xref: /illumos-gate/usr/src/uts/common/fs/zfs/rrwlock.c (revision 7800901e60d340b6af88e94a2149805dcfcaaf56)
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 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #include <sys/refcount.h>
29 #include <sys/rrwlock.h>
30 
31 /*
32  * This file contains the implementation of a re-entrant read
33  * reader/writer lock (aka "rrwlock").
34  *
35  * This is a normal reader/writer lock with the additional feature
36  * of allowing threads who have already obtained a read lock to
37  * re-enter another read lock (re-entrant read) - even if there are
38  * waiting writers.
39  *
40  * Callers who have not obtained a read lock give waiting writers priority.
41  *
42  * The rrwlock_t lock does not allow re-entrant writers, nor does it
43  * allow a re-entrant mix of reads and writes (that is, it does not
44  * allow a caller who has already obtained a read lock to be able to
45  * then grab a write lock without first dropping all read locks, and
46  * vice versa).
47  *
48  * The rrwlock_t uses tsd (thread specific data) to keep a list of
49  * nodes (rrw_node_t), where each node keeps track of which specific
50  * lock (rrw_node_t::rn_rrl) the thread has grabbed.  Since re-entering
51  * should be rare, a thread that grabs multiple reads on the same rrwlock_t
52  * will store multiple rrw_node_ts of the same 'rrn_rrl'. Nodes on the
53  * tsd list can represent a different rrwlock_t.  This allows a thread
54  * to enter multiple and unique rrwlock_ts for read locks at the same time.
55  *
56  * Since using tsd exposes some overhead, the rrwlock_t only needs to
57  * keep tsd data when writers are waiting.  If no writers are waiting, then
58  * a reader just bumps the anonymous read count (rr_anon_rcount) - no tsd
59  * is needed.  Once a writer attempts to grab the lock, readers then
60  * keep tsd data and bump the linked readers count (rr_linked_rcount).
61  *
62  * If there are waiting writers and there are anonymous readers, then a
63  * reader doesn't know if it is a re-entrant lock. But since it may be one,
64  * we allow the read to proceed (otherwise it could deadlock).  Since once
65  * waiting writers are active, readers no longer bump the anonymous count,
66  * the anonymous readers will eventually flush themselves out.  At this point,
67  * readers will be able to tell if they are a re-entrant lock (have a
68  * rrw_node_t entry for the lock) or not. If they are a re-entrant lock, then
69  * we must let the proceed.  If they are not, then the reader blocks for the
70  * waiting writers.  Hence, we do not starve writers.
71  */
72 
73 /* global key for TSD */
74 uint_t rrw_tsd_key;
75 
76 typedef struct rrw_node {
77 	struct rrw_node	*rn_next;
78 	rrwlock_t	*rn_rrl;
79 } rrw_node_t;
80 
81 static rrw_node_t *
82 rrn_find(rrwlock_t *rrl)
83 {
84 	rrw_node_t *rn;
85 
86 	if (refcount_count(&rrl->rr_linked_rcount) == 0)
87 		return (NULL);
88 
89 	for (rn = tsd_get(rrw_tsd_key); rn != NULL; rn = rn->rn_next) {
90 		if (rn->rn_rrl == rrl)
91 			return (rn);
92 	}
93 	return (NULL);
94 }
95 
96 /*
97  * Add a node to the head of the singly linked list.
98  */
99 static void
100 rrn_add(rrwlock_t *rrl)
101 {
102 	rrw_node_t *rn;
103 
104 	rn = kmem_alloc(sizeof (*rn), KM_SLEEP);
105 	rn->rn_rrl = rrl;
106 	rn->rn_next = tsd_get(rrw_tsd_key);
107 	VERIFY(tsd_set(rrw_tsd_key, rn) == 0);
108 }
109 
110 /*
111  * If a node is found for 'rrl', then remove the node from this
112  * thread's list and return TRUE; otherwise return FALSE.
