> So you are not using clock_nanosleep() but something like a POSIX timer? No, I was using timerfd_create() and then calling read() in a loop in my timer thread. Is it recommended to use clock_nanosleep() instead? -Mike On Wed, Jun 18, 2025 at 8:16 AM Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx> wrote: > > On 2025-06-18 07:34:25 [-0400], Michael Schanne wrote: > > I found my issue, I am using several SCHED_FIFO threads, one was spinning > > waiting on a timer to update, another thread which was supposed to update > > the timer count is higher priority, but with the RT patch applied there is > > also a ktimers thread set to RT priority 1, so it was never getting a > > chance to run and my timer would never fire. I adjusted the ktimers thread > > to a higher priority and things started working again. This thread does > > not exist in the vanilla kernel even with CONFIG_PREEMPT_RT enabled, so > > that is why I wasn't seeing it there. The i915 driver was a red herring. > > Okay not the i915. > The ktimers thread wakes only SCHED_OTHER tasks. The SCHED_FIFO/RR > threads should be woken up directly (without the need of the ktimers > thread). > So you are not using clock_nanosleep() but something like a POSIX timer? > > > Thanks, > > Mike > > Sebastian
