| blob | 0bae036831e2ca2aeeae94eb78d74f2aa769b54d |
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/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>
26 {
27 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
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 */
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 */
65 /* check return value from reiserfs_delete_object after
66 * ending the transaction
67 */
71 /* all items of file are deleted, so we can remove "save" link */
73 * about an error here */
75 /* no object items are in the tree */
76 ;
77 }
78 out:
85 no_delete:
88 }
92 {
100 }
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 */
106 {
110 length);
111 }
113 //
114 // when key is 0, do not set version and short key
115 //
120 {
125 }
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
134 }
136 //
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.
151 **
152 ** The size of the cache could be dynamic based on the size of the file.
153 **
154 ** I'd also like to see us cache the location the stat data item, since
155 ** we are needlessly researching for that frequently.
156 **
157 ** --chris
158 */
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
167 */
169 {
179 }
182 }
183 }
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 */
190 {
197 }
200 {
202 }
206 {
208 }
210 //
211 // files which were created in the earlier version can not be longer,
212 // than 2 gb
213 //
215 {
221 }
225 {
235 /* we cannot restart while nested */
238 }
245 }
247 }
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.
259 {
264 b_blocknr_t blocknr;
272 // prepare the key to look for the 'block'-th block of file
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.
289 }
291 }
292 //
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
300 access to file */
308 }
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.
315 }
321 }
322 // requested data are in direct item(s)
324 // we are called by bmap. FIXME: we can not map block of file
325 // when it is stored in direct item(s)
330 }
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
335 */
339 /*
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
344 */
348 }
349 // read file tail into part of page
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
356 ** kmap schedules
357 */
366 }
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
371 */
375 chars =
381 }
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
400 // i/o error most likely
409 finished:
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
417 */
421 }
423 // this is called to create file map. So, _get_block_create_0 will not
424 // read direct item
427 {
432 /* do not read the direct item */
436 }
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
443 ** is also returned.
444 **
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.
449 **
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.
454 */
458 {
460 }
462 /* This is special helper for reiserfs_get_block in case we are executing
463 direct_IO request. */
465 sector_t iblock,
468 {
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() */
483 /* don't allow direct io onto tail pages */
485 /* make sure future calls to the direct io funcs for this offset
486 ** in the file fail by unmapping the buffer
487 */
490 }
491 /* Possible unpacked tail. Flush the data before pages have
492 disappeared */
505 }
506 out:
508 }
510 /*
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
515 **
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
518 ** call this.
519 */
522 loff_t tail_offset)
523 {
534 /* always try to read until the end of the block */
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
541 tail page */
547 }
550 }
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.
555 **
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.
560 */
567 /* tail conversion might change the data in the page */
572 unlock:
576 }
577 out:
579 }
582 sector_t block,
586 {
589 #ifdef REISERFS_PREALLOCATE
593 }
594 #endif
596 block);
597 }
601 {
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 */
625 loff_t new_offset =
634 }
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
638 */
641 /* find number of block-th logical block of the file */
646 }
647 /*
648 * if we're already in a transaction, make sure to close
649 * any new transactions we start in this func
650 */
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
657 */
664 /* set the key of the first byte in the 'block'-th block of file */
667 start_trans:
672 }
674 }
675 research:
681 }
693 /* we have to allocate block for the unformatted node */
697 }
699 repeat =
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
707 */
712 repeat =
718 }
721 else
724 }
729 }
730 }
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) */
739 /* use allocated block to plug the hole */
744 bh);
746 }
755 }
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
766 ** transaction
767 */
769 }
774 }
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 */
785 /* indirect item has to be inserted */
791 /* we are going to add 'block'-th block to the file. Use
792 allocated block for that */
798 }
803 retval =
810 }
811 //mark_tail_converted (inode);
813 /* direct item has to be converted */
814 loff_t tail_offset;
816 tail_offset =
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 */
