Re: [PATCH v4 net-next 08/10] net/ncsi: Support NCSI packet generation

[Gavin Shan <gwshan@linux.vnet.ibm.com>]

On Thu, May 04, 2017 at 02:00:35PM +0200, Andrew Lunn wrote:
>On Thu, May 04, 2017 at 04:31:57PM +1000, Gavin Shan wrote:
>> On Wed, May 03, 2017 at 02:52:54PM +0200, Andrew Lunn wrote:
>> >On Wed, May 03, 2017 at 02:44:39PM +1000, Gavin Shan wrote:
>> >> This introduces /sys/kernel/debug/ncsi/eth0/pkt. The debugfs entry
>> >> can accept parameters to produce NCSI command packet. The received
>> >> NCSI response packet is dumped on read. Below is an example to send
>> >> CIS command and dump its response.
>> >> 
>> >>    # echo CIS,0,0 > /sys/kernel/debug/ncsi/eth0/pkt
>> >>    # cat /sys/kernel/debug/ncsi/eth0/pkt
>> >>    NCSI response [CIS] packet received
>> >> 
>> >>    00 01 dd 80 00 0004 0000 0000
>> >
>> >Could this be done with a raw socket for Tx and
>> >libpcap/tcpdump/wireshart for Rx?
>> >
>> 
>> Andrew, it's really good question. Unfortunately, I don't think it can
>> be done solely by raw socket to transmit NCSI command packet because the
>> [packet sequence number] field in the packet can't be same to any used ones.
>> Otherwise the remote NIC will be confused and start to reponse abnormally.
>
>Just to make sure i'm on the same page. We are talking about:
>
>https://www.dmtf.org/sites/default/files/standards/documents/DSP0222_1.0.1.pdf
>
>and the sequence number is the Instance ID. This is an 8-bit number,
>and if it receives a message with the previously used IID, it should
>assume it is a re-transmit of the request and send back the old
>response. It is a very simple scheme, no windowing, only one
>outstanding command/response, just one response buffered in case of a
>re-transmit.
>
>> We could reserve some sequence number to be used by raw socket.
>
>I don't think that works. You can only have one command
>'inflight'. Packets from a raw socket would have to go through your
>state machine. Which makes it complex.
>
>libpcap should however still work. So you should probably do all the
>receive side in user space.
>

Andrew, yeah, we're on same page about the sequence number. Yes, I
agree it's going to make thing complex to transmit packet through
raw socket. So lets keep using debugfs as we had.

I need to dig how libpcap receives packets. It's appreciated if you
can give some hints about that. However, I don't see the benefit to
receive packets by libpcap, could you claim?

Cheers,
Gavin