Re: [PATCH v5 07/15] locking/lockdep: Free lock classes that are no longer in use

[Peter Zijlstra]

On Thu, Jan 10, 2019 at 11:31:11AM -0800, Bart Van Assche wrote:
> On Thu, 2019-01-10 at 16:28 +0100, Peter Zijlstra wrote:
> > On Mon, Dec 17, 2018 at 01:29:54PM -0800, Bart Van Assche wrote:
> > > +static bool inside_selftest(void)
> > > +{
> > > +	return current == lockdep_selftest_task_struct;
> > > +}
> > > +void lockdep_free_key_range(void *start, unsigned long size)
> > > +{
> > > +	init_data_structures_once();
> > > +
> > > +	if (inside_selftest())
> > > +		lockdep_free_key_range_imm(start, size);
> > > +	else
> > > +		lockdep_free_key_range_reg(start, size);
> > >  }
> > 
> > That is .... unfortunate. The whole reason that whole immediate thing
> > works at all is because there is no concurrency what so ever that early,
> > right?
> > 
> > Should we maybe key off of: 'system_state == SYSTEM_BOOTING' instead?
> 
> Hi Peter,
> 
> I agree that it is unfortunate that the selftests require a different code
> path. I have not been able to find any way to avoid this. Using the test
> system_state == SYSTEM_BOOTING should work and will allow to remove the
> lockdep_selftest_task_struct variable. Do you want me to make that change?

I can do it; another approach is simply using the _imm (maybe renamed to
_selftest) functions directly, and getting rid of this dynamic choice.

I'm undecided as of yet.