[B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Vinay Tharigopula]

Hello..

I succesfully configured a batman-adv mesh to add access points to our 
existing wired lan.  I am using Openwrt 10.04 and batman-adv 2010.1.0. 
DHCP and everything works as expected. Any wired or wireless client 
connecting to the network gets an IP from a DHCP server hooked into the 
switch etc etc.

Our Network configuration:

8 batman-adv routers with a wireless mesh  with two or more of the 
routers having a wired connection directly to the switch. In this case, 
the 8 routers decide to pick one of the "plugged in" routers as a 
gateway to avoid bridge loops. This is fine.

I have been running into issues with wired clients however.

When a wired client plugged into the switch tries to contact a wired 
client plugged into the same switch, the packets manage to get lost. I 
believe they are are going through the wireless routers plugged into the 
same switch.

Instead of going like this..

PC 1 ----> switch1 ---> PC 2

they seem to be going like this... (my guess).

PC 1 ---> switch1 ---> wireless router 1 ----(wireless mesh link)----> 
wireless router 2 ----> switch1 --->  PC 2.

I think this is because the translation tables on the bat node are 
announcing that even the wired clients are connected to them.

This is causing extreme delays even in wired clients. Any help will be 
appreciated. Should I enable STP in on the switch ports ?

- Vinay

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Linus Lüssing]

Hi Vinay,

On Thu, Feb 24, 2011 at 12:02:55PM -0500, Vinay Tharigopula wrote:
> Hello..
> 
> I succesfully configured a batman-adv mesh to add access points to
> our existing wired lan.  I am using Openwrt 10.04 and batman-adv
> 2010.1.0. DHCP and everything works as expected. Any wired or
> wireless client connecting to the network gets an IP from a DHCP
> server hooked into the switch etc etc.
The bridge-loop-avoidance feature has been added in 2011.0.0 and
is not available in 2010.1.0.

If switching to 2011.0.0 helps, don't read any further :).


Could you explain a little more which interfaces you've added to
batman-adv and which interfaces you are bridging with bat0?
(ifconfig, brctl show, batctl if output from one router would be
helpful)

> 
> Our Network configuration:
> 
> 8 batman-adv routers with a wireless mesh  with two or more of the
> routers having a wired connection directly to the switch. In this
> case, the 8 routers decide to pick one of the "plugged in" routers
> as a gateway to avoid bridge loops. This is fine.
> 
> I have been running into issues with wired clients however.
> 
> When a wired client plugged into the switch tries to contact a wired
> client plugged into the same switch, the packets manage to get lost.
> I believe they are are going through the wireless routers plugged
> into the same switch.
> 
> Instead of going like this..
> 
> PC 1 ----> switch1 ---> PC 2
> 
> they seem to be going like this... (my guess).
> 
> PC 1 ---> switch1 ---> wireless router 1 ----(wireless mesh
> link)----> wireless router 2 ----> switch1 --->  PC 2.
Could you please verify with "batctl td" (or tcpdump/wireshark) if
the packets are leaving on the wireless router 1's wifi interface?

> 
> I think this is because the translation tables on the bat node are
> announcing that even the wired clients are connected to them.
Could you check with batctl tl and batctl tg if these tables are changing
rapidly? If you do not have any client roaming, then these tables are
expected to stay rather static.

> 
> This is causing extreme delays even in wired clients. Any help will
> be appreciated. Should I enable STP in on the switch ports ?
It does not sound like having disabled STP is your problem.

> 
> - Vinay
> 

Cheers, Linus

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Vinay Tharigopula]

Linus, Thanks for replying....

I am having trouble finding binaries of batman-adv 2011.0 for OpenWRT . 
Are they available or do I have to build them myself  ?

Interfaces:

On my router, totally there are 4 interfaces.

wlan0 - The infrastructure -type network - with Test1 SSID
wlan1-  The adhoc network with BSSID "xx:xx:xx:xx:xx:xx"
eth0 - the 4 lan ports.
eth1 - the wan port.

bat0 has wlan1 added to it.

