Heartbleed explanation: This line copies payload bytes from pl to bp. But payload contains the length that the client says that pl has, it has not been checked that the client doesn't lie about the length of pl. So if the client sent less data than it said it would, whatever lies after pl in memory also gets copied to bp. A few lines below, buffer (which is bp and its header) gets sent back to the client. Thanks to http://blog.existentialize.com/diagnosis-of-the-openssl-heartbleed-bug.html for explaining this.

In defense of the person who committed this code, it can be argued that a bad practice in another place in the code base made this bug much harder to find. http://article.gmane.org/gmane.os.openbsd.misc/211963

What about the very similar code in ssl/t1_lib.c below, is something subtle making it safe or is that also part of the bug?