Hi Luiz, Good news... On 02/03/2022 06:55, Luiz Augusto von Dentz wrote: <snip> >> >> Looks like I was able to reproduce but I still don't know the cause, >> anyway the symptom is the following: >> >> [23412.856410] remove_uuid:2273: hci0: sock 0000000049dcd70a >> [23412.856467] mgmt_class_complete:2174: hci0: err 0 >> [23412.856470] mgmt_cmd_complete:176: sock 00000000d63e046a >> >> It looks like the cmd->sk is wrong/corrupted, what is even more >> strange is that the socket pointer does seem to match previously >> connected MGMT clients and after a few attempts with btmgmt> power on >> it does come back to life, very bizarre... >> >> Can you try to enable some kernel debugs before you start bluetoothd: >> >> echo "file net/bluetooth/mgmt.c +pfl" > /sys/kernel/debug/dynamic_debug/control >> echo "file net/bluetooth/mgmt_util.c +pfl" > >> /sys/kernel/debug/dynamic_debug/control >> >> Maybe that should give us a clue what triggers it. > > Here is an attempt to fix the problem: > > https://patchwork.kernel.org/project/bluetooth/patch/20220302064351.225212-1-luiz.dentz@xxxxxxxxx/ > > Also it probably make more sense to test with the following tree since > we are using it to push regression fixes: > > https://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth.git/ > I've cloned bluetooth.git, applied the patch and installed the resultant kernel. I've booted into the kernel eight times and on each occasion , my bluetooth devices connected successfully. Additionally, I've the latest updates from Linus' tree into my local clone and applied your patch to that. That too has resulted in a working bluetooth service. Thanks and well done to you and your colleagues. For the patch above: Tested-by: Chris Clayton <chris2553@xxxxxxxxxxxxxx> >>>>>>> >>>>>>> -- >>>>>>> Luiz Augusto von Dentz >>>>>> >>>>>> >>>>>> >>>> >>>> >>>> >>>> -- >>>> Luiz Augusto von Dentz >>> >>> >>> >>> -- >>> Luiz Augusto von Dentz >> >> >> >> -- >> Luiz Augusto von Dentz > > >
