How to make a kernel thread sleep for a short amount of time?

I am re-working the net/core/pktgen code to be a kernel thread.

It is basically working, but I am having trouble making the thread efficiently sleep for durations in the milisecond and micro-second range.

I have looked at the udelay and mdelay methods, but they busy wait.

I do not need absolute real-time precision, so if I ask the thread to sleep for 100 micro-seconds, it is not a big deal if it does not wake up for 5000us. On average, it should be very close to 100us.

I believe the answer may be to use some sort of timer and have my thread sleep on this timer, but I cannot find any examples or documentation on how to do this on the web.

If anyone can point me to some example code or documentation, I would appreciate it.

Thanks,
Ben

Ben Greear <[email protected]>       <Ben_Greear AT excite.com>
President of Candela Technologies Inc      http://www.candelatech.com
ScryMUD:  http://scry.wanfear.com     http://scry.wanfear.com/~greear

The only generally available timer is the timer interrupt, sadly, which ticks once every 10ms (or soon once every ms, according to Linus' bitkeeper tree) on i386.

At OLS I was told of the existence of 'firm timers':
http://www.cse.ogi.edu/~luca/firm.html

These might have use in the shaping world too.

Regards,

bert hubert

http://www.PowerDNS.com          Versatile DNS Software & Services
http://www.tk                              the dot in .tk
http://lartc.org           Linux Advanced Routing & Traffic Control HOWTO

I am using high-res timers (http://sourceforge.net/projects/high-res-timers) in a driver to do sub-millisecond timing for a driver I am developing. With high-res timers, I have some code that looks like:

#ifdef CONFIG_HIGH_RES_TIMERS
       /* If the state machine asks for a short delay, then shorten
          the timer timeout. */
       if (kcs_result == KCS_CALL_WITH_DELAY) {
               kcs_timer.sub_expires
                       += usec_to_arch_cycles(KCS_SHORT_TIMEOUT_USEC);
               while (kcs_timer.sub_expires >= cycles_per_jiffies) {
                       kcs_timer.expires++;
                       kcs_timer.sub_expires -= cycles_per_jiffies;
               }
       } else {
               kcs_timer.expires += KCS_TIMEOUT_JIFFIES;
        }
#else
       kcs_timer.expires += KCS_TIMEOUT_JIFFIES;

#endif

But the high-res timers are not in the kernel right now, it's a patch you have to add, and the user has to have it configured.

-Corey