On Mon, Sep 13, 2021 at 09:04:12PM +0800, Yongji Xie wrote: > On Mon, Sep 13, 2021 at 8:43 PM Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > > > > On Mon, Sep 13, 2021 at 07:25:57PM +0800, Xie Yongji wrote: > > > + invalidate_gendisk(nbd->disk); > > > + if (nbd->config->bytesize) > > > + kobject_uevent(&nbd_to_dev(nbd)->kobj, KOBJ_CHANGE); > > > > I wonder if the invalidate helper should just use > > set_capacity_and_notify to take care of the notification in the proper > > way. This adds an uevent to loop, and adds the RESIZE=1 argument to > > nbd, but it feels like the right thing to do. > > Looks like set_capacity_and_notify() would not do notification if we > set capacity to zero. True. > How about calling kobject_uevent() directly in > the helper? That's probably and improvement over letting the driver do it, so let's go with that for now instead of adding ever more work to your plate.
