diff options
Diffstat (limited to 'drivers/cpufreq/cppc_cpufreq.c')
| -rw-r--r-- | drivers/cpufreq/cppc_cpufreq.c | 204 |
1 files changed, 186 insertions, 18 deletions
diff --git a/drivers/cpufreq/cppc_cpufreq.c b/drivers/cpufreq/cppc_cpufreq.c index a1c3025f9df7..30f302149730 100644 --- a/drivers/cpufreq/cppc_cpufreq.c +++ b/drivers/cpufreq/cppc_cpufreq.c @@ -20,6 +20,7 @@ #include <linux/cpu.h> #include <linux/cpufreq.h> #include <linux/dmi.h> +#include <linux/time.h> #include <linux/vmalloc.h> #include <asm/unaligned.h> @@ -41,9 +42,6 @@ */ static struct cppc_cpudata **all_cpu_data; -/* Capture the max KHz from DMI */ -static u64 cppc_dmi_max_khz; - /* Callback function used to retrieve the max frequency from DMI */ static void cppc_find_dmi_mhz(const struct dmi_header *dm, void *private) { @@ -74,6 +72,64 @@ static u64 cppc_get_dmi_max_khz(void) return (1000 * mhz); } +/* + * If CPPC lowest_freq and nominal_freq registers are exposed then we can + * use them to convert perf to freq and vice versa + * + * If the perf/freq point lies between Nominal and Lowest, we can treat + * (Low perf, Low freq) and (Nom Perf, Nom freq) as 2D co-ordinates of a line + * and extrapolate the rest + * For perf/freq > Nominal, we use the ratio perf:freq at Nominal for conversion + */ +static unsigned int cppc_cpufreq_perf_to_khz(struct cppc_cpudata *cpu, + unsigned int perf) +{ + static u64 max_khz; + struct cppc_perf_caps *caps = &cpu->perf_caps; + u64 mul, div; + + if (caps->lowest_freq && caps->nominal_freq) { + if (perf >= caps->nominal_perf) { + mul = caps->nominal_freq; + div = caps->nominal_perf; + } else { + mul = caps->nominal_freq - caps->lowest_freq; + div = caps->nominal_perf - caps->lowest_perf; + } + } else { + if (!max_khz) + max_khz = cppc_get_dmi_max_khz(); + mul = max_khz; + div = cpu->perf_caps.highest_perf; + } + return (u64)perf * mul / div; +} + +static unsigned int cppc_cpufreq_khz_to_perf(struct cppc_cpudata *cpu, + unsigned int freq) +{ + static u64 max_khz; + struct cppc_perf_caps *caps = &cpu->perf_caps; + u64 mul, div; + + if (caps->lowest_freq && caps->nominal_freq) { + if (freq >= caps->nominal_freq) { + mul = caps->nominal_perf; + div = caps->nominal_freq; + } else { + mul = caps->lowest_perf; + div = caps->lowest_freq; + } + } else { + if (!max_khz) + max_khz = cppc_get_dmi_max_khz(); + mul = cpu->perf_caps.highest_perf; + div = max_khz; + } + + return (u64)freq * mul / div; +} + static int cppc_cpufreq_set_target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation) @@ -85,7 +141,7 @@ static int cppc_cpufreq_set_target(struct cpufreq_policy *policy, cpu = all_cpu_data[policy->cpu]; - desired_perf = (u64)target_freq * cpu->perf_caps.highest_perf / cppc_dmi_max_khz; + desired_perf = cppc_cpufreq_khz_to_perf(cpu, target_freq); /* Return if it is exactly the same perf */ if (desired_perf == cpu->perf_ctrls.desired_perf) return ret; @@ -125,6 +181,49 @@ static void cppc_cpufreq_stop_cpu(struct cpufreq_policy *policy) cpu->perf_caps.lowest_perf, cpu_num, ret); } +/* + * The PCC subspace describes the rate at which platform can accept commands + * on the shared PCC channel (including READs which do not count towards freq + * trasition requests), so ideally we need to use the PCC values as a fallback + * if we don't have a platform specific transition_delay_us + */ +#ifdef CONFIG_ARM64 +#include <asm/cputype.h> + +static unsigned int cppc_cpufreq_get_transition_delay_us(int cpu) +{ + unsigned long implementor = read_cpuid_implementor(); + unsigned long part_num = read_cpuid_part_number(); + unsigned int delay_us = 0; + + switch (implementor) { + case ARM_CPU_IMP_QCOM: + switch (part_num) { + case QCOM_CPU_PART_FALKOR_V1: + case QCOM_CPU_PART_FALKOR: + delay_us = 10000; + break; + default: + delay_us = cppc_get_transition_latency(cpu) / NSEC_PER_USEC; + break; + } + break; + default: + delay_us = cppc_get_transition_latency(cpu) / NSEC_PER_USEC; + break; + } + + return delay_us; +} + +#else + +static unsigned int cppc_cpufreq_get_transition_delay_us(int cpu) +{ + return cppc_get_transition_latency(cpu) / NSEC_PER_USEC; +} +#endif + static int cppc_cpufreq_cpu_init(struct cpufreq_policy *policy) { struct cppc_cpudata *cpu; @@ -142,31 +241,41 @@ static int cppc_cpufreq_cpu_init(struct cpufreq_policy *policy) return ret; } - cppc_dmi_max_khz = cppc_get_dmi_max_khz(); + /* Convert the lowest and nominal freq from MHz to