From patchwork Mon Jun 6 14:47:33 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Peter Maydell X-Patchwork-Id: 9158541 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id ABCDF60572 for ; Mon, 6 Jun 2016 15:07:51 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 9A88920410 for ; Mon, 6 Jun 2016 15:07:51 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 8CED426E5D; Mon, 6 Jun 2016 15:07:51 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-6.9 required=2.0 tests=BAYES_00,RCVD_IN_DNSWL_HI autolearn=ham version=3.3.1 Received: from lists.gnu.org (lists.gnu.org [208.118.235.17]) (using TLSv1 with cipher AES256-SHA (256/256 bits)) (No client certificate requested) by mail.wl.linuxfoundation.org (Postfix) with ESMTPS id EF06420410 for ; Mon, 6 Jun 2016 15:07:50 +0000 (UTC) Received: from localhost ([::1]:43026 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1b9w86-00009e-4d for patchwork-qemu-devel@patchwork.kernel.org; Mon, 06 Jun 2016 11:07:50 -0400 Received: from eggs.gnu.org ([2001:4830:134:3::10]:37784) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1b9voq-0008Vm-UP for qemu-devel@nongnu.org; Mon, 06 Jun 2016 10:47:59 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1b9voo-0000QS-CQ for qemu-devel@nongnu.org; Mon, 06 Jun 2016 10:47:55 -0400 Received: from orth.archaic.org.uk ([2001:8b0:1d0::2]:57503) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1b9voo-0000Oc-3V for qemu-devel@nongnu.org; Mon, 06 Jun 2016 10:47:54 -0400 Received: from pm215 by orth.archaic.org.uk with local (Exim 4.84_2) (envelope-from ) id 1b9von-0008EP-HP for qemu-devel@nongnu.org; Mon, 06 Jun 2016 15:47:53 +0100 From: Peter Maydell To: qemu-devel@nongnu.org Date: Mon, 6 Jun 2016 15:47:33 +0100 Message-Id: <1465224465-21998-17-git-send-email-peter.maydell@linaro.org> X-Mailer: git-send-email 1.9.1 In-Reply-To: <1465224465-21998-1-git-send-email-peter.maydell@linaro.org> References: <1465224465-21998-1-git-send-email-peter.maydell@linaro.org> X-detected-operating-system: by eggs.gnu.org: GNU/Linux 2.2.x-3.x [generic] X-Received-From: 2001:8b0:1d0::2 Subject: [Qemu-devel] [PULL 16/28] hw/ptimer: Fix issues caused by the adjusted timer limit value X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.21 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: qemu-devel-bounces+patchwork-qemu-devel=patchwork.kernel.org@nongnu.org Sender: "Qemu-devel" X-Virus-Scanned: ClamAV using ClamSMTP From: Dmitry Osipenko Multiple issues here related to the timer with a adjusted .limit value: 1) ptimer_get_count() returns incorrect counter value for the disabled timer after loading the counter with a small value, because adjusted limit value is used instead of the original. For instance: 1) ptimer_stop(t) 2) ptimer_set_period(t, 1) 3) ptimer_set_limit(t, 0, 1) 4) ptimer_get_count(t) <-- would return 10000 instead of 0 2) ptimer_get_count() might return incorrect value for the timer running with a adjusted limit value. For instance: 1) ptimer_stop(t) 2) ptimer_set_period(t, 1) 3) ptimer_set_limit(t, 10, 1) 4) ptimer_run(t) 5) ptimer_get_count(t) <-- might return value > 10 3) Neither ptimer_set_period() nor ptimer_set_freq() are adjusting the limit value, so it is still possible to make timer timeout value arbitrary small. For instance: 1) ptimer_set_period(t, 10000) 2) ptimer_set_limit(t, 1, 0) 3) ptimer_set_period(t, 1) <-- bypass limit correction Fix all of the above issues by adjusting timer period instead of the limit. Perform