Dell r730xd with PERC H730 hangs under some load (depending on the workload pattern).
Machine runs recent -CURRENT, zroot is mirror pool.
When doing buildworld with -j16, the lockup occurs and the only way out is to reboot, but also ddb can be entered.
Using ddb and "alltrace" I've been able to trace that most of the processes are in sched_switch() and mrsas_ocr0 kernel process sleeps in mrsas_ocr_thread() (mrsas_ocr_thread+0x3d0).
Using "examine/i" from within ddb i've been able to trace that mrsas_ocr_thread+0x3d0 points somewhere into the function (mrsas_ocr_thread()) to a place inside for(;;) loop.
*SOMETIMES* after certain amount of time, the machine runs for minute or two, then the lockup occurs again.
What else can I try? Here my kernel-fu ends :-(
I'm not sure whether this is caused by scheduler problem(s), interrupt(s) problem(s) or mrsas driver as such.
Any help is appreciated. I repeat again that if there is a need for a debug session, we can arrange one.
You can get a crash dump from ddb prompt.
Then the problem would be easier to analyze.
Were there any significant changes recently? The problem seems to be gone now.
I can see that there were some changes to threading and/or scheduler.