1 ===================== 1 ============================
2 LINUX KERNEL MEMORY B 2 LINUX KERNEL MEMORY BARRIERS
3 ===================== 3 ============================
4 4
5 By: David Howells <dhowells@redhat.com> 5 By: David Howells <dhowells@redhat.com>
6 Paul E. McKenney <paulmck@linux.ibm.com> !! 6 Paul E. McKenney <paulmck@linux.vnet.ibm.com>
7 Will Deacon <will.deacon@arm.com> 7 Will Deacon <will.deacon@arm.com>
8 Peter Zijlstra <peterz@infradead.org> 8 Peter Zijlstra <peterz@infradead.org>
9 9
10 ========== 10 ==========
11 DISCLAIMER 11 DISCLAIMER
12 ========== 12 ==========
13 13
14 This document is not a specification; it is in 14 This document is not a specification; it is intentionally (for the sake of
15 brevity) and unintentionally (due to being hum 15 brevity) and unintentionally (due to being human) incomplete. This document is
16 meant as a guide to using the various memory b 16 meant as a guide to using the various memory barriers provided by Linux, but
17 in case of any doubt (and there are many) plea 17 in case of any doubt (and there are many) please ask. Some doubts may be
18 resolved by referring to the formal memory con 18 resolved by referring to the formal memory consistency model and related
19 documentation at tools/memory-model/. Neverth 19 documentation at tools/memory-model/. Nevertheless, even this memory
20 model should be viewed as the collective opini 20 model should be viewed as the collective opinion of its maintainers rather
21 than as an infallible oracle. 21 than as an infallible oracle.
22 22
23 To repeat, this document is not a specificatio 23 To repeat, this document is not a specification of what Linux expects from
24 hardware. 24 hardware.
25 25
26 The purpose of this document is twofold: 26 The purpose of this document is twofold:
27 27
28 (1) to specify the minimum functionality that 28 (1) to specify the minimum functionality that one can rely on for any
29 particular barrier, and 29 particular barrier, and
30 30
31 (2) to provide a guide as to how to use the b 31 (2) to provide a guide as to how to use the barriers that are available.
32 32
33 Note that an architecture can provide more tha 33 Note that an architecture can provide more than the minimum requirement
34 for any particular barrier, but if the archite 34 for any particular barrier, but if the architecture provides less than
35 that, that architecture is incorrect. 35 that, that architecture is incorrect.
36 36
37 Note also that it is possible that a barrier m 37 Note also that it is possible that a barrier may be a no-op for an
38 architecture because the way that arch works r 38 architecture because the way that arch works renders an explicit barrier
39 unnecessary in that case. 39 unnecessary in that case.
40 40
41 41
42 ======== 42 ========
43 CONTENTS 43 CONTENTS
44 ======== 44 ========
45 45
46 (*) Abstract memory access model. 46 (*) Abstract memory access model.
47 47
48 - Device operations. 48 - Device operations.
49 - Guarantees. 49 - Guarantees.
50 50
51 (*) What are memory barriers? 51 (*) What are memory barriers?
52 52
53 - Varieties of memory barrier. 53 - Varieties of memory barrier.
54 - What may not be assumed about memory ba 54 - What may not be assumed about memory barriers?
55 - Address-dependency barriers (historical !! 55 - Data dependency barriers (historical).
56 - Control dependencies. 56 - Control dependencies.
57 - SMP barrier pairing. 57 - SMP barrier pairing.
58 - Examples of memory barrier sequences. 58 - Examples of memory barrier sequences.
59 - Read memory barriers vs load speculatio 59 - Read memory barriers vs load speculation.
60 - Multicopy atomicity. 60 - Multicopy atomicity.
61 61
62 (*) Explicit kernel barriers. 62 (*) Explicit kernel barriers.
63 63
64 - Compiler barrier. 64 - Compiler barrier.
65 - CPU memory barriers. 65 - CPU memory barriers.
>> 66 - MMIO write barrier.
66 67
67 (*) Implicit kernel memory barriers. 68 (*) Implicit kernel memory barriers.
68 69
69 - Lock acquisition functions. 70 - Lock acquisition functions.
70 - Interrupt disabling functions. 71 - Interrupt disabling functions.
71 - Sleep and wake-up functions. 72 - Sleep and wake-up functions.
72 - Miscellaneous functions. 73 - Miscellaneous functions.
73 74
74 (*) Inter-CPU acquiring barrier effects. 75 (*) Inter-CPU acquiring barrier effects.
75 76
76 - Acquires vs memory accesses. 77 - Acquires vs memory accesses.
>> 78 - Acquires vs I/O accesses.
77 79
78 (*) Where are memory barriers needed? 80 (*) Where are memory barriers needed?
79 81
80 - Interprocessor interaction. 82 - Interprocessor interaction.
81 - Atomic operations. 83 - Atomic operations.
82 - Accessing devices. 84 - Accessing devices.
83 - Interrupts. 85 - Interrupts.
84 86
85 (*) Kernel I/O barrier effects. 87 (*) Kernel I/O barrier effects.
86 88
87 (*) Assumed minimum execution ordering model. 89 (*) Assumed minimum execution ordering model.
88 90
89 (*) The effects of the cpu cache. 91 (*) The effects of the cpu cache.
90 92
91 - Cache coherency. 93 - Cache coherency.
92 - Cache coherency vs DMA. 94 - Cache coherency vs DMA.
93 - Cache coherency vs MMIO. 95 - Cache coherency vs MMIO.
94 96
95 (*) The things CPUs get up to. 97 (*) The things CPUs get up to.
96 98
97 - And then there's the Alpha. 99 - And then there's the Alpha.
98 - Virtual Machine Guests. 100 - Virtual Machine Guests.
99 101
100 (*) Example uses. 102 (*) Example uses.
101 103
102 - Circular buffers. 104 - Circular buffers.
103 105
104 (*) References. 106 (*) References.
105 107
106 108
107 ============================ 109 ============================
108 ABSTRACT MEMORY ACCESS MODEL 110 ABSTRACT MEMORY ACCESS MODEL
109 ============================ 111 ============================
110 112
111 Consider the following abstract model of the s 113 Consider the following abstract model of the system:
112 114
113 : : 115 : :
114 : : 116 : :
115 : : 117 : :
116 +-------+ : +--------+ : 118 +-------+ : +--------+ : +-------+
117 | | : | | : 119 | | : | | : | |
118 | | : | | : 120 | | : | | : | |
119 | CPU 1 |<----->| Memory |<--- 121 | CPU 1 |<----->| Memory |<----->| CPU 2 |
120 | | : | | : 122 | | : | | : | |
121 | | : | | : 123 | | : | | : | |
122 +-------+ : +--------+ : 124 +-------+ : +--------+ : +-------+
123 ^ : ^ : 125 ^ : ^ : ^
124 | : | : 126 | : | : |
125 | : | : 127 | : | : |
126 | : v : 128 | : v : |
127 | : +--------+ : 129 | : +--------+ : |
128 | : | | : 130 | : | | : |
129 | : | | : 131 | : | | : |
130 +---------->| Device |<--- 132 +---------->| Device |<----------+
131 : | | : 133 : | | :
132 : | | : 134 : | | :
133 : +--------+ : 135 : +--------+ :
134 : : 136 : :
135 137
136 Each CPU executes a program that generates mem 138 Each CPU executes a program that generates memory access operations. In the
137 abstract CPU, memory operation ordering is ver 139 abstract CPU, memory operation ordering is very relaxed, and a CPU may actually
138 perform the memory operations in any order it 140 perform the memory operations in any order it likes, provided program causality
139 appears to be maintained. Similarly, the comp 141 appears to be maintained. Similarly, the compiler may also arrange the
140 instructions it emits in any order it likes, p 142 instructions it emits in any order it likes, provided it doesn't affect the
141 apparent operation of the program. 143 apparent operation of the program.
142 144
143 So in the above diagram, the effects of the me 145 So in the above diagram, the effects of the memory operations performed by a
144 CPU are perceived by the rest of the system as 146 CPU are perceived by the rest of the system as the operations cross the
145 interface between the CPU and rest of the syst 147 interface between the CPU and rest of the system (the dotted lines).
146 148
147 149
148 For example, consider the following sequence o 150 For example, consider the following sequence of events:
149 151
150 CPU 1 CPU 2 152 CPU 1 CPU 2
151 =============== =============== 153 =============== ===============
152 { A == 1; B == 2 } 154 { A == 1; B == 2 }
153 A = 3; x = B; 155 A = 3; x = B;
154 B = 4; y = A; 156 B = 4; y = A;
155 157
156 The set of accesses as seen by the memory syst 158 The set of accesses as seen by the memory system in the middle can be arranged
157 in 24 different combinations: 159 in 24 different combinations:
158 160
159 STORE A=3, STORE B=4, y=LOAD 161 STORE A=3, STORE B=4, y=LOAD A->3, x=LOAD B->4
160 STORE A=3, STORE B=4, x=LOAD 162 STORE A=3, STORE B=4, x=LOAD B->4, y=LOAD A->3
161 STORE A=3, y=LOAD A->3, STORE 163 STORE A=3, y=LOAD A->3, STORE B=4, x=LOAD B->4
162 STORE A=3, y=LOAD A->3, x=LOAD 164 STORE A=3, y=LOAD A->3, x=LOAD B->2, STORE B=4
163 STORE A=3, x=LOAD B->2, STORE 165 STORE A=3, x=LOAD B->2, STORE B=4, y=LOAD A->3
164 STORE A=3, x=LOAD B->2, y=LOAD 166 STORE A=3, x=LOAD B->2, y=LOAD A->3, STORE B=4
165 STORE B=4, STORE A=3, y=LOAD 167 STORE B=4, STORE A=3, y=LOAD A->3, x=LOAD B->4
166 STORE B=4, ... 168 STORE B=4, ...
167 ... 169 ...
168 170
169 and can thus result in four different combinat 171 and can thus result in four different combinations of values:
170 172
171 x == 2, y == 1 173 x == 2, y == 1
172 x == 2, y == 3 174 x == 2, y == 3
173 x == 4, y == 1 175 x == 4, y == 1
174 x == 4, y == 3 176 x == 4, y == 3
175 177
176 178
177 Furthermore, the stores committed by a CPU to 179 Furthermore, the stores committed by a CPU to the memory system may not be
178 perceived by the loads made by another CPU in 180 perceived by the loads made by another CPU in the same order as the stores were
179 committed. 181 committed.
180 182
181 183
182 As a further example, consider this sequence o 184 As a further example, consider this sequence of events:
183 185
184 CPU 1 CPU 2 186 CPU 1 CPU 2
185 =============== =============== 187 =============== ===============
186 { A == 1, B == 2, C == 3, P == &A, Q = 188 { A == 1, B == 2, C == 3, P == &A, Q == &C }
187 B = 4; Q = P; 189 B = 4; Q = P;
188 P = &B; D = *Q; !! 190 P = &B D = *Q;
189 191
190 There is an obvious address dependency here, a !! 192 There is an obvious data dependency here, as the value loaded into D depends on
191 on the address retrieved from P by CPU 2. At !! 193 the address retrieved from P by CPU 2. At the end of the sequence, any of the
192 the following results are possible: !! 194 following results are possible:
193 195
194 (Q == &A) and (D == 1) 196 (Q == &A) and (D == 1)
195 (Q == &B) and (D == 2) 197 (Q == &B) and (D == 2)
196 (Q == &B) and (D == 4) 198 (Q == &B) and (D == 4)
197 199
198 Note that CPU 2 will never try and load C into 200 Note that CPU 2 will never try and load C into D because the CPU will load P
199 into Q before issuing the load of *Q. 201 into Q before issuing the load of *Q.
200 202
201 203
202 DEVICE OPERATIONS 204 DEVICE OPERATIONS
203 ----------------- 205 -----------------
204 206
205 Some devices present their control interfaces 207 Some devices present their control interfaces as collections of memory
206 locations, but the order in which the control 208 locations, but the order in which the control registers are accessed is very
207 important. For instance, imagine an ethernet 209 important. For instance, imagine an ethernet card with a set of internal
208 registers that are accessed through an address 210 registers that are accessed through an address port register (A) and a data
209 port register (D). To read internal register 211 port register (D). To read internal register 5, the following code might then
210 be used: 212 be used:
211 213
212 *A = 5; 214 *A = 5;
213 x = *D; 215 x = *D;
214 216
215 but this might show up as either of the follow 217 but this might show up as either of the following two sequences:
216 218
217 STORE *A = 5, x = LOAD *D 219 STORE *A = 5, x = LOAD *D
218 x = LOAD *D, STORE *A = 5 220 x = LOAD *D, STORE *A = 5
219 221
220 the second of which will almost certainly resu 222 the second of which will almost certainly result in a malfunction, since it set
221 the address _after_ attempting to read the reg 223 the address _after_ attempting to read the register.
222 224
223 225
224 GUARANTEES 226 GUARANTEES
225 ---------- 227 ----------
226 228
227 There are some minimal guarantees that may be 229 There are some minimal guarantees that may be expected of a CPU:
228 230
229 (*) On any given CPU, dependent memory access 231 (*) On any given CPU, dependent memory accesses will be issued in order, with
230 respect to itself. This means that for: 232 respect to itself. This means that for:
231 233
232 Q = READ_ONCE(P); D = READ_ONCE(*Q); 234 Q = READ_ONCE(P); D = READ_ONCE(*Q);
233 235
234 the CPU will issue the following memory o 236 the CPU will issue the following memory operations:
235 237
236 Q = LOAD P, D = LOAD *Q 238 Q = LOAD P, D = LOAD *Q
237 239
238 and always in that order. However, on DE 240 and always in that order. However, on DEC Alpha, READ_ONCE() also
239 emits a memory-barrier instruction, so th 241 emits a memory-barrier instruction, so that a DEC Alpha CPU will
240 instead issue the following memory operat 242 instead issue the following memory operations:
241 243
242 Q = LOAD P, MEMORY_BARRIER, D = LOAD * 244 Q = LOAD P, MEMORY_BARRIER, D = LOAD *Q, MEMORY_BARRIER
243 245
244 Whether on DEC Alpha or not, the READ_ONC 246 Whether on DEC Alpha or not, the READ_ONCE() also prevents compiler
245 mischief. 247 mischief.
246 248
247 (*) Overlapping loads and stores within a par 249 (*) Overlapping loads and stores within a particular CPU will appear to be
248 ordered within that CPU. This means that 250 ordered within that CPU. This means that for:
249 251
250 a = READ_ONCE(*X); WRITE_ONCE(*X, b); 252 a = READ_ONCE(*X); WRITE_ONCE(*X, b);
251 253
252 the CPU will only issue the following seq 254 the CPU will only issue the following sequence of memory operations:
253 255
254 a = LOAD *X, STORE *X = b 256 a = LOAD *X, STORE *X = b
255 257
256 And for: 258 And for:
257 259
258 WRITE_ONCE(*X, c); d = READ_ONCE(*X); 260 WRITE_ONCE(*X, c); d = READ_ONCE(*X);
259 261
260 the CPU will only issue: 262 the CPU will only issue:
261 263
262 STORE *X = c, d = LOAD *X 264 STORE *X = c, d = LOAD *X
263 265
264 (Loads and stores overlap if they are tar 266 (Loads and stores overlap if they are targeted at overlapping pieces of
265 memory). 267 memory).
266 268
267 And there are a number of things that _must_ o 269 And there are a number of things that _must_ or _must_not_ be assumed:
268 270
269 (*) It _must_not_ be assumed that the compile 271 (*) It _must_not_ be assumed that the compiler will do what you want
270 with memory references that are not prote 272 with memory references that are not protected by READ_ONCE() and
271 WRITE_ONCE(). Without them, the compiler 273 WRITE_ONCE(). Without them, the compiler is within its rights to
272 do all sorts of "creative" transformation 274 do all sorts of "creative" transformations, which are covered in
273 the COMPILER BARRIER section. 275 the COMPILER BARRIER section.
274 276
275 (*) It _must_not_ be assumed that independent 277 (*) It _must_not_ be assumed that independent loads and stores will be issued
276 in the order given. This means that for: 278 in the order given. This means that for:
277 279
278 X = *A; Y = *B; *D = Z; 280 X = *A; Y = *B; *D = Z;
279 281
280 we may get any of the following sequences 282 we may get any of the following sequences:
281 283
282 X = LOAD *A, Y = LOAD *B, STORE *D = 284 X = LOAD *A, Y = LOAD *B, STORE *D = Z
283 X = LOAD *A, STORE *D = Z, Y = LOAD * 285 X = LOAD *A, STORE *D = Z, Y = LOAD *B
284 Y = LOAD *B, X = LOAD *A, STORE *D = 286 Y = LOAD *B, X = LOAD *A, STORE *D = Z
285 Y = LOAD *B, STORE *D = Z, X = LOAD * 287 Y = LOAD *B, STORE *D = Z, X = LOAD *A
286 STORE *D = Z, X = LOAD *A, Y = LOAD * 288 STORE *D = Z, X = LOAD *A, Y = LOAD *B
287 STORE *D = Z, Y = LOAD *B, X = LOAD * 289 STORE *D = Z, Y = LOAD *B, X = LOAD *A
288 290
289 (*) It _must_ be assumed that overlapping mem 291 (*) It _must_ be assumed that overlapping memory accesses may be merged or
290 discarded. This means that for: 292 discarded. This means that for:
291 293
292 X = *A; Y = *(A + 4); 294 X = *A; Y = *(A + 4);
293 295
294 we may get any one of the following seque 296 we may get any one of the following sequences:
295 297
296 X = LOAD *A; Y = LOAD *(A + 4); 298 X = LOAD *A; Y = LOAD *(A + 4);
297 Y = LOAD *(A + 4); X = LOAD *A; 299 Y = LOAD *(A + 4); X = LOAD *A;
298 {X, Y} = LOAD {*A, *(A + 4) }; 300 {X, Y} = LOAD {*A, *(A + 4) };
299 301
300 And for: 302 And for:
301 303
302 *A = X; *(A + 4) = Y; 304 *A = X; *(A + 4) = Y;
303 305
304 we may get any of: 306 we may get any of:
305 307
306 STORE *A = X; STORE *(A + 4) = Y; 308 STORE *A = X; STORE *(A + 4) = Y;
307 STORE *(A + 4) = Y; STORE *A = X; 309 STORE *(A + 4) = Y; STORE *A = X;
308 STORE {*A, *(A + 4) } = {X, Y}; 310 STORE {*A, *(A + 4) } = {X, Y};
309 311
310 And there are anti-guarantees: 312 And there are anti-guarantees:
311 313
312 (*) These guarantees do not apply to bitfield 314 (*) These guarantees do not apply to bitfields, because compilers often
313 generate code to modify these using non-a 315 generate code to modify these using non-atomic read-modify-write
314 sequences. Do not attempt to use bitfiel 316 sequences. Do not attempt to use bitfields to synchronize parallel
315 algorithms. 317 algorithms.
316 318
317 (*) Even in cases where bitfields are protect 319 (*) Even in cases where bitfields are protected by locks, all fields
318 in a given bitfield must be protected by 320 in a given bitfield must be protected by one lock. If two fields
319 in a given bitfield are protected by diff 321 in a given bitfield are protected by different locks, the compiler's
320 non-atomic read-modify-write sequences ca 322 non-atomic read-modify-write sequences can cause an update to one
321 field to corrupt the value of an adjacent 323 field to corrupt the value of an adjacent field.
322 324
323 (*) These guarantees apply only to properly a 325 (*) These guarantees apply only to properly aligned and sized scalar
324 variables. "Properly sized" currently me 326 variables. "Properly sized" currently means variables that are
325 the same size as "char", "short", "int" a 327 the same size as "char", "short", "int" and "long". "Properly
326 aligned" means the natural alignment, thu 328 aligned" means the natural alignment, thus no constraints for
327 "char", two-byte alignment for "short", f 329 "char", two-byte alignment for "short", four-byte alignment for
328 "int", and either four-byte or eight-byte 330 "int", and either four-byte or eight-byte alignment for "long",
329 on 32-bit and 64-bit systems, respectivel 331 on 32-bit and 64-bit systems, respectively. Note that these
330 guarantees were introduced into the C11 s 332 guarantees were introduced into the C11 standard, so beware when
331 using older pre-C11 compilers (for exampl 333 using older pre-C11 compilers (for example, gcc 4.6). The portion
332 of the standard containing this guarantee 334 of the standard containing this guarantee is Section 3.14, which
333 defines "memory location" as follows: 335 defines "memory location" as follows:
334 336
335 memory location 337 memory location
336 either an object of scalar typ 338 either an object of scalar type, or a maximal sequence
337 of adjacent bit-fields all hav 339 of adjacent bit-fields all having nonzero width
338 340
339 NOTE 1: Two threads of executi 341 NOTE 1: Two threads of execution can update and access
340 separate memory locations with 342 separate memory locations without interfering with
341 each other. 343 each other.
342 344
343 NOTE 2: A bit-field and an adj 345 NOTE 2: A bit-field and an adjacent non-bit-field member
344 are in separate memory locatio 346 are in separate memory locations. The same applies
345 to two bit-fields, if one is d 347 to two bit-fields, if one is declared inside a nested
346 structure declaration and the 348 structure declaration and the other is not, or if the two
347 are separated by a zero-length 349 are separated by a zero-length bit-field declaration,
348 or if they are separated by a 350 or if they are separated by a non-bit-field member
349 declaration. It is not safe to 351 declaration. It is not safe to concurrently update two
350 bit-fields in the same structu 352 bit-fields in the same structure if all members declared
351 between them are also bit-fiel 353 between them are also bit-fields, no matter what the
352 sizes of those intervening bit 354 sizes of those intervening bit-fields happen to be.
353 355
354 356
355 ========================= 357 =========================
356 WHAT ARE MEMORY BARRIERS? 358 WHAT ARE MEMORY BARRIERS?
357 ========================= 359 =========================
358 360
359 As can be seen above, independent memory opera 361 As can be seen above, independent memory operations are effectively performed
360 in random order, but this can be a problem for 362 in random order, but this can be a problem for CPU-CPU interaction and for I/O.
361 What is required is some way of intervening to 363 What is required is some way of intervening to instruct the compiler and the
362 CPU to restrict the order. 364 CPU to restrict the order.
363 365
364 Memory barriers are such interventions. They 366 Memory barriers are such interventions. They impose a perceived partial
365 ordering over the memory operations on either 367 ordering over the memory operations on either side of the barrier.
366 368
367 Such enforcement is important because the CPUs 369 Such enforcement is important because the CPUs and other devices in a system
368 can use a variety of tricks to improve perform 370 can use a variety of tricks to improve performance, including reordering,
369 deferral and combination of memory operations; 371 deferral and combination of memory operations; speculative loads; speculative
370 branch prediction and various types of caching 372 branch prediction and various types of caching. Memory barriers are used to
371 override or suppress these tricks, allowing th 373 override or suppress these tricks, allowing the code to sanely control the
372 interaction of multiple CPUs and/or devices. 374 interaction of multiple CPUs and/or devices.
373 375
374 376
375 VARIETIES OF MEMORY BARRIER 377 VARIETIES OF MEMORY BARRIER
376 --------------------------- 378 ---------------------------
377 379
378 Memory barriers come in four basic varieties: 380 Memory barriers come in four basic varieties:
379 381
380 (1) Write (or store) memory barriers. 382 (1) Write (or store) memory barriers.
381 383
382 A write memory barrier gives a guarantee 384 A write memory barrier gives a guarantee that all the STORE operations
383 specified before the barrier will appear 385 specified before the barrier will appear to happen before all the STORE
384 operations specified after the barrier wi 386 operations specified after the barrier with respect to the other
385 components of the system. 387 components of the system.
386 388
387 A write barrier is a partial ordering on 389 A write barrier is a partial ordering on stores only; it is not required
388 to have any effect on loads. 390 to have any effect on loads.
389 391
390 A CPU can be viewed as committing a seque 392 A CPU can be viewed as committing a sequence of store operations to the
391 memory system as time progresses. All st 393 memory system as time progresses. All stores _before_ a write barrier
392 will occur _before_ all the stores after 394 will occur _before_ all the stores after the write barrier.
393 395
394 [!] Note that write barriers should norma !! 396 [!] Note that write barriers should normally be paired with read or data
395 address-dependency barriers; see the "SMP !! 397 dependency barriers; see the "SMP barrier pairing" subsection.
396 398
397 399
398 (2) Address-dependency barriers (historical). !! 400 (2) Data dependency barriers.
399 [!] This section is marked as HISTORICAL: !! 401
400 smp_read_barrier_depends() macro, the sem !! 402 A data dependency barrier is a weaker form of read barrier. In the case
401 implicit in all marked accesses. For mor !! 403 where two loads are performed such that the second depends on the result
402 including how compiler transformations ca !! 404 of the first (eg: the first load retrieves the address to which the second
403 dependencies, see Documentation/RCU/rcu_d !! 405 load will be directed), a data dependency barrier would be required to
404 !! 406 make sure that the target of the second load is updated after the address
405 An address-dependency barrier is a weaker !! 407 obtained by the first load is accessed.
406 case where two loads are performed such t !! 408
407 result of the first (eg: the first load r !! 409 A data dependency barrier is a partial ordering on interdependent loads
408 the second load will be directed), an add !! 410 only; it is not required to have any effect on stores, independent loads
409 be required to make sure that the target !! 411 or overlapping loads.
410 after the address obtained by the first l <<
411 <<
412 An address-dependency barrier is a partia <<
413 loads only; it is not required to have an <<
414 loads or overlapping loads. <<
415 412
416 As mentioned in (1), the other CPUs in th 413 As mentioned in (1), the other CPUs in the system can be viewed as
417 committing sequences of stores to the mem 414 committing sequences of stores to the memory system that the CPU being
418 considered can then perceive. An address !! 415 considered can then perceive. A data dependency barrier issued by the CPU
419 the CPU under consideration guarantees th !! 416 under consideration guarantees that for any load preceding it, if that
420 if that load touches one of a sequence of !! 417 load touches one of a sequence of stores from another CPU, then by the
421 by the time the barrier completes, the ef !! 418 time the barrier completes, the effects of all the stores prior to that
422 that touched by the load will be percepti !! 419 touched by the load will be perceptible to any loads issued after the data
423 the address-dependency barrier. !! 420 dependency barrier.
