Coverity discovered that the free(domain->os_info.fv.arch) and then
usage later on during get_default_machine() and get_default_emulator()
calls could result in using free()'d memory.
If the 'cu_get_str_prop() failed or capsinfo == NULL, then the fv.arch
wouldn't necessarily be strdup()'d.
Passing a NULL os_info.fv_arch into the get*() API's is fine since
they'll call findDomainInfo() which can handle a NULL arch value.
Also added an initialization of val just to be safe. I don't think it's
necessary though.
---
NOTE:
I found this during a Coverity run applying the endianness patches. For
some reason Coverity "woke up" and saw this even though it hasn't found
this issue in a couple months of runs since the changes to this module
were made. See commit id '117dabb9'.
src/Virt_VirtualSystemManagementService.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/Virt_VirtualSystemManagementService.c b/src/Virt_VirtualSystemManagementService.c
index d51f230..9f8b5b9 100644
--- a/src/Virt_VirtualSystemManagementService.c
+++ b/src/Virt_VirtualSystemManagementService.c
@@ -464,7 +464,7 @@ static int fv_vssd_to_domain(CMPIInstance *inst,
{
int ret = 1;
int retr;
- const char *val;
+ const char *val = NULL;
const char *domtype = NULL;
const char *ostype = "hvm";
struct capabilities *capsinfo = NULL;
@@ -494,6 +494,7 @@ static int fv_vssd_to_domain(CMPIInstance *inst,
}
free(domain->os_info.fv.arch);
+ domain->os_info.fv.arch = NULL;
retr = cu_get_str_prop(inst, "Arch", &val);
if (retr != CMPI_RC_OK) {
if (capsinfo != NULL) { /* set default */
--
1.8.3.1
