Assuming that passives need verbal morphology, which languages commonly said to have a passive do not actually count?

Haspelmath's point is really one of terminology. He says that passives have to be morphologically marked and it is not enough for a construction to have a passive meaning with some possible corresponding change in syntax (word order). He does NOT actually say that English does not have a passive. Only that some constructions considered passive because of meaning/syntax such as 'destruction of the city by the enemy' are not actually passives. This is fairly sensible in the cases where additional constructions can claim the passive label - there are lots of constructions in English with passive meanings but no passive morphology ("they killed him" when they stands in for a generic de-emphasised agent) or in Slavic languages where things like reflexives are used in a similar way. Although, Haspelmath would accept the reflexive (or other) morphology as a possible way of morphologically marking the passive (a case of functional polysemy). But he does not make it clear how he classified languages where there are multiple morphological ways of expressing the passive but one canonical passive.

I'm not sure, that this is an all that interesting thing to ask. It would be somewhat interesting to see how many languages have had 'passive' identified simply under the theoretical assumption that all languages have a passive but it's ultimately a question of terminology. All those languages probably express some passive-like semantics through other means but those constructions do not have all the features of a passive but that is a question of how the term is defined - not one of what actually happens in the language which, as far as I can, tell is not really in dispute in most of these cases.