From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754557AbdKIWa5 (ORCPT ); Thu, 9 Nov 2017 17:30:57 -0500 Received: from mail-oi0-f66.google.com ([209.85.218.66]:45633 "EHLO mail-oi0-f66.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751225AbdKIWaz (ORCPT ); Thu, 9 Nov 2017 17:30:55 -0500 X-Google-Smtp-Source: AGs4zMb2v9RerTaw5/L5VINfnuo0a8XPu+NSDULXzhSxuRTGkWLw9rI62vOoGPCLCXlYb2acJIM6hAth0F+9GS2kOoY= MIME-Version: 1.0 In-Reply-To: <6583ed1f-3ea3-c7fd-7e69-1430c8387e54@intel.com> References: <20171109103814.70688-1-chao.wang@ucloud.cn> <6583ed1f-3ea3-c7fd-7e69-1430c8387e54@intel.com> From: "Rafael J. Wysocki" Date: Thu, 9 Nov 2017 23:30:54 +0100 X-Google-Sender-Auth: BuD2celQ04k6LRHNPL2se0g01bQ Message-ID: Subject: Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again To: Linus Torvalds Cc: WANG Chao , Linux Kernel Mailing List , Thomas Gleixner , Ingo Molnar , "H. Peter Anvin" , Vikas Shivappa , Kate Stewart , Len Brown , Greg Kroah-Hartman , Philippe Ombredanne , Mathias Krause , "the arch/x86 maintainers" , "Rafael J. Wysocki" , Linux PM , "Rafael J. Wysocki" Content-Type: text/plain; charset="UTF-8" Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Nov 9, 2017 at 5:06 PM, Rafael J. Wysocki wrote: > Hi Linus, > > On 11/9/2017 11:38 AM, WANG Chao wrote: >> >> Commit 941f5f0f6ef5 (x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo) caused >> a serious performance issue when reading from /proc/cpuinfo on system >> with aperfmperf. >> >> For each cpu, arch_freq_get_on_cpu() sleeps 20ms to get its frequency. >> On a system with 64 cpus, it takes 1.5s to finish running `cat >> /proc/cpuinfo`, while it previously was done in 15ms. > > Honestly, I'm not sure what to do to address this ATM. > > The last requested frequency is only available in the non-HWP case, so it > cannot be used universally. OK, here's an idea. c_start() can run aperfmperf_snapshot_khz() on all CPUs upfront (say in parallel), then wait for a while (say 5 ms; the current 20 ms wait is overkill) and then aperfmperf_snapshot_khz() can be run once on each CPU in show_cpuinfo() without taking the "stale cache" threshold into account. I'm going to try that and see how far I can get with it. Thanks, Rafael