[PATCH v2 0/9] panic: introduce panic status function family

[PATCH v2 0/9] panic: introduce panic status function family

[Jinchao Wang]

This series introduces a family of helper functions to manage panic state and
updates existing code to use them.

Before this series, panic state helpers were scattered and inconsistent. For
example, panic_in_progress() was defined in printk/printk.c, not in panic.c or
panic.h. As a result, developers had to look in unexpected places to understand
or re-use panic state logic. Other checks were open- coded, duplicating logic
across panic, crash, and watchdog paths.

The new helpers centralize the functionality in panic.c/panic.h:
  - panic_try_start()
  - panic_reset()
  - panic_in_progress()
  - panic_on_this_cpu()
  - panic_on_other_cpu()

Patches 1–8 add the helpers and convert panic/crash and printk/nbcon code to
use them.

Patch 9 fixes a bug in the watchdog subsystem by skipping checks when a panic
is in progress, avoiding interference with the panic CPU.

Together, this makes panic state handling simpler, more discoverable, and more
robust.

-- 

Changes from v1:

  - The commit messages have been re-formatted based on Baoquan He's suggestions.
  - The panic_in_progress() check has been moved to be the first check in
    watchdog_overflow_callback(), as suggested by Yury Norov.

Link to v1:
https://lore.kernel.org/all/20250820091702.512524-1-wangjinchao600@gmail.com/

Jinchao Wang (9):
  panic: Introduce helper functions for panic state
  fbdev: Use panic_in_progress() helper
  crash_core: use panic_try_start() in crash_kexec()
  panic: use panic_try_start() in nmi_panic()
  panic: use panic_try_start() in vpanic()
  printk/nbcon: use panic_on_this_cpu() helper
  panic/printk: replace this_cpu_in_panic() with panic_on_this_cpu()
  panic/printk: replace other_cpu_in_panic() with panic_on_other_cpu()
  watchdog: skip checks when panic is in progress

 drivers/video/fbdev/core/fbcon.c  |  9 +---
 include/linux/panic.h             |  6 +++
 include/linux/printk.h            |  2 -
 kernel/crash_core.c               | 15 ++-----
 kernel/panic.c                    | 71 +++++++++++++++++++++++++------
 kernel/printk/internal.h          |  1 -
 kernel/printk/nbcon.c             | 14 +++---
 kernel/printk/printk.c            | 37 +++-------------
 kernel/printk/printk_ringbuffer.c |  2 +-
 kernel/watchdog.c                 |  6 +++
 kernel/watchdog_perf.c            |  4 ++
 lib/dump_stack.c                  |  2 +-
 12 files changed, 91 insertions(+), 78 deletions(-)

-- 
2.43.0

[PATCH v2 1/9] panic: Introduce helper functions for panic state

[Jinchao Wang]

This patch introduces four new helper functions to abstract the
management of the panic_cpu variable. These functions will be used in
subsequent patches to refactor existing code.

The direct use of panic_cpu can be error-prone and ambiguous, as it
requires manual checks to determine which CPU is handling the panic.
The new helpers clarify intent:

panic_try_start():
Atomically sets the current CPU as the panicking CPU.

panic_reset():
Reset panic_cpu to PANIC_CPU_INVALID.

panic_in_progress():
Checks if a panic has been triggered.

panic_on_this_cpu():
Returns true if the current CPU is the panic originator.

panic_on_other_cpu():
Returns true if a panic is on another CPU.

This change lays the groundwork for improved code readability
and robustness in the panic handling subsystem.

Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
---
 include/linux/panic.h  |  6 +++++
 kernel/panic.c         | 53 ++++++++++++++++++++++++++++++++++++++++++
 kernel/printk/printk.c |  5 ----
 3 files changed, 59 insertions(+), 5 deletions(-)

diff --git a/include/linux/panic.h b/include/linux/panic.h
index 7be742628c25..6f972a66c13e 100644
--- a/include/linux/panic.h
+++ b/include/linux/panic.h
@@ -43,6 +43,12 @@ void abort(void);
 extern atomic_t panic_cpu;
 #define PANIC_CPU_INVALID	-1
 
+bool panic_try_start(void);
+void panic_reset(void);
+bool panic_in_progress(void);
+bool panic_on_this_cpu(void);
+bool panic_on_other_cpu(void);
+
 /*
  * Only to be used by arch init code. If the user over-wrote the default
  * CONFIG_PANIC_TIMEOUT, honor it.
diff --git a/kernel/panic.c b/kernel/panic.c
index 72fcbb5a071b..eacb0c972110 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -294,6 +294,59 @@ void __weak crash_smp_send_stop(void)
 
 atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
 
+bool panic_try_start(void)
+{
+	int old_cpu, this_cpu;
+
+	/*
+	 * Only one CPU is allowed to execute the crash_kexec() code as with
+	 * panic().  Otherwise parallel calls of panic() and crash_kexec()
+	 * may stop each other.  To exclude them, we use panic_cpu here too.
+	 */
+	old_cpu = PANIC_CPU_INVALID;
+	this_cpu = raw_smp_processor_id();
+
+	return atomic_try_cmpxchg(&panic_cpu, &old_cpu, this_cpu);
+}
+EXPORT_SYMBOL(panic_try_start);
+
+void panic_reset(void)
+{
+	atomic_set(&panic_cpu, PANIC_CPU_INVALID);
+}
+EXPORT_SYMBOL(panic_reset);
+
+bool panic_in_progress(void)
+{
+	return unlikely(atomic_read(&panic_cpu) != PANIC_CPU_INVALID);
+}
+EXPORT_SYMBOL(panic_in_progress);
+
+/* Return true if a panic is in progress on the current CPU. */
+bool panic_on_this_cpu(void)
+{
+	/*
+	 * We can use raw_smp_processor_id() here because it is impossible for
+	 * the task to be migrated to the panic_cpu, or away from it. If
+	 * panic_cpu has already been set, and we're not currently executing on
+	 * that CPU, then we never will be.
+	 */
+	return unlikely(atomic_read(&panic_cpu) == raw_smp_processor_id());
+}
+EXPORT_SYMBOL(panic_on_this_cpu);
+
+/*
+ * Return true if a panic is in progress on a remote CPU.
+ *
+ * On true, the local CPU should immediately release any printing resources
+ * that may be needed by the panic CPU.
+ */
+bool panic_on_other_cpu(void)
+{
+	return (panic_in_progress() && !this_cpu_in_panic());
+}
+EXPORT_SYMBOL(panic_on_other_cpu);
+
 /*
  * A variant of panic() called from NMI context. We return if we've already
  * panicked on this CPU. If another CPU already panicked, loop in
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 0efbcdda9aab..5fe35f377b79 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -345,11 +345,6 @@ static void __up_console_sem(unsigned long ip)
 }
 #define up_console_sem() __up_console_sem(_RET_IP_)
 
-static bool panic_in_progress(void)
-{
-	return unlikely(atomic_read(&panic_cpu) != PANIC_CPU_INVALID);
-}
-
 /* Return true if a panic is in progress on the current CPU. */
 bool this_cpu_in_panic(void)
 {
-- 
2.43.0

[PATCH v2 2/9] fbdev: Use panic_in_progress() helper

[Jinchao Wang]

This patch updates the fbcon_skip_panic() function to use the
panic_in_progress() helper.

The previous direct access to panic_cpu is less readable and is being
replaced by a dedicated function that more clearly expresses the intent.

This change is part of a series to refactor the kernel's panic handling
logic for better clarity and robustness.

Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
Acked-by Qianqiang Liu <qianqiang.liu@163.com>
---
 drivers/video/fbdev/core/fbcon.c | 9 +--------
 1 file changed, 1 insertion(+), 8 deletions(-)

diff --git a/drivers/video/fbdev/core/fbcon.c b/drivers/video/fbdev/core/fbcon.c
index 55f5731e94c3..b062b05f4128 100644
--- a/drivers/video/fbdev/core/fbcon.c
+++ b/drivers/video/fbdev/core/fbcon.c
@@ -279,14 +279,7 @@ static int fbcon_get_rotate(struct fb_info *info)
 
