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 <vm/vm_domain.h> 58 59 #include <machine/frame.h> 60 61 #include <security/audit/audit.h> 62 63 static SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0, 64 "thread allocation"); 65 66 static int max_threads_per_proc = 1500; 67 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_per_proc, CTLFLAG_RW, 68 &max_threads_per_proc, 0, "Limit on threads per proc"); 69 70 static int max_threads_hits; 71 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_hits, CTLFLAG_RD, 72 &max_threads_hits, 0, "kern.threads.max_threads_per_proc hit count"); 73 74 #ifdef COMPAT_FREEBSD32 75 76 static inline int 77 suword_lwpid(void *addr, lwpid_t lwpid) 78 { 79 int error; 80 81 if (SV_CURPROC_FLAG(SV_LP64)) 82 error = suword(addr, lwpid); 83 else 84 error = suword32(addr, lwpid); 85 return (error); 86 } 87 88 #else 89 #define suword_lwpid suword 90 #endif 91 92 /* 93 * System call interface. 94 */ 95 96 struct thr_create_initthr_args { 97 ucontext_t ctx; 98 long *tid; 99 }; 100 101 static int 102 thr_create_initthr(struct thread *td, void *thunk) 103 { 104 struct thr_create_initthr_args *args; 105 106 /* Copy out the child tid. */ 107 args = thunk; 108 if (args->tid != NULL && suword_lwpid(args->tid, td->td_tid)) 109 return (EFAULT); 110 111 return (set_mcontext(td, &args->ctx.uc_mcontext)); 112 } 113 114 int 115 sys_thr_create(struct thread *td, struct thr_create_args *uap) 116 /* ucontext_t *ctx, long *id, int flags */ 117 { 118 struct thr_create_initthr_args args; 119 int error; 120 121 if ((error = copyin(uap->ctx, &args.ctx, sizeof(args.ctx)))) 122 return (error); 123 args.tid = uap->id; 124 return (thread_create(td, NULL, thr_create_initthr, &args)); 125 } 126 127 int 128 sys_thr_new(struct thread *td, struct thr_new_args *uap) 129 /* struct thr_param * */ 130 { 131 struct thr_param param; 132 int error; 133 134 if (uap->param_size < 0 || uap->param_size > sizeof(param)) 135 return (EINVAL); 136 bzero(¶m, sizeof(param)); 137 if ((error = copyin(uap->param, ¶m, uap->param_size))) 138 return (error); 139 return (kern_thr_new(td, ¶m)); 140 } 141 142 static int 143 thr_new_initthr(struct thread *td, void *thunk) 144 { 145 stack_t stack; 146 struct thr_param *param; 147 148 /* 149 * Here we copy out tid to two places, one for child and one 150 * for parent, because pthread can create a detached thread, 151 * if parent wants to safely access child tid, it has to provide 152 * its storage, because child thread may exit quickly and 153 * memory is freed before parent thread can access it. 154 */ 155 param = thunk; 156 if ((param->child_tid != NULL && 157 suword_lwpid(param->child_tid, td->td_tid)) || 158 (param->parent_tid != NULL && 159 suword_lwpid(param->parent_tid, td->td_tid))) 160 return (EFAULT); 161 162 /* Set up our machine context. */ 163 stack.ss_sp = param->stack_base; 164 stack.ss_size = param->stack_size; 165 /* Set upcall address to user thread entry function. */ 166 cpu_set_upcall_kse(td, param->start_func, param->arg, &stack); 167 /* Setup user TLS address and TLS pointer register. */ 168 return (cpu_set_user_tls(td, param->tls_base)); 169 } 170 171 int 172 kern_thr_new(struct thread *td, struct thr_param *param) 173 { 174 struct rtprio rtp, *rtpp; 175 int error; 176 177 rtpp = NULL; 178 if (param->rtp != 0) { 179 error = copyin(param->rtp, &rtp, sizeof(struct rtprio)); 180 if (error) 181 return (error); 182 rtpp = &rtp; 183 } 184 return (thread_create(td, rtpp, thr_new_initthr, param)); 185 } 186 187 int 188 thread_create(struct thread *td, struct rtprio *rtp, 189 int (*initialize_thread)(struct thread *, void *), void *thunk) 190 { 191 struct thread *newtd; 192 struct proc *p; 193 int error; 194 195 p = td->td_proc; 196 197 if (rtp != NULL) { 198 switch(rtp->type) { 199 case RTP_PRIO_REALTIME: 200 case RTP_PRIO_FIFO: 201 /* Only root can set scheduler policy */ 202 if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0) 203 return (EPERM); 