Re: [PATCH v5 0/5] arm_mpam: resctrl: Counter Assignment (ABMC)

[Gavin Shan]

On 7/8/26 2:24 AM, Ben Horgan wrote:
> Version 5 of this series addresses a few review comments but is pretty
> similar to v4. The precursors in x86/cache have been merged so there is no
> longer a need for this to go through there. It's likely easiest if this
> goes through arm64.
>
> This series adds support for memory bandwidth monitoring. Counter assignment
> will be used for all platforms with memory bandwidth monitors rather than just
> those with insufficient counters as originally proposed.
>
> rfc v1: https://lore.kernel.org/lkml/20260225205436.3571756-1-ben.horgan@xxxxxxx/
> rfc v2: https://lore.kernel.org/lkml/20260319165540.381410-1-ben.horgan@xxxxxxx/
> v3: https://lore.kernel.org/linux-arm-kernel/20260511154147.557481-1-ben.horgan@xxxxxxx/
> v4: https://lore.kernel.org/lkml/20260520212458.1797221-1-ben.horgan@xxxxxxx/
>
> The code can be found at:
> https://gitlab.arm.com/linux-arm/linux-bh.git mpam_abmc_v5
>
> Ben Horgan (2):
>    arm_mpam: resctrl: Pre-allocate assignable monitors
>    arm64: mpam: Add memory bandwidth usage (MBWU) documentation
>
> James Morse (3):
>    arm_mpam: resctrl: Pick classes for use as MBM counters
>    arm_mpam: resctrl: Add resctrl_arch_config_cntr() for ABMC use
>    arm_mpam: resctrl: Add resctrl_arch_cntr_read() &
>      resctrl_arch_reset_cntr()
>
>   Documentation/arch/arm64/mpam.rst |  22 +++
>   drivers/resctrl/mpam_internal.h   |   6 +-
>   drivers/resctrl/mpam_resctrl.c    | 318 +++++++++++++++++++++++++++---
>   3 files changed, 323 insertions(+), 23 deletions(-)

The tests on NVidia's grace-hopper machine looks good: (a) no error or warning is seen
from the kunit-tests; (b) L3 cache partitioning and MBW (soft) limiting aren't broken,
still working to me; (c) The counter 'mbm_total_bytes' can be assigned to newly created
group and it gets increased when workload runs in the group.

Tested-by: Gavin Shan <gshan@xxxxxxxxxx>

Thanks,
Gavin