| blob | 41656d40dc5c87fc8bcfd9ec5aea4bc635ebc092 |
1 /*
2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
3 */
5 #include <linux/time.h>
6 #include <linux/fs.h>
7 #include <linux/reiserfs_fs.h>
8 #include <linux/reiserfs_acl.h>
9 #include <linux/reiserfs_xattr.h>
10 #include <linux/exportfs.h>
11 #include <linux/smp_lock.h>
12 #include <linux/pagemap.h>
13 #include <linux/highmem.h>
14 #include <linux/slab.h>
15 #include <asm/uaccess.h>
16 #include <asm/unaligned.h>
17 #include <linux/buffer_head.h>
18 #include <linux/mpage.h>
19 #include <linux/writeback.h>
20 #include <linux/quotaops.h>
21 #include <linux/swap.h>
27 {
28 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
45 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
46 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
57 /* Do quota update inside a transaction for journaled quotas. We must do that
58 * after delete_object so that quota updates go into the same transaction as
59 * stat data deletion */
66 /* check return value from reiserfs_delete_object after
67 * ending the transaction
68 */
72 /* all items of file are deleted, so we can remove "save" link */
74 * about an error here */
76 /* no object items are in the tree */
77 ;
78 }
79 out:
86 no_delete:
89 }
93 {
101 }
103 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
104 offset and type of key */
107 {
111 length);
112 }
114 //
115 // when key is 0, do not set version and short key
116 //
121 {
126 }
131 /* set_ih_free_space (ih, 0); */
132 // for directory items it is entry count, for directs and stat
133 // datas - 0xffff, for indirects - 0
135 }
137 //
138 // FIXME: we might cache recently accessed indirect item
140 // Ugh. Not too eager for that....
141 // I cut the code until such time as I see a convincing argument (benchmark).
142 // I don't want a bloated inode struct..., and I don't like code complexity....
144 /* cutting the code is fine, since it really isn't in use yet and is easy
145 ** to add back in. But, Vladimir has a really good idea here. Think
146 ** about what happens for reading a file. For each page,
147 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
148 ** an indirect item. This indirect item has X number of pointers, where
149 ** X is a big number if we've done the block allocation right. But,
150 ** we only use one or two of these pointers during each call to readpage,
151 ** needlessly researching again later on.
152 **
153 ** The size of the cache could be dynamic based on the size of the file.
154 **
155 ** I'd also like to see us cache the location the stat data item, since
156 ** we are needlessly researching for that frequently.
157 **
158 ** --chris
159 */
161 /* If this page has a file tail in it, and
162 ** it was read in by get_block_create_0, the page data is valid,
163 ** but tail is still sitting in a direct item, and we can't write to
164 ** it. So, look through this page, and check all the mapped buffers
165 ** to make sure they have valid block numbers. Any that don't need
166 ** to be unmapped, so that __block_write_begin will correctly call
167 ** reiserfs_get_block to convert the tail into an unformatted node
168 */
170 {
180 }
183 }
184 }
186 /* reiserfs_get_block does not need to allocate a block only if it has been
187 done already or non-hole position has been found in the indirect item */
191 {
198 }
201 {
203 }
207 {
209 }
211 //
212 // files which were created in the earlier version can not be longer,
213 // than 2 gb
214 //
216 {
222 }
226 {
236 /* we cannot restart while nested */
239 }
246 }
248 }
250 // it is called by get_block when create == 0. Returns block number
251 // for 'block'-th logical block of file. When it hits direct item it
252 // returns 0 (being called from bmap) or read direct item into piece
253 // of page (bh_result)
255 // Please improve the english/clarity in the comment above, as it is
256 // hard to understand.
260 {
265 b_blocknr_t blocknr;
273 // prepare the key to look for the 'block'-th block of file
285 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
286 // That there is some MMAPED data associated with it that is yet to be written to disk.
290 }
292 }
293 //
299 /* FIXME: here we could cache indirect item or part of it in
300 the inode to avoid search_by_key in case of subsequent
301 access to file */
309 }
311 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
312 // That there is some MMAPED data associated with it that is yet to be written to disk.
