tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: btuart API change - rfc

On Sun, 6 Apr 2008, KIYOHARA Takashi wrote:

> Moreover, bcsp(4) and csr of btuart(4) use BCCMD (BlueCore Command) for
> the speed initialization.  Is this treated in the userland?

for those unaware, BCSP is a different encapsulation of bluetooth packets
on a serial line, designed by CSR to handle errors better (the btuart
'protocol' has no error handling, it will just crash and burn)

I've not looked in detail at bcsp.c but currently it does not require the
setup ioctls, nor changes the line speed when it is activated, and these
are the complexity I'm trying to reduce before we must support them.

> It just like the BlueZ of Linux.

the BCSP protocol does not say what should happen when the link is finally
considered failed (9.3.3 says "notify higher level code"), and there is a
separate document for link establishment which is handled in the userland
part of BlueZ/Linux. Does the kernel part handle that also?

I think that bcsp(4) so far just restarts the link establishment process
on link-failure but I'm not so sure thats a good idea, what happens when
it cannot establish the link, will it just keep trying?

It could be better that the line disciplines (btuart and bcsp) should just
halt after detecting a link failure, and signal the controlling process if
possible (HUP?). Let the userland tool decide if it wants to restart or
notify the operator..


Home | Main Index | Thread Index | Old Index