Re: [dpdk-dev] [PATCH] app/testpmd: add support for forced ethernet speed

[Ajit Khaparde <ajit.khaparde@broadcom.com>]

[-- Attachment #1: Type: text/plain, Size: 7997 bytes --]

On Thu, Feb 25, 2021 at 10:25 AM Ferruh Yigit <ferruh.yigit@intel.com>
wrote:

> On 2/22/2021 7:18 PM, Ajit Khaparde wrote:
> > Add support for forced ethernet speed setting.
> > Currently testpmd tries to configure the Ethernet port in autoneg mode.
> > It is not possible to set the Ethernet port to a specific speed while
> > starting testpmd. In some cases capability to configure a forced speed
> > for the Ethernet port during initialization may be necessary. This patch
> > tries to add this support.
> >
> > The patch assumes full duplex setting and does not attempt to change
> that.
> > So speeds like 10M, 100M are not configurable using this method.
> >
> > The command line to configure a forced speed of 10G:
> > dpdk-testpmd -c 0xff  -- -i  --eth-link-speed  10000
> >
> > The command line to configure a forced speed of 50G:
> > dpdk-testpmd -c 0xff  -- -i  --eth-link-speed  50000
> >
> > Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> > ---
> >   app/test-pmd/parameters.c             | 42 +++++++++++++++++++++++++++
> >   app/test-pmd/testpmd.c                |  4 +++
> >   app/test-pmd/testpmd.h                |  1 +
> >   doc/guides/testpmd_app_ug/run_app.rst | 11 +++++++
> >   4 files changed, 58 insertions(+)
>
> Can you also update the release notes to document the new parameter?
>
Done


>
> >
> > diff --git a/app/test-pmd/parameters.c b/app/test-pmd/parameters.c
> > index c8acd5d1b7..e10f7d38fb 100644
> > --- a/app/test-pmd/parameters.c
> > +++ b/app/test-pmd/parameters.c
> > @@ -224,6 +224,7 @@ usage(char* progname)
> >       printf("  --hairpin-mode=0xXX: bitmask set the hairpin port
> mode.\n "
> >              "    0x10 - explicit Tx rule, 0x02 - hairpin ports paired\n"
> >              "    0x01 - hairpin ports loop, 0x00 - hairpin port
> self\n");
> > +     printf("  --eth-link-speed: forced link speed.\n");
> >   }
> >
> >   #ifdef RTE_LIB_CMDLINE
> > @@ -485,6 +486,41 @@ parse_event_printing_config(const char *optarg, int
> enable)
> >       return 0;
> >   }
> >
> > +static int
> > +parse_link_speed(int n)
> > +{
> > +     uint32_t speed;
> > +
> > +     switch (n) {
>
> OK to not support "10M, 100M", not sure if anybody really uses them, but
> what do
> you think checking them and return an error?
>
Ok. Added this.


>
> > +     case 1000:
> > +             speed = ETH_LINK_SPEED_1G;
> > +             break;
> > +     case 10000:
> > +             speed = ETH_LINK_SPEED_10G;
> > +             break;
> > +     case 25000:
> > +             speed = ETH_LINK_SPEED_25G;
> > +             break;
> > +     case 40000:
> > +             speed = ETH_LINK_SPEED_40G;
> > +             break;
> > +     case 50000:
> > +             speed = ETH_LINK_SPEED_50G;
> > +             break;
> > +     case 100000:
> > +             speed = ETH_LINK_SPEED_100G;
> > +             break;
> > +     case 200000:
> > +             speed = ETH_LINK_SPEED_200G;
> > +             break;
> > +     default:
> > +             speed = ETH_LINK_SPEED_AUTONEG;
> > +             break;
>
> Isn't this function to set a fixed link speed, why falling back to autoneg?
>
Good point. I removed it.


>
> Also shouldn't this function set 'ETH_LINK_SPEED_FIXED' too?
>
Yes. Done in v2.


>
> > +     }
> > +
> > +     return speed;
> > +}
> > +
> >   void
> >   launch_args_parse(int argc, char** argv)
> >   {
> > @@ -605,6 +641,7 @@ launch_args_parse(int argc, char** argv)
> >               { "rx-mq-mode",                 1, 0, 0 },
> >               { "record-core-cycles",         0, 0, 0 },
> >               { "record-burst-stats",         0, 0, 0 },
> > +             { "eth-link-speed",             1, 0, 0 },
> >               { 0, 0, 0, 0 },
> >       };
> >
> > @@ -1366,6 +1403,11 @@ launch_args_parse(int argc, char** argv)
> >                               record_core_cycles = 1;
> >                       if (!strcmp(lgopts[opt_idx].name,
> "record-burst-stats"))
> >                               record_burst_stats = 1;
> > +                     if (!strcmp(lgopts[opt_idx].name,
> "eth-link-speed")) {
> > +                             n = atoi(optarg);
> > +                             if (n >= 0 && parse_link_speed(n) >= 0)
> > +                                     eth_link_speed =
> parse_link_speed(n);
> > +                     }
> >                       break;
> >               case 'h':
> >                       usage(argv[0]);
> > diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
> > index caa711d6f3..9434e335a0 100644
> > --- a/app/test-pmd/testpmd.c
> > +++ b/app/test-pmd/testpmd.c
> > @@ -535,6 +535,8 @@ uint16_t gso_max_segment_size = RTE_ETHER_MAX_LEN -
> RTE_ETHER_CRC_LEN;
> >   /* Holds the registered mbuf dynamic flags names. */
> >   char dynf_names[64][RTE_MBUF_DYN_NAMESIZE];
> >
> > +uint32_t eth_link_speed;
> > +
>
> Can you please move this a little up, at the beginning of the 'testpmd.c'
> there
> are global config variables, this can go next to them.
>
Done in v2.


>
> And can you please add a single line comment to the variable, as done in
> all
> other ones.
>
Sure. Done in v2.


>
> >   /*
> >    * Helper function to check if socket is already discovered.
> >    * If yes, return positive value. If not, return zero.
> > @@ -1484,6 +1486,8 @@ init_config(void)
> >                       port->tx_conf[k].offloads =
> >                               port->dev_conf.txmode.offloads;
> >
> > +             port->dev_conf.link_speeds = eth_link_speed;
> > +
>
> This is set even user doesn't provide '--eth-link-speed' at all, in that
> case I
> assume it relies on the fact that variable default value (0) is the same
> as
> 'AUTONEG' value (0). So it sets speed to autoneg in that case.
> But what do you think to be more explicit, and set this only if the user
> provided the device argument, like:
>         if (eth_link_speed)
>                 port->dev_conf.link_speeds = eth_link_speed;
>
Done in v2.


>
>
> This sets the same link speed for all ports, do you think may we need
> capability
> to set link speed per port? Does it worth the complexity it brings?
>
It is turning out to be a bit complex. For this round I think let's keep it
simple.


>
> >               /* set flag to initialize port/queue */
> >               port->need_reconfig = 1;
> >               port->need_reconfig_queues = 1;
> > diff --git a/app/test-pmd/testpmd.h b/app/test-pmd/testpmd.h
> > index 60ddeb8f13..a3cd4a0e16 100644
> > --- a/app/test-pmd/testpmd.h
> > +++ b/app/test-pmd/testpmd.h
> > @@ -351,6 +351,7 @@ extern bool setup_on_probe_event; /**< disabled by
> port setup-on iterator */
> >   extern uint8_t hot_plug; /**< enable by "--hot-plug" parameter */
> >   extern int do_mlockall; /**< set by "--mlockall" or "--no-mlockall"
> parameter */
> >   extern uint8_t clear_ptypes; /**< disabled by set ptype cmd */
> > +extern uint32_t eth_link_speed;
> >
> >   #ifdef RTE_LIBRTE_IXGBE_BYPASS
> >   extern uint32_t bypass_timeout; /**< Store the NIC bypass watchdog
> timeout */
> > diff --git a/doc/guides/testpmd_app_ug/run_app.rst
> b/doc/guides/testpmd_app_ug/run_app.rst
> > index 6745072329..a856f52736 100644
> > --- a/doc/guides/testpmd_app_ug/run_app.rst
> > +++ b/doc/guides/testpmd_app_ug/run_app.rst
> > @@ -536,3 +536,14 @@ The command line options are:
> >       bit 1 - two hairpin ports paired
> >       bit 0 - two hairpin ports loop
> >       The default value is 0. Hairpin will use single port mode and
> implicit Tx flow mode.
> > +
> > +*   ``--eth-link-speed``
> > +
>
> Should this document that "10M, 100M" is not supported?
>
Done in v2.


>
> > +    Set a forced link speed to the ethernet port.
> > +    1000 - 1Gbps
> > +    10000 - 10Gbps
> > +    25000 - 25Gbps
> > +    40000 - 40Gbps
> > +    50000 - 50Gbps
> > +    100000 - 100Gbps
> > +    200000 - 200Gbps
>
> What do you think to put something like "..." to very bottom, to clarify
> this is
> not full list, so that we won't need to update this each time we add a new
> speed?
>
Done.