Basic Multicast Optimizations
This is the fifth revision of the basic multicast optimization patches.
It fixes two bugs:
For one thing packets to the IPv6 all-hosts-link-scope multicast address (ff02::1) were optimized, too. However this is the one link local address which we should not never optimize (see RFC4541, section 3 [1]).
For another since commit 87d4a282 ("batman-adv: remove useless find_router look up" the unicast-forwarding of multicast packets was broken and I had
only runtime tested the patches based on an older commit
(before 87d4a282 such unicast-forwarding was "only" broken if
a gateway-server was present).
Also note that rebasing on top of the TT-VLAN patches wasn't trivial:
Simply resolving the rebase complaints resulted in subtle runtime bugs. A few lines in the counting methods added to translation-table.c needed changes.
Cheers, Linus
[1]: Considerations for Internet Group Management Protocol (IGMP)
and Multicast Listener Discovery (MLD) Snooping Switches -> "Packets with the all hosts link-scope address should be forwarded
on all ports."
https://tools.ietf.org/html/rfc4541
PATCH v5 1/3 batman-adv: Multicast Listener Announcements via Translation Table
With this patch a node which has no bridge interface on top of its soft interface announces its local multicast listeners via the translation table.
--- Makefile | 2 + Makefile.kbuild | 1 + compat.h | 20 ++++- gen-compat-autoconf.sh | 1 + main.c | 6 ++ main.h | 1 + multicast.c | 210 ++++++++++++++++++++++++++++++++++++++++++++++++ multicast.h | 43 ++++++++++ soft-interface.c | 3 + sysfs.c | 6 ++ translation-table.c | 22 ++++- types.h | 12 +++ 12 files changed, 322 insertions(+), 5 deletions(-) create mode 100644 multicast.c create mode 100644 multicast.h
Makefile +2 -0
@@ -27,6 +27,8 @@ export CONFIG_BATMAN_ADV_BLA=y export CONFIG_BATMAN_ADV_DAT=y # B.A.T.M.A.N network coding (catwoman): export CONFIG_BATMAN_ADV_NC=n +# B.A.T.M.A.N. multicast optimizations: +export CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS=y PWD:=$(shell pwd) KERNELPATH ?= /lib/modules/$(shell uname -r)/build
Makefile.kbuild +1 -0
@@ -38,3 +38,4 @@ batman-adv-y += send.o batman-adv-y += soft-interface.o batman-adv-y += sysfs.o batman-adv-y += translation-table.o +batman-adv-$(CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS) += multicast.o
compat.h +19 -1
@@ -103,13 +103,31 @@ static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,#define pr_warn pr_warning +#undef netdev_for_each_mc_addr +#define netdev_for_each_mc_addr (mclist, dev) \ + for (mclist = (struct bat_dev_addr_list *)dev->mc_list; mclist; \ + mclist = (struct bat_dev_addr_list *)mclist->next) + +/* Note, that this breaks the usage of the normal 'struct netdev_hw_addr' +* for kernels < 2.6.35 in batman-adv! +*/ +#define netdev_hw_addr batadv_dev_addr_list +struct batadv_dev_addr_list { +struct dev_addr_list *next; +u8 addr[MAX_ADDR_LEN ]; +u8 da_addrlen; +u8 da_synced; +int da_users; +int da_gusers; +}; +#endif /* < KERNEL_VERSION(2, 6, 35) */ #if LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,36 )#define __rcu -#define IFF_BRIDGE_PORT 0 || (hard_iface->net_dev->br_port ? 1 : 0) +#define IFF_BRIDGE_PORT 0 || (soft_iface->br_port ? 1 : 0) struct kernel_param_ops {/* Returns 0, or -errno. arg is in kp->arg. */
gen-compat-autoconf.sh +1 -0
