| blob | 3a28e7751b3c714da6e3b2c0b7fe9ad89e82a87b |
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 <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>
28 {
29 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
40 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
41 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
52 /* Do quota update inside a transaction for journaled quotas. We must do that
53 * after delete_object so that quota updates go into the same transaction as
54 * stat data deletion */
61 /* check return value from reiserfs_delete_object after
62 * ending the transaction
63 */
67 /* all items of file are deleted, so we can remove "save" link */
69 * about an error here */
71 /* no object items are in the tree */
72 ;
73 }
74 out:
78 }
82 {
90 }
92 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
93 offset and type of key */
96 {
100 length);
101 }
103 //
104 // when key is 0, do not set version and short key
105 //
110 {
115 }
120 /* set_ih_free_space (ih, 0); */
121 // for directory items it is entry count, for directs and stat
122 // datas - 0xffff, for indirects - 0
124 }
126 //
127 // FIXME: we might cache recently accessed indirect item
129 // Ugh. Not too eager for that....
130 // I cut the code until such time as I see a convincing argument (benchmark).
131 // I don't want a bloated inode struct..., and I don't like code complexity....
133 /* cutting the code is fine, since it really isn't in use yet and is easy
134 ** to add back in. But, Vladimir has a really good idea here. Think
135 ** about what happens for reading a file. For each page,
136 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
137 ** an indirect item. This indirect item has X number of pointers, where
138 ** X is a big number if we've done the block allocation right. But,
139 ** we only use one or two of these pointers during each call to readpage,
140 ** needlessly researching again later on.
141 **
142 ** The size of the cache could be dynamic based on the size of the file.
143 **
144 ** I'd also like to see us cache the location the stat data item, since
145 ** we are needlessly researching for that frequently.
146 **
147 ** --chris
148 */
150 /* If this page has a file tail in it, and
151 ** it was read in by get_block_create_0, the page data is valid,
152 ** but tail is still sitting in a direct item, and we can't write to
153 ** it. So, look through this page, and check all the mapped buffers
154 ** to make sure they have valid block numbers. Any that don't need
155 ** to be unmapped, so that block_prepare_write will correctly call
156 ** reiserfs_get_block to convert the tail into an unformatted node
157 */
159 {
169 }
172 }
173 }
175 /* reiserfs_get_block does not need to allocate a block only if it has been
176 done already or non-hole position has been found in the indirect item */
180 {
187 }
190 {
192 }
196 {
198 }
200 //
201 // files which were created in the earlier version can not be longer,
202 // than 2 gb
203 //
205 {
211 }
215 {
225 /* we cannot restart while nested */
228 }
235 }
237 }
239 // it is called by get_block when create == 0. Returns block number
240 // for 'block'-th logical block of file. When it hits direct item it
241 // returns 0 (being called from bmap) or read direct item into piece
242 // of page (bh_result)
244 // Please improve the english/clarity in the comment above, as it is
245 // hard to understand.
249 {
254 b_blocknr_t blocknr;
262 // prepare the key to look for the 'block'-th block of file
274 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
275 // That there is some MMAPED data associated with it that is yet to be written to disk.
279 }
281 }
282 //
288 /* FIXME: here we could cache indirect item or part of it in
289 the inode to avoid search_by_key in case of subsequent
290 access to file */
298 }
300 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
301 // That there is some MMAPED data associated with it that is yet to be written to disk.
305 }
311 }
312 // requested data are in direct item(s)
314 // we are called by bmap. FIXME: we can not map block of file
315 // when it is stored in direct item(s)
320 }
322 /* if we've got a direct item, and the buffer or page was uptodate,
323 ** we don't want to pull data off disk again. skip to the
324 ** end, where we map the buffer and return
325 */
329 /*
330 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
331 ** pages without any buffers. If the page is up to date, we don't want
332 ** read old data off disk. Set the up to date bit on the buffer instead
333 ** and jump to the end
334 */
338 }
339 // read file tail into part of page
343 /* we only want to kmap if we are reading the tail into the page.
