Message ID | 20170922020429.20941-1-kalyan.kinthada@alliedtelesis.co.nz |
---|---|
State | Superseded |
Headers | show |
Series | dt-bindings: i2c: Add armada-38x i2c binding | expand |
Hi Kalyan, On ven., sept. 22 2017, Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> wrote: > All armada-38x variants(380, 385, 388) SoCs have an issue > in i2c controller which violates the i2c repeated start timing. Did you had an issue with one of the SoC ? If yes what was your setting ? On my side I never managed to have this sort of issue on Armada 38x. > > This errata is fixed in the i2c-mv64xxx driver but enabled > only for devices with compatible string "marvell,mv78230-i2c". In this case why not just adding the "marvell,mv78230-i2c" as compatible string in the appropriate device tree files ? Gregory > > This patch introduces a new compatible string > "marvell,armada-38x-i2c" for the i2c controller on armada-38x SoCs > so that the workaound is enabled. > > Signed-off-by: Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> > --- > Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt > index 5c30026921..fc100b6bdb 100644 > --- a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt > +++ b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt > @@ -14,6 +14,7 @@ Required properties : > very rare, initial version of the SoC which > had broken offload support. Linux > auto-detects this and sets it appropriately. > + - "marvell,armada-38x-i2c" > - interrupts : The interrupt number > > Optional properties : > -- > 2.14.1 >
Hi Gregory, I got this information from Armada-38x functional errata document. I can add the "marvell,mv78230-i2c" compatible string to the appropriate device tree files but the i2c-mv64xxx driver enables an additional feature (offload i2c transactions) based on the compatible string "marvell,mv78230-i2c". I am not sure if this feature (offload i2c transactions) should be enabled for Armada-38x devices. That is the reason I felt for the need of a new compatible string specifically for Armada-38x SoCs. Thanks, Kalyan
Hi Kalyan, Please try avoid top-posting. On dim., sept. 24 2017, Kalyan Kinthada <Kalyan.Kinthada@alliedtelesis.co.nz> wrote: > Hi Gregory, > > I got this information from Armada-38x functional errata document. OK but in any case just adding a new compatible was not enough you have to update the driver in the same time, however for this case we won't need it, see below. > > I can add the "marvell,mv78230-i2c" compatible string to the appropriate device tree files > but the i2c-mv64xxx driver enables an additional feature (offload i2c transactions) > based on the compatible string "marvell,mv78230-i2c". > > I am not sure if this feature (offload i2c transactions) should be enabled for Armada-38x devices. > That is the reason I felt for the need of a new compatible string > specifically for Armada-38x SoCs. Indeed the Armada-38x SoCs does not support hardware offloading (at least according the datasheet). But it happens that in the earlier version of the Armada XP the hardware offloading was buggy, so we introduced a compatible for this case: marvell,mv78230-a0-i2c. This compatible enable the errata fix but not the offloading feature. That means that it is exactly the compatible you need for Armada 38x (and Armada 39x and 375 I think). Gregory > > Thanks, > Kalyan > ________________________________________ > From: Gregory CLEMENT <gregory.clement@free-electrons.com> > Sent: Friday, 22 September 2017 7:30 p.m. > To: Kalyan Kinthada > Cc: devicetree@vger.kernel.org; wsa@the-dreams.de; robh+dt@kernel.org; > pawel.moll@arm.com; mark.rutland@arm.com; > ijc+devicetree@hellion.org.uk; galak@codeaurora.org; > linux-i2c@vger.kernel.org; linux-kernel@vger.kernel.org; Chris Packham > Subject: Re: [PATCH] dt-bindings: i2c: Add armada-38x i2c binding > > Hi Kalyan, > > On ven., sept. 22 2017, Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> wrote: > >> All armada-38x variants(380, 385, 388) SoCs have an issue >> in i2c controller which violates the i2c repeated start timing. > > > Did you had an issue with one of the SoC ? > If yes what was your setting ? > On my side I never managed to have this sort of issue on Armada 38x. > >> >> This errata is fixed in the i2c-mv64xxx driver but enabled >> only for devices with compatible string "marvell,mv78230-i2c". > > In this case why not