none
Bluetooth LE driver memory DoS RRS feed

  • Question

  • The bluetooth LE driver as reported in this question :

    https://social.msdn.microsoft.com/Forums/en-US/3b9e8468-ed88-4b2a-aad8-b032ce6450ee/bluetooth-le-driver-memory-leaks?forum=wdk

    performs a denial of service on the computer if run for a large period of time. Although the reply claims this is not a leak, it hoards memory to the point where all the available system RAM ends up being held by the Bluetooth driver. For me this means if I am testing my devices I will use up all my RAM within 8 hours (4GiBytes). This means I cannot test my devices for resilience because Windows is the first to give up.

    The proposed solution to reboot or remove and reattach the hardware is impractical and avoids the issue. This is also non-paged memory being held as reported by poolmon.exe under the BthL tag, so it blocks actual/real RAM usage eventually killing the usability of the computer.

    Is there going to be a bug fix for this?

    Bluetooth Smart development on windows is already fraught* with issues without this meaning that we cannot use it as a long term platform for monitoring Bluetooth smart devices.

    I hope someone can see this is a bug and not merely the correct way to implement it on windows.

    Yours,

    Peter Myerscough-Jackopson

    *can not detect disconnect reconnects cycles where notification CCCDs will be cleared on the device, large timeouts for some bluetooth smart implementations, intermittent pairing on devices with slower advertising periods, pairing is required again if you change bluetooth dongles/interface, notification callbacks stay registered in between closing and opening the same device.

    Tuesday, October 27, 2015 5:19 PM