1 .. _whatisrcu_doc:
2
3 What is RCU? -- "Read, Copy, Update"
4 ======================================
5
6 Please note that the "What is RCU?" LWN series is an excellent place
7 to start learning about RCU:
8
9 | 1. What is RCU, Fundamentally? https://lwn.net/Articles/262464/
10 | 2. What is RCU? Part 2: Usage https://lwn.net/Articles/263130/
11 | 3. RCU part 3: the RCU API https://lwn.net/Articles/264090/
12 | 4. The RCU API, 2010 Edition https://lwn.net/Articles/418853/
13 | 2010 Big API Table https://lwn.net/Articles/419086/
14 | 5. The RCU API, 2014 Edition https://lwn.net/Articles/609904/
15 | 2014 Big API Table https://lwn.net/Articles/609973/
16 | 6. The RCU API, 2019 Edition https://lwn.net/Articles/777036/
17 | 2019 Big API Table https://lwn.net/Articles/777165/
18
19 For those preferring video:
20
21 | 1. Unraveling RCU Mysteries: Fundamentals https://www.linuxfoundation.org/webinars/unraveling-rcu-usage-mysteries
22 | 2. Unraveling RCU Mysteries: Additional Use Cases https://www.linuxfoundation.org/webinars/unraveling-rcu-usage-mysteries-additional-use-cases
23
24
25 What is RCU?
26
27 RCU is a synchronization mechanism that was added to the Linux kernel
28 during the 2.5 development effort that is optimized for read-mostly
29 situations. Although RCU is actually quite simple, making effective use
30 of it requires you to think differently about your code. Another part
31 of the problem is the mistaken assumption that there is "one true way" to
32 describe and to use RCU. Instead, the experience has been that different
33 people must take different paths to arrive at an understanding of RCU,
34 depending on their experiences and use cases. This document provides
35 several different paths, as follows:
36
37 :ref:`1. RCU OVERVIEW <1_whatisRCU>`
38
39 :ref:`2. WHAT IS RCU'S CORE API? <2_whatisRCU>`
40
41 :ref:`3. WHAT ARE SOME EXAMPLE USES OF CORE RCU API? <3_whatisRCU>`
42
43 :ref:`4. WHAT IF MY UPDATING THREAD CANNOT BLOCK? <4_whatisRCU>`
44
45 :ref:`5. WHAT ARE SOME SIMPLE IMPLEMENTATIONS OF RCU? <5_whatisRCU>`
46
47 :ref:`6. ANALOGY WITH READER-WRITER LOCKING <6_whatisRCU>`
48
49 :ref:`7. ANALOGY WITH REFERENCE COUNTING <7_whatisRCU>`
50
51 :ref:`8. FULL LIST OF RCU APIs <8_whatisRCU>`
52
53 :ref:`9. ANSWERS TO QUICK QUIZZES <9_whatisRCU>`
54
55 People who prefer starting with a conceptual overview should focus on
56 Section 1, though most readers will profit by reading this section at
57 some point. People who prefer to start with an API that they can then
58 experiment with should focus on Section 2. People who prefer to start
59 with example uses should focus on Sections 3 and 4. People who need to
60 understand the RCU implementation should focus on Section 5, then dive
61 into the kernel source code. People who reason best by analogy should
62 focus on Section 6 and 7. Section 8 serves as an index to the docbook
63 API documentation, and Section 9 is the traditional answer key.
64
65 So, start with the section that makes the most sense to you and your
66 preferred method of learning. If you need to know everything about
67 everything, feel free to read the whole thing -- but if you are really
68 that type of person, you have perused the source code and will therefore
69 never need this document anyway. ;-)
70
71 .. _1_whatisRCU:
72
73 1. RCU OVERVIEW
74 ----------------
75
76 The basic idea behind RCU is to split updates into "removal" and
77 "reclamation" phases. The removal phase removes references to data items
78 within a data structure (possibly by replacing them with references to
79 new versions of these data items), and can run concurrently with readers.
80 The reason that it is safe to run the removal phase concurrently with
81 readers is the semantics of modern CPUs guarantee that readers will see
82 either the old or the new version of the data structure rather than a
83 partially updated reference. The reclamation phase does the work of reclaiming
84 (e.g., freeing) the data items removed from the data structure during the
85 removal phase. Because reclaiming data items can disrupt any readers
86 concurrently referencing those data items, the reclamation phase must
87 not start until readers no longer hold references to those data items.
88
89 Splitting the update into removal and reclamation phases permits the
90 updater to perform the removal phase immediately, and to defer the
91 reclamation phase until all readers active during the removal phase have
92 completed, either by blocking until they finish or by registering a
93 callback that is invoked after they finish. Only readers that are active
94 during the removal phase need be considered, because any reader starting
95 after the removal phase will be unable to gain a reference to the removed
96 data items, and therefore cannot be disrupted by the reclamation phase.
97
98 So the typical RCU update sequence goes something like the following:
99
100 a. Remove pointers to a data structure, so that subsequent
101 readers cannot gain a reference to it.
102
103 b. Wait for all previous readers to complete their RCU read-side
104 critical sections.
105
106 c. At this point, there cannot be any readers who hold references
107 to the data structure, so it now may safely be reclaimed
108 (e.g., kfree()d).
109
110 Step (b) above is the key idea underlying RCU's deferred destruction.
111 The ability to wait until all readers are done allows RCU readers to
112 use much lighter-weight synchronization, in some cases, absolutely no
113 synchronization at all. In contrast, in more conventional lock-based
114 schemes, readers must use heavy-weight synchronization in order to
115 prevent an updater from deleting the data structure out from under them.
116 This is because lock-based updaters typically update data items in place,
117 and must therefore exclude readers. In contrast, RCU-based updaters
118 typically take advantage of the fact that writes to single aligned
119 pointers are atomic on modern CPUs, allowing atomic insertion, removal,
120 and replacement of data items in a linked structure without disrupting
121 readers. Concurrent RCU readers can then continue accessing the old
122 versions, and can dispense with the atomic operations, memory barriers,
123 and communications cache misses that are so expensive on present-day
124 SMP computer systems, even in absence of lock contention.
125
126 In the three-step procedure shown above, the updater is performing both
127 the removal and the reclamation step, but it is often helpful for an
128 entirely different thread to do the reclamation, as is in fact the case
129 in the Linux kernel's directory-entry cache (dcache). Even if the same
130 thread performs both the update step (step (a) above) and the reclamation
131 step (step (c) above), it is often helpful to think of them separately.
132 For example, RCU readers and updaters need not communicate at all,
133 but RCU provides implicit low-overhead communication between readers
134 and reclaimers, namely, in step (b) above.
135
136 So how the heck can a reclaimer tell when a reader is done, given
137 that readers are not doing any sort of synchronization operations???
138 Read on to learn about how RCU's API makes this easy.
139
140 .. _2_whatisRCU:
141
142 2. WHAT IS RCU'S CORE API?
143 ---------------------------
144
145 The core RCU API is quite small:
146
147 a. rcu_read_lock()
148 b. rcu_read_unlock()
149 c. synchronize_rcu() / call_rcu()
150 d. rcu_assign_pointer()
151 e. rcu_dereference()
152
153 There are many other members of the RCU API, but the rest can be
154 expressed in terms of these five, though most implementations instead
155 express synchronize_rcu() in terms of the call_rcu() callback API.
156
157 The five core RCU APIs are described below, the other 18 will be enumerated
158 later. See the kernel docbook documentation for more info, or look directly
159 at the function header comments.
160
161 rcu_read_lock()
162 ^^^^^^^^^^^^^^^
163 void rcu_read_lock(void);
164
165 This temporal primitive is used by a reader to inform the
166 reclaimer that the reader is entering an RCU read-side critical
167 section. It is illegal to block while in an RCU read-side
168 critical section, though kernels built with CONFIG_PREEMPT_RCU
169 can preempt RCU read-side critical sections. Any RCU-protected
170 data structure accessed during an RCU read-side critical section
171 is guaranteed to remain unreclaimed for the full duration of that
172 critical section. Reference counts may be used in conjunction
173 with RCU to maintain longer-term references to data structures.
174
175 Note that anything that disables bottom halves, preemption,
176 or interrupts also enters an RCU read-side critical section.
177 Acquiring a spinlock also enters an RCU read-side critical
178 sections, even for spinlocks that do not disable preemption,
179 as is the case in kernels built with CONFIG_PREEMPT_RT=y.
180 Sleeplocks do *not* enter RCU read-side critical sections.
181
182 rcu_read_unlock()
183 ^^^^^^^^^^^^^^^^^
184 void rcu_read_unlock(void);
185
186 This temporal primitives is used by a reader to inform the
187 reclaimer that the reader is exiting an RCU read-side critical
188 section. Anything that enables bottom halves, preemption,
189 or interrupts also exits an RCU read-side critical section.
190 Releasing a spinlock also exits an RCU read-side critical section.
191
192 Note that RCU read-side critical sections may be nested and/or
193 overlapping.
194
195 synchronize_rcu()
196 ^^^^^^^^^^^^^^^^^
197 void synchronize_rcu(void);
198
199 This temporal primitive marks the end of updater code and the
200 beginning of reclaimer code. It does this by blocking until
201 all pre-existing RCU read-side critical sections on all CPUs
202 have completed. Note that synchronize_rcu() will **not**
203 necessarily wait for any subsequent RCU read-side critical
204 sections to complete. For example, consider the following
205 sequence of events::
206
207 CPU 0 CPU 1 CPU 2
208 ----------------- ------------------------- ---------------
209 1. rcu_read_lock()
210 2. enters synchronize_rcu()
211 3. rcu_read_lock()
212 4. rcu_read_unlock()
213 5. exits synchronize_rcu()
214 6. rcu_read_unlock()
215
216 To reiterate, synchronize_rcu() waits only for ongoing RCU
217 read-side critical sections to complete, not necessarily for
218 any that begin after synchronize_rcu() is invoked.
219
220 Of course, synchronize_rcu() does not necessarily return
221 **immediately** after the last pre-existing RCU read-side critical
222 section completes. For one thing, there might well be scheduling
223 delays. For another thing, many RCU implementations process
224 requests in batches in order to improve efficiencies, which can
225 further delay synchronize_rcu().
226
227 Since synchronize_rcu() is the API that must figure out when
228 readers are done, its implementation is key to RCU. For RCU
229 to be useful in all but the most read-intensive situations,
230 synchronize_rcu()'s overhead must also be quite small.
231
232 The call_rcu() API is an asynchronous callback form of
233 synchronize_rcu(), and is described in more detail in a later
234 section. Instead of blocking, it registers a function and
235 argument which are invoked after all ongoing RCU read-side
236 critical sections have completed. This callback variant is
237 particularly useful in situations where it is illegal to block
238 or where update-side performance is critically important.
239
240 However, the call_rcu() API should not be used lightly, as use
241 of the synchronize_rcu() API generally results in simpler code.
242 In addition, the synchronize_rcu() API has the nice property
243 of automatically limiting update rate should grace periods
244 be delayed. This property results in system resilience in face
245 of denial-of-service attacks. Code using call_rcu() should limit
246 update rate in order to gain this same sort of resilience. See
247 checklist.rst for some approaches to limiting the update rate.
248
249 rcu_assign_pointer()
250 ^^^^^^^^^^^^^^^^^^^^
251 void rcu_assign_pointer(p, typeof(p) v);
252
253 Yes, rcu_assign_pointer() **is** implemented as a macro, though
254 it would be cool to be able to declare a function in this manner.
255 (And there has been some discussion of adding overloaded functions
256 to the C language, so who knows?)
257
258 The updater uses this spatial macro to assign a new value to an
259 RCU-protected pointer, in order to safely communicate the change
260 in value from the updater to the reader. This is a spatial (as
261 opposed to temporal) macro. It does not evaluate to an rvalue,
262 but it does provide any compiler directives and memory-barrier
263 instructions required for a given compile or CPU architecture.
264 Its ordering properties are that of a store-release operation,
265 that is, any prior loads and stores required to initialize the
266 structure are ordered before the store that publishes the pointer
267 to that structure.
268
269 Perhaps just as important, rcu_assign_pointer() serves to document
270 (1) which pointers are protected by RCU and (2) the point at which
271 a given structure becomes accessible to other CPUs. That said,
272 rcu_assign_pointer() is most frequently used indirectly, via
273 the _rcu list-manipulation primitives such as list_add_rcu().
274
275 rcu_dereference()
276 ^^^^^^^^^^^^^^^^^
277 typeof(p) rcu_dereference(p);
278
279 Like rcu_assign_pointer(), rcu_dereference() must be implemented
280 as a macro.
281
282 The reader uses the spatial rcu_dereference() macro to fetch
283 an RCU-protected pointer, which returns a value that may
284 then be safely dereferenced. Note that rcu_dereference()
285 does not actually dereference the pointer, instead, it
286 protects the pointer for later dereferencing. It also
287 executes any needed memory-barrier instructions for a given
288 CPU architecture. Currently, only Alpha needs memory barriers
289 within rcu_dereference() -- on other CPUs, it compiles to a
290 volatile load. However, no mainstream C compilers respect
291 address dependencies, so rcu_dereference() uses volatile casts,
292 which, in combination with the coding guidelines listed in
293 rcu_dereference.rst, prevent current compilers from breaking
294 these dependencies.
295
296 Common coding practice uses rcu_dereference() to copy an
297 RCU-protected pointer to a local variable, then dereferences
298 this local variable, for example as follows::
299
300 p = rcu_dereference(head.next);
301 return p->data;
302
303 However, in this case, one could just as easily combine these
304 into one statement::
305
306 return rcu_dereference(head.next)->data;
307
308 If you are going to be fetching multiple fields from the
309 RCU-protected structure, using the local variable is of
310 course preferred. Repeated rcu_dereference() calls look
311 ugly, do not guarantee that the same pointer will be returned
312 if an update happened while in the critical section, and incur
313 unnecessary overhead on Alpha CPUs.
314
315 Note that the value returned by rcu_dereference() is valid
316 only within the enclosing RCU read-side critical section [1]_.
317 For example, the following is **not** legal::
318
319 rcu_read_lock();
320 p = rcu_dereference(head.next);
321 rcu_read_unlock();
322 x = p->address; /* BUG!!! */
323 rcu_read_lock();
324 y = p->data; /* BUG!!! */
325 rcu_read_unlock();
326
327 Holding a reference from one RCU read-side critical section
328 to another is just as illegal as holding a reference from
329 one lock-based critical section to another! Similarly,
330 using a reference outside of the critical section in which
331 it was acquired is just as illegal as doing so with normal
332 locking.
333
334 As with rcu_assign_pointer(), an important function of
335 rcu_dereference() is to document which pointers are protected by
336 RCU, in particular, flagging a pointer that is subject to changing
337 at any time, including immediately after the rcu_dereference().
338 And, again like rcu_assign_pointer(), rcu_dereference() is
339 typically used indirectly, via the _rcu list-manipulation
340 primitives, such as list_for_each_entry_rcu() [2]_.
341
342 .. [1] The variant rcu_dereference_protected() can be used outside
343 of an RCU read-side critical section as long as the usage is
344 protected by locks acquired by the update-side code. This variant
345 avoids the lockdep warning that would happen when using (for
346 example) rcu_dereference() without rcu_read_lock() protection.
347 Using rcu_dereference_protected() also has the advantage
348 of permitting compiler optimizations that rcu_dereference()
349 must prohibit. The rcu_dereference_protected() variant takes
350 a lockdep expression to indicate which locks must be acquired
351 by the caller. If the indicated protection is not provided,
352 a lockdep splat is emitted. See Design/Requirements/Requirements.rst
353 and the API's code comments for more details and example usage.
354
355 .. [2] If the list_for_each_entry_rcu() instance might be used by
356 update-side code as well as by RCU readers, then an additional
357 lockdep expression can be added to its list of arguments.
358 For example, given an additional "lock_is_held(&mylock)" argument,
359 the RCU lockdep code would complain only if this instance was
360 invoked outside of an RCU read-side critical section and without
361 the protection of mylock.
362
363 The following diagram shows how each API communicates among the
364 reader, updater, and reclaimer.
365 ::
366
367
368 rcu_assign_pointer()
369 +--------+
370 +---------------------->| reader |---------+
371 | +--------+ |
372 | | |
373 | | | Protect:
374 | | | rcu_read_lock()
375 | | | rcu_read_unlock()
376 | rcu_dereference() | |
377 +---------+ | |
378 | updater |<----------------+ |
379 +---------+ V
380 | +-----------+
381 +----------------------------------->| reclaimer |
382 +-----------+
383 Defer:
384 synchronize_rcu() & call_rcu()
385
386
387 The RCU infrastructure observes the temporal sequence of rcu_read_lock(),
388 rcu_read_unlock(), synchronize_rcu(), and call_rcu() invocations in
389 order to determine when (1) synchronize_rcu() invocations may return
390 to their callers and (2) call_rcu() callbacks may be invoked. Efficient
391 implementations of the RCU infrastructure make heavy use of batching in
392 order to amortize their overhead over many uses of the corresponding APIs.
393 The rcu_assign_pointer() and rcu_dereference() invocations communicate
394 spatial changes via stores to and loads from the RCU-protected pointer in
395 question.
396
397 There are at least three flavors of RCU usage in the Linux kernel. The diagram
398 above shows the most common one. On the updater side, the rcu_assign_pointer(),
399 synchronize_rcu() and call_rcu() primitives used are the same for all three
400 flavors. However for protection (on the reader side), the primitives used vary
401 depending on the flavor:
402
403 a. rcu_read_lock() / rcu_read_unlock()
404 rcu_dereference()
405
406 b. rcu_read_lock_bh() / rcu_read_unlock_bh()
407 local_bh_disable() / local_bh_enable()
408 rcu_dereference_bh()
409
410 c. rcu_read_lock_sched() / rcu_read_unlock_sched()
411 preempt_disable() / preempt_enable()
412 local_irq_save() / local_irq_restore()
413 hardirq enter / hardirq exit
414 NMI enter / NMI exit
415 rcu_dereference_sched()
416
417 These three flavors are used as follows:
418
419 a. RCU applied to normal data structures.
420
421 b. RCU applied to networking data structures that may be subjected
422 to remote denial-of-service attacks.
423
424 c. RCU applied to scheduler and interrupt/NMI-handler tasks.
425
426 Again, most uses will be of (a). The (b) and (c) cases are important
427 for specialized uses, but are relatively uncommon. The SRCU, RCU-Tasks,
428 RCU-Tasks-Rude, and RCU-Tasks-Trace have similar relationships among
429 their assorted primitives.
430
431 .. _3_whatisRCU:
432
433 3. WHAT ARE SOME EXAMPLE USES OF CORE RCU API?
434 -----------------------------------------------
435
436 This section shows a simple use of the core RCU API to protect a
437 global pointer to a dynamically allocated structure. More-typical
438 uses of RCU may be found in listRCU.rst and NMI-RCU.rst.
439 ::
440
441 struct foo {
442 int a;
443 char b;
444 long c;
445 };
446 DEFINE_SPINLOCK(foo_mutex);
447
448 struct foo __rcu *gbl_foo;
449
450 /*
451 * Create a new struct foo that is the same as the one currently
452 * pointed to by gbl_foo, except that field "a" is replaced
453 * with "new_a". Points gbl_foo to the new structure, and
454 * frees up the old structure after a grace period.
455 *
456 * Uses rcu_assign_pointer() to ensure that concurrent readers
457 * see the initialized version of the new structure.
458 *
459 * Uses synchronize_rcu() to ensure that any readers that might
460 * have references to the old structure complete before freeing
461 * the old structure.