424 421
425 See the "Examples of memory barrier seque 422 See the "Examples of memory barrier sequences" subsection for diagrams
426 showing the ordering constraints. 423 showing the ordering constraints.
427 424
428 [!] Note that the first load really has t !! 425 [!] Note that the first load really has to have a _data_ dependency and
429 not a control dependency. If the address 426 not a control dependency. If the address for the second load is dependent
430 on the first load, but the dependency is 427 on the first load, but the dependency is through a conditional rather than
431 actually loading the address itself, then 428 actually loading the address itself, then it's a _control_ dependency and
432 a full read barrier or better is required 429 a full read barrier or better is required. See the "Control dependencies"
433 subsection for more information. 430 subsection for more information.
434 431
435 [!] Note that address-dependency barriers !! 432 [!] Note that data dependency barriers should normally be paired with
436 write barriers; see the "SMP barrier pair 433 write barriers; see the "SMP barrier pairing" subsection.
437 434
438 [!] Kernel release v5.9 removed kernel AP <<
439 dependency barriers. Nowadays, APIs for <<
440 variables such as READ_ONCE() and rcu_der <<
441 address-dependency barriers. <<
442 435
443 (3) Read (or load) memory barriers. 436 (3) Read (or load) memory barriers.
444 437
445 A read barrier is an address-dependency b !! 438 A read barrier is a data dependency barrier plus a guarantee that all the
446 the LOAD operations specified before the !! 439 LOAD operations specified before the barrier will appear to happen before
447 before all the LOAD operations specified !! 440 all the LOAD operations specified after the barrier with respect to the
448 the other components of the system. !! 441 other components of the system.
449 442
450 A read barrier is a partial ordering on l 443 A read barrier is a partial ordering on loads only; it is not required to
451 have any effect on stores. 444 have any effect on stores.
452 445
453 Read memory barriers imply address-depend !! 446 Read memory barriers imply data dependency barriers, and so can substitute
454 substitute for them. !! 447 for them.
455 448
456 [!] Note that read barriers should normal 449 [!] Note that read barriers should normally be paired with write barriers;
457 see the "SMP barrier pairing" subsection. 450 see the "SMP barrier pairing" subsection.
458 451
459 452
460 (4) General memory barriers. 453 (4) General memory barriers.
461 454
462 A general memory barrier gives a guarante 455 A general memory barrier gives a guarantee that all the LOAD and STORE
463 operations specified before the barrier w 456 operations specified before the barrier will appear to happen before all
464 the LOAD and STORE operations specified a 457 the LOAD and STORE operations specified after the barrier with respect to
465 the other components of the system. 458 the other components of the system.
466 459
467 A general memory barrier is a partial ord 460 A general memory barrier is a partial ordering over both loads and stores.
468 461
469 General memory barriers imply both read a 462 General memory barriers imply both read and write memory barriers, and so
470 can substitute for either. 463 can substitute for either.
471 464
472 465
473 And a couple of implicit varieties: 466 And a couple of implicit varieties:
474 467
475 (5) ACQUIRE operations. 468 (5) ACQUIRE operations.
476 469
477 This acts as a one-way permeable barrier. 470 This acts as a one-way permeable barrier. It guarantees that all memory
478 operations after the ACQUIRE operation wi 471 operations after the ACQUIRE operation will appear to happen after the
479 ACQUIRE operation with respect to the oth 472 ACQUIRE operation with respect to the other components of the system.
480 ACQUIRE operations include LOCK operation 473 ACQUIRE operations include LOCK operations and both smp_load_acquire()
481 and smp_cond_load_acquire() operations. !! 474 and smp_cond_acquire() operations. The later builds the necessary ACQUIRE
>> 475 semantics from relying on a control dependency and smp_rmb().
482 476
483 Memory operations that occur before an AC 477 Memory operations that occur before an ACQUIRE operation may appear to
484 happen after it completes. 478 happen after it completes.
485 479
486 An ACQUIRE operation should almost always 480 An ACQUIRE operation should almost always be paired with a RELEASE
487 operation. 481 operation.
488 482
489 483
490 (6) RELEASE operations. 484 (6) RELEASE operations.
491 485
492 This also acts as a one-way permeable bar 486 This also acts as a one-way permeable barrier. It guarantees that all
493 memory operations before the RELEASE oper 487 memory operations before the RELEASE operation will appear to happen
494 before the RELEASE operation with respect 488 before the RELEASE operation with respect to the other components of the
495 system. RELEASE operations include UNLOCK 489 system. RELEASE operations include UNLOCK operations and
496 smp_store_release() operations. 490 smp_store_release() operations.
497 491
498 Memory operations that occur after a RELE 492 Memory operations that occur after a RELEASE operation may appear to
499 happen before it completes. 493 happen before it completes.
500 494
501 The use of ACQUIRE and RELEASE operations 495 The use of ACQUIRE and RELEASE operations generally precludes the need
502 for other sorts of memory barrier. In ad !! 496 for other sorts of memory barrier (but note the exceptions mentioned in
503 -not- guaranteed to act as a full memory !! 497 the subsection "MMIO write barrier"). In addition, a RELEASE+ACQUIRE
504 ACQUIRE on a given variable, all memory a !! 498 pair is -not- guaranteed to act as a full memory barrier. However, after
>> 499 an ACQUIRE on a given variable, all memory accesses preceding any prior
505 RELEASE on that same variable are guarant 500 RELEASE on that same variable are guaranteed to be visible. In other
506 words, within a given variable's critical 501 words, within a given variable's critical section, all accesses of all
507 previous critical sections for that varia 502 previous critical sections for that variable are guaranteed to have
508 completed. 503 completed.
509 504
510 This means that ACQUIRE acts as a minimal 505 This means that ACQUIRE acts as a minimal "acquire" operation and
511 RELEASE acts as a minimal "release" opera 506 RELEASE acts as a minimal "release" operation.
512 507
513 A subset of the atomic operations described in 508 A subset of the atomic operations described in atomic_t.txt have ACQUIRE and
514 RELEASE variants in addition to fully-ordered 509 RELEASE variants in addition to fully-ordered and relaxed (no barrier
515 semantics) definitions. For compound atomics 510 semantics) definitions. For compound atomics performing both a load and a
516 store, ACQUIRE semantics apply only to the loa 511 store, ACQUIRE semantics apply only to the load and RELEASE semantics apply
517 only to the store portion of the operation. 512 only to the store portion of the operation.
518 513
519 Memory barriers are only required where there' 514 Memory barriers are only required where there's a possibility of interaction
520 between two CPUs or between a CPU and a device 515 between two CPUs or between a CPU and a device. If it can be guaranteed that
521 there won't be any such interaction in any par 516 there won't be any such interaction in any particular piece of code, then
522 memory barriers are unnecessary in that piece 517 memory barriers are unnecessary in that piece of code.
523 518
524 519
525 Note that these are the _minimum_ guarantees. 520 Note that these are the _minimum_ guarantees. Different architectures may give
526 more substantial guarantees, but they may _not 521 more substantial guarantees, but they may _not_ be relied upon outside of arch
527 specific code. 522 specific code.
528 523
529 524
530 WHAT MAY NOT BE ASSUMED ABOUT MEMORY BARRIERS? 525 WHAT MAY NOT BE ASSUMED ABOUT MEMORY BARRIERS?
531 ---------------------------------------------- 526 ----------------------------------------------
532 527
533 There are certain things that the Linux kernel 528 There are certain things that the Linux kernel memory barriers do not guarantee:
534 529
535 (*) There is no guarantee that any of the mem 530 (*) There is no guarantee that any of the memory accesses specified before a
536 memory barrier will be _complete_ by the 531 memory barrier will be _complete_ by the completion of a memory barrier
537 instruction; the barrier can be considere 532 instruction; the barrier can be considered to draw a line in that CPU's
538 access queue that accesses of the appropr 533 access queue that accesses of the appropriate type may not cross.
539 534
540 (*) There is no guarantee that issuing a memo 535 (*) There is no guarantee that issuing a memory barrier on one CPU will have
541 any direct effect on another CPU or any o 536 any direct effect on another CPU or any other hardware in the system. The
542 indirect effect will be the order in whic 537 indirect effect will be the order in which the second CPU sees the effects
543 of the first CPU's accesses occur, but se 538 of the first CPU's accesses occur, but see the next point:
544 539
545 (*) There is no guarantee that a CPU will see 540 (*) There is no guarantee that a CPU will see the correct order of effects
546 from a second CPU's accesses, even _if_ t 541 from a second CPU's accesses, even _if_ the second CPU uses a memory
547 barrier, unless the first CPU _also_ uses 542 barrier, unless the first CPU _also_ uses a matching memory barrier (see
548 the subsection on "SMP Barrier Pairing"). 543 the subsection on "SMP Barrier Pairing").
549 544
550 (*) There is no guarantee that some interveni 545 (*) There is no guarantee that some intervening piece of off-the-CPU
551 hardware[*] will not reorder the memory a 546 hardware[*] will not reorder the memory accesses. CPU cache coherency
552 mechanisms should propagate the indirect 547 mechanisms should propagate the indirect effects of a memory barrier
553 between CPUs, but might not do so in orde 548 between CPUs, but might not do so in order.
554 549
555 [*] For information on bus mastering D 550 [*] For information on bus mastering DMA and coherency please read:
556 551
557 Documentation/driver-api/pci/pci.r !! 552 Documentation/PCI/pci.txt
558 Documentation/core-api/dma-api-how !! 553 Documentation/DMA-API-HOWTO.txt
559 Documentation/core-api/dma-api.rst !! 554 Documentation/DMA-API.txt
>> 555
>> 556
>> 557 DATA DEPENDENCY BARRIERS (HISTORICAL)
>> 558 -------------------------------------
>> 559
>> 560 As of v4.15 of the Linux kernel, an smp_read_barrier_depends() was
>> 561 added to READ_ONCE(), which means that about the only people who
>> 562 need to pay attention to this section are those working on DEC Alpha
>> 563 architecture-specific code and those working on READ_ONCE() itself.
>> 564 For those who need it, and for those who are interested in the history,
>> 565 here is the story of data-dependency barriers.
560 566
561 !! 567 The usage requirements of data dependency barriers are a little subtle, and
562 ADDRESS-DEPENDENCY BARRIERS (HISTORICAL) <<
563 ---------------------------------------- <<
564 [!] This section is marked as HISTORICAL: it c <<
565 smp_read_barrier_depends() macro, the semantic <<
566 in all marked accesses. For more up-to-date i <<
567 how compiler transformations can sometimes bre <<
568 see Documentation/RCU/rcu_dereference.rst. <<
569 <<
570 As of v4.15 of the Linux kernel, an smp_mb() w <<
571 DEC Alpha, which means that about the only peo <<
572 to this section are those working on DEC Alpha <<
573 and those working on READ_ONCE() itself. For <<
574 those who are interested in the history, here <<
575 address-dependency barriers. <<
576 <<
577 [!] While address dependencies are observed in <<
578 load-to-store relations, address-dependency ba <<
579 for load-to-store situations. <<
580 <<
581 The requirement of address-dependency barriers <<
582 it's not always obvious that they're needed. 568 it's not always obvious that they're needed. To illustrate, consider the
583 following sequence of events: 569 following sequence of events:
584 570
585 CPU 1 CPU 2 571 CPU 1 CPU 2
586 =============== =============== 572 =============== ===============
587 { A == 1, B == 2, C == 3, P == &A, Q = 573 { A == 1, B == 2, C == 3, P == &A, Q == &C }
588 B = 4; 574 B = 4;
589 <write barrier> 575 <write barrier>
590 WRITE_ONCE(P, &B); !! 576 WRITE_ONCE(P, &B)
591 Q = READ_ONCE_OL !! 577 Q = READ_ONCE(P);
592 D = *Q; 578 D = *Q;
593 579
594 [!] READ_ONCE_OLD() corresponds to READ_ONCE() !! 580 There's a clear data dependency here, and it would seem that by the end of the
595 doesn't imply an address-dependency barrier. !! 581 sequence, Q must be either &A or &B, and that:
596 <<
597 There's a clear address dependency here, and i <<
598 the sequence, Q must be either &A or &B, and t <<
599 582
600 (Q == &A) implies (D == 1) 583 (Q == &A) implies (D == 1)
601 (Q == &B) implies (D == 4) 584 (Q == &B) implies (D == 4)
602 585
603 But! CPU 2's perception of P may be updated _ 586 But! CPU 2's perception of P may be updated _before_ its perception of B, thus
604 leading to the following situation: 587 leading to the following situation:
605 588
606 (Q == &B) and (D == 2) ???? 589 (Q == &B) and (D == 2) ????
607 590
608 While this may seem like a failure of coherenc !! 591 Whilst this may seem like a failure of coherency or causality maintenance, it
609 isn't, and this behaviour can be observed on c 592 isn't, and this behaviour can be observed on certain real CPUs (such as the DEC
610 Alpha). 593 Alpha).
611 594
612 To deal with this, READ_ONCE() provides an imp !! 595 To deal with this, a data dependency barrier or better must be inserted
613 since kernel release v4.15: !! 596 between the address load and the data load:
614 597
615 CPU 1 CPU 2 598 CPU 1 CPU 2
616 =============== =============== 599 =============== ===============
617 { A == 1, B == 2, C == 3, P == &A, Q = 600 { A == 1, B == 2, C == 3, P == &A, Q == &C }
618 B = 4; 601 B = 4;
619 <write barrier> 602 <write barrier>
620 WRITE_ONCE(P, &B); 603 WRITE_ONCE(P, &B);
621 Q = READ_ONCE(P) 604 Q = READ_ONCE(P);
622 <implicit addres !! 605 <data dependency barrier>
623 D = *Q; 606 D = *Q;
624 607
625 This enforces the occurrence of one of the two 608 This enforces the occurrence of one of the two implications, and prevents the
626 third possibility from arising. 609 third possibility from arising.
627 610
628 611
629 [!] Note that this extremely counterintuitive 612 [!] Note that this extremely counterintuitive situation arises most easily on
630 machines with split caches, so that, for examp 613 machines with split caches, so that, for example, one cache bank processes
631 even-numbered cache lines and the other bank p 614 even-numbered cache lines and the other bank processes odd-numbered cache
632 lines. The pointer P might be stored in an od 615 lines. The pointer P might be stored in an odd-numbered cache line, and the
633 variable B might be stored in an even-numbered 616 variable B might be stored in an even-numbered cache line. Then, if the
634 even-numbered bank of the reading CPU's cache 617 even-numbered bank of the reading CPU's cache is extremely busy while the
635 odd-numbered bank is idle, one can see the new 618 odd-numbered bank is idle, one can see the new value of the pointer P (&B),
636 but the old value of the variable B (2). 619 but the old value of the variable B (2).
637 620
638 621
639 An address-dependency barrier is not required !! 622 A data-dependency barrier is not required to order dependent writes
640 because the CPUs that the Linux kernel support !! 623 because the CPUs that the Linux kernel supports don't do writes
641 are certain (1) that the write will actually h !! 624 until they are certain (1) that the write will actually happen, (2)
642 the write, and (3) of the value to be written. !! 625 of the location of the write, and (3) of the value to be written.
643 But please carefully read the "CONTROL DEPENDE 626 But please carefully read the "CONTROL DEPENDENCIES" section and the
644 Documentation/RCU/rcu_dereference.rst file: T !! 627 Documentation/RCU/rcu_dereference.txt file: The compiler can and does
645 dependencies in a great many highly creative w !! 628 break dependencies in a great many highly creative ways.
646 629
647 CPU 1 CPU 2 630 CPU 1 CPU 2
648 =============== =============== 631 =============== ===============
649 { A == 1, B == 2, C = 3, P == &A, Q == 632 { A == 1, B == 2, C = 3, P == &A, Q == &C }
650 B = 4; 633 B = 4;
651 <write barrier> 634 <write barrier>
652 WRITE_ONCE(P, &B); 635 WRITE_ONCE(P, &B);
653 Q = READ_ONCE_OL !! 636 Q = READ_ONCE(P);
654 WRITE_ONCE(*Q, 5 637 WRITE_ONCE(*Q, 5);
655 638
656 Therefore, no address-dependency barrier is re !! 639 Therefore, no data-dependency barrier is required to order the read into
657 Q with the store into *Q. In other words, thi 640 Q with the store into *Q. In other words, this outcome is prohibited,
658 even without an implicit address-dependency ba !! 641 even without a data-dependency barrier:
659 642
660 (Q == &B) && (B == 4) 643 (Q == &B) && (B == 4)
661 644
662 Please note that this pattern should be rare. 645 Please note that this pattern should be rare. After all, the whole point
663 of dependency ordering is to -prevent- writes 646 of dependency ordering is to -prevent- writes to the data structure, along
664 with the expensive cache misses associated wit 647 with the expensive cache misses associated with those writes. This pattern
665 can be used to record rare error conditions an 648 can be used to record rare error conditions and the like, and the CPUs'
666 naturally occurring ordering prevents such rec 649 naturally occurring ordering prevents such records from being lost.
667 650
668 651
669 Note well that the ordering provided by an add !! 652 Note well that the ordering provided by a data dependency is local to
670 the CPU containing it. See the section on "Mu 653 the CPU containing it. See the section on "Multicopy atomicity" for
671 more information. 654 more information.
672 655
673 656
674 The address-dependency barrier is very importa !! 657 The data dependency barrier is very important to the RCU system,
675 for example. See rcu_assign_pointer() and rcu 658 for example. See rcu_assign_pointer() and rcu_dereference() in
676 include/linux/rcupdate.h. This permits the cu 659 include/linux/rcupdate.h. This permits the current target of an RCU'd
677 pointer to be replaced with a new modified tar 660 pointer to be replaced with a new modified target, without the replacement
678 target appearing to be incompletely initialise 661 target appearing to be incompletely initialised.
679 662
680 See also the subsection on "Cache Coherency" f 663 See also the subsection on "Cache Coherency" for a more thorough example.
681 664
682 665
683 CONTROL DEPENDENCIES 666 CONTROL DEPENDENCIES
684 -------------------- 667 --------------------
685 668
686 Control dependencies can be a bit tricky becau 669 Control dependencies can be a bit tricky because current compilers do
687 not understand them. The purpose of this sect 670 not understand them. The purpose of this section is to help you prevent
688 the compiler's ignorance from breaking your co 671 the compiler's ignorance from breaking your code.
689 672
690 A load-load control dependency requires a full 673 A load-load control dependency requires a full read memory barrier, not
691 simply an (implicit) address-dependency barrie !! 674 simply a data dependency barrier to make it work correctly. Consider the
692 Consider the following bit of code: !! 675 following bit of code:
693 676
694 q = READ_ONCE(a); 677 q = READ_ONCE(a);
695 <implicit address-dependency barrier> <<
696 if (q) { 678 if (q) {
697 /* BUG: No address dependency! !! 679 <data dependency barrier> /* BUG: No data dependency!!! */
698 p = READ_ONCE(b); 680 p = READ_ONCE(b);
699 } 681 }
700 682
701 This will not have the desired effect because !! 683 This will not have the desired effect because there is no actual data
702 dependency, but rather a control dependency th 684 dependency, but rather a control dependency that the CPU may short-circuit
703 by attempting to predict the outcome in advanc 685 by attempting to predict the outcome in advance, so that other CPUs see
704 the load from b as having happened before the !! 686 the load from b as having happened before the load from a. In such a
705 what's actually required is: !! 687 case what's actually required is:
706 688
707 q = READ_ONCE(a); 689 q = READ_ONCE(a);
708 if (q) { 690 if (q) {
709 <read barrier> 691 <read barrier>
710 p = READ_ONCE(b); 692 p = READ_ONCE(b);
711 } 693 }
712 694
713 However, stores are not speculated. This mean 695 However, stores are not speculated. This means that ordering -is- provided
714 for load-store control dependencies, as in the 696 for load-store control dependencies, as in the following example:
715 697
716 q = READ_ONCE(a); 698 q = READ_ONCE(a);
717 if (q) { 699 if (q) {
718 WRITE_ONCE(b, 1); 700 WRITE_ONCE(b, 1);
719 } 701 }
720 702
721 Control dependencies pair normally with other 703 Control dependencies pair normally with other types of barriers.
722 That said, please note that neither READ_ONCE( 704 That said, please note that neither READ_ONCE() nor WRITE_ONCE()
723 are optional! Without the READ_ONCE(), the com 705 are optional! Without the READ_ONCE(), the compiler might combine the
724 load from 'a' with other loads from 'a'. With 706 load from 'a' with other loads from 'a'. Without the WRITE_ONCE(),
725 the compiler might combine the store to 'b' wi 707 the compiler might combine the store to 'b' with other stores to 'b'.
726 Either can result in highly counterintuitive e 708 Either can result in highly counterintuitive effects on ordering.
727 709
728 Worse yet, if the compiler is able to prove (s 710 Worse yet, if the compiler is able to prove (say) that the value of
729 variable 'a' is always non-zero, it would be w 711 variable 'a' is always non-zero, it would be well within its rights
730 to optimize the original example by eliminatin 712 to optimize the original example by eliminating the "if" statement
731 as follows: 713 as follows:
732 714
733 q = a; 715 q = a;
734 b = 1; /* BUG: Compiler and CPU can b 716 b = 1; /* BUG: Compiler and CPU can both reorder!!! */
735 717
736 So don't leave out the READ_ONCE(). 718 So don't leave out the READ_ONCE().
737 719
738 It is tempting to try to enforce ordering on i 720 It is tempting to try to enforce ordering on identical stores on both
739 branches of the "if" statement as follows: 721 branches of the "if" statement as follows:
740 722
741 q = READ_ONCE(a); 723 q = READ_ONCE(a);
742 if (q) { 724 if (q) {
743 barrier(); 725 barrier();
744 WRITE_ONCE(b, 1); 726 WRITE_ONCE(b, 1);
745 do_something(); 727 do_something();
746 } else { 728 } else {
747 barrier(); 729 barrier();
748 WRITE_ONCE(b, 1); 730 WRITE_ONCE(b, 1);
749 do_something_else(); 731 do_something_else();
750 } 732 }
751 733
752 Unfortunately, current compilers will transfor 734 Unfortunately, current compilers will transform this as follows at high
753 optimization levels: 735 optimization levels:
754 736
755 q = READ_ONCE(a); 737 q = READ_ONCE(a);
756 barrier(); 738 barrier();
757 WRITE_ONCE(b, 1); /* BUG: No ordering 739 WRITE_ONCE(b, 1); /* BUG: No ordering vs. load from a!!! */
758 if (q) { 740 if (q) {
759 /* WRITE_ONCE(b, 1); -- moved 741 /* WRITE_ONCE(b, 1); -- moved up, BUG!!! */
760 do_something(); 742 do_something();
761 } else { 743 } else {
762 /* WRITE_ONCE(b, 1); -- moved 744 /* WRITE_ONCE(b, 1); -- moved up, BUG!!! */
763 do_something_else(); 745 do_something_else();
764 } 746 }
765 747
766 Now there is no conditional between the load f 748 Now there is no conditional between the load from 'a' and the store to
767 'b', which means that the CPU is within its ri 749 'b', which means that the CPU is within its rights to reorder them:
768 The conditional is absolutely required, and mu 750 The conditional is absolutely required, and must be present in the
769 assembly code even after all compiler optimiza 751 assembly code even after all compiler optimizations have been applied.
770 Therefore, if you need ordering in this exampl 752 Therefore, if you need ordering in this example, you need explicit
771 memory barriers, for example, smp_store_releas 753 memory barriers, for example, smp_store_release():
772 754
773 q = READ_ONCE(a); 755 q = READ_ONCE(a);
774 if (q) { 756 if (q) {
775 smp_store_release(&b, 1); 757 smp_store_release(&b, 1);
776 do_something(); 758 do_something();
777 } else { 759 } else {
778 smp_store_release(&b, 1); 760 smp_store_release(&b, 1);
779 do_something_else(); 761 do_something_else();
780 } 762 }
781 763
782 In contrast, without explicit memory barriers, 764 In contrast, without explicit memory barriers, two-legged-if control
783 ordering is guaranteed only when the stores di 765 ordering is guaranteed only when the stores differ, for example:
784 766
785 q = READ_ONCE(a); 767 q = READ_ONCE(a);
786 if (q) { 768 if (q) {
787 WRITE_ONCE(b, 1); 769 WRITE_ONCE(b, 1);
788 do_something(); 770 do_something();
789 } else { 771 } else {
790 WRITE_ONCE(b, 2); 772 WRITE_ONCE(b, 2);
791 do_something_else(); 773 do_something_else();
792 } 774 }
793 775
794 The initial READ_ONCE() is still required to p 776 The initial READ_ONCE() is still required to prevent the compiler from
795 proving the value of 'a'. 777 proving the value of 'a'.
796 778
797 In addition, you need to be careful what you d 779 In addition, you need to be careful what you do with the local variable 'q',
798 otherwise the compiler might be able to guess 780 otherwise the compiler might be able to guess the value and again remove
799 the needed conditional. For example: 781 the needed conditional. For example:
800 782
801 q = READ_ONCE(a); 783 q = READ_ONCE(a);
802 if (q % MAX) { 784 if (q % MAX) {
803 WRITE_ONCE(b, 1); 785 WRITE_ONCE(b, 1);
804 do_something(); 786 do_something();
805 } else { 787 } else {
806 WRITE_ONCE(b, 2); 788 WRITE_ONCE(b, 2);
807 do_something_else(); 789 do_something_else();
808 } 790 }
809 791
810 If MAX is defined to be 1, then the compiler k 792 If MAX is defined to be 1, then the compiler knows that (q % MAX) is
811 equal to zero, in which case the compiler is w 793 equal to zero, in which case the compiler is within its rights to
812 transform the above code into the following: 794 transform the above code into the following:
813 795
814 q = READ_ONCE(a); 796 q = READ_ONCE(a);
815 WRITE_ONCE(b, 2); 797 WRITE_ONCE(b, 2);
816 do_something_else(); 798 do_something_else();
817 799
818 Given this transformation, the CPU is not requ 800 Given this transformation, the CPU is not required to respect the ordering
819 between the load from variable 'a' and the sto 801 between the load from variable 'a' and the store to variable 'b'. It is
820 tempting to add a barrier(), but this does not 802 tempting to add a barrier(), but this does not help. The conditional
821 is gone, and the barrier won't bring it back. 803 is gone, and the barrier won't bring it back. Therefore, if you are
822 relying on this ordering, you should make sure 804 relying on this ordering, you should make sure that MAX is greater than
823 one, perhaps as follows: 805 one, perhaps as follows:
824 806
825 q = READ_ONCE(a); 807 q = READ_ONCE(a);
826 BUILD_BUG_ON(MAX <= 1); /* Order load 808 BUILD_BUG_ON(MAX <= 1); /* Order load from a with store to b. */
827 if (q % MAX) { 809 if (q % MAX) {
828 WRITE_ONCE(b, 1); 810 WRITE_ONCE(b, 1);
829 do_something(); 811 do_something();
830 } else { 812 } else {
831 WRITE_ONCE(b, 2); 813 WRITE_ONCE(b, 2);
832 do_something_else(); 814 do_something_else();
833 } 815 }
834 816
835 Please note once again that the stores to 'b' 817 Please note once again that the stores to 'b' differ. If they were
836 identical, as noted earlier, the compiler coul 818 identical, as noted earlier, the compiler could pull this store outside
837 of the 'if' statement. 819 of the 'if' statement.
838 820
839 You must also be careful not to rely too much 821 You must also be careful not to rely too much on boolean short-circuit
840 evaluation. Consider this example: 822 evaluation. Consider this example:
841 823
842 q = READ_ONCE(a); 824 q = READ_ONCE(a);
843 if (q || 1 > 0) 825 if (q || 1 > 0)
844 WRITE_ONCE(b, 1); 826 WRITE_ONCE(b, 1);
845 827
846 Because the first condition cannot fault and t 828 Because the first condition cannot fault and the second condition is
847 always true, the compiler can transform this e 829 always true, the compiler can transform this example as following,
848 defeating control dependency: 830 defeating control dependency:
849 831
850 q = READ_ONCE(a); 832 q = READ_ONCE(a);
851 WRITE_ONCE(b, 1); 833 WRITE_ONCE(b, 1);
852 834
853 This example underscores the need to ensure th 835 This example underscores the need to ensure that the compiler cannot
854 out-guess your code. More generally, although 836 out-guess your code. More generally, although READ_ONCE() does force
855 the compiler to actually emit code for a given 837 the compiler to actually emit code for a given load, it does not force
856 the compiler to use the results. 838 the compiler to use the results.
857 839
858 In addition, control dependencies apply only t 840 In addition, control dependencies apply only to the then-clause and
859 else-clause of the if-statement in question. 841 else-clause of the if-statement in question. In particular, it does
860 not necessarily apply to code following the if 842 not necessarily apply to code following the if-statement:
861 843
862 q = READ_ONCE(a); 844 q = READ_ONCE(a);
863 if (q) { 845 if (q) {
864 WRITE_ONCE(b, 1); 846 WRITE_ONCE(b, 1);
865 } else { 847 } else {
866 WRITE_ONCE(b, 2); 848 WRITE_ONCE(b, 2);
867 } 849 }
868 WRITE_ONCE(c, 1); /* BUG: No ordering 850 WRITE_ONCE(c, 1); /* BUG: No ordering against the read from 'a'. */
869 851
870 It is tempting to argue that there in fact is 852 It is tempting to argue that there in fact is ordering because the
871 compiler cannot reorder volatile accesses and 853 compiler cannot reorder volatile accesses and also cannot reorder
872 the writes to 'b' with the condition. Unfortu 854 the writes to 'b' with the condition. Unfortunately for this line
873 of reasoning, the compiler might compile the t 855 of reasoning, the compiler might compile the two writes to 'b' as
874 conditional-move instructions, as in this fanc 856 conditional-move instructions, as in this fanciful pseudo-assembly
875 language: 857 language:
876 858
877 ld r1,a 859 ld r1,a
878 cmp r1,$0 860 cmp r1,$0
879 cmov,ne r4,$1 861 cmov,ne r4,$1
880 cmov,eq r4,$2 862 cmov,eq r4,$2
881 st r4,b 863 st r4,b
882 st $1,c 864 st $1,c
883 865
884 A weakly ordered CPU would have no dependency 866 A weakly ordered CPU would have no dependency of any sort between the load
885 from 'a' and the store to 'c'. The control de 867 from 'a' and the store to 'c'. The control dependencies would extend
886 only to the pair of cmov instructions and the 868 only to the pair of cmov instructions and the store depending on them.
887 In short, control dependencies apply only to t 869 In short, control dependencies apply only to the stores in the then-clause
888 and else-clause of the if-statement in questio 870 and else-clause of the if-statement in question (including functions
889 invoked by those two clauses), not to code fol 871 invoked by those two clauses), not to code following that if-statement.
890 872
891 873
892 Note well that the ordering provided by a cont 874 Note well that the ordering provided by a control dependency is local
893 to the CPU containing it. See the section on 875 to the CPU containing it. See the section on "Multicopy atomicity"
894 for more information. 876 for more information.
895 877
896 878
897 In summary: 879 In summary:
898 880
899 (*) Control dependencies can order prior loa 881 (*) Control dependencies can order prior loads against later stores.
900 However, they do -not- guarantee any oth 882 However, they do -not- guarantee any other sort of ordering:
901 Not prior loads against later loads, nor 883 Not prior loads against later loads, nor prior stores against
902 later anything. If you need these other 884 later anything. If you need these other forms of ordering,
903 use smp_rmb(), smp_wmb(), or, in the cas 885 use smp_rmb(), smp_wmb(), or, in the case of prior stores and
904 later loads, smp_mb(). 886 later loads, smp_mb().
905 887
906 (*) If both legs of the "if" statement begin 888 (*) If both legs of the "if" statement begin with identical stores to
907 the same variable, then those stores mus 889 the same variable, then those stores must be ordered, either by
908 preceding both of them with smp_mb() or 890 preceding both of them with smp_mb() or by using smp_store_release()
909 to carry out the stores. Please note th 891 to carry out the stores. Please note that it is -not- sufficient
910 to use barrier() at beginning of each le 892 to use barrier() at beginning of each leg of the "if" statement
911 because, as shown by the example above, 893 because, as shown by the example above, optimizing compilers can
912 destroy the control dependency while res 894 destroy the control dependency while respecting the letter of the
913 barrier() law. 895 barrier() law.
914 896
915 (*) Control dependencies require at least on 897 (*) Control dependencies require at least one run-time conditional
916 between the prior load and the subsequen 898 between the prior load and the subsequent store, and this
917 conditional must involve the prior load. 899 conditional must involve the prior load. If the compiler is able
918 to optimize the conditional away, it wil 900 to optimize the conditional away, it will have also optimized
919 away the ordering. Careful use of READ_ 901 away the ordering. Careful use of READ_ONCE() and WRITE_ONCE()
920 can help to preserve the needed conditio 902 can help to preserve the needed conditional.
921 903
922 (*) Control dependencies require that the co 904 (*) Control dependencies require that the compiler avoid reordering the
923 dependency into nonexistence. Careful u 905 dependency into nonexistence. Careful use of READ_ONCE() or
924 atomic{,64}_read() can help to preserve 906 atomic{,64}_read() can help to preserve your control dependency.
925 Please see the COMPILER BARRIER section 907 Please see the COMPILER BARRIER section for more information.
926 908
927 (*) Control dependencies apply only to the t 909 (*) Control dependencies apply only to the then-clause and else-clause
928 of the if-statement containing the contr 910 of the if-statement containing the control dependency, including
929 any functions that these two clauses cal 911 any functions that these two clauses call. Control dependencies
930 do -not- apply to code following the if- 912 do -not- apply to code following the if-statement containing the
931 control dependency. 913 control dependency.
932 914
933 (*) Control dependencies pair normally with 915 (*) Control dependencies pair normally with other types of barriers.
934 916
935 (*) Control dependencies do -not- provide mu 917 (*) Control dependencies do -not- provide multicopy atomicity. If you
936 need all the CPUs to see a given store a 918 need all the CPUs to see a given store at the same time, use smp_mb().
937 919
938 (*) Compilers do not understand control depe 920 (*) Compilers do not understand control dependencies. It is therefore
939 your job to ensure that they do not brea 921 your job to ensure that they do not break your code.
940 922
941 923
942 SMP BARRIER PAIRING 924 SMP BARRIER PAIRING
943 ------------------- 925 -------------------
944 926
945 When dealing with CPU-CPU interactions, certai 927 When dealing with CPU-CPU interactions, certain types of memory barrier should
946 always be paired. A lack of appropriate pairi 928 always be paired. A lack of appropriate pairing is almost certainly an error.
947 929
948 General barriers pair with each other, though 930 General barriers pair with each other, though they also pair with most
949 other types of barriers, albeit without multic 931 other types of barriers, albeit without multicopy atomicity. An acquire
950 barrier pairs with a release barrier, but both 932 barrier pairs with a release barrier, but both may also pair with other
951 barriers, including of course general barriers 933 barriers, including of course general barriers. A write barrier pairs
952 with an address-dependency barrier, a control !! 934 with a data dependency barrier, a control dependency, an acquire barrier,
953 a release barrier, a read barrier, or a genera 935 a release barrier, a read barrier, or a general barrier. Similarly a
954 read barrier, control dependency, or an addres !! 936 read barrier, control dependency, or a data dependency barrier pairs
955 with a write barrier, an acquire barrier, a re 937 with a write barrier, an acquire barrier, a release barrier, or a
956 general barrier: 938 general barrier:
957 939
958 CPU 1 CPU 2 940 CPU 1 CPU 2
959 =============== =============== 941 =============== ===============
960 WRITE_ONCE(a, 1); 942 WRITE_ONCE(a, 1);
961 <write barrier> 943 <write barrier>
962 WRITE_ONCE(b, 2); x = READ_ONCE(b) 944 WRITE_ONCE(b, 2); x = READ_ONCE(b);
963 <read barrier> 945 <read barrier>
964 y = READ_ONCE(a) 946 y = READ_ONCE(a);
965 947
966 Or: 948 Or:
967 949
968 CPU 1 CPU 2 950 CPU 1 CPU 2
969 =============== ================ 951 =============== ===============================
970 a = 1; 952 a = 1;
971 <write barrier> 953 <write barrier>
972 WRITE_ONCE(b, &a); x = READ_ONCE(b) 954 WRITE_ONCE(b, &a); x = READ_ONCE(b);
973 <implicit addres !! 955 <data dependency barrier>
974 y = *x; 956 y = *x;
975 957
976 Or even: 958 Or even:
977 959
978 CPU 1 CPU 2 960 CPU 1 CPU 2
979 =============== ================ 961 =============== ===============================
980 r1 = READ_ONCE(y); 962 r1 = READ_ONCE(y);
981 <general barrier> 963 <general barrier>
982 WRITE_ONCE(x, 1); if (r2 = READ_ON 964 WRITE_ONCE(x, 1); if (r2 = READ_ONCE(x)) {
983 <implicit con 965 <implicit control dependency>
984 WRITE_ONCE(y, 966 WRITE_ONCE(y, 1);
985 } 967 }
986 968
987 assert(r1 == 0 || r2 == 0); 969 assert(r1 == 0 || r2 == 0);
988 970
989 Basically, the read barrier always has to be t 971 Basically, the read barrier always has to be there, even though it can be of
990 the "weaker" type. 972 the "weaker" type.
991 973
992 [!] Note that the stores before the write barr 974 [!] Note that the stores before the write barrier would normally be expected to
993 match the loads after the read barrier or the !! 975 match the loads after the read barrier or the data dependency barrier, and vice
994 vice versa: !! 976 versa:
995 977
996 CPU 1 CP 978 CPU 1 CPU 2
997 =================== == 979 =================== ===================
998 WRITE_ONCE(a, 1); }---- --->{ v 980 WRITE_ONCE(a, 1); }---- --->{ v = READ_ONCE(c);
999 WRITE_ONCE(b, 2); } \ / { w 981 WRITE_ONCE(b, 2); } \ / { w = READ_ONCE(d);
1000 <write barrier> \ < 982 <write barrier> \ <read barrier>
1001 WRITE_ONCE(c, 3); } / \ { x 983 WRITE_ONCE(c, 3); } / \ { x = READ_ONCE(a);
1002 WRITE_ONCE(d, 4); }---- --->{ y 984 WRITE_ONCE(d, 4); }---- --->{ y = READ_ONCE(b);
1003 985
1004 986
1005 EXAMPLES OF MEMORY BARRIER SEQUENCES 987 EXAMPLES OF MEMORY BARRIER SEQUENCES
1006 ------------------------------------ 988 ------------------------------------
1007 989
1008 Firstly, write barriers act as partial orderi 990 Firstly, write barriers act as partial orderings on store operations.
1009 Consider the following sequence of events: 991 Consider the following sequence of events:
1010 992
1011 CPU 1 993 CPU 1
1012 ======================= 994 =======================
1013 STORE A = 1 995 STORE A = 1
1014 STORE B = 2 996 STORE B = 2
1015 STORE C = 3 997 STORE C = 3
1016 <write barrier> 998 <write barrier>
1017 STORE D = 4 999 STORE D = 4
1018 STORE E = 5 1000 STORE E = 5
1019 1001
1020 This sequence of events is committed to the m 1002 This sequence of events is committed to the memory coherence system in an order
1021 that the rest of the system might perceive as 1003 that the rest of the system might perceive as the unordered set of { STORE A,
1022 STORE B, STORE C } all occurring before the u 1004 STORE B, STORE C } all occurring before the unordered set of { STORE D, STORE E
1023 }: 1005 }:
1024 1006
1025 +-------+ : : 1007 +-------+ : :
1026 | | +------+ 1008 | | +------+
1027 | |------>| C=3 | } /\ 1009 | |------>| C=3 | } /\
1028 | | : +------+ }----- 1010 | | : +------+ }----- \ -----> Events perceptible to
1029 | | : | A=1 | } 1011 | | : | A=1 | } \/ the rest of the system
1030 | | : +------+ } 1012 | | : +------+ }
1031 | CPU 1 | : | B=2 | } 1013 | CPU 1 | : | B=2 | }
1032 | | +------+ } 1014 | | +------+ }
1033 | | wwwwwwwwwwwwwwww } <--- 1015 | | wwwwwwwwwwwwwwww } <--- At this point the write barrier
1034 | | +------+ } 1016 | | +------+ } requires all stores prior to the
1035 | | : | E=5 | } 1017 | | : | E=5 | } barrier to be committed before
1036 | | : +------+ } 1018 | | : +------+ } further stores may take place
1037 | |------>| D=4 | } 1019 | |------>| D=4 | }
1038 | | +------+ 1020 | | +------+
1039 +-------+ : : 1021 +-------+ : :
1040 | 1022 |
1041 | Sequence in whic 1023 | Sequence in which stores are committed to the
1042 | memory system by 1024 | memory system by CPU 1
1043 V 1025 V
1044 1026
1045 1027
1046 Secondly, address-dependency barriers act as !! 1028 Secondly, data dependency barriers act as partial orderings on data-dependent
1047 dependent loads. Consider the following sequ !! 1029 loads. Consider the following sequence of events:
1048 1030
1049 CPU 1 CPU 2 1031 CPU 1 CPU 2
1050 ======================= ============= 1032 ======================= =======================
1051 { B = 7; X = 9; Y = 8; C = &Y 1033 { B = 7; X = 9; Y = 8; C = &Y }
1052 STORE A = 1 1034 STORE A = 1
1053 STORE B = 2 1035 STORE B = 2
1054 <write barrier> 1036 <write barrier>
1055 STORE C = &B LOAD X 1037 STORE C = &B LOAD X
1056 STORE D = 4 LOAD C (gets 1038 STORE D = 4 LOAD C (gets &B)
1057 LOAD *C (read 1039 LOAD *C (reads B)
1058 1040
1059 Without intervention, CPU 2 may perceive the 1041 Without intervention, CPU 2 may perceive the events on CPU 1 in some
1060 effectively random order, despite the write b 1042 effectively random order, despite the write barrier issued by CPU 1:
1061 1043
1062 +-------+ : : 1044 +-------+ : : : :
1063 | | +------+ 1045 | | +------+ +-------+ | Sequence of update
1064 | |------>| B=2 |----- - 1046 | |------>| B=2 |----- --->| Y->8 | | of perception on
1065 | | : +------+ \ 1047 | | : +------+ \ +-------+ | CPU 2
1066 | CPU 1 | : | A=1 | \ - 1048 | CPU 1 | : | A=1 | \ --->| C->&Y | V
1067 | | +------+ | 1049 | | +------+ | +-------+
1068 | | wwwwwwwwwwwwwwww | 1050 | | wwwwwwwwwwwwwwww | : :
1069 | | +------+ | 1051 | | +------+ | : :
1070 | | : | C=&B |--- | 1052 | | : | C=&B |--- | : : +-------+
1071 | | : +------+ \ | 1053 | | : +------+ \ | +-------+ | |
1072 | |------>| D=4 | --------- 1054 | |------>| D=4 | ----------->| C->&B |------>| |
1073 | | +------+ | 1055 | | +------+ | +-------+ | |
1074 +-------+ : : | 1056 +-------+ : : | : : | |
1075 | 1057 | : : | |
1076 | 1058 | : : | CPU 2 |
1077 | 1059 | +-------+ | |
1078 Apparently incorrect ---> | 1060 Apparently incorrect ---> | | B->7 |------>| |
1079 perception of B (!) | 1061 perception of B (!) | +-------+ | |
1080 | 1062 | : : | |
1081 | 1063 | +-------+ | |
1082 The load of X holds ---> \ 1064 The load of X holds ---> \ | X->9 |------>| |
1083 up the maintenance \ 1065 up the maintenance \ +-------+ | |
1084 of coherence of B --- 1066 of coherence of B ----->| B->2 | +-------+
1085 1067 +-------+
1086 1068 : :
1087 1069
1088 1070
1089 In the above example, CPU 2 perceives that B 1071 In the above example, CPU 2 perceives that B is 7, despite the load of *C
1090 (which would be B) coming after the LOAD of C 1072 (which would be B) coming after the LOAD of C.
1091 1073
1092 If, however, an address-dependency barrier we !! 1074 If, however, a data dependency barrier were to be placed between the load of C
1093 of C and the load of *C (ie: B) on CPU 2: !! 1075 and the load of *C (ie: B) on CPU 2:
1094 1076
1095 CPU 1 CPU 2 1077 CPU 1 CPU 2
1096 ======================= ============= 1078 ======================= =======================
1097 { B = 7; X = 9; Y = 8; C = &Y 1079 { B = 7; X = 9; Y = 8; C = &Y }
1098 STORE A = 1 1080 STORE A = 1
1099 STORE B = 2 1081 STORE B = 2
1100 <write barrier> 1082 <write barrier>
1101 STORE C = &B LOAD X 1083 STORE C = &B LOAD X
1102 STORE D = 4 LOAD C (gets 1084 STORE D = 4 LOAD C (gets &B)
1103 <address-depe !! 1085 <data dependency barrier>
1104 LOAD *C (read 1086 LOAD *C (reads B)
1105 1087
1106 then the following will occur: 1088 then the following will occur:
1107 1089
1108 +-------+ : : 1090 +-------+ : : : :
1109 | | +------+ 1091 | | +------+ +-------+
1110 | |------>| B=2 |----- - 1092 | |------>| B=2 |----- --->| Y->8 |
1111 | | : +------+ \ 1093 | | : +------+ \ +-------+
1112 | CPU 1 | : | A=1 | \ - 1094 | CPU 1 | : | A=1 | \ --->| C->&Y |
1113 | | +------+ | 1095 | | +------+ | +-------+
1114 | | wwwwwwwwwwwwwwww | 1096 | | wwwwwwwwwwwwwwww | : :
1115 | | +------+ | 1097 | | +------+ | : :
1116 | | : | C=&B |--- | 1098 | | : | C=&B |--- | : : +-------+
1117 | | : +------+ \ | 1099 | | : +------+ \ | +-------+ | |
1118 | |------>| D=4 | --------- 1100 | |------>| D=4 | ----------->| C->&B |------>| |
1119 | | +------+ | 1101 | | +------+ | +-------+ | |
1120 +-------+ : : | 1102 +-------+ : : | : : | |
1121 | 1103 | : : | |
1122 | 1104 | : : | CPU 2 |
1123 | 1105 | +-------+ | |
1124 | 1106 | | X->9 |------>| |
1125 | 1107 | +-------+ | |
1126 Makes sure all effects ---> \ a !! 1108 Makes sure all effects ---> \ ddddddddddddddddd | |
1127 prior to the store of C \ 1109 prior to the store of C \ +-------+ | |
1128 are perceptible to --- 1110 are perceptible to ----->| B->2 |------>| |
1129 subsequent loads 1111 subsequent loads +-------+ | |
1130 1112 : : +-------+
1131 1113
1132 1114
1133 And thirdly, a read barrier acts as a partial 1115 And thirdly, a read barrier acts as a partial order on loads. Consider the
1134 following sequence of events: 1116 following sequence of events:
1135 1117
1136 CPU 1 CPU 2 1118 CPU 1 CPU 2
1137 ======================= ============= 1119 ======================= =======================
1138 { A = 0, B = 9 } 1120 { A = 0, B = 9 }
1139 STORE A=1 1121 STORE A=1
1140 <write barrier> 1122 <write barrier>
1141 STORE B=2 1123 STORE B=2
1142 LOAD B 1124 LOAD B
1143 LOAD A 1125 LOAD A
1144 1126
1145 Without intervention, CPU 2 may then choose t 1127 Without intervention, CPU 2 may then choose to perceive the events on CPU 1 in
1146 some effectively random order, despite the wr 1128 some effectively random order, despite the write barrier issued by CPU 1:
1147 1129
1148 +-------+ : : 1130 +-------+ : : : :
1149 | | +------+ 1131 | | +------+ +-------+
1150 | |------>| A=1 |------ - 1132 | |------>| A=1 |------ --->| A->0 |
1151 | | +------+ \ 1133 | | +------+ \ +-------+
1152 | CPU 1 | wwwwwwwwwwwwwwww \ - 1134 | CPU 1 | wwwwwwwwwwwwwwww \ --->| B->9 |
1153 | | +------+ | 1135 | | +------+ | +-------+
1154 | |------>| B=2 |--- | 1136 | |------>| B=2 |--- | : :
1155 | | +------+ \ | 1137 | | +------+ \ | : : +-------+
1156 +-------+ : : \ | 1138 +-------+ : : \ | +-------+ | |
1157 -------- 1139 ---------->| B->2 |------>| |
1158 | 1140 | +-------+ | CPU 2 |
1159 | 1141 | | A->0 |------>| |
1160 | 1142 | +-------+ | |
1161 | 1143 | : : +-------+
1162 \ 1144 \ : :
1163 \ 1145 \ +-------+
1164 -- 1146 ---->| A->1 |
1165 1147 +-------+
1166 1148 : :
1167 1149
1168 1150
1169 If, however, a read barrier were to be placed 1151 If, however, a read barrier were to be placed between the load of B and the
1170 load of A on CPU 2: 1152 load of A on CPU 2:
1171 1153
1172 CPU 1 CPU 2 1154 CPU 1 CPU 2
1173 ======================= ============= 1155 ======================= =======================
1174 { A = 0, B = 9 } 1156 { A = 0, B = 9 }
1175 STORE A=1 1157 STORE A=1
1176 <write barrier> 1158 <write barrier>
1177 STORE B=2 1159 STORE B=2
1178 LOAD B 1160 LOAD B
1179 <read barrier 1161 <read barrier>
1180 LOAD A 1162 LOAD A
1181 1163
1182 then the partial ordering imposed by CPU 1 wi 1164 then the partial ordering imposed by CPU 1 will be perceived correctly by CPU
1183 2: 1165 2:
1184 1166
1185 +-------+ : : 1167 +-------+ : : : :
1186 | | +------+ 1168 | | +------+ +-------+
1187 | |------>| A=1 |------ - 1169 | |------>| A=1 |------ --->| A->0 |
1188 | | +------+ \ 1170 | | +------+ \ +-------+
1189 | CPU 1 | wwwwwwwwwwwwwwww \ - 1171 | CPU 1 | wwwwwwwwwwwwwwww \ --->| B->9 |
1190 | | +------+ | 1172 | | +------+ | +-------+
1191 | |------>| B=2 |--- | 1173 | |------>| B=2 |--- | : :
1192 | | +------+ \ | 1174 | | +------+ \ | : : +-------+
1193 +-------+ : : \ | 1175 +-------+ : : \ | +-------+ | |
1194 -------- 1176 ---------->| B->2 |------>| |
1195 | 1177 | +-------+ | CPU 2 |
1196 | 1178 | : : | |
1197 | 1179 | : : | |
1198 At this point the read ----> \ r 1180 At this point the read ----> \ rrrrrrrrrrrrrrrrr | |
1199 barrier causes all effects \ 1181 barrier causes all effects \ +-------+ | |
1200 prior to the storage of B -- 1182 prior to the storage of B ---->| A->1 |------>| |
1201 to be perceptible to CPU 2 1183 to be perceptible to CPU 2 +-------+ | |
1202 1184 : : +-------+
1203 1185
1204 1186
1205 To illustrate this more completely, consider 1187 To illustrate this more completely, consider what could happen if the code
1206 contained a load of A either side of the read 1188 contained a load of A either side of the read barrier:
1207 1189
1208 CPU 1 CPU 2 1190 CPU 1 CPU 2
1209 ======================= ============= 1191 ======================= =======================
1210 { A = 0, B = 9 } 1192 { A = 0, B = 9 }
1211 STORE A=1 1193 STORE A=1
1212 <write barrier> 1194 <write barrier>
1213 STORE B=2 1195 STORE B=2
1214 LOAD B 1196 LOAD B
1215 LOAD A [first 1197 LOAD A [first load of A]
1216 <read barrier 1198 <read barrier>
1217 LOAD A [secon 1199 LOAD A [second load of A]
1218 1200
1219 Even though the two loads of A both occur aft 1201 Even though the two loads of A both occur after the load of B, they may both
1220 come up with different values: 1202 come up with different values:
1221 1203
1222 +-------+ : : 1204 +-------+ : : : :
1223 | | +------+ 1205 | | +------+ +-------+
1224 | |------>| A=1 |------ - 1206 | |------>| A=1 |------ --->| A->0 |
1225 | | +------+ \ 1207 | | +------+ \ +-------+
1226 | CPU 1 | wwwwwwwwwwwwwwww \ - 1208 | CPU 1 | wwwwwwwwwwwwwwww \ --->| B->9 |
1227 | | +------+ | 1209 | | +------+ | +-------+
1228 | |------>| B=2 |--- | 1210 | |------>| B=2 |--- | : :
1229 | | +------+ \ | 1211 | | +------+ \ | : : +-------+
1230 +-------+ : : \ | 1212 +-------+ : : \ | +-------+ | |
1231 -------- 1213 ---------->| B->2 |------>| |
1232 | 1214 | +-------+ | CPU 2 |
1233 | 1215 | : : | |
1234 | 1216 | : : | |
1235 | 1217 | +-------+ | |
1236 | 1218 | | A->0 |------>| 1st |
1237 | 1219 | +-------+ | |
1238 At this point the read ----> \ r 1220 At this point the read ----> \ rrrrrrrrrrrrrrrrr | |
1239 barrier causes all effects \ 1221 barrier causes all effects \ +-------+ | |
1240 prior to the storage of B -- 1222 prior to the storage of B ---->| A->1 |------>| 2nd |
1241 to be perceptible to CPU 2 1223 to be perceptible to CPU 2 +-------+ | |
1242 1224 : : +-------+
1243 1225
1244 1226
1245 But it may be that the update to A from CPU 1 1227 But it may be that the update to A from CPU 1 becomes perceptible to CPU 2
1246 before the read barrier completes anyway: 1228 before the read barrier completes anyway:
1247 1229
1248 +-------+ : : 1230 +-------+ : : : :
1249 | | +------+ 1231 | | +------+ +-------+
1250 | |------>| A=1 |------ - 1232 | |------>| A=1 |------ --->| A->0 |
1251 | | +------+ \ 1233 | | +------+ \ +-------+
1252 | CPU 1 | wwwwwwwwwwwwwwww \ - 1234 | CPU 1 | wwwwwwwwwwwwwwww \ --->| B->9 |
1253 | | +------+ | 1235 | | +------+ | +-------+
1254 | |------>| B=2 |--- | 1236 | |------>| B=2 |--- | : :
1255 | | +------+ \ | 1237 | | +------+ \ | : : +-------+
1256 +-------+ : : \ | 1238 +-------+ : : \ | +-------+ | |
1257 -------- 1239 ---------->| B->2 |------>| |
1258 | 1240 | +-------+ | CPU 2 |
1259 | 1241 | : : | |
1260 \ 1242 \ : : | |
1261 \ 1243 \ +-------+ | |
1262 -- 1244 ---->| A->1 |------>| 1st |
1263 1245 +-------+ | |
1264 r 1246 rrrrrrrrrrrrrrrrr | |
1265 1247 +-------+ | |
1266 1248 | A->1 |------>| 2nd |
1267 1249 +-------+ | |
1268 1250 : : +-------+
1269 1251
1270 1252
1271 The guarantee is that the second load will al 1253 The guarantee is that the second load will always come up with A == 1 if the
1272 load of B came up with B == 2. No such guara 1254 load of B came up with B == 2. No such guarantee exists for the first load of
1273 A; that may come up with either A == 0 or A = 1255 A; that may come up with either A == 0 or A == 1.
1274 1256
1275 1257
1276 READ MEMORY BARRIERS VS LOAD SPECULATION 1258 READ MEMORY BARRIERS VS LOAD SPECULATION
1277 ---------------------------------------- 1259 ----------------------------------------
1278 1260
1279 Many CPUs speculate with loads: that is they 1261 Many CPUs speculate with loads: that is they see that they will need to load an
1280 item from memory, and they find a time where 1262 item from memory, and they find a time where they're not using the bus for any
1281 other loads, and so do the load in advance - 1263 other loads, and so do the load in advance - even though they haven't actually
1282 got to that point in the instruction executio 1264 got to that point in the instruction execution flow yet. This permits the
1283 actual load instruction to potentially comple 1265 actual load instruction to potentially complete immediately because the CPU
1284 already has the value to hand. 1266 already has the value to hand.
