1 .. _rcu_barrier: 1 .. _rcu_barrier:
2 2
3 RCU and Unloadable Modules 3 RCU and Unloadable Modules
4 ========================== 4 ==========================
5 5
6 [Originally published in LWN Jan. 14, 2007: ht 6 [Originally published in LWN Jan. 14, 2007: http://lwn.net/Articles/217484/]
7 7
8 RCU updaters sometimes use call_rcu() to initi 8 RCU updaters sometimes use call_rcu() to initiate an asynchronous wait for
9 a grace period to elapse. This primitive take 9 a grace period to elapse. This primitive takes a pointer to an rcu_head
10 struct placed within the RCU-protected data st 10 struct placed within the RCU-protected data structure and another pointer
11 to a function that may be invoked later to fre 11 to a function that may be invoked later to free that structure. Code to
12 delete an element p from the linked list from 12 delete an element p from the linked list from IRQ context might then be
13 as follows:: 13 as follows::
14 14
15 list_del_rcu(p); 15 list_del_rcu(p);
16 call_rcu(&p->rcu, p_callback); 16 call_rcu(&p->rcu, p_callback);
17 17
18 Since call_rcu() never blocks, this code can s 18 Since call_rcu() never blocks, this code can safely be used from within
19 IRQ context. The function p_callback() might b 19 IRQ context. The function p_callback() might be defined as follows::
20 20
21 static void p_callback(struct rcu_head 21 static void p_callback(struct rcu_head *rp)
22 { 22 {
23 struct pstruct *p = container_ 23 struct pstruct *p = container_of(rp, struct pstruct, rcu);
24 24
25 kfree(p); 25 kfree(p);
26 } 26 }
27 27
28 28
29 Unloading Modules That Use call_rcu() 29 Unloading Modules That Use call_rcu()
30 ------------------------------------- 30 -------------------------------------
31 31
32 But what if the p_callback() function is defin 32 But what if the p_callback() function is defined in an unloadable module?
33 33
34 If we unload the module while some RCU callbac 34 If we unload the module while some RCU callbacks are pending,
35 the CPUs executing these callbacks are going t 35 the CPUs executing these callbacks are going to be severely
36 disappointed when they are later invoked, as f 36 disappointed when they are later invoked, as fancifully depicted at
37 http://lwn.net/images/ns/kernel/rcu-drop.jpg. 37 http://lwn.net/images/ns/kernel/rcu-drop.jpg.
38 38
39 We could try placing a synchronize_rcu() in th 39 We could try placing a synchronize_rcu() in the module-exit code path,
40 but this is not sufficient. Although synchroni 40 but this is not sufficient. Although synchronize_rcu() does wait for a
41 grace period to elapse, it does not wait for t 41 grace period to elapse, it does not wait for the callbacks to complete.
42 42
43 One might be tempted to try several back-to-ba 43 One might be tempted to try several back-to-back synchronize_rcu()
44 calls, but this is still not guaranteed to wor 44 calls, but this is still not guaranteed to work. If there is a very
45 heavy RCU-callback load, then some of the call 45 heavy RCU-callback load, then some of the callbacks might be deferred in
46 order to allow other processing to proceed. Fo 46 order to allow other processing to proceed. For but one example, such
47 deferral is required in realtime kernels in or 47 deferral is required in realtime kernels in order to avoid excessive
48 scheduling latencies. 48 scheduling latencies.
49 49
50 50
51 rcu_barrier() 51 rcu_barrier()
52 ------------- 52 -------------
53 53
54 This situation can be handled by the rcu_barri 54 This situation can be handled by the rcu_barrier() primitive. Rather
55 than waiting for a grace period to elapse, rcu 55 than waiting for a grace period to elapse, rcu_barrier() waits for all
56 outstanding RCU callbacks to complete. Please 56 outstanding RCU callbacks to complete. Please note that rcu_barrier()
57 does **not** imply synchronize_rcu(), in parti 57 does **not** imply synchronize_rcu(), in particular, if there are no RCU
58 callbacks queued anywhere, rcu_barrier() is wi 58 callbacks queued anywhere, rcu_barrier() is within its rights to return
59 immediately, without waiting for anything, let 59 immediately, without waiting for anything, let alone a grace period.
