[dpdk-dev] [PATCH] vfio: fix device unplug when several devices per vfio group

[Burakov, Anatoly]

> From: Alejandro Lucero [mailto:alejandro.lucero at netronome.com]
> Sent: Wednesday, April 26, 2017 11:50 AM
> To: dev at dpdk.org
> Cc: Burakov, Anatoly <anatoly.burakov at intel.com>
> Subject: [PATCH] vfio: fix device unplug when several devices per vfio group
> 
> VFIO allows a secure way of assigning devices to user space and those
> devices which can not be isolated from other ones are set in same VFIO
> group. Releasing or unplugging a device should be aware of remaining
> devices is the same group for avoiding to close such a group.
> 
> Fixes: 94c0776b1bad ("vfio: support hotplug")
> 
> Signed-off-by: Alejandro Lucero <alejandro.lucero at netronome.com>
> ---
>  lib/librte_eal/linuxapp/eal/eal_vfio.c | 32 ++++++++++++++++++++++++----
> ----  lib/librte_eal/linuxapp/eal/eal_vfio.h |  1 +
>  2 files changed, 25 insertions(+), 8 deletions(-)
> 
> diff --git a/lib/librte_eal/linuxapp/eal/eal_vfio.c
> b/lib/librte_eal/linuxapp/eal/eal_vfio.c
> index 6e2e84c..6e24273 100644
> --- a/lib/librte_eal/linuxapp/eal/eal_vfio.c
> +++ b/lib/librte_eal/linuxapp/eal/eal_vfio.c
> @@ -184,6 +184,7 @@
>  			if (vfio_cfg.vfio_groups[i].fd == vfio_group_fd) {
>  				vfio_cfg.vfio_groups[i].group_no = -1;
>  				vfio_cfg.vfio_groups[i].fd = -1;
> +				vfio_cfg.vfio_groups[i].devices = 0;
>  				vfio_cfg.vfio_active_groups--;
>  				return 0;
>  			}
> @@ -353,6 +354,7 @@
>  		clear_group(vfio_group_fd);
>  		return -1;
>  	}
> +	vfio_cfg.vfio_groups[vfio_group_fd].devices++;
> 
>  	return 0;
>  }
> @@ -390,17 +392,30 @@
>  	 * code will unset the container and the IOMMU mappings.
>  	 */
> 
> -	if (close(vfio_group_fd) < 0)
> -		RTE_LOG(INFO, EAL, "Error when closing vfio_group_fd for
> %s\n",
> -				   dev_addr);
> -
> -	if (close(vfio_dev_fd) < 0)
> +	/* Closing a device */
> +	if (close(vfio_dev_fd) < 0) {
>  		RTE_LOG(INFO, EAL, "Error when closing vfio_dev_fd for
> %s\n",
>  				   dev_addr);
> +		return -1;
> +	}
> 
> -	if (clear_group(vfio_group_fd) < 0)
> -		RTE_LOG(INFO, EAL, "Error when clearing group for %s\n",
> -				   dev_addr);
> +	/* An VFIO group can have several devices attached. Just when there
> is
> +	 * no devices remaining should the group be closed.
> +	 */
> +	if (--vfio_cfg.vfio_groups[vfio_group_fd].devices == 0) {
> +
> +		if (close(vfio_group_fd) < 0) {
> +			RTE_LOG(INFO, EAL, "Error when closing
> vfio_group_fd for %s\n",
> +				dev_addr);
> +			return -1;
> +		}
> +
> +		if (clear_group(vfio_group_fd) < 0) {
> +			RTE_LOG(INFO, EAL, "Error when clearing group for
> %s\n",
> +					   dev_addr);
> +			return -1;
> +		}
> +	}
> 
>  	return 0;
>  }
> @@ -415,6 +430,7 @@
>  	for (i = 0; i < VFIO_MAX_GROUPS; i++) {
>  		vfio_cfg.vfio_groups[i].fd = -1;
>  		vfio_cfg.vfio_groups[i].group_no = -1;
> +		vfio_cfg.vfio_groups[i].devices = 0;
>  	}
> 
>  	/* inform the user that we are probing for VFIO */ diff --git
> a/lib/librte_eal/linuxapp/eal/eal_vfio.h
> b/lib/librte_eal/linuxapp/eal/eal_vfio.h
> index 7fcec2c..2c7cb94 100644
> --- a/lib/librte_eal/linuxapp/eal/eal_vfio.h
> +++ b/lib/librte_eal/linuxapp/eal/eal_vfio.h
> @@ -103,6 +103,7 @@ struct vfio_iommu_spapr_tce_remove {  struct
> vfio_group {
>  	int group_no;
>  	int fd;
> +	int devices;
>  };
> 
>  struct vfio_config {
> --
> 1.9.1

I have tested this on my setup on an old kernel with multiple attach/detaches, and it works (whereas it fails without this patch).

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