arm: Use _rcuidle tracepoint to allow use from idle
authorPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Sun, 24 Apr 2016 21:30:16 +0000 (14:30 -0700)
committerPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Fri, 29 Apr 2016 20:29:02 +0000 (13:29 -0700)
Testing on ARM encountered the following pair of lockdep-RCU splats:

------------------------------------------------------------------------

===============================
[ INFO: suspicious RCU usage. ]
4.6.0-rc4-next-20160422 #1 Not tainted
-------------------------------
include/trace/events/power.h:328 suspicious rcu_dereference_check() usage!

other info that might help us debug this:

RCU used illegally from idle CPU!
rcu_scheduler_active = 1, debug_locks = 0
RCU used illegally from extended quiescent state!
no locks held by swapper/0/0.

stack backtrace:
CPU: 0 PID: 0 Comm: swapper/0 Not tainted 4.6.0-rc4-next-20160422 #1
Hardware name: Generic OMAP3-GP (Flattened Device Tree)
[<c010f55c>] (unwind_backtrace) from [<c010b64c>] (show_stack+0x10/0x14)
[<c010b64c>] (show_stack) from [<c047acbc>] (dump_stack+0xa8/0xe0)
[<c047acbc>] (dump_stack) from [<c012bc10>] (pwrdm_set_next_pwrst+0xf8/0x1cc)
[<c012bc10>] (pwrdm_set_next_pwrst) from [<c01269fc>] (omap3_enter_idle_bm+0x1b8/0x1e8)
[<c01269fc>] (omap3_enter_idle_bm) from [<c05fa0b8>] (cpuidle_enter_state+0x84/0x408)
[<c05fa0b8>] (cpuidle_enter_state) from [<c0182c1c>] (cpu_startup_entry+0x1c8/0x3f0)
[<c0182c1c>] (cpu_startup_entry) from [<c0b00c20>] (start_kernel+0x354/0x3cc)

------------------------------------------------------------------------

[<c010f55c>] (unwind_backtrace) from [<c010b64c>] (show_stack+0x10/0x14)
[<c010b64c>] (show_stack) from [<c047ac3c>] (dump_stack+0xa8/0xe0)
[<c047ac3c>] (dump_stack) from [<c012c340>] (_pwrdm_state_switch+0x188/0x32c)
[<c012c340>] (_pwrdm_state_switch) from [<c012c4f0>] (_pwrdm_post_transition_cb+0xc/0x14)
[<c012c4f0>] (_pwrdm_post_transition_cb) from [<c012ba74>] (pwrdm_for_each+0x30/0x5c)
[<c012ba74>] (pwrdm_for_each) from [<c012c72c>] (pwrdm_post_transition+0x24/0x30)
[<c012c72c>] (pwrdm_post_transition) from [<c012548c>] (omap_sram_idle+0xfc/0x240)
[<c012548c>] (omap_sram_idle) from [<c0126934>] (omap3_enter_idle_bm+0xf0/0x1e8)
[<c0126934>] (omap3_enter_idle_bm) from [<c05fa038>] (cpuidle_enter_state+0x84/0x408)
[<c05fa038>] (cpuidle_enter_state) from [<c0182b90>] (cpu_startup_entry+0x1c8/0x3f0)
[<c0182b90>] (cpu_startup_entry) from [<c0b00c20>] (start_kernel+0x354/0x3cc)

------------------------------------------------------------------------

These are caused by event tracing from the idle loop, and they were
exposed by commit 293e2421fe25 ("rcu: Remove superfluous versions of
rcu_read_lock_sched_held()"), which suppressed some false negatives.
The current commit therefore adds the _rcuidle suffix to make RCU aware
of this implicit use of RCU by event tracing, thus preventing both splats.

Reported-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Tony Lindgren <tony@atomide.com>
Cc: Russell King <linux@arm.linux.org.uk>
Reviewed-by: Steven Rostedt <rostedt@goodmis.org>
Cc: <linux-omap@vger.kernel.org>
Cc: <linux-arm-kernel@lists.infradead.org>
arch/arm/mach-omap2/powerdomain.c

index 78af6d8cf2e208811f2d6b31e1aa837c3b585565..daf2753de7aa2a4681c690e49fb00d874c856a90 100644 (file)
@@ -186,8 +186,9 @@ static int _pwrdm_state_switch(struct powerdomain *pwrdm, int flag)
                        trace_state = (PWRDM_TRACE_STATES_FLAG |
                                       ((next & OMAP_POWERSTATE_MASK) << 8) |
                                       ((prev & OMAP_POWERSTATE_MASK) << 0));
-                       trace_power_domain_target(pwrdm->name, trace_state,
-                                                 smp_processor_id());
+                       trace_power_domain_target_rcuidle(pwrdm->name,
+                                                         trace_state,
+                                                         smp_processor_id());
                }
                break;
        default:
@@ -523,8 +524,8 @@ int pwrdm_set_next_pwrst(struct powerdomain *pwrdm, u8 pwrst)
 
        if (arch_pwrdm && arch_pwrdm->pwrdm_set_next_pwrst) {
                /* Trace the pwrdm desired target state */
-               trace_power_domain_target(pwrdm->name, pwrst,
-                                         smp_processor_id());
+               trace_power_domain_target_rcuidle(pwrdm->name, pwrst,
+                                                 smp_processor_id());
                /* Program the pwrdm desired target state */
                ret = arch_pwrdm->pwrdm_set_next_pwrst(pwrdm, pwrst);
        }