Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 27 Jul 2003 22:03:57 +0200
From:      Bernd Walter <ticso@cicely12.cicely.de>
To:        Marcel Moolenaar <marcel@xcllnt.net>
Cc:        alpha@freebsd.org
Subject:   Re: NFS hangs appears dc(4) newbusification related
Message-ID:  <20030727200356.GL42805@cicely12.cicely.de>
In-Reply-To: <20030727194616.GA608@athlon.pn.xcllnt.net>
References:  <20030725220247.GA1653@dhcp01.pn.xcllnt.net> <20030726074052.GA93951@freebie.xs4all.nl> <20030726081120.GA7470@dhcp01.pn.xcllnt.net> <20030726163016.GF95189@freebie.xs4all.nl> <20030726183007.GA575@dhcp01.pn.xcllnt.net> <20030726184509.GA4440@freebie.xs4all.nl> <20030726192133.GB575@dhcp01.pn.xcllnt.net> <20030727081715.GB6392@freebie.xs4all.nl> <20030727091517.GA468@freebie.xs4all.nl> <20030727194616.GA608@athlon.pn.xcllnt.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Jul 27, 2003 at 12:46:16PM -0700, Marcel Moolenaar wrote:
> I think the mystery is solved:
> 
> After some triaging it became clear that the problem was not the
> newbusification of dc(4). I got the same lockups with sources
> before that change, but after the mbuf changes sibly@ made. The
> comment by ticso@ that tcp doesn't work anyway was not true
> before, but is now. That misalignment existed before is possible.

It is true since day one and therefor TCP NFS never worked reliable
on alpha.
If and how often you trigger the bug depends on your usage.
If you see anything else then an alignment panic then it's something
new.

> It never caused any problems for me though. Removing tcp from the
> mount options in fstab(5) solved the problem with -current sources,
> but could be be used prior to that.
> 
> So:
> o  Using TCP based NFS mounts is indeed broken and should at least
>    be mentioned somewhere, because the breakage is only visible on
>    alpha. Ideally it should be fixed.

The alignemt bug is a problem on other strong alignment platforms as
well.
I can't speak about the new problem.

> o  rpc.statd(8) still fails for msync(), which probably needs to
>    addressed as well, because it's specific to alpha too.
> o  If misalignment is the root cause of why TCP based NFS mounts
>    don't work, then why doesn't the kernel panic? All misalignments
>    in the kernel should be fatal, not cause "random" failures.

They do panic for shure.

-- 
B.Walter                   BWCT                http://www.bwct.de
ticso@bwct.de                                  info@bwct.de



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030727200356.GL42805>