From owner-freebsd-smp Sun Apr 20 19:13:33 1997 Return-Path: Received: (from root@localhost) by freefall.freebsd.org (8.8.5/8.8.5) id TAA06802 for smp-outgoing; Sun, 20 Apr 1997 19:13:33 -0700 (PDT) Received: from kvikk.uit.no (kvikk.Uit.No [129.242.4.32]) by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id TAA06793 for ; Sun, 20 Apr 1997 19:13:29 -0700 (PDT) Received: from sprint.cc.uit.no (sprint.Cc.Uit.No [129.242.5.198]) by kvikk.uit.no (8.8.5/8.8.5) with ESMTP id EAA08056; Mon, 21 Apr 1997 04:13:23 +0200 (MDT) Received: from slibo.cc.uit.no (slibo.Cc.Uit.No [129.242.5.36]) by sprint.cc.uit.no (8.8.5/8.8.5) with ESMTP id EAA08669; Mon, 21 Apr 1997 04:13:22 +0200 (METDST) Received: from localhost (terjem@localhost) by slibo.cc.uit.no (8.8.5/8.8.5) with ESMTP id EAA03641; Mon, 21 Apr 1997 04:13:21 +0200 (METDST) Message-Id: <199704210213.EAA03641@slibo.cc.uit.no> X-Mailer: exmh version 2.0gamma 1/27/96 To: Steve Passe cc: freebsd-smp@freefall.freebsd.org Subject: Re: Current runs fine on 4 CPU's as well In-reply-to: Your message of "Sun, 20 Apr 1997 19:28:09 METDST." <199704210128.TAA08535@Ilsa.StevesCafe.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Mon, 21 Apr 1997 04:13:21 +0200 From: Terje Normann Marthinussen Sender: owner-smp@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk >Hi, > >> Just sup'ed current, made the world and a new SMP kernel (suped 1 hour ago). >> ... >> It's been running on 4 CPU's now since december, and I've only had two >> crashes since then. It's not heavily loaded, but when it once in a while >> is loaded, it quite loaded chewing data that needs 50-100MB of memory >> and sometimes causing heavy paging as well (got 128MB on it). > >Cyrus recently identified a dead-lock in the APIC_IO version. The symptom >is a complete freeze of the system while doing heavy IO. Yes, I've followed the discussion. However, I've never seen the symptom, and it has made me wonder, probably just luck. In most cases the things this machine have done so far while running FreeBSD, has been tuned to fit into memory. However, I had one case while running the last pre-lite code, where I totally miscalculated memory need for an application here. I should run 3 instances in parallell and the code frequently traversed some large arrays. I thought all three processes should run nice in 90MB, however, I did something very stupid so each process used 60MB instead of 30. Had it trashing like mad on two 7200rpm FW HP disks while I was away for 4 days. It didn't manage to do much work due to the trashing, but it didn't crash either. On the other side, this machine is standing in a server room. And since it hasn't yet been put to the work it should do, it seldom has to deal with more than a couple of shell users, and only does some chewing when we don't have spare disk/cpu other more powerfull boxes. So, I'll have to admit that the usage it's probably not the best to find bugs. >Are you using the APIC_IO version? Used APIC_IO since before christmas. >When you say "crash", do you mean spontanious reboot, panic, or freeze? One reboot during the night and one panic that I never cared to follow up. Might have been other reboots, but at least I haven't seen them. Well, I should admit, there is one more, it was when I tested the famous FP exception problem. But running code that one knows will crash the system hardly counts in this case. >Earlier today I commited a "bandaid" that should help with this problem. >Did you get the version of mpapic.h that has "#define APICIPI_BANDAID" in it? Yes. Terje Marthinussen terjem@cc.uit.no