xref: /titanic_41/usr/src/uts/common/fs/zfs/txg.c (revision 7a286c471efbab8562f7655a82931904703fffe0)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #include <sys/zfs_context.h>
27 #include <sys/txg_impl.h>
28 #include <sys/dmu_impl.h>
29 #include <sys/dmu_tx.h>
30 #include <sys/dsl_pool.h>
31 #include <sys/callb.h>
32 
33 /*
34  * Pool-wide transaction groups.
35  */
36 
37 static void txg_sync_thread(dsl_pool_t *dp);
38 static void txg_quiesce_thread(dsl_pool_t *dp);
39 
40 int zfs_txg_timeout = 30;	/* max seconds worth of delta per txg */
41 
42 /*
43  * Prepare the txg subsystem.
44  */
45 void
46 txg_init(dsl_pool_t *dp, uint64_t txg)
47 {
48 	tx_state_t *tx = &dp->dp_tx;
49 	int c;
50 	bzero(tx, sizeof (tx_state_t));
51 
52 	tx->tx_cpu = kmem_zalloc(max_ncpus * sizeof (tx_cpu_t), KM_SLEEP);
53 
54 	for (c = 0; c < max_ncpus; c++) {
55 		int i;
56 
57 		mutex_init(&tx->tx_cpu[c].tc_lock, NULL, MUTEX_DEFAULT, NULL);
58 		for (i = 0; i < TXG_SIZE; i++) {
59 			cv_init(&tx->tx_cpu[c].tc_cv[i], NULL, CV_DEFAULT,
60 			    NULL);
61 			list_create(&tx->tx_cpu[c].tc_callbacks[i],
62 			    sizeof (dmu_tx_callback_t),
63 			    offsetof(dmu_tx_callback_t, dcb_node));
64 		}
65 	}
66 
67 	rw_init(&tx->tx_suspend, NULL, RW_DEFAULT, NULL);
68 	mutex_init(&tx->tx_sync_lock, NULL, MUTEX_DEFAULT, NULL);
69 
70 	cv_init(&tx->tx_sync_more_cv, NULL, CV_DEFAULT, NULL);
71 	cv_init(&tx->tx_sync_done_cv, NULL, CV_DEFAULT, NULL);
72 	cv_init(&tx->tx_quiesce_more_cv, NULL, CV_DEFAULT, NULL);
73 	cv_init(&tx->tx_quiesce_done_cv, NULL, CV_DEFAULT, NULL);
74 	cv_init(&tx->tx_exit_cv, NULL, CV_DEFAULT, NULL);
75 
76 	tx->tx_open_txg = txg;
77 }
78 
79 /*
80  * Close down the txg subsystem.
81  */
82 void
83 txg_fini(dsl_pool_t *dp)
84 {
85 	tx_state_t *tx = &dp->dp_tx;
86 	int c;
87 
88 	ASSERT(tx->tx_threads == 0);
89 
90 	rw_destroy(&tx->tx_suspend);
91 	mutex_destroy(&tx->tx_sync_lock);
92 
93 	cv_destroy(&tx->tx_sync_more_cv);
94 	cv_destroy(&tx->tx_sync_done_cv);
95 	cv_destroy(&tx->tx_quiesce_more_cv);
96 	cv_destroy(&tx->tx_quiesce_done_cv);
97 	cv_destroy(&tx->tx_exit_cv);
98 
99 	for (c = 0; c < max_ncpus; c++) {
100 		int i;
101 
102 		mutex_destroy(&tx->tx_cpu[c].tc_lock);
103 		for (i = 0; i < TXG_SIZE; i++) {
104 			cv_destroy(&tx->tx_cpu[c].tc_cv[i]);
105 			list_destroy(&tx->tx_cpu[c].tc_callbacks[i]);
106 		}
107 	}
108 
109 	if (tx->tx_commit_cb_taskq != NULL)
110 		taskq_destroy(tx->tx_commit_cb_taskq);
111 
112 	kmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));
113 
114 	bzero(tx, sizeof (tx_state_t));
115 }
116 
117 /*
118  * Start syncing transaction groups.
