Re: [PATCH 11/11] fsverity: use a hashtable to find the fsverity_info

[Christoph Hellwig <hch@lst.de>]

On Thu, Jan 22, 2026 at 02:04:20PM -0800, Darrick J. Wong wrote:
> On Thu, Jan 22, 2026 at 09:22:07AM +0100, Christoph Hellwig wrote:
> > Use the kernel's resizable hash table to find the fsverity_info.  This
> 
> Oh is that what the 'r' stands for?  I thought it was rcu.  Maybe it's
> both. :P

From the lib/rhashtable.c:

 * Resizable, Scalable, Concurrent Hash Table

> > Because insertation into the hash table now happens before S_VERITY is
> > set, fsverity just becomes a barrier and a flag check and doesn't have
> > to look up the fsverity_info at all, so there is only one two two
> 
> "one two two" <confused>?

one or two, sorry.  The cover letter actually explains this in more
detail, which this should be updated to.

> > +static const struct rhashtable_params fsverity_info_hash_params = {
> > +	.key_len		= sizeof(struct inode *),
> 
> 	.key_len		= sizeof_field(struct fsverity_info, inode),
> 
> Perhaps?

That should work, yes.

> > -	kfree(vi->tree_params.hashstate);
> > -	kvfree(vi->hash_block_verified);
> > -	kmem_cache_free(fsverity_info_cachep, vi);
> > +	return rhashtable_lookup_fast(&fsverity_info_hash, &inode,
> > +			fsverity_info_hash_params);
> 
> Hrm.  The rhashtable stores a pointer to the rhash_head, but now we're
> returning that as if it were a fsverity_info pointer.  Can I be pedantic
> and ask for a proper container_of() to avoid leaving a landmine if the
> struct layout ever changes?

rhashtable_lookup_fast returns the struct containing the rhash_head.
The paramters store the rhead_offset for that purpose.  See rht_obj
as used by rhashtable_lookup.

> > @@ -323,9 +323,9 @@ fsverity_init_verification_context(struct fsverity_verification_context *ctx,
> >  				   struct fsverity_info *vi)
> >  {
> >  	ctx->inode = inode;
> > -	ctx->vi = vi;
> 
> Can this function drop its @vi argument?

No..

> 
> > +	ctx->vi = fsverity_get_info(inode);

... but this extra lookup should have been removed and got messed up by a
rebase, causing a pointless extra lookup.  (Which still is completely
in the noise in my runs).