@@ -39,6 +39,7 @@ gen_config() { gen_config 'CONFIG_BATMAN_ADV_DEBUG' ${CONFIG_BATMAN_ADV_DEBUG:="n"} >> "${TMP}" gen_config 'CONFIG_BATMAN_ADV_BLA' ${CONFIG_BATMAN_ADV_BLA:="y"} >> "${TMP}" gen_config 'CONFIG_BATMAN_ADV_DAT' ${CONFIG_BATMAN_ADV_DAT:="y"} >> "${TMP}" +gen_config 'CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS' ${CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS:="y"} >> "${TMP}" gen_config 'CONFIG_BATMAN_ADV_NC' ${CONFIG_BATMAN_ADV_NC:="n"} >> "${TMP}" # only regenerate compat-autoconf.h when config was changed
main.c +6 -0
@@ -33,6 +33,7 @@#include "gateway_client.h" #include "bridge_loop_avoidance.h" #include "distributed-arp-table.h" +#include "multicast.h" #include "gateway_common.h" #include "hash.h" #include "bat_algo.h"
@@ -116,6 +117,9 @@ int batadv_mesh_init(struct net_device *soft_iface)INIT_LIST_HEAD (&bat_priv->tt.changes_list);INIT_LIST_HEAD (&bat_priv->tt.req_list);INIT_LIST_HEAD (&bat_priv->tt.roam_list); +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS +INIT_LIST_HEAD (&bat_priv->mcast.mla_list); +#endif INIT_HLIST_HEAD (&bat_priv->tvlv.container_list);INIT_HLIST_HEAD (&bat_priv->tvlv.handler_list);
@@ -169,6 +173,8 @@ void batadv_mesh_free(struct net_device *soft_iface)batadv_dat_free (bat_priv);batadv_bla_free (bat_priv); +batadv_mcast_free (bat_priv); +/* Free the TT and the originator tables only after having terminated * all the other depending components which may use these structures for * their purposes.
main.h +1 -0
@@ -164,6 +164,7 @@ enum batadv_uev_type {#include <linux/percpu.h> #include <linux/slab.h> #include <net/sock.h> /* struct sock */ +#include <net/addrconf.h> /* ipv6 address stuff */ #include <net/rtnetlink.h> #include <linux/jiffies.h> #include <linux/seq_file.h>
multicast.c new file +210 -0
210 lines not on this page · raw message
multicast.h new file +43 -0
43 lines not on this page · raw message
soft-interface.c +3 -0
@@ -471,6 +471,9 @@ static int batadv_softif_init_late(struct net_device *dev)#ifdef CONFIG_BATMAN_ADV_DAT atomic_set (&bat_priv->distributed_arp_table,1 );#endif +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS +atomic_set (&bat_priv->mcast_group_awareness,1 ); +#endif atomic_set (&bat_priv->ap_isolation,0 );atomic_set (&bat_priv->gw_mode,BATADV_GW_MODE_OFF );atomic_set (&bat_priv->gw_sel_class,20 );
sysfs.c +6 -0
@@ -369,6 +369,9 @@ BATADV_ATTR_SIF_UINT(gw_sel_class, S_IRUGO | S_IWUSR, 1, BATADV_TQ_MAX_VALUE, batadv_post_gw_deselect); static BATADV_ATTR(gw_bandwidth, S_IRUGO | S_IWUSR, batadv_show_gw_bwidth, batadv_store_gw_bwidth); +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS +BATADV_ATTR_SIF_BOOL (mcast_group_awareness,S_IRUGO |S_IWUSR ,NULL ); +#endif #ifdef CONFIG_BATMAN_ADV_DEBUG BATADV_ATTR_SIF_UINT (log_level,S_IRUGO |S_IWUSR ,0 ,BATADV_DBG_ALL ,NULL );#endif
@@ -386,6 +389,9 @@ static struct batadv_attribute *batadv_mesh_attrs[] = {#ifdef CONFIG_BATMAN_ADV_DAT &batadv_attr_distributed_arp_table,#endif +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS + &batadv_attr_mcast_group_awareness, +#endif &batadv_attr_fragmentation, &batadv_attr_ap_isolation, &batadv_attr_routing_algo,
translation-table.c +18 -4
@@ -26,6 +26,7 @@#include "originator.h" #include "routing.h" #include "bridge_loop_avoidance.h" +#include "multicast.h" #include <linux/crc32c.h>