113  */
114 static boolean_t
115 rrn_find_and_remove(rrwlock_t *rrl)
116 {
117 	rrw_node_t *rn;
118 	rrw_node_t *prev = NULL;
119 
120 	if (refcount_count(&rrl->rr_linked_rcount) == 0)
121 		return (NULL);
122 
123 	for (rn = tsd_get(rrw_tsd_key); rn != NULL; rn = rn->rn_next) {
124 		if (rn->rn_rrl == rrl) {
125 			if (prev)
126 				prev->rn_next = rn->rn_next;
127 			else
128 				VERIFY(tsd_set(rrw_tsd_key, rn->rn_next) == 0);
129 			kmem_free(rn, sizeof (*rn));
130 			return (B_TRUE);
131 		}
132 		prev = rn;
133 	}
134 	return (B_FALSE);
135 }
136 
137 void
138 rrw_init(rrwlock_t *rrl)
139 {
140 	mutex_init(&rrl->rr_lock, NULL, MUTEX_DEFAULT, NULL);
141 	cv_init(&rrl->rr_cv, NULL, CV_DEFAULT, NULL);
142 	rrl->rr_writer = NULL;
143 	refcount_create(&rrl->rr_anon_rcount);
144 	refcount_create(&rrl->rr_linked_rcount);
145 	rrl->rr_writer_wanted = B_FALSE;
146 }
147 
148 void
149 rrw_destroy(rrwlock_t *rrl)
150 {
151 	mutex_destroy(&rrl->rr_lock);
152 	cv_destroy(&rrl->rr_cv);
153 	ASSERT(rrl->rr_writer == NULL);
154 	refcount_destroy(&rrl->rr_anon_rcount);
155 	refcount_destroy(&rrl->rr_linked_rcount);
156 }
157 
158 static void
159 rrw_enter_read(rrwlock_t *rrl, void *tag)
160 {
161 	mutex_enter(&rrl->rr_lock);
162 	ASSERT(rrl->rr_writer != curthread);
163 	ASSERT(refcount_count(&rrl->rr_anon_rcount) >= 0);
164 
165 	while (rrl->rr_writer || (rrl->rr_writer_wanted &&
166 	    refcount_is_zero(&rrl->rr_anon_rcount) &&
167 	    rrn_find(rrl) == NULL))
168 		cv_wait(&rrl->rr_cv, &rrl->rr_lock);
169 
170 	if (rrl->rr_writer_wanted) {
171 		/* may or may not be a re-entrant enter */
172 		rrn_add(rrl);
173 		(void) refcount_add(&rrl->rr_linked_rcount, tag);
174 	} else {
175 		(void) refcount_add(&rrl->rr_anon_rcount, tag);
176 	}
177 	ASSERT(rrl->rr_writer == NULL);
178 	mutex_exit(&rrl->rr_lock);
179 }
180 
181 static void
182 rrw_enter_write(rrwlock_t *rrl)
183 {
184 	mutex_enter(&rrl->rr_lock);
185 	ASSERT(rrl->rr_writer != curthread);
186 
187 	while (refcount_count(&rrl->rr_anon_rcount) > 0 ||
188 	    refcount_count(&rrl->rr_linked_rcount) > 0 ||
189 	    rrl->rr_writer != NULL) {
190 		rrl->rr_writer_wanted = B_TRUE;
191 		cv_wait(&rrl->rr_cv, &rrl->rr_lock);
192 	}
193 	rrl->rr_writer_wanted = B_FALSE;
194 	rrl->rr_writer = curthread;
195 	mutex_exit(&rrl->rr_lock);
196 }
197 
198 void
199 rrw_enter(rrwlock_t *rrl, krw_t rw, void *tag)
200 {
201 	if (rw == RW_READER)
202 		rrw_enter_read(rrl, tag);
203 	else
204 		rrw_enter_write(rrl);
205 }
206 
207 void
208 rrw_exit(rrwlock_t *rrl, void *tag)
209 {
210 	mutex_enter(&rrl->rr_lock);
211 	ASSERT(!refcount_is_zero(&rrl->rr_anon_rcount) ||
212 	    !refcount_is_zero(&rrl->rr_linked_rcount) ||
213 	    rrl->rr_writer != NULL);
214 
215 	if (rrl->rr_writer == NULL) {
216 		if (rrn_find_and_remove(rrl)) {
217 			if (refcount_remove(&rrl->rr_linked_rcount, tag) == 0)
218 				cv_broadcast(&rrl->rr_cv);
219 
220 		} else {
221 			if (refcount_remove(&rrl->rr_anon_rcount, tag) == 0)
222 				cv_broadcast(&rrl->rr_cv);
223 		}
224 	} else {
225 		ASSERT(rrl->rr_writer == curthread);
226 		ASSERT(refcount_is_zero(&rrl->rr_anon_rcount) &&
227 		    refcount_is_zero(&rrl->rr_linked_rcount));
228 		rrl->rr_writer = NULL;
229 		cv_broadcast(&rrl->rr_cv);
230 	}
231 	mutex_exit(&rrl->rr_lock);
232 }
233 
234 boolean_t
235 rrw_held(rrwlock_t *rrl, krw_t rw)
236 {
237 	boolean_t held;
238 
239 	mutex_enter(&rrl->rr_lock);
240 	if (rw == RW_WRITER) {
241 		held = (rrl->rr_writer == curthread);
242 	} else {
243 		held = (!refcount_is_zero(&rrl->rr_anon_rcount) ||
244 		    !refcount_is_zero(&rrl->rr_linked_rcount));
245 	}
246 	mutex_exit(&rrl->rr_lock);
247 
248 	return (held);
249 }
250