Hi,
I have three machines running FC1. On those three, when I run top to check system usage, I see the load averages at sane values.
The lines displaying CPU usage are totally off. Either they're all at 0%, or irq, softirq and iowait are sharing everything among themselves, like this:
CPU states: cpu user nice system irq softirq iowait idle total 0,0% 0,0% 0,0% 33,3% 33,4% 33,1% 0,0%
Is anyone else seeing this ? Is this a known bug in top ? The only thing similar on the three machines is that I'm running stuff niced as well, but it shouldn't matter...
Thomas
Same thing for me.
Mine is fine.
OK, then maybe I should detail what I'm running a little bit :)
My distro is a Fedora Core Test 3 updated at 90% to Fedora Core, and also has some packages from third-party repositories and rawhide...
It even runs with a 2.4.20 kernel instead of the 2.4.22 kernel that comes with Fedora because of a kernel bug that made my hard drive go crazy for no reason.
I run this patchwork OS on a Fujitsu/Siemens Amilo D laptop with an Intel Pentium 4 CPU, in case that helps.
On Fri, Dec 05, 2003 at 06:22:59PM +0100, Thomas Vander Stichele wrote:
The lines displaying CPU usage are totally off. Either they're all at 0%, or irq, softirq and iowait are sharing everything among themselves, like this:
CPU states: cpu user nice system irq softirq iowait idle total 0,0% 0,0% 0,0% 33,3% 33,4% 33,1% 0,0%
Is anyone else seeing this ? Is this a known bug in top ? The only thing similar on the three machines is that I'm running stuff niced as well, but it shouldn't matter...
I'm seeing it too, on both smp and uniprocessor systems. I didn't notice it until you asked. If it's not already in bugzilla, please to file a bug report there...
michaelkjohnson
"He that composes himself is wiser than he that composes a book." Linux Application Development -- Ben Franklin http://people.redhat.com/johnsonm/lad/
On Fri, 2003-12-05 at 11:21, Michael K. Johnson wrote:
I'm seeing it too, on both smp and uniprocessor systems. I didn't notice it until you asked. If it's not already in bugzilla, please to file a bug report there...
I've noticed it too, but I've seen it reported before, so I assumed it was already in Bugzilla.
Wil
On Fri, Dec 05, 2003 at 06:22:59PM +0100, Thomas Vander Stichele wrote:
The lines displaying CPU usage are totally off. Either they're all at 0%, or irq, softirq and iowait are sharing everything among themselves, like this:
Heh, I noticed this a while back on AMD64 systems, and thought it was an issue with my kernel. I later forgot about it, and it is still there (the error that I thought was causing it is fixed). Now that I check my systems, I noticed I get it on some and not on others:
I get it on: IBM T23 Laptop, PIII 1GHz AMD Athlon64, Gigabyte MB, NF3 chipset AMD Athlon XP 2100+, Soltek MB, NF2 Chipset
I do not get it on: SMP Intel PIII 2 x 600MHz, Supermicro board AMD AthlonXP 1700+, Soyo MB, Via KT266A chipset
If you popped this in Bugzilla, let me know, and I will add this to the comments.
Justin M. Forbes
Hi,
please see
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=109484
for details and a patch for those who might need a quick solution.
The difference between the 2.4 and 2.6 kernels are that 2.6 is returning more information about system statuses in /proc. The information is not available in 2.4 and therefore not returned when reading from /proc/stat.
The problem in top is: uninitialized variables of the values to be read. The second patch in the issue above initializes the values before use.
Stephen
Am Fr, den 05.12.2003 schrieb Thomas Vander Stichele um 18:22:
Hi,
I have three machines running FC1. On those three, when I run top to check system usage, I see the load averages at sane values.
The lines displaying CPU usage are totally off. Either they're all at 0%, or irq, softirq and iowait are sharing everything among themselves, like this:
CPU states: cpu user nice system irq softirq iowait idle total 0,0% 0,0% 0,0% 33,3% 33,4% 33,1% 0,0%
Is anyone else seeing this ? Is this a known bug in top ? The only thing similar on the three machines is that I'm running stuff niced as well, but it shouldn't matter...
Thomas
-- fedora-devel-list mailing list fedora-devel-list@redhat.com http://www.redhat.com/mailman/listinfo/fedora-devel-list
On Sun, Dec 07, 2003 at 11:06:34AM +0100, Stephen Reindl wrote:
The problem in top is: uninitialized variables of the values to be read. The second patch in the issue above initializes the values before use.
Thank you! I attached yet another patch which is functionally equivalent but a little simpler -- just use calloc to get zeroed memory instead of calling malloc and memset.
michaelkjohnson
"He that composes himself is wiser than he that composes a book." Linux Application Development -- Ben Franklin http://people.redhat.com/johnsonm/lad/