1285 1267
1286 It may turn out that the CPU didn't actually 1268 It may turn out that the CPU didn't actually need the value - perhaps because a
1287 branch circumvented the load - in which case 1269 branch circumvented the load - in which case it can discard the value or just
1288 cache it for later use. 1270 cache it for later use.
1289 1271
1290 Consider: 1272 Consider:
1291 1273
1292 CPU 1 CPU 2 1274 CPU 1 CPU 2
1293 ======================= ============= 1275 ======================= =======================
1294 LOAD B 1276 LOAD B
1295 DIVIDE 1277 DIVIDE } Divide instructions generally
1296 DIVIDE 1278 DIVIDE } take a long time to perform
1297 LOAD A 1279 LOAD A
1298 1280
1299 Which might appear as this: 1281 Which might appear as this:
1300 1282
1301 1283 : : +-------+
1302 1284 +-------+ | |
1303 - 1285 --->| B->2 |------>| |
1304 1286 +-------+ | CPU 2 |
1305 1287 : :DIVIDE | |
1306 1288 +-------+ | |
1307 The CPU being busy doing a ---> - 1289 The CPU being busy doing a ---> --->| A->0 |~~~~ | |
1308 division speculates on the 1290 division speculates on the +-------+ ~ | |
1309 LOAD of A 1291 LOAD of A : : ~ | |
1310 1292 : :DIVIDE | |
1311 1293 : : ~ | |
1312 Once the divisions are complete --> 1294 Once the divisions are complete --> : : ~-->| |
1313 the CPU can then perform the 1295 the CPU can then perform the : : | |
1314 LOAD with immediate effect 1296 LOAD with immediate effect : : +-------+
1315 1297
1316 1298
1317 Placing a read barrier or an address-dependen !! 1299 Placing a read barrier or a data dependency barrier just before the second
1318 load: 1300 load:
1319 1301
1320 CPU 1 CPU 2 1302 CPU 1 CPU 2
1321 ======================= ============= 1303 ======================= =======================
1322 LOAD B 1304 LOAD B
1323 DIVIDE 1305 DIVIDE
1324 DIVIDE 1306 DIVIDE
1325 <read barrier 1307 <read barrier>
1326 LOAD A 1308 LOAD A
1327 1309
1328 will force any value speculatively obtained t 1310 will force any value speculatively obtained to be reconsidered to an extent
1329 dependent on the type of barrier used. If th 1311 dependent on the type of barrier used. If there was no change made to the
1330 speculated memory location, then the speculat 1312 speculated memory location, then the speculated value will just be used:
1331 1313
1332 1314 : : +-------+
1333 1315 +-------+ | |
1334 - 1316 --->| B->2 |------>| |
1335 1317 +-------+ | CPU 2 |
1336 1318 : :DIVIDE | |
1337 1319 +-------+ | |
1338 The CPU being busy doing a ---> - 1320 The CPU being busy doing a ---> --->| A->0 |~~~~ | |
1339 division speculates on the 1321 division speculates on the +-------+ ~ | |
1340 LOAD of A 1322 LOAD of A : : ~ | |
1341 1323 : :DIVIDE | |
1342 1324 : : ~ | |
1343 1325 : : ~ | |
1344 r 1326 rrrrrrrrrrrrrrrr~ | |
1345 1327 : : ~ | |
1346 1328 : : ~-->| |
1347 1329 : : | |
1348 1330 : : +-------+
1349 1331
1350 1332
1351 but if there was an update or an invalidation 1333 but if there was an update or an invalidation from another CPU pending, then
1352 the speculation will be cancelled and the val 1334 the speculation will be cancelled and the value reloaded:
1353 1335
1354 1336 : : +-------+
1355 1337 +-------+ | |
1356 - 1338 --->| B->2 |------>| |
1357 1339 +-------+ | CPU 2 |
1358 1340 : :DIVIDE | |
1359 1341 +-------+ | |
1360 The CPU being busy doing a ---> - 1342 The CPU being busy doing a ---> --->| A->0 |~~~~ | |
1361 division speculates on the 1343 division speculates on the +-------+ ~ | |
1362 LOAD of A 1344 LOAD of A : : ~ | |
1363 1345 : :DIVIDE | |
1364 1346 : : ~ | |
1365 1347 : : ~ | |
1366 r 1348 rrrrrrrrrrrrrrrrr | |
1367 1349 +-------+ | |
1368 The speculation is discarded ---> - 1350 The speculation is discarded ---> --->| A->1 |------>| |
1369 and an updated value is 1351 and an updated value is +-------+ | |
1370 retrieved 1352 retrieved : : +-------+
1371 1353
1372 1354
1373 MULTICOPY ATOMICITY 1355 MULTICOPY ATOMICITY
1374 -------------------- 1356 --------------------
1375 1357
1376 Multicopy atomicity is a deeply intuitive not 1358 Multicopy atomicity is a deeply intuitive notion about ordering that is
1377 not always provided by real computer systems, 1359 not always provided by real computer systems, namely that a given store
1378 becomes visible at the same time to all CPUs, 1360 becomes visible at the same time to all CPUs, or, alternatively, that all
1379 CPUs agree on the order in which all stores b 1361 CPUs agree on the order in which all stores become visible. However,
1380 support of full multicopy atomicity would rul 1362 support of full multicopy atomicity would rule out valuable hardware
1381 optimizations, so a weaker form called ``othe 1363 optimizations, so a weaker form called ``other multicopy atomicity''
1382 instead guarantees only that a given store be 1364 instead guarantees only that a given store becomes visible at the same
1383 time to all -other- CPUs. The remainder of t 1365 time to all -other- CPUs. The remainder of this document discusses this
1384 weaker form, but for brevity will call it sim 1366 weaker form, but for brevity will call it simply ``multicopy atomicity''.
1385 1367
1386 The following example demonstrates multicopy 1368 The following example demonstrates multicopy atomicity:
1387 1369
1388 CPU 1 CPU 2 1370 CPU 1 CPU 2 CPU 3
1389 ======================= ============= 1371 ======================= ======================= =======================
1390 { X = 0, Y = 0 } 1372 { X = 0, Y = 0 }
1391 STORE X=1 r1=LOAD X (re 1373 STORE X=1 r1=LOAD X (reads 1) LOAD Y (reads 1)
1392 <general barr 1374 <general barrier> <read barrier>
1393 STORE Y=r1 1375 STORE Y=r1 LOAD X
1394 1376
1395 Suppose that CPU 2's load from X returns 1, w 1377 Suppose that CPU 2's load from X returns 1, which it then stores to Y,
1396 and CPU 3's load from Y returns 1. This indi 1378 and CPU 3's load from Y returns 1. This indicates that CPU 1's store
1397 to X precedes CPU 2's load from X and that CP 1379 to X precedes CPU 2's load from X and that CPU 2's store to Y precedes
1398 CPU 3's load from Y. In addition, the memory 1380 CPU 3's load from Y. In addition, the memory barriers guarantee that
1399 CPU 2 executes its load before its store, and 1381 CPU 2 executes its load before its store, and CPU 3 loads from Y before
1400 it loads from X. The question is then "Can C 1382 it loads from X. The question is then "Can CPU 3's load from X return 0?"
1401 1383
1402 Because CPU 3's load from X in some sense com 1384 Because CPU 3's load from X in some sense comes after CPU 2's load, it
1403 is natural to expect that CPU 3's load from X 1385 is natural to expect that CPU 3's load from X must therefore return 1.
1404 This expectation follows from multicopy atomi 1386 This expectation follows from multicopy atomicity: if a load executing
1405 on CPU B follows a load from the same variabl 1387 on CPU B follows a load from the same variable executing on CPU A (and
1406 CPU A did not originally store the value whic 1388 CPU A did not originally store the value which it read), then on
1407 multicopy-atomic systems, CPU B's load must r 1389 multicopy-atomic systems, CPU B's load must return either the same value
1408 that CPU A's load did or some later value. H 1390 that CPU A's load did or some later value. However, the Linux kernel
1409 does not require systems to be multicopy atom 1391 does not require systems to be multicopy atomic.
1410 1392
1411 The use of a general memory barrier in the ex 1393 The use of a general memory barrier in the example above compensates
1412 for any lack of multicopy atomicity. In the 1394 for any lack of multicopy atomicity. In the example, if CPU 2's load
1413 from X returns 1 and CPU 3's load from Y retu 1395 from X returns 1 and CPU 3's load from Y returns 1, then CPU 3's load
1414 from X must indeed also return 1. 1396 from X must indeed also return 1.
1415 1397
1416 However, dependencies, read barriers, and wri 1398 However, dependencies, read barriers, and write barriers are not always
1417 able to compensate for non-multicopy atomicit 1399 able to compensate for non-multicopy atomicity. For example, suppose
1418 that CPU 2's general barrier is removed from 1400 that CPU 2's general barrier is removed from the above example, leaving
1419 only the data dependency shown below: 1401 only the data dependency shown below:
1420 1402
1421 CPU 1 CPU 2 1403 CPU 1 CPU 2 CPU 3
1422 ======================= ============= 1404 ======================= ======================= =======================
1423 { X = 0, Y = 0 } 1405 { X = 0, Y = 0 }
1424 STORE X=1 r1=LOAD X (re 1406 STORE X=1 r1=LOAD X (reads 1) LOAD Y (reads 1)
1425 <data depende 1407 <data dependency> <read barrier>
1426 STORE Y=r1 1408 STORE Y=r1 LOAD X (reads 0)
1427 1409
1428 This substitution allows non-multicopy atomic 1410 This substitution allows non-multicopy atomicity to run rampant: in
1429 this example, it is perfectly legal for CPU 2 1411 this example, it is perfectly legal for CPU 2's load from X to return 1,
1430 CPU 3's load from Y to return 1, and its load 1412 CPU 3's load from Y to return 1, and its load from X to return 0.
1431 1413
1432 The key point is that although CPU 2's data d 1414 The key point is that although CPU 2's data dependency orders its load
1433 and store, it does not guarantee to order CPU 1415 and store, it does not guarantee to order CPU 1's store. Thus, if this
1434 example runs on a non-multicopy-atomic system 1416 example runs on a non-multicopy-atomic system where CPUs 1 and 2 share a
1435 store buffer or a level of cache, CPU 2 might 1417 store buffer or a level of cache, CPU 2 might have early access to CPU 1's
1436 writes. General barriers are therefore requi 1418 writes. General barriers are therefore required to ensure that all CPUs
1437 agree on the combined order of multiple acces 1419 agree on the combined order of multiple accesses.
1438 1420
1439 General barriers can compensate not only for 1421 General barriers can compensate not only for non-multicopy atomicity,
1440 but can also generate additional ordering tha 1422 but can also generate additional ordering that can ensure that -all-
1441 CPUs will perceive the same order of -all- op 1423 CPUs will perceive the same order of -all- operations. In contrast, a
1442 chain of release-acquire pairs do not provide 1424 chain of release-acquire pairs do not provide this additional ordering,
1443 which means that only those CPUs on the chain 1425 which means that only those CPUs on the chain are guaranteed to agree
1444 on the combined order of the accesses. For e 1426 on the combined order of the accesses. For example, switching to C code
1445 in deference to the ghost of Herman Hollerith 1427 in deference to the ghost of Herman Hollerith:
1446 1428
1447 int u, v, x, y, z; 1429 int u, v, x, y, z;
1448 1430
1449 void cpu0(void) 1431 void cpu0(void)
1450 { 1432 {
1451 r0 = smp_load_acquire(&x); 1433 r0 = smp_load_acquire(&x);
1452 WRITE_ONCE(u, 1); 1434 WRITE_ONCE(u, 1);
1453 smp_store_release(&y, 1); 1435 smp_store_release(&y, 1);
1454 } 1436 }
1455 1437
1456 void cpu1(void) 1438 void cpu1(void)
1457 { 1439 {
1458 r1 = smp_load_acquire(&y); 1440 r1 = smp_load_acquire(&y);
1459 r4 = READ_ONCE(v); 1441 r4 = READ_ONCE(v);
1460 r5 = READ_ONCE(u); 1442 r5 = READ_ONCE(u);
1461 smp_store_release(&z, 1); 1443 smp_store_release(&z, 1);
1462 } 1444 }
1463 1445
1464 void cpu2(void) 1446 void cpu2(void)
1465 { 1447 {
1466 r2 = smp_load_acquire(&z); 1448 r2 = smp_load_acquire(&z);
1467 smp_store_release(&x, 1); 1449 smp_store_release(&x, 1);
1468 } 1450 }
1469 1451
1470 void cpu3(void) 1452 void cpu3(void)
1471 { 1453 {
1472 WRITE_ONCE(v, 1); 1454 WRITE_ONCE(v, 1);
1473 smp_mb(); 1455 smp_mb();
1474 r3 = READ_ONCE(u); 1456 r3 = READ_ONCE(u);
1475 } 1457 }
1476 1458
1477 Because cpu0(), cpu1(), and cpu2() participat 1459 Because cpu0(), cpu1(), and cpu2() participate in a chain of
1478 smp_store_release()/smp_load_acquire() pairs, 1460 smp_store_release()/smp_load_acquire() pairs, the following outcome
1479 is prohibited: 1461 is prohibited:
1480 1462
1481 r0 == 1 && r1 == 1 && r2 == 1 1463 r0 == 1 && r1 == 1 && r2 == 1
1482 1464
1483 Furthermore, because of the release-acquire r 1465 Furthermore, because of the release-acquire relationship between cpu0()
1484 and cpu1(), cpu1() must see cpu0()'s writes, 1466 and cpu1(), cpu1() must see cpu0()'s writes, so that the following
1485 outcome is prohibited: 1467 outcome is prohibited:
1486 1468
1487 r1 == 1 && r5 == 0 1469 r1 == 1 && r5 == 0
1488 1470
1489 However, the ordering provided by a release-a 1471 However, the ordering provided by a release-acquire chain is local
1490 to the CPUs participating in that chain and d 1472 to the CPUs participating in that chain and does not apply to cpu3(),
1491 at least aside from stores. Therefore, the f 1473 at least aside from stores. Therefore, the following outcome is possible:
1492 1474
1493 r0 == 0 && r1 == 1 && r2 == 1 && r3 = 1475 r0 == 0 && r1 == 1 && r2 == 1 && r3 == 0 && r4 == 0
1494 1476
1495 As an aside, the following outcome is also po 1477 As an aside, the following outcome is also possible:
1496 1478
1497 r0 == 0 && r1 == 1 && r2 == 1 && r3 = 1479 r0 == 0 && r1 == 1 && r2 == 1 && r3 == 0 && r4 == 0 && r5 == 1
1498 1480
1499 Although cpu0(), cpu1(), and cpu2() will see 1481 Although cpu0(), cpu1(), and cpu2() will see their respective reads and
1500 writes in order, CPUs not involved in the rel 1482 writes in order, CPUs not involved in the release-acquire chain might
1501 well disagree on the order. This disagreemen 1483 well disagree on the order. This disagreement stems from the fact that
1502 the weak memory-barrier instructions used to 1484 the weak memory-barrier instructions used to implement smp_load_acquire()
1503 and smp_store_release() are not required to o 1485 and smp_store_release() are not required to order prior stores against
1504 subsequent loads in all cases. This means th 1486 subsequent loads in all cases. This means that cpu3() can see cpu0()'s
1505 store to u as happening -after- cpu1()'s load 1487 store to u as happening -after- cpu1()'s load from v, even though
1506 both cpu0() and cpu1() agree that these two o 1488 both cpu0() and cpu1() agree that these two operations occurred in the
1507 intended order. 1489 intended order.
1508 1490
1509 However, please keep in mind that smp_load_ac 1491 However, please keep in mind that smp_load_acquire() is not magic.
1510 In particular, it simply reads from its argum 1492 In particular, it simply reads from its argument with ordering. It does
1511 -not- ensure that any particular value will b 1493 -not- ensure that any particular value will be read. Therefore, the
1512 following outcome is possible: 1494 following outcome is possible:
1513 1495
1514 r0 == 0 && r1 == 0 && r2 == 0 && r5 = 1496 r0 == 0 && r1 == 0 && r2 == 0 && r5 == 0
1515 1497
1516 Note that this outcome can happen even on a m 1498 Note that this outcome can happen even on a mythical sequentially
1517 consistent system where nothing is ever reord 1499 consistent system where nothing is ever reordered.
1518 1500
1519 To reiterate, if your code requires full orde 1501 To reiterate, if your code requires full ordering of all operations,
1520 use general barriers throughout. 1502 use general barriers throughout.
1521 1503
1522 1504
1523 ======================== 1505 ========================
1524 EXPLICIT KERNEL BARRIERS 1506 EXPLICIT KERNEL BARRIERS
1525 ======================== 1507 ========================
1526 1508
1527 The Linux kernel has a variety of different b 1509 The Linux kernel has a variety of different barriers that act at different
1528 levels: 1510 levels:
1529 1511
1530 (*) Compiler barrier. 1512 (*) Compiler barrier.
1531 1513
1532 (*) CPU memory barriers. 1514 (*) CPU memory barriers.
1533 1515
>> 1516 (*) MMIO write barrier.
>> 1517
1534 1518
1535 COMPILER BARRIER 1519 COMPILER BARRIER
1536 ---------------- 1520 ----------------
1537 1521
1538 The Linux kernel has an explicit compiler bar 1522 The Linux kernel has an explicit compiler barrier function that prevents the
1539 compiler from moving the memory accesses eith 1523 compiler from moving the memory accesses either side of it to the other side:
1540 1524
1541 barrier(); 1525 barrier();
1542 1526
1543 This is a general barrier -- there are no rea 1527 This is a general barrier -- there are no read-read or write-write
1544 variants of barrier(). However, READ_ONCE() 1528 variants of barrier(). However, READ_ONCE() and WRITE_ONCE() can be
1545 thought of as weak forms of barrier() that af 1529 thought of as weak forms of barrier() that affect only the specific
1546 accesses flagged by the READ_ONCE() or WRITE_ 1530 accesses flagged by the READ_ONCE() or WRITE_ONCE().
1547 1531
1548 The barrier() function has the following effe 1532 The barrier() function has the following effects:
1549 1533
1550 (*) Prevents the compiler from reordering ac 1534 (*) Prevents the compiler from reordering accesses following the
1551 barrier() to precede any accesses preced 1535 barrier() to precede any accesses preceding the barrier().
1552 One example use for this property is to 1536 One example use for this property is to ease communication between
1553 interrupt-handler code and the code that 1537 interrupt-handler code and the code that was interrupted.
1554 1538
1555 (*) Within a loop, forces the compiler to lo 1539 (*) Within a loop, forces the compiler to load the variables used
1556 in that loop's conditional on each pass 1540 in that loop's conditional on each pass through that loop.
1557 1541
1558 The READ_ONCE() and WRITE_ONCE() functions ca 1542 The READ_ONCE() and WRITE_ONCE() functions can prevent any number of
1559 optimizations that, while perfectly safe in s 1543 optimizations that, while perfectly safe in single-threaded code, can
1560 be fatal in concurrent code. Here are some e 1544 be fatal in concurrent code. Here are some examples of these sorts
1561 of optimizations: 1545 of optimizations:
1562 1546
1563 (*) The compiler is within its rights to reo 1547 (*) The compiler is within its rights to reorder loads and stores
1564 to the same variable, and in some cases, 1548 to the same variable, and in some cases, the CPU is within its
1565 rights to reorder loads to the same vari 1549 rights to reorder loads to the same variable. This means that
1566 the following code: 1550 the following code:
1567 1551
1568 a[0] = x; 1552 a[0] = x;
1569 a[1] = x; 1553 a[1] = x;
1570 1554
1571 Might result in an older value of x stor 1555 Might result in an older value of x stored in a[1] than in a[0].
1572 Prevent both the compiler and the CPU fr 1556 Prevent both the compiler and the CPU from doing this as follows:
1573 1557
1574 a[0] = READ_ONCE(x); 1558 a[0] = READ_ONCE(x);
1575 a[1] = READ_ONCE(x); 1559 a[1] = READ_ONCE(x);
1576 1560
1577 In short, READ_ONCE() and WRITE_ONCE() p 1561 In short, READ_ONCE() and WRITE_ONCE() provide cache coherence for
1578 accesses from multiple CPUs to a single 1562 accesses from multiple CPUs to a single variable.
1579 1563
1580 (*) The compiler is within its rights to mer 1564 (*) The compiler is within its rights to merge successive loads from
1581 the same variable. Such merging can cau 1565 the same variable. Such merging can cause the compiler to "optimize"
1582 the following code: 1566 the following code:
1583 1567
1584 while (tmp = a) 1568 while (tmp = a)
1585 do_something_with(tmp); 1569 do_something_with(tmp);
1586 1570
1587 into the following code, which, although 1571 into the following code, which, although in some sense legitimate
1588 for single-threaded code, is almost cert 1572 for single-threaded code, is almost certainly not what the developer
1589 intended: 1573 intended:
1590 1574
1591 if (tmp = a) 1575 if (tmp = a)
1592 for (;;) 1576 for (;;)
1593 do_something_with(tmp 1577 do_something_with(tmp);
1594 1578
1595 Use READ_ONCE() to prevent the compiler 1579 Use READ_ONCE() to prevent the compiler from doing this to you:
1596 1580
1597 while (tmp = READ_ONCE(a)) 1581 while (tmp = READ_ONCE(a))
1598 do_something_with(tmp); 1582 do_something_with(tmp);
1599 1583
1600 (*) The compiler is within its rights to rel 1584 (*) The compiler is within its rights to reload a variable, for example,
1601 in cases where high register pressure pr 1585 in cases where high register pressure prevents the compiler from
1602 keeping all data of interest in register 1586 keeping all data of interest in registers. The compiler might
1603 therefore optimize the variable 'tmp' ou 1587 therefore optimize the variable 'tmp' out of our previous example:
1604 1588
1605 while (tmp = a) 1589 while (tmp = a)
1606 do_something_with(tmp); 1590 do_something_with(tmp);
1607 1591
1608 This could result in the following code, 1592 This could result in the following code, which is perfectly safe in
1609 single-threaded code, but can be fatal i 1593 single-threaded code, but can be fatal in concurrent code:
1610 1594
1611 while (a) 1595 while (a)
1612 do_something_with(a); 1596 do_something_with(a);
1613 1597
1614 For example, the optimized version of th 1598 For example, the optimized version of this code could result in
1615 passing a zero to do_something_with() in 1599 passing a zero to do_something_with() in the case where the variable
1616 a was modified by some other CPU between 1600 a was modified by some other CPU between the "while" statement and
1617 the call to do_something_with(). 1601 the call to do_something_with().
