Subversion Repositories HelenOS-doc

<?xml version="1.0" encoding="UTF-8"?>

<?xml version="1.0" encoding="UTF-8"?>

<chapter id="time">

<chapter id="time">

  <?dbhtml filename="time.html"?>

  <?dbhtml filename="time.html"?>

  <title>Time Management</title>

  <title>Time Management</title>

  Knowledge of time makes it possible for the scheduler to preemptively plan

  Knowledge of time makes it possible for the scheduler to preemptively plan

  threads for execution. Different parts of the kernel can request execution

  threads for execution. Different parts of the kernel can request execution

  kernel code is the synchronization subsystem which uses this functionality

  kernel code is the synchronization subsystem which uses this functionality

  to implement timeouting versions of synchronization primitives.</para>

  to implement timeouting versions of synchronization primitives.</para>

  <section>

  <section>

    <title>System Clock</title>

    <title>System Clock</title>

    <para>Every hardware architecture supported by HelenOS must support some

    <para>Every hardware architecture supported by HelenOS must support some

    kind of a device that can be programmed to yield periodic time signals

    kind of a device that can be programmed to yield periodic time signals

    (i.e. clock interrupts). Some architectures have external clock that is

    (i.e. clock interrupts). Some architectures have external clock that is

    merely programmed by the kernel to interrupt the processor multiple times

    merely programmed by the kernel to interrupt the processor multiple times

    in a second. This is the case of ia32 and amd64 architectures<footnote>

    in a second. This is the case of ia32 and amd64 architectures<footnote>

        <para>When running in uniprocessor mode.</para>

        <para>When running in uniprocessor mode.</para>

      </footnote>, which use i8254 or a compatible chip to achieve the

      </footnote>, which use i8254 or a compatible chip to achieve the

    goal.</para>

    goal.</para>

    <para>Other architectures' processors typically contain two registers. The

    <para>Other architectures' processors typically contain two registers. The

    first register is usually called a compare or a match register and can be

    first register is usually called a compare or a match register and can be

    set to an arbitrary value by the operating system. The contents of the

    set to an arbitrary value by the operating system. The contents of the

    compare register then stays unaltered until it is written by the kernel

    compare register then stays unaltered until it is written by the kernel

    again. The second register, often called a counter register, can be also

    again. The second register, often called a counter register, can be also

    written by the kernel, but the processor automatically increments it after

    written by the kernel, but the processor automatically increments it after

    every executed instruction or in some fixed relation to processor speed.

    every executed instruction or in some fixed relation to processor speed.

