Hello!
Attached are 2 fixes for reiserfs in 2.4.17-rc2
map_block_for_writepage_highmem_fix-1.diff fixes oops on incorerct access to (possibly) unkmapped highmem pages.
mmaped_data_loss_fix.diff fixes a problem where some data may be lost if mmap-writed data is subsequently appended by
data written with write(2), and because of that mmapped data is lost. (Originally found by fsx tool)
Bye,
Oleg
--qMm9M+Fa2AknHoGS
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename="map_block_for_writepage_highmem_fix-1.diff"
--- linux/fs/reiserfs/inode.c.orig Wed Dec 19 16:26:05 2001
+++ linux/fs/reiserfs/inode.c Wed Dec 19 16:32:02 2001
@@ -1772,6 +1772,7 @@
int bytes_copied = 0 ;
int copy_size ;
+ kmap(bh_result->b_page) ;
start_over:
lock_kernel() ;
journal_begin(&th, inode->i_sb, jbegin_count) ;
@@ -1844,10 +1845,8 @@
/* this is where we fill in holes in the file. */
if (use_get_block) {
- kmap(bh_result->b_page) ;
retval = reiserfs_get_block(inode, block, bh_result,
GET_BLOCK_CREATE | GET_BLOCK_NO_ISEM) ;
- kunmap(bh_result->b_page) ;
if (!retval) {
if (!buffer_mapped(bh_result) || bh_result->b_blocknr == 0) {
/* get_block failed to find a mapped unformatted node. */
@@ -1856,6 +1855,7 @@
}
}
}
+ kunmap(bh_result->b_page) ;
return retval ;
}
--qMm9M+Fa2AknHoGS
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename="mmaped_data_loss_fix.diff"
--- linux/fs/reiserfs/inode.c.orig Tue Dec 18 14:50:24 2001
+++ linux/fs/reiserfs/inode.c Tue Dec 18 14:52:04 2001
@@ -273,7 +273,9 @@
pathrelse (&path);
if (p)
kunmap(bh_result->b_page) ;
- if ((args & GET_BLOCK_NO_HOLE)) {
+ // We do not return -ENOENT if there is a hole but page is uptodate, because it means
+ // That there is some MMAPED data associated with it that is yet to be written to disk.
+ if ((args & GET_BLOCK_NO_HOLE) && !Page_Uptodate(bh_result->b_page) ) {
return -ENOENT ;
}
return 0 ;
@@ -294,9 +296,13 @@
bh_result->b_dev = inode->i_dev;
bh_result->b_blocknr = blocknr;
bh_result->b_state |= (1UL << BH_Mapped);
- } else if ((args & GET_BLOCK_NO_HOLE)) {
- ret = -ENOENT ;
- }
+ } else
+ // We do not return -ENOENT if there is a hole but page is uptodate, because it means
+ // That there is some MMAPED data associated with it that is yet to be written to disk.
+ if ((args & GET_BLOCK_NO_HOLE) && !Page_Uptodate(bh_result->b_page) ) {
+ ret = -ENOENT ;
+ }
+
pathrelse (&path);
if (p)
kunmap(bh_result->b_page) ;
--qMm9M+Fa2AknHoGS--
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/