| blob | ad62bdbb451ee77f1cfcee9be8cae70537168667 |
1 /*
2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
3 */
5 #include <linux/time.h>
6 #include <linux/fs.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>
21 #include <linux/aio.h>
27 {
28 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
42 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
43 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 */
64 }
69 /* check return value from reiserfs_delete_object after
70 * ending the transaction
71 */
75 /* all items of file are deleted, so we can remove "save" link */
77 * about an error here */
78 out:
81 /* no object items are in the tree */
82 ;
83 }
89 no_delete:
92 }
96 {
104 }
106 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
107 offset and type of key */
110 {
114 length);
115 }
117 //
118 // when key is 0, do not set version and short key
119 //
124 {
129 }
134 /* set_ih_free_space (ih, 0); */
135 // for directory items it is entry count, for directs and stat
136 // datas - 0xffff, for indirects - 0
138 }
140 //
141 // FIXME: we might cache recently accessed indirect item
143 // Ugh. Not too eager for that....
144 // I cut the code until such time as I see a convincing argument (benchmark).
145 // I don't want a bloated inode struct..., and I don't like code complexity....
147 /* cutting the code is fine, since it really isn't in use yet and is easy
148 ** to add back in. But, Vladimir has a really good idea here. Think
149 ** about what happens for reading a file. For each page,
150 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
151 ** an indirect item. This indirect item has X number of pointers, where
152 ** X is a big number if we've done the block allocation right. But,
153 ** we only use one or two of these pointers during each call to readpage,
154 ** needlessly researching again later on.
155 **
156 ** The size of the cache could be dynamic based on the size of the file.
157 **
158 ** I'd also like to see us cache the location the stat data item, since
159 ** we are needlessly researching for that frequently.
160 **
161 ** --chris
162 */
164 /* If this page has a file tail in it, and
165 ** it was read in by get_block_create_0, the page data is valid,
166 ** but tail is still sitting in a direct item, and we can't write to
167 ** it. So, look through this page, and check all the mapped buffers
168 ** to make sure they have valid block numbers. Any that don't need
169 ** to be unmapped, so that __block_write_begin will correctly call
170 ** reiserfs_get_block to convert the tail into an unformatted node
171 */
173 {
183 }
186 }
187 }
189 /* reiserfs_get_block does not need to allocate a block only if it has been
190 done already or non-hole position has been found in the indirect item */
194 {
201 }
204 {
206 }
210 {
212 }
214 //
215 // files which were created in the earlier version can not be longer,
216 // than 2 gb
217 //
219 {
225 }
229 {
239 /* we cannot restart while nested */
242 }
249 }
251 }
253 // it is called by get_block when create == 0. Returns block number
254 // for 'block'-th logical block of file. When it hits direct item it
255 // returns 0 (being called from bmap) or read direct item into piece
256 // of page (bh_result)
258 // Please improve the english/clarity in the comment above, as it is
259 // hard to understand.
263 {
268 b_blocknr_t blocknr;
276 // prepare the key to look for the 'block'-th block of file
288 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
289 // That there is some MMAPED data associated with it that is yet to be written to disk.
293 }
295 }
296 //
302 /* FIXME: here we could cache indirect item or part of it in
303 the inode to avoid search_by_key in case of subsequent
304 access to file */
312 }
314 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
315 // That there is some MMAPED data associated with it that is yet to be written to disk.
319 }
325 }
326 // requested data are in direct item(s)
328 // we are called by bmap. FIXME: we can not map block of file
329 // when it is stored in direct item(s)
334 }
336 /* if we've got a direct item, and the buffer or page was uptodate,
337 ** we don't want to pull data off disk again. skip to the
338 ** end, where we map the buffer and return
339 */
343 /*
344 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
345 ** pages without any buffers. If the page is up to date, we don't want
346 ** read old data off disk. Set the up to date bit on the buffer instead
347 ** and jump to the end
348 */
352 }
353 // read file tail into part of page
357 /* we only want to kmap if we are reading the tail into the page.
358 ** this is not the common case, so we don't kmap until we are
359 ** sure we need to. But, this means the item might move if
360 ** kmap schedules
361 */
370 }
371 /* make sure we don't read more bytes than actually exist in
372 ** the file. This can happen in odd cases where i_size isn't
373 ** correct, and when direct item padding results in a few
374 ** extra bytes at the end of the direct item
375 */
379 chars =
385 }
394 // we done, if read direct item is not the last item of
395 // node FIXME: we could try to check right delimiting key
396 // to see whether direct item continues in the right
397 // neighbor or rely on i_size
400 // update key to look for the next piece
404 // i/o error most likely
413 finished:
419 /* this buffer has valid data, but isn't valid for io. mapping it to
420 * block #0 tells the rest of reiserfs it just has a tail in it
421 */
425 }
427 // this is called to create file map. So, _get_block_create_0 will not
428 // read direct item
431 {
436 /* do not read the direct item */
440 }
442 /* special version of get_block that is only used by grab_tail_page right
443 ** now. It is sent to __block_write_begin, and when you try to get a
444 ** block past the end of the file (or a block from a hole) it returns
445 ** -ENOENT instead of a valid buffer. __block_write_begin expects to
446 ** be able to do i/o on the buffers returned, unless an error value
447 ** is also returned.
448 **
449 ** So, this allows __block_write_begin to be used for reading a single block
450 ** in a page. Where it does not produce a valid page for holes, or past the
451 ** end of the file. This turns out to be exactly what we need for reading
452 ** tails for conversion.
453 **
454 ** The point of the wrapper is forcing a certain value for create, even
455 ** though the VFS layer is calling this function with create==1. If you
456 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
457 ** don't use this function.
458 */
462 {
464 }
466 /* This is special helper for reiserfs_get_block in case we are executing
467 direct_IO request. */
469 sector_t iblock,
472 {
477 /* We set the b_size before reiserfs_get_block call since it is
478 referenced in convert_tail_for_hole() that may be called from
479 reiserfs_get_block() */
487 /* don't allow direct io onto tail pages */
489 /* make sure future calls to the direct io funcs for this offset
490 ** in the file fail by unmapping the buffer
491 */
494 }
495 /* Possible unpacked tail. Flush the data before pages have
496 disappeared */
509 }
510 out:
512 }
514 /*
515 ** helper function for when reiserfs_get_block is called for a hole
516 ** but the file tail is still in a direct item
517 ** bh_result is the buffer head for the hole
518 ** tail_offset is the offset of the start of the tail in the file
519 **
520 ** This calls prepare_write, which will start a new transaction
521 ** you should not be in a transaction, or have any paths held when you
522 ** call this.
523 */
526 loff_t tail_offset)
527 {
538 /* always try to read until the end of the block */
543 /* hole_page can be zero in case of direct_io, we are sure
544 that we cannot get here if we write with O_DIRECT into
545 tail page */
551 }
554 }
556 /* we don't have to make sure the conversion did not happen while
557 ** we were locking the page because anyone that could convert
558 ** must first take i_mutex.
559 **
560 ** We must fix the tail page for writing because it might have buffers
561 ** that are mapped, but have a block number of 0. This indicates tail
562 ** data that has been read directly into the page, and
563 ** __block_write_begin won't trigger a get_block in this case.
564 */
571 /* tail conversion might change the data in the page */
576 unlock:
580 }
581 out:
583 }
586 sector_t block,
590 {
593 #ifdef REISERFS_PREALLOCATE
597 }
598 #endif
600 block);
601 }
605 {
617 /* space reserved in transaction batch:
618 . 3 balancings in direct->indirect conversion
619 . 1 block involved into reiserfs_update_sd()
620 XXX in practically impossible worst case direct2indirect()
621 can incur (much) more than 3 balancings.
