diff --git a/drivers/infiniband/hw/qib/qib.h b/drivers/infiniband/hw/qib/qib.h
index f044430..c00ae09 100644
--- a/drivers/infiniband/hw/qib/qib.h
+++ b/drivers/infiniband/hw/qib/qib.h
@@ -1186,7 +1186,7 @@
 void qib_set_ctxtcnt(struct qib_devdata *);
 int qib_create_ctxts(struct qib_devdata *dd);
 struct qib_ctxtdata *qib_create_ctxtdata(struct qib_pportdata *, u32, int);
-void qib_init_pportdata(struct qib_pportdata *, struct qib_devdata *, u8, u8);
+int qib_init_pportdata(struct qib_pportdata *, struct qib_devdata *, u8, u8);
 void qib_free_ctxtdata(struct qib_devdata *, struct qib_ctxtdata *);
 
 u32 qib_kreceive(struct qib_ctxtdata *, u32 *, u32 *);
diff --git a/drivers/infiniband/hw/qib/qib_iba6120.c b/drivers/infiniband/hw/qib/qib_iba6120.c
index b9bea2e..d68266a 100644
--- a/drivers/infiniband/hw/qib/qib_iba6120.c
+++ b/drivers/infiniband/hw/qib/qib_iba6120.c
@@ -3265,7 +3265,9 @@
 
 	dd->eep_st_masks[2].errs_to_log = ERR_MASK(ResetNegated);
 
-	qib_init_pportdata(ppd, dd, 0, 1);
+	ret = qib_init_pportdata(ppd, dd, 0, 1);
+	if (ret)
+		goto bail;
 	ppd->link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
 	ppd->link_speed_supported = QIB_IB_SDR;
 	ppd->link_width_enabled = IB_WIDTH_4X;
diff --git a/drivers/infiniband/hw/qib/qib_iba7220.c b/drivers/infiniband/hw/qib/qib_iba7220.c
index 28063d4..7dec89f 100644
--- a/drivers/infiniband/hw/qib/qib_iba7220.c
+++ b/drivers/infiniband/hw/qib/qib_iba7220.c
@@ -4059,7 +4059,9 @@
 	init_waitqueue_head(&cpspec->autoneg_wait);
 	INIT_DELAYED_WORK(&cpspec->autoneg_work, autoneg_7220_work);
 
-	qib_init_pportdata(ppd, dd, 0, 1);
+	ret = qib_init_pportdata(ppd, dd, 0, 1);
+	if (ret)
+		goto bail;
 	ppd->link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
 	ppd->link_speed_supported = QIB_IB_SDR | QIB_IB_DDR;
 
diff --git a/drivers/infiniband/hw/qib/qib_iba7322.c b/drivers/infiniband/hw/qib/qib_iba7322.c
index 8441579..a7eb325 100644
--- a/drivers/infiniband/hw/qib/qib_iba7322.c
+++ b/drivers/infiniband/hw/qib/qib_iba7322.c
@@ -6544,7 +6544,11 @@
 		}
 
 		dd->num_pports++;
-		qib_init_pportdata(ppd, dd, pidx, dd->num_pports);
+		ret = qib_init_pportdata(ppd, dd, pidx, dd->num_pports);
+		if (ret) {
+			dd->num_pports--;
+			goto bail;
+		}
 
 		ppd->link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
 		ppd->link_width_enabled = IB_WIDTH_4X;
diff --git a/drivers/infiniband/hw/qib/qib_init.c b/drivers/infiniband/hw/qib/qib_init.c
index 6d26299..7fbf466 100644
--- a/drivers/infiniband/hw/qib/qib_init.c
+++ b/drivers/infiniband/hw/qib/qib_init.c
@@ -233,7 +233,7 @@
 /*
  * Common code for initializing the physical port structure.
  */
-void qib_init_pportdata(struct qib_pportdata *ppd, struct qib_devdata *dd,
+int qib_init_pportdata(struct qib_pportdata *ppd, struct qib_devdata *dd,
 			u8 hw_pidx, u8 port)
 {
 	int size;
@@ -243,6 +243,7 @@
 
