How is a manual IPv6-in-IPv4 (6in4) tunnel configured, and what do you see in a packet capture of it?

Question

Accepted Answer

You create a virtual tunnel interface with tunnel mode ipv6ip, give it IPv4 source/destination endpoints and an IPv6 address, then route IPv6 over it — captures show each IPv6 packet wrapped inside an IPv4 packet.
A manual 6in4 tunnel lets two IPv6 sites talk across an IPv4-only path:
R1(config)# interface tunnel 0
R1(config-if)# tunnel source gig0/0            ! local IPv4 interface
R1(config-if)# tunnel destination 209.165.135.226   ! remote IPv4 endpoint
R1(config-if)# tunnel mode ipv6ip              ! manual 6in4 encapsulation
R1(config-if)# ipv6 address 2001:db8:acad:c::1/64
R1(config)# ipv6 route 2001:db8:acad:b::/64 tunnel 0   ! route IPv6 via tunnel

The tunnel endpoints are IPv4 addresses; the tunnel itself carries IPv6. When you ping the far IPv6 network and watch Wireshark, you see the tell-tale nesting (Verschachtelung): an outer IPv4 header (protocol 41 = IPv6) wrapping the inner IPv6 packet.
Tip: tunnel mode ipv6ip = static 6in4; the IPv4 protocol number for "IPv6 inside" is 41. Spotting protocol 41 in an IPv4 capture is a dead giveaway that someone is tunnelling IPv6.
Go deeper:

6in4 (Wikipedia) — the IPv4-header-wrapping-IPv6 packet format and the protocol-41 marker you spot in a capture.