[PATCH v2 0/5] Improvements and fixes on SpinLockIrq patch series

[PATCH v2 0/5] Improvements and fixes on SpinLockIrq patch series

[Boqun Feng]

Hi Peter,

I realized that v1 [1] was left on the list so I just want to respin it.
Just to refresh the memory, v1 was requested [2] as separate patches for
changes between v4 and v5 in the main series. I also include the
refcount_interrupt kunit test Kconfig change in this version as well.
Based on v7.3-rc5.

Please take a look. Thanks!

[1]: https://lore.kernel.org/rust-for-linux/20260811041714.27679-1-boqun@kernel.org/
[2]: https://lore.kernel.org/rust-for-linux/20260810085255.GE1642880@noisy.programming.kicks-ass.net/

Regards,
Boqun

Boqun Feng (4):
  irq: Add {over,under}flow detection for
    local_interrupt_{enable,disable}
  selftests/bpf: Use the new NMI_BITS definition
  irq: Explain better on NMI_MASK overflow condition
  irq: Add max local_interrupt_disable() nesting level kunit test case

Kuan-Wei Chiu (1):
  irq: Make refcount_interrupt kunit test selectable

 include/linux/hardirq.h                       |  2 +-
 include/linux/interrupt_rc.h                  | 13 +++++++++++
 kernel/irq/Kconfig                            | 12 ++++++++++
 kernel/irq/Makefile                           |  2 +-
 kernel/irq/refcount_interrupt_test.c          | 22 +++++++++++++++++++
 .../testing/selftests/bpf/bpf_experimental.h  |  2 +-
 6 files changed, 50 insertions(+), 3 deletions(-)

-- 
2.51.0

[PATCH v2 1/5] irq: Make refcount_interrupt kunit test selectable

[Boqun Feng]

From: Kuan-Wei Chiu <visitorckw@gmail.com>

Currently, refcount_interrupt_test is built unconditionally when
CONFIG_KUNIT is enabled, causing it to run unexpectedly during boot.

Fix this by introducing CONFIG_REFCOUNT_INTERRUPT_KUNIT_TEST so the
test can be configured independently, following standard kunit
practices.

Fixes: 07a88e2bcd5b ("irq: Add KUnit test for refcounted interrupt enable/disable")
Signed-off-by: Kuan-Wei Chiu <visitorckw@gmail.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Reviewed-by: Radu Rendec <radu@rendec.net>
Reviewed-by: Boqun Feng <boqun@kernel.org>
---
 kernel/irq/Kconfig  | 12 ++++++++++++
 kernel/irq/Makefile |  2 +-
 2 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/kernel/irq/Kconfig b/kernel/irq/Kconfig
index 05cba4e16dad..6923f37eaab4 100644
--- a/kernel/irq/Kconfig
+++ b/kernel/irq/Kconfig
@@ -150,6 +150,18 @@ config IRQ_KUNIT_TEST
 
 	  If unsure, say N.
 
+config REFCOUNT_INTERRUPT_KUNIT_TEST
+	tristate "Test refcounted interrupt enable/disable" if !KUNIT_ALL_TESTS
+	depends on KUNIT
+	default KUNIT_ALL_TESTS
+	help
+	  This builds the kunit tests for the refcounted interrupt
+	  infrastructure. It verifies the correctness of single, nested,
+	  and multiple interrupt enable/disable state changes and ensures
+	  that the underlying reference counting mechanisms work as expected.
+
+	  If unsure, say N.
+
 endmenu
 
 config GENERIC_IRQ_MULTI_HANDLER
diff --git a/kernel/irq/Makefile b/kernel/irq/Makefile
index 44c4d6fc502a..0e5df962a149 100644
--- a/kernel/irq/Makefile
+++ b/kernel/irq/Makefile
@@ -16,4 +16,4 @@ obj-$(CONFIG_SMP) += affinity.o
 obj-$(CONFIG_GENERIC_IRQ_DEBUGFS) += debugfs.o
 obj-$(CONFIG_GENERIC_IRQ_MATRIX_ALLOCATOR) += matrix.o
 obj-$(CONFIG_IRQ_KUNIT_TEST) += irq_test.o
-obj-$(CONFIG_KUNIT) += refcount_interrupt_test.o
+obj-$(CONFIG_REFCOUNT_INTERRUPT_KUNIT_TEST) += refcount_interrupt_test.o
-- 
2.51.0

[PATCH v2 2/5] irq: Add {over,under}flow detection for local_interrupt_{enable,disable}

[Boqun Feng]

Currently there is no way to detect overflow and underflow inside the
HARDIRQ_DISABLE_MASK part of preempt_count. Add these detection guarded
by DEBUG_PREEMPT similar to the detection for PREEMPT_MASK
{over,under}flow.

