Re: [PATCH v5 09/10] mmap locking API: add mmap_assert_locked

[Michel Lespinasse]

On Tue, Apr 21, 2020 at 07:18:50PM -0700, Matthew Wilcox wrote:
> On Tue, Apr 21, 2020 at 07:10:39PM -0700, Michel Lespinasse wrote:
> > On Tue, Apr 21, 2020 at 5:14 PM Michel Lespinasse <walken@xxxxxxxxxx> wrote:
> > > +static inline void mmap_assert_locked(struct mm_struct *mm)
> > > +{
> > > +       if (IS_ENABLED(CONFIG_LOCKDEP) && debug_locks)
> > > +               VM_BUG_ON_MM(!lockdep_is_held(&mm->mmap_sem), mm);
> > > +       else
> > > +               VM_BUG_ON_MM(!rwsem_is_locked(&mm->mmap_sem), mm);
> > > +}
> > 
> > Turns out this function definition does not work for !CONFIG_LOCKDEP
> > as lockdep_is_held is not defined in that case.
> 
> Oops, sorry.  It only defines
> 
> #define lockdep_is_held_type(l, r)              (1)
> #define lockdep_assert_held(l)                  do { (void)(l); } while (0)
> #define lockdep_assert_held_write(l)    do { (void)(l); } while (0)
> #define lockdep_assert_held_read(l)             do { (void)(l); } while (0)
> #define lockdep_assert_held_once(l)             do { (void)(l); } while (0)
> 
> which seems like an oversight, but not one that you should be fixing.
> 
> > The following should work instead:
> > 
> > static inline void mmap_assert_locked(struct mm_struct *mm)
> > {
> > #ifdef CONFIG_LOCKDEP
> >         if (debug_locks) {
> >                 VM_BUG_ON_MM(!lockdep_is_held(&mm->mmap_lock), mm);
> >                 return;
> >         }
> > #endif
> >         VM_BUG_ON_MM(!rwsem_is_locked(&mm->mmap_lock), mm);
> > }
> 
> Yes, I agree.

Sent an updated version of this
(also integrating your feedback on patch 10/10)

Please look for it under subject: [PATCH v5.5 09/10] mmap locking API: add mmap_assert_locked() and mmap_assert_write_locked()

-- 
Michel "Walken" Lespinasse
A program is never fully debugged until the last user dies.