Hello!
Certain disk corruptions and i/o errors may cause lookup() to panic, which is wrong.
This patch fixes the problem.
Please apply.
Bye,
Oleg
On Thursday, January 03, 2002 10:18:30 AM +0300 Oleg Drokin
<[email protected]> wrote:
> Hello!
>
> Certain disk corruptions and i/o errors may cause lookup() to panic,
> which is wrong. This patch fixes the problem.
> Please apply.
Hmmm, none of the callers of reiserfs_find_entry have been changed to check
for IO_ERROR. We should at least change reiserfs_add_entry to check for
IO_ERROR, so it doesn't try to create a name after getting io error during
the lookup.
-chris
--- linux/fs/reiserfs/namei.c.orig Tue Dec 25 16:27:27 2001
+++ linux/fs/reiserfs/namei.c Tue Dec 25 16:29:13 2001
@@ -309,9 +309,10 @@
while (1) {
retval = search_by_entry_key (dir->i_sb, &key_to_search, path_to_entry, de);
- if (retval == IO_ERROR)
- // FIXME: still has to be dealt with
- reiserfs_panic (dir->i_sb, "zam-7001: io error in " __FUNCTION__ "\n");
+ if (retval == IO_ERROR) {
+ reiserfs_warning ("zam-7001: io error in " __FUNCTION__ "\n");
+ return IO_ERROR;
+ }
/* compare names for all entries having given hash value */
retval = linear_search_in_dir_item (&key_to_search, de, name, namelen);
Hello!
On Fri, Jan 11, 2002 at 12:46:26PM -0500, Chris Mason wrote:
> > Certain disk corruptions and i/o errors may cause lookup() to panic,
> > which is wrong. This patch fixes the problem.
> > Please apply.
> Hmmm, none of the callers of reiserfs_find_entry have been changed to check
> for IO_ERROR. We should at least change reiserfs_add_entry to check for
> IO_ERROR, so it doesn't try to create a name after getting io error during
> the lookup.
Well, in fact reiserfs_add_entry won't do that anyway, consider this code:
retval = reiserfs_find_entry (dir, name, namelen, &path, &de);
if( retval != NAME_NOT_FOUND ) {
if (buffer != small_buf)
reiserfs_kfree (buffer, buflen, dir->i_sb);
pathrelse (&path);
if (retval != NAME_FOUND) {
reiserfs_warning ("zam-7002:" __FUNCTION__ ": \"reiserfs_find_entry\" has returned"
" unexpected value (%d)\n", retval);
}
return -EEXIST;
}
Though I see other places where code is not that smart. I'll come up with additional patch later today.
Thanks for noticing.
Bye,
Oleg