622 quota update for user, group */
628 loff_t new_offset =
637 }
639 /* if !create, we aren't changing the FS, so we don't need to
640 ** log anything, so we don't need to start a transaction
641 */
644 /* find number of block-th logical block of the file */
649 }
650 /*
651 * if we're already in a transaction, make sure to close
652 * any new transactions we start in this func
653 */
658 /* If file is of such a size, that it might have a tail and tails are enabled
659 ** we should mark it as possibly needing tail packing on close
660 */
667 /* set the key of the first byte in the 'block'-th block of file */
670 start_trans:
675 }
677 }
678 research:
684 }
696 /* we have to allocate block for the unformatted node */
700 }
702 repeat =
707 /* restart the transaction to give the journal a chance to free
708 ** some blocks. releases the path, so we have to go back to
709 ** research if we succeed on the second try
710 */
715 repeat =
721 }
724 else
727 }
732 }
733 }
736 b_blocknr_t unfm_ptr;
737 /* 'block'-th block is in the file already (there is
738 corresponding cell in some indirect item). But it may be
739 zero unformatted node pointer (hole) */
742 /* use allocated block to plug the hole */
747 bh);
749 }
758 }
767 /* the item was found, so new blocks were not added to the file
768 ** there is no need to make sure the inode is updated with this
769 ** transaction
770 */
772 }
777 }
779 /* desired position is not found or is in the direct item. We have
780 to append file with holes up to 'block'-th block converting
781 direct items to indirect one if necessary */
788 /* indirect item has to be inserted */
794 /* we are going to add 'block'-th block to the file. Use
795 allocated block for that */
801 }
806 retval =
813 }
814 //mark_tail_converted (inode);
816 /* direct item has to be converted */
817 loff_t tail_offset;
819 tail_offset =
823 /* direct item we just found fits into block we have
824 to map. Convert it into unformatted node: use
825 bh_result for the conversion */
831 /* we have to padd file tail stored in direct item(s)
832 up to block size and convert it to unformatted
833 node. FIXME: this should also get into page cache */
836 /*
837 * ugly, but we can only end the transaction if
838 * we aren't nested
839 */
842 retval =
843 reiserfs_end_persistent_transaction
848 }
850 retval =
852 tail_offset);
857 "convert tail failed "
860 retval);
862 /* the bitmap, the super, and the stat data == 3 */
867 inode,
868 allocated_block_nr,
870 }
872 }
874 }
875 retval =
877 tail_offset);
883 }
884 /* it is important the set_buffer_uptodate is done after
885 ** the direct2indirect. The buffer might contain valid
886 ** data newer than the data on disk (read by readpage, changed,
887 ** and then sent here by writepage). direct2indirect needs
888 ** to know if unbh was already up to date, so it can decide
889 ** if the data in unbh needs to be replaced with data from
890 ** the disk
891 */
894 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
895 buffer will disappear shortly, so it should not be added to
896 */
898 /* we've converted the tail, so we must
899 ** flush unbh before the transaction commits
900 */
903 /* mark it dirty now to prevent commit_write from adding
904 ** this buffer to the inode's dirty buffer list
905 */
906 /*
907 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
908 * It's still atomic, but it sets the page dirty too,
909 * which makes it eligible for writeback at any time by the
910 * VM (which was also the case with __mark_buffer_dirty())
911 */
913 }
915 /* append indirect item with holes if needed, when appending
916 pointer to 'block'-th block use block, which is already
917 allocated */
920 // one block which is a fastpath
922 __u64 max_to_insert =
924 UNFM_P_SIZE;
925 __u64 blocks_needed;
929 /* indirect item has to be appended, set up key of that position */
935 //pos_in_item * inode->i_sb->s_blocksize,
940 blocks_needed =
944 s_blocksize_bits);
954 }
955 }
957 /* we are going to add target block to the file. Use allocated
958 block for that */
966 /* paste hole to the indirect item */
967 /* If kmalloc failed, max_to_insert becomes zero and it means we
968 only have space for one block */
969 blocks_needed =
971 }
972 retval =
975 UNFM_P_SIZE *
976 blocks_needed);
985 }
987 /* We need to mark new file size in case this function will be
988 interrupted/aborted later on. And we may do this only for
989 holes. */
992 }
993 }
998 /* this loop could log more blocks than we had originally asked
999 ** for. So, we have to allow the transaction to end if it is
1000 ** too big or too full. Update the inode so things are
1001 ** consistent if we crash before the function returns
1002 **
1003 ** release the path so that anybody waiting on the path before
1004 ** ending their transaction will be able to continue.
1005 */
1010 }
1011 /*
1012 * inserting indirect pointers for a hole can take a
1013 * long time. reschedule if needed and also release the write
1014 * lock for others.
1015 */
1022 }
1032 }
1041 failure:
1049 }
1054 }
1056 static int
1059 {
1061 }
1063 /* Compute real number of used bytes by file
1064 * Following three functions can go away when we'll have enough space in stat item
1065 */
1067 {
1074 /* End of file is also in full block with indirect reference, so round
1075 ** up to the next block.
1076 **
1077 ** there is just no way to know if the tail is actually packed
1078 ** on the file, so we have to assume it isn't. When we pack the
1079 ** tail, we add 4 bytes to pretend there really is an unformatted
1080 ** node pointer
1081 */
1082 bytes =
1085 sd_size;
1087 }
1091 {
1095 }
1098 }
1100 /* Compute number of blocks used by file in ReiserFS counting */
1102 {
1106 /* keeps fsck and non-quota versions of reiserfs happy */
1109 }
1111 /* files from before the quota patch might i_blocks such that
1112 ** bytes < real_space. Deal with that here to prevent it from
1113 ** going negative.
1114 */
1118 }
1120 //
1121 // BAD: new directories have stat data of new type and all other items
1122 // of old type. Version stored in the inode says about body items, so
1123 // in update_stat_data we can not rely on inode, but have to check
1124 // item version directly
1125 //
1127 // called by read_locked_inode
1129 {
1132 __u32 rdev;
1133 //int version = ITEM_VERSION_1;
1172 // there was a bug in <=3.5.23 when i_blocks could take negative
1173 // values. Starting from 3.5.17 this value could even be stored in
1174 // stat data. For such files we set i_blocks based on file
1175 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1176 // only updated if file's inode will ever change
1178 }
1183 /* an early bug in the quota code can give us an odd number for the
1184 ** block count. This is incorrect, fix it here.
1185 */
1188 }
1191 SD_V1_SIZE));
1192 /* nopack is initially zero for v1 objects. For v2 objects,
1193 nopack is initialised from sd_attrs */
1196 // new stat data found, but object may have old items
1197 // (directories and symlinks)
1216 else
1221 else
1227 SD_V2_SIZE));
1228 /* read persistent inode attributes from sd and initialise
1229 generic inode flags from them */
1232 }
1249 }
1250 }
1252 // update new stat data with inode fields
1254 {
1256 __u16 flags;
1269 else
1274 }
1276 // used to copy inode's fields to old stat data
1278 {
1292 else
1295 // Sigh. i_first_direct_byte is back
1298 }
1300 /* NOTE, you must prepare the buffer head before sending it here,
1301 ** and then log it after the call
1302 */
1304 loff_t size)
1305 {
1317 // path points to old stat data
1321 }
1324 }
1328 {
1342 /* look for the object's stat data */
1346 "i/o failure occurred trying to "
1349 }
1354 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1356 }
1358 "stat data of object %k (nlink == %d) "
1361 pos);
1364 }
1366 /* sigh, prepare_for_journal might schedule. When it schedules the
1367 ** FS might change. We have to detect that, and loop back to the
1368 ** search if the stat data item has moved
1369 */
1379 }
1381 }
1386 }
1388 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1389 ** does a make_bad_inode when things go wrong. But, we need to make sure
1390 ** and clear the key in the private portion of the inode, otherwise a
1391 ** corresponding iput might try to delete whatever object the inode last
1392 ** represented.
