linux_emul.c revision 166930 
1/*-
2 * Copyright (c) 2006 Roman Divacky
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 *    in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 *    derived from this software without specific prior written permission
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: head/sys/compat/linux/linux_emul.c 166930 2007-02-23 22:29:24Z netchild $");
31
32#include "opt_compat.h"
33
34#include <sys/param.h>
35#include <sys/systm.h>
36#include <sys/imgact.h>
37#include <sys/kernel.h>
38#include <sys/lock.h>
39#include <sys/malloc.h>
40#include <sys/mutex.h>
41#include <sys/sx.h>
42#include <sys/proc.h>
43#include <sys/syscallsubr.h>
44#include <sys/sysproto.h>
45#include <sys/unistd.h>
46
47#include <compat/linux/linux_emul.h>
48#include <compat/linux/linux_futex.h>
49
50#ifdef COMPAT_LINUX32
51#include <machine/../linux32/linux.h>
52#include <machine/../linux32/linux32_proto.h>
53#else
54#include <machine/../linux/linux.h>
55#include <machine/../linux/linux_proto.h>
56#endif
57
58struct sx emul_shared_lock;
59struct sx emul_lock;
60
61/* this returns locked reference to the emuldata entry (if found) */
62struct linux_emuldata *
63em_find(struct proc *p, int locked)
64{
65	struct linux_emuldata *em;
66
67	if (locked == EMUL_DOLOCK)
68		EMUL_LOCK(&emul_lock);
69
70	em = p->p_emuldata;
71
72	if (em == NULL && locked == EMUL_DOLOCK)
73		EMUL_UNLOCK(&emul_lock);
74
75	return (em);
76}
77
78int
79linux_proc_init(struct thread *td, pid_t child, int flags)
80{
81	struct linux_emuldata *em, *p_em;
82	struct proc *p;
83
84	if (child != 0) {
85		/* non-exec call */
86		em = malloc(sizeof *em, M_LINUX, M_WAITOK | M_ZERO);
87		em->pid = child;
88		em->pdeath_signal = 0;
89		if (flags & CLONE_THREAD) {
90			/* handled later in the code */
91		} else {
92			struct linux_emuldata_shared *s;
93
94			s = malloc(sizeof *s, M_LINUX, M_WAITOK | M_ZERO);
95			s->refs = 1;
96			s->group_pid = child;
97
98			LIST_INIT(&s->threads);
99			em->shared = s;
100		}
101	} else {
102		/* lookup the old one */
103		em = em_find(td->td_proc, EMUL_DOLOCK);
104		KASSERT(em != NULL, ("proc_init: emuldata not found in exec case.\n"));
105	}
106
107	em->child_clear_tid = NULL;
108	em->child_set_tid = NULL;
109
110	/*
111	 * allocate the shared struct only in clone()/fork cases in the case
112	 * of clone() td = calling proc and child = pid of the newly created
113	 * proc
114	 */
115	if (child != 0) {
116		if (flags & CLONE_THREAD) {
117			/* lookup the parent */
118			/*
119			 * we dont have to lock the p_em because
120			 * its waiting for us in linux_clone so
121			 * there is no chance of it changing the
122			 * p_em->shared address
123			 */
124			p_em = em_find(td->td_proc, EMUL_DONTLOCK);
125			KASSERT(p_em != NULL, ("proc_init: parent emuldata not found for CLONE_THREAD\n"));
126			em->shared = p_em->shared;
127			EMUL_SHARED_WLOCK(&emul_shared_lock);
128			em->shared->refs++;
129			EMUL_SHARED_WUNLOCK(&emul_shared_lock);
130		} else {
131			/*
132			 * handled earlier to avoid malloc(M_WAITOK) with
133			 * rwlock held
134			 */
135		}
136	}
137	if (child != 0) {
138		EMUL_SHARED_WLOCK(&emul_shared_lock);
139		LIST_INSERT_HEAD(&em->shared->threads, em, threads);
140		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
141
142		p = pfind(child);
143		KASSERT(p != NULL, ("process not found in proc_init\n"));
144		p->p_emuldata = em;
145		PROC_UNLOCK(p);
146	} else
147		EMUL_UNLOCK(&emul_lock);
148
149	return (0);
150}
151
152void
153linux_proc_exit(void *arg __unused, struct proc *p)
154{
155	struct linux_emuldata *em;
156	int error;
157	struct thread *td = FIRST_THREAD_IN_PROC(p);
158	int *child_clear_tid;
159	struct proc *q, *nq;
160
161	if (__predict_true(p->p_sysent != &elf_linux_sysvec))
162		return;
163
164	/* find the emuldata */
165	em = em_find(p, EMUL_DOLOCK);
166
167	KASSERT(em != NULL, ("proc_exit: emuldata not found.\n"));
168
169	/* reparent all procs that are not a thread leader to initproc */
170	if (em->shared->group_pid != p->p_pid) {
171		child_clear_tid = em->child_clear_tid;
172		EMUL_UNLOCK(&emul_lock);
173		sx_xlock(&proctree_lock);
174		wakeup(initproc);
175		PROC_LOCK(p);
176		proc_reparent(p, initproc);
