path: root/kernel/Kconfig.preempt

choice
	prompt "Preemption Mode"
	default PREEMPT_RT

config PREEMPT_NONE
	bool "No Forced Preemption (Server)"
	help
	  This is the traditional Linux preemption model geared towards
	  throughput. It will still provide good latencies most of the
	  time but there are no guarantees and occasional long delays
	  are possible.

	  Select this option if you are building a kernel for a server or
	  scientific/computation system, or if you want to maximize the
	  raw processing power of the kernel, irrespective of scheduling
	  latencies.

config PREEMPT_VOLUNTARY
	bool "Voluntary Kernel Preemption (Desktop)"
	help
	  This option reduces the latency of the kernel by adding more
	  "explicit preemption points" to the kernel code. These new
	  preemption points have been selected to minimize the maximum
	  latency of rescheduling, providing faster application reactions,
	  at the cost of slightly lower throughput.

	  This allows reaction to interactive events by allowing a
	  low priority process to voluntarily preempt itself even if it
	  is in kernel mode executing a system call. This allows
	  applications to run more 'smoothly' even when the system is
	  under load.

	  Select this if you are building a kernel for a desktop system.

config PREEMPT_DESKTOP
	bool "Preemptible Kernel (Low-Latency Desktop)"
	help
	  This option reduces the latency of the kernel by making
	  all kernel code that is not executing in a critical section
	  preemptible.  This allows reaction to interactive events by
	  permitting a low priority process to be preempted involuntarily
	  even if it is in kernel mode executing a system call and would
	  otherwise not about to reach a preemption point.  This allows
	  applications to run more 'smoothly' even when the system is
	  under load, at the cost of slighly lower throughput and a
	  slight runtime overhead to kernel code.

	  (According to profiles, when this mode is selected then even
	  during kernel-intense workloads the system is in an immediately
	  preemptible state more than 50% of the time.)

	  Select this if you are building a kernel for a desktop or
	  embedded system with latency requirements in the milliseconds
	  range.

config PREEMPT_RT
	bool "Complete Preemption (Real-Time)"
	select PREEMPT_SOFTIRQS
	select PREEMPT_HARDIRQS
	select PREEMPT_RCU
	select RT_MUTEXES
	help
	  This option further reduces the scheduling latency of the
	  kernel by replacing almost every spinlock used by the kernel
	  with preemptible mutexes and thus making all but the most
	  critical kernel code involuntarily preemptible. The remaining
	  handful of lowlevel non-preemptible codepaths are short and
	  have a deterministic latency of a couple of tens of
	  microseconds (depending on the hardware).  This also allows
	  applications to run more 'smoothly' even when the system is
	  under load, at the cost of lower throughput and runtime
	  overhead to kernel code.

	  (According to profiles, when this mode is selected then even
	  during kernel-intense workloads the system is in an immediately
	  preemptible state more than 95% of the time.)

	  Select this if you are building a kernel for a desktop,
	  embedded or real-time system with guaranteed latency
	  requirements of 100 usecs or lower.

endchoice

config PREEMPT
	bool
	default y
	depends on PREEMPT_DESKTOP || PREEMPT_RT

config PREEMPT_SOFTIRQS
	bool "Thread Softirqs"
	default n
#	depends on PREEMPT
	help
	  This option reduces the latency of the kernel by 'threading'
          soft interrupts. This means that all softirqs will execute
          in softirqd's context. While this helps latency, it can also
          reduce performance.

          The threading of softirqs can also be controlled via
          /proc/sys/kernel/softirq_preemption runtime flag and the
          sofirq-preempt=0/1 boot-time option.

	  Say N if you are unsure.

config PREEMPT_HARDIRQS
	bool "Thread Hardirqs"
	default n
	depends on GENERIC_HARDIRQS_NO__DO_IRQ
	select PREEMPT_SOFTIRQS
	help
	  This option reduces the latency of the kernel by 'threading'
          hardirqs. This means that all (or selected) hardirqs will run
          in their own kernel thread context. While this helps latency,
          this feature can also reduce performance.

          The threading of hardirqs can also be controlled via the
          /proc/sys/kernel/hardirq_preemption runtime flag and the
          hardirq-preempt=0/1 boot-time option. Per-irq threading can
          be enabled/disable via the /proc/irq/<IRQ>/<handler>/threaded
          runtime flags.

	  Say N if you are unsure.