Hello, After upgrading to a 5.10-rc3 kernel (from a 5.4 kernel), I have found bluetooth was not working on our board. /sys/class/bluetooth/hci0 existed but the userland tools weren't able to find the HCI, presumably because it had the HCI_SETUP bit still set due to the error occuring, described immediately below. I have traced that to an error being reported by the HCI hardware with bluetooth code code of 0x1a (translated to errno -EPROTONOSUPPORT) during a call to __hci_req_sync() with function hci_init3_req() in __hci_init(). In the hci_init3_req() following commit b2cc23398e81 (Bluetooth: Enable RPA Timeout), the code checks that the locally supported command byte 35 has bit 6 set (mask of 0x40) before sending HCI_OP_LE_SET_RPA_TIMEOUT. Reverting this commit made bluetooth work again. I tried looking at the Bluetooth spec (revision date 2019-12-31), and on page 1921 it states that byte 35, bit 6 is indicating support for HCI_LE_Set_PHY. On that same byte, bit 2 (so mask 0x04, indicates support for HCI_LE_Set_Resolvable_Private_Address_Timeout, which seems to be the appropriate bit for HCI_OP_LE_SET_RPA_TIMEOUT. My HCI reports 0xf8 at byte 35 of the hdev->commands array, meaning that HCI_LE_Set_PHY is set, but not HCI_LE_Set_Resolvable_Private_Address_Timeout. Changing the mask from 0x40 to 0x04 made bluetooth work again as well. Am I correct in thinking that the commit mentioned above had the mask wrong? I can send a proper patch if my theory ends up being proven correct. Best regards, -- Nicolas Schichan
