linux-kernel - [PATCH 2/8] timekeeping: optimize ns_to

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

Date:   Fri,  8 Nov 2019 21:34:25 +0100
From:   Arnd Bergmann <arnd@...db.de>
To:     y2038@...ts.linaro.org, John Stultz <john.stultz@...aro.org>,
        Thomas Gleixner <tglx@...utronix.de>
Cc:     linux-kernel@...r.kernel.org, Arnd Bergmann <arnd@...db.de>,
        Stephen Boyd <sboyd@...nel.org>,
        Deepa Dinamani <deepa.kernel@...il.com>
Subject: [PATCH 2/8] timekeeping: optimize ns_to_timespec64

I noticed that ns_to_timespec64() calls div_s64_rem(), which is
a rather slow function on 32-bit architectures, as it cannot take
advantage of the do_div() optimizations for constant arguments.

This open-codes the div_s64_rem() function here, so we can pass
a constant divider into the optimized div_u64_rem() function.

Signed-off-by: Arnd Bergmann <arnd@...db.de>
---
 kernel/time/time.c | 21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)

diff --git a/kernel/time/time.c b/kernel/time/time.c
index 5c54ca632d08..45a358953f09 100644
--- a/kernel/time/time.c
+++ b/kernel/time/time.c
@@ -550,18 +550,21 @@ EXPORT_SYMBOL(set_normalized_timespec64);
  */
 struct timespec64 ns_to_timespec64(const s64 nsec)
 {
-	struct timespec64 ts;
+	struct timespec64 ts = { 0, 0 };
 	s32 rem;
 
-	if (!nsec)
-		return (struct timespec64) {0, 0};
-
-	ts.tv_sec = div_s64_rem(nsec, NSEC_PER_SEC, &rem);
-	if (unlikely(rem < 0)) {
-		ts.tv_sec--;
-		rem += NSEC_PER_SEC;
+	if (likely(nsec > 0)) {
+		ts.tv_sec = div_u64_rem(nsec, NSEC_PER_SEC, &rem);
+		ts.tv_nsec = rem;
+	} else if (nsec < 0) {
+		/*
+		 * With negative times, tv_sec points to the earlier
+		 * second, and tv_nsec counts the nanoseconds since
+		 * then, so tv_nsec is always a positive number.
+		 */
+		ts.tv_sec = -div_u64_rem(-nsec - 1, NSEC_PER_SEC, &rem) - 1;
+		ts.tv_nsec = NSEC_PER_SEC - rem - 1;
 	}
-	ts.tv_nsec = rem;
 
 	return ts;
 }
-- 
2.20.0