source: vbox/trunk/doc/manual/en_US/user_Technical.xml@ 32590

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook XML V4.4//EN"
  "http://www.oasis-open.org/docbook/xml/4.4/docbookx.dtd">
<chapter id="TechnicalBackground">
  <title>Technical background</title>

  <para>The contents of this chapter are not required to use VirtualBox
  successfully. The following is provided as additional information for
  readers who are more familiar with computer architecture and technology and
  wish to find out more about how VirtualBox works "under the hood".</para>

  <sect1>
    <title>VirtualBox executables and components</title>

    <para>VirtualBox was designed to be modular and flexible. When the
    VirtualBox graphical user interface (GUI) is opened and a VM is started,
    at least three processes are running:<orderedlist>
        <listitem>
          <para><computeroutput>VBoxSVC</computeroutput>, the VirtualBox
          service process which always runs in the background. This process is
          started automatically by the first VirtualBox client process (the
          GUI, <computeroutput>VBoxManage</computeroutput>,
          <computeroutput>VBoxHeadless</computeroutput>, the web service or
          others) and exits a short time after the last client exits. The
          service is responsible for bookkeeping, maintaining the state of all
          VMs, and for providing communication between VirtualBox components.
          This communication is implemented via COM/XPCOM.<note>
              <para>When we refer to "clients" here, we mean the local clients
              of a particular <computeroutput>VBoxSVC</computeroutput> server
              process, not clients in a network. VirtualBox employs its own
              client/server design to allow its processes to cooperate, but
              all these processes run under the same user account on the host
              operating system, and this is totally transparent to the
              user.</para>
            </note></para>
        </listitem>

        <listitem>
          <para>The GUI process, <computeroutput>VirtualBox</computeroutput>,
          a client application based on the cross-platform Qt library. When
          started without the <computeroutput>--startvm</computeroutput>
          option, this application acts as the VirtualBox main window,
          displaying the VMs and their settings. It then communicates settings
          and state changes to <computeroutput>VBoxSVC</computeroutput> and
          also reflects changes effected through other means, e.g.,
          <computeroutput>VBoxManage</computeroutput>.</para>
        </listitem>

        <listitem>
          <para>If the <computeroutput>VirtualBox</computeroutput> client
          application is started with the
          <computeroutput>--startvm</computeroutput> argument, it loads the
          VMM library which includes the actual hypervisor and then runs a
          virtual machine and provides the input and output for the
          guest.</para>
        </listitem>
      </orderedlist></para>

    <para>Any VirtualBox front-end (client) will communicate with the service
    process and can both control and reflect the current state. For example,
    either the VM selector or the VM window or VBoxManage can be used to pause
    the running VM, and other components will always reflect the changed
    state.</para>

    <para>The VirtualBox GUI application is only one of several available
    front-ends (clients). The complete list shipped with VirtualBox
    is:<orderedlist>
        <listitem>
          <para><computeroutput>VirtualBox</computeroutput>, the Qt GUI front
          end mentioned earlier.</para>
        </listitem>

        <listitem>
          <para><computeroutput>VBoxManage</computeroutput>, a less
          user-friendly but more powerful alternative to the GUI described in
          <xref linkend="vboxmanage" />.</para>
        </listitem>

        <listitem>
          <para><computeroutput>VBoxSDL</computeroutput>, a simple graphical
          front end based on the SDL library; see <xref
          linkend="vboxsdl" />.</para>
        </listitem>

        <listitem>
          <para><computeroutput>VBoxHeadless</computeroutput>, a VM front end
          which does not directly provide any video output and keyboard/mouse
          input, but allows redirection via VRDP; see <xref
          linkend="vboxheadless" />.</para>
        </listitem>

        <listitem>
          <para><computeroutput>vboxwebsrv</computeroutput>, the VirtualBox
          web service process which allows for controlling a VirtualBox host
          remotely. This is described in detail in the VirtualBox Software
          Development Kit (SDK) reference; please see <xref
          linkend="VirtualBoxAPI" /> for details.</para>
        </listitem>

        <listitem>
          <para>The VirtualBox Python shell, a Python alternative to
          VBoxManage. This is also described in the SDK reference.</para>
        </listitem>
      </orderedlist></para>

