B.A.T.M.A.N January 2011

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

20 participants
35 discussions

[B.A.T.M.A.N.] [PATCH] Even Batman should not dereference NULL pointers
by Jesper Juhl 13 Jan '11

13 Jan '11

There's a problem in net/batman-adv/unicast.c::frag_send_skb(). dev_alloc_skb() allocates memory and may fail, thus returning NULL. If this happens we'll pass a NULL pointer on to skb_split() which in turn hands it to skb_split_inside_header() from where it gets passed to skb_put() that lets skb_tail_pointer() play with it and that function dereferences it. And thus the bat dies. While I was at it I also moved the call to dev_alloc_skb() above the assignment to 'unicast_packet' since there's no reason to do that assignment if the memory allocation fails. Signed-off-by: Jesper Juhl <jj(a)chaosbits.net> --- unicast.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/net/batman-adv/unicast.c b/net/batman-adv/unicast.c index dc2e28b..ee41fef 100644 --- a/net/batman-adv/unicast.c +++ b/net/batman-adv/unicast.c @@ -229,10 +229,12 @@ int frag_send_skb(struct sk_buff *skb, struct bat_priv *bat_priv, if (!bat_priv->primary_if) goto dropped; - unicast_packet = (struct unicast_packet *) skb->data; + frag_skb = dev_alloc_skb(data_len - (data_len / 2) + ucf_hdr_len); + if (!frag_skb) + goto dropped; + unicast_packet = (struct unicast_packet *) skb->data; memcpy(&tmp_uc, unicast_packet, uc_hdr_len); - frag_skb = dev_alloc_skb(data_len - (data_len / 2) + ucf_hdr_len); skb_split(skb, frag_skb, data_len / 2); if (my_skb_head_push(skb, ucf_hdr_len - uc_hdr_len) < 0 || -- Jesper Juhl <jj(a)chaosbits.net> http://www.chaosbits.net/ Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html Plain text mails only, please.

2 1

[B.A.T.M.A.N.] bitarray.c / packet-counting / window questions
by Linus Lüssing 13 Jan '11

13 Jan '11

Hi everyone, For the NDP patches, the usage of the methods for determening the link qualities with the window is the code I'm most uncertain about (right after the fancy rcu-locking/refcounting ;) ). Therefore I'd like to ask some questions about code in bitarray.c and routing.c that came to my mind while reading some of those parts and which parts confused me, partly due to their naming style I think. It's great though, that the purpose of the methods in bitarray.c are well explained, I'm just missing some minor cases that are not explained in their comment. - for get_bit_status() it's especially not clear to me, what is meant with "in range" in its comment. is it supposed to be "in range" if it is after or before the last_seqno? - bit_get_packet(): just for clarification, if set_mark is set, that means the window can still be slided, but the bit for the according seqno in the window is not being set, right? (what do you think about adding a sentence about its purpose in this methods preceding comment?) - what does bit_get_packet() return in case of a not very old, but just old packet which is still inside of the window? (the comment does not seem to cover this case) - what do you think of renaming bit_get_packet() to something like update_window()? With function names having a "get" in their name I'd usually think that they'd not modify anything. And for instance renaming set_mark to "slide_only" or something else that might explain a little more the difference when setting or not setting it? - for the need_update in the code, this is just refering to an updated last_real_seqno, right? need_update could also be 0 in case of a out-of-order and not too old ogm, right? (maybe renaming it to something like window_slided or update_last_rq_seqno?) - in count_real_packets(), why are the window update functions (bit_get_packet(), bit_packet_count()) still called in case of is_duplicate == 1? Do you think it might make sense to add some more abstraction in routing.c to reduce the number of functions in bitarray.c you have to look at and their purpose you'd have to get your head around? To avoid having to think of the window operations in routing.c and abstract that part? And one more question for the TQ_LOCAL_WINDOW_SIZE, when i.e. I'd change that to a lower value, for instance 4, I'd have to set WORD_BIT_SIZE to (sizeof(unsigned long) / 2), right? Would that cause any trouble on the code side (and yeah, I know that this is causing unusable link quality measurements, but I wanted to reduce that to play a little with an exponentially weighted moving average on top of that). Cheers, Linus

