From owner-freebsd-hackers Tue Jul 11 15:29:01 1995 Return-Path: hackers-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id PAA07730 for hackers-outgoing; Tue, 11 Jul 1995 15:29:01 -0700 Received: from giant.mindlink.net (root@giant.mindlink.net [204.174.18.2]) by freefall.cdrom.com (8.6.10/8.6.6) with SMTP id PAA07724 for ; Tue, 11 Jul 1995 15:28:56 -0700 Received: by giant.mindlink.net (Smail3.1.28.1 #5) id m0sVnnD-0001ftC; Tue, 11 Jul 95 15:28 PDT Message-Id: From: a00776@giant.mindlink.net (Toomas Losin) Subject: RAM sizing doesn't quite work correctly To: freebsd-hackers@freebsd.org Date: Tue, 11 Jul 1995 15:28:37 -0700 (PDT) X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Content-Length: 836 Sender: hackers-owner@freebsd.org Precedence: bulk I recently upgraded our news server from 64 megs to 128, which required using MAXMEM to specify 128 megs before the kernel would recognize it. This worked properly, however in testing I found that specifying *more* RAM than you actually had led to problems with paging. What I did was recompile the kernel on a machine with limited memory (a 486 with 8 megs) with MAXMEM set to 128 megs and LARGEMEM defined. This kernel correctly sized the memory *but* something goes wrong with the paging because starting up a make depend on the kernel source renders the machine unusable after a few minutes. After the make depend has run for about 10 minutes, vmstat -s reports 30 million pages examined by the page daemon vs. 30 thousand with the old kernel. Setting MAXMEM to 8 megs produced a kernel that worked correctly. Interesting, no?