[Bluez-devel] hciattach killed

[Bluez-devel] hciattach killed

[Kutschenreuter Matthias]

[-- Attachment #1: Type: text/plain, Size: 1057 bytes --]

Hi,
 
I run Bluez (PAN) on an arm processor device with a kernel with all bluez
drivers built-in.
I can perfectly PAN (ping, ftp, etc.) with other devices.
 
The problem is now, when I stop any program  (l2ping, ping or higher
application e.g. upnp)
with <Ctrl-C>,  the BT chip is always detached :
"hciconfig" shows only 00:00:00:00:00:00 MAC-adress
I have to hciattach again.
If I stop programs regular, e.g. ftp with "quit", then it is NOT detached,
only
Crtl-C causes the problem.
(I also observed similar "automatic detach" when devices (e.g. a Zaurus) go
into power save 
mode, but I think this is due to a voltage drop at the BT chip, which Ctrl-C
shouldnt
cause.)
 
I didnt experience this on arm processor devices, where the bluez modules
(l2cap,
hci_uart, bnep) havent been built in, so my only idea at the moment is, that
the
problem is due to the fact, that all is built-in. Can it be, that killing a
program 
with Ctrl-C kills also a socket / port / whatever  used by the
build-in-hciattach. 
 
Do you have any idea?
 
Kind regards, Matthias
 

[-- Attachment #2: Type: text/html, Size: 4073 bytes --]

Re: [Bluez-devel] hciattach killed

[Marcel Holtmann]

Hi Matthias,

> I run Bluez (PAN) on an arm processor device with a kernel with all
> bluez drivers built-in.
> I can perfectly PAN (ping, ftp, etc.) with other devices.
>  
> The problem is now, when I stop any program  (l2ping, ping or higher
> application e.g. upnp)
> with <Ctrl-C>,  the BT chip is always detached :
> "hciconfig" shows only 00:00:00:00:00:00 MAC-adress
> I have to hciattach again.
> If I stop programs regular, e.g. ftp with "quit", then it is NOT
> detached, only
> Crtl-C causes the problem.
> (I also observed similar "automatic detach" when devices (e.g. a
> Zaurus) go into power save 
> mode, but I think this is due to a voltage drop at the BT chip, which
> Ctrl-C shouldnt
> cause.)
>  
> I didnt experience this on arm processor devices, where the bluez
> modules (l2cap,
> hci_uart, bnep) havent been built in, so my only idea at the moment
> is, that the
> problem is due to the fact, that all is built-in. Can it be, that
> killing a program 
> with Ctrl-C kills also a socket / port / whatever  used by the
> build-in-hciattach. 

the strange thing is that if hciattach is killed, the Bluetooth device
must disappear and hciconfig couldn't show it anymore. Check in detail
what is going on there. And btw the hciattach uses the TTY ldisc of the
hci_uart driver to attach the serial port.

Regards

Marcel




-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] hciattach killed

[Marcel Holtmann]

Hi Matthias,

> sorry I mixed it with another problem we had: hciconfig shows 
> NOTHING, not even 00:00:00:00:00:00.
> so I think, hciattach IS killed.
> 
> How can I check what is going on there? 

I would start with a binary hcidump of the connection that causes the
problem and then add some debugging printk's to the hci_uart driver. And
add debug code to hciattach, so you can be sure that the ldisc is closed
by the kernel module and that there is not segfault in the hciattach
itself. If you think this is a module versus built-in problem check it
with an identical setup, but only different kernel's.

And btw what hciattach command line and what module do you use?

Regards

Marcel




-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] hciattach killed

[Marcel Holtmann]

Hi Matthias,

> we use the latest available bluez-modules on a 2.4.21
> kernel with the latest available patch.

my fault. I meant what kind of Bluetooth module do you use?

> OK, we did some changes in the hciattach, because we do 
> weired things like attaching a BT Chip that has no EPROM.
> so we have to tell the device its MAC address with 
> "hciattach ttyS0 <special> 115200 <MAC>"
> 
> But we think, the Ctrl-C problem isnt due to our changes, 
> we only adapted the initialisation of the chip, like demanded
> by the chip. 

Show us the code and we will see what you have done. If you really
changed so much then go back and try it with a standard Bluetooth module
and see if it also shows the same problems.

Regards

Marcel




-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel