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 61 SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0, "thread allocation"); 62 63 static int max_threads_per_proc = 1500; 64 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_per_proc, CTLFLAG_RW, 65 &max_threads_per_proc, 0, "Limit on threads per proc"); 66 67 static int max_threads_hits; 68 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_hits, CTLFLAG_RD, 69 &max_threads_hits, 0, ""); 70 71 #ifdef COMPAT_FREEBSD32 72 73 static inline int 74 suword_lwpid(void *addr, lwpid_t lwpid) 75 { 76 int error; 77 78 if (SV_CURPROC_FLAG(SV_LP64)) 79 error = suword(addr, lwpid); 80 else 81 error = suword32(addr, lwpid); 82 return (error); 83 } 84 85 #else 86 #define suword_lwpid suword 87 #endif 88 89 static int create_thread(struct thread *td, mcontext_t *ctx, 90 void (*start_func)(void *), void *arg, 91 char *stack_base, size_t stack_size, 92 char *tls_base, 93 long *child_tid, long *parent_tid, 94 int flags, struct rtprio *rtp); 95 96 /* 97 * System call interface. 98 */ 99 int 100 thr_create(struct thread *td, struct thr_create_args *uap) 101 /* ucontext_t *ctx, long *id, int flags */ 102 { 103 ucontext_t ctx; 104 int error; 105 106 if ((error = copyin(uap->ctx, &ctx, sizeof(ctx)))) 107 return (error); 108 109 error = create_thread(td, &ctx.uc_mcontext, NULL, NULL, 110 NULL, 0, NULL, uap->id, NULL, uap->flags, NULL); 111 return (error); 112 } 113 114 int 115 thr_new(struct thread *td, struct thr_new_args *uap) 116 /* struct thr_param * */ 117 { 118 struct thr_param param; 119 int error; 120 121 if (uap->param_size < 0 || uap->param_size > sizeof(param)) 122 return (EINVAL); 123 bzero(¶m, sizeof(param)); 124 if ((error = copyin(uap->param, ¶m, uap->param_size))) 125 return (error); 126 return (kern_thr_new(td, ¶m)); 127 } 128 129 int 130 kern_thr_new(struct thread *td, struct thr_param *param) 131 { 132 struct rtprio rtp, *rtpp; 133 int error; 134 135 rtpp = NULL; 136 if (param->rtp != 0) { 137 error = copyin(param->rtp, &rtp, sizeof(struct rtprio)); 138 if (error) 139 return (error); 140 rtpp = &rtp; 141 } 142 error = create_thread(td, NULL, param->start_func, param->arg, 143 param->stack_base, param->stack_size, param->tls_base, 144 param->child_tid, param->parent_tid, param->flags, 145 rtpp); 146 return (error); 147 } 148 149 static int 150 create_thread(struct thread *td, mcontext_t *ctx, 151 void (*start_func)(void *), void *arg, 152 char *stack_base, size_t stack_size, 153 char *tls_base, 154 long *child_tid, long *parent_tid, 155 int flags, struct rtprio *rtp) 156 { 157 stack_t stack; 158 struct thread *newtd; 159 struct proc *p; 160 int error; 161 162 p = td->td_proc; 163 164 /* Have race condition but it is cheap. */ 165 if (p->p_numthreads >= max_threads_per_proc) { 166 ++max_threads_hits; 167 return (EPROCLIM); 168 } 169 170 if (rtp != NULL) { 171 switch(rtp->type) { 172 case RTP_PRIO_REALTIME: 173 case RTP_PRIO_FIFO: 174 /* Only root can set scheduler policy */ 175 if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0) 176 return (EPERM); 177 if (rtp->prio > RTP_PRIO_MAX) 178 return (EINVAL); 179 break; 180 case RTP_PRIO_NORMAL: 181 rtp->prio = 0; 182 break; 183 default: 184 return (EINVAL); 185 } 186 } 187 188 PROC_LOCK(td->td_proc); 189 error = racct_add(p, RACCT_NTHR, 1); 190 PROC_UNLOCK(td->td_proc); 191 if (error != 0) 192 return (EPROCLIM); 193 194 /* Initialize our td */ 195 newtd = thread_alloc(0); 196 if (newtd == NULL) { 197 error = ENOMEM; 198 goto fail; 199 } 200 201 /* 202 * Try the copyout as soon as we allocate the td so we don't 203 * have to tear things down in a failure case below. 