2 2

[B.A.T.M.A.N.] NDP patches v4
by Linus Lüssing 12 Jan '11

12 Jan '11

Hi everyone, thanks for all the feedback and reviewing again. The following updated patches address Marek's comments and suggestions: * Introducing a commit for renaming only * batman_packet_ogm -> ogm_packet, batman_packet_ndp -> ndp_packet * BAT_UNICAST followed by BAT_UNICAST_FRAG * structs introduced in the patches where they're used, not earlier (ndp_packet, neigh_entry) * moved own_ndp_send_time to ndp.c * checkpatch cleaning * avoid losing ndp-interval setting on interface down/up * fixed a potential memory leak in error cases * skb_copy instead of skb_clone * rebased to trunk * removal of unnecessary include * hlist instead of doubly linked list * added a break in one list traversal * secure neigh_list with rcu-locking + ref-counting * rename rq_real_bits to ndp_rq_window * use msecs_to_jiffies instead of manual HZ multiplication * fix get_batman_if_by_netdev usage / refcounting * adding licenses Furthermore I've changed the following: * use dev_kfree_skb instead of kfree_skb in case of successful ndp packet reception and processing (so no simple NET_RX_DROP return value) * adding include guards for ndp.h * use "= seqno - 1" in ndp_create_neighbor(), otherwise ndp_update_neighbor_lq() does not update tq/rq/last_valid on first ndp packet Would be great if someone could check the usage of rcu-locking + refcounting. I was also a little confused because in "Documentation/RCU/listRCU.txt" list_del_rcu() and list_add_rcu() are not protected with a spinlock for the list here, but in the batman-adv code we are usually having those extra locks. Do I have to leave those spinlocks or can I remove them for adding/deleting entries in the neigh_list? Cheers, Linus

4 14

[B.A.T.M.A.N.] open-mesh feature request submission location?
by David Sutherland 06 Jan '11

06 Jan '11

Happy new year all. Would someone point me to the best place to submit a request for a feature for a version of the NG Open-Mesh firmware if it is not here? I am very excited about the feature described here: http://support.open-mesh.com/activekb/questions.php?questionid=35 > "Bridge mode for SSID#2. This allows users on the private, encrypted SSID the ability to access LAN resources as it will be the LAN that assigns all client DHCP addresses. This is selected on the SSID#2 tab." I was thinking bridging would be the new feature that pushes me to sell off my Meraki Pro equipment and move all to Open-Mesh but then I realized today that there is no AAA on SSID #2. So if there any chance this Bridge Mode can be implemented on the SSID#1 where the captive portal also works, or is this a technical impossibility? *** Right now I've got a small non-profit training center where the students would like to be able to browse and find printers and shares on the network, but I also want each user to have private login to get on the network. So I don't currently see any solutions. Any suggestions? *** Thanks! David Sutherland Lindale, TX -- David C. Sutherland | (310) 729-6411 (c)

2 1

[B.A.T.M.A.N.] Mesh network topology problem
by Michael Y. Gluhovchenko 05 Jan '11

05 Jan '11

Hello All; I have a problem organizing mesh, each wireless router has a Wifi link and 2 wired links. The problem is that there is the parallel corridor with the same infrastructure. Routers can see each other across the corridor, the signal quality in the corridor is better than the main line. I want to use Linux bridge + stp interfaces bat0, eth0, eth1 wired links have advantages. When break wire links or failure one of the routers need to maintain communication through the mesh. Is it possible to configure mesh so that the link in another corridor with the best quality work in the last turn? What problems can occur for that network topology. With best regards, Michael;

2 2

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

B.A.T.M.A.N January 2011