[PATCH 1/4] drm/exynos: dsim: fix to control mipi phy register

[Inki Dae]

On 2015년 02월 09일 19:57, Sylwester Nawrocki wrote:
> On 07/02/15 12:53, Inki Dae wrote:
>> This patch fixes the issue that the try to get a phy object is failed
>> to enable mipi phy.
>>
>> System and power management unit registers should be controlled by
>> syscon framework. So this patch removes existing phy framework based
>> codes and adds syscon support instead. However, we should support
>> legacy device tree binding so consider the legacy binding for compatibility.
>>
>> In addition, we need to remove below device node and relevant properties,
>> 	mipi_phy: video-phy at 10020710 {
>> 		compatible = "samsung,s5pv210-mipi-video-phy";
>> 		reg = <0x10020710 8>;
>> 		#phy-cells = <1>;
>> 	};
>>
>> Now camera device node uses mipi_phy node relevant properties like below,
>> 	camera {
>> 		...
>> 		csis_0: csis at 11880000 {
>> 			...
>> 			phys = <&mipi_phy 0>;
>> 			phy-names = "csis";
>> 			...
>> 		};
>> 		csis_1: csis at 11890000 {
>> 			...
>> 			phys = <&mipi_phy 2>;
>> 			phy-names = "csis";
>> 			...
>> 		};
>> 		...
>> 	};
>>
>> With above, we will find below message while booting,
>>      can't request region for resource [mem 0x10020710-0x10020717]
> 
> I'm afraid this approach won't work because MIPI DSI Master and MIPI CSI
> Slave devices share a control bit in the register and it seems impossible
> to ensure proper locking with current regmap/syscon API.
> 
> I have submitted patches to fix this issue [1] and they should be already
> available in linux-next and can be found on linux-samsung-soc ML:
> 
> [PATCH 1/2] phy: exynos-video-mipi: Fix regression by adding support for PMU regmap
> [PATCH 2/2] ARM: dts: Add syscon phandle to the video-phy node for Exynos4
> 
> The other issue with your approach is that we are moving the PMU details
> to the MIPI DSIM driver and similar changes would need to be done in
> the MIPI CSIS driver.
> 
> Instead I just added syscon support to the PHY layer, it's not perfect
> but fixes the issue for both DSI and CSI and  doesn't strip the PHY layer
> which could potentially be useful.

Ah, Right. I didn't check your patch set. Your way is a better idea than
my one. With this, we don't need to change device drivers, MIPI DSI and CSI.

Then, what is the meaning that it's not perfect?

Thanks,
Inki Dae

>