Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
awilliam
GitHub Repository: awilliam/linux-vfio
Path: blob/master/Documentation/RCU/checklist.txt
10821 views
1
Review Checklist for RCU Patches
2
3
4
This document contains a checklist for producing and reviewing patches
5
that make use of RCU. Violating any of the rules listed below will
6
result in the same sorts of problems that leaving out a locking primitive
7
would cause. This list is based on experiences reviewing such patches
8
over a rather long period of time, but improvements are always welcome!
9
10
0. Is RCU being applied to a read-mostly situation? If the data
11
structure is updated more than about 10% of the time, then you
12
should strongly consider some other approach, unless detailed
13
performance measurements show that RCU is nonetheless the right
14
tool for the job. Yes, RCU does reduce read-side overhead by
15
increasing write-side overhead, which is exactly why normal uses
16
of RCU will do much more reading than updating.
17
18
Another exception is where performance is not an issue, and RCU
19
provides a simpler implementation. An example of this situation
20
is the dynamic NMI code in the Linux 2.6 kernel, at least on
21
architectures where NMIs are rare.
22
23
Yet another exception is where the low real-time latency of RCU's
24
read-side primitives is critically important.
25
26
1. Does the update code have proper mutual exclusion?
27
28
RCU does allow -readers- to run (almost) naked, but -writers- must
29
still use some sort of mutual exclusion, such as:
30
31
a. locking,
32
b. atomic operations, or
33
c. restricting updates to a single task.
34
35
If you choose #b, be prepared to describe how you have handled
36
memory barriers on weakly ordered machines (pretty much all of
37
them -- even x86 allows later loads to be reordered to precede
38
earlier stores), and be prepared to explain why this added
39
complexity is worthwhile. If you choose #c, be prepared to
40
explain how this single task does not become a major bottleneck on
41
big multiprocessor machines (for example, if the task is updating
42
information relating to itself that other tasks can read, there
43
by definition can be no bottleneck).
44
45
2. Do the RCU read-side critical sections make proper use of
46
rcu_read_lock() and friends? These primitives are needed
47
to prevent grace periods from ending prematurely, which
48
could result in data being unceremoniously freed out from
49
under your read-side code, which can greatly increase the
50
actuarial risk of your kernel.
51
52
As a rough rule of thumb, any dereference of an RCU-protected
53
pointer must be covered by rcu_read_lock(), rcu_read_lock_bh(),
54
rcu_read_lock_sched(), or by the appropriate update-side lock.
55
Disabling of preemption can serve as rcu_read_lock_sched(), but
56
is less readable.
57
58
3. Does the update code tolerate concurrent accesses?
59
60
The whole point of RCU is to permit readers to run without
61
any locks or atomic operations. This means that readers will
62
be running while updates are in progress. There are a number
63
of ways to handle this concurrency, depending on the situation:
64
65
a. Use the RCU variants of the list and hlist update
66
primitives to add, remove, and replace elements on
67
an RCU-protected list. Alternatively, use the other
68
RCU-protected data structures that have been added to
69
the Linux kernel.
70
71
This is almost always the best approach.
72
73
b. Proceed as in (a) above, but also maintain per-element
74
locks (that are acquired by both readers and writers)
75
that guard per-element state. Of course, fields that
76
the readers refrain from accessing can be guarded by
77
some other lock acquired only by updaters, if desired.
78
79
This works quite well, also.
80
81
c. Make updates appear atomic to readers. For example,
82
pointer updates to properly aligned fields will
83
appear atomic, as will individual atomic primitives.
84
Sequences of perations performed under a lock will -not-
85
appear to be atomic to RCU readers, nor will sequences
86
of multiple atomic primitives.
87
88
This can work, but is starting to get a bit tricky.
89
90
d. Carefully order the updates and the reads so that
91
readers see valid data at all phases of the update.
92
This is often more difficult than it sounds, especially
93
given modern CPUs' tendency to reorder memory references.
94
One must usually liberally sprinkle memory barriers
95
(smp_wmb(), smp_rmb(), smp_mb()) through the code,
96
making it difficult to understand and to test.
97
98
It is usually better to group the changing data into
99
a separate structure, so that the change may be made
100
to appear atomic by updating a pointer to reference
101
a new structure containing updated values.
