xref: /freebsd/sys/compat/linux/linux_fork.c (revision 0957b409) 
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2004 Tim J. Robbins
5  * Copyright (c) 2002 Doug Rabson
6  * Copyright (c) 2000 Marcel Moolenaar
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer
14  *    in this position and unchanged.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
20  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
21  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
22  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
24  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
28  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29  */
30 
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
33 
34 #include "opt_compat.h"
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/imgact.h>
39 #include <sys/ktr.h>
40 #include <sys/lock.h>
41 #include <sys/mutex.h>
42 #include <sys/proc.h>
43 #include <sys/ptrace.h>
44 #include <sys/racct.h>
45 #include <sys/sched.h>
46 #include <sys/syscallsubr.h>
47 #include <sys/sx.h>
48 #include <sys/umtx.h>
49 #include <sys/unistd.h>
50 #include <sys/wait.h>
51 
52 #include <vm/vm.h>
53 #include <vm/pmap.h>
54 #include <vm/vm_map.h>
55 
56 #ifdef COMPAT_LINUX32
57 #include <machine/../linux32/linux.h>
58 #include <machine/../linux32/linux32_proto.h>
59 #else
60 #include <machine/../linux/linux.h>
61 #include <machine/../linux/linux_proto.h>
62 #endif
63 #include <compat/linux/linux_emul.h>
64 #include <compat/linux/linux_futex.h>
65 #include <compat/linux/linux_misc.h>
66 #include <compat/linux/linux_util.h>
67 
68 #ifdef LINUX_LEGACY_SYSCALLS
69 int
70 linux_fork(struct thread *td, struct linux_fork_args *args)
71 {
72 	struct fork_req fr;
73 	int error;
74 	struct proc *p2;
75 	struct thread *td2;
76 
77 #ifdef DEBUG
78 	if (ldebug(fork))
79 		printf(ARGS(fork, ""));
80 #endif
81 
82 	bzero(&fr, sizeof(fr));
83 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
84 	fr.fr_procp = &p2;
85 	if ((error = fork1(td, &fr)) != 0)
86 		return (error);
87 
88 	td2 = FIRST_THREAD_IN_PROC(p2);
89 
90 	linux_proc_init(td, td2, 0);
91 
92 	td->td_retval[0] = p2->p_pid;
93 
94 	/*
95 	 * Make this runnable after we are finished with it.
96 	 */
97 	thread_lock(td2);
98 	TD_SET_CAN_RUN(td2);
99 	sched_add(td2, SRQ_BORING);
100 	thread_unlock(td2);
101 
102 	return (0);
103 }
104 
105 int
106 linux_vfork(struct thread *td, struct linux_vfork_args *args)
107 {
108 	struct fork_req fr;
109 	int error;
110 	struct proc *p2;
111 	struct thread *td2;
112 
113 #ifdef DEBUG
114 	if (ldebug(vfork))
115 		printf(ARGS(vfork, ""));
116 #endif
117 
118 	bzero(&fr, sizeof(fr));
119 	fr.fr_flags = RFFDG | RFPROC | RFMEM | RFPPWAIT | RFSTOPPED;
120 	fr.fr_procp = &p2;
121 	if ((error = fork1(td, &fr)) != 0)
122 		return (error);
123 
124 	td2 = FIRST_THREAD_IN_PROC(p2);
125 
126 	linux_proc_init(td, td2, 0);
127 
128 	td->td_retval[0] = p2->p_pid;
129 
130 	/*
131 	 * Make this runnable after we are finished with it.
132 	 */
133 	thread_lock(td2);
134 	TD_SET_CAN_RUN(td2);
135 	sched_add(td2, SRQ_BORING);
136 	thread_unlock(td2);
137 
138 	return (0);
139 }
140 #endif
141 
142 static int
143 linux_clone_proc(struct thread *td, struct linux_clone_args *args)
144 {
145 	struct fork_req fr;
146 	int error, ff = RFPROC | RFSTOPPED;
147 	struct proc *p2;
148 	struct thread *td2;
149 	int exit_signal;
150 	struct linux_emuldata *em;
151 
152 #ifdef DEBUG
153 	if (ldebug(clone)) {
154 		printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, "
155 		    "child tid: %p"), (unsigned)args->flags,
156 		    args->stack, args->parent_tidptr, args->child_tidptr);
157 	}
158 #endif
159 
160 	exit_signal = args->flags & 0x000000ff;
161 	if (LINUX_SIG_VALID(exit_signal)) {
162 		exit_signal = linux_to_bsd_signal(exit_signal);
163 	} else if (exit_signal != 0)
164 		return (EINVAL);
165 
166 	if (args->flags & LINUX_CLONE_VM)
167 		ff |= RFMEM;
168 	if (args->flags & LINUX_CLONE_SIGHAND)
169 		ff |= RFSIGSHARE;
170 	/*
171 	 * XXX: In Linux, sharing of fs info (chroot/cwd/umask)
172 	 * and open files is independent.  In FreeBSD, its in one
173 	 * structure but in reality it does not cause any problems
174 	 * because both of these flags are usually set together.
175 	 */
176 	if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS)))
177 		ff |= RFFDG;
178 
179 	if (args->flags & LINUX_CLONE_PARENT_SETTID)
180 		if (args->parent_tidptr == NULL)
181 			return (EINVAL);
182 
183 	if (args->flags & LINUX_CLONE_VFORK)
184 		ff |= RFPPWAIT;
185 
186 	bzero(&fr, sizeof(fr));
187 	fr.fr_flags = ff;
188 	fr.fr_procp = &p2;
189 	error = fork1(td, &fr);
190 	if (error)
191 		return (error);
192 
193 	td2 = FIRST_THREAD_IN_PROC(p2);
194 
195 	/* create the emuldata */
196 	linux_proc_init(td, td2, args->flags);
197 
198 	em = em_find(td2);
199 	KASSERT(em != NULL, ("clone_proc: emuldata not found.\n"));
200 
201 	if (args->flags & LINUX_CLONE_CHILD_SETTID)
202 		em->child_set_tid = args->child_tidptr;
203 	else
204 		em->child_set_tid = NULL;
205 
206 	if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
207 		em->child_clear_tid = args->child_tidptr;
208 	else
209 		em->child_clear_tid = NULL;
210 
211 	if (args->flags & LINUX_CLONE_PARENT_SETTID) {
212 		error = copyout(&p2->p_pid, args->parent_tidptr,
213 		    sizeof(p2->p_pid));
214 		if (error)
215 			printf(LMSG("copyout failed!"));
216 	}
217 
218 	PROC_LOCK(p2);
219 	p2->p_sigparent = exit_signal;
220 	PROC_UNLOCK(p2);
221 	/*
222 	 * In a case of stack = NULL, we are supposed to COW calling process
223 	 * stack. This is what normal fork() does, so we just keep tf_rsp arg
224 	 * intact.
225 	 */
226 	linux_set_upcall_kse(td2, PTROUT(args->stack));
227 
228 	if (args->flags & LINUX_CLONE_SETTLS)
229 		linux_set_cloned_tls(td2, args->tls);
230 
231 	/*
232 	 * If CLONE_PARENT is set, then the parent of the new process will be
233 	 * the same as that of the calling process.
234 	 */
235 	if (args->flags & LINUX_CLONE_PARENT) {
236 		sx_xlock(&proctree_lock);
237 		PROC_LOCK(p2);
238 		proc_reparent(p2, td->td_proc->p_pptr, true);
239 		PROC_UNLOCK(p2);
240 		sx_xunlock(&proctree_lock);
241 	}
242 
243 #ifdef DEBUG
244 	if (ldebug(clone))
245 		printf(LMSG("clone: successful rfork to %d, "
246 		    "stack %p sig = %d"), (int)p2->p_pid, args->stack,
247 		    exit_signal);
248 #endif
249 
250 	/*
251 	 * Make this runnable after we are finished with it.
