source: vbox/trunk/doc/manual/en_US/user_AdvancedTopics.xml@ 37817

<?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="AdvancedTopics">
  <title>Advanced topics</title>

  <sect1 id="vboxsdl">
    <title>VBoxSDL, the simplified VM displayer</title>

    <sect2>
      <title>Introduction</title>

      <para>VBoxSDL is a simple graphical user interface (GUI) that lacks the
      nice point-and-click support which VirtualBox, our main GUI, provides.
      VBoxSDL is currently primarily used internally for debugging VirtualBox
      and therefore not officially supported. Still, you may find it useful
      for environments where the virtual machines are not necessarily
      controlled by the same person that uses the virtual machine.<note>
          <para>VBoxSDL is not available on the Mac OS X host platform.</para>
        </note></para>

      <para>As you can see in the following screenshot, VBoxSDL does indeed
      only provide a simple window that contains only the "pure" virtual
      machine, without menus or other controls to click upon and no additional
      indicators of virtual machine activity:</para>

      <para><mediaobject>
          <imageobject>
            <imagedata align="center" fileref="images/vbox-sdl.png"
                       width="10cm" />
          </imageobject>
        </mediaobject></para>

      <para>To start a virtual machine with VBoxSDL instead of the VirtualBox
      GUI, enter the following on a command line:<screen>VBoxSDL --startvm &lt;vm&gt;</screen></para>

      <para>where <computeroutput>&lt;vm&gt;</computeroutput> is, as usual
      with VirtualBox command line parameters, the name or UUID of an existing
      virtual machine.</para>
    </sect2>

    <sect2>
      <title>Secure labeling with VBoxSDL</title>

      <para>When running guest operating systems in fullscreen mode, the guest
      operating system usually has control over the whole screen. This could
      present a security risk as the guest operating system might fool the
      user into thinking that it is either a different system (which might
      have a higher security level) or it might present messages on the screen
      that appear to stem from the host operating system.</para>

      <para>In order to protect the user against the above mentioned security
      risks, the secure labeling feature has been developed. Secure labeling
      is currently available only for VBoxSDL. When enabled, a portion of the
      display area is reserved for a label in which a user defined message is
      displayed. The label height in set to 20 pixels in VBoxSDL. The label
      font color and background color can be optionally set as hexadecimal RGB
      color values. The following syntax is used to enable secure
      labeling:</para>

      <screen>VBoxSDL --startvm "VM name"
      --securelabel --seclabelfnt ~/fonts/arial.ttf
      --seclabelsiz 14 --seclabelfgcol 00FF00 --seclabelbgcol 00FFFF</screen>

      <para>In addition to enabling secure labeling, a TrueType font has to be
      supplied. To use another font size than 12 point use the parameter
      <computeroutput>--seclabelsiz</computeroutput>.</para>

      <para>The label text can be set with <screen>VBoxManage setextradata "VM name" "VBoxSDL/SecureLabel" "The Label"</screen>
      Changing this label will take effect immediately.</para>

      <para>Typically, full screen resolutions are limited to certain
      "standard" geometries such as 1024 x 768. Increasing this by twenty
      lines is not usually feasible, so in most cases, VBoxSDL will chose the
      next higher resolution, e.g. 1280 x 1024 and the guest's screen will not
      cover the whole display surface. If VBoxSDL is unable to choose a higher
      resolution, the secure label will be painted on top of the guest's
      screen surface. In order to address the problem of the bottom part of
      the guest screen being hidden, VBoxSDL can provide custom video modes to
      the guest that are reduced by the height of the label. For Windows
      guests and recent Solaris and Linux guests, the VirtualBox Guest
      Additions automatically provide the reduced video modes. Additionally,
      the VESA BIOS has been adjusted to duplicate its standard mode table
      with adjusted resolutions. The adjusted mode IDs can be calculated using
      the following formula:</para>

      <screen>reduced_modeid = modeid + 0x30</screen>

      <para>For example, in order to start Linux with 1024 x 748 x 16, the
      standard mode 0x117 (1024 x 768 x 16) is used as a base. The Linux video
      mode kernel parameter can then be calculated using:</para>

      <screen>vga = 0x200 | 0x117 + 0x30
vga = 839</screen>

      <para>The reason for duplicating the standard modes instead of only
      supplying the adjusted modes is that most guest operating systems
      require the standard VESA modes to be fixed and refuse to start with
      different modes.</para>

      <para>When using the X.org VESA driver, custom modelines have to be
      calculated and added to the configuration (usually in
      <literal>/etc/X11/xorg.conf</literal>. A handy tool to determine
      modeline entries can be found at <literal><ulink
      url="http://www.tkk.fi/Misc/Electronics/faq/vga2rgb/calc.html">http://www.tkk.fi/Misc/Electronics/faq/vga2rgb/calc.html</ulink></literal>.)</para>
    </sect2>

    <sect2>
      <title>Releasing modifiers with VBoxSDL on Linux</title>

      <para>When switching from a X virtual terminal (VT) to another VT using
      Ctrl-Alt-Fx while the VBoxSDL window has the input focus, the guest will
      receive Ctrl and Alt keypress events without receiving the corresponding
      key release events. This is an architectural limitation of Linux. In
      order to reset the modifier keys, it is possible to send
      <computeroutput>SIGUSR1</computeroutput> to the VBoxSDL main thread
      (first entry in the <computeroutput>ps</computeroutput> list). For
      example, when switching away to another VT and saving the virtual
      machine from this terminal, the following sequence can be used to make
      sure the VM is not saved with stuck modifiers:</para>

      <para><screen>kill -usr1 &lt;pid&gt;
VBoxManage controlvm "Windows 2000" savestate</screen></para>
    </sect2>
  </sect1>

  <sect1>
    <title id="autologon">Automated guest logons</title>

    <para>VirtualBox provides Guest Addition modules for Windows, Linux and
    Solaris to enable automated logons on the guest.</para>

    <para>When a guest operating system is running in a virtual machine, it
    might be desirable to perform coordinated and automated logons using
    credentials from a master logon system. (With "credentials", we are
    referring to logon information consisting of user name, password and
    domain name, where each value might be empty.)</para>

    <sect2 id="autologon_win">
      <title>Automated Windows guest logons</title>

      <para>Since Windows NT, Windows has provided a modular system logon
      subsystem ("Winlogon") which can be customized and extended by means of
      so-called GINA modules (Graphical Identification and Authentication).
      With Windows Vista and Windows 7, the GINA modules were replaced with a
      new mechanism called "credential providers". The VirtualBox Guest
      Additions for Windows come with both, a GINA and a credential provider
      module, and therefore enable any Windows guest to perform automated
      logons.</para>

      <para>To activate the VirtualBox GINA or credential provider module,
      install the Guest Additions with using the command line switch
      <computeroutput>/with_autologon</computeroutput>. All the following
      manual steps required for installing these modules will be then done by
      the installer.</para>

      <para>To manually install the VirtualBox GINA module, extract the Guest
      Additions (see <xref linkend="windows-guest-file-extraction" />) and
      copy the file <computeroutput>VBoxGINA.dll</computeroutput> to the
      Windows <computeroutput>SYSTEM32</computeroutput> directory. Then, in
      the registry, create the following key: <screen>HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Winlogon\GinaDLL</screen>
      with a value of <computeroutput>VBoxGINA.dll</computeroutput>.</para>

      <note>
        <para>The VirtualBox GINA module is implemented as a wrapper around
        the standard Windows GINA module
        (<computeroutput>MSGINA.DLL</computeroutput>). As a result, it will
        most likely not work correctly with 3rd party GINA modules.</para>
      </note>

      <para>To manually install the VirtualBox credential provider module, extract the
      Guest Additions (see <xref linkend="windows-guest-file-extraction" />)
      and copy the file <computeroutput>VBoxCredProv.dll</computeroutput> to
      the Windows <computeroutput>SYSTEM32</computeroutput> directory. Then,
      in the registry, create the following keys:<screen>HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\
           Authentication\Credential Providers\{275D3BCC-22BB-4948-A7F6-3A3054EBA92B}

HKEY_CLASSES_ROOT\CLSID\{275D3BCC-22BB-4948-A7F6-3A3054EBA92B}

HKEY_CLASSES_ROOT\CLSID\{275D3BCC-22BB-4948-A7F6-3A3054EBA92B}\InprocServer32</screen></para>

      <para>with all default values (the key named
      <computeroutput>(Default)</computeroutput> in each key) set to
      <computeroutput>VBoxCredProv</computeroutput>. After that a new string
      named <screen>HKEY_CLASSES_ROOT\CLSID\{275D3BCC-22BB-4948-A7F6-3A3054EBA92B}\InprocServer32\ThreadingModel</screen>
      with a value of <computeroutput>Apartment</computeroutput> has to be
      created.</para>

