Research ethics: At least around mathematics, self-plagiarism is defined much more narrowly than plagiarism. Essentially, in order to self-plagiarize, you have to write a paper that contains no new results whatsoever (compared to your older work); just having some overlap is not a problem. Verbatim self-repetition is not regarded as problematic (although it will make people think you are lazy or a bad writer) as long as it does not pretend to be new material. Repeating research papers in lecture notes or textbooks is not considered wrong.
Some even take the point of view that publishing the same result many times (first sloppily, then better, then really well) is legitimate; see Rota's "Publish the Same Result Several Times" advice. (Note: This advice is probably controversial, particularly these days, when there are few "obscure" journals, and when there is no reason to publish a crappy draft when you can just post it on the arXiv. But a second publication that subsumes and improves on a previous paper won't usually be considered improper. Also, you can publish a draft of ongoing research in conference proceedings and then re-publish it in a proper paper; that's what conferences are for.)
Copyright: There is, as always, a noticeable disconnect between what is legally required and what is actually done. In terms of the former, there may well be problems. In terms of the latter, you shouldn't worry. Unless you copy and republish a whole book, no one will sue you. (And even annoyances such as takedown requests to your university tend to be rare -- and don't always correlate with actual breaches of copyright.) Academics disregard copyright six ways from Sunday; if they suddenly started to care, research and teaching would probably slow down to a grind for a year or so until the most important sources are rewritten and published as open content. You should almost always ask yourself "what will the reader think", not "what will Elsevier think", at least if you are planning to get hired by academics rather than by Elsevier :)