improve startup and runtime performance by
1) moving common startup procedures out of hotplug script when called
from mwan3 start
2) reducing calls to iptables to check status of rules
3) consolidating iptables updates and updating with iptables-restore
4) do not wait for kill if nothing was killed
5) running interface hotplug scripts in parallel
6) eliminate operations in hotplug script that check status on every
single interface unnecessarily
7) consolidate how mwan3track makes hotplug calls
8) do not restart mwan3track on connected events
This is a significant refactor, but should not result in any breaking
changes or require users to update their configurations.
version bump to 2.9.0
Signed-off-by: Aaron Goodman <aaronjg@stanford.edu>
use only committed uci changes for updating routing table
use functions.sh functions rather than uci command line tool
to find interfaces for routing table.
consolidate rtmon_ipv4 and rtmon_ipv6 functions into a single function
Signed-off-by: Aaron Goodman <aaronjg@stanford.edu>
This adjustment of the locks fixes the race condition when a mwan3
hotplug script and a mwan3 command are running at the same time.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
With this change it is now possible to combine interface action events.
If an interface action is generated by netifd or mwan3 for example ifup,
ifdown, connectd or disconnected and this action is configured in the inteface
uci section, then the conntrack table is flushed by mwan3.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
If we set the option "local_source" in the globals mwan3 section to "none",
traffic generated by the router it self will always use the default route from
the wan interface with the lowest metric. If this interface is down
the router traffic still uses the connection with the lowest metric but
this is disconnected. Load balancing and failover from the lan site is
still possible. Only router generated traffic is not load balanced and
could not use failover.
To solve this issue with router initiated traffic add the additional
option "online_metric" to the mwan3 interface section.
If the interface is connected then this lower "online metric" is set in the
default routing table.
With this change we have at least a failover with router initiated
traffic.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
With the list param "rt_table_lookup" in the mwan3 section globals,
it is now possible to add a additional routing table numbers which would get
also parsed and will be added to the connected network.
So mwan3 will treat them as they are directly connected to this device.
This could be usefull if we use ipsec.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
If an interface is not tracked by mwan3 or enabled and this interface is
setup by netifd, then the connected ipset is not update by mwan3.
To fix this also call connected ipset update code even if the interface
is not tracked or enabled by mwan3.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
Move all shell commands which are executed during /lib/mwan3/mwan3.sh
sourceing into a seperate init function which must be called at first.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
Interfaces of some PtP protocols do not have a real gateway. In that
case ubus may fill them with '0.0.0.0' or even leave it blank. This
will cause error when adding new routing rule.
Signed-off-by: David Yang <mmyangfl@gmail.com>
This commit fixed what 6d99b602 was supposed to fix without affecting
interface-bound traffic.
Before 6d99b602 interface-bound traffic was working normally as long
as at least one interface was online. However when the last interface
went offline, it was impossible to ping and such state was
unrecoverable.
Commit 6d99b602 fixed unrecoverable offline state problem (it was
possible to ping -I iface) but messed inteface-bound traffic. Traffic
with interface source address was not working if the interface was in
"offline" state, even if another interface was online.
The problem was caused by an inconsistent "offline" interface state:
iptables-related rules were kept while routing table and policy were
deleted.
The idea behind this commit is to:
1. Keep all the rules for each interface (iptables, routing table,
policy) regardless of its state. This ensures consistency,
2. Make interface state hotplug events affect only iptables'
mwan3_policy_* rules. Interface-related iptables, routing table
and policy is removed only when mwan3 is manually stopped.
To make such changes possible, it's necessary to change the way
mwan3_policy_* rule generator keeps track of interface state hotplug
events.
Until now, it checked for the existence of custom interface-related
routing table (table id 1, 2, 3, ...). Clearly we can no longer rely
on that so each interface state is stored explicitly in file.
Signed-off-by: Marcin Jurkowski <marcin1j@gmail.com>
Use common function to toggle ip address in /var/state/mwan3.
