Bug 837621

Summary:	sa -u cause Segmentation fault
Product:	Red Hat Enterprise Linux 6	Reporter:	Branislav Náter <bnater>
Component:	psacct	Assignee:	Jaromír Cápík <jcapik>
Status:	CLOSED CURRENTRELEASE	QA Contact:	Branislav Náter <bnater>
Severity:	high	Docs Contact:
Priority:	high
Version:	6.3	CC:	jpallich, ovasik
Target Milestone:	rc	Keywords:	ZStream
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2013-11-14 05:41:11 EST	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---
Bug Depends On:
Bug Blocks:	838998

Description Branislav Náter 2012-07-04 08:09:56 EDT

Description of problem:
Running sa -u (print also user id) command cause Segmentation fault on i386 and s390x archs and produce incorect output on x86_64.

Version-Release number of selected component (if applicable):
psacct-6.3.2-63.el6_3.2

How reproducible:
always

Steps to Reproduce:
1. run "sa -u"
  
Actual results on different archs:
ppc64:  root       0.00 cpu       62k mem accton
x86_64: 13179672   0.00 cpu      979k mem accton
i386:   Segmentation fault
s390x:  Segmentation fault

Expected results:
root       0.00 cpu       62k mem accton

Additional info:
Output seems to be correct only on ppc64 architecture.

Traceback from i386:
(gdb) t a a bt

Thread 1 (process 14151):
#0  0x0804a3f4 in parse_acct_entries () at sa.c:1254
#1  0x0804abea in main (argc=2, argv=0xbffff364) at sa.c:585

Traceback from s390x:
(gdb) t a a bt

Thread 1 (process 8385):
#0  0x0000000080002b96 in parse_acct_entries () at sa.c:1254
#1  0x0000000080003312 in main (argc=<value optimized out>, argv=<value optimized out>) at sa.c:585

Comment 1 Jaromír Cápík 2012-07-10 10:00:21 EDT

rec->ac_uid is invalid for some reason ... needs a deeper analysis.

Comment 2 Jaromír Cápík 2012-07-10 10:19:56 EDT

Some parts of the code are really very strange ...

----------------------------------------------------

#if __BYTE_ORDER == __BIG_ENDIAN
# define ACCT_BYTEORDER 0x80        /* Accounting file is big endian.  */
#else
# define ACCT_BYTEORDER 0x00        /* Accounting file is little endian.  */
#endif

...
...

if ((version % ACCT_BYTEORDER) != 3) {

----------------------------------------------------

Anything modulo 0 must throw SIGFPE ... 

Fortunately the structure seems to be BIG ENDIAN even on LITTLE_ENDIAN architectures ... this has been redesigned in the new upstream version.

Comment 3 Jaromír Cápík 2012-07-10 10:30:49 EDT

It seems I've found the issue. The ac_uid field needs to be read from the rec_v3 variable in case of acct_v3 structures. The acct structure version detection is a bit nonsense, but it doesn't cause the trouble. It was a human factor ... a cut'n'paste issue. Gonna test that on other architectures.