[PATCH] IB/ipath: fixes to performance get counters for IB compliance

This patch fixes some problems uncovered during IB compliance testing to return the right values for error counters returned by the Performance Get Counters packet. Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com> Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com> Cc: "Michael S. Tsirkin" <mst@mellanox.co.il> Cc: Roland Dreier <rolandd@cisco.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB/ipath: fixes to performance get counters for IB compliance
This patch fixes some problems uncovered during IB compliance testing to return the right values for error counters returned by the Performance Get Counters packet. Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com> Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com> Cc: "Michael S. Tsirkin" <mst@mellanox.co.il> Cc: Roland Dreier <rolandd@cisco.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Bryan O'Sullivan · Linus Torvalds
1 parent 85322947d7
Showing 9 changed files with 93 additions and 27 deletions Side-by-side Diff
drivers/infiniband/hw/ipath/ipath_driver.c
drivers/infiniband/hw/ipath/ipath_intr.c
drivers/infiniband/hw/ipath/ipath_kernel.h
drivers/infiniband/hw/ipath/ipath_layer.c
drivers/infiniband/hw/ipath/ipath_layer.h
drivers/infiniband/hw/ipath/ipath_mad.c
drivers/infiniband/hw/ipath/ipath_ud.c
drivers/infiniband/hw/ipath/ipath_verbs.c
drivers/infiniband/hw/ipath/ipath_verbs.h
@@ -460,6 +460,8 @@
 	 * by ipath_setup_htconfig.
 	 */
 	dd->ipath_flags = 0;
+	dd->ipath_lli_counter = 0;
+	dd->ipath_lli_errors = 0;
  
 	if (dd->ipath_f_bus(dd, pdev))
 		ipath_dev_err(dd, "Failed to setup config space; "
@@ -942,6 +944,18 @@
 				   "tlen=%x opcode=%x egridx=%x: %s\n",
 				   eflags, l, etype, tlen, bthbytes[0],
 				   ips_get_index((__le32 *) rc), emsg);
+			/* Count local link integrity errors. */
+			if (eflags & (INFINIPATH_RHF_H_ICRCERR |
+				      INFINIPATH_RHF_H_VCRCERR)) {
+				u8 n = (dd->ipath_ibcctrl >>
+					INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT) &
+					INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK;
+
+				if (++dd->ipath_lli_counter > n) {
+					dd->ipath_lli_counter = 0;
+					dd->ipath_lli_errors++;
+				}
+			}
 		} else if (etype == RCVHQ_RCV_TYPE_NON_KD) {
 				int ret = __ipath_verbs_rcv(dd, rc + 1,
 							    ebuf, tlen);
@@ -949,6 +963,9 @@
 					ipath_cdbg(VERBOSE,
 						   "received IB packet, "
 						   "not SMA (QP=%x)\n", qp);
+				if (dd->ipath_lli_counter)
+					dd->ipath_lli_counter--;
+
 		} else if (etype == RCVHQ_RCV_TYPE_EAGER) {
 			if (qp == IPATH_KD_QP &&
 			    bthbytes[0] == ipath_layer_rcv_opcode &&
@@ -262,6 +262,7 @@
 				     | IPATH_LINKACTIVE |
 				     IPATH_LINKARMED);
 		*dd->ipath_statusp &= ~IPATH_STATUS_IB_READY;
+		dd->ipath_lli_counter = 0;
 		if (!noprint) {
 			if (((dd->ipath_lastibcstat >>
 			      INFINIPATH_IBCS_LINKSTATE_SHIFT) &
@@ -507,6 +507,11 @@
 	u8 ipath_pci_cacheline;
 	/* LID mask control */
 	u8 ipath_lmc;
+
+	/* local link integrity counter */
+	u32 ipath_lli_counter;
+	/* local link integrity errors */
+	u32 ipath_lli_errors;
 };
  
 extern struct list_head ipath_dev_list;
@@ -1032,19 +1032,22 @@
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_ibsymbolerrcnt);
 	cntrs->link_error_recovery_counter =
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_iblinkerrrecovcnt);
+	/*
+	 * The link downed counter counts when the other side downs the
+	 * connection.  We add in the number of times we downed the link
+	 * due to local link integrity errors to compensate.
