Bug 810613

Summary:	Memory leak in virInitialize
Product:	[Community] Virtualization Tools	Reporter:	Richard W.M. Jones <rjones>
Component:	libvirt	Assignee:	Libvirt Maintainers <libvirt-maint>
Status:	CLOSED NOTABUG	QA Contact:
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	unspecified	CC:	acathrow, berrange, crobinso, dallan, mbooth, rjones, virt-maint, xen-maint
Target Milestone:	---
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2013-12-13 12:41:08 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Richard W.M. Jones 2012-04-07 08:25:02 UTC

Description of problem:

==24471== Memcheck, a memory error detector
==24471== Copyright (C) 2002-2011, and GNU GPL'd, by Julian Seward et al.
==24471== Using Valgrind-3.7.0 and LibVEX; rerun with -h for copyright info
==24471== Command: /home/rjones/d/libguestfs-1.16/df/.libs/lt-virt-df
==24471== Parent PID: 24470
==24471== 
==24471== 
==24471== HEAP SUMMARY:
==24471==     in use at exit: 130,697 bytes in 1,295 blocks
==24471==   total heap usage: 4,367 allocs, 3,072 frees, 104,142,842 bytes allocated
==24471== 
==24471== 21 bytes in 1 blocks are definitely lost in loss record 70 of 185
==24471==    at 0x4A0884D: malloc (vg_replace_malloc.c:263)
==24471==    by 0x3B16885B51: strdup (in /usr/lib64/libc-2.15.so)
==24471==    by 0x3B28805CBB: p11_kit_registered_module_to_name (in /usr/lib64/libp11-kit.so.0.0.0)
==24471==    by 0x3B2A048294: gnutls_pkcs11_init (in /usr/lib64/libgnutls.so.26.22.0)
==24471==    by 0x3B2A031CE1: gnutls_global_init (in /usr/lib64/libgnutls.so.26.22.0)
==24471==    by 0x3B326CE219: virInitialize (in /usr/lib64/libvirt.so.0.9.10)
==24471==    by 0x3B326CEF56: virConnectOpenReadOnly (in /usr/lib64/libvirt.so.0.9.10)
==24471==    by 0x4045B6: get_domains_from_libvirt (domains.c:111)
==24471==    by 0x40308A: main (main.c:264)
==24471== 
==24471== LEAK SUMMARY:
==24471==    definitely lost: 21 bytes in 1 blocks
==24471==    indirectly lost: 0 bytes in 0 blocks
==24471==      possibly lost: 0 bytes in 0 blocks
==24471==    still reachable: 130,327 bytes in 1,276 blocks
==24471==         suppressed: 349 bytes in 18 blocks
==24471== Reachable blocks (those to which a pointer was found) are not shown.
==24471== To see them, rerun with: --leak-check=full --show-reachable=yes
==24471== 
==24471== For counts of detected and suppressed errors, rerun with: -v
==24471== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 20 from 20)


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

$ rpm -qf /usr/lib64/libp11-kit.so.0.0.0
p11-kit-0.12-1.fc17.x86_64
$ rpm -qf /usr/lib64/libgnutls.so.26.22.0
gnutls-2.12.17-1.fc17.x86_64
$ rpm -qf /usr/lib64/libvirt.so.0.9.10
libvirt-client-0.9.10-2.fc17.x86_64
$ ./run ./fish/guestfish version
major: 1
minor: 16
release: 15
extra: 

(also happens on 1.17.25)

How reproducible:

100% on a machine that has 1 libvirt domain
0% on a basically identical machine that has > 1 domains

Comment 1 Richard W.M. Jones 2012-04-07 08:31:49 UTC

This is the reproducing command:

./run valgrind --leak-check=full --suppressions=tests/extra/suppressions ./df/virt-df -c qemu:///system

Comment 2 Richard W.M. Jones 2012-04-07 09:04:46 UTC

Bug also occurs with
libvirt-client-0.9.11-1.fc17.x86_64
from Fedora updates-testing.

Comment 3 Richard W.M. Jones 2012-04-07 09:38:12 UTC

Looks like a bug in libvirt.  I raised the issue on libvir-list:

https://www.redhat.com/archives/libvir-list/2012-April/msg00309.html

Comment 4 Richard W.M. Jones 2012-04-11 17:19:20 UTC

Reassigning to libvirt (note that the real bug is lack
of thread safety in gnutls).

For discussion, see the thread in comment 3.

Comment 5 Daniel Berrangé 2012-04-11 17:46:53 UTC

Well there's nothing we can really do about this in libvirt.  My suggestion is to add a wildcard against  'gnutls_global_init' to your valgrind suppressions file.

Comment 6 Dave Allan 2012-04-12 14:01:34 UTC

Should this be moved to gnutls or closed as CANTFIX?