8133666: OperatingSystemMXBean reports abnormally high machine CPU consumption on Linux
Reviewed-by: sla, mgronlun
/*
* Copyright (c) 2011, 2015, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation. Oracle designates this
* particular file as subject to the "Classpath" exception as provided
* by Oracle in the LICENSE file that accompanied this code.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*/
#include <stdio.h>
#include <stdint.h>
#include <stdarg.h>
#include <unistd.h>
#include <errno.h>
#include <string.h>
#include <sys/resource.h>
#include <sys/types.h>
#include <dirent.h>
#include <stdlib.h>
#include <dlfcn.h>
#include <pthread.h>
#include <inttypes.h>
#include "com_sun_management_internal_OperatingSystemImpl.h"
struct ticks {
uint64_t used;
uint64_t usedKernel;
uint64_t total;
};
typedef struct ticks ticks;
typedef enum {
CPU_LOAD_VM_ONLY,
CPU_LOAD_GLOBAL,
} CpuLoadTarget;
static struct perfbuf {
int nProcs;
ticks jvmTicks;
ticks cpuTicks;
ticks *cpus;
} counters;
#define DEC_64 "%"SCNd64
static void next_line(FILE *f) {
while (fgetc(f) != '\n');
}
/**
* Return the total number of ticks since the system was booted.
* If the usedTicks parameter is not NULL, it will be filled with
* the number of ticks spent on actual processes (user, system or
* nice processes) since system boot. Note that this is the total number
* of "executed" ticks on _all_ CPU:s, that is on a n-way system it is
* n times the number of ticks that has passed in clock time.
*
* Returns a negative value if the reading of the ticks failed.
*/
static int get_totalticks(int which, ticks *pticks) {
FILE *fh;
uint64_t userTicks, niceTicks, systemTicks, idleTicks;
uint64_t iowTicks = 0, irqTicks = 0, sirqTicks= 0;
int n;
if((fh = fopen("/proc/stat", "r")) == NULL) {
return -1;
}
n = fscanf(fh, "cpu " DEC_64 " " DEC_64 " " DEC_64 " " DEC_64 " " DEC_64 " "
DEC_64 " " DEC_64,
&userTicks, &niceTicks, &systemTicks, &idleTicks,
&iowTicks, &irqTicks, &sirqTicks);
// Move to next line
next_line(fh);
//find the line for requested cpu faster to just iterate linefeeds?
if (which != -1) {
int i;
for (i = 0; i < which; i++) {
if (fscanf(fh, "cpu%*d " DEC_64 " " DEC_64 " " DEC_64 " " DEC_64 " "
DEC_64 " " DEC_64 " " DEC_64,
&userTicks, &niceTicks, &systemTicks, &idleTicks,
&iowTicks, &irqTicks, &sirqTicks) < 4) {
fclose(fh);
return -2;
}
next_line(fh);
}
n = fscanf(fh, "cpu%*d " DEC_64 " " DEC_64 " " DEC_64 " " DEC_64 " "
DEC_64 " " DEC_64 " " DEC_64 "\n",
&userTicks, &niceTicks, &systemTicks, &idleTicks,
&iowTicks, &irqTicks, &sirqTicks);
}
fclose(fh);
if (n < 4) {
return -2;
}
pticks->used = userTicks + niceTicks;
pticks->usedKernel = systemTicks + irqTicks + sirqTicks;
pticks->total = userTicks + niceTicks + systemTicks + idleTicks +
iowTicks + irqTicks + sirqTicks;
return 0;
}
static int vread_statdata(const char *procfile, const char *fmt, va_list args) {
FILE *f;
int n;
char buf[2048];
if ((f = fopen(procfile, "r")) == NULL) {
return -1;
}
if ((n = fread(buf, 1, sizeof(buf), f)) != -1) {
char *tmp;
buf[n-1] = '\0';
/** skip through pid and exec name. the exec name _could be wacky_ (renamed) and
* make scanf go mupp.
*/
if ((tmp = strrchr(buf, ')')) != NULL) {
// skip the ')' and the following space but check that the buffer is long enough
tmp += 2;
if (tmp < buf + n) {
n = vsscanf(tmp, fmt, args);
}
}
}
fclose(f);
return n;
}
static int read_statdata(const char *procfile, const char *fmt, ...) {
int n;
va_list args;
va_start(args, fmt);
n = vread_statdata(procfile, fmt, args);
va_end(args);
return n;
}
/** read user and system ticks from a named procfile, assumed to be in 'stat' format then. */
static int read_ticks(const char *procfile, uint64_t *userTicks, uint64_t *systemTicks) {
return read_statdata(procfile, "%*c %*d %*d %*d %*d %*d %*u %*u %*u %*u %*u "DEC_64" "DEC_64,
userTicks, systemTicks
);
}
/**
* Return the number of ticks spent in any of the processes belonging
* to the JVM on any CPU.