just adding the "marvell,mv78230-i2c" as compatible > string in the appropriate device tree files ? > > Gregory > >> >> This patch introduces a new compatible string >> "marvell,armada-38x-i2c" for the i2c controller on armada-38x SoCs >> so that the workaound is enabled. > > >> >> Signed-off-by: Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> >> --- >> Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt | 1 + >> 1 file changed, 1 insertion(+) >> >> diff --git a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >> index 5c30026921..fc100b6bdb 100644 >> --- a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >> +++ b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >> @@ -14,6 +14,7 @@ Required properties : >> very rare, initial version of the SoC which >> had broken offload support. Linux >> auto-detects this and sets it appropriately. >> + - "marvell,armada-38x-i2c" >> - interrupts : The interrupt number >> >> Optional properties : >> -- >> 2.14.1 >> > > -- > Gregory Clement, Free Electrons > Kernel, drivers, real-time and embedded Linux > development, consulting, training and support. > http://free-electrons.com
On 27/09/17 00:55, Gregory CLEMENT wrote: > Hi Kalyan, > > Please try avoid top-posting. I am sorry. Fixed it now. > > On dim., sept. 24 2017, Kalyan Kinthada <Kalyan.Kinthada@alliedtelesis.co.nz> wrote: > >> Hi Gregory, >> >> I got this information from Armada-38x functional errata document. > OK but in any case just adding a new compatible was not enough you have > to update the driver in the same time, however for this case we won't > need it, see below. > >> I can add the "marvell,mv78230-i2c" compatible string to the appropriate device tree files >> but the i2c-mv64xxx driver enables an additional feature (offload i2c transactions) >> based on the compatible string "marvell,mv78230-i2c". >> >> I am not sure if this feature (offload i2c transactions) should be enabled for Armada-38x devices. >> That is the reason I felt for the need of a new compatible string >> specifically for Armada-38x SoCs. > Indeed the Armada-38x SoCs does not support hardware offloading (at > least according the datasheet). But it happens that in the earlier > version of the Armada XP the hardware offloading was buggy, so we > introduced a compatible for this case: marvell,mv78230-a0-i2c. This > compatible enable the errata fix but not the offloading feature. That > means that it is exactly the compatible you need for Armada 38x (and > Armada 39x and 375 I think). Ok. I will add "marvell,mv78230-a0-i2c" to the Armada-38x device tree file. I will send a new patch with these changes. Thanks for your time. Regards, Kalyan > > Gregory > >> Thanks, >> Kalyan >> ________________________________________ >> From: Gregory CLEMENT <gregory.clement@free-electrons.com> >> Sent: Friday, 22 September 2017 7:30 p.m. >> To: Kalyan Kinthada >> Cc: devicetree@vger.kernel.org; wsa@the-dreams.de; robh+dt@kernel.org; >> pawel.moll@arm.com; mark.rutland@arm.com; >> ijc+devicetree@hellion.org.uk; galak@codeaurora.org; >> linux-i2c@vger.kernel.org; linux-kernel@vger.kernel.org; Chris Packham >> Subject: Re: [PATCH] dt-bindings: i2c: Add armada-38x i2c binding >> >> Hi Kalyan, >> >> On ven., sept. 22 2017, Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> wrote: >> >>> All armada-38x variants(380, 385, 388) SoCs have an issue >>> in i2c controller which violates the i2c repeated start timing. >> >> Did you had an issue with one of the SoC ? >> If yes what was your setting ? >> On my side I never managed to have this sort of issue on Armada 38x. >> >>> This errata is fixed in the i2c-mv64xxx driver but enabled >>> only for devices with compatible string "marvell,mv78230-i2c". >> In this case why not just adding the "marvell,mv78230-i2c" as compatible >> string in the appropriate device tree files ? >> >> Gregory >> >>> This patch introduces a new compatible string >>> "marvell,armada-38x-i2c" for the i2c controller on armada-38x SoCs >>> so that the workaound is enabled. >> >>> Signed-off-by: Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> >>> --- >>> Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt | 1 + >>> 1 file changed, 1 insertion(+) >>> >>> diff --git a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >>> index 5c30026921..fc100b6bdb 100644 >>> --- a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >>> +++ b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >>> @@ -14,6 +14,7 @@ Required properties : >>> very rare, initial version of the SoC which >>> had broken offload support. Linux >>> auto-detects this and sets it appropriately. >>> + - "marvell,armada-38x-i2c" >>> - interrupts : The interrupt number >>> >>> Optional properties : >>> -- >>> 2.14.1 >>> >> -- >> Gregory Clement, Free Electrons >> Kernel, drivers, real-time and embedded Linux >> development, consulting, training and support. >> http://free-electrons.com
