Re: [PATCH] x86/Intel: avoid UB with NMI watchdog on family 15 CPUs

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 19/11/2020 16:37, Jan Beulich wrote:
> On 19.11.2020 17:10, Andrew Cooper wrote:
>> On 19/11/2020 15:57, Jan Beulich wrote:
>>> Found by looking for patterns similar to the one Julien did spot in
>>> pci_vtd_quirks().
>>>
>>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
>> Subject is wonky.  Is it P4 (Intel), or Fam15 (AMD) ?  I'd be tempted to
>> have the prefix as x86/nmi: either way.
> With this code:
>
>     case X86_VENDOR_INTEL:
>         switch (boot_cpu_data.x86) {
>         case 6:
>             setup_p6_watchdog((boot_cpu_data.x86_model < 14) 
>                               ? P6_EVENT_CPU_CLOCKS_NOT_HALTED
>                               : CORE_EVENT_CPU_CLOCKS_NOT_HALTED);
>             break;
>         case 15:
>             if (!setup_p4_watchdog())
>
> I think qualifying it like I did is quite reasonable. Hence ...
>
>> With that suitably adjusted, Acked-by: Andrew Cooper
>> <andrew.cooper3@xxxxxxxxxx>
> ... I'd prefer to keep it as is - please clarify.

Oh - original Xeon's.  I'd honestly forgotten that quirk of history.

I'd recommend "x86/nmi: Avoid UB in for P4-era watchdogs" to avoid the
ambiguity altogether.

~Andrew