depth
= reiserfs_write_lock_once(inode
->i_sb
);
2101 if (inode
->i_size
> 0) {
2102 error
= grab_tail_page(inode
, &page
, &bh
);
2104 // -ENOENT means we truncated past the end of the file,
2105 // and get_block_create_0 could not find a block to read in,
2107 if (error
!= -ENOENT
)
2108 reiserfs_error(inode
->i_sb
, "clm-6001",
2109 "grab_tail_page failed %d",
2116 /* so, if page != NULL, we have a buffer head for the offset at
2117 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2118 ** then we have an unformatted node. Otherwise, we have a direct item,
2119 ** and no zeroing is required on disk. We zero after the truncate,
2120 ** because the truncate might pack the item anyway
2121 ** (it will unmap bh if it packs).
2123 /* it is enough to reserve space in transaction for 2 balancings:
2124 one for "save" link adding and another for the first
2125 cut_from_item. 1 is for update_sd */
2126 error
= journal_begin(&th
, inode
->i_sb
,
2127 JOURNAL_PER_BALANCE_CNT
* 2 + 1);
2130 reiserfs_update_inode_transaction(inode
);
2131 if (update_timestamps
)
2132 /* we are doing real truncate: if the system crashes before the last
2133 transaction of truncating gets committed - on reboot the file
2134 either appears truncated properly or not truncated at all */
2135 add_save_link(&th
, inode
, 1);
2136 err2
= reiserfs_do_truncate(&th
, inode
, page
, update_timestamps
);
2138 journal_end(&th
, inode
->i_sb
, JOURNAL_PER_BALANCE_CNT
* 2 + 1);
2142 /* check reiserfs_do_truncate after ending the transaction */
2148 if (update_timestamps
) {
2149 error
= remove_save_link(inode
, 1 /* truncate */);
2155 length
= offset
& (blocksize
- 1);
2156 /* if we are not on a block boundary */
2158 length
= blocksize
- length
;
2159 zero_user(page
, offset
, length
);
2160 if (buffer_mapped(bh
) && bh
->b_blocknr
!= 0) {
2161 mark_buffer_dirty(bh
);
2165 page_cache_release(page
);
2168 reiserfs_write_unlock_once(inode
->i_sb
, lock_depth
);
2174 page_cache_release(page
);
2177 reiserfs_write_unlock_once(inode
->i_sb
, lock_depth
);
2182 static int map_block_for_writepage(struct inode
*inode
,
2183 struct buffer_head
*bh_result
,
2184 unsigned long block
)
2186 struct reiserfs_transaction_handle th
;
2188 struct item_head tmp_ih
;
2189 struct item_head
*ih
;
2190 struct buffer_head
*bh
;
2193 INITIALIZE_PATH(path
);
2195 int jbegin_count
= JOURNAL_PER_BALANCE_CNT
;
2196 loff_t byte_offset
= ((loff_t
)block
<< inode
->i_sb
->s_blocksize_bits
)+1;
2198 int use_get_block
= 0;
2199 int bytes_copied
= 0;
2201 int trans_running
= 0;
2203 /* catch places below that try to log something without starting a trans */
2206 if (!buffer_uptodate(bh_result
)) {
2210 kmap(bh_result
->b_page
);
2212 reiserfs_write_lock(inode
->i_sb
);
2213 make_cpu_key(&key
, inode
, byte_offset
, TYPE_ANY
, 3);
2216 retval
= search_for_position_by_key(inode
->i_sb
, &key
, &path
);
2217 if (retval
!= POSITION_FOUND
) {
2222 bh
= get_last_bh(&path
);
2224 item
= get_item(&path
);
2225 pos_in_item
= path
.pos_in_item
;
2227 /* we've found an unformatted node */
2228 if (indirect_item_found(retval
, ih
)) {
2229 if (bytes_copied
> 0) {
2230 reiserfs_warning(inode
->i_sb
, "clm-6002",
2231 "bytes_copied %d", bytes_copied
);
2233 if (!get_block_num(item
, pos_in_item
)) {
2234 /* crap, we are writing to a hole */
