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 help 12 This option selects the RCU implementation that is 13 designed for very large SMP system with hundreds or 14 thousands of CPUs. It also scales down nicely to 15 smaller systems. 16 17config PREEMPT_RCU 18 bool 19 default y if PREEMPTION 20 select TREE_RCU 21 help 22 This option selects the RCU implementation that is 23 designed for very large SMP systems with hundreds or 24 thousands of CPUs, but for which real-time response 25 is also required. It also scales down nicely to 26 smaller systems. 27 28 Select this option if you are unsure. 29 30config TINY_RCU 31 bool 32 default y if !PREEMPTION && !SMP 33 help 34 This option selects the RCU implementation that is 35 designed for UP systems from which real-time response 36 is not required. This option greatly reduces the 37 memory footprint of RCU. 38 39config RCU_EXPERT 40 bool "Make expert-level adjustments to RCU configuration" 41 default n 42 help 43 This option needs to be enabled if you wish to make 44 expert-level adjustments to RCU configuration. By default, 45 no such adjustments can be made, which has the often-beneficial 46 side-effect of preventing "make oldconfig" from asking you all 47 sorts of detailed questions about how you would like numerous 48 obscure RCU options to be set up. 49 50 Say Y if you need to make expert-level adjustments to RCU. 51 52 Say N if you are unsure. 53 54config 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 61config TINY_SRCU 62 bool 63 default y if SRCU && TINY_RCU 64 help 65 This option selects the single-CPU non-preemptible version of SRCU. 66 67config TREE_SRCU 68 bool 69 default y if SRCU && !TINY_RCU 70 help 71 This option selects the full-fledged version of SRCU. 72 73config TASKS_RCU_GENERIC 74 def_bool TASKS_RCU || TASKS_RUDE_RCU || TASKS_TRACE_RCU 75 select SRCU 76 help 77 This option enables generic infrastructure code supporting 78 task-based RCU implementations. Not for manual selection. 79 80config TASKS_RCU 81 def_bool PREEMPTION 82 help 83 This option enables a task-based RCU implementation that uses 84 only voluntary context switch (not preemption!), idle, and 85 user-mode execution as quiescent states. Not for manual selection. 86 87config TASKS_RUDE_RCU 88 def_bool 0 89 select IRQ_WORK 90 help 91 This option enables a task-based RCU implementation that uses 92 only context switch (including preemption) and user-mode 93 execution as quiescent states. It forces IPIs and context 94 switches on all online CPUs, including idle ones, so use 95 with caution. 96 97config TASKS_TRACE_RCU 98 def_bool 0 99 select IRQ_WORK 100 help 101 This option enables a task-based RCU implementation that uses 102 explicit rcu_read_lock_trace() read-side markers, and allows 103 these readers to appear in the idle loop as well as on the CPU 104 hotplug code paths. It can force IPIs on online CPUs, including 105 idle ones, so use with caution. 106 107config RCU_STALL_COMMON 108 def_bool TREE_RCU 109 help 110 This option enables RCU CPU stall code that is common between 111 the TINY and TREE variants of RCU. The purpose is to allow 112 the tiny variants to disable RCU CPU stall warnings, while 113 making these warnings mandatory for the tree variants. 114 115config RCU_NEED_SEGCBLIST 116 def_bool ( TREE_RCU || TREE_SRCU ) 117 118config RCU_FANOUT 119 int "Tree-based hierarchical RCU fanout value" 120 range 2 64 if 64BIT 121 range 2 32 if !64BIT 122 depends on TREE_RCU && RCU_EXPERT 123 default 64 if 64BIT 124 default 32 if !64BIT 125 help 126 This option controls the fanout of hierarchical implementations 127 of RCU, allowing RCU to work efficiently on machines with 128 large numbers of CPUs. This value must be at least the fourth 129 root of NR_CPUS, which allows NR_CPUS to be insanely large. 130 The default value of RCU_FANOUT should be used for production 131 systems, but if you are stress-testing the RCU implementation 132 itself, small RCU_FANOUT values allow you to test large-system 133 code paths on small(er) systems. 