The Linux VLAN HOWTO
  Kristjan Kotkas
  Ben Greear & Others


  1. META
    1. This is the first HOWTO for the "802.1Q VLAN implementation for Linux"

      Homepage: Mailing List:

    2. Copyright

      This document is part of the Linux HOWTO project. The copyright notice is the following: Unless otherwise stated, Linux HOWTO documents are copyrighted by their respective authors. Linux HOWTO documents may be reproduced and distributed in whole or in part, in any medium physical or electronic, as long as this copyright notice is retained on all copies. Commercial redistribution is allowed and encouraged; however, the author would like to be notified of any such distributions. All translations, derivative works, or aggregate works incorporating any Linux HOWTO documents must be covered under this copyright notice. That is, you may not produce a derivative work from a HOWTO and impose additional restrictions on its distribution. Exceptions to these rules may be granted under certain conditions; please contact the Linux HOWTO coordinator at the address given below. In short, we wish to promote dissemination of this information through as many channels as possible. However, we do wish to retain copyright on the HOWTO documents, and would like to be notified of any plans to redistribute the HOWTOs. If you have questions, please contact Tim Bynum, the Linux HOWTO coordinator, at via email.

    3. Disclaimer

      As usual: the author IS NOT responsible for any damage. For the correct wording, see the relevant part of the GNU GPL 0.1.1

    4. Credits

      Thanks to Ben Greear for the VLAN project and also to all the people who have contributed to the project.

    5. General

      What is VLAN is not described in this document. Info on the VLAN protocol in general can be found at Other info can also be found at

  2. Software/Hardware (Cisco-specific setup, with some general info as well.)

    1. VLAN installation & Configuration on the Linux Side.

      I have:
      * Linux kernel 2.2.14
      * Vlan 0.0.10 Patched into it
      * Cisco Catalyst 2900XL
      * 3Com 3C509B NIC using patched driver 3c59x
      Currently VLAN is not part of the kernel distribution so you need to patch
      it into a supported Linux kernel and re-compile.
      You need the kernel source.  If you don't have it already, you can get from or from one of its mirrors.  If this scares you, read the
      It is assumed that you have the linux kernel source extracted, and found at:
      $HOME/linux  If your setup is different, then some of these commands may
      need to be slightly different.
      Download the VLAN package from the vlan homepage and extract it's contents.
      tar -xvzf vlan*.tar.gz
      Go to the vlan directory, build the vlan tools by just typing:
      After this you get a programm named <vconfig>. This program manages all VLAN
      specific configurations.
      Now patch the kernel.
      Go to the linux directory 
      cd linux
      (if you installed the kernel source from some rpm based distribution it is 
      something like /usr/src/linux)
      patch the kernel by typing:
      patch -p 1 < $HOME/vlan/vlan.patch (patch is in the vlan directory)
      Time to compile your kernel. Use the make menuconfig command in your
      linux directory to select your kernel options. The option related to
      802.1Q VLANs is found under the Networking options.
      Additional help for kernel compilation can be found in KERNEL-HOWTO
      Assuming your kernel compiled cleanly, you are now ready to use it. 
      Install your kernel in the normal manner (fix up your /etc/lilo.conf file 
      appropriately and run lilo as root.) 
      Reboot your computer and choose your new kernel. 
      As your computer comes back to life, there will be little sign that you are
      now 802.1Q capable.
      You should see something like this:
              802.1Q VLAN Support v0.10  Ben Greear <>
              vlan Initialization complete.
      Your system is now vlan ready, lets configure some vlans:
      I'm assuming that your VLAN capable network card is eth0.
      First, set the eth0 state to down:
      ifconfig eth0 down
      Ben's Note:   Regarding the next section, you can run plain ethernet and VLAN over the same NIC, but you may not want to..
      Whatever your previous netconf was, you should move everything to vlans.
      This means, that you don't set ip address to the real interface, but set it to
      vlan interface. To set your eth0 with no ip:
      ifconfig eth0 up
      YOU MUST SET THE ETH0 TO UP, or it wont work. (ifconfig eth0 up)
      Add some vlans; goto your vlan directory where you previously compiled 
      vconfig and type:
      vconfig add eth0 2
      ! Little note about VLAN 1. In Cisco systems it is the default VLAN
      so you MUST start using vlans from 2.
      This will create device vlan0002 to your system. Linux will think, that it
      is just another network device, so you can configure it like any other. Also
      you should see the interface by typing
      ifconfig -a
      Lets make some conf on the vlan then:
      ifconfig -i vlan0002 broadcast netmask up
      This ends the configuration at the linux side.
    2. Specific Extreme Networks Configuration

        From: Craig Metz:
      Extreme configuration example:
        create vlan v42
        config vlan v42 tag 42
        config vlan v42 add port 10 tagged
        ... will create a vlan named v42, whose 802.1Q tag is 42, and connect port
      10 (tag 42) to that vlan.
    3. Cisco-specific configuration

      Cisco Conf
      configure the port you want to use as the trunk:
      telnet switch or use the console port
      (will prompt for password, so have it ready)
      conf t
      interface FastEthernet0/24 (it doesnt have to be 0/24)
       duplex full
       speed 100
       switchport trunk encapsulation dot1q
       switchport trunk allowed vlan 2
       switchport mode trunk
      This conf will do the following:
      Set the port to full duplex mode; force the port to 100Mb mode; set the port
      vlan encapsulation to support 802.1Q; tell the
      switch that the port is allowed to run vlans through (even if you set just
      VLAN 2, cisco will automatically add VLAN 1 and VLAN 1002-1005) to the port 
      and set the port to trunk mode aswell. Trunk mode tells the switch that 
      a number of VLANS can go through it.
      Last line is usually the mother of all screw-ups. If you forget that, you
      won't get your VLAN working. Simple as that.
      Now configure some other port to be used as the destination for the vlan:
      conf t
      interface FastEthernet0/1
       duplex half
       speed 10
       switchport access vlan 2
      Here we tell the switch to force the port 1 to half duplex 10Mb mode (normal
      10 Mb NIC) and only traffic from interface VLAN 2 can go through this port.
      also you can use a number of ports with VLAN 2, like a HUB ;)
      You should now connect some other device to port 1.
      Let it have an ip of eg. mask
      Ping linux from it
      If it replies scream: "YESS!!" This means, that VLAN is working.
      Hard truth: It's not over, till its over.
      if this works, then you are out of the woods, if not, well I hear that
      tcpdump is a good tool ;-) and tcpdump that came with the vlan package even
      better tool. (if you want to dump, use the one that came with vlan package)
      NOTE: Ethereal also supports VLANs, and is much more beautiful than tcpdump, if you have GUI capabilities.
      If you can ping the linux and from linux the host, you should try the
      following at linux side:
      ping -s 1476
      If there is no reply, there is something foggy with the NIC. and you
      should start debugging.  If ping -s 100 works, then it is most likely
      an MTU problem with your NIC/Driver.
  3. Specific patches and work-arounds for various configurations.

    1. My Tulip-based card has MTU problems.

      Here is a patch sent in by Ben McKeegan:

