Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'pm-6.12-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull power management fixes from Rafael Wysocki:
"Update cpufreq documentation to match the code after recent changes
(Christian Loehle), fix a units conversion issue in the CPPC cpufreq
driver (liwei), and fix an error check in the dtpm_devfreq power
capping driver (Yuan Can)"
* tag 'pm-6.12-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
cpufreq: CPPC: fix perf_to_khz/khz_to_perf conversion exception
powercap: dtpm_devfreq: Fix error check against dev_pm_qos_add_request()
cpufreq: docs: Reflect latency changes in docs
+28
-16
+10
-10
Documentation/admin-guide/pm/cpufreq.rst
+10
-10
Documentation/admin-guide/pm/cpufreq.rst
···
425
425
426
426
``rate_limit_us``
427
427
Minimum time (in microseconds) that has to pass between two consecutive
428
-
runs of governor computations (default: 1000 times the scaling driver's
429
-
transition latency).
428
+
runs of governor computations (default: 1.5 times the scaling driver's
429
+
transition latency or the maximum 2ms).
430
430
431
431
The purpose of this tunable is to reduce the scheduler context overhead
432
432
of the governor which might be excessive without it.
···
474
474
This is how often the governor's worker routine should run, in
475
475
microseconds.
476
476
477
-
Typically, it is set to values of the order of 10000 (10 ms). Its
478
-
default value is equal to the value of ``cpuinfo_transition_latency``
479
-
for each policy this governor is attached to (but since the unit here
480
-
is greater by 1000, this means that the time represented by
481
-
``sampling_rate`` is 1000 times greater than the transition latency by
482
-
default).
477
+
Typically, it is set to values of the order of 2000 (2 ms). Its
478
+
default value is to add a 50% breathing room
479
+
to ``cpuinfo_transition_latency`` on each policy this governor is
480
+
attached to. The minimum is typically the length of two scheduler
481
+
ticks.
483
482
484
483
If this tunable is per-policy, the following shell command sets the time
485
-
represented by it to be 750 times as high as the transition latency::
484
+
represented by it to be 1.5 times as high as the transition latency
485
+
(the default)::
486
486
487
-
# echo `$(($(cat cpuinfo_transition_latency) * 750 / 1000)) > ondemand/sampling_rate
487
+
# echo `$(($(cat cpuinfo_transition_latency) * 3 / 2)) > ondemand/sampling_rate
488
488
489
489
``up_threshold``
490
490
If the estimated CPU load is above this value (in percent), the governor
+17
-5
drivers/acpi/cppc_acpi.c
+17
-5
drivers/acpi/cppc_acpi.c
···
1916
1916
u64 mul, div;
1917
1917
1918
1918
if (caps->lowest_freq && caps->nominal_freq) {
1919
-
mul = caps->nominal_freq - caps->lowest_freq;
1919
+
/* Avoid special case when nominal_freq is equal to lowest_freq */
1920
+
if (caps->lowest_freq == caps->nominal_freq) {
1921
+
mul = caps->nominal_freq;
1922
+
div = caps->nominal_perf;
1923
+
} else {
1924
+
mul = caps->nominal_freq - caps->lowest_freq;
1925
+
div = caps->nominal_perf - caps->lowest_perf;
1926
+
}
1920
1927
mul *= KHZ_PER_MHZ;
1921
-
div = caps->nominal_perf - caps->lowest_perf;
1922
1928
offset = caps->nominal_freq * KHZ_PER_MHZ -
1923
1929
div64_u64(caps->nominal_perf * mul, div);
1924
1930
} else {
···
1945
1939
{
1946
1940
s64 retval, offset = 0;
1947
1941
static u64 max_khz;
1948
-
u64 mul, div;
1942
+
u64 mul, div;
1949
1943
1950
1944
if (caps->lowest_freq && caps->nominal_freq) {
1951
-
mul = caps->nominal_perf - caps->lowest_perf;
1952
-
div = caps->nominal_freq - caps->lowest_freq;
1945
+
/* Avoid special case when nominal_freq is equal to lowest_freq */
1946
+
if (caps->lowest_freq == caps->nominal_freq) {
1947
+
mul = caps->nominal_perf;
1948
+
div = caps->nominal_freq;
1949
+
} else {
1950
+
mul = caps->nominal_perf - caps->lowest_perf;
1951
+
div = caps->nominal_freq - caps->lowest_freq;
1952
+
}
1953
1953
/*
1954
1954
* We don't need to convert to kHz for computing offset and can
1955
1955
* directly use nominal_freq and lowest_freq as the div64_u64
+1
-1
drivers/powercap/dtpm_devfreq.c
+1
-1
drivers/powercap/dtpm_devfreq.c