60 60
61 Pseudo-code using rcu_barrier() is as follows: 61 Pseudo-code using rcu_barrier() is as follows:
62 62
63 1. Prevent any new RCU callbacks from being 63 1. Prevent any new RCU callbacks from being posted.
64 2. Execute rcu_barrier(). 64 2. Execute rcu_barrier().
65 3. Allow the module to be unloaded. 65 3. Allow the module to be unloaded.
66 66
67 There is also an srcu_barrier() function for S 67 There is also an srcu_barrier() function for SRCU, and you of course
68 must match the flavor of srcu_barrier() with t 68 must match the flavor of srcu_barrier() with that of call_srcu().
69 If your module uses multiple srcu_struct struc 69 If your module uses multiple srcu_struct structures, then it must also
70 use multiple invocations of srcu_barrier() whe 70 use multiple invocations of srcu_barrier() when unloading that module.
71 For example, if it uses call_rcu(), call_srcu( 71 For example, if it uses call_rcu(), call_srcu() on srcu_struct_1, and
72 call_srcu() on srcu_struct_2, then the followi 72 call_srcu() on srcu_struct_2, then the following three lines of code
73 will be required when unloading:: 73 will be required when unloading::
74 74
75 1 rcu_barrier(); 75 1 rcu_barrier();
76 2 srcu_barrier(&srcu_struct_1); 76 2 srcu_barrier(&srcu_struct_1);
77 3 srcu_barrier(&srcu_struct_2); 77 3 srcu_barrier(&srcu_struct_2);
78 78
79 If latency is of the essence, workqueues could 79 If latency is of the essence, workqueues could be used to run these
80 three functions concurrently. 80 three functions concurrently.
81 81
82 An ancient version of the rcutorture module ma 82 An ancient version of the rcutorture module makes use of rcu_barrier()
83 in its exit function as follows:: 83 in its exit function as follows::
84 84
85 1 static void 85 1 static void
86 2 rcu_torture_cleanup(void) 86 2 rcu_torture_cleanup(void)
87 3 { 87 3 {
88 4 int i; 88 4 int i;
89 5 89 5
90 6 fullstop = 1; 90 6 fullstop = 1;
91 7 if (shuffler_task != NULL) { 91 7 if (shuffler_task != NULL) {
92 8 VERBOSE_PRINTK_STRING("Stopping rcu_t 92 8 VERBOSE_PRINTK_STRING("Stopping rcu_torture_shuffle task");
93 9 kthread_stop(shuffler_task); 93 9 kthread_stop(shuffler_task);
94 10 } 94 10 }
95 11 shuffler_task = NULL; 95 11 shuffler_task = NULL;
96 12 96 12
97 13 if (writer_task != NULL) { 97 13 if (writer_task != NULL) {
98 14 VERBOSE_PRINTK_STRING("Stopping rcu_t 98 14 VERBOSE_PRINTK_STRING("Stopping rcu_torture_writer task");
99 15 kthread_stop(writer_task); 99 15 kthread_stop(writer_task);
100 16 } 100 16 }
101 17 writer_task = NULL; 101 17 writer_task = NULL;
102 18 102 18
103 19 if (reader_tasks != NULL) { 103 19 if (reader_tasks != NULL) {
104 20 for (i = 0; i < nrealreaders; i++) { 104 20 for (i = 0; i < nrealreaders; i++) {
105 21 if (reader_tasks[i] != NULL) { 105 21 if (reader_tasks[i] != NULL) {
106 22 VERBOSE_PRINTK_STRING( 106 22 VERBOSE_PRINTK_STRING(
107 23 "Stopping rcu_torture_reader ta 107 23 "Stopping rcu_torture_reader task");
108 24 kthread_stop(reader_tasks[i]); 108 24 kthread_stop(reader_tasks[i]);
109 25 } 109 25 }
110 26 reader_tasks[i] = NULL; 110 26 reader_tasks[i] = NULL;
111 27 } 111 27 }
112 28 kfree(reader_tasks); 112 28 kfree(reader_tasks);
113 29 reader_tasks = NULL; 113 29 reader_tasks = NULL;
114 30 } 114 30 }
115 31 rcu_torture_current = NULL; 115 31 rcu_torture_current = NULL;
116 32 116 32
117 33 if (fakewriter_tasks != NULL) { 117 33 if (fakewriter_tasks != NULL) {
118 34 for (i = 0; i < nfakewriters; i++) { 118 34 for (i = 0; i < nfakewriters; i++) {
119 35 if (fakewriter_tasks[i] != NULL) { 119 35 if (fakewriter_tasks[i] != NULL) {
120 36 VERBOSE_PRINTK_STRING( 120 36 VERBOSE_PRINTK_STRING(
121 37 "Stopping rcu_torture_fakewrite 121 37 "Stopping rcu_torture_fakewriter task");
122 38 kthread_stop(fakewriter_tasks[i]) 122 38 kthread_stop(fakewriter_tasks[i]);
123 39 } 123 39 }
124 40 fakewriter_tasks[i] = NULL; 124 40 fakewriter_tasks[i] = NULL;
125 41 } 125 41 }
126 42 kfree(fakewriter_tasks); 126 42 kfree(fakewriter_tasks);
127 43 fakewriter_tasks = NULL; 127 43 fakewriter_tasks = NULL;
128 44 } 128 44 }
129 45 129 45
130 46 if (stats_task != NULL) { 130 46 if (stats_task != NULL) {
131 47 VERBOSE_PRINTK_STRING("Stopping rcu_t 131 47 VERBOSE_PRINTK_STRING("Stopping rcu_torture_stats task");
132 48 kthread_stop(stats_task); 132 48 kthread_stop(stats_task);
133 49 } 133 49 }
134 50 stats_task = NULL; 134 50 stats_task = NULL;
135 51 135 51
136 52 /* Wait for all RCU callbacks to fire. 136 52 /* Wait for all RCU callbacks to fire. */
137 53 rcu_barrier(); 137 53 rcu_barrier();
138 54 138 54
139 55 rcu_torture_stats_print(); /* -After- t 139 55 rcu_torture_stats_print(); /* -After- the stats thread is stopped! */
140 56 140 56
141 57 if (cur_ops->cleanup != NULL) 141 57 if (cur_ops->cleanup != NULL)
142 58 cur_ops->cleanup(); 142 58 cur_ops->cleanup();
143 59 if (atomic_read(&n_rcu_torture_error)) 143 59 if (atomic_read(&n_rcu_torture_error))
144 60 rcu_torture_print_module_parms("End o 144 60 rcu_torture_print_module_parms("End of test: FAILURE");
145 61 else 145 61 else
146 62 rcu_torture_print_module_parms("End o 146 62 rcu_torture_print_module_parms("End of test: SUCCESS");
147 63 } 147 63 }
148 148
149 Line 6 sets a global variable that prevents an 149 Line 6 sets a global variable that prevents any RCU callbacks from
150 re-posting themselves. This will not be necess 150 re-posting themselves. This will not be necessary in most cases, since
151 RCU callbacks rarely include calls to call_rcu 151 RCU callbacks rarely include calls to call_rcu(). However, the rcutorture
152 module is an exception to this rule, and there 152 module is an exception to this rule, and therefore needs to set this
153 global variable. 153 global variable.
154 154
155 Lines 7-50 stop all the kernel tasks associate 155 Lines 7-50 stop all the kernel tasks associated with the rcutorture
156 module. Therefore, once execution reaches line 156 module. Therefore, once execution reaches line 53, no more rcutorture
157 RCU callbacks will be posted. The rcu_barrier( 157 RCU callbacks will be posted. The rcu_barrier() call on line 53 waits
158 for any pre-existing callbacks to complete. 158 for any pre-existing callbacks to complete.
159 159
160 Then lines 55-62 print status and do operation 160 Then lines 55-62 print status and do operation-specific cleanup, and
161 then return, permitting the module-unload oper 161 then return, permitting the module-unload operation to be completed.
162 162
163 .. _rcubarrier_quiz_1: 163 .. _rcubarrier_quiz_1:
164 164
165 Quick Quiz #1: 165 Quick Quiz #1:
166 Is there any other situation where rcu 166 Is there any other situation where rcu_barrier() might
167 be required? 167 be required?
