On Thu, Apr 29, 2021 at 08:56:31AM +0200, Greg Kroah-Hartman wrote: > On Thu, Apr 29, 2021 at 08:53:15AM +0200, Christoph Hellwig wrote: > > On Wed, Apr 28, 2021 at 09:53:21AM -0300, Jason Gunthorpe wrote: > > > The Linux standard is one patch one change. It is inapporiate for me > > > to backdoor sneak revert the VFIO communities past decisions on > > > licensing inside some unrelated cleanup patch. > > > > That's not what you are doing. You are removing weird condom code > > that could never work, and remove the sneak attempt of an nvidia employee > > to create a derived work that has no legal standing. > > > > > Otherwise this patch changes nothing - what existed today continues to > > > exist, and nothing new is being allowed. > > > > No, it changes the existing exports, which is a complete no-go. > > Agreed, Jason, please do not change the existing exports. I respect both of your positions on this topic, but.. I can't be part of a licensing discussion here. This is a cleanup project from the Mellanox BU at NVIDIA to get VFIO more in line with kernel design patterns. Mellanox is fully open source for all our kernel work and has no stake in these licensing topics. As Christoph notes, it seems some other BU at NVIDIA has an interest here. I hope you'll both understand that I can't get involved in a licensing topic between the community and some other BU at NVIDIA. Since none of the past discussions on EXPORT_SYMBOL resulted in any concrete guidelines to follow, I feel the basic "don't change things" (in the pragmatic, it worked before, so don't break it sense) guideline should be applied here. Since that is not agreeable I will shrink this patch series to remove the ccw conversion that already has complex feedback and drop this patch. I'll sadly shelve the rest of the work until something changes. This will at least allow the coming Intel IDXD mdev driver to be implemented more cleanly from the start. Regards, Jason
