Subversion Repositories HelenOS-doc

<?xml version="1.0" encoding="UTF-8"?>
<chapter id="scheduling">
  <?dbhtml filename="scheduling.html"?>

  <title>Scheduling</title>

  <para>One of the key aims of the operating system is to create and support
  the impression that several activities are executing contemporarily. This is
  true for both uniprocessor as well as multiprocessor systems. In the case of
  multiprocessor systems, the activities are trully happening in parallel. The
  scheduler helps to materialize this impression by planning threads on as
  many processors as possible and, where this means reaches its limits, by
  quickly switching among threads executing on a single processor.</para>

  <section>
    <title>Contexts</title>

    <para>The term context refers to the set of processor resources that
    define the current state of the computation or the environment and the
    kernel understands it in several more or less narrow sences:</para>

    <itemizedlist>
      <listitem>
        <para>synchronous register context,</para>
      </listitem>

      <listitem>
        <para>asynchronous register context,</para>
      </listitem>

      <listitem>
        <para>FPU context and</para>
      </listitem>

      <listitem>
        <para>memory management context.</para>
      </listitem>
    </itemizedlist>

    <para>The most narrow sence refers to the the synchronous register
    context. It includes all the preserved registers as defined by the
    architecture. To highlight some, the program counter and stack pointer
    take part in the synchronous register context. These are the registers
    that must be preserved across a procedure call and during synchronous
    context switches.</para>

    <para>The next type of the context understood by the kernel is the
    asynchronous register context. On an interrupt, the interrupted execution
    flow's state must be guaranteed to be eventually completely restored.
    Therefore the interrupt context includes, among other things, the scratch
    registers as defined by the architecture. As a special optimization and if
    certain conditions are met, it need not include the architecture's
    preserved registers. The condition mentioned in the previous sentence is
    that the low-level assembly language interrupt routines don't modify the
    preserved registers. The handlers usually call a higher-level C routine.
    The preserved registers are then saved on the stack by the compiler
    generated code of the higher-level function. In HelenOS, several
    architectures can be compiled with this optimization.</para>

    <para>Although the kernel does not do any floating point
    arithmetics<footnote>
        <para>Some architectures (e.g. ia64) inevitably use a fixed set of
        floating point registers to carry out its normal operations.</para>
      </footnote>, it must protect FPU context of userspace threads against
    destruction by other threads. Moreover, only a fraction of userspace
    programs use the floating point unit. HelenOS contains a generic framework
    for switching FPU context only when the switch is forced.</para>

    <para>The last member of the context family is the memory management
    context. It includes memory management registers that identify address
    spaces on hardware level (i.e. ASIDs and page tables pointers).</para>

    <section>
      <title>Synchronous context switches</title>

      <para>The scheduler, but also other pieces of the kernel, make heavy use
      of synchronous context switches, because it is a natural vehicle not
      only for changes in control flow, but also for switching between two
      kernel stacks. Two functions figure in a synchronous context switch
      implementation: <code>context_save</code> and
      <code>context_restore</code>. Note that these two functions break the
      natural perception of the linear C code execution flow starting at
      function's entry point and ending on one of the function's exit
      points.</para>

      <para>When the <code>context_save</code> function is called, the
      synchronous context is saved in a memory structure passed to it. After
      executing <code>context_save</code>, the caller is returned 1 as a
      return value. The execution of instructions continues as normally until
      <code>context_restore</code> is called. For the caller, it seems like
      the call never returns<footnote>
          <para>Which might be a source of problems with variable liveliness
          after <code>context_restore</code>.</para>
        </footnote>. Nevertheless, a synchronous register context, which is
      saved in a memory structure passed to <code>context_restore,</code> is
      restored, thus transfering the control flow to the place of occurrence
      of the corresponding call to <code>context_save</code>. From the
      perspective of the caller of the corresponding
      <code>context_save</code>, it looks as though a return from
      <code>context_save</code>. However, this time a return value of 0 is
      returned.</para>
    </section>
  </section>

  <section>
    <title>Threads</title>

    <para>A thread is the basic executable entity with some code and stack.
    While the code, implemented by a C language function, can be shared by
    several threads, the stack is always private to each instance of the
    thread. Each thread belongs to exactly one task through which it shares
    address space with its sibling threads. Threads that execute purely in the
    kernel don't have any userspace memory allocated. However, when a thread
    has ambitions to run in userspace, it must be allocated a userspace stack.
    The distinction between the purely kernel threads and threads running also
    in userspace is made by refering to the former group as to kernel threads
    and to the latter group as to userspace threads. Both kernel and userspace
    threads are visible to the scheduler and can become a subject of kernel
    preemption and thread migration during times when preemption is
    possible.</para>

    <para>HelenOS userspace layer knows even smaller units of execution. Each
    userspace thread can make use of an arbitrary number of pseudo threads.
    These pseudo threads have their own synchronous register context,
    userspace code and stack. They live their own life within the userspace
    thread and the scheduler does not have any idea about them because they
    are completely implemented by the userspace library. This implies several
    things:</para>

    <itemizedlist>
      <listitem>
        <para>pseudothreads schedule themselves cooperatively within the time
        slice given to their userspace thread,</para>
      </listitem>

      <listitem>
        <para>pseudothreads share FPU context of their containing thread
        and</para>
      </listitem>

      <listitem>
        <para>all pseudothreads of one userspace thread block when one of them
        goes to sleep.</para>
      </listitem>
    </itemizedlist>

    <para></para>
  </section>
</chapter>