rtld_lock.c revision 137785 
1/*-
2 * Copyright 1999, 2000 John D. Polstra.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
17 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
18 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
19 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
20 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
21 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
22 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
23 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
24 *
25 *	from: FreeBSD: src/libexec/rtld-elf/sparc64/lockdflt.c,v 1.3 2002/10/09
26 * $FreeBSD: head/libexec/rtld-elf/rtld_lock.c 137785 2004-11-16 20:45:51Z jhb $
27 */
28
29/*
30 * Thread locking implementation for the dynamic linker.
31 *
32 * We use the "simple, non-scalable reader-preference lock" from:
33 *
34 *   J. M. Mellor-Crummey and M. L. Scott. "Scalable Reader-Writer
35 *   Synchronization for Shared-Memory Multiprocessors." 3rd ACM Symp. on
36 *   Principles and Practice of Parallel Programming, April 1991.
37 *
38 * In this algorithm the lock is a single word.  Its low-order bit is
39 * set when a writer holds the lock.  The remaining high-order bits
40 * contain a count of readers desiring the lock.  The algorithm requires
41 * atomic "compare_and_store" and "add" operations, which we implement
42 * using assembly language sequences in "rtld_start.S".
43 */
44
45#include <signal.h>
46#include <stdlib.h>
47#include <time.h>
48
49#include "debug.h"
50#include "rtld.h"
51#include "rtld_machdep.h"
52
53#define WAFLAG		0x1	/* A writer holds the lock */
54#define RC_INCR		0x2	/* Adjusts count of readers desiring lock */
55
56typedef struct Struct_Lock {
57	volatile int lock;
58	void *base;
59} Lock;
60
61static sigset_t fullsigmask, oldsigmask;
62static int thread_flag;
63
64static void *
65def_lock_create()
66{
67    void *base;
68    char *p;
69    uintptr_t r;
70    Lock *l;
71
72    /*
73     * Arrange for the lock to occupy its own cache line.  First, we
74     * optimistically allocate just a cache line, hoping that malloc
75     * will give us a well-aligned block of memory.  If that doesn't
76     * work, we allocate a larger block and take a well-aligned cache
77     * line from it.
78     */
79    base = xmalloc(CACHE_LINE_SIZE);
80    p = (char *)base;
81    if ((uintptr_t)p % CACHE_LINE_SIZE != 0) {
82	free(base);
83	base = xmalloc(2 * CACHE_LINE_SIZE);
84	p = (char *)base;
85	if ((r = (uintptr_t)p % CACHE_LINE_SIZE) != 0)
86	    p += CACHE_LINE_SIZE - r;
87    }
88    l = (Lock *)p;
89    l->base = base;
90    l->lock = 0;
91    return l;
92}
93
94static void
95def_lock_destroy(void *lock)
96{
97    Lock *l = (Lock *)lock;
98
99    free(l->base);
100}
101
102static void
103def_rlock_acquire(void *lock)
104{
105    Lock *l = (Lock *)lock;
106
107    atomic_add_acq_int(&l->lock, RC_INCR);
108    while (l->lock & WAFLAG)
109	    ;	/* Spin */
110}
111
112static void
113def_wlock_acquire(void *lock)
114{
115    Lock *l = (Lock *)lock;
116    sigset_t tmp_oldsigmask;
117
118    for ( ; ; ) {
119	sigprocmask(SIG_BLOCK, &fullsigmask, &tmp_oldsigmask);
120	if (atomic_cmpset_acq_int(&l->lock, 0, WAFLAG))
121	    break;
122	sigprocmask(SIG_SETMASK, &tmp_oldsigmask, NULL);
123    }
124    oldsigmask = tmp_oldsigmask;
125}
126
127static void
128def_lock_release(void *lock)
129{
130    Lock *l = (Lock *)lock;
131
132    if ((l->lock & WAFLAG) == 0)
133    	atomic_add_rel_int(&l->lock, -RC_INCR);
134    else {
135    	atomic_add_rel_int(&l->lock, -WAFLAG);
136    	sigprocmask(SIG_SETMASK, &oldsigmask, NULL);
137    }
138}
139
140static int
141def_thread_set_flag(int mask)
142{
143	int old_val = thread_flag;
144	thread_flag |= mask;
145	return (old_val);
146}
147
148static int
149def_thread_clr_flag(int mask)
150{
151	int old_val = thread_flag;
152	thread_flag &= ~mask;
153	return (old_val);
154}
155
156/*
157 * Public interface exposed to the rest of the dynamic linker.
