FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_thr.c
1 /*-
2 * Copyright (c) 2003, Jeffrey Roberson <jeff@freebsd.org>
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 unmodified, this list of conditions, and the following
10 * disclaimer.
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 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 */
26
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD: releng/11.0/sys/kern/kern_thr.c 302195 2016-06-25 11:31:25Z kib $");
29
30 #include "opt_compat.h"
31 #include "opt_posix.h"
32 #include <sys/param.h>
33 #include <sys/kernel.h>
34 #include <sys/lock.h>
35 #include <sys/mutex.h>
36 #include <sys/priv.h>
37 #include <sys/proc.h>
38 #include <sys/posix4.h>
39 #include <sys/racct.h>
40 #include <sys/resourcevar.h>
41 #include <sys/rwlock.h>
42 #include <sys/sched.h>
43 #include <sys/sysctl.h>
44 #include <sys/smp.h>
45 #include <sys/syscallsubr.h>
46 #include <sys/sysent.h>
47 #include <sys/systm.h>
48 #include <sys/sysproto.h>
49 #include <sys/signalvar.h>
50 #include <sys/sysctl.h>
51 #include <sys/ucontext.h>
52 #include <sys/thr.h>
53 #include <sys/rtprio.h>
54 #include <sys/umtx.h>
55 #include <sys/limits.h>
56
57 #include <vm/vm_domain.h>
58
59 #include <machine/frame.h>
60
61 #include <security/audit/audit.h>
62
63 static SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0,
64 "thread allocation");
65
66 static int max_threads_per_proc = 1500;
67 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_per_proc, CTLFLAG_RW,
68 &max_threads_per_proc, 0, "Limit on threads per proc");
69
70 static int max_threads_hits;
71 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_hits, CTLFLAG_RD,
72 &max_threads_hits, 0, "kern.threads.max_threads_per_proc hit count");
73
74 #ifdef COMPAT_FREEBSD32
75
76 static inline int
77 suword_lwpid(void *addr, lwpid_t lwpid)
78 {
79 int error;
80
81 if (SV_CURPROC_FLAG(SV_LP64))
82 error = suword(addr, lwpid);
83 else
84 error = suword32(addr, lwpid);
85 return (error);
86 }
87
88 #else
89 #define suword_lwpid suword
90 #endif
91
92 /*
93 * System call interface.
94 */
95
96 struct thr_create_initthr_args {
97 ucontext_t ctx;
98 long *tid;
99 };
100
101 static int
102 thr_create_initthr(struct thread *td, void *thunk)
103 {
104 struct thr_create_initthr_args *args;
105
106 /* Copy out the child tid. */
107 args = thunk;
108 if (args->tid != NULL && suword_lwpid(args->tid, td->td_tid))
109 return (EFAULT);
110
111 return (set_mcontext(td, &args->ctx.uc_mcontext));
112 }
113
114 int
115 sys_thr_create(struct thread *td, struct thr_create_args *uap)
116 /* ucontext_t *ctx, long *id, int flags */
117 {
118 struct thr_create_initthr_args args;
119 int error;
120
121 if ((error = copyin(uap->ctx, &args.ctx, sizeof(args.ctx))))
122 return (error);
123 args.tid = uap->id;
124 return (thread_create(td, NULL, thr_create_initthr, &args));
125 }
126
127 int
128 sys_thr_new(struct thread *td, struct thr_new_args *uap)
129 /* struct thr_param * */
130 {
131 struct thr_param param;
132 int error;
133
134 if (uap->param_size < 0 || uap->param_size > sizeof(param))
135 return (EINVAL);
136 bzero(¶m, sizeof(param));
137 if ((error = copyin(uap->param, ¶m, uap->param_size)))
138 return (error);
139 return (kern_thr_new(td, ¶m));
140 }
141
142 static int
143 thr_new_initthr(struct thread *td, void *thunk)
144 {
145 stack_t stack;
146 struct thr_param *param;
147
148 /*
149 * Here we copy out tid to two places, one for child and one
150 * for parent, because pthread can create a detached thread,
151 * if parent wants to safely access child tid, it has to provide
152 * its storage, because child thread may exit quickly and
153 * memory is freed before parent thread can access it.
154 */
155 param = thunk;
156 if ((param->child_tid != NULL &&
157 suword_lwpid(param->child_tid, td->td_tid)) ||
158 (param->parent_tid != NULL &&
159 suword_lwpid(param->parent_tid, td->td_tid)))
160 return (EFAULT);
161
162 /* Set up our machine context. */
163 stack.ss_sp = param->stack_base;
164 stack.ss_size = param->stack_size;
165 /* Set upcall address to user thread entry function. */
166 cpu_set_upcall(td, param->start_func, param->arg, &stack);
167 /* Setup user TLS address and TLS pointer register. */
168 return (cpu_set_user_tls(td, param->tls_base));
169 }
170
171 int
172 kern_thr_new(struct thread *td, struct thr_param *param)
173 {
174 struct rtprio rtp, *rtpp;
175 int error;
176
177 rtpp = NULL;
178 if (param->rtp != 0) {
179 error = copyin(param->rtp, &rtp, sizeof(struct rtprio));
180 if (error)
181 return (error);
182 rtpp = &rtp;
183 }
184 return (thread_create(td, rtpp, thr_new_initthr, param));
185 }
186
187 int
188 thread_create(struct thread *td, struct rtprio *rtp,
189 int (*initialize_thread)(struct thread *, void *), void *thunk)
190 {
191 struct thread *newtd;
192 struct proc *p;
193 int error;
194
195 p = td->td_proc;
196
197 if (rtp != NULL) {
198 switch(rtp->type) {
199 case RTP_PRIO_REALTIME:
200 case RTP_PRIO_FIFO:
201 /* Only root can set scheduler policy */
202 if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0)
203 return (EPERM);
204 if (rtp->prio > RTP_PRIO_MAX)
205 return (EINVAL);
206 break;
207 case RTP_PRIO_NORMAL:
208 rtp->prio = 0;
209 break;
210 default:
211 return (EINVAL);
212 }
213 }
214
215 #ifdef RACCT
216 if (racct_enable) {
217 PROC_LOCK(p);
218 error = racct_add(p, RACCT_NTHR, 1);
219 PROC_UNLOCK(p);
220 if (error != 0)
221 return (EPROCLIM);
222 }
223 #endif
224
225 /* Initialize our td */
226 error = kern_thr_alloc(p, 0, &newtd);
227 if (error)
228 goto fail;
229
230 cpu_copy_thread(newtd, td);
231
232 bzero(&newtd->td_startzero,
233 __rangeof(struct thread, td_startzero, td_endzero));
234 bcopy(&td->td_startcopy, &newtd->td_startcopy,
235 __rangeof(struct thread, td_startcopy, td_endcopy));
236 newtd->td_proc = td->td_proc;
237 newtd->td_rb_list = newtd->td_rbp_list = newtd->td_rb_inact = 0;
238 thread_cow_get(newtd, td);
239
240 error = initialize_thread(newtd, thunk);
241 if (error != 0) {
242 thread_cow_free(newtd);
243 thread_free(newtd);
244 goto fail;
245 }
246
247 PROC_LOCK(p);
248 p->p_flag |= P_HADTHREADS;
249 thread_link(newtd, p);
250 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
251 thread_lock(td);
252 /* let the scheduler know about these things. */
253 sched_fork_thread(td, newtd);
254 thread_unlock(td);
255 if (P_SHOULDSTOP(p))
256 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
257 if (p->p_flag2 & P2_LWP_EVENTS)
258 newtd->td_dbgflags |= TDB_BORN;
259
260 /*
261 * Copy the existing thread VM policy into the new thread.
262 */
263 vm_domain_policy_localcopy(&newtd->td_vm_dom_policy,
264 &td->td_vm_dom_policy);
265
266 PROC_UNLOCK(p);
267
268 tidhash_add(newtd);
269
270 thread_lock(newtd);
271 if (rtp != NULL) {
272 if (!(td->td_pri_class == PRI_TIMESHARE &&
273 rtp->type == RTP_PRIO_NORMAL)) {
274 rtp_to_pri(rtp, newtd);
275 sched_prio(newtd, newtd->td_user_pri);
276 } /* ignore timesharing class */
277 }
278 TD_SET_CAN_RUN(newtd);
279 sched_add(newtd, SRQ_BORING);
280 thread_unlock(newtd);
281
282 return (0);
283
284 fail:
285 #ifdef RACCT
286 if (racct_enable) {
287 PROC_LOCK(p);
288 racct_sub(p, RACCT_NTHR, 1);
289 PROC_UNLOCK(p);
290 }
291 #endif
292 return (error);
293 }
294
295 int
296 sys_thr_self(struct thread *td, struct thr_self_args *uap)
297 /* long *id */
298 {
299 int error;
300
301 error = suword_lwpid(uap->id, (unsigned)td->td_tid);
302 if (error == -1)
303 return (EFAULT);
304 return (0);
305 }
306
307 int
308 sys_thr_exit(struct thread *td, struct thr_exit_args *uap)
309 /* long *state */
310 {
311
312 umtx_thread_exit(td);
313
314 /* Signal userland that it can free the stack. */
315 if ((void *)uap->state != NULL) {
316 suword_lwpid(uap->state, 1);
317 kern_umtx_wake(td, uap->state, INT_MAX, 0);
318 }
319
320 return (kern_thr_exit(td));
321 }
322
323 int
324 kern_thr_exit(struct thread *td)
325 {
326 struct proc *p;
327
328 p = td->td_proc;
329
330 /*
331 * If all of the threads in a process call this routine to
332 * exit (e.g. all threads call pthread_exit()), exactly one
333 * thread should return to the caller to terminate the process
334 * instead of the thread.