Hi Gregory, On 27/09/17 00:56, Gregory CLEMENT wrote: > Hi Kalyan, > > Please try avoid top-posting. > > On dim., sept. 24 2017, Kalyan Kinthada <Kalyan.Kinthada@alliedtelesis.co.nz> wrote: > >> Hi Gregory, >> >> I got this information from Armada-38x functional errata document. > > OK but in any case just adding a new compatible was not enough you have > to update the driver in the same time, however for this case we won't > need it, see below. > >> >> I can add the "marvell,mv78230-i2c" compatible string to the appropriate device tree files >> but the i2c-mv64xxx driver enables an additional feature (offload i2c transactions) >> based on the compatible string "marvell,mv78230-i2c". >> >> I am not sure if this feature (offload i2c transactions) should be enabled for Armada-38x devices. >> That is the reason I felt for the need of a new compatible string >> specifically for Armada-38x SoCs. > > Indeed the Armada-38x SoCs does not support hardware offloading (at > least according the datasheet). But it happens that in the earlier > version of the Armada XP the hardware offloading was buggy, so we > introduced a compatible for this case: marvell,mv78230-a0-i2c. This > compatible enable the errata fix but not the offloading feature. That > means that it is exactly the compatible you need for Armada 38x (and > Armada 39x and 375 I think). The "mv78230-a0-i2c" dt-binding has the following note Note: Only use "marvell,mv78230-a0-i2c" for a very rare, initial version of the SoC which had broken offload support. Linux auto-detects this and sets it appropriately. If we are going to re-use this binding for armada-38x we should probably remove this note. Personally my preference would be an armada-38x compatible string (or 370 if that's the common base of these SoCs). But of course we'll go with whatever your preference is as maintainer. > > Gregory > >> >> Thanks, >> Kalyan >> ________________________________________ >> From: Gregory CLEMENT <gregory.clement@free-electrons.com> >> Sent: Friday, 22 September 2017 7:30 p.m. >> To: Kalyan Kinthada >> Cc: devicetree@vger.kernel.org; wsa@the-dreams.de; robh+dt@kernel.org; >> pawel.moll@arm.com; mark.rutland@arm.com; >> ijc+devicetree@hellion.org.uk; galak@codeaurora.org; >> linux-i2c@vger.kernel.org; linux-kernel@vger.kernel.org; Chris Packham >> Subject: Re: [PATCH] dt-bindings: i2c: Add armada-38x i2c binding >> >> Hi Kalyan, >> >> On ven., sept. 22 2017, Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> wrote: >> >>> All armada-38x variants(380, 385, 388) SoCs have an issue >>> in i2c controller which violates the i2c repeated start timing. >> >> >> Did you had an issue with one of the SoC ? >> If yes what was your setting ? >> On my side I never managed to have this sort of issue on Armada 38x. >> >>> >>> This errata is fixed in the i2c-mv64xxx driver but enabled >>> only for devices with compatible string "marvell,mv78230-i2c". >> >> In this case why not just adding the "marvell,mv78230-i2c" as compatible >> string in the appropriate device tree files ? >> >> Gregory >> >>> >>> This patch introduces a new compatible string >>> "marvell,armada-38x-i2c" for the i2c controller on armada-38x SoCs >>> so that the workaound is enabled. >> >> >>> >>> Signed-off-by: Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> >>> --- >>> Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt | 1 + >>> 1 file changed, 1 insertion(+) >>> >>> diff --git a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >>> index 5c30026921..fc100b6bdb 100644 >>> --- a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >>> +++ b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >>> @@ -14,6 +14,7 @@ Required properties : >>> very rare, initial version of the SoC which >>> had broken offload support. Linux >>> auto-detects this and sets it appropriately. >>> + - "marvell,armada-38x-i2c" >>> - interrupts : The interrupt number >>> >>> Optional properties : >>> -- >>> 2.14.1 >>> >> >> -- >> Gregory Clement, Free Electrons >> Kernel, drivers, real-time and embedded Linux >> development, consulting, training and support. >> http://free-electrons.com >
