The variables can be empty if not set in the UCI config.
Reported-by: Petr Novák <petrn@me.com>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
This bug was introduced since dd206b7d0b
mwan3_remon_ipv4 and mwan3_remon_ipv6 is command to run not a variable
I add some comments on them hopefully people will notice it
Signed-off-by: Chen Minqiang <ptpt52@gmail.com>
The Go compiler can now manage the build dependencies by itself, as
obfs4proxy has been ported to a Go module.
Signed-off-by: Jeffery To <jeffery.to@gmail.com>
* fix a json related ressource leak
* add a reload trigger when the wireless config gets changed
* set an interface default 'trm_wwan' (like the LuCI frontend)
* reordered nested loops to optimize the connection handling
Signed-off-by: Dirk Brenken <dev@brenken.org>
Yggdrasil builds end-to-end encrypted networks with IPv6. Beyond the
similarities with cjdns is a different routing algorithm. This
globally-agreed spanning tree uses greedy routing in a metric space.
Back-pressure routing techniques allow advanced link aggregation bonding
on per-stream basis. In turn, a single stream will span across multiple
network interfaces simultaneously with much greater throughput.
Authored by: William Fleurant <meshnet@protonmail.com>
Signed-off-by: Paul Spooren <mail@aparcar.org>
This one contains only a few CVEs + bugfixes.
* CVE-2019-8381 memory access in do_checksum() (#538)
* CVE-2019-8376 NULL pointer dereference get_layer4_v6() (#537)
* CVE-2019-8377 NULL pointer dereference get_ipv6_l4proto() (#536)
* Rename Ethereal to Wireshark (#545)
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
trafficshaper create QoS rules to limit (or reserve) traffic used
by classes of clients.
Uplink and downlink can be controled (or not controlled) independently.
Client classes are defined by its network addresses (IPv4 or IPv6). Each
client class can define absolute or relative (to wan) bandwith, and also
the use (or not) of spare wan bandwidth when avaiable.
Signed-off-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Using shorewall-lite {en|dis}able instead of completely restarting
Shorewall is much more efficient.
But it also makes sense to move the starting of Shorewall from init
to an interface hotplug event. The "lan" interface should be a good
indicator that networking it ready. Besides, Shorewall won't start
until br-lan is available.
Signed-off-by: Brian J. Murrell <brian@interlinx.bc.ca>
When the server hostname resolved to both IPv4 and IPv6 addresses,
connecting would fail with nothing in syslog. This corrects that oversight.
Signed-off-by: Daniel Gimpelevich <daniel@gimpelevich.san-francisco.ca.us>
* fix for #8357
* fix unexpectedy calling option_cb() during wireless config_load
* react immediately when the current active uplink connection
gets deleted
Signed-off-by: Dirk Brenken <dev@brenken.org>
The following patches are in upstream now
0100-netdev-linux-Use-unsigned-int-for-ifi_flags.patch
0103-ovs-ctl-fix-setting-hostname.patch
0106-ovs-save-compatible-with-busybox-ip-command.patch
0107-datapath-use-KARCH-when-building-linux-datapath-modu.patch
As for 0001-musl-compatibility.patch, the net/if_packet.h part does not
apply anymore. And musl is not relevant as we use libatomic from gcc
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Josef Schlehofer
William Fleurant
Peter Wagner
Toke Høiland-Jørgensen
Chen Minqiang
Deng Qingfang
Jeffery To
Toni Uhlig
Dirk Brenken
Brian J. Murrell
Paul Spooren
Alexandru Ardelean
Luiz Angelo Daros de Luca
W. van den Akker
Jannik Vieten
Jan Pavlinec
W. Michael Petullo
Nikos Mavrogiannopoulos
Maxim Storchak
Yousong Zhou
Eric Luehrsen
Hannu Nyman
Rosy Song
Daniel Salzman
Daniel Gimpelevich
Daniel Engberg
Ken Keys
Daniel Golle