On 04/08/2013 10:45 AM, Daniel P. Berrange wrote:
On Mon, Apr 08, 2013 at 12:35:40PM -0400, Milos Vyletel wrote:
Even though http://libvirt.org/formatdomain.html#elementsMetadata states that it requires RFC4122 compliance UUIDs that are generated by virUUIDGenerate() are not. Neither does virUUIDIsValid() check for RFC4122 compliance. Following patch modifies virUUIDGenerate() to generate valid UUIDs and adds check to virUUIDIsValid() to validate UUIDs.
+ /* + * RFC4122 defines version 1 to 5 (section 4.1.3) + * RFC4122 defined variant is desribed in section 4.1.1 + */ + version = (uuid[6] >> 4); + variant = (uuid[8] >> 6); + if (!(version > 0 && version <= 5) || variant != 2) + return 0; + c = uuid[0];
for (i = 1; i < VIR_UUID_BUFLEN; i++)
but NACk to this part
What you're checking here is just one possible valid scheme for UUIDs. We shouldn't reject UUIDs just because they use a different scheme than the one we do.
Furthermore, if we took this hunk, but a user is running a guest created by a previous version of libvirt that happened to generate an invalid UUID, we would lose the ability to manage that older guest. It _might_ be appropriate to warn the user when a UUID is not valid according to the schemes we recognize, but it must be a warning and not a fatal error; furthermore, if we do add such a warning, we'd need to recognize ALL of the schemes that are valid in the RFCs, not just the particular scheme we use when generating a uuid ourselves, so as to minimize false negative printouts of the warning. -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org