Re: [PATCH RFC 12/29] fsverity: expose merkle tree geometry to callers

[Christoph Hellwig <hch@lst.de>]

On Mon, Aug 11, 2025 at 08:38:22AM -0700, Darrick J. Wong wrote:
> > Just curious, why does xfs need this, but the existing file systems
> > don't?  That would be some good background information for the commit
> > message.
> 
> Hrmmm... the last time I sent this RFC, online fsck used it to check the
> validity of the merkle tree xattrs.

I saw a few users, so it does get used.  But patches exporting something
should in generaly document what the use case is.

> > > +	if (!IS_VERITY(inode))
> > > +		return -ENODATA;
> > > +
> > > +	error = ensure_verity_info(inode);
> > > +	if (error)
> > > +		return error;
> > > +
> > > +	vi = inode->i_verity_info;
> > 
> > Wouldn't it be a better interface to return the verity_ino from
> > ensure_verity_info (NULL for !IS_VERITY, ERR_PTR for real error)
> > and then just look at the fields directly?
> 
> They're private to fsverity_private.h.

Indeed.  Is ensure_verity_info ven the right thing here?  I.e.
should quering the paramters create the info if it wasn't there
yet?