Re: [PATCH v7 06/16] lockdep: Detect and handle hist_lock ring buffer overwrite

[Peter Zijlstra]

On Fri, Jul 14, 2017 at 03:42:10PM +0900, Byungchul Park wrote:
> On Thu, Jul 13, 2017 at 08:23:33PM +0900, Byungchul Park wrote:
> > On Thu, Jul 13, 2017 at 8:12 PM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> > > On Thu, Jul 13, 2017 at 12:29:05PM +0200, Peter Zijlstra wrote:
> > >> On Thu, Jul 13, 2017 at 07:09:53PM +0900, Byungchul Park wrote:
> > >> > On Thu, Jul 13, 2017 at 11:50:52AM +0200, Peter Zijlstra wrote:
> > >> > >   wait_for_completion(&C);
> > >> > >     atomic_inc_return();
> > >> > >
> > >> > >                                   mutex_lock(A1);
> > >> > >                                   mutex_unlock(A1);
> > >> > >
> > >> > >
> > >> > >                                   <IRQ>
> > >> > >                                     spin_lock(B1);
> > >> > >                                     spin_unlock(B1);
> > >> > >
> > >> > >                                     ...
> > >> > >
> > >> > >                                     spin_lock(B64);
> > >> > >                                     spin_unlock(B64);
> > >> > >                                   </IRQ>
> > >> > >
> > >> > >
> > >
> > > Also consider the alternative:
> > >
> > >                                         <IRQ>
> > >                                           spin_lock(D);
> > >                                           spin_unlock(D);
> > >
> > >                                           complete(&C);
> > >                                         </IRQ>
> > >
> > > in which case the context test will also not work.
> > 
> > Context tests are done on xhlock with the release context, _not_
> > acquisition context. For example, spin_lock(D) and complete(&C) are
> > in the same context, so the test would pass in this example.

The point was, this example will also link C to B*.

(/me copy paste from older email)

That gives:

        xhist[ 0] = A1
        xhist[ 1] = B1
        ...
        xhist[63] = B63

then we wrap and have:

        xhist[0] = B64

then we rewind to 1 and invalidate to arrive at:

        xhist[ 0] = B64
        xhist[ 1] = NULL   <-- idx
        xhist[ 2] = B2
        ...
        xhist[63] = B63


Then we do D and get

        xhist[ 0] = B64
        xhist[ 1] = D   <-- idx
        xhist[ 2] = B2
        ...
        xhist[63] = B63


And now there is nothing that will invalidate B*, after all, the
gen_id's are all after C's stamp, and the same_context_xhlock() test
will also pass because they're all from IRQ context (albeit not the
same, but it cannot tell).


Does this explain? Or am I still missing something?