462 */
463 void foo_update_a(int new_a)
464 {
465 struct foo *new_fp;
466 struct foo *old_fp;
467
468 new_fp = kmalloc(sizeof(*new_fp), GFP_KERNEL);
469 spin_lock(&foo_mutex);
470 old_fp = rcu_dereference_protected(gbl_foo, lockdep_is_held(&foo_mutex));
471 *new_fp = *old_fp;
472 new_fp->a = new_a;
473 rcu_assign_pointer(gbl_foo, new_fp);
474 spin_unlock(&foo_mutex);
475 synchronize_rcu();
476 kfree(old_fp);
477 }
478
479 /*
480 * Return the value of field "a" of the current gbl_foo
481 * structure. Use rcu_read_lock() and rcu_read_unlock()
482 * to ensure that the structure does not get deleted out
483 * from under us, and use rcu_dereference() to ensure that
484 * we see the initialized version of the structure (important
485 * for DEC Alpha and for people reading the code).
486 */
487 int foo_get_a(void)
488 {
489 int retval;
490
491 rcu_read_lock();
492 retval = rcu_dereference(gbl_foo)->a;
493 rcu_read_unlock();
494 return retval;
495 }
496
497 So, to sum up:
498
499 - Use rcu_read_lock() and rcu_read_unlock() to guard RCU
500 read-side critical sections.
501
502 - Within an RCU read-side critical section, use rcu_dereference()
503 to dereference RCU-protected pointers.
504
505 - Use some solid design (such as locks or semaphores) to
506 keep concurrent updates from interfering with each other.
507
508 - Use rcu_assign_pointer() to update an RCU-protected pointer.
509 This primitive protects concurrent readers from the updater,
510 **not** concurrent updates from each other! You therefore still
511 need to use locking (or something similar) to keep concurrent
512 rcu_assign_pointer() primitives from interfering with each other.
513
514 - Use synchronize_rcu() **after** removing a data element from an
515 RCU-protected data structure, but **before** reclaiming/freeing
516 the data element, in order to wait for the completion of all
517 RCU read-side critical sections that might be referencing that
518 data item.
519
520 See checklist.rst for additional rules to follow when using RCU.
521 And again, more-typical uses of RCU may be found in listRCU.rst
522 and NMI-RCU.rst.
523
524 .. _4_whatisRCU:
525
526 4. WHAT IF MY UPDATING THREAD CANNOT BLOCK?
527 --------------------------------------------
528
529 In the example above, foo_update_a() blocks until a grace period elapses.
530 This is quite simple, but in some cases one cannot afford to wait so
531 long -- there might be other high-priority work to be done.
532
533 In such cases, one uses call_rcu() rather than synchronize_rcu().
534 The call_rcu() API is as follows::
535
536 void call_rcu(struct rcu_head *head, rcu_callback_t func);
537
538 This function invokes func(head) after a grace period has elapsed.
539 This invocation might happen from either softirq or process context,
540 so the function is not permitted to block. The foo struct needs to
541 have an rcu_head structure added, perhaps as follows::
542
543 struct foo {
544 int a;
545 char b;
546 long c;
547 struct rcu_head rcu;
548 };
549
550 The foo_update_a() function might then be written as follows::
551
552 /*
553 * Create a new struct foo that is the same as the one currently
554 * pointed to by gbl_foo, except that field "a" is replaced
555 * with "new_a". Points gbl_foo to the new structure, and
556 * frees up the old structure after a grace period.
557 *
558 * Uses rcu_assign_pointer() to ensure that concurrent readers
559 * see the initialized version of the new structure.
560 *
561 * Uses call_rcu() to ensure that any readers that might have
562 * references to the old structure complete before freeing the
563 * old structure.
564 */
565 void foo_update_a(int new_a)
566 {
567 struct foo *new_fp;
568 struct foo *old_fp;
569
570 new_fp = kmalloc(sizeof(*new_fp), GFP_KERNEL);
571 spin_lock(&foo_mutex);
572 old_fp = rcu_dereference_protected(gbl_foo, lockdep_is_held(&foo_mutex));
573 *new_fp = *old_fp;
574 new_fp->a = new_a;
575 rcu_assign_pointer(gbl_foo, new_fp);
576 spin_unlock(&foo_mutex);
577 call_rcu(&old_fp->rcu, foo_reclaim);
578 }
579
580 The foo_reclaim() function might appear as follows::
581
582 void foo_reclaim(struct rcu_head *rp)
583 {
584 struct foo *fp = container_of(rp, struct foo, rcu);
585
586 foo_cleanup(fp->a);
587
588 kfree(fp);
589 }
590
591 The container_of() primitive is a macro that, given a pointer into a
592 struct, the type of the struct, and the pointed-to field within the
593 struct, returns a pointer to the beginning of the struct.
594
595 The use of call_rcu() permits the caller of foo_update_a() to
596 immediately regain control, without needing to worry further about the
597 old version of the newly updated element. It also clearly shows the
598 RCU distinction between updater, namely foo_update_a(), and reclaimer,
599 namely foo_reclaim().
600
601 The summary of advice is the same as for the previous section, except
602 that we are now using call_rcu() rather than synchronize_rcu():
603
604 - Use call_rcu() **after** removing a data element from an
605 RCU-protected data structure in order to register a callback
606 function that will be invoked after the completion of all RCU
607 read-side critical sections that might be referencing that
608 data item.
609
610 If the callback for call_rcu() is not doing anything more than calling
611 kfree() on the structure, you can use kfree_rcu() instead of call_rcu()
612 to avoid having to write your own callback::
613
614 kfree_rcu(old_fp, rcu);
615
616 If the occasional sleep is permitted, the single-argument form may
617 be used, omitting the rcu_head structure from struct foo.
618
619 kfree_rcu_mightsleep(old_fp);
620
621 This variant almost never blocks, but might do so by invoking
622 synchronize_rcu() in response to memory-allocation failure.
623
624 Again, see checklist.rst for additional rules governing the use of RCU.
625
626 .. _5_whatisRCU:
627
628 5. WHAT ARE SOME SIMPLE IMPLEMENTATIONS OF RCU?
629 ------------------------------------------------
630
631 One of the nice things about RCU is that it has extremely simple "toy"
632 implementations that are a good first step towards understanding the
633 production-quality implementations in the Linux kernel. This section
634 presents two such "toy" implementations of RCU, one that is implemented
635 in terms of familiar locking primitives, and another that more closely
636 resembles "classic" RCU. Both are way too simple for real-world use,
637 lacking both functionality and performance. However, they are useful
638 in getting a feel for how RCU works. See kernel/rcu/update.c for a
639 production-quality implementation, and see:
640
641 https://docs.google.com/document/d/1X0lThx8OK0ZgLMqVoXiR4ZrGURHrXK6NyLRbeXe3Xac/edit
642
643 for papers describing the Linux kernel RCU implementation. The OLS'01
644 and OLS'02 papers are a good introduction, and the dissertation provides
645 more details on the current implementation as of early 2004.
