On Tue, Nov 23, 2021 at 04:07:21PM -0800, Akilesh Kailash wrote: > Why not extend dm-snapshot; in that way you can have the existing > kernel COW format + support compression. In short, this is quite a focussed requirement and it doesn't seem worth compromising on the goals by attempting to shoe-horn it into a framework that was designed to fit around some quite different constraints twenty years ago. > I understand this for read-only but if dm-snapshot supports it, > existing users can have the > compression feature if required. I doubt that anyone wants to develop dm-snapshot any further. It's only really good for small short-term snapshots where performance and memory usage don't matter. The more-sophisticated thin provisioning approach replaced it. > Also, I am curious what are the other real world use case here apart > from Android ? There seems to be a general industry movement towards edge, attestation, trusted boot, nested systems, confidential computing, containers, <insert your favourite buzzwords> etc. and I think this type of device-mapper target might play a useful role as one of the low-level components involved in building up support for some of those scenarios. (Just as we recently began adding support for attestation.) Alasdair -- dm-devel mailing list dm-devel@xxxxxxxxxx https://listman.redhat.com/mailman/listinfo/dm-devel
