Messages in this thread | | | Date | Tue, 02 Oct 2012 12:54:20 +0100 | From | "Jan Beulich" <> | Subject | Re: linux-next: manual merge of the xen-two tree with Linus' tree |
| |
>>> On 02.10.12 at 13:45, Stefano Stabellini <stefano.stabellini@eu.citrix.com> wrote: >> > Considering that dbgp doesn't seem to be very useful without PCI at the >> > moment, could we just turn it into: >> > >> > dom0-$(CONFIG_PCI) += dbgp.o >> > >> > ? >> >> Better not - the code is specifically not PCI-only. And I can't see >> how it would be harmful to be compiled on e.g. ARM (so the >> merge perhaps really should use XEN_DOM0 alone, without >> X86. If anything (and that may indeed be a minor oversight of >> the original patch) one might want it to depend on USB_SUPPORT, >> as without that no in-tree debug port capable driver would be >> able to load (and hence interfere with Xen's use of the debug >> port). However, as long as it builds fine with USB_SUPPORT >> undefined (which I believe it does), having it in the shape it >> is allows for out-of-tree drivers as well (as long as they make >> use of the designated interface). > > OK for PCI. > Regarding USB_SUPPORT, considering that gbgp.c calls hcd_to_bus, I think > it would make sense to make it depend on it.
As said - I'd prefer to do that only if indeed needed to get things to build without that option. Since include/usb/* doesn't reference CONFIG_USB_SUPPORT, I'm in favor of supporting at least those eventual out-of-tree drivers that do use the pre-existing data structures (and others shouldn't be calling pre-existing APIs anyway). But I wouldn't NAK a patch doing what you suggest either.
Jan
| |