Re: [patch V3 01/32] mm/slab: Fix broken stack trace storage

[Andy Lutomirski]

On Sun, Apr 14, 2019 at 9:34 AM Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
>
> On Sun, 14 Apr 2019, Andy Lutomirski wrote:
> > > +               struct stack_trace trace = {
> > > +                       .max_entries    = size - 4;
> > > +                       .entries        = addr;
> > > +                       .skip           = 3;
> > > +               };
> >
> > This looks correct, but I think that it would have been clearer if you
> > left the size -= 3 above.  You're still incrementing addr, but you're
> > not decrementing size, so they're out of sync and the resulting code
> > is hard to follow.
>
> What about the below?
>
> --- a/mm/slab.c
> +++ b/mm/slab.c
> @@ -1480,10 +1480,12 @@ static void store_stackinfo(struct kmem_
>         *addr++ = 0x12345678;
>         *addr++ = caller;
>         *addr++ = smp_processor_id();
> +       size -= 3;
>  #ifdef CONFIG_STACKTRACE
>         {
>                 struct stack_trace trace = {
> -                       .max_entries    = size - 4;
> +                       /* Leave one for the end marker below */
> +                       .max_entries    = size - 1;
>                         .entries        = addr;
>                         .skip           = 3;
>                 };

Looks good to me.