158 */
159static struct RtldLockInfo lockinfo;
160static struct RtldLockInfo deflockinfo;
161
162static __inline int
163thread_mask_set(int mask)
164{
165	return lockinfo.thread_set_flag(mask);
166}
167
168static __inline void
169thread_mask_clear(int mask)
170{
171	lockinfo.thread_clr_flag(mask);
172}
173
174#define	RTLD_LOCK_CNT	2
175struct rtld_lock {
176	void	*handle;
177	int	 mask;
178} rtld_locks[RTLD_LOCK_CNT];
179
180rtld_lock_t	rtld_bind_lock = &rtld_locks[0];
181rtld_lock_t	rtld_libc_lock = &rtld_locks[1];
182
183int
184rlock_acquire(rtld_lock_t lock)
185{
186	if (thread_mask_set(lock->mask)) {
187	    dbg("rlock_acquire: recursed");
188	    return (0);
189	}
190	lockinfo.rlock_acquire(lock->handle);
191	return (1);
192}
193
194int
195wlock_acquire(rtld_lock_t lock)
196{
197	if (thread_mask_set(lock->mask)) {
198	    dbg("wlock_acquire: recursed");
199	    return (0);
200	}
201	lockinfo.wlock_acquire(lock->handle);
202	return (1);
203}
204
205void
206rlock_release(rtld_lock_t lock, int locked)
207{
208	if (locked == 0)
209	    return;
210	thread_mask_clear(lock->mask);
211	lockinfo.lock_release(lock->handle);
212}
213
214void
215wlock_release(rtld_lock_t lock, int locked)
216{
217	if (locked == 0)
218	    return;
219	thread_mask_clear(lock->mask);
220	lockinfo.lock_release(lock->handle);
221}
222
223void
224lockdflt_init()
225{
226    int i;
227
228    deflockinfo.rtli_version  = RTLI_VERSION;
229    deflockinfo.lock_create   = def_lock_create;
230    deflockinfo.lock_destroy  = def_lock_destroy;
231    deflockinfo.rlock_acquire = def_rlock_acquire;
232    deflockinfo.wlock_acquire = def_wlock_acquire;
233    deflockinfo.lock_release  = def_lock_release;
234    deflockinfo.thread_set_flag = def_thread_set_flag;
235    deflockinfo.thread_clr_flag = def_thread_clr_flag;
236    deflockinfo.at_fork = NULL;
237
238    for (i = 0; i < RTLD_LOCK_CNT; i++) {
239	    rtld_locks[i].mask   = (1 << i);
240	    rtld_locks[i].handle = NULL;
241    }
242
243    memcpy(&lockinfo, &deflockinfo, sizeof(lockinfo));
244    _rtld_thread_init(NULL);
245    /*
246     * Construct a mask to block all signals except traps which might
247     * conceivably be generated within the dynamic linker itself.
248     */
249    sigfillset(&fullsigmask);
250    sigdelset(&fullsigmask, SIGILL);
251    sigdelset(&fullsigmask, SIGTRAP);
252    sigdelset(&fullsigmask, SIGABRT);
253    sigdelset(&fullsigmask, SIGEMT);
254    sigdelset(&fullsigmask, SIGFPE);
255    sigdelset(&fullsigmask, SIGBUS);
256    sigdelset(&fullsigmask, SIGSEGV);
257    sigdelset(&fullsigmask, SIGSYS);
258}
259
260/*
261 * Callback function to allow threads implementation to
262 * register their own locking primitives if the default
263 * one is not suitable.
264 * The current context should be the only context
265 * executing at the invocation time.
266 */
267void
268_rtld_thread_init(struct RtldLockInfo *pli)
269{
270	int flags, i;
271	void *locks[RTLD_LOCK_CNT];
272
273	/* disable all locking while this function is running */
274	flags =	thread_mask_set(~0);
275
276	if (pli == NULL)
277		pli = &deflockinfo;
278
279
280	for (i = 0; i < RTLD_LOCK_CNT; i++)
281		if ((locks[i] = pli->lock_create()) == NULL)
282			break;
283
284	if (i < RTLD_LOCK_CNT) {
285		while (--i >= 0)
286			pli->lock_destroy(locks[i]);
287		abort();
288	}
289
290	for (i = 0; i < RTLD_LOCK_CNT; i++) {
291		if (rtld_locks[i].handle == NULL)
292			continue;
293		if (flags & rtld_locks[i].mask)
294			lockinfo.lock_release(rtld_locks[i].handle);
295		lockinfo.lock_destroy(rtld_locks[i].handle);
296	}
297
298	for (i = 0; i < RTLD_LOCK_CNT; i++) {
299		rtld_locks[i].handle = locks[i];
300		if (flags & rtld_locks[i].mask)
301			pli->wlock_acquire(rtld_locks[i].handle);
302	}
303
304	lockinfo.lock_create = pli->lock_create;
305	lockinfo.lock_destroy = pli->lock_destroy;
306	lockinfo.rlock_acquire = pli->rlock_acquire;
307	lockinfo.wlock_acquire = pli->wlock_acquire;
308	lockinfo.lock_release  = pli->lock_release;
309	lockinfo.thread_set_flag = pli->thread_set_flag;
310	lockinfo.thread_clr_flag = pli->thread_clr_flag;
311	lockinfo.at_fork = pli->at_fork;
312
313	/* restore thread locking state, this time with new locks */
314	thread_mask_clear(~0);
315	thread_mask_set(flags);
316	dbg("_rtld_thread_init: done");
317}
318