> That said, I think the claim in the man page is actually wrong now. > contigmalloc(M_WAITOK) certainly used to be able to return NULL, but I > think that might no longer be true now that kmem_alloc_contig() is > implemented using domainset iterators. If we try all domains, we'll end > up sleeping and retrying indefinitely, or at least it appears to be that > way.
Hmm, contigmalloc does check kmem_alloc_contig's return for NULL, so if this is the case we should probably make that a KASSERT instead.