      <para>To set credentials, use the following command on a
      <emphasis>running</emphasis> VM:</para>

      <screen>VBoxManage controlvm "Windows XP" setcredentials "John Doe" "secretpassword" "DOMTEST"</screen>

      <para>While the VM is running, the credentials can be queried by the
      VirtualBox logon modules (GINA or credential provider) using the
      VirtualBox Guest Additions device driver. When Windows is in "logged
      out" mode, the logon modules will constantly poll for credentials and if
      they are present, a logon will be attempted. After retrieving the
      credentials, the logon modules will erase them so that the above command
      will have to be repeated for subsequent logons.</para>

      <para>For security reasons, credentials are not stored in any persistent
      manner and will be lost when the VM is reset. Also, the credentials are
      "write-only", i.e. there is no way to retrieve the credentials from the
      host side. Credentials can be reset from the host side by setting empty
      values.</para>

      <para>Depending on the particular variant of the Windows guest, the
      following restrictions apply: <orderedlist>
          <listitem>
            <para>For <emphasis role="bold">Windows XP guests,</emphasis> the
            logon subsystem needs to be configured to use the classic logon
            dialog as the VirtualBox GINA module does not support the XP-style
            welcome dialog.</para>
          </listitem>

          <listitem>
            <para>For <emphasis role="bold">Windows Vista and Windows 7
            guests,</emphasis> the logon subsystem does not support the
            so-called Secure Attention Sequence
            (<computeroutput>CTRL+ALT+DEL</computeroutput>). As a result, the
            guest's group policy settings need to be changed to not use the
            Secure Attention Sequence. Also, the user name given is only
            compared to the true user name, not the user friendly name. This
            means that when you rename a user, you still have to supply the
            original user name (internally, Windows never renames user
            accounts).</para>
          </listitem>

          <listitem>
            <para>Auto-logon handling of the built-in Windows Remote Desktop Service
              (formerly known as Terminal Services) is disabled by default. To enable
              it, create the registry key
              <screen>HKEY_LOCAL_MACHINE\SOFTWARE\Oracle\VirtualBox Guest Additions\AutoLogon</screen>
              with a <computeroutput>DWORD</computeroutput> value of <computeroutput>1</computeroutput>.</para>
          </listitem>
        </orderedlist></para>

      <para>The following command forces VirtualBox to keep the credentials
      after they were read by the guest and on VM reset: <screen>VBoxManage setextradata "Windows XP" VBoxInternal/Devices/VMMDev/0/Config/KeepCredentials 1</screen>Note
      that this is a potential security risk as a malicious application
      running on the guest could request this information using the proper
      interface.</para>
    </sect2>

    <sect2 id="autologon_unix">
      <title>Automated Linux/Unix guest logons</title>

      <para>Starting with version 3.2, VirtualBox provides a custom PAM module
      (Pluggable Authentication Module) which can be used to perform automated
      guest logons on platforms which support this framework. Virtually all
      modern Linux/Unix distributions rely on PAM.</para>

      <para>The <computeroutput>pam_vbox.so</computeroutput> module itself
      <emphasis role="bold">does not</emphasis> do an actual verification of
      the credentials passed to the guest OS; instead it relies on other
      modules such as <computeroutput>pam_unix.so</computeroutput> or
      <computeroutput>pam_unix2.so</computeroutput> down in the PAM stack to
      do the actual validation using the credentials retrieved by
      <computeroutput>pam_vbox.so</computeroutput>. Therefore
      <computeroutput>pam_vbox.so</computeroutput> has to be on top of the
      authentication PAM service list.</para>

      <note>
        <para>The <computeroutput>pam_vbox.so</computeroutput> only supports
        the <computeroutput>auth</computeroutput> primitive. Other primitives
        such as <computeroutput>account</computeroutput>,
        <computeroutput>session</computeroutput> or
        <computeroutput>password</computeroutput> are not supported.</para>
      </note>

      <para>The <computeroutput>pam_vbox.so</computeroutput> module is shipped
      as part of the Guest Additions but it is not installed and/or activated
      on the guest OS by default. In order to install it, it has to be copied
      from
      <computeroutput>/opt/VBoxGuestAdditions-&lt;version&gt;/lib/VBoxGuestAdditions/</computeroutput>
      to the security modules directory, usually
      <computeroutput>/lib/security/</computeroutput> on 32-bit guest Linuxes or
      <computeroutput>/lib64/security/</computeroutput> on 64-bit ones. Please refer to your
      guest OS documentation for the correct PAM module directory.</para>

      <para>For example, to use <computeroutput>pam_vbox.so</computeroutput>
      with a Ubuntu Linux guest OS and GDM (the GNOME Desktop Manager) to
      logon users automatically with the credentials passed by the host, the
      guest OS has to be configured like the following:</para>

      <orderedlist>
        <listitem>
          <para>The <computeroutput>pam_vbox.so</computeroutput> module has to
          be copied to the security modules directory, in this case it is
          <computeroutput>/lib/security</computeroutput>.</para>
        </listitem>

        <listitem>
          <para>Edit the PAM configuration file for GDM found at
          <computeroutput>/etc/pam.d/gdm</computeroutput>, adding the line
          <computeroutput>auth requisite pam_vbox.so</computeroutput> at the
          top. Additionaly, in most Linux distributions there is a file called
          <computeroutput>/etc/pam.d/common-auth</computeroutput>. This file
          is included in many other services (like the GDM file mentioned
          above). There you also have to add the line <computeroutput>auth
          requisite pam_vbox.so</computeroutput>.</para>
        </listitem>

        <listitem>
          <para>If authentication against the shadow database using
          <computeroutput>pam_unix.so</computeroutput> or
          <computeroutput>pam_unix2.so</computeroutput> is desired, the
          argument <computeroutput>try_first_pass</computeroutput> for
          <computeroutput>pam_unix.so</computeroutput> or
          <computeroutput>use_first_pass</computeroutput> for
          <computeroutput>pam_unix2.so</computeroutput> is needed
          in order to pass the credentials from the VirtualBox module to the
          shadow database authentication module. For Ubuntu, this needs to be
          added to <computeroutput>/etc/pam.d/common-auth</computeroutput>, to
          the end of the line referencing
          <computeroutput>pam_unix.so</computeroutput>. This argument tells
          the PAM module to use credentials already present in the stack, i.e.
          the ones provided by the VirtualBox PAM module.</para>
        </listitem>
      </orderedlist>

      <para><warning>
          <para>An incorrectly configured PAM stack can effectively prevent
          you from logging into your guest system!</para>
        </warning></para>

      <para>To make deployment easier, you can pass the argument
      <computeroutput>debug</computeroutput> right after the
      <computeroutput>pam_vbox.so</computeroutput> statement. Debug log output
      will then be recorded using syslog.</para>

      <para><warning>
          <para>At present, the GDM display manager only retrieves credentials
          at startup so unless the credentials have been supplied to the guest
          before GDM starts, automatic logon will not work. This limitation
          needs to be addressed by the GDM developers or another display
          manager must be used.</para>
        </warning></para>
    </sect2>
  </sect1>

  <sect1>
    <title>Advanced configuration for Windows guests</title>

    <sect2 id="sysprep">
      <title>Automated Windows system preparation</title>

      <para>Beginning with Windows NT 4.0, Microsoft offers a "system
      preparation" tool (in short: Sysprep) to prepare a Windows system for
      deployment or redistribution. Whereas Windows 2000 and XP ship with
      Sysprep on the installation medium, the tool also is available for
      download on the Microsoft web site. In a standard installation of
      Windows Vista and 7, Sysprep is already included. Sysprep mainly
      consists of an executable called
      <computeroutput>sysprep.exe</computeroutput> which is invoked by the
      user to put the Windows installation into preparation mode.</para>

      <para>Starting with VirtualBox 3.2.2, the Guest Additions offer a way to
      launch a system preparation on the guest operating system in an
      automated way, controlled from the host system. To achieve that, see
      <xref linkend="guestadd-guestcontrol" /> for using the feature with the
      special identifier <computeroutput>sysprep</computeroutput> as the
      program to execute, along with the user name
      <computeroutput>sysprep</computeroutput> and password
      <computeroutput>sysprep</computeroutput> for the credentials. Sysprep
      then gets launched with the required system rights.</para>

      <note>
        <para>Specifying the location of "sysprep.exe" is <emphasis
        role="bold">not possible</emphasis> -- instead the following paths are
        used (based on the operating system): <itemizedlist>
            <listitem>
              <para><computeroutput>C:\sysprep\sysprep.exe</computeroutput>
              for Windows NT 4.0, 2000 and XP</para>
            </listitem>

            <listitem>
              <para><computeroutput>%WINDIR%\System32\Sysprep\sysprep.exe</computeroutput>
              for Windows Vista, 2008 Server and 7</para>
            </listitem>
          </itemizedlist> The Guest Additions will automatically use the
        appropriate path to execute the system preparation tool.</para>
      </note>
    </sect2>
  </sect1>

