xref: /titanic_50/usr/src/uts/common/fs/zfs/txg.c (revision 540db9a98e48e044a5fb290242f3ebb8cc3afc36)
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 	mutex_init(&tx->tx_sync_lock, NULL, MUTEX_DEFAULT, NULL);
68 
69 	cv_init(&tx->tx_sync_more_cv, NULL, CV_DEFAULT, NULL);
70 	cv_init(&tx->tx_sync_done_cv, NULL, CV_DEFAULT, NULL);
71 	cv_init(&tx->tx_quiesce_more_cv, NULL, CV_DEFAULT, NULL);
72 	cv_init(&tx->tx_quiesce_done_cv, NULL, CV_DEFAULT, NULL);
73 	cv_init(&tx->tx_exit_cv, NULL, CV_DEFAULT, NULL);
74 
75 	tx->tx_open_txg = txg;
76 }
77 
78 /*
79  * Close down the txg subsystem.
80  */
81 void
82 txg_fini(dsl_pool_t *dp)
83 {
84 	tx_state_t *tx = &dp->dp_tx;
85 	int c;
86 
87 	ASSERT(tx->tx_threads == 0);
88 
89 	mutex_destroy(&tx->tx_sync_lock);
90 
91 	cv_destroy(&tx->tx_sync_more_cv);
92 	cv_destroy(&tx->tx_sync_done_cv);
93 	cv_destroy(&tx->tx_quiesce_more_cv);
94 	cv_destroy(&tx->tx_quiesce_done_cv);
95 	cv_destroy(&tx->tx_exit_cv);
96 
97 	for (c = 0; c < max_ncpus; c++) {
98 		int i;
99 
100 		mutex_destroy(&tx->tx_cpu[c].tc_lock);
101 		for (i = 0; i < TXG_SIZE; i++) {
102 			cv_destroy(&tx->tx_cpu[c].tc_cv[i]);
103 			list_destroy(&tx->tx_cpu[c].tc_callbacks[i]);
104 		}
105 	}
106 
107 	if (tx->tx_commit_cb_taskq != NULL)
108 		taskq_destroy(tx->tx_commit_cb_taskq);
109 
110 	kmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));
111 
112 	bzero(tx, sizeof (tx_state_t));
113 }
114 
115 /*
116  * Start syncing transaction groups.
117  */
118 void
119 txg_sync_start(dsl_pool_t *dp)
120 {
121 	tx_state_t *tx = &dp->dp_tx;
122 
123 	mutex_enter(&tx->tx_sync_lock);
124 
125 	dprintf("pool %p\n", dp);
126 
127 	ASSERT(tx->tx_threads == 0);
128 
129 	tx->tx_threads = 2;
130 
131 	tx->tx_quiesce_thread = thread_create(NULL, 0, txg_quiesce_thread,
132 	    dp, 0, &p0, TS_RUN, minclsyspri);
133 
134 	/*
135 	 * The sync thread can need a larger-than-default stack size on
136 	 * 32-bit x86.  This is due in part to nested pools and
137 	 * scrub_visitbp() recursion.
138 	 */
139 	tx->tx_sync_thread = thread_create(NULL, 12<<10, txg_sync_thread,
140 	    dp, 0, &p0, TS_RUN, minclsyspri);
141 
142 	mutex_exit(&tx->tx_sync_lock);
143 }
144 
145 static void
146 txg_thread_enter(tx_state_t *tx, callb_cpr_t *cpr)
147 {
148 	CALLB_CPR_INIT(cpr, &tx->tx_sync_lock, callb_generic_cpr, FTAG);
149 	mutex_enter(&tx->tx_sync_lock);
150 }
151 
152 static void
153 txg_thread_exit(tx_state_t *tx, callb_cpr_t *cpr, kthread_t **tpp)
154 {
155 	ASSERT(*tpp != NULL);
156 	*tpp = NULL;
157 	tx->tx_threads--;
158 	cv_broadcast(&tx->tx_exit_cv);
159 	CALLB_CPR_EXIT(cpr);		/* drops &tx->tx_sync_lock */
160 	thread_exit();
161 }
162 
163 static void
164 txg_thread_wait(tx_state_t *tx, callb_cpr_t *cpr, kcondvar_t *cv, uint64_t time)
165 {
166 	CALLB_CPR_SAFE_BEGIN(cpr);
167 
168 	if (time)
169 		(void) cv_timedwait(cv, &tx->tx_sync_lock, lbolt + time);
170 	else
171 		cv_wait(cv, &tx->tx_sync_lock);
172 
173 	CALLB_CPR_SAFE_END(cpr, &tx->tx_sync_lock);
174 }
175 
176 /*
177  * Stop syncing transaction groups.
178  */
179 void
180 txg_sync_stop(dsl_pool_t *dp)
181 {
182 	tx_state_t *tx = &dp->dp_tx;
183 
184 	dprintf("pool %p\n", dp);
185 	/*
186 	 * Finish off any work in progress.
187 	 */
188 	ASSERT(tx->tx_threads == 2);
189 
190 	/*
191 	 * We need to ensure that we've vacated the deferred space_maps.
192 	 */
193 	txg_wait_synced(dp, tx->tx_open_txg + TXG_DEFER_SIZE);
194 
195 	/*
196 	 * Wake all sync threads and wait for them to die.
197 	 */
198 	mutex_enter(&tx->tx_sync_lock);
199 
200 	ASSERT(tx->tx_threads == 2);
201 
202 	tx->tx_exiting = 1;
203 
204 	cv_broadcast(&tx->tx_quiesce_more_cv);
205 	cv_broadcast(&tx->tx_quiesce_done_cv);
206 	cv_broadcast(&tx->tx_sync_more_cv);
207 
208 	while (tx->tx_threads != 0)
209 		cv_wait(&tx->tx_exit_cv, &tx->tx_sync_lock);
210 
211 	tx->tx_exiting = 0;
212 
213 	mutex_exit(&tx->tx_sync_lock);
214 }
215 
216 uint64_t
217 txg_hold_open(dsl_pool_t *dp, txg_handle_t *th)
218 {
219 	tx_state_t *tx = &dp->dp_tx;
220 	tx_cpu_t *tc = &tx->tx_cpu[CPU_SEQID];
221 	uint64_t txg;
222 
223 	mutex_enter(&tc->tc_lock);
224 
225 	txg = tx->tx_open_txg;
226 	tc->tc_count[txg & TXG_MASK]++;
227 
228 	th->th_cpu = tc;
229 	th->th_txg = txg;
230 
231 	return (txg);
232 }
233 
234 void
235 txg_rele_to_quiesce(txg_handle_t *th)
236 {
237 	tx_cpu_t *tc = th->th_cpu;
238 
239 	mutex_exit(&tc->tc_lock);
240 }
241 
242 void
243 txg_register_callbacks(txg_handle_t *th, list_t *tx_callbacks)
244 {
245 	tx_cpu_t *tc = th->th_cpu;
246 	int g = th->th_txg & TXG_MASK;
247 
248 	mutex_enter(&tc->tc_lock);
249 	list_move_tail(&tc->tc_callbacks[g], tx_callbacks);
250 	mutex_exit(&tc->tc_lock);
251 }
252 
253 void
254 txg_rele_to_sync(txg_handle_t *th)
255 {
256 	tx_cpu_t *tc = th->th_cpu;
257 	int g = th->th_txg & TXG_MASK;
258 
259 	mutex_enter(&tc->tc_lock);
260 	ASSERT(tc->tc_count[g] != 0);
261 	if (--tc->tc_count[g] == 0)
262 		cv_broadcast(&tc->tc_cv[g]);
263 	mutex_exit(&tc->tc_lock);
264 
265 	th->th_cpu = NULL;	/* defensive */
266 }
267 
268 static void
269 txg_quiesce(dsl_pool_t *dp, uint64_t txg)
270 {
271 	tx_state_t *tx = &dp->dp_tx;
272 	int g = txg & TXG_MASK;
273 	int c;
274 
275 	/*
276 	 * Grab all tx_cpu locks so nobody else can get into this txg.
