On Tue, May 13, 2025 at 02:24:46PM -0700, Josh Poimboeuf wrote:
> On Tue, May 13, 2025 at 09:07:32AM -0400, Kent Overstreet wrote:
> > Static keys can now be a module parameter, e.g.
> >
> > module_param_named(foo, foo.key, static_key_t, 0644)
> >
> > bcachefs is now using this.
> >
> > Cc: Luis Chamberlain <mcg...@kernel.org>
> > Cc: Petr Pavlu <petr.pa...@suse.com>
> > Cc: Sami Tolvanen <samitolva...@google.com>
> > Cc: Daniel Gomez <da.go...@samsung.com>
> > Cc: linux-modu...@vger.kernel.org
> > Cc: Peter Zijlstra <pet...@infradead.org>
> > Cc: Josh Poimboeuf <jpoim...@kernel.org>
> > Cc: Jason Baron <jba...@akamai.com>
> > Cc: Steven Rostedt <rost...@goodmis.org>
> > Cc: Ard Biesheuvel <a...@kernel.org>
> > Signed-off-by: Kent Overstreet <kent.overstr...@linux.dev>
> > ---
> > include/linux/jump_label.h | 2 ++
> > include/linux/moduleparam.h | 7 +++++++
> > kernel/params.c | 35 +++++++++++++++++++++++++++++++++++
> > 3 files changed, 44 insertions(+)
> >
> > diff --git a/include/linux/jump_label.h b/include/linux/jump_label.h
> > index fdb79dd1ebd8..0fc9b71db56f 100644
> > --- a/include/linux/jump_label.h
> > +++ b/include/linux/jump_label.h
> > @@ -107,6 +107,8 @@ struct static_key {
> > #endif /* CONFIG_JUMP_LABEL */
> > };
> >
> > +typedef struct static_key static_key_t;
> > +
> > #endif /* __ASSEMBLY__ */
> >
> > #ifdef CONFIG_JUMP_LABEL
> > diff --git a/include/linux/moduleparam.h b/include/linux/moduleparam.h
> > index bfb85fd13e1f..11e8d5c57435 100644
> > --- a/include/linux/moduleparam.h
> > +++ b/include/linux/moduleparam.h
> > @@ -122,6 +122,7 @@ struct kparam_array
> > * charp: a character pointer
> > * bool: a bool, values 0/1, y/n, Y/N.
> > * invbool: the above, only sense-reversed (N = true).
> > + * static_key_t: same as bool, but updates a 'struct static_key'
>
> The static_key_*() interfaces are deprecated because they're really easy
> to use incorrectly. I don't think we want to propagate that misuse to
> modules.
>
> It would be better to have type(s) based on static_key_false and/or
> static_key_true, depending on whatever the default is.
Except those are just wrappers around struct static_key, so why does
that matter here?