    <para>Internally, VirtualBox consists of many more or less separate
    components. You may encounter these when analyzing VirtualBox internal
    error messages or log files. These include:</para>

    <itemizedlist>
      <listitem>
        <para>IPRT, a portable runtime library which abstracts file access,
        threading, string manipulation, etc. Whenever VirtualBox accesses host
        operating features, it does so through this library for cross-platform
        portability.</para>
      </listitem>

      <listitem>
        <para>VMM (Virtual Machine Monitor), the heart of the
        hypervisor.</para>
      </listitem>

      <listitem>
        <para>EM (Execution Manager), controls execution of guest code.</para>
      </listitem>

      <listitem>
        <para>REM (Recompiled Execution Monitor), provides software emulation
        of CPU instructions.</para>
      </listitem>

      <listitem>
        <para>TRPM (Trap Manager), intercepts and processes guest traps and
        exceptions.</para>
      </listitem>

      <listitem>
        <para>HWACCM (Hardware Acceleration Manager), provides support for
        VT-x and AMD-V.</para>
      </listitem>

      <listitem>
        <para>PDM (Pluggable Device Manager), an abstract interface between
        the VMM and emulated devices which separates device implementations
        from VMM internals and makes it easy to add new emulated devices.
        Through PDM, third-party developers can add new virtual devices to
        VirtualBox without having to change VirtualBox itself.</para>
      </listitem>

      <listitem>
        <para>PGM (Page Manager), a component controlling guest paging.</para>
      </listitem>

      <listitem>
        <para>PATM (Patch Manager), patches guest code to improve and speed up
        software virtualization.</para>
      </listitem>

      <listitem>
        <para>TM (Time Manager), handles timers and all aspects of time inside
        guests.</para>
      </listitem>

      <listitem>
        <para>CFGM (Configuration Manager), provides a tree structure which
        holds configuration settings for the VM and all emulated
        devices.</para>
      </listitem>

      <listitem>
        <para>SSM (Saved State Manager), saves and loads VM state.</para>
      </listitem>

      <listitem>
        <para>VUSB (Virtual USB), a USB layer which separates emulated USB
        controllers from the controllers on the host and from USB devices;
        this also enables remote USB.</para>
      </listitem>

      <listitem>
        <para>DBGF (Debug Facility), a built-in VM debuger.</para>
      </listitem>

      <listitem>
        <para>VirtualBox emulates a number of devices to provide the hardware
        environment that various guests need. Most of these are standard
        devices found in many PC compatible machines and widely supported by
        guest operating systems. For network and storage devices in
        particular, there are several options for the emulated devices to
        access the underlying hardware. These devices are managed by
        PDM.</para>
      </listitem>

      <listitem>
        <para>Guest Additions for various guest operating systems. This is
        code that is installed from within a virtual machine; see <xref
        linkend="guestadditions" />.</para>
      </listitem>

      <listitem>
        <para>The "Main" component is special: it ties all the above bits
        together and is the only public API that VirtualBox provides. All the
        client processes listed above use only this API and never access the
        hypervisor components directly. As a result, third-party applications
        that use the VirtualBox Main API can rely on the fact that it is
        always well-tested and that all capabilities of VirtualBox are fully
        exposed. It is this API that is described in the VirtualBox SDK
        mentioned above (again, see <xref linkend="VirtualBoxAPI" />).</para>
      </listitem>
    </itemizedlist>
  </sect1>

  <sect1 id="hwvirt">
    <title>Hardware vs. software virtualization</title>

    <para>VirtualBox allows software in the virtual machine to run directly on
    the processor of the host, but an array of complex techniques is employed
    to intercept operations that would interfere with your host. Whenever the
    guest attempts to do something that could be harmful to your computer and
    its data, VirtualBox steps in and takes action. In particular, for lots of
    hardware that the guest believes to be accessing, VirtualBox simulates a
    certain "virtual" environment according to how you have configured a
    virtual machine. For example, when the guest attempts to access a hard
    disk, VirtualBox redirects these requests to whatever you have configured
    to be the virtual machine's virtual hard disk -- normally, an image file
    on your host.</para>

