On Thu, Mar 14, 2019 at 15:16:56 +0100, Peter Krempa wrote:
On Thu, Mar 14, 2019 at 14:56:49 +0100, Michal Privoznik wrote:
> On 3/14/19 2:06 PM, Peter Krempa wrote:
> > On Thu, Mar 14, 2019 at 13:22:37 +0100, Michal Privoznik wrote:
[...]
> > Callers need to be fixed that in case when they choose not to ignore the
> > error they report some error.
> >
>
> You mean callers after I do what you suggest in 1/5 (where zpci will call
> qemuMonitorDelDevice() directly)? Because after 4/5 there is only one caller
> which doesn't want the error to be reported. And I guess you did not mean to
> add some code here only to remove it in 4/5, did you?
I still consider the case when 'device_del' fails but the event is not
delivered to be a reporting-worthy problem due to a possible race when
deleting backends of the removed device.
Okay, this can't happen as I've figured out, so you are right in the
regard that you'd need to add code which would be deleted later.
ACK as I don't have a better option as the ZPCI stuff will also require
this functionality.