There is a bridge "br-lan" with bat0 eth0 eth1 and wlan0.

On the "chosen gateway" router. I have observed this.

batctl tg

Globally announced HNAs received via the mesh bat0
  * 7c:c5:37:6a:ca:cd via 06:24:a5:d8:5d:19
  * e2:45:34:51:32:1c via 06:24:a5:d8:5c:16
  * 00:24:a5:d8:5c:16 via 06:24:a5:d8:5c:16
  * 06:b0:29:d9:45:e5 via 06:24:a5:af:f9:fb
  * 7c:c5:37:78:3c:6c via 06:24:a5:d8:5d:19
  * 00:24:a5:af:f9:fb via 06:24:a5:af:f9:fb
  * 00:1e:c1:c7:53:c2 via 06:24:a5:af:f9:fb
  * 00:24:a5:d8:5d:19 via 06:24:a5:d8:5d:19
  * 92:b3:f4:d5:c6:f1 via 06:24:a5:d8:5d:19

Note: This is from a group of 4 routers with 2 plugged in via wire.

batctl tl.

It is announcing around 60 hosts and constantly increasing , pretty much 
all the hosts connected to the switch that it is plugged into. !

-Vinay

On 2/24/2011 2:44 PM, Linus Lüssing wrote:
> Hi Vinay,
>
> On Thu, Feb 24, 2011 at 12:02:55PM -0500, Vinay Tharigopula wrote:
>> Hello..
>>
>> I succesfully configured a batman-adv mesh to add access points to
>> our existing wired lan.  I am using Openwrt 10.04 and batman-adv
>> 2010.1.0. DHCP and everything works as expected. Any wired or
>> wireless client connecting to the network gets an IP from a DHCP
>> server hooked into the switch etc etc.
> The bridge-loop-avoidance feature has been added in 2011.0.0 and
> is not available in 2010.1.0.
>
> If switching to 2011.0.0 helps, don't read any further :).
>
>
> Could you explain a little more which interfaces you've added to
> batman-adv and which interfaces you are bridging with bat0?
> (ifconfig, brctl show, batctl if output from one router would be
> helpful)
>
>> Our Network configuration:
>>
>> 8 batman-adv routers with a wireless mesh  with two or more of the
>> routers having a wired connection directly to the switch. In this
>> case, the 8 routers decide to pick one of the "plugged in" routers
>> as a gateway to avoid bridge loops. This is fine.
>>
>> I have been running into issues with wired clients however.
>>
>> When a wired client plugged into the switch tries to contact a wired
>> client plugged into the same switch, the packets manage to get lost.
>> I believe they are are going through the wireless routers plugged
>> into the same switch.
>>
>> Instead of going like this..
>>
>> PC 1 ---->  switch1 --->  PC 2
>>
>> they seem to be going like this... (my guess).
>>
>> PC 1 --->  switch1 --->  wireless router 1 ----(wireless mesh
>> link)---->  wireless router 2 ---->  switch1 --->   PC 2.
> Could you please verify with "batctl td" (or tcpdump/wireshark) if
> the packets are leaving on the wireless router 1's wifi interface?
>
>> I think this is because the translation tables on the bat node are
>> announcing that even the wired clients are connected to them.
> Could you check with batctl tl and batctl tg if these tables are changing
> rapidly? If you do not have any client roaming, then these tables are
> expected to stay rather static.
>
>> This is causing extreme delays even in wired clients. Any help will
>> be appreciated. Should I enable STP in on the switch ports ?
> It does not sound like having disabled STP is your problem.
>
>> - Vinay
>>
> Cheers, Linus

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Marek Lindner]

Hi Vinay,

> I am having trouble finding binaries of batman-adv 2011.0 for OpenWRT . 
> Are they available or do I have to build them myself  ?

as Linus already pointed out - bridge loop avoidance is not available before 
batman-adv 2011.0.0. That is what you need to address first.

