Re: [PATCH v6 20/22] x86/resctrl: Enable AMD ABMC feature by default when supported

[Moger, Babu]

Hi James,

On 8/16/24 11:32, James Morse wrote:
> Hi Babu,
> 
> On 06/08/2024 23:00, Babu Moger wrote:
>> Enable ABMC by default when supported during the boot up.
>>
>> Users will not see any difference in the behavior when resctrl is
>> mounted. With automatic assignment everything will work as running
>> in the legacy monitor mode.
> 
>> diff --git a/arch/x86/kernel/cpu/resctrl/core.c b/arch/x86/kernel/cpu/resctrl/core.c
>> index 6fb0cfdb5529..a7980f84c487 100644
>> --- a/arch/x86/kernel/cpu/resctrl/core.c
>> +++ b/arch/x86/kernel/cpu/resctrl/core.c
>> @@ -599,6 +599,7 @@ static void domain_add_cpu_mon(int cpu, struct rdt_resource *r)
>>  		d = container_of(hdr, struct rdt_mon_domain, hdr);
>>  
>>  		cpumask_set_cpu(cpu, &d->hdr.cpu_mask);
>> +		resctrl_arch_mbm_cntr_assign_configure();
>>  		return;
>>  	}
>>  
>> @@ -620,6 +621,7 @@ static void domain_add_cpu_mon(int cpu, struct rdt_resource *r)
>>  	arch_mon_domain_online(r, d);
>>  
>>  	resctrl_mbm_evt_config_init(hw_dom);
>> +	resctrl_arch_mbm_cntr_assign_configure();
>>  
>>  	if (arch_domain_mbm_alloc(r->mon.num_rmid, hw_dom)) {
>>  		mon_domain_free(hw_dom);
> 
>> diff --git a/arch/x86/kernel/cpu/resctrl/rdtgroup.c b/arch/x86/kernel/cpu/resctrl/rdtgroup.c
>> index 66febff2a3d3..d15fd1bde5f4 100644
>> --- a/arch/x86/kernel/cpu/resctrl/rdtgroup.c
>> +++ b/arch/x86/kernel/cpu/resctrl/rdtgroup.c
>> @@ -2756,6 +2756,23 @@ void resctrl_arch_mbm_cntr_assign_disable(void)
>>  	}
>>  }
>>  
>> +void resctrl_arch_mbm_cntr_assign_configure(void)
>> +{
>> +	struct rdt_resource *r = &rdt_resources_all[RDT_RESOURCE_L3].r_resctrl;
>> +	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(r);
>> +	bool enable = true;
>> +
>> +	mutex_lock(&rdtgroup_mutex);
> 
> As before - this lock isn't available to the architecture code after the filesystem code
> moves to /fs/. To prevent concurrent calls to resctrl_abmc_set_one_amd() I think you need
> your own mutex.

These calls are originated from domain_add_cpu_mon sequence. It holds the
lock already. We should be good. I will remove the lock.

> 
> 
>> +	if (r->mon.mbm_cntr_assignable) {
>> +		if (!hw_res->mbm_cntr_assign_enabled)
>> +			hw_res->mbm_cntr_assign_enabled = true;
>> +		resctrl_abmc_set_one_amd(&enable);
>> +	}
>> +
>> +	mutex_unlock(&rdtgroup_mutex);
>> +}
> 
> Neither of this functions callers are in filesystem code, could you drop the 'arch' from
> the name - it isn't part of the fs/arch interface.

Yes. Sure.

> 
> 
> Thanks,
> 
> James
> 

-- 
Thanks
Babu Moger