1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2003, Jeffrey Roberson <jeff@freebsd.org> 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice unmodified, this list of conditions, and the following 12 * disclaimer. 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. 16 * 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. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include "opt_posix.h" 33 #include "opt_hwpmc_hooks.h" 34 #include <sys/param.h> 35 #include <sys/kernel.h> 36 #include <sys/lock.h> 37 #include <sys/mutex.h> 38 #include <sys/priv.h> 39 #include <sys/proc.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> 47 #include <sys/smp.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> 55 #include <sys/thr.h> 56 #include <sys/rtprio.h> 57 #include <sys/umtx.h> 58 #include <sys/limits.h> 59 #ifdef HWPMC_HOOKS 60 #include <sys/pmckern.h> 61 #endif 62 63 #include <machine/frame.h> 64 65 #include <security/audit/audit.h> 66 67 static SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0, 68 "thread allocation"); 69 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"); 73 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"); 77 78 #ifdef COMPAT_FREEBSD32 79 80 static inline int 81 suword_lwpid(void *addr, lwpid_t lwpid) 82 { 83 int error; 84 85 if (SV_CURPROC_FLAG(SV_LP64)) 86 error = suword(addr, lwpid); 87 else 88 error = suword32(addr, lwpid); 89 return (error); 90 } 91 92 #else 93 #define suword_lwpid suword 94 #endif 95 96 /* 97 * System call interface. 98 */ 99 100 struct thr_create_initthr_args { 101 ucontext_t ctx; 102 long *tid; 103 }; 104 105 static int 106 thr_create_initthr(struct thread *td, void *thunk) 107 { 108 struct thr_create_initthr_args *args; 109 110 /* Copy out the child tid. */ 111 args = thunk; 112 if (args->tid != NULL && suword_lwpid(args->tid, td->td_tid)) 113 return (EFAULT); 114 115 return (set_mcontext(td, &args->ctx.uc_mcontext)); 116 } 117 118 int 119 sys_thr_create(struct thread *td, struct thr_create_args *uap) 120 /* ucontext_t *ctx, long *id, int flags */ 121 { 122 struct thr_create_initthr_args args; 123 int error; 124 125 if ((error = copyin(uap->ctx, &args.ctx, sizeof(args.ctx)))) 126 return (error); 127 args.tid = uap->id; 128 return (thread_create(td, NULL, thr_create_initthr, &args)); 129 } 130 131 int 132 sys_thr_new(struct thread *td, struct thr_new_args *uap) 133 /* struct thr_param * */ 134 { 135 struct thr_param param; 136 int error; 137 138 if (uap->param_size < 0 || uap->param_size > sizeof(param)) 139 return (EINVAL); 140 bzero(¶m, sizeof(param)); 141 if ((error = copyin(uap->param, ¶m, uap->param_size))) 142 return (error); 143 return (kern_thr_new(td, ¶m)); 144 } 145 146 static int 147 thr_new_initthr(struct thread *td, void *thunk) 148 { 149 stack_t stack; 150 struct thr_param *param; 151 152 /* 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. 158 */ 159 param = thunk; 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))) 164 return (EFAULT); 165 166 /* 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)); 173 } 174 175 int 176 kern_thr_new(struct thread *td, struct thr_param *param) 177 { 178 struct rtprio rtp, *rtpp; 179 int error; 180 181 rtpp = NULL; 182 if (param->rtp != 0) { 183 error = copyin(param->rtp, &rtp, sizeof(struct rtprio)); 184 if (error) 185 return (error); 186 rtpp = &rtp; 187 } 188 return (thread_create(td, rtpp, thr_new_initthr, param)); 189 } 190 191 int 192 thread_create(struct thread *td, struct rtprio *rtp, 193 int (*initialize_thread)(struct thread *, void *), void *thunk) 194 { 195 struct thread *newtd; 196 struct proc *p; 197 int error; 198 199 p = td->td_proc; 200 201 if (rtp != NULL) { 202 switch(rtp->type) { 203 case RTP_PRIO_REALTIME: 204 case RTP_PRIO_FIFO: 205 /* Only root can set scheduler policy */ 206 if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0) 207 return (EPERM); 208 if (rtp->prio > RTP_PRIO_MAX) 209 return (EINVAL); 210 break; 211 case RTP_PRIO_NORMAL: 212 rtp->prio = 0; 213 break; 214 default: 215 return (EINVAL); 216 } 217 } 218 219 #ifdef RACCT 220 if (racct_enable) { 221 PROC_LOCK(p); 222 error = racct_add(p, RACCT_NTHR, 1); 223 PROC_UNLOCK(p); 224 if (error != 0) 225 return (EPROCLIM); 226 } 227 #endif 228 229 /* Initialize our td */ 230 error = kern_thr_alloc(p, 0, &newtd); 231 if (error) 232 goto fail; 233 234 cpu_copy_thread(newtd, td); 235 236 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); 243 244 error = initialize_thread(newtd, thunk); 245 if (error != 0) { 246 thread_cow_free(newtd); 247 thread_free(newtd); 248 goto fail; 249 } 250 251 PROC_LOCK(p); 252 p->p_flag |= P_HADTHREADS; 253 thread_link(newtd, p); 254 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); 255 thread_lock(td); 256 /* let the scheduler know about these things. */ 257 sched_fork_thread(td, newtd); 258 thread_unlock(td); 259 if (P_SHOULDSTOP(p)) 260 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; 261 if (p->p_ptevents & PTRACE_LWP) 262 newtd->td_dbgflags |= TDB_BORN; 263 264 PROC_UNLOCK(p); 265 #ifdef HWPMC_HOOKS 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); 270 #endif 271 272 tidhash_add(newtd); 273 274 thread_lock(newtd); 275 if (rtp != 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 */ 281 } 282 TD_SET_CAN_RUN(newtd); 283 sched_add(newtd, SRQ_BORING); 284 thread_unlock(newtd); 285 286 return (0); 287 288 fail: 289 #ifdef RACCT 290 if (racct_enable) { 291 PROC_LOCK(p); 292 racct_sub(p, RACCT_NTHR, 1); 293 PROC_UNLOCK(p); 294 } 295 #endif 296 return (error); 297 } 298 299 int 300 sys_thr_self(struct thread *td, struct thr_self_args *uap) 301 /* long *id */ 302 { 303 int error; 304 305 error = suword_lwpid(uap->id, (unsigned)td->td_tid); 306 if (error == -1) 307 return (EFAULT); 308 return (0); 309 } 310 311 int 312 sys_thr_exit(struct thread *td, struct thr_exit_args *uap) 313 /* long *state */ 314 { 315 316 umtx_thread_exit(td); 317 318 /* 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); 322 } 323 324 return (kern_thr_exit(td)); 325 } 326 327 int 328 kern_thr_exit(struct thread *td) 329 { 330 struct proc *p; 331 332 p = td->td_proc; 333 334 /* 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. 339 * 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. 347 */ 348 PROC_LOCK(p); 349 if (p->p_numthreads == p->p_pendingexits + 1) { 350 /* 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. 354 */ 355 PROC_UNLOCK(p); 356 return (0); 357 } 358 359 p->p_pendingexits++; 360 td->td_dbgflags |= TDB_EXIT; 361 if (p->p_ptevents & PTRACE_LWP) 362 ptracestop(td, SIGTRAP, NULL); 363 PROC_UNLOCK(p); 364 tidhash_remove(td); 365 PROC_LOCK(p); 366 p->p_pendingexits--; 367 368 /* 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 372 * current thread. 373 */ 374 KASSERT(p->p_numthreads > 1, ("too few threads")); 375 racct_sub(p, RACCT_NTHR, 1); 376 tdsigcleanup(td); 377 378 #ifdef AUDIT 379 AUDIT_SYSCALL_EXIT(0, td); 380 #endif 381 382 PROC_SLOCK(p); 383 thread_stopped(p); 384 thread_exit(); 385 /* NOTREACHED */ 386 } 387 388 int 389 sys_thr_kill(struct thread *td, struct thr_kill_args *uap) 390 /* long id, int sig */ 391 { 392 ksiginfo_t ksi; 393 struct thread *ttd; 394 struct proc *p; 395 int error; 396 397 p = td->td_proc; 398 ksiginfo_init(&ksi); 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; 403 if (uap->id == -1) { 404 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 405 error = EINVAL; 406 } else { 407 error = ESRCH; 408 PROC_LOCK(p); 409 FOREACH_THREAD_IN_PROC(p, ttd) { 410 if (ttd != td) { 411 error = 0; 412 if (uap->sig == 0) 413 break; 414 tdksignal(ttd, uap->sig, &ksi); 415 } 416 } 417 PROC_UNLOCK(p); 418 } 419 } else { 420 error = 0; 421 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 422 if (ttd == NULL) 423 return (ESRCH); 424 if (uap->sig == 0) 425 ; 426 else if (!