1393 */
1395 {
1398 }
1400 //
1401 // initially this function was derived from minix or ext2's analog and
1402 // evolved as the prototype did
1403 //
1406 {
1411 }
1413 /* looks for stat data in the tree, and fills up the fields of in-core
1414 inode stat data fields */
1417 {
1425 /* set version 1, version 2 could be used too, because stat data
1426 key is the same in both versions */
1433 /* look for the object's stat data */
1437 "i/o failure occurred trying to find "
1441 }
1443 /* a stale NFS handle can trigger this without it being an error */
1448 }
1452 /* It is possible that knfsd is trying to access inode of a file
1453 that is being removed from the disk by some other thread. As we
1454 update sd on unlink all that is required is to check for nlink
1455 here. This bug was first found by Sizif when debugging
1456 SquidNG/Butterfly, forgotten, and found again after Philippe
1457 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1459 More logical fix would require changes in fs/inode.c:iput() to
1460 remove inode from hash-table _after_ fs cleaned disk stuff up and
1461 in iget() to return NULL if I_FREEING inode is found in
1462 hash-table. */
1463 /* Currently there is one place where it's ok to meet inode with
1464 nlink==0: processing of open-unlinked and half-truncated files
1465 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1469 "dead inode read from disk %K. "
1473 }
1477 /*
1478 * Stat data v1 doesn't support ACLs.
1479 */
1482 }
1484 /**
1485 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1486 *
1487 * @inode: inode from hash table to check
1488 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1489 *
1490 * This function is called by iget5_locked() to distinguish reiserfs inodes
1491 * having the same inode numbers. Such inodes can only exist due to some
1492 * error condition. One of them should be bad. Inodes with identical
1493 * inode numbers (objectids) are distinguished by parent directory ids.
1494 *
1495 */
1497 {
1501 /* args is already in CPU order */
1504 }
1507 {
1525 }
1528 /* either due to i/o error or a stale NFS handle */
1531 }
1533 }
1538 {
1550 }
1554 }
1558 {
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.
1563 * Valid types are:
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
1570 */
1577 }
1583 }
1587 {
1597 }
1601 {
1610 }
1623 }
1624 }
1626 }
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.
1634 */
1636 {
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.
1646 */
1652 }
1654 }
1656 }
1658 /* stat data of new object is inserted already, this inserts the item
1659 containing "." and ".." entries */
1664 {
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 */
1696 }
1698 /* look for place in the tree for new item */
1704 }
1711 }
1713 /* insert item, that is empty directory item */
1715 }
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 */
1723 {
1738 /* look for place in the tree for new item */
1744 }
1751 }
1753 /* insert item, that is body of symlink */
1755 }
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. */
1768 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1769 strlen (symname) for symlinks) */
1773 {
1794 }
1796 /* item head of new item */
1802 }
1807 else
1820 }
1823 /* not a perfect generation count, as object ids can be reused, but
1824 ** this is as good as reiserfs can do right now.
1825 ** note that the private part of inode isn't filled in yet, we have
1826 ** to use the directory.
1827 */
1829 else
1830 #if defined( USE_INODE_GENERATION_COUNTER )
1833 #else
1835 #endif
1837 /* fill stat data */
1840 /* uid and gid must already be set by the caller for quota init */
1842 /* symlink cannot be immutable or append only, right? */
1864 /* key to search for correct place for new stat data */
1869 /* find proper place for inserting of stat data */
1874 }
1879 }
1883 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1886 }
1890 }
1891 // store in in-core inode the key of stat data and version all
1892 // object items will have (directory items will have old offset
1893 // format, other new objects will consist of new items)
1896 else
1900 else
1903 /* insert the stat data into the tree */
1904 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1907 #endif
1908 retval =
1915 }
1916 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1919 #endif
1921 /* insert item with "." and ".." */
1922 retval =
1924 }
1927 /* insert body of symlink */
1930 retval =
1932 i_size);
1933 }
1939 }
1950 }
1953 "ACLs aren't enabled in the fs, "
1970 }
1971 }
1978 /* it looks like you can easily compress these two goto targets into
1979 * one. Keeping it like this doesn't actually hurt anything, and they
1980 * are place holders for what the quota code actually needs.