+	 */
 	cntrs->link_downed_counter =
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_iblinkdowncnt);
 	cntrs->port_rcv_errors =
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rxdroppktcnt) +
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvovflcnt) +
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_portovflcnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_errrcvflowctrlcnt) +
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_err_rlencnt) +
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_invalidrlencnt) +
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_erricrccnt) +
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_errvcrccnt) +
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_errlpcrccnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_errlinkcnt) +
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_badformatcnt);
 	cntrs->port_rcv_remphys_errors =
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvebpcnt);
@@ -1058,6 +1061,8 @@
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktsendcnt);
 	cntrs->port_rcv_packets =
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktrcvcnt);
+	cntrs->local_link_integrity_errors = dd->ipath_lli_errors;
+	cntrs->excessive_buffer_overrun_errors = 0; /* XXX */
  
 	ret = 0;
  
@@ -55,6 +55,8 @@
 	u64 port_rcv_data;
 	u64 port_xmit_packets;
 	u64 port_rcv_packets;
+	u32 local_link_integrity_errors;
+	u32 excessive_buffer_overrun_errors;
 };
  
 /*
@@ -613,6 +613,9 @@
 #define IB_PMA_SEL_PORT_RCV_ERRORS		__constant_htons(0x0008)
 #define IB_PMA_SEL_PORT_RCV_REMPHYS_ERRORS	__constant_htons(0x0010)
 #define IB_PMA_SEL_PORT_XMIT_DISCARDS		__constant_htons(0x0040)
+#define IB_PMA_SEL_LOCAL_LINK_INTEGRITY_ERRORS	__constant_htons(0x0200)
+#define IB_PMA_SEL_EXCESSIVE_BUFFER_OVERRUNS	__constant_htons(0x0400)
+#define IB_PMA_SEL_PORT_VL15_DROPPED		__constant_htons(0x0800)
 #define IB_PMA_SEL_PORT_XMIT_DATA		__constant_htons(0x1000)
 #define IB_PMA_SEL_PORT_RCV_DATA		__constant_htons(0x2000)
 #define IB_PMA_SEL_PORT_XMIT_PACKETS		__constant_htons(0x4000)
@@ -859,6 +862,10 @@
 	cntrs.port_rcv_data -= dev->z_port_rcv_data;
 	cntrs.port_xmit_packets -= dev->z_port_xmit_packets;
 	cntrs.port_rcv_packets -= dev->z_port_rcv_packets;
+	cntrs.local_link_integrity_errors -=
+		dev->z_local_link_integrity_errors;
+	cntrs.excessive_buffer_overrun_errors -=
+		dev->z_excessive_buffer_overrun_errors;
  
 	memset(pmp->data, 0, sizeof(pmp->data));
  
@@ -896,6 +903,16 @@
 	else
 		p->port_xmit_discards =
 			cpu_to_be16((u16)cntrs.port_xmit_discards);
+	if (cntrs.local_link_integrity_errors > 0xFUL)
+		cntrs.local_link_integrity_errors = 0xFUL;
+	if (cntrs.excessive_buffer_overrun_errors > 0xFUL)
+		cntrs.excessive_buffer_overrun_errors = 0xFUL;
+	p->lli_ebor_errors = (cntrs.local_link_integrity_errors << 4) |
+		cntrs.excessive_buffer_overrun_errors;
+	if (dev->n_vl15_dropped > 0xFFFFUL)
+		p->vl15_dropped = __constant_cpu_to_be16(0xFFFF);
+	else
+		p->vl15_dropped = cpu_to_be16((u16)dev->n_vl15_dropped);
 	if (cntrs.port_xmit_data > 0xFFFFFFFFUL)
 		p->port_xmit_data = __constant_cpu_to_be32(0xFFFFFFFF);
 	else
@@ -990,6 +1007,17 @@
 	if (p->counter_select & IB_PMA_SEL_PORT_XMIT_DISCARDS)
 		dev->z_port_xmit_discards = cntrs.port_xmit_discards;
  
+	if (p->counter_select & IB_PMA_SEL_LOCAL_LINK_INTEGRITY_ERRORS)
+		dev->z_local_link_integrity_errors =
+			cntrs.local_link_integrity_errors;
+
+	if (p->counter_select & IB_PMA_SEL_EXCESSIVE_BUFFER_OVERRUNS)
+		dev->z_excessive_buffer_overrun_errors =
+			cntrs.excessive_buffer_overrun_errors;
+
+	if (p->counter_select & IB_PMA_SEL_PORT_VL15_DROPPED)
+		dev->n_vl15_dropped = 0;
+
 	if (p->counter_select & IB_PMA_SEL_PORT_XMIT_DATA)
 		dev->z_port_xmit_data = cntrs.port_xmit_data;
  
  
@@ -1275,32 +1303,8 @@
 		      struct ib_wc *in_wc, struct ib_grh *in_grh,
 		      struct ib_mad *in_mad, struct ib_mad *out_mad)
 {
-	struct ipath_ibdev *dev = to_idev(ibdev);
 	int ret;
  
-	/*
-	 * Snapshot current HW counters to "clear" them.
-	 * This should be done when the driver is loaded except that for
-	 * some reason we get a zillion errors when brining up the link.
-	 */
-	if (dev->rcv_errors == 0) {
-		struct ipath_layer_counters cntrs;
-
-		ipath_layer_get_counters(to_idev(ibdev)->dd, &cntrs);
-		dev->rcv_errors++;
-		dev->z_symbol_error_counter = cntrs.symbol_error_counter;
-		dev->z_link_error_recovery_counter =
-			cntrs.link_error_recovery_counter;
-		dev->z_link_downed_counter = cntrs.link_downed_counter;
-		dev->z_port_rcv_errors = cntrs.port_rcv_errors + 1;
-		dev->z_port_rcv_remphys_errors =
-			cntrs.port_rcv_remphys_errors;
-		dev->z_port_xmit_discards = cntrs.port_xmit_discards;
-		dev->z_port_xmit_data = cntrs.port_xmit_data;
-		dev->z_port_rcv_data = cntrs.port_rcv_data;
-		dev->z_port_xmit_packets = cntrs.port_xmit_packets;
-		dev->z_port_rcv_packets = cntrs.port_rcv_packets;
-	}
 	switch (in_mad->mad_hdr.mgmt_class) {
 	case IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE:
 	case IB_MGMT_CLASS_SUBN_LID_ROUTED:
@@ -560,7 +560,16 @@
 	spin_lock_irqsave(&rq->lock, flags);
 	if (rq->tail == rq->head) {
 		spin_unlock_irqrestore(&rq->lock, flags);
-		dev->n_pkt_drops++;
+		/*
+		 * Count VL15 packets dropped due to no receive buffer.
+		 * Otherwise, count them as buffer overruns since usually,
+		 * the HW will be able to receive packets even if there are
+		 * no QPs with posted receive buffers.
+		 */
+		if (qp->ibqp.qp_num == 0)
+			dev->n_vl15_dropped++;
+		else
+			dev->rcv_errors++;
 		goto bail;
 	}
 	/* Silently drop packets which are too big. */
@@ -981,6 +981,7 @@
  */
 static void *ipath_register_ib_device(int unit, struct ipath_devdata *dd)
 {
+	struct ipath_layer_counters cntrs;
 	struct ipath_ibdev *idev;
 	struct ib_device *dev;
 	int ret;
@@ -1030,6 +1031,25 @@
 	idev->pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
 	idev->pma_counter_select[5] = IB_PMA_PORT_XMIT_WAIT;
 	idev->link_width_enabled = 3;	/* 1x or 4x */
+
+	/* Snapshot current HW counters to "clear" them. */
+	ipath_layer_get_counters(dd, &cntrs);
+	idev->z_symbol_error_counter = cntrs.symbol_error_counter;
+	idev->z_link_error_recovery_counter =
+		cntrs.link_error_recovery_counter;
+	idev->z_link_downed_counter = cntrs.link_downed_counter;
+	idev->z_port_rcv_errors = cntrs.port_rcv_errors;
+	idev->z_port_rcv_remphys_errors =
+		cntrs.port_rcv_remphys_errors;
+	idev->z_port_xmit_discards = cntrs.port_xmit_discards;
+	idev->z_port_xmit_data = cntrs.port_xmit_data;
+	idev->z_port_rcv_data = cntrs.port_rcv_data;
+	idev->z_port_xmit_packets = cntrs.port_xmit_packets;
+	idev->z_port_rcv_packets = cntrs.port_rcv_packets;
+	idev->z_local_link_integrity_errors =
+		cntrs.local_link_integrity_errors;
+	idev->z_excessive_buffer_overrun_errors =
+		cntrs.excessive_buffer_overrun_errors;
  
 	/*
 	 * The system image GUID is supposed to be the same for all
@@ -460,6 +460,8 @@
 	u64 z_port_xmit_packets;		/* starting count for PMA */
 	u64 z_port_rcv_packets;			/* starting count for PMA */
 	u32 z_pkey_violations;			/* starting count for PMA */
+	u32 z_local_link_integrity_errors;	/* starting count for PMA */
+	u32 z_excessive_buffer_overrun_errors;	/* starting count for PMA */
 	u32 n_rc_resends;
 	u32 n_rc_acks;
 	u32 n_rc_qacks;
@@ -469,6 +471,7 @@
 	u32 n_other_naks;
 	u32 n_timeouts;
 	u32 n_pkt_drops;
+	u32 n_vl15_dropped;
 	u32 n_wqe_errs;
 	u32 n_rdma_dup_busy;
 	u32 n_piowait;
...	...	@@ -460,6 +460,8 @@
460	460	* by ipath_setup_htconfig.
461	461	*/
462	462	dd->ipath_flags = 0;
	463	+ dd->ipath_lli_counter = 0;
	464	+ dd->ipath_lli_errors = 0;
463	465
464	466	if (dd->ipath_f_bus(dd, pdev))
465	467	ipath_dev_err(dd, "Failed to setup config space; "
...	...	@@ -942,6 +944,18 @@
942	944	"tlen=%x opcode=%x egridx=%x: %s\n",
943	945	eflags, l, etype, tlen, bthbytes[0],
944	946	ips_get_index((__le32 *) rc), emsg);
	947	+ /* Count local link integrity errors. */
	948	+ if (eflags & (INFINIPATH_RHF_H_ICRCERR \|
	949	+ INFINIPATH_RHF_H_VCRCERR)) {
	950	+ u8 n = (dd->ipath_ibcctrl >>
	951	+ INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT) &
	952	+ INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK;
	953	+
	954	+ if (++dd->ipath_lli_counter > n) {
	955	+ dd->ipath_lli_counter = 0;
	956	+ dd->ipath_lli_errors++;
	957	+ }
	958	+ }
945	959	} else if (etype == RCVHQ_RCV_TYPE_NON_KD) {
946	960	int ret = __ipath_verbs_rcv(dd, rc + 1,
947	961	ebuf, tlen);
...	...	@@ -949,6 +963,9 @@
949	963	ipath_cdbg(VERBOSE,
950	964	"received IB packet, "
951	965	"not SMA (QP=%x)\n", qp);
	966	+ if (dd->ipath_lli_counter)
	967	+ dd->ipath_lli_counter--;
	968	+
952	969	} else if (etype == RCVHQ_RCV_TYPE_EAGER) {
953	970	if (qp == IPATH_KD_QP &&
954	971	bthbytes[0] == ipath_layer_rcv_opcode &&
...	...	@@ -262,6 +262,7 @@
262	262	\| IPATH_LINKACTIVE \|
263	263	IPATH_LINKARMED);
264	264	*dd->ipath_statusp &= ~IPATH_STATUS_IB_READY;
	265	+ dd->ipath_lli_counter = 0;
265	266	if (!noprint) {
266	267	if (((dd->ipath_lastibcstat >>
267	268	INFINIPATH_IBCS_LINKSTATE_SHIFT) &
...	...	@@ -507,6 +507,11 @@
507	507	u8 ipath_pci_cacheline;
508	508	/* LID mask control */
509	509	u8 ipath_lmc;
	510	+
	511	+ /* local link integrity counter */
	512	+ u32 ipath_lli_counter;
	513	+ /* local link integrity errors */
	514	+ u32 ipath_lli_errors;
510	515	};
511	516
512	517	extern struct list_head ipath_dev_list;
...	...	@@ -1032,19 +1032,22 @@
1032	1032	ipath_snap_cntr(dd, dd->ipath_cregs->cr_ibsymbolerrcnt);
1033	1033	cntrs->link_error_recovery_counter =
1034	1034	ipath_snap_cntr(dd, dd->ipath_cregs->cr_iblinkerrrecovcnt);
	1035	+ /*
	1036	+ * The link downed counter counts when the other side downs the
	1037	+ * connection. We add in the number of times we downed the link
	1038	+ * due to local link integrity errors to compensate.