168 168
169 :ref:`Answer to Quick Quiz #1 <answer_rcubarri 169 :ref:`Answer to Quick Quiz #1 <answer_rcubarrier_quiz_1>`
170 170
171 Your module might have additional complication 171 Your module might have additional complications. For example, if your
172 module invokes call_rcu() from timers, you wil 172 module invokes call_rcu() from timers, you will need to first refrain
173 from posting new timers, cancel (or wait for) 173 from posting new timers, cancel (or wait for) all the already-posted
174 timers, and only then invoke rcu_barrier() to 174 timers, and only then invoke rcu_barrier() to wait for any remaining
175 RCU callbacks to complete. 175 RCU callbacks to complete.
176 176
177 Of course, if your module uses call_rcu(), you 177 Of course, if your module uses call_rcu(), you will need to invoke
178 rcu_barrier() before unloading. Similarly, if 178 rcu_barrier() before unloading. Similarly, if your module uses
179 call_srcu(), you will need to invoke srcu_barr 179 call_srcu(), you will need to invoke srcu_barrier() before unloading,
180 and on the same srcu_struct structure. If you 180 and on the same srcu_struct structure. If your module uses call_rcu()
181 **and** call_srcu(), then (as noted above) you 181 **and** call_srcu(), then (as noted above) you will need to invoke
182 rcu_barrier() **and** srcu_barrier(). 182 rcu_barrier() **and** srcu_barrier().
183 183
184 184
185 Implementing rcu_barrier() 185 Implementing rcu_barrier()
186 -------------------------- 186 --------------------------
187 187
188 Dipankar Sarma's implementation of rcu_barrier 188 Dipankar Sarma's implementation of rcu_barrier() makes use of the fact
189 that RCU callbacks are never reordered once qu 189 that RCU callbacks are never reordered once queued on one of the per-CPU
190 queues. His implementation queues an RCU callb 190 queues. His implementation queues an RCU callback on each of the per-CPU
191 callback queues, and then waits until they hav 191 callback queues, and then waits until they have all started executing, at
192 which point, all earlier RCU callbacks are gua 192 which point, all earlier RCU callbacks are guaranteed to have completed.
193 193
194 The original code for rcu_barrier() was roughl 194 The original code for rcu_barrier() was roughly as follows::
195 195
196 1 void rcu_barrier(void) 196 1 void rcu_barrier(void)
197 2 { 197 2 {
198 3 BUG_ON(in_interrupt()); 198 3 BUG_ON(in_interrupt());
199 4 /* Take cpucontrol mutex to protect aga 199 4 /* Take cpucontrol mutex to protect against CPU hotplug */
200 5 mutex_lock(&rcu_barrier_mutex); 200 5 mutex_lock(&rcu_barrier_mutex);
201 6 init_completion(&rcu_barrier_completion 201 6 init_completion(&rcu_barrier_completion);
202 7 atomic_set(&rcu_barrier_cpu_count, 1); 202 7 atomic_set(&rcu_barrier_cpu_count, 1);
203 8 on_each_cpu(rcu_barrier_func, NULL, 0, 203 8 on_each_cpu(rcu_barrier_func, NULL, 0, 1);
204 9 if (atomic_dec_and_test(&rcu_barrier_cp 204 9 if (atomic_dec_and_test(&rcu_barrier_cpu_count))
205 10 complete(&rcu_barrier_completion); 205 10 complete(&rcu_barrier_completion);
206 11 wait_for_completion(&rcu_barrier_comple 206 11 wait_for_completion(&rcu_barrier_completion);
207 12 mutex_unlock(&rcu_barrier_mutex); 207 12 mutex_unlock(&rcu_barrier_mutex);
208 13 } 208 13 }
209 209
210 Line 3 verifies that the caller is in process 210 Line 3 verifies that the caller is in process context, and lines 5 and 12
211 use rcu_barrier_mutex to ensure that only one 211 use rcu_barrier_mutex to ensure that only one rcu_barrier() is using the
212 global completion and counters at a time, whic 212 global completion and counters at a time, which are initialized on lines
213 6 and 7. Line 8 causes each CPU to invoke rcu_ 213 6 and 7. Line 8 causes each CPU to invoke rcu_barrier_func(), which is
214 shown below. Note that the final "1" in on_eac 214 shown below. Note that the final "1" in on_each_cpu()'s argument list
215 ensures that all the calls to rcu_barrier_func 215 ensures that all the calls to rcu_barrier_func() will have completed
216 before on_each_cpu() returns. Line 9 removes t 216 before on_each_cpu() returns. Line 9 removes the initial count from
217 rcu_barrier_cpu_count, and if this count is no 217 rcu_barrier_cpu_count, and if this count is now zero, line 10 finalizes
218 the completion, which prevents line 11 from bl 218 the completion, which prevents line 11 from blocking. Either way,
219 line 11 then waits (if needed) for the complet 219 line 11 then waits (if needed) for the completion.