  <sect1>
    <title>Advanced configuration for Linux and Solaris guests</title>

      <sect2>
        <title>Manual setup of selected guest services on Linux</title>

        <para>The VirtualBox Guest Additions contain several different
        drivers. If for any reason you do not wish to set them all up, you can
        install the Guest Additions using the following command:</para>

        <screen>  sh ./VBoxLinuxAdditions.run no_setup</screen>

        <para>After this, you will need to at least compile the kernel modules
        by running the command <screen>  /usr/lib/VBoxGuestAdditions/vboxadd setup</screen>
        as root (you will need to replace <emphasis>lib</emphasis> by
        <emphasis>lib64</emphasis> on some 64bit guests), and on older guests
        without the udev service you will need to add the
        <emphasis>vboxadd</emphasis> service to the default runlevel to ensure
        that the modules get loaded.</para>

        <para>To setup the time synchronization service, run the command
        <screen>  /usr/lib/VBoxGuestAdditions/vboxadd-service setup</screen>
        and add the service vboxadd-service to the default runlevel. To set up
        the X11 and OpenGL part of the Guest Additions, run the command
        <screen>  /usr/lib/VBoxGuestAdditions/vboxadd-x11 setup</screen> (you
        do not need to enable any services for this).</para>

        <para>To recompile the guest kernel modules, use this command:
        <screen>  /usr/lib/VBoxGuestAdditions/vboxadd setup</screen> After
        compilation you should reboot your guest to ensure that the new
        modules are actually used.</para>
      </sect2>

    <sect2 id="guestxorgsetup">
      <title>Guest graphics and mouse driver setup in depth</title>

        <para>This section assumes that you are familiar with configuring
        the X.Org server using xorg.conf and optionally the newer mechanisms
        using hal or udev and xorg.conf.d. If not you can learn about
        them by studying the documentation which comes with X.Org.</para>

        <para>The VirtualBox Guest Additions come with drivers for X.Org
        versions
          <itemizedlist>
            <listitem>X11R6.8/X11R6.9 and XFree86 version 4.3
            (vboxvideo_drv_68.o and vboxmouse_drv_68.o)</listitem>
            <listitem>X11R7.0 (vboxvideo_drv_70.so and vboxmouse_drv_70.so)
            </listitem>
            <listitem>X11R7.1 (vboxvideo_drv_71.so and vboxmouse_drv_71.so)
            </listitem>
            <listitem>X.Org Server versions 1.3 and later (vboxvideo_drv_13.so
            and vboxmouse_drv_13.so and so on).</listitem>
          </itemizedlist>
        By default these drivers can be found in the directory</para>
        <para>
        <computeroutput>/opt/VBoxGuestAdditions-&lt;version&gt;/lib/VBoxGuestAdditions</computeroutput>
        </para>
        <para>and the correct versions for the X server are symbolically linked
        into the X.Org driver directories.</para>

        <para>For graphics integration to work correctly, the X server must
        load the vboxvideo driver (many recent X server versions look for it
        automatically if they see that they are running in VirtualBox) and for
        an optimal user experience the guest kernel drivers must be loaded and
        the Guest Additions tool VBoxClient must be running as a client in the
        X session. For mouse integration to work correctly, the guest kernel
        drivers must be loaded and in addition, in X servers from X.Org X11R6.8
        to X11R7.1 and in XFree86 version 4.3 the right vboxmouse driver must
        be loaded and associated with /dev/mouse or /dev/psaux; in X.Org server
        1.3 or later a driver for a PS/2 mouse must be loaded and the right
        vboxmouse driver must be associated with /dev/vboxguest.</para>

        <para>The VirtualBox guest graphics driver can use any graphics
        configuration for which the virtual resolution fits into the virtual
        video memory allocated to the virtual machine (minus a small amount
        used by the guest driver) as described in
        <xref linkend="settings-display" />. The driver will offer a range of
        standard modes at least up to the default guest resolution for all
        active guest monitors.  In X.Org Server 1.3 and later the default mode
        can be changed by setting the output property VBOX_MODE to
        "&lt;width&gt;x&lt;height&gt;" for any guest monitor. When VBoxClient
        and the kernel drivers are active this is done automatically when the
        host requests a mode change. The driver for older versions can only
        receive new modes by querying the host for requests at regular
        intervals.</para>

        <para>With pre-1.3 X Servers you can also add your own modes to the X
        server configuration file. You simply need to add them to the "Modes"
        list in the "Display" subsection of the "Screen" section. For example,
        the section shown here has a custom 2048x800 resolution mode added:
        </para>

        <screen>Section "Screen"
        Identifier    "Default Screen"
        Device        "VirtualBox graphics card"
        Monitor       "Generic Monitor"
        DefaultDepth  24
        SubSection "Display"
                Depth         24
                Modes         "2048x800" "800x600" "640x480"
        EndSubSection
EndSection</screen>
    </sect2>
  </sect1>

  <sect1 id="cpuhotplug">
    <title>CPU hot-plugging</title>

    <para>With virtual machines running modern server operating systems,
    VirtualBox supports CPU hot-plugging.<footnote>
        <para>Support for CPU hot-plugging was introduced with VirtualBox
        3.2.</para>
      </footnote> Whereas on a physical computer this would mean that a CPU
    can be added or removed while the machine is running, VirtualBox supports
    adding and removing virtual CPUs while a virtual machine is
    running.</para>

    <para>CPU hot-plugging works only with guest operating systems that
    support it. So far this applies only to Linux and Windows Server 2008 x64
    Data Center Edition. Windows supports only hot-add while Linux supports
    hot-add and hot-remove but to use this feature with more than 8 CPUs a
    64bit Linux guest is required.</para>

    <para>At this time, CPU hot-plugging requires using the VBoxManage
    command-line interface. First, hot-plugging needs to be enabled for a
    virtual machine:<screen>VBoxManage modifyvm "VM name" --cpuhotplug on</screen></para>

    <para>After that, the --cpus option specifies the maximum number of CPUs
    that the virtual machine can have:<screen>VBoxManage modifyvm "VM name" --cpus 8</screen>When
    the VM is off, you can then add and remove virtual CPUs with the modifyvm
    --plugcpu and --unplugcpu subcommands, which take the number of the
    virtual CPU as a parameter, like this:<screen>VBoxManage modifyvm "VM name" --plugcpu 3
VBoxManage modifyvm "VM name" --unplugcpu 3</screen>Note that CPU 0 can never
    be removed.</para>

    <para>While the VM is running, CPUs can be added with the
    <computeroutput>controlvm plugcpu/unplugcpu</computeroutput> commands
    instead:<screen>VBoxManage controlvm "VM name" plugcpu 3
VBoxManage controlvm "VM name" unplugcpu 3</screen></para>

    <para>See <xref linkend="vboxmanage-modifyvm" /> and <xref
    linkend="vboxmanage-controlvm" /> for details.</para>

    <para>With Linux guests, the following applies: To prevent ejection while
    the CPU is still used it has to be ejected from within the guest before.
    The Linux Guest Additions contain a service which receives hot-remove
    events and ejects the CPU. Also, after a CPU is added to the VM it is not
    automatically used by Linux. The Linux Guest Additions service will take
    care of that if installed. If not a CPU can be started with the following
    command:<screen>echo 1 &gt; /sys/devices/system/cpu/cpu&lt;id&gt;/online</screen></para>
  </sect1>

  <sect1 id="pcipassthrough">
    <title>PCI passthrough</title>

    <para>When running on Linux hosts, with a recent enough kernel (at least version
      <computeroutput>2.6.31</computeroutput>) experimental host PCI devices
      passthrough is available.<footnote>
        <para>Experimental support for PCI passthrough was introduced with VirtualBox
        4.1.</para>
    </footnote> Essentially this feature allows to directly use physical PCI
      devices on the host by the guest even if host doesn't have drivers for this
      particular device. Both, regular PCI and some PCI Express cards, are
      supported. AGP and certain PCI Express cards are not supported at the
      moment if they rely on GART (Graphics Address Remapping Table) unit
      programming for texture management as it does rather nontrivial
      operations with pages remapping interfering with IOMMU.
      This limitation may be lifted in future releases.</para>

    <para>To be fully functional, PCI passthrough support in VirtualBox depends upon
      an IOMMU hardware unit which is not yet too widely available. If the device uses
      bus mastering (i.e. it performs DMA to the OS memory on its
      own), then an IOMMU is required, otherwise such DMA transactions may write to
      the wrong physical memory address as the device DMA engine is programmed using
      a device-specific protocol to perform memory transactions. The IOMMU functions
      as translation unit mapping physical memory access requests from the device
      using knowledge of the guest physical address to host physical addresses translation
      rules.</para>

