Merge pull request #99 from BarbarossaTM/feature/bonding

Add support for bonding / LAGs.
This commit is contained in:
Ariadne Conill 2020-10-05 14:27:07 -06:00 committed by GitHub
commit a66e8465d1
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
6 changed files with 273 additions and 4 deletions

View file

@ -143,7 +143,8 @@ MANPAGES_8 = \
doc/ifctrstat.8
MANPAGES_5 = \
doc/interfaces.5
doc/interfaces.5 \
doc/interfaces-bond.5
MANPAGES_7 = \
doc/ifupdown-executor.7

204
doc/interfaces-bond.scd Normal file
View file

@ -0,0 +1,204 @@
interfaces-bond(5)
# NAME
*interfaces-bond* - Bonding/LAG extensions for the interfaces(5) file format
# DESCRIPTION
The Linux implementation for Ling Aggregation Groups (LAGs) is called
_bonding_, whereas a LAG interface is called _bond_. The Linux bonding
implementation supports active/passive setups, classical EtherChannels
as well as LACP (802.3ad).
The following options set up bonding/LAG interfaces with ifupdown-ng.
See https://www.kernel.org/doc/Documentation/networking/bonding.rst and
for more information.
# BOND-RELATED OPTIONS
A bond interface must have at least one member port set. All other
options are optional.
*bond-members* _list of interfaces_
Denotes the physical member interfaces to form this LAG. For
compatiblity to ifupdown1 and ifupdown2 _slaves_ as well as
_bond-slaves_ are an alias for this option. This option is
required.
*bond-mode* _mode_
Denotes the mode for this LAG. The _mode_ can be given as string
or as numerical value. Valid values are _balance-rr_ (0),
_active-backup_ (1), _balance-xor_ (2), _broadcast_ (3),
_802.3ad_ (4), _balance-tlb_ (5), _balance-alb_ (6).
The default is _balance-rr_.
*bond-xmit-hash-policy* _policy_
Denotes the hash policy/algorithm used to distribute packets
across the physical links. This only applies for modes
_balance-alb_, _balance-tlb_, _balance-xor_, and _802.3ad_.
The _policy_ can be given as string or as numerical value.
Valid values are _layer2_ (0), _layer3+4_ (1), _layer2+3_ (2),
_encap2+3_ (3), and _encap3+4_ (4). The default is _layer2_.
*bond-min-links* _number_
Denotes the minimum number of available links before turning on
carrier.
*bond-miimon* _interval_
Denotes the MII link monitoring frequency in milliseconds.
This determines how often the link state of each slave is
inspected for link failures. A value of zero disables MII
link monitoring. The default is 0.
*bond-use-carrier* _bool_
Denotes wether miimon uses MII or ethtool ioctls vs. the
netif_carrier_ok() call to determine member link status.
A value of 1 enables the use of netif_carrier_ok(), a value of
0 will use the deprecated MII / ETHTOOL ioctls. The default
is 1.
*bond-updelay* _delay_
Denotes the delay in milliseconds before considering link up,
in milliseconds. The default is 0.
*bond-downdelay* _delay_
Denotes the delay in milliseconds before considering link down,
in milliseconds. The default is 0.
*bond-all-slaves-active* _bool_
Denotes wether duplicate frames (received on inactive ports)
should be dropped (0) or delivered (1). The default is 0.
*bond-packets-per-slave* _num_packets_
Denotes the number of packets to transmit through a member
before moving to the next one. When set to 0 then a slave is
chosen at random. The valid range is 0 - 65535; the default
value is 1. This option has effect only in balance-rr mode.
*bond-lp-interval* _interval_
Denotes the interval in seconds between sending learning packets
to each members peer switch. The valid range is 1 - 0x7fffffff;
the default value is 1. This option has effect only in modes
balance-tlb and balance-alb.
*bond-resend-igmp* _number_
Denotes the number of IGMP membership reports to send after a
link failover happend. The valid range is 0 - 255; a value of
0 prevents the IGMP membership report from being issued in
response to the failover event. The default is 1.
This option is useful for bonding modes balance-rr, active-backup
balance-tlb and balance-alb, in which a failover can switch the
IGMP traffic from one slave to another.
# LACP-RELATED OPTIONS
The following options are only valid in LACP (802.3ad) mode.
*bond-lacp-rate* _rate_
Denotes the _rate_ of LACPDU requested from the peer. The _rate_
can be given as string or as numerical value. Valid values are
slow (0) and fast (1). The default is slow.
*bond-ad-select* _mode_
Denotes the 802.3ad aggregation selection logic. The _mode_ can
be given as string or as numerical value. Valid values are
_stable_ (0), _bandwidth_ (1) and _cound_ (2). The default is
_stable_.
*bond-ad-actor-sys-prio* _priority_
Denotes the LACP system priority. The allowed range is 1 - 65535.
The default value is 65535.
*bond-ad-user-port-key* _key_
Denotes the upper 10 bits of the port-key. he values can be from
0 - 1023. The default is 0.
# ACTIVE/BACKUP-RELATED OPTIONS
The following options are only valid in active/passive setups.
*bond-primary* _interface_
Denotes the primary member interface The specified device will
always be the active slave while it is available. The primary
option is only valid for active-backup, balance-tlb and
balance-alb mode.
*bond-primary-reselect* _policy_
Denotes the reselection policy for the primary member interface.
Valid values are _always_ (0), _better_ (1), and _failure_ (2).
The default is _always_.
*bond-fail-over-mac* _mode_
Denotes whether active-backup mode should set all member
interfaces to the same MAC address at enslavement (the
traditional behavior), or, when enabled, perform special
handling of the bond's MAC address in accordance with the
selected policy. Valid values are _none_ (0), _active_ (1),
_follow_ (2). The default is _none_.
*bond-num-grat-arp* _count_
Denotes the number of peer notifications (gratuitous ARPs and
unsolicited IPv6 Neighbor Advertisements) to be issued after a
failover event. The valid range is 0 - 255; the default is 1.
*bond-num-unsol-na* _count_
This is an alias for _bond-num-grat-arp_
*bond-peer-notif-delay* _interval_
Denotes the interval in milliseconds, between each peer
notification (gratuitous ARP and unsolicited IPv6 Neighbor
Advertisement) issued after a failover event. The default
is 0 which means to match the value of the link monitor
interval.
# ARP-RELATED OPTIONS
The following options configure ARP link monitoring.
The ARP monitor works by periodically checking the slave
devices to determine whether they have sent or received
traffic recently. Regular traffic is generated via ARP
probes issued for the addresses specified by the
_bond-arp-ip-target_ option.
*bond-arp-interval* _interval_
Denotes the frequency in milliseconds to send ARP probes.
*bond-arp-ip-target* _IPv4 address_
Denotes the IP addresses to use as ARP monitoring peers when
_bond-arp-interval_ is > 0.
*bond-arp-validate* _mode_
Specifies whether or not ARP probes and replies should be
validated in any mode that supports arp monitoring, or whether
non-ARP traffic should be filtered (disregarded) for link
monitoring purposes. Valid values are _none_ (0), _active_ (1),
_backup_ (2), _all_ (3), _filter_ (4), _filter_active_ (5), and
_filter_backup_ (6). The default is _none_.
*bond-arp-all-targets* _mode_
Denotes the number of _bond-arp-ip-targets_ that have to be
reachable to consider the member interface to be up. Valid
options are _any_ (0) and _all_ (1). The default is _any_.
# EXAMPLES
A bond using two links and LACP (802.3ad):
```
auto bond0
iface bond0
bond-members eth0 eth1
bond-mode 802.3ad
bond-xmit-hash-policy layer3+4
bond-min-links 1
#
address 192.0.2.42/24
address 2001:db8::42/64
```
# AUTHORS
Maximilian Wilhelm <max@sdn.clinic>