1618 1602
1619 Again, use READ_ONCE() to prevent the co 1603 Again, use READ_ONCE() to prevent the compiler from doing this:
1620 1604
1621 while (tmp = READ_ONCE(a)) 1605 while (tmp = READ_ONCE(a))
1622 do_something_with(tmp); 1606 do_something_with(tmp);
1623 1607
1624 Note that if the compiler runs short of 1608 Note that if the compiler runs short of registers, it might save
1625 tmp onto the stack. The overhead of thi 1609 tmp onto the stack. The overhead of this saving and later restoring
1626 is why compilers reload variables. Doin 1610 is why compilers reload variables. Doing so is perfectly safe for
1627 single-threaded code, so you need to tel 1611 single-threaded code, so you need to tell the compiler about cases
1628 where it is not safe. 1612 where it is not safe.
1629 1613
1630 (*) The compiler is within its rights to omi 1614 (*) The compiler is within its rights to omit a load entirely if it knows
1631 what the value will be. For example, if 1615 what the value will be. For example, if the compiler can prove that
1632 the value of variable 'a' is always zero 1616 the value of variable 'a' is always zero, it can optimize this code:
1633 1617
1634 while (tmp = a) 1618 while (tmp = a)
1635 do_something_with(tmp); 1619 do_something_with(tmp);
1636 1620
1637 Into this: 1621 Into this:
1638 1622
1639 do { } while (0); 1623 do { } while (0);
1640 1624
1641 This transformation is a win for single- 1625 This transformation is a win for single-threaded code because it
1642 gets rid of a load and a branch. The pr 1626 gets rid of a load and a branch. The problem is that the compiler
1643 will carry out its proof assuming that t 1627 will carry out its proof assuming that the current CPU is the only
1644 one updating variable 'a'. If variable 1628 one updating variable 'a'. If variable 'a' is shared, then the
1645 compiler's proof will be erroneous. Use 1629 compiler's proof will be erroneous. Use READ_ONCE() to tell the
1646 compiler that it doesn't know as much as 1630 compiler that it doesn't know as much as it thinks it does:
1647 1631
1648 while (tmp = READ_ONCE(a)) 1632 while (tmp = READ_ONCE(a))
1649 do_something_with(tmp); 1633 do_something_with(tmp);
1650 1634
1651 But please note that the compiler is als 1635 But please note that the compiler is also closely watching what you
1652 do with the value after the READ_ONCE(). 1636 do with the value after the READ_ONCE(). For example, suppose you
1653 do the following and MAX is a preprocess 1637 do the following and MAX is a preprocessor macro with the value 1:
1654 1638
1655 while ((tmp = READ_ONCE(a)) % MAX) 1639 while ((tmp = READ_ONCE(a)) % MAX)
1656 do_something_with(tmp); 1640 do_something_with(tmp);
1657 1641
1658 Then the compiler knows that the result 1642 Then the compiler knows that the result of the "%" operator applied
1659 to MAX will always be zero, again allowi 1643 to MAX will always be zero, again allowing the compiler to optimize
1660 the code into near-nonexistence. (It wi 1644 the code into near-nonexistence. (It will still load from the
1661 variable 'a'.) 1645 variable 'a'.)
1662 1646
1663 (*) Similarly, the compiler is within its ri 1647 (*) Similarly, the compiler is within its rights to omit a store entirely
1664 if it knows that the variable already ha 1648 if it knows that the variable already has the value being stored.
1665 Again, the compiler assumes that the cur 1649 Again, the compiler assumes that the current CPU is the only one
1666 storing into the variable, which can cau 1650 storing into the variable, which can cause the compiler to do the
1667 wrong thing for shared variables. For e 1651 wrong thing for shared variables. For example, suppose you have
1668 the following: 1652 the following:
1669 1653
1670 a = 0; 1654 a = 0;
1671 ... Code that does not store to varia 1655 ... Code that does not store to variable a ...
1672 a = 0; 1656 a = 0;
1673 1657
1674 The compiler sees that the value of vari 1658 The compiler sees that the value of variable 'a' is already zero, so
1675 it might well omit the second store. Th 1659 it might well omit the second store. This would come as a fatal
1676 surprise if some other CPU might have st 1660 surprise if some other CPU might have stored to variable 'a' in the
1677 meantime. 1661 meantime.
1678 1662
1679 Use WRITE_ONCE() to prevent the compiler 1663 Use WRITE_ONCE() to prevent the compiler from making this sort of
1680 wrong guess: 1664 wrong guess:
1681 1665
1682 WRITE_ONCE(a, 0); 1666 WRITE_ONCE(a, 0);
1683 ... Code that does not store to varia 1667 ... Code that does not store to variable a ...
1684 WRITE_ONCE(a, 0); 1668 WRITE_ONCE(a, 0);
1685 1669
1686 (*) The compiler is within its rights to reo 1670 (*) The compiler is within its rights to reorder memory accesses unless
1687 you tell it not to. For example, consid 1671 you tell it not to. For example, consider the following interaction
1688 between process-level code and an interr 1672 between process-level code and an interrupt handler:
1689 1673
1690 void process_level(void) 1674 void process_level(void)
1691 { 1675 {
1692 msg = get_message(); 1676 msg = get_message();
1693 flag = true; 1677 flag = true;
1694 } 1678 }
1695 1679
1696 void interrupt_handler(void) 1680 void interrupt_handler(void)
1697 { 1681 {
1698 if (flag) 1682 if (flag)
1699 process_message(msg); 1683 process_message(msg);
1700 } 1684 }
1701 1685
1702 There is nothing to prevent the compiler 1686 There is nothing to prevent the compiler from transforming
1703 process_level() to the following, in fac 1687 process_level() to the following, in fact, this might well be a
1704 win for single-threaded code: 1688 win for single-threaded code:
1705 1689
1706 void process_level(void) 1690 void process_level(void)
1707 { 1691 {
1708 flag = true; 1692 flag = true;
1709 msg = get_message(); 1693 msg = get_message();
1710 } 1694 }
1711 1695
1712 If the interrupt occurs between these tw 1696 If the interrupt occurs between these two statement, then
1713 interrupt_handler() might be passed a ga 1697 interrupt_handler() might be passed a garbled msg. Use WRITE_ONCE()
1714 to prevent this as follows: 1698 to prevent this as follows:
1715 1699
1716 void process_level(void) 1700 void process_level(void)
1717 { 1701 {
1718 WRITE_ONCE(msg, get_message() 1702 WRITE_ONCE(msg, get_message());
1719 WRITE_ONCE(flag, true); 1703 WRITE_ONCE(flag, true);
1720 } 1704 }
1721 1705
1722 void interrupt_handler(void) 1706 void interrupt_handler(void)
1723 { 1707 {
1724 if (READ_ONCE(flag)) 1708 if (READ_ONCE(flag))
1725 process_message(READ_ 1709 process_message(READ_ONCE(msg));
1726 } 1710 }
1727 1711
1728 Note that the READ_ONCE() and WRITE_ONCE 1712 Note that the READ_ONCE() and WRITE_ONCE() wrappers in
1729 interrupt_handler() are needed if this i 1713 interrupt_handler() are needed if this interrupt handler can itself
1730 be interrupted by something that also ac 1714 be interrupted by something that also accesses 'flag' and 'msg',
1731 for example, a nested interrupt or an NM 1715 for example, a nested interrupt or an NMI. Otherwise, READ_ONCE()
1732 and WRITE_ONCE() are not needed in inter 1716 and WRITE_ONCE() are not needed in interrupt_handler() other than
1733 for documentation purposes. (Note also 1717 for documentation purposes. (Note also that nested interrupts
1734 do not typically occur in modern Linux k 1718 do not typically occur in modern Linux kernels, in fact, if an
1735 interrupt handler returns with interrupt 1719 interrupt handler returns with interrupts enabled, you will get a
1736 WARN_ONCE() splat.) 1720 WARN_ONCE() splat.)
1737 1721
1738 You should assume that the compiler can 1722 You should assume that the compiler can move READ_ONCE() and
1739 WRITE_ONCE() past code not containing RE 1723 WRITE_ONCE() past code not containing READ_ONCE(), WRITE_ONCE(),
1740 barrier(), or similar primitives. 1724 barrier(), or similar primitives.
1741 1725
1742 This effect could also be achieved using 1726 This effect could also be achieved using barrier(), but READ_ONCE()
1743 and WRITE_ONCE() are more selective: Wi 1727 and WRITE_ONCE() are more selective: With READ_ONCE() and
1744 WRITE_ONCE(), the compiler need only for 1728 WRITE_ONCE(), the compiler need only forget the contents of the
1745 indicated memory locations, while with b 1729 indicated memory locations, while with barrier() the compiler must
1746 discard the value of all memory location !! 1730 discard the value of all memory locations that it has currented
1747 cached in any machine registers. Of cou 1731 cached in any machine registers. Of course, the compiler must also
1748 respect the order in which the READ_ONCE 1732 respect the order in which the READ_ONCE()s and WRITE_ONCE()s occur,
1749 though the CPU of course need not do so. 1733 though the CPU of course need not do so.
1750 1734
1751 (*) The compiler is within its rights to inv 1735 (*) The compiler is within its rights to invent stores to a variable,
1752 as in the following example: 1736 as in the following example:
1753 1737
1754 if (a) 1738 if (a)
1755 b = a; 1739 b = a;
1756 else 1740 else
1757 b = 42; 1741 b = 42;
1758 1742
1759 The compiler might save a branch by opti 1743 The compiler might save a branch by optimizing this as follows:
1760 1744
1761 b = 42; 1745 b = 42;
1762 if (a) 1746 if (a)
1763 b = a; 1747 b = a;
1764 1748
1765 In single-threaded code, this is not onl 1749 In single-threaded code, this is not only safe, but also saves
1766 a branch. Unfortunately, in concurrent 1750 a branch. Unfortunately, in concurrent code, this optimization
1767 could cause some other CPU to see a spur 1751 could cause some other CPU to see a spurious value of 42 -- even
1768 if variable 'a' was never zero -- when l 1752 if variable 'a' was never zero -- when loading variable 'b'.
1769 Use WRITE_ONCE() to prevent this as foll 1753 Use WRITE_ONCE() to prevent this as follows:
1770 1754
1771 if (a) 1755 if (a)
1772 WRITE_ONCE(b, a); 1756 WRITE_ONCE(b, a);
1773 else 1757 else
1774 WRITE_ONCE(b, 42); 1758 WRITE_ONCE(b, 42);
1775 1759
1776 The compiler can also invent loads. The 1760 The compiler can also invent loads. These are usually less
1777 damaging, but they can result in cache-l 1761 damaging, but they can result in cache-line bouncing and thus in
1778 poor performance and scalability. Use R 1762 poor performance and scalability. Use READ_ONCE() to prevent
1779 invented loads. 1763 invented loads.
1780 1764
1781 (*) For aligned memory locations whose size 1765 (*) For aligned memory locations whose size allows them to be accessed
1782 with a single memory-reference instructi 1766 with a single memory-reference instruction, prevents "load tearing"
1783 and "store tearing," in which a single l 1767 and "store tearing," in which a single large access is replaced by
1784 multiple smaller accesses. For example, 1768 multiple smaller accesses. For example, given an architecture having
1785 16-bit store instructions with 7-bit imm 1769 16-bit store instructions with 7-bit immediate fields, the compiler
1786 might be tempted to use two 16-bit store 1770 might be tempted to use two 16-bit store-immediate instructions to
1787 implement the following 32-bit store: 1771 implement the following 32-bit store:
1788 1772
1789 p = 0x00010002; 1773 p = 0x00010002;
1790 1774
1791 Please note that GCC really does use thi 1775 Please note that GCC really does use this sort of optimization,
1792 which is not surprising given that it wo 1776 which is not surprising given that it would likely take more
1793 than two instructions to build the const 1777 than two instructions to build the constant and then store it.
1794 This optimization can therefore be a win 1778 This optimization can therefore be a win in single-threaded code.
1795 In fact, a recent bug (since fixed) caus 1779 In fact, a recent bug (since fixed) caused GCC to incorrectly use
1796 this optimization in a volatile store. 1780 this optimization in a volatile store. In the absence of such bugs,
1797 use of WRITE_ONCE() prevents store teari 1781 use of WRITE_ONCE() prevents store tearing in the following example:
1798 1782
1799 WRITE_ONCE(p, 0x00010002); 1783 WRITE_ONCE(p, 0x00010002);
1800 1784
1801 Use of packed structures can also result 1785 Use of packed structures can also result in load and store tearing,
1802 as in this example: 1786 as in this example:
1803 1787
1804 struct __attribute__((__packed__)) fo 1788 struct __attribute__((__packed__)) foo {
1805 short a; 1789 short a;
1806 int b; 1790 int b;
1807 short c; 1791 short c;
1808 }; 1792 };
1809 struct foo foo1, foo2; 1793 struct foo foo1, foo2;
1810 ... 1794 ...
1811 1795
1812 foo2.a = foo1.a; 1796 foo2.a = foo1.a;
1813 foo2.b = foo1.b; 1797 foo2.b = foo1.b;
1814 foo2.c = foo1.c; 1798 foo2.c = foo1.c;
1815 1799
1816 Because there are no READ_ONCE() or WRIT 1800 Because there are no READ_ONCE() or WRITE_ONCE() wrappers and no
1817 volatile markings, the compiler would be 1801 volatile markings, the compiler would be well within its rights to
1818 implement these three assignment stateme 1802 implement these three assignment statements as a pair of 32-bit
1819 loads followed by a pair of 32-bit store 1803 loads followed by a pair of 32-bit stores. This would result in
1820 load tearing on 'foo1.b' and store teari 1804 load tearing on 'foo1.b' and store tearing on 'foo2.b'. READ_ONCE()
1821 and WRITE_ONCE() again prevent tearing i 1805 and WRITE_ONCE() again prevent tearing in this example:
1822 1806
1823 foo2.a = foo1.a; 1807 foo2.a = foo1.a;
1824 WRITE_ONCE(foo2.b, READ_ONCE(foo1.b)) 1808 WRITE_ONCE(foo2.b, READ_ONCE(foo1.b));
1825 foo2.c = foo1.c; 1809 foo2.c = foo1.c;
1826 1810
1827 All that aside, it is never necessary to use 1811 All that aside, it is never necessary to use READ_ONCE() and
1828 WRITE_ONCE() on a variable that has been mark 1812 WRITE_ONCE() on a variable that has been marked volatile. For example,
1829 because 'jiffies' is marked volatile, it is n 1813 because 'jiffies' is marked volatile, it is never necessary to
1830 say READ_ONCE(jiffies). The reason for this 1814 say READ_ONCE(jiffies). The reason for this is that READ_ONCE() and
1831 WRITE_ONCE() are implemented as volatile cast 1815 WRITE_ONCE() are implemented as volatile casts, which has no effect when
1832 its argument is already marked volatile. 1816 its argument is already marked volatile.
1833 1817
1834 Please note that these compiler barriers have 1818 Please note that these compiler barriers have no direct effect on the CPU,
1835 which may then reorder things however it wish 1819 which may then reorder things however it wishes.
1836 1820
1837 1821
1838 CPU MEMORY BARRIERS 1822 CPU MEMORY BARRIERS
1839 ------------------- 1823 -------------------
1840 1824
1841 The Linux kernel has seven basic CPU memory b !! 1825 The Linux kernel has eight basic CPU memory barriers:
1842 1826
1843 TYPE MANDATORY !! 1827 TYPE MANDATORY SMP CONDITIONAL
1844 ======================= ============= !! 1828 =============== ======================= ===========================
1845 GENERAL mb() !! 1829 GENERAL mb() smp_mb()
1846 WRITE wmb() !! 1830 WRITE wmb() smp_wmb()
1847 READ rmb() !! 1831 READ rmb() smp_rmb()
1848 ADDRESS DEPENDENCY !! 1832 DATA DEPENDENCY READ_ONCE()
1849 1833
1850 1834
1851 All memory barriers except the address-depend !! 1835 All memory barriers except the data dependency barriers imply a compiler
1852 barrier. Address dependencies do not impose !! 1836 barrier. Data dependencies do not impose any additional compiler ordering.
1853 1837
1854 Aside: In the case of address dependencies, t !! 1838 Aside: In the case of data dependencies, the compiler would be expected
1855 to issue the loads in the correct order (eg. 1839 to issue the loads in the correct order (eg. `a[b]` would have to load
1856 the value of b before loading a[b]), however 1840 the value of b before loading a[b]), however there is no guarantee in
1857 the C specification that the compiler may not 1841 the C specification that the compiler may not speculate the value of b
1858 (eg. is equal to 1) and load a[b] before b (e !! 1842 (eg. is equal to 1) and load a before b (eg. tmp = a[1]; if (b != 1)
1859 tmp = a[b]; ). There is also the problem of 1843 tmp = a[b]; ). There is also the problem of a compiler reloading b after
1860 having loaded a[b], thus having a newer copy 1844 having loaded a[b], thus having a newer copy of b than a[b]. A consensus
1861 has not yet been reached about these problems 1845 has not yet been reached about these problems, however the READ_ONCE()
1862 macro is a good place to start looking. 1846 macro is a good place to start looking.
1863 1847
1864 SMP memory barriers are reduced to compiler b 1848 SMP memory barriers are reduced to compiler barriers on uniprocessor compiled
1865 systems because it is assumed that a CPU will 1849 systems because it is assumed that a CPU will appear to be self-consistent,
1866 and will order overlapping accesses correctly 1850 and will order overlapping accesses correctly with respect to itself.
1867 However, see the subsection on "Virtual Machi 1851 However, see the subsection on "Virtual Machine Guests" below.
1868 1852
1869 [!] Note that SMP memory barriers _must_ be u 1853 [!] Note that SMP memory barriers _must_ be used to control the ordering of
1870 references to shared memory on SMP systems, t 1854 references to shared memory on SMP systems, though the use of locking instead
1871 is sufficient. 1855 is sufficient.
1872 1856
1873 Mandatory barriers should not be used to cont 1857 Mandatory barriers should not be used to control SMP effects, since mandatory
1874 barriers impose unnecessary overhead on both 1858 barriers impose unnecessary overhead on both SMP and UP systems. They may,
1875 however, be used to control MMIO effects on a 1859 however, be used to control MMIO effects on accesses through relaxed memory I/O
1876 windows. These barriers are required even on 1860 windows. These barriers are required even on non-SMP systems as they affect
1877 the order in which memory operations appear t 1861 the order in which memory operations appear to a device by prohibiting both the
1878 compiler and the CPU from reordering them. 1862 compiler and the CPU from reordering them.
1879 1863
1880 1864
1881 There are some more advanced barrier function 1865 There are some more advanced barrier functions:
1882 1866
1883 (*) smp_store_mb(var, value) 1867 (*) smp_store_mb(var, value)
1884 1868
1885 This assigns the value to the variable a 1869 This assigns the value to the variable and then inserts a full memory
1886 barrier after it. It isn't guaranteed t 1870 barrier after it. It isn't guaranteed to insert anything more than a
1887 compiler barrier in a UP compilation. 1871 compiler barrier in a UP compilation.
1888 1872
1889 1873
1890 (*) smp_mb__before_atomic(); 1874 (*) smp_mb__before_atomic();
1891 (*) smp_mb__after_atomic(); 1875 (*) smp_mb__after_atomic();
1892 1876
1893 These are for use with atomic RMW functi !! 1877 These are for use with atomic (such as add, subtract, increment and
1894 barriers, but where the code needs a mem !! 1878 decrement) functions that don't return a value, especially when used for
1895 RMW functions that do not imply a memory !! 1879 reference counting. These functions do not imply memory barriers.
1896 subtract, (failed) conditional operation <<
1897 but not atomic_read or atomic_set. A com <<
1898 barrier may be required is when atomic o <<
1899 counting. <<
1900 1880
1901 These are also used for atomic RMW bitop !! 1881 These are also used for atomic bitop functions that do not return a
1902 memory barrier (such as set_bit and clea !! 1882 value (such as set_bit and clear_bit).
1903 1883
1904 As an example, consider a piece of code 1884 As an example, consider a piece of code that marks an object as being dead
1905 and then decrements the object's referen 1885 and then decrements the object's reference count:
1906 1886
1907 obj->dead = 1; 1887 obj->dead = 1;
1908 smp_mb__before_atomic(); 1888 smp_mb__before_atomic();
1909 atomic_dec(&obj->ref_count); 1889 atomic_dec(&obj->ref_count);
1910 1890
1911 This makes sure that the death mark on t 1891 This makes sure that the death mark on the object is perceived to be set
1912 *before* the reference counter is decrem 1892 *before* the reference counter is decremented.
1913 1893
1914 See Documentation/atomic_{t,bitops}.txt 1894 See Documentation/atomic_{t,bitops}.txt for more information.
1915 1895
1916 1896
1917 (*) dma_wmb(); 1897 (*) dma_wmb();
1918 (*) dma_rmb(); 1898 (*) dma_rmb();
1919 (*) dma_mb(); <<
1920 1899
1921 These are for use with consistent memory 1900 These are for use with consistent memory to guarantee the ordering
1922 of writes or reads of shared memory acce 1901 of writes or reads of shared memory accessible to both the CPU and a
1923 DMA capable device. See Documentation/co !! 1902 DMA capable device.
1924 information about consistent memory. <<
1925 1903
1926 For example, consider a device driver th 1904 For example, consider a device driver that shares memory with a device
1927 and uses a descriptor status value to in 1905 and uses a descriptor status value to indicate if the descriptor belongs
1928 to the device or the CPU, and a doorbell 1906 to the device or the CPU, and a doorbell to notify it when new
1929 descriptors are available: 1907 descriptors are available:
1930 1908
1931 if (desc->status != DEVICE_OWN) { 1909 if (desc->status != DEVICE_OWN) {
1932 /* do not read data until we 1910 /* do not read data until we own descriptor */
1933 dma_rmb(); 1911 dma_rmb();
1934 1912
1935 /* read/modify data */ 1913 /* read/modify data */
1936 read_data = desc->data; 1914 read_data = desc->data;
1937 desc->data = write_data; 1915 desc->data = write_data;
1938 1916
1939 /* flush modifications before 1917 /* flush modifications before status update */
1940 dma_wmb(); 1918 dma_wmb();
1941 1919
1942 /* assign ownership */ 1920 /* assign ownership */
1943 desc->status = DEVICE_OWN; 1921 desc->status = DEVICE_OWN;
1944 1922
1945 /* Make descriptor status vis !! 1923 /* notify device of new descriptors */
1946 * notify device of new descr <<
1947 */ <<
1948 writel(DESC_NOTIFY, doorbell) 1924 writel(DESC_NOTIFY, doorbell);
1949 } 1925 }
1950 1926
1951 The dma_rmb() allows us to guarantee tha !! 1927 The dma_rmb() allows us guarantee the device has released ownership
1952 before we read the data from the descrip 1928 before we read the data from the descriptor, and the dma_wmb() allows
1953 us to guarantee the data is written to t 1929 us to guarantee the data is written to the descriptor before the device
1954 can see it now has ownership. The dma_m !! 1930 can see it now has ownership. Note that, when using writel(), a prior
1955 a dma_wmb(). !! 1931 wmb() is not needed to guarantee that the cache coherent memory writes
>> 1932 have completed before writing to the MMIO region. The cheaper
>> 1933 writel_relaxed() does not provide this guarantee and must not be used
>> 1934 here.
>> 1935
>> 1936 See the subsection "Kernel I/O barrier effects" for more information on
>> 1937 relaxed I/O accessors and the Documentation/DMA-API.txt file for more
>> 1938 information on consistent memory.
>> 1939
>> 1940
>> 1941 MMIO WRITE BARRIER
>> 1942 ------------------
>> 1943
>> 1944 The Linux kernel also has a special barrier for use with memory-mapped I/O
>> 1945 writes:
>> 1946
>> 1947 mmiowb();
>> 1948
>> 1949 This is a variation on the mandatory write barrier that causes writes to weakly
>> 1950 ordered I/O regions to be partially ordered. Its effects may go beyond the
>> 1951 CPU->Hardware interface and actually affect the hardware at some level.
>> 1952
>> 1953 See the subsection "Acquires vs I/O accesses" for more information.
1956 1954
1957 Note that the dma_*() barriers do not pr <<
1958 accesses to MMIO regions. See the later <<
1959 subsection for more information about I/ <<
1960 <<
1961 (*) pmem_wmb(); <<
1962 <<
1963 This is for use with persistent memory t <<
1964 modifications are written to persistent <<
1965 durability domain. <<
1966 <<
1967 For example, after a non-temporal write <<
1968 to ensure that stores have reached a pla <<
1969 that stores have updated persistent stor <<
1970 data transfer caused by subsequent instr <<
1971 in addition to the ordering done by wmb( <<
1972 <<
1973 For load from persistent memory, existin <<
1974 to ensure read ordering. <<
1975 <<
1976 (*) io_stop_wc(); <<
1977 <<
1978 For memory accesses with write-combining <<
1979 by ioremap_wc()), the CPU may wait for p <<
1980 subsequent ones. io_stop_wc() can be use <<
1981 write-combining memory accesses before t <<
1982 such wait has performance implications. <<
1983 1955
1984 =============================== 1956 ===============================
1985 IMPLICIT KERNEL MEMORY BARRIERS 1957 IMPLICIT KERNEL MEMORY BARRIERS
1986 =============================== 1958 ===============================
1987 1959
1988 Some of the other functions in the linux kern 1960 Some of the other functions in the linux kernel imply memory barriers, amongst
1989 which are locking and scheduling functions. 1961 which are locking and scheduling functions.
1990 1962
1991 This specification is a _minimum_ guarantee; 1963 This specification is a _minimum_ guarantee; any particular architecture may
1992 provide more substantial guarantees, but thes 1964 provide more substantial guarantees, but these may not be relied upon outside
1993 of arch specific code. 1965 of arch specific code.
