When DPDK copies flow items with the `rte_flow_conv` call, it assumes
that size of PMD private data has pointer size - see patch [1]
MLX5 PMD defined `struct mlx5_rte_flow_item_sq` as 32 bits.
As the result, on 64 bits systems, when DPDK copied
MLX5_RTE_FLOW_ITEM_TYPE_SQ item, the target buffer was assigned
additional 32 bits.
The patch expands size of `struct mlx5_rte_flow_item_sq` to 64 bits
on 64 bits systems.
[1]:
commit 6cf72047332b ("ethdev: support flow elements with variable length")
Fixes: 75a00812b18f ("net/mlx5: add hardware steering item translation")
Cc: sta...@dpdk.org
Signed-off-by: Gregory Etelson <getel...@nvidia.com>
Acked-by: Dariusz Sosnowski <dsosnow...@nvidia.com>
---
drivers/net/mlx5/mlx5_flow.h | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
index db56ae051d..9cf54c3a6a 100644
--- a/drivers/net/mlx5/mlx5_flow.h
+++ b/drivers/net/mlx5/mlx5_flow.h
@@ -168,6 +168,9 @@ struct mlx5_flow_action_copy_mreg {
/* Matches on source queue. */
struct mlx5_rte_flow_item_sq {
uint32_t queue; /* DevX SQ number */
+#ifdef RTE_ARCH_64
+ uint32_t reserved;
+#endif
};
/* Map from registers to modify fields. */
--
2.43.0
