Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma

Pull rdma updates from Doug Ledford:
 "Initial roundup of 4.5 merge window patches

   - Remove usage of ib_query_device and instead store attributes in
     ib_device struct

   - Move iopoll out of block and into lib, rename to irqpoll, and use
     in several places in the rdma stack as our new completion queue
     polling library mechanism.  Update the other block drivers that
     already used iopoll to use the new mechanism too.

   - Replace the per-entry GID table locks with a single GID table lock

   - IPoIB multicast cleanup

   - Cleanups to the IB MR facility

   - Add support for 64bit extended IB counters

   - Fix for netlink oops while parsing RDMA nl messages

   - RoCEv2 support for the core IB code

   - mlx4 RoCEv2 support

   - mlx5 RoCEv2 support

   - Cross Channel support for mlx5

   - Timestamp support for mlx5

   - Atomic support for mlx5

   - Raw QP support for mlx5

   - MAINTAINERS update for mlx4/mlx5

   - Misc ocrdma, qib, nes, usNIC, cxgb3, cxgb4, mlx4, mlx5 updates

   - Add support for remote invalidate to the iSER driver (pushed
     through the RDMA tree due to dependencies, acknowledged by nab)

   - Update to NFSoRDMA (pushed through the RDMA tree due to
     dependencies, acknowledged by Bruce)"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma: (169 commits)
  IB/mlx5: Unify CQ create flags check
  IB/mlx5: Expose Raw Packet QP to user space consumers
  {IB, net}/mlx5: Move the modify QP operation table to mlx5_ib
  IB/mlx5: Support setting Ethernet priority for Raw Packet QPs
  IB/mlx5: Add Raw Packet QP query functionality
  IB/mlx5: Add create and destroy functionality for Raw Packet QP
  IB/mlx5: Refactor mlx5_ib_qp to accommodate other QP types
  IB/mlx5: Allocate a Transport Domain for each ucontext
  net/mlx5_core: Warn on unsupported events of QP/RQ/SQ
  net/mlx5_core: Add RQ and SQ event handling
  net/mlx5_core: Export transport objects
  IB/mlx5: Expose CQE version to user-space
  IB/mlx5: Add CQE version 1 support to user QPs and SRQs
  IB/mlx5: Fix data validation in mlx5_ib_alloc_ucontext
  IB/sa: Fix netlink local service GFP crash
  IB/srpt: Remove redundant wc array
  IB/qib: Improve ipoib UD performance
  IB/mlx4: Advertise RoCE v2 support
  IB/mlx4: Create and use another QP1 for RoCEv2
  IB/mlx4: Enable send of RoCE QP1 packets with IP/UDP headers
  ...

5 files changed, 56 insertions, 92 deletions

diff --git a/drivers/infiniband/hw/qib/qib_mr.c b/drivers/infiniband/hw/qib/qib_mr.c
index 294f5c7..5f53304 100644
--- a/drivers/infiniband/hw/qib/qib_mr.c
+++ b/drivers/infiniband/hw/qib/qib_mr.c
@@ -150,10 +150,7 @@ static struct qib_mr *alloc_mr(int count, struct ib_pd *pd)
 	rval = init_qib_mregion(&mr->mr, pd, count);
 	if (rval)
 		goto bail;
-	/*
-	 * ib_reg_phys_mr() will initialize mr->ibmr except for
-	 * lkey and rkey.
-	 */
+
 	rval = qib_alloc_lkey(&mr->mr, 0);
 	if (rval)
 		goto bail_mregion;
@@ -171,52 +168,6 @@ bail:
 }
 
