From owner-freebsd-hackers Tue Feb 16 11:15:54 1999 Received: from apollo.backplane.com (apollo.backplane.com [209.157.86.2]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id LAA22536 for ; Tue, 16 Feb 1999 11:15:52 -0800 (PST) (envelope-from dillon@apollo.backplane.com) Received: (from dillon@localhost) by apollo.backplane.com (8.9.3/8.9.1) id LAA37772; Tue, 16 Feb 1999 11:15:44 -0800 (PST) (envelope-from dillon) Date: Tue, 16 Feb 1999 11:15:44 -0800 (PST) From: Matthew Dillon Message-Id: <199902161915.LAA37772@apollo.backplane.com> To: perlsta Cc: "John S. Dyson" , freebsd-hackers@FreeBSD.ORG Subject: Re: inode / exec_map interlock ? (follow up) References: Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG : :I've noticed that the 'old' swapper or system seemed to leave a LOT of :swap still used en it wasn't trully needed. The new system seems to :reclaim these regeons as soon as they are swapped in. I've noticed the :new swapper is a bit more 'peppy' but i'm concerned that it is dooing what :John says. : :What's the deal here? Matt, even though your swapper lists pages as :'free' does it actually keep them around for reuse? What happens when a :page is READ faulted in, is the backing swap kept allocated to save on IO :later? The new swapper fixes a bunch of things, but the main thing you are probably seeing is the on-the-fly reallocation of swap backing store when swapping out dirty pages and the semi-on-the-fly deallocation of swap backing store when a page is dirtied again. The async I/O swamping problem is so minor it's hardly worth 4 hours of flying felder carp. It took 5 minutes to fix once someone ( other then John ) figured out what the problem was. It should be noted that the original code was documented *solely* as solving a low memory lockup condition. It said absolutely nothing about limiting parallel I/O. Anywhere. In fact, in a large-memory configuration the original code *wouldn't* limit the I/O load because it was scaled to the free page margin rather then scaled to I/O load. What a complete and utter waste of time. -Matt Matthew Dillon :One other thing, I has some trouble getting to sleep last night and :decided to venture into src/sys/vm, the comments are VERY helpful. The :kind of documentation going on here will really help people get into :systems programming, it is MUCH appreciated. : :-Alfred : : To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message