277 	 */
278 	for (c = 0; c < max_ncpus; c++)
279 		mutex_enter(&tx->tx_cpu[c].tc_lock);
280 
281 	ASSERT(txg == tx->tx_open_txg);
282 	tx->tx_open_txg++;
283 
284 	/*
285 	 * Now that we've incremented tx_open_txg, we can let threads
286 	 * enter the next transaction group.
287 	 */
288 	for (c = 0; c < max_ncpus; c++)
289 		mutex_exit(&tx->tx_cpu[c].tc_lock);
290 
291 	/*
292 	 * Quiesce the transaction group by waiting for everyone to txg_exit().
293 	 */
294 	for (c = 0; c < max_ncpus; c++) {
295 		tx_cpu_t *tc = &tx->tx_cpu[c];
296 		mutex_enter(&tc->tc_lock);
297 		while (tc->tc_count[g] != 0)
298 			cv_wait(&tc->tc_cv[g], &tc->tc_lock);
299 		mutex_exit(&tc->tc_lock);
300 	}
301 }
302 
303 static void
304 txg_do_callbacks(list_t *cb_list)
305 {
306 	dmu_tx_do_callbacks(cb_list, 0);
307 
308 	list_destroy(cb_list);
309 
310 	kmem_free(cb_list, sizeof (list_t));
311 }
312 
313 /*
314  * Dispatch the commit callbacks registered on this txg to worker threads.
315  */
316 static void
317 txg_dispatch_callbacks(dsl_pool_t *dp, uint64_t txg)
318 {
319 	int c;
320 	tx_state_t *tx = &dp->dp_tx;
321 	list_t *cb_list;
322 
323 	for (c = 0; c < max_ncpus; c++) {
324 		tx_cpu_t *tc = &tx->tx_cpu[c];
325 		/* No need to lock tx_cpu_t at this point */
326 
327 		int g = txg & TXG_MASK;
328 
329 		if (list_is_empty(&tc->tc_callbacks[g]))
330 			continue;
331 
332 		if (tx->tx_commit_cb_taskq == NULL) {
333 			/*
334 			 * Commit callback taskq hasn't been created yet.
335 			 */
336 			tx->tx_commit_cb_taskq = taskq_create("tx_commit_cb",
337 			    max_ncpus, minclsyspri, max_ncpus, max_ncpus * 2,
338 			    TASKQ_PREPOPULATE);
339 		}
340 
341 		cb_list = kmem_alloc(sizeof (list_t), KM_SLEEP);
342 		list_create(cb_list, sizeof (dmu_tx_callback_t),
343 		    offsetof(dmu_tx_callback_t, dcb_node));
344 
345 		list_move_tail(&tc->tc_callbacks[g], cb_list);
346 
347 		(void) taskq_dispatch(tx->tx_commit_cb_taskq, (task_func_t *)
348 		    txg_do_callbacks, cb_list, TQ_SLEEP);
349 	}
350 }
351 
352 static void
353 txg_sync_thread(dsl_pool_t *dp)
354 {
355 	tx_state_t *tx = &dp->dp_tx;
356 	callb_cpr_t cpr;
357 	uint64_t start, delta;
358 
359 	txg_thread_enter(tx, &cpr);
360 
361 	start = delta = 0;
362 	for (;;) {
363 		uint64_t timer, timeout = zfs_txg_timeout * hz;
364 		uint64_t txg;
365 
366 		/*
367 		 * We sync when we're scrubbing, there's someone waiting
368 		 * on us, or the quiesce thread has handed off a txg to
369 		 * us, or we have reached our timeout.
