Hi,
I am using the latest 4.9.6-rt4 patch. I can reproduce the issue with
4.9.4, too.
The machine is an imx6 solo, and the kernel is the vanilla one.
I have CONFIG_PREEMPT_RT_FULL
I have basically taken the trivial hello_world example, and just added a
check on the wakeup time (see below):
The problem is that I am getting errors.
The process should wakeup every 20 milliseconds, but sometimes wakes up
2 to 4 msecs too late.
To reproduce this very easily, I generate interrupts by just keeping
the return key pressed in the serial console.
It it also reproducible, by performing a 'ping -f' from another machine.
I have tried to investigate this, with lttng at first.
In the control flow, I can see kworker/u:2 that can use 100% CPU during
about 27 msecs
After having activated some debug traces, I found out that the work
function for the kworker is 'write_to_ldisc' , which seems
pertinent since the interrupts come from the serial console.
But I still unable to tell why it starves the CPU for so much time.
Another fact is that test shows less fails, with CONFIG_PREEMPT, than
with CONFIG_PREEMPT_RT_FULL. And with CONFIG_PREEMPT, the late delay in
never bigger than 2 msecs.
As anyone encountered such an issue ?
Thanks !
-----------------
#include <stdlib.h>
#include <stdio.h>
#include <time.h>
#include <sched.h>
#include <sys/mman.h>
#include <string.h>
#include <stdint.h>
#define MY_PRIORITY (49) /* 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
guaranteed safe to access without
faulting */
#define NSEC_PER_SEC (1000000000) /* The number of nsecs per sec. */
void stack_prefault(void) {
unsigned char dummy[MAX_SAFE_STACK];
memset(dummy, 0, MAX_SAFE_STACK);
return;
}
uint64_t timespec2msec(const struct timespec * t) {
return (t->tv_sec*1000 + t->tv_nsec / 1000000.0);
}
int main(int argc, char* argv[])
{
struct timespec t;
struct sched_param param;
int interval = 20000000; /* 20 milliseconds */
/* 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();
clock_gettime(CLOCK_MONOTONIC ,&t);
/* start after one second */
t.tv_sec++;
while(1) {
struct timespec now;
int ret;
/* wait until next shot */
ret = clock_nanosleep(CLOCK_MONOTONIC, TIMER_ABSTIME, &t, NULL);
clock_gettime(CLOCK_MONOTONIC, &now);
uint64_t late = timespec2msec(&now) - timespec2msec(&t);
if (ret < 0)
printf("ERROR\n");
if (late >= 2) {
printf("Too late: %lld\n", late);
}
/* do the stuff */
/* calculate next shot */
t.tv_nsec += interval;
while (t.tv_nsec >= NSEC_PER_SEC) {
t.tv_nsec -= NSEC_PER_SEC;
t.tv_sec++;
}
}
}
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
