[dpdk-stable] [PATCH 19.11] net/mlx4: fix device detach

[Christian Ehrhardt]

On Mon, Feb 15, 2021 at 1:20 PM Christian Ehrhardt
<christian.ehrhardt at canonical.com> wrote:
>
> On Fri, Feb 12, 2021 at 3:33 PM Viacheslav Ovsiienko
> <viacheslavo at nvidia.com> wrote:
> >
> > From: Michael Baum <michaelba at nvidia.com>
> >
> > [ upstream commit 8e1630e0f1985beb7e48429e3a7614f4732b0e68 ]
>
> Thanks, added
>
> > When mlx4 device is probed, 2 different ethdev ports may be created for
> > the 2 physical ports of the device.
> >
> > Wrongly, when the device is removed, the created ports are not released.
> >
> > Close and release the ethdev ports in remove process.
> >
> > Bugzilla ID: 488
> > Fixes: 7fae69eeff13 ("mlx4: new poll mode driver")
> > Cc: stable at dpdk.org
> >
> > Reported-by: David Marchand <david.marchand at redhat.com>
> > Signed-off-by: Michael Baum <michaelba at nvidia.com>
> > Acked-by: Matan Azrad <matan at nvidia.com>
> > Tested-by: David Marchand <david.marchand at redhat.com>
> > ---
> >  drivers/net/mlx4/mlx4.c | 35 +++++++++++++++++++++++++++++++++++
> >  1 file changed, 35 insertions(+)
> >
> > diff --git a/drivers/net/mlx4/mlx4.c b/drivers/net/mlx4/mlx4.c
> > index 4479022a4..3d2cef335 100644
> > --- a/drivers/net/mlx4/mlx4.c
> > +++ b/drivers/net/mlx4/mlx4.c
> > @@ -376,6 +376,10 @@ mlx4_dev_close(struct rte_eth_dev *dev)
> >         struct mlx4_priv *priv = dev->data->dev_private;
> >         unsigned int i;
> >
> > +       if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
> > +               rte_eth_dev_release_port(dev);
> > +               return 0;


I've found on tests that this won't work.
Back in 19.11 this was a void function and therefore "return 0" breaks
on some rather strict compiler settings.
I'll make it just a "return;" but wanted to ask you to double check if
that should overall still work fine in your opinion.

> > +       }
> >         DEBUG("%p: closing device \"%s\"",
> >               (void *)dev,
> >               ((priv->ctx != NULL) ? priv->ctx->device->name : ""));
> > @@ -1131,6 +1135,36 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
> >         return -err;
> >  }
> >
> > +/**
> > + * DPDK callback to remove a PCI device.
> > + *
> > + * This function removes all Ethernet devices belong to a given PCI device.
> > + *
> > + * @param[in] pci_dev
> > + *   Pointer to the PCI device.
> > + *
> > + * @return
> > + *   0 on success, the function cannot fail.
> > + */
> > +static int
> > +mlx4_pci_remove(struct rte_pci_device *pci_dev)
> > +{
> > +       uint16_t port_id;
> > +       int ret = 0;
> > +
> > +       RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device) {
> > +               /*
> > +                * mlx4_dev_close() is not registered to secondary process,
> > +                * call the close function explicitly for secondary process.
> > +                */
> > +               if (rte_eal_process_type() == RTE_PROC_SECONDARY)
> > +                       ret |= mlx4_dev_close(&rte_eth_devices[port_id]);
> > +               else
> > +                       ret |= rte_eth_dev_close(port_id);
> > +       }
> > +       return ret == 0 ? 0 : -EIO;
> > +}
> > +
> >  static const struct rte_pci_id mlx4_pci_id_map[] = {
> >         {
> >                 RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX,
> > @@ -1155,6 +1189,7 @@ static struct rte_pci_driver mlx4_driver = {
> >         },
> >         .id_table = mlx4_pci_id_map,
> >         .probe = mlx4_pci_probe,
> > +       .remove = mlx4_pci_remove,
> >         .drv_flags = RTE_PCI_DRV_INTR_LSC | RTE_PCI_DRV_INTR_RMV,
> >  };
> >
> > --
> > 2.18.1
> >
>
>
> --
> Christian Ehrhardt
> Staff Engineer, Ubuntu Server
> Canonical Ltd



-- 
Christian Ehrhardt
Staff Engineer, Ubuntu Server
Canonical Ltd