Re: Rethinking sigcontext's xfeatures slightly for PKRU's benefit?

From: Andy Lutomirski <luto@amacapital.net>
To: "H. Peter Anvin" <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Oleg Nesterov <oleg@redhat.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@kernel.org>, Borislav Petkov <bp@alien8.de>,
	Brian Gerst <brgerst@gmail.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	Christoph Hellwig <hch@lst.de>
Subject: Re: Rethinking sigcontext's xfeatures slightly for PKRU's benefit?
Date: Fri, 18 Dec 2015 13:02:50 -0800	[thread overview]
Message-ID: <CALCETrVsLX0vBKSRhC8WsbJvDSakPpP9+sPSPfaf22Mrm0a7yw@mail.gmail.com> (raw)
In-Reply-To: <56747360.3040909@zytor.com>

On Fri, Dec 18, 2015 at 12:58 PM, H. Peter Anvin <hpa@zytor.com> wrote:
> On 12/18/2015 12:49 PM, Andy Lutomirski wrote:
>>
>> IOW, I like my idea in which signal delivery always sets PKRU to the
>> application-requested-by-syscall values and sigreturn restores it.
>> Kinda like sigaltstack, but applies to all signals and affects PKRU
>> instead of RSP.
>>
>
> I think this is the only sensible option, with the default being all zero.
>

Or not quite all zero if we do Dave's experimental PROT_EXEC thing.

Actually, I want to introduce a set of per-mm "incompatible" bits.  By
default, they'd be zero.  We can, as needed, define bits that do
something nice but break old code.  I want one of the bits to turn
vsyscalls off entirely.  Another bit could say that the kernel is
allowed to steal a protection key for PROT_EXEC.

These bits would be read and written by prctl, but there could also be
an ELF note mechanism to initialize them on execve without a syscall.

--Andy