From 35ff032e65ab5cc03bbba46cefece7376c7c562f Mon Sep 17 00:00:00 2001
From: Ralph Campbell <ralph.campbell@qlogic.com>
Date: Fri, 27 Apr 2007 11:11:11 -0700
Subject: IB/ipath: Don't call spin_lock_irq() from interrupt context

This patch fixes the problem reported by Bernd Schubert <bs@q-leap.de>
with kernel debug options enabled:

    BUG: at kernel/lockdep.c:1860 trace_hardirqs_on()

This was caused by using spin_lock_irq()/spin_unlock_irq() from
interrupt context.  Fix all the places that might be called from
interrupts to use spin_lock_irqsave()/spin_unlock_irqrestore().

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c
index b4b88d0..e3e5332 100644
--- a/drivers/infiniband/hw/ipath/ipath_rc.c
+++ b/drivers/infiniband/hw/ipath/ipath_rc.c
@@ -587,6 +587,7 @@ static void send_rc_ack(struct ipath_qp *qp)
 	u32 hwords;
 	struct ipath_ib_header hdr;
 	struct ipath_other_headers *ohdr;
+	unsigned long flags;
 
 	/* Don't send ACK or NAK if a RDMA read or atomic is pending. */
 	if (qp->r_head_ack_queue != qp->s_tail_ack_queue)
@@ -640,11 +641,11 @@ static void send_rc_ack(struct ipath_qp *qp)
 	dev->n_rc_qacks++;
 
 queue_ack:
-	spin_lock_irq(&qp->s_lock);
+	spin_lock_irqsave(&qp->s_lock, flags);
 	qp->s_flags |= IPATH_S_ACK_PENDING;
 	qp->s_nak_state = qp->r_nak_state;
 	qp->s_ack_psn = qp->r_ack_psn;
-	spin_unlock_irq(&qp->s_lock);
+	spin_unlock_irqrestore(&qp->s_lock, flags);
 
 	/* Call ipath_do_rc_send() in another thread. */
 	tasklet_hi_schedule(&qp->s_task);
@@ -1294,6 +1295,7 @@ static inline int ipath_rc_rcv_error(struct ipath_ibdev *dev,
 	struct ipath_ack_entry *e;
 	u8 i, prev;
 	int old_req;
+	unsigned long flags;
 
 	if (diff > 0) {
 		/*
@@ -1327,7 +1329,7 @@ static inline int ipath_rc_rcv_error(struct ipath_ibdev *dev,
 	psn &= IPATH_PSN_MASK;
 	e = NULL;
 	old_req = 1;
-	spin_lock_irq(&qp->s_lock);
+	spin_lock_irqsave(&qp->s_lock, flags);
 	for (i = qp->r_head_ack_queue; ; i = prev) {
 		if (i == qp->s_tail_ack_queue)
 			old_req = 0;
@@ -1425,7 +1427,7 @@ static inline int ipath_rc_rcv_error(struct ipath_ibdev *dev,
 		 * after all the previous RDMA reads and atomics.
 		 */
 		if (i == qp->r_head_ack_queue) {
-			spin_unlock_irq(&qp->s_lock);
+			spin_unlock_irqrestore(&qp->s_lock, flags);
 			qp->r_nak_state = 0;
 			qp->r_ack_psn = qp->r_psn - 1;
 			goto send_ack;
@@ -1443,7 +1445,7 @@ static inline int ipath_rc_rcv_error(struct ipath_ibdev *dev,
 	tasklet_hi_schedule(&qp->s_task);
 
 unlock_done:
-	spin_unlock_irq(&qp->s_lock);
+	spin_unlock_irqrestore(&qp->s_lock, flags);
 done:
 	return 1;
 
@@ -1453,10 +1455,12 @@ send_ack:
 
 static void ipath_rc_error(struct ipath_qp *qp, enum ib_wc_status err)
 {
-	spin_lock_irq(&qp->s_lock);
+	unsigned long flags;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
 	qp->state = IB_QPS_ERR;
 	ipath_error_qp(qp, err);
-	spin_unlock_irq(&qp->s_lock);
+	spin_unlock_irqrestore(&qp->s_lock, flags);
 }
 
 /**
-- 
cgit v0.10.2


From c3af664adbe06803931dbc7a3c8588982d72fac1 Mon Sep 17 00:00:00 2001
From: Ralph Campbell <ralphc@pathscale.com>
Date: Fri, 27 Apr 2007 11:08:40 -0700
Subject: IB/ipath: Don't put QP in timeout queue if waiting to send

This fixes a problem which causes too many RC timeouts and
retransmits.

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c
index e3e5332..9e68c91 100644
--- a/drivers/infiniband/hw/ipath/ipath_rc.c
+++ b/drivers/infiniband/hw/ipath/ipath_rc.c
@@ -228,18 +228,13 @@ int ipath_make_rc_req(struct ipath_qp *qp,
 		goto done;
 
 	if (!(ib_ipath_state_ops[qp->state] & IPATH_PROCESS_SEND_OK) ||
-	    qp->s_rnr_timeout)
+	    qp->s_rnr_timeout || qp->s_wait_credit)
 		goto bail;
 
 	/* Limit the number of packets sent without an ACK. */
 	if (ipath_cmp24(qp->s_psn, qp->s_last_psn + IPATH_PSN_CREDIT) > 0) {
 		qp->s_wait_credit = 1;
 		dev->n_rc_stalls++;
-		spin_lock(&dev->pending_lock);
-		if (list_empty(&qp->timerwait))
-			list_add_tail(&qp->timerwait,
-				      &dev->pending[dev->pending_index]);
-		spin_unlock(&dev->pending_lock);
 		goto bail;
 	}
 
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
index 7c4929f..c62f9c5 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h
@@ -422,7 +422,7 @@ struct ipath_qp {
 #define IPATH_S_RDMAR_PENDING	0x04
 #define IPATH_S_ACK_PENDING	0x08
 
-#define IPATH_PSN_CREDIT	2048
+#define IPATH_PSN_CREDIT	512
 
 /*
  * Since struct ipath_swqe is not a fixed size, we can't simply index into
-- 
cgit v0.10.2


From 9ba6d5529dd919b442eedf5bef1dd28aca2ee9fe Mon Sep 17 00:00:00 2001
From: "Michael S. Tsirkin" <mst@mellanox.co.il>
Date: Thu, 12 Apr 2007 18:10:25 +0300
Subject: IB/mthca: Work around kernel QP starvation

With mthca, RC QPs can starve each other and even UD QPs on the same
hardware schedule queue.  As a result, userspace MPI can starve
e.g. IPoIB traffic, with netdev watchdog warnings getting printed out,
and TCP connections getting stuck or failing.

Reduce the chance of this happening by using three separate hardware
schedule queues: one for userspace RC QPs, one for kernel RC QPs, and
one for all other QPs.

Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c
index 8fe6fee..fee60c8 100644
--- a/drivers/infiniband/hw/mthca/mthca_qp.c
+++ b/drivers/infiniband/hw/mthca/mthca_qp.c
@@ -701,6 +701,19 @@ int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask,
 		qp_param->opt_param_mask |= cpu_to_be32(MTHCA_QP_OPTPAR_PRIMARY_ADDR_PATH);
 	}
 
+	if (ibqp->qp_type == IB_QPT_RC &&
+	    cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) {
+		u8 sched_queue = ibqp->uobject ? 0x2 : 0x1;
+
+		if (mthca_is_memfree(dev))
+			qp_context->rlkey_arbel_sched_queue |= sched_queue;
+		else
+			qp_context->tavor_sched_queue |= cpu_to_be32(sched_queue);
+
+		qp_param->opt_param_mask |=
+			cpu_to_be32(MTHCA_QP_OPTPAR_SCHED_QUEUE);
+	}
+
 	if (attr_mask & IB_QP_TIMEOUT) {
 		qp_context->pri_path.ackto = attr->timeout << 3;
 		qp_param->opt_param_mask |= cpu_to_be32(MTHCA_QP_OPTPAR_ACK_TIMEOUT);
-- 
cgit v0.10.2


From 6b66b2da1e821181a001c00b04a807724ad803cd Mon Sep 17 00:00:00 2001
From: Robert Walsh <rjwalsh@pathscale.com>
Date: Fri, 27 Apr 2007 21:07:23 -0700
Subject: IB/ipath: Don't corrupt pending mmap list when unmapped objects are
 freed

Fix the pending mmap code so it doesn't corrupt the list of pending
mmaps and crash the machine when pending mmaps are destroyed without
first being mapped.  Also, remove an unused variable, and use standard
kernel lists instead of our own homebrewed linked list implementation
to keep the pending mmap list.

Signed-off-by: Robert Walsh <robert.walsh@qlogic.com>
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_cq.c b/drivers/infiniband/hw/ipath/ipath_cq.c
index ea78e6d..4715f89 100644
--- a/drivers/infiniband/hw/ipath/ipath_cq.c
+++ b/drivers/infiniband/hw/ipath/ipath_cq.c
@@ -243,33 +243,21 @@ struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries,
 	 * See ipath_mmap() for details.
 	 */
 	if (udata && udata->outlen >= sizeof(__u64)) {
-		struct ipath_mmap_info *ip;
-		__u64 offset = (__u64) wc;
 		int err;
+		u32 s = sizeof *wc + sizeof(struct ib_wc) * entries;
 
-		err = ib_copy_to_udata(udata, &offset, sizeof(offset));
-		if (err) {
-			ret = ERR_PTR(err);
+		cq->ip = ipath_create_mmap_info(dev, s, context, wc);
+		if (!cq->ip) {
+			ret = ERR_PTR(-ENOMEM);
 			goto bail_wc;
 		}
 
-		/* Allocate info for ipath_mmap(). */
-		ip = kmalloc(sizeof(*ip), GFP_KERNEL);
-		if (!ip) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail_wc;
+		err = ib_copy_to_udata(udata, &cq->ip->offset,
+				       sizeof(cq->ip->offset));
+		if (err) {
+			ret = ERR_PTR(err);
+			goto bail_ip;
 		}
-		cq->ip = ip;
-		ip->context = context;
-		ip->obj = wc;
-		kref_init(&ip->ref);
-		ip->mmap_cnt = 0;
-		ip->size = PAGE_ALIGN(sizeof(*wc) +
-				      sizeof(struct ib_wc) * entries);
-		spin_lock_irq(&dev->pending_lock);
-		ip->next = dev->pending_mmaps;
-		dev->pending_mmaps = ip;
-		spin_unlock_irq(&dev->pending_lock);
 	} else
 		cq->ip = NULL;
 
@@ -277,12 +265,18 @@ struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries,
 	if (dev->n_cqs_allocated == ib_ipath_max_cqs) {
 		spin_unlock(&dev->n_cqs_lock);
 		ret = ERR_PTR(-ENOMEM);
-		goto bail_wc;
+		goto bail_ip;
 	}
 
 	dev->n_cqs_allocated++;
 	spin_unlock(&dev->n_cqs_lock);
 
+	if (cq->ip) {
+		spin_lock_irq(&dev->pending_lock);
+		list_add(&cq->ip->pending_mmaps, &dev->pending_mmaps);
+		spin_unlock_irq(&dev->pending_lock);
+	}
+
 	/*
 	 * ib_create_cq() will initialize cq->ibcq except for cq->ibcq.cqe.
 	 * The number of entries should be >= the number requested or return
@@ -301,12 +295,12 @@ struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries,
 
 	goto done;
 
+bail_ip:
+	kfree(cq->ip);
 bail_wc:
 	vfree(wc);
-
 bail_cq:
 	kfree(cq);
-
 done:
 	return ret;
 }
@@ -443,13 +437,12 @@ int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 	if (cq->ip) {
 		struct ipath_ibdev *dev = to_idev(ibcq->device);
 		struct ipath_mmap_info *ip = cq->ip;
+		u32 s = sizeof *wc + sizeof(struct ib_wc) * cqe;
 
-		ip->obj = wc;
-		ip->size = PAGE_ALIGN(sizeof(*wc) +
-				      sizeof(struct ib_wc) * cqe);
+		ipath_update_mmap_info(dev, ip, s, wc);
 		spin_lock_irq(&dev->pending_lock);
-		ip->next = dev->pending_mmaps;
-		dev->pending_mmaps = ip;
+		if (list_empty(&ip->pending_mmaps))
+			list_add(&ip->pending_mmaps, &dev->pending_mmaps);
 		spin_unlock_irq(&dev->pending_lock);
 	}
 
diff --git a/drivers/infiniband/hw/ipath/ipath_mmap.c b/drivers/infiniband/hw/ipath/ipath_mmap.c
index a82157d..937bc33 100644
--- a/drivers/infiniband/hw/ipath/ipath_mmap.c
+++ b/drivers/infiniband/hw/ipath/ipath_mmap.c
@@ -46,6 +46,11 @@ void ipath_release_mmap_info(struct kref *ref)
 {
 	struct ipath_mmap_info *ip =
 		container_of(ref, struct ipath_mmap_info, ref);
+	struct ipath_ibdev *dev = to_idev(ip->context->device);
+
+	spin_lock_irq(&dev->pending_lock);
+	list_del(&ip->pending_mmaps);
+	spin_unlock_irq(&dev->pending_lock);
 
 	vfree(ip->obj);
 	kfree(ip);
@@ -60,14 +65,12 @@ static void ipath_vma_open(struct vm_area_struct *vma)
 	struct ipath_mmap_info *ip = vma->vm_private_data;
 
 	kref_get(&ip->ref);
-	ip->mmap_cnt++;
 }
 
 static void ipath_vma_close(struct vm_area_struct *vma)
 {
 	struct ipath_mmap_info *ip = vma->vm_private_data;
 
-	ip->mmap_cnt--;
 	kref_put(&ip->ref, ipath_release_mmap_info);
 }
 
@@ -87,7 +90,7 @@ int ipath_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
 	struct ipath_ibdev *dev = to_idev(context->device);
 	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
 	unsigned long size = vma->vm_end - vma->vm_start;
-	struct ipath_mmap_info *ip, **pp;
+	struct ipath_mmap_info *ip, *pp;
 	int ret = -EINVAL;
 
 	/*
@@ -96,15 +99,16 @@ int ipath_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
 	 * CQ, QP, or SRQ is soon followed by a call to mmap().
 	 */
 	spin_lock_irq(&dev->pending_lock);
-	for (pp = &dev->pending_mmaps; (ip = *pp); pp = &ip->next) {
+	list_for_each_entry_safe(ip, pp, &dev->pending_mmaps,
+				 pending_mmaps) {
 		/* Only the creator is allowed to mmap the object */
-		if (context != ip->context || (void *) offset != ip->obj)
+		if (context != ip->context || (__u64) offset != ip->offset)
 			continue;
 		/* Don't allow a mmap larger than the object. */
 		if (size > ip->size)
 			break;
 
-		*pp = ip->next;
+		list_del_init(&ip->pending_mmaps);
 		spin_unlock_irq(&dev->pending_lock);
 
 		ret = remap_vmalloc_range(vma, ip->obj, 0);
@@ -119,3 +123,51 @@ int ipath_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
 done:
 	return ret;
 }
+
+/*
+ * Allocate information for ipath_mmap
+ */
+struct ipath_mmap_info *ipath_create_mmap_info(struct ipath_ibdev *dev,
+					       u32 size,
+					       struct ib_ucontext *context,
+					       void *obj) {
+	struct ipath_mmap_info *ip;
+
+	ip = kmalloc(sizeof *ip, GFP_KERNEL);
+	if (!ip)
+		goto bail;
+
+	size = PAGE_ALIGN(size);
+
+	spin_lock_irq(&dev->mmap_offset_lock);
+	if (dev->mmap_offset == 0)
+		dev->mmap_offset = PAGE_SIZE;
+	ip->offset = dev->mmap_offset;
+	dev->mmap_offset += size;
+	spin_unlock_irq(&dev->mmap_offset_lock);
+
+	INIT_LIST_HEAD(&ip->pending_mmaps);
+	ip->size = size;
+	ip->context = context;
+	ip->obj = obj;
+	kref_init(&ip->ref);
+
+bail:
+	return ip;
+}
+
+void ipath_update_mmap_info(struct ipath_ibdev *dev,
+			    struct ipath_mmap_info *ip,
+			    u32 size, void *obj) {
+	size = PAGE_ALIGN(size);
+
+	spin_lock_irq(&dev->mmap_offset_lock);
+	if (dev->mmap_offset == 0)
+		dev->mmap_offset = PAGE_SIZE;
+	ip->offset = dev->mmap_offset;
+	dev->mmap_offset += size;
+	spin_unlock_irq(&dev->mmap_offset_lock);
+
+	ip->size = size;
+	ip->obj = obj;
+}
diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c
index 16db9ac..bfef08e 100644
--- a/drivers/infiniband/hw/ipath/ipath_qp.c
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c
@@ -844,34 +844,36 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd,
 	 * See ipath_mmap() for details.
 	 */
 	if (udata && udata->outlen >= sizeof(__u64)) {
-		struct ipath_mmap_info *ip;
-		__u64 offset = (__u64) qp->r_rq.wq;
 		int err;
 
-		err = ib_copy_to_udata(udata, &offset, sizeof(offset));
-		if (err) {
-			ret = ERR_PTR(err);
-			goto bail_rwq;
-		}
+		if (!qp->r_rq.wq) {
+			__u64 offset = 0;
 
-		if (qp->r_rq.wq) {
-			/* Allocate info for ipath_mmap(). */
-			ip = kmalloc(sizeof(*ip), GFP_KERNEL);
-			if (!ip) {
+			err = ib_copy_to_udata(udata, &offset,
+					       sizeof(offset));
+			if (err) {
+				ret = ERR_PTR(err);
+				goto bail_rwq;
+			}
+		} else {
+			u32 s = sizeof(struct ipath_rwq) +
+				qp->r_rq.size * sz;
+
+			qp->ip =
+			    ipath_create_mmap_info(dev, s,
+						   ibpd->uobject->context,
+						   qp->r_rq.wq);
+			if (!qp->ip) {
 				ret = ERR_PTR(-ENOMEM);
 				goto bail_rwq;
 			}
-			qp->ip = ip;
-			ip->context = ibpd->uobject->context;
-			ip->obj = qp->r_rq.wq;
-			kref_init(&ip->ref);
-			ip->mmap_cnt = 0;
-			ip->size = PAGE_ALIGN(sizeof(struct ipath_rwq) +
-					      qp->r_rq.size * sz);
-			spin_lock_irq(&dev->pending_lock);
-			ip->next = dev->pending_mmaps;
-			dev->pending_mmaps = ip;
-			spin_unlock_irq(&dev->pending_lock);
+
+			err = ib_copy_to_udata(udata, &(qp->ip->offset),
+					       sizeof(qp->ip->offset));
+			if (err) {
+				ret = ERR_PTR(err);
+				goto bail_ip;
+			}
 		}
 	}
 
@@ -885,6 +887,12 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd,
 	dev->n_qps_allocated++;
 	spin_unlock(&dev->n_qps_lock);
 
+	if (qp->ip) {
+		spin_lock_irq(&dev->pending_lock);
+		list_add(&qp->ip->pending_mmaps, &dev->pending_mmaps);
+		spin_unlock_irq(&dev->pending_lock);
+	}
+
 	ret = &qp->ibqp;
 	goto bail;
 
diff --git a/drivers/infiniband/hw/ipath/ipath_srq.c b/drivers/infiniband/hw/ipath/ipath_srq.c
index 9403350..03acae6 100644
--- a/drivers/infiniband/hw/ipath/ipath_srq.c
+++ b/drivers/infiniband/hw/ipath/ipath_srq.c
@@ -139,33 +139,24 @@ struct ib_srq *ipath_create_srq(struct ib_pd *ibpd,
 	 * See ipath_mmap() for details.
 	 */
 	if (udata && udata->outlen >= sizeof(__u64)) {
-		struct ipath_mmap_info *ip;
-		__u64 offset = (__u64) srq->rq.wq;
 		int err;
+		u32 s = sizeof(struct ipath_rwq) + srq->rq.size * sz;
 
-		err = ib_copy_to_udata(udata, &offset, sizeof(offset));
-		if (err) {
-			ret = ERR_PTR(err);
+		srq->ip =
+		    ipath_create_mmap_info(dev, s,
+					   ibpd->uobject->context,
+					   srq->rq.wq);
+		if (!srq->ip) {
+			ret = ERR_PTR(-ENOMEM);
 			goto bail_wq;
 		}
 
-		/* Allocate info for ipath_mmap(). */
-		ip = kmalloc(sizeof(*ip), GFP_KERNEL);
-		if (!ip) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail_wq;
+		err = ib_copy_to_udata(udata, &srq->ip->offset,
+				       sizeof(srq->ip->offset));
+		if (err) {
+			ret = ERR_PTR(err);
+			goto bail_ip;
 		}
-		srq->ip = ip;
-		ip->context = ibpd->uobject->context;
-		ip->obj = srq->rq.wq;
-		kref_init(&ip->ref);
-		ip->mmap_cnt = 0;
-		ip->size = PAGE_ALIGN(sizeof(struct ipath_rwq) +
-				      srq->rq.size * sz);
-		spin_lock_irq(&dev->pending_lock);
-		ip->next = dev->pending_mmaps;
-		dev->pending_mmaps = ip;
-		spin_unlock_irq(&dev->pending_lock);
 	} else
 		srq->ip = NULL;
 
@@ -181,21 +172,27 @@ struct ib_srq *ipath_create_srq(struct ib_pd *ibpd,
 	if (dev->n_srqs_allocated == ib_ipath_max_srqs) {
 		spin_unlock(&dev->n_srqs_lock);
 		ret = ERR_PTR(-ENOMEM);
-		goto bail_wq;
+		goto bail_ip;
 	}
 
  	dev->n_srqs_allocated++;
 	spin_unlock(&dev->n_srqs_lock);
 
+	if (srq->ip) {
+		spin_lock_irq(&dev->pending_lock);
+		list_add(&srq->ip->pending_mmaps, &dev->pending_mmaps);
+		spin_unlock_irq(&dev->pending_lock);
+	}
+
 	ret = &srq->ibsrq;
 	goto done;
 
+bail_ip:
+	kfree(srq->ip);
 bail_wq:
 	vfree(srq->rq.wq);
-
 bail_srq:
 	kfree(srq);
-
 done:
 	return ret;
 }
@@ -312,13 +309,13 @@ int ipath_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 		if (srq->ip) {
 			struct ipath_mmap_info *ip = srq->ip;
 			struct ipath_ibdev *dev = to_idev(srq->ibsrq.device);
+			u32 s = sizeof(struct ipath_rwq) + size * sz;
 
-			ip->obj = wq;
-			ip->size = PAGE_ALIGN(sizeof(struct ipath_rwq) +
-					      size * sz);
+			ipath_update_mmap_info(dev, ip, s, wq);
 			spin_lock_irq(&dev->pending_lock);
-			ip->next = dev->pending_mmaps;
-			dev->pending_mmaps = ip;
+			if (list_empty(&ip->pending_mmaps))
+				list_add(&ip->pending_mmaps,
+					 &dev->pending_mmaps);
 			spin_unlock_irq(&dev->pending_lock);
 		}
 	} else if (attr_mask & IB_SRQ_LIMIT) {
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
index 18c6df2..b676ea8 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c
@@ -1476,7 +1476,10 @@ int ipath_register_ib_device(struct ipath_devdata *dd)
 		ret = -ENOMEM;
 		goto err_lk;
 	}
+	INIT_LIST_HEAD(&idev->pending_mmaps);
 	spin_lock_init(&idev->pending_lock);
+	idev->mmap_offset = PAGE_SIZE;
+	spin_lock_init(&idev->mmap_offset_lock);
 	INIT_LIST_HEAD(&idev->pending[0]);
 	INIT_LIST_HEAD(&idev->pending[1]);
 	INIT_LIST_HEAD(&idev->pending[2]);
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
index c62f9c5..ac66c00 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h
@@ -173,12 +173,12 @@ struct ipath_ah {
  * this as its vm_private_data.
  */
 struct ipath_mmap_info {
-	struct ipath_mmap_info *next;
+	struct list_head pending_mmaps;
 	struct ib_ucontext *context;
 	void *obj;
+	__u64 offset;
 	struct kref ref;
 	unsigned size;
-	unsigned mmap_cnt;
 };
 
 /*
@@ -485,9 +485,10 @@ struct ipath_opcode_stats {
 
 struct ipath_ibdev {
 	struct ib_device ibdev;
-	struct list_head dev_list;
 	struct ipath_devdata *dd;
-	struct ipath_mmap_info *pending_mmaps;
+	struct list_head pending_mmaps;
+	spinlock_t mmap_offset_lock;
+	u32 mmap_offset;
 	int ib_unit;		/* This is the device number */
 	u16 sm_lid;		/* in host order */
 	u8 sm_sl;
@@ -768,6 +769,15 @@ int ipath_dealloc_fmr(struct ib_fmr *ibfmr);
 
 void ipath_release_mmap_info(struct kref *ref);
 
+struct ipath_mmap_info *ipath_create_mmap_info(struct ipath_ibdev *dev,
+					       u32 size,
+					       struct ib_ucontext *context,
+					       void *obj);
+
+void ipath_update_mmap_info(struct ipath_ibdev *dev,
+			    struct ipath_mmap_info *ip,
+			    u32 size, void *obj);
+
 int ipath_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
 
 void ipath_no_bufs_available(struct ipath_qp *qp, struct ipath_ibdev *dev);
-- 
cgit v0.10.2


From 347fcfbed261fdd11f46fa03d524e1bddddab3a6 Mon Sep 17 00:00:00 2001
From: "Michael S. Tsirkin" <mst@dev.mellanox.co.il>
Date: Mon, 30 Apr 2007 17:30:28 -0700
Subject: IPoIB/cm: Fix error handling in ipoib_cm_dev_open()

If skb allocation fails when we start the device, we call
ipoib_cm_dev_stop() even though ipoib_cm_dev_open() did not run to
completion, so we pass an invalid pointer to ib_destroy_cm_id and get
an oops.

Fix by clearing cm.id on error, and testing it in cm_dev_stop().
This fixes <https://bugs.openfabrics.org/show_bug.cgi?id=561>

Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
index 0c4e59b..7671741 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
@@ -592,7 +592,9 @@ int ipoib_cm_dev_open(struct net_device *dev)
 	priv->cm.id = ib_create_cm_id(priv->ca, ipoib_cm_rx_handler, dev);
 	if (IS_ERR(priv->cm.id)) {
 		printk(KERN_WARNING "%s: failed to create CM ID\n", priv->ca->name);
-		return IS_ERR(priv->cm.id);
+		ret = PTR_ERR(priv->cm.id);
+		priv->cm.id = NULL;
+		return ret;
 	}
 
 	ret = ib_cm_listen(priv->cm.id, cpu_to_be64(IPOIB_CM_IETF_ID | priv->qp->qp_num),
@@ -601,6 +603,7 @@ int ipoib_cm_dev_open(struct net_device *dev)
 		printk(KERN_WARNING "%s: failed to listen on ID 0x%llx\n", priv->ca->name,
 		       IPOIB_CM_IETF_ID | priv->qp->qp_num);
 		ib_destroy_cm_id(priv->cm.id);
+		priv->cm.id = NULL;
 		return ret;
 	}
 	return 0;
@@ -611,10 +614,11 @@ void ipoib_cm_dev_stop(struct net_device *dev)
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_cm_rx *p;
 
-	if (!IPOIB_CM_SUPPORTED(dev->dev_addr))
+	if (!IPOIB_CM_SUPPORTED(dev->dev_addr) || !priv->cm.id)
 		return;
 
 	ib_destroy_cm_id(priv->cm.id);
+	priv->cm.id = NULL;
 	spin_lock_irq(&priv->lock);
 	while (!list_empty(&priv->cm.passive_ids)) {
 		p = list_entry(priv->cm.passive_ids.next, typeof(*p), list);
-- 
cgit v0.10.2


From 4a97d47ef7946cf31b76945c3199b0b5cad6a8ed Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Thu, 26 Apr 2007 15:21:02 -0500
Subject: RDMA/cxgb3: Fix TERM codes

Fix TERMINATE layer, type, and ecode values based on
conformance testing.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c
index 0a472c9..714dddb 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_qp.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c
@@ -471,43 +471,62 @@ int iwch_bind_mw(struct ib_qp *qp,
 	return err;
 }
 
-static void build_term_codes(int t3err, u8 *layer_type, u8 *ecode, int tagged)
+static inline void build_term_codes(struct respQ_msg_t *rsp_msg,
+				    u8 *layer_type, u8 *ecode)
 {
-	switch (t3err) {
+	int status = TPT_ERR_INTERNAL_ERR;
+	int tagged = 0;
+	int opcode = -1;
+	int rqtype = 0;
+	int send_inv = 0;
+
+	if (rsp_msg) {
+		status = CQE_STATUS(rsp_msg->cqe);
+		opcode = CQE_OPCODE(rsp_msg->cqe);
+		rqtype = RQ_TYPE(rsp_msg->cqe);
+		send_inv = (opcode == T3_SEND_WITH_INV) ||
+		           (opcode == T3_SEND_WITH_SE_INV);
+		tagged = (opcode == T3_RDMA_WRITE) ||
+			 (rqtype && (opcode == T3_READ_RESP));
+	}
+
+	switch (status) {
 	case TPT_ERR_STAG:
-		if (tagged == 1) {
-			*layer_type = LAYER_DDP|DDP_TAGGED_ERR;
-			*ecode = DDPT_INV_STAG;
-		} else if (tagged == 2) {
+		if (send_inv) {
+			*layer_type = LAYER_RDMAP|RDMAP_REMOTE_OP;
+			*ecode = RDMAP_CANT_INV_STAG;
+		} else {
 			*layer_type = LAYER_RDMAP|RDMAP_REMOTE_PROT;
 			*ecode = RDMAP_INV_STAG;
 		}
 		break;
 	case TPT_ERR_PDID:
+		*layer_type = LAYER_RDMAP|RDMAP_REMOTE_PROT;
+		if ((opcode == T3_SEND_WITH_INV) ||
+		    (opcode == T3_SEND_WITH_SE_INV))
+			*ecode = RDMAP_CANT_INV_STAG;
+		else
+			*ecode = RDMAP_STAG_NOT_ASSOC;
+		break;
 	case TPT_ERR_QPID:
+		*layer_type = LAYER_RDMAP|RDMAP_REMOTE_PROT;
+		*ecode = RDMAP_STAG_NOT_ASSOC;
+		break;
 	case TPT_ERR_ACCESS:
-		if (tagged == 1) {
-			*layer_type = LAYER_DDP|DDP_TAGGED_ERR;
-			*ecode = DDPT_STAG_NOT_ASSOC;
-		} else if (tagged == 2) {
-			*layer_type = LAYER_RDMAP|RDMAP_REMOTE_PROT;
-			*ecode = RDMAP_STAG_NOT_ASSOC;
-		}
+		*layer_type = LAYER_RDMAP|RDMAP_REMOTE_PROT;
+		*ecode = RDMAP_ACC_VIOL;
 		break;
 	case TPT_ERR_WRAP:
 		*layer_type = LAYER_RDMAP|RDMAP_REMOTE_PROT;
 		*ecode = RDMAP_TO_WRAP;
 		break;
 	case TPT_ERR_BOUND:
-		if (tagged == 1) {
+		if (tagged) {
 			*layer_type = LAYER_DDP|DDP_TAGGED_ERR;
 			*ecode = DDPT_BASE_BOUNDS;
-		} else if (tagged == 2) {
+		} else {
 			*layer_type = LAYER_RDMAP|RDMAP_REMOTE_PROT;
 			*ecode = RDMAP_BASE_BOUNDS;
-		} else {
-			*layer_type = LAYER_DDP|DDP_UNTAGGED_ERR;
-			*ecode = DDPU_MSG_TOOBIG;
 		}
 		break;
 	case TPT_ERR_INVALIDATE_SHARED_MR:
@@ -591,8 +610,6 @@ int iwch_post_terminate(struct iwch_qp *qhp, struct respQ_msg_t *rsp_msg)
 {
 	union t3_wr *wqe;
 	struct terminate_message *term;
-	int status;
-	int tagged = 0;
 	struct sk_buff *skb;
 
 	PDBG("%s %d\n", __FUNCTION__, __LINE__);
@@ -610,17 +627,7 @@ int iwch_post_terminate(struct iwch_qp *qhp, struct respQ_msg_t *rsp_msg)
 
 	/* immediate data starts here. */
 	term = (struct terminate_message *)wqe->send.sgl;
-	if (rsp_msg) {
-		status = CQE_STATUS(rsp_msg->cqe);
-		if (CQE_OPCODE(rsp_msg->cqe) == T3_RDMA_WRITE)
-			tagged = 1;
-		if ((CQE_OPCODE(rsp_msg->cqe) == T3_READ_REQ) ||
-		    (CQE_OPCODE(rsp_msg->cqe) == T3_READ_RESP))
-			tagged = 2;
-	} else {
-		status = TPT_ERR_INTERNAL_ERR;
-	}
-	build_term_codes(status, &term->layer_etype, &term->ecode, tagged);
+	build_term_codes(rsp_msg, &term->layer_etype, &term->ecode);
 	build_fw_riwrh((void *)wqe, T3_WR_SEND,
 		       T3_COMPLETION_FLAG | T3_NOTIFY_FLAG, 1,
 		       qhp->ep->hwtid, 5);
-- 
cgit v0.10.2


From 1860cdf802310e4a988e0b8fca41cc97da36f779 Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Thu, 26 Apr 2007 15:21:09 -0500
Subject: RDMA/cxgb3: Fail qp creation if the requested max_inline is too large

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/cxgb3/cxio_wr.h b/drivers/infiniband/hw/cxgb3/cxio_wr.h
index 90d7b89..ff7290e 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_wr.h
+++ b/drivers/infiniband/hw/cxgb3/cxio_wr.h
@@ -38,6 +38,7 @@
 #include "firmware_exports.h"
 
 #define T3_MAX_SGE      4
+#define T3_MAX_INLINE	64
 
 #define Q_EMPTY(rptr,wptr) ((rptr)==(wptr))
 #define Q_FULL(rptr,wptr,size_log2)  ( (((wptr)-(rptr))>>(size_log2)) && \
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index af28a31..93038c0 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -780,6 +780,9 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd,
 	if (rqsize > T3_MAX_RQ_SIZE)
 		return ERR_PTR(-EINVAL);
 
+	if (attrs->cap.max_inline_data > T3_MAX_INLINE)
+		return ERR_PTR(-EINVAL);
+
 	/*
 	 * NOTE: The SQ and total WQ sizes don't need to be
 	 * a power of two.  However, all the code assumes
-- 
cgit v0.10.2


From 60be4b5966e22040f97db9dada72841bf90479d1 Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Thu, 26 Apr 2007 15:21:15 -0500
Subject: RDMA/cxgb3: Initialize cpu_idx field in cpl_close_listserv_req
 message

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/cxgb3/iwch_cm.c b/drivers/infiniband/hw/cxgb3/iwch_cm.c
index 3b4b0ac..a02be40 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_cm.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_cm.c
@@ -1189,6 +1189,7 @@ static int listen_stop(struct iwch_listen_ep *ep)
 	}
 	req = (struct cpl_close_listserv_req *) skb_put(skb, sizeof(*req));
 	req->wr.wr_hi = htonl(V_WR_OP(FW_WROPCODE_FORWARD));
+	req->cpu_idx = 0;
 	OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_CLOSE_LISTSRV_REQ, ep->stid));
 	skb->priority = 1;
 	ep->com.tdev->send(ep->com.tdev, skb);
-- 
cgit v0.10.2


From aff9e39d97585486764572ab2f3bf5dfce18c660 Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Thu, 26 Apr 2007 15:21:20 -0500
Subject: RDMA/cxgb3: Support for new abort logic

The HW now posts 2 ABORT_RPL and/or PEER_ABORT_REQ messages.  We need
to handle them by silenty dropping the 1st but mark that we're ready
for the final message.  This plugs some close races between the uP and
HW.  Also update the minimum required firmware version.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/cxgb3/iwch_cm.c b/drivers/infiniband/hw/cxgb3/iwch_cm.c
index a02be40..b2faff5 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_cm.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_cm.c
@@ -1109,6 +1109,15 @@ static int abort_rpl(struct t3cdev *tdev, struct sk_buff *skb, void *ctx)
 
 	PDBG("%s ep %p\n", __FUNCTION__, ep);
 
+	/*
+	 * We get 2 abort replies from the HW.  The first one must
+	 * be ignored except for scribbling that we need one more.
+	 */
+	if (!(ep->flags & ABORT_REQ_IN_PROGRESS)) {
+		ep->flags |= ABORT_REQ_IN_PROGRESS;
+		return CPL_RET_BUF_DONE;
+	}
+
 	close_complete_upcall(ep);
 	state_set(&ep->com, DEAD);
 	release_ep_resources(ep);
@@ -1476,6 +1485,15 @@ static int peer_abort(struct t3cdev *tdev, struct sk_buff *skb, void *ctx)
 	int ret;
 	int state;
 
+	/*
+	 * We get 2 peer aborts from the HW.  The first one must
+	 * be ignored except for scribbling that we need one more.
+	 */
+	if (!(ep->flags & PEER_ABORT_IN_PROGRESS)) {
+		ep->flags |= PEER_ABORT_IN_PROGRESS;
+		return CPL_RET_BUF_DONE;
+	}
+
 	if (is_neg_adv_abort(req->status)) {
 		PDBG("%s neg_adv_abort ep %p tid %d\n", __FUNCTION__, ep,
 		     ep->hwtid);
diff --git a/drivers/infiniband/hw/cxgb3/iwch_cm.h b/drivers/infiniband/hw/cxgb3/iwch_cm.h
index 0c6f281..21a388c 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_cm.h
+++ b/drivers/infiniband/hw/cxgb3/iwch_cm.h
@@ -143,6 +143,11 @@ enum iwch_ep_state {
 	DEAD,
 };
 
+enum iwch_ep_flags {
+	PEER_ABORT_IN_PROGRESS	= (1 << 0),
+	ABORT_REQ_IN_PROGRESS	= (1 << 1),
+};
+
 struct iwch_ep_common {
 	struct iw_cm_id *cm_id;
 	struct iwch_qp *qp;
@@ -181,6 +186,7 @@ struct iwch_ep {
 	u16 plen;
 	u32 ird;
 	u32 ord;
+	u32 flags;
 };
 
 static inline struct iwch_ep *to_ep(struct iw_cm_id *cm_id)
diff --git a/drivers/net/cxgb3/version.h b/drivers/net/cxgb3/version.h
index 042e27e..b112317 100644
--- a/drivers/net/cxgb3/version.h
+++ b/drivers/net/cxgb3/version.h
@@ -38,7 +38,7 @@
 #define DRV_VERSION "1.0-ko"
 
 /* Firmware version */
-#define FW_VERSION_MAJOR 3
-#define FW_VERSION_MINOR 3
+#define FW_VERSION_MAJOR 4
+#define FW_VERSION_MINOR 0
 #define FW_VERSION_MICRO 0
 #endif				/* __CHELSIO_VERSION_H */
-- 
cgit v0.10.2


From d6ef7d68f6f51c5b9de01c542dab8b90067a9c27 Mon Sep 17 00:00:00 2001
From: "Michael S. Tsirkin" <mst@dev.mellanox.co.il>
Date: Wed, 2 May 2007 15:31:12 +0300
Subject: IPoIB/cm: Don't crash if remote side uses one QP for both directions

The IPoIB CM spec allows the use of a single connection in both
active->passive and passive->active directions.  The current Linux
code uses one connection for both directions, but if another node only
uses one connection for both directions, we oops when we try to look
up the passive connection.  Fix by checking that qp_context is
non-NULL before dereferencing it.

Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il>

diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
index 7671741..1e27930 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
@@ -370,7 +370,7 @@ void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 
 	if (!likely(wr_id & IPOIB_CM_RX_UPDATE_MASK)) {
 		p = wc->qp->qp_context;
-		if (time_after_eq(jiffies, p->jiffies + IPOIB_CM_RX_UPDATE_TIME)) {
+		if (p && time_after_eq(jiffies, p->jiffies + IPOIB_CM_RX_UPDATE_TIME)) {
 			spin_lock_irqsave(&priv->lock, flags);
 			p->jiffies = jiffies;
 			/* Move this entry to list head, but do
-- 
cgit v0.10.2


From 3633b3d096286cf21bc07b16fa6265fb006d0844 Mon Sep 17 00:00:00 2001
From: Ishai Rabinovitz <ishai@mellanox.co.il>
Date: Sun, 6 May 2007 21:18:11 -0700
Subject: IB/srp: Add orig_dgid sysfs attribute to scsi_host

Add an orig_dgid attribute in sysfs for SRP scsi_hosts, so that
userspace can tell what the original dgid value written to the
add_target file was, even if the connection is redirected to a
different port while connecting.

Signed-off-by: Ishai Rabinovitz <ishai@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c
index 5e8ac57..df526cb 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.c
+++ b/drivers/infiniband/ulp/srp/ib_srp.c
@@ -1468,6 +1468,25 @@ static ssize_t show_dgid(struct class_device *cdev, char *buf)
 		       be16_to_cpu(((__be16 *) target->path.dgid.raw)[7]));
 }
 
+static ssize_t show_orig_dgid(struct class_device *cdev, char *buf)
+{
+	struct srp_target_port *target = host_to_target(class_to_shost(cdev));
+
+	if (target->state == SRP_TARGET_DEAD ||
+	    target->state == SRP_TARGET_REMOVED)
+		return -ENODEV;
+
+	return sprintf(buf, "%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x\n",
+		       be16_to_cpu(target->orig_dgid[0]),
+		       be16_to_cpu(target->orig_dgid[1]),
+		       be16_to_cpu(target->orig_dgid[2]),
+		       be16_to_cpu(target->orig_dgid[3]),
+		       be16_to_cpu(target->orig_dgid[4]),
+		       be16_to_cpu(target->orig_dgid[5]),
+		       be16_to_cpu(target->orig_dgid[6]),
+		       be16_to_cpu(target->orig_dgid[7]));
+}
+
 static ssize_t show_zero_req_lim(struct class_device *cdev, char *buf)
 {
 	struct srp_target_port *target = host_to_target(class_to_shost(cdev));
@@ -1498,6 +1517,7 @@ static CLASS_DEVICE_ATTR(ioc_guid,	  S_IRUGO, show_ioc_guid,	 NULL);
 static CLASS_DEVICE_ATTR(service_id,	  S_IRUGO, show_service_id,	 NULL);
 static CLASS_DEVICE_ATTR(pkey,		  S_IRUGO, show_pkey,		 NULL);
 static CLASS_DEVICE_ATTR(dgid,		  S_IRUGO, show_dgid,		 NULL);
+static CLASS_DEVICE_ATTR(orig_dgid,	  S_IRUGO, show_orig_dgid,	 NULL);
 static CLASS_DEVICE_ATTR(zero_req_lim,	  S_IRUGO, show_zero_req_lim,	 NULL);
 static CLASS_DEVICE_ATTR(local_ib_port,   S_IRUGO, show_local_ib_port,	 NULL);
 static CLASS_DEVICE_ATTR(local_ib_device, S_IRUGO, show_local_ib_device, NULL);
@@ -1508,6 +1528,7 @@ static struct class_device_attribute *srp_host_attrs[] = {
 	&class_device_attr_service_id,
 	&class_device_attr_pkey,
 	&class_device_attr_dgid,
+	&class_device_attr_orig_dgid,
 	&class_device_attr_zero_req_lim,
 	&class_device_attr_local_ib_port,
 	&class_device_attr_local_ib_device,
@@ -1662,6 +1683,7 @@ static int srp_parse_options(const char *buf, struct srp_target_port *target)
 				target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16);
 			}
 			kfree(p);
+			memcpy(target->orig_dgid, target->path.dgid.raw, 16);
 			break;
 
 		case SRP_OPT_PKEY:
diff --git a/drivers/infiniband/ulp/srp/ib_srp.h b/drivers/infiniband/ulp/srp/ib_srp.h
index 2f3319c..1d53c7b 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.h
+++ b/drivers/infiniband/ulp/srp/ib_srp.h
@@ -129,6 +129,7 @@ struct srp_target_port {
 	unsigned int		scsi_id;
 
 	struct ib_sa_path_rec	path;
+	__be16			orig_dgid[8];
 	struct ib_sa_query     *path_query;
 	int			path_query_id;
 
-- 
cgit v0.10.2


From b7f008fdc92e498af34671048556fd17ddfe9be9 Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Sun, 6 May 2007 21:18:11 -0700
Subject: IB/srp: Set proc_name

Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c
index df526cb..3468ae1 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.c
+++ b/drivers/infiniband/ulp/srp/ib_srp.c
@@ -1537,7 +1537,8 @@ static struct class_device_attribute *srp_host_attrs[] = {
 
 static struct scsi_host_template srp_template = {
 	.module				= THIS_MODULE,
-	.name				= DRV_NAME,
+	.name				= "InfiniBand SRP initiator",
+	.proc_name			= DRV_NAME,
 	.info				= srp_target_info,
 	.queuecommand			= srp_queuecommand,
 	.eh_abort_handler		= srp_abort,
-- 
cgit v0.10.2


From 1a70a05d9d2b30db3e56f8cfbebb175663b41bad Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Sun, 6 May 2007 21:18:11 -0700
Subject: IB/fmr_pool: Add prefix to all printks

Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/core/fmr_pool.c b/drivers/infiniband/core/fmr_pool.c
index 1d796e7..a06bcc6 100644
--- a/drivers/infiniband/core/fmr_pool.c
+++ b/drivers/infiniband/core/fmr_pool.c
@@ -43,6 +43,8 @@
 
 #include "core_priv.h"
 
+#define PFX "fmr_pool: "
+
 enum {
 	IB_FMR_MAX_REMAPS = 32,
 
@@ -150,7 +152,7 @@ static void ib_fmr_batch_release(struct ib_fmr_pool *pool)
 
 #ifdef DEBUG
 		if (fmr->ref_count !=0) {
-			printk(KERN_WARNING "Unmapping FMR 0x%08x with ref count %d",
+			printk(KERN_WARNING PFX "Unmapping FMR 0x%08x with ref count %d",
 			       fmr, fmr->ref_count);
 		}
 #endif
@@ -168,7 +170,7 @@ static void ib_fmr_batch_release(struct ib_fmr_pool *pool)
 
 	ret = ib_unmap_fmr(&fmr_list);
 	if (ret)
-		printk(KERN_WARNING "ib_unmap_fmr returned %d", ret);
+		printk(KERN_WARNING PFX "ib_unmap_fmr returned %d", ret);
 
 	spin_lock_irq(&pool->pool_lock);
 	list_splice(&unmap_list, &pool->free_list);
@@ -226,20 +228,20 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd             *pd,
 	device = pd->device;
 	if (!device->alloc_fmr    || !device->dealloc_fmr  ||
 	    !device->map_phys_fmr || !device->unmap_fmr) {
-		printk(KERN_WARNING "Device %s does not support fast memory regions",
+		printk(KERN_INFO PFX "Device %s does not support FMRs\n",
 		       device->name);
 		return ERR_PTR(-ENOSYS);
 	}
 
 	attr = kmalloc(sizeof *attr, GFP_KERNEL);
 	if (!attr) {
-		printk(KERN_WARNING "couldn't allocate device attr struct");
+		printk(KERN_WARNING PFX "couldn't allocate device attr struct");
 		return ERR_PTR(-ENOMEM);
 	}
 
 	ret = ib_query_device(device, attr);
 	if (ret) {
-		printk(KERN_WARNING "couldn't query device");
+		printk(KERN_WARNING PFX "couldn't query device: %d", ret);
 		kfree(attr);
 		return ERR_PTR(ret);
 	}
@@ -253,7 +255,7 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd             *pd,
 
 	pool = kmalloc(sizeof *pool, GFP_KERNEL);
 	if (!pool) {
-		printk(KERN_WARNING "couldn't allocate pool struct");
+		printk(KERN_WARNING PFX "couldn't allocate pool struct");
 		return ERR_PTR(-ENOMEM);
 	}
 
