linaro-dev August 2014

linaro-dev@lists.linaro.org

27 participants
43 discussions

[PATCH] devfreq: add devfreq devices stats support

by Sanjay Singh Rawat

add window to show frequency stats for devfreq devices Signed-off-by: Rajagopal Venkat <rajagopal.venkat(a)gmail.com> Signed-off-by: Sanjay Singh Rawat <sanjay.rawat(a)linaro.org> --- src/Makefile.am | 2 +- src/devices/devfreq.cpp | 351 ++++++++++++++++++++++++++++++++++++++++++++++ src/devices/devfreq.h | 75 ++++++++++ src/main.cpp | 9 ++ src/report/report-maker.h | 1 + 5 files changed, 437 insertions(+), 1 deletion(-) create mode 100644 src/devices/devfreq.cpp create mode 100644 src/devices/devfreq.h diff --git a/src/Makefile.am b/src/Makefile.am index 6886388..fee1ffa 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -32,7 +32,7 @@ powertop_SOURCES = parameters/persistent.cpp parameters/learn.cpp parameters/par report/report-formatter-csv.cpp report/report-formatter-csv.h \ report/report-formatter-html.cpp report/report-formatter-html.h \ report/report-data-html.cpp report/report-data-html.h \ - main.cpp css.h powertop.css cpu/intel_gpu.cpp \ + main.cpp css.h powertop.css cpu/intel_gpu.cpp devices/devfreq.cpp \ cpu/rapl/rapl_interface.cpp cpu/cpu_rapl_device.cpp cpu/rapl/rapl_interface.h\ cpu/dram_rapl_device.cpp devices/gpu_rapl_device.cpp cpu/cpu_rapl_device.h \ cpu/dram_rapl_device.h devices/gpu_rapl_device.h diff --git a/src/devices/devfreq.cpp b/src/devices/devfreq.cpp new file mode 100644 index 0000000..29d7790 --- /dev/null +++ b/src/devices/devfreq.cpp @@ -0,0 +1,351 @@ +/* + * Copyright 2012, Linaro + * + * This file is part of PowerTOP + * + * This program file is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License as published by the + * Free Software Foundation; version 2 of the License. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program in a file named COPYING; if not, write to the + * Free Software Foundation, Inc, + * 51 Franklin Street, Fifth Floor, + * Boston, MA 02110-1301 USA + * or just google for it. + * + * Authors: + * Rajagopal Venkat <rajagopal.venkat(a)linaro.org> + */ + +#include <iostream> +#include <fstream> + +#include <dirent.h> +#include <stdlib.h> +#include <time.h> +#include <unistd.h> + +#include "device.h" +#include "devfreq.h" +#include "../display.h" +#include "../cpu/cpu.h" +#include "../report/report.h" +#include "../report/report-maker.h" + +static bool is_enabled = true; + +static vector<class devfreq *> all_devfreq; + +devfreq::devfreq(const char* dpath): device() +{ + strncpy(dir_name, dpath, sizeof(dir_name)); +} + +uint64_t devfreq::parse_freq_time(char* pchr) +{ + char *cptr, *pptr = pchr; + uint64_t ctime; + + cptr = strtok(pchr, " :"); + while (cptr != NULL) { + cptr = strtok(NULL, " :"); + if (cptr ) + pptr = cptr; + } + + ctime = strtoull(pptr, NULL, 10); + return ctime; +} + +void devfreq::process_time_stamps() +{ + unsigned int i; + uint64_t active_time = 0; + + sample_time = (1000000.0 * (stamp_after.tv_sec - stamp_before.tv_sec)) + + ((stamp_after.tv_usec - stamp_before.tv_usec) ); + + for (i=0; i < dstates.size()-1; i++) { + struct frequency *state = dstates[i]; + state->time_after = 1000 * (state->time_after - state->time_before); + active_time += state->time_after; + } + /* Compute idle time for the device */ + dstates[i]->time_after = sample_time - active_time; +} + +void devfreq::add_devfreq_freq_state(uint64_t freq, uint64_t time) +{ + struct frequency *state; + + state = new(std::nothrow) struct frequency; + if (!state) + return; + + memset(state, 0, sizeof(*state)); + dstates.push_back(state); + + state->freq = freq; + if (freq == 0) + strcpy(state->human_name, "Idle"); + else + hz_to_human(freq, state->human_name); + state->time_before = time; +} + +void devfreq::update_devfreq_freq_state(uint64_t freq, uint64_t time) +{ + unsigned int i; + struct frequency *state = NULL; + + for(i=0; i < dstates.size(); i++) { + if (freq == dstates[i]->freq) + state = dstates[i]; + } + + if (state == NULL) { + add_devfreq_freq_state(freq, time); + return; + } + + state->time_after = time; +} + +void devfreq::parse_devfreq_trans_stat(char *dname) +{ + ifstream file; + char filename[256]; + + sprintf(filename, "/sys/class/devfreq/%s/trans_stat", dir_name); + file.open(filename); + + if (!file) + return; + + char line[1024]; + char *c; + + while (file) { + uint64_t freq; + uint64_t time; + char *pchr; + + memset(line, 0, sizeof(line)); + file.getline(line, sizeof(line)); + + pchr = strchr(line, '*'); + pchr = (pchr != NULL) ? pchr+1 : line; + + freq = strtoull(pchr, &c, 10); + if (!freq) + continue; + + time = parse_freq_time(pchr); + update_devfreq_freq_state(freq, time); + } + file.close(); +} + +void devfreq::start_measurement(void) +{ + unsigned int i; + ifstream file; + + for (i=0; i < dstates.size(); i++) + delete dstates[i]; + dstates.resize(0); + sample_time = 0; + + gettimeofday(&stamp_before, NULL); + parse_devfreq_trans_stat(dir_name); + /* add device idle state */ + update_devfreq_freq_state(0, 0); +} + +void devfreq::end_measurement(void) +{ + parse_devfreq_trans_stat(dir_name); + gettimeofday(&stamp_after, NULL); + process_time_stamps(); +} + +double devfreq::power_usage(struct result_bundle *result, struct parameter_bundle *bundle) +{ + return 0; +} + +double devfreq::utilization(void) +{ + return 0; +} + +void devfreq::fill_freq_utilization(unsigned int idx, char *buf) +{ + buf[0] = 0; + + if (idx < dstates.size() && dstates[idx]) { + struct frequency *state = dstates[idx]; + sprintf(buf, " %5.1f%% ", percentage(1.0 * state->time_after / sample_time)); + } +} + +void devfreq::fill_freq_name(unsigned int idx, char *buf) +{ + buf[0] = 0; + + if (idx < dstates.size() && dstates[idx]) { + sprintf(buf, "%-15s", dstates[idx]->human_name); + } +} + +void start_devfreq_measurement(void) +{ + unsigned int i; + + for (i=0; i<all_devfreq.size(); i++) + all_devfreq[i]->start_measurement(); +} + +void end_devfreq_measurement(void) +{ + unsigned int i; + + for (i=0; i<all_devfreq.size(); i++) + all_devfreq[i]->end_measurement(); +} + +static void devfreq_dev_callback(const char *d_name) +{ + devfreq *df = new(std::nothrow) class devfreq(d_name); + if (df) + all_devfreq.push_back(df); +} + +void create_all_devfreq_devices(void) +{ + DIR *dir; + std::string p = "/sys/class/devfreq/"; + dir = opendir(p.c_str()); + if (dir == NULL) { + fprintf(stderr, "Devfreq not enabled\n"); + is_enabled = false; + return; + } + + callback fn = &devfreq_dev_callback; + process_directory(p.c_str(), fn); +} + +void initialize_devfreq(void) +{ + if (is_enabled) + create_tab("Device Freq stats", _("Device Freq stats")); +} + +void display_devfreq_devices(void) +{ + unsigned int i, j; + WINDOW *win; + char fline[1024]; + char buf[128]; + + win = get_ncurses_win("Device Freq stats"); + if (!win) + return; + + wclear(win); + wmove(win, 2,0); + + if (!is_enabled) { + wprintw(win, _(" Devfreq is not enabled")); + return; + } + + if (!all_devfreq.size()) { + wprintw(win, _(" No devfreq devices available")); + return; + } + + for (i=0; i<all_devfreq.size(); i++) { + + class devfreq *df = all_devfreq[i]; + wprintw(win, "\n%s\n", df->device_name()); + + for(j=0; j < df->dstates.size(); j++) { + memset(fline, 0, sizeof(fline)); + strcpy(fline, "\t"); + df->fill_freq_name(j, buf); + strcat(fline, buf); + df->fill_freq_utilization(j, buf); + strcat(fline, buf); + strcat(fline, "\n"); + wprintw(win, fline); + } + wprintw(win, "\n"); + } +} + +void report_devfreq_devices(void) +{ + if (!is_enabled) { + return; + } + +/* todo: adapt to new report format */ +#if 0 + char buffer[512]; + unsigned int i, j; + + report.begin_section(SECTION_DEVFREQ); + report.add_header("Device Frequency Report"); + + report.begin_table(TABLE_WIDE); + if (!all_devfreq.size()) { + report.begin_row(); + report.add(" No devfreq devices available"); + return; + } + + for (i = 0; i < all_devfreq.size(); i++) { + buffer[0] = 0; + class devfreq *df = all_devfreq[i]; + + report.begin_row(); + report.begin_cell(CELL_CPU_PSTATE_HEADER); + report.addf("%s", df->device_name()); + + for (j = 0; j < df->dstates.size(); j++) { + report.begin_row(); + report.begin_cell(CELL_CPU_STATE_VALUE); + df->fill_freq_name(j, buffer); + report.add(buffer); + report.begin_cell(CELL_CPU_STATE_VALUE); + df->fill_freq_utilization(j, buffer); + report.add(buffer); + } + } +#endif + +} + +void clear_all_devfreq() +{ + unsigned int i, j; + + for (i=0; i < all_devfreq.size(); i++) { + class devfreq *df = all_devfreq[i]; + + for(j=0; j < df->dstates.size(); j++) + delete df->dstates[j]; + + df->dstates.resize(0); + delete df; + } + all_devfreq.clear(); +} diff --git a/src/devices/devfreq.h b/src/devices/devfreq.h new file mode 100644 index 0000000..8ab5705 --- /dev/null +++ b/src/devices/devfreq.h @@ -0,0 +1,75 @@ +/* + * Copyright 2012, Linaro + * + * This file is part of PowerTOP + * + * This program file is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License as published by the + * Free Software Foundation; version 2 of the License. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program in a file named COPYING; if not, write to the + * Free Software Foundation, Inc, + * 51 Franklin Street, Fifth Floor, + * Boston, MA 02110-1301 USA + * or just google for it. + * + * Authors: + * Rajagopal Venkat <rajagopal.venkat(a)linaro.org> + */ +#ifndef _INCLUDE_GUARD_DEVFREQ_H +#define _INCLUDE_GUARD_DEVFREQ_H + +#include "device.h" +#include "../parameters/parameters.h" + +struct frequency; + +class devfreq: public device { + char dir_name[4096]; + struct timeval stamp_before, stamp_after; + double sample_time; + + uint64_t