316 }
322 }
323 // requested data are in direct item(s)
325 // we are called by bmap. FIXME: we can not map block of file
326 // when it is stored in direct item(s)
331 }
333 /* if we've got a direct item, and the buffer or page was uptodate,
334 ** we don't want to pull data off disk again. skip to the
335 ** end, where we map the buffer and return
336 */
340 /*
341 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
342 ** pages without any buffers. If the page is up to date, we don't want
343 ** read old data off disk. Set the up to date bit on the buffer instead
344 ** and jump to the end
345 */
349 }
350 // read file tail into part of page
354 /* we only want to kmap if we are reading the tail into the page.
355 ** this is not the common case, so we don't kmap until we are
356 ** sure we need to. But, this means the item might move if
357 ** kmap schedules
358 */
367 }
368 /* make sure we don't read more bytes than actually exist in
369 ** the file. This can happen in odd cases where i_size isn't
370 ** correct, and when direct item padding results in a few
371 ** extra bytes at the end of the direct item
372 */
376 chars =
382 }
391 // we done, if read direct item is not the last item of
392 // node FIXME: we could try to check right delimiting key
393 // to see whether direct item continues in the right
394 // neighbor or rely on i_size
397 // update key to look for the next piece
401 // i/o error most likely
410 finished:
416 /* this buffer has valid data, but isn't valid for io. mapping it to
417 * block #0 tells the rest of reiserfs it just has a tail in it
418 */
422 }
424 // this is called to create file map. So, _get_block_create_0 will not
425 // read direct item
428 {
433 /* do not read the direct item */
437 }
439 /* special version of get_block that is only used by grab_tail_page right
440 ** now. It is sent to __block_write_begin, and when you try to get a
441 ** block past the end of the file (or a block from a hole) it returns
442 ** -ENOENT instead of a valid buffer. __block_write_begin expects to
443 ** be able to do i/o on the buffers returned, unless an error value
444 ** is also returned.
445 **
446 ** So, this allows __block_write_begin to be used for reading a single block
447 ** in a page. Where it does not produce a valid page for holes, or past the
448 ** end of the file. This turns out to be exactly what we need for reading
449 ** tails for conversion.
450 **
451 ** The point of the wrapper is forcing a certain value for create, even
452 ** though the VFS layer is calling this function with create==1. If you
453 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
454 ** don't use this function.
455 */
459 {
461 }
463 /* This is special helper for reiserfs_get_block in case we are executing
464 direct_IO request. */
466 sector_t iblock,
469 {
474 /* We set the b_size before reiserfs_get_block call since it is
475 referenced in convert_tail_for_hole() that may be called from
476 reiserfs_get_block() */
484 /* don't allow direct io onto tail pages */
486 /* make sure future calls to the direct io funcs for this offset
487 ** in the file fail by unmapping the buffer
488 */
491 }
492 /* Possible unpacked tail. Flush the data before pages have
493 disappeared */
506 }
507 out:
509 }
511 /*
512 ** helper function for when reiserfs_get_block is called for a hole
513 ** but the file tail is still in a direct item
514 ** bh_result is the buffer head for the hole
515 ** tail_offset is the offset of the start of the tail in the file
516 **
517 ** This calls prepare_write, which will start a new transaction
518 ** you should not be in a transaction, or have any paths held when you
519 ** call this.
520 */
523 loff_t tail_offset)
524 {
535 /* always try to read until the end of the block */
540 /* hole_page can be zero in case of direct_io, we are sure
541 that we cannot get here if we write with O_DIRECT into
542 tail page */
548 }
551 }
553 /* we don't have to make sure the conversion did not happen while
554 ** we were locking the page because anyone that could convert
555 ** must first take i_mutex.
556 **
557 ** We must fix the tail page for writing because it might have buffers
558 ** that are mapped, but have a block number of 0. This indicates tail
559 ** data that has been read directly into the page, and
560 ** __block_write_begin won't trigger a get_block in this case.
561 */
568 /* tail conversion might change the data in the page */
573 unlock:
577 }
578 out:
580 }
583 sector_t block,
587 {
590 #ifdef REISERFS_PREALLOCATE
594 }
595 #endif
597 block);
598 }
602 {
615 /* space reserved in transaction batch:
616 . 3 balancings in direct->indirect conversion
617 . 1 block involved into reiserfs_update_sd()
618 XXX in practically impossible worst case direct2indirect()
619 can incur (much) more than 3 balancings.
