| Message ID | 20230411064743.273388-1-changhuang.liang@starfivetech.com |
|---|---|
| Headers | show |
| Series | Add JH7110 DPHY PMU support | expand |
On 2023/4/12 19:29, Krzysztof Kozlowski wrote: > On 12/04/2023 11:42, Conor Dooley wrote: >> On Wed, Apr 12, 2023 at 04:51:16PM +0800, Changhuang Liang wrote: >>> >>> >>> On 2023/4/12 16:35, Krzysztof Kozlowski wrote: >>>> On 11/04/2023 08:47, Changhuang Liang wrote: >>>>> When use "starfive,jh7110-pmu-dphy" compatible, do not need the reg and >>>>> interrupts properties. >>> [...] >>>>> >>>>> description: | >>>>> StarFive JH7110 SoC includes support for multiple power domains which can be >>>>> @@ -17,6 +18,7 @@ properties: >>>>> compatible: >>>>> enum: >>>>> - starfive,jh7110-pmu >>>>> + - starfive,jh7110-pmu-dphy >>>> >>>> You do here much more than commit msg says. >>>> >>>> Isn'y DPHY a phy? Why is it in power? >>>> >>> >>> OK, I will add more description. This is a power framework used to turn on/off >>> DPHY. So it in power, not a phy. >> >> Perhaps tie it less to its role w/ the phy, and more to do with its >> location, say "jh7110-aon-pmu"? >> There's already "aon"/"sys"/"stg" stuff used in clock-controller and >> syscon compatibles etc. >> >> Krzysztof, what do you think of that? (if you remember the whole >> discussion we previously had about using those identifiers a few weeks >> ago). > > Depends whether this is the same case or not. AFAIR, for AON/SYS/STG > these were blocks with few features, not only clock controller. > > This sounds like just phy. Powering on/off phy is still a job of phy > controller... unless it is a power domain controller. > Best regards, > Krzysztof > Hi, Coner and Krzysztof, Next version I will change commit message: dt-bindings: power: Add JH7110 DPHY PMU support. Add DPHY PMU for StarFive JH7110 SoC, it can be used to turn on/off DPHY rx/tx power switch, and it don't need the reg and interrupt properties. I think this commit message will helpful for you to understand it. Best regards, Changhuang
On 2023/4/12 19:29, Krzysztof Kozlowski wrote: > On 12/04/2023 11:42, Conor Dooley wrote: >> On Wed, Apr 12, 2023 at 04:51:16PM +0800, Changhuang Liang wrote: >>> >>> >>> On 2023/4/12 16:35, Krzysztof Kozlowski wrote: >>>> On 11/04/2023 08:47, Changhuang Liang wrote: >>>>> When use "starfive,jh7110-pmu-dphy" compatible, do not need the reg and >>>>> interrupts properties. >>> [...] >>>>> >>>>> description: | >>>>> StarFive JH7110 SoC includes support for multiple power domains which can be >>>>> @@ -17,6 +18,7 @@ properties: >>>>> compatible: >>>>> enum: >>>>> - starfive,jh7110-pmu >>>>> + - starfive,jh7110-pmu-dphy >>>> >>>> You do here much more than commit msg says. >>>> >>>> Isn'y DPHY a phy? Why is it in power? >>>> >>> >>> OK, I will add more description. This is a power framework used to turn on/off >>> DPHY. So it in power, not a phy. I found something wrong with my description here, not turn on/off DPHY, is turn on/off DPHY power switch. >> >> Perhaps tie it less to its role w/ the phy, and more to do with its >> location, say "jh7110-aon-pmu"? >> There's already "aon"/"sys"/"stg" stuff used in clock-controller and >> syscon compatibles etc. >> >> Krzysztof, what do you think of that? (if you remember the whole >> discussion we previously had about using those identifiers a few weeks >> ago). > > Depends whether this is the same case or not. AFAIR, for AON/SYS/STG > these were blocks with few features, not only clock controller. > > This sounds like just phy. Powering on/off phy is still a job of phy > controller... unless it is a power domain controller. > Best regards, > Krzysztof > So, next version the compatible can be changed to "jh7110-aon-pmu"?
On Wed, Apr 12, 2023 at 01:29:57PM +0200, Krzysztof Kozlowski wrote: > On 12/04/2023 11:42, Conor Dooley wrote: > > On Wed, Apr 12, 2023 at 04:51:16PM +0800, Changhuang Liang wrote: > >> > >> > >> On 2023/4/12 16:35, Krzysztof Kozlowski wrote: > >>> On 11/04/2023 08:47, Changhuang Liang wrote: > >>>> When use "starfive,jh7110-pmu-dphy" compatible, do not need the reg and > >>>> interrupts properties. > >> [...] > >>>> > >>>> description: | > >>>> StarFive JH7110 SoC includes support for multiple power domains which can be > >>>> @@ -17,6 +18,7 @@ properties: > >>>> compatible: > >>>> enum: > >>>> - starfive,jh7110-pmu > >>>> + - starfive,jh7110-pmu-dphy > >>> > >>> You do here much more than commit msg says. > >>> > >>> Isn'y DPHY a phy? Why is it in power? > >>> > >> > >> OK, I will add more description. This is a power framework used to turn on/off > >> DPHY. So it in power, not a phy. > > > > Perhaps tie it less to its role w/ the phy, and more to do with its > > location, say "jh7110-aon-pmu"? > > There's already "aon"/"sys"/"stg" stuff used in clock-controller and > > syscon compatibles etc. > > > > Krzysztof, what do you think of that? (if you remember the whole > > discussion we previously had about using those identifiers a few weeks > > ago). > > Depends whether this is the same case or not. > AFAIR, for AON/SYS/STG > these were blocks with few features, not only clock controller. Correct, yes. In the dts, this "pmu-dphy" node is a child node of the aon syscon, so this pmu stuff would be one of the several features.
