We are currently assuming that CEDE(0) has exit latency 10us, since
there is no way for us to query from the platform. However, if the
wakeup latency of an Extended CEDE state is smaller than 10us, then we
can be sure that the exit latency of CEDE(0) cannot be more than that.
In this patch, we fix the exit latency of CEDE(0) if we discover an
Extended CEDE state with wakeup latency smaller than 10us.
Benchmark results:
On POWER8, this patch does not have any impact since the advertized
latency of Extended CEDE (1) is 30us which is higher than the default
latency of CEDE (0) which is 10us.
On POWER9 we see improvement the single-threaded performance of
ebizzy, and no regression in the wakeup latency or the number of
context-switches.
ebizzy:
2 ebizzy threads bound to the same big-core. 25% improvement in the
avg records/s with patch.
x without_patch
* with_patch
N Min Max Median Avg Stddev
x 10 2491089 5834307 5398375 4244335 1596244.9
* 10 2893813 5834474 5832448 5327281.3 1055941.4
context_switch2:
There is no major regression observed with this patch as seen from the
context_switch2 benchmark.
context_switch2 across CPU0 CPU1 (Both belong to same big-core, but
different small cores). We observe a minor 0.14% regression in the
number of context-switches (higher is better).
x without_patch
* with_patch
N Min Max Median Avg Stddev
x 500 348872 362236 354712 354745.69 2711.827
* 500 349422 361452 353942 354215.4 2576.9258
Difference at 99.0% confidence
-530.288 +/- 430.963
-0.149484% +/- 0.121485%
(Student's t, pooled s = 2645.24)
context_switch2 across CPU0 CPU8 (Different big-cores). We observe a
0.37% improvement in the number of context-switches (higher is
better).
x without_patch
* with_patch
N Min Max Median Avg Stddev
x 500 287956 294940 288896 288977.23 646.59295
* 500 288300 294646 289582 290064.76 1161.9992
Difference at 99.0% confidence
1087.53 +/- 153.194
0.376337% +/- 0.0530125%
(Student's t, pooled s = 940.299)
schbench:
No major difference could be seen until the 99.9th percentile.
Without-patch:
Latency percentiles (usec)
50.0th: 29
75.0th: 39
90.0th: 49
95.0th: 59
*99.0th: 13104
99.5th: 14672
99.9th: 15824
min=0, max=17993
With-patch:
Latency percentiles (usec)
50.0th: 29
75.0th: 40
90.0th: 50
95.0th: 61
*99.0th: 13648
99.5th: 14768
99.9th: 15664
min=0, max=29812
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
[mpe: Minor formatting]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1596087177-30329-4-git-send-email-ego@linux.vnet.ibm.com
Currently we use CEDE with latency-hint 0 as the only other idle state
on a dedicated LPAR apart from the polling "snooze" state.
The platform might support additional extended CEDE idle states, which
can be discovered through the "ibm,get-system-parameter" rtas-call
made with CEDE_LATENCY_TOKEN.
This patch adds a function to obtain information about the extended
CEDE idle states from the platform and parse the contents to populate
an array of extended CEDE states. These idle states thus discovered
will be added to the cpuidle framework in the next patch.
dmesg on a POWER8 and POWER9 LPAR, demonstrating the output of parsing
the extended CEDE latency parameters are as follows
POWER8
[ 10.093279] xcede : xcede_record_size = 10
[ 10.093285] xcede : Record 0 : hint = 1, latency = 0x3c00 tb ticks, Wake-on-irq = 1
[ 10.093291] xcede : Record 1 : hint = 2, latency = 0x4e2000 tb ticks, Wake-on-irq = 0
[ 10.093297] cpuidle : Skipping the 2 Extended CEDE idle states
POWER9
[ 5.913180] xcede : xcede_record_size = 10
[ 5.913183] xcede : Record 0 : hint = 1, latency = 0x400 tb ticks, Wake-on-irq = 1
[ 5.913188] xcede : Record 1 : hint = 2, latency = 0x3e8000 tb ticks, Wake-on-irq = 0
[ 5.913193] cpuidle : Skipping the 2 Extended CEDE idle states
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
[mpe: Make space for 16 records, drop memset, minor cleanup & formatting]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1596087177-30329-3-git-send-email-ego@linux.vnet.ibm.com
As per the PAPR, each H_CEDE call is associated with a latency-hint to
be passed in the VPA field "cede_latency_hint". The CEDE states that
we were implicitly entering so far is CEDE with latency-hint = 0.
