netdev - Re: [PATCH net-next] dpaa2-eth: Don't use netif_receive_skb

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]

Message-ID: <49d021a6-46de-f37e-8097-2836d0d31930@gmail.com>
Date:   Wed, 24 Jul 2019 09:10:20 +0200
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     Ioana Radulescu <ruxandra.radulescu@....com>,
        netdev@...r.kernel.org, davem@...emloft.net
Cc:     ioana.ciornei@....com, vladimir.oltean@....com
Subject: Re: [PATCH net-next] dpaa2-eth: Don't use netif_receive_skb_list for
 TCP frames



On 7/23/19 7:28 PM, Ioana Radulescu wrote:
> Using Rx skb bulking for all frames may negatively impact the
> performance in some TCP termination scenarios, as it effectively
> bypasses GRO.

>  
> -	list_add_tail(&skb->list, ch->rx_list);
> +	if (frame_is_tcp(fd, fas))
> +		napi_gro_receive(&ch->napi, skb);
> +	else
> +		list_add_tail(&skb->list, ch->rx_list);
>  
>  	return;
>  


This is really bad.

This is exactly why I suggested to add the batching capability to GRO,
instead having to change all drivers.

Edward Cree is working on this.