620 quota update for user, group */
626 loff_t new_offset =
635 }
637 /* if !create, we aren't changing the FS, so we don't need to
638 ** log anything, so we don't need to start a transaction
639 */
642 /* find number of block-th logical block of the file */
647 }
648 /*
649 * if we're already in a transaction, make sure to close
650 * any new transactions we start in this func
651 */
656 /* If file is of such a size, that it might have a tail and tails are enabled
657 ** we should mark it as possibly needing tail packing on close
658 */
665 /* set the key of the first byte in the 'block'-th block of file */
668 start_trans:
673 }
675 }
676 research:
682 }
694 /* we have to allocate block for the unformatted node */
698 }
700 repeat =
705 /* restart the transaction to give the journal a chance to free
706 ** some blocks. releases the path, so we have to go back to
707 ** research if we succeed on the second try
708 */
713 repeat =
719 }
722 else
725 }
730 }
731 }
734 b_blocknr_t unfm_ptr;
735 /* 'block'-th block is in the file already (there is
736 corresponding cell in some indirect item). But it may be
737 zero unformatted node pointer (hole) */
740 /* use allocated block to plug the hole */
745 bh);
747 }
756 }
765 /* the item was found, so new blocks were not added to the file
766 ** there is no need to make sure the inode is updated with this
767 ** transaction
768 */
770 }
775 }
777 /* desired position is not found or is in the direct item. We have
778 to append file with holes up to 'block'-th block converting
779 direct items to indirect one if necessary */
786 /* indirect item has to be inserted */
792 /* we are going to add 'block'-th block to the file. Use
793 allocated block for that */
799 }
804 retval =
811 }
812 //mark_tail_converted (inode);
814 /* direct item has to be converted */
815 loff_t tail_offset;
817 tail_offset =
821 /* direct item we just found fits into block we have
822 to map. Convert it into unformatted node: use
823 bh_result for the conversion */
829 /* we have to padd file tail stored in direct item(s)
830 up to block size and convert it to unformatted
831 node. FIXME: this should also get into page cache */
834 /*
835 * ugly, but we can only end the transaction if
836 * we aren't nested
837 */
840 retval =
841 reiserfs_end_persistent_transaction
846 }
848 retval =
850 tail_offset);
855 "convert tail failed "
858 retval);
860 /* the bitmap, the super, and the stat data == 3 */
865 inode,
866 allocated_block_nr,
868 }
870 }
872 }
873 retval =
875 tail_offset);
881 }
882 /* it is important the set_buffer_uptodate is done after
883 ** the direct2indirect. The buffer might contain valid
884 ** data newer than the data on disk (read by readpage, changed,
885 ** and then sent here by writepage). direct2indirect needs
886 ** to know if unbh was already up to date, so it can decide
887 ** if the data in unbh needs to be replaced with data from
888 ** the disk
889 */
892 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
893 buffer will disappear shortly, so it should not be added to
894 */
896 /* we've converted the tail, so we must
897 ** flush unbh before the transaction commits
898 */
901 /* mark it dirty now to prevent commit_write from adding
902 ** this buffer to the inode's dirty buffer list
903 */
904 /*
905 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
906 * It's still atomic, but it sets the page dirty too,
907 * which makes it eligible for writeback at any time by the
908 * VM (which was also the case with __mark_buffer_dirty())
909 */
911 }
913 /* append indirect item with holes if needed, when appending
914 pointer to 'block'-th block use block, which is already
915 allocated */
918 // one block which is a fastpath
920 __u64 max_to_insert =
922 UNFM_P_SIZE;
923 __u64 blocks_needed;
927 /* indirect item has to be appended, set up key of that position */
933 //pos_in_item * inode->i_sb->s_blocksize,
938 blocks_needed =
942 s_blocksize_bits);
952 }
953 }
955 /* we are going to add target block to the file. Use allocated
956 block for that */
964 /* paste hole to the indirect item */
965 /* If kmalloc failed, max_to_insert becomes zero and it means we
966 only have space for one block */
967 blocks_needed =
969 }
970 retval =
973 UNFM_P_SIZE *
974 blocks_needed);
983 }
985 /* We need to mark new file size in case this function will be
986 interrupted/aborted later on. And we may do this only for
987 holes. */
990 }
991 }
996 /* this loop could log more blocks than we had originally asked
997 ** for. So, we have to allow the transaction to end if it is
998 ** too big or too full. Update the inode so things are
999 ** consistent if we crash before the function returns
1000 **
1001 ** release the path so that anybody waiting on the path before
1002 ** ending their transaction will be able to continue.