Hi Chris, On mer., sept. 27 2017, Chris Packham <Chris.Packham@alliedtelesis.co.nz> wrote: > Hi Gregory, > > On 27/09/17 00:56, Gregory CLEMENT wrote: >> Hi Kalyan, >> >> Please try avoid top-posting. >> >> On dim., sept. 24 2017, Kalyan Kinthada <Kalyan.Kinthada@alliedtelesis.co.nz> wrote: >> >>> Hi Gregory, >>> >>> I got this information from Armada-38x functional errata document. >> >> OK but in any case just adding a new compatible was not enough you have >> to update the driver in the same time, however for this case we won't >> need it, see below. >> >>> >>> I can add the "marvell,mv78230-i2c" compatible string to the appropriate device tree files >>> but the i2c-mv64xxx driver enables an additional feature (offload i2c transactions) >>> based on the compatible string "marvell,mv78230-i2c". >>> >>> I am not sure if this feature (offload i2c transactions) should be enabled for Armada-38x devices. >>> That is the reason I felt for the need of a new compatible string >>> specifically for Armada-38x SoCs. >> >> Indeed the Armada-38x SoCs does not support hardware offloading (at >> least according the datasheet). But it happens that in the earlier >> version of the Armada XP the hardware offloading was buggy, so we >> introduced a compatible for this case: marvell,mv78230-a0-i2c. This >> compatible enable the errata fix but not the offloading feature. That >> means that it is exactly the compatible you need for Armada 38x (and >> Armada 39x and 375 I think). > > The "mv78230-a0-i2c" dt-binding has the following note > > Note: Only use "marvell,mv78230-a0-i2c" for a > very rare, initial version of the SoC which > had broken offload support. Linux > auto-detects this and sets it appropriately. > > If we are going to re-use this binding for armada-38x we should probably > remove this note. Personally my preference would be an armada-38x Updating the documentation is the right thing to do yes. > compatible string (or 370 if that's the common base of these SoCs). But > of course we'll go with whatever your preference is as maintainer. If the IP is compatible then there is no reason to add a new one, else we will end with a compatible for each SoC and the compatible property will just loose its meaning. But an other reason to reuse "marvell,mv78230-a0-i2c" is that thanks to this you will benefit to this even with the old kernel by just updating the device tree, else you will have to update both the kernel and the dtb. Gregory > >> >> Gregory >> >>> >>> Thanks, >>> Kalyan >>> ________________________________________ >>> From: Gregory CLEMENT <gregory.clement@free-electrons.com> >>> Sent: Friday, 22 September 2017 7:30 p.m. >>> To: Kalyan Kinthada >>> Cc: devicetree@vger.kernel.org; wsa@the-dreams.de; robh+dt@kernel.org; >>> pawel.moll@arm.com; mark.rutland@arm.com; >>> ijc+devicetree@hellion.org.uk; galak@codeaurora.org; >>> linux-i2c@vger.kernel.org; linux-kernel@vger.kernel.org; Chris Packham >>> Subject: Re: [PATCH] dt-bindings: i2c: Add armada-38x i2c binding >>> >>> Hi Kalyan, >>> >>> On ven., sept. 22 2017, Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> wrote: >>> >>>> All armada-38x variants(380, 385, 388) SoCs have an issue >>>> in i2c controller which violates the i2c repeated start timing. >>> >>> >>> Did you had an issue with one of the SoC ? >>> If yes what was your setting ? >>> On my side I never managed to have this sort of issue on Armada 38x. >>> >>>> >>>> This errata is fixed in the i2c-mv64xxx driver but enabled >>>> only for devices with compatible string "marvell,mv78230-i2c". >>> >>> In this case why not just adding the "marvell,mv78230-i2c" as compatible >>> string in the appropriate device tree files ? >>> >>> Gregory >>> >>>> >>>> This patch introduces a new compatible string >>>> "marvell,armada-38x-i2c" for the i2c controller on armada-38x SoCs >>>> so that the workaound is enabled. >>> >>> >>>> >>>> Signed-off-by: Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> >>>> --- >>>> Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt | 1 + >>>> 1 file changed, 1 