Initial revision

[bacula/rescue] / rescue / linux / cdrom / yaird-0.0.5 / doc / kernel.xml

<section id="kernel">
  <title>The interface between kernel and image</title>

  <para>
    The initial boot image is supposed to load enough modules to let
    the real root device be mounted cleanly.  It starts up in a
    <emphasis>very</emphasis> bare environment and it has to do tricky
    stuff like juggling root filesystems; to pull that off successfully
    it makes sense to take a close look at the environment that the
    kernel creates for the image and what the kernel expects it to do.
    This section contains raw design notes based on kernel 2.6.8.
  </para>

  <para>
    The processing of the image starts even before the kernel is
    activated.  The bootloader, grub or lilo for example, reads two
    files from the boot file system into ram: the kernel and image.
    The bootloader somehow manages to set two variables in the kernel:
    <code>initrd_start</code> and <code>initrd_end</code>; these variables
    point to the copy of the image in ram.  The bootloader now
    hands over control to the kernel.
  </para>

  <para>
    During setup, the kernel creates a special file system, rootfs.
    This mostly reuses ramfs code, but there are a few twists: it can
    never be mounted from userspace, there's only one copy, and it's not
    mounted on top of anything else.  The existence of rootfs means that
    the rest of the kernel always can assume there's a place to mount
    other file systems.  It also is a place where temporary files can
    be created during the boot sequence.
  </para>

    <para>
      In <code>initramfs.c:populate_rootfs()</code>, there are two
      possibilities.  If the image looks like a cpio.gz file, it is
      unpacked into rootfs.  If the file <filename>/init</filename> is
      among the files unpacked from the cpio file, the initramfs model
      is used; otherwise we get a more complex interaction between kernel
      and initrd, discussed in <xref linkend="initrd"/>.
    </para>

  <simplesect>
    <title>Booting with Initramfs</title>
    <para>
      If the image was a cpio file, and it contains a file
      <filename>/init</filename>, the initram model is used.
      The kernel does some basic setup and hands over control to
      <filename>/init</filename>; it is then up to
      <filename>/init</filename> to make a real root available and to
      transfer control to the <filename>/sbin/init</filename> command
      on the real root.
    </para>

    <para>
      The tricky part is to do that in such a way that there
      is no way for user processes to gain access to the rootfs
      filesystem; and in such a way that rootfs remains empty and
      hidden under the user root file system.  This is best done
      using some C code; <application>yaird</application> uses
      <application>run_init</application>, a small tool based on
      <application>klibc</application>.
      <programlisting>
        # invoked as last command in /init, with no other processes running,
        # as follows:
        # exec run_init /newroot /sbin/init "$@"
        - chdir /newroot
        # following after lots of sanity checks and not across mounts:
        - rm -rf /*
        - mount --move . /
        - chroot .
        - chdir /
        - open /dev/console
        - exec /sbin/init "$@"
      </programlisting>
    </para>

  </simplesect>

  <simplesect id="initrd">
    <title>Booting with initrd</title>
    <para>
      If the image was not a cpio file, the kernel copies the
      initrd image from where ever the boot loader left it to
      <filename>rootfs:/initrd.image</filename>, and frees the ram used
      by the bootloader for the initrd image.
    </para>

    <para>
      After reading initrd, the kernel does more setup to the point where
      we have:
      <itemizedlist>

        <listitem>
          <para>
              working CPU and memory management
          </para>
        </listitem>

        <listitem>
          <para>
              working process management
          </para>
        </listitem>

        <listitem>
          <para>
              compiled in drivers activated
          </para>
        </listitem>

        <listitem>
          <para>
              a number of support processes such as ksoftirqd are created.
              (These processes have the rootfs as root; they can get a new
              root when the <code>pivot_root()</code> system call is used.)
          </para>
        </listitem>

        <listitem>
          <para>
              something like a console.  <code>Console_init()</code> is
              called before PCI or USB probes, so expect only compiled in
              console devices to work.
          </para>
        </listitem>

      </itemizedlist>
    </para>

    <para>
      At this point, in <code>do_mounts.c:prepare_namespace()</code>,
      the kernel looks for a root filesystem to mount.  That root file
      system can come from a number of places: NFS, a raid device, a plain
      disk or an initrd.  If it's an initrd, the sequence is as follows
      (where devfs can fail if it's not compiled into the kernel)

      <programlisting>
      - mount -t devfs devfs /dev
      - md_run_setup()
      - process initrd
      - umount /dev
      - mount --move . /
      - chroot .
      - mount -t devfs devfs /dev
      </programlisting>

