B.A.T.M.A.N May 2011

b.a.t.m.a.n@lists.open-mesh.org

14 participants
59 discussions

Re: [B.A.T.M.A.N.] mesh
by Sven Eckelmann 06 May '11

06 May '11

On Friday 06 May 2011 19:02:48 Jose Armando wrote: > Question so openwrt do not work with Batman? U saying I need to creater > package for my router? Diego to much Question but I'm need I was think just > install openwrt on the router u can install any package ? You still didn't answer what you want.... and the gibberish at the end doesn't help. You can try to join #batman on irc.freenode.net and maybe we can find out what you want and how this can be done. Kind regards, Sven

1 0

Re: [B.A.T.M.A.N.] mesh
by Sven Eckelmann 06 May '11

06 May '11

On Friday 06 May 2011 18:11:21 Jose Armando wrote: > Hello Sven > > Thank you so much for u info i will play with it > Note > Question if for one reason i cant make work can u help to install for > me i can open telnet on my router and u can do for me i will pay you > for u time and i will be greatfull to you No, this will not work because you must create a package on another system and ensure that you flash your router with the corresponding openwrt version. You can join #batman on irc.freenode.net to ask your questions during the installation phase. Kind regards, Sven

1 0

[B.A.T.M.A.N.] mesh
by Jose Armando 06 May '11

06 May '11

Hello Can any one tell me how to install batman on openwrt my router has openwrt install with luci but i cant install batman im need on this mesh or any one can build for me and i will pay for u time thank you Armando Dieguez Phone 714.292.0156

2 1

[B.A.T.M.A.N.] [PATCH] batman-adv: Fix rcu_lock imbalance in find_router
by Marek Lindner 05 May '11

05 May '11

Signed-off-by: Marek Lindner <lindner_marek(a)yahoo.de> --- routing.c | 14 +++++++++----- 1 files changed, 9 insertions(+), 5 deletions(-) diff --git a/routing.c b/routing.c index 49f5715..d8cde2b 100644 --- a/routing.c +++ b/routing.c @@ -1213,7 +1213,7 @@ struct neigh_node *find_router(struct bat_priv *bat_priv, router = orig_node_get_router(orig_node); if (!router) - return NULL; + goto err; /* without bonding, the first node should * always choose the default router. */ @@ -1222,10 +1222,8 @@ struct neigh_node *find_router(struct bat_priv *bat_priv, rcu_read_lock(); /* select default router to output */ router_orig = router->orig_node; - if (!router_orig) { - rcu_read_unlock(); - return NULL; - } + if (!router_orig) + goto err_unlock; if ((!recv_if) && (!bonding_enabled)) goto return_router; @@ -1268,6 +1266,12 @@ struct neigh_node *find_router(struct bat_priv *bat_priv, return_router: rcu_read_unlock(); return router; +err_unlock: + rcu_read_unlock(); +err: + if (router) + neigh_node_free_ref(router); + return NULL; } static int check_unicast_packet(struct sk_buff *skb, int hdr_size) -- 1.7.2.3

