[dpdk-dev] [PATCH] nfp: extend speed capabilities advertised

[Marc]

On 19 December 2016 at 17:18, Alejandro Lucero <
alejandro.lucero at netronome.com> wrote:

> On Mon, Dec 19, 2016 at 3:05 PM, Ferruh Yigit <ferruh.yigit at intel.com>
> wrote:
>
> > On 12/19/2016 3:02 PM, Alejandro Lucero wrote:
> > >
> > >
> > > On Mon, Dec 19, 2016 at 2:36 PM, Ferruh Yigit <ferruh.yigit at intel.com
> > > <mailto:ferruh.yigit at intel.com>> wrote:
> > >
> > >     Hi Alejandro,
> > >
> > >
> > > Hi,
> > >
> > >
> > >     On 12/19/2016 12:05 PM, Alejandro Lucero wrote:
> > >     > NFP supports more speeds than just 40 and 100GB, which were
> > >     > what was advertised before.
> > >     >
> > >     > Signed-off-by: Alejandro Lucero <alejandro.lucero at netronome.com
> > <mailto:alejandro.lucero at netronome.com>>
> > >     > ---
> > >     >  drivers/net/nfp/nfp_net.c | 4 +++-
> > >     >  1 file changed, 3 insertions(+), 1 deletion(-)
> > >     >
> > >     > diff --git a/drivers/net/nfp/nfp_net.c
> b/drivers/net/nfp/nfp_net.c
> > >     > index 27afbfd..77015c4 100644
> > >     > --- a/drivers/net/nfp/nfp_net.c
> > >     > +++ b/drivers/net/nfp/nfp_net.c
> > >     > @@ -1077,7 +1077,9 @@ static void nfp_net_read_mac(struct
> > nfp_net_hw *hw)
> > >     >       dev_info->reta_size = NFP_NET_CFG_RSS_ITBL_SZ;
> > >     >       dev_info->hash_key_size = NFP_NET_CFG_RSS_KEY_SZ;
> > >     >
> > >     > -     dev_info->speed_capa = ETH_LINK_SPEED_40G |
> > ETH_LINK_SPEED_100G;
> > >     > +     dev_info->speed_capa = ETH_SPEED_NUM_1G |
> ETH_LINK_SPEED_10G
> > |
> > >     > +                            ETH_SPEED_NUM_25G |
> ETH_SPEED_NUM_40G
> > |
> > >     > +                            ETH_SPEED_NUM_50G |
> > ETH_LINK_SPEED_100G;
> > >
> > >     Does all devices driver by this driver supports all these speeds?
> > >
> > >     I am aware of at least one exception to this, from previous patch
> > [1],
> > >     should we take that into account?
> > >
> > >
> > > So we have different NFP devices and different firmwares.
> > > NFP by design support all those speeds, but the PMD relies on the
> > > firmware for being able to know which is the current configured speed
> > > after link negotiation. PMD development was done with a specific
> > > firmware, and I was told to just report such speed by default. Last
> > > firmware versions give that speed info, but old firmware versions do
> not.
> > >
> > > So, all devices support such a speed range, indeed PMD works with any
> of
> > > them, but reported speed is always 40G with old firmware. This is a
> > > firmware limitation but we have to support old and new firmware.
> >
> > But this information to the application will be wrong for some (old) FW.
> > What do you think checking the FW version here and report capability
> > based on what FW supports?
> >
> >
> The driver advertises the right speed range supported. The problem is with
> the report about the current link speed configured.
> Maybe, is the right thing to do here to not report the current link speed
> because the driver really does not know about it?
>
> If you agree with this, I'm afraid the just accepted patch about the link
> report needs to be modified.
>

Alejandro,

If negociated link state has to be changed, then struct rte_eth_dev_data
dev_link field is where to do it.

As Ferruh was saying, dev_info->speed_capa contains the speed capabilties
of the particular NIC in use, not the driver (detecting firmware version
would be the best here).

marc


>
>
>
> > >
> > >
> > >
> > >     Also other than that exception, can you please confirm all other
> > devices
> > >     support all above speeds?
> > >
> > >     [1]
> > >     +       if ((NFD_CFG_MAJOR_VERSION_of(hw->ver) < 4) ||
> > >     +           ((NFD_CFG_MINOR_VERSION_of(hw->ver) == 4) &&
> > >     +           (NFD_CFG_MINOR_VERSION_of(hw->ver) == 0)))
> > >     +               link.link_speed = ETH_SPEED_NUM_40G;
> > >
> > >
> > >     >  }
> > >     >
> > >     >  static const uint32_t *
> > >     >
> > >
> > >
> >
> >
>