Skip to contentSkip to conversation

lockdep: PROVE_RAW_LOCK_NESTING=y fixes

PATCH Applied Peter Zijlstra 3 participants 8 messages Jun 17, 2021 – Jun 23, 2021

Subsystems (2)
lore
Share & tools
lore

Peter Zijlstra

author maintainer
#0 lore raw

cover lockdep: PROVE_RAW_LOCK_NESTING=y fixes

Hi,

Boris said, that Joerg said the locking selftests showed a few FAILED. They were right.

author maintainer
#1 lore raw

PATCH 1/3 lockdep: Fix wait-type for empty stack

Even the very first lock can violate the wait-context check, consider the various IRQ contexts.

Fixes de8f5e4f2dc1 ("lockdep: Introduce wait-type checks")
---
 kernel/locking/lockdep.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/locking/lockdep.c +1 -1
@@ -4577,7 +4577,7 @@ static int check_wait_context(struct tas
 	u8 curr_inner;
 	int depth;
 
-	if (!curr->lockdep_depth || !next_inner || next->trylock)
+	if (!next_inner || next->trylock)
 		return 0;
 
 	if (!next_outer)
merge bot Applied
#2 lore raw

[tip: locking/core] lockdep: Fix wait-type for empty stack

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

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

Commit-ID:     f8b298cc39f0619544c607eaef09fd0b2afd10f3
Gitweb:        https://git.kernel.org/tip/f8b298cc39f0619544c607eaef09fd0b2afd10f3
Author:        Peter Zijlstra <[email protected]>
AuthorDate:    Thu, 17 Jun 2021 20:57:18 +02:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Tue, 22 Jun 2021 16:42:08 +02:00

lockdep: Fix wait-type for empty stack

Even the very first lock can violate the wait-context check, consider the various IRQ contexts.

Fixes de8f5e4f2dc1 ("lockdep: Introduce wait-type checks")
---
 kernel/locking/lockdep.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/locking/lockdep.c +1 -1
index 6ff1e84..0584b20 100644
@@ -4688,7 +4688,7 @@ static int check_wait_context(struct task_struct *curr, struct held_lock *next)
 	u8 curr_inner;
 	int depth;
 
-	if (!curr->lockdep_depth || !next_inner || next->trylock)
+	if (!next_inner || next->trylock)
 		return 0;
 
 	if (!next_outer)
author maintainer
#3 lore raw

PATCH 2/3 lockdep/selftests: Fix selftests vs PROVE_RAW_LOCK_NESTING

In reply to Peter Zijlstra

When PROVE_RAW_LOCK_NESTING=y many of the selftests FAILED because HARDIRQ context is out-of-bounds for spinlocks. Instead make the default hardware context the threaded hardirq context, which preserves the old locking rules.

The wait-type specific locking selftests will have a non-threaded HARDIRQ variant.

Fixes de8f5e4f2dc1 ("lockdep: Introduce wait-type checks")
---
 lib/locking-selftest.c |    1 +
 1 file changed, 1 insertion(+)
lib/locking-selftest.c +1 -0
@@ -194,6 +194,7 @@ static void init_shared_classes(void)
 #define HARDIRQ_ENTER()				\
 	local_irq_disable();			\
 	__irq_enter();				\
+	lockdep_hardirq_threaded();		\
 	WARN_ON(!in_irq());
 
 #define HARDIRQ_EXIT()				\
merge bot Applied
#4 lore raw

[tip: locking/core] lockdep/selftests: Fix selftests vs PROVE_RAW_LOCK_NESTING

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

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

Commit-ID:     c0c2c0dad6a06e0c05e9a52d65f932bd54364c97
Gitweb:        https://git.kernel.org/tip/c0c2c0dad6a06e0c05e9a52d65f932bd54364c97
Author:        Peter Zijlstra <[email protected]>
AuthorDate:    Thu, 17 Jun 2021 20:57:19 +02:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Tue, 22 Jun 2021 16:42:08 +02:00

lockdep/selftests: Fix selftests vs PROVE_RAW_LOCK_NESTING

When PROVE_RAW_LOCK_NESTING=y many of the selftests FAILED because HARDIRQ context is out-of-bounds for spinlocks. Instead make the default hardware context the threaded hardirq context, which preserves the old locking rules.

