From b3eaf14f4c63fd6abc7b68c6d7a07c5680a6d8e5 Mon Sep 17 00:00:00 2001 From: Shinji Nomoto Date: Thu, 22 May 2025 15:10:58 +0900 Subject: [PATCH 1/3] cpupower: Fix a bug where the -t option of the set subcommand was not working. The set subcommand's -t option is documented as being available for boost configuration, but it was not actually functioning due to a bug in the option handling. Link: https://lore.kernel.org/r/20250522061122.2149188-2-fj5851bi@fujitsu.com Signed-off-by: Shinji Nomoto Signed-off-by: Shuah Khan --- tools/power/cpupower/utils/cpupower-set.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/power/cpupower/utils/cpupower-set.c b/tools/power/cpupower/utils/cpupower-set.c index 0677b58374ab..59ace394cf3e 100644 --- a/tools/power/cpupower/utils/cpupower-set.c +++ b/tools/power/cpupower/utils/cpupower-set.c @@ -62,8 +62,8 @@ int cmd_set(int argc, char **argv) params.params = 0; /* parameter parsing */ - while ((ret = getopt_long(argc, argv, "b:e:m:", - set_opts, NULL)) != -1) { + while ((ret = getopt_long(argc, argv, "b:e:m:t:", + set_opts, NULL)) != -1) { switch (ret) { case 'b': if (params.perf_bias) From 357d1fc38aad2cf4ea6626138cbf68299d20170c Mon Sep 17 00:00:00 2001 From: Shinji Nomoto Date: Thu, 22 May 2025 15:10:59 +0900 Subject: [PATCH 2/3] cpupower: Allow control of boost feature on non-x86 based systems with boost support. The cpufreq subsystem has a generic sysfs interface for controlling boost (/sys/devices/system/cpu/cpufreq/boost). The sysfs interface can be used to enable boost control from the cpupower command on non-x86 platforms as well. So, allow boost controlling on non-x86 system if boost sysfs file exists. The set subcommand enables/disables the boost feature using the following syntax: cpupower set --boost 1 cpupower set --boost 0 The --boost option is an alias for --turbo-boost. We provided the neutral option name because the name "turbo boost" is specific to Intel technology. The frequency-info subcommand displays the enabled/disabled state of the boost feature as follows: boost state support: Active: yes (or no) Link: https://lore.kernel.org/r/20250522061122.2149188-3-fj5851bi@fujitsu.com Signed-off-by: Shinji Nomoto Signed-off-by: Shuah Khan --- tools/power/cpupower/man/cpupower-set.1 | 7 +- tools/power/cpupower/utils/cpufreq-info.c | 16 ++++- tools/power/cpupower/utils/cpupower-set.c | 1 + tools/power/cpupower/utils/helpers/helpers.h | 14 ++-- tools/power/cpupower/utils/helpers/misc.c | 76 ++++++++++++++------ 5 files changed, 81 insertions(+), 33 deletions(-) diff --git a/tools/power/cpupower/man/cpupower-set.1 b/tools/power/cpupower/man/cpupower-set.1 index 500653ef98c7..8ac82b6f9189 100644 --- a/tools/power/cpupower/man/cpupower-set.1 +++ b/tools/power/cpupower/man/cpupower-set.1 @@ -81,10 +81,11 @@ Refer to the AMD P-State kernel documentation for further information. .RE .PP -\-\-turbo\-boost, \-t +\-\-turbo\-boost, \-\-boost, \-t .RS 4 -This option is used to enable or disable the turbo boost feature on -supported Intel and AMD processors. +This option is used to enable or disable the boost feature on +supported Intel and AMD processors, and other boost supported systems. +(The --boost option is an alias for the --turbo-boost option) This option takes as parameter either \fB1\fP to enable, or \fB0\fP to disable the feature. diff --git a/tools/power/cpupower/utils/cpufreq-info.c b/tools/power/cpupower/utils/cpufreq-info.c index fc750e127404..7d3732f5f2f6 100644 --- a/tools/power/cpupower/utils/cpufreq-info.c +++ b/tools/power/cpupower/utils/cpufreq-info.c @@ -128,7 +128,7 @@ static int get_boost_mode_x86(unsigned int cpu) /* ToDo: Make this more global */ unsigned long pstates[MAX_HW_PSTATES] = {0,}; - ret = cpufreq_has_boost_support(cpu, &support, &active, &b_states); + ret = cpufreq_has_x86_boost_support(cpu, &support, &active, &b_states); if (ret) { printf(_("Error while evaluating Boost Capabilities" " on CPU %d -- are you root?