On 12/9/24 8:33 AM, Krzysztof Kozlowski wrote: >> I'm thinking of using the same driver for both cases, and differentiate >> between the two by compatible and `of_device_id.data`. Thus I propose to >> have a "google,gs101-acpm-mbox" compatible for the ACPM SRAM case and in >> the future we may add a "google,gs101-mbox" compatible for the messages >> passed via the controller's data register case. > Good that you pointed it out, I was indeed wondering why this is > "acpm-mbox", not "mbox in compatible. > > This needs to be fixed - you cannot have two compatibles for the same > device. Will fix. I followed arm,mhu, which differentiates the transfer mode, data or doorbell, via compatible. For the fix I'll use "#mbox-cells" as <&phandle type channel>, where type specifies doorbel or data type. Clients will use: mboxes = <&ap2apm_mailbox DOORBELL 2>; or mboxes = <&ap2apm_mailbox DATA 3>; arm,mhu3 and fsl,mu pass the transfer mode in a similar way. >> Given this, I shall use the more generic name for the bindings, thus >> maybe "google,gs101-mbox.yaml"? But then exynos850 has the same >> controller, shouldn't we just use "samsung,exynos.yaml"? > If exynos850 has the same controller, then add it to the binding. Anyway > then use samsung,exynos850-mbox, because samsung,exynos is way too generic. Looks the same, yes, it differs by the number of how many data registers each has. But I'll stick to "google,gs101-mbox.yaml", as I can't test exynos850 and I assume we can rename the file when we'll need it.
