On 02/24/2012 12:18 PM, Patrick Ohly wrote:
On Wed, 2012-02-22 at 17:24 +0100, Daniel Wagner wrote:
> On 22.02.2012 08:33, Mikel Astiz wrote:
>>> Perhaps that's why it now fails:
>>> Feb 21 16:11:52 pohly-mobl1 bluetoothd: connect(): Permission
>>> denied (13)
>>> dbus.exceptions.DBusException: org.ofono.Error.Failed: Operation failed
>>> How do check which kind of pairing is used? I re-paired via the GNOME
>>> Bluetooth UI, it automatically suggested a PIN and asked whether the
>>> device showed the same one.
>> Well, if you had to enter a pin code, then it's legacy pairing, which
>> works fine. If you just confirmed a pin code that was given, then it's
>> SSP, which is affected by a known bug in the kernel.
>> You can either (a) patch your kernel, (b) disable SSP, or (c) initiate
>> the connection from the phone. I suggest the last one since it is
>> Also note that you need a running BlueZ agent. You can use the
>> test/simple-agent script in BlueZ, which will probably ask for
>> confirmation when the phone tries to initiate the connection.
I noticed that if I have no running BlueZ agent, my version of
obex-client fails internally, but does not return an error for the D-Bus
request. In that case the PBAB backend just hangs.
$ /usr/lib/obexd/obex-client -d -s
obex-client: client/main.c:main() Entering main loop
obex-client: client/session.c:obc_session_create() driver PBAP
obex-client: client/session.c:obc_session_ref() 0x1ac3780: ref=2
obex-client: client/session.c:manager_reply() adapter path /org/bluez/15697/hci0
obex-client: manager replied with an error: org.bluez.Error.AgentNotAvailable,
Agent Not Available
obex-client: client/session.c:obc_session_unref() 0x1ac3780: ref=1
> You seem to suffer from the SSP bug. You can fall back to legacy pairing
> as Mikel indicated with:
> sudo hciconfig hci0 sspmode 0
I tried that and indeed, it fixed the "cannot browse files" problem. But
PBAP and HFP (enable-modem) still get a "permission denied".
I wonder whether the "N900: fails to synchronize with some distros" bug
) is related to it.
Something to try out...
Does anyone have a link to the SSP bug report and/or more information
when it was introduced and when it might get fixed?
There were several independent bug reports (I just found mine again, but
there were more.)
I have no access to a kernel git tree at the moment, but IIRC this bug
is present for longer then a half a year. Hopefully we get the fix into
3.3. then we can ask Greg to add it to the stable tree 3.0.