xref: /openbmc/linux/kernel/rcu/Kconfig (revision ac3cb6de)
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 TINY_SRCU
57	bool
58	default y if TINY_RCU
59	help
60	  This option selects the single-CPU non-preemptible version of SRCU.
61
62config TREE_SRCU
63	bool
64	default y if !TINY_RCU
65	help
66	  This option selects the full-fledged version of SRCU.
67
68config NEED_SRCU_NMI_SAFE
69	def_bool HAVE_NMI && !ARCH_HAS_NMI_SAFE_THIS_CPU_OPS && !TINY_SRCU
70
71config TASKS_RCU_GENERIC
72	def_bool TASKS_RCU || TASKS_RUDE_RCU || TASKS_TRACE_RCU
73	help
74	  This option enables generic infrastructure code supporting
75	  task-based RCU implementations.  Not for manual selection.
76
77config FORCE_TASKS_RCU
78	bool "Force selection of TASKS_RCU"
79	depends on RCU_EXPERT
80	select TASKS_RCU
81	default n
82	help
83	  This option force-enables a task-based RCU implementation
84	  that uses only voluntary context switch (not preemption!),
85	  idle, and user-mode execution as quiescent states.  Not for
86	  manual selection in most cases.
87
88config TASKS_RCU
89	bool
90	default n
91	select IRQ_WORK
92
93config FORCE_TASKS_RUDE_RCU
94	bool "Force selection of Tasks Rude RCU"
95	depends on RCU_EXPERT
96	select TASKS_RUDE_RCU
97	default n
98	help
99	  This option force-enables a task-based RCU implementation
100	  that uses only context switch (including preemption) and
101	  user-mode execution as quiescent states.  It forces IPIs and
102	  context switches on all online CPUs, including idle ones,
103	  so use with caution.	Not for manual selection in most cases.
104
105config TASKS_RUDE_RCU
106	bool
107	default n
108	select IRQ_WORK
109
110config FORCE_TASKS_TRACE_RCU
111	bool "Force selection of Tasks Trace RCU"
112	depends on RCU_EXPERT
113	select TASKS_TRACE_RCU
114	default n
115	help
116	  This option enables a task-based RCU implementation that uses
117	  explicit rcu_read_lock_trace() read-side markers, and allows
118	  these readers to appear in the idle loop as well as on the
119	  CPU hotplug code paths.  It can force IPIs on online CPUs,
120	  including idle ones, so use with caution.  Not for manual
121	  selection in most cases.
122
123config TASKS_TRACE_RCU
124	bool
125	default n
126	select IRQ_WORK
127
128config RCU_STALL_COMMON
129	def_bool TREE_RCU
130	help
131	  This option enables RCU CPU stall code that is common between
132	  the TINY and TREE variants of RCU.  The purpose is to allow
133	  the tiny variants to disable RCU CPU stall warnings, while
134	  making these warnings mandatory for the tree variants.
135
136config RCU_NEED_SEGCBLIST
137	def_bool ( TREE_RCU || TREE_SRCU || TASKS_RCU_GENERIC )
138
139config RCU_FANOUT
140	int "Tree-based hierarchical RCU fanout value"
141	range 2 64 if 64BIT
142	range 2 32 if !64BIT
143	depends on TREE_RCU && RCU_EXPERT
144	default 64 if 64BIT
145	default 32 if !64BIT
146	help
147	  This option controls the fanout of hierarchical implementations
148	  of RCU, allowing RCU to work efficiently on machines with
149	  large numbers of CPUs.  This value must be at least the fourth
150	  root of NR_CPUS, which allows NR_CPUS to be insanely large.
151	  The default value of RCU_FANOUT should be used for production
152	  systems, but if you are stress-testing the RCU implementation
153	  itself, small RCU_FANOUT values allow you to test large-system
154	  code paths on small(er) systems.
