| Message ID | 20230822-topic-sm8550-rng-v1-1-8e10055165d1@linaro.org |
|---|---|
| State | New |
| Headers | show |
| Series | arm64: qcom: sm8550: enable RNG | expand |
On 22/08/2023 16:11, Neil Armstrong wrote: > Document SM8550 compatible for Pseudo Random Generator, > like SM8450 doesn't require clocks setup done by the secure > firmware. > > Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org> > --- Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Best regards, Krzysztof
PRNG Block on most of newer target from Qualcomm have some configuration where clock is configured by security firmware. Adding separate compatible string for each platform is overhead. We need to introduce common compatible string that can be used for all platforms with same configuration. I would suggest to use "qcom,rng-ee" for newer platform, dropping "p" also signifies it is not a Pseudo Random Number Generator. On 8/22/2023 7:58 PM, Konrad Dybcio wrote: > On 22.08.2023 16:11, Neil Armstrong wrote: >> Document SM8550 compatible for Pseudo Random Generator, >> like SM8450 doesn't require clocks setup done by the secure >> firmware. >> >> Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org> >> --- > + CC Om > > As mentioned in [1], perhaps we should rethink the compatible as > it may be a TRNG and not a PRNG? > > Konrad > > [1] https://lore.kernel.org/linux-arm-msm/d93902ee-c305-42cb-9d0d-1f0971ab3a70@quicinc.com/
On 22.08.2023 16:54, Om Prakash Singh wrote: > PRNG Block on most of newer target from Qualcomm have some configuration where clock is configured by security firmware. > > Adding separate compatible string for each platform is overhead. > > We need to introduce common compatible string that can be used for all platforms with same configuration. > > I would suggest to use "qcom,rng-ee" for newer platform, dropping "p" also signifies it is not a Pseudo Random Number Generator. Please reply inline and don't top-post. Is this what you're trying to say? 1. sort out the clock requirements for designs where Linux manages it vs where the FW does so 2. introduce a new compatible for SoCs implementing a TRNG 3. for SoCs in 2., register the TRNG as a hwrng device ? Konrad
On 8/22/2023 9:34 PM, Konrad Dybcio wrote: > On 22.08.2023 16:54, Om Prakash Singh wrote: >> PRNG Block on most of newer target from Qualcomm have some configuration where clock is configured by security firmware. >> >> Adding separate compatible string for each platform is overhead. >> >> We need to introduce common compatible string that can be used for all platforms with same configuration. >> >> I would suggest to use "qcom,rng-ee" for newer platform, dropping "p" also signifies it is not a Pseudo Random Number Generator. > Please reply inline and don't top-post. > > > Is this what you're trying to say? > > 1. sort out the clock requirements for designs where Linux manages it > vs where the FW does so > > 2. introduce a new compatible for SoCs implementing a TRNG > > 3. for SoCs in 2., register the TRNG as a hwrng device Yes to all > > > ? > > Konrad Thanks, Om
Hi, On 23/08/2023 02:10, Om Prakash Singh wrote: > > > On 8/22/2023 9:34 PM, Konrad Dybcio wrote: >> On 22.08.2023 16:54, Om Prakash Singh wrote: >>> PRNG Block on most of newer target from Qualcomm have some configuration where clock is configured by security firmware. >>> >>> Adding separate compatible string for each platform is overhead. >>> >>> We need to introduce common compatible string that can be used for all platforms with same configuration. >>> >>> I would suggest to use "qcom,rng-ee" for newer platform, dropping "p" also signifies it is not a Pseudo Random Number Generator. >> Please reply inline and don't top-post. >> >> >> Is this what you're trying to say? >> >> 1. sort out the clock requirements for designs where Linux manages it >> vs where the FW does so > >> 2. introduce a new compatible for SoCs implementing a TRNG >> >> 3. for SoCs in 2., register the TRNG as a hwrng device > > Yes to all I can send a proposal, but that means writing a new driver for this compatible in drivers/char/hw_random/ right ? Neil > >> >> >> ? >> >> Konrad > > Thanks, > Om
