[master][blivet][PATCH] Consider existing None format nonresizable (#1033778)

Vratislav Podzimek vpodzime at redhat.com
Thu Nov 28 06:17:44 UTC 2013


On Wed, 2013-11-27 at 10:30 -0600, David Lehman wrote:
> On Wed, 2013-11-27 at 15:37 +0100, Vratislav Podzimek wrote:
> > On Tue, 2013-11-26 at 10:43 -0600, David Lehman wrote:
> > > On Tue, 2013-11-26 at 14:57 +0100, Vratislav Podzimek wrote:
> > > > If there is an existing format, but we cannot recognize it we shouldn't consider
> > > > it resizable.
> > > 
> > > This is going to need a slightly more complicated fix. If a device is
> > > truly unformatted we should be able to resize it. See blivet commit
> > > e411bd44f47d85, which I now believe I did wrong. I think I should have
> > > set the name attribute -- not the type attr -- to the udev-provided
> > > fstype so all the places where we check for a type of None would still
> > > work and the UI stuff, which should use name instead of type, would all
> > > show something other than "Unknown".
> > Patch changing that sent to the list.
> > 
> > But still, how could we deal with the cases like the one described in
> > that bug? If there is a format we don't recognize how can we resize it?
> > If the device was truly unformatted, would it's format has the 'exists'
> > attribute set to True? I thought that is what makes a difference between
> > an unformatted device and a device with format we don't recognize.
> 
> The format exists in either case. It is ridiculous to refuse resize of
> an unformatted device (IMO). The issue is this: it is impossible to
> differentiate with 100% certainty an unformatted device from a device
> with formatting we (or blkid) cannot understand. There are some formats
> that blkid does not understand, in which case it will report an
> unformatted device. In other cases, blkid reports a format type, but we
> don't have a class for that type. In either case, the device ends up
> with a DeviceFormat. So -- in the case of a device whose formatting is
> reported to us by blkid as something we don't know what to do with,
> meaning it ends up with a DeviceFormat -- that leaves us one good
> option: present any type blkid reported to us in the GUI and still allow
> resize -- after all, the user should be paying attention when they
> repartition their disks. I do not care how many people whine to me that
> "you should have hit me over the head harder while I intentionally
> destroyed all of my data!" -- it is not my responsibility to keep you
> from being reckless or careless.
> 
> > 
> > Or would it be worth to mark devices that have the minSize = size as
> > "Not resizable"? That would also resolve that bug though in a completely
> > different way.
> > 
> 
> I think we should show the type reported by blkid (if any), allow
> resize, and let people take responsibility for their own actions.
That sound like a good approach to me. Thanks for the explanation! If I
understand it correctly, the change setting fmt._name instead of
fmt._type and using format.name in the GUI does everything we need,
right?

-- 
Vratislav Podzimek

Anaconda Rider | Red Hat, Inc. | Brno - Czech Republic



More information about the anaconda-patches mailing list