* [linux-2.6.26.8-rt14] modified "square wave" example program
@ 2009-02-25 14:44 Lukasz Majewski
0 siblings, 0 replies; only message in thread
From: Lukasz Majewski @ 2009-02-25 14:44 UTC (permalink / raw)
To: linux-rt-users; +Cc: majess1982
Hello.
I'm trying to solve problem with my RT application.
I've downloaded "square.c" example program from rt-wiki web page.
I've modified it a bit and set 20 ms period of my RT task and busy wait
for 15ms on this task. I'm "yielding" CPU for other tasks for 5ms. The
whole source code is pasted below.
This busy waiting simulates my "real" application in which I'm
performing some calculations (signal filtering).
I've unchanged priorities of other tasks (as also shown below).
On my target, root file system is mounted via NFS. I'm not using any ram
disk for startup purposes, just mounting root file system via NFS from
my host machine (root=/dev/nfs on my kernel's cmd line).
I'm using brand new arm-v5te-linux-gnueabi toolchain with glibc 2.8, gcc
4.3.2, binutils 2.18 (special thanks to Mark and Robert from
pengutronix.de for support :-) )
I have to use static linking during building of my application, because
on my NFS mounted root file system I've got other toolchain built (gcc
3.2 with OABI) and shared libraries from it aren't working with my
program. Due to this, linked binary has 3.2 MB and after stripping
around 460KB.
When I start my program, it runs once and exit normally. When I want to
start it again ,it hangs and target is not responsive anymore.
I'd like to ask if there is any limit on the amount of time that RT
process/ thread can consume only for it? To be more precise , if I have
20 ms period for RT process/thread is it safe to busy waiting(i.e.
executing some operations) for 15 ms in it? How can I assure that other,
necessary processes will not starve (especially those responsible for
NFS root file system )?
For the original "square wave" example RT thread wakes up with period of
50us performs it's operation for maybe 5us and then goes sleep. It means
(in a big simplification) that it only consumes 10% of processor time.
In my modified example it's consuming around 75% of CPU time (
nonetheless top utility shows always up to 50%).
Are there any special guidelines for writing programs employing cyclic
schedule?
Regards,
Lukasz
p.s. All comments on pasted below application are very welcomed. :-)
Modified "square wave " application source code.
#include <stdlib.h>
#include <stdio.h>
#include <time.h>
#include <sched.h>
#include <sys/mman.h>
#include <string.h>
#include <errno.h> /*Error number definitions*/
#define MY_PRIORITY (71) /* we use 49 as the PRREMPT_RT use 50
as the priority of kernel tasklets
and interrupt handler by default */
#define MAX_SAFE_STACK (8*1024) /* The maximum stack size which is
guranteed safe to access without
faulting */
#define NSEC_PER_SEC (1000000000) /* The number of nsecs per sec. */
/* Time measurement definition */
static struct timespec tim_start, tim;
void init_reftime(void) {
int ret;
ret = clock_gettime(CLOCK_MONOTONIC, &tim_start);
if (ret == -1) {
perror("clock_gettime");
}
}
unsigned long reftime(void) { /* usek resolution */
int ret;
static unsigned long long reftime;
static long tmp_reftime;
ret = clock_gettime(CLOCK_MONOTONIC, &tim);
if (ret == -1) {
perror("clock_gettime");
}
/* Calculating time stamp at us resolution */
tmp_reftime = tim.tv_sec - tim_start.tv_sec;
reftime = (unsigned long long) tmp_reftime * 1000000000; /* sec to ns */
tmp_reftime = tim.tv_nsec - tim_start.tv_nsec;
reftime += (unsigned long long) tmp_reftime;
/* ns to us */
reftime /= 1000;
return reftime;
}
void stack_prefault(void) {
unsigned char dummy[MAX_SAFE_STACK];
memset(&dummy, 0, MAX_SAFE_STACK);
return;
}
int main(int argc, char* argv[])
{
static int cnt;
static unsigned long tt,ta;
struct timespec t;
struct sched_param param;
int interval = 20000000; /* 20ms*/
/* Declare ourself as a real time task */
param.sched_priority = MY_PRIORITY;
if(sched_setscheduler(0, SCHED_FIFO, ¶m) == -1) {
perror("sched_setscheduler failed");
exit(-1);
}
/* Lock memory */
if(mlockall(MCL_CURRENT|MCL_FUTURE) == -1) {
perror("mlockall failed");
exit(-2);
}
/* Pre-fault our stack */
stack_prefault();
init_reftime();
clock_gettime(CLOCK_MONOTONIC ,&t);
/* start after one second */
t.tv_sec++;
for(cnt = 0; cnt < 10000; cnt++) {
/* wait until next shot */
clock_nanosleep(CLOCK_MONOTONIC, TIMER_ABSTIME, &t, NULL);
/* do the stuff */
/* busy waiting for 15 ms -> period is 20ms */
tt = reftime();
ta = tt;
while (ta <= (tt+15000)) {
ta = reftime();
}
/* calculate next shot */
t.tv_nsec += interval;
while (t.tv_nsec >= NSEC_PER_SEC) {
t.tv_nsec -= NSEC_PER_SEC;
t.tv_sec++;
}
}
return 0;
}
Output from ps -eH -o pid,rtptio,sched,cmd command executed on my target
(rt_test is the name of my application):
172:~# ps -eH -o pid,rtprio,sched,cmd
PID RTPRIO SCH CMD
2 - 0 [kthreadd]
3 99 1 [posixcputmr/0]
4 50 1 [sirq-high/0]
5 50 1 [sirq-timer/0]
6 50 1 [sirq-net-tx/0]
7 50 1 [sirq-net-rx/0]
8 50 1 [sirq-block/0]
9 50 1 [sirq-tasklet/0]
10 50 1 [sirq-sched/0]
11 50 1 [sirq-hrtimer/0]
12 50 1 [sirq-rcu/0]
13 - 0 [desched/0]
14 - 0 [rcu_sched_grace]
15 1 1 [events/0]
16 - 0 [khelper]
63 1 1 [krcupreemptd]
64 - 0 [pdflush]
65 - 0 [pdflush]
66 - 0 [kswapd0]
67 - 0 [aio/0]
68 - 0 [nfsiod]
660 50 1 [IRQ-5]
695 - 0 [rpciod/0]
696 50 1 [loadavg]
701 50 1 [IRQ-1]
1 - 0 init [2]
719 - 0 /usr/sbin/sshd
735 - 0 sshd: root@pts/0
738 - 0 -bash
743 - 0 ps -eH -o pid,rtprio,sched,cmd
721 - 0 /bin/bash
734 71 1 ./rt_test
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2009-02-25 14:44 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-02-25 14:44 [linux-2.6.26.8-rt14] modified "square wave" example program Lukasz Majewski
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).