204 * Here we copy out tid to two places, one for child and one 205 * for parent, because pthread can create a detached thread, 206 * if parent wants to safely access child tid, it has to provide 207 * its storage, because child thread may exit quickly and 208 * memory is freed before parent thread can access it. 209 */ 210 if ((child_tid != NULL && 211 suword_lwpid(child_tid, newtd->td_tid)) || 212 (parent_tid != NULL && 213 suword_lwpid(parent_tid, newtd->td_tid))) { 214 thread_free(newtd); 215 error = EFAULT; 216 goto fail; 217 } 218 219 bzero(&newtd->td_startzero, 220 __rangeof(struct thread, td_startzero, td_endzero)); 221 bcopy(&td->td_startcopy, &newtd->td_startcopy, 222 __rangeof(struct thread, td_startcopy, td_endcopy)); 223 newtd->td_proc = td->td_proc; 224 newtd->td_ucred = crhold(td->td_ucred); 225 226 cpu_set_upcall(newtd, td); 227 228 if (ctx != NULL) { /* old way to set user context */ 229 error = set_mcontext(newtd, ctx); 230 if (error != 0) { 231 thread_free(newtd); 232 crfree(td->td_ucred); 233 goto fail; 234 } 235 } else { 236 /* Set up our machine context. */ 237 stack.ss_sp = stack_base; 238 stack.ss_size = stack_size; 239 /* Set upcall address to user thread entry function. */ 240 cpu_set_upcall_kse(newtd, start_func, arg, &stack); 241 /* Setup user TLS address and TLS pointer register. */ 242 error = cpu_set_user_tls(newtd, tls_base); 243 if (error != 0) { 244 thread_free(newtd); 245 crfree(td->td_ucred); 246 goto fail; 247 } 248 } 249 250 PROC_LOCK(td->td_proc); 251 td->td_proc->p_flag |= P_HADTHREADS; 252 newtd->td_sigmask = td->td_sigmask; 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 PROC_UNLOCK(p); 262 263 tidhash_add(newtd); 264 265 thread_lock(newtd); 266 if (rtp != NULL) { 267 if (!(td->td_pri_class == PRI_TIMESHARE && 268 rtp->type == RTP_PRIO_NORMAL)) { 269 rtp_to_pri(rtp, newtd); 270 sched_prio(newtd, newtd->td_user_pri); 271 } /* ignore timesharing class */ 272 } 273 TD_SET_CAN_RUN(newtd); 274 sched_add(newtd, SRQ_BORING); 275 thread_unlock(newtd); 276 277 return (0); 278 279 fail: 280 PROC_LOCK(p); 281 racct_sub(p, RACCT_NTHR, 1); 282 PROC_UNLOCK(p); 283 return (error); 284 } 285 286 int 287 thr_self(struct thread *td, struct thr_self_args *uap) 288 /* long *id */ 289 { 290 int error; 291 292 error = suword_lwpid(uap->id, (unsigned)td->td_tid); 293 if (error == -1) 294 return (EFAULT); 295 return (0); 296 } 297 298 int 299 thr_exit(struct thread *td, struct thr_exit_args *uap) 300 /* long *state */ 301 { 302 struct proc *p; 303 304 p = td->td_proc; 305 306 /* Signal userland that it can free the stack. */ 307 if ((void *)uap->state != NULL) { 308 suword_lwpid(uap->state, 1); 309 kern_umtx_wake(td, uap->state, INT_MAX, 0); 310 } 311 312 rw_wlock(&tidhash_lock); 313 314 PROC_LOCK(p); 315 racct_sub(p, RACCT_NTHR, 1); 316 317 /* 318 * Shutting down last thread in the proc. This will actually 319 * call exit() in the trampoline when it returns. 320 */ 321 if (p->p_numthreads != 1) { 322 LIST_REMOVE(td, td_hash); 323 rw_wunlock(&tidhash_lock); 324 tdsigcleanup(td); 325 PROC_SLOCK(p); 326 thread_stopped(p); 327 thread_exit(); 328 /* NOTREACHED */ 329 } 330 PROC_UNLOCK(p); 331 rw_wunlock(&tidhash_lock); 332 return (0); 333 } 334 335 int 336 thr_kill(struct thread *td, struct thr_kill_args *uap) 337 /* long id, int sig */ 338 { 339 ksiginfo_t ksi; 340 struct thread *ttd; 341 struct proc *p; 342 int error; 343 344 p = td->td_proc; 345 ksiginfo_init(&ksi); 346 ksi.ksi_signo = uap->sig; 347 ksi.ksi_code = SI_LWP; 348 ksi.ksi_pid = p->p_pid; 349 ksi.ksi_uid = td->td_ucred->cr_ruid; 350 if (uap->id == -1) { 351 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 352 error = EINVAL; 353 } else { 354 error = ESRCH; 355 PROC_LOCK(p); 356 FOREACH_THREAD_IN_PROC(p, ttd) { 357 if (ttd != td) { 358 error = 0; 359 if (uap->sig == 0) 360 break; 361 tdksignal(ttd, uap->sig, &ksi); 362 } 363 } 364 PROC_UNLOCK(p); 365 } 366 } else { 367 error = 0; 368 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 369 if (ttd == NULL) 370 return (ESRCH); 371 if (uap->sig == 0) 372 ; 373 else if (!