linux-kernel - Re: [patch V2 02/52] x86/fpu: Fix copy_xstate_to

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, 17 Jun 2021 00:02:42 +0200
From:   Thomas Gleixner <tglx@...utronix.de>
To:     LKML <linux-kernel@...r.kernel.org>
Cc:     Andy Lutomirski <luto@...nel.org>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        Fenghua Yu <fenghua.yu@...el.com>,
        Tony Luck <tony.luck@...el.com>,
        Yu-cheng Yu <yu-cheng.yu@...el.com>,
        Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
        Borislav Petkov <bp@...e.de>,
        Peter Zijlstra <peterz@...radead.org>,
        Kan Liang <kan.liang@...ux.intel.com>
Subject: Re: [patch V2 02/52] x86/fpu: Fix copy_xstate_to_kernel() gap handling

On Mon, Jun 14 2021 at 17:44, Thomas Gleixner wrote:
> The gap handling in copy_xstate_to_kernel() is wrong in two aspects when
> XSAVES is in use.
>
>   1) Copying of xstate.i387.xmm_space is only copied when the SSE feature
>      bit is set. This is not correct because YMM (AVX) shares the XMM space
>      and that state must also be copied if only the YMM feature bit set
>      like already done for MXCSR.

Thinking more about it. That'd be broken in hardware. When YMM is not in
init state then SSE cannot be in init state.

Of course you can use xsave, then clear the SSE bit and XRSTOR which
blows away the SSE state. Or clear the bit in the sigframe. So copying
it over is silly as XRSTOR will ignore it anyway. If so, user space can
keep the pieces. Let me take that out.

This stuff drives me nuts.

Thanks,

        tglx