Jean Tourrilhes <[EMAIL PROTECTED]> writes:

>       Hi,
> 
>       At 115200 kb/s or lower, connect to another Linux box and then
> do a 'cat /proc/net/irda/irlap'. Anything wrong there ?
>       I've tried to see where the bug is, but I can't find it !

The bug was in irlap_adjust_qos_settings() in qos.c ;-) I've fixed it and
will make a new patch.

One thing I should mention is that the -irda8 patch is a test where I for
the first time follow the IrDA spec for limiting the sending window. That
is the reason why the Tx window is only 2 (you cannot send more than 2 full
2K frames at 115200 bps with 500 ms turn time).

Previously I kept the window full sized, and calculated how many bytes I
could send in the window. So for Linux it was possible to send 2 full sized
frames, or 7 small frames (or something in between). I still think the
Linux way is better, so I'll probably go back to this way of doing it in
the next patch. Then we only need to check for a data size larger than the
maximum line capacity since the window is changed dynamically based on how
many and large frames that have already been send in the current
window. For more info check out section 6.6.11 (page 44) in
IrLAP. Comments?

>       Of course, the performance is abysmal, IrLAN doesn't work at
> 9600 and STREAM_SEQPACKET crash the machine at every try...

Hmm, I'm still not sure if everything is OK now, so I'll try out some low
speed connections later with SAR later.

-- Dag 

-- 
   / Dag Brattli                   | The Linux-IrDA Project               /
  // University of Tromsoe, Norway | Infrared communication for Linux    //
 /// http://www.cs.uit.no/~dagb    | http://www.cs.uit.no/linux-irda/   ///

_______________________________________________
Linux-IrDA mailing list  -  [EMAIL PROTECTED]
http://www4.pasta.cs.UiT.No/mailman/listinfo/linux-irda

Reply via email to