 static bool fbcon_skip_panic(struct fb_info *info)
 {
-/* panic_cpu is not exported, and can't be used if built as module. Use
- * oops_in_progress instead, but non-fatal oops won't be printed.
- */
-#if defined(MODULE)
-	return (info->skip_panic && unlikely(oops_in_progress));
-#else
-	return (info->skip_panic && unlikely(atomic_read(&panic_cpu) != PANIC_CPU_INVALID));
-#endif
+	return (info->skip_panic && unlikely(panic_in_progress()));
 }
 
 static inline bool fbcon_is_active(struct vc_data *vc, struct fb_info *info)
-- 
2.43.0

[PATCH v2 3/9] crash_core: use panic_try_start() in crash_kexec()

[Jinchao Wang]

crash_kexec() had its own code to exclude parallel execution by setting
panic_cpu. This is already handled by panic_try_start(). Switch to
panic_try_start() to remove the duplication and keep the logic consistent.

Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
---
 kernel/crash_core.c | 15 +++------------
 1 file changed, 3 insertions(+), 12 deletions(-)

diff --git a/kernel/crash_core.c b/kernel/crash_core.c
index a4ef79591eb2..bb38bbaf3a26 100644
--- a/kernel/crash_core.c
+++ b/kernel/crash_core.c
@@ -4,6 +4,7 @@
  * Copyright (C) 2002-2004 Eric Biederman  <ebiederm@xmission.com>
  */
 
+#include "linux/panic.h"
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
 #include <linux/buildid.h>
@@ -143,17 +144,7 @@ STACK_FRAME_NON_STANDARD(__crash_kexec);
 
 __bpf_kfunc void crash_kexec(struct pt_regs *regs)
 {
-	int old_cpu, this_cpu;
-
-	/*
-	 * Only one CPU is allowed to execute the crash_kexec() code as with
-	 * panic().  Otherwise parallel calls of panic() and crash_kexec()
-	 * may stop each other.  To exclude them, we use panic_cpu here too.
-	 */
-	old_cpu = PANIC_CPU_INVALID;
-	this_cpu = raw_smp_processor_id();
-
-	if (atomic_try_cmpxchg(&panic_cpu, &old_cpu, this_cpu)) {
+	if (panic_try_start()) {
 		/* This is the 1st CPU which comes here, so go ahead. */
 		__crash_kexec(regs);
 
@@ -161,7 +152,7 @@ __bpf_kfunc void crash_kexec(struct pt_regs *regs)
 		 * Reset panic_cpu to allow another panic()/crash_kexec()
 		 * call.
 		 */
-		atomic_set(&panic_cpu, PANIC_CPU_INVALID);
+		panic_reset();
 	}
 }
 
-- 
2.43.0

Re: [PATCH v2 3/9] crash_core: use panic_try_start() in crash_kexec()

[Qianqiang Liu]

On Mon, Aug 25, 2025 at 10:29:31AM +0800, Jinchao Wang wrote:
> crash_kexec() had its own code to exclude parallel execution by setting
> panic_cpu. This is already handled by panic_try_start(). Switch to
> panic_try_start() to remove the duplication and keep the logic consistent.
> 
> Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
> ---
>  kernel/crash_core.c | 15 +++------------
>  1 file changed, 3 insertions(+), 12 deletions(-)
> 
> diff --git a/kernel/crash_core.c b/kernel/crash_core.c
> index a4ef79591eb2..bb38bbaf3a26 100644
> --- a/kernel/crash_core.c
> +++ b/kernel/crash_core.c
> @@ -4,6 +4,7 @@
>   * Copyright (C) 2002-2004 Eric Biederman  <ebiederm@xmission.com>
>   */
>  
> +#include "linux/panic.h"

Can you change this line of code to #include <linux/panic.h>?

