[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CALCETrXYTSsxYUmxb6FmvQrSaMPJ77jqgnfzG_7yRGxLUw3_QQ@mail.gmail.com>
Date: Thu, 8 Nov 2018 10:25:24 -0800
From: Andy Lutomirski <luto@...nel.org>
To: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Cc: LKML <linux-kernel@...r.kernel.org>, X86 ML <x86@...nel.org>,
Andrew Lutomirski <luto@...nel.org>,
Paolo Bonzini <pbonzini@...hat.com>,
Radim Krcmar <rkrcmar@...hat.com>,
kvm list <kvm@...r.kernel.org>,
"Jason A. Donenfeld" <Jason@...c4.com>,
Rik van Riel <riel@...riel.com>,
Dave Hansen <dave.hansen@...ux.intel.com>
Subject: Re: [PATCH 22/23] x86/fpu: Don't restore the FPU state directly from
userland in __fpu__restore_sig()
On Wed, Nov 7, 2018 at 11:49 AM Sebastian Andrzej Siewior
<bigeasy@...utronix.de> wrote:
>
> __fpu__restore_sig() restores the CPU's FPU state directly from
> userland. If we restore registers on return to userland then we can't
> load them directly from userland because a context switch/BH could
> destroy them.
>
> Restore the FPU registers after they have been copied from userland.
> __fpregs_changes_begin() ensures that they are not modified while beeing
> worked on. TIF_NEED_FPU_LOAD is clreared we want to keep our state, not
> the saved state.
I'm conceptually okay with this change, but what happens if the
registers that are copied into the kernel are garbage? We used to
fail the restore and presumably kill the task. What happens now?
Powered by blists - more mailing lists