@@ -270,7 +272,7 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd             *pd,
 			kmalloc(IB_FMR_HASH_SIZE * sizeof *pool->cache_bucket,
 				GFP_KERNEL);
 		if (!pool->cache_bucket) {
-			printk(KERN_WARNING "Failed to allocate cache in pool");
+			printk(KERN_WARNING PFX "Failed to allocate cache in pool");
 			ret = -ENOMEM;
 			goto out_free_pool;
 		}
@@ -294,7 +296,7 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd             *pd,
 				      "ib_fmr(%s)",
 				      device->name);
 	if (IS_ERR(pool->thread)) {
-		printk(KERN_WARNING "couldn't start cleanup thread");
+		printk(KERN_WARNING PFX "couldn't start cleanup thread");
 		ret = PTR_ERR(pool->thread);
 		goto out_free_pool;
 	}
@@ -311,8 +313,8 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd             *pd,
 			fmr = kmalloc(sizeof *fmr + params->max_pages_per_fmr * sizeof (u64),
 				      GFP_KERNEL);
 			if (!fmr) {
-				printk(KERN_WARNING "failed to allocate fmr struct "
-				       "for FMR %d", i);
+				printk(KERN_WARNING PFX "failed to allocate fmr "
+				       "struct for FMR %d", i);
 				goto out_fail;
 			}
 
@@ -323,7 +325,8 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd             *pd,
 
 			fmr->fmr = ib_alloc_fmr(pd, params->access, &fmr_attr);
 			if (IS_ERR(fmr->fmr)) {
-				printk(KERN_WARNING "fmr_create failed for FMR %d", i);
+				printk(KERN_WARNING PFX "fmr_create failed "
+				       "for FMR %d", i);
 				kfree(fmr);
 				goto out_fail;
 			}
@@ -378,7 +381,7 @@ void ib_destroy_fmr_pool(struct ib_fmr_pool *pool)
 	}
 
 	if (i < pool->pool_size)
-		printk(KERN_WARNING "pool still has %d regions registered",
+		printk(KERN_WARNING PFX "pool still has %d regions registered",
 		       pool->pool_size - i);
 
 	kfree(pool->cache_bucket);
@@ -463,8 +466,7 @@ struct ib_pool_fmr *ib_fmr_pool_map_phys(struct ib_fmr_pool *pool_handle,
 		list_add(&fmr->list, &pool->free_list);
 		spin_unlock_irqrestore(&pool->pool_lock, flags);
 
-		printk(KERN_WARNING "fmr_map returns %d\n",
-		       result);
+		printk(KERN_WARNING PFX "fmr_map returns %d\n", result);
 
 		return ERR_PTR(result);
 	}
@@ -516,7 +518,7 @@ int ib_fmr_pool_unmap(struct ib_pool_fmr *fmr)
 
 #ifdef DEBUG
 	if (fmr->ref_count < 0)
-		printk(KERN_WARNING "FMR %p has ref count %d < 0",
+		printk(KERN_WARNING PFX "FMR %p has ref count %d < 0",
 		       fmr, fmr->ref_count);
 #endif
 
-- 
cgit v0.10.2


From 6ed89b9574776d4178f1ad754d20e4f1e5a4b6c8 Mon Sep 17 00:00:00 2001
From: Ralph Campbell <ralphc@pathscale.com>
Date: Thu, 3 May 2007 12:40:51 -0700
Subject: IB/ipath: Fix two more spin lock problems

Fix a missing unlock in ipath_rc_rcv_resp() and remove an extra unlock
from ipath_rc_rcv_error().

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c
index 9e68c91..f4d729d 100644
--- a/drivers/infiniband/hw/ipath/ipath_rc.c
+++ b/drivers/infiniband/hw/ipath/ipath_rc.c
@@ -1257,6 +1257,7 @@ ack_err:
 	wc.dlid_path_bits = 0;
 	wc.port_num = 0;
 	ipath_sqerror_qp(qp, &wc);
+	spin_unlock_irqrestore(&qp->s_lock, flags);
 bail:
 	return;
 }
@@ -1436,7 +1437,6 @@ static inline int ipath_rc_rcv_error(struct ipath_ibdev *dev,
 		break;
 	}
 	qp->r_nak_state = 0;
-	spin_unlock_irq(&qp->s_lock);
 	tasklet_hi_schedule(&qp->s_task);
 
 unlock_done:
-- 
cgit v0.10.2


From 154257f3626ea6dd96781fac0896c3f27fe2b0a1 Mon Sep 17 00:00:00 2001
From: Ralph Campbell <ralph.campbell@qlogic.com>
Date: Thu, 3 May 2007 12:43:03 -0700
Subject: IB/ipath: Fix a race condition when generating ACKs

Fix a problem where simple ACKs can be sent ahead of RDMA read
responses thus implicitly NAKing the RDMA read.

Signed-off-by: Ralph Campbell <ralph.cambpell@qlogic.com>
Signed-off-by: Robert Walsh <robert.walsh@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c
index f4d729d..1915771 100644
--- a/drivers/infiniband/hw/ipath/ipath_rc.c
+++ b/drivers/infiniband/hw/ipath/ipath_rc.c
@@ -98,13 +98,21 @@ static int ipath_make_rc_ack(struct ipath_qp *qp,
 	case OP(RDMA_READ_RESPONSE_LAST):
 	case OP(RDMA_READ_RESPONSE_ONLY):
 	case OP(ATOMIC_ACKNOWLEDGE):
-		qp->s_ack_state = OP(ACKNOWLEDGE);
+		/*
+		 * We can increment the tail pointer now that the last
+		 * response has been sent instead of only being
+		 * constructed.
+		 */
+		if (++qp->s_tail_ack_queue > IPATH_MAX_RDMA_ATOMIC)
+			qp->s_tail_ack_queue = 0;
 		/* FALLTHROUGH */
+	case OP(SEND_ONLY):
 	case OP(ACKNOWLEDGE):
 		/* Check for no next entry in the queue. */
 		if (qp->r_head_ack_queue == qp->s_tail_ack_queue) {
 			if (qp->s_flags & IPATH_S_ACK_PENDING)
 				goto normal;
+			qp->s_ack_state = OP(ACKNOWLEDGE);
 			goto bail;
 		}
 
@@ -117,12 +125,8 @@ static int ipath_make_rc_ack(struct ipath_qp *qp,
 			if (len > pmtu) {
 				len = pmtu;
 				qp->s_ack_state = OP(RDMA_READ_RESPONSE_FIRST);
-			} else {
+			} else
 				qp->s_ack_state = OP(RDMA_READ_RESPONSE_ONLY);
-				if (++qp->s_tail_ack_queue >
-				    IPATH_MAX_RDMA_ATOMIC)
-					qp->s_tail_ack_queue = 0;
-			}
 			ohdr->u.aeth = ipath_compute_aeth(qp);
 			hwords++;
 			qp->s_ack_rdma_psn = e->psn;
@@ -139,8 +143,6 @@ static int ipath_make_rc_ack(struct ipath_qp *qp,
 				cpu_to_be32(e->atomic_data);
 			hwords += sizeof(ohdr->u.at) / sizeof(u32);
 			bth2 = e->psn;
-			if (++qp->s_tail_ack_queue > IPATH_MAX_RDMA_ATOMIC)
-				qp->s_tail_ack_queue = 0;
 		}
 		bth0 = qp->s_ack_state << 24;
 		break;
@@ -156,8 +158,6 @@ static int ipath_make_rc_ack(struct ipath_qp *qp,
 			ohdr->u.aeth = ipath_compute_aeth(qp);
 			hwords++;
 			qp->s_ack_state = OP(RDMA_READ_RESPONSE_LAST);
-			if (++qp->s_tail_ack_queue > IPATH_MAX_RDMA_ATOMIC)
-				qp->s_tail_ack_queue = 0;
 		}
 		bth0 = qp->s_ack_state << 24;
 		bth2 = qp->s_ack_rdma_psn++ & IPATH_PSN_MASK;
@@ -171,7 +171,7 @@ static int ipath_make_rc_ack(struct ipath_qp *qp,
 		 * the ACK before setting s_ack_state to ACKNOWLEDGE
 		 * (see above).
 		 */
-		qp->s_ack_state = OP(ATOMIC_ACKNOWLEDGE);
+		qp->s_ack_state = OP(SEND_ONLY);
 		qp->s_flags &= ~IPATH_S_ACK_PENDING;
 		qp->s_cur_sge = NULL;
 		if (qp->s_nak_state)
@@ -223,7 +223,7 @@ int ipath_make_rc_req(struct ipath_qp *qp,
 	/* Sending responses has higher priority over sending requests. */
 	if ((qp->r_head_ack_queue != qp->s_tail_ack_queue ||
 	     (qp->s_flags & IPATH_S_ACK_PENDING) ||
-	     qp->s_ack_state != IB_OPCODE_RC_ACKNOWLEDGE) &&
+	     qp->s_ack_state != OP(ACKNOWLEDGE)) &&
 	    ipath_make_rc_ack(qp, ohdr, pmtu, bth0p, bth2p))
 		goto done;
 
@@ -585,7 +585,9 @@ static void send_rc_ack(struct ipath_qp *qp)
 	unsigned long flags;
 
 	/* Don't send ACK or NAK if a RDMA read or atomic is pending. */
-	if (qp->r_head_ack_queue != qp->s_tail_ack_queue)
+	if (qp->r_head_ack_queue != qp->s_tail_ack_queue ||
+	    (qp->s_flags & IPATH_S_ACK_PENDING) ||
+	    qp->s_ack_state != OP(ACKNOWLEDGE))
 		goto queue_ack;
 
 	/* Construct the header. */
-- 
cgit v0.10.2


From f4fd0b224d60044d2da5ca02f8f2b5150c1d8731 Mon Sep 17 00:00:00 2001
From: "Michael S. Tsirkin" <mst@dev.mellanox.co.il>
Date: Thu, 3 May 2007 13:48:47 +0300
Subject: IB: Add CQ comp_vector support

Add a num_comp_vectors member to struct ib_device and extend
ib_create_cq() to pass in a comp_vector parameter -- this parallels
the userspace libibverbs API.  Update all hardware drivers to set
num_comp_vectors to 1 and have all ULPs pass 0 for the comp_vector
value.  Pass the value of num_comp_vectors to userspace rather than
hard-coding a value of 1.

We want multiple CQ event vector support (via MSI-X or similar for
adapters that can generate multiple interrupts), but it's not clear
how many vectors we want, or how we want to deal with policy issues
such as how to decide which vector to use or how to set up interrupt
affinity.  This patch is useful for experimenting, since no core
changes will be necessary when updating a driver to support multiple
vectors, and we know that we want to make at least these changes
anyway.

Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index 6edfecf..85ccf13 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -2771,7 +2771,7 @@ static int ib_mad_port_open(struct ib_device *device,
 	cq_size = (IB_MAD_QP_SEND_SIZE + IB_MAD_QP_RECV_SIZE) * 2;
 	port_priv->cq = ib_create_cq(port_priv->device,
 				     ib_mad_thread_completion_handler,
-				     NULL, port_priv, cq_size);
+				     NULL, port_priv, cq_size, 0);
 	if (IS_ERR(port_priv->cq)) {
 		printk(KERN_ERR PFX "Couldn't create ib_mad CQ\n");
 		ret = PTR_ERR(port_priv->cq);
diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c
index 4fd75af..bab6676 100644
--- a/drivers/infiniband/core/uverbs_cmd.c
+++ b/drivers/infiniband/core/uverbs_cmd.c
@@ -802,6 +802,7 @@ ssize_t ib_uverbs_create_cq(struct ib_uverbs_file *file,
 	INIT_LIST_HEAD(&obj->async_list);
 
 	cq = file->device->ib_dev->create_cq(file->device->ib_dev, cmd.cqe,
+					     cmd.comp_vector,
 					     file->ucontext, &udata);
 	if (IS_ERR(cq)) {
 		ret = PTR_ERR(cq);
diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c
index f8bc822..d44e547 100644
--- a/drivers/infiniband/core/uverbs_main.c
+++ b/drivers/infiniband/core/uverbs_main.c
@@ -752,7 +752,7 @@ static void ib_uverbs_add_one(struct ib_device *device)
 	spin_unlock(&map_lock);
 
 	uverbs_dev->ib_dev           = device;
-	uverbs_dev->num_comp_vectors = 1;
+	uverbs_dev->num_comp_vectors = device->num_comp_vectors;
 
 	uverbs_dev->dev = cdev_alloc();
 	if (!uverbs_dev->dev)
diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
index ccdf93d..86ed8af 100644
--- a/drivers/infiniband/core/verbs.c
+++ b/drivers/infiniband/core/verbs.c
@@ -609,11 +609,11 @@ EXPORT_SYMBOL(ib_destroy_qp);
 struct ib_cq *ib_create_cq(struct ib_device *device,
 			   ib_comp_handler comp_handler,
 			   void (*event_handler)(struct ib_event *, void *),
-			   void *cq_context, int cqe)
+			   void *cq_context, int cqe, int comp_vector)
 {
 	struct ib_cq *cq;
 
-	cq = device->create_cq(device, cqe, NULL, NULL);
+	cq = device->create_cq(device, cqe, comp_vector, NULL, NULL);
 
 	if (!IS_ERR(cq)) {
 		cq->device        = device;
diff --git a/drivers/infiniband/hw/amso1100/c2_provider.c b/drivers/infiniband/hw/amso1100/c2_provider.c
index 607c09b..1091662 100644
--- a/drivers/infiniband/hw/amso1100/c2_provider.c
+++ b/drivers/infiniband/hw/amso1100/c2_provider.c
@@ -290,7 +290,7 @@ static int c2_destroy_qp(struct ib_qp *ib_qp)
 	return 0;
 }
 
-static struct ib_cq *c2_create_cq(struct ib_device *ibdev, int entries,
+static struct ib_cq *c2_create_cq(struct ib_device *ibdev, int entries, int vector,
 				  struct ib_ucontext *context,
 				  struct ib_udata *udata)
 {
@@ -795,6 +795,7 @@ int c2_register_device(struct c2_dev *dev)
 	memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid));
 	memcpy(&dev->ibdev.node_guid, dev->pseudo_netdev->dev_addr, 6);
 	dev->ibdev.phys_port_cnt = 1;
+	dev->ibdev.num_comp_vectors = 1;
 	dev->ibdev.dma_device = &dev->pcidev->dev;
 	dev->ibdev.query_device = c2_query_device;
 	dev->ibdev.query_port = c2_query_port;
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 93038c0..78a495f 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -139,7 +139,7 @@ static int iwch_destroy_cq(struct ib_cq *ib_cq)
 	return 0;
 }
 
-static struct ib_cq *iwch_create_cq(struct ib_device *ibdev, int entries,
+static struct ib_cq *iwch_create_cq(struct ib_device *ibdev, int entries, int vector,
 			     struct ib_ucontext *ib_context,
 			     struct ib_udata *udata)
 {
@@ -1110,6 +1110,7 @@ int iwch_register_device(struct iwch_dev *dev)
 	dev->ibdev.node_type = RDMA_NODE_RNIC;
 	memcpy(dev->ibdev.node_desc, IWCH_NODE_DESC, sizeof(IWCH_NODE_DESC));
 	dev->ibdev.phys_port_cnt = dev->rdev.port_info.nports;
+	dev->ibdev.num_comp_vectors = 1;
 	dev->ibdev.dma_device = &(dev->rdev.rnic_info.pdev->dev);
 	dev->ibdev.query_device = iwch_query_device;
 	dev->ibdev.query_port = iwch_query_port;
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c b/drivers/infiniband/hw/ehca/ehca_cq.c
index e2cdc1a..67f0670 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -113,7 +113,7 @@ struct ehca_qp* ehca_cq_get_qp(struct ehca_cq *cq, int real_qp_num)
 	return ret;
 }
 
-struct ib_cq *ehca_create_cq(struct ib_device *device, int cqe,
+struct ib_cq *ehca_create_cq(struct ib_device *device, int cqe, int comp_vector,
 			     struct ib_ucontext *context,
 			     struct ib_udata *udata)
 {
diff --git a/drivers/infiniband/hw/ehca/ehca_iverbs.h b/drivers/infiniband/hw/ehca/ehca_iverbs.h
index 95fd59f..aff96ac 100644
--- a/drivers/infiniband/hw/ehca/ehca_iverbs.h
+++ b/drivers/infiniband/hw/ehca/ehca_iverbs.h
@@ -123,7 +123,7 @@ int ehca_destroy_eq(struct ehca_shca *shca, struct ehca_eq *eq);
 void *ehca_poll_eq(struct ehca_shca *shca, struct ehca_eq *eq);
 
 
-struct ib_cq *ehca_create_cq(struct ib_device *device, int cqe,
+struct ib_cq *ehca_create_cq(struct ib_device *device, int cqe, int comp_vector,
 			     struct ib_ucontext *context,
 			     struct ib_udata *udata);
 
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c b/drivers/infiniband/hw/ehca/ehca_main.c
index 3b23d677..77bb36b 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -313,6 +313,7 @@ int ehca_init_device(struct ehca_shca *shca)
 
 	shca->ib_device.node_type           = RDMA_NODE_IB_CA;
 	shca->ib_device.phys_port_cnt       = shca->num_ports;
+	shca->ib_device.num_comp_vectors    = 1;
 	shca->ib_device.dma_device          = &shca->ibmebus_dev->ofdev.dev;
 	shca->ib_device.query_device        = ehca_query_device;
 	shca->ib_device.query_port          = ehca_query_port;
@@ -375,7 +376,7 @@ static int ehca_create_aqp1(struct ehca_shca *shca, u32 port)
 		return -EPERM;
 	}
 
-	ibcq = ib_create_cq(&shca->ib_device, NULL, NULL, (void*)(-1), 10);
+	ibcq = ib_create_cq(&shca->ib_device, NULL, NULL, (void*)(-1), 10, 0);
 	if (IS_ERR(ibcq)) {
 		ehca_err(&shca->ib_device, "Cannot create AQP1 CQ.");
 		return PTR_ERR(ibcq);
diff --git a/drivers/infiniband/hw/ipath/ipath_cq.c b/drivers/infiniband/hw/ipath/ipath_cq.c
index 4715f89..00d3eb9 100644
--- a/drivers/infiniband/hw/ipath/ipath_cq.c
+++ b/drivers/infiniband/hw/ipath/ipath_cq.c
@@ -204,7 +204,7 @@ static void send_complete(unsigned long data)
  *
  * Called by ib_create_cq() in the generic verbs code.
  */
-struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries,
+struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries, int comp_vector,
 			      struct ib_ucontext *context,
 			      struct ib_udata *udata)
 {
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
index b676ea8..12933e7 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c
@@ -1561,6 +1561,7 @@ int ipath_register_ib_device(struct ipath_devdata *dd)
 		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
 	dev->node_type = RDMA_NODE_IB_CA;
 	dev->phys_port_cnt = 1;
+	dev->num_comp_vectors = 1;
 	dev->dma_device = &dd->pcidev->dev;
 	dev->query_device = ipath_query_device;
 	dev->modify_device = ipath_modify_device;
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
index ac66c00..2d734fb 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h
@@ -735,7 +735,7 @@ int ipath_destroy_srq(struct ib_srq *ibsrq);
 
 int ipath_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry);
 
-struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries,
+struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries, int comp_vector,
 			      struct ib_ucontext *context,
 			      struct ib_udata *udata);
 
diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c
index 47e6fd46d..1c05486 100644
--- a/drivers/infiniband/hw/mthca/mthca_provider.c
+++ b/drivers/infiniband/hw/mthca/mthca_provider.c
@@ -663,6 +663,7 @@ static int mthca_destroy_qp(struct ib_qp *qp)
 }
 
 static struct ib_cq *mthca_create_cq(struct ib_device *ibdev, int entries,
+				     int comp_vector,
 				     struct ib_ucontext *context,
 				     struct ib_udata *udata)
 {
@@ -1292,6 +1293,7 @@ int mthca_register_device(struct mthca_dev *dev)
 		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST);
 	dev->ib_dev.node_type            = RDMA_NODE_IB_CA;
 	dev->ib_dev.phys_port_cnt        = dev->limits.num_ports;
+	dev->ib_dev.num_comp_vectors     = 1;
 	dev->ib_dev.dma_device           = &dev->pdev->dev;
 	dev->ib_dev.query_device         = mthca_query_device;
 	dev->ib_dev.query_port           = mthca_query_port;
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
index 1e27930..b8089a0 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
@@ -793,7 +793,7 @@ static int ipoib_cm_tx_init(struct ipoib_cm_tx *p, u32 qpn,
 	}
 
 	p->cq = ib_create_cq(priv->ca, ipoib_cm_tx_completion, NULL, p,
-			     ipoib_sendq_size + 1);
+			     ipoib_sendq_size + 1, 0);
 	if (IS_ERR(p->cq)) {
 		ret = PTR_ERR(p->cq);
 		ipoib_warn(priv, "failed to allocate tx cq: %d\n", ret);
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c
index 7f3ec20..5c3c6a4 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c
@@ -187,7 +187,7 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 	if (!ret)
 		size += ipoib_recvq_size;
 
-	priv->cq = ib_create_cq(priv->ca, ipoib_ib_completion, NULL, dev, size);
+	priv->cq = ib_create_cq(priv->ca, ipoib_ib_completion, NULL, dev, size, 0);
 	if (IS_ERR(priv->cq)) {
 		printk(KERN_WARNING "%s: failed to create CQ\n", ca->name);
 		goto out_free_mr;
diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c
index 1fc9674..89d6008 100644
--- a/drivers/infiniband/ulp/iser/iser_verbs.c
+++ b/drivers/infiniband/ulp/iser/iser_verbs.c
@@ -76,7 +76,7 @@ static int iser_create_device_ib_res(struct iser_device *device)
 				  iser_cq_callback,
 				  iser_cq_event_callback,
 				  (void *)device,
-				  ISER_MAX_CQ_LEN);
+				  ISER_MAX_CQ_LEN, 0);
 	if (IS_ERR(device->cq))
 		goto cq_err;
 
diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c
index 3468ae1..39bf057 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.c
+++ b/drivers/infiniband/ulp/srp/ib_srp.c
@@ -197,7 +197,7 @@ static int srp_create_target_ib(struct srp_target_port *target)
 		return -ENOMEM;
 
 	target->cq = ib_create_cq(target->srp_host->dev->dev, srp_completion,
-				  NULL, target, SRP_CQ_SIZE);
+				  NULL, target, SRP_CQ_SIZE, 0);
 	if (IS_ERR(target->cq)) {
 		ret = PTR_ERR(target->cq);
 		goto out;
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 765589f..17cc309 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -912,6 +912,8 @@ struct ib_device {
 
 	u32                           flags;
 
+	int			      num_comp_vectors;
+
 	struct iw_cm_verbs	     *iwcm;
 
 	int		           (*query_device)(struct ib_device *device,
@@ -978,6 +980,7 @@ struct ib_device {
 						struct ib_recv_wr *recv_wr,
 						struct ib_recv_wr **bad_recv_wr);
 	struct ib_cq *             (*create_cq)(struct ib_device *device, int cqe,
+						int comp_vector,
 						struct ib_ucontext *context,
 						struct ib_udata *udata);
 	int                        (*destroy_cq)(struct ib_cq *cq);
@@ -1358,13 +1361,15 @@ static inline int ib_post_recv(struct ib_qp *qp,
  * @cq_context: Context associated with the CQ returned to the user via
  *   the associated completion and event handlers.
  * @cqe: The minimum size of the CQ.
+ * @comp_vector - Completion vector used to signal completion events.
+ *     Must be >= 0 and < context->num_comp_vectors.
  *
  * Users can examine the cq structure to determine the actual CQ size.
  */
 struct ib_cq *ib_create_cq(struct ib_device *device,
 			   ib_comp_handler comp_handler,
 			   void (*event_handler)(struct ib_event *, void *),
-			   void *cq_context, int cqe);
+			   void *cq_context, int cqe, int comp_vector);
 
 /**
  * ib_resize_cq - Modifies the capacity of the CQ.
-- 
cgit v0.10.2


From ed23a72778f3dbd465e55b06fe31629e7e1dd2f3 Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Sun, 6 May 2007 21:02:48 -0700
Subject: IB: Return "maybe missed event" hint from ib_req_notify_cq()

The semantics defined by the InfiniBand specification say that
completion events are only generated when a completions is added to a
completion queue (CQ) after completion notification is requested.  In
other words, this means that the following race is possible:

	while (CQ is not empty)
		ib_poll_cq(CQ);
	// new completion is added after while loop is exited
	ib_req_notify_cq(CQ);
	// no event is generated for the existing completion

To close this race, the IB spec recommends doing another poll of the
CQ after requesting notification.

