On Tue, Mar 10, 2026 at 4:34 AM Danilo Machado
<[email protected]> wrote:
>
> Hello,
> I am reporting a regression affecting my AMD Radeon R9 380 (Tonga / GCN 1.2) 
> related to HDMI output after system suspend. I am sharing detailed 
> observations in case they help identify the cause.
> This issue has also been reported in the Ubuntu bug tracker:
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2142389
> ________________________________
> System information
> GPU: AMD Radeon R9 380 (Tonga / GCN 1.2)
> CPU: AMD Ryzen 5 5500
> Motherboard: Gigabyte B450 AORUS PRO WIFI
> Connection: HDMI directly to monitor
> Distribution: Zorin OS 18 (Ubuntu-based)
> Desktop environments tested: GNOME (Wayland and X11 sessions)
> ________________________________
> Summary of the issue
> Since kernel 6.13, my system occasionally resumes from suspend with no HDMI 
> signal.
> The system itself resumes normally, but the monitor remains black and reports 
> no input signal.
> This behavior appears related to a failure in EDID detection or HDMI link 
> reinitialization after suspend.
> Kernel 6.12 and earlier do not show this behavior.
> ________________________________
> Kernel behavior comparison
> Kernel 6.12.x
> Stable operation
> Suspend and resume work correctly
> HDMI output always restored
> Kernel 6.13.x and newer
> Regression observed
> After suspend, the system resumes but HDMI signal is sometimes not restored

Can you bisect?  If 6.12 works and 6.13 is broken, you should be able
to bisect between them to identify the bad commit.

https://docs.kernel.org/admin-guide/bug-bisect.html

Alex
> ________________________________
> Relevant kernel log excerpt
> During resume when the issue occurs, the following messages appear in dmesg:
> [drm] ERROR HDMI-A-1: probed a monitor but no|invalid EDID
> amdgpu 0000:0b:00.0: amdgpu: failed to read EDID from connector HDMI-A-1
> amdgpu 0000:0b:00.0: amdgpu: link training failed during resume
> These messages suggest that EDID detection or HDMI link reinitialization may 
> be failing during the resume sequence.
> The complete logs and additional system information are available in the 
> Launchpad bug report linked above.
> ________________________________
> Wayland vs X11 behavior
> I also tested different display sessions to help isolate the issue.
> Wayland session:
> Suspend and resume work reliably in my tests. The monitor wakes normally and 
> HDMI output is restored.
> X11 session:
> Short suspend cycles usually resume correctly. However, after longer suspend 
> periods (approximately 10–15 minutes or more), the system resumes but the 
> monitor receives no signal.
> The system itself continues running normally in the background.
> This suggests that the issue may involve display reinitialization during 
> deeper power-state transitions or differences in display management between 
> Wayland compositors and the X11 stack.
> ________________________________
> Troubleshooting attempts
> Before reporting this bug, I performed several tests and troubleshooting 
> steps:
> • Tested multiple kernel versions
> • Confirmed stable behavior with kernel 6.12
> • Reproduced the issue with newer kernels
> • Tested both Wayland and X11 sessions
> • Verified HDMI cable and monitor behavior
> • Compared suspend durations to identify patterns
> • Collected logs and monitored dmesg output
> Because kernel 6.12 remains stable on this system, I have temporarily stayed 
> on the LTS kernel (6.8 series) to maintain reliability.
> ________________________________
> Background
> I was previously a long-time Windows user (from Windows XP through Windows 
> 11), but I recently migrated fully to Linux and do not plan to return to 
> Windows.
> My goal is to maintain a stable Linux system and eventually upgrade to newer 
> kernels once this regression is fully resolved.
> I appreciate the work of the kernel and AMDGPU developers and hope this 
> report helps identify the issue.
> If additional logs, kernel parameters, or test kernels would be helpful, I 
> would be glad to assist with further testing.
> Best regards,
> Danilo Machado
>

Reply via email to