1 .. SPDX-License-Identifier: GPL-2.0 OR GFDL-1. 1 .. SPDX-License-Identifier: GPL-2.0 OR GFDL-1.2-no-invariants-only
2 2
3 =========================== 3 ===========================
4 Lockless Ring Buffer Design 4 Lockless Ring Buffer Design
5 =========================== 5 ===========================
6 6
7 Copyright 2009 Red Hat Inc. 7 Copyright 2009 Red Hat Inc.
8 8
9 :Author: Steven Rostedt <srostedt@redhat.com> 9 :Author: Steven Rostedt <srostedt@redhat.com>
10 :License: The GNU Free Documentation License, 10 :License: The GNU Free Documentation License, Version 1.2
11 (dual licensed under the GPL v2) 11 (dual licensed under the GPL v2)
12 :Reviewers: Mathieu Desnoyers, Huang Ying, Hi 12 :Reviewers: Mathieu Desnoyers, Huang Ying, Hidetoshi Seto,
13 and Frederic Weisbecker. 13 and Frederic Weisbecker.
14 14
15 15
16 Written for: 2.6.31 16 Written for: 2.6.31
17 17
18 Terminology used in this Document 18 Terminology used in this Document
19 --------------------------------- 19 ---------------------------------
20 20
21 tail 21 tail
22 - where new writes happen in the ring 22 - where new writes happen in the ring buffer.
23 23
24 head 24 head
25 - where new reads happen in the ring b 25 - where new reads happen in the ring buffer.
26 26
27 producer 27 producer
28 - the task that writes into the ring b 28 - the task that writes into the ring buffer (same as writer)
29 29
30 writer 30 writer
31 - same as producer 31 - same as producer
32 32
33 consumer 33 consumer
34 - the task that reads from the buffer 34 - the task that reads from the buffer (same as reader)
35 35
36 reader 36 reader
37 - same as consumer. 37 - same as consumer.
38 38
39 reader_page 39 reader_page
40 - A page outside the ring buffer used 40 - A page outside the ring buffer used solely (for the most part)
41 by the reader. 41 by the reader.
42 42
43 head_page 43 head_page
44 - a pointer to the page that the reade 44 - a pointer to the page that the reader will use next
45 45
46 tail_page 46 tail_page
47 - a pointer to the page that will be w 47 - a pointer to the page that will be written to next
48 48
49 commit_page 49 commit_page
50 - a pointer to the page with the last 50 - a pointer to the page with the last finished non-nested write.
51 51
52 cmpxchg 52 cmpxchg
53 - hardware-assisted atomic transaction 53 - hardware-assisted atomic transaction that performs the following::
54 54
55 A = B if previous A == C 55 A = B if previous A == C
56 56
57 R = cmpxchg(A, C, B) is saying tha 57 R = cmpxchg(A, C, B) is saying that we replace A with B if and only
58 if current A is equal to C, an 58 if current A is equal to C, and we put the old (current)
59 A into R 59 A into R
60 60
61 R gets the previous A regardless i 61 R gets the previous A regardless if A is updated with B or not.
62 62
63 To see if the update was successful 63 To see if the update was successful a compare of ``R == C``
64 may be used. 64 may be used.
65 65
66 The Generic Ring Buffer 66 The Generic Ring Buffer
67 ----------------------- 67 -----------------------
68 68
69 The ring buffer can be used in either an overw 69 The ring buffer can be used in either an overwrite mode or in
70 producer/consumer mode. 70 producer/consumer mode.
71 71
72 Producer/consumer mode is where if the produce 72 Producer/consumer mode is where if the producer were to fill up the
73 buffer before the consumer could free up anyth 73 buffer before the consumer could free up anything, the producer
74 will stop writing to the buffer. This will los 74 will stop writing to the buffer. This will lose most recent events.
75 75
76 Overwrite mode is where if the producer were t 76 Overwrite mode is where if the producer were to fill up the buffer
77 before the consumer could free up anything, th 77 before the consumer could free up anything, the producer will
78 overwrite the older data. This will lose the o 78 overwrite the older data. This will lose the oldest events.
79 79
80 No two writers can write at the same time (on 80 No two writers can write at the same time (on the same per-cpu buffer),
81 but a writer may interrupt another writer, but 81 but a writer may interrupt another writer, but it must finish writing
82 before the previous writer may continue. This 82 before the previous writer may continue. This is very important to the
83 algorithm. The writers act like a "stack". The 83 algorithm. The writers act like a "stack". The way interrupts works
84 enforces this behavior:: 84 enforces this behavior::
85 85
86 86
87 writer1 start 87 writer1 start
88 <preempted> writer2 start 88 <preempted> writer2 start
89 <preempted> writer3 start 89 <preempted> writer3 start
90 writer3 finishes 90 writer3 finishes
91 writer2 finishes 91 writer2 finishes
92 writer1 finishes 92 writer1 finishes
93 93
94 This is very much like a writer being preempte 94 This is very much like a writer being preempted by an interrupt and
95 the interrupt doing a write as well. 95 the interrupt doing a write as well.
96 96
97 Readers can happen at any time. But no two rea 97 Readers can happen at any time. But no two readers may run at the
98 same time, nor can a reader preempt/interrupt 98 same time, nor can a reader preempt/interrupt another reader. A reader
99 cannot preempt/interrupt a writer, but it may 99 cannot preempt/interrupt a writer, but it may read/consume from the
100 buffer at the same time as a writer is writing 100 buffer at the same time as a writer is writing, but the reader must be
101 on another processor to do so. A reader may re 101 on another processor to do so. A reader may read on its own processor
102 and can be preempted by a writer. 102 and can be preempted by a writer.
103 103
104 A writer can preempt a reader, but a reader ca 104 A writer can preempt a reader, but a reader cannot preempt a writer.
105 But a reader can read the buffer at the same t 105 But a reader can read the buffer at the same time (on another processor)
106 as a writer. 106 as a writer.
107 107
108 The ring buffer is made up of a list of pages 108 The ring buffer is made up of a list of pages held together by a linked list.
109 109
110 At initialization a reader page is allocated f 110 At initialization a reader page is allocated for the reader that is not
111 part of the ring buffer. 111 part of the ring buffer.
112 112
113 The head_page, tail_page and commit_page are a 113 The head_page, tail_page and commit_page are all initialized to point
114 to the same page. 114 to the same page.
115 115
116 The reader page is initialized to have its nex 116 The reader page is initialized to have its next pointer pointing to
117 the head page, and its previous pointer pointi 117 the head page, and its previous pointer pointing to a page before
118 the head page. 118 the head page.
119 119
120 The reader has its own page to use. At start u 120 The reader has its own page to use. At start up time, this page is
121 allocated but is not attached to the list. Whe 121 allocated but is not attached to the list. When the reader wants
122 to read from the buffer, if its page is empty 122 to read from the buffer, if its page is empty (like it is on start-up),
123 it will swap its page with the head_page. The 123 it will swap its page with the head_page. The old reader page will
124 become part of the ring buffer and the head_pa 124 become part of the ring buffer and the head_page will be removed.
