Merge git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

48 files changed, 2234 insertions, 886 deletions

diff --git a/drivers/infiniband/hw/cxgb3/Makefile b/drivers/infiniband/hw/cxgb3/Makefile
index 36b98989b15e..7e7b5a66f042 100644
--- a/drivers/infiniband/hw/cxgb3/Makefile
+++ b/drivers/infiniband/hw/cxgb3/Makefile
@@ -1,5 +1,4 @@
-EXTRA_CFLAGS += -I$(TOPDIR)/drivers/net/cxgb3 \
-		-I$(TOPDIR)/drivers/infiniband/hw/cxgb3/core
+EXTRA_CFLAGS += -Idrivers/net/cxgb3
 
 obj-$(CONFIG_INFINIBAND_CXGB3) += iw_cxgb3.o
 
diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c
index eec6a30840ca..03c5ff62889a 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.c
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c
@@ -179,7 +179,7 @@ int cxio_create_cq(struct cxio_rdev *rdev_p, struct t3_cq *cq)
 	setup.size = 1UL << cq->size_log2;
 	setup.credits = 65535;
 	setup.credit_thres = 1;
-	if (rdev_p->t3cdev_p->type == T3B)
+	if (rdev_p->t3cdev_p->type != T3A)
 		setup.ovfl_mode = 0;
 	else
 		setup.ovfl_mode = 1;
@@ -584,7 +584,7 @@ static int cxio_hal_ctrl_qp_write_mem(struct cxio_rdev *rdev_p, u32 addr,
 {
 	u32 i, nr_wqe, copy_len;
 	u8 *copy_data;
-	u8 wr_len, utx_len;	/* lenght in 8 byte flit */
+	u8 wr_len, utx_len;	/* length in 8 byte flit */
 	enum t3_wr_flags flag;
 	__be64 *wqe;
 	u64 utx_cmd;
diff --git a/drivers/infiniband/hw/cxgb3/cxio_wr.h b/drivers/infiniband/hw/cxgb3/cxio_wr.h
index c84d4ac49355..969d4d928455 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_wr.h
+++ b/drivers/infiniband/hw/cxgb3/cxio_wr.h
@@ -315,7 +315,7 @@ struct t3_rdma_init_wr {
 	__be32 ird;
 	__be64 qp_dma_addr;	/* 7 */
 	__be32 qp_dma_size;	/* 8 */
-	u32 irs;
+	__be32 irs;
 };
 
 struct t3_genbit {
@@ -324,7 +324,8 @@ struct t3_genbit {
 };
 
 enum rdma_init_wr_flags {
-	RECVS_POSTED = 1,
+	RECVS_POSTED = (1<<0),
+	PRIV_QP = (1<<1),
 };
 
 union t3_wr {
diff --git a/drivers/infiniband/hw/cxgb3/iwch_cm.c b/drivers/infiniband/hw/cxgb3/iwch_cm.c
index 20ba372dd182..e9a08fa3dffe 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_cm.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_cm.c
@@ -332,7 +332,7 @@ static struct rtable *find_route(struct t3cdev *dev, __be32 local_ip,
 			  }
 	};
 
-	if (ip_route_output_flow(&rt, &fl, NULL, 0))
+	if (ip_route_output_flow(&init_net, &rt, &fl, NULL, 0))
 		return NULL;
 	return rt;
 }
@@ -1118,7 +1118,7 @@ static int act_open_rpl(struct t3cdev *tdev, struct sk_buff *skb, void *ctx)
 	     status2errno(rpl->status));
 	connect_reply_upcall(ep, status2errno(rpl->status));
 	state_set(&ep->com, DEAD);
-	if (ep->com.tdev->type == T3B && act_open_has_tid(rpl->status))
+	if (ep->com.tdev->type != T3A && act_open_has_tid(rpl->status))
 		release_tid(ep->com.tdev, GET_TID(rpl), NULL);
 	cxgb3_free_atid(ep->com.tdev, ep->atid);
 	dst_release(ep->dst);
@@ -1249,7 +1249,7 @@ static void reject_cr(struct t3cdev *tdev, u32 hwtid, __be32 peer_ip,
 	skb_trim(skb, sizeof(struct cpl_tid_release));
 	skb_get(skb);
 
-	if (tdev->type == T3B)
+	if (tdev->type != T3A)
 		release_tid(tdev, hwtid, skb);
 	else {
 		struct cpl_pass_accept_rpl *rpl;
diff --git a/drivers/infiniband/hw/cxgb3/iwch_mem.c b/drivers/infiniband/hw/cxgb3/iwch_mem.c
index a6c2c4ba29e6..73bfd1656f86 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_mem.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_mem.c
@@ -122,6 +122,13 @@ int build_phys_page_list(struct ib_phys_buf *buffer_list,
 		*total_size += buffer_list[i].size;
 		if (i > 0)
 			mask |= buffer_list[i].addr;
+		else
+			mask |= buffer_list[i].addr & PAGE_MASK;
+		if (i != num_phys_buf - 1)
+			mask |= buffer_list[i].addr + buffer_list[i].size;
+		else
+			mask |= (buffer_list[i].addr + buffer_list[i].size +
+				PAGE_SIZE - 1) & PAGE_MASK;
 	}
 
 	if (*total_size > 0xFFFFFFFFULL)
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index f0c777589374..df1838f8f94d 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -39,6 +39,7 @@
 #include <linux/list.h>
 #include <linux/spinlock.h>
 #include <linux/ethtool.h>
+#include <linux/rtnetlink.h>
 
 #include <asm/io.h>
 #include <asm/irq.h>
@@ -645,7 +646,7 @@ static struct ib_mr *iwch_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	if (err)
 		goto err;
 
-	if (udata && t3b_device(rhp)) {
+	if (udata && !t3a_device(rhp)) {
 		uresp.pbl_addr = (mhp->attr.pbl_addr -
 	                         rhp->rdev.rnic_info.pbl_base) >> 3;
 		PDBG("%s user resp pbl_addr 0x%x\n", __FUNCTION__,
@@ -1000,6 +1001,7 @@ static int iwch_query_device(struct ib_device *ibdev,
 	props->max_sge = dev->attr.max_sge_per_wr;
 	props->max_sge_rd = 1;
 	props->max_qp_rd_atom = dev->attr.max_rdma_reads_per_qp;
+	props->max_qp_init_rd_atom = dev->attr.max_rdma_reads_per_qp;
 	props->max_cq = dev->attr.max_cqs;
 	props->max_cqe = dev->attr.max_cqes_per_cq;
 	props->max_mr = dev->attr.max_mem_regs;
@@ -1052,7 +1054,9 @@ static ssize_t show_fw_ver(struct class_device *cdev, char *buf)
 	struct net_device *lldev = dev->rdev.t3cdev_p->lldev;
 
 	PDBG("%s class dev 0x%p\n", __FUNCTION__, cdev);
+	rtnl_lock();
 	lldev->ethtool_ops->get_drvinfo(lldev, &info);
+	rtnl_unlock();
 	return sprintf(buf, "%s\n", info.fw_version);
 }
 
@@ -1064,7 +1068,9 @@ static ssize_t show_hca(struct class_device *cdev, char *buf)
 	struct net_device *lldev = dev->rdev.t3cdev_p->lldev;
 
 	PDBG("%s class dev 0x%p\n", __FUNCTION__, cdev);
+	rtnl_lock();
 	lldev->ethtool_ops->get_drvinfo(lldev, &info);
+	rtnl_unlock();
 	return sprintf(buf, "%s\n", info.driver);
 }
 
diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c
index dd89b6b91f9c..ea2cdd73dd85 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_qp.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c
@@ -208,36 +208,19 @@ static int iwch_sgl2pbl_map(struct iwch_dev *rhp, struct ib_sge *sg_list,
 static int iwch_build_rdma_recv(struct iwch_dev *rhp, union t3_wr *wqe,
 				struct ib_recv_wr *wr)
 {
-	int i, err = 0;
-	u32 pbl_addr[4];
-	u8 page_size[4];
+	int i;
 	if (wr->num_sge > T3_MAX_SGE)
 		return -EINVAL;
-	err = iwch_sgl2pbl_map(rhp, wr->sg_list, wr->num_sge, pbl_addr,
-			       page_size);
-	if (err)
-		return err;
-	wqe->recv.pagesz[0] = page_size[0];
-	wqe->recv.pagesz[1] = page_size[1];
-	wqe->recv.pagesz[2] = page_size[2];
-	wqe->recv.pagesz[3] = page_size[3];
 	wqe->recv.num_sgle = cpu_to_be32(wr->num_sge);
 	for (i = 0; i < wr->num_sge; i++) {
 		wqe->recv.sgl[i].stag = cpu_to_be32(wr->sg_list[i].lkey);
 		wqe->recv.sgl[i].len = cpu_to_be32(wr->sg_list[i].length);
-
-		/* to in the WQE == the offset into the page */
-		wqe->recv.sgl[i].to = cpu_to_be64(((u32) wr->sg_list[i].addr) %
-				(1UL << (12 + page_size[i])));
-
-		/* pbl_addr is the adapters address in the PBL */
-		wqe->recv.pbl_addr[i] = cpu_to_be32(pbl_addr[i]);
+		wqe->recv.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr);
 	}
 	for (; i < T3_MAX_SGE; i++) {
 		wqe->recv.sgl[i].stag = 0;
 		wqe->recv.sgl[i].len = 0;
 		wqe->recv.sgl[i].to = 0;
-		wqe->recv.pbl_addr[i] = 0;
 	}
 	return 0;
 }
@@ -659,6 +642,7 @@ static void __flush_qp(struct iwch_qp *qhp, unsigned long *flag)
 	cxio_flush_rq(&qhp->wq, &rchp->cq, count);
 	spin_unlock(&qhp->lock);
 	spin_unlock_irqrestore(&rchp->lock, *flag);
+	(*rchp->ibcq.comp_handler)(&rchp->ibcq, rchp->ibcq.cq_context);
 
 	/* locking heirarchy: cq lock first, then qp lock. */
 	spin_lock_irqsave(&schp->lock, *flag);
@@ -668,6 +652,7 @@ static void __flush_qp(struct iwch_qp *qhp, unsigned long *flag)
 	cxio_flush_sq(&qhp->wq, &schp->cq, count);
 	spin_unlock(&qhp->lock);
 	spin_unlock_irqrestore(&schp->lock, *flag);
+	(*schp->ibcq.comp_handler)(&schp->ibcq, schp->ibcq.cq_context);
 
 	/* deref */
 	if (atomic_dec_and_test(&qhp->refcnt))
@@ -678,7 +663,7 @@ static void __flush_qp(struct iwch_qp *qhp, unsigned long *flag)
 
 static void flush_qp(struct iwch_qp *qhp, unsigned long *flag)
 {
-	if (t3b_device(qhp->rhp))
+	if (qhp->ibqp.uobject)
 		cxio_set_wq_in_error(&qhp->wq);
 	else
 		__flush_qp(qhp, flag);
@@ -732,6 +717,7 @@ static int rdma_init(struct iwch_dev *rhp, struct iwch_qp *qhp,
 	init_attr.qp_dma_addr = qhp->wq.dma_addr;
 	init_attr.qp_dma_size = (1UL << qhp->wq.size_log2);
 	init_attr.flags = rqes_posted(qhp) ? RECVS_POSTED : 0;
+	init_attr.flags |= capable(CAP_NET_BIND_SERVICE) ? PRIV_QP : 0;
 	init_attr.irs = qhp->ep->rcv_seq;
 	PDBG("%s init_attr.rq_addr 0x%x init_attr.rq_size = %d "
 	     "flags 0x%x qpcaps 0x%x\n", __FUNCTION__,
@@ -847,10 +833,11 @@ int iwch_modify_qp(struct iwch_dev *rhp, struct iwch_qp *qhp,
 				disconnect = 1;
 				ep = qhp->ep;
 			}
+			flush_qp(qhp, &flag);
 			break;
 		case IWCH_QP_STATE_TERMINATE:
 			qhp->attr.state = IWCH_QP_STATE_TERMINATE;
-			if (t3b_device(qhp->rhp))
+			if (qhp->ibqp.uobject)
 				cxio_set_wq_in_error(&qhp->wq);
 			if (!internal)
 				terminate = 1;
diff --git a/drivers/infiniband/hw/ehca/ehca_av.c b/drivers/infiniband/hw/ehca/ehca_av.c
index 97d108634c58..194c1c30cf63 100644
--- a/drivers/infiniband/hw/ehca/ehca_av.c
+++ b/drivers/infiniband/hw/ehca/ehca_av.c
@@ -1,7 +1,7 @@
 /*
  *  IBM eServer eHCA Infiniband device driver for Linux on POWER
  *
- *  adress vector functions
+ *  address vector functions
  *
  *  Authors: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
  *           Khadija Souissi <souissik@de.ibm.com>
@@ -50,6 +50,42 @@
 
 static struct kmem_cache *av_cache;
 
+int ehca_calc_ipd(struct ehca_shca *shca, int port,
+		  enum ib_rate path_rate, u32 *ipd)
+{
+	int path = ib_rate_to_mult(path_rate);
+	int link, ret;
+	struct ib_port_attr pa;
+
+	if (path_rate == IB_RATE_PORT_CURRENT) {
+		*ipd = 0;
+		return 0;
+	}
+
+	if (unlikely(path < 0)) {
+		ehca_err(&shca->ib_device, "Invalid static rate! path_rate=%x",
+			 path_rate);
+		return -EINVAL;
+	}
+
+	ret = ehca_query_port(&shca->ib_device, port, &pa);
+	if (unlikely(ret < 0)) {
+		ehca_err(&shca->ib_device, "Failed to query port  ret=%i", ret);
+		return ret;
+	}
+
+	link = ib_width_enum_to_int(pa.active_width) * pa.active_speed;
+
+	if (path >= link)
+		/* no need to throttle if path faster than link */
+		*ipd = 0;
+	else
+		/* IPD = round((link / path) - 1) */
+		*ipd = ((link + (path >> 1)) / path) - 1;
+
+	return 0;
+}
+
 struct ib_ah *ehca_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr)
 {
 	int ret;
@@ -69,15 +105,13 @@ struct ib_ah *ehca_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr)
 	av->av.slid_path_bits = ah_attr->src_path_bits;
 
 	if (ehca_static_rate < 0) {
-		int ah_mult = ib_rate_to_mult(ah_attr->static_rate);
-		int ehca_mult =
-			ib_rate_to_mult(shca->sport[ah_attr->port_num].rate );
-
-		if (ah_mult >= ehca_mult)
-			av->av.ipd = 0;
-		else
-			av->av.ipd = (ah_mult > 0) ?
-				((ehca_mult - 1) / ah_mult) : 0;
+		u32 ipd;
+		if (ehca_calc_ipd(shca, ah_attr->port_num,
+				  ah_attr->static_rate, &ipd)) {
+			ret = -EINVAL;
+			goto create_ah_exit1;
+		}
+		av->av.ipd = ipd;
 	} else
 		av->av.ipd = ehca_static_rate;
 
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h b/drivers/infiniband/hw/ehca/ehca_classes.h
index 2d660ae189e5..f281d16040f5 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -94,8 +94,11 @@ struct ehca_sma_attr {
 
 struct ehca_sport {
 	struct ib_cq *ibcq_aqp1;
-	struct ib_qp *ibqp_aqp1;
-	enum ib_rate  rate;
+	struct ib_qp *ibqp_sqp[2];
+	/* lock to serialze modify_qp() calls for sqp in normal
+	 * and irq path (when event PORT_ACTIVE is received first time)
+	 */
+	spinlock_t mod_sqp_lock;
 	enum ib_port_state port_state;
 	struct ehca_sma_attr saved_attr;
 };
@@ -142,6 +145,14 @@ enum ehca_ext_qp_type {
 	EQPT_SRQ       = 3,
 };
 
+/* struct to cache modify_qp()'s parms for GSI/SMI qp */
+struct ehca_mod_qp_parm {
+	int mask;
+	struct ib_qp_attr attr;
+};
+
+#define EHCA_MOD_QP_PARM_MAX 4
+
 struct ehca_qp {
 	union {
 		struct ib_qp ib_qp;
@@ -165,10 +176,18 @@ struct ehca_qp {
 	struct ehca_cq *recv_cq;
 	unsigned int sqerr_purgeflag;
 	struct hlist_node list_entries;
+	/* array to cache modify_qp()'s parms for GSI/SMI qp */
+	struct ehca_mod_qp_parm *mod_qp_parm;
+	int mod_qp_parm_idx;
 	/* mmap counter for resources mapped into user space */
 	u32 mm_count_squeue;
 	u32 mm_count_rqueue;
 	u32 mm_count_galpa;
+	/* unsolicited ack circumvention */
+	int unsol_ack_circ;
+	int mtu_shift;
+	u32 message_count;
+	u32 packet_count;
 };
 
 #define IS_SRQ(qp) (qp->ext_type == EQPT_SRQ)
@@ -323,6 +342,8 @@ extern int ehca_static_rate;
 extern int ehca_port_act_time;
 extern int ehca_use_hp_mr;
 extern int ehca_scaling_code;
+extern int ehca_lock_hcalls;
+extern int ehca_nr_ports;
 
 struct ipzu_queue_resp {
 	u32 qe_size;      /* queue entry size */
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c b/drivers/infiniband/hw/ehca/ehca_cq.c
index 79c25f51c21e..0467c158d4a9 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -246,7 +246,7 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int cqe, int comp_vector,
 		} else {
 			if (h_ret != H_PAGE_REGISTERED) {
 				ehca_err(device, "Registration of page failed "
-					 "ehca_cq=%p cq_num=%x h_ret=%li"
+					 "ehca_cq=%p cq_num=%x h_ret=%li "
 					 "counter=%i act_pages=%i",
 					 my_cq, my_cq->cq_number,
 					 h_ret, counter, param.act_pages);
diff --git a/drivers/infiniband/hw/ehca/ehca_hca.c b/drivers/infiniband/hw/ehca/ehca_hca.c
index 15806d140461..5bd7b591987e 100644
--- a/drivers/infiniband/hw/ehca/ehca_hca.c
+++ b/drivers/infiniband/hw/ehca/ehca_hca.c
@@ -151,7 +151,6 @@ int ehca_query_port(struct ib_device *ibdev,
 	}
 
 	memset(props, 0, sizeof(struct ib_port_attr));
-	props->state = rblock->state;
 
 	switch (rblock->max_mtu) {
 	case 0x1:
@@ -188,11 +187,20 @@ int ehca_query_port(struct ib_device *ibdev,
 	props->subnet_timeout  = rblock->subnet_timeout;
 	props->init_type_reply = rblock->init_type_reply;
 
-	props->active_width    = IB_WIDTH_12X;
-	props->active_speed    = 0x1;
-
-	/* at the moment (logical) link state is always LINK_UP */
-	props->phys_state      = 0x5;
+	if (rblock->state && rblock->phys_width) {
+		props->phys_state      = rblock->phys_pstate;
+		props->state           = rblock->phys_state;
+		props->active_width    = rblock->phys_width;
+		props->active_speed    = rblock->phys_speed;
+	} else {
+		/* old firmware releases don't report physical
+		 * port info, so use default values
+		 */
+		props->phys_state      = 5;
+		props->state           = rblock->state;
+		props->active_width    = IB_WIDTH_12X;
+		props->active_speed    = 0x1;
+	}
 
 query_port1:
 	ehca_free_fw_ctrlblock(rblock);
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c b/drivers/infiniband/hw/ehca/ehca_irq.c
index 3f617b27b954..863b34fa9ff9 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -62,6 +62,7 @@
 #define NEQE_PORT_NUMBER       EHCA_BMASK_IBM( 8, 15)
 #define NEQE_PORT_AVAILABILITY EHCA_BMASK_IBM(16, 16)
 #define NEQE_DISRUPTIVE        EHCA_BMASK_IBM(16, 16)
+#define NEQE_SPECIFIC_EVENT    EHCA_BMASK_IBM(16, 23)
 
 #define ERROR_DATA_LENGTH      EHCA_BMASK_IBM(52, 63)
 #define ERROR_DATA_TYPE        EHCA_BMASK_IBM( 0,  7)
@@ -354,17 +355,34 @@ static void parse_ec(struct ehca_shca *shca, u64 eqe)
 {
 	u8 ec   = EHCA_BMASK_GET(NEQE_EVENT_CODE, eqe);
 	u8 port = EHCA_BMASK_GET(NEQE_PORT_NUMBER, eqe);
+	u8 spec_event;
+	struct ehca_sport *sport = &shca->sport[port - 1];
+	unsigned long flags;
 
 	switch (ec) {
 	case 0x30: /* port availability change */
 		if (EHCA_BMASK_GET(NEQE_PORT_AVAILABILITY, eqe)) {
-			shca->sport[port - 1].port_state = IB_PORT_ACTIVE;
+			int suppress_event;
+			/* replay modify_qp for sqps */
+			spin_lock_irqsave(&sport->mod_sqp_lock, flags);
+			suppress_event = !sport->ibqp_sqp[IB_QPT_GSI];
+			if (sport->ibqp_sqp[IB_QPT_SMI])
+				ehca_recover_sqp(sport->ibqp_sqp[IB_QPT_SMI]);
+			if (!suppress_event)
+				ehca_recover_sqp(sport->ibqp_sqp[IB_QPT_GSI]);
+			spin_unlock_irqrestore(&sport->mod_sqp_lock, flags);
+
+			/* AQP1 was destroyed, ignore this event */
+			if (suppress_event)
+				break;
+
+			sport->port_state = IB_PORT_ACTIVE;
 			dispatch_port_event(shca, port, IB_EVENT_PORT_ACTIVE,
 					    "is active");
 			ehca_query_sma_attr(shca, port,
-					    &shca->sport[port - 1].saved_attr);
+					    &sport->saved_attr);
 		} else {
-			shca->sport[port - 1].port_state = IB_PORT_DOWN;
+			sport->port_state = IB_PORT_DOWN;
 			dispatch_port_event(shca, port, IB_EVENT_PORT_ERR,
 					    "is inactive");
 		}
@@ -378,11 +396,11 @@ static void parse_ec(struct ehca_shca *shca, u64 eqe)
 			ehca_warn(&shca->ib_device, "disruptive port "
 				  "%d configuration change", port);
 
-			shca->sport[port - 1].port_state = IB_PORT_DOWN;
+			sport->port_state = IB_PORT_DOWN;
 			dispatch_port_event(shca, port, IB_EVENT_PORT_ERR,
 					    "is inactive");
 
-			shca->sport[port - 1].port_state = IB_PORT_ACTIVE;
+			sport->port_state = IB_PORT_ACTIVE;
 			dispatch_port_event(shca, port, IB_EVENT_PORT_ACTIVE,
 					    "is active");
 		} else
@@ -394,6 +412,16 @@ static void parse_ec(struct ehca_shca *shca, u64 eqe)
 	case 0x33:  /* trace stopped */
 		ehca_err(&shca->ib_device, "Traced stopped.");
 		break;
+	case 0x34: /* util async event */
+		spec_event = EHCA_BMASK_GET(NEQE_SPECIFIC_EVENT, eqe);
+		if (spec_event == 0x80) /* client reregister required */
+			dispatch_port_event(shca, port,
+					    IB_EVENT_CLIENT_REREGISTER,
+					    "client reregister req.");
+		else
+			ehca_warn(&shca->ib_device, "Unknown util async "
+				  "event %x on port %x", spec_event, port);
+		break;
 	default:
 		ehca_err(&shca->ib_device, "Unknown event code: %x on %s.",
 			 ec, shca->ib_device.name);
diff --git a/drivers/infiniband/hw/ehca/ehca_iverbs.h b/drivers/infiniband/hw/ehca/ehca_iverbs.h
index dce503bb7d6b..c469bfde2708 100644
--- a/drivers/infiniband/hw/ehca/ehca_iverbs.h
+++ b/drivers/infiniband/hw/ehca/ehca_iverbs.h
@@ -189,6 +189,9 @@ int ehca_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
 
 void ehca_poll_eqs(unsigned long data);
 
+int ehca_calc_ipd(struct ehca_shca *shca, int port,
+		  enum ib_rate path_rate, u32 *ipd);
+
 #ifdef CONFIG_PPC_64K_PAGES
 void *ehca_alloc_fw_ctrlblock(gfp_t flags);
 void ehca_free_fw_ctrlblock(void *ptr);
@@ -197,4 +200,6 @@ void ehca_free_fw_ctrlblock(void *ptr);
 #define ehca_free_fw_ctrlblock(ptr) free_page((unsigned long)(ptr))
 #endif
 
+void ehca_recover_sqp(struct ib_qp *sqp);
+
 #endif
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c b/drivers/infiniband/hw/ehca/ehca_main.c
index c6cd38c5321f..84c9b7b8669b 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -43,13 +43,14 @@
 #ifdef CONFIG_PPC_64K_PAGES
 #include <linux/slab.h>
 #endif
+
 #include "ehca_classes.h"
 #include "ehca_iverbs.h"
 #include "ehca_mrmw.h"
 #include "ehca_tools.h"
 #include "hcp_if.h"
 
-#define HCAD_VERSION "0024"
+#define HCAD_VERSION "0025"
 
 MODULE_LICENSE("Dual BSD/GPL");
 MODULE_AUTHOR("Christoph Raisch <raisch@de.ibm.com>");
@@ -66,6 +67,7 @@ int ehca_poll_all_eqs  = 1;
 int ehca_static_rate   = -1;
 int ehca_scaling_code  = 0;
 int ehca_mr_largepage  = 1;
+int ehca_lock_hcalls   = -1;
 
 module_param_named(open_aqp1,     ehca_open_aqp1,     int, S_IRUGO);
 module_param_named(debug_level,   ehca_debug_level,   int, S_IRUGO);
@@ -77,6 +79,7 @@ module_param_named(poll_all_eqs,  ehca_poll_all_eqs,  int, S_IRUGO);
 module_param_named(static_rate,   ehca_static_rate,   int, S_IRUGO);
 module_param_named(scaling_code,  ehca_scaling_code,  int, S_IRUGO);
 module_param_named(mr_largepage,  ehca_mr_largepage,  int, S_IRUGO);
+module_param_named(lock_hcalls,   ehca_lock_hcalls,   bool, S_IRUGO);
 
 MODULE_PARM_DESC(open_aqp1,
 		 "AQP1 on startup (0: no (default), 1: yes)");
@@ -87,7 +90,8 @@ MODULE_PARM_DESC(hw_level,
 		 "hardware level"
 		 " (0: autosensing (default), 1: v. 0.20, 2: v. 0.21)");
 MODULE_PARM_DESC(nr_ports,
-		 "number of connected ports (default: 2)");
+		 "number of connected ports (-1: autodetect, 1: port one only, "
+		 "2: two ports (default)");
 MODULE_PARM_DESC(use_hp_mr,
 		 "high performance MRs (0: no (default), 1: yes)");
 MODULE_PARM_DESC(port_act_time,
@@ -102,6 +106,9 @@ MODULE_PARM_DESC(scaling_code,
 MODULE_PARM_DESC(mr_largepage,
 		 "use large page for MR (0: use PAGE_SIZE (default), "
 		 "1: use large page depending on MR size");
+MODULE_PARM_DESC(lock_hcalls,
+		 "serialize all hCalls made by the driver "
+		 "(default: autodetect)");
 
 DEFINE_RWLOCK(ehca_qp_idr_lock);
 DEFINE_RWLOCK(ehca_cq_idr_lock);
@@ -258,6 +265,7 @@ static struct cap_descr {
 	{ HCA_CAP_UD_LL_QP, "HCA_CAP_UD_LL_QP" },
 	{ HCA_CAP_RESIZE_MR, "HCA_CAP_RESIZE_MR" },
 	{ HCA_CAP_MINI_QP, "HCA_CAP_MINI_QP" },
+	{ HCA_CAP_H_ALLOC_RES_SYNC, "HCA_CAP_H_ALLOC_RES_SYNC" },
 };
 
 static int ehca_sense_attributes(struct ehca_shca *shca)
@@ -327,15 +335,18 @@ static int ehca_sense_attributes(struct ehca_shca *shca)
 		shca->hw_level = ehca_hw_level;
 	ehca_gen_dbg(" ... hardware level=%x", shca->hw_level);
 
-	shca->sport[0].rate = IB_RATE_30_GBPS;
-	shca->sport[1].rate = IB_RATE_30_GBPS;
-
 	shca->hca_cap = rblock->hca_cap_indicators;
 	ehca_gen_dbg(" ... HCA capabilities:");
 	for (i = 0; i < ARRAY_SIZE(hca_cap_descr); i++)
 		if (EHCA_BMASK_GET(hca_cap_descr[i].mask, shca->hca_cap))
 			ehca_gen_dbg("   %s", hca_cap_descr[i].descr);
 
+	/* Autodetect hCall locking -- the "H_ALLOC_RESOURCE synced" flag is
+	 * a firmware property, so it's valid across all adapters
+	 */
+	if (ehca_lock_hcalls == -1)
+		ehca_lock_hcalls = !(shca->hca_cap & HCA_CAP_H_ALLOC_RES_SYNC);
+
 	/* translate supported MR page sizes; always support 4K */
 	shca->hca_cap_mr_pgsize = EHCA_PAGESIZE;
 	if (ehca_mr_largepage) { /* support extra sizes only if enabled */
@@ -501,7 +512,7 @@ static int ehca_create_aqp1(struct ehca_shca *shca, u32 port)
 	}
 	sport->ibcq_aqp1 = ibcq;
 
-	if (sport->ibqp_aqp1) {
+	if (sport->ibqp_sqp[IB_QPT_GSI]) {
 		ehca_err(&shca->ib_device, "AQP1 QP is already created.");
 		ret = -EPERM;
 		goto create_aqp1;
@@ -527,7 +538,7 @@ static int ehca_create_aqp1(struct ehca_shca *shca, u32 port)
 		ret = PTR_ERR(ibqp);
 		goto create_aqp1;
 	}
-	sport->ibqp_aqp1 = ibqp;
+	sport->ibqp_sqp[IB_QPT_GSI] = ibqp;
 
 	return 0;
 
@@ -540,7 +551,7 @@ static int ehca_destroy_aqp1(struct ehca_sport *sport)
 {
 	int ret;
 
-	ret = ib_destroy_qp(sport->ibqp_aqp1);
+	ret = ib_destroy_qp(sport->ibqp_sqp[IB_QPT_GSI]);
 	if (ret) {
 		ehca_gen_err("Cannot destroy AQP1 QP. ret=%i", ret);
 		return ret;
@@ -580,6 +591,11 @@ static struct attribute_group ehca_drv_attr_grp = {
 	.attrs = ehca_drv_attrs
 };
 
+static struct attribute_group *ehca_drv_attr_groups[] = {
+	&ehca_drv_attr_grp,
+	NULL,
+};
+
 #define EHCA_RESOURCE_ATTR(name)                                           \
 static ssize_t  ehca_show_##name(struct device *dev,                       \
 				 struct device_attribute *attr,            \
@@ -678,7 +694,7 @@ static int __devinit ehca_probe(struct of_device *dev,
 	struct ehca_shca *shca;
 	const u64 *handle;
 	struct ib_pd *ibpd;
-	int ret;
+	int ret, i;
 
 	handle = of_get_property(dev->node, "ibm,hca-handle", NULL);
 	if (!handle) {
@@ -699,6 +715,8 @@ static int __devinit ehca_probe(struct of_device *dev,
 		return -ENOMEM;
 	}
 	mutex_init(&shca->modify_mutex);
+	for (i = 0; i < ARRAY_SIZE(shca->sport); i++)
+		spin_lock_init(&shca->sport[i].mod_sqp_lock);
 
 	shca->ofdev = dev;
 	shca->ipz_hca_handle.handle = *handle;
@@ -889,6 +907,9 @@ static struct of_platform_driver ehca_driver = {
 	.match_table = ehca_device_table,
 	.probe       = ehca_probe,
 	.remove      = ehca_remove,
+	.driver	     = {
+		.groups = ehca_drv_attr_groups,
+	},
 };
 
 void ehca_poll_eqs(unsigned long data)
@@ -916,7 +937,7 @@ void ehca_poll_eqs(unsigned long data)
 				ehca_process_eq(shca, 0);
 		}
 	}
-	mod_timer(&poll_eqs_timer, jiffies + HZ);
+	mod_timer(&poll_eqs_timer, round_jiffies(jiffies + HZ));
 	spin_unlock(&shca_list_lock);
 }
 
@@ -947,10 +968,6 @@ int __init ehca_module_init(void)
 		goto module_init2;
 	}
 
-	ret = sysfs_create_group(&ehca_driver.driver.kobj, &ehca_drv_attr_grp);
-	if (ret) /* only complain; we can live without attributes */
-		ehca_gen_err("Cannot create driver attributes  ret=%d", ret);
-
 	if (ehca_poll_all_eqs != 1) {
 		ehca_gen_err("WARNING!!!");
 		ehca_gen_err("It is possible to lose interrupts.");
@@ -976,7 +993,6 @@ void __exit ehca_module_exit(void)
 	if (ehca_poll_all_eqs == 1)
 		del_timer_sync(&poll_eqs_timer);
 
