From d253d2a52676cfa3d89b8f0737a08ce7db665207 Mon Sep 17 00:00:00 2001 From: Brennan Shacklett Date: Mon, 21 Oct 2013 09:20:32 -0700 Subject: intel_pstate: Improve accuracy by not truncating until final result This patch addresses Bug 60727 (https://bugzilla.kernel.org/show_bug.cgi?id=60727) which was due to the truncation of intermediate values in the calculations, which causes the code to consistently underestimate the current cpu frequency, specifically 100% cpu utilization was truncated down to the setpoint of 97%. This patch fixes the problem by keeping the results of all intermediate calculations as fixed point numbers rather scaling them back and forth between integers and fixed point. References: https://bugzilla.kernel.org/show_bug.cgi?id=60727 Signed-off-by: Brennan Shacklett Acked-by: Dirk Brandewie Signed-off-by: Rafael J. Wysocki diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c index badf620..8b8677f 100644 --- a/drivers/cpufreq/intel_pstate.c +++ b/drivers/cpufreq/intel_pstate.c @@ -48,7 +48,7 @@ static inline int32_t div_fp(int32_t x, int32_t y) } struct sample { - int core_pct_busy; + int32_t core_pct_busy; u64 aperf; u64 mperf; int freq; @@ -68,7 +68,7 @@ struct _pid { int32_t i_gain; int32_t d_gain; int deadband; - int last_err; + int32_t last_err; }; struct cpudata { @@ -153,16 +153,15 @@ static inline void pid_d_gain_set(struct _pid *pid, int percent) pid->d_gain = div_fp(int_tofp(percent), int_tofp(100)); } -static signed int pid_calc(struct _pid *pid, int busy) +static signed int pid_calc(struct _pid *pid, int32_t busy) { - signed int err, result; + signed int result; int32_t pterm, dterm, fp_error; int32_t integral_limit; - err = pid->setpoint - busy; - fp_error = int_tofp(err); + fp_error = int_tofp(pid->setpoint) - busy; - if (abs(err) <= pid->deadband) + if (abs(fp_error) <= int_tofp(pid->deadband)) return 0; pterm = mul_fp(pid->p_gain, fp_error); @@ -176,8 +175,8 @@ static signed int pid_calc(struct _pid *pid, int busy) if (pid->integral < -integral_limit) pid->integral = -integral_limit; - dterm = mul_fp(pid->d_gain, (err - pid->last_err)); - pid->last_err = err; + dterm = mul_fp(pid->d_gain, fp_error - pid->last_err); + pid->last_err = fp_error; result = pterm + mul_fp(pid->integral, pid->i_gain) + dterm; @@ -436,8 +435,9 @@ static inline void intel_pstate_calc_busy(struct cpudata *cpu, struct sample *sample) { u64 core_pct; - core_pct = div64_u64(sample->aperf * 100, sample->mperf); - sample->freq = cpu->pstate.max_pstate * core_pct * 1000; + core_pct = div64_u64(int_tofp(sample->aperf * 100), + sample->mperf); + sample->freq = fp_toint(cpu->pstate.max_pstate * core_pct * 1000); sample->core_pct_busy = core_pct; } @@ -469,22 +469,19 @@ static inline void intel_pstate_set_sample_time(struct cpudata *cpu) mod_timer_pinned(&cpu->timer, jiffies + delay); } -static inline int intel_pstate_get_scaled_busy(struct cpudata *cpu) +static inline int32_t intel_pstate_get_scaled_busy(struct cpudata *cpu) { - int32_t busy_scaled; int32_t core_busy, max_pstate, current_pstate; - core_busy = int_tofp(cpu->samples[cpu->sample_ptr].core_pct_busy); + core_busy = cpu->samples[cpu->sample_ptr].core_pct_busy; max_pstate = int_tofp(cpu->pstate.max_pstate); current_pstate = int_tofp(cpu->pstate.current_pstate); - busy_scaled = mul_fp(core_busy, div_fp(max_pstate, current_pstate)); - - return fp_toint(busy_scaled); + return mul_fp(core_busy, div_fp(max_pstate, current_pstate)); } static