About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Documentation / RCU / whatisRCU.txt




Custom Search

Based on kernel version 3.13. Page generated on 2014-01-20 22:04 EST.

1	Please note that the "What is RCU?" LWN series is an excellent place
2	to start learning about RCU:
3	
4	1.	What is RCU, Fundamentally?  http://lwn.net/Articles/262464/
5	2.	What is RCU? Part 2: Usage   http://lwn.net/Articles/263130/
6	3.	RCU part 3: the RCU API      http://lwn.net/Articles/264090/
7	4.	The RCU API, 2010 Edition    http://lwn.net/Articles/418853/
8	
9	
10	What is RCU?
11	
12	RCU is a synchronization mechanism that was added to the Linux kernel
13	during the 2.5 development effort that is optimized for read-mostly
14	situations.  Although RCU is actually quite simple once you understand it,
15	getting there can sometimes be a challenge.  Part of the problem is that
16	most of the past descriptions of RCU have been written with the mistaken
17	assumption that there is "one true way" to describe RCU.  Instead,
18	the experience has been that different people must take different paths
19	to arrive at an understanding of RCU.  This document provides several
20	different paths, as follows:
21	
22	1.	RCU OVERVIEW
23	2.	WHAT IS RCU'S CORE API?
24	3.	WHAT ARE SOME EXAMPLE USES OF CORE RCU API?
25	4.	WHAT IF MY UPDATING THREAD CANNOT BLOCK?
26	5.	WHAT ARE SOME SIMPLE IMPLEMENTATIONS OF RCU?
27	6.	ANALOGY WITH READER-WRITER LOCKING
28	7.	FULL LIST OF RCU APIs
29	8.	ANSWERS TO QUICK QUIZZES
30	
31	People who prefer starting with a conceptual overview should focus on
32	Section 1, though most readers will profit by reading this section at
33	some point.  People who prefer to start with an API that they can then
34	experiment with should focus on Section 2.  People who prefer to start
35	with example uses should focus on Sections 3 and 4.  People who need to
36	understand the RCU implementation should focus on Section 5, then dive
37	into the kernel source code.  People who reason best by analogy should
38	focus on Section 6.  Section 7 serves as an index to the docbook API
39	documentation, and Section 8 is the traditional answer key.
40	
41	So, start with the section that makes the most sense to you and your
42	preferred method of learning.  If you need to know everything about
43	everything, feel free to read the whole thing -- but if you are really
44	that type of person, you have perused the source code and will therefore
45	never need this document anyway.  ;-)
46	
47	
48	1.  RCU OVERVIEW
49	
50	The basic idea behind RCU is to split updates into "removal" and
51	"reclamation" phases.  The removal phase removes references to data items
52	within a data structure (possibly by replacing them with references to
53	new versions of these data items), and can run concurrently with readers.
54	The reason that it is safe to run the removal phase concurrently with
55	readers is the semantics of modern CPUs guarantee that readers will see
56	either the old or the new version of the data structure rather than a
57	partially updated reference.  The reclamation phase does the work of reclaiming
58	(e.g., freeing) the data items removed from the data structure during the
59	removal phase.  Because reclaiming data items can disrupt any readers
60	concurrently referencing those data items, the reclamation phase must
61	not start until readers no longer hold references to those data items.
62	
63	Splitting the update into removal and reclamation phases permits the
64	updater to perform the removal phase immediately, and to defer the
65	reclamation phase until all readers active during the removal phase have
66	completed, either by blocking until they finish or by registering a
67	callback that is invoked after they finish.  Only readers that are active
68	during the removal phase need be considered, because any reader starting
69	after the removal phase will be unable to gain a reference to the removed
70	data items, and therefore cannot be disrupted by the reclamation phase.
71	
72	So the typical RCU update sequence goes something like the following:
73	
74	a.	Remove pointers to a data structure, so that subsequent
75		readers cannot gain a reference to it.
76	
77	b.	Wait for all previous readers to complete their RCU read-side
78		critical sections.
79	
80	c.	At this point, there cannot be any readers who hold references
81		to the data structure, so it now may safely be reclaimed
82		(e.g., kfree()d).
83	
84	Step (b) above is the key idea underlying RCU's deferred destruction.
85	The ability to wait until all readers are done allows RCU readers to
86	use much lighter-weight synchronization, in some cases, absolutely no
87	synchronization at all.  In contrast, in more conventional lock-based
88	schemes, readers must use heavy-weight synchronization in order to
89	prevent an updater from deleting the data structure out from under them.
90	This is because lock-based updaters typically update data items in place,
91	and must therefore exclude readers.  In contrast, RCU-based updaters
92	typically take advantage of the fact that writes to single aligned
93	pointers are atomic on modern CPUs, allowing atomic insertion, removal,
94	and replacement of data items in a linked structure without disrupting
95	readers.  Concurrent RCU readers can then continue accessing the old
96	versions, and can dispense with the atomic operations, memory barriers,
97	and communications cache misses that are so expensive on present-day
98	SMP computer systems, even in absence of lock contention.
99	
100	In the three-step procedure shown above, the updater is performing both
101	the removal and the reclamation step, but it is often helpful for an
102	entirely different thread to do the reclamation, as is in fact the case
103	in the Linux kernel's directory-entry cache (dcache).  Even if the same
104	thread performs both the update step (step (a) above) and the reclamation
105	step (step (c) above), it is often helpful to think of them separately.
