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