At Thu, 20 Feb 2014 00:57:58 +0900, Mark Brown wrote:
On Wed, Feb 19, 2014 at 04:21:05PM +0100, Takashi Iwai wrote:
Mark Brown wrote:
But lots of architectures do actually have these operations, I suspect looking at some of the existing users depending on PCI is enough if excessively strict (this will improve build coverage which tends to be useful even if the driver can't be run).
Yes. I guess including linux/io.h should fix the build issue.
It should for PowerPC (and ought to be done anyway since an implicit include is going to break eventually on x86 too) but it won't for architectures that don't have the function at all.
memcpy_toio() is supposed to be defined in all architecture. If the arch doesn't support it properly, it still should take from asm-generic/io.h. So, it's a good test coverage for such archs ;)
Takashi