1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
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/param.h>
32 #include <sys/ktr.h>
33 #include <sys/lock.h>
34 #include <sys/mutex.h>
35 #include <sys/proc.h>
36 #include <sys/ptrace.h>
37 #include <sys/racct.h>
38 #include <sys/sched.h>
39 #include <sys/syscallsubr.h>
40 #include <sys/sx.h>
41 #include <sys/umtxvar.h>
42 #include <sys/unistd.h>
43
44 #include <vm/vm.h>
45 #include <vm/pmap.h>
46 #include <vm/vm_map.h>
47
48 #ifdef COMPAT_LINUX32
49 #include <machine/../linux32/linux.h>
50 #include <machine/../linux32/linux32_proto.h>
51 #else
52 #include <machine/../linux/linux.h>
53 #include <machine/../linux/linux_proto.h>
54 #endif
55 #include <compat/linux/linux.h>
56 #include <compat/linux/linux_emul.h>
57 #include <compat/linux/linux_fork.h>
58 #include <compat/linux/linux_futex.h>
59 #include <compat/linux/linux_mib.h>
60 #include <compat/linux/linux_misc.h>
61 #include <compat/linux/linux_util.h>
62
63 #ifdef LINUX_LEGACY_SYSCALLS
64 int
linux_fork(struct thread * td,struct linux_fork_args * args)65 linux_fork(struct thread *td, struct linux_fork_args *args)
66 {
67 struct fork_req fr;
68 int error;
69 struct proc *p2;
70 struct thread *td2;
71
72 bzero(&fr, sizeof(fr));
73 fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
74 fr.fr_procp = &p2;
75 if ((error = fork1(td, &fr)) != 0)
76 return (error);
77
78 td2 = FIRST_THREAD_IN_PROC(p2);
79
80 linux_proc_init(td, td2, false);
81
82 td->td_retval[0] = p2->p_pid;
83
84 /*
85 * Make this runnable after we are finished with it.
86 */
87 thread_lock(td2);
88 TD_SET_CAN_RUN(td2);
89 sched_add(td2, SRQ_BORING);
90
91 return (0);
92 }
93
94 int
linux_vfork(struct thread * td,struct linux_vfork_args * args)95 linux_vfork(struct thread *td, struct linux_vfork_args *args)
96 {
97 struct fork_req fr;
98 int error;
99 struct proc *p2;
100 struct thread *td2;
101
102 bzero(&fr, sizeof(fr));
103 fr.fr_flags = RFFDG | RFPROC | RFMEM | RFPPWAIT | RFSTOPPED;
104 fr.fr_procp = &p2;
105 if ((error = fork1(td, &fr)) != 0)
106 return (error);
107
108 td2 = FIRST_THREAD_IN_PROC(p2);
109
110 linux_proc_init(td, td2, false);
111
112 td->td_retval[0] = p2->p_pid;
113
114 /*
115 * Make this runnable after we are finished with it.
116 */
117 thread_lock(td2);
118 TD_SET_CAN_RUN(td2);
119 sched_add(td2, SRQ_BORING);
120
121 return (0);
122 }
123 #endif
124
125 static int
linux_clone_proc(struct thread * td,struct l_clone_args * args)126 linux_clone_proc(struct thread *td, struct l_clone_args *args)
127 {
128 struct fork_req fr;
129 int error, ff, f2;
130 struct proc *p2;
131 struct thread *td2;
132 int exit_signal;
133 struct linux_emuldata *em;
134
135 f2 = 0;
136 ff = RFPROC | RFSTOPPED;
137 if (LINUX_SIG_VALID(args->exit_signal)) {
138 exit_signal = linux_to_bsd_signal(args->exit_signal);
139 } else if (args->exit_signal != 0)
140 return (EINVAL);
141 else
142 exit_signal = 0;
143
144 if (args->flags & LINUX_CLONE_VM)
145 ff |= RFMEM;
146 if (args->flags & LINUX_CLONE_SIGHAND)
147 ff |= RFSIGSHARE;
148 if ((args->flags & LINUX_CLONE_CLEAR_SIGHAND) != 0)
149 f2 |= FR2_DROPSIG_CAUGHT;
150 if (args->flags & LINUX_CLONE_FILES) {
151 if (!(args->flags & LINUX_CLONE_FS))
152 f2 |= FR2_SHARE_PATHS;
153 } else {
154 ff |= RFFDG;
155 if (args->flags & LINUX_CLONE_FS)
156 f2 |= FR2_SHARE_PATHS;
157 }
158
159 if (args->flags & LINUX_CLONE_PARENT_SETTID)
160 if (args->parent_tid == NULL)
161 return (EINVAL);
162
163 if (args->flags & LINUX_CLONE_VFORK)
164 ff |= RFPPWAIT;
165
166 bzero(&fr, sizeof(fr));
167 fr.fr_flags = ff;
168 fr.fr_flags2 = f2;
169 fr.fr_procp = &p2;
170 error = fork1(td, &fr);
171 if (error)
172 return (error);
173
174 td2 = FIRST_THREAD_IN_PROC(p2);
175
176 /* create the emuldata */
177 linux_proc_init(td, td2, false);
178
179 em = em_find(td2);
180 KASSERT(em != NULL, ("clone_proc: emuldata not found.\n"));
181
182 if (args->flags & LINUX_CLONE_CHILD_SETTID)
183 em->child_set_tid = args->child_tid;
184 else
185 em->child_set_tid = NULL;
186
187 if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
188 em->child_clear_tid = args->child_tid;
189 else
190 em->child_clear_tid = NULL;
191
192 if (args->flags & LINUX_CLONE_PARENT_SETTID) {
193 error = copyout(&p2->p_pid, args->parent_tid,
194 sizeof(p2->p_pid));
195 if (error)
196 linux_msg(td, "copyout p_pid failed!");
197 }
198
199 PROC_LOCK(p2);
200 p2->p_sigparent = exit_signal;
201 PROC_UNLOCK(p2);
202 /*
203 * In a case of stack = NULL, we are supposed to COW calling process
204 * stack. This is what normal fork() does, so we just keep tf_rsp arg
205 * intact.
206 */
207 linux_set_upcall(td2, args->stack);
208
209 if (args->flags & LINUX_CLONE_SETTLS)
210 linux_set_cloned_tls(td2, PTRIN(args->tls));
211
212 /*
213 * If CLONE_PARENT is set, then the parent of the new process will be
214 * the same as that of the calling process.
215 */
216 if (args->flags & LINUX_CLONE_PARENT) {
217 sx_xlock(&proctree_lock);
218 PROC_LOCK(p2);
219 proc_reparent(p2, td->td_proc->p_pptr, true);
220 PROC_UNLOCK(p2);
221 sx_xunlock(&proctree_lock);
222 }
223
224 /*
225 * Make this runnable after we are finished with it.
226 */
227 thread_lock(td2);
228 TD_SET_CAN_RUN(td2);
229 sched_add(td2, SRQ_BORING);
230
231 td->td_retval[0] = p2->p_pid;
232
233 return (0);
234 }
235
236 static int
linux_clone_thread(struct thread * td,struct l_clone_args * args)237 linux_clone_thread(struct thread *td, struct l_clone_args *args)
238 {
239 struct linux_emuldata *em;
240 struct thread *newtd;
241 struct proc *p;
242 int error;
243
244 LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p",
245 td->td_tid, (unsigned)args->flags,
246 args->parent_tid, args->child_tid);
247
248 if ((args->flags & LINUX_CLONE_PARENT) != 0)
249 return (EINVAL);
250 if (args->flags & LINUX_CLONE_PARENT_SETTID)
251 if (args->parent_tid == NULL)
252 return (EINVAL);
253
254 /* Threads should be created with own stack */
255 if (PTRIN(args->stack) == NULL)
256 return (EINVAL);
257
258 p = td->td_proc;
259
260 #ifdef RACCT
261 if (racct_enable) {
262 PROC_LOCK(p);
263 error = racct_add(p, RACCT_NTHR, 1);
264 PROC_UNLOCK(p);
265 if (error != 0)
266 return (EPROCLIM);
267 }
268 #endif
269
270 /* Initialize our td */
271 error = kern_thr_alloc(p, 0, &newtd);
272 if (error)
273 goto fail;
274
275 bzero(&newtd->td_startzero,
276 __rangeof(struct thread, td_startzero, td_endzero));
277 bcopy(&td->td_startcopy, &newtd->td_startcopy,
278 __rangeof(struct thread, td_startcopy, td_endcopy));
279
280 newtd->td_proc = p;
281 thread_cow_get(newtd, td);
282
283 cpu_copy_thread(newtd, td);
284
285 /* create the emuldata */
286 linux_proc_init(td, newtd, true);
287
288 em = em_find(newtd);
289 KASSERT(em != NULL, ("clone_thread: emuldata not found.\n"));
290
291 if (args->flags & LINUX_CLONE_SETTLS)
292 linux_set_cloned_tls(newtd, PTRIN(args->tls));
293
294 if (args->flags & LINUX_CLONE_CHILD_SETTID)
295 em->child_set_tid = args->child_tid;
296 else
297 em->child_set_tid = NULL;
298
299 if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
300 em->child_clear_tid = args->child_tid;
301 else
302 em->child_clear_tid = NULL;
303
304 cpu_thread_clean(newtd);
305
306 linux_set_upcall(newtd, args->stack);
307
308 PROC_LOCK(p);
309 p->p_flag |= P_HADTHREADS;
310 thread_link(newtd, p);
311 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
312
313 thread_lock(td);
314 /* let the scheduler know about these things. */
315 sched_fork_thread(td, newtd);
316 thread_unlock(td);
317 if (P_SHOULDSTOP(p))
318 ast_sched(newtd, TDA_SUSPEND);
319
320 if (p->p_ptevents & PTRACE_LWP)
321 newtd->td_dbgflags |= TDB_BORN;
322 PROC_UNLOCK(p);
323
324 tidhash_add(newtd);
325
326 LINUX_CTR2(clone_thread, "thread(%d) successful clone to %d",
327 td->td_tid, newtd->td_tid);
328
329 if (args->flags & LINUX_CLONE_PARENT_SETTID) {
330 error = copyout(&newtd->td_tid, args->parent_tid,
331 sizeof(newtd->td_tid));
332 if (error)
333 linux_msg(td, "clone_thread: copyout td_tid failed!");
334 }
335
336 /*
337 * Make this runnable after we are finished with it.
338 */
339 thread_lock(newtd);
340 TD_SET_CAN_RUN(newtd);
341 sched_add(newtd, SRQ_BORING);
342
343 td->td_retval[0] = newtd->td_tid;
344
345 return (0);
346
347 fail:
348 #ifdef RACCT
349 if (racct_enable) {
350 PROC_LOCK(p);
351 racct_sub(p, RACCT_NTHR, 1);
352 PROC_UNLOCK(p);
353 }
354 #endif
355 return (error);
356 }
357
358 int
linux_clone(struct thread * td,struct linux_clone_args * args)359 linux_clone(struct thread *td, struct linux_clone_args *args)
360 {
361 struct l_clone_args ca = {
362 .flags = (lower_32_bits(args->flags) & ~LINUX_CSIGNAL),
363 .child_tid = args->child_tidptr,
364 .parent_tid = args->parent_tidptr,
365 .exit_signal = (lower_32_bits(args->flags) & LINUX_CSIGNAL),
366 .stack = args->stack,
367 .tls = args->tls,
368 };
369
370 if (args->flags & LINUX_CLONE_THREAD)
371 return (linux_clone_thread(td, &ca));
372 else
373 return (linux_clone_proc(td, &ca));
374 }
375
376
377 static int
linux_clone3_args_valid(struct l_user_clone_args * uca)378 linux_clone3_args_valid(struct l_user_clone_args *uca)
379 {
380
381 /* Verify that no unknown flags are passed along. */
382 if ((uca->flags & ~(LINUX_CLONE_LEGACY_FLAGS |
383 LINUX_CLONE_CLEAR_SIGHAND | LINUX_CLONE_INTO_CGROUP)) != 0)
384 return (EINVAL);
385 if ((uca->flags & (LINUX_CLONE_DETACHED | LINUX_CSIGNAL)) != 0)
386 return (EINVAL);
387
388 if ((uca->flags & (LINUX_CLONE_SIGHAND | LINUX_CLONE_CLEAR_SIGHAND)) ==
389 (LINUX_CLONE_SIGHAND | LINUX_CLONE_CLEAR_SIGHAND))
390 return (EINVAL);
391 if ((uca->flags & (LINUX_CLONE_THREAD | LINUX_CLONE_PARENT)) != 0 &&
392 uca->exit_signal != 0)
393 return (EINVAL);
394
395 /* We don't support set_tid, only validate input. */
396 if (uca->set_tid_size > LINUX_MAX_PID_NS_LEVEL)
397 return (EINVAL);
398 if (uca->set_tid == 0 && uca->set_tid_size > 0)
399 return (EINVAL);
400 if (uca->set_tid != 0 && uca->set_tid_size == 0)
401 return (EINVAL);
402
403 if (uca->stack == 0 && uca->stack_size > 0)
404 return (EINVAL);
405 if (uca->stack != 0 && uca->stack_size == 0)
406 return (EINVAL);
407
408 /* Verify that higher 32bits of exit_signal are unset. */
409 if ((uca->exit_signal & ~(uint64_t)LINUX_CSIGNAL) != 0)
410 return (EINVAL);
411
412 /* Verify that no unsupported flags are passed along. */
413 if ((uca->flags & LINUX_CLONE_NEWTIME) != 0) {
414 LINUX_RATELIMIT_MSG("unsupported clone3 option CLONE_NEWTIME");
415 return (ENOSYS);
416 }
417 if ((uca->flags & LINUX_CLONE_INTO_CGROUP) != 0) {
418 LINUX_RATELIMIT_MSG("unsupported clone3 option CLONE_INTO_CGROUP");
419 return (ENOSYS);
420 }
421 if (uca->set_tid != 0 || uca->set_tid_size != 0) {
422 LINUX_RATELIMIT_MSG("unsupported clone3 set_tid");
423 return (ENOSYS);
424 }
425
426 return (0);
427 }
428
429 int
linux_clone3(struct thread * td,struct linux_clone3_args * args)430 linux_clone3(struct thread *td, struct linux_clone3_args *args)
431 {
432 struct l_user_clone_args *uca;
433 struct l_clone_args *ca;
434 size_t size;
435 int error;
436
437 if (args->usize > PAGE_SIZE)
438 return (E2BIG);
439 if (args->usize < LINUX_CLONE_ARGS_SIZE_VER0)
440 return (EINVAL);
441
442 /*
443 * usize can be less than size of struct clone_args, to avoid using
444 * of uninitialized data of struct clone_args, allocate at least
445 * sizeof(struct clone_args) storage and zero it.
