1 .. _whatisrcu_doc: 1 .. _whatisrcu_doc:
2 2
3 What is RCU? -- "Read, Copy, Update" 3 What is RCU? -- "Read, Copy, Update"
4 ====================================== 4 ======================================
5 5
6 Please note that the "What is RCU?" LWN series 6 Please note that the "What is RCU?" LWN series is an excellent place
7 to start learning about RCU: 7 to start learning about RCU:
8 8
9 | 1. What is RCU, Fundamentally? https://l 9 | 1. What is RCU, Fundamentally? https://lwn.net/Articles/262464/
10 | 2. What is RCU? Part 2: Usage https://l 10 | 2. What is RCU? Part 2: Usage https://lwn.net/Articles/263130/
11 | 3. RCU part 3: the RCU API https://l 11 | 3. RCU part 3: the RCU API https://lwn.net/Articles/264090/
12 | 4. The RCU API, 2010 Edition https://l 12 | 4. The RCU API, 2010 Edition https://lwn.net/Articles/418853/
13 | 2010 Big API Table https://l 13 | 2010 Big API Table https://lwn.net/Articles/419086/
14 | 5. The RCU API, 2014 Edition https://l 14 | 5. The RCU API, 2014 Edition https://lwn.net/Articles/609904/
15 | 2014 Big API Table https://l 15 | 2014 Big API Table https://lwn.net/Articles/609973/
16 | 6. The RCU API, 2019 Edition https://l 16 | 6. The RCU API, 2019 Edition https://lwn.net/Articles/777036/
17 | 2019 Big API Table https://l 17 | 2019 Big API Table https://lwn.net/Articles/777165/
18 18
19 For those preferring video: 19 For those preferring video:
20 20
21 | 1. Unraveling RCU Mysteries: Fundamentals 21 | 1. Unraveling RCU Mysteries: Fundamentals https://www.linuxfoundation.org/webinars/unraveling-rcu-usage-mysteries
22 | 2. Unraveling RCU Mysteries: Additional U 22 | 2. Unraveling RCU Mysteries: Additional Use Cases https://www.linuxfoundation.org/webinars/unraveling-rcu-usage-mysteries-additional-use-cases
23 23
24 24
25 What is RCU? 25 What is RCU?
26 26
27 RCU is a synchronization mechanism that was ad 27 RCU is a synchronization mechanism that was added to the Linux kernel
28 during the 2.5 development effort that is opti 28 during the 2.5 development effort that is optimized for read-mostly
29 situations. Although RCU is actually quite si 29 situations. Although RCU is actually quite simple, making effective use
30 of it requires you to think differently about 30 of it requires you to think differently about your code. Another part
31 of the problem is the mistaken assumption that 31 of the problem is the mistaken assumption that there is "one true way" to
32 describe and to use RCU. Instead, the experie 32 describe and to use RCU. Instead, the experience has been that different
33 people must take different paths to arrive at 33 people must take different paths to arrive at an understanding of RCU,
34 depending on their experiences and use cases. 34 depending on their experiences and use cases. This document provides
35 several different paths, as follows: 35 several different paths, as follows:
36 36
37 :ref:`1. RCU OVERVIEW <1_whatisRCU>` 37 :ref:`1. RCU OVERVIEW <1_whatisRCU>`
38 38
39 :ref:`2. WHAT IS RCU'S CORE API? <2_wha 39 :ref:`2. WHAT IS RCU'S CORE API? <2_whatisRCU>`
40 40
41 :ref:`3. WHAT ARE SOME EXAMPLE USES OF 41 :ref:`3. WHAT ARE SOME EXAMPLE USES OF CORE RCU API? <3_whatisRCU>`
42 42
43 :ref:`4. WHAT IF MY UPDATING THREAD CAN 43 :ref:`4. WHAT IF MY UPDATING THREAD CANNOT BLOCK? <4_whatisRCU>`
44 44
45 :ref:`5. WHAT ARE SOME SIMPLE IMPLEMENT 45 :ref:`5. WHAT ARE SOME SIMPLE IMPLEMENTATIONS OF RCU? <5_whatisRCU>`
46 46
47 :ref:`6. ANALOGY WITH READER-WRITER LOC 47 :ref:`6. ANALOGY WITH READER-WRITER LOCKING <6_whatisRCU>`
48 48
49 :ref:`7. ANALOGY WITH REFERENCE COUNTIN 49 :ref:`7. ANALOGY WITH REFERENCE COUNTING <7_whatisRCU>`
50 50
51 :ref:`8. FULL LIST OF RCU APIs <8_whati 51 :ref:`8. FULL LIST OF RCU APIs <8_whatisRCU>`
52 52
53 :ref:`9. ANSWERS TO QUICK QUIZZES <9_wh 53 :ref:`9. ANSWERS TO QUICK QUIZZES <9_whatisRCU>`
54 54
55 People who prefer starting with a conceptual o 55 People who prefer starting with a conceptual overview should focus on
56 Section 1, though most readers will profit by 56 Section 1, though most readers will profit by reading this section at
57 some point. People who prefer to start with a 57 some point. People who prefer to start with an API that they can then
58 experiment with should focus on Section 2. Pe 58 experiment with should focus on Section 2. People who prefer to start
59 with example uses should focus on Sections 3 a 59 with example uses should focus on Sections 3 and 4. People who need to
60 understand the RCU implementation should focus 60 understand the RCU implementation should focus on Section 5, then dive
61 into the kernel source code. People who reaso 61 into the kernel source code. People who reason best by analogy should
62 focus on Section 6 and 7. Section 8 serves as 62 focus on Section 6 and 7. Section 8 serves as an index to the docbook
63 API documentation, and Section 9 is the tradit 63 API documentation, and Section 9 is the traditional answer key.
64 64
65 So, start with the section that makes the most 65 So, start with the section that makes the most sense to you and your
66 preferred method of learning. If you need to 66 preferred method of learning. If you need to know everything about
67 everything, feel free to read the whole thing 67 everything, feel free to read the whole thing -- but if you are really
68 that type of person, you have perused the sour 68 that type of person, you have perused the source code and will therefore
69 never need this document anyway. ;-) 69 never need this document anyway. ;-)
70 70
71 .. _1_whatisRCU: 71 .. _1_whatisRCU:
72 72
73 1. RCU OVERVIEW 73 1. RCU OVERVIEW
74 ---------------- 74 ----------------
75 75
76 The basic idea behind RCU is to split updates 76 The basic idea behind RCU is to split updates into "removal" and
77 "reclamation" phases. The removal phase remov 77 "reclamation" phases. The removal phase removes references to data items
78 within a data structure (possibly by replacing 78 within a data structure (possibly by replacing them with references to
79 new versions of these data items), and can run 79 new versions of these data items), and can run concurrently with readers.
80 The reason that it is safe to run the removal 80 The reason that it is safe to run the removal phase concurrently with
81 readers is the semantics of modern CPUs guaran 81 readers is the semantics of modern CPUs guarantee that readers will see
82 either the old or the new version of the data 82 either the old or the new version of the data structure rather than a
83 partially updated reference. The reclamation 83 partially updated reference. The reclamation phase does the work of reclaiming
84 (e.g., freeing) the data items removed from th 84 (e.g., freeing) the data items removed from the data structure during the
85 removal phase. Because reclaiming data items 85 removal phase. Because reclaiming data items can disrupt any readers
86 concurrently referencing those data items, the 86 concurrently referencing those data items, the reclamation phase must
87 not start until readers no longer hold referen 87 not start until readers no longer hold references to those data items.
88 88
89 Splitting the update into removal and reclamat 89 Splitting the update into removal and reclamation phases permits the
90 updater to perform the removal phase immediate 90 updater to perform the removal phase immediately, and to defer the
91 reclamation phase until all readers active dur 91 reclamation phase until all readers active during the removal phase have
92 completed, either by blocking until they finis 92 completed, either by blocking until they finish or by registering a
93 callback that is invoked after they finish. O 93 callback that is invoked after they finish. Only readers that are active
94 during the removal phase need be considered, b 94 during the removal phase need be considered, because any reader starting
95 after the removal phase will be unable to gain 95 after the removal phase will be unable to gain a reference to the removed
96 data items, and therefore cannot be disrupted 96 data items, and therefore cannot be disrupted by the reclamation phase.
97 97
98 So the typical RCU update sequence goes someth 98 So the typical RCU update sequence goes something like the following:
99 99
100 a. Remove pointers to a data structure, s 100 a. Remove pointers to a data structure, so that subsequent
101 readers cannot gain a reference to it. 101 readers cannot gain a reference to it.
102 102
103 b. Wait for all previous readers to compl 103 b. Wait for all previous readers to complete their RCU read-side
104 critical sections. 104 critical sections.
105 105
106 c. At this point, there cannot be any rea 106 c. At this point, there cannot be any readers who hold references
107 to the data structure, so it now may s 107 to the data structure, so it now may safely be reclaimed
108 (e.g., kfree()d). 108 (e.g., kfree()d).
109 109
110 Step (b) above is the key idea underlying RCU' 110 Step (b) above is the key idea underlying RCU's deferred destruction.
111 The ability to wait until all readers are done 111 The ability to wait until all readers are done allows RCU readers to
112 use much lighter-weight synchronization, in so 112 use much lighter-weight synchronization, in some cases, absolutely no
113 synchronization at all. In contrast, in more 113 synchronization at all. In contrast, in more conventional lock-based
114 schemes, readers must use heavy-weight synchro 114 schemes, readers must use heavy-weight synchronization in order to
115 prevent an updater from deleting the data stru 115 prevent an updater from deleting the data structure out from under them.
116 This is because lock-based updaters typically 116 This is because lock-based updaters typically update data items in place,
117 and must therefore exclude readers. In contra 117 and must therefore exclude readers. In contrast, RCU-based updaters
118 typically take advantage of the fact that writ 118 typically take advantage of the fact that writes to single aligned
119 pointers are atomic on modern CPUs, allowing a 119 pointers are atomic on modern CPUs, allowing atomic insertion, removal,
120 and replacement of data items in a linked stru 120 and replacement of data items in a linked structure without disrupting
121 readers. Concurrent RCU readers can then cont 121 readers. Concurrent RCU readers can then continue accessing the old
122 versions, and can dispense with the atomic ope 122 versions, and can dispense with the atomic operations, memory barriers,
123 and communications cache misses that are so ex 123 and communications cache misses that are so expensive on present-day
124 SMP computer systems, even in absence of lock 124 SMP computer systems, even in absence of lock contention.
125 125
126 In the three-step procedure shown above, the u 126 In the three-step procedure shown above, the updater is performing both
127 the removal and the reclamation step, but it i 127 the removal and the reclamation step, but it is often helpful for an
128 entirely different thread to do the reclamatio 128 entirely different thread to do the reclamation, as is in fact the case
129 in the Linux kernel's directory-entry cache (d 129 in the Linux kernel's directory-entry cache (dcache). Even if the same
130 thread performs both the update step (step (a) 130 thread performs both the update step (step (a) above) and the reclamation
131 step (step (c) above), it is often helpful to 131 step (step (c) above), it is often helpful to think of them separately.
132 For example, RCU readers and updaters need not 132 For example, RCU readers and updaters need not communicate at all,
133 but RCU provides implicit low-overhead communi 133 but RCU provides implicit low-overhead communication between readers
134 and reclaimers, namely, in step (b) above. 134 and reclaimers, namely, in step (b) above.
135 135
136 So how the heck can a reclaimer tell when a re 136 So how the heck can a reclaimer tell when a reader is done, given
137 that readers are not doing any sort of synchro 137 that readers are not doing any sort of synchronization operations???
138 Read on to learn about how RCU's API makes thi 138 Read on to learn about how RCU's API makes this easy.
139 139
140 .. _2_whatisRCU: 140 .. _2_whatisRCU:
141 141
142 2. WHAT IS RCU'S CORE API? 142 2. WHAT IS RCU'S CORE API?
143 --------------------------- 143 ---------------------------
144 144
145 The core RCU API is quite small: 145 The core RCU API is quite small:
146 146
147 a. rcu_read_lock() 147 a. rcu_read_lock()
148 b. rcu_read_unlock() 148 b. rcu_read_unlock()
149 c. synchronize_rcu() / call_rcu() 149 c. synchronize_rcu() / call_rcu()
150 d. rcu_assign_pointer() 150 d. rcu_assign_pointer()
151 e. rcu_dereference() 151 e. rcu_dereference()
152 152
153 There are many other members of the RCU API, b 153 There are many other members of the RCU API, but the rest can be
154 expressed in terms of these five, though most 154 expressed in terms of these five, though most implementations instead
155 express synchronize_rcu() in terms of the call 155 express synchronize_rcu() in terms of the call_rcu() callback API.
156 156
157 The five core RCU APIs are described below, th 157 The five core RCU APIs are described below, the other 18 will be enumerated
158 later. See the kernel docbook documentation f 158 later. See the kernel docbook documentation for more info, or look directly
159 at the function header comments. 159 at the function header comments.
160 160
161 rcu_read_lock() 161 rcu_read_lock()
162 ^^^^^^^^^^^^^^^ 162 ^^^^^^^^^^^^^^^
163 void rcu_read_lock(void); 163 void rcu_read_lock(void);
164 164
165 This temporal primitive is used by a r 165 This temporal primitive is used by a reader to inform the
166 reclaimer that the reader is entering 166 reclaimer that the reader is entering an RCU read-side critical
167 section. It is illegal to block while 167 section. It is illegal to block while in an RCU read-side
168 critical section, though kernels built 168 critical section, though kernels built with CONFIG_PREEMPT_RCU
169 can preempt RCU read-side critical sec 169 can preempt RCU read-side critical sections. Any RCU-protected
170 data structure accessed during an RCU 170 data structure accessed during an RCU read-side critical section
171 is guaranteed to remain unreclaimed fo 171 is guaranteed to remain unreclaimed for the full duration of that
172 critical section. Reference counts ma 172 critical section. Reference counts may be used in conjunction
173 with RCU to maintain longer-term refer 173 with RCU to maintain longer-term references to data structures.
174 174
175 Note that anything that disables botto 175 Note that anything that disables bottom halves, preemption,
176 or interrupts also enters an RCU read- 176 or interrupts also enters an RCU read-side critical section.
177 Acquiring a spinlock also enters an RC 177 Acquiring a spinlock also enters an RCU read-side critical
178 sections, even for spinlocks that do n 178 sections, even for spinlocks that do not disable preemption,
179 as is the case in kernels built with C 179 as is the case in kernels built with CONFIG_PREEMPT_RT=y.
180 Sleeplocks do *not* enter RCU read-sid 180 Sleeplocks do *not* enter RCU read-side critical sections.
181 181
182 rcu_read_unlock() 182 rcu_read_unlock()
183 ^^^^^^^^^^^^^^^^^ 183 ^^^^^^^^^^^^^^^^^
184 void rcu_read_unlock(void); 184 void rcu_read_unlock(void);
185 185
186 This temporal primitives is used by a 186 This temporal primitives is used by a reader to inform the
187 reclaimer that the reader is exiting a 187 reclaimer that the reader is exiting an RCU read-side critical
188 section. Anything that enables bottom 188 section. Anything that enables bottom halves, preemption,
189 or interrupts also exits an RCU read-s 189 or interrupts also exits an RCU read-side critical section.
190 Releasing a spinlock also exits an RCU 190 Releasing a spinlock also exits an RCU read-side critical section.
191 191
192 Note that RCU read-side critical secti 192 Note that RCU read-side critical sections may be nested and/or
193 overlapping. 193 overlapping.
194 194
195 synchronize_rcu() 195 synchronize_rcu()
196 ^^^^^^^^^^^^^^^^^ 196 ^^^^^^^^^^^^^^^^^
197 void synchronize_rcu(void); 197 void synchronize_rcu(void);
198 198
199 This temporal primitive marks the end 199 This temporal primitive marks the end of updater code and the
200 beginning of reclaimer code. It does 200 beginning of reclaimer code. It does this by blocking until
201 all pre-existing RCU read-side critica 201 all pre-existing RCU read-side critical sections on all CPUs
202 have completed. Note that synchronize 202 have completed. Note that synchronize_rcu() will **not**
203 necessarily wait for any subsequent RC 203 necessarily wait for any subsequent RCU read-side critical
204 sections to complete. For example, co 204 sections to complete. For example, consider the following
205 sequence of events:: 205 sequence of events::
206 206
207 CPU 0 CPU 1 207 CPU 0 CPU 1 CPU 2
208 ----------------- --------------- 208 ----------------- ------------------------- ---------------
209 1. rcu_read_lock() 209 1. rcu_read_lock()
210 2. enters synchron 210 2. enters synchronize_rcu()
211 3. 211 3. rcu_read_lock()
212 4. rcu_read_unlock() 212 4. rcu_read_unlock()
213 5. exits synchron 213 5. exits synchronize_rcu()
214 6. 214 6. rcu_read_unlock()
215 215
216 To reiterate, synchronize_rcu() waits 216 To reiterate, synchronize_rcu() waits only for ongoing RCU
217 read-side critical sections to complet 217 read-side critical sections to complete, not necessarily for
218 any that begin after synchronize_rcu() 218 any that begin after synchronize_rcu() is invoked.
219 219
220 Of course, synchronize_rcu() does not 220 Of course, synchronize_rcu() does not necessarily return
221 **immediately** after the last pre-exi 221 **immediately** after the last pre-existing RCU read-side critical
222 section completes. For one thing, the 222 section completes. For one thing, there might well be scheduling
223 delays. For another thing, many RCU i 223 delays. For another thing, many RCU implementations process
224 requests in batches in order to improv 224 requests in batches in order to improve efficiencies, which can
225 further delay synchronize_rcu(). 225 further delay synchronize_rcu().
226 226
227 Since synchronize_rcu() is the API tha 227 Since synchronize_rcu() is the API that must figure out when
228 readers are done, its implementation i 228 readers are done, its implementation is key to RCU. For RCU
229 to be useful in all but the most read- 229 to be useful in all but the most read-intensive situations,
230 synchronize_rcu()'s overhead must also 230 synchronize_rcu()'s overhead must also be quite small.
231 231
232 The call_rcu() API is an asynchronous 232 The call_rcu() API is an asynchronous callback form of
233 synchronize_rcu(), and is described in 233 synchronize_rcu(), and is described in more detail in a later
234 section. Instead of blocking, it regi 234 section. Instead of blocking, it registers a function and
235 argument which are invoked after all o 235 argument which are invoked after all ongoing RCU read-side
236 critical sections have completed. Thi 236 critical sections have completed. This callback variant is
237 particularly useful in situations wher 237 particularly useful in situations where it is illegal to block
238 or where update-side performance is cr 238 or where update-side performance is critically important.
239 239
240 However, the call_rcu() API should not 240 However, the call_rcu() API should not be used lightly, as use
241 of the synchronize_rcu() API generally 241 of the synchronize_rcu() API generally results in simpler code.
242 In addition, the synchronize_rcu() API 242 In addition, the synchronize_rcu() API has the nice property
243 of automatically limiting update rate 243 of automatically limiting update rate should grace periods
244 be delayed. This property results in 244 be delayed. This property results in system resilience in face
245 of denial-of-service attacks. Code us 245 of denial-of-service attacks. Code using call_rcu() should limit
246 update rate in order to gain this same 246 update rate in order to gain this same sort of resilience. See
247 checklist.rst for some approaches to l 247 checklist.rst for some approaches to limiting the update rate.
248 248
249 rcu_assign_pointer() 249 rcu_assign_pointer()
250 ^^^^^^^^^^^^^^^^^^^^ 250 ^^^^^^^^^^^^^^^^^^^^
251 void rcu_assign_pointer(p, typeof(p) v 251 void rcu_assign_pointer(p, typeof(p) v);
252 252
253 Yes, rcu_assign_pointer() **is** imple !! 253 Yes, rcu_assign_pointer() **is** implemented as a macro, though it
254 it would be cool to be able to declare !! 254 would be cool to be able to declare a function in this manner.
255 (And there has been some discussion of !! 255 (Compiler experts will no doubt disagree.)
256 to the C language, so who knows?) <<
257 256
258 The updater uses this spatial macro to 257 The updater uses this spatial macro to assign a new value to an
259 RCU-protected pointer, in order to saf 258 RCU-protected pointer, in order to safely communicate the change
260 in value from the updater to the reade 259 in value from the updater to the reader. This is a spatial (as
261 opposed to temporal) macro. It does n 260 opposed to temporal) macro. It does not evaluate to an rvalue,
262 but it does provide any compiler direc !! 261 but it does execute any memory-barrier instructions required
263 instructions required for a given comp !! 262 for a given CPU architecture. Its ordering properties are that
264 Its ordering properties are that of a !! 263 of a store-release operation.
265 that is, any prior loads and stores re <<
266 structure are ordered before the store <<
267 to that structure. <<
268 264
269 Perhaps just as important, rcu_assign_ !! 265 Perhaps just as important, it serves to document (1) which
270 (1) which pointers are protected by RC !! 266 pointers are protected by RCU and (2) the point at which a
271 a given structure becomes accessible t !! 267 given structure becomes accessible to other CPUs. That said,
272 rcu_assign_pointer() is most frequentl 268 rcu_assign_pointer() is most frequently used indirectly, via
273 the _rcu list-manipulation primitives 269 the _rcu list-manipulation primitives such as list_add_rcu().
274 270
275 rcu_dereference() 271 rcu_dereference()
276 ^^^^^^^^^^^^^^^^^ 272 ^^^^^^^^^^^^^^^^^
277 typeof(p) rcu_dereference(p); 273 typeof(p) rcu_dereference(p);
278 274
279 Like rcu_assign_pointer(), rcu_derefer 275 Like rcu_assign_pointer(), rcu_dereference() must be implemented
280 as a macro. 276 as a macro.
281 277
282 The reader uses the spatial rcu_derefe 278 The reader uses the spatial rcu_dereference() macro to fetch
283 an RCU-protected pointer, which return 279 an RCU-protected pointer, which returns a value that may
284 then be safely dereferenced. Note tha 280 then be safely dereferenced. Note that rcu_dereference()
285 does not actually dereference the poin 281 does not actually dereference the pointer, instead, it
286 protects the pointer for later derefer 282 protects the pointer for later dereferencing. It also
287 executes any needed memory-barrier ins 283 executes any needed memory-barrier instructions for a given
288 CPU architecture. Currently, only Alp 284 CPU architecture. Currently, only Alpha needs memory barriers
289 within rcu_dereference() -- on other C 285 within rcu_dereference() -- on other CPUs, it compiles to a
290 volatile load. However, no mainstream !! 286 volatile load.
291 address dependencies, so rcu_dereferen <<
292 which, in combination with the coding <<
293 rcu_dereference.rst, prevent current c <<
294 these dependencies. <<
295 287
296 Common coding practice uses rcu_derefe 288 Common coding practice uses rcu_dereference() to copy an
297 RCU-protected pointer to a local varia 289 RCU-protected pointer to a local variable, then dereferences
298 this local variable, for example as fo 290 this local variable, for example as follows::
299 291
300 p = rcu_dereference(head.next) 292 p = rcu_dereference(head.next);
301 return p->data; 293 return p->data;
302 294
303 However, in this case, one could just 295 However, in this case, one could just as easily combine these
304 into one statement:: 296 into one statement::
305 297
306 return rcu_dereference(head.ne 298 return rcu_dereference(head.next)->data;
307 299
308 If you are going to be fetching multip 300 If you are going to be fetching multiple fields from the
309 RCU-protected structure, using the loc 301 RCU-protected structure, using the local variable is of
310 course preferred. Repeated rcu_derefe 302 course preferred. Repeated rcu_dereference() calls look
311 ugly, do not guarantee that the same p 303 ugly, do not guarantee that the same pointer will be returned
312 if an update happened while in the cri 304 if an update happened while in the critical section, and incur
313 unnecessary overhead on Alpha CPUs. 305 unnecessary overhead on Alpha CPUs.
