From owner-freebsd-hackers  Fri Jan 22 20:26:15 1999
Return-Path: <owner-freebsd-hackers@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id UAA10117
          for freebsd-hackers-outgoing; Fri, 22 Jan 1999 20:26:15 -0800 (PST)
          (envelope-from owner-freebsd-hackers@FreeBSD.ORG)
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 UAA10112
          for <hackers@FreeBSD.ORG>; Fri, 22 Jan 1999 20:26:13 -0800 (PST)
          (envelope-from dillon@apollo.backplane.com)
Received: (from dillon@localhost)
	by apollo.backplane.com (8.9.2/8.9.1) id UAA58213;
	Fri, 22 Jan 1999 20:25:56 -0800 (PST)
	(envelope-from dillon)
Date: Fri, 22 Jan 1999 20:25:56 -0800 (PST)
From: Matthew Dillon <dillon@apollo.backplane.com>
Message-Id: <199901230425.UAA58213@apollo.backplane.com>
To: "John S. Dyson" <dyson@iquest.net>
Cc: rcarter@pinyon.org (Russell L. Carter), dufault@hda.com,
        hackers@FreeBSD.ORG
Subject: Re: Error in vm_fault change
References:  <199901230415.XAA62055@y.dyson.net>
Sender: owner-freebsd-hackers@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

    Guys, guys... when I said 'priority' I simply meant 'some sort
    of scheduling mechanism'.    I didn't mean 'priority' in the
    sense of some arbitrary static number, nor do I infer that 
    we intentionally block any processes when memory *is* available.

    And when I said 'scheduler' I didn't mean scheduling for cpu,
    ok?  I meant scheduling who the available memory is given to.
    As in deciding intelligently.

    There are lots of ways to do this.  The easiest is to change
    FreeBSD's notion of when to block waiting for memory.  Right
    now it's notion is ( from vm_page_alloc() ):

        case VM_ALLOC_NORMAL:
                if (cnt.v_free_count >= cnt.v_free_reserved) {
                        m = vm_page_select_free(object, pindex, PQ_FREE);
                        KASSERT(m != NULL, ("vm_page_alloc(NORMAL): missing page
 on free queue\n"));
                } else {
                        m = vm_page_select_cache(object, pindex);
			...
		}

    So, for example, one way to schedule memory is to adjust
    the notion of cnt.v_free_reserved based on some dynamic entity
    associated with the process:

                if (cnt.v_free_count >= cnt.v_free_reserved + curproc->fubar) {

    And then make curproc->fubar a function of something.  That is ALL
    I am saying here.

					-Matt

					Matthew Dillon 
					<dillon@backplane.com>

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message