[PATCH] vfio: add get device info API

[Burakov, Anatoly]

On 11/14/2023 11:48 AM, Mingjin Ye wrote:
> This patch adds an API to support getting device information.
> 
> The driver can use the "rte_vfio_get_device_info" helper to get
> device information from EAL.
> 
> Cc: stable at dpdk.org
> 
> Signed-off-by: Mingjin Ye <mingjinx.ye at intel.com>
> ---
>   lib/eal/include/rte_vfio.h | 26 ++++++++++++++++++++++++++
>   lib/eal/linux/eal_vfio.c   | 19 +++++++++++++++++++
>   lib/eal/version.map        |  1 +
>   3 files changed, 46 insertions(+)
> 
> diff --git a/lib/eal/include/rte_vfio.h b/lib/eal/include/rte_vfio.h
> index 3487c4f2a2..b3f55963e3 100644
> --- a/lib/eal/include/rte_vfio.h
> +++ b/lib/eal/include/rte_vfio.h
> @@ -212,6 +212,32 @@ int
>   rte_vfio_get_group_num(const char *sysfs_base,
>   		      const char *dev_addr, int *iommu_group_num);
>   
> +/**
> + * Get device information
> + *
> + * This function is only relevant to linux and will return
> + * an error on BSD.
> + *
> + * @param sysfs_base
> + *   sysfs path prefix.
> + *
> + * @param dev_addr
> + *   device location.
> + *
> + * @param vfio_dev_fd
> + *   VFIO fd.
> + *
> + * @param device_info
> + *   Device information.
> + *
> + * @return
> + *   0 on success.
> + *   <0 on failure.
> + */
> +int
> +rte_vfio_get_device_info(const char *sysfs_base, const char *dev_addr,
> +			int *vfio_dev_fd, struct vfio_device_info *device_info);
> +
>   /**
>    * Open a new VFIO container fd
>    *
> diff --git a/lib/eal/linux/eal_vfio.c b/lib/eal/linux/eal_vfio.c
> index ad3c1654b2..5810d9fcd7 100644
> --- a/lib/eal/linux/eal_vfio.c
> +++ b/lib/eal/linux/eal_vfio.c
> @@ -1222,6 +1222,25 @@ vfio_set_iommu_type(int vfio_container_fd)
>   	return NULL;
>   }
>   
> +int
> +rte_vfio_get_device_info(const char *sysfs_base, const char *dev_addr,
> +			int *vfio_dev_fd, struct vfio_device_info *device_info)
> +{
> +	if (!device_info || *vfio_dev_fd < 0)
> +		return -1;

Do we not need to check other parameters? E.g. sysfs_base, dev_addr, 
vfio_dev_fd itself (could be NULL dereference).

> +
> +	if (*vfio_dev_fd == 0) {
> +		if (rte_vfio_setup_device(sysfs_base, dev_addr,
> +				vfio_dev_fd, device_info))
> +			return -1;
> +	} else {
> +		if (ioctl(*vfio_dev_fd, VFIO_DEVICE_GET_INFO, &device_info))
> +			return -1;
> +	}

Couldn't this just be an `else if`?

Would also be nice to have some DEBUG output here.

With all of the above addressed,

Acked-by: Anatoly Burakov <anatoly.burakov at intel.com>

> +
> +	return 0;
> +}
> +
>   int
>   vfio_has_supported_extensions(int vfio_container_fd)
>   {
> diff --git a/lib/eal/version.map b/lib/eal/version.map
> index e00a844805..8b4a9c237e 100644
> --- a/lib/eal/version.map
> +++ b/lib/eal/version.map
> @@ -300,6 +300,7 @@ DPDK_24 {
>   	rte_vfio_noiommu_is_enabled; # WINDOWS_NO_EXPORT
>   	rte_vfio_release_device; # WINDOWS_NO_EXPORT
>   	rte_vfio_setup_device; # WINDOWS_NO_EXPORT
> +	rte_vfio_get_device_info; # WINDOWS_NO_EXPORT
>   	rte_zmalloc;
>   	rte_zmalloc_socket;
>   

-- 
Thanks,
Anatoly