From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andreas Hartmann <andihartmann@01019freenet.de>
Subject: AMD iommu: Difference between setting iommu=pt and not setting this
 option
Date: Mon, 02 Apr 2012 09:56:24 +0200
Message-ID: <4F795BA8.7020404@01019freenet.de>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-15
Content-Transfer-Encoding: 7bit
To: KVM <kvm@vger.kernel.org>,
	Kernel-Mailingliste <linux-kernel@vger.kernel.org>
Return-path: <linux-kernel-owner@vger.kernel.org>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: kvm.vger.kernel.org

Hello,

please, could somebody explain the difference in behaviour between
setting the boot option iommu=pt and not setting it?

I can see the following differences:

w/ iommu=pt:

[    0.832946] AMD-Vi: Enabling IOMMU at 0000:00:00.2 cap 0x40
[    0.883983] AMD-Vi: Initialized for Passthrough Mode


w/o iommu=pt:

[    0.834905] AMD-Vi: Enabling IOMMU at 0000:00:00.2 cap 0x40
[    0.894172] AMD-Vi: Lazy IO/TLB flushing enabled



In both cases, I can pass through a PCIe device but not a PCI device (->
no problem with IRQ-sharing - the PCI device has its own exclusive IRQ).

If I use iommu=pt, the passed PCIe device is broken after the VM has
been shutdown and started again.



Thank you,
kind regards,
Andreas