119  */
120 void
121 txg_sync_start(dsl_pool_t *dp)
122 {
123 	tx_state_t *tx = &dp->dp_tx;
124 
125 	mutex_enter(&tx->tx_sync_lock);
126 
127 	dprintf("pool %p\n", dp);
128 
129 	ASSERT(tx->tx_threads == 0);
130 
131 	tx->tx_threads = 2;
132 
133 	tx->tx_quiesce_thread = thread_create(NULL, 0, txg_quiesce_thread,
134 	    dp, 0, &p0, TS_RUN, minclsyspri);
135 
136 	/*
137 	 * The sync thread can need a larger-than-default stack size on
138 	 * 32-bit x86.  This is due in part to nested pools and
139 	 * scrub_visitbp() recursion.
140 	 */
141 	tx->tx_sync_thread = thread_create(NULL, 12<<10, txg_sync_thread,
142 	    dp, 0, &p0, TS_RUN, minclsyspri);
143 
144 	mutex_exit(&tx->tx_sync_lock);
145 }
146 
147 static void
148 txg_thread_enter(tx_state_t *tx, callb_cpr_t *cpr)
149 {
150 	CALLB_CPR_INIT(cpr, &tx->tx_sync_lock, callb_generic_cpr, FTAG);
151 	mutex_enter(&tx->tx_sync_lock);
152 }
153 
154 static void
155 txg_thread_exit(tx_state_t *tx, callb_cpr_t *cpr, kthread_t **tpp)
156 {
157 	ASSERT(*tpp != NULL);
158 	*tpp = NULL;
159 	tx->tx_threads--;
160 	cv_broadcast(&tx->tx_exit_cv);
161 	CALLB_CPR_EXIT(cpr);		/* drops &tx->tx_sync_lock */
162 	thread_exit();
163 }
164 
165 static void
166 txg_thread_wait(tx_state_t *tx, callb_cpr_t *cpr, kcondvar_t *cv, uint64_t time)
167 {
168 	CALLB_CPR_SAFE_BEGIN(cpr);
169 
170 	if (time)
171 		(void) cv_timedwait(cv, &tx->tx_sync_lock, lbolt + time);
172 	else
173 		cv_wait(cv, &tx->tx_sync_lock);
174 
175 	CALLB_CPR_SAFE_END(cpr, &tx->tx_sync_lock);
176 }
177 
178 /*
179  * Stop syncing transaction groups.
180  */
181 void
182 txg_sync_stop(dsl_pool_t *dp)
183 {
184 	tx_state_t *tx = &dp->dp_tx;
185 
186 	dprintf("pool %p\n", dp);
187 	/*
188 	 * Finish off any work in progress.
189 	 */
190 	ASSERT(tx->tx_threads == 2);
191 	txg_wait_synced(dp, 0);
192 
193 	/*
194 	 * Wake all sync threads and wait for them to die.
195 	 */
196 	mutex_enter(&tx->tx_sync_lock);
197 
198 	ASSERT(tx->tx_threads == 2);
199 
200 	tx->tx_exiting = 1;
201 
202 	cv_broadcast(&tx->tx_quiesce_more_cv);
203 	cv_broadcast(&tx->tx_quiesce_done_cv);
204 	cv_broadcast(&tx->tx_sync_more_cv);
205 
206 	while (tx->tx_threads != 0)
207 		cv_wait(&tx->tx_exit_cv, &tx->tx_sync_lock);
208 
209 	tx->tx_exiting = 0;
210 
211 	mutex_exit(&tx->tx_sync_lock);
212 }
213 
214 uint64_t
215 txg_hold_open(dsl_pool_t *dp, txg_handle_t *th)
216 {
217 	tx_state_t *tx = &dp->dp_tx;
218 	tx_cpu_t *tc = &tx->tx_cpu[CPU_SEQID];
219 	uint64_t txg;
220 
221 	mutex_enter(&tc->tc_lock);
222 
223 	txg = tx->tx_open_txg;
224 	tc->tc_count[txg & TXG_MASK]++;
225 
226 	th->th_cpu = tc;
227 	th->th_txg = txg;
228 
229 	return (txg);
230 }
231 
232 void
233 txg_rele_to_quiesce(txg_handle_t *th)
234 {
235 	tx_cpu_t *tc = th->th_cpu;
236 
237 	mutex_exit(&tc->tc_lock);
238 }
239 
240 void
241 txg_register_callbacks(txg_handle_t *th, list_t *tx_callbacks)
242 {
243 	tx_cpu_t *tc = th->th_cpu;
244 	int g = th->th_txg & TXG_MASK;
245 
246 	mutex_enter(&tc->tc_lock);
247 	list_move_tail(&tc->tc_callbacks[g], tx_callbacks);
248 	mutex_exit(&tc->tc_lock);
249 }
250 
251 void
252 txg_rele_to_sync(txg_handle_t *th)
253 {
254 	tx_cpu_t *tc = th->th_cpu;
255 	int g = th->th_txg & TXG_MASK;
256 
257 	mutex_enter(&tc->tc_lock);
258 	ASSERT(tc->tc_count[g] != 0);
259 	if (--tc->tc_count[g] == 0)
260 		cv_broadcast(&tc->tc_cv[g]);
261 	mutex_exit(&tc->tc_lock);
262 
263 	th->th_cpu = NULL;	/* defensive */
264 }
265 
266 static void
267 txg_quiesce(dsl_pool_t *dp, uint64_t txg)
268 {
269 	tx_state_t *tx = &dp->dp_tx;
270 	int g = txg & TXG_MASK;
271 	int c;
272 
273 	/*
274 	 * Grab all tx_cpu locks so nobody else can get into this txg.
275 	 */
276 	for (c = 0; c < max_ncpus; c++)
277 		mutex_enter(&tx->tx_cpu[c].tc_lock);
278 
279 	ASSERT(txg == tx->tx_open_txg);
280 	tx->tx_open_txg++;
281 
282 	/*
283 	 * Now that we've incremented tx_open_txg, we can let threads
284 	 * enter the next transaction group.
285 	 */
286 	for (c = 0; c < max_ncpus; c++)
287 		mutex_exit(&tx->tx_cpu[c].tc_lock);
288 
289 	/*
290 	 * Quiesce the transaction group by waiting for everyone to txg_exit().
291 	 */
292 	for (c = 0; c < max_ncpus; c++) {
293 		tx_cpu_t *tc = &tx->tx_cpu[c];
294 		mutex_enter(&tc->tc_lock);
295 		while (tc->tc_count[g] != 0)
296 			cv_wait(&tc->tc_cv[g], &tc->tc_lock);
297 		mutex_exit(&tc->tc_lock);
298 	}
299 }
300 
301 static void
302 txg_do_callbacks(list_t *cb_list)
303 {
304 	dmu_tx_do_callbacks(cb_list, 0);
305 
306 	list_destroy(cb_list);
307 
308 	kmem_free(cb_list, sizeof (list_t));
309 }
310 
311 /*
312  * Dispatch the commit callbacks registered on this txg to worker threads.
313  */
314 static void
315 txg_dispatch_callbacks(dsl_pool_t *dp, uint64_t txg)
316 {
317 	int c;
318 	tx_state_t *tx = &dp->dp_tx;
319 	list_t *cb_list;
320 
321 	for (c = 0; c < max_ncpus; c++) {
322 		tx_cpu_t *tc = &tx->tx_cpu[c];
323 		/* No need to lock tx_cpu_t at this point */
324 
325 		int g = txg & TXG_MASK;
326 
327 		if (list_is_empty(&tc->tc_callbacks[g]))
328 			continue;
329 
330 		if (tx->tx_commit_cb_taskq == NULL) {
331 			/*
332 			 * Commit callback taskq hasn't been created yet.
333 			 */
334 			tx->tx_commit_cb_taskq = taskq_create("tx_commit_cb",
335 			    max_ncpus, minclsyspri, max_ncpus, max_ncpus * 2,
336 			    TASKQ_PREPOPULATE);
337 		}
338 
339 		cb_list = kmem_alloc(sizeof (list_t), KM_SLEEP);
340 		list_create(cb_list, sizeof (dmu_tx_callback_t),
341 		    offsetof(dmu_tx_callback_t, dcb_node));
342 
343 		list_move_tail(&tc->tc_callbacks[g], cb_list);
344 
345 		(void) taskq_dispatch(tx->tx_commit_cb_taskq, (task_func_t *)
346 		    txg_do_callbacks, cb_list, TQ_SLEEP);
347 	}
348 }
349 
350 static void
351 txg_sync_thread(dsl_pool_t *dp)
352 {
353 	tx_state_t *tx = &dp->dp_tx;
354 	callb_cpr_t cpr;
355 	uint64_t start, delta;
356 
357 	txg_thread_enter(tx, &cpr);
358 
359 	start = delta = 0;
360 	for (;;) {
361 		uint64_t timer, timeout = zfs_txg_timeout * hz;
362 		uint64_t txg;
363 
364 		/*
365 		 * We sync when we're scrubbing, there's someone waiting
366 		 * on us, or the quiesce thread has handed off a txg to
367 		 * us, or we have reached our timeout.
368 		 */
369 		timer = (delta >= timeout ? 0 : timeout - delta);
370 		while ((dp->dp_scrub_func == SCRUB_FUNC_NONE ||
371 		    spa_shutting_down(dp->dp_spa)) &&
372 		    !tx->tx_exiting && timer > 0 &&
373 		    tx->tx_synced_txg >= tx->tx_sync_txg_waiting &&
374 		    tx->tx_quiesced_txg == 0) {
375 			dprintf("waiting; tx_synced=%llu waiting=%llu dp=%p\n",
376 			    tx->tx_synced_txg, tx->tx_sync_txg_waiting, dp);
377 			txg_thread_wait(tx, &cpr, &tx->tx_sync_more_cv, timer);
378 			delta = lbolt - start;
379 			timer = (delta > timeout ? 0 : timeout - delta);
380 		}
381 
382 		/*
383 		 * Wait until the quiesce thread hands off a txg to us,
384 		 * prompting it to do so if necessary.
385 		 */
386 		while (!tx->tx_exiting && tx->tx_quiesced_txg == 0) {
387 			if (tx->tx_quiesce_txg_waiting < tx->tx_open_txg+1)
388 				tx->tx_quiesce_txg_waiting = tx->tx_open_txg+1;
389 			cv_broadcast(&tx->tx_quiesce_more_cv);
390 			txg_thread_wait(tx, &cpr, &tx->tx_quiesce_done_cv, 0);
391 		}
392 
393 		if (tx->tx_exiting)
394 			txg_thread_exit(tx, &cpr, &tx->tx_sync_thread);
395 
396 		rw_enter(&tx->tx_suspend, RW_WRITER);
397 
398 		/*
399 		 * Consume the quiesced txg which has been handed off to
400 		 * us.  This may cause the quiescing thread to now be
401 		 * able to quiesce another txg, so we must signal it.
402 		 */
403 		txg = tx->tx_quiesced_txg;
404 		tx->tx_quiesced_txg = 0;
405 		tx->tx_syncing_txg = txg;
406 		cv_broadcast(&tx->tx_quiesce_more_cv);
407 		rw_exit(&tx->tx_suspend);
408 
409 		dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
410 		    txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
411 		mutex_exit(&tx->tx_sync_lock);
412 
413 		start = lbolt;
414 		spa_sync(dp->dp_spa, txg);
415 		delta = lbolt - start;
416 
417 		mutex_enter(&tx->tx_sync_lock);
418 		rw_enter(&tx->tx_suspend, RW_WRITER);
419 		tx->tx_synced_txg = txg;
420 		tx->tx_syncing_txg = 0;
421 		rw_exit(&tx->tx_suspend);
422 		cv_broadcast(&tx->tx_sync_done_cv);
423 
424 		/*
425 		 * Dispatch commit callbacks to worker threads.
426 		 */
427 		txg_dispatch_callbacks(dp, txg);
428 	}
429 }
430 
431 static void
432 txg_quiesce_thread(dsl_pool_t *dp)
433 {
434 	tx_state_t *tx = &dp->dp_tx;
435 	callb_cpr_t cpr;
436 
437 	txg_thread_enter(tx, &cpr);
438 
439 	for (;;) {
440 		uint64_t txg;
441 
442 		/*
443 		 * We quiesce when there's someone waiting on us.
444 		 * However, we can only have one txg in "quiescing" or
445 		 * "quiesced, waiting to sync" state.  So we wait until
446 		 * the "quiesced, waiting to sync" txg has been consumed
447 		 * by the sync thread.
448 		 */
449 		while (!tx->tx_exiting &&
450 		    (tx->tx_open_txg >= tx->tx_quiesce_txg_waiting ||
451 		    tx->tx_quiesced_txg != 0))
452 			txg_thread_wait(tx, &cpr, &tx->tx_quiesce_more_cv, 0);
453 
454 		if (tx->tx_exiting)
455 			txg_thread_exit(tx, &cpr, &tx->tx_quiesce_thread);
456 
457 		txg = tx->tx_open_txg;
458 		dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
459 		    txg, tx->tx_quiesce_txg_waiting,
460 		    tx->tx_sync_txg_waiting);
461 		mutex_exit(&tx->tx_sync_lock);
462 		txg_quiesce(dp, txg);
463 		mutex_enter(&tx->tx_sync_lock);
464 
465 		/*
466 		 * Hand this txg off to the sync thread.
467 		 */
468 		dprintf("quiesce done, handing off txg %llu\n", txg);
469 		tx->tx_quiesced_txg = txg;
470 		cv_broadcast(&tx->tx_sync_more_cv);
471 		cv_broadcast(&tx->tx_quiesce_done_cv);
472 	}
473 }
474 
475 /*
476  * Delay this thread by 'ticks' if we are still in the open transaction
477  * group and there is already a waiting txg quiesing or quiesced.  Abort
478  * the delay if this txg stalls or enters the quiesing state.
479  */
480 void
481 txg_delay(dsl_pool_t *dp, uint64_t txg, int ticks)
482 {
483 	tx_state_t *tx = &dp->dp_tx;
484 	int timeout = lbolt + ticks;
485 
486 	/* don't delay if this txg could transition to quiesing immediately */
487 	if (tx->tx_open_txg > txg ||
488 	    tx->tx_syncing_txg == txg-1 || tx->tx_synced_txg == txg-1)
489 		return;
490 
491 	mutex_enter(&tx->tx_sync_lock);
492 	if (tx->tx_open_txg > txg || tx->tx_synced_txg == txg-1) {
493 		mutex_exit(&tx->tx_sync_lock);
494 		return;
495 	}
496 
497 	while (lbolt < timeout &&
498 	    tx->tx_syncing_txg < txg-1 && !txg_stalled(dp))
499 		(void) cv_timedwait(&tx->tx_quiesce_more_cv, &tx->tx_sync_lock,
500 		    timeout);
501 
502 	mutex_exit(&tx->tx_sync_lock);
503 }
504 
505 void
506 txg_wait_synced(dsl_pool_t *dp, uint64_t txg)
507 {
508 	tx_state_t *tx = &dp->dp_tx;
509 
510 	mutex_enter(&tx->tx_sync_lock);
511 	ASSERT(tx->tx_threads == 2);
512 	if (txg == 0)
513 		txg = tx->tx_open_txg;
514 	if (tx->tx_sync_txg_waiting < txg)
515 		tx->tx_sync_txg_waiting = txg;
516 	dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
517 	    txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
518 	while (tx->tx_synced_txg < txg) {
519 		dprintf("broadcasting sync more "
520 		    "tx_synced=%llu waiting=%llu dp=%p\n",
521 		    tx->tx_synced_txg, tx->tx_sync_txg_waiting, dp);
522 		cv_broadcast(&tx->tx_sync_more_cv);
523 		cv_wait(&tx->tx_sync_done_cv, &tx->tx_sync_lock);
524 	}
525 	mutex_exit(&tx->tx_sync_lock);
526 }
527 
528 void
529 txg_wait_open(dsl_pool_t *dp, uint64_t txg)
530 {
531 	tx_state_t *tx = &dp->dp_tx;
532 
533 	mutex_enter(&tx->tx_sync_lock);
534 	ASSERT(tx->tx_threads == 2);
535 	if (txg == 0)
536 		txg = tx->tx_open_txg + 1;
537 	if (tx->tx_quiesce_txg_waiting < txg)
538 		tx->tx_quiesce_txg_waiting = txg;
539 	dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
540 	    txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
541 	while (tx->tx_open_txg < txg) {
542 		cv_broadcast(&tx->tx_quiesce_more_cv);
543 		cv_wait(&tx->tx_quiesce_done_cv, &tx->tx_sync_lock);
544 	}
545 	mutex_exit(&tx->tx_sync_lock);
546 }
547 
548 boolean_t
549 txg_stalled(dsl_pool_t *dp)
550 {
551 	tx_state_t *tx = &dp->dp_tx;
552 	return (tx->tx_quiesce_txg_waiting > tx->tx_open_txg);
553 }
554 
555 boolean_t
556 txg_sync_waiting(dsl_pool_t *dp)
557 {
558 	tx_state_t *tx = &dp->dp_tx;
559 
560 	return (tx->tx_syncing_txg <= tx->tx_sync_txg_waiting ||
561 	    tx->tx_quiesced_txg != 0);
562 }
563 
564 void
565 txg_suspend(dsl_pool_t *dp)
566 {
567 	tx_state_t *tx = &dp->dp_tx;
568 	/* XXX some code paths suspend when they are already suspended! */
569 	rw_enter(&tx->tx_suspend, RW_READER);
570 }
571 
572 void
573 txg_resume(dsl_pool_t *dp)
574 {
575 	tx_state_t *tx = &dp->dp_tx;
576 	rw_exit(&tx->tx_suspend);
577 }
578 
579 /*
580  * Per-txg object lists.
581  */
582 void
583 txg_list_create(txg_list_t *tl, size_t offset)
584 {
585 	int t;
586 
587 	mutex_init(&tl->tl_lock, NULL, MUTEX_DEFAULT, NULL);
588 
589 	tl->tl_offset = offset;
590 
591 	for (t = 0; t < TXG_SIZE; t++)
592 		tl->tl_head[t] = NULL;
593 }
594 
595 void
596 txg_list_destroy(txg_list_t *tl)
597 {
598 	int t;
599 
600 	for (t = 0; t < TXG_SIZE; t++)
601 		ASSERT(txg_list_empty(tl, t));
602 
603 	mutex_destroy(&tl->tl_lock);
604 }
605 
606 int
607 txg_list_empty(txg_list_t *tl, uint64_t txg)
608 {
609 	return (tl->tl_head[txg & TXG_MASK] == NULL);
610 }
611 
612 /*
613  * Add an entry to the list.
614  * Returns 0 if it's a new entry, 1 if it's already there.
615  */
616 int
617 txg_list_add(txg_list_t *tl, void *p, uint64_t txg)
618 {
619 	int t = txg & TXG_MASK;
620 	txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
621 	int already_on_list;
622 
623 	mutex_enter(&tl->tl_lock);
624 	already_on_list = tn->tn_member[t];
625 	if (!already_on_list) {
626 		tn->tn_member[t] = 1;
627 		tn->tn_next[t] = tl->tl_head[t];
628 		tl->tl_head[t] = tn;
629 	}
630 	mutex_exit(&tl->tl_lock);
631 
632 	return (already_on_list);
633 }
634 
635 /*
636  * Remove the head of the list and return it.
637  */
638 void *
639 txg_list_remove(txg_list_t *tl, uint64_t txg)
640 {
641 	int t = txg & TXG_MASK;
642 	txg_node_t *tn;
643 	void *p = NULL;
644 
645 	mutex_enter(&tl->tl_lock);
646 	if ((tn = tl->tl_head[t]) != NULL) {
647 		p = (char *)tn - tl->tl_offset;
648 		tl->tl_head[t] = tn->tn_next[t];
649 		tn->tn_next[t] = NULL;
650 		tn->tn_member[t] = 0;
651 	}
652 	mutex_exit(&tl->tl_lock);
653 
654 	return (p);
655 }
656 
657 /*
658  * Remove a specific item from the list and return it.
659  */
660 void *
661 txg_list_remove_this(txg_list_t *tl, void *p, uint64_t txg)
662 {
663 	int t = txg & TXG_MASK;
664 	txg_node_t *tn, **tp;
665 
666 	mutex_enter(&tl->tl_lock);
667 
668 	for (tp = &tl->tl_head[t]; (tn = *tp) != NULL; tp = &tn->tn_next[t]) {
669 		if ((char *)tn - tl->tl_offset == p) {
670 			*tp = tn->tn_next[t];
671 			tn->tn_next[t] = NULL;
672 			tn->tn_member[t] = 0;
673 			mutex_exit(&tl->tl_lock);
674 			return (p);
675 		}
676 	}
677 
678 	mutex_exit(&tl->tl_lock);
679 
680 	return (NULL);
681 }
682 
683 int
684 txg_list_member(txg_list_t *tl, void *p, uint64_t txg)
685 {
686 	int t = txg & TXG_MASK;
687 	txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
688 
689 	return (tn->tn_member[t]);
690 }
691 
692 /*
693  * Walk a txg list -- only safe if you know it's not changing.
694  */
695 void *
696 txg_list_head(txg_list_t *tl, uint64_t txg)
697 {
698 	int t = txg & TXG_MASK;
699 	txg_node_t *tn = tl->tl_head[t];
700 
701 	return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
702 }
703 
704 void *
705 txg_list_next(txg_list_t *tl, void *p, uint64_t txg)
706 {
707 	int t = txg & TXG_MASK;
708 	txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
709 
710 	tn = tn->tn_next[t];
711 
712 	return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
713 }
714