125 The page after the inserted page (old reader_p 125 The page after the inserted page (old reader_page) will become the
126 new head page. 126 new head page.
127 127
128 Once the new page is given to the reader, the 128 Once the new page is given to the reader, the reader could do what
129 it wants with it, as long as a writer has left 129 it wants with it, as long as a writer has left that page.
130 130
131 A sample of how the reader page is swapped: No 131 A sample of how the reader page is swapped: Note this does not
132 show the head page in the buffer, it is for de 132 show the head page in the buffer, it is for demonstrating a swap
133 only. 133 only.
134 134
135 :: 135 ::
136 136
137 +------+ 137 +------+
138 |reader| RING BUFFER 138 |reader| RING BUFFER
139 |page | 139 |page |
140 +------+ 140 +------+
141 +---+ +---+ +---+ 141 +---+ +---+ +---+
142 | |-->| |-->| | 142 | |-->| |-->| |
143 | |<--| |<--| | 143 | |<--| |<--| |
144 +---+ +---+ +---+ 144 +---+ +---+ +---+
145 ^ | ^ | 145 ^ | ^ |
146 | +-------------+ | 146 | +-------------+ |
147 +-----------------+ 147 +-----------------+
148 148
149 149
150 +------+ 150 +------+
151 |reader| RING BUFFER 151 |reader| RING BUFFER
152 |page |-------------------+ 152 |page |-------------------+
153 +------+ v 153 +------+ v
154 | +---+ +---+ +---+ 154 | +---+ +---+ +---+
155 | | |-->| |-->| | 155 | | |-->| |-->| |
156 | | |<--| |<--| |<-+ 156 | | |<--| |<--| |<-+
157 | +---+ +---+ +---+ | 157 | +---+ +---+ +---+ |
158 | ^ | ^ | | 158 | ^ | ^ | |
159 | | +-------------+ | | 159 | | +-------------+ | |
160 | +-----------------+ | 160 | +-----------------+ |
161 +------------------------------------+ 161 +------------------------------------+
162 162
163 +------+ 163 +------+
164 |reader| RING BUFFER 164 |reader| RING BUFFER
165 |page |-------------------+ 165 |page |-------------------+
166 +------+ <---------------+ v 166 +------+ <---------------+ v
167 | ^ +---+ +---+ +---+ 167 | ^ +---+ +---+ +---+
168 | | | |-->| |-->| | 168 | | | |-->| |-->| |
169 | | | | | |<--| |<-+ 169 | | | | | |<--| |<-+
170 | | +---+ +---+ +---+ | 170 | | +---+ +---+ +---+ |
171 | | | ^ | | 171 | | | ^ | |
172 | | +-------------+ | | 172 | | +-------------+ | |
173 | +-----------------------------+ | 173 | +-----------------------------+ |
174 +------------------------------------+ 174 +------------------------------------+
175 175
176 +------+ 176 +------+
177 |buffer| RING BUFFER 177 |buffer| RING BUFFER
178 |page |-------------------+ 178 |page |-------------------+
179 +------+ <---------------+ v 179 +------+ <---------------+ v
180 | ^ +---+ +---+ +---+ 180 | ^ +---+ +---+ +---+
181 | | | | | |-->| | 181 | | | | | |-->| |
182 | | New | | | |<--| |<-+ 182 | | New | | | |<--| |<-+
183 | | Reader +---+ +---+ +---+ | 183 | | Reader +---+ +---+ +---+ |
184 | | page ----^ | | 184 | | page ----^ | |
185 | | | | 185 | | | |
186 | +-----------------------------+ | 186 | +-----------------------------+ |
187 +------------------------------------+ 187 +------------------------------------+
188 188
189 189
190 190
191 It is possible that the page swapped is the co 191 It is possible that the page swapped is the commit page and the tail page,
192 if what is in the ring buffer is less than wha 192 if what is in the ring buffer is less than what is held in a buffer page.
193 193
194 :: 194 ::
195 195
196 reader page commit page tail 196 reader page commit page tail page
197 | | | 197 | | |
198 v | | 198 v | |
199 +---+ | | 199 +---+ | |
200 | |<----------+ | 200 | |<----------+ |
201 | |<------------------------+ 201 | |<------------------------+
202 | |------+ 202 | |------+
203 +---+ | 203 +---+ |
204 | 204 |
205 v 205 v
206 +---+ +---+ +---+ +---+ 206 +---+ +---+ +---+ +---+
207 <---| |--->| |--->| |--->| |---> 207 <---| |--->| |--->| |--->| |--->
208 --->| |<---| |<---| |<---| |<--- 208 --->| |<---| |<---| |<---| |<---
209 +---+ +---+ +---+ +---+ 209 +---+ +---+ +---+ +---+
210 210
211 This case is still valid for this algorithm. 211 This case is still valid for this algorithm.
212 When the writer leaves the page, it simply goe 212 When the writer leaves the page, it simply goes into the ring buffer
213 since the reader page still points to the next 213 since the reader page still points to the next location in the ring
214 buffer. 214 buffer.
215 215
216 216
217 The main pointers: 217 The main pointers:
218 218
219 reader page 219 reader page
220 - The page used solely by the read 220 - The page used solely by the reader and is not part
221 of the ring buffer (may be swapp 221 of the ring buffer (may be swapped in)
222 222
223 head page 223 head page
224 - the next page in the ring buffer 224 - the next page in the ring buffer that will be swapped
225 with the reader page. 225 with the reader page.
226 226
227 tail page 227 tail page
228 - the page where the next write wi 228 - the page where the next write will take place.
229 229
230 commit page 230 commit page
231 - the page that last finished a wr 231 - the page that last finished a write.
232 232
233 The commit page only is updated by the outermo 233 The commit page only is updated by the outermost writer in the
234 writer stack. A writer that preempts another w 234 writer stack. A writer that preempts another writer will not move the
235 commit page. 235 commit page.
236 236
237 When data is written into the ring buffer, a p 237 When data is written into the ring buffer, a position is reserved
238 in the ring buffer and passed back to the writ 238 in the ring buffer and passed back to the writer. When the writer
239 is finished writing data into that position, i 239 is finished writing data into that position, it commits the write.
240 240
241 Another write (or a read) may take place at an 241 Another write (or a read) may take place at anytime during this
242 transaction. If another write happens it must 242 transaction. If another write happens it must finish before continuing
243 with the previous write. 243 with the previous write.