1994 1966
1995 1967
1996 LOCK ACQUISITION FUNCTIONS 1968 LOCK ACQUISITION FUNCTIONS
1997 -------------------------- 1969 --------------------------
1998 1970
1999 The Linux kernel has a number of locking cons 1971 The Linux kernel has a number of locking constructs:
2000 1972
2001 (*) spin locks 1973 (*) spin locks
2002 (*) R/W spin locks 1974 (*) R/W spin locks
2003 (*) mutexes 1975 (*) mutexes
2004 (*) semaphores 1976 (*) semaphores
2005 (*) R/W semaphores 1977 (*) R/W semaphores
2006 1978
2007 In all cases there are variants on "ACQUIRE" 1979 In all cases there are variants on "ACQUIRE" operations and "RELEASE" operations
2008 for each construct. These operations all imp 1980 for each construct. These operations all imply certain barriers:
2009 1981
2010 (1) ACQUIRE operation implication: 1982 (1) ACQUIRE operation implication:
2011 1983
2012 Memory operations issued after the ACQUI 1984 Memory operations issued after the ACQUIRE will be completed after the
2013 ACQUIRE operation has completed. 1985 ACQUIRE operation has completed.
2014 1986
2015 Memory operations issued before the ACQU 1987 Memory operations issued before the ACQUIRE may be completed after
2016 the ACQUIRE operation has completed. 1988 the ACQUIRE operation has completed.
2017 1989
2018 (2) RELEASE operation implication: 1990 (2) RELEASE operation implication:
2019 1991
2020 Memory operations issued before the RELE 1992 Memory operations issued before the RELEASE will be completed before the
2021 RELEASE operation has completed. 1993 RELEASE operation has completed.
2022 1994
2023 Memory operations issued after the RELEA 1995 Memory operations issued after the RELEASE may be completed before the
2024 RELEASE operation has completed. 1996 RELEASE operation has completed.
2025 1997
2026 (3) ACQUIRE vs ACQUIRE implication: 1998 (3) ACQUIRE vs ACQUIRE implication:
2027 1999
2028 All ACQUIRE operations issued before ano 2000 All ACQUIRE operations issued before another ACQUIRE operation will be
2029 completed before that ACQUIRE operation. 2001 completed before that ACQUIRE operation.
2030 2002
2031 (4) ACQUIRE vs RELEASE implication: 2003 (4) ACQUIRE vs RELEASE implication:
2032 2004
2033 All ACQUIRE operations issued before a R 2005 All ACQUIRE operations issued before a RELEASE operation will be
2034 completed before the RELEASE operation. 2006 completed before the RELEASE operation.
2035 2007
2036 (5) Failed conditional ACQUIRE implication: 2008 (5) Failed conditional ACQUIRE implication:
2037 2009
2038 Certain locking variants of the ACQUIRE 2010 Certain locking variants of the ACQUIRE operation may fail, either due to
2039 being unable to get the lock immediately 2011 being unable to get the lock immediately, or due to receiving an unblocked
2040 signal while asleep waiting for the lock !! 2012 signal whilst asleep waiting for the lock to become available. Failed
2041 locks do not imply any sort of barrier. 2013 locks do not imply any sort of barrier.
2042 2014
2043 [!] Note: one of the consequences of lock ACQ 2015 [!] Note: one of the consequences of lock ACQUIREs and RELEASEs being only
2044 one-way barriers is that the effects of instr 2016 one-way barriers is that the effects of instructions outside of a critical
2045 section may seep into the inside of the criti 2017 section may seep into the inside of the critical section.
2046 2018
2047 An ACQUIRE followed by a RELEASE may not be a 2019 An ACQUIRE followed by a RELEASE may not be assumed to be full memory barrier
2048 because it is possible for an access precedin 2020 because it is possible for an access preceding the ACQUIRE to happen after the
2049 ACQUIRE, and an access following the RELEASE 2021 ACQUIRE, and an access following the RELEASE to happen before the RELEASE, and
2050 the two accesses can themselves then cross: 2022 the two accesses can themselves then cross:
2051 2023
2052 *A = a; 2024 *A = a;
2053 ACQUIRE M 2025 ACQUIRE M
2054 RELEASE M 2026 RELEASE M
2055 *B = b; 2027 *B = b;
2056 2028
2057 may occur as: 2029 may occur as:
2058 2030
2059 ACQUIRE M, STORE *B, STORE *A, RELEAS 2031 ACQUIRE M, STORE *B, STORE *A, RELEASE M
2060 2032
2061 When the ACQUIRE and RELEASE are a lock acqui 2033 When the ACQUIRE and RELEASE are a lock acquisition and release,
2062 respectively, this same reordering can occur 2034 respectively, this same reordering can occur if the lock's ACQUIRE and
2063 RELEASE are to the same lock variable, but on 2035 RELEASE are to the same lock variable, but only from the perspective of
2064 another CPU not holding that lock. In short, 2036 another CPU not holding that lock. In short, a ACQUIRE followed by an
2065 RELEASE may -not- be assumed to be a full mem 2037 RELEASE may -not- be assumed to be a full memory barrier.
2066 2038
2067 Similarly, the reverse case of a RELEASE foll 2039 Similarly, the reverse case of a RELEASE followed by an ACQUIRE does
2068 not imply a full memory barrier. Therefore, 2040 not imply a full memory barrier. Therefore, the CPU's execution of the
2069 critical sections corresponding to the RELEAS 2041 critical sections corresponding to the RELEASE and the ACQUIRE can cross,
2070 so that: 2042 so that:
2071 2043
2072 *A = a; 2044 *A = a;
2073 RELEASE M 2045 RELEASE M
2074 ACQUIRE N 2046 ACQUIRE N
2075 *B = b; 2047 *B = b;
2076 2048
2077 could occur as: 2049 could occur as:
2078 2050
2079 ACQUIRE N, STORE *B, STORE *A, RELEAS 2051 ACQUIRE N, STORE *B, STORE *A, RELEASE M
2080 2052
2081 It might appear that this reordering could in 2053 It might appear that this reordering could introduce a deadlock.
2082 However, this cannot happen because if such a 2054 However, this cannot happen because if such a deadlock threatened,
2083 the RELEASE would simply complete, thereby av 2055 the RELEASE would simply complete, thereby avoiding the deadlock.
2084 2056
2085 Why does this work? 2057 Why does this work?
2086 2058
2087 One key point is that we are only tal 2059 One key point is that we are only talking about the CPU doing
2088 the reordering, not the compiler. If 2060 the reordering, not the compiler. If the compiler (or, for
2089 that matter, the developer) switched 2061 that matter, the developer) switched the operations, deadlock
2090 -could- occur. 2062 -could- occur.
2091 2063
2092 But suppose the CPU reordered the ope 2064 But suppose the CPU reordered the operations. In this case,
2093 the unlock precedes the lock in the a 2065 the unlock precedes the lock in the assembly code. The CPU
2094 simply elected to try executing the l 2066 simply elected to try executing the later lock operation first.
2095 If there is a deadlock, this lock ope 2067 If there is a deadlock, this lock operation will simply spin (or
2096 try to sleep, but more on that later) 2068 try to sleep, but more on that later). The CPU will eventually
2097 execute the unlock operation (which p 2069 execute the unlock operation (which preceded the lock operation
2098 in the assembly code), which will unr 2070 in the assembly code), which will unravel the potential deadlock,
2099 allowing the lock operation to succee 2071 allowing the lock operation to succeed.
2100 2072
2101 But what if the lock is a sleeplock? 2073 But what if the lock is a sleeplock? In that case, the code will
2102 try to enter the scheduler, where it 2074 try to enter the scheduler, where it will eventually encounter
2103 a memory barrier, which will force th 2075 a memory barrier, which will force the earlier unlock operation
2104 to complete, again unraveling the dea 2076 to complete, again unraveling the deadlock. There might be
2105 a sleep-unlock race, but the locking 2077 a sleep-unlock race, but the locking primitive needs to resolve
2106 such races properly in any case. 2078 such races properly in any case.
2107 2079
2108 Locks and semaphores may not provide any guar 2080 Locks and semaphores may not provide any guarantee of ordering on UP compiled
2109 systems, and so cannot be counted on in such 2081 systems, and so cannot be counted on in such a situation to actually achieve
2110 anything at all - especially with respect to 2082 anything at all - especially with respect to I/O accesses - unless combined
2111 with interrupt disabling operations. 2083 with interrupt disabling operations.
2112 2084
2113 See also the section on "Inter-CPU acquiring 2085 See also the section on "Inter-CPU acquiring barrier effects".
2114 2086
2115 2087
2116 As an example, consider the following: 2088 As an example, consider the following:
2117 2089
2118 *A = a; 2090 *A = a;
2119 *B = b; 2091 *B = b;
2120 ACQUIRE 2092 ACQUIRE
2121 *C = c; 2093 *C = c;
2122 *D = d; 2094 *D = d;
2123 RELEASE 2095 RELEASE
2124 *E = e; 2096 *E = e;
2125 *F = f; 2097 *F = f;
2126 2098
2127 The following sequence of events is acceptabl 2099 The following sequence of events is acceptable:
2128 2100
2129 ACQUIRE, {*F,*A}, *E, {*C,*D}, *B, RE 2101 ACQUIRE, {*F,*A}, *E, {*C,*D}, *B, RELEASE
2130 2102
2131 [+] Note that {*F,*A} indicates a com 2103 [+] Note that {*F,*A} indicates a combined access.
2132 2104
2133 But none of the following are: 2105 But none of the following are:
2134 2106
2135 {*F,*A}, *B, ACQUIRE, *C, *D, 2107 {*F,*A}, *B, ACQUIRE, *C, *D, RELEASE, *E
2136 *A, *B, *C, ACQUIRE, *D, 2108 *A, *B, *C, ACQUIRE, *D, RELEASE, *E, *F
2137 *A, *B, ACQUIRE, *C, 2109 *A, *B, ACQUIRE, *C, RELEASE, *D, *E, *F
2138 *B, ACQUIRE, *C, *D, 2110 *B, ACQUIRE, *C, *D, RELEASE, {*F,*A}, *E
2139 2111
2140 2112
2141 2113
2142 INTERRUPT DISABLING FUNCTIONS 2114 INTERRUPT DISABLING FUNCTIONS
2143 ----------------------------- 2115 -----------------------------
2144 2116
2145 Functions that disable interrupts (ACQUIRE eq 2117 Functions that disable interrupts (ACQUIRE equivalent) and enable interrupts
2146 (RELEASE equivalent) will act as compiler bar 2118 (RELEASE equivalent) will act as compiler barriers only. So if memory or I/O
2147 barriers are required in such a situation, th 2119 barriers are required in such a situation, they must be provided from some
2148 other means. 2120 other means.
2149 2121
2150 2122
2151 SLEEP AND WAKE-UP FUNCTIONS 2123 SLEEP AND WAKE-UP FUNCTIONS
2152 --------------------------- 2124 ---------------------------
2153 2125
2154 Sleeping and waking on an event flagged in gl 2126 Sleeping and waking on an event flagged in global data can be viewed as an
2155 interaction between two pieces of data: the t 2127 interaction between two pieces of data: the task state of the task waiting for
2156 the event and the global data used to indicat 2128 the event and the global data used to indicate the event. To make sure that
2157 these appear to happen in the right order, th 2129 these appear to happen in the right order, the primitives to begin the process
2158 of going to sleep, and the primitives to init 2130 of going to sleep, and the primitives to initiate a wake up imply certain
2159 barriers. 2131 barriers.
2160 2132
2161 Firstly, the sleeper normally follows somethi 2133 Firstly, the sleeper normally follows something like this sequence of events:
2162 2134
2163 for (;;) { 2135 for (;;) {
2164 set_current_state(TASK_UNINTE 2136 set_current_state(TASK_UNINTERRUPTIBLE);
2165 if (event_indicated) 2137 if (event_indicated)
2166 break; 2138 break;
2167 schedule(); 2139 schedule();
2168 } 2140 }
2169 2141
2170 A general memory barrier is interpolated auto 2142 A general memory barrier is interpolated automatically by set_current_state()
2171 after it has altered the task state: 2143 after it has altered the task state:
2172 2144
2173 CPU 1 2145 CPU 1
2174 =============================== 2146 ===============================
2175 set_current_state(); 2147 set_current_state();
2176 smp_store_mb(); 2148 smp_store_mb();
2177 STORE current->state 2149 STORE current->state
2178 <general barrier> 2150 <general barrier>
2179 LOAD event_indicated 2151 LOAD event_indicated
2180 2152
2181 set_current_state() may be wrapped by: 2153 set_current_state() may be wrapped by:
2182 2154
2183 prepare_to_wait(); 2155 prepare_to_wait();
2184 prepare_to_wait_exclusive(); 2156 prepare_to_wait_exclusive();
2185 2157
2186 which therefore also imply a general memory b 2158 which therefore also imply a general memory barrier after setting the state.
2187 The whole sequence above is available in vari 2159 The whole sequence above is available in various canned forms, all of which
2188 interpolate the memory barrier in the right p 2160 interpolate the memory barrier in the right place:
2189 2161
2190 wait_event(); 2162 wait_event();
2191 wait_event_interruptible(); 2163 wait_event_interruptible();
2192 wait_event_interruptible_exclusive(); 2164 wait_event_interruptible_exclusive();
2193 wait_event_interruptible_timeout(); 2165 wait_event_interruptible_timeout();
2194 wait_event_killable(); 2166 wait_event_killable();
2195 wait_event_timeout(); 2167 wait_event_timeout();
2196 wait_on_bit(); 2168 wait_on_bit();
2197 wait_on_bit_lock(); 2169 wait_on_bit_lock();
2198 2170
2199 2171
2200 Secondly, code that performs a wake up normal 2172 Secondly, code that performs a wake up normally follows something like this:
2201 2173
2202 event_indicated = 1; 2174 event_indicated = 1;
2203 wake_up(&event_wait_queue); 2175 wake_up(&event_wait_queue);
2204 2176
2205 or: 2177 or:
2206 2178
2207 event_indicated = 1; 2179 event_indicated = 1;
2208 wake_up_process(event_daemon); 2180 wake_up_process(event_daemon);
2209 2181
2210 A general memory barrier is executed by wake_ 2182 A general memory barrier is executed by wake_up() if it wakes something up.
2211 If it doesn't wake anything up then a memory 2183 If it doesn't wake anything up then a memory barrier may or may not be
2212 executed; you must not rely on it. The barri 2184 executed; you must not rely on it. The barrier occurs before the task state
2213 is accessed, in particular, it sits between t 2185 is accessed, in particular, it sits between the STORE to indicate the event
2214 and the STORE to set TASK_RUNNING: 2186 and the STORE to set TASK_RUNNING:
2215 2187
2216 CPU 1 (Sleeper) CPU 2 2188 CPU 1 (Sleeper) CPU 2 (Waker)
2217 =============================== ===== 2189 =============================== ===============================
2218 set_current_state(); STORE 2190 set_current_state(); STORE event_indicated
2219 smp_store_mb(); wake_ 2191 smp_store_mb(); wake_up();
2220 STORE current->state ... 2192 STORE current->state ...
2221 <general barrier> <ge 2193 <general barrier> <general barrier>
2222 LOAD event_indicated if 2194 LOAD event_indicated if ((LOAD task->state) & TASK_NORMAL)
2223 S 2195 STORE task->state
2224 2196
2225 where "task" is the thread being woken up and 2197 where "task" is the thread being woken up and it equals CPU 1's "current".
2226 2198
2227 To repeat, a general memory barrier is guaran 2199 To repeat, a general memory barrier is guaranteed to be executed by wake_up()
2228 if something is actually awakened, but otherw 2200 if something is actually awakened, but otherwise there is no such guarantee.
2229 To see this, consider the following sequence 2201 To see this, consider the following sequence of events, where X and Y are both
2230 initially zero: 2202 initially zero:
2231 2203
2232 CPU 1 CPU 2 2204 CPU 1 CPU 2
2233 =============================== ===== 2205 =============================== ===============================
2234 X = 1; Y = 1 2206 X = 1; Y = 1;
2235 smp_mb(); wake_ 2207 smp_mb(); wake_up();
2236 LOAD Y LOAD 2208 LOAD Y LOAD X
2237 2209
2238 If a wakeup does occur, one (at least) of the 2210 If a wakeup does occur, one (at least) of the two loads must see 1. If, on
2239 the other hand, a wakeup does not occur, both 2211 the other hand, a wakeup does not occur, both loads might see 0.
2240 2212
2241 wake_up_process() always executes a general m 2213 wake_up_process() always executes a general memory barrier. The barrier again
2242 occurs before the task state is accessed. In 2214 occurs before the task state is accessed. In particular, if the wake_up() in
2243 the previous snippet were replaced by a call 2215 the previous snippet were replaced by a call to wake_up_process() then one of
2244 the two loads would be guaranteed to see 1. 2216 the two loads would be guaranteed to see 1.
2245 2217
2246 The available waker functions include: 2218 The available waker functions include:
2247 2219
2248 complete(); 2220 complete();
2249 wake_up(); 2221 wake_up();
2250 wake_up_all(); 2222 wake_up_all();
2251 wake_up_bit(); 2223 wake_up_bit();
2252 wake_up_interruptible(); 2224 wake_up_interruptible();
2253 wake_up_interruptible_all(); 2225 wake_up_interruptible_all();
2254 wake_up_interruptible_nr(); 2226 wake_up_interruptible_nr();
2255 wake_up_interruptible_poll(); 2227 wake_up_interruptible_poll();
2256 wake_up_interruptible_sync(); 2228 wake_up_interruptible_sync();
2257 wake_up_interruptible_sync_poll(); 2229 wake_up_interruptible_sync_poll();
2258 wake_up_locked(); 2230 wake_up_locked();
2259 wake_up_locked_poll(); 2231 wake_up_locked_poll();
2260 wake_up_nr(); 2232 wake_up_nr();
2261 wake_up_poll(); 2233 wake_up_poll();
2262 wake_up_process(); 2234 wake_up_process();
2263 2235
2264 In terms of memory ordering, these functions 2236 In terms of memory ordering, these functions all provide the same guarantees of
2265 a wake_up() (or stronger). 2237 a wake_up() (or stronger).
2266 2238
2267 [!] Note that the memory barriers implied by 2239 [!] Note that the memory barriers implied by the sleeper and the waker do _not_
2268 order multiple stores before the wake-up with 2240 order multiple stores before the wake-up with respect to loads of those stored
2269 values after the sleeper has called set_curre 2241 values after the sleeper has called set_current_state(). For instance, if the
2270 sleeper does: 2242 sleeper does:
2271 2243
2272 set_current_state(TASK_INTERRUPTIBLE) 2244 set_current_state(TASK_INTERRUPTIBLE);
2273 if (event_indicated) 2245 if (event_indicated)
2274 break; 2246 break;
2275 __set_current_state(TASK_RUNNING); 2247 __set_current_state(TASK_RUNNING);
2276 do_something(my_data); 2248 do_something(my_data);
2277 2249
2278 and the waker does: 2250 and the waker does:
2279 2251
2280 my_data = value; 2252 my_data = value;
2281 event_indicated = 1; 2253 event_indicated = 1;
2282 wake_up(&event_wait_queue); 2254 wake_up(&event_wait_queue);
2283 2255
2284 there's no guarantee that the change to event 2256 there's no guarantee that the change to event_indicated will be perceived by
2285 the sleeper as coming after the change to my_ 2257 the sleeper as coming after the change to my_data. In such a circumstance, the
2286 code on both sides must interpolate its own m 2258 code on both sides must interpolate its own memory barriers between the
2287 separate data accesses. Thus the above sleep 2259 separate data accesses. Thus the above sleeper ought to do:
2288 2260
2289 set_current_state(TASK_INTERRUPTIBLE) 2261 set_current_state(TASK_INTERRUPTIBLE);
2290 if (event_indicated) { 2262 if (event_indicated) {
2291 smp_rmb(); 2263 smp_rmb();
2292 do_something(my_data); 2264 do_something(my_data);
2293 } 2265 }
2294 2266
2295 and the waker should do: 2267 and the waker should do:
2296 2268
2297 my_data = value; 2269 my_data = value;
2298 smp_wmb(); 2270 smp_wmb();
2299 event_indicated = 1; 2271 event_indicated = 1;
2300 wake_up(&event_wait_queue); 2272 wake_up(&event_wait_queue);
2301 2273
2302 2274
2303 MISCELLANEOUS FUNCTIONS 2275 MISCELLANEOUS FUNCTIONS
2304 ----------------------- 2276 -----------------------
2305 2277
2306 Other functions that imply barriers: 2278 Other functions that imply barriers:
2307 2279
2308 (*) schedule() and similar imply full memory 2280 (*) schedule() and similar imply full memory barriers.
2309 2281
2310 2282
2311 =================================== 2283 ===================================
2312 INTER-CPU ACQUIRING BARRIER EFFECTS 2284 INTER-CPU ACQUIRING BARRIER EFFECTS
2313 =================================== 2285 ===================================
2314 2286
2315 On SMP systems locking primitives give a more 2287 On SMP systems locking primitives give a more substantial form of barrier: one
2316 that does affect memory access ordering on ot 2288 that does affect memory access ordering on other CPUs, within the context of
2317 conflict on any particular lock. 2289 conflict on any particular lock.
2318 2290
2319 2291
2320 ACQUIRES VS MEMORY ACCESSES 2292 ACQUIRES VS MEMORY ACCESSES
2321 --------------------------- 2293 ---------------------------
2322 2294
2323 Consider the following: the system has a pair 2295 Consider the following: the system has a pair of spinlocks (M) and (Q), and
2324 three CPUs; then should the following sequenc 2296 three CPUs; then should the following sequence of events occur:
2325 2297
2326 CPU 1 CPU 2 2298 CPU 1 CPU 2
2327 =============================== ===== 2299 =============================== ===============================
2328 WRITE_ONCE(*A, a); WRITE 2300 WRITE_ONCE(*A, a); WRITE_ONCE(*E, e);
2329 ACQUIRE M ACQUI 2301 ACQUIRE M ACQUIRE Q
2330 WRITE_ONCE(*B, b); WRITE 2302 WRITE_ONCE(*B, b); WRITE_ONCE(*F, f);
2331 WRITE_ONCE(*C, c); WRITE 2303 WRITE_ONCE(*C, c); WRITE_ONCE(*G, g);
2332 RELEASE M RELEA 2304 RELEASE M RELEASE Q
2333 WRITE_ONCE(*D, d); WRITE 2305 WRITE_ONCE(*D, d); WRITE_ONCE(*H, h);
2334 2306
2335 Then there is no guarantee as to what order C 2307 Then there is no guarantee as to what order CPU 3 will see the accesses to *A
2336 through *H occur in, other than the constrain 2308 through *H occur in, other than the constraints imposed by the separate locks
2337 on the separate CPUs. It might, for example, 2309 on the separate CPUs. It might, for example, see:
2338 2310
2339 *E, ACQUIRE M, ACQUIRE Q, *G, *C, *F, 2311 *E, ACQUIRE M, ACQUIRE Q, *G, *C, *F, *A, *B, RELEASE Q, *D, *H, RELEASE M
2340 2312
2341 But it won't see any of: 2313 But it won't see any of:
2342 2314
2343 *B, *C or *D preceding ACQUIRE M 2315 *B, *C or *D preceding ACQUIRE M
2344 *A, *B or *C following RELEASE M 2316 *A, *B or *C following RELEASE M
2345 *F, *G or *H preceding ACQUIRE Q 2317 *F, *G or *H preceding ACQUIRE Q
2346 *E, *F or *G following RELEASE Q 2318 *E, *F or *G following RELEASE Q
2347 2319
2348 2320
>> 2321
>> 2322 ACQUIRES VS I/O ACCESSES
>> 2323 ------------------------
>> 2324
>> 2325 Under certain circumstances (especially involving NUMA), I/O accesses within
>> 2326 two spinlocked sections on two different CPUs may be seen as interleaved by the
>> 2327 PCI bridge, because the PCI bridge does not necessarily participate in the
>> 2328 cache-coherence protocol, and is therefore incapable of issuing the required
>> 2329 read memory barriers.
>> 2330
>> 2331 For example:
>> 2332
>> 2333 CPU 1 CPU 2
>> 2334 =============================== ===============================
>> 2335 spin_lock(Q)
>> 2336 writel(0, ADDR)
>> 2337 writel(1, DATA);
>> 2338 spin_unlock(Q);
>> 2339 spin_lock(Q);
>> 2340 writel(4, ADDR);
>> 2341 writel(5, DATA);
>> 2342 spin_unlock(Q);
>> 2343
>> 2344 may be seen by the PCI bridge as follows:
>> 2345
>> 2346 STORE *ADDR = 0, STORE *ADDR = 4, STORE *DATA = 1, STORE *DATA = 5
>> 2347
>> 2348 which would probably cause the hardware to malfunction.
>> 2349
>> 2350
>> 2351 What is necessary here is to intervene with an mmiowb() before dropping the
>> 2352 spinlock, for example:
>> 2353
>> 2354 CPU 1 CPU 2
>> 2355 =============================== ===============================
>> 2356 spin_lock(Q)
>> 2357 writel(0, ADDR)
>> 2358 writel(1, DATA);
>> 2359 mmiowb();
>> 2360 spin_unlock(Q);
>> 2361 spin_lock(Q);
>> 2362 writel(4, ADDR);
>> 2363 writel(5, DATA);
>> 2364 mmiowb();
>> 2365 spin_unlock(Q);
>> 2366
>> 2367 this will ensure that the two stores issued on CPU 1 appear at the PCI bridge
>> 2368 before either of the stores issued on CPU 2.
>> 2369
>> 2370
>> 2371 Furthermore, following a store by a load from the same device obviates the need
>> 2372 for the mmiowb(), because the load forces the store to complete before the load
>> 2373 is performed:
>> 2374
>> 2375 CPU 1 CPU 2
>> 2376 =============================== ===============================
>> 2377 spin_lock(Q)
>> 2378 writel(0, ADDR)
>> 2379 a = readl(DATA);
>> 2380 spin_unlock(Q);
>> 2381 spin_lock(Q);
>> 2382 writel(4, ADDR);
>> 2383 b = readl(DATA);
>> 2384 spin_unlock(Q);
>> 2385
>> 2386
>> 2387 See Documentation/driver-api/device-io.rst for more information.
>> 2388
>> 2389
2349 ================================= 2390 =================================
2350 WHERE ARE MEMORY BARRIERS NEEDED? 2391 WHERE ARE MEMORY BARRIERS NEEDED?
2351 ================================= 2392 =================================
2352 2393
2353 Under normal operation, memory operation reor 2394 Under normal operation, memory operation reordering is generally not going to
2354 be a problem as a single-threaded linear piec 2395 be a problem as a single-threaded linear piece of code will still appear to
2355 work correctly, even if it's in an SMP kernel 2396 work correctly, even if it's in an SMP kernel. There are, however, four
2356 circumstances in which reordering definitely 2397 circumstances in which reordering definitely _could_ be a problem:
2357 2398
2358 (*) Interprocessor interaction. 2399 (*) Interprocessor interaction.
