ISO socket bugs/lack of support

* ISO socket bugs/lack of support
@ 2023-03-03 14:52 Iulia Tanasescu
  2023-03-03 17:48 ` Luiz Augusto von Dentz
  0 siblings, 1 reply; 3+ messages in thread
From: Iulia Tanasescu @ 2023-03-03 14:52 UTC (permalink / raw)
  To: linux-bluetooth

Hello,

Earlier this week I submitted a BlueZ patch ([1]) containing an initial implementation of the Broadcast Audio Scan Service.

In order to perform some of the BASS procedures, the BASS Server is required to send specific HCI commands and to handle HCI events accordingly - for example, the "Set Broadcast_Code" operation of the Broadcast Audio Scan Control Point characteristic requires the BASS Server to issue the LE BIG Create Sync command with the Broadcast_Code parameter set to the characteristic value written by the BASS Client.

My approach was to open an HCI socket and to send the command directly from the BASS handler function. After my proposed implementation was reviewed, I was told I should use an ISO socket instead and let the Bluetooth kernel perform the BIG synchronization procedure. I investigated this solution and the problem that I encountered was that the kernel sends HCI commands with 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 parameter 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 send the LE Periodic Advertising Create Sync command with the Sync_Timeout parameter 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 Controller 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 the problem.

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) plen 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.

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 - is there a way to set the parameters from BlueZ? Or should this support be implemented in the Bluetooth kernel?

Thank you in advance

Iulia

[1]	[BlueZ,0/2] Add initial BASS support - Patchwork (kernel.org)

^ permalink raw reply	[flat|nested] 3+ messages in thread