Bug 529028

Summary:	bfd_get_section_by_name() not working correctly on i686 machine
Product:	Red Hat Enterprise Linux 5	Reporter:	William Cohen <wcohen>
Component:	binutils	Assignee:	Andreas Schwab <schwab>
Status:	CLOSED ERRATA	QA Contact:	qe-baseos-tools-bugs
Severity:	medium	Docs Contact:
Priority:	medium
Version:	5.4	CC:	ebachalo, jan.kratochvil, mnowak, mschmidt, pmuller
Target Milestone:	rc	Keywords:	Regression
Target Release:	5.5
Hardware:	All
OS:	Linux
Whiteboard:
Fixed In Version:	binutils-2.17.50.0.6-13.el5	Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2010-03-30 09:04:36 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:
Bug Depends On:
Bug Blocks:	527679

Description William Cohen 2009-10-14 17:14:56 UTC

Description of problem:

OProfile builds with binutils-devel. On the i686 (32-bit) platform bfd_get_section_by_name() is returning incorrect information about the debuginfo section.

There is a regression in binutils-devel. Rebuilt oprofile with older 
binutils-devel-2.17.50.0.6-9.el5.i386.rpm and "opreport -l" works fine. Build oprofile with binutils-devel-2.17.50.0.6-12.el5 and "opreport -l" doesn't work.


Version-Release number of selected component (if applicable):
binutils-devel.i386 0:2.17.50.0.6-12.el5
oprofile-0.9.4-11.el5

How reproducible:

Run the following script:
opcontrol --init
opcontrol --setup --no-vmlinux
opcontrol --start
sleep 3
opcontrol --dump
opreport -l && echo "opreport ended successfully" || echo "ERROR: opreport
returned $?"
opcontrol --shutdown > /dev/null

Actual results:
The output ends with:
...
CPU: P4 / Xeon, speed 2400 MHz (estimated)
Counted GLOBAL_POWER_EVENTS events (time during which processor is not stopped)
with a unit mask of 0x01 (mandatory) count 100000
bfd_get_section_contents:get_debug:: Bad value
ERROR: opreport returned 2

Expected results:
A long detailed per-symbol profile and "opreport ended successfully" at the
end.


Additional info:

Can set a break point in get_debug_link_info() in opreport and then step through the function. After bfd_get_section_by_name() runs can examine the contents of sect. This has a length of 0 which is not correct.

Comment 1 William Cohen 2009-10-15 21:47:08 UTC

The binutils-devel-2.17.50.0.6-10 seems to be where the problem first appears, so something in the following changes caused oprofile to break:

PatchSet 124 
Date: 2009/04/08 15:16:10
Author: dvlasenk
Branch: HEAD
Tag: binutils-2_17_50_0_6-10_el5 
Branches: 
Log:
- Improve support of ELF64, particularly on 32-bit architectures.
  Resolves: RHBZ #457189.
- Fix rare linker corner case when it produces large zero-filled hole
  in the executable.
  Resolves: RHBZ #458301.
- Make "strings -n non_number" error message clearer.
  Resolves: RHBZ #480009.
- Fix man page of c++filt to show --[no-]strip-underscore option correctly.
  Resolves: RHBZ #485194.
- While at it, fix "makeinfo >= 4.4" check in configure.

Members: 
	binutils-2.17.50.0.6-build-fix-configure.patch:INITIAL->1.1 
	binutils-2.17.50.0.6-rh457189-minimal-64bit-i386.patch:INITIAL->1.1 
	binutils-2.17.50.0.6-rh458301-2m-hole.patch:INITIAL->1.1 
	binutils-2.17.50.0.6-rh480009.patch:INITIAL->1.1 
	binutils-2.17.50.0.6-rh485194.patch:INITIAL->1.1 
	binutils.spec:1.113->1.114

Comment 2 Andreas Schwab 2009-10-16 16:07:45 UTC

This specfile patch fixes the bug (oprofile needs to be recompiled afterwards):

--- binutils.spec.~1.116.~	2009-04-17 14:44:09.000000000 +0200
+++ binutils.spec	2009-10-16 18:02:59.000000000 +0200
@@ -185,9 +185,6 @@ rm -rf %{buildroot}%{_prefix}/%{_target_
 
 %ifarch %{ix86} x86_64 ppc ppc64 s390 s390x sparc sparc64
 sed -i -e '/^#include "ansidecl.h"/{p;s~^.*$~#include <bits/wordsize.h>~;}' \
-%ifarch %{ix86} x86_64
-    -e 's/^#define BFD_ARCH_SIZE \(32\|64\) *$/#define BFD_ARCH_SIZE __WORDSIZE/' \
-%endif
     -e 's/^#define BFD_DEFAULT_TARGET_SIZE \(32\|64\) *$/#define BFD_DEFAULT_TARGET_SIZE __WORDSIZE/' \
     -e 's/^#define BFD_HOST_64BIT_LONG [01] *$/#define BFD_HOST_64BIT_LONG (__WORDSIZE == 64)/' \
     -e 's/^#define BFD_HOST_64_BIT \(long \)\?long *$/#if __WORDSIZE == 32\

Comment 3 William Cohen 2009-10-16 16:58:43 UTC

Yes, this change in the spec file resolves the problem in oprofile.

Comment 4 Eric Bachalo 2009-10-22 16:11:28 UTC

This patch is required to fix bz 527679, a OProfile BZ targeted for RHEL 5.5

Comment 15 errata-xmlrpc 2010-03-30 09:04:36 UTC

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHBA-2010-0304.html