linux/arch
Rafael J. Wysocki d11a1d08a0 cpufreq: ACPI: Update arch scale-invariance max perf ratio if CPPC is not there
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 db865272d9 ("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: db865272d9 ("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>
2021-02-08 13:45:51 +01:00
..
alpha local64.h: make <asm/local64.h> mandatory 2020-12-29 15:36:49 -08:00
arc ARC: [hsdk]: Enable FPU_SAVE_RESTORE 2021-01-08 13:46:58 -08:00
arm ARM fixes for 5.11: 2021-02-06 15:07:51 -08:00
arm64 x86 has lots of small bugfixes, mostly one liners. It's quite late in 2021-02-05 10:03:01 -08:00
c6x tif-task_work.arch-2020-12-14 2020-12-16 12:33:35 -08:00
csky local64.h: make <asm/local64.h> mandatory 2020-12-29 15:36:49 -08:00
h8300 local64.h: make <asm/local64.h> mandatory 2020-12-29 15:36:49 -08:00
hexagon local64.h: make <asm/local64.h> mandatory 2020-12-29 15:36:49 -08:00
ia64 Kbuild fixes for v5.11 (2nd) 2021-02-07 09:37:37 -08:00
m68k local64.h: make <asm/local64.h> mandatory 2020-12-29 15:36:49 -08:00
microblaze local64.h: make <asm/local64.h> mandatory 2020-12-29 15:36:49 -08:00
mips mips/mm/highmem: use set_pte() for kmap_local() 2021-01-24 10:34:52 -08:00
nds32 local64.h: make <asm/local64.h> mandatory 2020-12-29 15:36:49 -08:00
nios2 tif-task_work.arch-2020-12-14 2020-12-16 12:33:35 -08:00
openrisc openrisc: io: Add missing __iomem annotation to iounmap() 2021-01-20 06:14:26 +09:00
parisc parisc: Enable -mlong-calls gcc option by default when !CONFIG_MODULES 2021-01-26 20:16:21 +01:00
powerpc powerpc fixes for 5.11 #7 2021-02-06 15:14:39 -08:00
riscv RISC-V: Define MAXPHYSMEM_1GB only for RV32 2021-02-02 18:36:42 -08:00
s390 s390: uv: Fix sysfs max number of VCPUs reporting 2021-01-27 13:00:04 +01:00
sh Cleanup and warning fixes. 2021-01-24 13:52:02 -08:00
sparc sparc/mm/highmem: flush cache and TLB 2021-01-24 10:34:52 -08:00
um um: time: fix initialization in time-travel mode 2021-01-26 22:11:38 +01:00
x86 cpufreq: ACPI: Update arch scale-invariance max perf ratio if CPPC is not there 2021-02-08 13:45:51 +01:00
xtensa local64.h: make <asm/local64.h> mandatory 2020-12-29 15:36:49 -08:00
.gitignore
Kconfig fanotify: Fix sys_fanotify_mark() on native x86-32 2020-12-28 11:58:59 +01:00