[dpdk-dev] Cannot use MLX4 with igb_uio driver

[Royce Niu]

Hi, Adrien,

Actually, I tested using the original kernel module without binding. It
works.

However, it is only 6Mpps for 64B in pkt-gen, which is so slow for a 40Gbps
NIC.

Is that right?


On Mon, Jan 9, 2017 at 11:13 PM, Adrien Mazarguil <
adrien.mazarguil at 6wind.com> wrote:

> Hi Royce,
>
> On Mon, Jan 09, 2017 at 10:53:37PM +0800, Royce Niu wrote:
> > Dear all,
> >
> > I cannot use my Mellanox 3 Pro, after I binded it with igb_uio driver.
> >
> > It always shows when I use my DPDK application.
> >
> > EAL: Detected 32 lcore(s)
> > EAL: Probing VFIO support...
> > PMD: bnxt_rte_pmd_init() called for (null)
> > EAL: PCI device 0000:02:00.0 on NUMA socket 0
> > EAL:   probe driver: 8086:1521 rte_igb_pmd
> > EAL: PCI device 0000:02:00.1 on NUMA socket 0
> > EAL:   probe driver: 8086:1521 rte_igb_pmd
> > EAL: PCI device 0000:02:00.2 on NUMA socket 0
> > EAL:   probe driver: 8086:1521 rte_igb_pmd
> > EAL: PCI device 0000:02:00.3 on NUMA socket 0
> > EAL:   probe driver: 8086:1521 rte_igb_pmd
> > EAL: PCI device 0000:81:00.0 on NUMA socket 1
> > EAL:   probe driver: 15b3:1007 librte_pmd_mlx4
> > PMD: librte_pmd_mlx4: cannot access device, is mlx4_ib loaded?
> > EAL: Error - exiting with code: 1
> >   Cause: Cannot create mbuf pool
> >
> > ---------------
> > I have added CONFIG_RTE_LIBRTE_MLX4_PMD=y in .config, and
> > install MLNX_OFED_LINUX-3.4-2.0.0.0.
>
> The mlx4 PMD does not operate through igb_uio (see mlx4 documentation [1]),
> PCI devices must remain bound to their original kernel module (mlx4_core),
> however you have to additionally load mlx4_ib, mlx4_en and ib_uverbs [2].
>
> [1] http://dpdk.org/doc/guides/nics/mlx4.html
> [2] http://dpdk.org/doc/guides/nics/mlx4.html#prerequisites
>
> --
> Adrien Mazarguil
> 6WIND
>



-- 
Regards,

Royce