On Fri, Mar 2, 2012 at 6:38 PM, Kai Meyer <kai.me...@storagecraft.com> wrote: > > ________________________________________ > From: Paolo Bonzini [pbonz...@redhat.com] > Sent: Thursday, March 01, 2012 11:54 PM > To: Kai Meyer > Cc: Anthony Liguori; Stefan Weil; qemu-devel@nongnu.org; Nate Bushman > Subject: Re: Add support for new image type > > Il 01/03/2012 22:14, Kai Meyer ha scritto: >> If we can't use qemu in general use-cases (since we can't push support >> for our images up stream), we could still really benefit from a targeted >> use-cases (like a Rescue Environment.) > > It does not matter whether it is upstream or not. > > When you distribute your modified QEMU binary, anyone who receives it > has the right to ask you for the complete corresponding source code. > > I also suggest that you write a wrapper around your library that exports > the contents as iSCSI or NBD. > > Paolo > > > Well, yes. I was assuming that there was potential for us to be able to > distribute qemu modifications that would not require us opening up our > library. The more we look at it, and some past precedence we've experienced, > it looks like it's not going to happen. > > I think I agree that iSCSI is really our only option (legally). Too bad. The > code was so simple to integrate directly into qemu. I have to give you guys > kudos again for already having such a simple integration point.
If iSCSI is too much effort, I suggest looking at NBD which Paolo mentioned earlier. It's a small and simple protocol and QEMU already has the "glue code" to speak it. In other words, you would write an NBD server that understands your image format and QEMU would use its existing block/nbd.c driver to talk to it over a standard protocol. That even opens the possibility of attaching the image file as a Linux block device using the nbd driver in Linux. Stefan
