linux_fork.c revision 297305
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 297305 2016-03-27 06:55:49Z 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/ktr.h>
39#include <sys/lock.h>
40#include <sys/mutex.h>
41#include <sys/proc.h>
42#include <sys/racct.h>
43#include <sys/sched.h>
44#include <sys/syscallsubr.h>
45#include <sys/sx.h>
46#include <sys/unistd.h>
47#include <sys/wait.h>
48
49#include <vm/vm.h>
50#include <vm/pmap.h>
51#include <vm/vm_map.h>
52
53#ifdef COMPAT_LINUX32
54#include <machine/../linux32/linux.h>
55#include <machine/../linux32/linux32_proto.h>
56#else
57#include <machine/../linux/linux.h>
58#include <machine/../linux/linux_proto.h>
59#endif
60#include <compat/linux/linux_emul.h>
61#include <compat/linux/linux_futex.h>
62#include <compat/linux/linux_misc.h>
63#include <compat/linux/linux_util.h>
64
65int
66linux_fork(struct thread *td, struct linux_fork_args *args)
67{
68	int error;
69	struct proc *p2;
70	struct thread *td2;
71
72#ifdef DEBUG
73	if (ldebug(fork))
74		printf(ARGS(fork, ""));
75#endif
76
77	if ((error = fork1(td, RFFDG | RFPROC | RFSTOPPED, 0, &p2, NULL, 0))
78	    != 0)
79		return (error);
80
81	td2 = FIRST_THREAD_IN_PROC(p2);
82
83	linux_proc_init(td, td2, 0);
84
85	td->td_retval[0] = p2->p_pid;
86
87	/*
88	 * Make this runnable after we are finished with it.
89	 */
90	thread_lock(td2);
91	TD_SET_CAN_RUN(td2);
92	sched_add(td2, SRQ_BORING);
93	thread_unlock(td2);
94
95	return (0);
96}
97
98int
99linux_vfork(struct thread *td, struct linux_vfork_args *args)
100{
101	int error;
102	struct proc *p2;
103	struct thread *td2;
104
105#ifdef DEBUG
106	if (ldebug(vfork))
107		printf(ARGS(vfork, ""));
108#endif
109
110	if ((error = fork1(td, RFFDG | RFPROC | RFMEM | RFPPWAIT | RFSTOPPED,
111	    0, &p2, NULL, 0)) != 0)
112		return (error);
113
114	td2 = FIRST_THREAD_IN_PROC(p2);
115
116	linux_proc_init(td, td2, 0);
117
118   	td->td_retval[0] = p2->p_pid;
119
120	/*
121	 * Make this runnable after we are finished with it.
122	 */
123	thread_lock(td2);
124	TD_SET_CAN_RUN(td2);
125	sched_add(td2, SRQ_BORING);
126	thread_unlock(td2);
127
128	return (0);
129}
130
131static int
132linux_clone_proc(struct thread *td, struct linux_clone_args *args)
133{
134	int error, ff = RFPROC | RFSTOPPED;
135	struct proc *p2;
136	struct thread *td2;
137	int exit_signal;
138	struct linux_emuldata *em;
139
140#ifdef DEBUG
141	if (ldebug(clone)) {
142		printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, "
143		    "child tid: %p"), (unsigned)args->flags,
144		    args->stack, args->parent_tidptr, args->child_tidptr);
145	}
146#endif
147
148	exit_signal = args->flags & 0x000000ff;
149	if (LINUX_SIG_VALID(exit_signal)) {
150		exit_signal = linux_to_bsd_signal(exit_signal);
151	} else if (exit_signal != 0)
152		return (EINVAL);
153
154	if (args->flags & LINUX_CLONE_VM)
155		ff |= RFMEM;
156	if (args->flags & LINUX_CLONE_SIGHAND)
157		ff |= RFSIGSHARE;
158	/*
159	 * XXX: In Linux, sharing of fs info (chroot/cwd/umask)
160	 * and open files is independant.  In FreeBSD, its in one
161	 * structure but in reality it does not cause any problems
162	 * because both of these flags are usually set together.
