1 Using RCU's CPU Stall Detector
3 The rcu_cpu_stall_suppress module parameter enables RCU's CPU stall
4 detector, which detects conditions that unduly delay RCU grace periods.
5 This module parameter enables CPU stall detection by default, but
6 may be overridden via boot-time parameter or at runtime via sysfs.
7 The stall detector's idea of what constitutes "unduly delayed" is
8 controlled by a set of kernel configuration variables and cpp macros:
10 CONFIG_RCU_CPU_STALL_TIMEOUT
12 This kernel configuration parameter defines the period of time
13 that RCU will wait from the beginning of a grace period until it
14 issues an RCU CPU stall warning. This time period is normally
17 This configuration parameter may be changed at runtime via the
18 /sys/module/rcutree/parameters/rcu_cpu_stall_timeout, however
19 this parameter is checked only at the beginning of a cycle.
20 So if you are 10 seconds into a 40-second stall, setting this
21 sysfs parameter to (say) five will shorten the timeout for the
22 -next- stall, or the following warning for the current stall
23 (assuming the stall lasts long enough). It will not affect the
24 timing of the next warning for the current stall.
26 Stall-warning messages may be enabled and disabled completely via
27 /sys/module/rcutree/parameters/rcu_cpu_stall_suppress.
29 CONFIG_RCU_CPU_STALL_VERBOSE
31 This kernel configuration parameter causes the stall warning to
32 also dump the stacks of any tasks that are blocking the current
33 RCU-preempt grace period.
35 CONFIG_RCU_CPU_STALL_INFO
37 This kernel configuration parameter causes the stall warning to
38 print out additional per-CPU diagnostic information, including
39 information on scheduling-clock ticks and RCU's idle-CPU tracking.
43 Although the lockdep facility is extremely useful, it does add
44 some overhead. Therefore, under CONFIG_PROVE_RCU, the
45 RCU_STALL_DELAY_DELTA macro allows five extra seconds before
46 giving an RCU CPU stall warning message. (This is a cpp
47 macro, not a kernel configuration parameter.)
51 The CPU stall detector tries to make the offending CPU print its
52 own warnings, as this often gives better-quality stack traces.
53 However, if the offending CPU does not detect its own stall in
54 the number of jiffies specified by RCU_STALL_RAT_DELAY, then
55 some other CPU will complain. This delay is normally set to
56 two jiffies. (This is a cpp macro, not a kernel configuration
59 When a CPU detects that it is stalling, it will print a message similar
62 INFO: rcu_sched_state detected stall on CPU 5 (t=2500 jiffies)
64 This message indicates that CPU 5 detected that it was causing a stall,
65 and that the stall was affecting RCU-sched. This message will normally be
66 followed by a stack dump of the offending CPU. On TREE_RCU kernel builds,
67 RCU and RCU-sched are implemented by the same underlying mechanism,
68 while on TREE_PREEMPT_RCU kernel builds, RCU is instead implemented
71 On the other hand, if the offending CPU fails to print out a stall-warning
72 message quickly enough, some other CPU will print a message similar to
75 INFO: rcu_bh_state detected stalls on CPUs/tasks: { 3 5 } (detected by 2, 2502 jiffies)
77 This message indicates that CPU 2 detected that CPUs 3 and 5 were both
78 causing stalls, and that the stall was affecting RCU-bh. This message
79 will normally be followed by stack dumps for each CPU. Please note that
80 TREE_PREEMPT_RCU builds can be stalled by tasks as well as by CPUs,
81 and that the tasks will be indicated by PID, for example, "P3421".
82 It is even possible for a rcu_preempt_state stall to be caused by both
83 CPUs -and- tasks, in which case the offending CPUs and tasks will all
84 be called out in the list.
86 Finally, if the grace period ends just as the stall warning starts
87 printing, there will be a spurious stall-warning message:
89 INFO: rcu_bh_state detected stalls on CPUs/tasks: { } (detected by 4, 2502 jiffies)
91 This is rare, but does happen from time to time in real life. It is also
92 possible for a zero-jiffy stall to be flagged in this case, depending
93 on how the stall warning and the grace-period initialization happen to
94 interact. Please note that it is not possible to entirely eliminate this
95 sort of false positive without resorting to things like stop_machine(),
96 which is overkill for this sort of problem.
98 If the CONFIG_RCU_CPU_STALL_INFO kernel configuration parameter is set,
99 more information is printed with the stall-warning message, for example:
101 INFO: rcu_preempt detected stall on CPU
102 0: (63959 ticks this GP) idle=241/3fffffffffffffff/0 softirq=82/543
105 In kernels with CONFIG_RCU_FAST_NO_HZ, even more information is
108 INFO: rcu_preempt detected stall on CPU
109 0: (64628 ticks this GP) idle=dd5/3fffffffffffffff/0 softirq=82/543 last_accelerate: a345/d342 nonlazy_posted: 25 .D
112 The "(64628 ticks this GP)" indicates that this CPU has taken more
113 than 64,000 scheduling-clock interrupts during the current stalled
114 grace period. If the CPU was not yet aware of the current grace
115 period (for example, if it was offline), then this part of the message
116 indicates how many grace periods behind the CPU is.
