Messages in this thread | | | Date | Thu, 26 Oct 2023 11:21:34 +0800 | Subject | Re: [PATCH 2/3] cpufreq: CPPC: Keep the target core awake when reading its cpufreq rate | From | Zeng Heng <> |
| |
在 2023/10/25 18:54, Mark Rutland 写道: > [adding Ionela] > > On Wed, Oct 25, 2023 at 05:38:46PM +0800, Zeng Heng wrote: >> As ARM AMU's document says, all counters are subject to any changes >> in clock frequency, including clock stopping caused by the WFI and WFE >> instructions. >> >> Therefore, using smp_call_on_cpu() to trigger target CPU to >> read self's AMU counters, which ensures the counters are working >> properly while cstate feature is enabled. > IIUC there's a pretty deliberate split with all the actual reading of the AMU > living in arch/arm64/kernel/topolgy.c, and the driver code being (relatively) > generic. > > We already have code in arch/arm64/kernel/topolgy.c to read counters on a > specific CPU; why can't e reuse that (and avoid exporting cpu_has_amu_feat())? > > Mark.
In this scenario, both topology.c and cppc_acpi.c do not provide an API to keep the AMU online
during the whole sampling period. Just using cpc_read_ffh at the start and end of the sampling
period is not enough.
However, I can propose cpc_ffh_supported() function to replace the cpu_has_amu_feat() as v2
if you think this patch set is still valuable.
Thanks,
Zeng Heng
>> Reported-by: Sumit Gupta <sumitg@nvidia.com> >> Link: https://lore.kernel.org/all/20230418113459.12860-7-sumitg@nvidia.com/ >> Signed-off-by: Zeng Heng <zengheng4@huawei.com> >> --- >> drivers/cpufreq/cppc_cpufreq.c | 39 ++++++++++++++++++++++++++-------- >> 1 file changed, 30 insertions(+), 9 deletions(-) >> >> diff --git a/drivers/cpufreq/cppc_cpufreq.c b/drivers/cpufreq/cppc_cpufreq.c >> index fe08ca419b3d..321a9dc9484d 100644 >> --- a/drivers/cpufreq/cppc_cpufreq.c >> +++ b/drivers/cpufreq/cppc_cpufreq.c >> @@ -90,6 +90,12 @@ static int cppc_perf_from_fbctrs(struct cppc_cpudata *cpu_data, >> struct cppc_perf_fb_ctrs *fb_ctrs_t0, >> struct cppc_perf_fb_ctrs *fb_ctrs_t1); >> >> +struct fb_ctr_pair { >> + u32 cpu; >> + struct cppc_perf_fb_ctrs fb_ctrs_t0; >> + struct cppc_perf_fb_ctrs fb_ctrs_t1; >> +}; >> + >> /** >> * cppc_scale_freq_workfn - CPPC arch_freq_scale updater for frequency invariance >> * @work: The work item. >> @@ -840,9 +846,24 @@ static int cppc_perf_from_fbctrs(struct cppc_cpudata *cpu_data, >> return (reference_perf * delta_delivered) / delta_reference; >> } >> >> +static int cppc_get_perf_ctrs_pair(void *val) >> +{ >> + struct fb_ctr_pair *fb_ctrs = val; >> + int cpu = fb_ctrs->cpu; >> + int ret; >> + >> + ret = cppc_get_perf_ctrs(cpu, &fb_ctrs->fb_ctrs_t0); >> + if (ret) >> + return ret; >> + >> + udelay(2); /* 2usec delay between sampling */ >> + >> + return cppc_get_perf_ctrs(cpu, &fb_ctrs->fb_ctrs_t1); >> +} >> + >> static unsigned int cppc_cpufreq_get_rate(unsigned int cpu) >> { >> - struct cppc_perf_fb_ctrs fb_ctrs_t0 = {0}, fb_ctrs_t1 = {0}; >> + struct fb_ctr_pair fb_ctrs = { .cpu = cpu, }; >> struct cpufreq_policy *policy = cpufreq_cpu_get(cpu); >> struct cppc_cpudata *cpu_data = policy->driver_data; >> u64 delivered_perf; >> @@ -850,18 +871,18 @@ static unsigned int cppc_cpufreq_get_rate(unsigned int cpu) >> >> cpufreq_cpu_put(policy); >> >> - ret = cppc_get_perf_ctrs(cpu, &fb_ctrs_t0); >> - if (ret) >> - return 0; >> - >> - udelay(2); /* 2usec delay between sampling */ >> + if (cpu_has_amu_feat(cpu)) >> + ret = smp_call_on_cpu(cpu, cppc_get_perf_ctrs_pair, >> + &fb_ctrs, false); >> + else >> + ret = cppc_get_perf_ctrs_pair(&fb_ctrs); >> >> - ret = cppc_get_perf_ctrs(cpu, &fb_ctrs_t1); >> if (ret) >> return 0; >> >> - delivered_perf = cppc_perf_from_fbctrs(cpu_data, &fb_ctrs_t0, >> - &fb_ctrs_t1); >> + delivered_perf = cppc_perf_from_fbctrs(cpu_data, >> + &fb_ctrs.fb_ctrs_t0, >> + &fb_ctrs.fb_ctrs_t1); >> >> return cppc_cpufreq_perf_to_khz(cpu_data, delivered_perf); >> } >> -- >> 2.25.1 >>
| |