Re: My vote against eepro* removal

[Adrian Bunk]

On Thu, Jan 19, 2006 at 11:26:51AM +0100, kus Kusche Klaus wrote:
> > From: Lee Revell
> > On Thu, 2006-01-19 at 08:19 +0100, kus Kusche Klaus wrote:
> > > Last time I tested (around 2.6.12), eepro100 worked much better 
> > > in -rt kernels w.r.t. latencies than e100:
> > > 
> > > e100 caused a periodic latency of about 500 microseconds
> > > exactly every 2 seconds, no matter what the load on the interface
> > > was (i.e. even on an idle interface).
> > > 
> > > eepro100 did not show any latencies that long, it worked much
> > > smoother w.r.t. latencies.
> > > 
> > > Of course I would prefer to have e100 fixed over keeping eepro100
> > > around forever, but the last time I checked, it still wasn't fixed.
> > 
> > Please provide latency traces to illustrate the problematic code path.
> 
> It's not a "latency": As far as I can tell, interrupts or preemption
> are not disabled, the latency tracer doesn't show anything.
> 
> I just noticed that low-pri rt processes did not get scheduled for
> about 500 microseconds when e100 was active (even if the net was
> idle), and that there were no such breaks with eepro100.
> 
> I didn't analyze it in detail at that time, I believed that the e100
> interrupt handler thread was running every 2 seconds for 500 
> microseconds, because the interrupt count of eth0 incremented every
> 2 seconds, exactly when my rt processes paused.
> 
> This would be bad: That irq thread is at rt prio 47 on my system, 
> above many importent things.
> 
> However, I checked more closely now, and found out that only a small
> portion of the 500 microseconds is spent in the irq thread. Most of
> it is spent in the timer thread, at rt prio 1, so the whole thing
> is a much smaller problem than I originally believed.
> 
> Must be the function e100_watchdog.
>...

Is this with 2.6.12 or 2.6.16-rc1?

If it's the former, please check whether the problem is still presnt in 
the latter.

If it's the latter, I'm sure the e100 developers (Cc'ed) are interested 
in your problem.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/