@@ -351,14 +352,16 @@ void batadv_tt_local_add(struct net_device *soft_iface, const uint8_t *addr, {struct batadv_priv *bat_priv =netdev_priv (soft_iface);struct batadv_tt_local_entry *tt_local; -struct batadv_tt_global_entry *tt_global; +struct batadv_tt_global_entry *tt_global =NULL ;struct hlist_head *head;struct batadv_tt_orig_list_entry *orig_entry;int hash_added;bool roamed_back = false; tt_local =batadv_tt_local_hash_find (bat_priv, addr, vid); - tt_global =batadv_tt_global_hash_find (bat_priv, addr, vid); + +if (!is_multicast_ether_addr (addr)) + tt_global =batadv_tt_global_hash_find (bat_priv, addr, vid);if (tt_local) { tt_local->last_seen = jiffies;
@@ -412,8 +415,11 @@ void batadv_tt_local_add(struct net_device *soft_iface, const uint8_t *addr, tt_local->last_seen = jiffies; tt_local->common.added_at = tt_local->last_seen; -/* the batman interface mac address should never be purged */ -if (batadv_compare_eth (addr, soft_iface->dev_addr)) +/* the batman interface mac and multicast addresses should never be +* purged +*/ +if (batadv_compare_eth (addr, soft_iface->dev_addr) || +is_multicast_ether_addr (addr)) tt_local->common.flags |=BATADV_TT_CLIENT_NOPURGE ; hash_added = batadv_hash_add(bat_priv->tt.local_hash, batadv_compare_tt,
@@ -1007,6 +1013,11 @@ add_orig_entry: ret = true; out_remove: +/* Do not remove multicast addresses from the local hash on +* global additions +*/ +if (is_multicast_ether_addr (tt_addr)) + goto out;/* remove address from local hash if present */ local_flags = batadv_tt_local_remove(bat_priv, tt_addr, vid,
@@ -2521,6 +2532,9 @@ void batadv_tt_local_commit_changes(struct batadv_priv *bat_priv) {uint16_t changed_num =0 ; +/* Update multicast addresses in local translation table */ +batadv_mcast_mla_tt_update (bat_priv); +if (atomic_read (&bat_priv->tt.local_changes) <1 ) {if (!batadv_atomic_dec_not_zero (&bat_priv->tt.ogm_append_cnt))batadv_tt_tvlv_container_update (bat_priv);
types.h +12 -0
@@ -501,6 +501,12 @@ struct batadv_priv_dat { };#endif +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS +struct batadv_priv_mcast { +struct list_head mla_list; +}; +#endif +/** * struct batadv_priv_nc - per mesh interface network coding private data * @work: work queue callback item for cleanup
@@ -591,6 +597,9 @@ struct batadv_priv {#ifdef CONFIG_BATMAN_ADV_DAT atomic_t distributed_arp_table;#endif +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS +atomic_t mcast_group_awareness; +#endif atomic_t gw_mode;atomic_t gw_sel_class;atomic_t orig_interval;
@@ -625,6 +634,9 @@ struct batadv_priv {#ifdef CONFIG_BATMAN_ADV_DAT struct batadv_priv_dat dat;#endif +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS +struct batadv_priv_mcast mcast; +#endif #ifdef CONFIG_BATMAN_ADV_NC atomic_t network_coding;struct batadv_priv_nc nc;
Signature · 1 line
1.7.10.4
PATCH v5 2/3 batman-adv: Announce new capability via multicast TVLV
In reply to Linus LüssingIf the soft interface of a node is not part of a bridge then a node announces a new multicast TVLV: The according flag (BATADV_MCAST_LISTENER_ANNOUNCEMENT) signalizes that this node is announcing all of its multicast listeners via the translation table infrastructure. More precisely, all multicast listeners of scope greater than link-local for IPv4 and of scope greater
or equal to link-local for IPv6.