1981 */
1982 out_bad_inode:
1983 /* Invalidate the object, nothing was inserted yet */
1986 /* Quota change must be inside a transaction for journaling */
1991 out_end_trans:
1993 /* Drop can be outside and it needs more credits so it's better to have it outside */
2000 out_inserted_sd:
2006 }
2008 /*
2009 ** finds the tail page in the page cache,
2010 ** reads the last block in.
2011 **
2012 ** On success, page_result is set to a locked, pinned page, and bh_result
2013 ** is set to an up to date buffer for the last block in the file. returns 0.
2014 **
2015 ** tail conversion is not done, so bh_result might not be valid for writing
2016 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2017 ** trying to write the block.
2018 **
2019 ** on failure, nonzero is returned, page_result and bh_result are untouched.
2020 */
2024 {
2026 /* we want the page with the last byte in the file,
2027 ** not the page that will hold the next byte for appending
2028 */
2039 /* we know that we are only called with inode->i_size > 0.
2040 ** we also know that a file tail can never be as big as a block
2041 ** If i_size % blocksize == 0, our file is currently block aligned
2042 ** and it won't need converting or zeroing after a truncate.
2043 */
2046 }
2051 }
2052 /* start within the page of the last block in the file */
2056 reiserfs_get_block_create_0);
2065 }
2071 /* note, this should never happen, prepare_write should
2072 ** be taking care of this for us. If the buffer isn't up to date,
2073 ** I've screwed up the code to find the buffer, or the code to
2074 ** call prepare_write
2075 */
2080 }
2084 out:
2087 unlock:
2091 }
2093 /*
2094 ** vfs version of truncate file. Must NOT be called with
2095 ** a transaction already started.
2096 **
2097 ** some code taken from block_truncate_page
2098 */
2100 {
2102 /* we want the offset for the first byte after the end of the file */
2116 // -ENOENT means we truncated past the end of the file,
2117 // and get_block_create_0 could not find a block to read in,
2118 // which is ok.
2122 error);
2125 }
2126 }
2128 /* so, if page != NULL, we have a buffer head for the offset at
2129 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2130 ** then we have an unformatted node. Otherwise, we have a direct item,
2131 ** and no zeroing is required on disk. We zero after the truncate,
2132 ** because the truncate might pack the item anyway
2133 ** (it will unmap bh if it packs).
2134 */
2135 /* it is enough to reserve space in transaction for 2 balancings:
2136 one for "save" link adding and another for the first
2137 cut_from_item. 1 is for update_sd */
2144 /* we are doing real truncate: if the system crashes before the last
2145 transaction of truncating gets committed - on reboot the file
2146 either appears truncated properly or not truncated at all */
2149 error =
2154 /* check reiserfs_do_truncate after ending the transaction */
2158 }
2164 }
2168 /* if we are not on a block boundary */
2174 }
2175 }
2178 }
2183 out:
2187 }
2192 }
2197 {
2215 /* catch places below that try to log something without starting a trans */
2220 }
2223 start_over:
2227 research:
2232 }
2239 /* we've found an unformatted node */
2244 }
2246 /* crap, we are writing to a hole */
2249 }
2262 /* vs-3050 is gone, no need to drop the path */
2271 bh);
2273 }
2274 }
2282 }
2285 copy_size);
2291 /* are there still bytes left? */
2296 copy_size);
2298 }
2304 }
2307 out:
2314 }
2317 /* this is where we fill in holes in the file. */
2320 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2325 /* get_block failed to find a mapped unformatted node. */
2328 }
2329 }
2330 }
2334 /* we've copied data from the page into the direct item, so the
2335 * buffer in the page is now clean, mark it to reflect that.
