Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:4 as permitted sender) client-ip=2620:137:e000::3:4;
Date:   Tue, 10 Oct 2023 13:41:19 -0700
From:   Josh Poimboeuf <jpoimboe@kernel.org>
To:     "Kaplan, David" <David.Kaplan@amd.com>
Cc:     "x86@kernel.org" <x86@kernel.org>,
        "luto@kernel.org" <luto@kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 3/3] x86/retpoline: Ensure default return thunk isn't
 used at runtime
Message-ID: <20231010204119.76i7vwecmeo6ex6d@treble>
References: <20231010171020.462211-1-david.kaplan@amd.com>
 <20231010171020.462211-4-david.kaplan@amd.com>
 <20231010193643.su6iqjniuxqqke6d@treble>
 <SN6PR12MB2702315F5C39E5354D63E68E94CDA@SN6PR12MB2702.namprd12.prod.outlook.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <SN6PR12MB2702315F5C39E5354D63E68E94CDA@SN6PR12MB2702.namprd12.prod.outlook.com>
Precedence: bulk

On Tue, Oct 10, 2023 at 08:14:33PM +0000, Kaplan, David wrote:
> [AMD Official Use Only - General]
> 
> > -----Original Message-----
> > From: Josh Poimboeuf <jpoimboe@kernel.org>
> > Sent: Tuesday, October 10, 2023 2:37 PM
> > To: Kaplan, David <David.Kaplan@amd.com>
> > Cc: x86@kernel.org; luto@kernel.org; linux-kernel@vger.kernel.org
> > Subject: Re: [PATCH 3/3] x86/retpoline: Ensure default return thunk isn't used
> > at runtime
> >
> > >   *
> > > - * This code is only used during kernel boot or module init.  All
> > > + * This code is only used during kernel boot.  All
> > >   * 'JMP __x86_return_thunk' sites are changed to something else by
> > >   * apply_returns().
> > > + *
> > > + * This thunk is turned into a ud2 to ensure it is never used at runtime.
> > > + * Alternative instructions are applied after apply_returns().
> > >   */
> > >  SYM_CODE_START(__x86_return_thunk)
> > >       UNWIND_HINT_FUNC
> > >       ANNOTATE_NOENDBR
> > > -     ANNOTATE_UNRET_SAFE
> > > -     ret
> > > +     ALTERNATIVE __stringify(ANNOTATE_UNRET_SAFE;ret),"ud2",
> > > + X86_FEATURE_RETHUNK
> >
> > If it's truly never used after boot (even for non-rethunk cases) then can we use
> > X86_FEATURE_ALWAYS?
> >
> 
> I think that could work.  There is one subtlety though I'll point out:
> 
> The use of __x86_return_thunk when X86_FEATURE_RETHUNK is set is a
> potential security issue, as it means the required return thunk is not
> being used.  The use of __x86_return_thunk when X86_FEATURE_RETHUNK is
> not set is only a performance issue, as it means there is a return
> that was not rewritten to be an inline 'ret' by apply_returns().
> 
> The ud2 was primarily intended to capture cases where there is a
> potential security hole, while it is a bit overkill just to point out
> a return that was not optimized.

Even if it's not a security hole, I'd still view it as a major BUG() as
it would directly contradict our understanding (and the comments above)
and could cause performance or other correctness issues that would
otherwise go unnoticed.

So I think an unconditional UD2 is warranted.

-- 
Josh