> Have not checked it under 2.2.11 but it seems it does not hang on 2.2.2 and
> 2.2.10.
> I'm not THAT sure because this server is newly installed and it has /var
> separated.
> I have a lot of servers running 2.2.10 (previously 2.2.2) and they have
> quotas and chown works
> Ok. And updatedb also works ok.
So I might found out what was the problem. It was the deadlock on
write_dquot() and it is there for ages. I thought I fixed it but it is still
there only harder to trigger. Try following patch. The problem should go
away.
Honza.
--8t9RHnE3ZwKMSgU+
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename="dquot.c.diff"
--- linux/fs/dquot.c Mon Oct 25 13:54:16 1999
+++ linux/fs/dquot.c Mon Oct 25 14:00:10 1999
@@ -1341,7 +1341,6 @@
dquot_incr_blocks(transfer_to[cnt], blocks);
unlock_dquot(transfer_from[cnt]);
- dqput(transfer_from[cnt]);
if (inode->i_dquot[cnt] != NODQUOT) {
struct dquot *temp = inode->i_dquot[cnt];
inode->i_dquot[cnt] = transfer_to[cnt];
@@ -1351,20 +1350,20 @@
unlock_dquot(transfer_to[cnt]);
dqput(transfer_to[cnt]);
}
+ dqput(transfer_from[cnt]);
}
return 0;
put_all:
for (disc = 0; disc < cnt; disc++) {
/* There should be none or both pointers set but... */
- if (transfer_to[disc] != NODQUOT) {
+ if (transfer_to[disc] != NODQUOT)
unlock_dquot(transfer_to[disc]);
- dqput(transfer_to[disc]);
- }
- if (transfer_from[disc] != NODQUOT) {
+ if (transfer_from[disc] != NODQUOT)
unlock_dquot(transfer_from[disc]);
- dqput(transfer_from[disc]);
- }
+ /* dqput() tests for NODQUOT itself... */
+ dqput(transfer_from[disc]);
+ dqput(transfer_to[disc]);
}
return error;
}
--8t9RHnE3ZwKMSgU+--
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/