102
103
4. Weakly ordered CPUs pose special challenges. Almost all CPUs
104
are weakly ordered -- even x86 CPUs allow later loads to be
105
reordered to precede earlier stores. RCU code must take all of
106
the following measures to prevent memory-corruption problems:
107
108
a. Readers must maintain proper ordering of their memory
109
accesses. The rcu_dereference() primitive ensures that
110
the CPU picks up the pointer before it picks up the data
111
that the pointer points to. This really is necessary
112
on Alpha CPUs. If you don't believe me, see:
113
114
http://www.openvms.compaq.com/wizard/wiz_2637.html
115
116
The rcu_dereference() primitive is also an excellent
117
documentation aid, letting the person reading the code
118
know exactly which pointers are protected by RCU.
119
Please note that compilers can also reorder code, and
120
they are becoming increasingly aggressive about doing
121
just that. The rcu_dereference() primitive therefore
122
also prevents destructive compiler optimizations.
123
124
The rcu_dereference() primitive is used by the
125
various "_rcu()" list-traversal primitives, such
126
as the list_for_each_entry_rcu(). Note that it is
127
perfectly legal (if redundant) for update-side code to
128
use rcu_dereference() and the "_rcu()" list-traversal
129
primitives. This is particularly useful in code that
130
is common to readers and updaters. However, lockdep
131
will complain if you access rcu_dereference() outside
132
of an RCU read-side critical section. See lockdep.txt
133
to learn what to do about this.
134
135
Of course, neither rcu_dereference() nor the "_rcu()"
136
list-traversal primitives can substitute for a good
137
concurrency design coordinating among multiple updaters.
138
139
b. If the list macros are being used, the list_add_tail_rcu()
140
and list_add_rcu() primitives must be used in order
141
to prevent weakly ordered machines from misordering
142
structure initialization and pointer planting.
143
Similarly, if the hlist macros are being used, the
144
hlist_add_head_rcu() primitive is required.
145
146
c. If the list macros are being used, the list_del_rcu()
147
primitive must be used to keep list_del()'s pointer
148
poisoning from inflicting toxic effects on concurrent
149
readers. Similarly, if the hlist macros are being used,
150
the hlist_del_rcu() primitive is required.
151
152
The list_replace_rcu() and hlist_replace_rcu() primitives
153
may be used to replace an old structure with a new one
154
in their respective types of RCU-protected lists.
155
156
d. Rules similar to (4b) and (4c) apply to the "hlist_nulls"
157
type of RCU-protected linked lists.
158
159
e. Updates must ensure that initialization of a given
160
structure happens before pointers to that structure are
161
publicized. Use the rcu_assign_pointer() primitive
162
when publicizing a pointer to a structure that can
163
be traversed by an RCU read-side critical section.
164
165
5. If call_rcu(), or a related primitive such as call_rcu_bh() or
166
call_rcu_sched(), is used, the callback function must be
167
written to be called from softirq context. In particular,
168
it cannot block.
169
170
6. Since synchronize_rcu() can block, it cannot be called from
171
any sort of irq context. The same rule applies for
172
synchronize_rcu_bh(), synchronize_sched(), synchronize_srcu(),
173
synchronize_rcu_expedited(), synchronize_rcu_bh_expedited(),
174
synchronize_sched_expedite(), and synchronize_srcu_expedited().
175
176
The expedited forms of these primitives have the same semantics
177
as the non-expedited forms, but expediting is both expensive
178
and unfriendly to real-time workloads. Use of the expedited
179
primitives should be restricted to rare configuration-change
180
operations that would not normally be undertaken while a real-time
181
workload is running.
182
183
7. If the updater uses call_rcu() or synchronize_rcu(), then the
184
corresponding readers must use rcu_read_lock() and
185
rcu_read_unlock(). If the updater uses call_rcu_bh() or
186
synchronize_rcu_bh(), then the corresponding readers must
187
use rcu_read_lock_bh() and rcu_read_unlock_bh(). If the
188
updater uses call_rcu_sched() or synchronize_sched(), then
189
the corresponding readers must disable preemption, possibly
190
by calling rcu_read_lock_sched() and rcu_read_unlock_sched().
191
If the updater uses synchronize_srcu(), the the corresponding
192
readers must use srcu_read_lock() and srcu_read_unlock(),
193
and with the same srcu_struct. The rules for the expedited
194
primitives are the same as for their non-expedited counterparts.
195
Mixing things up will result in confusion and broken kernels.
196
197
One exception to this rule: rcu_read_lock() and rcu_read_unlock()
198
may be substituted for rcu_read_lock_bh() and rcu_read_unlock_bh()
199
in cases where local bottom halves are already known to be
200
disabled, for example, in irq or softirq context. Commenting
201
such cases is a must, of course! And the jury is still out on
202
whether the increased speed is worth it.
203
204
8. Although synchronize_rcu() is slower than is call_rcu(), it
205
usually results in simpler code. So, unless update performance
206
is critically important or the updaters cannot block,
207
synchronize_rcu() should be used in preference to call_rcu().
