svn commit: r355102 - head/sys/netinet6

[Gleb Smirnoff]

  Ryan,

On Mon, Nov 25, 2019 at 10:25:10PM +0000, Ryan Libby wrote:
R> Author: rlibby
R> Date: Mon Nov 25 22:25:10 2019
R> New Revision: 355102
R> URL: https://svnweb.freebsd.org/changeset/base/355102
R> 
R> Log:
R>   in6_joingroup_locked: need if_addr_lock around in6m_disconnect_locked
R>   
R>   It looks like the call that requires the lock was introduced in r337866.
R>   
R>   Reviewed by:	hselasky
R>   Sponsored by:	Dell EMC Isilon
R>   Differential Revision:	https://reviews.freebsd.org/D20739
R> 
R> Modified:
R>   head/sys/netinet6/in6_mcast.c
R> 
R> Modified: head/sys/netinet6/in6_mcast.c
R> ==============================================================================
R> --- head/sys/netinet6/in6_mcast.c	Mon Nov 25 21:21:37 2019	(r355101)
R> +++ head/sys/netinet6/in6_mcast.c	Mon Nov 25 22:25:10 2019	(r355102)
R> @@ -1267,6 +1267,7 @@ out_in6m_release:
R>  		struct epoch_tracker et;
R>  
R>  		CTR2(KTR_MLD, "%s: dropping ref on %p", __func__, inm);
R> +		IF_ADDR_WLOCK(ifp);
R>  		NET_EPOCH_ENTER(et);
R>  		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
R>  			if (ifma->ifma_protospec == inm) {
R> @@ -1277,6 +1278,7 @@ out_in6m_release:
R>  		in6m_disconnect_locked(&inmh, inm);
R>  		in6m_rele_locked(&inmh, inm);
R>  		NET_EPOCH_EXIT(et);
R> +		IF_ADDR_WUNLOCK(ifp);
R>  	} else {
R>  		*pinm = inm;
R>  	}

It looks like after this change the epoch entrance in this block is superfluous
and can be removed.

-- 
Gleb Smirnoff