On Thu, 19 Mar 2020 23:18:29 +0900 <masahiroy@xxxxxxxxxx> wrote: > On Thu, Mar 19, 2020 at 4:51 PM Kunihiko Hayashi > <hayashi.kunihiko@xxxxxxxxxxxxx> wrote: > > > > The address of the extension register region in example is incorrect, > > however, this extension register region is optional > > > On which SoC is it optional? > > In your previous DT submission, every reg was, > like this: > > reg = <0x5fc10000 0x1000>, <0x5fc20000 0x800>; > > > > and you meant > > reg = <0x5fc10000 0x1000>, <0x5fc12000 0x800>; > > ? Yes. 'Optional' might not be appropriate because all SoCs has the region. > > and isn't currently > > referred from the driver, so the description of the region should be > > suppressed until referred by the driver. > > This sounds like you plan to get it back > as you extend the driver. Right, however, it isn't desiable that the description of the region is changed by extending the driver. > I checked the datasheet. This controller has more registers, > so you split the reg into small chunks, the final form will look scary: > > reg = <0x5fc10000 0x1000>, <0x5fc12000 0x800>, > <0x5fc14000 0x100>, <0x5fc15000 0x100>; > > > My question is why you did not use a single reg tuple > that covers the whole registers. > > Is any other hardware reg interleaved in between? No, there is no other hardware reg between each region. Surely it seems pointless to divide all tuples individually. I'll rewrite it to cover entire xdmac reg region. Thank you, --- Best Regards, Kunihiko Hayashi
