From owner-cvs-all Mon Jun 25 23:29:55 2001 Delivered-To: cvs-all@freebsd.org Received: from earth.backplane.com (earth-nat-cw.backplane.com [208.161.114.67]) by hub.freebsd.org (Postfix) with ESMTP id 2C94537B413; Mon, 25 Jun 2001 23:29:23 -0700 (PDT) (envelope-from dillon@earth.backplane.com) Received: (from dillon@localhost) by earth.backplane.com (8.11.3/8.11.2) id f5Q6T9P20451; Mon, 25 Jun 2001 23:29:09 -0700 (PDT) (envelope-from dillon) Date: Mon, 25 Jun 2001 23:29:09 -0700 (PDT) From: Matt Dillon Message-Id: <200106260629.f5Q6T9P20451@earth.backplane.com> To: mikea Cc: Mikhail Teterin , "Daniel O'Connor" , cvs-all@FreeBSD.ORG, cvs-committers@FreeBSD.ORG, jhb@FreeBSD.ORG Subject: Re: cvs commit: src/sys/vm vm_pageout.c References: <200106221613.f5MGDGA80305@aldan.algebra.com> <20010622113200.B72711@mikea.ath.cx> Sender: owner-cvs-all@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG :... :> :> Yes, on a number of occasions. Its un-orthogonal. I can turn swapping :> on, but I can not turn it off without rebooting. Yackk. See also the :> BUGS sections of swapon(8). : :Also yes, more than a few times. When I ran a mainframe shop :(MVS, then OS/390) on IBM iron, we had the ability (after lots of :people griped because it wasn't there) to turn off paging and/or :swapping on a particular page/swap file so that we could take the :device offline for maintenance. The IBM developers asked why we :would ever want to do that; apparently they had never heard of :hardware failures or of device monitors (e.g., SMART in the PC :world, IOS in the MVS mainframe world) that warned of them being :impending. : :In my humble opinion, backed up by 37 years of fighting :computers to a draw, this needs to be added to FreeBSD. : :-- :Mike Andrews :mikea@mikea.ath.cx Well, go for it! No, I'm not kidding... well, ok, maybe I am. There was some work done on this a long while back but I don't remember it ever being 100% completed. It isn't impossible to do, but it isn't trivial either. You basically have to reserve all remaining swap space associated with the device in question and then force a page-in of all the data swapped to that device, freeing and reserving the backing swap as you go. When you've got the whole device's space reserved (assuming you haven't thrashed the machine to death getting there :-)), you can remove the device from consideration. The scanning required is actually fairly easy to do with the new swap code's consolidated hash table. Paging the data and reserving the space as you go is harder. It would be an interesting 'Intermediate Kernel Hacker' project (verses 'Junior Kernel Hacker' project). -Matt To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message