First question is, which version of the PREEMPT_RT patch did you use? For v5.15.7 you should be using the series released back in December of 2021: v5.15.7-rt23. It looks like things went bad when booting the RT kernel while initializing the cpuidle driver. I'm suspicious of the nested locking done in the genpd_* functions (and how they work with rtmutexes). If I get some time next week I'll see if I can make sense of the locking done there. Clark On Fri, Sep 30, 2022 at 3:44 PM Brian Wrenn <dcbrianw@xxxxxxxxx> wrote: > > Hi, > > I am building a kernel from source derived from Linaro 21.12, based on > linux kernel 5.15.7. It runs on a Variscite DART SD410 dev board. > When I build that kernel, it will boots successfully. When I build > the same kernel source with the PREEMPT-RT patch, it aborts during > boot. I attached console output for each case. > > An incompatibility between the Variscite patches and PREEMPT-RT could > cause this abort, but just to cover all my bases, I'm try to rule out > the PREEMPT-RT patch itself. > > The call trace includes RT specific routines, so I'm wondering if > there are any known problems with routines in this call state, and if > any fix for them exists in some way applicable to the v23 patch. > > Thank you for any information or leads. > > -Brian