This patch explicitly sets the latency hint corresponding to the CEDE
state that we are currently entering. While at it, we save the
previous hint, to be restored once we wakeup from CEDE. This will be
required in the future when we expose extended-cede states through the
cpuidle framework, where each of them will have a different
cede-latency hint.
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
[mpe: Make cede_latency_hint static]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1596087177-30329-2-git-send-email-ego@linux.vnet.ibm.com
The sparse tool complains as follows:
drivers/cpuidle/cpuidle-pseries.c:25:23: warning:
symbol 'pseries_idle_driver' was not declared. Should it be static?
'pseries_idle_driver' is not used outside of this file, so marks
it static.
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200714142424.66648-1-weiyongjun1@huawei.com
Currently when CPU goes idle, we take a snapshot of PURR via
pseries_idle_prolog() which is used at the CPU idle exit to compute
the idle PURR cycles via the function pseries_idle_epilog(). Thus,
the value of idle PURR cycle thus read before pseries_idle_prolog() and
after pseries_idle_epilog() is always correct.
However, if we were to read the idle PURR cycles from an interrupt
context between pseries_idle_prolog() and pseries_idle_epilog() (this
will be done in a future patch), then, the value of the idle PURR thus
read will not include the cycles spent in the most recent idle period.
Thus, in that interrupt context, we will need access to the snapshot
of the PURR before going idle, in order to compute the idle PURR
cycles for the latest idle duration.
In this patch, we save the snapshot of PURR in pseries_idle_prolog()
in a per-cpu variable, instead of on the stack, so that it can be
accessed from an interrupt context.
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/1586249263-14048-3-git-send-email-ego@linux.vnet.ibm.com
Currently prior to entering an idle state on a Linux Guest, the
pseries cpuidle driver implement an idle_loop_prolog() and
idle_loop_epilog() functions which ensure that idle_purr is correctly
computed, and the hypervisor is informed that the CPU cycles have been
donated.
These prolog and epilog functions are also required in the default
idle call, i.e pseries_lpar_idle(). Hence move these accessor
functions to a common header file and call them from
pseries_lpar_idle(). Since the existing header files such as
asm/processor.h have enough clutter, create a new header file
asm/idle.h. Finally rename idle_loop_prolog() and idle_loop_epilog()
to pseries_idle_prolog() and pseries_idle_epilog() as they are only
relavent for on pseries guests.
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/1586249263-14048-2-git-send-email-ego@linux.vnet.ibm.com
When booting a pseries kernel with PREEMPT enabled, it dumps the
following warning:
BUG: using smp_processor_id() in preemptible [00000000] code: swapper/0/1
caller is pseries_processor_idle_init+0x5c/0x22c
CPU: 13 PID: 1 Comm: swapper/0 Not tainted 4.20.0-rc3-00090-g12201a0128bc-dirty #828
Call Trace:
[c000000429437ab0] [c0000000009c8878] dump_stack+0xec/0x164 (unreliable)
[c000000429437b00] [c0000000005f2f24] check_preemption_disabled+0x154/0x160
[c000000429437b90] [c000000000cab8e8] pseries_processor_idle_init+0x5c/0x22c
[c000000429437c10] [c000000000010ed4] do_one_initcall+0x64/0x300
[c000000429437ce0] [c000000000c54500] kernel_init_freeable+0x3f0/0x500
[c000000429437db0] [c0000000000112dc] kernel_init+0x2c/0x160
[c000000429437e20] [c00000000000c1d0] ret_from_kernel_thread+0x5c/0x6c
This happens because the code calls get_lppaca() which calls
get_paca() and it checks if preemption is disabled through
check_preemption_disabled().
