[dpdk-dev] [PATCH] ethdev: additional parameter in RX callback

[Bruce Richardson]

On Thu, Mar 12, 2015 at 03:15:40PM -0400, Neil Horman wrote:
> On Thu, Mar 12, 2015 at 04:54:27PM +0000, John McNamara wrote:
> > 
> > This patch is a minor extension to the recent patchset for RX/TX callbacks
> > based on feedback from users implementing solutions based on it.
> > 
> > The patch adds a new parameter to the RX callback to pass in the number of
> > available RX packets in addition to the number of dequeued packets.
> > This provides the RX callback functions with additional information
> > that can be used to decide how packets from a burst are handled.
> > 
> > The TX callback doesn't require this additional parameter so the RX
> > and TX callbacks no longer have the same function parameters. As such
> > the single RX/TX callback has been refactored into two separate callbacks.
> > 
> > Since this is an API change we hope it can be included in 2.0.0 to avoid
> > changing the API in a subsequent release.    
> > 
> > 
> > John McNamara (1):
> >   ethdev: added additional packet count parameter to RX callbacks
> > 
> >  examples/rxtx_callbacks/main.c |    3 +-
> >  lib/librte_ether/rte_ethdev.c  |    8 ++--
> >  lib/librte_ether/rte_ethdev.h  |   74 ++++++++++++++++++++++++++--------------
> >  3 files changed, 54 insertions(+), 31 deletions(-)
> > 
> > -- 
> > 1.7.4.1
> > 
> > 
> 
> 
> Well, we're well past the new feature phase of this cycle, so I would say NACK.
> I would also suggest that you don't need to modify ABI to accomodate this
> feature.  Instead just document the pkts array to be terminated by a reserved
> value, so that the callback can determine its size dynamically.  You could
> alternatively create a new api call that allows you to retrieve that information
> from the context of the callback.
> 
> Neil
> 

Yes, I would agree we are past the new feature phase. However, given that we
are making a change to the API, and a fairly small change too - adding one extra
parameter - we think that the benefit of including this now outweighs any risk
of merging the patch. It seems a bit crazy to ship a release with a new API and
then immediately change the API straight after release. Is it not better to
take the received feedback on the API and fix/improve it pre-release before it
gets set-in-stone?

/Bruce