Where did you get the binaries for batman-adv 2010.1.0 ? Does this source also 
provide newer binaries ? Unlike user space application a kernel module heavily 
depends on your system (kernel version / patches / etc). Which means nobody 
can give you binaries without having access to exactly the same system you are 
using.

Regards,
Marek

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Vinay Tharigopula]

Marek..

I got the binaries from the openwrt prebuilt packages. However, now I 
rebuilt the image and built batman-adv 2011.myself. I will try this now 
and let you know.

However, now I am running into a much bigger issue. If two routers are 
plugged in to the same switch, bridge loop avoidance doesnt work and 
both of them are stopping responding to pings.

Here is my config.

bat0 if:
wlan1: active.

wlan1: Mesh Network.

wlan0: Infrastructure network.
eth0 : 4 lan ports
eth1: 1 wan port

br-lan : wlan0 eth0 eth1 bat0


In case anybody else wants the binaries for Openwrt 10.03.1 RC 4 , let 
me know I can send it to you.


-Vinay

On 2/24/2011 7:30 PM, Marek Lindner wrote:
> Hi Vinay,
>
>> I am having trouble finding binaries of batman-adv 2011.0 for OpenWRT .
>> Are they available or do I have to build them myself  ?
> as Linus already pointed out - bridge loop avoidance is not available before
> batman-adv 2011.0.0. That is what you need to address first.
>
> Where did you get the binaries for batman-adv 2010.1.0 ? Does this source also
> provide newer binaries ? Unlike user space application a kernel module heavily
> depends on your system (kernel version / patches / etc). Which means nobody
> can give you binaries without having access to exactly the same system you are
> using.
>
> Regards,
> Marek

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Vinay Tharigopula]

Update on this...

As I said previously, after updating to the latest batman-adv,  both 
routers connected to the lan stopped responding to pings.

I was playing around with the switch settings and setting " Edge + Root 
protection with STP" on the ports connected to the routers wired 
connection brought them back.  Now they act according to the bridge-loop 
avoidance document and pick one of the routere to be the primary router 
to the lan.

Any idea why this is happening ?

Thanks

Vinay


On 2/25/2011 8:44 AM, Vinay Tharigopula wrote:
> Marek..
>
> I got the binaries from the openwrt prebuilt packages. However, now I 
> rebuilt the image and built batman-adv 2011.myself. I will try this 
> now and let you know.
>
> However, now I am running into a much bigger issue. If two routers are 
> plugged in to the same switch, bridge loop avoidance doesnt work and 
> both of them are stopping responding to pings.
>
> Here is my config.
>
> bat0 if:
> wlan1: active.
>
> wlan1: Mesh Network.
>
> wlan0: Infrastructure network.
> eth0 : 4 lan ports
> eth1: 1 wan port
>
> br-lan : wlan0 eth0 eth1 bat0
>
>
> In case anybody else wants the binaries for Openwrt 10.03.1 RC 4 , let 
> me know I can send it to you.
>
>
> -Vinay
>
> On 2/24/2011 7:30 PM, Marek Lindner wrote:
>> Hi Vinay,
>>
>>> I am having trouble finding binaries of batman-adv 2011.0 for OpenWRT .
>>> Are they available or do I have to build them myself  ?
>> as Linus already pointed out - bridge loop avoidance is not available 
>> before
>> batman-adv 2011.0.0. That is what you need to address first.
>>
>> Where did you get the binaries for batman-adv 2010.1.0 ? Does this 
>> source also
>> provide newer binaries ? Unlike user space application a kernel 
>> module heavily
>> depends on your system (kernel version / patches / etc). Which means 
>> nobody
>> can give you binaries without having access to exactly the same 
>> system you are
>> using.
>>
>> Regards,
>> Marek
>

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Vinay Tharigopula]

Update on this...

As I said previously, after updating to the latest batman-adv,  both 
routers connected to the lan stopped responding to pings.

I was playing around with the switch settings and setting " Edge + Root 
protection with STP" on the ports connected to the routers wired 
connection brought them back.  Now they act according to the bridge-loop 
avoidance document and pick one of the routere to be the primary router 
to the lan.