>  #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>  
>  #include <linux/buildid.h>
> @@ -143,17 +144,7 @@ STACK_FRAME_NON_STANDARD(__crash_kexec);
>  
>  __bpf_kfunc void crash_kexec(struct pt_regs *regs)
>  {
> -	int old_cpu, this_cpu;
> -
> -	/*
> -	 * Only one CPU is allowed to execute the crash_kexec() code as with
> -	 * panic().  Otherwise parallel calls of panic() and crash_kexec()
> -	 * may stop each other.  To exclude them, we use panic_cpu here too.
> -	 */
> -	old_cpu = PANIC_CPU_INVALID;
> -	this_cpu = raw_smp_processor_id();
> -
> -	if (atomic_try_cmpxchg(&panic_cpu, &old_cpu, this_cpu)) {
> +	if (panic_try_start()) {
>  		/* This is the 1st CPU which comes here, so go ahead. */
>  		__crash_kexec(regs);
>  
> @@ -161,7 +152,7 @@ __bpf_kfunc void crash_kexec(struct pt_regs *regs)
>  		 * Reset panic_cpu to allow another panic()/crash_kexec()
>  		 * call.
>  		 */
> -		atomic_set(&panic_cpu, PANIC_CPU_INVALID);
> +		panic_reset();
>  	}
>  }
>  
> -- 
> 2.43.0

-- 
Best,
Qianqiang Liu

Re: [PATCH v2 3/9] crash_core: use panic_try_start() in crash_kexec()

[Jinchao Wang]

On 8/29/25 10:30, Qianqiang Liu wrote:
> On Mon, Aug 25, 2025 at 10:29:31AM +0800, Jinchao Wang wrote:
>> crash_kexec() had its own code to exclude parallel execution by setting
>> panic_cpu. This is already handled by panic_try_start(). Switch to
>> panic_try_start() to remove the duplication and keep the logic consistent.
>>
>> Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
>> ---
>>   kernel/crash_core.c | 15 +++------------
>>   1 file changed, 3 insertions(+), 12 deletions(-)
>>
>> diff --git a/kernel/crash_core.c b/kernel/crash_core.c
>> index a4ef79591eb2..bb38bbaf3a26 100644
>> --- a/kernel/crash_core.c
>> +++ b/kernel/crash_core.c
>> @@ -4,6 +4,7 @@
>>    * Copyright (C) 2002-2004 Eric Biederman  <ebiederm@xmission.com>
>>    */
>>   
>> +#include "linux/panic.h"
> 
> Can you change this line of code to #include <linux/panic.h>?
Yes, I will.
Thanks for the feedback.>
>>   #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>>   
>>   #include <linux/buildid.h>
>> @@ -143,17 +144,7 @@ STACK_FRAME_NON_STANDARD(__crash_kexec);
>>   
>>   __bpf_kfunc void crash_kexec(struct pt_regs *regs)
>>   {
>> -	int old_cpu, this_cpu;
>> -
>> -	/*
>> -	 * Only one CPU is allowed to execute the crash_kexec() code as with
>> -	 * panic().  Otherwise parallel calls of panic() and crash_kexec()
>> -	 * may stop each other.  To exclude them, we use panic_cpu here too.
>> -	 */
>> -	old_cpu = PANIC_CPU_INVALID;
>> -	this_cpu = raw_smp_processor_id();
>> -
>> -	if (atomic_try_cmpxchg(&panic_cpu, &old_cpu, this_cpu)) {
>> +	if (panic_try_start()) {
>>   		/* This is the 1st CPU which comes here, so go ahead. */
>>   		__crash_kexec(regs);
>>   
>> @@ -161,7 +152,7 @@ __bpf_kfunc void crash_kexec(struct pt_regs *regs)
>>   		 * Reset panic_cpu to allow another panic()/crash_kexec()
>>   		 * call.
>>   		 */
>> -		atomic_set(&panic_cpu, PANIC_CPU_INVALID);
>> +		panic_reset();
>>   	}
>>   }
>>   
>> -- 
>> 2.43.0
> 


-- 
Best regards,
Jinchao

[PATCH v2 4/9] panic: use panic_try_start() in nmi_panic()

[Jinchao Wang]

nmi_panic() duplicated the logic to claim panic_cpu with atomic_try_cmpxchg.
This is already wrapped in panic_try_start().

Replace the open-coded logic with panic_try_start(), and use
panic_on_other_cpu() for the fallback path.

