linaro-dev December 2012

linaro-dev@lists.linaro.org

55 participants
57 discussions

by Riku Voipio

Hi, The following code fails to build with OE Aarch64 toolchain with current kernel headers. While ugly, the code is a reduced testcase from fuse build failure ( https://bugs.launchpad.net/linaro-oe/+bug/1087757 ) and the same fuse code compiles on all other architectures. Before I send a workaround for upstream, I'd like to know how we can end up with different definitions for int64_t when that happens on no other architectures - something wrong with the generic kernel headers? Testcase: #include <sys/types.h> #define __s64 int64_t #include <signal.h> int main(int argc, char **argv) { int64_t x=4; return x; } Failure: /data/oe/build/tmp-eglibc/sysroots/x86_64-linux/usr/bin/aarch64-oe-linux/aarch64-oe-linux-gcc -save-temps --sysroot=/data/oe/build/tmp-eglibc/sysroots/genericarmv8 -o test test.c In file included from /data/oe/build/tmp-eglibc/sysroots/genericarmv8/usr/include/asm-generic/types.h:7:0, from /data/oe/build/tmp-eglibc/sysroots/genericarmv8/usr/include/asm/types.h:1, from /data/oe/build/tmp-eglibc/sysroots/genericarmv8/usr/include/linux/types.h:4, from /data/oe/build/tmp-eglibc/sysroots/genericarmv8/usr/include/asm/sigcontext.h:19, from /data/oe/build/tmp-eglibc/sysroots/genericarmv8/usr/include/bits/sigcontext.h:27, from /data/oe/build/tmp-eglibc/sysroots/genericarmv8/usr/include/signal.h:338, from test.c:4: /data/oe/build/tmp-eglibc/sysroots/genericarmv8/usr/include/asm-generic/int-ll64.h:29:44: error: conflicting types for 'int64_t' In file included from test.c:2:0: /data/oe/build/tmp-eglibc/sysroots/genericarmv8/usr/include/sys/types.h:197:13: note: previous declaration of 'int64_t' was here

11 years, 4 months

[PATCH] sched/fair: Set se->vruntime directly in place_entity()

by Viresh Kumar

We are first storing the new vruntime in a variable and then storing it in se->vruntime. Simply update se->vruntime directly. Signed-off-by: Viresh Kumar <viresh.kumar(a)linaro.org> --- kernel/sched/fair.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index a319d56c..820a2f1 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -1454,9 +1454,7 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial) } /* ensure we never gain time by being placed backwards. */ - vruntime = max_vruntime(se->vruntime, vruntime); - - se->vruntime = vruntime; + se->vruntime = max_vruntime(se->vruntime, vruntime); } static void check_enqueue_throttle(struct cfs_rq *cfs_rq); -- 1.7.12.rc2.18.g61b472e