On 8/23/2023 1:25 PM, Neil Armstrong wrote: > Hi, > > On 23/08/2023 02:10, Om Prakash Singh wrote: >> >> >> On 8/22/2023 9:34 PM, Konrad Dybcio wrote: >>> On 22.08.2023 16:54, Om Prakash Singh wrote: >>>> PRNG Block on most of newer target from Qualcomm have some >>>> configuration where clock is configured by security firmware. >>>> >>>> Adding separate compatible string for each platform is overhead. >>>> >>>> We need to introduce common compatible string that can be used for >>>> all platforms with same configuration. >>>> >>>> I would suggest to use "qcom,rng-ee" for newer platform, dropping >>>> "p" also signifies it is not a Pseudo Random Number Generator. >>> Please reply inline and don't top-post. >>> >>> >>> Is this what you're trying to say? >>> >>> 1. sort out the clock requirements for designs where Linux manages it >>> vs where the FW does so > >>> 2. introduce a new compatible for SoCs implementing a TRNG >>> >>> 3. for SoCs in 2., register the TRNG as a hwrng device >> >> Yes to all > > I can send a proposal, but that means writing a new driver for this > compatible in drivers/char/hw_random/ right ? We can add hwrng support in same driver like drivers/crypto/hisilicon/trng/trng.c As Krzysztof is suggesting we need to have platform specific compatible string, we can go with your change. for hwrng support I will send separate patches. > > Neil > >> >>> >>> >>> ? >>> >>> Konrad >> >> Thanks, >> Om >
On 8/22/2023 7:41 PM, Neil Armstrong wrote: > Document SM8550 compatible for Pseudo Random Generator, > like SM8450 doesn't require clocks setup done by the secure > firmware. > > Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org> > --- Acked-by: Om Prakash Singh <quic_omprsing@quicinc.com> > Documentation/devicetree/bindings/crypto/qcom,prng.yaml | 8 ++++++-- > 1 file changed, 6 insertions(+), 2 deletions(-) > > diff --git a/Documentation/devicetree/bindings/crypto/qcom,prng.yaml b/Documentation/devicetree/bindings/crypto/qcom,prng.yaml > index 36b0ebd9a44b..60fc9f261b83 100644 > --- a/Documentation/devicetree/bindings/crypto/qcom,prng.yaml > +++ b/Documentation/devicetree/bindings/crypto/qcom,prng.yaml > @@ -16,7 +16,9 @@ properties: > - qcom,prng # 8916 etc. > - qcom,prng-ee # 8996 and later using EE > - items: > - - const: qcom,sm8450-prng-ee > + - enum: > + - qcom,sm8450-prng-ee > + - qcom,sm8550-prng-ee > - const: qcom,prng-ee > > reg: > @@ -39,7 +41,9 @@ allOf: > properties: > compatible: > contains: > - const: qcom,sm8450-prng-ee > + enum: > + - qcom,sm8450-prng-ee > + - qcom,sm8550-prng-ee > then: > required: > - clocks >
On 24/08/2023 01:32, Om Prakash Singh wrote: > > > On 8/23/2023 1:25 PM, Neil Armstrong wrote: >> Hi, >> >> On 23/08/2023 02:10, Om Prakash Singh wrote: >>> >>> >>> On 8/22/2023 9:34 PM, Konrad Dybcio wrote: >>>> On 22.08.2023 16:54, Om Prakash Singh wrote: >>>>> PRNG Block on most of newer target from Qualcomm have some >>>>> configuration where clock is configured by security firmware. >>>>> >>>>> Adding separate compatible string for each platform is overhead. >>>>> >>>>> We need to introduce common compatible string that can be used for >>>>> all platforms with same configuration. >>>>> >>>>> I would suggest to use "qcom,rng-ee" for newer platform, dropping >>>>> "p" also signifies it is not a Pseudo Random Number Generator. >>>> Please reply inline and don't top-post. >>>> >>>> >>>> Is this what you're trying to say? >>>> >>>> 1. sort out the clock requirements for designs where Linux manages it >>>> vs where the FW does so > >>>> 2. introduce a new compatible for SoCs implementing a TRNG >>>> >>>> 3. for SoCs in 2., register the TRNG as a hwrng device >>> >>> Yes to all >> >> I can send a proposal, but that means writing a new driver for this >> compatible in drivers/char/hw_random/ right ? > > We can add hwrng support in same driver like > drivers/crypto/hisilicon/trng/trng.c > > As Krzysztof is suggesting we need to have platform specific compatible That's independent question > string, we can go with your change. for hwrng support I will send > separate patches. Any bindings decision should be made now. We don't produce knowingly incomplete bindings just to change them later. Therefore now you need to decide whether you call it prng-ee or something else. Best regards, Krzysztof
On 24.08.2023 08:37, Krzysztof Kozlowski wrote: > On 24/08/2023 01:32, Om Prakash Singh wrote: >> >> >> On 8/23/2023 1:25 PM, Neil Armstrong wrote: >>> Hi, >>> >>> On 23/08/2023 02:10, Om Prakash Singh wrote: >>>> >>>> >>>> On 8/22/2023 9:34 PM, Konrad Dybcio wrote: >>>>> On 22.08.2023 16:54, Om Prakash Singh wrote: >>>>>> PRNG Block on most of newer target from Qualcomm have some >>>>>> configuration where clock is configured by security firmware. >>>>>> >>>>>> Adding separate compatible string for each platform is overhead. >>>>>> >>>>>> We need to introduce common compatible string that can be used for >>>>>> all platforms with same configuration. >>>>>> >>>>>> I would suggest to use "qcom,rng-ee" for newer platform, dropping >>>>>> "p" also signifies it is not a Pseudo Random Number Generator. >>>>> Please reply inline and don't top-post. >>>>> >>>>> >>>>> Is this what you're trying to say? >>>>> >>>>> 1. sort out the clock requirements for designs where Linux manages it >>>>> vs where the