208
209
An especially important property of the synchronize_rcu()
210
primitive is that it automatically self-limits: if grace periods
211
are delayed for whatever reason, then the synchronize_rcu()
212
primitive will correspondingly delay updates. In contrast,
213
code using call_rcu() should explicitly limit update rate in
214
cases where grace periods are delayed, as failing to do so can
215
result in excessive realtime latencies or even OOM conditions.
216
217
Ways of gaining this self-limiting property when using call_rcu()
218
include:
219
220
a. Keeping a count of the number of data-structure elements
221
used by the RCU-protected data structure, including
222
those waiting for a grace period to elapse. Enforce a
223
limit on this number, stalling updates as needed to allow
224
previously deferred frees to complete. Alternatively,
225
limit only the number awaiting deferred free rather than
226
the total number of elements.
227
228
One way to stall the updates is to acquire the update-side
229
mutex. (Don't try this with a spinlock -- other CPUs
230
spinning on the lock could prevent the grace period
231
from ever ending.) Another way to stall the updates
232
is for the updates to use a wrapper function around
233
the memory allocator, so that this wrapper function
234
simulates OOM when there is too much memory awaiting an
235
RCU grace period. There are of course many other
236
variations on this theme.
237
238
b. Limiting update rate. For example, if updates occur only
239
once per hour, then no explicit rate limiting is required,
240
unless your system is already badly broken. The dcache
241
subsystem takes this approach -- updates are guarded
242
by a global lock, limiting their rate.
243
244
c. Trusted update -- if updates can only be done manually by
245
superuser or some other trusted user, then it might not
246
be necessary to automatically limit them. The theory
247
here is that superuser already has lots of ways to crash
248
the machine.
249
250
d. Use call_rcu_bh() rather than call_rcu(), in order to take
251
advantage of call_rcu_bh()'s faster grace periods.
252
253
e. Periodically invoke synchronize_rcu(), permitting a limited
254
number of updates per grace period.
255
256
The same cautions apply to call_rcu_bh() and call_rcu_sched().
257
258
9. All RCU list-traversal primitives, which include
259
rcu_dereference(), list_for_each_entry_rcu(),
260
list_for_each_continue_rcu(), and list_for_each_safe_rcu(),
261
must be either within an RCU read-side critical section or
262
must be protected by appropriate update-side locks. RCU
263
read-side critical sections are delimited by rcu_read_lock()
264
and rcu_read_unlock(), or by similar primitives such as
265
rcu_read_lock_bh() and rcu_read_unlock_bh(), in which case
266
the matching rcu_dereference() primitive must be used in order
267
to keep lockdep happy, in this case, rcu_dereference_bh().
268
269
The reason that it is permissible to use RCU list-traversal
270
primitives when the update-side lock is held is that doing so
271
can be quite helpful in reducing code bloat when common code is
272
shared between readers and updaters. Additional primitives
273
are provided for this case, as discussed in lockdep.txt.
274
275
10. Conversely, if you are in an RCU read-side critical section,
276
and you don't hold the appropriate update-side lock, you -must-
277
use the "_rcu()" variants of the list macros. Failing to do so
278
will break Alpha, cause aggressive compilers to generate bad code,
279
and confuse people trying to read your code.
280
281
11. Note that synchronize_rcu() -only- guarantees to wait until
282
all currently executing rcu_read_lock()-protected RCU read-side
283
critical sections complete. It does -not- necessarily guarantee
284
that all currently running interrupts, NMIs, preempt_disable()
285
code, or idle loops will complete. Therefore, if you do not have
286
rcu_read_lock()-protected read-side critical sections, do -not-
287
use synchronize_rcu().
288
289
Similarly, disabling preemption is not an acceptable substitute
290
for rcu_read_lock(). Code that attempts to use preemption
291
disabling where it should be using rcu_read_lock() will break
292
in real-time kernel builds.
293
294
If you want to wait for interrupt handlers, NMI handlers, and
295
code under the influence of preempt_disable(), you instead
296
need to use synchronize_irq() or synchronize_sched().
297
298
12. Any lock acquired by an RCU callback must be acquired elsewhere
299
with softirq disabled, e.g., via spin_lock_irqsave(),
300
spin_lock_bh(), etc. Failing to disable irq on a given
301
acquisition of that lock will result in deadlock as soon as
302
the RCU softirq handler happens to run your RCU callback while
303
interrupting that acquisition's critical section.