2238 set_block_dev_mapped(bh_result
,
2239 get_block_num(item
, pos_in_item
), inode
);
2240 } else if (is_direct_le_ih(ih
)) {
2242 p
= page_address(bh_result
->b_page
);
2243 p
+= (byte_offset
- 1) & (PAGE_CACHE_SIZE
- 1);
2244 copy_size
= ih_item_len(ih
) - pos_in_item
;
2246 fs_gen
= get_generation(inode
->i_sb
);
2247 copy_item_head(&tmp_ih
, ih
);
2249 if (!trans_running
) {
2250 /* vs-3050 is gone, no need to drop the path */
2251 retval
= journal_begin(&th
, inode
->i_sb
, jbegin_count
);
2254 reiserfs_update_inode_transaction(inode
);
2256 if (fs_changed(fs_gen
, inode
->i_sb
)
2257 && item_moved(&tmp_ih
, &path
)) {
2258 reiserfs_restore_prepared_buffer(inode
->i_sb
,
2264 reiserfs_prepare_for_journal(inode
->i_sb
, bh
, 1);
2266 if (fs_changed(fs_gen
, inode
->i_sb
)
2267 && item_moved(&tmp_ih
, &path
)) {
2268 reiserfs_restore_prepared_buffer(inode
->i_sb
, bh
);
2272 memcpy(B_I_PITEM(bh
, ih
) + pos_in_item
, p
+ bytes_copied
,
2275 journal_mark_dirty(&th
, inode
->i_sb
, bh
);
2276 bytes_copied
+= copy_size
;
2277 set_block_dev_mapped(bh_result
, 0, inode
);
2279 /* are there still bytes left? */
2280 if (bytes_copied
< bh_result
->b_size
&&
2281 (byte_offset
+ bytes_copied
) < inode
->i_size
) {
2282 set_cpu_key_k_offset(&key
,
2283 cpu_key_k_offset(&key
) +
2288 reiserfs_warning(inode
->i_sb
, "clm-6003",
2289 "bad item inode %lu", inode
->i_ino
);
2297 if (trans_running
) {
2298 int err
= journal_end(&th
, inode
->i_sb
, jbegin_count
);
2303 reiserfs_write_unlock(inode
->i_sb
);
2305 /* this is where we fill in holes in the file. */
2306 if (use_get_block
) {
2307 retval
= reiserfs_get_block(inode
, block
, bh_result
,
2308 GET_BLOCK_CREATE
| GET_BLOCK_NO_IMUX
2309 | GET_BLOCK_NO_DANGLE
);
2311 if (!buffer_mapped(bh_result
)
2312 || bh_result
->b_blocknr
== 0) {
2313 /* get_block failed to find a mapped unformatted node. */
2319 kunmap(bh_result
->b_page
);
2321 if (!retval
&& buffer_mapped(bh_result
) && bh_result
->b_blocknr
== 0) {
2322 /* we've copied data from the page into the direct item, so the
2323 * buffer in the page is now clean, mark it to reflect that.
2325 lock_buffer(bh_result
);
2326 clear_buffer_dirty(bh_result
);
2327 unlock_buffer(bh_result
);
2333 * mason@suse.com: updated in 2.5.54 to follow the same general io
2334 * start/recovery path as __block_write_full_page, along with special
2335 * code to handle reiserfs tails.
2337 static int reiserfs_write_full_page(struct page
*page
,
2338 struct writeback_control
*wbc
)
2340 struct inode
*inode
= page
->mapping
->host
;
2341 unsigned long end_index
= inode
->i_size
>> PAGE_CACHE_SHIFT
;
2343 unsigned long block
;
2344 sector_t last_block
;
2345 struct buffer_head
*head
, *bh
;
2348 int checked
= PageChecked(page
);
2349 struct reiserfs_transaction_handle th
;
2350 struct super_block
*s
= inode
->i_sb
;
2351 int bh_per_page
= PAGE_CACHE_SIZE
/ s
->s_blocksize
;
2354 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2355 if (checked
&& (current
->flags
& PF_MEMALLOC
)) {
2356 redirty_page_for_writepage(wbc
, page
);
2361 /* The page dirty bit is cleared before writepage is called, which
2362 * means we have to tell create_empty_buffers to make dirty buffers
2363 * The page really should be up to date at this point, so tossing
2364 * in the BH_Uptodate is just a sanity check.