      --- drivers/net/tulip.c.orig    Thu May  4 01:16:43 2000
      +++ drivers/net/tulip.c Sat Nov 25 02:39:03 2000
      @@ -59,7 +59,7 @@
       /* Set the copy breakpoint for the copy-only-tiny-buffer Rx structure. */
       #ifdef __alpha__
      -static int rx_copybreak = 1518;
      +static int rx_copybreak = 1522;
       static int rx_copybreak = 100;
      @@ -89,7 +89,7 @@
       /* Operational parameters that usually are not changed. */
       /* Time in jiffies before concluding the transmitter is hung. */
       #define TX_TIMEOUT  (4*HZ)
      -#define PKT_BUF_SZ             1536                    /* Size of each temporary Rx buffer.*/
      +#define PKT_BUF_SZ             1540                    /* Size of each temporary Rx buffer.*/
       /* This is a mysterious value that can be written to CSR11 in the 21040 (only)
          to support a pre-NWay full-duplex signaling mechanism using short frames.
          No one knows what it should be, but if left at its default value some
      @@ -457,7 +457,7 @@
       enum desc_status_bits {
      -       DescOwned=0x80000000, RxDescFatalErr=0x8000, RxWholePkt=0x0300,
      +       DescOwned=0x80000000, RxDescFatalErr=0x4842, RxWholePkt=0x0300,
       /* Ring-wrap flag in length field, use for last ring entry.
      @@ -2848,8 +2848,8 @@
                                         dev->name, entry, status);
                      if (--rx_work_limit < 0)
      -               if ((status & 0x38008300) != 0x0300) {
      -                       if ((status & 0x38000300) != 0x0300) {
      +               if ((status & (0x38000000 | RxDescFatalErr | RxWholePkt)) != RxWholePkt) {
      +                       if ((status & (0x38000000 | RxWholePkt)) != RxWholePkt) {
                                      /* Ingore earlier buffers. */
                                      if ((status & 0xffff) != 0x7fff) {
                                              if (tulip_debug > 1)
      @@ -2875,10 +2875,10 @@
                              struct sk_buff *skb;
       #ifndef final_version
      -                       if (pkt_len > 1518) {
      +                       if (pkt_len > 1522) {
                                      printk(KERN_WARNING "%s: Bogus packet size of %d (%#x).\n",
                                                 dev->name, pkt_len, pkt_len);
      -                               pkt_len = 1518;
      +                               pkt_len = 1522;
    2. My eepro has MTU problems.

      Here is a patch sent in by

      --- linux/drivers/net/eepro100.c        Tue Oct 26 20:53:40 1999
      +++ linux1/drivers/net/eepro100.c       Sun May 14 07:47:34 2000
      @@ -377,12 +377,12 @@
       const char i82557_config_cmd[22] = {
              22, 0x08, 0, 0,  0, 0x80, 0x32, 0x03,  1, /* 1=Use MII  0=Use AUI */
              0, 0x2E, 0,  0x60, 0,
      -       0xf2, 0x48,   0, 0x40, 0xf2, 0x80,              /* 0x40=Force full-duplex */
      +       0xf2, 0x48,   0, 0x40, 0xfa, 0x80,              /* 0x40=Force full-duplex */
              0x3f, 0x05, };
       const char i82558_config_cmd[22] = {
              22, 0x08, 0, 1,  0, 0x80, 0x22, 0x03,  1, /* 1=Use MII  0=Use AUI */
              0, 0x2E, 0,  0x60, 0x08, 0x88,
      -       0x68, 0, 0x40, 0xf2, 0xBD,              /* 0xBD->0xFD=Force full-duplex */
      +       0x68, 0, 0x40, 0xfa, 0xBD,              /* 0xBD->0xFD=Force full-duplex */
              0x31, 0x05, };
       /* PHY media interface chips. */
    3. My SysKonnect sk98lin doesn't work (submitted by: Patrick Schaaf <>)

      Here's a piece needed to get SysKonnect sk98lin driven cards to play nice; they recognize and drop incoming VLAN tagged frames in the driver, the patch below removes that check. Tested a bit with a Cisco Catalyst 6509 on the other side, and a fibre link, works like a charm. The card and driver already supports MTUs up to over 9000, so no problem on that side.

      diff -urN linux/drivers/net/sk98lin/skge.c blues/drivers/net/sk98lin/skge.c
      --- linux/drivers/net/sk98lin/skge.c    Mon Jun 19 20:42:38 2000
      +++ blues/drivers/net/sk98lin/skge.c    Mon Aug  7 09:43:18 2000
      @@ -1948,7 +1948,7 @@
                      if ((Control & RX_CTRL_STAT_VALID) == RX_CTRL_STAT_VALID &&
                              (FrameStat & 
      -                       (XMR_FS_ANY_ERR | XMR_FS_1L_VLAN | XMR_FS_2L_VLAN))
      +                       (XMR_FS_ANY_ERR /*| XMR_FS_1L_VLAN*/ | XMR_FS_2L_VLAN))
                               == 0) {
                              SK_DBG_MSG(NULL, SK_DBGMOD_DRV,
    4. My 3C59X has MTU problems.

