MIME-Version: 1.0
References: <AS8PR04MB8898256F1388824FC19F67BFECB39@AS8PR04MB8898.eurprd04.prod.outlook.com>
In-Reply-To: <AS8PR04MB8898256F1388824FC19F67BFECB39@AS8PR04MB8898.eurprd04.prod.outlook.com>
From:   Luiz Augusto von Dentz <luiz.dentz@gmail.com>
Date:   Fri, 3 Mar 2023 09:48:43 -0800
Message-ID: <CABBYNZJpzvDWf89DYGBfErc=Y6-euzUCdjUydJqSq+5NDExOvA@mail.gmail.com>
Subject: Re: ISO socket bugs/lack of support
To:     Iulia Tanasescu <iulia.tanasescu@nxp.com>
Cc:     "linux-bluetooth@vger.kernel.org" <linux-bluetooth@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Hi Iulia,

On Fri, Mar 3, 2023 at 7:02=E2=80=AFAM Iulia Tanasescu <iulia.tanasescu@nxp=
.com> wrote:
>
> Hello,
>
> Earlier this week I submitted a BlueZ patch ([1]) containing an initial i=
mplementation of the Broadcast Audio Scan Service.
>
> In order to perform some of the BASS procedures, the BASS Server is requi=
red to send specific HCI commands and to handle HCI events accordingly - fo=
r example, the "Set Broadcast_Code" operation of the Broadcast Audio Scan C=
ontrol Point characteristic requires the BASS Server to issue the LE BIG Cr=
eate Sync command with the Broadcast_Code parameter set to the characterist=
ic value written by the BASS Client.
>
> My approach was to open an HCI socket and to send the command directly fr=
om the BASS handler function. After my proposed implementation was reviewed=
, I was told I should use an ISO socket instead and let the Bluetooth kerne=
l perform the BIG synchronization procedure. I investigated this solution a=
nd the problem that I encountered was that the kernel sends HCI commands wi=
th fixed parameters - calling the "listen" API on an ISO socket will always=
 attempt to send the LE BIG Create Sync command with the Broadcast_Code par=
ameter set to 0. I couldn't find a way to provide my desired parameters to =
the HCI command.
>
> I also discovered 2 kernel issues when calling "listen" on an ISO socket:
>
> The first issue is that the "hci_pa_create_sync" function attempts to sen=
d the LE Periodic Advertising Create Sync command with the Sync_Timeout par=
ameter set to 0 - according to the Bluetooth Core specification version 5.3=
, this parameter should be in the 0x000A - 0x4000 range. This is why the Co=
ntroller will reject this command, as shown in the following btmon capture:
>
>
> < HCI Command: LE Periodic Advertising Create Sync (0x08|0x0044) plen 14
>         Options: 0x0000
>         Use advertising SID, Advertiser Address Type and address
>         Reporting initially enabled
>         SID: 0x01
>         Adv address type: Public (0x00)
>         Adv address: C0:07:E8:8B:69:F2 (OUI C0-07-E8)
>         Skip: 0x0000
>         Sync timeout: 0 msec (0x0000)
>         Sync CTE type: 0x0000
>
> > HCI Event: Command Status (0x0f) plen 4
>       LE Periodic Advertising Create Sync (0x08|0x0044) ncmd 1
>         Status: Invalid HCI Command Parameters (0x12)
>
> Updating the "hci_pa_create_sync" function and setting the "sync_timeout"=
 field of the command parameters to a value in the required range solves th=
e problem.

Broadcast support is in its very early stages of development, and yes
the sync_timeout probably need fixing as it using invalid range right
now, we should probably a way to configure it via use of BT_ISO_QOS,
currently both unicast and broadcast are using the same structure but
perhaps it would be a good idea to differentiate them with more
dedicated options.

> The same issue appears when the kernel attempts to send the LE BIG Create=
 Sync command with the BIG_Sync_Timeout parameter set to 0:
>
>
> < HCI Command: LE Broadcast Isochronous Group Create Sync (0x08|0x006b) p=
len 25
>         BIG Handle: 0x00
>         BIG Sync Handle: 0x0000
>         Encryption: Encrypted (0x00)
>         Broadcast Code: 00000000000000000000000000000000
>         Maximum Number Subevents: 0x00
>         Timeout: 0 ms (0x0000)
>         Number of BIS: 1
>         BIS ID: 0x01
>
> > HCI Event: Command Status (0x0f) plen 4
>       LE Broadcast Isochronous Group Create Sync (0x08|0x006b) ncmd 1
>         Status: Parameter Out Of Manadatory Range (0x30)
>
> Updating the "hci_le_big_create_sync" function and setting the "timeout" =
field of the command parameters to a valid value solves the issue.

Patches are welcome to fix this as well.

> Hopefully someone can confirm the kernel issues I described and the fixes=
 can be committed. I would also like to ask for some guidance regarding the=
 ISO socket usage when it comes to personalizing HCI command parameters - i=
s there a way to set the parameters from BlueZ? Or should this support be i=
mplemented in the Bluetooth kernel?

The setup of Qos is done via BT_ISO_QOS:

https://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth-next.gi=
t/tree/include/net/bluetooth/bluetooth.h#n174

As you can see currently the structure is targeting unicast
parameters, we should probably split this and have a dedicated one for
broadcast.

> Thank you in advance
>
> Iulia
>
> [1]     [BlueZ,0/2] Add initial BASS support - Patchwork (kernel.org)
>


--=20
Luiz Augusto von Dentz