2359 2400
2360 (*) Atomic operations. 2401 (*) Atomic operations.
2361 2402
2362 (*) Accessing devices. 2403 (*) Accessing devices.
2363 2404
2364 (*) Interrupts. 2405 (*) Interrupts.
2365 2406
2366 2407
2367 INTERPROCESSOR INTERACTION 2408 INTERPROCESSOR INTERACTION
2368 -------------------------- 2409 --------------------------
2369 2410
2370 When there's a system with more than one proc 2411 When there's a system with more than one processor, more than one CPU in the
2371 system may be working on the same data set at 2412 system may be working on the same data set at the same time. This can cause
2372 synchronisation problems, and the usual way o 2413 synchronisation problems, and the usual way of dealing with them is to use
2373 locks. Locks, however, are quite expensive, 2414 locks. Locks, however, are quite expensive, and so it may be preferable to
2374 operate without the use of a lock if at all p 2415 operate without the use of a lock if at all possible. In such a case
2375 operations that affect both CPUs may have to 2416 operations that affect both CPUs may have to be carefully ordered to prevent
2376 a malfunction. 2417 a malfunction.
2377 2418
2378 Consider, for example, the R/W semaphore slow 2419 Consider, for example, the R/W semaphore slow path. Here a waiting process is
2379 queued on the semaphore, by virtue of it havi 2420 queued on the semaphore, by virtue of it having a piece of its stack linked to
2380 the semaphore's list of waiting processes: 2421 the semaphore's list of waiting processes:
2381 2422
2382 struct rw_semaphore { 2423 struct rw_semaphore {
2383 ... 2424 ...
2384 spinlock_t lock; 2425 spinlock_t lock;
2385 struct list_head waiters; 2426 struct list_head waiters;
2386 }; 2427 };
2387 2428
2388 struct rwsem_waiter { 2429 struct rwsem_waiter {
2389 struct list_head list; 2430 struct list_head list;
2390 struct task_struct *task; 2431 struct task_struct *task;
2391 }; 2432 };
2392 2433
2393 To wake up a particular waiter, the up_read() 2434 To wake up a particular waiter, the up_read() or up_write() functions have to:
2394 2435
2395 (1) read the next pointer from this waiter's 2436 (1) read the next pointer from this waiter's record to know as to where the
2396 next waiter record is; 2437 next waiter record is;
2397 2438
2398 (2) read the pointer to the waiter's task st 2439 (2) read the pointer to the waiter's task structure;
2399 2440
2400 (3) clear the task pointer to tell the waite 2441 (3) clear the task pointer to tell the waiter it has been given the semaphore;
2401 2442
2402 (4) call wake_up_process() on the task; and 2443 (4) call wake_up_process() on the task; and
2403 2444
2404 (5) release the reference held on the waiter 2445 (5) release the reference held on the waiter's task struct.
2405 2446
2406 In other words, it has to perform this sequen 2447 In other words, it has to perform this sequence of events:
2407 2448
2408 LOAD waiter->list.next; 2449 LOAD waiter->list.next;
2409 LOAD waiter->task; 2450 LOAD waiter->task;
2410 STORE waiter->task; 2451 STORE waiter->task;
2411 CALL wakeup 2452 CALL wakeup
2412 RELEASE task 2453 RELEASE task
2413 2454
2414 and if any of these steps occur out of order, 2455 and if any of these steps occur out of order, then the whole thing may
2415 malfunction. 2456 malfunction.
2416 2457
2417 Once it has queued itself and dropped the sem 2458 Once it has queued itself and dropped the semaphore lock, the waiter does not
2418 get the lock again; it instead just waits for 2459 get the lock again; it instead just waits for its task pointer to be cleared
2419 before proceeding. Since the record is on th 2460 before proceeding. Since the record is on the waiter's stack, this means that
2420 if the task pointer is cleared _before_ the n 2461 if the task pointer is cleared _before_ the next pointer in the list is read,
2421 another CPU might start processing the waiter 2462 another CPU might start processing the waiter and might clobber the waiter's
2422 stack before the up*() function has a chance 2463 stack before the up*() function has a chance to read the next pointer.
2423 2464
2424 Consider then what might happen to the above 2465 Consider then what might happen to the above sequence of events:
2425 2466
2426 CPU 1 CPU 2 2467 CPU 1 CPU 2
2427 =============================== ===== 2468 =============================== ===============================
2428 down_ 2469 down_xxx()
2429 Queue 2470 Queue waiter
2430 Sleep 2471 Sleep
2431 up_yyy() 2472 up_yyy()
2432 LOAD waiter->task; 2473 LOAD waiter->task;
2433 STORE waiter->task; 2474 STORE waiter->task;
2434 Woken 2475 Woken up by other event
2435 <preempt> 2476 <preempt>
2436 Resum 2477 Resume processing
2437 down_ 2478 down_xxx() returns
2438 call 2479 call foo()
2439 foo() 2480 foo() clobbers *waiter
2440 </preempt> 2481 </preempt>
2441 LOAD waiter->list.next; 2482 LOAD waiter->list.next;
2442 --- OOPS --- 2483 --- OOPS ---
2443 2484
2444 This could be dealt with using the semaphore 2485 This could be dealt with using the semaphore lock, but then the down_xxx()
2445 function has to needlessly get the spinlock a 2486 function has to needlessly get the spinlock again after being woken up.
2446 2487
2447 The way to deal with this is to insert a gene 2488 The way to deal with this is to insert a general SMP memory barrier:
2448 2489
2449 LOAD waiter->list.next; 2490 LOAD waiter->list.next;
2450 LOAD waiter->task; 2491 LOAD waiter->task;
2451 smp_mb(); 2492 smp_mb();
2452 STORE waiter->task; 2493 STORE waiter->task;
2453 CALL wakeup 2494 CALL wakeup
2454 RELEASE task 2495 RELEASE task
2455 2496
2456 In this case, the barrier makes a guarantee t 2497 In this case, the barrier makes a guarantee that all memory accesses before the
2457 barrier will appear to happen before all the 2498 barrier will appear to happen before all the memory accesses after the barrier
2458 with respect to the other CPUs on the system. 2499 with respect to the other CPUs on the system. It does _not_ guarantee that all
2459 the memory accesses before the barrier will b 2500 the memory accesses before the barrier will be complete by the time the barrier
2460 instruction itself is complete. 2501 instruction itself is complete.
2461 2502
2462 On a UP system - where this wouldn't be a pro 2503 On a UP system - where this wouldn't be a problem - the smp_mb() is just a
2463 compiler barrier, thus making sure the compil 2504 compiler barrier, thus making sure the compiler emits the instructions in the
2464 right order without actually intervening in t 2505 right order without actually intervening in the CPU. Since there's only one
2465 CPU, that CPU's dependency ordering logic wil 2506 CPU, that CPU's dependency ordering logic will take care of everything else.
2466 2507
2467 2508
2468 ATOMIC OPERATIONS 2509 ATOMIC OPERATIONS
2469 ----------------- 2510 -----------------
2470 2511
2471 While they are technically interprocessor int !! 2512 Whilst they are technically interprocessor interaction considerations, atomic
2472 operations are noted specially as some of the 2513 operations are noted specially as some of them imply full memory barriers and
2473 some don't, but they're very heavily relied o 2514 some don't, but they're very heavily relied on as a group throughout the
2474 kernel. 2515 kernel.
2475 2516
2476 See Documentation/atomic_t.txt for more infor 2517 See Documentation/atomic_t.txt for more information.
2477 2518
2478 2519
2479 ACCESSING DEVICES 2520 ACCESSING DEVICES
2480 ----------------- 2521 -----------------
2481 2522
2482 Many devices can be memory mapped, and so app 2523 Many devices can be memory mapped, and so appear to the CPU as if they're just
2483 a set of memory locations. To control such a 2524 a set of memory locations. To control such a device, the driver usually has to
2484 make the right memory accesses in exactly the 2525 make the right memory accesses in exactly the right order.
2485 2526
2486 However, having a clever CPU or a clever comp 2527 However, having a clever CPU or a clever compiler creates a potential problem
2487 in that the carefully sequenced accesses in t 2528 in that the carefully sequenced accesses in the driver code won't reach the
2488 device in the requisite order if the CPU or t 2529 device in the requisite order if the CPU or the compiler thinks it is more
2489 efficient to reorder, combine or merge access 2530 efficient to reorder, combine or merge accesses - something that would cause
2490 the device to malfunction. 2531 the device to malfunction.
2491 2532
2492 Inside of the Linux kernel, I/O should be don 2533 Inside of the Linux kernel, I/O should be done through the appropriate accessor
2493 routines - such as inb() or writel() - which 2534 routines - such as inb() or writel() - which know how to make such accesses
2494 appropriately sequential. While this, for th !! 2535 appropriately sequential. Whilst this, for the most part, renders the explicit
2495 use of memory barriers unnecessary, if the ac !! 2536 use of memory barriers unnecessary, there are a couple of situations where they
2496 to an I/O memory window with relaxed memory a !! 2537 might be needed:
2497 memory barriers are required to enforce order !! 2538
>> 2539 (1) On some systems, I/O stores are not strongly ordered across all CPUs, and
>> 2540 so for _all_ general drivers locks should be used and mmiowb() must be
>> 2541 issued prior to unlocking the critical section.
>> 2542
>> 2543 (2) If the accessor functions are used to refer to an I/O memory window with
>> 2544 relaxed memory access properties, then _mandatory_ memory barriers are
>> 2545 required to enforce ordering.
2498 2546
2499 See Documentation/driver-api/device-io.rst fo 2547 See Documentation/driver-api/device-io.rst for more information.
2500 2548
2501 2549
2502 INTERRUPTS 2550 INTERRUPTS
2503 ---------- 2551 ----------
2504 2552
2505 A driver may be interrupted by its own interr 2553 A driver may be interrupted by its own interrupt service routine, and thus the
2506 two parts of the driver may interfere with ea 2554 two parts of the driver may interfere with each other's attempts to control or
2507 access the device. 2555 access the device.
2508 2556
2509 This may be alleviated - at least in part - b 2557 This may be alleviated - at least in part - by disabling local interrupts (a
2510 form of locking), such that the critical oper 2558 form of locking), such that the critical operations are all contained within
2511 the interrupt-disabled section in the driver. !! 2559 the interrupt-disabled section in the driver. Whilst the driver's interrupt
2512 routine is executing, the driver's core may n 2560 routine is executing, the driver's core may not run on the same CPU, and its
2513 interrupt is not permitted to happen again un 2561 interrupt is not permitted to happen again until the current interrupt has been
2514 handled, thus the interrupt handler does not 2562 handled, thus the interrupt handler does not need to lock against that.
2515 2563
2516 However, consider a driver that was talking t 2564 However, consider a driver that was talking to an ethernet card that sports an
2517 address register and a data register. If tha 2565 address register and a data register. If that driver's core talks to the card
2518 under interrupt-disablement and then the driv 2566 under interrupt-disablement and then the driver's interrupt handler is invoked:
2519 2567
2520 LOCAL IRQ DISABLE 2568 LOCAL IRQ DISABLE
2521 writew(ADDR, 3); 2569 writew(ADDR, 3);
2522 writew(DATA, y); 2570 writew(DATA, y);
2523 LOCAL IRQ ENABLE 2571 LOCAL IRQ ENABLE
2524 <interrupt> 2572 <interrupt>
2525 writew(ADDR, 4); 2573 writew(ADDR, 4);
2526 q = readw(DATA); 2574 q = readw(DATA);
2527 </interrupt> 2575 </interrupt>
2528 2576
2529 The store to the data register might happen a 2577 The store to the data register might happen after the second store to the
2530 address register if ordering rules are suffic 2578 address register if ordering rules are sufficiently relaxed:
2531 2579
2532 STORE *ADDR = 3, STORE *ADDR = 4, STO 2580 STORE *ADDR = 3, STORE *ADDR = 4, STORE *DATA = y, q = LOAD *DATA
2533 2581
2534 2582
2535 If ordering rules are relaxed, it must be ass 2583 If ordering rules are relaxed, it must be assumed that accesses done inside an
2536 interrupt disabled section may leak outside o 2584 interrupt disabled section may leak outside of it and may interleave with
2537 accesses performed in an interrupt - and vice 2585 accesses performed in an interrupt - and vice versa - unless implicit or
2538 explicit barriers are used. 2586 explicit barriers are used.
2539 2587
2540 Normally this won't be a problem because the 2588 Normally this won't be a problem because the I/O accesses done inside such
2541 sections will include synchronous load operat 2589 sections will include synchronous load operations on strictly ordered I/O
2542 registers that form implicit I/O barriers. !! 2590 registers that form implicit I/O barriers. If this isn't sufficient then an
>> 2591 mmiowb() may need to be used explicitly.
2543 2592
2544 2593
2545 A similar situation may occur between an inte 2594 A similar situation may occur between an interrupt routine and two routines
2546 running on separate CPUs that communicate wit 2595 running on separate CPUs that communicate with each other. If such a case is
2547 likely, then interrupt-disabling locks should 2596 likely, then interrupt-disabling locks should be used to guarantee ordering.
2548 2597
2549 2598
2550 ========================== 2599 ==========================
2551 KERNEL I/O BARRIER EFFECTS 2600 KERNEL I/O BARRIER EFFECTS
2552 ========================== 2601 ==========================
2553 2602
2554 Interfacing with peripherals via I/O accesses !! 2603 When accessing I/O memory, drivers should use the appropriate accessor
2555 specific. Therefore, drivers which are inhere !! 2604 functions:
2556 specific behaviours of their target systems i !! 2605
2557 in the most lightweight manner possible. For !! 2606 (*) inX(), outX():
2558 between multiple architectures and bus implem !! 2607
2559 series of accessor functions that provide var !! 2608 These are intended to talk to I/O space rather than memory space, but
2560 guarantees: !! 2609 that's primarily a CPU-specific concept. The i386 and x86_64 processors
>> 2610 do indeed have special I/O space access cycles and instructions, but many
>> 2611 CPUs don't have such a concept.
>> 2612
>> 2613 The PCI bus, amongst others, defines an I/O space concept which - on such
>> 2614 CPUs as i386 and x86_64 - readily maps to the CPU's concept of I/O
>> 2615 space. However, it may also be mapped as a virtual I/O space in the CPU's
>> 2616 memory map, particularly on those CPUs that don't support alternate I/O
>> 2617 spaces.
>> 2618
>> 2619 Accesses to this space may be fully synchronous (as on i386), but
>> 2620 intermediary bridges (such as the PCI host bridge) may not fully honour
>> 2621 that.
>> 2622
>> 2623 They are guaranteed to be fully ordered with respect to each other.
>> 2624
>> 2625 They are not guaranteed to be fully ordered with respect to other types of
>> 2626 memory and I/O operation.
2561 2627
2562 (*) readX(), writeX(): 2628 (*) readX(), writeX():
2563 2629
2564 The readX() and writeX() MMIO accesso !! 2630 Whether these are guaranteed to be fully ordered and uncombined with
2565 peripheral being accessed as an __iom !! 2631 respect to each other on the issuing CPU depends on the characteristics
2566 mapped with the default I/O attribute !! 2632 defined for the memory window through which they're accessing. On later
2567 ioremap()), the ordering guarantees a !! 2633 i386 architecture machines, for example, this is controlled by way of the
2568 !! 2634 MTRR registers.
2569 1. All readX() and writeX() accesses !! 2635
2570 with respect to each other. This e !! 2636 Ordinarily, these will be guaranteed to be fully ordered and uncombined,
2571 by the same CPU thread to a partic !! 2637 provided they're not accessing a prefetchable device.
2572 order. !! 2638
2573 !! 2639 However, intermediary hardware (such as a PCI bridge) may indulge in
2574 2. A writeX() issued by a CPU thread !! 2640 deferral if it so wishes; to flush a store, a load from the same location
2575 before a writeX() to the same peri !! 2641 is preferred[*], but a load from the same device or from configuration
2576 issued after a later acquisition o !! 2642 space should suffice for PCI.
2577 that MMIO register writes to a par !! 2643
2578 a spinlock will arrive in an order !! 2644 [*] NOTE! attempting to load from the same location as was written to may
2579 the lock. !! 2645 cause a malfunction - consider the 16550 Rx/Tx serial registers for
2580 !! 2646 example.
2581 3. A writeX() by a CPU thread to the !! 2647
2582 completion of all prior writes to !! 2648 Used with prefetchable I/O memory, an mmiowb() barrier may be required to
2583 propagated to, the same thread. Th !! 2649 force stores to be ordered.
2584 to an outbound DMA buffer allocate !! 2650
2585 visible to a DMA engine when the C !! 2651 Please refer to the PCI specification for more information on interactions
2586 register to trigger the transfer. !! 2652 between PCI transactions.
2587 !! 2653
2588 4. A readX() by a CPU thread from the !! 2654 (*) readX_relaxed(), writeX_relaxed()
2589 any subsequent reads from memory b !! 2655
2590 ensures that reads by the CPU from !! 2656 These are similar to readX() and writeX(), but provide weaker memory
2591 by dma_alloc_coherent() will not s !! 2657 ordering guarantees. Specifically, they do not guarantee ordering with
2592 the DMA engine's MMIO status regis !! 2658 respect to normal memory accesses (e.g. DMA buffers) nor do they guarantee
2593 transfer has completed. !! 2659 ordering with respect to LOCK or UNLOCK operations. If the latter is
2594 !! 2660 required, an mmiowb() barrier can be used. Note that relaxed accesses to
2595 5. A readX() by a CPU thread from the !! 2661 the same peripheral are guaranteed to be ordered with respect to each
2596 any subsequent delay() loop can be !! 2662 other.
2597 This ensures that two MMIO registe <<
2598 will arrive at least 1us apart if <<
2599 back with readX() and udelay(1) is <<
2600 writeX(): <<
2601 <<
2602 writel(42, DEVICE_REGISTER_0) <<
2603 readl(DEVICE_REGISTER_0); <<
2604 udelay(1); <<
2605 writel(42, DEVICE_REGISTER_1) <<
2606 <<
2607 The ordering properties of __iomem po <<
2608 attributes (e.g. those returned by io <<
2609 underlying architecture and therefore <<
2610 generally be relied upon for accesses <<
2611 <<
2612 (*) readX_relaxed(), writeX_relaxed(): <<
2613 <<
2614 These are similar to readX() and writ <<
2615 ordering guarantees. Specifically, th <<
2616 respect to locking, normal memory acc <<
2617 bullets 2-5 above) but they are still <<
2618 respect to other accesses from the sa <<
2619 peripheral when operating on __iomem <<
2620 I/O attributes. <<
2621 <<
2622 (*) readsX(), writesX(): <<
2623 <<
2624 The readsX() and writesX() MMIO acces <<
2625 register-based, memory-mapped FIFOs r <<
2626 capable of performing DMA. Consequent <<
2627 guarantees of readX_relaxed() and wri <<
2628 2663
2629 (*) inX(), outX(): !! 2664 (*) ioreadX(), iowriteX()
2630 2665
2631 The inX() and outX() accessors are in !! 2666 These will perform appropriately for the type of access they're actually
2632 I/O peripherals, which may require sp !! 2667 doing, be it inX()/outX() or readX()/writeX().
2633 architectures (notably x86). The port <<
2634 accessed is passed as an argument. <<
2635 <<
2636 Since many CPU architectures ultimate <<
2637 internal virtual memory mapping, the <<
2638 provided by inX() and outX() are the <<
2639 and writeX() respectively when access <<
2640 attributes. <<
2641 <<
2642 Device drivers may expect outX() to e <<
2643 that waits for a completion response <<
2644 returning. This is not guaranteed by <<
2645 not part of the portable ordering sem <<
2646 <<
2647 (*) insX(), outsX(): <<
2648 <<
2649 As above, the insX() and outsX() acce <<
2650 guarantees as readsX() and writesX() <<
2651 mapping with the default I/O attribut <<
2652 <<
2653 (*) ioreadX(), iowriteX(): <<
2654 <<
2655 These will perform appropriately for <<
2656 doing, be it inX()/outX() or readX()/ <<
2657 <<
2658 With the exception of the string accessors (i <<
2659 writesX()), all of the above assume that the <<
2660 little-endian and will therefore perform byte <<
2661 architectures. <<
2662 2668
2663 2669
2664 ======================================== 2670 ========================================
2665 ASSUMED MINIMUM EXECUTION ORDERING MODEL 2671 ASSUMED MINIMUM EXECUTION ORDERING MODEL
2666 ======================================== 2672 ========================================
2667 2673
2668 It has to be assumed that the conceptual CPU 2674 It has to be assumed that the conceptual CPU is weakly-ordered but that it will
2669 maintain the appearance of program causality 2675 maintain the appearance of program causality with respect to itself. Some CPUs
2670 (such as i386 or x86_64) are more constrained 2676 (such as i386 or x86_64) are more constrained than others (such as powerpc or
2671 frv), and so the most relaxed case (namely DE 2677 frv), and so the most relaxed case (namely DEC Alpha) must be assumed outside
2672 of arch-specific code. 2678 of arch-specific code.
2673 2679
2674 This means that it must be considered that th 2680 This means that it must be considered that the CPU will execute its instruction
2675 stream in any order it feels like - or even i 2681 stream in any order it feels like - or even in parallel - provided that if an
2676 instruction in the stream depends on an earli 2682 instruction in the stream depends on an earlier instruction, then that
2677 earlier instruction must be sufficiently comp 2683 earlier instruction must be sufficiently complete[*] before the later
2678 instruction may proceed; in other words: prov 2684 instruction may proceed; in other words: provided that the appearance of
2679 causality is maintained. 2685 causality is maintained.
2680 2686
2681 [*] Some instructions have more than one eff 2687 [*] Some instructions have more than one effect - such as changing the
2682 condition codes, changing registers or c 2688 condition codes, changing registers or changing memory - and different
2683 instructions may depend on different eff 2689 instructions may depend on different effects.
2684 2690
2685 A CPU may also discard any instruction sequen 2691 A CPU may also discard any instruction sequence that winds up having no
2686 ultimate effect. For example, if two adjacen 2692 ultimate effect. For example, if two adjacent instructions both load an
2687 immediate value into the same register, the f 2693 immediate value into the same register, the first may be discarded.
2688 2694
2689 2695
2690 Similarly, it has to be assumed that compiler 2696 Similarly, it has to be assumed that compiler might reorder the instruction
2691 stream in any way it sees fit, again provided 2697 stream in any way it sees fit, again provided the appearance of causality is
2692 maintained. 2698 maintained.
2693 2699
2694 2700
2695 ============================ 2701 ============================
2696 THE EFFECTS OF THE CPU CACHE 2702 THE EFFECTS OF THE CPU CACHE
2697 ============================ 2703 ============================
2698 2704
2699 The way cached memory operations are perceive 2705 The way cached memory operations are perceived across the system is affected to
2700 a certain extent by the caches that lie betwe 2706 a certain extent by the caches that lie between CPUs and memory, and by the
2701 memory coherence system that maintains the co 2707 memory coherence system that maintains the consistency of state in the system.
2702 2708
2703 As far as the way a CPU interacts with anothe 2709 As far as the way a CPU interacts with another part of the system through the
2704 caches goes, the memory system has to include 2710 caches goes, the memory system has to include the CPU's caches, and memory
2705 barriers for the most part act at the interfa 2711 barriers for the most part act at the interface between the CPU and its cache
2706 (memory barriers logically act on the dotted 2712 (memory barriers logically act on the dotted line in the following diagram):
2707 2713
2708 <--- CPU ---> : <-- 2714 <--- CPU ---> : <----------- Memory ----------->
2709 : 2715 :
2710 +--------+ +--------+ : +------ 2716 +--------+ +--------+ : +--------+ +-----------+
2711 | | | | : | 2717 | | | | : | | | | +--------+
2712 | CPU | | Memory | : | CPU 2718 | CPU | | Memory | : | CPU | | | | |
2713 | Core |--->| Access |----->| Cache 2719 | Core |--->| Access |----->| Cache |<-->| | | |
2714 | | | Queue | : | 2720 | | | Queue | : | | | |--->| Memory |
2715 | | | | : | 2721 | | | | : | | | | | |
2716 +--------+ +--------+ : +------ 2722 +--------+ +--------+ : +--------+ | | | |
2717 : 2723 : | Cache | +--------+
2718 : 2724 : | Coherency |
2719 : 2725 : | Mechanism | +--------+
2720 +--------+ +--------+ : +------ 2726 +--------+ +--------+ : +--------+ | | | |
2721 | | | | : | 2727 | | | | : | | | | | |
2722 | CPU | | Memory | : | CPU 2728 | CPU | | Memory | : | CPU | | |--->| Device |
2723 | Core |--->| Access |----->| Cache 2729 | Core |--->| Access |----->| Cache |<-->| | | |
2724 | | | Queue | : | 2730 | | | Queue | : | | | | | |
2725 | | | | : | 2731 | | | | : | | | | +--------+
2726 +--------+ +--------+ : +------ 2732 +--------+ +--------+ : +--------+ +-----------+
2727 : 2733 :
2728 : 2734 :
2729 2735
2730 Although any particular load or store may not 2736 Although any particular load or store may not actually appear outside of the
2731 CPU that issued it since it may have been sat 2737 CPU that issued it since it may have been satisfied within the CPU's own cache,
2732 it will still appear as if the full memory ac 2738 it will still appear as if the full memory access had taken place as far as the
2733 other CPUs are concerned since the cache cohe 2739 other CPUs are concerned since the cache coherency mechanisms will migrate the
2734 cacheline over to the accessing CPU and propa 2740 cacheline over to the accessing CPU and propagate the effects upon conflict.
2735 2741
2736 The CPU core may execute instructions in any 2742 The CPU core may execute instructions in any order it deems fit, provided the
2737 expected program causality appears to be main 2743 expected program causality appears to be maintained. Some of the instructions
2738 generate load and store operations which then 2744 generate load and store operations which then go into the queue of memory
2739 accesses to be performed. The core may place 2745 accesses to be performed. The core may place these in the queue in any order
2740 it wishes, and continue execution until it is 2746 it wishes, and continue execution until it is forced to wait for an instruction
2741 to complete. 2747 to complete.