344 ** this is not the common case, so we don't kmap until we are
345 ** sure we need to. But, this means the item might move if
346 ** kmap schedules
347 */
356 }
357 /* make sure we don't read more bytes than actually exist in
358 ** the file. This can happen in odd cases where i_size isn't
359 ** correct, and when direct item padding results in a few
360 ** extra bytes at the end of the direct item
361 */
365 chars =
371 }
380 // we done, if read direct item is not the last item of
381 // node FIXME: we could try to check right delimiting key
382 // to see whether direct item continues in the right
383 // neighbor or rely on i_size
386 // update key to look for the next piece
390 // i/o error most likely
399 finished:
405 /* this buffer has valid data, but isn't valid for io. mapping it to
406 * block #0 tells the rest of reiserfs it just has a tail in it
407 */
411 }
413 // this is called to create file map. So, _get_block_create_0 will not
414 // read direct item
417 {
422 /* do not read the direct item */
426 }
428 /* special version of get_block that is only used by grab_tail_page right
429 ** now. It is sent to block_prepare_write, and when you try to get a
430 ** block past the end of the file (or a block from a hole) it returns
431 ** -ENOENT instead of a valid buffer. block_prepare_write expects to
432 ** be able to do i/o on the buffers returned, unless an error value
433 ** is also returned.
434 **
435 ** So, this allows block_prepare_write to be used for reading a single block
436 ** in a page. Where it does not produce a valid page for holes, or past the
437 ** end of the file. This turns out to be exactly what we need for reading
438 ** tails for conversion.
439 **
440 ** The point of the wrapper is forcing a certain value for create, even
441 ** though the VFS layer is calling this function with create==1. If you
442 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
443 ** don't use this function.
444 */
448 {
450 }
452 /* This is special helper for reiserfs_get_block in case we are executing
453 direct_IO request. */
455 sector_t iblock,
458 {
463 /* We set the b_size before reiserfs_get_block call since it is
464 referenced in convert_tail_for_hole() that may be called from
465 reiserfs_get_block() */
473 /* don't allow direct io onto tail pages */
475 /* make sure future calls to the direct io funcs for this offset
476 ** in the file fail by unmapping the buffer
477 */
480 }
481 /* Possible unpacked tail. Flush the data before pages have
482 disappeared */
495 }
496 out:
498 }
500 /*
501 ** helper function for when reiserfs_get_block is called for a hole
502 ** but the file tail is still in a direct item
503 ** bh_result is the buffer head for the hole
504 ** tail_offset is the offset of the start of the tail in the file
505 **
506 ** This calls prepare_write, which will start a new transaction
507 ** you should not be in a transaction, or have any paths held when you
508 ** call this.
509 */
512 loff_t tail_offset)
513 {
524 /* always try to read until the end of the block */
529 /* hole_page can be zero in case of direct_io, we are sure
530 that we cannot get here if we write with O_DIRECT into
531 tail page */
537 }
540 }
542 /* we don't have to make sure the conversion did not happen while
543 ** we were locking the page because anyone that could convert
544 ** must first take i_mutex.
545 **
546 ** We must fix the tail page for writing because it might have buffers
547 ** that are mapped, but have a block number of 0. This indicates tail
548 ** data that has been read directly into the page, and block_prepare_write
549 ** won't trigger a get_block in this case.
550 */
556 /* tail conversion might change the data in the page */
561 unlock:
565 }
566 out:
568 }
571 sector_t block,
575 {
578 #ifdef REISERFS_PREALLOCATE
582 }
583 #endif
585 block);
586 }
590 {
603 /* space reserved in transaction batch:
604 . 3 balancings in direct->indirect conversion
605 . 1 block involved into reiserfs_update_sd()
606 XXX in practically impossible worst case direct2indirect()
607 can incur (much) more than 3 balancings.
