From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755682Ab0IBRpO (ORCPT ); Thu, 2 Sep 2010 13:45:14 -0400 Received: from n1.taur.dk ([217.198.219.102]:33875 "EHLO n1.taur.dk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750761Ab0IBRpM (ORCPT ); Thu, 2 Sep 2010 13:45:12 -0400 X-Greylist: delayed 3051 seconds by postgrey-1.27 at vger.kernel.org; Thu, 02 Sep 2010 13:45:10 EDT Message-ID: <4C7FD6B8.7080401@kasperkp.dk> Date: Thu, 02 Sep 2010 18:54:16 +0200 From: Kasper Pedersen User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.9) Gecko/20100430 Fedora/3.0.4-2.fc11 Thunderbird/3.0.4 MIME-Version: 1.0 To: Thomas Gleixner CC: John Stultz , linux-kernel@vger.kernel.org Subject: [PATCH] time: compensate for rounding on odd-frequency clocksources Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org When the clocksource is not a multiple of HZ, the clock will be off. For acpi_pm, HZ=1000 the error is 127.111 ppm: The rounding of cycle_interval ends up generating a false error term in ntp_error accumulation since xtime_interval is not exactly 1/HZ. So, we subtract out the error caused by the rounding. This has been visible since 2.6.32-rc2 commit a092ff0f90cae22b2ac8028ecd2c6f6c1a9e4601 time: Implement logarithmic time accumulation That commit raised NTP_INTERVAL_FREQ and exposed the rounding error. testing tool: http://n1.taur.dk/permanent/testpmt.c Also tested with ntpd and a frequency counter. patch against 2.6.36-rc3-next. Signed-off-by: Kasper Pedersen --- kernel/time/timekeeping.c | 9 +++++++-- 1 files changed, 7 insertions(+), 2 deletions(-) diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c index 77e930d..5ee5dea 100644 --- a/kernel/time/timekeeping.c +++ b/kernel/time/timekeeping.c @@ -32,6 +32,8 @@ struct timekeeper { cycle_t cycle_interval; /* Number of clock shifted nano seconds in one NTP interval. */ u64 xtime_interval; + /* shifted nano seconds left over when rounding cycle_interval */ + s64 xtime_remainder; /* Raw nano seconds accumulated per NTP interval. */ u32 raw_interval; @@ -62,7 +64,7 @@ struct timekeeper timekeeper; static void timekeeper_setup_internals(struct clocksource *clock) { cycle_t interval; - u64 tmp; + u64 tmp, ntpinterval; timekeeper.clock = clock; clock->cycle_last = clock->read(clock); @@ -70,6 +72,7 @@ static void timekeeper_setup_internals(struct clocksource *clock) /* Do the ns -> cycle conversion first, using original mult */ tmp = NTP_INTERVAL_LENGTH; tmp <<= clock->shift; + ntpinterval = tmp; tmp += clock->mult/2; do_div(tmp, clock->mult); if (tmp == 0) @@ -80,6 +83,7 @@ static void timekeeper_setup_internals(struct clocksource *clock) /* Go back from cycles -> shifted ns */ timekeeper.xtime_interval = (u64) interval * clock->mult; + timekeeper.xtime_remainder = ntpinterval - timekeeper.xtime_interval; timekeeper.raw_interval = ((u64) interval * clock->mult) >> clock->shift; @@ -746,7 +750,8 @@ static cycle_t logarithmic_accumulation(cycle_t offset, int shift) /* Accumulate error between NTP and clock interval */ timekeeper.ntp_error += tick_length << shift; - timekeeper.ntp_error -= timekeeper.xtime_interval << + timekeeper.ntp_error -= + (timekeeper.xtime_interval + timekeeper.xtime_remainder) << (timekeeper.ntp_error_shift + shift); return offset;