linux_fork.c revision 293492
1/*-
2 * Copyright (c) 2004 Tim J. Robbins
3 * Copyright (c) 2002 Doug Rabson
4 * Copyright (c) 2000 Marcel Moolenaar
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer
12 *    in this position and unchanged.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
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: stable/10/sys/compat/linux/linux_fork.c 293492 2016-01-09 15:12:31Z dchagin $");
31
32#include "opt_compat.h"
33#include "opt_kdtrace.h"
34
35#include <sys/param.h>
36#include <sys/systm.h>
37#include <sys/imgact.h>
38#include <sys/lock.h>
39#include <sys/mutex.h>
40#include <sys/proc.h>
41#include <sys/sched.h>
42#include <sys/sdt.h>
43#include <sys/sx.h>
44#include <sys/unistd.h>
45#include <sys/wait.h>
46
47#ifdef COMPAT_LINUX32
48#include <machine/../linux32/linux.h>
49#include <machine/../linux32/linux32_proto.h>
50#else
51#include <machine/../linux/linux.h>
52#include <machine/../linux/linux_proto.h>
53#endif
54#include <compat/linux/linux_dtrace.h>
55#include <compat/linux/linux_signal.h>
56#include <compat/linux/linux_emul.h>
57#include <compat/linux/linux_misc.h>
58
59/* DTrace init */
60LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
61
62/* Linuxulator-global DTrace probes */
63LIN_SDT_PROBE_DECLARE(locks, emul_lock, locked);
64LIN_SDT_PROBE_DECLARE(locks, emul_lock, unlock);
65
66
67int
68linux_fork(struct thread *td, struct linux_fork_args *args)
69{
70	int error;
71	struct proc *p2;
72	struct thread *td2;
73
74#ifdef DEBUG
75	if (ldebug(fork))
76		printf(ARGS(fork, ""));
77#endif
78
79	if ((error = fork1(td, RFFDG | RFPROC | RFSTOPPED, 0, &p2, NULL, 0))
80	    != 0)
81		return (error);
82
83	td->td_retval[0] = p2->p_pid;
84	td->td_retval[1] = 0;
85
86	error = linux_proc_init(td, td->td_retval[0], 0);
87	if (error)
88		return (error);
89
90	td2 = FIRST_THREAD_IN_PROC(p2);
91
92	/*
93	 * Make this runnable after we are finished with it.
94	 */
95	thread_lock(td2);
96	TD_SET_CAN_RUN(td2);
97	sched_add(td2, SRQ_BORING);
98	thread_unlock(td2);
99
100	return (0);
101}
102
103int
104linux_vfork(struct thread *td, struct linux_vfork_args *args)
105{
106	int error;
107	struct proc *p2;
108	struct thread *td2;
109
110#ifdef DEBUG
111	if (ldebug(vfork))
112		printf(ARGS(vfork, ""));
113#endif
114
115	/* Exclude RFPPWAIT */
116	if ((error = fork1(td, RFFDG | RFPROC | RFMEM | RFSTOPPED, 0, &p2,
117	    NULL, 0)) != 0)
118		return (error);
119
120   	td->td_retval[0] = p2->p_pid;
121
122	error = linux_proc_init(td, td->td_retval[0], 0);
123	if (error)
124		return (error);
125
126	PROC_LOCK(p2);
127	p2->p_flag |= P_PPWAIT;
128	PROC_UNLOCK(p2);
129
130	td2 = FIRST_THREAD_IN_PROC(p2);
131
132	/*
133	 * Make this runnable after we are finished with it.
134	 */
135	thread_lock(td2);
136	TD_SET_CAN_RUN(td2);
137	sched_add(td2, SRQ_BORING);
138	thread_unlock(td2);
139
140	/* wait for the children to exit, ie. emulate vfork */
141	PROC_LOCK(p2);
142	while (p2->p_flag & P_PPWAIT)
143		cv_wait(&p2->p_pwait, &p2->p_mtx);
144	PROC_UNLOCK(p2);
145
146	return (0);
147}
148
149int
150linux_clone(struct thread *td, struct linux_clone_args *args)
151{
152	int error, ff = RFPROC | RFSTOPPED;
153	struct proc *p2;
154	struct thread *td2;
155	int exit_signal;
156	struct linux_emuldata *em;
157
158#ifdef DEBUG
159	if (ldebug(clone)) {
160		printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, "
161		    "child tid: %p"), (unsigned)args->flags,
162		    args->stack, args->parent_tidptr, args->child_tidptr);
163	}
164#endif
165
166	exit_signal = args->flags & 0x000000ff;
167	if (LINUX_SIG_VALID(exit_signal)) {
168		if (exit_signal <= LINUX_SIGTBLSZ)
169			exit_signal =
170			    linux_to_bsd_signal[_SIG_IDX(exit_signal)];
171	} else if (exit_signal != 0)
172		return (EINVAL);
173
174	if (args->flags & LINUX_CLONE_VM)
175		ff |= RFMEM;
176	if (args->flags & LINUX_CLONE_SIGHAND)
177		ff |= RFSIGSHARE;
178	/*
179	 * XXX: In Linux, sharing of fs info (chroot/cwd/umask)
180	 * and open files is independant.  In FreeBSD, its in one
181	 * structure but in reality it does not cause any problems
182	 * because both of these flags are usually set together.
