--/6/NiKJuzyDL1e1D
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Wed, Mar 08, 2006 at 02:05:02AM +0000, segv@netctl.net wrote:
> On Wed, 8 Mar 2006 03:06:26 +0100
> Quentin Garnier <cube@cubidou.net> wrote:
>=20
> > On Tue, Mar 07, 2006 at 07:16:44PM -0600, Peter Seebach wrote:
> > > Hi!  This is NOT a NetBSD bug, but I'm really curious as to whether a=
nyone's
> > > seen it.
> > >=20
> > > This program reliably hangs three different AMD64 machines, all of th=
em
> > > Athlons, and all with nForce4 chipsets.  I would love to know more ca=
ses
> > > where it works or doesn't.  Hangs have been observed under 64-bit Lin=
ux
> > > and 32-bit Windows.
> > >=20
> > >         #include <string.h>
> > >         unsigned char a[4*1024*1024];
> > >=20
> > >         int
> > >         main(void) {
> > >                 while (1)
> > >                         memset(a, 0, 4*1024*1024);
> > >                 return 0;
> > >         }
> > >=20
> > > This program 100% reliably hangs my Athlon64 box.  It takes under 10 =
seconds
> > > under Linux, several minutes on Windows.  And no, that doesn't make a=
ny
> > > sense. I don't have any EM64T boxes handy for testing.
> >=20
> > The memory is probably failing.
> >=20
>=20
> On three different machines at the same time?

It happened to me last week with two wholefully different machines.
Each of them saw their buffer cache corrupted by one bit over a QA
session at work, affecting two different but very close files (which of
course made all the remaining tests of the QA fail).  I had a hard time
believing failing memory could be the reason in both case, yet memtest
proved me wrong.

But you should note that one of the machines ran memtest86 for a whole
week-end and only got four one-bit errors that all happened within a
single second.

So yes, now I'm more inclined to consider memory the problem these days
for that kind of hair-pulling issue.  And no, one pass of memtest86 is
not enough.

--=20
Quentin Garnier - cube@cubidou.net - cube@NetBSD.org
"When I find the controls, I'll go where I like, I'll know where I want
to be, but maybe for now I'll stay right here on a silent sea."
KT Tunstall, Silent Sea, Eye to the Telescope, 2004.

--/6/NiKJuzyDL1e1D
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (NetBSD)

iQEVAwUBRA6PTtgoQloHrPnoAQKZqwf/VErOKda3cdp1RL462CdLCdqVHtnfxbbe
D3oYWt7gn1gMfjPFlLrslkUvLGOTpt9ZA7pmgyG2QvKGmPlgc0aIg4sbjZXb6XI3
nlDrFylHeq9xF6AkdjZeeRPACdJ7hGa4E4vc9NHZKihfXJ+vrDWjoJE5lDLLa0gI
AWEwwjMVYPfE9SURV6s7LlDK0i4+vnZMCq/HAcF5PJWrpi1GTnP6QflqFsrzJpbY
CT7HdX0/4qBSUGat456io58ZQ+aFQxnpkBi40CJuwn3LIw1lsc+arlMJGQSh3xqi
fDMfOxjxvo1RuDJNAOIgMkMmPLp28Pmcx0qUuEqkSf9bsduENofbjQ==
=1AzU
-----END PGP SIGNATURE-----

--/6/NiKJuzyDL1e1D--