From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sasha Levin Subject: Re: [PATCH 1/2] kvm: Remove ability to assign a device without iommu support Date: Tue, 20 Dec 2011 11:08:48 +0200 Message-ID: <1324372128.22797.6.camel@lappy> References: <20111220030826.11829.9141.stgit@bling.home> <20111220031947.11829.66292.stgit@bling.home> <1324370959.22797.3.camel@lappy> <4EF04F54.9010500@siemens.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Cc: Alex Williamson , "avi@redhat.com" , "kvm@vger.kernel.org" , "linux-kernel@vger.kernel.org" To: Jan Kiszka Return-path: Received: from mail-ww0-f44.google.com ([74.125.82.44]:58533 "EHLO mail-ww0-f44.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751387Ab1LTJIz (ORCPT ); Tue, 20 Dec 2011 04:08:55 -0500 In-Reply-To: <4EF04F54.9010500@siemens.com> Sender: kvm-owner@vger.kernel.org List-ID: On Tue, 2011-12-20 at 10:03 +0100, Jan Kiszka wrote: > On 2011-12-20 09:49, Sasha Levin wrote: > > On Mon, 2011-12-19 at 20:19 -0700, Alex Williamson wrote: > >> This option has no users and it exposes a security hole that we > >> can allow devices to be assigned without iommu protection. Make > >> KVM_DEV_ASSIGN_ENABLE_IOMMU a mandatory option. > >> > >> Signed-off-by: Alex Williamson > >> --- > >> > >> virt/kvm/assigned-dev.c | 18 +++++++++--------- > >> 1 files changed, 9 insertions(+), 9 deletions(-) > >> > >> diff --git a/virt/kvm/assigned-dev.c b/virt/kvm/assigned-dev.c > >> index 3ad0925..a251a28 100644 > >> --- a/virt/kvm/assigned-dev.c > >> +++ b/virt/kvm/assigned-dev.c > >> @@ -487,6 +487,9 @@ static int kvm_vm_ioctl_assign_device(struct kvm *kvm, > >> struct kvm_assigned_dev_kernel *match; > >> struct pci_dev *dev; > >> > >> + if (!(assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU)) > >> + return -EINVAL; > > > > Could we just drop KVM_DEV_ASSIGN_ENABLE_IOMMU and do it by default? > > calling KVM_ASSIGN_PCI_DEVICE without that flag set it pretty > > meaningless. > > There is that thing called "backward compatibility". :) Well, Alex suggested skipping deprecation period because there are currently no users of KVM_ASSIGN_PCI_DEVICE without KVM_DEV_ASSIGN_ENABLE_IOMMU, so it should be fine to just make it the default behavior, no? We can leave KVM_DEV_ASSIGN_ENABLE_IOMMU itself so userspace won't break, but theres no reason to enforce it being set in the kernel code. -- Sasha.