*/
static int get_jvmticks(ticks *pticks) {
uint64_t userTicks;
uint64_t systemTicks;
if (read_ticks("/proc/self/stat", &userTicks, &systemTicks) < 0) {
return -1;
}
// get the total
if (get_totalticks(-1, pticks) < 0) {
return -1;
}
pticks->used = userTicks;
pticks->usedKernel = systemTicks;
return 0;
}
/**
* This method must be called first, before any data can be gathererd.
*/
int perfInit() {
static int initialized=1;
if (!initialized) {
int i;
int n = sysconf(_SC_NPROCESSORS_ONLN);
if (n <= 0) {
n = 1;
}
counters.cpus = calloc(n,sizeof(ticks));
if (counters.cpus != NULL) {
// For the CPU load
get_totalticks(-1, &counters.cpuTicks);
for (i = 0; i < n; i++) {
get_totalticks(i, &counters.cpus[i]);
}
// For JVM load
get_jvmticks(&counters.jvmTicks);
initialized = 1;
}
}
return initialized ? 0 : -1;
}
#define MAX(a,b) (a>b?a:b)
#define MIN(a,b) (a<b?a:b)
static pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
/**
* Return the load of the CPU as a double. 1.0 means the CPU process uses all
* available time for user or system processes, 0.0 means the CPU uses all time
* being idle.
*
* Returns a negative value if there is a problem in determining the CPU load.
*/
static double get_cpuload_internal(int which, double *pkernelLoad, CpuLoadTarget target) {
uint64_t udiff, kdiff, tdiff;
ticks *pticks, tmp;
double user_load = -1.0;
int failed = 0;
*pkernelLoad = 0.0;
pthread_mutex_lock(&lock);
if(perfInit() == 0) {
if (target == CPU_LOAD_VM_ONLY) {
pticks = &counters.jvmTicks;
} else if (which == -1) {
pticks = &counters.cpuTicks;
} else {
pticks = &counters.cpus[which];
}
tmp = *pticks;
if (target == CPU_LOAD_VM_ONLY) {
if (get_jvmticks(pticks) != 0) {
failed = 1;
}
} else if (get_totalticks(which, pticks) < 0) {
failed = 1;
}
if(!failed) {
// seems like we sometimes end up with less kernel ticks when
// reading /proc/self/stat a second time, timing issue between cpus?
if (pticks->usedKernel < tmp.usedKernel) {
kdiff = 0;
} else {
kdiff = pticks->usedKernel - tmp.usedKernel;
}
tdiff = pticks->total - tmp.total;
udiff = pticks->used - tmp.used;
if (tdiff == 0) {
user_load = 0;
} else {
if (tdiff < (udiff + kdiff)) {
tdiff = udiff + kdiff;
}
*pkernelLoad = (kdiff / (double)tdiff);
// BUG9044876, normalize return values to sane values
*pkernelLoad = MAX(*pkernelLoad, 0.0);
*pkernelLoad = MIN(*pkernelLoad, 1.0);
user_load = (udiff / (double)tdiff);
user_load = MAX(user_load, 0.0);
user_load = MIN(user_load, 1.0);
}
}
}
pthread_mutex_unlock(&lock);
return user_load;
}
double get_cpu_load(int which) {
double u, s;
u = get_cpuload_internal(which, &s, CPU_LOAD_GLOBAL);
if (u < 0) {
return -1.0;
}
// Cap total systemload to 1.0
return MIN((u + s), 1.0);
}
double get_process_load() {
double u, s;
u = get_cpuload_internal(-1, &s, CPU_LOAD_VM_ONLY);
if (u < 0) {
return -1.0;
}
return u + s;
}
JNIEXPORT jdouble JNICALL
Java_com_sun_management_internal_OperatingSystemImpl_getSystemCpuLoad0
(JNIEnv *env, jobject dummy)
{
if(perfInit() == 0) {
return get_cpu_load(-1);
} else {
return -1.0;
}
}
JNIEXPORT jdouble JNICALL
Java_com_sun_management_internal_OperatingSystemImpl_getProcessCpuLoad0
(JNIEnv *env, jobject dummy)
{
if(perfInit() == 0) {
return get_process_load();
} else {
return -1.0;
}
}