[PATCH] workqueue: Control the frequency of intensive warning through cmdline

From: Xuewen Yan
Date: Mon Feb 19 2024 - 02:50:04 EST


When CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel will report
the work functions which violate the intensive_threshold_us repeatedly.
And now, only when the violate times exceed 4 and is a power of 2,
the kernel warning could be triggered.

However, sometimes we want to print the warning every time when the work
executed too long. Because sometimes, even if a long work execution time
occurs only once, it may cause other work to be delayed for a long time.

In order to freely control the frequency of printing, a boot argument
is added so that the user can control the warning to be printed
only after a certain number of work timeouts.

Default, it would print warning every 4 times.

Signed-off-by: Xuewen Yan <xuewen.yan@xxxxxxxxxx>
---
Documentation/admin-guide/kernel-parameters.txt | 9 +++++++++
kernel/workqueue.c | 10 ++++++++--
2 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 31b3a25680d0..599fc59fcf70 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -7225,6 +7225,15 @@
threshold repeatedly. They are likely good
candidates for using WQ_UNBOUND workqueues instead.

+ workqueue.cpu_intensive_warning_per_count=
+ If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel
+ will report the work functions which violate the
+ intensive_threshold_us repeatedly. In order to prevent
+ the kernel log from being printed too frequently.
+ Control the frequency.
+
+ Default, it will print one warning per 4 times.
+
workqueue.power_efficient
Per-cpu workqueues are generally preferred because
they show better performance thanks to cache
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 7b482a26d741..8e8cccf5329a 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -359,6 +359,10 @@ static const char *wq_affn_names[WQ_AFFN_NR_TYPES] = {
*/
static unsigned long wq_cpu_intensive_thresh_us = ULONG_MAX;
module_param_named(cpu_intensive_thresh_us, wq_cpu_intensive_thresh_us, ulong, 0644);
+#ifdef CONFIG_WQ_CPU_INTENSIVE_REPORT
+static unsigned int wq_cpu_intensive_warning_per_count = 4;
+module_param_named(cpu_intensive_warning_per_count, wq_cpu_intensive_warning_per_count, uint, 0644);
+#endif

/* see the comment above the definition of WQ_POWER_EFFICIENT */
static bool wq_power_efficient = IS_ENABLED(CONFIG_WQ_POWER_EFFICIENT_DEFAULT);
@@ -1202,7 +1206,7 @@ static void wq_cpu_intensive_report(work_func_t func)
* exponentially.
*/
cnt = atomic64_inc_return_relaxed(&ent->cnt);
- if (cnt >= 4 && is_power_of_2(cnt))
+ if (wq_cpu_intensive_warning_per_count && !(cnt % wq_cpu_intensive_warning_per_count))
printk_deferred(KERN_WARNING "workqueue: %ps hogged CPU for >%luus %llu times, consider switching to WQ_UNBOUND\n",
ent->func, wq_cpu_intensive_thresh_us,
atomic64_read(&ent->cnt));
@@ -1231,10 +1235,12 @@ static void wq_cpu_intensive_report(work_func_t func)

ent = &wci_ents[wci_nr_ents++];
ent->func = func;
- atomic64_set(&ent->cnt, 1);
+ atomic64_set(&ent->cnt, 0);
hash_add_rcu(wci_hash, &ent->hash_node, (unsigned long)func);

raw_spin_unlock(&wci_lock);
+
+ goto restart;
}

#else /* CONFIG_WQ_CPU_INTENSIVE_REPORT */
--
2.25.1