REDUCING OS JITTER DUE TO PER-CPU KTHREADS
This document lists per-CPU kthreads in the Linux kernel and presents
options to control their OS jitter. Note that non-per-CPU kthreads are
not listed here. To reduce OS jitter from non-per-CPU kthreads, bind
them to a "housekeeping" CPU dedicated to such work.
REFERENCES
o Documentation/IRQ-affinity.txt: Binding interrupts to sets of CPUs.
o Documentation/cgroup-v1: Using cgroups to bind tasks to sets of CPUs.
o man taskset: Using the taskset command to bind tasks to sets
of CPUs.
o man sched_setaffinity: Using the sched_setaffinity() system
call to bind tasks to sets of CPUs.
o /sys/devices/system/cpu/cpuN/online: Control CPU N's hotplug state,
writing "0" to offline and "1" to online.
o In order to locate kernel-generated OS jitter on CPU N:
cd /sys/kernel/debug/tracing
echo 1 > max_graph_depth # Increase the "1" for more detail
echo function_graph > current_tracer
# run workload
cat per_cpu/cpuN/trace
KTHREADS
Name: ehca_comp/%u
Purpose: Periodically process Infiniband-related work.
To reduce its OS jitter, do any of the following:
1. Don't use eHCA Infiniband hardware, instead choosing hardware
that does not require per-CPU kthreads. This will prevent these
kthreads from being created in the first place. (This will
work for most people, as this hardware, though important, is
relatively old and is produced in relatively low unit volumes.)
2. Do all eHCA-Infiniband-related work on other CPUs, including
interrupts.
3. Rework the eHCA driver so that its per-CPU kthreads are
provisioned only on selected CPUs.
Name: irq/%d-%s
Purpose: Handle threaded interrupts.
To reduce its OS jitter, do the following:
1. Use irq affinity to force the irq threads to execute on
some other CPU.
Name: kcmtpd_ctr_%d
Purpose: Handle Bluetooth work.
To reduce its OS jitter, do one of the following:
1. Don't use Bluetooth, in which case these kthreads won't be
created in the first place.
2. Use irq affinity to force Bluetooth-related interrupts to
occur on some other CPU and furthermore initiate all
Bluetooth activity on some other CPU.
Name: ksoftirqd/%u
Purpose: Execute softirq handlers when threaded or when under heavy load.
To reduce its OS jitter, each softirq vector must be handled
separately as follows:
TIMER_SOFTIRQ: Do all of the following:
1. To the extent possible, keep the CPU out of the kernel when it
is non-idle, for example, by avoiding system calls and by forcing
both kernel threads and interrupts to execute elsewhere.
2. Build with CONFIG_HOTPLUG_CPU=y. After boot completes, force
the CPU offline, then bring it back online. This forces
recurring timers to migrate elsewhere. If you are concerned
with multiple CPUs, force them all offline before bringing the
first one back online. Once you have onlined the CPUs in question,
do not offline any other CPUs, because doing so could force the
timer back onto one of the CPUs in question.
NET_TX_SOFTIRQ and NET_RX_SOFTIRQ: Do all of the following:
1. Force networking interrupts onto other CPUs.
2. Initiate any network I/O on other CPUs.
3. Once your application has started, prevent CPU-hotplug operations
from being initiated from tasks that might run on the CPU to
be de-jittered. (It is OK to force this CPU offline and then
bring it back online before you start your application.)
BLOCK_SOFTIRQ: Do all of the following:
1. Force block-device interrupts onto some other CPU.
2. Initiate any block I/O on other CPUs.
3. Once your application has started, prevent CPU-hotplug operations
from being initiated from tasks that might run on the CPU to
be de-jittered. (It is OK to force this CPU offline and then
bring it back online before you start your application.)
IRQ_POLL_SOFTIRQ: Do all of the following:
1. Force block-device interrupts onto some other CPU.
2. Initiate any block I/O and block-I/O polling on other CPUs.
3. Once your application has started, prevent CPU-hotplug operations
from being initiated from tasks that might run on the CPU to
be de-jittered. (It is OK to force this CPU offline and then
bring it back online before you start your application.)
TASKLET_SOFTIRQ: Do one or more of the following:
1. Avoid use of drivers that use tasklets. (Such drivers will contain
calls to things like tasklet_schedule().)
2. Convert all drivers that you must use from tasklets to workqueues.
3. Force interrupts for drivers using tasklets onto other CPUs,
and also do I/O involving these drivers on other CPUs.
