Re: [PATCH] Migration of Standard Timers

[Lee Revell <rlrevell@joe-job.com>]

On Tue, 2006-09-19 at 11:41 -0500, Dimitri Sivanich wrote:
> On Tue, Sep 19, 2006 at 12:33:37PM -0400, Lee Revell wrote:
> > Which driver or subsystem is doing 100s of usecs of work in a timer?
> 
> The longest one I've captured so far results from:
> 
> rsp                rip                Function (args)
>  ======================= <nmi>
> 0xffff810257822fd8 0xffffffff803a0e94 rt_check_expire+0x8c
>  ======================= <interrupt>  
> 0xffff81025781fee8 0xffffffff803a0e08 rt_check_expire
> 0xffff81025781ff08 0xffffffff802386b3 run_timer_softirq+0x133
> 0xffff81025781ff38 0xffffffff80235262 __do_softirq+0x5e
> 0xffff81025781ff68 0xffffffff8020a958 call_softirq+0x1c
> 0xffff81025781ff80 0xffffffff8020bea7 do_softirq+0x2c
> 0xffff81025781ff90 0xffffffff80235142 irq_exit+0x48
> 

Ah, I remember that one.  Eric Dumazet had some suggestions to fix it
6-12 months ago which never went anywhere - the thread was called "RCU
latency regression in 2.6.16-rc1".

That one is especially annoying as there's no workaround (shrinking the
route cache or reducing the GC interval via net.ipv4.route.* sysctls has
no effect)

Lee

-
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/