matthew green a écrit :
i've just commited a workaround that should get rid of all the xcall() failed to ping messages. i've also commited a change to the way we handle interrupt evcnt(9) to be more per-cpu. please update and see if it helps this. you can monitor if this workaround has triggered by monitoring the kernel evcnt's that match "IPI mutex_trylock fail" (see: vmstat -e.)
I don't have time enough to test your patch until next saturday. I'll come back when I'll test it.
Regards,
JKB