[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20100525102603.GA11494@gondor.apana.org.au>
Date: Tue, 25 May 2010 20:26:03 +1000
From: Herbert Xu <herbert@...dor.apana.org.au>
To: jamal <hadi@...erus.ca>
Cc: Jiri Pirko <jpirko@...hat.com>, netdev@...r.kernel.org,
davem@...emloft.net, kaber@...sh.net
Subject: Re: Question about an assignment in handle_ing()
On Tue, May 25, 2010 at 05:51:07AM -0400, jamal wrote:
> The code is correct.
>
> Main reason for the else condn is driven by optimization:
> If you are running tcpdump or other af packet type code, the "if" condn
> is hit and matching actions are not allowed trample on that same packet
> data. They have to make a private copy; otherwise, the "else" is hit and
> (for optimization reason) you give ok to the actions that follow to
> munge the packet. Essentially, you dont want actions to alloc/copy every
> single time when you are not running tcpdump for example; reason is that
> most of the time you run tcpdump it is for debugging.
> [I had seen very observable differences on some old mips board back in
> the day on whether you avoided copy every time vs when debugging by
> running tcpdump and copied every packet.]
In that case you should be checking whether the skb is cloned.
After all, tcpdump might have simply filtered the packet out.
BTW, this is the case whenever you run a DHCP client/server. So
on most boxes your optimisation will never kick in as is. Also
the skb may still be cloned even if there is no AF_PACKET listener.
In that case your optimisation may be incorrect.
Cheers,
--
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <herbert@...dor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
--
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