View file

@ -180,6 +180,7 @@ iface eth0
# SEE ALSO
*interfaces-bond*(5)
*interfaces-vxlan*(5)
# AUTHORS

64
executor-scripts/linux/bond Executable file
View file

@ -0,0 +1,64 @@
#!/bin/sh
#
# This executor is responsible for setting up bond/LAG interfaces.
#
# Sat, 03 Oct 2020 20:42:19 +0200
# -- Maximilian Wilhelm <max@sdn.clinic>
#
set -e
[ -n "$VERBOSE" ] && set -x
get_bond_options() {
env | while IFS="=" read opt value; do
# We only care for options of format IF_BOND_<OPTION_NAME>
if ! echo "${opt}" | grep -q '^IF_BOND_[A-Z0-9_]\+$'; then
continue
fi
# Members are handled seperately
if [ "${opt}" = "IF_BOND_MEMBERS" ]; then
continue
fi
# Convert options for the actual name
real_opt=$(echo "${opt}" | sed -e 's/^IF_BOND_\([A-Z0-9_]\+\)/\1/' | tr '[A-Z]' '[a-z]')
echo -n " ${real_opt} ${value}"
done
}
case "$PHASE" in
depend)
echo "${IF_BOND_MEMBERS}"
;;
create)
if [ -d "/sys/class/net/${IFACE}" ]; then
exit 0
fi
# Gather bonding options for this interface
options=$(get_bond_options)
# Create bond
${MOCK} ip link add "${IFACE}" type bond ${options}
# Add members to bundle
for member_iface in ${IF_BOND_MEMBERS}; do
# Work around the current execution order
ip link set "${member_iface}" down
ip link set master "${IFACE}" "${member_iface}"
ip link set "${member_iface}" up
done
;;
destroy)
if [ -z "${MOCK}" -a ! -d "/sys/class/net/${IFACE}" ]; then
exit 0
fi
${MOCK} ip link del "${IFACE}"
;;
esac

View file

@ -28,6 +28,8 @@ struct remap_token {
/* this list must be in alphabetical order for bsearch */
static const struct remap_token tokens[] = {
{"bond-ad-sys-priority", "bond-ad-actor-sys-prio"}, /* ifupdown2 */
{"bond-slaves", "bond-members"}, /* legacy ifupdown, ifupdown2 */
{"driver-message-level", "ethtool-msglvl"}, /* Debian ethtool integration */
{"endpoint", "tunnel-remote"}, /* legacy ifupdown */
{"ethernet-autoneg", "ethtool-ethernet-autoneg"}, /* Debian ethtool integration */

View file

@ -163,9 +163,6 @@ build_environment(char **envp[], const struct lif_execute_opts *opts, const stru
{
if (iface->is_bridge)
lif_environment_push(envp, "IF_BRIDGE_PORTS", (const char *) entry->data);
if (iface->is_bond)
lif_environment_push(envp, "IF_BOND_SLAVES", (const char *) entry->data);
}
char envkey[4096] = "IF_";