Problem with timecounters and memory model

[Sebastian Huber]

On 29/05/15 15:41, Poul-Henning Kamp wrote:
> --------
> In message <55686C0F.9000101 at embedded-brains.de>, Sebastian Huber writes:
>
>> Ok, what makes me wonder is why this problem didn't show up in FreeBSD.=20
>> In our case we observed backward jumps in the supposed to be monotonic=20
>> uptime.
> How often do you "wind" the timecounter ?
>

More often than necessary. About 1000 times a second with only one 
timehand (this is a hack to support legacy systems). The test case is:

     while (1) {
         struct timespec newtime;

         if (clock_gettime(CLOCK_MONOTONIC, &newtime) != 0) {
             fprintf(stderr, "clock_gettime() failed: \"%s\"\n", 
strerror(errno));

             return (NULL);
         }


         if (!(newtime.tv_sec > lasttime.tv_sec
               || (newtime.tv_sec == lasttime.tv_sec
                   && newtime.tv_nsec >= lasttime.tv_nsec))) {
             printf("Time went backwards, from {%ld, %ld} to {%ld, 
%ld}\n", lasttime.tv_sec, lasttime.tv_nsec, newtime.tv_sec, 
newtime.tv_nsec);
         }

         lasttime = newtime;
     }

-- 
Sebastian Huber, embedded brains GmbH

Address : Dornierstr. 4, D-82178 Puchheim, Germany
Phone   : +49 89 189 47 41-16
Fax     : +49 89 189 47 41-09
E-Mail  : sebastian.huber at embedded-brains.de
PGP     : Public key available on request.

Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.