183	 */
184	if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS)))
185		ff |= RFFDG;
186
187	/*
188	 * Attempt to detect when linux_clone(2) is used for creating
189	 * kernel threads. Unfortunately despite the existence of the
190	 * CLONE_THREAD flag, version of linuxthreads package used in
191	 * most popular distros as of beginning of 2005 doesn't make
192	 * any use of it. Therefore, this detection relies on
193	 * empirical observation that linuxthreads sets certain
194	 * combination of flags, so that we can make more or less
195	 * precise detection and notify the FreeBSD kernel that several
196	 * processes are in fact part of the same threading group, so
197	 * that special treatment is necessary for signal delivery
198	 * between those processes and fd locking.
199	 */
200	if ((args->flags & 0xffffff00) == LINUX_THREADING_FLAGS)
201		ff |= RFTHREAD;
202
203	if (args->flags & LINUX_CLONE_PARENT_SETTID)
204		if (args->parent_tidptr == NULL)
205			return (EINVAL);
206
207	error = fork1(td, ff, 0, &p2, NULL, 0);
208	if (error)
209		return (error);
210
211	if (args->flags & (LINUX_CLONE_PARENT | LINUX_CLONE_THREAD)) {
212	   	sx_xlock(&proctree_lock);
213		PROC_LOCK(p2);
214		proc_reparent(p2, td->td_proc->p_pptr);
215		PROC_UNLOCK(p2);
216		sx_xunlock(&proctree_lock);
217	}
218
219	/* create the emuldata */
220	error = linux_proc_init(td, p2->p_pid, args->flags);
221	/* reference it - no need to check this */
222	em = em_find(p2, EMUL_DOLOCK);
223	KASSERT(em != NULL, ("clone: emuldata not found."));
224	/* and adjust it */
225
226	if (args->flags & LINUX_CLONE_THREAD) {
227#ifdef notyet
228	   	PROC_LOCK(p2);
229	   	p2->p_pgrp = td->td_proc->p_pgrp;
230	   	PROC_UNLOCK(p2);
231#endif
232		exit_signal = 0;
233	}
234
235	if (args->flags & LINUX_CLONE_CHILD_SETTID)
236		em->child_set_tid = args->child_tidptr;
237	else
238	   	em->child_set_tid = NULL;
239
240	if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
241		em->child_clear_tid = args->child_tidptr;
242	else
243	   	em->child_clear_tid = NULL;
244
245	EMUL_UNLOCK(&emul_lock);
246
247	if (args->flags & LINUX_CLONE_PARENT_SETTID) {
248		error = copyout(&p2->p_pid, args->parent_tidptr,
249		    sizeof(p2->p_pid));
250		if (error)
251			printf(LMSG("copyout failed!"));
252	}
253
254	PROC_LOCK(p2);
255	p2->p_sigparent = exit_signal;
256	PROC_UNLOCK(p2);
257	td2 = FIRST_THREAD_IN_PROC(p2);
258	/*
259	 * In a case of stack = NULL, we are supposed to COW calling process
260	 * stack. This is what normal fork() does, so we just keep tf_rsp arg
261	 * intact.
262	 */
263	if (args->stack)
264		linux_set_upcall_kse(td2, PTROUT(args->stack));
265
266	if (args->flags & LINUX_CLONE_SETTLS)
267		linux_set_cloned_tls(td2, args->tls);
268
269#ifdef DEBUG
270	if (ldebug(clone))
271		printf(LMSG("clone: successful rfork to %d, "
272		    "stack %p sig = %d"), (int)p2->p_pid, args->stack,
273		    exit_signal);
274#endif
275	if (args->flags & LINUX_CLONE_VFORK) {
276	   	PROC_LOCK(p2);
277	   	p2->p_flag |= P_PPWAIT;
278	   	PROC_UNLOCK(p2);
279	}
280
281	/*
282	 * Make this runnable after we are finished with it.
283	 */
284	thread_lock(td2);
285	TD_SET_CAN_RUN(td2);
286	sched_add(td2, SRQ_BORING);
287	thread_unlock(td2);
288
289	td->td_retval[0] = p2->p_pid;
290	td->td_retval[1] = 0;
291
292	if (args->flags & LINUX_CLONE_VFORK) {
293		/* wait for the children to exit, ie. emulate vfork */
294		PROC_LOCK(p2);
295		while (p2->p_flag & P_PPWAIT)
296			cv_wait(&p2->p_pwait, &p2->p_mtx);
297		PROC_UNLOCK(p2);
298	}
299
300	return (0);
301}
302
303int
304linux_exit(struct thread *td, struct linux_exit_args *args)
305{
306
307#ifdef DEBUG
308	if (ldebug(exit))
309		printf(ARGS(exit, "%d"), args->rval);
310#endif
311
312	exit1(td, W_EXITCODE(args->rval, 0));
313		/* NOTREACHED */
314}
315