the adjustment for periodic timer only. Use the delta value instead of the limit to make decision whether adjustment is required, as limit could be altered while timer is running, resulting in incorrect value returned by ptimer_get_count. Signed-off-by: Dmitry Osipenko Reviewed-by: Peter Crosthwaite Message-id: cd141f74f5737480ec586b9c7d18cce1d69884e2.1464367869.git.digetx@gmail.com Signed-off-by: Peter Maydell --- hw/core/ptimer.c | 51 +++++++++++++++++++++++++++++++-------------------- 1 file changed, 31 insertions(+), 20 deletions(-) diff --git a/hw/core/ptimer.c b/hw/core/ptimer.c index 153c835..16d7dd5 100644 --- a/hw/core/ptimer.c +++ b/hw/core/ptimer.c @@ -35,6 +35,9 @@ static void ptimer_trigger(ptimer_state *s) static void ptimer_reload(ptimer_state *s) { + uint32_t period_frac = s->period_frac; + uint64_t period = s->period; + if (s->delta == 0) { ptimer_trigger(s); s->delta = s->limit; @@ -45,10 +48,24 @@ static void ptimer_reload(ptimer_state *s) return; } + /* + * Artificially limit timeout rate to something + * achievable under QEMU. Otherwise, QEMU spends all + * its time generating timer interrupts, and there + * is no forward progress. + * About ten microseconds is the fastest that really works + * on the current generation of host machines. + */ + + if (s->enabled == 1 && (s->delta * period < 10000) && !use_icount) { + period = 10000 / s->delta; + period_frac = 0; + } + s->last_event = s->next_event; - s->next_event = s->last_event + s->delta * s->period; - if (s->period_frac) { - s->next_event += ((int64_t)s->period_frac * s->delta) >> 32; + s->next_event = s->last_event + s->delta * period; + if (period_frac) { + s->next_event += ((int64_t)period_frac * s->delta) >> 32; } timer_mod(s->timer, s->next_event); } @@ -83,6 +100,13 @@ uint64_t ptimer_get_count(ptimer_state *s) uint64_t div; int clz1, clz2; int shift; + uint32_t period_frac = s->period_frac; + uint64_t period = s->period; + + if ((s->enabled == 1) && !use_icount && (s->delta * period < 10000)) { + period = 10000 / s->delta; + period_frac = 0; + } /* We need to divide time by period, where time is stored in rem (64-bit integer) and period is stored in period/period_frac @@ -95,7 +119,7 @@ uint64_t ptimer_get_count(ptimer_state *s) */ rem = s->next_event - now; - div = s->period; + div = period; clz1 = clz64(rem); clz2 = clz64(div); @@ -104,13 +128,13 @@ uint64_t ptimer_get_count(ptimer_state *s) rem <<= shift; div <<= shift; if (shift >= 32) { - div |= ((uint64_t)s->period_frac << (shift - 32)); + div |= ((uint64_t)period_frac << (shift - 32)); } else { if (shift != 0) - div |= (s->period_frac >> (32 - shift)); + div |= (period_frac >> (32 - shift)); /* Look at remaining bits of period_frac and round div up if necessary. */ - if ((uint32_t)(s->period_frac << shift)) + if ((uint32_t)(period_frac << shift)) div += 1; } counter = rem / div; @@ -182,19 +206,6 @@ void ptimer_set_freq(ptimer_state *s, uint32_t freq) count = limit. */ void ptimer_set_limit(ptimer_state *s, uint64_t limit, int reload) { - /* - * Artificially limit timeout rate to something - * achievable under QEMU. Otherwise, QEMU spends all - * its time generating timer interrupts, and there - * is no forward progress. - * About ten microseconds is the fastest that really works - * on the current generation of host machines. - */ - - if (!use_icount && limit * s->period < 10000 && s->period) { - limit = 10000 / s->period; - } - s->limit = limit; if (reload) s->delta = limit;