2742 2748
2743 What memory barriers are concerned with is co 2749 What memory barriers are concerned with is controlling the order in which
2744 accesses cross from the CPU side of things to 2750 accesses cross from the CPU side of things to the memory side of things, and
2745 the order in which the effects are perceived 2751 the order in which the effects are perceived to happen by the other observers
2746 in the system. 2752 in the system.
2747 2753
2748 [!] Memory barriers are _not_ needed within a 2754 [!] Memory barriers are _not_ needed within a given CPU, as CPUs always see
2749 their own loads and stores as if they had hap 2755 their own loads and stores as if they had happened in program order.
2750 2756
2751 [!] MMIO or other device accesses may bypass 2757 [!] MMIO or other device accesses may bypass the cache system. This depends on
2752 the properties of the memory window through w 2758 the properties of the memory window through which devices are accessed and/or
2753 the use of any special device communication i 2759 the use of any special device communication instructions the CPU may have.
2754 2760
2755 2761
>> 2762 CACHE COHERENCY
>> 2763 ---------------
>> 2764
>> 2765 Life isn't quite as simple as it may appear above, however: for while the
>> 2766 caches are expected to be coherent, there's no guarantee that that coherency
>> 2767 will be ordered. This means that whilst changes made on one CPU will
>> 2768 eventually become visible on all CPUs, there's no guarantee that they will
>> 2769 become apparent in the same order on those other CPUs.
>> 2770
>> 2771
>> 2772 Consider dealing with a system that has a pair of CPUs (1 & 2), each of which
>> 2773 has a pair of parallel data caches (CPU 1 has A/B, and CPU 2 has C/D):
>> 2774
>> 2775 :
>> 2776 : +--------+
>> 2777 : +---------+ | |
>> 2778 +--------+ : +--->| Cache A |<------->| |
>> 2779 | | : | +---------+ | |
>> 2780 | CPU 1 |<---+ | |
>> 2781 | | : | +---------+ | |
>> 2782 +--------+ : +--->| Cache B |<------->| |
>> 2783 : +---------+ | |
>> 2784 : | Memory |
>> 2785 : +---------+ | System |
>> 2786 +--------+ : +--->| Cache C |<------->| |
>> 2787 | | : | +---------+ | |
>> 2788 | CPU 2 |<---+ | |
>> 2789 | | : | +---------+ | |
>> 2790 +--------+ : +--->| Cache D |<------->| |
>> 2791 : +---------+ | |
>> 2792 : +--------+
>> 2793 :
>> 2794
>> 2795 Imagine the system has the following properties:
>> 2796
>> 2797 (*) an odd-numbered cache line may be in cache A, cache C or it may still be
>> 2798 resident in memory;
>> 2799
>> 2800 (*) an even-numbered cache line may be in cache B, cache D or it may still be
>> 2801 resident in memory;
>> 2802
>> 2803 (*) whilst the CPU core is interrogating one cache, the other cache may be
>> 2804 making use of the bus to access the rest of the system - perhaps to
>> 2805 displace a dirty cacheline or to do a speculative load;
>> 2806
>> 2807 (*) each cache has a queue of operations that need to be applied to that cache
>> 2808 to maintain coherency with the rest of the system;
>> 2809
>> 2810 (*) the coherency queue is not flushed by normal loads to lines already
>> 2811 present in the cache, even though the contents of the queue may
>> 2812 potentially affect those loads.
>> 2813
>> 2814 Imagine, then, that two writes are made on the first CPU, with a write barrier
>> 2815 between them to guarantee that they will appear to reach that CPU's caches in
>> 2816 the requisite order:
>> 2817
>> 2818 CPU 1 CPU 2 COMMENT
>> 2819 =============== =============== =======================================
>> 2820 u == 0, v == 1 and p == &u, q == &u
>> 2821 v = 2;
>> 2822 smp_wmb(); Make sure change to v is visible before
>> 2823 change to p
>> 2824 <A:modify v=2> v is now in cache A exclusively
>> 2825 p = &v;
>> 2826 <B:modify p=&v> p is now in cache B exclusively
>> 2827
>> 2828 The write memory barrier forces the other CPUs in the system to perceive that
>> 2829 the local CPU's caches have apparently been updated in the correct order. But
>> 2830 now imagine that the second CPU wants to read those values:
>> 2831
>> 2832 CPU 1 CPU 2 COMMENT
>> 2833 =============== =============== =======================================
>> 2834 ...
>> 2835 q = p;
>> 2836 x = *q;
>> 2837
>> 2838 The above pair of reads may then fail to happen in the expected order, as the
>> 2839 cacheline holding p may get updated in one of the second CPU's caches whilst
>> 2840 the update to the cacheline holding v is delayed in the other of the second
>> 2841 CPU's caches by some other cache event:
>> 2842
>> 2843 CPU 1 CPU 2 COMMENT
>> 2844 =============== =============== =======================================
>> 2845 u == 0, v == 1 and p == &u, q == &u
>> 2846 v = 2;
>> 2847 smp_wmb();
>> 2848 <A:modify v=2> <C:busy>
>> 2849 <C:queue v=2>
>> 2850 p = &v; q = p;
>> 2851 <D:request p>
>> 2852 <B:modify p=&v> <D:commit p=&v>
>> 2853 <D:read p>
>> 2854 x = *q;
>> 2855 <C:read *q> Reads from v before v updated in cache
>> 2856 <C:unbusy>
>> 2857 <C:commit v=2>
>> 2858
>> 2859 Basically, whilst both cachelines will be updated on CPU 2 eventually, there's
>> 2860 no guarantee that, without intervention, the order of update will be the same
>> 2861 as that committed on CPU 1.
>> 2862
>> 2863
>> 2864 To intervene, we need to interpolate a data dependency barrier or a read
>> 2865 barrier between the loads (which as of v4.15 is supplied unconditionally
>> 2866 by the READ_ONCE() macro). This will force the cache to commit its
>> 2867 coherency queue before processing any further requests:
>> 2868
>> 2869 CPU 1 CPU 2 COMMENT
>> 2870 =============== =============== =======================================
>> 2871 u == 0, v == 1 and p == &u, q == &u
>> 2872 v = 2;
>> 2873 smp_wmb();
>> 2874 <A:modify v=2> <C:busy>
>> 2875 <C:queue v=2>
>> 2876 p = &v; q = p;
>> 2877 <D:request p>
>> 2878 <B:modify p=&v> <D:commit p=&v>
>> 2879 <D:read p>
>> 2880 smp_read_barrier_depends()
>> 2881 <C:unbusy>
>> 2882 <C:commit v=2>
>> 2883 x = *q;
>> 2884 <C:read *q> Reads from v after v updated in cache
>> 2885
>> 2886
>> 2887 This sort of problem can be encountered on DEC Alpha processors as they have a
>> 2888 split cache that improves performance by making better use of the data bus.
>> 2889 Whilst most CPUs do imply a data dependency barrier on the read when a memory
>> 2890 access depends on a read, not all do, so it may not be relied on.
>> 2891
>> 2892 Other CPUs may also have split caches, but must coordinate between the various
>> 2893 cachelets for normal memory accesses. The semantics of the Alpha removes the
>> 2894 need for hardware coordination in the absence of memory barriers, which
>> 2895 permitted Alpha to sport higher CPU clock rates back in the day. However,
>> 2896 please note that (again, as of v4.15) smp_read_barrier_depends() should not
>> 2897 be used except in Alpha arch-specific code and within the READ_ONCE() macro.
>> 2898
>> 2899
2756 CACHE COHERENCY VS DMA 2900 CACHE COHERENCY VS DMA
2757 ---------------------- 2901 ----------------------
2758 2902
2759 Not all systems maintain cache coherency with 2903 Not all systems maintain cache coherency with respect to devices doing DMA. In
2760 such cases, a device attempting DMA may obtai 2904 such cases, a device attempting DMA may obtain stale data from RAM because
2761 dirty cache lines may be resident in the cach 2905 dirty cache lines may be resident in the caches of various CPUs, and may not
2762 have been written back to RAM yet. To deal w 2906 have been written back to RAM yet. To deal with this, the appropriate part of
2763 the kernel must flush the overlapping bits of 2907 the kernel must flush the overlapping bits of cache on each CPU (and maybe
2764 invalidate them as well). 2908 invalidate them as well).
2765 2909
2766 In addition, the data DMA'd to RAM by a devic 2910 In addition, the data DMA'd to RAM by a device may be overwritten by dirty
2767 cache lines being written back to RAM from a 2911 cache lines being written back to RAM from a CPU's cache after the device has
2768 installed its own data, or cache lines presen 2912 installed its own data, or cache lines present in the CPU's cache may simply
2769 obscure the fact that RAM has been updated, u 2913 obscure the fact that RAM has been updated, until at such time as the cacheline
2770 is discarded from the CPU's cache and reloade 2914 is discarded from the CPU's cache and reloaded. To deal with this, the
2771 appropriate part of the kernel must invalidat 2915 appropriate part of the kernel must invalidate the overlapping bits of the
2772 cache on each CPU. 2916 cache on each CPU.
2773 2917
2774 See Documentation/core-api/cachetlb.rst for m !! 2918 See Documentation/core-api/cachetlb.rst for more information on cache management.
2775 management. <<
2776 2919
2777 2920
2778 CACHE COHERENCY VS MMIO 2921 CACHE COHERENCY VS MMIO
2779 ----------------------- 2922 -----------------------
2780 2923
2781 Memory mapped I/O usually takes place through 2924 Memory mapped I/O usually takes place through memory locations that are part of
2782 a window in the CPU's memory space that has d 2925 a window in the CPU's memory space that has different properties assigned than
2783 the usual RAM directed window. 2926 the usual RAM directed window.
2784 2927
2785 Amongst these properties is usually the fact 2928 Amongst these properties is usually the fact that such accesses bypass the
2786 caching entirely and go directly to the devic 2929 caching entirely and go directly to the device buses. This means MMIO accesses
2787 may, in effect, overtake accesses to cached m 2930 may, in effect, overtake accesses to cached memory that were emitted earlier.
2788 A memory barrier isn't sufficient in such a c 2931 A memory barrier isn't sufficient in such a case, but rather the cache must be
2789 flushed between the cached memory write and t 2932 flushed between the cached memory write and the MMIO access if the two are in
2790 any way dependent. 2933 any way dependent.
2791 2934
2792 2935
2793 ========================= 2936 =========================
2794 THE THINGS CPUS GET UP TO 2937 THE THINGS CPUS GET UP TO
2795 ========================= 2938 =========================
2796 2939
2797 A programmer might take it for granted that t 2940 A programmer might take it for granted that the CPU will perform memory
2798 operations in exactly the order specified, so 2941 operations in exactly the order specified, so that if the CPU is, for example,
2799 given the following piece of code to execute: 2942 given the following piece of code to execute:
2800 2943
2801 a = READ_ONCE(*A); 2944 a = READ_ONCE(*A);
2802 WRITE_ONCE(*B, b); 2945 WRITE_ONCE(*B, b);
2803 c = READ_ONCE(*C); 2946 c = READ_ONCE(*C);
2804 d = READ_ONCE(*D); 2947 d = READ_ONCE(*D);
2805 WRITE_ONCE(*E, e); 2948 WRITE_ONCE(*E, e);
2806 2949
2807 they would then expect that the CPU will comp 2950 they would then expect that the CPU will complete the memory operation for each
2808 instruction before moving on to the next one, 2951 instruction before moving on to the next one, leading to a definite sequence of
2809 operations as seen by external observers in t 2952 operations as seen by external observers in the system:
2810 2953
2811 LOAD *A, STORE *B, LOAD *C, LOAD *D, 2954 LOAD *A, STORE *B, LOAD *C, LOAD *D, STORE *E.
2812 2955
2813 2956
2814 Reality is, of course, much messier. With ma 2957 Reality is, of course, much messier. With many CPUs and compilers, the above
2815 assumption doesn't hold because: 2958 assumption doesn't hold because:
2816 2959
2817 (*) loads are more likely to need to be comp 2960 (*) loads are more likely to need to be completed immediately to permit
2818 execution progress, whereas stores can o 2961 execution progress, whereas stores can often be deferred without a
2819 problem; 2962 problem;
2820 2963
2821 (*) loads may be done speculatively, and the 2964 (*) loads may be done speculatively, and the result discarded should it prove
2822 to have been unnecessary; 2965 to have been unnecessary;
2823 2966
2824 (*) loads may be done speculatively, leading 2967 (*) loads may be done speculatively, leading to the result having been fetched
2825 at the wrong time in the expected sequen 2968 at the wrong time in the expected sequence of events;
2826 2969
2827 (*) the order of the memory accesses may be 2970 (*) the order of the memory accesses may be rearranged to promote better use
2828 of the CPU buses and caches; 2971 of the CPU buses and caches;
2829 2972
2830 (*) loads and stores may be combined to impr 2973 (*) loads and stores may be combined to improve performance when talking to
2831 memory or I/O hardware that can do batch 2974 memory or I/O hardware that can do batched accesses of adjacent locations,
2832 thus cutting down on transaction setup c 2975 thus cutting down on transaction setup costs (memory and PCI devices may
2833 both be able to do this); and 2976 both be able to do this); and
2834 2977
2835 (*) the CPU's data cache may affect the orde !! 2978 (*) the CPU's data cache may affect the ordering, and whilst cache-coherency
2836 mechanisms may alleviate this - once the 2979 mechanisms may alleviate this - once the store has actually hit the cache
2837 - there's no guarantee that the coherenc 2980 - there's no guarantee that the coherency management will be propagated in
2838 order to other CPUs. 2981 order to other CPUs.
2839 2982
2840 So what another CPU, say, might actually obse 2983 So what another CPU, say, might actually observe from the above piece of code
2841 is: 2984 is:
2842 2985
2843 LOAD *A, ..., LOAD {*C,*D}, STORE *E, 2986 LOAD *A, ..., LOAD {*C,*D}, STORE *E, STORE *B
2844 2987
2845 (Where "LOAD {*C,*D}" is a combined l 2988 (Where "LOAD {*C,*D}" is a combined load)
2846 2989
2847 2990
2848 However, it is guaranteed that a CPU will be 2991 However, it is guaranteed that a CPU will be self-consistent: it will see its
2849 _own_ accesses appear to be correctly ordered 2992 _own_ accesses appear to be correctly ordered, without the need for a memory
2850 barrier. For instance with the following cod 2993 barrier. For instance with the following code:
2851 2994
2852 U = READ_ONCE(*A); 2995 U = READ_ONCE(*A);
2853 WRITE_ONCE(*A, V); 2996 WRITE_ONCE(*A, V);
2854 WRITE_ONCE(*A, W); 2997 WRITE_ONCE(*A, W);
2855 X = READ_ONCE(*A); 2998 X = READ_ONCE(*A);
2856 WRITE_ONCE(*A, Y); 2999 WRITE_ONCE(*A, Y);
2857 Z = READ_ONCE(*A); 3000 Z = READ_ONCE(*A);
2858 3001
2859 and assuming no intervention by an external i 3002 and assuming no intervention by an external influence, it can be assumed that
2860 the final result will appear to be: 3003 the final result will appear to be:
2861 3004
2862 U == the original value of *A 3005 U == the original value of *A
2863 X == W 3006 X == W
2864 Z == Y 3007 Z == Y
2865 *A == Y 3008 *A == Y
2866 3009
2867 The code above may cause the CPU to generate 3010 The code above may cause the CPU to generate the full sequence of memory
2868 accesses: 3011 accesses:
2869 3012
2870 U=LOAD *A, STORE *A=V, STORE *A=W, X= 3013 U=LOAD *A, STORE *A=V, STORE *A=W, X=LOAD *A, STORE *A=Y, Z=LOAD *A
2871 3014
2872 in that order, but, without intervention, the 3015 in that order, but, without intervention, the sequence may have almost any
2873 combination of elements combined or discarded 3016 combination of elements combined or discarded, provided the program's view
2874 of the world remains consistent. Note that R 3017 of the world remains consistent. Note that READ_ONCE() and WRITE_ONCE()
2875 are -not- optional in the above example, as t 3018 are -not- optional in the above example, as there are architectures
2876 where a given CPU might reorder successive lo 3019 where a given CPU might reorder successive loads to the same location.
2877 On such architectures, READ_ONCE() and WRITE_ 3020 On such architectures, READ_ONCE() and WRITE_ONCE() do whatever is
2878 necessary to prevent this, for example, on It 3021 necessary to prevent this, for example, on Itanium the volatile casts
2879 used by READ_ONCE() and WRITE_ONCE() cause GC 3022 used by READ_ONCE() and WRITE_ONCE() cause GCC to emit the special ld.acq
2880 and st.rel instructions (respectively) that p 3023 and st.rel instructions (respectively) that prevent such reordering.
2881 3024
2882 The compiler may also combine, discard or def 3025 The compiler may also combine, discard or defer elements of the sequence before
2883 the CPU even sees them. 3026 the CPU even sees them.
2884 3027
2885 For instance: 3028 For instance:
2886 3029
2887 *A = V; 3030 *A = V;
2888 *A = W; 3031 *A = W;
2889 3032
2890 may be reduced to: 3033 may be reduced to:
2891 3034
2892 *A = W; 3035 *A = W;
2893 3036
2894 since, without either a write barrier or an W 3037 since, without either a write barrier or an WRITE_ONCE(), it can be
2895 assumed that the effect of the storage of V t 3038 assumed that the effect of the storage of V to *A is lost. Similarly:
2896 3039
2897 *A = Y; 3040 *A = Y;
2898 Z = *A; 3041 Z = *A;
2899 3042
2900 may, without a memory barrier or an READ_ONCE 3043 may, without a memory barrier or an READ_ONCE() and WRITE_ONCE(), be
2901 reduced to: 3044 reduced to:
2902 3045
2903 *A = Y; 3046 *A = Y;
2904 Z = Y; 3047 Z = Y;
2905 3048
2906 and the LOAD operation never appear outside o 3049 and the LOAD operation never appear outside of the CPU.
2907 3050
2908 3051
2909 AND THEN THERE'S THE ALPHA 3052 AND THEN THERE'S THE ALPHA
2910 -------------------------- 3053 --------------------------
2911 3054
2912 The DEC Alpha CPU is one of the most relaxed 3055 The DEC Alpha CPU is one of the most relaxed CPUs there is. Not only that,
2913 some versions of the Alpha CPU have a split d 3056 some versions of the Alpha CPU have a split data cache, permitting them to have
2914 two semantically-related cache lines updated 3057 two semantically-related cache lines updated at separate times. This is where
2915 the address-dependency barrier really becomes !! 3058 the data dependency barrier really becomes necessary as this synchronises both
2916 both caches with the memory coherence system, !! 3059 caches with the memory coherence system, thus making it seem like pointer
2917 changes vs new data occur in the right order. 3060 changes vs new data occur in the right order.
2918 3061
2919 The Alpha defines the Linux kernel's memory m 3062 The Alpha defines the Linux kernel's memory model, although as of v4.15
2920 the Linux kernel's addition of smp_mb() to RE !! 3063 the Linux kernel's addition of smp_read_barrier_depends() to READ_ONCE()
2921 reduced its impact on the memory model. !! 3064 greatly reduced Alpha's impact on the memory model.
>> 3065
>> 3066 See the subsection on "Cache Coherency" above.
2922 3067
2923 3068
2924 VIRTUAL MACHINE GUESTS 3069 VIRTUAL MACHINE GUESTS
2925 ---------------------- 3070 ----------------------
2926 3071
2927 Guests running within virtual machines might 3072 Guests running within virtual machines might be affected by SMP effects even if
2928 the guest itself is compiled without SMP supp 3073 the guest itself is compiled without SMP support. This is an artifact of
2929 interfacing with an SMP host while running an 3074 interfacing with an SMP host while running an UP kernel. Using mandatory
2930 barriers for this use-case would be possible 3075 barriers for this use-case would be possible but is often suboptimal.
2931 3076
2932 To handle this case optimally, low-level virt 3077 To handle this case optimally, low-level virt_mb() etc macros are available.
2933 These have the same effect as smp_mb() etc wh 3078 These have the same effect as smp_mb() etc when SMP is enabled, but generate
2934 identical code for SMP and non-SMP systems. 3079 identical code for SMP and non-SMP systems. For example, virtual machine guests
2935 should use virt_mb() rather than smp_mb() whe 3080 should use virt_mb() rather than smp_mb() when synchronizing against a
2936 (possibly SMP) host. 3081 (possibly SMP) host.
2937 3082
2938 These are equivalent to smp_mb() etc counterp 3083 These are equivalent to smp_mb() etc counterparts in all other respects,
2939 in particular, they do not control MMIO effec 3084 in particular, they do not control MMIO effects: to control
2940 MMIO effects, use mandatory barriers. 3085 MMIO effects, use mandatory barriers.
2941 3086
2942 3087
2943 ============ 3088 ============
2944 EXAMPLE USES 3089 EXAMPLE USES
2945 ============ 3090 ============
2946 3091
2947 CIRCULAR BUFFERS 3092 CIRCULAR BUFFERS
2948 ---------------- 3093 ----------------
2949 3094
2950 Memory barriers can be used to implement circ 3095 Memory barriers can be used to implement circular buffering without the need
2951 of a lock to serialise the producer with the 3096 of a lock to serialise the producer with the consumer. See:
2952 3097
2953 Documentation/core-api/circular-buffe 3098 Documentation/core-api/circular-buffers.rst
2954 3099
2955 for details. 3100 for details.
2956 3101
2957 3102
2958 ========== 3103 ==========
2959 REFERENCES 3104 REFERENCES
2960 ========== 3105 ==========
2961 3106
2962 Alpha AXP Architecture Reference Manual, Seco 3107 Alpha AXP Architecture Reference Manual, Second Edition (Sites & Witek,
2963 Digital Press) 3108 Digital Press)
2964 Chapter 5.2: Physical Address Space C 3109 Chapter 5.2: Physical Address Space Characteristics
2965 Chapter 5.4: Caches and Write Buffers 3110 Chapter 5.4: Caches and Write Buffers
2966 Chapter 5.5: Data Sharing 3111 Chapter 5.5: Data Sharing
2967 Chapter 5.6: Read/Write Ordering 3112 Chapter 5.6: Read/Write Ordering
2968 3113
2969 AMD64 Architecture Programmer's Manual Volume 3114 AMD64 Architecture Programmer's Manual Volume 2: System Programming
2970 Chapter 7.1: Memory-Access Ordering 3115 Chapter 7.1: Memory-Access Ordering
2971 Chapter 7.4: Buffering and Combining 3116 Chapter 7.4: Buffering and Combining Memory Writes
2972 3117
2973 ARM Architecture Reference Manual (ARMv8, for 3118 ARM Architecture Reference Manual (ARMv8, for ARMv8-A architecture profile)
2974 Chapter B2: The AArch64 Application L 3119 Chapter B2: The AArch64 Application Level Memory Model
2975 3120
2976 IA-32 Intel Architecture Software Developer's 3121 IA-32 Intel Architecture Software Developer's Manual, Volume 3:
2977 System Programming Guide 3122 System Programming Guide
2978 Chapter 7.1: Locked Atomic Operations 3123 Chapter 7.1: Locked Atomic Operations
2979 Chapter 7.2: Memory Ordering 3124 Chapter 7.2: Memory Ordering
2980 Chapter 7.4: Serializing Instructions 3125 Chapter 7.4: Serializing Instructions
2981 3126
2982 The SPARC Architecture Manual, Version 9 3127 The SPARC Architecture Manual, Version 9
2983 Chapter 8: Memory Models 3128 Chapter 8: Memory Models
2984 Appendix D: Formal Specification of t 3129 Appendix D: Formal Specification of the Memory Models
2985 Appendix J: Programming with the Memo 3130 Appendix J: Programming with the Memory Models
2986 3131
2987 Storage in the PowerPC (Stone and Fitzgerald) 3132 Storage in the PowerPC (Stone and Fitzgerald)
2988 3133
2989 UltraSPARC Programmer Reference Manual 3134 UltraSPARC Programmer Reference Manual
2990 Chapter 5: Memory Accesses and Cachea 3135 Chapter 5: Memory Accesses and Cacheability
2991 Chapter 15: Sparc-V9 Memory Models 3136 Chapter 15: Sparc-V9 Memory Models
2992 3137
2993 UltraSPARC III Cu User's Manual 3138 UltraSPARC III Cu User's Manual
2994 Chapter 9: Memory Models 3139 Chapter 9: Memory Models
2995 3140
2996 UltraSPARC IIIi Processor User's Manual 3141 UltraSPARC IIIi Processor User's Manual
2997 Chapter 8: Memory Models 3142 Chapter 8: Memory Models
2998 3143
2999 UltraSPARC Architecture 2005 3144 UltraSPARC Architecture 2005
3000 Chapter 9: Memory 3145 Chapter 9: Memory
3001 Appendix D: Formal Specifications of 3146 Appendix D: Formal Specifications of the Memory Models
3002 3147
3003 UltraSPARC T1 Supplement to the UltraSPARC Ar 3148 UltraSPARC T1 Supplement to the UltraSPARC Architecture 2005
3004 Chapter 8: Memory Models 3149 Chapter 8: Memory Models
3005 Appendix F: Caches and Cache Coherenc 3150 Appendix F: Caches and Cache Coherency
3006 3151
3007 Solaris Internals, Core Kernel Architecture, 3152 Solaris Internals, Core Kernel Architecture, p63-68:
3008 Chapter 3.3: Hardware Considerations 3153 Chapter 3.3: Hardware Considerations for Locks and
3009 Synchronization 3154 Synchronization
3010 3155
3011 Unix Systems for Modern Architectures, Symmet 3156 Unix Systems for Modern Architectures, Symmetric Multiprocessing and Caching
3012 for Kernel Programmers: 3157 for Kernel Programmers:
3013 Chapter 13: Other Memory Models 3158 Chapter 13: Other Memory Models
3014 3159
3015 Intel Itanium Architecture Software Developer 3160 Intel Itanium Architecture Software Developer's Manual: Volume 1:
3016 Section 2.6: Speculation 3161 Section 2.6: Speculation
3017 Section 4.4: Memory Access 3162 Section 4.4: Memory Access
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.