Preemption should be disabled because the per CPU variable may make no
sense if there is a preemption (and a CPU switch) after it reads the
per CPU data and when it is used.
In this device driver specifically, it is not a problem, because this
code just needs to have access to one lppaca struct, and it does not
matter if it is the current per CPU lppaca struct or not (i.e. when
there is a preemption and a CPU migration).
That said, the most appropriate fix seems to be related to avoiding
the debug_smp_processor_id() call at get_paca(), instead of calling
preempt_disable() before get_paca().
Signed-off-by: Breno Leitao <leitao@debian.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
For shared processor guests (e.g., KVM), add an idle polling mode rather
than immediately returning to the hypervisor when the guest CPU goes
idle.
Test setup is a 2 socket POWER9 with 4 guests running, each with vCPUs
equal to 1/2 of real of CPUs. Saturated each guest with tbench. Using
polling idle gives about 1.4x throughput.
Kernel compile speed was not changed significantly.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Since e1689795a7 ("cpuidle: Add common time keeping and irq enabling"),
cpuidle drivers are expected to return from ->enter with irqs disabled.
Update the cpuidle-powernv snooze and cede loops to disable irqs before
returning.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Many source files in the tree are missing licensing information, which
makes it harder for compliance tools to determine the correct license.
By default all files without license information are under the default
license of the kernel, which is GPL version 2.
Update the files which contain no license information with the 'GPL-2.0'
SPDX license identifier. The SPDX identifier is a legally binding
shorthand, which can be used instead of the full boiler plate text.
This patch is based on work done by Thomas Gleixner and Kate Stewart and
Philippe Ombredanne.
How this work was done:
Patches were generated and checked against linux-4.14-rc6 for a subset of
the use cases:
- file had no licensing information it it.
- file was a */uapi/* one with no licensing information in it,
- file was a */uapi/* one with existing licensing information,
Further patches will be generated in subsequent months to fix up cases
where non-standard license headers were used, and references to license
had to be inferred by heuristics based on keywords.
The analysis to determine which SPDX License Identifier to be applied to
a file was done in a spreadsheet of side by side results from of the
output of two independent scanners (ScanCode & Windriver) producing SPDX
tag:value files created by Philippe Ombredanne. Philippe prepared the
base worksheet, and did an initial spot review of a few 1000 files.
The 4.13 kernel was the starting point of the analysis with 60,537 files
assessed. Kate Stewart did a file by file comparison of the scanner
results in the spreadsheet to determine which SPDX license identifier(s)
to be applied to the file. She confirmed any determination that was not
immediately clear with lawyers working with the Linux Foundation.
Criteria used to select files for SPDX license identifier tagging was:
- Files considered eligible had to be source code files.
- Make and config files were included as candidates if they contained >5
lines of source
- File already had some variant of a license header in it (even if <5
lines).
All documentation files were explicitly excluded.
The following heuristics were used to determine which SPDX license
identifiers to apply.
- when both scanners couldn't find any license traces, file was
considered to have no license information in it, and the top level
COPYING file license applied.
For non */uapi/* files that summary was:
SPDX license identifier # files
---------------------------------------------------|-------
GPL-2.0 11139
and resulted in the first patch in this series.
If that file was a */uapi/* path one, it was "GPL-2.0 WITH
Linux-syscall-note" otherwise it was "GPL-2.0". Results of that was:
SPDX license identifier # files
---------------------------------------------------|-------
GPL-2.0 WITH Linux-syscall-note 930
and resulted in the second patch in this series.
