+++ This bug was initially created as a clone of Bug #137927 +++ Description of problem: The top utility is reporting incorrect values for memory usage of several long running processes. We have Linux rac5 2.4.21-27.EL #1 SMP Wed Dec 1 21:54:21 EST 2004 ia64 ia64 ia64 GNU/Linux Applied all the patches, and still having the issue. How reproducible: reported memory usage grows the longer the processes are running. Actual results: 16:32:50 up 14 days, 3:54, 23 users, load average: 0.33, 0.17, 0.06 133 processes: 132 sleeping, 1 running, 0 zombie, 0 stopped CPU states: cpu user nice system irq softirq iowait idle total 10.4% 0.6% 5.2% 0.2% 0.0% 0.2% 182.8% cpu00 0.6% 0.0% 0.3% 0.0% 0.0% 0.0% 99.0% cpu01 9.9% 0.6% 4.9% 0.3% 0.0% 0.3% 83.8% Mem: 509876k av, 495460k used, 14416k free, 0k shrd, 15048k buff 339952k actv, 62768k in_d, 6932k in_c Swap: 1044216k av, 398044k used, 646172k free 106360k cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 24717 smithj4 15 0 1591M 1.6G 3720 S 4.6 319.4 248:32 1 gkrellm 5148 smithj4 15 0 905M 894M 14876 S 0.0 179.7 109:55 0 mozilla-bin 6232 smithj4 15 0 805M 785M 5460 S 0.3 157.7 99:32 1 gnome-terminal 4625 root 15 0 870M 582M 7512 S 3.7 117.0 647:44 0 X 4726 smithj4 15 0 467M 463M 5184 S 0.3 93.0 35:02 0 gnome-panel 4472 root 15 0 98968 96M 540 S 0.0 19.4 0:01 0 crond Expected results: The ps command shows that gkrellm for example is using around 17MB of memory instead of the nearly 1.6GB that top shows: # ps aux | grep gkrellm smithj4 24717 2.3 0.9 17080 4924 ? S Oct26 248:50 gkrellm Additional info: Compiled the attachment shown here: https://bugzilla.redhat.com/bugzilla/attachment.cgi?id=106011&action=view and ran the executable: # ./pagesize getpagesize()=4096, PAGE_SHIFT: 12, pgshift: 2
> Applied all the patches, and still having the issue. Applied what patches?
Exactly how does one reproduce this problem? And what does the "pagesize" program have to do with it? Was the version whose output you show compiled on a 32-bit x86, and then run on an ia64? Are the programs with the ps/top memory-size inequity 32-bit executables? I'm completely confused...
Hmm.. it's not perfect bug report... I think clone a bug is not enough, better is clone -> think & edit -> commit new report bug :-) Please, ignore the program "pagesize" -- don't forget it's "clone of bug #137927" where it was used during bug investigation only. The original bug was possible detect by: http://people.redhat.com/kzak/procps/proc-mem-test.py For details see bug #137927.
Ok, can we confirm that this is an ia64-specific bug?
I an running 2.4.21-32 (the latest proposed RHEL3-U5 kernel) on an ia64, and the proc-mem-test.py test returns: "The resident memory information test PASS" every time I run it.
Also, running a VM-test program with a large RSS, I don't see any discrepancies between the output of top vs. ps, regardless of how long it runs. Can somebody please post a test program that shows the problem on 2.4.21-32?
This bug is filed against RHEL 3, which is in maintenance phase. During the maintenance phase, only security errata and select mission critical bug fixes will be released for enterprise products. Since this bug does not meet that criteria, it is now being closed. For more information of the RHEL errata support policy, please visit: http://www.redhat.com/security/updates/errata/ If you feel this bug is indeed mission critical, please contact your support representative. You may be asked to provide detailed information on how this bug is affecting you.