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 | 89 +++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 89 insertions(+)
diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index 81a8862295d6..71d5b6dfa358 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -16,6 +16,8 @@
#include <linux/cpu.h>
#include <linux/nmi.h>
#include <linux/init.h>
+#include <linux/kernel_stat.h>
+#include <linux/math64.h>
#include <linux/module.h>
#include <linux/sysctl.h>
#include <linux/tick.h>
@@ -333,6 +335,90 @@ __setup("watchdog_thresh=", watchdog_thresh_setup);
static void __lockup_detector_cleanup(void);
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+#define NUM_STATS_GROUPS 5
+#define NUM_STATS_PER_GROUP 4
+enum stats_per_group {
+ STATS_SYSTEM,
+ STATS_SOFTIRQ,
+ STATS_HARDIRQ,
+ STATS_IDLE,
+};
+static const enum cpu_usage_stat tracked_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_util[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_ns)
+{
+ return data_ns >> 24LL; /* 2^24ns ~= 16.8ms */
+}
+
+static void update_cpustat(void)
+{
+ int i;
+ u8 util;
+ u16 old_stat, new_stat;
+ struct kernel_cpustat kcpustat;
+ u64 *cpustat = kcpustat.cpustat;
+ u8 tail = __this_cpu_read(cpustat_tail);
+ u16 sample_period_16 = get_16bit_precision(sample_period);
+
+ kcpustat_cpu_fetch(&kcpustat, smp_processor_id());
+ for (i = 0; i < NUM_STATS_PER_GROUP; i++) {
+ old_stat = __this_cpu_read(cpustat_old[i]);
+ new_stat = get_16bit_precision(cpustat[tracked_stats[i]]);
+ util = DIV_ROUND_UP(100 * (new_stat - old_stat), sample_period_16);
+ __this_cpu_write(cpustat_util[tail][i], util);
+ __this_cpu_write(cpustat_old[i], new_stat);
+ }
+ __this_cpu_write(cpustat_tail, (tail + 1) % NUM_STATS_GROUPS);
+}
+
+static void print_cpustat(void)
+{
+ int i, group;
+ 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 (i = 0; i < NUM_STATS_GROUPS; i++) {
+ group = (tail + i) % NUM_STATS_GROUPS;
+ printk(KERN_CRIT "\t#%d: %3u%% system,\t%3u%% softirq,\t"
+ "%3u%% hardirq,\t%3u%% idle\n", i+1,
+ __this_cpu_read(cpustat_util[group][STATS_SYSTEM]),
+ __this_cpu_read(cpustat_util[group][STATS_SOFTIRQ]),
+ __this_cpu_read(cpustat_util[group][STATS_HARDIRQ]),
+ __this_cpu_read(cpustat_util[group][STATS_IDLE]));
+ }
+}
+
+static void report_cpu_status(void)
+{
+ print_cpustat();
+}
+#else
+static inline void update_cpustat(void) { }
+static inline void report_cpu_status(void) { }
+#endif
+
/*
* Hard-lockup warnings should be triggered after just a few seconds. Soft-
* lockups can have false positives under extreme conditions. So we generally
@@ -504,6 +590,8 @@ 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 +627,7 @@ 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));
+ report_cpu_status();
print_modules();
print_irqtrace_events(current);
if (regs)
--
2.37.1 (Apple Git-137.1)
Hi,
On Tue, Feb 6, 2024 at 1:59 AM Bitao Hu <yaoma@linux.alibaba.com> wrote:
>
> 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 | 89 +++++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 89 insertions(+)
On v4 you got Liu Song's Reviewed-by and I don't think this is
massively different than v4. I would have expected you to carry the
tag forward. In any case ,I guess Liu Song can give it again...
> diff --git a/kernel/watchdog.c b/kernel/watchdog.c
> index 81a8862295d6..71d5b6dfa358 100644
> --- a/kernel/watchdog.c
> +++ b/kernel/watchdog.c
> @@ -16,6 +16,8 @@
> #include <linux/cpu.h>
> #include <linux/nmi.h>
> #include <linux/init.h>
> +#include <linux/kernel_stat.h>
> +#include <linux/math64.h>
> #include <linux/module.h>
> #include <linux/sysctl.h>
> #include <linux/tick.h>
> @@ -333,6 +335,90 @@ __setup("watchdog_thresh=", watchdog_thresh_setup);
>
> static void __lockup_detector_cleanup(void);
>
> +#ifdef CONFIG_IRQ_TIME_ACCOUNTING
> +#define NUM_STATS_GROUPS 5
> +#define NUM_STATS_PER_GROUP 4
> +enum stats_per_group {
> + STATS_SYSTEM,
> + STATS_SOFTIRQ,
> + STATS_HARDIRQ,
> + STATS_IDLE,
nit: I still would have left "NUM_STATS_PER_GROUP" here instead of as
a separate #define.
> +static void print_cpustat(void)
> +{
> + int i, group;
> + u8 tail = __this_cpu_read(cpustat_tail);
Sorry for not noticing before, but why are you using
"__this_cpu_read()" instead of "this_cpu_read()"? In other words, why
do you need the double-underscore version everywhere? I don't think
you do, do you?
> + 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 (i = 0; i < NUM_STATS_GROUPS; i++) {
> + group = (tail + i) % NUM_STATS_GROUPS;
> + printk(KERN_CRIT "\t#%d: %3u%% system,\t%3u%% softirq,\t"
> + "%3u%% hardirq,\t%3u%% idle\n", i+1,
nit: though I don't care too much in this case, I think kernel folks
slightly prefer "i + 1" instead of "i+1". Running
"./scripts/checkpatch.pl --strict" will give a warning about this, for
instance. Actually, "./scripts/checkpatch.pl --strict" has a few extra
style nits that you could consider fixing.
