Re: [PATCH v2 14/19] ARM: dts: Add bus nodes using VDD_INT for Exynos4x12

[Krzysztof Kozlowski]

On 10.12.2015 15:43, Chanwoo Choi wrote:
> On 2015ë 12ì 10ì 15:32, Krzysztof Kozlowski wrote:
>> On 10.12.2015 15:08, Chanwoo Choi wrote:
>>> On 2015ë 12ì 10ì 14:57, Krzysztof Kozlowski wrote:
>>>> On 09.12.2015 13:08, Chanwoo Choi wrote:
>>>>> This patch adds the bus noes using VDD_INT for Exynos4x12 SoC.
>>>>> Exynos4x12 has the following AXI buses to translate data between
>>>>> DRAM and sub-blocks.
>>>>>
>>>>> Following list specifies the detailed relation between DRAM and sub-blocks:
>>>>> - ACLK100 clock for PERIL/PERIR/MFC(PCLK)
>>>>> - ACLK160 clock for CAM/TV/LCD
>>>>> : The minimum clock of ACLK160 should be over 160MHz.
>>>>>   When drop the clock under 160MHz, show the broken image.
>>>>> - ACLK133 clock for FSYS
>>>>> - GDL clock for LEFTBUS
>>>>> - GDR clock for RIGHTBUS
>>>>> - SCLK_MFC clock for MFC
>>>>>
>>>>> Signed-off-by: Chanwoo Choi <cw00.choi@xxxxxxxxxxx>
>>>>> ---
>>>>>  arch/arm/boot/dts/exynos4x12.dtsi | 112 ++++++++++++++++++++++++++++++++++++++
>>>>>  1 file changed, 112 insertions(+)
>>>>>
>>>>> diff --git a/arch/arm/boot/dts/exynos4x12.dtsi b/arch/arm/boot/dts/exynos4x12.dtsi
>>>>> index 3bcf0939755e..8bc4aee156b5 100644
>>>>> --- a/arch/arm/boot/dts/exynos4x12.dtsi
>>>>> +++ b/arch/arm/boot/dts/exynos4x12.dtsi
>>>>> @@ -354,6 +354,118 @@
>>>>>  			opp-microvolt = <950000>;
>>>>>  		};
>>>>>  	};
>>>>> +
>>>>> +	bus_leftbus: bus_leftbus {
>>>>> +		compatible = "samsung,exynos-bus";
>>>>> +		clocks = <&clock CLK_DIV_GDL>;
>>>>> +		clock-names = "bus";
>>>>> +		operating-points-v2 = <&bus_leftbus_opp_table>;
>>>>> +		status = "disabled";
>>>>> +	};
>>>>> +
>>>>> +	bus_rightbus: bus_rightbus {
>>>>> +		compatible = "samsung,exynos-bus";
>>>>> +		clocks = <&clock CLK_DIV_GDR>;
>>>>> +		clock-names = "bus";
>>>>> +		operating-points-v2 = <&bus_leftbus_opp_table>;
>>>>> +		status = "disabled";
>>>>> +	};
>>>>
>>>> These two nodes are symmetrical. The MFC below and other buses in other
>>>> DTS share opps. How about changing the binding so multiple clocks could
>>>> be specified at once ("bus0", "bus1")? I think there is no need for a
>>>> bus device for each clock.
>>>
>>> The your commented method is possible.
>>>
>>> But, I focus on implementing the generic bus frequency driver.
>>>
>>> If specific bus device-tree node includes the one more clocks,
>>> when adding the new Exynos SoC, the exynos-bus.c should be added
>>> for new Exynos SoC. Because, each Exynos SoC has the different
>>> set of bus device.
>>>
>>> If we use my approach, we don't need to modify the exynos-bus.c
>>> driver to support for the bus frequency of new Exynos SoC.
>>
>> This won't change. The driver will just support from 1 to N clocks for
>> given bus device and set the same OPP to all of them. This will only
>> limit the number of duplicated entries. This won't affect the generic
>> approach of driver itself.
> 
> You're right aspect of only implementation of device driver.
> 
> But,
> If we use your commented approach, we can show the information
> of only parent device through sysfs. We cannot see the information
> of passive device. The some information includes the current
> frequency and correlation of parent device. (But, current patchset
> don' include the topology information between parent device and
> passive device. I'll do it on later patches).
> 
> For example, 
> We can see the following bus device through /sys/class/devfreq.
> 
> drwxr-xr-x  2 root root 0 Dec 31 17:00 .
> drwxr-xr-x 44 root root 0 Dec 31 17:00 ..
> lrwxrwxrwx  1 root root 0 Dec 31 17:00 bus_display -> ../../devices/platform/bus_display/devfreq/bus_display
> lrwxrwxrwx  1 root root 0 Dec 31 17:00 bus_fsys -> ../../devices/platform/bus_fsys/devfreq/bus_fsys
> lrwxrwxrwx  1 root root 0 Dec 31 17:00 bus_leftbus -> ../../devices/platform/bus_leftbus/devfreq/bus_leftbus
> lrwxrwxrwx  1 root root 0 Dec 31 17:00 bus_peri -> ../../devices/platform/bus_peri/devfreq/bus_peri
> 
> 
> We don't see the following bus device because of following bus device
> has the same frequency table with bus_leftbus device.
> lrwxrwxrwx  1 root root 0 Dec 31 17:00 bus_mfc -> ../../devices/platform/bus_mfc/devfreq/bus_mfc
> lrwxrwxrwx  1 root root 0 Dec 31 17:00 bus_rightbus -> ../../devices/platform/bus_rightbus/devfreq/bus_rightbus

Right, but why do you want to see these bus devices? AFAIU, they will
always behave exactly the same as LEFTBUS. Their PPMU counters will be
the same... or not? The MFC does not have its own PPMU counter. There
are separate counters for left and right bus... but they are attached to
the "&bus_leftbus" node. The "&bus_rightbus" does not use the PPMU
counter and it follows the parent governor decisions... so this is
purely an artificial creation just to handle one clock.

I just can't see the benefit of such additional bus device.

Best regards,
Krzysztof

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/