From 030587c28c7e90e7bd4df4cb87d392c6a398b7f7 Mon Sep 17 00:00:00 2001 From: Peter Jung Date: Sun, 18 May 2025 16:17:31 +0200 Subject: [PATCH 1/9] amd-pstate Signed-off-by: Peter Jung --- arch/x86/include/asm/msr-index.h | 20 +- arch/x86/kernel/acpi/cppc.c | 4 +- drivers/cpufreq/amd-pstate-trace.h | 13 +- drivers/cpufreq/amd-pstate-ut.c | 211 ++++----- drivers/cpufreq/amd-pstate.c | 674 ++++++++++++----------------- drivers/cpufreq/amd-pstate.h | 63 +-- include/linux/cpufreq.h | 3 + include/linux/sched/topology.h | 6 + kernel/sched/debug.c | 4 + kernel/sched/fair.c | 5 +- kernel/sched/topology.c | 58 +++ 11 files changed, 507 insertions(+), 554 deletions(-) diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h index d4308e78a009..904dd8c0408c 100644 --- a/arch/x86/include/asm/msr-index.h +++ b/arch/x86/include/asm/msr-index.h @@ -709,15 +709,17 @@ #define MSR_AMD_CPPC_REQ 0xc00102b3 #define MSR_AMD_CPPC_STATUS 0xc00102b4 -#define AMD_CPPC_LOWEST_PERF(x) (((x) >> 0) & 0xff) -#define AMD_CPPC_LOWNONLIN_PERF(x) (((x) >> 8) & 0xff) -#define AMD_CPPC_NOMINAL_PERF(x) (((x) >> 16) & 0xff) -#define AMD_CPPC_HIGHEST_PERF(x) (((x) >> 24) & 0xff) - -#define AMD_CPPC_MAX_PERF(x) (((x) & 0xff) << 0) -#define AMD_CPPC_MIN_PERF(x) (((x) & 0xff) << 8) -#define AMD_CPPC_DES_PERF(x) (((x) & 0xff) << 16) -#define AMD_CPPC_ENERGY_PERF_PREF(x) (((x) & 0xff) << 24) +/* Masks for use with MSR_AMD_CPPC_CAP1 */ +#define AMD_CPPC_LOWEST_PERF_MASK GENMASK(7, 0) +#define AMD_CPPC_LOWNONLIN_PERF_MASK GENMASK(15, 8) +#define AMD_CPPC_NOMINAL_PERF_MASK GENMASK(23, 16) +#define AMD_CPPC_HIGHEST_PERF_MASK GENMASK(31, 24) + +/* Masks for use with MSR_AMD_CPPC_REQ */ +#define AMD_CPPC_MAX_PERF_MASK GENMASK(7, 0) +#define AMD_CPPC_MIN_PERF_MASK GENMASK(15, 8) +#define AMD_CPPC_DES_PERF_MASK GENMASK(23, 16) +#define AMD_CPPC_EPP_PERF_MASK GENMASK(31, 24) /* AMD Performance Counter Global Status and Control MSRs */ #define MSR_AMD64_PERF_CNTR_GLOBAL_STATUS 0xc0000300 diff --git a/arch/x86/kernel/acpi/cppc.c b/arch/x86/kernel/acpi/cppc.c index d745dd586303..77bfb846490c 100644 --- a/arch/x86/kernel/acpi/cppc.c +++ b/arch/x86/kernel/acpi/cppc.c @@ -4,6 +4,8 @@ * Copyright (c) 2016, Intel Corporation. */ +#include + #include #include #include @@ -149,7 +151,7 @@ int amd_get_highest_perf(unsigned int cpu, u32 *highest_perf) if (ret) goto out; - val = AMD_CPPC_HIGHEST_PERF(val); + val = FIELD_GET(AMD_CPPC_HIGHEST_PERF_MASK, val); } else { ret = cppc_get_highest_perf(cpu, &val); if (ret) diff --git a/drivers/cpufreq/amd-pstate-trace.h b/drivers/cpufreq/amd-pstate-trace.h index f457d4af2c62..32e1bdc588c5 100644 --- a/drivers/cpufreq/amd-pstate-trace.h +++ b/drivers/cpufreq/amd-pstate-trace.h @@ -90,7 +90,8 @@ TRACE_EVENT(amd_pstate_epp_perf, u8 epp, u8 min_perf, u8 max_perf, - bool boost + bool boost, + bool changed ), TP_ARGS(cpu_id, @@ -98,7 +99,8 @@ TRACE_EVENT(amd_pstate_epp_perf, epp, min_perf, max_perf, - boost), + boost, + changed), TP_STRUCT__entry( __field(unsigned int, cpu_id) @@ -107,6 +109,7 @@ TRACE_EVENT(amd_pstate_epp_perf, __field(u8, min_perf) __field(u8, max_perf) __field(bool, boost) + __field(bool, changed) ), TP_fast_assign( @@ -116,15 +119,17 @@ TRACE_EVENT(amd_pstate_epp_perf, __entry->min_perf = min_perf; __entry->max_perf = max_perf; __entry->boost = boost; + __entry->changed = changed; ), - TP_printk("cpu%u: [%hhu<->%hhu]/%hhu, epp=%hhu, boost=%u", + TP_printk("cpu%u: [%hhu<->%hhu]/%hhu, epp=%hhu, boost=%u, changed=%u", (unsigned int)__entry->cpu_id, (u8)__entry->min_perf, (u8)__entry->max_perf, (u8)__entry->highest_perf, (u8)__entry->epp, - (bool)__entry->boost + (bool)__entry->boost, + (bool)__entry->changed ) ); diff --git a/drivers/cpufreq/amd-pstate-ut.c b/drivers/cpufreq/amd-pstate-ut.c index 3a0a380c3590..e671bc7d1550 100644 --- a/drivers/cpufreq/amd-pstate-ut.c +++ b/drivers/cpufreq/amd-pstate-ut.c @@ -22,39 +22,31 @@ #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt +#include #include #include #include #include +#include #include #include "amd-pstate.h" -/* - * Abbreviations: - * amd_pstate_ut: used as a shortform for AMD P-State unit test. - * It helps to keep variable names smaller, simpler - */ -enum amd_pstate_ut_result { - AMD_PSTATE_UT_RESULT_PASS, - AMD_PSTATE_UT_RESULT_FAIL, -}; struct amd_pstate_ut_struct { const char *name; - void (*func)(u32 index); - enum amd_pstate_ut_result result; + int (*func)(u32 index); }; /* * Kernel module for testing the AMD P-State unit test */ -static void amd_pstate_ut_acpi_cpc_valid(u32 index); -static void amd_pstate_ut_check_enabled(u32 index); -static void amd_pstate_ut_check_perf(u32 index); -static void amd_pstate_ut_check_freq(u32 index); -static void amd_pstate_ut_check_driver(u32 index); +static int amd_pstate_ut_acpi_cpc_valid(u32 index); +static int amd_pstate_ut_check_enabled(u32 index); +static int amd_pstate_ut_check_perf(u32 index); +static int amd_pstate_ut_check_freq(u32 index); +static int amd_pstate_ut_check_driver(u32 index); static struct amd_pstate_ut_struct amd_pstate_ut_cases[] = { {"amd_pstate_ut_acpi_cpc_valid", amd_pstate_ut_acpi_cpc_valid }, @@ -77,71 +69,67 @@ static bool get_shared_mem(void) /* * check the _CPC object is present in SBIOS. */ -static void amd_pstate_ut_acpi_cpc_valid(u32 index) +static int amd_pstate_ut_acpi_cpc_valid(u32 index) { - if (acpi_cpc_valid()) - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_PASS; - else { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; + if (!acpi_cpc_valid()) { pr_err("%s the _CPC object is not present in SBIOS!\n", __func__); + return -EINVAL; } + + return 0; } -static void amd_pstate_ut_pstate_enable(u32 index) +/* + * check if amd pstate is enabled + */ +static int amd_pstate_ut_check_enabled(u32 index) { - int ret = 0; u64 cppc_enable = 0; + int ret; + + if (get_shared_mem()) + return 0; ret = rdmsrl_safe(MSR_AMD_CPPC_ENABLE, &cppc_enable); if (ret) { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; pr_err("%s rdmsrl_safe MSR_AMD_CPPC_ENABLE ret=%d error!\n", __func__, ret); - return; + return ret; } - if (cppc_enable) - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_PASS; - else { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; + + if (!cppc_enable) { pr_err("%s amd pstate must be enabled!\n", __func__); + return -EINVAL; } -} -/* - * check if amd pstate is enabled - */ -static void amd_pstate_ut_check_enabled(u32 index) -{ - if (get_shared_mem()) - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_PASS; - else - amd_pstate_ut_pstate_enable(index); + return 0; } /* * check if performance values are reasonable. * highest_perf >= nominal_perf > lowest_nonlinear_perf > lowest_perf > 0 */ -static void amd_pstate_ut_check_perf(u32 index) +static int amd_pstate_ut_check_perf(u32 index) { int cpu = 0, ret = 0; u32 highest_perf = 0, nominal_perf = 0, lowest_nonlinear_perf = 0, lowest_perf = 0; u64 cap1 = 0; struct cppc_perf_caps cppc_perf; - struct cpufreq_policy *policy = NULL; - struct amd_cpudata *cpudata = NULL; + union perf_cached cur_perf; + + for_each_online_cpu(cpu) { + struct cpufreq_policy *policy __free(put_cpufreq_policy) = NULL; + struct amd_cpudata *cpudata; - for_each_possible_cpu(cpu) { policy = cpufreq_cpu_get(cpu); if (!policy) - break; + continue; cpudata = policy->driver_data; if (get_shared_mem()) { ret = cppc_get_perf_caps(cpu, &cppc_perf); if (ret) { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; pr_err("%s cppc_get_perf_caps ret=%d error!\n", __func__, ret); - goto skip_test; + return ret; } highest_perf = cppc_perf.highest_perf; @@ -151,50 +139,44 @@ static void amd_pstate_ut_check_perf(u32 index) } else { ret = rdmsrl_safe_on_cpu(cpu, MSR_AMD_CPPC_CAP1, &cap1); if (ret) { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; pr_err("%s read CPPC_CAP1 ret=%d error!