11 years, 4 months

[PATCH] cpufreq: interactive: introduce hispeed2 for big LITTLE systems

by Viresh Kumar

Latest big LITTLE systems for ARM SoC's have following kind of ranges of freqs: LITTLE CPU ranges: x MHz -> y MHz big CPU ranges: y + delta -> z MHz For these, we use hispeed_freq as y MHz, to avoid a switch of clusters from LITTLE to big when load burst comes. Similar to non big.LITTLE systems, we also need to avoid highest OPPs for load bursts and so need another "speed breaker". Hence, we introduce hispeed2_freq here. This would normally be set to z - delta MHz. For, non big.LITTLE systems if we don't set hispeed2_freq, it would be set as policy->max frequency. And so, behavior would remain same as existing code. Based on Initial work from Sudeep, hence his SOB. Signed-off-by: Sudeep KarkadaNagesha <sudeep.karkadanagesha(a)arm.com> Signed-off-by: Viresh Kumar <viresh.kumar(a)linaro.org> --- drivers/cpufreq/cpufreq_interactive.c | 119 +++++++++++++++++++++++++++++++--- 1 file changed, 111 insertions(+), 8 deletions(-) diff --git a/drivers/cpufreq/cpufreq_interactive.c b/drivers/cpufreq/cpufreq_interactive.c index 4dc7f04..0c5bb2c 100644 --- a/drivers/cpufreq/cpufreq_interactive.c +++ b/drivers/cpufreq/cpufreq_interactive.c @@ -68,6 +68,13 @@ static unsigned int hispeed_freq; #define DEFAULT_GO_HISPEED_LOAD 85 static unsigned long go_hispeed_load; +/* Hi speed2 to bump to from Hi speed when load bursts over hispeed freq */ +static unsigned int hispeed2_freq; + +/* Go to hi speed2 when CPU load at or above this value. */ +#define DEFAULT_GO_HISPEED2_LOAD 95 +static unsigned long go_hispeed2_load; + /* Target load. Lower values result in higher CPU speeds. */ #define DEFAULT_TARGET_LOAD 90 static unsigned int default_target_loads[] = {DEFAULT_TARGET_LOAD}; @@ -88,12 +95,15 @@ static unsigned long min_sample_time; static unsigned long timer_rate; /* - * Wait this long before raising speed above hispeed, by default a single - * timer interval. + * Wait this long before raising speed above hispeed/hispeed2, by default a + * single timer interval. */ #define DEFAULT_ABOVE_HISPEED_DELAY DEFAULT_TIMER_RATE static unsigned long above_hispeed_delay_val; +#define DEFAULT_ABOVE_HISPEED2_DELAY DEFAULT_TIMER_RATE +static unsigned long above_hispeed2_delay_val; + /* Non-zero means indefinite speed boost active */ static int boost_val; /* Duration of a boot pulse in usecs */ @@ -272,7 +282,7 @@ static void cpufreq_interactive_timer(unsigned long data) int cpu_load; struct cpufreq_interactive_cpuinfo *pcpu = &per_cpu(cpuinfo, data); - unsigned int new_freq; + unsigned int new_freq = 0, min_freq = 0; unsigned int loadadjfreq; unsigned int index; unsigned long flags; @@ -301,13 +311,22 @@ static void cpufreq_interactive_timer(unsigned long data) if (pcpu->target_freq < hispeed_freq) { new_freq = hispeed_freq; } else { - new_freq = choose_freq(pcpu, loadadjfreq); - - if (new_freq < hispeed_freq) - new_freq = hispeed_freq; + if (cpu_load >= go_hispeed2_load) { + if (pcpu->target_freq < hispeed2_freq) + new_freq = hispeed2_freq; + else + min_freq = hispeed2_freq; + } else { + min_freq = hispeed_freq; + } } - } else { + } + + if (!new_freq) { new_freq = choose_freq(pcpu, loadadjfreq); + + if (new_freq < min_freq) + new_freq = min_freq; } if (pcpu->target_freq >= hispeed_freq && @@ -319,6 +338,15 @@ static void cpufreq_interactive_timer(unsigned long data) goto rearm; } + if (pcpu->target_freq >= hispeed2_freq && + new_freq > pcpu->target_freq && + now - pcpu->hispeed_validate_time < above_hispeed2_delay_val) { + trace_cpufreq_interactive_notyet( + data, cpu_load, pcpu->target_freq, + pcpu->policy->cur, new_freq); + goto rearm; + } + pcpu->hispeed_validate_time = now; if (cpufreq_frequency_table_target(pcpu->policy, pcpu->freq_table, @@ -688,6 +716,51 @@ static ssize_t store_go_hispeed_load(struct kobject *kobj, static struct global_attr go_hispeed_load_attr = __ATTR(go_hispeed_load, 0644, show_go_hispeed_load, store_go_hispeed_load); +static ssize_t show_hispeed2_freq(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + return sprintf(buf, "%u\n", hispeed2_freq); +} + +static ssize_t store_hispeed2_freq(struct kobject *kobj, + struct attribute *attr, const char *buf, + size_t count) +{ + int ret; + long unsigned int val; + + ret = strict_strtoul(buf, 0, &val); + if (ret < 0) + return ret; + hispeed2_freq = val; + return count; +} + +static struct global_attr hispeed2_freq_attr = __ATTR(hispeed2_freq, 0644, + show_hispeed2_freq, store_hispeed2_freq); + +static ssize_t show_go_hispeed2_load(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + return sprintf(buf, "%lu\n", go_hispeed2_load); +} + +static ssize_t store_go_hispeed2_load(struct kobject *kobj, + struct attribute *attr, const char *buf, size_t count) +{ + int ret; + unsigned long val; + + ret = strict_strtoul(buf, 0, &val); + if (ret < 0) + return ret; + go_hispeed2_load = val; + return count; +} + +static struct global_attr go_hispeed2_load_attr = __ATTR(go_hispeed2_load, 0644, + show_go_hispeed2_load, store_go_hispeed2_load); + static ssize_t show_min_sample_time(struct kobject *kobj, struct attribute *attr, char *buf) { @@ -732,6 +805,28 @@ static ssize_t store_above_hispeed_delay(struct kobject *kobj, define_one_global_rw(above_hispeed_delay); +static ssize_t show_above_hispeed2_delay(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + return sprintf(buf, "%lu\n", above_hispeed2_delay_val); +} + +static ssize_t store_above_hispeed2_delay(struct kobject *kobj, + struct attribute *attr, + const char *buf, size_t count) +{ + int ret; + unsigned long val; + + ret = strict_strtoul(buf, 0, &val); + if (ret < 0) + return ret; + above_hispeed2_delay_val = val; + return count; +} + +define_one_global_rw(above_hispeed2_delay); + static ssize_t show_timer_rate(struct kobject *kobj, struct attribute *attr, char *buf) { @@ -854,6 +949,9 @@ static struct attribute *interactive_attributes[] = { &hispeed_freq_attr.attr, &go_hispeed_load_attr.attr, &above_hispeed_delay.attr, + &hispeed2_freq_attr.attr, + &go_hispeed2_load_attr.attr, + &above_hispeed2_delay.attr, &min_sample_time_attr.attr, &timer_rate_attr.attr, &timer_slack.attr, @@ -906,6 +1004,9 @@ static int cpufreq_governor_interactive(struct cpufreq_policy *policy, if (!hispeed_freq) hispeed_freq = policy->max; + if (!hispeed2_freq || hispeed2_freq > policy->max) + hispeed2_freq = policy->max; + for_each_cpu(j, policy->cpus) { unsigned long expires; @@ -994,6 +1095,8 @@ static int __init cpufreq_interactive_init(void) go_hispeed_load = DEFAULT_GO_HISPEED_LOAD; min_sample_time = DEFAULT_MIN_SAMPLE_TIME; above_hispeed_delay_val = DEFAULT_ABOVE_HISPEED_DELAY; + go_hispeed2_load = DEFAULT_GO_HISPEED2_LOAD; + above_hispeed2_delay_val = DEFAULT_ABOVE_HISPEED2_DELAY; timer_rate = DEFAULT_TIMER_RATE; /* Initalize per-cpu timers */ -- 1.7.12.rc2.18.g61b472e

