btpand problem

[Iain Hibbert]

On Sun, 14 Oct 2012, Andreas Longwitz wrote:

> Hi, thanks for the patch.
>
> > > but, btpand should also ensure that the buffer sizes are suitable. The
> > > server_init() function already does this in FreeBSD, though I also added
> > > some code to increase the RCVBUF size in NetBSD code, and also added (but
> > > didn't yet commit) the similar for client code.. I will prepare a patch
> >
> > patch for btpand attached
>
> I tried your patch and it works, I do not need the kernel patch for the
> default size of NG_BTSOCKET_L2CAP_SENDSPACE anymore.

as I said, I think 672 is a bit low, though I used 4096 for NetBSD (with a
sysctl for runtime adjustment if necessary) and that is still potentially
prone to error with larger packets (L2CAP MTU between 48-65535 bytes is
valid) so I guess that requiring the application to set the buffer size is
best practice..

> Now I describe another problem with my ping test, I am not sure if this
> problem concerns btpand too. If the ICMP packet must be fragmented, then the
> first fragment ist lost on the tap interface (and can not be found by tcpdump
> running on this tap) if an ARP request is necessary.
>    ping -c 1 -s 1400 panserver
> works all the time,
>    ping -c 1 -s 1600 panserver
> only works if the ip address of panserver is in the arp cache.

does the ARP request get sent out?

> I did nothing special with the tap interface, it is included in my
> "cloned_interfaces" in rc.conf and opened by btpand.
>
> Any ideas ?

No, I confess.. I am not a networking expert :)

btpand does change the address on the tap, so in the first instance the
address will be unknown to the rest of the kernel I guess.. should it
generate a notification of the address change somehow?

iain