On Thu, Mar 19, 2026 at 9:39 AM Jacob Pan <[email protected]> wrote: > > Hi David, > > On Mon, 16 Mar 2026 11:11:20 -0700 > David Matlack <[email protected]> wrote: > > > On Fri, Mar 13, 2026 at 8:39 AM Jacob Pan > > <[email protected]> wrote: > > > On Thu, 12 Mar 2026 23:39:45 +0000 > > > David Matlack <[email protected]> wrote: > > > > > > +config VFIO_PCI_LIVEUPDATE > > > > + bool "VFIO PCI support for Live Update (EXPERIMENTAL)" > > > > + depends on LIVEUPDATE && VFIO_PCI > > > > + help > > > > + Support for preserving devices bound to vfio-pci across > > > > a Live > > > > + Update. The eventual goal is that preserved devices can > > > > run > > > > + uninterrupted during a Live Update, including DMA to > > > > preserved > > > > + memory buffers and P2P. However there are many steps > > > > still needed to > > > > + achieve this, including: > > > > + > > > > + - Preservation of iommufd files > > > > + - Preservation of IOMMU driver state > > > In the interim, what do you think about moving forward with noiommu > > > mode without preserving IOMMU driver state? > > > > I haven't gotten a chance to look at your noiommu patches yet, but I > > do think noiommu will probably be a good way to develop the VFIO and > > PCI support for Live Update without blocking on IOMMU driver support. > > > > > I’ve tested your > > > vfio_pci_liveupdate_kexec_test with the noiommu cdev, and it works > > > as expected. > > > > Would you be able to send patches to add noiommu support to VFIO > > selftests? That would be very helpful toward allowing us to develop > > and test Live Update support with noiommu. > > It is included in the noiommu patchset; the only difference is that the > VFIO cdev name is prefixed with noiommu, which is not strictly required > and is done out of caution. > > /dev/vfio/ > |-- devices > | `-- noiommu-vfio0 > > https://lore.kernel.org/linux-iommu/[email protected]/
This one flew under my radar. I will take a look and respond there, thanks!