608 quota update for user, group */
614 loff_t new_offset =
623 }
625 /* if !create, we aren't changing the FS, so we don't need to
626 ** log anything, so we don't need to start a transaction
627 */
630 /* find number of block-th logical block of the file */
635 }
636 /*
637 * if we're already in a transaction, make sure to close
638 * any new transactions we start in this func
639 */
644 /* If file is of such a size, that it might have a tail and tails are enabled
645 ** we should mark it as possibly needing tail packing on close
646 */
653 /* set the key of the first byte in the 'block'-th block of file */
656 start_trans:
661 }
663 }
664 research:
670 }
682 /* we have to allocate block for the unformatted node */
686 }
688 repeat =
693 /* restart the transaction to give the journal a chance to free
694 ** some blocks. releases the path, so we have to go back to
695 ** research if we succeed on the second try
696 */
701 repeat =
707 }
710 else
713 }
718 }
719 }
722 b_blocknr_t unfm_ptr;
723 /* 'block'-th block is in the file already (there is
724 corresponding cell in some indirect item). But it may be
725 zero unformatted node pointer (hole) */
728 /* use allocated block to plug the hole */
733 bh);
735 }
744 }
753 /* the item was found, so new blocks were not added to the file
754 ** there is no need to make sure the inode is updated with this
755 ** transaction
756 */
758 }
763 }
765 /* desired position is not found or is in the direct item. We have
766 to append file with holes up to 'block'-th block converting
767 direct items to indirect one if necessary */
774 /* indirect item has to be inserted */
780 /* we are going to add 'block'-th block to the file. Use
781 allocated block for that */
787 }
792 retval =
799 }
800 //mark_tail_converted (inode);
802 /* direct item has to be converted */
803 loff_t tail_offset;
805 tail_offset =
809 /* direct item we just found fits into block we have
810 to map. Convert it into unformatted node: use
811 bh_result for the conversion */
817 /* we have to padd file tail stored in direct item(s)
818 up to block size and convert it to unformatted
819 node. FIXME: this should also get into page cache */
822 /*
823 * ugly, but we can only end the transaction if
824 * we aren't nested
825 */
828 retval =
829 reiserfs_end_persistent_transaction
834 }
836 retval =
838 tail_offset);
843 "convert tail failed "
846 retval);
848 /* the bitmap, the super, and the stat data == 3 */
853 inode,
854 allocated_block_nr,
856 }
858 }
860 }
861 retval =
863 tail_offset);
869 }
870 /* it is important the set_buffer_uptodate is done after
871 ** the direct2indirect. The buffer might contain valid
872 ** data newer than the data on disk (read by readpage, changed,
873 ** and then sent here by writepage). direct2indirect needs
874 ** to know if unbh was already up to date, so it can decide
875 ** if the data in unbh needs to be replaced with data from
876 ** the disk
877 */
880 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
881 buffer will disappear shortly, so it should not be added to
882 */
884 /* we've converted the tail, so we must
885 ** flush unbh before the transaction commits
886 */
889 /* mark it dirty now to prevent commit_write from adding
890 ** this buffer to the inode's dirty buffer list
891 */
892 /*
893 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
894 * It's still atomic, but it sets the page dirty too,
895 * which makes it eligible for writeback at any time by the
896 * VM (which was also the case with __mark_buffer_dirty())
897 */
899 }
901 /* append indirect item with holes if needed, when appending
902 pointer to 'block'-th block use block, which is already
903 allocated */
906 // one block which is a fastpath
908 __u64 max_to_insert =
910 UNFM_P_SIZE;
911 __u64 blocks_needed;
915 /* indirect item has to be appended, set up key of that position */
921 //pos_in_item * inode->i_sb->s_blocksize,
926 blocks_needed =
930 s_blocksize_bits);
940 }
941 }
943 /* we are going to add target block to the file. Use allocated
944 block for that */
952 /* paste hole to the indirect item */
953 /* If kmalloc failed, max_to_insert becomes zero and it means we
954 only have space for one block */
955 blocks_needed =
957 }
958 retval =
961 UNFM_P_SIZE *
962 blocks_needed);
971 }
973 /* We need to mark new file size in case this function will be
974 interrupted/aborted later on. And we may do this only for
975 holes. */
978 }
979 }
984 /* this loop could log more blocks than we had originally asked
985 ** for. So, we have to allow the transaction to end if it is
986 ** too big or too full. Update the inode so things are
987 ** consistent if we crash before the function returns
988 **
989 ** release the path so that anybody waiting on the path before
990 ** ending their transaction will be able to continue.
991 */
996 }
997 /*
998 * inserting indirect pointers for a hole can take a
999 * long time. reschedule if needed and also release the write
1000 * lock for others.
1001 */
1006 }
1012 }
1022 }
1031 failure:
1039 }
1044 }
1046 static int
1049 {
1051 }
1053 /* Compute real number of used bytes by file
1054 * Following three functions can go away when we'll have enough space in stat item
1055 */
1057 {
1064 /* End of file is also in full block with indirect reference, so round
1065 ** up to the next block.
1066 **
1067 ** there is just no way to know if the tail is actually packed
1068 ** on the file, so we have to assume it isn't. When we pack the
1069 ** tail, we add 4 bytes to pretend there really is an unformatted
1070 ** node pointer
1071 */
1072 bytes =
1075 sd_size;
1077 }
1081 {
1085 }
1088 }
1090 /* Compute number of blocks used by file in ReiserFS counting */
1092 {
1096 /* keeps fsck and non-quota versions of reiserfs happy */
1099 }
1101 /* files from before the quota patch might i_blocks such that
1102 ** bytes < real_space. Deal with that here to prevent it from
1103 ** going negative.
1104 */
1108 }
1110 //
1111 // BAD: new directories have stat data of new type and all other items
1112 // of old type. Version stored in the inode says about body items, so
1113 // in update_stat_data we can not rely on inode, but have to check
1114 // item version directly
1115 //
1117 // called by read_locked_inode
1119 {
1122 __u32 rdev;
1123 //int version = ITEM_VERSION_1;
1163 // there was a bug in <=3.5.23 when i_blocks could take negative
1164 // values. Starting from 3.5.17 this value could even be stored in
1165 // stat data. For such files we set i_blocks based on file
1166 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1167 // only updated if file's inode will ever change
1169 }
1174 /* an early bug in the quota code can give us an odd number for the
1175 ** block count. This is incorrect, fix it here.
1176 */
1179 }
1182 SD_V1_SIZE));
1183 /* nopack is initially zero for v1 objects. For v2 objects,
1184 nopack is initialised from sd_attrs */
1187 // new stat data found, but object may have old items
1188 // (directories and symlinks)
1207 else
1212 else
1218 SD_V2_SIZE));
1219 /* read persistent inode attributes from sd and initalise
1220 generic inode flags from them */
1223 }
1240 }
1241 }
1243 // update new stat data with inode fields
1245 {
1247 __u16 flags;
1260 else
1265 }
1267 // used to copy inode's fields to old stat data
1269 {
1283 else
1286 // Sigh. i_first_direct_byte is back
1289 }
1291 /* NOTE, you must prepare the buffer head before sending it here,
1292 ** and then log it after the call
1293 */
1295 loff_t size)
1296 {
1308 // path points to old stat data
1312 }
1315 }
1319 {
1333 /* look for the object's stat data */
1337 "i/o failure occurred trying to "
1340 }
1345 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1347 }
1349 "stat data of object %k (nlink == %d) "
1352 pos);
1355 }
1357 /* sigh, prepare_for_journal might schedule. When it schedules the
1358 ** FS might change. We have to detect that, and loop back to the
1359 ** search if the stat data item has moved
1360 */
1370 }
1372 }
1377 }
1379 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1380 ** does a make_bad_inode when things go wrong. But, we need to make sure
1381 ** and clear the key in the private portion of the inode, otherwise a
1382 ** corresponding iput might try to delete whatever object the inode last
1383 ** represented.
