Re: [PATCH v6 0/8] platform/x86/amd/hsmp: Family 1Ah Model 50h-5Fh HSMP and metrics

[Ilpo Järvinen]

On Fri, 12 Jun 2026, Muralidhara M K wrote:

> Hi,
> 
> This series extends the AMD HSMP platform driver and UAPI for Family 1Ah,
> Model 50h-5Fh: new HSMP message IDs (including PQoS-related IDs in the
> first patch), Zen 6 metric table layout in UAPI, relaxed response_sz
> handling against the descriptor table, firmware-reported metric table
> sizing for ioremap and reads, HSMP_IOCTL_GET_TELEMETRY_DATA on the
> existing character device for a single-shot full-table copy (avoiding
> sysfs PAGE_SIZE limits and torn reads), ACPI enablement for protocol
> version 7 tables, Spectre-v1 hardening on user-controlled indices, and
> coordinated metric-table access plus ACPI HSMP probe/remove lifetime
> fixes on the shared global socket array.  Documentation is updated in
> Documentation/arch/x86/amd_hsmp.rst.
> 
> v5 submission (lore.kernel.org)
> 
>   https://lore.kernel.org/platform-driver-x86/20260611052919.1095549-1-muralidhara.mk@xxxxxxx/T/#t
> 
> Changes since v5
> 
> * Patch 6/8: Tighten the Spectre-v1 changelog and subject around
>   clamping ioctl and hsmp_send_message() indices; wrap the commit
>   description so scripts/checkpatch.pl stays quiet on line length.
> 
> * Patch 8/8: Broaden the final fix beyond sysfs metric locking alone.
>   Serialize metric table refresh and memcpy_fromio(), refcount ACPI
>   HSMP sockets with a probe mutex across init_acpi(), misc_register(),
>   and refcount_inc(), replace devm metric DRAM maps with explicit
>   ioremap()/mutex_init plus hsmp_metric_tbl_unmap_all() on last ACPI
>   remove, add hsmp_acpi_free_sock_alloc_if_idle() for early probe
>   failures, and order legacy platform remove so misc is deregistered
>   before metric unmap (matching ACPI teardown).
> 
> The series is bisect-safe: each commit was built with
>   make M=drivers/platform/x86/amd/hsmp
> after the previous commits, and scripts/checkpatch.pl -g <rev> was run on
> each change (0 errors / 0 warnings).
> 
> Thanks,
> Muralidhara M K
> 
> Muralidhara M K (8):
>   platform/x86/amd/hsmp: Add HSMP messages for Family 1Ah, Model 50h-5Fh
>   platform/x86/amd/hsmp: Add UAPI structures for Family 1Ah Model
>     50h-5Fh metrics table
>   platform/x86/amd/hsmp: Unify response_sz validation to an upper-bound
>     check
>   platform/x86/amd/hsmp: Source metric-table size from firmware
>   platform/x86/amd/hsmp: Add IOCTL_GET_TELEMETRY_DATA for metric table
>     reads
>   platform/x86/amd/hsmp: Clamp ioctl/send_message indices (Spectre v1)
>   platform/x86/amd/hsmp: Enable HSMP_PROTO_VER7 metric tables on the
>     ACPI driver via the IOCTL
>   platform/x86/amd/hsmp: Serialize metric table access and ACPI HSMP
>     teardown

Hi,

Could you please do a fresh iteration of this. It no longer seems to apply 
cleanly and maybe sashiko now figures we should have the locking in place
as well so we get a fresh review which hopefully does take the latest 
improvements into account.

-- 
 i.