244 244
245 245
246 Write reserve:: 246 Write reserve::
247 247
248 Buffer page 248 Buffer page
249 +---------+ 249 +---------+
250 |written | 250 |written |
251 +---------+ <--- given back to writer ( 251 +---------+ <--- given back to writer (current commit)
252 |reserved | 252 |reserved |
253 +---------+ <--- tail pointer 253 +---------+ <--- tail pointer
254 | empty | 254 | empty |
255 +---------+ 255 +---------+
256 256
257 Write commit:: 257 Write commit::
258 258
259 Buffer page 259 Buffer page
260 +---------+ 260 +---------+
261 |written | 261 |written |
262 +---------+ 262 +---------+
263 |written | 263 |written |
264 +---------+ <--- next position for writ 264 +---------+ <--- next position for write (current commit)
265 | empty | 265 | empty |
266 +---------+ 266 +---------+
267 267
268 268
269 If a write happens after the first reserve:: 269 If a write happens after the first reserve::
270 270
271 Buffer page 271 Buffer page
272 +---------+ 272 +---------+
273 |written | 273 |written |
274 +---------+ <-- current commit 274 +---------+ <-- current commit
275 |reserved | 275 |reserved |
276 +---------+ <--- given back to second w 276 +---------+ <--- given back to second writer
277 |reserved | 277 |reserved |
278 +---------+ <--- tail pointer 278 +---------+ <--- tail pointer
279 279
280 After second writer commits:: 280 After second writer commits::
281 281
282 282
283 Buffer page 283 Buffer page
284 +---------+ 284 +---------+
285 |written | 285 |written |
286 +---------+ <--(last full commit) 286 +---------+ <--(last full commit)
287 |reserved | 287 |reserved |
288 +---------+ 288 +---------+
289 |pending | 289 |pending |
290 |commit | 290 |commit |
291 +---------+ <--- tail pointer 291 +---------+ <--- tail pointer
292 292
293 When the first writer commits:: 293 When the first writer commits::
294 294
295 Buffer page 295 Buffer page
296 +---------+ 296 +---------+
297 |written | 297 |written |
298 +---------+ 298 +---------+
299 |written | 299 |written |
300 +---------+ 300 +---------+
301 |written | 301 |written |
302 +---------+ <--(last full commit and ta 302 +---------+ <--(last full commit and tail pointer)
303 303
304 304
305 The commit pointer points to the last write lo 305 The commit pointer points to the last write location that was
306 committed without preempting another write. Wh 306 committed without preempting another write. When a write that
307 preempted another write is committed, it only 307 preempted another write is committed, it only becomes a pending commit
308 and will not be a full commit until all writes 308 and will not be a full commit until all writes have been committed.
309 309
310 The commit page points to the page that has th 310 The commit page points to the page that has the last full commit.
311 The tail page points to the page with the last 311 The tail page points to the page with the last write (before
312 committing). 312 committing).
313 313
314 The tail page is always equal to or after the 314 The tail page is always equal to or after the commit page. It may
315 be several pages ahead. If the tail page catch 315 be several pages ahead. If the tail page catches up to the commit
316 page then no more writes may take place (regar 316 page then no more writes may take place (regardless of the mode
317 of the ring buffer: overwrite and produce/cons 317 of the ring buffer: overwrite and produce/consumer).
318 318
319 The order of pages is:: 319 The order of pages is::
320 320
321 head page 321 head page
322 commit page 322 commit page
323 tail page 323 tail page
324 324
325 Possible scenario:: 325 Possible scenario::
326 326
327 tail page 327 tail page
328 head page commit page | 328 head page commit page |
329 | | | 329 | | |
330 v v v 330 v v v
331 +---+ +---+ +---+ +---+ 331 +---+ +---+ +---+ +---+
332 <---| |--->| |--->| |--->| |---> 332 <---| |--->| |--->| |--->| |--->
333 --->| |<---| |<---| |<---| |<--- 333 --->| |<---| |<---| |<---| |<---
334 +---+ +---+ +---+ +---+ 334 +---+ +---+ +---+ +---+
335 335
336 There is a special case that the head page is 336 There is a special case that the head page is after either the commit page
337 and possibly the tail page. That is when the c 337 and possibly the tail page. That is when the commit (and tail) page has been
338 swapped with the reader page. This is because 338 swapped with the reader page. This is because the head page is always
339 part of the ring buffer, but the reader page i 339 part of the ring buffer, but the reader page is not. Whenever there
340 has been less than a full page that has been c 340 has been less than a full page that has been committed inside the ring buffer,
341 and a reader swaps out a page, it will be swap 341 and a reader swaps out a page, it will be swapping out the commit page.
342 342
343 :: 343 ::
344 344
345 reader page commit page tail 345 reader page commit page tail page
346 | | | 346 | | |
347 v | | 347 v | |
348 +---+ | | 348 +---+ | |
349 | |<----------+ | 349 | |<----------+ |
350 | |<------------------------+ 350 | |<------------------------+
351 | |------+ 351 | |------+
352 +---+ | 352 +---+ |
353 | 353 |
354 v 354 v
355 +---+ +---+ +---+ +---+ 355 +---+ +---+ +---+ +---+
356 <---| |--->| |--->| |--->| |---> 356 <---| |--->| |--->| |--->| |--->
357 --->| |<---| |<---| |<---| |<--- 357 --->| |<---| |<---| |<---| |<---
358 +---+ +---+ +---+ +---+ 358 +---+ +---+ +---+ +---+
359 ^ 359 ^
360 | 360 |
361 head page 361 head page
362 362
363 363
364 In this case, the head page will not move when 364 In this case, the head page will not move when the tail and commit
365 move back into the ring buffer. 365 move back into the ring buffer.
366 366
367 The reader cannot swap a page into the ring bu 367 The reader cannot swap a page into the ring buffer if the commit page
368 is still on that page. If the read meets the l 368 is still on that page. If the read meets the last commit (real commit
369 not pending or reserved), then there is nothin 369 not pending or reserved), then there is nothing more to read.
370 The buffer is considered empty until another f 370 The buffer is considered empty until another full commit finishes.
371 371
372 When the tail meets the head page, if the buff 372 When the tail meets the head page, if the buffer is in overwrite mode,
373 the head page will be pushed ahead one. If the 373 the head page will be pushed ahead one. If the buffer is in producer/consumer
374 mode, the write will fail. 374 mode, the write will fail.
