linux-kernel - [PATCH] watchdog: Fix watchdog may detect false positive of softlockup

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <20250416013922.2905051-1-luogengkun@huaweicloud.com>
Date: Wed, 16 Apr 2025 01:39:22 +0000
From: Luo Gengkun <luogengkun@...weicloud.com>
To: akpm@...ux-foundation.org
Cc: joel.granados@...nel.org,
	song@...nel.org,
	dianders@...omium.org,
	tglx@...utronix.de,
	linux-kernel@...r.kernel.org,
	luogengkun@...weicloud.com
Subject: [PATCH] watchdog: Fix watchdog may detect false positive of softlockup

The watchdog may dectect false positive of softlockup because of stop
softlockup after update watchdog_thresh. The problem can be described as
follow:

 # We asuume previous watchdog_thresh is 60, so the timer is coming every
 # 24s.
echo 10 > /proc/sys/kernel/watchdog_thresh (User space)
|
+------>+ update watchdog_thresh (We are in kernel now)
	|
	|
	+------>+ watchdog hrtimer (irq context: detect softlockup)
		|
		|
	+-------+
	|
	|
	+ softlockup_stop_all

As showed above, there is a window between update watchdog_thresh and
softlockup_stop_all. During this window, if a timer is coming, a false
positive of softlockup will happen. To fix this problem, use a shadow
variable to store the new value and write back to watchdog_thresh after
softlockup_stop_all.

Signed-off-by: Luo Gengkun <luogengkun@...weicloud.com>
---
 kernel/watchdog.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index 9fa2af9dbf2c..80e5a77e92fd 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -47,6 +47,7 @@ int __read_mostly watchdog_user_enabled = 1;
 static int __read_mostly watchdog_hardlockup_user_enabled = WATCHDOG_HARDLOCKUP_DEFAULT;
 static int __read_mostly watchdog_softlockup_user_enabled = 1;
 int __read_mostly watchdog_thresh = 10;
+static int __read_mostly watchdog_thresh_shadow;
 static int __read_mostly watchdog_hardlockup_available;
 
 struct cpumask watchdog_cpumask __read_mostly;
@@ -876,6 +877,7 @@ static void __lockup_detector_reconfigure(void)
 	watchdog_hardlockup_stop();
 
 	softlockup_stop_all();
+	watchdog_thresh = READ_ONCE(watchdog_thresh_shadow);
 	set_sample_period();
 	lockup_detector_update_enable();
 	if (watchdog_enabled && watchdog_thresh)
@@ -1035,10 +1037,12 @@ static int proc_watchdog_thresh(const struct ctl_table *table, int write,
 
 	mutex_lock(&watchdog_mutex);
 
-	old = READ_ONCE(watchdog_thresh);
+	watchdog_thresh_shadow = READ_ONCE(watchdog_thresh);
+
+	old = watchdog_thresh_shadow;
 	err = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
 
-	if (!err && write && old != READ_ONCE(watchdog_thresh))
+	if (!err && write && old != READ_ONCE(watchdog_thresh_shadow))
 		proc_watchdog_update();
 
 	mutex_unlock(&watchdog_mutex);
@@ -1080,7 +1084,7 @@ static const struct ctl_table watchdog_sysctls[] = {
 	},
 	{
 		.procname	= "watchdog_thresh",
-		.data		= &watchdog_thresh,
+		.data		= &watchdog_thresh_shadow,
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= proc_watchdog_thresh,
-- 
2.34.1