2023. július 12., szerda 0:21 keltezéssel, Barnabás Pőcze <pobrn@xxxxxxxxxxxxxx> írta: > [...] > > > > Those generally should be harmless, as the hci_conn may be destroyed > > earlier than when controller stops sending SCO packets for it. > > > > It's not clear those should be printed as errors, same also for the > > "corrupted SCO packet" messages which are also printed when the > > hci_conn has been already destroyed. You get lots of that spam during > > normal SCO operation. > > > > For this crash, maybe you can try apply the following patches: > > > > https://lore.kernel.org/linux-bluetooth/490b5c6a0e13047fd1bea42d3184b46623adc359.1689003801.git.pav@xxxxxx/ > > > > https://lore.kernel.org/linux-bluetooth/20230711131353.40500-1-code@xxxxxxxx/ > > > > Basically the ISO sockets had similar crashes before, which are now > > fixed and SCO seems to need similar fixes. > > Thanks. I will try them. Any idea as to how I could trigger the issue? I haven't > noticed any pattern yet... Seemingly `hci_send_sco()` is only hit when I set > the profile to HSP/HFP, however, every time this issue was triggered, the device > was in the A2DP profile (at least as far as I can tell). > [...] I have not seen the oops since applying those two patches on 2023-07-12, so hopefully the issue is fixed. Regards, Barnabás Pőcze
