From patchwork Fri Jun 14 13:33:35 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Eliezer Tamir X-Patchwork-Id: 251427 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 409642C009A for ; Fri, 14 Jun 2013 23:34:08 +1000 (EST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752858Ab3FNNdo (ORCPT ); Fri, 14 Jun 2013 09:33:44 -0400 Received: from mga02.intel.com ([134.134.136.20]:7901 "EHLO mga02.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751381Ab3FNNdm (ORCPT ); Fri, 14 Jun 2013 09:33:42 -0400 Received: from orsmga002.jf.intel.com ([10.7.209.21]) by orsmga101.jf.intel.com with ESMTP; 14 Jun 2013 06:33:41 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.87,866,1363158000"; d="scan'208";a="353445729" Received: from ladj378.jer.intel.com ([10.12.232.220]) by orsmga002.jf.intel.com with ESMTP; 14 Jun 2013 06:33:36 -0700 From: Eliezer Tamir Subject: [PATCH v5 net-next 2/4] net: convert low latency sockets to sched_clock() To: David Miller Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Jesse Brandeburg , Don Skidmore , e1000-devel@lists.sourceforge.net, Willem de Bruijn , Eric Dumazet , Ben Hutchings , Andi Kleen , HPA , Eilon Greenstien , Or Gerlitz , Amir Vadai , Alex Rosenbaum , Avner Ben Hanoch , Or Kehati , sockperf-dev@googlegroups.com, Eliezer Tamir Date: Fri, 14 Jun 2013 16:33:35 +0300 Message-ID: <20130614133335.6126.8092.stgit@ladj378.jer.intel.com> In-Reply-To: <20130614133314.6126.77315.stgit@ladj378.jer.intel.com> References: <20130614133314.6126.77315.stgit@ladj378.jer.intel.com> User-Agent: StGIT/0.14.3 MIME-Version: 1.0 Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org Use sched_clock() instead of get_cycles(). We can use sched_clock() because we don't care much about accuracy. Remove the dependency on X86_TSC Signed-off-by: Eliezer Tamir --- include/net/ll_poll.h | 33 +++++++++++++++++---------------- net/Kconfig | 1 - 2 files changed, 17 insertions(+), 17 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html diff --git a/include/net/ll_poll.h b/include/net/ll_poll.h index 44e2f70..6930cbd 100644 --- a/include/net/ll_poll.h +++ b/include/net/ll_poll.h @@ -21,10 +21,6 @@ * e1000-devel Mailing List */ -/* - * For now this depends on CONFIG_X86_TSC - */ - #ifndef _LINUX_NET_LL_POLL_H #define _LINUX_NET_LL_POLL_H @@ -40,13 +36,19 @@ extern unsigned int sysctl_net_ll_poll __read_mostly; #define LL_FLUSH_FAILED -1 #define LL_FLUSH_BUSY -2 -/* we don't mind a ~2.5% imprecision */ -#define TSC_MHZ (tsc_khz >> 10) - -static inline cycles_t ll_end_time(void) +/* we can use sched_clock() because we don't care much about precision + * we only care that the average is bounded + */ +static inline u64 ll_end_time(void) { - return (cycles_t)TSC_MHZ * ACCESS_ONCE(sysctl_net_ll_poll) - + get_cycles(); + u64 end_time = ACCESS_ONCE(sysctl_net_ll_poll); + + /* we don't mind a ~2.5% imprecision + * sysctl_net_ll_poll is a u_int so this can't overflow + */ + end_time = (end_time << 10) + sched_clock(); + + return end_time; } static inline bool sk_valid_ll(struct sock *sk) @@ -55,16 +57,15 @@ static inline bool sk_valid_ll(struct sock *sk) !need_resched() && !signal_pending(current); } -static inline bool can_poll_ll(cycles_t end_time) +static inline bool can_poll_ll(u64 end_time) { - return !time_after((unsigned long)get_cycles(), - (unsigned long)end_time); + return !time_after64(sched_clock(), end_time); } static inline bool sk_poll_ll(struct sock *sk, int nonblock) { - cycles_t end_time = ll_end_time(); const struct net_device_ops *ops; + u64 end_time = ll_end_time(); struct napi_struct *napi; int rc = false; @@ -117,7 +118,7 @@ static inline void sk_mark_ll(struct sock *sk, struct sk_buff *skb) #else /* CONFIG_NET_LL_RX_POLL */ -static inline cycles_t ll_end_time(void) +static inline u64 ll_end_time(void) { return 0; } @@ -140,7 +141,7 @@ static inline void sk_mark_ll(struct sock *sk, struct sk_buff *skb) { } -static inline bool can_poll_ll(cycles_t end_time) +static inline bool can_poll_ll(u64 end_time) { return false; } diff --git a/net/Kconfig b/net/Kconfig index d6a9ce6..e591668 100644 --- a/net/Kconfig +++ b/net/Kconfig @@ -245,7 +245,6 @@ config NETPRIO_CGROUP config NET_LL_RX_POLL bool "Low Latency Receive Poll" - depends on X86_TSC default n ---help--- Support Low Latency Receive Queue Poll.