| Age | Commit message (Collapse) | Author |
|---|
|
* pm-cpufreq:
cpufreq: Fix typo in kerneldoc comment
cpufreq: schedutil: Remove update_lock comment from struct sugov_policy definition
cpufreq: schedutil: Remove needless sg_policy parameter from ignore_dl_rate_limit()
cpufreq: ACPI: Set cpuinfo.max_freq directly if max boost is known
cpufreq: qcom-hw: drop devm_xxx() calls from init/exit hooks
* pm-opp:
opp: Don't skip freq update for different frequency
|
|
Change 'Terget' to 'Target'.
Should be Target.
Signed-off-by: Yue Hu <huyue2@yulong.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
[ rjw: Subject edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm
Pull ARM cpufreq fix for 5.12 from Viresh Kumar:
"Single patch to fix issue with cpu hotplug and policy recreation for
qcom-cpufreq-hw driver."
* 'cpufreq/arm/linux-next' of git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm:
cpufreq: qcom-hw: drop devm_xxx() calls from init/exit hooks
|
|
Commit 3c55e94c0ade ("cpufreq: ACPI: Extend frequency tables to cover
boost frequencies") attempted to address a performance issue involving
acpi-cpufreq, the schedutil governor and scale-invariance on x86 by
extending the frequency tables created by acpi-cpufreq to cover the
entire range of "turbo" (or "boost") frequencies, but that caused
frequencies reported via /proc/cpuinfo and the scaling_cur_freq
attribute in sysfs to change which may confuse users and monitoring
tools.
For this reason, revert the part of commit 3c55e94c0ade adding the
extra entry to the frequency table and use the observation that
in principle cpuinfo.max_freq need not be equal to the maximum
frequency listed in the frequency table for the given policy.
Namely, modify cpufreq_frequency_table_cpuinfo() to allow cpufreq
drivers to set their own cpuinfo.max_freq above that frequency and
change acpi-cpufreq to set cpuinfo.max_freq to the maximum boost
frequency found via CPPC.
This should be sufficient to let all of the cpufreq subsystem know
the real maximum frequency of the CPU without changing frequency
reporting.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=211305
Fixes: 3c55e94c0ade ("cpufreq: ACPI: Extend frequency tables to cover boost frequencies")
Reported-by: Matt McDonald <gardotd426@gmail.com>
Tested-by: Matt McDonald <gardotd426@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Tested-by: Michael Larabel <Michael@phoronix.com>
Cc: 5.11+ <stable@vger.kernel.org> # 5.11+
|
|
Commit f17b3e44320b ("cpufreq: qcom-hw: Use
devm_platform_ioremap_resource() to simplify code") introduces
a regression on platforms using the driver, by failing to initialise
a policy, when one is created post hotplug.
When all the CPUs of a policy are hoptplugged out, the call to .exit()
and later to devm_iounmap() does not release the memory region that was
requested during devm_platform_ioremap_resource(). Therefore,
a subsequent call to .init() will result in the following error, which
will prevent a new policy to be initialised:
[ 3395.915416] CPU4: shutdown
[ 3395.938185] psci: CPU4 killed (polled 0 ms)
[ 3399.071424] CPU5: shutdown
[ 3399.094316] psci: CPU5 killed (polled 0 ms)
[ 3402.139358] CPU6: shutdown
[ 3402.161705] psci: CPU6 killed (polled 0 ms)
[ 3404.742939] CPU7: shutdown
[ 3404.765592] psci: CPU7 killed (polled 0 ms)
[ 3411.492274] Detected VIPT I-cache on CPU4
[ 3411.492337] GICv3: CPU4: found redistributor 400 region 0:0x0000000017ae0000
[ 3411.492448] CPU4: Booted secondary processor 0x0000000400 [0x516f802d]
[ 3411.503654] qcom-cpufreq-hw 17d43000.cpufreq: can't request region for resource [mem 0x17d45800-0x17d46bff]
With that being said, the original code was tricky and skipping memory
region request intentionally to hide this issue. The true cause is that
those devm_xxx() device managed functions shouldn't be used for cpufreq
init/exit hooks, because &pdev->dev is alive across the hooks and will
not trigger auto resource free-up. Let's drop the use of device managed
functions and manually allocate/free resources, so that the issue can be
fixed properly.
Cc: v5.10+ <stable@vger.kernel.org> # v5.10+
Fixes: f17b3e44320b ("cpufreq: qcom-hw: Use devm_platform_ioremap_resource() to simplify code")
Suggested-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
* pm-opp: (37 commits)
PM / devfreq: Add required OPPs support to passive governor
PM / devfreq: Cache OPP table reference in devfreq
OPP: Add function to look up required OPP's for a given OPP
opp: Replace ENOTSUPP with EOPNOTSUPP
opp: Fix "foo * bar" should be "foo *bar"
opp: Don't ignore clk_get() errors other than -ENOENT
opp: Update bandwidth requirements based on scaling up/down
opp: Allow lazy-linking of required-opps
opp: Remove dev_pm_opp_set_bw()
devfreq: tegra30: Migrate to dev_pm_opp_set_opp()
drm: msm: Migrate to dev_pm_opp_set_opp()
cpufreq: qcom: Migrate to dev_pm_opp_set_opp()
opp: Implement dev_pm_opp_set_opp()
opp: Update parameters of _set_opp_custom()
opp: Allow _generic_set_opp_clk_only() to work for non-freq devices
opp: Allow _generic_set_opp_regulator() to work for non-freq devices
opp: Allow _set_opp() to work for non-freq devices
opp: Split _set_opp() out of dev_pm_opp_set_rate()
opp: Keep track of currently programmed OPP
opp: No need to check clk for errors
...
|
|
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm
Pull ARM cpufreq changes for v5.12 from Viresh Kumar:
"- Removal of Tango driver as the platform got removed (Arnd Bergmann).
- Use resource managed APIs for tegra20 (Dmitry Osipenko).
- Generic cleanups for brcmstb (Christophe JAILLET).
- Enable boost support for qcom-hw (Shawn Guo)."
* 'cpufreq/arm/linux-next' of git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm:
cpufreq: remove tango driver
cpufreq: brcmstb-avs-cpufreq: Fix resource leaks in ->remove()
cpufreq: brcmstb-avs-cpufreq: Free resources in error path
cpufreq: qcom-hw: enable boost support
cpufreq: tegra20: Use resource-managed API
|
|
If the maximum performance level taken for computing the
arch_max_freq_ratio value used in the x86 scale-invariance code is
higher than the one corresponding to the cpuinfo.max_freq value
coming from the acpi_cpufreq driver, the scale-invariant utilization
falls below 100% even if the CPU runs at cpuinfo.max_freq or slightly
faster, which causes the schedutil governor to select a frequency
below cpuinfo.max_freq. That frequency corresponds to a frequency
table entry below the maximum performance level necessary to get to
the "boost" range of CPU frequencies which prevents "boost"
frequencies from being used in some workloads.
While this issue is related to scale-invariance, it may be amplified
by commit db865272d9c4 ("cpufreq: Avoid configuring old governors as
default with intel_pstate") from the 5.10 development cycle which
made it extremely easy to default to schedutil even if the preferred
driver is acpi_cpufreq as long as intel_pstate is built too, because
the mere presence of the latter effectively removes the ondemand
governor from the defaults. Distro kernels are likely to include
both intel_pstate and acpi_cpufreq on x86, so their users who cannot
use intel_pstate or choose to use acpi_cpufreq may easily be
affectecd by this issue.
If CPPC is available, it can be used to address this issue by
extending the frequency tables created by acpi_cpufreq to cover the
entire available frequency range (including "boost" frequencies) for
each CPU, but if CPPC is not there, acpi_cpufreq has no idea what
the maximum "boost" frequency is and the frequency tables created by
it cannot be extended in a meaningful way, so in that case make it
ask the arch scale-invariance code to to use the "nominal" performance
level for CPU utilization scaling in order to avoid the issue at hand.
Fixes: db865272d9c4 ("cpufreq: Avoid configuring old governors as default with intel_pstate")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
|
|
A severe performance regression on AMD EPYC processors when using
the schedutil scaling governor was discovered by Phoronix.com and
attributed to the following commits:
41ea667227ba ("x86, sched: Calculate frequency invariance for AMD
systems")
976df7e5730e ("x86, sched: Use midpoint of max_boost and max_P for
frequency invariance on AMD EPYC")
The source of the problem is that the maximum performance level taken
for computing the arch_max_freq_ratio value used in the x86 scale-
invariance code is higher than the one corresponding to the
cpuinfo.max_freq value coming from the acpi_cpufreq driver.
This effectively causes the scale-invariant utilization to fall below
100% even if the CPU runs at cpuinfo.max_freq or slightly faster, so
the schedutil governor selects a frequency below cpuinfo.max_freq
then. That frequency corresponds to a frequency table entry below
the maximum performance level necessary to get to the "boost" range
of CPU frequencies.
