diff mbox series

[PATCHv2,1/2] watchdog/softlockup: low-overhead detection of interrupt storm

Message ID 20240130074744.45759-2-yaoma@linux.alibaba.com
State New
Headers
Series *** Detect interrupt storm in softlockup *** |

Commit Message

Bitao Hu Jan. 30, 2024, 7:47 a.m. UTC 
  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 | 72 +++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 72 insertions(+)

Comments

Liu Song Jan. 31, 2024, 1:19 a.m. UTC | #1 
在 2024/1/30 15:47, Bitao Hu 写道:
> 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 | 72 +++++++++++++++++++++++++++++++++++++++++++++++
>   1 file changed, 72 insertions(+)
>
> diff --git a/kernel/watchdog.c b/kernel/watchdog.c
> index 81a8862295d6..0efe9604c3c2 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 +443,73 @@ static int is_softlockup(unsigned long touch_ts,
>   	return 0;
>   }
>   
> +#ifdef CONFIG_IRQ_TIME_ACCOUNTING
> +#define NUM_STATS_GROUPS	5
> +#define STATS_SYSTEM		0
> +#define STATS_SOFTIRQ		1
> +#define STATS_HARDIRQ		2
> +#define STATS_IDLE		3
> +#define NUM_STATS_PER_GROUP	4
This is a set of related numbers; wouldn't it be better to use an enum?
> +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 enum cpu_usage_stat idx_to_stat[NUM_STATS_PER_GROUP] = {
> +	CPUTIME_SYSTEM, CPUTIME_SOFTIRQ, CPUTIME_IRQ, CPUTIME_IDLE
> +};
To be honest, I'm not particularly fond of the name 'idx_to_stat' as the 
concept of an
index is already implied by the nature of an array, so adding 'idx' is 
redundant.
I suggest shortening the name.

> +
> +static void update_cpustat(void)
> +{
> +	u8 i;
> +	u16 *old = this_cpu_ptr(cpustat_old);
> +	u8 (*utilization)[NUM_STATS_PER_GROUP] = this_cpu_ptr(cpustat_utilization);
> +	u8 tail = this_cpu_read(cpustat_tail);
> +	struct kernel_cpustat kcpustat;
> +	u64 *cpustat = kcpustat.cpustat;
> +	u16 sample_period_ms = sample_period >> 24LL; /* 2^24ns ~= 16.8ms */

There are two instances where right shift operations are used; it is 
suggested to employ a helper macro for a more comfortable look.


> +
> +	kcpustat_cpu_fetch(&kcpustat, smp_processor_id());
> +	for (i = STATS_SYSTEM; i < NUM_STATS_PER_GROUP; i++) {
> +		/*
> +		 * 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
> +		 */
> +		cpustat[idx_to_stat[i]] = lower_16_bits(cpustat[idx_to_stat[i]] >> 24LL);
> +		utilization[tail][i] = 100 * (u16)(cpustat[idx_to_stat[i]] - old[i])
> +					/ sample_period_ms;
> +		old[i] = cpustat[idx_to_stat[i]];
> +	}
> +	this_cpu_write(cpustat_tail, (tail + 1) % NUM_STATS_GROUPS);
> +}
> +
> +static void print_cpustat(void)
> +{
> +	u8 i, j;
> +	u8 (*utilization)[NUM_STATS_PER_GROUP] = this_cpu_ptr(cpustat_utilization);
> +	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,
> +			utilization[i][STATS_SYSTEM], utilization[i][STATS_SOFTIRQ],
> +			utilization[i][STATS_HARDIRQ], 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 +573,8 @@ 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 +610,7 @@ 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)
kernel test robot Jan. 31, 2024, 4:28 a.m. UTC | #2 
Hi Bitao,

kernel test robot noticed the following build warnings:

[auto build test WARNING on akpm-mm/mm-everything]
[also build test WARNING on linus/master v6.8-rc2 next-20240131]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Bitao-Hu/watchdog-softlockup-low-overhead-detection-of-interrupt-storm/20240130-155029
base:   https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-everything
patch link:    https://lore.kernel.org/r/20240130074744.45759-2-yaoma%40linux.alibaba.com
patch subject: [PATCHv2 1/2] watchdog/softlockup: low-overhead detection of interrupt storm
config: x86_64-randconfig-122-20240130 (https://download.01.org/0day-ci/archive/20240131/202401311206.nS3XM5Li-lkp@intel.com/config)
compiler: clang version 17.0.6 (https://github.com/llvm/llvm-project 6009708b4367171ccdbf4b5905cb6a803753fe18)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240131/202401311206.nS3XM5Li-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202401311206.nS3XM5Li-lkp@intel.com/

sparse warnings: (new ones prefixed by >>)
>> kernel/watchdog.c:464:50: sparse: sparse: incorrect type in initializer (different modifiers) @@     expected unsigned char [usertype] ( *utilization )[4] @@     got unsigned char [noderef] [usertype] ( * )[4] @@
   kernel/watchdog.c:464:50: sparse:     expected unsigned char [usertype] ( *utilization )[4]
   kernel/watchdog.c:464:50: sparse:     got unsigned char [noderef] [usertype] ( * )[4]
   kernel/watchdog.c:489:50: sparse: sparse: incorrect type in initializer (different modifiers) @@     expected unsigned char [usertype] ( *utilization )[4] @@     got unsigned char [noderef] [usertype] ( * )[4] @@
   kernel/watchdog.c:489:50: sparse:     expected unsigned char [usertype] ( *utilization )[4]
   kernel/watchdog.c:489:50: sparse:     got unsigned char [noderef] [usertype] ( * )[4]

vim +464 kernel/watchdog.c

   459	
   460	static void update_cpustat(void)
   461	{
   462		u8 i;
   463		u16 *old = this_cpu_ptr(cpustat_old);
 > 464		u8 (*utilization)[NUM_STATS_PER_GROUP] = this_cpu_ptr(cpustat_utilization);
   465		u8 tail = this_cpu_read(cpustat_tail);
   466		struct kernel_cpustat kcpustat;
   467		u64 *cpustat = kcpustat.cpustat;
   468		u16 sample_period_ms = sample_period >> 24LL; /* 2^24ns ~= 16.8ms */
   469	
   470		kcpustat_cpu_fetch(&kcpustat, smp_processor_id());
   471		for (i = STATS_SYSTEM; i < NUM_STATS_PER_GROUP; i++) {
   472			/*
   473			 * We don't need nanosecond resolution. A granularity of 16ms is
   474			 * sufficient for our precision, allowing us to use u16 to store
   475			 * cpustats, which will roll over roughly every ~1000 seconds.
   476			 * 2^24 ~= 16 * 10^6
   477			 */
   478			cpustat[idx_to_stat[i]] = lower_16_bits(cpustat[idx_to_stat[i]] >> 24LL);
   479			utilization[tail][i] = 100 * (u16)(cpustat[idx_to_stat[i]] - old[i])
   480						/ sample_period_ms;
   481			old[i] = cpustat[idx_to_stat[i]];
   482		}
   483		this_cpu_write(cpustat_tail, (tail + 1) % NUM_STATS_GROUPS);
   484	}
   485
Bitao Hu Jan. 31, 2024, 11:06 a.m. UTC | #3 
On 2024/1/31 09:19, Liu Song wrote:
> 
> 在 2024/1/30 15:47, Bitao Hu 写道:
>> 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 | 72 +++++++++++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 72 insertions(+)
>>
>> diff --git a/kernel/watchdog.c b/kernel/watchdog.c
>> index 81a8862295d6..0efe9604c3c2 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 +443,73 @@ static int is_softlockup(unsigned long touch_ts,
>>       return 0;
>>   }
>> +#ifdef CONFIG_IRQ_TIME_ACCOUNTING
>> +#define NUM_STATS_GROUPS    5
>> +#define STATS_SYSTEM        0
>> +#define STATS_SOFTIRQ        1
>> +#define STATS_HARDIRQ        2
>> +#define STATS_IDLE        3
>> +#define NUM_STATS_PER_GROUP    4
> This is a set of related numbers; wouldn't it be better to use an enum?
Agree.
>> +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 enum cpu_usage_stat idx_to_stat[NUM_STATS_PER_GROUP] = {
>> +    CPUTIME_SYSTEM, CPUTIME_SOFTIRQ, CPUTIME_IRQ, CPUTIME_IDLE
>> +};
> To be honest, I'm not particularly fond of the name 'idx_to_stat' as the 
> concept of an
> index is already implied by the nature of an array, so adding 'idx' is 
> redundant.
> I suggest shortening the name.
OK, the name 'stats' is clear enough here.
> 
>> +
>> +static void update_cpustat(void)
>> +{
>> +    u8 i;
>> +    u16 *old = this_cpu_ptr(cpustat_old);
>> +    u8 (*utilization)[NUM_STATS_PER_GROUP] = 
>> this_cpu_ptr(cpustat_utilization);
>> +    u8 tail = this_cpu_read(cpustat_tail);
>> +    struct kernel_cpustat kcpustat;
>> +    u64 *cpustat = kcpustat.cpustat;
>> +    u16 sample_period_ms = sample_period >> 24LL; /* 2^24ns ~= 16.8ms */
> 
> There are two instances where right shift operations are used; it is 
> suggested to employ a helper macro for a more comfortable look.
OK.
> 
> 
>> +
>> +    kcpustat_cpu_fetch(&kcpustat, smp_processor_id());
>> +    for (i = STATS_SYSTEM; i < NUM_STATS_PER_GROUP; i++) {
>> +        /*
>> +         * 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
>> +         */
>> +        cpustat[idx_to_stat[i]] = 
>> lower_16_bits(cpustat[idx_to_stat[i]] >> 24LL);
>> +        utilization[tail][i] = 100 * (u16)(cpustat[idx_to_stat[i]] - 
>> old[i])
>> +                    / sample_period_ms;
>> +        old[i] = cpustat[idx_to_stat[i]];
>> +    }
>> +    this_cpu_write(cpustat_tail, (tail + 1) % NUM_STATS_GROUPS);
>> +}
>> +
>> +static void print_cpustat(void)
>> +{
>> +    u8 i, j;
>> +    u8 (*utilization)[NUM_STATS_PER_GROUP] = 
>> this_cpu_ptr(cpustat_utilization);
>> +    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,
>> +            utilization[i][STATS_SYSTEM], utilization[i][STATS_SOFTIRQ],
>> +            utilization[i][STATS_HARDIRQ], 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 +573,8 @@ 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 +610,7 @@ 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)
diff mbox series

Patch

diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index 81a8862295d6..0efe9604c3c2 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 +443,73 @@  static int is_softlockup(unsigned long touch_ts,
 	return 0;
 }
 
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+#define NUM_STATS_GROUPS	5
+#define STATS_SYSTEM		0
+#define STATS_SOFTIRQ		1
+#define STATS_HARDIRQ		2
+#define STATS_IDLE		3
+#define NUM_STATS_PER_GROUP	4
+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 enum cpu_usage_stat idx_to_stat[NUM_STATS_PER_GROUP] = {
+	CPUTIME_SYSTEM, CPUTIME_SOFTIRQ, CPUTIME_IRQ, CPUTIME_IDLE
+};
+
+static void update_cpustat(void)
+{
+	u8 i;
+	u16 *old = this_cpu_ptr(cpustat_old);
+	u8 (*utilization)[NUM_STATS_PER_GROUP] = this_cpu_ptr(cpustat_utilization);
+	u8 tail = this_cpu_read(cpustat_tail);
+	struct kernel_cpustat kcpustat;
+	u64 *cpustat = kcpustat.cpustat;
+	u16 sample_period_ms = sample_period >> 24LL; /* 2^24ns ~= 16.8ms */
+
+	kcpustat_cpu_fetch(&kcpustat, smp_processor_id());
+	for (i = STATS_SYSTEM; i < NUM_STATS_PER_GROUP; i++) {
+		/*
+		 * 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
+		 */
+		cpustat[idx_to_stat[i]] = lower_16_bits(cpustat[idx_to_stat[i]] >> 24LL);
+		utilization[tail][i] = 100 * (u16)(cpustat[idx_to_stat[i]] - old[i])
+					/ sample_period_ms;
+		old[i] = cpustat[idx_to_stat[i]];
+	}
+	this_cpu_write(cpustat_tail, (tail + 1) % NUM_STATS_GROUPS);
+}
+
+static void print_cpustat(void)
+{
+	u8 i, j;
+	u8 (*utilization)[NUM_STATS_PER_GROUP] = this_cpu_ptr(cpustat_utilization);
+	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,
+			utilization[i][STATS_SYSTEM], utilization[i][STATS_SOFTIRQ],
+			utilization[i][STATS_HARDIRQ], 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 +573,8 @@  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 +610,7 @@  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)