Re: [PATCH] KVM: x86: Allow suppressing prints on RDMSR/WRMSR of unhandled MSRs

[Paolo Bonzini <pbonzini@redhat.com>]

On 09/11/2017 16:22, Radim Krčmář wrote:
> 2017-11-08 14:32+0200, Eyal Moscovici:
>> Some guests use these unhandled MSRs very frequently.
>> This cause dmesg to be populated with lots of aggregated messages on
>> usage of ignored MSRs. As ignore_msrs=true means that the user is
>> well-aware his guest use ignored MSRs, allow to also disable the
>> prints on their usage.
>>
>> An example of such guest is ESXi which tends to access a lot to MSR
>> 0x34 (MSR_SMI_COUNT) very frequently.
>>
>> In addition, we have observed this to cause unnecessary delays to
>> guest execution. Such an example is ESXi which experience networking
>> delays in it's guests (L2 guests) because of these prints (even when
>> prints are rate-limited). This can easily be reproduced by pinging
>> from one L2 guest to another.  Once in a while, a peak in ping RTT
>> will be observed. Removing these unhandled MSR prints solves the
>> issue.
>>
>> Because these prints can help diagnose issues with guests,
>> this commit only suppress them by a module parameter instead of
>> removing them from code entirely.
>>
>> Signed-off-by: Eyal Moscovici <eyal.moscovici@oracle.com>
>> Reviewed-by: Liran Alon <liran.alon@oracle.com>
>> Reviewed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
>> Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
>> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
>> ---
>>  arch/x86/kvm/x86.c | 17 +++++++++++++----
>>  1 file changed, 13 insertions(+), 4 deletions(-)
>>
>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>> index 03869eb..21c0059 100644
>> --- a/arch/x86/kvm/x86.c
>> +++ b/arch/x86/kvm/x86.c
>> @@ -107,6 +107,9 @@
>>  static bool __read_mostly ignore_msrs = 0;
>>  module_param(ignore_msrs, bool, S_IRUGO | S_IWUSR);
>>  
>> +static bool __read_mostly suppress_ignore_msrs_prints = false;
>> +module_param(suppress_ignore_msrs_prints, bool, S_IRUGO | S_IWUSR);
> 
> What about calling it 'print_ignored_msrs' and default to true?

Makes sense! s/print/report/ perhaps?

Thanks,

Paolo

> (I can change that when applying.)
> 
> Thanks.
>