KHz */ + cpu->perf_caps.lowest_freq *= 1000; + cpu->perf_caps.nominal_freq *= 1000; /* * Set min to lowest nonlinear perf to avoid any efficiency penalty (see * Section 8.4.7.1.1.5 of ACPI 6.1 spec) */ - policy->min = cpu->perf_caps.lowest_nonlinear_perf * cppc_dmi_max_khz / - cpu->perf_caps.highest_perf; - policy->max = cppc_dmi_max_khz; + policy->min = cppc_cpufreq_perf_to_khz(cpu, cpu->perf_caps.lowest_nonlinear_perf); + policy->max = cppc_cpufreq_perf_to_khz(cpu, cpu->perf_caps.highest_perf); /* * Set cpuinfo.min_freq to Lowest to make the full range of performance * available if userspace wants to use any perf between lowest & lowest * nonlinear perf */ - policy->cpuinfo.min_freq = cpu->perf_caps.lowest_perf * cppc_dmi_max_khz / - cpu->perf_caps.highest_perf; - policy->cpuinfo.max_freq = cppc_dmi_max_khz; + policy->cpuinfo.min_freq = cppc_cpufreq_perf_to_khz(cpu, cpu->perf_caps.lowest_perf); + policy->cpuinfo.max_freq = cppc_cpufreq_perf_to_khz(cpu, cpu->perf_caps.highest_perf); - policy->cpuinfo.transition_latency = cppc_get_transition_latency(cpu_num); + policy->transition_delay_us = cppc_cpufreq_get_transition_delay_us(cpu_num); policy->shared_type = cpu->shared_type; - if (policy->shared_type == CPUFREQ_SHARED_TYPE_ANY) + if (policy->shared_type == CPUFREQ_SHARED_TYPE_ANY) { + int i; + cpumask_copy(policy->cpus, cpu->shared_cpu_map); - else if (policy->shared_type == CPUFREQ_SHARED_TYPE_ALL) { + + for_each_cpu(i, policy->cpus) { + if (unlikely(i == policy->cpu)) + continue; + + memcpy(&all_cpu_data[i]->perf_caps, &cpu->perf_caps, + sizeof(cpu->perf_caps)); + } + } else if (policy->shared_type == CPUFREQ_SHARED_TYPE_ALL) { /* Support only SW_ANY for now. */ pr_debug("Unsupported CPU co-ord type\n"); return -EFAULT; @@ -175,7 +284,8 @@ static int cppc_cpufreq_cpu_init(struct cpufreq_policy *policy) cpu->cur_policy = policy; /* Set policy->cur to max now. The governors will adjust later. */ - policy->cur = cppc_dmi_max_khz; + policy->cur = cppc_cpufreq_perf_to_khz(cpu, + cpu->perf_caps.highest_perf); cpu->perf_ctrls.desired_perf = cpu->perf_caps.highest_perf; ret = cppc_set_perf(cpu_num, &cpu->perf_ctrls); @@ -186,10 +296,62 @@ static int cppc_cpufreq_cpu_init(struct cpufreq_policy *policy) return ret; } +static inline u64 get_delta(u64 t1, u64 t0) +{ + if (t1 > t0 || t0 > ~(u32)0) + return t1 - t0; + + return (u32)t1 - (u32)t0; +} + +static int cppc_get_rate_from_fbctrs(struct cppc_cpudata *cpu, + struct cppc_perf_fb_ctrs fb_ctrs_t0, + struct cppc_perf_fb_ctrs fb_ctrs_t1) +{ + u64 delta_reference, delta_delivered; + u64 reference_perf, delivered_perf; + + reference_perf = fb_ctrs_t0.reference_perf; + + delta_reference = get_delta(fb_ctrs_t1.reference, + fb_ctrs_t0.reference); + delta_delivered = get_delta(fb_ctrs_t1.delivered, + fb_ctrs_t0.delivered); + + /* Check to avoid divide-by zero */ + if (delta_reference || delta_delivered) + delivered_perf = (reference_perf * delta_delivered) / + delta_reference; + else + delivered_perf = cpu->perf_ctrls.desired_perf; + + return cppc_cpufreq_perf_to_khz(cpu, delivered_perf); +} + +static unsigned int cppc_cpufreq_get_rate(unsigned int cpunum) +{ + struct cppc_perf_fb_ctrs fb_ctrs_t0 = {0}, fb_ctrs_t1 = {0}; + struct cppc_cpudata *cpu = all_cpu_data[cpunum]; + int ret; + + ret = cppc_get_perf_ctrs(cpunum, &fb_ctrs_t0); + if (ret) + return ret; + + udelay(2); /* 2usec delay between sampling */ + + ret = cppc_get_perf_ctrs(cpunum, &fb_ctrs_t1); + if (ret) + return ret; + + return cppc_get_rate_from_fbctrs(cpu, fb_ctrs_t0, fb_ctrs_t1); +} + static struct cpufreq_driver cppc_cpufreq_driver = { .flags = CPUFREQ_CONST_LOOPS, .verify = cppc_verify_policy, .target = cppc_cpufreq_set_target, + .get = cppc_cpufreq_get_rate, .init = cppc_cpufreq_cpu_init, .stop_cpu = cppc_cpufreq_stop_cpu, .name = "cppc_cpufreq", @@ -203,7 +365,8 @@ static int __init cppc_cpufreq_init(void) if (acpi_disabled) return -ENODEV; - all_cpu_data = kzalloc(sizeof(void *) * num_possible_cpus(), GFP_KERNEL); + all_cpu_data = kcalloc(num_possible_cpus(), sizeof(void *), + GFP_KERNEL); if (!all_cpu_data) return -ENOMEM; @@ -230,8 +393,13 @@ static int __init cppc_cpufreq_init(void) return ret; out: - for_each_possible_cpu(i) - kfree(all_cpu_data[i]); + for_each_possible_cpu(i) { + cpu = all_cpu_data[i]; + if (!cpu) + break; + free_cpumask_var(cpu->shared_cpu_map); + kfree(cpu); + } kfree(all_cpu_data); return -ENODEV; |