Re: [PATCH v3 2/2] eal: warn user that VFIO is disabled

[Ferruh Yigit <ferruh.yigit@intel.com>]

On 10/30/2017 9:36 AM, Gaetan Rivet wrote:
> If VFIO is not compatible with the kernel used, print a soft warning to
> the user if RTE_EAL_VFIO was still enabled in the configuration.
> 
> Signed-off-by: Gaetan Rivet <gaetan.rivet@6wind.com>
> ---
>  lib/librte_eal/linuxapp/eal/eal_vfio.h | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/lib/librte_eal/linuxapp/eal/eal_vfio.h b/lib/librte_eal/linuxapp/eal/eal_vfio.h
> index 766d004..fadd274 100644
> --- a/lib/librte_eal/linuxapp/eal/eal_vfio.h
> +++ b/lib/librte_eal/linuxapp/eal/eal_vfio.h
> @@ -183,6 +183,8 @@ int vfio_mp_sync_setup(void);
>  #define SOCKET_ERR 0xFF
>  
>  #define VFIO_PRESENT
> +#else
> +#pragma message("VFIO configured but not supported by this kernel, disabling.")

There is a flaw here, but it seems it was already there, nothing coming for this
patch.

There is a RTE_KERNELDIR variable, to compile DPDK with a given kernel, instead
of default one. I am using this a lot for kni and igb_uio.

Default RTE_KERNELDIR is:
RTE_KERNELDIR ?= /lib/modules/$(shell uname -r)/build

For vfio case, it is not taking this variable into account and always using the
default one, so RTE_KERNELDIR is broken here.

>  #endif /* kernel version */
>  #endif /* RTE_EAL_VFIO */
>  
>