Hi there!
I've upgraded my laptop and I'm testing the following debian package:
- linux-image-3.0.0-1-686-pae 3.0.0-3
The first impression is the clocksource problem is apparently fixed, but the
system is unstable. The first problem is a high resource comsumption with
nfs-kernel-server ( 1:1.2.4-1 ) :
Linux 3.0.0-1-686-pae (tomberi) 30/08/11 _i686_ (1 CPU)
18:48:21 CPU %usr %nice %sys %iowait %irq %soft %steal
%guest %idle
18:48:23 all 0,50 0,00 64,50 1,50 0,00 8,50 0,00
0,00 25,00
18:48:25 all 0,50 0,00 35,32 0,00 0,00 5,47 0,00
0,00 58,71
18:48:27 all 1,00 0,00 1,50 3,50 0,00 0,00 0,00
0,00 94,00
18:48:29 all 1,01 0,00 39,39 2,53 0,00 8,59 0,00
0,00 48,48
18:48:31 all 0,50 0,00 63,50 1,50 0,00 8,50 0,00
0,00 26,00
18:48:33 all 1,49 0,00 68,16 1,00 0,00 5,47 0,00
0,00 23,88
18:48:35 all 0,50 0,00 67,34 1,51 0,00 5,53 0,00
0,00 25,13
18:48:37 all 1,00 0,00 66,00 1,00 0,00 9,00 0,00
0,00 23,00
18:48:39 all 0,50 0,00 65,50 2,00 0,00 5,50 0,00
0,00 26,50
18:48:41 all 0,50 0,00 67,00 0,50 0,00 8,00 0,00
0,00 24,00
top - 19:02:49 up 4 min, 1 user, load average: 0.07, 0.14, 0.06
Tasks: 98 total, 2 running, 96 sleeping, 0 stopped, 0 zombie
Cpu(s): 7.6%us, 26.3%sy, 0.8%ni, 51.4%id, 10.8%wa, 0.0%hi, 3.1%si,
0.0%st
Mem: 1000592k total, 701852k used, 298740k free, 20928k
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
1373 root 20 0 0 0 0 R 23.1 0.0 0:12.59 nfsd
1374 root 20 0 0 0 0 S 19.2 0.0 0:09.74 nfsd
1371 root 20 0 0 0 0 S 17.3 0.0 0:08.29 nfsd
1370 root 20 0 0 0 0 S 9.6 0.0 0:09.46 nfsd
This happens with clocksource=acpi_pm ( the default ) and jiffies on the
Linux 3.0 kernel. Now, finished copying files the system is at the same load
with the nsfd using nearly at 100%:
1369 root 20 0 0 0 0 R 10.6 0.0 1:54.41 nfsd
1370 root 20 0 0 0 0 R 10.6 0.0 1:35.82 nfsd
1372 root 20 0 0 0 0 R 10.6 0.0 1:20.20 nfsd
1373 root 20 0 0 0 0 R 10.6 0.0 2:12.25 nfsd
1374 root 20 0 0 0 0 R 10.6 0.0 2:19.37 nfsd
With 2.6.32 and the jiffies option, the system load while copying files is
the folliwing:
Linux 2.6.32-5-686 (tomberi) 30/08/11 _i686_ (1 CPU)
18:56:11 CPU %usr %nice %sys %iowait %irq %soft %steal
%guest %idle
18:56:13 all 14,93 0,00 0,00 1,99 0,00 0,00 0,00
0,00 83,08
18:56:15 all 8,04 0,00 0,00 4,52 0,00 0,00 0,00
0,00 87,44
18:56:17 all 7,96 0,00 0,00 0,00 0,00 0,00 0,00
0,00 92,04
18:56:19 all 3,52 0,00 0,50 1,51 0,00 0,00 0,00
0,00 94,47
18:56:21 all 10,00 0,00 0,00 3,00 0,00 0,50 0,00
0,00 86,50
18:56:23 all 16,92 0,00 1,49 0,00 0,00 0,00 0,00
0,00 81,59
18:56:25 all 0,00 0,00 0,00 1,50 0,00 0,00 0,00
0,00 98,50
18:56:27 all 7,54 0,00 0,00 0,00 0,00 0,00 0,00
0,00 92,46
18:56:29 all 7,96 0,00 0,00 2,99 0,00 0,00 0,00
0,00 89,05
18:56:31 all 7,54 0,00 0,00 0,00 0,00 0,00 0,00
0,00 92,46
18:56:33 all 0,00 0,00 0,00 0,00 0,00 0,00 0,00
0,00 100,00
18:56:35 all 7,50 0,00 0,00 2,50 0,00 0,00 0,00
0,00 90,00
18:56:37 all 1,00 0,00 0,50 0,00 0,00 0,00 0,00
0,00 98,50
18:56:39 all 2,00 0,00 0,00 0,00 0,00 0,00 0,00
0,00 98,00
18:56:41 all 7,50 0,00 0,50 3,00 0,00 0,00 0,00
0,00 89,00
18:56:43 all 0,00 0,00 0,00 0,00 0,00 0,00 0,00
0,00 100,00
18:56:45 all 0,50 0,00 0,50 0,00 0,00 0,00 0,00
0,00 99,00
The net throughtput and the clock drift seems fixed, but now my silent
laptop is now a noisy and hot.
I'll try to investigate this problem and I'll reopen the bug if the problem
strikes back.
Thanks for your great work.
On Thu, Aug 25, 2011 at 12:18 AM, Jonathan Nieder <jrnie...@gmail.com>wrote:
> Hi,
>
> Jose Luis Salas Carrascosa wrote:
>
> > I think that the bug is caused by changes in the clocksource between
> kernels
> > and the powernow-k8 module.
> >
> > One time I tested booting the xen flavour of the linux kernel ( the
> > 2.6.32-5-xen-686 ) and the problem dissapeared, the net worked at full
> speed
> > but there was no cpu scaling.
> >
> > Now, I boot the kernel with the clocksource=jiffies option and the
> > powersaved daemon to do the frequency scaling, the net works al full
> speed
> > and the notebook is less hot, and me happier.
>
> That's quite believable.
>
> I don't see any relevant fixes in stable recently (except maybe
> v2.6.32.43~31, "clocksource: Make watchdog robust vs. interruption"
> which is not in squeeze yet), while upstream of course there has been
> more activity. Would it be possible to test a v3.0.x kernel from
> unstable?
>
> Thanks for a clear report,
> Jonathan
>
