What´s the best Practice to Build own Kernel

Ryder.Lee · May 22, 2018, 9:06am

Please apply these patches:

https://patchwork.kernel.org/patch/10344913/ https://patchwork.kernel.org/patch/10341993/

ed1ba70.diff (9.2 KB) 915c49f.diff (4.0 KB)

frank-w · May 22, 2018, 4:42pm

added all so far to my 4.16-hdmi-branch, compiles ok, if anyone will test it

uploaded 3 versions of 4.16 to my gdrive for testing:

main (with usb-fix)
hdmi (hdmi-patches)
wlan (driver for internal wifi)

https://drive.google.com/open?id=1dJPz1REuIQiVfY4XliGOmGSZVDgT80Yc

garywang · May 23, 2018, 12:21am

Hi frank

Thanks for your sharing, I’d like to know if the HDMI interface is working fine on your branch.

DeadMeat · May 23, 2018, 3:35am

Hello, Frank.

I’ve tested your hdmi-patched kernel, it runs with a lot of errors. Looks like modules version magic is different from that kernel expects, please see attached log. (I didn’t tryed to run other versions)

4.16-hdmi-frank-w.log (29.0 КБ)

After i’ve rebuilt the kernel from hdmi branch with latest commits it has starded with only one hdmi related error:

[    1.051102] mediatek-dpi 14014000.dpi: Found bridge node: /hdmi@14015000
[    1.052741] random: crng init done
[    1.060319] mediatek-drm 14000000.dispsys: Adding component match for /ovl@14007000
[    1.070253] mediatek-drm 14000000.dispsys: Adding component match for /rdma@14008000
[    1.078065] mediatek-drm 14000000.dispsys: Adding component match for /color@1400b000
[    1.085947] mediatek-drm 14000000.dispsys: Adding component match for /rdma@14012000
[    1.093662] mediatek-drm 14000000.dispsys: Adding component match for /dpi@14014000
[    1.102368] mediatek-hdmi-phy 10209100.phy: Using default TX DRV impedance: 4.2k/36
[    1.111339] clk: failed to reparent hdmi_sel to hdmitx_dig_cts: -22
[    1.117727] [drm] hdmi-audio-codec driver bound to HDMI

See full log: 4.16-hdmi-self_built.log (20.7 КБ)

P.S. HDMI doesn’t work.

Ryder.Lee · May 23, 2018, 5:59am

clk: failed to reparent hdmi_sel to hdmitx_dig_cts: -22

Did you apply this patch https://patchwork.kernel.org/patch/10344913/ ?

DeadMeat · May 23, 2018, 6:04am

After i’ve added hdmitx_dig_cts to hdmi_parents (just like in 4.4) the error has gone.

clk-mt2701.diff

Now after kernel boots, there are no more hdmi errors, monitor doesn’t sows any messages about ‘no hdmi signal’, and does not suspends, but there is no image. The /dev/fb0 framebuffer does not creates.

Ryder.Lee · May 23, 2018, 6:09am

We use DRM framework so I guess you need to enable some configs for use.

DeadMeat · May 23, 2018, 6:15am

Yes, it’s applied.

Ok, thank You!

Ryder.Lee · May 23, 2018, 6:45am

It’s odd. Did you apply all of the patches? As we don’t need to add hdmitx_dig_cts to hdmi_parents.

https://lkml.org/lkml/2018/5/7/843

frank-w · May 23, 2018, 6:53am

here you can see patches i have applied: the 7 part patchset from Bibby Hsieh, your 4 patches and the iommu-patch

i used the v1 from this patchset from may 14 instead of the old from may 8, but on a quick look i see no difference in part 7/7

Ryder.Lee · May 23, 2018, 7:18am

ohhh, you don’t need to add this. This is a harmless error.

DeadMeat · May 23, 2018, 7:28am

i’ve traced initialisation from mtk_hdmi_phy.c:

static int mtk_hdmi_phy_probe(struct platform_device *pdev)
....
        return of_clk_add_provider(dev->of_node, of_clk_src_simple_get,
                                   hdmi_phy->pll);

to drivers/clk/clk.c:

static int clk_fetch_parent_index(struct clk_core *core,
				  struct clk_core *parent)
        ....
	for (i = 0; i < core->num_parents; i++)
		if (clk_core_get_parent_by_index(core, i) == parent)
			return i;

	return -EINVAL;

and the only way to fix it, i’ve found, was to add hdmitx_dig_cts to hdmi_parents.

Another good news! After a’ve added CONFIG_DRM_MALI_DISPLAY=y option the kernel became able to start Xorg server. But we steel have no framebuffer console. I didn’t found any gegeral fb device like VGA or VESA on x86 platform, so i don’t know what to add.

DeadMeat · May 23, 2018, 7:37am

Ryder.Lee · May 23, 2018, 7:41am

Looks Good . But I guess you could ignore that harmless error.

DeadMeat · May 23, 2018, 7:46am

Sorry, didn’t get it. I’m completely new to hardware drivers, so it’s possible that i can miss some obvious errors

frank-w · May 23, 2018, 7:47am

nice what have you changed in addition to my repo?

Ryder.Lee · May 23, 2018, 7:56am

Blockquote[quote=“DeadMeat, post:155, topic:3937”] Sorry, didn’t get it. I’m completely new to hardware drivers, so it’s possible that i can miss some obvious errors [/quote]

That is, I’m afraid the clock topology will be broken (not sure), and we will fix this .Can you try to adjust the resolution?

DeadMeat · May 23, 2018, 8:05am

Hello, Frank, i’ve applided diff --git a/drivers/clk/mediatek/clk-mt2701.c b/drivers/clk/mediatek/clk-mt2701.c index 4dda898…c560ac6 100644 — a/drivers/clk/mediatek/clk-mt2701.c +++ b/drivers/clk/mediatek/clk-mt2701.c @@ -315,9 +315,11 @@ static const char * const tve_parents[] = {

 static const char * const hdmi_parents[] = {
        "clk26m",
+       "hdmitx_dig_cts",
        "hdmipll_ck",
        "hdmipll_d2",
        "hdmipll_d3"
+
 };
 
 static const char * const apll_parents[] = {

Also i’ve enabled CONFIG_REGULATOR_FIXED_VOLTAGE=y, CONFIG_DRM_MALI_DISPLAY=y, and some other not hdmi related stuff. I can make diff between my and standart .config . I didn’t changed mt7623n_evb_fwu_defconfig

DeadMeat · May 23, 2018, 8:08am

I’ve tried few random resolution setting. Everything seems ok.

frank-w · May 23, 2018, 8:10am

CONFIG_REGULATOR_FIXED_VOLTAGE=y is only for usb, or is it needed for hdmi (found no reference in dts-changes)?

can you try without the “hdmitx_dig_cts”-patch, because ryder says it is not neccessary

so actually only CONFIG_DRM_MALI_DISPLAY=y is missing in defconfig, right?