[PATCH 0/4][VTD] kvm vt-d support kernel changes

[PATCH 0/4][VTD] kvm vt-d support kernel changes

[Kay, Allen M]

Following four patches contains changes for enabling VT-d PCI
passthrough.  The patches are located at:

git.kernel.org/pub/scm/linux/kernel/git/amit/kvm.git vtd

I have incorporated most of the feedbacks from the last RFC submission.
It was tested with passthrough an E1000 NIC to a linux guest using
irqhook interrupt injection mechanism.

1) intel_iommu_move.patch: move intel-iommu.h/iova.h to include/linux.
2) intel_iommu_mods.patch: kvm modifications to intel-iommu.c.
3) kvm_generic.patch: vt-d hooks in generic KVM sources.
4) kvm_vtd.patch: vt-d specific files in KVM.

Signed-off-by: Allen M. Kay <allen.m.kay@intel.com>

Re: [PATCH 0/4][VTD] kvm vt-d support kernel changes

[Muli Ben-Yehuda]

On Mon, Jun 09, 2008 at 05:39:12PM -0700, Kay, Allen M wrote:
> Following four patches contains changes for enabling VT-d PCI
> passthrough.  The patches are located at:
> 
> git.kernel.org/pub/scm/linux/kernel/git/amit/kvm.git vtd

Allen, which userspace tree did you test this against (branch and
commit)? I am running into incompatibilities between these patches and
the current heads of both master and pvdma branches of
git.kernel.org/pub/scm/linux/kernel/git/amit/kvm-userspace.git.

Cheers,
Muli

RE: [PATCH 0/4][VTD] kvm vt-d support kernel changes

[Kay, Allen M]

Muli,

The userpsace branch is located at:

git.kernel.org/pub/scm/linux/kernel/git/amit/kvm-userspace.git vtd

I forgot to send out the userspace patch last night. :-)

Allen

>-----Original Message-----
>From: Muli Ben-Yehuda [mailto:muli@il.ibm.com] 
>Sent: Tuesday, June 10, 2008 5:53 AM
>To: Kay, Allen M
>Cc: kvm@vger.kernel.org; Amit Shah; Ben-Ami Yassour1; Avi 
>Kivity; Anthony Liguori; Chris Wright; Han, Weidong
>Subject: Re: [PATCH 0/4][VTD] kvm vt-d support kernel changes
>
>On Mon, Jun 09, 2008 at 05:39:12PM -0700, Kay, Allen M wrote:
>> Following four patches contains changes for enabling VT-d PCI
>> passthrough.  The patches are located at:
>> 
>> git.kernel.org/pub/scm/linux/kernel/git/amit/kvm.git vtd
>
>Allen, which userspace tree did you test this against (branch and
>commit)? I am running into incompatibilities between these patches and
>the current heads of both master and pvdma branches of
>git.kernel.org/pub/scm/linux/kernel/git/amit/kvm-userspace.git.
>
>Cheers,
>Muli
>

Re: [PATCH 0/4][VTD] kvm vt-d support kernel changes

[Muli Ben-Yehuda]

On Tue, Jun 10, 2008 at 12:45:30PM -0700, Kay, Allen M wrote:
> Muli,
> 
> The userpsace branch is located at:
> 
> git.kernel.org/pub/scm/linux/kernel/git/amit/kvm-userspace.git vtd
> 
> I forgot to send out the userspace patch last night. :-)

Thanks Allen. I am trying to get the latest rev to work with the irq
injection method rather than irq chip. Amit's latest code overloads
KVM_ASSIGN_PCI_PT_DEV to inform the host kernel of guest IRQ changes,
so we either need something like the attachd patch to not map the
guest multiple times, or we need to introduce
KVM_PCI_PT_DEV_IRQ_CHANGE in additino to KVM_ASSIGN_PCI_PT_DEV.

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 1b50ba3..2d61375 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -305,6 +305,14 @@ static int kvm_vm_ioctl_pci_pt_dev(struct kvm *kvm,
 			goto out_free;
 		}
 	}
+
+	if (kvm_intel_iommu_found()) {
+		r = kvm_iommu_map_guest(kvm, pci_pt_dev);
+		if (r)
+			goto out_free;
+	}
+
 	write_lock_irqsave(&kvm_pci_pt_lock, flags);
 
 	INIT_WORK(&match->pt_dev.int_work.work, kvm_pci_pt_work_fn);
@@ -1962,11 +1970,6 @@ long kvm_arch_vm_ioctl(struct file *filp,
 		r = kvm_vm_ioctl_pci_pt_dev(kvm, &pci_pt_dev);
 		if (r)
 			goto out;
-		if (kvm_intel_iommu_found()) {
-			r = kvm_iommu_map_guest(kvm, &pci_pt_dev);
-			if (r)
-				goto out;
-		}
 		break;
 	}
 	case KVM_GET_PIT: {

Cheers,
Muli

Re: [PATCH 0/4][VTD] kvm vt-d support kernel changes

[Avi Kivity]

Kay, Allen M wrote:
> Following four patches contains changes for enabling VT-d PCI
> passthrough.  The patches are located at:
>
> git.kernel.org/pub/scm/linux/kernel/git/amit/kvm.git vtd
>
>   

Please attach patches as text/plain or inline them.  It's very annoying 
to review application/octet-stream patches.

> I have incorporated most of the feedbacks from the last RFC submission.
> It was tested with passthrough an E1000 NIC to a linux guest using
> irqhook interrupt injection mechanism.
>
> 1) intel_iommu_move.patch: move intel-iommu.h/iova.h to include/linux.
> 2) intel_iommu_mods.patch: kvm modifications to intel-iommu.c.
>   

What's the upstream path for these (who's the subsystem maintainer)?

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

RE: [PATCH 0/4][VTD] kvm vt-d support kernel changes

[Han, Weidong]

Avi Kivity wrote:
> Kay, Allen M wrote:
>> Following four patches contains changes for enabling VT-d PCI
>> passthrough.  The patches are located at:
>> 
>> git.kernel.org/pub/scm/linux/kernel/git/amit/kvm.git vtd
>> 
>> 
> 
> Please attach patches as text/plain or inline them.  It's very
> annoying to review application/octet-stream patches.
> 
>> I have incorporated most of the feedbacks from the last RFC
>> submission. It was tested with passthrough an E1000 NIC to a linux
>> guest using irqhook interrupt injection mechanism.
>> 
>> 1) intel_iommu_move.patch: move intel-iommu.h/iova.h to
>> include/linux. 2) intel_iommu_mods.patch: kvm modifications to
>> intel-iommu.c. 
>> 
> 
> What's the upstream path for these (who's the subsystem maintainer)?

Now Gross Mark from Intel is the kernel VT-d maintainer. I have
contacted him and asked him to review these patches.

Randy (Weidong)