linux-kernel - Re: [PATCH 0/2] x86/fpu: Extend kernel_fpu_begin

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

Date: Thu, 2 May 2024 14:35:25 -0700
From: "Chang S. Bae" <chang.seok.bae@...el.com>
To: <linux-kernel@...r.kernel.org>
CC: <x86@...nel.org>, <platform-driver-x86@...r.kernel.org>,
	<tglx@...utronix.de>, <mingo@...hat.com>, <bp@...en8.de>,
	<dave.hansen@...ux.intel.com>, <hdegoede@...hat.com>,
	<ilpo.jarvinen@...ux.intel.com>, <tony.luck@...el.com>,
	<ashok.raj@...el.com>, <jithu.joseph@...el.com>
Subject: Re: [PATCH 0/2] x86/fpu: Extend kernel_fpu_begin_mask() for the
 In-Field Scan driver

On 5/2/2024 1:52 PM, Chang S. Bae wrote:
> 
> This brief mention may prompt questions about why the hardware must 
> refuse to run the test under these conditions. Because this is part of 
> its internal, we'd go back to folks who wrote this test and will grab 
> their write-up to provide the logic behind this requirement.

Okay, I got the write-up sooner than expected:

In-Field Scan (IFS) is a destructive test meaning that it will overwrite 
the existing state to achieve the goal of testing the logic. IFS must 
ensure that it is able to return the CPU back to the OS in the same 
architectural state as it was prior to IFS start. To do this 
architectural state (register state visible to the OS) is saved in a 
cache [*]. The cache has a limited size. When AMX is in use, the state 
of the AMX logic that needs to be saved is too large to be accommodated 
in the cache. Therefore OS must clear the AMX logic from being in use 
prior to IFS start.

[*] This cache is software-inaccessible space.

Thanks,
Chang