- if a file had some form of licensing information in it, and was one
of the */uapi/* ones, it was denoted with the Linux-syscall-note if
any GPL family license was found in the file or had no licensing in
it (per prior point). Results summary:
SPDX license identifier # files
---------------------------------------------------|------
GPL-2.0 WITH Linux-syscall-note 270
GPL-2.0+ WITH Linux-syscall-note 169
((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) 21
((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause) 17
LGPL-2.1+ WITH Linux-syscall-note 15
GPL-1.0+ WITH Linux-syscall-note 14
((GPL-2.0+ WITH Linux-syscall-note) OR BSD-3-Clause) 5
LGPL-2.0+ WITH Linux-syscall-note 4
LGPL-2.1 WITH Linux-syscall-note 3
((GPL-2.0 WITH Linux-syscall-note) OR MIT) 3
((GPL-2.0 WITH Linux-syscall-note) AND MIT) 1
and that resulted in the third patch in this series.
- when the two scanners agreed on the detected license(s), that became
the concluded license(s).
- when there was disagreement between the two scanners (one detected a
license but the other didn't, or they both detected different
licenses) a manual inspection of the file occurred.
- In most cases a manual inspection of the information in the file
resulted in a clear resolution of the license that should apply (and
which scanner probably needed to revisit its heuristics).
- When it was not immediately clear, the license identifier was
confirmed with lawyers working with the Linux Foundation.
- If there was any question as to the appropriate license identifier,
the file was flagged for further research and to be revisited later
in time.
In total, over 70 hours of logged manual review was done on the
spreadsheet to determine the SPDX license identifiers to apply to the
source files by Kate, Philippe, Thomas and, in some cases, confirmation
by lawyers working with the Linux Foundation.
Kate also obtained a third independent scan of the 4.13 code base from
FOSSology, and compared selected files where the other two scanners
disagreed against that SPDX file, to see if there was new insights. The
Windriver scanner is based on an older version of FOSSology in part, so
they are related.
Thomas did random spot checks in about 500 files from the spreadsheets
for the uapi headers and agreed with SPDX license identifier in the
files he inspected. For the non-uapi files Thomas did random spot checks
in about 15000 files.
In initial set of patches against 4.14-rc6, 3 files were found to have
copy/paste license identifier errors, and have been fixed to reflect the
correct identifier.
Additionally Philippe spent 10 hours this week doing a detailed manual
inspection and review of the 12,461 patched files from the initial patch
version early this week with:
- a full scancode scan run, collecting the matched texts, detected
license ids and scores
- reviewing anything where there was a license detected (about 500+
files) to ensure that the applied SPDX license was correct
- reviewing anything where there was no detection but the patch license
was not GPL-2.0 WITH Linux-syscall-note to ensure that the applied
SPDX license was correct
This produced a worksheet with 20 files needing minor correction. This
worksheet was then exported into 3 different .csv files for the
different types of files to be modified.
These .csv files were then reviewed by Greg. Thomas wrote a script to
parse the csv files and add the proper SPDX tag to the file, in the
format that the file expected. This script was further refined by Greg
based on the output to detect more types of files automatically and to
distinguish between header and source .c files (which need different
comment types.) Finally Greg ran the script using the .csv files to
generate the patches.
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Reviewed-by: Philippe Ombredanne <pombredanne@nexb.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A memory barrier is not required after the task wakes up,
only if we clear the polling flag before waking. The case
where we have work to do is the important one, so optimise
for it.
Reviewed-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Ensure these don't get put into bouncing cachelines.
Reviewed-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
local_irq_enable can cause interrupts to be taken which could
take significant amount of processing time. The idle process
should set its polling flag before this, so another process that
wakes it during this time will not have to send an IPI.
Expand the TIF_POLLING_NRFLAG coverage to as large as possible.
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The idle cpus which stay in snooze for a long period can degrade the
perfomance of the sibling cpus. If the cpu stays in snooze for more
than target residency of the next available idle state, then exit from
snooze. This gives a chance to the cpuidle governor to re-evaluate the
last idle state of the cpu to promote it to deeper idle states.
Signed-off-by: Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com>
Reviewed-by: Preeti U Murthy <preeti@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The only place where the time is invalid is when the ACPI_CSTATE_FFH entry
method is not set. Otherwise for all the drivers, the time can be correctly
measured.
