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("$FreeBSD$");
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 <machine/frame.h>
58
59#include <security/audit/audit.h>
60
61static SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0,
62    "thread allocation");
63
64static int max_threads_per_proc = 1500;
65SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_per_proc, CTLFLAG_RW,
66    &max_threads_per_proc, 0, "Limit on threads per proc");
67
68static int max_threads_hits;
69SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_hits, CTLFLAG_RD,
70    &max_threads_hits, 0, "kern.threads.max_threads_per_proc hit count");
71
72#ifdef COMPAT_FREEBSD32
73
74static inline int
75suword_lwpid(void *addr, lwpid_t lwpid)
76{
77	int error;
78
79	if (SV_CURPROC_FLAG(SV_LP64))
80		error = suword(addr, lwpid);
81	else
82		error = suword32(addr, lwpid);
83	return (error);
84}
85
86#else
87#define suword_lwpid	suword
88#endif
89
90static int create_thread(struct thread *td, mcontext_t *ctx,
91			 void (*start_func)(void *), void *arg,
92			 char *stack_base, size_t stack_size,
93			 char *tls_base,
94			 long *child_tid, long *parent_tid,
95			 int flags, struct rtprio *rtp);
96
97/*
98 * System call interface.
99 */
100int
101sys_thr_create(struct thread *td, struct thr_create_args *uap)
102    /* ucontext_t *ctx, long *id, int flags */
103{
104	ucontext_t ctx;
105	int error;
106
107	if ((error = copyin(uap->ctx, &ctx, sizeof(ctx))))
108		return (error);
109
110	error = create_thread(td, &ctx.uc_mcontext, NULL, NULL,
111		NULL, 0, NULL, uap->id, NULL, uap->flags, NULL);
112	return (error);
113}
114
115int
116sys_thr_new(struct thread *td, struct thr_new_args *uap)
117    /* struct thr_param * */
118{
119	struct thr_param param;
120	int error;
121
122	if (uap->param_size < 0 || uap->param_size > sizeof(param))
123		return (EINVAL);
124	bzero(&param, sizeof(param));
125	if ((error = copyin(uap->param, &param, uap->param_size)))
126		return (error);
127	return (kern_thr_new(td, &param));
128}
129
130int
131kern_thr_new(struct thread *td, struct thr_param *param)
132{
133	struct rtprio rtp, *rtpp;
134	int error;
135
136	rtpp = NULL;
137	if (param->rtp != 0) {
138		error = copyin(param->rtp, &rtp, sizeof(struct rtprio));
139		if (error)
140			return (error);
141		rtpp = &rtp;
142	}
143	error = create_thread(td, NULL, param->start_func, param->arg,
144		param->stack_base, param->stack_size, param->tls_base,
145		param->child_tid, param->parent_tid, param->flags,
146		rtpp);
147	return (error);
148}
149
150static int
151create_thread(struct thread *td, mcontext_t *ctx,
152	    void (*start_func)(void *), void *arg,
153	    char *stack_base, size_t stack_size,
154	    char *tls_base,
155	    long *child_tid, long *parent_tid,
156	    int flags, struct rtprio *rtp)
157{
158	stack_t stack;
159	struct thread *newtd;
160	struct proc *p;
161	int error;
162
163	p = td->td_proc;
164
165	/* Have race condition but it is cheap. */
166	if (p->p_numthreads >= max_threads_per_proc) {
167		++max_threads_hits;
168		return (EPROCLIM);
169	}
170
171	if (rtp != NULL) {
172		switch(rtp->type) {
173		case RTP_PRIO_REALTIME:
174		case RTP_PRIO_FIFO:
175			/* Only root can set scheduler policy */
176			if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0)
177				return (EPERM);
178			if (rtp->prio > RTP_PRIO_MAX)
179				return (EINVAL);
180			break;
181		case RTP_PRIO_NORMAL:
182			rtp->prio = 0;
183			break;
184		default:
185			return (EINVAL);
186		}
187	}
188
189#ifdef RACCT
190	PROC_LOCK(td->td_proc);
191	error = racct_add(p, RACCT_NTHR, 1);
192	PROC_UNLOCK(td->td_proc);
193	if (error != 0)
194		return (EPROCLIM);
195#endif
196
197	/* Initialize our td */
198	newtd = thread_alloc(0);
199	if (newtd == NULL) {
200		error = ENOMEM;
201		goto fail;
202	}
203
204	cpu_set_upcall(newtd, td);
205
206	/*
207	 * Try the copyout as soon as we allocate the td so we don't
208	 * have to tear things down in a failure case below.
