On Fri, Jun 23, 2017 at 06:50:41PM -0700, Palmer Dabbelt wrote: > While upstreaming the RISC-V port, it was pointed out that multiple > architectures have copied the mostly empty versions of at least one of these > functions. This defines weakly bound versions of the common functions and > removes the now obselete functions from other ports. > > This has been split out from the RISC-V submission so we can decouple all these > generic changes from our port review process. There's some discussion on an > earlier version of the patch here > > https://lkml.org/lkml/2017/6/6/998 > > but I'm afraid a lot of this is really out of my wheelhouse (and I'm pretty > slammed trying to get the RISC-V port in better shape), so I'm afraid I'm not > going to be able to perform the full cleanup suggested. Applied to pci/resource for v4.14, thanks! There's room for future improvement, e.g., using per-host callbacks, etc., but I think this is a clear improvement as-is. There's some folklore about the evils of weak symbols, but I don't know the details. It would be useful to have them codified somewhere in Documentation/ or near the __weak definition.