2336 */
2340 }
2342 }
2344 /*
2345 * mason@suse.com: updated in 2.5.54 to follow the same general io
2346 * start/recovery path as __block_write_full_page, along with special
2347 * code to handle reiserfs tails.
2348 */
2351 {
2356 sector_t last_block;
2366 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2371 }
2373 /* The page dirty bit is cleared before writepage is called, which
2374 * means we have to tell create_empty_buffers to make dirty buffers
2375 * The page really should be up to date at this point, so tossing
2376 * in the BH_Uptodate is just a sanity check.
2377 */
2381 }
2384 /* last page in the file, zero out any contents past the
2385 ** last byte in the file
2386 */
2391 /* no file contents in this page */
2395 }
2397 }
2401 /* first map all the buffers, logging any direct items we find */
2404 /*
2405 * This can happen when the block size is less than
2406 * the page size. The corresponding bytes in the page
2407 * were zero filled above
2408 */
2415 /* not mapped yet, or it points to a direct item, search
2416 * the btree for the mapping info, and log any direct
2417 * items found
2418 */
2421 }
2422 }
2424 block++;
2427 /*
2428 * we start the transaction after map_block_for_writepage,
2429 * because it can create holes in the file (an unbounded operation).
2430 * starting it here, we can make a reliable estimate for how many
2431 * blocks we're going to log
2432 */
2440 }
2442 }
2443 /* now go through and lock any dirty buffers on the page */
2455 }
2456 /* from this point on, we know the buffer is mapped to a
2457 * real block and not a direct item
2458 */
2465 }
2466 }
2471 }
2479 }
2484 /*
2485 * since any buffer might be the only dirty buffer on the page,
2486 * the first submit_bh can bring the page out of writeback.
2487 * be careful with the buffers.
2488 */
2493 nr++;
2494 }
2500 done:
2502 /*
2503 * if this page only had a direct item, it is very possible for
2504 * no io to be required without there being an error. Or,
2505 * someone else could have locked them and sent them down the
2506 * pipe without locking the page
2507 */
2513 }
2519 }
2522 fail:
2523 /* catches various errors, we need to make sure any valid dirty blocks
2524 * get to the media. The page is currently locked and not marked for
2525 * writeback
2526 */
2535 /*
2536 * clear any dirty bits that might have come from getting
2537 * attached to a dirty page
2538 */
2540 }
2552 nr++;
2553 }
2558 }
2561 {
2563 }
2566 {
2570 }
2573 {
2576 }
2582 {
2585 pgoff_t index;
2593 pos ++;
2595 }
2608 journal_info;
2613 }
2617 /* this gets a little ugly. If reiserfs_get_block returned an
2618 * error and left a transacstion running, we've got to close it,
2619 * and we've got to free handle if it was a persistent transaction.
2620 *
2621 * But, if we had nested into an existing transaction, we need
2622 * to just drop the ref count on the handle.
2623 *
2624 * If old_ref == 0, the transaction is from reiserfs_get_block,
2625 * and it was a persistent trans. Otherwise, it was nested above.
2626 */
2637 }
2638 }
2639 }
2643 /* Truncate allocated blocks */
2645 }
2647 }
2650 {
2664 journal_info;
2669 }
2674 /* this gets a little ugly. If reiserfs_get_block returned an
2675 * error and left a transacstion running, we've got to close it,
2676 * and we've got to free handle if it was a persistent transaction.
2677 *
2678 * But, if we had nested into an existing transaction, we need
2679 * to just drop the ref count on the handle.
2680 *
2681 * If old_ref == 0, the transaction is from reiserfs_get_block,
2682 * and it was a persistent trans. Otherwise, it was nested above.
2683 */
2694 }
2695 }
2696 }
2699 }
2702 {
2704 }
2709 {
2718 pos ++;
2723 else
2732 }
2737 /* generic_commit_write does this for us, but does not update the
2738 ** transaction tracking stuff when the size changes. So, we have
2739 ** to do the i_size updates here.
