| Message ID | 20180822120412.20285-1-bgodavar@codeaurora.org (mailing list archive) |
|---|---|
| Headers | show |
| Series | Fix for proto races in hci_serdev. | expand |
Hi Balakrishna, > In recent testing we found that while removing hci_uart, we have seen > execution of hci_uart_write_work() after calling vendor specific > proto close. As we are freeing the vendor specific Tx and Rx buffers > in vendor close, execution of functions i.e. Rx or Tx functions may cause > a crash. > > we already have a commit for hci_ldisc.c "e508e6026b19" and "048e1bd3a27f" > to overcome the race condition. > > Changes of v1: > > * clearing flag HCI_UART_PROTO_READY while mnodule deinit such that > we will not have any tractions further on Tx or Rx. > * added check of HCI_UART_PROTO_READY while dequeuing an packet. > > > Balakrishna Godavarthi (2): > Bluetooth: hci_serdev: clear HCI_UART_PROTO_READY to avoid closing > proto races > Bluetooth: hci_serdev: Add protocol check in hci_uart_dequeue(). > > drivers/bluetooth/hci_serdev.c | 8 +++++--- > 1 file changed, 5 insertions(+), 3 deletions(-) patch has been applied to bluetooth-next tree. Regards Marcel
In recent testing we found that while removing hci_uart, we have seen execution of hci_uart_write_work() after calling vendor specific proto close. As we are freeing the vendor specific Tx and Rx buffers in vendor close, execution of functions i.e. Rx or Tx functions may cause a crash. we already have a commit for hci_ldisc.c "e508e6026b19" and "048e1bd3a27f" to overcome the race condition. Changes of v1: * clearing flag HCI_UART_PROTO_READY while mnodule deinit such that we will not have any tractions further on Tx or Rx. * added check of HCI_UART_PROTO_READY while dequeuing an packet. Balakrishna Godavarthi (2): Bluetooth: hci_serdev: clear HCI_UART_PROTO_READY to avoid closing proto races Bluetooth: hci_serdev: Add protocol check in hci_uart_dequeue(). drivers/bluetooth/hci_serdev.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)