646
647
648 5A. "TOY" IMPLEMENTATION #1: LOCKING
649 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
650 This section presents a "toy" RCU implementation that is based on
651 familiar locking primitives. Its overhead makes it a non-starter for
652 real-life use, as does its lack of scalability. It is also unsuitable
653 for realtime use, since it allows scheduling latency to "bleed" from
654 one read-side critical section to another. It also assumes recursive
655 reader-writer locks: If you try this with non-recursive locks, and
656 you allow nested rcu_read_lock() calls, you can deadlock.
657
658 However, it is probably the easiest implementation to relate to, so is
659 a good starting point.
660
661 It is extremely simple::
662
663 static DEFINE_RWLOCK(rcu_gp_mutex);
664
665 void rcu_read_lock(void)
666 {
667 read_lock(&rcu_gp_mutex);
668 }
669
670 void rcu_read_unlock(void)
671 {
672 read_unlock(&rcu_gp_mutex);
673 }
674
675 void synchronize_rcu(void)
676 {
677 write_lock(&rcu_gp_mutex);
678 smp_mb__after_spinlock();
679 write_unlock(&rcu_gp_mutex);
680 }
681
682 [You can ignore rcu_assign_pointer() and rcu_dereference() without missing
683 much. But here are simplified versions anyway. And whatever you do,
684 don't forget about them when submitting patches making use of RCU!]::
685
686 #define rcu_assign_pointer(p, v) \
687 ({ \
688 smp_store_release(&(p), (v)); \
689 })
690
691 #define rcu_dereference(p) \
692 ({ \
693 typeof(p) _________p1 = READ_ONCE(p); \
694 (_________p1); \
695 })
696
697
698 The rcu_read_lock() and rcu_read_unlock() primitive read-acquire
699 and release a global reader-writer lock. The synchronize_rcu()
700 primitive write-acquires this same lock, then releases it. This means
701 that once synchronize_rcu() exits, all RCU read-side critical sections
702 that were in progress before synchronize_rcu() was called are guaranteed
703 to have completed -- there is no way that synchronize_rcu() would have
704 been able to write-acquire the lock otherwise. The smp_mb__after_spinlock()
705 promotes synchronize_rcu() to a full memory barrier in compliance with
706 the "Memory-Barrier Guarantees" listed in:
707
708 Design/Requirements/Requirements.rst
709
710 It is possible to nest rcu_read_lock(), since reader-writer locks may
711 be recursively acquired. Note also that rcu_read_lock() is immune
712 from deadlock (an important property of RCU). The reason for this is
713 that the only thing that can block rcu_read_lock() is a synchronize_rcu().
714 But synchronize_rcu() does not acquire any locks while holding rcu_gp_mutex,
715 so there can be no deadlock cycle.
716
717 .. _quiz_1:
718
719 Quick Quiz #1:
720 Why is this argument naive? How could a deadlock
721 occur when using this algorithm in a real-world Linux
722 kernel? How could this deadlock be avoided?
723
724 :ref:`Answers to Quick Quiz <9_whatisRCU>`
725
726 5B. "TOY" EXAMPLE #2: CLASSIC RCU
727 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
728 This section presents a "toy" RCU implementation that is based on
729 "classic RCU". It is also short on performance (but only for updates) and
730 on features such as hotplug CPU and the ability to run in CONFIG_PREEMPTION
731 kernels. The definitions of rcu_dereference() and rcu_assign_pointer()
732 are the same as those shown in the preceding section, so they are omitted.
733 ::
734
735 void rcu_read_lock(void) { }
736
737 void rcu_read_unlock(void) { }
738
739 void synchronize_rcu(void)
740 {
741 int cpu;
742
743 for_each_possible_cpu(cpu)
744 run_on(cpu);
745 }
746
747 Note that rcu_read_lock() and rcu_read_unlock() do absolutely nothing.
748 This is the great strength of classic RCU in a non-preemptive kernel:
749 read-side overhead is precisely zero, at least on non-Alpha CPUs.
750 And there is absolutely no way that rcu_read_lock() can possibly
751 participate in a deadlock cycle!
752
753 The implementation of synchronize_rcu() simply schedules itself on each
754 CPU in turn. The run_on() primitive can be implemented straightforwardly
755 in terms of the sched_setaffinity() primitive. Of course, a somewhat less
756 "toy" implementation would restore the affinity upon completion rather
757 than just leaving all tasks running on the last CPU, but when I said
758 "toy", I meant **toy**!
759
760 So how the heck is this supposed to work???
761
762 Remember that it is illegal to block while in an RCU read-side critical
763 section. Therefore, if a given CPU executes a context switch, we know
764 that it must have completed all preceding RCU read-side critical sections.
765 Once **all** CPUs have executed a context switch, then **all** preceding
766 RCU read-side critical sections will have completed.
767
768 So, suppose that we remove a data item from its structure and then invoke
769 synchronize_rcu(). Once synchronize_rcu() returns, we are guaranteed
770 that there are no RCU read-side critical sections holding a reference
771 to that data item, so we can safely reclaim it.
772
773 .. _quiz_2:
774
775 Quick Quiz #2:
776 Give an example where Classic RCU's read-side
777 overhead is **negative**.
778
779 :ref:`Answers to Quick Quiz <9_whatisRCU>`
780
781 .. _quiz_3:
782
783 Quick Quiz #3:
784 If it is illegal to block in an RCU read-side
785 critical section, what the heck do you do in
786 CONFIG_PREEMPT_RT, where normal spinlocks can block???
787
788 :ref:`Answers to Quick Quiz <9_whatisRCU>`
789
790 .. _6_whatisRCU:
791
792 6. ANALOGY WITH READER-WRITER LOCKING
793 --------------------------------------
794
795 Although RCU can be used in many different ways, a very common use of
796 RCU is analogous to reader-writer locking. The following unified
797 diff shows how closely related RCU and reader-writer locking can be.
