On Wed, Mar 08, 2017 at 03:37:48PM -0800, Brian Norris wrote: > I haven't quite figured out the right way to invert pci_remap_iospace(). > I guess no one supports this yet?
Jeffy Chen pointed out to me that there's a pci_unmap_iospace() as of 4.8. Looks like that should probably do the job. I'll rework this and send it out sometime. The first 2 patches are still relevant though, and the first one is a bugfix. Brian