IB/core: Add an unbound WQ type to the new CQ API

commit f794809a7259dfaa3d47d90ef5a86007cf48b1ce upstream.

The upstream kernel commit cited below modified the workqueue in the
new CQ API to be bound to a specific CPU (instead of being unbound).
This caused ALL users of the new CQ API to use the same bound WQ.

Specifically, MAD handling was severely delayed when the CPU bound
to the WQ was busy handling (higher priority) interrupts.

This caused a delay in the MAD "heartbeat" response handling,
which resulted in ports being incorrectly classified as "down".

To fix this, add a new "unbound" WQ type to the new CQ API, so that users
have the option to choose either a bound WQ or an unbound WQ.

For MADs, choose the new "unbound" WQ.

Fixes: b7363e67b23e ("IB/device: Convert ib-comp-wq to be CPU-bound")
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.m>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

3 files changed, 21 insertions, 4 deletions

diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c
index 757d308bebe8..88c54db16f29 100644
--- a/drivers/infiniband/core/cq.c
+++ b/drivers/infiniband/core/cq.c
@@ -112,12 +112,12 @@ static void ib_cq_poll_work(struct work_struct *work)
 				    IB_POLL_BATCH);
 	if (completed >= IB_POLL_BUDGET_WORKQUEUE ||
 	    ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0)
-		queue_work(ib_comp_wq, &cq->work);
+		queue_work(cq->comp_wq, &cq->work);
 }
 
 static void ib_cq_completion_workqueue(struct ib_cq *cq, void *private)
 {
-	queue_work(ib_comp_wq, &cq->work);
+	queue_work(cq->comp_wq, &cq->work);
 }
 
 /**
@@ -169,9 +169,12 @@ struct ib_cq *ib_alloc_cq(struct ib_device *dev, void *private,
 		ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
 		break;
 	case IB_POLL_WORKQUEUE:
+	case IB_POLL_UNBOUND_WORKQUEUE:
 		cq->comp_handler = ib_cq_completion_workqueue;
 		INIT_WORK(&cq->work, ib_cq_poll_work);
 		ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
+		cq->comp_wq = (cq->poll_ctx == IB_POLL_WORKQUEUE) ?
+				ib_comp_wq : ib_comp_unbound_wq;
 		break;
 	default:
 		ret = -EINVAL;
@@ -206,6 +209,7 @@ void ib_free_cq(struct ib_cq *cq)
 		irq_poll_disable(&cq->iop);
 		break;
 	case IB_POLL_WORKQUEUE:
+	case IB_POLL_UNBOUND_WORKQUEUE:
 		cancel_work_sync(&cq->work);
 		break;
 	default:
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index 4dff06ab771e..61ade4b3e7bb 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -61,6 +61,7 @@ struct ib_client_data {
 };
 
 struct workqueue_struct *ib_comp_wq;
+struct workqueue_struct *ib_comp_unbound_wq;
 struct workqueue_struct *ib_wq;
 EXPORT_SYMBOL_GPL(ib_wq);
 
@@ -1202,10 +1203,19 @@ static int __init ib_core_init(void)
 		goto err;
 	}
 
+	ib_comp_unbound_wq =
+		alloc_workqueue("ib-comp-unb-wq",
+				WQ_UNBOUND | WQ_HIGHPRI | WQ_MEM_RECLAIM |
+				WQ_SYSFS, WQ_UNBOUND_MAX_ACTIVE);
+	if (!ib_comp_unbound_wq) {
+		ret = -ENOMEM;
+		goto err_comp;
+	}
+
 	ret = class_register(&ib_class);
 	if (ret) {
 		pr_warn("Couldn't create InfiniBand device class\n");
-		goto err_comp;
+		goto err_comp_unbound;
 	}
 
 	ret = rdma_nl_init();
@@ -1254,6 +1264,8 @@ err_ibnl:
 	rdma_nl_exit();
 err_sysfs:
 	class_unregister(&ib_class);
+err_comp_unbound:
+	destroy_workqueue(ib_comp_unbound_wq);
 err_comp:
 	destroy_workqueue(ib_comp_wq);
 err:
@@ -1272,6 +1284,7 @@ static void __exit ib_core_cleanup(void)
 	addr_cleanup();
 	rdma_nl_exit();
 	class_unregister(&ib_class);
+	destroy_workqueue(ib_comp_unbound_wq);
 	destroy_workqueue(ib_comp_wq);
 	/* Make sure that any pending umem accounting work is done. */
 	destroy_workqueue(ib_wq);
diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index 49b6da1d990f..e4339b9e43a5 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -3178,7 +3178,7 @@ static int ib_mad_port_open(struct ib_device *device,
 	}
 
 	port_priv->cq = ib_alloc_cq(port_priv->device, port_priv, cq_size, 0,
-			IB_POLL_WORKQUEUE);
+			IB_POLL_UNBOUND_WORKQUEUE);
 	if (IS_ERR(port_priv->cq)) {
 		dev_err(&device->dev, "Couldn't create ib_mad CQ\n");
 		ret = PTR_ERR(port_priv->cq);