Any idea why this is happening ?

Thanks

Vinay

On 2/25/2011 8:44 AM, Vinay Tharigopula wrote:
> Marek..
>
> I got the binaries from the openwrt prebuilt packages. However, now I 
> rebuilt the image and built batman-adv 2011.myself. I will try this 
> now and let you know.
>
> However, now I am running into a much bigger issue. If two routers are 
> plugged in to the same switch, bridge loop avoidance doesnt work and 
> both of them are stopping responding to pings.
>
> Here is my config.
>
> bat0 if:
> wlan1: active.
>
> wlan1: Mesh Network.
>
> wlan0: Infrastructure network.
> eth0 : 4 lan ports
> eth1: 1 wan port
>
> br-lan : wlan0 eth0 eth1 bat0
>
>
> In case anybody else wants the binaries for Openwrt 10.03.1 RC 4 , let 
> me know I can send it to you.
>
>
> -Vinay
>
> On 2/24/2011 7:30 PM, Marek Lindner wrote:
>> Hi Vinay,
>>
>>> I am having trouble finding binaries of batman-adv 2011.0 for OpenWRT .
>>> Are they available or do I have to build them myself  ?
>> as Linus already pointed out - bridge loop avoidance is not available 
>> before
>> batman-adv 2011.0.0. That is what you need to address first.
>>
>> Where did you get the binaries for batman-adv 2010.1.0 ? Does this 
>> source also
>> provide newer binaries ? Unlike user space application a kernel 
>> module heavily
>> depends on your system (kernel version / patches / etc). Which means 
>> nobody
>> can give you binaries without having access to exactly the same 
>> system you are
>> using.
>>
>> Regards,
>> Marek
>

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Linus Lüssing]

Hi Vinay,

On Fri, Feb 25, 2011 at 08:44:33AM -0500, Vinay Tharigopula wrote:
> Marek..
> 
> I got the binaries from the openwrt prebuilt packages. However, now
> I rebuilt the image and built batman-adv 2011.myself. I will try
> this now and let you know.
> 
> However, now I am running into a much bigger issue. If two routers
> are plugged in to the same switch, bridge loop avoidance doesnt work
> and both of them are stopping responding to pings.
Could you try "batctl ping" too and check whether this is working
or not.

If that's working but the normal pinging is not, then it should be
a layer 3 problem - in that case we'd need the outputs of "ip
route show" and "ip address show".

If batctl ping is already not working then "batctl o" and "batctl
if" output will be interesting.
> 
> Here is my config.
> 
> bat0 if:
> wlan1: active.
> 
> wlan1: Mesh Network.
> 
> wlan0: Infrastructure network.
> eth0 : 4 lan ports
> eth1: 1 wan port
> 
> br-lan : wlan0 eth0 eth1 bat0
Hmm, that looks sane and correct, should work like that.

Is there anything weird in the output of "dmesg"? Any errors or
warnings?

Cheers, Linus

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Linus Lüssing]

Hi Vinay,

On Fri, Feb 25, 2011 at 08:44:33AM -0500, Vinay Tharigopula wrote:
> Marek..
>
> I got the binaries from the openwrt prebuilt packages. However,
> now
> I rebuilt the image and built batman-adv 2011.myself. I will try
> this now and let you know.
>
> However, now I am running into a much bigger issue. If two
> routers
> are plugged in to the same switch, bridge loop avoidance doesnt
> work
> and both of them are stopping responding to pings.
Could you try "batctl ping" too and check whether this is working
or not.

If that's working but the normal pinging is not, then it should be
a layer 3 problem - in that case we'd need the outputs of "ip
route show" and "ip address show".

If batctl ping is already not working then "batctl o" and "batctl
if" output will be interesting.
>
> Here is my config.
>
> bat0 if:
> wlan1: active.
>
> wlan1: Mesh Network.
>
> wlan0: Infrastructure network.
> eth0 : 4 lan ports
> eth1: 1 wan port
>
> br-lan : wlan0 eth0 eth1 bat0
Hmm, that looks sane and correct, should work like that.

