On Thu, Feb 07, 2013 at 03:01:50PM +0000, David Woodhouse wrote:
On Thu, 2013-02-07 at 08:40 +0000, David Woodhouse wrote:
On Wed, 2013-02-06 at 23:13 -0500, Kevin O'Connor wrote:
Restoring flags on exit is assuredly more correct, and if I move timer_setup() to the prepare-to-boot phase this problem goes away for me. So, I think this is more evidence that the timer can not be enabled in the init-yourself phase.
Yeah, I'll change that. Thanks.
Ah, I see you already did. It breaks the pmtimer support though; we end up calling timer_setup() and thus setting cpu_khz *after* pmtimer_setup() has set it. Was that why you disabled CONFIG_PMTIMER for the CSM build in your tree? I've fixed that up in my tree now, as well as reinstating the pmtimer support for Xen and Coreboot builds.
Thanks. I rolled up your changes and my changes into a new patch 17 and pushed it out along with your patches 4, 6, 7, 13, 14 (modified) 15, and 16. In addition to the reloc changes and fixes we discussed, I modified the core csm patch (patch 17) to not depend on pmtimer and to remove some Kconfig changes. I wanted to merge the core CSM code and then add those on top.
So, at this point, I think still pending is: Kconfig changes, pmtimer, UmbStart/UmbEnd, and unaligned PCIR.
Kconfig - I'm not sure we should have a qemu vs coreboot vs csm selection as it's possible we can have qemu+csm (or even qemu+coreboot+csm) and I think Kconfig should reflect that reality. I'll put together a patch to demonstrate what I'm thinking.
pmtimer - the patches are fine - the coreboot/xen parts just need to be tested before merging. I may be able to test the coreboot part on my e350m1.
UmbStart/UmbEnd - I'd like to hear back from the OVMF team before merging this.
unaligned PCIR - I'm not sure we can merge this as it would break QEMU users with a current "lgpl vgabios".