	1039	+ */
1035	1040	cntrs->link_downed_counter =
1036	1041	ipath_snap_cntr(dd, dd->ipath_cregs->cr_iblinkdowncnt);
1037	1042	cntrs->port_rcv_errors =
1038	1043	ipath_snap_cntr(dd, dd->ipath_cregs->cr_rxdroppktcnt) +
1039	1044	ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvovflcnt) +
1040	1045	ipath_snap_cntr(dd, dd->ipath_cregs->cr_portovflcnt) +
1041		- ipath_snap_cntr(dd, dd->ipath_cregs->cr_errrcvflowctrlcnt) +
1042	1046	ipath_snap_cntr(dd, dd->ipath_cregs->cr_err_rlencnt) +
1043	1047	ipath_snap_cntr(dd, dd->ipath_cregs->cr_invalidrlencnt) +
1044	1048	ipath_snap_cntr(dd, dd->ipath_cregs->cr_erricrccnt) +
1045	1049	ipath_snap_cntr(dd, dd->ipath_cregs->cr_errvcrccnt) +
1046	1050	ipath_snap_cntr(dd, dd->ipath_cregs->cr_errlpcrccnt) +
1047		- ipath_snap_cntr(dd, dd->ipath_cregs->cr_errlinkcnt) +
1048	1051	ipath_snap_cntr(dd, dd->ipath_cregs->cr_badformatcnt);
1049	1052	cntrs->port_rcv_remphys_errors =
1050	1053	ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvebpcnt);
...	...	@@ -1058,6 +1061,8 @@
1058	1061	ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktsendcnt);
1059	1062	cntrs->port_rcv_packets =
1060	1063	ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktrcvcnt);
	1064	+ cntrs->local_link_integrity_errors = dd->ipath_lli_errors;
	1065	+ cntrs->excessive_buffer_overrun_errors = 0; /* XXX */
1061	1066
1062	1067	ret = 0;
1063	1068
...	...	@@ -55,6 +55,8 @@
55	55	u64 port_rcv_data;
56	56	u64 port_xmit_packets;
57	57	u64 port_rcv_packets;
	58	+ u32 local_link_integrity_errors;
	59	+ u32 excessive_buffer_overrun_errors;
58	60	};
59	61
60	62	/*
...	...	@@ -613,6 +613,9 @@
613	613	#define IB_PMA_SEL_PORT_RCV_ERRORS __constant_htons(0x0008)
614	614	#define IB_PMA_SEL_PORT_RCV_REMPHYS_ERRORS __constant_htons(0x0010)
615	615	#define IB_PMA_SEL_PORT_XMIT_DISCARDS __constant_htons(0x0040)
	616	+#define IB_PMA_SEL_LOCAL_LINK_INTEGRITY_ERRORS __constant_htons(0x0200)
	617	+#define IB_PMA_SEL_EXCESSIVE_BUFFER_OVERRUNS __constant_htons(0x0400)
	618	+#define IB_PMA_SEL_PORT_VL15_DROPPED __constant_htons(0x0800)
616	619	#define IB_PMA_SEL_PORT_XMIT_DATA __constant_htons(0x1000)
617	620	#define IB_PMA_SEL_PORT_RCV_DATA __constant_htons(0x2000)
618	621	#define IB_PMA_SEL_PORT_XMIT_PACKETS __constant_htons(0x4000)
...	...	@@ -859,6 +862,10 @@
859	862	cntrs.port_rcv_data -= dev->z_port_rcv_data;
860	863	cntrs.port_xmit_packets -= dev->z_port_xmit_packets;
861	864	cntrs.port_rcv_packets -= dev->z_port_rcv_packets;
	865	+ cntrs.local_link_integrity_errors -=
	866	+ dev->z_local_link_integrity_errors;
	867	+ cntrs.excessive_buffer_overrun_errors -=
	868	+ dev->z_excessive_buffer_overrun_errors;
862	869
863	870	memset(pmp->data, 0, sizeof(pmp->data));
864	871
...	...	@@ -896,6 +903,16 @@
896	903	else
897	904	p->port_xmit_discards =
898	905	cpu_to_be16((u16)cntrs.port_xmit_discards);
	906	+ if (cntrs.local_link_integrity_errors > 0xFUL)
	907	+ cntrs.local_link_integrity_errors = 0xFUL;