However, if the cpuinfo.max_freq value coming from acpi_cpufreq was
higher, the schedutil governor would select higher frequencies which
in turn would allow acpi_cpufreq to set more adequate performance
levels and to get to the "boost" range of CPU frequencies more often.
This issue affects any systems where acpi_cpufreq is used and the
"boost" (or "turbo") frequencies are enabled, not just AMD EPYC.
Moreover, commit db865272d9c4 ("cpufreq: Avoid configuring old
governors as default with intel_pstate") from the 5.10 development
cycle made it extremely easy to default to schedutil even if the
preferred driver is acpi_cpufreq as long as intel_pstate is built
too, because the mere presence of the latter effectively removes the
ondemand governor from the defaults. Distro kernels are likely to
include both intel_pstate and acpi_cpufreq on x86, so their users
who cannot use intel_pstate or choose to use acpi_cpufreq may
easily be affectecd by this issue.
To address this issue, extend the frequency table constructed by
acpi_cpufreq for each CPU to cover the entire range of available
frequencies (including the "boost" ones) if CPPC is available and
indicates that "boost" (or "turbo") frequencies are enabled. That
causes cpuinfo.max_freq to become the maximum "boost" frequency of
the given CPU (instead of the maximum frequency returned by the ACPI
_PSS object that corresponds to the "nominal" performance level).
Fixes: 41ea667227ba ("x86, sched: Calculate frequency invariance for AMD systems")
Fixes: 976df7e5730e ("x86, sched: Use midpoint of max_boost and max_P for frequency invariance on AMD EPYC")
Fixes: db865272d9c4 ("cpufreq: Avoid configuring old governors as default with intel_pstate")
Link: https://www.phoronix.com/scan.php?page=article&item=linux511-amd-schedutil&num=1
Link: https://lore.kernel.org/linux-pm/20210203135321.12253-2-ggherdovich@suse.cz/
Reported-by: Michael Larabel <Michael@phoronix.com>
Diagnosed-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Reviewed-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Tested-by: Michael Larabel <Michael@phoronix.com>
|
|
This flag is set by one of the drivers but it isn't used in the code
otherwise. Remove the unused flag and update the driver.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
During cpufreq driver's registration, if the ->init() callback for all
the CPUs fail then there is not much point in keeping the driver around
as it will only account for more of unnecessary noise, for example
cpufreq core will try to suspend/resume the driver which never got
registered properly.
The removal of such a driver is avoided if the driver carries the
CPUFREQ_STICKY flag. This was added way back [1] in 2004 and perhaps no
one should ever need it now. A lot of drivers do set this flag, probably
because they just copied it from other drivers.
This was added earlier for some platforms [2] because their cpufreq
drivers were getting registered before the CPUs were registered with
subsys framework. And hence they used to fail.
The same isn't true anymore though. The current code flow in the kernel
is:
start_kernel()
-> kernel_init()
-> kernel_init_freeable()
-> do_basic_setup()
-> driver_init()
-> cpu_dev_init()
-> subsys_system_register() //For CPUs
-> do_initcalls()
-> cpufreq_register_driver()
Clearly, the CPUs will always get registered with subsys framework
before any cpufreq driver can get probed. Remove the flag and update the
relevant drivers.
Link: https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/include/linux/cpufreq.h?id=7cc9f0d9a1ab04cedc60d64fd8dcf7df224a3b4d # [1]
Link: https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/arch/arm/mach-sa1100/cpu-sa1100.c?id=f59d3bbe35f6268d729f51be82af8325d62f20f5 # [2]
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
dev_pm_opp_set_bw() is getting removed and dev_pm_opp_set_opp() should
be used instead. Migrate to the new API.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Tested-by: Dmitry Osipenko <digetx@gmail.com>
|
|
In the comment for trace in passive mode there is an
unnecessary "the". Eradicate it.
Signed-off-by: Nigel Christian <nigel.l.christian@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
The tango platform is getting removed, so the driver is no
longer needed.
Cc: Marc Gonzalez <marc.w.gonzalez@free.fr>
Cc: Mans Rullgard <mans@mansr.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
[ Viresh: Update cpufreq-dt-platdev.c as well ]
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
If 'cpufreq_unregister_driver()' fails, just WARN and continue, so that
other resources are freed.
Fixes: de322e085995 ("cpufreq: brcmstb-avs-cpufreq: AVS CPUfreq driver for Broadcom STB SoCs")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
[ Viresh: Updated Subject ]
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
If 'cpufreq_register_driver()' fails, we must release the resources
allocated in 'brcm_avs_prepare_init()' as already done in the remove
function.
To do that, introduce a new function 'brcm_avs_prepare_uninit()' in order
to avoid code duplication. This also makes the code more readable (IMHO).
Fixes: de322e085995 ("cpufreq: brcmstb-avs-cpufreq: AVS CPUfreq driver for Broadcom STB SoCs")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
[ Viresh: Updated Subject ]
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
At least on sdm850, the 2956800 khz is detected as a boost frequency in
function qcom_cpufreq_hw_read_lut(). Let's enable boost support by
calling cpufreq_enable_boost_support(), so that we can get the boost
frequency by switching it on via 'boost' sysfs entry like below.
$ echo 1 > /sys/devices/system/cpu/cpufreq/boost
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Tested-by: Steev Klimaszewski <steev@kali.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
Switch cpufreq-tegra20 driver to use resource-managed API.
This removes the need to get opp_table pointer using
dev_pm_opp_get_opp_table() in order to release OPP table that
was requested by dev_pm_opp_set_supported_hw(), making the code
a bit more straightforward.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
available
Currently, when turbo is disabled (either by BIOS or by the user),
the intel_pstate driver reads the max non-turbo frequency from the
package-wide MSR_PLATFORM_INFO(0xce) register.
However, on asymmetric platforms it is possible in theory that small
and big core with HWP enabled might have different max non-turbo CPU
frequency, because MSR_HWP_CAPABILITIES is per-CPU scope according
to Intel Software Developer Manual.
The turbo max freq is already per-CPU in current code, so make
similar change to the max non-turbo frequency as well.
Reported-by: Wendy Wang <wendy.wang@intel.com>
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
[ rjw: Subject and changelog edits ]
Cc: 4.18+ <stable@vger.kernel.org> # 4.18+: a45ee4d4e13b: cpufreq: intel_pstate: Change intel_pstate_get_hwp_max() argument
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Rename intel_cpufreq_adjust_hwp() and intel_cpufreq_adjust_perf_ctl()
to intel_cpufreq_hwp_update() and intel_cpufreq_perf_ctl_update(),
respectively, to avoid possible confusion with the ->adjist_perf()
callback function, intel_cpufreq_adjust_perf().
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Chen Yu <yu.c.chen@intel.com>
|
|
All of the callers of intel_pstate_get_hwp_max() access the struct
cpudata object that corresponds to the given CPU already and the
function itself needs to access that object (in order to update
hwp_cap_cached), so modify the code to pass a struct cpudata pointer
to it instead of the CPU number.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Chen Yu <yu.c.chen@intel.com>
|
|
Because intel_pstate_get_hwp_max() which updates hwp_cap_cached
may run in parallel with the readers of it, annotate all of the
read accesses to it with READ_ONCE().
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Chen Yu <yu.c.chen@intel.com>
|
|
percent_fp() was used in intel_pstate_pid_reset(), which was removed in
commit 9d0ef7af1f2d ("cpufreq: intel_pstate: Do not use PID-based P-state
selection") and hence, percent_fp() is unused since then.
percent_ext_fp() was last used in intel_pstate_update_perf_limits(), which
was refactored in commit 1a4fe38add8b ("cpufreq: intel_pstate: Remove
max/min fractions to limit performance"), and hence, percent_ext_fp() is
unused since then.
make CC=clang W=1 points us those unused functions:
drivers/cpufreq/intel_pstate.c:79:23: warning: unused function 'percent_fp' [-Wunused-function]
static inline int32_t percent_fp(int percent)
^
drivers/cpufreq/intel_pstate.c:94:23: warning: unused function 'percent_ext_fp' [-Wunused-function]
static inline int32_t percent_ext_fp(int percent)
^
Remove those obsolete functions.
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Currently there is an unlikely case where cpufreq_cpu_get() returns a
NULL policy and this will cause a NULL pointer dereference later on.
Fix this by passing the policy to transition_frequency_fidvid() from
the caller and hence eliminating the need for the cpufreq_cpu_get()
and cpufreq_cpu_put().
Thanks to Viresh Kumar for suggesting the fix.
Addresses-Coverity: ("Dereference null return")
Fixes: b43a7ffbf33b ("cpufreq: Notify all policy->cpus in cpufreq_notify_transition()")
Suggested-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
If turbo P-states cannot be used, either due to the configuration of
the processor, or because intel_pstate is not allowed to used them,
the maximum available P-state with HWP enabled corresponds to the
HWP_CAP.GUARANTEED value which is not static. It can be adjusted by
an out-of-band agent or during an Intel Speed Select performance
level change, so long as it remains less than or equal to
HWP_CAP.MAX.