_SIG_VALID(uap->sig)) 374 error = EINVAL; 375 else 376 tdksignal(ttd, uap->sig, &ksi); 377 PROC_UNLOCK(ttd->td_proc); 378 } 379 return (error); 380 } 381 382 int 383 thr_kill2(struct thread *td, struct thr_kill2_args *uap) 384 /* pid_t pid, long id, int sig */ 385 { 386 ksiginfo_t ksi; 387 struct thread *ttd; 388 struct proc *p; 389 int error; 390 391 AUDIT_ARG_SIGNUM(uap->sig); 392 393 ksiginfo_init(&ksi); 394 ksi.ksi_signo = uap->sig; 395 ksi.ksi_code = SI_LWP; 396 ksi.ksi_pid = td->td_proc->p_pid; 397 ksi.ksi_uid = td->td_ucred->cr_ruid; 398 if (uap->id == -1) { 399 if ((p = pfind(uap->pid)) == NULL) 400 return (ESRCH); 401 AUDIT_ARG_PROCESS(p); 402 error = p_cansignal(td, p, uap->sig); 403 if (error) { 404 PROC_UNLOCK(p); 405 return (error); 406 } 407 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 408 error = EINVAL; 409 } else { 410 error = ESRCH; 411 FOREACH_THREAD_IN_PROC(p, ttd) { 412 if (ttd != td) { 413 error = 0; 414 if (uap->sig == 0) 415 break; 416 tdksignal(ttd, uap->sig, &ksi); 417 } 418 } 419 } 420 PROC_UNLOCK(p); 421 } else { 422 ttd = tdfind((lwpid_t)uap->id, uap->pid); 423 if (ttd == NULL) 424 return (ESRCH); 425 p = ttd->td_proc; 426 AUDIT_ARG_PROCESS(p); 427 error = p_cansignal(td, p, uap->sig); 428 if (uap->sig == 0) 429 ; 430 else if (!_SIG_VALID(uap->sig)) 431 error = EINVAL; 432 else 433 tdksignal(ttd, uap->sig, &ksi); 434 PROC_UNLOCK(p); 435 } 436 return (error); 437 } 438 439 int 440 thr_suspend(struct thread *td, struct thr_suspend_args *uap) 441 /* const struct timespec *timeout */ 442 { 443 struct timespec ts, *tsp; 444 int error; 445 446 tsp = NULL; 447 if (uap->timeout != NULL) { 448 error = copyin((const void *)uap->timeout, (void *)&ts, 449 sizeof(struct timespec)); 450 if (error != 0) 451 return (error); 452 tsp = &ts; 453 } 454 455 return (kern_thr_suspend(td, tsp)); 456 } 457 458 int 459 kern_thr_suspend(struct thread *td, struct timespec *tsp) 460 { 461 struct proc *p = td->td_proc; 462 struct timeval tv; 463 int error = 0; 464 int timo = 0; 465 466 if (td->td_pflags & TDP_WAKEUP) { 467 td->td_pflags &= ~TDP_WAKEUP; 468 return (0); 469 } 470 471 if (tsp != NULL) { 472 if (tsp->tv_nsec < 0 || tsp->tv_nsec > 1000000000) 473 return (EINVAL); 474 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) 475 error = EWOULDBLOCK; 476 else { 477 TIMESPEC_TO_TIMEVAL(&tv, tsp); 478 timo = tvtohz(&tv); 479 } 480 } 481 482 PROC_LOCK(p); 483 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0) 484 error = msleep((void *)td, &p->p_mtx, 485 PCATCH, "lthr", timo); 486 487 if (td->td_flags & TDF_THRWAKEUP) { 488 thread_lock(td); 489 td->td_flags &= ~TDF_THRWAKEUP; 490 thread_unlock(td); 491 PROC_UNLOCK(p); 492 return (0); 493 } 494 PROC_UNLOCK(p); 495 if (error == EWOULDBLOCK) 496 error = ETIMEDOUT; 497 else if (error == ERESTART) { 498 if (timo != 0) 499 error = EINTR; 500 } 501 return (error); 502 } 503 504 int 505 thr_wake(struct thread *td, struct thr_wake_args *uap) 506 /* long id */ 507 { 508 struct proc *p; 509 struct thread *ttd; 510 511 if (uap->id == td->td_tid) { 512 td->td_pflags |= TDP_WAKEUP; 513 return (0); 514 } 515 516 p = td->td_proc; 517 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 518 if (ttd == NULL) 519 return (ESRCH); 520 thread_lock(ttd); 521 ttd->td_flags |= TDF_THRWAKEUP; 522 thread_unlock(ttd); 523 wakeup((void *)ttd); 524 PROC_UNLOCK(p); 525 return (0); 526 } 527 528 int 529 thr_set_name(struct thread *td, struct thr_set_name_args *uap) 530 { 531 struct proc *p; 532 char name[MAXCOMLEN + 1]; 533 struct thread *ttd; 534 int error; 535 536 error = 0; 537 name[0] = '\0'; 538 if (uap->name != NULL) { 539 error = copyinstr(uap->name, name, sizeof(name), 540 NULL); 541 if (error) 542 return (error); 543 } 544 p = td->td_proc; 545 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 546 if (ttd == NULL) 547 return (ESRCH); 548 strcpy(ttd->td_name, name); 549 PROC_UNLOCK(p); 550 return (error); 551 } 552