cvs commit: src/sys/i386/i386 machdep.c

[John Baldwin]

On Tuesday 03 February 2004 10:50 am, Bjoern A. Zeeb wrote:
> On Tue, 3 Feb 2004, John Baldwin wrote:
> > On Tuesday 03 February 2004 01:18 am, Bjoern A. Zeeb wrote:
> > > On Mon, 2 Feb 2004, John Baldwin wrote:
> > >
> > > Hi,
> > >
> > > > jhb         2004/02/02 15:22:24 PST
> > > >
> > > >   FreeBSD src repository
> > > >
> > > >   Modified files:
> > > >     sys/i386/i386        machdep.c
> > > >   Log:
> > > >   Set PCPU_GET(curpcb) for the BSP to thread0's pcb.  Otherwise, the
> > > > boot CPU doesn't have a pcb until after it's first context switch. 
> > > > This can cause secondary panics if a page fault happens during
> > > > bootup.
> > > >
> > > >   Revision  Changes    Path
> > > >   1.586     +1 -0      src/sys/i386/i386/machdep.c
> > >
> > > this gave me a fatal trap 9 gen. protection fault while in km
> > > last time I tried with your patch.
> >
> > You have a double panic and this fixed the first panic.  You are still
> > getting another panic that I'm not sure how to fix.
>
> *hmm*
>
> while this is true w/o ACPI enabled with ACPI enabled I do not get
> a panic while booting a kernel prior to this commit (which is
> equal to the patch you had sent me).
>
> With your patch (thus also with this commit I guess) I do get the
> panic with and without ACPI.
>
> Conclusion: with this patch/commit I will no longer be able to boot
> this problematic machine here.

This patch cannot possibly cause a panic.  It now takes a variable that was 
set to NULL, and gives it an initial value.  Before this change, places would 
panic with a page fault when they tried to access this variable before it was 
initialized.

-- 
John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org