linux-kernel - Re: [PATCH v2] arm64/mpam: Support partial-core boot for MPAM

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <59c21340-4e3d-4c48-9d9f-c0f78a023bac@arm.com>
Date: Wed, 11 Feb 2026 12:22:14 +0000
From: Ben Horgan <ben.horgan@....com>
To: Zeng Heng <zengheng4@...wei.com>, james.morse@....com,
 fenghuay@...dia.com, reinette.chatre@...el.com
Cc: wangkefeng.wang@...wei.com, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] arm64/mpam: Support partial-core boot for MPAM

Hi Zeng,

On 2/3/26 09:31, Zeng Heng wrote:
> Some MPAM MSCs (like L2 MSC) shares the same power domain with its
> associated CPUs. Therefore, in scenarios where only partial cores power
> up, the MSCs belonging to the un-powered cores don't need and should not
> be accessed, otherwise bus-access fault would occur.
> 
> In such non-full core boot scenarios, the MSCs corresponding to offline
> CPUs should skip. If the MSC's accessibility mask doesn't contain any
> online CPU, this MSC remains uninitialized.
> 
> During initialization of class->props, skip any MSC that is not powered
> up, so that ensure the class->props member unaffected from uninitialized
> vmsc->props in mpam_enable_init_class_features() and
> mpam_enable_merge_{class|vmsc}_features().

I'm sorry but I don't think we can incorporate this kind of change into
the MPAM driver. The usable features of MPAM need to be determined
before MPAM is enabled. Any unprobed MSC may restrict the usable
features of MPAM or even render it unusable by the driver. On many
systems the feature can be inferred by symmetry (or prior knowledge of
the platform) but the driver cannot have this knowledge.

For systems where the cpus are faulty and disabled in firmware meaning
the msc is inaccessible, and not used at all, I would recommend marking
it as disabled in the acpi table. This can be done from version 3 of the
MPAM acpi by setting the MMIO size to 0.

For systems where you don't want to boot with all cpus for each msc you
will need to transitively bring online a cpu that is affine to msc.

> 
> Signed-off-by: Zeng Heng <zengheng4@...wei.com>
> ---
> Change in v2:
> - Add check for the MSC probe status in mpam_enable_merge_class_features()
> 
> v1: https://lore.kernel.org/all/20260107031336.3599175-1-zengheng4@huawei.com/
> 

Thanks,

Ben