209	 * Here we copy out tid to two places, one for child and one
210	 * for parent, because pthread can create a detached thread,
211	 * if parent wants to safely access child tid, it has to provide
212	 * its storage, because child thread may exit quickly and
213	 * memory is freed before parent thread can access it.
214	 */
215	if ((child_tid != NULL &&
216	    suword_lwpid(child_tid, newtd->td_tid)) ||
217	    (parent_tid != NULL &&
218	    suword_lwpid(parent_tid, newtd->td_tid))) {
219		thread_free(newtd);
220		error = EFAULT;
221		goto fail;
222	}
223
224	bzero(&newtd->td_startzero,
225	    __rangeof(struct thread, td_startzero, td_endzero));
226	bcopy(&td->td_startcopy, &newtd->td_startcopy,
227	    __rangeof(struct thread, td_startcopy, td_endcopy));
228	newtd->td_proc = td->td_proc;
229	newtd->td_ucred = crhold(td->td_ucred);
230
231	if (ctx != NULL) { /* old way to set user context */
232		error = set_mcontext(newtd, ctx);
233		if (error != 0) {
234			thread_free(newtd);
235			crfree(td->td_ucred);
236			goto fail;
237		}
238	} else {
239		/* Set up our machine context. */
240		stack.ss_sp = stack_base;
241		stack.ss_size = stack_size;
242		/* Set upcall address to user thread entry function. */
243		cpu_set_upcall_kse(newtd, start_func, arg, &stack);
244		/* Setup user TLS address and TLS pointer register. */
245		error = cpu_set_user_tls(newtd, tls_base);
246		if (error != 0) {
247			thread_free(newtd);
248			crfree(td->td_ucred);
249			goto fail;
250		}
251	}
252
253	PROC_LOCK(td->td_proc);
254	td->td_proc->p_flag |= P_HADTHREADS;
255	thread_link(newtd, p);
256	bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
257	thread_lock(td);
258	/* let the scheduler know about these things. */
259	sched_fork_thread(td, newtd);
260	thread_unlock(td);
261	if (P_SHOULDSTOP(p))
262		newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
263	PROC_UNLOCK(p);
264
265	tidhash_add(newtd);
266
267	thread_lock(newtd);
268	if (rtp != NULL) {
269		if (!(td->td_pri_class == PRI_TIMESHARE &&
270		      rtp->type == RTP_PRIO_NORMAL)) {
271			rtp_to_pri(rtp, newtd);
272			sched_prio(newtd, newtd->td_user_pri);
273		} /* ignore timesharing class */
274	}
275	TD_SET_CAN_RUN(newtd);
276	sched_add(newtd, SRQ_BORING);
277	thread_unlock(newtd);
278
279	return (0);
280
281fail:
282#ifdef RACCT
283	PROC_LOCK(p);
284	racct_sub(p, RACCT_NTHR, 1);
285	PROC_UNLOCK(p);
286#endif
287	return (error);
288}
289
290int
291sys_thr_self(struct thread *td, struct thr_self_args *uap)
292    /* long *id */
293{
294	int error;
295
296	error = suword_lwpid(uap->id, (unsigned)td->td_tid);
297	if (error == -1)
298		return (EFAULT);
299	return (0);
300}
301
302int
303sys_thr_exit(struct thread *td, struct thr_exit_args *uap)
304    /* long *state */
305{
306	struct proc *p;
307
308	p = td->td_proc;
309
310	/* Signal userland that it can free the stack. */
311	if ((void *)uap->state != NULL) {
312		suword_lwpid(uap->state, 1);
313		kern_umtx_wake(td, uap->state, INT_MAX, 0);
314	}
315
316	rw_wlock(&tidhash_lock);
317
318	PROC_LOCK(p);
319
320	/*
321	 * Shutting down last thread in the proc.  This will actually
322	 * call exit() in the trampoline when it returns.
