Re: [PATCH v16 08/34] x86,fs/resctrl: Detect Assignable Bandwidth Monitoring feature details

["Moger, Babu" <babu.moger@amd.com>]

Hi Reinette,

On 7/30/25 14:49, Reinette Chatre wrote:
> Hi Babu,
> 
> On 7/25/25 11:29 AM, Babu Moger wrote:
>> ABMC feature details are reported via CPUID Fn8000_0020_EBX_x5.
>> Bits Description
>> 15:0 MAX_ABMC Maximum Supported Assignable Bandwidth
>>      Monitoring Counter ID + 1
>>
>> The feature details are documented in APM listed below [1].
>> [1] AMD64 Architecture Programmer's Manual Volume 2: System Programming
>> Publication # 24593 Revision 3.41 section 19.3.3.3 Assignable Bandwidth
>> Monitoring (ABMC).
>>
>> Detect the feature and number of assignable counters supported. For
>> backward compatibility, upon detecting the assignable counter feature,
>> enable the mbm_total_bytes and mbm_local_bytes events that users are
>> familiar with as part of original L3 MBM support.
>>
>> Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537
>> Signed-off-by: Babu Moger <babu.moger@amd.com>
>> ---
> 
> ...
> 
>> diff --git a/arch/x86/kernel/cpu/resctrl/core.c b/arch/x86/kernel/cpu/resctrl/core.c
>> index 267e9206a999..09cb5a70b1cb 100644
>> --- a/arch/x86/kernel/cpu/resctrl/core.c
>> +++ b/arch/x86/kernel/cpu/resctrl/core.c
>> @@ -883,6 +883,8 @@ static __init bool get_rdt_mon_resources(void)
>>  		resctrl_enable_mon_event(QOS_L3_MBM_LOCAL_EVENT_ID);
>>  		ret = true;
>>  	}
>> +	if (rdt_cpu_has(X86_FEATURE_ABMC))
>> +		ret = true;
>>  
>>  	if (!ret)
>>  		return false;
>> @@ -990,7 +992,8 @@ void resctrl_cpu_detect(struct cpuinfo_x86 *c)
>>  
> 
> To complement the change below, shouldn't the snippet that precedes it look like:
> 	if (!cpu_has(c, X86_FEATURE_CQM_LLC) && !cpu_has(c, X86_FEATURE_ABMC)) {
> 		...
> 		return;
> 	}


Sure. Added now.

> 
>>  	if (cpu_has(c, X86_FEATURE_CQM_OCCUP_LLC) ||
>>  	    cpu_has(c, X86_FEATURE_CQM_MBM_TOTAL) ||
>> -	    cpu_has(c, X86_FEATURE_CQM_MBM_LOCAL)) {
>> +	    cpu_has(c, X86_FEATURE_CQM_MBM_LOCAL) ||
>> +	    cpu_has(c, X86_FEATURE_ABMC)) {
>>  		u32 eax, ebx, ecx, edx;
>>  
>>  		/* QoS sub-leaf, EAX=0Fh, ECX=1 */
>> diff --git a/arch/x86/kernel/cpu/resctrl/monitor.c b/arch/x86/kernel/cpu/resctrl/monitor.c
>> index 2558b1bdef8b..0a695ce68f46 100644
>> --- a/arch/x86/kernel/cpu/resctrl/monitor.c
>> +++ b/arch/x86/kernel/cpu/resctrl/monitor.c
>> @@ -339,6 +339,7 @@ int __init rdt_get_mon_l3_config(struct rdt_resource *r)
>>  	unsigned int mbm_offset = boot_cpu_data.x86_cache_mbm_width_offset;
>>  	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(r);
>>  	unsigned int threshold;
>> +	u32 eax, ebx, ecx, edx;
>>  
>>  	snc_nodes_per_l3_cache = snc_get_config();
>>  
>> @@ -368,14 +369,18 @@ int __init rdt_get_mon_l3_config(struct rdt_resource *r)
>>  	 */
>>  	resctrl_rmid_realloc_threshold = resctrl_arch_round_mon_val(threshold);
>>  
>> -	if (rdt_cpu_has(X86_FEATURE_BMEC)) {
>> -		u32 eax, ebx, ecx, edx;
>> -
>> +	if (rdt_cpu_has(X86_FEATURE_BMEC) || rdt_cpu_has(X86_FEATURE_ABMC)) {
>>  		/* Detect list of bandwidth sources that can be tracked */
>>  		cpuid_count(0x80000020, 3, &eax, &ebx, &ecx, &edx);
>>  		r->mon.mbm_cfg_mask = ecx & MAX_EVT_CONFIG_BITS;
> 
> I interpret this mbm_cfg_mask initialization that an ABMC system will report which of
> the memory transactions can be monitored. 
> In patch #15 "fs/resctrl: Introduce event configuration field in struct mon_evt"
> the event configurations of memory transactions that should be monitored are hardcoded
> as below without taking into account what the system supports:
> 
> 	resctrl_mon_resource_init() {
> 		...
> 		mon_event_all[QOS_L3_MBM_TOTAL_EVENT_ID].evt_cfg = MAX_EVT_CONFIG_BITS;
> 		mon_event_all[QOS_L3_MBM_LOCAL_EVENT_ID].evt_cfg = READS_TO_LOCAL_MEM |
> 								   READS_TO_LOCAL_S_MEM |
> 								   NON_TEMP_WRITE_TO_LOCAL_MEM;
> 		...
> 	}

That is correct.
Changed the assignment.

mon_event_all[QOS_L3_MBM_TOTAL_EVENT_ID].evt_cfg = r->mon.mbm_cfg_mask;
mon_event_all[QOS_L3_MBM_LOCAL_EVENT_ID].evt_cfg =
r->mon.mbm_cfg_mask & (READS_TO_LOCAL_MEM | READS_TO_LOCAL_S_MEM |
NON_TEMP_WRITE_TO_LOCAL_MEM);



> 
> It may thus be that a system may not support all memory transactions it is configured to
> monitor. It seems to me that the initialization done in resctrl_mon_resource_init() needs
> to take r->mon.mbm_cfg_mask (what the system supports) into account? If so, then
> the same hardcoding done by patch #32 in resctrl_mbm_assign_mode_write() should
> also be changed.

Yes. Sure.

-- 
Thanks
Babu Moger