On 07/03/2018 12:58 PM, Luck, Tony wrote:
> On Tue, Jul 03, 2018 at 12:48:44PM -0400, Prarit Bhargava wrote:
>> On systems where a runtime microcode update has occurred the
>> microcode version is wrong because boot_cpu_data.microcode is
>> not updated during runtime.
>>
>> Use the per-CPU microcode version in the MCE message.
>
> Other option would be to fix the microcode update code to make
> sure that boot_cpu_data.microcode is right.
I thought about that too [1] but TBH I've always been a bit confused by
boot_cpu_data vs. cpu_data(). At one point I thought that boot_cpu_data is
valid only up to a specific point in boot and from that point on the cpu_data()
should be used but I'm no longer sure if that's correct.
If that is not correct when should boot_cpu_data be used vs cpu_data()? Is
boot_cpu_data truly supposed to represent the general state of the boot cpu at
boot time?
Look at the same function mce_setup() that I modified in my patch. There is
another reference to boot_cpu_data.x86_vendor (which obviously will never change
during runtime) that likely should be cpu_data() if my understanding above is
correct.
P.
[1] Since the kernel now uses a cpu rendezvous model to update the microcode it
would be trivial to set boot_cpu_data.microcode to the new value.
>
> But if we go this route we should add:
>
> Fixes: fa94d0c6e0f3 ("x86/MCE: Save microcode revision in machine check
> records")
> Cc: sta...@vger.kernel.org
>
> Can also get a:
>
> Reviewed-by: Tony Luck <tony.l...@intel.com>
>
> -Tony Luck
>