2366 if (!page_has_buffers(page
)) {
2367 create_empty_buffers(page
, s
->s_blocksize
,
2368 (1 << BH_Dirty
) | (1 << BH_Uptodate
));
2370 head
= page_buffers(page
);
2372 /* last page in the file, zero out any contents past the
2373 ** last byte in the file
2375 if (page
->index
>= end_index
) {
2376 unsigned last_offset
;
2378 last_offset
= inode
->i_size
& (PAGE_CACHE_SIZE
- 1);
2379 /* no file contents in this page */
2380 if (page
->index
>= end_index
+ 1 || !last_offset
) {
2384 zero_user_segment(page
, last_offset
, PAGE_CACHE_SIZE
);
2387 block
= page
->index
<< (PAGE_CACHE_SHIFT
- s
->s_blocksize_bits
);
2388 last_block
= (i_size_read(inode
) - 1) >> inode
->i_blkbits
;
2389 /* first map all the buffers, logging any direct items we find */
2391 if (block
> last_block
) {
2393 * This can happen when the block size is less than
2394 * the page size. The corresponding bytes in the page
2395 * were zero filled above
2397 clear_buffer_dirty(bh
);
2398 set_buffer_uptodate(bh
);
2399 } else if ((checked
|| buffer_dirty(bh
)) &&
2400 (!buffer_mapped(bh
) || (buffer_mapped(bh
)
2403 /* not mapped yet, or it points to a direct item, search
2404 * the btree for the mapping info, and log any direct
2407 if ((error
= map_block_for_writepage(inode
, bh
, block
))) {
2411 bh
= bh
->b_this_page
;
2413 } while (bh
!= head
);
2416 * we start the transaction after map_block_for_writepage,
2417 * because it can create holes in the file (an unbounded operation).
2418 * starting it here, we can make a reliable estimate for how many
2419 * blocks we're going to log
2422 ClearPageChecked(page
);
2423 reiserfs_write_lock(s
);
2424 error
= journal_begin(&th
, s
, bh_per_page
+ 1);
2426 reiserfs_write_unlock(s
);
2429 reiserfs_update_inode_transaction(inode
);
2431 /* now go through and lock any dirty buffers on the page */
2434 if (!buffer_mapped(bh
))
2436 if (buffer_mapped(bh
) && bh
->b_blocknr
== 0)
2440 reiserfs_prepare_for_journal(s
, bh
, 1);
2441 journal_mark_dirty(&th
, s
, bh
);
2444 /* from this point on, we know the buffer is mapped to a
2445 * real block and not a direct item
2447 if (wbc
->sync_mode
!= WB_SYNC_NONE
) {
2450 if (!trylock_buffer(bh
)) {
2451 redirty_page_for_writepage(wbc
, page
);
2455 if (test_clear_buffer_dirty(bh
)) {
2456 mark_buffer_async_write(bh
);
2460 } while ((bh
= bh
->b_this_page
) != head
);
2463 error
= journal_end(&th
, s
, bh_per_page
+ 1);
2464 reiserfs_write_unlock(s
);
2468 BUG_ON(PageWriteback(page
));
2469 set_page_writeback(page
);
2473 * since any buffer might be the only dirty buffer on the page,
2474 * the first submit_bh can bring the page out of writeback.
2475 * be careful with the buffers.
2478 struct buffer_head
*next
= bh
->b_this_page
;
2479 if (buffer_async_write(bh
)) {
2480 submit_bh(WRITE
, bh
);
2485 } while (bh
!= head
);
2491 * if this page only had a direct item, it is very possible for
2492 * no io to be required without there being an error. Or,
2493 * someone else could have locked them and sent them down the
2494 * pipe without locking the page
2498 if (!buffer_uptodate(bh
)) {
2502 bh
= bh
->b_this_page
;
2503 } while (bh
!= head
);
2505 SetPageUptodate(page
);
2506 end_page_writeback(page
);
2511 /* catches various errors, we need to make sure any valid dirty blocks
2512 * get to the media. The page is currently locked and not marked for
2515 ClearPageUptodate(page
);
2519 if (buffer_mapped(bh
) && buffer_dirty(bh
) && bh
->b_blocknr
) {
2521 mark_buffer_async_write(bh
);
2524 * clear any dirty bits that might have come from getting
2525 * attached to a dirty page
2527 clear_buffer_dirty(bh
);
2529 bh
= bh
->b_this_page
;
2530 } while (bh
!= head
);
2532 BUG_ON(PageWriteback(page
));
2533 set_page_writeback(page
);
2536 struct buffer_head
*next
= bh
->b_this_page
;
2537 if (buffer_async_write(bh
)) {
2538 clear_buffer_dirty(bh
);
2539 submit_bh(WRITE
, bh
);
2544 } while (bh
!= head
);
2548 static int reiserfs_readpage(struct file
*f
, struct page
*page
)
2550 return block_read_full_page(page
, reiserfs_get_block
);
2553 static int reiserfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2555 struct inode
*inode
= page
->mapping
->host
;
2556 reiserfs_wait_on_write_block(inode
->i_sb
);
2557 return reiserfs_write_full_page(page
, wbc
);
2560 static void reiserfs_truncate_failed_write(struct inode
*inode
)
2562 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2563 reiserfs_truncate_file(inode
, 0);
2566 static int reiserfs_write_begin(struct file
*file
,
2567 struct address_space
*mapping
,
2568 loff_t pos
, unsigned len
, unsigned flags
,
2569 struct page
**pagep
, void **fsdata
)
2571 struct inode
*inode
;
2577 inode
= mapping
->host
;
2579 if (flags
& AOP_FLAG_CONT_EXPAND
&&
2580 (pos
& (inode
->i_sb
->s_blocksize
- 1)) == 0) {
2582 *fsdata
= (void *)(unsigned long)flags
;
2585 index
= pos
>> PAGE_CACHE_SHIFT
;
2586 page
= grab_cache_page_write_begin(mapping
, index
, flags
);
2591 reiserfs_wait_on_write_block(inode
->i_sb
);
2592 fix_tail_page_for_writing(page
);
2593 if (reiserfs_transaction_running(inode
->i_sb
)) {
2594 struct reiserfs_transaction_handle
*th
;
2595 th
= (struct reiserfs_transaction_handle
*)current
->
2597 BUG_ON(!th
->t_refcount
);
2598 BUG_ON(!th
->t_trans_id
);
2599 old_ref
= th
->t_refcount
;
2602 ret
= __block_write_begin(page
, pos
, len
, reiserfs_get_block
);
2603 if (ret
&& reiserfs_transaction_running(inode
->i_sb
)) {
2604 struct reiserfs_transaction_handle
*th
= current
->journal_info
;
2605 /* this gets a little ugly. If reiserfs_get_block returned an
2606 * error and left a transacstion running, we've got to close it,
2607 * and we've got to free handle if it was a persistent transaction.
2609 * But, if we had nested into an existing transaction, we need
2610 * to just drop the ref count on the handle.
2612 * If old_ref == 0, the transaction is from reiserfs_get_block,
2613 * and it was a persistent trans. Otherwise, it was nested above.
2615 if (th
->t_refcount
> old_ref
) {
2620 reiserfs_write_lock(inode
->i_sb
);
2621 err
= reiserfs_end_persistent_transaction(th
);
2622 reiserfs_write_unlock(inode
->i_sb
);
2630 page_cache_release(page
);
2631 /* Truncate allocated blocks */
2632 reiserfs_truncate_failed_write(inode
);
2637 int __reiserfs_write_begin(struct page
*page
, unsigned from
, unsigned len
)
2639 struct inode
*inode
= page
->mapping
->host
;
2643 reiserfs_write_unlock(inode
->i_sb
);
2644 reiserfs_wait_on_write_block(inode
->i_sb
);
2645 reiserfs_write_lock(inode
->i_sb
);
2647 fix_tail_page_for_writing(page
);
2648 if (reiserfs_transaction_running(inode
->i_sb
)) {
2649 struct reiserfs_transaction_handle
*th
;
2650 th
= (struct reiserfs_transaction_handle
*)current
->
2652 BUG_ON(!th
->t_refcount
);
2653 BUG_ON(!th
->t_trans_id
);
2654 old_ref
= th
->t_refcount
;
2658 ret
= __block_write_begin(page
, from
, len
, reiserfs_get_block
);
2659 if (ret
&& reiserfs_transaction_running(inode
->i_sb
)) {
2660 struct reiserfs_transaction_handle
*th
= current
->journal_info
;
2661 /* this gets a little ugly. If reiserfs_get_block returned an
2662 * error and left a transacstion running, we've got to close it,
2663 * and we've got to free handle if it was a persistent transaction.
2665 * But, if we had nested into an existing transaction, we need
2666 * to just drop the ref count on the handle.
2668 * If old_ref == 0, the transaction is from reiserfs_get_block,
2669 * and it was a persistent trans. Otherwise, it was nested above.
