Re: drivers/macintosh/Kconfig (was: Re: Linux 2.6.3-rc1)

On Mon, Feb 09, 2004 at 09:22:35AM +1100, Benjamin Herrenschmidt wrote:

I agree that the abuse of adb_request should go, though replacing
it isn't easy at this stage. It's not worth bothering with a
common interface for things like poweroff, reset and RTC imho,
it would only solve a small part of the problem, there are too
many places where we actually need to send directly a PMU command.

Based on my reading of the code, most of the places sending PMU
commands directly are actually calling pmu_request, so I was
planning on leaving that alone as a hook directly into the PMU
driver, and although it's related to ADB, it isn't going through
the current ADB framework, either.

Part of the problem is that the PMU driver low level state machine
it tied to the format of the adb_request structure. I don't think
I will fix any of that for 2.6. For 2.7, I may define a low level
pmu_request structure _without_ embedded buffers and have the ADB
request handling allocate one of those atomically.

I don't see where that is an issue. The PMU driver internally can
continue to be based on the same structure, and interface with
the new ADB layer just for ADB messages. I suppose it will be clearer
once I get that part of the code finished. I've been concentrating
more on the 68k side of things for this, since I don't care if I
confuse one of my older boxes.

	Brad Boyer
	flar@allandria.com


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/