370 		 */
371 		timer = (delta >= timeout ? 0 : timeout - delta);
372 		while ((dp->dp_scrub_func == SCRUB_FUNC_NONE ||
373 		    spa_shutting_down(dp->dp_spa)) &&
374 		    !tx->tx_exiting && timer > 0 &&
375 		    tx->tx_synced_txg >= tx->tx_sync_txg_waiting &&
376 		    tx->tx_quiesced_txg == 0) {
377 			dprintf("waiting; tx_synced=%llu waiting=%llu dp=%p\n",
378 			    tx->tx_synced_txg, tx->tx_sync_txg_waiting, dp);
379 			txg_thread_wait(tx, &cpr, &tx->tx_sync_more_cv, timer);
380 			delta = lbolt - start;
381 			timer = (delta > timeout ? 0 : timeout - delta);
382 		}
383 
384 		/*
385 		 * Wait until the quiesce thread hands off a txg to us,
386 		 * prompting it to do so if necessary.
387 		 */
388 		while (!tx->tx_exiting && tx->tx_quiesced_txg == 0) {
389 			if (tx->tx_quiesce_txg_waiting < tx->tx_open_txg+1)
390 				tx->tx_quiesce_txg_waiting = tx->tx_open_txg+1;
391 			cv_broadcast(&tx->tx_quiesce_more_cv);
392 			txg_thread_wait(tx, &cpr, &tx->tx_quiesce_done_cv, 0);
393 		}
394 
395 		if (tx->tx_exiting)
396 			txg_thread_exit(tx, &cpr, &tx->tx_sync_thread);
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 
408 		dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
409 		    txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
410 		mutex_exit(&tx->tx_sync_lock);
411 
412 		start = lbolt;
413 		spa_sync(dp->dp_spa, txg);
414 		delta = lbolt - start;
415 
416 		mutex_enter(&tx->tx_sync_lock);
417 		tx->tx_synced_txg = txg;
418 		tx->tx_syncing_txg = 0;
419 		cv_broadcast(&tx->tx_sync_done_cv);
420 
421 		/*
422 		 * Dispatch commit callbacks to worker threads.
423 		 */
424 		txg_dispatch_callbacks(dp, txg);
425 	}
426 }
427 
428 static void
429 txg_quiesce_thread(dsl_pool_t *dp)
430 {
431 	tx_state_t *tx = &dp->dp_tx;
432 	callb_cpr_t cpr;
433 
434 	txg_thread_enter(tx, &cpr);
435 
436 	for (;;) {
437 		uint64_t txg;
438 
439 		/*
440 		 * We quiesce when there's someone waiting on us.
441 		 * However, we can only have one txg in "quiescing" or
442 		 * "quiesced, waiting to sync" state.  So we wait until
443 		 * the "quiesced, waiting to sync" txg has been consumed
444 		 * by the sync thread.
445 		 */
446 		while (!tx->tx_exiting &&
447 		    (tx->tx_open_txg >= tx->tx_quiesce_txg_waiting ||
448 		    tx->tx_quiesced_txg != 0))
449 			txg_thread_wait(tx, &cpr, &tx->tx_quiesce_more_cv, 0);
450 
451 		if (tx->tx_exiting)
452 			txg_thread_exit(tx, &cpr, &tx->tx_quiesce_thread);
453 
454 		txg = tx->tx_open_txg;
455 		dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
456 		    txg, tx->tx_quiesce_txg_waiting,
457 		    tx->tx_sync_txg_waiting);
458 		mutex_exit(&tx->tx_sync_lock);
459 		txg_quiesce(dp, txg);
460 		mutex_enter(&tx->tx_sync_lock);
461 
462 		/*
463 		 * Hand this txg off to the sync thread.
464 		 */
465 		dprintf("quiesce done, handing off txg %llu\n", txg);
466 		tx->tx_quiesced_txg = txg;
467 		cv_broadcast(&tx->tx_sync_more_cv);
468 		cv_broadcast(&tx->tx_quiesce_done_cv);
469 	}
470 }
471 
472 /*
473  * Delay this thread by 'ticks' if we are still in the open transaction
474  * group and there is already a waiting txg quiesing or quiesced.  Abort
475  * the delay if this txg stalls or enters the quiesing state.
