Messages in this thread | | | Date | Fri, 18 Jul 2008 18:05:35 +0900 | From | "Magnus Damm" <> | Subject | Re: [PATCH 02/05] resource: add resource_type() and IORESOURCE_TYPE_BITS |
| |
On Fri, Jul 18, 2008 at 5:33 PM, Ben Dooks <ben-linux@fluff.org> wrote: > On Fri, Jul 18, 2008 at 05:24:59PM +0900, Magnus Damm wrote: >> On Fri, Jul 18, 2008 at 4:56 PM, Ben Dooks <ben-linux@fluff.org> wrote: >> > You are changing a simple test to a mask and compare, is anyone going >> > to produce resources with an IORESOURCE_MEM and an IORESOURCE_IO >> > together? >> >> Actually, I'd like to replace the one-bit-per-type strategy with a >> N-bit counter. But that is not very compatible with the case you are >> pointing out. I'm not sure if that's a combination we really want to >> support though. Both IRQ and DMA doesn't make much sense to me. =) > > I'm not saying it is a bad idea, I just do not know if anyone is > currently relying on this to work...
In V1 I posted both a mega patch that went through and converted arch/ and also a patch that converted the type into a N-bit counter. In V2 I've taken more of a step-by-step approach and not converted into a N-bit counter.
I'm thinking that this patch shouldn't break anything since the bits are left exactly like before. And the code in drivers/base/platform.c seems to treat the bits as only one should be set anyway. For instance platform_device_add() seems to prioritize IORESOURCE_MEM over IORESOURCE_IO.
Or did I change the logic in drivers/base/platform without realizing it?
Cheers,
/ magnus
| |