-	sysfs_remove_group(&ehca_driver.driver.kobj, &ehca_drv_attr_grp);
 	ibmebus_unregister_driver(&ehca_driver);
 
 	ehca_destroy_slab_caches();
diff --git a/drivers/infiniband/hw/ehca/ehca_qp.c b/drivers/infiniband/hw/ehca/ehca_qp.c
index de182648b282..1012f15a7140 100644
--- a/drivers/infiniband/hw/ehca/ehca_qp.c
+++ b/drivers/infiniband/hw/ehca/ehca_qp.c
@@ -592,10 +592,8 @@ static struct ehca_qp *internal_create_qp(
 		goto create_qp_exit1;
 	}
 
-	if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
-		parms.sigtype = HCALL_SIGT_EVERY;
-	else
-		parms.sigtype = HCALL_SIGT_BY_WQE;
+	/* Always signal by WQE so we can hide circ. WQEs */
+	parms.sigtype = HCALL_SIGT_BY_WQE;
 
 	/* UD_AV CIRCUMVENTION */
 	max_send_sge = init_attr->cap.max_send_sge;
@@ -618,6 +616,10 @@ static struct ehca_qp *internal_create_qp(
 	parms.squeue.max_sge = max_send_sge;
 	parms.rqueue.max_sge = max_recv_sge;
 
+	/* RC QPs need one more SWQE for unsolicited ack circumvention */
+	if (qp_type == IB_QPT_RC)
+		parms.squeue.max_wr++;
+
 	if (EHCA_BMASK_GET(HCA_CAP_MINI_QP, shca->hca_cap)) {
 		if (HAS_SQ(my_qp))
 			ehca_determine_small_queue(
@@ -650,6 +652,8 @@ static struct ehca_qp *internal_create_qp(
 			parms.squeue.act_nr_sges = 1;
 			parms.rqueue.act_nr_sges = 1;
 		}
+		/* hide the extra WQE */
+		parms.squeue.act_nr_wqes--;
 		break;
 	case IB_QPT_UD:
 	case IB_QPT_GSI:
@@ -729,12 +733,31 @@ static struct ehca_qp *internal_create_qp(
 	init_attr->cap.max_send_wr = parms.squeue.act_nr_wqes;
 	my_qp->init_attr = *init_attr;
 
+	if (qp_type == IB_QPT_SMI || qp_type == IB_QPT_GSI) {
+		shca->sport[init_attr->port_num - 1].ibqp_sqp[qp_type] =
+			&my_qp->ib_qp;
+		if (ehca_nr_ports < 0) {
+			/* alloc array to cache subsequent modify qp parms
+			 * for autodetect mode
+			 */
+			my_qp->mod_qp_parm =
+				kzalloc(EHCA_MOD_QP_PARM_MAX *
+					sizeof(*my_qp->mod_qp_parm),
+					GFP_KERNEL);
+			if (!my_qp->mod_qp_parm) {
+				ehca_err(pd->device,
+					 "Could not alloc mod_qp_parm");
+				goto create_qp_exit4;
+			}
+		}
+	}
+
 	/* NOTE: define_apq0() not supported yet */
 	if (qp_type == IB_QPT_GSI) {
 		h_ret = ehca_define_sqp(shca, my_qp, init_attr);
 		if (h_ret != H_SUCCESS) {
 			ret = ehca2ib_return_code(h_ret);
-			goto create_qp_exit4;
+			goto create_qp_exit5;
 		}
 	}
 
@@ -743,7 +766,7 @@ static struct ehca_qp *internal_create_qp(
 		if (ret) {
 			ehca_err(pd->device,
 				 "Couldn't assign qp to send_cq ret=%i", ret);
-			goto create_qp_exit4;
+			goto create_qp_exit5;
 		}
 	}
 
@@ -769,12 +792,18 @@ static struct ehca_qp *internal_create_qp(
 		if (ib_copy_to_udata(udata, &resp, sizeof resp)) {
 			ehca_err(pd->device, "Copy to udata failed");
 			ret = -EINVAL;
-			goto create_qp_exit4;
+			goto create_qp_exit6;
 		}
 	}
 
 	return my_qp;
 
+create_qp_exit6:
+	ehca_cq_unassign_qp(my_qp->send_cq, my_qp->real_qp_num);
+
+create_qp_exit5:
+	kfree(my_qp->mod_qp_parm);
+
 create_qp_exit4:
 	if (HAS_RQ(my_qp))
 		ipz_queue_dtor(my_pd, &my_qp->ipz_rqueue);
@@ -838,7 +867,7 @@ struct ib_srq *ehca_create_srq(struct ib_pd *pd,
 
 	/* copy back return values */
 	srq_init_attr->attr.max_wr = qp_init_attr.cap.max_recv_wr;
-	srq_init_attr->attr.max_sge = qp_init_attr.cap.max_recv_sge;
+	srq_init_attr->attr.max_sge = 3;
 
 	/* drive SRQ into RTR state */
 	mqpcb = ehca_alloc_fw_ctrlblock(GFP_KERNEL);
@@ -858,7 +887,7 @@ struct ib_srq *ehca_create_srq(struct ib_pd *pd,
 				update_mask,
 				mqpcb, my_qp->galpas.kernel);
 	if (hret != H_SUCCESS) {
-		ehca_err(pd->device, "Could not modify SRQ to INIT"
+		ehca_err(pd->device, "Could not modify SRQ to INIT "
 			 "ehca_qp=%p qp_num=%x h_ret=%li",
 			 my_qp, my_qp->real_qp_num, hret);
 		goto create_srq2;
@@ -872,7 +901,7 @@ struct ib_srq *ehca_create_srq(struct ib_pd *pd,
 				update_mask,
 				mqpcb, my_qp->galpas.kernel);
 	if (hret != H_SUCCESS) {
-		ehca_err(pd->device, "Could not enable SRQ"
+		ehca_err(pd->device, "Could not enable SRQ "
 			 "ehca_qp=%p qp_num=%x h_ret=%li",
 			 my_qp, my_qp->real_qp_num, hret);
 		goto create_srq2;
@@ -886,7 +915,7 @@ struct ib_srq *ehca_create_srq(struct ib_pd *pd,
 				update_mask,
 				mqpcb, my_qp->galpas.kernel);
 	if (hret != H_SUCCESS) {
-		ehca_err(pd->device, "Could not modify SRQ to RTR"
+		ehca_err(pd->device, "Could not modify SRQ to RTR "
 			 "ehca_qp=%p qp_num=%x h_ret=%li",
 			 my_qp, my_qp->real_qp_num, hret);
 		goto create_srq2;
@@ -992,7 +1021,7 @@ static int internal_modify_qp(struct ib_qp *ibqp,
 	unsigned long flags = 0;
 
 	/* do query_qp to obtain current attr values */
-	mqpcb = ehca_alloc_fw_ctrlblock(GFP_KERNEL);
+	mqpcb = ehca_alloc_fw_ctrlblock(GFP_ATOMIC);
 	if (!mqpcb) {
 		ehca_err(ibqp->device, "Could not get zeroed page for mqpcb "
 			 "ehca_qp=%p qp_num=%x ", my_qp, ibqp->qp_num);
@@ -1180,6 +1209,8 @@ static int internal_modify_qp(struct ib_qp *ibqp,
 		update_mask |= EHCA_BMASK_SET(MQPCB_MASK_PRIM_P_KEY_IDX, 1);
 	}
 	if (attr_mask & IB_QP_PORT) {
+		struct ehca_sport *sport;
+		struct ehca_qp *aqp1;
 		if (attr->port_num < 1 || attr->port_num > shca->num_ports) {
 			ret = -EINVAL;
 			ehca_err(ibqp->device, "Invalid port=%x. "
@@ -1188,6 +1219,29 @@ static int internal_modify_qp(struct ib_qp *ibqp,
 				 shca->num_ports);
 			goto modify_qp_exit2;
 		}
+		sport = &shca->sport[attr->port_num - 1];
+		if (!sport->ibqp_sqp[IB_QPT_GSI]) {
+			/* should not occur */
+			ret = -EFAULT;
+			ehca_err(ibqp->device, "AQP1 was not created for "
+				 "port=%x", attr->port_num);
+			goto modify_qp_exit2;
+		}
+		aqp1 = container_of(sport->ibqp_sqp[IB_QPT_GSI],
+				    struct ehca_qp, ib_qp);
+		if (ibqp->qp_type != IB_QPT_GSI &&
+		    ibqp->qp_type != IB_QPT_SMI &&
+		    aqp1->mod_qp_parm) {
+			/*
+			 * firmware will reject this modify_qp() because
+			 * port is not activated/initialized fully
+			 */
+			ret = -EFAULT;
+			ehca_warn(ibqp->device, "Couldn't modify qp port=%x: "
+				  "either port is being activated (try again) "
+				  "or cabling issue", attr->port_num);
+			goto modify_qp_exit2;
+		}
 		mqpcb->prim_phys_port = attr->port_num;
 		update_mask |= EHCA_BMASK_SET(MQPCB_MASK_PRIM_PHYS_PORT, 1);
 	}
@@ -1196,10 +1250,6 @@ static int internal_modify_qp(struct ib_qp *ibqp,
 		update_mask |= EHCA_BMASK_SET(MQPCB_MASK_QKEY, 1);
 	}
 	if (attr_mask & IB_QP_AV) {
-		int ah_mult = ib_rate_to_mult(attr->ah_attr.static_rate);
-		int ehca_mult = ib_rate_to_mult(shca->sport[my_qp->
-						init_attr.port_num].rate);
-
 		mqpcb->dlid = attr->ah_attr.dlid;
 		update_mask |= EHCA_BMASK_SET(MQPCB_MASK_DLID, 1);
 		mqpcb->source_path_bits = attr->ah_attr.src_path_bits;
@@ -1207,11 +1257,12 @@ static int internal_modify_qp(struct ib_qp *ibqp,
 		mqpcb->service_level = attr->ah_attr.sl;
 		update_mask |= EHCA_BMASK_SET(MQPCB_MASK_SERVICE_LEVEL, 1);
 
-		if (ah_mult < ehca_mult)
-			mqpcb->max_static_rate = (ah_mult > 0) ?
-			((ehca_mult - 1) / ah_mult) : 0;
-		else
-			mqpcb->max_static_rate = 0;
+		if (ehca_calc_ipd(shca, mqpcb->prim_phys_port,
+				  attr->ah_attr.static_rate,
+				  &mqpcb->max_static_rate)) {
+			ret = -EINVAL;
+			goto modify_qp_exit2;
+		}
 		update_mask |= EHCA_BMASK_SET(MQPCB_MASK_MAX_STATIC_RATE, 1);
 
 		/*
@@ -1247,6 +1298,8 @@ static int internal_modify_qp(struct ib_qp *ibqp,
 	}
 
 	if (attr_mask & IB_QP_PATH_MTU) {
+		/* store ld(MTU) */
+		my_qp->mtu_shift = attr->path_mtu + 7;
 		mqpcb->path_mtu = attr->path_mtu;
 		update_mask |= EHCA_BMASK_SET(MQPCB_MASK_PATH_MTU, 1);
 	}
@@ -1280,10 +1333,6 @@ static int internal_modify_qp(struct ib_qp *ibqp,
 			(MQPCB_MASK_RDMA_ATOMIC_OUTST_DEST_QP, 1);
 	}
 	if (attr_mask & IB_QP_ALT_PATH) {
-		int ah_mult = ib_rate_to_mult(attr->alt_ah_attr.static_rate);
-		int ehca_mult = ib_rate_to_mult(
-			shca->sport[my_qp->init_attr.port_num].rate);
-
 		if (attr->alt_port_num < 1
 		    || attr->alt_port_num > shca->num_ports) {
 			ret = -EINVAL;
@@ -1309,10 +1358,12 @@ static int internal_modify_qp(struct ib_qp *ibqp,
 		mqpcb->source_path_bits_al = attr->alt_ah_attr.src_path_bits;
 		mqpcb->service_level_al = attr->alt_ah_attr.sl;
 
-		if (ah_mult > 0 && ah_mult < ehca_mult)
-			mqpcb->max_static_rate_al = (ehca_mult - 1) / ah_mult;
-		else
-			mqpcb->max_static_rate_al = 0;
+		if (ehca_calc_ipd(shca, mqpcb->alt_phys_port,
+				  attr->alt_ah_attr.static_rate,
+				  &mqpcb->max_static_rate_al)) {
+			ret = -EINVAL;
+			goto modify_qp_exit2;
+		}
 
 		/* OpenIB doesn't support alternate retry counts - copy them */
 		mqpcb->retry_count_al = mqpcb->retry_count;
@@ -1472,6 +1523,8 @@ modify_qp_exit1:
 int ehca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask,
 		   struct ib_udata *udata)
 {
+	struct ehca_shca *shca = container_of(ibqp->device, struct ehca_shca,
+					      ib_device);
 	struct ehca_qp *my_qp = container_of(ibqp, struct ehca_qp, ib_qp);
 	struct ehca_pd *my_pd = container_of(my_qp->ib_qp.pd, struct ehca_pd,
 					     ib_pd);
@@ -1484,9 +1537,100 @@ int ehca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask,
 		return -EINVAL;
 	}
 
+	/* The if-block below caches qp_attr to be modified for GSI and SMI
+	 * qps during the initialization by ib_mad. When the respective port
+	 * is activated, ie we got an event PORT_ACTIVE, we'll replay the
+	 * cached modify calls sequence, see ehca_recover_sqs() below.
+	 * Why that is required:
+	 * 1) If one port is connected, older code requires that port one
+	 *    to be connected and module option nr_ports=1 to be given by
+	 *    user, which is very inconvenient for end user.
+	 * 2) Firmware accepts modify_qp() only if respective port has become
+	 *    active. Older code had a wait loop of 30sec create_qp()/
+	 *    define_aqp1(), which is not appropriate in practice. This
+	 *    code now removes that wait loop, see define_aqp1(), and always
+	 *    reports all ports to ib_mad resp. users. Only activated ports
+	 *    will then usable for the users.
+	 */
+	if (ibqp->qp_type == IB_QPT_GSI || ibqp->qp_type == IB_QPT_SMI) {
+		int port = my_qp->init_attr.port_num;
+		struct ehca_sport *sport = &shca->sport[port - 1];
+		unsigned long flags;
+		spin_lock_irqsave(&sport->mod_sqp_lock, flags);
+		/* cache qp_attr only during init */
+		if (my_qp->mod_qp_parm) {
+			struct ehca_mod_qp_parm *p;
+			if (my_qp->mod_qp_parm_idx >= EHCA_MOD_QP_PARM_MAX) {
+				ehca_err(&shca->ib_device,
+					 "mod_qp_parm overflow state=%x port=%x"
+					 " type=%x", attr->qp_state,
+					 my_qp->init_attr.port_num,
+					 ibqp->qp_type);
+				spin_unlock_irqrestore(&sport->mod_sqp_lock,
+						       flags);
+				return -EINVAL;
+			}
+			p = &my_qp->mod_qp_parm[my_qp->mod_qp_parm_idx];
+			p->mask = attr_mask;
+			p->attr = *attr;
+			my_qp->mod_qp_parm_idx++;
+			ehca_dbg(&shca->ib_device,
+				 "Saved qp_attr for state=%x port=%x type=%x",
+				 attr->qp_state, my_qp->init_attr.port_num,
+				 ibqp->qp_type);
+			spin_unlock_irqrestore(&sport->mod_sqp_lock, flags);
+			return 0;
+		}
+		spin_unlock_irqrestore(&sport->mod_sqp_lock, flags);
+	}
+
 	return internal_modify_qp(ibqp, attr, attr_mask, 0);
 }
 
+void ehca_recover_sqp(struct ib_qp *sqp)
+{
+	struct ehca_qp *my_sqp = container_of(sqp, struct ehca_qp, ib_qp);
+	int port = my_sqp->init_attr.port_num;
+	struct ib_qp_attr attr;
+	struct ehca_mod_qp_parm *qp_parm;
+	int i, qp_parm_idx, ret;
+	unsigned long flags, wr_cnt;
+
+	if (!my_sqp->mod_qp_parm)
+		return;
+	ehca_dbg(sqp->device, "SQP port=%x qp_num=%x", port, sqp->qp_num);
+
+	qp_parm = my_sqp->mod_qp_parm;
+	qp_parm_idx = my_sqp->mod_qp_parm_idx;
+	for (i = 0; i < qp_parm_idx; i++) {
+		attr = qp_parm[i].attr;
+		ret = internal_modify_qp(sqp, &attr, qp_parm[i].mask, 0);
+		if (ret) {
+			ehca_err(sqp->device, "Could not modify SQP port=%x "
+				 "qp_num=%x ret=%x", port, sqp->qp_num, ret);
+			goto free_qp_parm;
+		}
+		ehca_dbg(sqp->device, "SQP port=%x qp_num=%x in state=%x",
+			 port, sqp->qp_num, attr.qp_state);
+	}
+
+	/* re-trigger posted recv wrs */
+	wr_cnt =  my_sqp->ipz_rqueue.current_q_offset /
+		my_sqp->ipz_rqueue.qe_size;
+	if (wr_cnt) {
+		spin_lock_irqsave(&my_sqp->spinlock_r, flags);
+		hipz_update_rqa(my_sqp, wr_cnt);
+		spin_unlock_irqrestore(&my_sqp->spinlock_r, flags);
+		ehca_dbg(sqp->device, "doorbell port=%x qp_num=%x wr_cnt=%lx",
+			 port, sqp->qp_num, wr_cnt);
+	}
+
+free_qp_parm:
+	kfree(qp_parm);
+	/* this prevents subsequent calls to modify_qp() to cache qp_attr */
+	my_sqp->mod_qp_parm = NULL;
+}
+
 int ehca_query_qp(struct ib_qp *qp,
 		  struct ib_qp_attr *qp_attr,
 		  int qp_attr_mask, struct ib_qp_init_attr *qp_init_attr)
@@ -1755,7 +1899,7 @@ int ehca_query_srq(struct ib_srq *srq, struct ib_srq_attr *srq_attr)
 	}
 
 	srq_attr->max_wr = qpcb->max_nr_outst_recv_wr - 1;
-	srq_attr->max_sge = qpcb->actual_nr_sges_in_rq_wqe;
+	srq_attr->max_sge = 3;
 	srq_attr->srq_limit = EHCA_BMASK_GET(
 		MQPCB_CURR_SRQ_LIMIT, qpcb->curr_srq_limit);
 
@@ -1774,6 +1918,7 @@ static int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
 	struct ehca_shca *shca = container_of(dev, struct ehca_shca, ib_device);
 	struct ehca_pd *my_pd = container_of(my_qp->ib_qp.pd, struct ehca_pd,
 					     ib_pd);
+	struct ehca_sport *sport = &shca->sport[my_qp->init_attr.port_num - 1];
 	u32 cur_pid = current->tgid;
 	u32 qp_num = my_qp->real_qp_num;
 	int ret;
@@ -1820,6 +1965,14 @@ static int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
 	port_num = my_qp->init_attr.port_num;
 	qp_type  = my_qp->init_attr.qp_type;
 
+	if (qp_type == IB_QPT_SMI || qp_type == IB_QPT_GSI) {
+		spin_lock_irqsave(&sport->mod_sqp_lock, flags);
+		kfree(my_qp->mod_qp_parm);
+		my_qp->mod_qp_parm = NULL;
+		shca->sport[port_num - 1].ibqp_sqp[qp_type] = NULL;
+		spin_unlock_irqrestore(&sport->mod_sqp_lock, flags);
+	}
+
 	/* no support for IB_QPT_SMI yet */
 	if (qp_type == IB_QPT_GSI) {
 		struct ib_event event;
diff --git a/drivers/infiniband/hw/ehca/ehca_reqs.c b/drivers/infiniband/hw/ehca/ehca_reqs.c
index ea91360835d3..3aacc8cf1e44 100644
--- a/drivers/infiniband/hw/ehca/ehca_reqs.c
+++ b/drivers/infiniband/hw/ehca/ehca_reqs.c
@@ -50,6 +50,9 @@
 #include "hcp_if.h"
 #include "hipz_fns.h"
 
+/* in RC traffic, insert an empty RDMA READ every this many packets */
+#define ACK_CIRC_THRESHOLD 2000000
+
 static inline int ehca_write_rwqe(struct ipz_queue *ipz_rqueue,
 				  struct ehca_wqe *wqe_p,
 				  struct ib_recv_wr *recv_wr)
@@ -81,7 +84,7 @@ static inline int ehca_write_rwqe(struct ipz_queue *ipz_rqueue,
 	if (ehca_debug_level) {
 		ehca_gen_dbg("RECEIVE WQE written into ipz_rqueue=%p",
 			     ipz_rqueue);
-		ehca_dmp( wqe_p, 16*(6 + wqe_p->nr_of_data_seg), "recv wqe");
+		ehca_dmp(wqe_p, 16*(6 + wqe_p->nr_of_data_seg), "recv wqe");
 	}
 
 	return 0;
@@ -135,7 +138,8 @@ static void trace_send_wr_ud(const struct ib_send_wr *send_wr)
 
 static inline int ehca_write_swqe(struct ehca_qp *qp,
 				  struct ehca_wqe *wqe_p,
-				  const struct ib_send_wr *send_wr)
+				  const struct ib_send_wr *send_wr,
+				  int hidden)
 {
 	u32 idx;
 	u64 dma_length;
@@ -176,7 +180,9 @@ static inline int ehca_write_swqe(struct ehca_qp *qp,
 
 	wqe_p->wr_flag = 0;
 
-	if (send_wr->send_flags & IB_SEND_SIGNALED)
+	if ((send_wr->send_flags & IB_SEND_SIGNALED ||
+	    qp->init_attr.sq_sig_type == IB_SIGNAL_ALL_WR)
+	    && !hidden)
 		wqe_p->wr_flag |= WQE_WRFLAG_REQ_SIGNAL_COM;
 
 	if (send_wr->opcode == IB_WR_SEND_WITH_IMM ||
@@ -199,7 +205,7 @@ static inline int ehca_write_swqe(struct ehca_qp *qp,
 
 		wqe_p->destination_qp_number = send_wr->wr.ud.remote_qpn << 8;
 		wqe_p->local_ee_context_qkey = remote_qkey;
-		if (!send_wr->wr.ud.ah) {
+		if (unlikely(!send_wr->wr.ud.ah)) {
 			ehca_gen_err("wr.ud.ah is NULL. qp=%p", qp);
 			return -EINVAL;
 		}
@@ -255,6 +261,15 @@ static inline int ehca_write_swqe(struct ehca_qp *qp,
 		} /* eof idx */
 		wqe_p->u.nud.atomic_1st_op_dma_len = dma_length;
 
+		/* unsolicited ack circumvention */
+		if (send_wr->opcode == IB_WR_RDMA_READ) {
+			/* on RDMA read, switch on and reset counters */
+			qp->message_count = qp->packet_count = 0;
+			qp->unsol_ack_circ = 1;
+		} else
+			/* else estimate #packets */
+			qp->packet_count += (dma_length >> qp->mtu_shift) + 1;
+
 		break;
 
 	default:
@@ -355,13 +370,49 @@ static inline void map_ib_wc_status(u32 cqe_status,
 		*wc_status = IB_WC_SUCCESS;
 }
 
+static inline int post_one_send(struct ehca_qp *my_qp,
+			 struct ib_send_wr *cur_send_wr,
+			 struct ib_send_wr **bad_send_wr,
+			 int hidden)
+{
+	struct ehca_wqe *wqe_p;
+	int ret;
+	u64 start_offset = my_qp->ipz_squeue.current_q_offset;
+
+	/* get pointer next to free WQE */
+	wqe_p = ipz_qeit_get_inc(&my_qp->ipz_squeue);
+	if (unlikely(!wqe_p)) {
+		/* too many posted work requests: queue overflow */
+		if (bad_send_wr)
+			*bad_send_wr = cur_send_wr;
+		ehca_err(my_qp->ib_qp.device, "Too many posted WQEs "
+			 "qp_num=%x", my_qp->ib_qp.qp_num);
+		return -ENOMEM;
+	}
+	/* write a SEND WQE into the QUEUE */
+	ret = ehca_write_swqe(my_qp, wqe_p, cur_send_wr, hidden);
+	/*
+	 * if something failed,
+	 * reset the free entry pointer to the start value
+	 */
+	if (unlikely(ret)) {
+		my_qp->ipz_squeue.current_q_offset = start_offset;
+		if (bad_send_wr)
+			*bad_send_wr = cur_send_wr;
+		ehca_err(my_qp->ib_qp.device, "Could not write WQE "
+			 "qp_num=%x", my_qp->ib_qp.qp_num);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 int ehca_post_send(struct ib_qp *qp,
 		   struct ib_send_wr *send_wr,
 		   struct ib_send_wr **bad_send_wr)
 {
 	struct ehca_qp *my_qp = container_of(qp, struct ehca_qp, ib_qp);
 	struct ib_send_wr *cur_send_wr;
-	struct ehca_wqe *wqe_p;
 	int wqe_cnt = 0;
 	int ret = 0;
 	unsigned long flags;
@@ -369,37 +420,33 @@ int ehca_post_send(struct ib_qp *qp,
 	/* LOCK the QUEUE */
 	spin_lock_irqsave(&my_qp->spinlock_s, flags);
 
+	/* Send an empty extra RDMA read if:
+	 *  1) there has been an RDMA read on this connection before
+	 *  2) no RDMA read occurred for ACK_CIRC_THRESHOLD link packets
+	 *  3) we can be sure that any previous extra RDMA read has been
+	 *     processed so we don't overflow the SQ
+	 */
+	if (unlikely(my_qp->unsol_ack_circ &&
+		     my_qp->packet_count > ACK_CIRC_THRESHOLD &&
+		     my_qp->message_count > my_qp->init_attr.cap.max_send_wr)) {
+		/* insert an empty RDMA READ to fix up the remote QP state */
+		struct ib_send_wr circ_wr;
+		memset(&circ_wr, 0, sizeof(circ_wr));
+		circ_wr.opcode = IB_WR_RDMA_READ;
+		post_one_send(my_qp, &circ_wr, NULL, 1); /* ignore retcode */
+		wqe_cnt++;
+		ehca_dbg(qp->device, "posted circ wr  qp_num=%x", qp->qp_num);
+		my_qp->message_count = my_qp->packet_count = 0;
+	}
+
 	/* loop processes list of send reqs */
 	for (cur_send_wr = send_wr; cur_send_wr != NULL;
 	     cur_send_wr = cur_send_wr->next) {
-		u64 start_offset = my_qp->ipz_squeue.current_q_offset;
-		/* get pointer next to free WQE */
-		wqe_p = ipz_qeit_get_inc(&my_qp->ipz_squeue);
-		if (unlikely(!wqe_p)) {
-			/* too many posted work requests: queue overflow */
-			if (bad_send_wr)
-				*bad_send_wr = cur_send_wr;
-			if (wqe_cnt == 0) {
-				ret = -ENOMEM;
-				ehca_err(qp->device, "Too many posted WQEs "
-					 "qp_num=%x", qp->qp_num);
-			}
-			goto post_send_exit0;
-		}
-		/* write a SEND WQE into the QUEUE */
-		ret = ehca_write_swqe(my_qp, wqe_p, cur_send_wr);
-		/*
-		 * if something failed,
-		 * reset the free entry pointer to the start value
-		 */
+		ret = post_one_send(my_qp, cur_send_wr, bad_send_wr, 0);
 		if (unlikely(ret)) {
-			my_qp->ipz_squeue.current_q_offset = start_offset;
-			*bad_send_wr = cur_send_wr;
-			if (wqe_cnt == 0) {
-				ret = -EINVAL;
-				ehca_err(qp->device, "Could not write WQE "
-					 "qp_num=%x", qp->qp_num);
-			}
+			/* if one or more WQEs were successful, don't fail */
+			if (wqe_cnt)
+				ret = 0;
 			goto post_send_exit0;
 		}
 		wqe_cnt++;
@@ -410,6 +457,7 @@ int ehca_post_send(struct ib_qp *qp,
 post_send_exit0:
 	iosync(); /* serialize GAL register access */
 	hipz_update_sqa(my_qp, wqe_cnt);
+	my_qp->message_count += wqe_cnt;
 	spin_unlock_irqrestore(&my_qp->spinlock_s, flags);
 	return ret;
 }
diff --git a/drivers/infiniband/hw/ehca/ehca_sqp.c b/drivers/infiniband/hw/ehca/ehca_sqp.c
index f0792e5fbd02..79e72b25b252 100644
--- a/drivers/infiniband/hw/ehca/ehca_sqp.c
+++ b/drivers/infiniband/hw/ehca/ehca_sqp.c
@@ -40,11 +40,8 @@
  */
 
 
-#include <linux/module.h>
-#include <linux/err.h>
 #include "ehca_classes.h"
 #include "ehca_tools.h"
-#include "ehca_qes.h"
 #include "ehca_iverbs.h"
 #include "hcp_if.h"
 
@@ -93,6 +90,9 @@ u64 ehca_define_sqp(struct ehca_shca *shca,
 		return H_PARAMETER;
 	}
 
+	if (ehca_nr_ports < 0) /* autodetect mode */
+		return H_SUCCESS;
+
 	for (counter = 0;
 	     shca->sport[port - 1].port_state != IB_PORT_ACTIVE &&
 		     counter < ehca_port_act_time;
diff --git a/drivers/infiniband/hw/ehca/hcp_if.c b/drivers/infiniband/hw/ehca/hcp_if.c
index c16a21374bb5..7029aa653751 100644
--- a/drivers/infiniband/hw/ehca/hcp_if.c
+++ b/drivers/infiniband/hw/ehca/hcp_if.c
@@ -120,26 +120,21 @@ static long ehca_plpar_hcall_norets(unsigned long opcode,
 				    unsigned long arg7)
 {
 	long ret;
-	int i, sleep_msecs, do_lock;
-	unsigned long flags;
+	int i, sleep_msecs;
+	unsigned long flags = 0;
 
 	ehca_gen_dbg("opcode=%lx " HCALL7_REGS_FORMAT,
 		     opcode, arg1, arg2, arg3, arg4, arg5, arg6, arg7);
 
-	/* lock H_FREE_RESOURCE(MR) against itself and H_ALLOC_RESOURCE(MR) */
-	if ((opcode == H_FREE_RESOURCE) && (arg7 == 5)) {
-		arg7 = 0; /* better not upset firmware */
-		do_lock = 1;
-	}
-
 	for (i = 0; i < 5; i++) {
-		if (do_lock)
+		/* serialize hCalls to work around firmware issue */
+		if (ehca_lock_hcalls)
 			spin_lock_irqsave(&hcall_lock, flags);
 
 		ret = plpar_hcall_norets(opcode, arg1, arg2, arg3, arg4,
 					 arg5, arg6, arg7);
 
-		if (do_lock)
+		if (ehca_lock_hcalls)
 			spin_unlock_irqrestore(&hcall_lock, flags);
 
 		if (H_IS_LONG_BUSY(ret)) {
@@ -174,24 +169,22 @@ static long ehca_plpar_hcall9(unsigned long opcode,
 			      unsigned long arg9)
 {
 	long ret;
-	int i, sleep_msecs, do_lock;
+	int i, sleep_msecs;
 	unsigned long flags = 0;
 
 	ehca_gen_dbg("INPUT -- opcode=%lx " HCALL9_REGS_FORMAT, opcode,
 		     arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8, arg9);
 
-	/* lock H_ALLOC_RESOURCE(MR) against itself and H_FREE_RESOURCE(MR) */
-	do_lock = ((opcode == H_ALLOC_RESOURCE) && (arg2 == 5));
-
 	for (i = 0; i < 5; i++) {
-		if (do_lock)
+		/* serialize hCalls to work around firmware issue */
+		if (ehca_lock_hcalls)
 			spin_lock_irqsave(&hcall_lock, flags);
 
 		ret = plpar_hcall9(opcode, outs,
 				   arg1, arg2, arg3, arg4, arg5,
 				   arg6, arg7, arg8, arg9);
 
-		if (do_lock)
+		if (ehca_lock_hcalls)
 			spin_unlock_irqrestore(&hcall_lock, flags);
 
 		if (H_IS_LONG_BUSY(ret)) {
@@ -821,7 +814,7 @@ u64 hipz_h_free_resource_mr(const struct ipz_adapter_handle adapter_handle,
 	return ehca_plpar_hcall_norets(H_FREE_RESOURCE,
 				       adapter_handle.handle,    /* r4 */
 				       mr->ipz_mr_handle.handle, /* r5 */
-				       0, 0, 0, 0, 5);
+				       0, 0, 0, 0, 0);
 }
 
 u64 hipz_h_reregister_pmr(const struct ipz_adapter_handle adapter_handle,
diff --git a/drivers/infiniband/hw/ehca/hipz_hw.h b/drivers/infiniband/hw/ehca/hipz_hw.h
index d9739e554515..bf996c7acc42 100644
--- a/drivers/infiniband/hw/ehca/hipz_hw.h
+++ b/drivers/infiniband/hw/ehca/hipz_hw.h
@@ -378,6 +378,7 @@ struct hipz_query_hca {
 #define HCA_CAP_UD_LL_QP              EHCA_BMASK_IBM(16, 16)
 #define HCA_CAP_RESIZE_MR             EHCA_BMASK_IBM(17, 17)
 #define HCA_CAP_MINI_QP               EHCA_BMASK_IBM(18, 18)
+#define HCA_CAP_H_ALLOC_RES_SYNC      EHCA_BMASK_IBM(19, 19)
 
 /* query port response block */
 struct hipz_query_port {
@@ -402,7 +403,11 @@ struct hipz_query_port {
 	u64 max_msg_sz;
 	u32 max_mtu;
 	u32 vl_cap;
-	u8  reserved2[1900];
+	u32 phys_pstate;
+	u32 phys_state;
+	u32 phys_speed;
+	u32 phys_width;
+	u8  reserved2[1884];
 	u64 guid_entries[255];
 } __attribute__ ((packed));
 
diff --git a/drivers/infiniband/hw/ipath/ipath_common.h b/drivers/infiniband/hw/ipath/ipath_common.h
index 851df8a75e79..414621095540 100644
--- a/drivers/infiniband/hw/ipath/ipath_common.h
+++ b/drivers/infiniband/hw/ipath/ipath_common.h
@@ -82,6 +82,16 @@
 #define IPATH_IB_LINK_EXTERNAL	7 /* normal, disable local loopback */
 
 /*
+ * These 3 values (SDR and DDR may be ORed for auto-speed
+ * negotiation) are used for the 3rd argument to path_f_set_ib_cfg
+ * with cmd IPATH_IB_CFG_SPD_ENB, by direct calls or via sysfs.  They
+ * are also the the possible values for ipath_link_speed_enabled and active
+ * The values were chosen to match values used within the IB spec.
+ */
+#define IPATH_IB_SDR 1
+#define IPATH_IB_DDR 2
+
+/*
  * stats maintained by the driver.  For now, at least, this is global
  * to all minor devices.
  */
@@ -433,8 +443,9 @@ struct ipath_user_info {
 #define IPATH_CMD_UNUSED_2	26
 #define IPATH_CMD_PIOAVAILUPD	27	/* force an update of PIOAvail reg */
 #define IPATH_CMD_POLL_TYPE	28	/* set the kind of polling we want */
+#define IPATH_CMD_ARMLAUNCH_CTRL	29 /* armlaunch detection control */
 
-#define IPATH_CMD_MAX		28
+#define IPATH_CMD_MAX		29
 
 /*
  * Poll types
@@ -477,6 +488,8 @@ struct ipath_cmd {
 		__u64 port_info;
 		/* enable/disable receipt of packets */
 		__u32 recv_ctrl;
+		/* enable/disable armlaunch errors (non-zero to enable) */
+		__u32 armlaunch_ctrl;
 		/* partition key to set */
 		__u16 part_key;
 		/* user address of __u32 bitmask of active slaves */
@@ -579,7 +592,7 @@ struct ipath_flash {
 struct infinipath_counters {
 	__u64 LBIntCnt;
 	__u64 LBFlowStallCnt;
-	__u64 Reserved1;
+	__u64 TxSDmaDescCnt;	/* was Reserved1 */
 	__u64 TxUnsupVLErrCnt;
 	__u64 TxDataPktCnt;
 	__u64 TxFlowPktCnt;
@@ -615,12 +628,26 @@ struct infinipath_counters {
 	__u64 RxP6HdrEgrOvflCnt;
 	__u64 RxP7HdrEgrOvflCnt;
 	__u64 RxP8HdrEgrOvflCnt;
-	__u64 Reserved6;
-	__u64 Reserved7;
+	__u64 RxP9HdrEgrOvflCnt;	/* was Reserved6 */
+	__u64 RxP10HdrEgrOvflCnt;	/* was Reserved7 */
+	__u64 RxP11HdrEgrOvflCnt;	/* new for IBA7220 */
+	__u64 RxP12HdrEgrOvflCnt;	/* new for IBA7220 */
+	__u64 RxP13HdrEgrOvflCnt;	/* new for IBA7220 */
+	__u64 RxP14HdrEgrOvflCnt;	/* new for IBA7220 */
+	__u64 RxP15HdrEgrOvflCnt;	/* new for IBA7220 */
+	__u64 RxP16HdrEgrOvflCnt;	/* new for IBA7220 */
 	__u64 IBStatusChangeCnt;
 	__u64 IBLinkErrRecoveryCnt;
 	__u64 IBLinkDownedCnt;
 	__u64 IBSymbolErrCnt;
+	/* The following are new for IBA7220 */
+	__u64 RxVL15DroppedPktCnt;
+	__u64 RxOtherLocalPhyErrCnt;
+	__u64 PcieRetryBufDiagQwordCnt;
+	__u64 ExcessBufferOvflCnt;
+	__u64 LocalLinkIntegrityErrCnt;
+	__u64 RxVlErrCnt;
+	__u64 RxDlidFltrCnt;
 };
 
 /*
diff --git a/drivers/infiniband/hw/ipath/ipath_cq.c b/drivers/infiniband/hw/ipath/ipath_cq.c
index 645ed71fd797..a03bd28d9b48 100644
--- a/drivers/infiniband/hw/ipath/ipath_cq.c
+++ b/drivers/infiniband/hw/ipath/ipath_cq.c
@@ -395,16 +395,13 @@ int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 		goto bail;
 	}
 
-	/*
-	 * Return the address of the WC as the offset to mmap.
-	 * See ipath_mmap() for details.
-	 */
+	/* Check that we can write the offset to mmap. */
 	if (udata && udata->outlen >= sizeof(__u64)) {
-		__u64 offset = (__u64) wc;
+		__u64 offset = 0;
 
 		ret = ib_copy_to_udata(udata, &offset, sizeof(offset));
 		if (ret)
-			goto bail;
+			goto bail_free;
 	}
 
 	spin_lock_irq(&cq->lock);
@@ -424,10 +421,8 @@ int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 	else
 		n = head - tail;
 	if (unlikely((u32)cqe < n)) {
-		spin_unlock_irq(&cq->lock);
-		vfree(wc);
-		ret = -EOVERFLOW;
-		goto bail;
+		ret = -EINVAL;
+		goto bail_unlock;
 	}
 	for (n = 0; tail != head; n++) {
 		if (cq->ip)
@@ -452,6 +447,18 @@ int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 		struct ipath_mmap_info *ip = cq->ip;
 
 		ipath_update_mmap_info(dev, ip, sz, wc);
+
+		/*
+		 * Return the offset to mmap.
+		 * See ipath_mmap() for details.
+		 */
+		if (udata && udata->outlen >= sizeof(__u64)) {
+			ret = ib_copy_to_udata(udata, &ip->offset,
+					       sizeof(ip->offset));
+			if (ret)
+				goto bail;
+		}
+
 		spin_lock_irq(&dev->pending_lock);
 		if (list_empty(&ip->pending_mmaps))
 			list_add(&ip->pending_mmaps, &dev->pending_mmaps);
@@ -459,7 +466,12 @@ int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 	}
 
 	ret = 0;
+	goto bail;
 
+bail_unlock:
+	spin_unlock_irq(&cq->lock);
+bail_free:
+	vfree(wc);
 bail:
 	return ret;
 }
diff --git a/drivers/infiniband/hw/ipath/ipath_debug.h b/drivers/infiniband/hw/ipath/ipath_debug.h
index 19c56e6491eb..d6f69532d83f 100644
--- a/drivers/infiniband/hw/ipath/ipath_debug.h
+++ b/drivers/infiniband/hw/ipath/ipath_debug.h
@@ -55,7 +55,7 @@
 #define __IPATH_PKTDBG      0x80	/* print packet data */
 /* print process startup (init)/exit messages */
 #define __IPATH_PROCDBG     0x100
-/* print mmap/nopage stuff, not using VDBG any more */
+/* print mmap/fault stuff, not using VDBG any more */
 #define __IPATH_MMDBG       0x200
 #define __IPATH_ERRPKTDBG   0x400
 #define __IPATH_USER_SEND   0x1000	/* use user mode send */
@@ -81,7 +81,7 @@
 #define __IPATH_VERBDBG   0x0	/* very verbose debug */
 #define __IPATH_PKTDBG    0x0	/* print packet data */
 #define __IPATH_PROCDBG   0x0	/* process startup (init)/exit messages */
-/* print mmap/nopage stuff, not using VDBG any more */
+/* print mmap/fault stuff, not using VDBG any more */
 #define __IPATH_MMDBG     0x0
 #define __IPATH_EPKTDBG   0x0	/* print ethernet packet data */
 #define __IPATH_IPATHDBG  0x0	/* Ethernet (IPATH) table dump on */
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
index 1f152ded1e3c..d5ff6ca2db30 100644
--- a/drivers/infiniband/hw/ipath/ipath_driver.c
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c
@@ -121,6 +121,9 @@ static struct pci_driver ipath_driver = {
 	.probe = ipath_init_one,
 	.remove = __devexit_p(ipath_remove_one),
 	.id_table = ipath_pci_tbl,
+	.driver = {
+		.groups = ipath_driver_attr_groups,
+	},
 };
 
 static void ipath_check_status(struct work_struct *work)
@@ -331,6 +334,8 @@ static void ipath_verify_pioperf(struct ipath_devdata *dd)
 		udelay(1);
 	}
 
+	ipath_disable_armlaunch(dd);
+
 	writeq(0, piobuf); /* length 0, no dwords actually sent */
 	ipath_flush_wc();
 
@@ -362,6 +367,7 @@ static void ipath_verify_pioperf(struct ipath_devdata *dd)
 done:
 	/* disarm piobuf, so it's available again */
 	ipath_disarm_piobufs(dd, pbnum, 1);
+	ipath_enable_armlaunch(dd);
 }
 
 static int __devinit ipath_init_one(struct pci_dev *pdev,
@@ -800,31 +806,37 @@ void ipath_disarm_piobufs(struct ipath_devdata *dd, unsigned first,
 			  unsigned cnt)
 {
 	unsigned i, last = first + cnt;
-	u64 sendctrl, sendorig;
+	unsigned long flags;
 
 	ipath_cdbg(PKT, "disarm %u PIObufs first=%u\n", cnt, first);
-	sendorig = dd->ipath_sendctrl;
 	for (i = first; i < last; i++) {
-		sendctrl = sendorig  | INFINIPATH_S_DISARM |
-			(i << INFINIPATH_S_DISARMPIOBUF_SHIFT);
+		spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
+		/*
+		 * The disarm-related bits are write-only, so it
+		 * is ok to OR them in with our copy of sendctrl
+		 * while we hold the lock.
+		 */
 		ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
-				 sendctrl);
+			dd->ipath_sendctrl | INFINIPATH_S_DISARM |
+			(i << INFINIPATH_S_DISARMPIOBUF_SHIFT));
+		/* can't disarm bufs back-to-back per iba7220 spec */
+		ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+		spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
 	}
 
 	/*
-	 * Write it again with current value, in case ipath_sendctrl changed
-	 * while we were looping; no critical bits that would require
-	 * locking.
-	 *
-	 * disable PIOAVAILUPD, then re-enable, reading scratch in
+	 * Disable PIOAVAILUPD, then re-enable, reading scratch in
 	 * between.  This seems to avoid a chip timing race that causes
-	 * pioavail updates to memory to stop.
+	 * pioavail updates to memory to stop.  We xor as we don't
+	 * know the state of the bit when we're called.
 	 */
+	spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
-			 sendorig & ~INFINIPATH_S_PIOBUFAVAILUPD);
-	sendorig = ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+		dd->ipath_sendctrl ^ INFINIPATH_S_PIOBUFAVAILUPD);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
 			 dd->ipath_sendctrl);
+	spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
 }
 
 /**
@@ -1000,12 +1012,10 @@ static void get_rhf_errstring(u32 err, char *msg, size_t len)
  * ipath_get_egrbuf - get an eager buffer
  * @dd: the infinipath device
  * @bufnum: the eager buffer to get
- * @err: unused
  *
  * must only be called if ipath_pd[port] is known to be allocated
  */
-static inline void *ipath_get_egrbuf(struct ipath_devdata *dd, u32 bufnum,
-				     int err)
+static inline void *ipath_get_egrbuf(struct ipath_devdata *dd, u32 bufnum)
 {
 	return dd->ipath_port0_skbinfo ?
 		(void *) dd->ipath_port0_skbinfo[bufnum].skb->data : NULL;
@@ -1097,13 +1107,14 @@ static void ipath_rcv_hdrerr(struct ipath_devdata *dd,
 
 /*
  * ipath_kreceive - receive a packet
- * @dd: the infinipath device
+ * @pd: the infinipath port
  *
  * called from interrupt handler for errors or receive interrupt
  */
-void ipath_kreceive(struct ipath_devdata *dd)
+void ipath_kreceive(struct ipath_portdata *pd)
 {
 	u64 *rc;
+	struct ipath_devdata *dd = pd->port_dd;
 	void *ebuf;
 	const u32 rsize = dd->ipath_rcvhdrentsize;	/* words */
 	const u32 maxcnt = dd->ipath_rcvhdrcnt * rsize;	/* words */
@@ -1118,8 +1129,8 @@ void ipath_kreceive(struct ipath_devdata *dd)
 		goto bail;
 	}
 
-	l = dd->ipath_port0head;
-	hdrqtail = (u32) le64_to_cpu(*dd->ipath_hdrqtailptr);
+	l = pd->port_head;
+	hdrqtail = ipath_get_rcvhdrtail(pd);
 	if (l == hdrqtail)
 		goto bail;
 
@@ -1128,7 +1139,7 @@ reloop:
 		u32 qp;
 		u8 *bthbytes;
 
-		rc = (u64 *) (dd->ipath_pd[0]->port_rcvhdrq + (l << 2));
+		rc = (u64 *) (pd->port_rcvhdrq + (l << 2));
 		hdr = (struct ipath_message_header *)&rc[1];
 		/*
 		 * could make a network order version of IPATH_KD_QP, and
@@ -1153,7 +1164,7 @@ reloop:
 			etail = ipath_hdrget_index((__le32 *) rc);
 			if (tlen > sizeof(*hdr) ||
 			    etype == RCVHQ_RCV_TYPE_NON_KD)
-				ebuf = ipath_get_egrbuf(dd, etail, 0);
+				ebuf = ipath_get_egrbuf(dd, etail);
 		}
 
 		/*
@@ -1188,7 +1199,7 @@ reloop:
 				  be32_to_cpu(hdr->bth[0]) & 0xff);
 		else {
 			/*
-			 * error packet, type of error	unknown.
+			 * error packet, type of error unknown.
 			 * Probably type 3, but we don't know, so don't
 			 * even try to print the opcode, etc.
 			 */
@@ -1238,7 +1249,7 @@ reloop:
 		 * earlier packets, we "almost" guarantee we have covered
 		 * that case.
 		 */
-		u32 hqtail = (u32)le64_to_cpu(*dd->ipath_hdrqtailptr);
+		u32 hqtail = ipath_get_rcvhdrtail(pd);
 		if (hqtail != hdrqtail) {
 			hdrqtail = hqtail;
 			reloop = 1; /* loop 1 extra time at most */
@@ -1248,7 +1259,7 @@ reloop:
 
 	pkttot += i;
 
-	dd->ipath_port0head = l;
+	pd->port_head = l;
 
 	if (pkttot > ipath_stats.sps_maxpkts_call)
 		ipath_stats.sps_maxpkts_call = pkttot;
@@ -1332,14 +1343,9 @@ static void ipath_update_pio_bufs(struct ipath_devdata *dd)
 		/*
 		 * Chip Errata: bug 6641; even and odd qwords>3 are swapped
 		 */
-		if (i > 3) {
-			if (i & 1)
-				piov = le64_to_cpu(
-					dd->ipath_pioavailregs_dma[i - 1]);
-			else
-				piov = le64_to_cpu(
-					dd->ipath_pioavailregs_dma[i + 1]);
-		} else
+		if (i > 3 && (dd->ipath_flags & IPATH_SWAP_PIOBUFS))
+			piov = le64_to_cpu(dd->ipath_pioavailregs_dma[i ^ 1]);
+		else
 			piov = le64_to_cpu(dd->ipath_pioavailregs_dma[i]);
 		pchg = _IPATH_ALL_CHECKBITS &
 			~(dd->ipath_pioavailshadow[i] ^ piov);
@@ -1598,7 +1604,8 @@ int ipath_create_rcvhdrq(struct ipath_devdata *dd,
 
 	/* clear for security and sanity on each use */
 	memset(pd->port_rcvhdrq, 0, pd->port_rcvhdrq_size);
-	memset(pd->port_rcvhdrtail_kvaddr, 0, PAGE_SIZE);
+	if (pd->port_rcvhdrtail_kvaddr)
+		memset(pd->port_rcvhdrtail_kvaddr, 0, PAGE_SIZE);
 
 	/*
 	 * tell chip each time we init it, even if we are re-using previous
@@ -1614,77 +1621,6 @@ bail:
 	return ret;
 }
 
-int ipath_waitfor_complete(struct ipath_devdata *dd, ipath_kreg reg_id,
-			   u64 bits_to_wait_for, u64 * valp)
-{
-	unsigned long timeout;
-	u64 lastval, val;
-	int ret;
-
-	lastval = ipath_read_kreg64(dd, reg_id);
-	/* wait a ridiculously long time */
-	timeout = jiffies + msecs_to_jiffies(5);
-	do {
-		val = ipath_read_kreg64(dd, reg_id);
-		/* set so they have something, even on failures. */
-		*valp = val;
-		if ((val & bits_to_wait_for) == bits_to_wait_for) {
-			ret = 0;
-			break;
-		}
-		if (val != lastval)
-			ipath_cdbg(VERBOSE, "Changed from %llx to %llx, "
-				   "waiting for %llx bits\n",
-				   (unsigned long long) lastval,
-				   (unsigned long long) val,
-				   (unsigned long long) bits_to_wait_for);
-		cond_resched();
-		if (time_after(jiffies, timeout)) {
-			ipath_dbg("Didn't get bits %llx in register 0x%x, "
-				  "got %llx\n",
-				  (unsigned long long) bits_to_wait_for,
-				  reg_id, (unsigned long long) *valp);
-			ret = -ENODEV;
-			break;
-		}
-	} while (1);
-
-	return ret;
-}
-
-/**
- * ipath_waitfor_mdio_cmdready - wait for last command to complete
- * @dd: the infinipath device
- *
- * Like ipath_waitfor_complete(), but we wait for the CMDVALID bit to go
- * away indicating the last command has completed.  It doesn't return data
- */
-int ipath_waitfor_mdio_cmdready(struct ipath_devdata *dd)
-{
-	unsigned long timeout;
-	u64 val;
-	int ret;
-
-	/* wait a ridiculously long time */
-	timeout = jiffies + msecs_to_jiffies(5);
-	do {
-		val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_mdio);
-		if (!(val & IPATH_MDIO_CMDVALID)) {
-			ret = 0;
-			break;
-		}
-		cond_resched();
-		if (time_after(jiffies, timeout)) {
-			ipath_dbg("CMDVALID stuck in mdio reg? (%llx)\n",
-				  (unsigned long long) val);
-			ret = -ENODEV;
-			break;
-		}
-	} while (1);
-
-	return ret;
-}
-
 
 /*
  * Flush all sends that might be in the ready to send state, as well as any
@@ -2053,6 +1989,8 @@ void ipath_set_led_override(struct ipath_devdata *dd, unsigned int val)
  */
 void ipath_shutdown_device(struct ipath_devdata *dd)
 {
+	unsigned long flags;
+
 	ipath_dbg("Shutting down the device\n");
 
 	dd->ipath_flags |= IPATH_LINKUNK;
@@ -2073,9 +2011,13 @@ void ipath_shutdown_device(struct ipath_devdata *dd)
 	 * gracefully stop all sends allowing any in progress to trickle out
 	 * first.
 	 */
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, 0ULL);
+	spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
+	dd->ipath_sendctrl = 0;
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, dd->ipath_sendctrl);
 	/* flush it */
 	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
+
 	/*
 	 * enough for anything that's going to trickle out to have actually
 	 * done so.
@@ -2217,25 +2159,15 @@ static int __init infinipath_init(void)
 		goto bail_unit;
 	}
 
-	ret = ipath_driver_create_group(&ipath_driver.driver);
-	if (ret < 0) {
-		printk(KERN_ERR IPATH_DRV_NAME ": Unable to create driver "
-		       "sysfs entries: error %d\n", -ret);
-		goto bail_pci;
-	}
-
 	ret = ipath_init_ipathfs();
 	if (ret < 0) {
 		printk(KERN_ERR IPATH_DRV_NAME ": Unable to create "
 		       "ipathfs: error %d\n", -ret);
-		goto bail_group;
+		goto bail_pci;
 	}
 
 	goto bail;
 
-bail_group:
-	ipath_driver_remove_group(&ipath_driver.driver);
-
 bail_pci:
 	pci_unregister_driver(&ipath_driver);
 
@@ -2250,8 +2182,6 @@ static void __exit infinipath_cleanup(void)
 {
 	ipath_exit_ipathfs();
 
-	ipath_driver_remove_group(&ipath_driver.driver);
-
 	ipath_cdbg(VERBOSE, "Unregistering pci driver\n");
 	pci_unregister_driver(&ipath_driver);
 
@@ -2344,5 +2274,34 @@ int ipath_set_rx_pol_inv(struct ipath_devdata *dd, u8 new_pol_inv)
 	}
 	return 0;
 }
+
+/*
+ * Disable and enable the armlaunch error.  Used for PIO bandwidth testing on
+ * the 7220, which is count-based, rather than trigger-based.  Safe for the
+ * driver check, since it's at init.   Not completely safe when used for
+ * user-mode checking, since some error checking can be lost, but not
+ * particularly risky, and only has problematic side-effects in the face of
+ * very buggy user code.  There is no reference counting, but that's also
+ * fine, given the intended use.
+ */
+void ipath_enable_armlaunch(struct ipath_devdata *dd)
+{
+	dd->ipath_lasterror &= ~INFINIPATH_E_SPIOARMLAUNCH;
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_errorclear,
+		INFINIPATH_E_SPIOARMLAUNCH);
+	dd->ipath_errormask |= INFINIPATH_E_SPIOARMLAUNCH;
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_errormask,
+		dd->ipath_errormask);
+}
+
+void ipath_disable_armlaunch(struct ipath_devdata *dd)
+{
+	/* so don't re-enable if already set */
+	dd->ipath_maskederrs &= ~INFINIPATH_E_SPIOARMLAUNCH;
+	dd->ipath_errormask &= ~INFINIPATH_E_SPIOARMLAUNCH;
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_errormask,
+		dd->ipath_errormask);
+}
+
 module_init(infinipath_init);
 module_exit(infinipath_cleanup);
diff --git a/drivers/infiniband/hw/ipath/ipath_eeprom.c b/drivers/infiniband/hw/ipath/ipath_eeprom.c
index bcfa3ccb555f..e28a42f53769 100644
--- a/drivers/infiniband/hw/ipath/ipath_eeprom.c
+++ b/drivers/infiniband/hw/ipath/ipath_eeprom.c
@@ -510,10 +510,10 @@ int ipath_eeprom_read(struct ipath_devdata *dd, u8 eeprom_offset,
 {
 	int ret;
 
-	ret = down_interruptible(&dd->ipath_eep_sem);
+	ret = mutex_lock_interruptible(&dd->ipath_eep_lock);
 	if (!ret) {
 		ret = ipath_eeprom_internal_read(dd, eeprom_offset, buff, len);
-		up(&dd->ipath_eep_sem);
+		mutex_unlock(&dd->ipath_eep_lock);
 	}
 
 	return ret;
@@ -524,10 +524,10 @@ int ipath_eeprom_write(struct ipath_devdata *dd, u8 eeprom_offset,
 {
 	int ret;
 
-	ret = down_interruptible(&dd->ipath_eep_sem);
+	ret = mutex_lock_interruptible(&dd->ipath_eep_lock);
 	if (!ret) {
 		ret = ipath_eeprom_internal_write(dd, eeprom_offset, buff, len);
-		up(&dd->ipath_eep_sem);
+		mutex_unlock(&dd->ipath_eep_lock);
 	}
 
 	return ret;
@@ -538,7 +538,15 @@ static u8 flash_csum(struct ipath_flash *ifp, int adjust)
 	u8 *ip = (u8 *) ifp;
 	u8 csum = 0, len;
 
-	for (len = 0; len < ifp->if_length; len++)
+	/*
+	 * Limit length checksummed to max length of actual data.
+	 * Checksum of erased eeprom will still be bad, but we avoid
+	 * reading past the end of the buffer we were passed.
+	 */
+	len = ifp->if_length;
+	if (len > sizeof(struct ipath_flash))
+		len = sizeof(struct ipath_flash);
+	while (len--)
 		csum += *ip++;
 	csum -= ifp->if_csum;
 	csum = ~csum;
@@ -566,7 +574,7 @@ void ipath_get_eeprom_info(struct ipath_devdata *dd)
 	struct ipath_devdata *dd0 = ipath_lookup(0);
 
 	if (t && dd0->ipath_nguid > 1 && t <= dd0->ipath_nguid) {
-		u8 *bguid, oguid;
+		u8 oguid;
 		dd->ipath_guid = dd0->ipath_guid;
 		bguid = (u8 *) & dd->ipath_guid;
 
@@ -608,9 +616,9 @@ void ipath_get_eeprom_info(struct ipath_devdata *dd)
 		goto bail;
 	}
 
-	down(&dd->ipath_eep_sem);
+	mutex_lock(&dd->ipath_eep_lock);
 	eep_stat = ipath_eeprom_internal_read(dd, 0, buf, len);
-	up(&dd->ipath_eep_sem);
+	mutex_unlock(&dd->ipath_eep_lock);
 
 	if (eep_stat) {
 		ipath_dev_err(dd, "Failed reading GUID from eeprom\n");
@@ -666,7 +674,6 @@ void ipath_get_eeprom_info(struct ipath_devdata *dd)
 		 * elsewhere for backward-compatibility.
 		 */
 		char *snp = dd->ipath_serial;
-		int len;
 		memcpy(snp, ifp->if_sprefix, sizeof ifp->if_sprefix);
 		snp[sizeof ifp->if_sprefix] = '\0';
 		len = strlen(snp);
@@ -756,14 +763,14 @@ int ipath_update_eeprom_log(struct ipath_devdata *dd)
 	/* Grab semaphore and read current EEPROM. If we get an
 	 * error, let go, but if not, keep it until we finish write.
 	 */
-	ret = down_interruptible(&dd->ipath_eep_sem);
+	ret = mutex_lock_interruptible(&dd->ipath_eep_lock);
 	if (ret) {
 		ipath_dev_err(dd, "Unable to acquire EEPROM for logging\n");
 		goto free_bail;
 	}
 	ret = ipath_eeprom_internal_read(dd, 0, buf, len);
 	if (ret) {
-		up(&dd->ipath_eep_sem);
+		mutex_unlock(&dd->ipath_eep_lock);
 		ipath_dev_err(dd, "Unable read EEPROM for logging\n");
 		goto free_bail;
 	}
@@ -771,7 +778,7 @@ int ipath_update_eeprom_log(struct ipath_devdata *dd)
 
 	csum = flash_csum(ifp, 0);
 	if (csum != ifp->if_csum) {
-		up(&dd->ipath_eep_sem);
+		mutex_unlock(&dd->ipath_eep_lock);
 		ipath_dev_err(dd, "EEPROM cks err (0x%02X, S/B 0x%02X)\n",
 				csum, ifp->if_csum);
 		ret = 1;
@@ -841,7 +848,7 @@ int ipath_update_eeprom_log(struct ipath_devdata *dd)
 		csum = flash_csum(ifp, 1);
 		ret = ipath_eeprom_internal_write(dd, 0, buf, hi_water + 1);
 	}
-	up(&dd->ipath_eep_sem);
+	mutex_unlock(&dd->ipath_eep_lock);
 	if (ret)
 		ipath_dev_err(dd, "Failed updating EEPROM\n");
 
diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
index 5de3243a47c3..7e025c8e01b6 100644
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c
@@ -169,7 +169,7 @@ static int ipath_get_base_info(struct file *fp,
 		kinfo->spi_piocnt = dd->ipath_pbufsport;
 		kinfo->spi_piobufbase = (u64) pd->port_piobufs;
 		kinfo->__spi_uregbase = (u64) dd->ipath_uregbase +
-			dd->ipath_palign * pd->port_port;
+			dd->ipath_ureg_align * pd->port_port;
 	} else if (master) {
 		kinfo->spi_piocnt = (dd->ipath_pbufsport / subport_cnt) +
 				    (dd->ipath_pbufsport % subport_cnt);
@@ -186,7 +186,7 @@ static int ipath_get_base_info(struct file *fp,
 	}
 	if (shared) {
 		kinfo->spi_port_uregbase = (u64) dd->ipath_uregbase +
-			dd->ipath_palign * pd->port_port;
+			dd->ipath_ureg_align * pd->port_port;
 		kinfo->spi_port_rcvegrbuf = kinfo->spi_rcv_egrbufs;
 		kinfo->spi_port_rcvhdr_base = kinfo->spi_rcvhdr_base;
 		kinfo->spi_port_rcvhdr_tailaddr = kinfo->spi_rcvhdr_tailaddr;
@@ -742,11 +742,12 @@ static int ipath_manage_rcvq(struct ipath_portdata *pd, unsigned subport,
 		 * updated and correct itself, even in the face of software
 		 * bugs.
 		 */
-		*(volatile u64 *)pd->port_rcvhdrtail_kvaddr = 0;
-		set_bit(INFINIPATH_R_PORTENABLE_SHIFT + pd->port_port,
+		if (pd->port_rcvhdrtail_kvaddr)
+			ipath_clear_rcvhdrtail(pd);
+		set_bit(dd->ipath_r_portenable_shift + pd->port_port,
 			&dd->ipath_rcvctrl);
 	} else
-		clear_bit(INFINIPATH_R_PORTENABLE_SHIFT + pd->port_port,
+		clear_bit(dd->ipath_r_portenable_shift + pd->port_port,
 			  &dd->ipath_rcvctrl);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
 			 dd->ipath_rcvctrl);
@@ -881,7 +882,7 @@ static int ipath_create_user_egr(struct ipath_portdata *pd)
 
 	egrcnt = dd->ipath_rcvegrcnt;
 	/* TID number offset for this port */
-	egroff = pd->port_port * egrcnt;
+	egroff = (pd->port_port - 1) * egrcnt + dd->ipath_p0_rcvegrcnt;
 	egrsize = dd->ipath_rcvegrbufsize;
 	ipath_cdbg(VERBOSE, "Allocating %d egr buffers, at egrtid "
 		   "offset %x, egrsize %u\n", egrcnt, egroff, egrsize);
@@ -1049,11 +1050,6 @@ static int mmap_piobufs(struct vm_area_struct *vma,
 
 	phys = dd->ipath_physaddr + piobufs;
 
-	/*
-	 * Don't mark this as non-cached, or we don't get the
-	 * write combining behavior we want on the PIO buffers!
-	 */
-
 #if defined(__powerpc__)
 	/* There isn't a generic way to specify writethrough mappings */
 	pgprot_val(vma->vm_page_prot) |= _PAGE_NO_CACHE;
@@ -1120,33 +1116,24 @@ bail:
 }
 
 /*
- * ipath_file_vma_nopage - handle a VMA page fault.
+ * ipath_file_vma_fault - handle a VMA page fault.
  */
-static struct page *ipath_file_vma_nopage(struct vm_area_struct *vma,
-					  unsigned long address, int *type)
+static int ipath_file_vma_fault(struct vm_area_struct *vma,
+					struct vm_fault *vmf)
 {
-	unsigned long offset = address - vma->vm_start;
-	struct page *page = NOPAGE_SIGBUS;
-	void *pageptr;
+	struct page *page;
 
-	/*
-	 * Convert the vmalloc address into a struct page.
-	 */
-	pageptr = (void *)(offset + (vma->vm_pgoff << PAGE_SHIFT));
-	page = vmalloc_to_page(pageptr);
+	page = vmalloc_to_page((void *)(vmf->pgoff << PAGE_SHIFT));
 	if (!page)
-		goto out;
-
-	/* Increment the reference count. */
+		return VM_FAULT_SIGBUS;
 	get_page(page);
-	if (type)
-		*type = VM_FAULT_MINOR;
-out:
-	return page;
+	vmf->page = page;
+
+	return 0;
 }
 
 static struct vm_operations_struct ipath_file_vm_ops = {
-	.nopage = ipath_file_vma_nopage,
+	.fault = ipath_file_vma_fault,
 };
 
 static int mmap_kvaddr(struct vm_area_struct *vma, u64 pgaddr,
@@ -1284,7 +1271,7 @@ static int ipath_mmap(struct file *fp, struct vm_area_struct *vma)
 		goto bail;
 	}
 