375 375
376 Overwrite mode:: 376 Overwrite mode::
377 377
378 tail page 378 tail page
379 | 379 |
380 v 380 v
381 +---+ +---+ +---+ +---+ 381 +---+ +---+ +---+ +---+
382 <---| |--->| |--->| |--->| |---> 382 <---| |--->| |--->| |--->| |--->
383 --->| |<---| |<---| |<---| |<--- 383 --->| |<---| |<---| |<---| |<---
384 +---+ +---+ +---+ +---+ 384 +---+ +---+ +---+ +---+
385 ^ 385 ^
386 | 386 |
387 head page 387 head page
388 388
389 389
390 tail page 390 tail page
391 | 391 |
392 v 392 v
393 +---+ +---+ +---+ +---+ 393 +---+ +---+ +---+ +---+
394 <---| |--->| |--->| |--->| |---> 394 <---| |--->| |--->| |--->| |--->
395 --->| |<---| |<---| |<---| |<--- 395 --->| |<---| |<---| |<---| |<---
396 +---+ +---+ +---+ +---+ 396 +---+ +---+ +---+ +---+
397 ^ 397 ^
398 | 398 |
399 head page 399 head page
400 400
401 401
402 tail page 402 tail page
403 | 403 |
404 v 404 v
405 +---+ +---+ +---+ +---+ 405 +---+ +---+ +---+ +---+
406 <---| |--->| |--->| |--->| |---> 406 <---| |--->| |--->| |--->| |--->
407 --->| |<---| |<---| |<---| |<--- 407 --->| |<---| |<---| |<---| |<---
408 +---+ +---+ +---+ +---+ 408 +---+ +---+ +---+ +---+
409 ^ 409 ^
410 | 410 |
411 head page 411 head page
412 412
413 Note, the reader page will still point to the 413 Note, the reader page will still point to the previous head page.
414 But when a swap takes place, it will use the m 414 But when a swap takes place, it will use the most recent head page.
415 415
416 416
417 Making the Ring Buffer Lockless: 417 Making the Ring Buffer Lockless:
418 -------------------------------- 418 --------------------------------
419 419
420 The main idea behind the lockless algorithm is 420 The main idea behind the lockless algorithm is to combine the moving
421 of the head_page pointer with the swapping of 421 of the head_page pointer with the swapping of pages with the reader.
422 State flags are placed inside the pointer to t 422 State flags are placed inside the pointer to the page. To do this,
423 each page must be aligned in memory by 4 bytes 423 each page must be aligned in memory by 4 bytes. This will allow the 2
424 least significant bits of the address to be us 424 least significant bits of the address to be used as flags, since
425 they will always be zero for the address. To g 425 they will always be zero for the address. To get the address,
426 simply mask out the flags:: 426 simply mask out the flags::
427 427
428 MASK = ~3 428 MASK = ~3
429 429
430 address & MASK 430 address & MASK
431 431
432 Two flags will be kept by these two bits: 432 Two flags will be kept by these two bits:
433 433
434 HEADER 434 HEADER
435 - the page being pointed to is a head 435 - the page being pointed to is a head page
436 436
437 UPDATE 437 UPDATE
438 - the page being pointed to is being u 438 - the page being pointed to is being updated by a writer
439 and was or is about to be a head pag 439 and was or is about to be a head page.
440 440
441 :: 441 ::
442 442
443 reader page 443 reader page
444 | 444 |
445 v 445 v
446 +---+ 446 +---+
447 | |------+ 447 | |------+
448 +---+ | 448 +---+ |
449 | 449 |
450 v 450 v
451 +---+ +---+ +---+ +---+ 451 +---+ +---+ +---+ +---+
452 <---| |--->| |-H->| |--->| |---> 452 <---| |--->| |-H->| |--->| |--->
453 --->| |<---| |<---| |<---| |<--- 453 --->| |<---| |<---| |<---| |<---
454 +---+ +---+ +---+ +---+ 454 +---+ +---+ +---+ +---+
455 455
456 456
457 The above pointer "-H->" would have the HEADER 457 The above pointer "-H->" would have the HEADER flag set. That is
458 the next page is the next page to be swapped o 458 the next page is the next page to be swapped out by the reader.
459 This pointer means the next page is the head p 459 This pointer means the next page is the head page.
460 460
461 When the tail page meets the head pointer, it 461 When the tail page meets the head pointer, it will use cmpxchg to
462 change the pointer to the UPDATE state:: 462 change the pointer to the UPDATE state::
463 463
464 464
465 tail page 465 tail page
466 | 466 |
467 v 467 v
468 +---+ +---+ +---+ +---+ 468 +---+ +---+ +---+ +---+
469 <---| |--->| |-H->| |--->| |---> 469 <---| |--->| |-H->| |--->| |--->
470 --->| |<---| |<---| |<---| |<--- 470 --->| |<---| |<---| |<---| |<---
471 +---+ +---+ +---+ +---+ 471 +---+ +---+ +---+ +---+
472 472
473 tail page 473 tail page
474 | 474 |
475 v 475 v
476 +---+ +---+ +---+ +---+ 476 +---+ +---+ +---+ +---+
477 <---| |--->| |-U->| |--->| |---> 477 <---| |--->| |-U->| |--->| |--->
478 --->| |<---| |<---| |<---| |<--- 478 --->| |<---| |<---| |<---| |<---
479 +---+ +---+ +---+ +---+ 479 +---+ +---+ +---+ +---+
480 480
481 "-U->" represents a pointer in the UPDATE stat 481 "-U->" represents a pointer in the UPDATE state.
482 482
483 Any access to the reader will need to take som 483 Any access to the reader will need to take some sort of lock to serialize
484 the readers. But the writers will never take a 484 the readers. But the writers will never take a lock to write to the
485 ring buffer. This means we only need to worry 485 ring buffer. This means we only need to worry about a single reader,
486 and writes only preempt in "stack" formation. 486 and writes only preempt in "stack" formation.
487 487
488 When the reader tries to swap the page with th 488 When the reader tries to swap the page with the ring buffer, it
489 will also use cmpxchg. If the flag bit in the 489 will also use cmpxchg. If the flag bit in the pointer to the
490 head page does not have the HEADER flag set, t 490 head page does not have the HEADER flag set, the compare will fail
491 and the reader will need to look for the new h 491 and the reader will need to look for the new head page and try again.
