1 files changed, 92 insertions, 43 deletions

diff --git a/drivers/infiniband/hw/amso1100/c2_qp.c b/drivers/infiniband/hw/amso1100/c2_qp.c
index 12261132b07..86708dee58b 100644
--- a/drivers/infiniband/hw/amso1100/c2_qp.c
+++ b/drivers/infiniband/hw/amso1100/c2_qp.c
@@ -35,6 +35,9 @@
  *
  */
 
+#include <linux/delay.h>
+#include <linux/gfp.h>
+
 #include "c2.h"
 #include "c2_vq.h"
 #include "c2_status.h"
@@ -119,7 +122,7 @@ void c2_set_qp_state(struct c2_qp *qp, int c2_state)
 	int new_state = to_ib_state(c2_state);
 
 	pr_debug("%s: qp[%p] state modify %s --> %s\n",
-	       __FUNCTION__,
+	       __func__,
 		qp,
 		to_ib_state_str(qp->state),
 		to_ib_state_str(new_state));
@@ -139,7 +142,7 @@ int c2_qp_modify(struct c2_dev *c2dev, struct c2_qp *qp,
 	int err;
 
 	pr_debug("%s:%d qp=%p, %s --> %s\n",
-		__FUNCTION__, __LINE__,
+		__func__, __LINE__,
 		qp,
 		to_ib_state_str(qp->state),
 		to_ib_state_str(attr->qp_state));
@@ -159,8 +162,10 @@ int c2_qp_modify(struct c2_dev *c2dev, struct c2_qp *qp,
 
 	if (attr_mask & IB_QP_STATE) {
 		/* Ensure the state is valid */
-		if (attr->qp_state < 0 || attr->qp_state > IB_QPS_ERR)
-			return -EINVAL;
+		if (attr->qp_state < 0 || attr->qp_state > IB_QPS_ERR) {
+			err = -EINVAL;
+			goto bail0;
+		}
 
 		wr.next_qp_state = cpu_to_be32(to_c2_state(attr->qp_state));
 
@@ -182,9 +187,10 @@ int c2_qp_modify(struct c2_dev *c2dev, struct c2_qp *qp,
 		if (attr->cur_qp_state != IB_QPS_RTR &&
 		    attr->cur_qp_state != IB_QPS_RTS &&
 		    attr->cur_qp_state != IB_QPS_SQD &&
-		    attr->cur_qp_state != IB_QPS_SQE)
-			return -EINVAL;
-		else
+		    attr->cur_qp_state != IB_QPS_SQE) {
+			err = -EINVAL;
+			goto bail0;
+		} else
 			wr.next_qp_state =
 			    cpu_to_be32(to_c2_state(attr->cur_qp_state));
 
@@ -219,7 +225,7 @@ int c2_qp_modify(struct c2_dev *c2dev, struct c2_qp *qp,
 		qp->state = next_state;
 #ifdef DEBUG
 	else
-		pr_debug("%s: c2_errno=%d\n", __FUNCTION__, err);
+		pr_debug("%s: c2_errno=%d\n", __func__, err);
 #endif
 	/*
 	 * If we're going to error and generating the event here, then
@@ -238,7 +244,7 @@ int c2_qp_modify(struct c2_dev *c2dev, struct c2_qp *qp,
 	vq_req_free(c2dev, vq_req);
 
 	pr_debug("%s:%d qp=%p, cur_state=%s\n",
-		__FUNCTION__, __LINE__,
+		__func__, __LINE__,
 		qp,
 		to_ib_state_str(qp->state));
 	return err;
@@ -376,14 +382,16 @@ static int c2_alloc_qpn(struct c2_dev *c2dev, struct c2_qp *qp)
 {
 	int ret;
 
-        do {
-		spin_lock_irq(&c2dev->qp_table.lock);
-		ret = idr_get_new_above(&c2dev->qp_table.idr, qp,
-					c2dev->qp_table.last++, &qp->qpn);
-		spin_unlock_irq(&c2dev->qp_table.lock);
-        } while ((ret == -EAGAIN) &&
-	 	 idr_pre_get(&c2dev->qp_table.idr, GFP_KERNEL));
-	return ret;
+	idr_preload(GFP_KERNEL);
+	spin_lock_irq(&c2dev->qp_table.lock);
+
+	ret = idr_alloc_cyclic(&c2dev->qp_table.idr, qp, 0, 0, GFP_NOWAIT);
+	if (ret >= 0)
+		qp->qpn = ret;
+
+	spin_unlock_irq(&c2dev->qp_table.lock);
+	idr_preload_end();
+	return ret < 0 ? ret : 0;
 }
 
 static void c2_free_qpn(struct c2_dev *c2dev, int qpn)
@@ -501,6 +509,7 @@ int c2_alloc_qp(struct c2_dev *c2dev,
 	qp->send_sgl_depth = qp_attrs->cap.max_send_sge;
 	qp->rdma_write_sgl_depth = qp_attrs->cap.max_send_sge;
 	qp->recv_sgl_depth = qp_attrs->cap.max_recv_sge;
+	init_waitqueue_head(&qp->wait);
 