155
156	  Select a specific number if testing RCU itself.
157	  Take the default if unsure.
158
159config RCU_FANOUT_LEAF
160	int "Tree-based hierarchical RCU leaf-level fanout value"
161	range 2 64 if 64BIT && !RCU_STRICT_GRACE_PERIOD
162	range 2 32 if !64BIT && !RCU_STRICT_GRACE_PERIOD
163	range 2 3 if RCU_STRICT_GRACE_PERIOD
164	depends on TREE_RCU && RCU_EXPERT
165	default 16 if !RCU_STRICT_GRACE_PERIOD
166	default 2 if RCU_STRICT_GRACE_PERIOD
167	help
168	  This option controls the leaf-level fanout of hierarchical
169	  implementations of RCU, and allows trading off cache misses
170	  against lock contention.  Systems that synchronize their
171	  scheduling-clock interrupts for energy-efficiency reasons will
172	  want the default because the smaller leaf-level fanout keeps
173	  lock contention levels acceptably low.  Very large systems
174	  (hundreds or thousands of CPUs) will instead want to set this
175	  value to the maximum value possible in order to reduce the
176	  number of cache misses incurred during RCU's grace-period
177	  initialization.  These systems tend to run CPU-bound, and thus
178	  are not helped by synchronized interrupts, and thus tend to
179	  skew them, which reduces lock contention enough that large
180	  leaf-level fanouts work well.  That said, setting leaf-level
181	  fanout to a large number will likely cause problematic
182	  lock contention on the leaf-level rcu_node structures unless
183	  you boot with the skew_tick kernel parameter.
184
185	  Select a specific number if testing RCU itself.
186
187	  Select the maximum permissible value for large systems, but
188	  please understand that you may also need to set the skew_tick
189	  kernel boot parameter to avoid contention on the rcu_node
190	  structure's locks.
191
192	  Take the default if unsure.
193
194config RCU_BOOST
195	bool "Enable RCU priority boosting"
196	depends on (RT_MUTEXES && PREEMPT_RCU && RCU_EXPERT) || PREEMPT_RT
197	default y if PREEMPT_RT
198	help
199	  This option boosts the priority of preempted RCU readers that
200	  block the current preemptible RCU grace period for too long.
201	  This option also prevents heavy loads from blocking RCU
202	  callback invocation.
203
204	  Say Y here if you are working with real-time apps or heavy loads
205	  Say N here if you are unsure.
206
207config RCU_BOOST_DELAY
208	int "Milliseconds to delay boosting after RCU grace-period start"
209	range 0 3000
210	depends on RCU_BOOST
211	default 500
212	help
213	  This option specifies the time to wait after the beginning of
214	  a given grace period before priority-boosting preempted RCU
215	  readers blocking that grace period.  Note that any RCU reader
216	  blocking an expedited RCU grace period is boosted immediately.
217
218	  Accept the default if unsure.
219
220config RCU_EXP_KTHREAD
221	bool "Perform RCU expedited work in a real-time kthread"
222	depends on RCU_BOOST && RCU_EXPERT
223	default !PREEMPT_RT && NR_CPUS <= 32
224	help
225	  Use this option to further reduce the latencies of expedited
226	  grace periods at the expense of being more disruptive.
227
228	  This option is disabled by default on PREEMPT_RT=y kernels which
229	  disable expedited grace periods after boot by unconditionally
230	  setting rcupdate.rcu_normal_after_boot=1.
231
232	  Accept the default if unsure.
233
234config RCU_NOCB_CPU
235	bool "Offload RCU callback processing from boot-selected CPUs"
236	depends on TREE_RCU
237	depends on RCU_EXPERT || NO_HZ_FULL
238	default n
239	help
240	  Use this option to reduce OS jitter for aggressive HPC or
241	  real-time workloads.	It can also be used to offload RCU
242	  callback invocation to energy-efficient CPUs in battery-powered
243	  asymmetric multiprocessors.  The price of this reduced jitter
244	  is that the overhead of call_rcu() increases and that some
245	  workloads will incur significant increases in context-switch
246	  rates.