446 */
447 size = max(args->usize, sizeof(*uca));
448 uca = malloc(size, M_LINUX, M_WAITOK | M_ZERO);
449 error = copyin(args->uargs, uca, args->usize);
450 if (error != 0)
451 goto out;
452 error = linux_clone3_args_valid(uca);
453 if (error != 0)
454 goto out;
455 ca = malloc(sizeof(*ca), M_LINUX, M_WAITOK | M_ZERO);
456 ca->flags = uca->flags;
457 ca->child_tid = PTRIN(uca->child_tid);
458 ca->parent_tid = PTRIN(uca->parent_tid);
459 ca->exit_signal = uca->exit_signal;
460 ca->stack = uca->stack + uca->stack_size;
461 ca->stack_size = uca->stack_size;
462 ca->tls = uca->tls;
463
464 if ((ca->flags & LINUX_CLONE_THREAD) != 0)
465 error = linux_clone_thread(td, ca);
466 else
467 error = linux_clone_proc(td, ca);
468 free(ca, M_LINUX);
469 out:
470 free(uca, M_LINUX);
471 return (error);
472 }
473
474 int
linux_exit(struct thread * td,struct linux_exit_args * args)475 linux_exit(struct thread *td, struct linux_exit_args *args)
476 {
477 struct linux_emuldata *em __diagused;
478
479 em = em_find(td);
480 KASSERT(em != NULL, ("exit: emuldata not found.\n"));
481
482 LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval);
483
484 linux_thread_detach(td);
485
486 /*
487 * XXX. When the last two threads of a process
488 * exit via pthread_exit() try thr_exit() first.
489 */
490 kern_thr_exit(td);
491 exit1(td, args->rval, 0);
492 /* NOTREACHED */
493 }
494
495 int
linux_set_tid_address(struct thread * td,struct linux_set_tid_address_args * args)496 linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args)
497 {
498 struct linux_emuldata *em;
499
500 em = em_find(td);
501 KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n"));
502
503 em->child_clear_tid = args->tidptr;
504
505 td->td_retval[0] = em->em_tid;
506
507 LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d",
508 em->em_tid, args->tidptr, td->td_retval[0]);
509
510 return (0);
511 }
512
513 void
linux_thread_detach(struct thread * td)514 linux_thread_detach(struct thread *td)
515 {
516 struct linux_emuldata *em;
517 int *child_clear_tid;
518 int error;
519
520 em = em_find(td);
521 KASSERT(em != NULL, ("thread_detach: emuldata not found.\n"));
522
523 LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid);
524
525 release_futexes(td, em);
526
527 child_clear_tid = em->child_clear_tid;
528
529 if (child_clear_tid != NULL) {
530 LINUX_CTR2(thread_detach, "thread(%d) %p",
531 em->em_tid, child_clear_tid);
532
533 error = suword32(child_clear_tid, 0);
534 if (error != 0)
535 return;
536
537 error = futex_wake(td, child_clear_tid, 1, false);
538 /*
539 * this cannot happen at the moment and if this happens it
540 * probably means there is a user space bug
541 */
542 if (error != 0)
543 linux_msg(td, "futex stuff in thread_detach failed.");
544 }
545
546 /*
547 * Do not rely on the robust list which is maintained by userspace,
548 * cleanup remaining pi (if any) after release_futexes anyway.
549 */
550 umtx_thread_exit(td);
551 }
552