    <para>Intel's solution for IOMMU is marketed as "Intel Virtualization Technology for
      Directed I/O" (VT-d), and AMD's one is called AMD-Vi. So please check if your
      motherboard datasheet has appropriate technology.
      Even if your hardware doesn't have a IOMMU, certain PCI cards may work
      (such as serial PCI adapters), but the guest will show a warning on boot and
      the VM execution will terminate if the guest driver will attempt to enable card
      bus mastering.</para>

    <para>
      It is very common that the BIOS or the host OS disables the IOMMU by default.
      So before any attempt to use it please make sure that
      <orderedlist>
        <listitem>
            Your motherboard has an IOMMU unit.
        </listitem>
        <listitem>
            Your CPU supports the IOMMU.
        </listitem>
        <listitem>
            The IOMMU is enabled in the BIOS.
        </listitem>
        <listitem>
            Your Linux kernel was compiled with IOMMU support (including DMA remapping, see
            <computeroutput>CONFIG_DMAR</computeroutput> kernel compilation option). The
            PCI stub driver (<computeroutput>CONFIG_PCI_STUB</computeroutput>) is required
            as well.
        </listitem>
        <listitem>
          Your Linux kernel recognizes and uses the IOMMU unit
          (<computeroutput>intel_iommu=on</computeroutput>
           boot option could be needed). Search for DMAR and PCI-DMA in kernel boot log.
        </listitem>
      </orderedlist>
    </para>

    <para>Once you made sure that the host kernel supports the IOMMU, the next step is
      to select the PCI card and attach it to the guest. To figure out the list of
      available PCI devices, use the <computeroutput>lspci</computeroutput> command.
      The output will look like this
      <screen>
        01:00.0 VGA compatible controller: ATI Technologies Inc Cedar PRO [Radeon HD 5450]
        01:00.1 Audio device: ATI Technologies Inc Manhattan HDMI Audio [Mobility Radeon HD 5000 Series]
        02:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168B PCI Express Gigabit Ethernet controller (rev 03)
        03:00.0 SATA controller: JMicron Technology Corp. JMB362/JMB363 Serial ATA Controller (rev 03)
        03:00.1 IDE interface: JMicron Technology Corp. JMB362/JMB363 Serial ATA Controller (rev 03)
        06:00.0 VGA compatible controller: nVidia Corporation G86 [GeForce 8500 GT] (rev a1)
      </screen>
      The first column is a PCI address (in format <computeroutput>bus:device.function</computeroutput>).
      This address could be used to identify the device for further operations.
      For example, to attach a PCI network controller on the system listed above
      to the second PCI bus in the guest, as device 5, function 0, use the following command:
      <screen>VBoxManage modifyvm "VM name" --pciattach 02:00.0@01:05.0</screen>
      To detach same device, use
      <screen>VBoxManage modifyvm "VM name" --pcidetach 02:00.0</screen>
      Please note that both host and guest could freely assign a different PCI address to
      the card attached during runtime, so those addresses only apply to the address of
      the card at the moment of attachment (host), and during BIOS PCI init (guest).
    </para>

    <para>If the virtual machine has a PCI device attached, certain limitations apply:
      <orderedlist>
         <listitem>
          Only PCI cards with non-shared interrupts (such as using MSI on host) are
          supported at the moment.
        </listitem>
        <listitem>
          No guest state can be reliably saved/restored (as the internal state of the PCI
          card could not be retrieved).
        </listitem>
        <listitem>
          Teleportation (live migration) doesn't work (for the same reason).
        </listitem>
        <listitem>
          No lazy physical memory allocation. The host will preallocate the whole RAM
          required for the VM on startup (as we cannot catch physical hardware accesses
          to the physical memory).
        </listitem>
      </orderedlist>
    </para>

  </sect1>


  <sect1>
    <title>Advanced display configuration</title>

    <sect2>
      <title>Custom VESA resolutions</title>

      <para>Apart from the standard VESA resolutions, the VirtualBox VESA BIOS
      allows you to add up to 16 custom video modes which will be reported to
      the guest operating system. When using Windows guests with the
      VirtualBox Guest Additions, a custom graphics driver will be used
      instead of the fallback VESA solution so this information does not
      apply.</para>

      <para>Additional video modes can be configured for each VM using the
      extra data facility. The extra data key is called
      <literal>CustomVideoMode&lt;x&gt;</literal> with <literal>x</literal>
      being a number from 1 to 16. Please note that modes will be read from 1
      until either the following number is not defined or 16 is reached. The
      following example adds a video mode that corresponds to the native
      display resolution of many notebook computers:</para>

      <screen>VBoxManage setextradata "VM name" "CustomVideoMode1" "1400x1050x16"</screen>

      <para>The VESA mode IDs for custom video modes start at
      <literal>0x160</literal>. In order to use the above defined custom video
      mode, the following command line has be supplied to Linux:</para>

      <screen>vga = 0x200 | 0x160
vga = 864</screen>

      <para>For guest operating systems with VirtualBox Guest Additions, a
      custom video mode can be set using the video mode hint feature.</para>
    </sect2>

    <sect2>
      <title>Configuring the maximum resolution of guests when using the
      graphical frontend</title>

      <para>When guest systems with the Guest Additions installed are started
      using the graphical frontend (the normal VirtualBox application), they
      will not be allowed to use screen resolutions greater than the host's
      screen size unless the user manually resizes them by dragging the
      window, switching to fullscreen or seamless mode or sending a video mode
      hint using VBoxManage. This behavior is what most users will want, but
      if you have different needs, it is possible to change it by issuing one
      of the following commands from the command line:</para>

      <screen>VBoxManage setextradata global GUI/MaxGuestResolution any</screen>

      <para>will remove all limits on guest resolutions.</para>

      <screen>VBoxManage setextradata global GUI/MaxGuestResolution &gt;width,height&lt;</screen>

      <para>manually specifies a maximum resolution.</para>

      <screen>VBoxManage setextradata global GUI/MaxGuestResolution auto</screen>

      <para>restores the default settings. Note that these settings apply
      globally to all guest systems, not just to a single machine.</para>
    </sect2>

  </sect1>

  <sect1>
    <title>Advanced storage configuration</title>

    <sect2 id="rawdisk">
      <title>Using a raw host hard disk from a guest</title>

      <para>Starting with version 1.4, as an alternative to using virtual disk
      images (as described in detail in <xref linkend="storage" />),
      VirtualBox can also present either entire physical hard disks or
      selected partitions thereof as virtual disks to virtual machines.</para>

      <para>With VirtualBox, this type of access is called "raw hard disk
      access"; it allows a guest operating system to access its virtual hard
      disk without going through the host OS file system. The actual
      performance difference for image files vs. raw disk varies greatly
      depending on the overhead of the host file system, whether dynamically
      growing images are used and on host OS caching strategies. The caching
      indirectly also affects other aspects such as failure behavior, i.e.
      whether the virtual disk contains all data written before a host OS
      crash. Consult your host OS documentation for details on this.</para>

      <para><warning>
          <para>Raw hard disk access is for expert users only. Incorrect use
          or use of an outdated configuration can lead to <emphasis
          role="bold">total loss of data </emphasis>on the physical disk. Most
          importantly, <emphasis>do not</emphasis> attempt to boot the
          partition with the currently running host operating system in a
          guest. This will lead to severe data corruption.</para>
        </warning></para>

      <para>Raw hard disk access -- both for entire disks and individual
      partitions -- is implemented as part of the VMDK image format support.
      As a result, you will need to create a special VMDK image file which
      defines where the data will be stored. After creating such a special
      VMDK image, you can use it like a regular virtual disk image. For
      example, you can use the Virtual Media Manager (<xref linkend="vdis" />)
      or <computeroutput>VBoxManage</computeroutput> to assign the image to a
      virtual machine.</para>

      <sect3>
        <title>Access to entire physical hard disk</title>

        <para>While this variant is the simplest to set up, you must be aware
        that this will give a guest operating system direct and full access to
        an <emphasis>entire physical disk</emphasis>. If your
        <emphasis>host</emphasis> operating system is also booted from this
        disk, please take special care to not access the partition from the
        guest at all. On the positive side, the physical disk can be
        repartitioned in arbitrary ways without having to recreate the image
        file that gives access to the raw disk.</para>

        <para>To create an image that represents an entire physical hard disk
        (which will not contain any actual data, as this will all be stored on
        the physical disk), on a Linux host, use the command<screen>VBoxManage internalcommands createrawvmdk -filename /path/to/file.vmdk
      -rawdisk /dev/sda</screen>This creates the image
        <code>/path/to/file.vmdk</code> (must be absolute), and all data will
        be read and written from <code>/dev/sda</code>.</para>

