svn commit: r290662 - head/sys/kern
Josh Paetzel
jpaetzel at FreeBSD.org
Tue Nov 10 14:14:34 UTC 2015
Author: jpaetzel
Date: Tue Nov 10 14:14:32 2015
New Revision: 290662
URL: https://svnweb.freebsd.org/changeset/base/290662
Log:
Fix a bug in the CPU % limiting code
If you attempt to set a pcpu limit that is higher than
110% using rctl (for instance, you want a jail to be
able to use 2 cores on your system so you set pcpu to
200%) the thing you are trying to limit becomes unthrottled.
PR: 189870
Submitted by: dustinwenz at ebureau.com
Reviewed by: trasz
MFC after: 1 week
Modified:
head/sys/kern/kern_racct.c
Modified: head/sys/kern/kern_racct.c
==============================================================================
--- head/sys/kern/kern_racct.c Tue Nov 10 13:47:28 2015 (r290661)
+++ head/sys/kern/kern_racct.c Tue Nov 10 14:14:32 2015 (r290662)
@@ -517,16 +517,16 @@ racct_adjust_resource(struct racct *racc
/*
* There are some cases where the racct %cpu resource would grow
- * beyond 100%.
- * For example in racct_proc_exit() we add the process %cpu usage
- * to the ucred racct containers. If too many processes terminated
- * in a short time span, the ucred %cpu resource could grow too much.
- * Also, the 4BSD scheduler sometimes returns for a thread more than
- * 100% cpu usage. So we set a boundary here to 100%.
+ * beyond 100% per core. For example in racct_proc_exit() we add
+ * the process %cpu usage to the ucred racct containers. If too
+ * many processes terminated in a short time span, the ucred %cpu
+ * resource could grow too much. Also, the 4BSD scheduler sometimes
+ * returns for a thread more than 100% cpu usage. So we set a sane
+ * boundary here to 100% * the maxumum number of CPUs.
*/
if ((resource == RACCT_PCTCPU) &&
- (racct->r_resources[RACCT_PCTCPU] > 100 * 1000000))
- racct->r_resources[RACCT_PCTCPU] = 100 * 1000000;
+ (racct->r_resources[RACCT_PCTCPU] > 100 * 1000000 * (int64_t)MAXCPU))
+ racct->r_resources[RACCT_PCTCPU] = 100 * 1000000 * (int64_t)MAXCPU;
}
static int
More information about the svn-src-all
mailing list