323	 */
324	if (p->p_numthreads != 1) {
325		racct_sub(p, RACCT_NTHR, 1);
326		LIST_REMOVE(td, td_hash);
327		rw_wunlock(&tidhash_lock);
328		tdsigcleanup(td);
329		PROC_SLOCK(p);
330		thread_stopped(p);
331		thread_exit();
332		/* NOTREACHED */
333	}
334	PROC_UNLOCK(p);
335	rw_wunlock(&tidhash_lock);
336	return (0);
337}
338
339int
340sys_thr_kill(struct thread *td, struct thr_kill_args *uap)
341    /* long id, int sig */
342{
343	ksiginfo_t ksi;
344	struct thread *ttd;
345	struct proc *p;
346	int error;
347
348	p = td->td_proc;
349	ksiginfo_init(&ksi);
350	ksi.ksi_signo = uap->sig;
351	ksi.ksi_code = SI_LWP;
352	ksi.ksi_pid = p->p_pid;
353	ksi.ksi_uid = td->td_ucred->cr_ruid;
354	if (uap->id == -1) {
355		if (uap->sig != 0 && !_SIG_VALID(uap->sig)) {
356			error = EINVAL;
357		} else {
358			error = ESRCH;
359			PROC_LOCK(p);
360			FOREACH_THREAD_IN_PROC(p, ttd) {
361				if (ttd != td) {
362					error = 0;
363					if (uap->sig == 0)
364						break;
365					tdksignal(ttd, uap->sig, &ksi);
366				}
367			}
368			PROC_UNLOCK(p);
369		}
370	} else {
371		error = 0;
372		ttd = tdfind((lwpid_t)uap->id, p->p_pid);
373		if (ttd == NULL)
374			return (ESRCH);
375		if (uap->sig == 0)
376			;
377		else if (!_SIG_VALID(uap->sig))
378			error = EINVAL;
379		else
380			tdksignal(ttd, uap->sig, &ksi);
381		PROC_UNLOCK(ttd->td_proc);
382	}
383	return (error);
384}
385
386int
387sys_thr_kill2(struct thread *td, struct thr_kill2_args *uap)
388    /* pid_t pid, long id, int sig */
389{
390	ksiginfo_t ksi;
391	struct thread *ttd;
392	struct proc *p;
393	int error;
394
395	AUDIT_ARG_SIGNUM(uap->sig);
396
397	ksiginfo_init(&ksi);
398	ksi.ksi_signo = uap->sig;
399	ksi.ksi_code = SI_LWP;
400	ksi.ksi_pid = td->td_proc->p_pid;
401	ksi.ksi_uid = td->td_ucred->cr_ruid;
402	if (uap->id == -1) {
403		if ((p = pfind(uap->pid)) == NULL)
404			return (ESRCH);
405		AUDIT_ARG_PROCESS(p);
406		error = p_cansignal(td, p, uap->sig);
407		if (error) {
408			PROC_UNLOCK(p);
409			return (error);
410		}
411		if (uap->sig != 0 && !_SIG_VALID(uap->sig)) {
412			error = EINVAL;
413		} else {
414			error = ESRCH;
415			FOREACH_THREAD_IN_PROC(p, ttd) {
416				if (ttd != td) {
417					error = 0;
418					if (uap->sig == 0)
419						break;
420					tdksignal(ttd, uap->sig, &ksi);
421				}
422			}
423		}
424		PROC_UNLOCK(p);
425	} else {
426		ttd = tdfind((lwpid_t)uap->id, uap->pid);
427		if (ttd == NULL)
428			return (ESRCH);
429		p = ttd->td_proc;
430		AUDIT_ARG_PROCESS(p);
431		error = p_cansignal(td, p, uap->sig);
432		if (uap->sig == 0)
433			;
434		else if (!_SIG_VALID(uap->sig))
