Re: [PATCH] Preemption Latency Measurement Tool

[Robert Love]

On Sat, 2001-09-22 at 17:09, Dieter Nützel wrote:
> Yes, I do like I've posted in all the related threads.
> 
> Here is some more input for you. I hope it could help nail it down.
> 
> Regards,
> 	Dieter
> 
> BTW I'll now switching to 2.4.10-pre14
>
> SunWave1#time tar xIf /Pakete/Linux/linux-2.4.9.tar.bz2
> 34.570u 5.350s 0:47.84 83.4%    0+0k 0+0io 295pf+0w
> 
> I "hear" some have disk activities (disk trashing).
> 
> SunWave1#sync
> 
> Runs for ages!!!

It has to flush that whole tar job you just did, which it did in memory
:)

> User	 CPU  0%
> System	 CPU  0%
> Idle	 CPU 99%
> 
> So where did it wait???
> 
> Here comes what latencytimes show:
> 
> Worst 20 latency times of 5061 measured in this period.
>   usec      cause     mask   start line/file      address   end line/file
> 322263   reacqBKL        1  1375/sched.c         c01138b4   697/sched.c
> 216174        BKL        1    30/inode.c         c016b971    52/inode.c
> 158328        BKL        9   742/block_dev.c     c0144d51   697/sched.c

OK, these three are not acceptable...adding them to the "list".

-- 
Robert M. Love
rml at ufl.edu
rml at tech9.net