1 # SPDX-License-Identifier: GPL-2.0-only 1 # SPDX-License-Identifier: GPL-2.0-only
2 # 2 #
3 # RCU-related configuration options 3 # RCU-related configuration options
4 # 4 #
5 5
6 menu "RCU Subsystem" 6 menu "RCU Subsystem"
7 7
8 config TREE_RCU 8 config TREE_RCU
9 bool 9 bool
10 default y if SMP 10 default y if SMP
11 # Dynticks-idle tracking <<
12 select CONTEXT_TRACKING_IDLE <<
13 help 11 help
14 This option selects the RCU implemen 12 This option selects the RCU implementation that is
15 designed for very large SMP system w 13 designed for very large SMP system with hundreds or
16 thousands of CPUs. It also scales d 14 thousands of CPUs. It also scales down nicely to
17 smaller systems. 15 smaller systems.
18 16
19 config PREEMPT_RCU 17 config PREEMPT_RCU
20 bool 18 bool
21 default y if PREEMPTION 19 default y if PREEMPTION
22 select TREE_RCU 20 select TREE_RCU
23 help 21 help
24 This option selects the RCU implemen 22 This option selects the RCU implementation that is
25 designed for very large SMP systems 23 designed for very large SMP systems with hundreds or
26 thousands of CPUs, but for which rea 24 thousands of CPUs, but for which real-time response
27 is also required. It also scales do 25 is also required. It also scales down nicely to
28 smaller systems. 26 smaller systems.
29 27
30 Select this option if you are unsure 28 Select this option if you are unsure.
31 29
32 config TINY_RCU 30 config TINY_RCU
33 bool 31 bool
34 default y if !PREEMPT_RCU && !SMP !! 32 default y if !PREEMPTION && !SMP
35 help 33 help
36 This option selects the RCU implemen 34 This option selects the RCU implementation that is
37 designed for UP systems from which r 35 designed for UP systems from which real-time response
38 is not required. This option greatl 36 is not required. This option greatly reduces the
39 memory footprint of RCU. 37 memory footprint of RCU.
40 38
41 config RCU_EXPERT 39 config RCU_EXPERT
42 bool "Make expert-level adjustments to 40 bool "Make expert-level adjustments to RCU configuration"
43 default n 41 default n
44 help 42 help
45 This option needs to be enabled if y 43 This option needs to be enabled if you wish to make
46 expert-level adjustments to RCU conf 44 expert-level adjustments to RCU configuration. By default,
47 no such adjustments can be made, whi 45 no such adjustments can be made, which has the often-beneficial
48 side-effect of preventing "make oldc 46 side-effect of preventing "make oldconfig" from asking you all
49 sorts of detailed questions about ho 47 sorts of detailed questions about how you would like numerous
50 obscure RCU options to be set up. 48 obscure RCU options to be set up.
51 49
52 Say Y if you need to make expert-lev 50 Say Y if you need to make expert-level adjustments to RCU.
53 51
54 Say N if you are unsure. 52 Say N if you are unsure.
55 53
>> 54 config SRCU
>> 55 bool
>> 56 help
>> 57 This option selects the sleepable version of RCU. This version
>> 58 permits arbitrary sleeping or blocking within RCU read-side critical
>> 59 sections.
>> 60
56 config TINY_SRCU 61 config TINY_SRCU
57 bool 62 bool
58 default y if TINY_RCU !! 63 default y if SRCU && TINY_RCU
59 help 64 help
60 This option selects the single-CPU n 65 This option selects the single-CPU non-preemptible version of SRCU.
61 66
62 config TREE_SRCU 67 config TREE_SRCU
63 bool 68 bool
64 default y if !TINY_RCU !! 69 default y if SRCU && !TINY_RCU
65 help 70 help
66 This option selects the full-fledged 71 This option selects the full-fledged version of SRCU.
67 72
68 config NEED_SRCU_NMI_SAFE <<
69 def_bool HAVE_NMI && !ARCH_HAS_NMI_SAF <<
70 <<
71 config TASKS_RCU_GENERIC 73 config TASKS_RCU_GENERIC
72 def_bool TASKS_RCU || TASKS_RUDE_RCU | 74 def_bool TASKS_RCU || TASKS_RUDE_RCU || TASKS_TRACE_RCU
>> 75 select SRCU
73 help 76 help
74 This option enables generic infrastr 77 This option enables generic infrastructure code supporting
75 task-based RCU implementations. Not 78 task-based RCU implementations. Not for manual selection.