435			error = EINVAL;
436		else
437			tdksignal(ttd, uap->sig, &ksi);
438		PROC_UNLOCK(p);
439	}
440	return (error);
441}
442
443int
444sys_thr_suspend(struct thread *td, struct thr_suspend_args *uap)
445	/* const struct timespec *timeout */
446{
447	struct timespec ts, *tsp;
448	int error;
449
450	tsp = NULL;
451	if (uap->timeout != NULL) {
452		error = umtx_copyin_timeout(uap->timeout, &ts);
453		if (error != 0)
454			return (error);
455		tsp = &ts;
456	}
457
458	return (kern_thr_suspend(td, tsp));
459}
460
461int
462kern_thr_suspend(struct thread *td, struct timespec *tsp)
463{
464	struct proc *p = td->td_proc;
465	struct timeval tv;
466	int error = 0;
467	int timo = 0;
468
469	if (td->td_pflags & TDP_WAKEUP) {
470		td->td_pflags &= ~TDP_WAKEUP;
471		return (0);
472	}
473
474	if (tsp != NULL) {
475		if (tsp->tv_sec == 0 && tsp->tv_nsec == 0)
476			error = EWOULDBLOCK;
477		else {
478			TIMESPEC_TO_TIMEVAL(&tv, tsp);
479			timo = tvtohz(&tv);
480		}
481	}
482
483	PROC_LOCK(p);
484	if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0)
485		error = msleep((void *)td, &p->p_mtx,
486			 PCATCH, "lthr", timo);
487
488	if (td->td_flags & TDF_THRWAKEUP) {
489		thread_lock(td);
490		td->td_flags &= ~TDF_THRWAKEUP;
491		thread_unlock(td);
492		PROC_UNLOCK(p);
493		return (0);
494	}
495	PROC_UNLOCK(p);
496	if (error == EWOULDBLOCK)
497		error = ETIMEDOUT;
498	else if (error == ERESTART) {
499		if (timo != 0)
500			error = EINTR;
501	}
502	return (error);
503}
504
505int
506sys_thr_wake(struct thread *td, struct thr_wake_args *uap)
507	/* long id */
508{
509	struct proc *p;
510	struct thread *ttd;
511
512	if (uap->id == td->td_tid) {
513		td->td_pflags |= TDP_WAKEUP;
514		return (0);
515	}
516
517	p = td->td_proc;
518	ttd = tdfind((lwpid_t)uap->id, p->p_pid);
519	if (ttd == NULL)
520		return (ESRCH);
521	thread_lock(ttd);
522	ttd->td_flags |= TDF_THRWAKEUP;
523	thread_unlock(ttd);
524	wakeup((void *)ttd);
525	PROC_UNLOCK(p);
526	return (0);
527}
528
529int
530sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap)
531{
532	struct proc *p;
533	char name[MAXCOMLEN + 1];
534	struct thread *ttd;
535	int error;
536
537	error = 0;
538	name[0] = '\0';
539	if (uap->name != NULL) {
540		error = copyinstr(uap->name, name, sizeof(name),
541			NULL);
542		if (error)
543			return (error);
544	}
545	p = td->td_proc;
546	ttd = tdfind((lwpid_t)uap->id, p->p_pid);
547	if (ttd == NULL)
548		return (ESRCH);
549	strcpy(ttd->td_name, name);
550#ifdef KTR
551	sched_clear_tdname(ttd);
552#endif
553	PROC_UNLOCK(p);
554	return (error);
555}
556