linux/drivers/cpuidle
Gautham R. Shenoy 50741b70b0 cpuidle: pseries: Fixup CEDE0 latency only for POWER10 onwards
Commit d947fb4c96 ("cpuidle: pseries: Fixup exit latency for
CEDE(0)") sets the exit latency of CEDE(0) based on the latency values
of the Extended CEDE states advertised by the platform

On POWER9 LPARs, the firmwares advertise a very low value of 2us for
CEDE1 exit latency on a Dedicated LPAR. The latency advertized by the
PHYP hypervisor corresponds to the latency required to wakeup from the
underlying hardware idle state. However the wakeup latency from the
LPAR perspective should include

1. The time taken to transition the CPU from the Hypervisor into the
   LPAR post wakeup from platform idle state

2. Time taken to send the IPI from the source CPU (waker) to the idle
   target CPU (wakee).

1. can be measured via timer idle test, where we queue a timer, say
for 1ms, and enter the CEDE state. When the timer fires, in the timer
handler we compute how much extra timer over the expected 1ms have we
consumed. On a a POWER9 LPAR the numbers are

CEDE latency measured using a timer (numbers in ns)
N       Min      Median   Avg       90%ile  99%ile    Max    Stddev
400     2601     5677     5668.74    5917    6413     9299   455.01

1. and 2. combined can be determined by an IPI latency test where we
send an IPI to an idle CPU and in the handler compute the time
difference between when the IPI was sent and when the handler ran. We
see the following numbers on POWER9 LPAR.

CEDE latency measured using an IPI (numbers in ns)
N       Min      Median   Avg       90%ile  99%ile    Max    Stddev
400     711      7564     7369.43   8559    9514      9698   1200.01

Suppose, we consider the 99th percentile latency value measured using
the IPI to be the wakeup latency, the value would be 9.5us This is in
the ballpark of the default value of 10us.

Hence, use the exit latency of CEDE(0) based on the latency values
advertized by platform only from POWER10 onwards. The values
advertized on POWER10 platforms is more realistic and informed by the
latency measurements. For earlier platforms stick to the default value
of 10us. The fix was suggested by Michael Ellerman.

Fixes: d947fb4c96 ("cpuidle: pseries: Fixup exit latency for CEDE(0)")
Reported-by: Enrico Joedecke <joedecke@de.ibm.com>
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1626676399-15975-2-git-send-email-ego@linux.vnet.ibm.com
2021-08-03 22:33:19 +10:00
..
governors Power management updates for 5.14-rc1 2021-06-29 13:36:06 -07:00
coupled.c smp: Cleanup smp_call_function*() 2020-11-24 16:47:49 +01:00
cpuidle-arm.c thermal: cpuidle: Register cpuidle cooling device 2020-05-19 12:58:07 +02:00
cpuidle-at91.c drivers/cpuidle: Convert non-modular drivers to use builtin_platform_driver 2015-06-16 14:12:38 -04:00
cpuidle-big_little.c cpuidle: big.LITTLE: enable driver only on Peach-Pit/Pi Chromebooks 2020-10-26 10:15:24 +01:00
cpuidle-calxeda.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 201 2019-05-30 11:29:52 -07:00
cpuidle-clps711x.c cpuidle: clps711x: convert to devm_platform_ioremap_resource() 2019-12-20 10:04:27 +01:00
cpuidle-cps.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 152 2019-05-30 11:26:32 -07:00
cpuidle-exynos.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 500 2019-06-19 17:09:55 +02:00
cpuidle-haltpoll.c cpuidle-haltpoll: Fix small typo 2020-04-08 14:35:45 +02:00
cpuidle-kirkwood.c cpuidle: kirkwood: convert to devm_platform_ioremap_resource() 2019-12-20 10:05:44 +01:00
cpuidle-mvebu-v7.c cpuidle: mvebu: disable the bind/unbind attributes and use builtin_platform_driver 2015-10-23 12:40:48 +02:00
cpuidle-powernv.c powerpc/powernv/idle: add a basic stop 0-3 driver for POWER10 2020-09-15 22:13:38 +10:00
cpuidle-psci-domain.c cpuidle: psci: Enable suspend-to-idle for PSCI OSI mode 2020-11-10 20:42:01 +01:00
cpuidle-psci.c cpuidle: psci: Enable suspend-to-idle for PSCI OSI mode 2020-11-10 20:42:01 +01:00
cpuidle-psci.h cpuidle: psci: Fix error path via converting to a platform driver 2020-07-29 18:38:30 +02:00
cpuidle-pseries.c cpuidle: pseries: Fixup CEDE0 latency only for POWER10 onwards 2021-08-03 22:33:19 +10:00
cpuidle-qcom-spm.c cpuidle: qcom: Add SPM register data for MSM8226 2021-06-16 20:03:26 +02:00
cpuidle-tegra.c cpuidle: tegra: Remove do_idle firmware call 2021-04-08 19:54:14 +02:00
cpuidle-ux500.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 500 2019-06-19 17:09:55 +02:00
cpuidle-zynq.c treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 201 2019-05-30 11:29:52 -07:00
cpuidle.c cpuidle: Select polling interval based on a c-state with a longer target residency 2020-12-01 17:59:18 +01:00
cpuidle.h cpuidle: allow governor switch on cpuidle_register_driver() 2019-09-11 17:36:30 +02:00
driver.c cpuidle: Use s64 as exit_latency_ns and target_residency_ns data type 2021-04-07 19:26:44 +02:00
dt_idle_states.c cpuidle: dt: Support hierarchical CPU idle states 2020-01-02 16:50:08 +01:00
dt_idle_states.h License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
governor.c cpuidle: Call cpu_latency_qos_limit() instead of pm_qos_request() 2020-02-13 11:26:57 +01:00
Kconfig cpuidle: Fix Kconfig indentation 2019-11-29 11:47:35 +01:00
Kconfig.arm cpuidle: Fix ARM_QCOM_SPM_CPUIDLE configuration 2021-04-08 19:54:14 +02:00
Kconfig.mips treewide: Add SPDX license identifier - Makefile/Kconfig 2019-05-21 10:50:46 +02:00
Kconfig.powerpc treewide: Add SPDX license identifier - Makefile/Kconfig 2019-05-21 10:50:46 +02:00
Makefile cpuidle: psci: Split into two separate build objects 2020-07-29 18:38:30 +02:00
poll_state.c cpuidle: Drop disabled field from struct cpuidle_state 2019-11-29 11:48:39 +01:00
sysfs.c cpuidle: record state entry rejection statistics 2020-09-23 14:10:31 +02:00