Documentation/RCU/torture.txt

   1 RCU Torture Test Operation
   2
   3
   4 CONFIG_RCU_TORTURE_TEST
   5
   6 The CONFIG_RCU_TORTURE_TEST config option is available for all RCU
   7 implementations.  It creates an rcutorture kernel module that can
   8 be loaded to run a torture test.  The test periodically outputs
   9 status messages via printk(), which can be examined via the dmesg
  10 command (perhaps grepping for "torture").  The test is started
  11 when the module is loaded, and stops when the module is unloaded.
  12
  13 CONFIG_RCU_TORTURE_TEST_RUNNABLE
  14
  15 It is also possible to specify CONFIG_RCU_TORTURE_TEST=y, which will
  16 result in the tests being loaded into the base kernel.  In this case,
  17 the CONFIG_RCU_TORTURE_TEST_RUNNABLE config option is used to specify
  18 whether the RCU torture tests are to be started immediately during
  19 boot or whether the /proc/sys/kernel/rcutorture_runnable file is used
  20 to enable them.  This /proc file can be used to repeatedly pause and
  21 restart the tests, regardless of the initial state specified by the
  22 CONFIG_RCU_TORTURE_TEST_RUNNABLE config option.
  23
  24 You 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
  26 this can sometimes be useful in finding boot-time bugs.
  27
  28
  29 MODULE PARAMETERS
  30
  31 This module has the following parameters:
  32
  33 fqs_duration    Duration (in microseconds) of artificially induced bursts
  34                 of force_quiescent_state() invocations.  In RCU
  35                 implementations having force_quiescent_state(), these
  36                 bursts help force races between forcing a given grace
  37                 period and that grace period ending on its own.
  38
  39 fqs_holdoff     Holdoff time (in microseconds) between consecutive calls
  40                 to force_quiescent_state() within a burst.
  41
  42 fqs_stutter     Wait time (in seconds) between consecutive bursts
  43                 of calls to force_quiescent_state().
  44
  45 irqreader       Says to invoke RCU readers from irq level.  This is currently
  46                 done via timers.  Defaults to "1" for variants of RCU that
  47                 permit this.  (Or, more accurately, variants of RCU that do
  48                 -not- permit this know to ignore this variable.)
  49
  50 nfakewriters    This is the number of RCU fake writer threads to run.  Fake
  51                 writer threads repeatedly use the synchronous "wait for
  52                 current readers" function of the interface selected by
  53                 torture_type, with a delay between calls to allow for various
  54                 different numbers of writers running in parallel.
  55                 nfakewriters defaults to 4, which provides enough parallelism
  56                 to trigger special cases caused by multiple writers, such as
  57                 the synchronize_srcu() early return optimization.
  58
  59 nreaders        This is the number of RCU reading threads supported.
  60                 The default is twice the number of CPUs.  Why twice?
  61                 To properly exercise RCU implementations with preemptible
  62                 read-side critical sections.
  63
  64 onoff_interval
  65                 The number of seconds between each attempt to execute a
  66                 randomly selected CPU-hotplug operation.  Defaults to
  67                 zero, which disables CPU hotplugging.  In HOTPLUG_CPU=n
  68                 kernels, rcutorture will silently refuse to do any
  69                 CPU-hotplug operations regardless of what value is
  70                 specified for onoff_interval.
  71
  72 onoff_holdoff   The number of seconds to wait until starting CPU-hotplug
  73                 operations.  This would normally only be used when
  74                 rcutorture was built into the kernel and started
  75                 automatically at boot time, in which case it is useful
  76                 in order to avoid confusing boot-time code with CPUs
  77                 coming and going.
  78
  79 shuffle_interval
  80                 The number of seconds to keep the test threads affinitied
  81                 to a particular subset of the CPUs, defaults to 3 seconds.
  82                 Used in conjunction with test_no_idle_hz.
  83
  84 shutdown_secs   The number of seconds to run the test before terminating
  85                 the test and powering off the system.  The default is
  86                 zero, which disables test termination and system shutdown.
  87                 This capability is useful for automated testing.
  88
  89 stall_cpu       The number of seconds that a CPU should be stalled while
  90                 within both an rcu_read_lock() and a preempt_disable().
  91                 This stall happens only once per rcutorture run.
  92                 If you need multiple stalls, use modprobe and rmmod to
  93                 repeatedly run rcutorture.  The default for stall_cpu
  94                 is zero, which prevents rcutorture from stalling a CPU.
  95
  96                 Note that attempts to rmmod rcutorture while the stall
  97                 is ongoing will hang, so be careful what value you
  98                 choose for this module parameter!  In addition, too-large
  99                 values for stall_cpu might well induce failures and
 100                 warnings in other parts of the kernel.  You have been
 101                 warned!
 102
 103 stall_cpu_holdoff
 104                 The number of seconds to wait after rcutorture starts
 105                 before stalling a CPU.  Defaults to 10 seconds.
