Re: [PATCH V2 2/3] target/i386: call cpu_exec_realizefn before x86_cpu_filter_features

[Chen, Zide]

On 5/30/2024 11:30 PM, Zhao Liu wrote:
> Hi Zide,
> 
> On Fri, May 24, 2024 at 01:00:16PM -0700, Zide Chen wrote:
>> Date: Fri, 24 May 2024 13:00:16 -0700
>> From: Zide Chen <zide.chen@intel.com>
>> Subject: [PATCH V2 2/3] target/i386: call cpu_exec_realizefn before
>>  x86_cpu_filter_features
>> X-Mailer: git-send-email 2.34.1
>>
>> cpu_exec_realizefn which calls the accel-specific realizefn may expand
>> features.  e.g., some accel-specific options may require extra features
>> to be enabled, and it's appropriate to expand these features in accel-
>> specific realizefn.
>>
>> One such example is the cpu-pm option, which may add CPUID_EXT_MONITOR.
>>
>> Thus, call cpu_exec_realizefn before x86_cpu_filter_features to ensure
>> that it won't expose features not supported by the host.
>>
>> Fixes: 662175b91ff2 ("i386: reorder call to cpu_exec_realizefn")
>> Suggested-by: Xiaoyao Li <xiaoyao.li@intel.com>
>> Signed-off-by: Zide Chen <zide.chen@intel.com>
>> ---
>>  target/i386/cpu.c         | 24 ++++++++++++------------
>>  target/i386/kvm/kvm-cpu.c |  1 -
>>  2 files changed, 12 insertions(+), 13 deletions(-)
>>
>> diff --git a/target/i386/cpu.c b/target/i386/cpu.c
>> index bc2dceb647fa..a1c1c785bd2f 100644
>> --- a/target/i386/cpu.c
>> +++ b/target/i386/cpu.c
>> @@ -7604,6 +7604,18 @@ static void x86_cpu_realizefn(DeviceState *dev, Error 
>> **errp)
>>          }
>>      }
>>  
>> +    /*
>> +     * note: the call to the framework needs to happen after feature 
>> expansion,
>> +     * but before the checks/modifications to ucode_rev, mwait, phys_bits.
>> +     * These may be set by the accel-specific code,
>> +     * and the results are subsequently checked / assumed in this function.
>> +     */
>> +    cpu_exec_realizefn(cs, &local_err);
>> +    if (local_err != NULL) {
>> +        error_propagate(errp, local_err);
>> +        return;
>> +    }
>> +
>>      x86_cpu_filter_features(cpu, cpu->check_cpuid || cpu->enforce_cpuid);
> 
> For your case, which sets cpu-pm=on via overcommit, then
> x86_cpu_filter_features() will complain that mwait is not supported.
> 
> Such warning is not necessary, because the purpose of overcommit (from
> code) is only to support mwait when possible, not to commit to support
> mwait in Guest.
> 
> Additionally, I understand x86_cpu_filter_features() is primarily
> intended to filter features configured by the user, 

Yes, that's why this patches intends to let x86_cpu_filter_features()
filter out the MWAIT bit which is set from the overcommit option.

> and the changes of
> CPUID after x86_cpu_filter_features() should by default be regarded like
> "QEMU knows what it is doing".

Sure, we can add feature bits after x86_cpu_filter_features(), but I
think moving cpu_exec_realizefn() before x86_cpu_filter_features() is
more generic, and actually this is what QEMU did before commit 662175b91ff2.

- Less redundant code. Specifically, no need to call
x86_cpu_get_supported_feature_word() again.
- Potentially there could be other features could be added from the
accel-specific realizefn, kvm_cpu_realizefn() for example.  And these
features need to be checked against the host availability.

> 
> I feel adding a check for the CPUID mwait bit in host_cpu_realizefn()
> is enough, after all, this bit should be present if host supports mwait
> and enable_cpu_pm (in kvm_arch_get_supported_cpuid()).

Besides the above reasons, it seems to me expanding env->features in
host-cpu.c is confusing.

> Thanks,
> Zhao
>