Re: [PATCH v5 1/2] Renesas Ethernet AVB driver proper

[Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>]

Hello.

On 06/04/2015 06:12 AM, David Miller wrote:

>> +		/* Received network control queue */
>> +		if (ris0 & RIS0_FRF1) {
>> +			ravb_write(ndev, ~RIS0_FRF1, RIS0);
>> +			/* Timestamp of network control packets that is based
>> +			 * on IEEE802.1AS, is used time synchronization of PTP.
>> +			 * It should not be handled by NAPI scheduling, because
>> +			 * it needs to be received as soon as possible.
>> +			 */
>> +			ravb_rx(ndev, NULL, RAVB_NC);
>> +			result = IRQ_HANDLED;
>> +		}

> Nobody else makes this distinction, all packets should be processed
> via your NAPI path.

    I see.

> When I see people conditionally invoking netif_rx()
> vs. netif_receive_skb() in their packet receive routine, like
> conditional locking, it's a big red flag.

> Furthermore, you should pass the NAPI context into ravb_rx() and use

    I guess I should have one NAPI context per RX queue?

> it so that you can invoke napi_gro_receive() on all of the packets and
> therefore support GRO.

    Hmm, I don't think I have a hardware support for generic RX offload...
I'm still unsure what it is anyway -- is it used for IP packet defragmentation?

WBR, Sergei