On 08/01/2018 11:43 AM, Borislav Petkov wrote:
> (dropping stable@)
>
> On Tue, Jul 31, 2018 at 07:27:39AM -0400, Prarit Bhargava wrote:
>> I tested this on AMD Ryzen & Intel Broadwell system and dumped the
>> boot_cpu_data before and after a microcode update. On the Intel
>> system I also did a fatal MCE using mce-inject to confirm the output
>> from the mce handling code.
>
> Ok, sending the two patches I've massaged into submission as a reply to
> this message - please run them on your boxes. You can also inject MCEs
> on the AMD box to verify - simply enable CONFIG_X86_MCE_INJECT, go to
> debugfs and do, for example:
>
> cd /sys/kernel/debug/mce-inject/
> echo 10 > /sys/devices/system/machinecheck/machinecheck0/check_interval
> echo 0x9c7d410092080813 > status; echo 0x000000006d3d483b > addr; echo 2 >
> cpu; echo 4 > bank; echo hw > flags
> echo 0xbc00200000010015 > status; echo 0x00000000942e4d98 > addr; echo 6 >
> cpu; echo 2 > bank; echo hw > flags
> echo 0xfe000010000b0c0f > status; echo 0xe1101add1e550012 > addr; echo
> 0xcafebeef > misc; echo 2 > cpu; echo hw > flags; echo 4 > bank
>
Borislav, I'm not sure if you saw his comments but Oleksandr, cc'd, has an
objection to overwriting the stored microcode version in boot_cpu_data. He has
found that information useful in debugging microcode issues.
P.
> Thanks!
>