118 The "idle=" portion of the message prints the dyntick-idle state.
119 The hex number before the first "/" is the low-order 12 bits of the
120 dynticks counter, which will have an even-numbered value if the CPU is
121 in dyntick-idle mode and an odd-numbered value otherwise. The hex
122 number between the two "/"s is the value of the nesting, which will
123 be a small positive number if in the idle loop and a very large positive
124 number (as shown above) otherwise.
126 The "softirq=" portion of the message tracks the number of RCU softirq
127 handlers that the stalled CPU has executed. The number before the "/"
128 is the number that had executed since boot at the time that this CPU
129 last noted the beginning of a grace period, which might be the current
130 (stalled) grace period, or it might be some earlier grace period (for
131 example, if the CPU might have been in dyntick-idle mode for an extended
132 time period. The number after the "/" is the number that have executed
133 since boot until the current time. If this latter number stays constant
134 across repeated stall-warning messages, it is possible that RCU's softirq
135 handlers are no longer able to execute on this CPU. This can happen if
136 the stalled CPU is spinning with interrupts are disabled, or, in -rt
137 kernels, if a high-priority process is starving RCU's softirq handler.
139 For CONFIG_RCU_FAST_NO_HZ kernels, the "last_accelerate:" prints the
140 low-order 16 bits (in hex) of the jiffies counter when this CPU last
141 invoked rcu_try_advance_all_cbs() from rcu_needs_cpu() or last invoked
142 rcu_accelerate_cbs() from rcu_prepare_for_idle(). The "nonlazy_posted:"
143 prints the number of non-lazy callbacks posted since the last call to
144 rcu_needs_cpu(). Finally, an "L" indicates that there are currently
145 no non-lazy callbacks ("." is printed otherwise, as shown above) and
146 "D" indicates that dyntick-idle processing is enabled ("." is printed
147 otherwise, for example, if disabled via the "nohz=" kernel boot parameter).
150 Multiple Warnings From One Stall
152 If a stall lasts long enough, multiple stall-warning messages will be
153 printed for it. The second and subsequent messages are printed at
154 longer intervals, so that the time between (say) the first and second
155 message will be about three times the interval between the beginning
156 of the stall and the first message.
159 What Causes RCU CPU Stall Warnings?
161 So your kernel printed an RCU CPU stall warning. The next question is
162 "What caused it?" The following problems can result in RCU CPU stall
165 o A CPU looping in an RCU read-side critical section.
167 o A CPU looping with interrupts disabled. This condition can
168 result in RCU-sched and RCU-bh stalls.
170 o A CPU looping with preemption disabled. This condition can
171 result in RCU-sched stalls and, if ksoftirqd is in use, RCU-bh
174 o A CPU looping with bottom halves disabled. This condition can
175 result in RCU-sched and RCU-bh stalls.
177 o For !CONFIG_PREEMPT kernels, a CPU looping anywhere in the kernel
178 without invoking schedule().
180 o A CPU-bound real-time task in a CONFIG_PREEMPT kernel, which might
181 happen to preempt a low-priority task in the middle of an RCU
182 read-side critical section. This is especially damaging if
183 that low-priority task is not permitted to run on any other CPU,
184 in which case the next RCU grace period can never complete, which
185 will eventually cause the system to run out of memory and hang.
186 While the system is in the process of running itself out of
187 memory, you might see stall-warning messages.
189 o A CPU-bound real-time task in a CONFIG_PREEMPT_RT kernel that
190 is running at a higher priority than the RCU softirq threads.
191 This will prevent RCU callbacks from ever being invoked,
192 and in a CONFIG_TREE_PREEMPT_RCU kernel will further prevent
193 RCU grace periods from ever completing. Either way, the
194 system will eventually run out of memory and hang. In the
195 CONFIG_TREE_PREEMPT_RCU case, you might see stall-warning
198 o A hardware or software issue shuts off the scheduler-clock
199 interrupt on a CPU that is not in dyntick-idle mode. This
200 problem really has happened, and seems to be most likely to
201 result in RCU CPU stall warnings for CONFIG_NO_HZ_COMMON=n kernels.
203 o A bug in the RCU implementation.
205 o A hardware failure. This is quite unlikely, but has occurred
206 at least once in real life. A CPU failed in a running system,
207 becoming unresponsive, but not causing an immediate crash.
208 This resulted in a series of RCU CPU stall warnings, eventually
209 leading the realization that the CPU had failed.
211 The RCU, RCU-sched, and RCU-bh implementations have CPU stall warning.
212 SRCU does not have its own CPU stall warnings, but its calls to
213 synchronize_sched() will result in RCU-sched detecting RCU-sched-related
214 CPU stalls. Please note that RCU only detects CPU stalls when there is
215 a grace period in progress. No grace period, no CPU stall warnings.
217 To diagnose the cause of the stall, inspect the stack traces.
218 The offending function will usually be near the top of the stack.
219 If you have a series of stall warnings from a single extended stall,
220 comparing the stack traces can often help determine where the stall
221 is occurring, which will usually be in the function nearest the top of
222 that portion of the stack which remains the same from trace to trace.
223 If you can reliably trigger the stall, ftrace can be quite helpful.
225 RCU bugs can often be debugged with the help of CONFIG_RCU_TRACE
226 and with RCU's event tracing. For information on RCU's event tracing,
227 see include/trace/events/rcu.h.