Peter Zijlstra wrote:
> On Mon, Mar 26, 2018 at 07:18:33PM +0900, Tetsuo Handa wrote:
> > [ 628.863629] 2 locks held by a.out/1165:
> > [ 628.867533] #0: [ffffa3b438472e48] (&mm->mmap_sem){++++}, at:
> > __do_page_fault+0x16f/0x4d0
> > [ 628.873570] #1: [ffffa3b4f2c52ac0] (&mapping->i_mmap_rwsem){++++}, at:
> > rmap_walk_file+0x1d9/0x2a0
>
> Maybe change the string a little, because from the above it's not at all
> effident that the [] thing is the lock instance.
>
> > diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
> > index 12a2805..7835233 100644
> > --- a/kernel/locking/lockdep.c
> > +++ b/kernel/locking/lockdep.c
> > @@ -556,9 +556,9 @@ static void print_lock(struct held_lock *hlock)
> > return;
> > }
> >
> > + printk(KERN_CONT "[%px]", hlock->instance);
>
> And yeah, what Michal said, that wants to be %p, we're fine with the
> thing being hashed, all we want to do is equivalience, which can be done
> with hashed pinters too.
>
> > print_lock_name(lock_classes + class_idx - 1);
> > - printk(KERN_CONT ", at: [<%px>] %pS\n",
> > - (void *)hlock->acquire_ip, (void *)hlock->acquire_ip);
> > + printk(KERN_CONT ", at: %pS\n", (void *)hlock->acquire_ip);
> > }
>
> Otherwise no real objection to the patch.
>
I see. What about plain
- printk(KERN_CONT "[%px]", hlock->instance);
+ printk(KERN_CONT "%p", hlock->instance);
because we don't need to use [] ?
I'm trying to remove "[<%px>]" for hlock->acquire_ip field in order to
reduce amount of output, for debug_show_all_locks() prints a lot.