inline void intel_pstate_adjust_busy_pstate(struct cpudata *cpu) { - int busy_scaled; + int32_t busy_scaled; struct _pid *pid; signed int ctl = 0; int steps; -- cgit v0.10.2 From 7244cb62d96e735847dc9d08f870550df896898c Mon Sep 17 00:00:00 2001 From: Dirk Brandewie Date: Mon, 21 Oct 2013 09:20:33 -0700 Subject: intel_pstate: Correct calculation of min pstate value The minimum pstate is supposed to be a percentage of the maximum P state available. Calculate min using max pstate and not the current max which may have been limited by the user Signed-off-by: Dirk Brandewie Signed-off-by: Rafael J. Wysocki diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c index 8b8677f..eb3fdc7 100644 --- a/drivers/cpufreq/intel_pstate.c +++ b/drivers/cpufreq/intel_pstate.c @@ -366,12 +366,13 @@ static int intel_pstate_turbo_pstate(void) static void intel_pstate_get_min_max(struct cpudata *cpu, int *min, int *max) { int max_perf = cpu->pstate.turbo_pstate; + int max_perf_adj; int min_perf; if (limits.no_turbo) max_perf = cpu->pstate.max_pstate; - max_perf = fp_toint(mul_fp(int_tofp(max_perf), limits.max_perf)); - *max = clamp_t(int, max_perf, + max_perf_adj = fp_toint(mul_fp(int_tofp(max_perf), limits.max_perf)); + *max = clamp_t(int, max_perf_adj, cpu->pstate.min_pstate, cpu->pstate.turbo_pstate); min_perf = fp_toint(mul_fp(int_tofp(max_perf), limits.min_perf)); -- cgit v0.10.2 From d3c345dbc7c083414ef74eb22ff26ba2bd100759 Mon Sep 17 00:00:00 2001 From: Russ Dill Date: Thu, 24 Oct 2013 14:25:26 +0100 Subject: PM / hibernate: Move software_resume to late_initcall_sync software_resume is being called after deferred_probe_initcall in drivers base. If the probing of the device that contains the resume image is deferred, and the system has been instructed to wait for it to show up, this wait will occur in software_resume. This causes a deadlock. Move software_resume into late_initcall_sync so that it happens after all the other late_initcalls. Signed-off-by: Russ Dill Acked-by: Pavel Machek Signed-off-by: Rafael J. Wysocki diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c index c9c759d..0121dab 100644 --- a/kernel/power/hibernate.c +++ b/kernel/power/hibernate.c @@ -846,7 +846,7 @@ static int software_resume(void) goto Finish; } -late_initcall(software_resume); +late_initcall_sync(software_resume); static const char * const hibernation_modes[] = { -- cgit v0.10.2 From 75c0758137c7ac647927b4b12bb5cfca96a0e4e6 Mon Sep 17 00:00:00 2001 From: "Rafael J. Wysocki" Date: Fri, 25 Oct 2013 16:22:47 +0200 Subject: acpi-cpufreq: Fail initialization if driver cannot be registered Make acpi_cpufreq_init() return error codes when the driver cannot be registered so that the module doesn't stay useless in memory and so that acpi_cpufreq_exit() doesn't attempt to unregister things that have never been registered when the module is unloaded. Signed-off-by: Rafael J. Wysocki Acked-by: Viresh Kumar diff --git a/drivers/cpufreq/acpi-cpufreq.c b/drivers/cpufreq/acpi-cpufreq.c index d2c3253..506fd23 100644 --- a/drivers/cpufreq/acpi-cpufreq.c +++ b/drivers/cpufreq/acpi-cpufreq.c @@ -986,12 +986,12 @@ static int __init acpi_cpufreq_init(void) { int ret; + if (acpi_disabled) + return -ENODEV; + /* don't keep reloading if cpufreq_driver exists */ if (cpufreq_get_current_driver()) - return 0; - - if (acpi_disabled) - return 0; + return -EEXIST; pr_debug("acpi_cpufreq_init\n"); -- cgit v0.10.2