Re: [PATCH RFC net-next] pktgen: introduce 'rx' mode

[Alexei Starovoitov <ast@plumgrid.com>]

On 4/28/15 9:14 PM, Eric Dumazet wrote:
> On Tue, 2015-04-28 at 19:11 -0700, Alexei Starovoitov wrote:
>
>> +	if (pkt_dev->flags & F_DO_RX) {
>> +		local_bh_disable();
>> +		atomic_add(burst, &pkt_dev->skb->users);
>> +		do {
>> +			ret = netif_receive_skb(pkt_dev->skb);
>> +			if (ret == NET_RX_DROP)
>> +				pkt_dev->errors++;
>> +			pkt_dev->last_ok = 1;
>> +			pkt_dev->sofar++;
>> +			pkt_dev->seq_num++;
>> +		} while (--burst > 0);
>> +		local_bh_enable();
>> +		goto out;
>> +	}
>> +
>
> This looks buggy.
>
> skb can be put on a queue, so skb->next and skb->prev cannot be reused,
> or queues will be corrupted.

don't see the bug yet.
Any layer that wants to do such queueing should do skb_share_check
first. Just like ip_rcv does. So everything in IP world should
work fine, because it will be operating on clean cloned skb.

> Note that on TX, it is possible to have the same issue if you use a
> virtual device like bonding, and skb is queued on a slave qdisc.
>
> (Thats why we have this IFF_TX_SKB_SHARING flag)

yep, that's why xmit into veth is not useful for benchmarking of rx,
since the hottest functions are alloc_skb and fill_packet, while
netif_receive_skb is not even seen in top 10.