linux-kernel - Re: [PATCH v2 8/8] x86/fpu/xstate: Restore supervisor xstates for __fpu__restore

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: <20200228165021.76pec2cdudmtpxeu@linutronix.de>
Date:   Fri, 28 Feb 2020 17:50:21 +0100
From:   Sebastian Andrzej Siewior <bigeasy@...utronix.de>
To:     Yu-cheng Yu <yu-cheng.yu@...el.com>
Cc:     Borislav Petkov <bp@...en8.de>, linux-kernel@...r.kernel.org,
        x86@...nel.org, "H. Peter Anvin" <hpa@...or.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        Tony Luck <tony.luck@...el.com>,
        Andy Lutomirski <luto@...nel.org>,
        Rik van Riel <riel@...riel.com>,
        "Ravi V. Shankar" <ravi.v.shankar@...el.com>,
        Fenghua Yu <fenghua.yu@...el.com>,
        Peter Zijlstra <peterz@...radead.org>
Subject: Re: [PATCH v2 8/8] x86/fpu/xstate: Restore supervisor xstates for
 __fpu__restore_sig()

On 2020-02-28 08:20:27 [-0800], Yu-cheng Yu wrote:
> On Fri, 2020-02-28 at 17:23 +0100, Borislav Petkov wrote:
> > On Fri, Feb 28, 2020 at 07:53:38AM -0800, Yu-cheng Yu wrote:
> > > Yes, saving only supervisor states is optimal, but doing XSAVES with a
> > > partial RFBM changes xcomp_bv.
> > 
> > ... and that means what exactly in plain english?
> 
> When XSAVES writes to an xsave buffer, xsave->header.xcomp_bv is set to
> include only saved components, effectively changing the buffer's format.

How large is this supervisor state at most? I guess saving the AVX512
state just to get the 2 bytes of the supervisor state at the right spot
is not really optimal.
But this is the performance division…

> I will include this in the comments.

If you do so, please state that the first hunk is only interested in the
supervisor-state bits and everything else is ignored.

> Yu-cheng

Sebastian