linux-kernel - Re: [PATCH v2] MIPS: disable MMID when not supported by the hardware

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: <aHfqG2yhp3-rVUms@alpha.franken.de>
Date: Wed, 16 Jul 2025 20:06:19 +0200
From: Thomas Bogendoerfer <tsbogend@...ha.franken.de>
To: Gregory CLEMENT <gregory.clement@...tlin.com>
Cc: Jiaxun Yang <jiaxun.yang@...goat.com>,
	Vladimir Kondratiev <vladimir.kondratiev@...ileye.com>,
	Théo Lebrun <theo.lebrun@...tlin.com>,
	Tawfik Bayouk <tawfik.bayouk@...ileye.com>,
	Thomas Petazzoni <thomas.petazzoni@...tlin.com>,
	linux-mips@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] MIPS: disable MMID when not supported by the hardware

On Thu, Jul 10, 2025 at 04:54:42PM +0200, Gregory CLEMENT wrote:
> It is possible that MMID is supported at the CPU level, but its
> integration in a SoC prevents its usage. For instance, if the
> System-level Interconnect (also known as Network on Chip) does not
> support global invalidation, then the MMID feature is not usable. The
> current implementation of MMID relies on the GINV* instructions.
> 
> This patch allows the disabling of MMID based on a device tree
> property, as this issue cannot be detected at runtime.
> 
> MMID is set up very early during the boot process, even before device
> tree data can be accessed. Therefore, when we determine whether MMID
> needs to be disabled, some MMID setup has already been performed for
> the boot CPU. Consequently, we must revert the MMID setup on the first
> CPU before disabling the feature for the subsequent CPUs that will be
> initialized later.
> 
> Signed-off-by: Gregory CLEMENT <gregory.clement@...tlin.com>
> ---
> Since the issue around MMID support cannot be detected at runtime,
> this patch now relies on a device tree property instead of a Kconfig
> option, as requested by Jiaxun and Thomas. However, as I pointed out
> in the commit log, MMID is enabled very early in the boot
> process. Therefore, I had to track all the parts of the kernel where
> MMID was set up before being able to access the device tree.
> 
> I found that MMID was only set up for the first CPU, and I also
> ensured that we can revert the MMID configuration. Indeed, the MMID
> feature is not actively used until paging is set up, which occurs
> later in the boot process.
> 
> Gregory
> ---
> Changes in v2:
> - Remove Kconfig dependencies
> - Rely on device tree property
> - Link to v1: https://lore.kernel.org/r/20250625-mmid_disable_no_ginv_on_noc-v1-1-38a3902607a7@bootlin.com
> ---
>  arch/mips/include/asm/cpu-info.h |  1 +
>  arch/mips/kernel/cpu-probe.c     | 42 ++++++++++++++++++++++++++++++++++++++--
>  arch/mips/kernel/mips-cm.c       |  5 +++++
>  3 files changed, 46 insertions(+), 2 deletions(-)

applied to mips-next.

Thomas.

-- 
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]