kernel - Tear out selwakeup()

[dragonfly.git] / nrelease / root / README

                            DRAGONFLY CDROM README FILE

    This CDROM boots DragonFly BSD.  Basically what you get is a full base
    system on CD with certain critical directories, such as /tmp, remounted
    read-write using MFS.  Your existing hard drive is not affected by
    booting this CDROM.

    NOTE!!! DRAGONFLY IS UNDERGOING DEVELOPMENT AND IS CONSIDERED
    EXPERIMENTAL!  BSD RELATED EXPERIENCE IS RECOMMENDED WHEN USING
    THIS CDROM.

    If you just want to play with DragonFly and not mess with your hard disk,
    this CDROM boots into a fully operational console-based system, though
    without swap it should be noted that you are limited by available memory.
    It is a good idea to test your hardware for compatibility from a CD boot
    before spending time installing the dist on your hard disk.


                            AUTOMATIC INSTALLATION

    There are currently two installation tools available - the installer, and
    rconfig.

    The installer can be run with a text-based (curses) user interface
    from the serial console or a VTY, and provides a straightforward method
    for installing DragonFly on your HD.  To start it, just login with the
    username 'installer'.

    The installer can also be run with a web-based (CGI) user interface.
    To set this up manually is a bit of work, but much of it can be automated
    by writing a couple of lines into a configuration file on a floppy disk
    or USB pendrive, and inserting or attaching that to the computer before
    booting the CDROM.  See the file /etc/defaults/pfi.conf for more info.

    rconfig is a client/server protocol which requires a server (typically on
    the same network).  An example server setup can be found in
    /usr/share/examples/rconfig.  If you have multiple machines you can setup
    an installation script and run rconfig on a server and then install the
    clients from CD boot with network connectivity (e.g. dhclient <blah>)
    and then, typically, 'rconfig -a'.

    You can also just boot from the CD, login as 'root' to get a shell
    prompt, copy the sample script to /tmp, edit, and run it directly
    (assuming that blowing away your existing disk is ok).


                                  CONSOLE OPERATION

    The second stage boot (boot2) and third stage boot (loader) default
    to dual serial & video console I/O.  The system will also throw a login
    prompt up on ttyd0 by default.  You can direct the boot output
    to just the serial port by creating the file /boot.config with the
    line '-h', or to just the screen using '-V'.  If you wish to leave
    boot2 in dual I/O mode but want the third stage to use just one or the
    other, you can set the 'console' environment variable in /boot/loader.conf
    to either 'console=vidconsole' or 'console=comconsole'.

    The dual serial port operation might have to be disabled if you use
    the serial port for things like UPSs.  Also note that by default
    the CD will run a login prompt on the serial port after booting is
    complete.  This can be disabled by editing the 'ttyd0' line in /etc/ttys
    after installation is complete.

    Note that the kernel itself currently only supports one console or the
    other.  If both are enabled, the kernel will use the video console or
    the last one for which input was received.


                                MANUAL INSTALLATION

    Manual installation of DragonFly onto an HD involves the following sequence
    of commands.  You must be familiar with BSD style UNIX systems to do
    installations manually.  The primary IDE hard drive is typically 'ad0'
    and DragonFly is typically installed onto the first free slice
    (ad0s1 if disk is empty, ad0s2 if your first slice contains
    another OS, etc).  Be careful to substitute the correct partition name
    in the steps below.

        # OPTIONAL STEP: If your disk is already partitioned and you
        # have a spare primary partition on which you want to install
        # Dragonfly, skip this step.  However, sometimes old boot
        # blocks or cruft in the boot area can interfere with the
        # initialization process.  A cure is to zero out the start of
        # the disk before running fdisk.
        #
        # WARNING: This COMPLETELY WIPES and repartitions your hard drive.
        #
        dd if=/dev/zero of=/dev/ad0 bs=32k count=16
        fdisk -I ad0
        fdisk -B ad0

        # If you didn't zero the disk as above, but have a spare slice
        # whose partition type you want to change to UFS, use fdisk(8).