335 *
336 * Checking p_numthreads alone is not sufficient since threads
337 * might be committed to terminating while the PROC_LOCK is
338 * dropped in either ptracestop() or while removing this thread
339 * from the tidhash. Instead, the p_pendingexits field holds
340 * the count of threads in either of those states and a thread
341 * is considered the "last" thread if all of the other threads
342 * in a process are already terminating.
343 */
344 PROC_LOCK(p);
345 if (p->p_numthreads == p->p_pendingexits + 1) {
346 /*
347 * Ignore attempts to shut down last thread in the
348 * proc. This will actually call _exit(2) in the
349 * usermode trampoline when it returns.
350 */
351 PROC_UNLOCK(p);
352 return (0);
353 }
354
355 p->p_pendingexits++;
356 td->td_dbgflags |= TDB_EXIT;
357 if (p->p_flag & P_TRACED && p->p_flag2 & P2_LWP_EVENTS)
358 ptracestop(td, SIGTRAP);
359 PROC_UNLOCK(p);
360 tidhash_remove(td);
361 PROC_LOCK(p);
362 p->p_pendingexits--;
363
364 /*
365 * The check above should prevent all other threads from this
366 * process from exiting while the PROC_LOCK is dropped, so
367 * there must be at least one other thread other than the
368 * current thread.
369 */
370 KASSERT(p->p_numthreads > 1, ("too few threads"));
371 racct_sub(p, RACCT_NTHR, 1);
372 tdsigcleanup(td);
373 PROC_SLOCK(p);
374 thread_stopped(p);
375 thread_exit();
376 /* NOTREACHED */
377 }
378
379 int
380 sys_thr_kill(struct thread *td, struct thr_kill_args *uap)
381 /* long id, int sig */
382 {
383 ksiginfo_t ksi;
384 struct thread *ttd;
385 struct proc *p;
386 int error;
387
388 p = td->td_proc;
389 ksiginfo_init(&ksi);
390 ksi.ksi_signo = uap->sig;
391 ksi.ksi_code = SI_LWP;
392 ksi.ksi_pid = p->p_pid;
393 ksi.ksi_uid = td->td_ucred->cr_ruid;
394 if (uap->id == -1) {
395 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) {
396 error = EINVAL;
397 } else {
398 error = ESRCH;
399 PROC_LOCK(p);
400 FOREACH_THREAD_IN_PROC(p, ttd) {
401 if (ttd != td) {
402 error = 0;
403 if (uap->sig == 0)
404 break;
405 tdksignal(ttd, uap->sig, &ksi);
406 }
407 }
408 PROC_UNLOCK(p);
409 }
410 } else {
411 error = 0;
412 ttd = tdfind((lwpid_t)uap->id, p->p_pid);
413 if (ttd == NULL)
414 return (ESRCH);
415 if (uap->sig == 0)
416 ;
417 else if (!_SIG_VALID(uap->sig))
418 error = EINVAL;
419 else
420 tdksignal(ttd, uap->sig, &ksi);
421 PROC_UNLOCK(ttd->td_proc);
422 }
423 return (error);
424 }
425
426 int
427 sys_thr_kill2(struct thread *td, struct thr_kill2_args *uap)
428 /* pid_t pid, long id, int sig */
429 {
430 ksiginfo_t ksi;
431 struct thread *ttd;
432 struct proc *p;
433 int error;
434
435 AUDIT_ARG_SIGNUM(uap->sig);
436
437 ksiginfo_init(&ksi);
438 ksi.ksi_signo = uap->sig;
439 ksi.ksi_code = SI_LWP;
440 ksi.ksi_pid = td->td_proc->p_pid;
441 ksi.ksi_uid = td->td_ucred->cr_ruid;
442 if (uap->id == -1) {
443 if ((p = pfind(uap->pid)) == NULL)
444 return (ESRCH);
445 AUDIT_ARG_PROCESS(p);
446 error = p_cansignal(td, p, uap->sig);
447 if (error) {
448 PROC_UNLOCK(p);
449 return (error);
450 }
451 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) {
452 error = EINVAL;
453 } else {
454 error = ESRCH;
455 FOREACH_THREAD_IN_PROC(p, ttd) {
456 if (ttd != td) {
457 error = 0;
458 if (uap->sig == 0)
459 break;
