RDMA: Convert CQ allocations to be under core responsibility

Ensure that CQ is allocated and freed by IB/core and not by drivers.

Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Acked-by: Gal Pressman <galpress@amazon.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Tested-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

6 files changed, 31 insertions, 59 deletions

diff --git a/drivers/infiniband/sw/rdmavt/cq.c b/drivers/infiniband/sw/rdmavt/cq.c
index 8e76036fad4a..b46714a92b7a 100644
--- a/drivers/infiniband/sw/rdmavt/cq.c
+++ b/drivers/infiniband/sw/rdmavt/cq.c
@@ -166,43 +166,37 @@ static void send_complete(struct work_struct *work)
 
 /**
  * rvt_create_cq - create a completion queue
- * @ibdev: the device this completion queue is attached to
+ * @ibcq: Allocated CQ
  * @attr: creation attributes
  * @udata: user data for libibverbs.so
  *
  * Called by ib_create_cq() in the generic verbs code.
  *
- * Return: pointer to the completion queue or negative errno values
- * for failure.
+ * Return: 0 on success
  */
-struct ib_cq *rvt_create_cq(struct ib_device *ibdev,
-			    const struct ib_cq_init_attr *attr,
-			    struct ib_udata *udata)
+int rvt_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
+		  struct ib_udata *udata)
 {
+	struct ib_device *ibdev = ibcq->device;
 	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
-	struct rvt_cq *cq;
+	struct rvt_cq *cq = container_of(ibcq, struct rvt_cq, ibcq);
 	struct rvt_cq_wc *wc;
-	struct ib_cq *ret;
 	u32 sz;
 	unsigned int entries = attr->cqe;
 	int comp_vector = attr->comp_vector;
+	int err;
 
 	if (attr->flags)
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 
 	if (entries < 1 || entries > rdi->dparms.props.max_cqe)
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 
 	if (comp_vector < 0)
 		comp_vector = 0;
 
 	comp_vector = comp_vector % rdi->ibdev.num_comp_vectors;
 
-	/* Allocate the completion queue structure. */
-	cq = kzalloc_node(sizeof(*cq), GFP_KERNEL, rdi->dparms.node);
-	if (!cq)
-		return ERR_PTR(-ENOMEM);
-
 	/*
 	 * Allocate the completion queue entries and head/tail pointers.
 	 * This is allocated separately so that it can be resized and
@@ -218,36 +212,29 @@ struct ib_cq *rvt_create_cq(struct ib_device *ibdev,
 	wc = udata ?
 		vmalloc_user(sz) :
 		vzalloc_node(sz, rdi->dparms.node);
-	if (!wc) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail_cq;
-	}
-
+	if (!wc)
+		return -ENOMEM;
 	/*
 	 * Return the address of the WC as the offset to mmap.
 	 * See rvt_mmap() for details.
 	 */
 	if (udata && udata->outlen >= sizeof(__u64)) {
-		int err;
-
 		cq->ip = rvt_create_mmap_info(rdi, sz, udata, wc);
 		if (!cq->ip) {
-			ret = ERR_PTR(-ENOMEM);
+			err = -ENOMEM;
 			goto bail_wc;
 		}
 
 		err = ib_copy_to_udata(udata, &cq->ip->offset,
 				       sizeof(cq->ip->offset));
-		if (err) {
-			ret = ERR_PTR(err);
+		if (err)
 			goto bail_ip;
-		}
 	}
 
 	spin_lock_irq(&rdi->n_cqs_lock);
 	if (rdi->n_cqs_allocated == rdi->dparms.props.max_cq) {
 		spin_unlock_irq(&rdi->n_cqs_lock);
-		ret = ERR_PTR(-ENOMEM);
+		err = -ENOMEM;
 		goto bail_ip;
 	}
 
@@ -279,19 +266,14 @@ struct ib_cq *rvt_create_cq(struct ib_device *ibdev,
 	INIT_WORK(&cq->comptask, send_complete);
 	cq->queue = wc;
 
-	ret = &cq->ibcq;
-
 	trace_rvt_create_cq(cq, attr);
-	goto done;
+	return 0;
 
 bail_ip:
 	kfree(cq->ip);
 bail_wc:
 	vfree(wc);
-bail_cq:
-	kfree(cq);
-done:
-	return ret;
+	return err;
 }
 
 /**
@@ -314,7 +296,6 @@ void rvt_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata)
 		kref_put(&cq->ip->ref, rvt_release_mmap_info);
 	else
 		vfree(cq->queue);
-	kfree(cq);
 }
 
 /**
diff --git a/drivers/infiniband/sw/rdmavt/cq.h b/drivers/infiniband/sw/rdmavt/cq.h
index 495d8c3e6580..5e26a2eb19a4 100644
--- a/drivers/infiniband/sw/rdmavt/cq.h
+++ b/drivers/infiniband/sw/rdmavt/cq.h
@@ -51,9 +51,8 @@
 #include <rdma/rdma_vt.h>
 #include <rdma/rdmavt_cq.h>
 
