[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [Nbd] [PATCH v2] nbd: Issue a uevent when disconnecting



[lkml dropped, no longer relevant]

On Fri, Feb 12, 2016 at 03:49:32PM +0100, Markus Pargmann wrote:
> Hi Wouter,
> 
> On Friday, February 05, 2016 10:42:41 AM Wouter Verhelst wrote:
> > Hi Markus,
> > 
> > On Fri, Feb 05, 2016 at 10:03:41AM +0100, Markus Pargmann wrote:
> > > Hi Wouter,
> > > 
> > > On Tuesday, February 02, 2016 04:27:00 PM Wouter Verhelst wrote:
> > > > There already is a uevent by default when closing a device upon connect
> > > > of the device. However, the same isn't true on disconnect.
> > > > 
> > > > This makes it hard for userspace to keep track of whether a device is
> > > > connected, since we are notified when the connection is created, but not
> > > > when it is removed again.
> > > > 
> > > > Explicitly issue a "change" uevent to remedy.
> > > > 
> > > > [v2: incorporate feedback from udev maintainer and make sure that we
> > > > issue a uevent upon connect as well as disconnect, rather than connect
> > > > only]
> > > 
> > > The systemd people had the same feedback.
> > 
> > Well, the systemd people and the udev people are the same people these
> > days :-)
> > 
> > > Basically the device is not marked as ready until a uevent is received. My
> > > idea was to use the size property consistently. If the device is connected
> > > the size is correctly changed and a uevent is created. On disconnect the size
> > > is set to 0 with another uevent.
> > > 
> > > Does this work for you as well? I will send the patch as reply.
> > 
> > Sure. What matters most is that the uevent is done; how it's done, less
> > so.
> > 
> > I should add that this was meant to go with systemd #2422. I wasn't
> > aware of #358, or I would've talked to you more :-)
> 
> Ah I see, thanks.
> 
> I wrote some tests yesterday to ensure that I don't break other things
> with the new size calculations. I will send the pull request with this
> patch for 4.6.

Noted. Thank you.

For reference: the reason I'm interested in this is that I'm working on a
systemd unit for nbd-client, which is needed because of
<https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=796633>.

Regards,

-- 
It is easy to love a country that is famous for chocolate and beer

  -- Barack Obama, speaking in Brussels, Belgium, 2014-03-26



Reply to: