From owner-freebsd-current Tue Sep 12 14:33:06 1995 Return-Path: current-owner Received: (from majordom@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id OAA00941 for current-outgoing; Tue, 12 Sep 1995 14:33:06 -0700 Received: from ns1.win.net (ns1.win.net [204.215.209.3]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id OAA00933 for ; Tue, 12 Sep 1995 14:33:04 -0700 Received: (from bugs@localhost) by ns1.win.net (8.6.11/8.6.9) id RAA21489 for current@freebsd.org; Tue, 12 Sep 1995 17:37:16 -0400 From: Mark Hittinger Message-Id: <199509122137.RAA21489@ns1.win.net> Subject: Re: Progress so far on the Sig-11 problem (fwd) To: current@freebsd.org Date: Tue, 12 Sep 1995 17:37:16 -0400 (EDT) X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 1889 Sender: current-owner@freebsd.org Precedence: bulk > > > > It seems that John Dyson said: > > > I am having problems reproducing the problem. It takes quite-a-while to > > > get it to happen, and only appears when I am running in less than 8MB. > > > > I have 32 MB and while sed used to die at startup, now xconsole and xload > > die when I log with xdm. emacs die pretty often too. > > > Well, I have come across something interesting :-). I can get -current > to Sig-11, but my "-current" copy on freefall won't. It seems that > changes were made that I had not tracked -- but the two versions > (my freefall copy, and a new, fresh one) have IDENTICAL vm stuff (plus > or minus prototyping changes, etc.) I am looking further.... > I'm playing with this also. Thanks to John for working on it. I got the problem to go away last week with a make world, but today's make world has brought the problem back. I am beginning to suspect that the problem is related to the way the system boots, or the order that the boot happens. The box that sig 11's has the older rc setup without the sysconfig. If I manually start things without the automatic rc I can get the system to sort of run ok. If I do the automatic startup then you can't even log in. I did the shared library ldconfig last. Hopefully I am not on another wild goose chase, but it looks like the sequence of bootstrap operations, and the kinds of things which are done have some effect on the sig 11 issue. *BTW this is a box that has the dma problem and must use bounce buffers* Could the clustering be broken on a box that needs bounce buffers? How about a wrap around condition at the boundary? (Just blue skying here) When I play with the order of starting things the sig 11's do not occur. I'm posting this in the hopes that it might give John that magic "oh yeah!!". I will play with this some more tommorrow. Regards, Mark Hittinger bugs@win.net