Ok, so all the reiserfs tail bugs weren't quite fixed yet, the last
tail fix can cause problems with highmem turned on. Both bugs are
in fs/reiserfs/inode.c:_get_block_create_0
When reading the tail in, if the buffer was already up to date,
we skip the disk i/o and return. But the cleanup code assumes the
page was kmap'd, which isn't right.
Also, there was a chance to double kmap the page if kmap scheduled a
nd the tree balanced while we slept. This bug has been there for
a long time.
Anyway, this was tested with Andrea's HIGHMEM_DEBUG_MERE_MORTALS
patch to force highmem on my 128MB machine. It works for me, but
more testers are always good.
-chris
against 2.4.4-pre6, should work against 2.4.3 or higher.
diff -Nru a/fs/reiserfs/inode.c b/fs/reiserfs/inode.c
--- a/fs/reiserfs/inode.c Wed Apr 25 23:15:14 2001
+++ b/fs/reiserfs/inode.c Wed Apr 25 23:15:14 2001
@@ -374,9 +374,11 @@
** sure we need to. But, this means the item might move if
** kmap schedules
*/
- p = (char *)kmap(bh_result->b_page) ;
- if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
- goto research;
+ if (!p) {
+ p = (char *)kmap(bh_result->b_page) ;
+ if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
+ goto research;
+ }
}
p += offset ;
memset (p, 0, inode->i_sb->s_blocksize);
@@ -420,14 +422,15 @@
ih = get_ih (&path);
} while (1);
+ flush_dcache_page(bh_result->b_page) ;
+ kunmap(bh_result->b_page) ;
+
finished:
pathrelse (&path);
bh_result->b_blocknr = 0 ;
bh_result->b_dev = inode->i_dev;
mark_buffer_uptodate (bh_result, 1);
bh_result->b_state |= (1UL << BH_Mapped);
- flush_dcache_page(bh_result->b_page) ;
- kunmap(bh_result->b_page) ;
return 0;
}