106	For example, RCU readers and updaters need not communicate at all,
107	but RCU provides implicit low-overhead communication between readers
108	and reclaimers, namely, in step (b) above.
109	
110	So how the heck can a reclaimer tell when a reader is done, given
111	that readers are not doing any sort of synchronization operations???
112	Read on to learn about how RCU's API makes this easy.
113	
114	
115	2.  WHAT IS RCU'S CORE API?
116	
117	The core RCU API is quite small:
118	
119	a.	rcu_read_lock()
120	b.	rcu_read_unlock()
121	c.	synchronize_rcu() / call_rcu()
122	d.	rcu_assign_pointer()
123	e.	rcu_dereference()
124	
125	There are many other members of the RCU API, but the rest can be
126	expressed in terms of these five, though most implementations instead
127	express synchronize_rcu() in terms of the call_rcu() callback API.
128	
129	The five core RCU APIs are described below, the other 18 will be enumerated
130	later.  See the kernel docbook documentation for more info, or look directly
131	at the function header comments.
132	
133	rcu_read_lock()
134	
135		void rcu_read_lock(void);
136	
137		Used by a reader to inform the reclaimer that the reader is
138		entering an RCU read-side critical section.  It is illegal
139		to block while in an RCU read-side critical section, though
140		kernels built with CONFIG_TREE_PREEMPT_RCU can preempt RCU
141		read-side critical sections.  Any RCU-protected data structure
142		accessed during an RCU read-side critical section is guaranteed to
143		remain unreclaimed for the full duration of that critical section.
144		Reference counts may be used in conjunction with RCU to maintain
145		longer-term references to data structures.
146	
147	rcu_read_unlock()
148	
149		void rcu_read_unlock(void);
150	
151		Used by a reader to inform the reclaimer that the reader is
152		exiting an RCU read-side critical section.  Note that RCU
153		read-side critical sections may be nested and/or overlapping.
154	
155	synchronize_rcu()
156	
157		void synchronize_rcu(void);
158	
159		Marks the end of updater code and the beginning of reclaimer
160		code.  It does this by blocking until all pre-existing RCU
161		read-side critical sections on all CPUs have completed.
162		Note that synchronize_rcu() will -not- necessarily wait for
163		any subsequent RCU read-side critical sections to complete.
164		For example, consider the following sequence of events:
165	
166		         CPU 0                  CPU 1                 CPU 2
167		     ----------------- ------------------------- ---------------
168		 1.  rcu_read_lock()
169		 2.                    enters synchronize_rcu()
170		 3.                                               rcu_read_lock()
171		 4.  rcu_read_unlock()
172		 5.                     exits synchronize_rcu()
173		 6.                                              rcu_read_unlock()
174	
175		To reiterate, synchronize_rcu() waits only for ongoing RCU
176		read-side critical sections to complete, not necessarily for
177		any that begin after synchronize_rcu() is invoked.
178	
179		Of course, synchronize_rcu() does not necessarily return
180		-immediately- after the last pre-existing RCU read-side critical
181		section completes.  For one thing, there might well be scheduling
182		delays.  For another thing, many RCU implementations process
183		requests in batches in order to improve efficiencies, which can
184		further delay synchronize_rcu().
185	
186		Since synchronize_rcu() is the API that must figure out when
187		readers are done, its implementation is key to RCU.  For RCU
188		to be useful in all but the most read-intensive situations,
189		synchronize_rcu()'s overhead must also be quite small.
190	
191		The call_rcu() API is a callback form of synchronize_rcu(),
192		and is described in more detail in a later section.  Instead of
193		blocking, it registers a function and argument which are invoked
194		after all ongoing RCU read-side critical sections have completed.
195		This callback variant is particularly useful in situations where
196		it is illegal to block or where update-side performance is
197		critically important.
198	
199		However, the call_rcu() API should not be used lightly, as use
200		of the synchronize_rcu() API generally results in simpler code.
201		In addition, the synchronize_rcu() API has the nice property
202		of automatically limiting update rate should grace periods
203		be delayed.  This property results in system resilience in face
204		of denial-of-service attacks.  Code using call_rcu() should limit
205		update rate in order to gain this same sort of resilience.  See
206		checklist.txt for some approaches to limiting the update rate.
207	
208	rcu_assign_pointer()
209	
210		typeof(p) rcu_assign_pointer(p, typeof(p) v);
211	
212		Yes, rcu_assign_pointer() -is- implemented as a macro, though it
213		would be cool to be able to declare a function in this manner.
214		(Compiler experts will no doubt disagree.)
215	
216		The updater uses this function to assign a new value to an
217		RCU-protected pointer, in order to safely communicate the change
218		in value from the updater to the reader.  This function returns
219		the new value, and also executes any memory-barrier instructions
220		required for a given CPU architecture.