163	 */
164	if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS)))
165		ff |= RFFDG;
166
167	if (args->flags & LINUX_CLONE_PARENT_SETTID)
168		if (args->parent_tidptr == NULL)
169			return (EINVAL);
170
171	if (args->flags & LINUX_CLONE_VFORK)
172		ff |= RFPPWAIT;
173
174	error = fork1(td, ff, 0, &p2, NULL, 0);
175	if (error)
176		return (error);
177
178	td2 = FIRST_THREAD_IN_PROC(p2);
179
180	/* create the emuldata */
181	linux_proc_init(td, td2, args->flags);
182
183	em = em_find(td2);
184	KASSERT(em != NULL, ("clone_proc: emuldata not found.\n"));
185
186	if (args->flags & LINUX_CLONE_CHILD_SETTID)
187		em->child_set_tid = args->child_tidptr;
188	else
189	   	em->child_set_tid = NULL;
190
191	if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
192		em->child_clear_tid = args->child_tidptr;
193	else
194	   	em->child_clear_tid = NULL;
195
196	if (args->flags & LINUX_CLONE_PARENT_SETTID) {
197		error = copyout(&p2->p_pid, args->parent_tidptr,
198		    sizeof(p2->p_pid));
199		if (error)
200			printf(LMSG("copyout failed!"));
201	}
202
203	PROC_LOCK(p2);
204	p2->p_sigparent = exit_signal;
205	PROC_UNLOCK(p2);
206	/*
207	 * In a case of stack = NULL, we are supposed to COW calling process
208	 * stack. This is what normal fork() does, so we just keep tf_rsp arg
209	 * intact.
210	 */
211	linux_set_upcall_kse(td2, PTROUT(args->stack));
212
213	if (args->flags & LINUX_CLONE_SETTLS)
214		linux_set_cloned_tls(td2, args->tls);
215
216	/*
217	 * If CLONE_PARENT is set, then the parent of the new process will be
218	 * the same as that of the calling process.
219	 */
220	if (args->flags & LINUX_CLONE_PARENT) {
221		sx_xlock(&proctree_lock);
222		PROC_LOCK(p2);
223		proc_reparent(p2, td->td_proc->p_pptr);
224		PROC_UNLOCK(p2);
225		sx_xunlock(&proctree_lock);
226	}
227
228#ifdef DEBUG
229	if (ldebug(clone))
230		printf(LMSG("clone: successful rfork to %d, "
231		    "stack %p sig = %d"), (int)p2->p_pid, args->stack,
232		    exit_signal);
233#endif
234
235	/*
236	 * Make this runnable after we are finished with it.
237	 */
238	thread_lock(td2);
239	TD_SET_CAN_RUN(td2);
240	sched_add(td2, SRQ_BORING);
241	thread_unlock(td2);
242
243	td->td_retval[0] = p2->p_pid;
244
245	return (0);
246}
247
248static int
249linux_clone_thread(struct thread *td, struct linux_clone_args *args)
250{
251	struct linux_emuldata *em;
252	struct thread *newtd;
253	struct proc *p;
254	int error;
255
256#ifdef DEBUG
257	if (ldebug(clone)) {
258		printf(ARGS(clone, "thread: flags %x, stack %p, parent tid: %p, "
259		    "child tid: %p"), (unsigned)args->flags,
260		    args->stack, args->parent_tidptr, args->child_tidptr);
261	}
262#endif
263
264	LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p",
265	    td->td_tid, (unsigned)args->flags,
266	    args->parent_tidptr, args->child_tidptr);
267
268	if (args->flags & LINUX_CLONE_PARENT_SETTID)
269		if (args->parent_tidptr == NULL)
270			return (EINVAL);
271
272	/* Threads should be created with own stack */
273	if (args->stack == NULL)
274		return (EINVAL);
275
276	p = td->td_proc;
277
278#ifdef RACCT
279	if (racct_enable) {
280		PROC_LOCK(p);
281		error = racct_add(p, RACCT_NTHR, 1);
282		PROC_UNLOCK(p);
283		if (error != 0)
284			return (EPROCLIM);
285	}
286#endif
287
288	/* Initialize our td */
289	error = kern_thr_alloc(p, 0, &newtd);
290	if (error)
291		goto fail;
292
293	cpu_set_upcall(newtd, td);
294
295	bzero(&newtd->td_startzero,
296	    __rangeof(struct thread, td_startzero, td_endzero));
297	bcopy(&td->td_startcopy, &newtd->td_startcopy,
298	    __rangeof(struct thread, td_startcopy, td_endcopy));
299
300	newtd->td_proc = p;
301	newtd->td_ucred = crhold(td->td_ucred);
302
303	/* create the emuldata */
304	linux_proc_init(td, newtd, args->flags);
305
306	em = em_find(newtd);
307	KASSERT(em != NULL, ("clone_thread: emuldata not found.\n"));
308
309	if (args->flags & LINUX_CLONE_SETTLS)
310		linux_set_cloned_tls(newtd, args->tls);
311
312	if (args->flags & LINUX_CLONE_CHILD_SETTID)