      Edit the 3C59x.c, find the place where it says:
      static const int mtu = 1500;
      and replace 1500 with
      static const int mtu = 1504;
    5. My 3C905B has MTU problems.

      From Matti Aarnio <matti.aarnio AT>

              This is a story from the war-front of how I got
              that card to behave properly and to allow full
              ethernet sized MTUs for my VLAN channels.
              To give the salient details:
              Driver has  'static int mtu = 1500;'  line,  which
              variable isn't listed as module parameter (not that
              that obvious fix helped - somehow module parameter
              for this didn't catch at my modprobes..)
              My kludgy solution in the end was just to increase
              that statically compiled in value to   1504,  which
              however made all IP interfaces to show MTU 1504,
              which is -- bogus..
              A bit better solution (I think) is to set max received
              packet size to 1500+14+4 bytes, and not to mess with
              externally visible interface MTU size.
              Reading 3COM documentation about Wn3_MaxPktSize register
              persistance, it seems to need be re-set after every time
              the RxReset is issued.
              Further reading of 3c905C documentation seems to indicate
              that this version of the card (ASICs) will be happy to
              accept VLAN-tagged frames even without driver fixes.  Only
              3c905B (which I have) will need the change (and plain 3c905
              doesn't have MaxPktSize register at all).
              Perhaps simplest for 3c90* series cards is just define fixed
              literal telling that oversize frames are acceptable:
      -       /* Set the full-duplex bit. */
      -       outb(((vp->info1 & 0x8000) || vp->full_duplex ? 0x20 : 0) |
      -                (dev->mtu > 1500 ? 0x40 : 0), ioaddr + Wn3_MAC_Ctrl);
      +       /* Set the full-duplex bit, and allowLargePackets bit too! */
      +       outb(((vp->info1 & 0x8000) || vp->full_duplex ? 0x20 : 0) |
      +                0x40, ioaddr + Wn3_MAC_Ctrl);
              (and similar code at  vortex_timer(),  look for "> 1500")
              If the incoming packet is larger than can fit info buffer,
              we get an overflow error.
              An yet another alternate is to treat plain "OversizedFrame"
              indication as a non-error at packet reception:
                      if ((rx_status & 0x001f4000) == 0x00104000)
                              rx_status &= ~0x0010400;
              and continue with current processing at  boomerang_rx().
              At the  vortex_rx()  this code needs to be a bit different,
              likely some code re-arrangements are called for.
      /Matti Aarnio <matti.aarnio AT>
    6. How to make my PCMCIA ethernet card work with VLANs?

      Per Peter Stuge:

      The problem was that the VLAN code in kernel header files weren't included properly when compiling the PCMCIA stuff. Exactly why? I'm not sure, might be because the PCMCIA stuff isn't the actual kernel and that means it's missing defines that trigger the VLAN #ifdefs.

      Commenting the #ifdefs in linux/netdevice.h and adding code to clear out the struct vlan_dev_info* vlan_dev after having created the new network device in the PCMCIA client for the networking card did the trick if I remember correctly. (The reason it doesn't work out-of-the-box is that the kernel VLAN code has garbage data for the VLAN fields in struct device since the device creator (PCMCIA client driver) doesn't know about them.)

      Ben Adds:

      To clear out the garbage, the PCMCIA driver needs to mset the net_device structure to zero (it should do this anyway..) If anyone has a patch, please send it to me and the owners of the PCMCIA code...