314 306
315 Note that the value returned by rcu_de 307 Note that the value returned by rcu_dereference() is valid
316 only within the enclosing RCU read-sid 308 only within the enclosing RCU read-side critical section [1]_.
317 For example, the following is **not** 309 For example, the following is **not** legal::
318 310
319 rcu_read_lock(); 311 rcu_read_lock();
320 p = rcu_dereference(head.next) 312 p = rcu_dereference(head.next);
321 rcu_read_unlock(); 313 rcu_read_unlock();
322 x = p->address; /* BUG!!! */ 314 x = p->address; /* BUG!!! */
323 rcu_read_lock(); 315 rcu_read_lock();
324 y = p->data; /* BUG!!! */ 316 y = p->data; /* BUG!!! */
325 rcu_read_unlock(); 317 rcu_read_unlock();
326 318
327 Holding a reference from one RCU read- 319 Holding a reference from one RCU read-side critical section
328 to another is just as illegal as holdi 320 to another is just as illegal as holding a reference from
329 one lock-based critical section to ano 321 one lock-based critical section to another! Similarly,
330 using a reference outside of the criti 322 using a reference outside of the critical section in which
331 it was acquired is just as illegal as 323 it was acquired is just as illegal as doing so with normal
332 locking. 324 locking.
333 325
334 As with rcu_assign_pointer(), an impor 326 As with rcu_assign_pointer(), an important function of
335 rcu_dereference() is to document which 327 rcu_dereference() is to document which pointers are protected by
336 RCU, in particular, flagging a pointer 328 RCU, in particular, flagging a pointer that is subject to changing
337 at any time, including immediately aft 329 at any time, including immediately after the rcu_dereference().
338 And, again like rcu_assign_pointer(), 330 And, again like rcu_assign_pointer(), rcu_dereference() is
339 typically used indirectly, via the _rc 331 typically used indirectly, via the _rcu list-manipulation
340 primitives, such as list_for_each_entr 332 primitives, such as list_for_each_entry_rcu() [2]_.
341 333
342 .. [1] The variant rcu_dereference_protec 334 .. [1] The variant rcu_dereference_protected() can be used outside
343 of an RCU read-side critical section a 335 of an RCU read-side critical section as long as the usage is
344 protected by locks acquired by the upd 336 protected by locks acquired by the update-side code. This variant
345 avoids the lockdep warning that would 337 avoids the lockdep warning that would happen when using (for
346 example) rcu_dereference() without rcu 338 example) rcu_dereference() without rcu_read_lock() protection.
347 Using rcu_dereference_protected() also 339 Using rcu_dereference_protected() also has the advantage
348 of permitting compiler optimizations t 340 of permitting compiler optimizations that rcu_dereference()
349 must prohibit. The rcu_dereference_pr 341 must prohibit. The rcu_dereference_protected() variant takes
350 a lockdep expression to indicate which 342 a lockdep expression to indicate which locks must be acquired
351 by the caller. If the indicated protec 343 by the caller. If the indicated protection is not provided,
352 a lockdep splat is emitted. See Desig 344 a lockdep splat is emitted. See Design/Requirements/Requirements.rst
353 and the API's code comments for more d 345 and the API's code comments for more details and example usage.
354 346
355 .. [2] If the list_for_each_entry_rcu() i 347 .. [2] If the list_for_each_entry_rcu() instance might be used by
356 update-side code as well as by RCU rea 348 update-side code as well as by RCU readers, then an additional
357 lockdep expression can be added to its 349 lockdep expression can be added to its list of arguments.
358 For example, given an additional "lock 350 For example, given an additional "lock_is_held(&mylock)" argument,
359 the RCU lockdep code would complain on 351 the RCU lockdep code would complain only if this instance was
360 invoked outside of an RCU read-side cr 352 invoked outside of an RCU read-side critical section and without
361 the protection of mylock. 353 the protection of mylock.
362 354
363 The following diagram shows how each API commu 355 The following diagram shows how each API communicates among the
364 reader, updater, and reclaimer. 356 reader, updater, and reclaimer.
365 :: 357 ::
366 358
367 359
368 rcu_assign_pointer() 360 rcu_assign_pointer()
369 +--------+ 361 +--------+
370 +---------------------->| reader | 362 +---------------------->| reader |---------+
371 | +--------+ 363 | +--------+ |
372 | | 364 | | |
373 | | 365 | | | Protect:
374 | | 366 | | | rcu_read_lock()
375 | | 367 | | | rcu_read_unlock()
376 | rcu_dereference() | 368 | rcu_dereference() | |
377 +---------+ | 369 +---------+ | |
378 | updater |<----------------+ 370 | updater |<----------------+ |
379 +---------+ 371 +---------+ V
380 | 372 | +-----------+
381 +--------------------------------- 373 +----------------------------------->| reclaimer |
382 374 +-----------+
383 Defer: 375 Defer:
384 synchronize_rcu() & call_rcu() 376 synchronize_rcu() & call_rcu()
385 377
386 378
387 The RCU infrastructure observes the temporal s 379 The RCU infrastructure observes the temporal sequence of rcu_read_lock(),
388 rcu_read_unlock(), synchronize_rcu(), and call 380 rcu_read_unlock(), synchronize_rcu(), and call_rcu() invocations in
389 order to determine when (1) synchronize_rcu() 381 order to determine when (1) synchronize_rcu() invocations may return
390 to their callers and (2) call_rcu() callbacks 382 to their callers and (2) call_rcu() callbacks may be invoked. Efficient
391 implementations of the RCU infrastructure make 383 implementations of the RCU infrastructure make heavy use of batching in
392 order to amortize their overhead over many use 384 order to amortize their overhead over many uses of the corresponding APIs.
393 The rcu_assign_pointer() and rcu_dereference() 385 The rcu_assign_pointer() and rcu_dereference() invocations communicate
394 spatial changes via stores to and loads from t 386 spatial changes via stores to and loads from the RCU-protected pointer in
395 question. 387 question.
396 388
397 There are at least three flavors of RCU usage 389 There are at least three flavors of RCU usage in the Linux kernel. The diagram
398 above shows the most common one. On the update 390 above shows the most common one. On the updater side, the rcu_assign_pointer(),
399 synchronize_rcu() and call_rcu() primitives us 391 synchronize_rcu() and call_rcu() primitives used are the same for all three
400 flavors. However for protection (on the reader 392 flavors. However for protection (on the reader side), the primitives used vary
401 depending on the flavor: 393 depending on the flavor:
402 394
403 a. rcu_read_lock() / rcu_read_unlock() 395 a. rcu_read_lock() / rcu_read_unlock()
404 rcu_dereference() 396 rcu_dereference()
405 397
406 b. rcu_read_lock_bh() / rcu_read_unlock_b 398 b. rcu_read_lock_bh() / rcu_read_unlock_bh()
407 local_bh_disable() / local_bh_enable() 399 local_bh_disable() / local_bh_enable()
408 rcu_dereference_bh() 400 rcu_dereference_bh()
409 401
410 c. rcu_read_lock_sched() / rcu_read_unloc 402 c. rcu_read_lock_sched() / rcu_read_unlock_sched()
411 preempt_disable() / preempt_enable() 403 preempt_disable() / preempt_enable()
412 local_irq_save() / local_irq_restore() 404 local_irq_save() / local_irq_restore()
413 hardirq enter / hardirq exit 405 hardirq enter / hardirq exit
414 NMI enter / NMI exit 406 NMI enter / NMI exit
415 rcu_dereference_sched() 407 rcu_dereference_sched()
416 408
417 These three flavors are used as follows: 409 These three flavors are used as follows:
418 410
419 a. RCU applied to normal data structures. 411 a. RCU applied to normal data structures.
420 412
421 b. RCU applied to networking data structu 413 b. RCU applied to networking data structures that may be subjected
422 to remote denial-of-service attacks. 414 to remote denial-of-service attacks.
423 415
424 c. RCU applied to scheduler and interrupt 416 c. RCU applied to scheduler and interrupt/NMI-handler tasks.
425 417
426 Again, most uses will be of (a). The (b) and 418 Again, most uses will be of (a). The (b) and (c) cases are important
427 for specialized uses, but are relatively uncom 419 for specialized uses, but are relatively uncommon. The SRCU, RCU-Tasks,
428 RCU-Tasks-Rude, and RCU-Tasks-Trace have simil 420 RCU-Tasks-Rude, and RCU-Tasks-Trace have similar relationships among
429 their assorted primitives. 421 their assorted primitives.
430 422
431 .. _3_whatisRCU: 423 .. _3_whatisRCU:
432 424
433 3. WHAT ARE SOME EXAMPLE USES OF CORE RCU API 425 3. WHAT ARE SOME EXAMPLE USES OF CORE RCU API?
434 ---------------------------------------------- 426 -----------------------------------------------
435 427
436 This section shows a simple use of the core RC 428 This section shows a simple use of the core RCU API to protect a
437 global pointer to a dynamically allocated stru 429 global pointer to a dynamically allocated structure. More-typical
438 uses of RCU may be found in listRCU.rst and NM 430 uses of RCU may be found in listRCU.rst and NMI-RCU.rst.
439 :: 431 ::
440 432
441 struct foo { 433 struct foo {
442 int a; 434 int a;
443 char b; 435 char b;
444 long c; 436 long c;
445 }; 437 };
446 DEFINE_SPINLOCK(foo_mutex); 438 DEFINE_SPINLOCK(foo_mutex);
447 439
448 struct foo __rcu *gbl_foo; 440 struct foo __rcu *gbl_foo;
449 441
450 /* 442 /*
451 * Create a new struct foo that is the 443 * Create a new struct foo that is the same as the one currently
452 * pointed to by gbl_foo, except that 444 * pointed to by gbl_foo, except that field "a" is replaced
453 * with "new_a". Points gbl_foo to th 445 * with "new_a". Points gbl_foo to the new structure, and
454 * frees up the old structure after a 446 * frees up the old structure after a grace period.
455 * 447 *
456 * Uses rcu_assign_pointer() to ensure 448 * Uses rcu_assign_pointer() to ensure that concurrent readers
457 * see the initialized version of the 449 * see the initialized version of the new structure.
458 * 450 *
459 * Uses synchronize_rcu() to ensure th 451 * Uses synchronize_rcu() to ensure that any readers that might
460 * have references to the old structur 452 * have references to the old structure complete before freeing
461 * the old structure. 453 * the old structure.
462 */ 454 */
463 void foo_update_a(int new_a) 455 void foo_update_a(int new_a)
464 { 456 {
465 struct foo *new_fp; 457 struct foo *new_fp;
466 struct foo *old_fp; 458 struct foo *old_fp;
467 459
468 new_fp = kmalloc(sizeof(*new_f 460 new_fp = kmalloc(sizeof(*new_fp), GFP_KERNEL);
469 spin_lock(&foo_mutex); 461 spin_lock(&foo_mutex);
470 old_fp = rcu_dereference_prote 462 old_fp = rcu_dereference_protected(gbl_foo, lockdep_is_held(&foo_mutex));
471 *new_fp = *old_fp; 463 *new_fp = *old_fp;
472 new_fp->a = new_a; 464 new_fp->a = new_a;
473 rcu_assign_pointer(gbl_foo, ne 465 rcu_assign_pointer(gbl_foo, new_fp);
474 spin_unlock(&foo_mutex); 466 spin_unlock(&foo_mutex);
475 synchronize_rcu(); 467 synchronize_rcu();
476 kfree(old_fp); 468 kfree(old_fp);
477 } 469 }
478 470
479 /* 471 /*
480 * Return the value of field "a" of th 472 * Return the value of field "a" of the current gbl_foo
481 * structure. Use rcu_read_lock() and 473 * structure. Use rcu_read_lock() and rcu_read_unlock()
482 * to ensure that the structure does n 474 * to ensure that the structure does not get deleted out
483 * from under us, and use rcu_derefere 475 * from under us, and use rcu_dereference() to ensure that
484 * we see the initialized version of t 476 * we see the initialized version of the structure (important
485 * for DEC Alpha and for people readin 477 * for DEC Alpha and for people reading the code).
486 */ 478 */
487 int foo_get_a(void) 479 int foo_get_a(void)
488 { 480 {
489 int retval; 481 int retval;
490 482
491 rcu_read_lock(); 483 rcu_read_lock();
492 retval = rcu_dereference(gbl_f 484 retval = rcu_dereference(gbl_foo)->a;
493 rcu_read_unlock(); 485 rcu_read_unlock();
494 return retval; 486 return retval;
495 } 487 }
496 488
497 So, to sum up: 489 So, to sum up:
498 490
499 - Use rcu_read_lock() and rcu_read_unloc 491 - Use rcu_read_lock() and rcu_read_unlock() to guard RCU
500 read-side critical sections. 492 read-side critical sections.
501 493
502 - Within an RCU read-side critical secti 494 - Within an RCU read-side critical section, use rcu_dereference()
503 to dereference RCU-protected pointers. 495 to dereference RCU-protected pointers.
504 496
505 - Use some solid design (such as locks o 497 - Use some solid design (such as locks or semaphores) to
506 keep concurrent updates from interferi 498 keep concurrent updates from interfering with each other.
507 499
508 - Use rcu_assign_pointer() to update an 500 - Use rcu_assign_pointer() to update an RCU-protected pointer.
509 This primitive protects concurrent rea 501 This primitive protects concurrent readers from the updater,
510 **not** concurrent updates from each o 502 **not** concurrent updates from each other! You therefore still
511 need to use locking (or something simi 503 need to use locking (or something similar) to keep concurrent
512 rcu_assign_pointer() primitives from i 504 rcu_assign_pointer() primitives from interfering with each other.
513 505
514 - Use synchronize_rcu() **after** removi 506 - Use synchronize_rcu() **after** removing a data element from an
515 RCU-protected data structure, but **be 507 RCU-protected data structure, but **before** reclaiming/freeing
516 the data element, in order to wait for 508 the data element, in order to wait for the completion of all
517 RCU read-side critical sections that m 509 RCU read-side critical sections that might be referencing that
518 data item. 510 data item.
519 511
520 See checklist.rst for additional rules to foll 512 See checklist.rst for additional rules to follow when using RCU.
521 And again, more-typical uses of RCU may be fou 513 And again, more-typical uses of RCU may be found in listRCU.rst
522 and NMI-RCU.rst. 514 and NMI-RCU.rst.
523 515
524 .. _4_whatisRCU: 516 .. _4_whatisRCU:
525 517
526 4. WHAT IF MY UPDATING THREAD CANNOT BLOCK? 518 4. WHAT IF MY UPDATING THREAD CANNOT BLOCK?
527 -------------------------------------------- 519 --------------------------------------------
528 520
529 In the example above, foo_update_a() blocks un 521 In the example above, foo_update_a() blocks until a grace period elapses.
530 This is quite simple, but in some cases one ca 522 This is quite simple, but in some cases one cannot afford to wait so
531 long -- there might be other high-priority wor 523 long -- there might be other high-priority work to be done.
532 524
533 In such cases, one uses call_rcu() rather than 525 In such cases, one uses call_rcu() rather than synchronize_rcu().
534 The call_rcu() API is as follows:: 526 The call_rcu() API is as follows::
535 527
536 void call_rcu(struct rcu_head *head, r 528 void call_rcu(struct rcu_head *head, rcu_callback_t func);
537 529
538 This function invokes func(head) after a grace 530 This function invokes func(head) after a grace period has elapsed.
539 This invocation might happen from either softi 531 This invocation might happen from either softirq or process context,
540 so the function is not permitted to block. Th 532 so the function is not permitted to block. The foo struct needs to
541 have an rcu_head structure added, perhaps as f 533 have an rcu_head structure added, perhaps as follows::
542 534
543 struct foo { 535 struct foo {
544 int a; 536 int a;
545 char b; 537 char b;
546 long c; 538 long c;
547 struct rcu_head rcu; 539 struct rcu_head rcu;
548 }; 540 };
549 541
550 The foo_update_a() function might then be writ 542 The foo_update_a() function might then be written as follows::
551 543
552 /* 544 /*
553 * Create a new struct foo that is the 545 * Create a new struct foo that is the same as the one currently
554 * pointed to by gbl_foo, except that 546 * pointed to by gbl_foo, except that field "a" is replaced
555 * with "new_a". Points gbl_foo to th 547 * with "new_a". Points gbl_foo to the new structure, and
556 * frees up the old structure after a 548 * frees up the old structure after a grace period.
557 * 549 *
558 * Uses rcu_assign_pointer() to ensure 550 * Uses rcu_assign_pointer() to ensure that concurrent readers
559 * see the initialized version of the 551 * see the initialized version of the new structure.
560 * 552 *
561 * Uses call_rcu() to ensure that any 553 * Uses call_rcu() to ensure that any readers that might have
562 * references to the old structure com 554 * references to the old structure complete before freeing the
563 * old structure. 555 * old structure.
564 */ 556 */
565 void foo_update_a(int new_a) 557 void foo_update_a(int new_a)
566 { 558 {
567 struct foo *new_fp; 559 struct foo *new_fp;
568 struct foo *old_fp; 560 struct foo *old_fp;
569 561
570 new_fp = kmalloc(sizeof(*new_f 562 new_fp = kmalloc(sizeof(*new_fp), GFP_KERNEL);
571 spin_lock(&foo_mutex); 563 spin_lock(&foo_mutex);
572 old_fp = rcu_dereference_prote 564 old_fp = rcu_dereference_protected(gbl_foo, lockdep_is_held(&foo_mutex));
573 *new_fp = *old_fp; 565 *new_fp = *old_fp;
574 new_fp->a = new_a; 566 new_fp->a = new_a;
575 rcu_assign_pointer(gbl_foo, ne 567 rcu_assign_pointer(gbl_foo, new_fp);
576 spin_unlock(&foo_mutex); 568 spin_unlock(&foo_mutex);
577 call_rcu(&old_fp->rcu, foo_rec 569 call_rcu(&old_fp->rcu, foo_reclaim);
578 } 570 }
579 571
580 The foo_reclaim() function might appear as fol 572 The foo_reclaim() function might appear as follows::
581 573
582 void foo_reclaim(struct rcu_head *rp) 574 void foo_reclaim(struct rcu_head *rp)
583 { 575 {
584 struct foo *fp = container_of( 576 struct foo *fp = container_of(rp, struct foo, rcu);
585 577
586 foo_cleanup(fp->a); 578 foo_cleanup(fp->a);
587 579
588 kfree(fp); 580 kfree(fp);
589 } 581 }
590 582
591 The container_of() primitive is a macro that, 583 The container_of() primitive is a macro that, given a pointer into a
592 struct, the type of the struct, and the pointe 584 struct, the type of the struct, and the pointed-to field within the
593 struct, returns a pointer to the beginning of 585 struct, returns a pointer to the beginning of the struct.
594 586
595 The use of call_rcu() permits the caller of fo 587 The use of call_rcu() permits the caller of foo_update_a() to
596 immediately regain control, without needing to 588 immediately regain control, without needing to worry further about the
597 old version of the newly updated element. It 589 old version of the newly updated element. It also clearly shows the
598 RCU distinction between updater, namely foo_up 590 RCU distinction between updater, namely foo_update_a(), and reclaimer,
599 namely foo_reclaim(). 591 namely foo_reclaim().
600 592
601 The summary of advice is the same as for the p 593 The summary of advice is the same as for the previous section, except
602 that we are now using call_rcu() rather than s 594 that we are now using call_rcu() rather than synchronize_rcu():
603 595
604 - Use call_rcu() **after** removing a da 596 - Use call_rcu() **after** removing a data element from an
605 RCU-protected data structure in order 597 RCU-protected data structure in order to register a callback
606 function that will be invoked after th 598 function that will be invoked after the completion of all RCU
607 read-side critical sections that might 599 read-side critical sections that might be referencing that
608 data item. 600 data item.
609 601
610 If the callback for call_rcu() is not doing an 602 If the callback for call_rcu() is not doing anything more than calling
611 kfree() on the structure, you can use kfree_rc 603 kfree() on the structure, you can use kfree_rcu() instead of call_rcu()
612 to avoid having to write your own callback:: 604 to avoid having to write your own callback::
613 605
614 kfree_rcu(old_fp, rcu); 606 kfree_rcu(old_fp, rcu);
615 607
616 If the occasional sleep is permitted, the sing 608 If the occasional sleep is permitted, the single-argument form may
617 be used, omitting the rcu_head structure from 609 be used, omitting the rcu_head structure from struct foo.
618 610
619 kfree_rcu_mightsleep(old_fp); 611 kfree_rcu_mightsleep(old_fp);
620 612
621 This variant almost never blocks, but might do 613 This variant almost never blocks, but might do so by invoking
622 synchronize_rcu() in response to memory-alloca 614 synchronize_rcu() in response to memory-allocation failure.
623 615
624 Again, see checklist.rst for additional rules 616 Again, see checklist.rst for additional rules governing the use of RCU.
625 617
626 .. _5_whatisRCU: 618 .. _5_whatisRCU:
627 619
628 5. WHAT ARE SOME SIMPLE IMPLEMENTATIONS OF RC 620 5. WHAT ARE SOME SIMPLE IMPLEMENTATIONS OF RCU?
629 ---------------------------------------------- 621 ------------------------------------------------
630 622
631 One of the nice things about RCU is that it ha 623 One of the nice things about RCU is that it has extremely simple "toy"
632 implementations that are a good first step tow 624 implementations that are a good first step towards understanding the
633 production-quality implementations in the Linu 625 production-quality implementations in the Linux kernel. This section
634 presents two such "toy" implementations of RCU 626 presents two such "toy" implementations of RCU, one that is implemented
635 in terms of familiar locking primitives, and a 627 in terms of familiar locking primitives, and another that more closely
636 resembles "classic" RCU. Both are way too sim 628 resembles "classic" RCU. Both are way too simple for real-world use,
637 lacking both functionality and performance. H 629 lacking both functionality and performance. However, they are useful
638 in getting a feel for how RCU works. See kern 630 in getting a feel for how RCU works. See kernel/rcu/update.c for a
639 production-quality implementation, and see: 631 production-quality implementation, and see:
640 632
641 https://docs.google.com/document/d/1X0 633 https://docs.google.com/document/d/1X0lThx8OK0ZgLMqVoXiR4ZrGURHrXK6NyLRbeXe3Xac/edit
642 634
643 for papers describing the Linux kernel RCU imp 635 for papers describing the Linux kernel RCU implementation. The OLS'01
644 and OLS'02 papers are a good introduction, and 636 and OLS'02 papers are a good introduction, and the dissertation provides
645 more details on the current implementation as 637 more details on the current implementation as of early 2004.
646 638
647 639
648 5A. "TOY" IMPLEMENTATION #1: LOCKING 640 5A. "TOY" IMPLEMENTATION #1: LOCKING
649 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 641 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
650 This section presents a "toy" RCU implementati 642 This section presents a "toy" RCU implementation that is based on
651 familiar locking primitives. Its overhead mak 643 familiar locking primitives. Its overhead makes it a non-starter for
652 real-life use, as does its lack of scalability 644 real-life use, as does its lack of scalability. It is also unsuitable
653 for realtime use, since it allows scheduling l 645 for realtime use, since it allows scheduling latency to "bleed" from
654 one read-side critical section to another. It 646 one read-side critical section to another. It also assumes recursive
655 reader-writer locks: If you try this with non 647 reader-writer locks: If you try this with non-recursive locks, and
656 you allow nested rcu_read_lock() calls, you ca 648 you allow nested rcu_read_lock() calls, you can deadlock.
657 649
658 However, it is probably the easiest implementa 650 However, it is probably the easiest implementation to relate to, so is
659 a good starting point. 651 a good starting point.
660 652
661 It is extremely simple:: 653 It is extremely simple::
662 654
663 static DEFINE_RWLOCK(rcu_gp_mutex); 655 static DEFINE_RWLOCK(rcu_gp_mutex);
664 656
665 void rcu_read_lock(void) 657 void rcu_read_lock(void)
666 { 658 {
667 read_lock(&rcu_gp_mutex); 659 read_lock(&rcu_gp_mutex);
668 } 660 }
669 661
670 void rcu_read_unlock(void) 662 void rcu_read_unlock(void)
671 { 663 {
672 read_unlock(&rcu_gp_mutex); 664 read_unlock(&rcu_gp_mutex);
673 } 665 }
674 666
675 void synchronize_rcu(void) 667 void synchronize_rcu(void)
676 { 668 {
677 write_lock(&rcu_gp_mutex); 669 write_lock(&rcu_gp_mutex);
678 smp_mb__after_spinlock(); 670 smp_mb__after_spinlock();
679 write_unlock(&rcu_gp_mutex); 671 write_unlock(&rcu_gp_mutex);
680 } 672 }
681 673
682 [You can ignore rcu_assign_pointer() and rcu_d 674 [You can ignore rcu_assign_pointer() and rcu_dereference() without missing
683 much. But here are simplified versions anyway 675 much. But here are simplified versions anyway. And whatever you do,
684 don't forget about them when submitting patche 676 don't forget about them when submitting patches making use of RCU!]::
685 677
686 #define rcu_assign_pointer(p, v) \ 678 #define rcu_assign_pointer(p, v) \
687 ({ \ 679 ({ \
688 smp_store_release(&(p), (v)); 680 smp_store_release(&(p), (v)); \
689 }) 681 })
690 682
691 #define rcu_dereference(p) \ 683 #define rcu_dereference(p) \
692 ({ \ 684 ({ \
693 typeof(p) _________p1 = READ_O 685 typeof(p) _________p1 = READ_ONCE(p); \
694 (_________p1); \ 686 (_________p1); \
695 }) 687 })
696 688
697 689
698 The rcu_read_lock() and rcu_read_unlock() prim 690 The rcu_read_lock() and rcu_read_unlock() primitive read-acquire
699 and release a global reader-writer lock. The 691 and release a global reader-writer lock. The synchronize_rcu()
700 primitive write-acquires this same lock, then 692 primitive write-acquires this same lock, then releases it. This means
701 that once synchronize_rcu() exits, all RCU rea 693 that once synchronize_rcu() exits, all RCU read-side critical sections
702 that were in progress before synchronize_rcu() 694 that were in progress before synchronize_rcu() was called are guaranteed
703 to have completed -- there is no way that sync 695 to have completed -- there is no way that synchronize_rcu() would have
704 been able to write-acquire the lock otherwise. 696 been able to write-acquire the lock otherwise. The smp_mb__after_spinlock()
705 promotes synchronize_rcu() to a full memory ba 697 promotes synchronize_rcu() to a full memory barrier in compliance with
706 the "Memory-Barrier Guarantees" listed in: 698 the "Memory-Barrier Guarantees" listed in:
707 699
708 Design/Requirements/Requirements.rst 700 Design/Requirements/Requirements.rst
709 701
710 It is possible to nest rcu_read_lock(), since 702 It is possible to nest rcu_read_lock(), since reader-writer locks may
711 be recursively acquired. Note also that rcu_r 703 be recursively acquired. Note also that rcu_read_lock() is immune
712 from deadlock (an important property of RCU). 704 from deadlock (an important property of RCU). The reason for this is
713 that the only thing that can block rcu_read_lo 705 that the only thing that can block rcu_read_lock() is a synchronize_rcu().
714 But synchronize_rcu() does not acquire any loc 706 But synchronize_rcu() does not acquire any locks while holding rcu_gp_mutex,
715 so there can be no deadlock cycle. 707 so there can be no deadlock cycle.
716 708
717 .. _quiz_1: 709 .. _quiz_1:
718 710
719 Quick Quiz #1: 711 Quick Quiz #1:
720 Why is this argument naive? H 712 Why is this argument naive? How could a deadlock
721 occur when using this algorith 713 occur when using this algorithm in a real-world Linux
722 kernel? How could this deadlo 714 kernel? How could this deadlock be avoided?
723 715
724 :ref:`Answers to Quick Quiz <9_whatisRCU>` 716 :ref:`Answers to Quick Quiz <9_whatisRCU>`
725 717
726 5B. "TOY" EXAMPLE #2: CLASSIC RCU 718 5B. "TOY" EXAMPLE #2: CLASSIC RCU
727 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 719 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
728 This section presents a "toy" RCU implementati 720 This section presents a "toy" RCU implementation that is based on
729 "classic RCU". It is also short on performanc 721 "classic RCU". It is also short on performance (but only for updates) and
730 on features such as hotplug CPU and the abilit 722 on features such as hotplug CPU and the ability to run in CONFIG_PREEMPTION
731 kernels. The definitions of rcu_dereference() 723 kernels. The definitions of rcu_dereference() and rcu_assign_pointer()
732 are the same as those shown in the preceding s 724 are the same as those shown in the preceding section, so they are omitted.
733 :: 725 ::
734 726
735 void rcu_read_lock(void) { } 727 void rcu_read_lock(void) { }
736 728
737 void rcu_read_unlock(void) { } 729 void rcu_read_unlock(void) { }
738 730
739 void synchronize_rcu(void) 731 void synchronize_rcu(void)
740 { 732 {
741 int cpu; 733 int cpu;
742 734
743 for_each_possible_cpu(cpu) 735 for_each_possible_cpu(cpu)
744 run_on(cpu); 736 run_on(cpu);
745 } 737 }
746 738
747 Note that rcu_read_lock() and rcu_read_unlock( 739 Note that rcu_read_lock() and rcu_read_unlock() do absolutely nothing.
748 This is the great strength of classic RCU in a 740 This is the great strength of classic RCU in a non-preemptive kernel:
749 read-side overhead is precisely zero, at least 741 read-side overhead is precisely zero, at least on non-Alpha CPUs.
750 And there is absolutely no way that rcu_read_l 742 And there is absolutely no way that rcu_read_lock() can possibly
751 participate in a deadlock cycle! 743 participate in a deadlock cycle!
752 744
753 The implementation of synchronize_rcu() simply 745 The implementation of synchronize_rcu() simply schedules itself on each
754 CPU in turn. The run_on() primitive can be im 746 CPU in turn. The run_on() primitive can be implemented straightforwardly
755 in terms of the sched_setaffinity() primitive. 747 in terms of the sched_setaffinity() primitive. Of course, a somewhat less
756 "toy" implementation would restore the affinit 748 "toy" implementation would restore the affinity upon completion rather
757 than just leaving all tasks running on the las 749 than just leaving all tasks running on the last CPU, but when I said
758 "toy", I meant **toy**! 750 "toy", I meant **toy**!
759 751
760 So how the heck is this supposed to work??? 752 So how the heck is this supposed to work???
761 753
762 Remember that it is illegal to block while in 754 Remember that it is illegal to block while in an RCU read-side critical
763 section. Therefore, if a given CPU executes a 755 section. Therefore, if a given CPU executes a context switch, we know
764 that it must have completed all preceding RCU 756 that it must have completed all preceding RCU read-side critical sections.
765 Once **all** CPUs have executed a context swit 757 Once **all** CPUs have executed a context switch, then **all** preceding
766 RCU read-side critical sections will have comp 758 RCU read-side critical sections will have completed.
767 759
768 So, suppose that we remove a data item from it 760 So, suppose that we remove a data item from its structure and then invoke
769 synchronize_rcu(). Once synchronize_rcu() ret 761 synchronize_rcu(). Once synchronize_rcu() returns, we are guaranteed
770 that there are no RCU read-side critical secti 762 that there are no RCU read-side critical sections holding a reference
771 to that data item, so we can safely reclaim it 763 to that data item, so we can safely reclaim it.
772 764
773 .. _quiz_2: 765 .. _quiz_2:
774 766
775 Quick Quiz #2: 767 Quick Quiz #2:
776 Give an example where Classic 768 Give an example where Classic RCU's read-side
777 overhead is **negative**. 769 overhead is **negative**.
778 770
779 :ref:`Answers to Quick Quiz <9_whatisRCU>` 771 :ref:`Answers to Quick Quiz <9_whatisRCU>`
780 772
781 .. _quiz_3: 773 .. _quiz_3:
782 774
783 Quick Quiz #3: 775 Quick Quiz #3:
784 If it is illegal to block in a 776 If it is illegal to block in an RCU read-side
785 critical section, what the hec 777 critical section, what the heck do you do in
786 CONFIG_PREEMPT_RT, where norma 778 CONFIG_PREEMPT_RT, where normal spinlocks can block???
787 779
788 :ref:`Answers to Quick Quiz <9_whatisRCU>` 780 :ref:`Answers to Quick Quiz <9_whatisRCU>`
789 781
790 .. _6_whatisRCU: 782 .. _6_whatisRCU:
791 783
792 6. ANALOGY WITH READER-WRITER LOCKING 784 6. ANALOGY WITH READER-WRITER LOCKING
793 -------------------------------------- 785 --------------------------------------
794 786
795 Although RCU can be used in many different way 787 Although RCU can be used in many different ways, a very common use of
796 RCU is analogous to reader-writer locking. Th 788 RCU is analogous to reader-writer locking. The following unified
797 diff shows how closely related RCU and reader- 789 diff shows how closely related RCU and reader-writer locking can be.