1384 */
1386 {
1389 }
1391 //
1392 // initially this function was derived from minix or ext2's analog and
1393 // evolved as the prototype did
1394 //
1397 {
1402 }
1404 /* looks for stat data in the tree, and fills up the fields of in-core
1405 inode stat data fields */
1408 {
1416 /* set version 1, version 2 could be used too, because stat data
1417 key is the same in both versions */
1424 /* look for the object's stat data */
1428 "i/o failure occurred trying to find "
1432 }
1434 /* a stale NFS handle can trigger this without it being an error */
1439 }
1443 /* It is possible that knfsd is trying to access inode of a file
1444 that is being removed from the disk by some other thread. As we
1445 update sd on unlink all that is required is to check for nlink
1446 here. This bug was first found by Sizif when debugging
1447 SquidNG/Butterfly, forgotten, and found again after Philippe
1448 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1450 More logical fix would require changes in fs/inode.c:iput() to
1451 remove inode from hash-table _after_ fs cleaned disk stuff up and
1452 in iget() to return NULL if I_FREEING inode is found in
1453 hash-table. */
1454 /* Currently there is one place where it's ok to meet inode with
1455 nlink==0: processing of open-unlinked and half-truncated files
1456 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1460 "dead inode read from disk %K. "
1464 }
1468 }
1470 /**
1471 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1472 *
1473 * @inode: inode from hash table to check
1474 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1475 *
1476 * This function is called by iget5_locked() to distinguish reiserfs inodes
1477 * having the same inode numbers. Such inodes can only exist due to some
1478 * error condition. One of them should be bad. Inodes with identical
1479 * inode numbers (objectids) are distinguished by parent directory ids.
1480 *
1481 */
1483 {
1487 /* args is already in CPU order */
1490 }
1493 {
1508 }
1511 /* either due to i/o error or a stale NFS handle */
1514 }
1516 }
1521 {
1533 }
1537 }
1541 {
1542 /* fhtype happens to reflect the number of u32s encoded.
1543 * due to a bug in earlier code, fhtype might indicate there
1544 * are more u32s then actually fitted.
1545 * so if fhtype seems to be more than len, reduce fhtype.
1546 * Valid types are:
1547 * 2 - objectid + dir_id - legacy support
1548 * 3 - objectid + dir_id + generation
1549 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1550 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1551 * 6 - as above plus generation of directory
1552 * 6 does not fit in NFSv2 handles
1553 */
1560 }
1564 }
1568 {
1576 }
1580 {
1591 /* no room for directory info? return what we've stored so far */
1603 }
1606 }
1608 /* looks for stat data, then copies fields to it, marks the buffer
1609 containing stat data as dirty */
1610 /* reiserfs inodes are never really dirty, since the dirty inode call
1611 ** always logs them. This call allows the VFS inode marking routines
1612 ** to properly mark inodes for datasync and such, but only actually
1613 ** does something when called for a synchronous update.
1614 */
1616 {
1622 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1623 ** these cases are just when the system needs ram, not when the
1624 ** inode needs to reach disk for safety, and they can safely be
1625 ** ignored because the altered inode has already been logged.
1626 */
1632 }
1634 }
1636 }
1638 /* stat data of new object is inserted already, this inserts the item
1639 containing "." and ".." entries */
1644 {
1657 /* compose item head for new item. Directories consist of items of
1658 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1659 is done by reiserfs_new_inode */
1676 }
1678 /* look for place in the tree for new item */
1684 }
1691 }
1693 /* insert item, that is empty directory item */
1695 }
1697 /* stat data of object has been inserted, this inserts the item
1698 containing the body of symlink */
1699 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1703 {
1718 /* look for place in the tree for new item */
1724 }
1731 }
1733 /* insert item, that is body of symlink */
1735 }
1737 /* inserts the stat data into the tree, and then calls
1738 reiserfs_new_directory (to insert ".", ".." item if new object is
1739 directory) or reiserfs_new_symlink (to insert symlink body if new
1740 object is symlink) or nothing (if new object is regular file)
1742 NOTE! uid and gid must already be set in the inode. If we return
1743 non-zero due to an error, we have to drop the quota previously allocated
1744 for the fresh inode. This can only be done outside a transaction, so
1745 if we return non-zero, we also end the transaction. */
1748 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1749 strlen (symname) for symlinks) */
1753 {
1768 }
1772 }
1776 /* item head of new item */
1782 }
1787 else
1796 }
1798 /* not a perfect generation count, as object ids can be reused, but
1799 ** this is as good as reiserfs can do right now.
1800 ** note that the private part of inode isn't filled in yet, we have
1801 ** to use the directory.