insertion(+) >>>> >>>> diff --git a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >>>> index 5c30026921..fc100b6bdb 100644 >>>> --- a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >>>> +++ b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt >>>> @@ -14,6 +14,7 @@ Required properties : >>>> very rare, initial version of the SoC which >>>> had broken offload support. Linux >>>> auto-detects this and sets it appropriately. >>>> + - "marvell,armada-38x-i2c" >>>> - interrupts : The interrupt number >>>> >>>> Optional properties : >>>> -- >>>> 2.14.1 >>>> >>> >>> -- >>> Gregory Clement, Free Electrons >>> Kernel, drivers, real-time and embedded Linux >>> development, consulting, training and support. >>> http://free-electrons.com >> >
On Wed, Sep 27, 2017 at 09:33:00AM +0200, Gregory CLEMENT wrote: > Hi Chris, > > On mer., sept. 27 2017, Chris Packham <Chris.Packham@alliedtelesis.co.nz> wrote: > > > Hi Gregory, > > > > On 27/09/17 00:56, Gregory CLEMENT wrote: > >> Hi Kalyan, > >> > >> Please try avoid top-posting. > >> > >> On dim., sept. 24 2017, Kalyan Kinthada <Kalyan.Kinthada@alliedtelesis.co.nz> wrote: > >> > >>> Hi Gregory, > >>> > >>> I got this information from Armada-38x functional errata document. > >> > >> OK but in any case just adding a new compatible was not enough you have > >> to update the driver in the same time, however for this case we won't > >> need it, see below. > >> > >>> > >>> I can add the "marvell,mv78230-i2c" compatible string to the appropriate device tree files > >>> but the i2c-mv64xxx driver enables an additional feature (offload i2c transactions) > >>> based on the compatible string "marvell,mv78230-i2c". > >>> > >>> I am not sure if this feature (offload i2c transactions) should be enabled for Armada-38x devices. > >>> That is the reason I felt for the need of a new compatible string > >>> specifically for Armada-38x SoCs. > >> > >> Indeed the Armada-38x SoCs does not support hardware offloading (at > >> least according the datasheet). But it happens that in the earlier > >> version of the Armada XP the hardware offloading was buggy, so we > >> introduced a compatible for this case: marvell,mv78230-a0-i2c. This > >> compatible enable the errata fix but not the offloading feature. That > >> means that it is exactly the compatible you need for Armada 38x (and > >> Armada 39x and 375 I think). > > > > The "mv78230-a0-i2c" dt-binding has the following note > > > > Note: Only use "marvell,mv78230-a0-i2c" for a > > very rare, initial version of the SoC which > > had broken offload support. Linux > > auto-detects this and sets it appropriately. > > > > If we are going to re-use this binding for armada-38x we should probably > > remove this note. Personally my preference would be an armada-38x > > Updating the documentation is the right thing to do yes. > > > > compatible string (or 370 if that's the common base of these SoCs). But > > of course we'll go with whatever your preference is as maintainer. > > If the IP is compatible then there is no reason to add a new one, else > we will end with a compatible for each SoC and the compatible property > will just loose its meaning. If you all had added compatibles for each SoC in the first place, then we wouldn't be having this dicussion. Rob
Hi Rob, On mar., oct. 03 2017, Rob Herring <robh@kernel.org> wrote: > On Wed, Sep 27, 2017 at 09:33:00AM +0200, Gregory CLEMENT wrote: >> Hi Chris, >> >> On mer., sept. 27 2017, Chris Packham <Chris.Packham@alliedtelesis.co.nz> wrote: >> >> > Hi Gregory, >> > >> > On 27/09/17 00:56, Gregory CLEMENT wrote: >> >> Hi Kalyan, >> >> >> >> Please try avoid top-posting. >> >> >> >> On dim., sept. 24 2017, Kalyan Kinthada <Kalyan.Kinthada@alliedtelesis.co.nz> wrote: >> >> >> >>> Hi Gregory, >> >>> >> >>> I got this information from Armada-38x functional errata document. >> >> >> >> OK but in any case just adding a new compatible was not enough you have >> >> to update the driver in the same time, however for this case we won't >> >> need it, see below. >> >> >> >>> >> >>> I can add the "marvell,mv78230-i2c" compatible string to the appropriate device tree files >> >>> but the i2c-mv64xxx driver enables an additional feature (offload i2c transactions) >> >>> based on the compatible string "marvell,mv78230-i2c". >> >>> >> >>> I am not sure if this feature (offload i2c transactions) should