linux-kernel - Re: [PATCH 0/6] Introduce CET supervisor state support

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]

Message-ID: <67c5a358-0e40-4b2f-b679-33dd0dfe73fb@intel.com>
Date: Tue, 9 Jul 2024 11:17:57 +0800
From: "Yang, Weijiang" <weijiang.yang@...el.com>
To: <tglx@...utronix.de>, <dave.hansen@...el.com>, <x86@...nel.org>,
	<seanjc@...gle.com>, <pbonzini@...hat.com>, <linux-kernel@...r.kernel.org>,
	<kvm@...r.kernel.org>
CC: <peterz@...radead.org>, <chao.gao@...el.com>,
	<rick.p.edgecombe@...el.com>, <mlevitsk@...hat.com>, <john.allen@....com>
Subject: Re: [PATCH 0/6] Introduce CET supervisor state support

Hi, maintainers,
I recently did some tests for this series, the benchmark is here:
https://github.com/antonblanchard/will-it-scale/blob/master/tests/context_switch1.c

The purpose is to check what's the overall performance impact in thread/process context-
switch when CET supervisor state bit, i.e., RFBM[12], is fixed with 0 or 1 in xsaves/xrstors.

3 cases are tested:
case 1: stock v6.10-rc1 kernel.
case 2: v6.10-rc1 kernel + this patch series so that RFBM[12] == 0 for normal thread/process.
case 3: v6.10-rc1 kernel + this patch series and with patch 3(Introduce XFEATURE_MASK_KERNEL_DYNAMIC
xfeature set) reverted so that RFBM[12] == 1 for normal thread process.

Run below command 10 times in each case:

./context_switch1_processes -s 20 -t 50 -n

Trim the results by removing the top and down 10% of the data in each case, and I got below numbers:

case 1:16444675.45Set as 1

case 2:16412285.61~99.80%

case 3:16405716.19~99.76%

As you can see from the results, in case 2 with optimization based on XFEATURE_MASK_KERNEL_DYNAMIC,
the performance gain is ~0.2%.

So I'm not sure whether XFEATURE_MASK_KERNEL_DYNAMIC and related changes are worth or not
for this series.

Could you share your thoughts?

Thanks a lot!