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 (PREEMPT || PREEMPT_RT || PREEMPT_DYNAMIC) 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 !PREEMPT_RCU && !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 TINY_SRCU 57 bool 58 default y if TINY_RCU 59 help 60 This option selects the single-CPU non-preemptible version of SRCU. 61 62config TREE_SRCU 63 bool 64 default y if !TINY_RCU 65 help 66 This option selects the full-fledged version of SRCU. 67 68config FORCE_NEED_SRCU_NMI_SAFE 69 bool "Force selection of NEED_SRCU_NMI_SAFE" 70 depends on !TINY_SRCU 71 select NEED_SRCU_NMI_SAFE 72 default n 73 help 74 This option forces selection of the NEED_SRCU_NMI_SAFE 75 Kconfig option, allowing testing of srcu_read_lock_nmisafe() 76 and srcu_read_unlock_nmisafe() on architectures (like x86) 77 that select the ARCH_HAS_NMI_SAFE_THIS_CPU_OPS Kconfig option. 78 79config NEED_SRCU_NMI_SAFE 80 def_bool HAVE_NMI && !ARCH_HAS_NMI_SAFE_THIS_CPU_OPS && !TINY_SRCU 81 82config TASKS_RCU_GENERIC 83 def_bool TASKS_RCU || TASKS_RUDE_RCU || TASKS_TRACE_RCU 84 help 85 This option enables generic infrastructure code supporting 86 task-based RCU implementations. Not for manual selection. 87 88config FORCE_TASKS_RCU 89 bool "Force selection of TASKS_RCU" 90 depends on RCU_EXPERT 91 select TASKS_RCU 92 default n 93 help 94 This option force-enables a task-based RCU implementation 95 that uses only voluntary context switch (not preemption!), 96 idle, and user-mode execution as quiescent states. Not for 97 manual selection in most cases. 98 99config NEED_TASKS_RCU 100 bool 101 default n 102 103config TASKS_RCU 104 bool 105 default NEED_TASKS_RCU && PREEMPTION 106 select IRQ_WORK 107 108config FORCE_TASKS_RUDE_RCU 109 bool "Force selection of Tasks Rude RCU" 110 depends on RCU_EXPERT 111 select TASKS_RUDE_RCU 112 default n 113 help 114 This option force-enables a task-based RCU implementation 115 that uses only context switch (including preemption) and 116 user-mode execution as quiescent states. It forces IPIs and 117 context switches on all online CPUs, including idle ones, 118 so use with caution. Not for manual selection in most cases. 119 120config TASKS_RUDE_RCU 121 bool 122 default n 123 select IRQ_WORK 124 125config FORCE_TASKS_TRACE_RCU 126 bool "Force selection of Tasks Trace RCU" 127 depends on RCU_EXPERT 128 select TASKS_TRACE_RCU 129 default n 130 help 131 This option enables a task-based RCU implementation that uses 132 explicit rcu_read_lock_trace() read-side markers, and allows 133 these readers to appear in the idle loop as well as on the 134 CPU hotplug code paths. It can force IPIs on online CPUs, 135 including idle ones, so use with caution. Not for manual 136 selection in most cases. 137 138config TASKS_TRACE_RCU 139 bool 140 default n 141 select IRQ_WORK 142 143config RCU_STALL_COMMON 144 def_bool TREE_RCU 145 help 146 This option enables RCU CPU stall code that is common between 147 the TINY and TREE variants of RCU. The purpose is to allow 148 the tiny variants to disable RCU CPU stall warnings, while 149 making these warnings mandatory for the tree variants. 150 151config RCU_NEED_SEGCBLIST 152 def_bool ( TREE_RCU || TREE_SRCU || TASKS_RCU_GENERIC ) 153 154config RCU_FANOUT 155 int "Tree-based hierarchical RCU fanout value" 156 range 2 64 if 64BIT 157 range 2 32 if !64BIT 158 depends on TREE_RCU && RCU_EXPERT 159 default 64 if 64BIT 160 default 32 if !64BIT 161 help 162 This option controls the fanout of hierarchical implementations 163 of RCU, allowing RCU to work efficiently on machines with 164 large numbers of CPUs. This value must be at least the fourth 165 root of NR_CPUS, which allows NR_CPUS to be insanely large. 166 The default value of RCU_FANOUT should be used for production 167 systems, but if you are stress-testing the RCU implementation 168 itself, small RCU_FANOUT values allow you to test large-system 169 code paths on small(er) systems. 170 171 Select a specific number if testing RCU itself. 172 Take the default if unsure. 173 174config RCU_FANOUT_LEAF 175 int "Tree-based hierarchical RCU leaf-level fanout value" 176 range 2 64 if 64BIT && !RCU_STRICT_GRACE_PERIOD 177 range 2 32 if !64BIT && !RCU_STRICT_GRACE_PERIOD 178 range 2 3 if RCU_STRICT_GRACE_PERIOD 179 depends on TREE_RCU && RCU_EXPERT 180 default 16 if !RCU_STRICT_GRACE_PERIOD 181 default 2 if RCU_STRICT_GRACE_PERIOD 182 help 183 This option controls the leaf-level fanout of hierarchical 184 implementations of RCU, and allows trading off cache misses 185 against lock contention. Systems that synchronize their 186 scheduling-clock interrupts for energy-efficiency reasons will 187 want the default because the smaller leaf-level fanout keeps 188 lock contention levels acceptably low. Very large systems 189 (hundreds or thousands of CPUs) will instead want to set this 190 value to the maximum value possible in order to reduce the 191 number of cache misses incurred during RCU's grace-period 192 initialization. These systems tend to run CPU-bound, and thus 193 are not helped by synchronized interrupts, and thus tend to 194 skew them, which reduces lock contention enough that large 195 leaf-level fanouts work well. That said, setting leaf-level 196 fanout to a large number will likely cause problematic 197 lock contention on the leaf-level rcu_node structures unless 198 you boot with the skew_tick kernel parameter. 199 200 Select a specific number if testing RCU itself. 201 202 Select the maximum permissible value for large systems, but 203 please understand that you may also need to set the skew_tick 204 kernel boot parameter to avoid contention on the rcu_node 205 structure's locks. 206 207 Take the default if unsure. 208 209config RCU_BOOST 210 bool "Enable RCU priority boosting" 211 depends on (RT_MUTEXES && PREEMPT_RCU && RCU_EXPERT) || PREEMPT_RT 212 default y if PREEMPT_RT 213 help 214 This option boosts the priority of preempted RCU readers that 215 block the current preemptible RCU grace period for too long. 216 This option also prevents heavy loads from blocking RCU 217 callback invocation. 218 219 Say Y here if you are working with real-time apps or heavy loads 220 Say N here if you are unsure. 221 222config RCU_BOOST_DELAY 223 int "Milliseconds to delay boosting after RCU grace-period start" 224 range 0 3000 225 depends on RCU_BOOST 226 default 500 227 help 228 This option specifies the time to wait after the beginning of 229 a given grace period before priority-boosting preempted RCU 230 readers blocking that grace period. Note that any RCU reader 231 blocking an expedited RCU grace period is boosted immediately. 232 233 Accept the default if unsure. 234 235config RCU_EXP_KTHREAD 236 bool "Perform RCU expedited work in a real-time kthread" 237 depends on RCU_BOOST && RCU_EXPERT 238 default !PREEMPT_RT && NR_CPUS <= 32 239 help 240 Use this option to further reduce the latencies of expedited 241 grace periods at the expense of being more disruptive. 242 243 This option is disabled by default on PREEMPT_RT=y kernels which 244 disable expedited grace periods after boot by unconditionally 245 setting rcupdate.rcu_normal_after_boot=1. 246 247 Accept the default if unsure. 248 249config RCU_NOCB_CPU 250 bool "Offload RCU callback processing from boot-selected CPUs" 251 depends on TREE_RCU 252 depends on RCU_EXPERT || NO_HZ_FULL 253 default n 254 help 255 Use this option to reduce OS jitter for aggressive HPC or 256 real-time workloads. It can also be used to offload RCU 257 callback invocation to energy-efficient CPUs in battery-powered 258 asymmetric multiprocessors. The price of this reduced jitter 259 is that the overhead of call_rcu() increases and that some 260 workloads will incur significant increases in context-switch 261 rates. 262 263 This option offloads callback invocation from the set of 264 CPUs specified at boot time by the rcu_nocbs parameter. 265 For each such CPU, a kthread ("rcuox/N") will be created to 266 invoke callbacks, where the "N" is the CPU being offloaded, 267 and where the "x" is "p" for RCU-preempt (PREEMPTION kernels) 268 and "s" for RCU-sched (!PREEMPTION kernels). This option 269 also creates another kthread for each sqrt(nr_cpu_ids) CPUs 270 ("rcuog/N", where N is the first CPU in that group to come 271 online), which handles grace periods for its group. Nothing 272 prevents these kthreads from running on the specified CPUs, 273 but (1) the kthreads may be preempted between each callback, 274 and (2) affinity or cgroups can be used to force the kthreads 275 to run on whatever set of CPUs is desired. 276 277 The sqrt(nr_cpu_ids) grouping may be overridden using the 278 rcutree.rcu_nocb_gp_stride kernel boot parameter. This can 279 be especially helpful for smaller numbers of CPUs, where 280 sqrt(nr_cpu_ids) can be a bit of a blunt instrument. 281 282 Say Y here if you need reduced OS jitter, despite added overhead. 283 Say N here if you are unsure. 284 285config RCU_NOCB_CPU_DEFAULT_ALL 286 bool "Offload RCU callback processing from all CPUs by default" 287 depends on RCU_NOCB_CPU 288 default n 289 help 290 Use this option to offload callback processing from all CPUs 291 by default, in the absence of the rcu_nocbs or nohz_full boot 292 parameter. This also avoids the need to use any boot parameters 293 to achieve the effect of offloading all CPUs on boot. 294 295 Say Y here if you want offload all CPUs by default on boot. 296 Say N here if you are unsure. 297 298config RCU_NOCB_CPU_CB_BOOST 299 bool "Offload RCU callback from real-time kthread" 300 depends on RCU_NOCB_CPU && RCU_BOOST 301 default y if PREEMPT_RT 302 help 303 Use this option to invoke offloaded callbacks as SCHED_FIFO 304 to avoid starvation by heavy SCHED_OTHER background load. 305 Of course, running as SCHED_FIFO during callback floods will 306 cause the rcuo[ps] kthreads to monopolize the CPU for hundreds 307 of milliseconds or more. Therefore, when enabling this option, 308 it is your responsibility to ensure that latency-sensitive 309 tasks either run with higher priority or run on some other CPU. 310 311 Say Y here if you want to set RT priority for offloading kthreads. 312 Say N here if you are building a !PREEMPT_RT kernel and are unsure. 313 314config TASKS_TRACE_RCU_READ_MB 315 bool "Tasks Trace RCU readers use memory barriers in user and idle" 316 depends on RCU_EXPERT && TASKS_TRACE_RCU 317 default PREEMPT_RT || NR_CPUS < 8 318 help 319 Use this option to further reduce the number of IPIs sent 320 to CPUs executing in userspace or idle during tasks trace 321 RCU grace periods. Given that a reasonable setting of 322 the rcupdate.rcu_task_ipi_delay kernel boot parameter 323 eliminates such IPIs for many workloads, proper setting 324 of this Kconfig option is important mostly for aggressive 325 real-time installations and for battery-powered devices, 326 hence the default chosen above. 327 328 Say Y here if you hate IPIs. 329 Say N here if you hate read-side memory barriers. 330 Take the default if you are unsure. 331 332config RCU_LAZY 333 bool "RCU callback lazy invocation functionality" 334 depends on RCU_NOCB_CPU 335 default n 336 help 337 To save power, batch RCU callbacks and delay starting the 338 corresponding grace period for multiple seconds. The grace 339 period will be started after this delay, in case of memory 340 pressure, or if the corresponding CPU's callback list grows 341 too large. 342 343 These delays happen only on rcu_nocbs CPUs, that is, CPUs 344 whose callbacks have been offloaded. 345 346 Use the rcutree.enable_rcu_lazy=0 kernel-boot parameter to 347 globally disable these delays. 348 349config RCU_LAZY_DEFAULT_OFF 350 bool "Turn RCU lazy invocation off by default" 351 depends on RCU_LAZY 352 default n 353 help 354 Build the kernel with CONFIG_RCU_LAZY=y, but cause the kernel 355 to boot with these energy-efficiency delays disabled. Use the 356 rcutree.enable_rcu_lazy=0 kernel-boot parameter to override 357 the this option at boot time, thus re-enabling these delays. 358 359config RCU_DOUBLE_CHECK_CB_TIME 360 bool "RCU callback-batch backup time check" 361 depends on RCU_EXPERT 362 default n 363 help 364 Use this option to provide more precise enforcement of the 365 rcutree.rcu_resched_ns module parameter in situations where 366 a single RCU callback might run for hundreds of microseconds, 367 thus defeating the 32-callback batching used to amortize the 368 cost of the fine-grained but expensive local_clock() function. 369 370 This option rounds rcutree.rcu_resched_ns up to the next 371 jiffy, and overrides the 32-callback batching if this limit 372 is exceeded. 373 374 Say Y here if you need tighter callback-limit enforcement. 375 Say N here if you are unsure. 376 377endmenu # "RCU Subsystem" 378