-struct ib_cq *rvt_create_cq(struct ib_device *ibdev,
-			    const struct ib_cq_init_attr *attr,
-			    struct ib_udata *udata);
+int rvt_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
+		  struct ib_udata *udata);
 void rvt_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata);
 int rvt_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags);
 int rvt_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata);
diff --git a/drivers/infiniband/sw/rdmavt/vt.c b/drivers/infiniband/sw/rdmavt/vt.c
index 639ef8ac5400..18da1e1ea979 100644
--- a/drivers/infiniband/sw/rdmavt/vt.c
+++ b/drivers/infiniband/sw/rdmavt/vt.c
@@ -429,6 +429,7 @@ static const struct ib_device_ops rvt_dev_ops = {
 	.unmap_fmr = rvt_unmap_fmr,
 
 	INIT_RDMA_OBJ_SIZE(ib_ah, rvt_ah, ibah),
+	INIT_RDMA_OBJ_SIZE(ib_cq, rvt_cq, ibcq),
 	INIT_RDMA_OBJ_SIZE(ib_pd, rvt_pd, ibpd),
 	INIT_RDMA_OBJ_SIZE(ib_srq, rvt_srq, ibsrq),
 	INIT_RDMA_OBJ_SIZE(ib_ucontext, rvt_ucontext, ibucontext),
diff --git a/drivers/infiniband/sw/rxe/rxe_pool.c b/drivers/infiniband/sw/rxe/rxe_pool.c
index 56cf18af016a..fbcbac52290b 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.c
+++ b/drivers/infiniband/sw/rxe/rxe_pool.c
@@ -72,6 +72,7 @@ struct rxe_type_info rxe_type_info[RXE_NUM_TYPES] = {
 	[RXE_TYPE_CQ] = {
 		.name		= "rxe-cq",
 		.size		= sizeof(struct rxe_cq),
+		.flags          = RXE_POOL_NO_ALLOC,
 		.cleanup	= rxe_cq_cleanup,
 	},
 	[RXE_TYPE_MR] = {
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index b14881decbee..4ebdfcf4d33e 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -778,45 +778,34 @@ err1:
 	return err;
 }
 
-static struct ib_cq *rxe_create_cq(struct ib_device *dev,
-				   const struct ib_cq_init_attr *attr,
-				   struct ib_udata *udata)
+static int rxe_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
+			 struct ib_udata *udata)
 {
 	int err;
+	struct ib_device *dev = ibcq->device;
 	struct rxe_dev *rxe = to_rdev(dev);
-	struct rxe_cq *cq;
+	struct rxe_cq *cq = to_rcq(ibcq);
 	struct rxe_create_cq_resp __user *uresp = NULL;
 
 	if (udata) {
 		if (udata->outlen < sizeof(*uresp))
-			return ERR_PTR(-EINVAL);
+			return -EINVAL;
 		uresp = udata->outbuf;
 	}
 
 	if (attr->flags)
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 
 	err = rxe_cq_chk_attr(rxe, NULL, attr->cqe, attr->comp_vector);
 	if (err)
-		goto err1;
-
-	cq = rxe_alloc(&rxe->cq_pool);
-	if (!cq) {
-		err = -ENOMEM;
-		goto err1;
-	}
+		return err;
 
 	err = rxe_cq_from_init(rxe, cq, attr->cqe, attr->comp_vector, udata,
 			       uresp);
 	if (err)
-		goto err2;
-
-	return &cq->ibcq;
+		return err;
 
-err2:
-	rxe_drop_ref(cq);
-err1:
-	return ERR_PTR(err);
+	return rxe_add_to_pool(&rxe->cq_pool, &cq->pelem);
 }
 
 static void rxe_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata)
@@ -1160,6 +1149,7 @@ static const struct ib_device_ops rxe_dev_ops = {
 	.resize_cq = rxe_resize_cq,
 
 	INIT_RDMA_OBJ_SIZE(ib_ah, rxe_ah, ibah),
+	INIT_RDMA_OBJ_SIZE(ib_cq, rxe_cq, ibcq),
 	INIT_RDMA_OBJ_SIZE(ib_pd, rxe_pd, ibpd),
 	INIT_RDMA_OBJ_SIZE(ib_srq, rxe_srq, ibsrq),
 	INIT_RDMA_OBJ_SIZE(ib_ucontext, rxe_ucontext, ibuc),
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
index e8be7f44e3be..6c997d39a418 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
@@ -85,8 +85,8 @@ struct rxe_cqe {
 };
 
 struct rxe_cq {
-	struct rxe_pool_entry	pelem;
 	struct ib_cq		ibcq;
+	struct rxe_pool_entry	pelem;
 	struct rxe_queue	*queue;
 	spinlock_t		cq_lock;
 	u8			notify;