 	/* Initialize the SQ MQ */
 	q_size = be32_to_cpu(reply->sq_depth);
@@ -562,6 +571,32 @@ int c2_alloc_qp(struct c2_dev *c2dev,
 	return err;
 }
 
+static inline void c2_lock_cqs(struct c2_cq *send_cq, struct c2_cq *recv_cq)
+{
+	if (send_cq == recv_cq)
+		spin_lock_irq(&send_cq->lock);
+	else if (send_cq > recv_cq) {
+		spin_lock_irq(&send_cq->lock);
+		spin_lock_nested(&recv_cq->lock, SINGLE_DEPTH_NESTING);
+	} else {
+		spin_lock_irq(&recv_cq->lock);
+		spin_lock_nested(&send_cq->lock, SINGLE_DEPTH_NESTING);
+	}
+}
+
+static inline void c2_unlock_cqs(struct c2_cq *send_cq, struct c2_cq *recv_cq)
+{
+	if (send_cq == recv_cq)
+		spin_unlock_irq(&send_cq->lock);
+	else if (send_cq > recv_cq) {
+		spin_unlock(&recv_cq->lock);
+		spin_unlock_irq(&send_cq->lock);
+	} else {
+		spin_unlock(&send_cq->lock);
+		spin_unlock_irq(&recv_cq->lock);
+	}
+}
+
 void c2_free_qp(struct c2_dev *c2dev, struct c2_qp *qp)
 {
 	struct c2_cq *send_cq;
@@ -574,18 +609,12 @@ void c2_free_qp(struct c2_dev *c2dev, struct c2_qp *qp)
 	 * Lock CQs here, so that CQ polling code can do QP lookup
 	 * without taking a lock.
 	 */
-	spin_lock_irq(&send_cq->lock);
-	if (send_cq != recv_cq)
-		spin_lock(&recv_cq->lock);
-
+	c2_lock_cqs(send_cq, recv_cq);
 	c2_free_qpn(c2dev, qp->qpn);
-
-	if (send_cq != recv_cq)
-		spin_unlock(&recv_cq->lock);
-	spin_unlock_irq(&send_cq->lock);
+	c2_unlock_cqs(send_cq, recv_cq);
 
 	/*
-	 * Destory qp in the rnic...
+	 * Destroy qp in the rnic...
 	 */
 	destroy_qp(c2dev, qp);
 
@@ -705,10 +734,8 @@ static inline void c2_activity(struct c2_dev *c2dev, u32 mq_index, u16 shared)
 	 * cannot get on the bus and the card and system hang in a
 	 * deadlock -- thus the need for this code. [TOT]
 	 */
-	while (readl(c2dev->regs + PCI_BAR0_ADAPTER_HINT) & 0x80000000) {
-		set_current_state(TASK_UNINTERRUPTIBLE);
-		schedule_timeout(0);
-	}
+	while (readl(c2dev->regs + PCI_BAR0_ADAPTER_HINT) & 0x80000000)
+		udelay(10);
 
 	__raw_writel(C2_HINT_MAKE(mq_index, shared),
 		     c2dev->regs + PCI_BAR0_ADAPTER_HINT);
@@ -766,6 +793,7 @@ int c2_post_send(struct ib_qp *ibqp, struct ib_send_wr *ib_wr,
 	struct c2_dev *c2dev = to_c2dev(ibqp->device);
 	struct c2_qp *qp = to_c2qp(ibqp);
 	union c2wr wr;
+	unsigned long lock_flags;
 	int err = 0;
 
 	u32 flags;
@@ -773,8 +801,10 @@ int c2_post_send(struct ib_qp *ibqp, struct ib_send_wr *ib_wr,
 	u8 actual_sge_count;
 	u32 msg_size;
 
