hcitool -a shows "invalid option -- a".
This is the bluetooth chip of an iPaq 3970. It's running Familiar
stable, so it must be a pretty old bluez version. Don't know if it
helps...

Ok, I knew I had to give more details... I create links between 3 iPaqs.
I tried using higher layer protocols to do so. Never worked. I'm just
trying to do a TCP/IP network so I can run XML-RPC on top of it, ssh
etc...

That's all I need. I tried to use pan, never worked, I tried to use dun
over pan, never worked, I tried to use rfcomm never worked. Always the
same problem:
I could connect A to B and B to C, never B to A and C to A. 

Now I went the hcitool route. Created my own connections very very
easily. I connected B to A and C to A (A master). Ran dund _after_
hcitool connections and everything is running great.

hcitool con on A gives me two master connections. These connections have
been initialised from B and C as slave. So I can have two master
connections, but I can't create them from A. First one works, second one
doesn't, nothing happen. It's just out of curiosity.

--
Pierre



On Fri, 2004-04-02 at 20:10, Marcel Holtmann wrote:
> Hi Pierre,
> 
> > I've been trying to setup ip net with 3 identical devices connected
> > simultanously over bluetooth. It works after a long
> > while of fiddling around. I had to go the hcitool route manually to set
> > the initial connections and then I could use dund no problem. My
> > question is: why can't I initiate a connection (hcitool cc) from a
> > device which is already connected?
> 
> for some setups this depends on your Bluetooth chip. So what does
> "hciconfig -a" show? But I am not sure about what kind of network you
> are talking, because between two device there can be only _one_ ACL link
> and this means you can only use "hcitool cc" once. Actually there is no
> need for creating the ACL link manually with hcitool, because every
> higher layer will create it on demand.
> 
> Regards
> 
> Marcel
> 
>