1. Forum
  2. Usenet
  3. LINUX.DEBIAN.KERNEL

  • Re: Bug#1108598: amdgpu: incorrectly redrawn screen after resuming from

    From Uwe =?utf-8?Q?Kleine-K=C3=B6nig?=@1:229/2 to Ivan Krylov on Wed Jul 2 18:20:01 2025
    XPost: linux.debian.bugs.dist
    From: u.kleine-koenig@baylibre.com

    tag -1 + moreinfo

    Hello,

    On Tue, Jul 01, 2025 at 10:21:22AM +0300, Ivan Krylov wrote:
    Package: src:linux
    Version: 6.1.140-1
    Severity: important
    Tags: upstream
    X-Debbugs-Cc: ikrylov@disroot.org

    Dear Maintainer,

    After upgrading from linux-image-6.1.0-34-amd64 to linux-image-6.1.0-37-amd64, the system no longer correctly resumes from hibernation. When the resume process completes, the screen freezes in
    the VGA text mode (the cursor stops blinking). At this point,
    Ctrl+Alt+F1 can be used to switch to the text console (and cause the
    video mode to switch to the native resolution), and after that, Alt+F7
    can bring me back to X, but in both cases, only the upper-left corner
    of the screen is drawn. The rest of the screen is filled with the
    contents of the edges of that area.

    The amdgpu WARN_ON_ONCE is new; the libata WARNING has been there for a
    long time without any apparent consequences.

    [...]
    [ 71.485255] amdgpu 0000:06:00.0: drm_WARN_ON_ONCE(!dev->mode_config.poll_enabled)

    I would expect the libata warning to be unrelated. And while we had a
    bunch of bug reports about amdgpu in the nearer past, this warning is
    new to me.

    Can you reliably trigger that problem?

    Can you please test a 6.12 kernel (from testing or backports) and/or
    6.15 (from experimental)? Depending on if these work better or not, we
    need to report upstram or check which patches to backport to 6.1.x (or
    convince you to switch to the newer kernel :-)

    Best regards
    Uwe

    -----BEGIN PGP SIGNATURE-----

    iQEzBAABCgAdFiEEP4GsaTp6HlmJrf7Tj4D7WH0S/k4FAmhlWuAACgkQj4D7WH0S /k79TAf/fcA3rNDqNciTQeMNtP4z/zbYZZgXgdU4vCw1H2cpN/xQHtiNvbx5S1z9 gFWdiMjv5GBknj/VpsD3OFKygDplQBhoXiua4SXLeruj7+2XlARM7qdG0khAZRwX osU79Tqx9zG86UTTw8ylGHj9Id3rch9kVqCzWwi7M/+bUzbdhZH112ZAEkGJ2/eQ gE04azXk5IcAsgULQtnBXctV31ZIIuYiKq4NtW8vCnt+GpNEfvQtZLMv4osnvsc6 PbC9SBDF24EIP9W/r2NEppzOQP6VR+WYLVcgXosi3xcdS+jP3fXikVsH8ZwLBFeK y2jyEj2OLlNHS2kjoTAjAbxVwVR/ow==
    =4TuR
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From Ivan Krylov@1:229/2 to u.kleine-koenig@baylibre.com on Fri Jul 4 00:00:01 2025
    XPost: linux.debian.bugs.dist
    From: ikrylov@disroot.org

    Hello Uwe and thank you for the prompt response!

    On Wed, 2 Jul 2025 18:14:27 +0200
    Uwe Kleine-König <u.kleine-koenig@baylibre.com> wrote:

    Can you reliably trigger that problem?

    The issue happened to me 4 times out of 4 times I resumed 6.1.0-37 from hibernation.

    Can you please test a 6.12 kernel (from testing or backports) and/or
    6.15 (from experimental)?

    6.12.32+bpo just resumed without a problem after echo reboot >
    /sys/power/disk; echo disk > /sys/power/state. The drm_WARN_ON_ONCE(!dev->mode_config.poll_enabled) message is gone too.

    Bug#1107579 does indeed look similar. Since it seems to work for me, I
    wouldn't mind running a kernel from backports as a workaround, but if
    you need more information from me to help narrow it down, I can try to
    gather it.

    --
    Best regards,
    Ivan

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)