FW does so > >>>>> 2. introduce a new compatible for SoCs implementing a TRNG >>>>> >>>>> 3. for SoCs in 2., register the TRNG as a hwrng device >>>> >>>> Yes to all >>> >>> I can send a proposal, but that means writing a new driver for this >>> compatible in drivers/char/hw_random/ right ? >> >> We can add hwrng support in same driver like >> drivers/crypto/hisilicon/trng/trng.c >> >> As Krzysztof is suggesting we need to have platform specific compatible > > That's independent question > >> string, we can go with your change. for hwrng support I will send >> separate patches. > > Any bindings decision should be made now. We don't produce knowingly > incomplete bindings just to change them later. Therefore now you need to > decide whether you call it prng-ee or something else. Herbert already picked up the 8450 compatible last week or so. If we decide quickly, perhaps it can be reverted and substituted with the non-*P*RNG one. It would theoretically be an ABI break, but: a) it would be very very prompt b) the dts patch hasn't been merged so there are no users I'd be fine with that, not sure about the rest of you guys. Konrad
On 24/08/2023 10:40, Konrad Dybcio wrote: > On 24.08.2023 08:37, Krzysztof Kozlowski wrote: >> On 24/08/2023 01:32, Om Prakash Singh wrote: >>> >>> >>> On 8/23/2023 1:25 PM, Neil Armstrong wrote: >>>> Hi, >>>> >>>> On 23/08/2023 02:10, Om Prakash Singh wrote: >>>>> >>>>> >>>>> On 8/22/2023 9:34 PM, Konrad Dybcio wrote: >>>>>> On 22.08.2023 16:54, Om Prakash Singh wrote: >>>>>>> PRNG Block on most of newer target from Qualcomm have some >>>>>>> configuration where clock is configured by security firmware. >>>>>>> >>>>>>> Adding separate compatible string for each platform is overhead. >>>>>>> >>>>>>> We need to introduce common compatible string that can be used for >>>>>>> all platforms with same configuration. >>>>>>> >>>>>>> I would suggest to use "qcom,rng-ee" for newer platform, dropping >>>>>>> "p" also signifies it is not a Pseudo Random Number Generator. >>>>>> Please reply inline and don't top-post. >>>>>> >>>>>> >>>>>> Is this what you're trying to say? >>>>>> >>>>>> 1. sort out the clock requirements for designs where Linux manages it >>>>>> vs where the FW does so > >>>>>> 2. introduce a new compatible for SoCs implementing a TRNG >>>>>> >>>>>> 3. for SoCs in 2., register the TRNG as a hwrng device >>>>> >>>>> Yes to all >>>> >>>> I can send a proposal, but that means writing a new driver for this >>>> compatible in drivers/char/hw_random/ right ? >>> >>> We can add hwrng support in same driver like >>> drivers/crypto/hisilicon/trng/trng.c >>> >>> As Krzysztof is suggesting we need to have platform specific compatible >> >> That's independent question >> >>> string, we can go with your change. for hwrng support I will send >>> separate patches. >> >> Any bindings decision should be made now. We don't produce knowingly >> incomplete bindings just to change them later. Therefore now you need to >> decide whether you call it prng-ee or something else. > Herbert already picked up the 8450 compatible last week or so. > If we decide quickly, perhaps it can be reverted and substituted > with the non-*P*RNG one. It would theoretically be an ABI break, > but: > > a) it would be very very prompt > b) the dts patch hasn't been merged so there are no users > > I'd be fine with that, not sure about the rest of you guys. I'm fine for that aswell, this can be done quickly without the hwrng part I can quickly refresh this serie with : 1) introduce a new "qcom,trng" and move "qcom,sm8450-prng-ee" to "qcom,sm8450-trng" 2) add qcom,sm8550-prng-ee 3) add "qcom,trng" to the driver compatible list then afterwards, the hwrng part can be added in a separate serie. Neil > > Konrad
diff --git a/Documentation/devicetree/bindings/crypto/qcom,prng.yaml b/Documentation/devicetree/bindings/crypto/qcom,prng.yaml index 36b0ebd9a44b..60fc9f261b83 100644 --- a/Documentation/devicetree/bindings/crypto/qcom,prng.yaml +++ b/Documentation/devicetree/bindings/crypto/qcom,prng.yaml @@ -16,7 +16,9 @@ properties: - qcom,prng # 8916 etc. - qcom,prng-ee # 8996 and later using EE - items: - - const: qcom,sm8450-prng-ee + - enum: + - qcom,sm8450-prng-ee + - qcom,sm8550-prng-ee - const: qcom,prng-ee reg: @@ -39,7 +41,9 @@ allOf: properties: compatible: contains: - const: qcom,sm8450-prng-ee + enum: + - qcom,sm8450-prng-ee + - qcom,sm8550-prng-ee then: required: - clocks
Document SM8550 compatible for Pseudo Random Generator, like SM8450 doesn't require clocks setup done by the secure firmware. Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org> --- Documentation/devicetree/bindings/crypto/qcom,prng.yaml | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)