Re: dtc warnings

Re: dtc warnings

[Nishanth Menon]

On 04/01/2016 02:53 AM, Krzysztof Kozlowski wrote:
> On Fri, Apr 1, 2016 at 4:49 PM, Tomi Valkeinen <tomi.valkeinen@gmail.com> wrote:
>>
>> Hi Rob,
>>
>> On 01/04/16 04:40, Rob Herring wrote:
>>> You may have noticed that linux-next had gotten noisy with dtc
>>> warnings lately. I dropped the change for a bit, but added it back
>>> today except now it is disabled unless building with "W=1".
>>>
>>> There's ~25K (2500 unique) warnings generated from the ARM dts files.
>>> Here's the ranking of warnings by dtb. OMAP is the clear winner (based
>>> on the similar counts, probably lots of duplicates). Please help
>>> remind contributors to test with W=1 and start to fix these.
>>>
>>> At least for memory nodes, I plan to whitelist allowing no
>>> unit-address. There could be others, but none that I've seen so far.
>>
>> What's the correct way to fix nodes for display platform devices? For
>> example, omap4-panda-common.dtsi has two connector nodes:
>>
>> dvi0: connector@0 {
>>         compatible = "dvi-connector";
>>         label = "dvi";
>>         ...
>> };
>>
>>
>> hdmi0: connector@1 {
>>         compatible = "hdmi-connector";
>>         label = "hdmi";
>>         ...
>> };
> 
> I have the same doubts. The ePAPR says in that case "the node-name
> alone differentiates the node from other nodes at the same level in
> the tree.". But which is preferred? Differentiating by number or by
> type?
> 
> Similarly, what to do with the opp modes (a lot of warnings) in
> Documentation/devicetree/bindings/opp/opp.txt?
> 
+ linux-pm and Viresh for opp.txt


-- 
Regards,
Nishanth Menon

Re: dtc warnings

[Viresh Kumar]

On 01-04-16, 08:57, Nishanth Menon wrote:
> On 04/01/2016 02:53 AM, Krzysztof Kozlowski wrote:
> > On Fri, Apr 1, 2016 at 4:49 PM, Tomi Valkeinen <tomi.valkeinen@gmail.com> wrote:
> >>
> >> Hi Rob,
> >>
> >> On 01/04/16 04:40, Rob Herring wrote:
> >>> You may have noticed that linux-next had gotten noisy with dtc
> >>> warnings lately. I dropped the change for a bit, but added it back
> >>> today except now it is disabled unless building with "W=1".
> >>>
> >>> There's ~25K (2500 unique) warnings generated from the ARM dts files.
> >>> Here's the ranking of warnings by dtb. OMAP is the clear winner (based
> >>> on the similar counts, probably lots of duplicates). Please help
> >>> remind contributors to test with W=1 and start to fix these.
> >>>
> >>> At least for memory nodes, I plan to whitelist allowing no
> >>> unit-address. There could be others, but none that I've seen so far.
> >>
> >> What's the correct way to fix nodes for display platform devices? For
> >> example, omap4-panda-common.dtsi has two connector nodes:
> >>
> >> dvi0: connector@0 {
> >>         compatible = "dvi-connector";
> >>         label = "dvi";
> >>         ...
> >> };
> >>
> >>
> >> hdmi0: connector@1 {
> >>         compatible = "hdmi-connector";
> >>         label = "hdmi";
> >>         ...
> >> };
> > 
> > I have the same doubts. The ePAPR says in that case "the node-name
> > alone differentiates the node from other nodes at the same level in
> > the tree.". But which is preferred? Differentiating by number or by
> > type?
> > 
> > Similarly, what to do with the opp modes (a lot of warnings) in
> > Documentation/devicetree/bindings/opp/opp.txt?

What warnings are you talking about ?

-- 
viresh

Re: dtc warnings

[Krzysztof Kozlowski]