76 79
77 config 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-bas <<
84 that uses only voluntary context swi <<
85 idle, and user-mode execution as qui <<
86 manual selection in most cases. <<
87 <<
88 config NEED_TASKS_RCU <<
89 bool <<
90 default n <<
91 <<
92 config TASKS_RCU 80 config TASKS_RCU
93 bool !! 81 def_bool PREEMPTION
94 default NEED_TASKS_RCU && (PREEMPTION <<
95 select IRQ_WORK <<
96 <<
97 config FORCE_TASKS_RUDE_RCU <<
98 bool "Force selection of Tasks Rude RC <<
99 depends on RCU_EXPERT <<
100 select TASKS_RUDE_RCU <<
101 default n <<
102 help 82 help
103 This option force-enables a task-bas !! 83 This option enables a task-based RCU implementation that uses
104 that uses only context switch (inclu !! 84 only voluntary context switch (not preemption!), idle, and
105 user-mode execution as quiescent sta !! 85 user-mode execution as quiescent states. Not for manual selection.
106 context switches on all online CPUs, <<
107 so use with caution. Not for manual <<
108 86
109 config TASKS_RUDE_RCU 87 config TASKS_RUDE_RCU
110 bool !! 88 def_bool 0
111 default n <<
112 select IRQ_WORK <<
113 <<
114 config FORCE_TASKS_TRACE_RCU <<
115 bool "Force selection of Tasks Trace R <<
116 depends on RCU_EXPERT <<
117 select TASKS_TRACE_RCU <<
118 default n <<
119 help 89 help
120 This option enables a task-based RCU 90 This option enables a task-based RCU implementation that uses
121 explicit rcu_read_lock_trace() read- !! 91 only context switch (including preemption) and user-mode
122 these readers to appear in the idle !! 92 execution as quiescent states. It forces IPIs and context
123 CPU hotplug code paths. It can forc !! 93 switches on all online CPUs, including idle ones, so use
124 including idle ones, so use with cau !! 94 with caution.
125 selection in most cases. <<
126 95
127 config TASKS_TRACE_RCU 96 config TASKS_TRACE_RCU
128 bool !! 97 def_bool 0
129 default n !! 98 help
130 select IRQ_WORK !! 99 This option enables a task-based RCU implementation that uses
>> 100 explicit rcu_read_lock_trace() read-side markers, and allows
>> 101 these readers to appear in the idle loop as well as on the CPU
>> 102 hotplug code paths. It can force IPIs on online CPUs, including
>> 103 idle ones, so use with caution.
131 104
132 config RCU_STALL_COMMON 105 config RCU_STALL_COMMON
133 def_bool TREE_RCU 106 def_bool TREE_RCU
134 help 107 help
135 This option enables RCU CPU stall co 108 This option enables RCU CPU stall code that is common between
136 the TINY and TREE variants of RCU. 109 the TINY and TREE variants of RCU. The purpose is to allow
137 the tiny variants to disable RCU CPU 110 the tiny variants to disable RCU CPU stall warnings, while
138 making these warnings mandatory for 111 making these warnings mandatory for the tree variants.
139 112
140 config RCU_NEED_SEGCBLIST 113 config RCU_NEED_SEGCBLIST
141 def_bool ( TREE_RCU || TREE_SRCU || TA !! 114 def_bool ( TREE_RCU || TREE_SRCU )
142 115
143 config RCU_FANOUT 116 config RCU_FANOUT
144 int "Tree-based hierarchical RCU fanou 117 int "Tree-based hierarchical RCU fanout value"
145 range 2 64 if 64BIT 118 range 2 64 if 64BIT
146 range 2 32 if !64BIT 119 range 2 32 if !64BIT
147 depends on TREE_RCU && RCU_EXPERT 120 depends on TREE_RCU && RCU_EXPERT
148 default 64 if 64BIT 121 default 64 if 64BIT
149 default 32 if !64BIT 122 default 32 if !64BIT
150 help 123 help
151 This option controls the fanout of h 124 This option controls the fanout of hierarchical implementations
152 of RCU, allowing RCU to work efficie 125 of RCU, allowing RCU to work efficiently on machines with
153 large numbers of CPUs. This value m 126 large numbers of CPUs. This value must be at least the fourth
154 root of NR_CPUS, which allows NR_CPU 127 root of NR_CPUS, which allows NR_CPUS to be insanely large.
