Re: [PATCH 4/4] mm/kasan: make kasan=on|off take effect for all three modes

[Baoquan He <bhe@redhat.com>]

On 08/06/25 at 05:26pm, Lorenzo Stoakes wrote:
> On Tue, Aug 05, 2025 at 10:22:31PM -0700, SeongJae Park wrote:
> > Hello Baoqua,
> >
> > On Tue,  5 Aug 2025 14:23:33 +0800 Baoquan He <bhe@redhat.com> wrote:
> >
> > > Now everything is ready, set kasan=off can disable kasan for all
> > > three modes.
> > >
> > > Signed-off-by: Baoquan He <bhe@redhat.com>
> > > ---
> > >  include/linux/kasan-enabled.h | 11 +----------
> > >  1 file changed, 1 insertion(+), 10 deletions(-)
> > >
> > > diff --git a/include/linux/kasan-enabled.h b/include/linux/kasan-enabled.h
> > > index 32f2d19f599f..b5857e15ef14 100644
> > > --- a/include/linux/kasan-enabled.h
> > > +++ b/include/linux/kasan-enabled.h
> > > @@ -8,30 +8,21 @@ extern bool kasan_arg_disabled;
> > >
> > >  DECLARE_STATIC_KEY_FALSE(kasan_flag_enabled);
> > >
> > > -#ifdef CONFIG_KASAN_HW_TAGS
> > > -
> > >  static __always_inline bool kasan_enabled(void)
> > >  {
> > >  	return static_branch_likely(&kasan_flag_enabled);
> > >  }
> >
> > I found mm-new build fails when CONFIG_KASAN is unset as below, and 'git
> > bisect' points this patch.
> 
> Yup just hit this + bisected here.

Sorry for the trouble and thanks for reporting.

> 
> >
> >       LD      .tmp_vmlinux1
> >     ld: lib/stackdepot.o:(__jump_table+0x8): undefined reference to `kasan_flag_enabled'
> >
> > Since kasna_flag_enabled is defined in mm/kasan/common.c, I confirmed diff like
> > below fixes this.  I think it may not be a correct fix though, since I didn't
> > read this patchset thoroughly.
> >
> > diff --git a/include/linux/kasan-enabled.h b/include/linux/kasan-enabled.h
> > index b5857e15ef14..a53d112b1020 100644
> > --- a/include/linux/kasan-enabled.h
> > +++ b/include/linux/kasan-enabled.h
> > @@ -8,11 +8,22 @@ extern bool kasan_arg_disabled;
> >
> >  DECLARE_STATIC_KEY_FALSE(kasan_flag_enabled);
> >
> > +#ifdef CONFIG_KASAN
> > +
> 
> Shouldn't we put this above the static key declaration?
> 
> Feels like the whole header should be included really.

You are right, kasan_flag_enabled should be included in CONFIG_KASAN
ifdeffery scope.

Since CONFIG_KASAN_HW_TAGS depends on CONFIG_KASAN, we may not need
include below CONFIG_KASAN_HW_TAGS ifdeffery into CONFIG_KASAN ifdeffery
scope. Not sure if this is incorrect.

Thanks a lot for checking this.
> 
> >  static __always_inline bool kasan_enabled(void)
> >  {
> >  	return static_branch_likely(&kasan_flag_enabled);
> >  }
> >
> > +#else /* CONFIG_KASAN */
> > +
> > +static inline bool kasan_enabled(void)
> > +{
> > +	return false;
> > +}
> > +
> > +#endif
> > +
> >  #ifdef CONFIG_KASAN_HW_TAGS
> >  static inline bool kasan_hw_tags_enabled(void)
> >  {
> >
> >
> > [...]
> >
> > Thanks,
> > SJ
> >
> 
> Cheers, Lorenzo
>