Change also to use toggle function and not set function. If ip address
is often changed every change is saved to /var/state/mwan3.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
Fixes the following bugs introduced in commit 815e83d4:
- hotplug: invalid parameter order when initial interface state is "online",
mwan3track expects initial state to be the third argument
- hotplug: missing source ip address when initial interface state is "offline"
- mwan3track: source ip address should be the fourth argument
Signed-off-by: Marcin Jurkowski <marcin1j@gmail.com>
Add new globals config section with option local_source.
With this config option the self interface generation will be done now
automatically on hotplug event. You can specify which interface (ip)
sould be used for router traffic. To replace the self intereface in the
config set local_source to "lan".
The default option is none, so it will not change default behavior if a
"self" interface is configured in the network section.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
Add new interface config option "inital_state".
If interface comeing up the first time(mwan3 start, boot),
there are now two option for interface behaviour:
- online (default as is now)
Set up interface regardless wether tracking ip are reachable or not.
- offline
Set up interface first to ping tracking ip and if they are reachable set up
the interface completely.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
Even though error was fixed the interface checks still fails, if last_resort
was set to blackhole or unreachable.
To fix this issue do not remove failure interface from iptables change on
down event.
Reported-by: Colby Whitney <colby.whitney@luxul.com>
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
Adds support for interface tracking using either ping, arping or
httping. This allows to track interface status on networks with filtered
ICMP traffic or simply to monitor data link layer etc.
To facilitate binding to a specified interface its IP address is passed
as a new mwan3track parameter. It's currently required by httping
and possibly by other tools that may be added in the future.
Signed-off-by: Marcin Jurkowski <marcin1j@gmail.com>
If more then one interface get up/down at once mwan3 could be in a
undefined state, because more then one mwan3 hotplug script are running
and editing the iptables.
Lock the critical section should solve this issue.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
If netifd set an interface up/down which is not tracked by mwan3 the
connected network of that interface should regardless be added/removed to the
mwan3_connected ipset.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
If interface is getting down by netifd (unplug ethernet cable)
mwan3track will not recognize this change. It will also generate an
additional down event when he notice does his tracking interface is offline.
Mwan3track will now be informed by a signal (trap) USR1 during down event
that the interface is already down. An additional down event will not be
generated.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
On interfaces with interface proto "qmi|ncm" and option dhcp
enabled the interface will be changed to $iface_4 | $iface_6. The
interface mapping in mwan3 will not recognize this interface. To fix
this issue an ubus call will check if an upper interface exists.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
User scripts conversion from luci-app-wan3 will be simplified if the variables
are passed by env instead of arguments.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
On dynamic interface proto (dhcp/pppoe) the hotplug will not execude (exit 9)
because the gateway is already released. The check will now only be made
on a ifup ACTION event.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
Local variable declarations outside of functions are illegal since the Busybox
update to v1.25.0, therfore remove them from the appropriate places.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Fix issue with sticky sessions not working correctly
Fix issue where user created ipsets were not applied
Signed-off-by: Jeroen Louwes <jeroen.louwes@gmail.com>
Fixed issue in mwan3 status output with mwan3 interfaces not yet configuerd in network config
Removed nexthop argument as it is no longer used
Signed-off-by: Jeroen Louwes <jeroen.louwes@gmail.com>
Fix issue where config file would be overwritten in some cases
Fix issue where local address of wan point-to-point links were unreachable from lan
Signed-off-by: Jeroen Louwes <jeroen.louwes@gmail.com>
Fixed issue where an manual ifup-ed interface would immediatly go down again
Remove from init as mwan3 is not a service
Signed-off-by: Jeroen Louwes <jeroen.louwes@gmail.com>
Fixed issue where mwan3 would not immediately set interface down on link-loss event
Added feature to define last-resort action on policies with no members
Signed-off-by: Jeroen Louwes <jeroen.louwes@gmail.com>
Aaron Goodman
Florian Eckert
Chen Minqiang
David Yang
Marcin Jurkowski
Jo-Philipp Wich
Jeroen Louwes