[drm:amdgpu_dm_atomic_commit_tail [amdgpu]] *ERROR* Waiting for fences timed out or interrupted! happens every time when a сutscene showed in Max Payne 3

Submitted by Mikhail Gavrilov @Mikhail

Assigned to Default DRI bug account

Link to original bug (#111124)

Description

Created attachment 144775
save game should be placed in ".steam/steam/steamapps/compatdata/204100/pfx/drive_c/users/steamuser/My Documents/Rockstar Games" directory

Here is a demonstration: https://youtu.be/5-peqm82_qI

Kernel: 5.3.0 commit 5450e8a316a6
Mesa: 19.2.0 commit f1e0a45d
libdrm: 2.4.99 commit 331e51e3
llvm: 8.0.0

Attachment 144775, "save game should be placed in ".steam/steam/steamapps/compatdata/204100/pfx/drive_c/users/steamuser/My Documents/Rockstar Games" directory":
Max_Payne_3.zip

added AMDgpu bugzilla labels

Mikhail Gavrilov @Mikhail uploaded an attachment:

Attachment 144776, "dmesg":
dmesg.txt

Mikhail Gavrilov @Mikhail uploaded an attachment:

Attachment 144777, "trace-cmd start -e dma_fence -e gpu_scheduler -e amdgpu -v -e "amdgpu:amdgpu_mm_rreg" -e "amdgpu:amdgpu_mm_wreg" -e "amdgpu:amdgpu_iv"":
event_dump.txt

Mikhail Gavrilov @Mikhail uploaded an attachment:

Attachment 144778, "./umr -O halt_waves -wa":
halt_waves.txt

Mikhail Gavrilov @Mikhail uploaded an attachment:

Attachment 144779, "./umr -R gfx[.]":
gfx.tar.xz

Mikhail Gavrilov @Mikhail uploaded an attachment:

Attachment 144780, "./umr -O many,bits -r ..mmGRBM_STATUS*":
mmGRBM_STATUS.txt

Mikhail Gavrilov @Mikhail uploaded an attachment:

Attachment 144781, "./umr -O many,bits -r ..mmCP_EOP_*":
mmCP_EOP.txt

Mikhail Gavrilov @Mikhail uploaded an attachment:

Attachment 144782, "./umr -O many,bits -r ..mmCP_PFP_HEADER_DUMP":
mmCP_PFP_HEADER_DUMP.txt

Mikhail Gavrilov @Mikhail uploaded an attachment:

Attachment 144783, "./umr -O many,bits -r ..mmCP_ME_HEADER_DUMP":
mmCP_ME_HEADER_DUMP.txt

Mikhail Gavrilov @Mikhail said:

GPU: AMD Radeon VII

Harald Linden said:

I can report the same problem running Kerbal Space Program (ksp) 1.4.5. Occasionally, during EVAs on planetary surfaces, the games freezes for ten seconds and the error message above shows up in the kernel log.

Distro: Debian GNU/Linux bullseye/sid
Kernel: 5.3.0-1-amd64 #1 (closed) SMP Debian 5.3.7-1 (2019-10-19) x86_64
Mesa: Mesa 19.2.3
libdrm: 2.4.99-1
llvm: 8.0.1-3+b1
GPU: "AMD Radeon (TM) RX 480 Graphics" (ChipID = 0x67df)

Harald Linden uploaded an attachment:

Attachment 145919, "dmesg (KSP)":
dmesg_filtered.txt

Harald Linden uploaded an attachment:

Attachment 145920, "(KSP) trace-cmd record -e dma_fence -e gpu_scheduler -e amdgpu -v -e "amdgpu:amdgpu_mm_rreg" -e "amdgpu:amdgpu_mm_wreg" -e "amdgpu:amdgpu_iv"":
trace.dat.xz

Harald Linden uploaded an attachment:

Attachment 145921, "(KSP) trace-cmd report":
report.txt.xz

Harald Linden uploaded an attachment:

Attachment 145922, "(KSP) Xorg.log":
Xorg_filtered.log

Im having similar issues with dolphin-emu on fedora. Distro: Fedora 31 Kernel: Linux scotty 5.3.11-300.fc31.x86_64 #1 (closed) SMP Tue Nov 12 19:08:07 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux Mesa: 19.2.6-1 libdrm: 2.4.100-1 llvm: 9.0.0-1 GPU: AMD Radeon (TM) RX 480 Graphicsgpu.log

This issue not reproduced with the latest LLVM/Mesa.
Tested on Mesa 20.0.0 git 11f736a and LLVM 10.0.0 git d9ae493
I am author original bug report, but due to loss of authorship during migration to gitlab, I can't close this bug report.

@hakzsam can you help me to close this bug report?

@Mikhail Does it also work with mesa-19.3.0-rc6 and LLVM 9?

I just checked on Mesa 19.3.0~rc6 + LLVM 9.0.0 and didn't see that my case from the first message is reproduced.

@Mikhail Thanks confirming. I guess we can close this then?

yes

closed

This still happens on 5.3.16 and 5.4.x on AMD Ryzen 5 3400G with Radeon Vega Graphics. I can quite easily reproduce this issue.

This still happens on 5.3.16 and 5.4.x

Which is Mesa version? Please retest it with the actual 19.3 version.

I can quite easily reproduce this issue.

If you could reproduce the issue with the latest Mesa version please fill a new bugreport with your reproduction case.

We use git mesa. We have #934

mentioned in issue #934