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_PREEMPTION 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_PREEMPTION 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.  It can also result in
96	the ``rcu_.*timer wakeup didn't happen for`` console-log message,
97	which will include additional debugging information.
98
99-	A low-level kernel issue that either fails to invoke one of the
100	variants of rcu_user_enter(), rcu_user_exit(), rcu_idle_enter(),
101	rcu_idle_exit(), rcu_irq_enter(), or rcu_irq_exit() on the one
102	hand, or that invokes one of them too many times on the other.
103	Historically, the most frequent issue has been an omission
104	of either irq_enter() or irq_exit(), which in turn invoke
105	rcu_irq_enter() or rcu_irq_exit(), respectively.  Building your
106	kernel with CONFIG_RCU_EQS_DEBUG=y can help track down these types
107	of issues, which sometimes arise in architecture-specific code.
108
109-	A bug in the RCU implementation.
110
111-	A hardware failure.  This is quite unlikely, but has occurred
112	at least once in real life.  A CPU failed in a running system,
113	becoming unresponsive, but not causing an immediate crash.
114	This resulted in a series of RCU CPU stall warnings, eventually
115	leading the realization that the CPU had failed.
116
117The RCU, RCU-sched, and RCU-tasks implementations have CPU stall warning.
118Note that SRCU does *not* have CPU stall warnings.  Please note that
119RCU only detects CPU stalls when there is a grace period in progress.
120No grace period, no CPU stall warnings.
121
122To diagnose the cause of the stall, inspect the stack traces.
123The offending function will usually be near the top of the stack.
124If you have a series of stall warnings from a single extended stall,
125comparing the stack traces can often help determine where the stall
126is occurring, which will usually be in the function nearest the top of
127that portion of the stack which remains the same from trace to trace.
128If you can reliably trigger the stall, ftrace can be quite helpful.
129
130RCU bugs can often be debugged with the help of CONFIG_RCU_TRACE
131and with RCU's event tracing.  For information on RCU's event tracing,
132see include/trace/events/rcu.h.
133
134
135Fine-Tuning the RCU CPU Stall Detector
136======================================
137
138The rcuupdate.rcu_cpu_stall_suppress module parameter disables RCU's
139CPU stall detector, which detects conditions that unduly delay RCU grace
140periods.  This module parameter enables CPU stall detection by default,
141but may be overridden via boot-time parameter or at runtime via sysfs.
142The stall detector's idea of what constitutes "unduly delayed" is
143controlled by a set of kernel configuration variables and cpp macros:
144
145CONFIG_RCU_CPU_STALL_TIMEOUT
146----------------------------
147
148	This kernel configuration parameter defines the period of time
149	that RCU will wait from the beginning of a grace period until it
150	issues an RCU CPU stall warning.  This time period is normally
151	21 seconds.
152
153	This configuration parameter may be changed at runtime via the
154	/sys/module/rcupdate/parameters/rcu_cpu_stall_timeout, however
155	this parameter is checked only at the beginning of a cycle.
156	So if you are 10 seconds into a 40-second stall, setting this
157	sysfs parameter to (say) five will shorten the timeout for the
158	*next* stall, or the following warning for the current stall
159	(assuming the stall lasts long enough).  It will not affect the
160	timing of the next warning for the current stall.
161
162	Stall-warning messages may be enabled and disabled completely via
163	/sys/module/rcupdate/parameters/rcu_cpu_stall_suppress.
164
165RCU_STALL_DELAY_DELTA
166---------------------
167
168	Although the lockdep facility is extremely useful, it does add
169	some overhead.  Therefore, under CONFIG_PROVE_RCU, the
170	RCU_STALL_DELAY_DELTA macro allows five extra seconds before
171	giving an RCU CPU stall warning message.  (This is a cpp
172	macro, not a kernel configuration parameter.)
173
174RCU_STALL_RAT_DELAY
175-------------------
176
177	The CPU stall detector tries to make the offending CPU print its
178	own warnings, as this often gives better-quality stack traces.
179	However, if the offending CPU does not detect its own stall in
180	the number of jiffies specified by RCU_STALL_RAT_DELAY, then
181	some other CPU will complain.  This delay is normally set to
182	two jiffies.  (This is a cpp macro, not a kernel configuration
183	parameter.)