    <para>Unfortunately, the x86 platform was never designed to be
    virtualized. Detecting situations in which VirtualBox needs to take
    control over the guest code that is executing, as described above, is
    difficult. There are two ways in which to achive this:<itemizedlist>
        <listitem>
          <para>Since 2006, Intel and AMD processors have had support for
          so-called <emphasis role="bold">"hardware
          virtualization"</emphasis>. This means that these processors can
          help VirtualBox to intercept potentially dangerous operations that a
          guest operating system may be attempting and also makes it easier to
          present virtual hardware to a virtual machine.</para>

          <para>These hardware features differ between Intel and AMD
          processors. Intel named its technology <emphasis
          role="bold">VT-x</emphasis>; AMD calls theirs <emphasis
          role="bold">AMD-V</emphasis>. The Intel and AMD support for
          virtualization is very different in detail, but not very different
          in principle.<note>
              <para>On many systems, the hardware virtualization features
              first need to be enabled in the BIOS before VirtualBox can use
              them.</para>
            </note></para>
        </listitem>

        <listitem>
          <para>As opposed to other virtualization software, for many usage
          scenarios, VirtualBox does not <emphasis>require</emphasis> hardware
          virtualization features to be present. Through sophisticated
          techniques, VirtualBox virtualizes many guest operating systems
          entirely in <emphasis role="bold">software</emphasis>. This means
          that you can run virtual machines even on older processors which do
          not support hardware virtualization.</para>
        </listitem>
      </itemizedlist></para>

    <para>Even though VirtualBox does not always require hardware
    virtualization, enabling it is <emphasis>required</emphasis> in the
    following scenarios:<itemizedlist>
        <listitem>
          <para>Certain rare guest operating systems like OS/2 make use of
          very esoteric processor instructions that are not supported with our
          software virtualization. For virtual machines that are configured to
          contain such an operating system, hardware virtualization is enabled
          automatically.</para>
        </listitem>

        <listitem>
          <para>VirtualBox's 64-bit guest support (added with version 2.0) and
          multiprocessing (SMP, added with version 3.0) both require hardware
          virtualization to be enabled. (This is not much of a limitation
          since the vast majority of today's 64-bit and multicore CPUs ship
          with hardware virtualization anyway; the exceptions to this rule are
          e.g. older Intel Celeron and AMD Opteron CPUs.)</para>
        </listitem>
      </itemizedlist></para>

    <warning>
      <para>Do not run other hypervisors (open-source or commercial
      virtualization products) together with VirtualBox! While several
      hypervisors can normally be <emphasis>installed</emphasis> in parallel,
      do not attempt to <emphasis>run</emphasis> several virtual machines from
      competing hypervisors at the same time. VirtualBox cannot track what
      another hypervisor is currently attempting to do on the same host, and
      especially if several products attempt to use hardware virtualization
      features such as VT-x, this can crash the entire host. Also, within
      VirtualBox, you can mix software and hardware virtualization when
      running multiple VMs. In certain cases a small performance penalty will
      be unavoidable when mixing VT-x and software virtualization VMs. We
      recommend not mixing virtualization modes if maximum performance and low
      overhead are essential. This does <emphasis>not</emphasis> apply to
      AMD-V.</para>
    </warning>
  </sect1>

  <sect1>
    <title>Details about software virtualization</title>

    <para>Implementing virtualization on x86 CPUs with no hardware
    virtualization support is an extraordinarily complex task because the CPU
    architecture was not designed to be virtualized. The problems can usually
    be solved, but at the cost of reduced performance. Thus, there is a
    constant clash between virtualization performance and accuracy.</para>

    <para>The x86 instruction set was originally designed in the 1970s and
    underwent significant changes with the addition of protected mode in the
    1980s with the 286 CPU architecture and then again with the Intel 386 and
    its 32-bit architecture. Whereas the 386 did have limited virtualization
    support for real mode operation (V86 mode, as used by the "DOS Box" of
    Windows 3.x and OS/2 2.x), no support was provided for virtualizing the
    entire architecture.</para>

    <para>In theory, software virtualization is not overly complex. In
    addition to the four privilege levels ("rings") provided by the hardware
    (of which typically only two are used: ring 0 for kernel mode and ring 3
    for user mode), one needs to differentiate between "host context" and
    "guest context".</para>

