On Wed, Jan 23, 2013 at 11:18:04AM -0700, Eric Blake wrote:
On 01/23/2013 11:13 AM, Daniel P. Berrange wrote:
> On Wed, Jan 23, 2013 at 11:12:03AM -0700, Eric Blake wrote:
>> On 01/23/2013 11:02 AM, Richard W.M. Jones wrote:
>>> On Wed, Jan 23, 2013 at 05:06:35PM +0000, Daniel P. Berrange wrote:
>>>>> + memset(nodeinfo, 0, sizeof(nodeinfo));
>>>>> +
>>>>
>>>> Of course i meant sizeof(*nodeinfo)
>>>
>>> Yikes. Doesn't libvirt have some type of VIR_* macro to either catch
>>> that error systematically or to allocate pre-cleared memory?
>>
>> VIR_MALLOC guarantees pre-cleared memory, but in this case, nodeinfo was
>> passed in by the caller rather than something that we control directly.
>
> Of course there are plenty of places where we stack allocate
> stuff which we then memset(), so I believe a VIR_ZERO is worth
> while
Agreed. Also, in the case of virNodeGetInfo(), I wonder if we should
just hoist the VIR_ZERO into the libvirt.c entry point, rather than
making every driver have to worry about clearing out potentially
uninitialized incoming data on all error paths.
Makes sense - that's what we do for virDomainGetInfo. Might be other
APIs neeeding similar change
Daniel
--
|:
http://berrange.com -o-
http://www.flickr.com/photos/dberrange/ :|
|:
http://libvirt.org -o-
http://virt-manager.org :|
|:
http://autobuild.org -o-
http://search.cpan.org/~danberr/ :|
|:
http://entangle-photo.org -o-
http://live.gnome.org/gtk-vnc :|