184
185rcupdate.rcu_task_stall_timeout
186-------------------------------
187
188	This boot/sysfs parameter controls the RCU-tasks stall warning
189	interval.  A value of zero or less suppresses RCU-tasks stall
190	warnings.  A positive value sets the stall-warning interval
191	in seconds.  An RCU-tasks stall warning starts with the line:
192
193		INFO: rcu_tasks detected stalls on tasks:
194
195	And continues with the output of sched_show_task() for each
196	task stalling the current RCU-tasks grace period.
197
198
199Interpreting RCU's CPU Stall-Detector "Splats"
200==============================================
201
202For non-RCU-tasks flavors of RCU, when a CPU detects that some other
203CPU is stalling, it will print a message similar to the following::
204
205	INFO: rcu_sched detected stalls on CPUs/tasks:
206	2-...: (3 GPs behind) idle=06c/0/0 softirq=1453/1455 fqs=0
207	16-...: (0 ticks this GP) idle=81c/0/0 softirq=764/764 fqs=0
208	(detected by 32, t=2603 jiffies, g=7075, q=625)
209
210This message indicates that CPU 32 detected that CPUs 2 and 16 were both
211causing stalls, and that the stall was affecting RCU-sched.  This message
212will normally be followed by stack dumps for each CPU.  Please note that
213PREEMPT_RCU builds can be stalled by tasks as well as by CPUs, and that
214the tasks will be indicated by PID, for example, "P3421".  It is even
215possible for an rcu_state stall to be caused by both CPUs *and* tasks,
216in which case the offending CPUs and tasks will all be called out in the list.
217In some cases, CPUs will detect themselves stalling, which will result
218in a self-detected stall.
219
220CPU 2's "(3 GPs behind)" indicates that this CPU has not interacted with
221the RCU core for the past three grace periods.  In contrast, CPU 16's "(0
222ticks this GP)" indicates that this CPU has not taken any scheduling-clock
223interrupts during the current stalled grace period.
224
225The "idle=" portion of the message prints the dyntick-idle state.
226The hex number before the first "/" is the low-order 12 bits of the
227dynticks counter, which will have an even-numbered value if the CPU
228is in dyntick-idle mode and an odd-numbered value otherwise.  The hex
229number between the two "/"s is the value of the nesting, which will be
230a small non-negative number if in the idle loop (as shown above) and a
231very large positive number otherwise.
232
233The "softirq=" portion of the message tracks the number of RCU softirq
234handlers that the stalled CPU has executed.  The number before the "/"
235is the number that had executed since boot at the time that this CPU
236last noted the beginning of a grace period, which might be the current
237(stalled) grace period, or it might be some earlier grace period (for
238example, if the CPU might have been in dyntick-idle mode for an extended
239time period).  The number after the "/" is the number that have executed
240since boot until the current time.  If this latter number stays constant
241across repeated stall-warning messages, it is possible that RCU's softirq
242handlers are no longer able to execute on this CPU.  This can happen if
243the stalled CPU is spinning with interrupts are disabled, or, in -rt
244kernels, if a high-priority process is starving RCU's softirq handler.
245
246The "fqs=" shows the number of force-quiescent-state idle/offline
247detection passes that the grace-period kthread has made across this
248CPU since the last time that this CPU noted the beginning of a grace
249period.
250
251The "detected by" line indicates which CPU detected the stall (in this
252case, CPU 32), how many jiffies have elapsed since the start of the grace
253period (in this case 2603), the grace-period sequence number (7075), and
254an estimate of the total number of RCU callbacks queued across all CPUs
255(625 in this case).
256
257If the grace period ends just as the stall warning starts printing,
258there will be a spurious stall-warning message, which will include
259the following::
260
261	INFO: Stall ended before state dump start
262
263This is rare, but does happen from time to time in real life.  It is also
264possible for a zero-jiffy stall to be flagged in this case, depending
265on how the stall warning and the grace-period initialization happen to
266interact.  Please note that it is not possible to entirely eliminate this
267sort of false positive without resorting to things like stop_machine(),
268which is overkill for this sort of problem.
