Bug 508701

Summary:	"ps -e" does not show nscd (but "ps ax" does)
Product:	Red Hat Enterprise Linux 5	Reporter:	Matteo Corti <matteo>
Component:	glibc	Assignee:	Andreas Schwab <schwab>
Status:	CLOSED DUPLICATE	QA Contact:	qe-baseos-tools-bugs
Severity:	medium	Docs Contact:
Priority:	low
Version:	5.3	CC:	albert, drepper, fweimer, jbastian, matteo
Target Milestone:	rc
Target Release:	---
Hardware:	All
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2010-02-02 17:03:32 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Matteo Corti 2009-06-29 14:38:04 UTC

Description of problem:

ps -e -o comm or (ps axo comm) are not showing all the processes

Version-Release number of selected component (if applicable):

procps-3.2.7-11.1.el5

How reproducible:

Always. Start nscd and search for it with ps -e -o comm: the process is not listed (although running).
The very same works without any problem on Fedora

Steps to Reproduce:
1. start nscd (service nscd start)
2. check that nscd is running using BSD syntax (ps ax | grep nscd)
3. check that nscd is running using standard syntac (ps -e | grep nscd)
  
Actual results:

$ ps -e | grep nscd
$ 

Expected results:

$ ps -e | grep nscd
14175 ?        00:00:03 nscd
$

Additional info:

The same command works without problems on Fedora (F11, procps-3.2.7-27.fc11.i586)

Comment 1 Matteo Corti 2009-06-30 14:17:37 UTC

After further investigation I found out that the entry in /proc for ntsc is strange:

RH5 $ cat /proc/PID/stat

28793 (exe) S 1 28793 28793 0 -1 4202816 2918 0 0 0 22 253 0 0 15 0 7 0 37808079 115728384 798 4294967295 7360512 7463540 3214996912 3214995252 4903938 0 0 3674112 16391 4294967295 0 0 17 1 0 0 0

tcomm is exe instead of nscd
/proc/PID/exe is a link to /usr/sbin/nscd

ps reads 'exe' and lists exe with the 'comm' argument

This does not happen on Fedora:

F11 $ cat /proc/PID/stat

14175 (nscd) S 1 14175 14175 0 -1 4194624 37521 0 5 0 467 1853 0 0 20 0 12 0 1878751 153812992 310 4294967295 1843200 1977488 3219997600 3219991756 6214692 0 0 3674112 16390 4294967295 0 0 17 0 0 0 0 0 0

Here tcomm corresponds to 'nscd'

Now: ist this a bug of ps that when exe is encountered should follow /proc/PID/exe or is it a kernel problem and 'exe' in tcomm field is wrong?

Comment 2 Albert Cahalan 2009-06-30 19:10:05 UTC

It's possible that nscd called prctl to change the name to exe.
Use strace to find out.

If not, this appears to be a kernel bug.

Comment 3 Matteo Corti 2009-07-07 05:41:08 UTC

Ok figured out, it is an nscd problem. nscd does:

  execv ("/proc/self/exe", argv);

which changes the name to exe

This does not occur on Fedora as there nscd tries to read the symlink and use the real name:

  /* Try to exec the real nscd program so the process name (as reported
     in /proc/PID/status) will be 'nscd', but fall back to /proc/self/exe
     if readlink fails */
  ssize_t n = readlink ("/proc/self/exe", pathbuf, sizeof (pathbuf) - 1);
  if (n == -1)
    execv ("/proc/self/exe", argv);
  else
    {
      pathbuf[n] = '\0';
      execv (pathbuf, argv);
    }

Thanks

Comment 4 Subhendu Ghosh 2009-10-23 16:44:14 UTC

Thanks for the update.
Moving to glibc component as nscd is part of glibc