 	spin_lock_init(&ppd->sdma_lock);
 	spin_lock_init(&ppd->lflags_lock);
+	spin_lock_init(&ppd->cc_shadow_lock);
 	init_waitqueue_head(&ppd->state_wait);
 
 	init_timer(&ppd->symerr_clear_timer);
@@ -250,8 +251,10 @@
 	ppd->symerr_clear_timer.data = (unsigned long)ppd;
 
 	ppd->qib_wq = NULL;
-
-	spin_lock_init(&ppd->cc_shadow_lock);
+	ppd->ibport_data.pmastats =
+		alloc_percpu(struct qib_pma_counters);
+	if (!ppd->ibport_data.pmastats)
+		return -ENOMEM;
 
 	if (qib_cc_table_size < IB_CCT_MIN_ENTRIES)
 		goto bail;
@@ -299,7 +302,7 @@
 		goto bail_3;
 	}
 
-	return;
+	return 0;
 
 bail_3:
 	kfree(ppd->ccti_entries_shadow);
@@ -313,7 +316,7 @@
 bail:
 	/* User is intentionally disabling the congestion control agent */
 	if (!qib_cc_table_size)
-		return;
+		return 0;
 
 	if (qib_cc_table_size < IB_CCT_MIN_ENTRIES) {
 		qib_cc_table_size = 0;
@@ -324,7 +327,7 @@
 
 	qib_dev_err(dd, "Congestion Control Agent disabled for port %d\n",
 		port);
-	return;
+	return 0;
 }
 
 static int init_pioavailregs(struct qib_devdata *dd)
@@ -635,6 +638,12 @@
 	return -ENOMEM;
 }
 
+static void qib_free_pportdata(struct qib_pportdata *ppd)
+{
+	free_percpu(ppd->ibport_data.pmastats);
+	ppd->ibport_data.pmastats = NULL;
+}
+
 /**
  * qib_init - do the actual initialization sequence on the chip
  * @dd: the qlogic_ib device
@@ -922,6 +931,7 @@
 			destroy_workqueue(ppd->qib_wq);
 			ppd->qib_wq = NULL;
 		}
+		qib_free_pportdata(ppd);
 	}
 
 	qib_update_eeprom_log(dd);
diff --git a/drivers/infiniband/hw/qib/qib_mad.c b/drivers/infiniband/hw/qib/qib_mad.c
index ccb1191..edad991 100644
--- a/drivers/infiniband/hw/qib/qib_mad.c
+++ b/drivers/infiniband/hw/qib/qib_mad.c
@@ -1634,6 +1634,23 @@
 	return reply((struct ib_smp *)pmp);
 }
 
+static void qib_snapshot_pmacounters(
+	struct qib_ibport *ibp,
+	struct qib_pma_counters *pmacounters)
+{
+	struct qib_pma_counters *p;
+	int cpu;
+
+	memset(pmacounters, 0, sizeof(*pmacounters));
+	for_each_possible_cpu(cpu) {
+		p = per_cpu_ptr(ibp->pmastats, cpu);
+		pmacounters->n_unicast_xmit += p->n_unicast_xmit;
+		pmacounters->n_unicast_rcv += p->n_unicast_rcv;
+		pmacounters->n_multicast_xmit += p->n_multicast_xmit;
+		pmacounters->n_multicast_rcv += p->n_multicast_rcv;
+	}
+}
+
 static int pma_get_portcounters_ext(struct ib_pma_mad *pmp,
 				    struct ib_device *ibdev, u8 port)
 {
@@ -1642,6 +1659,7 @@
 	struct qib_ibport *ibp = to_iport(ibdev, port);
 	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
 	u64 swords, rwords, spkts, rpkts, xwait;
+	struct qib_pma_counters pma;
 	u8 port_select = p->port_select;
 
