On Sun, Jul 08, 2007 at 09:38:42PM +0200, Marcus Better wrote:
> Hi,
>
> > I'm a bit confused here. You cloned the bug and reassigned one part against
> > nfs-utils -- this would usually indicate you mean _both_ nfs-utils and
> > e2fsprogs are buggy.
>
> I should have explained my intentions better. I have problems with nfs-utils
> as described. A hint from upstream made me suspect that the discussion [2]
> has relevance for this bug.
>
> My *guess* is that [2] describes the bug in rpc.mountd that causes this
> problem.
The two bugs are not related at all. Bug #413661 is a memory leak.
It causes rpc.mountd to use more and more memory over time; and it
wasn't detected right away since most users of blkid do their thing
and then exit; they aren't a long-term daemon. that bug was fixed in
e2fsprogs version 1.39+1.40-WIP-2007.04.07+dfsg-1.
You are looking at a bug that involves lots and lots of calls to
ioctl(), and that's clearly something else.
OK, so could you please do the following and then open a new bug
against libblkid1, if you are sure the fault lies there. Save a copy
of /etc/blkid.tab first (cp /etc/blkid.tab /tmp/blkid.tab.1). Next,
set the environment variable BLKID_DEBUG to the value 0xffff ("export
BLKID_DEBUG=0xffff") and then run rpc.mountd in the foreground with
the output redirected to a file (rpc.mountd -F > /tmp/mountd.log);
make sure you kill off the rpc.mound which is running as a daemon
first. Finally save another a copy of /etc/blkid.tab (cp
/etc/blkid.tab /tmp/blkid.tab.2)
Now trigger a mount request, and check to see that you see blkid debug
messages showing up in /tmp/mountd.log. Then send me (via the new bug
report) the /tmp/mountd.log, /tmp/blkid.tab.1, and /tmp/blkid.tab.2
files.
Thanks!
- Ted
P.S. In the future, unless you're absolutely certain that a bug you
are seeing is related, please don't claim a bug is found. There is a
rather large difference between a memory leak and lots of md ioctl's.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
