No monitor detected on RX 570 with kernel 6.4

I am running Fedora 37 on a system with an RX 570 card, connected via DisplayPort to a Dell U3223QE monitor. When I upgrade to any Fedora 6.4 kernel (tried 6.4.4-100.fc37, 6.4.6-100.fc37, 6.4.7-100.fc37, and 6.4.8-100.fc37), the monitor goes to sleep for no input detected as soon as the kernel switches from the bootup text console to the framebuffer (when removing "rhgb quiet" from boot) or when X starts (otherwise). I don't see anything in the kernel log that looks different from a working (kernel 6.3.x) boot, but with 6.4, I do see this in the Xorg.0.log:

[ 7.620] (EE) AMDGPU(0): failed to set mode: Invalid argument [ 7.620] (WW) AMDGPU(0): Failed to set mode on CRTC 0 [ 7.620] (EE) AMDGPU(0): Failed to enable any CRTC

It seems like with 6.4, the monitor is no longer detected, and it is a problem in the kernel (since framebuffer doesn't work, not just X).

System info: $ inxi -GxxS System: Host: kosh.cmadams.net Kernel: 6.3.12-100.fc37.x86_64 arch: x86_64 bits: 64 compiler: gcc v: 2.38-27.fc37 Desktop: MATE v: 1.26.1 wm: marco dm: LightDM Distro: Fedora release 37 (Thirty Seven) Graphics: Device-1: AMD Ellesmere [Radeon RX 470/480/570/570X/580/580X/590] vendor: Sapphire driver: amdgpu v: kernel arch: GCN-4 pcie: speed: 8 GT/s lanes: 16 ports: active: DP-3 empty: DP-1, DP-2, DP-4, DVI-D-1, HDMI-A-1, HDMI-A-2 bus-ID: 0a:00.0 chip-ID: 1002:67df temp: 47.0 C Display: x11 server: X.Org v: 1.20.14 compositor: marco v: 1.26.2 driver: X: loaded: amdgpu unloaded: fbdev,modesetting,vesa dri: radeonsi gpu: amdgpu display-ID: :0 screens: 1 Screen-1: 0 s-res: 3840x2160 s-dpi: 96 Monitor-1: DP-3 mapped: DisplayPort-2 model: Dell U3223QE res: 3840x2160 dpi: 140 diag: 801mm (31.5") API: OpenGL v: 4.6 Mesa 23.0.3 renderer: AMD Radeon RX 570 Series (polaris10 LLVM 15.0.7 DRM 3.52 6.3.12-100.fc37.x86_64) direct-render: Yes

added DC label

added Polaris label

Please attach your dmesg output and xorg log.

Here you go - let me know if there's anything else I can provide.

dmesgXorg.0.log

Given it's polaris, and it looks like your display was using MST, can you try this: https://patchwork.freedesktop.org/patch/550066/

I added that patch and rebuilt the Fedora 6.4.8-100.fc37 kernel, got the same result. In case it helps, here's logs from 6.3.12 (works) and also 6.4.8-local (with patch, doesn't work):

dmesg-6.3.12Xorg.0.log-6.3.12 dmesg-6.4.8-localXorg.0.log-6.4.8-local

OK. If it's not that; would you be able to bisect it?

I'll see what I can do - have to admit, it's been probably 20 years since I regularly built kernels rather than using vendor kernels. Might be this weekend before I get to it.

I bisected it, although I'm not sure if the commit make sense as an error:

Also, I was getting boots with no video and no network (just had to hit RESET), so I set up a USB serial console and started over. Every boot that didn't work had an identical oops, like the one attached.dmesg-oops

Ignore my messed-up bisect... I just tried Fedora's 6.4.9 kernel plus the patch you posted, and it works (while Fedora 6.4.9-100.fc37 did not work). I must have also missed getting the patch applied before. One of those weeks...

Thanks. Will that patch be added in a 6.4.x kernel?

marked this issue as related to #2671 (closed)

Yes; it will land in 6.5-rc6 and backport to 6.4.y after that. If you want it sooner you can ask Fedora guys to pick it up in a Fedora bug.

changed milestone to %linux-6.4.y

mentioned in issue #2795 (closed)

It's landed in 6.4.11.

closed

marked this issue as a duplicate of #2671 (closed)

marked #2795 (closed) as a duplicate of this issue

marked this issue as related to #2795 (closed)