[PATCH] workqueue: update wq_watchdog_touched for unbound lockup checking

From: Wang Qing
Date: Fri Mar 19 2021 - 03:25:37 EST


When touch_softlockup_watchdog() is called, only wq_watchdog_touched_cpu
updated, while the unbound worker_pool running on its core uses
wq_watchdog_touched to determine whether locked up. This may be mischecked.

My suggestion is to update both when touch_softlockup_watchdog() is called,
use wq_watchdog_touched_cpu to check bound, and use wq_watchdog_touched
to check unbound worker_pool.

Signed-off-by: Wang Qing <wangqing@xxxxxxxx>
---
kernel/workqueue.c | 17 ++++++-----------
1 file changed, 6 insertions(+), 11 deletions(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 0d150da..be08295
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -5787,22 +5787,17 @@ static void wq_watchdog_timer_fn(struct timer_list *unused)
continue;

/* get the latest of pool and touched timestamps */
+ if (pool->cpu >= 0)
+ touched = READ_ONCE(per_cpu(wq_watchdog_touched_cpu, pool->cpu));
+ else
+ touched = READ_ONCE(wq_watchdog_touched);
pool_ts = READ_ONCE(pool->watchdog_ts);
- touched = READ_ONCE(wq_watchdog_touched);

if (time_after(pool_ts, touched))
ts = pool_ts;
else
ts = touched;

- if (pool->cpu >= 0) {
- unsigned long cpu_touched =
- READ_ONCE(per_cpu(wq_watchdog_touched_cpu,
- pool->cpu));
- if (time_after(cpu_touched, ts))
- ts = cpu_touched;
- }
-
/* did we stall? */
if (time_after(jiffies, ts + thresh)) {
lockup_detected = true;
@@ -5826,8 +5821,8 @@ notrace void wq_watchdog_touch(int cpu)
{
if (cpu >= 0)
per_cpu(wq_watchdog_touched_cpu, cpu) = jiffies;
- else
- wq_watchdog_touched = jiffies;
+
+ wq_watchdog_touched = jiffies;
}

static void wq_watchdog_set_thresh(unsigned long thresh)
--
2.7.4