    The point is that a clock interrupt is generated whenever the values of

    The point is that a clock interrupt is generated whenever the values of

    the counter and the compare registers match. Sometimes, the scheme of two

    the counter and the compare registers match. Sometimes, the scheme of two

    registers is modified so that only one register is needed. Such a

    registers is modified so that only one register is needed. Such a

    register, called a decrementer, then counts towards zero and an interrupt

    register, called a decrementer, then counts towards zero and an interrupt

    is generated when zero is reached.</para>

    is generated when zero is reached.</para>

    <para>In any case, the initial value of the decrementer or the initial

    <para>In any case, the initial value of the decrementer or the initial

    difference between the counter and the compare registers, respectively,

    difference between the counter and the compare registers, respectively,

    must be set accordingly to a known relation between the real time and the

    must be set accordingly to a known relation between the real time and the

    speed of the decrementer or the counter register, respectively.</para>

    speed of the decrementer or the counter register, respectively.</para>

    <para>The rest of this section will, for the sake of clarity, focus on the

    <para>The rest of this section will, for the sake of clarity, focus on the

    two-register scheme. The decrementer scheme is very similar.</para>

    two-register scheme. The decrementer scheme is very similar.</para>

    <para>The kernel must reinitialize one of the two registers after each

    <para>The kernel must reinitialize one of the two registers after each

    clock interrupt in order to schedule next interrupt. However this step is

    clock interrupt in order to schedule next interrupt. However this step is

    tricky and must be done with caution. Imagine that the clock interrupt is

    tricky and must be done with caution. Imagine that the clock interrupt is

    masked either because the kernel is servicing another interrupt or because

    masked either because the kernel is servicing another interrupt or because

    the processor locally disabled interrupts for a while. If the clock

    the processor locally disabled interrupts for a while. If the clock

    <para>Let us assume that the kernel wants each clock interrupt be

    <para>Let us assume that the kernel wants each clock interrupt be

    generated every <constant>TICKCONST</constant> ticks. This value

    generated every <constant>TICKCONST</constant> ticks. This value

    represents the ideal number of ticks between two non-delayed clock

    represents the ideal number of ticks between two non-delayed clock

    interrupts and has some known relation to real time. On each clock

    interrupts and has some known relation to real time. On each clock

    interrupt, the kernel computes and writes down the expected value of the

    interrupt, the kernel computes and writes down the expected value of the

    counter register as it hopes to read it on the next clock interrupt. When

    counter register as it hopes to read it on the next clock interrupt. When

    that interrupt comes, the kernel reads the counter register again and

    that interrupt comes, the kernel reads the counter register again and

    compares it with the written down value. If the difference is smaller than

    compares it with the written down value. If the difference is smaller than

    or equal to <constant>TICKCONST</constant>, then the time drift is none or

    or equal to <constant>TICKCONST</constant>, then the time drift is none or

    small and the next interrupt is scheduled earlier with a penalty of so

    small and the next interrupt is scheduled earlier with a penalty of so

    many ticks as is the value of the difference. However, if the difference

    many ticks as is the value of the difference. However, if the difference

    is bigger, then at least one clock signal was missed. In that case, the

    is bigger, then at least one clock signal was missed. In that case, the

    missed clock signal is remembered in the special counter. If there are

    missed clock signal is remembered in the special counter. If there are

    more missed signals, each of them is recorded there. The next interrupt is

    more missed signals, each of them is recorded there. The next interrupt is

    scheduled with respect to the difference similarily to the former case.

    scheduled with respect to the difference similarily to the former case.

    This time, the penalty is taken modulo <constant>TICKCONST</constant>. The

    This time, the penalty is taken modulo <constant>TICKCONST</constant>. The

    effect of missed clock signals is remedied in the generic clock interrupt

    effect of missed clock signals is remedied in the generic clock interrupt

    handler.</para>

    handler.</para>

  </section>

  </section>

  <section>

  <section>

    <title>Timeouts</title>

    <title>Timeouts</title>

    <para>Kernel subsystems can register a callback function to be executed

    <para>Kernel subsystems can register a callback function to be executed

    with a specified delay. Such a registration is represented by a kernel

    with a specified delay. Such a registration is represented by a kernel

    structure called <classname>timeout</classname>. Timeouts are registered

    structure called <classname>timeout</classname>. Timeouts are registered

    via <code>timeout_register</code> function. This function takes a pointer

    via <code>timeout_register</code> function. This function takes a pointer

    to a timeout structure, a callback function, a parameter of the callback

    to a timeout structure, a callback function, a parameter of the callback

    function and a delay in microseconds as parameters. After the structure is

    function and a delay in microseconds as parameters. After the structure is

    initialized with all these values, it is sorted into the processor's list

    initialized with all these values, it is sorted into the processor's list

    of active timeouts, according to the number of clock interrupts remaining

    of active timeouts, according to the number of clock interrupts remaining

    to their expiration and relatively to already listed timeouts.</para>

    to their expiration and relatively to already listed timeouts.</para>

    <para>Timeouts can be unregistered via <code>timeout_unregister</code>.

    <para>Timeouts can be unregistered via <code>timeout_unregister</code>.

