Re: 2.6.14-rc4-rt7

[Steven Rostedt <rostedt@goodmis.org>]

On Thu, 2005-10-27 at 15:01 -0700, William Weston wrote:

> Hi Steven.  I think you fixed it!
> 
> The xeon-smt box has been up for over 30 minutes with this patch, and no
> warnings as of yet.  Also, 'rtc_wakeup -f 1024' is reporting a max jitter
> of 131us (decent for this box considering its hardware induced latencies)
> instead of the >500us jitter seen earlier.
> 
> I'll try the patch out on the athlon box (which does mostly audio/midi)
> when I get home.

Yeah, I finally got a machine available that I could run Ingo's RT patch
on.  And with out this fix, I get the warning messages with the
following program, and with the fix I don't.  So I guess that solves it.

-- Steve

#include <stdio.h>
#include <time.h>

/* I'm sure there's a compare for this, but I was to lazy to look */
static inline int comparets(struct timespec *a, struct timespec *b)
{
	return (a->tv_sec < b->tv_sec) ? -1 :
		(a->tv_sec > b->tv_sec) ? 1 :
		(a->tv_nsec < b->tv_nsec) ? -1 :
		(a->tv_nsec > b->tv_nsec) ? 1 :
		0;
}

int main(int argc, char **argv)
{
	struct timespec ts, oldts;
	int i;

	clock_gettime(CLOCK_MONOTONIC, &oldts);
	for (i=0; i < 1000000; i++) {
		clock_gettime(CLOCK_MONOTONIC, &ts);
		if (comparets(&ts,&oldts) < 0) {
			printf("time went backwards from %ld.%09ld to %ld.%09ld\n",
				oldts.tv_sec, oldts.tv_nsec,
				ts.tv_sec, ts.tv_nsec);
		}
		oldts = ts;
	}
	return 0;
}