Is there anything weird in the output of "dmesg"? Any errors or
warnings?

Cheers, Linus

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Marek Lindner]

Hi,

Linus has some problems sending mails to the list, so I forward his reply.

Cheers,
Marek


============================================


Hi Vinay,

On Fri, Feb 25, 2011 at 08:44:33AM -0500, Vinay Tharigopula wrote:
> Marek..
>
> I got the binaries from the openwrt prebuilt packages. However,
> now
> I rebuilt the image and built batman-adv 2011.myself. I will try
> this now and let you know.
>
> However, now I am running into a much bigger issue. If two
> routers
> are plugged in to the same switch, bridge loop avoidance doesnt
> work
> and both of them are stopping responding to pings.
Could you try "batctl ping" too and check whether this is working
or not.

If that's working but the normal pinging is not, then it should be
a layer 3 problem - in that case we'd need the outputs of "ip
route show" and "ip address show".

If batctl ping is already not working then "batctl o" and "batctl
if" output will be interesting.
>
> Here is my config.
>
> bat0 if:
> wlan1: active.
>
> wlan1: Mesh Network.
>
> wlan0: Infrastructure network.
> eth0 : 4 lan ports
> eth1: 1 wan port
>
> br-lan : wlan0 eth0 eth1 bat0
Hmm, that looks sane and correct, should work like that.

Is there anything weird in the output of "dmesg"? Any errors or
warnings?

Cheers, Linus

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Vinay Tharigopula]

Linus , Marek...

I think the List had issues with all of us getting multiple emails for 
some reason. I wanted to update you on this status and let others know 
about my solution.

Linus, in the problem that I described after updating to batman-adv 
2011.0, I would lose connection to both the routers that were plugged 
in. Any SSH connection already established also would timeout. Therefore 
I can't get the output of batctl ping or dmesg at that exact instant. I 
did try to do a tcp dump but got nowhere because of the disconnects.

Now here is how I (kinda) solved it.

I had to enable STP on the lan-switches.  STP type: STP and Mode: Root 
Protection. This can be enabled on all ports safely without causing 
disruption to the clients plugged into the switch.

After this step was done, bridge-loop-avoidance worked properly and if 
more than 1 router in a mesh was plugged in, all the routers would pick 
one as a gateway.

Extra Bonus Step:

If more than one is plugged in, the higher MAC router will drop all lan 
packets which will result in slower throughput to clients to that router 
because it is hopping to another router via wireless and neglecting the 
fact that it's own wired link would take it faster to the intended 
destination.

My "solution" to this was to enable STP in the router too. In the UCI 
page for networking, under the br-lan, I had this option.

option 'stp' '1'.

After this setting, the 2nd router would choose the wired link as its 
preferred method of linking to the network.

The customary "warning" : I am no network expert and I am just offering 
a solution for anybody else reading this.

Thanks for all your help Linus and Marek. Please do let me know if you 
have any explanation for this :) . If you want me to send some of the 
binaries I built or my OpenWRT config files to the list, I can do so . 
Let me know.

Thanks

Vinay.

My specific setup:

Routers   : 8  x  Buffalo WZR - HP -G 300NH
Switches : 6  x 3Com 5500G-EI.