parse_freq_time(char *ptr); + void add_devfreq_freq_state(uint64_t freq, uint64_t time); + void update_devfreq_freq_state(uint64_t freq, uint64_t time); + void parse_devfreq_trans_stat(char *dname); + void process_time_stamps(); + +public: + + vector<struct frequency *> dstates; + + devfreq(const char *c); + void fill_freq_utilization(unsigned int idx, char *buf); + void fill_freq_name(unsigned int idx, char *buf); + + virtual void start_measurement(void); + virtual void end_measurement(void); + + virtual double utilization(void); /* percentage */ + + virtual const char * class_name(void) { return "devfreq";}; + + virtual const char * device_name(void) { return dir_name;}; + virtual const char * human_name(void) { return "devfreq";}; + virtual double power_usage(struct result_bundle *result, struct parameter_bundle *bundle); + virtual const char * util_units(void) { return " rpm"; }; + virtual int power_valid(void) { return 0; /*utilization_power_valid(r_index);*/}; + virtual int grouping_prio(void) { return 1; }; +}; + +extern void create_all_devfreq_devices(void); +extern void clear_all_devfreq(void); +extern void display_devfreq_devices(void); +extern void report_devfreq_devices(void); +extern void initialize_devfreq(void); +extern void start_devfreq_measurement(void); +extern void end_devfreq_measurement(void); + +#endif diff --git a/src/main.cpp b/src/main.cpp index cf4e547..d33eaed 100644 --- a/src/main.cpp +++ b/src/main.cpp @@ -48,6 +48,7 @@ #include "devices/device.h" +#include "devices/devfreq.h" #include "devices/usb.h" #include "devices/ahci.h" #include "measurement/measurement.h" @@ -194,6 +195,7 @@ void one_measurement(int seconds, char *workload) create_all_usb_devices(); start_power_measurement(); devices_start_measurement(); + start_devfreq_measurement(); start_process_measurement(); start_cpu_measurement(); @@ -206,6 +208,7 @@ void one_measurement(int seconds, char *workload) end_cpu_measurement(); end_process_measurement(); collect_open_devices(); + end_devfreq_measurement(); devices_end_measurement(); end_power_measurement(); @@ -233,6 +236,9 @@ void one_measurement(int seconds, char *workload) report_show_open_devices(); report_devices(); + display_devfreq_devices(); + report_devfreq_devices(); + ahci_create_device_stats_table(); store_results(measurement_time); end_cpu_data(); @@ -344,6 +350,7 @@ static void powertop_init(void) enumerate_cpus(); create_all_devices(); + create_all_devfreq_devices(); detect_power_meters(); register_parameter("base power", 100, 0.5); @@ -457,6 +464,7 @@ int main(int argc, char **argv) exit(0); } init_display(); + initialize_devfreq(); initialize_tuning(); /* first one is short to not let the user wait too long */ one_measurement(1, NULL); @@ -491,6 +499,7 @@ int main(int argc, char **argv) clean_open_devices(); clear_all_devices(); + clear_all_devfreq(); clear_all_cpus(); return 0; diff --git a/src/report/report-maker.h b/src/report/report-maker.h index 423568a..bda4cef 100644 --- a/src/report/report-maker.h +++ b/src/report/report-maker.h @@ -89,6 +89,7 @@ enum section_type { SECTION_SYSINFO, SECTION_CPUIDLE, SECTION_CPUFREQ, + SECTION_DEVFREQ, SECTION_DEVPOWER, SECTION_SOFTWARE, SECTION_SUMMARY, -- 1.8.3.2