        <para>On a Windows host, instead of the above device specification,
        use e.g. <code>\\.\PhysicalDrive0</code>. On a Mac OS X host, instead
        of the above device specification use e.g. <code>/dev/disk1</code>.
        Note that on OS X you can only get access to an entire disk if no
        volume is mounted from it.</para>

        <para>Creating the image requires read/write access for the given
        device. Read/write access is also later needed when using the image
        from a virtual machine.</para>

        <para>Just like with regular disk images, this does not automatically
        attach the newly created image to a virtual machine. This can be done
        with e.g. <screen>VBoxManage storageattach WindowsXP --storagectl "IDE Controller"
      --port 0 --device 0 --type hdd --medium /path/to/file.vmdk</screen>When
        this is done the selected virtual machine will boot from the specified
        physical disk.</para>
      </sect3>

      <sect3>
        <title>Access to individual physical hard disk partitions</title>

        <para>This "raw partition support" is quite similar to the "full hard
        disk" access described above. However, in this case, any partitioning
        information will be stored inside the VMDK image, so you can e.g.
        install a different boot loader in the virtual hard disk without
        affecting the host's partitioning information. While the guest will be
        able to <emphasis>see</emphasis> all partitions that exist on the
        physical disk, access will be filtered in that reading from partitions
        for which no access is allowed the partitions will only yield zeroes,
        and all writes to them are ignored.</para>

        <para>To create a special image for raw partition support (which will
        contain a small amount of data, as already mentioned), on a Linux
        host, use the command<screen>VBoxManage internalcommands createrawvmdk -filename /path/to/file.vmdk
      -rawdisk /dev/sda -partitions 1,5</screen></para>

        <para>As you can see, the command is identical to the one for "full
        hard disk" access, except for the additional
        <computeroutput>-partitions</computeroutput> parameter. This example
        would create the image <code>/path/to/file.vmdk</code> (which, again,
        must be absolute), and partitions 1 and 5 of <code>/dev/sda</code>
        would be made accessible to the guest.</para>

        <para>VirtualBox uses the same partition numbering as your Linux host.
        As a result, the numbers given in the above example would refer to the
        first primary partition and the first logical drive in the extended
        partition, respectively.</para>

        <para>On a Windows host, instead of the above device specification,
        use e.g. <code>\\.\PhysicalDrive0</code>. On a Mac OS X host, instead
        of the above device specification use e.g. <code>/dev/disk1</code>.
        Note that on OS X you can only use partitions which are not mounted
        (eject the respective volume first). Partition numbers are the same on
        Linux, Windows and Mac OS X hosts.</para>

        <para>The numbers for the list of partitions can be taken from the
        output of<screen>VBoxManage internalcommands listpartitions -rawdisk /dev/sda</screen>The
        output lists the partition types and sizes to give the user enough
        information to identify the partitions necessary for the guest.</para>

        <para>Images which give access to individual partitions are specific
        to a particular host disk setup. You cannot transfer these images to
        another host; also, whenever the host partitioning changes, the image
        <emphasis>must be recreated</emphasis>.</para>

        <para>Creating the image requires read/write access for the given
        device. Read/write access is also later needed when using the image
        from a virtual machine. If this is not feasible, there is a special
        variant for raw partition access (currently only available on Linux
        hosts) that avoids having to give the current user access to the
        entire disk. To set up such an image, use<screen>VBoxManage internalcommands createrawvmdk -filename /path/to/file.vmdk
      -rawdisk /dev/sda -partitions 1,5 -relative</screen>When used from a
        virtual machine, the image will then refer not to the entire disk, but
        only to the individual partitions (in the example
        <code>/dev/sda1</code> and <code>/dev/sda5</code>). As a consequence,
        read/write access is only required for the affected partitions, not
        for the entire disk. During creation however, read-only access to the
        entire disk is required to obtain the partitioning information.</para>

        <para>In some configurations it may be necessary to change the MBR
        code of the created image, e.g. to replace the Linux boot loader that
        is used on the host by another boot loader. This allows e.g. the guest
        to boot directly to Windows, while the host boots Linux from the
        "same" disk. For this purpose the
        <computeroutput>-mbr</computeroutput> parameter is provided. It
        specifies a file name from which to take the MBR code. The partition
        table is not modified at all, so a MBR file from a system with totally
        different partitioning can be used. An example of this is<screen>VBoxManage internalcommands createrawvmdk -filename /path/to/file.vmdk
      -rawdisk /dev/sda -partitions 1,5 -mbr winxp.mbr</screen>The modified
        MBR will be stored inside the image, not on the host disk.</para>

        <para>The created image can be attached to a storage controller in
        a VM configuration as usual.</para>
      </sect3>
    </sect2>

    <sect2 id="changevpd">
      <title>Configuring the hard disk vendor product data (VPD)</title>

      <para>VirtualBox reports vendor product data for its virtual hard disks
      which consist of hard disk serial number, firmware revision and model
      number. These can be changed using the following commands:</para>

      <screen>VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/ahci/0/Config/Port0/SerialNumber" "serial"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/ahci/0/Config/Port0/FirmwareRevision" "firmware"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/ahci/0/Config/Port0/ModelNumber" "model"</screen>

      <para>The serial number is a 20 byte alphanumeric string, the firmware
      revision an 8 byte alphanumeric string and the model number a 40 byte
      alphanumeric string. Instead of "Port0" (referring to the first port),
      specify the desired SATA hard disk port.</para>

      <para>The above commands apply to virtual machines with an AHCI (SATA)
      controller. The commands for virtual machines with an IDE controller
      are:</para>

      <screen>VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/piix3ide/0/Config/PrimaryMaster/SerialNumber" "serial"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/piix3ide/0/Config/PrimaryMaster/FirmwareRevision" "firmware"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/piix3ide/0/Config/PrimaryMaster/ModelNumber" "model"</screen>

      <para>For hard disks it's also possible (experimental!) to mark the drive
      as having a non-rotational medium with:</para>

      <screen>VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/ahci/0/Config/Port0/NonRotational" "1"</screen>

      <para>Additional three parameters are needed for CD/DVD drives to report
      the vendor product data:</para>

      <screen>VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/ahci/0/Config/Port0/ATAPIVendorId" "vendor"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/ahci/0/Config/Port0/ATAPIProductId" "product"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/ahci/0/Config/Port0/ATAPIRevision" "revision"</screen>

      <para>The vendor id is an 8 byte alphanumeric string, the product id an
      16 byte alphanumeric string and the revision a 4 byte alphanumeric
      string. Instead of "Port0" (referring to the first port), specify the
      desired SATA hard disk port.</para>
    </sect2>

    <sect2>
      <title id="iscsi-intnet">Access iSCSI targets via Internal
      Networking</title>

      <para>As an experimental feature, VirtualBox allows for accessing an
      iSCSI target running in a virtual machine which is configured for using
      Internal Networking mode. Please see <xref linkend="storage-iscsi" />;
      <xref linkend="network_internal" />; and <xref
      linkend="vboxmanage-storageattach" /> for additional information.</para>

      <para>The IP stack accessing Internal Networking must be configured in
      the virtual machine which accesses the iSCSI target. A free static IP
      and a MAC address not used by other virtual machines must be chosen. In
      the example below, adapt the name of the virtual machine, the MAC
      address, the IP configuration and the Internal Networking name
      ("MyIntNet") according to your needs. The following seven commands must
      first be issued:<screen>VBoxManage setextradata "VM name" VBoxInternal/Devices/IntNetIP/0/Trusted 1
VBoxManage setextradata "VM name" VBoxInternal/Devices/IntNetIP/0/Config/MAC 08:00:27:01:02:0f
VBoxManage setextradata "VM name" VBoxInternal/Devices/IntNetIP/0/Config/IP 10.0.9.1
VBoxManage setextradata "VM name" VBoxInternal/Devices/IntNetIP/0/Config/Netmask 255.255.255.0
VBoxManage setextradata "VM name" VBoxInternal/Devices/IntNetIP/0/LUN#0/Driver IntNet
VBoxManage setextradata "VM name" VBoxInternal/Devices/IntNetIP/0/LUN#0/Config/Network MyIntNet
VBoxManage setextradata "VM name" VBoxInternal/Devices/IntNetIP/0/LUN#0/Config/IsService 1</screen></para>

      <para>Finally the iSCSI disk must be attached with the
      <computeroutput>--intnet</computeroutput> option to tell the iSCSI
      initiator to use internal networking:<screen>VBoxManage storageattach ... --medium iscsi
         --server 10.0.9.30 --target iqn.2008-12.com.sun:sampletarget --intnet</screen></para>

      <para>Compared to a "regular" iSCSI setup, IP address of the target
      <emphasis>must</emphasis> be specified as a numeric IP address, as there
      is no DNS resolver for internal networking.</para>