On Sun, Apr 03, 2016 at 12:29:09PM +0530, Viresh Kumar wrote:
> On 01-04-16, 08:57, Nishanth Menon wrote:
> > On 04/01/2016 02:53 AM, Krzysztof Kozlowski wrote:
> > > On Fri, Apr 1, 2016 at 4:49 PM, Tomi Valkeinen <tomi.valkeinen@gmail.com> wrote:
> > >>
> > >> Hi Rob,
> > >>
> > >> On 01/04/16 04:40, Rob Herring wrote:
> > >>> You may have noticed that linux-next had gotten noisy with dtc
> > >>> warnings lately. I dropped the change for a bit, but added it back
> > >>> today except now it is disabled unless building with "W=1".
> > >>>
> > >>> There's ~25K (2500 unique) warnings generated from the ARM dts files.
> > >>> Here's the ranking of warnings by dtb. OMAP is the clear winner (based
> > >>> on the similar counts, probably lots of duplicates). Please help
> > >>> remind contributors to test with W=1 and start to fix these.
> > >>>
> > >>> At least for memory nodes, I plan to whitelist allowing no
> > >>> unit-address. There could be others, but none that I've seen so far.
> > >>
> > >> What's the correct way to fix nodes for display platform devices? For
> > >> example, omap4-panda-common.dtsi has two connector nodes:
> > >>
> > >> dvi0: connector@0 {
> > >>         compatible = "dvi-connector";
> > >>         label = "dvi";
> > >>         ...
> > >> };
> > >>
> > >>
> > >> hdmi0: connector@1 {
> > >>         compatible = "hdmi-connector";
> > >>         label = "hdmi";
> > >>         ...
> > >> };
> > > 
> > > I have the same doubts. The ePAPR says in that case "the node-name
> > > alone differentiates the node from other nodes at the same level in
> > > the tree.". But which is preferred? Differentiating by number or by
> > > type?
> > > 
> > > Similarly, what to do with the opp modes (a lot of warnings) in
> > > Documentation/devicetree/bindings/opp/opp.txt?
> 
> What warnings are you talking about ?

Warnings coming from recent linux-next:

  DTC     arch/arm/boot/dts/exynos5800-peach-pi.dtb
Warning (unit_address_vs_reg): Node /opp_table0/opp@1800000000 has a unit name, but no reg property
Warning (unit_address_vs_reg): Node /opp_table0/opp@1700000000 has a unit name, but no reg property
...

Best regards,
Krzysztof

Re: dtc warnings

[Viresh Kumar]

On 3 April 2016 at 13:38, Krzysztof Kozlowski <k.kozlowski@samsung.com> wrote:
> Warnings coming from recent linux-next:
>
>   DTC     arch/arm/boot/dts/exynos5800-peach-pi.dtb
> Warning (unit_address_vs_reg): Node /opp_table0/opp@1800000000 has a unit name, but no reg property
> Warning (unit_address_vs_reg): Node /opp_table0/opp@1700000000 has a unit name, but no reg property

@Rob: What should we be doing about these warnings?
OPP-entries don't have a reg property obviously and we
discussed earlier if naming them based on freq is fine or
not.

--
viresh

Re: dtc warnings

[Rob Herring]

On Sun, Apr 03, 2016 at 05:57:00PM +0530, Viresh Kumar wrote:
> On 3 April 2016 at 13:38, Krzysztof Kozlowski <k.kozlowski@samsung.com> wrote:
> > Warnings coming from recent linux-next:
> >
> >   DTC     arch/arm/boot/dts/exynos5800-peach-pi.dtb
> > Warning (unit_address_vs_reg): Node /opp_table0/opp@1800000000 has a unit name, but no reg property
> > Warning (unit_address_vs_reg): Node /opp_table0/opp@1700000000 has a unit name, but no reg property
> 
> @Rob: What should we be doing about these warnings?
> OPP-entries don't have a reg property obviously and we
> discussed earlier if naming them based on freq is fine or
> not.

We can whitelist opp nodes in dtc, use oppN instead, or ignore the 
warnings for now. 

Rob

Re: dtc warnings

[Viresh Kumar]

On 04-04-16, 00:16, Rob Herring wrote:
> On Sun, Apr 03, 2016 at 05:57:00PM +0530, Viresh Kumar wrote:
> > On 3 April 2016 at 13:38, Krzysztof Kozlowski <k.kozlowski@samsung.com> wrote:
> > > Warnings coming from recent linux-next:
> > >
> > >   DTC     arch/arm/boot/dts/exynos5800-peach-pi.dtb
> > > Warning (unit_address_vs_reg): Node /opp_table0/opp@1800000000 has a unit name, but no reg property
> > > Warning (unit_address_vs_reg): Node /opp_table0/opp@1700000000 has a unit name, but no reg property

Something isn't right in my setup, as I am not hitting these:

$ make CROSS_COMPILE="arm-linux-gnueabihf-" ARCH=arm O=../bexynos dtbs
make[1]: Entering directory '/mnt/ssd/all/work/repos/devel/bexynos'
  CHK     include/config/kernel.release
  Using /mnt/ssd/all/work/repos/devel/linux as source for kernel
  GEN     ./Makefile
  CHK     include/generated/uapi/linux/version.h
  CHK     include/generated/utsrelease.h
make[2]: 'include/generated/mach-types.h' is up to date.
  CHK     include/generated/bounds.h
  CHK     include/generated/timeconst.h
  CHK     include/generated/asm-offsets.h
  CALL    /mnt/ssd/all/work/repos/devel/linux/scripts/checksyscalls.sh
<stdin>:1310:2: warning: #warning syscall preadv2 not implemented [-Wcpp]
<stdin>:1313:2: warning: #warning syscall pwritev2 not implemented [-Wcpp]
  DTC     arch/arm/boot/dts/exynos3250-artik5-eval.dtb
  DTC     arch/arm/boot/dts/exynos3250-monk.dtb
  DTC     arch/arm/boot/dts/exynos3250-rinato.dtb
  DTC     arch/arm/boot/dts/exynos4210-origen.dtb
  DTC     arch/arm/boot/dts/exynos4210-smdkv310.dtb
  DTC     arch/arm/boot/dts/exynos4210-trats.dtb
  DTC     arch/arm/boot/dts/exynos4210-universal_c210.dtb
  DTC     arch/arm/boot/dts/exynos4412-odroidu3.dtb
  DTC     arch/arm/boot/dts/exynos4412-odroidx.dtb
  DTC     arch/arm/boot/dts/exynos4412-odroidx2.dtb
  DTC     arch/arm/boot/dts/exynos4412-origen.dtb
  DTC     arch/arm/boot/dts/exynos4412-smdk4412.dtb
  DTC     arch/arm/boot/dts/exynos4412-tiny4412.dtb
  DTC     arch/arm/boot/dts/exynos4412-trats2.dtb
  DTC     arch/arm/boot/dts/exynos5250-arndale.dtb
  DTC     arch/arm/boot/dts/exynos5250-smdk5250.dtb
  DTC     arch/arm/boot/dts/exynos5250-snow.dtb
  DTC     arch/arm/boot/dts/exynos5250-snow-rev5.dtb
  DTC     arch/arm/boot/dts/exynos5250-spring.dtb
  DTC     arch/arm/boot/dts/exynos5260-xyref5260.dtb
  DTC     arch/arm/boot/dts/exynos5410-smdk5410.dtb
  DTC     arch/arm/boot/dts/exynos5420-arndale-octa.dtb
  DTC     arch/arm/boot/dts/exynos5420-peach-pit.dtb
  DTC     arch/arm/boot/dts/exynos5420-smdk5420.dtb
  DTC     arch/arm/boot/dts/exynos5422-odroidxu3.dtb
  DTC     arch/arm/boot/dts/exynos5422-odroidxu3-lite.dtb
  DTC     arch/arm/boot/dts/exynos5422-odroidxu4.dtb
  DTC     arch/arm/boot/dts/exynos5440-sd5v1.dtb
  DTC     arch/arm/boot/dts/exynos5440-ssdk5440.dtb
  DTC     arch/arm/boot/dts/exynos5800-peach-pi.dtb
make[1]: Leaving directory '/mnt/ssd/all/work/repos/devel/bexynos'

> > 
> > @Rob: What should we be doing about these warnings?
> > OPP-entries don't have a reg property obviously and we
> > discussed earlier if naming them based on freq is fine or
> > not.
> 
> We can whitelist opp nodes in dtc, use oppN instead, or ignore the 
> warnings for now. 

I will be required to update check_unit_address_vs_reg() for that, right ? Or
is there a easier way for doing this ? :)

-- 
viresh

Re: dtc warnings

[Krzysztof Kozlowski]

On 04.04.2016 14:50, Viresh Kumar wrote:
> On 04-04-16, 00:16, Rob Herring wrote:
>> On Sun, Apr 03, 2016 at 05:57:00PM +0530, Viresh Kumar wrote:
>>> On 3 April 2016 at 13:38, Krzysztof Kozlowski <k.kozlowski@samsung.com> wrote:
>>>> Warnings coming from recent linux-next:
>>>>
>>>>   DTC     arch/arm/boot/dts/exynos5800-peach-pi.dtb
>>>> Warning (unit_address_vs_reg): Node /opp_table0/opp@1800000000 has a unit name, but no reg property
>>>> Warning (unit_address_vs_reg): Node /opp_table0/opp@1700000000 has a unit name, but no reg property
> 
> Something isn't right in my setup, as I am not hitting these:
> 
> $ make CROSS_COMPILE="arm-linux-gnueabihf-" ARCH=arm O=../bexynos dtbs

You need W=1 (that was recent change by Rob to not pollute everyone's
builds).

Best regards,
Krzysztof