[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20151030203201.GB27879@netboy>
Date: Fri, 30 Oct 2015 21:32:01 +0100
From: Richard Cochran <richardcochran@...il.com>
To: Stefan Sørensen
<stefan.sorensen@...ctralink.com>
Cc: davem@...emloft.net, netdev@...r.kernel.org
Subject: Re: [PATCH net-next 1/5] dp83640: Include hash in timestamp/packet
matching
On Fri, Oct 30, 2015 at 01:14:00PM +0100, Stefan Sørensen wrote:
> Only using the message type and sequence id for matching timestamps with
> packets is error prone, particularly if packets can be reordered. Fix by
> extending the check to include the hash of bytes 20-29 (source id in PTPv2)
> that is provided with the timestamps.
The example of reordered packets is bogus, since the sequence numbers
are not affected. The one case that benefits from the hash check is
when the port is in the master role and multiple clients send
Delay_Req messages all with the same sequence number by chance.
> @@ -819,11 +820,18 @@ static int match(struct sk_buff *skb, unsigned int type, struct rxts *rxts)
> msgtype = data + offset + OFF_PTP_CONTROL;
> else
> msgtype = data + offset;
> + if (rxts->msgtype != (*msgtype & 0xf))
> + return 0;
>
> seqid = (u16 *)(data + offset + OFF_PTP_SEQUENCE_ID);
> + if (rxts->seqid != ntohs(*seqid))
> + return 0;
> +
> + hash = ether_crc(10, data + offset + 20) >> 20;
This could use a macro instead of magic 20.
Thanks,
Richard
> + if (rxts->hash != hash)
> + return 0;
>
> - return rxts->msgtype == (*msgtype & 0xf) &&
> - rxts->seqid == ntohs(*seqid);
> + return 1;
> }
>
> static void decode_rxts(struct dp83640_private *dp83640,
> --
> 2.5.0
>
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists