Neil Armstrong

Is this only on Le Potato ? This should not happen, it fails while registering a clock.

@ning exact, the OSD scaler is shared between the OSD1 & OSD1, but we use the scaler to update automatically the field when in interlace mode, if we wanted to enable OSD2 we should switch the field manually in the vsync irq. It's doable but I don't plan on doing it For A311D, Amlogic did a brand new blender module for OSD and VD planes, but I don't have the exact capabilities of the blender and the number of available scalers, when I'll have the info, I may expose more OSD planes and eventually a Cursor plane.

No, sorry. The OSD2 plane does not have a scaler, so we cannot use it in all mode settings (especially interlaced outputs)

S805X doesn’t support 4K, even if it works some time, it’s not stable enough to consider functional, I sent fixes to Da for that, but he’s still testing them. any (binary) edid dump would help ! and for 4K development, the hdmi2 yuv420 4k60 mode is only used on Cheap 4K TVs, workout this patchset you’ll have 4k30 just fine

Indeed you need to write fip/u-boot.bin at the second block (512bytes)

@Igorthe firnware are stored here https://github.com/LibreELEC/meson-firmware until we manage to get them into linux-firmware...

@Igor The Debian stretch image works fine, I looked in the /boot directory of the Ubuntu images and it's empty

@TonyMac32 I'm getting a failure in u-boot with the latest 4.19 bionic desktop LePotato image I get from armbian.com :

@TonyMac32 the fixup was not right (it worked, but was not the right way), can you check if https://patchwork.freedesktop.org/patch/277316/ does the same effect ? I'm also checking on my side

Good to read !

@TonyMac32I found a regression, can you test this fix ? https://github.com/superna9999/linux/commit/b9edb9c9eec654ba57f8da7966e4b1b81a6d7c7b

@TonyMac32 What did you change between the 4.19.9-meson64 and 4.19.10-meson64 kernel ? the only upstream change is about adding the HDMI 1080p25 mode

@rino so you connect the same screen via HDMI and "HDMI to DVI" and only the "HDMI to DVI" fails ? What is the resolution selected via HDMI and "HDMI to DVI" ? is that 1920x1200, 1920x1080 or 1600*1200 ? On 4.19 I added a "generic" support for all monitors, but I waited for feedback to know if it failed on some monitors. Let me check

@rinoAudio is only supported out of HDMI now, we need funding (and time) to finish audio properly and have the 3,5mm Jack output and S/PDIF with data passthrough

And this behavior in only since 4.19 ?

@TonyMac32 With witch screen do you still have an issue you didn't had with the previous kernel releases ?

@Tommy21You won't have OpenGL acceleration, but yes should be able with the actual images. With a Bionic ubuntu desktop image, you should be able to select wayland gnome at the login prompt.

@Tommy21 No need to wait, it already works

Seems you found the solution !

We plan to use https://github.com/superna9999/pyamlboot for LaFrite, but it works on LePotato aswell, and it will be supported by the next U-boot version (in master right now https://github.com/u-boot/u-boot/commit/d96a782d09dbdc4a28ece3d18dc17a572e39d4f2) We still need to generate a generic initramfs+kernel image to flash anything.

@TonyMac32the overalloc is for the fbdev variant only, which won't work at all on 4.19 (needs a dirty quirk to enable it back which will land on 4.20 https://www.spinics.net/lists/dri-devel/msg191151.html), if you plan to use the X11, Wayland or GBM variant you should use the CONFIG_DRM_FBDEV_OVERALLOC=100 default config For USB, this patch https://github.com/chewitt/LibreELEC.tv/blob/amlogic/projects/Amlogic/patches/linux/linux-9999-dwc2.patch should solve your issue for the K2 and C2.

Yes this is normal since there is no HW OpenGL acceleration yet ! (But Lima is beginnning to be usable !) @TonyMac32i think I the last issue for your 4K display : drop the CONFIG_DRM_FBDEV_OVERALLOC=300 to CONFIG_DRM_FBDEV_OVERALLOC=100 this 300 value is only used for the libMali fbdev, otherwise it’s useless and a waste of memory !

@martinayotte well the person who posted the patch didn’t answer back to repost it in a correct way and we have very few users of s905 using this port...

@TonyMac32 weird stuff, could you dump the DRI driver status in /sys/kernel/debug/dri at each step ? same for EDID, EDID+status when freshly booted and non-working screen, and the same with it's working Thanks ! @guidol can you dump the output of read-edid in binary format ? the parse-edid gives Xorg format output, which is not the best :-p Neil

Oh yeah for the Ethernet, you should really take this fix https://patchwork.kernel.org/patch/10712159/ The IRQ type was wrong since the beginning... not it works correctly...