[PATCH] watchdog: Force console verbose when reporting lockups

[PATCH] watchdog: Force console verbose when reporting lockups

[Tzung-Bi Shih]

When a hard or soft lockup is detected, the watchdog prints a header at
KERN_EMERG (<0>), followed by diagnostic details (modules, registers,
and backtraces) at KERN_WARNING/KERN_DEFAULT (<4>) and KERN_INFO (<6>).
If console_loglevel is quiet (<= 4), the console skips those <4>/<6>
diagnostic messages before vpanic() has a chance to call
console_verbose().  With the buddy detector, this leaves only the
detector CPU's panic stack on the console and nothing about the locked
CPU:

  <0>[ 1535.991833][    C6] watchdog: Watchdog detected hard LOCKUP on cpu 7
  <0>[ 1535.995960][    C6] Kernel panic - not syncing: Hard LOCKUP
  <4>[ 1535.995970][    C6] CPU: 6 UID: 0 PID: 0 Comm: swapper/6 ...
  <4>[ 1535.995997][    C6] Call trace:
  ...
  <4>[ 1535.996063][    C6]  watchdog_hardlockup_check+0x1e0/0x288
  <4>[ 1535.996075][    C6]  watchdog_buddy_check_hardlockup+0xb4/0xc0

Similar to commit 168e06f7937d ("kernel/hung_task.c: force console
verbose before panic"), call console_verbose() before printing the
lockup report so the diagnostic output is not suppressed.

Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org>
---
 kernel/watchdog.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index e5134ad7b663..64da2e500144 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -255,6 +255,9 @@ void watchdog_hardlockup_check(unsigned int cpu, struct pt_regs *regs)
 			return;
 	}
 
+	if (hardlockup_panic)
+		console_verbose();
+
 	/*
 	 * NOTE: we call printk_cpu_sync_get_irqsave() after printing
 	 * the lockup message. While it would be nice to serialize
@@ -880,6 +883,10 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
 		/* Start period for the next softlockup warning. */
 		update_report_ts();
 
+		thresh_count = duration / get_softlockup_thresh();
+		if (softlockup_panic && thresh_count >= softlockup_panic)
+			console_verbose();
+
 		printk_cpu_sync_get_irqsave(flags);
 		pr_emerg("BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n",
 			smp_processor_id(), duration,
@@ -901,7 +908,6 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
 
 		add_taint(TAINT_SOFTLOCKUP, LOCKDEP_STILL_OK);
 		sys_info(softlockup_si_mask & ~SYS_INFO_ALL_BT);
-		thresh_count = duration / get_softlockup_thresh();
 
 		if (softlockup_panic && thresh_count >= softlockup_panic)
 			panic("softlockup: hung tasks");
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

Re: [PATCH] watchdog: Force console verbose when reporting lockups

[Bradley Morgan]

On 7 October 2026 06:17:26 BST, Tzung-Bi Shih <tzungbi@kernel.org> wrote:
>When a hard or soft lockup is detected, the watchdog prints a header at
>KERN_EMERG (<0>), followed by diagnostic details (modules, registers,
>and backtraces) at KERN_WARNING/KERN_DEFAULT (<4>) and KERN_INFO (<6>).
>If console_loglevel is quiet (<= 4), the console skips those <4>/<6>
>diagnostic messages before vpanic() has a chance to call
>console_verbose().  With the buddy detector, this leaves only the
>detector CPU's panic stack on the console and nothing about the locked
>CPU:
>
>  <0>[ 1535.991833][    C6] watchdog: Watchdog detected hard LOCKUP on cpu 7
>  <0>[ 1535.995960][    C6] Kernel panic - not syncing: Hard LOCKUP
>  <4>[ 1535.995970][    C6] CPU: 6 UID: 0 PID: 0 Comm: swapper/6 ...
>  <4>[ 1535.995997][    C6] Call trace:
>  ...
>  <4>[ 1535.996063][    C6]  watchdog_hardlockup_check+0x1e0/0x288
>  <4>[ 1535.996075][    C6]  watchdog_buddy_check_hardlockup+0xb4/0xc0
>
>Similar to commit 168e06f7937d ("kernel/hung_task.c: force console
>verbose before panic"), call console_verbose() before printing the
>lockup report so the diagnostic output is not suppressed.
>

