There are a sources of confusion when it comes to such examples and the binding theory. At the point in time when Principle C and the traditional binding theory were established in the late 1970s and early 1980s, nominal groups were widely assumed to be NPs. By the late 1980s, however, the DP analysis of nominal groups had gained many adherents. The question is concerned, I believe, with this distinction in part. If the possessive determiner his is put in the head D position of DP, that would mean that it does in fact c-command out of the subject and over everything to its right. If one assumes the traditional NP analysis of nominal groups, though, then it clearly does not do this. Or if one puts it in the specifier position of DP (instead of the head position), then it also does not do this. Thus, the question can be answered in a clear way only if one clarifies which analysis of nominal groups one is assuming.
Let us assume the traditional NP analysis of nominal groups, the dominant one at the point in time when Conditions A, B, and C of the binding theory were established. On this approach, his does not c-command out of the subject phrase because it is not the head of the phrase, and it hence does not c-command John, which means it cannot bind John. The name John is therefore definitely free in this regard and the indicated reading (established by the indices) is predicted to be available.
There is another aspect of such sentences that muddies the water. This aspect is that when you actually test such cases on informants, the coreferential reading is judged to be rather unlikely. I know this because I myself have tested these cases extensively. Note that if you switch the order of pronoun and name, the marginality disappears, e.g.
(1) John(i)'s friends love him(i).
The contrast in acceptability across such pairs demonstrates for me that linear order plays a significant role in determining pronoun distribution -- contrary to the traditional binding theory, which rejects a role for linear order. Consider in this area that the traditional binding theory dates from the 1980s. The theory of binding has progressed far beyond that account in the intervening four decades.