From owner-freebsd-current Sun Nov 12 09:17:49 1995 Return-Path: owner-current Received: (from root@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id JAA15737 for current-outgoing; Sun, 12 Nov 1995 09:17:49 -0800 Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id JAA15731 for ; Sun, 12 Nov 1995 09:17:45 -0800 Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id EAA17041; Mon, 13 Nov 1995 04:17:27 +1100 Date: Mon, 13 Nov 1995 04:17:27 +1100 From: Bruce Evans Message-Id: <199511121717.EAA17041@godzilla.zeta.org.au> To: current@freebsd.org, imb@scgt.oz.au Subject: Re: calcru: negative time: %qd usec Sender: owner-current@freebsd.org Precedence: bulk >What causes "calcru: negative time: %qd usec" to appear all over the >console. I'm running -current sup'd today on a 486DX2/66. I know it comes It is caused by the clock running backwards. `%qd' is supposed to show how much the clock is wrong by, but the kernel printf doesn't support quads. >from ../sys/kern/microtime.as but did something change to break it ? The message comes from kern_resource.c. The bogus time [difference] probably comes from a bug in microtime(). Perhaps the 8254 counter is not being read correctly, or a time daemon is fiddling with the clock. You can delete the message without losing much (first change it to print the value as 2 longs). The negative times are now replaced by 0. In 2.0.5R, a negative time of -1 usec gave a resource usage of 2^32-1 usec. I wonder if this was responsible for the "freezes". The scheduler should give a very low priority to a process that has consumed 2^32-1 usec of time resources in a few usec of real time :-). Bruce