| Message ID | 1291761150-29818-18-git-send-email-linus.luessing@saxnet.de |
|---|---|
| State | Superseded, archived |
| Headers | show |
On Tue, Dec 07, 2010 at 11:32:28PM +0100, Linus L??ssing wrote: > We may only optimize the multicast packet flow, if an mcast_mode has been > activated and if we are a multicast receiver of the same group. Otherwise > flood the multicast packet without optimizations. > > This allows us to still flood multicast packets of protocols where it is > not easily possible for a multicast sender to be a multicast receiver of > the same group instead of dropping them (for instance IPv6 NDP). > > This commit therefore also makes IPv6 usable again, if the > proact_tracking multicast mode has been activated. > > Signed-off-by: Linus L??ssing <linus.luessing@saxnet.de> > --- > multicast.c | 25 +++++++++++++++++++++++++ > multicast.h | 1 + > soft-interface.c | 2 +- > 3 files changed, 27 insertions(+), 1 deletions(-) > > diff --git a/multicast.c b/multicast.c > index 042d392..1f84f7c 100644 > --- a/multicast.c > +++ b/multicast.c > @@ -107,6 +107,31 @@ void mcast_tracker_reset(struct bat_priv *bat_priv) > start_mcast_tracker(bat_priv); > } > > +inline int mcast_may_optimize(uint8_t *dest, struct net_device *soft_iface) { > + MC_LIST *mc_entry; > + unsigned long flags; > + struct bat_priv *bat_priv = netdev_priv(soft_iface); > + int mcast_mode = atomic_read(&bat_priv->mcast_mode); > + > --- a/soft-interface.c > +++ b/soft-interface.c > @@ -391,7 +391,7 @@ int interface_tx(struct sk_buff *skb, struct net_device *soft_iface) > > if (is_broadcast_ether_addr(ethhdr->h_dest)) > bcast_dst = true; > - else if (atomic_read(&bat_priv->mcast_mode) == MCAST_MODE_PROACT_TRACKING) > + else if (mcast_may_optimize(ethhdr->h_dest, soft_iface)) > mcast_dst = true; > else > bcast_dst = true; You define mcast_may_optimize as inline, and then use it from a different file. This makes the inline pointless, as far as i know. Andrew
diff --git a/multicast.c b/multicast.c index 042d392..1f84f7c 100644 --- a/multicast.c +++ b/multicast.c @@ -107,6 +107,31 @@ void mcast_tracker_reset(struct bat_priv *bat_priv) start_mcast_tracker(bat_priv); } +inline int mcast_may_optimize(uint8_t *dest, struct net_device *soft_iface) { + MC_LIST *mc_entry; + unsigned long flags; + struct bat_priv *bat_priv = netdev_priv(soft_iface); + int mcast_mode = atomic_read(&bat_priv->mcast_mode); + + if (mcast_mode != MCAST_MODE_PROACT_TRACKING) + return 0; + + /* Still allow flooding of multicast packets of protocols where it is + * not easily possible for a multicast sender to be a multicast + * receiver of the same group (for instance IPv6 NDP) */ + MC_LIST_LOCK(soft_iface, flags); + netdev_for_each_mc_addr(mc_entry, soft_iface) { + if (memcmp(dest, mc_entry->MC_LIST_ADDR, ETH_ALEN)) + continue; + + MC_LIST_UNLOCK(soft_iface, flags); + return 1; + } + MC_LIST_UNLOCK(soft_iface, flags); + + return 0; +} + static inline int get_remaining_timeout( struct mcast_forw_nexthop_entry *nexthop_entry, struct bat_priv *bat_priv) diff --git a/multicast.h b/multicast.h index 6dcf537..630a0ae 100644 --- a/multicast.h +++ b/multicast.h @@ -27,6 +27,7 @@ int mcast_tracker_interval_set(struct net_device *net_dev, char *buff, int mcast_tracker_timeout_set(struct net_device *net_dev, char *buff, size_t count); void mcast_tracker_reset(struct bat_priv *bat_priv); +int mcast_may_optimize(uint8_t *dest, struct net_device *soft_iface); void route_mcast_tracker_packet( struct mcast_tracker_packet *tracker_packet, int tracker_packet_len, struct bat_priv *bat_priv); diff --git a/soft-interface.c b/soft-interface.c index 2a5a728..444028d 100644 --- a/soft-interface.c +++ b/soft-interface.c @@ -391,7 +391,7 @@ int interface_tx(struct sk_buff *skb, struct net_device *soft_iface) if (is_broadcast_ether_addr(ethhdr->h_dest)) bcast_dst = true; - else if (atomic_read(&bat_priv->mcast_mode) == MCAST_MODE_PROACT_TRACKING) + else if (mcast_may_optimize(ethhdr->h_dest, soft_iface)) mcast_dst = true; else bcast_dst = true;
We may only optimize the multicast packet flow, if an mcast_mode has been activated and if we are a multicast receiver of the same group. Otherwise flood the multicast packet without optimizations. This allows us to still flood multicast packets of protocols where it is not easily possible for a multicast sender to be a multicast receiver of the same group instead of dropping them (for instance IPv6 NDP). This commit therefore also makes IPv6 usable again, if the proact_tracking multicast mode has been activated. Signed-off-by: Linus Lüssing <linus.luessing@saxnet.de> --- multicast.c | 25 +++++++++++++++++++++++++ multicast.h | 1 + soft-interface.c | 2 +- 3 files changed, 27 insertions(+), 1 deletions(-)