The wait-type specific locking selftests will have a non-threaded HARDIRQ variant.

Fixes de8f5e4f2dc1 ("lockdep: Introduce wait-type checks")
---
 lib/locking-selftest.c | 1 +
 1 file changed, 1 insertion(+)
lib/locking-selftest.c +1 -0
index 5c50b09..af12e84 100644
@@ -195,6 +195,7 @@ static void init_shared_classes(void)
 #define HARDIRQ_ENTER()				\
 	local_irq_disable();			\
 	__irq_enter();				\
+	lockdep_hardirq_threaded();		\
 	WARN_ON(!in_irq());
 
 #define HARDIRQ_EXIT()				\
author maintainer
#5 lore raw

PATCH 3/3 lockdep/selftest: Remove wait-type RCU_CALLBACK tests

In reply to Peter Zijlstra

The problem is that rcu_callback_map doesn't have wait_types defined, and doing so would make it indistinguishable from SOFTIRQ in any case. Remove it.

Fixes 9271a40d2a14 ("lockdep/selftest: Add wait context selftests")
---
 lib/locking-selftest.c |   17 -----------------
 1 file changed, 17 deletions(-)
lib/locking-selftest.c +0 -17
@@ -2492,16 +2492,6 @@ static void rcu_sched_exit(int *_)
 	int rcu_sched_guard_##name __guard(rcu_sched_exit);	\
 	rcu_read_lock_sched();
 
-static void rcu_callback_exit(int *_)
-{
-	rcu_lock_release(&rcu_callback_map);
-}
-
-#define RCU_CALLBACK_CONTEXT(name, ...)					\
-	int rcu_callback_guard_##name __guard(rcu_callback_exit);	\
-	rcu_lock_acquire(&rcu_callback_map);
-
-
 static void raw_spinlock_exit(raw_spinlock_t **lock)
 {
 	raw_spin_unlock(*lock);
@@ -2558,8 +2548,6 @@ static void __maybe_unused inner##_in_##
  * ---------------+-------+----------+------+-------
  * RCU_BH         |   o   |    o     |  o   |  x
  * ---------------+-------+----------+------+-------
- * RCU_CALLBACK   |   o   |    o     |  o   |  x
- * ---------------+-------+----------+------+-------
  * RCU_SCHED      |   o   |    o     |  x   |  x
  * ---------------+-------+----------+------+-------
  * RAW_SPIN       |   o   |    o     |  x   |  x
@@ -2576,7 +2564,6 @@ GENERATE_2_CONTEXT_TESTCASE(NOTTHREADED_
 GENERATE_2_CONTEXT_TESTCASE(SOFTIRQ, , inner, inner_lock)			\
 GENERATE_2_CONTEXT_TESTCASE(RCU, , inner, inner_lock)				\
 GENERATE_2_CONTEXT_TESTCASE(RCU_BH, , inner, inner_lock)			\
-GENERATE_2_CONTEXT_TESTCASE(RCU_CALLBACK, , inner, inner_lock)			\
 GENERATE_2_CONTEXT_TESTCASE(RCU_SCHED, , inner, inner_lock)			\
 GENERATE_2_CONTEXT_TESTCASE(RAW_SPINLOCK, raw_lock_A, inner, inner_lock)	\
 GENERATE_2_CONTEXT_TESTCASE(SPINLOCK, lock_A, inner, inner_lock)		\
@@ -2638,10 +2625,6 @@ static void wait_context_tests(void)
 	DO_CONTEXT_TESTCASE_OUTER_LIMITED_PREEMPTIBLE(RCU_BH);
 	pr_cont("\n");
 
-	print_testname("in RCU callback context");
-	DO_CONTEXT_TESTCASE_OUTER_LIMITED_PREEMPTIBLE(RCU_CALLBACK);
-	pr_cont("\n");
-
 	print_testname("in RCU-sched context");
 	DO_CONTEXT_TESTCASE_OUTER_NOT_PREEMPTIBLE(RCU_SCHED);
 	pr_cont("\n");
merge bot Applied
#6 lore raw

