L2CAP MTU and packet fragmentation?

Hi,

Forgive me if this has been answered time and again, but I
can't seem to
access the archives just now. Google's not helping me much
either.

I have a bit of BlueZ code, which I have to make use of
L2CAP sockets to
communicate between devices. I've seen that I can alter the
MTU for
packets, but I'm surprised that packets don't get
fragmented as per
plain UDP when a packet larger than the MTU is passed to
send().

Have I missed a configuration option somewhere?

I could live with the MTU being 64k with no fragmentation,
but I seem to
get quite a lot of corrupt packets over around 2kB.


Cheers,
Stephen.



Using Tomcat but need to do more? Need to support web
services, security?
Get stuff done quickly with pre-integrated technology to
make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on
Apache Geronimo
http://sel.as-us.falkag.net/
sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-users mailing list
Bluez-userslists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-users

Hi Stephen,

> Forgive me if this has been answered time and again,
but I can't seem to
> access the archives just now. Google's not helping me
much either.
> 
> I have a bit of BlueZ code, which I have to make use of
L2CAP sockets to
> communicate between devices. I've seen that I can
alter the MTU for
> packets, but I'm surprised that packets don't get
fragmented as per
> plain UDP when a packet larger than the MTU is passed
to send().
> 
> Have I missed a configuration option somewhere?

the L2CAP socket is a SOCK_SEQPACKET and you have to honor
that MTU. It
is not a stream or a connection-less socket.

> I could live with the MTU being 64k with no
fragmentation, but I seem to
> get quite a lot of corrupt packets over around 2kB.

I have no idea what you are talking about. The L2CAP layer
is reliable
and the underlaying ACL link, too.

Regards

Marcel



Using Tomcat but need to do more? Need to support web
services, security?
Get stuff done quickly with pre-integrated technology to
make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on
Apache Geronimo
http://sel.as-us.falkag.net/
sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-users mailing list
Bluez-userslists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-users

Hi,

On Mon, 2006-06-26 at 14:17 +0200, Marcel Holtmann wrote:
> > I could live with the MTU being 64k with no
fragmentation, but I seem to
> > get quite a lot of corrupt packets over around
2kB.
> 
> I have no idea what you are talking about. The L2CAP
layer is reliable
> and the underlaying ACL link, too.

Sorry, correction, I get a lot of lost packets. I assume
either because
they've become corrupted, or perhaps because they never
arrive.

I get between 20-40% packet loss with packet sizes around
3000 bytes,
and more loss as packet sizes increase. l2ping says 56% loss
at 3500
bytes.

This loss is what prompted my question. This behaviour,
actually, is why
I'm using L2CAP -- because it was likened to UDP in what
I'd read
previously. Ironically, perhaps, it's just a little *too*
lossy  

I'll either bolt on some packet segmentation to my app, or
use RFCOMM
for now.


Cheers,
S.



Using Tomcat but need to do more? Need to support web
services, security?
Get stuff done quickly with pre-integrated technology to
make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on
Apache Geronimo
http://sel.as-us.falkag.net/
sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-users mailing list
Bluez-userslists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-users

Hi Stephen,

> > > I could live with the MTU being 64k with no
fragmentation, but I seem to
> > > get quite a lot of corrupt packets over
around 2kB.
> > 
> > I have no idea what you are talking about. The
L2CAP layer is reliable
> > and the underlaying ACL link, too.
> 
> Sorry, correction, I get a lot of lost packets. I
assume either because
> they've become corrupted, or perhaps because they
never arrive.
> 
> I get between 20-40% packet loss with packet sizes
around 3000 bytes,
> and more loss as packet sizes increase. l2ping says 56%
loss at 3500
> bytes.
> 
> This loss is what prompted my question. This behaviour,
actually, is why
> I'm using L2CAP -- because it was likened to UDP in
what I'd read
> previously. Ironically, perhaps, it's just a little
*too* lossy  

with l2ping you use the L2CAP signal channel and it happens
that some
remote implementation don't handle this very well. Hence
you might see a
loss percentage. This actually means unanswered echo
requests.

If you actually use L2CAP data channel with SOCK_SEQPACKET
then you
won't see a loss at all. It is a reliable channel.

> I'll either bolt on some packet segmentation to my
app, or use RFCOMM
> for now.

Or help Martin Roehricht to extend L2CAP with retransmission
and flow
control. Once this has been implemented it will be possible
to have a
SOCK_STREAM on L2CAP like you have with RFCOMM.

Regards

Marcel



Using Tomcat but need to do more? Need to support web
services, security?
Get stuff done quickly with pre-integrated technology to
make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on
Apache Geronimo
http://sel.as-us.falkag.net/
sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-users mailing list
Bluez-userslists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-users