\n"), cpu); @@ -204,6 +204,18 @@ static int get_boost_mode_x86(unsigned int cpu) return 0; } +static int get_boost_mode_generic(unsigned int cpu) +{ + bool active; + + if (!cpufreq_has_generic_boost_support(&active)) { + printf(_(" boost state support:\n")); + printf(_(" Active: %s\n"), active ? _("yes") : _("no")); + } + + return 0; +} + /* --boost / -b */ static int get_boost_mode(unsigned int cpu) @@ -214,6 +226,8 @@ static int get_boost_mode(unsigned int cpu) cpupower_cpu_info.vendor == X86_VENDOR_HYGON || cpupower_cpu_info.vendor == X86_VENDOR_INTEL) return get_boost_mode_x86(cpu); + else + get_boost_mode_generic(cpu); freqs = cpufreq_get_boost_frequencies(cpu); if (freqs) { diff --git a/tools/power/cpupower/utils/cpupower-set.c b/tools/power/cpupower/utils/cpupower-set.c index 59ace394cf3e..c2117e5650dd 100644 --- a/tools/power/cpupower/utils/cpupower-set.c +++ b/tools/power/cpupower/utils/cpupower-set.c @@ -21,6 +21,7 @@ static struct option set_opts[] = { {"epp", required_argument, NULL, 'e'}, {"amd-pstate-mode", required_argument, NULL, 'm'}, {"turbo-boost", required_argument, NULL, 't'}, + {"boost", required_argument, NULL, 't'}, { }, }; diff --git a/tools/power/cpupower/utils/helpers/helpers.h b/tools/power/cpupower/utils/helpers/helpers.h index 95749b8ee475..82ea62bdf5a2 100644 --- a/tools/power/cpupower/utils/helpers/helpers.h +++ b/tools/power/cpupower/utils/helpers/helpers.h @@ -103,6 +103,9 @@ extern struct cpupower_cpu_info cpupower_cpu_info; /* cpuid and cpuinfo helpers **************************/ +int cpufreq_has_generic_boost_support(bool *active); +int cpupower_set_turbo_boost(int turbo_boost); + /* X86 ONLY ****************************************/ #if defined(__i386__) || defined(__x86_64__) @@ -118,7 +121,6 @@ extern unsigned long long msr_intel_get_turbo_ratio(unsigned int cpu); extern int cpupower_set_epp(unsigned int cpu, char *epp); extern int cpupower_set_amd_pstate_mode(char *mode); -extern int cpupower_set_turbo_boost(int turbo_boost); /* Read/Write msr ****************************/ @@ -139,8 +141,8 @@ extern int decode_pstates(unsigned int cpu, int boost_states, /* AMD HW pstate decoding **************************/ -extern int cpufreq_has_boost_support(unsigned int cpu, int *support, - int *active, int * states); +int cpufreq_has_x86_boost_support(unsigned int cpu, int *support, + int *active, int *states); /* AMD P-State stuff **************************/ bool cpupower_amd_pstate_enabled(void); @@ -181,13 +183,11 @@ static inline int cpupower_set_epp(unsigned int cpu, char *epp) { return -1; }; static inline int cpupower_set_amd_pstate_mode(char *mode) { return -1; }; -static inline int cpupower_set_turbo_boost(int turbo_boost) -{ return -1; }; /* Read/Write msr ****************************/ -static inline int cpufreq_has_boost_support(unsigned int cpu, int *support, - int *active, int * states) +static inline int cpufreq_has_x86_boost_support(unsigned int cpu, int *support, + int *active, int *states) { return -1; } static inline bool cpupower_amd_pstate_enabled(void) diff --git a/tools/power/cpupower/utils/helpers/misc.c b/tools/power/cpupower/utils/helpers/misc.c index 76e461ff4f74..166dc1e470ea 100644 --- a/tools/power/cpupower/utils/helpers/misc.c +++ b/tools/power/cpupower/utils/helpers/misc.c @@ -8,15 +8,14 @@ #include "helpers/helpers.h" #include "helpers/sysfs.h" #include "cpufreq.h" +#include "cpupower_intern.h" #if defined(__i386__) || defined(__x86_64__) -#include "cpupower_intern.h" - #define MSR_AMD_HWCR 0xc0010015 -int cpufreq_has_boost_support(unsigned int cpu, int *support, int *active, - int *states) +int cpufreq_has_x86_boost_support(unsigned int cpu, int *support, int *active, + int *states) { int ret; unsigned long long val; @@ -124,24 +123,6 @@ int cpupower_set_amd_pstate_mode(char *mode) return 0; } -int cpupower_set_turbo_boost(int turbo_boost) -{ - char path[SYSFS_PATH_MAX]; - char linebuf[2] = {}; - - snprintf(path, sizeof(path), PATH_TO_CPU "cpufreq/boost"); - - if (!is_valid_path(path)) - return -1; - - snprintf(linebuf, sizeof(linebuf), "%d", turbo_boost); - - if (cpupower_write_sysfs(path, linebuf, 2) <= 0) - return -1; - - return 0; -} - bool cpupower_amd_pstate_enabled(void) { char *driver = cpufreq_get_driver(0); @@ -160,6 +141,39 @@ bool cpupower_amd_pstate_enabled(void) #endif /* #if defined(__i386__) || defined(__x86_64__) */ +int cpufreq_has_generic_boost_support(bool *active) +{ + char path[SYSFS_PATH_MAX]; + char linebuf[2] = {}; + unsigned long val; + char *endp; + + snprintf(path, sizeof(path), PATH_TO_CPU "cpufreq/boost"); + + if (!is_valid_path(path)) + return -EACCES; + + if (cpupower_read_sysfs(path, linebuf, 2) <= 0) + return -EINVAL; + + val = strtoul(linebuf, &endp, 0); + if (endp == linebuf || errno == ERANGE) + return -EINVAL; + + switch (val) { + case 0: + *active = false; + break; + case 1: + *active = true; + break; + default: + return -EINVAL; + } + + return 0; +} + /* get_cpustate * * Gather the information of all online CPUs into bitmask struct @@ -259,3 +273,21 @@ void print_speed(unsigned long speed, int no_rounding) } } } + +int cpupower_set_turbo_boost(int turbo_boost) +{ + char path[SYSFS_PATH_MAX]; + char linebuf[2] = {}; + + snprintf(path, sizeof(path), PATH_TO_CPU "cpufreq/boost"); + + if (!is_valid_path(path)) + return -1; + + snprintf(linebuf, sizeof(linebuf), "%d", turbo_boost); + + if (cpupower_write_sysfs(path, linebuf, 2) <= 0) + return -1; + + return 0; +} From 779b1a1cb13ae17028aeddb2fbbdba97357a1e15 Mon Sep 17 00:00:00 2001 From: "Rafael J. Wysocki" Date: Wed, 13 Aug 2025 12:25:58 +0200 Subject: [PATCH 3/3] cpuidle: governors: menu: Avoid selecting states with too much latency Occasionally, the exit latency of the idle state selected by the menu governor may exceed the PM QoS CPU wakeup latency limit. Namely, if the scheduler tick has been stopped already and predicted_ns is greater than the tick period length, the governor may return an idle state whose exit latency exceeds latency_req because that decision is made before checking the current idle state's exit latency. For instance, say that there are 3 idle states, 0, 1, and 2. For idle states 0 and 1, the exit latency is equal to the target residency and the values are 0 and 5 us, respectively. State 2 is deeper and has the exit latency and target residency of 200 us and 2 ms (which is greater than the tick period length), respectively. Say that predicted_ns is equal to TICK_NSEC and the PM QoS latency limit is 20 us. After the first two iterations of the main loop in menu_select(), idx becomes 1 and in the third iteration of it the target residency of the current state (state 2) is greater than predicted_ns. State 2 is not a polling one and predicted_ns is not less than TICK_NSEC, so the check on whether or not the tick has been stopped is done. Say that the tick