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/resourcevar.h> 40 #include <sys/sched.h> 41 #include <sys/sysctl.h> 42 #include <sys/smp.h> 43 #include <sys/syscallsubr.h> 44 #include <sys/sysent.h> 45 #include <sys/systm.h> 46 #include <sys/sysproto.h> 47 #include <sys/signalvar.h> 48 #include <sys/ucontext.h> 49 #include <sys/thr.h> 50 #include <sys/rtprio.h> 51 #include <sys/umtx.h> 52 #include <sys/limits.h> 53 54 #include <machine/frame.h> 55 56 #include <security/audit/audit.h> 57 58 #ifdef COMPAT_FREEBSD32 59 60 static inline int 61 suword_lwpid(void *addr, lwpid_t lwpid) 62 { 63 int error; 64 65 if (SV_CURPROC_FLAG(SV_LP64)) 66 error = suword(addr, lwpid); 67 else 68 error = suword32(addr, lwpid); 69 return (error); 70 } 71 72 #else 73 #define suword_lwpid suword 74 #endif 75 76 extern int max_threads_per_proc; 77 extern int max_threads_hits; 78 79 static int create_thread(struct thread *td, mcontext_t *ctx, 80 void (*start_func)(void *), void *arg, 81 char *stack_base, size_t stack_size, 82 char *tls_base, 83 long *child_tid, long *parent_tid, 84 int flags, struct rtprio *rtp); 85 86 /* 87 * System call interface. 88 */ 89 int 90 thr_create(struct thread *td, struct thr_create_args *uap) 91 /* ucontext_t *ctx, long *id, int flags */ 92 { 93 ucontext_t ctx; 94 int error; 95 96 if ((error = copyin(uap->ctx, &ctx, sizeof(ctx)))) 97 return (error); 98 99 error = create_thread(td, &ctx.uc_mcontext, NULL, NULL, 100 NULL, 0, NULL, uap->id, NULL, uap->flags, NULL); 101 return (error); 102 } 103 104 int 105 thr_new(struct thread *td, struct thr_new_args *uap) 106 /* struct thr_param * */ 107 { 108 struct thr_param param; 109 int error; 110 111 if (uap->param_size < 0 || uap->param_size > sizeof(param)) 112 return (EINVAL); 113 bzero(¶m, sizeof(param)); 114 if ((error = copyin(uap->param, ¶m, uap->param_size))) 115 return (error); 116 return (kern_thr_new(td, ¶m)); 117 } 118 119 int 120 kern_thr_new(struct thread *td, struct thr_param *param) 121 { 122 struct rtprio rtp, *rtpp; 123 int error; 124 125 rtpp = NULL; 126 if (param->rtp != 0) { 127 error = copyin(param->rtp, &rtp, sizeof(struct rtprio)); 128 if (error) 129 return (error); 130 rtpp = &rtp; 131 } 132 error = create_thread(td, NULL, param->start_func, param->arg, 133 param->stack_base, param->stack_size, param->tls_base, 134 param->child_tid, param->parent_tid, param->flags, 135 rtpp); 136 return (error); 137 } 138 139 static int 140 create_thread(struct thread *td, mcontext_t *ctx, 141 void (*start_func)(void *), void *arg, 142 char *stack_base, size_t stack_size, 143 char *tls_base, 144 long *child_tid, long *parent_tid, 145 int flags, struct rtprio *rtp) 146 { 147 stack_t stack; 148 struct thread *newtd; 149 struct proc *p; 150 int error; 151 152 p = td->td_proc; 153 154 /* Have race condition but it is cheap. */ 155 if (p->p_numthreads >= max_threads_per_proc) { 156 ++max_threads_hits; 157 return (EPROCLIM); 158 } 159 160 if (rtp != NULL) { 161 switch(rtp->type) { 162 case RTP_PRIO_REALTIME: 163 case RTP_PRIO_FIFO: 164 /* Only root can set scheduler policy */ 165 if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0) 166 return (EPERM); 167 if (rtp->prio > RTP_PRIO_MAX) 168 return (EINVAL); 169 break; 170 case RTP_PRIO_NORMAL: 171 rtp->prio = 0; 172 break; 173 default: 174 return (EINVAL); 175 } 176 } 177 178 /* Initialize our td */ 179 newtd = thread_alloc(0); 180 if (newtd == NULL) 181 return (ENOMEM); 182 183 /* 184 * Try the copyout as soon as we allocate the td so we don't 185 * have to tear things down in a failure case below. 