2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2003, Jeffrey Roberson <jeff@freebsd.org>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice unmodified, this list of conditions, and the following
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include "opt_posix.h"
33 #include "opt_hwpmc_hooks.h"
34 #include <sys/param.h>
35 #include <sys/kernel.h>
37 #include <sys/mutex.h>
40 #include <sys/posix4.h>
41 #include <sys/ptrace.h>
42 #include <sys/racct.h>
43 #include <sys/resourcevar.h>
44 #include <sys/rwlock.h>
45 #include <sys/sched.h>
46 #include <sys/sysctl.h>
48 #include <sys/syscallsubr.h>
49 #include <sys/sysent.h>
50 #include <sys/systm.h>
51 #include <sys/sysproto.h>
52 #include <sys/signalvar.h>
53 #include <sys/sysctl.h>
54 #include <sys/ucontext.h>
56 #include <sys/rtprio.h>
58 #include <sys/limits.h>
60 #include <sys/pmckern.h>
63 #include <machine/frame.h>
65 #include <security/audit/audit.h>
67 static SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0,
70 static int max_threads_per_proc = 1500;
71 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_per_proc, CTLFLAG_RW,
72 &max_threads_per_proc, 0, "Limit on threads per proc");
74 static int max_threads_hits;
75 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_hits, CTLFLAG_RD,
76 &max_threads_hits, 0, "kern.threads.max_threads_per_proc hit count");
78 #ifdef COMPAT_FREEBSD32
81 suword_lwpid(void *addr, lwpid_t lwpid)
85 if (SV_CURPROC_FLAG(SV_LP64))
86 error = suword(addr, lwpid);
88 error = suword32(addr, lwpid);
93 #define suword_lwpid suword
97 * System call interface.
100 struct thr_create_initthr_args {
106 thr_create_initthr(struct thread *td, void *thunk)
108 struct thr_create_initthr_args *args;
110 /* Copy out the child tid. */
112 if (args->tid != NULL && suword_lwpid(args->tid, td->td_tid))
115 return (set_mcontext(td, &args->ctx.uc_mcontext));
119 sys_thr_create(struct thread *td, struct thr_create_args *uap)
120 /* ucontext_t *ctx, long *id, int flags */
122 struct thr_create_initthr_args args;
125 if ((error = copyin(uap->ctx, &args.ctx, sizeof(args.ctx))))
128 return (thread_create(td, NULL, thr_create_initthr, &args));
132 sys_thr_new(struct thread *td, struct thr_new_args *uap)
133 /* struct thr_param * */
135 struct thr_param param;
138 if (uap->param_size < 0 || uap->param_size > sizeof(param))
140 bzero(¶m, sizeof(param));
141 if ((error = copyin(uap->param, ¶m, uap->param_size)))
143 return (kern_thr_new(td, ¶m));
147 thr_new_initthr(struct thread *td, void *thunk)
150 struct thr_param *param;
153 * Here we copy out tid to two places, one for child and one
154 * for parent, because pthread can create a detached thread,
155 * if parent wants to safely access child tid, it has to provide
156 * its storage, because child thread may exit quickly and
157 * memory is freed before parent thread can access it.
