netdev - Re: [PATCH net 1/1] ice: identify aRFS flows using L3/L4 dissector info

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Sun, 9 Apr 2023 13:45:29 +0300
From:   Leon Romanovsky <leon@...nel.org>
To:     Tony Nguyen <anthony.l.nguyen@...el.com>
Cc:     davem@...emloft.net, kuba@...nel.org, pabeni@...hat.com,
        edumazet@...gle.com, netdev@...r.kernel.org,
        Ahmed Zaki <ahmed.zaki@...el.com>,
        Arpana Arland <arpanax.arland@...el.com>
Subject: Re: [PATCH net 1/1] ice: identify aRFS flows using L3/L4 dissector
 info

On Fri, Apr 07, 2023 at 02:08:20PM -0700, Tony Nguyen wrote:
> From: Ahmed Zaki <ahmed.zaki@...el.com>
> 
> The flow ID passed to ice_rx_flow_steer() is computed like this:
> 
>     flow_id = skb_get_hash(skb) & flow_table->mask;
> 
> With smaller aRFS tables (for example, size 256) and higher number of
> flows, there is a good chance of flow ID collisions where two or more
> different flows are using the same flow ID. This results in the aRFS
> destination queue constantly changing for all flows sharing that ID.
> 
> Use the full L3/L4 flow dissector info to identify the steered flow
> instead of the passed flow ID.
> 
> Fixes: 28bf26724fdb ("ice: Implement aRFS")
> Signed-off-by: Ahmed Zaki <ahmed.zaki@...el.com>
> Tested-by: Arpana Arland <arpanax.arland@...el.com> (A Contingent worker at Intel)
> Signed-off-by: Tony Nguyen <anthony.l.nguyen@...el.com>
> ---
>  drivers/net/ethernet/intel/ice/ice_arfs.c | 44 +++++++++++++++++++++--
>  1 file changed, 41 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/ice/ice_arfs.c b/drivers/net/ethernet/intel/ice/ice_arfs.c
> index fba178e07600..d7ae64d21e01 100644
> --- a/drivers/net/ethernet/intel/ice/ice_arfs.c
> +++ b/drivers/net/ethernet/intel/ice/ice_arfs.c
> @@ -345,6 +345,44 @@ ice_arfs_build_entry(struct ice_vsi *vsi, const struct flow_keys *fk,
>  	return arfs_entry;
>  }
>  
> +/**
> + * ice_arfs_cmp - compare flow to a saved ARFS entry's filter info
> + * @fltr_info: filter info of the saved ARFS entry
> + * @fk: flow dissector keys
> + *
> + * Caller must hold arfs_lock if @fltr_info belongs to arfs_fltr_list
> + */
> +static bool
> +ice_arfs_cmp(struct ice_fdir_fltr *fltr_info, const struct flow_keys *fk)
> +{
> +	bool is_ipv4;
> +
> +	if (!fltr_info || !fk)
> +		return false;
> +
> +	is_ipv4 = (fltr_info->flow_type == ICE_FLTR_PTYPE_NONF_IPV4_UDP ||
> +		fltr_info->flow_type == ICE_FLTR_PTYPE_NONF_IPV4_TCP);
> +
> +	if (fk->basic.n_proto == htons(ETH_P_IP) && is_ipv4)
> +		return (fltr_info->ip.v4.proto == fk->basic.ip_proto &&
> +			fltr_info->ip.v4.src_port == fk->ports.src &&
> +			fltr_info->ip.v4.dst_port == fk->ports.dst &&
> +			fltr_info->ip.v4.src_ip == fk->addrs.v4addrs.src &&
> +			fltr_info->ip.v4.dst_ip == fk->addrs.v4addrs.dst);
> +	else if (fk->basic.n_proto == htons(ETH_P_IPV6) && !is_ipv4)
> +		return (fltr_info->ip.v6.proto == fk->basic.ip_proto &&
> +			fltr_info->ip.v6.src_port == fk->ports.src &&
> +			fltr_info->ip.v6.dst_port == fk->ports.dst &&
> +			!memcmp(&fltr_info->ip.v6.src_ip,
> +				&fk->addrs.v6addrs.src,
> +				sizeof(struct in6_addr)) &&
> +			!memcmp(&fltr_info->ip.v6.dst_ip,
> +				&fk->addrs.v6addrs.dst,
> +				sizeof(struct in6_addr)));

I'm confident that you can write this function more clear with
comparisons in one "return ..." instruction.

Thanks

> +
> +	return false;
> +}
> +
>  /**
>   * ice_arfs_is_perfect_flow_set - Check to see if perfect flow is set
>   * @hw: pointer to HW structure
> @@ -436,17 +474,17 @@ ice_rx_flow_steer(struct net_device *netdev, const struct sk_buff *skb,
>  
>  	/* choose the aRFS list bucket based on skb hash */
>  	idx = skb_get_hash_raw(skb) & ICE_ARFS_LST_MASK;
> +
>  	/* search for entry in the bucket */
>  	spin_lock_bh(&vsi->arfs_lock);
>  	hlist_for_each_entry(arfs_entry, &vsi->arfs_fltr_list[idx],
>  			     list_entry) {
> -		struct ice_fdir_fltr *fltr_info;
> +		struct ice_fdir_fltr *fltr_info = &arfs_entry->fltr_info;
>  
>  		/* keep searching for the already existing arfs_entry flow */
> -		if (arfs_entry->flow_id != flow_id)
> +		if (!ice_arfs_cmp(fltr_info, &fk))
>  			continue;
>  
> -		fltr_info = &arfs_entry->fltr_info;
>  		ret = fltr_info->fltr_id;
>  
>  		if (fltr_info->q_index == rxq_idx ||
> -- 
> 2.38.1
>