 106
 107 stat_interval   The number of seconds between output of torture
 108                 statistics (via printk()).  Regardless of the interval,
 109                 statistics are printed when the module is unloaded.
 110                 Setting the interval to zero causes the statistics to
 111                 be printed -only- when the module is unloaded, and this
 112                 is the default.
 113
 114 stutter         The length of time to run the test before pausing for this
 115                 same period of time.  Defaults to "stutter=5", so as
 116                 to run and pause for (roughly) five-second intervals.
 117                 Specifying "stutter=0" causes the test to run continuously
 118                 without pausing, which is the old default behavior.
 119
 120 test_boost      Whether or not to test the ability of RCU to do priority
 121                 boosting.  Defaults to "test_boost=1", which performs
 122                 RCU priority-inversion testing only if the selected
 123                 RCU implementation supports priority boosting.  Specifying
 124                 "test_boost=0" never performs RCU priority-inversion
 125                 testing.  Specifying "test_boost=2" performs RCU
 126                 priority-inversion testing even if the selected RCU
 127                 implementation does not support RCU priority boosting,
 128                 which can be used to test rcutorture's ability to
 129                 carry out RCU priority-inversion testing.
 130
 131 test_boost_interval
 132                 The number of seconds in an RCU priority-inversion test
 133                 cycle.  Defaults to "test_boost_interval=7".  It is
 134                 usually wise for this value to be relatively prime to
 135                 the value selected for "stutter".
 136
 137 test_boost_duration
 138                 The number of seconds to do RCU priority-inversion testing
 139                 within any given "test_boost_interval".  Defaults to
 140                 "test_boost_duration=4".
 141
 142 test_no_idle_hz Whether or not to test the ability of RCU to operate in
 143                 a kernel that disables the scheduling-clock interrupt to
 144                 idle CPUs.  Boolean parameter, "1" to test, "0" otherwise.
 145                 Defaults to omitting this test.
 146
 147 torture_type    The type of RCU to test, with string values as follows:
 148
 149                 "rcu":  rcu_read_lock(), rcu_read_unlock() and call_rcu().
 150
 151                 "rcu_sync":  rcu_read_lock(), rcu_read_unlock(), and
 152                         synchronize_rcu().
 153
 154                 "rcu_expedited": rcu_read_lock(), rcu_read_unlock(), and
 155                         synchronize_rcu_expedited().
 156
 157                 "rcu_bh": rcu_read_lock_bh(), rcu_read_unlock_bh(), and
 158                         call_rcu_bh().
 159
 160                 "rcu_bh_sync": rcu_read_lock_bh(), rcu_read_unlock_bh(),
 161                         and synchronize_rcu_bh().
 162
 163                 "rcu_bh_expedited": rcu_read_lock_bh(), rcu_read_unlock_bh(),
 164                         and synchronize_rcu_bh_expedited().
 165
 166                 "srcu": srcu_read_lock(), srcu_read_unlock() and
 167                         synchronize_srcu().
 168
 169                 "srcu_expedited": srcu_read_lock(), srcu_read_unlock() and
 170                         synchronize_srcu_expedited().
 171
 172                 "sched": preempt_disable(), preempt_enable(), and
 173                         call_rcu_sched().
 174
 175                 "sched_sync": preempt_disable(), preempt_enable(), and
 176                         synchronize_sched().
 177
 178                 "sched_expedited": preempt_disable(), preempt_enable(), and
 179                         synchronize_sched_expedited().
 180
 181                 Defaults to "rcu".
 182
 183 verbose         Enable debug printk()s.  Default is disabled.
 184
 185
 186 OUTPUT
 187
 188 The statistics output is as follows:
 189
 190         rcu-torture:--- Start of test: nreaders=16 nfakewriters=4 stat_interval=30 verbose=0 test_no_idle_hz=1 shuffle_interval=3 stutter=5 irqreader=1 fqs_duration=0 fqs_holdoff=0 fqs_stutter=3 test_boost=1/0 test_boost_interval=7 test_boost_duration=4
 191         rcu-torture: rtc:           (null) ver: 155441 tfle: 0 rta: 155441 rtaf: 8884 rtf: 155440 rtmbe: 0 rtbke: 0 rtbre: 0 rtbf: 0 rtb: 0 nt: 3055767
 192         rcu-torture: Reader Pipe:  727860534 34213 0 0 0 0 0 0 0 0 0
 193         rcu-torture: Reader Batch:  727877838 17003 0 0 0 0 0 0 0 0 0
 194         rcu-torture: Free-Block Circulation:  155440 155440 155440 155440 155440 155440 155440 155440 155440 155440 0
 195         rcu-torture:--- End of test: SUCCESS: nreaders=16 nfakewriters=4 stat_interval=30 verbose=0 test_no_idle_hz=1 shuffle_interval=3 stutter=5 irqreader=1 fqs_duration=0 fqs_holdoff=0 fqs_stutter=3 test_boost=1/0 test_boost_interval=7 test_boost_duration=4
 196
 197 The command "dmesg | grep torture:" will extract this information on
 198 most systems.  On more esoteric configurations, it may be necessary to
 199 use other commands to access the output of the printk()s used by
 200 the RCU torture test.  The printk()s use KERN_ALERT, so they should
 201 be evident.  ;-)
 202
 203 The first and last lines show the rcutorture module parameters, and the
 204 last line shows either "SUCCESS" or "FAILURE", based on rcutorture's
 205 automatic determination as to whether RCU operated correctly.
