Re: [PATCH] fuse: add max_stack_depth to fuse_init_in

[Bernd Schubert <bernd@bsbernd.com>]

On 5/14/25 14:14, Allison Karlitskaya wrote:
> FILESYSTEM_MAX_STACK_DEPTH is defined privately inside of the kernel,
> but you need to know its value to properly implement fd passthrough on a
> FUSE filesystem.  So far most users have been assuming its current value
> of 2, but there's nothing that says that it won't change.
> 
> Use one of the unused fields in fuse_init_in to add a max_stack_depth
> uint32_t (matching the max_stack_depth uint32_t in fuse_init_out). If
> CONFIG_FUSE_PASSTHROUGH is configured then this is set to the maximum
> value that the kernel will accept for the corresponding field in
> fuse_init_out (ie: FILESYSTEM_MAX_STACK_DEPTH).
> 
> Let's not treat this as an ABI change: this struct is zero-initialized
> and the maximum max_stack_depth is non-zero (and always will be) so
> userspace can easily find out for itself if the value is present in the
> struct or not.
> 
> Signed-off-by: Allison Karlitskaya <allison.karlitskaya@redhat.com>
> ---
>   fs/fuse/inode.c           | 4 +++-
>   include/uapi/linux/fuse.h | 3 ++-
>   2 files changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
> index fd48e8d37f2e..46fd37eec9ae 100644
> --- a/fs/fuse/inode.c
> +++ b/fs/fuse/inode.c
> @@ -1497,8 +1497,10 @@ void fuse_send_init(struct fuse_mount *fm)
>   #endif
>   	if (fm->fc->auto_submounts)
>   		flags |= FUSE_SUBMOUNTS;
> -	if (IS_ENABLED(CONFIG_FUSE_PASSTHROUGH))
> +	if (IS_ENABLED(CONFIG_FUSE_PASSTHROUGH)) {
>   		flags |= FUSE_PASSTHROUGH;
> +		ia->in.max_stack_depth = FILESYSTEM_MAX_STACK_DEPTH;
> +	}
>   
>   	/*
>   	 * This is just an information flag for fuse server. No need to check
> diff --git a/include/uapi/linux/fuse.h b/include/uapi/linux/fuse.h
> index 5ec43ecbceb7..eb5d77d50176 100644
> --- a/include/uapi/linux/fuse.h
> +++ b/include/uapi/linux/fuse.h
> @@ -895,7 +895,8 @@ struct fuse_init_in {
>   	uint32_t	max_readahead;
>   	uint32_t	flags;
>   	uint32_t	flags2;
> -	uint32_t	unused[11];
> +	uint32_t	max_stack_depth;

Objections to make this a uint8_t? In fuse_init_out it just had slipped 
through in review. (And I wonder a bit if we should post change it in 
fuse_init_out.


Thanks,
Bernd