be enabled for Armada-38x devices. >> >>> That is the reason I felt for the need of a new compatible string >> >>> specifically for Armada-38x SoCs. >> >> >> >> Indeed the Armada-38x SoCs does not support hardware offloading (at >> >> least according the datasheet). But it happens that in the earlier >> >> version of the Armada XP the hardware offloading was buggy, so we >> >> introduced a compatible for this case: marvell,mv78230-a0-i2c. This >> >> compatible enable the errata fix but not the offloading feature. That >> >> means that it is exactly the compatible you need for Armada 38x (and >> >> Armada 39x and 375 I think). >> > >> > The "mv78230-a0-i2c" dt-binding has the following note >> > >> > Note: Only use "marvell,mv78230-a0-i2c" for a >> > very rare, initial version of the SoC which >> > had broken offload support. Linux >> > auto-detects this and sets it appropriately. >> > >> > If we are going to re-use this binding for armada-38x we should probably >> > remove this note. Personally my preference would be an armada-38x >> >> Updating the documentation is the right thing to do yes. >> >> >> > compatible string (or 370 if that's the common base of these SoCs). But >> > of course we'll go with whatever your preference is as maintainer. >> >> If the IP is compatible then there is no reason to add a new one, else >> we will end with a compatible for each SoC and the compatible property >> will just loose its meaning. > > If you all had added compatibles for each SoC in the first place, then > we wouldn't be having this dicussion. Really?? For Armada 38x we have already 6 flavor 381, 382, 383, 385, 388. Then we have 9 SoC families on mvebu: orion5x, kirkwood, dove, Armada 370, Armada XP, Armada 375, Armada 39x, Armada 7K, Armada 8K. Of course in each family we have several flavors. But then Allwiner also use this driver in 8 SoC families: sun4i, sun5i, sun6i, sun7i, sun8i, sun9i, sunxi, sun50i. Here again each family have their own flavor. So you are just suggesting to blot the i2c driver with more than 50 compatible string! That also mean updating the i2c driver each time a new SoC flavor appear, so more or less for each release. Really this just don't scale. Gregory > > Rob
On Wed, Oct 4, 2017 at 3:33 AM, Gregory CLEMENT <gregory.clement@free-electrons.com> wrote: > Hi Rob, > > On mar., oct. 03 2017, Rob Herring <robh@kernel.org> wrote: > >> On Wed, Sep 27, 2017 at 09:33:00AM +0200, Gregory CLEMENT wrote: >>> Hi Chris, >>> >>> On mer., sept. 27 2017, Chris Packham <Chris.Packham@alliedtelesis.co.nz> wrote: >>> >>> > Hi Gregory, >>> > >>> > On 27/09/17 00:56, Gregory CLEMENT wrote: >>> >> Hi Kalyan, >>> >> >>> >> Please try avoid top-posting. >>> >> >>> >> On dim., sept. 24 2017, Kalyan Kinthada <Kalyan.Kinthada@alliedtelesis.co.nz> wrote: >>> >> >>> >>> Hi Gregory, >>> >>> >>> >>> I got this information from Armada-38x functional errata document. >>> >> >>> >> OK but in any case just adding a new compatible was not enough you have >>> >> to update the driver in the same time, however for this case we won't >>> >> need it, see below. >>> >> >>> >>> >>> >>> I can add the "marvell,mv78230-i2c" compatible string to the appropriate device tree files >>> >>> but the i2c-mv64xxx driver enables an additional feature (offload i2c transactions) >>> >>> based on the compatible string "marvell,mv78230-i2c". >>> >>> >>> >>> I am not sure if this feature (offload i2c transactions) should be enabled for Armada-38x devices. >>> >>> That is the reason I felt for the need of a new compatible string >>> >>> specifically for Armada-38x SoCs. >>> >> >>> >> Indeed the Armada-38x SoCs does not support hardware offloading (at >>> >> least according the datasheet). But it happens that in the earlier >>> >> version of the Armada XP the hardware offloading was buggy, so we >>> >> introduced a compatible for this case: marvell,mv78230-a0-i2c. This >>> >> compatible enable the errata fix but not the offloading feature. That >>> >> means that it is exactly the compatible you need for Armada 38x (and >>> >> Armada 39x and 375 I think). >>> > >>> > The "mv78230-a0-i2c" dt-binding has the following note >>> > >>> > Note: Only use "marvell,mv78230-a0-i2c" for a >>> > very rare, initial version of the SoC which >>> > had broken offload support. Linux >>> > auto-detects this and sets it appropriately. >>> > >>> > If we are