[PATCH v2 1/2] net/i40e: fix MAC address removal

[Bruce Richardson]

On Mon, Sep 07, 2026 at 03:21:27PM +0200, David Marchand wrote:
> On Mon, 7 Sept 2026 at 14:57, Bruce Richardson
> <bruce.richardson at intel.com> wrote:
> >
> > On Mon, Sep 07, 2026 at 01:17:23PM +0200, David Marchand wrote:
> > > MAC addresses removal was tied with VMDq pools even when not used.
> > > So if VMDq is not enabled, no address would be ever removed.
> > >
> > > Fixes: 9de506a6c781 ("ethdev: skip VMDq pools unless configured")
> > >
> > > Signed-off-by: David Marchand <david.marchand at redhat.com>
> > > ---
> > > I could not test the change as my setups are KO atm.
> > >
> > > Probably worth squashing in 9de506a6c781 before pulling to main
> > >
> > > Changes since v1:
> > > - fixed compilation,
> > >
> > > ---
> > >  drivers/net/intel/i40e/i40e_ethdev.c | 7 ++++++-
> > >  1 file changed, 6 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/net/intel/i40e/i40e_ethdev.c b/drivers/net/intel/i40e/i40e_ethdev.c
> > > index b6b2d291ee..3c01354c39 100644
> > > --- a/drivers/net/intel/i40e/i40e_ethdev.c
> > > +++ b/drivers/net/intel/i40e/i40e_ethdev.c
> > > @@ -4509,11 +4509,16 @@ i40e_macaddr_remove(struct rte_eth_dev *dev, uint32_t index)
> > >       struct rte_ether_addr *macaddr;
> > >       int ret;
> > >       uint32_t i;
> > > +     bool vmdq;
> > >       uint64_t pool_sel;
> > >
> > >       macaddr = &(data->mac_addrs[index]);
> > >
> > > -     pool_sel = dev->data->mac_pool_sel[index];
> > > +     vmdq = (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_VMDQ_FLAG) != 0;
> > > +     if (!vmdq)
> > > +             pool_sel = 1;
> > > +     else
> > > +             pool_sel = dev->data->mac_pool_sel[index];
> > >
> > The explanation makes sense, but do we really need the new temporary
> > variable, rather than just checking the flag directly in the "if"?
> >
> > Also, the rest of the code in this function, and the add function about it,
> > uses "pf->flags & I40E_FLAG_VMDQ" as a check for vmdq support. Is there a
> > reason we can't use that flag also here, rather than checking the rx
> > mq_mode flags?
> 
> No, I just did not notice this internal flag and I simply used the
> ethdev level config.
> 
> What do you think of:
> 
> $ git diff next-net/for-main -- drivers/net/intel/i40e
> diff --git a/drivers/net/intel/i40e/i40e_ethdev.c
> b/drivers/net/intel/i40e/i40e_ethdev.c
> index b6b2d291ee..0d914d86ea 100644
> --- a/drivers/net/intel/i40e/i40e_ethdev.c
> +++ b/drivers/net/intel/i40e/i40e_ethdev.c
> @@ -4513,6 +4513,12 @@ i40e_macaddr_remove(struct rte_eth_dev *dev,
> uint32_t index)
> 
>         macaddr = &(data->mac_addrs[index]);
> 
> +       if (!(pf->flags & I40E_FLAG_VMDQ)) {
> +               if (i40e_vsi_delete_mac(pf->main_vsi, macaddr) != 0)
> +                       PMD_DRV_LOG(ERR, "Failed to remove MACVLAN filter");
> +               return;
> +       }
> +
>         pool_sel = dev->data->mac_pool_sel[index];
> 
>         for (i = 0; i < sizeof(pool_sel) * CHAR_BIT; i++) {
> @@ -4521,8 +4527,7 @@ i40e_macaddr_remove(struct rte_eth_dev *dev,
> uint32_t index)
>                                 vsi = pf->main_vsi;
>                         else {
>                                 /* No VMDQ pool enabled or configured */
> -                               if (!(pf->flags & I40E_FLAG_VMDQ) ||
> -                                       (i > pf->nb_cfg_vmdq_vsi)) {
> +                               if (i > pf->nb_cfg_vmdq_vsi) {
>                                         PMD_DRV_LOG(ERR,
>                                                 "No VMDQ pool
> enabled/configured");
>                                         return;
> 
> 
That looks simpler. Thanks.

/Bruce