Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 18 Mar 2022 11:49:58 +0100
From:   Peter Zijlstra <peterz@infradead.org>
To:     Josh Poimboeuf <jpoimboe@redhat.com>
Cc:     David Laight <David.Laight@aculab.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "linux-tip-commits@vger.kernel.org" 
        <linux-tip-commits@vger.kernel.org>,
        "x86@kernel.org" <x86@kernel.org>
Subject: Re: [tip: x86/core] objtool: Find unused ENDBR instructions
Message-ID: <YjRj1lmqPFdkslpr@hirez.programming.kicks-ass.net>
References: <20220308154319.763643193@infradead.org>
 <164734101940.16921.11639161864874862247.tip-bot2@tip-bot2>
 <a5fa50d9f00542de8a6ad7a3fe0c49b3@AcuMS.aculab.com>
 <20220317222254.lm2f2337jejcf3uu@treble>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220317222254.lm2f2337jejcf3uu@treble>
Precedence: bulk

On Thu, Mar 17, 2022 at 03:22:54PM -0700, Josh Poimboeuf wrote:
> On Tue, Mar 15, 2022 at 03:39:52PM +0000, David Laight wrote:
> > From: Peter Zijlstra
> > > 
> > > objtool: Find unused ENDBR instructions
> > > 
> > > Find all ENDBR instructions which are never referenced and stick them
> > > in a section such that the kernel can poison them, sealing the
> > > functions from ever being an indirect call target.
> > 
> > Thought, what happens if the only indirect call is from
> > code in a module?
> 
> Then <boom>, I guess.  Is it safe to assume in-tree modules don't need
> to do indirect calls to exported functions?  I guess we'll find out :-)

So exported functions will keep their ENDBR. Specifically, their address
is taken by the EXPORT_SYMBOL thing.

Sealing them might work, but let's not do that just now ;-)

Any unexported function discovered through kallsyms OTOH... those
deservedly will go *boom*.