xref: /openbmc/linux/kernel/rcu/Kconfig (revision e40573a4)
1# SPDX-License-Identifier: GPL-2.0-only
2#
3# RCU-related configuration options
4#
5
6menu "RCU Subsystem"
7
8config TREE_RCU
9	bool
10	default y if SMP
11	# Dynticks-idle tracking
12	select CONTEXT_TRACKING_IDLE
13	help
14	  This option selects the RCU implementation that is
15	  designed for very large SMP system with hundreds or
16	  thousands of CPUs.  It also scales down nicely to
17	  smaller systems.
18
19config PREEMPT_RCU
20	bool
21	default y if PREEMPTION
22	select TREE_RCU
23	help
24	  This option selects the RCU implementation that is
25	  designed for very large SMP systems with hundreds or
26	  thousands of CPUs, but for which real-time response
27	  is also required.  It also scales down nicely to
28	  smaller systems.
29
30	  Select this option if you are unsure.
31
32config TINY_RCU
33	bool
34	default y if !PREEMPTION && !SMP
35	help
36	  This option selects the RCU implementation that is
37	  designed for UP systems from which real-time response
38	  is not required.  This option greatly reduces the
39	  memory footprint of RCU.
40
41config RCU_EXPERT
42	bool "Make expert-level adjustments to RCU configuration"
43	default n
44	help
45	  This option needs to be enabled if you wish to make
46	  expert-level adjustments to RCU configuration.  By default,
47	  no such adjustments can be made, which has the often-beneficial
48	  side-effect of preventing "make oldconfig" from asking you all
49	  sorts of detailed questions about how you would like numerous
50	  obscure RCU options to be set up.
51
52	  Say Y if you need to make expert-level adjustments to RCU.
53
54	  Say N if you are unsure.
55
56config SRCU
57	bool
58	help
59	  This option selects the sleepable version of RCU. This version
60	  permits arbitrary sleeping or blocking within RCU read-side critical
61	  sections.
62
63config TINY_SRCU
64	bool
65	default y if SRCU && TINY_RCU
66	help
67	  This option selects the single-CPU non-preemptible version of SRCU.
68
69config TREE_SRCU
70	bool
71	default y if SRCU && !TINY_RCU
72	help
73	  This option selects the full-fledged version of SRCU.
74
75config TASKS_RCU_GENERIC
76	def_bool TASKS_RCU || TASKS_RUDE_RCU || TASKS_TRACE_RCU
77	select SRCU
78	help
79	  This option enables generic infrastructure code supporting
80	  task-based RCU implementations.  Not for manual selection.
81
82config FORCE_TASKS_RCU
83	bool "Force selection of TASKS_RCU"
84	depends on RCU_EXPERT
85	select TASKS_RCU
86	default n
87	help
88	  This option force-enables a task-based RCU implementation
89	  that uses only voluntary context switch (not preemption!),
90	  idle, and user-mode execution as quiescent states.  Not for
91	  manual selection in most cases.
92
93config TASKS_RCU
94	bool
95	default n
96	select IRQ_WORK
97
98config FORCE_TASKS_RUDE_RCU
99	bool "Force selection of Tasks Rude RCU"
100	depends on RCU_EXPERT
101	select TASKS_RUDE_RCU
102	default n
103	help
104	  This option force-enables a task-based RCU implementation
105	  that uses only context switch (including preemption) and
106	  user-mode execution as quiescent states.  It forces IPIs and
107	  context switches on all online CPUs, including idle ones,
108	  so use with caution.	Not for manual selection in most cases.
109
110config TASKS_RUDE_RCU
111	bool
112	default n
113	select IRQ_WORK
114
115config FORCE_TASKS_TRACE_RCU
116	bool "Force selection of Tasks Trace RCU"
117	depends on RCU_EXPERT
118	select TASKS_TRACE_RCU
119	default n
120	help
121	  This option enables a task-based RCU implementation that uses
122	  explicit rcu_read_lock_trace() read-side markers, and allows
123	  these readers to appear in the idle loop as well as on the
124	  CPU hotplug code paths.  It can force IPIs on online CPUs,
125	  including idle ones, so use with caution.  Not for manual
126	  selection in most cases.
