Index: sys/amd64/include/counter.h
===================================================================
--- sys/amd64/include/counter.h
+++ sys/amd64/include/counter.h
@@ -33,7 +33,7 @@
 
 #include <sys/pcpu.h>
 
-#define	EARLY_COUNTER	&temp_bsp_pcpu.pc_early_dummy_counter
+#define	EARLY_COUNTER	(void *)__offsetof(struct pcpu, pc_early_dummy_counter)
 
 #define	counter_enter()	do {} while (0)
 #define	counter_exit()	do {} while (0)
@@ -43,6 +43,7 @@
 counter_u64_read_one(counter_u64_t c, int cpu)
 {
 
+	MPASS(c != EARLY_COUNTER);
 	return (*zpcpu_get_cpu(c, cpu));
 }
 
@@ -65,6 +66,7 @@
 	counter_u64_t c;
 
 	c = arg;
+	MPASS(c != EARLY_COUNTER);
 	*(zpcpu_get(c)) = 0;
 }
 
@@ -86,7 +88,7 @@
 	KASSERT(IS_BSP() || c != EARLY_COUNTER, ("EARLY_COUNTER used on AP"));
 	__asm __volatile("addq\t%1,%%gs:(%0)"
 	    :
-	    : "r" ((char *)c - (char *)&__pcpu[0]), "ri" (inc)
+	    : "r" (c), "ri" (inc)
 	    : "memory", "cc");
 }
 
Index: sys/amd64/include/pcpu.h
===================================================================
--- sys/amd64/include/pcpu.h
+++ sys/amd64/include/pcpu.h
@@ -240,6 +240,10 @@
 
 #define	IS_BSP()	(PCPU_GET(cpuid) == 0)
 
+#define zpcpu_offset_cpu(cpu)	((uintptr_t)&__pcpu[0] + UMA_PCPU_ALLOC_SIZE * cpu)
+#define zpcpu_base_to_offset(base) (void *)((uintptr_t)(base) - (uintptr_t)&__pcpu[0])
+#define zpcpu_offset_to_base(base) (void *)((uintptr_t)(base) + (uintptr_t)&__pcpu[0])
+
 #else /* !__GNUCLIKE_ASM || !__GNUCLIKE___TYPEOF */
 
 #error "this file needs to be ported to your compiler"
Index: sys/kern/subr_pcpu.c
===================================================================
--- sys/kern/subr_pcpu.c
+++ sys/kern/subr_pcpu.c
@@ -95,6 +95,7 @@
 	cpu_pcpu_init(pcpu, cpuid, size);
 	pcpu->pc_rm_queue.rmq_next = &pcpu->pc_rm_queue;
 	pcpu->pc_rm_queue.rmq_prev = &pcpu->pc_rm_queue;
+	pcpu->pc_zpcpu_offset = zpcpu_offset_cpu(cpuid);
 }
 
 void
Index: sys/sys/pcpu.h
===================================================================
--- sys/sys/pcpu.h
+++ sys/sys/pcpu.h
@@ -194,6 +194,7 @@
 	struct rm_queue	pc_rm_queue;		/* rmlock list of trackers */
 	uintptr_t	pc_dynamic;		/* Dynamic per-cpu data area */
 	uint64_t	pc_early_dummy_counter;	/* Startup time counter(9) */
+	uintptr_t	pc_zpcpu_offset;	/* Offset into zpcpu allocs */
 
 	/*
 	 * Keep MD fields last, so that CPU-specific variations on a
@@ -227,14 +228,28 @@
 #endif
 #define	curproc		(curthread->td_proc)
 
+#ifndef zpcpu_offset_cpu
+#define zpcpu_offset_cpu(cpu)	(UMA_PCPU_ALLOC_SIZE * cpu)
+#endif
+#ifndef zpcpu_offset
+#define zpcpu_offset()		(PCPU_GET(zpcpu_offset))
+#endif
+
+#ifndef zpcpu_base_to_offset
+#define zpcpu_base_to_offset(base) base
+#endif
+#ifndef zpcpu_offset_to_base
+#define zpcpu_offset_to_base(base) base
+#endif
+
 /* Accessor to elements allocated via UMA_ZONE_PCPU zone. */
 #define zpcpu_get(base) ({								\
-	__typeof(base) _ptr = (void *)((char *)(base) + UMA_PCPU_ALLOC_SIZE * curcpu);	\
+	__typeof(base) _ptr = (void *)((char *)(base) + zpcpu_offset());		\
 	_ptr;										\
 })
 
 #define zpcpu_get_cpu(base, cpu) ({							\
-	__typeof(base) _ptr = (void *)((char *)(base) +	UMA_PCPU_ALLOC_SIZE * cpu);	\
+	__typeof(base) _ptr = (void *)((char *)(base) +	zpcpu_offset_cpu(cpu));		\
 	_ptr;										\
 })
 
Index: sys/vm/uma_core.c
===================================================================
--- sys/vm/uma_core.c
+++ sys/vm/uma_core.c
@@ -2944,34 +2944,39 @@
 void *
 uma_zalloc_pcpu_arg(uma_zone_t zone, void *udata, int flags)
 {
-	void *item;
+	void *item, *pcpu_item;
 #ifdef SMP
 	int i;
 
 	MPASS(zone->uz_flags & UMA_ZONE_PCPU);
 #endif
 	item = uma_zalloc_arg(zone, udata, flags & ~M_ZERO);
-	if (item != NULL && (flags & M_ZERO)) {
+	if (item == NULL)
+		return (NULL);
+	pcpu_item = zpcpu_base_to_offset(item);
+	if (flags & M_ZERO) {
 #ifdef SMP
 		for (i = 0; i <= mp_maxid; i++)
-			bzero(zpcpu_get_cpu(item, i), zone->uz_size);
+			bzero(zpcpu_get_cpu(pcpu_item, i), zone->uz_size);
 #else
 		bzero(item, zone->uz_size);
 #endif
 	}
-	return (item);
+	return (pcpu_item);
 }
 
 /*
  * A stub while both regular and pcpu cases are identical.
  */
 void
-uma_zfree_pcpu_arg(uma_zone_t zone, void *item, void *udata)
+uma_zfree_pcpu_arg(uma_zone_t zone, void *pcpu_item, void *udata)
 {
+	void *item;
 
 #ifdef SMP
 	MPASS(zone->uz_flags & UMA_ZONE_PCPU);
 #endif
+	item = zpcpu_offset_to_base(pcpu_item);
 	uma_zfree_arg(zone, item, udata);
 }