Re: [PATCH 1/1] net: hold sock reference while processing tx timestamps

[Richard Cochran <richardcochran@gmail.com>]

On Wed, Oct 12, 2011 at 09:27:53PM +0200, Johannes Berg wrote:
> On Wed, 2011-10-12 at 20:36 +0200, Richard Cochran wrote:
> > The pair of functions,
> > 
> >  * skb_clone_tx_timestamp()
> >  * skb_complete_tx_timestamp()
...
> 
> This still needs a fix to the PHY driver right? It has a case that can
> kfree_skb() the skb instead of passing it back to
> complete_tx_timestamp().

Yes, right, will do.

> > -	if (!hwtstamps)
> > +	if (!hwtstamps) {
> > +		sock_put(sk);
> > +		kfree_skb(skb);
> >  		return;
> > +	}
> 
> Is that right w/o skb->sk = NULL?

I think it is okay. In clone/complete, we use skb->sk in a special
way, just to remember the socket. Within kfree_skb, only the
skb->destructor would possibly use skb->sk, and that function pointer
is NULL (after the clone in the Tx path).

> FWIW I just decided to do it the other way around in mac80211 -- keep
> the original SKB that's charged to the socket for the error queue, and
> use a clone to actually do the TX.

In this case, we must clone, I think. The Ethernet MAC driver will
call the clone_tx_timetstamp hook just before freeing the original
skb.

Thanks,
Richard