Re: [PATCH v3 3/6] x86/sev: Add support to perform RMP optimizations asynchronously

["Kalra, Ashish" <ashish.kalra@amd.com>]

Hello Dave,

On 3/30/2026 7:46 PM, Kalra, Ashish wrote:
> 
> On 3/30/2026 6:22 PM, Dave Hansen wrote:

>>
>>>  static __init void configure_and_enable_rmpopt(void)
>>>  {
>>>  	phys_addr_t pa_start = ALIGN_DOWN(PFN_PHYS(min_low_pfn), SZ_1G);
>>> @@ -499,6 +582,37 @@ static __init void configure_and_enable_rmpopt(void)
>>>  	 */
>>>  	for_each_online_cpu(cpu)
>>>  		wrmsrq_on_cpu(cpu, MSR_AMD64_RMPOPT_BASE, rmpopt_base);
>>
>> What is the scope of MSR_AMD64_RMPOPT_BASE? Can you have it enabled on
>> one thread and not the other? Could they be different values both for
>> enabling and the rmpopt_base value?
>>
>> If it's not per-thread, then why is it being initialized for each thread?
>>
> 
> Only one logical thread per core needs to set RMPOPT_BASE MSR as it is per-core,
> so i will use the "primary_threads_cpumask" here to use it for programming this
> MSR.
> 
> Just another reason, to set the "primary_threads_cpumask" here in this function 
> and then re-use it for the RMPOPT worker.
> 

Coming back to this ...

For using the "primary_thread_cpumask" i will need to use something like
on_each_cpu_mask() similar to what i was doing in v2.

In v2, i was programming the RMPOPT_BASE MSR using on_each_cpu_mask(),
that required using a callback function to do the WRMSR: 

+static void __configure_rmpopt(void *val)
+{
+       u64 rmpopt_base = ((u64)val & PUD_MASK) | MSR_AMD64_RMPOPT_ENABLE;
+
+       wrmsrq(MSR_AMD64_RMPOPT_BASE, rmpopt_base);
+}
+

+       on_each_cpu_mask(cpu_online_mask, __configure_rmpopt, (void *)pa_start, true);


But, that required using the (void *) casting, which you objected to and you 
suggested the use of for_each_online_cpu() and wrmsrq_on_cpu(), and i has replied
that i need to do it (only) once on each thread per core, and that's why i may need
to use on_each_cpu_mask() and then you had suggested that if you *need* performance  
then i can implement/add something like wrmsrq_on_cpumask(). 

For programming the RMPOPT_BASE MSR performance is not really that important as 
it is for issuing the RMPOPT instruction on only thread per core, and as we are
programming the RMPOPT_BASE MSRs on all CPUs/threads to the same (starting) physical
address to support all RAM up-to 2TB for RMP optimizations, therefore, i don't
think it is that critical to implement wrmsrq_on_cpumask() and instead we can continue
to program the RMPOPT_BASE MSR on all CPUs (threads).

Thanks,
Ashish