<inode>
Kwiboo: anything in particular you'd like to check in terms of those patches? i've applied them on v5.5-rc7 and so far i can see a few additional modes listed as being available according to libdrm's modetest
<inode>
in the meantime, i'll paste some drm debug (0x0e) output and the modetest output to a bin for comparison
<inode>
and here's the output from the same commands when running the same kernel version without your patches: https://pastebin.com/TJBhD3Yz
ldevulder_ has joined #linux-rockchip
ldevulder has quit [Ping timeout: 265 seconds]
<inode>
i usually have "drm.edid_firmware=edid/1920x1080.bin" added to my command line - and without the edid override and without your patches, when i run programs like kmscube, glmark2-es2-drm or Xorg: my monitor freaks out about the input mode generated (attempts at 1024x768 presumably due to the best mode drm seems to have probed at the time)
<inode>
but without the edid override and your patches applied - the monitor accepts the mode and the output of the programs appear as 1024x768 (FPS is about 2x better than when running at 1920x1080, and the pilar box is obvious)
lkcl has quit [Ping timeout: 258 seconds]
<inode>
in all cases, the framebuffer console appears at 1920x1080 - but the edid override is the only way to convince drm to use something better than 1024x768
<inode>
at least in the current state of the code and my configuration
<inode>
i guess there is something wrong or missing on my end because after the first two calls to drm_helper_probe_single_connector_modes - the probed mode list is drastically reduced
<Kwiboo>
inode: thanks for testing, it looks strange that the first drm_helper_probe_single_connector_modes returns more modes and include 1920x1080 compared to the modetest run
vicencb has quit [Quit: Leaving.]
<inode>
i wonder if DDC is configured correctly
<Kwiboo>
the edid reported by modetest looks valid and includes the first set of modes reported in kernel log, I have mainly tested on a TV and not a monitor
<Kwiboo>
it is very stange that modetest and initial kernel modeline parse return different modes, and also some of the new modes should have been filtered out due to missing rates in inno hdmi phy pll table
<Kwiboo>
was the device tree patch applied and your dtb file updated? do you have the cat /sys/kernel/debug/clk/clk_summary output, maybe I selected wrong vpll clock or the parent clock is wrong :-)
<inode>
i'll check
<Kwiboo>
I will also do some more tests on a regular monitor on my end, to be sure it works as expected on rk3328
anarsoul has quit [Remote host closed the connection]
<inode>
doesn't look like i rebuilt the dtb, i'll try again
anarsoul has joined #linux-rockchip
<Kwiboo>
thanks, without an updated dtb all modes will be reported okay, and with an updated dtb with hdmi vpll clock, the modes not supported by inno hdmi phy pll table should be filtered out and reported as BAD/CLOCK_RANGE