This removes duplication and keeps panic handling code consistent.

Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
---
 kernel/panic.c | 10 ++--------
 1 file changed, 2 insertions(+), 8 deletions(-)

diff --git a/kernel/panic.c b/kernel/panic.c
index eacb0c972110..cd86d37d124c 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -355,15 +355,9 @@ EXPORT_SYMBOL(panic_on_other_cpu);
  */
 void nmi_panic(struct pt_regs *regs, const char *msg)
 {
-	int old_cpu, this_cpu;
-
-	old_cpu = PANIC_CPU_INVALID;
-	this_cpu = raw_smp_processor_id();
-
-	/* atomic_try_cmpxchg updates old_cpu on failure */
-	if (atomic_try_cmpxchg(&panic_cpu, &old_cpu, this_cpu))
+	if (panic_try_start())
 		panic("%s", msg);
-	else if (old_cpu != this_cpu)
+	else if (panic_on_other_cpu())
 		nmi_panic_self_stop(regs);
 }
 EXPORT_SYMBOL(nmi_panic);
-- 
2.43.0

[PATCH v2 5/9] panic: use panic_try_start() in vpanic()

[Jinchao Wang]

vpanic() had open-coded logic to claim panic_cpu with atomic_try_cmpxchg.
This is already handled by panic_try_start().

Switch to panic_try_start() and use panic_on_other_cpu() for the fallback
path.

This removes duplicate code and makes panic handling consistent across
functions.

Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
---
 kernel/panic.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/kernel/panic.c b/kernel/panic.c
index cd86d37d124c..5266e195f5ac 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -415,7 +415,6 @@ void vpanic(const char *fmt, va_list args)
 	static char buf[1024];
 	long i, i_next = 0, len;
 	int state = 0;
-	int old_cpu, this_cpu;
 	bool _crash_kexec_post_notifiers = crash_kexec_post_notifiers;
 
 	if (panic_on_warn) {
@@ -452,13 +451,10 @@ void vpanic(const char *fmt, va_list args)
 	 * `old_cpu == this_cpu' means we came from nmi_panic() which sets
 	 * panic_cpu to this CPU.  In this case, this is also the 1st CPU.
 	 */
-	old_cpu = PANIC_CPU_INVALID;
-	this_cpu = raw_smp_processor_id();
-
 	/* atomic_try_cmpxchg updates old_cpu on failure */
-	if (atomic_try_cmpxchg(&panic_cpu, &old_cpu, this_cpu)) {
+	if (panic_try_start()) {
 		/* go ahead */
-	} else if (old_cpu != this_cpu)
+	} else if (panic_on_other_cpu())
 		panic_smp_self_stop();
 
 	console_verbose();
-- 
2.43.0

[PATCH v2 6/9] printk/nbcon: use panic_on_this_cpu() helper

[Jinchao Wang]

nbcon_context_try_acquire() compared panic_cpu directly with
smp_processor_id().  This open-coded check is now provided by
panic_on_this_cpu().

Switch to panic_on_this_cpu() to simplify the code and improve readability.

Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
---
 kernel/printk/nbcon.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 646801813415..7490865e2f44 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -2,6 +2,7 @@
 // Copyright (C) 2022 Linutronix GmbH, John Ogness
 // Copyright (C) 2022 Intel, Thomas Gleixner
 
+#include "linux/panic.h"
 #include <linux/atomic.h>
 #include <linux/bug.h>
 #include <linux/console.h>
@@ -589,7 +590,6 @@ static struct printk_buffers panic_nbcon_pbufs;
  */
 static bool nbcon_context_try_acquire(struct nbcon_context *ctxt, bool is_reacquire)
 {
-	unsigned int cpu = smp_processor_id();
 	struct console *con = ctxt->console;
 	struct nbcon_state cur;
 	int err;
@@ -614,7 +614,7 @@ static bool nbcon_context_try_acquire(struct nbcon_context *ctxt, bool is_reacqu
 	/* Acquire succeeded. */
 
 	/* Assign the appropriate buffer for this context. */
-	if (atomic_read(&panic_cpu) == cpu)
+	if (panic_on_this_cpu())
 		ctxt->pbufs = &panic_nbcon_pbufs;
 	else
 		ctxt->pbufs = con->pbufs;
-- 
2.43.0

[PATCH v2 7/9] panic/printk: replace this_cpu_in_panic() with panic_on_this_cpu()

[Jinchao Wang]

The helper this_cpu_in_panic() duplicated logic already provided by
panic_on_this_cpu().