460 tdksignal(ttd, uap->sig, &ksi);
461 }
462 }
463 }
464 PROC_UNLOCK(p);
465 } else {
466 ttd = tdfind((lwpid_t)uap->id, uap->pid);
467 if (ttd == NULL)
468 return (ESRCH);
469 p = ttd->td_proc;
470 AUDIT_ARG_PROCESS(p);
471 error = p_cansignal(td, p, uap->sig);
472 if (uap->sig == 0)
473 ;
474 else if (!_SIG_VALID(uap->sig))
475 error = EINVAL;
476 else
477 tdksignal(ttd, uap->sig, &ksi);
478 PROC_UNLOCK(p);
479 }
480 return (error);
481 }
482
483 int
484 sys_thr_suspend(struct thread *td, struct thr_suspend_args *uap)
485 /* const struct timespec *timeout */
486 {
487 struct timespec ts, *tsp;
488 int error;
489
490 tsp = NULL;
491 if (uap->timeout != NULL) {
492 error = umtx_copyin_timeout(uap->timeout, &ts);
493 if (error != 0)
494 return (error);
495 tsp = &ts;
496 }
497
498 return (kern_thr_suspend(td, tsp));
499 }
500
501 int
502 kern_thr_suspend(struct thread *td, struct timespec *tsp)
503 {
504 struct proc *p = td->td_proc;
505 struct timeval tv;
506 int error = 0;
507 int timo = 0;
508
509 if (td->td_pflags & TDP_WAKEUP) {
510 td->td_pflags &= ~TDP_WAKEUP;
511 return (0);
512 }
513
514 if (tsp != NULL) {
515 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0)
516 error = EWOULDBLOCK;
517 else {
518 TIMESPEC_TO_TIMEVAL(&tv, tsp);
519 timo = tvtohz(&tv);
520 }
521 }
522
523 PROC_LOCK(p);
524 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0)
525 error = msleep((void *)td, &p->p_mtx,
526 PCATCH, "lthr", timo);
527
528 if (td->td_flags & TDF_THRWAKEUP) {
529 thread_lock(td);
530 td->td_flags &= ~TDF_THRWAKEUP;
531 thread_unlock(td);
532 PROC_UNLOCK(p);
533 return (0);
534 }
535 PROC_UNLOCK(p);
536 if (error == EWOULDBLOCK)
537 error = ETIMEDOUT;
538 else if (error == ERESTART) {
539 if (timo != 0)
540 error = EINTR;
541 }
542 return (error);
543 }
544
545 int
546 sys_thr_wake(struct thread *td, struct thr_wake_args *uap)
547 /* long id */
548 {
549 struct proc *p;
550 struct thread *ttd;
551
552 if (uap->id == td->td_tid) {
553 td->td_pflags |= TDP_WAKEUP;
554 return (0);
555 }
556
557 p = td->td_proc;
558 ttd = tdfind((lwpid_t)uap->id, p->p_pid);
559 if (ttd == NULL)
560 return (ESRCH);
561 thread_lock(ttd);
562 ttd->td_flags |= TDF_THRWAKEUP;
563 thread_unlock(ttd);
564 wakeup((void *)ttd);
565 PROC_UNLOCK(p);
566 return (0);
567 }
568
569 int
570 sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap)
571 {
572 struct proc *p;
573 char name[MAXCOMLEN + 1];
574 struct thread *ttd;
575 int error;
576
577 error = 0;
578 name[0] = '\0';
579 if (uap->name != NULL) {
580 error = copyinstr(uap->name, name, sizeof(name),
581 NULL);
582 if (error)
583 return (error);
584 }
585 p = td->td_proc;
586 ttd = tdfind((lwpid_t)uap->id, p->p_pid);
587 if (ttd == NULL)
588 return (ESRCH);
589 strcpy(ttd->td_name, name);
590 #ifdef KTR
591 sched_clear_tdname(ttd);
592 #endif
593 PROC_UNLOCK(p);
594 return (error);
595 }
596
597 int
598 kern_thr_alloc(struct proc *p, int pages, struct thread **ntd)
599 {
600
601 /* Have race condition but it is cheap. */
602 if (p->p_numthreads >= max_threads_per_proc) {
603 ++max_threads_hits;
604 return (EPROCLIM);
605 }
606
607 *ntd = thread_alloc(pages);
608 if (*ntd == NULL)
609 return (ENOMEM);
610
611 return (0);
612 }
Cache object: 0851026045cccdd9df191f6c238a4a4e
|