476  */
477 void
478 txg_delay(dsl_pool_t *dp, uint64_t txg, int ticks)
479 {
480 	tx_state_t *tx = &dp->dp_tx;
481 	int timeout = lbolt + ticks;
482 
483 	/* don't delay if this txg could transition to quiesing immediately */
484 	if (tx->tx_open_txg > txg ||
485 	    tx->tx_syncing_txg == txg-1 || tx->tx_synced_txg == txg-1)
486 		return;
487 
488 	mutex_enter(&tx->tx_sync_lock);
489 	if (tx->tx_open_txg > txg || tx->tx_synced_txg == txg-1) {
490 		mutex_exit(&tx->tx_sync_lock);
491 		return;
492 	}
493 
494 	while (lbolt < timeout &&
495 	    tx->tx_syncing_txg < txg-1 && !txg_stalled(dp))
496 		(void) cv_timedwait(&tx->tx_quiesce_more_cv, &tx->tx_sync_lock,
497 		    timeout);
498 
499 	mutex_exit(&tx->tx_sync_lock);
500 }
501 
502 void
503 txg_wait_synced(dsl_pool_t *dp, uint64_t txg)
504 {
505 	tx_state_t *tx = &dp->dp_tx;
506 
507 	mutex_enter(&tx->tx_sync_lock);
508 	ASSERT(tx->tx_threads == 2);
509 	if (txg == 0)
510 		txg = tx->tx_open_txg + TXG_DEFER_SIZE;
511 	if (tx->tx_sync_txg_waiting < txg)
512 		tx->tx_sync_txg_waiting = txg;
513 	dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
514 	    txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
515 	while (tx->tx_synced_txg < txg) {
516 		dprintf("broadcasting sync more "
517 		    "tx_synced=%llu waiting=%llu dp=%p\n",
518 		    tx->tx_synced_txg, tx->tx_sync_txg_waiting, dp);
519 		cv_broadcast(&tx->tx_sync_more_cv);
520 		cv_wait(&tx->tx_sync_done_cv, &tx->tx_sync_lock);
521 	}
522 	mutex_exit(&tx->tx_sync_lock);
523 }
524 
525 void
526 txg_wait_open(dsl_pool_t *dp, uint64_t txg)
527 {
528 	tx_state_t *tx = &dp->dp_tx;
529 
530 	mutex_enter(&tx->tx_sync_lock);
531 	ASSERT(tx->tx_threads == 2);
532 	if (txg == 0)
533 		txg = tx->tx_open_txg + 1;
534 	if (tx->tx_quiesce_txg_waiting < txg)
535 		tx->tx_quiesce_txg_waiting = txg;
536 	dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
537 	    txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
538 	while (tx->tx_open_txg < txg) {
539 		cv_broadcast(&tx->tx_quiesce_more_cv);
540 		cv_wait(&tx->tx_quiesce_done_cv, &tx->tx_sync_lock);
541 	}
542 	mutex_exit(&tx->tx_sync_lock);
543 }
544 
545 boolean_t
546 txg_stalled(dsl_pool_t *dp)
547 {
548 	tx_state_t *tx = &dp->dp_tx;
549 	return (tx->tx_quiesce_txg_waiting > tx->tx_open_txg);
550 }
551 
552 boolean_t
553 txg_sync_waiting(dsl_pool_t *dp)
554 {
555 	tx_state_t *tx = &dp->dp_tx;
556 
557 	return (tx->tx_syncing_txg <= tx->tx_sync_txg_waiting ||
558 	    tx->tx_quiesced_txg != 0);
559 }
560 
561 /*
562  * Per-txg object lists.