      <para>The virtual machine with the iSCSI target should be started before
      the VM using it is powered on. If a virtual machine using an iSCSI disk
      is started without having the iSCSI target powered up, it can take up to
      200 seconds to detect this situation. The VM will fail to power
      up.</para>
    </sect2>
  </sect1>

  <sect1>
    <title>Launching more than 120 VMs on Solaris hosts</title>

    <para>Solaris hosts have a fixed number of IPC semaphores IDs per process
    preventing users from starting more than 120 VMs. While trying to launch
    more VMs you would be shown a "Cannot create IPC semaphore" error.</para>

    <para>In order to run more VMs, you will need to bump the semaphore ID
    limit of the VBoxSVC process. Execute as root the
    <computeroutput>prctl</computeroutput> command as shown below. The process
    ID of VBoxSVC can be obtained using the
    <computeroutput>ps</computeroutput> list command.</para>

    <para><screen>prctl -r -n project.max-sem-ids -v 2048 &lt;pid-of-VBoxSVC&gt;</screen></para>
  </sect1>

  <sect1>
    <title>Legacy commands for using serial ports</title>

    <para>Starting with version 1.4, VirtualBox provided support for virtual
    serial ports, which, at the time, was rather complicated to set up with a
    sequence of <computeroutput>VBoxManage setextradata</computeroutput>
    statements. Since version 1.5, that way of setting up serial ports is no
    longer necessary and <emphasis>deprecated.</emphasis> To set up virtual
    serial ports, use the methods now described in <xref
    linkend="serialports" />.<note>
        <para>For backwards compatibility, the old
        <computeroutput>setextradata</computeroutput> statements, whose
        description is retained below from the old version of the manual, take
        <emphasis>precedence</emphasis> over the new way of configuring serial
        ports. As a result, if configuring serial ports the new way doesn't
        work, make sure the VM in question does not have old configuration
        data such as below still active.</para>
      </note></para>

    <para>The old sequence of configuring a serial port used the following 6
    commands:</para>

    <screen>VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/serial/0/Config/IRQ" 4
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/serial/0/Config/IOBase" 0x3f8
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/serial/0/LUN#0/Driver" Char
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/serial/0/LUN#0/AttachedDriver/Driver" NamedPipe
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/serial/0/LUN#0/AttachedDriver/Config/Location" "\\.\pipe\vboxCOM1"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/serial/0/LUN#0/AttachedDriver/Config/IsServer" 1</screen>

    <para>This sets up a serial port in the guest with the default settings
    for COM1 (IRQ 4, I/O address 0x3f8) and the
    <computeroutput>Location</computeroutput> setting assumes that this
    configuration is used on a Windows host, because the Windows named pipe
    syntax is used. Keep in mind that on Windows hosts a named pipe must
    always start with <computeroutput>\\.\pipe\</computeroutput>. On Linux the
    same config settings apply, except that the path name for the
    <computeroutput>Location</computeroutput> can be chosen more freely. Local
    domain sockets can be placed anywhere, provided the user running
    VirtualBox has the permission to create a new file in the directory. The
    final command above defines that VirtualBox acts as a server, i.e. it
    creates the named pipe itself instead of connecting to an already existing
    one.</para>
  </sect1>

  <sect1 id="changenat">
    <title>Fine-tuning the VirtualBox NAT engine</title>

    <sect2>
      <title>Configuring the address of a NAT network interface</title>

      <para>In NAT mode, the guest network interface is assigned to the IPv4
      range <computeroutput>10.0.x.0/24</computeroutput> by default where
      <computeroutput>x</computeroutput> corresponds to the instance of the
      NAT interface +2. So <computeroutput>x</computeroutput> is 2 when there
      is only one NAT instance active. In that case the guest is assigned to
      the address <computeroutput>10.0.2.15</computeroutput>, the gateway is
      set to <computeroutput>10.0.2.2</computeroutput> and the name server can
      be found at <computeroutput>10.0.2.3</computeroutput>.</para>

      <para>If, for any reason, the NAT network needs to be changed, this can
      be achieved with the following command:</para>

      <screen>VBoxManage modifyvm "VM name" --natnet1 "192.168/16"</screen>

      <para>This command would reserve the network addresses from
      <computeroutput>192.168.0.0</computeroutput> to
      <computeroutput>192.168.254.254</computeroutput> for the first NAT
      network instance of "VM name". The guest IP would be assigned to
      <computeroutput>192.168.0.15</computeroutput> and the default gateway
      could be found at <computeroutput>192.168.0.2</computeroutput>.</para>
    </sect2>

    <sect2 id="nat-adv-tftp">
      <title>Configuring the boot server (next server) of a NAT network
      interface</title>

      <para>For network booting in NAT mode, by default VirtualBox uses a
      built-in TFTP server at the IP address 10.0.2.3. This default behavior
      should work fine for typical remote-booting scenarios. However, it is
      possible to change the boot server IP and the location of the boot image
      with the following commands: <screen>VBoxManage modifyvm "VM name" --nattftpserver1 10.0.2.2
VBoxManage modifyvm "VM name" --nattftpfile1 /srv/tftp/boot/MyPXEBoot.pxe</screen></para>
    </sect2>

    <sect2 id="nat-adv-settings">
      <title>Tuning TCP/IP buffers for NAT</title>

      <para>The VirtualBox NAT stack performance is often determined by its
      interaction with the host's TCP/IP stack and the size of several buffers
      (<computeroutput>SO_RCVBUF</computeroutput> and
      <computeroutput>SO_SNDBUF</computeroutput>). For certain setups users
      might want to adjust the buffer size for a better performance. This can
      by achieved using the following commands (values are in kilobytes and
      can range from 8 to 1024): <screen>VBoxManage modifyvm "VM name" --natsettings1 16000,128,128,0,0</screen>
      This example illustrates tuning the NAT settings. The first parameter is
      the MTU, then the size of the socket's send buffer and the size of the
      socket's receive buffer, the initial size of the TCP send window, and
      lastly the initial size of the TCP receive window. Note that specifying
      zero means fallback to the default value.</para>

      <para>Each of these buffers has a default size of 64KB and default MTU
      is 1500.</para>
    </sect2>

    <sect2>
      <title>Binding NAT sockets to a specific interface</title>

      <para>By default, VirtualBox's NAT engine will route TCP/IP packets
      through the default interface assigned by the host's TCP/IP stack. (The
      technical reason for this is that the NAT engine uses sockets for
      communication.) If, for some reason, you want to change this behavior,
      you can tell the NAT engine to bind to a particular IP address instead.
      Use the following command: <screen>VBoxManage modifyvm "VM name" --natbindip1 "10.45.0.2"</screen></para>

      <para>After this, all outgoing traffic will be sent through the
      interface with the IP address 10.45.0.2. Please make sure that this
      interface is up and running prior to this assignment.</para>
    </sect2>

    <sect2 id="nat-adv-dns">
      <title>Enabling DNS proxy in NAT mode</title>

      <para>The NAT engine by default offers the same DNS servers to the guest
      that are configured on the host. In some scenarios, it can be desirable
      to hide the DNS server IPs from the guest, for example when this
      information can change on the host due to expiring DHCP leases. In this
      case, you can tell the NAT engine to act as DNS proxy using the
      following command: <screen>VBoxManage modifyvm "VM name" --natdnsproxy1 on</screen></para>
    </sect2>

    <sect2 id="nat_host_resolver_proxy">
      <title>Using the host's resolver as a DNS proxy in NAT mode</title>

      <para>For resolving network names, the DHCP server of the NAT engine
      offers a list of registered DNS servers of the host. If for some reason
      you need to hide this DNS server list and use the host's resolver
      settings, thereby forcing the VirtualBox NAT engine to intercept DNS
      requests and forward them to host's resolver, use the following command:
      <screen>VBoxManage modifyvm "VM name" --natdnshostresolver1 on</screen>
      Note that this setting is similar to the DNS proxy mode, however whereas
      the proxy mode just forwards DNS requests to the appropriate servers,
      the resolver mode will interpret the DNS requests and use the host's DNS
      API to query the information and return it to the guest.</para>
    </sect2>

    <sect2 id="nat-adv-alias">
      <title>Configuring aliasing of the NAT engine</title>

      <para>By default, the NAT core uses aliasing and uses random ports when
      generating an alias for a connection. This works well for the most
      protocols like SSH, FTP and so on. Though some protocols might need a
      more transparent behavior or may depend on the real port number the
      packet was sent from. It is possible to change the NAT mode via the
      VBoxManage frontend with the following commands: <screen>VBoxManage modifyvm "VM name" --nataliasmode proxyonly</screen>
      and <screen>VBoxManage modifyvm "Linux Guest" --nataliasmode sameports</screen>
      The first example disables aliasing and switches NAT into transparent
      mode, the second example enforces preserving of port values. These modes
      can be combined if necessary.</para>
    </sect2>
  </sect1>