798 ::
799
800 @@ -5,5 +5,5 @@ struct el {
801 int data;
802 /* Other data fields */
803 };
804 -rwlock_t listmutex;
805 +spinlock_t listmutex;
806 struct el head;
807
808 @@ -13,15 +14,15 @@
809 struct list_head *lp;
810 struct el *p;
811
812 - read_lock(&listmutex);
813 - list_for_each_entry(p, head, lp) {
814 + rcu_read_lock();
815 + list_for_each_entry_rcu(p, head, lp) {
816 if (p->key == key) {
817 *result = p->data;
818 - read_unlock(&listmutex);
819 + rcu_read_unlock();
820 return 1;
821 }
822 }
823 - read_unlock(&listmutex);
824 + rcu_read_unlock();
825 return 0;
826 }
827
828 @@ -29,15 +30,16 @@
829 {
830 struct el *p;
831
832 - write_lock(&listmutex);
833 + spin_lock(&listmutex);
834 list_for_each_entry(p, head, lp) {
835 if (p->key == key) {
836 - list_del(&p->list);
837 - write_unlock(&listmutex);
838 + list_del_rcu(&p->list);
839 + spin_unlock(&listmutex);
840 + synchronize_rcu();
841 kfree(p);
842 return 1;
843 }
844 }
845 - write_unlock(&listmutex);
846 + spin_unlock(&listmutex);
847 return 0;
848 }
849
850 Or, for those who prefer a side-by-side listing::
851
852 1 struct el { 1 struct el {
853 2 struct list_head list; 2 struct list_head list;
854 3 long key; 3 long key;
855 4 spinlock_t mutex; 4 spinlock_t mutex;
856 5 int data; 5 int data;
857 6 /* Other data fields */ 6 /* Other data fields */
858 7 }; 7 };
859 8 rwlock_t listmutex; 8 spinlock_t listmutex;
860 9 struct el head; 9 struct el head;
861
862 ::
863
864 1 int search(long key, int *result) 1 int search(long key, int *result)
865 2 { 2 {
866 3 struct list_head *lp; 3 struct list_head *lp;
867 4 struct el *p; 4 struct el *p;
868 5 5
869 6 read_lock(&listmutex); 6 rcu_read_lock();
870 7 list_for_each_entry(p, head, lp) { 7 list_for_each_entry_rcu(p, head, lp) {
871 8 if (p->key == key) { 8 if (p->key == key) {
872 9 *result = p->data; 9 *result = p->data;
873 10 read_unlock(&listmutex); 10 rcu_read_unlock();
874 11 return 1; 11 return 1;
875 12 } 12 }
876 13 } 13 }
877 14 read_unlock(&listmutex); 14 rcu_read_unlock();
878 15 return 0; 15 return 0;
879 16 } 16 }
880
881 ::
882
883 1 int delete(long key) 1 int delete(long key)
884 2 { 2 {
885 3 struct el *p; 3 struct el *p;
886 4 4
887 5 write_lock(&listmutex); 5 spin_lock(&listmutex);
888 6 list_for_each_entry(p, head, lp) { 6 list_for_each_entry(p, head, lp) {
889 7 if (p->key == key) { 7 if (p->key == key) {
890 8 list_del(&p->list); 8 list_del_rcu(&p->list);
891 9 write_unlock(&listmutex); 9 spin_unlock(&listmutex);
892 10 synchronize_rcu();
893 10 kfree(p); 11 kfree(p);
894 11 return 1; 12 return 1;
895 12 } 13 }
896 13 } 14 }
897 14 write_unlock(&listmutex); 15 spin_unlock(&listmutex);
898 15 return 0; 16 return 0;
899 16 } 17 }
900
901 Either way, the differences are quite small. Read-side locking moves
902 to rcu_read_lock() and rcu_read_unlock, update-side locking moves from
903 a reader-writer lock to a simple spinlock, and a synchronize_rcu()
904 precedes the kfree().
905
906 However, there is one potential catch: the read-side and update-side
907 critical sections can now run concurrently. In many cases, this will
908 not be a problem, but it is necessary to check carefully regardless.
909 For example, if multiple independent list updates must be seen as
910 a single atomic update, converting to RCU will require special care.
911
912 Also, the presence of synchronize_rcu() means that the RCU version of
913 delete() can now block. If this is a problem, there is a callback-based
914 mechanism that never blocks, namely call_rcu() or kfree_rcu(), that can
915 be used in place of synchronize_rcu().
916
917 .. _7_whatisRCU:
918
919 7. ANALOGY WITH REFERENCE COUNTING
920 -----------------------------------
921
922 The reader-writer analogy (illustrated by the previous section) is not
923 always the best way to think about using RCU. Another helpful analogy
924 considers RCU an effective reference count on everything which is
925 protected by RCU.
926
927 A reference count typically does not prevent the referenced object's
928 values from changing, but does prevent changes to type -- particularly the
929 gross change of type that happens when that object's memory is freed and
930 re-allocated for some other purpose. Once a type-safe reference to the
931 object is obtained, some other mechanism is needed to ensure consistent
932 access to the data in the object. This could involve taking a spinlock,
933 but with RCU the typical approach is to perform reads with SMP-aware
934 operations such as smp_load_acquire(), to perform updates with atomic
935 read-modify-write operations, and to provide the necessary ordering.
936 RCU provides a number of support functions that embed the required
937 operations and ordering, such as the list_for_each_entry_rcu() macro
938 used in the previous section.
939
940 A more focused view of the reference counting behavior is that,
941 between rcu_read_lock() and rcu_read_unlock(), any reference taken with
942 rcu_dereference() on a pointer marked as ``__rcu`` can be treated as
943 though a reference-count on that object has been temporarily increased.
944 This prevents the object from changing type. Exactly what this means
945 will depend on normal expectations of objects of that type, but it
946 typically includes that spinlocks can still be safely locked, normal
947 reference counters can be safely manipulated, and ``__rcu`` pointers
948 can be safely dereferenced.
949
950 Some operations that one might expect to see on an object for
951 which an RCU reference is held include:
952
953 - Copying out data that is guaranteed to be stable by the object's type.
954 - Using kref_get_unless_zero() or similar to get a longer-term
955 reference. This may fail of course.
956 - Acquiring a spinlock in the object, and checking if the object still
957 is the expected object and if so, manipulating it freely.
958
959 The understanding that RCU provides a reference that only prevents a
960 change of type is particularly visible with objects allocated from a
961 slab cache marked ``SLAB_TYPESAFE_BY_RCU``. RCU operations may yield a
962 reference to an object from such a cache that has been concurrently freed
963 and the memory reallocated to a completely different object, though of
964 the same type. In this case RCU doesn't even protect the identity of the
965 object from changing, only its type. So the object found may not be the
966 one expected, but it will be one where it is safe to take a reference
967 (and then potentially acquiring a spinlock), allowing subsequent code
968 to check whether the identity matches expectations. It is tempting
969 to simply acquire the spinlock without first taking the reference, but
970 unfortunately any spinlock in a ``SLAB_TYPESAFE_BY_RCU`` object must be
971 initialized after each and every call to kmem_cache_alloc(), which renders
972 reference-free spinlock acquisition completely unsafe. Therefore, when
973 using ``SLAB_TYPESAFE_BY_RCU``, make proper use of a reference counter.
974 (Those willing to initialize their locks in a kmem_cache constructor
975 may also use locking, including cache-friendly sequence locking.)
