From owner-p4-projects@FreeBSD.ORG Tue Aug 1 08:23:16 2006
Return-Path: <owner-p4-projects@FreeBSD.ORG>
X-Original-To: p4-projects@freebsd.org
Delivered-To: p4-projects@freebsd.org
Received: by hub.freebsd.org (Postfix, from userid 32767)
id D658216A4E0; Tue, 1 Aug 2006 08:23:15 +0000 (UTC)
X-Original-To: perforce@FreeBSD.org
Delivered-To: perforce@FreeBSD.org
Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125])
by hub.freebsd.org (Postfix) with ESMTP id 80CF416A4DD
for <perforce@FreeBSD.org>; Tue, 1 Aug 2006 08:23:15 +0000 (UTC)
(envelope-from cdjones@FreeBSD.org)
Received: from repoman.freebsd.org (repoman.freebsd.org [216.136.204.115])
by mx1.FreeBSD.org (Postfix) with ESMTP id DEAE943D5A
for <perforce@FreeBSD.org>; Tue, 1 Aug 2006 08:23:14 +0000 (GMT)
(envelope-from cdjones@FreeBSD.org)
Received: from repoman.freebsd.org (localhost [127.0.0.1])
by repoman.freebsd.org (8.13.6/8.13.6) with ESMTP id k718NEFd047331
for <perforce@FreeBSD.org>; Tue, 1 Aug 2006 08:23:14 GMT
(envelope-from cdjones@FreeBSD.org)
Received: (from perforce@localhost)
by repoman.freebsd.org (8.13.6/8.13.4/Submit) id k718NEZg047328
for perforce@freebsd.org; Tue, 1 Aug 2006 08:23:14 GMT
(envelope-from cdjones@FreeBSD.org)
Date: Tue, 1 Aug 2006 08:23:14 GMT
Message-Id: <200608010823.k718NEZg047328@repoman.freebsd.org>
X-Authentication-Warning: repoman.freebsd.org: perforce set sender to
cdjones@FreeBSD.org using -f
From: Chris Jones <cdjones@FreeBSD.org>
To: Perforce Change Reviews <perforce@FreeBSD.org>
Cc:
Subject: PERFORCE change 102900 for review
X-BeenThere: p4-projects@freebsd.org
X-Mailman-Version: 2.1.5
Precedence: list
List-Id: p4 projects tree changes <p4-projects.freebsd.org>
List-Unsubscribe: <http://lists.freebsd.org/mailman/listinfo/p4-projects>,
<mailto:p4-projects-request@freebsd.org?subject=unsubscribe>
List-Archive: <http://lists.freebsd.org/pipermail/p4-projects>
List-Post: <mailto:p4-projects@freebsd.org>
List-Help: <mailto:p4-projects-request@freebsd.org?subject=help>
List-Subscribe: <http://lists.freebsd.org/mailman/listinfo/p4-projects>,
<mailto:p4-projects-request@freebsd.org?subject=subscribe>
X-List-Received-Date: Tue, 01 Aug 2006 08:23:16 -0000
http://perforce.freebsd.org/chv.cgi?CH=102900
Change 102900 by cdjones@cdjones-impulse on 2006/08/01 08:22:36
Memory limitation now works.
Implementation: a kernel thread (per jail) checks to see whether the jail's RSS exceeds the jail's memory limit. If so, the kernel thread tries to page out 5% of each jailed process until the RSS is no longer over the memory limit. This could probably use tuning / more intelligence.
Patch to go on wiki & note to mailing list for beta testers coming later today. Next up, scheduling.
Affected files ...
.. //depot/projects/soc2006/cdjones_jail/src/sys/kern/kern_jail.c#18 edit
Differences ...
==== //depot/projects/soc2006/cdjones_jail/src/sys/kern/kern_jail.c#18 (text+ko) ====
@@ -128,13 +128,13 @@
struct proc *p;
struct prison *pr;
struct thread *td;
- vm_pindex_t limit, size, usage;
+ long limit, cursize, newsize, usage;
int breakout;
pr = arg;
printf("Starting jpager/%d with memory limit %ld bytes\n",
- pr->pr_id, (long) prison_memory_limit(pr));
+ pr->pr_id, (long) prison_memory_limit(pr));
for (;;) {
if (pr->pr_pager_flags & J_PAGER_TD_DIE)
@@ -144,8 +144,8 @@
* pushing back when we approach the limit, rather than
* when we hit it.
*/
- limit = prison_memory_limit(pr);
- usage = prison_memory(pr);
+ limit = (long) prison_memory_limit(pr);
+ usage = (long) prison_memory(pr);
/* The logic from vm_daemon() really needs to go here.
* Problem: we want to push things below their rlimits.
@@ -153,13 +153,13 @@
* TODO: refactor vm_daemon to optionally act on specific jails?
*/
- printf("jthread/%d: memory %ld / %ld bytes\n",
- pr->pr_id, (long) usage, (long) limit);
+ printf("jpager/%d: memory %ld / %ld bytes\n",
+ pr->pr_id, usage, limit);
if ((usage - limit) > 0) {
- printf("jthread/%d: overcommitted by %ld bytes (%d %%)\n",
- pr->pr_id, (long) (usage - limit),
- (int) (100 * (usage - limit) / limit));
+ printf("jpager/%d: overcommitted by %ld bytes (%lf percent)\n",
+ pr->pr_id, usage - limit,
+ (double) 100 * ((double) (usage - limit) / (double) limit));
sx_slock(&allproc_lock);
LIST_FOREACH(p, &allproc, p_list) {
@@ -197,25 +197,28 @@
*/
/* TODO: this arbitrarily reduces each process's space by
- * one page (until it's completely swapped out) while
+ * 5% (until it's completely swapped out) while
* we're under memory pressure. A better way would be
* to either hit large processes first, or to hit the
* least-active processes first, or go proportionally,
- * ....
+ * or ....
*/
- size = vmspace_resident_count(p->p_vmspace) - 1;
- if (size < 0)
- size = 0;
- printf("jpager_td: squeezing process %d to %ld\n", p->p_pid, (long) size);
- vm_pageout_map_deactivate_pages(&p->p_vmspace->vm_map, (long) size);
-
- sx_sunlock(&allproc_lock);
+ newsize = cursize = (long) vmspace_resident_count(p->p_vmspace);
+ newsize -= newsize / 20;
+ if (cursize < 0)
+ newsize = 0;
+ PROC_UNLOCK(p);
+ printf("jpager/%d: squeezing process %d from %ld to %ld\n",
+ pr->pr_id, p->p_pid, cursize, newsize);
+ vm_pageout_map_deactivate_pages(&p->p_vmspace->vm_map, newsize);
} /* end LIST_FOREACH procs */
+ sx_sunlock(&allproc_lock);
}
/* TODO --- make interval into a sysctl. */
+ /* 6 seconds because VM recomputes totals every 5. */
printf("jpager_td sleeping\n");
- tsleep(pr, 0, "-", 3 * hz);
+ tsleep(pr, 0, "-", 6 * hz);
}
printf("Exiting jpager_td\n");