Re: ddb(4) and LOCKDEBUG incompatibility

[Andrew Doran <ad%netbsd.org@localhost>]

On Sat, May 31, 2008 at 10:20:47AM -0700, Matt Thomas wrote:

> On May 30, 2008, at 3:42 AM, Andrew Doran wrote:
> 
> >On Thu, May 29, 2008 at 11:23:42PM -0500, David Young wrote:
> >
> >>I hit the magic keys on a NetBSD router to break into ddb, and I got
> >>the LOCKDEBUG panic and page fault loop (see below).  I'm  
> >>wondering if
> >>there is a prayer of using ddb and LOCKDEBUG together, since it  
> >>can be
> >>awfully useful.
> >>
> >>It seems that we can sometimes punt on synchronization when ddb is  
> >>active,
> >>since that implies that the other CPUs on the system are suspended.
> >>Thoughts?
> >
> >There's no way we can resolve page faults from interrupt context,  
> >so I have
> >changed x86 to not call uvm_fault() if an interrupt is active. I  
> >think that
> >should fix it.
> 
> That than changing all the ports (powerpc has done that check since  
> '03),
> now that we have cpu_intr_p(), make uvm_fault check it.

Good idea, I'll look into doing that in the next few days.

Andrew