Re: [RFC V2 12/18] famfs_fuse: Plumb the GET_FMAP message/response

["Darrick J. Wong" <djwong@kernel.org>]

On Fri, Aug 15, 2025 at 11:53:16AM -0500, John Groves wrote:
> On 25/08/14 04:36PM, Miklos Szeredi wrote:
> > On Thu, 14 Aug 2025 at 15:36, Miklos Szeredi <miklos@szeredi.hu> wrote:
> > 
> > > I'm still hoping some common ground would benefit both interfaces.
> > > Just not sure what it should be.
> > 
> > Something very high level:
> > 
> >  - allow several map formats: say a plain one with a list of extents
> > and a famfs one
> >  - allow several types of backing files: say regular and dax dev
> >  - querying maps has a common protocol, format of maps is opaque to this
> >  - maps are cached by a common facility
> >  - each type of mapping has a decoder module
> >  - each type of backing file has a module for handling I/O
> > 
> > Does this make sense?
> > 
> > This doesn't have to be implemented in one go, but for example
> > GET_FMAP could be renamed to GET_READ_MAP with an added offset and
> > size parameter.  For famfs the offset/size would be set to zero/inf.
> > I'd be content with that for now.
> 
> Maybe GET_FILE_MAP or GET_FILE_IOMAP if we want to keep overloading 
> the term iomap. Maps are to backing-dev for regular file systems,
> and to device memory (devdax) for famfs - in all cases both read
> and write (when write is allowed).

The calling model for fuse-iomap is the same as fs/iomap -- there's an
IOMAP_BEGIN upcall to get a mapping from the filesystem, and an
IOMAP_END upcall to tell the fuse server whatever it did with the
mapping.  Some filesystems will reserve delayed allocation reservations
in iomap_begin for a pagecache write, and need to cancel those
reservations if the write fails.

For a pagecache write you need both a read and a write mapping because
the caller's file range isn't guaranteed to be fsblock-aligned.  famfs
mappings are a subcase of iomappings -- the read & write mappings are
the same, and they're always FUSE_IOMAP_TYPE_MAPPED.

IOWs, I don't want "GET_FILE_IOMAP" because that's not how iomap works.
(There's a separate FUSE_IOMAP_IOEND to pass along IO completions from
storage)

Given that famfs just calls dax_iomap_rw with an iomap_ops struct, I
seriously wonder if I should just wire up fsdax for RFC v5 and then
let's see how much code famfs actually needs on top of that.

--D

> Thanks,
> John
> 
>