    This function can, as opposed to <code>timeout_register</code>, fail when

    This function can, as opposed to <code>timeout_register</code>, fail when

    it is too late to remove the timeout from the list of active

    it is too late to remove the timeout from the list of active

    timeouts.</para>

    timeouts.</para>

    <para>Timeouts are nearing their expiration in the list of active timeouts

    <para>Timeouts are nearing their expiration in the list of active timeouts

    which exists on every processor in the system. The expiration counters are

    which exists on every processor in the system. The expiration counters are

    decremented on each clock interrupt by the generic clock interrupt

    decremented on each clock interrupt by the generic clock interrupt

    handler. Due to the relative ordering of timeouts in the list, it is

    handler. Due to the relative ordering of timeouts in the list, it is

    sufficient to decrement expiration counter only of the first timeout in

    sufficient to decrement expiration counter only of the first timeout in

    the list. Timeouts with expiration counter equal to zero are removed from

    the list. Timeouts with expiration counter equal to zero are removed from

    the list and their callback function is called with respective

    the list and their callback function is called with respective

    parameter.</para>

    parameter.</para>

  </section>

  </section>

  <section>

  <section>

    <title>Generic Clock Interrupt Handler</title>

    <title>Generic Clock Interrupt Handler</title>

    <para>On each clock interrupt, the architecture specific part of the clock

    <para>On each clock interrupt, the architecture specific part of the clock

    interrupt handler makes a call to the generic clock interrupt handler

    interrupt handler makes a call to the generic clock interrupt handler

    implemented by the <code>clock</code> function. The generic handler takes

    implemented by the <code>clock</code> function. The generic handler takes

    care of several mission critical goals:</para>

    care of several mission critical goals:</para>

    <itemizedlist>

    <itemizedlist>

      <listitem>

      <listitem>

        <para>expiration of timeouts,</para>

        <para>expiration of timeouts,</para>

      </listitem>

      </listitem>

      <listitem>

      <listitem>

        <para>updating time of the day counters for userspace and</para>

        <para>updating time of the day counters for userspace and</para>

      </listitem>

      </listitem>

      <listitem>

      <listitem>

        <para>preemption of threads.</para>

        <para>preemption of threads.</para>

      </listitem>

      </listitem>

    </itemizedlist>

    </itemizedlist>

    <para>The <code>clock</code> function checks for expired timeouts and

    <para>The <code>clock</code> function checks for expired timeouts and

    decrements unexpired timeout expiration counters exactly one more times

    decrements unexpired timeout expiration counters exactly one more times

    than is the number of missed clock signals (i.e. at least once and

    than is the number of missed clock signals (i.e. at least once and

    possibly more times, depending on the missed clock signals counter). The

    possibly more times, depending on the missed clock signals counter). The

    time of the day counters are also updated one more times than is the

    time of the day counters are also updated one more times than is the

    number of missed clock signals. And finally, the remaining timeslice of

    number of missed clock signals. And finally, the remaining timeslice of

    the running thread is decremented with respect to this counter as well. By

    the running thread is decremented with respect to this counter as well. By

    considering its value, the kernel performs actions that would otherwise be

    considering its value, the kernel performs actions that would otherwise be

    lost due to an occasional excessive time drift described in previous

    lost due to an occasional excessive time drift described in previous

    paragraphs.</para>

    paragraphs.</para>

  </section>

  </section>

  <section>

  <section>

    <title>Time Source for Userspace</title>

    <title>Time Source for Userspace</title>

    <para>In HelenOS, userspace tasks don't communicate with the kernel in

    <para>In HelenOS, userspace tasks don't communicate with the kernel in

    order to read the system time. Instead, a mechanism that shares kernel

    order to read the system time. Instead, a mechanism that shares kernel

    time of the the day counters with userspace address spaces is deployed. On

    time of the the day counters with userspace address spaces is deployed. On

    the kernel side, during system initialization, HelenOS allocates a frame

    the kernel side, during system initialization, HelenOS allocates a frame

    of physical memory and stores the time of the day counters there. The