304
305
13. RCU callbacks can be and are executed in parallel. In many cases,
306
the callback code simply wrappers around kfree(), so that this
307
is not an issue (or, more accurately, to the extent that it is
308
an issue, the memory-allocator locking handles it). However,
309
if the callbacks do manipulate a shared data structure, they
310
must use whatever locking or other synchronization is required
311
to safely access and/or modify that data structure.
312
313
RCU callbacks are -usually- executed on the same CPU that executed
314
the corresponding call_rcu(), call_rcu_bh(), or call_rcu_sched(),
315
but are by -no- means guaranteed to be. For example, if a given
316
CPU goes offline while having an RCU callback pending, then that
317
RCU callback will execute on some surviving CPU. (If this was
318
not the case, a self-spawning RCU callback would prevent the
319
victim CPU from ever going offline.)
320
321
14. SRCU (srcu_read_lock(), srcu_read_unlock(), srcu_dereference(),
322
synchronize_srcu(), and synchronize_srcu_expedited()) may only
323
be invoked from process context. Unlike other forms of RCU, it
324
-is- permissible to block in an SRCU read-side critical section
325
(demarked by srcu_read_lock() and srcu_read_unlock()), hence the
326
"SRCU": "sleepable RCU". Please note that if you don't need
327
to sleep in read-side critical sections, you should be using
328
RCU rather than SRCU, because RCU is almost always faster and
329
easier to use than is SRCU.
330
331
Also unlike other forms of RCU, explicit initialization
332
and cleanup is required via init_srcu_struct() and
333
cleanup_srcu_struct(). These are passed a "struct srcu_struct"
334
that defines the scope of a given SRCU domain. Once initialized,
335
the srcu_struct is passed to srcu_read_lock(), srcu_read_unlock()
336
synchronize_srcu(), and synchronize_srcu_expedited(). A given
337
synchronize_srcu() waits only for SRCU read-side critical
338
sections governed by srcu_read_lock() and srcu_read_unlock()
339
calls that have been passed the same srcu_struct. This property
340
is what makes sleeping read-side critical sections tolerable --
341
a given subsystem delays only its own updates, not those of other
342
subsystems using SRCU. Therefore, SRCU is less prone to OOM the
343
system than RCU would be if RCU's read-side critical sections
344
were permitted to sleep.
345
346
The ability to sleep in read-side critical sections does not
347
come for free. First, corresponding srcu_read_lock() and
348
srcu_read_unlock() calls must be passed the same srcu_struct.
349
Second, grace-period-detection overhead is amortized only
350
over those updates sharing a given srcu_struct, rather than
351
being globally amortized as they are for other forms of RCU.
352
Therefore, SRCU should be used in preference to rw_semaphore
353
only in extremely read-intensive situations, or in situations
354
requiring SRCU's read-side deadlock immunity or low read-side
355
realtime latency.
356
357
Note that, rcu_assign_pointer() relates to SRCU just as they do
358
to other forms of RCU.
359
360
15. The whole point of call_rcu(), synchronize_rcu(), and friends
361
is to wait until all pre-existing readers have finished before
362
carrying out some otherwise-destructive operation. It is
363
therefore critically important to -first- remove any path
364
that readers can follow that could be affected by the
365
destructive operation, and -only- -then- invoke call_rcu(),
366
synchronize_rcu(), or friends.
367
368
Because these primitives only wait for pre-existing readers, it
369
is the caller's responsibility to guarantee that any subsequent
370
readers will execute safely.
371
372
16. The various RCU read-side primitives do -not- necessarily contain
373
memory barriers. You should therefore plan for the CPU
374
and the compiler to freely reorder code into and out of RCU
375
read-side critical sections. It is the responsibility of the
376
RCU update-side primitives to deal with this.
377
378
17. Use CONFIG_PROVE_RCU, CONFIG_DEBUG_OBJECTS_RCU_HEAD, and
379
the __rcu sparse checks to validate your RCU code. These
380
can help find problems as follows:
381
382
CONFIG_PROVE_RCU: check that accesses to RCU-protected data
383
structures are carried out under the proper RCU
384
read-side critical section, while holding the right
385
combination of locks, or whatever other conditions
386
are appropriate.
387
388
CONFIG_DEBUG_OBJECTS_RCU_HEAD: check that you don't pass the
389
same object to call_rcu() (or friends) before an RCU
390
grace period has elapsed since the last time that you
391
passed that same object to call_rcu() (or friends).
392
393
__rcu sparse checks: tag the pointer to the RCU-protected data
394
structure with __rcu, and sparse will warn you if you
395
access that pointer without the services of one of the
396
variants of rcu_dereference().
397
398
These debugging aids can help you find problems that are
399
otherwise extremely difficult to spot.
400
401