 /**
- * qib_reg_phys_mr - register a physical memory region
- * @pd: protection domain for this memory region
- * @buffer_list: pointer to the list of physical buffers to register
- * @num_phys_buf: the number of physical buffers to register
- * @iova_start: the starting address passed over IB which maps to this MR
- *
- * Returns the memory region on success, otherwise returns an errno.
- */
-struct ib_mr *qib_reg_phys_mr(struct ib_pd *pd,
-			      struct ib_phys_buf *buffer_list,
-			      int num_phys_buf, int acc, u64 *iova_start)
-{
-	struct qib_mr *mr;
-	int n, m, i;
-	struct ib_mr *ret;
-
-	mr = alloc_mr(num_phys_buf, pd);
-	if (IS_ERR(mr)) {
-		ret = (struct ib_mr *)mr;
-		goto bail;
-	}
-
-	mr->mr.user_base = *iova_start;
-	mr->mr.iova = *iova_start;
-	mr->mr.access_flags = acc;
-
-	m = 0;
-	n = 0;
-	for (i = 0; i < num_phys_buf; i++) {
-		mr->mr.map[m]->segs[n].vaddr = (void *) buffer_list[i].addr;
-		mr->mr.map[m]->segs[n].length = buffer_list[i].size;
-		mr->mr.length += buffer_list[i].size;
-		n++;
-		if (n == QIB_SEGSZ) {
-			m++;
-			n = 0;
-		}
-	}
-
-	ret = &mr->ibmr;
-
-bail:
-	return ret;
-}
-
-/**
  * qib_reg_user_mr - register a userspace memory region
  * @pd: protection domain for this memory region
  * @start: starting userspace address
diff --git a/drivers/infiniband/hw/qib/qib_qp.c b/drivers/infiniband/hw/qib/qib_qp.c
index 40f85bb..3eff35c 100644
--- a/drivers/infiniband/hw/qib/qib_qp.c
+++ b/drivers/infiniband/hw/qib/qib_qp.c
@@ -100,9 +100,10 @@ static u32 credit_table[31] = {
 	32768                   /* 1E */
 };
 
-static void get_map_page(struct qib_qpn_table *qpt, struct qpn_map *map)
+static void get_map_page(struct qib_qpn_table *qpt, struct qpn_map *map,
+			 gfp_t gfp)
 {
-	unsigned long page = get_zeroed_page(GFP_KERNEL);
+	unsigned long page = get_zeroed_page(gfp);
 
 	/*
 	 * Free the page if someone raced with us installing it.
@@ -121,7 +122,7 @@ static void get_map_page(struct qib_qpn_table *qpt, struct qpn_map *map)
  * zero/one for QP type IB_QPT_SMI/IB_QPT_GSI.
  */
 static int alloc_qpn(struct qib_devdata *dd, struct qib_qpn_table *qpt,
-		     enum ib_qp_type type, u8 port)
+		     enum ib_qp_type type, u8 port, gfp_t gfp)
 {
 	u32 i, offset, max_scan, qpn;
 	struct qpn_map *map;
@@ -151,7 +152,7 @@ static int alloc_qpn(struct qib_devdata *dd, struct qib_qpn_table *qpt,
 	max_scan = qpt->nmaps - !offset;
 	for (i = 0;;) {
 		if (unlikely(!map->page)) {
-			get_map_page(qpt, map);
+			get_map_page(qpt, map, gfp);
 			if (unlikely(!map->page))
 				break;
 		}
@@ -983,13 +984,21 @@ struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
 	size_t sz;
 	size_t sg_list_sz;
 	struct ib_qp *ret;
+	gfp_t gfp;
+
 
 	if (init_attr->cap.max_send_sge > ib_qib_max_sges ||
 	    init_attr->cap.max_send_wr > ib_qib_max_qp_wrs ||
-	    init_attr->create_flags) {
-		ret = ERR_PTR(-EINVAL);
-		goto bail;
-	}
+	    init_attr->create_flags & ~(IB_QP_CREATE_USE_GFP_NOIO))
+		return ERR_PTR(-EINVAL);
+
+	/* GFP_NOIO is applicable in RC QPs only */
+	if (init_attr->create_flags & IB_QP_CREATE_USE_GFP_NOIO &&
+	    init_attr->qp_type != IB_QPT_RC)
+		return ERR_PTR(-EINVAL);
+
+	gfp = init_attr->create_flags & IB_QP_CREATE_USE_GFP_NOIO ?
+			GFP_NOIO : GFP_KERNEL;
 