However, if turbo P-states cannot be used, intel_cpufreq_adjust_perf()
always uses pstate.max_pstate (set during the initialization of the
driver only) as the maximum available P-state, so it may miss a change
of the HWP_CAP.GUARANTEED value.
Prevent that from happening by modifyig intel_cpufreq_adjust_perf()
to always read the "guaranteed" and "maximum turbo" performance
levels from the cached HWP_CAP value.
Fixes: a365ab6b9dfb ("cpufreq: intel_pstate: Implement the ->adjust_perf() callback")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
|
|
When sugov_update_single_perf() falls back to the "frequency"
path due to the missing scale-invariance, it will call
cpufreq_driver_fast_switch() via sugov_fast_switch()
and the driver's ->fast_switch() callback will be invoked,
so it must not be NULL.
However, after commit a365ab6b9dfb ("cpufreq: intel_pstate: Implement
the ->adjust_perf() callback") intel_pstate sets ->fast_switch() to
NULL when it is going to use intel_cpufreq_adjust_perf(), which is a
mistake, because on x86 the scale-invariance may be turned off
dynamically, so modify it to retain the original ->adjust_perf()
callback pointer.
Fixes: a365ab6b9dfb ("cpufreq: intel_pstate: Implement the ->adjust_perf() callback")
Reported-by: Kenneth R. Crudup <kenny@panix.com>
Tested-by: Kenneth R. Crudup <kenny@panix.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
* pm-cpufreq:
cpufreq: intel_pstate: Use most recent guaranteed performance values
cpufreq: intel_pstate: Implement the ->adjust_perf() callback
cpufreq: Add special-purpose fast-switching callback for drivers
cpufreq: schedutil: Add util to struct sg_cpu
cppc_cpufreq: replace per-cpu data array with a list
cppc_cpufreq: expose information on frequency domains
cppc_cpufreq: clarify support for coordination types
cppc_cpufreq: use policy->cpu as driver of frequency setting
ACPI: processor: fix NONE coordination for domain mapping failure
ACPI: processor: Drop duplicate setting of shared_cpu_map
|
|
When turbo has been disabled by the BIOS, but HWP_CAP.GUARANTEED is
changed later, user space may want to take advantage of this increased
guaranteed performance.
HWP_CAP.GUARANTEED is not a static value. It can be adjusted by an
out-of-band agent or during an Intel Speed Select performance level
change. The HWP_CAP.MAX is still the maximum achievable performance
with turbo disabled by the BIOS, so HWP_CAP.GUARANTEED can still
change as long as it remains less than or equal to HWP_CAP.MAX.
When HWP_CAP.GUARANTEED is changed, the sysfs base_frequency
attribute shows the most recent guaranteed frequency value. This
attribute can be used by user space software to update the scaling
min/max limits of the CPU.
Currently, the ->setpolicy() callback already uses the latest
HWP_CAP values when setting HWP_REQ, but the ->verify() callback will
restrict the user settings to the to old guaranteed performance value
which prevents user space from making use of the extra CPU capacity
theoretically available to it after increasing HWP_CAP.GUARANTEED.
To address this, read HWP_CAP in intel_pstate_verify_cpu_policy()
to obtain the maximum P-state that can be used and use that to
confine the policy max limit instead of using the cached and
possibly stale pstate.max_freq value for this purpose.
For consistency, update intel_pstate_update_perf_limits() to use the
maximum available P-state returned by intel_pstate_get_hwp_max() to
compute the maximum frequency instead of using the return value of
intel_pstate_get_max_freq() which, again, may be stale.
This issue is a side-effect of fixing the scaling frequency limits in
commit eacc9c5a927e ("cpufreq: intel_pstate: Fix intel_pstate_get_hwp_max()
for turbo disabled") which corrected the setting of the reduced scaling
frequency values, but caused stale HWP_CAP.GUARANTEED to be used in
the case at hand.
Fixes: eacc9c5a927e ("cpufreq: intel_pstate: Fix intel_pstate_get_hwp_max() for turbo disabled")
Reported-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Tested-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Cc: 5.8+ <stable@vger.kernel.org> # 5.8+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Make intel_pstate expose the ->adjust_perf() callback when it
operates in the passive mode with HWP enabled which causes the
schedutil governor to use that callback instead of ->fast_switch().