204 if (rtp->prio > RTP_PRIO_MAX) 205 return (EINVAL); 206 break; 207 case RTP_PRIO_NORMAL: 208 rtp->prio = 0; 209 break; 210 default: 211 return (EINVAL); 212 } 213 } 214 215 #ifdef RACCT 216 if (racct_enable) { 217 PROC_LOCK(p); 218 error = racct_add(p, RACCT_NTHR, 1); 219 PROC_UNLOCK(p); 220 if (error != 0) 221 return (EPROCLIM); 222 } 223 #endif 224 225 /* Initialize our td */ 226 error = kern_thr_alloc(p, 0, &newtd); 227 if (error) 228 goto fail; 229 230 cpu_set_upcall(newtd, td); 231 232 bzero(&newtd->td_startzero, 233 __rangeof(struct thread, td_startzero, td_endzero)); 234 bcopy(&td->td_startcopy, &newtd->td_startcopy, 235 __rangeof(struct thread, td_startcopy, td_endcopy)); 236 newtd->td_proc = td->td_proc; 237 thread_cow_get(newtd, td); 238 239 error = initialize_thread(newtd, thunk); 240 if (error != 0) { 241 thread_cow_free(newtd); 242 thread_free(newtd); 243 goto fail; 244 } 245 246 PROC_LOCK(p); 247 p->p_flag |= P_HADTHREADS; 248 thread_link(newtd, p); 249 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); 250 thread_lock(td); 251 /* let the scheduler know about these things. */ 252 sched_fork_thread(td, newtd); 253 thread_unlock(td); 254 if (P_SHOULDSTOP(p)) 255 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; 256 257 /* 258 * Copy the existing thread VM policy into the new thread. 259 */ 260 vm_domain_policy_localcopy(&newtd->td_vm_dom_policy, 261 &td->td_vm_dom_policy); 262 263 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 if (racct_enable) { 284 PROC_LOCK(p); 285 racct_sub(p, RACCT_NTHR, 1); 286 PROC_UNLOCK(p); 287 } 288 #endif 289 return (error); 290 } 291 292 int 293 sys_thr_self(struct thread *td, struct thr_self_args *uap) 294 /* long *id */ 295 { 296 int error; 297 298 error = suword_lwpid(uap->id, (unsigned)td->td_tid); 299 if (error == -1) 300 return (EFAULT); 301 return (0); 302 } 303 304 int 305 sys_thr_exit(struct thread *td, struct thr_exit_args *uap) 306 /* long *state */ 307 { 308 309 /* Signal userland that it can free the stack. */ 310 if ((void *)uap->state != NULL) { 311 suword_lwpid(uap->state, 1); 312 kern_umtx_wake(td, uap->state, INT_MAX, 0); 313 } 314 315 return (kern_thr_exit(td)); 316 } 317 318 int 319 kern_thr_exit(struct thread *td) 320 { 321 struct proc *p; 322 323 p = td->td_proc; 324 325 rw_wlock(&tidhash_lock); 326 PROC_LOCK(p); 327 328 if (p->p_numthreads != 1) { 329 racct_sub(p, RACCT_NTHR, 1); 330 LIST_REMOVE(td, td_hash); 331 rw_wunlock(&tidhash_lock); 332 tdsigcleanup(td); 333 umtx_thread_exit(td); 334 PROC_SLOCK(p); 335 thread_stopped(p); 336 thread_exit(); 337 /* NOTREACHED */ 338 } 339 340 /* 341 * Ignore attempts to shut down last thread in the proc. This 342 * will actually call _exit(2) in the usermode trampoline when 343 * it returns. 344 */ 345 PROC_UNLOCK(p); 346 rw_wunlock(&tidhash_lock); 347 return (0); 348 } 349 350 int 351 sys_thr_kill(struct thread *td, struct thr_kill_args *uap) 352 /* long id, int sig */ 353 { 354 ksiginfo_t ksi; 355 struct thread *ttd; 356 struct proc *p; 357 int error; 358 359 p = td->td_proc; 360 ksiginfo_init(&ksi); 361 ksi.ksi_signo = uap->sig; 362 ksi.ksi_code = SI_LWP; 363 ksi.ksi_pid = p->p_pid; 364 ksi.ksi_uid = td->td_ucred->cr_ruid; 365 if (uap->id == -1) { 366 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 367 error = EINVAL; 368 } else { 369 error = ESRCH; 370 PROC_LOCK(p); 371 FOREACH_THREAD_IN_PROC(p, ttd) { 372 if (ttd != td) { 373 error = 0; 374 if (uap->sig == 0) 375 break; 376 tdksignal(ttd, uap->sig, &ksi); 377 } 378 } 379 PROC_UNLOCK(p); 380 } 381 } else { 382 error = 0; 383 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 384 if (ttd == NULL) 385 return (ESRCH); 386 if (uap->sig == 0) 387 ; 388 else if (!