Linux makes the impossible, possible - Linux

This is a discussion on Linux makes the impossible, possible - Linux ; "...sometimes the system freezes even though I didn't even touch the mouse or the keyboard." http://www.murga-linux.com/puppy/vie...323145&t=27550...

+ Reply to Thread
Results 1 to 2 of 2

Thread: Linux makes the impossible, possible

  1. Linux makes the impossible, possible

    "...sometimes the system freezes even though I didn't even touch the mouse
    or the keyboard."
    http://www.murga-linux.com/puppy/vie...323145&t=27550



  2. Re: Linux makes the impossible, possible

    On Oct 23, 8:37 pm, "DFS" wrote:
    > "...sometimes the system freezes even though I didn't even touch the mouse
    > or the keyboard."http://www.murga-linux.com/puppy/viewtopic.php?search_id=1213323145&t...


    That is correct; Linux (or a module therein) will freeze if
    certain conditions are met. These conditions may be very
    hard to identify, but one simplistic one for example is
    a wayward interrupt from a card that Linux misidentified
    or misconfigured. If the code for that card is written
    in a certain fashion one might get a call to panic()
    and the entire system basically shuts up and goes south.

    (One very ancient example: on PCs prior to about the 386
    era I would routinely see, when DOS was running, a quick
    "Divide by zero" fault on the screen when I shut the
    PC off. Clearly something wasn't programmed to handle
    a power fault -- assuming a power fault indication was
    available at all -- and the OS misinterpreted it as a
    divide by zero error, printing it out just before the
    machine ceased operations.)

    It is *not* (purely) a hardware error; it is a combination
    of hardware and software (hardware by itself doesn't do
    much, after all). As someone once said, in QA:

    - Test the means.
    - Test the extremes.
    - Test the exceptions.

    In order to properly test these, one should implement
    them. Clearly, there's a driver making a bad assumption,
    causing a system freeze. Faulty hardware might test that
    assumption, or it might not. Identifying which driver
    is rather important; regrettably, the OP merely gives us
    a list of candidates. (Another poster shows his naivete:
    trying to execute the command '!/bin/sh' doesn't work
    very well. ;-) )

    At least with Linux, one might have a fighting chance at
    compiling a debug version of the module and attempting to
    figure out why the machine is freezing up. Best idea I
    have in that area is to set up another machine with remote
    logging capabilities, and send messages to that machine
    from the kernel. When the first machine freezes up, one
    might have a record on the second machine as to what Linux
    was "thinking" at the time.

    The rest of the thread speculates that the problem might
    be caused by a bad BIOS setting on a fan unit (though
    they don't identify such that specifically; the OP merely
    states that he "manually set the fans to higher RPMs"
    somehow), or more simply because the unit overheats.
    Anyone familiar with "thermal runaway" in transistor units
    gets very worried at this point, and this could happen with
    most systems if the fans get clogged with dust, of course.

    Even magical mystical Microsoft Windows Vista can't quite fix
    that problem.

    On a somewhat related note: my nx9010 has no thermal
    problems as such, but it does have some interesting
    varispeed fan capability; I do not believe this capability
    is directly controlled by Linux, though with acpi I
    can monitor the temperature of the CPU readily enough.
    Briefly, the fans have 4 speeds (though it's possible they
    simply switch on and off instead, with all three being on
    for "high speed" or something):

    - off, for nice quiet operation -- for a short time
    - low, when the unit is relatively cool and generally idle
    - medium, when I'm compiling something or playing a game,
    which tends to heat up things
    - high, when I'm compiling something rather intensively and
    the CPU goes over 65 degrees Celsius (it can tolerate
    73 from what docs I've been able to find, so there's a
    certain margin here, fortunately).

    I suspect other laptops have similar capabilities.

    Note that even when one isn't touching things, the CPU is
    executing occasional cycles; after all, clocks and crontab
    need to run, and one might get occasional interrupts from
    other devices such as NIC packets from one's upstream
    router. Get the wrong combination of packets and
    a slightly buggy driver, and things can easily go south.

+ Reply to Thread