Re: [PATCH] x86/entry/64: randomize kernel stack offset upon syscall

From: Kees Cook <keescook@chromium.org>
To: "Reshetova, Elena" <elena.reshetova@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>, Andy Lutomirski <luto@kernel.org>,
	Theodore Ts'o <tytso@mit.edu>,
	David Laight <David.Laight@aculab.com>,
	Eric Biggers <ebiggers3@gmail.com>,
	"ebiggers@google.com" <ebiggers@google.com>,
	"herbert@gondor.apana.org.au" <herbert@gondor.apana.org.au>,
	Peter Zijlstra <peterz@infradead.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"jpoimboe@redhat.com" <jpoimboe@redhat.com>,
	"jannh@google.com" <jannh@google.com>,
	"Perla, Enrico" <enrico.perla@intel.com>,
	"mingo@redhat.com" <mingo@redhat.com>,
	"bp@alien8.de" <bp@alien8.de>,
	"tglx@linutronix.de" <tglx@linutronix.de>,
	"gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>,
	"Edgecombe, Rick P" <rick.p.edgecombe@intel.com>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Peter Zijlstra <a.p.zijlstra@chello.nl>
Subject: Re: [PATCH] x86/entry/64: randomize kernel stack offset upon syscall
Date: Tue, 30 Jul 2019 11:07:57 -0700	[thread overview]
Message-ID: <201907301102.9339710C6@keescook> (raw)
In-Reply-To: <2236FBA76BA1254E88B949DDB74E612BA4D4BFCA@IRSMSX102.ger.corp.intel.com>

On Mon, Jul 29, 2019 at 11:41:11AM +0000, Reshetova, Elena wrote:
> I want to summarize here the data (including the performance numbers)
> and reasoning for the in-stack randomization feature. I have organized
> it in a simple set of Q&A below.

Thanks for these!

> The in-stack randomization is really a very small change both code wise and
> logic wise.
> It does not affect real workloads and does not require enablement of other
> features (such as GCC plugins).
> So, I think we should really reconsider its inclusion.

I'd agree: the code is tiny and while the benefit can't point to a
specific issue, it does point to the general weakness of the stack
offset being predictable which has been a core observation for many
stack-based attacks.

If we're going to save state between syscalls (like the 4096 random
bytes pool), how about instead we just use a single per-CPU long mixed
with rdtsc saved at syscall exit. That should be a reasonable balance
between all the considerations and make it trivial for the feature to
be a boot flag without the extra page of storage, etc.

-- 
Kees Cook