252 	 */
253 	thread_lock(td2);
254 	TD_SET_CAN_RUN(td2);
255 	sched_add(td2, SRQ_BORING);
256 	thread_unlock(td2);
257 
258 	td->td_retval[0] = p2->p_pid;
259 
260 	return (0);
261 }
262 
263 static int
264 linux_clone_thread(struct thread *td, struct linux_clone_args *args)
265 {
266 	struct linux_emuldata *em;
267 	struct thread *newtd;
268 	struct proc *p;
269 	int error;
270 
271 #ifdef DEBUG
272 	if (ldebug(clone)) {
273 		printf(ARGS(clone, "thread: flags %x, stack %p, parent tid: %p, "
274 		    "child tid: %p"), (unsigned)args->flags,
275 		    args->stack, args->parent_tidptr, args->child_tidptr);
276 	}
277 #endif
278 
279 	LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p",
280 	    td->td_tid, (unsigned)args->flags,
281 	    args->parent_tidptr, args->child_tidptr);
282 
283 	if (args->flags & LINUX_CLONE_PARENT_SETTID)
284 		if (args->parent_tidptr == NULL)
285 			return (EINVAL);
286 
287 	/* Threads should be created with own stack */
288 	if (args->stack == NULL)
289 		return (EINVAL);
290 
291 	p = td->td_proc;
292 
293 #ifdef RACCT
294 	if (racct_enable) {
295 		PROC_LOCK(p);
296 		error = racct_add(p, RACCT_NTHR, 1);
297 		PROC_UNLOCK(p);
298 		if (error != 0)
299 			return (EPROCLIM);
300 	}
301 #endif
302 
303 	/* Initialize our td */
304 	error = kern_thr_alloc(p, 0, &newtd);
305 	if (error)
306 		goto fail;
307 
308 	cpu_copy_thread(newtd, td);
309 
310 	bzero(&newtd->td_startzero,
311 	    __rangeof(struct thread, td_startzero, td_endzero));
312 	bcopy(&td->td_startcopy, &newtd->td_startcopy,
313 	    __rangeof(struct thread, td_startcopy, td_endcopy));
314 
315 	newtd->td_proc = p;
316 	thread_cow_get(newtd, td);
317 
318 	/* create the emuldata */
319 	linux_proc_init(td, newtd, args->flags);
320 
321 	em = em_find(newtd);
322 	KASSERT(em != NULL, ("clone_thread: emuldata not found.\n"));
323 
324 	if (args->flags & LINUX_CLONE_SETTLS)
325 		linux_set_cloned_tls(newtd, args->tls);
326 
327 	if (args->flags & LINUX_CLONE_CHILD_SETTID)
328 		em->child_set_tid = args->child_tidptr;
329 	else
330 		em->child_set_tid = NULL;
331 
332 	if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
333 		em->child_clear_tid = args->child_tidptr;
334 	else
335 		em->child_clear_tid = NULL;
336 
337 	cpu_thread_clean(newtd);
338 
339 	linux_set_upcall_kse(newtd, PTROUT(args->stack));
340 
341 	PROC_LOCK(p);
342 	p->p_flag |= P_HADTHREADS;
343 	bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
344 
345 	if (args->flags & LINUX_CLONE_PARENT)
346 		thread_link(newtd, p->p_pptr);
347 	else
348 		thread_link(newtd, p);
349 
350 	thread_lock(td);
351 	/* let the scheduler know about these things. */
352 	sched_fork_thread(td, newtd);
353 	thread_unlock(td);
354 	if (P_SHOULDSTOP(p))
355 		newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
356 
357 	if (p->p_ptevents & PTRACE_LWP)
358 		newtd->td_dbgflags |= TDB_BORN;
359 	PROC_UNLOCK(p);
360 
361 	tidhash_add(newtd);
362 
363 #ifdef DEBUG
364 	if (ldebug(clone))
365 		printf(ARGS(clone, "successful clone to %d, stack %p"),
366 		(int)newtd->td_tid, args->stack);
367 #endif
368 
369 	LINUX_CTR2(clone_thread, "thread(%d) successful clone to %d",
370 	    td->td_tid, newtd->td_tid);
371 
372 	if (args->flags & LINUX_CLONE_PARENT_SETTID) {
373 		error = copyout(&newtd->td_tid, args->parent_tidptr,
374 		    sizeof(newtd->td_tid));
375 		if (error)
376 			printf(LMSG("clone_thread: copyout failed!"));
377 	}
378 
379 	/*
380 	 * Make this runnable after we are finished with it.