        # This installs boot blocks onto the HD and verifies their 
        # installation.  See note just above the 'reboot' below for
        # things to try if it does not boot from your HD.  If you
        # already have a multi-OS bootloader installed you can skip
        # this step.
        #
        boot0cfg -B ad0
        boot0cfg -v ad0

        # This creates an initial label on the chosen slice of the HD.  If
        # you have problems booting you could try wiping the first 32 blocks
        # of the slice with dd and then reinstalling the label.  Replace
        # 'ad0s1' with the chosen slice.
        #
        # dd if=/dev/zero of=/dev/ad0s1 bs=32k count=16
        disklabel -B -r -w ad0s1 auto

        # Edit the label.  Create various standard partitions.  The typical
        # configuration is:
        #
        #       ad0s1a  256m            This will be your /
        #       ad0s1b  1024m           This will be your swap
        #       ad0s1c                  (leave alone)
        #       ad0s1d  256m            This will be your /var
        #       ad0s1e  256m            This will be your /tmp
        #       ad0s1f  8192m           This will be your /usr (min 4096m)
        #       ad0s1g  *               All remaining space to your /home
        #
        # An example disklabel can be found in /etc/disklabel.ad0s1.
        #
        disklabel -e ad0s1

        # Newfs (format) the various filesystems.  Softupdates is not 
        # normally enabled on the root filesystem because large kernel or
        # world installs/upgrades can run it out of space due to softupdate's
        # delayed bitmap freeing code.
        #
        newfs /dev/ad0s1a
        newfs -U /dev/ad0s1d
        newfs -U /dev/ad0s1e
        newfs -U /dev/ad0s1f
        newfs -U /dev/ad0s1g

        # Mount the filesystems.
        #
        mount /dev/ad0s1a /mnt
        mkdir /mnt/var
        mkdir /mnt/tmp
        mkdir /mnt/usr
        mkdir /mnt/home
        mount /dev/ad0s1d /mnt/var
        mount /dev/ad0s1e /mnt/tmp
        mount /dev/ad0s1f /mnt/usr
        mount /dev/ad0s1g /mnt/home

        # Copy the CDRom onto the target.  cpdup won't cross mount boundaries
        # on the source (e.g. the MFS remounts) so it takes a few commands.
        #
        # Note that /etc contains the config files used for booting from the
        # CDROM itself, and /etc.hdd contains those for booting off a
        # hard disk.  So it's the latter that you want to copy to /mnt/etc.
        #
        cpdup / /mnt
        cpdup /var /mnt/var
        cpdup /etc.hdd /mnt/etc
        cpdup /usr /mnt/usr

        # Cleanup.  Also, with /tmp a partition it is usually reasonable
        # to make /var/tmp a softlink to /tmp.
        #
        chmod 1777 /mnt/tmp
        rm -rf /mnt/var/tmp
        ln -s /tmp /mnt/var/tmp

        # Edit /mnt/etc/fstab to reflect the new mounts.  An example fstab
        # file based on the above parameters exists as /mnt/etc/fstab.example
        # which you can rename to /mnt/etc/fstab.
        #
        mv /mnt/etc/fstab.example /mnt/etc/fstab
        vi /mnt/etc/fstab

        # Save out your disklabel just in case.  It's a good idea to save
        # it to /etc so you can get at it from your backups.  You do intend
        # to backup your system, yah? :-)  (This isn't critical but it's a
        # good idea).
        # 
        disklabel ad0s1 > /mnt/etc/disklabel.ad0s1


                        MISC CLEANUPS BEFORE REBOOTING

    Once you've duplicated the CD onto your HD you have to make some edits
    so the system boots properly from your HD.  Primarily you must remove
    or edit /mnt/boot/loader.conf, which exists on the CD to tell the kernel
    to mount the CD's root partition.

        # Remove or edit /mnt/boot/loader.conf so the kernel does not try
        # to obtain the root filesystem from the CD, and remove the other
        # cruft that was sitting on the CD that you don't need on the HD.
        #
        rm /mnt/boot/loader.conf
        rm /mnt/README* /mnt/autorun* /mnt/index.html /mnt/dflybsd.ico
        rm /mnt/boot.catalog
        rm -r /mnt/rr_moved

    At this point it should be possible to reboot.  The CD may be locked
    since it is currently mounted.  To remove the CD, type 'halt' instead
    of 'reboot', wait for the machine to halt, then the CD door should be
    unlocked.  Remove the CD and hit any key to reboot.

    Be careful of the CD drawer closing on you if you try to remove the CD
    while the machine is undergoing a reboot or reset.

    WARNING: Do not just hit reset; the kernel may not have written out
    all the pending data to your HD.  Either unmount the HD partitions
    or type halt or reboot.