    </para>

    <para>
      Once that returns, in <code>init/main.c:init()</code>,
      initialisation memory is freed and <filename>/sbin/init</filename>
      is executed with <code>/dev/console</code> as file  descriptor 0, 1
      and 2.  <filename>/sbin/init</filename> can be overruled with
      an <code>init=/usr/bin/firefox</code> parameter passed to the
      boot loader; if <filename>/sbin/init</filename> is not found,
      <filename>/etc/init</filename> and a number of other fallbacks
      are tried.  We're in business.
    </para>

    <para>
      The processing of initrd starts in
      <code>do_mounts_initrd.c:initrd_load()</code>.  It creates
      <filename>rootfs:/dev/ram</filename>, then copies
      <filename>rootfs:/initrd.image</filename> there and unlinks
      <filename>rootfs:/initrd.image</filename>.  Now we have the initrd
      image in a block device, which is good for mounting.  It calls
      <code>handle_initrd()</code>, which does:

      <programlisting>
      # make another block special file for ram0
      - mknod /dev/root.old b 1 0
      # try mounting initrd with all known file systems,
      # optionally read-only
      - mount -t xxx /dev/root.old /root
      - mkdir rootfs:/old
      - cd /root
      - mount --move . /
      - chroot .
      - mount -t devfs devfs /dev
      - system ("/linuxrc");
      - cd rootfs:/old
      - mount --move / .
      - cd rootfs:/
      - chroot .
      - umount rootfs:/old/dev
      - ... more ...
      </programlisting>

    </para>

    <para>
      So <filename>initrd:/linuxrc</filename> runs in an environment where
      initrd is the root, with devfs mounted if available, and rootfs is
      invisible (except that there are open file handles to directories
      in rootfs, needed to change back to the old environment).
    </para>

    <para>
      Now the idea seems to have been that <filename>/linuxrc</filename>
      would mount the real root and <code>pivot_root</code> into it, then start
      <filename>/sbin/init</filename>.  Thus, linuxrc would never return.
      However, <code>main.c:init()</code> does some usefull stuff only
      after linuxrc returns: freeing init memory segments and starting numa
      policy, so in eg Debian and Fedora, <filename>/linuxrc</filename>
      will end, and <filename>/sbin/init</filename>
      is started by <code>main.c:init()</code>.
    </para>

    <para>
      After linuxrc returns, the variable <code>real_root_dev</code>
      determines what happens.  This variable can be read and written
      via <filename>/proc/sys/kernel/real-root-dev</filename>.  If it
      is 0x0100 (the device number of <filename>/dev/ram0</filename>)
      or something equivalent, <code>handle_initrd()</code> will change
      directory to <filename>/old</filename> and return.  If it is
      something else, <code>handle_initrd()</code> will decode it, mount
      it as root, mount initrd as <filename>/root/initrd</filename>,
      and again start <filename>/sbin/init</filename>.  (if mounting as
      <filename>/root/initrd</filename> fails, the block device is freed.)
    </para>

    <para>
      Remember <code>handle_initrd()</code> was called via
      <code>load_initrd()</code> from <code>prepare_namespace()</code>,
      and <code>prepare_namespace()</code> ends by chrooting into the
      current directory: <filename>rootfs:/old</filename>.
    </para>

    <para>
      Note that <filename>rootfs:/old</filename> was move-mounted
      from '/' after <filename>/linuxrc</filename> returned.
      When <filename>/linuxrc</filename> started, the root was
      initrd, but <filename>/linuxrc</filename> may have done a
      <code>pivot_root()</code>, replacing the root with a real root,
      say <filename>/dev/hda1</filename>.
    </para>

    <para>
      Thus:
      <itemizedlist>

        <listitem>
          <para>
              <filename>/linuxrc</filename> is started with initrd
              mounted as root.
          </para>
        </listitem>

        <listitem>
          <para>
              There is working memory management, processes, compiled
              in drivers, and stdin/out/err are connected to a console,
              if the relevant drivers are compiled in.
          </para>
        </listitem>

        <listitem>
          <para>
              Devfs may be mounted on <filename>/dev</filename>.
          </para>
        </listitem>

        <listitem>
          <para>
              <filename>/linuxrc</filename> can <code>pivot_root</code>.
          </para>
        </listitem>

        <listitem>
          <para>
              If you echo 0x0100 to
              <filename>/proc/sys/kernel/real-root-dev</filename>,
              the <code>pivot_root</code> will remain in effect after
              <filename>/linuxrc</filename> ends.
          </para>
        </listitem>

        <listitem>
          <para>
              After <filename>/linuxrc</filename> returns,
              <filename>/dev</filename> may be unmounted and replaced
              with devfs.
          </para>
        </listitem>

