On Mon, Nov 14, 2022 at 11:34:06AM +0100, Thierry Reding wrote: > On Fri, Nov 11, 2022 at 11:01:58AM -0600, Rob Herring wrote: > > On Fri, Nov 11, 2022 at 5:47 AM Thierry Reding <thierry.reding@xxxxxxxxx> wrote: > > > > > > From: Thierry Reding <treding@xxxxxxxxxx> > > > > > > The #dma-address-cells and #dma-size-cells properties can be used to > > > override their non-DMA counterparts (#address-cells and #size-cells) > > > for cases where the control bus (defined by the "reg" and "ranges" > > > properties) has different addressing requirements from the DMA bus. > > > > > > The "dma-ranges" property needs to be sized depending on these DMA > > > bus properties rather than the control bus properties, so adjust the > > > check accordingly. > > > > I assume I'll be seeing a spec and schema addition too. > > Yeah, I was looking around to see where else we may need changes. I had > looked at dt-schema but couldn't find a good place to add them since > #address-cells and #size-cells seem to be mostly handled in the library > code rather than in the json-schema definitions. So if you could provide > some pointers as to how you think this should be added, that'd be great. > > I can look at writing an update to the spec, but to be frank could use > some guidance on that as well. So, I was typing up the spec changes for this and now I'm having second thoughts. The only reason why I can justify the existence of these properties is because we don't want to touch #address-cells and #size-cells. But there's really no good reason to not do that. Yes, it will be slightly painful to do this, but it's not like we can't. Given all the weird special cases that this is going to add, I'm no longer sure this is a good idea. Thierry
Attachment:
signature.asc
Description: PGP signature