155 The default value of RCU_FANOUT shou 128 The default value of RCU_FANOUT should be used for production
156 systems, but if you are stress-testi 129 systems, but if you are stress-testing the RCU implementation
157 itself, small RCU_FANOUT values allo 130 itself, small RCU_FANOUT values allow you to test large-system
158 code paths on small(er) systems. 131 code paths on small(er) systems.
159 132
160 Select a specific number if testing 133 Select a specific number if testing RCU itself.
161 Take the default if unsure. 134 Take the default if unsure.
162 135
163 config RCU_FANOUT_LEAF 136 config RCU_FANOUT_LEAF
164 int "Tree-based hierarchical RCU leaf- 137 int "Tree-based hierarchical RCU leaf-level fanout value"
165 range 2 64 if 64BIT && !RCU_STRICT_GRA 138 range 2 64 if 64BIT && !RCU_STRICT_GRACE_PERIOD
166 range 2 32 if !64BIT && !RCU_STRICT_GR 139 range 2 32 if !64BIT && !RCU_STRICT_GRACE_PERIOD
167 range 2 3 if RCU_STRICT_GRACE_PERIOD 140 range 2 3 if RCU_STRICT_GRACE_PERIOD
168 depends on TREE_RCU && RCU_EXPERT 141 depends on TREE_RCU && RCU_EXPERT
169 default 16 if !RCU_STRICT_GRACE_PERIOD 142 default 16 if !RCU_STRICT_GRACE_PERIOD
170 default 2 if RCU_STRICT_GRACE_PERIOD 143 default 2 if RCU_STRICT_GRACE_PERIOD
171 help 144 help
172 This option controls the leaf-level 145 This option controls the leaf-level fanout of hierarchical
173 implementations of RCU, and allows t 146 implementations of RCU, and allows trading off cache misses
174 against lock contention. Systems th 147 against lock contention. Systems that synchronize their
175 scheduling-clock interrupts for ener 148 scheduling-clock interrupts for energy-efficiency reasons will
176 want the default because the smaller 149 want the default because the smaller leaf-level fanout keeps
177 lock contention levels acceptably lo 150 lock contention levels acceptably low. Very large systems
178 (hundreds or thousands of CPUs) will 151 (hundreds or thousands of CPUs) will instead want to set this
179 value to the maximum value possible 152 value to the maximum value possible in order to reduce the
180 number of cache misses incurred duri 153 number of cache misses incurred during RCU's grace-period
181 initialization. These systems tend 154 initialization. These systems tend to run CPU-bound, and thus
182 are not helped by synchronized inter 155 are not helped by synchronized interrupts, and thus tend to
183 skew them, which reduces lock conten 156 skew them, which reduces lock contention enough that large
184 leaf-level fanouts work well. That 157 leaf-level fanouts work well. That said, setting leaf-level
185 fanout to a large number will likely 158 fanout to a large number will likely cause problematic
186 lock contention on the leaf-level rc 159 lock contention on the leaf-level rcu_node structures unless
187 you boot with the skew_tick kernel p 160 you boot with the skew_tick kernel parameter.
188 161
189 Select a specific number if testing 162 Select a specific number if testing RCU itself.
190 163
191 Select the maximum permissible value 164 Select the maximum permissible value for large systems, but
192 please understand that you may also 165 please understand that you may also need to set the skew_tick
193 kernel boot parameter to avoid conte 166 kernel boot parameter to avoid contention on the rcu_node
194 structure's locks. 167 structure's locks.
195 168
196 Take the default if unsure. 169 Take the default if unsure.
197 170
>> 171 config RCU_FAST_NO_HZ
>> 172 bool "Accelerate last non-dyntick-idle CPU's grace periods"
>> 173 depends on NO_HZ_COMMON && SMP && RCU_EXPERT
>> 174 default n
>> 175 help
>> 176 This option permits CPUs to enter dynticks-idle state even if
>> 177 they have RCU callbacks queued, and prevents RCU from waking
>> 178 these CPUs up more than roughly once every four jiffies (by
>> 179 default, you can adjust this using the rcutree.rcu_idle_gp_delay
>> 180 parameter), thus improving energy efficiency. On the other
>> 181 hand, this option increases the duration of RCU grace periods,
>> 182 for example, slowing down synchronize_rcu().
>> 183
>> 184 Say Y if energy efficiency is critically important, and you
>> 185 don't care about increased grace-period durations.