--- main.c | 4 ++++ multicast.c | 68 ++++++++++++++++++++++++++++++++++++++++++++++++++++-- multicast.h | 14 +++++++++++ originator.c | 6 +++++ packet.h | 7 ++++++ soft-interface.c | 1 + types.h | 4 ++++ 7 files changed, 102 insertions(+), 2 deletions(-)
main.c +4 -0
@@ -146,6 +146,10 @@ int batadv_mesh_init(struct net_device *soft_iface)if (ret <0 ) goto err; + ret =batadv_mcast_init (bat_priv); +if (ret <0 ) + goto err; + ret =batadv_gw_init (bat_priv);if (ret <0 ) goto err;
multicast.c +66 -2
96 lines not on this page · raw message
multicast.h +14 -0
@@ -24,8 +24,12 @@void batadv_mcast_mla_tt_update (struct batadv_priv *bat_priv); +int batadv_mcast_init (struct batadv_priv *bat_priv); +void batadv_mcast_free (struct batadv_priv *bat_priv); +void batadv_mcast_purge_orig (struct batadv_orig_node *orig_node); +#else static inline void batadv_mcast_mla_tt_update (struct batadv_priv *bat_priv)
@@ -33,11 +37,21 @@ static inline void batadv_mcast_mla_tt_update(struct batadv_priv *bat_priv)return ; } +static inline int batadv_mcast_init (struct batadv_priv *bat_priv) +{ +return 0 ; +} +static inline void batadv_mcast_free (struct batadv_priv *bat_priv) {return ; } +static inline void batadv_mcast_purge_orig (struct batadv_orig_node *orig_node) +{ +return ; +} +#endif /* CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS */ #endif /* _NET_BATMAN_ADV_MULTICAST_H_ */
originator.c +6 -0
@@ -29,6 +29,7 @@#include "bridge_loop_avoidance.h" #include "network-coding.h" #include "fragmentation.h" +#include "multicast.h" /* hash class keys */ static struct lock_class_key batadv_orig_hash_lock_class_key;
@@ -143,6 +144,8 @@ static void batadv_orig_node_free_rcu(struct rcu_head *rcu)spin_unlock_bh (&orig_node->neigh_list_lock); +batadv_mcast_purge_orig (orig_node); +/* Free nc_nodes */ batadv_nc_purge_orig (orig_node->bat_priv, orig_node,NULL );
@@ -258,6 +261,9 @@ struct batadv_orig_node *batadv_get_orig_node(struct batadv_priv *bat_priv, reset_time = jiffies -1 -msecs_to_jiffies (BATADV_RESET_PROTECTION_MS ); orig_node->bcast_seqno_reset = reset_time; orig_node->batman_seqno_reset = reset_time; +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS + orig_node->mcast_flags =BATADV_MCAST_LISTENER_ANNOUNCEMENT ; +#endif atomic_set (&orig_node->bond_candidates,0 );
packet.h +7 -0
@@ -91,6 +91,11 @@ enum batadv_icmp_packettype { BATADV_PARAMETER_PROBLEM = 12, }; +/* multicast capabilities */ +enum batadv_mcast_flags { +BATADV_MCAST_LISTENER_ANNOUNCEMENT =BIT (0 ), +}; +/* tt data subtypes */ #define BATADV_TT_DATA_TYPE_MASK 0x0F
@@ -145,6 +150,7 @@ enum batadv_bla_claimframe {* @BATADV_TVLV_NC: network coding tvlv * @BATADV_TVLV_TT: translation table tvlv * @BATADV_TVLV_ROAM: roaming advertisement tvlv +* @BATADV_TVLV_MCAST: multicast capability tvlv */ enum batadv_tvlv_type {BATADV_TVLV_GW =0x01 ,
@@ -152,6 +158,7 @@ enum batadv_tvlv_type {BATADV_TVLV_NC =0x03 ,BATADV_TVLV_TT =0x04 ,BATADV_TVLV_ROAM =0x05 , + BATADV_TVLV_MCAST = 0x06, };/* the destination hardware field in the ARP frame is used to
soft-interface.c +1 -0
@@ -473,6 +473,7 @@ static int batadv_softif_init_late(struct net_device *dev)#endif #ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS atomic_set (&bat_priv->mcast_group_awareness,1 ); +atomic_set (&bat_priv->mcast.num_no_mla,0 );#endif atomic_set (&bat_priv->ap_isolation,0 );atomic_set (&bat_priv->gw_mode,BATADV_GW_MODE_OFF );
types.h +4 -0
@@ -163,6 +163,9 @@ struct batadv_orig_node {unsigned long last_seen;unsigned long bcast_seqno_reset;unsigned long batman_seqno_reset; +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS +uint8_t mcast_flags; +#endif uint8_t capabilities;atomic_t last_ttvn;uint32_t tt_crc;