 206
 207 The entries are as follows:
 208
 209 o       "rtc": The hexadecimal address of the structure currently visible
 210         to readers.
 211
 212 o       "ver": The number of times since boot that the RCU writer task
 213         has changed the structure visible to readers.
 214
 215 o       "tfle": If non-zero, indicates that the "torture freelist"
 216         containing structures to be placed into the "rtc" area is empty.
 217         This condition is important, since it can fool you into thinking
 218         that RCU is working when it is not.  :-/
 219
 220 o       "rta": Number of structures allocated from the torture freelist.
 221
 222 o       "rtaf": Number of allocations from the torture freelist that have
 223         failed due to the list being empty.  It is not unusual for this
 224         to be non-zero, but it is bad for it to be a large fraction of
 225         the value indicated by "rta".
 226
 227 o       "rtf": Number of frees into the torture freelist.
 228
 229 o       "rtmbe": A non-zero value indicates that rcutorture believes that
 230         rcu_assign_pointer() and rcu_dereference() are not working
 231         correctly.  This value should be zero.
 232
 233 o       "rtbke": rcutorture was unable to create the real-time kthreads
 234         used to force RCU priority inversion.  This value should be zero.
 235
 236 o       "rtbre": Although rcutorture successfully created the kthreads
 237         used to force RCU priority inversion, it was unable to set them
 238         to the real-time priority level of 1.  This value should be zero.
 239
 240 o       "rtbf": The number of times that RCU priority boosting failed
 241         to resolve RCU priority inversion.
 242
 243 o       "rtb": The number of times that rcutorture attempted to force
 244         an RCU priority inversion condition.  If you are testing RCU
 245         priority boosting via the "test_boost" module parameter, this
 246         value should be non-zero.
 247
 248 o       "nt": The number of times rcutorture ran RCU read-side code from
 249         within a timer handler.  This value should be non-zero only
 250         if you specified the "irqreader" module parameter.
 251
 252 o       "Reader Pipe": Histogram of "ages" of structures seen by readers.
 253         If any entries past the first two are non-zero, RCU is broken.
 254         And rcutorture prints the error flag string "!!!" to make sure
 255         you notice.  The age of a newly allocated structure is zero,
 256         it becomes one when removed from reader visibility, and is
 257         incremented once per grace period subsequently -- and is freed
 258         after passing through (RCU_TORTURE_PIPE_LEN-2) grace periods.
 259
 260         The output displayed above was taken from a correctly working
 261         RCU.  If you want to see what it looks like when broken, break
 262         it yourself.  ;-)
 263
 264 o       "Reader Batch": Another histogram of "ages" of structures seen
 265         by readers, but in terms of counter flips (or batches) rather
 266         than in terms of grace periods.  The legal number of non-zero
 267         entries is again two.  The reason for this separate view is that
 268         it is sometimes easier to get the third entry to show up in the
 269         "Reader Batch" list than in the "Reader Pipe" list.
 270
 271 o       "Free-Block Circulation": Shows the number of torture structures
 272         that have reached a given point in the pipeline.  The first element
 273         should closely correspond to the number of structures allocated,
 274         the second to the number that have been removed from reader view,
 275         and all but the last remaining to the corresponding number of
 276         passes through a grace period.  The last entry should be zero,
 277         as it is only incremented if a torture structure's counter
 278         somehow gets incremented farther than it should.
 279
 280 Different implementations of RCU can provide implementation-specific
 281 additional information.  For example, SRCU provides the following
 282 additional line:
 283
 284         srcu-torture: per-CPU(idx=1): 0(0,1) 1(0,1) 2(0,0) 3(0,1)
 285
 286 This line shows the per-CPU counter state.  The numbers in parentheses are
 287 the values of the "old" and "current" counters for the corresponding CPU.
 288 The "idx" value maps the "old" and "current" values to the underlying
 289 array, and is useful for debugging.
 290
 291
 292 USAGE
 293
 294 The following script may be used to torture RCU:
 295
 296         #!/bin/sh
 297
 298         modprobe rcutorture
 299         sleep 3600
 300         rmmod rcutorture
 301         dmesg | grep torture:
 302
 303 The output can be manually inspected for the error flag of "!!!".
 304 One could of course create a more elaborate script that automatically
 305 checked for such errors.  The "rmmod" command forces a "SUCCESS",
 306 "FAILURE", or "RCU_HOTPLUG" indication to be printk()ed.  The first
 307 two are self-explanatory, while the last indicates that while there
 308 were no RCU failures, CPU-hotplug problems were detected.