_SIG_VALID(uap->sig)) 427 error = EINVAL; 428 else 429 tdksignal(ttd, uap->sig, &ksi); 430 PROC_UNLOCK(ttd->td_proc); 431 } 432 return (error); 433 } 434 435 int 436 sys_thr_kill2(struct thread *td, struct thr_kill2_args *uap) 437 /* pid_t pid, long id, int sig */ 438 { 439 ksiginfo_t ksi; 440 struct thread *ttd; 441 struct proc *p; 442 int error; 443 444 AUDIT_ARG_SIGNUM(uap->sig); 445 446 ksiginfo_init(&ksi); 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; 451 if (uap->id == -1) { 452 if ((p = pfind(uap->pid)) == NULL) 453 return (ESRCH); 454 AUDIT_ARG_PROCESS(p); 455 error = p_cansignal(td, p, uap->sig); 456 if (error) { 457 PROC_UNLOCK(p); 458 return (error); 459 } 460 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 461 error = EINVAL; 462 } else { 463 error = ESRCH; 464 FOREACH_THREAD_IN_PROC(p, ttd) { 465 if (ttd != td) { 466 error = 0; 467 if (uap->sig == 0) 468 break; 469 tdksignal(ttd, uap->sig, &ksi); 470 } 471 } 472 } 473 PROC_UNLOCK(p); 474 } else { 475 ttd = tdfind((lwpid_t)uap->id, uap->pid); 476 if (ttd == NULL) 477 return (ESRCH); 478 p = ttd->td_proc; 479 AUDIT_ARG_PROCESS(p); 480 error = p_cansignal(td, p, uap->sig); 481 if (uap->sig == 0) 482 ; 483 else if (!_SIG_VALID(uap->sig)) 484 error = EINVAL; 485 else 486 tdksignal(ttd, uap->sig, &ksi); 487 PROC_UNLOCK(p); 488 } 489 return (error); 490 } 491 492 int 493 sys_thr_suspend(struct thread *td, struct thr_suspend_args *uap) 494 /* const struct timespec *timeout */ 495 { 496 struct timespec ts, *tsp; 497 int error; 498 499 tsp = NULL; 500 if (uap->timeout != NULL) { 501 error = umtx_copyin_timeout(uap->timeout, &ts); 502 if (error != 0) 503 return (error); 504 tsp = &ts; 505 } 506 507 return (kern_thr_suspend(td, tsp)); 508 } 509 510 int 511 kern_thr_suspend(struct thread *td, struct timespec *tsp) 512 { 513 struct proc *p = td->td_proc; 514 struct timeval tv; 515 int error = 0; 516 int timo = 0; 517 518 if (td->td_pflags & TDP_WAKEUP) { 519 td->td_pflags &= ~TDP_WAKEUP; 520 return (0); 521 } 522 523 if (tsp != NULL) { 524 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) 525 error = EWOULDBLOCK; 526 else { 527 TIMESPEC_TO_TIMEVAL(&tv, tsp); 528 timo = tvtohz(&tv); 529 } 530 } 531 532 PROC_LOCK(p); 533 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0) 534 error = msleep((void *)td, &p->p_mtx, 535 PCATCH, "lthr", timo); 536 537 if (td->td_flags & TDF_THRWAKEUP) { 538 thread_lock(td); 539 td->td_flags &= ~TDF_THRWAKEUP; 540 thread_unlock(td); 541 PROC_UNLOCK(p); 542 return (0); 543 } 544 PROC_UNLOCK(p); 545 if (error == EWOULDBLOCK) 546 error = ETIMEDOUT; 547 else if (error == ERESTART) { 548 if (timo != 0) 549 error = EINTR; 550 } 551 return (error); 552 } 553 554 int 555 sys_thr_wake(struct thread *td, struct thr_wake_args *uap) 556 /* long id */ 557 { 558 struct proc *p; 559 struct thread *ttd; 560 561 if (uap->id == td->td_tid) { 562 td->td_pflags |= TDP_WAKEUP; 563 return (0); 564 } 565 566 p = td->td_proc; 567 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 568 if (ttd == NULL) 569 return (ESRCH); 570 thread_lock(ttd); 571 ttd->td_flags |= TDF_THRWAKEUP; 572 thread_unlock(ttd); 573 wakeup((void *)ttd); 574 PROC_UNLOCK(p); 575 return (0); 576 } 577 578 int 579 sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap) 580 { 581 struct proc *p; 582 char name[MAXCOMLEN + 1]; 583 struct thread *ttd; 584 int error; 585 586 error = 0; 587 name[0] = '\0'; 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'; 593 } 594 if (error) 595 return (error); 596 } 597 p = td->td_proc; 598 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 599 if (ttd == NULL) 600 return (ESRCH); 601 strcpy(ttd->td_name, name); 602 #ifdef HWPMC_HOOKS 603 if (PMC_PROC_IS_USING_PMCS(p) || PMC_SYSTEM_SAMPLING_ACTIVE()) 604 PMC_CALL_HOOK_UNLOCKED(ttd, PMC_FN_THR_CREATE_LOG, NULL); 605 #endif 606 #ifdef KTR 607 sched_clear_tdname(ttd); 608 #endif 609 PROC_UNLOCK(p); 610 return (error); 611 } 612 613 int 614 kern_thr_alloc(struct proc *p, int pages, struct thread **ntd) 615 { 616 617 /* Have race condition but it is cheap. */ 618 if (p->p_numthreads >= max_threads_per_proc) { 619 ++max_threads_hits; 620 return (EPROCLIM); 621 } 622 623 *ntd = thread_alloc(pages); 624 if (*ntd == NULL) 625 return (ENOMEM); 626 627 return (0); 628 } 629