On Sun, Mar 06, 2022 at 12:15:39PM +0100, Krzysztof Kozlowski wrote: > On 06/03/2022 11:41, Frank Wunderlich wrote: > >> Gesendet: Sonntag, 06. März 2022 um 11:27 Uhr > >> Von: "Krzysztof Kozlowski" <krzysztof.kozlowski@xxxxxxxxxxxxx> > >>> add compatibles used together with generic-ahci > >>> - marvell,berlin2-ahci > >> > >> This is fine, just mention it in commit msg. > >> > >>> - qcom,apq8064-ahci > >>> - qcom,ipq806x-ahci > >> > >> These you need to consult with qcom-sata.txt. This could be a following > >> commit which will integrate qcom-sata.txt and remove it. > > > > this depends on Robs opinion > > Then maybe precise the question for Rob... I would leave qcom separate, but the warnings should be fixed. For that you need a custom 'select' schema that lists everything but 'generic-ahci'. Adding the berlin compatible looks right. > >> Either you have > >> binding document for all devices or you create a common part, like for UFS: > >> https://lore.kernel.org/linux-devicetree/20220222145854.358646-1-krzysztof.kozlowski@xxxxxxxxxxxxx/ > >> https://github.com/krzk/linux/commits/n/dt-bindings-ufs-v2 > >> > >> The choice depends more or less on complexity of bindings, IOW, how big > >> and complicated bindings would be if you combine everything to one YAML. > >> > >> In the case of UFS, the devices differ - by clocks, resets, phys and > >> sometimes supplies. Therefore it easier to have one common shared part > >> and several device bindings. > >> > >> AHCI looks more consistent - except that Qualcomm - so maybe better to > >> have one document. > >> > >>> increase reg-count to 2 (used in omap5-l4.dtsi) > >>> increase clock-count to 5 (used in qcom-apq8064.dtsi) > >> > >> This would need allOf+if. > > > > if i get ok from rob i add only the berlin-compatible and skip the > > qcom+reg/clock-change in the first applied version. Adding the > > allOf/if (and making it right) will only delay the sata-binding/dts-change. > > I don't get what is the problem with delaying this patch for the time > needed to make the bindings correct? Especially that alternative is to > add bindings document which soon will need to be modified, e.g. split > into common part. Is there a particular hurry with these bindings > conversion? Qcom doesn't use sata-port nodes, so I don't think there is anything to share. And if it did, that's already in sata-common.yaml. Rob
