[PATCH v3] timekeeping: Use time_after() in timekeeping_check_update()

Chen Yufan posted 1 patch 1 year, 5 months ago
include/linux/timekeeper_internal.h | 2 +-
kernel/time/timekeeping.c           | 4 ++--
2 files changed, 3 insertions(+), 3 deletions(-)
[PATCH v3] timekeeping: Use time_after() in timekeeping_check_update()
Posted by Chen Yufan 1 year, 5 months ago
Use time_after because the open coded comparison 
does not handle wrap arounds correctly.

Signed-off-by: Chen Yufan <chenyufan@vivo.com>
---
The modifications made compared to the previous version are as follows:
1. change the type of timekeeper::last_warning to
unsigned long because time_after() requires its
parameters to be unsigned long.
---
 include/linux/timekeeper_internal.h | 2 +-
 kernel/time/timekeeping.c           | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/include/linux/timekeeper_internal.h b/include/linux/timekeeper_internal.h
index 84ff2844df2a..928a9258b7dd 100644
--- a/include/linux/timekeeper_internal.h
+++ b/include/linux/timekeeper_internal.h
@@ -125,7 +125,7 @@ struct timekeeper {
 	/* Flag used to avoid updating NTP twice with same second */
 	u32			skip_second_overflow;
 #ifdef CONFIG_DEBUG_TIMEKEEPING
-	long			last_warning;
+	unsigned long			last_warning;
 	/*
 	 * These simple flag variables are managed
 	 * without locks, which is racy, but they are
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index 5391e4167d60..6cda65dbe18f 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -217,7 +217,7 @@ static void timekeeping_check_update(struct timekeeper *tk, u64 offset)
 	}
 
 	if (tk->underflow_seen) {
-		if (jiffies - tk->last_warning > WARNING_FREQ) {
+		if (time_after(jiffies, tk->last_warning + WARNING_FREQ)) {
 			printk_deferred("WARNING: Underflow in clocksource '%s' observed, time update ignored.\n", name);
 			printk_deferred("         Please report this, consider using a different clocksource, if possible.\n");
 			printk_deferred("         Your kernel is probably still fine.\n");
@@ -227,7 +227,7 @@ static void timekeeping_check_update(struct timekeeper *tk, u64 offset)
 	}
 
 	if (tk->overflow_seen) {
-		if (jiffies - tk->last_warning > WARNING_FREQ) {
+		if (time_after(jiffies, tk->last_warning + WARNING_FREQ)) {
 			printk_deferred("WARNING: Overflow in clocksource '%s' observed, time update capped.\n", name);
 			printk_deferred("         Please report this, consider using a different clocksource, if possible.\n");
 			printk_deferred("         Your kernel is probably still fine.\n");
-- 
2.39.0
Re: [PATCH v3] timekeeping: Use time_after() in timekeeping_check_update()
Posted by Thomas Gleixner 1 year, 5 months ago
On Fri, Aug 30 2024 at 09:54, Chen Yufan wrote:
> Use time_after because the open coded comparison 
> does not handle wrap arounds correctly.
>
> Signed-off-by: Chen Yufan <chenyufan@vivo.com>
> ---
> The modifications made compared to the previous version are as follows:
> 1. change the type of timekeeper::last_warning to
> unsigned long because time_after() requires its
> parameters to be unsigned long.

The change of the struct member wants to be explained in the change log
and not in the version history. I told you so, no?

Thanks,

        tglx