[libvirt] [PATCH] network: avoid memory leak on cleanup

[Matthias Bolte]

2011/7/15 Alex Jia <ajia at redhat.com>:
> On 07/15/2011 03:03 PM, Alex Jia wrote:
>
> On 07/15/2011 02:49 PM, Matthias Bolte wrote:
>
> 2011/7/15  <ajia at redhat.com>:
>
> * src/network/bridge_driver.c: Fix memory leak on cleanup section from
>  networkGetBridgeName function.
> ---
>  src/network/bridge_driver.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletions(-)
>
> diff --git a/src/network/bridge_driver.c b/src/network/bridge_driver.c
> index 0a12bc0..59e780d 100644
> --- a/src/network/bridge_driver.c
> +++ b/src/network/bridge_driver.c
> @@ -2474,7 +2474,8 @@ static char *networkGetBridgeName(virNetworkPtr net) {
>  cleanup:
>     if (network)
>         virNetworkObjUnlock(network);
> -    return bridge;
> +    VIR_FREE(bridge);
> +    return NULL;
>  }
>
>  static int networkGetAutostart(virNetworkPtr net,
>
> NACK. Now networkGetBridgeName returns NULL always, that's wrong.
>
> Why do you think that there is a leak?
>
> Detected in valgrind run:
> ==9020== 7 bytes in 1 blocks are definitely lost in loss record 1 of 26
> ==9020==    at 0x4A05FDE: malloc (vg_replace_malloc.c:236)
> ==9020==    by 0x386A314B3D: xdr_string (in /lib64/libc-2.12.so)
> ==9020==    by 0x4CFC0AD: xdr_remote_nonnull_string (remote_protocol.c:30)
> ==9020==    by 0x4CFCC08: xdr_remote_network_get_bridge_name_ret
> (remote_protocol.c:1999)
> ==9020==    by 0x4D06FC1: virNetMessageDecodePayload (virnetmessage.c:286)
> ==9020==    by 0x4D03235: virNetClientProgramCall
> (virnetclientprogram.c:318)
> ==9020==    by 0x4CE7262: call (remote_driver.c:3925)
> ==9020==    by 0x4CED8D2: remoteNetworkGetBridgeName
> (remote_client_bodies.h:3384)
> ==9020==    by 0x4CC494E: virNetworkGetBridgeName (libvirt.c:8503)
> ==9020==    by 0x40F654: cmdNetworkInfo (virsh.c:5015)
> ==9020==    by 0x410D02: vshCommandRun (virsh.c:12738)
> ==9020==    by 0x41F2D5: main (virsh.c:14084)

The caller of virNetworkGetBridgeName has to free the allocated string:

diff --git a/tools/virsh.c b/tools/virsh.c
index bd6fea7..5c48f0c 100644
--- a/tools/virsh.c
+++ b/tools/virsh.c
@@ -5011,6 +5011,7 @@ cmdNetworkInfo(vshControl *ctl, const vshCmd *cmd)
     if (bridge)
         vshPrint(ctl, "%-15s %s\n", _("Bridge:"), bridge);

+    VIR_FREE(bridge);
     virNetworkFree(network);
     return true;
 }

> ==9020== 10 bytes in 1 blocks are definitely lost in loss record 3 of 26
> ==9020==    at 0x4A05FDE: malloc (vg_replace_malloc.c:236)
> ==9020==    by 0x386A27F8A1: strdup (in /lib64/libc-2.12.so)
> ==9020==    by 0x4CF508B: doRemoteOpen (remote_driver.c:364)
> ==9020==    by 0x4CF6E8F: remoteOpen (remote_driver.c:800)
> ==9020==    by 0x4CCB384: do_open (libvirt.c:1054)
> ==9020==    by 0x4CCBEB5: virConnectOpenAuth (libvirt.c:1280)
> ==9020==    by 0x410BC0: vshReconnect (virsh.c:589)
> ==9020==    by 0x410DCF: vshCommandRun (virsh.c:12733)
> ==9020==    by 0x41F2D5: main (virsh.c:14084)
> ==9020==
> ==9020== 56 (24 direct, 32 indirect) bytes in 1 blocks are definitely lost
> in loss record 17 of 26
> ==9020==    at 0x4A04A28: calloc (vg_replace_malloc.c:467)
> ==9020==    by 0x4C686ED: virAlloc (memory.c:101)
> ==9020==    by 0x4C96870: virDomainEventStateNew (domain_event.c:578)
> ==9020==    by 0x4CF5A8E: doRemoteOpen (remote_driver.c:658)
> ==9020==    by 0x4CF6E8F: remoteOpen (remote_driver.c:800)
> ==9020==    by 0x4CCB384: do_open (libvirt.c:1054)
> ==9020==    by 0x4CCBEB5: virConnectOpenAuth (libvirt.c:1280)
> ==9020==    by 0x410BC0: vshReconnect (virsh.c:589)
> ==9020==    by 0x410DCF: vshCommandRun (virsh.c:12733)
> ==9020==    by 0x41F2D5: main (virsh.c:14084)
>
> Note: remote memory leak are a known bug:
> https://bugzilla.redhat.com/show_bug.cgi?id=690734,
>       so this patch only fix network memory leak.

Judging by a quick look at the code this two leaks should not be
possible? Are you working on current git? Anyway, I'll leave those to
you.

-- 
Matthias Bolte
http://photron.blogspot.com