However, it is not always possible to arrange code this way (for
example, we have found that NAPI for IPoIB cannot poll after
requesting notification).  Also, some hardware (eg Mellanox HCAs)
actually will generate an event for completions added before the call
to ib_req_notify_cq() -- which is allowed by the spec, since there's
no way for any upper-layer consumer to know exactly when a completion
was really added -- so the extra poll of the CQ is just a waste.

Motivated by this, we add a new flag "IB_CQ_REPORT_MISSED_EVENTS" for
ib_req_notify_cq() so that it can return a hint about whether the a
completion may have been added before the request for notification.
The return value of ib_req_notify_cq() is extended so:

	 < 0	means an error occurred while requesting notification
	== 0	means notification was requested successfully, and if
		IB_CQ_REPORT_MISSED_EVENTS was passed in, then no
		events were missed and it is safe to wait for another
		event.
	 > 0	is only returned if IB_CQ_REPORT_MISSED_EVENTS was
		passed in.  It means that the consumer must poll the
		CQ again to make sure it is empty to avoid the race
		described above.

We add a flag to enable this behavior rather than turning it on
unconditionally, because checking for missed events may incur
significant overhead for some low-level drivers, and consumers that
don't care about the results of this test shouldn't be forced to pay
for the test.

Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/amso1100/c2.h b/drivers/infiniband/hw/amso1100/c2.h
index 04a9db5..fa58200 100644
--- a/drivers/infiniband/hw/amso1100/c2.h
+++ b/drivers/infiniband/hw/amso1100/c2.h
@@ -519,7 +519,7 @@ extern void c2_free_cq(struct c2_dev *c2dev, struct c2_cq *cq);
 extern void c2_cq_event(struct c2_dev *c2dev, u32 mq_index);
 extern void c2_cq_clean(struct c2_dev *c2dev, struct c2_qp *qp, u32 mq_index);
 extern int c2_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry);
-extern int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify);
+extern int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags);
 
 /* CM */
 extern int c2_llp_connect(struct iw_cm_id *cm_id,
diff --git a/drivers/infiniband/hw/amso1100/c2_cq.c b/drivers/infiniband/hw/amso1100/c2_cq.c
index 5175c99..d2b3366 100644
--- a/drivers/infiniband/hw/amso1100/c2_cq.c
+++ b/drivers/infiniband/hw/amso1100/c2_cq.c
@@ -217,17 +217,19 @@ int c2_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
 	return npolled;
 }
 
-int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
+int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
 {
 	struct c2_mq_shared __iomem *shared;
 	struct c2_cq *cq;
+	unsigned long flags;
+	int ret = 0;
 
 	cq = to_c2cq(ibcq);
 	shared = cq->mq.peer;
 
-	if (notify == IB_CQ_NEXT_COMP)
+	if ((notify_flags & IB_CQ_SOLICITED_MASK) == IB_CQ_NEXT_COMP)
 		writeb(C2_CQ_NOTIFICATION_TYPE_NEXT, &shared->notification_type);
-	else if (notify == IB_CQ_SOLICITED)
+	else if ((notify_flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED)
 		writeb(C2_CQ_NOTIFICATION_TYPE_NEXT_SE, &shared->notification_type);
 	else
 		return -EINVAL;
@@ -241,7 +243,13 @@ int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
 	 */
 	readb(&shared->armed);
 
-	return 0;
+	if (notify_flags & IB_CQ_REPORT_MISSED_EVENTS) {
+		spin_lock_irqsave(&cq->lock, flags);
+		ret = !c2_mq_empty(&cq->mq);
+		spin_unlock_irqrestore(&cq->lock, flags);
+	}
+
+	return ret;
 }
 
 static void c2_free_cq_buf(struct c2_dev *c2dev, struct c2_mq *mq)
diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c
index f5e9aee..76049af 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.c
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c
@@ -114,7 +114,10 @@ int cxio_hal_cq_op(struct cxio_rdev *rdev_p, struct t3_cq *cq,
 				return -EIO;
 			}
 		}
+
+		return 1;
 	}
+
 	return 0;
 }
 
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 78a495f..a891493 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -292,7 +292,7 @@ static int iwch_resize_cq(struct ib_cq *cq, int cqe, struct ib_udata *udata)
 #endif
 }
 
-static int iwch_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
+static int iwch_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
 {
 	struct iwch_dev *rhp;
 	struct iwch_cq *chp;
@@ -303,7 +303,7 @@ static int iwch_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
 
 	chp = to_iwch_cq(ibcq);
 	rhp = chp->rhp;
-	if (notify == IB_CQ_SOLICITED)
+	if ((flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED)
 		cq_op = CQ_ARM_SE;
 	else
 		cq_op = CQ_ARM_AN;
@@ -317,9 +317,11 @@ static int iwch_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
 	PDBG("%s rptr 0x%x\n", __FUNCTION__, chp->cq.rptr);
 	err = cxio_hal_cq_op(&rhp->rdev, &chp->cq, cq_op, 0);
 	spin_unlock_irqrestore(&chp->lock, flag);
-	if (err)
+	if (err < 0)
 		printk(KERN_ERR MOD "Error %d rearming CQID 0x%x\n", err,
 		       chp->cq.cqid);
+	if (err > 0 && !(flags & IB_CQ_REPORT_MISSED_EVENTS))
+		err = 0;
 	return err;
 }
 
diff --git a/drivers/infiniband/hw/ehca/ehca_iverbs.h b/drivers/infiniband/hw/ehca/ehca_iverbs.h
index aff96ac..e14b029 100644
--- a/drivers/infiniband/hw/ehca/ehca_iverbs.h
+++ b/drivers/infiniband/hw/ehca/ehca_iverbs.h
@@ -135,7 +135,7 @@ int ehca_poll_cq(struct ib_cq *cq, int num_entries, struct ib_wc *wc);
 
 int ehca_peek_cq(struct ib_cq *cq, int wc_cnt);
 
-int ehca_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify cq_notify);
+int ehca_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify_flags notify_flags);
 
 struct ib_qp *ehca_create_qp(struct ib_pd *pd,
 			     struct ib_qp_init_attr *init_attr,
diff --git a/drivers/infiniband/hw/ehca/ehca_reqs.c b/drivers/infiniband/hw/ehca/ehca_reqs.c
index 08d3f89..caec9de 100644
--- a/drivers/infiniband/hw/ehca/ehca_reqs.c
+++ b/drivers/infiniband/hw/ehca/ehca_reqs.c
@@ -634,11 +634,13 @@ poll_cq_exit0:
 	return ret;
 }
 
-int ehca_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify cq_notify)
+int ehca_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify_flags notify_flags)
 {
 	struct ehca_cq *my_cq = container_of(cq, struct ehca_cq, ib_cq);
+	unsigned long spl_flags;
+	int ret = 0;
 
-	switch (cq_notify) {
+	switch (notify_flags & IB_CQ_SOLICITED_MASK) {
 	case IB_CQ_SOLICITED:
 		hipz_set_cqx_n0(my_cq, 1);
 		break;
@@ -649,5 +651,11 @@ int ehca_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify cq_notify)
 		return -EINVAL;
 	}
 
-	return 0;
+	if (notify_flags & IB_CQ_REPORT_MISSED_EVENTS) {
+		spin_lock_irqsave(&my_cq->spinlock, spl_flags);
+		ret = ipz_qeit_is_valid(&my_cq->ipz_queue);
+		spin_unlock_irqrestore(&my_cq->spinlock, spl_flags);
+	}
+
+	return ret;
 }
diff --git a/drivers/infiniband/hw/ehca/ipz_pt_fn.h b/drivers/infiniband/hw/ehca/ipz_pt_fn.h
index 8199c45..57f141a 100644
--- a/drivers/infiniband/hw/ehca/ipz_pt_fn.h
+++ b/drivers/infiniband/hw/ehca/ipz_pt_fn.h
@@ -140,6 +140,14 @@ static inline void *ipz_qeit_get_inc_valid(struct ipz_queue *queue)
 	return cqe;
 }
 
+static inline int ipz_qeit_is_valid(struct ipz_queue *queue)
+{
+	struct ehca_cqe *cqe = ipz_qeit_get(queue);
+	u32 cqe_flags = cqe->cqe_flags;
+
+	return cqe_flags >> 7 == (queue->toggle_state & 1);
+}
+
 /*
  * returns and resets Queue Entry iterator
  * returns address (kv) of first Queue Entry
diff --git a/drivers/infiniband/hw/ipath/ipath_cq.c b/drivers/infiniband/hw/ipath/ipath_cq.c
index 00d3eb9..3e9241b 100644
--- a/drivers/infiniband/hw/ipath/ipath_cq.c
+++ b/drivers/infiniband/hw/ipath/ipath_cq.c
@@ -334,17 +334,18 @@ int ipath_destroy_cq(struct ib_cq *ibcq)
 /**
  * ipath_req_notify_cq - change the notification type for a completion queue
  * @ibcq: the completion queue
- * @notify: the type of notification to request
+ * @notify_flags: the type of notification to request
  *
  * Returns 0 for success.
  *
  * This may be called from interrupt context.  Also called by
  * ib_req_notify_cq() in the generic verbs code.
  */
-int ipath_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
+int ipath_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
 {
 	struct ipath_cq *cq = to_icq(ibcq);
 	unsigned long flags;
+	int ret = 0;
 
 	spin_lock_irqsave(&cq->lock, flags);
 	/*
@@ -352,9 +353,15 @@ int ipath_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
 	 * any other transitions (see C11-31 and C11-32 in ch. 11.4.2.2).
 	 */
 	if (cq->notify != IB_CQ_NEXT_COMP)
-		cq->notify = notify;
+		cq->notify = notify_flags & IB_CQ_SOLICITED_MASK;
+
+	if ((notify_flags & IB_CQ_REPORT_MISSED_EVENTS) &&
+	    cq->queue->head != cq->queue->tail)
+		ret = 1;
+
 	spin_unlock_irqrestore(&cq->lock, flags);
-	return 0;
+
+	return ret;
 }
 
 /**
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
index 2d734fb..7064fc2 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h
@@ -741,7 +741,7 @@ struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries, int comp_vec
 
 int ipath_destroy_cq(struct ib_cq *ibcq);
 
-int ipath_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify notify);
+int ipath_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags);
 
 int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata);
 
diff --git a/drivers/infiniband/hw/mthca/mthca_cq.c b/drivers/infiniband/hw/mthca/mthca_cq.c
index efd79ef..cf0868f 100644
--- a/drivers/infiniband/hw/mthca/mthca_cq.c
+++ b/drivers/infiniband/hw/mthca/mthca_cq.c
@@ -726,11 +726,12 @@ repoll:
 	return err == 0 || err == -EAGAIN ? npolled : err;
 }
 
-int mthca_tavor_arm_cq(struct ib_cq *cq, enum ib_cq_notify notify)
+int mthca_tavor_arm_cq(struct ib_cq *cq, enum ib_cq_notify_flags flags)
 {
 	__be32 doorbell[2];
 
-	doorbell[0] = cpu_to_be32((notify == IB_CQ_SOLICITED ?
+	doorbell[0] = cpu_to_be32(((flags & IB_CQ_SOLICITED_MASK) ==
+				   IB_CQ_SOLICITED ?
 				   MTHCA_TAVOR_CQ_DB_REQ_NOT_SOL :
 				   MTHCA_TAVOR_CQ_DB_REQ_NOT)      |
 				  to_mcq(cq)->cqn);
@@ -743,7 +744,7 @@ int mthca_tavor_arm_cq(struct ib_cq *cq, enum ib_cq_notify notify)
 	return 0;
 }
 
