1 # SPDX-License-Identifier: GPL-2.0-only
4 prompt "Preemption Model"
5 default PREEMPT_NONE_BEHAVIOUR
7 config PREEMPT_NONE_BEHAVIOUR
8 bool "No Forced Preemption (Server)"
9 select PREEMPT_NONE if !PREEMPT_DYNAMIC
11 This is the traditional Linux preemption model, geared towards
12 throughput. It will still provide good latencies most of the
13 time, but there are no guarantees and occasional longer delays
16 Select this option if you are building a kernel for a server or
17 scientific/computation system, or if you want to maximize the
18 raw processing power of the kernel, irrespective of scheduling
21 config PREEMPT_VOLUNTARY_BEHAVIOUR
22 bool "Voluntary Kernel Preemption (Desktop)"
23 depends on !ARCH_NO_PREEMPT
24 select PREEMPT_VOLUNTARY if !PREEMPT_DYNAMIC
26 This option reduces the latency of the kernel by adding more
27 "explicit preemption points" to the kernel code. These new
28 preemption points have been selected to reduce the maximum
29 latency of rescheduling, providing faster application reactions,
30 at the cost of slightly lower throughput.
32 This allows reaction to interactive events by allowing a
33 low priority process to voluntarily preempt itself even if it
34 is in kernel mode executing a system call. This allows
35 applications to run more 'smoothly' even when the system is
38 Select this if you are building a kernel for a desktop system.
40 config PREEMPT_BEHAVIOUR
41 bool "Preemptible Kernel (Low-Latency Desktop)"
42 depends on !ARCH_NO_PREEMPT
45 This option reduces the latency of the kernel by making
46 all kernel code (that is not executing in a critical section)
47 preemptible. This allows reaction to interactive events by
48 permitting a low priority process to be preempted involuntarily
49 even if it is in kernel mode executing a system call and would
50 otherwise not be about to reach a natural preemption point.
51 This allows applications to run more 'smoothly' even when the
52 system is under load, at the cost of slightly lower throughput
53 and a slight runtime overhead to kernel code.
55 Select this if you are building a kernel for a desktop or
56 embedded system with latency requirements in the milliseconds
60 bool "Fully Preemptible Kernel (Real-Time)"
61 depends on EXPERT && ARCH_SUPPORTS_RT && !PREEMPT_DYNAMIC
64 This option turns the kernel into a real-time kernel by replacing
65 various locking primitives (spinlocks, rwlocks, etc.) with
66 preemptible priority-inheritance aware variants, enforcing
67 interrupt threading and introducing mechanisms to break up long
68 non-preemptible sections. This makes the kernel, except for very
69 low level and critical code paths (entry code, scheduler, low
70 level interrupt handling) fully preemptible and brings most
71 execution contexts under scheduler control.
73 Select this if you are building a kernel for systems which
74 require real-time guarantees.
81 config PREEMPT_VOLUNTARY
87 select UNINLINE_SPIN_UNLOCK if !ARCH_INLINE_SPIN_UNLOCK
96 config PREEMPT_DYNAMIC
97 bool "Preemption behaviour defined on boot"
98 depends on HAVE_PREEMPT_DYNAMIC
102 This option allows to define the preemption model on the kernel
103 command line parameter and thus override the default preemption
104 model defined during compile time.
106 The feature is primarily interesting for Linux distributions which
107 provide a pre-built kernel binary to reduce the number of kernel
108 flavors they offer while still offering different usecases.
110 The runtime overhead is negligible with HAVE_STATIC_CALL_INLINE enabled
111 but if runtime patching is not available for the specific architecture
112 then the potential overhead should be considered.
114 Interesting if you want the same pre-built kernel should be used for
115 both Server and Desktop workloads.
118 bool "Core Scheduling for SMT"
121 This option permits Core Scheduling, a means of coordinated task
122 selection across SMT siblings. When enabled -- see
123 prctl(PR_SCHED_CORE) -- task selection ensures that all SMT siblings
124 will execute a task from the same 'core group', forcing idle when no
125 matching task is found.
127 Use of this feature includes:
128 - mitigation of some (not all) SMT side channels;
129 - limiting SMT interference to improve determinism and/or performance.
131 SCHED_CORE is default disabled. When it is enabled and unused,
132 which is the likely usage by Linux distributions, there should
133 be no measurable impact on performance.