492 Note, the flags UPDATE and HEADER are never se 492 Note, the flags UPDATE and HEADER are never set at the same time.
493 493
494 The reader swaps the reader page as follows:: 494 The reader swaps the reader page as follows::
495 495
496 +------+ 496 +------+
497 |reader| RING BUFFER 497 |reader| RING BUFFER
498 |page | 498 |page |
499 +------+ 499 +------+
500 +---+ +---+ +---+ 500 +---+ +---+ +---+
501 | |--->| |--->| | 501 | |--->| |--->| |
502 | |<---| |<---| | 502 | |<---| |<---| |
503 +---+ +---+ +---+ 503 +---+ +---+ +---+
504 ^ | ^ | 504 ^ | ^ |
505 | +---------------+ | 505 | +---------------+ |
506 +-----H-------------+ 506 +-----H-------------+
507 507
508 The reader sets the reader page next pointer a 508 The reader sets the reader page next pointer as HEADER to the page after
509 the head page:: 509 the head page::
510 510
511 511
512 +------+ 512 +------+
513 |reader| RING BUFFER 513 |reader| RING BUFFER
514 |page |-------H-----------+ 514 |page |-------H-----------+
515 +------+ v 515 +------+ v
516 | +---+ +---+ +---+ 516 | +---+ +---+ +---+
517 | | |--->| |--->| | 517 | | |--->| |--->| |
518 | | |<---| |<---| |<-+ 518 | | |<---| |<---| |<-+
519 | +---+ +---+ +---+ | 519 | +---+ +---+ +---+ |
520 | ^ | ^ | | 520 | ^ | ^ | |
521 | | +---------------+ | | 521 | | +---------------+ | |
522 | +-----H-------------+ | 522 | +-----H-------------+ |
523 +--------------------------------------+ 523 +--------------------------------------+
524 524
525 It does a cmpxchg with the pointer to the prev 525 It does a cmpxchg with the pointer to the previous head page to make it
526 point to the reader page. Note that the new po 526 point to the reader page. Note that the new pointer does not have the HEADER
527 flag set. This action atomically moves the he 527 flag set. This action atomically moves the head page forward::
528 528
529 +------+ 529 +------+
530 |reader| RING BUFFER 530 |reader| RING BUFFER
531 |page |-------H-----------+ 531 |page |-------H-----------+
532 +------+ v 532 +------+ v
533 | ^ +---+ +---+ +---+ 533 | ^ +---+ +---+ +---+
534 | | | |-->| |-->| | 534 | | | |-->| |-->| |
535 | | | |<--| |<--| |<-+ 535 | | | |<--| |<--| |<-+
536 | | +---+ +---+ +---+ | 536 | | +---+ +---+ +---+ |
537 | | | ^ | | 537 | | | ^ | |
538 | | +-------------+ | | 538 | | +-------------+ | |
539 | +-----------------------------+ | 539 | +-----------------------------+ |
540 +------------------------------------+ 540 +------------------------------------+
541 541
542 After the new head page is set, the previous p 542 After the new head page is set, the previous pointer of the head page is
543 updated to the reader page:: 543 updated to the reader page::
544 544
545 +------+ 545 +------+
546 |reader| RING BUFFER 546 |reader| RING BUFFER
547 |page |-------H-----------+ 547 |page |-------H-----------+
548 +------+ <---------------+ v 548 +------+ <---------------+ v
549 | ^ +---+ +---+ +---+ 549 | ^ +---+ +---+ +---+
550 | | | |-->| |-->| | 550 | | | |-->| |-->| |
551 | | | | | |<--| |<-+ 551 | | | | | |<--| |<-+
552 | | +---+ +---+ +---+ | 552 | | +---+ +---+ +---+ |
553 | | | ^ | | 553 | | | ^ | |
554 | | +-------------+ | | 554 | | +-------------+ | |
555 | +-----------------------------+ | 555 | +-----------------------------+ |
556 +------------------------------------+ 556 +------------------------------------+
557 557
558 +------+ 558 +------+
559 |buffer| RING BUFFER 559 |buffer| RING BUFFER
560 |page |-------H-----------+ <--- New head 560 |page |-------H-----------+ <--- New head page
561 +------+ <---------------+ v 561 +------+ <---------------+ v
562 | ^ +---+ +---+ +---+ 562 | ^ +---+ +---+ +---+
563 | | | | | |-->| | 563 | | | | | |-->| |
564 | | New | | | |<--| |<-+ 564 | | New | | | |<--| |<-+
565 | | Reader +---+ +---+ +---+ | 565 | | Reader +---+ +---+ +---+ |
566 | | page ----^ | | 566 | | page ----^ | |
567 | | | | 567 | | | |
568 | +-----------------------------+ | 568 | +-----------------------------+ |
569 +------------------------------------+ 569 +------------------------------------+
570 570
571 Another important point: The page that the rea 571 Another important point: The page that the reader page points back to
572 by its previous pointer (the one that now poin 572 by its previous pointer (the one that now points to the new head page)
573 never points back to the reader page. That is 573 never points back to the reader page. That is because the reader page is
574 not part of the ring buffer. Traversing the ri 574 not part of the ring buffer. Traversing the ring buffer via the next pointers
575 will always stay in the ring buffer. Traversin 575 will always stay in the ring buffer. Traversing the ring buffer via the
576 prev pointers may not. 576 prev pointers may not.
577 577
578 Note, the way to determine a reader page is si 578 Note, the way to determine a reader page is simply by examining the previous
579 pointer of the page. If the next pointer of th 579 pointer of the page. If the next pointer of the previous page does not
580 point back to the original page, then the orig 580 point back to the original page, then the original page is a reader page::
581 581
582 582
583 +--------+ 583 +--------+
584 | reader | next +----+ 584 | reader | next +----+
585 | page |-------->| |<====== 585 | page |-------->| |<====== (buffer page)
586 +--------+ +----+ 586 +--------+ +----+
587 | | ^ 587 | | ^
588 | v | next 588 | v | next
589 prev | +----+ 589 prev | +----+
590 +------------->| | 590 +------------->| |
591 +----+ 591 +----+
592 592
593 The way the head page moves forward: 593 The way the head page moves forward:
594 594
595 When the tail page meets the head page and the 595 When the tail page meets the head page and the buffer is in overwrite mode
596 and more writes take place, the head page must 596 and more writes take place, the head page must be moved forward before the
597 writer may move the tail page. The way this is 597 writer may move the tail page. The way this is done is that the writer
598 performs a cmpxchg to convert the pointer to t 598 performs a cmpxchg to convert the pointer to the head page from the HEADER
599 flag to have the UPDATE flag set. Once this is 599 flag to have the UPDATE flag set. Once this is done, the reader will
600 not be able to swap the head page from the buf 600 not be able to swap the head page from the buffer, nor will it be able to
601 move the head page, until the writer is finish 601 move the head page, until the writer is finished with the move.