    <para>In "host context", everything is as if no hypervisor was active.
    This might be the active mode if another application on your host has been
    scheduled CPU time; in that case, there is a host ring 3 mode and a host
    ring 0 mode. The hypervisor is not involved.</para>

    <para>In "guest context", however, a virtual machine is active. So long as
    the guest code is running in ring 3, this is not much of a problem since a
    hypervisor can set up the page tables properly and run that code natively
    on the processor. The problems mostly lie in how to intercept what the
    guest's kernel does.</para>

    <para>There are several possible solutions to these problems. One approach
    is full software emulation, usually involving recompilation. That is, all
    code to be run by the guest is analyzed, transformed into a form which
    will not allow the guest to either modify or see the true state of the
    CPU, and only then executed. This process is obviously highly complex and
    costly in terms of performance. (VirtualBox contains a recompiler based on
    QEMU which can be used for pure software emulation, but the recompiler is
    only activated in special situations, described below.)</para>

    <para>Another possible solution is paravirtualization, in which only
    specially modified guest OSes are allowed to run. This way, most of the
    hardware access is abstracted and any functions which would normally
    access the hardware or privileged CPU state are passed on to the
    hypervisor instead. Paravirtualization can achieve good functionality and
    performance on standard x86 CPUs, but it can only work if the guest OS can
    actually be modified, which is obviously not always the case.</para>

    <para>VirtualBox chooses a different approach. When starting a virtual
    machine, through its ring-0 support kernel driver, VirtualBox has set up
    the host system so that it can run most of the guest code natively, but it
    has inserted itself at the "bottom" of the picture. It can then assume
    control when needed -- if a privileged instruction is executed, the guest
    traps (in particular because an I/O register was accessed and a device
    needs to be virtualized) or external interrupts occur. VirtualBox may then
    handle this and either route a request to a virtual device or possibly
    delegate handling such things to the guest or host OS. In guest context,
    VirtualBox can therefore be in one of three states:</para>

    <para><itemizedlist>
        <listitem>
          <para>Guest ring 3 code is run unmodified, at full speed, as much as
          possible. The number of faults will generally be low (unless the
          guest allows port I/O from ring 3, something we cannot do as we
          don't want the guest to be able to access real ports). This is also
          referred to as "raw mode", as the guest ring-3 code runs
          unmodified.</para>
        </listitem>

        <listitem>
          <para>For guest code in ring 0, VirtualBox employs a nasty trick: it
          actually reconfigures the guest so that its ring-0 code is run in
          ring 1 instead (which is normally not used in x86 operating
          systems). As a result, when guest ring-0 code (actually running in
          ring 1) such as a guest device driver attempts to write to an I/O
          register or execute a privileged instruction, the VirtualBox
          hypervisor in "real" ring 0 can take over.</para>
        </listitem>

        <listitem>
          <para>The hypervisor (VMM) can be active. Every time a fault occurs,
          VirtualBox looks at the offending instruction and can relegate it to
          a virtual device or the host OS or the guest OS or run it in the
          recompiler.</para>

          <para>In particular, the recompiler is used when guest code disables
          interrupts and VirtualBox cannot figure out when they will be
          switched back on (in these situations, VirtualBox actually analyzes
          the guest code using its own disassembler). Also, certain privileged
          instructions such as LIDT need to be handled specially. Finally, any
          real-mode or protected-mode code (e.g. BIOS code, a DOS guest, or
          any operating system startup) is run in the recompiler
          entirely.</para>
        </listitem>
      </itemizedlist></para>

    <para>Unfortunately this only works to a degree. Among others, the
    following situations require special handling:</para>

    <para><orderedlist>
        <listitem>
          <para>Running ring 0 code in ring 1 causes a lot of additional
          instruction faults, as ring 1 is not allowed to execute any
          privileged instructions (of which guest's ring-0 contains plenty).
          With each of these faults, the VMM must step in and emulate the code
          to achieve the desired behavior. While this works, emulating
          thousands of these faults is very expensive and severely hurts the
          performance of the virtualized guest.</para>
        </listitem>