\n", __func__, ret); - goto skip_test; + return ret; } - highest_perf = AMD_CPPC_HIGHEST_PERF(cap1); - nominal_perf = AMD_CPPC_NOMINAL_PERF(cap1); - lowest_nonlinear_perf = AMD_CPPC_LOWNONLIN_PERF(cap1); - lowest_perf = AMD_CPPC_LOWEST_PERF(cap1); + highest_perf = FIELD_GET(AMD_CPPC_HIGHEST_PERF_MASK, cap1); + nominal_perf = FIELD_GET(AMD_CPPC_NOMINAL_PERF_MASK, cap1); + lowest_nonlinear_perf = FIELD_GET(AMD_CPPC_LOWNONLIN_PERF_MASK, cap1); + lowest_perf = FIELD_GET(AMD_CPPC_LOWEST_PERF_MASK, cap1); } - if (highest_perf != READ_ONCE(cpudata->highest_perf) && !cpudata->hw_prefcore) { + cur_perf = READ_ONCE(cpudata->perf); + if (highest_perf != cur_perf.highest_perf && !cpudata->hw_prefcore) { pr_err("%s cpu%d highest=%d %d highest perf doesn't match\n", - __func__, cpu, highest_perf, cpudata->highest_perf); - goto skip_test; + __func__, cpu, highest_perf, cur_perf.highest_perf); + return -EINVAL; } - if ((nominal_perf != READ_ONCE(cpudata->nominal_perf)) || - (lowest_nonlinear_perf != READ_ONCE(cpudata->lowest_nonlinear_perf)) || - (lowest_perf != READ_ONCE(cpudata->lowest_perf))) { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; + if (nominal_perf != cur_perf.nominal_perf || + (lowest_nonlinear_perf != cur_perf.lowest_nonlinear_perf) || + (lowest_perf != cur_perf.lowest_perf)) { pr_err("%s cpu%d nominal=%d %d lowest_nonlinear=%d %d lowest=%d %d, they should be equal!\n", - __func__, cpu, nominal_perf, cpudata->nominal_perf, - lowest_nonlinear_perf, cpudata->lowest_nonlinear_perf, - lowest_perf, cpudata->lowest_perf); - goto skip_test; + __func__, cpu, nominal_perf, cur_perf.nominal_perf, + lowest_nonlinear_perf, cur_perf.lowest_nonlinear_perf, + lowest_perf, cur_perf.lowest_perf); + return -EINVAL; } if (!((highest_perf >= nominal_perf) && (nominal_perf > lowest_nonlinear_perf) && - (lowest_nonlinear_perf > lowest_perf) && + (lowest_nonlinear_perf >= lowest_perf) && (lowest_perf > 0))) { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; pr_err("%s cpu%d highest=%d >= nominal=%d > lowest_nonlinear=%d > lowest=%d > 0, the formula is incorrect!\n", __func__, cpu, highest_perf, nominal_perf, lowest_nonlinear_perf, lowest_perf); - goto skip_test; + return -EINVAL; } - cpufreq_cpu_put(policy); } - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_PASS; - return; -skip_test: - cpufreq_cpu_put(policy); + return 0; } /* @@ -202,59 +184,50 @@ static void amd_pstate_ut_check_perf(u32 index) * max_freq >= nominal_freq > lowest_nonlinear_freq > min_freq > 0 * check max freq when set support boost mode. */ -static void amd_pstate_ut_check_freq(u32 index) +static int amd_pstate_ut_check_freq(u32 index) { int cpu = 0; - struct cpufreq_policy *policy = NULL; - struct amd_cpudata *cpudata = NULL; - for_each_possible_cpu(cpu) { + for_each_online_cpu(cpu) { + struct cpufreq_policy *policy __free(put_cpufreq_policy) = NULL; + struct amd_cpudata *cpudata; + policy = cpufreq_cpu_get(cpu); if (!policy) - break; + continue; cpudata = policy->driver_data; - if (!((cpudata->max_freq >= cpudata->nominal_freq) && + if (!((policy->cpuinfo.max_freq >= cpudata->nominal_freq) && (cpudata->nominal_freq > cpudata->lowest_nonlinear_freq) && - (cpudata->lowest_nonlinear_freq > cpudata->min_freq) && - (cpudata->min_freq > 0))) { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; + (cpudata->lowest_nonlinear_freq >= policy->cpuinfo.min_freq) && + (policy->cpuinfo.min_freq > 0))) { pr_err("%s cpu%d max=%d >= nominal=%d > lowest_nonlinear=%d > min=%d > 0, the formula is incorrect!\n", - __func__, cpu, cpudata->max_freq, cpudata->nominal_freq, - cpudata->lowest_nonlinear_freq, cpudata->min_freq); - goto skip_test; + __func__, cpu, policy->cpuinfo.max_freq, cpudata->nominal_freq, + cpudata->lowest_nonlinear_freq, policy->cpuinfo.min_freq); + return -EINVAL; } if (cpudata->lowest_nonlinear_freq != policy->min) { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; pr_err("%s cpu%d cpudata_lowest_nonlinear_freq=%d policy_min=%d, they should be equal!\n", __func__, cpu, cpudata->lowest_nonlinear_freq, policy->min); - goto skip_test; + return -EINVAL; } if (cpudata->boost_supported) { - if ((policy->max == cpudata->max_freq) || - (policy->max == cpudata->nominal_freq)) - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_PASS; - else { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; + if ((policy->max != policy->cpuinfo.max_freq) && + (policy->max != cpudata->nominal_freq)) { pr_err("%s cpu%d policy_max=%d should be equal cpu_max=%d or cpu_nominal=%d !\n", - __func__, cpu, policy->max, cpudata->max_freq, + __func__, cpu, policy->max, policy->cpuinfo.max_freq, cpudata->nominal_freq); - goto skip_test; + return -EINVAL; } } else { - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_FAIL; pr_err("%s cpu%d must support boost!\n", __func__, cpu); - goto skip_test; + return -EINVAL; } - cpufreq_cpu_put(policy); } - amd_pstate_ut_cases[index].result = AMD_PSTATE_UT_RESULT_PASS; - return; -skip_test: - cpufreq_cpu_put(policy); + return 0; } static int amd_pstate_set_mode(enum amd_pstate_mode mode) @@ -266,32 +239,28 @@ static int amd_pstate_set_mode(enum amd_pstate_mode mode) return amd_pstate_update_status(mode_str, strlen(mode_str)); } -static void amd_pstate_ut_check_driver(u32 index) +static int amd_pstate_ut_check_driver(u32 index) { enum amd_pstate_mode mode1, mode2 = AMD_PSTATE_DISABLE; - int ret; for (mode1 = AMD_PSTATE_DISABLE; mode1 < AMD_PSTATE_MAX; mode1++) { - ret = amd_pstate_set_mode(mode1); + int ret = amd_pstate_set_mode(mode1); if (ret) - goto out; + return ret; for (mode2 = AMD_PSTATE_DISABLE; mode2 < AMD_PSTATE_MAX; mode2++) { if (mode1 == mode2) continue; ret = amd_pstate_set_mode(mode2); - if (ret) - goto out; + if (ret) { + pr_err("%s: failed to update status for %s->%s\n", __func__, + amd_pstate_get_mode_string(mode1), + amd_pstate_get_mode_string(mode2)); + return ret; + } } } -out: - if (ret) - pr_warn("%s: failed to update status for %s->%s: %d\n", __func__, - amd_pstate_get_mode_string(mode1), - amd_pstate_get_mode_string(mode2), ret); - - amd_pstate_ut_cases[index].result = ret ? - AMD_PSTATE_UT_RESULT_FAIL : - AMD_PSTATE_UT_RESULT_PASS; + + return 0; } static int __init amd_pstate_ut_init(void) @@ -299,16 +268,12 @@ static int __init amd_pstate_ut_init(void) u32 i = 0, arr_size = ARRAY_SIZE(amd_pstate_ut_cases); for (i = 0; i < arr_size; i++) { - amd_pstate_ut_cases[i].func(i); - switch (amd_pstate_ut_cases[i].result) { - case AMD_PSTATE_UT_RESULT_PASS: + int ret = amd_pstate_ut_cases[i].func(i); + + if (ret) + pr_err("%-4d %-20s\t fail: %d!\n", i+1, amd_pstate_ut_cases[i].name, ret); + else pr_info("%-4d %-20s\t success!\n", i+1, amd_pstate_ut_cases[i].name); - break; - case AMD_PSTATE_UT_RESULT_FAIL: - default: - pr_info("%-4d %-20s\t fail!