The minimum and target performance-level values passed by the
governor to ->adjust_perf() are converted to HWP.REQ.MIN and
HWP.REQ.DESIRED, respectively, which allows the processor to
adjust its configuration to maximize energy-efficiency while
providing sufficient capacity.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
First off, some cpufreq drivers (eg. intel_pstate) can pass hints
beyond the current target frequency to the hardware and there are no
provisions for doing that in the cpufreq framework. In particular,
today the driver has to assume that it should not allow the frequency
to fall below the one requested by the governor (or the required
capacity may not be provided) which may not be the case and which may
lead to excessive energy usage in some scenarios.
Second, the hints passed by these drivers to the hardware need not be
in terms of the frequency, so representing the utilization numbers
coming from the scheduler as frequency before passing them to those
drivers is not really useful.
Address the two points above by adding a special-purpose replacement
for the ->fast_switch callback, called ->adjust_perf, allowing the
governor to pass abstract performance level (rather than frequency)
values for the minimum (required) and target (desired) performance
along with the CPU capacity to compare them to.
Also update the schedutil governor to use the new callback instead
of ->fast_switch if present and if the utilization mertics are
frequency-invariant (that is requisite for the direct mapping
between the utilization and the CPU performance levels to be a
reasonable approximation).
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
The cppc_cpudata per-cpu storage was inefficient (1) additional to causing
functional issues (2) when CPUs are hotplugged out, due to per-cpu data
being improperly initialised.
(1) The amount of information needed for CPPC performance control in its
cpufreq driver depends on the domain (PSD) coordination type:
ANY: One set of CPPC control and capability data (e.g desired
performance, highest/lowest performance, etc) applies to all
CPUs in the domain.
ALL: Same as ANY. To be noted that this type is not currently
supported. When supported, information about which CPUs
belong to a domain is needed in order for frequency change
requests to be sent to each of them.
HW: It's necessary to store CPPC control and capability
information for all the CPUs. HW will then coordinate the
performance state based on their limitations and requests.
NONE: Same as HW. No HW coordination is expected.
Despite this, the previous initialisation code would indiscriminately
allocate memory for all CPUs (all_cpu_data) and unnecessarily
duplicate performance capabilities and the domain sharing mask and type
for each possible CPU.
(2) With the current per-cpu structure, when having ANY coordination,
the cppc_cpudata cpu information is not initialised (will remain 0)
for all CPUs in a policy, other than policy->cpu. When policy->cpu is
hotplugged out, the driver will incorrectly use the uninitialised (0)
value of the other CPUs when making frequency changes. Additionally,
the previous values stored in the perf_ctrls.desired_perf will be
lost when policy->cpu changes.
Therefore replace the array of per cpu data with a list. The memory for
each structure is allocated at policy init, where a single structure
can be allocated per policy, not per cpu. In order to accommodate the
struct list_head node in the cppc_cpudata structure, the now unused cpu
and cur_policy variables are removed.
For example, on a arm64 Juno platform with 6 CPUs: (0, 1, 2, 3) in PSD1,
(4, 5) in PSD2 - ANY coordination, the memory allocation comparison shows:
Before patch:
- ANY coordination:
total slack req alloc/free caller
0 0 0 0/1 _kernel_size_le_hi32+0x0xffff800008ff7810
0 0 0 0/6 _kernel_size_le_hi32+0x0xffff800008ff7808
128 80 48 1/0 _kernel_size_le_hi32+0x0xffff800008ffc070
768 0 768 6/0 _kernel_size_le_hi32+0x0xffff800008ffc0e4
After patch:
- ANY coordination:
total slack req alloc/free caller
256 0 256 2/0 _kernel_size_le_hi32+0x0xffff800008fed410
0 0 0 0/2 _kernel_size_le_hi32+0x0xffff800008fed274
Additional notes:
- A pointer to the policy's cppc_cpudata is stored in policy->driver_data
- Driver registration is skipped if _CPC entries are not present.
Signed-off-by: Ionela Voinescu <ionela.voinescu@arm.com>
Tested-by: Mian Yousaf Kaukab <ykaukab@suse.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Use the existing sysfs attribute "freqdomain_cpus" to expose
information to userspace about CPUs in the same frequency domain.
Signed-off-by: Ionela Voinescu <ionela.voinescu@arm.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Tested-by: Mian Yousaf Kaukab <ykaukab@suse.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
The previous coordination type handling in the cppc_cpufreq init code
created some confusion: the comment mentioned "Support only SW_ANY for
now" while only the SW_ALL/ALL case resulted in a failure. The other
coordination types (HW_ALL/HW, NONE) were silently supported.
