linux-kernel - Re: [PATCH] intel_iommu: Disable vfio and kvm interrupt assignment when unsafe

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]

Date:	Thu, 7 Feb 2013 09:53:45 -0800
From:	Andy Lutomirski <luto@...capital.net>
To:	Joerg Roedel <joro@...tes.org>
Cc:	Gleb Natapov <gleb@...hat.com>,
	LKML <linux-kernel@...r.kernel.org>, x86@...nel.org,
	"H. Peter Anvin" <hpa@...or.com>,
	Alex Williamson <alex.williamson@...hat.com>,
	Don Zickus <dzickus@...hat.com>,
	Prarit Bhargava <prarit@...hat.com>,
	David Woodhouse <dwmw2@...radead.org>
Subject: Re: [PATCH] intel_iommu: Disable vfio and kvm interrupt assignment
 when unsafe

On Thu, Feb 7, 2013 at 9:27 AM, Joerg Roedel <joro@...tes.org> wrote:
> On Thu, Feb 07, 2013 at 08:29:42AM -0800, Andy Lutomirski wrote:
>> On Thu, Feb 7, 2013 at 3:33 AM, Joerg Roedel <joro@...tes.org> wrote:
>> > On Wed, Feb 06, 2013 at 07:08:24PM -0800, Andy Lutomirski wrote:
>> >> -     if (x2apic_present)
>> >> -             WARN(1, KERN_WARNING
>> >> -                     "Failed to enable irq remapping.  You are vulnerable to irq-injection attacks.\n");
>> >> -
>> >> +     irq_remapping_is_secure = 0;
>> >>       return -1;
>> >>  }
>> >
>> > Why do you remove this warning? It seems unrelated to the rest of the
>> > patch.
>>
>> The idea is that setting irq_remapping_is_secure = 0 makes you (much
>> less) vulnerable to irq-injection attacks: you're vulnerable to
>> malicious hardware but not to attack via vfio or kvm, because those
>> paths are disabled.
>>
>> I'd have no problem leaving the warning in and letting whoever manages
>> to trigger it and get annoyed fix it.  FWIW, it's actually likely to
>> be interesting if the warning hits.
>
> Hmm, looking into the intel_irq_remapping.c version in the tip tree
> makes me wonder even more.

Is this the version I'm based on (intel_irq_remapping: Clean up x2apic
optout security warning mess), or something else?

>
> First, I wonder why the warning only hits when an x2apic is present. The
> function is not x2apic-specific and the vulnerability also exists in
> xapic mode. So that dependency can be removed.
>
> Second, I think that it should be a pr_warn instead of a full WARN. When
> IRQ remapping could not be enabled it's most likely because of the BIOS
> or the hardware. So a message in the kernel log will do and the
> backtrace provides no additional value.
>

Which warning are you referring to?  Unless I'm failing at reading
code this morning, the result of this patch has no such warning.

> Same is true for the warning in the function iommu_set_irq_remapping():
>
>        if (sts & DMA_GSTS_CFIS)
>                 WARN(1, KERN_WARNING
>                         "Compatibility-format IRQs enabled despite intr remapping;\n"
>                         "you are vulnerable to IRQ injection.\n");
>
> From what I can see this condition depends only on the hardware too. So
> a simple pr_warn() provides the same amount of information.
>

What's the general rule here?  If this warning hits, then my
understanding of the Intel VT-d spec is wrong, and I don't think that
firmware can cause it.  A buggy hypervisor could, I suppose.

--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/