Remove this_cpu_in_panic() and switch all users to panic_on_this_cpu().

This simplifies the code and avoids having two helpers for the same check.

Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
---
 include/linux/printk.h            |  2 --
 kernel/panic.c                    |  2 +-
 kernel/printk/nbcon.c             |  2 +-
 kernel/printk/printk.c            | 15 ++-------------
 kernel/printk/printk_ringbuffer.c |  2 +-
 lib/dump_stack.c                  |  2 +-
 6 files changed, 6 insertions(+), 19 deletions(-)

diff --git a/include/linux/printk.h b/include/linux/printk.h
index 5d22b803f51e..45c663124c9b 100644
--- a/include/linux/printk.h
+++ b/include/linux/printk.h
@@ -330,8 +330,6 @@ static inline bool pr_flush(int timeout_ms, bool reset_on_progress)
 
 #endif
 
-bool this_cpu_in_panic(void);
-
 #ifdef CONFIG_SMP
 extern int __printk_cpu_sync_try_get(void);
 extern void __printk_cpu_sync_wait(void);
diff --git a/kernel/panic.c b/kernel/panic.c
index 5266e195f5ac..8415e4073da1 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -343,7 +343,7 @@ EXPORT_SYMBOL(panic_on_this_cpu);
  */
 bool panic_on_other_cpu(void)
 {
-	return (panic_in_progress() && !this_cpu_in_panic());
+	return (panic_in_progress() && !panic_on_this_cpu());
 }
 EXPORT_SYMBOL(panic_on_other_cpu);
 
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 7490865e2f44..c6d1a4a747e9 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -1394,7 +1394,7 @@ enum nbcon_prio nbcon_get_default_prio(void)
 {
 	unsigned int *cpu_emergency_nesting;
 
-	if (this_cpu_in_panic())
+	if (panic_on_this_cpu())
 		return NBCON_PRIO_PANIC;
 
 	cpu_emergency_nesting = nbcon_get_cpu_emergency_nesting();
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 5fe35f377b79..faa8b1f0585b 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -17,6 +17,7 @@
  *	01Mar01 Andrew Morton
  */
 
+#include "linux/panic.h"
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
 #include <linux/kernel.h>
@@ -345,18 +346,6 @@ static void __up_console_sem(unsigned long ip)
 }
 #define up_console_sem() __up_console_sem(_RET_IP_)
 
-/* Return true if a panic is in progress on the current CPU. */
-bool this_cpu_in_panic(void)
-{
-	/*
-	 * We can use raw_smp_processor_id() here because it is impossible for
-	 * the task to be migrated to the panic_cpu, or away from it. If
-	 * panic_cpu has already been set, and we're not currently executing on
-	 * that CPU, then we never will be.
-	 */
-	return unlikely(atomic_read(&panic_cpu) == raw_smp_processor_id());
-}
-
 /*
  * Return true if a panic is in progress on a remote CPU.
  *
@@ -365,7 +354,7 @@ bool this_cpu_in_panic(void)
  */
 bool other_cpu_in_panic(void)
 {
-	return (panic_in_progress() && !this_cpu_in_panic());
+	return (panic_in_progress() && !panic_on_this_cpu());
 }
 
