On Fri, Feb 27, 2009 at 4:10 PM, David Miller <davem@davemloft.net> wrote:
quoted text
> From: "Duyck, Alexander H" <alexander.h.duyck@intel.com>
> Date: Fri, 27 Feb 2009 08:39:30 -0800
>
>> The fact is, if you are planning to kill it further up the stack,
>> you can put that down as another reason for removing the call.  The
>> timesync / killing skb->sk is a separate discussion that I don't
>> know if I even really need to be involved in.  The patch doesn't
>> break anything, and removes a call to skb_orphan that shouldn't
>> really be there in the first place.
>
> That's not true at all.
>
> One reason we haven't removed it yet is because there are some
> semantic issues to address, which are also effected by drivers
> doing this kind of by-hand skb_orphan()'ing.
>
> For one thing, if you skb_orphan() in your transmit handler,
> a UDP application can take over your network card by just
> transmitting in an endless tight loop.  It would be able to
> nearly keep all other users from using the card.
>
> That happens because if you skb_orphan(), nothing controls the
> pace at which datagram applications can send.  Currently
> we need the socket send buffer limits to help us out here,
> but if you skb_orphan() you completely subvert that proptection.
>
> So I will not apply this patch.
> --

So the removal of the skb_orphan() in the driver should be a good
thing, the only issue is with the patch description?  By cleaning up
the patch description, will that make this patch acceptable for
everyone?

-- 
Cheers,
Jeff
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html