602 602
603 This eliminates any races that the reader can 603 This eliminates any races that the reader can have on the writer. The reader
604 must spin, and this is why the reader cannot p 604 must spin, and this is why the reader cannot preempt the writer::
605 605
606 tail page 606 tail page
607 | 607 |
608 v 608 v
609 +---+ +---+ +---+ +---+ 609 +---+ +---+ +---+ +---+
610 <---| |--->| |-H->| |--->| |---> 610 <---| |--->| |-H->| |--->| |--->
611 --->| |<---| |<---| |<---| |<--- 611 --->| |<---| |<---| |<---| |<---
612 +---+ +---+ +---+ +---+ 612 +---+ +---+ +---+ +---+
613 613
614 tail page 614 tail page
615 | 615 |
616 v 616 v
617 +---+ +---+ +---+ +---+ 617 +---+ +---+ +---+ +---+
618 <---| |--->| |-U->| |--->| |---> 618 <---| |--->| |-U->| |--->| |--->
619 --->| |<---| |<---| |<---| |<--- 619 --->| |<---| |<---| |<---| |<---
620 +---+ +---+ +---+ +---+ 620 +---+ +---+ +---+ +---+
621 621
622 The following page will be made into the new h 622 The following page will be made into the new head page::
623 623
624 tail page 624 tail page
625 | 625 |
626 v 626 v
627 +---+ +---+ +---+ +---+ 627 +---+ +---+ +---+ +---+
628 <---| |--->| |-U->| |-H->| |---> 628 <---| |--->| |-U->| |-H->| |--->
629 --->| |<---| |<---| |<---| |<--- 629 --->| |<---| |<---| |<---| |<---
630 +---+ +---+ +---+ +---+ 630 +---+ +---+ +---+ +---+
631 631
632 After the new head page has been set, we can s 632 After the new head page has been set, we can set the old head page
633 pointer back to NORMAL:: 633 pointer back to NORMAL::
634 634
635 tail page 635 tail page
636 | 636 |
637 v 637 v
638 +---+ +---+ +---+ +---+ 638 +---+ +---+ +---+ +---+
639 <---| |--->| |--->| |-H->| |---> 639 <---| |--->| |--->| |-H->| |--->
640 --->| |<---| |<---| |<---| |<--- 640 --->| |<---| |<---| |<---| |<---
641 +---+ +---+ +---+ +---+ 641 +---+ +---+ +---+ +---+
642 642
643 After the head page has been moved, the tail p 643 After the head page has been moved, the tail page may now move forward::
644 644
645 tail page 645 tail page
646 | 646 |
647 v 647 v
648 +---+ +---+ +---+ +---+ 648 +---+ +---+ +---+ +---+
649 <---| |--->| |--->| |-H->| |---> 649 <---| |--->| |--->| |-H->| |--->
650 --->| |<---| |<---| |<---| |<--- 650 --->| |<---| |<---| |<---| |<---
651 +---+ +---+ +---+ +---+ 651 +---+ +---+ +---+ +---+
652 652
653 653
654 The above are the trivial updates. Now for the 654 The above are the trivial updates. Now for the more complex scenarios.
655 655
656 656
657 As stated before, if enough writes preempt the 657 As stated before, if enough writes preempt the first write, the
658 tail page may make it all the way around the b 658 tail page may make it all the way around the buffer and meet the commit
659 page. At this time, we must start dropping wri 659 page. At this time, we must start dropping writes (usually with some kind
660 of warning to the user). But what happens if t 660 of warning to the user). But what happens if the commit was still on the
661 reader page? The commit page is not part of th 661 reader page? The commit page is not part of the ring buffer. The tail page
662 must account for this:: 662 must account for this::
663 663
664 664
665 reader page commit page 665 reader page commit page
666 | | 666 | |
667 v | 667 v |
668 +---+ | 668 +---+ |
669 | |<----------+ 669 | |<----------+
670 | | 670 | |
671 | |------+ 671 | |------+
672 +---+ | 672 +---+ |
673 | 673 |
674 v 674 v
675 +---+ +---+ +---+ +---+ 675 +---+ +---+ +---+ +---+
676 <---| |--->| |-H->| |--->| |---> 676 <---| |--->| |-H->| |--->| |--->
677 --->| |<---| |<---| |<---| |<--- 677 --->| |<---| |<---| |<---| |<---
678 +---+ +---+ +---+ +---+ 678 +---+ +---+ +---+ +---+
679 ^ 679 ^
680 | 680 |
681 tail page 681 tail page
682 682
683 If the tail page were to simply push the head 683 If the tail page were to simply push the head page forward, the commit when
684 leaving the reader page would not be pointing 684 leaving the reader page would not be pointing to the correct page.
685 685
686 The solution to this is to test if the commit 686 The solution to this is to test if the commit page is on the reader page
687 before pushing the head page. If it is, then i 687 before pushing the head page. If it is, then it can be assumed that the
688 tail page wrapped the buffer, and we must drop 688 tail page wrapped the buffer, and we must drop new writes.
689 689
690 This is not a race condition, because the comm 690 This is not a race condition, because the commit page can only be moved
691 by the outermost writer (the writer that was p 691 by the outermost writer (the writer that was preempted).
692 This means that the commit will not move while 692 This means that the commit will not move while a writer is moving the
693 tail page. The reader cannot swap the reader p 693 tail page. The reader cannot swap the reader page if it is also being
694 used as the commit page. The reader can simply 694 used as the commit page. The reader can simply check that the commit
695 is off the reader page. Once the commit page l 695 is off the reader page. Once the commit page leaves the reader page
696 it will never go back on it unless a reader do 696 it will never go back on it unless a reader does another swap with the
697 buffer page that is also the commit page. 697 buffer page that is also the commit page.
698 698
699 699
700 Nested writes 700 Nested writes
701 ------------- 701 -------------
702 702
703 In the pushing forward of the tail page we mus 703 In the pushing forward of the tail page we must first push forward
704 the head page if the head page is the next pag 704 the head page if the head page is the next page. If the head page
705 is not the next page, the tail page is simply 705 is not the next page, the tail page is simply updated with a cmpxchg.
706 706
707 Only writers move the tail page. This must be 707 Only writers move the tail page. This must be done atomically to protect
708 against nested writers:: 708 against nested writers::
709 709
710 temp_page = tail_page 710 temp_page = tail_page
711 next_page = temp_page->next 711 next_page = temp_page->next
712 cmpxchg(tail_page, temp_page, next_page) 712 cmpxchg(tail_page, temp_page, next_page)
713 713
714 The above will update the tail page if it is s 714 The above will update the tail page if it is still pointing to the expected
715 page. If this fails, a nested write pushed it 715 page. If this fails, a nested write pushed it forward, the current write
716 does not need to push it:: 716 does not need to push it::
717 717
718 718
719 temp page 719 temp page
720 | 720 |
721 v 721 v
722 tail page 722 tail page
723 | 723 |
724 v 724 v
725 +---+ +---+ +---+ +---+ 725 +---+ +---+ +---+ +---+
726 <---| |--->| |--->| |--->| |---> 726 <---| |--->| |--->| |--->| |--->
727 --->| |<---| |<---| |<---| |<--- 727 --->| |<---| |<---| |<---| |<---
728 +---+ +---+ +---+ +---+ 728 +---+ +---+ +---+ +---+
729 729
730 Nested write comes in and moves the tail page 730 Nested write comes in and moves the tail page forward::
731 731
732 tail page (moved by nest 732 tail page (moved by nested writer)
733 temp page | 733 temp page |
734 | | 734 | |
735 v v 735 v v
736 +---+ +---+ +---+ +---+ 736 +---+ +---+ +---+ +---+
737 <---| |--->| |--->| |--->| |---> 737 <---| |--->| |--->| |--->| |--->
738 --->| |<---| |<---| |<---| |<--- 738 --->| |<---| |<---| |<---| |<---
739 +---+ +---+ +---+ +---+ 739 +---+ +---+ +---+ +---+
740 740
741 The above would fail the cmpxchg, but since th 741 The above would fail the cmpxchg, but since the tail page has already
742 been moved forward, the writer will just try a 742 been moved forward, the writer will just try again to reserve storage
743 on the new tail page. 743 on the new tail page.