11 years, 4 months

[Resend Patch v2 0/3] EXYNOS: Add support for Exynos4x12

by Chander Kashyap

This patch series popultes Register addresses, clock structure and gpio structure for Exynos4x12. Rebase to latest u-boot-samsung tree. Chander Kashyap (3): EXYNOS: EXYNOS4X12: Populate Exynos4x12 register addresses EXYNOS: EXYNOS4X12: Add clock structure for EXYNOS4X12 EXYNOS: EXYNOS4X12: Add gpio structure for EXYNOS4X12 arch/arm/include/asm/arch-exynos/clock.h | 276 ++++++++++++++++++++++++++++++ arch/arm/include/asm/arch-exynos/cpu.h | 48 +++++- arch/arm/include/asm/arch-exynos/gpio.h | 85 +++++++++ 3 files changed, 402 insertions(+), 7 deletions(-) -- 1.7.9.5

11 years, 4 months

[RFC PATCH v2 0/6] sched: packing small tasks

by Vincent Guittot

Hi, This patchset takes advantage of the new per-task load tracking that is available in the kernel for packing the small tasks in as few as possible CPU/Cluster/Core. The main goal of packing small tasks is to reduce the power consumption by minimizing the number of power domain that are enabled. The packing is done in 2 steps: The 1st step looks for the best place to pack tasks in a system according to its topology and it defines a pack buddy CPU for each CPU if there is one available. The policy for defining a buddy CPU is that we pack at all levels where a group of CPU can be power gated independently from others. For describing this capability, a new flag has been introduced SD_SHARE_POWERDOMAIN that is used to indicate whether the groups of CPUs of a scheduling domain are sharing their power state. By default, this flag has been set in all sched_domain in order to keep unchanged the current behavior of the scheduler. In a 2nd step, the scheduler checks the load average of a task which wakes up as well as the load average of the buddy CPU and can decide to migrate the task on the buddy. This check is done during the wake up because small tasks tend to wake up between load balance and asynchronously to each other which prevents the default mechanism to catch and migrate them efficiently. Change since V1: Patch 2/6 - Change the flag name which was not clear. The new name is SD_SHARE_POWERDOMAIN. - Create an architecture dependent function to tune the sched_domain flags Patch 3/6 - Fix issues in the algorithm that looks for the best buddy CPU - Use pr_debug instead of pr_info - Fix for uniprocessor Patch 4/6 - Remove the use of usage_avg_sum which has not been merged Patch 5/6 - Change the way the coherency of runnable_avg_sum and runnable_avg_period is ensured Patch 6/6 - Use the arch dependent function to set/clear SD_SHARE_POWERDOMAIN for ARM platform New results for V2: This series has been tested with MP3 play back on ARM platform: TC2 HMP (dual CA-15 and 3xCA-7 cluster). The measurements have been done on an Ubuntu image during 60 seconds of playback and the result has been normalized to 100. | CA15 | CA7 | total | ------------------------------------- default | 81 | 97 | 178 | pack | 13 | 100 | 113 | ------------------------------------- Previous result for V1: The patch-set has been tested on ARM platforms: quad CA-9 SMP and TC2 HMP (dual CA-15 and 3xCA-7 cluster). For ARM platform, the results have demonstrated that it's worth packing small tasks at all topology levels. The performance tests have been done on both platforms with sysbench. The results don't show any performance regressions. These results are aligned with the policy which uses the normal behavior with heavy use cases. test: sysbench --test=cpu --num-threads=N --max-requests=R run Results below is the average duration of 3 tests on the quad CA-9. default is the current scheduler behavior (pack buddy CPU is -1) pack is the scheduler with the pack mechanism | default | pack | ----------------------------------- N=8; R=200 | 3.1999 | 3.1921 | N=8; R=2000 | 31.4939 | 31.4844 | N=12; R=200 | 3.2043 | 3.2084 | N=12; R=2000 | 31.4897 | 31.4831 | N=16; R=200 | 3.1774 | 3.1824 | N=16; R=2000 | 31.4899 | 31.4897 | ----------------------------------- The power consumption tests have been done only on TC2 platform which has got accessible power lines and I have used cyclictest to simulate small tasks. The tests show some power consumption improvements. test: cyclictest -t 8 -q -e 1000000 -D 20 & cyclictest -t 8 -q -e 1000000 -D 20 The measurements have been done during 16 seconds and the result has been normalized to 100 | CA15 | CA7 | total | ------------------------------------- default | 100 | 40 | 140 | pack | <1 | 45 | <46 | ------------------------------------- The A15 cluster is less power efficient than the A7 cluster but if we assume that the tasks is well spread on both clusters, we can guest estimate that the power consumption on a dual cluster of CA7 would have been for a default kernel: Vincent Guittot (6): Revert "sched: introduce temporary FAIR_GROUP_SCHED dependency for load-tracking" sched: add a new SD SHARE_POWERLINE flag for sched_domain sched: pack small tasks sched: secure access to other CPU statistics sched: pack the idle load balance ARM: sched: clear SD_SHARE_POWERLINE arch/arm/kernel/topology.c | 9 +++ arch/ia64/include/asm/topology.h | 1 + arch/tile/include/asm/topology.h | 1 + include/linux/sched.h | 9 +-- include/linux/topology.h | 4 ++ kernel/sched/core.c | 14 ++-- kernel/sched/fair.c | 134 +++++++++++++++++++++++++++++++++++++- kernel/sched/sched.h | 14 ++-- 8 files changed, 163 insertions(+), 23 deletions(-) -- 1.7.9.5

