tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1RCU Torture Test Operation
2
3
4CONFIG_RCU_TORTURE_TEST
5
6The CONFIG_RCU_TORTURE_TEST config option is available for all RCU
7implementations. It creates an rcutorture kernel module that can
8be loaded to run a torture test. The test periodically outputs
9status messages via printk(), which can be examined via the dmesg
10command (perhaps grepping for "torture"). The test is started
11when the module is loaded, and stops when the module is unloaded.
12
13CONFIG_RCU_TORTURE_TEST_RUNNABLE
14
15It is also possible to specify CONFIG_RCU_TORTURE_TEST=y, which will
16result in the tests being loaded into the base kernel. In this case,
17the CONFIG_RCU_TORTURE_TEST_RUNNABLE config option is used to specify
18whether the RCU torture tests are to be started immediately during
19boot or whether the /proc/sys/kernel/rcutorture_runnable file is used
20to enable them. This /proc file can be used to repeatedly pause and
21restart the tests, regardless of the initial state specified by the
22CONFIG_RCU_TORTURE_TEST_RUNNABLE config option.
23
24You will normally -not- want to start the RCU torture tests during boot
25(and thus the default is CONFIG_RCU_TORTURE_TEST_RUNNABLE=n), but doing
26this can sometimes be useful in finding boot-time bugs.
27
28
29MODULE PARAMETERS
30
31This module has the following parameters:
32
33irqreaders Says to invoke RCU readers from irq level. This is currently
34 done via timers. Defaults to "1" for variants of RCU that
35 permit this. (Or, more accurately, variants of RCU that do
36 -not- permit this know to ignore this variable.)
37
38nfakewriters This is the number of RCU fake writer threads to run. Fake
39 writer threads repeatedly use the synchronous "wait for
40 current readers" function of the interface selected by
41 torture_type, with a delay between calls to allow for various
42 different numbers of writers running in parallel.
43 nfakewriters defaults to 4, which provides enough parallelism
44 to trigger special cases caused by multiple writers, such as
45 the synchronize_srcu() early return optimization.
46
47nreaders This is the number of RCU reading threads supported.
48 The default is twice the number of CPUs. Why twice?
49 To properly exercise RCU implementations with preemptible
50 read-side critical sections.
51
52shuffle_interval
53 The number of seconds to keep the test threads affinitied
54 to a particular subset of the CPUs, defaults to 3 seconds.
55 Used in conjunction with test_no_idle_hz.
56
57stat_interval The number of seconds between output of torture
58 statistics (via printk()). Regardless of the interval,
59 statistics are printed when the module is unloaded.
60 Setting the interval to zero causes the statistics to
61 be printed -only- when the module is unloaded, and this
62 is the default.
63
64stutter The length of time to run the test before pausing for this
65 same period of time. Defaults to "stutter=5", so as
66 to run and pause for (roughly) five-second intervals.
67 Specifying "stutter=0" causes the test to run continuously
68 without pausing, which is the old default behavior.
69
70test_no_idle_hz Whether or not to test the ability of RCU to operate in
71 a kernel that disables the scheduling-clock interrupt to
72 idle CPUs. Boolean parameter, "1" to test, "0" otherwise.
73 Defaults to omitting this test.
74
75torture_type The type of RCU to test: "rcu" for the rcu_read_lock() API,
76 "rcu_sync" for rcu_read_lock() with synchronous reclamation,
77 "rcu_bh" for the rcu_read_lock_bh() API, "rcu_bh_sync" for
78 rcu_read_lock_bh() with synchronous reclamation, "srcu" for
79 the "srcu_read_lock()" API, and "sched" for the use of
80 preempt_disable() together with synchronize_sched().
81
82verbose Enable debug printk()s. Default is disabled.
