Bug 110713

Summary: A Hat in Time: GPU hang with gallium nine
Product: Mesa Reporter: Franc[e]sco <lolisamurai>
Component: Drivers/Gallium/radeonsiAssignee: Default DRI bug account <dri-devel>
Status: RESOLVED MOVED QA Contact: Default DRI bug account <dri-devel>
Severity: normal    
Priority: medium CC: kiwii
Version: git   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:

Description Franc[e]sco 2019-05-19 20:06:49 UTC 
when using gallium nine, A Hat in Time locks up the entire system (requires a hard reboot) on certain scenes.

I'm running a r9 270x with the amdgpu driver and mesa-git 110950.fc455797c1b-1 + llvm-svn 360675-1

this issue was discussed with the gallium nine guys here: https://github.com/iXit/wine-nine-standalone/issues/35 and they narrowed it down to radeonsi

using AMD_DEBUG=mono fixes the issue

this is the apitrace for gallium nine https://mega.nz/#!ZJhRhYaR!cVsMkRYkl6kPl0f43sMKu8NI4pvA4eE85cbLJyoQ-jE

the hang happens on the very last part of the trace, when the green goo in the middle of the room is on screen. the area is chapter3, act6 and all other challenges that take place there

playing back the trace also causes the same hang.

playing back the trace with AMD_DEBUG=mono works correctly without any hangs

other users report that turning off dynamic lights also fixes the issue

another user reported a similar hang in chapter 1, act 4 with this apitrace https://kiwii.casa/downloads/d3d9.trace.xz
Comment 1 kiwii 2019-05-20 17:03:22 UTC 
I'd like to add, besides the R9 270X, this also affects Vega (both traces hang my Vega 64) and Polaris (one user reports the chapter 1, act 4 hang to be 100% reproducible on his RX 580).

I can also confirm both traces to play fine on my card with AMD_DEBUG=mono on mesa-git and R600_DEBUG=mono on mesa 19.0.4.

With the NIR backend (R600_DEBUG=nir) it still hangs the same way as with default TGSI on both 19.0.4 and git.

Note: Both traces are recorded on x64 and may need to be replayed with x64 apitrace to prevent running out of memory during shader compilation.
Comment 2 GitLab Migration User 2019-09-25 18:50:02 UTC 
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/1408.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.