2009-06-26 14:04:17

by Gallus

[permalink] [raw]
Subject: Kernel hard_start_xmit performance vs SOCK_RAW performance

Hi,
what path would you choose to rapidly generate packets:
1. Develop kernel space module and user space client tool.
The tool would call the module through ioctl() passing the packet
contents. The module would then construct sbk buffer and call
hard_start_xmit.

2. Develop only user space client tool
The tool would just sent packets through raw sockets.


I'm wondering if amount of system calls in (1) solution will not be a problem?

Which path is generally better?

--
Regards,
Gallus


2009-06-26 14:07:29

by Alan

[permalink] [raw]
Subject: Re: Kernel hard_start_xmit performance vs SOCK_RAW performance

On Fri, 26 Jun 2009 16:04:04 +0200
Gallus <[email protected]> wrote:

> Hi,
> what path would you choose to rapidly generate packets

Documentation/networking/pktgen.txt

2009-06-26 14:35:40

by Gallus

[permalink] [raw]
Subject: Re: Kernel hard_start_xmit performance vs SOCK_RAW performance

2009/6/26 Alan Cox <[email protected]>:
> On Fri, 26 Jun 2009 16:04:04 +0200
> Gallus <[email protected]> wrote:
>
>> Hi,
>> what path would you choose to rapidly generate packets
>
> Documentation/networking/pktgen.txt

Thank you Alan for the answer.

However, my goal is to generate packets with certain *content* in them.

--
Regards,
Gallus
please CC me

2009-06-26 14:48:07

by Peter Chacko

[permalink] [raw]
Subject: Re: Kernel hard_start_xmit performance vs SOCK_RAW performance

Did you try using PACKET socket interface from your module ?
(essentially as fast as hard-start-transmit() path ...there will be
tools that can be used for this purpose....I am not sure of any such
tool though...:-) you need to skip the scheduling (qdiscs) path also..

I think, Linux-net mailing list would be more appropriate to have this
discussion.

Thanks,
Peter

NetDiox computing systems
Bangalore-india
http://www.netdiox.com
080 2664 0708

On Fri, Jun 26, 2009 at 8:05 PM, Gallus<[email protected]> wrote:
> 2009/6/26 Alan Cox <[email protected]>:
>> On Fri, 26 Jun 2009 16:04:04 +0200
>> Gallus <[email protected]> wrote:
>>
>>> Hi,
>>> what path would you choose to rapidly generate packets
>>
>> Documentation/networking/pktgen.txt
>
> Thank you Alan for the answer.
>
> However, my goal is to generate packets with certain *content* in them.
>
> --
> Regards,
> Gallus
> please CC me
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at ?http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at ?http://www.tux.org/lkml/
>