 /*
diff --git a/kernel/printk/printk_ringbuffer.c b/kernel/printk/printk_ringbuffer.c
index d9fb053cff67..e2a1b2d34d2b 100644
--- a/kernel/printk/printk_ringbuffer.c
+++ b/kernel/printk/printk_ringbuffer.c
@@ -2143,7 +2143,7 @@ static bool _prb_read_valid(struct printk_ringbuffer *rb, u64 *seq,
 			 * But it would have the sequence number returned
 			 * by "prb_next_reserve_seq() - 1".
 			 */
-			if (this_cpu_in_panic() &&
+			if (panic_on_this_cpu() &&
 			    (!debug_non_panic_cpus || legacy_allow_panic_sync) &&
 			    ((*seq + 1) < prb_next_reserve_seq(rb))) {
 				(*seq)++;
diff --git a/lib/dump_stack.c b/lib/dump_stack.c
index b3a85fe8b673..f0c78b5b5324 100644
--- a/lib/dump_stack.c
+++ b/lib/dump_stack.c
@@ -102,7 +102,7 @@ static void __dump_stack(const char *log_lvl)
  */
 asmlinkage __visible void dump_stack_lvl(const char *log_lvl)
 {
-	bool in_panic = this_cpu_in_panic();
+	bool in_panic = panic_on_this_cpu();
 	unsigned long flags;
 
 	/*
-- 
2.43.0

[PATCH v2 8/9] panic/printk: replace other_cpu_in_panic() with panic_on_other_cpu()

[Jinchao Wang]

The helper other_cpu_in_panic() duplicated logic already provided by
panic_on_other_cpu().

Remove other_cpu_in_panic() and update all users to call
panic_on_other_cpu() instead.

This removes redundant code and makes panic handling consistent.

Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
---
 kernel/printk/internal.h |  1 -
 kernel/printk/nbcon.c    |  8 ++++----
 kernel/printk/printk.c   | 19 ++++---------------
 3 files changed, 8 insertions(+), 20 deletions(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index ef282001f200..f72bbfa266d6 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -332,7 +332,6 @@ struct printk_message {
 	unsigned long		dropped;
 };
 
-bool other_cpu_in_panic(void);
 bool printk_get_next_message(struct printk_message *pmsg, u64 seq,
 			     bool is_extended, bool may_supress);
 
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index c6d1a4a747e9..171480135830 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -255,7 +255,7 @@ static int nbcon_context_try_acquire_direct(struct nbcon_context *ctxt,
 		 * opportunity to perform any necessary cleanup if they were
 		 * interrupted by the panic CPU while printing.
 		 */
-		if (other_cpu_in_panic() &&
+		if (panic_on_other_cpu() &&
 		    (!is_reacquire || cur->unsafe_takeover)) {
 			return -EPERM;
 		}
@@ -310,7 +310,7 @@ static bool nbcon_waiter_matches(struct nbcon_state *cur, int expected_prio)
 	 * Event #2 implies the new context is PANIC.
 	 * Event #3 occurs when panic() has flushed the console.
 	 * Event #4 occurs when a non-panic CPU reacquires.
-	 * Event #5 is not possible due to the other_cpu_in_panic() check
+	 * Event #5 is not possible due to the panic_on_other_cpu() check
 	 *          in nbcon_context_try_acquire_handover().
 	 */
 
@@ -349,7 +349,7 @@ static int nbcon_context_try_acquire_requested(struct nbcon_context *ctxt,
 	struct nbcon_state new;
 
 	/* Note that the caller must still remove the request! */
-	if (other_cpu_in_panic())
+	if (panic_on_other_cpu())
 		return -EPERM;
 
 	/*
@@ -447,7 +447,7 @@ static int nbcon_context_try_acquire_handover(struct nbcon_context *ctxt,
 	 * nbcon_waiter_matches(). In particular, the assumption that
 	 * lower priorities are ignored during panic.
 	 */
-	if (other_cpu_in_panic())
+	if (panic_on_other_cpu())
 		return -EPERM;
 
 	/* Handover is not possible on the same CPU. */
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index faa8b1f0585b..236f03937107 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -346,17 +346,6 @@ static void __up_console_sem(unsigned long ip)
 }
 #define up_console_sem() __up_console_sem(_RET_IP_)
 