83
84
85OUTPUT
86
87The statistics output is as follows:
88
89 rcu-torture: --- Start of test: nreaders=16 stat_interval=0 verbose=0
90 rcu-torture: rtc: 0000000000000000 ver: 1916 tfle: 0 rta: 1916 rtaf: 0 rtf: 1915
91 rcu-torture: Reader Pipe: 1466408 9747 0 0 0 0 0 0 0 0 0
92 rcu-torture: Reader Batch: 1464477 11678 0 0 0 0 0 0 0 0
93 rcu-torture: Free-Block Circulation: 1915 1915 1915 1915 1915 1915 1915 1915 1915 1915 0
94 rcu-torture: --- End of test
95
96The command "dmesg | grep torture:" will extract this information on
97most systems. On more esoteric configurations, it may be necessary to
98use other commands to access the output of the printk()s used by
99the RCU torture test. The printk()s use KERN_ALERT, so they should
100be evident. ;-)
101
102The entries are as follows:
103
104o "rtc": The hexadecimal address of the structure currently visible
105 to readers.
106
107o "ver": The number of times since boot that the rcutw writer task
108 has changed the structure visible to readers.
109
110o "tfle": If non-zero, indicates that the "torture freelist"
111 containing structure to be placed into the "rtc" area is empty.
112 This condition is important, since it can fool you into thinking
113 that RCU is working when it is not. :-/
114
115o "rta": Number of structures allocated from the torture freelist.
116
117o "rtaf": Number of allocations from the torture freelist that have
118 failed due to the list being empty.
119
120o "rtf": Number of frees into the torture freelist.
121
122o "Reader Pipe": Histogram of "ages" of structures seen by readers.
123 If any entries past the first two are non-zero, RCU is broken.
124 And rcutorture prints the error flag string "!!!" to make sure
125 you notice. The age of a newly allocated structure is zero,
126 it becomes one when removed from reader visibility, and is
127 incremented once per grace period subsequently -- and is freed
128 after passing through (RCU_TORTURE_PIPE_LEN-2) grace periods.
129
130 The output displayed above was taken from a correctly working
131 RCU. If you want to see what it looks like when broken, break
132 it yourself. ;-)
133
134o "Reader Batch": Another histogram of "ages" of structures seen
135 by readers, but in terms of counter flips (or batches) rather
136 than in terms of grace periods. The legal number of non-zero
137 entries is again two. The reason for this separate view is that
138 it is sometimes easier to get the third entry to show up in the
139 "Reader Batch" list than in the "Reader Pipe" list.
140
141o "Free-Block Circulation": Shows the number of torture structures
142 that have reached a given point in the pipeline. The first element
143 should closely correspond to the number of structures allocated,
144 the second to the number that have been removed from reader view,
145 and all but the last remaining to the corresponding number of
146 passes through a grace period. The last entry should be zero,
147 as it is only incremented if a torture structure's counter
148 somehow gets incremented farther than it should.
149
150Different implementations of RCU can provide implementation-specific
151additional information. For example, SRCU provides the following:
152
153 srcu-torture: rtc: f8cf46a8 ver: 355 tfle: 0 rta: 356 rtaf: 0 rtf: 346 rtmbe: 0
154 srcu-torture: Reader Pipe: 559738 939 0 0 0 0 0 0 0 0 0
155 srcu-torture: Reader Batch: 560434 243 0 0 0 0 0 0 0 0
156 srcu-torture: Free-Block Circulation: 355 354 353 352 351 350 349 348 347 346 0
157 srcu-torture: per-CPU(idx=1): 0(0,1) 1(0,1) 2(0,0) 3(0,1)
158
159The first four lines are similar to those for RCU. The last line shows
160the per-CPU counter state. The numbers in parentheses are the values
161of the "old" and "current" counters for the corresponding CPU. The
162"idx" value maps the "old" and "current" values to the underlying array,
163and is useful for debugging.
164
165
166USAGE
167
168The following script may be used to torture RCU:
169
170 #!/bin/sh
171
172 modprobe rcutorture
173 sleep 100
174 rmmod rcutorture
175 dmesg | grep torture:
176
177The output can be manually inspected for the error flag of "!!!".
178One could of course create a more elaborate script that automatically
179checked for such errors. The "rmmod" command forces a "SUCCESS" or
180"FAILURE" indication to be printk()ed.