220 220
221 .. _rcubarrier_quiz_2: 221 .. _rcubarrier_quiz_2:
222 222
223 Quick Quiz #2: 223 Quick Quiz #2:
224 Why doesn't line 8 initialize rcu_barr 224 Why doesn't line 8 initialize rcu_barrier_cpu_count to zero,
225 thereby avoiding the need for lines 9 225 thereby avoiding the need for lines 9 and 10?
226 226
227 :ref:`Answer to Quick Quiz #2 <answer_rcubarri 227 :ref:`Answer to Quick Quiz #2 <answer_rcubarrier_quiz_2>`
228 228
229 This code was rewritten in 2008 and several ti 229 This code was rewritten in 2008 and several times thereafter, but this
230 still gives the general idea. 230 still gives the general idea.
231 231
232 The rcu_barrier_func() runs on each CPU, where 232 The rcu_barrier_func() runs on each CPU, where it invokes call_rcu()
233 to post an RCU callback, as follows:: 233 to post an RCU callback, as follows::
234 234
235 1 static void rcu_barrier_func(void *notuse 235 1 static void rcu_barrier_func(void *notused)
236 2 { 236 2 {
237 3 int cpu = smp_processor_id(); 237 3 int cpu = smp_processor_id();
238 4 struct rcu_data *rdp = &per_cpu(rcu_dat 238 4 struct rcu_data *rdp = &per_cpu(rcu_data, cpu);
239 5 struct rcu_head *head; 239 5 struct rcu_head *head;
240 6 240 6
241 7 head = &rdp->barrier; 241 7 head = &rdp->barrier;
242 8 atomic_inc(&rcu_barrier_cpu_count); 242 8 atomic_inc(&rcu_barrier_cpu_count);
243 9 call_rcu(head, rcu_barrier_callback); 243 9 call_rcu(head, rcu_barrier_callback);
244 10 } 244 10 }
245 245
246 Lines 3 and 4 locate RCU's internal per-CPU rc 246 Lines 3 and 4 locate RCU's internal per-CPU rcu_data structure,
247 which contains the struct rcu_head that needed 247 which contains the struct rcu_head that needed for the later call to
248 call_rcu(). Line 7 picks up a pointer to this 248 call_rcu(). Line 7 picks up a pointer to this struct rcu_head, and line
249 8 increments the global counter. This counter 249 8 increments the global counter. This counter will later be decremented
250 by the callback. Line 9 then registers the rcu 250 by the callback. Line 9 then registers the rcu_barrier_callback() on
251 the current CPU's queue. 251 the current CPU's queue.
252 252
253 The rcu_barrier_callback() function simply ato 253 The rcu_barrier_callback() function simply atomically decrements the
254 rcu_barrier_cpu_count variable and finalizes t 254 rcu_barrier_cpu_count variable and finalizes the completion when it
255 reaches zero, as follows:: 255 reaches zero, as follows::
256 256
257 1 static void rcu_barrier_callback(struct r 257 1 static void rcu_barrier_callback(struct rcu_head *notused)
258 2 { 258 2 {
259 3 if (atomic_dec_and_test(&rcu_barrier_cp 259 3 if (atomic_dec_and_test(&rcu_barrier_cpu_count))
260 4 complete(&rcu_barrier_completion); 260 4 complete(&rcu_barrier_completion);
261 5 } 261 5 }
262 262
263 .. _rcubarrier_quiz_3: 263 .. _rcubarrier_quiz_3:
264 264
265 Quick Quiz #3: 265 Quick Quiz #3:
266 What happens if CPU 0's rcu_barrier_fu 266 What happens if CPU 0's rcu_barrier_func() executes
267 immediately (thus incrementing rcu_bar 267 immediately (thus incrementing rcu_barrier_cpu_count to the
268 value one), but the other CPU's rcu_ba 268 value one), but the other CPU's rcu_barrier_func() invocations
269 are delayed for a full grace period? C 269 are delayed for a full grace period? Couldn't this result in
270 rcu_barrier() returning prematurely? 270 rcu_barrier() returning prematurely?
