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 sys_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 sys_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 #ifdef RACCT 189 PROC_LOCK(td->td_proc); 190 error = racct_add(p, RACCT_NTHR, 1); 191 PROC_UNLOCK(td->td_proc); 192 if (error != 0) 193 return (EPROCLIM); 194 #endif 195 196 /* Initialize our td */ 197 newtd = thread_alloc(0); 198 if (newtd == NULL) { 199 error = ENOMEM; 200 goto fail; 201 } 202 203 /* 204 * Try the copyout as soon as we allocate the td so we don't 205 * have to tear things down in a failure case below. 206 * Here we copy out tid to two places, one for child and one 207 * for parent, because pthread can create a detached thread, 208 * if parent wants to safely access child tid, it has to provide 209 * its storage, because child thread may exit quickly and 210 * memory is freed before parent thread can access it. 211 */ 212 if ((child_tid != NULL && 213 suword_lwpid(child_tid, newtd->td_tid)) || 214 (parent_tid != NULL && 215 suword_lwpid(parent_tid, newtd->td_tid))) { 216 thread_free(newtd); 217 error = EFAULT; 218 goto fail; 219 } 220 221 bzero(&newtd->td_startzero, 222 __rangeof(struct thread, td_startzero, td_endzero)); 223 bcopy(&td->td_startcopy, &newtd->td_startcopy, 224 __rangeof(struct thread, td_startcopy, td_endcopy)); 225 newtd->td_proc = td->td_proc; 226 newtd->td_ucred = crhold(td->td_ucred); 227 228 cpu_set_upcall(newtd, td); 229 230 if (ctx != NULL) { /* old way to set user context */ 231 error = set_mcontext(newtd, ctx); 232 if (error != 0) { 233 thread_free(newtd); 234 crfree(td->td_ucred); 235 goto fail; 236 } 237 } else { 238 /* Set up our machine context. */ 239 stack.ss_sp = stack_base; 240 stack.ss_size = stack_size; 241 /* Set upcall address to user thread entry function. */ 242 cpu_set_upcall_kse(newtd, start_func, arg, &stack); 243 /* Setup user TLS address and TLS pointer register. */ 244 error = cpu_set_user_tls(newtd, tls_base); 245 if (error != 0) { 246 thread_free(newtd); 247 crfree(td->td_ucred); 248 goto fail; 249 } 250 } 251 252 PROC_LOCK(td->td_proc); 253 td->td_proc->p_flag |= P_HADTHREADS; 254 newtd->td_sigmask = td->td_sigmask; 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); 280 281 fail: 282 #ifdef RACCT 283 PROC_LOCK(p); 284 racct_sub(p, RACCT_NTHR, 1); 285 PROC_UNLOCK(p); 286 #endif 287 return (error); 288 } 289 290 int 291 sys_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 } 301 302 int 303 sys_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 racct_sub(p, RACCT_NTHR, 1); 320 321 /* 322 * Shutting down last thread in the proc. This will actually 323 * call exit() in the trampoline when it returns. 324 */ 325 if (p->p_numthreads != 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 } 338 339 int 340 sys_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 } 385 386 int 387 sys_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 } 442 443 int 444 sys_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 = copyin((const void *)uap->timeout, (void *)&ts, 453 sizeof(struct timespec)); 454 if (error != 0) 455 return (error); 456 tsp = &ts; 457 } 458 459 return (kern_thr_suspend(td, tsp)); 460 } 461 462 int 463 kern_thr_suspend(struct thread *td, struct timespec *tsp) 464 { 465 struct proc *p = td->td_proc; 466 struct timeval tv; 467 int error = 0; 468 int timo = 0; 469 470 if (td->td_pflags & TDP_WAKEUP) { 471 td->td_pflags &= ~TDP_WAKEUP; 472 return (0); 473 } 474 475 if (tsp != NULL) { 476 if (tsp->tv_nsec < 0 || tsp->tv_nsec > 1000000000) 477 return (EINVAL); 478 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) 479 error = EWOULDBLOCK; 480 else { 481 TIMESPEC_TO_TIMEVAL(&tv, tsp); 482 timo = tvtohz(&tv); 483 } 484 } 485 486 PROC_LOCK(p); 487 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0) 488 error = msleep((void *)td, &p->p_mtx, 489 PCATCH, "lthr", timo); 490 491 if (td->td_flags & TDF_THRWAKEUP) { 492 thread_lock(td); 493 td->td_flags &= ~TDF_THRWAKEUP; 494 thread_unlock(td); 495 PROC_UNLOCK(p); 496 return (0); 497 } 498 PROC_UNLOCK(p); 499 if (error == EWOULDBLOCK) 500 error = ETIMEDOUT; 501 else if (error == ERESTART) { 502 if (timo != 0) 503 error = EINTR; 504 } 505 return (error); 506 } 507 508 int 509 sys_thr_wake(struct thread *td, struct thr_wake_args *uap) 510 /* long id */ 511 { 512 struct proc *p; 513 struct thread *ttd; 514 515 if (uap->id == td->td_tid) { 516 td->td_pflags |= TDP_WAKEUP; 517 return (0); 518 } 519 520 p = td->td_proc; 521 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 522 if (ttd == NULL) 523 return (ESRCH); 524 thread_lock(ttd); 525 ttd->td_flags |= TDF_THRWAKEUP; 526 thread_unlock(ttd); 527 wakeup((void *)ttd); 528 PROC_UNLOCK(p); 529 return (0); 530 } 531 532 int 533 sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap) 534 { 535 struct proc *p; 536 char name[MAXCOMLEN + 1]; 537 struct thread *ttd; 538 int error; 539 540 error = 0; 541 name[0] = '\0'; 542 if (uap->name != NULL) { 543 error = copyinstr(uap->name, name, sizeof(name), 544 NULL); 545 if (error) 546 return (error); 547 } 548 p = td->td_proc; 549 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 550 if (ttd == NULL) 551 return (ESRCH); 552 strcpy(ttd->td_name, name); 553 PROC_UNLOCK(p); 554 return (error); 555 } 556