I know this is painful to change at this point, but I don't like the mc_ prefix. It seems too terse. mchain_ is only four characters longer, and is consistent with mbufq_.

Assert that the chain is empty?

I was thinking of mc_foo() functions to be substitute for m_foo() functions in many places, with local variable named mc instead if two variables named m and tail, seeing mchain as an alternative to mbuf itself, rather than to mbufq. Also, to be fully consistent with mbufq the prefix should be mbufc_.

Let me work on other things you noticed and let's come back to that later.

Why? The supposed usage of this KPI is that mc has just been allocated on stack. A prologue of a function that makes use of mchain, but was called by old code:

foo(struct mbuf *m)
{
      struct mchain mc;

      mc_init_m(&mc, m);