Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-30 Thread Gleb Natapov
On Sun, Sep 29, 2013 at 09:55:59AM -0600, Alex Williamson wrote: On Sun, 2013-09-29 at 17:44 +0300, Gleb Natapov wrote: On Sun, Sep 29, 2013 at 07:52:28AM -0600, Alex Williamson wrote: On Sun, 2013-09-29 at 16:16 +0300, Gleb Natapov wrote: On Thu, Sep 12, 2013 at 03:23:15PM -0600, Alex

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-29 Thread Gleb Natapov
On Thu, Sep 12, 2013 at 03:23:15PM -0600, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any important point where that breaks down. Intel VT-d hardware may or may not support snoop control. When snoop control is available,

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-29 Thread Alex Williamson
On Sun, 2013-09-29 at 16:16 +0300, Gleb Natapov wrote: On Thu, Sep 12, 2013 at 03:23:15PM -0600, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any important point where that breaks down. Intel VT-d hardware may or may not

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-29 Thread Gleb Natapov
On Sun, Sep 29, 2013 at 07:52:28AM -0600, Alex Williamson wrote: On Sun, 2013-09-29 at 16:16 +0300, Gleb Natapov wrote: On Thu, Sep 12, 2013 at 03:23:15PM -0600, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any important

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-29 Thread Alex Williamson
On Sun, 2013-09-29 at 17:44 +0300, Gleb Natapov wrote: On Sun, Sep 29, 2013 at 07:52:28AM -0600, Alex Williamson wrote: On Sun, 2013-09-29 at 16:16 +0300, Gleb Natapov wrote: On Thu, Sep 12, 2013 at 03:23:15PM -0600, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-25 Thread Alex Williamson
On Sun, 2013-09-15 at 22:40 +1000, Alexey Kardashevskiy wrote: On 09/14/2013 02:25 AM, Alex Williamson wrote: On Fri, 2013-09-13 at 18:49 +1000, Alexey Kardashevskiy wrote: On 09/13/2013 07:23 AM, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-25 Thread Alexey Kardashevskiy
On 09/26/2013 06:19 AM, Alex Williamson wrote: On Sun, 2013-09-15 at 22:40 +1000, Alexey Kardashevskiy wrote: On 09/14/2013 02:25 AM, Alex Williamson wrote: On Fri, 2013-09-13 at 18:49 +1000, Alexey Kardashevskiy wrote: On 09/13/2013 07:23 AM, Alex Williamson wrote: So far we've succeeded at

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-15 Thread Alexey Kardashevskiy
On 09/14/2013 02:25 AM, Alex Williamson wrote: On Fri, 2013-09-13 at 18:49 +1000, Alexey Kardashevskiy wrote: On 09/13/2013 07:23 AM, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any important point where that breaks down.

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-13 Thread Alexey Kardashevskiy
On 09/13/2013 07:23 AM, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any important point where that breaks down. Intel VT-d hardware may or may not support snoop control. When snoop control is available, intel-iommu promotes

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-13 Thread Michael S. Tsirkin
On Thu, Sep 12, 2013 at 03:23:15PM -0600, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any important point where that breaks down. Intel VT-d hardware may or may not support snoop control. When snoop control is available,

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-13 Thread Alex Williamson
On Fri, 2013-09-13 at 15:39 +0300, Michael S. Tsirkin wrote: On Thu, Sep 12, 2013 at 03:23:15PM -0600, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any important point where that breaks down. Intel VT-d hardware may or may

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-13 Thread Michael S. Tsirkin
On Fri, Sep 13, 2013 at 08:13:40AM -0600, Alex Williamson wrote: On Fri, 2013-09-13 at 15:39 +0300, Michael S. Tsirkin wrote: On Thu, Sep 12, 2013 at 03:23:15PM -0600, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-13 Thread Alex Williamson
On Fri, 2013-09-13 at 17:52 +0300, Michael S. Tsirkin wrote: On Fri, Sep 13, 2013 at 08:13:40AM -0600, Alex Williamson wrote: On Fri, 2013-09-13 at 15:39 +0300, Michael S. Tsirkin wrote: On Thu, Sep 12, 2013 at 03:23:15PM -0600, Alex Williamson wrote: So far we've succeeded at making KVM

Re: [RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-13 Thread Alex Williamson
On Fri, 2013-09-13 at 18:49 +1000, Alexey Kardashevskiy wrote: On 09/13/2013 07:23 AM, Alex Williamson wrote: So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any important point where that breaks down. Intel VT-d hardware may or may not support snoop

[RFC PATCH 3/3] kvm: Add VFIO device for handling IOMMU cache coherency

2013-09-12 Thread Alex Williamson
So far we've succeeded at making KVM and VFIO mostly unaware of each other, but there's any important point where that breaks down. Intel VT-d hardware may or may not support snoop control. When snoop control is available, intel-iommu promotes No-Snoop transactions on PCIe to be cache coherent.