 	/* Check receive queue parameters if no SRQ is specified. */
 	if (!init_attr->srq) {
@@ -1021,7 +1030,8 @@ struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
 		sz = sizeof(struct qib_sge) *
 			init_attr->cap.max_send_sge +
 			sizeof(struct qib_swqe);
-		swq = vmalloc((init_attr->cap.max_send_wr + 1) * sz);
+		swq = __vmalloc((init_attr->cap.max_send_wr + 1) * sz,
+				gfp, PAGE_KERNEL);
 		if (swq == NULL) {
 			ret = ERR_PTR(-ENOMEM);
 			goto bail;
@@ -1037,13 +1047,13 @@ struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
 		} else if (init_attr->cap.max_recv_sge > 1)
 			sg_list_sz = sizeof(*qp->r_sg_list) *
 				(init_attr->cap.max_recv_sge - 1);
-		qp = kzalloc(sz + sg_list_sz, GFP_KERNEL);
+		qp = kzalloc(sz + sg_list_sz, gfp);
 		if (!qp) {
 			ret = ERR_PTR(-ENOMEM);
 			goto bail_swq;
 		}
 		RCU_INIT_POINTER(qp->next, NULL);
-		qp->s_hdr = kzalloc(sizeof(*qp->s_hdr), GFP_KERNEL);
+		qp->s_hdr = kzalloc(sizeof(*qp->s_hdr), gfp);
 		if (!qp->s_hdr) {
 			ret = ERR_PTR(-ENOMEM);
 			goto bail_qp;
@@ -1058,8 +1068,16 @@ struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
 			qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
 			sz = (sizeof(struct ib_sge) * qp->r_rq.max_sge) +
 				sizeof(struct qib_rwqe);
-			qp->r_rq.wq = vmalloc_user(sizeof(struct qib_rwq) +
-						   qp->r_rq.size * sz);
+			if (gfp != GFP_NOIO)
+				qp->r_rq.wq = vmalloc_user(
+						sizeof(struct qib_rwq) +
+						qp->r_rq.size * sz);
+			else
+				qp->r_rq.wq = __vmalloc(
+						sizeof(struct qib_rwq) +
+						qp->r_rq.size * sz,
+						gfp, PAGE_KERNEL);
+
 			if (!qp->r_rq.wq) {
 				ret = ERR_PTR(-ENOMEM);
 				goto bail_qp;
@@ -1090,7 +1108,7 @@ struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
 		dev = to_idev(ibpd->device);
 		dd = dd_from_dev(dev);
 		err = alloc_qpn(dd, &dev->qpn_table, init_attr->qp_type,
-				init_attr->port_num);
+				init_attr->port_num, gfp);
 		if (err < 0) {
 			ret = ERR_PTR(err);
 			vfree(qp->r_rq.wq);
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index de6cb6f..baf1e42 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -346,6 +346,7 @@ static int qib_post_one_send(struct qib_qp *qp, struct ib_send_wr *wr,
 	unsigned long flags;
 	struct qib_lkey_table *rkt;
 	struct qib_pd *pd;
+	int avoid_schedule = 0;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
 
@@ -438,11 +439,15 @@ static int qib_post_one_send(struct qib_qp *qp, struct ib_send_wr *wr,
 	    qp->ibqp.qp_type == IB_QPT_RC) {
 		if (wqe->length > 0x80000000U)
 			goto bail_inval_free;
+		if (wqe->length <= qp->pmtu)
+			avoid_schedule = 1;
 	} else if (wqe->length > (dd_from_ibdev(qp->ibqp.device)->pport +
-				  qp->port_num - 1)->ibmtu)
+				  qp->port_num - 1)->ibmtu) {
 		goto bail_inval_free;
-	else
+	} else {
 		atomic_inc(&to_iah(ud_wr(wr)->ah)->refcount);
+		avoid_schedule = 1;
+	}
 	wqe->ssn = qp->s_ssn++;
 	qp->s_head = next;
 
@@ -458,7 +463,7 @@ bail_inval_free:
 bail_inval:
 	ret = -EINVAL;
 bail:
-	if (!ret && !wr->next &&
+	if (!ret && !wr->next && !avoid_schedule &&
 	 !qib_sdma_empty(
 	   dd_from_ibdev(qp->ibqp.device)->pport + qp->port_num - 1)) {
 		qib_schedule_send(qp);
@@ -2256,7 +2261,6 @@ int qib_register_ib_device(struct qib_devdata *dd)
 	ibdev->poll_cq = qib_poll_cq;
 	ibdev->req_notify_cq = qib_req_notify_cq;
 	ibdev->get_dma_mr = qib_get_dma_mr;
-	ibdev->reg_phys_mr = qib_reg_phys_mr;
 	ibdev->reg_user_mr = qib_reg_user_mr;
 	ibdev->dereg_mr = qib_dereg_mr;
 	ibdev->alloc_mr = qib_alloc_mr;
diff --git a/drivers/infiniband/hw/qib/qib_verbs.h b/drivers/infiniband/hw/qib/qib_verbs.h
index bc803f3..6c5e777 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.h
+++ b/drivers/infiniband/hw/qib/qib_verbs.h
@@ -1032,10 +1032,6 @@ int qib_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata);
 
 struct ib_mr *qib_get_dma_mr(struct ib_pd *pd, int acc);
 
-struct ib_mr *qib_reg_phys_mr(struct ib_pd *pd,
-			      struct ib_phys_buf *buffer_list,
-			      int num_phys_buf, int acc, u64 *iova_start);
-
 struct ib_mr *qib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 			      u64 virt_addr, int mr_access_flags,
 			      struct ib_udata *udata);
diff --git a/drivers/infiniband/hw/qib/qib_verbs_mcast.c b/drivers/infiniband/hw/qib/qib_verbs_mcast.c
index f8ea069..b2fb528 100644
--- a/drivers/infiniband/hw/qib/qib_verbs_mcast.c
+++ b/drivers/infiniband/hw/qib/qib_verbs_mcast.c
@@ -286,15 +286,13 @@ int qib_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 	struct qib_ibdev *dev = to_idev(ibqp->device);
 	struct qib_ibport *ibp = to_iport(ibqp->device, qp->port_num);
 	struct qib_mcast *mcast = NULL;
-	struct qib_mcast_qp *p, *tmp;
+	struct qib_mcast_qp *p, *tmp, *delp = NULL;
 	struct rb_node *n;
 	int last = 0;
 	int ret;
 
-	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET) {
-		ret = -EINVAL;
-		goto bail;
-	}
+	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET)
+		return -EINVAL;
 
 	spin_lock_irq(&ibp->lock);
 
@@ -303,8 +301,7 @@ int qib_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 	while (1) {
 		if (n == NULL) {
 			spin_unlock_irq(&ibp->lock);
-			ret = -EINVAL;
-			goto bail;
+			return -EINVAL;
 		}
 
 		mcast = rb_entry(n, struct qib_mcast, rb_node);
@@ -328,6 +325,7 @@ int qib_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 		 */
 		list_del_rcu(&p->list);
 		mcast->n_attached--;
+		delp = p;
 
 		/* If this was the last attached QP, remove the GID too. */
 		if (list_empty(&mcast->qp_list)) {
@@ -338,15 +336,16 @@ int qib_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 	}
 
 	spin_unlock_irq(&ibp->lock);
+	/* QP not attached */
+	if (!delp)
+		return -EINVAL;
+	/*
+	 * Wait for any list walkers to finish before freeing the
+	 * list element.
+	 */
+	wait_event(mcast->wait, atomic_read(&mcast->refcount) <= 1);
+	qib_mcast_qp_free(delp);
 
-	if (p) {
-		/*
-		 * Wait for any list walkers to finish before freeing the
-		 * list element.
-		 */
-		wait_event(mcast->wait, atomic_read(&mcast->refcount) <= 1);
-		qib_mcast_qp_free(p);
-	}
 	if (last) {
 		atomic_dec(&mcast->refcount);
 		wait_event(mcast->wait, !atomic_read(&mcast->refcount));
@@ -355,11 +354,7 @@ int qib_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 		dev->n_mcast_grps_allocated--;
 		spin_unlock_irq(&dev->n_mcast_grps_lock);
 	}
-
-	ret = 0;
-
-bail:
-	return ret;
+	return 0;
 }
 
 int qib_mcast_tree_empty(struct qib_ibport *ibp)