From owner-freebsd-current Mon Nov 4 20:19:23 1996 Return-Path: owner-current Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id UAA29207 for current-outgoing; Mon, 4 Nov 1996 20:19:23 -0800 (PST) Received: from dyson.iquest.net ([198.70.144.127]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id UAA29192 for ; Mon, 4 Nov 1996 20:19:16 -0800 (PST) Received: (from root@localhost) by dyson.iquest.net (8.8.2/8.6.9) id XAA00188 for current@freebsd.org; Mon, 4 Nov 1996 23:19:12 -0500 (EST) From: "John S. Dyson" Message-Id: <199611050419.XAA00188@dyson.iquest.net> Subject: Please test the new contigmalloc To: current@freebsd.org Date: Mon, 4 Nov 1996 23:19:10 -0500 (EST) Reply-To: dyson@freebsd.org X-Mailer: ELM [version 2.4 PL24 ME8] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk This is to let you know that there is a new contigmalloc committed tonight. This new code increases dramatically the probability that a sound driver or it's like can be loaded at bootup time after the system starts, and on large systems might even be able to be loaded sometime after bootup time. Specifically, it is best to load the sound LKM *IMMEDIATELY* after the mount of the root filesystem. Also, there are still problems on systems < 8MBytes with this new code, in that the driver might not get enough memory in that case. It appears that just after the mount of root and swap space, the new contigmalloc will succeed for a 64K allocation. Perhaps it would be wise to back-off to a smaller allocation upon an allocation failure. In the case of failure, the major thing that still should be done is that the buffer cache should probably be minimized. Otherwise, this code does about as much as can be done given the current architecture of page allocation. John