[tip: locking/core] lockdep/selftest: Remove wait-type RCU_CALLBACK tests

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

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

Commit-ID:     1a8122960484b19d8d887fb32e1cf42be5647533
Gitweb:        https://git.kernel.org/tip/1a8122960484b19d8d887fb32e1cf42be5647533
Author:        Peter Zijlstra <[email protected]>
AuthorDate:    Thu, 17 Jun 2021 20:57:20 +02:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Tue, 22 Jun 2021 16:42:08 +02:00

lockdep/selftest: Remove wait-type RCU_CALLBACK tests

The problem is that rcu_callback_map doesn't have wait_types defined, and doing so would make it indistinguishable from SOFTIRQ in any case. Remove it.

Fixes 9271a40d2a14 ("lockdep/selftest: Add wait context selftests")
---
 lib/locking-selftest.c | 17 -----------------
 1 file changed, 17 deletions(-)
lib/locking-selftest.c +0 -17
index af12e84..161108e 100644
@@ -2494,16 +2494,6 @@ static void rcu_sched_exit(int *_)
 	int rcu_sched_guard_##name __guard(rcu_sched_exit);	\
 	rcu_read_lock_sched();
 
-static void rcu_callback_exit(int *_)
-{
-	rcu_lock_release(&rcu_callback_map);
-}
-
-#define RCU_CALLBACK_CONTEXT(name, ...)					\
-	int rcu_callback_guard_##name __guard(rcu_callback_exit);	\
-	rcu_lock_acquire(&rcu_callback_map);
-
-
 static void raw_spinlock_exit(raw_spinlock_t **lock)
 {
 	raw_spin_unlock(*lock);
@@ -2560,8 +2550,6 @@ static void __maybe_unused inner##_in_##outer(void)				\
  * ---------------+-------+----------+------+-------
  * RCU_BH         |   o   |    o     |  o   |  x
  * ---------------+-------+----------+------+-------
- * RCU_CALLBACK   |   o   |    o     |  o   |  x
- * ---------------+-------+----------+------+-------
  * RCU_SCHED      |   o   |    o     |  x   |  x
  * ---------------+-------+----------+------+-------
  * RAW_SPIN       |   o   |    o     |  x   |  x
@@ -2578,7 +2566,6 @@ GENERATE_2_CONTEXT_TESTCASE(NOTTHREADED_HARDIRQ, , inner, inner_lock)		\
 GENERATE_2_CONTEXT_TESTCASE(SOFTIRQ, , inner, inner_lock)			\
 GENERATE_2_CONTEXT_TESTCASE(RCU, , inner, inner_lock)				\
 GENERATE_2_CONTEXT_TESTCASE(RCU_BH, , inner, inner_lock)			\
-GENERATE_2_CONTEXT_TESTCASE(RCU_CALLBACK, , inner, inner_lock)			\
 GENERATE_2_CONTEXT_TESTCASE(RCU_SCHED, , inner, inner_lock)			\
 GENERATE_2_CONTEXT_TESTCASE(RAW_SPINLOCK, raw_lock_A, inner, inner_lock)	\
 GENERATE_2_CONTEXT_TESTCASE(SPINLOCK, lock_A, inner, inner_lock)		\
@@ -2640,10 +2627,6 @@ static void wait_context_tests(void)
 	DO_CONTEXT_TESTCASE_OUTER_LIMITED_PREEMPTIBLE(RCU_BH);
 	pr_cont("\n");
 
-	print_testname("in RCU callback context");
-	DO_CONTEXT_TESTCASE_OUTER_LIMITED_PREEMPTIBLE(RCU_CALLBACK);
-	pr_cont("\n");
-
 	print_testname("in RCU-sched context");
 	DO_CONTEXT_TESTCASE_OUTER_NOT_PREEMPTIBLE(RCU_SCHED);
 	pr_cont("\n");
Tested-by on the v1 series
#7 lore raw
In reply to Peter Zijlstra

Hi Peter,

Quoted 2 lines · Peter Zijlstra

Boris said, that Joerg said the locking selftests showed a few FAILED. They were right.

Thanks for looking into this. I tested the patches on my SEV-ES debug kernel and can confirm that the failures in the locking selftests went away, so