  <sect1 id="changedmi">
    <title>Configuring the BIOS DMI information</title>

    <para>The DMI data VirtualBox provides to guests can be changed for a
    specific VM. Use the following commands to configure the DMI BIOS
    information:</para>

    <screen>VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiBIOSVendor"        "BIOS Vendor"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiBIOSVersion"       "BIOS Version"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiBIOSReleaseDate"   "BIOS Release Date"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiBIOSReleaseMajor"  1
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiBIOSReleaseMinor"  2
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiBIOSFirmwareMajor" 3
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiBIOSFirmwareMinor" 4
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiSystemVendor"      "System Vendor"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiSystemProduct"     "System Product"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiSystemVersion"     "System Version"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiSystemSerial"      "System Serial"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiSystemSKU"         "System SKU"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiSystemFamily"      "System Family"
VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiSystemUuid"
                                               "9852bf98-b83c-49db-a8de-182c42c7226b"</screen>

    <para>If a DMI string is not set, the default value of VirtualBox is used.
    To set an empty string use
    <computeroutput>"&lt;EMPTY&gt;"</computeroutput>.</para>

    <para>Note that in the above list, all quoted parameters (DmiBIOSVendor,
    DmiBIOSVersion but not DmiBIOSReleaseMajor) are expected to be strings. If
    such a string is a valid number, the parameter is treated as number and
    the VM will most probably refuse to start with an
    <computeroutput>VERR_CFGM_NOT_STRING</computeroutput> error. In that case,
    use <computeroutput>"string:&lt;value&gt;"</computeroutput>, for instance
    <screen>VBoxManage setextradata "VM name"
      "VBoxInternal/Devices/pcbios/0/Config/DmiSystemSerial"      "string:1234"</screen></para>

    <para>Changing this information can be necessary to provide the DMI
    information of the host to the guest to prevent Windows from asking for a
    new product key. On Linux hosts the DMI BIOS information can be obtained
    with <screen>dmidecode -t0</screen>and the DMI system information can be
    obtained with <screen>dmidecode -t1</screen></para>
  </sect1>

  <sect1>
    <title>Fine-tuning timers and time synchronization</title>

    <sect2 id="changetscmode">
      <title>Configuring the guest time stamp counter (TSC) to reflect guest
      execution</title>

      <para>By default, VirtualBox keeps all sources of time visible to the
      guest synchronized to a single time source, the monotonic host time.
      This reflects the assumptions of many guest operating systems, which
      expect all time sources to reflect "wall clock" time. In special
      circumstances it may be useful however to make the TSC (time stamp
      counter) in the guest reflect the time actually spent executing the
      guest.</para>

      <para>This special TSC handling mode can be enabled on a per-VM basis,
      and for best results must be used only in combination with hardware
      virtualization. To enable this mode use the following command:</para>

      <screen>VBoxManage setextradata "VM name" "VBoxInternal/TM/TSCTiedToExecution" 1</screen>

      <para>To revert to the default TSC handling mode use:</para>

      <screen>VBoxManage setextradata "VM name" "VBoxInternal/TM/TSCTiedToExecution"</screen>

      <para>Note that if you use the special TSC handling mode with a guest
      operating system which is very strict about the consistency of time
      sources you may get a warning or error message about the timing
      inconsistency. It may also cause clocks to become unreliable with some
      guest operating systems depending on they use the TSC.</para>
    </sect2>

    <sect2 id="warpguest">
      <title>Accelerate or slow down the guest clock</title>

      <para>For certain purposes it can be useful to accelerate or to slow
      down the (virtual) guest clock. This can be achieved as follows:</para>

      <screen>VBoxManage setextradata "VM name" "VBoxInternal/TM/WarpDrivePercentage" 200</screen>

      <para>The above example will double the speed of the guest clock
      while</para>

      <screen>VBoxManage setextradata "VM name" "VBoxInternal/TM/WarpDrivePercentage" 50</screen>

      <para>will halve the speed of the guest clock. Note that changing the
      rate of the virtual clock can confuse the guest and can even lead to
      abnormal guest behavior. For instance, a higher clock rate means shorter
      timeouts for virtual devices with the result that a slightly increased
      response time of a virtual device due to an increased host load can
      cause guest failures. Note further that any time synchronization
      mechanism will frequently try to resynchronize the guest clock with the
      reference clock (which is the host clock if the VirtualBox Guest
      Additions are active). Therefore any time synchronization should be
      disabled if the rate of the guest clock is changed as described above
      (see <xref linkend="changetimesync" />).</para>
    </sect2>

    <sect2 id="changetimesync">
      <title>Tuning the Guest Additions time synchronization
      parameters</title>

      <para>The VirtualBox Guest Additions ensure that the guest's system time
      is synchronized with the host time. There are several parameters which
      can be tuned. The parameters can be set for a specific VM using the
      following command:</para>

      <screen>VBoxManage guestproperty set VM_NAME "/VirtualBox/GuestAdd/VBoxService/PARAMETER" VALUE</screen>

      <para>where <computeroutput>PARAMETER</computeroutput> is one of the
      following:</para>

      <para><glosslist>
          <glossentry>
            <glossterm><computeroutput>--timesync-interval</computeroutput></glossterm>

            <glossdef>
              <para>Specifies the interval at which to synchronize the time
              with the host. The default is 10000 ms (10 seconds).</para>
            </glossdef>
          </glossentry>

          <glossentry>
            <glossterm><computeroutput>--timesync-min-adjust</computeroutput></glossterm>

            <glossdef>
              <para>The minimum absolute drift value measured in milliseconds
              to make adjustments for. The default is 1000 ms on OS/2 and 100
              ms elsewhere.</para>
            </glossdef>
          </glossentry>

          <glossentry>
            <glossterm><computeroutput>--timesync-latency-factor</computeroutput></glossterm>

            <glossdef>
              <para>The factor to multiply the time query latency with to
              calculate the dynamic minimum adjust time. The default is 8
              times, that means in detail: Measure the time it takes to
              determine the host time (the guest has to contact the VM host
              service which may take some time), multiply this value by 8 and
              do an adjustment only if the time difference between host and
              guest is bigger than this value. Don't do any time adjustment
              otherwise.</para>
            </glossdef>
          </glossentry>

          <glossentry>
            <glossterm><computeroutput>--timesync-max-latency</computeroutput></glossterm>

            <glossdef>
              <para>The max host timer query latency to accept. The default is
              250 ms.</para>
            </glossdef>
          </glossentry>

          <glossentry>
            <glossterm><computeroutput>--timesync-set-threshold</computeroutput></glossterm>

            <glossdef>
              <para>The absolute drift threshold, given as milliseconds where
              to start setting the time instead of trying to smoothly adjust
              it. The default is 20 minutes.</para>
            </glossdef>
          </glossentry>

          <glossentry>
            <glossterm><computeroutput>--timesync-set-start</computeroutput></glossterm>

            <glossdef>
              <para>Set the time when starting the time sync service.</para>
            </glossdef>
          </glossentry>

          <glossentry>
            <glossterm><computeroutput>--timesync-set-on-restore
            0|1</computeroutput></glossterm>

            <glossdef>
              <para>Set the time after the VM was restored from a saved state
              when passing 1 as parameter (default). Disable by passing 0. In
              the latter case, the time will be adjusted smoothly which can
              take a long time.</para>
            </glossdef>
          </glossentry>
        </glosslist></para>

      <para>All these parameters can be specified as command line parameters
      to VBoxService as well.</para>
    </sect2>
  </sect1>

  <sect1 id="addhostonlysolaris">
    <title>Configuring multiple host-only network interfaces on Solaris
    hosts</title>

    <para>By default VirtualBox provides you with one host-only network
    interface. Adding more host-only network interfaces on Solaris hosts
    requires manual configuration. Here's how to add two more host-only
    network interfaces.</para>

    <para>You first need to stop all running VMs and unplumb all existing
    "vboxnet" interfaces. Execute the following commands as root:</para>

    <screen>ifconfig vboxnet0 unplumb</screen>

    <para>Once you make sure all vboxnet interfaces are unplumbed, remove the
    driver using:</para>

    <para><screen>rem_drv vboxnet</screen>then edit the file
    <computeroutput>/platform/i86pc/kernel/drv/vboxnet.conf</computeroutput>
    and add a line for the new interfaces:</para>

    <para><screen>name="vboxnet" parent="pseudo" instance=1;
name="vboxnet" parent="pseudo" instance=2;</screen>Add as many of these lines
    as required and make sure "instance" number is uniquely incremented. Next
    reload the vboxnet driver using:</para>

    <para><screen>add_drv vboxnet</screen>Now plumb all the interfaces using
    <computeroutput>ifconfig vboxnetX plumb</computeroutput> (where X can be
    0, 1 or 2 in this case) and once plumbed you can then configure the
    interface like any other network interface.</para>

