Re: [PATCH v2 net-next 1/3] geneve: pass geneve_config pointer to helper functions

[Simon Horman <horms@kernel.org>]

On Tue, Jul 07, 2026 at 02:53:28PM +0000, Eric Dumazet wrote:
> In preparation for converting geneve->cfg to an RCU-protected pointer,
> update helper functions to explicitly accept a const struct geneve_config
> pointer instead of dereferencing geneve->cfg directly.
> 
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Suggested-by: Paolo Abeni <pabeni@redhat.com>
> ---
>  drivers/net/geneve.c | 140 +++++++++++++++++++++++--------------------
>  1 file changed, 76 insertions(+), 64 deletions(-)
> 
> diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c

...

> @@ -1306,14 +1313,14 @@ static int geneve_build_skb(struct dst_entry *dst, struct sk_buff *skb,
>  }
>  
>  static u8 geneve_get_dsfield(struct sk_buff *skb, struct net_device *dev,
> +			     const struct geneve_config *cfg,
>  			     const struct ip_tunnel_info *info,
>  			     bool *use_cache)
>  {
> -	struct geneve_dev *geneve = netdev_priv(dev);
>  	u8 dsfield;
>  
>  	dsfield = info->key.tos;
> -	if (dsfield == 1 && !geneve->cfg.collect_md) {
> +	if (cfg && dsfield == 1 && !cfg->collect_md) {

Hi Eric,

Sashiko.dev seems to think that the cfg guard is purely defensive.
Which does seem to be the case.
Perhaps it is an artifact from the development of v2?

>  		dsfield = ip_tunnel_get_dsfield(ip_hdr(skb), skb);
>  		*use_cache = false;
>  	}

...