-/*
- * Return true if a panic is in progress on a remote CPU.
- *
- * On true, the local CPU should immediately release any printing resources
- * that may be needed by the panic CPU.
- */
-bool other_cpu_in_panic(void)
-{
-	return (panic_in_progress() && !panic_on_this_cpu());
-}
-
 /*
  * This is used for debugging the mess that is the VT code by
  * keeping track if we have the console semaphore held. It's
@@ -2391,7 +2380,7 @@ asmlinkage int vprintk_emit(int facility, int level,
 	 * non-panic CPUs are generating any messages, they will be
 	 * silently dropped.
 	 */
-	if (other_cpu_in_panic() &&
+	if (panic_on_other_cpu() &&
 	    !debug_non_panic_cpus &&
 	    !panic_triggering_all_cpu_backtrace)
 		return 0;
@@ -2827,7 +2816,7 @@ void console_lock(void)
 	might_sleep();
 
 	/* On panic, the console_lock must be left to the panic cpu. */
-	while (other_cpu_in_panic())
+	while (panic_on_other_cpu())
 		msleep(1000);
 
 	down_console_sem();
@@ -2847,7 +2836,7 @@ EXPORT_SYMBOL(console_lock);
 int console_trylock(void)
 {
 	/* On panic, the console_lock must be left to the panic cpu. */
-	if (other_cpu_in_panic())
+	if (panic_on_other_cpu())
 		return 0;
 	if (down_trylock_console_sem())
 		return 0;
@@ -3227,7 +3216,7 @@ static bool console_flush_all(bool do_cond_resched, u64 *next_seq, bool *handove
 			any_progress = true;
 
 			/* Allow panic_cpu to take over the consoles safely. */
-			if (other_cpu_in_panic())
+			if (panic_on_other_cpu())
 				goto abandon;
 
 			if (do_cond_resched)
-- 
2.43.0

[PATCH v2 9/9] watchdog: skip checks when panic is in progress

[Jinchao Wang]

This issue was found when an EFI pstore was configured for kdump
logging with the NMI hard lockup detector enabled. The efi-pstore
write operation was slow, and with a large number of logs, the
pstore dump callback within kmsg_dump() took a long time.

This delay triggered the NMI watchdog, leading to a nested panic.
The call flow demonstrates how the secondary panic caused an
emergency_restart() to be triggered before the initial pstore
operation could finish, leading to a failure to dump the logs:

  real panic() {
	kmsg_dump() {
		...
		pstore_dump() {
			start_dump();
			... // long time operation triggers NMI watchdog
			nmi panic() {
				...
				emergency_restart(); // pstore unfinished
			}
			...
			finish_dump(); // never reached
		}
	}
  }

Both watchdog_buddy_check_hardlockup() and watchdog_overflow_callback() may
trigger during a panic. This can lead to recursive panic handling.

Add panic_in_progress() checks so watchdog activity is skipped once a panic
has begun.

This prevents recursive panic and keeps the panic path more reliable.

Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
Reviewed-by: Yury Norov (NVIDIA) <yury.norov@gmail.com>
---
 kernel/watchdog.c      | 6 ++++++
 kernel/watchdog_perf.c | 4 ++++
 2 files changed, 10 insertions(+)

diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index 80b56c002c7f..597c0d947c93 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -740,6 +740,12 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
 	if (!watchdog_enabled)
 		return HRTIMER_NORESTART;
 
+	/*
+	 * pass the buddy check if a panic is in process
+	 */
+	if (panic_in_progress())
+		return HRTIMER_NORESTART;
+
 	watchdog_hardlockup_kick();
 
 	/* kick the softlockup detector */
diff --git a/kernel/watchdog_perf.c b/kernel/watchdog_perf.c
index 9c58f5b4381d..d3ca70e3c256 100644
--- a/kernel/watchdog_perf.c
+++ b/kernel/watchdog_perf.c
@@ -12,6 +12,7 @@
 
 #define pr_fmt(fmt) "NMI watchdog: " fmt
 
+#include <linux/panic.h>
 #include <linux/nmi.h>
 #include <linux/atomic.h>
 #include <linux/module.h>
@@ -108,6 +109,9 @@ static void watchdog_overflow_callback(struct perf_event *event,
 	/* Ensure the watchdog never gets throttled */
 	event->hw.interrupts = 0;
 
+	if (panic_in_progress())
+		return;
+
 	if (!watchdog_check_timestamp())
 		return;
 
-- 
2.43.0