Skip to contentSkip to conversation

timekeeping: Use READ_ONCE/WRITE_ONCE for xtime_sec to prevent tearing

PATCH Applied Dennis Moshegov 2 participants 2 messages Jul 24 – Aug 10 Timekeeping, clocksource core, NTP, alarmtimer

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 timekeeping: Use READ_ONCE/WRITE_ONCE for xtime_sec to prevent tearing

The timekeeper update path uses a bulk memcpy() to synchronize the timekeeper structure, which is not guaranteed to be atomic. This allows for torn reads in ktime_get_real_seconds() on 64-bit systems, where the sequence counter protection is bypassed for performance.

To prevent reading a torn 64-bit xtime_sec value, enforce atomic-like access by using WRITE_ONCE() for the critical field before the bulk memcpy() in timekeeping_update_from_shadow(). Correspondingly, use READ_ONCE() in ktime_get_real_seconds() to ensure a fresh, consistent load from memory.

Reported-by [email protected]
---
 kernel/time/timekeeping.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
kernel/time/timekeeping.c +4 -3
index b1b5ec43c0f2..bb1c020c0e8b 100644
@@ -832,6 +832,7 @@ static void timekeeping_update_from_shadow(struct tk_data *tkd, unsigned int act
 	 * the cacheline optimized data layout of the timekeeper and requires
 	 * another indirection.
 	 */
+	WRITE_ONCE(tkd->timekeeper.xtime_sec, tk->xtime_sec);
 	memcpy(&tkd->timekeeper, tk, sizeof(*tk));
 	write_seqcount_end(&tkd->seq);
 }
@@ -1159,11 +1160,11 @@ time64_t ktime_get_real_seconds(void)
 	unsigned int seq;
 
 	if (IS_ENABLED(CONFIG_64BIT))
-		return tk->xtime_sec;
+		return READ_ONCE(tk->xtime_sec);
 
 	do {
 		seq = read_seqcount_begin(&tk_core.seq);
-		seconds = tk->xtime_sec;
+		seconds = READ_ONCE(tk->xtime_sec);
 
 	} while (read_seqcount_retry(&tk_core.seq, seq));
 
@@ -1185,7 +1186,7 @@ noinstr time64_t __ktime_get_real_seconds(void)
 {
 	struct timekeeper *tk = &tk_core.timekeeper;
 
-	return tk->xtime_sec;
+	return READ_ONCE(tk->xtime_sec);
 }
 
 static inline u64 tk_clock_read_snapshot(const struct tk_read_base *tkr,
Signature · 1 line
2.53.0.windows.1
#1 lore raw

[tip: timers/core] timekeeping: Use READ_ONCE/WRITE_ONCE() for xtime_sec to prevent tearing

The following commit has been merged into the timers/core branch of tip:

The following commit has been merged into the timers/core branch of tip:

Commit-ID:     d7fc133bf91f713df72facebd41ae9dfc83e35f7
Gitweb:        https://git.kernel.org/tip/d7fc133bf91f713df72facebd41ae9dfc83e35f7
Author:        Dennis Moshegov <[email protected]>
AuthorDate:    Fri, 24 Jul 2026 16:43:55 +01:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Mon, 10 Aug 2026 17:52:19 +02:00

timekeeping: Use READ_ONCE/WRITE_ONCE() for xtime_sec to prevent tearing

The timekeeper update path uses a bulk memcpy() to synchronize the timekeeper structure, which is not guaranteed to be atomic. This allows for torn reads in ktime_get_real_seconds() on 64-bit systems, where the sequence counter protection is bypassed for performance.

To prevent reading a torn 64-bit xtime_sec value, enforce atomic-like access by using WRITE_ONCE() for the critical field before the bulk memcpy() in timekeeping_update_from_shadow(). Correspondingly, use READ_ONCE() in ktime_get_real_seconds() to ensure a fresh, consistent load from memory.

[ tglx: Format changelog and add comment ]

---
 kernel/time/timekeeping.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
kernel/time/timekeeping.c +7 -3
index 97db2e9..c4230f4 100644
@@ -858,7 +858,11 @@ static void timekeeping_update_from_shadow(struct tk_data *tkd, unsigned int act
 	 * the downside that the reader side does not longer benefit from
 	 * the cacheline optimized data layout of the timekeeper and requires
 	 * another indirection.
+	 *
+	 * Write xtime_sec first so that even if the memcpy() tears the store
+	 * data integrity is provided for ktime_get_real_seconds().
 	 */
+	WRITE_ONCE(tkd->timekeeper.xtime_sec, tk->xtime_sec);
 	memcpy(&tkd->timekeeper, tk, sizeof(*tk));
 	write_seqcount_end(&tkd->seq);
 }
@@ -1186,11 +1190,11 @@ time64_t ktime_get_real_seconds(void)
 	unsigned int seq;
 
 	if (IS_ENABLED(CONFIG_64BIT))
-		return tk->xtime_sec;
+		return READ_ONCE(tk->xtime_sec);
 
 	do {
 		seq = read_seqcount_begin(&tk_core.seq);
-		seconds = tk->xtime_sec;
+		seconds = READ_ONCE(tk->xtime_sec);
 
 	} while (read_seqcount_retry(&tk_core.seq, seq));
 
@@ -1212,7 +1216,7 @@ noinstr time64_t __ktime_get_real_seconds(void)
 {
 	struct timekeeper *tk = &tk_core.timekeeper;
 
-	return tk->xtime_sec;
+	return READ_ONCE(tk->xtime_sec);
 }
 
 static inline u64 tk_clock_read_snapshot(const struct tk_read_base *tkr,