Clarify support for coordination types while describing in comments the
intended behavior.
Signed-off-by: Ionela Voinescu <ionela.voinescu@arm.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Tested-by: Mian Yousaf Kaukab <ykaukab@suse.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Considering only the currently supported coordination types (ANY, HW,
NONE), this change only makes a difference for the ANY type, when
policy->cpu is hotplugged out. In that case the new policy->cpu will
be different from ((struct cppc_cpudata *)policy->driver_data)->cpu.
While in this case the controls of *ANY* CPU could be used to drive
frequency changes, it's more consistent to use policy->cpu as the
leading CPU, as used in all other cppc_cpufreq functions. Additionally,
the debug prints in cppc_set_perf() would no longer create confusion
when referring to a CPU that is hotplugged out.
Signed-off-by: Ionela Voinescu <ionela.voinescu@arm.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Tested-by: Mian Yousaf Kaukab <ykaukab@suse.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
* pm-cpufreq: (31 commits)
cpufreq: Fix cpufreq_online() return value on errors
cpufreq: Fix up several kerneldoc comments
cpufreq: stats: Use local_clock() instead of jiffies
cpufreq: schedutil: Simplify sugov_update_next_freq()
cpufreq: intel_pstate: Simplify intel_cpufreq_update_pstate()
cpufreq: arm_scmi: Discover the power scale in performance protocol
firmware: arm_scmi: Add power_scale_mw_get() interface
cpufreq: tegra194: Rename tegra194_get_speed_common function
cpufreq: tegra194: Remove unnecessary frequency calculation
cpufreq: tegra186: Simplify cluster information lookup
cpufreq: tegra186: Fix sparse 'incorrect type in assignment' warning
cpufreq: imx: fix NVMEM_IMX_OCOTP dependency
cpufreq: vexpress-spc: Add missing MODULE_ALIAS
cpufreq: scpi: Add missing MODULE_ALIAS
cpufreq: loongson1: Add missing MODULE_ALIAS
cpufreq: sun50i: Add missing MODULE_DEVICE_TABLE
cpufreq: st: Add missing MODULE_DEVICE_TABLE
cpufreq: qcom: Add missing MODULE_DEVICE_TABLE
cpufreq: mediatek: Add missing MODULE_DEVICE_TABLE
cpufreq: highbank: Add missing MODULE_DEVICE_TABLE
...
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm
Pull ARM cpufreq updates for 5.11-rc1 from Viresh Kumar:
"This contains the following updates:
- Fix imx's NVMEM_IMX_OCOTP dependency (Arnd Bergmann).
- Add support for mt8167 and blacklist mt8516 (Fabien Parent).
- Some ->get() callback related cleanups to the tegra194 driver and
some optimizations in tegra186 driver (Jon Hunter and Sumit Gupta).
- Power scale improvements to arm_scmi driver (Lukasz Luba).
- Add missing MODULE_DEVICE_TABLE and MODULE_ALIAS to several drivers
(Pali Rohár).
- Fix error path in mediatek driver (Qinglang Miao).
- Fix memleak in ST's cpufreq driver (Yangtao Li)."
* 'cpufreq/arm/linux-next' of git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm: (22 commits)
cpufreq: arm_scmi: Discover the power scale in performance protocol
firmware: arm_scmi: Add power_scale_mw_get() interface
cpufreq: tegra194: Rename tegra194_get_speed_common function
cpufreq: tegra194: Remove unnecessary frequency calculation
cpufreq: tegra186: Simplify cluster information lookup
cpufreq: tegra186: Fix sparse 'incorrect type in assignment' warning
cpufreq: imx: fix NVMEM_IMX_OCOTP dependency
cpufreq: vexpress-spc: Add missing MODULE_ALIAS
cpufreq: scpi: Add missing MODULE_ALIAS
cpufreq: loongson1: Add missing MODULE_ALIAS
cpufreq: sun50i: Add missing MODULE_DEVICE_TABLE
cpufreq: st: Add missing MODULE_DEVICE_TABLE
cpufreq: qcom: Add missing MODULE_DEVICE_TABLE
cpufreq: mediatek: Add missing MODULE_DEVICE_TABLE
cpufreq: highbank: Add missing MODULE_DEVICE_TABLE
cpufreq: ap806: Add missing MODULE_DEVICE_TABLE
cpufreq: mediatek: add missing platform_driver_unregister() on error in mtk_cpufreq_driver_init
cpufreq: tegra194: get consistent cpuinfo_cur_freq
cpufreq: blacklist mt8516 in cpufreq-dt-platdev
cpufreq: mediatek: Add support for mt8167
...