-	if (qp->state > IB_QPS_RTS)
-		return -EINVAL;
+	if (qp->state > IB_QPS_RTS) {
+		err = -EINVAL;
+		goto out;
+	}
 
 	while (ib_wr) {
 
@@ -786,16 +816,24 @@ int c2_post_send(struct ib_qp *ibqp, struct ib_send_wr *ib_wr,
 
 		switch (ib_wr->opcode) {
 		case IB_WR_SEND:
-			if (ib_wr->send_flags & IB_SEND_SOLICITED) {
-				c2_wr_set_id(&wr, C2_WR_TYPE_SEND_SE);
-				msg_size = sizeof(struct c2wr_send_req);
+		case IB_WR_SEND_WITH_INV:
+			if (ib_wr->opcode == IB_WR_SEND) {
+				if (ib_wr->send_flags & IB_SEND_SOLICITED)
+					c2_wr_set_id(&wr, C2_WR_TYPE_SEND_SE);
+				else
+					c2_wr_set_id(&wr, C2_WR_TYPE_SEND);
+				wr.sqwr.send.remote_stag = 0;
 			} else {
-				c2_wr_set_id(&wr, C2_WR_TYPE_SEND);
-				msg_size = sizeof(struct c2wr_send_req);
+				if (ib_wr->send_flags & IB_SEND_SOLICITED)
+					c2_wr_set_id(&wr, C2_WR_TYPE_SEND_SE_INV);
+				else
+					c2_wr_set_id(&wr, C2_WR_TYPE_SEND_INV);
+				wr.sqwr.send.remote_stag =
+					cpu_to_be32(ib_wr->ex.invalidate_rkey);
 			}
 
-			wr.sqwr.send.remote_stag = 0;
-			msg_size += sizeof(struct c2_data_addr) * ib_wr->num_sge;
+			msg_size = sizeof(struct c2wr_send_req) +
+				sizeof(struct c2_data_addr) * ib_wr->num_sge;
 			if (ib_wr->num_sge > qp->send_sgl_depth) {
 				err = -EINVAL;
 				break;
@@ -881,8 +919,10 @@ int c2_post_send(struct ib_qp *ibqp, struct ib_send_wr *ib_wr,
 		/*
 		 * Post the puppy!
 		 */
+		spin_lock_irqsave(&qp->lock, lock_flags);
 		err = qp_wr_post(&qp->sq_mq, &wr, qp, msg_size);
 		if (err) {
+			spin_unlock_irqrestore(&qp->lock, lock_flags);
 			break;
 		}
 
@@ -890,10 +930,12 @@ int c2_post_send(struct ib_qp *ibqp, struct ib_send_wr *ib_wr,
 		 * Enqueue mq index to activity FIFO.
 		 */
 		c2_activity(c2dev, qp->sq_mq.index, qp->sq_mq.hint_count);
+		spin_unlock_irqrestore(&qp->lock, lock_flags);
 
 		ib_wr = ib_wr->next;
 	}
 
+out:
 	if (err)
 		*bad_wr = ib_wr;
 	return err;
@@ -905,10 +947,13 @@ int c2_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *ib_wr,
 	struct c2_dev *c2dev = to_c2dev(ibqp->device);
 	struct c2_qp *qp = to_c2qp(ibqp);
 	union c2wr wr;
+	unsigned long lock_flags;
 	int err = 0;
 
-	if (qp->state > IB_QPS_RTS)
-		return -EINVAL;
+	if (qp->state > IB_QPS_RTS) {
+		err = -EINVAL;
+		goto out;
+	}
 
 	/*
 	 * Try and post each work request
@@ -945,8 +990,10 @@ int c2_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *ib_wr,
 			break;
 		}
 
+		spin_lock_irqsave(&qp->lock, lock_flags);
 		err = qp_wr_post(&qp->rq_mq, &wr, qp, qp->rq_mq.msg_size);
 		if (err) {
+			spin_unlock_irqrestore(&qp->lock, lock_flags);
 			break;
 		}
 
@@ -954,22 +1001,24 @@ int c2_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *ib_wr,
 		 * Enqueue mq index to activity FIFO
 		 */
 		c2_activity(c2dev, qp->rq_mq.index, qp->rq_mq.hint_count);
+		spin_unlock_irqrestore(&qp->lock, lock_flags);
 
 		ib_wr = ib_wr->next;
 	}
 
+out:
 	if (err)
 		*bad_wr = ib_wr;
 	return err;
 }
 
-void __devinit c2_init_qp_table(struct c2_dev *c2dev)
+void c2_init_qp_table(struct c2_dev *c2dev)
 {
 	spin_lock_init(&c2dev->qp_table.lock);
 	idr_init(&c2dev->qp_table.idr);
 }
 
-void __devexit c2_cleanup_qp_table(struct c2_dev *c2dev)
+void c2_cleanup_qp_table(struct c2_dev *c2dev)
 {
 	idr_destroy(&c2dev->qp_table.idr);
 }