271 271
272 :ref:`Answer to Quick Quiz #3 <answer_rcubarri 272 :ref:`Answer to Quick Quiz #3 <answer_rcubarrier_quiz_3>`
273 273
274 The current rcu_barrier() implementation is mo 274 The current rcu_barrier() implementation is more complex, due to the need
275 to avoid disturbing idle CPUs (especially on b 275 to avoid disturbing idle CPUs (especially on battery-powered systems)
276 and the need to minimally disturb non-idle CPU 276 and the need to minimally disturb non-idle CPUs in real-time systems.
277 In addition, a great many optimizations have b 277 In addition, a great many optimizations have been applied. However,
278 the code above illustrates the concepts. 278 the code above illustrates the concepts.
279 279
280 280
281 rcu_barrier() Summary 281 rcu_barrier() Summary
282 --------------------- 282 ---------------------
283 283
284 The rcu_barrier() primitive is used relatively 284 The rcu_barrier() primitive is used relatively infrequently, since most
285 code using RCU is in the core kernel rather th 285 code using RCU is in the core kernel rather than in modules. However, if
286 you are using RCU from an unloadable module, y 286 you are using RCU from an unloadable module, you need to use rcu_barrier()
287 so that your module may be safely unloaded. 287 so that your module may be safely unloaded.
288 288
289 289
290 Answers to Quick Quizzes 290 Answers to Quick Quizzes
291 ------------------------ 291 ------------------------
292 292
293 .. _answer_rcubarrier_quiz_1: 293 .. _answer_rcubarrier_quiz_1:
294 294
295 Quick Quiz #1: 295 Quick Quiz #1:
296 Is there any other situation where rcu 296 Is there any other situation where rcu_barrier() might
297 be required? 297 be required?
298 298
299 Answer: 299 Answer:
300 Interestingly enough, rcu_barrier() wa 300 Interestingly enough, rcu_barrier() was not originally
301 implemented for module unloading. Niki 301 implemented for module unloading. Nikita Danilov was using
302 RCU in a filesystem, which resulted in 302 RCU in a filesystem, which resulted in a similar situation at
303 filesystem-unmount time. Dipankar Sarm 303 filesystem-unmount time. Dipankar Sarma coded up rcu_barrier()
304 in response, so that Nikita could invo 304 in response, so that Nikita could invoke it during the
305 filesystem-unmount process. 305 filesystem-unmount process.
306 306
307 Much later, yours truly hit the RCU mo 307 Much later, yours truly hit the RCU module-unload problem when
308 implementing rcutorture, and found tha 308 implementing rcutorture, and found that rcu_barrier() solves
309 this problem as well. 309 this problem as well.
310 310
311 :ref:`Back to Quick Quiz #1 <rcubarrier_quiz_1 311 :ref:`Back to Quick Quiz #1 <rcubarrier_quiz_1>`
312 312
313 .. _answer_rcubarrier_quiz_2: 313 .. _answer_rcubarrier_quiz_2:
314 314
315 Quick Quiz #2: 315 Quick Quiz #2:
316 Why doesn't line 8 initialize rcu_barr 316 Why doesn't line 8 initialize rcu_barrier_cpu_count to zero,
317 thereby avoiding the need for lines 9 317 thereby avoiding the need for lines 9 and 10?
318 318
319 Answer: 319 Answer:
320 Suppose that the on_each_cpu() functio 320 Suppose that the on_each_cpu() function shown on line 8 was
321 delayed, so that CPU 0's rcu_barrier_f 321 delayed, so that CPU 0's rcu_barrier_func() executed and
322 the corresponding grace period elapsed 322 the corresponding grace period elapsed, all before CPU 1's
323 rcu_barrier_func() started executing. 323 rcu_barrier_func() started executing. This would result in
324 rcu_barrier_cpu_count being decremente 324 rcu_barrier_cpu_count being decremented to zero, so that line
325 11's wait_for_completion() would retur 325 11's wait_for_completion() would return immediately, failing to
326 wait for CPU 1's callbacks to be invok 326 wait for CPU 1's callbacks to be invoked.