134 135 Select a specific number if testing RCU itself. 136 Take the default if unsure. 137 138config RCU_FANOUT_LEAF 139 int "Tree-based hierarchical RCU leaf-level fanout value" 140 range 2 64 if 64BIT && !RCU_STRICT_GRACE_PERIOD 141 range 2 32 if !64BIT && !RCU_STRICT_GRACE_PERIOD 142 range 2 3 if RCU_STRICT_GRACE_PERIOD 143 depends on TREE_RCU && RCU_EXPERT 144 default 16 if !RCU_STRICT_GRACE_PERIOD 145 default 2 if RCU_STRICT_GRACE_PERIOD 146 help 147 This option controls the leaf-level fanout of hierarchical 148 implementations of RCU, and allows trading off cache misses 149 against lock contention. Systems that synchronize their 150 scheduling-clock interrupts for energy-efficiency reasons will 151 want the default because the smaller leaf-level fanout keeps 152 lock contention levels acceptably low. Very large systems 153 (hundreds or thousands of CPUs) will instead want to set this 154 value to the maximum value possible in order to reduce the 155 number of cache misses incurred during RCU's grace-period 156 initialization. These systems tend to run CPU-bound, and thus 157 are not helped by synchronized interrupts, and thus tend to 158 skew them, which reduces lock contention enough that large 159 leaf-level fanouts work well. That said, setting leaf-level 160 fanout to a large number will likely cause problematic 161 lock contention on the leaf-level rcu_node structures unless 162 you boot with the skew_tick kernel parameter. 163 164 Select a specific number if testing RCU itself. 165 166 Select the maximum permissible value for large systems, but 167 please understand that you may also need to set the skew_tick 168 kernel boot parameter to avoid contention on the rcu_node 169 structure's locks. 170 171 Take the default if unsure. 172 173config RCU_FAST_NO_HZ 174 bool "Accelerate last non-dyntick-idle CPU's grace periods" 175 depends on NO_HZ_COMMON && SMP && RCU_EXPERT 176 default n 177 help 178 This option permits CPUs to enter dynticks-idle state even if 179 they have RCU callbacks queued, and prevents RCU from waking 180 these CPUs up more than roughly once every four jiffies (by 181 default, you can adjust this using the rcutree.rcu_idle_gp_delay 182 parameter), thus improving energy efficiency. On the other 183 hand, this option increases the duration of RCU grace periods, 184 for example, slowing down synchronize_rcu(). 185 186 Say Y if energy efficiency is critically important, and you 187 don't care about increased grace-period durations. 188 189 Say N if you are unsure. 190 191config RCU_BOOST 192 bool "Enable RCU priority boosting" 193 depends on (RT_MUTEXES && PREEMPT_RCU && RCU_EXPERT) || PREEMPT_RT 194 default y if PREEMPT_RT 195 help 196 This option boosts the priority of preempted RCU readers that 197 block the current preemptible RCU grace period for too long. 198 This option also prevents heavy loads from blocking RCU 199 callback invocation. 200 201 Say Y here if you are working with real-time apps or heavy loads 202 Say N here if you are unsure. 203 204config RCU_BOOST_DELAY 205 int "Milliseconds to delay boosting after RCU grace-period start" 206 range 0 3000 207 depends on RCU_BOOST 208 default 500 209 help 210 This option specifies the time to wait after the beginning of 211 a given grace period before priority-boosting preempted RCU 212 readers blocking that grace period. Note that any RCU reader 213 blocking an expedited RCU grace period is boosted immediately. 214 215 Accept the default if unsure. 216 217config RCU_EXP_KTHREAD 218 bool "Perform RCU expedited work in a real-time kthread" 219 depends on RCU_BOOST && RCU_EXPERT 220 default !PREEMPT_RT && NR_CPUS <= 32 221 help 222 Use this option to further reduce the latencies of expedited 223 grace periods at the expense of being more disruptive. 224 225 This option is disabled by default on PREEMPT_RT=y kernels which 226 disable expedited grace periods after boot by unconditionally 227 setting rcupdate.rcu_normal_after_boot=1. 228 229 Accept the default if unsure. 230 231config RCU_NOCB_CPU 232 bool "Offload RCU callback processing from boot-selected CPUs" 233 depends on TREE_RCU 234 depends on RCU_EXPERT || NO_HZ_FULL 235 default n 236 help 237 Use this option to reduce OS jitter for aggressive HPC or 238 real-time workloads. It can also be used to offload RCU 239 callback invocation to energy-efficient CPUs in battery-powered 240 asymmetric multiprocessors. The price of this reduced jitter 241 is that the overhead of call_rcu() increases and that some 242 workloads will incur significant increases in context-switch 243 rates. 244 245 This option offloads callback invocation from the set of CPUs 246 specified at boot time by the rcu_nocbs parameter. For each 247 such CPU, a kthread ("rcuox/N") will be created to invoke 248 callbacks, where the "N" is the CPU being offloaded, and where 249 the "x" is "p" for RCU-preempt (PREEMPTION kernels) and "s" for 250 RCU-sched (!PREEMPTION kernels). Nothing prevents this kthread 251 from running on the specified CPUs, but (1) the kthreads may be 252 preempted between each callback, and (2) affinity or cgroups can 253 be used to force the kthreads to run on whatever set of CPUs is 254 desired. 255 256 Say Y here if you need reduced OS jitter, despite added overhead. 257 Say N here if you are unsure. 258 259config TASKS_TRACE_RCU_READ_MB 260 bool "Tasks Trace RCU readers use memory barriers in user and idle" 261 depends on RCU_EXPERT 262 default PREEMPT_RT || NR_CPUS < 8 263 help 264 Use this option to further reduce the number of IPIs sent 265 to CPUs executing in userspace or idle during tasks trace 266 RCU grace periods. Given that a reasonable setting of 267 the rcupdate.rcu_task_ipi_delay kernel boot parameter 268 eliminates such IPIs for many workloads, proper setting 269 of this Kconfig option is important mostly for aggressive 270 real-time installations and for battery-powered devices, 271 hence the default chosen above. 272 273 Say Y here if you hate IPIs. 274 Say N here if you hate read-side memory barriers. 275 Take the default if you are unsure. 276 277config RCU_LAZY 278 bool "RCU callback lazy invocation functionality" 279 depends on RCU_NOCB_CPU 280 default n 281 help 282 To save power, batch RCU callbacks and flush after delay, memory 283 pressure, or callback list growing too big. 284 Requires rcu_nocbs=all to be set. 285 286 Use rcutree.enable_rcu_lazy=0 to turn it off at boot time. 287 288config RCU_LAZY_DEFAULT_OFF 289 bool "Turn RCU lazy invocation off by default" 290 depends on RCU_LAZY 291 default n 292 help 293 Allows building the kernel with CONFIG_RCU_LAZY=y yet keep it default 294 off. Boot time param rcutree.enable_rcu_lazy=1 can be used to switch 295 it back on. 296 297config RCU_BOOT_END_DELAY 298 int "Minimum time before RCU may consider in-kernel boot as completed" 299 range 0 120000 300 default 20000 301 help 302 Default value of the minimum time in milliseconds from the start of boot 303 that must elapse before the boot sequence can be marked complete from RCU's 304 perspective, after which RCU's behavior becomes more relaxed. 305 Userspace can also mark the boot as completed sooner than this default 306 by writing the time in milliseconds, say once userspace considers 307 the system as booted, to: /sys/module/rcupdate/parameters/rcu_boot_end_delay. 308 Or even just writing a value of 0 to this sysfs node. The sysfs node can 309 also be used to extend the delay to be larger than the default, assuming 310 the marking of boot completion has not yet occurred. 311 312 The actual delay for RCU's view of the system to be marked as booted can be 313 higher than this value if the kernel takes a long time to initialize but it 314 will never be smaller than this value. 315 316 Accept the default if unsure. 317 318endmenu # "RCU Subsystem" 319