[PATCH RFC v2] apei: Add ACPI APEI event notification support

Jon Masters jcm at redhat.com
Thu Nov 19 07:05:16 PST 2015

Top post - quick reply - you are correct. This is a vendor specific implementation. APEI is generic, and in the general case there should not be a need to do anything with a special driver. Some will need their own special solution in the early days, but everyone else should be able to go generic. In particular I expect those implementing EL3 to do firmware first handling and to signal the event independently of any OS specific hooks.

Computer Architect | Sent from my 64-bit #ARM Powered phone

> On Nov 19, 2015, at 08:18, Will Deacon <will.deacon at arm.com> wrote:
> [Thanks Boris for CC'ing me]
> Hi all,
> Sorry for the delay on this, I had to do some research and this isn't
> really my area of expertise.
>> On Thu, Oct 22, 2015 at 11:57:16AM +0200, Borislav Petkov wrote:
>> I'm top-posting here because I'm adding some more ARM people to CC and
>> would like for them to see the whole thing.
>> Ok, so what's the strategy here?
>> I know Tomasz did some untangling of GHES stuff to make it load on ARM
>> too and be arch-agnostic. The registration code in it is more than the
>> tiny edac_apei_irq_handler().
>> So why is this thing a separate driver? It is called EDAC_APEI although
>> it is ARM-specific.
>> Why can't it be part of ghes_edac.c with ARM-specific section, if
>> absolutely needed?
>> If this is going to implement the ACPI spec, then I don't see anything
>> vendor-, or arch-specific getting in the way except maybe that APMC0D51
>> id.
>> Hmmm?
> I think you've hit the nail on the head. As far as I can tell, this is
> *not* conformant to ACPI APEI (which should be platform and architecture
> agnostic) and is actually an implementation-specific interface. However,
> I'm happy to be proven wrong if somebody can point me at a document
> describing ACPI APEI in a way that matches this implementation.
>>> +static const struct acpi_device_id edac_apei_match[] = {
>>> +    { "APMC0D51", 0},
>>> +    {},
> This, in particular, gives the game away methinks.
> Will

More information about the linux-arm-kernel mailing list