Ahh... seems the culprit is softclock_thread added 2016/09/22 (kern/kern_timeout.c mpi@).
I guess next question is whether OpenBSD will live (and document) this change (increase) in load average, or some sort of counter measure will be taken in uvm_loadav() to suppress the extra bump in the value? Cheers, --patrick On Fri, Nov 11, 2016 at 01:23:26PM -0800, patrick keshishian wrote: > Hi, > > For a long time I was running an amd64 snapshot from 2016/03/30, > then I installed a snapshot from 2016/10/09 on four machines > (two laptops, a desktop and a 1U supermicro). > > One apparent difference was that the load average on all four > machines idled at slightly above 1 (1.12, 1.18, ...), about 1 > more than previous snapshots (up and including March 30th). > > Initially I assumed maybe something in the snapshot(s), which > would eventually go away with future updates. > > I built a couple of kernels from source since 2016/10/09 snapshot > and each still runs at load avg of 1 more than expected when the > box is completely idle. > > I briefly looked at cvs history (cvs log -d '2016/03/30<') and > failed to see any obvious change that would attribute to this > apparent display/reporting "glitch?" > > Can someone enlighten me on this issue? > > Thanks in advance, > > --patrick > > > > > Index: kern_pledge.c > =================================================================== > RCS file: /cvs/obsd/src/sys/kern/kern_pledge.c,v > retrieving revision 1.187 > diff -u -p -u -p -r1.187 kern_pledge.c > --- kern_pledge.c 27 Oct 2016 10:48:25 -0000 1.187 > +++ kern_pledge.c 11 Nov 2016 20:56:21 -0000 > @@ -1046,7 +1046,7 @@ pledge_sysctl(struct proc *p, int miblen > if (miblen == 2 && /* hw.ncpu */ > mib[0] == CTL_HW && mib[1] == HW_NCPU) > return (0); > - if (miblen == 2 && /* kern.loadavg / getloadavg(3) */ > + if (miblen == 2 && /* vm.loadavg / getloadavg(3) */ > mib[0] == CTL_VM && mib[1] == VM_LOADAVG) > return (0);
