Bug 14648 - [2.6.x regression] proc fs busy/usage statistics during IO operations makes no sense
Summary: [2.6.x regression] proc fs busy/usage statistics during IO operations makes n...
Status: RESOLVED INVALID
Alias: None
Product: Other
Classification: Unclassified
Component: Other (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: other_other
URL:
Keywords:
Depends on: 14531
Blocks:
  Show dependency tree
 
Reported: 2009-11-20 19:37 UTC by Artem S. Tashkinov
Modified: 2010-05-27 19:24 UTC (History)
0 users

See Also:
Kernel Version:
Subsystem:
Regression: Yes
Bisected commit-id:


Attachments

Description Artem S. Tashkinov 2009-11-20 19:37:53 UTC
I really dislike the fact that during intense IO operations (like copying a file from one partition to another one) you cannot see the actual CPU usage because kernel (via proc fs) reports that one of the CPUs is 100% busy. Most people, when faced with such statistics, will judge with all confidence, that CPU is indeed 100% busy which is not true.

One could think that this is probably a userspace bug, but I tend not to think this way, because both top and htop utitities report the same data. Furthermore, this simple test:

cat big_file > /dev/null

will raise load average to 1.00 which means that at least one of the CPUs in a system is 100% busy and load average is a _direct_ kernel counter, so, top and htop cannot be wrong.

That doesn't make sense. Please, fix this bug. If a CPU is indeed busy waiting for IO operations to complete, then a real IO wait time should be reported.

P.S. Kernel 2.4.x did not have this problem.
Comment 1 Artem S. Tashkinov 2010-05-27 19:24:24 UTC
Gnome System Monitor doesn't have this problem, closing.

Note You need to log in before you can comment on or make changes to this bug.