[Bug 240685] netgraph/ng_vlan_rotate: IEEE 802.1ad VLAN manipulation netgraph node type (new type)
bugzilla-noreply at freebsd.org
bugzilla-noreply at freebsd.org
Thu Sep 19 13:02:29 UTC 2019
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=240685
--- Comment #1 from lutz at donnerhacke.de ---
For the sake of reasoning I'll provide the formatted man page as a comment:
NG_VLAN_ROTATE(4) FreeBSD Kernel Interfaces Manual NG_VLAN_ROTATE(4)
NAME
ng_vlan_rotate - IEEE 802.1ad VLAN manipulation netgraph node type
SYNOPSIS
#include <sys/types.h>
#include <netgraph.h>
#include <netgraph/ng_vlan_rotate.h>
DESCRIPTION
The vlan_rotate node type manipulates the order of VLAN tags of frames
tagged according to the IEEE 802.1ad (an extension of IEEE 802.1Q)
standard between different hooks.
Each node has four special hooks, original, ordered, excessive, and
incomplete.
A frame tagged with an arbitrary number of ETHERTYPE_VLAN,
ETHERTYPE_QINQ, and 0x9100 tags received on the original hook will be
rearranged to a new order of those tags and is sent out the "ordered"
hook. After successful processing the histogram counter for the observed
stack size increments.
If it contains fewer VLANs in the stack, than the configured min limit,
the frame is send out to the incomplete hook and the incomplete counter
increments.
If it contains more VLANs in the stack, than the configured max limit,
the frame is send out to the excessive hook and the excessive counter
increments.
If any destination hook is not connected, the frame is dropped and the
drops counter increments.
For Ethernet frames received on the ordered hook, the transformation is
reversed and is passed to the original hook. Please note, that this
process is identical to the one described above, besides the
ordered/original hooks and the transformation are swapped.
An Ethernet frame received on incomplede or excessive hook is forwarded
to the original hook without any modification.
This node supports only one operation at the moment: Rotation of the
VLANs in the stack. Setting the configuration parameter rot to a positive
value, the stack will roll up by this amount. Negative values will roll
down. A typical scenario is setting the value to 1 in order to bring the
innermost VLAN tag to the outmost level. Rotation includes the VLAN id
and the ethertype, but the QOS paramenters pcp and cfi stay in place.
Typical QOS handling refers to the outmost setting, so this scheme keeps
QOS intact.
HOOKS
This node type supports the following hooks:
original Typically this hook would be connected to a ng_ether(4) node,
using the lower hook connected to a carrier network.
ordered Typically this hook would be connected to a ng_vlan(4) type
node using the downstream hook in order to seperate services.
excessive see below.
incomplete Typically those hooks would be attached to an ng_eiface(4)
type node using the ether hook for anomaly monitoring
purposes.
CONTROL MESSAGES
This node type supports the generic control messages, plus the following:
NGM_VLANROTATE_GET_CONF (getconf)
Read the current configuration.
NGM_VLANROTATE_SET_CONF (setconf)
Set the current configuration.
NGM_VLANROTATE_GET_STAT (getstat)
Read the current statistics.
NGM_VLANROTATE_CLR_STAT (clrstat)
Zeroize the statistics.
NGM_VLANROTATE_GETCLR_STAT (getclrstat)
Read the current statistics and zeroize it in one step.
EXAMPLES
The first example demonstrates how to rotate double or triple tagged
frames so, that the innermost C-VLAN can be used as service
discriminator. The single or double tagged frames (C-VLAN removed) are
send out the an interface pointing to different infrastucture.
#!/bin/sh
BNG_IF=ixl3
VOIP_IF=bge2
ngctl -f- <<EOF
mkpeer ${BNG_IF}: vlan_rotate lower original
name ${BNG_IF}:lower rotate
msg rotate: setconf { min=2 max=3 rot=1 }
mkpeer rotate: vlan ordered downstream
name rotate:ordered services
connect services: ${VOIP_IF} voip lower
msg services: addfilter { vlan=123 hook="voip" }
EOF
Let's inject the following sample frame on the BNG_IF interface:
00:00:00:00:01:01 > 00:01:02:03:04:05,
ethertype 802.1Q-9100 (0x9100), length 110: vlan 2, p 1,
ethertype 802.1Q-QinQ, vlan 101, p 0,
ethertype 802.1Q, vlan 123, p 7,
ethertype IPv4, (tos 0x0, ttl 64, id 15994, offset 0, flags [none],
proto ICMP (1), length 84) 192.168.140.101 > 192.168.140.1:
ICMP echo request, id 40234, seq 0, length 64
The frame ejected on the ordered hook will look like this:
00:00:00:00:01:01 > 00:01:02:03:04:05,
ethertype 802.1Q (0x8100), length 110: vlan 123, p 1,
ethertype 802.1Q-9100, vlan 2, p 0,
ethertype 802.1Q-QinQ, vlan 101, p 7,
ethertype IPv4, (tos 0x0, ttl 64, id 15994, offset 0, flags [none],
proto ICMP (1), length 84) 192.168.140.101 > 192.168.140.1:
ICMP echo request, id 40234, seq 0, length 64
Hence the frame pushed out to the VOIP_IF will have this form:
00:00:00:00:01:01 > 00:01:02:03:04:05,
ethertype 802.1Q-9100, vlan 2, p 0,
ethertype 802.1Q-QinQ, vlan 101, p 7,
ethertype IPv4, (tos 0x0, ttl 64, id 15994, offset 0, flags [none],
proto ICMP (1), length 84) 192.168.140.101 > 192.168.140.1:
ICMP echo request, id 40234, seq 0, length 64
The second example distinguish between double tagged and single tagged
frames. Frames with more VLAN tags are dropped.
#!/bin/sh
IN_IF=bge1
ngctl -f- <<EOF
mkpeer ${IN_IF}: vlan_rotate lower original
name ${IN_IF}:lower separate
msg separate: setconf { min=1 max=1 rot=0 }
mkpeer separate: eiface incomplete ether
name separate:incomplete untagged
mkpeer separate: eiface ordered ether
name separate:ordered tagged
EOF
Setting the rot parameter to zero (or omitting it) does not change the
order of the tags within the frame.
SHUTDOWN
This node shuts down upon receipt of a NGM_SHUTDOWN control message, or
when all hooks have been disconnected.
SEE ALSO
netgraph(4), ng_eiface(4), ng_ether(4), ng_vlan(4), ngctl(8)
HISTORY
The ng_vlan_rotate node type appeared in FreeBSD 12.1-PRERELEASE.
AUTHORS
Lutz Donnerhacke <lutz at donnerhacke.de>
--
You are receiving this mail because:
You are the assignee for the bug.
More information about the freebsd-bugs
mailing list