________________________________
De: Danilo Machado <[email protected]>
Enviado: quinta-feira, 26 de março de 2026 18:44
Para: Alex Deucher <[email protected]>
Assunto: RE: [REGRESSION][bisected] amdgpu (Tonga) HDMI no signal after suspend
Hi all,
Thanks again for your feedback.
I took another careful look at the bisect results and the system behavior, and
I’d like to share a more complete picture in case it helps narrow things down.
Even though the first bad commit points to a KVM merge, all commits inside that
range were consistently tested as GOOD. The issue only appears starting from
that merge point, which suggests this is not a direct regression in that code,
but rather something indirectly triggered (possibly timing or ordering during
resume).
What seems consistent across all tests:
* GPU: AMD Radeon R9 380 (Tonga, GCN 3)
* Kernel range tested: 6.3 (good) → 6.4 (bad)
* Issue reproducible
Behavior:
* After suspend/resume:
* HDMI output does not recover ("no signal")
* System may freeze under X11
* Wayland does not show the same hard failure
* Using s2idle instead of deep avoids the full system lockup, but the
graphical session may still return in a partially broken state
This seems to point more toward a problem in the display resume path rather
than the original bisected commit itself.
Based on the behavior differences:
* deep → full GPU/display reinit → hard failure
* s2idle → partial resume → no hard lock, but inconsistent display/session
state
This makes me suspect something around:
* DC state restore after resume
* HDMI link training
* EDID re-read
* atomic modeset state reconstruction
Possibly a latent issue being exposed by subtle timing/order changes introduced
during that merge window.
I completely understand that the bisected commit itself may not be directly
relevant, but the transition point seems reliable as a trigger.
If this line of investigation makes sense, I’d be glad to help further by:
* providing full bisect logs
* capturing detailed dmesg/journalctl before and after resume
* testing specific patches or debug options
* narrowing the range further if needed
I really appreciate the work you do on AMDGPU, and I’m happy to help within my
limits to get this issue understood.
Thanks again for your time.
Best regards,
Danilo
________________________________
De: Danilo Machado <[email protected]>
Enviado: terça-feira, 24 de março de 2026 16:10
Para: Alex Deucher <[email protected]>
Assunto: [REGRESSION][bisected] amdgpu (Tonga) HDMI no signal after suspend
Hi,
Just following up on this report.
I completed a git bisect and identified the first bad commit:
danilo@danilo-B450-AORUS-PRO-WIFI:~/kernel-bisect/linux$ git bisect good
b3c98052d46948a8d65d2778c7f306ff38366aac is the first bad commit
commit b3c98052d46948a8d65d2778c7f306ff38366aac
Merge: 4a5fd419952e 4984563823f0
Author: Paolo Bonzini <[email protected]>
Date: Wed Apr 26 15:56:51 2023 -0400
Merge tag 'kvm-x86-vmx-6.4' of https://github.com/kvm-x86/linux into HEAD
KVM VMX changes for 6.4:
- Fix a bug in emulation of ENCLS in compatibility mode
- Allow emulation of NOP and PAUSE for L2
- Misc cleanups
arch/x86/kvm/vmx/nested.c | 112 ++++++++++++++++++++++++++++++----------------
arch/x86/kvm/vmx/sgx.c | 4 +-
arch/x86/kvm/vmx/vmx.c | 15 +++++++
3 files changed, 91 insertions(+), 40 deletions(-)
danilo@danilo-B450-AORUS-PRO-WIFI:~/kernel-bisect/linux$
________________________________________
De: Danilo Machado <[email protected]>
Enviado: quinta-feira, 19 de março de 2026 11:18
Para: Alex Deucher <[email protected]>
Assunto: RE: AMDGPU Tonga suspend regression identified between kernel 6.3 and
6.4
Hello Alex,
I hope you're doing well.
First
of all, thank you for your guidance and encouragement regarding the
bisect process — it was very helpful and allowed me to narrow down the
issue precisely.
I
would like to report a regression affecting my GPU, an AMD Radeon R9
380 (Tonga), where HDMI output stops working after suspend/resume. The
issue is fully reproducible and was not present in Linux 6.3.
________________________________________
🔍 Summary
GPU: AMD Radeon R9 380 (Tonga)
CPU: AMD Ryzen 5 5500
RAM: 16 GB
Display: HDMI
Driver: amdgpu (RADV Tonga)
Display server behavior:
X11 → ❌ HDMI “no signal” after resume
Wayland → ✅ Works normally
Kernel status:
Linux 6.3 → ✅ Working correctly
Linux 6.4+ → ❌ Regression present (X11 only)
________________________________________
⚠️ Problem Description
After
resuming from suspend under X11, the system continues running normally,
but the HDMI display loses signal completely ("no signal").
The issue is 100% reproducible and requires a reboot to restore video output.
This problem does not occur under Wayland, which strongly suggests an issue in
the X11/DRM interaction path or display reinitialization specific to that stack.
________________________________________
🧪 Bisect Result
A full git bisect was performed between Linux 6.3 (good) and Linux 6.4 (bad).
Result:
First bad commit:
b3c98052d46948a8d65d2778c7f306ff38366aac
Merge tag 'kvm-x86-vmx-6.4'
Last known good:
4a5fd419952e49ef870e9259d815718062871cc1
________________________________________
🧠 Additional Notes
All intermediate commits tested during bisect (mostly KVM-related) were marked
as GOOD.
The regression appears to be introduced indirectly via this merge rather than a
direct amdgpu/DRM change.
The behavior strongly suggests a failure in display reinitialization or EDID
handling after resume, specifically affecting X11.
Full bisect log available upon request.
________________________________________
💡 Real-world relevance
Despite being an older GPU, the R9 380 (Tonga) remains a capable and widely
used card.
As
a practical example, this GPU is still capable of running modern titles
(e.g., Resident Evil Requiem via Proton), demonstrating that continued
stability and support improvements remain highly relevant.
Given
the high cost of newer GPUs, maintaining reliability and fixing
regressions on this class of hardware is very important for users.
________________________________________
📎 Reproduction Steps
Boot system with HDMI connected
Log into an X11 session
Suspend the system
Resume
HDMI signal is lost ("no signal")
________________________________________
🙏 Closing
I hope this bisect result helps narrow down the issue.
Please let me know if I can assist further with testing or provide logs — I
would be happy to help in any way possible.
Thank you again for your support and for your work on AMD GPU support.
Best regards,
Danilo Machado
This regression affects:
* AMD Radeon R9 380 (Tonga)
* HDMI output (no signal after resume)
* System freeze under X11
Important:
* Issue does NOT occur under Wayland
Please let me know if further testing or logs are needed.
Thanks
