kern_turnstile.c revision 1.4 1 /* $NetBSD: kern_turnstile.c,v 1.4 2007/02/26 09:20:53 yamt Exp $ */
2
3 /*-
4 * Copyright (c) 2002, 2006, 2007 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Jason R. Thorpe and Andrew Doran.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the NetBSD
21 * Foundation, Inc. and its contributors.
22 * 4. Neither the name of The NetBSD Foundation nor the names of its
23 * contributors may be used to endorse or promote products derived
24 * from this software without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 * POSSIBILITY OF SUCH DAMAGE.
37 */
38
39 /*
40 * Turnstiles are described in detail in:
41 *
42 * Solaris Internals: Core Kernel Architecture, Jim Mauro and
43 * Richard McDougall.
44 *
45 * Turnstiles are kept in a hash table. There are likely to be many more
46 * synchronisation objects than there are threads. Since a thread can block
47 * on only one lock at a time, we only need one turnstile per thread, and
48 * so they are allocated at thread creation time.
49 *
50 * When a thread decides it needs to block on a lock, it looks up the
51 * active turnstile for that lock. If no active turnstile exists, then
52 * the process lends its turnstile to the lock. If there is already an
53 * active turnstile for the lock, the thread places its turnstile on a
54 * list of free turnstiles, and references the active one instead.
55 *
56 * The act of looking up the turnstile acquires an interlock on the sleep
57 * queue. If a thread decides it doesn't need to block after all, then this
58 * interlock must be released by explicitly aborting the turnstile
59 * operation.
60 *
61 * When a thread is awakened, it needs to get its turnstile back. If there
62 * are still other threads waiting in the active turnstile, the the thread
63 * grabs a free turnstile off the free list. Otherwise, it can take back
64 * the active turnstile from the lock (thus deactivating the turnstile).
65 *
66 * Turnstiles are the place to do priority inheritence.
67 */
68
69 #include <sys/cdefs.h>
70 __KERNEL_RCSID(0, "$NetBSD: kern_turnstile.c,v 1.4 2007/02/26 09:20:53 yamt Exp $");
71
72 #include "opt_lockdebug.h"
73 #include "opt_multiprocessor.h"
74 #include "opt_ktrace.h"
75 #include "opt_ddb.h"
76
77 #include <sys/param.h>
78 #include <sys/lock.h>
79 #include <sys/lockdebug.h>
80 #include <sys/pool.h>
81 #include <sys/proc.h>
82 #include <sys/sleepq.h>
83 #include <sys/systm.h>
84
85 #include <uvm/uvm_extern.h>
86
87 #define TS_HASH_SIZE 64
88 #define TS_HASH_MASK (TS_HASH_SIZE - 1)
89 #define TS_HASH(obj) (((uintptr_t)(obj) >> 3) & TS_HASH_MASK)
90
91 tschain_t turnstile_tab[TS_HASH_SIZE];
92
93 struct pool turnstile_pool;
94 struct pool_cache turnstile_cache;
95
96 int turnstile_ctor(void *, void *, int);
97
98 extern turnstile_t turnstile0;
99
100 /*
101 * turnstile_init:
102 *
103 * Initialize the turnstile mechanism.
104 */
105 void
106 turnstile_init(void)
107 {
108 tschain_t *tc;
109 int i;
110
111 for (i = 0; i < TS_HASH_SIZE; i++) {
112 tc = &turnstile_tab[i];
113 LIST_INIT(&tc->tc_chain);
114 #if defined(MULTIPROCESSOR) || defined(LOCKDEBUG)
115 mutex_init(&tc->tc_mutexstore, MUTEX_SPIN, IPL_SCHED);
116 tc->tc_mutex = &tc->tc_mutexstore;
117 #else
118 tc->tc_mutex = &sched_mutex;
119 #endif
120 }
121
122 pool_init(&turnstile_pool, sizeof(turnstile_t), 0, 0, 0,
123 "tstilepl", &pool_allocator_nointr);
124 pool_cache_init(&turnstile_cache, &turnstile_pool,
125 turnstile_ctor, NULL, NULL);
126
127 (void)turnstile_ctor(NULL, &turnstile0, 0);
128 }
129
130 /*
131 * turnstile_ctor:
132 *
133 * Constructor for turnstiles.
134 */
135 int
136 turnstile_ctor(void *arg, void *obj, int flags)
137 {
138 turnstile_t *ts = obj;
139
140 memset(ts, 0, sizeof(*ts));
141 sleepq_init(&ts->ts_sleepq[TS_READER_Q], NULL);
142 sleepq_init(&ts->ts_sleepq[TS_WRITER_Q], NULL);
143 return (0);
144 }
145
146 /*
147 * turnstile_remove:
148 *
149 * Remove an LWP from a turnstile sleep queue and wake it.
150 */
151 static inline int
152 turnstile_remove(turnstile_t *ts, struct lwp *l, sleepq_t *sq)
153 {
154 turnstile_t *nts;
155
156 KASSERT(l->l_ts == ts);
157
158 /*
159 * This process is no longer using the active turnstile.
160 * Find an inactive one on the free list to give to it.
161 */
162 if ((nts = ts->ts_free) != NULL) {
163 KASSERT(TS_ALL_WAITERS(ts) > 1);
164 l->l_ts = nts;
165 ts->ts_free = nts->ts_free;
166 nts->ts_free = NULL;
167 } else {
168 /*
169 * If the free list is empty, this is the last
170 * waiter.
171 */
172 KASSERT(TS_ALL_WAITERS(ts) == 1);
173 LIST_REMOVE(ts, ts_chain);
174 }
175
176 return sleepq_remove(sq, l);
177 }
178
179 /*
180 * turnstile_lookup:
181 *
182 * Look up the turnstile for the specified lock. This acquires and
183 * holds the turnstile chain lock (sleep queue interlock).
184 */
185 turnstile_t *
186 turnstile_lookup(wchan_t obj)
187 {
188 turnstile_t *ts;
189 tschain_t *tc;
190
191 tc = &turnstile_tab[TS_HASH(obj)];
192 mutex_spin_enter(tc->tc_mutex);
193
194 LIST_FOREACH(ts, &tc->tc_chain, ts_chain)
195 if (ts->ts_obj == obj)
196 return (ts);
197
198 /*
199 * No turnstile yet for this lock. No problem, turnstile_block()
200 * handles this by fetching the turnstile from the blocking thread.
201 */
202 return (NULL);
203 }
204
205 /*
206 * turnstile_exit:
207 *
208 * Abort a turnstile operation.
209 */
210 void
211 turnstile_exit(wchan_t obj)
212 {
213 tschain_t *tc;
214
215 tc = &turnstile_tab[TS_HASH(obj)];
216 mutex_spin_exit(tc->tc_mutex);
217 }
218
219 /*
220 * turnstile_block:
221 *
222 * Enter an object into the turnstile chain and prepare the current
223 * LWP for sleep.
224 */
225 void
226 turnstile_block(turnstile_t *ts, int q, wchan_t obj, syncobj_t *sobj)
227 {
228 struct lwp *l;
229 struct lwp *cur; /* cached curlwp */
230 struct lwp *owner;
231 turnstile_t *ots;
232 tschain_t *tc;
233 sleepq_t *sq;
234 u_char prio;
235
236 tc = &turnstile_tab[TS_HASH(obj)];
237 l = cur = curlwp;
238
239 KASSERT(q == TS_READER_Q || q == TS_WRITER_Q);
240 KASSERT(mutex_owned(tc->tc_mutex));
241 KASSERT(l != NULL && l->l_ts != NULL);
242
243 if (ts == NULL) {
244 /*
245 * We are the first thread to wait for this object;
246 * lend our turnstile to it.
247 */
248 ts = l->l_ts;
249 KASSERT(TS_ALL_WAITERS(ts) == 0);
250 KASSERT(TAILQ_EMPTY(&ts->ts_sleepq[TS_READER_Q].sq_queue) &&
251 TAILQ_EMPTY(&ts->ts_sleepq[TS_WRITER_Q].sq_queue));
252 ts->ts_obj = obj;
253 ts->ts_inheritor = NULL;
254 ts->ts_sleepq[TS_READER_Q].sq_mutex = tc->tc_mutex;
255 ts->ts_sleepq[TS_WRITER_Q].sq_mutex = tc->tc_mutex;
256 LIST_INSERT_HEAD(&tc->tc_chain, ts, ts_chain);
257 } else {
258 /*
259 * Object already has a turnstile. Put our turnstile
260 * onto the free list, and reference the existing
261 * turnstile instead.
262 */
263 ots = l->l_ts;
264 ots->ts_free = ts->ts_free;
265 ts->ts_free = ots;
266 l->l_ts = ts;
267
268 KASSERT(ts->ts_obj == obj);
269 KASSERT(TS_ALL_WAITERS(ts) != 0);
270 KASSERT(!TAILQ_EMPTY(&ts->ts_sleepq[TS_READER_Q].sq_queue) ||
271 !TAILQ_EMPTY(&ts->ts_sleepq[TS_WRITER_Q].sq_queue));
272 }
273
274 sq = &ts->ts_sleepq[q];
275 sleepq_enter(sq, l);
276 LOCKDEBUG_BARRIER(tc->tc_mutex, 1);
277 prio = lwp_eprio(l);
278 sleepq_enqueue(sq, prio, obj, "tstile", sobj);
279
280 /*
281 * lend our priority to lwps on the blocking chain.
282 */
283
284 for (;;) {
285 bool dolock;
286
287 if (l->l_wchan == NULL)
288 break;
289
290 owner = (*l->l_syncobj->sobj_owner)(l->l_wchan);
291 if (owner == NULL)
292 break;
293
294 KASSERT(l != owner);
295 KASSERT(cur != owner);
296
297 if (l->l_mutex != owner->l_mutex)
298 dolock = true;
299 else
300 dolock = false;
301 if (dolock && !lwp_trylock(owner)) {
302 /*
303 * restart from curlwp.
304 */
305 lwp_unlock(l);
306 l = cur;
307 lwp_lock(l);
308 prio = lwp_eprio(l);
309 continue;
310 }
311 if (prio >= lwp_eprio(owner)) {
312 if (dolock)
313 lwp_unlock(owner);
314 break;
315 }
316 ts = l->l_ts;
317 KASSERT(ts->ts_inheritor == owner || ts->ts_inheritor == NULL);
318 if (ts->ts_inheritor == NULL) {
319 ts->ts_inheritor = owner;
320 ts->ts_eprio = prio;
321 SLIST_INSERT_HEAD(&owner->l_pi_lenders, ts, ts_pichain);
322 lwp_lendpri(owner, prio);
323 } else if (prio < ts->ts_eprio) {
324 ts->ts_eprio = prio;
325 lwp_lendpri(owner, prio);
326 }
327 if (dolock)
328 lwp_unlock(l);
329 l = owner;
330 }
331 LOCKDEBUG_BARRIER(l->l_mutex, 1);
332 if (cur->l_mutex != l->l_mutex) {
333 lwp_unlock(l);
334 lwp_lock(cur);
335 }
336 LOCKDEBUG_BARRIER(cur->l_mutex, 1);
337
338 sleepq_switch(0, 0);
339 }
340
341 /*
342 * turnstile_wakeup:
343 *
344 * Wake up the specified number of threads that are blocked
345 * in a turnstile.
346 */
347 void
348 turnstile_wakeup(turnstile_t *ts, int q, int count, struct lwp *nl)
349 {
350 sleepq_t *sq;
351 tschain_t *tc;
352 struct lwp *l;
353 int swapin;
354
355 tc = &turnstile_tab[TS_HASH(ts->ts_obj)];
356 sq = &ts->ts_sleepq[q];
357 swapin = 0;
358
359 KASSERT(q == TS_READER_Q || q == TS_WRITER_Q);
360 KASSERT(count > 0 && count <= TS_WAITERS(ts, q));
361 KASSERT(mutex_owned(tc->tc_mutex) && sq->sq_mutex == tc->tc_mutex);
362 KASSERT(ts->ts_inheritor == curlwp || ts->ts_inheritor == NULL);
363
364 /*
365 * restore inherited priority if necessary.
366 */
367
368 if (ts->ts_inheritor != NULL) {
369 turnstile_t *iter;
370 turnstile_t *next;
371 turnstile_t *prev = NULL;
372 u_char prio;
373
374 ts->ts_inheritor = NULL;
375 l = curlwp;
376 lwp_lock(l);
377
378 /*
379 * the following loop does two things.
380 *
381 * - remove ts from the list.
382 *
383 * - from the rest of the list, find the highest priority.
384 */
385
386 prio = MAXPRI;
387 KASSERT(!SLIST_EMPTY(&l->l_pi_lenders));
388 for (iter = SLIST_FIRST(&l->l_pi_lenders);
389 iter != NULL; iter = next) {
390 KASSERT(lwp_eprio(l) <= ts->ts_eprio);
391 next = SLIST_NEXT(iter, ts_pichain);
392 if (iter == ts) {
393 if (prev == NULL) {
394 SLIST_REMOVE_HEAD(&l->l_pi_lenders,
395 ts_pichain);
396 } else {
397 SLIST_REMOVE_AFTER(prev, ts_pichain);
398 }
399 } else if (prio > iter->ts_eprio) {
400 prio = iter->ts_eprio;
401 }
402 prev = iter;
403 }
404
405 lwp_lendpri(l, prio);
406 lwp_unlock(l);
407 }
408
409 if (nl != NULL) {
410 #if defined(DEBUG) || defined(LOCKDEBUG)
411 TAILQ_FOREACH(l, &sq->sq_queue, l_sleepchain) {
412 if (l == nl)
413 break;
414 }
415 if (l == NULL)
416 panic("turnstile_wakeup: nl not on sleepq");
417 #endif
418 swapin |= turnstile_remove(ts, nl, sq);
419 } else {
420 while (count-- > 0) {
421 l = TAILQ_FIRST(&sq->sq_queue);
422 KASSERT(l != NULL);
423 swapin |= turnstile_remove(ts, l, sq);
424 }
425 }
426 mutex_spin_exit(tc->tc_mutex);
427
428 /*
429 * If there are newly awakend threads that need to be swapped in,
430 * then kick the swapper into action.
431 */
432 if (swapin)
433 uvm_kick_scheduler();
434 }
435
436 /*
437 * turnstile_unsleep:
438 *
439 * Remove an LWP from the turnstile. This is called when the LWP has
440 * not been awoken normally but instead interrupted: for example, if it
441 * has received a signal. It's not a valid action for turnstiles,
442 * since LWPs blocking on a turnstile are not interruptable.
443 */
444 void
445 turnstile_unsleep(struct lwp *l)
446 {
447
448 lwp_unlock(l);
449 panic("turnstile_unsleep");
450 }
451
452 /*
453 * turnstile_changepri:
454 *
455 * Adjust the priority of an LWP residing on a turnstile.
456 */
457 void
458 turnstile_changepri(struct lwp *l, int pri)
459 {
460
461 /* XXX priority inheritance */
462 sleepq_changepri(l, pri);
463 }
464
465 #if defined(LOCKDEBUG)
466 /*
467 * turnstile_print:
468 *
469 * Given the address of a lock object, print the contents of a
470 * turnstile.
471 */
472 void
473 turnstile_print(volatile void *obj, void (*pr)(const char *, ...))
474 {
475 turnstile_t *ts;
476 tschain_t *tc;
477 sleepq_t *rsq, *wsq;
478 struct lwp *l;
479
480 tc = &turnstile_tab[TS_HASH(obj)];
481
482 LIST_FOREACH(ts, &tc->tc_chain, ts_chain)
483 if (ts->ts_obj == obj)
484 break;
485
486 (*pr)("Turnstile chain at %p with tc_mutex at %p.\n", tc, tc->tc_mutex);
487 if (ts == NULL) {
488 (*pr)("=> No active turnstile for this lock.\n");
489 return;
490 }
491
492 rsq = &ts->ts_sleepq[TS_READER_Q];
493 wsq = &ts->ts_sleepq[TS_WRITER_Q];
494
495 (*pr)("=> Turnstile at %p (wrq=%p, rdq=%p).\n", ts, rsq, wsq);
496
497 (*pr)("=> %d waiting readers:", rsq->sq_waiters);
498 TAILQ_FOREACH(l, &rsq->sq_queue, l_sleepchain) {
499 (*pr)(" %p", l);
500 }
501 (*pr)("\n");
502
503 (*pr)("=> %d waiting writers:", wsq->sq_waiters);
504 TAILQ_FOREACH(l, &wsq->sq_queue, l_sleepchain) {
505 (*pr)(" %p", l);
506 }
507 (*pr)("\n");
508 }
509 #endif /* LOCKDEBUG */
510