On Wed, Apr 23, 2025 at 10:19:56AM +0300, Dan Carpenter wrote:
> On Wed, Apr 23, 2025 at 10:17:16AM +0300, Dan Carpenter wrote:
> > Hello Joel Fernandes,
> >
> > Commit bd57ec707441 ("rcutorture: Perform more frequent testing of
> > ->gpwrap") from Feb 16, 2025 (linux-next), leads to the following
> > Smatch static checker warning:
> >
> > kernel/rcu/rcutorture.c:4586 rcu_torture_init()
> > warn: missing error code 'firsterr'
> >
> > kernel/rcu/rcutorture.c
> > 4576 if (torture_init_error(firsterr))
> > 4577 goto unwind;
> > 4578 }
> > 4579 if (object_debug)
> > 4580 rcu_test_debug_objects();
> > 4581 torture_init_end();
> ^^^^^^^^^^^^^^^^^^^
> Also:
>
> kernel/rcu/rcutorture.c:4591 rcu_torture_init() warn: double unlock 'global
> &fullstop_mutex' (orig line 4581)
You lost me on this one. This mutex is acquired by the earlier call to
torture_init_begin(), but only if that function returns true. If that
function returns false, yes, it releases fullstop_mutex, but in that case
rcu_torture_init() returns immediately, thus avoiding this invocation
of torture_init_end().
Or am I missing something subtle here? Or missing something obvious,
for that matter! ;-)
Thanx, Paul
> > 4582 if (cur_ops->gp_slow_register &&
> > !WARN_ON_ONCE(!cur_ops->gp_slow_unregister))
> > 4583 cur_ops->gp_slow_register(&rcu_fwd_cb_nodelay);
> > 4584
> > 4585 if (cur_ops->set_gpwrap_lag && rcu_gpwrap_lag_init())
> > --> 4586 goto unwind;
> >
> > set an error code?
> >
> > 4587
> > 4588 return 0;
> > 4589
> > 4590 unwind:
> > 4591 torture_init_end();
> ^^^^^^^^^^^^^^^^^^^
>
> regards,
> dan carpenter
>
> > 4592 rcu_torture_cleanup();
> > 4593 if (shutdown_secs) {
> > 4594 WARN_ON(!IS_MODULE(CONFIG_RCU_TORTURE_TEST));
> > 4595 kernel_power_off();
> > 4596 }
> > 4597 return firsterr;
> > 4598 }
> >
> > regards,
> > dan carpenter
>
