Re: [PATCHv4 net-next 1/4] net: add a common function to compute features from lowers devices

[Simon Horman <horms@kernel.org>]

On Tue, Oct 14, 2025 at 08:02:14AM +0000, Hangbin Liu wrote:

...

> diff --git a/net/core/dev.c b/net/core/dev.c
> index a64cef2c537e..54f0e792fbd2 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -12616,6 +12616,101 @@ netdev_features_t netdev_increment_features(netdev_features_t all,
>  }
>  EXPORT_SYMBOL(netdev_increment_features);
>  
> +/**
> + *	netdev_compute_features_from_lowers - compute feature from lowers
> + *	@dev: the upper device
> + *	@update_header: whether to update upper device's header_len/headroom/tailroom
> + *
> + *	Recompute the upper device's feature based on all lower devices.
> + */
> +void netdev_compute_features_from_lowers(struct net_device *dev, bool update_header)
> +{
> +	unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM;
> +	netdev_features_t gso_partial_features = VIRTUAL_DEV_GSO_PARTIAL_FEATURES;
> +#ifdef CONFIG_XFRM_OFFLOAD
> +	netdev_features_t xfrm_features = VIRTUAL_DEV_XFRM_FEATURES;
> +#endif

Hi Hangbin,

It would be nice to avoid the #ifdefs in this function.

Could xfrm_features be declared unconditoinally.
And then used behind if(IS_ENABLED(CONFIG_XFRM_OFFLOAD)) conditions?
This would increase compile coverage (and readability IMHO).

> +	netdev_features_t mpls_features = VIRTUAL_DEV_MPLS_FEATURES;
> +	netdev_features_t vlan_features = VIRTUAL_DEV_VLAN_FEATURES;
> +	netdev_features_t enc_features = VIRTUAL_DEV_ENC_FEATURES;
> +	unsigned short max_header_len = ETH_HLEN;
> +	unsigned int tso_max_size = TSO_MAX_SIZE;
> +	u16 tso_max_segs = TSO_MAX_SEGS;
> +	struct net_device *lower_dev;
> +	unsigned short max_headroom;
> +	unsigned short max_tailroom;
> +	struct list_head *iter;
> +
> +	mpls_features = netdev_base_features(mpls_features);
> +	vlan_features = netdev_base_features(vlan_features);
> +	enc_features = netdev_base_features(enc_features);
> +
> +	netdev_for_each_lower_dev(dev, lower_dev, iter) {
> +		gso_partial_features = netdev_increment_features(gso_partial_features,
> +								 lower_dev->gso_partial_features,
> +								 VIRTUAL_DEV_GSO_PARTIAL_FEATURES);
> +
> +		vlan_features = netdev_increment_features(vlan_features,
> +							  lower_dev->vlan_features,
> +							  VIRTUAL_DEV_VLAN_FEATURES);
> +
> +		enc_features = netdev_increment_features(enc_features,
> +							 lower_dev->hw_enc_features,
> +							 VIRTUAL_DEV_ENC_FEATURES);
> +
> +#ifdef CONFIG_XFRM_OFFLOAD
> +		xfrm_features = netdev_increment_features(xfrm_features,
> +							  lower_dev->hw_enc_features,
> +							  VIRTUAL_DEV_XFRM_FEATURES);
> +#endif
> +
> +		mpls_features = netdev_increment_features(mpls_features,
> +							  lower_dev->mpls_features,
> +							  VIRTUAL_DEV_MPLS_FEATURES);
> +
> +		dst_release_flag &= lower_dev->priv_flags;
> +
> +		if (update_header) {
> +			max_header_len = max_t(unsigned short, max_header_len,
> +					lower_dev->hard_header_len);

Both the type of max_header_len and .hard_header_len is unsigned short.
So I think max() can be used here instead of max_t(). Likewise for the
following two lines.

> +			max_headroom = max_t(unsigned short, max_headroom,
> +					lower_dev->needed_headroom);

Max Headroom [1] is used uninitialised the first time we reach here.
Likewise for max_tailroom below.

[1] https://en.wikipedia.org/wiki/Max_Headroom

Flagged by Smatch.

> +			max_tailroom = max_t(unsigned short, max_tailroom,
> +					lower_dev->needed_tailroom);
> +		}
> +
> +		tso_max_size = min(tso_max_size, lower_dev->tso_max_size);
> +		tso_max_segs = min(tso_max_segs, lower_dev->tso_max_segs);
> +	}
> +
> +	dev->gso_partial_features = gso_partial_features;
> +	dev->vlan_features = vlan_features;
> +	dev->hw_enc_features = enc_features | NETIF_F_GSO_ENCAP_ALL |
> +				    NETIF_F_HW_VLAN_CTAG_TX |
> +				    NETIF_F_HW_VLAN_STAG_TX;
> +#ifdef CONFIG_XFRM_OFFLOAD
> +	dev->hw_enc_features |= xfrm_features;
> +#endif
> +	dev->mpls_features = mpls_features;
> +
> +	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
> +	if ((dev->priv_flags & IFF_XMIT_DST_RELEASE_PERM) &&
> +	    dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM))
> +		dev->priv_flags |= IFF_XMIT_DST_RELEASE;
> +
> +	if (update_header) {
> +		dev->hard_header_len = max_header_len;
> +		dev->needed_headroom = max_headroom;
> +		dev->needed_tailroom = max_tailroom;

Also, maybe it can't happen in practice. But I think that max_headroom and
max_tailroom will may be used uninitialised here if the previous
'update_header' condition is never reached/met.

Also flagged by Smatch.

> +	}
> +
> +	netif_set_tso_max_segs(dev, tso_max_segs);
> +	netif_set_tso_max_size(dev, tso_max_size);
> +
> +	netdev_change_features(dev);
> +}
> +EXPORT_SYMBOL(netdev_compute_features_from_lowers);
> +

...