2 * Copyright (c) 2003, Jeffrey Roberson <jeff@freebsd.org>
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice unmodified, this list of conditions, and the following
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.
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.
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
30 #include "opt_compat.h"
31 #include "opt_posix.h"
32 #include <sys/param.h>
33 #include <sys/kernel.h>
35 #include <sys/mutex.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>
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>
53 #include <sys/rtprio.h>
55 #include <sys/limits.h>
57 #include <machine/frame.h>
59 #include <security/audit/audit.h>
61 static SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0,
64 static int max_threads_per_proc = 1500;
65 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_per_proc, CTLFLAG_RW,
66 &max_threads_per_proc, 0, "Limit on threads per proc");
68 static int max_threads_hits;
69 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_hits, CTLFLAG_RD,
70 &max_threads_hits, 0, "kern.threads.max_threads_per_proc hit count");
72 #ifdef COMPAT_FREEBSD32
75 suword_lwpid(void *addr, lwpid_t lwpid)
79 if (SV_CURPROC_FLAG(SV_LP64))
80 error = suword(addr, lwpid);
82 error = suword32(addr, lwpid);
87 #define suword_lwpid suword
91 * System call interface.
94 struct thr_create_initthr_args {
100 thr_create_initthr(struct thread *td, void *thunk)
102 struct thr_create_initthr_args *args;
104 /* Copy out the child tid. */
106 if (args->tid != NULL && suword_lwpid(args->tid, td->td_tid))
109 return (set_mcontext(td, &args->ctx.uc_mcontext));
113 sys_thr_create(struct thread *td, struct thr_create_args *uap)
114 /* ucontext_t *ctx, long *id, int flags */
116 struct thr_create_initthr_args args;
119 if ((error = copyin(uap->ctx, &args.ctx, sizeof(args.ctx))))
122 return (thread_create(td, NULL, thr_create_initthr, &args));
126 sys_thr_new(struct thread *td, struct thr_new_args *uap)
127 /* struct thr_param * */
129 struct thr_param param;
132 if (uap->param_size < 0 || uap->param_size > sizeof(param))
134 bzero(¶m, sizeof(param));
135 if ((error = copyin(uap->param, ¶m, uap->param_size)))
137 return (kern_thr_new(td, ¶m));
141 thr_new_initthr(struct thread *td, void *thunk)
144 struct thr_param *param;
147 * Here we copy out tid to two places, one for child and one
148 * for parent, because pthread can create a detached thread,
149 * if parent wants to safely access child tid, it has to provide
150 * its storage, because child thread may exit quickly and
151 * memory is freed before parent thread can access it.
154 if ((param->child_tid != NULL &&
155 suword_lwpid(param->child_tid, td->td_tid)) ||
156 (param->parent_tid != NULL &&
157 suword_lwpid(param->parent_tid, td->td_tid)))
160 /* Set up our machine context. */
161 stack.ss_sp = param->stack_base;
162 stack.ss_size = param->stack_size;
163 /* Set upcall address to user thread entry function. */
164 cpu_set_upcall_kse(td, param->start_func, param->arg, &stack);
165 /* Setup user TLS address and TLS pointer register. */
166 return (cpu_set_user_tls(td, param->tls_base));
170 kern_thr_new(struct thread *td, struct thr_param *param)
172 struct rtprio rtp, *rtpp;
176 if (param->rtp != 0) {
177 error = copyin(param->rtp, &rtp, sizeof(struct rtprio));
182 return (thread_create(td, rtpp, thr_new_initthr, param));
186 thread_create(struct thread *td, struct rtprio *rtp,
187 int (*initialize_thread)(struct thread *, void *), void *thunk)
189 struct thread *newtd;
197 case RTP_PRIO_REALTIME:
199 /* Only root can set scheduler policy */
200 if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0)
202 if (rtp->prio > RTP_PRIO_MAX)
205 case RTP_PRIO_NORMAL:
214 PROC_LOCK(td->td_proc);
215 error = racct_add(p, RACCT_NTHR, 1);
216 PROC_UNLOCK(td->td_proc);
221 /* Initialize our td */
222 error = kern_thr_alloc(p, 0, &newtd);
226 cpu_set_upcall(newtd, td);
228 bzero(&newtd->td_startzero,
229 __rangeof(struct thread, td_startzero, td_endzero));
231 bcopy(&td->td_startcopy, &newtd->td_startcopy,
232 __rangeof(struct thread, td_startcopy, td_endcopy));
233 newtd->td_proc = td->td_proc;
234 newtd->td_ucred = crhold(td->td_ucred);
235 newtd->td_dbg_sc_code = td->td_dbg_sc_code;
236 newtd->td_dbg_sc_narg = td->td_dbg_sc_narg;
238 error = initialize_thread(newtd, thunk);
241 crfree(td->td_ucred);
245 PROC_LOCK(td->td_proc);
246 td->td_proc->p_flag |= P_HADTHREADS;
247 thread_link(newtd, p);
248 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
250 /* let the scheduler know about these things. */
251 sched_fork_thread(td, newtd);
254 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
255 if (p->p_flag2 & P2_LWP_EVENTS)
256 newtd->td_dbgflags |= TDB_BORN;
263 if (!(td->td_pri_class == PRI_TIMESHARE &&
264 rtp->type == RTP_PRIO_NORMAL)) {
265 rtp_to_pri(rtp, newtd);
266 sched_prio(newtd, newtd->td_user_pri);
267 } /* ignore timesharing class */
269 TD_SET_CAN_RUN(newtd);
270 sched_add(newtd, SRQ_BORING);
271 thread_unlock(newtd);
279 racct_sub(p, RACCT_NTHR, 1);
287 sys_thr_self(struct thread *td, struct thr_self_args *uap)
292 error = suword_lwpid(uap->id, (unsigned)td->td_tid);
299 sys_thr_exit(struct thread *td, struct thr_exit_args *uap)
303 /* Signal userland that it can free the stack. */
304 if ((void *)uap->state != NULL) {
305 suword_lwpid(uap->state, 1);
306 kern_umtx_wake(td, uap->state, INT_MAX, 0);
309 return (kern_thr_exit(td));
313 kern_thr_exit(struct thread *td)
320 * If all of the threads in a process call this routine to
321 * exit (e.g. all threads call pthread_exit()), exactly one
322 * thread should return to the caller to terminate the process
323 * instead of the thread.
