Re: [PATCH BUNDLE v7] famfs: Fabric-Attached Memory File System

[John Groves]

On 26/01/20 09:12AM, Alireza Sanaee wrote:
> On Sun, 18 Jan 2026 22:29:18 +0000
> John Groves <john@xxxxxxxxxxxxxx> wrote:
> 
> Hi John,
> 
> I wonder if these new patches sent recently have been reflected on the github repo readme files. It seems it is not, is it?
> 

[ ... ]

> > 
> > References
> > ----------
> > [1] https://lore.kernel.org/linux-cxl/cover.1708709155.git.john@xxxxxxxxxx/
> > [2] https://lore.kernel.org/linux-cxl/cover.1714409084.git.john@xxxxxxxxxx/
> > [3] https://lwn.net/Articles/983105/ (LSFMM 2024)
> > [4] https://lwn.net/Articles/1020170/ (LSFMM 2025)
> > [5] https://famfs.org (famfs user space)
> > [6] https://lore.kernel.org/linux-cxl/20250703185032.46568-1-john@xxxxxxxxxx/ (V2)
> > [7] https://lore.kernel.org/linux-fsdevel/20260107153244.64703-1-john@xxxxxxxxxx/T/#m0000d8c00290f48c086b8b176c7525e410f8508c (related ndctl series)
> > --

Hi Ali,

[5] points to the main famfs user space repo; I haven't updated documentation
there yet. The master branch there works with this patch set, and also
remains compatible with famfs kernels back to 6.8 (both fuse and standalone),
but I recommend this latest version (which is the famfs-v7 tag in my kernel
repos).

Some people are still running standalone famfs, and for that I recommend the
famfs_dualv3 branch, which supports both fuse and standalone mounts in a
6.14 kernel. I don't currently plan to forward-port standalone famfs to
6.19, because fuse is the path forward.

We're working on a performance regression test suite now, but early
indications are the fuse version is equivalent performance to standalone -
except for open, which is slower due to the fuse kernel/server interaction.
Most of our use cases involve large data sets, so we think this is OK - but
there is an opportunity later optimization of open.

Hope this is helpful,
John