MIPI port from EdgeV to VIM3L issue

numbqq · January 4, 2020, 10:16am

@quatro Does the uboot logo display ok?

quatro · January 4, 2020, 7:06pm

no, it does not turn on any pixel actually…

quatro · January 4, 2020, 7:33pm

I’m seeing LCD initialization code in lcd.c for various displays… why are there so many unneeded displays, can this be interfering with my own display at uboot?

quatro · January 4, 2020, 7:36pm

why is this code in lcd.c? shouldn’t it be taken from the DTB?

quatro · January 6, 2020, 2:31am

Still can’t find the reason why it does not work on VIM3L like it does in EdgeV

quatro · January 7, 2020, 4:23am

Can you help me with this please? the lcd still looks like this no matter what I do…

numbqq · January 7, 2020, 6:08am

I don’t know what’s your LCD looks like, if your LCD has the same PIN definition as TS050, then you don’t need to modify the power control. If you use normal GPIOs, it will be more easier.

quatro · January 7, 2020, 6:56am

pinout is no problem, in fact it works with EdgeV perfectly… but in VIM3L the image breaks in pieces, see my second post…

I have set the same parameters, sync, clock, init mipi commands, everything is the same as EdgeV but it gets broken in VIM3L. I need to know what it is happening inside VIM3L because I already tried everything and it does not get fixed.

The closes I have come, is resetting the same attributes via /sys/class/lcd/change and the image gets a bit better…

what changes between VIM3L and EdgeV in the DSI that makes it not work correctly?

numbqq · January 7, 2020, 6:59am

Please try to resolve the problem step by step. I think the first step you need to make sure the u-boot logo display correct then is the kernel.

quatro · January 7, 2020, 7:01am

is it necessary to load the lcd driver in uboot for it to work ?

quatro · January 8, 2020, 12:44am

I have “solved” uboot display… the image was missing
I downloaded the official test image to the board:
http://www.mediafire.com/file/xoobk7gc3t5bo00/khadas.bmp/file

And this is what I get: (this behaviour also happens with TS050, the image is not blurry, this is an effect happening in the display)

numbqq · January 8, 2020, 1:57am

What image you test? I checked the LCD is OK with latest V191231 release.

quatro · January 8, 2020, 1:58am

the image is generated by fenix 0.7

numbqq · January 8, 2020, 1:59am

Please use the latest source code. Or download the image from our server.

quatro · January 8, 2020, 2:49am

I fixed it by running “fbset -fb /dev/fb0 -g 720 1280 720 2560 32”… can you guide me to a better and more “stable” way to set these parameters? anywhere in the linux source ?

numbqq · January 8, 2020, 6:27am

Maybe you can try to change the default fb setting in source code.

github.com

khadas/linux/blob/khadas-vims-4.9.y/drivers/amlogic/media/osd/osd_fb.c#L219


	},
	{
		COLOR_INDEX_32_ARGB, 1, 5,
		16, 8, 0, 8, 8, 0, 0, 8, 0, 24, 8, 0,
		FB_VISUAL_TRUECOLOR, 32
	},
	/*YUV color*/
	{COLOR_INDEX_YUV_422, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 16},
};


static struct fb_var_screeninfo fb_def_var[] = {
	{
		.xres            = 1920,
		.yres            = 1080,
		.xres_virtual    = 1920,
		.yres_virtual    = 2160,
		.xoffset         = 0,
		.yoffset         = 0,
		.bits_per_pixel = 32,
		.grayscale       = 0,
		.red             = {16, 8, 0},

By the way, I have updated the GPIO expander, you can check here is you have interest.(just fot testing, not final version)

quatro · January 8, 2020, 8:50am

I have rebuild the kernel but it does not take the default parameters… do you know where else to make the change ?

fbset output:

# fbset

mode "1088x1920"
    geometry 1088 1920 1088 3840 32
    timings 0 0 0 0 0 0 0
    nonstd 1
    rgba 8/16,8/8,8/0,0/24
endmode

numbqq · January 8, 2020, 8:52am

You also need to modify this file and rebuilt the image.

github.com

khadas/fenix/blob/master/archives/filesystem/special/VIM-COMMON/etc/initramfs-tools/scripts/local-top/khadas_init.sh#L32


      
          			bpp=${x#*=}
          			;;
          		vout=*)
          			vout=${x#*=}
          			;;
          		panel_exist=*)
          			panel_exist=${x#*=}
          			;;
          		panel_type=*)
          			panel_type=${x#*=}
          			;;
          	esac
          done
          
          display_device=`echo $vout | awk -F ',' '{print $1}'`
          
          if [ $panel_exist -eq 1 ] && [ $display_device = panel ]; then
          
          	#echo null > /sys/class/display/mode
          
          	echo panel > /sys/class/display/mode

quatro · January 8, 2020, 8:56am

that was one that I found soon after finding out the framebuffer was the problem…
I have also found panel_setup.sh

but if I modify them in the running OS in VIM3L board it does not do anything, this is something that needs to run during build of kernel, am I right ?

numbqq · January 8, 2020, 8:59am

You can’t simplely modify them. Because them are in initramfs not in rootfs. Suggest you to use dpkg to install the image deb package after you modify the script, so that new initramfs will be generated.

If you don’t know how to do this, you can modify the script in fenix source code and rebuilt the image.