On Tue, Sep 28, 2021 at 5:55 AM Anatolij Gustschin agust@denx.de wrote:
Hey Tim,
On Mon, 27 Sep 2021 17:25:58 -0700 Tim Harvey tharvey@gateworks.com wrote:
Anatolij,
Since commit d37618d18d49 ("imx: convert gwventana board to DM_VIDEO") video support for IMX6 based Ventana boards has been broken.
Back then I've tested similar DM_VIDEO conversion changes on i.mx6q nitrogen6q board and on i.mx6d/i.mx6s wandboards, it was okay if the board configuration uses the video console output during the boot sequence (i.e. configured to show splash screen or to output strings on vidconsole).
I find that while the bind function for fsl_imx6q_ipu is called the probe never is (ipuv3_video_probe). Do you know why this is?
Is video console output used during the boot? If not, then it might be the reason. With DM, the devices are probed when a subsystem actually tries to use them. You can try to trigger probing via switching to the video console output, i.e.:
=> setenv stdout vidconsole
Yes, this calls ipuv3_video_probe. I guess I expected the display to just work by default as it did before.
I looked over doc/README.console. It may be out of date as it refers to 'video' instead of 'vidconsole'. What is the difference?
How do I get back to the state where a splash-screen is shown on the display by default?
I see that with commit 57f065fee2a4 ("video: ipuv3: add DM_VIDEO support") you mention that DTS files must include 'u-boot,dm-pre-reloc' for soc/ipu nodes to enable driver binding to ipu device but I haven't been able to get that to make a difference nor have I found a board that does this. You did add those props to imx6qdl.dtsi at one point but they are no longer there.
Does your board dts include imx6qdl-u-boot.dtsi ?
The u-boot,dm-pre-reloc properties were moved to this U-Boot specific dtsi in 7932b1c9fdb73393aa110249c89bd426533c0649 (imx: imx6qdl: dtsi: move U-Boot specific change to u-boot.dtsi)
No, I'm not including that but that is the right place for it.
Strangely, I found that if I don't have 'u-boot,dm-pre-reloc' for soc/ipu nodes enabling video via 'setenv stdout serial,vidconsole' works. Is there some reason why the prop is no longer required?
There must have been several other IMX6 boards with video support that were affected by this so perhaps I'm missing something simple.
I did not test it recently, it looks I'll have to do it before the new release.
I would think the other board maintainers would have raised a red flag if things broke for their boards, I just haven't tested it in quite some time for gwventana_*_defconfig.
How do you go about testing this by the way without having all the boards and displays?
Best regards,
Tim