160 if ((param->child_tid != NULL &&
161 suword_lwpid(param->child_tid, td->td_tid)) ||
162 (param->parent_tid != NULL &&
163 suword_lwpid(param->parent_tid, td->td_tid)))
166 /* Set up our machine context. */
167 stack.ss_sp = param->stack_base;
168 stack.ss_size = param->stack_size;
169 /* Set upcall address to user thread entry function. */
170 cpu_set_upcall(td, param->start_func, param->arg, &stack);
171 /* Setup user TLS address and TLS pointer register. */
172 return (cpu_set_user_tls(td, param->tls_base));
176 kern_thr_new(struct thread *td, struct thr_param *param)
178 struct rtprio rtp, *rtpp;
182 if (param->rtp != 0) {
183 error = copyin(param->rtp, &rtp, sizeof(struct rtprio));
188 return (thread_create(td, rtpp, thr_new_initthr, param));
192 thread_create(struct thread *td, struct rtprio *rtp,
193 int (*initialize_thread)(struct thread *, void *), void *thunk)
195 struct thread *newtd;
203 case RTP_PRIO_REALTIME:
205 /* Only root can set scheduler policy */
206 if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0)
208 if (rtp->prio > RTP_PRIO_MAX)
211 case RTP_PRIO_NORMAL:
222 error = racct_add(p, RACCT_NTHR, 1);
229 /* Initialize our td */
230 error = kern_thr_alloc(p, 0, &newtd);
234 cpu_copy_thread(newtd, td);
236 bzero(&newtd->td_startzero,
237 __rangeof(struct thread, td_startzero, td_endzero));
238 bcopy(&td->td_startcopy, &newtd->td_startcopy,
239 __rangeof(struct thread, td_startcopy, td_endcopy));
240 newtd->td_proc = td->td_proc;
241 newtd->td_rb_list = newtd->td_rbp_list = newtd->td_rb_inact = 0;
242 thread_cow_get(newtd, td);
244 error = initialize_thread(newtd, thunk);
246 thread_cow_free(newtd);
252 p->p_flag |= P_HADTHREADS;
253 thread_link(newtd, p);
254 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
256 /* let the scheduler know about these things. */
257 sched_fork_thread(td, newtd);
260 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
261 if (p->p_ptevents & PTRACE_LWP)
262 newtd->td_dbgflags |= TDB_BORN;
266 if (PMC_PROC_IS_USING_PMCS(p))
267 PMC_CALL_HOOK(newtd, PMC_FN_THR_CREATE, NULL);
268 else if (PMC_SYSTEM_SAMPLING_ACTIVE())
269 PMC_CALL_HOOK_UNLOCKED(newtd, PMC_FN_THR_CREATE_LOG, NULL);
276 if (!(td->td_pri_class == PRI_TIMESHARE &&
277 rtp->type == RTP_PRIO_NORMAL)) {
278 rtp_to_pri(rtp, newtd);
279 sched_prio(newtd, newtd->td_user_pri);
280 } /* ignore timesharing class */
282 TD_SET_CAN_RUN(newtd);
283 sched_add(newtd, SRQ_BORING);
284 thread_unlock(newtd);
292 racct_sub(p, RACCT_NTHR, 1);
300 sys_thr_self(struct thread *td, struct thr_self_args *uap)
305 error = suword_lwpid(uap->id, (unsigned)td->td_tid);
312 sys_thr_exit(struct thread *td, struct thr_exit_args *uap)
316 umtx_thread_exit(td);
318 /* Signal userland that it can free the stack. */
319 if ((void *)uap->state != NULL) {
320 suword_lwpid(uap->state, 1);
321 kern_umtx_wake(td, uap->state, INT_MAX, 0);
324 return (kern_thr_exit(td));
328 kern_thr_exit(struct thread *td)
335 * If all of the threads in a process call this routine to
336 * exit (e.g. all threads call pthread_exit()), exactly one
337 * thread should return to the caller to terminate the process
338 * instead of the thread.
340 * Checking p_numthreads alone is not sufficient since threads
341 * might be committed to terminating while the PROC_LOCK is
342 * dropped in either ptracestop() or while removing this thread
343 * from the tidhash. Instead, the p_pendingexits field holds
344 * the count of threads in either of those states and a thread
345 * is considered the "last" thread if all of the other threads
346 * in a process are already terminating.
349 if (p->p_numthreads == p->p_pendingexits + 1) {
351 * Ignore attempts to shut down last thread in the
352 * proc. This will actually call _exit(2) in the
353 * usermode trampoline when it returns.