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm
Pull OPP (Operating Performance Points) updates for 5.11-rc1 from
Viresh Kumar:
"This contains the following updates:
- Allow empty (node-less) OPP tables in DT for passing just the
dependency related information (Nicola Mazzucato).
- Fix a potential lockdep in OPP core and other OPP core cleanups
(Viresh Kumar).
- Don't abuse dev_pm_opp_get_opp_table() to create an OPP table, fix
cpufreq-dt driver for the same (Viresh Kumar).
- dev_pm_opp_put_regulators() accepts a NULL argument now, updates to
all the users as well (Viresh Kumar)."
* 'opp/linux-next' of git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm:
opp: of: Allow empty opp-table with opp-shared
dt-bindings: opp: Allow empty OPP tables
media: venus: dev_pm_opp_put_*() accepts NULL argument
drm/panfrost: dev_pm_opp_put_*() accepts NULL argument
drm/lima: dev_pm_opp_put_*() accepts NULL argument
PM / devfreq: exynos: dev_pm_opp_put_*() accepts NULL argument
cpufreq: qcom-cpufreq-nvmem: dev_pm_opp_put_*() accepts NULL argument
cpufreq: dt: dev_pm_opp_put_regulators() accepts NULL argument
opp: Allow dev_pm_opp_put_*() APIs to accept NULL opp_table
opp: Don't create an OPP table from dev_pm_opp_get_opp_table()
cpufreq: dt: Don't (ab)use dev_pm_opp_get_opp_table() to create OPP table
opp: Reduce the size of critical section in _opp_kref_release()
opp: Don't return opp_dev from _find_opp_dev()
opp: Allocate the OPP table outside of opp_table_lock
opp: Always add entries in dev_list with opp_table->lock held
|
|
Make cpufreq_online() return negative error codes on all errors that
cause the policy to be destroyed, as appropriate.
Signed-off-by: Wang ShaoBo <bobo.shaobowang@huawei.com>
[ rjw: Subject and changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Fix up the remaining kerneldoc comments that don't adhere to the
expected format and clarify some of them a bit.
No functional changes.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
local_clock() has better precision and accuracy as compared to jiffies,
lets use it for time management in cpufreq stats.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Avoid doing the same assignment in both branches of a conditional,
do it after the whole conditional instead.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
|
|
The dev_pm_opp_put_*() APIs now accepts a NULL opp_table pointer and so
there is no need for us to carry the extra checks. Drop them.
Reviewed-by: Ilia Lin <ilia.lin@kernel.org>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
The dev_pm_opp_put_*() APIs now accepts a NULL opp_table pointer and so
there is no need for us to carry the extra checks. Drop them.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
Initially, the helper dev_pm_opp_get_opp_table() was supposed to be used
only for the OPP core's internal use (it tries to find an existing OPP
table and if it doesn't find one, then it allocates the OPP table).
Sometime back, the cpufreq-dt driver started using it to make sure all
the relevant resources required by the OPP core are available earlier
during initialization process to properly propagate -EPROBE_DEFER.
It worked but it also abused the API to create an OPP table, which
should be created with the help of other helpers provided by the OPP
core.
The OPP core will be updated in a later commit to limit the scope of
dev_pm_opp_get_opp_table() to only finding an existing OPP table and not
create one. This commit updates the cpufreq-dt driver before that
happens.
Now the cpufreq-dt driver creates the OPP and cpufreq tables for all the
CPUs from driver's init callback itself.
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
|
|
Add mechanism to discover the power scale present in the performance
protocol for all domains. Provide this information to Energy Model,
which then can be checked in other frameworks, e.g. thermal.
Suggested-by: Morten Rasmussen <morten.rasmussen@arm.com>
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
The function tegra194_get_speed_common() uses hardware timers to
calculate the current CPUFREQ and so rename this function to be
tegra194_calculate_speed() to reflect what it does.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
The Tegra194 CPUFREQ driver sets the CPUFREQ_NEED_INITIAL_FREQ_CHECK
flag which means that the CPUFREQ framework will call the 'get' callback
on boot to determine the current frequency of the CPUs. Therefore, it is
not necessary for the Tegra194 CPUFREQ driver to internally call the
tegra194_get_speed_common() during initialisation to query the current
frequency as well. Fix this by removing the call to the
tegra194_get_speed_common() during initialisation and simplify the code.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