327 327
328 Note that this was not a problem when 328 Note that this was not a problem when the rcu_barrier() code
329 was first added back in 2005. This is 329 was first added back in 2005. This is because on_each_cpu()
330 disables preemption, which acted as an 330 disables preemption, which acted as an RCU read-side critical
331 section, thus preventing CPU 0's grace 331 section, thus preventing CPU 0's grace period from completing
332 until on_each_cpu() had dealt with all 332 until on_each_cpu() had dealt with all of the CPUs. However,
333 with the advent of preemptible RCU, rc 333 with the advent of preemptible RCU, rcu_barrier() no longer
334 waited on nonpreemptible regions of co 334 waited on nonpreemptible regions of code in preemptible kernels,
335 that being the job of the new rcu_barr 335 that being the job of the new rcu_barrier_sched() function.
336 336
337 However, with the RCU flavor consolida 337 However, with the RCU flavor consolidation around v4.20, this
338 possibility was once again ruled out, 338 possibility was once again ruled out, because the consolidated
339 RCU once again waits on nonpreemptible 339 RCU once again waits on nonpreemptible regions of code.
340 340
341 Nevertheless, that extra count might s 341 Nevertheless, that extra count might still be a good idea.
342 Relying on these sort of accidents of 342 Relying on these sort of accidents of implementation can result
343 in later surprise bugs when the implem 343 in later surprise bugs when the implementation changes.
344 344
345 :ref:`Back to Quick Quiz #2 <rcubarrier_quiz_2 345 :ref:`Back to Quick Quiz #2 <rcubarrier_quiz_2>`
346 346
347 .. _answer_rcubarrier_quiz_3: 347 .. _answer_rcubarrier_quiz_3:
348 348
349 Quick Quiz #3: 349 Quick Quiz #3:
350 What happens if CPU 0's rcu_barrier_fu 350 What happens if CPU 0's rcu_barrier_func() executes
351 immediately (thus incrementing rcu_bar 351 immediately (thus incrementing rcu_barrier_cpu_count to the
352 value one), but the other CPU's rcu_ba 352 value one), but the other CPU's rcu_barrier_func() invocations
353 are delayed for a full grace period? C 353 are delayed for a full grace period? Couldn't this result in
354 rcu_barrier() returning prematurely? 354 rcu_barrier() returning prematurely?
355 355
356 Answer: 356 Answer:
357 This cannot happen. The reason is that 357 This cannot happen. The reason is that on_each_cpu() has its last
358 argument, the wait flag, set to "1". T 358 argument, the wait flag, set to "1". This flag is passed through
359 to smp_call_function() and further to 359 to smp_call_function() and further to smp_call_function_on_cpu(),
360 causing this latter to spin until the 360 causing this latter to spin until the cross-CPU invocation of
361 rcu_barrier_func() has completed. This 361 rcu_barrier_func() has completed. This by itself would prevent
362 a grace period from completing on non- 362 a grace period from completing on non-CONFIG_PREEMPTION kernels,
363 since each CPU must undergo a context 363 since each CPU must undergo a context switch (or other quiescent
364 state) before the grace period can com 364 state) before the grace period can complete. However, this is
365 of no use in CONFIG_PREEMPTION kernels 365 of no use in CONFIG_PREEMPTION kernels.
366 366
367 Therefore, on_each_cpu() disables pree 367 Therefore, on_each_cpu() disables preemption across its call
368 to smp_call_function() and also across 368 to smp_call_function() and also across the local call to
369 rcu_barrier_func(). Because recent RCU 369 rcu_barrier_func(). Because recent RCU implementations treat
370 preemption-disabled regions of code as 370 preemption-disabled regions of code as RCU read-side critical
371 sections, this prevents grace periods 371 sections, this prevents grace periods from completing. This
372 means that all CPUs have executed rcu_ 372 means that all CPUs have executed rcu_barrier_func() before
373 the first rcu_barrier_callback() can p 373 the first rcu_barrier_callback() can possibly execute, in turn
374 preventing rcu_barrier_cpu_count from 374 preventing rcu_barrier_cpu_count from prematurely reaching zero.
375 375
376 But if on_each_cpu() ever decides to f 376 But if on_each_cpu() ever decides to forgo disabling preemption,
377 as might well happen due to real-time 377 as might well happen due to real-time latency considerations,
378 initializing rcu_barrier_cpu_count to 378 initializing rcu_barrier_cpu_count to one will save the day.
379 379
380 :ref:`Back to Quick Quiz #3 <rcubarrier_quiz_3 380 :ref:`Back to Quick Quiz #3 <rcubarrier_quiz_3>`
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.