I currently have the following setup

DSL-300---AP~~~~~AP---Linux-router

and I am having terrible lag / throughput problems when running PPPoE over this link.

Is there any way to check if the AP bridge is dropping packets? Both APs are currently running 2.39 and are in Point to Point bridge mode. The odd thing about this is that when I had the APs acting as a bridge between the Linux-router/file-server and a client PC, I didn't have any lag issues. However that was straight TCP/IP and UDP/IP traffic rather it being tunneled over PPPoE.

Any hints? Is it worth trying an earlier firmware release?

Don't worry guys, I found the problem and it wasn't the APs.

When going over everything again I found that mldonkey was running on my router. After stopping it everything came good. I had forgotten that I had previously stopped the daemon but hadn't uninstalled it. Powering the router down while moving it of course meant it started backup when the router did.

mea culpa on this one.

Well after running the setup for a couple of days I am still getting packet loss and unable to run my 512/128k DSL connection at full speed.

Any suggestions?

none at the moment.

I did you setup for a little while when testing PPPoE over Point to Point mode.

I was running 2.39 same as you and have not see this issue.

You sure there is nothing else running which could be causing problems?

Screwball

positive, I shutdown all other software and routing for my client PCs.

I then ran a wget to a large fine on my isp's mirror server it kept pausing and dropping back in speed before coming back up again. The same thing with out the bridges gives me full line speed.

I also ran iptraf to enusre nothing else was running or trying to DOS me etc.

I will give 2.34 a spin later today and see if it makes a difference.

Well emperical testing so far have found the following.

• 2.34 doesn't make any difference
• Short or long preambles makes no difference
• Clamping the pppoe MSS seems to help

I've reduced my pppoe TCP MSS (maximum Segment Size) from 1452 to 1412. I don't know if it has completely fixed the apparent packet loss issue but it has definatly helped. I'll do some more testing and report back.

If you have any suggestions of how to test this idea further or to workout the optimum MSS I would appreicate it.

Is there any way to lock the auto sense on the MNWAPB to either crossover or straight through mode?

Both the DSL-300 and the MNWAPB can do this and I have noticed that the ethernet link light will dropout briefly every 20-30 secs. It happens with both crossover and straight through cables.

While I feel like I'm clutching at straws, I want to eliminate as many variables as possible.

No I belive that this has been requested however that feature is not available as of yet.

Hopefully we will get an answer soon on whether there is another firmware release or not for this unit. We have raised our concerns to the OEM, and the list is large.

Hopefully if they take notice we should have a few more feature, but there are no promises which I wish I could give.

The part about the MTU is an interesting one you mentioned also. Did you set this in your ethernet cards or on the ADSL modem?

Another option might be to slow down the ethernet card you have in your pc so that they are only half duplex, this may also work and speeds should not be too affected.

Screwball

Neither. I can set MTU/MRU limits within ppp and MSS within pppoe.

I did this to ensure the PPPoE encapsulation of the IP packets was not pushing it over this size of the the ethernet frame + what ever tagging the 802.11b bridge may happen to do.

It is a shame they don't just setup a telnetd server, it would make life so much easier.

Edit: I have managed to find a spare 10mb hub at work, I'll run some tests to tonight to see if using it to force both the dsl-300 and the MAWAPB to 10mb half/duplex will fix the link dropout.

From past testing, I found that the minitars had significant issues with dealing with MTUs larger than 512 bytes. I was testing with a minitar in adhoc station mode ad the time. results of pinging the management IP of the AP went along the lines of

512 bytes - 0% packetloss
700 bytes - 5-10% p/s
1000 bytes - 30-40% pl
1300 bytes - 40-60% p/l

I dont know why, or how. Thats how it was. Running in it AP mode did not give this problem

Thanks for that information.

Unfortunately that means the APs are useless for my intended mode of operation. PPPoE has a minimum MSS size of 536 bytes, plus you have the ppp encapsulation on top of that.

Does either 2.34 or 2.39 support PPPoE in the client <-> AP mode of operation?

Also so forcing the AP and the DSL-300 down to 10mb half-duplex by using a 10mb hub has not made any difference in performance of the link