Suggested-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Reviewed-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Signed-off-by: Boqun Feng <boqun@kernel.org>
---
 include/linux/interrupt_rc.h | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/include/linux/interrupt_rc.h b/include/linux/interrupt_rc.h
index a9ed937a80e7..d385ab05a2b6 100644
--- a/include/linux/interrupt_rc.h
+++ b/include/linux/interrupt_rc.h
@@ -11,6 +11,7 @@
  * make it easier to include without hitting circular header dependencies.
  */
 
+#include <linux/debug_locks.h>
 #include <linux/irqflags.h>
 #include <linux/preempt.h>
 #include <linux/processor.h>
@@ -65,6 +66,13 @@ static inline void local_interrupt_disable(void)
 	local_irq_save(flags);
 	new_count = hardirq_disable_enter();
 
+	/* Is hardirq disable count overflow soon? */
+	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT) &&
+	    DEBUG_LOCKS_WARN_ON(((new_count & HARDIRQ_DISABLE_MASK) +
+				 (10 << HARDIRQ_DISABLE_SHIFT)) >=
+				HARDIRQ_DISABLE_MASK))
+		return;
+
 	if ((new_count & HARDIRQ_DISABLE_MASK) == HARDIRQ_DISABLE_OFFSET)
 		_local_interrupt_save_state(flags);
 }
@@ -73,6 +81,11 @@ static inline void local_interrupt_enable(void)
 {
 	int new_count;
 
+	/* Unpaired local_interrupt_enable()? Warn and abort. */
+	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT) &&
+	    DEBUG_LOCKS_WARN_ON((preempt_count() & HARDIRQ_DISABLE_MASK) == 0))
+		return;
+
 	new_count = hardirq_disable_exit();
 
 	if ((new_count & HARDIRQ_DISABLE_MASK) == 0)
-- 
2.51.0

[PATCH v2 3/5] selftests/bpf: Use the new NMI_BITS definition

[Boqun Feng]

The definition of NMI_BITS was out of sync after the introduction of
HAS_SEPARATE_PREEMPT_RESCHED_BITS, hence adjust it.

Reported-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Reviewed-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Signed-off-by: Boqun Feng <boqun@kernel.org>
---
 tools/testing/selftests/bpf/bpf_experimental.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/testing/selftests/bpf/bpf_experimental.h b/tools/testing/selftests/bpf/bpf_experimental.h
index 2893bf06ff25..4bf7390994bf 100644
--- a/tools/testing/selftests/bpf/bpf_experimental.h
+++ b/tools/testing/selftests/bpf/bpf_experimental.h
@@ -371,7 +371,7 @@ extern int bpf_sock_read_xattr(struct socket *sock, const char *name__str,
 #define SOFTIRQ_BITS	8
 #define HARDIRQ_DISABLE_BITS	8
 #define HARDIRQ_BITS	4
-#define NMI_BITS	1
+#define NMI_BITS	(1 + 3*IS_ENABLED(CONFIG_HAS_SEPARATE_PREEMPT_RESCHED_BITS))
 
 #define PREEMPT_SHIFT	0
 #define SOFTIRQ_SHIFT	(PREEMPT_SHIFT + PREEMPT_BITS)
-- 
2.51.0

[PATCH v2 4/5] irq: Explain better on NMI_MASK overflow condition

[Boqun Feng]

Currently __preempt_count_nmi_enter() uses 15 as the maximum NMI nesting
level, and it's because 4 bits are used to represent the nesting level
of NMI. Replace the tautological comment with a more direct mention to
the 4 bits limitation to help readers understand this condition.

Suggested-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Reviewed-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Signed-off-by: Boqun Feng <boqun@kernel.org>
---
 include/linux/hardirq.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
index 73b48dd9f135..2648c49cd8f6 100644
--- a/include/linux/hardirq.h
+++ b/include/linux/hardirq.h
@@ -108,7 +108,7 @@ DECLARE_PER_CPU(unsigned int, nmi_nesting);
 #define __preempt_count_nmi_enter()				\
 	do {							\
 		__preempt_count_add(HARDIRQ_OFFSET);		\
-		/* Maximum NMI nesting is 15. */		\
+		/* NMI nesting is represented in 4 bits. */	\
 		BUG_ON(__this_cpu_read(nmi_nesting) >= 15);	\
 		__this_cpu_inc(nmi_nesting);			\
 		preempt_count_set(preempt_count() | NMI_MASK);  \
-- 
2.51.0

[PATCH v2 5/5] irq: Add max local_interrupt_disable() nesting level kunit test case

[Boqun Feng]

To confirm the max nesting level of local_interrupt_disable() works, a
kunit test is added to the whole test suite.

