Re: [dpdk-dev] [PATCH] app/testpmd: Document what the application does.

[Ben Pfaff <blp@ovn.org>]

On Tue, Sep 14, 2021 at 08:58:31AM +0100, Ferruh Yigit wrote:
> On 9/10/2021 6:57 PM, Ben Pfaff wrote:
> > I could not find anything in the documentation that says what
> > testpmd does.  This should save other people time trying to
> > figure that out in the future.
> > 
> > Signed-off-by: Ben Pfaff <blp@ovn.org>
> > ---
> >  doc/guides/testpmd_app_ug/run_app.rst | 6 ++++++
> >  1 file changed, 6 insertions(+)
> > 
> > diff --git a/doc/guides/testpmd_app_ug/run_app.rst b/doc/guides/testpmd_app_ug/run_app.rst
> > index 6061674239..7c3406f72b 100644
> > --- a/doc/guides/testpmd_app_ug/run_app.rst
> > +++ b/doc/guides/testpmd_app_ug/run_app.rst
> > @@ -15,6 +15,12 @@ a list of available EAL command-line options.
> >  Testpmd Command-line Options
> >  ----------------------------
> >  
> > +By default, testpmd receives packets on each configured port and
> > +forwards the received packets to its paired port.  Ports 0 and 1 are
> > +paired, as are ports 2 and 3, and so on.  With an odd number of ports,
> > +the last port is paired with itself: packets received on the port are
> > +sent back out on the same port.
> Hi Ben,
> 
> testpmd has the concept of 'forwarding engine' (struct fwd_engine), which is
> decoupled from testpmd logic and can be changed in the runtime. What you
> described above is the 'io' (default) forwarding engine.
> There are forwarding engines like 'flowgen' that generates and send L3 packets,
> so behaves like a very simple packet generator, etc...
> 
> And even for the 'io' forwarding engine, the paired port description above is
> controlled by "--port-topology=<paired|chained|loop>", so for example if the
> 'chained' topology is selected, packets will be forwarded as 0 -> 1 -> 2 -> 3 ..
> 
> 
> Overall, I understand the intention to briefly describe the testpmd, but I think
> it is not correct to reduce the description to packet forwarding, although that
> is an important function of testpmd, it is not only function and testpmd does
> much more, it has many control path functions.

I think that someone who maintains testpmd needs to take this patch
over, then.  I don't know it well enough.