Am 25.02.2013 um 16:01 hat Markus Armbruster geschrieben: > >> The behaviour then would be nop on no medium during commit all. But if > >> you specifically tried to just run commit() on a single device that > >> had no medium, you would receive ENOMEDIUM. That seems logical to me. > > > > I think, even "commit <blockdev-without-medium>" should not return > > -ENOMEDIUM. That would be more consistent IMHO. > > "commit FOO" makes sense only when FOO is a COW backend. When it isn't, > we can either ignore the non-sensical request silently, or complain. I > think both ways are defensible. We just need to pick one, and stick to > it consistently. > > "commit all" should obviously apply only to COW backends. I don't think > that makes a "commit FOO" that complains inconsistent. I simply read > "all" as "all COW backends". > > That said, I don't really care which way we go. Kevin or Stefan, got a > preference?
I don't have a very strong opinion on it, but my expectation would have been that an explicit commit on an empty or non-COW drive would give me an error. Kevin
