From: Daniel Walker Subject: [PATCH -rt] cycles_to_usecs rounding fix Date: Wed, 12 Dec 2007 12:15:22 -0800 Message-ID: <20071212201522.061896981@mvista.com> Cc: mingo@elte.hu, linux-kernel@vger.kernel.org, linux-rt-users@vger.kernel.org, johnstul@us.ibm.com To: rostedt@goodmis.org Return-path: Received: from homer.mvista.com ([63.81.120.158]:16422 "EHLO dwalker1.mvista.com" rhost-flags-OK-FAIL-OK-FAIL) by vger.kernel.org with ESMTP id S1754585AbXLLU1i (ORCPT ); Wed, 12 Dec 2007 15:27:38 -0500 Message-Id: <20071212201258.890003246@mvista.com> Date: Wed, 12 Dec 2007 12:12:58 -0800 Sender: linux-rt-users-owner@vger.kernel.org List-ID: do_div() rounds down, so we add have the divisor to round up. This effected my change to preempt_max_latency. Each time you read preempt_max_latency it gets rounded lower. Signed-off-by: Daniel Walker --- kernel/time/timekeeping.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) Index: linux-2.6.23/kernel/time/timekeeping.c =================================================================== --- linux-2.6.23.orig/kernel/time/timekeeping.c +++ linux-2.6.23/kernel/time/timekeeping.c @@ -144,7 +144,8 @@ unsigned long notrace cycles_to_usecs(cy { u64 ret = cyc2ns(clock, cycles); - do_div(ret, 1000); + ret += NSEC_PER_USEC/2; /* For rounding in do_div() */ + do_div(ret, NSEC_PER_USEC); return ret; } -- --