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