563  */
564 void
565 txg_list_create(txg_list_t *tl, size_t offset)
566 {
567 	int t;
568 
569 	mutex_init(&tl->tl_lock, NULL, MUTEX_DEFAULT, NULL);
570 
571 	tl->tl_offset = offset;
572 
573 	for (t = 0; t < TXG_SIZE; t++)
574 		tl->tl_head[t] = NULL;
575 }
576 
577 void
578 txg_list_destroy(txg_list_t *tl)
579 {
580 	int t;
581 
582 	for (t = 0; t < TXG_SIZE; t++)
583 		ASSERT(txg_list_empty(tl, t));
584 
585 	mutex_destroy(&tl->tl_lock);
586 }
587 
588 int
589 txg_list_empty(txg_list_t *tl, uint64_t txg)
590 {
591 	return (tl->tl_head[txg & TXG_MASK] == NULL);
592 }
593 
594 /*
595  * Add an entry to the list.
596  * Returns 0 if it's a new entry, 1 if it's already there.
597  */
598 int
599 txg_list_add(txg_list_t *tl, void *p, uint64_t txg)
600 {
601 	int t = txg & TXG_MASK;
602 	txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
603 	int already_on_list;
604 
605 	mutex_enter(&tl->tl_lock);
606 	already_on_list = tn->tn_member[t];
607 	if (!already_on_list) {
608 		tn->tn_member[t] = 1;
609 		tn->tn_next[t] = tl->tl_head[t];
610 		tl->tl_head[t] = tn;
611 	}
612 	mutex_exit(&tl->tl_lock);
613 
614 	return (already_on_list);
615 }
616 
617 /*
618  * Remove the head of the list and return it.
619  */
620 void *
621 txg_list_remove(txg_list_t *tl, uint64_t txg)
622 {
623 	int t = txg & TXG_MASK;
624 	txg_node_t *tn;
625 	void *p = NULL;
626 
627 	mutex_enter(&tl->tl_lock);
628 	if ((tn = tl->tl_head[t]) != NULL) {
629 		p = (char *)tn - tl->tl_offset;
630 		tl->tl_head[t] = tn->tn_next[t];
631 		tn->tn_next[t] = NULL;
632 		tn->tn_member[t] = 0;
633 	}
634 	mutex_exit(&tl->tl_lock);
635 
636 	return (p);
637 }
638 
639 /*
640  * Remove a specific item from the list and return it.
641  */
642 void *
643 txg_list_remove_this(txg_list_t *tl, void *p, uint64_t txg)
644 {
645 	int t = txg & TXG_MASK;
646 	txg_node_t *tn, **tp;
647 
648 	mutex_enter(&tl->tl_lock);
649 
650 	for (tp = &tl->tl_head[t]; (tn = *tp) != NULL; tp = &tn->tn_next[t]) {
651 		if ((char *)tn - tl->tl_offset == p) {
652 			*tp = tn->tn_next[t];
653 			tn->tn_next[t] = NULL;
654 			tn->tn_member[t] = 0;
655 			mutex_exit(&tl->tl_lock);
656 			return (p);
657 		}
658 	}
659 
660 	mutex_exit(&tl->tl_lock);
661 
662 	return (NULL);
663 }
664 
665 int
666 txg_list_member(txg_list_t *tl, void *p, uint64_t txg)
667 {
668 	int t = txg & TXG_MASK;
669 	txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
670 
671 	return (tn->tn_member[t]);
672 }
673 
674 /*
675  * Walk a txg list -- only safe if you know it's not changing.
676  */
677 void *
678 txg_list_head(txg_list_t *tl, uint64_t txg)
679 {
680 	int t = txg & TXG_MASK;
681 	txg_node_t *tn = tl->tl_head[t];
682 
683 	return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
684 }
685 
686 void *
687 txg_list_next(txg_list_t *tl, void *p, uint64_t txg)
688 {
689 	int t = txg & TXG_MASK;
690 	txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
691 
692 	tn = tn->tn_next[t];
693 
694 	return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
695 }
696