has been stopped already and there are no imminent timers (that is, delta_tick is greater than the target residency of state 2). In that case, idx becomes 2 and it is returned immediately, but the exit latency of state 2 exceeds the latency limit. Address this issue by modifying the code to compare the exit latency of the current idle state (idle state i) with the latency limit before comparing its target residency with predicted_ns, which allows one more exit_latency_ns check that becomes redundant to be dropped. However, after the above change, latency_req cannot take the predicted_ns value any more, which takes place after commit 38f83090f515 ("cpuidle: menu: Remove iowait influence"), because it may cause a polling state to be returned prematurely. In the context of the previous example say that predicted_ns is 3000 and the PM QoS latency limit is still 20 us. Additionally, say that idle state 0 is a polling one. Moving the exit_latency_ns check before the target_residency_ns one causes the loop to terminate in the second iteration, before the target_residency_ns check, so idle state 0 will be returned even though previously state 1 would be returned if there were no imminent timers. For this reason, remove the assignment of the predicted_ns value to latency_req from the code. Fixes: 5ef499cd571c ("cpuidle: menu: Handle stopped tick more aggressively") Cc: 4.17+ # 4.17+ Signed-off-by: Rafael J. Wysocki Reviewed-by: Christian Loehle Link: https://patch.msgid.link/5043159.31r3eYUQgx@rafael.j.wysocki --- drivers/cpuidle/governors/menu.c | 29 ++++++++++++----------------- 1 file changed, 12 insertions(+), 17 deletions(-) diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c index 81306612a5c6..b2e3d0b0a116 100644 --- a/drivers/cpuidle/governors/menu.c +++ b/drivers/cpuidle/governors/menu.c @@ -287,20 +287,15 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev, return 0; } - if (tick_nohz_tick_stopped()) { - /* - * If the tick is already stopped, the cost of possible short - * idle duration misprediction is much higher, because the CPU - * may be stuck in a shallow idle state for a long time as a - * result of it. In that case say we might mispredict and use - * the known time till the closest timer event for the idle - * state selection. - */ - if (predicted_ns < TICK_NSEC) - predicted_ns = data->next_timer_ns; - } else if (latency_req > predicted_ns) { - latency_req = predicted_ns; - } + /* + * If the tick is already stopped, the cost of possible short idle + * duration misprediction is much higher, because the CPU may be stuck + * in a shallow idle state for a long time as a result of it. In that + * case, say we might mispredict and use the known time till the closest + * timer event for the idle state selection. + */ + if (tick_nohz_tick_stopped() && predicted_ns < TICK_NSEC) + predicted_ns = data->next_timer_ns; /* * Find the idle state with the lowest power while satisfying @@ -316,13 +311,15 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev, if (idx == -1) idx = i; /* first enabled state */ + if (s->exit_latency_ns > latency_req) + break; + if (s->target_residency_ns > predicted_ns) { /* * Use a physical idle state, not busy polling, unless * a timer is going to trigger soon enough. */ if ((drv->states[idx].flags & CPUIDLE_FLAG_POLLING) && - s->exit_latency_ns <= latency_req && s->target_residency_ns <= data->next_timer_ns) { predicted_ns = s->target_residency_ns; idx = i; @@ -354,8 +351,6 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev, return idx; } - if (s->exit_latency_ns > latency_req) - break; idx = i; }