Magnetic permeability is confusing**. When dealing with permeability, there are actually THREE related field quantities: The B-field, the H-field, and the M-field. It's difficult to define these all in a way that is both intuitive and technically correct, so I won't try except to say that external fields are generally thought of as H whereas internal fields are generally thought of as B (although there is an internal H field as well). People argue about which one, B or H, is more "fundamental" but I leave those discussions to the philosophers. It's like stress and strain -- different but intertwined. Either way, the B vs H or B vs M curve of a material determines the magnetic permeability of the material. When a "soft" magnetic material (meaning highly permeable but NOT permanently magnetized, approximated by constant permeability $\mu$) is placed in an external field, the internal H-field tends to approach zero (depending also on geometry and value of $\mu$) and is more or less replaced by the M field, representing the magnetization effects on the material. The B field can be thought of as the total of both these fields. Integrating the value of M over the volume of the material gives you the effective magnetic moment $m$. But when you take away the external field, $M$ and $m$ go to zero (as do B and H).
For a situation where the H field of the material "opposes" the B field, you have to consider "hard" magnetic materials with permanent magnetization. In that case, the B vs H curve is more like a loop - not linear at all, and $\mu$ varies depending on the point of interest. There does not need to be an external H field to magnetize the material because it is already magnetized (although an external H was needed to achieve that). To describe the state of magnetization, the relevant point on the B-H curve is in the second quadrant with positive B and negative H, where the H can be thought of as "internal." The magnetization M is positive as before.
** To add to the confusion, there are at least three different unit systems in use, so you may see equations like $B = H$ instead of $B = \mu H$ or factors of $4 \pi$ floating around in some equations but not others. This makes the subject especially hard to self-teach, in my opinion.