11 years, 4 months

Test Result Summary of Linaro 12.12 Release for Linux Linaro ubuntu Quantal.

by Botao Sun

Linaro 12.12 Release (Calendar Week 51): Here is test result summary for Linux Linaro ubuntu Quantal image on following boards: 1) ARM Versatile Express A9; 2) Samsung Origen; 3) TI Panda 4430; 4) TI Panda 4460; 5) ST Ericsson Snowball. Synopsis: For the major issues, they are: ST Ericsson Snowball boots failed, and no network connection on Samsung Origen. The other boards work well on most of features. 1. ARM Versatile Express A9 + Linux Linaro Quantal (Column E): https://docs.google.com/a/linaro.org/spreadsheet/ccc?key=0AroPySpr4FnEdFNmV… It keeps exactly same status as last week: only "Halt" & "Device Tree" test failed, all other features work well. 2. Samsung Origen + Linux Linaro Quantal (Column E): https://docs.google.com/a/linaro.org/spreadsheet/ccc?key=0AroPySpr4FnEdEowN… Exactly same test result as last week: network connection is unavailable, "reboot" & "halt" test failed.. Power Management test would hang system when run test "cpuhotplug_02". 3. TI Panda 4430 + Linux Linaro Quantal (Column E): https://docs.google.com/a/linaro.org/spreadsheet/ccc?key=0AroPySpr4FnEdEwwZ… ARM DS-5 backs to work. All 3 tests, data streaming, remote SSH connection and on target application debug passed. Device Tree is unavailable, the directory "/proc/device-tree" is empty. A new issue here is power management test "cpuhotplug_07" failed. However, this test passed on TI Panda 4460. 4. TI Panda 4460 + Linux Linaro Quantal (Column E): https://docs.google.com/a/linaro.org/spreadsheet/ccc?key=0AroPySpr4FnEdEwwZ… ARM DS-5 backs to work, now only Device Tree issue - empty directory "/proc/device-tree" exists. All other features work well. 5. ST Ericsson Snowball + Linux Linaro Quantal (Column E): https://docs.google.com/a/linaro.org/spreadsheet/ccc?key=0AroPySpr4FnEdFJ4X… Same as last week, system boots failed. For the previous week (Calendar week 50) summary, please refer to attachment. Thank you. Best Regards Botao Sun

11 years, 4 months

[PATCH] Powertop: Fix script for android

by Chander Kashyap

Following issues fixed: 1. Name of generated rcport file is powertop*.csv not PowerTop*.csv 2. sudo not available in android so dropped it. 3. ls -1 not supported in android ls command so changed it to plain ls. Signed-off-by: Chander Kashyap <chander.kashyap(a)linaro.org> --- powertop/powertop_01.sh | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/powertop/powertop_01.sh b/powertop/powertop_01.sh index cf243f1..d2e78f2 100755 --- a/powertop/powertop_01.sh +++ b/powertop/powertop_01.sh @@ -1,4 +1,4 @@ -#!/bin/bash +#!/bin/bas # # PM-QA validation test suite for the power management on Linux # @@ -33,14 +33,14 @@ run_powertop() { local report=csv local seconds=10 local iterations=2 - local report_name=PowerTOP*.csv + local report_name=powertop*.csv # remove old reports if exists rm -f $report_name # run powertop for $(iterations) in report generation mode start_time=`date +%s` - sudo $bin_path --$report --time=$seconds --iteration=$iterations + $bin_path --$report --time=$seconds --iteration=$iterations end_time=`date +%s` # check if powertop run for desired time @@ -50,7 +50,7 @@ run_powertop() { check "if powertop run for $expected_time sec" "test $actual_time -ge $expected_time" # check if $(iterations) number of reports are generated - check "if reports are generated" "test $(ls -1 $report_name | wc -l) -eq $iterations" + check "if reports are generated" "test $(ls $report_name | wc -l) -eq $iterations" return 0 } -- 1.7.9.5