SCHED_SOFTIRQ: Do all of the following:
1. Avoid sending scheduler IPIs to the CPU to be de-jittered,
for example, ensure that at most one runnable kthread is present
on that CPU. If a thread that expects to run on the de-jittered
CPU awakens, the scheduler will send an IPI that can result in
a subsequent SCHED_SOFTIRQ.
2. Build with CONFIG_RCU_NOCB_CPU=y, CONFIG_RCU_NOCB_CPU_ALL=y,
CONFIG_NO_HZ_FULL=y, and, in addition, ensure that the CPU
to be de-jittered is marked as an adaptive-ticks CPU using the
"nohz_full=" boot parameter. This reduces the number of
scheduler-clock interrupts that the de-jittered CPU receives,
minimizing its chances of being selected to do the load balancing
work that runs in SCHED_SOFTIRQ context.
3. To the extent possible, keep the CPU out of the kernel when it
is non-idle, for example, by avoiding system calls and by
forcing both kernel threads and interrupts to execute elsewhere.
This further reduces the number of scheduler-clock interrupts
received by the de-jittered CPU.
HRTIMER_SOFTIRQ: Do all of the following:
1. To the extent possible, keep the CPU out of the kernel when it
is non-idle. For example, avoid system calls and force both
kernel threads and interrupts to execute elsewhere.
2. Build with CONFIG_HOTPLUG_CPU=y. Once boot completes, force the
CPU offline, then bring it back online. This forces recurring
timers to migrate elsewhere. If you are concerned with multiple
CPUs, force them all offline before bringing the first one
back online. Once you have onlined the CPUs in question, do not
offline any other CPUs, because doing so could force the timer
back onto one of the CPUs in question.
RCU_SOFTIRQ: Do at least one of the following:
1. Offload callbacks and keep the CPU in either dyntick-idle or
adaptive-ticks state by doing all of the following:
a. Build with CONFIG_RCU_NOCB_CPU=y, CONFIG_RCU_NOCB_CPU_ALL=y,
CONFIG_NO_HZ_FULL=y, and, in addition ensure that the CPU
to be de-jittered is marked as an adaptive-ticks CPU using
the "nohz_full=" boot parameter. Bind the rcuo kthreads
to housekeeping CPUs, which can tolerate OS jitter.
b. To the extent possible, keep the CPU out of the kernel
when it is non-idle, for example, by avoiding system
calls and by forcing both kernel threads and interrupts
to execute elsewhere.
2. Enable RCU to do its processing remotely via dyntick-idle by
doing all of the following:
a. Build with CONFIG_NO_HZ=y and CONFIG_RCU_FAST_NO_HZ=y.
b. Ensure that the CPU goes idle frequently, allowing other
CPUs to detect that it has passed through an RCU quiescent
state. If the kernel is built with CONFIG_NO_HZ_FULL=y,
userspace execution also allows other CPUs to detect that
the CPU in question has passed through a quiescent state.
c. To the extent possible, keep the CPU out of the kernel
when it is non-idle, for example, by avoiding system
calls and by forcing both kernel threads and interrupts
to execute elsewhere.
Name: kworker/%u:%d%s (cpu, id, priority)
Purpose: Execute workqueue requests
To reduce its OS jitter, do any of the following:
1. Run your workload at a real-time priority, which will allow
preempting the kworker daemons.
2. A given workqueue can be made visible in the sysfs filesystem
by passing the WQ_SYSFS to that workqueue's alloc_workqueue().
Such a workqueue can be confined to a given subset of the
CPUs using the /sys/devices/virtual/workqueue/*/cpumask sysfs
files. The set of WQ_SYSFS workqueues can be displayed using
"ls sys/devices/virtual/workqueue". That said, the workqueues
maintainer would like to caution people against indiscriminately
sprinkling WQ_SYSFS across all the workqueues. The reason for
caution is that it is easy to add WQ_SYSFS, but because sysfs is
part of the formal user/kernel API, it can be nearly impossible
to remove it, even if its addition was a mistake.
3. Do any of the following needed to avoid jitter that your
application cannot tolerate:
a. Build your kernel with CONFIG_SLUB=y rather than
CONFIG_SLAB=y, thus avoiding the slab allocator's periodic
use of each CPU's workqueues to run its cache_reap()
function.
b. Avoid using oprofile, thus avoiding OS jitter from
wq_sync_buffer().
c. Limit your CPU frequency so that a CPU-frequency
governor is not required, possibly enlisting the aid of
special heatsinks or other cooling technologies. If done
correctly, and if you CPU architecture permits, you should
be able to build your kernel with CONFIG_CPU_FREQ=n to
avoid the CPU-frequency governor periodically running
on each CPU, including cs_dbs_timer() and od_dbs_timer().
