Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Sun, 21 Jan 2018 21:17:53 +0100
From:   Borislav Petkov <bp@alien8.de>
To:     David Woodhouse <dwmw2@infradead.org>
Cc:     arjan@linux.intel.com, tglx@linutronix.de, karahmed@amazon.de,
        x86@kernel.org, linux-kernel@vger.kernel.org,
        tim.c.chen@linux.intel.com, peterz@infradead.org,
        pbonzini@redhat.com, ak@linux.intel.com,
        torvalds@linux-foundation.org, gregkh@linux-foundation.org
Subject: Re: [PATCH v2 5/8] x86/speculation: Add basic support for IBPB
Message-ID: <20180121201753.hvcrs7dvhk4vmrrv@pd.tnic>
References: <1516528149-9370-1-git-send-email-dwmw@amazon.co.uk>
 <1516528149-9370-6-git-send-email-dwmw@amazon.co.uk>
 <20180121180621.ufmc5m7nr6v4tjvc@pd.tnic>
 <1516560862.9814.47.camel@infradead.org>
 <20180121190430.ss22poeakvwzvcal@pd.tnic>
 <1516563099.9814.54.camel@infradead.org>
 <20180121195407.hbahkbmc37yrrvi4@pd.tnic>
 <1516565226.9814.63.camel@infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <1516565226.9814.63.camel@infradead.org>
User-Agent: NeoMutt/20170609 (1.8.3)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sun, Jan 21, 2018 at 08:07:06PM +0000, David Woodhouse wrote:
> That bug is the *reason* we're arguing about static_cpu_has vs.
> ALTERNATIVE. 
> 
> A conditional branch that the CPU sees can be speculated over...

Lemme see if I understand it correctly: you don't want to have any such
constructs:

	 testb $1,boot_cpu_data+50(%rip)	#, MEM[(const char *)&boot_cpu_data + 50B]
	 jnz .L707	#

which could cause any speculation.

Instead, you want to have unconditional code which is slapped in by
alternatives.

Oh, and you don't want to have that at privilege crossing sites, like
VMEXIT or so.

Am I close?

> Now, Andrew is right that in a number of cases there will be another
> serialising instruction before we ever hit a problematic indirect
> branch. But as I just said elsewhere, I'd really like the *primitives*
> to support unconditional operation.

In this particular case, WRMSR is already serializing too. So it is hard
to exploit stuff here :-)

-- 
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.