Note that when DEBUG_PREEMPT=y, it'll generate a warning which is
expected.

Suggested-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Reviewed-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Signed-off-by: Boqun Feng <boqun@kernel.org>
---
 kernel/irq/refcount_interrupt_test.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/kernel/irq/refcount_interrupt_test.c b/kernel/irq/refcount_interrupt_test.c
index ca904dba24b9..38dfccbaa4d4 100644
--- a/kernel/irq/refcount_interrupt_test.c
+++ b/kernel/irq/refcount_interrupt_test.c
@@ -52,6 +52,27 @@ static void test_multiple_irq_change(struct kunit *test)
 	TEST_IRQ_ON();
 }
 
+static void test_max_nesting_irq_change(struct kunit *test)
+{
+	for (int i = 0; i < __IRQ_MASK(HARDIRQ_DISABLE_BITS); i++) {
+		local_interrupt_disable();
+		TEST_IRQ_OFF();
+	}
+
+
+	for (int i = 0; i < __IRQ_MASK(HARDIRQ_DISABLE_BITS); i++) {
+		TEST_IRQ_OFF();
+		local_interrupt_enable();
+	}
+
+	TEST_IRQ_ON();
+
+	local_interrupt_disable();
+	TEST_IRQ_OFF();
+	local_interrupt_enable();
+	TEST_IRQ_ON();
+}
+
 static void test_irq_save(struct kunit *test)
 {
 	unsigned long flags;
@@ -79,6 +100,7 @@ static struct kunit_case test_cases[] = {
 	KUNIT_CASE(test_single_irq_change),
 	KUNIT_CASE(test_nested_irq_change),
 	KUNIT_CASE(test_multiple_irq_change),
+	KUNIT_CASE(test_max_nesting_irq_change),
 	KUNIT_CASE(test_irq_save),
 	{},
 };
-- 
2.51.0

Re: [PATCH v2 3/5] selftests/bpf: Use the new NMI_BITS definition

[Boqun Feng]

On Thu, Oct 01, 2026 at 10:25:17AM +0200, Boqun Feng wrote:
> The definition of NMI_BITS was out of sync after the introduction of
> HAS_SEPARATE_PREEMPT_RESCHED_BITS, hence adjust it.
> 
> Reported-by: Shrikanth Hegde <sshegde@linux.ibm.com>
> Reviewed-by: Shrikanth Hegde <sshegde@linux.ibm.com>
> Signed-off-by: Boqun Feng <boqun@kernel.org>
> ---
>  tools/testing/selftests/bpf/bpf_experimental.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/bpf/bpf_experimental.h b/tools/testing/selftests/bpf/bpf_experimental.h
> index 2893bf06ff25..4bf7390994bf 100644
> --- a/tools/testing/selftests/bpf/bpf_experimental.h
> +++ b/tools/testing/selftests/bpf/bpf_experimental.h

Per sashiko's report, seems I need the following to make it work:

diff --git a/tools/testing/selftests/bpf/bpf_experimental.h b/tools/testing/selftests/bpf/bpf_experimental.h
index 4bf7390994bf..94143cf0ab0d 100644
--- a/tools/testing/selftests/bpf/bpf_experimental.h
+++ b/tools/testing/selftests/bpf/bpf_experimental.h
@@ -367,6 +367,9 @@ extern int bpf_cgroup_read_xattr(struct cgroup *cgroup, const char *name__str,
 extern int bpf_sock_read_xattr(struct socket *sock, const char *name__str,
                               struct bpf_dynptr *value_p) __weak __ksym;

+extern bool CONFIG_HAS_SEPARATE_PREEMPT_RESCHED_BITS __kconfig __weak;
+#define IS_ENABLED(cfg) (cfg)
+
 #define PREEMPT_BITS   8
 #define SOFTIRQ_BITS   8
 #define HARDIRQ_DISABLE_BITS   8

?

[Cc Leon and Alexi]

Regards,
Boqun

> @@ -371,7 +371,7 @@ extern int bpf_sock_read_xattr(struct socket *sock, const char *name__str,
>  #define SOFTIRQ_BITS	8
>  #define HARDIRQ_DISABLE_BITS	8
>  #define HARDIRQ_BITS	4
> -#define NMI_BITS	1
> +#define NMI_BITS	(1 + 3*IS_ENABLED(CONFIG_HAS_SEPARATE_PREEMPT_RESCHED_BITS))
>  
>  #define PREEMPT_SHIFT	0
>  #define SOFTIRQ_SHIFT	(PREEMPT_SHIFT + PREEMPT_BITS)
> -- 
> 2.51.0
>