1003 */
1008 }
1009 /*
1010 * inserting indirect pointers for a hole can take a
1011 * long time. reschedule if needed and also release the write
1012 * lock for others.
1013 */
1018 }
1024 }
1034 }
1043 failure:
1051 }
1056 }
1058 static int
1061 {
1063 }
1065 /* Compute real number of used bytes by file
1066 * Following three functions can go away when we'll have enough space in stat item
1067 */
1069 {
1076 /* End of file is also in full block with indirect reference, so round
1077 ** up to the next block.
1078 **
1079 ** there is just no way to know if the tail is actually packed
1080 ** on the file, so we have to assume it isn't. When we pack the
1081 ** tail, we add 4 bytes to pretend there really is an unformatted
1082 ** node pointer
1083 */
1084 bytes =
1087 sd_size;
1089 }
1093 {
1097 }
1100 }
1102 /* Compute number of blocks used by file in ReiserFS counting */
1104 {
1108 /* keeps fsck and non-quota versions of reiserfs happy */
1111 }
1113 /* files from before the quota patch might i_blocks such that
1114 ** bytes < real_space. Deal with that here to prevent it from
1115 ** going negative.
1116 */
1120 }
1122 //
1123 // BAD: new directories have stat data of new type and all other items
1124 // of old type. Version stored in the inode says about body items, so
1125 // in update_stat_data we can not rely on inode, but have to check
1126 // item version directly
1127 //
1129 // called by read_locked_inode
1131 {
1134 __u32 rdev;
1135 //int version = ITEM_VERSION_1;
1174 // there was a bug in <=3.5.23 when i_blocks could take negative
1175 // values. Starting from 3.5.17 this value could even be stored in
1176 // stat data. For such files we set i_blocks based on file
1177 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1178 // only updated if file's inode will ever change
1180 }
1185 /* an early bug in the quota code can give us an odd number for the
1186 ** block count. This is incorrect, fix it here.
1187 */
1190 }
1193 SD_V1_SIZE));
1194 /* nopack is initially zero for v1 objects. For v2 objects,
1195 nopack is initialised from sd_attrs */
1198 // new stat data found, but object may have old items
1199 // (directories and symlinks)
1218 else
1223 else
1229 SD_V2_SIZE));
1230 /* read persistent inode attributes from sd and initialise
1231 generic inode flags from them */
1234 }
1251 }
1252 }
1254 // update new stat data with inode fields
1256 {
1258 __u16 flags;
1271 else
1276 }
1278 // used to copy inode's fields to old stat data
1280 {
1294 else
1297 // Sigh. i_first_direct_byte is back
1300 }
1302 /* NOTE, you must prepare the buffer head before sending it here,
1303 ** and then log it after the call
1304 */
1306 loff_t size)
1307 {
1319 // path points to old stat data
1323 }
1326 }
1330 {
1344 /* look for the object's stat data */
1348 "i/o failure occurred trying to "
1351 }
1356 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1358 }
1360 "stat data of object %k (nlink == %d) "
1363 pos);
1366 }
1368 /* sigh, prepare_for_journal might schedule. When it schedules the
1369 ** FS might change. We have to detect that, and loop back to the
1370 ** search if the stat data item has moved
1371 */
1381 }
1383 }
1388 }
1390 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1391 ** does a make_bad_inode when things go wrong. But, we need to make sure
1392 ** and clear the key in the private portion of the inode, otherwise a
1393 ** corresponding iput might try to delete whatever object the inode last
1394 ** represented.