360 td->td_dbgflags |= TDB_EXIT;
361 if (p->p_ptevents & PTRACE_LWP)
362 ptracestop(td, SIGTRAP, NULL);
369 * The check above should prevent all other threads from this
370 * process from exiting while the PROC_LOCK is dropped, so
371 * there must be at least one other thread other than the
374 KASSERT(p->p_numthreads > 1, ("too few threads"));
375 racct_sub(p, RACCT_NTHR, 1);
379 AUDIT_SYSCALL_EXIT(0, td);
389 sys_thr_kill(struct thread *td, struct thr_kill_args *uap)
390 /* long id, int sig */
399 ksi.ksi_signo = uap->sig;
400 ksi.ksi_code = SI_LWP;
401 ksi.ksi_pid = p->p_pid;
402 ksi.ksi_uid = td->td_ucred->cr_ruid;
404 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) {
409 FOREACH_THREAD_IN_PROC(p, ttd) {
414 tdksignal(ttd, uap->sig, &ksi);
421 ttd = tdfind((lwpid_t)uap->id, p->p_pid);
426 else if (!_SIG_VALID(uap->sig))
429 tdksignal(ttd, uap->sig, &ksi);
430 PROC_UNLOCK(ttd->td_proc);
436 sys_thr_kill2(struct thread *td, struct thr_kill2_args *uap)
437 /* pid_t pid, long id, int sig */
444 AUDIT_ARG_SIGNUM(uap->sig);
447 ksi.ksi_signo = uap->sig;
448 ksi.ksi_code = SI_LWP;
449 ksi.ksi_pid = td->td_proc->p_pid;
450 ksi.ksi_uid = td->td_ucred->cr_ruid;
452 if ((p = pfind(uap->pid)) == NULL)
454 AUDIT_ARG_PROCESS(p);
455 error = p_cansignal(td, p, uap->sig);
460 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) {
464 FOREACH_THREAD_IN_PROC(p, ttd) {
469 tdksignal(ttd, uap->sig, &ksi);
475 ttd = tdfind((lwpid_t)uap->id, uap->pid);
479 AUDIT_ARG_PROCESS(p);
480 error = p_cansignal(td, p, uap->sig);
483 else if (!_SIG_VALID(uap->sig))
486 tdksignal(ttd, uap->sig, &ksi);
493 sys_thr_suspend(struct thread *td, struct thr_suspend_args *uap)
494 /* const struct timespec *timeout */
496 struct timespec ts, *tsp;
500 if (uap->timeout != NULL) {
501 error = umtx_copyin_timeout(uap->timeout, &ts);
507 return (kern_thr_suspend(td, tsp));
511 kern_thr_suspend(struct thread *td, struct timespec *tsp)
513 struct proc *p = td->td_proc;
518 if (td->td_pflags & TDP_WAKEUP) {
519 td->td_pflags &= ~TDP_WAKEUP;
524 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0)
527 TIMESPEC_TO_TIMEVAL(&tv, tsp);
533 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0)
534 error = msleep((void *)td, &p->p_mtx,
535 PCATCH, "lthr", timo);
537 if (td->td_flags & TDF_THRWAKEUP) {
539 td->td_flags &= ~TDF_THRWAKEUP;
545 if (error == EWOULDBLOCK)
547 else if (error == ERESTART) {
555 sys_thr_wake(struct thread *td, struct thr_wake_args *uap)
561 if (uap->id == td->td_tid) {
562 td->td_pflags |= TDP_WAKEUP;
567 ttd = tdfind((lwpid_t)uap->id, p->p_pid);
571 ttd->td_flags |= TDF_THRWAKEUP;
579 sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap)
582 char name[MAXCOMLEN + 1];
588 if (uap->name != NULL) {
589 error = copyinstr(uap->name, name, sizeof(name), NULL);
590 if (error == ENAMETOOLONG) {
591 error = copyin(uap->name, name, sizeof(name) - 1);
592 name[sizeof(name) - 1] = '\0';
598 ttd = tdfind((lwpid_t)uap->id, p->p_pid);
601 strcpy(ttd->td_name, name);
603 if (PMC_PROC_IS_USING_PMCS(p) || PMC_SYSTEM_SAMPLING_ACTIVE())
604 PMC_CALL_HOOK_UNLOCKED(ttd, PMC_FN_THR_CREATE_LOG, NULL);
607 sched_clear_tdname(ttd);
614 kern_thr_alloc(struct proc *p, int pages, struct thread **ntd)
617 /* Have race condition but it is cheap. */
618 if (p->p_numthreads >= max_threads_per_proc) {
623 *ntd = thread_alloc(pages);