11 years, 4 months

[PATCH v2] Powertop: Fix script for android

by Chander Kashyap

Following issues fixed: 1. Name of generated rcport file is powertop*.csv not PowerTop*.csv 2. sudo not available in android so dropped it. 3. "ls -1" not supported in android "ls" command so changed it to plain "ls". Signed-off-by: Chander Kashyap <chander.kashyap(a)linaro.org> --- Changes in v2: fixed typo powertop/powertop_01.sh | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/powertop/powertop_01.sh b/powertop/powertop_01.sh index cf243f1..6f2511f 100755 --- a/powertop/powertop_01.sh +++ b/powertop/powertop_01.sh @@ -33,14 +33,14 @@ run_powertop() { local report=csv local seconds=10 local iterations=2 - local report_name=PowerTOP*.csv + local report_name=powertop*.csv # remove old reports if exists rm -f $report_name # run powertop for $(iterations) in report generation mode start_time=`date +%s` - sudo $bin_path --$report --time=$seconds --iteration=$iterations + $bin_path --$report --time=$seconds --iteration=$iterations end_time=`date +%s` # check if powertop run for desired time @@ -50,7 +50,7 @@ run_powertop() { check "if powertop run for $expected_time sec" "test $actual_time -ge $expected_time" # check if $(iterations) number of reports are generated - check "if reports are generated" "test $(ls -1 $report_name | wc -l) -eq $iterations" + check "if reports are generated" "test $(ls $report_name | wc -l) -eq $iterations" return 0 } -- 1.7.9.5

