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;
Content-Type: text/plain;
        charset=us-ascii
Mime-Version: 1.0 (1.0)
Subject: Re: [RFC 09/10] x86/enter: Create macros to restrict/unrestrict Indirect Branch Speculation
From:   Andy Lutomirski <luto@amacapital.net>
In-Reply-To: <0575AF4FD06DD142AD198903C74E1CC87A5F0AC5@ORSMSX103.amr.corp.intel.com>
Date:   Tue, 23 Jan 2018 19:25:23 -0800
Cc:     Andy Lutomirski <luto@kernel.org>,
        Tim Chen <tim.c.chen@linux.intel.com>,
        "Woodhouse, David" <dwmw@amazon.co.uk>,
        Andi Kleen <ak@linux.intel.com>,
        Tom Lendacky <thomas.lendacky@amd.com>,
        KarimAllah Ahmed <karahmed@amazon.de>,
        LKML <linux-kernel@vger.kernel.org>,
        Andrea Arcangeli <aarcange@redhat.com>,
        Arjan van de Ven <arjan@linux.intel.com>,
        "Raj, Ashok" <ashok.raj@intel.com>,
        "Mallick, Asit K" <asit.k.mallick@intel.com>,
        Borislav Petkov <bp@suse.de>,
        "Williams, Dan J" <dan.j.williams@intel.com>,
        "Hansen, Dave" <dave.hansen@intel.com>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        "H . Peter Anvin" <hpa@zytor.com>, Ingo Molnar <mingo@redhat.com>,
        Janakarajan Natarajan <Janakarajan.Natarajan@amd.com>,
        Joerg Roedel <joro@8bytes.org>,
        "Nakajima, Jun" <jun.nakajima@intel.com>,
        Laura Abbott <labbott@redhat.com>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        =?utf-8?Q?Radim_Krcm=C3=A1r?= <rkrcmar@redhat.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        kvm list <kvm@vger.kernel.org>, X86 ML <x86@kernel.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <BA8CD56D-CB26-415B-9ABA-6E6552900916@amacapital.net>
References: <1516476182-5153-1-git-send-email-karahmed@amazon.de> <1516476182-5153-10-git-send-email-karahmed@amazon.de> <243BE571-AF73-44B3-8D17-193F9E07686A@amacapital.net> <4e01a7a9-29e4-adcc-3f53-550fb7f3d370@amd.com> <1516724457.9521.156.camel@amazon.co.uk> <fb2b5589-e14c-470e-2de0-d9da1667bace@amd.com> <20180123224956.GQ7844@tassilo.jf.intel.com> <1516749276.13558.25.camel@amazon.co.uk> <cac69284-e76f-6081-62db-adb4dc123045@linux.intel.com> <CALCETrUTjw+KHFA3a4eL+ds=HOoK_BFWZPdkeOkfcn=Oq-F2+w@mail.gmail.com> <0575AF4FD06DD142AD198903C74E1CC87A5F0AC5@ORSMSX103.amr.corp.intel.com>
To:     "Van De Ven, Arjan" <arjan.van.de.ven@intel.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


> On Jan 23, 2018, at 5:59 PM, Van De Ven, Arjan <arjan.van.de.ven@intel.com=
> wrote:
>=20
>=20
>>> It is a reasonable approach.  Let a process who needs max security
>>> opt in with disabled dumpable. It can have a flush with IBPB clear befor=
e
>>> starting to run, and have STIBP set while running.
>>>=20
>>=20
>> Do we maybe want a separate opt in?  I can easily imagine things like
>> web browsers that *don't* want to be non-dumpable but do want this
>> opt-in.
>=20
> eventually we need something better. Probably in addition.
> dumpable is used today for things that want this.
>=20
>>=20
>> Also, what's the performance hit of STIBP?
>=20
> pretty steep, but it depends on the CPU generation, for some it's cheaper t=
han others. (yes I realize this is a vague answer, but the range is really f=
rom just about zero to oh my god)
>=20
> I'm not a fan of doing this right now to be honest. We really need to not p=
iece meal some of this, and come up with a better concept of protection on a=
 higher level.
> For example, you mention web browsers, but the threat model for browsers i=
s generally internet content. For V2 to work you need to get some "evil poin=
ter" into the app from the observer and browsers usually aren't doing that.
> The most likely user would be some software-TPM-like service that has magi=
c keys.
>=20
> And for keys we want something else... we want an madvice() sort of thing t=
hat does a few things, like equivalent of mlock (so the key does not end up i=
n swap),

I'd love to see a slight variant: encrypt that page against some ephemeral k=
ey if it gets swapped.

> not having the page (but potentially the rest) end up in core dumps, and t=
he kernel making sure that if the program exits (say for segv) that the key p=
age gets zeroed before going into the free pool. Once you do that as feature=
, making the key speculation safe is not too hard (intel and arm have cpu op=
tions to mark pages for that)
>=20
>=20

How do we do that on Intel?  Make it UC?=