I personally love this patch, it's like incredibly useful:

Reviewed-by: Bradley Morgan <brads@mainlining.org>

I tested it on two machines:


Tested-by: Bradley Morgan <brads@mainlining.org> # Power10 / ARM64


>Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org>
>---
> kernel/watchdog.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
>diff --git a/kernel/watchdog.c b/kernel/watchdog.c
>index e5134ad7b663..64da2e500144 100644
>--- a/kernel/watchdog.c
>+++ b/kernel/watchdog.c
>@@ -255,6 +255,9 @@ void watchdog_hardlockup_check(unsigned int cpu, struct pt_regs *regs)
> 			return;
> 	}
> 
>+	if (hardlockup_panic)
>+		console_verbose();
>+
> 	/*
> 	 * NOTE: we call printk_cpu_sync_get_irqsave() after printing
> 	 * the lockup message. While it would be nice to serialize
>@@ -880,6 +883,10 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
> 		/* Start period for the next softlockup warning. */
> 		update_report_ts();
> 
>+		thresh_count = duration / get_softlockup_thresh();
>+		if (softlockup_panic && thresh_count >= softlockup_panic)
>+			console_verbose();

Makes sense.

>+
> 		printk_cpu_sync_get_irqsave(flags);
> 		pr_emerg("BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n",
> 			smp_processor_id(), duration,
>@@ -901,7 +908,6 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
> 
> 		add_taint(TAINT_SOFTLOCKUP, LOCKDEP_STILL_OK);
> 		sys_info(softlockup_si_mask & ~SYS_INFO_ALL_BT);
>-		thresh_count = duration / get_softlockup_thresh();
> 
> 		if (softlockup_panic && thresh_count >= softlockup_panic)
> 			panic("softlockup: hung tasks");
>

--- Thanks!
"I'm not a very positive person" - Linus torvalds

Re: [PATCH] watchdog: Force console verbose when reporting lockups

[Doug Anderson]

Hi,

On Tue, Oct 6, 2026 at 10:17 PM Tzung-Bi Shih <tzungbi@kernel.org> wrote:
>
> When a hard or soft lockup is detected, the watchdog prints a header at
> KERN_EMERG (<0>), followed by diagnostic details (modules, registers,
> and backtraces) at KERN_WARNING/KERN_DEFAULT (<4>) and KERN_INFO (<6>).
> If console_loglevel is quiet (<= 4), the console skips those <4>/<6>
> diagnostic messages before vpanic() has a chance to call
> console_verbose().  With the buddy detector, this leaves only the
> detector CPU's panic stack on the console and nothing about the locked
> CPU:
>
>   <0>[ 1535.991833][    C6] watchdog: Watchdog detected hard LOCKUP on cpu 7
>   <0>[ 1535.995960][    C6] Kernel panic - not syncing: Hard LOCKUP
>   <4>[ 1535.995970][    C6] CPU: 6 UID: 0 PID: 0 Comm: swapper/6 ...
>   <4>[ 1535.995997][    C6] Call trace:
>   ...
>   <4>[ 1535.996063][    C6]  watchdog_hardlockup_check+0x1e0/0x288
>   <4>[ 1535.996075][    C6]  watchdog_buddy_check_hardlockup+0xb4/0xc0
>
> Similar to commit 168e06f7937d ("kernel/hung_task.c: force console
> verbose before panic"), call console_verbose() before printing the
> lockup report so the diagnostic output is not suppressed.
>
> Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org>
> ---
>  kernel/watchdog.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)

Reviewed-by: Douglas Anderson <dianders@chromium.org>