    of physical memory and stores the time of the day counters there. The

    counters have the following structure:</para>

    counters have the following structure:</para>

    <itemizedlist>

    <itemizedlist>

      <listitem>

      <listitem>

        <para>first 32-bit counter for seconds,</para>

        <para>first 32-bit counter for seconds,</para>

      </listitem>

      </listitem>

      <listitem>

      <listitem>

        <para>32-bit counter for microseconds and</para>

        <para>32-bit counter for microseconds and</para>

      </listitem>

      </listitem>

      <listitem>

      <listitem>

        <para>second 32-bit counter for seconds.</para>

        <para>second 32-bit counter for seconds.</para>

      </listitem>

      </listitem>

    </itemizedlist>

    </itemizedlist>

    <para>One of the userspace tasks with capabilities of memory manager (e.g.

    <para>One of the userspace tasks with capabilities of memory manager (e.g.

    ns) asks the kernel to map this frame into its address space. Other

    ns) asks the kernel to map this frame into its address space. Other

    non-privileged tasks then use IPC to receive read-only share of this

    non-privileged tasks then use IPC to receive read-only share of this

    memory. Reading time in a userspace task is therefore just a matter of

    memory. Reading time in a userspace task is therefore just a matter of

    reading memory.</para>

    reading memory.</para>

    <para>There are two interesting points about this. First, the counters are

    <para>There are two interesting points about this. First, the counters are

    32-bit even on 64-bit machines. The goal is to provide subsecond precision

    32-bit even on 64-bit machines. The goal is to provide subsecond precision

    with the possibility to span roughly 136 years. Note that a single 64-bit

    with the possibility to span roughly 136 years. Note that a single 64-bit

    microsecond counter could not be usually read atomically on 32-bit

    microsecond counter could not be usually read atomically on 32-bit

    platforms. Unfortunately, on 32-bit platforms it is usually impossible to

    platforms. Unfortunately, on 32-bit platforms it is usually impossible to

    read atomically two 32-bit counters either. However, a generic protocol is

    read atomically two 32-bit counters either. However, a generic protocol is

    used to guarantee that sequentially read times will create a

    used to guarantee that sequentially read times will create a

    non-decreasing sequence.</para>

    non-decreasing sequence.</para>

    <para>The problematic part is incrementing seconds counter and clearing

    <para>The problematic part is incrementing seconds counter and clearing

    microseconds counter together once every second. Seconds must be

    microseconds counter together once every second. Seconds must be

    incremented and microseconds must be reset. However, without any

    incremented and microseconds must be reset. However, without any

    synchronization, the two kernel stores and the two userspace reads can

    synchronization, the two kernel stores and the two userspace reads can

    arbitrarily interleave. Furthemore, the reader has no chance to detect

    arbitrarily interleave. Furthemore, the reader has no chance to detect

    that the counters were updated only paritally. Therefore three counters

    that the counters were updated only paritally. Therefore three counters

    are used in HelenOS.</para>

    are used in HelenOS.</para>

    <para>If seconds need to be updated, the kernel increments the first

    <para>If seconds need to be updated, the kernel increments the first

    second counter, issues a write memory barrier operation, updates the

    second counter, issues a write memory barrier operation, updates the

    microsecond counter, issues another write memory barrier operation and

    microsecond counter, issues another write memory barrier operation and

    increments the second second counter. When only microseconds needs to be

    increments the second second counter. When only microseconds needs to be

    updated, no special action is taken by the kernel. On the other hand, the

    updated, no special action is taken by the kernel. On the other hand, the

    userspace task must always read all three counters in reversed order. A

    userspace task must always read all three counters in reversed order. A

    read memory barrier operation must be issued between each two reads. A

    read memory barrier operation must be issued between each two reads. A

    non-atomic read is detected when the two second counters differ. The

    non-atomic read is detected when the two second counters differ. The

    userspace library solves this situation by returning higher of them with

    userspace library solves this situation by returning higher of them with

    microseconds set to zero.</para>

    microseconds set to zero.</para>

  </section>

  </section>

</chapter>

</chapter>