1802 */
1804 else
1805 #if defined( USE_INODE_GENERATION_COUNTER )
1808 #else
1810 #endif
1812 /* fill stat data */
1815 /* uid and gid must already be set by the caller for quota init */
1817 /* symlink cannot be immutable or append only, right? */
1840 /* key to search for correct place for new stat data */
1845 /* find proper place for inserting of stat data */
1850 }
1855 }
1859 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1862 }
1866 }
1867 // store in in-core inode the key of stat data and version all
1868 // object items will have (directory items will have old offset
1869 // format, other new objects will consist of new items)
1872 else
1876 else
1879 /* insert the stat data into the tree */
1880 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1883 #endif
1884 retval =
1891 }
1892 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1895 #endif
1897 /* insert item with "." and ".." */
1898 retval =
1900 }
1903 /* insert body of symlink */
1906 retval =
1908 i_size);
1909 }
1915 }
1924 }
1927 "ACLs aren't enabled in the fs, "
1942 }
1943 }
1950 /* it looks like you can easily compress these two goto targets into
1951 * one. Keeping it like this doesn't actually hurt anything, and they
1952 * are place holders for what the quota code actually needs.
1953 */
1954 out_bad_inode:
1955 /* Invalidate the object, nothing was inserted yet */
1958 /* Quota change must be inside a transaction for journaling */
1961 out_end_trans:
1963 /* Drop can be outside and it needs more credits so it's better to have it outside */
1968 out_inserted_sd:
1974 }
1976 /*
1977 ** finds the tail page in the page cache,
1978 ** reads the last block in.
1979 **
1980 ** On success, page_result is set to a locked, pinned page, and bh_result
1981 ** is set to an up to date buffer for the last block in the file. returns 0.
1982 **
1983 ** tail conversion is not done, so bh_result might not be valid for writing
1984 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1985 ** trying to write the block.
1986 **
1987 ** on failure, nonzero is returned, page_result and bh_result are untouched.
1988 */
1992 {
1994 /* we want the page with the last byte in the file,
1995 ** not the page that will hold the next byte for appending
1996 */
2007 /* we know that we are only called with inode->i_size > 0.
2008 ** we also know that a file tail can never be as big as a block
2009 ** If i_size % blocksize == 0, our file is currently block aligned
2010 ** and it won't need converting or zeroing after a truncate.
2011 */
2014 }
2019 }
2020 /* start within the page of the last block in the file */
2024 reiserfs_get_block_create_0);
2033 }
2039 /* note, this should never happen, prepare_write should
2040 ** be taking care of this for us. If the buffer isn't up to date,
2041 ** I've screwed up the code to find the buffer, or the code to
2042 ** call prepare_write
2043 */
2048 }
2052 out:
2055 unlock:
2059 }
2061 /*
2062 ** vfs version of truncate file. Must NOT be called with
2063 ** a transaction already started.
2064 **
2065 ** some code taken from block_truncate_page
2066 */
2068 {
2070 /* we want the offset for the first byte after the end of the file */
2085 // -ENOENT means we truncated past the end of the file,
2086 // and get_block_create_0 could not find a block to read in,
2087 // which is ok.
2091 error);
2094 }
2095 }
2097 /* so, if page != NULL, we have a buffer head for the offset at
2098 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2099 ** then we have an unformatted node. Otherwise, we have a direct item,
2100 ** and no zeroing is required on disk. We zero after the truncate,
2101 ** because the truncate might pack the item anyway
2102 ** (it will unmap bh if it packs).
2103 */
2104 /* it is enough to reserve space in transaction for 2 balancings:
2105 one for "save" link adding and another for the first
2106 cut_from_item. 1 is for update_sd */
2113 /* we are doing real truncate: if the system crashes before the last
2114 transaction of truncating gets committed - on reboot the file
2115 either appears truncated properly or not truncated at all */
2118 error =
2123 /* check reiserfs_do_truncate after ending the transaction */
2127 }
2133 }
2137 /* if we are not on a block boundary */
2143 }
2144 }
2147 }
2152 out:
2156 }
2161 }
2166 {
2184 /* catch places below that try to log something without starting a trans */
2189 }
2192 start_over:
2196 research:
2201 }
2208 /* we've found an unformatted node */
2213 }
2215 /* crap, we are writing to a hole */
2218 }
2231 /* vs-3050 is gone, no need to drop the path */
2240 bh);
2242 }
2243 }
2251 }
2254 copy_size);
2260 /* are there still bytes left? */
2265 copy_size);
2267 }
2273 }
2276 out:
2283 }
2286 /* this is where we fill in holes in the file. */
2289 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2294 /* get_block failed to find a mapped unformatted node. */
2297 }
2298 }
2299 }
2303 /* we've copied data from the page into the direct item, so the
2304 * buffer in the page is now clean, mark it to reflect that.