798 :: 790 ::
799 791
800 @@ -5,5 +5,5 @@ struct el { 792 @@ -5,5 +5,5 @@ struct el {
801 int data; 793 int data;
802 /* Other data fields */ 794 /* Other data fields */
803 }; 795 };
804 -rwlock_t listmutex; 796 -rwlock_t listmutex;
805 +spinlock_t listmutex; 797 +spinlock_t listmutex;
806 struct el head; 798 struct el head;
807 799
808 @@ -13,15 +14,15 @@ 800 @@ -13,15 +14,15 @@
809 struct list_head *lp; 801 struct list_head *lp;
810 struct el *p; 802 struct el *p;
811 803
812 - read_lock(&listmutex); 804 - read_lock(&listmutex);
813 - list_for_each_entry(p, head, l 805 - list_for_each_entry(p, head, lp) {
814 + rcu_read_lock(); 806 + rcu_read_lock();
815 + list_for_each_entry_rcu(p, hea 807 + list_for_each_entry_rcu(p, head, lp) {
816 if (p->key == key) { 808 if (p->key == key) {
817 *result = p->d 809 *result = p->data;
818 - read_unlock(&l 810 - read_unlock(&listmutex);
819 + rcu_read_unloc 811 + rcu_read_unlock();
820 return 1; 812 return 1;
821 } 813 }
822 } 814 }
823 - read_unlock(&listmutex); 815 - read_unlock(&listmutex);
824 + rcu_read_unlock(); 816 + rcu_read_unlock();
825 return 0; 817 return 0;
826 } 818 }
827 819
828 @@ -29,15 +30,16 @@ 820 @@ -29,15 +30,16 @@
829 { 821 {
830 struct el *p; 822 struct el *p;
831 823
832 - write_lock(&listmutex); 824 - write_lock(&listmutex);
833 + spin_lock(&listmutex); 825 + spin_lock(&listmutex);
834 list_for_each_entry(p, head, l 826 list_for_each_entry(p, head, lp) {
835 if (p->key == key) { 827 if (p->key == key) {
836 - list_del(&p->l 828 - list_del(&p->list);
837 - write_unlock(& 829 - write_unlock(&listmutex);
838 + list_del_rcu(& 830 + list_del_rcu(&p->list);
839 + spin_unlock(&l 831 + spin_unlock(&listmutex);
840 + synchronize_rc 832 + synchronize_rcu();
841 kfree(p); 833 kfree(p);
842 return 1; 834 return 1;
843 } 835 }
844 } 836 }
845 - write_unlock(&listmutex); 837 - write_unlock(&listmutex);
846 + spin_unlock(&listmutex); 838 + spin_unlock(&listmutex);
847 return 0; 839 return 0;
848 } 840 }
849 841
850 Or, for those who prefer a side-by-side listin 842 Or, for those who prefer a side-by-side listing::
851 843
852 1 struct el { 1 stru 844 1 struct el { 1 struct el {
853 2 struct list_head list; 2 st 845 2 struct list_head list; 2 struct list_head list;
854 3 long key; 3 lo 846 3 long key; 3 long key;
855 4 spinlock_t mutex; 4 sp 847 4 spinlock_t mutex; 4 spinlock_t mutex;
856 5 int data; 5 in 848 5 int data; 5 int data;
857 6 /* Other data fields */ 6 /* 849 6 /* Other data fields */ 6 /* Other data fields */
858 7 }; 7 }; 850 7 }; 7 };
859 8 rwlock_t listmutex; 8 spin 851 8 rwlock_t listmutex; 8 spinlock_t listmutex;
860 9 struct el head; 9 stru 852 9 struct el head; 9 struct el head;
861 853
862 :: 854 ::
863 855
864 1 int search(long key, int *result) 1 int 856 1 int search(long key, int *result) 1 int search(long key, int *result)
865 2 { 2 { 857 2 { 2 {
866 3 struct list_head *lp; 3 s 858 3 struct list_head *lp; 3 struct list_head *lp;
867 4 struct el *p; 4 s 859 4 struct el *p; 4 struct el *p;
868 5 5 860 5 5
869 6 read_lock(&listmutex); 6 r 861 6 read_lock(&listmutex); 6 rcu_read_lock();
870 7 list_for_each_entry(p, head, lp) { 7 l 862 7 list_for_each_entry(p, head, lp) { 7 list_for_each_entry_rcu(p, head, lp) {
871 8 if (p->key == key) { 8 863 8 if (p->key == key) { 8 if (p->key == key) {
872 9 *result = p->data; 9 864 9 *result = p->data; 9 *result = p->data;
873 10 read_unlock(&listmutex); 10 865 10 read_unlock(&listmutex); 10 rcu_read_unlock();
874 11 return 1; 11 866 11 return 1; 11 return 1;
875 12 } 12 867 12 } 12 }
876 13 } 13 } 868 13 } 13 }
877 14 read_unlock(&listmutex); 14 r 869 14 read_unlock(&listmutex); 14 rcu_read_unlock();
878 15 return 0; 15 r 870 15 return 0; 15 return 0;
879 16 } 16 } 871 16 } 16 }
880 872
881 :: 873 ::
882 874
883 1 int delete(long key) 1 int 875 1 int delete(long key) 1 int delete(long key)
884 2 { 2 { 876 2 { 2 {
885 3 struct el *p; 3 s 877 3 struct el *p; 3 struct el *p;
886 4 4 878 4 4
887 5 write_lock(&listmutex); 5 s 879 5 write_lock(&listmutex); 5 spin_lock(&listmutex);
888 6 list_for_each_entry(p, head, lp) { 6 l 880 6 list_for_each_entry(p, head, lp) { 6 list_for_each_entry(p, head, lp) {
889 7 if (p->key == key) { 7 881 7 if (p->key == key) { 7 if (p->key == key) {
890 8 list_del(&p->list); 8 882 8 list_del(&p->list); 8 list_del_rcu(&p->list);
891 9 write_unlock(&listmutex); 9 883 9 write_unlock(&listmutex); 9 spin_unlock(&listmutex);
892 10 884 10 synchronize_rcu();
893 10 kfree(p); 11 885 10 kfree(p); 11 kfree(p);
894 11 return 1; 12 886 11 return 1; 12 return 1;
895 12 } 13 887 12 } 13 }
896 13 } 14 } 888 13 } 14 }
897 14 write_unlock(&listmutex); 15 s 889 14 write_unlock(&listmutex); 15 spin_unlock(&listmutex);
898 15 return 0; 16 r 890 15 return 0; 16 return 0;
899 16 } 17 } 891 16 } 17 }
900 892
901 Either way, the differences are quite small. 893 Either way, the differences are quite small. Read-side locking moves
902 to rcu_read_lock() and rcu_read_unlock, update 894 to rcu_read_lock() and rcu_read_unlock, update-side locking moves from
903 a reader-writer lock to a simple spinlock, and 895 a reader-writer lock to a simple spinlock, and a synchronize_rcu()
904 precedes the kfree(). 896 precedes the kfree().
905 897
906 However, there is one potential catch: the rea 898 However, there is one potential catch: the read-side and update-side
907 critical sections can now run concurrently. I 899 critical sections can now run concurrently. In many cases, this will
908 not be a problem, but it is necessary to check 900 not be a problem, but it is necessary to check carefully regardless.
909 For example, if multiple independent list upda 901 For example, if multiple independent list updates must be seen as
910 a single atomic update, converting to RCU will 902 a single atomic update, converting to RCU will require special care.
911 903
912 Also, the presence of synchronize_rcu() means 904 Also, the presence of synchronize_rcu() means that the RCU version of
913 delete() can now block. If this is a problem, 905 delete() can now block. If this is a problem, there is a callback-based
914 mechanism that never blocks, namely call_rcu() 906 mechanism that never blocks, namely call_rcu() or kfree_rcu(), that can
915 be used in place of synchronize_rcu(). 907 be used in place of synchronize_rcu().
916 908
917 .. _7_whatisRCU: 909 .. _7_whatisRCU:
918 910
919 7. ANALOGY WITH REFERENCE COUNTING 911 7. ANALOGY WITH REFERENCE COUNTING
920 ----------------------------------- 912 -----------------------------------
921 913
922 The reader-writer analogy (illustrated by the 914 The reader-writer analogy (illustrated by the previous section) is not
923 always the best way to think about using RCU. 915 always the best way to think about using RCU. Another helpful analogy
924 considers RCU an effective reference count on 916 considers RCU an effective reference count on everything which is
925 protected by RCU. 917 protected by RCU.
926 918
927 A reference count typically does not prevent t 919 A reference count typically does not prevent the referenced object's
928 values from changing, but does prevent changes 920 values from changing, but does prevent changes to type -- particularly the
929 gross change of type that happens when that ob 921 gross change of type that happens when that object's memory is freed and
930 re-allocated for some other purpose. Once a t 922 re-allocated for some other purpose. Once a type-safe reference to the
931 object is obtained, some other mechanism is ne 923 object is obtained, some other mechanism is needed to ensure consistent
932 access to the data in the object. This could 924 access to the data in the object. This could involve taking a spinlock,
933 but with RCU the typical approach is to perfor 925 but with RCU the typical approach is to perform reads with SMP-aware
934 operations such as smp_load_acquire(), to perf 926 operations such as smp_load_acquire(), to perform updates with atomic
935 read-modify-write operations, and to provide t 927 read-modify-write operations, and to provide the necessary ordering.
936 RCU provides a number of support functions tha 928 RCU provides a number of support functions that embed the required
937 operations and ordering, such as the list_for_ 929 operations and ordering, such as the list_for_each_entry_rcu() macro
938 used in the previous section. 930 used in the previous section.
939 931
940 A more focused view of the reference counting 932 A more focused view of the reference counting behavior is that,
941 between rcu_read_lock() and rcu_read_unlock(), 933 between rcu_read_lock() and rcu_read_unlock(), any reference taken with
942 rcu_dereference() on a pointer marked as ``__r 934 rcu_dereference() on a pointer marked as ``__rcu`` can be treated as
943 though a reference-count on that object has be 935 though a reference-count on that object has been temporarily increased.
944 This prevents the object from changing type. 936 This prevents the object from changing type. Exactly what this means
945 will depend on normal expectations of objects 937 will depend on normal expectations of objects of that type, but it
946 typically includes that spinlocks can still be 938 typically includes that spinlocks can still be safely locked, normal
947 reference counters can be safely manipulated, 939 reference counters can be safely manipulated, and ``__rcu`` pointers
948 can be safely dereferenced. 940 can be safely dereferenced.
949 941
950 Some operations that one might expect to see o 942 Some operations that one might expect to see on an object for
951 which an RCU reference is held include: 943 which an RCU reference is held include:
952 944
953 - Copying out data that is guaranteed to be s 945 - Copying out data that is guaranteed to be stable by the object's type.
954 - Using kref_get_unless_zero() or similar to 946 - Using kref_get_unless_zero() or similar to get a longer-term
955 reference. This may fail of course. 947 reference. This may fail of course.
956 - Acquiring a spinlock in the object, and che 948 - Acquiring a spinlock in the object, and checking if the object still
957 is the expected object and if so, manipulat 949 is the expected object and if so, manipulating it freely.
958 950
959 The understanding that RCU provides a referenc 951 The understanding that RCU provides a reference that only prevents a
960 change of type is particularly visible with ob 952 change of type is particularly visible with objects allocated from a
961 slab cache marked ``SLAB_TYPESAFE_BY_RCU``. R 953 slab cache marked ``SLAB_TYPESAFE_BY_RCU``. RCU operations may yield a
962 reference to an object from such a cache that 954 reference to an object from such a cache that has been concurrently freed
963 and the memory reallocated to a completely dif 955 and the memory reallocated to a completely different object, though of
964 the same type. In this case RCU doesn't even 956 the same type. In this case RCU doesn't even protect the identity of the
965 object from changing, only its type. So the o 957 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 s 958 one expected, but it will be one where it is safe to take a reference
967 (and then potentially acquiring a spinlock), a 959 (and then potentially acquiring a spinlock), allowing subsequent code
968 to check whether the identity matches expectat 960 to check whether the identity matches expectations. It is tempting
969 to simply acquire the spinlock without first t 961 to simply acquire the spinlock without first taking the reference, but
970 unfortunately any spinlock in a ``SLAB_TYPESAF 962 unfortunately any spinlock in a ``SLAB_TYPESAFE_BY_RCU`` object must be
971 initialized after each and every call to kmem_ 963 initialized after each and every call to kmem_cache_alloc(), which renders
972 reference-free spinlock acquisition completely 964 reference-free spinlock acquisition completely unsafe. Therefore, when
973 using ``SLAB_TYPESAFE_BY_RCU``, make proper us 965 using ``SLAB_TYPESAFE_BY_RCU``, make proper use of a reference counter.
974 (Those willing to initialize their locks in a 966 (Those willing to initialize their locks in a kmem_cache constructor
975 may also use locking, including cache-friendly 967 may also use locking, including cache-friendly sequence locking.)
