[nos-bbs] ugly ax25 linux bug ? no idea SAMB gets 2 UA replies ...

[M Langelaar]

An update.

The user has recompiled and relinked their application, so ensuring the
netax25 calls are matching what is actually installed on the linux distro.

So far so good, it's been over an hour, and I've been hammering away at
their system off and on, and it's not happening as of this email ...

Will know better in a day or so, stay tuned.

Maiko

On 10/12/2020 7:30 p.m., M Langelaar wrote:

> This is a bit technical, but I have stumbled across a problem where if I
> connect to a node that depends on AX25 Linux stack I am receiving 2
> UA to my SABM, the first UA JNOS takes and moves it to LAPB_CONNECTED,
> but the second UA JNOS takes and tries to re-establish the link and 
> instantly
> we get a loop of SABM / UA over and over, just loops and does not 
> stop, basically
> creating a DOS attack. It's unreal to watch.
>
> My instant fix for this (and I confirm it works) is to ignore UA FRMR 
> if we are
> in a LAPB_CONNECTED state, which probably it should. I can fix that on my
> end no problem. I can't find in the AX25 specs what a system should do 
> it it
> receives a UA while in LAPB_CONNECTED, but I never see that situation 
> ever
> happen until recently.
>
> Is anyone aware of an AX25 utils or linux libs bug that might be 
> causing this ?
>
> Maiko / VE4KLM
>
>