@@ -504,6 +507,7 @@ struct batadv_priv_dat {#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS struct batadv_priv_mcast {struct list_head mla_list; +atomic_t num_no_mla; };#endif
Signature · 1 line
1.7.10.4
PATCH v5 3/3 batman-adv: Modified forwarding behaviour for multicast packets
In reply to Linus LüssingWith this patch a multicast packet is not always simply flooded anymore, the bevahiour for the following cases is changed to reduce unnecessary overhead:
If all nodes within the horizon of a certain node have signalized multicast listener announcement capability (BATADV_MCAST_LISTENER_ANNOUNCEMENT) then an IPv6 multicast packet with a destination of IPv6 link-local scope coming from the upstream of this node...
* ...is dropped if there is no according multicast listener in the translation table,
* ...is forwarded via unicast if there is a single node with interested multicast listeners
* ...and otherwise still gets flooded.
--- distributed-arp-table.c | 2 +- multicast.c | 56 +++++++++++++++++++++++++++++++++++++ multicast.h | 23 +++++++++++++++ send.c | 5 ++-- send.h | 9 +++--- soft-interface.c | 18 ++++++++++-- translation-table.c | 71 +++++++++++++++++++++++++++++++++++++++++++++++ translation-table.h | 2 ++ 8 files changed, 177 insertions(+), 9 deletions(-)
distributed-arp-table.c +1 -1
@@ -1030,7 +1030,7 @@ bool batadv_dat_snoop_incoming_arp_request(struct batadv_priv *bat_priv,BATADV_P_DAT_CACHE_REPLY , vid); else - err =batadv_send_skb_unicast (bat_priv, skb_new, vid); + err =batadv_send_skb_unicast (bat_priv, skb_new, vid, false);if (!err) {batadv_inc_counter (bat_priv,BATADV_CNT_DAT_CACHED_REPLY_TX );
multicast.c +56 -0
68 lines not on this page · raw message
multicast.h +23 -0
@@ -20,10 +20,27 @@#ifndef _NET_BATMAN_ADV_MULTICAST_H_#define _NET_BATMAN_ADV_MULTICAST_H_ +/** +* batadv_forw_mode - the way a packet should be forwarded as +* @BATADV_FORW_ALL: forward the packet to all nodes +* (currently via classic flooding) +* @BATADV_FORW_SINGLE: forward the packet to a single node +* (currently via the BATMAN_IV unicast routing protocol) +* @BATADV_FORW_NONE: don't forward, drop it +*/ +enum batadv_forw_mode { +BATADV_FORW_ALL , +BATADV_FORW_SINGLE , +BATADV_FORW_NONE , +}; +#ifdef CONFIG_BATMAN_ADV_MCAST_OPTIMIZATIONS void batadv_mcast_mla_tt_update (struct batadv_priv *bat_priv); +enum batadv_forw_mode +batadv_mcast_forw_mode (struct sk_buff *skb,struct batadv_priv *bat_priv); +int batadv_mcast_init (struct batadv_priv *bat_priv);void batadv_mcast_free (struct batadv_priv *bat_priv);
@@ -37,6 +54,12 @@ static inline void batadv_mcast_mla_tt_update(struct batadv_priv *bat_priv)return ; } +static inline enum batadv_forw_mode +batadv_mcast_forw_mode (struct sk_buff *skb,struct batadv_priv *bat_priv) +{ +return BATADV_FORW_ALL ; +} +static inline int batadv_mcast_init (struct batadv_priv *bat_priv) {return 0 ;
send.c +3 -2
@@ -242,13 +242,14 @@ out:* @packet_subtype: the unicast 4addr packet subtype (only relevant for unicast * 4addr packets) * @vid: the vid to be used to search the translation table +* @forw_to_gw: whether to forward to a gateway (or via TT otherwise) * * Returns 1 in case of error or 0 otherwise. */ int batadv_send_skb_generic_unicast (struct batadv_priv *bat_priv,struct sk_buff *skb,int packet_type,int packet_subtype, -unsigned short vid) +unsigned short vid,bool forw_to_gw) {struct ethhdr *ethhdr = (struct ethhdr *)skb->data;struct batadv_unicast_packet *unicast_packet;
@@ -256,7 +257,7 @@ int batadv_send_skb_generic_unicast(struct batadv_priv *bat_priv,int ret =NET_RX_DROP ;/* get routing information */ -if (is_multicast_ether_addr (ethhdr->h_dest)) +if (forw_to_gw) orig_node =batadv_gw_get_selected_orig (bat_priv);else /* check for tt host - increases orig_node refcount.
send.h +5 -4
@@ -41,22 +41,23 @@ bool batadv_send_skb_prepare_unicast_4addr(struct batadv_priv *bat_priv,int batadv_send_skb_generic_unicast (struct batadv_priv *bat_priv,struct sk_buff *skb,int packet_type,int packet_subtype, -unsigned short vid); +unsigned short vid,bool forw_to_gw);/** * batadv_send_unicast_skb - send the skb encapsulated in a unicast packet * @bat_priv: the bat priv with all the soft interface information * @skb: the payload to send * @vid: the vid to be used to search the translation table +* @forw_to_gw: whether to forward to a gateway (or via TT otherwise) * * Returns 1 in case of error or 0 otherwise. */ static inline int batadv_send_skb_unicast (struct batadv_priv *bat_priv,struct sk_buff *skb, -unsigned short vid) +unsigned short vid,bool forw_to_gw) {return batadv_send_skb_generic_unicast (bat_priv, skb,BATADV_UNICAST , -0 , vid); +0 , vid, forw_to_gw); }/**
@@ -76,7 +77,7 @@ static inline int batadv_send_skb_unicast_4addr(struct batadv_priv *bat_priv, {return batadv_send_skb_generic_unicast (bat_priv, skb,BATADV_UNICAST_4ADDR , - packet_subtype, vid); + packet_subtype, vid, false); }#endif /* _NET_BATMAN_ADV_SEND_H_ */
soft-interface.c +16 -2
@@ -34,6 +34,7 @@#include <linux/ethtool.h> #include <linux/etherdevice.h> #include <linux/if_vlan.h> +#include "multicast.h" #include "bridge_loop_avoidance.h" #include "network-coding.h"
@@ -169,6 +170,8 @@ static int batadv_interface_tx(struct sk_buff *skb,bool do_bcast = false;unsigned short vid;uint32_t seqno; +enum batadv_forw_mode mode; +bool forw_to_gw = false;if (atomic_read (&bat_priv->mesh_state) !=BATADV_MESH_ACTIVE ) goto dropped;
@@ -226,13 +229,24 @@ static int batadv_interface_tx(struct sk_buff *skb,* via unicast to their gateway */ ret =batadv_gw_is_dhcp_target (skb, &header_len); -if (ret) +if (ret) { do_bcast = false; + forw_to_gw = true; + }break ;case BATADV_GW_MODE_OFF :default :break ; } + +if (do_bcast && !is_broadcast_ether_addr (ethhdr->h_dest)) { + mode =batadv_mcast_forw_mode (skb, bat_priv); +if (mode ==BATADV_FORW_NONE ) + goto dropped; + +if (mode ==BATADV_FORW_SINGLE ) + do_bcast = false; + } }/* ethernet packet should be broadcasted */
@@ -289,7 +303,7 @@ static int batadv_interface_tx(struct sk_buff *skb,batadv_dat_snoop_outgoing_arp_reply (bat_priv, skb); - ret =batadv_send_skb_unicast (bat_priv, skb, vid); + ret =batadv_send_skb_unicast (bat_priv, skb, vid, forw_to_gw);if (ret !=0 ) goto dropped_freed; }
translation-table.c +71 -0
77 lines not on this page · raw message
translation-table.h +2 -0
@@ -31,6 +31,8 @@ int batadv_tt_global_seq_print_text(struct seq_file *seq, void *offset);void batadv_tt_global_del_orig (struct batadv_priv *bat_priv,struct batadv_orig_node *orig_node,const char *message); +int batadv_tt_global_hash_count (struct batadv_priv *bat_priv, +const uint8_t *addr,unsigned short vid); struct batadv_orig_node *batadv_transtable_search(struct batadv_priv *bat_priv, const uint8_t *src, const uint8_t *addr,
Signature · 1 line
1.7.10.4