WARNING: Please check your CPU specifications to
make sure that this is safe on your particular system.
d. As of v3.18, Christoph Lameter's on-demand vmstat workers
commit prevents OS jitter due to vmstat_update() on
CONFIG_SMP=y systems. Before v3.18, is not possible
to entirely get rid of the OS jitter, but you can
decrease its frequency by writing a large value to
/proc/sys/vm/stat_interval. The default value is HZ,
for an interval of one second. Of course, larger values
will make your virtual-memory statistics update more
slowly. Of course, you can also run your workload at
a real-time priority, thus preempting vmstat_update(),
but if your workload is CPU-bound, this is a bad idea.
However, there is an RFC patch from Christoph Lameter
(based on an earlier one from Gilad Ben-Yossef) that
reduces or even eliminates vmstat overhead for some
workloads at https://lkml.org/lkml/2013/9/4/379.
e. Boot with "elevator=noop" to avoid workqueue use by
the block layer.
f. If running on high-end powerpc servers, build with
CONFIG_PPC_RTAS_DAEMON=n. This prevents the RTAS
daemon from running on each CPU every second or so.
(This will require editing Kconfig files and will defeat
this platform's RAS functionality.) This avoids jitter
due to the rtas_event_scan() function.
WARNING: Please check your CPU specifications to
make sure that this is safe on your particular system.
g. If running on Cell Processor, build your kernel with
CBE_CPUFREQ_SPU_GOVERNOR=n to avoid OS jitter from
spu_gov_work().
WARNING: Please check your CPU specifications to
make sure that this is safe on your particular system.
h. If running on PowerMAC, build your kernel with
CONFIG_PMAC_RACKMETER=n to disable the CPU-meter,
avoiding OS jitter from rackmeter_do_timer().
Name: rcuc/%u
Purpose: Execute RCU callbacks in CONFIG_RCU_BOOST=y kernels.
To reduce its OS jitter, do at least one of the following:
1. Build the kernel with CONFIG_PREEMPT=n. This prevents these
kthreads from being created in the first place, and also obviates
the need for RCU priority boosting. This approach is feasible
for workloads that do not require high degrees of responsiveness.
2. Build the kernel with CONFIG_RCU_BOOST=n. This prevents these
kthreads from being created in the first place. This approach
is feasible only if your workload never requires RCU priority
boosting, for example, if you ensure frequent idle time on all
CPUs that might execute within the kernel.
3. Build with CONFIG_RCU_NOCB_CPU=y and CONFIG_RCU_NOCB_CPU_ALL=y,
which offloads all RCU callbacks to kthreads that can be moved
off of CPUs susceptible to OS jitter. This approach prevents the
rcuc/%u kthreads from having any work to do, so that they are
never awakened.
4. Ensure that the CPU never enters the kernel, and, in particular,
avoid initiating any CPU hotplug operations on this CPU. This is
another way of preventing any callbacks from being queued on the
CPU, again preventing the rcuc/%u kthreads from having any work
to do.
Name: rcuob/%d, rcuop/%d, and rcuos/%d
Purpose: Offload RCU callbacks from the corresponding CPU.
To reduce its OS jitter, do at least one of the following:
1. Use affinity, cgroups, or other mechanism to force these kthreads
to execute on some other CPU.
2. Build with CONFIG_RCU_NOCB_CPU=n, which will prevent these
kthreads from being created in the first place. However, please
note that this will not eliminate OS jitter, but will instead
shift it to RCU_SOFTIRQ.
Name: watchdog/%u
Purpose: Detect software lockups on each CPU.
To reduce its OS jitter, do at least one of the following:
1. Build with CONFIG_LOCKUP_DETECTOR=n, which will prevent these
kthreads from being created in the first place.
2. Boot with "nosoftlockup=0", which will also prevent these kthreads
from being created. Other related watchdog and softlockup boot
parameters may be found in Documentation/kernel-parameters.txt
and Documentation/watchdog/watchdog-parameters.txt.
3. Echo a zero to /proc/sys/kernel/watchdog to disable the
watchdog timer.
4. Echo a large number of /proc/sys/kernel/watchdog_thresh in
order to reduce the frequency of OS jitter due to the watchdog
timer down to a level that is acceptable for your workload.