976 968
977 With traditional reference counting -- such as 969 With traditional reference counting -- such as that implemented by the
978 kref library in Linux -- there is typically co 970 kref library in Linux -- there is typically code that runs when the last
979 reference to an object is dropped. With kref, 971 reference to an object is dropped. With kref, this is the function
980 passed to kref_put(). When RCU is being used, 972 passed to kref_put(). When RCU is being used, such finalization code
981 must not be run until all ``__rcu`` pointers r 973 must not be run until all ``__rcu`` pointers referencing the object have
982 been updated, and then a grace period has pass 974 been updated, and then a grace period has passed. Every remaining
983 globally visible pointer to the object must be 975 globally visible pointer to the object must be considered to be a
984 potential counted reference, and the finalizat 976 potential counted reference, and the finalization code is typically run
985 using call_rcu() only after all those pointers 977 using call_rcu() only after all those pointers have been changed.
986 978
987 To see how to choose between these two analogi 979 To see how to choose between these two analogies -- of RCU as a
988 reader-writer lock and RCU as a reference coun 980 reader-writer lock and RCU as a reference counting system -- it is useful
989 to reflect on the scale of the thing being pro 981 to reflect on the scale of the thing being protected. The reader-writer
990 lock analogy looks at larger multi-part object 982 lock analogy looks at larger multi-part objects such as a linked list
991 and shows how RCU can facilitate concurrency w 983 and shows how RCU can facilitate concurrency while elements are added
992 to, and removed from, the list. The reference 984 to, and removed from, the list. The reference-count analogy looks at
993 the individual objects and looks at how they c 985 the individual objects and looks at how they can be accessed safely
994 within whatever whole they are a part of. 986 within whatever whole they are a part of.
995 987
996 .. _8_whatisRCU: 988 .. _8_whatisRCU:
997 989
998 8. FULL LIST OF RCU APIs 990 8. FULL LIST OF RCU APIs
999 ------------------------- 991 -------------------------
1000 992
1001 The RCU APIs are documented in docbook-format 993 The RCU APIs are documented in docbook-format header comments in the
1002 Linux-kernel source code, but it helps to hav 994 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 995 APIs, since there does not appear to be a way to categorize them
1004 in docbook. Here is the list, by category. 996 in docbook. Here is the list, by category.
1005 997
1006 RCU list traversal:: 998 RCU list traversal::
1007 999
1008 list_entry_rcu 1000 list_entry_rcu
1009 list_entry_lockless 1001 list_entry_lockless
1010 list_first_entry_rcu 1002 list_first_entry_rcu
1011 list_next_rcu 1003 list_next_rcu
1012 list_for_each_entry_rcu 1004 list_for_each_entry_rcu
1013 list_for_each_entry_continue_rcu 1005 list_for_each_entry_continue_rcu
1014 list_for_each_entry_from_rcu 1006 list_for_each_entry_from_rcu
1015 list_first_or_null_rcu 1007 list_first_or_null_rcu
1016 list_next_or_null_rcu 1008 list_next_or_null_rcu
1017 hlist_first_rcu 1009 hlist_first_rcu
1018 hlist_next_rcu 1010 hlist_next_rcu
1019 hlist_pprev_rcu 1011 hlist_pprev_rcu
1020 hlist_for_each_entry_rcu 1012 hlist_for_each_entry_rcu
1021 hlist_for_each_entry_rcu_bh 1013 hlist_for_each_entry_rcu_bh
1022 hlist_for_each_entry_from_rcu 1014 hlist_for_each_entry_from_rcu
1023 hlist_for_each_entry_continue_rcu 1015 hlist_for_each_entry_continue_rcu
1024 hlist_for_each_entry_continue_rcu_bh 1016 hlist_for_each_entry_continue_rcu_bh
1025 hlist_nulls_first_rcu 1017 hlist_nulls_first_rcu
1026 hlist_nulls_for_each_entry_rcu 1018 hlist_nulls_for_each_entry_rcu
1027 hlist_bl_first_rcu 1019 hlist_bl_first_rcu
1028 hlist_bl_for_each_entry_rcu 1020 hlist_bl_for_each_entry_rcu
1029 1021
1030 RCU pointer/list update:: 1022 RCU pointer/list update::
1031 1023
1032 rcu_assign_pointer 1024 rcu_assign_pointer
1033 list_add_rcu 1025 list_add_rcu
1034 list_add_tail_rcu 1026 list_add_tail_rcu
1035 list_del_rcu 1027 list_del_rcu
1036 list_replace_rcu 1028 list_replace_rcu
1037 hlist_add_behind_rcu 1029 hlist_add_behind_rcu
1038 hlist_add_before_rcu 1030 hlist_add_before_rcu
1039 hlist_add_head_rcu 1031 hlist_add_head_rcu
1040 hlist_add_tail_rcu 1032 hlist_add_tail_rcu
1041 hlist_del_rcu 1033 hlist_del_rcu
1042 hlist_del_init_rcu 1034 hlist_del_init_rcu
1043 hlist_replace_rcu 1035 hlist_replace_rcu
1044 list_splice_init_rcu 1036 list_splice_init_rcu
1045 list_splice_tail_init_rcu 1037 list_splice_tail_init_rcu
1046 hlist_nulls_del_init_rcu 1038 hlist_nulls_del_init_rcu
1047 hlist_nulls_del_rcu 1039 hlist_nulls_del_rcu
1048 hlist_nulls_add_head_rcu 1040 hlist_nulls_add_head_rcu
1049 hlist_bl_add_head_rcu 1041 hlist_bl_add_head_rcu
1050 hlist_bl_del_init_rcu 1042 hlist_bl_del_init_rcu
1051 hlist_bl_del_rcu 1043 hlist_bl_del_rcu
1052 hlist_bl_set_first_rcu 1044 hlist_bl_set_first_rcu
1053 1045
1054 RCU:: 1046 RCU::
1055 1047
1056 Critical sections Grace period 1048 Critical sections Grace period Barrier
1057 1049
1058 rcu_read_lock synchronize_n 1050 rcu_read_lock synchronize_net rcu_barrier
1059 rcu_read_unlock synchronize_r 1051 rcu_read_unlock synchronize_rcu
1060 rcu_dereference synchronize_r 1052 rcu_dereference synchronize_rcu_expedited
1061 rcu_read_lock_held call_rcu 1053 rcu_read_lock_held call_rcu
1062 rcu_dereference_check kfree_rcu 1054 rcu_dereference_check kfree_rcu
1063 rcu_dereference_protected 1055 rcu_dereference_protected
1064 1056
1065 bh:: 1057 bh::
1066 1058
1067 Critical sections Grace period 1059 Critical sections Grace period Barrier
1068 1060
1069 rcu_read_lock_bh call_rcu 1061 rcu_read_lock_bh call_rcu rcu_barrier
1070 rcu_read_unlock_bh synchronize_r 1062 rcu_read_unlock_bh synchronize_rcu
1071 [local_bh_disable] synchronize_r 1063 [local_bh_disable] synchronize_rcu_expedited
1072 [and friends] 1064 [and friends]
1073 rcu_dereference_bh 1065 rcu_dereference_bh
1074 rcu_dereference_bh_check 1066 rcu_dereference_bh_check
1075 rcu_dereference_bh_protected 1067 rcu_dereference_bh_protected
1076 rcu_read_lock_bh_held 1068 rcu_read_lock_bh_held
1077 1069
1078 sched:: 1070 sched::
1079 1071
1080 Critical sections Grace period 1072 Critical sections Grace period Barrier
1081 1073
1082 rcu_read_lock_sched call_rcu 1074 rcu_read_lock_sched call_rcu rcu_barrier
1083 rcu_read_unlock_sched synchronize_r 1075 rcu_read_unlock_sched synchronize_rcu
1084 [preempt_disable] synchronize_r 1076 [preempt_disable] synchronize_rcu_expedited
1085 [and friends] 1077 [and friends]
1086 rcu_read_lock_sched_notrace 1078 rcu_read_lock_sched_notrace
1087 rcu_read_unlock_sched_notrace 1079 rcu_read_unlock_sched_notrace
1088 rcu_dereference_sched 1080 rcu_dereference_sched
1089 rcu_dereference_sched_check 1081 rcu_dereference_sched_check
1090 rcu_dereference_sched_protected 1082 rcu_dereference_sched_protected
1091 rcu_read_lock_sched_held 1083 rcu_read_lock_sched_held
1092 1084
1093 1085
1094 RCU-Tasks:: 1086 RCU-Tasks::
1095 1087
1096 Critical sections Grace period 1088 Critical sections Grace period Barrier
1097 1089
1098 N/A call_rcu_task 1090 N/A call_rcu_tasks rcu_barrier_tasks
1099 synchronize_r 1091 synchronize_rcu_tasks
1100 1092
1101 1093
1102 RCU-Tasks-Rude:: 1094 RCU-Tasks-Rude::
1103 1095
1104 Critical sections Grace period 1096 Critical sections Grace period Barrier
1105 1097
1106 N/A call_rcu_task 1098 N/A call_rcu_tasks_rude rcu_barrier_tasks_rude
1107 synchronize_r 1099 synchronize_rcu_tasks_rude
1108 1100
1109 1101
1110 RCU-Tasks-Trace:: 1102 RCU-Tasks-Trace::
1111 1103
1112 Critical sections Grace period 1104 Critical sections Grace period Barrier
1113 1105
1114 rcu_read_lock_trace call_rcu_task 1106 rcu_read_lock_trace call_rcu_tasks_trace rcu_barrier_tasks_trace
1115 rcu_read_unlock_trace synchronize_r 1107 rcu_read_unlock_trace synchronize_rcu_tasks_trace
1116 1108
1117 1109
1118 SRCU:: 1110 SRCU::
1119 1111
1120 Critical sections Grace period 1112 Critical sections Grace period Barrier
1121 1113
1122 srcu_read_lock call_srcu 1114 srcu_read_lock call_srcu srcu_barrier
1123 srcu_read_unlock synchronize_s 1115 srcu_read_unlock synchronize_srcu
1124 srcu_dereference synchronize_s 1116 srcu_dereference synchronize_srcu_expedited
1125 srcu_dereference_check 1117 srcu_dereference_check
1126 srcu_read_lock_held 1118 srcu_read_lock_held
1127 1119
1128 SRCU: Initialization/cleanup:: 1120 SRCU: Initialization/cleanup::
1129 1121
1130 DEFINE_SRCU 1122 DEFINE_SRCU
1131 DEFINE_STATIC_SRCU 1123 DEFINE_STATIC_SRCU
1132 init_srcu_struct 1124 init_srcu_struct
1133 cleanup_srcu_struct 1125 cleanup_srcu_struct
1134 1126
1135 All: lockdep-checked RCU utility APIs:: 1127 All: lockdep-checked RCU utility APIs::
1136 1128
1137 RCU_LOCKDEP_WARN 1129 RCU_LOCKDEP_WARN
1138 rcu_sleep_check 1130 rcu_sleep_check
1139 1131
1140 All: Unchecked RCU-protected pointer access:: 1132 All: Unchecked RCU-protected pointer access::
1141 1133
1142 rcu_dereference_raw 1134 rcu_dereference_raw
1143 1135
1144 All: Unchecked RCU-protected pointer access w 1136 All: Unchecked RCU-protected pointer access with dereferencing prohibited::
1145 1137
1146 rcu_access_pointer 1138 rcu_access_pointer
1147 1139
1148 See the comment headers in the source code (o 1140 See the comment headers in the source code (or the docbook generated
1149 from them) for more information. 1141 from them) for more information.
1150 1142
1151 However, given that there are no fewer than f 1143 However, given that there are no fewer than four families of RCU APIs
1152 in the Linux kernel, how do you choose which 1144 in the Linux kernel, how do you choose which one to use? The following
1153 list can be helpful: 1145 list can be helpful:
1154 1146
1155 a. Will readers need to block? If so, y 1147 a. Will readers need to block? If so, you need SRCU.