_SIG_VALID(uap->sig)) 389 error = EINVAL; 390 else 391 tdksignal(ttd, uap->sig, &ksi); 392 PROC_UNLOCK(ttd->td_proc); 393 } 394 return (error); 395 } 396 397 int 398 sys_thr_kill2(struct thread *td, struct thr_kill2_args *uap) 399 /* pid_t pid, long id, int sig */ 400 { 401 ksiginfo_t ksi; 402 struct thread *ttd; 403 struct proc *p; 404 int error; 405 406 AUDIT_ARG_SIGNUM(uap->sig); 407 408 ksiginfo_init(&ksi); 409 ksi.ksi_signo = uap->sig; 410 ksi.ksi_code = SI_LWP; 411 ksi.ksi_pid = td->td_proc->p_pid; 412 ksi.ksi_uid = td->td_ucred->cr_ruid; 413 if (uap->id == -1) { 414 if ((p = pfind(uap->pid)) == NULL) 415 return (ESRCH); 416 AUDIT_ARG_PROCESS(p); 417 error = p_cansignal(td, p, uap->sig); 418 if (error) { 419 PROC_UNLOCK(p); 420 return (error); 421 } 422 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 423 error = EINVAL; 424 } else { 425 error = ESRCH; 426 FOREACH_THREAD_IN_PROC(p, ttd) { 427 if (ttd != td) { 428 error = 0; 429 if (uap->sig == 0) 430 break; 431 tdksignal(ttd, uap->sig, &ksi); 432 } 433 } 434 } 435 PROC_UNLOCK(p); 436 } else { 437 ttd = tdfind((lwpid_t)uap->id, uap->pid); 438 if (ttd == NULL) 439 return (ESRCH); 440 p = ttd->td_proc; 441 AUDIT_ARG_PROCESS(p); 442 error = p_cansignal(td, p, uap->sig); 443 if (uap->sig == 0) 444 ; 445 else if (!_SIG_VALID(uap->sig)) 446 error = EINVAL; 447 else 448 tdksignal(ttd, uap->sig, &ksi); 449 PROC_UNLOCK(p); 450 } 451 return (error); 452 } 453 454 int 455 sys_thr_suspend(struct thread *td, struct thr_suspend_args *uap) 456 /* const struct timespec *timeout */ 457 { 458 struct timespec ts, *tsp; 459 int error; 460 461 tsp = NULL; 462 if (uap->timeout != NULL) { 463 error = umtx_copyin_timeout(uap->timeout, &ts); 464 if (error != 0) 465 return (error); 466 tsp = &ts; 467 } 468 469 return (kern_thr_suspend(td, tsp)); 470 } 471 472 int 473 kern_thr_suspend(struct thread *td, struct timespec *tsp) 474 { 475 struct proc *p = td->td_proc; 476 struct timeval tv; 477 int error = 0; 478 int timo = 0; 479 480 if (td->td_pflags & TDP_WAKEUP) { 481 td->td_pflags &= ~TDP_WAKEUP; 482 return (0); 483 } 484 485 if (tsp != NULL) { 486 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) 487 error = EWOULDBLOCK; 488 else { 489 TIMESPEC_TO_TIMEVAL(&tv, tsp); 490 timo = tvtohz(&tv); 491 } 492 } 493 494 PROC_LOCK(p); 495 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0) 496 error = msleep((void *)td, &p->p_mtx, 497 PCATCH, "lthr", timo); 498 499 if (td->td_flags & TDF_THRWAKEUP) { 500 thread_lock(td); 501 td->td_flags &= ~TDF_THRWAKEUP; 502 thread_unlock(td); 503 PROC_UNLOCK(p); 504 return (0); 505 } 506 PROC_UNLOCK(p); 507 if (error == EWOULDBLOCK) 508 error = ETIMEDOUT; 509 else if (error == ERESTART) { 510 if (timo != 0) 511 error = EINTR; 512 } 513 return (error); 514 } 515 516 int 517 sys_thr_wake(struct thread *td, struct thr_wake_args *uap) 518 /* long id */ 519 { 520 struct proc *p; 521 struct thread *ttd; 522 523 if (uap->id == td->td_tid) { 524 td->td_pflags |= TDP_WAKEUP; 525 return (0); 526 } 527 528 p = td->td_proc; 529 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 530 if (ttd == NULL) 531 return (ESRCH); 532 thread_lock(ttd); 533 ttd->td_flags |= TDF_THRWAKEUP; 534 thread_unlock(ttd); 535 wakeup((void *)ttd); 536 PROC_UNLOCK(p); 537 return (0); 538 } 539 540 int 541 sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap) 542 { 543 struct proc *p; 544 char name[MAXCOMLEN + 1]; 545 struct thread *ttd; 546 int error; 547 548 error = 0; 549 name[0] = '\0'; 550 if (uap->name != NULL) { 551 error = copyinstr(uap->name, name, sizeof(name), 552 NULL); 553 if (error) 554 return (error); 555 } 556 p = td->td_proc; 557 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 558 if (ttd == NULL) 559 return (ESRCH); 560 strcpy(ttd->td_name, name); 561 #ifdef KTR 562 sched_clear_tdname(ttd); 563 #endif 564 PROC_UNLOCK(p); 565 return (error); 566 } 567 568 int 569 kern_thr_alloc(struct proc *p, int pages, struct thread **ntd) 570 { 571 572 /* Have race condition but it is cheap. */ 573 if (p->p_numthreads >= max_threads_per_proc) { 574 ++max_threads_hits; 575 return (EPROCLIM); 576 } 577 578 *ntd = thread_alloc(pages); 579 if (*ntd == NULL) 580 return (ENOMEM); 581 582 return (0); 583 } 584