Re: [PATCH v4 3/7] x86/sev: Initialize RMPOPT configuration MSRs

[Kalra, Ashish]

Hello Ackerley,

On 5/1/2026 1:12 PM, Ackerley Tng wrote:
> Ashish Kalra <Ashish.Kalra@xxxxxxx> writes:
> 
>>
>> [...snip...]
>>
>> +void snp_setup_rmpopt(void)
>> +{
>> +	u64 rmpopt_base;
>> +	int cpu;
>> +
>> +	if (!cpu_feature_enabled(X86_FEATURE_RMPOPT))
>> +		return;
>> +
>> +	/*
>> +	 * RMPOPT_BASE MSR is per-core, so only one thread per core needs to
>> +	 * setup RMPOPT_BASE MSR.
>> +	 */
>> +
>> +	for_each_online_cpu(cpu) {
> 
> Dave mentioned hotplug in v3 [1], which led me to check. From this
> series, it looks like RMPOPT won't ever be performed for newly-plugged
> cpus, is that okay?

Yes, with this current series, RMPOPT won't be performed on newly-plugged
CPUs, i was computing and storing the cpumask for threads to fire RMPOPT on,
statically during the setup code to save computing it again on the
worker thread, but i believe the simple fix for handling hotplugged CPUs
would be to compute the cpumask every time the worker thread executes.

> 
>> +		if (!topology_is_primary_thread(cpu))
>> +			continue;
>> +
>> +		cpumask_set_cpu(cpu, &rmpopt_cpumask);
> 
> nit: perhaps flipping the condition is easier to read:

Yes.

> 
>     if (topology_is_primary_thread(cpu))
>     	cpumask_set_cpu()
> 
>> +	}
>> +
>> +	rmpopt_pa_start = ALIGN_DOWN(PFN_PHYS(min_low_pfn), SZ_1G);
>> +	rmpopt_base = rmpopt_pa_start | MSR_AMD64_RMPOPT_ENABLE;
>> +
>> +	/*
>> +	 * Per-CPU RMPOPT tables support at most 2 TB of addressable memory
>> +	 * for RMP optimizations. Initialize the per-CPU RMPOPT table base
>> +	 * to the starting physical address to enable RMP optimizations for
>> +	 * up to 2 TB of system RAM on all CPUs.
>> +	 */
>> +	wrmsrq_on_cpus(&rmpopt_cpumask, MSR_AMD64_RMPOPT_BASE, rmpopt_base);
>> +}
> 
> [1] https://lore.kernel.org/all/ab41b1d8-e464-4ad6-ac07-7318686db10e@xxxxxxxxx/
> 
>>
>> [...snip...]
>>