]> git.ipfire.org Git - thirdparty/libvirt.git/commit
remoteClientCloseFunc: Don't mangle connection object refcount
authorMichal Privoznik <mprivozn@redhat.com>
Thu, 3 Sep 2015 10:11:53 +0000 (12:11 +0200)
committerMichal Privoznik <mprivozn@redhat.com>
Thu, 3 Sep 2015 15:46:14 +0000 (17:46 +0200)
commit4268b1f1027eab225d78d8a842fee1c48f4bd8ac
tree953c941301d249aa3449ac2f6cb55b5a990bcaa5
parentd055989083df4bf68eb1388d327ebffb3501bb83
remoteClientCloseFunc: Don't mangle connection object refcount

Well, in 8ad126e6 we tried to fix a memory corruption problem.
However, the fix was not as good as it could be. I mean, the
commit has one line more than it should. I've noticed this output
just recently:

  # ./run valgrind --leak-check=full --show-reachable=yes ./tools/virsh domblklist gentoo
  ==17019== Memcheck, a memory error detector
  ==17019== Copyright (C) 2002-2013, and GNU GPL'd, by Julian Seward et al.
  ==17019== Using Valgrind-3.10.1 and LibVEX; rerun with -h for copyright info
  ==17019== Command: /home/zippy/work/libvirt/libvirt.git/tools/.libs/virsh domblklist gentoo
  ==17019==
  Target     Source
  ------------------------------------------------
  fda        /var/lib/libvirt/images/fd.img
  vda        /var/lib/libvirt/images/gentoo.qcow2
  hdc        /home/zippy/tmp/install-amd64-minimal-20150402.iso

  ==17019== Thread 2:
  ==17019== Invalid read of size 4
  ==17019==    at 0x4EFF5B4: virObjectUnref (virobject.c:258)
  ==17019==    by 0x5038CFF: remoteClientCloseFunc (remote_driver.c:552)
  ==17019==    by 0x5069D57: virNetClientCloseLocked (virnetclient.c:685)
  ==17019==    by 0x506C848: virNetClientIncomingEvent (virnetclient.c:1852)
  ==17019==    by 0x5082136: virNetSocketEventHandle (virnetsocket.c:1913)
  ==17019==    by 0x4ECD64E: virEventPollDispatchHandles (vireventpoll.c:509)
  ==17019==    by 0x4ECDE02: virEventPollRunOnce (vireventpoll.c:658)
  ==17019==    by 0x4ECBF00: virEventRunDefaultImpl (virevent.c:308)
  ==17019==    by 0x130386: vshEventLoop (vsh.c:1864)
  ==17019==    by 0x4F1EB07: virThreadHelper (virthread.c:206)
  ==17019==    by 0xA8462D3: start_thread (in /lib64/libpthread-2.20.so)
  ==17019==    by 0xAB441FC: clone (in /lib64/libc-2.20.so)
  ==17019==  Address 0x139023f4 is 4 bytes inside a block of size 240 free'd
  ==17019==    at 0x4C2B1F0: free (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
  ==17019==    by 0x4EA8949: virFree (viralloc.c:582)
  ==17019==    by 0x4EFF6D0: virObjectUnref (virobject.c:273)
  ==17019==    by 0x4FE74D6: virConnectClose (libvirt.c:1390)
  ==17019==    by 0x13342A: virshDeinit (virsh.c:406)
  ==17019==    by 0x134A37: main (virsh.c:950)

The problem is, when registering remoteClientCloseFunc(), it's
conn->closeCallback which is ref'd. But in the function itself
it's conn->closeCallback->conn what is unref'd. This is causing
imbalance in reference counting. Moreover, there's no need for
the remote driver to increase/decrease conn refcount since it's
not used anywhere. It's just merely passed to client registered
callback. And for that purpose it's correctly ref'd in
virConnectRegisterCloseCallback() and then unref'd in
virConnectUnregisterCloseCallback().

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
(cherry picked from commit e68930077034f786e219bdb015f8880dbc5a246f)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
src/remote/remote_driver.c