On 2/26/2011 5:49 PM, Marek Lindner wrote:
> Hi,
>
> Linus has some problems sending mails to the list, so I forward his reply.
>
> Cheers,
> Marek
>
>
> ============================================
>
>
> Hi Vinay,
>
> On Fri, Feb 25, 2011 at 08:44:33AM -0500, Vinay Tharigopula wrote:
>> Marek..
>>
>> I got the binaries from the openwrt prebuilt packages. However,
>> now
>> I rebuilt the image and built batman-adv 2011.myself. I will try
>> this now and let you know.
>>
>> However, now I am running into a much bigger issue. If two
>> routers
>> are plugged in to the same switch, bridge loop avoidance doesnt
>> work
>> and both of them are stopping responding to pings.
> Could you try "batctl ping" too and check whether this is working
> or not.
>
> If that's working but the normal pinging is not, then it should be
> a layer 3 problem - in that case we'd need the outputs of "ip
> route show" and "ip address show".
>
> If batctl ping is already not working then "batctl o" and "batctl
> if" output will be interesting.
>> Here is my config.
>>
>> bat0 if:
>> wlan1: active.
>>
>> wlan1: Mesh Network.
>>
>> wlan0: Infrastructure network.
>> eth0 : 4 lan ports
>> eth1: 1 wan port
>>
>> br-lan : wlan0 eth0 eth1 bat0
> Hmm, that looks sane and correct, should work like that.
>
> Is there anything weird in the output of "dmesg"? Any errors or
> warnings?
>
> Cheers, Linus

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Marek Lindner]

Hi,

> I think the List had issues with all of us getting multiple emails for
> some reason. I wanted to update you on this status and let others know
> about my solution.

yes, mailman was standing still due to a deadlock. Sven had to restart mailman 
before all the sent mails were processed. Seems some people were getting 
impatient ..  ;-)


> I had to enable STP on the lan-switches.  STP type: STP and Mode: Root
> Protection. This can be enabled on all ports safely without causing
> disruption to the clients plugged into the switch.
> 
> After this step was done, bridge-loop-avoidance worked properly and if
> more than 1 router in a mesh was plugged in, all the routers would pick
> one as a gateway.

I don't know much about STP (root protection does not tell my anything) but 
using STP seems like a very bad idea. STP does not know a thing about your 
mesh network and the qualities of the links. To avoid the loop it will simply 
turn off "ports" in the router which certainly is not what you want. At some 
point in time it might decide to turn off a port you are using ...

I re-read all the mails but I still can't understand your logic: You have 2 
batman-adv gateways that announce the wired clients via HNA. This happens to 
allow wireless clients connected to the mesh to communicated with your wired 
clients. No problem here. Why do you think the packets go through the mesh ?


> If more than one is plugged in, the higher MAC router will drop all lan
> packets which will result in slower throughput to clients to that router
> because it is hopping to another router via wireless and neglecting the
> fact that it's own wired link would take it faster to the intended
> destination.

Are you talking about the bridge-loop-avoidance ? What makes you think they 
are using the wireless interface to forward the traffic to the other gateway ?


> My "solution" to this was to enable STP in the router too.

Again, using STP on top of batman-adv will create more problems than it 
solves.

Regards,
Marek

Re: [B.A.T.M.A.N.] batman-adv mesh slows down wired clients (bridge-loop-avoidance)

[Marek Lindner]

Hi,

Linus has some problems sending mails to the list, so I forward his reply.

Cheers,
Marek


============================================


Hi Vinay,

On Fri, Feb 25, 2011 at 08:44:33AM -0500, Vinay Tharigopula wrote:
> Marek..
>
> I got the binaries from the openwrt prebuilt packages. However,
> now
> I rebuilt the image and built batman-adv 2011.myself. I will try
> this now and let you know.
>
> However, now I am running into a much bigger issue. If two
> routers
> are plugged in to the same switch, bridge loop avoidance doesnt
> work
> and both of them are stopping responding to pings.
Could you try "batctl ping" too and check whether this is working
or not.

If that's working but the normal pinging is not, then it should be
a layer 3 problem - in that case we'd need the outputs of "ip
route show" and "ip address show".

If batctl ping is already not working then "batctl o" and "batctl
if" output will be interesting.
>
> Here is my config.
>
> bat0 if:
> wlan1: active.
>
> wlan1: Mesh Network.
>
> wlan0: Infrastructure network.
> eth0 : 4 lan ports
> eth1: 1 wan port
>
> br-lan : wlan0 eth0 eth1 bat0
Hmm, that looks sane and correct, should work like that.

Is there anything weird in the output of "dmesg"? Any errors or
warnings?

Cheers, Linus