127
128config TASKS_TRACE_RCU
129	bool
130	default n
131	select IRQ_WORK
132
133config RCU_STALL_COMMON
134	def_bool TREE_RCU
135	help
136	  This option enables RCU CPU stall code that is common between
137	  the TINY and TREE variants of RCU.  The purpose is to allow
138	  the tiny variants to disable RCU CPU stall warnings, while
139	  making these warnings mandatory for the tree variants.
140
141config RCU_NEED_SEGCBLIST
142	def_bool ( TREE_RCU || TREE_SRCU || TASKS_RCU_GENERIC )
143
144config RCU_FANOUT
145	int "Tree-based hierarchical RCU fanout value"
146	range 2 64 if 64BIT
147	range 2 32 if !64BIT
148	depends on TREE_RCU && RCU_EXPERT
149	default 64 if 64BIT
150	default 32 if !64BIT
151	help
152	  This option controls the fanout of hierarchical implementations
153	  of RCU, allowing RCU to work efficiently on machines with
154	  large numbers of CPUs.  This value must be at least the fourth
155	  root of NR_CPUS, which allows NR_CPUS to be insanely large.
156	  The default value of RCU_FANOUT should be used for production
157	  systems, but if you are stress-testing the RCU implementation
158	  itself, small RCU_FANOUT values allow you to test large-system
159	  code paths on small(er) systems.
160
161	  Select a specific number if testing RCU itself.
162	  Take the default if unsure.
163
164config RCU_FANOUT_LEAF
165	int "Tree-based hierarchical RCU leaf-level fanout value"
166	range 2 64 if 64BIT && !RCU_STRICT_GRACE_PERIOD
167	range 2 32 if !64BIT && !RCU_STRICT_GRACE_PERIOD
168	range 2 3 if RCU_STRICT_GRACE_PERIOD
169	depends on TREE_RCU && RCU_EXPERT
170	default 16 if !RCU_STRICT_GRACE_PERIOD
171	default 2 if RCU_STRICT_GRACE_PERIOD
172	help
173	  This option controls the leaf-level fanout of hierarchical
174	  implementations of RCU, and allows trading off cache misses
175	  against lock contention.  Systems that synchronize their
176	  scheduling-clock interrupts for energy-efficiency reasons will
177	  want the default because the smaller leaf-level fanout keeps
178	  lock contention levels acceptably low.  Very large systems
179	  (hundreds or thousands of CPUs) will instead want to set this
180	  value to the maximum value possible in order to reduce the
181	  number of cache misses incurred during RCU's grace-period
182	  initialization.  These systems tend to run CPU-bound, and thus
183	  are not helped by synchronized interrupts, and thus tend to
184	  skew them, which reduces lock contention enough that large
185	  leaf-level fanouts work well.  That said, setting leaf-level
186	  fanout to a large number will likely cause problematic
187	  lock contention on the leaf-level rcu_node structures unless
188	  you boot with the skew_tick kernel parameter.
189
190	  Select a specific number if testing RCU itself.
191
192	  Select the maximum permissible value for large systems, but
193	  please understand that you may also need to set the skew_tick
194	  kernel boot parameter to avoid contention on the rcu_node
195	  structure's locks.
196
197	  Take the default if unsure.
198
199config RCU_BOOST
200	bool "Enable RCU priority boosting"
201	depends on (RT_MUTEXES && PREEMPT_RCU && RCU_EXPERT) || PREEMPT_RT
202	default y if PREEMPT_RT
203	help
204	  This option boosts the priority of preempted RCU readers that
205	  block the current preemptible RCU grace period for too long.
206	  This option also prevents heavy loads from blocking RCU
207	  callback invocation.
208
209	  Say Y here if you are working with real-time apps or heavy loads
210	  Say N here if you are unsure.
211
212config RCU_BOOST_DELAY
213	int "Milliseconds to delay boosting after RCU grace-period start"
214	range 0 3000
215	depends on RCU_BOOST
216	default 500
217	help
218	  This option specifies the time to wait after the beginning of
219	  a given grace period before priority-boosting preempted RCU
220	  readers blocking that grace period.  Note that any RCU reader
221	  blocking an expedited RCU grace period is boosted immediately.
