virtio-iommu: Fix migration regression

[PATCH] virtio-iommu: Fix migration regression

Zhenzhong Duan posted 1 patch 1 year, 11 months ago

Download mbox

Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20220624093740.3525267-1-zhenzhong.duan@intel.com

Maintainers: "Michael S. Tsirkin" <mst@redhat.com>, Eric Auger <eric.auger@redhat.com>

hw/virtio/virtio-iommu.c | 8 ++++++++
1 file changed, 8 insertions(+)

Expand all Fold all

[PATCH] virtio-iommu: Fix migration regression

Posted by Zhenzhong Duan 1 year, 11 months ago

We also need to switch to the right address space on dest side
after loading the device status. DMA to wrong address space is
destructive.

Fixes: 3facd774962fd ("virtio-iommu: Add bypass mode support to assigned device")
Suggested-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
---
 hw/virtio/virtio-iommu.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/hw/virtio/virtio-iommu.c b/hw/virtio/virtio-iommu.c
index 08b227e828f8..281152d338f4 100644
--- a/hw/virtio/virtio-iommu.c
+++ b/hw/virtio/virtio-iommu.c
@@ -1322,6 +1322,14 @@ static int iommu_post_load(void *opaque, int version_id)
     VirtIOIOMMU *s = opaque;
 
     g_tree_foreach(s->domains, reconstruct_endpoints, s);
+
+    /*
+     * Memory regions are dynamically turned on/off depending on
+     * 'config.bypass' and attached domain type if there is. After
+     * migration, we need to make sure the memory regions are
+     * still correct.
+     */
+    virtio_iommu_switch_address_space_all(s);
     return 0;
 }
 
-- 
2.25.1

Re: [PATCH] virtio-iommu: Fix migration regression

Posted by Eric Auger 1 year, 10 months ago

Hi,

On 6/24/22 11:37, Zhenzhong Duan wrote:
> We also need to switch to the right address space on dest side
> after loading the device status. DMA to wrong address space is
> destructive.
>
> Fixes: 3facd774962fd ("virtio-iommu: Add bypass mode support to assigned device")
> Suggested-by: Eric Auger <eric.auger@redhat.com>
> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>

Eric
> ---
>  hw/virtio/virtio-iommu.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
>
> diff --git a/hw/virtio/virtio-iommu.c b/hw/virtio/virtio-iommu.c
> index 08b227e828f8..281152d338f4 100644
> --- a/hw/virtio/virtio-iommu.c
> +++ b/hw/virtio/virtio-iommu.c
> @@ -1322,6 +1322,14 @@ static int iommu_post_load(void *opaque, int version_id)
>      VirtIOIOMMU *s = opaque;
>  
>      g_tree_foreach(s->domains, reconstruct_endpoints, s);
> +
> +    /*
> +     * Memory regions are dynamically turned on/off depending on
> +     * 'config.bypass' and attached domain type if there is. After
> +     * migration, we need to make sure the memory regions are
> +     * still correct.
> +     */
> +    virtio_iommu_switch_address_space_all(s);
>      return 0;
>  }
>