\n", i+1, amd_pstate_ut_cases[i].name); - break; - } } return 0; diff --git a/drivers/cpufreq/amd-pstate.c b/drivers/cpufreq/amd-pstate.c index 1b26845703f6..f62dd09c1e4c 100644 --- a/drivers/cpufreq/amd-pstate.c +++ b/drivers/cpufreq/amd-pstate.c @@ -85,15 +85,9 @@ static struct cpufreq_driver *current_pstate_driver; static struct cpufreq_driver amd_pstate_driver; static struct cpufreq_driver amd_pstate_epp_driver; static int cppc_state = AMD_PSTATE_UNDEFINED; -static bool cppc_enabled; static bool amd_pstate_prefcore = true; static struct quirk_entry *quirks; -#define AMD_CPPC_MAX_PERF_MASK GENMASK(7, 0) -#define AMD_CPPC_MIN_PERF_MASK GENMASK(15, 8) -#define AMD_CPPC_DES_PERF_MASK GENMASK(23, 16) -#define AMD_CPPC_EPP_PERF_MASK GENMASK(31, 24) - /* * AMD Energy Preference Performance (EPP) * The EPP is used in the CCLK DPM controller to drive @@ -142,6 +136,19 @@ static struct quirk_entry quirk_amd_7k62 = { .lowest_freq = 550, }; +static inline u8 freq_to_perf(union perf_cached perf, u32 nominal_freq, unsigned int freq_val) +{ + u32 perf_val = DIV_ROUND_UP_ULL((u64)freq_val * perf.nominal_perf, nominal_freq); + + return (u8)clamp(perf_val, perf.lowest_perf, perf.highest_perf); +} + +static inline u32 perf_to_freq(union perf_cached perf, u32 nominal_freq, u8 perf_val) +{ + return DIV_ROUND_UP_ULL((u64)nominal_freq * perf_val, + perf.nominal_perf); +} + static int __init dmi_matched_7k62_bios_bug(const struct dmi_system_id *dmi) { /** @@ -183,7 +190,6 @@ static inline int get_mode_idx_from_str(const char *str, size_t size) return -EINVAL; } -static DEFINE_MUTEX(amd_pstate_limits_lock); static DEFINE_MUTEX(amd_pstate_driver_lock); static u8 msr_get_epp(struct amd_cpudata *cpudata) @@ -221,9 +227,10 @@ static u8 shmem_get_epp(struct amd_cpudata *cpudata) return FIELD_GET(AMD_CPPC_EPP_PERF_MASK, epp); } -static int msr_update_perf(struct amd_cpudata *cpudata, u8 min_perf, +static int msr_update_perf(struct cpufreq_policy *policy, u8 min_perf, u8 des_perf, u8 max_perf, u8 epp, bool fast_switch) { + struct amd_cpudata *cpudata = policy->driver_data; u64 value, prev; value = prev = READ_ONCE(cpudata->cppc_req_cached); @@ -235,6 +242,18 @@ static int msr_update_perf(struct amd_cpudata *cpudata, u8 min_perf, value |= FIELD_PREP(AMD_CPPC_MIN_PERF_MASK, min_perf); value |= FIELD_PREP(AMD_CPPC_EPP_PERF_MASK, epp); + if (trace_amd_pstate_epp_perf_enabled()) { + union perf_cached perf = READ_ONCE(cpudata->perf); + + trace_amd_pstate_epp_perf(cpudata->cpu, + perf.highest_perf, + epp, + min_perf, + max_perf, + policy->boost_enabled, + value != prev); + } + if (value == prev) return 0; @@ -249,24 +268,24 @@ static int msr_update_perf(struct amd_cpudata *cpudata, u8 min_perf, } WRITE_ONCE(cpudata->cppc_req_cached, value); - WRITE_ONCE(cpudata->epp_cached, epp); return 0; } DEFINE_STATIC_CALL(amd_pstate_update_perf, msr_update_perf); -static inline int amd_pstate_update_perf(struct amd_cpudata *cpudata, +static inline int amd_pstate_update_perf(struct cpufreq_policy *policy, u8 min_perf, u8 des_perf, u8 max_perf, u8 epp, bool fast_switch) { - return static_call(amd_pstate_update_perf)(cpudata, min_perf, des_perf, + return static_call(amd_pstate_update_perf)(policy, min_perf, des_perf, max_perf, epp, fast_switch); } -static int msr_set_epp(struct amd_cpudata *cpudata, u8 epp) +static int msr_set_epp(struct cpufreq_policy *policy, u8 epp) { + struct amd_cpudata *cpudata = policy->driver_data; u64 value, prev; int ret; @@ -274,6 +293,19 @@ static int msr_set_epp(struct amd_cpudata *cpudata, u8 epp) value &= ~AMD_CPPC_EPP_PERF_MASK; value |= FIELD_PREP(AMD_CPPC_EPP_PERF_MASK, epp); + if (trace_amd_pstate_epp_perf_enabled()) { + union perf_cached perf = cpudata->perf; + + trace_amd_pstate_epp_perf(cpudata->cpu, perf.highest_perf, + epp, + FIELD_GET(AMD_CPPC_MIN_PERF_MASK, + cpudata->cppc_req_cached), + FIELD_GET(AMD_CPPC_MAX_PERF_MASK, + cpudata->cppc_req_cached), + policy->boost_enabled, + value != prev); + } + if (value == prev) return 0; @@ -284,7 +316,6 @@ static int msr_set_epp(struct amd_cpudata *cpudata, u8 epp) } /* update both so that msr_update_perf() can effectively check */ - WRITE_ONCE(cpudata->epp_cached, epp); WRITE_ONCE(cpudata->cppc_req_cached, value); return ret; @@ -292,17 +323,35 @@ static int msr_set_epp(struct amd_cpudata *cpudata, u8 epp) DEFINE_STATIC_CALL(amd_pstate_set_epp, msr_set_epp); -static inline int amd_pstate_set_epp(struct amd_cpudata *cpudata, u8 epp) +static inline int amd_pstate_set_epp(struct cpufreq_policy *policy, u8 epp) { - return static_call(amd_pstate_set_epp)(cpudata, epp); + return static_call(amd_pstate_set_epp)(policy, epp); } -static int shmem_set_epp(struct amd_cpudata *cpudata, u8 epp) +static int shmem_set_epp(struct cpufreq_policy *policy, u8 epp) { - int ret; + struct amd_cpudata *cpudata = policy->driver_data; struct cppc_perf_ctrls perf_ctrls; + u8 epp_cached; + u64 value; + int ret; - if (epp == cpudata->epp_cached) + + epp_cached = FIELD_GET(AMD_CPPC_EPP_PERF_MASK, cpudata->cppc_req_cached); + if (trace_amd_pstate_epp_perf_enabled()) { + union perf_cached perf = cpudata->perf; + + trace_amd_pstate_epp_perf(cpudata->cpu, perf.highest_perf, + epp, + FIELD_GET(AMD_CPPC_MIN_PERF_MASK, + cpudata->cppc_req_cached), + FIELD_GET(AMD_CPPC_MAX_PERF_MASK, + cpudata->cppc_req_cached), + policy->boost_enabled, + epp != epp_cached); + } + + if (epp == epp_cached) return 0; perf_ctrls.energy_perf = epp; @@ -311,106 +360,35 @@ static int shmem_set_epp(struct amd_cpudata *cpudata, u8 epp) pr_debug("failed to set energy perf value (%d)\n", ret); return ret; } - WRITE_ONCE(cpudata->epp_cached, epp); - - return ret; -} - -static int amd_pstate_set_energy_pref_index(struct cpufreq_policy *policy, - int pref_index) -{ - struct amd_cpudata *cpudata = policy->driver_data; - u8 epp; - - if (!pref_index) - epp = cpudata->epp_default; - else - epp = epp_values[pref_index]; - - if (epp > 0 && cpudata->policy == CPUFREQ_POLICY_PERFORMANCE) { - pr_debug("EPP cannot be set under performance policy\n"); - return -EBUSY; - } - if (trace_amd_pstate_epp_perf_enabled()) { - trace_amd_pstate_epp_perf(cpudata->cpu, cpudata->highest_perf, - epp, - FIELD_GET(AMD_CPPC_MIN_PERF_MASK, cpudata->cppc_req_cached), - FIELD_GET(AMD_CPPC_MAX_PERF_MASK, cpudata->cppc_req_cached), - policy->boost_enabled); - } + value = READ_ONCE(cpudata->cppc_req_cached); + value &= ~AMD_CPPC_EPP_PERF_MASK; + value |= FIELD_PREP(AMD_CPPC_EPP_PERF_MASK, epp); + WRITE_ONCE(cpudata->cppc_req_cached, value); - return amd_pstate_set_epp(cpudata, epp); + return ret; } -static inline int msr_cppc_enable(bool enable) +static inline int msr_cppc_enable(struct cpufreq_policy *policy) { - int ret, cpu; - unsigned long logical_proc_id_mask = 0; - - /* - * MSR_AMD_CPPC_ENABLE is write-once, once set it cannot be cleared. - */ - if (!enable) - return 0; - - if (enable == cppc_enabled) - return 0; - - for_each_present_cpu(cpu) { - unsigned long logical_id = topology_logical_package_id(cpu); - - if (test_bit(logical_id, &logical_proc_id_mask)) - continue; - - set_bit(logical_id, &logical_proc_id_mask); - - ret = wrmsrl_safe_on_cpu(cpu, MSR_AMD_CPPC_ENABLE, - enable); - if (ret) - return ret; - } - - cppc_enabled = enable; - return 0; + return wrmsrl_safe_on_cpu(policy->cpu, MSR_AMD_CPPC_ENABLE, 1); } -static int shmem_cppc_enable(bool enable) +static int shmem_cppc_enable(struct cpufreq_policy *policy) { - int cpu, ret = 0; - struct cppc_perf_ctrls perf_ctrls; - - if (enable == cppc_enabled) - return 0; - - for_each_present_cpu(cpu) { - ret = cppc_set_enable(cpu, enable); - if (ret) - return ret; - - /* Enable autonomous mode for EPP */ - if (cppc_state == AMD_PSTATE_ACTIVE) { - /* Set desired perf as zero to allow EPP firmware control */ - perf_ctrls.desired_perf = 0; - ret = cppc_set_perf(cpu, &perf_ctrls); - if (ret) - return ret; - } - } - - cppc_enabled = enable; - return ret; + return cppc_set_enable(policy->cpu, 1); } DEFINE_STATIC_CALL(amd_pstate_cppc_enable, msr_cppc_enable); -static inline int amd_pstate_cppc_enable(bool enable) +static inline int amd_pstate_cppc_enable(struct cpufreq_policy *policy) { - return static_call(amd_pstate_cppc_enable)(enable); + return static_call(amd_pstate_cppc_enable)(policy); } static int msr_init_perf(struct amd_cpudata *cpudata) { + union perf_cached perf = READ_ONCE(cpudata->perf); u64 cap1, numerator; int ret = rdmsrl_safe_on_cpu(cpudata->cpu, MSR_AMD_CPPC_CAP1, @@ -422,19 +400,22 @@ static int msr_init_perf(struct amd_cpudata *cpudata) if (ret) return ret; - WRITE_ONCE(cpudata->highest_perf, numerator); - WRITE_ONCE(cpudata->max_limit_perf, numerator); - WRITE_ONCE(cpudata->nominal_perf, AMD_CPPC_NOMINAL_PERF(cap1)); - WRITE_ONCE(cpudata->lowest_nonlinear_perf, AMD_CPPC_LOWNONLIN_PERF(cap1)); - WRITE_ONCE(cpudata->lowest_perf, AMD_CPPC_LOWEST_PERF(cap1)); - WRITE_ONCE(cpudata->prefcore_ranking, AMD_CPPC_HIGHEST_PERF(cap1)); - WRITE_ONCE(cpudata->min_limit_perf, AMD_CPPC_LOWEST_PERF(cap1)); + perf.highest_perf = numerator; + perf.max_limit_perf = numerator; + perf.min_limit_perf = FIELD_GET(AMD_CPPC_LOWEST_PERF_MASK, cap1); + perf.nominal_perf = FIELD_GET(AMD_CPPC_NOMINAL_PERF_MASK, cap1); + perf.lowest_nonlinear_perf = FIELD_GET(AMD_CPPC_LOWNONLIN_PERF_MASK, cap1); + perf.lowest_perf = FIELD_GET(AMD_CPPC_LOWEST_PERF_MASK, cap1); + WRITE_ONCE(cpudata->perf, perf); + WRITE_ONCE(cpudata->prefcore_ranking, FIELD_GET(AMD_CPPC_HIGHEST_PERF_MASK, cap1)); + return 0; } static int shmem_init_perf(struct amd_cpudata *cpudata) { struct cppc_perf_caps cppc_perf; + union perf_cached perf = READ_ONCE(cpudata->perf); u64 numerator; int ret = cppc_get_perf_caps(cpudata->cpu, &cppc_perf); @@ -445,14 +426,14 @@ static int shmem_init_perf(struct amd_cpudata *cpudata) if (ret) return ret; - WRITE_ONCE(cpudata->highest_perf, numerator); - WRITE_ONCE(cpudata->max_limit_perf, numerator); - WRITE_ONCE(cpudata->nominal_perf, cppc_perf.nominal_perf); - WRITE_ONCE(cpudata->lowest_nonlinear_perf, - cppc_perf.lowest_nonlinear_perf); - WRITE_ONCE(cpudata->lowest_perf, cppc_perf.lowest_perf); + perf.highest_perf = numerator; + perf.max_limit_perf = numerator; + perf.min_limit_perf = cppc_perf.lowest_perf; + perf.nominal_perf = cppc_perf.nominal_perf; + perf.lowest_nonlinear_perf = cppc_perf.lowest_nonlinear_perf; + perf.lowest_perf = cppc_perf.lowest_perf; + WRITE_ONCE(cpudata->perf, perf); WRITE_ONCE(cpudata->prefcore_ranking, cppc_perf.highest_perf); - WRITE_ONCE(cpudata->min_limit_perf, cppc_perf.lowest_perf); if (cppc_state == AMD_PSTATE_ACTIVE) return 0; @@ -479,23 +460,56 @@ static inline int amd_pstate_init_perf(struct amd_cpudata *cpudata) return static_call(amd_pstate_init_perf)(cpudata); } -static int shmem_update_perf(struct amd_cpudata *cpudata, u8 min_perf, +static int shmem_update_perf(struct cpufreq_policy *policy, u8 min_perf, u8 des_perf, u8 max_perf, u8 epp, bool fast_switch) { + struct amd_cpudata *cpudata = policy->driver_data; struct cppc_perf_ctrls perf_ctrls; + u64 value, prev; + int ret; if (cppc_state == AMD_PSTATE_ACTIVE) { - int ret = shmem_set_epp(cpudata, epp); + int ret = shmem_set_epp(policy, epp); if (ret) return ret; } + value = prev = READ_ONCE(cpudata->cppc_req_cached); + + value &= ~(AMD_CPPC_MAX_PERF_MASK | AMD_CPPC_MIN_PERF_MASK | + AMD_CPPC_DES_PERF_MASK | AMD_CPPC_EPP_PERF_MASK); + value |= FIELD_PREP(AMD_CPPC_MAX_PERF_MASK, max_perf); + value |= FIELD_PREP(AMD_CPPC_DES_PERF_MASK, des_perf); + value |= FIELD_PREP(AMD_CPPC_MIN_PERF_MASK, min_perf); + value |= FIELD_PREP(AMD_CPPC_EPP_PERF_MASK, epp); + + if (trace_amd_pstate_epp_perf_enabled()) { + union perf_cached perf = READ_ONCE(cpudata->perf); + + trace_amd_pstate_epp_perf(cpudata->cpu, + perf.highest_perf, + epp, + min_perf, + max_perf, + policy->boost_enabled, + value != prev); + } + + if (value == prev) + return 0; + perf_ctrls.max_perf = max_perf; perf_ctrls.min_perf = min_perf; perf_ctrls.desired_perf = des_perf; - return cppc_set_perf(cpudata->cpu, &perf_ctrls); + ret = cppc_set_perf(cpudata->cpu, &perf_ctrls); + if (ret) + return ret; + + WRITE_ONCE(cpudata->cppc_req_cached, value); + + return 0; } static inline bool amd_pstate_sample(struct amd_cpudata *cpudata) @@ -534,17 +548,15 @@ static inline bool amd_pstate_sample(struct amd_cpudata *cpudata) static void amd_pstate_update(struct amd_cpudata *cpudata, u8 min_perf, u8 des_perf, u8 max_perf, bool fast_switch, int gov_flags) { - unsigned long max_freq; - struct cpufreq_policy *policy = cpufreq_cpu_get(cpudata->cpu); - u8 nominal_perf = READ_ONCE(cpudata->nominal_perf); + struct cpufreq_policy *policy __free(put_cpufreq_policy) = cpufreq_cpu_get(cpudata->cpu); + union perf_cached perf = READ_ONCE(cpudata->perf); if (!policy) return; des_perf = clamp_t(u8, des_perf, min_perf, max_perf); - max_freq = READ_ONCE(cpudata->max_limit_freq); - policy->cur = div_u64(des_perf * max_freq, max_perf); + policy->cur = perf_to_freq(perf, cpudata->nominal_freq, des_perf); if ((cppc_state == AMD_PSTATE_GUIDED) && (gov_flags & CPUFREQ_GOV_DYNAMIC_SWITCHING)) { min_perf = des_perf; @@ -553,7 +565,7 @@ static void amd_pstate_update(struct amd_cpudata *cpudata, u8 min_perf, /* limit the max perf when core performance boost feature is disabled */ if (!cpudata->boost_supported) - max_perf = min_t(u8, nominal_perf, max_perf); + max_perf = min_t(u8, perf.nominal_perf, max_perf); if (trace_amd_pstate_perf_enabled() && amd_pstate_sample(cpudata)) { trace_amd_pstate_perf(min_perf, des_perf, max_perf, cpudata->freq, @@ -561,9 +573,7 @@ static void amd_pstate_update(struct amd_cpudata *cpudata, u8 min_perf, cpudata->cpu, fast_switch); } - amd_pstate_update_perf(cpudata, min_perf, des_perf, max_perf, 0, fast_switch); - - cpufreq_cpu_put(policy); + amd_pstate_update_perf(policy, min_perf, des_perf, max_perf, 0, fast_switch); } static int amd_pstate_verify(struct cpufreq_policy_data *policy_data) @@ -575,7 +585,8 @@ static int amd_pstate_verify(struct cpufreq_policy_data *policy_data) * amd-pstate qos_requests. */ if (policy_data->min == FREQ_QOS_MIN_DEFAULT_VALUE) { - struct cpufreq_policy *policy = cpufreq_cpu_get(policy_data->cpu); + struct cpufreq_policy *policy __free(put_cpufreq_policy) = + cpufreq_cpu_get(policy_data->cpu); struct amd_cpudata *cpudata; if (!policy) @@ -583,57 +594,51 @@ static int amd_pstate_verify(struct cpufreq_policy_data *policy_data) cpudata = policy->driver_data; policy_data->min = cpudata->lowest_nonlinear_freq; - cpufreq_cpu_put(policy); } cpufreq_verify_within_cpu_limits(policy_data); - pr_debug("policy_max =%d, policy_min=%d\n", policy_data->max, policy_data->min); return 0; } -static int amd_pstate_update_min_max_limit(struct cpufreq_policy *policy) +static void amd_pstate_update_min_max_limit(struct cpufreq_policy *policy) { - u8 max_limit_perf, min_limit_perf, max_perf; - u32 max_freq; struct amd_cpudata *cpudata = policy->driver_data; + union perf_cached perf = READ_ONCE(cpudata->perf); - max_perf = READ_ONCE(cpudata->highest_perf); - max_freq = READ_ONCE(cpudata->max_freq); - max_limit_perf = div_u64(policy->max * max_perf, max_freq); - min_limit_perf = div_u64(policy->min * max_perf, max_freq); - - if (cpudata->policy == CPUFREQ_POLICY_PERFORMANCE) - min_limit_perf = min(cpudata->nominal_perf, max_limit_perf); - - WRITE_ONCE(cpudata->max_limit_perf, max_limit_perf); - WRITE_ONCE(cpudata->min_limit_perf, min_limit_perf); + perf.max_limit_perf = freq_to_perf(perf, cpudata->nominal_freq, policy->max); WRITE_ONCE(cpudata->max_limit_freq, policy->max); - WRITE_ONCE(cpudata->min_limit_freq, policy->min); - return 0; + if (cpudata->policy == CPUFREQ_POLICY_PERFORMANCE) { + perf.min_limit_perf = min(perf.nominal_perf, perf.max_limit_perf); + WRITE_ONCE(cpudata->min_limit_freq, min(cpudata->nominal_freq, cpudata->max_limit_freq)); + } else { + perf.min_limit_perf = freq_to_perf(perf, cpudata->nominal_freq, policy->min); + WRITE_ONCE(cpudata->min_limit_freq, policy->min); + } + + WRITE_ONCE(cpudata->perf, perf); } static int amd_pstate_update_freq(struct cpufreq_policy *policy, unsigned int target_freq, bool fast_switch) { struct cpufreq_freqs freqs; - struct amd_cpudata *cpudata = policy->driver_data; - u8 des_perf, cap_perf; + struct amd_cpudata *cpudata; + union perf_cached perf; + u8 des_perf; - if (!cpudata->max_freq) - return -ENODEV; + cpudata = policy->driver_data; if (policy->min != cpudata->min_limit_freq || policy->max != cpudata->max_limit_freq) amd_pstate_update_min_max_limit(policy); - cap_perf = READ_ONCE(cpudata->highest_perf); + perf = READ_ONCE(cpudata->perf); freqs.old = policy->cur; freqs.new = target_freq; - des_perf = DIV_ROUND_CLOSEST(target_freq * cap_perf, - cpudata->max_freq); + des_perf = freq_to_perf(perf, cpudata->nominal_freq, target_freq); WARN_ON(fast_switch && !policy->fast_switch_enabled); /* @@ -644,8 +649,8 @@ static int amd_pstate_update_freq(struct cpufreq_policy *policy, if (!fast_switch) cpufreq_freq_transition_begin(policy, &freqs); - amd_pstate_update(cpudata, cpudata->min_limit_perf, des_perf, - cpudata->max_limit_perf, fast_switch, + amd_pstate_update(cpudata, perf.min_limit_perf, des_perf, + perf.max_limit_perf, fast_switch, policy->governor->flags); if (!fast_switch) @@ -674,9 +679,10 @@ static void amd_pstate_adjust_perf(unsigned int cpu, unsigned long target_perf, unsigned long capacity) { - u8 max_perf, min_perf, des_perf, cap_perf, min_limit_perf; - struct cpufreq_policy *policy = cpufreq_cpu_get(cpu); + u8 max_perf, min_perf, des_perf, cap_perf; + struct cpufreq_policy *policy __free(put_cpufreq_policy) = cpufreq_cpu_get(cpu); struct amd_cpudata *cpudata; + union perf_cached perf; if (!policy) return; @@ -686,8 +692,8 @@ static void amd_pstate_adjust_perf(unsigned int cpu, if (policy->min != cpudata->min_limit_freq || policy->max != cpudata->max_limit_freq) amd_pstate_update_min_max_limit(policy); - cap_perf = READ_ONCE(cpudata->highest_perf); - min_limit_perf = READ_ONCE(cpudata->min_limit_perf); + perf = READ_ONCE(cpudata->perf); + cap_perf = perf.highest_perf; des_perf = cap_perf; if (target_perf < capacity) @@ -698,28 +704,26 @@ static void amd_pstate_adjust_perf(unsigned int cpu, else min_perf = cap_perf; - if (min_perf < min_limit_perf) - min_perf = min_limit_perf; + if (min_perf < perf.min_limit_perf) + min_perf = perf.min_limit_perf; - max_perf = cpudata->max_limit_perf; + max_perf = perf.max_limit_perf; if (max_perf < min_perf) max_perf = min_perf; - des_perf = clamp_t(unsigned long, des_perf, min_perf, max_perf); - amd_pstate_update(cpudata, min_perf, des_perf, max_perf, true, policy->governor->flags); - cpufreq_cpu_put(policy); } static int amd_pstate_cpu_boost_update(struct cpufreq_policy *policy, bool on) { struct amd_cpudata *cpudata = policy->driver_data; + union perf_cached perf = READ_ONCE(cpudata->perf); u32 nominal_freq, max_freq; int ret = 0; nominal_freq = READ_ONCE(cpudata->nominal_freq); - max_freq = READ_ONCE(cpudata->max_freq); + max_freq = perf_to_freq(perf, cpudata->nominal_freq, perf.highest_perf); if (on) policy->cpuinfo.max_freq = max_freq; @@ -746,7 +750,6 @@ static int amd_pstate_set_boost(struct cpufreq_policy *policy, int state) pr_err("Boost mode is not supported by this processor or SBIOS\n"); return -EOPNOTSUPP; } - guard(mutex)(&amd_pstate_driver_lock); ret = amd_pstate_cpu_boost_update(policy, state); refresh_frequency_limits(policy); @@ -791,19 +794,9 @@ static void amd_perf_ctl_reset(unsigned int cpu) wrmsrl_on_cpu(cpu, MSR_AMD_PERF_CTL, 0); } -/* - * Set amd-pstate preferred core enable can't be done directly from cpufreq callbacks - * due to locking, so queue the work for later. - */ -static void amd_pstste_sched_prefcore_workfn(struct work_struct *work) -{ - sched_set_itmt_support(); -} -static DECLARE_WORK(sched_prefcore_work, amd_pstste_sched_prefcore_workfn); - #define CPPC_MAX_PERF U8_MAX -static void amd_pstate_init_prefcore(struct amd_cpudata *cpudata) +static void amd_pstate_init_asym_prio(struct amd_cpudata *cpudata) { /* user disabled or not detected */ if (!amd_pstate_prefcore) @@ -811,54 +804,38 @@ static void amd_pstate_init_prefcore(struct amd_cpudata *cpudata) cpudata->hw_prefcore = true; - /* - * The priorities can be set regardless of whether or not - * sched_set_itmt_support(true) has been called and it is valid to - * update them at any time after it has been called. - */ + /* The priorities must be initialized before ITMT support can be toggled on. */ sched_set_itmt_core_prio((int)READ_ONCE(cpudata->prefcore_ranking), cpudata->cpu); - - schedule_work(&sched_prefcore_work); } static void amd_pstate_update_limits(unsigned int cpu) { - struct cpufreq_policy *policy = NULL; + struct cpufreq_policy *policy __free(put_cpufreq_policy) = cpufreq_cpu_get(cpu); struct amd_cpudata *cpudata; u32 prev_high = 0, cur_high = 0; - int ret; bool highest_perf_changed = false; if (!amd_pstate_prefcore) return; - policy = cpufreq_cpu_get(cpu); if (!policy) return; - cpudata = policy->driver_data; - - guard(mutex)(&amd_pstate_driver_lock); - - ret = amd_get_highest_perf(cpu, &cur_high); - if (ret) { - cpufreq_cpu_put(policy); + if (amd_get_highest_perf(cpu, &cur_high)) return; - } + + cpudata = policy->driver_data; prev_high = READ_ONCE(cpudata->prefcore_ranking); highest_perf_changed = (prev_high != cur_high); if (highest_perf_changed) { WRITE_ONCE(cpudata->prefcore_ranking, cur_high); - if (cur_high < CPPC_MAX_PERF) + if (cur_high < CPPC_MAX_PERF) { sched_set_itmt_core_prio((int)cur_high, cpu); + sched_update_asym_prefer_cpu(cpu, prev_high, cur_high); + } } - cpufreq_cpu_put(policy); - - if (!highest_perf_changed) - cpufreq_update_policy(cpu); - } /* @@ -896,48 +873,45 @@ static u32 amd_pstate_get_transition_latency(unsigned int cpu) } /* - * amd_pstate_init_freq: Initialize the max_freq, min_freq, - * nominal_freq and lowest_nonlinear_freq for - * the @cpudata object. + * amd_pstate_init_freq: Initialize the nominal_freq and lowest_nonlinear_freq + * for the @cpudata object. * - * Requires: highest_perf, lowest_perf, nominal_perf and - * lowest_nonlinear_perf members of @cpudata to be - * initialized. + * Requires: all perf members of @cpudata to be initialized. * - * Returns 0 on success, non-zero value on failure. + * Returns 0 on success, non-zero value on failure. */ static int amd_pstate_init_freq(struct amd_cpudata *cpudata) { - int ret; - u32 min_freq, max_freq; - u8 highest_perf, nominal_perf, lowest_nonlinear_perf; - u32 nominal_freq, lowest_nonlinear_freq; + u32 min_freq, max_freq, nominal_freq, lowest_nonlinear_freq; struct cppc_perf_caps cppc_perf; + union perf_cached perf; + int ret; ret = cppc_get_perf_caps(cpudata->cpu, &cppc_perf); if (ret) return ret; + perf = READ_ONCE(cpudata->perf); - if (quirks && quirks->lowest_freq) + if (quirks && quirks->nominal_freq) + nominal_freq = quirks->nominal_freq; + else + nominal_freq = cppc_perf.nominal_freq; + nominal_freq *= 1000; + + if (quirks && quirks->lowest_freq) { min_freq = quirks->lowest_freq; - else + perf.lowest_perf = freq_to_perf(perf, nominal_freq, min_freq); + WRITE_ONCE(cpudata->perf, perf); + } else min_freq = cppc_perf.lowest_freq; - if (quirks && quirks->nominal_freq) - nominal_freq = quirks->nominal_freq; - else - nominal_freq = cppc_perf.nominal_freq; + min_freq *= 1000; - highest_perf = READ_ONCE(cpudata->highest_perf); - nominal_perf = READ_ONCE(cpudata->nominal_perf); - max_freq = div_u64((u64)highest_perf * nominal_freq, nominal_perf); + WRITE_ONCE(cpudata->nominal_freq, nominal_freq); - lowest_nonlinear_perf = READ_ONCE(cpudata->lowest_nonlinear_perf); - lowest_nonlinear_freq = div_u64((u64)nominal_freq * lowest_nonlinear_perf, nominal_perf); - WRITE_ONCE(cpudata->min_freq, min_freq * 1000); - WRITE_ONCE(cpudata->lowest_nonlinear_freq, lowest_nonlinear_freq * 1000); - WRITE_ONCE(cpudata->nominal_freq, nominal_freq * 1000); - WRITE_ONCE(cpudata->max_freq, max_freq * 1000); + max_freq = perf_to_freq(perf, nominal_freq, perf.highest_perf); + lowest_nonlinear_freq = perf_to_freq(perf, nominal_freq, perf.lowest_nonlinear_perf); + WRITE_ONCE(cpudata->lowest_nonlinear_freq, lowest_nonlinear_freq); /** * Below values need to be initialized correctly, otherwise driver will fail to load @@ -962,9 +936,10 @@ static int amd_pstate_init_freq(struct amd_cpudata *cpudata) static int amd_pstate_cpu_init(struct cpufreq_policy *policy) { - int min_freq, max_freq, ret; - struct device *dev; struct amd_cpudata *cpudata; + union perf_cached perf; + struct device *dev; + int ret; /* * Resetting PERF_CTL_MSR will put the CPU in P0 frequency, @@ -985,7 +960,7 @@ static int amd_pstate_cpu_init(struct cpufreq_policy *policy) if (ret) goto free_cpudata1; - amd_pstate_init_prefcore(cpudata); + amd_pstate_init_asym_prio(cpudata); ret = amd_pstate_init_freq(cpudata); if (ret) @@ -995,17 +970,21 @@ static int amd_pstate_cpu_init(struct cpufreq_policy *policy) if (ret) goto free_cpudata1; - min_freq = READ_ONCE(cpudata->min_freq); - max_freq = READ_ONCE(cpudata->max_freq); - policy->cpuinfo.transition_latency = amd_pstate_get_transition_latency(policy->cpu); policy->transition_delay_us = amd_pstate_get_transition_delay_us(policy->cpu); - policy->min = min_freq; - policy->max = max_freq; + perf = READ_ONCE(cpudata->perf); - policy->cpuinfo.min_freq = min_freq; - policy->cpuinfo.max_freq = max_freq; + policy->cpuinfo.min_freq = policy->min = perf_to_freq(perf, + cpudata->nominal_freq, + perf.lowest_perf); + policy->cpuinfo.max_freq = policy->max = perf_to_freq(perf, + cpudata->nominal_freq, + perf.highest_perf); + + ret = amd_pstate_cppc_enable(policy); + if (ret) + goto free_cpudata1; policy->boost_enabled = READ_ONCE(cpudata->boost_supported); @@ -1029,9 +1008,6 @@ static int amd_pstate_cpu_init(struct cpufreq_policy *policy) goto free_cpudata2; } - cpudata->max_limit_freq = max_freq; - cpudata->min_limit_freq = min_freq; - policy->driver_data = cpudata; if (!current_pstate_driver->adjust_perf) @@ -1042,6 +1018,7 @@ static int amd_pstate_cpu_init(struct cpufreq_policy *policy) free_cpudata2: freq_qos_remove_request(&cpudata->req[0]); free_cpudata1: + pr_warn("Failed to initialize CPU %d: %d\n", policy->cpu, ret); kfree(cpudata); return ret; } @@ -1056,28 +1033,6 @@ static void amd_pstate_cpu_exit(struct cpufreq_policy *policy) kfree(cpudata); } -static int amd_pstate_cpu_resume(struct cpufreq_policy *policy) -{ - int ret; - - ret = amd_pstate_cppc_enable(true); - if (ret) - pr_err("failed to enable amd-pstate during resume, return %d\n", ret); - - return ret; -} - -static int amd_pstate_cpu_suspend(struct cpufreq_policy *policy) -{ - int ret; - - ret = amd_pstate_cppc_enable(false); - if (ret) - pr_err("failed to disable amd-pstate during suspend, return %d\n", ret); - - return ret; -} - /* Sysfs attributes */ /* @@ -1088,27 +1043,27 @@ static int amd_pstate_cpu_suspend(struct cpufreq_policy *policy) static ssize_t show_amd_pstate_max_freq(struct cpufreq_policy *policy, char *buf) { - int max_freq; - struct amd_cpudata *cpudata = policy->driver_data; + struct amd_cpudata *cpudata; + union perf_cached perf; - max_freq = READ_ONCE(cpudata->max_freq); - if (max_freq < 0) - return max_freq; + cpudata = policy->driver_data; + perf = READ_ONCE(cpudata->perf); - return sysfs_emit(buf, "%u\n", max_freq); + return sysfs_emit(buf, "%u\n", + perf_to_freq(perf, cpudata->nominal_freq, perf.highest_perf)); } static ssize_t show_amd_pstate_lowest_nonlinear_freq(struct cpufreq_policy *policy, char *buf) { - int freq; - struct amd_cpudata *cpudata = policy->driver_data; + struct amd_cpudata *cpudata; + union perf_cached perf; - freq = READ_ONCE(cpudata->lowest_nonlinear_freq); - if (freq < 0) - return freq; + cpudata = policy->driver_data; + perf = READ_ONCE(cpudata->perf); - return sysfs_emit(buf, "%u\n", freq); + return sysfs_emit(buf, "%u\n", + perf_to_freq(perf, cpudata->nominal_freq, perf.lowest_nonlinear_perf)); } /* @@ -1118,12 +1073,11 @@ static ssize_t show_amd_pstate_lowest_nonlinear_freq(struct cpufreq_policy *poli static ssize_t show_amd_pstate_highest_perf(struct cpufreq_policy *policy, char *buf) { - u8 perf; - struct amd_cpudata *cpudata = policy->driver_data; + struct amd_cpudata *cpudata; - perf = READ_ONCE(cpudata->highest_perf); + cpudata = policy->driver_data; - return sysfs_emit(buf, "%u\n", perf); + return sysfs_emit(buf, "%u\n", cpudata->perf.highest_perf); } static ssize_t show_amd_pstate_prefcore_ranking(struct cpufreq_policy *policy, @@ -1170,8 +1124,10 @@ static ssize_t show_energy_performance_available_preferences( static ssize_t store_energy_performance_preference( struct cpufreq_policy *policy, const char *buf, size_t count) { + struct amd_cpudata *cpudata = policy->driver_data; char str_preference[21]; ssize_t ret; + u8 epp; ret = sscanf(buf, "%20s", str_preference); if (ret != 1) @@ -1181,9 +1137,17 @@ static ssize_t store_energy_performance_preference( if (ret < 0) return -EINVAL; - guard(mutex)(&amd_pstate_limits_lock); + if (!ret) + epp = cpudata->epp_default; + else + epp = epp_values[ret]; + + if (epp > 0 && policy->policy == CPUFREQ_POLICY_PERFORMANCE) { + pr_debug("EPP cannot be set under performance policy\n"); + return -EBUSY; + } - ret = amd_pstate_set_energy_pref_index(policy, ret); + ret = amd_pstate_set_epp(policy, epp); return ret ? ret : count; } @@ -1192,9 +1156,11 @@ static ssize_t show_energy_performance_preference( struct cpufreq_policy *policy, char *buf) { struct amd_cpudata *cpudata = policy->driver_data; - u8 preference; + u8 preference, epp; - switch (cpudata->epp_cached) { + epp = FIELD_GET(AMD_CPPC_EPP_PERF_MASK, cpudata->cppc_req_cached); + + switch (epp) { case AMD_CPPC_EPP_PERFORMANCE: preference = EPP_INDEX_PERFORMANCE; break; @@ -1216,7 +1182,6 @@ static ssize_t show_energy_performance_preference( static void amd_pstate_driver_cleanup(void) { - amd_pstate_cppc_enable(false); cppc_state = AMD_PSTATE_DISABLE; current_pstate_driver = NULL; } @@ -1250,14 +1215,6 @@ static int amd_pstate_register_driver(int mode) cppc_state = mode; - ret = amd_pstate_cppc_enable(true); - if (ret) { - pr_err("failed to enable cppc during amd-pstate driver registration, return %d\n", - ret); - amd_pstate_driver_cleanup(); - return ret; - } - /* at least one CPU supports CPB */ current_pstate_driver->boost_enabled = cpu_feature_enabled(X86_FEATURE_CPB); @@ -1355,8 +1312,10 @@ int amd_pstate_update_status(const char *buf, size_t size) if (mode_idx < 0 || mode_idx >= AMD_PSTATE_MAX) return -EINVAL; - if (mode_state_machine[cppc_state][mode_idx]) + if (mode_state_machine[cppc_state][mode_idx]) { + guard(mutex)(&amd_pstate_driver_lock); return mode_state_machine[cppc_state][mode_idx](mode_idx); + } return 0; } @@ -1377,7 +1336,6 @@ static ssize_t status_store(struct device *a, struct device_attribute *b, char *p = memchr(buf, '\n', count); int ret; - guard(mutex)(&amd_pstate_driver_lock); ret = amd_pstate_update_status(buf, p ? p - buf : count); return ret < 0 ? ret : count; @@ -1453,10 +1411,11 @@ static bool amd_pstate_acpi_pm_profile_undefined(void) static int amd_pstate_epp_cpu_init(struct cpufreq_policy *policy) { - int min_freq, max_freq, ret; struct amd_cpudata *cpudata; + union perf_cached perf; struct device *dev; u64 value; + int ret; /* * Resetting PERF_CTL_MSR will put the CPU in P0 frequency, @@ -1477,7 +1436,7 @@ static int amd_pstate_epp_cpu_init(struct cpufreq_policy *policy) if (ret) goto free_cpudata1; - amd_pstate_init_prefcore(cpudata); + amd_pstate_init_asym_prio(cpudata); ret = amd_pstate_init_freq(cpudata); if (ret) @@ -1487,18 +1446,23 @@ static int amd_pstate_epp_cpu_init(struct cpufreq_policy *policy) if (ret) goto