Re: Odd pci_iounmap() declaration rules..

On Sun, Sep 19, 2021 at 3:44 PM Linus Torvalds
[off-list ref] wrote:

The fix seems to be to just move that odd code from the header file to
lib/pci_iomap.c, and that should make it all JustWork(tm).

I'm not 100% happy about the end result, and in particular I think
that the new generic pci_iounmap() function for the
ARCH_WANTS_GENERIC_PCI_IOUNMAP case should do the "iounmap(p)" thing
even if ARCH_HAS_GENERIC_IOPORT_MAP wasn't true, but I tried to keep
the old rules, even if they seemed broken.

arm and arm64 build for me, as did sparc64 and alpha. At least in the
configs I tested.

And the code _does_ make a bit more sense than it used to. It still
has crazy corners, but moving the pci_iounmap() code out of the header
file should make it easier to fix up in the future.

          Linus