On Thursday 14 February 2008 20:48, Roland Dreier wrote:
> by the way, it seems we never release ICM table entries when freeing
> MPTs. Does the patch below make sense to you?
>
> diff --git a/drivers/net/mlx4/mr.c b/drivers/net/mlx4/mr.c
> index 679dfdb..3ffce7a 100644
> --- a/drivers/net/mlx4/mr.c
> +++ b/drivers/net/mlx4/mr.c
> @@ -287,6 +287,8 @@ void mlx4_mr_free(struct mlx4_dev *dev, struct mlx4_mr
> *mr)
> (dev->caps.num_mpts - 1));
> if (err)
> mlx4_warn(dev, "HW2SW_MPT failed (%d)\n", err);
> +
> + mlx4_table_put(dev, &mr_table->dmpt_table,
> key_to_hw_index(mr->key));
> }
>
> mlx4_mtt_cleanup(dev, &mr->mtt);
>
I'm nervous about freeing ICM table entries when HW2SW fails (hw may still
access the
memory.
How about the patch below ?
(P.S., please note the priv-> in the patch).
- Jack
---
Index: ofed_kernel/drivers/net/mlx4/mr.c
===================================================================
--- ofed_kernel.orig/drivers/net/mlx4/mr.c 2008-02-18 09:49:28.000000000
+0200
+++ ofed_kernel/drivers/net/mlx4/mr.c 2008-02-18 10:10:21.956745000 +0200
@@ -287,6 +287,9 @@ void mlx4_mr_free(struct mlx4_dev *dev,
(dev->caps.num_mpts - 1));
if (err)
mlx4_warn(dev, "HW2SW_MPT failed (%d)\n", err);
+ else
+ mlx4_table_put(dev, &priv->mr_table.dmpt_table,
+ key_to_hw_index(mr->key);
}
mlx4_mtt_cleanup(dev, &mr->mtt);
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