-	ureg = dd->ipath_uregbase + dd->ipath_palign * pd->port_port;
+	ureg = dd->ipath_uregbase + dd->ipath_ureg_align * pd->port_port;
 	if (!pd->port_subport_cnt) {
 		/* port is not shared */
 		piocnt = dd->ipath_pbufsport;
@@ -1400,7 +1387,10 @@ static unsigned int ipath_poll_next(struct ipath_portdata *pd,
 	pollflag = ipath_poll_hdrqfull(pd);
 
 	head = ipath_read_ureg32(dd, ur_rcvhdrhead, pd->port_port);
-	tail = *(volatile u64 *)pd->port_rcvhdrtail_kvaddr;
+	if (pd->port_rcvhdrtail_kvaddr)
+		tail = ipath_get_rcvhdrtail(pd);
+	else
+		tail = ipath_read_ureg32(dd, ur_rcvhdrtail, pd->port_port);
 
 	if (head != tail)
 		pollflag |= POLLIN | POLLRDNORM;
@@ -1410,7 +1400,7 @@ static unsigned int ipath_poll_next(struct ipath_portdata *pd,
 		/* flush waiting flag so we don't miss an event */
 		wmb();
 
-		set_bit(pd->port_port + INFINIPATH_R_INTRAVAIL_SHIFT,
+		set_bit(pd->port_port + dd->ipath_r_intravail_shift,
 			&dd->ipath_rcvctrl);
 
 		ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
@@ -1790,6 +1780,7 @@ static int find_shared_port(struct file *fp,
 			}
 			port_fp(fp) = pd;
 			subport_fp(fp) = pd->port_cnt++;
+			pd->port_subpid[subport_fp(fp)] = current->pid;
 			tidcursor_fp(fp) = 0;
 			pd->active_slaves |= 1 << subport_fp(fp);
 			ipath_cdbg(PROC,
@@ -1920,8 +1911,7 @@ static int ipath_do_user_init(struct file *fp,
 	 */
 	head32 = ipath_read_ureg32(dd, ur_rcvegrindextail, pd->port_port);
 	ipath_write_ureg(dd, ur_rcvegrindexhead, head32, pd->port_port);
-	dd->ipath_lastegrheads[pd->port_port] = -1;
-	dd->ipath_lastrcvhdrqtails[pd->port_port] = -1;
+	pd->port_lastrcvhdrqtail = -1;
 	ipath_cdbg(VERBOSE, "Wrote port%d egrhead %x from tail regs\n",
 		pd->port_port, head32);
 	pd->port_tidcursor = 0;	/* start at beginning after open */
@@ -1941,11 +1931,13 @@ static int ipath_do_user_init(struct file *fp,
 	 * We explictly set the in-memory copy to 0 beforehand, so we don't
 	 * have to wait to be sure the DMA update has happened.
 	 */
-	*(volatile u64 *)pd->port_rcvhdrtail_kvaddr = 0ULL;
-	set_bit(INFINIPATH_R_PORTENABLE_SHIFT + pd->port_port,
+	if (pd->port_rcvhdrtail_kvaddr)
+		ipath_clear_rcvhdrtail(pd);
+	set_bit(dd->ipath_r_portenable_shift + pd->port_port,
 		&dd->ipath_rcvctrl);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
-			 dd->ipath_rcvctrl & ~INFINIPATH_R_TAILUPD);
+			dd->ipath_rcvctrl &
+			~(1ULL << dd->ipath_r_tailupd_shift));
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
 			 dd->ipath_rcvctrl);
 	/* Notify any waiting slaves */
@@ -2022,6 +2014,7 @@ static int ipath_close(struct inode *in, struct file *fp)
 		 * the slave(s) don't wait for receive data forever.
 		 */
 		pd->active_slaves &= ~(1 << fd->subport);
+		pd->port_subpid[fd->subport] = 0;
 		mutex_unlock(&ipath_mutex);
 		goto bail;
 	}
@@ -2054,9 +2047,9 @@ static int ipath_close(struct inode *in, struct file *fp)
 	if (dd->ipath_kregbase) {
 		int i;
 		/* atomically clear receive enable port and intr avail. */
-		clear_bit(INFINIPATH_R_PORTENABLE_SHIFT + port,
+		clear_bit(dd->ipath_r_portenable_shift + port,
 			  &dd->ipath_rcvctrl);
-		clear_bit(pd->port_port + INFINIPATH_R_INTRAVAIL_SHIFT,
+		clear_bit(pd->port_port + dd->ipath_r_intravail_shift,
 			  &dd->ipath_rcvctrl);
 		ipath_write_kreg( dd, dd->ipath_kregs->kr_rcvctrl,
 			dd->ipath_rcvctrl);
@@ -2149,11 +2142,15 @@ static int ipath_get_slave_info(struct ipath_portdata *pd,
 
 static int ipath_force_pio_avail_update(struct ipath_devdata *dd)
 {
-	u64 reg = dd->ipath_sendctrl;
+	unsigned long flags;
 
-	clear_bit(IPATH_S_PIOBUFAVAILUPD, &reg);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, reg);
+	spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
+		dd->ipath_sendctrl & ~INFINIPATH_S_PIOBUFAVAILUPD);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, dd->ipath_sendctrl);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
 
 	return 0;
 }
@@ -2227,6 +2224,11 @@ static ssize_t ipath_write(struct file *fp, const char __user *data,
 		dest = &cmd.cmd.poll_type;
 		src = &ucmd->cmd.poll_type;
 		break;
+	case IPATH_CMD_ARMLAUNCH_CTRL:
+		copy = sizeof(cmd.cmd.armlaunch_ctrl);
+		dest = &cmd.cmd.armlaunch_ctrl;
+		src = &ucmd->cmd.armlaunch_ctrl;
+		break;
 	default:
 		ret = -EINVAL;
 		goto bail;
@@ -2302,6 +2304,12 @@ static ssize_t ipath_write(struct file *fp, const char __user *data,
 	case IPATH_CMD_POLL_TYPE:
 		pd->poll_type = cmd.cmd.poll_type;
 		break;
+	case IPATH_CMD_ARMLAUNCH_CTRL:
+		if (cmd.cmd.armlaunch_ctrl)
+			ipath_enable_armlaunch(pd->port_dd);
+		else
+			ipath_disable_armlaunch(pd->port_dd);
+		break;
 	}
 
 	if (ret >= 0)
diff --git a/drivers/infiniband/hw/ipath/ipath_fs.c b/drivers/infiniband/hw/ipath/ipath_fs.c
index 262c25db05cd..23faba9d21eb 100644
--- a/drivers/infiniband/hw/ipath/ipath_fs.c
+++ b/drivers/infiniband/hw/ipath/ipath_fs.c
@@ -108,21 +108,16 @@ static const struct file_operations atomic_stats_ops = {
 	.read = atomic_stats_read,
 };
 
-#define NUM_COUNTERS sizeof(struct infinipath_counters) / sizeof(u64)
-
 static ssize_t atomic_counters_read(struct file *file, char __user *buf,
 				    size_t count, loff_t *ppos)
 {
-	u64 counters[NUM_COUNTERS];
-	u16 i;
+	struct infinipath_counters counters;
 	struct ipath_devdata *dd;
 
 	dd = file->f_path.dentry->d_inode->i_private;
+	dd->ipath_f_read_counters(dd, &counters);
 
-	for (i = 0; i < NUM_COUNTERS; i++)
-		counters[i] = ipath_snap_cntr(dd, i);
-
-	return simple_read_from_buffer(buf, count, ppos, counters,
+	return simple_read_from_buffer(buf, count, ppos, &counters,
 				       sizeof counters);
 }
 
@@ -243,8 +238,7 @@ static int create_device_files(struct super_block *sb,
 
 	snprintf(unit, sizeof unit, "%02d", dd->ipath_unit);
 	ret = create_file(unit, S_IFDIR|S_IRUGO|S_IXUGO, sb->s_root, &dir,
-			  (struct file_operations *) &simple_dir_operations,
-			  dd);
+			  &simple_dir_operations, dd);
 	if (ret) {
 		printk(KERN_ERR "create_file(%s) failed: %d\n", unit, ret);
 		goto bail;
diff --git a/drivers/infiniband/hw/ipath/ipath_iba6110.c b/drivers/infiniband/hw/ipath/ipath_iba6110.c
index ddbebe4bdb27..9e2ced3cdc5e 100644
--- a/drivers/infiniband/hw/ipath/ipath_iba6110.c
+++ b/drivers/infiniband/hw/ipath/ipath_iba6110.c
@@ -148,10 +148,57 @@ struct _infinipath_do_not_use_kernel_regs {
 	unsigned long long ReservedSW2[4];
 };
 
-#define IPATH_KREG_OFFSET(field) (offsetof(struct \
-    _infinipath_do_not_use_kernel_regs, field) / sizeof(u64))
+struct _infinipath_do_not_use_counters {
+	__u64 LBIntCnt;
+	__u64 LBFlowStallCnt;
+	__u64 Reserved1;
+	__u64 TxUnsupVLErrCnt;
+	__u64 TxDataPktCnt;
+	__u64 TxFlowPktCnt;
+	__u64 TxDwordCnt;
+	__u64 TxLenErrCnt;
+	__u64 TxMaxMinLenErrCnt;
+	__u64 TxUnderrunCnt;
+	__u64 TxFlowStallCnt;
+	__u64 TxDroppedPktCnt;
+	__u64 RxDroppedPktCnt;
+	__u64 RxDataPktCnt;
+	__u64 RxFlowPktCnt;
+	__u64 RxDwordCnt;
+	__u64 RxLenErrCnt;
+	__u64 RxMaxMinLenErrCnt;
+	__u64 RxICRCErrCnt;
+	__u64 RxVCRCErrCnt;
+	__u64 RxFlowCtrlErrCnt;
+	__u64 RxBadFormatCnt;
+	__u64 RxLinkProblemCnt;
+	__u64 RxEBPCnt;
+	__u64 RxLPCRCErrCnt;
+	__u64 RxBufOvflCnt;
+	__u64 RxTIDFullErrCnt;
+	__u64 RxTIDValidErrCnt;
+	__u64 RxPKeyMismatchCnt;
+	__u64 RxP0HdrEgrOvflCnt;
+	__u64 RxP1HdrEgrOvflCnt;
+	__u64 RxP2HdrEgrOvflCnt;
+	__u64 RxP3HdrEgrOvflCnt;
+	__u64 RxP4HdrEgrOvflCnt;
+	__u64 RxP5HdrEgrOvflCnt;
+	__u64 RxP6HdrEgrOvflCnt;
+	__u64 RxP7HdrEgrOvflCnt;
+	__u64 RxP8HdrEgrOvflCnt;
+	__u64 Reserved6;
+	__u64 Reserved7;
+	__u64 IBStatusChangeCnt;
+	__u64 IBLinkErrRecoveryCnt;
+	__u64 IBLinkDownedCnt;
+	__u64 IBSymbolErrCnt;
+};
+
+#define IPATH_KREG_OFFSET(field) (offsetof( \
+	struct _infinipath_do_not_use_kernel_regs, field) / sizeof(u64))
 #define IPATH_CREG_OFFSET(field) (offsetof( \
-    struct infinipath_counters, field) / sizeof(u64))
+	struct _infinipath_do_not_use_counters, field) / sizeof(u64))
 
 static const struct ipath_kregs ipath_ht_kregs = {
 	.kr_control = IPATH_KREG_OFFSET(Control),
@@ -282,6 +329,9 @@ static const struct ipath_cregs ipath_ht_cregs = {
 #define INFINIPATH_HWE_HTAPLL_RFSLIP        0x1000000000000000ULL
 #define INFINIPATH_HWE_SERDESPLLFAILED      0x2000000000000000ULL
 
+#define IBA6110_IBCS_LINKTRAININGSTATE_MASK 0xf
+#define IBA6110_IBCS_LINKSTATE_SHIFT 4
+
 /* kr_extstatus bits */
 #define INFINIPATH_EXTS_FREQSEL 0x2
 #define INFINIPATH_EXTS_SERDESSEL 0x4
@@ -296,6 +346,12 @@ static const struct ipath_cregs ipath_ht_cregs = {
 #define INFINIPATH_RT_BUFSIZE_MASK 0x3FFFULL
 #define INFINIPATH_RT_BUFSIZE_SHIFT 48
 
+#define INFINIPATH_R_INTRAVAIL_SHIFT 16
+#define INFINIPATH_R_TAILUPD_SHIFT 31
+
+/* kr_xgxsconfig bits */
+#define INFINIPATH_XGXS_RESET          0x7ULL
+
 /*
  * masks and bits that are different in different chips, or present only
  * in one
@@ -652,7 +708,6 @@ static int ipath_ht_boardname(struct ipath_devdata *dd, char *name,
 			      "with ID %u\n", boardrev);
 		snprintf(name, namelen, "Unknown_InfiniPath_QHT7xxx_%u",
 			 boardrev);
-		ret = 1;
 		break;
 	}
 	if (n)
@@ -686,6 +741,13 @@ static int ipath_ht_boardname(struct ipath_devdata *dd, char *name,
 			      dd->ipath_htspeed);
 	ret = 0;
 
+	/*
+	 * set here, not in ipath_init_*_funcs because we have to do
+	 * it after we can read chip registers.
+	 */
+	dd->ipath_ureg_align =
+		ipath_read_kreg32(dd, dd->ipath_kregs->kr_pagealign);
+
 bail:
 	return ret;
 }
@@ -969,7 +1031,8 @@ static int ipath_setup_ht_config(struct ipath_devdata *dd,
 	do {
 		u8 cap_type;
 
-		/* the HT capability type byte is 3 bytes after the
+		/*
+		 * The HT capability type byte is 3 bytes after the
 		 * capability byte.
 		 */
 		if (pci_read_config_byte(pdev, pos + 3, &cap_type)) {
@@ -982,6 +1045,8 @@ static int ipath_setup_ht_config(struct ipath_devdata *dd,
 	} while ((pos = pci_find_next_capability(pdev, pos,
 						 PCI_CAP_ID_HT)));
 
+	dd->ipath_flags |= IPATH_SWAP_PIOBUFS;
+
 bail:
 	return ret;
 }
@@ -1074,11 +1139,55 @@ static void ipath_setup_ht_setextled(struct ipath_devdata *dd,
 
 static void ipath_init_ht_variables(struct ipath_devdata *dd)
 {
+	/*
+	 * setup the register offsets, since they are different for each
+	 * chip
+	 */
+	dd->ipath_kregs = &ipath_ht_kregs;
+	dd->ipath_cregs = &ipath_ht_cregs;
+
 	dd->ipath_gpio_sda_num = _IPATH_GPIO_SDA_NUM;
 	dd->ipath_gpio_scl_num = _IPATH_GPIO_SCL_NUM;
 	dd->ipath_gpio_sda = IPATH_GPIO_SDA;
 	dd->ipath_gpio_scl = IPATH_GPIO_SCL;
 
+	/*
+	 * Fill in data for field-values that change in newer chips.
+	 * We dynamically specify only the mask for LINKTRAININGSTATE
+	 * and only the shift for LINKSTATE, as they are the only ones
+	 * that change.  Also precalculate the 3 link states of interest
+	 * and the combined mask.
+	 */
+	dd->ibcs_ls_shift = IBA6110_IBCS_LINKSTATE_SHIFT;
+	dd->ibcs_lts_mask = IBA6110_IBCS_LINKTRAININGSTATE_MASK;
+	dd->ibcs_mask = (INFINIPATH_IBCS_LINKSTATE_MASK <<
+		dd->ibcs_ls_shift) | dd->ibcs_lts_mask;
+	dd->ib_init = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
+		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
+		(INFINIPATH_IBCS_L_STATE_INIT << dd->ibcs_ls_shift);
+	dd->ib_arm = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
+		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
+		(INFINIPATH_IBCS_L_STATE_ARM << dd->ibcs_ls_shift);
+	dd->ib_active = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
+		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
+		(INFINIPATH_IBCS_L_STATE_ACTIVE << dd->ibcs_ls_shift);
+
+	/*
+	 * Fill in data for ibcc field-values that change in newer chips.
+	 * We dynamically specify only the mask for LINKINITCMD
+	 * and only the shift for LINKCMD and MAXPKTLEN, as they are
+	 * the only ones that change.
+	 */
+	dd->ibcc_lic_mask = INFINIPATH_IBCC_LINKINITCMD_MASK;
+	dd->ibcc_lc_shift = INFINIPATH_IBCC_LINKCMD_SHIFT;
+	dd->ibcc_mpl_shift = INFINIPATH_IBCC_MAXPKTLEN_SHIFT;
+
+	/* Fill in shifts for RcvCtrl. */
+	dd->ipath_r_portenable_shift = INFINIPATH_R_PORTENABLE_SHIFT;
+	dd->ipath_r_intravail_shift = INFINIPATH_R_INTRAVAIL_SHIFT;
+	dd->ipath_r_tailupd_shift = INFINIPATH_R_TAILUPD_SHIFT;
+	dd->ipath_r_portcfg_shift = 0; /* Not on IBA6110 */
+
 	dd->ipath_i_bitsextant =
 		(INFINIPATH_I_RCVURG_MASK << INFINIPATH_I_RCVURG_SHIFT) |
 		(INFINIPATH_I_RCVAVAIL_MASK <<
@@ -1135,6 +1244,8 @@ static void ipath_init_ht_variables(struct ipath_devdata *dd)
 
 	dd->ipath_i_rcvavail_mask = INFINIPATH_I_RCVAVAIL_MASK;
 	dd->ipath_i_rcvurg_mask = INFINIPATH_I_RCVURG_MASK;
+	dd->ipath_i_rcvavail_shift = INFINIPATH_I_RCVAVAIL_SHIFT;
+	dd->ipath_i_rcvurg_shift = INFINIPATH_I_RCVURG_SHIFT;
 
 	/*
 	 * EEPROM error log 0 is TXE Parity errors. 1 is RXE Parity.
@@ -1148,9 +1259,17 @@ static void ipath_init_ht_variables(struct ipath_devdata *dd)
 		INFINIPATH_HWE_RXEMEMPARITYERR_MASK <<
 		INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT;
 
-	dd->ipath_eep_st_masks[2].errs_to_log =
-		INFINIPATH_E_INVALIDADDR | INFINIPATH_E_RESET;
+	dd->ipath_eep_st_masks[2].errs_to_log = INFINIPATH_E_RESET;
 
+	dd->delay_mult = 2; /* SDR, 4X, can't change */
+
+	dd->ipath_link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
+	dd->ipath_link_speed_supported = IPATH_IB_SDR;
+	dd->ipath_link_width_enabled = IB_WIDTH_4X;
+	dd->ipath_link_speed_enabled = dd->ipath_link_speed_supported;
+	/* these can't change for this chip, so set once */
+	dd->ipath_link_width_active = dd->ipath_link_width_enabled;
+	dd->ipath_link_speed_active = dd->ipath_link_speed_enabled;
 }
 
 /**
@@ -1205,14 +1324,16 @@ static void ipath_ht_init_hwerrors(struct ipath_devdata *dd)
 	val &= ~INFINIPATH_HWE_HTCMISCERR4;
 
 	/*
-	 * PLL ignored because MDIO interface has a logic problem
-	 * for reads, on Comstock and Ponderosa.  BRINGUP
+	 * PLL ignored because unused MDIO interface has a logic problem
 	 */
 	if (dd->ipath_boardrev == 4 || dd->ipath_boardrev == 9)
 		val &= ~INFINIPATH_HWE_SERDESPLLFAILED;
 	dd->ipath_hwerrmask = val;
 }
 
+
+
+
 /**
  * ipath_ht_bringup_serdes - bring up the serdes
  * @dd: the infinipath device
@@ -1284,16 +1405,6 @@ static int ipath_ht_bringup_serdes(struct ipath_devdata *dd)
 	}
 
 	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
-	if (((val >> INFINIPATH_XGXS_MDIOADDR_SHIFT) &
-	     INFINIPATH_XGXS_MDIOADDR_MASK) != 3) {
-		val &= ~(INFINIPATH_XGXS_MDIOADDR_MASK <<
-			 INFINIPATH_XGXS_MDIOADDR_SHIFT);
-		/*
-		 * we use address 3
-		 */
-		val |= 3ULL << INFINIPATH_XGXS_MDIOADDR_SHIFT;
-		change = 1;
-	}
 	if (val & INFINIPATH_XGXS_RESET) {
 		/* normally true after boot */
 		val &= ~INFINIPATH_XGXS_RESET;
@@ -1329,21 +1440,6 @@ static int ipath_ht_bringup_serdes(struct ipath_devdata *dd)
 		   (unsigned long long)
 		   ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig));
 
-	if (!ipath_waitfor_mdio_cmdready(dd)) {
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_mdio,
-				 ipath_mdio_req(IPATH_MDIO_CMD_READ, 31,
-						IPATH_MDIO_CTRL_XGXS_REG_8,
-						0));
-		if (ipath_waitfor_complete(dd, dd->ipath_kregs->kr_mdio,
-					   IPATH_MDIO_DATAVALID, &val))
-			ipath_dbg("Never got MDIO data for XGXS status "
-				  "read\n");
-		else
-			ipath_cdbg(VERBOSE, "MDIO Read reg8, "
-				   "'bank' 31 %x\n", (u32) val);
-	} else
-		ipath_dbg("Never got MDIO cmdready for XGXS status read\n");
-
 	return ret;		/* for now, say we always succeeded */
 }
 
@@ -1396,6 +1492,7 @@ static void ipath_ht_put_tid(struct ipath_devdata *dd,
 			pa |= lenvalid | INFINIPATH_RT_VALID;
 		}
 	}
+
 	writeq(pa, tidptr);
 }
 
@@ -1526,8 +1623,7 @@ static int ipath_ht_early_init(struct ipath_devdata *dd)
 	}
 
 	ipath_get_eeprom_info(dd);
-	if (dd->ipath_boardrev == 5 && dd->ipath_serial[0] == '1' &&
-		dd->ipath_serial[1] == '2' && dd->ipath_serial[2] == '8') {
+	if (dd->ipath_boardrev == 5) {
 		/*
 		 * Later production QHT7040 has same changes as QHT7140, so
 		 * can use GPIO interrupts.  They have serial #'s starting
@@ -1602,6 +1698,210 @@ static void ipath_ht_free_irq(struct ipath_devdata *dd)
 	dd->ipath_intconfig = 0;
 }
 
+static struct ipath_message_header *
+ipath_ht_get_msgheader(struct ipath_devdata *dd, __le32 *rhf_addr)
+{
+	return (struct ipath_message_header *)
+		&rhf_addr[sizeof(u64) / sizeof(u32)];
+}
+
+static void ipath_ht_config_ports(struct ipath_devdata *dd, ushort cfgports)
+{
+	dd->ipath_portcnt =
+		ipath_read_kreg32(dd, dd->ipath_kregs->kr_portcnt);
+	dd->ipath_p0_rcvegrcnt =
+		ipath_read_kreg32(dd, dd->ipath_kregs->kr_rcvegrcnt);
+}
+
+static void ipath_ht_read_counters(struct ipath_devdata *dd,
+				   struct infinipath_counters *cntrs)
+{
+	cntrs->LBIntCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(LBIntCnt));
+	cntrs->LBFlowStallCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(LBFlowStallCnt));
+	cntrs->TxSDmaDescCnt = 0;
+	cntrs->TxUnsupVLErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxUnsupVLErrCnt));
+	cntrs->TxDataPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxDataPktCnt));
+	cntrs->TxFlowPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxFlowPktCnt));
+	cntrs->TxDwordCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxDwordCnt));
+	cntrs->TxLenErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxLenErrCnt));
+	cntrs->TxMaxMinLenErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxMaxMinLenErrCnt));
+	cntrs->TxUnderrunCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxUnderrunCnt));
+	cntrs->TxFlowStallCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxFlowStallCnt));
+	cntrs->TxDroppedPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxDroppedPktCnt));
+	cntrs->RxDroppedPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxDroppedPktCnt));
+	cntrs->RxDataPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxDataPktCnt));
+	cntrs->RxFlowPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxFlowPktCnt));
+	cntrs->RxDwordCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxDwordCnt));
+	cntrs->RxLenErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxLenErrCnt));
+	cntrs->RxMaxMinLenErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxMaxMinLenErrCnt));
+	cntrs->RxICRCErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxICRCErrCnt));
+	cntrs->RxVCRCErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxVCRCErrCnt));
+	cntrs->RxFlowCtrlErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxFlowCtrlErrCnt));
+	cntrs->RxBadFormatCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxBadFormatCnt));
+	cntrs->RxLinkProblemCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxLinkProblemCnt));
+	cntrs->RxEBPCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxEBPCnt));
+	cntrs->RxLPCRCErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxLPCRCErrCnt));
+	cntrs->RxBufOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxBufOvflCnt));
+	cntrs->RxTIDFullErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxTIDFullErrCnt));
+	cntrs->RxTIDValidErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxTIDValidErrCnt));
+	cntrs->RxPKeyMismatchCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxPKeyMismatchCnt));
+	cntrs->RxP0HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP0HdrEgrOvflCnt));
+	cntrs->RxP1HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP1HdrEgrOvflCnt));
+	cntrs->RxP2HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP2HdrEgrOvflCnt));
+	cntrs->RxP3HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP3HdrEgrOvflCnt));
+	cntrs->RxP4HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP4HdrEgrOvflCnt));
+	cntrs->RxP5HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP5HdrEgrOvflCnt));
+	cntrs->RxP6HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP6HdrEgrOvflCnt));
+	cntrs->RxP7HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP7HdrEgrOvflCnt));
+	cntrs->RxP8HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP8HdrEgrOvflCnt));
+	cntrs->RxP9HdrEgrOvflCnt = 0;
+	cntrs->RxP10HdrEgrOvflCnt = 0;
+	cntrs->RxP11HdrEgrOvflCnt = 0;
+	cntrs->RxP12HdrEgrOvflCnt = 0;
+	cntrs->RxP13HdrEgrOvflCnt = 0;
+	cntrs->RxP14HdrEgrOvflCnt = 0;
+	cntrs->RxP15HdrEgrOvflCnt = 0;
+	cntrs->RxP16HdrEgrOvflCnt = 0;
+	cntrs->IBStatusChangeCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBStatusChangeCnt));
+	cntrs->IBLinkErrRecoveryCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBLinkErrRecoveryCnt));
+	cntrs->IBLinkDownedCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBLinkDownedCnt));
+	cntrs->IBSymbolErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBSymbolErrCnt));
+	cntrs->RxVL15DroppedPktCnt = 0;
+	cntrs->RxOtherLocalPhyErrCnt = 0;
+	cntrs->PcieRetryBufDiagQwordCnt = 0;
+	cntrs->ExcessBufferOvflCnt = dd->ipath_overrun_thresh_errs;
+	cntrs->LocalLinkIntegrityErrCnt =
+		(dd->ipath_flags & IPATH_GPIO_ERRINTRS) ?
+		dd->ipath_lli_errs : dd->ipath_lli_errors;
+	cntrs->RxVlErrCnt = 0;
+	cntrs->RxDlidFltrCnt = 0;
+}
+
+
+/* no interrupt fallback for these chips */
+static int ipath_ht_nointr_fallback(struct ipath_devdata *dd)
+{
+	return 0;
+}
+
+
+/*
+ * reset the XGXS (between serdes and IBC).  Slightly less intrusive
+ * than resetting the IBC or external link state, and useful in some
+ * cases to cause some retraining.  To do this right, we reset IBC
+ * as well.
+ */
+static void ipath_ht_xgxs_reset(struct ipath_devdata *dd)
+{
+	u64 val, prev_val;
+
+	prev_val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
+	val = prev_val | INFINIPATH_XGXS_RESET;
+	prev_val &= ~INFINIPATH_XGXS_RESET; /* be sure */
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_control,
+			 dd->ipath_control & ~INFINIPATH_C_LINKENABLE);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
+	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, prev_val);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_control,
+			 dd->ipath_control);
+}
+
+
+static int ipath_ht_get_ib_cfg(struct ipath_devdata *dd, int which)
+{
+	int ret;
+
+	switch (which) {
+	case IPATH_IB_CFG_LWID:
+		ret = dd->ipath_link_width_active;
+		break;
+	case IPATH_IB_CFG_SPD:
+		ret = dd->ipath_link_speed_active;
+		break;
+	case IPATH_IB_CFG_LWID_ENB:
+		ret = dd->ipath_link_width_enabled;
+		break;
+	case IPATH_IB_CFG_SPD_ENB:
+		ret = dd->ipath_link_speed_enabled;
+		break;
+	default:
+		ret =  -ENOTSUPP;
+		break;
+	}
+	return ret;
+}
+
+
+/* we assume range checking is already done, if needed */
+static int ipath_ht_set_ib_cfg(struct ipath_devdata *dd, int which, u32 val)
+{
+	int ret = 0;
+
+	if (which == IPATH_IB_CFG_LWID_ENB)
+		dd->ipath_link_width_enabled = val;
+	else if (which == IPATH_IB_CFG_SPD_ENB)
+		dd->ipath_link_speed_enabled = val;
+	else
+		ret = -ENOTSUPP;
+	return ret;
+}
+
+
+static void ipath_ht_config_jint(struct ipath_devdata *dd, u16 a, u16 b)
+{
+}
+
+
+static int ipath_ht_ib_updown(struct ipath_devdata *dd, int ibup, u64 ibcs)
+{
+	ipath_setup_ht_setextled(dd, ipath_ib_linkstate(dd, ibcs),
+		ipath_ib_linktrstate(dd, ibcs));
+	return 0;
+}
+
+
 /**
  * ipath_init_iba6110_funcs - set up the chip-specific function pointers
  * @dd: the infinipath device
@@ -1626,22 +1926,19 @@ void ipath_init_iba6110_funcs(struct ipath_devdata *dd)
 	dd->ipath_f_setextled = ipath_setup_ht_setextled;
 	dd->ipath_f_get_base_info = ipath_ht_get_base_info;
 	dd->ipath_f_free_irq = ipath_ht_free_irq;
-
-	/*
-	 * initialize chip-specific variables
-	 */
 	dd->ipath_f_tidtemplate = ipath_ht_tidtemplate;
+	dd->ipath_f_intr_fallback = ipath_ht_nointr_fallback;
+	dd->ipath_f_get_msgheader = ipath_ht_get_msgheader;
+	dd->ipath_f_config_ports = ipath_ht_config_ports;
+	dd->ipath_f_read_counters = ipath_ht_read_counters;
+	dd->ipath_f_xgxs_reset = ipath_ht_xgxs_reset;
+	dd->ipath_f_get_ib_cfg = ipath_ht_get_ib_cfg;
+	dd->ipath_f_set_ib_cfg = ipath_ht_set_ib_cfg;
+	dd->ipath_f_config_jint = ipath_ht_config_jint;
+	dd->ipath_f_ib_updown = ipath_ht_ib_updown;
 
 	/*
-	 * setup the register offsets, since they are different for each
-	 * chip
-	 */
-	dd->ipath_kregs = &ipath_ht_kregs;
-	dd->ipath_cregs = &ipath_ht_cregs;
-
-	/*
-	 * do very early init that is needed before ipath_f_bus is
-	 * called
+	 * initialize chip-specific variables
 	 */
 	ipath_init_ht_variables(dd);
 }
diff --git a/drivers/infiniband/hw/ipath/ipath_iba6120.c b/drivers/infiniband/hw/ipath/ipath_iba6120.c
index 0103d6f4847b..c7a2f50824c0 100644
--- a/drivers/infiniband/hw/ipath/ipath_iba6120.c
+++ b/drivers/infiniband/hw/ipath/ipath_iba6120.c
@@ -145,10 +145,57 @@ struct _infinipath_do_not_use_kernel_regs {
 	unsigned long long Reserved12;
 };
 
