Re: [dpdk-dev] [PATCH 04/12] mbuf: add function to calculate a checksum

[Olivier Matz <olivier.matz@6wind.com>]

Dear Don,

On 07/21/2016 06:26 PM, Don Provan wrote:
>> -----Original Message-----
>> From: Ananyev, Konstantin [mailto:konstantin.ananyev@intel.com]
>> Sent: Thursday, July 21, 2016 3:51 AM
>> Subject: Re: [dpdk-dev] [PATCH 04/12] mbuf: add function to calculate a
>> checksum
>>
>> ...
>>> +  Added a new function ``rte_pktmbuf_cksum()`` to process the checksum
>>> + of  data embedded in an mbuf chain.
>>> ...
>>> +#include <rte_ip.h>
>>
>> As a nit, do we need to introduce a dependency for librte_mbuf on librte_net?
>> Might be better to put this functionality into librte_net?
> 
> That's not a nit at all. This is clearly a net function that has no place in the mbuf code.
> That should be obvious even before we notice this circular dependency.


The function is called rte_pktmbuf_cksum(), and takes a mbuf as a
parameter. You cannot haughtily say "it no place in the mbuf code".

As you can see, librte_net only contains headers files. The initial goal
of librte_net was to contain network headers and nothing more.
See:
http://dpdk.org/browse/dpdk/commit/lib/librte_net?id=af75078fece3615088e561357c1e97603e43a5fe

To me, the question of having a dependency in one direction or another
(librte_net needs librte_mbuf, or librte_mbuf needs librte_net) is an
open debate.

I've asked myself the same question when software packet type parsing,
that needs definitions of network headers. As packet_type is a pure mbuf
notion, my choice was to have this parse in mbuf library, and using
network headers definitions provided by librte_net.