IB/qib: Avoid returning EBUSY from MR deregister

A timing issue can occur where qib_mr_dereg can return -EBUSY if the
MR use count is not zero.

This can occur if the MR is de-registered while RDMA read response
packets are being progressed from the SDMA ring.  The suspicion is
that the peer sent an RDMA read request, which has already been copied
across to the peer.  The peer sees the completion of his request and
then communicates to the responder that the MR is not needed any
longer.  The responder tries to de-register the MR, catching some
responses remaining in the SDMA ring holding the MR use count.

The code now uses a get/put paradigm to track MR use counts and
coordinates with the MR de-registration process using a completion
when the count has reached zero.  A timeout on the delay is in place
to catch other EBUSY issues.

The reference count protocol is as follows:
- The return to the user counts as 1
- A reference from the lk_table or the qib_ibdev counts as 1.
- Transient I/O operations increase/decrease as necessary

A lot of code duplication has been folded into the new routines
init_qib_mregion() and deinit_qib_mregion().  Additionally, explicit
initialization of fields to zero is now handled by kzalloc().

Also, duplicated code 'while.*num_sge' that decrements reference
counts have been consolidated in qib_put_ss().

Reviewed-by: Ramkrishna Vepa <ramkrishna.vepa@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>

1 files changed, 52 insertions, 32 deletions

diff --git a/drivers/infiniband/hw/qib/qib_keys.c b/drivers/infiniband/hw/qib/qib_keys.c
index 8fd19a47df0..8b5ee3aa8e3 100644
--- a/drivers/infiniband/hw/qib/qib_keys.c
+++ b/drivers/infiniband/hw/qib/qib_keys.c
@@ -35,21 +35,40 @@
 
 /**
  * qib_alloc_lkey - allocate an lkey
- * @rkt: lkey table in which to allocate the lkey
  * @mr: memory region that this lkey protects
+ * @dma_region: 0->normal key, 1->restricted DMA key
+ *
+ * Returns 0 if successful, otherwise returns -errno.
+ *
+ * Increments mr reference count and sets published
+ * as required.
+ *
+ * Sets the lkey field mr for non-dma regions.
  *
- * Returns 1 if successful, otherwise returns 0.
  */
 
-int qib_alloc_lkey(struct qib_lkey_table *rkt, struct qib_mregion *mr)
+int qib_alloc_lkey(struct qib_mregion *mr, int dma_region)
 {
 	unsigned long flags;
 	u32 r;
 	u32 n;
-	int ret;
+	int ret = 0;
+	struct qib_ibdev *dev = to_idev(mr->pd->device);
+	struct qib_lkey_table *rkt = &dev->lk_table;
 
 	spin_lock_irqsave(&rkt->lock, flags);
 
+	/* special case for dma_mr lkey == 0 */
+	if (dma_region) {
+		/* should the dma_mr be relative to the pd? */
+		if (!dev->dma_mr) {
+			qib_get_mr(mr);
+			dev->dma_mr = mr;
+			mr->lkey_published = 1;
+		}
+		goto success;
+	}
+
 	/* Find the next available LKEY */
 	r = rkt->next;
 	n = r;
@@ -57,11 +76,8 @@ int qib_alloc_lkey(struct qib_lkey_table *rkt, struct qib_mregion *mr)
 		if (rkt->table[r] == NULL)
 			break;
 		r = (r + 1) & (rkt->max - 1);
-		if (r == n) {
-			spin_unlock_irqrestore(&rkt->lock, flags);
-			ret = 0;
+		if (r == n)
 			goto bail;
-		}
 	}
 	rkt->next = (r + 1) & (rkt->max - 1);
 	/*
@@ -76,46 +92,50 @@ int qib_alloc_lkey(struct qib_lkey_table *rkt, struct qib_mregion *mr)
 		mr->lkey |= 1 << 8;
 		rkt->gen++;
 	}
+	qib_get_mr(mr);
 	rkt->table[r] = mr;
+	mr->lkey_published = 1;
+success:
 	spin_unlock_irqrestore(&rkt->lock, flags);
-
-	ret = 1;
-
-bail:
+out:
 	return ret;
+bail:
+	spin_unlock_irqrestore(&rkt->lock, flags);
+	ret = -ENOMEM;
+	goto out;
 }
 
 /**
  * qib_free_lkey - free an lkey
- * @rkt: table from which to free the lkey
- * @lkey: lkey id to free
+ * @mr: mr to free from tables
  */
-int qib_free_lkey(struct qib_ibdev *dev, struct qib_mregion *mr)
+void qib_free_lkey(struct qib_mregion *mr)
 {
 	unsigned long flags;
 	u32 lkey = mr->lkey;
 	u32 r;
-	int ret;
+	struct qib_ibdev *dev = to_idev(mr->pd->device);
+	struct qib_lkey_table *rkt = &dev->lk_table;
+
+	spin_lock_irqsave(&rkt->lock, flags);
+	if (!mr->lkey_published)
+		goto out;
+	mr->lkey_published = 0;
+
 
 	spin_lock_irqsave(&dev->lk_table.lock, flags);
 	if (lkey == 0) {
 		if (dev->dma_mr && dev->dma_mr == mr) {
-			ret = atomic_read(&dev->dma_mr->refcount);
-			if (!ret)
-				dev->dma_mr = NULL;
-		} else
-			ret = 0;
+			qib_put_mr(dev->dma_mr);
+			dev->dma_mr = NULL;
+		}
 	} else {
 		r = lkey >> (32 - ib_qib_lkey_table_size);
-		ret = atomic_read(&dev->lk_table.table[r]->refcount);
-		if (!ret)
-			dev->lk_table.table[r] = NULL;
+		qib_put_mr(dev->dma_mr);
+		rkt->table[r] = NULL;
 	}
+out:
 	spin_unlock_irqrestore(&dev->lk_table.lock, flags);
-
-	if (ret)
-		ret = -EBUSY;
-	return ret;
 }
 
 /**
@@ -150,7 +170,7 @@ int qib_lkey_ok(struct qib_lkey_table *rkt, struct qib_pd *pd,
 			goto bail;
 		if (!dev->dma_mr)
 			goto bail;
-		atomic_inc(&dev->dma_mr->refcount);
+		qib_get_mr(dev->dma_mr);
 		spin_unlock_irqrestore(&rkt->lock, flags);
 
 		isge->mr = dev->dma_mr;
@@ -171,7 +191,7 @@ int qib_lkey_ok(struct qib_lkey_table *rkt, struct qib_pd *pd,
 		     off + sge->length > mr->length ||
 		     (mr->access_flags & acc) != acc))
 		goto bail;
-	atomic_inc(&mr->refcount);
+	qib_get_mr(mr);
 	spin_unlock_irqrestore(&rkt->lock, flags);
 
 	off += mr->offset;
@@ -245,7 +265,7 @@ int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
 			goto bail;
 		if (!dev->dma_mr)
 			goto bail;
-		atomic_inc(&dev->dma_mr->refcount);
+		qib_get_mr(dev->dma_mr);
 		spin_unlock_irqrestore(&rkt->lock, flags);
 
 		sge->mr = dev->dma_mr;
@@ -265,7 +285,7 @@ int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
 	if (unlikely(vaddr < mr->iova || off + len > mr->length ||
 		     (mr->access_flags & acc) == 0))
 		goto bail;
-	atomic_inc(&mr->refcount);
+	qib_get_mr(mr);
 	spin_unlock_irqrestore(&rkt->lock, flags);
 
 	off += mr->offset;