Question about a.out (why kernel tasks are in one a.out) - Minix

This is a discussion on Question about a.out (why kernel tasks are in one a.out) - Minix ; Hi! By looking at the kernel startup stuff, we recognized, that every task and process in the boot-image table (table.c) has a different a.out, except for the four kernel tasks, which have only one a.out, meaning they share stack segment ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: Question about a.out (why kernel tasks are in one a.out)

  1. Question about a.out (why kernel tasks are in one a.out)

    Hi!

    By looking at the kernel startup stuff, we
    recognized, that every task and process in
    the boot-image table (table.c) has a
    different a.out, except for the four kernel
    tasks, which have only one a.out, meaning they
    share stack segment etc. This creates an overhead
    in management (e.g., stack guard). So why is
    this done, instead of creating a single a.out
    for each of the kernel asks?

    Bernhard

  2. Re: Question about a.out (why kernel tasks are in one a.out)

    Bernhard Kast wrote:

    : By looking at the kernel startup stuff, we
    : recognized, that every task and process in
    : the boot-image table (table.c) has a
    : different a.out, except for the four kernel
    : tasks, which have only one a.out, meaning they
    : share stack segment etc. This creates an overhead
    : in management (e.g., stack guard). So why is
    : this done, instead of creating a single a.out
    : for each of the kernel asks?

    One of the four kernel tasks, HARDWARE, is a complete dummy. IDLE is
    almost a dummy, if you look at its code you'll see it executes only a
    few instructions before it halts the CPU. That leaves two kernel tasks
    that are real processes.

    SYSTEM needs to be compiled in the same binary as the kernel because it
    provides access to kernel data structures that is needed by, but not
    available to, other processes that are not part of the kernel binary.
    Finally, CLOCK is so essential to scheduling process execution it would be
    very inefficient to make it a totally independent process.

    Note: to reply to me about Minix-related topics please use the address
    below, even if I have posted or e-mailed from another address.
    --
    +----------------------------------+
    | Albert S. Woodhull |
    | awoodhull@hampshire.edu |
    | http://minix1.hampshire.edu/asw/ |
    +----------------------------------+

+ Reply to Thread