zero copy with PACKET_MMAP

zero copy with PACKET_MMAP

[ratheesh kannoth]

Hi ,

The link

https://www.kernel.org/doc/Documentation/networking/packet_mmap.txt
talks about  achieving zero copy. But i find a deep  packet copy ( of
length snaplen= skb->len )  here
http://lxr.free-electrons.com/source/net/packet/af_packet.c#L1841.

Is there anything missing/wrong in my understanding ?


Thanks,
Ratheesh

Re: zero copy with PACKET_MMAP

[Daniel Borkmann]

On 02/27/2014 07:40 AM, ratheesh kannoth wrote:
...
> https://www.kernel.org/doc/Documentation/networking/packet_mmap.txt
> talks about  achieving zero copy. But i find a deep  packet copy ( of
> length snaplen= skb->len )  here
> http://lxr.free-electrons.com/source/net/packet/af_packet.c#L1841.
>
> Is there anything missing/wrong in my understanding ?

The zero-copy is between user and kernel space.

zero copy with PACKET_MMAP

[Jeff Haran]

> -----Original Message-----
> From: kernelnewbies-bounces at kernelnewbies.org [mailto:kernelnewbies-bounces at kernelnewbies.org] On Behalf Of ratheesh
> kannoth
> Sent: Wednesday, February 26, 2014 11:28 PM
> To: kernelnewbies at kernelnewbies.org
> Subject: Fwd: zero copy with PACKET_MMAP
> 
> Hi ,
> 
> The link
> 
> https://www.kernel.org/doc/Documentation/networking/packet_mmap.txt
> talks about  achieving zero copy. But i find a deep  packet copy ( of
> length snaplen= skb->len )  here
> http://lxr.free-electrons.com/source/net/packet/af_packet.c#L1841.
> 
> Is there anything missing/wrong in my understanding ?
> 
> 
> Thanks,
> Ratheesh

It's "zero copy" in that it is not necessary to copy the contents of the kernel resident socket buffer to user space in order for user space to read it.

But you are not missing anything. There is still a copy involved in order to copy the packet to the socket ring buffer.

Jeff Haran