1395 */
1397 {
1400 }
1402 //
1403 // initially this function was derived from minix or ext2's analog and
1404 // evolved as the prototype did
1405 //
1408 {
1413 }
1415 /* looks for stat data in the tree, and fills up the fields of in-core
1416 inode stat data fields */
1419 {
1427 /* set version 1, version 2 could be used too, because stat data
1428 key is the same in both versions */
1435 /* look for the object's stat data */
1439 "i/o failure occurred trying to find "
1443 }
1445 /* a stale NFS handle can trigger this without it being an error */
1450 }
1454 /* It is possible that knfsd is trying to access inode of a file
1455 that is being removed from the disk by some other thread. As we
1456 update sd on unlink all that is required is to check for nlink
1457 here. This bug was first found by Sizif when debugging
1458 SquidNG/Butterfly, forgotten, and found again after Philippe
1459 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1461 More logical fix would require changes in fs/inode.c:iput() to
1462 remove inode from hash-table _after_ fs cleaned disk stuff up and
1463 in iget() to return NULL if I_FREEING inode is found in
1464 hash-table. */
1465 /* Currently there is one place where it's ok to meet inode with
1466 nlink==0: processing of open-unlinked and half-truncated files
1467 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1471 "dead inode read from disk %K. "
1475 }
1479 }
1481 /**
1482 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1483 *
1484 * @inode: inode from hash table to check
1485 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1486 *
1487 * This function is called by iget5_locked() to distinguish reiserfs inodes
1488 * having the same inode numbers. Such inodes can only exist due to some
1489 * error condition. One of them should be bad. Inodes with identical
1490 * inode numbers (objectids) are distinguished by parent directory ids.
1491 *
1492 */
1494 {
1498 /* args is already in CPU order */
1501 }
1504 {
1521 }
1524 /* either due to i/o error or a stale NFS handle */
1527 }
1529 }
1534 {
1546 }
1550 }
1554 {
1555 /* fhtype happens to reflect the number of u32s encoded.
1556 * due to a bug in earlier code, fhtype might indicate there
1557 * are more u32s then actually fitted.
1558 * so if fhtype seems to be more than len, reduce fhtype.
1559 * Valid types are:
1560 * 2 - objectid + dir_id - legacy support
1561 * 3 - objectid + dir_id + generation
1562 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1563 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1564 * 6 - as above plus generation of directory
1565 * 6 does not fit in NFSv2 handles
1566 */
1573 }
1577 }
1581 {
1589 }
1593 {
1604 /* no room for directory info? return what we've stored so far */
1616 }
1619 }
1621 /* looks for stat data, then copies fields to it, marks the buffer
1622 containing stat data as dirty */
1623 /* reiserfs inodes are never really dirty, since the dirty inode call
1624 ** always logs them. This call allows the VFS inode marking routines
1625 ** to properly mark inodes for datasync and such, but only actually
1626 ** does something when called for a synchronous update.
1627 */
1629 {
1635 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1636 ** these cases are just when the system needs ram, not when the
1637 ** inode needs to reach disk for safety, and they can safely be
1638 ** ignored because the altered inode has already been logged.
1639 */
1645 }
1647 }
1649 }
1651 /* stat data of new object is inserted already, this inserts the item
1652 containing "." and ".." entries */
1657 {
1670 /* compose item head for new item. Directories consist of items of
1671 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1672 is done by reiserfs_new_inode */
1689 }
1691 /* look for place in the tree for new item */
1697 }
1704 }
1706 /* insert item, that is empty directory item */
1708 }
1710 /* stat data of object has been inserted, this inserts the item
1711 containing the body of symlink */
1712 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1716 {
1731 /* look for place in the tree for new item */
1737 }
1744 }
1746 /* insert item, that is body of symlink */
1748 }
1750 /* inserts the stat data into the tree, and then calls
1751 reiserfs_new_directory (to insert ".", ".." item if new object is
1752 directory) or reiserfs_new_symlink (to insert symlink body if new
1753 object is symlink) or nothing (if new object is regular file)
1755 NOTE! uid and gid must already be set in the inode. If we return
1756 non-zero due to an error, we have to drop the quota previously allocated
1757 for the fresh inode. This can only be done outside a transaction, so
1758 if we return non-zero, we also end the transaction. */
1761 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1762 strlen (symname) for symlinks) */
1766 {
1785 }
1789 /* item head of new item */
1795 }
1800 else
1809 }
1811 /* not a perfect generation count, as object ids can be reused, but
1812 ** this is as good as reiserfs can do right now.
1813 ** note that the private part of inode isn't filled in yet, we have
1814 ** to use the directory.