-int mthca_arbel_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
+int mthca_arbel_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
 {
 	struct mthca_cq *cq = to_mcq(ibcq);
 	__be32 doorbell[2];
@@ -755,7 +756,8 @@ int mthca_arbel_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
 
 	doorbell[0] = ci;
 	doorbell[1] = cpu_to_be32((cq->cqn << 8) | (2 << 5) | (sn << 3) |
-				  (notify == IB_CQ_SOLICITED ? 1 : 2));
+				  ((flags & IB_CQ_SOLICITED_MASK) ==
+				   IB_CQ_SOLICITED ? 1 : 2));
 
 	mthca_write_db_rec(doorbell, cq->arm_db);
 
@@ -766,7 +768,7 @@ int mthca_arbel_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
 	wmb();
 
 	doorbell[0] = cpu_to_be32((sn << 28)                       |
-				  (notify == IB_CQ_SOLICITED ?
+				  ((flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED ?
 				   MTHCA_ARBEL_CQ_DB_REQ_NOT_SOL :
 				   MTHCA_ARBEL_CQ_DB_REQ_NOT)      |
 				  cq->cqn);
diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h
index b7e42ef..9bae3cc 100644
--- a/drivers/infiniband/hw/mthca/mthca_dev.h
+++ b/drivers/infiniband/hw/mthca/mthca_dev.h
@@ -495,8 +495,8 @@ void mthca_unmap_eq_icm(struct mthca_dev *dev);
 
 int mthca_poll_cq(struct ib_cq *ibcq, int num_entries,
 		  struct ib_wc *entry);
-int mthca_tavor_arm_cq(struct ib_cq *cq, enum ib_cq_notify notify);
-int mthca_arbel_arm_cq(struct ib_cq *cq, enum ib_cq_notify notify);
+int mthca_tavor_arm_cq(struct ib_cq *cq, enum ib_cq_notify_flags flags);
+int mthca_arbel_arm_cq(struct ib_cq *cq, enum ib_cq_notify_flags flags);
 int mthca_init_cq(struct mthca_dev *dev, int nent,
 		  struct mthca_ucontext *ctx, u32 pdn,
 		  struct mthca_cq *cq);
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 17cc309..5342ac6 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -431,9 +431,11 @@ struct ib_wc {
 	u8			port_num;	/* valid only for DR SMPs on switches */
 };
 
-enum ib_cq_notify {
-	IB_CQ_SOLICITED,
-	IB_CQ_NEXT_COMP
+enum ib_cq_notify_flags {
+	IB_CQ_SOLICITED			= 1 << 0,
+	IB_CQ_NEXT_COMP			= 1 << 1,
+	IB_CQ_SOLICITED_MASK		= IB_CQ_SOLICITED | IB_CQ_NEXT_COMP,
+	IB_CQ_REPORT_MISSED_EVENTS	= 1 << 2,
 };
 
 enum ib_srq_attr_mask {
@@ -990,7 +992,7 @@ struct ib_device {
 					      struct ib_wc *wc);
 	int                        (*peek_cq)(struct ib_cq *cq, int wc_cnt);
 	int                        (*req_notify_cq)(struct ib_cq *cq,
-						    enum ib_cq_notify cq_notify);
+						    enum ib_cq_notify_flags flags);
 	int                        (*req_ncomp_notif)(struct ib_cq *cq,
 						      int wc_cnt);
 	struct ib_mr *             (*get_dma_mr)(struct ib_pd *pd,
@@ -1419,14 +1421,34 @@ int ib_peek_cq(struct ib_cq *cq, int wc_cnt);
 /**
  * ib_req_notify_cq - Request completion notification on a CQ.
  * @cq: The CQ to generate an event for.
- * @cq_notify: If set to %IB_CQ_SOLICITED, completion notification will
- *   occur on the next solicited event. If set to %IB_CQ_NEXT_COMP,
- *   notification will occur on the next completion.
+ * @flags:
+ *   Must contain exactly one of %IB_CQ_SOLICITED or %IB_CQ_NEXT_COMP
+ *   to request an event on the next solicited event or next work
+ *   completion at any type, respectively. %IB_CQ_REPORT_MISSED_EVENTS
+ *   may also be |ed in to request a hint about missed events, as
+ *   described below.
+ *
+ * Return Value:
+ *    < 0 means an error occurred while requesting notification
+ *   == 0 means notification was requested successfully, and if
+ *        IB_CQ_REPORT_MISSED_EVENTS was passed in, then no events
+ *        were missed and it is safe to wait for another event.  In
+ *        this case is it guaranteed that any work completions added
+ *        to the CQ since the last CQ poll will trigger a completion
+ *        notification event.
+ *    > 0 is only returned if IB_CQ_REPORT_MISSED_EVENTS was passed
+ *        in.  It means that the consumer must poll the CQ again to
+ *        make sure it is empty to avoid missing an event because of a
+ *        race between requesting notification and an entry being
+ *        added to the CQ.  This return value means it is possible
+ *        (but not guaranteed) that a work completion has been added
+ *        to the CQ since the last poll without triggering a
+ *        completion notification event.
  */
 static inline int ib_req_notify_cq(struct ib_cq *cq,
-				   enum ib_cq_notify cq_notify)
+				   enum ib_cq_notify_flags flags)
 {
-	return cq->device->req_notify_cq(cq, cq_notify);
+	return cq->device->req_notify_cq(cq, flags);
 }
 
 /**
-- 
cgit v0.10.2


From 8d1cc86a6278687efbab7b8c294ab01efe4d4231 Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Sun, 6 May 2007 21:05:32 -0700
Subject: IPoIB: Convert to NAPI

Convert the IP-over-InfiniBand network device driver over to using
NAPI to handle completions for the main CQ.  This covers all receives
as well as datagram mode sends; send completions for connected mode
connections are still handled from interrupt context.

Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h
index fd55826..15867af 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib.h
+++ b/drivers/infiniband/ulp/ipoib/ipoib.h
@@ -311,6 +311,7 @@ extern struct workqueue_struct *ipoib_workqueue;
 
 /* functions */
 
+int ipoib_poll(struct net_device *dev, int *budget);
 void ipoib_ib_completion(struct ib_cq *cq, void *dev_ptr);
 
 struct ipoib_ah *ipoib_create_ah(struct net_device *dev,
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
index b8089a0..785bc85 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
@@ -416,7 +416,7 @@ void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 	skb->dev = dev;
 	/* XXX get correct PACKET_ type here */
 	skb->pkt_type = PACKET_HOST;
-	netif_rx_ni(skb);
+	netif_receive_skb(skb);
 
 repost:
 	if (unlikely(ipoib_cm_post_receive(dev, wr_id)))
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
index 1bdb910..68d72c6 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
@@ -226,7 +226,7 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 		skb->dev = dev;
 		/* XXX get correct PACKET_ type here */
 		skb->pkt_type = PACKET_HOST;
-		netif_rx_ni(skb);
+		netif_receive_skb(skb);
 	} else {
 		ipoib_dbg_data(priv, "dropping loopback packet\n");
 		dev_kfree_skb_any(skb);
@@ -280,28 +280,63 @@ static void ipoib_ib_handle_tx_wc(struct net_device *dev, struct ib_wc *wc)
 			   wc->status, wr_id, wc->vendor_err);
 }
 
-static void ipoib_ib_handle_wc(struct net_device *dev, struct ib_wc *wc)
+int ipoib_poll(struct net_device *dev, int *budget)
 {
-	if (wc->wr_id & IPOIB_CM_OP_SRQ)
-		ipoib_cm_handle_rx_wc(dev, wc);
-	else if (wc->wr_id & IPOIB_OP_RECV)
-		ipoib_ib_handle_rx_wc(dev, wc);
-	else
-		ipoib_ib_handle_tx_wc(dev, wc);
+	struct ipoib_dev_priv *priv = netdev_priv(dev);
+	int max = min(*budget, dev->quota);
+	int done;
+	int t;
+	int empty;
+	int n, i;
+
+	done  = 0;
+	empty = 0;
+
+	while (max) {
+		t = min(IPOIB_NUM_WC, max);
+		n = ib_poll_cq(priv->cq, t, priv->ibwc);
+
+		for (i = 0; i < n; ++i) {
+			struct ib_wc *wc = priv->ibwc + i;
+
+			if (wc->wr_id & IPOIB_CM_OP_SRQ) {
+				++done;
+				--max;
+				ipoib_cm_handle_rx_wc(dev, wc);
+			} else if (wc->wr_id & IPOIB_OP_RECV) {
+				++done;
+				--max;
+				ipoib_ib_handle_rx_wc(dev, wc);
+			} else
+				ipoib_ib_handle_tx_wc(dev, wc);
+		}
+
+		if (n != t) {
+			empty = 1;
+			break;
+		}
+	}
+
+	dev->quota -= done;
+	*budget    -= done;
+
+	if (empty) {
+		netif_rx_complete(dev);
+		if (unlikely(ib_req_notify_cq(priv->cq,
+					      IB_CQ_NEXT_COMP |
+					      IB_CQ_REPORT_MISSED_EVENTS)) &&
+		    netif_rx_reschedule(dev, 0))
+			return 1;
+
+		return 0;
+	}
+
+	return 1;
 }
 
 void ipoib_ib_completion(struct ib_cq *cq, void *dev_ptr)
 {
-	struct net_device *dev = (struct net_device *) dev_ptr;
-	struct ipoib_dev_priv *priv = netdev_priv(dev);
-	int n, i;
-
-	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
-	do {
-		n = ib_poll_cq(cq, IPOIB_NUM_WC, priv->ibwc);
-		for (i = 0; i < n; ++i)
-			ipoib_ib_handle_wc(dev, priv->ibwc + i);
-	} while (n == IPOIB_NUM_WC);
+	netif_rx_schedule(dev_ptr);
 }
 
 static inline int post_send(struct ipoib_dev_priv *priv,
@@ -514,9 +549,10 @@ int ipoib_ib_dev_stop(struct net_device *dev)
 	struct ib_qp_attr qp_attr;
 	unsigned long begin;
 	struct ipoib_tx_buf *tx_req;
-	int i;
+	int i, n;
 
 	clear_bit(IPOIB_FLAG_INITIALIZED, &priv->flags);
+	netif_poll_disable(dev);
 
 	ipoib_cm_dev_stop(dev);
 
@@ -568,6 +604,18 @@ int ipoib_ib_dev_stop(struct net_device *dev)
 			goto timeout;
 		}
 
+		do {
+			n = ib_poll_cq(priv->cq, IPOIB_NUM_WC, priv->ibwc);
+			for (i = 0; i < n; ++i) {
+				if (priv->ibwc[i].wr_id & IPOIB_CM_OP_SRQ)
+					ipoib_cm_handle_rx_wc(dev, priv->ibwc + i);
+				else if (priv->ibwc[i].wr_id & IPOIB_OP_RECV)
+					ipoib_ib_handle_rx_wc(dev, priv->ibwc + i);
+				else
+					ipoib_ib_handle_tx_wc(dev, priv->ibwc + i);
+			}
+		} while (n == IPOIB_NUM_WC);
+
 		msleep(1);
 	}
 
@@ -596,6 +644,9 @@ timeout:
 		msleep(1);
 	}
 
+	netif_poll_enable(dev);
+	ib_req_notify_cq(priv->cq, IB_CQ_NEXT_COMP);
+
 	return 0;
 }
 
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c
index b4c380c..0a428f2 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_main.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c
@@ -948,6 +948,8 @@ static void ipoib_setup(struct net_device *dev)
 	dev->hard_header 	 = ipoib_hard_header;
 	dev->set_multicast_list  = ipoib_set_mcast_list;
 	dev->neigh_setup         = ipoib_neigh_setup_dev;
+	dev->poll                = ipoib_poll;
+	dev->weight              = 100;
 
 	dev->watchdog_timeo 	 = HZ;
 
-- 
cgit v0.10.2