Re: Linux 2.6.9 latencies: scheduler bug?

[Nick Piggin]

Sami Farin wrote:
> On Mon, Oct 25, 2004 at 12:26:18AM +0300, Sami Farin wrote:
> ...
>
> > while that's running, "rtc_latencytest 1024" has max latencies of 253ms.
> > $BLOCKLIST has around two thousand lines.
> > http://safari.iki.fi/ip_tables_original_maxlat_253ms.png
> > from pressing sysrq+p while it was running, I guesstimated one evil spot
> > and added cond_resched() in there :)
> > http://safari.iki.fi/ip_tables_cond_resched.png
> > at 5s I started the script.
>
>
> forget this stupid ip_tables.c patch, latencies have nothing to do with
> netfilter code, but bad interaction between xmms, rtc_latencytest
> and iptables.  I now get at max 3.1s (yup, 3100000us) latencies.
> http://safari.iki.fi/2.6.9-xmms-fun-1.png
> if you want to reproduce this:
> 1) run "rtc_latencytest 1024" (can't reproduce with "rtc_latencytest 512")
> 2) press play in xmms
> 3) start iptables-script
>
> xmms has to be prepared first.
> a) put it in repeat mode
> b) start playing >= 2 files (like that short testcase.mp3 from lame)
> c) remove the files while xmms is playing
> d) wait till xmms has played all of the selected files
> e) press stop [now then you press play in 2), xmms does silly infinite
>    loop without delays while trying to open the songs]
>
> please give me patches to try, 3.1s is really evil 8-)

Don't think I've tried rtc_latency test. A quick search didn't turn
up its source code...

So... stupid question, is rtc_latencytest running with a realtime
scheduling policy?
-
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/