On Sat, Dec 19, 2015 at 07:09:02PM +0800, Antonio Quartulli wrote:
Am I wrong or also the commit message is misleading? Looks like it is "fixing" a bogus lockdep check...
Yes and no. The lockdep checks placed are not bogus as is. They are fine for batadv_iv_ogm_schedule() vs. batadv_update_mtu(), so no need to remove it.
but as far as I can say this patch is making sure that the lock is held in that particular point.
There is no locking issue from batadv_mcast_free() here as it is in the freeing phase of the soft-iface. The soft-iface is supposed to be out-of-scope for anything else by now. This patch therefore can't be a fix for a race regarding mcast.mla_list.
However it does fix an issue with a lockdep warning. Which makes it suitable for stable to at least remove that splat for v4.3 users.
Regards, Linus