In vm_phys_alloc_seg_contig, check alignment before checking orders of multiple memory blocks, to avoid such deeply nested code. In checking multiple memory blocks, check seg->end only once, against pa_end because there's no need to repeat the test for consecutive values of pa, and because the current code, in the case that pa reaches pa_end and passes seg->end in the same iteration, proceeds to allocate that memory.
Define vm_phys_paddr_seg_page to implement a common, complex expression and use it everywhere.