8 years, 11 months

[ACTIVITY] Week 34

by Bernhard Rosenkränzer

== Progress == * Fixed build failures caused by 2014.08 based Android toolchains * Rewrote NDK build scripts to build everything from source instead of patching the upstream release (allowing us to build from master branch) * Created reduced test cases for compiler issues * Worked on cbuild2 toolchain releases == Plans == * cbuild2 toolchain releases * NDK testing == Issues == * A cbuild2 showstopper bug seems to not be reproducible by TCWG

9 years

[ANN] Reminder about recommended Linaro git access procedure

by Paul Sokolovsky

Hello, Recently, we had DoS-like episodes on the main Linaro git server, http://git.linaro.org , which affected number of Linaro users, including users of Gerrit system, http://review.linaro.org . These episodes were related to unfriendly usage of native protocol, git:// (service port 9418). The implementation of this protocol is known to be resource-hungry and not scale to many connections and users. The issue itself is not new, it is something which affected us in waves over last 3 years, and a resolution for which was established a year ago, providing 2 HTTP-based protocols (so called "dump" and "smart" protocols) as more scalable replacement. So, this is a gentle reminder that use of git:// protocol by is discouraged for Linaro engineers, and completely unsupported(*1) for third parties. Based on the analysis and outcome of the current DoS-like activity, we may need to make git:// access more limited and strict. So, please kindly: 1. Check URLs you use for cloning and updating your local trees. If you use "ssh://" or "http(s)://" protocols, you're ok. If you use git://, please switch to using http-based protocol instead. In most cases, this requires just replacing "git://" schema with "http://". If in doubt, please visit gitweb page for your repositories, which lists all supported URLS to clone a repository, e.g.: https://git.linaro.org/arm/arm-trusted-firmware.git 2. If you set up of oversee CI or automated build jobs, please audit and apply similar changes to them. Thanks, Paul, on behalf of Linaro Systems Team and ITS (*1) Unsupported in the current context means that "git://" URLs are not published in up-to-date information, and there's no warranty that any 3rd party will be able to complete a clone successfully using this protocol. Linaro.org | Open source software for ARM SoCs Follow Linaro: http://www.facebook.com/pages/Linaro http://twitter.com/#!/linaroorg - http://www.linaro.org/linaro-blog

9 years

can I flush the icache on armv8 from user mode

by Edward Nevill

Hi, We have a need on OpenJDK to flush the entire icache. Basically this need occurs when we do a garbage collection which may include garbage collection and patching of the dynamically generated code. Problem is, the IC IALLUIS instruction which we need is only executable from EL1 or greater. Is there any kernel hook to flush then entire icache. Thanks for your help, Ed.

9 years

[ACTIVITY] Week 35

by Harigopal Gollamudi

== Progress == 1. CARD-1216: CI bring up: member build for TI-Vayu J6 platform is on Hold 2. Integrated Mali drivers to Nexus 10 android build. == Plans == Need to take up new assignment == Issues == Non e -- Thanks and Regards, HariGopal G Android Engineer

9 years

[ACTIVITY] Week 35

by Andrey Konovalov

== Progress == * vacations (99%) * linux-linaro-core-tracking moved to v3.17-rc2, but the Android topic isn't there yet (CARD-788, 1%) == Plans == * vacations (20%) * move linux-linaro tree to 3.17, add the 2014.08 Android topic back to linux-linaro[-core-tracking] (CARD-788) * sanity check the ILP32 topic branch for ll using LTP and LP64 model, identify the missing components needed for testing in ILP32 (CARD-1161)

9 years

[ACTIVITY] Week 34

by Andrey Konovalov

== Progress == * 2014.08 ll kernel release (CARD-788, 50%) * fixed the linux-linaro kernel CI job for ifc6410: CI builds boot OK now (CARD-788, 50%) == Plans == * vacations * move linux-linaro-core-tracking and linux-linaro trees to 3.17 (CARD-788) * sanity check the ILP32 topic branch for ll using LTP and LP64 model, identify the missing components needed for testing in ILP32 (CARD-1161)

9 years

[ACTIVITY] Week 33

by Andrey Konovalov

== Progress == * few more updates to linux-linaro-core-tracking and linux-linaro (CARD-788, 20%) * build failures for ifc6410 builds based on ll tree and multi_v7_defconfig: root cause found (CARD-788, 60%) * created linux-linaro kernel CI job variant using git based job configuration. Added linux-linaro-tracking-ll.yaml to ci/job/configs.git, but jenkins is not using it yet (CARD-788, 20%) == Plans == * 2014.08 ll kernel release (CARD-788) * sanity check the ILP32 topic branch for ll using LTP and LP64 model, identify the missing components needed for testing in ILP32 (CARD-1161)

