diff --git a/sys/kern/kern_malloc.c b/sys/kern/kern_malloc.c
--- a/sys/kern/kern_malloc.c
+++ b/sys/kern/kern_malloc.c
@@ -763,6 +763,19 @@
 	return (malloc_large(&size, mtp, ds, flags DEBUG_REDZONE_ARG));
 }
 
+void *
+malloc_domainset_aligned(size_t size, size_t align,
+    struct malloc_type *mtp, struct domainset *ds, int flags)
+{
+	MPASS(powerof2(align));
+	KASSERT(align <= kmemzones[nitems(kmemzones) - 2].kz_size,
+	    ("malloc_domainset_aligned: align %#zx too large", align));
+
+	if (size < align)
+		size = align;
+	return (malloc_domainset(size, mtp, ds, flags));
+}
+
 void *
 mallocarray(size_t nmemb, size_t size, struct malloc_type *type, int flags)
 {
@@ -1146,8 +1159,12 @@
 	for (i = 0, indx = 0; kmemzones[indx].kz_size != 0; indx++) {
 		int size = kmemzones[indx].kz_size;
 		const char *name = kmemzones[indx].kz_name;
+		size_t align;
 		int subzone;
 
+		align = UMA_ALIGN_PTR;
+		if (powerof2(size) && size > sizeof(void *))
+			align = size - 1;
 		for (subzone = 0; subzone < numzones; subzone++) {
 			kmemzones[indx].kz_zone[subzone] =
 			    uma_zcreate(name, size,
@@ -1156,7 +1173,7 @@
 #else
 			    NULL, NULL, NULL, NULL,
 #endif
-			    UMA_ALIGN_PTR, UMA_ZONE_MALLOC);
+			    align, UMA_ZONE_MALLOC);
 		}
 		for (;i <= size; i+= KMEM_ZBASE)
 			kmemsize[i >> KMEM_ZSHIFT] = indx;
diff --git a/sys/sys/malloc.h b/sys/sys/malloc.h
--- a/sys/sys/malloc.h
+++ b/sys/sys/malloc.h
@@ -261,6 +261,9 @@
 	    __result_use_check __alloc_size(2);
 void	*reallocf(void *addr, size_t size, struct malloc_type *type, int flags)
 	    __result_use_check __alloc_size(2);
+void	*malloc_domainset_aligned(size_t size, size_t align,
+	    struct malloc_type *mtp, struct domainset *ds, int flags)
+	    __malloc_like __result_use_check __alloc_size(1);
 
 struct malloc_type *malloc_desc2type(const char *desc);
 
diff --git a/sys/x86/x86/busdma_bounce.c b/sys/x86/x86/busdma_bounce.c
--- a/sys/x86/x86/busdma_bounce.c
+++ b/sys/x86/x86/busdma_bounce.c
@@ -445,19 +445,14 @@
 	 * else allocate a block of contiguous pages because one or more of the
 	 * constraints is something that only the contig allocator can fulfill.
 	 *
-	 * NOTE: The (dmat->common.alignment <= dmat->maxsize) check
-	 * below is just a quick hack. The exact alignment guarantees
-	 * of malloc(9) need to be nailed down, and the code below
-	 * should be rewritten to take that into account.
-	 *
-	 * In the meantime warn the user if malloc gets it wrong.
+	 * Warn the user if malloc gets it wrong.
 	 */
 	if (dmat->common.maxsize <= PAGE_SIZE &&
-	    dmat->common.alignment <= dmat->common.maxsize &&
 	    dmat->common.lowaddr >= ptoa((vm_paddr_t)Maxmem) &&
 	    attr == VM_MEMATTR_DEFAULT) {
-		*vaddr = malloc_domainset(dmat->common.maxsize, M_DEVBUF,
-		    DOMAINSET_PREF(dmat->common.domain), mflags);
+		*vaddr = malloc_domainset_aligned(dmat->common.maxsize,
+		    dmat->common.alignment,
+		    M_DEVBUF, DOMAINSET_PREF(dmat->common.domain), mflags);
 	} else if (dmat->common.nsegments >=
 	    howmany(dmat->common.maxsize, MIN(dmat->common.maxsegsz,
 	    PAGE_SIZE)) &&