	908	+ if (cntrs.excessive_buffer_overrun_errors > 0xFUL)
	909	+ cntrs.excessive_buffer_overrun_errors = 0xFUL;
	910	+ p->lli_ebor_errors = (cntrs.local_link_integrity_errors << 4) \|
	911	+ cntrs.excessive_buffer_overrun_errors;
	912	+ if (dev->n_vl15_dropped > 0xFFFFUL)
	913	+ p->vl15_dropped = __constant_cpu_to_be16(0xFFFF);
	914	+ else
	915	+ p->vl15_dropped = cpu_to_be16((u16)dev->n_vl15_dropped);
899	916	if (cntrs.port_xmit_data > 0xFFFFFFFFUL)
900	917	p->port_xmit_data = __constant_cpu_to_be32(0xFFFFFFFF);
901	918	else
...	...	@@ -990,6 +1007,17 @@
990	1007	if (p->counter_select & IB_PMA_SEL_PORT_XMIT_DISCARDS)
991	1008	dev->z_port_xmit_discards = cntrs.port_xmit_discards;
992	1009
	1010	+ if (p->counter_select & IB_PMA_SEL_LOCAL_LINK_INTEGRITY_ERRORS)
	1011	+ dev->z_local_link_integrity_errors =
	1012	+ cntrs.local_link_integrity_errors;
	1013	+
	1014	+ if (p->counter_select & IB_PMA_SEL_EXCESSIVE_BUFFER_OVERRUNS)
	1015	+ dev->z_excessive_buffer_overrun_errors =
	1016	+ cntrs.excessive_buffer_overrun_errors;
	1017	+
	1018	+ if (p->counter_select & IB_PMA_SEL_PORT_VL15_DROPPED)
	1019	+ dev->n_vl15_dropped = 0;
	1020	+
993	1021	if (p->counter_select & IB_PMA_SEL_PORT_XMIT_DATA)
994	1022	dev->z_port_xmit_data = cntrs.port_xmit_data;
995	1023
996	1024
...	...	@@ -1275,32 +1303,8 @@
1275	1303	struct ib_wc in_wc, struct ib_grh in_grh,
1276	1304	struct ib_mad in_mad, struct ib_mad out_mad)
1277	1305	{
1278		- struct ipath_ibdev *dev = to_idev(ibdev);
1279	1306	int ret;
1280	1307
1281		- /*
1282		- * Snapshot current HW counters to "clear" them.
1283		- * This should be done when the driver is loaded except that for
1284		- * some reason we get a zillion errors when brining up the link.
1285		- */
1286		- if (dev->rcv_errors == 0) {
1287		- struct ipath_layer_counters cntrs;
1288		-
1289		- ipath_layer_get_counters(to_idev(ibdev)->dd, &cntrs);
1290		- dev->rcv_errors++;
1291		- dev->z_symbol_error_counter = cntrs.symbol_error_counter;
1292		- dev->z_link_error_recovery_counter =
1293		- cntrs.link_error_recovery_counter;
1294		- dev->z_link_downed_counter = cntrs.link_downed_counter;
1295		- dev->z_port_rcv_errors = cntrs.port_rcv_errors + 1;
1296		- dev->z_port_rcv_remphys_errors =
1297		- cntrs.port_rcv_remphys_errors;
1298		- dev->z_port_xmit_discards = cntrs.port_xmit_discards;
1299		- dev->z_port_xmit_data = cntrs.port_xmit_data;
1300		- dev->z_port_rcv_data = cntrs.port_rcv_data;
1301		- dev->z_port_xmit_packets = cntrs.port_xmit_packets;
1302		- dev->z_port_rcv_packets = cntrs.port_rcv_packets;
1303		- }
1304	1308	switch (in_mad->mad_hdr.mgmt_class) {
1305	1309	case IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE:
1306	1310	case IB_MGMT_CLASS_SUBN_LID_ROUTED:
...	...	@@ -560,7 +560,16 @@
560	560	spin_lock_irqsave(&rq->lock, flags);
561	561	if (rq->tail == rq->head) {
562	562	spin_unlock_irqrestore(&rq->lock, flags);
563		- dev->n_pkt_drops++;
	563	+ /*
	564	+ * Count VL15 packets dropped due to no receive buffer.
