Curl usage (http GET command) and memory issue

Greetings,

When trying to perform an http GET command over LTE-M (using the "Curl" MoSh command, to trigger a http download of a 500Ko or 1Mo file) after a few successfull attempts, the following error message appears on the CLI (and the "Curl" command no longer works successfully):

curl: (27) Out of memory

I suspect this is just a memory saturation issue. How can itbe solved? Is there a command available to free some memory for the subsequent Curl attempt?

Best regards,

fdup78

  • Hi,

    Are you using the default MoHs configs?

    What modem firmware are you using?

    What version of the SDK are you using?

    Are you using other curl hooks?

    Regards,
    Jonathan

  • GReetings,

    See below the requested information:

    1) MoSh used:

    MOSH version: v1.8.0
    MOSH build id: custom
    MOSH build variant: dev

    2) Modem FW:

    mosh:~$ at at+cgmr
    at at+cgmr
    mfw_nrf9160_1.3.1
    OK

    3) DK version:

    mosh:~$ at at%hwversion

    at at%hwversion
    %HWVERSION: nRF9160 SICA B1A
    OK

    4) Other CUrl hooks?

    Not sure I understand the question... 

    See below a typical example of the pb (with a 500Ko file) :

    _mosh:~$  curl http://<IP@_zapped>/j_500Ko.txt  -v

    curl http://<IP@_zapped>/j_500Ko.txt  -v

    note: custom traces "AT%XMODEMTRACE=1,5" was set for testing

    note: use --curr-mdm-traces hook for the current ones

     

    curl: (27) Out of memory

    note: default traces "AT%XMODEMTRACE=1,2" was set

    [2022-06-09 15:42:07.746]

    DONE

    Hope this helps.

  • Hi,

    The issue seems to be due to some factor that limits the RAM for CUrL, so something in our application is limiting the RAM. Because seems that the "custom traces "AT%XMODEMTRACE=1,5" was set for testing" that means that there is custom config in use, i.e.

    CONFIG_NRF_MODEM_LIB_TRACE_ENABLED is in use and that results less RAM for curl usage.

    If you can share you config things would make more sense.... but the issue is that there is something lese limiting the CUrL RAM usage. 

    Regards,
    Jonathan
Related