1156 1148
1157 b. Will readers need to block and are yo 1149 b. Will readers need to block and are you doing tracing, for
1158 example, ftrace or BPF? If so, you n 1150 example, ftrace or BPF? If so, you need RCU-tasks,
1159 RCU-tasks-rude, and/or RCU-tasks-trac 1151 RCU-tasks-rude, and/or RCU-tasks-trace.
1160 1152
1161 c. What about the -rt patchset? If read 1153 c. What about the -rt patchset? If readers would need to block in
1162 an non-rt kernel, you need SRCU. If 1154 an non-rt kernel, you need SRCU. If readers would block when
1163 acquiring spinlocks in a -rt kernel, 1155 acquiring spinlocks in a -rt kernel, but not in a non-rt kernel,
1164 SRCU is not necessary. (The -rt patc 1156 SRCU is not necessary. (The -rt patchset turns spinlocks into
1165 sleeplocks, hence this distinction.) 1157 sleeplocks, hence this distinction.)
1166 1158
1167 d. Do you need to treat NMI handlers, ha 1159 d. Do you need to treat NMI handlers, hardirq handlers,
1168 and code segments with preemption dis 1160 and code segments with preemption disabled (whether
1169 via preempt_disable(), local_irq_save 1161 via preempt_disable(), local_irq_save(), local_bh_disable(),
1170 or some other mechanism) as if they w 1162 or some other mechanism) as if they were explicit RCU readers?
1171 If so, RCU-sched readers are the only 1163 If so, RCU-sched readers are the only choice that will work
1172 for you, but since about v4.20 you us 1164 for you, but since about v4.20 you use can use the vanilla RCU
1173 update primitives. 1165 update primitives.
1174 1166
1175 e. Do you need RCU grace periods to comp 1167 e. Do you need RCU grace periods to complete even in the face of
1176 softirq monopolization of one or more 1168 softirq monopolization of one or more of the CPUs? For example,
1177 is your code subject to network-based 1169 is your code subject to network-based denial-of-service attacks?
1178 If so, you should disable softirq acr 1170 If so, you should disable softirq across your readers, for
1179 example, by using rcu_read_lock_bh(). 1171 example, by using rcu_read_lock_bh(). Since about v4.20 you
1180 use can use the vanilla RCU update pr 1172 use can use the vanilla RCU update primitives.
1181 1173
1182 f. Is your workload too update-intensive 1174 f. Is your workload too update-intensive for normal use of
1183 RCU, but inappropriate for other sync 1175 RCU, but inappropriate for other synchronization mechanisms?
1184 If so, consider SLAB_TYPESAFE_BY_RCU 1176 If so, consider SLAB_TYPESAFE_BY_RCU (which was originally
1185 named SLAB_DESTROY_BY_RCU). But plea 1177 named SLAB_DESTROY_BY_RCU). But please be careful!
1186 1178
1187 g. Do you need read-side critical sectio 1179 g. Do you need read-side critical sections that are respected even
1188 on CPUs that are deep in the idle loo 1180 on CPUs that are deep in the idle loop, during entry to or exit
1189 from user-mode execution, or on an of 1181 from user-mode execution, or on an offlined CPU? If so, SRCU
1190 and RCU Tasks Trace are the only choi 1182 and RCU Tasks Trace are the only choices that will work for you,
1191 with SRCU being strongly preferred in 1183 with SRCU being strongly preferred in almost all cases.
1192 1184
1193 h. Otherwise, use RCU. 1185 h. Otherwise, use RCU.
1194 1186
1195 Of course, this all assumes that you have det 1187 Of course, this all assumes that you have determined that RCU is in fact
1196 the right tool for your job. 1188 the right tool for your job.
1197 1189
1198 .. _9_whatisRCU: 1190 .. _9_whatisRCU:
1199 1191
1200 9. ANSWERS TO QUICK QUIZZES 1192 9. ANSWERS TO QUICK QUIZZES
1201 ---------------------------- 1193 ----------------------------
1202 1194
1203 Quick Quiz #1: 1195 Quick Quiz #1:
1204 Why is this argument naive? 1196 Why is this argument naive? How could a deadlock
1205 occur when using this algorit 1197 occur when using this algorithm in a real-world Linux
1206 kernel? [Referring to the lo 1198 kernel? [Referring to the lock-based "toy" RCU
1207 algorithm.] 1199 algorithm.]
1208 1200
1209 Answer: 1201 Answer:
1210 Consider the following sequen 1202 Consider the following sequence of events:
1211 1203
1212 1. CPU 0 acquires some u 1204 1. CPU 0 acquires some unrelated lock, call it
1213 "problematic_lock", d 1205 "problematic_lock", disabling irq via
1214 spin_lock_irqsave(). 1206 spin_lock_irqsave().
1215 1207
1216 2. CPU 1 enters synchron 1208 2. CPU 1 enters synchronize_rcu(), write-acquiring
1217 rcu_gp_mutex. 1209 rcu_gp_mutex.
1218 1210
1219 3. CPU 0 enters rcu_read 1211 3. CPU 0 enters rcu_read_lock(), but must wait
1220 because CPU 1 holds r 1212 because CPU 1 holds rcu_gp_mutex.
1221 1213
1222 4. CPU 1 is interrupted, 1214 4. CPU 1 is interrupted, and the irq handler
1223 attempts to acquire p 1215 attempts to acquire problematic_lock.
1224 1216
1225 The system is now deadlocked. 1217 The system is now deadlocked.
1226 1218
1227 One way to avoid this deadloc 1219 One way to avoid this deadlock is to use an approach like
1228 that of CONFIG_PREEMPT_RT, wh 1220 that of CONFIG_PREEMPT_RT, where all normal spinlocks
1229 become blocking locks, and al 1221 become blocking locks, and all irq handlers execute in
1230 the context of special tasks. 1222 the context of special tasks. In this case, in step 4
1231 above, the irq handler would 1223 above, the irq handler would block, allowing CPU 1 to
1232 release rcu_gp_mutex, avoidin 1224 release rcu_gp_mutex, avoiding the deadlock.
1233 1225
1234 Even in the absence of deadlo 1226 Even in the absence of deadlock, this RCU implementation
1235 allows latency to "bleed" fro 1227 allows latency to "bleed" from readers to other
1236 readers through synchronize_r 1228 readers through synchronize_rcu(). To see this,
1237 consider task A in an RCU rea 1229 consider task A in an RCU read-side critical section
1238 (thus read-holding rcu_gp_mut 1230 (thus read-holding rcu_gp_mutex), task B blocked
1239 attempting to write-acquire r 1231 attempting to write-acquire rcu_gp_mutex, and
1240 task C blocked in rcu_read_lo 1232 task C blocked in rcu_read_lock() attempting to
1241 read_acquire rcu_gp_mutex. T 1233 read_acquire rcu_gp_mutex. Task A's RCU read-side
1242 latency is holding up task C, 1234 latency is holding up task C, albeit indirectly via
1243 task B. 1235 task B.
1244 1236
1245 Realtime RCU implementations 1237 Realtime RCU implementations therefore use a counter-based
1246 approach where tasks in RCU r 1238 approach where tasks in RCU read-side critical sections
1247 cannot be blocked by tasks ex 1239 cannot be blocked by tasks executing synchronize_rcu().
1248 1240
1249 :ref:`Back to Quick Quiz #1 <quiz_1>` 1241 :ref:`Back to Quick Quiz #1 <quiz_1>`
1250 1242
1251 Quick Quiz #2: 1243 Quick Quiz #2:
1252 Give an example where Classic 1244 Give an example where Classic RCU's read-side
1253 overhead is **negative**. 1245 overhead is **negative**.
1254 1246
1255 Answer: 1247 Answer:
1256 Imagine a single-CPU system w 1248 Imagine a single-CPU system with a non-CONFIG_PREEMPTION
1257 kernel where a routing table 1249 kernel where a routing table is used by process-context
1258 code, but can be updated by i 1250 code, but can be updated by irq-context code (for example,
1259 by an "ICMP REDIRECT" packet) 1251 by an "ICMP REDIRECT" packet). The usual way of handling
1260 this would be to have the pro 1252 this would be to have the process-context code disable
1261 interrupts while searching th 1253 interrupts while searching the routing table. Use of
1262 RCU allows such interrupt-dis 1254 RCU allows such interrupt-disabling to be dispensed with.
1263 Thus, without RCU, you pay th 1255 Thus, without RCU, you pay the cost of disabling interrupts,
1264 and with RCU you don't. 1256 and with RCU you don't.
1265 1257
1266 One can argue that the overhe 1258 One can argue that the overhead of RCU in this
1267 case is negative with respect 1259 case is negative with respect to the single-CPU
1268 interrupt-disabling approach. 1260 interrupt-disabling approach. Others might argue that
1269 the overhead of RCU is merely 1261 the overhead of RCU is merely zero, and that replacing
1270 the positive overhead of the 1262 the positive overhead of the interrupt-disabling scheme
1271 with the zero-overhead RCU sc 1263 with the zero-overhead RCU scheme does not constitute
1272 negative overhead. 1264 negative overhead.
1273 1265
1274 In real life, of course, thin 1266 In real life, of course, things are more complex. But
1275 even the theoretical possibil 1267 even the theoretical possibility of negative overhead for
1276 a synchronization primitive i 1268 a synchronization primitive is a bit unexpected. ;-)
1277 1269
1278 :ref:`Back to Quick Quiz #2 <quiz_2>` 1270 :ref:`Back to Quick Quiz #2 <quiz_2>`
1279 1271
1280 Quick Quiz #3: 1272 Quick Quiz #3:
1281 If it is illegal to block in 1273 If it is illegal to block in an RCU read-side
1282 critical section, what the he 1274 critical section, what the heck do you do in
1283 CONFIG_PREEMPT_RT, where norm 1275 CONFIG_PREEMPT_RT, where normal spinlocks can block???
1284 1276
1285 Answer: 1277 Answer:
1286 Just as CONFIG_PREEMPT_RT per 1278 Just as CONFIG_PREEMPT_RT permits preemption of spinlock
1287 critical sections, it permits 1279 critical sections, it permits preemption of RCU
1288 read-side critical sections. 1280 read-side critical sections. It also permits
1289 spinlocks blocking while in R 1281 spinlocks blocking while in RCU read-side critical
1290 sections. 1282 sections.
1291 1283
1292 Why the apparent inconsistenc 1284 Why the apparent inconsistency? Because it is
1293 possible to use priority boos 1285 possible to use priority boosting to keep the RCU
1294 grace periods short if need b 1286 grace periods short if need be (for example, if running
1295 short of memory). In contras 1287 short of memory). In contrast, if blocking waiting
1296 for (say) network reception, 1288 for (say) network reception, there is no way to know
1297 what should be boosted. Espe 1289 what should be boosted. Especially given that the
1298 process we need to boost migh 1290 process we need to boost might well be a human being
1299 who just went out for a pizza 1291 who just went out for a pizza or something. And although
1300 a computer-operated cattle pr 1292 a computer-operated cattle prod might arouse serious
1301 interest, it might also provo 1293 interest, it might also provoke serious objections.
1302 Besides, how does the compute 1294 Besides, how does the computer know what pizza parlor
1303 the human being went to??? 1295 the human being went to???
1304 1296
1305 :ref:`Back to Quick Quiz #3 <quiz_3>` 1297 :ref:`Back to Quick Quiz #3 <quiz_3>`
1306 1298
1307 ACKNOWLEDGEMENTS 1299 ACKNOWLEDGEMENTS
1308 1300
1309 My thanks to the people who helped make this 1301 My thanks to the people who helped make this human-readable, including
1310 Jon Walpole, Josh Triplett, Serge Hallyn, Suz 1302 Jon Walpole, Josh Triplett, Serge Hallyn, Suzanne Wood, and Alan Stern.
1311 1303
1312 1304
1313 For more information, see http://www.rdrop.co 1305 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.