	565	+ * Otherwise, count them as buffer overruns since usually,
	566	+ * the HW will be able to receive packets even if there are
	567	+ * no QPs with posted receive buffers.
	568	+ */
	569	+ if (qp->ibqp.qp_num == 0)
	570	+ dev->n_vl15_dropped++;
	571	+ else
	572	+ dev->rcv_errors++;
564	573	goto bail;
565	574	}
566	575	/* Silently drop packets which are too big. */
...	...	@@ -981,6 +981,7 @@
981	981	*/
982	982	static void ipath_register_ib_device(int unit, struct ipath_devdata dd)
983	983	{
	984	+ struct ipath_layer_counters cntrs;
984	985	struct ipath_ibdev *idev;
985	986	struct ib_device *dev;
986	987	int ret;
...	...	@@ -1030,6 +1031,25 @@
1030	1031	idev->pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
1031	1032	idev->pma_counter_select[5] = IB_PMA_PORT_XMIT_WAIT;
1032	1033	idev->link_width_enabled = 3; /* 1x or 4x */
	1034	+
	1035	+ /* Snapshot current HW counters to "clear" them. */
	1036	+ ipath_layer_get_counters(dd, &cntrs);
	1037	+ idev->z_symbol_error_counter = cntrs.symbol_error_counter;
	1038	+ idev->z_link_error_recovery_counter =
	1039	+ cntrs.link_error_recovery_counter;
	1040	+ idev->z_link_downed_counter = cntrs.link_downed_counter;
	1041	+ idev->z_port_rcv_errors = cntrs.port_rcv_errors;
	1042	+ idev->z_port_rcv_remphys_errors =
	1043	+ cntrs.port_rcv_remphys_errors;
	1044	+ idev->z_port_xmit_discards = cntrs.port_xmit_discards;
	1045	+ idev->z_port_xmit_data = cntrs.port_xmit_data;
	1046	+ idev->z_port_rcv_data = cntrs.port_rcv_data;
	1047	+ idev->z_port_xmit_packets = cntrs.port_xmit_packets;
	1048	+ idev->z_port_rcv_packets = cntrs.port_rcv_packets;
	1049	+ idev->z_local_link_integrity_errors =
	1050	+ cntrs.local_link_integrity_errors;
	1051	+ idev->z_excessive_buffer_overrun_errors =
	1052	+ cntrs.excessive_buffer_overrun_errors;
1033	1053
1034	1054	/*
1035	1055	* The system image GUID is supposed to be the same for all
...	...	@@ -460,6 +460,8 @@
460	460	u64 z_port_xmit_packets; /* starting count for PMA */
461	461	u64 z_port_rcv_packets; /* starting count for PMA */
462	462	u32 z_pkey_violations; /* starting count for PMA */
	463	+ u32 z_local_link_integrity_errors; /* starting count for PMA */
	464	+ u32 z_excessive_buffer_overrun_errors; /* starting count for PMA */
463	465	u32 n_rc_resends;
464	466	u32 n_rc_acks;
465	467	u32 n_rc_qacks;
...	...	@@ -469,6 +471,7 @@
469	471	u32 n_other_naks;
470	472	u32 n_timeouts;
471	473	u32 n_pkt_drops;
	474	+ u32 n_vl15_dropped;
472	475	u32 n_wqe_errs;
473	476	u32 n_rdma_dup_busy;
474	477	u32 n_piowait;