    <para>To make your newly added interfaces' settings persistent across
    reboots you will need to edit the files
    <computeroutput>/etc/netmasks</computeroutput>, and if you are using NWAM
    <computeroutput>/etc/nwam/llp</computeroutput> and add the appropriate
    entries to set the netmask and static IP for each of those interfaces. The
    VirtualBox installer only updates these configuration files for the one
    "vboxnet0" interface it creates by default.</para>
  </sect1>

  <sect1 id="solariscodedumper">
    <title>Configuring the VirtualBox CoreDumper on Solaris hosts</title>

    <para>VirtualBox is capable of producing its own core files when things go
    wrong and for more extensive debugging. Currently this is only available
    on Solaris hosts.</para>

    <para>The VirtualBox CoreDumper can be enabled using the following
    command:</para>

    <para><screen>VBoxManage setextradata "VM name" VBoxInternal2/CoreDumpEnabled 1</screen></para>

    <para>You can specify which directory to use for core dumps with this
    command:</para>

    <para><screen>VBoxManage setextradata "VM name" VBoxInternal2/CoreDumpDir &lt;path-to-directory&gt;</screen>Make
    sure the directory you specify is on a volume with sufficient free space
    and that the VirtualBox process has sufficient permissions to write files
    to this directory. If you skip this command and don't specify any core
    dump directory, the current directory of the VirtualBox executable will be
    used (which would most likely fail when writing cores as they are
    protected with root permissions). It is recommended you explicity set a
    core dump directory.</para>

    <para>You must specify when the VirtualBox CoreDumper should be triggered.
    This is done using the following commands:</para>

    <para><screen>VBoxManage setextradata "VM name" VBoxInternal2/CoreDumpReplaceSystemDump 1
VBoxManage setextradata "VM name" VBoxInternal2/CoreDumpLive 1</screen>At
    least one of the above two commands will have to be provided if you have
    enabled the VirtualBox CoreDumper.</para>

    <para>Setting <computeroutput>CoreDumpReplaceSystemDump</computeroutput>
    sets up the VM to override the host's core dumping mechanism and in the
    event of any crash only the VirtualBox CoreDumper would produce the core
    file.</para>

    <para>Setting <computeroutput>CoreDumpLive</computeroutput> sets up the VM
    to produce cores whenever the VM receives a
    <computeroutput>SIGUSR2</computeroutput> signal. After producing the core
    file, the VM will not be terminated and will continue to run. You can then
    take cores of the VM process using:</para>

    <para><screen>kill -s SIGUSR2 &lt;VM-process-id&gt;</screen></para>

    <para>Core files produced by the VirtualBox CoreDumper are of the form
    <computeroutput>core.vb.&lt;ProcessName&gt;.&lt;ProcessID&gt;</computeroutput>,
    e.g.<computeroutput>core.vb.VBoxHeadless.11321</computeroutput>.</para>
  </sect1>

  <sect1 id="guitweaks">
    <title>Locking down the VirtualBox manager GUI</title>

    <para>There are several advanced customization settings for locking down
    the VirtualBox manager, that is, removing some features that the user
    should not see.<screen>VBoxManage setextradata global GUI/Customizations OPTION[,OPTION...]</screen></para>

    <para>where <computeroutput>OPTION</computeroutput> is one of the
    following keywords:<glosslist>
        <glossentry>
          <glossterm><computeroutput>noSelector</computeroutput></glossterm>

          <glossdef>
            <para>Don't allow to start the VirtualBox manager. Trying to do so
            will show a window containing a proper error message.</para>
          </glossdef>
        </glossentry>

        <glossentry>
          <glossterm><computeroutput>noMenuBar</computeroutput></glossterm>

          <glossdef>
            <para>VM windows will not contain a menu bar.</para>
          </glossdef>
        </glossentry>

        <glossentry>
          <glossterm><computeroutput>noStatusBar</computeroutput></glossterm>

          <glossdef>
            <para>VM windows will not contain a status bar.</para>
          </glossdef>
        </glossentry>
      </glosslist></para>

    <para>To disable any GUI customization do <screen>VBoxManage setextradata global GUI/Customizations</screen></para>

    <para>To disable all host key combinations, open the preferences and
    change the host key to <emphasis>None</emphasis>. This might be useful
    when using VirtualBox in a kiosk mode.</para>

    <para>Furthermore, you can disallow certain actions when terminating a VM.
    To disallow specific actions, type:</para>

    <para><screen>VBoxManage setextradata "VM name" GUI/RestrictedCloseActions OPTION[,OPTION...]</screen></para>

    <para>where <computeroutput>OPTION</computeroutput> is one of the
    following keywords:<glosslist>
        <glossentry>
          <glossterm><computeroutput>SaveState</computeroutput></glossterm>

          <glossdef>
            <para>Don't allow the user to save the VM state when terminating
            the VM.</para>
          </glossdef>
        </glossentry>

        <glossentry>
          <glossterm><computeroutput>Shutdown</computeroutput></glossterm>

          <glossdef>
            <para>Don't allow the user to shutdown the VM by sending the ACPI
            power-off event to the guest.</para>
          </glossdef>
        </glossentry>

        <glossentry>
          <glossterm><computeroutput>PowerOff</computeroutput></glossterm>

          <glossdef>
            <para>Don't allow the user to power off the VM.</para>
          </glossdef>
        </glossentry>

        <glossentry>
          <glossterm><computeroutput>Restore</computeroutput></glossterm>

          <glossdef>
            <para>Don't allow the user to return to the last snapshot when
            powering off the VM.</para>
          </glossdef>
        </glossentry>
      </glosslist></para>

    <para>Any combination of the above is allowed. If all options are
    specified, the VM cannot be shut down at all.</para>
  </sect1>

  <sect1 id="vboxwebsrv-daemon">
    <title>Starting the VirtualBox web service automatically</title>

    <para>The VirtualBox web service
    (<computeroutput>vboxwebsrv</computeroutput>) is used for controlling
    VirtualBox remotely. It is documented in detail in the VirtualBox Software
    Development Kit (SDK); please see <xref linkend="VirtualBoxAPI" />. As the
    client base using this interface is growing, we added start scripts for
    the various operation systems we support. The following describes how to
    use them. <itemizedlist>
        <listitem>
          <para>On Mac OS X, launchd is used. An example configuration file
          can be found in
          <computeroutput>$HOME/Library/LaunchAgents/org.virtualbox.vboxwebsrv.plist</computeroutput>.
          It can be enabled by changing the
          <computeroutput>Disabled</computeroutput> key from
          <computeroutput>true</computeroutput> to
          <computeroutput>false</computeroutput>. To manually start the
          service use the following command: <screen>launchctl load ~/Library/LaunchAgents/org.virtualbox.vboxwebsrv.plist</screen>
          For additional information on how launchd services could be
          configured see <literal><ulink
          url="http://developer.apple.com/mac/library/documentation/MacOSX/Conceptual/BPSystemStartup/BPSystemStartup.html">http://developer.apple.com/mac/library/documentation/MacOSX/Conceptual/BPSystemStartup/BPSystemStartup.html</ulink></literal>.</para>
        </listitem>
      </itemizedlist></para>
  </sect1>

  <sect1 id="vboxballoonctrl">
    <title>Memory Ballooning Service</title>

    <para>Starting with VirtualBox 4.0.8 a new host executable called <computeroutput>VBoxBalloonCtrl</computeroutput> is
    available to automatically take care of a VM's configured memory balloon (see <xref linkend="guestadd-balloon" />
    for an introduction to memory ballooning). This is especially useful for server environments where VMs may dynamically
    require more or less memory during runtime.</para>

    <para>VBoxBalloonCtrl periodically checks a VM's current memory balloon and its free guest RAM and automatically adjusts
    the current memory balloon by inflating or deflating it accordingly. This handling only applies to running VMs having recent
    Guest Additions installed.</para>

    <para>To set up VBoxBalloonCtrl and adjust the maximum ballooning size a VM can reach the following parameters will be checked in the following
    order:
      <itemizedlist>
        <listitem>specified via VBoxBalloonCtrl command line parameter <computeroutput>--balloon-max</computeroutput></listitem>
        <listitem>per-VM parameter using <screen>VBoxManage setextradata "VM-Name" VBoxInternal/Guest/BalloonSizeMax &lt;Size in MB&gt;</screen></listitem>
        <listitem>global parameter for all VMs using <screen>VBoxManage setextradata global VBoxInternal/Guest/BalloonSizeMax &lt;Size in MB&gt;</screen></listitem>
      </itemizedlist>
      <note>
        <para>If no maximum ballooning size is specified by at least one of the parameters above, no ballooning will be performed at all.</para>
      </note>
    </para>

    <para>For more options and parameters check the built-in command line help accessible with <computeroutput>--help</computeroutput>.</para>
  </sect1>
</chapter>