The mutex_lock_io() API is identical in syntax to the mutex_lock() API; the only difference is that the kernel thinks that the wait time of the loser thread(s) is the same as waiting for I/O (the code comment in kernel/locking/mutex.c:mutex_lock_io() clearly documents this; take a look). This can matter accounting-wise.
You can find fairly exotic APIs such as mutex_lock[_interruptible]_nested() within the kernel, with the emphasis here being on the nested suffix. However, note that the Linux kernel does not prefer developers to use nested (or recursive) locking (as we mentioned in the Correctly using the mutex lock section). Also, these APIs only get compiled in the presence of the CONFIG_DEBUG_LOCK_ALLOC config option; in effect, the nested APIs were added to support the kernel lock validator mechanism. They should only be used in special circumstances (where a nesting level must be incorporated between...