381 	 */
382 	thread_lock(newtd);
383 	TD_SET_CAN_RUN(newtd);
384 	sched_add(newtd, SRQ_BORING);
385 	thread_unlock(newtd);
386 
387 	td->td_retval[0] = newtd->td_tid;
388 
389 	return (0);
390 
391 fail:
392 #ifdef RACCT
393 	if (racct_enable) {
394 		PROC_LOCK(p);
395 		racct_sub(p, RACCT_NTHR, 1);
396 		PROC_UNLOCK(p);
397 	}
398 #endif
399 	return (error);
400 }
401 
402 int
403 linux_clone(struct thread *td, struct linux_clone_args *args)
404 {
405 
406 	if (args->flags & LINUX_CLONE_THREAD)
407 		return (linux_clone_thread(td, args));
408 	else
409 		return (linux_clone_proc(td, args));
410 }
411 
412 int
413 linux_exit(struct thread *td, struct linux_exit_args *args)
414 {
415 	struct linux_emuldata *em;
416 
417 	em = em_find(td);
418 	KASSERT(em != NULL, ("exit: emuldata not found.\n"));
419 
420 	LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval);
421 
422 	umtx_thread_exit(td);
423 
424 	linux_thread_detach(td);
425 
426 	/*
427 	 * XXX. When the last two threads of a process
428 	 * exit via pthread_exit() try thr_exit() first.
429 	 */
430 	kern_thr_exit(td);
431 	exit1(td, args->rval, 0);
432 		/* NOTREACHED */
433 }
434 
435 int
436 linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args)
437 {
438 	struct linux_emuldata *em;
439 
440 	em = em_find(td);
441 	KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n"));
442 
443 	em->child_clear_tid = args->tidptr;
444 
445 	td->td_retval[0] = em->em_tid;
446 
447 	LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d",
448 	    em->em_tid, args->tidptr, td->td_retval[0]);
449 
450 	return (0);
451 }
452 
453 void
454 linux_thread_detach(struct thread *td)
455 {
456 	struct linux_sys_futex_args cup;
457 	struct linux_emuldata *em;
458 	int *child_clear_tid;
459 	int error;
460 
461 	em = em_find(td);
462 	KASSERT(em != NULL, ("thread_detach: emuldata not found.\n"));
463 
464 	LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid);
465 
466 	release_futexes(td, em);
467 
468 	child_clear_tid = em->child_clear_tid;
469 
470 	if (child_clear_tid != NULL) {
471 
472 		LINUX_CTR2(thread_detach, "thread(%d) %p",
473 		    em->em_tid, child_clear_tid);
474 
475 		error = suword32(child_clear_tid, 0);
476 		if (error != 0)
477 			return;
478 
479 		cup.uaddr = child_clear_tid;
480 		cup.op = LINUX_FUTEX_WAKE;
481 		cup.val = 1;		/* wake one */
482 		cup.timeout = NULL;
483 		cup.uaddr2 = NULL;
484 		cup.val3 = 0;
485 		error = linux_sys_futex(td, &cup);
486 		/*
487 		 * this cannot happen at the moment and if this happens it
488 		 * probably means there is a user space bug
489 		 */
490 		if (error != 0)
491 			linux_msg(td, "futex stuff in thread_detach failed.");
492 	}
493 }
494