Re: looking for a preempt-patch for 2.4.10-ac12

Robert Love (rml@tech9.net)
17 Oct 2001 15:53:54 -0400


On Wed, 2001-10-17 at 15:47, elko wrote:
> the patch is there, I applied it together with the stats-patch
> and my system is running like a charm right now, never have seen
> this kind of response in X.

Glad everything is working smooth...

> the only thing is, the perl-script at:
> http://www.tech9.net/rml/linux/top-latencies
>
> shows something this:
>
> ----[ SNIP ]----
> n min avg max cause mask start line/file address end line/file
> 14 9512 9590 9711 spin_lock 5 2111/tcp_ipv4.c c0226736119/softirq.c
> 89 9454 9559 9682 spin_lock 9 2111/tcp_ipv4.c c0226736119/softirq.c
> 2 9540 9551 9563 spin_lock 3 2111/tcp_ipv4.c c0226736119/softirq.c
> 3895 7708 9532 14296 spin_lock 1 2111/tcp_ipv4.c c0226736119/softirq.c
> 1 9513 9513 9513 spin_lock 1 2111/tcp_ipv4.c c02267362152/tcp_ipv4.c
> 363 3594 6166 9512 spin_lock 0 2111/tcp_ipv4.c c02267362152/tcp_ipv4.c
> ----[ SNIP ]----
>
> that 3895 number for '2111/tcp_ipv4.c c0226736119/softirq.c'
> keeps adding up, how should I translate that? big network
> latency, is that what it means? if so, any idea on how
> can I fix that??

the n column is the number of times the lock has been held. the lock is
probably held numerous times in a given second, so you see large n
values.

note that all those rows you showed could be in one row, but the
top-latencies tool keeps them separate since they have a different mask.

anyhow, the max recorded latency is 9.5ms which is not too bad. I'm not
looking at the code, but I would imagine its some TCP work done in a
BH. I wouldn't worry too much.

Robert Love

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