On Tue, May 04, 2021 at 11:38:57AM +0200, Christoph Hellwig wrote: > On Mon, May 03, 2021 at 02:32:20PM -0300, Jason Gunthorpe wrote: > > 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. > > Please don't. I'll happily takes on that this is the right work, and > should not be damaged by a bad actor (Nvidia corporate that has been > sneaking weird backdoors into Linux for a while) directing someone > that now works for them through an acquisition. If everyone can have a solid agreement on licensing for vfio-mdev modules then it is fine from my perspective. IMHO that needs to be settled outside this patch series. If it has to wait while that is done, then fine. I'm not being "directed" by NVIDIA. My limitation is I can't be involved in licensing discussions, and frankly after 25 years of this I'm tired of them anyhow. This licensing topic in particular never seems to go anywhere. Half the participants want EXPORT_SYMBOL() abolished and the other half view it as an existential requirement. The whole thing is toxic to the community. > And we realy need to put Nvidia in the watchlist unfortunately as they > have caused so much damage to Linux through all their crazy backdoors. Well that seems impractical. Check the lwn statistics. NVIDIA is fairly regularly the 10th largest changeset contributor. We have > 100 people in Mellanox writing kernel patches and we employ several kernel maintainers now. If the NVIDIA/ARM purchase goes ahead it will be get even bigger. All these big amalgamations of people seem to have their unique challenges, and I'm not convinced NVIDIA is significantly more damaging to the kernel than Intel, the Android world or other places. So, let's not paint > 100 developers with such a broad brush please. I prefer the optimistive view: Mellanox's continued open source success will be inspiring. Jason
