Hi Hiep-san,
On Wed, Apr 6, 2016 at 3:47 AM, Cao Minh Hiep <cm-hiep@xxxxxxxxxxx> wrote:
> On 04/05/2016 07:36 PM, Geert Uytterhoeven wrote:
>> On Tue, Apr 5, 2016 at 4:35 AM, Cao Minh Hiep <cm-hiep@xxxxxxxxxxx> wrote:
>>> On 04/04/2016 05:05 PM, Geert Uytterhoeven wrote:
>>>> On Fri, Apr 1, 2016 at 9:57 AM, Cao Minh Hiep <cm-hiep@xxxxxxxxxxx>
>>>> wrote:
>>>>> This Linux version v4.5, we can not show an image or picture on HDMI
>>>>> display.
>>>>> (There is no problem on v4.5-rc3, could show an image on HDMI display
>>>>> normally)
>> After moving Koelsch to the living room, and connecting it to a Full HD
>> Sony TV, I get:
>>
>> rcar-du feb00000.display: failed to allocate buffer with size 8294400
>> rcar-du feb00000.display: Failed to set initial hw configuration.
>> rcar-du feb00000.display: failed to initialize DRM/KMS (-12)
>> rcar-du: probe of feb00000.display failed with error -12
>>
>> Don't you get this? The DELL S2240L is also Full HD, so it should need a
>> buffer
>> of 1920 * 1080 * 4 bytes, too (unless its preferred mode is smaller).
>
> I also got exactly the same log with v4.5-stable.
> I attach the test logs + boot logs and uImage.dtb in three cases of v4.5-rc3
> and v4.5-stable
> and v4.5 stable after reverting the patch of b26a719.
Thanks for the logs, they were very helpful to understand what's going on!
1. Your good case (v4.5-rc3, or v4.5 with commit b26a719 reverted) has:
Console: switching to colour frame buffer device 128x48
rcar-du feb00000.display: fb0: frame buffer device
[drm] Initialized rcar-du 1.0.0 20130110 on minor 0
While your bad case (v4.5) has:
rcar-du feb00000.display: failed to allocate buffer with size 8294400
rcar-du feb00000.display: Failed to set initial hw configuration.
rcar-du feb00000.display: failed to initialize DRM/KMS (-12)
rcar-du: probe of feb00000.display failed with error -12
Actually I believe the former is the bad case (EDID retrieval through HDMI
fails, and the drivers falls back to the default resolution of 1024x768 =>
128x48 text console), and the latter is the good case (EDID retrieval
succeeded, and the driver detected the preferred resolution of 1920x1080).
Without commit b26a719 ("gpio: rcar: Add Runtime PM handling for
interrupts"), the GPIO interrupt used for HDMI EDID retrieval will not
trigger if the GPIO MSTP clock is disabled.
2. Your Koelsch has U-Boot 46367b1, while mine has b6af5fc. The former disables
many MSTP clocks (incl . GPIO), while the latter keeps them enabled, so
that's why v4.5-rc3 worked for me.
Usually I'm using local early kernel code to disable unused MSTP clocks,
but not this time, as you mentioned plain v4.5-rc3 and v4.5.
So the real fix is to enable CONFIG_CMA and CONFIG_DMA_CMA, but
this requires fixing ohci-pci first.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
