NFS hangs appears dc(4) newbusification related

[Bernd Walter]

On Sun, Jul 27, 2003 at 01:08:16PM -0700, Marcel Moolenaar wrote:
> On Sun, Jul 27, 2003 at 10:03:57PM +0200, Bernd Walter wrote:
> > 
> > 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 must be something new then. Which only means that a broken state
> has further regressed...
> 
> > > 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 don't see it here on ia64 nor sparc64.

It happens rarely under very high load.
There is a realignment code that works fine for a single missalignment
in the input mbuf cluster, but fails if there is a second one.
That also means that we need a condition with multiple transactions
in the input queue to trigger this bug.
Maybe it only triggers in server code, but the problem exists in the
client code as well.

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