11 years, 4 months

[GIT PULL] cpufreq-interactive-gov-master-v1

by Viresh Kumar

Hi Andrey, This is more of announcement than a PULL request. :) Most of us are working with interactive governor. But most of us use a different version of it, as it is not yet mainlined and there is no hope of it getting mainlined. http://lists.linaro.org/pipermail/linaro-kernel/2012-February/001122.html So, we have decided to keep the latest version within Linaro and make it part of our release, so that everybody works with the same version and use it in ubuntu as well. I am keeping it in my big LITTLE MP repository in separate branches: - cpufreq-interactive-gov-master-v*: Latest version of governor from google tree - cpufreq-interactive-gov-experimental-v*: Rebased over master with updates from Linaro. This is a PULL request for master branch: -------------- The following changes since commit 9489e9dcae718d5fde988e4a684a0f55b5f94d17: Linux 3.7-rc7 (2012-11-25 17:59:19 -0800) are available in the git repository at: git://git.linaro.org/arm/big.LITTLE/mp.git cpufreq-interactive-gov-master-v1 for you to fetch changes up to 1ad5d36179224fbcdddf8abf3cbfdd183da38b3e: cpufreq: interactive: kick timer on idle exit past expiry (2012-11-29 00:48:46 +0530) ---------------------------------------------------------------- John Stultz (1): cpufreq-interactive: Compile fixup Lianwei Wang (1): cpufreq: interactive: use deferrable timer by default Mike Chan (1): cpufreq: interactive: New 'interactive' governor Sam Leffler (2): cpufreq: interactive: take idle notifications only when active cpufreq: interactive: keep freezer happy when not current governor Todd Poynor (24): Move x86_64 idle notifiers to generic ARM: Call idle notifiers ARM: Move leds idle start/stop calls to idle notifiers cpufreq interactive governor: event tracing cpufreq: interactive: apply intermediate load to max speed not current cpufreq: interactive: set at least hispeed when above hispeed load cpufreq: interactive: don't drop speed if recently at higher load cpufreq: interactive: configurable delay before raising above hispeed cpufreq: interactive: adjust code and documentation to match cpufreq: interactive: base hispeed bump on target freq, not actual cpufreq: interactive: Separate speed target revalidate time and initial set time cpufreq: interactive: Boost frequency on touchscreen input cpufreq: interactive: remove unused target_validate_time_in_idle cpufreq: interactive: Add sysfs boost interface for hints from userspace cpufreq: interactive: set floor for boosted speed cpufreq: interactive: add boost pulse interface cpufreq: interactive: restart above_hispeed_delay at each hispeed load cpufreq: interactive: fixup trace of string params cpufreq: interactive: handle speed up and down in the realtime task cpufreq: interactive: remove input_boost handling cpufreq: interactive: always limit initial speed bump to hispeed cpufreq: interactive: run at fraction of hispeed_freq when load is low cpufreq: interactive: pin timers to associated CPU cpufreq: interactive: kick timer on idle exit past expiry Documentation/cpu-freq/governors.txt | 59 +++ arch/arm/kernel/process.c | 4 +- arch/x86/include/asm/idle.h | 7 - arch/x86/kernel/process.c | 17 +- drivers/cpufreq/Kconfig | 27 ++ drivers/cpufreq/Makefile | 1 + drivers/cpufreq/cpufreq_interactive.c | 754 +++++++++++++++++++++++++++++ drivers/leds/ledtrig-cpu.c | 25 + include/linux/cpu.h | 7 + include/linux/cpufreq.h | 3 + include/trace/events/cpufreq_interactive.h | 106 ++++ kernel/cpu.c | 20 + 12 files changed, 1006 insertions(+), 24 deletions(-) create mode 100644 drivers/cpufreq/cpufreq_interactive.c create mode 100644 include/trace/events/cpufreq_interactive.h

