Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 12 Oct 2003 04:12:27 -0700
From:      Kris Kennaway <kris@obsecurity.org>
To:        Tillman Hodgson <tillman@seekingfire.com>
Cc:        Kris Kennaway <kris@obsecurity.org>
Subject:   Re: X on sparc64?
Message-ID:  <20031012111227.GA27698@rot13.obsecurity.org>
In-Reply-To: <20031011194837.B564@seekingfire.com>
References:  <20031011114819.GA54814@freebie.xs4all.nl> <20031012003402.GA32486@rot13.obsecurity.org> <20031011193836.A564@seekingfire.com> <20031012014224.GA95189@rot13.obsecurity.org> <20031011194837.B564@seekingfire.com>

next in thread | previous in thread | raw e-mail | index | archive | help

--MGYHOYXEY6WxJCY8
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Sat, Oct 11, 2003 at 07:48:37PM -0600, Tillman Hodgson wrote:
> On Sat, Oct 11, 2003 at 06:42:24PM -0700, Kris Kennaway wrote:
> > On Sat, Oct 11, 2003 at 07:38:36PM -0600, Tillman Hodgson wrote:
> > > On Sat, Oct 11, 2003 at 05:34:02PM -0700, Kris Kennaway wrote:
> > > > Be aware that 4.x's NFS code doesn't like talking to sparc64 client=
s -
> > > > it causes the i386 server to double fault.  5.x as a server is
> > > > perfectly happy though.
> > >=20
> > > I have an Ultra 5 running a recent -current that uses an i386 -stable=
 as
> > > it's NFS server ... I've never seen any problems with it. Is this an
> > > issue that is only affected via certain NFS options or is otherwise
> > > possible to (unknowingly) avoid?
> >=20
> > It only occurs under load..if you've never put heavy (I mean heavy!)
> > loads on the server you've probably avoided it.  I was able to repeat
> > this on two separate networks.
>=20
> Load on which end? I dump all filesystems on 4 servers to the -stable
> box out of daily.local (thus simultaneously), which should swamp the NFS
> server fairly regularly. 14 out of 16 of my nfsd's show significant CPU
> usage, considering that my uptime is 11 days (my last -stable upgrade).
>=20
> I'm now somewhat nervous about the setup if you meant load on the NFS
> server rather than the sparc64 client :-)
>=20

BTW, if you want to test this, a good way to try would be to build the
fsx tool in /usr/src/tools/regression/fsx, mount the i386 4.x machine
rw on the sparc64 and run a dozen or so instances concurrently for a
day or so (or until it blows up).  I haven't tried this, but fsx is
typically a good way to exercise filesystem bugs.

Kris

--MGYHOYXEY6WxJCY8
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (FreeBSD)

iD8DBQE/iTcbWry0BWjoQKURAkFfAJ433AYT/0D3FFuGej+ujS/93tCDWACgiJ2h
ZwZZjXwGw6ihRfvyCbyONOg=
=cAI+
-----END PGP SIGNATURE-----

--MGYHOYXEY6WxJCY8--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20031012111227.GA27698>