Re: [B.A.T.M.A.N.] [PATCHv5 3/3] batman-adv: Modified forwarding behaviour for multicast packets
Quoted 62 lines · Linus Lüssing
/**
+ *batadv_tt_orig_entries_count- count the number of originators
+ * @head: a list of originators
+ *
+ * Return the number of originator entries in the given list.
+ *
+ * The caller needs to hold thercu_read_lock().
+ */translation-table.c +54
-0+static int batadv_tt_orig_entries_count (struct hlist_head *head) +{ +struct batadv_tt_orig_list_entry *orig_entry; +int count =0 ; + +hlist_for_each_entry_rcu (orig_entry, head, list) { +if (!atomic_read (&orig_entry->refcount)) +continue ; + + count++; + } + +return count; +} + +/** +* batadv_tt_global_hash_count - count the number of orig entries +* @hash: hash table containing the tt entries +* @addr: the mac address of the client to count entries for +* @vid: VLAN identifier +* +* Return the number of originators advertising the given address/data +* (excluding ourself). +*/ +int batadv_tt_global_hash_count (struct batadv_priv *bat_priv, +const uint8_t *addr,unsigned short vid) +{ +struct hlist_head *head, *orig_list; +struct batadv_tt_common_entry to_search, *tt_common_entry; +struct batadv_tt_global_entry *tt_global_entry; +uint32_t index; +int count =0 ; + +if (!bat_priv->tt.global_hash) + goto out; + +memcpy (to_search.addr, addr,ETH_ALEN ); + to_search.vid = vid; + + index =batadv_choose_tt (&to_search, bat_priv->tt.global_hash->size); + head = &bat_priv->tt.global_hash->table[index]; + +rcu_read_lock (); +hlist_for_each_entry_rcu (tt_common_entry, head, hash_entry) { +if (!batadv_compare_eth (tt_common_entry, addr)) +continue ; + +if (!atomic_read (&tt_common_entry->refcount)) +continue ; + + tt_global_entry =container_of (tt_common_entry, + struct batadv_tt_global_entry, + common); + orig_list = &tt_global_entry->orig_list;
translation-table.c batadv_tt_hash_find(struct batadv_hashtable *hash, const uint8_t *addr, · line 198
Quoted 10 lines
translation-table.c +10
-0+ count =batadv_tt_orig_entries_count (orig_list); +break ; + } +rcu_read_unlock (); + +out: +return count; +} + +/**
I've been thinking about this part ... it is not good to iterate over the whole hash for every multicast packet - this is the fastpath after all, so it's critical.
Can't you just use batadv_tt_global_hash_find(), count the list entries within struct batadv_tt_global_entry (or find out if it is empty, one entry or more than one entry) and be done? This should be much faster.
Apart from that, the patch series appears to be pretty mature now, I've done some more testing in my VMs and it looks good. Should be (hopefully) the last thing from my side. :)
Cheers,
Simon
Re: [B.A.T.M.A.N.] [PATCHv5 3/3] batman-adv: Modified forwarding behaviour for multicast packets
Quoted 69 lines · Simon Wunderlich
Quoted 67 lines · Linus Lüssing
/**
+ *batadv_tt_orig_entries_count- count the number of originators
+ * @head: a list of originators
+ *
+ * Return the number of originator entries in the given list.
