Subject: Re: CVS commit: syssrc/sys/arch/arm32/arm32
To: IWAMOTO Toshihiro <email@example.com>
From: Chris Gilbert <firstname.lastname@example.org>
Date: 06/27/2001 08:29:48
On Wednesday 27 June 2001 7:58 am, IWAMOTO Toshihiro wrote:
> At Wed, 27 Jun 2001 01:09:22 +0300 (EEST),
> Chris Gilbert wrote:
> > Modified Files:
> > syssrc/sys/arch/arm32/arm32: fault.c
> > Log Message:
> > Add checks (make non DIAGNOSTIC) to make sure that we never call
> > uvm_fault from an interrupt.
> > To generate a diff of this commit:
> > cvs rdiff -r1.55 -r1.56 syssrc/sys/arch/arm32/arm32/fault.c
> KASSERT expands to void when DIAGNOSTIC isn't defined. So I think the
> check in data_abort_handler is better to be kept inside DIAGNOSITC
> ifdef for consistency.
> It might be even better to use KASSERT for the both of checks.
We've some extra ddb goo in the data_abort_handler case, the KASSERT in
cowfault was added more out of a we should be checking, but I've no idea if
it might happen, I didn't have time last night to analyse how we get to
cowfault, although it mostly looks to be called from doing syscalls.
> BTW, can we move it to arm/arm32/fault.c (and abandon hpcarm version)
> to reduce maintainance load?
Yes, I noticed there's still a few files left in arm32 that should be moved,
potentially all of this lot (taken from file.cats):
file arch/arm32/arm32/procfs_machdep.c procfs
I'll look into doing that this evening for those that use arm32, I'll not do
hpcarm as I know that some of your copies have been changed for soft
interrupts (which perhaps we should port into the generic arm32 code)