-#define IPATH_KREG_OFFSET(field) (offsetof(struct \
-    _infinipath_do_not_use_kernel_regs, field) / sizeof(u64))
+struct _infinipath_do_not_use_counters {
+	__u64 LBIntCnt;
+	__u64 LBFlowStallCnt;
+	__u64 Reserved1;
+	__u64 TxUnsupVLErrCnt;
+	__u64 TxDataPktCnt;
+	__u64 TxFlowPktCnt;
+	__u64 TxDwordCnt;
+	__u64 TxLenErrCnt;
+	__u64 TxMaxMinLenErrCnt;
+	__u64 TxUnderrunCnt;
+	__u64 TxFlowStallCnt;
+	__u64 TxDroppedPktCnt;
+	__u64 RxDroppedPktCnt;
+	__u64 RxDataPktCnt;
+	__u64 RxFlowPktCnt;
+	__u64 RxDwordCnt;
+	__u64 RxLenErrCnt;
+	__u64 RxMaxMinLenErrCnt;
+	__u64 RxICRCErrCnt;
+	__u64 RxVCRCErrCnt;
+	__u64 RxFlowCtrlErrCnt;
+	__u64 RxBadFormatCnt;
+	__u64 RxLinkProblemCnt;
+	__u64 RxEBPCnt;
+	__u64 RxLPCRCErrCnt;
+	__u64 RxBufOvflCnt;
+	__u64 RxTIDFullErrCnt;
+	__u64 RxTIDValidErrCnt;
+	__u64 RxPKeyMismatchCnt;
+	__u64 RxP0HdrEgrOvflCnt;
+	__u64 RxP1HdrEgrOvflCnt;
+	__u64 RxP2HdrEgrOvflCnt;
+	__u64 RxP3HdrEgrOvflCnt;
+	__u64 RxP4HdrEgrOvflCnt;
+	__u64 RxP5HdrEgrOvflCnt;
+	__u64 RxP6HdrEgrOvflCnt;
+	__u64 RxP7HdrEgrOvflCnt;
+	__u64 RxP8HdrEgrOvflCnt;
+	__u64 Reserved6;
+	__u64 Reserved7;
+	__u64 IBStatusChangeCnt;
+	__u64 IBLinkErrRecoveryCnt;
+	__u64 IBLinkDownedCnt;
+	__u64 IBSymbolErrCnt;
+};
+
+#define IPATH_KREG_OFFSET(field) (offsetof( \
+	struct _infinipath_do_not_use_kernel_regs, field) / sizeof(u64))
 #define IPATH_CREG_OFFSET(field) (offsetof( \
-    struct infinipath_counters, field) / sizeof(u64))
+	struct _infinipath_do_not_use_counters, field) / sizeof(u64))
 
 static const struct ipath_kregs ipath_pe_kregs = {
 	.kr_control = IPATH_KREG_OFFSET(Control),
@@ -282,6 +329,9 @@ static const struct ipath_cregs ipath_pe_cregs = {
 #define INFINIPATH_HWE_PCIE0PLLFAILED       0x0800000000000000ULL
 #define INFINIPATH_HWE_SERDESPLLFAILED      0x1000000000000000ULL
 
+#define IBA6120_IBCS_LINKTRAININGSTATE_MASK 0xf
+#define IBA6120_IBCS_LINKSTATE_SHIFT 4
+
 /* kr_extstatus bits */
 #define INFINIPATH_EXTS_FREQSEL 0x2
 #define INFINIPATH_EXTS_SERDESSEL 0x4
@@ -296,6 +346,9 @@ static const struct ipath_cregs ipath_pe_cregs = {
 #define IPATH_GPIO_SCL (1ULL << \
 	(_IPATH_GPIO_SCL_NUM+INFINIPATH_EXTC_GPIOOE_SHIFT))
 
+#define INFINIPATH_R_INTRAVAIL_SHIFT 16
+#define INFINIPATH_R_TAILUPD_SHIFT 31
+
 /* 6120 specific hardware errors... */
 static const struct ipath_hwerror_msgs ipath_6120_hwerror_msgs[] = {
 	INFINIPATH_HWE_MSG(PCIEPOISONEDTLP, "PCIe Poisoned TLP"),
@@ -320,10 +373,28 @@ static const struct ipath_hwerror_msgs ipath_6120_hwerror_msgs[] = {
 		        INFINIPATH_HWE_TXEMEMPARITYERR_PIOPBC) \
 		        << INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT)
 
-static int ipath_pe_txe_recover(struct ipath_devdata *);
 static void ipath_pe_put_tid_2(struct ipath_devdata *, u64 __iomem *,
 			       u32, unsigned long);
 
+/*
+ * On platforms using this chip, and not having ordered WC stores, we
+ * can get TXE parity errors due to speculative reads to the PIO buffers,
+ * and this, due to a chip bug can result in (many) false parity error
+ * reports.  So it's a debug print on those, and an info print on systems
+ * where the speculative reads don't occur.
+ */
+static void ipath_pe_txe_recover(struct ipath_devdata *dd)
+{
+	if (ipath_unordered_wc())
+		ipath_dbg("Recovering from TXE PIO parity error\n");
+	else {
+		++ipath_stats.sps_txeparity;
+		dev_info(&dd->pcidev->dev,
+			"Recovering from TXE PIO parity error\n");
+	}
+}
+
+
 /**
  * ipath_pe_handle_hwerrors - display hardware errors.
  * @dd: the infinipath device
@@ -403,35 +474,11 @@ static void ipath_pe_handle_hwerrors(struct ipath_devdata *dd, char *msg,
 		 * occur if a processor speculative read is done to the PIO
 		 * buffer while we are sending a packet, for example.
 		 */
-		if ((hwerrs & TXE_PIO_PARITY) && ipath_pe_txe_recover(dd))
+		if (hwerrs & TXE_PIO_PARITY) {
+			ipath_pe_txe_recover(dd);
 			hwerrs &= ~TXE_PIO_PARITY;
-		if (hwerrs) {
-			/*
-			 * if any set that we aren't ignoring only make the
-			 * complaint once, in case it's stuck or recurring,
-			 * and we get here multiple times
-			 * Force link down, so switch knows, and
-			 * LEDs are turned off
-			 */
-			if (dd->ipath_flags & IPATH_INITTED) {
-				ipath_set_linkstate(dd, IPATH_IB_LINKDOWN);
-				ipath_setup_pe_setextled(dd,
-					INFINIPATH_IBCS_L_STATE_DOWN,
-					INFINIPATH_IBCS_LT_STATE_DISABLED);
-				ipath_dev_err(dd, "Fatal Hardware Error (freeze "
-					      "mode), no longer usable, SN %.16s\n",
-						  dd->ipath_serial);
-				isfatal = 1;
-			}
-			/*
-			 * Mark as having had an error for driver, and also
-			 * for /sys and status word mapped to user programs.
-			 * This marks unit as not usable, until reset
-			 */
-			*dd->ipath_statusp &= ~IPATH_STATUS_IB_READY;
-			*dd->ipath_statusp |= IPATH_STATUS_HWERROR;
-			dd->ipath_flags &= ~IPATH_INITTED;
-		} else {
+		}
+		if (!hwerrs) {
 			static u32 freeze_cnt;
 
 			freeze_cnt++;
@@ -485,7 +532,7 @@ static void ipath_pe_handle_hwerrors(struct ipath_devdata *dd, char *msg,
 
 	if (hwerrs & INFINIPATH_HWE_SERDESPLLFAILED) {
 		/*
-		 * If it occurs, it is left masked since the eternal
+		 * If it occurs, it is left masked since the external
 		 * interface is unused
 		 */
 		dd->ipath_hwerrmask &= ~INFINIPATH_HWE_SERDESPLLFAILED;
@@ -563,6 +610,14 @@ static int ipath_pe_boardname(struct ipath_devdata *dd, char *name,
 			dd->ipath_f_put_tid = ipath_pe_put_tid_2;
 	}
 
+
+	/*
+	 * set here, not in ipath_init_*_funcs because we have to do
+	 * it after we can read chip registers.
+	 */
+	dd->ipath_ureg_align =
+		ipath_read_kreg32(dd, dd->ipath_kregs->kr_pagealign);
+
 	return ret;
 }
 
@@ -667,17 +722,8 @@ static int ipath_pe_bringup_serdes(struct ipath_devdata *dd)
 
 	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
 	prev_val = val;
-	if (((val >> INFINIPATH_XGXS_MDIOADDR_SHIFT) &
-	     INFINIPATH_XGXS_MDIOADDR_MASK) != 3) {
-		val &=
-			~(INFINIPATH_XGXS_MDIOADDR_MASK <<
-			  INFINIPATH_XGXS_MDIOADDR_SHIFT);
-		/* MDIO address 3 */
-		val |= 3ULL << INFINIPATH_XGXS_MDIOADDR_SHIFT;
-	}
-	if (val & INFINIPATH_XGXS_RESET) {
+	if (val & INFINIPATH_XGXS_RESET)
 		val &= ~INFINIPATH_XGXS_RESET;
-	}
 	if (((val >> INFINIPATH_XGXS_RX_POL_SHIFT) &
 	     INFINIPATH_XGXS_RX_POL_MASK) != dd->ipath_rx_pol_inv ) {
 		/* need to compensate for Tx inversion in partner */
@@ -707,21 +753,6 @@ static int ipath_pe_bringup_serdes(struct ipath_devdata *dd)
 		   (unsigned long long)
 		   ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig));
 
-	if (!ipath_waitfor_mdio_cmdready(dd)) {
-		ipath_write_kreg(
-			dd, dd->ipath_kregs->kr_mdio,
-			ipath_mdio_req(IPATH_MDIO_CMD_READ, 31,
-				       IPATH_MDIO_CTRL_XGXS_REG_8, 0));
-		if (ipath_waitfor_complete(dd, dd->ipath_kregs->kr_mdio,
-					   IPATH_MDIO_DATAVALID, &val))
-			ipath_dbg("Never got MDIO data for XGXS "
-				  "status read\n");
-		else
-			ipath_cdbg(VERBOSE, "MDIO Read reg8, "
-				   "'bank' 31 %x\n", (u32) val);
-	} else
-		ipath_dbg("Never got MDIO cmdready for XGXS status read\n");
-
 	return ret;
 }
 
@@ -902,12 +933,27 @@ static int ipath_setup_pe_config(struct ipath_devdata *dd,
 	else
 		ipath_dev_err(dd, "Can't find PCI Express "
 			      "capability!\n");
+
+	dd->ipath_link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
+	dd->ipath_link_speed_supported = IPATH_IB_SDR;
+	dd->ipath_link_width_enabled = IB_WIDTH_4X;
+	dd->ipath_link_speed_enabled = dd->ipath_link_speed_supported;
+	/* these can't change for this chip, so set once */
+	dd->ipath_link_width_active = dd->ipath_link_width_enabled;
+	dd->ipath_link_speed_active = dd->ipath_link_speed_enabled;
 	return 0;
 }
 
 static void ipath_init_pe_variables(struct ipath_devdata *dd)
 {
 	/*
+	 * setup the register offsets, since they are different for each
+	 * chip
+	 */
+	dd->ipath_kregs = &ipath_pe_kregs;
+	dd->ipath_cregs = &ipath_pe_cregs;
+
+	/*
 	 * bits for selecting i2c direction and values,
 	 * used for I2C serial flash
 	 */
@@ -916,6 +962,43 @@ static void ipath_init_pe_variables(struct ipath_devdata *dd)
 	dd->ipath_gpio_sda = IPATH_GPIO_SDA;
 	dd->ipath_gpio_scl = IPATH_GPIO_SCL;
 
+	/*
+	 * Fill in data for field-values that change in newer chips.
+	 * We dynamically specify only the mask for LINKTRAININGSTATE
+	 * and only the shift for LINKSTATE, as they are the only ones
+	 * that change.  Also precalculate the 3 link states of interest
+	 * and the combined mask.
+	 */
+	dd->ibcs_ls_shift = IBA6120_IBCS_LINKSTATE_SHIFT;
+	dd->ibcs_lts_mask = IBA6120_IBCS_LINKTRAININGSTATE_MASK;
+	dd->ibcs_mask = (INFINIPATH_IBCS_LINKSTATE_MASK <<
+		dd->ibcs_ls_shift) | dd->ibcs_lts_mask;
+	dd->ib_init = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
+		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
+		(INFINIPATH_IBCS_L_STATE_INIT << dd->ibcs_ls_shift);
+	dd->ib_arm = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
+		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
+		(INFINIPATH_IBCS_L_STATE_ARM << dd->ibcs_ls_shift);
+	dd->ib_active = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
+		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
+		(INFINIPATH_IBCS_L_STATE_ACTIVE << dd->ibcs_ls_shift);
+
+	/*
+	 * Fill in data for ibcc field-values that change in newer chips.
+	 * We dynamically specify only the mask for LINKINITCMD
+	 * and only the shift for LINKCMD and MAXPKTLEN, as they are
+	 * the only ones that change.
+	 */
+	dd->ibcc_lic_mask = INFINIPATH_IBCC_LINKINITCMD_MASK;
+	dd->ibcc_lc_shift = INFINIPATH_IBCC_LINKCMD_SHIFT;
+	dd->ibcc_mpl_shift = INFINIPATH_IBCC_MAXPKTLEN_SHIFT;
+
+	/* Fill in shifts for RcvCtrl. */
+	dd->ipath_r_portenable_shift = INFINIPATH_R_PORTENABLE_SHIFT;
+	dd->ipath_r_intravail_shift = INFINIPATH_R_INTRAVAIL_SHIFT;
+	dd->ipath_r_tailupd_shift = INFINIPATH_R_TAILUPD_SHIFT;
+	dd->ipath_r_portcfg_shift = 0; /* Not on IBA6120 */
+
 	/* variables for sanity checking interrupt and errors */
 	dd->ipath_hwe_bitsextant =
 		(INFINIPATH_HWE_RXEMEMPARITYERR_MASK <<
@@ -963,6 +1046,8 @@ static void ipath_init_pe_variables(struct ipath_devdata *dd)
 
 	dd->ipath_i_rcvavail_mask = INFINIPATH_I_RCVAVAIL_MASK;
 	dd->ipath_i_rcvurg_mask = INFINIPATH_I_RCVURG_MASK;
+	dd->ipath_i_rcvavail_shift = INFINIPATH_I_RCVAVAIL_SHIFT;
+	dd->ipath_i_rcvurg_shift = INFINIPATH_I_RCVURG_SHIFT;
 
 	/*
 	 * EEPROM error log 0 is TXE Parity errors. 1 is RXE Parity.
@@ -984,6 +1069,7 @@ static void ipath_init_pe_variables(struct ipath_devdata *dd)
 		INFINIPATH_E_INVALIDADDR | INFINIPATH_E_RESET;
 
 
+	dd->delay_mult = 2; /* SDR, 4X, can't change */
 }
 
 /* setup the MSI stuff again after a reset.  I'd like to just call
@@ -1289,6 +1375,9 @@ static int ipath_pe_early_init(struct ipath_devdata *dd)
 	 */
 	dd->ipath_rcvhdrentsize = 24;
 	dd->ipath_rcvhdrsize = IPATH_DFLT_RCVHDRSIZE;
+	dd->ipath_rhf_offset = 0;
+	dd->ipath_egrtidbase = (u64 __iomem *)
+		((char __iomem *) dd->ipath_kregbase + dd->ipath_rcvegrbase);
 
 	/*
 	 * To truly support a 4KB MTU (for usermode), we need to
@@ -1359,34 +1448,204 @@ static void ipath_pe_free_irq(struct ipath_devdata *dd)
 	dd->ipath_irq = 0;
 }
 
+
+static struct ipath_message_header *
+ipath_pe_get_msgheader(struct ipath_devdata *dd, __le32 *rhf_addr)
+{
+	return (struct ipath_message_header *)
+		&rhf_addr[sizeof(u64) / sizeof(u32)];
+}
+
+static void ipath_pe_config_ports(struct ipath_devdata *dd, ushort cfgports)
+{
+	dd->ipath_portcnt =
+		ipath_read_kreg32(dd, dd->ipath_kregs->kr_portcnt);
+	dd->ipath_p0_rcvegrcnt =
+		ipath_read_kreg32(dd, dd->ipath_kregs->kr_rcvegrcnt);
+}
+
+static void ipath_pe_read_counters(struct ipath_devdata *dd,
+				   struct infinipath_counters *cntrs)
+{
+	cntrs->LBIntCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(LBIntCnt));
+	cntrs->LBFlowStallCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(LBFlowStallCnt));
+	cntrs->TxSDmaDescCnt = 0;
+	cntrs->TxUnsupVLErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxUnsupVLErrCnt));
+	cntrs->TxDataPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxDataPktCnt));
+	cntrs->TxFlowPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxFlowPktCnt));
+	cntrs->TxDwordCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxDwordCnt));
+	cntrs->TxLenErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxLenErrCnt));
+	cntrs->TxMaxMinLenErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxMaxMinLenErrCnt));
+	cntrs->TxUnderrunCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxUnderrunCnt));
+	cntrs->TxFlowStallCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxFlowStallCnt));
+	cntrs->TxDroppedPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxDroppedPktCnt));
+	cntrs->RxDroppedPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxDroppedPktCnt));
+	cntrs->RxDataPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxDataPktCnt));
+	cntrs->RxFlowPktCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxFlowPktCnt));
+	cntrs->RxDwordCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxDwordCnt));
+	cntrs->RxLenErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxLenErrCnt));
+	cntrs->RxMaxMinLenErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxMaxMinLenErrCnt));
+	cntrs->RxICRCErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxICRCErrCnt));
+	cntrs->RxVCRCErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxVCRCErrCnt));
+	cntrs->RxFlowCtrlErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxFlowCtrlErrCnt));
+	cntrs->RxBadFormatCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxBadFormatCnt));
+	cntrs->RxLinkProblemCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxLinkProblemCnt));
+	cntrs->RxEBPCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxEBPCnt));
+	cntrs->RxLPCRCErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxLPCRCErrCnt));
+	cntrs->RxBufOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxBufOvflCnt));
+	cntrs->RxTIDFullErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxTIDFullErrCnt));
+	cntrs->RxTIDValidErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxTIDValidErrCnt));
+	cntrs->RxPKeyMismatchCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxPKeyMismatchCnt));
+	cntrs->RxP0HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP0HdrEgrOvflCnt));
+	cntrs->RxP1HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP1HdrEgrOvflCnt));
+	cntrs->RxP2HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP2HdrEgrOvflCnt));
+	cntrs->RxP3HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP3HdrEgrOvflCnt));
+	cntrs->RxP4HdrEgrOvflCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP4HdrEgrOvflCnt));
+	cntrs->RxP5HdrEgrOvflCnt = 0;
+	cntrs->RxP6HdrEgrOvflCnt = 0;
+	cntrs->RxP7HdrEgrOvflCnt = 0;
+	cntrs->RxP8HdrEgrOvflCnt = 0;
+	cntrs->RxP9HdrEgrOvflCnt = 0;
+	cntrs->RxP10HdrEgrOvflCnt = 0;
+	cntrs->RxP11HdrEgrOvflCnt = 0;
+	cntrs->RxP12HdrEgrOvflCnt = 0;
+	cntrs->RxP13HdrEgrOvflCnt = 0;
+	cntrs->RxP14HdrEgrOvflCnt = 0;
+	cntrs->RxP15HdrEgrOvflCnt = 0;
+	cntrs->RxP16HdrEgrOvflCnt = 0;
+	cntrs->IBStatusChangeCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBStatusChangeCnt));
+	cntrs->IBLinkErrRecoveryCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBLinkErrRecoveryCnt));
+	cntrs->IBLinkDownedCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBLinkDownedCnt));
+	cntrs->IBSymbolErrCnt =
+		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBSymbolErrCnt));
+	cntrs->RxVL15DroppedPktCnt = 0;
+	cntrs->RxOtherLocalPhyErrCnt = 0;
+	cntrs->PcieRetryBufDiagQwordCnt = 0;
+	cntrs->ExcessBufferOvflCnt = dd->ipath_overrun_thresh_errs;
+	cntrs->LocalLinkIntegrityErrCnt = dd->ipath_lli_errs;
+	cntrs->RxVlErrCnt = 0;
+	cntrs->RxDlidFltrCnt = 0;
+}
+
+
+/* no interrupt fallback for these chips */
+static int ipath_pe_nointr_fallback(struct ipath_devdata *dd)
+{
+	return 0;
+}
+
+
 /*
- * On platforms using this chip, and not having ordered WC stores, we
- * can get TXE parity errors due to speculative reads to the PIO buffers,
- * and this, due to a chip bug can result in (many) false parity error
- * reports.  So it's a debug print on those, and an info print on systems
- * where the speculative reads don't occur.
- * Because we can get lots of false errors, we have no upper limit
- * on recovery attempts on those platforms.
+ * reset the XGXS (between serdes and IBC).  Slightly less intrusive
+ * than resetting the IBC or external link state, and useful in some
+ * cases to cause some retraining.  To do this right, we reset IBC
+ * as well.
  */
-static int ipath_pe_txe_recover(struct ipath_devdata *dd)
+static void ipath_pe_xgxs_reset(struct ipath_devdata *dd)
 {
-	if (ipath_unordered_wc())
-		ipath_dbg("Recovering from TXE PIO parity error\n");
-	else {
-		int cnt = ++ipath_stats.sps_txeparity;
-		if (cnt >= IPATH_MAX_PARITY_ATTEMPTS)  {
-			if (cnt == IPATH_MAX_PARITY_ATTEMPTS)
-				ipath_dev_err(dd,
-					"Too many attempts to recover from "
-					"TXE parity, giving up\n");
-			return 0;
-		}
-		dev_info(&dd->pcidev->dev,
-			"Recovering from TXE PIO parity error\n");
+	u64 val, prev_val;
+
+	prev_val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
+	val = prev_val | INFINIPATH_XGXS_RESET;
+	prev_val &= ~INFINIPATH_XGXS_RESET; /* be sure */
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_control,
+			 dd->ipath_control & ~INFINIPATH_C_LINKENABLE);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
+	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, prev_val);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_control,
+			 dd->ipath_control);
+}
+
+
+static int ipath_pe_get_ib_cfg(struct ipath_devdata *dd, int which)
+{
+	int ret;
+
+	switch (which) {
+	case IPATH_IB_CFG_LWID:
+		ret = dd->ipath_link_width_active;
+		break;
+	case IPATH_IB_CFG_SPD:
+		ret = dd->ipath_link_speed_active;
+		break;
+	case IPATH_IB_CFG_LWID_ENB:
+		ret = dd->ipath_link_width_enabled;
+		break;
+	case IPATH_IB_CFG_SPD_ENB:
+		ret = dd->ipath_link_speed_enabled;
+		break;
+	default:
+		ret =  -ENOTSUPP;
+		break;
 	}
-	return 1;
+	return ret;
+}
+
+
+/* we assume range checking is already done, if needed */
+static int ipath_pe_set_ib_cfg(struct ipath_devdata *dd, int which, u32 val)
+{
+	int ret = 0;
+
+	if (which == IPATH_IB_CFG_LWID_ENB)
+		dd->ipath_link_width_enabled = val;
+	else if (which == IPATH_IB_CFG_SPD_ENB)
+		dd->ipath_link_speed_enabled = val;
+	else
+		ret = -ENOTSUPP;
+	return ret;
 }
 
+static void ipath_pe_config_jint(struct ipath_devdata *dd, u16 a, u16 b)
+{
+}
+
+
+static int ipath_pe_ib_updown(struct ipath_devdata *dd, int ibup, u64 ibcs)
+{
+	ipath_setup_pe_setextled(dd, ipath_ib_linkstate(dd, ibcs),
+		ipath_ib_linktrstate(dd, ibcs));
+	return 0;
+}
+
+
 /**
  * ipath_init_iba6120_funcs - set up the chip-specific function pointers
  * @dd: the infinipath device
@@ -1407,7 +1666,7 @@ void ipath_init_iba6120_funcs(struct ipath_devdata *dd)
 	dd->ipath_f_bringup_serdes = ipath_pe_bringup_serdes;
 	dd->ipath_f_clear_tids = ipath_pe_clear_tids;
 	/*
-	 * this may get changed after we read the chip revision,
+	 * _f_put_tid may get changed after we read the chip revision,
 	 * but we start with the safe version for all revs
 	 */
 	dd->ipath_f_put_tid = ipath_pe_put_tid;
@@ -1415,17 +1674,19 @@ void ipath_init_iba6120_funcs(struct ipath_devdata *dd)
 	dd->ipath_f_setextled = ipath_setup_pe_setextled;
 	dd->ipath_f_get_base_info = ipath_pe_get_base_info;
 	dd->ipath_f_free_irq = ipath_pe_free_irq;
-
-	/* initialize chip-specific variables */
 	dd->ipath_f_tidtemplate = ipath_pe_tidtemplate;
+	dd->ipath_f_intr_fallback = ipath_pe_nointr_fallback;
+	dd->ipath_f_xgxs_reset = ipath_pe_xgxs_reset;
+	dd->ipath_f_get_msgheader = ipath_pe_get_msgheader;
+	dd->ipath_f_config_ports = ipath_pe_config_ports;
+	dd->ipath_f_read_counters = ipath_pe_read_counters;
+	dd->ipath_f_get_ib_cfg = ipath_pe_get_ib_cfg;
+	dd->ipath_f_set_ib_cfg = ipath_pe_set_ib_cfg;
+	dd->ipath_f_config_jint = ipath_pe_config_jint;
+	dd->ipath_f_ib_updown = ipath_pe_ib_updown;
 
-	/*
-	 * setup the register offsets, since they are different for each
-	 * chip
-	 */
-	dd->ipath_kregs = &ipath_pe_kregs;
-	dd->ipath_cregs = &ipath_pe_cregs;
 
+	/* initialize chip-specific variables */
 	ipath_init_pe_variables(dd);
 }
 
diff --git a/drivers/infiniband/hw/ipath/ipath_init_chip.c b/drivers/infiniband/hw/ipath/ipath_init_chip.c
index 9dd0bacf8461..4471674975cd 100644
--- a/drivers/infiniband/hw/ipath/ipath_init_chip.c
+++ b/drivers/infiniband/hw/ipath/ipath_init_chip.c
@@ -91,7 +91,7 @@ static int create_port0_egr(struct ipath_devdata *dd)
 	struct ipath_skbinfo *skbinfo;
 	int ret;
 
-	egrcnt = dd->ipath_rcvegrcnt;
+	egrcnt = dd->ipath_p0_rcvegrcnt;
 
 	skbinfo = vmalloc(sizeof(*dd->ipath_port0_skbinfo) * egrcnt);
 	if (skbinfo == NULL) {
@@ -244,8 +244,7 @@ static int init_chip_first(struct ipath_devdata *dd,
 	 * cfgports.  We do still check and report a difference, if
 	 * not same (should be impossible).
 	 */
-	dd->ipath_portcnt =
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_portcnt);
+	dd->ipath_f_config_ports(dd, ipath_cfgports);
 	if (!ipath_cfgports)
 		dd->ipath_cfgports = dd->ipath_portcnt;
 	else if (ipath_cfgports <= dd->ipath_portcnt) {
@@ -272,22 +271,7 @@ static int init_chip_first(struct ipath_devdata *dd,
 		goto done;
 	}
 
-	dd->ipath_lastegrheads = kzalloc(sizeof(*dd->ipath_lastegrheads)
-					 * dd->ipath_cfgports,
-					 GFP_KERNEL);
-	dd->ipath_lastrcvhdrqtails =
-		kzalloc(sizeof(*dd->ipath_lastrcvhdrqtails)
-			* dd->ipath_cfgports, GFP_KERNEL);
-
-	if (!dd->ipath_lastegrheads || !dd->ipath_lastrcvhdrqtails) {
-		ipath_dev_err(dd, "Unable to allocate head arrays, "
-			      "failing\n");
-		ret = -ENOMEM;
-		goto done;
-	}
-
 	pd = create_portdata0(dd);
-
 	if (!pd) {
 		ipath_dev_err(dd, "Unable to allocate portdata for port "
 			      "0, failing\n");
@@ -345,10 +329,10 @@ static int init_chip_first(struct ipath_devdata *dd,
 		       dd->ipath_piobcnt2k, dd->ipath_pio2kbase);
 
 	spin_lock_init(&dd->ipath_tid_lock);
-
+	spin_lock_init(&dd->ipath_sendctrl_lock);
 	spin_lock_init(&dd->ipath_gpio_lock);
 	spin_lock_init(&dd->ipath_eep_st_lock);
-	sema_init(&dd->ipath_eep_sem, 1);
+	mutex_init(&dd->ipath_eep_lock);
 
 done:
 	*pdp = pd;
@@ -372,9 +356,9 @@ static int init_chip_reset(struct ipath_devdata *dd,
 	*pdp = dd->ipath_pd[0];
 	/* ensure chip does no sends or receives while we re-initialize */
 	dd->ipath_control = dd->ipath_sendctrl = dd->ipath_rcvctrl = 0U;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl, 0);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, 0);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_control, 0);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl, dd->ipath_rcvctrl);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, dd->ipath_sendctrl);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_control, dd->ipath_control);
 
 	rtmp = ipath_read_kreg32(dd, dd->ipath_kregs->kr_portcnt);
 	if (dd->ipath_portcnt != rtmp)
@@ -487,6 +471,7 @@ static void enable_chip(struct ipath_devdata *dd,
 			struct ipath_portdata *pd, int reinit)
 {
 	u32 val;
+	unsigned long flags;
 	int i;
 
 	if (!reinit)
@@ -495,19 +480,21 @@ static void enable_chip(struct ipath_devdata *dd,
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
 			 dd->ipath_rcvctrl);
 
+	spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
 	/* Enable PIO send, and update of PIOavail regs to memory. */
 	dd->ipath_sendctrl = INFINIPATH_S_PIOENABLE |
 		INFINIPATH_S_PIOBUFAVAILUPD;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
-			 dd->ipath_sendctrl);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, dd->ipath_sendctrl);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
 
 	/*
 	 * enable port 0 receive, and receive interrupt.  other ports
 	 * done as user opens and inits them.
 	 */
-	dd->ipath_rcvctrl = INFINIPATH_R_TAILUPD |
-		(1ULL << INFINIPATH_R_PORTENABLE_SHIFT) |
-		(1ULL << INFINIPATH_R_INTRAVAIL_SHIFT);
+	dd->ipath_rcvctrl = (1ULL << dd->ipath_r_tailupd_shift) |
+		(1ULL << dd->ipath_r_portenable_shift) |
+		(1ULL << dd->ipath_r_intravail_shift);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
 			 dd->ipath_rcvctrl);
 
@@ -523,12 +510,11 @@ static void enable_chip(struct ipath_devdata *dd,
 	 */
 	val = ipath_read_ureg32(dd, ur_rcvegrindextail, 0);
 	(void)ipath_write_ureg(dd, ur_rcvegrindexhead, val, 0);
-	dd->ipath_port0head = ipath_read_ureg32(dd, ur_rcvhdrtail, 0);
 
 	/* Initialize so we interrupt on next packet received */
 	(void)ipath_write_ureg(dd, ur_rcvhdrhead,
 			       dd->ipath_rhdrhead_intr_off |
-			       dd->ipath_port0head, 0);
+			       dd->ipath_pd[0]->port_head, 0);
 
 	/*
 	 * by now pioavail updates to memory should have occurred, so
@@ -542,12 +528,8 @@ static void enable_chip(struct ipath_devdata *dd,
 		/*
 		 * Chip Errata bug 6641; even and odd qwords>3 are swapped.
 		 */
-		if (i > 3) {
-			if (i & 1)
-				val = dd->ipath_pioavailregs_dma[i - 1];
-			else
-				val = dd->ipath_pioavailregs_dma[i + 1];
-		}
+		if (i > 3 && (dd->ipath_flags & IPATH_SWAP_PIOBUFS))
+			val = dd->ipath_pioavailregs_dma[i ^ 1];
 		else
 			val = dd->ipath_pioavailregs_dma[i];
 		dd->ipath_pioavailshadow[i] = le64_to_cpu(val);
@@ -690,12 +672,13 @@ done:
  */
 int ipath_init_chip(struct ipath_devdata *dd, int reinit)
 {
-	int ret = 0, i;
+	int ret = 0;
 	u32 val32, kpiobufs;
 	u32 piobufs, uports;
 	u64 val;
 	struct ipath_portdata *pd = NULL; /* keep gcc4 happy */
 	gfp_t gfp_flags = GFP_USER | __GFP_COMP;
+	unsigned long flags;
 
 	ret = init_housekeeping(dd, &pd, reinit);
 	if (ret)
@@ -746,7 +729,7 @@ int ipath_init_chip(struct ipath_devdata *dd, int reinit)
 		kpiobufs = ipath_kpiobufs;
 
 	if (kpiobufs + (uports * IPATH_MIN_USER_PORT_BUFCNT) > piobufs) {
-		i = (int) piobufs -
+		int i = (int) piobufs -
 			(int) (uports * IPATH_MIN_USER_PORT_BUFCNT);
 		if (i < 0)
 			i = 0;
@@ -827,8 +810,12 @@ int ipath_init_chip(struct ipath_devdata *dd, int reinit)
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrclear,
 			 ~0ULL&~INFINIPATH_HWE_MEMBISTFAILED);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_control, 0ULL);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
-			 INFINIPATH_S_PIOENABLE);
+
+	spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
+	dd->ipath_sendctrl = INFINIPATH_S_PIOENABLE;
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, dd->ipath_sendctrl);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
 
 	/*
 	 * before error clears, since we expect serdes pll errors during
diff --git a/drivers/infiniband/hw/ipath/ipath_intr.c b/drivers/infiniband/hw/ipath/ipath_intr.c
index 6a5dd5cd773d..92e58c921522 100644
--- a/drivers/infiniband/hw/ipath/ipath_intr.c
+++ b/drivers/infiniband/hw/ipath/ipath_intr.c
@@ -453,7 +453,7 @@ skip_ibchange:
 }
 
 static void handle_supp_msgs(struct ipath_devdata *dd,
-			     unsigned supp_msgs, char msg[512])
+			     unsigned supp_msgs, char *msg, int msgsz)
 {
 	/*
 	 * Print the message unless it's ibc status change only, which
@@ -461,9 +461,9 @@ static void handle_supp_msgs(struct ipath_devdata *dd,
 	 */
 	if (dd->ipath_lasterror & ~INFINIPATH_E_IBSTATUSCHANGED) {
 		int iserr;
-		iserr = ipath_decode_err(msg, sizeof msg,
-				dd->ipath_lasterror &
-				~INFINIPATH_E_IBSTATUSCHANGED);
+		iserr = ipath_decode_err(msg, msgsz,
+					 dd->ipath_lasterror &
+					 ~INFINIPATH_E_IBSTATUSCHANGED);
 		if (dd->ipath_lasterror &
 			~(INFINIPATH_E_RRCVEGRFULL |
 			INFINIPATH_E_RRCVHDRFULL | INFINIPATH_E_PKTERRS))
@@ -492,8 +492,8 @@ static void handle_supp_msgs(struct ipath_devdata *dd,
 }
 
 static unsigned handle_frequent_errors(struct ipath_devdata *dd,
-				       ipath_err_t errs, char msg[512],
-				       int *noprint)
+				       ipath_err_t errs, char *msg,
+				       int msgsz, int *noprint)
 {
 	unsigned long nc;
 	static unsigned long nextmsg_time;
@@ -512,7 +512,7 @@ static unsigned handle_frequent_errors(struct ipath_devdata *dd,
 				nextmsg_time = nc + HZ * 3;
 		}
 		else if (supp_msgs) {
-			handle_supp_msgs(dd, supp_msgs, msg);
+			handle_supp_msgs(dd, supp_msgs, msg, msgsz);
 			supp_msgs = 0;
 			nmsgs = 0;
 		}
@@ -525,14 +525,14 @@ static unsigned handle_frequent_errors(struct ipath_devdata *dd,
 
 static int handle_errors(struct ipath_devdata *dd, ipath_err_t errs)
 {
-	char msg[512];
+	char msg[128];
 	u64 ignore_this_time = 0;
 	int i, iserr = 0;
 	int chkerrpkts = 0, noprint = 0;
 	unsigned supp_msgs;
 	int log_idx;
 
-	supp_msgs = handle_frequent_errors(dd, errs, msg, &noprint);
+	supp_msgs = handle_frequent_errors(dd, errs, msg, sizeof msg, &noprint);
 
 	/* don't report errors that are masked */
 	errs &= ~dd->ipath_maskederrs;
@@ -683,7 +683,7 @@ static int handle_errors(struct ipath_devdata *dd, ipath_err_t errs)
 		for (i = 0; i < dd->ipath_cfgports; i++) {
 			struct ipath_portdata *pd = dd->ipath_pd[i];
 			if (i == 0) {
-				hd = dd->ipath_port0head;
+				hd = pd->port_head;
 				tl = (u32) le64_to_cpu(
 					*dd->ipath_hdrqtailptr);
 			} else if (pd && pd->port_cnt &&
@@ -693,7 +693,7 @@ static int handle_errors(struct ipath_devdata *dd, ipath_err_t errs)
 				 * except kernel
 				 */
 				tl = *(u64 *) pd->port_rcvhdrtail_kvaddr;
-				if (tl == dd->ipath_lastrcvhdrqtails[i])
+				if (tl == pd->port_lastrcvhdrqtail)
 					continue;
 				hd = ipath_read_ureg32(dd, ur_rcvhdrhead,
 						       i);
@@ -703,7 +703,7 @@ static int handle_errors(struct ipath_devdata *dd, ipath_err_t errs)
 			    (!hd && tl == dd->ipath_hdrqlast)) {
 				if (i == 0)
 					chkerrpkts = 1;
-				dd->ipath_lastrcvhdrqtails[i] = tl;
+				pd->port_lastrcvhdrqtail = tl;
 				pd->port_hdrqfull++;
 				/* flush hdrqfull so that poll() sees it */
 				wmb();
@@ -712,6 +712,8 @@ static int handle_errors(struct ipath_devdata *dd, ipath_err_t errs)
 		}
 	}
 	if (errs & INFINIPATH_E_RRCVEGRFULL) {
+		struct ipath_portdata *pd = dd->ipath_pd[0];
+
 		/*
 		 * since this is of less importance and not likely to
 		 * happen without also getting hdrfull, only count
@@ -719,7 +721,7 @@ static int handle_errors(struct ipath_devdata *dd, ipath_err_t errs)
 		 * vs user)
 		 */
 		ipath_stats.sps_etidfull++;
-		if (dd->ipath_port0head !=
+		if (pd->port_head !=
 		    (u32) le64_to_cpu(*dd->ipath_hdrqtailptr))
 			chkerrpkts = 1;
 	}
@@ -795,6 +797,7 @@ void ipath_clear_freeze(struct ipath_devdata *dd)
 {
 	int i, im;
 	__le64 val;
+	unsigned long flags;
 
 	/* disable error interrupts, to avoid confusion */
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_errormask, 0ULL);
@@ -813,11 +816,14 @@ void ipath_clear_freeze(struct ipath_devdata *dd)
 			 dd->ipath_control);
 
 	/* ensure pio avail updates continue */
+	spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
 		 dd->ipath_sendctrl & ~INFINIPATH_S_PIOBUFAVAILUPD);
 	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
-		 dd->ipath_sendctrl);
+			 dd->ipath_sendctrl);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
 
 	/*
 	 * We just enabled pioavailupdate, so dma copy is almost certainly
@@ -825,8 +831,8 @@ void ipath_clear_freeze(struct ipath_devdata *dd)
 	 */
 	for (i = 0; i < dd->ipath_pioavregs; i++) {
 		/* deal with 6110 chip bug */
-		im = i > 3 ? ((i&1) ? i-1 : i+1) : i;
-		val = ipath_read_kreg64(dd, (0x1000/sizeof(u64))+im);
+		im = i > 3 ? i ^ 1 : i;
+		val = ipath_read_kreg64(dd, (0x1000 / sizeof(u64)) + im);
 		dd->ipath_pioavailregs_dma[i] = dd->ipath_pioavailshadow[i]
 			= le64_to_cpu(val);
 	}
@@ -849,7 +855,7 @@ void ipath_clear_freeze(struct ipath_devdata *dd)
 
 /* this is separate to allow for better optimization of ipath_intr() */
 
-static void ipath_bad_intr(struct ipath_devdata *dd, u32 * unexpectp)
+static noinline void ipath_bad_intr(struct ipath_devdata *dd, u32 *unexpectp)
 {
 	/*
 	 * sometimes happen during driver init and unload, don't want
@@ -877,7 +883,7 @@ static void ipath_bad_intr(struct ipath_devdata *dd, u32 * unexpectp)
 				dd->ipath_f_free_irq(dd);
 			}
 		}
-		if (ipath_read_kreg32(dd, dd->ipath_kregs->kr_intmask)) {
+		if (ipath_read_ireg(dd, dd->ipath_kregs->kr_intmask)) {
 			ipath_dev_err(dd, "%u unexpected interrupts, "
 				      "disabling interrupts completely\n",
 				      *unexpectp);
@@ -892,7 +898,7 @@ static void ipath_bad_intr(struct ipath_devdata *dd, u32 * unexpectp)
 			  "ignoring\n");
 }
 
-static void ipath_bad_regread(struct ipath_devdata *dd)
+static noinline void ipath_bad_regread(struct ipath_devdata *dd)
 {
 	static int allbits;
 
@@ -920,31 +926,9 @@ static void ipath_bad_regread(struct ipath_devdata *dd)
 	}
 }
 
-static void handle_port_pioavail(struct ipath_devdata *dd)
-{
-	u32 i;
-	/*
-	 * start from port 1, since for now port 0  is never using
-	 * wait_event for PIO
-	 */
-	for (i = 1; dd->ipath_portpiowait && i < dd->ipath_cfgports; i++) {
-		struct ipath_portdata *pd = dd->ipath_pd[i];
-
-		if (pd && pd->port_cnt &&
-		    dd->ipath_portpiowait & (1U << i)) {
-			clear_bit(i, &dd->ipath_portpiowait);
-			if (test_bit(IPATH_PORT_WAITING_PIO,
-				     &pd->port_flag)) {
-				clear_bit(IPATH_PORT_WAITING_PIO,
-					  &pd->port_flag);
-				wake_up_interruptible(&pd->port_wait);
-			}
-		}
-	}
-}
-
 static void handle_layer_pioavail(struct ipath_devdata *dd)
 {
+	unsigned long flags;
 	int ret;
 
 	ret = ipath_ib_piobufavail(dd->verbs_dev);
@@ -953,9 +937,12 @@ static void handle_layer_pioavail(struct ipath_devdata *dd)
 
 	return;
 set:
-	set_bit(IPATH_S_PIOINTBUFAVAIL, &dd->ipath_sendctrl);
+	spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
+	dd->ipath_sendctrl |= INFINIPATH_S_PIOINTBUFAVAIL;
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
 			 dd->ipath_sendctrl);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
 }
 
 /*
@@ -969,7 +956,15 @@ static void handle_urcv(struct ipath_devdata *dd, u32 istat)
 	int i;
 	int rcvdint = 0;
 
-	/* test_bit below needs this... */
+	/*
+	 * test_and_clear_bit(IPATH_PORT_WAITING_RCV) and
+	 * test_and_clear_bit(IPATH_PORT_WAITING_URG) below
+	 * would both like timely updates of the bits so that
+	 * we don't pass them by unnecessarily.  the rmb()
+	 * here ensures that we see them promptly -- the
+	 * corresponding wmb()'s are in ipath_poll_urgent()
+	 * and ipath_poll_next()...
+	 */
 	rmb();
 	portr = ((istat >> INFINIPATH_I_RCVAVAIL_SHIFT) &
 		 dd->ipath_i_rcvavail_mask)
@@ -980,7 +975,7 @@ static void handle_urcv(struct ipath_devdata *dd, u32 istat)
 		if (portr & (1 << i) && pd && pd->port_cnt) {
 			if (test_and_clear_bit(IPATH_PORT_WAITING_RCV,
 					       &pd->port_flag)) {
-				clear_bit(i + INFINIPATH_R_INTRAVAIL_SHIFT,
+				clear_bit(i + dd->ipath_r_intravail_shift,
 					  &dd->ipath_rcvctrl);
 				wake_up_interruptible(&pd->port_wait);
 				rcvdint = 1;
@@ -1039,7 +1034,7 @@ irqreturn_t ipath_intr(int irq, void *data)
 		goto bail;
 	}
 
-	istat = ipath_read_kreg32(dd, dd->ipath_kregs->kr_intstatus);
+	istat = ipath_read_ireg(dd, dd->ipath_kregs->kr_intstatus);
 
 	if (unlikely(!istat)) {
 		ipath_stats.sps_nullintr++;
@@ -1180,7 +1175,7 @@ irqreturn_t ipath_intr(int irq, void *data)
 	 * for receive are at the bottom.
 	 */
 	if (chk0rcv) {
-		ipath_kreceive(dd);
+		ipath_kreceive(dd->ipath_pd[0]);
 		istat &= ~port0rbits;
 	}
 
@@ -1191,12 +1186,14 @@ irqreturn_t ipath_intr(int irq, void *data)
 		handle_urcv(dd, istat);
 
 	if (istat & INFINIPATH_I_SPIOBUFAVAIL) {
-		clear_bit(IPATH_S_PIOINTBUFAVAIL, &dd->ipath_sendctrl);
+		unsigned long flags;
+
+		spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
+		dd->ipath_sendctrl &= ~INFINIPATH_S_PIOINTBUFAVAIL;
 		ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
 				 dd->ipath_sendctrl);
-
-		if (dd->ipath_portpiowait)
-			handle_port_pioavail(dd);
+		ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+		spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
 
 		handle_layer_pioavail(dd);
 	}
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
index 8786dd7922e4..4cc0f95ea877 100644
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h
@@ -41,6 +41,7 @@
 #include <linux/interrupt.h>
 #include <linux/pci.h>
 #include <linux/dma-mapping.h>
+#include <linux/mutex.h>
 #include <asm/io.h>
 #include <rdma/ib_verbs.h>
 
@@ -140,6 +141,11 @@ struct ipath_portdata {
 	u32 port_pionowait;
 	/* total number of rcvhdrqfull errors */
 	u32 port_hdrqfull;
+	/*
+	 * Used to suppress multiple instances of same
+	 * port staying stuck at same point.
+	 */
+	u32 port_lastrcvhdrqtail;
 	/* saved total number of rcvhdrqfull errors for poll edge trigger */
 	u32 port_hdrqfull_poll;
 	/* total number of polled urgent packets */
@@ -148,6 +154,7 @@ struct ipath_portdata {
 	u32 port_urgent_poll;
 	/* pid of process using this port */
 	pid_t port_pid;
+	pid_t port_subpid[INFINIPATH_MAX_SUBPORT];
 	/* same size as task_struct .comm[] */
 	char port_comm[16];
 	/* pkeys set by this use of this port */
@@ -166,6 +173,8 @@ struct ipath_portdata {
 	u32 active_slaves;
 	/* Type of packets or conditions we want to poll for */
 	u16 poll_type;
+	/* port rcvhdrq head offset */
+	u32 port_head;
 };
 
 struct sk_buff;
@@ -182,6 +191,22 @@ struct ipath_skbinfo {
 	dma_addr_t phys;
 };
 
+/*
+ * Possible IB config parameters for ipath_f_get/set_ib_cfg()
+ */
+#define IPATH_IB_CFG_LIDLMC 0 /* Get/set LID (LS16b) and Mask (MS16b) */
+#define IPATH_IB_CFG_HRTBT 1 /* Get/set Heartbeat off/enable/auto */
+#define IPATH_IB_HRTBT_ON 3 /* Heartbeat enabled, sent every 100msec */
+#define IPATH_IB_HRTBT_OFF 0 /* Heartbeat off */
+#define IPATH_IB_CFG_LWID_ENB 2 /* Get/set allowed Link-width */
+#define IPATH_IB_CFG_LWID 3 /* Get currently active Link-width */
+#define IPATH_IB_CFG_SPD_ENB 4 /* Get/set allowed Link speeds */
+#define IPATH_IB_CFG_SPD 5 /* Get current Link spd */
+#define IPATH_IB_CFG_RXPOL_ENB 6 /* Get/set Auto-RX-polarity enable */
+#define IPATH_IB_CFG_LREV_ENB 7 /* Get/set Auto-Lane-reversal enable */
+#define IPATH_IB_CFG_LINKLATENCY 8 /* Get Auto-Lane-reversal enable */
+
+
 struct ipath_devdata {
 	struct list_head ipath_list;
 
@@ -222,6 +247,8 @@ struct ipath_devdata {
 	struct _ipath_layer ipath_layer;
 	/* setup intr */
 	int (*ipath_f_intrsetup)(struct ipath_devdata *);
+	/* fallback to alternate interrupt type if possible */
+	int (*ipath_f_intr_fallback)(struct ipath_devdata *);
 	/* setup on-chip bus config */
 	int (*ipath_f_bus)(struct ipath_devdata *, struct pci_dev *);
 	/* hard reset chip */
@@ -244,6 +271,18 @@ struct ipath_devdata {
 	int (*ipath_f_get_base_info)(struct ipath_portdata *, void *);
 	/* free irq */
 	void (*ipath_f_free_irq)(struct ipath_devdata *);
+	struct ipath_message_header *(*ipath_f_get_msgheader)
+					(struct ipath_devdata *, __le32 *);
+	void (*ipath_f_config_ports)(struct ipath_devdata *, ushort);
+	int (*ipath_f_get_ib_cfg)(struct ipath_devdata *, int);
+	int (*ipath_f_set_ib_cfg)(struct ipath_devdata *, int, u32);
+	void (*ipath_f_config_jint)(struct ipath_devdata *, u16 , u16);
+	void (*ipath_f_read_counters)(struct ipath_devdata *,
+					struct infinipath_counters *);
+	void (*ipath_f_xgxs_reset)(struct ipath_devdata *);
+	/* per chip actions needed for IB Link up/down changes */
+	int (*ipath_f_ib_updown)(struct ipath_devdata *, int, u64);
+
 	struct ipath_ibdev *verbs_dev;
 	struct timer_list verbs_timer;
 	/* total dwords sent (summed from counter) */
@@ -313,22 +352,12 @@ struct ipath_devdata {
 	 * supports, less gives more pio bufs/port, etc.
 	 */
 	u32 ipath_cfgports;
-	/* port0 rcvhdrq head offset */
-	u32 ipath_port0head;
 	/* count of port 0 hdrqfull errors */
 	u32 ipath_p0_hdrqfull;
+	/* port 0 number of receive eager buffers */
+	u32 ipath_p0_rcvegrcnt;
 
 	/*
-	 * (*cfgports) used to suppress multiple instances of same
-	 * port staying stuck at same point
-	 */
-	u32 *ipath_lastrcvhdrqtails;
-	/*
-	 * (*cfgports) used to suppress multiple instances of same
-	 * port staying stuck at same point
-	 */
-	u32 *ipath_lastegrheads;
-	/*
 	 * index of last piobuffer we used.  Speeds up searching, by
 	 * starting at this point.  Doesn't matter if multiple cpu's use and
 	 * update, last updater is only write that matters.  Whenever it
@@ -367,14 +396,15 @@ struct ipath_devdata {
 	unsigned long ipath_wc_len;
 	/* ref count for each pkey */
 	atomic_t ipath_pkeyrefs[4];
-	/* shadow copy of all exptids physaddr; used only by funcsim */
-	u64 *ipath_tidsimshadow;
 	/* shadow copy of struct page *'s for exp tid pages */
 	struct page **ipath_pageshadow;
 	/* shadow copy of dma handles for exp tid pages */
 	dma_addr_t *ipath_physshadow;
-	/* lock to workaround chip bug 9437 */
+	u64 __iomem *ipath_egrtidbase;
+	/* lock to workaround chip bug 9437 and others */
+	spinlock_t ipath_kernel_tid_lock;
 	spinlock_t ipath_tid_lock;
+	spinlock_t ipath_sendctrl_lock;
 
 	/*
 	 * IPATH_STATUS_*,
@@ -395,6 +425,8 @@ struct ipath_devdata {
 	void *ipath_dummy_hdrq;	/* used after port close */
 	dma_addr_t ipath_dummy_hdrq_phys;
 
+	unsigned long ipath_ureg_align; /* user register alignment */
+
 	/*
 	 * Shadow copies of registers; size indicates read access size.
 	 * Most of them are readonly, but some are write-only register,
@@ -456,8 +488,6 @@ struct ipath_devdata {
 	unsigned long ipath_rcvctrl;
 	/* shadow kr_sendctrl */
 	unsigned long ipath_sendctrl;
-	/* ports waiting for PIOavail intr */
-	unsigned long ipath_portpiowait;
 	unsigned long ipath_lastcancel; /* to not count armlaunch after cancel */
 
 	/* value we put in kr_rcvhdrcnt */
@@ -550,12 +580,26 @@ struct ipath_devdata {
 	u8 ipath_minrev;
 	/* board rev, from ipath_revision */
 	u8 ipath_boardrev;
+
+	u8 ipath_r_portenable_shift;
+	u8 ipath_r_intravail_shift;
+	u8 ipath_r_tailupd_shift;
+	u8 ipath_r_portcfg_shift;
+
 	/* unit # of this chip, if present */
 	int ipath_unit;
 	/* saved for restore after reset */
 	u8 ipath_pci_cacheline;
 	/* LID mask control */
 	u8 ipath_lmc;
+	/* link width supported */
+	u8 ipath_link_width_supported;
+	/* link speed supported */
+	u8 ipath_link_speed_supported;
+	u8 ipath_link_width_enabled;
+	u8 ipath_link_speed_enabled;
+	u8 ipath_link_width_active;
+	u8 ipath_link_speed_active;
 	/* Rx Polarity inversion (compensate for ~tx on partner) */
 	u8 ipath_rx_pol_inv;
 
@@ -590,6 +634,8 @@ struct ipath_devdata {
 	 */
 	u32 ipath_i_rcvavail_mask;
 	u32 ipath_i_rcvurg_mask;
+	u16 ipath_i_rcvurg_shift;
+	u16 ipath_i_rcvavail_shift;
 
 	/*
 	 * Register bits for selecting i2c direction and values, used for
@@ -603,6 +649,29 @@ struct ipath_devdata {
 	/* lock for doing RMW of shadows/regs for ExtCtrl and GPIO */
 	spinlock_t ipath_gpio_lock;
 
+	/*
+	 * IB link and linktraining states and masks that vary per chip in
+	 * some way.  Set at init, to avoid each IB status change interrupt
+	 */
+	u8 ibcs_ls_shift;
+	u8 ibcs_lts_mask;
+	u32 ibcs_mask;
+	u32 ib_init;
+	u32 ib_arm;
+	u32 ib_active;
+
+	u16 ipath_rhf_offset; /* offset of RHF within receive header entry */
+
+	/*
+	 * shift/mask for linkcmd, linkinitcmd, maxpktlen in ibccontol
+	 * reg. Changes for IBA7220
+	 */
+	u8 ibcc_lic_mask; /* LinkInitCmd */
+	u8 ibcc_lc_shift; /* LinkCmd */
+	u8 ibcc_mpl_shift; /* Maxpktlen */
+
+	u8 delay_mult;
+
 	/* used to override LED behavior */
 	u8 ipath_led_override;  /* Substituted for normal value, if non-zero */
 	u16 ipath_led_override_timeoff; /* delta to next timer event */
@@ -616,7 +685,7 @@ struct ipath_devdata {
 	/* control access to actual counters, timer */
 	spinlock_t ipath_eep_st_lock;
 	/* control high-level access to EEPROM */
-	struct semaphore ipath_eep_sem;
+	struct mutex ipath_eep_lock;
 	/* Below inc'd by ipath_snap_cntrs(), locked by ipath_eep_st_lock */
 	uint64_t ipath_traffic_wds;
 	/* active time is kept in seconds, but logged in hours */
@@ -630,6 +699,10 @@ struct ipath_devdata {
 	 * each of the counters to increment.
 	 */
 	struct ipath_eep_log_mask ipath_eep_st_masks[IPATH_EEP_LOG_CNT];
+
+	/* interrupt mitigation reload register info */
+	u16 ipath_jint_idle_ticks;	/* idle clock ticks */
+	u16 ipath_jint_max_packets;	/* max packets across all ports */
 };
 
 /* Private data for file operations */
@@ -690,7 +763,7 @@ void ipath_free_pddata(struct ipath_devdata *, struct ipath_portdata *);
 
 int ipath_parse_ushort(const char *str, unsigned short *valp);
 
-void ipath_kreceive(struct ipath_devdata *);
+void ipath_kreceive(struct ipath_portdata *);
 int ipath_setrcvhdrsize(struct ipath_devdata *, unsigned);
 int ipath_reset_device(int);
 void ipath_get_faststats(unsigned long);
@@ -698,6 +771,8 @@ int ipath_set_linkstate(struct ipath_devdata *, u8);
 int ipath_set_mtu(struct ipath_devdata *, u16);
 int ipath_set_lid(struct ipath_devdata *, u32, u8);
 int ipath_set_rx_pol_inv(struct ipath_devdata *dd, u8 new_pol_inv);
+void ipath_enable_armlaunch(struct ipath_devdata *);
+void ipath_disable_armlaunch(struct ipath_devdata *);
 
 /* for use in system calls, where we want to know device type, etc. */
 #define port_fp(fp) ((struct ipath_filedata *)(fp)->private_data)->pd
@@ -744,9 +819,15 @@ int ipath_set_rx_pol_inv(struct ipath_devdata *dd, u8 new_pol_inv);
 		 * are 64bit */
 #define IPATH_32BITCOUNTERS 0x20000
 		/* can miss port0 rx interrupts */
+		/* Interrupt register is 64 bits */
+#define IPATH_INTREG_64     0x40000
 #define IPATH_DISABLED      0x80000 /* administratively disabled */
 		/* Use GPIO interrupts for new counters */
 #define IPATH_GPIO_ERRINTRS 0x100000
+#define IPATH_SWAP_PIOBUFS  0x200000
+		/* Suppress heartbeat, even if turning off loopback */
+#define IPATH_NO_HRTBT      0x1000000
+#define IPATH_HAS_MULT_IB_SPEED 0x8000000
 
 /* Bits in GPIO for the added interrupts */
 #define IPATH_GPIO_PORT0_BIT 2
@@ -758,8 +839,6 @@ int ipath_set_rx_pol_inv(struct ipath_devdata *dd, u8 new_pol_inv);
 /* portdata flag bit offsets */
 		/* waiting for a packet to arrive */
 #define IPATH_PORT_WAITING_RCV   2
-		/* waiting for a PIO buffer to be available */
-#define IPATH_PORT_WAITING_PIO   3
 		/* master has not finished initializing */
 #define IPATH_PORT_MASTER_UNINIT 4
 		/* waiting for an urgent packet to arrive */
@@ -767,8 +846,6 @@ int ipath_set_rx_pol_inv(struct ipath_devdata *dd, u8 new_pol_inv);
 
 /* free up any allocated data at closes */
 void ipath_free_data(struct ipath_portdata *dd);
-int ipath_waitfor_mdio_cmdready(struct ipath_devdata *);
-int ipath_waitfor_complete(struct ipath_devdata *, ipath_kreg, u64, u64 *);
 u32 __iomem *ipath_getpiobuf(struct ipath_devdata *, u32 *);
 void ipath_init_iba6120_funcs(struct ipath_devdata *);
 void ipath_init_iba6110_funcs(struct ipath_devdata *);
@@ -792,33 +869,6 @@ void ipath_set_led_override(struct ipath_devdata *dd, unsigned int val);
  */
 #define IPATH_DFLT_RCVHDRSIZE 9
 
-#define IPATH_MDIO_CMD_WRITE   1
-#define IPATH_MDIO_CMD_READ    2
-#define IPATH_MDIO_CLD_DIV     25	/* to get 2.5 Mhz mdio clock */
-#define IPATH_MDIO_CMDVALID    0x40000000	/* bit 30 */
-#define IPATH_MDIO_DATAVALID   0x80000000	/* bit 31 */
-#define IPATH_MDIO_CTRL_STD    0x0
-
-static inline u64 ipath_mdio_req(int cmd, int dev, int reg, int data)
-{
-	return (((u64) IPATH_MDIO_CLD_DIV) << 32) |
-		(cmd << 26) |
-		(dev << 21) |
-		(reg << 16) |
-		(data & 0xFFFF);
-}
-
-		/* signal and fifo status, in bank 31 */
-#define IPATH_MDIO_CTRL_XGXS_REG_8  0x8
-		/* controls loopback, redundancy */
-#define IPATH_MDIO_CTRL_8355_REG_1  0x10
-		/* premph, encdec, etc. */
-#define IPATH_MDIO_CTRL_8355_REG_2  0x11
-		/* Kchars, etc. */
-#define IPATH_MDIO_CTRL_8355_REG_6  0x15
-#define IPATH_MDIO_CTRL_8355_REG_9  0x18
-#define IPATH_MDIO_CTRL_8355_REG_10 0x1D
-
 int ipath_get_user_pages(unsigned long, size_t, struct page **);
 void ipath_release_user_pages(struct page **, size_t);
 void ipath_release_user_pages_on_close(struct page **, size_t);
@@ -863,7 +913,7 @@ static inline u32 ipath_read_ureg32(const struct ipath_devdata *dd,
 	return readl(regno + (u64 __iomem *)
 		     (dd->ipath_uregbase +
 		      (char __iomem *)dd->ipath_kregbase +
-		      dd->ipath_palign * port));
+		      dd->ipath_ureg_align * port));
 }
 
 /**
@@ -880,7 +930,7 @@ static inline void ipath_write_ureg(const struct ipath_devdata *dd,
 {
 	u64 __iomem *ubase = (u64 __iomem *)
 		(dd->ipath_uregbase + (char __iomem *) dd->ipath_kregbase +
-		 dd->ipath_palign * port);
+		 dd->ipath_ureg_align * port);
 	if (dd->ipath_kregbase)
 		writeq(value, &ubase[regno]);
 }
@@ -930,6 +980,53 @@ static inline u32 ipath_read_creg32(const struct ipath_devdata *dd,
 		      (char __iomem *)dd->ipath_kregbase));
 }
 
+static inline void ipath_write_creg(const struct ipath_devdata *dd,
+				    ipath_creg regno, u64 value)
+{
+	if (dd->ipath_kregbase)
+		writeq(value, regno + (u64 __iomem *)
+		       (dd->ipath_cregbase +
+			(char __iomem *)dd->ipath_kregbase));
+}
+
+static inline void ipath_clear_rcvhdrtail(const struct ipath_portdata *pd)
+{
+	*((u64 *) pd->port_rcvhdrtail_kvaddr) = 0ULL;
+}
+
+static inline u32 ipath_get_rcvhdrtail(const struct ipath_portdata *pd)
+{
+	return (u32) le64_to_cpu(*((volatile __le64 *)
+				pd->port_rcvhdrtail_kvaddr));
+}
+
+static inline u64 ipath_read_ireg(const struct ipath_devdata *dd, ipath_kreg r)
+{
+	return (dd->ipath_flags & IPATH_INTREG_64) ?
+		ipath_read_kreg64(dd, r) : ipath_read_kreg32(dd, r);
+}
+
+/*
+ * from contents of IBCStatus (or a saved copy), return linkstate
+ * Report ACTIVE_DEFER as ACTIVE, because we treat them the same
+ * everywhere, anyway (and should be, for almost all purposes).
+ */
+static inline u32 ipath_ib_linkstate(struct ipath_devdata *dd, u64 ibcs)
+{
+	u32 state = (u32)(ibcs >> dd->ibcs_ls_shift) &
+		INFINIPATH_IBCS_LINKSTATE_MASK;
+	if (state == INFINIPATH_IBCS_L_STATE_ACT_DEFER)
+		state = INFINIPATH_IBCS_L_STATE_ACTIVE;
+	return state;
+}
+
+/* from contents of IBCStatus (or a saved copy), return linktrainingstate */
+static inline u32 ipath_ib_linktrstate(struct ipath_devdata *dd, u64 ibcs)
+{
+	return (u32)(ibcs >> INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) &
+		dd->ibcs_lts_mask;
+}
+
 /*
  * sysfs interface.
  */
@@ -938,8 +1035,7 @@ struct device_driver;
 
 extern const char ib_ipath_version[];
 
-int ipath_driver_create_group(struct device_driver *);
-void ipath_driver_remove_group(struct device_driver *);
+extern struct attribute_group *ipath_driver_attr_groups[];
 
 int ipath_device_create_group(struct device *, struct ipath_devdata *);
 void ipath_device_remove_group(struct device *, struct ipath_devdata *);
diff --git a/drivers/infiniband/hw/ipath/ipath_keys.c b/drivers/infiniband/hw/ipath/ipath_keys.c
index 85a4aefc6c03..8f32b17a5eed 100644
--- a/drivers/infiniband/hw/ipath/ipath_keys.c
+++ b/drivers/infiniband/hw/ipath/ipath_keys.c
@@ -128,9 +128,8 @@ int ipath_lkey_ok(struct ipath_qp *qp, struct ipath_sge *isge,
 	int ret;
 
 	/*
-	 * We use LKEY == zero to mean a physical kmalloc() address.
-	 * This is a bit of a hack since we rely on dma_map_single()
-	 * being reversible by calling bus_to_virt().
+	 * We use LKEY == zero for kernel virtual addresses
+	 * (see ipath_get_dma_mr and ipath_dma.c).
 	 */
 	if (sge->lkey == 0) {
 		struct ipath_pd *pd = to_ipd(qp->ibqp.pd);
diff --git a/drivers/infiniband/hw/ipath/ipath_mad.c b/drivers/infiniband/hw/ipath/ipath_mad.c
index 3d1432d1e3f4..d98d5f103700 100644
--- a/drivers/infiniband/hw/ipath/ipath_mad.c
+++ b/drivers/infiniband/hw/ipath/ipath_mad.c
@@ -934,6 +934,7 @@ static int recv_pma_get_portsamplescontrol(struct ib_perf *pmp,
 	struct ib_pma_portsamplescontrol *p =
 		(struct ib_pma_portsamplescontrol *)pmp->data;
 	struct ipath_ibdev *dev = to_idev(ibdev);
+	struct ipath_cregs const *crp = dev->dd->ipath_cregs;
 	unsigned long flags;
 	u8 port_select = p->port_select;
 
