On 2025-06-15 22:06:31 [-0400], Michael Schanne wrote: > Hi all, Hi, > I am trying to use the 6.12.28-rt10 patch in order to have both the i915 > driver and CONFIG_PREEMPT_RT enabled. I'm encountering an issue where my > application's GUI thread is pegging its cpu core. The event loop keeps > waking up although there are no events to process. A vanilla 6.12.28 > kernel does not exhibit this issue (both with PREEMPT_RT/no i915, or with > i915/no PREEMPT_RT). Do you use any realtime priorities here or is this all "normal" SCHED_OTHER processing with no extra tweaks? > Some things I have tried: > - Blacklist i915 module (corrupts display) > - Pass i915.enable_psr=0 and i915.enable_dc=0 boot parameters (no effect) > - Force software rendering with LIBGL_ALWAYS_SOFTWARE=1 (no effect) > > The target system is an Intel Skylake i5-6500TE CPU (I also tried a > Kabylake i5-7500T and it also shows this lockup). > > I have attached my kernel config for reference. > > Does anyone have any suggestions on how to debug this lockup? If nothing pops up in dmesg, could add CONFIG_DEBUG_ATOMIC_SLEEP and later maybe CONFIG_PROVE_LOCKING to see if the kernel complains? > Thanks, > Mike Schanne Sebastian
