NetBSD 11 RC4 xrandr issue

r0ller

Hi All,

After installing RC4 my earlier issue with my AMD Radeon R9 Nano vanished since the amdgpu driver kicks in at last but the only resolution xrandr can identify and use is 1024x768. However, earlier I could use 1920x1080. I tried setting what cvt and gtf suggest without any success. The first error I get is:
xrandr: screen cannot be larger than 1024x768

Which I could not fix via xrandr but in arandr I managed to adjust the size. However, after that when selecting the new 1920x1080 mode, I get the error:
Configure crtc 0 failed

After some googling I bumped into a page where someone had the same issue:
https://commonplace.doubleloop.net/xrandr-configure-crtc-0-failed-problems

I also looked up my monitor's specs and it was indeed different from what cvt and gtf suggest:
`1920x1080 @ 60.00 Hz (GTF) hsync: 67.08 kHz; pclk: 172.80 MHz
Modeline "1920x1080_60.00" 172.80 1920 2040 2248 2576 1080 1081 1084 1118 -HSync +Vsync

1920x1080 59.96 Hz (CVT 2.07M9) hsync: 67.16 kHz; pclk: 173.00 MHz
Modeline "1920x1080_60.00" 173.00 1920 2048 2248 2576 1080 1083 1088 1120 -hsync +vsync`

In the specs pclk is 148.5 while hsync and vsync are +/+ instead of -/+. Unfortunately, no matter which of these I try to apply, I still get the same error mentioned above. Anyhow, this what xrandr verbose says now:
Screen 0: minimum 1024 x 768, current 1024 x 768, maximum 1920 x 1080 default connected 1024x768+0+0 (0x51c) normal (normal) 0mm x 0mm Identifier: 0x51b Timestamp: 4357110 Subpixel: unknown Gamma: 1.0:1.0:1.0 Brightness: 0.0 Clones: CRTC: 0 CRTCs: 0 Transform: 1.000000 0.000000 0.000000 0.000000 1.000000 0.000000 0.000000 0.000000 1.000000 filter: non-desktop: 0 supported: 0, 1 1024x768 (0x51c) 0.000MHz *current h: width 1024 start 0 end 0 total 1024 skew 0 clock 0.00KHz v: height 768 start 0 end 0 total 768 clock 0.00Hz 1920x1080 (0x520) 148.500MHz +HSync +VSync h: width 1920 start 2048 end 2248 total 2576 skew 0 clock 57.65KHz v: height 1080 start 1083 end 1088 total 1120 clock 51.47Hz

Any help is appreciated.

BR,
r0ller

rvp

r0ller

Can you show logs?
- dmesg output after boot -vx to see any kernel errors on mode changes; and,
- /var/log/Xorg.0.log for the X server side.
Can we see the EDID? (the hex should be visible in the Xorg.0.log file.)

Did you try the reduced blanking timings as those post suggested?

cvt -v -r 1920 1080 60

r0ller

rvp
This is the dmesg after boot -vx:
https://drive.google.com/file/d/1u_GqhrH6fkBlBNoClk0HdQnq-LqTP3BD/view?usp=sharing

This is the Xorg log:
https://drive.google.com/file/d/1Ho9SJvcDQn6RznrYSKItSSwhxbZgd8tK/view?usp=sharing

I don't see any EDID in it though. I also tried reduced blanking timings as well but that did not help either. What I spotted now in Xorg log is suspicious though:
[ 26.662] (EE) open /dev/dri/card0: No such file or directory [ 26.662] (WW) Falling back to old probe method for modesetting [ 26.662] (EE) open /dev/dri/card0: No such file or directory [ 26.662] vesa: Refusing to run, UEFI booted [ 26.662] (WW) Falling back to old probe method for wsfb [ 26.662] (II) wsfb(1): using default device [ 26.662] (WW) VGA arbiter: cannot open kernel arbiter, no multi-card support [ 26.662] (EE) Screen 0 deleted because of no matching config section. [ 26.662] (II) UnloadModule: "modesetting"
I remember some time back on NetBSD 9 or 10 there were some who had to create those manually. Do you think the remedy to be the same this time?

BR,
r0ller

rvp

r0ller I don't see any EDID in it though.

You won't because you aren't running the DRMKMS driver. You just have genfb the unaccelerated kernel framebuffer and the wsfb driver on the Xorg side:

dmesg output:

[     1.053637] genfb0 at pci9 dev 0 function 0: ATI Technologies Radeon R9 FURY / NANO Series (rev. 0xca)
[     1.053637] genfb0: recycling existing fb mapping at ffff8f04ad79c000
[     1.053637] genfb0: framebuffer at 0xe0000000, size 1024x768, depth 32, stride 4096
[     1.053637] genfb0: shadow framebuffer enabled, size 3072 KB
[     1.053637] wsdisplay0 at genfb0 kbdmux 1: console (default, vt100 emulation), using wskbd0
[     1.053637] wsmux1: connecting to wsdisplay0
[     1.053637] drm at genfb0 not configured

Xorg.0.log:

[    26.662] (II) UnloadModule: "modesetting"
[    26.662] (II) fboffset 0
[    26.662] (II) wsfb(0): Creating default Display subsection in Screen section
        "Builtin Default wsfb Screen 0" for depth/fbbpp 24/32
[    26.662] (==) wsfb(0): Depth 24, (==) framebuffer bpp 32
[    26.662] (II) fbi_flags: 0
[    26.662] (II) masks generated: 00ff0000 0000ff00 000000ff
[    26.662] (==) wsfb(0): RGB weight 888
[    26.662] (==) wsfb(0): Default visual is TrueColor
[    26.662] (==) wsfb(0): Using gamma correction (1.0, 1.0, 1.0)
[    26.662] (II) wsfb(0): Vidmem: 3072k
[    26.662] (==) wsfb(0): DPI set to (96, 96)
[    26.662] (==) wsfb(0): Using HW cursor
[    26.662] (**) wsfb(0): Using "Shadow Framebuffer"

wsfb doesn't support any mode-setting. (What's surprising is that 2 modes are being shown in your xrandr output, when I've only ever seen the one mode the system booted with---and this is set by the BIOS.)

r0ller It turns out that adding gop 0 [...]

Yes, that's the only way to set resolutions if you don't have working HW drivers.

You have a card which should be driven by the amdgpu DRMKMS driver. Since this driver isn't compiled into the GENERIC amd64 kernel (but, the modules do get installed), either

a) compile them into the kernel (preferred); or,
b) load the correct modules at boot-time in /boot.cfg

r0ller

It turns out that adding 'gop 0' (which is 1920x1080 when I do a 'gop list' at the boot prompt) to the /boot.cfg menu line like:
menu=Boot normally:gop 0; ... ;boot
makes it work.

r0ller

rvp

That's pretty much embarrassing. I did it quite some time back but could not recall what I need to do. On the other hand, it's also good news that wsfb got so much better in 11.0 that I thought (without looking into the logs) the amdgpu kicked in. It can flawlessly play fullscreen youtube videos in firefox unlike in 10.1. I'll try to find some time to build a custom kernel during the weekend. Sorry for being shallow and thanks for your suggestions!

BR,
r0ller

r0ller

I finally built a kernel with amdgpu enabled but the situation is the same as reported in PR58032 🙁