Bug 211956

Summary:	top goes into hard loop on xterm exit
Product:	Red Hat Enterprise Linux 4	Reporter:	Major <major>
Component:	procps	Assignee:	Karel Zak <kzak>
Status:	CLOSED WONTFIX	QA Contact:	Brian Brock <bbrock>
Severity:	medium	Docs Contact:
Priority:	medium
Version:	4.4
Target Milestone:	---
Target Release:	---
Hardware:	i686
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2006-10-24 10:26:22 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 Major 2006-10-24 04:16:51 UTC

Description of problem:
top run as root goes into hard loop if the xterm it is running under exits

Version-Release number of selected component (if applicable):
procps-3.2.3-8.4

How reproducible:
100%

Steps to Reproduce:
1. Start an xterm
2. su to root
3. run top
4. kill the xterm
  
Actual results:
top process goes into hard loop consuming 100% of available cpu

Expected results:
top process terminates

Additional info:

Comment 1 Karel Zak 2006-10-24 10:17:17 UTC

*** Bug 211957 has been marked as a duplicate of this bug. ***

Comment 2 Karel Zak 2006-10-24 10:26:22 UTC

Your xterm runs with non-root permissions and it can't send SIGHUP to the su
which runs with root permissions. This problem cannot be fixed. Sorry.

BTW, I you want to run the top with root permissions in a xterm, you can start
it by: su -c "xterm -e top".

Comment 3 Albert Cahalan 2006-12-25 19:08:15 UTC

This is actually a kernel bug.

The xterm itself DOES NOT send the SIGHUP. The SIGHUP is sent by the kernel
itself, automatically as the xterm dies.

On a real BSD system, where this whole idea originates AFAIK, the signal will
always get sent. Linux used to do this too, but broke when some other changes
went it. I believe it was the SE Linux merge that caused this, but note that the
bug shows up even when SE Linux is not compiled into the kernel. In other words,
I think the SE Linux changes broke non-SE systems.

I've been unable to find the bug from an examination of the kernel source. As
far as I can tell, the signal is sent in a way that will bypass permission
checking. Somehow though, this isn't working.

Please reopen this bug and assign it to the kernel.