Re: [PATCH net-next v9 05/15] net: homa: create homa_peer.h and homa_peer.c

[Simon Horman <horms@kernel.org>]

On Fri, Jun 13, 2025 at 10:12:11AM -0700, John Ousterhout wrote:
> (I have implemented all of your suggestions for which there are no
> responses below)
> 
> On Fri, Jun 13, 2025 at 7:40 AM Simon Horman <horms@kernel.org> wrote:
> > > +/**
> > > + * homa_peer_hash() - Hash function used for @peertab->ht.
> > > + * @data:    Pointer to key for which a hash is desired. Must actually
> > > + *           be a struct homa_peer_key.
> > > + * @dummy:   Not used
> > > + * @seed:    Seed for the hash.
> > > + * Return:   A 32-bit hash value for the given key.
> > > + */
> > > +static inline u32 homa_peer_hash(const void *data, u32 dummy, u32 seed)
> >
> > Sorry if this has already been asked but can homa reuse the code in
> > drivers/md/dm-vdo/murmurhash3.c:murmurhash3_128() (after moving it
> > somewhere else).
> 
> No problem; the question hasn't been asked before. I'd be happy to use
> an existing implementation of murmurhash3 but couldn't find one that
> was accessible. What do you mean by "moving it somewhere else"? Are
> you suggesting I add a new murmurhash3 implementation somewhere
> "public" in the kernel? I'm a little hesitant to do this because I'm
> not at all expert on murmurhash3: I'm not confident about getting this
> right. Also, I wouldn't feel comfortable taking on maintainer
> responsibility for this.

Hi John,

I'm suggesting moving murmurhash3_128(), say to lib/ at the top
of the Kernel tree. And I'm happy to assist with this.

Aside from being accessible, does murmurhash3_128() meet your
needs in it's current form?

> 
> > > +     const struct homa_peer *peer = obj;
> > > +     const struct homa_peer_key *key = arg->key;
> >
> > nit: Reverse xmas tree here please.
> >      Likewise elsewhere in this patchset.
> >
> >      This tool can he useful here
> >      https://github.com/ecree-solarflare/xmastree/commits/master/
> 
> Thanks for the pointer to xmastree.pl; I wasn't aware of it and it
> will be super-helpful. I've fixed all of the reverse xmas tree issues
> now.
> 
> -John-
>