entry: Fix the incorrect ordering of lockdep and RCU check

entry: Fix the incorrect ordering of lockdep and RCU check

[Thomas Gleixner]

When an exception/interrupt hits kernel space and the kernel is not
currently in the idle task then RCU must be watching.

irqentry_enter() validates this via rcu_irq_enter_check_tick(), which in
turn invokes lockdep when taking a lock. But at that point lockdep does not
yet know about the fact that interrupts have been disabled by the CPU,
which triggers a lockdep splat complaining about inconsistent state.

Invoking trace_hardirqs_off() before rcu_irq_enter_check_tick() defeats the
point of rcu_irq_enter_check_tick() because trace_hardirqs_off() uses RCU.

So use the same sequence as for the idle case and tell lockdep about the
irq state change first, invoke the RCU check and then do the lockdep and
tracer update.

Fixes: a5497bab5f72 ("entry: Provide generic interrupt entry/exit code")
Reported-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
---
 kernel/entry/common.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

--- a/kernel/entry/common.c
+++ b/kernel/entry/common.c
@@ -337,10 +337,10 @@ noinstr irqentry_state_t irqentry_enter(
 	 * already contains a warning when RCU is not watching, so no point
 	 * in having another one here.
 	 */
+	lockdep_hardirqs_off(CALLER_ADDR0);
 	instrumentation_begin();
 	rcu_irq_enter_check_tick();
-	/* Use the combo lockdep/tracing function */
-	trace_hardirqs_off();
+	trace_hardirqs_off_finish();
 	instrumentation_end();
 
 	return ret;

Re: entry: Fix the incorrect ordering of lockdep and RCU check

[Mark Rutland]

On Wed, Nov 04, 2020 at 02:06:23PM +0100, Thomas Gleixner wrote:
> When an exception/interrupt hits kernel space and the kernel is not
> currently in the idle task then RCU must be watching.
> 
> irqentry_enter() validates this via rcu_irq_enter_check_tick(), which in
> turn invokes lockdep when taking a lock. But at that point lockdep does not
> yet know about the fact that interrupts have been disabled by the CPU,
> which triggers a lockdep splat complaining about inconsistent state.
> 
> Invoking trace_hardirqs_off() before rcu_irq_enter_check_tick() defeats the
> point of rcu_irq_enter_check_tick() because trace_hardirqs_off() uses RCU.
> 
> So use the same sequence as for the idle case and tell lockdep about the
> irq state change first, invoke the RCU check and then do the lockdep and
> tracer update.
> 
> Fixes: a5497bab5f72 ("entry: Provide generic interrupt entry/exit code")
> Reported-by: Mark Rutland <mark.rutland@arm.com>
> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
> Cc: stable@vger.kernel.org

I just gave this a spin on x86_64 defconfig + PROVE_LOCKING +
DEBUG_LOCKDEP + NO_HZ_FULL + CONTEXT_TRACKING_FORCE, and it gets rid of
the boot-time splat. So FWIW:

Tested-by: Mark Rutland <mark.rutland@arm.com>

Mark.

> ---
>  kernel/entry/common.c |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> --- a/kernel/entry/common.c
> +++ b/kernel/entry/common.c
> @@ -337,10 +337,10 @@ noinstr irqentry_state_t irqentry_enter(
>  	 * already contains a warning when RCU is not watching, so no point
>  	 * in having another one here.
>  	 */
> +	lockdep_hardirqs_off(CALLER_ADDR0);
>  	instrumentation_begin();
>  	rcu_irq_enter_check_tick();
> -	/* Use the combo lockdep/tracing function */
> -	trace_hardirqs_off();
> +	trace_hardirqs_off_finish();
>  	instrumentation_end();
>  
>  	return ret;

[tip: core/urgent] entry: Fix the incorrect ordering of lockdep and RCU check

[tip-bot2 for Thomas Gleixner]

The following commit has been merged into the core/urgent branch of tip:

Commit-ID:     9d820f68b2bdba5b2e7bf135123c3f57c5051d05
Gitweb:        https://git.kernel.org/tip/9d820f68b2bdba5b2e7bf135123c3f57c5051d05
Author:        Thomas Gleixner <tglx@linutronix.de>
AuthorDate:    Wed, 04 Nov 2020 14:06:23 +01:00
Committer:     Thomas Gleixner <tglx@linutronix.de>
CommitterDate: Wed, 04 Nov 2020 18:06:14 +01:00

entry: Fix the incorrect ordering of lockdep and RCU check

When an exception/interrupt hits kernel space and the kernel is not
currently in the idle task then RCU must be watching.

irqentry_enter() validates this via rcu_irq_enter_check_tick(), which in
turn invokes lockdep when taking a lock. But at that point lockdep does not
yet know about the fact that interrupts have been disabled by the CPU,
which triggers a lockdep splat complaining about inconsistent state.

Invoking trace_hardirqs_off() before rcu_irq_enter_check_tick() defeats the
point of rcu_irq_enter_check_tick() because trace_hardirqs_off() uses RCU.

So use the same sequence as for the idle case and tell lockdep about the
irq state change first, invoke the RCU check and then do the lockdep and
tracer update.

Fixes: a5497bab5f72 ("entry: Provide generic interrupt entry/exit code")
Reported-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/87y2jhl19s.fsf@nanos.tec.linutronix.de

---
 kernel/entry/common.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/entry/common.c b/kernel/entry/common.c
index 2b83666..e9e2df3 100644
--- a/kernel/entry/common.c
+++ b/kernel/entry/common.c
@@ -337,10 +337,10 @@ noinstr irqentry_state_t irqentry_enter(struct pt_regs *regs)
 	 * already contains a warning when RCU is not watching, so no point
 	 * in having another one here.
 	 */
+	lockdep_hardirqs_off(CALLER_ADDR0);
 	instrumentation_begin();
 	rcu_irq_enter_check_tick();
-	/* Use the combo lockdep/tracing function */
-	trace_hardirqs_off();
+	trace_hardirqs_off_finish();
 	instrumentation_end();
 
 	return ret;