 	memset(pmp->data, 0, sizeof(pmp->data));
@@ -1664,10 +1682,17 @@
 	p->port_rcv_data = cpu_to_be64(rwords);
 	p->port_xmit_packets = cpu_to_be64(spkts);
 	p->port_rcv_packets = cpu_to_be64(rpkts);
-	p->port_unicast_xmit_packets = cpu_to_be64(ibp->n_unicast_xmit);
-	p->port_unicast_rcv_packets = cpu_to_be64(ibp->n_unicast_rcv);
-	p->port_multicast_xmit_packets = cpu_to_be64(ibp->n_multicast_xmit);
-	p->port_multicast_rcv_packets = cpu_to_be64(ibp->n_multicast_rcv);
+
+	qib_snapshot_pmacounters(ibp, &pma);
+
+	p->port_unicast_xmit_packets = cpu_to_be64(pma.n_unicast_xmit
+		- ibp->z_unicast_xmit);
+	p->port_unicast_rcv_packets = cpu_to_be64(pma.n_unicast_rcv
+		- ibp->z_unicast_rcv);
+	p->port_multicast_xmit_packets = cpu_to_be64(pma.n_multicast_xmit
+		- ibp->z_multicast_xmit);
+	p->port_multicast_rcv_packets = cpu_to_be64(pma.n_multicast_rcv
+		- ibp->z_multicast_rcv);
 
 bail:
 	return reply((struct ib_smp *) pmp);
@@ -1795,6 +1820,7 @@
 	struct qib_ibport *ibp = to_iport(ibdev, port);
 	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
 	u64 swords, rwords, spkts, rpkts, xwait;
+	struct qib_pma_counters pma;
 
 	qib_snapshot_counters(ppd, &swords, &rwords, &spkts, &rpkts, &xwait);
 
@@ -1810,17 +1836,19 @@
 	if (p->counter_select & IB_PMA_SELX_PORT_RCV_PACKETS)
 		ibp->z_port_rcv_packets = rpkts;
 
+	qib_snapshot_pmacounters(ibp, &pma);
+
 	if (p->counter_select & IB_PMA_SELX_PORT_UNI_XMIT_PACKETS)
-		ibp->n_unicast_xmit = 0;
+		ibp->z_unicast_xmit = pma.n_unicast_xmit;
 
 	if (p->counter_select & IB_PMA_SELX_PORT_UNI_RCV_PACKETS)
-		ibp->n_unicast_rcv = 0;
+		ibp->z_unicast_rcv = pma.n_unicast_rcv;
 
 	if (p->counter_select & IB_PMA_SELX_PORT_MULTI_XMIT_PACKETS)
-		ibp->n_multicast_xmit = 0;
+		ibp->z_multicast_xmit = pma.n_multicast_xmit;
 
 	if (p->counter_select & IB_PMA_SELX_PORT_MULTI_RCV_PACKETS)
-		ibp->n_multicast_rcv = 0;
+		ibp->z_multicast_rcv = pma.n_multicast_rcv;
 
 	return pma_get_portcounters_ext(pmp, ibdev, port);
 }
diff --git a/drivers/infiniband/hw/qib/qib_rc.c b/drivers/infiniband/hw/qib/qib_rc.c
index 3ab3413..2f25018 100644
--- a/drivers/infiniband/hw/qib/qib_rc.c
+++ b/drivers/infiniband/hw/qib/qib_rc.c
@@ -752,7 +752,7 @@
 	qib_flush_wc();
 	qib_sendbuf_done(dd, pbufn);
 
-	ibp->n_unicast_xmit++;
+	this_cpu_inc(ibp->pmastats->n_unicast_xmit);
 	goto done;
 
 queue_ack:
diff --git a/drivers/infiniband/hw/qib/qib_ruc.c b/drivers/infiniband/hw/qib/qib_ruc.c
index 357b6cfc..4c07a8b 100644
--- a/drivers/infiniband/hw/qib/qib_ruc.c
+++ b/drivers/infiniband/hw/qib/qib_ruc.c
@@ -703,6 +703,7 @@
 	ohdr->bth[0] = cpu_to_be32(bth0);
 	ohdr->bth[1] = cpu_to_be32(qp->remote_qpn);
 	ohdr->bth[2] = cpu_to_be32(bth2);
+	this_cpu_inc(ibp->pmastats->n_unicast_xmit);
 }
 
