On 5/21/26 3:33 PM, Kees Cook wrote: > Add macros that define a struct kernel_param_ops initializer through a > macro so the underlying field layout can evolve without touching every > call site. Three variants cover the three cases: > > DEFINE_KERNEL_PARAM_OPS(name, set, get) // basic > DEFINE_KERNEL_PARAM_OPS_NOARG(name, set, get) // set > KERNEL_PARAM_OPS_FL_NOARG > DEFINE_KERNEL_PARAM_OPS_FREE(name, set, get, free) // also set .free > > Callers prefix their own visibility qualifiers, e.g.: > > static DEFINE_KERNEL_PARAM_OPS(my_ops, my_set, my_get); > > Also update module_param_call() and STANDARD_PARAM_DEF() to use > DEFINE_KERNEL_PARAM_OPS internally so the generated ops table will go > through the same macro as everything else. > > Subsequent commits convert all open-coded struct kernel_param_ops > definitions to use these macros, in preparation for migrating to a > seq_buf .get API. > > Signed-off-by: Kees Cook <[email protected]> > --- > include/linux/moduleparam.h | 36 ++++++++++++++++++++++++++++++++++-- > kernel/params.c | 6 ++---- > 2 files changed, 36 insertions(+), 6 deletions(-) > > diff --git a/include/linux/moduleparam.h b/include/linux/moduleparam.h > index 075f28585074..26bf45b36d02 100644 > --- a/include/linux/moduleparam.h > +++ b/include/linux/moduleparam.h > @@ -68,6 +68,39 @@ struct kernel_param_ops { > void (*free)(void *arg); > }; > > +/* > + * Define a const struct kernel_param_ops initializer. Callers prefix with > + * any required visibility qualifiers (typically "static"): > + * > + * static DEFINE_KERNEL_PARAM_OPS(my_ops, my_set, my_get); > + * > + * Routing the @_set and @_get function pointers through the macro > + * (rather than naming the struct fields at every call site) lets the > + * field layout change in one place when callbacks are migrated to a > + * new signature. > + */
Nit: The newly introduced DEFINE_KERNEL_PARAM_OPS*() macros remain in place at the end of the series after the migration is complete and this comment is removed in patch 7. It would be helpful to describe in the commit message why these macros are generally preferable to defining kernel_param_ops instances directly. I assume the motivation is that the structure is simple enough and using macros then makes defining kernel_param_ops instances a bit more concise. A minor disadvantage is that some analysis tools, such as ctags, may no longer see the generated definition, but that is also the case for DEFINE_MUTEX() and other similar macros. -- Thanks, Petr