221	
222		Perhaps just as important, it serves to document (1) which
223		pointers are protected by RCU and (2) the point at which a
224		given structure becomes accessible to other CPUs.  That said,
225		rcu_assign_pointer() is most frequently used indirectly, via
226		the _rcu list-manipulation primitives such as list_add_rcu().
227	
228	rcu_dereference()
229	
230		typeof(p) rcu_dereference(p);
231	
232		Like rcu_assign_pointer(), rcu_dereference() must be implemented
233		as a macro.
234	
235		The reader uses rcu_dereference() to fetch an RCU-protected
236		pointer, which returns a value that may then be safely
237		dereferenced.  Note that rcu_deference() does not actually
238		dereference the pointer, instead, it protects the pointer for
239		later dereferencing.  It also executes any needed memory-barrier
240		instructions for a given CPU architecture.  Currently, only Alpha
241		needs memory barriers within rcu_dereference() -- on other CPUs,
242		it compiles to nothing, not even a compiler directive.
243	
244		Common coding practice uses rcu_dereference() to copy an
245		RCU-protected pointer to a local variable, then dereferences
246		this local variable, for example as follows:
247	
248			p = rcu_dereference(head.next);
249			return p->data;
250	
251		However, in this case, one could just as easily combine these
252		into one statement:
253	
254			return rcu_dereference(head.next)->data;
255	
256		If you are going to be fetching multiple fields from the
257		RCU-protected structure, using the local variable is of
258		course preferred.  Repeated rcu_dereference() calls look
259		ugly and incur unnecessary overhead on Alpha CPUs.
260	
261		Note that the value returned by rcu_dereference() is valid
262		only within the enclosing RCU read-side critical section.
263		For example, the following is -not- legal:
264	
265			rcu_read_lock();
266			p = rcu_dereference(head.next);
267			rcu_read_unlock();
268			x = p->address;	/* BUG!!! */
269			rcu_read_lock();
270			y = p->data;	/* BUG!!! */
271			rcu_read_unlock();
272	
273		Holding a reference from one RCU read-side critical section
274		to another is just as illegal as holding a reference from
275		one lock-based critical section to another!  Similarly,
276		using a reference outside of the critical section in which
277		it was acquired is just as illegal as doing so with normal
278		locking.
279	
280		As with rcu_assign_pointer(), an important function of
281		rcu_dereference() is to document which pointers are protected by
282		RCU, in particular, flagging a pointer that is subject to changing
283		at any time, including immediately after the rcu_dereference().
284		And, again like rcu_assign_pointer(), rcu_dereference() is
285		typically used indirectly, via the _rcu list-manipulation
286		primitives, such as list_for_each_entry_rcu().
287	
288	The following diagram shows how each API communicates among the
289	reader, updater, and reclaimer.
290	
291	
292		    rcu_assign_pointer()
293		    			    +--------+
294		    +---------------------->| reader |---------+
295		    |                       +--------+         |
296		    |                           |              |
297		    |                           |              | Protect:
298		    |                           |              | rcu_read_lock()
299		    |                           |              | rcu_read_unlock()
300		    |        rcu_dereference()  |              |
301	       +---------+                      |              |
302	       | updater |<---------------------+              |
303	       +---------+                                     V
304		    |                                    +-----------+
305		    +----------------------------------->| reclaimer |
306		    				         +-----------+
307		      Defer:
308		      synchronize_rcu() & call_rcu()
309	
310	
311	The RCU infrastructure observes the time sequence of rcu_read_lock(),
312	rcu_read_unlock(), synchronize_rcu(), and call_rcu() invocations in
313	order to determine when (1) synchronize_rcu() invocations may return
314	to their callers and (2) call_rcu() callbacks may be invoked.  Efficient
315	implementations of the RCU infrastructure make heavy use of batching in
316	order to amortize their overhead over many uses of the corresponding APIs.
317	
318	There are no fewer than three RCU mechanisms in the Linux kernel; the
319	diagram above shows the first one, which is by far the most commonly used.
320	The rcu_dereference() and rcu_assign_pointer() primitives are used for
321	all three mechanisms, but different defer and protect primitives are
322	used as follows:
323	
324		Defer			Protect
325	
326	a.	synchronize_rcu()	rcu_read_lock() / rcu_read_unlock()
327		call_rcu()		rcu_dereference()
328	
329	b.	call_rcu_bh()		rcu_read_lock_bh() / rcu_read_unlock_bh()
330					rcu_dereference_bh()
331	
332	c.	synchronize_sched()	rcu_read_lock_sched() / rcu_read_unlock_sched()
333					preempt_disable() / preempt_enable()
334					local_irq_save() / local_irq_restore()
335					hardirq enter / hardirq exit
336					NMI enter / NMI exit
337					rcu_dereference_sched()
338	
339	These three mechanisms are used as follows:
340	
341	a.	RCU applied to normal data structures.
342	
343	b.	RCU applied to networking data structures that may be subjected
344		to remote denial-of-service attacks.
345	
346	c.	RCU applied to scheduler and interrupt/NMI-handler tasks.
347	
348	Again, most uses will be of (a).  The (b) and (c) cases are important
349	for specialized uses, but are relatively uncommon.
350	
351	
352	3.  WHAT ARE SOME EXAMPLE USES OF CORE RCU API?
353	
354	This section shows a simple use of the core RCU API to protect a
355	global pointer to a dynamically allocated structure.  More-typical
356	uses of RCU may be found in listRCU.txt, arrayRCU.txt, and NMI-RCU.txt.
357	
358		struct foo {
359			int a;
360			char b;
361			long c;
362		};
363		DEFINE_SPINLOCK(foo_mutex);
364	
365		struct foo *gbl_foo;
366	
367		/*
368		 * Create a new struct foo that is the same as the one currently
369		 * pointed to by gbl_foo, except that field "a" is replaced
370		 * with "new_a".  Points gbl_foo to the new structure, and
371		 * frees up the old structure after a grace period.
372		 *
373		 * Uses rcu_assign_pointer() to ensure that concurrent readers
374		 * see the initialized version of the new structure.
375		 *
376		 * Uses synchronize_rcu() to ensure that any readers that might
377		 * have references to the old structure complete before freeing
378		 * the old structure.
379		 */
380		void foo_update_a(int new_a)
381		{
382			struct foo *new_fp;
383			struct foo *old_fp;
384	
385			new_fp = kmalloc(sizeof(*new_fp), GFP_KERNEL);
386			spin_lock(&foo_mutex);
387			old_fp = gbl_foo;
388			*new_fp = *old_fp;
389			new_fp->a = new_a;
390			rcu_assign_pointer(gbl_foo, new_fp);
391			spin_unlock(&foo_mutex);
392			synchronize_rcu();
393			kfree(old_fp);
394		}
395	
396		/*
397		 * Return the value of field "a" of the current gbl_foo
398		 * structure.  Use rcu_read_lock() and rcu_read_unlock()
399		 * to ensure that the structure does not get deleted out
400		 * from under us, and use rcu_dereference() to ensure that
401		 * we see the initialized version of the structure (important
402		 * for DEC Alpha and for people reading the code).
403		 */
404		int foo_get_a(void)
405		{
406			int retval;
407	
408			rcu_read_lock();
409			retval = rcu_dereference(gbl_foo)->a;
410			rcu_read_unlock();
411			return retval;
412		}
413	
414	So, to sum up:
415	
416	o	Use rcu_read_lock() and rcu_read_unlock() to guard RCU
417		read-side critical sections.
418	
419	o	Within an RCU read-side critical section, use rcu_dereference()
420		to dereference RCU-protected pointers.
421	
422	o	Use some solid scheme (such as locks or semaphores) to
423		keep concurrent updates from interfering with each other.
424	
425	o	Use rcu_assign_pointer() to update an RCU-protected pointer.
426		This primitive protects concurrent readers from the updater,
427		-not- concurrent updates from each other!  You therefore still
428		need to use locking (or something similar) to keep concurrent
429		rcu_assign_pointer() primitives from interfering with each other.
430	
431	o	Use synchronize_rcu() -after- removing a data element from an
432		RCU-protected data structure, but -before- reclaiming/freeing
433		the data element, in order to wait for the completion of all
434		RCU read-side critical sections that might be referencing that
435		data item.
436	
437	See checklist.txt for additional rules to follow when using RCU.
438	And again, more-typical uses of RCU may be found in listRCU.txt,
439	arrayRCU.txt, and NMI-RCU.txt.
440	
441	
442	4.  WHAT IF MY UPDATING THREAD CANNOT BLOCK?
443	
444	In the example above, foo_update_a() blocks until a grace period elapses.
445	This is quite simple, but in some cases one cannot afford to wait so
446	long -- there might be other high-priority work to be done.
447	
448	In such cases, one uses call_rcu() rather than synchronize_rcu().
449	The call_rcu() API is as follows:
450	
451		void call_rcu(struct rcu_head * head,
452			      void (*func)(struct rcu_head *head));
453	
454	This function invokes func(head) after a grace period has elapsed.
455	This invocation might happen from either softirq or process context,
456	so the function is not permitted to block.  The foo struct needs to
457	have an rcu_head structure added, perhaps as follows:
458	
459		struct foo {
460			int a;
461			char b;
462			long c;
463			struct rcu_head rcu;
464		};
465	
466	The foo_update_a() function might then be written as follows:
467	
468		/*
469		 * Create a new struct foo that is the same as the one currently
470		 * pointed to by gbl_foo, except that field "a" is replaced
471		 * with "new_a".  Points gbl_foo to the new structure, and
472		 * frees up the old structure after a grace period.
473		 *
474		 * Uses rcu_assign_pointer() to ensure that concurrent readers
475		 * see the initialized version of the new structure.
476		 *
477		 * Uses call_rcu() to ensure that any readers that might have
478		 * references to the old structure complete before freeing the
479		 * old structure.
480		 */
481		void foo_update_a(int new_a)
482		{
483			struct foo *new_fp;
484			struct foo *old_fp;
485	
486			new_fp = kmalloc(sizeof(*new_fp), GFP_KERNEL);
487			spin_lock(&foo_mutex);
488			old_fp = gbl_foo;
489			*new_fp = *old_fp;
490			new_fp->a = new_a;
491			rcu_assign_pointer(gbl_foo, new_fp);
492			spin_unlock(&foo_mutex);
493			call_rcu(&old_fp->rcu, foo_reclaim);
494		}
495	
496	The foo_reclaim() function might appear as follows:
497	
498		void foo_reclaim(struct rcu_head *rp)
499		{
500			struct foo *fp = container_of(rp, struct foo, rcu);
501	
502			foo_cleanup(fp->a);
503	
504			kfree(fp);
505		}
506	
507	The container_of() primitive is a macro that, given a pointer into a
508	struct, the type of the struct, and the pointed-to field within the
509	struct, returns a pointer to the beginning of the struct.
510	
511	The use of call_rcu() permits the caller of foo_update_a() to
512	immediately regain control, without needing to worry further about the
513	old version of the newly updated element.  It also clearly shows the
514	RCU distinction between updater, namely foo_update_a(), and reclaimer,
515	namely foo_reclaim().
516	
517	The summary of advice is the same as for the previous section, except
518	that we are now using call_rcu() rather than synchronize_rcu():
519	
520	o	Use call_rcu() -after- removing a data element from an
521		RCU-protected data structure in order to register a callback
522		function that will be invoked after the completion of all RCU
523		read-side critical sections that might be referencing that
524		data item.
525	
526	If the callback for call_rcu() is not doing anything more than calling
527	kfree() on the structure, you can use kfree_rcu() instead of call_rcu()
528	to avoid having to write your own callback:
529	
530		kfree_rcu(old_fp, rcu);
531	
532	Again, see checklist.txt for additional rules governing the use of RCU.
533	
534	
535	5.  WHAT ARE SOME SIMPLE IMPLEMENTATIONS OF RCU?
536	
537	One of the nice things about RCU is that it has extremely simple "toy"
538	implementations that are a good first step towards understanding the
539	production-quality implementations in the Linux kernel.  This section
540	presents two such "toy" implementations of RCU, one that is implemented
541	in terms of familiar locking primitives, and another that more closely
542	resembles "classic" RCU.  Both are way too simple for real-world use,
543	lacking both functionality and performance.  However, they are useful
544	in getting a feel for how RCU works.  See kernel/rcupdate.c for a
545	production-quality implementation, and see:
546	
547		http://www.rdrop.com/users/paulmck/RCU
548	
549	for papers describing the Linux kernel RCU implementation.  The OLS'01
550	and OLS'02 papers are a good introduction, and the dissertation provides
551	more details on the current implementation as of early 2004.
552	
553	
554	5A.  "TOY" IMPLEMENTATION #1: LOCKING
555	
556	This section presents a "toy" RCU implementation that is based on
557	familiar locking primitives.  Its overhead makes it a non-starter for
558	real-life use, as does its lack of scalability.  It is also unsuitable
559	for realtime use, since it allows scheduling latency to "bleed" from
560	one read-side critical section to another.
561	
562	However, it is probably the easiest implementation to relate to, so is
563	a good starting point.
564	
565	It is extremely simple:
566	
567		static DEFINE_RWLOCK(rcu_gp_mutex);
568	
569		void rcu_read_lock(void)
570		{
571			read_lock(&rcu_gp_mutex);
572		}
573	
574		void rcu_read_unlock(void)
575		{
576			read_unlock(&rcu_gp_mutex);
577		}
578	
579		void synchronize_rcu(void)
580		{
581			write_lock(&rcu_gp_mutex);
582			write_unlock(&rcu_gp_mutex);
583		}
584	
585	[You can ignore rcu_assign_pointer() and rcu_dereference() without
586	missing much.  But here they are anyway.  And whatever you do, don't
587	forget about them when submitting patches making use of RCU!]
588	
589		#define rcu_assign_pointer(p, v)	({ \
590								smp_wmb(); \
591								(p) = (v); \
592							})
593	
594		#define rcu_dereference(p)     ({ \
595						typeof(p) _________p1 = p; \
596						smp_read_barrier_depends(); \
597						(_________p1); \
598						})
599	
600	
601	The rcu_read_lock() and rcu_read_unlock() primitive read-acquire
602	and release a global reader-writer lock.  The synchronize_rcu()
603	primitive write-acquires this same lock, then immediately releases
604	it.  This means that once synchronize_rcu() exits, all RCU read-side
605	critical sections that were in progress before synchronize_rcu() was
606	called are guaranteed to have completed -- there is no way that
607	synchronize_rcu() would have been able to write-acquire the lock
608	otherwise.
609	
610	It is possible to nest rcu_read_lock(), since reader-writer locks may
611	be recursively acquired.  Note also that rcu_read_lock() is immune
612	from deadlock (an important property of RCU).  The reason for this is
613	that the only thing that can block rcu_read_lock() is a synchronize_rcu().
614	But synchronize_rcu() does not acquire any locks while holding rcu_gp_mutex,
615	so there can be no deadlock cycle.
616	
617	Quick Quiz #1:	Why is this argument naive?  How could a deadlock
618			occur when using this algorithm in a real-world Linux
619			kernel?  How could this deadlock be avoided?
620	
621	
622	5B.  "TOY" EXAMPLE #2: CLASSIC RCU
623	
624	This section presents a "toy" RCU implementation that is based on
625	"classic RCU".  It is also short on performance (but only for updates) and
626	on features such as hotplug CPU and the ability to run in CONFIG_PREEMPT
627	kernels.  The definitions of rcu_dereference() and rcu_assign_pointer()
628	are the same as those shown in the preceding section, so they are omitted.
629	
630		void rcu_read_lock(void) { }
631	
632		void rcu_read_unlock(void) { }
633	
634		void synchronize_rcu(void)
635		{
636			int cpu;
637	
638			for_each_possible_cpu(cpu)
639				run_on(cpu);
640		}
641	
642	Note that rcu_read_lock() and rcu_read_unlock() do absolutely nothing.
643	This is the great strength of classic RCU in a non-preemptive kernel:
644	read-side overhead is precisely zero, at least on non-Alpha CPUs.
645	And there is absolutely no way that rcu_read_lock() can possibly
646	participate in a deadlock cycle!
647	
648	The implementation of synchronize_rcu() simply schedules itself on each
649	CPU in turn.  The run_on() primitive can be implemented straightforwardly
650	in terms of the sched_setaffinity() primitive.  Of course, a somewhat less
651	"toy" implementation would restore the affinity upon completion rather
652	than just leaving all tasks running on the last CPU, but when I said
653	"toy", I meant -toy-!
654	
655	So how the heck is this supposed to work???
656	
657	Remember that it is illegal to block while in an RCU read-side critical
658	section.  Therefore, if a given CPU executes a context switch, we know
659	that it must have completed all preceding RCU read-side critical sections.
660	Once -all- CPUs have executed a context switch, then -all- preceding
661	RCU read-side critical sections will have completed.
662	
663	So, suppose that we remove a data item from its structure and then invoke
664	synchronize_rcu().  Once synchronize_rcu() returns, we are guaranteed
665	that there are no RCU read-side critical sections holding a reference
666	to that data item, so we can safely reclaim it.
667	
668	Quick Quiz #2:	Give an example where Classic RCU's read-side
669			overhead is -negative-.
670	
671	Quick Quiz #3:  If it is illegal to block in an RCU read-side
672			critical section, what the heck do you do in
673			PREEMPT_RT, where normal spinlocks can block???
674	
675	
676	6.  ANALOGY WITH READER-WRITER LOCKING
677	
678	Although RCU can be used in many different ways, a very common use of
679	RCU is analogous to reader-writer locking.  The following unified
680	diff shows how closely related RCU and reader-writer locking can be.
681	
682		@@ -13,15 +14,15 @@
683			struct list_head *lp;
684			struct el *p;
685	
686		-	read_lock();
687		-	list_for_each_entry(p, head, lp) {
688		+	rcu_read_lock();
689		+	list_for_each_entry_rcu(p, head, lp) {
690				if (p->key == key) {
691					*result = p->data;
692		-			read_unlock();
693		+			rcu_read_unlock();
694					return 1;
695				}
696			}
697		-	read_unlock();
698		+	rcu_read_unlock();
699			return 0;
700		 }
701	
702		@@ -29,15 +30,16 @@
703		 {
704			struct el *p;
705	
706		-	write_lock(&listmutex);
707		+	spin_lock(&listmutex);
708			list_for_each_entry(p, head, lp) {
709				if (p->key == key) {
710		-			list_del(&p->list);
711		-			write_unlock(&listmutex);
712		+			list_del_rcu(&p->list);
713		+			spin_unlock(&listmutex);
714		+			synchronize_rcu();
715					kfree(p);
716					return 1;
717				}
718			}
719		-	write_unlock(&listmutex);
720		+	spin_unlock(&listmutex);
721			return 0;
722		 }
723	
724	Or, for those who prefer a side-by-side listing:
725	
726	 1 struct el {                          1 struct el {
727	 2   struct list_head list;             2   struct list_head list;
728	 3   long key;                          3   long key;
729	 4   spinlock_t mutex;                  4   spinlock_t mutex;
730	 5   int data;                          5   int data;
731	 6   /* Other data fields */            6   /* Other data fields */
732	 7 };                                   7 };
733	 8 spinlock_t listmutex;                8 spinlock_t listmutex;
734	 9 struct el head;                      9 struct el head;
735	
736	 1 int search(long key, int *result)    1 int search(long key, int *result)
737	 2 {                                    2 {
738	 3   struct list_head *lp;              3   struct list_head *lp;
739	 4   struct el *p;                      4   struct el *p;
740	 5                                      5
741	 6   read_lock();                       6   rcu_read_lock();
742	 7   list_for_each_entry(p, head, lp) { 7   list_for_each_entry_rcu(p, head, lp) {
743	 8     if (p->key == key) {             8     if (p->key == key) {
744	 9       *result = p->data;             9       *result = p->data;
745	10       read_unlock();                10       rcu_read_unlock();
746	11       return 1;                     11       return 1;
747	12     }                               12     }
748	13   }                                 13   }
749	14   read_unlock();                    14   rcu_read_unlock();
750	15   return 0;                         15   return 0;
751	16 }                                   16 }
752	
753	 1 int delete(long key)                 1 int delete(long key)
754	 2 {                                    2 {
755	 3   struct el *p;                      3   struct el *p;
756	 4                                      4
757	 5   write_lock(&listmutex);            5   spin_lock(&listmutex);
758	 6   list_for_each_entry(p, head, lp) { 6   list_for_each_entry(p, head, lp) {
759	 7     if (p->key == key) {             7     if (p->key == key) {
760	 8       list_del(&p->list);            8       list_del_rcu(&p->list);
761	 9       write_unlock(&listmutex);      9       spin_unlock(&listmutex);
762	                                       10       synchronize_rcu();
763	10       kfree(p);                     11       kfree(p);
764	11       return 1;                     12       return 1;
765	12     }                               13     }
766	13   }                                 14   }
767	14   write_unlock(&listmutex);         15   spin_unlock(&listmutex);
768	15   return 0;                         16   return 0;
769	16 }                                   17 }
770	
771	Either way, the differences are quite small.  Read-side locking moves
772	to rcu_read_lock() and rcu_read_unlock, update-side locking moves from
773	a reader-writer lock to a simple spinlock, and a synchronize_rcu()
774	precedes the kfree().
775	
776	However, there is one potential catch: the read-side and update-side
777	critical sections can now run concurrently.  In many cases, this will
778	not be a problem, but it is necessary to check carefully regardless.
779	For example, if multiple independent list updates must be seen as
780	a single atomic update, converting to RCU will require special care.
781	
782	Also, the presence of synchronize_rcu() means that the RCU version of
783	delete() can now block.  If this is a problem, there is a callback-based
784	mechanism that never blocks, namely call_rcu() or kfree_rcu(), that can
785	be used in place of synchronize_rcu().
786	
787	
788	7.  FULL LIST OF RCU APIs
789	
790	The RCU APIs are documented in docbook-format header comments in the
791	Linux-kernel source code, but it helps to have a full list of the
792	APIs, since there does not appear to be a way to categorize them
793	in docbook.  Here is the list, by category.
794	
795	RCU list traversal:
796	
797		list_for_each_entry_rcu
798		hlist_for_each_entry_rcu
799		hlist_nulls_for_each_entry_rcu
800		list_for_each_entry_continue_rcu
801	
802	RCU pointer/list update:
803	
804		rcu_assign_pointer
805		list_add_rcu
806		list_add_tail_rcu
807		list_del_rcu
808		list_replace_rcu
809		hlist_del_rcu
810		hlist_add_after_rcu
811		hlist_add_before_rcu
812		hlist_add_head_rcu
813		hlist_replace_rcu
814		list_splice_init_rcu()
815	
816	RCU:	Critical sections	Grace period		Barrier
817	
818		rcu_read_lock		synchronize_net		rcu_barrier
819		rcu_read_unlock		synchronize_rcu
820		rcu_dereference		synchronize_rcu_expedited
821					call_rcu
822					kfree_rcu
823	
824	
825	bh:	Critical sections	Grace period		Barrier
826	
827		rcu_read_lock_bh	call_rcu_bh		rcu_barrier_bh
828		rcu_read_unlock_bh	synchronize_rcu_bh
829		rcu_dereference_bh	synchronize_rcu_bh_expedited
830	
831	
832	sched:	Critical sections	Grace period		Barrier
833	
834		rcu_read_lock_sched	synchronize_sched	rcu_barrier_sched
835		rcu_read_unlock_sched	call_rcu_sched
836		[preempt_disable]	synchronize_sched_expedited
837		[and friends]
838		rcu_dereference_sched
839	
840	
841	SRCU:	Critical sections	Grace period		Barrier
842	
843		srcu_read_lock		synchronize_srcu	srcu_barrier
844		srcu_read_unlock	call_srcu
845		srcu_dereference	synchronize_srcu_expedited
846	
847	SRCU:	Initialization/cleanup
848		init_srcu_struct
849		cleanup_srcu_struct
850	
851	All:  lockdep-checked RCU-protected pointer access
852	
853		rcu_dereference_check
854		rcu_dereference_protected
855		rcu_access_pointer
856	
857	See the comment headers in the source code (or the docbook generated
858	from them) for more information.
859	
860	However, given that there are no fewer than four families of RCU APIs
861	in the Linux kernel, how do you choose which one to use?  The following
862	list can be helpful:
863	
864	a.	Will readers need to block?  If so, you need SRCU.
865	
866	b.	What about the -rt patchset?  If readers would need to block
867		in an non-rt kernel, you need SRCU.  If readers would block
868		in a -rt kernel, but not in a non-rt kernel, SRCU is not
869		necessary.
870	
871	c.	Do you need to treat NMI handlers, hardirq handlers,
872		and code segments with preemption disabled (whether
873		via preempt_disable(), local_irq_save(), local_bh_disable(),
874		or some other mechanism) as if they were explicit RCU readers?
875		If so, RCU-sched is the only choice that will work for you.
876	
877	d.	Do you need RCU grace periods to complete even in the face
878		of softirq monopolization of one or more of the CPUs?  For
879		example, is your code subject to network-based denial-of-service
880		attacks?  If so, you need RCU-bh.
881	
882	e.	Is your workload too update-intensive for normal use of
883		RCU, but inappropriate for other synchronization mechanisms?
884		If so, consider SLAB_DESTROY_BY_RCU.  But please be careful!
885	
886	f.	Do you need read-side critical sections that are respected
887		even though they are in the middle of the idle loop, during
888		user-mode execution, or on an offlined CPU?  If so, SRCU is the
889		only choice that will work for you.
890	
891	g.	Otherwise, use RCU.
892	
893	Of course, this all assumes that you have determined that RCU is in fact
894	the right tool for your job.
895	
896	
897	8.  ANSWERS TO QUICK QUIZZES
898	
899	Quick Quiz #1:	Why is this argument naive?  How could a deadlock
900			occur when using this algorithm in a real-world Linux
901			kernel?  [Referring to the lock-based "toy" RCU
902			algorithm.]
903	
904	Answer:		Consider the following sequence of events:
905	
906			1.	CPU 0 acquires some unrelated lock, call it
907				"problematic_lock", disabling irq via
908				spin_lock_irqsave().
909	
910			2.	CPU 1 enters synchronize_rcu(), write-acquiring
911				rcu_gp_mutex.
912	
913			3.	CPU 0 enters rcu_read_lock(), but must wait
914				because CPU 1 holds rcu_gp_mutex.
915	
916			4.	CPU 1 is interrupted, and the irq handler
917				attempts to acquire problematic_lock.
918	
919			The system is now deadlocked.
920	
921			One way to avoid this deadlock is to use an approach like
922			that of CONFIG_PREEMPT_RT, where all normal spinlocks
923			become blocking locks, and all irq handlers execute in
924			the context of special tasks.  In this case, in step 4
925			above, the irq handler would block, allowing CPU 1 to
926			release rcu_gp_mutex, avoiding the deadlock.
927	
928			Even in the absence of deadlock, this RCU implementation
929			allows latency to "bleed" from readers to other
930			readers through synchronize_rcu().  To see this,
931			consider task A in an RCU read-side critical section
932			(thus read-holding rcu_gp_mutex), task B blocked
933			attempting to write-acquire rcu_gp_mutex, and
934			task C blocked in rcu_read_lock() attempting to
935			read_acquire rcu_gp_mutex.  Task A's RCU read-side
936			latency is holding up task C, albeit indirectly via
937			task B.
938	
939			Realtime RCU implementations therefore use a counter-based
940			approach where tasks in RCU read-side critical sections
941			cannot be blocked by tasks executing synchronize_rcu().
942	
943	Quick Quiz #2:	Give an example where Classic RCU's read-side
944			overhead is -negative-.
945	
946	Answer:		Imagine a single-CPU system with a non-CONFIG_PREEMPT
947			kernel where a routing table is used by process-context
948			code, but can be updated by irq-context code (for example,
949			by an "ICMP REDIRECT" packet).	The usual way of handling
950			this would be to have the process-context code disable
951			interrupts while searching the routing table.  Use of
952			RCU allows such interrupt-disabling to be dispensed with.
953			Thus, without RCU, you pay the cost of disabling interrupts,
954			and with RCU you don't.
955	
956			One can argue that the overhead of RCU in this
957			case is negative with respect to the single-CPU
958			interrupt-disabling approach.  Others might argue that
959			the overhead of RCU is merely zero, and that replacing
960			the positive overhead of the interrupt-disabling scheme
961			with the zero-overhead RCU scheme does not constitute
962			negative overhead.
963	
964			In real life, of course, things are more complex.  But
965			even the theoretical possibility of negative overhead for
966			a synchronization primitive is a bit unexpected.  ;-)
967	
968	Quick Quiz #3:  If it is illegal to block in an RCU read-side
969			critical section, what the heck do you do in
970			PREEMPT_RT, where normal spinlocks can block???
971	
972	Answer:		Just as PREEMPT_RT permits preemption of spinlock
973			critical sections, it permits preemption of RCU
974			read-side critical sections.  It also permits
975			spinlocks blocking while in RCU read-side critical
976			sections.
977	
978			Why the apparent inconsistency?  Because it is it
979			possible to use priority boosting to keep the RCU
980			grace periods short if need be (for example, if running
981			short of memory).  In contrast, if blocking waiting
982			for (say) network reception, there is no way to know
983			what should be boosted.  Especially given that the
984			process we need to boost might well be a human being
985			who just went out for a pizza or something.  And although
986			a computer-operated cattle prod might arouse serious
987			interest, it might also provoke serious objections.
988			Besides, how does the computer know what pizza parlor
989			the human being went to???
990	
991	
992	ACKNOWLEDGEMENTS
993	
994	My thanks to the people who helped make this human-readable, including
995	Jon Walpole, Josh Triplett, Serge Hallyn, Suzanne Wood, and Alan Stern.
996	
997	
998	For more information, see http://www.rdrop.com/users/paulmck/RCU.
Hide Line Numbers
About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Information is copyright its respective author. All material is available from the Linux Kernel Source distributed under a GPL License. This page is provided as a free service by mjmwired.net.