2305 */
2309 }
2311 }
2313 /*
2314 * mason@suse.com: updated in 2.5.54 to follow the same general io
2315 * start/recovery path as __block_write_full_page, along with special
2316 * code to handle reiserfs tails.
2317 */
2320 {
2325 sector_t last_block;
2335 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2340 }
2342 /* The page dirty bit is cleared before writepage is called, which
2343 * means we have to tell create_empty_buffers to make dirty buffers
2344 * The page really should be up to date at this point, so tossing
2345 * in the BH_Uptodate is just a sanity check.
2346 */
2350 }
2353 /* last page in the file, zero out any contents past the
2354 ** last byte in the file
2355 */
2360 /* no file contents in this page */
2364 }
2366 }
2370 /* first map all the buffers, logging any direct items we find */
2373 /*
2374 * This can happen when the block size is less than
2375 * the page size. The corresponding bytes in the page
2376 * were zero filled above
2377 */
2384 /* not mapped yet, or it points to a direct item, search
2385 * the btree for the mapping info, and log any direct
2386 * items found
2387 */
2390 }
2391 }
2393 block++;
2396 /*
2397 * we start the transaction after map_block_for_writepage,
2398 * because it can create holes in the file (an unbounded operation).
2399 * starting it here, we can make a reliable estimate for how many
2400 * blocks we're going to log
2401 */
2409 }
2411 }
2412 /* now go through and lock any dirty buffers on the page */
2424 }
2425 /* from this point on, we know the buffer is mapped to a
2426 * real block and not a direct item
2427 */
2434 }
2435 }
2440 }
2448 }
2453 /*
2454 * since any buffer might be the only dirty buffer on the page,
2455 * the first submit_bh can bring the page out of writeback.
2456 * be careful with the buffers.
2457 */
2462 nr++;
2463 }
2469 done:
2471 /*
2472 * if this page only had a direct item, it is very possible for
2473 * no io to be required without there being an error. Or,
2474 * someone else could have locked them and sent them down the
2475 * pipe without locking the page
2476 */
2482 }
2488 }
2491 fail:
2492 /* catches various errors, we need to make sure any valid dirty blocks
2493 * get to the media. The page is currently locked and not marked for
2494 * writeback
2495 */
2504 /*
2505 * clear any dirty bits that might have come from getting
2506 * attached to a dirty page
2507 */
2509 }
2521 nr++;
2522 }
2527 }
2530 {
2532 }
2535 {
2539 }
2545 {
2548 pgoff_t index;
2556 pos ++;
2558 }
2571 journal_info;
2576 }
2578 reiserfs_get_block);
2581 /* this gets a little ugly. If reiserfs_get_block returned an
2582 * error and left a transacstion running, we've got to close it,
2583 * and we've got to free handle if it was a persistent transaction.
2584 *
2585 * But, if we had nested into an existing transaction, we need
2586 * to just drop the ref count on the handle.
2587 *
2588 * If old_ref == 0, the transaction is from reiserfs_get_block,
2589 * and it was a persistent trans. Otherwise, it was nested above.
2590 */
2601 }
2602 }
2603 }
2607 }
2609 }
2613 {
2626 journal_info;
2631 }
2636 /* this gets a little ugly. If reiserfs_get_block returned an
2637 * error and left a transacstion running, we've got to close it,
2638 * and we've got to free handle if it was a persistent transaction.
2639 *
2640 * But, if we had nested into an existing transaction, we need
2641 * to just drop the ref count on the handle.
2642 *
2643 * If old_ref == 0, the transaction is from reiserfs_get_block,
2644 * and it was a persistent trans. Otherwise, it was nested above.
2645 */
2656 }
2657 }
2658 }
2661 }
2664 {
2666 }
2671 {
2681 pos ++;
2686 else
2695 }
2700 /* generic_commit_write does this for us, but does not update the
2701 ** transaction tracking stuff when the size changes. So, we have
2702 ** to do the i_size updates here.
2703 */
2710 /* If the file have grown beyond the border where it
2711 can have a tail, unmark it as needing a tail
2712 packing */
2725 /*
2726 * this will just nest into our transaction. It's important
2727 * to use mark_inode_dirty so the inode gets pushed around on the
2728 * dirty lists, and so that O_SYNC works as expected
2729 */
2736 }
2741 }
2747 }
2749 out:
2756 journal_error:
2763 }
2765 }
2769 {
2782 }
2785 /* generic_commit_write does this for us, but does not update the
2786 ** transaction tracking stuff when the size changes. So, we have
2787 ** to do the i_size updates here.
2788 */
2791 /* If the file have grown beyond the border where it
2792 can have a tail, unmark it as needing a tail
2793 packing */
2806 /*
2807 * this will just nest into our transaction. It's important
2808 * to use mark_inode_dirty so the inode gets pushed around on the
2809 * dirty lists, and so that O_SYNC works as expected
2810 */
2817 }
2824 }
2826 out:
2829 journal_error:
2834 }
2837 }
2840 {
2844 else
2848 else
2852 else
2856 else
2860 else
2862 }
2863 }
2866 {
2870 else
2874 else
2878 else
2882 else
2884 }
2885 }
2887 /* decide if this buffer needs to stay around for data logging or ordered
2888 ** write purposes
2889 */
2891 {
2899 }
2900 /* the page is locked, and the only places that log a data buffer
2901 * also lock the page.
2902 */
2904 /*
2905 * very conservative, leave the buffer pinned if
2906 * anyone might need it.
2907 */
2910 }
2915 /* why is this safe?
2916 * reiserfs_setattr updates i_size in the on disk
2917 * stat data before allowing vmtruncate to be called.
2918 *
2919 * If buffer was put onto the ordered list for this
2920 * transaction, we know for sure either this transaction
2921 * or an older one already has updated i_size on disk,
2922 * and this ordered data won't be referenced in the file
2923 * if we crash.
2924 *
2925 * if the buffer was put onto the ordered list for an older
2926 * transaction, we need to leave it around
2927 */
2931 }
2932 free_jh:
2935 }
2939 }
2941 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2943 {
2963 /*
2964 * is this block fully invalidated?
2965 */
2969 else
2971 }
2976 /*
2977 * We release buffers only if the entire page is being invalidated.
2978 * The get_block cached value has been unconditionally invalidated,
2979 * so real IO is not possible anymore.
2980 */
2983 /* maybe should BUG_ON(!ret); - neilb */
2984 }
2985 out:
2987 }
2990 {
2995 }
2997 }
2999 /*
3000 * Returns 1 if the page's buffers were dropped. The page is locked.
3001 *
3002 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3003 * in the buffers at page_buffers(page).
3004 *
3005 * even in -o notail mode, we can't be sure an old mount without -o notail
3006 * didn't create files with tails.
3007 */
3009 {
3027 }
3028 }
3035 }
3037 /* We thank Mingming Cao for helping us understand in great detail what
3038 to do in this section of the code. */
3042 {
3049 }
3052 {
3057 /* must be turned off for recursive notify_change calls */
3062 /* version 2 items will be caught by the s_maxbytes check
3063 ** done for us in vmtruncate
3064 */
3069 }
3070 /* fill in hole pointers in the expanding truncate case. */
3076 /* we're changing at most 2 bitmaps, inode + super */
3081 }
3084 }
3087 /*
3088 * file size is changed, ctime and mtime are
3089 * to be updated
3090 */
3092 }
3093 }
3098 /* stat data of format v3.5 has 16 bit uid and gid */
3101 }
3117 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3118 error =
3120 jbegin_count);
3123 error =
3127 jbegin_count);
3129 }
3130 /* Update corresponding info in inode so that everything is in
3131 * one transaction */
3137 error =
3139 }
3140 }
3143 }
3148 }
3150 out:
3153 }
3167 };