>> 186
>> 187 Say N if you are unsure.
>> 188
198 config RCU_BOOST 189 config RCU_BOOST
199 bool "Enable RCU priority boosting" 190 bool "Enable RCU priority boosting"
200 depends on (RT_MUTEXES && PREEMPT_RCU !! 191 depends on RT_MUTEXES && PREEMPT_RCU && RCU_EXPERT
201 default y if PREEMPT_RT !! 192 default n
202 help 193 help
203 This option boosts the priority of p 194 This option boosts the priority of preempted RCU readers that
204 block the current preemptible RCU gr 195 block the current preemptible RCU grace period for too long.
205 This option also prevents heavy load 196 This option also prevents heavy loads from blocking RCU
206 callback invocation. 197 callback invocation.
207 198
208 Say Y here if you are working with r 199 Say Y here if you are working with real-time apps or heavy loads
209 Say N here if you are unsure. 200 Say N here if you are unsure.
210 201
211 config RCU_BOOST_DELAY 202 config RCU_BOOST_DELAY
212 int "Milliseconds to delay boosting af 203 int "Milliseconds to delay boosting after RCU grace-period start"
213 range 0 3000 204 range 0 3000
214 depends on RCU_BOOST 205 depends on RCU_BOOST
215 default 500 206 default 500
216 help 207 help
217 This option specifies the time to wa 208 This option specifies the time to wait after the beginning of
218 a given grace period before priority 209 a given grace period before priority-boosting preempted RCU
219 readers blocking that grace period. 210 readers blocking that grace period. Note that any RCU reader
220 blocking an expedited RCU grace peri 211 blocking an expedited RCU grace period is boosted immediately.
221 212
222 Accept the default if unsure. 213 Accept the default if unsure.
223 214
224 config RCU_EXP_KTHREAD <<
225 bool "Perform RCU expedited work in a <<
226 depends on RCU_BOOST && RCU_EXPERT <<
227 default !PREEMPT_RT && NR_CPUS <= 32 <<
228 help <<
229 Use this option to further reduce th <<
230 grace periods at the expense of bein <<
231 <<
232 This option is disabled by default o <<
233 disable expedited grace periods afte <<
234 setting rcupdate.rcu_normal_after_bo <<
235 <<
236 Accept the default if unsure. <<
237 <<
238 config RCU_NOCB_CPU 215 config RCU_NOCB_CPU
239 bool "Offload RCU callback processing 216 bool "Offload RCU callback processing from boot-selected CPUs"
240 depends on TREE_RCU 217 depends on TREE_RCU
241 depends on RCU_EXPERT || NO_HZ_FULL 218 depends on RCU_EXPERT || NO_HZ_FULL
242 default n 219 default n
243 help 220 help
244 Use this option to reduce OS jitter 221 Use this option to reduce OS jitter for aggressive HPC or
245 real-time workloads. It can also be 222 real-time workloads. It can also be used to offload RCU
246 callback invocation to energy-effici 223 callback invocation to energy-efficient CPUs in battery-powered
247 asymmetric multiprocessors. The pri 224 asymmetric multiprocessors. The price of this reduced jitter
248 is that the overhead of call_rcu() i 225 is that the overhead of call_rcu() increases and that some
249 workloads will incur significant inc 226 workloads will incur significant increases in context-switch
250 rates. 227 rates.
251 228
252 This option offloads callback invoca 229 This option offloads callback invocation from the set of CPUs
253 specified at boot time by the rcu_no 230 specified at boot time by the rcu_nocbs parameter. For each
254 such CPU, a kthread ("rcuox/N") will 231 such CPU, a kthread ("rcuox/N") will be created to invoke
255 callbacks, where the "N" is the CPU 232 callbacks, where the "N" is the CPU being offloaded, and where
256 the "x" is "p" for RCU-preempt (PREE 233 the "x" is "p" for RCU-preempt (PREEMPTION kernels) and "s" for
257 RCU-sched (!PREEMPTION kernels). No 234 RCU-sched (!PREEMPTION kernels). Nothing prevents this kthread
258 from running on the specified CPUs, 235 from running on the specified CPUs, but (1) the kthreads may be
259 preempted between each callback, and 236 preempted between each callback, and (2) affinity or cgroups can
260 be used to force the kthreads to run 237 be used to force the kthreads to run on whatever set of CPUs is
261 desired. 238 desired.
