Date:   Thu, 23 Feb 2023 14:27:06 -0500
From:   Kent Overstreet <kent.overstreet@linux.dev>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     linux-kernel@vger.kernel.org, mingo@redhat.com,
        stern@rowland.harvard.edu
Subject: Re: [PATCH 1/2] lockdep: lock_set_lock_cmp_fn()
Message-ID: <Y/e+Ci5ptzJT84wq@moria.home.lan>
References: <20230218032117.2372071-1-kent.overstreet@linux.dev>
 <20230218032117.2372071-2-kent.overstreet@linux.dev>
 <Y/ONLYqVrWPFyboF@hirez.programming.kicks-ass.net>
 <Y/QHn+PecW2n2K5O@moria.home.lan>
 <Y/djxk1q5EiYHFfF@hirez.programming.kicks-ass.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Y/djxk1q5EiYHFfF@hirez.programming.kicks-ass.net>
Precedence: bulk

On Thu, Feb 23, 2023 at 02:01:58PM +0100, Peter Zijlstra wrote:
> I would suggest the same as last time; integrate this in the class
> setting zoo of functions.

I don't really see a particular zoo, things like pretty scattered to me.
With lockdep_init_map_type?

> If you insiste, please keep it one function
> and force print_fn when cmp_fn. We don't want people to skimp out on
> this.

That's easily done.

> 
> Other than that, I don't think this can fully replace subclasses, since
> subclasses would allow lock hierarchies with other classes inter-twined,
> while this really relies on pure class nesting.
> 
> That is:
> 
> 	A/0
> 	B
> 	A/1
> 
> is a valid subclass nesting set, but you can't achieve the same with
> this.

Why not just assign those As to two different classes?

> That said; it does seem like a very useful additional annotation for the
> more complex nesting sets.
> 
> Thanks!

Cheers