325 * Checking p_numthreads alone is not sufficient since threads
326 * might be committed to terminating while the PROC_LOCK is
327 * dropped in either ptracestop() or while removing this thread
328 * from the tidhash. Instead, the p_pendingexits field holds
329 * the count of threads in either of those states and a thread
330 * is considered the "last" thread if all of the other threads
331 * in a process are already terminating.
334 if (p->p_numthreads == p->p_pendingexits + 1) {
336 * Ignore attempts to shut down last thread in the
337 * proc. This will actually call _exit(2) in the
338 * usermode trampoline when it returns.
345 td->td_dbgflags |= TDB_EXIT;
346 if (p->p_flag & P_TRACED && p->p_flag2 & P2_LWP_EVENTS)
347 ptracestop(td, SIGTRAP);
354 * The check above should prevent all other threads from this
355 * process from exiting while the PROC_LOCK is dropped, so
356 * there must be at least one other thread other than the
359 KASSERT(p->p_numthreads > 1, ("too few threads"));
360 racct_sub(p, RACCT_NTHR, 1);
362 umtx_thread_exit(td);
370 sys_thr_kill(struct thread *td, struct thr_kill_args *uap)
371 /* long id, int sig */
380 ksi.ksi_signo = uap->sig;
381 ksi.ksi_code = SI_LWP;
382 ksi.ksi_pid = p->p_pid;
383 ksi.ksi_uid = td->td_ucred->cr_ruid;
385 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) {
390 FOREACH_THREAD_IN_PROC(p, ttd) {
395 tdksignal(ttd, uap->sig, &ksi);
402 ttd = tdfind((lwpid_t)uap->id, p->p_pid);
407 else if (!_SIG_VALID(uap->sig))
410 tdksignal(ttd, uap->sig, &ksi);
411 PROC_UNLOCK(ttd->td_proc);
417 sys_thr_kill2(struct thread *td, struct thr_kill2_args *uap)
418 /* pid_t pid, long id, int sig */
425 AUDIT_ARG_SIGNUM(uap->sig);
428 ksi.ksi_signo = uap->sig;
429 ksi.ksi_code = SI_LWP;
430 ksi.ksi_pid = td->td_proc->p_pid;
431 ksi.ksi_uid = td->td_ucred->cr_ruid;
433 if ((p = pfind(uap->pid)) == NULL)
435 AUDIT_ARG_PROCESS(p);
436 error = p_cansignal(td, p, uap->sig);
441 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) {
445 FOREACH_THREAD_IN_PROC(p, ttd) {
450 tdksignal(ttd, uap->sig, &ksi);
456 ttd = tdfind((lwpid_t)uap->id, uap->pid);
460 AUDIT_ARG_PROCESS(p);
461 error = p_cansignal(td, p, uap->sig);
464 else if (!_SIG_VALID(uap->sig))
467 tdksignal(ttd, uap->sig, &ksi);
474 sys_thr_suspend(struct thread *td, struct thr_suspend_args *uap)
475 /* const struct timespec *timeout */
477 struct timespec ts, *tsp;
481 if (uap->timeout != NULL) {
482 error = umtx_copyin_timeout(uap->timeout, &ts);
488 return (kern_thr_suspend(td, tsp));
492 kern_thr_suspend(struct thread *td, struct timespec *tsp)
494 struct proc *p = td->td_proc;
499 if (td->td_pflags & TDP_WAKEUP) {
500 td->td_pflags &= ~TDP_WAKEUP;
505 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0)
508 TIMESPEC_TO_TIMEVAL(&tv, tsp);
514 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0)
515 error = msleep((void *)td, &p->p_mtx,
516 PCATCH, "lthr", timo);
518 if (td->td_flags & TDF_THRWAKEUP) {
520 td->td_flags &= ~TDF_THRWAKEUP;
526 if (error == EWOULDBLOCK)
528 else if (error == ERESTART) {
536 sys_thr_wake(struct thread *td, struct thr_wake_args *uap)
542 if (uap->id == td->td_tid) {
543 td->td_pflags |= TDP_WAKEUP;
548 ttd = tdfind((lwpid_t)uap->id, p->p_pid);
552 ttd->td_flags |= TDF_THRWAKEUP;
560 sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap)
563 char name[MAXCOMLEN + 1];
569 if (uap->name != NULL) {
570 error = copyinstr(uap->name, name, sizeof(name),
576 ttd = tdfind((lwpid_t)uap->id, p->p_pid);
579 strcpy(ttd->td_name, name);
581 sched_clear_tdname(ttd);
588 kern_thr_alloc(struct proc *p, int pages, struct thread **ntd)
591 /* Have race condition but it is cheap. */
592 if (p->p_numthreads >= max_threads_per_proc) {
597 *ntd = thread_alloc(pages);