Skip to content

Radeon R5 Graphics, Radeon HD 7340 and others emit spurious HDMI unplug events

Brief summary of the problem:

After login on Chrome OS, an HDMI unplug event fires even though no unplug occurs. cat /proc/asound/card0/eld#0.0 shows the device as unplugged, though it was plugged prior to completing login. Notably Chrome OS more or less does a full graphics restart upon login, and though I'm not sure I suspect that's why I haven't been able to reproduce the issue so far on other distros. The issue is present on several radeon machines; the two I've done the most testing on are detailed below.

I noticed removing the call to radeon_audio_component_init() outright resolved the issue for me:

diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index afbb3a80c0c6..7913995486b8 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -1453,7 +1453,7 @@ int radeon_device_init(struct radeon_device *rdev,
                        goto failed;
        }
 
-       radeon_audio_component_init(rdev);
+       //radeon_audio_component_init(rdev);
 
        r = radeon_ib_ring_tests(rdev);
        if (r)
@@ -1517,7 +1517,7 @@ void radeon_device_fini(struct radeon_device *rdev)
        rdev->shutdown = true;
        /* evict vram memory */
        radeon_bo_evict_vram(rdev);
-       radeon_audio_component_fini(rdev);
+       //radeon_audio_component_fini(rdev);
        radeon_fini(rdev);
        if (!pci_is_thunderbolt_attached(rdev->pdev))
                vga_switcheroo_unregister_client(rdev->pdev);

Hardware description:

  • CPU: AMD E2-2000 APU with Radeon(tm) HD Graphics, also AMD A8-7100 Radeon R5
  • GPU: Radeon HD 7340, also Radeon R5 Graphics
  • System Memory: 8G, also 4G
  • Display(s): 1080p HDMI and internal laptop display for both
  • Type of Display Connection: HDMI for both

System information:

How to reproduce the issue:

The issue appears to occur when restarting a graphics session with an external HDMI monitor. The affected machines are both laptops.

Proposed Kernel change

I have a proposed change (not the above code) based on #1569 (comment 1603002) .

See https://lore.kernel.org/all/20241115-hdmi-audio-radeon-connectors-v1-1-d6d66d3128dd@chromium.org

Edited by Steve Kendall