Instead of duplicating the CPUIDLE_FLAG_TIME_VALID flag in all the drivers
for all the states, just invert the logic by replacing it by the flag
CPUIDLE_FLAG_TIME_INVALID, hence we can set this flag only for the acpi idle
driver, remove the former flag from all the drivers and invert the logic with
this flag in the different governor.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit d8c6ad3184 ("sched/idle, PPC: Remove redundant
cpuidle_idle_call()") reintroduced ppc64_runlatch_off/on() in the
pseries cpuidle backend driver. Hence the cleanup caused by the
commit "c0c4301c54adde05:pseries/cpuidle: Remove redundant call
to ppc64_runlatch_off() in cpu idle routines" in conjuction
with the commit d8c6ad3184 causes a build failure.
Signed-off-by: Preeti U Murthy <preeti@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Nicolas Pitre <nicolas.pitre@linaro.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Link: http://lkml.kernel.org/r/52FAFD2D.2090306@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The core idle loop now takes care of it. However a few things need
checking:
- Invocation of cpuidle_idle_call() in pseries_lpar_idle() happened
through arch_cpu_idle() and was therefore always preceded by a call
to ppc64_runlatch_off(). To preserve this property now that
cpuidle_idle_call() is invoked directly from core code, a call to
ppc64_runlatch_off() has been added to idle_loop_prolog() in
platforms/pseries/processor_idle.c.
- Similarly, cpuidle_idle_call() was followed by ppc64_runlatch_off()
so a call to the later has been added to idle_loop_epilog().
- And since arch_cpu_idle() always made sure to re-enable IRQs if they
were not enabled, this is now
done in idle_loop_epilog() as well.
The above was made in order to keep the execution flow close to the
original. I don't know if that was strictly necessary. Someone well
aquainted with the platform details might find some room for possible
optimizations.
Signed-off-by: Nicolas Pitre <nico@linaro.org>
Reviewed-by: Preeti U Murthy <preeti@linux.vnet.ibm.com>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linuxppc-dev@lists.ozlabs.org
Cc: linux-sh@vger.kernel.org
Cc: linux-pm@vger.kernel.org
Cc: Russell King <linux@arm.linux.org.uk>
Cc: linaro-kernel@lists.linaro.org
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/n/tip-47o4m03citrfg9y1vxic5asb@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
smt-snooze-delay was designed to disable NAP state or delay the entry
to the NAP state prior to adoption of cpuidle framework. This
is per-cpu variable. With the coming of CPUIDLE framework,
states can be disabled on per-cpu basis using the cpuidle/enable
sysfs entry.
Also, with the coming of cpuidle driver each state's target residency
is per-driver unlike earlier which was per-device. Therefore,
the per-cpu sysfs smt-snooze-delay which decides the target residency
of the idle state on a particular cpu causes more confusion to the user
as we cannot have different smt-snooze-delay (target residency)
values for each cpu.
In the current code, smt-snooze-delay functionality is completely broken.
It makes sense to remove smt-snooze-delay from idle driver with the
coming of cpuidle framework.
However, sysfs files are retained as ppc64_util currently
utilises it. Once we fix ppc64_util, propose to clean
up the kernel code.
Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This patch removes the usage of MAX_IDLE_STATE macro
and dead code around it. The number of states
are determined at run time based on the cpuidle
state table selected on a given platform
Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently cpuidle-pseries backend driver cannot be
built as a module due to dependencies wrt cpuidle framework.
This patch removes all the module related code in the driver.
Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This patch replaces the cpuidle driver and devices initialisation
calls with a single generic cpuidle_register() call
and also includes minor refactoring of the code around it.
Remove the cpu online check in snooze loop, as this code can
only locally run on a cpu only if it is online. Therefore,
this check is not required.
Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Move the file from arch specific pseries/processor_idle.c
to drivers/cpuidle/cpuidle-pseries.c
Make the relevant Makefile and Kconfig changes.
Also, introduce Kconfig.powerpc in drivers/cpuidle
for all powerpc cpuidle drivers.
Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Gautham R. Shenoy