 /**
diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c
index 3ad651c..aaf7039 100644
--- a/drivers/infiniband/hw/qib/qib_ud.c
+++ b/drivers/infiniband/hw/qib/qib_ud.c
@@ -280,11 +280,11 @@
 	ah_attr = &to_iah(wqe->wr.wr.ud.ah)->attr;
 	if (ah_attr->dlid >= QIB_MULTICAST_LID_BASE) {
 		if (ah_attr->dlid != QIB_PERMISSIVE_LID)
-			ibp->n_multicast_xmit++;
+			this_cpu_inc(ibp->pmastats->n_multicast_xmit);
 		else
-			ibp->n_unicast_xmit++;
+			this_cpu_inc(ibp->pmastats->n_unicast_xmit);
 	} else {
-		ibp->n_unicast_xmit++;
+		this_cpu_inc(ibp->pmastats->n_unicast_xmit);
 		lid = ah_attr->dlid & ~((1 << ppd->lmc) - 1);
 		if (unlikely(lid == ppd->lid)) {
 			/*
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index 092b0bb..1b00734 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -662,7 +662,7 @@
 		mcast = qib_mcast_find(ibp, &hdr->u.l.grh.dgid);
 		if (mcast == NULL)
 			goto drop;
-		ibp->n_multicast_rcv++;
+		this_cpu_inc(ibp->pmastats->n_multicast_rcv);
 		list_for_each_entry_rcu(p, &mcast->qp_list, list)
 			qib_qp_rcv(rcd, hdr, 1, data, tlen, p->qp);
 		/*
@@ -689,7 +689,7 @@
 			rcd->lookaside_qpn = qp_num;
 		} else
 			qp = rcd->lookaside_qp;
-		ibp->n_unicast_rcv++;
+		this_cpu_inc(ibp->pmastats->n_unicast_rcv);
 		qib_qp_rcv(rcd, hdr, lnh == QIB_LRH_GRH, data, tlen, qp);
 	}
 	return;
diff --git a/drivers/infiniband/hw/qib/qib_verbs.h b/drivers/infiniband/hw/qib/qib_verbs.h
index a01c7d2..bfc8948 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.h
+++ b/drivers/infiniband/hw/qib/qib_verbs.h
@@ -664,6 +664,13 @@
 	struct qib_opcode_stats stats[128];
 };
 
+struct qib_pma_counters {
+	u64 n_unicast_xmit;     /* total unicast packets sent */
+	u64 n_unicast_rcv;      /* total unicast packets received */
+	u64 n_multicast_xmit;   /* total multicast packets sent */
+	u64 n_multicast_rcv;    /* total multicast packets received */
+};
+
 struct qib_ibport {
 	struct qib_qp __rcu *qp0;
 	struct qib_qp __rcu *qp1;
@@ -680,10 +687,11 @@
 	__be64 mkey;
 	__be64 guids[QIB_GUIDS_PER_PORT	- 1];	/* writable GUIDs */
 	u64 tid;		/* TID for traps */
-	u64 n_unicast_xmit;     /* total unicast packets sent */
-	u64 n_unicast_rcv;      /* total unicast packets received */
-	u64 n_multicast_xmit;   /* total multicast packets sent */
-	u64 n_multicast_rcv;    /* total multicast packets received */
+	struct qib_pma_counters __percpu *pmastats;
+	u64 z_unicast_xmit;     /* starting count for PMA */
+	u64 z_unicast_rcv;      /* starting count for PMA */
+	u64 z_multicast_xmit;   /* starting count for PMA */
+	u64 z_multicast_rcv;    /* starting count for PMA */
 	u64 z_symbol_error_counter;             /* starting count for PMA */
 	u64 z_link_error_recovery_counter;      /* starting count for PMA */
 	u64 z_link_downed_counter;              /* starting count for PMA */