On 10/21/2010 12:43 PM, Ryan Harper wrote:
Thinking on this again; if we don't have a confirmation that the guest has removed the pci device (independent of whether we've revoked access to the underlying block device); then I really think we should fail the device removal; at the very least we cannot update libvirt state to indicate that the slot is free for re-assignment.
For the time being, I think we have to take the approach Dan suggested, and then revisit the entire API. The impact here is actually minimal because the chances of unplug failing when plug has succeeded is very low.
If it can fail, it changes the nature of the API in a fundamental way that probably requires a deep thought about what the API should look like in the future.
Regards, Anthony Liguori -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list