2 3

Re: [B.A.T.M.A.N.] Routing decisions
by Sven Eckelmann 05 May '11

05 May '11

Ed Okerson wrote: > On Thu, May 5, 2011 at 12:35 PM, Sven Eckelmann <sven(a)narfation.org> wrote: > > Ed Okerson wrote: > >> On Thu, May 5, 2011 at 10:44 AM, Marek Lindner <lindner_marek(a)yahoo.de> > > > > wrote: > >> > Hi, > >> > > >> >> We are evaluating using Batman in an environment where there could be > >> >> 200-300 devices in a single building. We started out setting up 10 > >> >> devices in our office to figure out how everything works and do some > >> >> throughput testing. We have noticed that the routing decisions > >> >> always send the packet to the node towards the destination with the > >> >> highest signal strength. This causes the packet to always traverse > >> >> the network with the maximum possible number of hops, which causes > >> >> performance to degrade quickly. Is it possible to use a different > >> >> routing algorithm? It would seem that sending to the node closest > >> >> to the destination that the source node can still communicate with > >> >> directly would minimize the number of hops. > >> > > >> > if you wish to minimize the number of hops you have to increase the > >> > hop penalty. Check the "hop penalty" section here: > >> > http://www.open-mesh.org/wiki/batman-adv/Tweaking > >> > >> That seems to indicate that it is a per node setting, i.e. "using this > >> node will incur a penalty of x". That is also not the desired > >> behavior. For our installation all nodes are in a fixed location, so > >> using a particular node as a next hop in the route may incur a penalty > >> for one source node, but not another. This should be dynamically > >> determined for each route from each source to each destination to > >> minimize hops. > > > > So you have to increase the hop penalty everywhere to force the routing > > algorithm to reduce the number of hops and prefer worse routes with less > > hops. > > If I increase the hop penalty on all nodes, they will all still be > equal. If all nodes have a hop penalty of 128, will it not make the > same routing decisions as it does if they are all 10? No, the wiki explains it quite well, but I'll try to explain it a little bit different. Lets assume that you have perfect links (quality = 255) between all nodes in a route. Each one would say: "hey, I can reach the following neighbors with the quality of 255".... now introduce a penalty in each hop (10 for example) which is just substracted from the incoming value. So after a distance of 5 hops only a quality of 205 would be announced. Therefore a node with only a distance of 2 (quality = 235) would be prefered as next hop in this case and not the one with the distance of 5. Target - X1 - X2 - X3 - X4 - X5 - (Announces route with quality 205 to Target) \ \ Y1 - Y2 - (Announces route with quality 235 to Target) This is a quite abstract and not 100% correct explanation of the hop penalty, but should help to understand the implications a little bit better. You have to do the calculations yourself to find a good value for hop penalty. You should have a look at send.c to find more information about it. You should see that it is not really substracted in a linear fashion and that the actual quality is calculation is also more complex,... but the idea is similar to the example mentioned before. Kind regards, Sven

1 0

[B.A.T.M.A.N.] Routing decisions
by Ed Okerson 05 May '11

05 May '11

Hi, We are evaluating using Batman in an environment where there could be 200-300 devices in a single building. We started out setting up 10 devices in our office to figure out how everything works and do some throughput testing. We have noticed that the routing decisions always send the packet to the node towards the destination with the highest signal strength. This causes the packet to always traverse the network with the maximum possible number of hops, which causes performance to degrade quickly. Is it possible to use a different routing algorithm? It would seem that sending to the node closest to the destination that the source node can still communicate with directly would minimize the number of hops. Ed Okerson

3 3

[B.A.T.M.A.N.] [PATCH] batman-adv: remove duplicate code from function is_bidirectional_neigh()
by Daniele Furlan 05 May '11