On Fri, Apr 14, 2023 at 10:20:31AM +0800, Changhuang Liang wrote: > > > On 2023/4/12 19:29, Krzysztof Kozlowski wrote: > > On 12/04/2023 11:42, Conor Dooley wrote: > >> On Wed, Apr 12, 2023 at 04:51:16PM +0800, Changhuang Liang wrote: > >>> > >>> > >>> On 2023/4/12 16:35, Krzysztof Kozlowski wrote: > >>>> On 11/04/2023 08:47, Changhuang Liang wrote: > >>>>> When use "starfive,jh7110-pmu-dphy" compatible, do not need the reg and > >>>>> interrupts properties. > >>> [...] > >>>>> > >>>>> description: | > >>>>> StarFive JH7110 SoC includes support for multiple power domains which can be > >>>>> @@ -17,6 +18,7 @@ properties: > >>>>> compatible: > >>>>> enum: > >>>>> - starfive,jh7110-pmu > >>>>> + - starfive,jh7110-pmu-dphy > >>>> > >>>> You do here much more than commit msg says. > >>>> > >>>> Isn'y DPHY a phy? Why is it in power? > >>>> > >>> > >>> OK, I will add more description. This is a power framework used to turn on/off > >>> DPHY. So it in power, not a phy. > > I found something wrong with my description here, not turn on/off DPHY, > is turn on/off DPHY power switch. > > >> > >> Perhaps tie it less to its role w/ the phy, and more to do with its > >> location, say "jh7110-aon-pmu"? > >> There's already "aon"/"sys"/"stg" stuff used in clock-controller and > >> syscon compatibles etc. > >> > >> Krzysztof, what do you think of that? (if you remember the whole > >> discussion we previously had about using those identifiers a few weeks > >> ago). > > > > Depends whether this is the same case or not. AFAIR, for AON/SYS/STG > > these were blocks with few features, not only clock controller. > > > > This sounds like just phy. Powering on/off phy is still a job of phy > > controller... unless it is a power domain controller. > > Best regards, > > Krzysztof > > > > So, next version the compatible can be changed to "jh7110-aon-pmu"? Hmm, is the dphy the only thing that's power is controlled by registers in the aon syscon? I tried looking in the "preliminary" TRM that I have, but it's not really got a proper register map so I could not tell. If there are, it'd help your case I think Changhuang Liang.
On 2023/4/18 2:55, Conor Dooley wrote: > On Fri, Apr 14, 2023 at 10:20:31AM +0800, Changhuang Liang wrote: >> >> >> On 2023/4/12 19:29, Krzysztof Kozlowski wrote: >>> On 12/04/2023 11:42, Conor Dooley wrote: >>>> On Wed, Apr 12, 2023 at 04:51:16PM +0800, Changhuang Liang wrote: >>>>> >>>>> >>>>> On 2023/4/12 16:35, Krzysztof Kozlowski wrote: >>>>>> On 11/04/2023 08:47, Changhuang Liang wrote: >>>>>>> When use "starfive,jh7110-pmu-dphy" compatible, do not need the reg and >>>>>>> interrupts properties. >>>>> [...] >>>>>>> >>>>>>> description: | >>>>>>> StarFive JH7110 SoC includes support for multiple power domains which can be >>>>>>> @@ -17,6 +18,7 @@ properties: >>>>>>> compatible: >>>>>>> enum: >>>>>>> - starfive,jh7110-pmu >>>>>>> + - starfive,jh7110-pmu-dphy >>>>>> >>>>>> You do here much more than commit msg says. >>>>>> >>>>>> Isn'y DPHY a phy? Why is it in power? >>>>>> >>>>> >>>>> OK, I will add more description. This is a power framework used to turn on/off >>>>> DPHY. So it in power, not a phy. >> >> I found something wrong with my description here, not turn on/off DPHY, >> is turn on/off DPHY power switch. >> >>>> >>>> Perhaps tie it less to its role w/ the phy, and more to do with its >>>> location, say "jh7110-aon-pmu"? >>>> There's already "aon"/"sys"/"stg" stuff used in clock-controller and >>>> syscon compatibles etc. >>>> >>>> Krzysztof, what do you think of that? (if you remember the whole >>>> discussion we previously had about using those identifiers a few weeks >>>> ago). >>> >>> Depends whether this is the same case or not. AFAIR, for AON/SYS/STG >>> these were blocks with few features, not only clock controller. >>> >>> This sounds like just phy. Powering on/off phy is still a job of phy >>> controller... unless it is a power domain controller. >>> Best regards, >>> Krzysztof >>> >> >> So, next version the compatible can be changed to "jh7110-aon-pmu"? > > Hmm, is the dphy the only thing that's power is controlled by registers > in the aon syscon? I tried looking in the "preliminary" TRM that I have, > but it's not really got a proper register map so I could not tell. > > If there are, it'd help your case I think Changhuang Liang. I made a discussion with Walker, We don't use other bit on the visionfive2 board. And I first naming by function. So I will change to "jh7110-aon-pmu" next version.