976
977 With traditional reference counting -- such as that implemented by the
978 kref library in Linux -- there is typically code that runs when the last
979 reference to an object is dropped. With kref, this is the function
980 passed to kref_put(). When RCU is being used, such finalization code
981 must not be run until all ``__rcu`` pointers referencing the object have
982 been updated, and then a grace period has passed. Every remaining
983 globally visible pointer to the object must be considered to be a
984 potential counted reference, and the finalization code is typically run
985 using call_rcu() only after all those pointers have been changed.
986
987 To see how to choose between these two analogies -- of RCU as a
988 reader-writer lock and RCU as a reference counting system -- it is useful
989 to reflect on the scale of the thing being protected. The reader-writer
990 lock analogy looks at larger multi-part objects such as a linked list
991 and shows how RCU can facilitate concurrency while elements are added
992 to, and removed from, the list. The reference-count analogy looks at
993 the individual objects and looks at how they can be accessed safely
994 within whatever whole they are a part of.
995
996 .. _8_whatisRCU:
997
998 8. FULL LIST OF RCU APIs
999 -------------------------
1000
1001 The RCU APIs are documented in docbook-format header comments in the
1002 Linux-kernel source code, but it helps to have a full list of the
1003 APIs, since there does not appear to be a way to categorize them
1004 in docbook. Here is the list, by category.
1005
1006 RCU list traversal::
1007
1008 list_entry_rcu
1009 list_entry_lockless
1010 list_first_entry_rcu
1011 list_next_rcu
1012 list_for_each_entry_rcu
1013 list_for_each_entry_continue_rcu
1014 list_for_each_entry_from_rcu
1015 list_first_or_null_rcu
1016 list_next_or_null_rcu
1017 hlist_first_rcu
1018 hlist_next_rcu
1019 hlist_pprev_rcu
1020 hlist_for_each_entry_rcu
1021 hlist_for_each_entry_rcu_bh
1022 hlist_for_each_entry_from_rcu
1023 hlist_for_each_entry_continue_rcu
1024 hlist_for_each_entry_continue_rcu_bh
1025 hlist_nulls_first_rcu
1026 hlist_nulls_for_each_entry_rcu
1027 hlist_bl_first_rcu
1028 hlist_bl_for_each_entry_rcu
1029
1030 RCU pointer/list update::
1031
1032 rcu_assign_pointer
1033 list_add_rcu
1034 list_add_tail_rcu
1035 list_del_rcu
1036 list_replace_rcu
1037 hlist_add_behind_rcu
1038 hlist_add_before_rcu
1039 hlist_add_head_rcu
1040 hlist_add_tail_rcu
1041 hlist_del_rcu
1042 hlist_del_init_rcu
1043 hlist_replace_rcu
1044 list_splice_init_rcu
1045 list_splice_tail_init_rcu
1046 hlist_nulls_del_init_rcu
1047 hlist_nulls_del_rcu
1048 hlist_nulls_add_head_rcu
1049 hlist_bl_add_head_rcu
1050 hlist_bl_del_init_rcu
1051 hlist_bl_del_rcu
1052 hlist_bl_set_first_rcu
1053
1054 RCU::
1055
1056 Critical sections Grace period Barrier
1057
1058 rcu_read_lock synchronize_net rcu_barrier
1059 rcu_read_unlock synchronize_rcu
1060 rcu_dereference synchronize_rcu_expedited
1061 rcu_read_lock_held call_rcu
1062 rcu_dereference_check kfree_rcu
1063 rcu_dereference_protected
1064
1065 bh::
1066
1067 Critical sections Grace period Barrier
1068
1069 rcu_read_lock_bh call_rcu rcu_barrier
1070 rcu_read_unlock_bh synchronize_rcu
1071 [local_bh_disable] synchronize_rcu_expedited
1072 [and friends]
1073 rcu_dereference_bh
1074 rcu_dereference_bh_check
1075 rcu_dereference_bh_protected
1076 rcu_read_lock_bh_held
1077
1078 sched::
1079
1080 Critical sections Grace period Barrier
1081
1082 rcu_read_lock_sched call_rcu rcu_barrier
1083 rcu_read_unlock_sched synchronize_rcu
1084 [preempt_disable] synchronize_rcu_expedited
1085 [and friends]
1086 rcu_read_lock_sched_notrace
1087 rcu_read_unlock_sched_notrace
1088 rcu_dereference_sched
1089 rcu_dereference_sched_check
1090 rcu_dereference_sched_protected
1091 rcu_read_lock_sched_held
1092
1093
1094 RCU-Tasks::
1095
1096 Critical sections Grace period Barrier
1097
1098 N/A call_rcu_tasks rcu_barrier_tasks
1099 synchronize_rcu_tasks
1100
1101
1102 RCU-Tasks-Rude::
1103
1104 Critical sections Grace period Barrier
1105
1106 N/A call_rcu_tasks_rude rcu_barrier_tasks_rude
1107 synchronize_rcu_tasks_rude
1108
1109
1110 RCU-Tasks-Trace::
1111
1112 Critical sections Grace period Barrier
1113
1114 rcu_read_lock_trace call_rcu_tasks_trace rcu_barrier_tasks_trace
1115 rcu_read_unlock_trace synchronize_rcu_tasks_trace
1116
1117
1118 SRCU::
1119
1120 Critical sections Grace period Barrier
1121
1122 srcu_read_lock call_srcu srcu_barrier
1123 srcu_read_unlock synchronize_srcu
1124 srcu_dereference synchronize_srcu_expedited
1125 srcu_dereference_check
1126 srcu_read_lock_held
1127
1128 SRCU: Initialization/cleanup::
1129
1130 DEFINE_SRCU
1131 DEFINE_STATIC_SRCU
1132 init_srcu_struct
1133 cleanup_srcu_struct
1134
1135 All: lockdep-checked RCU utility APIs::
1136
1137 RCU_LOCKDEP_WARN
1138 rcu_sleep_check
1139
1140 All: Unchecked RCU-protected pointer access::
1141
1142 rcu_dereference_raw
1143
1144 All: Unchecked RCU-protected pointer access with dereferencing prohibited::
1145
1146 rcu_access_pointer
1147
1148 See the comment headers in the source code (or the docbook generated
1149 from them) for more information.
1150
1151 However, given that there are no fewer than four families of RCU APIs
1152 in the Linux kernel, how do you choose which one to use? The following
1153 list can be helpful:
1154
1155 a. Will readers need to block? If so, you need SRCU.