269
270If all CPUs and tasks have passed through quiescent states, but the
271grace period has nevertheless failed to end, the stall-warning splat
272will include something like the following::
273
274	All QSes seen, last rcu_preempt kthread activity 23807 (4297905177-4297881370), jiffies_till_next_fqs=3, root ->qsmask 0x0
275
276The "23807" indicates that it has been more than 23 thousand jiffies
277since the grace-period kthread ran.  The "jiffies_till_next_fqs"
278indicates how frequently that kthread should run, giving the number
279of jiffies between force-quiescent-state scans, in this case three,
280which is way less than 23807.  Finally, the root rcu_node structure's
281->qsmask field is printed, which will normally be zero.
282
283If the relevant grace-period kthread has been unable to run prior to
284the stall warning, as was the case in the "All QSes seen" line above,
285the following additional line is printed::
286
287	rcu_sched kthread starved for 23807 jiffies! g7075 f0x0 RCU_GP_WAIT_FQS(3) ->state=0x1 ->cpu=5
288	Unless rcu_sched kthread gets sufficient CPU time, OOM is now expected behavior.
289
290Starving the grace-period kthreads of CPU time can of course result
291in RCU CPU stall warnings even when all CPUs and tasks have passed
292through the required quiescent states.  The "g" number shows the current
293grace-period sequence number, the "f" precedes the ->gp_flags command
294to the grace-period kthread, the "RCU_GP_WAIT_FQS" indicates that the
295kthread is waiting for a short timeout, the "state" precedes value of the
296task_struct ->state field, and the "cpu" indicates that the grace-period
297kthread last ran on CPU 5.
298
299If the relevant grace-period kthread does not wake from FQS wait in a
300reasonable time, then the following additional line is printed::
301
302	kthread timer wakeup didn't happen for 23804 jiffies! g7076 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402
303
304The "23804" indicates that kthread's timer expired more than 23 thousand
305jiffies ago.  The rest of the line has meaning similar to the kthread
306starvation case.
307
308Additionally, the following line is printed::
309
310	Possible timer handling issue on cpu=4 timer-softirq=11142
311
312Here "cpu" indicates that the grace-period kthread last ran on CPU 4,
313where it queued the fqs timer.  The number following the "timer-softirq"
314is the current ``TIMER_SOFTIRQ`` count on cpu 4.  If this value does not
315change on successive RCU CPU stall warnings, there is further reason to
316suspect a timer problem.
317
318These messages are usually followed by stack dumps of the CPUs and tasks
319involved in the stall.  These stack traces can help you locate the cause
320of the stall, keeping in mind that the CPU detecting the stall will have
321an interrupt frame that is mainly devoted to detecting the stall.
322
323
324Multiple Warnings From One Stall
325================================
326
327If a stall lasts long enough, multiple stall-warning messages will
328be printed for it.  The second and subsequent messages are printed at
329longer intervals, so that the time between (say) the first and second
330message will be about three times the interval between the beginning
331of the stall and the first message.  It can be helpful to compare the
332stack dumps for the different messages for the same stalled grace period.
333
334
335Stall Warnings for Expedited Grace Periods
336==========================================
337
338If an expedited grace period detects a stall, it will place a message
339like the following in dmesg::
340
341	INFO: rcu_sched detected expedited stalls on CPUs/tasks: { 7-... } 21119 jiffies s: 73 root: 0x2/.
342
343This indicates that CPU 7 has failed to respond to a reschedule IPI.
344The three periods (".") following the CPU number indicate that the CPU
345is online (otherwise the first period would instead have been "O"),
346that the CPU was online at the beginning of the expedited grace period
347(otherwise the second period would have instead been "o"), and that
348the CPU has been online at least once since boot (otherwise, the third
349period would instead have been "N").  The number before the "jiffies"
350indicates that the expedited grace period has been going on for 21,119
351jiffies.  The number following the "s:" indicates that the expedited
352grace-period sequence counter is 73.  The fact that this last value is
353odd indicates that an expedited grace period is in flight.  The number
354following "root:" is a bitmask that indicates which children of the root
355rcu_node structure correspond to CPUs and/or tasks that are blocking the
356current expedited grace period.  If the tree had more than one level,
357additional hex numbers would be printed for the states of the other
358rcu_node structures in the tree.
359
360As with normal grace periods, PREEMPT_RCU builds can be stalled by
361tasks as well as by CPUs, and that the tasks will be indicated by PID,
362for example, "P3421".
363
364It is entirely possible to see stall warnings from normal and from
365expedited grace periods at about the same time during the same run.
366