[Bug 221202] New: DisplayPort monitor (Samsung Odyssey G65B) sometimes wakes with invalid EDID and only exposes 640x480 until monitor power-cycle (NVIDIA + Wayland)

[bugzilla-daemon@kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=221202

            Bug ID: 221202
           Summary: DisplayPort monitor (Samsung Odyssey G65B) sometimes
                    wakes with invalid EDID and only exposes 640x480 until
                    monitor power-cycle (NVIDIA + Wayland)
           Product: Drivers
           Version: 2.5
          Hardware: All
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P3
         Component: Video(DRI - non Intel)
          Assignee: drivers_video-dri@kernel-bugs.osdl.org
          Reporter: uitham@gmail.com
        Regression: No

Created attachment 309601
  --> https://bugzilla.kernel.org/attachment.cgi?id=309601&action=edit
Archive containing .txt outputs of commands

Kernel: Linux 6.18.9-arch1-2
GPU: NVIDIA RTX 5080
Driver: nvidia-open-dkms 590.48.01-6 (nvidia-drm.modeset=1)
Desktop: KDE Plasma (Wayland)
Connection: DisplayPort
Monitor: Samsung Odyssey G65B

Problem:

After the system or monitor wakes from sleep, the Odyssey G65B sometimes
appears connected but exposes only a single fallback mode (640x480). The
monitor remains detected but its EDID appears unreadable or corrupted until the
monitor is power-cycled.

Power-cycling the monitor immediately restores the correct EDID and full mode
list without rebooting the system or reconnecting cables.

Observed behavior:

kscreen-doctor shows DP-3 connected but only a single mode:

Output: DP-3
Modes: 640x480@59.94*

xrandr --verbose shows only fallback modes:

DP-3 connected 640x480
640x480 *current +preferred
320x240
640x400
320x200
640x350

Kernel / userspace logs show EDID read failures:

Checking EDID failed after 2 tries
Unable to read EDID for /dev/i2c-20
Failed to find connector name for /dev/i2c-20

KWin also reports invalid EDID colorimetry:
EDID colorimetry xy(...) is invalid

EDID comparison:
When the monitor is working normally, the EDID decodes correctly and identifies
the display as:

Monitor name: Odyssey G65B

The EDID includes valid CTA extension blocks and detailed timings including:

- 2560x1440 @ ~60 Hz
- 1920x1080 @ 120 Hz
- HDR metadata

When the problem occurs, the monitor appears to return an invalid or truncated
EDID and only legacy fallback modes are exposed.

Reproduction pattern (intermittent):

1. System runs normally with monitor connected over DisplayPort
2. Monitor enters sleep
3. Wake the monitor/system
4. Occasionally the display returns with only 640x480 available
5. Power-cycle the monitor
6. Correct EDID and full mode list return immediately

Notes:

The connector remains detected during the failure (DP-3 shows as connected),
but EDID reads fail on the corresponding I²C bus (/dev/i2c-20). This suggests a
wake/reprobe issue in the DisplayPort EDID detection path rather than a
configuration issue in user space.

This system uses the NVIDIA open kernel modules (nvidia-open-dkms), which
provide the nvidia-drm DRM/KMS driver. The issue may therefore be within the
nvidia-drm driver stack. I am reporting it here in case the problem involves
generic DRM / DisplayPort EDID detection or wake-up handling.

Attachments:

- kscreen-doctor output (failed state)
- xrandr --verbose (failed state)
- journalctl EDID errors
- EDID decode (failed vs working state)

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching the assignee of the bug.