free_cpudata1; - min_freq = READ_ONCE(cpudata->min_freq); - max_freq = READ_ONCE(cpudata->max_freq); + perf = READ_ONCE(cpudata->perf); + + policy->cpuinfo.min_freq = policy->min = perf_to_freq(perf, + cpudata->nominal_freq, + perf.lowest_perf); + policy->cpuinfo.max_freq = policy->max = perf_to_freq(perf, + cpudata->nominal_freq, + perf.highest_perf); + policy->driver_data = cpudata; + + ret = amd_pstate_cppc_enable(policy); + if (ret) + goto free_cpudata1; - policy->cpuinfo.min_freq = min_freq; - policy->cpuinfo.max_freq = max_freq; /* It will be updated by governor */ policy->cur = policy->cpuinfo.min_freq; - policy->driver_data = cpudata; - - policy->min = policy->cpuinfo.min_freq; - policy->max = policy->cpuinfo.max_freq; policy->boost_enabled = READ_ONCE(cpudata->boost_supported); @@ -1520,13 +1484,8 @@ static int amd_pstate_epp_cpu_init(struct cpufreq_policy *policy) if (ret) return ret; WRITE_ONCE(cpudata->cppc_req_cached, value); - - ret = rdmsrl_on_cpu(cpudata->cpu, MSR_AMD_CPPC_CAP1, &value); - if (ret) - return ret; - WRITE_ONCE(cpudata->cppc_cap1_cached, value); } - ret = amd_pstate_set_epp(cpudata, cpudata->epp_default); + ret = amd_pstate_set_epp(policy, cpudata->epp_default); if (ret) return ret; @@ -1535,6 +1494,7 @@ static int amd_pstate_epp_cpu_init(struct cpufreq_policy *policy) return 0; free_cpudata1: + pr_warn("Failed to initialize CPU %d: %d\n", policy->cpu, ret); kfree(cpudata); return ret; } @@ -1554,24 +1514,21 @@ static void amd_pstate_epp_cpu_exit(struct cpufreq_policy *policy) static int amd_pstate_epp_update_limit(struct cpufreq_policy *policy) { struct amd_cpudata *cpudata = policy->driver_data; + union perf_cached perf; u8 epp; - amd_pstate_update_min_max_limit(policy); + if (policy->min != cpudata->min_limit_freq || policy->max != cpudata->max_limit_freq) + amd_pstate_update_min_max_limit(policy); if (cpudata->policy == CPUFREQ_POLICY_PERFORMANCE) epp = 0; else - epp = READ_ONCE(cpudata->epp_cached); + epp = FIELD_GET(AMD_CPPC_EPP_PERF_MASK, cpudata->cppc_req_cached); - if (trace_amd_pstate_epp_perf_enabled()) { - trace_amd_pstate_epp_perf(cpudata->cpu, cpudata->highest_perf, epp, - cpudata->min_limit_perf, - cpudata->max_limit_perf, - policy->boost_enabled); - } + perf = READ_ONCE(cpudata->perf); - return amd_pstate_update_perf(cpudata, cpudata->min_limit_perf, 0U, - cpudata->max_limit_perf, epp, false); + return amd_pstate_update_perf(policy, perf.min_limit_perf, 0U, + perf.max_limit_perf, epp, false); } static int amd_pstate_epp_set_policy(struct cpufreq_policy *policy) @@ -1582,9 +1539,6 @@ static int amd_pstate_epp_set_policy(struct cpufreq_policy *policy) if (!policy->cpuinfo.max_freq) return -ENODEV; - pr_debug("set_policy: cpuinfo.max %u policy->max %u\n", - policy->cpuinfo.max_freq, policy->max); - cpudata->policy = policy->policy; ret = amd_pstate_epp_update_limit(policy); @@ -1600,73 +1554,21 @@ static int amd_pstate_epp_set_policy(struct cpufreq_policy *policy) return 0; } -static int amd_pstate_epp_reenable(struct cpufreq_policy *policy) -{ - struct amd_cpudata *cpudata = policy->driver_data; - u8 max_perf; - int ret; - - ret = amd_pstate_cppc_enable(true); - if (ret) - pr_err("failed to enable amd pstate during resume, return %d\n", ret); - - max_perf = READ_ONCE(cpudata->highest_perf); - - if (trace_amd_pstate_epp_perf_enabled()) { - trace_amd_pstate_epp_perf(cpudata->cpu, cpudata->highest_perf, - cpudata->epp_cached, - FIELD_GET(AMD_CPPC_MIN_PERF_MASK, cpudata->cppc_req_cached), - max_perf, policy->boost_enabled); - } - - return amd_pstate_epp_update_limit(policy); -} - static int amd_pstate_epp_cpu_online(struct cpufreq_policy *policy) { - struct amd_cpudata *cpudata = policy->driver_data; - int ret; + pr_debug("AMD CPU Core %d going online\n", policy->cpu); - pr_debug("AMD CPU Core %d going online\n", cpudata->cpu); - - ret = amd_pstate_epp_reenable(policy); - if (ret) - return ret; - cpudata->suspended = false; - - return 0; + return amd_pstate_cppc_enable(policy); } static int amd_pstate_epp_cpu_offline(struct cpufreq_policy *policy) { - struct amd_cpudata *cpudata = policy->driver_data; - u8 min_perf; - - if (cpudata->suspended) - return 0; - - min_perf = READ_ONCE(cpudata->lowest_perf); - - guard(mutex)(&amd_pstate_limits_lock); - - if (trace_amd_pstate_epp_perf_enabled()) { - trace_amd_pstate_epp_perf(cpudata->cpu, cpudata->highest_perf, - AMD_CPPC_EPP_BALANCE_POWERSAVE, - min_perf, min_perf, policy->boost_enabled); - } - - return amd_pstate_update_perf(cpudata, min_perf, 0, min_perf, - AMD_CPPC_EPP_BALANCE_POWERSAVE, false); + return 0; } static int amd_pstate_epp_suspend(struct cpufreq_policy *policy) { struct amd_cpudata *cpudata = policy->driver_data; - int ret; - - /* avoid suspending when EPP is not enabled */ - if (cppc_state != AMD_PSTATE_ACTIVE) - return 0; /* invalidate to ensure it's rewritten during resume */ cpudata->cppc_req_cached = 0; @@ -1674,11 +1576,6 @@ static int amd_pstate_epp_suspend(struct cpufreq_policy *policy) /* set this flag to avoid setting core offline*/ cpudata->suspended = true; - /* disable CPPC in lowlevel firmware */ - ret = amd_pstate_cppc_enable(false); - if (ret) - pr_err("failed to suspend, return %d\n", ret); - return 0; } @@ -1687,10 +1584,12 @@ static int amd_pstate_epp_resume(struct cpufreq_policy *policy) struct amd_cpudata *cpudata = policy->driver_data; if (cpudata->suspended) { - guard(mutex)(&amd_pstate_limits_lock); + int ret; /* enable amd pstate from suspend state*/ - amd_pstate_epp_reenable(policy); + ret = amd_pstate_epp_update_limit(policy); + if (ret) + return ret; cpudata->suspended = false; } @@ -1705,8 +1604,6 @@ static struct cpufreq_driver amd_pstate_driver = { .fast_switch = amd_pstate_fast_switch, .init = amd_pstate_cpu_init, .exit = amd_pstate_cpu_exit, - .suspend = amd_pstate_cpu_suspend, - .resume = amd_pstate_cpu_resume, .set_boost = amd_pstate_set_boost, .update_limits = amd_pstate_update_limits, .name = "amd-pstate", @@ -1869,11 +1766,14 @@ static int __init amd_pstate_init(void) } } + /* Enable ITMT support once all CPUs have initialized their asym priorities. */ + if (amd_pstate_prefcore) + sched_set_itmt_support(); + return ret; global_attr_free: cpufreq_unregister_driver(current_pstate_driver); - amd_pstate_cppc_enable(false); return ret; } device_initcall(amd_pstate_init); diff --git a/drivers/cpufreq/amd-pstate.h b/drivers/cpufreq/amd-pstate.h index 19d405c6d805..fbe1c08d3f06 100644 --- a/drivers/cpufreq/amd-pstate.h +++ b/drivers/cpufreq/amd-pstate.h @@ -13,6 +13,36 @@ /********************************************************************* * AMD P-state INTERFACE * *********************************************************************/ + +/** + * union perf_cached - A union to cache performance-related data. + * @highest_perf: the maximum performance an individual processor may reach, + * assuming ideal conditions + * For platforms that support the preferred core feature, the highest_perf value maybe + * configured to any value in the range 166-255 by the firmware (because the preferred + * core ranking is encoded in the highest_perf value). To maintain consistency across + * all platforms, we split the highest_perf and preferred core ranking values into + * cpudata->perf.highest_perf and cpudata->prefcore_ranking. + * @nominal_perf: the maximum sustained performance level of the processor, + * assuming ideal operating conditions + * @lowest_nonlinear_perf: the lowest performance level at which nonlinear power + * savings are achieved + * @lowest_perf: the absolute lowest performance level of the processor + * @min_limit_perf: Cached value of the performance corresponding to policy->min + * @max_limit_perf: Cached value