Bluetooth keeps stopping issue

[ildar]

Following up the phhusson#376

Current Behavior:

BT stack is crashing regularly.

To Reproduce

1. Connect BT headphones
2. Connect BLE gadget (e.g. smartwatch)
3. Start some music through A2DP
4. CRASH of the Bluetooth stack, then OS restarts it automatically

Smartphone (please complete the following information):

• Device: AGM H3
• Build Flavor: lineage-20.0-20230115-UNOFFICIAL-arm64_bvS-vndklite
• Version: latest

Additional context

To solve the problem there was the phhusson/platform_system_bt#1
But it changes nothing for my device.

[Iceows]

pra-lx1 don't support android 13, pra is a-only. Can you confirm your phone model?

[ildar]

sure. My phone is different but the behavior is quite similar. I have AGM H3, it is MT6762V-based. I'm attaching my sysinfo. [13eng.crossg.20221109.001316.json.txt](https://github.com/TrebleDroid/treble_experimentations/files/11638352/13eng.crossg.20221109.001316.json.txt)

[ildar]

BTW I trace Phh's changes and am going to test the new fixes soon. Preferably when LOS-TD by Andy Yan is out.

[phhusson]

@ildar latest fix for bluetooth crash literally has a if so it won't apply to you.
Please send a log of your crash. Just what's in adb logcat -b crashshould be enough

[ildar]

The log is attached. I am sorry, the log contains some unrelated crashes. I decided to keep them all as I see two types of BT stack crashes. Please note that I also suffer from phhusson#2413 Thanks a lot!! [adb.logcat.txt](https://github.com/TrebleDroid/treble_experimentations/files/11642146/adb.logcat.txt)

[ildar]

is there anything useful in the log?

[ildar]

Can this help me? TrebleDroid/platform_packages_modules_Bluetooth@44f470e

…

I've just installed the latest TD based LOS by Andy and it still crashes.

[pineapple-academic]

I may have similar issues with Bluetooth and wireless earphones.
The Android GSI image is ArrowOS 13.1 v2023.07.05 (latest) and the device is tablet Alldocube kPad (Unisoc T610 UMS512; system-arm64-ab). "Use System Wide BT HAL" is enabled.
Earphones connect to the device but sometimes when hitting play in Spotify there is no audio (more often stops after Play -> Pause -> Play combo). And sometimes after few minutes the audio resumes. It's very sporadic behavior.

adb logcat -b crash ->
logcat_tablet_kpad_bt_audio_crash.txt

[phhusson]

@ildar
Okay so your log has "06-02 23:25:11.675 17290 17290 F DEBUG : Abort message: 'assertion 'waiting_command_ == op_code' failed - Waiting for 0x2012 (LE_REMOVE_DEVICE_FROM_FILTER_ACCEPT_LIST), got 0x200e (LE_CREATE_CONNECTION_CANCEL)'"

So, let's assume that sending "LE_REMOVE_DEVICE_FROM_FILTER_ACCEPT_LIST" is the issue (I'm honestly not sure in this case), and let's tell Android that the chip doesn't actually support it.

Looking for LE_REMOVE_DEVICE_FROM_FILTER_ACCEPT_LIST in packages/modules/bluetooth, we end up with:

packages/modules/Bluetooth/system/gd/hci/controller.cc
811:      OP_CODE_MAPPING(LE_REMOVE_DEVICE_FROM_FILTER_ACCEPT_LIST)

Looking at OP_CODE_MAPPING, we see it is referring to local_supported_commands_
which we can see refers to persist.sys.bt.unsupported.commands
Then we see 552: LE_REMOVE_DEVICE_FROM_FILTER_ACCEPT_LIST = 271,

So to tell Android this command is not supported, you should be able to do this:
setprop persist.sys.bt.unsupported.commands 271

[ildar]

@phhusson , HUGE thank you for this detailed algorithm: it may help a lot to me or any other one willing to fix a broken BT stack. I confirm I don't see LE_REMOVE_DEVICE_FROM_FILTER_ACCEPT_LIST-related crashes anymore. But I still experience these:

05-21 20:15:44.685 21032 21032 F DEBUG : Abort message: 'assertion 'false' failed - Done waiting for debug information after HCI timeout (READ_RSSI)'

Any idea how to overcome this one?

[ildar]

@AndyCGYan , would you consider adding this patch (0007-Add-properties-to-disable-some-features-commands-sta.patch) to your LOS-light build? It seems effective and lightweight. I couldn't quickly find its source, just see it in Releases ( https://github.com/TrebleDroid/treble_experimentations/releases/download/ci-20230718/patches-for-developers.zip ) (I'm on LOS-TD now but having fp issues so want to get back to LOS-light).

[AndyCGYan]

@AndyCGYan , would you consider adding this patch (0007-Add-properties-to-disable-some-features-commands-sta.patch) to your LOS-light build? It seems effective and lightweight. I couldn't quickly find its source, just see it in Releases ( https://github.com/TrebleDroid/treble_experimentations/releases/download/ci-20230718/patches-for-developers.zip ) (I'm on LOS-TD now but having fp issues so want to get back to LOS-light).

LOS upstream merged a similar commit (albeit with less coverage), gated by a different property bluetooth.hci.disabled_commands. On Light you might wanna try this?

[ildar]

Indeed. Thanks a lot, I'll try that.

[ildar]

On Mon, Aug 21, 2023 at 1:58 PM Ildar Mulyukov wrote: But I still experience these: > 05-21 20:15:44.685 21032 21032 F DEBUG : Abort message: 'assertion 'false' failed - Done waiting for debug information after HCI timeout (READ_RSSI)' Any idea how to overcome this one?

I am really sorry for this dump question. Actually READ_RSSI is just another command that needs to be masked and you already fully explained how to handle it. So for anyone having similar situations, please use this to find the right codes: https://github.com/TrebleDroid/platform_packages_modules_bluetooth/blob/android-13.0.0_r8-td/system/gd/hci/hci_packets.pdl So in my case I need this: # setprop persist.sys.bt.unsupported.commands 271,155 Or in LineageOS GSI: # setprop bluetooth.hci.disabled_commands 271,155 Again, thank you so much, @phhusson !!

[ildar]

Though it's not 100% success story: I just got the SIGABRT again with

Abort message: 'assertion 'false' failed - Done waiting for debug

information after HCI timeout (READ_RSSI)'. That's strange because I had `disabled_commands 271,155`

[AndyCGYan]

Though it's not 100% success story: I just got the SIGABRT again with
Abort message: 'assertion 'false' failed - Done waiting for debug
information after HCI timeout (READ_RSSI)'. That's strange because I had disabled_commands 271,155

Could be because LOS's has less coverage - I'll note this and perhaps switch to PHH's commit in the next update.

[ildar]

Andy, thanks a lot. PHH's patch covers not just commands but features etc, true. But that's doesn't matter in my case, I use just commands which are supported equally. (I suspect the LOS guy simply picked the PHH's code and put it in he's).

[PowerPlop]

@ildar was this problem fixed in the end?

[ildar]

unfortunately no, #5 (comment)