Serge and I discussed the new network parser we've merge a couple of days ago.
He pointed out that a bunch of use-cases we're currently supporting in the old
network parser would be broken by the new parser. As we've pointed out many
times before, we're strongly commited to backwards compatibility and not
breaking existing use-cases. That's why we decided to take a new approach.
Instead of trying to mangle the old parser and new parser to come up with
something that allows a smooth transition we will simply deprecate the old
configuration keys with LXC 3.0. In the meantime we will support the full-blown
old legacy parser and the new network parser. Specifically, this means that
we're deprecating:
lxc.network.*
in favor of
lxc.net.*
With LXC 2.1. defining networks using lxc.network.* keys will cause a
deprecation warning to be shown/logged. We strongly suggest that users upgrade
their existing configuration files to switch to the new network configuration
parser. Starting with LXC 3.0 we will remove all lxc.network.* keys and will
only support lxc.net.* style network configurations.
Note that the new network configuration parser will only accept index based
configuration keys, i.e. we are only support lxc.net.[i].* keys without an
index such as lxc.net.type are not supported anymore. The advantages of this
approach are vast. Not just internally, but also user-facing since it is much
clearer what configuration key belongs to what network.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
The new network parser will support specifying multiple networks in the old
format where each new non-indexed "lxc.network.type" line starts a new network
configuration. This way we don't break users. For now, we just print a
deprecation warning. We will KILL this in LXC 3.0.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Some of the checks were previously performed when parsing the network config.
But since we allow for a little more flexibility now it doesn't work anymore.
Instead, let's validate the network at creation time.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
lxc_get_netdev_by_idx() takes care of checking whether a given netdev struct
for a given index is already allocated. If so it returns a pointer to it to the
caller.
If it doesn't find it it will allocate a new netdev struct and insert it into
the network list at the right position. For example, let's say you have the
following networks defined in your config file:
lxc.network.0.type=veth
lxc.network.0.link=lxcbr0
lxc.network.0.name=eth0
lxc.network.0.flags=up
lxc.network.3.type=veth
lxc.network.3.link=lxcbr0
lxc.network.3.name=eth1
lxc.network.3.flags=up
lxc.network.3.hwaddr = 00:16:3e:5d:7d:65
lxc.network.1.type=veth
lxc.network.1.link=lxcbr0
lxc.network.1.name=eth0
lxc.network.1.flags=up
lxc_get_netdev_by_idx() will take care that the internal network list appears
in the order:
0
1
3
The ordering of the keys themselves can also be random, meaning you could do:
lxc.network.3.type=veth
lxc.network.0.link=lxcbr0
lxc.network.0.name=eth0
lxc.network.3.hwaddr = 00:16:3e:5d:7d:65
lxc.network.0.flags=up
lxc.network.3.flags=up
lxc.network.0.type=veth
lxc.network.3.link=lxcbr0
lxc.network.3.name=eth1
and lxc_get_netdev_by_idx() would still figure out the correct ordering since
it keeps track of the index.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