313		em->child_set_tid = args->child_tidptr;
314	else
315	   	em->child_set_tid = NULL;
316
317	if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
318		em->child_clear_tid = args->child_tidptr;
319	else
320	   	em->child_clear_tid = NULL;
321
322	cpu_thread_clean(newtd);
323
324	linux_set_upcall_kse(newtd, PTROUT(args->stack));
325
326	PROC_LOCK(p);
327	p->p_flag |= P_HADTHREADS;
328	bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
329
330	if (args->flags & LINUX_CLONE_PARENT)
331		thread_link(newtd, p->p_pptr);
332	else
333		thread_link(newtd, p);
334
335	thread_lock(td);
336	/* let the scheduler know about these things. */
337	sched_fork_thread(td, newtd);
338	thread_unlock(td);
339	if (P_SHOULDSTOP(p))
340		newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
341	PROC_UNLOCK(p);
342
343	tidhash_add(newtd);
344
345#ifdef DEBUG
346	if (ldebug(clone))
347		printf(ARGS(clone, "successful clone to %d, stack %p"),
348		(int)newtd->td_tid, args->stack);
349#endif
350
351	LINUX_CTR2(clone_thread, "thread(%d) successful clone to %d",
352	    td->td_tid, newtd->td_tid);
353
354	if (args->flags & LINUX_CLONE_PARENT_SETTID) {
355		error = copyout(&newtd->td_tid, args->parent_tidptr,
356		    sizeof(newtd->td_tid));
357		if (error)
358			printf(LMSG("clone_thread: copyout failed!"));
359	}
360
361	/*
362	 * Make this runnable after we are finished with it.
363	 */
364	thread_lock(newtd);
365	TD_SET_CAN_RUN(newtd);
366	sched_add(newtd, SRQ_BORING);
367	thread_unlock(newtd);
368
369	td->td_retval[0] = newtd->td_tid;
370
371	return (0);
372
373fail:
374#ifdef RACCT
375	if (racct_enable) {
376		PROC_LOCK(p);
377		racct_sub(p, RACCT_NTHR, 1);
378		PROC_UNLOCK(p);
379	}
380#endif
381	return (error);
382}
383
384int
385linux_clone(struct thread *td, struct linux_clone_args *args)
386{
387
388	if (args->flags & LINUX_CLONE_THREAD)
389		return (linux_clone_thread(td, args));
390	else
391		return (linux_clone_proc(td, args));
392}
393
394int
395linux_exit(struct thread *td, struct linux_exit_args *args)
396{
397	struct linux_emuldata *em;
398
399	em = em_find(td);
400	KASSERT(em != NULL, ("exit: emuldata not found.\n"));
401
402	LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval);
403
404	linux_thread_detach(td);
405
406	/*
407	 * XXX. When the last two threads of a process
408	 * exit via pthread_exit() try thr_exit() first.
409	 */
410	kern_thr_exit(td);
411	exit1(td, W_EXITCODE(args->rval, 0));
412		/* NOTREACHED */
413}
414
415int
416linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args)
417{
418	struct linux_emuldata *em;
419
420	em = em_find(td);
421	KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n"));
422
423	em->child_clear_tid = args->tidptr;
424
425	td->td_retval[0] = em->em_tid;
426
427	LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d",
428	    em->em_tid, args->tidptr, td->td_retval[0]);
429
430	return (0);
431}
432
433void
434linux_thread_detach(struct thread *td)
435{
436	struct linux_sys_futex_args cup;
437	struct linux_emuldata *em;
438	int *child_clear_tid;
439	int error;
440
441	em = em_find(td);
442	KASSERT(em != NULL, ("thread_detach: emuldata not found.\n"));
443
444	LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid);
445
446	release_futexes(td, em);
447
448	child_clear_tid = em->child_clear_tid;
449
450	if (child_clear_tid != NULL) {
451
452		LINUX_CTR2(thread_detach, "thread(%d) %p",
453		    em->em_tid, child_clear_tid);
454
455		error = suword32(child_clear_tid, 0);
456		if (error != 0)
457			return;
458
459		cup.uaddr = child_clear_tid;
460		cup.op = LINUX_FUTEX_WAKE;
461		cup.val = 1;		/* wake one */
462		cup.timeout = NULL;
463		cup.uaddr2 = NULL;
464		cup.val3 = 0;
465		error = linux_sys_futex(td, &cup);
466		/*
467		 * this cannot happen at the moment and if this happens it
468		 * probably means there is a user space bug
469		 */
470		if (error != 0)
471			linux_msg(td, "futex stuff in thread_detach failed.");
472	}
473}
474