      </itemizedlist>
    </para>

    <para>
      Thus a good strategy for <filename>/linuxrc</filename> is to
      do as little as possible, and defer the real initialisation
      to <filename>/sbin/init</filename> on the initrd; this
      <filename>/sbin/init</filename> can then <code>pivot_root</code>
      into the real root device.
      <programlisting>
        #!/bin/dash
        set -x
        mount -nt proc proc /proc
        # root=$(cat proc/sys/kernel/real-root-dev)
        echo 256 > proc/sys/kernel/real-root-dev
        umount -n /proc
      </programlisting>
    </para>

  </simplesect>

  <simplesect>
    <title>Kernel command line parameters</title>
    <para>
      The kernel passes more information than just an initial file system
      to the initrd or initramfs image; there also are the kernel boot
      parameters.  The bootloader passes these to the kernel, and the kernel
      in turn passes them on via <filename>/proc/cmdline</filename>.
    </para>

    <para>
      An old version of these parameters is documented in the
      <citerefentry>
        <!-- Sometimes I think docbook is overdoing this markup thing -->
        <refentrytitle>bootparam</refentrytitle>
        <manvolnum>7</manvolnum>
      </citerefentry> manual page; more recent information is in the kernel
      documentation file <citetitle>kernel-parameters.txt</citetitle>.
      Mostly, these parameters are used to configure non-modular drivers,
      and thus not very interesting to <application>yaird</application>.
      Then there are parameters such as <code>noapic</code>, which are
      interpreted by the kernel core and also irrelevant to
      <application>yaird</application>.
      Finally there are a few parameters which are used by the kernel
      to determine how to mount the root file system.
    </para>

    <para>
      Whether the initial image should emulate these options or ignore them
      is open to discussion; you can make a case that the flexibility these
      options offer has become irrelevant now that initrd/initramfs offers
      far more fine grained control over the way in which the system
      is booted.
      Support for these options is mostly a matter of tuning the
      distribution specific templates, but it is possible that the
      templates need an occassional hint from the planner.
      To find out just how much "mostly" is, we'll try to implement
      full support for these options and see where we run into
      limitations.
      An inventarisation of relevant options.
      <variablelist>

        <varlistentry>
          <term>
            ide
          </term>
          <listitem>
            <para>
              These are options for the modular ide-core driver.
              This could be supported by adding an attribute
              "isIdeCore" to insmod actions, and expanding the ide
              kernel options only for insmod actions where that
              attribute is true.
              It seems cleaner to support the options from
              <filename>/etc/modprobe.conf</filename>.
              Unsupported for now.
            </para>
          </listitem>
        </varlistentry>

        <varlistentry>
          <term>
            init
          </term>
          <listitem>
            <para>
              The first program to be started on the definitive root device,
              default <filename>/sbin/init</filename>.  Supported.
            </para>
          </listitem>
        </varlistentry>

        <varlistentry>
          <term>
            ro
          </term>
          <listitem>
            <para>
              Mount the definitive root device read only,
              so that it can be submitted to <application>fsck</application>.
              Supported; this is the default behaviour.
            </para>
          </listitem>
        </varlistentry>

        <varlistentry>
          <term>
            rw
          </term>
          <listitem>
            <para>
              Three guesses.  Supported.
            </para>
          </listitem>
        </varlistentry>

        <varlistentry>
          <term>
            resume, noresume
          </term>
          <listitem>
            <para>
              Which device (not) to use for software suspend.
              To be done.
            </para>
          </listitem>
        </varlistentry>

        <varlistentry>
          <term>
            root
          </term>
          <listitem>
            <para>
              The device to mount as root.  This is a nasty one:
              the planner by default only creates device nodes
              that are needed to mount the root device, and even
              if you were to put hotplug on the inital image
              to create all possible device nodes, there's still
              the matter of putting support for the proper file system
              on the initial image.
              We could make an option to
              <application>yaird</application> to specify a list
              of possible root devices and load the necessary
              modules for all of them.
              Unsupported until there's a clear need for it.
            </para>
          </listitem>
        </varlistentry>

        <varlistentry>
          <term>
            rootflags
          </term>
          <listitem>
            <para>
              Flags to use while mounting root file system.
              Implement together with root option.
            </para>
          </listitem>
        </varlistentry>

        <varlistentry>
          <term>
            rootfstype
          </term>
          <listitem>
            <para>
              File system type for root file system.
              Implement together with root option.
            </para>
          </listitem>
        </varlistentry>

        <varlistentry>
          <term>
            nfsaddrs
          </term>
          <listitem>
            <para>
              For diskless booting.
              Unclear whether we need this.  NFS booting is desirable,
              but I guess that will mostly be done under control of
              DHCP.  Unsupported for now.
            </para>
          </listitem>
        </varlistentry>

        <varlistentry>
          <term>
            nfsroot
          </term>
          <listitem>
            <para>
              More diskless booting.
              Unsupported for now.
            </para>
          </listitem>
        </varlistentry>

      </variablelist>

    </para>
  </simplesect>
</section>