It is a bit inefficient about my patch, and it is imperfect. I wish to see a better idea to fix this problem. And again, thank you for your correction to my patches. Thanks, Zongyong Wu > -----Original Message----- > From: Erik Skultety [mailto:eskultet@xxxxxxxxxx] > Sent: Thursday, January 11, 2018 6:16 PM > To: Wuzongyong (Euler Dept) <cordius.wu@xxxxxxxxxx> > Cc: libvir-list@xxxxxxxxxx; weijinfen <weijinfen@xxxxxxxxxx>; Wanzongshun > (Vincent) <wanzongshun@xxxxxxxxxx> > Subject: Re: [PATCH 3/3] nodedev: update caps before invoking nodedev > driver interfaces > > On Wed, Jan 10, 2018 at 08:14:51PM +0800, Wu Zongyong wrote: > > Some capabilities of node devices rely on what driver they bound to, > > and therefore, these capabilities may change when the driver change. > > So, it is necessary to manually update devices' capabilities each time > > before nodedev driver interfaces invoked. > > > > Signed-off-by: Wu Zongyong <cordius.wu@xxxxxxxxxx> > > --- > > Thank you for posting the patch, since I hadn't noticed the problem with > other APIs until I read it. > It was a sad realization that a driver change is not reflected by a > udev/kernel CHANGE event, that would make things much much simpler. I have > an idea to either make this patch shorter or not needed at all though, > we'll see when I finish my patch (it's a long-needed refactor). Despite I > haven't found any major flaws in this patch, let's just put it on hold for > a while until I finish my investigation/work on my patch and see we can > really do better here. > > Thanks, > Erik -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list