9 years

[ACTIVITY] Weeks33-35

by Wookey

[Been awful busy and got behind on these - 3 weeks-worth - ignore spurious repeat of week32] Debian-ports buildd maintenance (e.g perl 5.20 transition) Now at 87% built. qt4-x11 patches forward-ported, tested, upstreamed and uploaded DebBug #735488 (DEVPLAT-226) OpenJade autoconf updated so it builds, DebBug #748626 (DEVPLAT-226) Notmuch gdb issue fixed Debbug #757454 (DEVPLAT-221 5%) Keyutils build fixed: DebBug #757530( DEVPLAT-221 4%) Debian porterbox lvm+chroots setup (DEVPLAT-226: 5%) Porterbox accounts/keys for mono, libreoffice, qt + few other maintainers created Debian main buildds setup (with Steve McIntyre) (at 48% after 2 weeks) (CARD-1586) Bootstrap sprint hotel booking for all Bootstrap Sprint, Paris (4 days) 30% Build-profiles design improved and finalised Multiarch cross-compiler packaging designed and implemented Fixed guile-2.0 crossbuild failure on arm and arm64: DebBug #758463 (DEVPLAT-221) Sprint report written and sent https://wiki.debian.org/Sprints/2014/BootstrapSprint/Results gcc-cross-support uploaded to tools test repo (DEVPLAT-220) dpkg-cross autoconf vars redesigned into cross-support Cross-binutils updated (dpkg-vendor, new targets) (DEVPLAT-220) Wanna-build-sortblockers enhanced for BDI output Jobs set up to put w-b-sortblockers output online (debian and debian-ports) Buildd bootstrap tending. (CARD-1586 20%) Cross-gcc updated and uploaded (DEVPLAT-220 10%) Gcc-cross-support enhanced (with Dima Kogan) (DEVPLAT-220) Sbuild fixed to support multiarch builds. Tested on head and build-daemon branches. (DEVPLAT-220) Mass-bugfiling script for arm64 (autoreconf) written (CARD-1586) First 35 Mass bugs filed (CARD-1586 - 10%) New debian arches announced: https://lists.debian.org/debian-devel-announce/2014/08/msg00012.html Debootstrap from new arm buildds tested (DEVPLAT-219 5%) 'Can Debian change anything in less than 2 years' talk given at debconf 'State of bootstrap' talk given at debconf ARM BoF at debconf (with Steve McIntyre) Dak tested with cross-arch deps (OK) (DEVPLAT-220 2%) Design for britney cross-arch processing discussed (DEVPLAT-220 3%) debile (buildd) cross-support designed with developer Wookey -- Principal hats: Linaro, Emdebian, Wookware, Balloonboard, ARM http://wookware.org/

9 years

[ACTIVITY] Week33

by Wookey

== Progress == * Buildd-tending: Debian arm64 is at 84% built (DEVPLAT-226: 5%) http://buildd.debian-ports.org/stats/graph-quarter-big.png * binNMUed packages for arm64 debian-installer builds * rt-app accepted in debian. Asked rt-app authors to clarify GPL2/GPL2+ licencing * Fix and uploaded previously failed uploads: gtkgl2, exiv2 (DEVPLAT-226: 10%) * Unblocked blender cycle with opencolorio/openimageio * Doubled memory on 2nd buildd (but ghc builds still run out). (DEVPLAT-226: 2%) * Built ghc manually on bigger buildd (DEVPLAT-226: 5%) * Investigated persistently failing package builds (DEVPLAT-221: 15%) * Fixed libfm, rrdtool breakge (DEVPLAT-221: 10%) * Found rdeps of mpi-defaults. Did 15 binnmus +giveback to transition: (DEVPLAT-221: 10%) * Worked on openjade (crufty codebase need autofoo updates). (DEVPLAT-226: 20%) * Wrote card for official debian bootstrap * Investigated gcc-4.9 failure. Bug filed (DEVPLAT-226: 5%) * Tracked notmuch failure down to gdb/ptrace/kernel issue. Bug filed. (DEVPLAT-221: 10%) * Made current list/set of binaries to prime official new buildds with * GSOC catch-up with student after hols * BUilt updated cross-binutils packages - uploaded (DEVPLAT-220: 5%) * Worked on breakage in current cross-gcc packages for upcoming sprint: (DEVPLAT-220: 5%) == Plans == * Get cross-toolchain packages into buildable state * Add multiarch builds to sbuild * Fix/NMU remaining packages in SCC http://people.debian.org/~wookey/bootstrap/blockerlist == Issues == Wookey -- Principal hats: Linaro, Emdebian, Wookware, Balloonboard, ARM http://wookware.org/

9 years

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

linaro-dev August 2014