On Tue, 2022-11-08 at 09:54 -0400, Jason Gunthorpe wrote: > On Tue, Nov 08, 2022 at 08:50:53AM -0500, Matthew Rosato wrote: > > > FWIW, vfio-pci via s390 is working fine so far, though I'll put it > > through more paces over the next few weeks and report if I find > > anything. > > OK great > > > As far as mdev drivers... > > > > -ccw: Sounds like Eric is already aware there is an issue and is > > investigating (I see errors as well). I -think- the problem for -ccw is that the new vfio_pin_pages requires the input addresses to be page-aligned, and while most of ours are, the first one in any given transaction may not be. We never bothered to mask off the addresses since it was handled for us, and we needed to keep the offsets anyway. By happenstance, I had some code that would do the masking ourselves (for an unrelated reason); I'll see if I can get that fit on top and if it helps matters. After coffee. Eric > > > > -ap: I see the exact same issue that Christian mentioned... I'll > > talk to Tony & Jason about it. > > A clue what is going wrong might get a quick realization on the > problem? > > I'm guessing something in the vfio side more than the access > 'pinning' > part? > > > > If I recall there was some desire from the S390 platform team to > > > start > > > building on iommufd to create some vIOMMU acceleration for S390 > > > guests, this is a necessary first step. > > > > There's probably something here for -ccw in the future, but you > > might be thinking of s390 vfio-pci e.g. to implement the in-kernel > > handling of nested mappings on s390 -- yep, work in in progress > > here, not ready for sharing yet but I have been most recently > > basing > > my work on top of the nesting series > > https://github.com/yiliu1765/iommufd/tree/iommufd-v6.0-rc3-nesting > > The relation is that if vfio-pci wants to do the above then > vfio-ccw/ap will have to run on iommufd when used in the same VM as > vfio-pci. > > So we need it to work :) > > Jason
