Reading this thread, there seem to be some fundamental misconceptions here.
Going to first principles, an 802.11s mesh network is purely layer 2 and neither knows nor cares about ip routing. The mesh autonomously builds its own layer 2 mac-routed network that works just like a virtual layer 2 switch.
See:
In the use case here, if I understand correctly, there are a number of routers where the Internet feed will be supplied via a mesh network operating as a backhaul.
Consider the simple case where you have your isp router providing a "local area network" typically on the subnet 192.168.1.x.
If you connect one of the lan ports of the isp router to an ethernet, layer 2 switch, then the remaining ports on the switch will all act transparently as "extension" ports of the isp router's local area network.
If you have a second OpenWrt configured router that has a wan port and typically one or more lan ports plus a wireless interface giving you an SSID you can connect your phone/tablet/laptop etc to, then this second router will be configured with its own ip subnet, say 192.168.2.x, its own dhcp server and its own dns server in the normal way.
Now to get Internet connectivity from connections to this second router, all you need to do is connect the wan port on this second router to one of the switch ports that are acting as extension ports to the isp router lan.
In the same way, you can connect a third router, this time configured with subnet 192.168.3.x.
You can do this for a fourth router, and a fifth etc until you run out of free ports on the switch.
Now, if you replace the layer 2 switch in this example with an 802.11s mesh network, then everything will work in exactly the same way.
Obviously you will need to configure each additional router to use the mesh network as its "wan" port, instead of the physical ethernet wan port, but that is a relatively trivial config change to each router.