@@ -955,7 +956,10 @@ static int recv_pma_get_portsamplescontrol(struct ib_perf *pmp,
 	p->counter_width = 4;	/* 32 bit counters */
 	p->counter_mask0_9 = COUNTER_MASK0_9;
 	spin_lock_irqsave(&dev->pending_lock, flags);
-	p->sample_status = dev->pma_sample_status;
+	if (crp->cr_psstat)
+		p->sample_status = ipath_read_creg32(dev->dd, crp->cr_psstat);
+	else
+		p->sample_status = dev->pma_sample_status;
 	p->sample_start = cpu_to_be32(dev->pma_sample_start);
 	p->sample_interval = cpu_to_be32(dev->pma_sample_interval);
 	p->tag = cpu_to_be16(dev->pma_tag);
@@ -975,8 +979,9 @@ static int recv_pma_set_portsamplescontrol(struct ib_perf *pmp,
 	struct ib_pma_portsamplescontrol *p =
 		(struct ib_pma_portsamplescontrol *)pmp->data;
 	struct ipath_ibdev *dev = to_idev(ibdev);
+	struct ipath_cregs const *crp = dev->dd->ipath_cregs;
 	unsigned long flags;
-	u32 start;
+	u8 status;
 	int ret;
 
 	if (pmp->attr_mod != 0 ||
@@ -986,59 +991,67 @@ static int recv_pma_set_portsamplescontrol(struct ib_perf *pmp,
 		goto bail;
 	}
 
-	start = be32_to_cpu(p->sample_start);
-	if (start != 0) {
-		spin_lock_irqsave(&dev->pending_lock, flags);
-		if (dev->pma_sample_status == IB_PMA_SAMPLE_STATUS_DONE) {
-			dev->pma_sample_status =
-				IB_PMA_SAMPLE_STATUS_STARTED;
-			dev->pma_sample_start = start;
-			dev->pma_sample_interval =
-				be32_to_cpu(p->sample_interval);
-			dev->pma_tag = be16_to_cpu(p->tag);
-			if (p->counter_select[0])
-				dev->pma_counter_select[0] =
-					p->counter_select[0];
-			if (p->counter_select[1])
-				dev->pma_counter_select[1] =
-					p->counter_select[1];
-			if (p->counter_select[2])
-				dev->pma_counter_select[2] =
-					p->counter_select[2];
-			if (p->counter_select[3])
-				dev->pma_counter_select[3] =
-					p->counter_select[3];
-			if (p->counter_select[4])
-				dev->pma_counter_select[4] =
-					p->counter_select[4];
-		}
-		spin_unlock_irqrestore(&dev->pending_lock, flags);
+	spin_lock_irqsave(&dev->pending_lock, flags);
+	if (crp->cr_psstat)
+		status = ipath_read_creg32(dev->dd, crp->cr_psstat);
+	else
+		status = dev->pma_sample_status;
+	if (status == IB_PMA_SAMPLE_STATUS_DONE) {
+		dev->pma_sample_start = be32_to_cpu(p->sample_start);
+		dev->pma_sample_interval = be32_to_cpu(p->sample_interval);
+		dev->pma_tag = be16_to_cpu(p->tag);
+		dev->pma_counter_select[0] = p->counter_select[0];
+		dev->pma_counter_select[1] = p->counter_select[1];
+		dev->pma_counter_select[2] = p->counter_select[2];
+		dev->pma_counter_select[3] = p->counter_select[3];
+		dev->pma_counter_select[4] = p->counter_select[4];
+		if (crp->cr_psstat) {
+			ipath_write_creg(dev->dd, crp->cr_psinterval,
+					 dev->pma_sample_interval);
+			ipath_write_creg(dev->dd, crp->cr_psstart,
+					 dev->pma_sample_start);
+		} else
+			dev->pma_sample_status = IB_PMA_SAMPLE_STATUS_STARTED;
 	}
+	spin_unlock_irqrestore(&dev->pending_lock, flags);
+
 	ret = recv_pma_get_portsamplescontrol(pmp, ibdev, port);
 
 bail:
 	return ret;
 }
 
-static u64 get_counter(struct ipath_ibdev *dev, __be16 sel)
+static u64 get_counter(struct ipath_ibdev *dev,
+		       struct ipath_cregs const *crp,
+		       __be16 sel)
 {
 	u64 ret;
 
 	switch (sel) {
 	case IB_PMA_PORT_XMIT_DATA:
-		ret = dev->ipath_sword;
+		ret = (crp->cr_psxmitdatacount) ?
+			ipath_read_creg32(dev->dd, crp->cr_psxmitdatacount) :
+			dev->ipath_sword;
 		break;
 	case IB_PMA_PORT_RCV_DATA:
-		ret = dev->ipath_rword;
+		ret = (crp->cr_psrcvdatacount) ?
+			ipath_read_creg32(dev->dd, crp->cr_psrcvdatacount) :
+			dev->ipath_rword;
 		break;
 	case IB_PMA_PORT_XMIT_PKTS:
-		ret = dev->ipath_spkts;
+		ret = (crp->cr_psxmitpktscount) ?
+			ipath_read_creg32(dev->dd, crp->cr_psxmitpktscount) :
+			dev->ipath_spkts;
 		break;
 	case IB_PMA_PORT_RCV_PKTS:
-		ret = dev->ipath_rpkts;
+		ret = (crp->cr_psrcvpktscount) ?
+			ipath_read_creg32(dev->dd, crp->cr_psrcvpktscount) :
+			dev->ipath_rpkts;
 		break;
 	case IB_PMA_PORT_XMIT_WAIT:
-		ret = dev->ipath_xmit_wait;
+		ret = (crp->cr_psxmitwaitcount) ?
+			ipath_read_creg32(dev->dd, crp->cr_psxmitwaitcount) :
+			dev->ipath_xmit_wait;
 		break;
 	default:
 		ret = 0;
@@ -1053,14 +1066,21 @@ static int recv_pma_get_portsamplesresult(struct ib_perf *pmp,
 	struct ib_pma_portsamplesresult *p =
 		(struct ib_pma_portsamplesresult *)pmp->data;
 	struct ipath_ibdev *dev = to_idev(ibdev);
+	struct ipath_cregs const *crp = dev->dd->ipath_cregs;
+	u8 status;
 	int i;
 
 	memset(pmp->data, 0, sizeof(pmp->data));
 	p->tag = cpu_to_be16(dev->pma_tag);
-	p->sample_status = cpu_to_be16(dev->pma_sample_status);
+	if (crp->cr_psstat)
+		status = ipath_read_creg32(dev->dd, crp->cr_psstat);
+	else
+		status = dev->pma_sample_status;
+	p->sample_status = cpu_to_be16(status);
 	for (i = 0; i < ARRAY_SIZE(dev->pma_counter_select); i++)
-		p->counter[i] = cpu_to_be32(
-			get_counter(dev, dev->pma_counter_select[i]));
+		p->counter[i] = (status != IB_PMA_SAMPLE_STATUS_DONE) ? 0 :
+		    cpu_to_be32(
+			get_counter(dev, crp, dev->pma_counter_select[i]));
 
 	return reply((struct ib_smp *) pmp);
 }
@@ -1071,16 +1091,23 @@ static int recv_pma_get_portsamplesresult_ext(struct ib_perf *pmp,
 	struct ib_pma_portsamplesresult_ext *p =
 		(struct ib_pma_portsamplesresult_ext *)pmp->data;
 	struct ipath_ibdev *dev = to_idev(ibdev);
+	struct ipath_cregs const *crp = dev->dd->ipath_cregs;
+	u8 status;
 	int i;
 
 	memset(pmp->data, 0, sizeof(pmp->data));
 	p->tag = cpu_to_be16(dev->pma_tag);
-	p->sample_status = cpu_to_be16(dev->pma_sample_status);
+	if (crp->cr_psstat)
+		status = ipath_read_creg32(dev->dd, crp->cr_psstat);
+	else
+		status = dev->pma_sample_status;
+	p->sample_status = cpu_to_be16(status);
 	/* 64 bits */
 	p->extended_width = __constant_cpu_to_be32(0x80000000);
 	for (i = 0; i < ARRAY_SIZE(dev->pma_counter_select); i++)
-		p->counter[i] = cpu_to_be64(
-			get_counter(dev, dev->pma_counter_select[i]));
+		p->counter[i] = (status != IB_PMA_SAMPLE_STATUS_DONE) ? 0 :
+		    cpu_to_be64(
+			get_counter(dev, crp, dev->pma_counter_select[i]));
 
 	return reply((struct ib_smp *) pmp);
 }
@@ -1113,6 +1140,8 @@ static int recv_pma_get_portcounters(struct ib_perf *pmp,
 		dev->z_local_link_integrity_errors;
 	cntrs.excessive_buffer_overrun_errors -=
 		dev->z_excessive_buffer_overrun_errors;
+	cntrs.vl15_dropped -= dev->z_vl15_dropped;
+	cntrs.vl15_dropped += dev->n_vl15_dropped;
 
 	memset(pmp->data, 0, sizeof(pmp->data));
 
@@ -1156,10 +1185,10 @@ static int recv_pma_get_portcounters(struct ib_perf *pmp,
 		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)
+	if (cntrs.vl15_dropped > 0xFFFFUL)
 		p->vl15_dropped = __constant_cpu_to_be16(0xFFFF);
 	else
-		p->vl15_dropped = cpu_to_be16((u16)dev->n_vl15_dropped);
+		p->vl15_dropped = cpu_to_be16((u16)cntrs.vl15_dropped);
 	if (cntrs.port_xmit_data > 0xFFFFFFFFUL)
 		p->port_xmit_data = __constant_cpu_to_be32(0xFFFFFFFF);
 	else
@@ -1262,8 +1291,10 @@ static int recv_pma_set_portcounters(struct ib_perf *pmp,
 		dev->z_excessive_buffer_overrun_errors =
 			cntrs.excessive_buffer_overrun_errors;
 
-	if (p->counter_select & IB_PMA_SEL_PORT_VL15_DROPPED)
+	if (p->counter_select & IB_PMA_SEL_PORT_VL15_DROPPED) {
 		dev->n_vl15_dropped = 0;
+		dev->z_vl15_dropped = cntrs.vl15_dropped;
+	}
 
 	if (p->counter_select & IB_PMA_SEL_PORT_XMIT_DATA)
 		dev->z_port_xmit_data = cntrs.port_xmit_data;
@@ -1434,7 +1465,7 @@ static int process_subn(struct ib_device *ibdev, int mad_flags,
 		 * before checking for other consumers.
 		 * Just tell the caller to process it normally.
 		 */
-		ret = IB_MAD_RESULT_FAILURE;
+		ret = IB_MAD_RESULT_SUCCESS;
 		goto bail;
 	default:
 		smp->status |= IB_SMP_UNSUP_METHOD;
@@ -1516,7 +1547,7 @@ static int process_perf(struct ib_device *ibdev, u8 port_num,
 		 * before checking for other consumers.
 		 * Just tell the caller to process it normally.
 		 */
-		ret = IB_MAD_RESULT_FAILURE;
+		ret = IB_MAD_RESULT_SUCCESS;
 		goto bail;
 	default:
 		pmp->status |= IB_SMP_UNSUP_METHOD;
diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c
index 6a41fdbc8e57..80dc623cee40 100644
--- a/drivers/infiniband/hw/ipath/ipath_qp.c
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c
@@ -387,8 +387,8 @@ int ipath_error_qp(struct ipath_qp *qp, enum ib_wc_status err)
 	struct ib_wc wc;
 	int ret = 0;
 
-	ipath_dbg("QP%d/%d in error state\n",
-		  qp->ibqp.qp_num, qp->remote_qpn);
+	ipath_dbg("QP%d/%d in error state (%d)\n",
+		  qp->ibqp.qp_num, qp->remote_qpn, err);
 
 	spin_lock(&dev->pending_lock);
 	/* XXX What if its already removed by the timeout code? */
@@ -835,7 +835,8 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd,
 				      init_attr->qp_type);
 		if (err) {
 			ret = ERR_PTR(err);
-			goto bail_rwq;
+			vfree(qp->r_rq.wq);
+			goto bail_qp;
 		}
 		qp->ip = NULL;
 		ipath_reset_qp(qp);
@@ -854,8 +855,6 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd,
 	 * See ipath_mmap() for details.
 	 */
 	if (udata && udata->outlen >= sizeof(__u64)) {
-		int err;
-
 		if (!qp->r_rq.wq) {
 			__u64 offset = 0;
 
@@ -863,7 +862,7 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd,
 					       sizeof(offset));
 			if (err) {
 				ret = ERR_PTR(err);
-				goto bail_rwq;
+				goto bail_ip;
 			}
 		} else {
 			u32 s = sizeof(struct ipath_rwq) +
@@ -875,7 +874,7 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd,
 						   qp->r_rq.wq);
 			if (!qp->ip) {
 				ret = ERR_PTR(-ENOMEM);
-				goto bail_rwq;
+				goto bail_ip;
 			}
 
 			err = ib_copy_to_udata(udata, &(qp->ip->offset),
@@ -907,9 +906,11 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd,
 	goto bail;
 
 bail_ip:
-	kfree(qp->ip);
-bail_rwq:
-	vfree(qp->r_rq.wq);
+	if (qp->ip)
+		kref_put(&qp->ip->ref, ipath_release_mmap_info);
+	else
+		vfree(qp->r_rq.wq);
+	ipath_free_qp(&dev->qp_table, qp);
 bail_qp:
 	kfree(qp);
 bail_swq:
diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c
index 5c29b2bfea17..459e46e2c016 100644
--- a/drivers/infiniband/hw/ipath/ipath_rc.c
+++ b/drivers/infiniband/hw/ipath/ipath_rc.c
@@ -647,6 +647,7 @@ static void send_rc_ack(struct ipath_qp *qp)
 
 queue_ack:
 	spin_lock_irqsave(&qp->s_lock, flags);
+	dev->n_rc_qacks++;
 	qp->s_flags |= IPATH_S_ACK_PENDING;
 	qp->s_nak_state = qp->r_nak_state;
 	qp->s_ack_psn = qp->r_ack_psn;
@@ -798,11 +799,13 @@ bail:
 
 static inline void update_last_psn(struct ipath_qp *qp, u32 psn)
 {
-	if (qp->s_wait_credit) {
-		qp->s_wait_credit = 0;
-		tasklet_hi_schedule(&qp->s_task);
+	if (qp->s_last_psn != psn) {
+		qp->s_last_psn = psn;
+		if (qp->s_wait_credit) {
+			qp->s_wait_credit = 0;
+			tasklet_hi_schedule(&qp->s_task);
+		}
 	}
-	qp->s_last_psn = psn;
 }
 
 /**
@@ -959,8 +962,9 @@ static int do_rc_ack(struct ipath_qp *qp, u32 aeth, u32 psn, int opcode,
 		/* If this is a partial ACK, reset the retransmit timer. */
 		if (qp->s_last != qp->s_tail) {
 			spin_lock(&dev->pending_lock);
-			list_add_tail(&qp->timerwait,
-				      &dev->pending[dev->pending_index]);
+			if (list_empty(&qp->timerwait))
+				list_add_tail(&qp->timerwait,
+					&dev->pending[dev->pending_index]);
 			spin_unlock(&dev->pending_lock);
 			/*
 			 * If we get a partial ACK for a resent operation,
@@ -1652,13 +1656,6 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	case OP(SEND_FIRST):
 		if (!ipath_get_rwqe(qp, 0)) {
 		rnr_nak:
-			/*
-			 * A RNR NAK will ACK earlier sends and RDMA writes.
-			 * Don't queue the NAK if a RDMA read or atomic
-			 * is pending though.
-			 */
-			if (qp->r_nak_state)
-				goto done;
 			qp->r_nak_state = IB_RNR_NAK | qp->r_min_rnr_timer;
 			qp->r_ack_psn = qp->r_psn;
 			goto send_ack;
diff --git a/drivers/infiniband/hw/ipath/ipath_registers.h b/drivers/infiniband/hw/ipath/ipath_registers.h
index 708eba3165d7..6d2a17f9c1da 100644
--- a/drivers/infiniband/hw/ipath/ipath_registers.h
+++ b/drivers/infiniband/hw/ipath/ipath_registers.h
@@ -82,8 +82,7 @@
 
 /* kr_rcvctrl bits */
 #define INFINIPATH_R_PORTENABLE_SHIFT 0
-#define INFINIPATH_R_INTRAVAIL_SHIFT 16
-#define INFINIPATH_R_TAILUPD   0x80000000
+#define INFINIPATH_R_QPMAP_ENABLE (1ULL << 38)
 
 /* kr_intstatus, kr_intclear, kr_intmask bits */
 #define INFINIPATH_I_RCVURG_SHIFT 0
@@ -272,20 +271,6 @@
 #define INFINIPATH_EXTC_LEDGBLOK_ON          0x00000002ULL
 #define INFINIPATH_EXTC_LEDGBLERR_OFF        0x00000001ULL
 
-/* kr_mdio bits */
-#define INFINIPATH_MDIO_CLKDIV_MASK 0x7FULL
-#define INFINIPATH_MDIO_CLKDIV_SHIFT 32
-#define INFINIPATH_MDIO_COMMAND_MASK 0x7ULL
-#define INFINIPATH_MDIO_COMMAND_SHIFT 26
-#define INFINIPATH_MDIO_DEVADDR_MASK 0x1FULL
-#define INFINIPATH_MDIO_DEVADDR_SHIFT 21
-#define INFINIPATH_MDIO_REGADDR_MASK 0x1FULL
-#define INFINIPATH_MDIO_REGADDR_SHIFT 16
-#define INFINIPATH_MDIO_DATA_MASK 0xFFFFULL
-#define INFINIPATH_MDIO_DATA_SHIFT 0
-#define INFINIPATH_MDIO_CMDVALID    0x0000000040000000ULL
-#define INFINIPATH_MDIO_RDDATAVALID 0x0000000080000000ULL
-
 /* kr_partitionkey bits */
 #define INFINIPATH_PKEY_SIZE 16
 #define INFINIPATH_PKEY_MASK 0xFFFF
@@ -303,8 +288,6 @@
 
 /* kr_xgxsconfig bits */
 #define INFINIPATH_XGXS_RESET          0x7ULL
-#define INFINIPATH_XGXS_MDIOADDR_MASK  0xfULL
-#define INFINIPATH_XGXS_MDIOADDR_SHIFT 4
 #define INFINIPATH_XGXS_RX_POL_SHIFT 19
 #define INFINIPATH_XGXS_RX_POL_MASK 0xfULL
 
@@ -470,6 +453,20 @@ struct ipath_cregs {
 	ipath_creg cr_unsupvlcnt;
 	ipath_creg cr_wordrcvcnt;
 	ipath_creg cr_wordsendcnt;
+	ipath_creg cr_vl15droppedpktcnt;
+	ipath_creg cr_rxotherlocalphyerrcnt;
+	ipath_creg cr_excessbufferovflcnt;
+	ipath_creg cr_locallinkintegrityerrcnt;
+	ipath_creg cr_rxvlerrcnt;
+	ipath_creg cr_rxdlidfltrcnt;
+	ipath_creg cr_psstat;
+	ipath_creg cr_psstart;
+	ipath_creg cr_psinterval;
+	ipath_creg cr_psrcvdatacount;
+	ipath_creg cr_psrcvpktscount;
+	ipath_creg cr_psxmitdatacount;
+	ipath_creg cr_psxmitpktscount;
+	ipath_creg cr_psxmitwaitcount;
 };
 
 #endif				/* _IPATH_REGISTERS_H */
diff --git a/drivers/infiniband/hw/ipath/ipath_ruc.c b/drivers/infiniband/hw/ipath/ipath_ruc.c
index 4b6b7ee8e5c1..a59bdbd0ed87 100644
--- a/drivers/infiniband/hw/ipath/ipath_ruc.c
+++ b/drivers/infiniband/hw/ipath/ipath_ruc.c
@@ -98,11 +98,15 @@ void ipath_insert_rnr_queue(struct ipath_qp *qp)
 		while (qp->s_rnr_timeout >= nqp->s_rnr_timeout) {
 			qp->s_rnr_timeout -= nqp->s_rnr_timeout;
 			l = l->next;
-			if (l->next == &dev->rnrwait)
+			if (l->next == &dev->rnrwait) {
+				nqp = NULL;
 				break;
+			}
 			nqp = list_entry(l->next, struct ipath_qp,
 					 timerwait);
 		}
+		if (nqp)
+			nqp->s_rnr_timeout -= qp->s_rnr_timeout;
 		list_add(&qp->timerwait, l);
 	}
 	spin_unlock_irqrestore(&dev->pending_lock, flags);
@@ -479,9 +483,14 @@ done:
 
 static void want_buffer(struct ipath_devdata *dd)
 {
-	set_bit(IPATH_S_PIOINTBUFAVAIL, &dd->ipath_sendctrl);
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->ipath_sendctrl_lock, flags);
+	dd->ipath_sendctrl |= INFINIPATH_S_PIOINTBUFAVAIL;
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
 			 dd->ipath_sendctrl);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	spin_unlock_irqrestore(&dd->ipath_sendctrl_lock, flags);
 }
 
 /**
@@ -630,11 +639,8 @@ bail:;
 void ipath_send_complete(struct ipath_qp *qp, struct ipath_swqe *wqe,
 			 enum ib_wc_status status)
 {
-	u32 last = qp->s_last;
-
-	if (++last == qp->s_size)
-		last = 0;
-	qp->s_last = last;
+	unsigned long flags;
+	u32 last;
 
 	/* See ch. 11.2.4.1 and 10.7.3.1 */
 	if (!(qp->s_flags & IPATH_S_SIGNAL_REQ_WR) ||
@@ -658,4 +664,11 @@ void ipath_send_complete(struct ipath_qp *qp, struct ipath_swqe *wqe,
 		wc.port_num = 0;
 		ipath_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 0);
 	}
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+	last = qp->s_last;
+	if (++last >= qp->s_size)
+		last = 0;
+	qp->s_last = last;
+	spin_unlock_irqrestore(&qp->s_lock, flags);
 }
diff --git a/drivers/infiniband/hw/ipath/ipath_srq.c b/drivers/infiniband/hw/ipath/ipath_srq.c
index 40c36ec19016..f772102e4713 100644
--- a/drivers/infiniband/hw/ipath/ipath_srq.c
+++ b/drivers/infiniband/hw/ipath/ipath_srq.c
@@ -59,7 +59,7 @@ int ipath_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
 
 		if ((unsigned) wr->num_sge > srq->rq.max_sge) {
 			*bad_wr = wr;
-			ret = -ENOMEM;
+			ret = -EINVAL;
 			goto bail;
 		}
 
@@ -94,8 +94,8 @@ bail:
 /**
  * ipath_create_srq - create a shared receive queue
  * @ibpd: the protection domain of the SRQ to create
- * @attr: the attributes of the SRQ
- * @udata: not used by the InfiniPath verbs driver
+ * @srq_init_attr: the attributes of the SRQ
+ * @udata: data from libipathverbs when creating a user SRQ
  */
 struct ib_srq *ipath_create_srq(struct ib_pd *ibpd,
 				struct ib_srq_init_attr *srq_init_attr,
@@ -211,11 +211,11 @@ int ipath_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 		     struct ib_udata *udata)
 {
 	struct ipath_srq *srq = to_isrq(ibsrq);
+	struct ipath_rwq *wq;
 	int ret = 0;
 
 	if (attr_mask & IB_SRQ_MAX_WR) {
 		struct ipath_rwq *owq;
-		struct ipath_rwq *wq;
 		struct ipath_rwqe *p;
 		u32 sz, size, n, head, tail;
 
@@ -236,27 +236,20 @@ int ipath_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 			goto bail;
 		}
 
-		/*
-		 * Return the address of the RWQ as the offset to mmap.
-		 * See ipath_mmap() for details.
-		 */
+		/* Check that we can write the offset to mmap. */
 		if (udata && udata->inlen >= sizeof(__u64)) {
 			__u64 offset_addr;
-			__u64 offset = (__u64) wq;
+			__u64 offset = 0;
 
 			ret = ib_copy_from_udata(&offset_addr, udata,
 						 sizeof(offset_addr));
-			if (ret) {
-				vfree(wq);
-				goto bail;
-			}
+			if (ret)
+				goto bail_free;
 			udata->outbuf = (void __user *) offset_addr;
 			ret = ib_copy_to_udata(udata, &offset,
 					       sizeof(offset));
-			if (ret) {
-				vfree(wq);
-				goto bail;
-			}
+			if (ret)
+				goto bail_free;
 		}
 
 		spin_lock_irq(&srq->rq.lock);
@@ -277,10 +270,8 @@ int ipath_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 		else
 			n -= tail;
 		if (size <= n) {
-			spin_unlock_irq(&srq->rq.lock);
-			vfree(wq);
 			ret = -EINVAL;
-			goto bail;
+			goto bail_unlock;
 		}
 		n = 0;
 		p = wq->wq;
@@ -314,6 +305,18 @@ int ipath_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 			u32 s = sizeof(struct ipath_rwq) + size * sz;
 
 			ipath_update_mmap_info(dev, ip, s, wq);
+
+			/*
+			 * Return the offset to mmap.
+			 * See ipath_mmap() for details.
+			 */
+			if (udata && udata->inlen >= sizeof(__u64)) {
+				ret = ib_copy_to_udata(udata, &ip->offset,
+						       sizeof(ip->offset));
+				if (ret)
+					goto bail;
+			}
+
 			spin_lock_irq(&dev->pending_lock);
 			if (list_empty(&ip->pending_mmaps))
 				list_add(&ip->pending_mmaps,
@@ -328,7 +331,12 @@ int ipath_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 			srq->limit = attr->srq_limit;
 		spin_unlock_irq(&srq->rq.lock);
 	}
+	goto bail;
 
+bail_unlock:
+	spin_unlock_irq(&srq->rq.lock);
+bail_free:
+	vfree(wq);
 bail:
 	return ret;
 }
diff --git a/drivers/infiniband/hw/ipath/ipath_stats.c b/drivers/infiniband/hw/ipath/ipath_stats.c
index f0271415cd5b..d2725cd11bdc 100644
--- a/drivers/infiniband/hw/ipath/ipath_stats.c
+++ b/drivers/infiniband/hw/ipath/ipath_stats.c
@@ -133,15 +133,16 @@ bail:
 static void ipath_qcheck(struct ipath_devdata *dd)
 {
 	static u64 last_tot_hdrqfull;
+	struct ipath_portdata *pd = dd->ipath_pd[0];
 	size_t blen = 0;
 	char buf[128];
 
 	*buf = 0;
-	if (dd->ipath_pd[0]->port_hdrqfull != dd->ipath_p0_hdrqfull) {
+	if (pd->port_hdrqfull != dd->ipath_p0_hdrqfull) {
 		blen = snprintf(buf, sizeof buf, "port 0 hdrqfull %u",
-				dd->ipath_pd[0]->port_hdrqfull -
+				pd->port_hdrqfull -
 				dd->ipath_p0_hdrqfull);
-		dd->ipath_p0_hdrqfull = dd->ipath_pd[0]->port_hdrqfull;
+		dd->ipath_p0_hdrqfull = pd->port_hdrqfull;
 	}
 	if (ipath_stats.sps_etidfull != dd->ipath_last_tidfull) {
 		blen += snprintf(buf + blen, sizeof buf - blen,
@@ -173,7 +174,7 @@ static void ipath_qcheck(struct ipath_devdata *dd)
 	if (blen)
 		ipath_dbg("%s\n", buf);
 
-	if (dd->ipath_port0head != (u32)
+	if (pd->port_head != (u32)
 	    le64_to_cpu(*dd->ipath_hdrqtailptr)) {
 		if (dd->ipath_lastport0rcv_cnt ==
 		    ipath_stats.sps_port0pkts) {
@@ -181,7 +182,7 @@ static void ipath_qcheck(struct ipath_devdata *dd)
 				   "port0 hd=%llx tl=%x; port0pkts %llx\n",
 				   (unsigned long long)
 				   le64_to_cpu(*dd->ipath_hdrqtailptr),
-				   dd->ipath_port0head,
+				   pd->port_head,
 				   (unsigned long long)
 				   ipath_stats.sps_port0pkts);
 		}
@@ -237,7 +238,7 @@ static void ipath_chk_errormask(struct ipath_devdata *dd)
 void ipath_get_faststats(unsigned long opaque)
 {
 	struct ipath_devdata *dd = (struct ipath_devdata *) opaque;
-	u32 val;
+	int i;
 	static unsigned cnt;
 	unsigned long flags;
 	u64 traffic_wds;
@@ -321,12 +322,11 @@ void ipath_get_faststats(unsigned long opaque)
 
 	/* limit qfull messages to ~one per minute per port */
 	if ((++cnt & 0x10)) {
-		for (val = dd->ipath_cfgports - 1; ((int)val) >= 0;
-		     val--) {
-			if (dd->ipath_lastegrheads[val] != -1)
-				dd->ipath_lastegrheads[val] = -1;
-			if (dd->ipath_lastrcvhdrqtails[val] != -1)
-				dd->ipath_lastrcvhdrqtails[val] = -1;
+		for (i = (int) dd->ipath_cfgports; --i >= 0; ) {
+			struct ipath_portdata *pd = dd->ipath_pd[i];
+
+			if (pd && pd->port_lastrcvhdrqtail != -1)
+				pd->port_lastrcvhdrqtail = -1;
 		}
 	}
 
diff --git a/drivers/infiniband/hw/ipath/ipath_sysfs.c b/drivers/infiniband/hw/ipath/ipath_sysfs.c
index e1ad7cfc21fd..56dfc8a2344c 100644
--- a/drivers/infiniband/hw/ipath/ipath_sysfs.c
+++ b/drivers/infiniband/hw/ipath/ipath_sysfs.c
@@ -363,6 +363,60 @@ static ssize_t show_unit(struct device *dev,
 	return scnprintf(buf, PAGE_SIZE, "%u\n", dd->ipath_unit);
 }
 
+static ssize_t show_jint_max_packets(struct device *dev,
+				     struct device_attribute *attr,
+				     char *buf)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+
+	return scnprintf(buf, PAGE_SIZE, "%hu\n", dd->ipath_jint_max_packets);
+}
+
+static ssize_t store_jint_max_packets(struct device *dev,
+				      struct device_attribute *attr,
+				      const char *buf,
+				      size_t count)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	u16 v = 0;
+	int ret;
+
+	ret = ipath_parse_ushort(buf, &v);
+	if (ret < 0)
+		ipath_dev_err(dd, "invalid jint_max_packets.\n");
+	else
+		dd->ipath_f_config_jint(dd, dd->ipath_jint_idle_ticks, v);
+
+	return ret;
+}
+
+static ssize_t show_jint_idle_ticks(struct device *dev,
+				    struct device_attribute *attr,
+				    char *buf)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+
+	return scnprintf(buf, PAGE_SIZE, "%hu\n", dd->ipath_jint_idle_ticks);
+}
+
+static ssize_t store_jint_idle_ticks(struct device *dev,
+				     struct device_attribute *attr,
+				     const char *buf,
+				     size_t count)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	u16 v = 0;
+	int ret;
+
+	ret = ipath_parse_ushort(buf, &v);
+	if (ret < 0)
+		ipath_dev_err(dd, "invalid jint_idle_ticks.\n");
+	else
+		dd->ipath_f_config_jint(dd, v, dd->ipath_jint_max_packets);
+
+	return ret;
+}
+
 #define DEVICE_COUNTER(name, attr) \
 	static ssize_t show_counter_##name(struct device *dev, \
 					   struct device_attribute *attr, \
@@ -670,6 +724,257 @@ static ssize_t show_logged_errs(struct device *dev,
 	return count;
 }
 
