Re: Request for clarification about FILESYSTEM_MAX_STACK_DEPTH

[Miklos Szeredi <miklos@szeredi.hu>]

On Mon, 5 May 2025 at 23:02, Allison Karlitskaya <lis@redhat.com> wrote:
>
> hi,
>
> Here's another "would be nice to have this on the record" sort of
> thing that I haven't been able to find any other public statements
> about.
>
> FILESYSTEM_MAX_STACK_DEPTH is defined in a non-public header as 2.  As
> far as I can tell, there's no way to query the limit out of a running
> kernel, and there's no indication if this value might ever be
> increased.  Hopefully it won't be decreased.
>
> I'm trying to write a userspace binding layer for supporting
> passthrough fds in fuse and it's hard to validate user input for the
> stacking depth parameter.  libfuse hardcodes some constant values (0,
> 1) that the user might choose, but in an adjacent comment makes
> references to the "current" kernel, suggesting that it might change at
> some point.  It's also sort of difficult to determine if a value is
> valid by probing: choosing an invalid value simply disables
> passthrough fd support, which you won't find out about until you
> actually go and try to create a passthrough fd, at which point it
> fails with EPERM (but which can also happen for many other reasons).
>
> So, I guess:
>  - will 2 ever change?

It will never be smaller than 2, that's guaranteed.  I think there's a
good chance of increasing this in the future if there's use case for
it.

>  - if so, can we add some sort of API to get the current value?

Yes.

Thanks,
Miklos