Re: [PATCH mm 1/2] kasan: don't use read-only static keys

[Kees Cook <keescook@chromium.org>]

On Wed, Dec 09, 2020 at 07:49:36PM +0100, Marco Elver wrote:
> On Wed, 9 Dec 2020 at 19:24, Andrey Konovalov <andreyknvl@google.com> wrote:
> > __ro_after_init static keys are incompatible with usage in loadable kernel
> > modules and cause crashes. Don't use those, use normal static keys.
> >
> > Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
> 
> Reviewed-by: Marco Elver <elver@google.com>
> 
> > ---
> >
> > This fix can be squashed into
> > "kasan: add and integrate kasan boot parameters".
> >
> > ---
> >  mm/kasan/hw_tags.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/mm/kasan/hw_tags.c b/mm/kasan/hw_tags.c
> > index c91f2c06ecb5..55bd6f09c70f 100644
> > --- a/mm/kasan/hw_tags.c
> > +++ b/mm/kasan/hw_tags.c
> > @@ -43,11 +43,11 @@ static enum kasan_arg_stacktrace kasan_arg_stacktrace __ro_after_init;
> >  static enum kasan_arg_fault kasan_arg_fault __ro_after_init;
> >
> >  /* Whether KASAN is enabled at all. */
> > -DEFINE_STATIC_KEY_FALSE_RO(kasan_flag_enabled);
> > +DEFINE_STATIC_KEY_FALSE(kasan_flag_enabled);
> 
> Side-node: This appears to be just a bad interface; I think the macro
> DEFINE_STATIC_KEY_FALSE_RO() is error-prone, if it can't be guaranteed
> that this is always safe, since the presence of the macro encourages
> its use and we'll inevitably run into this problem again.
> 
> >  EXPORT_SYMBOL(kasan_flag_enabled);
> 
> DEFINE_STATIC_KEY_FALSE_RO() + EXPORT_SYMBOL() is an immediate bug.
> Given its use has not increased substantially since its introduction,
> it may be safer to consider its removal.

Right -- it seems the export is the problem, not the RO-ness. What is
actually trying to change the flag after __init?

-- 
Kees Cook