vdpa/ifc: fix reconnetion issue in SW assisted live migration
Checks
Commit Message
In the case using argument "sw-live-migration=1" to enable SW assisted live
migration, we take QEMU as front end for example, after source VM migrates
to destination VM, we keep vdpa process for source VM there, we kill the
QEMU process for source VM, and restart the QEMU process for source VM.
In this case, vdpa driver will not perform DMA map and data path will not
work properly.
The above case works fine in the case "sw-live-migration=0".
The root cause is that current code driver does not set running flag to 0.
Driver treats device as ruuning and does not perform DMA map.
Fixes: 4bb531e152d3 ("net/ifc: support SW assisted VDPA live migration")
Cc: stable@dpdk.org
Signed-off-by: Andy Pei <andy.pei@intel.com>
---
drivers/vdpa/ifc/ifcvf_vdpa.c | 2 ++
1 file changed, 2 insertions(+)
Comments
On 12/12/22 08:12, Andy Pei wrote:
> In the case using argument "sw-live-migration=1" to enable SW assisted live
> migration, we take QEMU as front end for example, after source VM migrates
> to destination VM, we keep vdpa process for source VM there, we kill the
> QEMU process for source VM, and restart the QEMU process for source VM.
> In this case, vdpa driver will not perform DMA map and data path will not
> work properly.
>
> The above case works fine in the case "sw-live-migration=0".
>
> The root cause is that current code driver does not set running flag to 0.
> Driver treats device as ruuning and does not perform DMA map.
>
> Fixes: 4bb531e152d3 ("net/ifc: support SW assisted VDPA live migration")
> Cc: stable@dpdk.org
>
> Signed-off-by: Andy Pei <andy.pei@intel.com>
> ---
> drivers/vdpa/ifc/ifcvf_vdpa.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/drivers/vdpa/ifc/ifcvf_vdpa.c b/drivers/vdpa/ifc/ifcvf_vdpa.c
> index 49d68ad..dc8600d 100644
> --- a/drivers/vdpa/ifc/ifcvf_vdpa.c
> +++ b/drivers/vdpa/ifc/ifcvf_vdpa.c
> @@ -1044,6 +1044,8 @@ struct rte_vdpa_dev_info {
>
> vdpa_disable_vfio_intr(internal);
>
> + rte_atomic32_set(&internal->running, 0);
> +
> ret = rte_vhost_host_notifier_ctrl(vid, RTE_VHOST_QUEUE_ALL, false);
> if (ret && ret != -ENOTSUP)
> goto error;
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Thanks,
Maxime
On 12/12/22 08:12, Andy Pei wrote:
> In the case using argument "sw-live-migration=1" to enable SW assisted live
> migration, we take QEMU as front end for example, after source VM migrates
> to destination VM, we keep vdpa process for source VM there, we kill the
> QEMU process for source VM, and restart the QEMU process for source VM.
> In this case, vdpa driver will not perform DMA map and data path will not
> work properly.
>
> The above case works fine in the case "sw-live-migration=0".
>
> The root cause is that current code driver does not set running flag to 0.
> Driver treats device as ruuning and does not perform DMA map.
>
> Fixes: 4bb531e152d3 ("net/ifc: support SW assisted VDPA live migration")
> Cc: stable@dpdk.org
>
> Signed-off-by: Andy Pei <andy.pei@intel.com>
> ---
> drivers/vdpa/ifc/ifcvf_vdpa.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/drivers/vdpa/ifc/ifcvf_vdpa.c b/drivers/vdpa/ifc/ifcvf_vdpa.c
> index 49d68ad..dc8600d 100644
> --- a/drivers/vdpa/ifc/ifcvf_vdpa.c
> +++ b/drivers/vdpa/ifc/ifcvf_vdpa.c
> @@ -1044,6 +1044,8 @@ struct rte_vdpa_dev_info {
>
> vdpa_disable_vfio_intr(internal);
>
> + rte_atomic32_set(&internal->running, 0);
> +
> ret = rte_vhost_host_notifier_ctrl(vid, RTE_VHOST_QUEUE_ALL, false);
> if (ret && ret != -ENOTSUP)
> goto error;
Applied to dpdk-next-virtio/main.
Thanks,
Maxime
@@ -1044,6 +1044,8 @@ struct rte_vdpa_dev_info {
vdpa_disable_vfio_intr(internal);
+ rte_atomic32_set(&internal->running, 0);
+
ret = rte_vhost_host_notifier_ctrl(vid, RTE_VHOST_QUEUE_ALL, false);
if (ret && ret != -ENOTSUP)
goto error;