05 May '11

In function is_bidirectional_neigh the code that find out the one hop neighbor is duplicated. This patch remove the duplicate code and adjust the comments. P.S. It's my first patch, let me know if there is any problem. Signed-off-by: Daniele Furlan <daniele.furlan(a)gmail.com> --- routing.c | 75 ++++++++++++++++++++---------------------------------------- 1 files changed, 25 insertions(+), 50 deletions(-) diff --git a/routing.c b/routing.c index 49f5715..c5dec8c 100644 --- a/routing.c +++ b/routing.c @@ -169,65 +169,40 @@ static int is_bidirectional_neigh(struct orig_node *orig_node, uint8_t orig_eq_count, neigh_rq_count, tq_own; int tq_asym_penalty, ret = 0; - if (orig_node == orig_neigh_node) { - rcu_read_lock(); - hlist_for_each_entry_rcu(tmp_neigh_node, node, - &orig_node->neigh_list, list) { + /* find corresponding one hop neighbor */ + rcu_read_lock(); + hlist_for_each_entry_rcu(tmp_neigh_node, node, + &orig_node->neigh_list, list) { - if (!compare_eth(tmp_neigh_node->addr, - orig_neigh_node->orig)) - continue; + if (!compare_eth(tmp_neigh_node->addr, + orig_neigh_node->orig)) + continue; - if (tmp_neigh_node->if_incoming != if_incoming) - continue; + if (tmp_neigh_node->if_incoming != if_incoming) + continue; - if (!atomic_inc_not_zero(&tmp_neigh_node->refcount)) - continue; + if (!atomic_inc_not_zero(&tmp_neigh_node->refcount)) + continue; - neigh_node = tmp_neigh_node; - } - rcu_read_unlock(); + neigh_node = tmp_neigh_node; + } + rcu_read_unlock(); - if (!neigh_node) - neigh_node = create_neighbor(orig_node, - orig_neigh_node, - orig_neigh_node->orig, - if_incoming); - if (!neigh_node) - goto out; + if (!neigh_node) + neigh_node = create_neighbor(orig_node, + orig_neigh_node, + orig_neigh_node->orig, + if_incoming); + if (!neigh_node) + goto out; + /* if orig_node is direct neighbour update neigh_node last_valid */ + if (orig_node == orig_neigh_node) { neigh_node->last_valid = jiffies; - } else { - /* find packet count of corresponding one hop neighbor */ - rcu_read_lock(); - hlist_for_each_entry_rcu(tmp_neigh_node, node, - &orig_neigh_node->neigh_list, list) { - - if (!compare_eth(tmp_neigh_node->addr, - orig_neigh_node->orig)) - continue; - - if (tmp_neigh_node->if_incoming != if_incoming) - continue; - - if (!atomic_inc_not_zero(&tmp_neigh_node->refcount)) - continue; - - neigh_node = tmp_neigh_node; - } - rcu_read_unlock(); - - if (!neigh_node) - neigh_node = create_neighbor(orig_neigh_node, - orig_neigh_node, - orig_neigh_node->orig, - if_incoming); - if (!neigh_node) - goto out; - } orig_node->last_valid = jiffies; + /* find packet count of corresponding one hop neighbor */ spin_lock_bh(&orig_node->ogm_cnt_lock); orig_eq_count = orig_neigh_node->bcast_own_sum[if_incoming->if_num]; neigh_rq_count = neigh_node->real_packet_count; -- 1.7.1 -- Daniele Furlan

1 0

[B.A.T.M.A.N.] [PATCH] batman-adv: Fix rcu_lock imbalance in softif_neigh_get
by Sven Eckelmann 04 May '11

04 May '11

Fix extra rcu_read_unlock after an failed softif_neigh_vid_get in softif_neigh_get introduced by 0281a6aab718965d7c0ef45c4ba22e305f6aefe7 Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- soft-interface.c | 7 ++++--- 1 files changed, 4 insertions(+), 3 deletions(-) diff --git a/soft-interface.c b/soft-interface.c index 1827734..8023c4e 100644 --- a/soft-interface.c +++ b/soft-interface.c @@ -175,12 +175,12 @@ static struct softif_neigh *softif_neigh_get(struct bat_priv *bat_priv, continue; softif_neigh->last_seen = jiffies; - goto out; + goto unlock; } softif_neigh = kzalloc(sizeof(struct softif_neigh), GFP_ATOMIC); if (!softif_neigh) - goto out; + goto unlock; memcpy(softif_neigh->addr, addr, ETH_ALEN); softif_neigh->last_seen = jiffies; @@ -193,8 +193,9 @@ static struct softif_neigh *softif_neigh_get(struct bat_priv *bat_priv, &softif_neigh_vid->softif_neigh_list); spin_unlock_bh(&bat_priv->softif_neigh_lock); -out: +unlock: rcu_read_unlock(); +out: if (softif_neigh_vid) softif_neigh_vid_free_ref(softif_neigh_vid); return softif_neigh; -- 1.7.4.4