        <listitem>
          <para>There are certain flaws in the implementation of ring 1 in the
          x86 architecture that were never fixed. Certain instructions that
          <emphasis>should</emphasis> trap in ring 1 don't. This affect for
          example the LGDT/SGDT, LIDT/SIDT, or POPF/PUSHF instruction pairs.
          Whereas the "load" operation is privileged and can therefore be
          trapped, the "store" instruction always succeed. If the guest is
          allowed to execute these, it will see the true state of the CPU, not
          the virtualized state. The CPUID instruction also has the same
          problem.</para>
        </listitem>

        <listitem>
          <para>A hypervisor typically needs to reserve some portion of the
          guest's address space (both linear address space and selectors) for
          its own use. This is not entirely transparent to the guest OS and
          may cause clashes.</para>
        </listitem>

        <listitem>
          <para>The SYSENTER instruction (used for system calls) executed by
          an application running in a guest OS always transitions to ring 0.
          But that is where the hypervisor runs, not the guest OS. In this
          case, the hypervisor must trap and emulate the instruction even when
          it is not desirable.</para>
        </listitem>

        <listitem>
          <para>The CPU segment registers contain a "hidden" descriptor cache
          which is not software-accessible. The hypervisor cannot read, save,
          or restore this state, but the guest OS may use it.</para>
        </listitem>

        <listitem>
          <para>Some resources must (and can) be trapped by the hypervisor,
          but the access is so frequent that this creates a significant
          performance overhead. An example is the TPR (Task Priority) register
          in 32-bit mode. Accesses to this register must be trapped by the
          hypervisor, but certain guest operating systems (notably Windows and
          Solaris) write this register very often, which adversely affects
          virtualization performance.</para>
        </listitem>
      </orderedlist></para>

    <para>To fix these performance and security issues, VirtualBox contains a
    Code Scanning and Analysis Manager (CSAM), which disassembles guest code,
    and the Patch Manager (PATM), which can replace it at runtime.</para>

    <para>Before executing ring 0 code, CSAM scans it recursively to discover
    problematic instructions. PATM then performs <emphasis>in-situ
    </emphasis>patching, i.e. it replaces the instruction with a jump to
    hypervisor memory where an integrated code generator has placed a more
    suitable implementation. In reality, this is a very complex task as there
    are lots of odd situations to be discovered and handled correctly. So,
    with its current complexity, one could argue that PATM is an advanced
    <emphasis>in-situ</emphasis> recompiler.</para>

    <para>In addition, every time a fault occurs, VirtualBox analyzes the
    offending code to determine if it is possible to patch it in order to
    prevent it from causing more faults in the future. This approach works
    well in practice and dramatically improves software virtualization
    performance.</para>
  </sect1>

  <sect1>
    <title>Details about hardware virtualization</title>

    <para>With Intel VT-x, there are two distinct modes of CPU operation: VMX
    root mode and non-root mode.<itemizedlist>
        <listitem>
          <para>In root mode, the CPU operates much like older generations of
          processors without VT-x support. There are four privilege levels
          ("rings"), and the same instruction set is supported, with the
          addition of several virtualization specific instruction. Root mode
          is what a host operating system without virtualization uses, and it
          is also used by a hypervisor when virtualization is active.</para>
        </listitem>

        <listitem>
          <para>In non-root mode, CPU operation is significantly different.
          There are still four privilege rings and the same instruction set,
          but a new structure called VMCS (Virtual Machine Control Structure)
          now controls the CPU operation and determines how certain
          instructions behave. Non-root mode is where guest systems
          run.</para>
        </listitem>
      </itemizedlist></para>

    <para>Switching from root mode to non-root mode is called "VM entry", the
    switch back is "VM exit". The VMCS includes a guest and host state area
    which is saved/restored at VM entry and exit. Most importantly, the VMCS
    controls which guest operations will cause VM exits.</para>

    <para>The VMCS provides fairly fine-grained control over what the guests
    can and can't do. For example, a hypervisor can allow a guest to write
    certain bits in shadowed control registers, but not others. This enables
    efficient virtualization in cases where guests can be allowed to write
    control bits without disrupting the hypervisor, while preventing them from
    altering control bits over which the hypervisor needs to retain full
    control. The VMCS also provides control over interrupt delivery and
    exceptions.</para>

