[PATCHv3 0/2] *** Detect interrupt storm in softlockup ***

[PATCHv3 0/2] *** Detect interrupt storm in softlockup ***

[Bitao Hu]

Hi, guys.
I have implemented a low-overhead method for detecting interrupt storm
in softlockup. Please review it, all comments are welcome.

Changes from v2 to v3:

- From Liu Song, using enum instead of macro for cpu_stats, shortening
the name 'idx_to_stat' to 'stats', adding 'get_16bit_precesion' instead
of using right shift operations, and using 'struct irq_counts'.

- From kernel robot test, using '__this_cpu_read' and '__this_cpu_write'
instead of accessing to an per-cpu array directly, in order to avoid
this warning.
'sparse: incorrect type in initializer (different modifiers)'.

Changes from v1 to v2:

- From Douglas, optimize the memory of cpustats. With the maximum number
of CPUs, that's now this.
2 * 8192 * 4 + 1 * 8192 * 5 * 4 + 1 * 8192 = 237,568 bytes.

- From Liu Song, refactor the code format and add necessary comments.

- From Douglas, use interrupt counts instead of interrupt time to
determine the cause of softlockup.

- Remove the cmdline parameter added in PATCHv1.

Bitao Hu (2):
  watchdog/softlockup: low-overhead detection of interrupt storm
  watchdog/softlockup: report the most frequent interrupts

 kernel/watchdog.c | 238 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 238 insertions(+)

-- 
2.37.1 (Apple Git-137.1)

[PATCHv3 1/2] watchdog/softlockup: low-overhead detection of interrupt storm

[Bitao Hu]

The following softlockup is caused by interrupt storm, but it cannot be
identified from the call tree. Because the call tree is just a snapshot
and doesn't fully capture the behavior of the CPU during the soft lockup.
  watchdog: BUG: soft lockup - CPU#28 stuck for 23s! [fio:83921]
  ...
  Call trace:
    __do_softirq+0xa0/0x37c
    __irq_exit_rcu+0x108/0x140
    irq_exit+0x14/0x20
    __handle_domain_irq+0x84/0xe0
    gic_handle_irq+0x80/0x108
    el0_irq_naked+0x50/0x58

Therefore，I think it is necessary to report CPU utilization during the
softlockup_thresh period (report once every sample_period, for a total
of 5 reportings), like this:
  watchdog: BUG: soft lockup - CPU#28 stuck for 23s! [fio:83921]
  CPU#28 Utilization every 4s during lockup:
    #1: 0% system, 0% softirq, 100% hardirq, 0% idle
    #2: 0% system, 0% softirq, 100% hardirq, 0% idle
    #3: 0% system, 0% softirq, 100% hardirq, 0% idle
    #4: 0% system, 0% softirq, 100% hardirq, 0% idle
    #5: 0% system, 0% softirq, 100% hardirq, 0% idle
  ...

This would be helpful in determining whether an interrupt storm has
occurred or in identifying the cause of the softlockup. The criteria for
determination are as follows:
  a. If the hardirq utilization is high, then interrupt storm should be
  considered and the root cause cannot be determined from the call tree.
  b. If the softirq utilization is high, then we could analyze the call
  tree but it may cannot reflect the root cause.
  c. If the system utilization is high, then we could analyze the root
  cause from the call tree.

Signed-off-by: Bitao Hu <yaoma@linux.alibaba.com>
---
 kernel/watchdog.c | 84 +++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 84 insertions(+)

diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index 81a8862295d6..046507be4eb5 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -23,6 +23,8 @@
 #include <linux/sched/debug.h>
 #include <linux/sched/isolation.h>
 #include <linux/stop_machine.h>
+#include <linux/kernel_stat.h>
+#include <linux/math64.h>
 
 #include <asm/irq_regs.h>
 #include <linux/kvm_para.h>
@@ -441,6 +443,85 @@ static int is_softlockup(unsigned long touch_ts,
 	return 0;
 }
 
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+#define NUM_STATS_GROUPS	5
+enum stats_per_group {
+	STATS_SYSTEM,
+	STATS_SOFTIRQ,
+	STATS_HARDIRQ,
+	STATS_IDLE,
+	NUM_STATS_PER_GROUP,
+};
+static enum cpu_usage_stat stats[NUM_STATS_PER_GROUP] = {
+	CPUTIME_SYSTEM,
+	CPUTIME_SOFTIRQ,
+	CPUTIME_IRQ,
+	CPUTIME_IDLE,
+};
+static DEFINE_PER_CPU(u16, cpustat_old[NUM_STATS_PER_GROUP]);
+static DEFINE_PER_CPU(u8, cpustat_utilization[NUM_STATS_GROUPS][NUM_STATS_PER_GROUP]);
+static DEFINE_PER_CPU(u8, cpustat_tail);
+
+/*
+ * We don't need nanosecond resolution. A granularity of 16ms is
+ * sufficient for our precision, allowing us to use u16 to store
+ * cpustats, which will roll over roughly every ~1000 seconds.
+ * 2^24 ~= 16 * 10^6
+ */
+static u16 get_16bit_precision(u64 data)
+{
+	return data >> 24LL; /* 2^24ns ~= 16.8ms */
+}
+
+static void update_cpustat(void)
+{
+	u8 i;
+	u16 old;
+	u8 utilization;
+	u8 tail = __this_cpu_read(cpustat_tail);
+	struct kernel_cpustat kcpustat;
+	u64 *cpustat = kcpustat.cpustat;
+	u16 sample_period_ms = get_16bit_precision(sample_period);
+
+	kcpustat_cpu_fetch(&kcpustat, smp_processor_id());
+	for (i = STATS_SYSTEM; i < NUM_STATS_PER_GROUP; i++) {
+		old = __this_cpu_read(cpustat_old[i]);
+		cpustat[stats[i]] = get_16bit_precision(cpustat[stats[i]]);
+		utilization = 100 * (u16)(cpustat[stats[i]] - old) / sample_period_ms;
+		__this_cpu_write(cpustat_utilization[tail][i], utilization);
+		__this_cpu_write(cpustat_old[i], cpustat[stats[i]]);
+	}
+	__this_cpu_write(cpustat_tail, (tail + 1) % NUM_STATS_GROUPS);
+}
+
+static void print_cpustat(void)
+{
+	u8 i, j;
+	u8 tail = __this_cpu_read(cpustat_tail);
+	u64 sample_period_second = sample_period;
+
+	do_div(sample_period_second, NSEC_PER_SEC);
+	/*
+	 * We do not want the "watchdog: " prefix on every line,
+	 * hence we use "printk" instead of "pr_crit".
+	 */
+	printk(KERN_CRIT "CPU#%d Utilization every %llus during lockup:\n",
+		smp_processor_id(), sample_period_second);
+	for (j = STATS_SYSTEM, i = tail; j < NUM_STATS_GROUPS;
+		j++, i = (i + 1) % NUM_STATS_GROUPS) {
+		printk(KERN_CRIT "\t#%d: %3u%% system,\t%3u%% softirq,\t"
+			"%3u%% hardirq,\t%3u%% idle\n", j+1,
+			__this_cpu_read(cpustat_utilization[i][STATS_SYSTEM]),
+			__this_cpu_read(cpustat_utilization[i][STATS_SOFTIRQ]),
+			__this_cpu_read(cpustat_utilization[i][STATS_HARDIRQ]),
+			__this_cpu_read(cpustat_utilization[i][STATS_IDLE]));
+	}
+}
+#else
+static inline void update_cpustat(void) { }
+static inline void print_cpustat(void) { }
+#endif
+
 /* watchdog detector functions */
 static DEFINE_PER_CPU(struct completion, softlockup_completion);
 static DEFINE_PER_CPU(struct cpu_stop_work, softlockup_stop_work);
@@ -504,6 +585,8 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
 	 */
 	period_ts = READ_ONCE(*this_cpu_ptr(&watchdog_report_ts));
 
+	update_cpustat();
+
 	/* Reset the interval when touched by known problematic code. */
 	if (period_ts == SOFTLOCKUP_DELAY_REPORT) {
 		if (unlikely(__this_cpu_read(softlockup_touch_sync))) {
@@ -539,6 +622,7 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
 		pr_emerg("BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n",
 			smp_processor_id(), duration,
 			current->comm, task_pid_nr(current));
+		print_cpustat();
 		print_modules();
 		print_irqtrace_events(current);
 		if (regs)
-- 
2.37.1 (Apple Git-137.1)

[PATCHv3 2/2] watchdog/softlockup: report the most frequent interrupts

[Bitao Hu]

When the watchdog determines that the current soft lockup is due
to an interrupt storm based on CPU utilization, reporting the
most frequent interrupts could be good enough for further
troubleshooting.

Below is an example of interrupt storm. The call tree does not
provide useful information, but we can analyze which interrupt
caused the soft lockup by comparing the counts of interrupts.

[ 2987.488075] watchdog: BUG: soft lockup - CPU#9 stuck for 23s! [kworker/9:1:214]
[ 2987.488607] CPU#9 Utilization every 4s during lockup:
[ 2987.488941]  #1:   0% system,          0% softirq,   100% hardirq,     0% idle
[ 2987.489357]  #2:   0% system,          0% softirq,   100% hardirq,     0% idle
[ 2987.489771]  #3:   0% system,          0% softirq,   100% hardirq,     0% idle
[ 2987.490186]  #4:   0% system,          0% softirq,   100% hardirq,     0% idle
[ 2987.490601]  #5:   0% system,          0% softirq,   100% hardirq,     0% idle
[ 2987.491034] CPU#9 Detect HardIRQ Time exceeds 50%. Most frequent HardIRQs:
[ 2987.491493]  #1: 330985      irq#7(IPI)
[ 2987.491743]  #2: 5000        irq#10(arch_timer)
[ 2987.492039]  #3: 9           irq#91(nvme0q2)
[ 2987.492318]  #4: 3           irq#118(virtio1-output.12)
...
[ 2987.492728] Call trace:
[ 2987.492729]  __do_softirq+0xa8/0x364

Signed-off-by: Bitao Hu <yaoma@linux.alibaba.com>
---
 kernel/watchdog.c | 154 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 154 insertions(+)

diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index 046507be4eb5..6841ecdc3053 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -25,6 +25,9 @@
 #include <linux/stop_machine.h>
 #include <linux/kernel_stat.h>
 #include <linux/math64.h>
+#include <linux/irq.h>
+#include <linux/bitops.h>
+#include <linux/irqdesc.h>
 
 #include <asm/irq_regs.h>
 #include <linux/kvm_para.h>
@@ -431,11 +434,15 @@ void touch_softlockup_watchdog_sync(void)
 	__this_cpu_write(watchdog_report_ts, SOFTLOCKUP_DELAY_REPORT);
 }
 
+static void set_potential_softlockup(unsigned long now, unsigned long touch_ts);
+
 static int is_softlockup(unsigned long touch_ts,
 			 unsigned long period_ts,
 			 unsigned long now)
 {
 	if ((watchdog_enabled & WATCHDOG_SOFTOCKUP_ENABLED) && watchdog_thresh) {
+		/* Softlockup may occur in the current period */
+		set_potential_softlockup(now, period_ts);
 		/* Warn about unreasonable delays. */
 		if (time_after(now, period_ts + get_softlockup_thresh()))
 			return now - touch_ts;
@@ -462,6 +469,8 @@ static DEFINE_PER_CPU(u16, cpustat_old[NUM_STATS_PER_GROUP]);
 static DEFINE_PER_CPU(u8, cpustat_utilization[NUM_STATS_GROUPS][NUM_STATS_PER_GROUP]);
 static DEFINE_PER_CPU(u8, cpustat_tail);
 
+static void print_hardirq_counts(void);
+
 /*
  * We don't need nanosecond resolution. A granularity of 16ms is
  * sufficient for our precision, allowing us to use u16 to store
@@ -516,7 +525,151 @@ static void print_cpustat(void)
 			__this_cpu_read(cpustat_utilization[i][STATS_HARDIRQ]),
 			__this_cpu_read(cpustat_utilization[i][STATS_IDLE]));
 	}
+	print_hardirq_counts();
+}
+
+#define HARDIRQ_PERCENT_THRESH		50
+#define NUM_HARDIRQ_REPORT		5
+static DECLARE_BITMAP(softlockup_hardirq_cpus, CONFIG_NR_CPUS);
+static DEFINE_PER_CPU(u32 *, hardirq_counts);
+
+struct irq_counts {
+	int irq;
+	u32 counts;
+};
+
+static void find_counts_top(struct irq_counts *irq_counts, int irq, u32 counts, int range)
+{
+	unsigned int i, j;
+
+	for (i = 0; i < range; i++) {
+		if (counts > irq_counts[i].counts) {
+			for (j = range - 1; j > i; j--) {
+				irq_counts[j].counts = irq_counts[j - 1].counts;
+				irq_counts[j].irq = irq_counts[j - 1].irq;
+			}
+			irq_counts[j].counts = counts;
+			irq_counts[j].irq = irq;
+			break;
+		}
+	}
+}
+
+/*
+ * If the proportion of time spent handling irq exceeds HARDIRQ_PERCENT_THRESH%
+ * during sample_period, then it is necessary to record the counts of each irq.
+ */
+static inline bool need_record_irq_counts(int type)
+{
+	int tail = this_cpu_read(cpustat_tail);
+	u8 utilization;
+
+	if (--tail == -1)
+		tail = 4;
+	utilization = this_cpu_read(cpustat_utilization[tail][type]);
+	return utilization > HARDIRQ_PERCENT_THRESH;
+}
+
+/*
+ * Mark softlockup as potentially caused by hardirq
+ */
+static void set_potential_softlockup_hardirq(void)
+{
+	u32 i;
+	u32 *counts = __this_cpu_read(hardirq_counts);
+	int cpu = smp_processor_id();
+	struct irq_desc *desc;
+
+	if (!need_record_irq_counts(STATS_HARDIRQ))
+		return;
+
+	if (!test_bit(cpu, softlockup_hardirq_cpus)) {
+		counts = kmalloc_array(nr_irqs, sizeof(u32), GFP_ATOMIC);
+		if (!counts)
+			return;
+		for_each_irq_desc(i, desc) {
+			if (!desc)
+				continue;
+			counts[i] = desc->kstat_irqs ?
+				*this_cpu_ptr(desc->kstat_irqs) : 0;
+		}
+		__this_cpu_write(hardirq_counts, counts);
+		set_bit(cpu, softlockup_hardirq_cpus);
+	}
+}
+
+static void clear_potential_softlockup_hardirq(void)
+{
+	u32 *counts = __this_cpu_read(hardirq_counts);
+	int cpu = smp_processor_id();
+
+	if (test_bit(cpu, softlockup_hardirq_cpus)) {
+		kfree(counts);
+		counts = NULL;
+		__this_cpu_write(hardirq_counts, counts);
+		clear_bit(cpu, softlockup_hardirq_cpus);
+	}
 }
+
+/*
+ * Mark that softlockup may occur
+ */
+static void set_potential_softlockup(unsigned long now, unsigned long period_ts)
+{
+	if (time_after_eq(now, period_ts + get_softlockup_thresh() / 5))
+		set_potential_softlockup_hardirq();
+}
+
+static void clear_potential_softlockup(void)
+{
+	clear_potential_softlockup_hardirq();
+}
+
+static void print_hardirq_counts(void)
+{
+	u32 i;
+	struct irq_desc *desc;
+	u32 counts_diff;
+	u32 *counts = __this_cpu_read(hardirq_counts);
+	int cpu = smp_processor_id();
+	struct irq_counts hardirq_counts_top[NUM_HARDIRQ_REPORT] = {
+		{-1, 0}, {-1, 0}, {-1, 0}, {-1, 0},
+	};
+
+	if (test_bit(cpu, softlockup_hardirq_cpus)) {
+		/* Find the top NUM_HARDIRQ_REPORT most frequent interrupts */
+		for_each_irq_desc(i, desc) {
+			if (!desc)
+				continue;
+			counts_diff = desc->kstat_irqs ?
+				*this_cpu_ptr(desc->kstat_irqs) - counts[i] : 0;
+			find_counts_top(hardirq_counts_top, i, counts_diff,
+					NUM_HARDIRQ_REPORT);
+		}
+		/*
+		 * We do not want the "watchdog: " prefix on every line,
+		 * hence we use "printk" instead of "pr_crit".
+		 */
+		printk(KERN_CRIT "CPU#%d Detect HardIRQ Time exceeds %d%%. Most frequent HardIRQs:\n",
+			smp_processor_id(), HARDIRQ_PERCENT_THRESH);
+		for (i = 0; i < NUM_HARDIRQ_REPORT; i++) {
+			if (hardirq_counts_top[i].irq == -1)
+				break;
+			desc = irq_to_desc(hardirq_counts_top[i].irq);
+			if (desc && desc->action)
+				printk(KERN_CRIT "\t#%u: %-10u\tirq#%d(%s)\n",
+					i+1, hardirq_counts_top[i].counts,
+					hardirq_counts_top[i].irq, desc->action->name);
+			else
+				printk(KERN_CRIT "\t#%u: %-10u\tirq#%d\n",
+					i+1, hardirq_counts_top[i].counts,
+					hardirq_counts_top[i].irq);
+		}
+		if (!need_record_irq_counts(STATS_HARDIRQ))
+			clear_potential_softlockup_hardirq();
+	}
+}
+
 #else
 static inline void update_cpustat(void) { }
 static inline void print_cpustat(void) { }
@@ -537,6 +690,7 @@ static DEFINE_PER_CPU(struct cpu_stop_work, softlockup_stop_work);
 static int softlockup_fn(void *data)
 {
 	update_touch_ts();
+	clear_potential_softlockup();
 	complete(this_cpu_ptr(&softlockup_completion));
 
 	return 0;
-- 
2.37.1 (Apple Git-137.1)

Re: [PATCHv3 2/2] watchdog/softlockup: report the most frequent interrupts

[Bitao Hu]

The version is wrong, please ignore this.

Best regards,

Bitao Hu

On 2024/2/1 00:45, Bitao Hu wrote:
> When the watchdog determines that the current soft lockup is due
> to an interrupt storm based on CPU utilization, reporting the
> most frequent interrupts could be good enough for further
> troubleshooting.
> 
> Below is an example of interrupt storm. The call tree does not
> provide useful information, but we can analyze which interrupt
> caused the soft lockup by comparing the counts of interrupts.
> 
> [ 2987.488075] watchdog: BUG: soft lockup - CPU#9 stuck for 23s! [kworker/9:1:214]
> [ 2987.488607] CPU#9 Utilization every 4s during lockup:
> [ 2987.488941]  #1:   0% system,          0% softirq,   100% hardirq,     0% idle
> [ 2987.489357]  #2:   0% system,          0% softirq,   100% hardirq,     0% idle
> [ 2987.489771]  #3:   0% system,          0% softirq,   100% hardirq,     0% idle
> [ 2987.490186]  #4:   0% system,          0% softirq,   100% hardirq,     0% idle
> [ 2987.490601]  #5:   0% system,          0% softirq,   100% hardirq,     0% idle
> [ 2987.491034] CPU#9 Detect HardIRQ Time exceeds 50%. Most frequent HardIRQs:
> [ 2987.491493]  #1: 330985      irq#7(IPI)
> [ 2987.491743]  #2: 5000        irq#10(arch_timer)
> [ 2987.492039]  #3: 9           irq#91(nvme0q2)
> [ 2987.492318]  #4: 3           irq#118(virtio1-output.12)
> ...
> [ 2987.492728] Call trace:
> [ 2987.492729]  __do_softirq+0xa8/0x364
> 
> Signed-off-by: Bitao Hu <yaoma@linux.alibaba.com>
> ---
>   kernel/watchdog.c | 154 ++++++++++++++++++++++++++++++++++++++++++++++
>   1 file changed, 154 insertions(+)
> 
> diff --git a/kernel/watchdog.c b/kernel/watchdog.c
> index 046507be4eb5..6841ecdc3053 100644
> --- a/kernel/watchdog.c
> +++ b/kernel/watchdog.c
> @@ -25,6 +25,9 @@
>   #include <linux/stop_machine.h>
>   #include <linux/kernel_stat.h>
>   #include <linux/math64.h>
> +#include <linux/irq.h>
> +#include <linux/bitops.h>
> +#include <linux/irqdesc.h>
>   
>   #include <asm/irq_regs.h>
>   #include <linux/kvm_para.h>
> @@ -431,11 +434,15 @@ void touch_softlockup_watchdog_sync(void)
>   	__this_cpu_write(watchdog_report_ts, SOFTLOCKUP_DELAY_REPORT);
>   }
>   
> +static void set_potential_softlockup(unsigned long now, unsigned long touch_ts);
> +
>   static int is_softlockup(unsigned long touch_ts,
>   			 unsigned long period_ts,
>   			 unsigned long now)
>   {
>   	if ((watchdog_enabled & WATCHDOG_SOFTOCKUP_ENABLED) && watchdog_thresh) {
> +		/* Softlockup may occur in the current period */
> +		set_potential_softlockup(now, period_ts);
>   		/* Warn about unreasonable delays. */
>   		if (time_after(now, period_ts + get_softlockup_thresh()))
>   			return now - touch_ts;
> @@ -462,6 +469,8 @@ static DEFINE_PER_CPU(u16, cpustat_old[NUM_STATS_PER_GROUP]);
>   static DEFINE_PER_CPU(u8, cpustat_utilization[NUM_STATS_GROUPS][NUM_STATS_PER_GROUP]);
>   static DEFINE_PER_CPU(u8, cpustat_tail);
>   
> +static void print_hardirq_counts(void);
> +
>   /*
>    * We don't need nanosecond resolution. A granularity of 16ms is
>    * sufficient for our precision, allowing us to use u16 to store
> @@ -516,7 +525,151 @@ static void print_cpustat(void)
>   			__this_cpu_read(cpustat_utilization[i][STATS_HARDIRQ]),
>   			__this_cpu_read(cpustat_utilization[i][STATS_IDLE]));
>   	}
> +	print_hardirq_counts();
> +}
> +
> +#define HARDIRQ_PERCENT_THRESH		50
> +#define NUM_HARDIRQ_REPORT		5
> +static DECLARE_BITMAP(softlockup_hardirq_cpus, CONFIG_NR_CPUS);
> +static DEFINE_PER_CPU(u32 *, hardirq_counts);
> +
> +struct irq_counts {
> +	int irq;
> +	u32 counts;
> +};
> +
> +static void find_counts_top(struct irq_counts *irq_counts, int irq, u32 counts, int range)
> +{
> +	unsigned int i, j;
> +
> +	for (i = 0; i < range; i++) {
> +		if (counts > irq_counts[i].counts) {
> +			for (j = range - 1; j > i; j--) {
> +				irq_counts[j].counts = irq_counts[j - 1].counts;
> +				irq_counts[j].irq = irq_counts[j - 1].irq;
> +			}
> +			irq_counts[j].counts = counts;
> +			irq_counts[j].irq = irq;
> +			break;
> +		}
> +	}
> +}
> +
> +/*
> + * If the proportion of time spent handling irq exceeds HARDIRQ_PERCENT_THRESH%
> + * during sample_period, then it is necessary to record the counts of each irq.
> + */
> +static inline bool need_record_irq_counts(int type)
> +{
> +	int tail = this_cpu_read(cpustat_tail);
> +	u8 utilization;
> +
> +	if (--tail == -1)
> +		tail = 4;
> +	utilization = this_cpu_read(cpustat_utilization[tail][type]);
> +	return utilization > HARDIRQ_PERCENT_THRESH;
> +}
> +
> +/*
> + * Mark softlockup as potentially caused by hardirq
> + */
> +static void set_potential_softlockup_hardirq(void)
> +{
> +	u32 i;
> +	u32 *counts = __this_cpu_read(hardirq_counts);
> +	int cpu = smp_processor_id();
> +	struct irq_desc *desc;
> +
> +	if (!need_record_irq_counts(STATS_HARDIRQ))
> +		return;
> +
> +	if (!test_bit(cpu, softlockup_hardirq_cpus)) {
> +		counts = kmalloc_array(nr_irqs, sizeof(u32), GFP_ATOMIC);
> +		if (!counts)
> +			return;
> +		for_each_irq_desc(i, desc) {
> +			if (!desc)
> +				continue;
> +			counts[i] = desc->kstat_irqs ?
> +				*this_cpu_ptr(desc->kstat_irqs) : 0;
> +		}
> +		__this_cpu_write(hardirq_counts, counts);
> +		set_bit(cpu, softlockup_hardirq_cpus);
> +	}
> +}
> +
> +static void clear_potential_softlockup_hardirq(void)
> +{
> +	u32 *counts = __this_cpu_read(hardirq_counts);
> +	int cpu = smp_processor_id();
> +
> +	if (test_bit(cpu, softlockup_hardirq_cpus)) {
> +		kfree(counts);
> +		counts = NULL;
> +		__this_cpu_write(hardirq_counts, counts);
> +		clear_bit(cpu, softlockup_hardirq_cpus);
> +	}
>   }
> +
> +/*
> + * Mark that softlockup may occur
> + */
> +static void set_potential_softlockup(unsigned long now, unsigned long period_ts)
> +{
> +	if (time_after_eq(now, period_ts + get_softlockup_thresh() / 5))
> +		set_potential_softlockup_hardirq();
> +}
> +
> +static void clear_potential_softlockup(void)
> +{
> +	clear_potential_softlockup_hardirq();
> +}
> +
> +static void print_hardirq_counts(void)
> +{
> +	u32 i;
> +	struct irq_desc *desc;
> +	u32 counts_diff;
> +	u32 *counts = __this_cpu_read(hardirq_counts);
> +	int cpu = smp_processor_id();
> +	struct irq_counts hardirq_counts_top[NUM_HARDIRQ_REPORT] = {
> +		{-1, 0}, {-1, 0}, {-1, 0}, {-1, 0},
> +	};
> +
> +	if (test_bit(cpu, softlockup_hardirq_cpus)) {
> +		/* Find the top NUM_HARDIRQ_REPORT most frequent interrupts */
> +		for_each_irq_desc(i, desc) {
> +			if (!desc)
> +				continue;
> +			counts_diff = desc->kstat_irqs ?
> +				*this_cpu_ptr(desc->kstat_irqs) - counts[i] : 0;
> +			find_counts_top(hardirq_counts_top, i, counts_diff,
> +					NUM_HARDIRQ_REPORT);
> +		}
> +		/*
> +		 * We do not want the "watchdog: " prefix on every line,
> +		 * hence we use "printk" instead of "pr_crit".
> +		 */
> +		printk(KERN_CRIT "CPU#%d Detect HardIRQ Time exceeds %d%%. Most frequent HardIRQs:\n",
> +			smp_processor_id(), HARDIRQ_PERCENT_THRESH);
> +		for (i = 0; i < NUM_HARDIRQ_REPORT; i++) {
> +			if (hardirq_counts_top[i].irq == -1)
> +				break;
> +			desc = irq_to_desc(hardirq_counts_top[i].irq);
> +			if (desc && desc->action)
> +				printk(KERN_CRIT "\t#%u: %-10u\tirq#%d(%s)\n",
> +					i+1, hardirq_counts_top[i].counts,
> +					hardirq_counts_top[i].irq, desc->action->name);
> +			else
> +				printk(KERN_CRIT "\t#%u: %-10u\tirq#%d\n",
> +					i+1, hardirq_counts_top[i].counts,
> +					hardirq_counts_top[i].irq);
> +		}
> +		if (!need_record_irq_counts(STATS_HARDIRQ))
> +			clear_potential_softlockup_hardirq();
> +	}
> +}
> +
>   #else
>   static inline void update_cpustat(void) { }
>   static inline void print_cpustat(void) { }
> @@ -537,6 +690,7 @@ static DEFINE_PER_CPU(struct cpu_stop_work, softlockup_stop_work);
>   static int softlockup_fn(void *data)
>   {
>   	update_touch_ts();
> +	clear_potential_softlockup();
>   	complete(this_cpu_ptr(&softlockup_completion));
>   
>   	return 0;

[PATCHv3 0/2] *** Detect interrupt storm in softlockup ***

[Bitao Hu]

Hi, guys.
I have implemented a low-overhead method for detecting interrupt storm
in softlockup. Please review it, all comments are welcome.

Changes from v2 to v3:

- From Liu Song, using enum instead of macro for cpu_stats, shortening
the name 'idx_to_stat' to 'stats', adding 'get_16bit_precesion' instead
of using right shift operations, and using 'struct irq_counts'.

- From kernel robot test, using '__this_cpu_read' and '__this_cpu_write'
instead of accessing to an per-cpu array directly, in order to avoid
this warning.
'sparse: incorrect type in initializer (different modifiers)'

Changes from v1 to v2:

- From Douglas, optimize the memory of cpustats. With the maximum number
of CPUs, that's now this.
2 * 8192 * 4 + 1 * 8192 * 5 * 4 + 1 * 8192 = 237,568 bytes.

- From Liu Song, refactor the code format and add necessary comments.

- From Douglas, use interrupt counts instead of interrupt time to
determine the cause of softlockup.

- Remove the cmdline parameter added in PATCHv1.

Bitao Hu (2):
  watchdog/softlockup: low-overhead detection of interrupt storm
  watchdog/softlockup: report the most frequent interrupts

 kernel/watchdog.c | 240 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 240 insertions(+)

-- 
2.37.1 (Apple Git-137.1)