[PATCH rdma-next] RDMA/core: Add driver APIs pre_destroy_cq() and post_destroy_cq()

[Leon Romanovsky <leon@kernel.org>]

From: Mark Zhang <markzhang@nvidia.com>

Currently in ib_free_cq, it disables IRQ or cancel the CQ work before
driver destroy_cq. This isn't good as a new IRQ or a CQ work can be
submitted immediately after disabling IRQ or canceling CQ work, which
may run concurrently with destroy_cq and cause crashes.
The right flow should be:
 1. Driver disables CQ to make sure no new CQ event will be submitted;
 2. Disables IRQ or Cancels CQ work in core layer, to make sure no CQ
    polling work is running;
 3. Free all resources to destroy the CQ.

This patch adds 2 driver APIs:
- pre_destroy_cq(): Disable a CQ to prevent it from generating any new
  work completions, but not free any kernel resources;
- post_destroy_cq(): Free all kernel resources.

In ib_free_cq, the IRQ is disabled or CQ work is canceled after
pre_destroy_cq, and before post_destroy_cq.

Fixes: 14d3a3b2498e ("IB: add a proper completion queue abstraction")
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/core/cq.c     | 12 ++++++++++--
 drivers/infiniband/core/device.c |  2 ++
 include/rdma/ib_verbs.h          |  9 +++++++++
 3 files changed, 21 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c
index a70876a0a231..584537c71545 100644
--- a/drivers/infiniband/core/cq.c
+++ b/drivers/infiniband/core/cq.c
@@ -317,13 +317,18 @@ EXPORT_SYMBOL(__ib_alloc_cq_any);
  */
 void ib_free_cq(struct ib_cq *cq)
 {
-	int ret;
+	int ret = 0;
 
 	if (WARN_ON_ONCE(atomic_read(&cq->usecnt)))
 		return;
 	if (WARN_ON_ONCE(cq->cqe_used))
 		return;
 
+	if (cq->device->ops.pre_destroy_cq) {
+		ret = cq->device->ops.pre_destroy_cq(cq);
+		WARN_ONCE(ret, "Disable of kernel CQ shouldn't fail");
+	}
+
 	switch (cq->poll_ctx) {
 	case IB_POLL_DIRECT:
 		break;
@@ -340,7 +345,10 @@ void ib_free_cq(struct ib_cq *cq)
 
 	rdma_dim_destroy(cq);
 	trace_cq_free(cq);
-	ret = cq->device->ops.destroy_cq(cq, NULL);
+	if (cq->device->ops.post_destroy_cq)
+		cq->device->ops.post_destroy_cq(cq);
+	else
+		ret = cq->device->ops.destroy_cq(cq, NULL);
 	WARN_ONCE(ret, "Destroy of kernel CQ shouldn't fail");
 	rdma_restrack_del(&cq->res);
 	kfree(cq->wc);
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index d4263385850a..468ed6bd4722 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -2763,8 +2763,10 @@ void ib_set_device_ops(struct ib_device *dev, const struct ib_device_ops *ops)
 	SET_DEVICE_OP(dev_ops, modify_srq);
 	SET_DEVICE_OP(dev_ops, modify_wq);
 	SET_DEVICE_OP(dev_ops, peek_cq);
+	SET_DEVICE_OP(dev_ops, pre_destroy_cq);
 	SET_DEVICE_OP(dev_ops, poll_cq);
 	SET_DEVICE_OP(dev_ops, port_groups);
+	SET_DEVICE_OP(dev_ops, post_destroy_cq);
 	SET_DEVICE_OP(dev_ops, post_recv);
 	SET_DEVICE_OP(dev_ops, post_send);
 	SET_DEVICE_OP(dev_ops, post_srq_recv);
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index af43a8d2a74a..38f68d245fa6 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -2489,6 +2489,15 @@ struct ib_device_ops {
 	int (*modify_cq)(struct ib_cq *cq, u16 cq_count, u16 cq_period);
 	int (*destroy_cq)(struct ib_cq *cq, struct ib_udata *udata);
 	int (*resize_cq)(struct ib_cq *cq, int cqe, struct ib_udata *udata);
+	/**
+	 * pre_destroy_cq - Prevent a cq from generating any new work
+	 * completions, but not free any kernel resources
+	 */
+	int (*pre_destroy_cq)(struct ib_cq *cq);
+	/**
+	 * post_destroy_cq - Free all kernel resources
+	 */
+	void (*post_destroy_cq)(struct ib_cq *cq);
 	struct ib_mr *(*get_dma_mr)(struct ib_pd *pd, int mr_access_flags);
 	struct ib_mr *(*reg_user_mr)(struct ib_pd *pd, u64 start, u64 length,
 				     u64 virt_addr, int mr_access_flags,
-- 
2.49.0