Re: [libvirt] [PATCH] virsh.c: avoid leak on OOM error path
Jim Meyering wrote:
No one really cares if we leak memory while dying, but who knows...
freeing that buffer may let us go down more gracefully.
FYI, the leak is triggered when virFileReadAll succeeds
(it allocates "buffer"), yet xmlNewDoc fails:
if (virFileReadAll(from, VIRSH_MAX_XML_FILE, &buffer) < 0)
return FALSE;
doc = xmlNewDoc(NULL);
if (doc == NULL)
goto no_memory;
>From 03c7a44e3d5b2e7c992bebc98fc8c6a7bf63881e Mon Sep 17 00:00:00 2001
From: Jim Meyering <meyering(a)redhat.com>
Date: Fri, 19 Feb 2010 18:03:41 +0100
Subject: [PATCH] virsh.c: avoid leak on OOM error path
* tools/virsh.c (cmdCPUBaseline): Also free "buffer" upon OOM.
---
tools/virsh.c | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/tools/virsh.c b/tools/virsh.c
index dd916f3..8756a7a 100644
--- a/tools/virsh.c
+++ b/tools/virsh.c
@@ -7139,6 +7139,7 @@ cleanup:
return ret;
no_memory:
+ VIR_FREE(buffer);
vshError(ctl, "%s", _("Out of memory"));
ret = FALSE;
return ret;
--
The above is correct, but there's another leak in the same function,
so I've amended the patch to also free the "list" buffer.
"list" is allocated in the for-loop.
If on the 2nd or subsequent iteration of that loop we take
the "goto no_memory", we'd leak that buffer.
From 3d97412799c1b5bfedc647059c7d3b18e763f0bc Mon Sep 17 00:00:00 2001
From: Jim Meyering <meyering(a)redhat.com>
Date: Fri, 19 Feb 2010 18:03:41 +0100
Subject: [PATCH] virsh.c: avoid leak on OOM error path
* tools/virsh.c (cmdCPUBaseline): Also free "buffer" and "list" upon
OOM.
---
tools/virsh.c | 2 ++
1 files changed, 2 insertions(+), 0 deletions(-)
diff --git a/tools/virsh.c b/tools/virsh.c
index dd916f3..c8ae9f2 100644
--- a/tools/virsh.c
+++ b/tools/virsh.c
@@ -7139,6 +7139,8 @@ cleanup:
return ret;
no_memory:
+ VIR_FREE(list);
+ VIR_FREE(buffer);
vshError(ctl, "%s", _("Out of memory"));
ret = FALSE;
return ret;
--
1.7.0.233.g05e1a