+ *
+ * The caller needs to hold thercu_read_lock().
+ */translation-table.c +59
-0+static int batadv_tt_orig_entries_count (struct hlist_head *head) +{ +struct batadv_tt_orig_list_entry *orig_entry; +int count =0 ; + +hlist_for_each_entry_rcu (orig_entry, head, list) { +if (!atomic_read (&orig_entry->refcount)) +continue ; + + count++; + } + +return count; +} + +/** +* batadv_tt_global_hash_count - count the number of orig entries +* @hash: hash table containing the tt entries +* @addr: the mac address of the client to count entries for +* @vid: VLAN identifier +* +* Return the number of originators advertising the given address/data +* (excluding ourself). +*/ +int batadv_tt_global_hash_count (struct batadv_priv *bat_priv, +const uint8_t *addr,unsigned short vid) +{ +struct hlist_head *head, *orig_list; +struct batadv_tt_common_entry to_search, *tt_common_entry; +struct batadv_tt_global_entry *tt_global_entry; +uint32_t index; +int count =0 ; + +if (!bat_priv->tt.global_hash) + goto out; + +memcpy (to_search.addr, addr,ETH_ALEN ); + to_search.vid = vid; + + index =batadv_choose_tt (&to_search, bat_priv->tt.global_hash->size); + head = &bat_priv->tt.global_hash->table[index]; + +rcu_read_lock (); +hlist_for_each_entry_rcu (tt_common_entry, head, hash_entry) { +if (!batadv_compare_eth (tt_common_entry, addr)) +continue ; + +if (!atomic_read (&tt_common_entry->refcount)) +continue ; + + tt_global_entry =container_of (tt_common_entry, + struct batadv_tt_global_entry, + common); + orig_list = &tt_global_entry->orig_list; + count =batadv_tt_orig_entries_count (orig_list); +break ; + } +rcu_read_unlock (); +
Quoted 8 lines
translation-table.c batadv_tt_hash_find(struct batadv_hashtable *hash, const uint8_t *addr, · line 198
Quoted 5 lines
translation-table.c +5
-0+out: +return count; +} + +/** I've been thinking about this part ... it is not good to iterate over the whole hash for every multicast packet - this is the fastpath after all, so it's critical.
This is what we do for any packet, it's a simple, fast hash-lookup (and for the rare case of two entries having the same hash result, then a small list lookup, too). Pretty much what we are doing with any batadv_transtable_search(). It's mostly a copy-paste of the batadv_tt_hash_find().
Except the batadv_tt_orig_entries_count(), that's a list lookup which we do not have with a common transtable_search(). I had thought about adding an atomic counter for the orig_list in and to the struct tt_global_entry (but didn't do it because I wanted to leave the original translation-table.c functions as untouched as possible and thought the performance gain wouldn't be that huge). Hm, but maybe you're right about the performance if we are going to have a large mesh network with many originators claiming a certain multicast address. I'm going to add that atomic counter.
Quoted 4 lines
Can't you just use
batadv_tt_global_hash_find(), count the list entries within structbatadv_tt_global_entry(or find out if it is empty, one entry or more than one entry) and be done? This should be much faster.
But looking at that function again now, you're right, because the lookup in batadv_tt_global_hash_count() is nearly identical now to batadv_tt_global_hash_find(), I can use that one instead. (Initially I needed to do such copying of the tt_global_hash_find() function because of my wrong assumption that there'd be a tt_global_entry for every originator announcing that address - but, yeah, now that's even easier :) ).
Will change that :).
Quoted 4 lines
Apart from that, the patch series appears to be pretty mature now, I've done some more testing in my VMs and it looks good. Should be (hopefully) the last thing from my side. :)
No worries, every comment welcome. And with just 3 instead of ~15 multicast patches it's a lot easier to fix and rebase things now :P.
Quoted 3 lines
Cheers,
Simon
Cheers, Linus