On Tue, 20 Sept 2022 at 16:46, Krzysztof Kozlowski
<krzysztof.kozlowski@xxxxxxxxxx> wrote:
>
> On 20/09/2022 12:48, Bhupesh Sharma wrote:
> > On Tue, 20 Sept 2022 at 15:09, Krzysztof Kozlowski
> > <krzysztof.kozlowski@xxxxxxxxxx> wrote:
> >>
> >> On 20/09/2022 10:57, Bhupesh Sharma wrote:
> >>>>> crypto: crypto@8e3a000 {
> >>>>> - compatible = "qcom,crypto-v5.1";
> >>>>> + compatible = "qcom,ipq4019-qce";
> >>>>
> >>>> There are few issues here:
> >>>> 1. Compatible is not documented.
> >>>
> >>> Its documented here:
> >>> https://lore.kernel.org/linux-arm-msm/30756e6f-952f-ccf2-b493-e515ba4f0a64@xxxxxxxxxx/
> >>>
> >>> [as mentioned in the dependency section in the cover letter :)]
> >>>
> >>>> 2. Compatible is not supported by old kernel - ABI break.
> >>
> >> You cannot fix this with dependencies/ordering.
> >>
> >>>> 3. Everything won't be bisectable...
> >>>
> >>> I think its a question of dependencies b/w the patchsets intended for
> >>> separate areas. Let me think more on how, I can resolve it in newer
> >>> versions.
> >>
> >> DTS always goes separately so this also cannot be fixed with ordering or
> >> dependencies. However if Bjorn is fine with it, it's good.
> >
> > Sure, I get your point. SInce I haven't sent out the crypto driver and
> > DMA driver subsets yet, let me stop and respin the series with the
> > dt-bindings changes clubbed with the crypto driver patches in a single
> > patchset. I can keep the DMA and dts patchsets separate and send them
> > out separately.
> >
> > I think that should help maintain the ABI and backward compatibility.
> > Please let me know if you think otherwise.
>
> I actually don't know what's in the drivers, so maybe there is no ABI
> break by kernel... but you are changing the compatibles in DTS thus any
> other project using them will be still broken.
I have sent out the crypto and dt-bindings clubbed together as one
patchset in the v7 version (see [1]).
[1]. https://lore.kernel.org/linux-arm-msm/20220920114051.1116441-1-bhupesh.sharma@xxxxxxxxxx/
Thanks,
Bhupesh