2 1

[B.A.T.M.A.N.] [PATCHv4 1/2] batman-adv: Remove unnecessary hardif_list_lock
by Sven Eckelmann 04 May '11

04 May '11

hardif_list_lock is unneccessary because we already ensure that no multiple admin operations can take place through rtnl_lock. hardif_list_lock only adds additional overhead and complexity. Critical functions now check whether they are called with rtnl_lock using ASSERT_RTNL. It indirectly fixes the problem that orig_hash_del_if() expects that only one interface is deleted from hardif_list at a time, but hardif_remove_interfaces() removes all at once and then calls orig_hash_del_if(). Reported-by: Linus Lüssing <linus.luessing(a)web.de> Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- Addressed Linus' comments: * Merged patch 1/3 and 2/3 again * Ensure that primary_if_select is only called with rtnl_lock through hardif_enable_interface (was fixed in patch 3/3... but now in a patch earlier) * Add a comment above hardif_list to inform about needed locking bat_sysfs.c | 2 ++ hard-interface.c | 30 +++++++----------------------- main.c | 3 +++ 3 files changed, 12 insertions(+), 23 deletions(-) diff --git a/bat_sysfs.c b/bat_sysfs.c index e449bf6..85ba20d 100644 --- a/bat_sysfs.c +++ b/bat_sysfs.c @@ -502,7 +502,9 @@ static ssize_t store_mesh_iface(struct kobject *kobj, struct attribute *attr, rtnl_unlock(); } + rtnl_lock(); ret = hardif_enable_interface(hard_iface, buff); + rtnl_unlock(); out: hardif_free_ref(hard_iface); diff --git a/hard-interface.c b/hard-interface.c index 3e888f1..7e2f772 100644 --- a/hard-interface.c +++ b/hard-interface.c @@ -31,9 +31,6 @@ #include <linux/if_arp.h> -/* protect update critical side of hardif_list - but not the content */ -static DEFINE_SPINLOCK(hardif_list_lock); - static int batman_skb_recv(struct sk_buff *skb, struct net_device *dev, @@ -136,7 +133,7 @@ static void primary_if_select(struct bat_priv *bat_priv, struct hard_iface *curr_hard_iface; struct batman_packet *batman_packet; - spin_lock_bh(&hardif_list_lock); + ASSERT_RTNL(); if (new_hard_iface && !atomic_inc_not_zero(&new_hard_iface->refcount)) new_hard_iface = NULL; @@ -148,7 +145,7 @@ static void primary_if_select(struct bat_priv *bat_priv, hardif_free_ref(curr_hard_iface); if (!new_hard_iface) - goto out; + return; batman_packet = (struct batman_packet *)(new_hard_iface->packet_buff); batman_packet->flags = PRIMARIES_FIRST_HOP; @@ -161,9 +158,6 @@ static void primary_if_select(struct bat_priv *bat_priv, * our new primary interface */ atomic_set(&bat_priv->hna_local_changed, 1); - -out: - spin_unlock_bh(&hardif_list_lock); } static bool hardif_is_iface_up(struct hard_iface *hard_iface) @@ -456,6 +450,8 @@ static struct hard_iface *hardif_add_interface(struct net_device *net_dev) struct hard_iface *hard_iface; int ret; + ASSERT_RTNL(); + ret = is_valid_iface(net_dev); if (ret != 1) goto out; @@ -482,10 +478,7 @@ static struct hard_iface *hardif_add_interface(struct net_device *net_dev) atomic_set(&hard_iface->refcount, 2); check_known_mac_addr(hard_iface->net_dev); - - spin_lock(&hardif_list_lock); list_add_tail_rcu(&hard_iface->list, &hardif_list); - spin_unlock(&hardif_list_lock); return hard_iface; @@ -499,6 +492,8 @@ out: static void hardif_remove_interface(struct hard_iface *hard_iface) { + ASSERT_RTNL(); + /* first deactivate interface */ if (hard_iface->if_status != IF_NOT_IN_USE) hardif_disable_interface(hard_iface); @@ -514,20 +509,11 @@ static void hardif_remove_interface(struct hard_iface *hard_iface) void hardif_remove_interfaces(void) { struct hard_iface *hard_iface, *hard_iface_tmp; - struct list_head if_queue; - INIT_LIST_HEAD(&if_queue); - - spin_lock(&hardif_list_lock); + rtnl_lock(); list_for_each_entry_safe(hard_iface, hard_iface_tmp, &hardif_list, list) { list_del_rcu(&hard_iface->list); - list_add_tail(&hard_iface->list, &if_queue); - } - spin_unlock(&hardif_list_lock); - - rtnl_lock(); - list_for_each_entry_safe(hard_iface, hard_iface_tmp, &if_queue, list) { hardif_remove_interface(hard_iface); } rtnl_unlock(); @@ -556,9 +542,7 @@ static int hard_if_event(struct notifier_block *this, hardif_deactivate_interface(hard_iface); break; case NETDEV_UNREGISTER: - spin_lock(&hardif_list_lock); list_del_rcu(&hard_iface->list); - spin_unlock(&hardif_list_lock); hardif_remove_interface(hard_iface); break; diff --git a/main.c b/main.c index 709b33b..39feb5a 100644 --- a/main.c +++ b/main.c @@ -33,6 +33,9 @@ #include "vis.h" #include "hash.h" + +/* List manipulations on hardif_list have to be rtnl_lock()'ed, + * list traversals just rcu-locked */ struct list_head hardif_list; unsigned char broadcast_addr[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff}; -- 1.7.4.4

3 8

[B.A.T.M.A.N.] [PATCH] batman-adv: cosmetic: Use ethtool ethtool_cmd_speed API
by Sven Eckelmann 04 May '11

04 May '11

From: David Decotigny <decot(a)google.com> This updates the network drivers so that they don't access the ethtool_cmd::speed field directly, but use ethtool_cmd_speed() instead. For most of the drivers, these changes are purely cosmetic and don't fix any problem, such as for those 1GbE/10GbE drivers that indirectly call their own ethtool get_settings()/mii_ethtool_gset(). The changes are meant to enforce code consistency and provide robustness with future larger throughputs, at the expense of a few CPU cycles for each ethtool operation. All drivers compiled with make allyesconfig ion x86_64 have been updated. Tested: make allyesconfig on x86_64 + e1000e/bnx2x work Signed-off-by: David Decotigny <decot(a)google.com> Signed-off-by: David S. Miller <davem(a)davemloft.net> [sven(a)narfation.org: compat function added] Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- This patch is now part of net-next-2.6 and will be in linux 2.6.40 compat.h | 3 +++ soft-interface.c | 2 +- 2 files changed, 4 insertions(+), 1 deletions(-) diff --git a/compat.h b/compat.h index 0c5ad82..c3fd2cd 100644 --- a/compat.h +++ b/compat.h @@ -183,6 +183,9 @@ static inline char *pack_hex_byte(char *buf, u8 byte) #if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 27) +#define ethtool_cmd_speed_set(_ep, _speed) \ + do { (_ep)->speed = (_speed); } while (0) + #ifndef dereference_function_descriptor #define dereference_function_descriptor(p) (p) #endif diff --git a/soft-interface.c b/soft-interface.c index 1772e2b..9e5fcd1 100644 --- a/soft-interface.c +++ b/soft-interface.c @@ -723,7 +723,7 @@ static int bat_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) { cmd->supported = 0; cmd->advertising = 0; - cmd->speed = SPEED_10; + ethtool_cmd_speed_set(cmd, SPEED_10); cmd->duplex = DUPLEX_FULL; cmd->port = PORT_TP; cmd->phy_address = 0; -- 1.7.4.4

2 1

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

B.A.T.M.A.N May 2011