222
223	  Accept the default if unsure.
224
225config RCU_EXP_KTHREAD
226	bool "Perform RCU expedited work in a real-time kthread"
227	depends on RCU_BOOST && RCU_EXPERT
228	default !PREEMPT_RT && NR_CPUS <= 32
229	help
230	  Use this option to further reduce the latencies of expedited
231	  grace periods at the expense of being more disruptive.
232
233	  This option is disabled by default on PREEMPT_RT=y kernels which
234	  disable expedited grace periods after boot by unconditionally
235	  setting rcupdate.rcu_normal_after_boot=1.
236
237	  Accept the default if unsure.
238
239config RCU_NOCB_CPU
240	bool "Offload RCU callback processing from boot-selected CPUs"
241	depends on TREE_RCU
242	depends on RCU_EXPERT || NO_HZ_FULL
243	default n
244	help
245	  Use this option to reduce OS jitter for aggressive HPC or
246	  real-time workloads.	It can also be used to offload RCU
247	  callback invocation to energy-efficient CPUs in battery-powered
248	  asymmetric multiprocessors.  The price of this reduced jitter
249	  is that the overhead of call_rcu() increases and that some
250	  workloads will incur significant increases in context-switch
251	  rates.
252
253	  This option offloads callback invocation from the set of CPUs
254	  specified at boot time by the rcu_nocbs parameter.  For each
255	  such CPU, a kthread ("rcuox/N") will be created to invoke
256	  callbacks, where the "N" is the CPU being offloaded, and where
257	  the "x" is "p" for RCU-preempt (PREEMPTION kernels) and "s" for
258	  RCU-sched (!PREEMPTION kernels).  Nothing prevents this kthread
259	  from running on the specified CPUs, but (1) the kthreads may be
260	  preempted between each callback, and (2) affinity or cgroups can
261	  be used to force the kthreads to run on whatever set of CPUs is
262	  desired.
263
264	  Say Y here if you need reduced OS jitter, despite added overhead.
265	  Say N here if you are unsure.
266
267config RCU_NOCB_CPU_DEFAULT_ALL
268	bool "Offload RCU callback processing from all CPUs by default"
269	depends on RCU_NOCB_CPU
270	default n
271	help
272	  Use this option to offload callback processing from all CPUs
273	  by default, in the absence of the rcu_nocbs or nohz_full boot
274	  parameter. This also avoids the need to use any boot parameters
275	  to achieve the effect of offloading all CPUs on boot.
276
277	  Say Y here if you want offload all CPUs by default on boot.
278	  Say N here if you are unsure.
279
280config RCU_NOCB_CPU_CB_BOOST
281	bool "Offload RCU callback from real-time kthread"
282	depends on RCU_NOCB_CPU && RCU_BOOST
283	default y if PREEMPT_RT
284	help
285	  Use this option to invoke offloaded callbacks as SCHED_FIFO
286	  to avoid starvation by heavy SCHED_OTHER background load.
287	  Of course, running as SCHED_FIFO during callback floods will
288	  cause the rcuo[ps] kthreads to monopolize the CPU for hundreds
289	  of milliseconds or more.  Therefore, when enabling this option,
290	  it is your responsibility to ensure that latency-sensitive
291	  tasks either run with higher priority or run on some other CPU.
292
293	  Say Y here if you want to set RT priority for offloading kthreads.
294	  Say N here if you are building a !PREEMPT_RT kernel and are unsure.
295
296config TASKS_TRACE_RCU_READ_MB
297	bool "Tasks Trace RCU readers use memory barriers in user and idle"
298	depends on RCU_EXPERT && TASKS_TRACE_RCU
299	default PREEMPT_RT || NR_CPUS < 8
300	help
301	  Use this option to further reduce the number of IPIs sent
302	  to CPUs executing in userspace or idle during tasks trace
303	  RCU grace periods.  Given that a reasonable setting of
304	  the rcupdate.rcu_task_ipi_delay kernel boot parameter
305	  eliminates such IPIs for many workloads, proper setting
306	  of this Kconfig option is important mostly for aggressive
307	  real-time installations and for battery-powered devices,
308	  hence the default chosen above.
309
310	  Say Y here if you hate IPIs.
311	  Say N here if you hate read-side memory barriers.
312	  Take the default if you are unsure.
313
314endmenu # "RCU Subsystem"
315