        # halt
        (let the machine halt)
        (remove CD when convenient, be careful of the CD drawer closing on you)
        (hit any key to reboot)


                                    THE ACPI ISSUE

    You will notice in the boot menu that you can choose to boot with or 
    without ACPI.  ACPI is an infrastructure designed to allow an operating
    system to configure hardware devices associated with the system.
    Unfortunately, as usual, PC BIOS makers have royally screwed up the
    standard and ACPI is as likely to hurt as it is to help.  Worse, some
    PCs cannot be booted without it, so there is no good 'default' choice.

    The system will use ACPI by default.  You can disable it in the default
    boot by adding the line 'hint.acpi.0.disabled=1' in /boot/loader.conf.
    If you boot without hitting any menu options the system will boot without
    ACPI.  To boot without ACPI no matter what, place 'unset acpi_load' in
    our /boot/loader.conf instead.  This is not recommended.


                        IF YOU HAVE PROBLEMS BOOTING FROM HD

    There are a couple of things to try.  Try booting from the CD again and
    use boot0cfg to turn off packet mode (boot0cfg -o nopacket ad0).  If you
    can select CHS or LBA mode in your BIOS, try changing the mode to LBA.
    Also try booting with and without ACPI (option 1 or 2 in the boot menu).

    Once you have a working HD based system you can clean up /etc/rc.conf
    to enable things like cron, sendmail, setup your networking, and so
    forth.  If 'ifconfig' does not show your networking device you could
    try to kldload it from /modules.  With a recognized network device
    you can ifconfig its IP address or, if you have a DHCP server on your
    network, use 'dhclient <interfacename>' to obtain an IP address from
    the network.


            USING GIT TO OBTAIN A SOURCE TREE AND DOING BUILDWORLDS

    Instructions on how to obtain and maintain DragonFly source code using
    git are in the development(7) manual page.

    To upgrade a DragonFly system from sources you run the following
    sequence:

    cd /usr/src
    make buildworld
    make buildkernel KERNCONF=<KERNELNAME>
    make installkernel KERNCONF=<KERNELNAME>
    make installworld

    You will also want to run the 'upgrade' target to upgrade your /etc
    and the rest of your system.  The upgrade target is aware of stale
    files created by older DragonFly installations and should delete them
    automatically.

    make upgrade

    See the build(7) manual page for further information.

    Once you've done a full build of the world and kernel you can do
    incremental upgrades of either by using the 'quickworld' and
    'quickkernel' targets instead of 'buildworld' and 'buildkernel'.  If
    you have any problems with the quick targets, try updating your repo
    first, and then a full buildworld and buildkernel as shown above, before
    asking for help.


                OBTAINING A PKGSRC TREE TO BUILD/INSTALL PACKAGES

    In order to obtain a reasonably current snapshot of the pkgsrc tree, use
    the tarball from NetBSD:

    fetch -o /tmp/pkgsrc.tar.gz ftp://ftp.NetBSD.org/pub/NetBSD/packages/pkgsrc.tar.gz
    cd /usr; tar -xzf /tmp/pkgsrc.tar.gz; chown -R root:wheel pkgsrc

    This tree can then be kept up to date with cvs update:

    cd /usr/pkgsrc; cvs up


                        EMERGENCY RECOVERY FROM THE CD

    Lets say you blew up your kernel or something else in / and you need to
    boot the CD to fix it.  Remember that you have a fully operational 
    system when booting the CD, but that you have to fsck and mount your
    hard drive (typically onto /mnt) to get at the contents of your HD.

    Your HD is typically an IDE hard drive, so the device is typically
    /dev/ad0.  DragonFly is typically on the first slice, which is
    /dev/ad0s1, and the root partition is always in partition 'a',
    which is /dev/ad0s1a.

    # fsck root before trying to mount it.  
    fsck /dev/ad0s1a
    # mount root read-write onto /mnt
    mount /dev/ad0s1a /mnt
    # copy files from the CD as appropriate to make it possible to boot
    # from your HD again.  Note that /mnt/kernel may be flags-protected.
    chflags noschg /mnt/kernel
    cp /kernel /mnt/kernel
    cp /modules/* /mnt/modules/

    If you want to mount other partitions from your HD but have forgotten
    what they are, simply cat /mnt/etc/fstab after mounting the root
    partition.