Re: [PATCH v3 04/10] nvmem: qfprom: Add support for secure reading on QDU1000/QRU1000

[Bjorn Andersson <andersson@kernel.org>]

On Mon, May 15, 2023 at 02:02:11PM +0530, Komal Bajaj wrote:
> 
> 
> On 5/12/2023 11:01 PM, Dmitry Baryshkov wrote:
> > On Fri, 12 May 2023 at 20:01, Krzysztof Kozlowski
> > <krzysztof.kozlowski@linaro.org> wrote:
> > > On 12/05/2023 14:21, Komal Bajaj wrote:
> > > > Add qfprom driver support for QDU1000/QRU1000 SOCs.
> > > > 
> > > > Signed-off-by: Komal Bajaj <quic_kbajaj@quicinc.com>
> > > > ---
> > > >   drivers/nvmem/qfprom.c | 5 +++++
> > > >   1 file changed, 5 insertions(+)
> > > > 
> > > > diff --git a/drivers/nvmem/qfprom.c b/drivers/nvmem/qfprom.c
> > > > index 20662e2d3732..12a7981a8a71 100644
> > > > --- a/drivers/nvmem/qfprom.c
> > > > +++ b/drivers/nvmem/qfprom.c
> > > > @@ -109,6 +109,10 @@ struct qfprom_soc_compatible_data {
> > > >        bool secure;
> > > >   };
> > > > 
> > > > +static const struct qfprom_soc_compatible_data qdu1000_qfprom = {
> > > > +     .secure = true
> > > > +};
> > > > +
> > > >   static const struct nvmem_keepout sc7180_qfprom_keepout[] = {
> > > >        {.start = 0x128, .end = 0x148},
> > > >        {.start = 0x220, .end = 0x228}
> > > > @@ -490,6 +494,7 @@ static int qfprom_probe(struct platform_device *pdev)
> > > > 
> > > >   static const struct of_device_id qfprom_of_match[] = {
> > > >        { .compatible = "qcom,qfprom",},
> > > > +     { .compatible = "qcom,qdu1000-qfprom", .data = &qdu1000_qfprom},
> > > >        { .compatible = "qcom,sc7180-qfprom", .data = &sc7180_qfprom},
> > > I have doubts that this is still compatible with qcom,qfprom. It uses
> > > entirely different read method. That's why generic fallbacks are bad,
> > > one more case to my growing list of awesome examples. :)
> Okay, will do that.
> > Yes, it looks like it should be 'qcom,qdu1000-qfprom",
> > "qcom,scm-qfprom". And possibly a separate driver for scm-qfprom.
> The only difference here is in read method, which can be controlled by a
> single property,
> do we really need to write a separate driver for just reading secure feature
> register.

I presume that if reads are hidden behind scm, then the most of the
driver - which deals with writing to qfprom - isn't going to be at all
applicable.

So, I actually think it would make sense to put that in a separate
qfprom-scm driver, which handles the generic fallback of
"qcom,qfprom-scm".

Regards,
Bjorn

> 
> Thanks,
> Komal
> > 
> > 
>