+/*
+ * New sysfs entries to control various IB config. These all turn into
+ * accesses via ipath_f_get/set_ib_cfg.
+ *
+ * Get/Set heartbeat enable. Or of 1=enabled, 2=auto
+ */
+static ssize_t show_hrtbt_enb(struct device *dev,
+			 struct device_attribute *attr,
+			 char *buf)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret;
+
+	ret = dd->ipath_f_get_ib_cfg(dd, IPATH_IB_CFG_HRTBT);
+	if (ret >= 0)
+		ret = scnprintf(buf, PAGE_SIZE, "%d\n", ret);
+	return ret;
+}
+
+static ssize_t store_hrtbt_enb(struct device *dev,
+			  struct device_attribute *attr,
+			  const char *buf,
+			  size_t count)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret, r;
+	u16 val;
+
+	ret = ipath_parse_ushort(buf, &val);
+	if (ret >= 0 && val > 3)
+		ret = -EINVAL;
+	if (ret < 0) {
+		ipath_dev_err(dd, "attempt to set invalid Heartbeat enable\n");
+		goto bail;
+	}
+
+	/*
+	 * Set the "intentional" heartbeat enable per either of
+	 * "Enable" and "Auto", as these are normally set together.
+	 * This bit is consulted when leaving loopback mode,
+	 * because entering loopback mode overrides it and automatically
+	 * disables heartbeat.
+	 */
+	r = dd->ipath_f_set_ib_cfg(dd, IPATH_IB_CFG_HRTBT, val);
+	if (r < 0)
+		ret = r;
+	else if (val == IPATH_IB_HRTBT_OFF)
+		dd->ipath_flags |= IPATH_NO_HRTBT;
+	else
+		dd->ipath_flags &= ~IPATH_NO_HRTBT;
+
+bail:
+	return ret;
+}
+
+/*
+ * Get/Set Link-widths enabled. Or of 1=1x, 2=4x (this is human/IB centric,
+ * _not_ the particular encoding of any given chip)
+ */
+static ssize_t show_lwid_enb(struct device *dev,
+			 struct device_attribute *attr,
+			 char *buf)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret;
+
+	ret = dd->ipath_f_get_ib_cfg(dd, IPATH_IB_CFG_LWID_ENB);
+	if (ret >= 0)
+		ret = scnprintf(buf, PAGE_SIZE, "%d\n", ret);
+	return ret;
+}
+
+static ssize_t store_lwid_enb(struct device *dev,
+			  struct device_attribute *attr,
+			  const char *buf,
+			  size_t count)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret, r;
+	u16 val;
+
+	ret = ipath_parse_ushort(buf, &val);
+	if (ret >= 0 && (val == 0 || val > 3))
+		ret = -EINVAL;
+	if (ret < 0) {
+		ipath_dev_err(dd,
+			"attempt to set invalid Link Width (enable)\n");
+		goto bail;
+	}
+
+	r = dd->ipath_f_set_ib_cfg(dd, IPATH_IB_CFG_LWID_ENB, val);
+	if (r < 0)
+		ret = r;
+
+bail:
+	return ret;
+}
+
+/* Get current link width */
+static ssize_t show_lwid(struct device *dev,
+			 struct device_attribute *attr,
+			 char *buf)
+
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret;
+
+	ret = dd->ipath_f_get_ib_cfg(dd, IPATH_IB_CFG_LWID);
+	if (ret >= 0)
+		ret = scnprintf(buf, PAGE_SIZE, "%d\n", ret);
+	return ret;
+}
+
+/*
+ * Get/Set Link-speeds enabled. Or of 1=SDR 2=DDR.
+ */
+static ssize_t show_spd_enb(struct device *dev,
+			 struct device_attribute *attr,
+			 char *buf)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret;
+
+	ret = dd->ipath_f_get_ib_cfg(dd, IPATH_IB_CFG_SPD_ENB);
+	if (ret >= 0)
+		ret = scnprintf(buf, PAGE_SIZE, "%d\n", ret);
+	return ret;
+}
+
+static ssize_t store_spd_enb(struct device *dev,
+			  struct device_attribute *attr,
+			  const char *buf,
+			  size_t count)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret, r;
+	u16 val;
+
+	ret = ipath_parse_ushort(buf, &val);
+	if (ret >= 0 && (val == 0 || val > (IPATH_IB_SDR | IPATH_IB_DDR)))
+		ret = -EINVAL;
+	if (ret < 0) {
+		ipath_dev_err(dd,
+			"attempt to set invalid Link Speed (enable)\n");
+		goto bail;
+	}
+
+	r = dd->ipath_f_set_ib_cfg(dd, IPATH_IB_CFG_SPD_ENB, val);
+	if (r < 0)
+		ret = r;
+
+bail:
+	return ret;
+}
+
+/* Get current link speed */
+static ssize_t show_spd(struct device *dev,
+			 struct device_attribute *attr,
+			 char *buf)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret;
+
+	ret = dd->ipath_f_get_ib_cfg(dd, IPATH_IB_CFG_SPD);
+	if (ret >= 0)
+		ret = scnprintf(buf, PAGE_SIZE, "%d\n", ret);
+	return ret;
+}
+
+/*
+ * Get/Set RX polarity-invert enable. 0=no, 1=yes.
+ */
+static ssize_t show_rx_polinv_enb(struct device *dev,
+			 struct device_attribute *attr,
+			 char *buf)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret;
+
+	ret = dd->ipath_f_get_ib_cfg(dd, IPATH_IB_CFG_RXPOL_ENB);
+	if (ret >= 0)
+		ret = scnprintf(buf, PAGE_SIZE, "%d\n", ret);
+	return ret;
+}
+
+static ssize_t store_rx_polinv_enb(struct device *dev,
+			  struct device_attribute *attr,
+			  const char *buf,
+			  size_t count)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret, r;
+	u16 val;
+
+	ret = ipath_parse_ushort(buf, &val);
+	if (ret < 0 || val > 1)
+		goto invalid;
+
+	r = dd->ipath_f_set_ib_cfg(dd, IPATH_IB_CFG_RXPOL_ENB, val);
+	if (r < 0) {
+		ret = r;
+		goto bail;
+	}
+
+	goto bail;
+invalid:
+	ipath_dev_err(dd, "attempt to set invalid Rx Polarity (enable)\n");
+bail:
+	return ret;
+}
+/*
+ * Get/Set RX lane-reversal enable. 0=no, 1=yes.
+ */
+static ssize_t show_lanerev_enb(struct device *dev,
+			 struct device_attribute *attr,
+			 char *buf)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret;
+
+	ret = dd->ipath_f_get_ib_cfg(dd, IPATH_IB_CFG_LREV_ENB);
+	if (ret >= 0)
+		ret = scnprintf(buf, PAGE_SIZE, "%d\n", ret);
+	return ret;
+}
+
+static ssize_t store_lanerev_enb(struct device *dev,
+			  struct device_attribute *attr,
+			  const char *buf,
+			  size_t count)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret, r;
+	u16 val;
+
+	ret = ipath_parse_ushort(buf, &val);
+	if (ret >= 0 && val > 1) {
+		ret = -EINVAL;
+		ipath_dev_err(dd,
+			"attempt to set invalid Lane reversal (enable)\n");
+		goto bail;
+	}
+
+	r = dd->ipath_f_set_ib_cfg(dd, IPATH_IB_CFG_LREV_ENB, val);
+	if (r < 0)
+		ret = r;
+
+bail:
+	return ret;
+}
+
 static DRIVER_ATTR(num_units, S_IRUGO, show_num_units, NULL);
 static DRIVER_ATTR(version, S_IRUGO, show_version, NULL);
 
@@ -683,6 +988,11 @@ static struct attribute_group driver_attr_group = {
 	.attrs = driver_attributes
 };
 
+struct attribute_group *ipath_driver_attr_groups[] = {
+	&driver_attr_group,
+	NULL,
+};
+
 static DEVICE_ATTR(guid, S_IWUSR | S_IRUGO, show_guid, store_guid);
 static DEVICE_ATTR(lmc, S_IWUSR | S_IRUGO, show_lmc, store_lmc);
 static DEVICE_ATTR(lid, S_IWUSR | S_IRUGO, show_lid, store_lid);
@@ -701,6 +1011,10 @@ static DEVICE_ATTR(unit, S_IRUGO, show_unit, NULL);
 static DEVICE_ATTR(rx_pol_inv, S_IWUSR, NULL, store_rx_pol_inv);
 static DEVICE_ATTR(led_override, S_IWUSR, NULL, store_led_override);
 static DEVICE_ATTR(logged_errors, S_IRUGO, show_logged_errs, NULL);
+static DEVICE_ATTR(jint_max_packets, S_IWUSR | S_IRUGO,
+		   show_jint_max_packets, store_jint_max_packets);
+static DEVICE_ATTR(jint_idle_ticks, S_IWUSR | S_IRUGO,
+		   show_jint_idle_ticks, store_jint_idle_ticks);
 
 static struct attribute *dev_attributes[] = {
 	&dev_attr_guid.attr,
@@ -727,6 +1041,34 @@ static struct attribute_group dev_attr_group = {
 	.attrs = dev_attributes
 };
 
+static DEVICE_ATTR(hrtbt_enable, S_IWUSR | S_IRUGO, show_hrtbt_enb,
+		   store_hrtbt_enb);
+static DEVICE_ATTR(link_width_enable, S_IWUSR | S_IRUGO, show_lwid_enb,
+		   store_lwid_enb);
+static DEVICE_ATTR(link_width, S_IRUGO, show_lwid, NULL);
+static DEVICE_ATTR(link_speed_enable, S_IWUSR | S_IRUGO, show_spd_enb,
+		   store_spd_enb);
+static DEVICE_ATTR(link_speed, S_IRUGO, show_spd, NULL);
+static DEVICE_ATTR(rx_pol_inv_enable, S_IWUSR | S_IRUGO, show_rx_polinv_enb,
+		   store_rx_polinv_enb);
+static DEVICE_ATTR(rx_lane_rev_enable, S_IWUSR | S_IRUGO, show_lanerev_enb,
+		   store_lanerev_enb);
+
+static struct attribute *dev_ibcfg_attributes[] = {
+	&dev_attr_hrtbt_enable.attr,
+	&dev_attr_link_width_enable.attr,
+	&dev_attr_link_width.attr,
+	&dev_attr_link_speed_enable.attr,
+	&dev_attr_link_speed.attr,
+	&dev_attr_rx_pol_inv_enable.attr,
+	&dev_attr_rx_lane_rev_enable.attr,
+	NULL
+};
+
+static struct attribute_group dev_ibcfg_attr_group = {
+	.attrs = dev_ibcfg_attributes
+};
+
 /**
  * ipath_expose_reset - create a device reset file
  * @dev: the device structure
@@ -753,24 +1095,9 @@ int ipath_expose_reset(struct device *dev)
 	return ret;
 }
 
-int ipath_driver_create_group(struct device_driver *drv)
-{
-	int ret;
-
-	ret = sysfs_create_group(&drv->kobj, &driver_attr_group);
-
-	return ret;
-}
-
-void ipath_driver_remove_group(struct device_driver *drv)
-{
-	sysfs_remove_group(&drv->kobj, &driver_attr_group);
-}
-
 int ipath_device_create_group(struct device *dev, struct ipath_devdata *dd)
 {
 	int ret;
-	char unit[5];
 
 	ret = sysfs_create_group(&dev->kobj, &dev_attr_group);
 	if (ret)
@@ -780,11 +1107,26 @@ int ipath_device_create_group(struct device *dev, struct ipath_devdata *dd)
 	if (ret)
 		goto bail_attrs;
 
-	snprintf(unit, sizeof(unit), "%02d", dd->ipath_unit);
-	ret = sysfs_create_link(&dev->driver->kobj, &dev->kobj, unit);
-	if (ret == 0)
-		goto bail;
+	if (dd->ipath_flags & IPATH_HAS_MULT_IB_SPEED) {
+		ret = device_create_file(dev, &dev_attr_jint_idle_ticks);
+		if (ret)
+			goto bail_counter;
+		ret = device_create_file(dev, &dev_attr_jint_max_packets);
+		if (ret)
+			goto bail_idle;
 
+		ret = sysfs_create_group(&dev->kobj, &dev_ibcfg_attr_group);
+		if (ret)
+			goto bail_max;
+	}
+
+	return 0;
+
+bail_max:
+	device_remove_file(dev, &dev_attr_jint_max_packets);
+bail_idle:
+	device_remove_file(dev, &dev_attr_jint_idle_ticks);
+bail_counter:
 	sysfs_remove_group(&dev->kobj, &dev_counter_attr_group);
 bail_attrs:
 	sysfs_remove_group(&dev->kobj, &dev_attr_group);
@@ -794,12 +1136,14 @@ bail:
 
 void ipath_device_remove_group(struct device *dev, struct ipath_devdata *dd)
 {
-	char unit[5];
+	sysfs_remove_group(&dev->kobj, &dev_counter_attr_group);
 
-	snprintf(unit, sizeof(unit), "%02d", dd->ipath_unit);
-	sysfs_remove_link(&dev->driver->kobj, unit);
+	if (dd->ipath_flags & IPATH_HAS_MULT_IB_SPEED) {
+		sysfs_remove_group(&dev->kobj, &dev_ibcfg_attr_group);
+		device_remove_file(dev, &dev_attr_jint_idle_ticks);
+		device_remove_file(dev, &dev_attr_jint_max_packets);
+	}
 
-	sysfs_remove_group(&dev->kobj, &dev_counter_attr_group);
 	sysfs_remove_group(&dev->kobj, &dev_attr_group);
 
 	device_remove_file(dev, &dev_attr_reset);
diff --git a/drivers/infiniband/hw/ipath/ipath_ud.c b/drivers/infiniband/hw/ipath/ipath_ud.c
index 16a2a938b520..de67eed08ed0 100644
--- a/drivers/infiniband/hw/ipath/ipath_ud.c
+++ b/drivers/infiniband/hw/ipath/ipath_ud.c
@@ -301,8 +301,6 @@ int ipath_make_ud_req(struct ipath_qp *qp)
 
 	/* header size in 32-bit words LRH+BTH+DETH = (8+12+8)/4. */
 	qp->s_hdrwords = 7;
-	if (wqe->wr.opcode == IB_WR_SEND_WITH_IMM)
-		qp->s_hdrwords++;
 	qp->s_cur_size = wqe->length;
 	qp->s_cur_sge = &qp->s_sge;
 	qp->s_wqe = wqe;
@@ -327,6 +325,7 @@ int ipath_make_ud_req(struct ipath_qp *qp)
 		ohdr = &qp->s_hdr.u.oth;
 	}
 	if (wqe->wr.opcode == IB_WR_SEND_WITH_IMM) {
+		qp->s_hdrwords++;
 		ohdr->u.ud.imm_data = wqe->wr.imm_data;
 		bth0 = IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE << 24;
 	} else
@@ -455,6 +454,28 @@ void ipath_ud_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 		}
 	}
 
+	/*
+	 * The opcode is in the low byte when its in network order
+	 * (top byte when in host order).
+	 */
+	opcode = be32_to_cpu(ohdr->bth[0]) >> 24;
+	if (qp->ibqp.qp_num > 1 &&
+	    opcode == IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE) {
+		if (header_in_data) {
+			wc.imm_data = *(__be32 *) data;
+			data += sizeof(__be32);
+		} else
+			wc.imm_data = ohdr->u.ud.imm_data;
+		wc.wc_flags = IB_WC_WITH_IMM;
+		hdrsize += sizeof(u32);
+	} else if (opcode == IB_OPCODE_UD_SEND_ONLY) {
+		wc.imm_data = 0;
+		wc.wc_flags = 0;
+	} else {
+		dev->n_pkt_drops++;
+		goto bail;
+	}
+
 	/* Get the number of bytes the message was padded by. */
 	pad = (be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
 	if (unlikely(tlen < (hdrsize + pad + 4))) {
@@ -482,28 +503,6 @@ void ipath_ud_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	wc.byte_len = tlen + sizeof(struct ib_grh);
 
 	/*
-	 * The opcode is in the low byte when its in network order
-	 * (top byte when in host order).
-	 */
-	opcode = be32_to_cpu(ohdr->bth[0]) >> 24;
-	if (qp->ibqp.qp_num > 1 &&
-	    opcode == IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE) {
-		if (header_in_data) {
-			wc.imm_data = *(__be32 *) data;
-			data += sizeof(__be32);
-		} else
-			wc.imm_data = ohdr->u.ud.imm_data;
-		wc.wc_flags = IB_WC_WITH_IMM;
-		hdrsize += sizeof(u32);
-	} else if (opcode == IB_OPCODE_UD_SEND_ONLY) {
-		wc.imm_data = 0;
-		wc.wc_flags = 0;
-	} else {
-		dev->n_pkt_drops++;
-		goto bail;
-	}
-
-	/*
 	 * Get the next work request entry to find where to put the data.
 	 */
 	if (qp->r_reuse_sge)
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
index 74f77e7c2c1b..32d8f882e56c 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c
@@ -302,8 +302,10 @@ static int ipath_post_one_send(struct ipath_qp *qp, struct ib_send_wr *wr)
 	next = qp->s_head + 1;
 	if (next >= qp->s_size)
 		next = 0;
-	if (next == qp->s_last)
-		goto bail_inval;
+	if (next == qp->s_last) {
+		ret = -ENOMEM;
+		goto bail;
+	}
 
 	wqe = get_swqe_ptr(qp, qp->s_head);
 	wqe->wr = *wr;
@@ -404,7 +406,7 @@ static int ipath_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 
 		if ((unsigned) wr->num_sge > qp->r_rq.max_sge) {
 			*bad_wr = wr;
-			ret = -ENOMEM;
+			ret = -EINVAL;
 			goto bail;
 		}
 
@@ -941,7 +943,7 @@ bail:
  * ipath_verbs_send - send a packet
  * @qp: the QP to send on
  * @hdr: the packet header
- * @hdrwords: the number of words in the header
+ * @hdrwords: the number of 32-bit words in the header
  * @ss: the SGE to send
  * @len: the length of the packet in bytes
  */
@@ -953,7 +955,10 @@ int ipath_verbs_send(struct ipath_qp *qp, struct ipath_ib_header *hdr,
 	int ret;
 	u32 dwords = (len + 3) >> 2;
 
-	/* +1 is for the qword padding of pbc */
+	/*
+	 * Calculate the send buffer trigger address.
+	 * The +1 counts for the pbc control dword following the pbc length.
+	 */
 	plen = hdrwords + dwords + 1;
 
 	/* Drop non-VL15 packets if we are not in the active state */
@@ -1128,20 +1133,34 @@ static int ipath_query_device(struct ib_device *ibdev,
 	return 0;
 }
 
-const u8 ipath_cvt_physportstate[16] = {
-	[INFINIPATH_IBCS_LT_STATE_DISABLED] = 3,
-	[INFINIPATH_IBCS_LT_STATE_LINKUP] = 5,
-	[INFINIPATH_IBCS_LT_STATE_POLLACTIVE] = 2,
-	[INFINIPATH_IBCS_LT_STATE_POLLQUIET] = 2,
-	[INFINIPATH_IBCS_LT_STATE_SLEEPDELAY] = 1,
-	[INFINIPATH_IBCS_LT_STATE_SLEEPQUIET] = 1,
-	[INFINIPATH_IBCS_LT_STATE_CFGDEBOUNCE] = 4,
-	[INFINIPATH_IBCS_LT_STATE_CFGRCVFCFG] = 4,
-	[INFINIPATH_IBCS_LT_STATE_CFGWAITRMT] = 4,
-	[INFINIPATH_IBCS_LT_STATE_CFGIDLE] = 4,
-	[INFINIPATH_IBCS_LT_STATE_RECOVERRETRAIN] = 6,
-	[INFINIPATH_IBCS_LT_STATE_RECOVERWAITRMT] = 6,
-	[INFINIPATH_IBCS_LT_STATE_RECOVERIDLE] = 6,
+const u8 ipath_cvt_physportstate[32] = {
+	[INFINIPATH_IBCS_LT_STATE_DISABLED] = IB_PHYSPORTSTATE_DISABLED,
+	[INFINIPATH_IBCS_LT_STATE_LINKUP] = IB_PHYSPORTSTATE_LINKUP,
+	[INFINIPATH_IBCS_LT_STATE_POLLACTIVE] = IB_PHYSPORTSTATE_POLL,
+	[INFINIPATH_IBCS_LT_STATE_POLLQUIET] = IB_PHYSPORTSTATE_POLL,
+	[INFINIPATH_IBCS_LT_STATE_SLEEPDELAY] = IB_PHYSPORTSTATE_SLEEP,
+	[INFINIPATH_IBCS_LT_STATE_SLEEPQUIET] = IB_PHYSPORTSTATE_SLEEP,
+	[INFINIPATH_IBCS_LT_STATE_CFGDEBOUNCE] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[INFINIPATH_IBCS_LT_STATE_CFGRCVFCFG] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[INFINIPATH_IBCS_LT_STATE_CFGWAITRMT] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[INFINIPATH_IBCS_LT_STATE_CFGIDLE] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[INFINIPATH_IBCS_LT_STATE_RECOVERRETRAIN] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[INFINIPATH_IBCS_LT_STATE_RECOVERWAITRMT] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[INFINIPATH_IBCS_LT_STATE_RECOVERIDLE] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[0x10] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x11] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x12] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x13] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x14] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x15] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x16] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x17] = IB_PHYSPORTSTATE_CFG_TRAIN
 };
 
 u32 ipath_get_cr_errpkey(struct ipath_devdata *dd)
@@ -1166,8 +1185,9 @@ static int ipath_query_port(struct ib_device *ibdev,
 	ibcstat = dd->ipath_lastibcstat;
 	props->state = ((ibcstat >> 4) & 0x3) + 1;
 	/* See phys_state_show() */
-	props->phys_state = ipath_cvt_physportstate[
-		dd->ipath_lastibcstat & 0xf];
+	props->phys_state = /* MEA: assumes shift == 0 */
+		ipath_cvt_physportstate[dd->ipath_lastibcstat &
+		dd->ibcs_lts_mask];
 	props->port_cap_flags = dev->port_cap_flags;
 	props->gid_tbl_len = 1;
 	props->max_msg_sz = 0x80000000;
@@ -1639,6 +1659,7 @@ int ipath_register_ib_device(struct ipath_devdata *dd)
 		cntrs.local_link_integrity_errors;
 	idev->z_excessive_buffer_overrun_errors =
 		cntrs.excessive_buffer_overrun_errors;
+	idev->z_vl15_dropped = cntrs.vl15_dropped;
 
 	/*
 	 * The system image GUID is supposed to be the same for all
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
index 6ccb54f104a3..3d59736b49b2 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h
@@ -554,6 +554,7 @@ struct ipath_ibdev {
 	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 z_vl15_dropped;			/* starting count for PMA */
 	u32 n_rc_resends;
 	u32 n_rc_acks;
 	u32 n_rc_qacks;
@@ -598,6 +599,7 @@ struct ipath_verbs_counters {
 	u64 port_rcv_packets;
 	u32 local_link_integrity_errors;
 	u32 excessive_buffer_overrun_errors;
+	u32 vl15_dropped;
 };
 
 static inline struct ipath_mr *to_imr(struct ib_mr *ibmr)
@@ -830,7 +832,17 @@ unsigned ipath_get_pkey(struct ipath_devdata *, unsigned);
 
 extern const enum ib_wc_opcode ib_ipath_wc_opcode[];
 
+/*
+ * Below converts HCA-specific LinkTrainingState to IB PhysPortState
+ * values.
+ */
 extern const u8 ipath_cvt_physportstate[];
+#define IB_PHYSPORTSTATE_SLEEP 1
+#define IB_PHYSPORTSTATE_POLL 2
+#define IB_PHYSPORTSTATE_DISABLED 3
+#define IB_PHYSPORTSTATE_CFG_TRAIN 4
+#define IB_PHYSPORTSTATE_LINKUP 5
+#define IB_PHYSPORTSTATE_LINK_ERR_RECOVER 6
 
 extern const int ib_ipath_state_ops[];
 
diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c
index 8bf44daf45ec..7950aa6e8184 100644
--- a/drivers/infiniband/hw/mlx4/cq.c
+++ b/drivers/infiniband/hw/mlx4/cq.c
@@ -313,6 +313,7 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq,
 	struct mlx4_ib_srq *srq;
 	int is_send;
 	int is_error;
+	u32 g_mlpath_rqpn;
 	u16 wqe_ctr;
 
 	cqe = next_cqe_sw(cq);
@@ -426,11 +427,11 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq,
 
 		wc->slid	   = be16_to_cpu(cqe->rlid);
 		wc->sl		   = cqe->sl >> 4;
-		wc->src_qp	   = be32_to_cpu(cqe->g_mlpath_rqpn) & 0xffffff;
-		wc->dlid_path_bits = (be32_to_cpu(cqe->g_mlpath_rqpn) >> 24) & 0x7f;
-		wc->wc_flags      |= be32_to_cpu(cqe->g_mlpath_rqpn) & 0x80000000 ?
-			IB_WC_GRH : 0;
-		wc->pkey_index     = be32_to_cpu(cqe->immed_rss_invalid) >> 16;
+		g_mlpath_rqpn	   = be32_to_cpu(cqe->g_mlpath_rqpn);
+		wc->src_qp	   = g_mlpath_rqpn & 0xffffff;
+		wc->dlid_path_bits = (g_mlpath_rqpn >> 24) & 0x7f;
+		wc->wc_flags	  |= g_mlpath_rqpn & 0x80000000 ? IB_WC_GRH : 0;
+		wc->pkey_index     = be32_to_cpu(cqe->immed_rss_invalid) & 0x7f;
 	}
 
 	return 0;
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 6b3322486b5e..8cba9c532e64 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -1282,7 +1282,7 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 	int size;
 	int i;
 
-	spin_lock_irqsave(&qp->rq.lock, flags);
+	spin_lock_irqsave(&qp->sq.lock, flags);
 
 	ind = qp->sq.head;
 
@@ -1448,7 +1448,7 @@ out:
 			       (qp->sq.wqe_cnt - 1));
 	}
 
-	spin_unlock_irqrestore(&qp->rq.lock, flags);
+	spin_unlock_irqrestore(&qp->sq.lock, flags);
 
 	return err;
 }
diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h
index 15aa32eb78b6..7bbdd1f4e6c7 100644
--- a/drivers/infiniband/hw/mthca/mthca_dev.h
+++ b/drivers/infiniband/hw/mthca/mthca_dev.h
@@ -60,13 +60,12 @@
 enum {
 	MTHCA_FLAG_DDR_HIDDEN = 1 << 1,
 	MTHCA_FLAG_SRQ        = 1 << 2,
-	MTHCA_FLAG_MSI        = 1 << 3,
-	MTHCA_FLAG_MSI_X      = 1 << 4,
-	MTHCA_FLAG_NO_LAM     = 1 << 5,
-	MTHCA_FLAG_FMR        = 1 << 6,
-	MTHCA_FLAG_MEMFREE    = 1 << 7,
-	MTHCA_FLAG_PCIE       = 1 << 8,
-	MTHCA_FLAG_SINAI_OPT  = 1 << 9
+	MTHCA_FLAG_MSI_X      = 1 << 3,
+	MTHCA_FLAG_NO_LAM     = 1 << 4,
+	MTHCA_FLAG_FMR        = 1 << 5,
+	MTHCA_FLAG_MEMFREE    = 1 << 6,
+	MTHCA_FLAG_PCIE       = 1 << 7,
+	MTHCA_FLAG_SINAI_OPT  = 1 << 8
 };
 
 enum {
diff --git a/drivers/infiniband/hw/mthca/mthca_eq.c b/drivers/infiniband/hw/mthca/mthca_eq.c
index b29de51b7f35..b60eb5df96e8 100644
--- a/drivers/infiniband/hw/mthca/mthca_eq.c
+++ b/drivers/infiniband/hw/mthca/mthca_eq.c
@@ -827,8 +827,7 @@ int mthca_init_eq_table(struct mthca_dev *dev)
 	if (err)
 		goto err_out_free;
 
-	if (dev->mthca_flags & MTHCA_FLAG_MSI ||
-	    dev->mthca_flags & MTHCA_FLAG_MSI_X) {
+	if (dev->mthca_flags & MTHCA_FLAG_MSI_X) {
 		dev->eq_table.clr_mask = 0;
 	} else {
 		dev->eq_table.clr_mask =
@@ -839,8 +838,7 @@ int mthca_init_eq_table(struct mthca_dev *dev)
 
 	dev->eq_table.arm_mask = 0;
 
-	intr = (dev->mthca_flags & MTHCA_FLAG_MSI) ?
-		128 : dev->eq_table.inta_pin;
+	intr = dev->eq_table.inta_pin;
 
 	err = mthca_create_eq(dev, dev->limits.num_cqs + MTHCA_NUM_SPARE_EQE,
 			      (dev->mthca_flags & MTHCA_FLAG_MSI_X) ? 128 : intr,
diff --git a/drivers/infiniband/hw/mthca/mthca_main.c b/drivers/infiniband/hw/mthca/mthca_main.c
index 60de6f93869e..5cf8250d4e16 100644
--- a/drivers/infiniband/hw/mthca/mthca_main.c
+++ b/drivers/infiniband/hw/mthca/mthca_main.c
@@ -65,14 +65,9 @@ static int msi_x = 1;
 module_param(msi_x, int, 0444);
 MODULE_PARM_DESC(msi_x, "attempt to use MSI-X if nonzero");
 
-static int msi = 0;
-module_param(msi, int, 0444);
-MODULE_PARM_DESC(msi, "attempt to use MSI if nonzero (deprecated, use MSI-X instead)");
-
 #else /* CONFIG_PCI_MSI */
 
 #define msi_x (0)
-#define msi   (0)
 
 #endif /* CONFIG_PCI_MSI */
 
@@ -816,13 +811,11 @@ static int mthca_setup_hca(struct mthca_dev *dev)
 
 	err = mthca_NOP(dev, &status);
 	if (err || status) {
-		if (dev->mthca_flags & (MTHCA_FLAG_MSI | MTHCA_FLAG_MSI_X)) {
+		if (dev->mthca_flags & MTHCA_FLAG_MSI_X) {
 			mthca_warn(dev, "NOP command failed to generate interrupt "
 				   "(IRQ %d).\n",
-				   dev->mthca_flags & MTHCA_FLAG_MSI_X ?
-				   dev->eq_table.eq[MTHCA_EQ_CMD].msi_x_vector :
-				   dev->pdev->irq);
-			mthca_warn(dev, "Trying again with MSI/MSI-X disabled.\n");
+				   dev->eq_table.eq[MTHCA_EQ_CMD].msi_x_vector);
+			mthca_warn(dev, "Trying again with MSI-X disabled.\n");
 		} else {
 			mthca_err(dev, "NOP command failed to generate interrupt "
 				  "(IRQ %d), aborting.\n",
@@ -1005,7 +998,7 @@ static struct {
 			   .flags     = 0 },
 	[ARBEL_COMPAT] = { .latest_fw = MTHCA_FW_VER(4, 8, 200),
 			   .flags     = MTHCA_FLAG_PCIE },
-	[ARBEL_NATIVE] = { .latest_fw = MTHCA_FW_VER(5, 2, 0),
+	[ARBEL_NATIVE] = { .latest_fw = MTHCA_FW_VER(5, 3, 0),
 			   .flags     = MTHCA_FLAG_MEMFREE |
 					MTHCA_FLAG_PCIE },
 	[SINAI]        = { .latest_fw = MTHCA_FW_VER(1, 2, 0),
@@ -1128,29 +1121,12 @@ static int __mthca_init_one(struct pci_dev *pdev, int hca_type)
 
 	if (msi_x && !mthca_enable_msi_x(mdev))
 		mdev->mthca_flags |= MTHCA_FLAG_MSI_X;
-	else if (msi) {
-		static int warned;
-
-		if (!warned) {
-			printk(KERN_WARNING PFX "WARNING: MSI support will be "
-			       "removed from the ib_mthca driver in January 2008.\n");
-			printk(KERN_WARNING "    If you are using MSI and cannot "
-			       "switch to MSI-X, please tell "
-			       "<general@lists.openfabrics.org>.\n");
-			++warned;
-		}
-
-		if (!pci_enable_msi(pdev))
-			mdev->mthca_flags |= MTHCA_FLAG_MSI;
-	}
 
 	err = mthca_setup_hca(mdev);
-	if (err == -EBUSY && (mdev->mthca_flags & (MTHCA_FLAG_MSI | MTHCA_FLAG_MSI_X))) {
+	if (err == -EBUSY && (mdev->mthca_flags & MTHCA_FLAG_MSI_X)) {
 		if (mdev->mthca_flags & MTHCA_FLAG_MSI_X)
 			pci_disable_msix(pdev);
-		if (mdev->mthca_flags & MTHCA_FLAG_MSI)
-			pci_disable_msi(pdev);
-		mdev->mthca_flags &= ~(MTHCA_FLAG_MSI_X | MTHCA_FLAG_MSI);
+		mdev->mthca_flags &= ~MTHCA_FLAG_MSI_X;
 
 		err = mthca_setup_hca(mdev);
 	}
@@ -1192,8 +1168,6 @@ err_cleanup:
 err_close:
 	if (mdev->mthca_flags & MTHCA_FLAG_MSI_X)
 		pci_disable_msix(pdev);
-	if (mdev->mthca_flags & MTHCA_FLAG_MSI)
-		pci_disable_msi(pdev);
 
 	mthca_close_hca(mdev);
 
@@ -1246,8 +1220,6 @@ static void __mthca_remove_one(struct pci_dev *pdev)
 
 		if (mdev->mthca_flags & MTHCA_FLAG_MSI_X)
 			pci_disable_msix(pdev);
-		if (mdev->mthca_flags & MTHCA_FLAG_MSI)
-			pci_disable_msi(pdev);
 
 		ib_dealloc_device(&mdev->ib_dev);
 		mthca_release_regions(pdev, mdev->mthca_flags &