B.A.T.M.A.N November 2010

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

16 participants
27 discussions

[B.A.T.M.A.N.] batman-adv 2010.2.0 released
by Marek Lindner 21 Nov '10

21 Nov '10

The B.A.T.M.A.N. team proudly presents the last batman-adv release in the year 2010, carrying the version number 2010.2.0. This release contains a mix of new features, Linux kernel compliance and bug fixes. As the kernel module always depends on the Linux kernel it was compiled against, it does not make sense to provide binaries on our website. As usual, you will find the signed tarballs in our download section: http://downloads.open-mesh.org/batman/releases/batman-adv-2010.2.0/ as well as prepackaged binaries in your distribution. Important changes ---------------- The batman-adv module now supports multiple mesh clouds simultaneously which made changes to the 'bat0' interface behavior necessary: Mesh interfaces (such as bat0) are created only after 'real' interfaces (e.g. wlan0) have been assigned. Mesh interfaces that have no real interfaces assigned are removed from the system. Furthermore the mesh interface can be freely named such as bat5, mesh0, etc, therefore batman-adv might be active even though bat0 does not exist. More details can be found below as well as in our documentation. Thanks ---------------- Thanks to all people sending in patches: * Andreas Langer <an.langer(a)gmx.de> * Andy Shevchenko <andy.shevchenko(a)gmail.com> * Chris Lang <clang(a)gateworks.com> * Linus Lüssing <linus.luessing(a)web.de> * Marek Lindner <lindner_marek(a)yahoo.de> * Simon Wunderlich <siwu(a)hrz.tu-chemnitz.de> * Sven Eckelmann <sven.eckelmann(a)gmx.de> * Tracey Dent <tdent48227(a)gmail.com> * Vasiliy Kulikov <segooon(a)gmail.com> and to all those that supported us with good advice or rigorous testing: * Dan Carpenter <error27(a)gmail.com> * David S. Miller <davem(a)davemloft.net> * Paul E. McKenney <paulmck(a)linux.vnet.ibm.com> * Rafal Lesniak <lesniak(a)eresi-project.org> * Sam Yeung <sam.cwyeung(a)gmail.com> batman-adv ---------------- The work that nearly a year ago began has finally come to an end: Multiple mesh cloud support has been completed. This allows one host to participate in multiple meshes at the same time via different interfaces. This often requested feature is particularly useful to segment large batman-adv installations, thereby reducing overhead. Each mesh cloud can be individually configured, deals with its traffic flow independently and possesses its own batX interface. Also added was a layer 2 fragmentation for unicast data flowing through the mesh which will allow to run batman-adv over interfaces / connections that don't allow to increase the MTU beyond the standard Ethernet packet size of 1500 bytes. This initial fragmentation only supports splitting the large packets at "the beginning of mesh" whereas an extension which allows fragmenting "on the fly" anywhere in the mesh network will be included in the following release. As usual, many improvements happened under the hood: The module uses the kernel supplied functions for network buffer manipulations instead of re- implementing them and takes advantage of the optimized kernel functionality for OGM counting. The RCU protected lists have been revamped and enhanced with referenced based counting which squashes quite some race conditions. The HNA handling needed some fixes too when it was stress tested with a very large local network. Many minor bugs such as false warnings, reduced memory usage and coding style issues were addressed. batctl ---------------- To support the "multiple mesh clouds" feature batctl received a new (optional) parameter '-m' to specify the mesh cloud interface in order to query the mesh cloud of interest. If '-m' is not specified batctl will assume 'bat0' as mesh cloud interface which maintains backward compatibility and suits most users. The newly added fragmentation on/off switch also was added as well as batctl tcpdump support for fragmented packets. In addition, the batctl tcpdump can deal with multiple header encapsulation (e.g. HTTP in VLAN in BAT UNICAST in VLAN) and the batctl ping utilities' '-i' option is not silently ignored any longer. Various cleanups took place and a crash fixed if the HOME environment variable is not set. Happy routing, The B.A.T.M.A.N. team

1 0

[B.A.T.M.A.N.] [PATCH] batman-adv: Rename if_list_queue to if_queue
by Sven Eckelmann 21 Nov '10