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 */
833 /*
834 * ugly, but we can only end the transaction if
835 * we aren't nested
836 */
839 retval =
840 reiserfs_end_persistent_transaction
845 }
847 retval =
849 tail_offset);
854 "convert tail failed "
857 retval);
859 /* the bitmap, the super, and the stat data == 3 */
864 inode,
865 allocated_block_nr,
867 }
869 }
871 }
872 retval =
874 tail_offset);
880 }
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
887 ** the disk
888 */
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
893 */
895 /* we've converted the tail, so we must
896 ** flush unbh before the transaction commits
897 */
900 /* mark it dirty now to prevent commit_write from adding
901 ** this buffer to the inode's dirty buffer list
902 */
903 /*
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())
908 */
910 }
912 /* append indirect item with holes if needed, when appending
913 pointer to 'block'-th block use block, which is already
914 allocated */
917 // one block which is a fastpath
919 __u64 max_to_insert =
921 UNFM_P_SIZE;
922 __u64 blocks_needed;
926 /* indirect item has to be appended, set up key of that position */
932 //pos_in_item * inode->i_sb->s_blocksize,
937 blocks_needed =
941 s_blocksize_bits);
951 }
952 }
954 /* we are going to add target block to the file. Use allocated
955 block for that */
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 */
966 blocks_needed =
968 }
969 retval =
972 UNFM_P_SIZE *
973 blocks_needed);
982 }
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
986 holes. */
989 }
990 }
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
999 **
1000 ** release the path so that anybody waiting on the path before
1001 ** ending their transaction will be able to continue.
1002 */
1007 }
1008 /*
1009 * inserting indirect pointers for a hole can take a
1010 * long time. reschedule if needed and also release the write
1011 * lock for others.
1012 */
1017 }
1023 }
1033 }
1042 failure:
1050 }
1055 }
1057 static int
1060 {
1062 }
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
1066 */
1068 {
1075 /* End of file is also in full block with indirect reference, so round
1076 ** up to the next block.
1077 **
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
1081 ** node pointer
1082 */
1083 bytes =
1086 sd_size;
1088 }
1092 {
1096 }
1099 }
1101 /* Compute number of blocks used by file in ReiserFS counting */
1103 {
1107 /* keeps fsck and non-quota versions of reiserfs happy */
1110 }
1112 /* files from before the quota patch might i_blocks such that
1113 ** bytes < real_space. Deal with that here to prevent it from
1114 ** going negative.
1115 */
1119 }
1121 //
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
1126 //
1128 // called by read_locked_inode
1130 {
1133 __u32 rdev;
1134 //int version = ITEM_VERSION_1;
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
1179 }
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.
1186 */
1189 }
1192 SD_V1_SIZE));
1193 /* nopack is initially zero for v1 objects. For v2 objects,
1194 nopack is initialised from sd_attrs */
1197 // new stat data found, but object may have old items
1198 // (directories and symlinks)
1217 else
1222 else
1228 SD_V2_SIZE));
1229 /* read persistent inode attributes from sd and initialise
1230 generic inode flags from them */
1233 }
1250 }
1251 }
1253 // update new stat data with inode fields
1255 {
1257 __u16 flags;
1270 else
1275 }
1277 // used to copy inode's fields to old stat data
1279 {
1293 else
1296 // Sigh. i_first_direct_byte is back
1299 }
1301 /* NOTE, you must prepare the buffer head before sending it here,
1302 ** and then log it after the call
1303 */
1305 loff_t size)
1306 {
1318 // path points to old stat data
1322 }
1325 }
1329 {
1343 /* look for the object's stat data */
1347 "i/o failure occurred trying to "
1350 }
1355 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1357 }
1359 "stat data of object %k (nlink == %d) "
1362 pos);
1365 }
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
1370 */
1380 }
1382 }
1387 }
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
1393 ** represented.
1394 */
1396 {
1399 }
1401 //
1402 // initially this function was derived from minix or ext2's analog and
1403 // evolved as the prototype did
1404 //
1407 {
1412 }
1414 /* looks for stat data in the tree, and fills up the fields of in-core
1415 inode stat data fields */
1418 {
1426 /* set version 1, version 2 could be used too, because stat data
1427 key is the same in both versions */
1434 /* look for the object's stat data */
1438 "i/o failure occurred trying to find "
1442 }
1444 /* a stale NFS handle can trigger this without it being an error */
1449 }
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
1463 hash-table. */
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()). */
1470 "dead inode read from disk %K. "
1474 }
1478 }
1480 /**
1481 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1482 *
1483 * @inode: inode from hash table to check
1484 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1485 *
1486 * This function is called by iget5_locked() to distinguish reiserfs inodes
1487 * having the same inode numbers. Such inodes can only exist due to some
1488 * error condition. One of them should be bad. Inodes with identical
1489 * inode numbers (objectids) are distinguished by parent directory ids.
1490 *
1491 */
1493 {
1497 /* args is already in CPU order */
1500 }
1503 {
1520 }
1523 /* either due to i/o error or a stale NFS handle */
1526 }
1528 }
1533 {
1545 }
1549 }
1553 {
1554 /* fhtype happens to reflect the number of u32s encoded.
1555 * due to a bug in earlier code, fhtype might indicate there
1556 * are more u32s then actually fitted.
1557 * so if fhtype seems to be more than len, reduce fhtype.
1558 * Valid types are:
1559 * 2 - objectid + dir_id - legacy support
1560 * 3 - objectid + dir_id + generation
1561 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1562 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1563 * 6 - as above plus generation of directory
1564 * 6 does not fit in NFSv2 handles
1565 */
1572 }
1576 }
1580 {
1588 }
1592 {
1603 /* no room for directory info? return what we've stored so far */
1615 }
1618 }
1620 /* looks for stat data, then copies fields to it, marks the buffer
1621 containing stat data as dirty */
1622 /* reiserfs inodes are never really dirty, since the dirty inode call
1623 ** always logs them. This call allows the VFS inode marking routines
1624 ** to properly mark inodes for datasync and such, but only actually
1625 ** does something when called for a synchronous update.
1626 */
1628 {
1634 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1635 ** these cases are just when the system needs ram, not when the
1636 ** inode needs to reach disk for safety, and they can safely be
1637 ** ignored because the altered inode has already been logged.
1638 */
1644 }
1646 }
1648 }
1650 /* stat data of new object is inserted already, this inserts the item
1651 containing "." and ".." entries */
1656 {
1669 /* compose item head for new item. Directories consist of items of
1670 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1671 is done by reiserfs_new_inode */
1688 }
1690 /* look for place in the tree for new item */
1696 }
1703 }
1705 /* insert item, that is empty directory item */
1707 }
1709 /* stat data of object has been inserted, this inserts the item
1710 containing the body of symlink */
1711 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1715 {
1730 /* look for place in the tree for new item */
1736 }
1743 }
1745 /* insert item, that is body of symlink */
1747 }
1749 /* inserts the stat data into the tree, and then calls
1750 reiserfs_new_directory (to insert ".", ".." item if new object is
1751 directory) or reiserfs_new_symlink (to insert symlink body if new
1752 object is symlink) or nothing (if new object is regular file)
1754 NOTE! uid and gid must already be set in the inode. If we return
1755 non-zero due to an error, we have to drop the quota previously allocated
1756 for the fresh inode. This can only be done outside a transaction, so
1757 if we return non-zero, we also end the transaction. */
1760 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1761 strlen (symname) for symlinks) */
1765 {
1784 }
1788 /* item head of new item */
1794 }
1799 else
1808 }
1810 /* not a perfect generation count, as object ids can be reused, but
1811 ** this is as good as reiserfs can do right now.
1812 ** note that the private part of inode isn't filled in yet, we have
1813 ** to use the directory.
1814 */
1816 else
1817 #if defined( USE_INODE_GENERATION_COUNTER )
1820 #else
1822 #endif
1824 /* fill stat data */
1827 /* uid and gid must already be set by the caller for quota init */
1829 /* symlink cannot be immutable or append only, right? */
1851 /* key to search for correct place for new stat data */
1856 /* find proper place for inserting of stat data */
1861 }
1866 }
1870 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1873 }
1877 }
1878 // store in in-core inode the key of stat data and version all
1879 // object items will have (directory items will have old offset
1880 // format, other new objects will consist of new items)
1883 else
1887 else
1890 /* insert the stat data into the tree */
1891 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1894 #endif
1895 retval =
1902 }
1903 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1906 #endif
1908 /* insert item with "." and ".." */
1909 retval =
1911 }
1914 /* insert body of symlink */
1917 retval =
1919 i_size);
1920 }
1926 }
1935 }
1938 "ACLs aren't enabled in the fs, "
1953 }
1954 }
1961 /* it looks like you can easily compress these two goto targets into
1962 * one. Keeping it like this doesn't actually hurt anything, and they
1963 * are place holders for what the quota code actually needs.
