Re: [PATCH net 4/4] mlx5: Don't accidentally set RTO_ONLINK before mlx5e_route_lookup_ipv4_get()

[Saeed Mahameed <saeed@kernel.org>]

On Wed, Jan 05, 2022 at 08:56:28PM +0100, Guillaume Nault wrote:
>Mask the ECN bits before calling mlx5e_route_lookup_ipv4_get(). The
>tunnel key might have the last ECN bit set. This interferes with the
>route lookup process as ip_route_output_key_hash() interpretes this bit
>specially (to restrict the route scope).
>
>Found by code inspection, compile tested only.
>
>Fixes: c7b9038d8af6 ("net/mlx5e: TC preparation refactoring for routing update event")
>Fixes: 9a941117fb76 ("net/mlx5e: Maximize ip tunnel key usage on the TC offloading path")
>Signed-off-by: Guillaume Nault <gnault@redhat.com>
>---
> drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c | 5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
>diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
>index a5e450973225..bc5f1dcb75e1 100644
>--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
>+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
>@@ -1,6 +1,7 @@
> /* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
> /* Copyright (c) 2018 Mellanox Technologies. */
>
>+#include <net/inet_ecn.h>
> #include <net/vxlan.h>
> #include <net/gre.h>
> #include <net/geneve.h>
>@@ -235,7 +236,7 @@ int mlx5e_tc_tun_create_header_ipv4(struct mlx5e_priv *priv,
> 	int err;
>
> 	/* add the IP fields */
>-	attr.fl.fl4.flowi4_tos = tun_key->tos;
>+	attr.fl.fl4.flowi4_tos = tun_key->tos & ~INET_ECN_MASK;

This is TC control path, why would ecn bits be ON in TC act->tunnel info ?
I don't believe these bits are on and if they were, TC tunnels layer should
clear them before calling the driver's offload callback.