1815 */
1817 else
1818 #if defined( USE_INODE_GENERATION_COUNTER )
1821 #else
1823 #endif
1825 /* fill stat data */
1828 /* uid and gid must already be set by the caller for quota init */
1830 /* symlink cannot be immutable or append only, right? */
1852 /* key to search for correct place for new stat data */
1857 /* find proper place for inserting of stat data */
1862 }
1867 }
1871 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1874 }
1878 }
1879 // store in in-core inode the key of stat data and version all
1880 // object items will have (directory items will have old offset
1881 // format, other new objects will consist of new items)
1884 else
1888 else
1891 /* insert the stat data into the tree */
1892 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1895 #endif
1896 retval =
1903 }
1904 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1907 #endif
1909 /* insert item with "." and ".." */
1910 retval =
1912 }
1915 /* insert body of symlink */
1918 retval =
1920 i_size);
1921 }
1927 }
1936 }
1939 "ACLs aren't enabled in the fs, "
1954 }
1955 }
1962 /* it looks like you can easily compress these two goto targets into
1963 * one. Keeping it like this doesn't actually hurt anything, and they
1964 * are place holders for what the quota code actually needs.
1965 */
1966 out_bad_inode:
1967 /* Invalidate the object, nothing was inserted yet */
1970 /* Quota change must be inside a transaction for journaling */
1973 out_end_trans:
1975 /* Drop can be outside and it needs more credits so it's better to have it outside */
1980 out_inserted_sd:
1986 }
1988 /*
1989 ** finds the tail page in the page cache,
1990 ** reads the last block in.
1991 **
1992 ** On success, page_result is set to a locked, pinned page, and bh_result
1993 ** is set to an up to date buffer for the last block in the file. returns 0.
1994 **
1995 ** tail conversion is not done, so bh_result might not be valid for writing
1996 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1997 ** trying to write the block.
1998 **
1999 ** on failure, nonzero is returned, page_result and bh_result are untouched.
2000 */
2004 {
2006 /* we want the page with the last byte in the file,
2007 ** not the page that will hold the next byte for appending
2008 */
2019 /* we know that we are only called with inode->i_size > 0.
2020 ** we also know that a file tail can never be as big as a block
2021 ** If i_size % blocksize == 0, our file is currently block aligned
2022 ** and it won't need converting or zeroing after a truncate.
2023 */
2026 }
2031 }
2032 /* start within the page of the last block in the file */
2036 reiserfs_get_block_create_0);
2045 }
2051 /* note, this should never happen, prepare_write should
2052 ** be taking care of this for us. If the buffer isn't up to date,
2053 ** I've screwed up the code to find the buffer, or the code to
2054 ** call prepare_write
2055 */
2060 }
2064 out:
2067 unlock:
2071 }
2073 /*
2074 ** vfs version of truncate file. Must NOT be called with
2075 ** a transaction already started.
2076 **
2077 ** some code taken from block_truncate_page
2078 */
2080 {
2082 /* we want the offset for the first byte after the end of the file */
2097 // -ENOENT means we truncated past the end of the file,
2098 // and get_block_create_0 could not find a block to read in,
2099 // which is ok.
2103 error);
2106 }
2107 }
2109 /* so, if page != NULL, we have a buffer head for the offset at
2110 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2111 ** then we have an unformatted node. Otherwise, we have a direct item,
2112 ** and no zeroing is required on disk. We zero after the truncate,
2113 ** because the truncate might pack the item anyway
2114 ** (it will unmap bh if it packs).
2115 */
2116 /* it is enough to reserve space in transaction for 2 balancings:
2117 one for "save" link adding and another for the first
2118 cut_from_item. 1 is for update_sd */
2125 /* we are doing real truncate: if the system crashes before the last
2126 transaction of truncating gets committed - on reboot the file
2127 either appears truncated properly or not truncated at all */
2130 error =
2135 /* check reiserfs_do_truncate after ending the transaction */
2139 }
2145 }
2149 /* if we are not on a block boundary */
2155 }
2156 }
2159 }
2164 out:
2168 }
2173 }
2178 {
2196 /* catch places below that try to log something without starting a trans */
2201 }
2204 start_over:
2208 research:
2213 }
2220 /* we've found an unformatted node */
2225 }
2227 /* crap, we are writing to a hole */
2230 }
2243 /* vs-3050 is gone, no need to drop the path */
2252 bh);
2254 }
2255 }
2263 }
2266 copy_size);
2272 /* are there still bytes left? */
2277 copy_size);
2279 }
2285 }
2288 out:
2295 }
2298 /* this is where we fill in holes in the file. */
2301 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2306 /* get_block failed to find a mapped unformatted node. */
2309 }
2310 }
2311 }
2315 /* we've copied data from the page into the direct item, so the
2316 * buffer in the page is now clean, mark it to reflect that.