2671 if (th
->t_refcount
> old_ref
) {
2676 reiserfs_write_lock(inode
->i_sb
);
2677 err
= reiserfs_end_persistent_transaction(th
);
2678 reiserfs_write_unlock(inode
->i_sb
);
2688 static sector_t
reiserfs_aop_bmap(struct address_space
*as
, sector_t block
)
2690 return generic_block_bmap(as
, block
, reiserfs_bmap
);
2693 static int reiserfs_write_end(struct file
*file
, struct address_space
*mapping
,
2694 loff_t pos
, unsigned len
, unsigned copied
,
2695 struct page
*page
, void *fsdata
)
2697 struct inode
*inode
= page
->mapping
->host
;
2700 struct reiserfs_transaction_handle
*th
;
2703 bool locked
= false;
2705 if ((unsigned long)fsdata
& AOP_FLAG_CONT_EXPAND
)
2708 reiserfs_wait_on_write_block(inode
->i_sb
);
2709 if (reiserfs_transaction_running(inode
->i_sb
))
2710 th
= current
->journal_info
;
2714 start
= pos
& (PAGE_CACHE_SIZE
- 1);
2715 if (unlikely(copied
< len
)) {
2716 if (!PageUptodate(page
))
2719 page_zero_new_buffers(page
, start
+ copied
, start
+ len
);
2721 flush_dcache_page(page
);
2723 reiserfs_commit_page(inode
, page
, start
, start
+ copied
);
2725 /* generic_commit_write does this for us, but does not update the
2726 ** transaction tracking stuff when the size changes. So, we have
2727 ** to do the i_size updates here.
2729 if (pos
+ copied
> inode
->i_size
) {
2730 struct reiserfs_transaction_handle myth
;
2731 lock_depth
= reiserfs_write_lock_once(inode
->i_sb
);
2733 /* If the file have grown beyond the border where it
2734 can have a tail, unmark it as needing a tail
2736 if ((have_large_tails(inode
->i_sb
)
2737 && inode
->i_size
> i_block_size(inode
) * 4)
2738 || (have_small_tails(inode
->i_sb
)
2739 && inode
->i_size
> i_block_size(inode
)))
2740 REISERFS_I(inode
)->i_flags
&= ~i_pack_on_close_mask
;
2742 ret
= journal_begin(&myth
, inode
->i_sb
, 1);
2746 reiserfs_update_inode_transaction(inode
);
2747 inode
->i_size
= pos
+ copied
;
2749 * this will just nest into our transaction. It's important
2750 * to use mark_inode_dirty so the inode gets pushed around on the
2751 * dirty lists, and so that O_SYNC works as expected
2753 mark_inode_dirty(inode
);
2754 reiserfs_update_sd(&myth
, inode
);
2756 ret
= journal_end(&myth
, inode
->i_sb
, 1);
2762 lock_depth
= reiserfs_write_lock_once(inode
->i_sb
);
2766 mark_inode_dirty(inode
);
2767 ret
= reiserfs_end_persistent_transaction(th
);
2774 reiserfs_write_unlock_once(inode
->i_sb
, lock_depth
);
2776 page_cache_release(page
);
2778 if (pos
+ len
> inode
->i_size
)
2779 reiserfs_truncate_failed_write(inode
);
2781 return ret
== 0 ? copied
: ret
;
2784 reiserfs_write_unlock_once(inode
->i_sb
, lock_depth
);
2788 reiserfs_update_sd(th
, inode
);
2789 ret
= reiserfs_end_persistent_transaction(th
);
2794 int reiserfs_commit_write(struct file
*f
, struct page
*page
,
2795 unsigned from
, unsigned to
)
2797 struct inode
*inode
= page
->mapping
->host
;
2798 loff_t pos
= ((loff_t
) page
->index
<< PAGE_CACHE_SHIFT
) + to
;
2801 struct reiserfs_transaction_handle
*th
= NULL
;
2803 reiserfs_write_unlock(inode
->i_sb
);
2804 reiserfs_wait_on_write_block(inode
->i_sb
);
2805 reiserfs_write_lock(inode
->i_sb
);
2807 if (reiserfs_transaction_running(inode
->i_sb
)) {
2808 th
= current
->journal_info
;
2810 reiserfs_commit_page(inode
, page
, from
, to
);
2812 /* generic_commit_write does this for us, but does not update the
2813 ** transaction tracking stuff when the size changes. So, we have
2814 ** to do the i_size updates here.
2816 if (pos
> inode
->i_size
) {
2817 struct reiserfs_transaction_handle myth
;
2818 /* If the file have grown beyond the border where it
2819 can have a tail, unmark it as needing a tail
2821 if ((have_large_tails(inode
->i_sb
)
2822 && inode
->i_size
> i_block_size(inode
) * 4)
2823 || (have_small_tails(inode
->i_sb
)
2824 && inode
->i_size
> i_block_size(inode
)))
2825 REISERFS_I(inode
)->i_flags
&= ~i_pack_on_close_mask
;
2827 ret
= journal_begin(&myth
, inode
->i_sb
, 1);
2831 reiserfs_update_inode_transaction(inode
);
2832 inode
->i_size
= pos
;
2834 * this will just nest into our transaction. It's important
2835 * to use mark_inode_dirty so the inode gets pushed around on the
2836 * dirty lists, and so that O_SYNC works as expected
2838 mark_inode_dirty(inode
);
2839 reiserfs_update_sd(&myth
, inode
);
2841 ret
= journal_end(&myth
, inode
->i_sb
, 1);
2847 mark_inode_dirty(inode
);
2848 ret
= reiserfs_end_persistent_transaction(th
);
2859 reiserfs_update_sd(th
, inode
);
2860 ret
= reiserfs_end_persistent_transaction(th
);
2866 void sd_attrs_to_i_attrs(__u16 sd_attrs
, struct inode
*inode
)
2868 if (reiserfs_attrs(inode
->i_sb
)) {
2869 if (sd_attrs
& REISERFS_SYNC_FL
)
2870 inode
->i_flags
|= S_SYNC
;
2872 inode
->i_flags
&= ~S_SYNC
;
2873 if (sd_attrs
& REISERFS_IMMUTABLE_FL
)
2874 inode
->i_flags
|= S_IMMUTABLE
;
2876 inode
->i_flags
&= ~S_IMMUTABLE
;
2877 if (sd_attrs
& REISERFS_APPEND_FL
)
2878 inode
->i_flags
|= S_APPEND
;
2880 inode
->i_flags
&= ~S_APPEND
;
2881 if (sd_attrs
& REISERFS_NOATIME_FL
)
2882 inode
->i_flags
|= S_NOATIME
;
2884 inode
->i_flags
&= ~S_NOATIME
;
2885 if (sd_attrs
& REISERFS_NOTAIL_FL
)
2886 REISERFS_I(inode
)->i_flags
|= i_nopack_mask
;
2888 REISERFS_I(inode
)->i_flags
&= ~i_nopack_mask
;
2892 void i_attrs_to_sd_attrs(struct inode
*inode
, __u16
* sd_attrs
)
2894 if (reiserfs_attrs(inode
->i_sb
)) {
2895 if (inode
->i_flags
& S_IMMUTABLE
)
2896 *sd_attrs
|= REISERFS_IMMUTABLE_FL
;
2898 *sd_attrs
&= ~REISERFS_IMMUTABLE_FL
;
2899 if (inode
->i_flags
& S_SYNC
)
2900 *sd_attrs
|= REISERFS_SYNC_FL
;
2902 *sd_attrs
&= ~REISERFS_SYNC_FL
;
2903 if (inode
->i_flags
& S_NOATIME
)
2904 *sd_attrs
|= REISERFS_NOATIME_FL
;
2906 *sd_attrs
&= ~REISERFS_NOATIME_FL
;
2907 if (REISERFS_I(inode
)->i_flags
& i_nopack_mask
)
2908 *sd_attrs
|= REISERFS_NOTAIL_FL
;
2910 *sd_attrs
&= ~REISERFS_NOTAIL_FL
;
2914 /* decide if this buffer needs to stay around for data logging or ordered
2917 static int invalidatepage_can_drop(struct inode
*inode
, struct buffer_head
*bh
)
2920 struct reiserfs_journal
*j
= SB_JOURNAL(inode
->i_sb
);
2923 spin_lock(&j
->j_dirty_buffers_lock
);
2924 if (!buffer_mapped(bh
)) {
2927 /* the page is locked, and the only places that log a data buffer
2928 * also lock the page.
2930 if (reiserfs_file_data_log(inode
)) {
2932 * very conservative, leave the buffer pinned if
2933 * anyone might need it.
2935 if (buffer_journaled(bh
) || buffer_journal_dirty(bh
)) {
2938 } else if (buffer_dirty(bh
)) {
2939 struct reiserfs_journal_list
*jl
;
2940 struct reiserfs_jh
*jh
= bh
->b_private
;
2942 /* why is this safe?
2943 * reiserfs_setattr updates i_size in the on disk
2944 * stat data before allowing vmtruncate to be called.
2946 * If buffer was put onto the ordered list for this
2947 * transaction, we know for sure either this transaction
2948 * or an older one already has updated i_size on disk,
2949 * and this ordered data won't be referenced in the file
2952 * if the buffer was put onto the ordered list for an older
2953 * transaction, we need to leave it around
2955 if (jh
&& (jl
= jh
->jl
)
2956 && jl
!= SB_JOURNAL(inode
->i_sb
)->j_current_jl
)
2960 if (ret
&& bh
->b_private
) {
2961 reiserfs_free_jh(bh
);
2963 spin_unlock(&j
->j_dirty_buffers_lock
);
2968 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2969 static void reiserfs_invalidatepage(struct page
*page
, unsigned long offset
)
2971 struct buffer_head
*head
, *bh
, *next
;
2972 struct inode
*inode
= page
->mapping
->host
;
2973 unsigned int curr_off
= 0;
2976 BUG_ON(!PageLocked(page
));
2979 ClearPageChecked(page
);
2981 if (!page_has_buffers(page
))
2984 head
= page_buffers(page
);
2987 unsigned int next_off
= curr_off
+ bh
->b_size
;
2988 next
= bh
->b_this_page
;
2991 * is this block fully invalidated?
2993 if (offset
<= curr_off
) {
2994 if (invalidatepage_can_drop(inode
, bh
))
2995 reiserfs_unmap_buffer(bh
);
2999 curr_off
= next_off
;
3001 } while (bh
!= head
);
3004 * We release buffers only if the entire page is being invalidated.
3005 * The get_block cached value has been unconditionally invalidated,
3006 * so real IO is not possible anymore.
3008 if (!offset
&& ret
) {
3009 ret
= try_to_release_page(page
, 0);
3010 /* maybe should BUG_ON(!ret); - neilb */
3016 static int reiserfs_set_page_dirty(struct page
*page
)
3018 struct inode
*inode
= page
->mapping
->host
;
3019 if (reiserfs_file_data_log(inode
)) {
3020 SetPageChecked(page
);
3021 return __set_page_dirty_nobuffers(page
);
3023 return __set_page_dirty_buffers(page
);
3027 * Returns 1 if the page's buffers were dropped. The page is locked.
3029 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3030 * in the buffers at page_buffers(page).
3032 * even in -o notail mode, we can't be sure an old mount without -o notail
3033 * didn't create files with tails.
3035 static int reiserfs_releasepage(struct page
*page
, gfp_t unused_gfp_flags
)
3037 struct inode
*inode
= page
->mapping
->host
;
3038 struct reiserfs_journal
*j
= SB_JOURNAL(inode
->i_sb
);
3039 struct buffer_head
*head
;
3040 struct buffer_head
*bh
;
3043 WARN_ON(PageChecked(page
));
3044 spin_lock(&j
->j_dirty_buffers_lock
);
3045 head
= page_buffers(page
);
3048 if (bh
->b_private
) {
3049 if (!buffer_dirty(bh
) && !buffer_locked(bh
)) {
3050 reiserfs_free_jh(bh
);
3056 bh
= bh
->b_this_page
;
3057 } while (bh
!= head
);
3059 ret
= try_to_free_buffers(page
);
3060 spin_unlock(&j
->j_dirty_buffers_lock
);
3064 /* We thank Mingming Cao for helping us understand in great detail what
3065 to do in this section of the code. */
3066 static ssize_t
reiserfs_direct_IO(int rw
, struct kiocb
*iocb
,
3067 const struct iovec
*iov
, loff_t offset
,
3068 unsigned long nr_segs
)
3070 struct file
*file
= iocb
->ki_filp
;
3071 struct inode
*inode
= file
->f_mapping
->host
;
3074 ret
= blockdev_direct_IO(rw
, iocb
, inode
, iov
, offset
, nr_segs
,
3075 reiserfs_get_blocks_direct_io
);
3078 * In case of error extending write may have instantiated a few
3079 * blocks outside i_size. Trim these off again.
