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