1964 */
1965 out_bad_inode:
1966 /* Invalidate the object, nothing was inserted yet */
1969 /* Quota change must be inside a transaction for journaling */
1972 out_end_trans:
1974 /* Drop can be outside and it needs more credits so it's better to have it outside */
1979 out_inserted_sd:
1985 }
1987 /*
1988 ** finds the tail page in the page cache,
1989 ** reads the last block in.
1990 **
1991 ** On success, page_result is set to a locked, pinned page, and bh_result
1992 ** is set to an up to date buffer for the last block in the file. returns 0.
1993 **
1994 ** tail conversion is not done, so bh_result might not be valid for writing
1995 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1996 ** trying to write the block.
1997 **
1998 ** on failure, nonzero is returned, page_result and bh_result are untouched.
1999 */
2003 {
2005 /* we want the page with the last byte in the file,
2006 ** not the page that will hold the next byte for appending
2007 */
2018 /* we know that we are only called with inode->i_size > 0.
2019 ** we also know that a file tail can never be as big as a block
2020 ** If i_size % blocksize == 0, our file is currently block aligned
2021 ** and it won't need converting or zeroing after a truncate.
2022 */
2025 }
2030 }
2031 /* start within the page of the last block in the file */
2035 reiserfs_get_block_create_0);
2044 }
2050 /* note, this should never happen, prepare_write should
2051 ** be taking care of this for us. If the buffer isn't up to date,
2052 ** I've screwed up the code to find the buffer, or the code to
2053 ** call prepare_write
2054 */
2059 }
2063 out:
2066 unlock:
2070 }
2072 /*
2073 ** vfs version of truncate file. Must NOT be called with
2074 ** a transaction already started.
2075 **
2076 ** some code taken from block_truncate_page
2077 */
2079 {
2081 /* we want the offset for the first byte after the end of the file */
2096 // -ENOENT means we truncated past the end of the file,
2097 // and get_block_create_0 could not find a block to read in,
2098 // which is ok.
2102 error);
2105 }
2106 }
2108 /* so, if page != NULL, we have a buffer head for the offset at
2109 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2110 ** then we have an unformatted node. Otherwise, we have a direct item,
2111 ** and no zeroing is required on disk. We zero after the truncate,
2112 ** because the truncate might pack the item anyway
2113 ** (it will unmap bh if it packs).
2114 */
2115 /* it is enough to reserve space in transaction for 2 balancings:
2116 one for "save" link adding and another for the first
2117 cut_from_item. 1 is for update_sd */
2124 /* we are doing real truncate: if the system crashes before the last
2125 transaction of truncating gets committed - on reboot the file
2126 either appears truncated properly or not truncated at all */
2129 error =
2134 /* check reiserfs_do_truncate after ending the transaction */
2138 }
2144 }
2148 /* if we are not on a block boundary */
2154 }
2155 }
2158 }
2163 out:
2167 }
2172 }
2177 {
2195 /* catch places below that try to log something without starting a trans */
2200 }
2203 start_over:
2207 research:
2212 }
2219 /* we've found an unformatted node */
2224 }
2226 /* crap, we are writing to a hole */
2229 }
2242 /* vs-3050 is gone, no need to drop the path */
2251 bh);
2253 }
2254 }
2262 }
2265 copy_size);
2271 /* are there still bytes left? */
2276 copy_size);
2278 }
2284 }
2287 out:
2294 }
2297 /* this is where we fill in holes in the file. */
2300 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2305 /* get_block failed to find a mapped unformatted node. */
2308 }
2309 }
2310 }
2314 /* we've copied data from the page into the direct item, so the
2315 * buffer in the page is now clean, mark it to reflect that.
2316 */
2320 }
2322 }
2324 /*
2325 * mason@suse.com: updated in 2.5.54 to follow the same general io
2326 * start/recovery path as __block_write_full_page, along with special
2327 * code to handle reiserfs tails.
2328 */
2331 {
2336 sector_t last_block;
2346 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2351 }
2353 /* The page dirty bit is cleared before writepage is called, which
2354 * means we have to tell create_empty_buffers to make dirty buffers
2355 * The page really should be up to date at this point, so tossing
2356 * in the BH_Uptodate is just a sanity check.
2357 */
2361 }
2364 /* last page in the file, zero out any contents past the
2365 ** last byte in the file
2366 */
2371 /* no file contents in this page */
2375 }
2377 }
2381 /* first map all the buffers, logging any direct items we find */
2384 /*
2385 * This can happen when the block size is less than
2386 * the page size. The corresponding bytes in the page
2387 * were zero filled above
2388 */
2395 /* not mapped yet, or it points to a direct item, search
2396 * the btree for the mapping info, and log any direct
2397 * items found
2398 */
2401 }
2402 }
2404 block++;
2407 /*
2408 * we start the transaction after map_block_for_writepage,
2409 * because it can create holes in the file (an unbounded operation).
2410 * starting it here, we can make a reliable estimate for how many
2411 * blocks we're going to log
2412 */
2420 }
2422 }
2423 /* now go through and lock any dirty buffers on the page */
2435 }
2436 /* from this point on, we know the buffer is mapped to a
2437 * real block and not a direct item
2438 */
2445 }
2446 }
2451 }
2459 }
2464 /*
2465 * since any buffer might be the only dirty buffer on the page,
2466 * the first submit_bh can bring the page out of writeback.
2467 * be careful with the buffers.
2468 */
2473 nr++;
2474 }
2480 done:
2482 /*
2483 * if this page only had a direct item, it is very possible for
2484 * no io to be required without there being an error. Or,
2485 * someone else could have locked them and sent them down the
2486 * pipe without locking the page
2487 */
2493 }
2499 }
2502 fail:
2503 /* catches various errors, we need to make sure any valid dirty blocks
2504 * get to the media. The page is currently locked and not marked for
2505 * writeback
2506 */
2515 /*
2516 * clear any dirty bits that might have come from getting
2517 * attached to a dirty page
2518 */
2520 }
2532 nr++;
2533 }
2538 }
2541 {
2543 }
2546 {
2550 }
2553 {
2556 }
2562 {
2565 pgoff_t index;
2573 pos ++;
2575 }
2588 journal_info;
2593 }
2597 /* this gets a little ugly. If reiserfs_get_block returned an
2598 * error and left a transacstion running, we've got to close it,
2599 * and we've got to free handle if it was a persistent transaction.
2600 *
2601 * But, if we had nested into an existing transaction, we need
2602 * to just drop the ref count on the handle.
2603 *
2604 * If old_ref == 0, the transaction is from reiserfs_get_block,
2605 * and it was a persistent trans. Otherwise, it was nested above.
2606 */
2617 }
2618 }
2619 }
2623 /* Truncate allocated blocks */
2625 }
2627 }
2630 {
2643 journal_info;
2648 }
2653 /* this gets a little ugly. If reiserfs_get_block returned an
2654 * error and left a transacstion running, we've got to close it,
2655 * and we've got to free handle if it was a persistent transaction.
2656 *
2657 * But, if we had nested into an existing transaction, we need
2658 * to just drop the ref count on the handle.
2659 *
2660 * If old_ref == 0, the transaction is from reiserfs_get_block,
2661 * and it was a persistent trans. Otherwise, it was nested above.
2662 */
2673 }
2674 }
2675 }
2678 }
2681 {
2683 }
2688 {
2698 pos ++;
2703 else
2712 }
2717 /* generic_commit_write does this for us, but does not update the
2718 ** transaction tracking stuff when the size changes. So, we have
2719 ** to do the i_size updates here.