3081 if (unlikely((rw
& WRITE
) && ret
< 0)) {
3082 loff_t isize
= i_size_read(inode
);
3083 loff_t end
= offset
+ iov_length(iov
, nr_segs
);
3086 vmtruncate(inode
, isize
);
3092 int reiserfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
3094 struct inode
*inode
= dentry
->d_inode
;
3095 unsigned int ia_valid
;
3099 error
= inode_change_ok(inode
, attr
);
3103 /* must be turned off for recursive notify_change calls */
3104 ia_valid
= attr
->ia_valid
&= ~(ATTR_KILL_SUID
|ATTR_KILL_SGID
);
3106 depth
= reiserfs_write_lock_once(inode
->i_sb
);
3107 if (is_quota_modification(inode
, attr
))
3108 dquot_initialize(inode
);
3110 if (attr
->ia_valid
& ATTR_SIZE
) {
3111 /* version 2 items will be caught by the s_maxbytes check
3112 ** done for us in vmtruncate
3114 if (get_inode_item_key_version(inode
) == KEY_FORMAT_3_5
&&
3115 attr
->ia_size
> MAX_NON_LFS
) {
3120 inode_dio_wait(inode
);
3122 /* fill in hole pointers in the expanding truncate case. */
3123 if (attr
->ia_size
> inode
->i_size
) {
3124 error
= generic_cont_expand_simple(inode
, attr
->ia_size
);
3125 if (REISERFS_I(inode
)->i_prealloc_count
> 0) {
3127 struct reiserfs_transaction_handle th
;
3128 /* we're changing at most 2 bitmaps, inode + super */
3129 err
= journal_begin(&th
, inode
->i_sb
, 4);
3131 reiserfs_discard_prealloc(&th
, inode
);
3132 err
= journal_end(&th
, inode
->i_sb
, 4);
3140 * file size is changed, ctime and mtime are
3143 attr
->ia_valid
|= (ATTR_MTIME
| ATTR_CTIME
);
3147 if ((((attr
->ia_valid
& ATTR_UID
) && (from_kuid(&init_user_ns
, attr
->ia_uid
) & ~0xffff)) ||
3148 ((attr
->ia_valid
& ATTR_GID
) && (from_kgid(&init_user_ns
, attr
->ia_gid
) & ~0xffff))) &&
3149 (get_inode_sd_version(inode
) == STAT_DATA_V1
)) {
3150 /* stat data of format v3.5 has 16 bit uid and gid */
3155 if ((ia_valid
& ATTR_UID
&& !uid_eq(attr
->ia_uid
, inode
->i_uid
)) ||
3156 (ia_valid
& ATTR_GID
&& !gid_eq(attr
->ia_gid
, inode
->i_gid
))) {
3157 struct reiserfs_transaction_handle th
;
3160 (REISERFS_QUOTA_INIT_BLOCKS(inode
->i_sb
) +
3161 REISERFS_QUOTA_DEL_BLOCKS(inode
->i_sb
)) +
3164 error
= reiserfs_chown_xattrs(inode
, attr
);
3169 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3170 error
= journal_begin(&th
, inode
->i_sb
, jbegin_count
);
3173 error
= dquot_transfer(inode
, attr
);
3175 journal_end(&th
, inode
->i_sb
, jbegin_count
);
3179 /* Update corresponding info in inode so that everything is in
3180 * one transaction */
3181 if (attr
->ia_valid
& ATTR_UID
)
3182 inode
->i_uid
= attr
->ia_uid
;
3183 if (attr
->ia_valid
& ATTR_GID
)
3184 inode
->i_gid
= attr
->ia_gid
;
3185 mark_inode_dirty(inode
);
3186 error
= journal_end(&th
, inode
->i_sb
, jbegin_count
);
3192 * Relax the lock here, as it might truncate the
3193 * inode pages and wait for inode pages locks.
3194 * To release such page lock, the owner needs the
3197 reiserfs_write_unlock_once(inode
->i_sb
, depth
);
3198 if ((attr
->ia_valid
& ATTR_SIZE
) &&
3199 attr
->ia_size
!= i_size_read(inode
))
3200 error
= vmtruncate(inode
, attr
->ia_size
);
3203 setattr_copy(inode
, attr
);
3204 mark_inode_dirty(inode
);
3206 depth
= reiserfs_write_lock_once(inode
->i_sb
);
3208 if (!error
&& reiserfs_posixacl(inode
->i_sb
)) {
3209 if (attr
->ia_valid
& ATTR_MODE
)
3210 error
= reiserfs_acl_chmod(inode
);
3214 reiserfs_write_unlock_once(inode
->i_sb
, depth
);
3219 const struct address_space_operations reiserfs_address_space_operations
= {
3220 .writepage
= reiserfs_writepage
,
3221 .readpage
= reiserfs_readpage
,
3222 .readpages
= reiserfs_readpages
,
3223 .releasepage
= reiserfs_releasepage
,
3224 .invalidatepage
= reiserfs_invalidatepage
,
3225 .write_begin
= reiserfs_write_begin
,
3226 .write_end
= reiserfs_write_end
,
3227 .bmap
= reiserfs_aop_bmap
,
3228 .direct_IO
= reiserfs_direct_IO
,
3229 .set_page_dirty
= reiserfs_set_page_dirty
,