On Fri, Mar 14, 2025 at 03:25:53PM +0100, Cédric Le Goater wrote: > > This is the 8th revision of the vfio-user client implementation. The > > vfio-user > > protocol allows for implementing (PCI) devices in another userspace process; > > SPDK is one example, which includes a virtual NVMe implementation. > > This series is going in the right direction. Thanks for all the > changes. However, I will postpone my review for the reasons below. > > VFIO initial commit in 2012 was a single file adding support for a > "vfio-pci" device. Since, the subsystem has changed a lot and it grew > to +16K line of code. Over the years, we've had numerous additions, > VFIO platform devices, s390x VFIO AP and CCW devices, migration > support with dirty tracking, support for different host IOMMU backend > devices, multifd, etc. All these additions never took the opportunity > to introduce a VFIO interface for external usage and PPC, s390x, > migration, backends subsystems happily and randomly peek and poke into > it. The subsystem is now a happy jumble of disparate functions, with > weak organization and names. The vfio-common.h header file is > good example of the result.
Right, I certainly agree! > It's time for a global code reshuffling. I'll take care of it (and > I've already started) and this means nothing new (fixes are OK) will > be merged before this is done. I hope I have made enough progress > before the start of the QEMU 10.1 cycle. There won't be any functional > changes, but there will be new files and new function names, so it's > probably a real earthquake for your series. It's had a couple of earthquakes already, very happy to bear the pain for a better end result! I will try to look out for your series and review where I can as well. thanks john