744 744
745 But the moving of the head page is a bit more 745 But the moving of the head page is a bit more complex::
746 746
747 tail page 747 tail page
748 | 748 |
749 v 749 v
750 +---+ +---+ +---+ +---+ 750 +---+ +---+ +---+ +---+
751 <---| |--->| |-H->| |--->| |---> 751 <---| |--->| |-H->| |--->| |--->
752 --->| |<---| |<---| |<---| |<--- 752 --->| |<---| |<---| |<---| |<---
753 +---+ +---+ +---+ +---+ 753 +---+ +---+ +---+ +---+
754 754
755 The write converts the head page pointer to UP 755 The write converts the head page pointer to UPDATE::
756 756
757 tail page 757 tail page
758 | 758 |
759 v 759 v
760 +---+ +---+ +---+ +---+ 760 +---+ +---+ +---+ +---+
761 <---| |--->| |-U->| |--->| |---> 761 <---| |--->| |-U->| |--->| |--->
762 --->| |<---| |<---| |<---| |<--- 762 --->| |<---| |<---| |<---| |<---
763 +---+ +---+ +---+ +---+ 763 +---+ +---+ +---+ +---+
764 764
765 But if a nested writer preempts here, it will 765 But if a nested writer preempts here, it will see that the next
766 page is a head page, but it is also nested. It 766 page is a head page, but it is also nested. It will detect that
767 it is nested and will save that information. T 767 it is nested and will save that information. The detection is the
768 fact that it sees the UPDATE flag instead of a 768 fact that it sees the UPDATE flag instead of a HEADER or NORMAL
769 pointer. 769 pointer.
770 770
771 The nested writer will set the new head page p 771 The nested writer will set the new head page pointer::
772 772
773 tail page 773 tail page
774 | 774 |
775 v 775 v
776 +---+ +---+ +---+ +---+ 776 +---+ +---+ +---+ +---+
777 <---| |--->| |-U->| |-H->| |---> 777 <---| |--->| |-U->| |-H->| |--->
778 --->| |<---| |<---| |<---| |<--- 778 --->| |<---| |<---| |<---| |<---
779 +---+ +---+ +---+ +---+ 779 +---+ +---+ +---+ +---+
780 780
781 But it will not reset the update back to norma 781 But it will not reset the update back to normal. Only the writer
782 that converted a pointer from HEAD to UPDATE w 782 that converted a pointer from HEAD to UPDATE will convert it back
783 to NORMAL:: 783 to NORMAL::
784 784
785 tail page 785 tail page
786 | 786 |
787 v 787 v
788 +---+ +---+ +---+ +---+ 788 +---+ +---+ +---+ +---+
789 <---| |--->| |-U->| |-H->| |---> 789 <---| |--->| |-U->| |-H->| |--->
790 --->| |<---| |<---| |<---| |<--- 790 --->| |<---| |<---| |<---| |<---
791 +---+ +---+ +---+ +---+ 791 +---+ +---+ +---+ +---+
792 792
793 After the nested writer finishes, the outermos 793 After the nested writer finishes, the outermost writer will convert
794 the UPDATE pointer to NORMAL:: 794 the UPDATE pointer to NORMAL::
795 795
796 796
797 tail page 797 tail page
798 | 798 |
799 v 799 v
800 +---+ +---+ +---+ +---+ 800 +---+ +---+ +---+ +---+
801 <---| |--->| |--->| |-H->| |---> 801 <---| |--->| |--->| |-H->| |--->
802 --->| |<---| |<---| |<---| |<--- 802 --->| |<---| |<---| |<---| |<---
803 +---+ +---+ +---+ +---+ 803 +---+ +---+ +---+ +---+
804 804
805 805
806 It can be even more complex if several nested 806 It can be even more complex if several nested writes came in and moved
807 the tail page ahead several pages:: 807 the tail page ahead several pages::
808 808
809 809
810 (first writer) 810 (first writer)
811 811
812 tail page 812 tail page
813 | 813 |
814 v 814 v
815 +---+ +---+ +---+ +---+ 815 +---+ +---+ +---+ +---+
816 <---| |--->| |-H->| |--->| |---> 816 <---| |--->| |-H->| |--->| |--->
817 --->| |<---| |<---| |<---| |<--- 817 --->| |<---| |<---| |<---| |<---
818 +---+ +---+ +---+ +---+ 818 +---+ +---+ +---+ +---+
819 819
820 The write converts the head page pointer to UP 820 The write converts the head page pointer to UPDATE::
821 821
822 tail page 822 tail page
823 | 823 |
824 v 824 v
825 +---+ +---+ +---+ +---+ 825 +---+ +---+ +---+ +---+
826 <---| |--->| |-U->| |--->| |---> 826 <---| |--->| |-U->| |--->| |--->
827 --->| |<---| |<---| |<---| |<--- 827 --->| |<---| |<---| |<---| |<---
828 +---+ +---+ +---+ +---+ 828 +---+ +---+ +---+ +---+
829 829
830 Next writer comes in, and sees the update and 830 Next writer comes in, and sees the update and sets up the new
831 head page:: 831 head page::
832 832
833 (second writer) 833 (second writer)
834 834
835 tail page 835 tail page
836 | 836 |
837 v 837 v
838 +---+ +---+ +---+ +---+ 838 +---+ +---+ +---+ +---+
839 <---| |--->| |-U->| |-H->| |---> 839 <---| |--->| |-U->| |-H->| |--->
840 --->| |<---| |<---| |<---| |<--- 840 --->| |<---| |<---| |<---| |<---
841 +---+ +---+ +---+ +---+ 841 +---+ +---+ +---+ +---+
842 842
843 The nested writer moves the tail page forward. 843 The nested writer moves the tail page forward. But does not set the old
844 update page to NORMAL because it is not the ou 844 update page to NORMAL because it is not the outermost writer::
845 845
846 tail page 846 tail page
847 | 847 |
848 v 848 v
849 +---+ +---+ +---+ +---+ 849 +---+ +---+ +---+ +---+
850 <---| |--->| |-U->| |-H->| |---> 850 <---| |--->| |-U->| |-H->| |--->
851 --->| |<---| |<---| |<---| |<--- 851 --->| |<---| |<---| |<---| |<---
852 +---+ +---+ +---+ +---+ 852 +---+ +---+ +---+ +---+
853 853
854 Another writer preempts and sees the page afte 854 Another writer preempts and sees the page after the tail page is a head page.
855 It changes it from HEAD to UPDATE:: 855 It changes it from HEAD to UPDATE::
856 856
857 (third writer) 857 (third writer)
858 858
859 tail page 859 tail page
860 | 860 |
861 v 861 v
862 +---+ +---+ +---+ +---+ 862 +---+ +---+ +---+ +---+
863 <---| |--->| |-U->| |-U->| |---> 863 <---| |--->| |-U->| |-U->| |--->
864 --->| |<---| |<---| |<---| |<--- 864 --->| |<---| |<---| |<---| |<---
865 +---+ +---+ +---+ +---+ 865 +---+ +---+ +---+ +---+
866 866
867 The writer will move the head page forward:: 867 The writer will move the head page forward::
868 868
869 869
870 (third writer) 870 (third writer)
871 871
872 tail page 872 tail page
873 | 873 |
874 v 874 v
875 +---+ +---+ +---+ +---+ 875 +---+ +---+ +---+ +---+
876 <---| |--->| |-U->| |-U->| |-H-> 876 <---| |--->| |-U->| |-U->| |-H->
877 --->| |<---| |<---| |<---| |<--- 877 --->| |<---| |<---| |<---| |<---
878 +---+ +---+ +---+ +---+ 878 +---+ +---+ +---+ +---+
879 879
880 But now that the third writer did change the H 880 But now that the third writer did change the HEAD flag to UPDATE it
881 will convert it to normal:: 881 will convert it to normal::
882 882
883 883
884 (third writer) 884 (third writer)
885 885
886 tail page 886 tail page
887 | 887 |
888 v 888 v
889 +---+ +---+ +---+ +---+ 889 +---+ +---+ +---+ +---+
890 <---| |--->| |-U->| |--->| |-H-> 890 <---| |--->| |-U->| |--->| |-H->
891 --->| |<---| |<---| |<---| |<--- 891 --->| |<---| |<---| |<---| |<---
892 +---+ +---+ +---+ +---+ 892 +---+ +---+ +---+ +---+
893 893
894 894
895 Then it will move the tail page, and return ba 895 Then it will move the tail page, and return back to the second writer::
896 896
897 897
898 (second writer) 898 (second writer)
899 899
900 tail page 900 tail page
901 | 901 |
902 v 902 v
903 +---+ +---+ +---+ +---+ 903 +---+ +---+ +---+ +---+
904 <---| |--->| |-U->| |--->| |-H-> 904 <---| |--->| |-U->| |--->| |-H->
905 --->| |<---| |<---| |<---| |<--- 905 --->| |<---| |<---| |<---| |<---
906 +---+ +---+ +---+ +---+ 906 +---+ +---+ +---+ +---+
907 907
908 908
909 The second writer will fail to move the tail p 909 The second writer will fail to move the tail page because it was already
910 moved, so it will try again and add its data t 910 moved, so it will try again and add its data to the new tail page.
911 It will return to the first writer:: 911 It will return to the first writer::
912 912
913 913
914 (first writer) 914 (first writer)
915 915
916 tail page 916 tail page
917 | 917 |
918 v 918 v
919 +---+ +---+ +---+ +---+ 919 +---+ +---+ +---+ +---+
920 <---| |--->| |-U->| |--->| |-H-> 920 <---| |--->| |-U->| |--->| |-H->
921 --->| |<---| |<---| |<---| |<--- 921 --->| |<---| |<---| |<---| |<---
922 +---+ +---+ +---+ +---+ 922 +---+ +---+ +---+ +---+
923 923
924 The first writer cannot know atomically if the 924 The first writer cannot know atomically if the tail page moved
925 while it updates the HEAD page. It will then u 925 while it updates the HEAD page. It will then update the head page to
926 what it thinks is the new head page:: 926 what it thinks is the new head page::
927 927
928 928
929 (first writer) 929 (first writer)
930 930
931 tail page 931 tail page
932 | 932 |
933 v 933 v
934 +---+ +---+ +---+ +---+ 934 +---+ +---+ +---+ +---+
935 <---| |--->| |-U->| |-H->| |-H-> 935 <---| |--->| |-U->| |-H->| |-H->
936 --->| |<---| |<---| |<---| |<--- 936 --->| |<---| |<---| |<---| |<---
937 +---+ +---+ +---+ +---+ 937 +---+ +---+ +---+ +---+
938 938
939 Since the cmpxchg returns the old value of the 939 Since the cmpxchg returns the old value of the pointer the first writer
940 will see it succeeded in updating the pointer 940 will see it succeeded in updating the pointer from NORMAL to HEAD.
941 But as we can see, this is not good enough. It 941 But as we can see, this is not good enough. It must also check to see
942 if the tail page is either where it use to be 942 if the tail page is either where it use to be or on the next page::
943 943
944 944
945 (first writer) 945 (first writer)
946 946
947 A B tail page 947 A B tail page
948 | | | 948 | | |
949 v v v 949 v v v
950 +---+ +---+ +---+ +---+ 950 +---+ +---+ +---+ +---+
951 <---| |--->| |-U->| |-H->| |-H-> 951 <---| |--->| |-U->| |-H->| |-H->
952 --->| |<---| |<---| |<---| |<--- 952 --->| |<---| |<---| |<---| |<---
953 +---+ +---+ +---+ +---+ 953 +---+ +---+ +---+ +---+
954 954
955 If tail page != A and tail page != B, then it 955 If tail page != A and tail page != B, then it must reset the pointer
956 back to NORMAL. The fact that it only needs to 956 back to NORMAL. The fact that it only needs to worry about nested
957 writers means that it only needs to check this 957 writers means that it only needs to check this after setting the HEAD page::
958 958
959 959
960 (first writer) 960 (first writer)
961 961
962 A B tail page 962 A B tail page
963 | | | 963 | | |
964 v v v 964 v v v
965 +---+ +---+ +---+ +---+ 965 +---+ +---+ +---+ +---+
966 <---| |--->| |-U->| |--->| |-H-> 966 <---| |--->| |-U->| |--->| |-H->
967 --->| |<---| |<---| |<---| |<--- 967 --->| |<---| |<---| |<---| |<---
968 +---+ +---+ +---+ +---+ 968 +---+ +---+ +---+ +---+
969 969
970 Now the writer can update the head page. This 970 Now the writer can update the head page. This is also why the head page must
971 remain in UPDATE and only reset by the outermo 971 remain in UPDATE and only reset by the outermost writer. This prevents
972 the reader from seeing the incorrect head page 972 the reader from seeing the incorrect head page::
973 973
974 974
975 (first writer) 975 (first writer)
976 976
977 A B tail page 977 A B tail page
978 | | | 978 | | |
979 v v v 979 v v v
980 +---+ +---+ +---+ +---+ 980 +---+ +---+ +---+ +---+
981 <---| |--->| |--->| |--->| |-H-> 981 <---| |--->| |--->| |--->| |-H->
982 --->| |<---| |<---| |<---| |<--- 982 --->| |<---| |<---| |<---| |<---
983 +---+ +---+ +---+ +---+ 983 +---+ +---+ +---+ +---+
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.