Re: [PATCH v6] app/testpmd: expand noisy neighbour forward mode support

[Mike Pattrick <mkp@redhat.com>]

On Thu, Jun 8, 2023 at 6:25 AM Ferruh Yigit <ferruh.yigit@amd.com> wrote:
>
> On 6/8/2023 10:59 AM, Mike Pattrick wrote:
> > Previously the noisy neighbour vnf simulation would only operate in io
> > mode, forwarding packets as is. However, this limited the usefulness of
> > noisy neighbour simulation.
> >
> > This feature has now been expanded to supporting mac, macswap, and
> > 5tswap modes. To facilitate adding this support, some new header files
> > were added.
> >
> > Signed-off-by: Mike Pattrick <mkp@redhat.com>
> >  ---
> >  v2: Reverted changes to random memory lookup
> >  v3: Refactored entire patch
> >  v4: Implemented recommended formatting changes
> >  v5: Corrected copyright statement and formatting changes
> >  v6: Reordered some variables, preserved noisy subtype for display
> > ---
> >  app/test-pmd/5tswap.c                 | 118 +----------------------
> >  app/test-pmd/5tswap.h                 | 130 ++++++++++++++++++++++++++
> >  app/test-pmd/config.c                 |  18 +++-
> >  app/test-pmd/macfwd.c                 |  33 +------
> >  app/test-pmd/macfwd.h                 |  45 +++++++++
> >  app/test-pmd/noisy_vnf.c              | 106 +++++++++++++++++----
> >  app/test-pmd/parameters.c             |  15 +++
> >  app/test-pmd/testpmd.c                |  14 +++
> >  app/test-pmd/testpmd.h                |  10 ++
> >  doc/guides/testpmd_app_ug/run_app.rst |   9 ++
> >  10 files changed, 334 insertions(+), 164 deletions(-)
> >  create mode 100644 app/test-pmd/5tswap.h
> >  create mode 100644 app/test-pmd/macfwd.h
> >
>
> <...>
>
> > diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
> > index 096c218c12..a3e2c2ac15 100644
> > --- a/app/test-pmd/config.c
> > +++ b/app/test-pmd/config.c
> > @@ -4052,9 +4052,16 @@ rxtx_config_display(void)
> >  {
> >       portid_t pid;
> >       queueid_t qid;
> > +     char buf[32];
> > +
> > +     if (cur_fwd_eng == &noisy_vnf_engine)
> > +             snprintf(buf, sizeof(buf), " (%s)", noisy_fwd_mode_desc[noisy_fwd_mode]);
> > +     else
> > +             buf[0] = '\0';
> >
> > -     printf("  %s packet forwarding%s packets/burst=%d\n",
> > +     printf("  %s%s packet forwarding%s packets/burst=%d\n",
> >              cur_fwd_eng->fwd_mode_name,
> > +            buf,
> >              retry_enabled == 0 ? "" : " with retry",
> >              nb_pkt_per_burst);
> >
> > @@ -4816,10 +4823,17 @@ pkt_fwd_config_display(struct fwd_config *cfg)
> >       struct fwd_stream *fs;
> >       lcoreid_t  lc_id;
> >       streamid_t sm_id;
> > +     char buf[32];
> > +
> > +     if (cfg->fwd_eng == &noisy_vnf_engine)
> > +             snprintf(buf, sizeof(buf), " (%s)", noisy_fwd_mode_desc[noisy_fwd_mode]);
> > +     else
> > +             buf[0] = '\0';
> >
> > -     printf("%s packet forwarding%s - ports=%d - cores=%d - streams=%d - "
> > +     printf("%s%s packet forwarding%s - ports=%d - cores=%d - streams=%d - "
> >               "NUMA support %s, MP allocation mode: %s\n",
> >               cfg->fwd_eng->fwd_mode_name,
> > +             buf,
> >               retry_enabled == 0 ? "" : " with retry",
> >               cfg->nb_fwd_ports, cfg->nb_fwd_lcores, cfg->nb_fwd_streams,
> >               numa_support == 1 ? "enabled" : "disabled",
>
>
> This works but I wonder if we can keep the display functions generic (as
> much as possible), without forwarding enginee specific checks.
>
> What about the idea to update '.fwd_mode_name' in 'noisy_fwd_begin()'?
> That way generic display code can stay as it is.

I worried that this would interfere with set_pkt_forwarding_mode()
among other functions. The .fwd_mode_name seems to play double duty as
view and model.

I was also thinking about adding another member to struct fwd_engine
to display subtype / internal status. But I couldn't think of a
compelling use case in the other modules. So I leaned towards this
implementation.

Which do you prefer?

>
> <...>
>
> > +static bool
> > +pkt_burst_mac(struct fwd_stream *fs)
> > +{
> > +     struct rte_mbuf  *pkts_burst[MAX_PKT_BURST];
> > +     uint16_t nb_rx;
> > +     uint16_t nb_tx;
> > +
> > +     nb_rx = common_fwd_stream_receive(fs, pkts_burst, nb_pkt_per_burst);
> > +     if (likely(nb_rx != 0))
> > +             do_macfwd(pkts_burst, nb_rx, fs);
> > +     nb_tx = noisy_eth_tx_burst(fs, nb_rx, pkts_burst);
> > +
> > +     return nb_rx > 0 || nb_tx > 0;
> > +}
> > +static bool
> > +pkt_burst_macswap(struct fwd_stream *fs)
> > +{
> > +     struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
> > +     uint16_t nb_rx;
> > +     uint16_t nb_tx;
> > +
> > +     nb_rx = common_fwd_stream_receive(fs, pkts_burst, nb_pkt_per_burst);
> > +     if (likely(nb_rx != 0))
> > +             do_macswap(pkts_burst, nb_rx, &ports[fs->tx_port]);
> > +     nb_tx = noisy_eth_tx_burst(fs, nb_rx, pkts_burst);
> > +
> > +     return nb_rx > 0 || nb_tx > 0;
> > +}
> > +static bool
> > +pkt_burst_5tswap(struct fwd_stream *fs)
> > +{
> > +     struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
> > +     uint16_t nb_rx;
> > +     uint16_t nb_tx;
> > +
> > +     nb_rx = common_fwd_stream_receive(fs, pkts_burst, nb_pkt_per_burst);
> > +     if (likely(nb_rx != 0))
> > +             do_5tswap(pkts_burst, nb_rx, fs);
> > +     nb_tx = noisy_eth_tx_burst(fs, nb_rx, pkts_burst);
> > +
> > +     return nb_rx > 0 || nb_tx > 0;
> > +}
> >
>
> Empty lines are missing between functions.
>
>