247
248	  This option offloads callback invocation from the set of CPUs
249	  specified at boot time by the rcu_nocbs parameter.  For each
250	  such CPU, a kthread ("rcuox/N") will be created to invoke
251	  callbacks, where the "N" is the CPU being offloaded, and where
252	  the "x" is "p" for RCU-preempt (PREEMPTION kernels) and "s" for
253	  RCU-sched (!PREEMPTION kernels).  Nothing prevents this kthread
254	  from running on the specified CPUs, but (1) the kthreads may be
255	  preempted between each callback, and (2) affinity or cgroups can
256	  be used to force the kthreads to run on whatever set of CPUs is
257	  desired.
258
259	  Say Y here if you need reduced OS jitter, despite added overhead.
260	  Say N here if you are unsure.
261
262config RCU_NOCB_CPU_DEFAULT_ALL
263	bool "Offload RCU callback processing from all CPUs by default"
264	depends on RCU_NOCB_CPU
265	default n
266	help
267	  Use this option to offload callback processing from all CPUs
268	  by default, in the absence of the rcu_nocbs or nohz_full boot
269	  parameter. This also avoids the need to use any boot parameters
270	  to achieve the effect of offloading all CPUs on boot.
271
272	  Say Y here if you want offload all CPUs by default on boot.
273	  Say N here if you are unsure.
274
275config RCU_NOCB_CPU_CB_BOOST
276	bool "Offload RCU callback from real-time kthread"
277	depends on RCU_NOCB_CPU && RCU_BOOST
278	default y if PREEMPT_RT
279	help
280	  Use this option to invoke offloaded callbacks as SCHED_FIFO
281	  to avoid starvation by heavy SCHED_OTHER background load.
282	  Of course, running as SCHED_FIFO during callback floods will
283	  cause the rcuo[ps] kthreads to monopolize the CPU for hundreds
284	  of milliseconds or more.  Therefore, when enabling this option,
285	  it is your responsibility to ensure that latency-sensitive
286	  tasks either run with higher priority or run on some other CPU.
287
288	  Say Y here if you want to set RT priority for offloading kthreads.
289	  Say N here if you are building a !PREEMPT_RT kernel and are unsure.
290
291config TASKS_TRACE_RCU_READ_MB
292	bool "Tasks Trace RCU readers use memory barriers in user and idle"
293	depends on RCU_EXPERT && TASKS_TRACE_RCU
294	default PREEMPT_RT || NR_CPUS < 8
295	help
296	  Use this option to further reduce the number of IPIs sent
297	  to CPUs executing in userspace or idle during tasks trace
298	  RCU grace periods.  Given that a reasonable setting of
299	  the rcupdate.rcu_task_ipi_delay kernel boot parameter
300	  eliminates such IPIs for many workloads, proper setting
301	  of this Kconfig option is important mostly for aggressive
302	  real-time installations and for battery-powered devices,
303	  hence the default chosen above.
304
305	  Say Y here if you hate IPIs.
306	  Say N here if you hate read-side memory barriers.
307	  Take the default if you are unsure.
308
309config RCU_LAZY
310	bool "RCU callback lazy invocation functionality"
311	depends on RCU_NOCB_CPU
312	default n
313	help
314	  To save power, batch RCU callbacks and flush after delay, memory
315	  pressure, or callback list growing too big.
316
317config RCU_DOUBLE_CHECK_CB_TIME
318	bool "RCU callback-batch backup time check"
319	depends on RCU_EXPERT
320	default n
321	help
322	  Use this option to provide more precise enforcement of the
323	  rcutree.rcu_resched_ns module parameter in situations where
324	  a single RCU callback might run for hundreds of microseconds,
325	  thus defeating the 32-callback batching used to amortize the
326	  cost of the fine-grained but expensive local_clock() function.
327
328	  This option rounds rcutree.rcu_resched_ns up to the next
329	  jiffy, and overrides the 32-callback batching if this limit
330	  is exceeded.
331
332	  Say Y here if you need tighter callback-limit enforcement.
333	  Say N here if you are unsure.
334
335endmenu # "RCU Subsystem"
336