177		p->p_sigparent = SIGCHLD;
178		PROC_UNLOCK(p);
179		sx_xunlock(&proctree_lock);
180	} else {
181		child_clear_tid = em->child_clear_tid;
182		EMUL_UNLOCK(&emul_lock);
183	}
184
185	EMUL_SHARED_WLOCK(&emul_shared_lock);
186	LIST_REMOVE(em, threads);
187
188	em->shared->refs--;
189	if (em->shared->refs == 0)
190		free(em->shared, M_LINUX);
191	EMUL_SHARED_WUNLOCK(&emul_shared_lock);
192
193	if (child_clear_tid != NULL) {
194		struct linux_sys_futex_args cup;
195		int null = 0;
196
197		error = copyout(&null, child_clear_tid, sizeof(null));
198		if (error) {
199			free(em, M_LINUX);
200			return;
201		}
202
203		/* futexes stuff */
204		cup.uaddr = child_clear_tid;
205		cup.op = LINUX_FUTEX_WAKE;
206		cup.val = 0x7fffffff;	/* Awake everyone */
207		cup.timeout = NULL;
208		cup.uaddr2 = NULL;
209		cup.val3 = 0;
210		error = linux_sys_futex(FIRST_THREAD_IN_PROC(p), &cup);
211		/*
212		 * this cannot happen at the moment and if this happens it
213		 * probably mean there is a userspace bug
214		 */
215		if (error)
216			printf(LMSG("futex stuff in proc_exit failed.\n"));
217	}
218
219	/* clean the stuff up */
220	free(em, M_LINUX);
221
222	/* this is a little weird but rewritten from exit1() */
223	sx_xlock(&proctree_lock);
224	q = LIST_FIRST(&p->p_children);
225	for (; q != NULL; q = nq) {
226		nq = LIST_NEXT(q, p_sibling);
227		if (q->p_flag & P_WEXIT)
228			continue;
229		if (__predict_false(q->p_sysent != &elf_linux_sysvec))
230			continue;
231		em = em_find(q, EMUL_DOLOCK);
232		KASSERT(em != NULL, ("linux_reparent: emuldata not found: %i\n", q->p_pid));
233		if (em->pdeath_signal != 0) {
234			PROC_LOCK(q);
235			psignal(q, em->pdeath_signal);
236			PROC_UNLOCK(q);
237		}
238		EMUL_UNLOCK(&emul_lock);
239	}
240	sx_xunlock(&proctree_lock);
241}
242
243/*
244 * This is used in a case of transition from FreeBSD binary execing to linux binary
245 * in this case we create linux emuldata proc entry with the pid of the currently running
246 * process.
247 */
248void
249linux_proc_exec(void *arg __unused, struct proc *p, struct image_params *imgp)
250{
251	if (__predict_false(imgp->sysent == &elf_linux_sysvec
252	    && p->p_sysent != &elf_linux_sysvec))
253		linux_proc_init(FIRST_THREAD_IN_PROC(p), p->p_pid, 0);
254	if (__predict_false(imgp->sysent != &elf_linux_sysvec
255	    && p->p_sysent == &elf_linux_sysvec)) {
256		struct linux_emuldata *em;
257
258		/*
259		 * XXX:There's a race because here we assign p->p_emuldata NULL
260		 * but the process is still counted as linux one for a short
261 		 * time so some other process might reference it and try to
262 		 * access its p->p_emuldata and panicing on a NULL reference.
263		 */
264		em = em_find(p, EMUL_DONTLOCK);
265
266		KASSERT(em != NULL, ("proc_exec: emuldata not found.\n"));
267
268		EMUL_SHARED_WLOCK(&emul_shared_lock);
269		LIST_REMOVE(em, threads);
270
271		PROC_LOCK(p);
272		p->p_emuldata = NULL;
273		PROC_UNLOCK(p);
274
275		em->shared->refs--;
276		if (em->shared->refs == 0)
277			free(em->shared, M_LINUX);
278		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
279
280		free(em, M_LINUX);
281	}
282}
283
284void
285linux_schedtail(void *arg __unused, struct proc *p)
286{
287	struct linux_emuldata *em;
288	int error = 0;
289	int *child_set_tid;
290
291	if (__predict_true(p->p_sysent != &elf_linux_sysvec))
292		return;
293
294	/* find the emuldata */
295	em = em_find(p, EMUL_DOLOCK);
296
297	KASSERT(em != NULL, ("linux_schedtail: emuldata not found.\n"));
298	child_set_tid = em->child_set_tid;
299	EMUL_UNLOCK(&emul_lock);
300
301	if (child_set_tid != NULL)
302		error = copyout(&p->p_pid, (int *)child_set_tid,
303		    sizeof(p->p_pid));
304
305	return;
306}
307
308int
309linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args)
310{
311	struct linux_emuldata *em;
312
313#ifdef DEBUG
314	if (ldebug(set_tid_address))
315		printf(ARGS(set_tid_address, "%p"), args->tidptr);
316#endif
317
318	/* find the emuldata */
319	em = em_find(td->td_proc, EMUL_DOLOCK);
320
321	KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n"));
322
323	em->child_clear_tid = args->tidptr;
324	td->td_retval[0] = td->td_proc->p_pid;
325
326	EMUL_UNLOCK(&emul_lock);
327	return 0;
328}
329