On Tue, May 06, 2014 at 07:51:10PM +0200, Max Reitz wrote: > On 06.05.2014 10:10, Stefan Hajnoczi wrote: > >On Mon, May 05, 2014 at 06:19:07PM +0200, Max Reitz wrote: > >>On 10.04.2014 20:43, Max Reitz wrote: > >>>This series adds a passthrough JSON protocol block driver. Its filenames > >>>are JSON objects prefixed by "json:". The objects are used as options > >>>for opening another block device which will be the child of the JSON > >>>device. Regarding this child device, the JSON driver behaves nearly the > >>>same as raw_bsd in that it is just a passthrough driver. The only > >>>difference is probably that the JSON driver identifies itself as a block > >>>filter, in contrast to raw_bsd. > >>> > >>>The purpose of this driver is that it may sometimes be desirable to > >>>specify options for a block device where only a filename can be given, > >>>e.g., for backing files. Using this should obviously be the exception, > >>>but it is nice to have if actually needed. > >>Ping – I do understand that Kevin has reservations against this > >>series, but as long as he doesn't explicitly ask me to reimplement > >>this in bdrv_open() without an own block driver (which I'd more or > >>less gladly do), I do not see issues why this series should not be > >>merged. > >I haven't reviewed it further because it seems like a kludge (that we > >have to keep supporting once it's merged). Was hoping you and Kevin > >will come up with a long-term fix instead. > > Okay, if you think the same, I guess I'll have to rewrite this > series. I agree that including this functionality in bdrv_open() is > the nicer alternative from the user's point of view, whereas I think > using a separate block driver results in nicer code. > > I'll rewrite this series and then we'll see how bad it actually looks. ;-)
Thanks! Stefan
