Re: [RFC 2.6.11.2 1/1] Add reboot fixup for gx1/cs5530a

["H. Peter Anvin" <hpa@zytor.com>]

jayalk@intworks.biz wrote:
> Hi Riley, Dave, Peter, i386 boot/workaround maintainers,
> 
> I ran into a problem getting reboot working with 2.6.11 on an embedded
> board. The board has a Geode GX1 with a CS5530A companion. What I observe on
> reboot is the "Restarting system" printk, and then a cpu stall/hang. I think
> the problem arises because the keyboard controller is disabled by the BIOS,
> so the traditional mach_reboot()'s output to port 0x64 is ignored. Then the
> 386 triple fault issued after mach_reboot() results in a shutdown (because
> the hardware doesn't have to detect the triple fault and issue a reset).
> That then gives the end result of a stalled cpu/hang. 
> 
> I found that the CS5530A in question has a "issue system wide reset" bit.
> The reboot works cleanly if I write that bit rather than do mach_reboot().
> So the following patch is my attempt to incorporate that change into 2.6.11
> by adding a X86_REBOOTFIXUPS option. In order to keep reboot.c free of hw
> specific fixups, I put it in another file, reboot_fixups.c. I tried to make
> it a bit generic so that if there are other reboot related fixups for other
> chipsets/boards, there'd be a clean place to put it. Please let me know what
> you think.
> 

This makes a lot of sense to me.  I appreciate the fact that you took 
the time to properly add detection/abstraction code.

	-hpa