    <para>Whenever an instruction or event causes a VM exit, the VMCS contains
    information about the exit reason, often with accompanying detail. For
    example, if a write to the CR0 register causes an exit, the offending
    instruction is recorded, along with the fact that a write access to a
    control register caused the exit, and information about source and
    destination register. Thus the hypervisor can efficiently handle the
    condition without needing advanced techniques such as CSAM and PATM
    described above.</para>

    <para>VT-x inherently avoids several of the problems which software
    virtualization faces. The guest has its own completely separate address
    space not shared with the hypervisor, which eliminates potential clashes.
    Additionally, guest OS kernel code runs at privilege ring 0 in VMX
    non-root mode, obviating the problems by running ring 0 code at less
    privileged levels. For example the SYSENTER instruction can transition to
    ring 0 without causing problems. Naturally, even at ring 0 in VMX non-root
    mode, any I/O access by guest code still causes a VM exit, allowing for
    device emulation.</para>

    <para>The biggest difference between VT-x and AMD-V is that AMD-V provides
    a more complete virtualization environment. VT-x requires the VMX non-root
    code to run with paging enabled, which precludes hardware virtualization
    of real-mode code and non-paged protected-mode software. This typically
    only includes firmware and OS loaders, but nevertheless complicates VT-x
    hypervisor implementation. AMD-V does not have this restriction.</para>

    <para>Of course hardware virtualization is not perfect. Compared to
    software virtualization, the overhead of VM exits is relatively high. This
    causes problems for devices whose emulation requires high number of traps.
    One example is the VGA device in 16-color modes, where not only every I/O
    port access but also every access to the framebuffer memory must be
    trapped.</para>
  </sect1>

  <sect1 id="nestedpaging">
    <title>Nested paging and VPIDs</title>

    <para>In addition to "plain" hardware virtualization, your processor may
    also support additional sophisticated techniques:<footnote>
        <para>VirtualBox 2.0 added support for AMD's nested paging; support
        for Intel's EPT and VPIDs was added with version 2.1.</para>
      </footnote><itemizedlist>
        <listitem>
          <para>A newer feature called <emphasis role="bold">"nested
          paging"</emphasis> implements some memory management in hardware,
          which can greatly accelerate hardware virtualization since these
          tasks no longer need to be performed by the virtualization
          software.</para>

          <para>With nested paging, the hardware provides another level of
          indirection when translating linear to physical addresses. Page
          tables function as before, but linear addresses are now translated
          to "guest physical" addresses first and not physical addresses
          directly. A new set of paging registers now exists under the
          traditional paging mechanism and translates from guest physical
          addresses to host physical addresses, which are used to access
          memory.</para>

          <para>Nested paging eliminates the overhead caused by VM exits and
          page table accesses. In essence, with nested page tables the guest
          can handle paging without intervention from the hypervisor. Nested
          paging thus significantly improves virtualization
          performance.</para>

          <para>On AMD processors, nested paging has been available starting
          with the Barcelona (K10) architecture; Intel added support for
          nested paging, which they call "extended page tables" (EPT), with
          their Core i7 (Nehalem) processors.</para>

          <para>If nested paging is enabled, the VirtualBox hypervisor can
          also use <emphasis role="bold">large pages</emphasis> to reduce TLB
          usage and overhead. This can yield a performance improvement of up
          to 5%. To enable this feature for a VM, you need to use the
          <computeroutput>VBoxManage modifyvm
          </computeroutput><computeroutput>--largepages</computeroutput>
          command; see <xref linkend="vboxmanage-modifyvm" />.</para>
        </listitem>

        <listitem>
          <para>On Intel CPUs, another hardware feature called <emphasis
          role="bold">"Virtual Processor Identifiers" (VPIDs)</emphasis> can
          greatly accelerate context switching by reducing the need for
          expensive flushing of the processor's Translation Lookaside Buffers
          (TLBs).</para>

          <para>To enable these features for a VM, you need to use the
          <computeroutput>VBoxManage modifyvm --vtxvpids</computeroutput> and
          <computeroutput>--largepages</computeroutput> commands; see <xref
          linkend="vboxmanage-modifyvm" />.</para>
        </listitem>
      </itemizedlist></para>
  </sect1>
</chapter>