RDMA/core: Stop DIM before destroying CQ (5d46b289) · Commits · 戴 / test

HW destroy operation should be last operation after all possible CQ users completed their work, so move DIM work cancellation before such destroy call. Fixes: ("RDMA/core: Provide RDMA DIM support for ULPs") Link: https://lore.kernel.org/r/20200730082719.1582397-3-leon@kernel.org Reviewed-by:

Max Gurtovoy <maxg@mellanox.com> Signed-off-by:

Leon Romanovsky <leonro@mellanox.com> Signed-off-by:

Jason Gunthorpe <jgg@nvidia.com>

drivers/infiniband/core/cq.c

+10 −3

Original line number	Diff line number	Diff line
		@@ -72,6 +72,15 @@ static void rdma_dim_init(struct ib_cq *cq)
		INIT_WORK(&dim->work, ib_cq_rdma_dim_work);
		}

		static void rdma_dim_destroy(struct ib_cq *cq)
		{
		if (!cq->dim)
		return;

		cancel_work_sync(&cq->dim->work);
		kfree(cq->dim);
		}

		static int __poll_cq(struct ib_cq cq, int num_entries, struct ib_wc wc)
		{
		int rc;
		@@ -331,12 +340,10 @@ void ib_free_cq_user(struct ib_cq cq, struct ib_udata udata)
		WARN_ON_ONCE(1);
		}

		rdma_dim_destroy(cq);
		trace_cq_free(cq);
		rdma_restrack_del(&cq->res);
		cq->device->ops.destroy_cq(cq, udata);
		if (cq->dim)
		cancel_work_sync(&cq->dim->work);
		kfree(cq->dim);
		kfree(cq->wc);
		kfree(cq);
		}

Admin message