Segfault on Release audio connection (port/linux-port)

[simo-scar-cob]

Describe the bug

Using linux-port, when running/debugging the hfp_ag_demo file, a segfault happens when the audio connection is established, and I try to close it, both with stdin process input command or CTRL+C.

To Reproduce

Steps to reproduce the behavior:

1. Run example hfp_ag_demo using linux-port
2. Connect to the headset
3. Establish ACL connection
4. Establish audio connection
5. Close it: default is 'B', or close the program with CTRL+C

Expected behavior

Audio connection should close without any errors

Additional context
Running on linux embedded custom platform

[mringwal]

Do you have a backtrace for this? A packet log might be helpful, too.
Does the same happen with modplayer demo on exit (it's playing music)?

[simo-scar-cob]

At the moment, I can't get a proper backtrace.
I have a packet log here.
hci_dump_hfp_ag_demo.zip

No, it doesn't happen on modplayer (a2dp source demo tested)

[mringwal]

Tricky. I'm currently don't have a laptop with a Linux vm, which would allow me to reproduce this. I'll get a new laptop in 1-2 weeks on which I'll setup a new Ubuntu VM. I hope this is not urgent and can wait a bit.

Until then: could you send me a packet log for the case where you just disable SCO audio with 'B'?
The segfault is irritating as the only active data source is the one to the kernel HCI socket, which should never get invalid.

[simo-scar-cob]

hci_dump_hfp_ag_demo.zip

Of course, in this log dump, I disabled SCO audio with 'B'.
Also, I noticed that after some retries of the code execution (and a segfault triggered), some files related to ALSA get corrupted (atm, I don't have any trace, simply because I don't want to break my system... if it happens, I'll post it here for sure), and opening the SCO audio connection is no longer possible. The only solution I found to solve the issue is to reset the filesystem entirely, which, of course, is not good.
Hope this helps!

[mringwal]

What do you mean by "files ... get corrupted"? BTstack has code to write: the packet log, persistent storage (by default into /tmp) and in this demo, a .wav file (into the current directory)

Thanks for the log. Nothing strange in there, just an SCO disconnect. I have currently no idea what's going on the there. I'm optimistic that this is easy to debug & fix with a debugger and hope that I can replicate it in a regular Ubuntu VM when I get the new laptop.

[simo-scar-cob]

I know that BTstack has code to write those files and nothing else. I'm just wondering if there's some unexpected behaviour, caused by the segfault, that causes the corruption of something on the fs. Anyway, let's just focus on the segfault. If I can replicate the issue or encounter any other problem, I'll let you know for sure.
Yes, I also think that should be easy to solve :)