On 29/06/2023 09:40, Zhenzhong Duan wrote:
> When vfio_realize() succeeds, hot unplug will call vfio_exitfn()
> to free resources allocated in vfio_realize(); when vfio_realize()
> fails, vfio_exitfn() is never called and we need to free resources
> in vfio_realize().
>
> In the case that vfio_migration_realize() fails,
> e.g: with -only-migratable & enable-migration=off, we see below:
>
> (qemu) device_add
> vfio-pci,host=81:11.1,id=vfio1,bus=root1,enable-migration=off
> 0000:81:11.1: Migration disabled
> Error: disallowing migration blocker (--only-migratable) for: 0000:81:11.1:
> Migration is disabled for VFIO device
>
> If we hotplug again we should see same log as above, but we see:
> (qemu) device_add
> vfio-pci,host=81:11.1,id=vfio1,bus=root1,enable-migration=off
> Error: vfio 0000:81:11.1: device is already attached
>
> That's because some references to VFIO device isn't released,
> we should check return value of vfio_migration_realize() and
> release the references, then VFIO device will be truely
> released when hotplug fails.
>
> Fixes: a22651053b59 ("vfio: Make vfio-pci device migration capable")
> Signed-off-by: Zhenzhong Duan <[email protected]>
> ---
> hw/vfio/pci.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c
> index 54a8179d1c64..dc69d3031b24 100644
> --- a/hw/vfio/pci.c
> +++ b/hw/vfio/pci.c
> @@ -3210,6 +3210,7 @@ static void vfio_realize(PCIDevice *pdev, Error **errp)
> ret = vfio_migration_realize(vbasedev, errp);
> if (ret) {
> error_report("%s: Migration disabled", vbasedev->name);
> + goto out_vfio_migration;
> }
> }
>
> @@ -3219,6 +3220,8 @@ static void vfio_realize(PCIDevice *pdev, Error **errp)
>
> return;
>
> +out_vfio_migration:
> + vfio_migration_exit(vbasedev);
> out_deregister:
> vfio_disable_interrupts(vdev);
> out_intx_disable:
I agree with the general sentiment behind the change.
Clearly vfio::migration and vfio::migration_blocker are leaking from inside the
migration_realize() function.
But it is kinda awkward semantic that vfio_migration_realize() (or any realize)
failures need to be accompanied with a vfio_migration_exit() that tears down
state *leaked* by its realize() failure.
It sounds to me that this should be inside the vfio_migration_realize() not on
the caller? Unless QEMU ::realize() is expected to do this.