11 years, 4 months

[PATCH] cpufreq: interactive: fix racy timer stopping

by Viresh Kumar

From: Nicolas Pitre <nicolas.pitre(a)linaro.org> Relying on pcpu->governor_enabled alone to ensure proper shutdown of the governor activities is flawed. The following race is perfectly possible: CPU0 |CPU2 ---- +---- |cpufreq_interactive_timer() cpufreq_governor_interactive(GOV_STOP) |if (!pcpu->governor_enabled) pcpu->governor_enabled = 0; | return; [untaken] smp_wmb(); |[...] del_timer_sync(&pcpu->cpu_timer); |cpufreq_interactive_timer_resched() |mod_timer_pinned(&pcpu->cpu_timer) Result: timer is pending again. It is also possible for del_timer_sync() to race against the invocation of the idle notifier callback which also has the potential to reactivate the timer. To fix this issue, a read-write semaphore is used. Multiple readers are allowed as long as pcpu->governor_enabled is true. However it can be moved to false only after taking a write semaphore which would wait for any on-going asynchronous activities to complete and prevent any more of those activities to be initiated. Signed-off-by: Nicolas Pitre <nicolas.pitre(a)linaro.org> --- drivers/cpufreq/cpufreq_interactive.c | 38 ++++++++++++++++++++++++++--------- 1 file changed, 28 insertions(+), 10 deletions(-) diff --git a/drivers/cpufreq/cpufreq_interactive.c b/drivers/cpufreq/cpufreq_interactive.c index c82d9fe..061af7f 100644 --- a/drivers/cpufreq/cpufreq_interactive.c +++ b/drivers/cpufreq/cpufreq_interactive.c @@ -21,14 +21,13 @@ #include <linux/cpufreq.h> #include <linux/module.h> #include <linux/moduleparam.h> -#include <linux/mutex.h> +#include <linux/rwsem.h> #include <linux/sched.h> #include <linux/tick.h> #include <linux/time.h> #include <linux/timer.h> #include <linux/workqueue.h> #include <linux/kthread.h> -#include <linux/mutex.h> #include <linux/slab.h> #include <asm/cputime.h> @@ -50,6 +49,7 @@ struct cpufreq_interactive_cpuinfo { unsigned int floor_freq; u64 floor_validate_time; u64 hispeed_validate_time; + struct rw_semaphore mutex; int governor_enabled; }; @@ -138,8 +138,8 @@ static void cpufreq_interactive_timer(unsigned long data) unsigned int index; unsigned long flags; - smp_rmb(); - + if (!down_read_trylock(&pcpu->mutex)) + return; if (!pcpu->governor_enabled) goto exit; @@ -258,6 +258,7 @@ rearm: } exit: + up_read(&pcpu->mutex); return; } @@ -267,8 +268,12 @@ static void cpufreq_interactive_idle_start(void) &per_cpu(cpuinfo, smp_processor_id()); int pending; - if (!pcpu->governor_enabled) + if (!down_read_trylock(&pcpu->mutex)) + return; + if (!pcpu->governor_enabled) { + up_read(&pcpu->mutex); return; + } pending = timer_pending(&pcpu->cpu_timer); @@ -298,6 +303,7 @@ static void cpufreq_interactive_idle_start(void) } } + up_read(&pcpu->mutex); } static void cpufreq_interactive_idle_end(void) @@ -305,8 +311,12 @@ static void cpufreq_interactive_idle_end(void) struct cpufreq_interactive_cpuinfo *pcpu = &per_cpu(cpuinfo, smp_processor_id()); - if (!pcpu->governor_enabled) + if (!down_read_trylock(&pcpu->mutex)) + return; + if (!pcpu->governor_enabled) { + up_read(&pcpu->mutex); return; + } /* Arm the timer for 1-2 ticks later if not already. */ if (!timer_pending(&pcpu->cpu_timer)) { @@ -317,6 +327,8 @@ static void cpufreq_interactive_idle_end(void) del_timer(&pcpu->cpu_timer); cpufreq_interactive_timer(smp_processor_id()); } + + up_read(&pcpu->mutex); } static int cpufreq_interactive_speedchange_task(void *data) @@ -351,10 +363,12 @@ static int cpufreq_interactive_speedchange_task(void *data) unsigned int max_freq = 0; pcpu = &per_cpu(cpuinfo, cpu); - smp_rmb(); - - if (!pcpu->governor_enabled) + if (!down_read_trylock(&pcpu->mutex)) + continue; + if (!pcpu->governor_enabled) { + up_read(&pcpu->mutex); continue; + } for_each_cpu(j, pcpu->policy->cpus) { struct cpufreq_interactive_cpuinfo *pjcpu = @@ -371,6 +385,8 @@ static int cpufreq_interactive_speedchange_task(void *data) trace_cpufreq_interactive_setspeed(cpu, pcpu->target_freq, pcpu->policy->cur); + + up_read(&pcpu->mutex); } } @@ -690,9 +706,10 @@ static int cpufreq_governor_interactive(struct cpufreq_policy *policy, case CPUFREQ_GOV_STOP: for_each_cpu(j, policy->cpus) { pcpu = &per_cpu(cpuinfo, j); + down_write(&pcpu->mutex); pcpu->governor_enabled = 0; - smp_wmb(); del_timer_sync(&pcpu->cpu_timer); + up_write(&pcpu->mutex); } if (atomic_dec_return(&active_count) > 0) @@ -736,6 +753,7 @@ static int __init cpufreq_interactive_init(void) init_timer_deferrable(&pcpu->cpu_timer); pcpu->cpu_timer.function = cpufreq_interactive_timer; pcpu->cpu_timer.data = i; + init_rwsem(&pcpu->mutex); } spin_lock_init(&speedchange_cpumask_lock); -- 1.7.12.rc2.18.g61b472e

11 years, 4 months

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

linaro-dev December 2012