going to re-use this binding for armada-38x we should probably >>> > remove this note. Personally my preference would be an armada-38x >>> >>> Updating the documentation is the right thing to do yes. >>> >>> >>> > compatible string (or 370 if that's the common base of these SoCs). But >>> > of course we'll go with whatever your preference is as maintainer. >>> >>> If the IP is compatible then there is no reason to add a new one, else >>> we will end with a compatible for each SoC and the compatible property >>> will just loose its meaning. >> >> If you all had added compatibles for each SoC in the first place, then >> we wouldn't be having this dicussion. > > Really?? > > For Armada 38x we have already 6 flavor 381, 382, 383, 385, 388. Then > we have 9 SoC families on mvebu: orion5x, kirkwood, dove, Armada 370, > Armada XP, Armada 375, Armada 39x, Armada 7K, Armada 8K. Of course in > each family we have several flavors. > > But then Allwiner also use this driver in 8 SoC families: sun4i, sun5i, > sun6i, sun7i, sun8i, sun9i, sunxi, sun50i. Here again each family have > their own flavor. > > So you are just suggesting to blot the i2c driver with more than 50 > compatible string! No. > That also mean updating the i2c driver each time a new SoC flavor > appear, so more or less for each release. No. You can have one or several fallbacks. Use the fallback as long as you think they are the same and then add the more specific one to the driver when you find differences. Rob
Hi Rob, On mer., oct. 04 2017, Rob Herring <robh@kernel.org> wrote: > On Wed, Oct 4, 2017 at 3:33 AM, Gregory CLEMENT > <gregory.clement@free-electrons.com> wrote: >> Hi Rob, >> >> On mar., oct. 03 2017, Rob Herring <robh@kernel.org> wrote: >> >>> On Wed, Sep 27, 2017 at 09:33:00AM +0200, Gregory CLEMENT wrote: >>>> Hi Chris, >>>> >>>> On mer., sept. 27 2017, Chris Packham <Chris.Packham@alliedtelesis.co.nz> wrote: >>>> >>>> > Hi Gregory, >>>> > >>>> > On 27/09/17 00:56, Gregory CLEMENT wrote: >>>> >> Hi Kalyan, >>>> >> >>>> >> Please try avoid top-posting. >>>> >> >>>> >> On dim., sept. 24 2017, Kalyan Kinthada <Kalyan.Kinthada@alliedtelesis.co.nz> wrote: >>>> >> >>>> >>> Hi Gregory, >>>> >>> >>>> >>> I got this information from Armada-38x functional errata document. >>>> >> >>>> >> OK but in any case just adding a new compatible was not enough you have >>>> >> to update the driver in the same time, however for this case we won't >>>> >> need it, see below. >>>> >> >>>> >>> >>>> >>> I can add the "marvell,mv78230-i2c" compatible string to the appropriate device tree files >>>> >>> but the i2c-mv64xxx driver enables an additional feature (offload i2c transactions) >>>> >>> based on the compatible string "marvell,mv78230-i2c". >>>> >>> >>>> >>> I am not sure if this feature (offload i2c transactions) should be enabled for Armada-38x devices. >>>> >>> That is the reason I felt for the need of a new compatible string >>>> >>> specifically for Armada-38x SoCs. >>>> >> >>>> >> Indeed the Armada-38x SoCs does not support hardware offloading (at >>>> >> least according the datasheet). But it happens that in the earlier >>>> >> version of the Armada XP the hardware offloading was buggy, so we >>>> >> introduced a compatible for this case: marvell,mv78230-a0-i2c. This >>>> >> compatible enable the errata fix but not the offloading feature. That >>>> >> means that it is exactly the compatible you need for Armada 38x (and >>>> >> Armada 39x and 375 I think). >>>> > >>>> > The "mv78230-a0-i2c" dt-binding has the following note >>>> > >>>> > Note: Only use "marvell,mv78230-a0-i2c" for a >>>> > very rare, initial version of the SoC which >>>> > had broken offload support. Linux >>>> > auto-detects this and sets it appropriately. >>>> > >>>> > If we are going to re-use this binding for armada-38x we should probably >>>> > remove this note. Personally my preference would be an armada-38x >>>> >>>> Updating the documentation is the right thing to do yes. >>>> >>>> >>>> > compatible string (or 370 if that's the common base of these SoCs). But >>>> > of course we'll go with whatever your preference is as maintainer. >>>> >>>> If the IP is compatible then there is no reason to add a new one, else >>>> we will end with a compatible for each SoC and the compatible property >>>> will just loose its meaning. >>> >>> If you all had added compatibles for