2317 */
2321 }
2323 }
2325 /*
2326 * mason@suse.com: updated in 2.5.54 to follow the same general io
2327 * start/recovery path as __block_write_full_page, along with special
2328 * code to handle reiserfs tails.
2329 */
2332 {
2337 sector_t last_block;
2347 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2352 }
2354 /* The page dirty bit is cleared before writepage is called, which
2355 * means we have to tell create_empty_buffers to make dirty buffers
2356 * The page really should be up to date at this point, so tossing
2357 * in the BH_Uptodate is just a sanity check.
2358 */
2362 }
2365 /* last page in the file, zero out any contents past the
2366 ** last byte in the file
2367 */
2372 /* no file contents in this page */
2376 }
2378 }
2382 /* first map all the buffers, logging any direct items we find */
2385 /*
2386 * This can happen when the block size is less than
2387 * the page size. The corresponding bytes in the page
2388 * were zero filled above
2389 */
2396 /* not mapped yet, or it points to a direct item, search
2397 * the btree for the mapping info, and log any direct
2398 * items found
2399 */
2402 }
2403 }
2405 block++;
2408 /*
2409 * we start the transaction after map_block_for_writepage,
2410 * because it can create holes in the file (an unbounded operation).
2411 * starting it here, we can make a reliable estimate for how many
2412 * blocks we're going to log
2413 */
2421 }
2423 }
2424 /* now go through and lock any dirty buffers on the page */
2436 }
2437 /* from this point on, we know the buffer is mapped to a
2438 * real block and not a direct item
2439 */
2446 }
2447 }
2452 }
2460 }
2465 /*
2466 * since any buffer might be the only dirty buffer on the page,
2467 * the first submit_bh can bring the page out of writeback.
2468 * be careful with the buffers.
2469 */
2474 nr++;
2475 }
2481 done:
2483 /*
2484 * if this page only had a direct item, it is very possible for
2485 * no io to be required without there being an error. Or,
2486 * someone else could have locked them and sent them down the
2487 * pipe without locking the page
2488 */
2494 }
2500 }
2503 fail:
2504 /* catches various errors, we need to make sure any valid dirty blocks
2505 * get to the media. The page is currently locked and not marked for
2506 * writeback
2507 */
2516 /*
2517 * clear any dirty bits that might have come from getting
2518 * attached to a dirty page
2519 */
2521 }
2533 nr++;
2534 }
2539 }
2542 {
2544 }
2547 {
2551 }
2554 {
2557 }
2563 {
2566 pgoff_t index;
2574 pos ++;
2576 }
2589 journal_info;
2594 }
2598 /* this gets a little ugly. If reiserfs_get_block returned an
2599 * error and left a transacstion running, we've got to close it,
2600 * and we've got to free handle if it was a persistent transaction.
2601 *
2602 * But, if we had nested into an existing transaction, we need
2603 * to just drop the ref count on the handle.
2604 *
2605 * If old_ref == 0, the transaction is from reiserfs_get_block,
2606 * and it was a persistent trans. Otherwise, it was nested above.
2607 */
2618 }
2619 }
2620 }
2624 /* Truncate allocated blocks */
2626 }
2628 }
2631 {
2644 journal_info;
2649 }
2654 /* this gets a little ugly. If reiserfs_get_block returned an
2655 * error and left a transacstion running, we've got to close it,
2656 * and we've got to free handle if it was a persistent transaction.
2657 *
2658 * But, if we had nested into an existing transaction, we need
2659 * to just drop the ref count on the handle.
2660 *
2661 * If old_ref == 0, the transaction is from reiserfs_get_block,
2662 * and it was a persistent trans. Otherwise, it was nested above.
2663 */
2674 }
2675 }
2676 }
2679 }
2682 {
2684 }
2689 {
2699 pos ++;
2704 else
2713 }
2718 /* generic_commit_write does this for us, but does not update the
2719 ** transaction tracking stuff when the size changes. So, we have
2720 ** to do the i_size updates here.
