On Thu, Mar 05, 2015 at 12:05:24 +0100, Michal Privoznik wrote:
This patch alone does not make much sense, I know. But it prepares ground for next patch which when looking up a network in the object list will not lock each network separately when accessing its definition. Therefore we must have all the places changing network definition lock the list.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> --- src/conf/network_conf.c | 9 ++++++++- src/conf/network_conf.h | 3 ++- src/network/bridge_driver.c | 4 ++-- 3 files changed, 12 insertions(+), 4 deletions(-)
diff --git a/src/conf/network_conf.c b/src/conf/network_conf.c index 3d318ce..007cebb 100644 --- a/src/conf/network_conf.c +++ b/src/conf/network_conf.c @@ -537,12 +537,19 @@ virNetworkObjSetDefTransient(virNetworkObjPtr network, bool live) * This *undoes* what virNetworkObjSetDefTransient did. */ void
I've looked through the next patch and you are basically trying to make the name and UUID pointers for domain immutable or at leas write locked ...
-virNetworkObjUnsetDefTransient(virNetworkObjPtr network) +virNetworkObjUnsetDefTransient(virNetworkObjListPtr nets, + virNetworkObjPtr network) { if (network->newDef) { + virObjectRef(network); + virObjectUnlock(network); + virObjectLock(nets); + virObjectLock(network); + virObjectUnref(network);
But I don't really like pulling in the complexity into this helper.
virNetworkDefFree(network->def); network->def = network->newDef; network->newDef = NULL; + virObjectUnlock(nets); } }
While I like the idea, I'd rather see a conversion to R/W locks or making of the name and UUID pointers immutable than this hack. Peter