Hello,
Who is the mainainer of ax25 in Linux ?
Has ax25 changed from 2.0.36 to 2.2.x ??
I think that I have found one error in the SDL diagrams in wich is based Linux
code, wich is causing rejects.
The conclusion is the following:
In the Timer recovery State (4), I frame MUST NOT BE TX when we receive
incoming Information frames (ore other frames wich are not Poll Responses) wich
moves the TX window. AX25 v2.0 say that "clearly" (you know, that's not very
clear, words and sentences are not so clean than a sdl diagram).
What is bad is the "I Frame pops off queue" in the SDL diagram page 98 in ax25
v2.2. This branch shouldn't be here.
I explain that:
In recovery state, you have send and Poll.
This poll was sent because of the timeout of t1.
But it is possible that if the peer has a "slow" persist/slottime, his frames
are "being sent" while your timeout happens.
When the timeout happens, you receive the frames that were "being sent", so
that moves your TX window, and you TX more frames (That is the error!!!)
The peer receive your Poll, and ack the frames that were BEFORE the last frames
you have just transmited.
You receive this poll and think that it has not received this last frames, and
"Invoke Retransmission" box is called !! You retransmit this frames !!
The peer receives duplicated set of frames ! So it send a REJ !!!
So:
1) You have transmitted frames twice
2) That takes time, it's possible that in the middle another timeout happens !!
3) Very low throughput in this situation.
4) I also detect that some frames are received OUT OF SEQUENCE, causing
Disconnects. I have not yet detected cause...
On the other hand, if the poll bit of the last I frame is activated (which I
think it must not, I have changed my sources to do that, and sincerly I'd like
it to be disactivated in the kernel), what happened was very much more howfull,
the same set of frames were TX a lot of times.
Opinions ???
--
Saludos de Juli�n
-.-