> +static void report_cpu_status(void)
> +{
> + print_cpustat();
> +}
I don't understand why you need the extra wrapper. You didn't have it
on v3 and I don't see any reason why you introduced it. Ah, I see, in
the next patch you add something to it. OK, I guess it's fine to
introduce it here.
-Doug
Hi,
On 2024/2/7 05:41, Doug Anderson wrote:
> Hi,
>
> On Tue, Feb 6, 2024 at 1:59 AM Bitao Hu <yaoma@linux.alibaba.com> wrote:
>>
>> 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 | 89 +++++++++++++++++++++++++++++++++++++++++++++++
>> 1 file changed, 89 insertions(+)
>
> On v4 you got Liu Song's Reviewed-by and I don't think this is
> massively different than v4. I would have expected you to carry the
> tag forward. In any case ,I guess Liu Song can give it again.. >
>
>> diff --git a/kernel/watchdog.c b/kernel/watchdog.c
>> index 81a8862295d6..71d5b6dfa358 100644
>> --- a/kernel/watchdog.c
>> +++ b/kernel/watchdog.c
>> @@ -16,6 +16,8 @@
>> #include <linux/cpu.h>
>> #include <linux/nmi.h>
>> #include <linux/init.h>
>> +#include <linux/kernel_stat.h>
>> +#include <linux/math64.h>
>> #include <linux/module.h>
>> #include <linux/sysctl.h>
>> #include <linux/tick.h>
>> @@ -333,6 +335,90 @@ __setup("watchdog_thresh=", watchdog_thresh_setup);
>>
>> static void __lockup_detector_cleanup(void);
>>
>> +#ifdef CONFIG_IRQ_TIME_ACCOUNTING
>> +#define NUM_STATS_GROUPS 5
>> +#define NUM_STATS_PER_GROUP 4
>> +enum stats_per_group {
>> + STATS_SYSTEM,
>> + STATS_SOFTIRQ,
>> + STATS_HARDIRQ,
>> + STATS_IDLE,
>
> nit: I still would have left "NUM_STATS_PER_GROUP" here instead of as
> a separate #define.
OK.
>
>
>> +static void print_cpustat(void)
>> +{
>> + int i, group;
>> + u8 tail = __this_cpu_read(cpustat_tail);
>
> Sorry for not noticing before, but why are you using
> "__this_cpu_read()" instead of "this_cpu_read()"? In other words, why
> do you need the double-underscore version everywhere? I don't think
> you do, do you?
I also struggled with which version of the operation to use. The one
without double-underscores provides preemption/interrupt protection,
but in watchdog.c, the version with double-underscores is used. I
analyzed that it is also safe to use the version without
preemption/interrupt protection in my code, so to maintain consistency
with watchdog.c, I ues the version with double-underscores.
Is my approach reasonable? If not, I will switch to using the
non-underscored version.
>
>
>> + 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 (i = 0; i < NUM_STATS_GROUPS; i++) {
>> + group = (tail + i) % NUM_STATS_GROUPS;
>> + printk(KERN_CRIT "\t#%d: %3u%% system,\t%3u%% softirq,\t"
>> + "%3u%% hardirq,\t%3u%% idle\n", i+1,
>
> nit: though I don't care too much in this case, I think kernel folks
> slightly prefer "i + 1" instead of "i+1". Running
> "./scripts/checkpatch.pl --strict" will give a warning about this, for
> instance. Actually, "./scripts/checkpatch.pl --strict" has a few extra
> style nits that you could consider fixing.
Thanks for your reminder. I will use "./scripts/checkpatch.pl --strict"
to check and correct these patches.
>
>
>> +static void report_cpu_status(void)
>> +{
>> + print_cpustat();
>> +}
>
> I don't understand why you need the extra wrapper. You didn't have it
> on v3 and I don't see any reason why you introduced it. Ah, I see, in
> the next patch you add something to it. OK, I guess it's fine to
> introduce it here.
Yes, I add this wrapper to prepare for the next patch, to avoid
predeclaring of "print_irq_counts".
>
> -Doug
Hi,
On Tue, Feb 6, 2024 at 10:18 PM Bitao Hu <yaoma@linux.alibaba.com> wrote:
>
> >> +static void print_cpustat(void)
> >> +{
> >> + int i, group;
> >> + u8 tail = __this_cpu_read(cpustat_tail);
> >
> > Sorry for not noticing before, but why are you using
> > "__this_cpu_read()" instead of "this_cpu_read()"? In other words, why
> > do you need the double-underscore version everywhere? I don't think
> > you do, do you?
> I also struggled with which version of the operation to use. The one
> without double-underscores provides preemption/interrupt protection,
> but in watchdog.c, the version with double-underscores is used. I
> analyzed that it is also safe to use the version without
> preemption/interrupt protection in my code, so to maintain consistency
> with watchdog.c, I ues the version with double-underscores.
>
> Is my approach reasonable? If not, I will switch to using the
> non-underscored version.
Ah, OK. I hadn't followed the macros all the way through to the
arch-specific defines and I didn't see the preemption disable. OK,
what you have seems fine to me, especially since the double-underscore
version still has double-checks that preemption is disabled. Thanks
for explaining!
-Doug
© 2016 - 2026 Red Hat, Inc.