Return-Path: <cyrus@holtmann.org>
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 8.1 \(1993\))
Subject: Re: GATT connection without BlueZ installed
From: Marcel Holtmann <marcel@holtmann.org>
In-Reply-To: <CAD81oNYnfe-5r8hDOqP86CTCRG8iaTt7HexzMOBSsSqcy8Z1NA@mail.gmail.com>
Date: Fri, 19 Dec 2014 11:20:25 +0100
Cc: "linux-bluetooth@vger.kernel.org" <linux-bluetooth@vger.kernel.org>
Message-Id: <8AB9692B-357D-4649-A62F-E770A78848FD@holtmann.org>
References: <CAD81oNYnfe-5r8hDOqP86CTCRG8iaTt7HexzMOBSsSqcy8Z1NA@mail.gmail.com>
To: Artem Rakhov <arakhov@chromium.org>
Sender: linux-bluetooth-owner@vger.kernel.org
List-ID: <linux-bluetooth.vger.kernel.org>

Hi Artem,

> I'm testing GATT connection: running btgatt-server on one machine and
> btgatt-client on the other one. Server machine doesn't have BlueZ
> installed, just kernel part of Bluetooth stack is present. However, I
> still use btgatt-server and btmgmt from tools/ on this machine.
> 
> Connection does not initiate correctly.
> 
> Server is stuck on listening:
> "Started listening on ATT channel. Waiting for connections"
> This means that bind() and listen() on socket are successful, but
> accept() didn't finish. It doesn't see the incoming connection.
> 
> Client initiates MTU exchange, but doesn't receive anything in response:
> btgatt-client: Opening L2CAP LE connection on ATT channel:
> src: 00:00:00:00:00:00
> dest: 24:0A:64:5A:AF:1B
> Connecting to device... Done
> [GATT client]# att: ATT op 0x02
> [GATT client]# att: < 02 17 00
> ...
> [GATT client]# att: Operation timed out: 0x02
> 
> btmon from the server:
>> HCI Event: LE Meta Event (0x3e) plen 19                                              [hci0] 82.760084
>      LE Connection Complete (0x01)
>        Status: Success (0x00)
>        Handle: 32
>        Role: Slave (0x01)
>        Peer address type: Public (0x00)
>        Peer address: 3C:77:E6:60:5C:B2 (OUI 3C-77-E6)
>        Connection interval: 67.50 msec (0x0036)
>        Connection latency: 0.00 msec (0x0000)
>        Supervision timeout: 420 msec (0x002a)
>        Master clock accuracy: 0x00
> @ Device Connected: 3C:77:E6:60:5C:B2 (1) flags 0x0000
>> ACL Data RX: Handle 32 flags 0x02 dlen 7                                             [hci0] 83.051725
>      ATT: Exchange MTU Request (0x02) len 2
>        Client RX MTU: 23
>> HCI Event: Vendor (0xff) plen 7                                                     [hci0] 115.115106
>        82 12 95 40 10 36 59                             ...@.6Y
>> HCI Event: Disconnect Complete (0x05) plen 4                                        [hci0] 115.116099
>        Status: Success (0x00)
>        Handle: 32
>        Reason: Remote User Terminated Connection (0x13)
> @ Device Disconnected: 3C:77:E6:60:5C:B2 (1) reason 3
> 
> When I change the roles (so that server runs btgatt-client, and on the
> other machine I stop bluetoothd and run btgatt-server), the connection
> works fine.
> When I use completely different software setup (BlueZ installed on
> both, but bluetoothd stopped), everything also works fine, which makes
> me believe that this is not a hardware problem.
> Kernel is 3.10 on server, 3.8 on client.
> 
> Does anyone have any clues why accept() could not succeed on the server?

I assume it is because bluetoothd does a bit more extended setup. Please keep in mind that mgmt settings are remembering during power down/up cycles. Might want to run btmgmt info and compare with how bluetoothd sets up the controller compared to your own setup.

Any chance you decode the vendor event into something meaningful. It might tell you why the connection gets disconnected. Might be a hint as well.

One other area of checking what is different would be debugfs /sys/kernel/debug/bluetooth/hci0/ since that gives you a bit of insights on some of the states of the state.

Regards

Marcel