Zephyr BTLE in infinite waiting loop

[AGeegee]

When a BTLE Connection is underway, calling bt_disable can result in an infinite hang in the Zephyr BTLE stack. Using ncs v3.0.0

Any ideas? There is some mention of: CONFIG_BT_HCI_ACL_FLOW_CONTROL causing a deadlock, not sure if that is it and also github.com/.../6464ffa3f94eb9d9203b56518cf53b141b1feea5

But the WDT (which is on a low priority thread) doesn't refresh the WDT, and I find that the WDT is not timing out to cause a reset. Thus any ideas why the WDT is not causing a reset as well? 

The WDT is only kicked in one place, and this deadlock doesn't cause a reset.


Thanks

sym_WVJLOWUIPBC2HDDMNEWOD4POANUBUKS6OYFXDSA 0x000000000002416a sym_DQFOGVCNU4NTOMWSL3IYDNCMIXT2LT7SZHL5IHI 0x000000000002373c sym_3LXCSEW4O4NZFGMXLMWXSJJVGEQV4VDRFCVRV4Q 0x0000000000017d3c sym_ONGROAY43I27VAHOOPEU4ALGBDKXPPQVPQ7CTTI 0x0000000000027d94 sym_PX6SXPYAZJ5TSPSKFFBO6FPMKZLYVT5KZ3EHMXA 0x0000000000027d6e sym_4W26TPGCKOCVVRKXQDBKGXGC2Y5LMZXNKS2QITA 0x00000000000192e4 sym_QA2HTUWXDQAGJZ32BSTQMVSMKKNETZBUHL37LYA 0x0000000000026fee sdc_hci_cmd_cb_reset 0x000000000001585e controller_and_baseband_cmd_put hci_internal.c:886 cmd_put hci_internal.c:1748 hci_internal_cmd_put hci_internal.c:1810 cmd_handle hci_driver.c:358 hci_driver_send hci_driver.c:432 bt_hci_send bluetooth.h:168 bt_send hci_core.c:4040 hci_core_send_cmd hci_core.c:3050 process_pending_cmd hci_core.c:4719 tx_processor hci_core.c:4730 work_queue_main work.c:684 z_thread_entry thread_entry.c:48 <unknown> 0x00000000aaaaaaaa

[Turbo J]

Some of the deadlocks won't keep any thread in running state, i.e. all threads are in one of the waiting functions similar to k_sem_take(). 

Thus its highly likely that your WDT thread was still able to run even if its at lowest priority. How do you know it did not refresh WDT?

[AGeegee]

I put a breakpoint in the WDT Kick area. Though it could be that most of the device is 'sleeping' / WFI as such a 5s WDT timeout might take 100's of seconds before it times out.

[Simon D-M]

Hi,

The link you sent is for a commit that, I believe, was merged in the NCS v3.1.0. So to see if it fixes your issue you can either try to patch your current SDK or try to migrate to a later version.

Also, you can check the "known Issues" page. It looks like NCSDK-31528 or NCSDK-30959 can be relevant. Did you already try the proposed workarounds ?

If the workarounds don't work, can you please share a bit more (code snippet, debug logs, ...)

Best regards,

Simon D-M