Re: [Xen-devel] [PATCH 1/2] xen/mce: Add mcelog support for Xen platform

["Liu, Jinsong" <jinsong.liu@xxxxxxxxx>]

Konrad Rzeszutek Wilk wrote:
>>>> That's vMCE injection logic.
>>> 
>>> Are you sure about it? The comments in it speak of piggybacking on
>>> the native MCE handling routines. But since that is not used anymore
>>> do you need to use a different mechanism?
>> 
>> What is not used anymore? what's your concern about
>> cvt_gate_to_trap? I really confused here. Could you elaborate more?
> 
> Well, the mce.c is not involved anymore. So if it we are piggybacking
> on the native MCE handling routines - those routines
> (do_machine_check) won't deliever the data to your driver anymore?
> B/c the do_machine_check is doing mce_log which spools data. But your
> driver is using a different system to de-spool the data - and it does
> not use the mcelog structure array.

native mce is still there, we didn't mask it.
what is not used is native mce_log, w/ guest info (like guest address) when 
running at xen platform,
instead we use xen_mce_log, a self-contained logic getting real physical error 
info from xen hypervisor.
that's just what we expected.

> 
> .. snip.
>>>> git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
>>>> c/s a01ee165a132fadb57659d26246e340d6ac53265
>>> 
>>> Which I think the tree is based on too.
>> 
>> So it would not be stuck?
> 
> I've no idea what you mean here. Could you elaborate please?

What I meant here is, it should not meet confliction when you patched our 
patches to latest linus tree.
(Per my understanding, you meet confliction when patched, right? any 
misunderstanding please tell me)

Thanks,
Jinsong
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel