Re: [PATCH v13 01/27] x86/cpufeatures: Add support for Assignable Bandwidth Monitoring Counters (ABMC)

[Moger, Babu]

Hi Reinette,

On 5/22/25 15:51, Reinette Chatre wrote:
> Hi Babu,
> 
> On 5/15/25 3:51 PM, Babu Moger wrote:
>> Users can create as many monitor groups as RMIDs supported by the hardware.
>> However, bandwidth monitoring feature on AMD system only guarantees that
>> RMIDs currently assigned to a processor will be tracked by hardware. The
>> counters of any other RMIDs which are no longer being tracked will be reset
>> to zero. The MBM event counters return "Unavailable" for the RMIDs that are
>> not tracked by hardware. So, there can be only limited number of groups
>> that can give guaranteed monitoring numbers. With ever changing
>> configurations there is no way to definitely know which of these groups are
>> being tracked for certain point of time. Users do not have the option to
>> monitor a group or set of groups for certain period of time without
>> worrying about RMID being reset in between.
>>
>> The ABMC feature provides an option to the user to assign a hardware
>> counter to an RMID, event pair and monitor the bandwidth as long as it is
>> assigned. The assigned RMID will be tracked by the hardware until the user
>> unassigns it manually. There is no need to worry about counters being reset
>> during this period. Additionally, the user can specify a bitmask
>> identifying the specific bandwidth types from the given source to track
>> with the counter.
>>
>> Without ABMC enabled, monitoring will work in current mode without
>> assignment option.
>>
>> The Linux resctrl subsystem provides an interface that allows monitoring of
>> up to two memory bandwidth events per group, selected from a combination of
>> available total and local events. When ABMC is enabled, two events will be
>> assigned to each group by default, in line with the current interface
>> design. Users will also have the option to configure which types of memory
>> transactions are counted by these events.
>>
>> Due to the limited number of available counters (32), users may quickly
>> exhaust the available counters. If the system runs out of assignable ABMC
>> counters, the kernel will report an error. In such cases, users will nee
>> dto unassign one or more active counters to free up countes for new
> 
> "nee dto" -> "need to"
> "countes" -> "counters"

Sure.

> 
>> assignments. The interface will provide options to assign or unassign
> 
> "The interface will" -> "resctrl will"?
> 

Sure.

>> events through the group-specific interface file.
>>
>> The feature can be detected via CPUID_Fn80000020_EBX_x00 bit 5.
> 
> "The feature can be detected" -> "The feature is detected"
> 

Sure.

>> Bits Description
>> 5    ABMC (Assignable Bandwidth Monitoring Counters)
>>
>> 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).
>>
>> Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537
>> Signed-off-by: Babu Moger <babu.moger@xxxxxxx>
>> ---
> 
> ...
>>  arch/x86/include/asm/cpufeatures.h | 1 +
>>  arch/x86/kernel/cpu/cpuid-deps.c   | 2 ++
>>  arch/x86/kernel/cpu/scattered.c    | 1 +
>>  3 files changed, 4 insertions(+)
>>
>> diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
>> index 6c2c152d8a67..d5c14dc678df 100644
>> --- a/arch/x86/include/asm/cpufeatures.h
>> +++ b/arch/x86/include/asm/cpufeatures.h
>> @@ -481,6 +481,7 @@
>>  #define X86_FEATURE_AMD_HETEROGENEOUS_CORES (21*32 + 6) /* Heterogeneous Core Topology */
>>  #define X86_FEATURE_AMD_WORKLOAD_CLASS	(21*32 + 7) /* Workload Classification */
>>  #define X86_FEATURE_PREFER_YMM		(21*32 + 8) /* Avoid ZMM registers due to downclocking */
>> +#define X86_FEATURE_ABMC		(21*32 + 9) /* Assignable Bandwidth Monitoring Counters */
>>  
>>  /*
>>   * BUG word(s)
>> diff --git a/arch/x86/kernel/cpu/cpuid-deps.c b/arch/x86/kernel/cpu/cpuid-deps.c
>> index a2fbea0be535..2f54831e04e5 100644
>> --- a/arch/x86/kernel/cpu/cpuid-deps.c
>> +++ b/arch/x86/kernel/cpu/cpuid-deps.c
>> @@ -71,6 +71,8 @@ static const struct cpuid_dep cpuid_deps[] = {
>>  	{ X86_FEATURE_CQM_MBM_LOCAL,		X86_FEATURE_CQM_LLC   },
>>  	{ X86_FEATURE_BMEC,			X86_FEATURE_CQM_MBM_TOTAL   },
>>  	{ X86_FEATURE_BMEC,			X86_FEATURE_CQM_MBM_LOCAL   },
>> +	{ X86_FEATURE_ABMC,			X86_FEATURE_CQM_MBM_TOTAL   },
>> +	{ X86_FEATURE_ABMC,			X86_FEATURE_CQM_MBM_LOCAL   },
> 
> Is this dependency still accurate now that the implementation switched to the 
> "extended event ID" variant of ABMC that no longer uses the event IDs associated
> with X86_FEATURE_CQM_MBM_TOTAL and X86_FEATURE_CQM_MBM_LOCAL?

That's a good question. Unfortunately, we may need to retain this
dependency for now, as a significant portion of the code relies on
functions like resctrl_is_mbm_event(), resctrl_is_mbm_enabled(),
resctrl_arch_is_mbm_total_enabled(), and others.


> 
>>  	{ X86_FEATURE_AVX512_BF16,		X86_FEATURE_AVX512VL  },
>>  	{ X86_FEATURE_AVX512_FP16,		X86_FEATURE_AVX512BW  },
>>  	{ X86_FEATURE_ENQCMD,			X86_FEATURE_XSAVES    },
>> diff --git a/arch/x86/kernel/cpu/scattered.c b/arch/x86/kernel/cpu/scattered.c
>> index 16f3ca30626a..3b72b72270f1 100644
>> --- a/arch/x86/kernel/cpu/scattered.c
>> +++ b/arch/x86/kernel/cpu/scattered.c
>> @@ -49,6 +49,7 @@ static const struct cpuid_bit cpuid_bits[] = {
>>  	{ X86_FEATURE_MBA,			CPUID_EBX,  6, 0x80000008, 0 },
>>  	{ X86_FEATURE_SMBA,			CPUID_EBX,  2, 0x80000020, 0 },
>>  	{ X86_FEATURE_BMEC,			CPUID_EBX,  3, 0x80000020, 0 },
>> +	{ X86_FEATURE_ABMC,			CPUID_EBX,  5, 0x80000020, 0 },
>>  	{ X86_FEATURE_AMD_WORKLOAD_CLASS,	CPUID_EAX, 22, 0x80000021, 0 },
>>  	{ X86_FEATURE_PERFMON_V2,		CPUID_EAX,  0, 0x80000022, 0 },
>>  	{ X86_FEATURE_AMD_LBR_V2,		CPUID_EAX,  1, 0x80000022, 0 },
> 
> Reinette
> 

-- 
Thanks
Babu Moger