of the performance corresponding to policy->max + */ +union perf_cached { + struct { + u8 highest_perf; + u8 nominal_perf; + u8 lowest_nonlinear_perf; + u8 lowest_perf; + u8 min_limit_perf; + u8 max_limit_perf; + }; + u64 val; +}; + /** * struct amd_aperf_mperf * @aperf: actual performance frequency clock count @@ -30,24 +60,11 @@ struct amd_aperf_mperf { * @cpu: CPU number * @req: constraint request to apply * @cppc_req_cached: cached performance request hints - * @highest_perf: the maximum performance an individual processor may reach, - * assuming ideal conditions - * For platforms that do not support the preferred core feature, the - * highest_pef may be configured with 166 or 255, to avoid max frequency - * calculated wrongly. we take the fixed value as the highest_perf. - * @nominal_perf: the maximum sustained performance level of the processor, - * assuming ideal operating conditions - * @lowest_nonlinear_perf: the lowest performance level at which nonlinear power - * savings are achieved - * @lowest_perf: the absolute lowest performance level of the processor + * @perf: cached performance-related data * @prefcore_ranking: the preferred core ranking, the higher value indicates a higher * priority. - * @min_limit_perf: Cached value of the performance corresponding to policy->min - * @max_limit_perf: Cached value of the performance corresponding to policy->max * @min_limit_freq: Cached value of policy->min (in khz) * @max_limit_freq: Cached value of policy->max (in khz) - * @max_freq: the frequency (in khz) that mapped to highest_perf - * @min_freq: the frequency (in khz) that mapped to lowest_perf * @nominal_freq: the frequency (in khz) that mapped to nominal_perf * @lowest_nonlinear_freq: the frequency (in khz) that mapped to lowest_nonlinear_perf * @cur: Difference of Aperf/Mperf/tsc count between last and current sample @@ -59,7 +76,6 @@ struct amd_aperf_mperf { * AMD P-State driver supports preferred core featue. * @epp_cached: Cached CPPC energy-performance preference value * @policy: Cpufreq policy value - * @cppc_cap1_cached Cached MSR_AMD_CPPC_CAP1 register value * * The amd_cpudata is key private data for each CPU thread in AMD P-State, and * represents all the attributes and goals that AMD P-State requests at runtime. @@ -70,18 +86,11 @@ struct amd_cpudata { struct freq_qos_request req[2]; u64 cppc_req_cached; - u8 highest_perf; - u8 nominal_perf; - u8 lowest_nonlinear_perf; - u8 lowest_perf; - u8 prefcore_ranking; - u8 min_limit_perf; - u8 max_limit_perf; - u32 min_limit_freq; - u32 max_limit_freq; + union perf_cached perf; - u32 max_freq; - u32 min_freq; + u8 prefcore_ranking; + u32 min_limit_freq; + u32 max_limit_freq; u32 nominal_freq; u32 lowest_nonlinear_freq; @@ -93,9 +102,7 @@ struct amd_cpudata { bool hw_prefcore; /* EPP feature related attributes*/ - u8 epp_cached; u32 policy; - u64 cppc_cap1_cached; bool suspended; u8 epp_default; }; diff --git a/include/linux/cpufreq.h b/include/linux/cpufreq.h index 73024830bd73..efad63f78785 100644 --- a/include/linux/cpufreq.h +++ b/include/linux/cpufreq.h @@ -213,6 +213,9 @@ static inline struct cpufreq_policy *cpufreq_cpu_get(unsigned int cpu) static inline void cpufreq_cpu_put(struct cpufreq_policy *policy) { } #endif +/* Scope based cleanup macro for cpufreq_policy kobject reference counting */ +DEFINE_FREE(put_cpufreq_policy, struct cpufreq_policy *, if (_T) cpufreq_cpu_put(_T)) + static inline bool policy_is_inactive(struct cpufreq_policy *policy) { return cpumask_empty(policy->cpus); diff --git a/include/linux/sched/topology.h b/include/linux/sched/topology.h index 7f3dbafe1817..85a772968af3 100644 --- a/include/linux/sched/topology.h +++ b/include/linux/sched/topology.h @@ -203,6 +203,8 @@ struct sched_domain_topology_level { }; extern void __init set_sched_topology(struct sched_domain_topology_level *tl); +extern void sched_update_asym_prefer_cpu(int cpu, int old_prio, int new_prio); + # define SD_INIT_NAME(type) .name = #type @@ -237,6 +239,10 @@ static inline bool cpus_share_resources(int this_cpu, int that_cpu) return true; } +static inline void sched_update_asym_prefer_cpu(int cpu, int old_prio, int new_prio) +{ +} + #endif /* !CONFIG_SMP */ #if defined(CONFIG_ENERGY_MODEL) && defined(CONFIG_CPU_FREQ_GOV_SCHEDUTIL) diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c index a0893a483d35..228d981dea5d 100644 --- a/kernel/sched/debug.c +++ b/kernel/sched/debug.c @@ -586,6 +586,10 @@ static void register_sd(struct sched_domain *sd, struct dentry *parent) debugfs_create_file("flags", 0444, parent, &sd->flags, &sd_flags_fops); debugfs_create_file("groups_flags", 0444, parent, &sd->groups->flags, &sd_flags_fops); debugfs_create_u32("level", 0444, parent, (u32 *)&sd->level); + + if (sd->flags & SD_ASYM_PACKING) + debugfs_create_u32("group_asym_prefer_cpu", 0444, parent, + (u32 *)&sd->groups->asym_prefer_cpu); } void update_sched_domain_debugfs(void) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 3d9b68a347b7..f67b79c473dd 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -10259,7 +10259,7 @@ sched_group_asym(struct lb_env *env, struct sg_lb_stats *sgs, struct sched_group (sgs->group_weight - sgs->idle_cpus != 1)) return false; - return sched_asym(env->sd, env->dst_cpu, group->asym_prefer_cpu); + return sched_asym(env->sd, env->dst_cpu, READ_ONCE(group->asym_prefer_cpu)); } /* One group has more than one SMT CPU while the other group does not */ @@ -10496,7 +10496,8 @@ static bool update_sd_pick_busiest(struct lb_env *env, case group_asym_packing: /* Prefer to move from lowest priority CPU's work */ - return sched_asym_prefer(sds->busiest->asym_prefer_cpu, sg->asym_prefer_cpu); + return sched_asym_prefer(READ_ONCE(sds->busiest->asym_prefer_cpu), + READ_ONCE(sg->asym_prefer_cpu)); case group_misfit_task: /* diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c index 363ad268a25b..879c1d6df5d8 100644 --- a/kernel/sched/topology.c +++ b/kernel/sched/topology.c @@ -1344,6 +1344,64 @@ static void init_sched_groups_capacity(int cpu, struct sched_domain *sd) update_group_capacity(sd, cpu); } +#ifdef CONFIG_SMP + +/* Update the "asym_prefer_cpu" when arch_asym_cpu_priority() changes. */ +void sched_update_asym_prefer_cpu(int cpu, int old_prio, int new_prio) +{ + int asym_prefer_cpu = cpu; + struct sched_domain *sd; + + guard(rcu)(); + + for_each_domain(cpu, sd) { + struct sched_group *sg; + int group_cpu; + + if (!(sd->flags & SD_ASYM_PACKING)) + continue; + + /* + * Groups of overlapping domain are replicated per NUMA + * node and will require updating "asym_prefer_cpu" on + * each local copy. + * + * If you are hitting this warning, consider moving + * "sg->asym_prefer_cpu" to "sg->sgc->asym_prefer_cpu" + * which is shared by all the overlapping groups. + */ + WARN_ON_ONCE(sd->flags & SD_OVERLAP); + + sg = sd->groups; + if (cpu != sg->asym_prefer_cpu) { + /* + * Since the parent is a superset of the current group, + * if the cpu is not the "asym_prefer_cpu" at the + * current level, it cannot be the preferred CPU at a + * higher levels either. + */ + if (!sched_asym_prefer(cpu, sg->asym_prefer_cpu)) + return; + + WRITE_ONCE(sg->asym_prefer_cpu, cpu); + continue; + } + + /* Ranking has improved; CPU is still the preferred one. */ + if (new_prio >= old_prio) + continue; + + for_each_cpu(group_cpu, sched_group_span(sg)) { + if (sched_asym_prefer(group_cpu, asym_prefer_cpu)) + asym_prefer_cpu = group_cpu; + } + + WRITE_ONCE(sg->asym_prefer_cpu, asym_prefer_cpu); + } +} + +#endif /* CONFIG_SMP */ + /* * Set of available CPUs grouped by their corresponding capacities * Each list entry contains a CPU mask reflecting CPUs that share the same -- 2.49.0.592.g1a8a4971cc