Re: [PATCH v3] x86/NMI: Allow processing unknown NMIs when watchdog is enabled

[Ross Lagerwall <ross.lagerwall@citrix.com>]

On 08/28/2014 03:07 PM, Andrew Cooper wrote:
>>   Run an NMI watchdog on each processor.  If a processor is stuck for
>> -longer than the **watchdog\_timeout**, a panic occurs.
>> +longer than the **watchdog\_timeout**, a panic occurs.  When `force` is
>> +specified, in addition to running an NMI watchdog on each processor,
>> +unknown NMIs will still be processed.
>>
>>   ### watchdog\_timeout
>>   > `= <integer>`
>> diff --git a/xen/arch/x86/nmi.c b/xen/arch/x86/nmi.c
>> index c4427a6..873427f 100644
>> --- a/xen/arch/x86/nmi.c
>> +++ b/xen/arch/x86/nmi.c
>> @@ -43,7 +43,18 @@ static DEFINE_PER_CPU(unsigned int, nmi_timer_ticks);
>>
>>   /* opt_watchdog: If true, run a watchdog NMI on each processor. */
>>   bool_t __initdata opt_watchdog = 0;
>> -boolean_param("watchdog", opt_watchdog);
>> +
>> +/* watchdog_force: If true, process unknown NMIs when running the watchdog. */
>> +bool_t watchdog_force = 0;
>> +
>> +static void __init parse_watchdog(char * s)
>> +{
>> +    opt_watchdog = !!parse_bool(s);
>
> A lot of code in Xen gets the use of parse_bool() wrong.  In this case,
> opt_watchdog will be set if garbage is passed in the parameter, which is
> a change in behaviour from before.
>

This is not a change in behavior from before. A bool param if specified 
on the command-line is set to true unless explicitly set to a false-like 
value (e.g. this is how a bare parameter like "watchdog" enables the 
watchdog).

As it has been committed, the behavior has been changed and means 
setting "watchdog" on the command-line does not result in the watchdog 
being enabled.

-- 
Ross Lagerwall