On Wed, Mar 23, 2022 at 04:23:34PM +0100, Miklos Szeredi wrote: > On Wed, 23 Mar 2022 at 14:38, Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote: > > > This has been proposed in the past a few times. Most recently by the > > KVM developers, which tried to create a "generic" api, but ended up just > > making something to work for KVM as they got tired of people ignoring > > their more intrusive patch sets. See virt/kvm/binary_stats.c for what > > they ended up with, and perhaps you can just use that same type of > > interface here as well? > > So this looks like a fixed set of statistics where each one has a > descriptor (a name, size, offset, flags, ...) that tells about the > piece of data to be exported. The stats are kept up to date in kernel > memory and copied to userspace on read. The copy can be selective, > since the read can specify the offset and size of data it would like > to retrieve. > > The interface is self descriptive and selective, but its structure is > fixed for a specific object type, there's no way this could be > extended to look up things like extended attributes. Maybe that's not > a problem, but the lack of a hierarchical namespace could turn out to > be a major drawback. > > I think people underestimate the usefulness of hierarchical > namespaces, even though we use them extensively in lots of well > established interfaces. I like the namespaces, they work well. If you want self-describing interfaces (which I think your patch does), then why not just use the varlink protocol? It's been implemented for the kernel already many years ago: https://github.com/varlink and specifically: https://github.com/varlink/linux-varlink It doesn't need a new syscall. thanks, greg k-h
