Skip to contentSkip to conversation

watchdog: Inform about watchdog timeout.

PATCH In review Mathieu Othacehe 3 participants 3 messages Jul 14, 2024 – Jul 18, 2024 Watchdog device drivers

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 watchdog: Inform about watchdog timeout.

When the kernel stops feeding the watchdog, inform the user that it is about to timeout in X seconds.

The message looks like: watchdog: watchdog0: timeout in 15 seconds

This is helpful to warn the user that, most likely, CONFIG_WATCHDOG_OPEN_TIMEOUT seconds have elapsed and the userspace hasn't start feeding the watchdog.

---
 drivers/watchdog/watchdog_dev.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
drivers/watchdog/watchdog_dev.c +12 -0
index e2bd266b1b5b3..0d95c91365993 100644
@@ -104,6 +104,7 @@ static ktime_t watchdog_next_keepalive(struct watchdog_device *wdd)
 	ktime_t last_heartbeat, latest_heartbeat;
 	ktime_t virt_timeout;
 	unsigned int hw_heartbeat_ms;
+	int delta_to_timeout;
 
 	if (watchdog_active(wdd))
 		virt_timeout = ktime_add(wd_data->last_keepalive,
@@ -121,6 +122,17 @@ static ktime_t watchdog_next_keepalive(struct watchdog_device *wdd)
 	 */
 	last_heartbeat = ktime_sub(virt_timeout, ms_to_ktime(hw_heartbeat_ms));
 	latest_heartbeat = ktime_sub(last_heartbeat, ktime_get());
+
+	/*
+	 * If we are past the last heartbeat, inform the user that we are
+	 * about to timeout.
+	 */
+	delta_to_timeout =
+		(int)ktime_ms_delta(virt_timeout, ktime_get()) / MSEC_PER_SEC;
+	if (latest_heartbeat <= 0 && delta_to_timeout > 0)
+		pr_info("watchdog%d: timeout in %d seconds\n", wdd->id,
+			delta_to_timeout);
+
 	if (ktime_before(latest_heartbeat, keepalive_interval))
 		return latest_heartbeat;
 	return keepalive_interval;
Signature · 1 line
2.45.2
maintainer
#1 lore raw
Quoted 10 lines · Mathieu Othacehe

When the kernel stops feeding the watchdog, inform the user that it is about to timeout in X seconds.

The message looks like: watchdog: watchdog0: timeout in 15 seconds

This is helpful to warn the user that, most likely, CONFIG_WATCHDOG_OPEN_TIMEOUT seconds have elapsed and the userspace hasn't start feeding the watchdog.

I don't like noisiness in the kernel, and I don't see value in this message.

Maybe Wim thinks otherwise, but for me this is NACK.

Guenter

maintainer
#2 lore raw

Hi Mathieu, Guenter,

Quoted 10 lines · Mathieu Othacehe
Quoted 9 lines

When the kernel stops feeding the watchdog, inform the user that it is about to timeout in X seconds.

The message looks like: watchdog: watchdog0: timeout in 15 seconds

This is helpful to warn the user that, most likely, CONFIG_WATCHDOG_OPEN_TIMEOUT seconds have elapsed and the userspace hasn't start feeding the watchdog.

Quoted 5 lines
Quoted 1 line

I don't like noisiness in the kernel, and I don't see value in this message.

Maybe Wim thinks otherwise, but for me this is NACK.

No, I was thinking the same. So also NACK for me.

Kind regards,
Wim.