  4. Scripts and Recipes.

    1. Mandrake (RedHat-style) startup script for VLANs.

      Contributed by: "MaxiM Basunov" <>

      # network       Bring up/down networking
      # chkconfig: 2345 10 90
      # description: Activates/Deactivates all network interfaces configured to \
      #              start at boot time.
      # probe: true
      # Source function library.
      . /etc/init.d/functions
      if [ ! -f /etc/sysconfig/network ]; then
          exit 0
      . /etc/sysconfig/network
      if [ -f /etc/sysconfig/pcmcia ]; then
       . /etc/sysconfig/pcmcia
      # Check that networking is up.
      [ ${NETWORKING} = "no" ] && exit 0
      [ -x /sbin/ifconfig ] || exit 0
      # Even if IPX is configured, without the utilities we can't do much
      [ ! -x /sbin/ipx_internal_net -o ! -x /sbin/ipx_configure ] && IPX=
      cd /etc/sysconfig/network-scripts
      # find all the interfaces besides loopback.
      # ignore aliases, alternative configurations, and editor backup files
      interfaces=`ls ifcfg* | egrep -v '(ifcfg-lo|:)' | \
                  egrep -v 'ifcfg-ippp[0-9]+$' | \
      >            egrep 'ifcfg-[a-z0-9\.]+$' | \
      <            egrep 'ifcfg-[a-z0-9]+$' | \
                  sed 's/^ifcfg-//g'`
      # See how we were called.
      case "$1" in
         action "Setting network parameters: " sysctl -p /etc/sysctl.conf
       action "Bringing up interface lo: " ./ifup ifcfg-lo
       case "$IPX" in
           /sbin/ipx_configure --auto_primary=$IPXAUTOPRIMARY \
           if [ "$IPXINTERNALNETNUM" != "0" ]; then
              /sbin/ipx_internal_net add $IPXINTERNALNETNUM $IPXINTERNALNODENUM
       # depreciated but we still use it.
       if [ -f /proc/sys/net/ipv4/ip_forward ] && [ "$FORWARD_IPV4" = "yes" ] ||
       "$FORWARD_IPV4" = "true" ];
        action "Enabling IPv4 packet forwarding" sysctl -w net.ipv4.ip_forward=1
      > action "Setting VLAN parameters: " vconfig set_name_type DEV_PLUS_VID
       for i in $interfaces; do
        if egrep -L "^ONBOOT=\"?[Nn][Oo]\"?" ifcfg-$i >/dev/null 2>&1; then
         # Probe module to preserve interface ordering
         /sbin/ifconfig $i >/dev/null 2>&1
      >   vlan=`echo $i | egrep -v '(lo|:)' | \
      >        egrep -v 'ippp[0-9]+$' | \
      >        egrep '[a-z0-9]+\.[0-9][0-9][0-9][0-9]$' | \
      >        sed "s/^[a-z0-9]*\.//g;s/^0*//g"`
      >   ifvlan=`echo $i | egrep -v '(lo|:)' | \
      >        egrep -v 'ippp[0-9]+$' | \
      >           egrep '[a-z0-9]+\.[0-9][0-9][0-9][0-9]$' | \
      >        sed "s/\.[a-z0-9]*$//g"`
      >   if [ -n "${vlan}" ]; then
      >    action "Enable ${vlan} on {$ifvlan}: " vconfig add ${ifvlan} ${vlan}
      >   fi
         action "Bringing up interface $i: " ./ifup $i boot
       # Add non interface-specific static-routes.
       if [ -f /etc/sysconfig/static-routes ]; then
          grep "^any" /etc/sysconfig/static-routes | while read ignore type dest
      netmask mask bogus args; do
             if [ "${bogus}" = "gw" ]; then
              /sbin/route add -$type $dest $netmask $mask $args
              /sbin/route add -$type $dest $netmask $mask $bogus $args
              touch /var/lock/subsys/network
         # If this is a final shutdown/halt, check for network FS,
       # and unmount them even if the user didn't turn on netfs
       if [ "$RUNLEVEL" = "6" -o "$RUNLEVEL" = "0" -o "$RUNLEVEL" = "1" ]; then
        NFSMTAB=`grep -v '^#' /proc/mounts | awk '{ if ($3 ~ /^nfs$/ ) print $2}'`
        SMBMTAB=`grep -v '^#' /proc/mounts | awk '{ if ($3 ~ /^smbfs$/ ) print