2740 */
2745 /* If the file have grown beyond the border where it
2746 can have a tail, unmark it as needing a tail
2747 packing */
2760 /*
2761 * this will just nest into our transaction. It's important
2762 * to use mark_inode_dirty so the inode gets pushed around on the
2763 * dirty lists, and so that O_SYNC works as expected
2764 */
2771 }
2776 }
2782 }
2784 out:
2795 journal_error:
2802 }
2804 }
2808 {
2822 }
2825 /* generic_commit_write does this for us, but does not update the
2826 ** transaction tracking stuff when the size changes. So, we have
2827 ** to do the i_size updates here.
2828 */
2831 /* If the file have grown beyond the border where it
2832 can have a tail, unmark it as needing a tail
2833 packing */
2846 /*
2847 * this will just nest into our transaction. It's important
2848 * to use mark_inode_dirty so the inode gets pushed around on the
2849 * dirty lists, and so that O_SYNC works as expected
2850 */
2857 }
2864 }
2866 out:
2869 journal_error:
2874 }
2877 }
2880 {
2884 else
2888 else
2892 else
2896 else
2900 else
2902 }
2903 }
2906 {
2910 else
2914 else
2918 else
2922 else
2924 }
2925 }
2927 /* decide if this buffer needs to stay around for data logging or ordered
2928 ** write purposes
2929 */
2931 {
2939 }
2940 /* the page is locked, and the only places that log a data buffer
2941 * also lock the page.
2942 */
2944 /*
2945 * very conservative, leave the buffer pinned if
2946 * anyone might need it.
2947 */
2950 }
2955 /* why is this safe?
2956 * reiserfs_setattr updates i_size in the on disk
2957 * stat data before allowing vmtruncate to be called.
2958 *
2959 * If buffer was put onto the ordered list for this
2960 * transaction, we know for sure either this transaction
2961 * or an older one already has updated i_size on disk,
2962 * and this ordered data won't be referenced in the file
2963 * if we crash.
2964 *
2965 * if the buffer was put onto the ordered list for an older
2966 * transaction, we need to leave it around
2967 */
2971 }
2972 free_jh:
2975 }
2979 }
2981 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2984 {
3009 /*
3010 * is this block fully invalidated?
3011 */
3015 else
3017 }
3022 /*
3023 * We release buffers only if the entire page is being invalidated.
3024 * The get_block cached value has been unconditionally invalidated,
3025 * so real IO is not possible anymore.
3026 */
3029 /* maybe should BUG_ON(!ret); - neilb */
3030 }
3031 out:
3033 }
3036 {
3041 }
3043 }
3045 /*
3046 * Returns 1 if the page's buffers were dropped. The page is locked.
3047 *
3048 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3049 * in the buffers at page_buffers(page).
3050 *
3051 * even in -o notail mode, we can't be sure an old mount without -o notail
3052 * didn't create files with tails.
3053 */
3055 {
3073 }
3074 }
3081 }
3083 /* We thank Mingming Cao for helping us understand in great detail what
3084 to do in this section of the code. */
3088 {
3091 ssize_t ret;
3094 reiserfs_get_blocks_direct_io);
3096 /*
3097 * In case of error extending write may have instantiated a few
3098 * blocks outside i_size. Trim these off again.
3099 */
3107 }
3108 }
3111 }
3114 {
3123 /* must be turned off for recursive notify_change calls */
3130 /* version 2 items will be caught by the s_maxbytes check
3131 ** done for us in vmtruncate
3132 */
3138 }
3142 /* fill in hole pointers in the expanding truncate case. */
3148 /* we're changing at most 2 bitmaps, inode + super */
3153 }
3156 }
3160 }
3161 /*
3162 * file size is changed, ctime and mtime are
3163 * to be updated
3164 */
3166 }
3167 }
3173 /* stat data of format v3.5 has 16 bit uid and gid */
3176 }
3192 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3204 }
3206 /* Update corresponding info in inode so that everything is in
3207 * one transaction */
3217 }
3225 }
3226 }
3231 }
3236 }
3238 out:
3240 }
3253 };