262 239
263 Say Y here if you need reduced OS ji 240 Say Y here if you need reduced OS jitter, despite added overhead.
264 Say N here if you are unsure. 241 Say N here if you are unsure.
265 242
266 config RCU_NOCB_CPU_DEFAULT_ALL <<
267 bool "Offload RCU callback processing <<
268 depends on RCU_NOCB_CPU <<
269 default n <<
270 help <<
271 Use this option to offload callback <<
272 by default, in the absence of the rc <<
273 parameter. This also avoids the need <<
274 to achieve the effect of offloading <<
275 <<
276 Say Y here if you want offload all C <<
277 Say N here if you are unsure. <<
278 <<
279 config RCU_NOCB_CPU_CB_BOOST <<
280 bool "Offload RCU callback from real-t <<
281 depends on RCU_NOCB_CPU && RCU_BOOST <<
282 default y if PREEMPT_RT <<
283 help <<
284 Use this option to invoke offloaded <<
285 to avoid starvation by heavy SCHED_O <<
286 Of course, running as SCHED_FIFO dur <<
287 cause the rcuo[ps] kthreads to monop <<
288 of milliseconds or more. Therefore, <<
289 it is your responsibility to ensure <<
290 tasks either run with higher priorit <<
291 <<
292 Say Y here if you want to set RT pri <<
293 Say N here if you are building a !PR <<
294 <<
295 config TASKS_TRACE_RCU_READ_MB 243 config TASKS_TRACE_RCU_READ_MB
296 bool "Tasks Trace RCU readers use memo 244 bool "Tasks Trace RCU readers use memory barriers in user and idle"
297 depends on RCU_EXPERT && TASKS_TRACE_R !! 245 depends on RCU_EXPERT
298 default PREEMPT_RT || NR_CPUS < 8 246 default PREEMPT_RT || NR_CPUS < 8
299 help 247 help
300 Use this option to further reduce th 248 Use this option to further reduce the number of IPIs sent
301 to CPUs executing in userspace or id 249 to CPUs executing in userspace or idle during tasks trace
302 RCU grace periods. Given that a rea 250 RCU grace periods. Given that a reasonable setting of
303 the rcupdate.rcu_task_ipi_delay kern 251 the rcupdate.rcu_task_ipi_delay kernel boot parameter
304 eliminates such IPIs for many worklo 252 eliminates such IPIs for many workloads, proper setting
305 of this Kconfig option is important 253 of this Kconfig option is important mostly for aggressive
306 real-time installations and for batt 254 real-time installations and for battery-powered devices,
307 hence the default chosen above. 255 hence the default chosen above.
308 256
309 Say Y here if you hate IPIs. 257 Say Y here if you hate IPIs.
310 Say N here if you hate read-side mem 258 Say N here if you hate read-side memory barriers.
311 Take the default if you are unsure. 259 Take the default if you are unsure.
312 <<
313 config RCU_LAZY <<
314 bool "RCU callback lazy invocation fun <<
315 depends on RCU_NOCB_CPU <<
316 default n <<
317 help <<
318 To save power, batch RCU callbacks a <<
319 pressure, or callback list growing t <<
320 <<
321 Requires rcu_nocbs=all to be set. <<
322 <<
323 Use rcutree.enable_rcu_lazy=0 to tur <<
324 <<
325 config RCU_LAZY_DEFAULT_OFF <<
326 bool "Turn RCU lazy invocation off by <<
327 depends on RCU_LAZY <<
328 default n <<
329 help <<
330 Allows building the kernel with CONF <<
331 off. Boot time param rcutree.enable_ <<
332 it back on. <<
333 <<
334 config RCU_DOUBLE_CHECK_CB_TIME <<
335 bool "RCU callback-batch backup time c <<
336 depends on RCU_EXPERT <<
337 default n <<
338 help <<
339 Use this option to provide more prec <<
340 rcutree.rcu_resched_ns module parame <<
341 a single RCU callback might run for <<
342 thus defeating the 32-callback batch <<
343 cost of the fine-grained but expensi <<
344 <<
345 This option rounds rcutree.rcu_resch <<
346 jiffy, and overrides the 32-callback <<
347 is exceeded. <<
348 <<
349 Say Y here if you need tighter callb <<
350 Say N here if you are unsure. <<
351 260
352 endmenu # "RCU Subsystem" 261 endmenu # "RCU Subsystem"
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.