2720 */
2725 /* If the file have grown beyond the border where it
2726 can have a tail, unmark it as needing a tail
2727 packing */
2740 /*
2741 * this will just nest into our transaction. It's important
2742 * to use mark_inode_dirty so the inode gets pushed around on the
2743 * dirty lists, and so that O_SYNC works as expected
2744 */
2751 }
2756 }
2762 }
2764 out:
2775 journal_error:
2782 }
2784 }
2788 {
2801 }
2804 /* generic_commit_write does this for us, but does not update the
2805 ** transaction tracking stuff when the size changes. So, we have
2806 ** to do the i_size updates here.
2807 */
2810 /* If the file have grown beyond the border where it
2811 can have a tail, unmark it as needing a tail
2812 packing */
2825 /*
2826 * this will just nest into our transaction. It's important
2827 * to use mark_inode_dirty so the inode gets pushed around on the
2828 * dirty lists, and so that O_SYNC works as expected
2829 */
2836 }
2843 }
2845 out:
2848 journal_error:
2853 }
2856 }
2859 {
2863 else
2867 else
2871 else
2875 else
2879 else
2881 }
2882 }
2885 {
2889 else
2893 else
2897 else
2901 else
2903 }
2904 }
2906 /* decide if this buffer needs to stay around for data logging or ordered
2907 ** write purposes
2908 */
2910 {
2918 }
2919 /* the page is locked, and the only places that log a data buffer
2920 * also lock the page.
2921 */
2923 /*
2924 * very conservative, leave the buffer pinned if
2925 * anyone might need it.
2926 */
2929 }
2934 /* why is this safe?
2935 * reiserfs_setattr updates i_size in the on disk
2936 * stat data before allowing vmtruncate to be called.
2937 *
2938 * If buffer was put onto the ordered list for this
2939 * transaction, we know for sure either this transaction
2940 * or an older one already has updated i_size on disk,
2941 * and this ordered data won't be referenced in the file
2942 * if we crash.
2943 *
2944 * if the buffer was put onto the ordered list for an older
2945 * transaction, we need to leave it around
2946 */
2950 }
2951 free_jh:
2954 }
2958 }
2960 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2962 {
2982 /*
2983 * is this block fully invalidated?
2984 */
2988 else
2990 }
2995 /*
2996 * We release buffers only if the entire page is being invalidated.
2997 * The get_block cached value has been unconditionally invalidated,
2998 * so real IO is not possible anymore.
2999 */
3002 /* maybe should BUG_ON(!ret); - neilb */
3003 }
3004 out:
3006 }
3009 {
3014 }
3016 }
3018 /*
3019 * Returns 1 if the page's buffers were dropped. The page is locked.
3020 *
3021 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3022 * in the buffers at page_buffers(page).
3023 *
3024 * even in -o notail mode, we can't be sure an old mount without -o notail
3025 * didn't create files with tails.
3026 */
3028 {
3046 }
3047 }
3054 }
3056 /* We thank Mingming Cao for helping us understand in great detail what
3057 to do in this section of the code. */
3061 {
3064 ssize_t ret;
3070 /*
3071 * In case of error extending write may have instantiated a few
3072 * blocks outside i_size. Trim these off again.
3073 */
3080 }
3083 }
3086 {
3096 /* must be turned off for recursive notify_change calls */
3104 /* version 2 items will be caught by the s_maxbytes check
3105 ** done for us in vmtruncate
3106 */
3111 }
3112 /* fill in hole pointers in the expanding truncate case. */
3118 /* we're changing at most 2 bitmaps, inode + super */
3123 }
3126 }
3129 /*
3130 * file size is changed, ctime and mtime are
3131 * to be updated
3132 */
3134 }
3135 }
3140 /* stat data of format v3.5 has 16 bit uid and gid */
3143 }
3159 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3167 }
3169 /* Update corresponding info in inode so that everything is in
3170 * one transaction */
3179 }
3181 /*
3182 * Relax the lock here, as it might truncate the
3183 * inode pages and wait for inode pages locks.
3184 * To release such page lock, the owner needs the
3185 * reiserfs lock
3186 */
3195 }
3201 }
3203 out:
3207 }
3221 };