2721 */
2726 /* If the file have grown beyond the border where it
2727 can have a tail, unmark it as needing a tail
2728 packing */
2741 /*
2742 * this will just nest into our transaction. It's important
2743 * to use mark_inode_dirty so the inode gets pushed around on the
2744 * dirty lists, and so that O_SYNC works as expected
2745 */
2752 }
2757 }
2763 }
2765 out:
2776 journal_error:
2783 }
2785 }
2789 {
2802 }
2805 /* generic_commit_write does this for us, but does not update the
2806 ** transaction tracking stuff when the size changes. So, we have
2807 ** to do the i_size updates here.
2808 */
2811 /* If the file have grown beyond the border where it
2812 can have a tail, unmark it as needing a tail
2813 packing */
2826 /*
2827 * this will just nest into our transaction. It's important
2828 * to use mark_inode_dirty so the inode gets pushed around on the
2829 * dirty lists, and so that O_SYNC works as expected
2830 */
2837 }
2844 }
2846 out:
2849 journal_error:
2854 }
2857 }
2860 {
2864 else
2868 else
2872 else
2876 else
2880 else
2882 }
2883 }
2886 {
2890 else
2894 else
2898 else
2902 else
2904 }
2905 }
2907 /* decide if this buffer needs to stay around for data logging or ordered
2908 ** write purposes
2909 */
2911 {
2919 }
2920 /* the page is locked, and the only places that log a data buffer
2921 * also lock the page.
2922 */
2924 /*
2925 * very conservative, leave the buffer pinned if
2926 * anyone might need it.
2927 */
2930 }
2935 /* why is this safe?
2936 * reiserfs_setattr updates i_size in the on disk
2937 * stat data before allowing vmtruncate to be called.
2938 *
2939 * If buffer was put onto the ordered list for this
2940 * transaction, we know for sure either this transaction
2941 * or an older one already has updated i_size on disk,
2942 * and this ordered data won't be referenced in the file
2943 * if we crash.
2944 *
2945 * if the buffer was put onto the ordered list for an older
2946 * transaction, we need to leave it around
2947 */
2951 }
2952 free_jh:
2955 }
2959 }
2961 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2963 {
2983 /*
2984 * is this block fully invalidated?
2985 */
2989 else
2991 }
2996 /*
2997 * We release buffers only if the entire page is being invalidated.
2998 * The get_block cached value has been unconditionally invalidated,
2999 * so real IO is not possible anymore.
3000 */
3003 /* maybe should BUG_ON(!ret); - neilb */
3004 }
3005 out:
3007 }
3010 {
3015 }
3017 }
3019 /*
3020 * Returns 1 if the page's buffers were dropped. The page is locked.
3021 *
3022 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3023 * in the buffers at page_buffers(page).
3024 *
3025 * even in -o notail mode, we can't be sure an old mount without -o notail
3026 * didn't create files with tails.
3027 */
3029 {
3047 }
3048 }
3055 }
3057 /* We thank Mingming Cao for helping us understand in great detail what
3058 to do in this section of the code. */
3062 {
3065 ssize_t ret;
3071 /*
3072 * In case of error extending write may have instantiated a few
3073 * blocks outside i_size. Trim these off again.
3074 */
3081 }
3084 }
3087 {
3097 /* must be turned off for recursive notify_change calls */
3105 /* version 2 items will be caught by the s_maxbytes check
3106 ** done for us in vmtruncate
3107 */
3112 }
3113 /* fill in hole pointers in the expanding truncate case. */
3119 /* we're changing at most 2 bitmaps, inode + super */
3124 }
3127 }
3130 /*
3131 * file size is changed, ctime and mtime are
3132 * to be updated
3133 */
3135 }
3136 }
3141 /* stat data of format v3.5 has 16 bit uid and gid */
3144 }
3160 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3168 }
3170 /* Update corresponding info in inode so that everything is in
3171 * one transaction */
3180 }
3182 /*
3183 * Relax the lock here, as it might truncate the
3184 * inode pages and wait for inode pages locks.
3185 * To release such page lock, the owner needs the
3186 * reiserfs lock
3187 */
3196 }
3202 }
3204 out:
3208 }
3222 };