Re: [libvirt] [PATCH 0/4] PCI multifunction partial assignment support
On Mon, 7 Oct 2019 18:11:32 -0300 Daniel Henrique Barboza <danielhb413 gmail com> wrote:
These series tries to solve the partial assignment of multifunction hostdev PCI devices by introducing a new hostdev attribute called 'assigned'. This is how it works:
- it is a boolean value that will be efffective just for multifunction hostdev PCI devices, since there's no other occurrence for this kind of use in Libvirt. Trying to declare assign='yes|no' in any other PCI hostdev device will cause parse errors;
I think this functionality should be available to all hostdev PCI devices and not just multifunction ones. You might have several devices that are part of the same IOMMU group that you don't mind giving up from the host, but don't want to supply all of them to the guest. This would be an alternative to using ACS patch in these situations. One example is having two nvme drives in a single IOMMU group, you will be able to pass a single one to the guest rather than being forced to pass both.
On 10/10/19 9:27 PM, Abdulla Bubshait wrote:
On Mon, 7 Oct 2019 18:11:32 -0300 Daniel Henrique Barboza <danielhb413 gmail com> wrote:
These series tries to solve the partial assignment of multifunction hostdev PCI devices by introducing a new hostdev attribute called 'assigned'. This is how it works:
- it is a boolean value that will be efffective just for multifunction hostdev PCI devices, since there's no other occurrence for this kind of use in Libvirt. Trying to declare assign='yes|no' in any other PCI hostdev device will cause parse errors; I think this functionality should be available to all hostdev PCI devices and not just multifunction ones. You might have several devices that are part of the same IOMMU group that you don't mind giving up from the host, but don't want to supply all of them to the guest. This would be an alternative to using ACS patch in these situations.
One example is having two nvme drives in a single IOMMU group, you will be able to pass a single one to the guest rather than being forced to pass both.
You're right. This can be extended to all PCI hostdev devices. The conditions that I'm handling here (detaching/reattaching all the devices that belongs in the same IOMMU) isn't exclusive to PCI multifunction devices. Thanks for the input. I'll re-spin this series considering all PCI hostdev devices this time. DHB
participants (2)
-
Abdulla Bubshait -
Daniel Henrique Barboza