Subject: Re: uvm_fault
To: Pavel Cahyna <email@example.com>
From: Chuck Silvers <firstname.lastname@example.org>
Date: 01/30/2006 08:48:05
On Mon, Jan 30, 2006 at 09:25:53AM +0100, Pavel Cahyna wrote:
> On Sun, Jan 29, 2006 at 08:38:52PM -0800, Chuck Silvers wrote:
> > On Sun, Jan 29, 2006 at 12:35:51AM +0100, Pavel Cahyna wrote:
> > > On Sat, Jan 28, 2006 at 12:53:27PM +0000, email@example.com wrote:
> > > > This is on dual Pentium 3 machine
> > > > Kernel was compiled with '-O3 -march=pentium3 -mmmx -msse -mfpmath=sse'
> > >
> > > Can you please recompile a kernel without those options (but still with
> > > makeoptions DEBUG="-g") and try again?
> > >
> > > If you repeat the problem, please also type "show registers" in ddb before
> > > you do "bt $ebp".
> > "show registers" in ddb after a panic shows you the registers of panic(),
> > not the registers at the time the event that triggered the panic
> > (in this case, a fatal page-fault trap) occured.
> eip was 0, how can it be a register in panic()?
> This also seems consistent with the message "Stopped in pid 543.1 (mountd) at 0: invalid address"
> Don't you think that the code called a NULL function pointer?
hmm, it looks like you're right. the current code appears to not do
the annoying thing in that case now. maybe it used to be worse,
or maybe I'm just misremembering. anyway, nevermind.