each SoC in the first place, then >>> we wouldn't be having this dicussion. >> >> Really?? >> >> For Armada 38x we have already 6 flavor 381, 382, 383, 385, 388. Then >> we have 9 SoC families on mvebu: orion5x, kirkwood, dove, Armada 370, >> Armada XP, Armada 375, Armada 39x, Armada 7K, Armada 8K. Of course in >> each family we have several flavors. >> >> But then Allwiner also use this driver in 8 SoC families: sun4i, sun5i, >> sun6i, sun7i, sun8i, sun9i, sunxi, sun50i. Here again each family have >> their own flavor. >> >> So you are just suggesting to blot the i2c driver with more than 50 >> compatible string! > > No. > >> That also mean updating the i2c driver each time a new SoC flavor >> appear, so more or less for each release. > > No. You can have one or several fallbacks. Use the fallback as long as > you think they are the same and then add the more specific one to the > driver when you find differences. So you mean adding all the possible variant of the SoC in the binding documentation, but only implement them in the driver when needed? Gregory > > Rob
On Wed, Oct 4, 2017 at 7:52 AM, Gregory CLEMENT <gregory.clement@free-electrons.com> wrote: > Hi Rob, > > On mer., oct. 04 2017, Rob Herring <robh@kernel.org> wrote: > >> On Wed, Oct 4, 2017 at 3:33 AM, Gregory CLEMENT >> <gregory.clement@free-electrons.com> wrote: >>> Hi Rob, >>> >>> On mar., oct. 03 2017, Rob Herring <robh@kernel.org> wrote: >>> >>>> On Wed, Sep 27, 2017 at 09:33:00AM +0200, Gregory CLEMENT wrote: >>>>> Hi Chris, >>>>> >>>>> On mer., sept. 27 2017, Chris Packham <Chris.Packham@alliedtelesis.co.nz> wrote: [...] >>>>> If the IP is compatible then there is no reason to add a new one, else >>>>> we will end with a compatible for each SoC and the compatible property >>>>> will just loose its meaning. >>>> >>>> If you all had added compatibles for each SoC in the first place, then >>>> we wouldn't be having this dicussion. >>> >>> Really?? >>> >>> For Armada 38x we have already 6 flavor 381, 382, 383, 385, 388. Then >>> we have 9 SoC families on mvebu: orion5x, kirkwood, dove, Armada 370, >>> Armada XP, Armada 375, Armada 39x, Armada 7K, Armada 8K. Of course in >>> each family we have several flavors. >>> >>> But then Allwiner also use this driver in 8 SoC families: sun4i, sun5i, >>> sun6i, sun7i, sun8i, sun9i, sunxi, sun50i. Here again each family have >>> their own flavor. >>> >>> So you are just suggesting to blot the i2c driver with more than 50 >>> compatible string! >> >> No. >> >>> That also mean updating the i2c driver each time a new SoC flavor >>> appear, so more or less for each release. >> >> No. You can have one or several fallbacks. Use the fallback as long as >> you think they are the same and then add the more specific one to the >> driver when you find differences. > > So you mean adding all the possible variant of the SoC in the binding > documentation, but only implement them in the driver when needed? Yes, and in the dts files. If some are the same die, but just different packaging or fusing then I would not distinguish those. But I realize you probably don't have that information being outside of Marvell. For the older parts, I'd guess they've been around long enough you've either found the differences by now or few people care anymore, so I'm not saying go back and add possible string now. Rob
diff --git a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt index 5c30026921..fc100b6bdb 100644 --- a/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt +++ b/Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt @@ -14,6 +14,7 @@ Required properties : very rare, initial version of the SoC which had broken offload support. Linux auto-detects this and sets it appropriately. + - "marvell,armada-38x-i2c" - interrupts : The interrupt number Optional properties :
All armada-38x variants(380, 385, 388) SoCs have an issue in i2c controller which violates the i2c repeated start timing. This errata is fixed in the i2c-mv64xxx driver but enabled only for devices with compatible string "marvell,mv78230-i2c". This patch introduces a new compatible string "marvell,armada-38x-i2c" for the i2c controller on armada-38x SoCs so that the workaound is enabled. Signed-off-by: Kalyan Kinthada <kalyan.kinthada@alliedtelesis.co.nz> --- Documentation/devicetree/bindings/i2c/i2c-mv64xxx.txt | 1 + 1 file changed, 1 insertion(+)