        NCPMTAB=`grep -v '^#' /proc/mounts | awk '{ if ($3 ~ /^ncpfs$/ ) print
        if [ -n "$NFSMTAB" -o -n "$SMBMTAB" -o -n "$NCPMTAB" ] ; then
         /etc/init.d/netfs stop
       for i in $interfaces ; do
               if ifconfig $i 2>/dev/null | grep -q "UP" >/dev/null 2>&1 ; then
           action "Shutting down interface $i: " ./ifdown $i boot
       case "$IPX" in
           if [ "$IPXINTERNALNETNUM" != "0" ]; then
              /sbin/ipx_internal_net del
       ./ifdown ifcfg-lo
       if [ -d /proc/sys/net/ipv4 ]; then
         if [ -f /proc/sys/net/ipv4/ip_forward ]; then
        if [ `cat /proc/sys/net/ipv4/ip_forward` != 0 ]; then
         action "Disabling IPv4 packet forwarding: " sysctl -w
         if [ -f /proc/sys/net/ipv4/ip_always_defrag ]; then
               if [ `cat /proc/sys/net/ipv4/ip_always_defrag` != 0 ]; then
                action "Disabling IPv4 automatic defragmentation: " sysctl -w
       if [ -f /proc/sys/net/ipv4/tcp_syncookies ];then
               if [ `cat /proc/sys/net/ipv4/tcp_syncookies` != 0 ]; then
            sysctl -w net.ipv4.tcp_syncookies=0
              rm -f /var/lock/subsys/network
       echo "Configured devices:"
       echo lo $interfaces
       if [ -x /bin/linuxconf ] ; then
        eval `/bin/linuxconf --hint netdev`
        echo "Devices that are down:"
        echo $DEV_UP
        echo "Devices with modified configuration:"
        echo $DEV_RECONF
        echo "Currently active devices:"
        echo `/sbin/ifconfig | grep ^[a-z] | awk '{print $1}'`
              cd $CWD
       $0 stop
       $0 start
       if [ -x /bin/linuxconf ] ; then
        eval `/bin/linuxconf --hint netdev`
        for device in $DEV_UP ; do
         action "Bringing up device $device: " ./ifup $device
        for device in $DEV_DOWN ; do
         action "Shutting down device $device: " ./ifdown $device
        for device in $DEV_RECONF ; do
         action "Shutting down device $device: " ./ifdown $device
         action "Bringing up device $device: " ./ifup $device
        for device in $DEV_RECONF_ALIASES ; do
         action "Briging up alias $device: "
      /etc/sysconfig/network-scripts/ifup-aliases $device
        for device in $DEV_RECONF_ROUTES ; do
         action "Bringing up route $device: "
      /etc/sysconfig/network-scripts/ifup-routes $device
        case $IPX in yes|true)
          case $IPXINTERNALNET in
         action "Deleting internal IPX network: " /sbin/ipx_internal_net del
         action "Adding internal IPX network $IPXINTERNALNETNUM
      $IPXINTERNALNODENUM: " /sbin/ipx_internal_net add $IPXINTERNALNETNUM \
         action "Adding internal IPX network $IPXINTERNALNETNUM
      $IPXINTERNALNODENUM: "/sbin/ipx_internal_net add $IPXINTERNALNETNUM \
         action "Deleting internal IPX network: " /sbin/ipx_internal_net del
               cd $CWD
        $0 restart
       if [ -x /bin/linuxconf ] ; then
        eval `/bin/linuxconf --hint netdev`
         echo reload
        exit 0
        # if linuxconf isn't around to figure stuff out for us,
        # we punt.  Probably better than completely reloading
        # networking if user isn't sure which to do.  If user
        # is sure, they would run restart or reload, not probe.
        exit 0
              echo "Usage: network {start|stop|restart|reload|status|probe}"
              exit 1
      exit 0

    Kristjan Kotkas 
    KPNQwest Estonia
    t + 372 62 66299 m + 372 51 60697 f + 372 62 66292

    Ben Greear
    Last modified: Sat Jan 13 12:52:41 MST 2001