1156
1157 b. Will readers need to block and are you doing tracing, for
1158 example, ftrace or BPF? If so, you need RCU-tasks,
1159 RCU-tasks-rude, and/or RCU-tasks-trace.
1160
1161 c. What about the -rt patchset? If readers would need to block in
1162 an non-rt kernel, you need SRCU. If readers would block when
1163 acquiring spinlocks in a -rt kernel, but not in a non-rt kernel,
1164 SRCU is not necessary. (The -rt patchset turns spinlocks into
1165 sleeplocks, hence this distinction.)
1166
1167 d. Do you need to treat NMI handlers, hardirq handlers,
1168 and code segments with preemption disabled (whether
1169 via preempt_disable(), local_irq_save(), local_bh_disable(),
1170 or some other mechanism) as if they were explicit RCU readers?
1171 If so, RCU-sched readers are the only choice that will work
1172 for you, but since about v4.20 you use can use the vanilla RCU
1173 update primitives.
1174
1175 e. Do you need RCU grace periods to complete even in the face of
1176 softirq monopolization of one or more of the CPUs? For example,
1177 is your code subject to network-based denial-of-service attacks?
1178 If so, you should disable softirq across your readers, for
1179 example, by using rcu_read_lock_bh(). Since about v4.20 you
1180 use can use the vanilla RCU update primitives.
1181
1182 f. Is your workload too update-intensive for normal use of
1183 RCU, but inappropriate for other synchronization mechanisms?
1184 If so, consider SLAB_TYPESAFE_BY_RCU (which was originally
1185 named SLAB_DESTROY_BY_RCU). But please be careful!
1186
1187 g. Do you need read-side critical sections that are respected even
1188 on CPUs that are deep in the idle loop, during entry to or exit
1189 from user-mode execution, or on an offlined CPU? If so, SRCU
1190 and RCU Tasks Trace are the only choices that will work for you,
1191 with SRCU being strongly preferred in almost all cases.
1192
1193 h. Otherwise, use RCU.
1194
1195 Of course, this all assumes that you have determined that RCU is in fact
1196 the right tool for your job.
1197
1198 .. _9_whatisRCU:
1199
1200 9. ANSWERS TO QUICK QUIZZES
1201 ----------------------------
1202
1203 Quick Quiz #1:
1204 Why is this argument naive? How could a deadlock
1205 occur when using this algorithm in a real-world Linux
1206 kernel? [Referring to the lock-based "toy" RCU
1207 algorithm.]
1208
1209 Answer:
1210 Consider the following sequence of events:
1211
1212 1. CPU 0 acquires some unrelated lock, call it
1213 "problematic_lock", disabling irq via
1214 spin_lock_irqsave().
1215
1216 2. CPU 1 enters synchronize_rcu(), write-acquiring
1217 rcu_gp_mutex.
1218
1219 3. CPU 0 enters rcu_read_lock(), but must wait
1220 because CPU 1 holds rcu_gp_mutex.
1221
1222 4. CPU 1 is interrupted, and the irq handler
1223 attempts to acquire problematic_lock.
1224
1225 The system is now deadlocked.
1226
1227 One way to avoid this deadlock is to use an approach like
1228 that of CONFIG_PREEMPT_RT, where all normal spinlocks
1229 become blocking locks, and all irq handlers execute in
1230 the context of special tasks. In this case, in step 4
1231 above, the irq handler would block, allowing CPU 1 to
1232 release rcu_gp_mutex, avoiding the deadlock.
1233
1234 Even in the absence of deadlock, this RCU implementation
1235 allows latency to "bleed" from readers to other
1236 readers through synchronize_rcu(). To see this,
1237 consider task A in an RCU read-side critical section
1238 (thus read-holding rcu_gp_mutex), task B blocked
1239 attempting to write-acquire rcu_gp_mutex, and
1240 task C blocked in rcu_read_lock() attempting to
1241 read_acquire rcu_gp_mutex. Task A's RCU read-side
1242 latency is holding up task C, albeit indirectly via
1243 task B.
1244
1245 Realtime RCU implementations therefore use a counter-based
1246 approach where tasks in RCU read-side critical sections
1247 cannot be blocked by tasks executing synchronize_rcu().
1248
1249 :ref:`Back to Quick Quiz #1 <quiz_1>`
1250
1251 Quick Quiz #2:
1252 Give an example where Classic RCU's read-side
1253 overhead is **negative**.
1254
1255 Answer:
1256 Imagine a single-CPU system with a non-CONFIG_PREEMPTION
1257 kernel where a routing table is used by process-context
1258 code, but can be updated by irq-context code (for example,
1259 by an "ICMP REDIRECT" packet). The usual way of handling
1260 this would be to have the process-context code disable
1261 interrupts while searching the routing table. Use of
1262 RCU allows such interrupt-disabling to be dispensed with.
1263 Thus, without RCU, you pay the cost of disabling interrupts,
1264 and with RCU you don't.
1265
1266 One can argue that the overhead of RCU in this
1267 case is negative with respect to the single-CPU
1268 interrupt-disabling approach. Others might argue that
1269 the overhead of RCU is merely zero, and that replacing
1270 the positive overhead of the interrupt-disabling scheme
1271 with the zero-overhead RCU scheme does not constitute
1272 negative overhead.
1273
1274 In real life, of course, things are more complex. But
1275 even the theoretical possibility of negative overhead for
1276 a synchronization primitive is a bit unexpected. ;-)
1277
1278 :ref:`Back to Quick Quiz #2 <quiz_2>`
1279
1280 Quick Quiz #3:
1281 If it is illegal to block in an RCU read-side
1282 critical section, what the heck do you do in
1283 CONFIG_PREEMPT_RT, where normal spinlocks can block???
1284
1285 Answer:
1286 Just as CONFIG_PREEMPT_RT permits preemption of spinlock
1287 critical sections, it permits preemption of RCU
1288 read-side critical sections. It also permits
1289 spinlocks blocking while in RCU read-side critical
1290 sections.
1291
1292 Why the apparent inconsistency? Because it is
1293 possible to use priority boosting to keep the RCU
1294 grace periods short if need be (for example, if running
1295 short of memory). In contrast, if blocking waiting
1296 for (say) network reception, there is no way to know
1297 what should be boosted. Especially given that the
1298 process we need to boost might well be a human being
1299 who just went out for a pizza or something. And although
1300 a computer-operated cattle prod might arouse serious
1301 interest, it might also provoke serious objections.
1302 Besides, how does the computer know what pizza parlor
1303 the human being went to???
1304
1305 :ref:`Back to Quick Quiz #3 <quiz_3>`
1306
1307 ACKNOWLEDGEMENTS
1308
1309 My thanks to the people who helped make this human-readable, including
1310 Jon Walpole, Josh Triplett, Serge Hallyn, Suzanne Wood, and Alan Stern.
1311
1312
1313 For more information, see http://www.rdrop.com/users/paulmck/RCU.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.