On Wednesday 30 July 2008, Patrick Ohly wrote:

quoted

I don't think that skb->sk is guaranteed to be around in hard_start_xmit.
But we should not need it, if we overload the skb->tstamp, right?

skb->sk is needed in skb_tx_timestamp() to send the copied skb with the
time stamp back to the application, isn't it? skb_tx_timestamp() won't
crash, but it won't be able to do the desired action either. Overloading
skb->tstamp doesn't help with that.

Yes you are right :)

I don't know where the connection to the originating socket is lost, but
in linux-2.6 as of one week ago (not 2.6.26, as I said above) the
skb->sk pointer is indeed NULL for the PTP broadcast packets - I just
checked again.

Thanks for letting me know, I will look into it.

tavi