21 Nov '10

The name "if_list_queue" would create lines longer than 80 chars which can be avoided by shortened version "if_queue". Signed-off-by: Sven Eckelmann <sven.eckelmann(a)gmx.de> --- batman-adv/hard-interface.c | 8 ++++---- 1 files changed, 4 insertions(+), 4 deletions(-) diff --git a/batman-adv/hard-interface.c b/batman-adv/hard-interface.c index bf98b90..4f95777 100644 --- a/batman-adv/hard-interface.c +++ b/batman-adv/hard-interface.c @@ -476,19 +476,19 @@ static void hardif_remove_interface(struct batman_if *batman_if) void hardif_remove_interfaces(void) { struct batman_if *batman_if, *batman_if_tmp; - struct list_head if_list_queue; + struct list_head if_queue; - INIT_LIST_HEAD(&if_list_queue); + INIT_LIST_HEAD(&if_queue); spin_lock(&if_list_lock); list_for_each_entry_safe(batman_if, batman_if_tmp, &if_list, list) { list_del_rcu(&batman_if->list); - list_add_tail(&batman_if->list, &if_list_queue); + list_add_tail(&batman_if->list, &if_queue); } spin_unlock(&if_list_lock); rtnl_lock(); - list_for_each_entry_safe(batman_if, batman_if_tmp, &if_list_queue, list) { + list_for_each_entry_safe(batman_if, batman_if_tmp, &if_queue, list) { hardif_remove_interface(batman_if); } rtnl_unlock(); -- 1.7.2.3

2 1

[B.A.T.M.A.N.] [PATCH] batman-adv: Don't remove interface with spinlock held
by Sven Eckelmann 21 Nov '10

21 Nov '10

We call a lot of the netdevice code when holding if_list_lock which will spin the whole time. This is not necessary because we only want to protect the access to the list to be serialized. An extra queue can be used which hold all interfaces which should be removed and then use that queue without any locks for netdevice cleanup. Signed-off-by: Sven Eckelmann <sven.eckelmann(a)gmx.de> --- batman-adv/hard-interface.c | 19 +++++++++++++------ 1 files changed, 13 insertions(+), 6 deletions(-) diff --git a/batman-adv/hard-interface.c b/batman-adv/hard-interface.c index e22f0dd..bf98b90 100644 --- a/batman-adv/hard-interface.c +++ b/batman-adv/hard-interface.c @@ -469,9 +469,6 @@ static void hardif_remove_interface(struct batman_if *batman_if) return; batman_if->if_status = IF_TO_BE_REMOVED; - - /* caller must take if_list_lock */ - list_del_rcu(&batman_if->list); sysfs_del_hardif(&batman_if->hardif_obj); call_rcu(&batman_if->rcu, hardif_free_rcu); } @@ -479,13 +476,21 @@ static void hardif_remove_interface(struct batman_if *batman_if) void hardif_remove_interfaces(void) { struct batman_if *batman_if, *batman_if_tmp; + struct list_head if_list_queue; + + INIT_LIST_HEAD(&if_list_queue); - rtnl_lock(); spin_lock(&if_list_lock); list_for_each_entry_safe(batman_if, batman_if_tmp, &if_list, list) { - hardif_remove_interface(batman_if); + list_del_rcu(&batman_if->list); + list_add_tail(&batman_if->list, &if_list_queue); } spin_unlock(&if_list_lock); + + rtnl_lock(); + list_for_each_entry_safe(batman_if, batman_if_tmp, &if_list_queue, list) { + hardif_remove_interface(batman_if); + } rtnl_unlock(); } @@ -512,8 +517,10 @@ static int hard_if_event(struct notifier_block *this, break; case NETDEV_UNREGISTER: spin_lock(&if_list_lock); - hardif_remove_interface(batman_if); + list_del_rcu(&batman_if->list); spin_unlock(&if_list_lock); + + hardif_remove_interface(batman_if); break; case NETDEV_CHANGEMTU: if (batman_if->soft_iface) -- 1.7.2.3

2 1

[B.A.T.M.A.N.] batman-adv crash on Intel 3945ABG & 4965 AGN
by Marek Lindner 21 Nov '10

21 Nov '10

Hi, 2 days ago I proposed a patch which fixes the crash you reported in ticket #141. Although you confirmed the patch solves the issue I haven't pushed it yet as the cause seems to be a bug within the Intel driver and should be fixed there (unless we do something horribly wrong). Could you please let us know the driver version that was used in your test setup ? A 'modinfo iwlagn | grep version' should do. We will then get in touch with the Intel wifi maintainers. Regards, Marek

2 2

[B.A.T.M.A.N.] performance patches
by Marek Lindner 20 Nov '10

20 Nov '10

Hi, after running a series of tests I noticed that batman-adv introduces a 10-20% performance overhead (depending on various factors) compared standard routing. I could identify the orig_hash locking as the main bottleneck as code all over the module requires to hold the lock causing a serious slowdown. During the last days I worked on a set of patches meant to increase the performance by replacing the orig_hash lock with rcu-locking and reference counting. At the moment these patches are highly experimental because they fundamentally alter main data structures batman-adv possesses since it was first released. It is very likely that they will crash your system and/or introduce other funny behavior. However, feedback is highly appreciated due to the complex nature of the changes. Known issues: * bonding / alternating candidates are not secured by refcounting (see find_router() * bcast_own / bcast_own_sum need protection, especially in orig_hash_add_if() / orig_hash_del_if() * NULL pointer dereference when an orig_node gets purged Cheers, Marek

2 10

Re: [B.A.T.M.A.N.] patch "staging: batman-adv: Use linux/etherdevice.h address helper functions" added to staging tree
by Marek Lindner 17 Nov '10

17 Nov '10

Greg, > The patch is only compile-tested. [..] > diff --git a/drivers/staging/batman-adv/main.c > b/drivers/staging/batman-adv/main.c index 0587940..6ea6420 100644 > --- a/drivers/staging/batman-adv/main.c > +++ b/drivers/staging/batman-adv/main.c > @@ -149,7 +149,7 @@ void dec_module_count(void) > > int compare_orig(void *data1, void *data2) > { > - return (memcmp(data1, data2, ETH_ALEN) == 0 ? 1 : 0); > + return (compare_ether_addr(data1, data2) == 0 ? 1 : 0); > } I'm wondering why you accepted this patch despite the raised objections regarding alignment problems. [1][2] Is there something you know that we overlooked ? We haven't heard from Thomas or you since the alignment question came on the table. Cheers, Marek [1]https://lists.open-mesh.org/pipermail/b.a.t.m.a.n/2010-November/003640.ht… [2]https://lists.open-mesh.org/pipermail/b.a.t.m.a.n/2010-November/003643.ht…

4 6

[B.A.T.M.A.N.] [PATCH 1/2] batman-adv: Use kernel version min macro
by Sven Eckelmann 17 Nov '10

17 Nov '10

The kernel headers already provide different versions of a min/max macro which should be used by all modules according to Documentation/CodingStyle. Signed-off-by: Sven Eckelmann <sven.eckelmann(a)gmx.de> --- batman-adv/hard-interface.c | 6 ++---- 1 files changed, 2 insertions(+), 4 deletions(-) diff --git a/batman-adv/hard-interface.c b/batman-adv/hard-interface.c index 5a1d631..e22f0dd 100644 --- a/batman-adv/hard-interface.c +++ b/batman-adv/hard-interface.c @@ -31,8 +31,6 @@ #include <linux/if_arp.h> -#define MIN(x, y) ((x) < (y) ? (x) : (y)) - /* protect update critical side of if_list - but not the content */ static DEFINE_SPINLOCK(if_list_lock); @@ -220,8 +218,8 @@ int hardif_min_mtu(struct net_device *soft_iface) if (batman_if->soft_iface != soft_iface) continue; - min_mtu = MIN(batman_if->net_dev->mtu - BAT_HEADER_LEN, - min_mtu); + min_mtu = min_t(int, batman_if->net_dev->mtu - BAT_HEADER_LEN, + min_mtu); } rcu_read_unlock(); out: -- 1.7.2.3

3 6

Re: [B.A.T.M.A.N.] patch "staging: batman-adv: Use linux/etherdevice.h address helper functions" added to staging tree
by Marek Lindner 15 Nov '10

15 Nov '10

Hi, > Hello, im a BATMAN user and also a member of the website. Im doing a > project using BATMAN for MANET.I actually plan to modify codes from > batman-0.3.2, to be able to modify the packet header and send the codes > between notebooks. Any kind soul able to point me in the direction to look > at? Which codes do i specifically modify? you wish to modify the batman daemon sources ? What kind of changes do you have in mind ? Note that HTML mails are not blocked on the batman mailing list. Cheers, Marek

1 0

Re: [B.A.T.M.A.N.] [Battlemesh] Another peer-reviewed assessment of Babel, Batman and OLSR
by Marek Lindner 15 Nov '10

15 Nov '10

Hi, > We are happy to answer any questions people might have with that paper. Quote from: http://bridgingthelayers.org/docs/murray_routing.pdf "Upon inspection of the BATMAN L2 and BATMAN L3 results, the only major performance difference between data link layer routing and network layer routing is the packet size. Data link layer routing protocols will not require a network layer IPv4/IPv6 header and may therefore be smaller." Please note that the packet size of the protocol packets might be of vital importance to the IETF and IEEE but for the batman project that was a rather insignificant side effect. Having full control over the traffic flowing through the mesh was the main motivation to go down to layer 2. It allowed us to implement things like broadcast support & broadcast optimization (ARQ), multicast support & optimization, interface alternating (consult the second part of this article: http://www.open-mesh.org/wiki/2010-06-13-wbm2010-bracciano), seamless roaming for non-mesh participants. Regards, Marek

1 0

[B.A.T.M.A.N.] batman-adv gateway documentation
by Marek Lindner 12 Nov '10

12 Nov '10

Hi folks, the batman-adv gateway support has left the experimental stage and will show up in one of the following releases (probably early 2011). To make this feature a really useful one I wrote a wiki article to explain how it works and how it is supposed to be used: http://www.open-mesh.org/wiki/batman-adv-gateways Feel free to demand clarification if the text is not comprehensible. ;-) Regards, Marek

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

B.A.T.M.A.N November 2010