none
Possible crash of 802.11a sample driver in Sora v1.1 RRS feed

  • General discussion

  • Hi all,

    We have identified an issue that the lastest released 802.11a sample driver may crash in a system that has less than 8 logic cores, for example Intel Core i5 (four cores with hyper-threading disabled). It is because of the hard-coded core assignment in the sample driver that may try to put a thread to a core that does not exist.

    There are 4 threads in the sample driver. The hard-coded assignment can be found as follows.
    1.       Send thread: core 2 sdr_mac_main.c line 85
    2.       Recv thread: core 4 sdr_mac_main.c line 93
    3.       Mac state machine: core 3 sdr_mac_main.c line 52
    4.       Veterbi thread: core 5 arx_control.c line line 17

    Please change the core assignment accordingly to fit in your system. The rules-of-thumb for core assignement are following:

    1. Assign each (realtime) ethread to a different core, i.e. MAC state machine and Viterbi, on different cores .
    2. Do not share an ethread with other threads.
    3. You may assign non-realtime, (in this case, snd/rcv thread), on the same core.

    Thanks,

    Sore Core Team

     


     

     

    • Changed type Kun Tan Saturday, February 12, 2011 3:49 AM
    • Changed type Kun Tan Saturday, February 12, 2011 3:50 AM no reason
    Saturday, February 12, 2011 3:47 AM

All replies

  • Hi Kun,

    I have performance problem with 802.11a on a Quad Core. (802.11b works fine)

    I changed core affinity of send and recv to core 2, and changed veterbi thread to core 4, for the driver to work.

    And now after enabling 802.11a driver, CPU goes very high (100% on core 3 and 4, 80-90% on core 1 2)

    and mouse move become very laggy.

     

    Any ideas why this is?

     

     

     

    Wednesday, March 9, 2011 6:22 PM