Date: Mon, 06 Apr 1998 12:03:43 +0800 From: Peter Wemm <peter@netplex.com.au> To: Peter Dufault <dufault@hda.com> Cc: smp@FreeBSD.ORG Subject: Re: beware! -current under SMP is "not looking good". Message-ID: <199804060403.MAA05339@spinner.netplex.com.au> In-Reply-To: Your message of "Sun, 05 Apr 1998 19:26:07 -0400." <199804052326.TAA24641@hda.hda.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Peter Dufault wrote: > > It seems that doing a 'setpriority()' is the last thing that happens > > before the hang with ntpdate. I wonder if the posix4 stuff is having an > > effect here? > > The latest round of changes were minor. The "maybe_resched" that went > in near the beginning of March would be more likely to cause problems - > that was what broke idle priority. I think the setitimer() problems turned out to be the culprit in this case, I should point out what I was concerned about.. maybe_resched() looks at the new process compared to the currently executing process on the current cpu. It looks like if the new process has a higher priority, then a reschedule is called. However, it doesn't check the other cpus.. Suppose the current cpu was be running a high priority process, and the "other" cpu could be running a max-nice process. If a new normal priority process was being activated, then a reschedule should be done so that the "other" cpu can pick up the new process. Otherwise, the intermediate priority process hangs around until the next roundrobin or whatever, and the max-nice process keeps on running for the rest of the period. However, this isn't what I'd call an earth-shattering problem. Anyway, things are looking a lot happier in -current at the moment... Cheers, -Peter -- Peter Wemm <peter@netplex.com.au> Netplex Consulting To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199804060403.MAA05339>