Hi Philippe,
while playing with maxcpus on x86, I ran into one issue again that I think I saw
already but forgot to dig into back then: When an RTDM driver happens to create
a kernel task (rtdm_task_init) on a support outside of xenomai.supported_cpus,
we will crash. That may happen because that command is run over module_init, or
any other context that is not a Xenomai-blessed Linux task with therefore
constraint affinity mask.
Any suggestions how to resolve that best? Restrict the affinity of the
rtdm_task_init caller? Permanently or maybe just temporarily? Or is there some
way to resolve that deeper inside xnthread_* APIs?
Jan
--
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux