1.. SPDX-License-Identifier: GPL-2.0 2 3============================== 4Using RCU's CPU Stall Detector 5============================== 6 7This document first discusses what sorts of issues RCU's CPU stall 8detector can locate, and then discusses kernel parameters and Kconfig 9options that can be used to fine-tune the detector's operation. Finally, 10this document explains the stall detector's "splat" format. 11 12 13What Causes RCU CPU Stall Warnings? 14=================================== 15 16So your kernel printed an RCU CPU stall warning. The next question is 17"What caused it?" The following problems can result in RCU CPU stall 18warnings: 19 20- A CPU looping in an RCU read-side critical section. 21 22- A CPU looping with interrupts disabled. 23 24- A CPU looping with preemption disabled. 25 26- A CPU looping with bottom halves disabled. 27 28- For !CONFIG_PREEMPT kernels, a CPU looping anywhere in the kernel 29 without invoking schedule(). If the looping in the kernel is 30 really expected and desirable behavior, you might need to add 31 some calls to cond_resched(). 32 33- Booting Linux using a console connection that is too slow to 34 keep up with the boot-time console-message rate. For example, 35 a 115Kbaud serial console can be -way- too slow to keep up 36 with boot-time message rates, and will frequently result in 37 RCU CPU stall warning messages. Especially if you have added 38 debug printk()s. 39 40- Anything that prevents RCU's grace-period kthreads from running. 41 This can result in the "All QSes seen" console-log message. 42 This message will include information on when the kthread last 43 ran and how often it should be expected to run. It can also 44 result in the ``rcu_.*kthread starved for`` console-log message, 45 which will include additional debugging information. 46 47- A CPU-bound real-time task in a CONFIG_PREEMPT kernel, which might 48 happen to preempt a low-priority task in the middle of an RCU 49 read-side critical section. This is especially damaging if 50 that low-priority task is not permitted to run on any other CPU, 51 in which case the next RCU grace period can never complete, which 52 will eventually cause the system to run out of memory and hang. 53 While the system is in the process of running itself out of 54 memory, you might see stall-warning messages. 55 56- A CPU-bound real-time task in a CONFIG_PREEMPT_RT kernel that 57 is running at a higher priority than the RCU softirq threads. 58 This will prevent RCU callbacks from ever being invoked, 59 and in a CONFIG_PREEMPT_RCU kernel will further prevent 60 RCU grace periods from ever completing. Either way, the 61 system will eventually run out of memory and hang. In the 62 CONFIG_PREEMPT_RCU case, you might see stall-warning 63 messages. 64 65 You can use the rcutree.kthread_prio kernel boot parameter to 66 increase the scheduling priority of RCU's kthreads, which can 67 help avoid this problem. However, please note that doing this 68 can increase your system's context-switch rate and thus degrade 69 performance. 70 71- A periodic interrupt whose handler takes longer than the time 72 interval between successive pairs of interrupts. This can 73 prevent RCU's kthreads and softirq handlers from running. 74 Note that certain high-overhead debugging options, for example 75 the function_graph tracer, can result in interrupt handler taking 76 considerably longer than normal, which can in turn result in 77 RCU CPU stall warnings. 78 79- Testing a workload on a fast system, tuning the stall-warning 80 timeout down to just barely avoid RCU CPU stall warnings, and then 81 running the same workload with the same stall-warning timeout on a 82 slow system. Note that thermal throttling and on-demand governors 83 can cause a single system to be sometimes fast and sometimes slow! 84 85- A hardware or software issue shuts off the scheduler-clock 86 interrupt on a CPU that is not in dyntick-idle mode. This 87 problem really has happened, and seems to be most likely to 88 result in RCU CPU stall warnings for CONFIG_NO_HZ_COMMON=n kernels. 89 90- A hardware or software issue that prevents time-based wakeups 91 from occurring. These issues can range from misconfigured or 92 buggy timer hardware through bugs in the interrupt or exception 93 path (whether hardware, firmware, or software) through bugs 94 in Linux's timer subsystem through bugs in the scheduler, and, 95 yes, even including bugs in RCU itself. 96 97- A bug in the RCU implementation. 98 99- A hardware failure. This is quite unlikely, but has occurred 100 at least once in real life. A CPU failed in a running system, 101 becoming unresponsive, but not causing an immediate crash. 102 This resulted in a series of RCU CPU stall warnings, eventually 103 leading the realization that the CPU had failed. 104 105The RCU, RCU-sched, and RCU-tasks implementations have CPU stall warning. 106Note that SRCU does -not- have CPU stall warnings. Please note that 107RCU only detects CPU stalls when there is a grace period in progress. 108No grace period, no CPU stall warnings. 109 110To diagnose the cause of the stall, inspect the stack traces. 111The offending function will usually be near the top of the stack. 112If you have a series of stall warnings from a single extended stall, 113comparing the stack traces can often help determine where the stall 114is occurring, which will usually be in the function nearest the top of 115that portion of the stack which remains the same from trace to trace. 116If you can reliably trigger the stall, ftrace can be quite helpful. 117 118RCU bugs can often be debugged with the help of CONFIG_RCU_TRACE 119and with RCU's event tracing. For information on RCU's event tracing, 120see include/trace/events/rcu.h. 121 122 123Fine-Tuning the RCU CPU Stall Detector 124====================================== 125 126The rcuupdate.rcu_cpu_stall_suppress module parameter disables RCU's 127CPU stall detector, which detects conditions that unduly delay RCU grace 128periods. This module parameter enables CPU stall detection by default, 129but may be overridden via boot-time parameter or at runtime via sysfs. 130The stall detector's idea of what constitutes "unduly delayed" is 131controlled by a set of kernel configuration variables and cpp macros: 132 133CONFIG_RCU_CPU_STALL_TIMEOUT 134---------------------------- 135 136 This kernel configuration parameter defines the period of time 137 that RCU will wait from the beginning of a grace period until it 138 issues an RCU CPU stall warning. This time period is normally 139 21 seconds. 140 141 This configuration parameter may be changed at runtime via the 142 /sys/module/rcupdate/parameters/rcu_cpu_stall_timeout, however 143 this parameter is checked only at the beginning of a cycle. 144 So if you are 10 seconds into a 40-second stall, setting this 145 sysfs parameter to (say) five will shorten the timeout for the 146 -next- stall, or the following warning for the current stall 147 (assuming the stall lasts long enough). It will not affect the 148 timing of the next warning for the current stall. 149 150 Stall-warning messages may be enabled and disabled completely via 151 /sys/module/rcupdate/parameters/rcu_cpu_stall_suppress. 152 153RCU_STALL_DELAY_DELTA 154--------------------- 155 156 Although the lockdep facility is extremely useful, it does add 157 some overhead. Therefore, under CONFIG_PROVE_RCU, the 158 RCU_STALL_DELAY_DELTA macro allows five extra seconds before 159 giving an RCU CPU stall warning message. (This is a cpp 160 macro, not a kernel configuration parameter.) 161 162RCU_STALL_RAT_DELAY 163------------------- 164 165 The CPU stall detector tries to make the offending CPU print its 166 own warnings, as this often gives better-quality stack traces. 167 However, if the offending CPU does not detect its own stall in 168 the number of jiffies specified by RCU_STALL_RAT_DELAY, then 169 some other CPU will complain. This delay is normally set to 170 two jiffies. (This is a cpp macro, not a kernel configuration 171 parameter.) 172 173rcupdate.rcu_task_stall_timeout 174------------------------------- 175 176 This boot/sysfs parameter controls the RCU-tasks stall warning 177 interval. A value of zero or less suppresses RCU-tasks stall 178 warnings. A positive value sets the stall-warning interval 179 in seconds. An RCU-tasks stall warning starts with the line: 180 181 INFO: rcu_tasks detected stalls on tasks: 182 183 And continues with the output of sched_show_task() for each 184 task stalling the current RCU-tasks grace period. 185 186 187Interpreting RCU's CPU Stall-Detector "Splats" 188============================================== 189 190For non-RCU-tasks flavors of RCU, when a CPU detects that it is stalling, 191it will print a message similar to the following:: 192 193 INFO: rcu_sched detected stalls on CPUs/tasks: 194 2-...: (3 GPs behind) idle=06c/0/0 softirq=1453/1455 fqs=0 195 16-...: (0 ticks this GP) idle=81c/0/0 softirq=764/764 fqs=0 196 (detected by 32, t=2603 jiffies, g=7075, q=625) 197 198This message indicates that CPU 32 detected that CPUs 2 and 16 were both 199causing stalls, and that the stall was affecting RCU-sched. This message 200will normally be followed by stack dumps for each CPU. Please note that 201PREEMPT_RCU builds can be stalled by tasks as well as by CPUs, and that 202the tasks will be indicated by PID, for example, "P3421". It is even 203possible for an rcu_state stall to be caused by both CPUs -and- tasks, 204in which case the offending CPUs and tasks will all be called out in the list. 205 206CPU 2's "(3 GPs behind)" indicates that this CPU has not interacted with 207the RCU core for the past three grace periods. In contrast, CPU 16's "(0 208ticks this GP)" indicates that this CPU has not taken any scheduling-clock 209interrupts during the current stalled grace period. 210 211The "idle=" portion of the message prints the dyntick-idle state. 212The hex number before the first "/" is the low-order 12 bits of the 213dynticks counter, which will have an even-numbered value if the CPU 214is in dyntick-idle mode and an odd-numbered value otherwise. The hex 215number between the two "/"s is the value of the nesting, which will be 216a small non-negative number if in the idle loop (as shown above) and a 217very large positive number otherwise. 218 219The "softirq=" portion of the message tracks the number of RCU softirq 220handlers that the stalled CPU has executed. The number before the "/" 221is the number that had executed since boot at the time that this CPU 222last noted the beginning of a grace period, which might be the current 223(stalled) grace period, or it might be some earlier grace period (for 224example, if the CPU might have been in dyntick-idle mode for an extended 225time period. The number after the "/" is the number that have executed 226since boot until the current time. If this latter number stays constant 227across repeated stall-warning messages, it is possible that RCU's softirq 228handlers are no longer able to execute on this CPU. This can happen if 229the stalled CPU is spinning with interrupts are disabled, or, in -rt 230kernels, if a high-priority process is starving RCU's softirq handler. 231 232The "fqs=" shows the number of force-quiescent-state idle/offline 233detection passes that the grace-period kthread has made across this 234CPU since the last time that this CPU noted the beginning of a grace 235period. 236 237The "detected by" line indicates which CPU detected the stall (in this 238case, CPU 32), how many jiffies have elapsed since the start of the grace 239period (in this case 2603), the grace-period sequence number (7075), and 240an estimate of the total number of RCU callbacks queued across all CPUs 241(625 in this case). 242 243In kernels with CONFIG_RCU_FAST_NO_HZ, more information is printed 244for each CPU:: 245 246 0: (64628 ticks this GP) idle=dd5/3fffffffffffffff/0 softirq=82/543 last_accelerate: a345/d342 dyntick_enabled: 1 247 248The "last_accelerate:" prints the low-order 16 bits (in hex) of the 249jiffies counter when this CPU last invoked rcu_try_advance_all_cbs() 250from rcu_needs_cpu() or last invoked rcu_accelerate_cbs() from 251rcu_prepare_for_idle(). "dyntick_enabled: 1" indicates that dyntick-idle 252processing is enabled. 253 254If the grace period ends just as the stall warning starts printing, 255there will be a spurious stall-warning message, which will include 256the following:: 257 258 INFO: Stall ended before state dump start 259 260This is rare, but does happen from time to time in real life. It is also 261possible for a zero-jiffy stall to be flagged in this case, depending 262on how the stall warning and the grace-period initialization happen to 263interact. Please note that it is not possible to entirely eliminate this 264sort of false positive without resorting to things like stop_machine(), 265which is overkill for this sort of problem. 266 267If all CPUs and tasks have passed through quiescent states, but the 268grace period has nevertheless failed to end, the stall-warning splat 269will include something like the following:: 270 271 All QSes seen, last rcu_preempt kthread activity 23807 (4297905177-4297881370), jiffies_till_next_fqs=3, root ->qsmask 0x0 272 273The "23807" indicates that it has been more than 23 thousand jiffies 274since the grace-period kthread ran. The "jiffies_till_next_fqs" 275indicates how frequently that kthread should run, giving the number 276of jiffies between force-quiescent-state scans, in this case three, 277which is way less than 23807. Finally, the root rcu_node structure's 278->qsmask field is printed, which will normally be zero. 279 280If the relevant grace-period kthread has been unable to run prior to 281the stall warning, as was the case in the "All QSes seen" line above, 282the following additional line is printed:: 283 284 kthread starved for 23807 jiffies! g7075 f0x0 RCU_GP_WAIT_FQS(3) ->state=0x1 ->cpu=5 285 286Starving the grace-period kthreads of CPU time can of course result 287in RCU CPU stall warnings even when all CPUs and tasks have passed 288through the required quiescent states. The "g" number shows the current 289grace-period sequence number, the "f" precedes the ->gp_flags command 290to the grace-period kthread, the "RCU_GP_WAIT_FQS" indicates that the 291kthread is waiting for a short timeout, the "state" precedes value of the 292task_struct ->state field, and the "cpu" indicates that the grace-period 293kthread last ran on CPU 5. 294 295 296Multiple Warnings From One Stall 297================================ 298 299If a stall lasts long enough, multiple stall-warning messages will be 300printed for it. The second and subsequent messages are printed at 301longer intervals, so that the time between (say) the first and second 302message will be about three times the interval between the beginning 303of the stall and the first message. 304 305 306Stall Warnings for Expedited Grace Periods 307========================================== 308 309If an expedited grace period detects a stall, it will place a message 310like the following in dmesg:: 311 312 INFO: rcu_sched detected expedited stalls on CPUs/tasks: { 7-... } 21119 jiffies s: 73 root: 0x2/. 313 314This indicates that CPU 7 has failed to respond to a reschedule IPI. 315The three periods (".") following the CPU number indicate that the CPU 316is online (otherwise the first period would instead have been "O"), 317that the CPU was online at the beginning of the expedited grace period 318(otherwise the second period would have instead been "o"), and that 319the CPU has been online at least once since boot (otherwise, the third 320period would instead have been "N"). The number before the "jiffies" 321indicates that the expedited grace period has been going on for 21,119 322jiffies. The number following the "s:" indicates that the expedited 323grace-period sequence counter is 73. The fact that this last value is 324odd indicates that an expedited grace period is in flight. The number 325following "root:" is a bitmask that indicates which children of the root 326rcu_node structure correspond to CPUs and/or tasks that are blocking the 327current expedited grace period. If the tree had more than one level, 328additional hex numbers would be printed for the states of the other 329rcu_node structures in the tree. 330 331As with normal grace periods, PREEMPT_RCU builds can be stalled by 332tasks as well as by CPUs, and that the tasks will be indicated by PID, 333for example, "P3421". 334 335It is entirely possible to see stall warnings from normal and from 336expedited grace periods at about the same time during the same run. 337