linux-kernel - Re: [RFC PATCH] platform/x86: Add sysfs interface for Intel TXT status

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20220309105343.GA14476@srcf.ucam.org>
Date:   Wed, 9 Mar 2022 10:53:43 +0000
From:   Matthew Garrett <mjg59@...f.ucam.org>
To:     Jonathan McDowell <noodles@...com>
Cc:     Dmitrii Okunev <xaionaro@...com>,
        Hans de Goede <hdegoede@...hat.com>,
        Mark Gross <markgross@...nel.org>,
        Qiaowei Ren <qiaowei.ren@...el.com>,
        Xiaoyan Zhang <xiaoyan.zhang@...el.com>,
        Pavel Machek <pavel@...x.de>,
        Greg Kroah-Hartman <greg@...ah.com>,
        "x86@...nel.org" <x86@...nel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "platform-driver-x86@...r.kernel.org" 
        <platform-driver-x86@...r.kernel.org>
Subject: Re: [RFC PATCH] platform/x86: Add sysfs interface for Intel TXT
 status

On Wed, Mar 09, 2022 at 10:40:03AM +0000, Jonathan McDowell wrote:

> This module provides read-only access to the Intel TXT (Trusted
> Execution Technology) status registers, allowing userspace to determine
> the status of measured boot and whether the dynamic root of trust for
> measurement (DRTM) has been fully enabled.

So there's the obvious issue that in the event that the system has been 
compromised this information is no longer trustworthy - is this expected 
to just be informative for diagnostic purposes rather than forming any 
part of security policy?

> +	  These registers provide details about the status of the platform's
> +	  measured launch and execution environment, allowing userspace to
> +	  make trust based decisions. See tboot

Mm. This makes it sound like it's expected that userspace make decisions 
based on this, which sounds like a bad plan?

> +/* Shows if TXT has been enabled */
> +static int txt_enabled_show(struct seq_file *m, void *v)
> +{
> +	/* If the BIOS has enabled TXT then the heap base will be set */

Sorry it's not that I want to say "Wait are you trusting that the BIOS 
will do the right thing here" but wait are you trusting that the BIOS 
will do the right thing here? Does setting the heap base guarantee that 
TXT was enabled (and, conversely, are there any scenarios where TXT was 
enabled and the BIOS could have cleared the heap base after a 
measurement event?)

> +/* Shows the 256 bit hash of the public key */
> +static int txt_key_show(struct seq_file *m, void *v)
> +{
> +	seq_printf(m, "%016llx%016llx%016llx%016llx\n",
> +		cpu_to_be64(*(u64 *)(txt_pub_regs + TXT_CR_PUBLIC_KEY)),
> +		cpu_to_be64(*(u64 *)(txt_pub_regs + TXT_CR_PUBLIC_KEY + 8)),
> +		cpu_to_be64(*(u64 *)(txt_pub_regs + TXT_CR_PUBLIC_KEY + 16)),
> +		cpu_to_be64(*(u64 *)(txt_pub_regs + TXT_CR_PUBLIC_KEY + 24)));

What's the expected consumer of this, and what are they expected to do 
with it?