186 * Here we copy out tid to two places, one for child and one 187 * for parent, because pthread can create a detached thread, 188 * if parent wants to safely access child tid, it has to provide 189 * its storage, because child thread may exit quickly and 190 * memory is freed before parent thread can access it. 191 */ 192 if ((child_tid != NULL && 193 suword_lwpid(child_tid, newtd->td_tid)) || 194 (parent_tid != NULL && 195 suword_lwpid(parent_tid, newtd->td_tid))) { 196 thread_free(newtd); 197 return (EFAULT); 198 } 199 200 bzero(&newtd->td_startzero, 201 __rangeof(struct thread, td_startzero, td_endzero)); 202 bcopy(&td->td_startcopy, &newtd->td_startcopy, 203 __rangeof(struct thread, td_startcopy, td_endcopy)); 204 newtd->td_proc = td->td_proc; 205 newtd->td_ucred = crhold(td->td_ucred); 206 207 cpu_set_upcall(newtd, td); 208 209 if (ctx != NULL) { /* old way to set user context */ 210 error = set_mcontext(newtd, ctx); 211 if (error != 0) { 212 thread_free(newtd); 213 crfree(td->td_ucred); 214 return (error); 215 } 216 } else { 217 /* Set up our machine context. */ 218 stack.ss_sp = stack_base; 219 stack.ss_size = stack_size; 220 /* Set upcall address to user thread entry function. */ 221 cpu_set_upcall_kse(newtd, start_func, arg, &stack); 222 /* Setup user TLS address and TLS pointer register. */ 223 error = cpu_set_user_tls(newtd, tls_base); 224 if (error != 0) { 225 thread_free(newtd); 226 crfree(td->td_ucred); 227 return (error); 228 } 229 } 230 231 PROC_LOCK(td->td_proc); 232 td->td_proc->p_flag |= P_HADTHREADS; 233 newtd->td_sigmask = td->td_sigmask; 234 thread_link(newtd, p); 235 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); 236 thread_lock(td); 237 /* let the scheduler know about these things. */ 238 sched_fork_thread(td, newtd); 239 thread_unlock(td); 240 if (P_SHOULDSTOP(p)) 241 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; 242 PROC_UNLOCK(p); 243 thread_lock(newtd); 244 if (rtp != NULL) { 245 if (!(td->td_pri_class == PRI_TIMESHARE && 246 rtp->type == RTP_PRIO_NORMAL)) { 247 rtp_to_pri(rtp, newtd); 248 sched_prio(newtd, newtd->td_user_pri); 249 } /* ignore timesharing class */ 250 } 251 TD_SET_CAN_RUN(newtd); 252 sched_add(newtd, SRQ_BORING); 253 thread_unlock(newtd); 254 255 return (0); 256 } 257 258 int 259 thr_self(struct thread *td, struct thr_self_args *uap) 260 /* long *id */ 261 { 262 int error; 263 264 error = suword_lwpid(uap->id, (unsigned)td->td_tid); 265 if (error == -1) 266 return (EFAULT); 267 return (0); 268 } 269 270 int 271 thr_exit(struct thread *td, struct thr_exit_args *uap) 272 /* long *state */ 273 { 274 struct proc *p; 275 276 p = td->td_proc; 277 278 /* Signal userland that it can free the stack. */ 279 if ((void *)uap->state != NULL) { 280 suword_lwpid(uap->state, 1); 281 kern_umtx_wake(td, uap->state, INT_MAX, 0); 282 } 283 284 PROC_LOCK(p); 285 tdsigcleanup(td); 286 PROC_SLOCK(p); 287 288 /* 289 * Shutting down last thread in the proc. This will actually 290 * call exit() in the trampoline when it returns. 291 */ 292 if (p->p_numthreads != 1) { 293 thread_stopped(p); 294 thread_exit(); 295 /* NOTREACHED */ 296 } 297 PROC_SUNLOCK(p); 298 PROC_UNLOCK(p); 299 return (0); 300 } 301 302 int 303 thr_kill(struct thread *td, struct thr_kill_args *uap) 304 /* long id, int sig */ 305 { 306 ksiginfo_t ksi; 307 struct thread *ttd; 308 struct proc *p; 309 int error; 310 311 p = td->td_proc; 312 error = 0; 313 ksiginfo_init(&ksi); 314 ksi.ksi_signo = uap->sig; 315 ksi.ksi_code = SI_USER; 316 ksi.ksi_pid = p->p_pid; 317 ksi.ksi_uid = td->td_ucred->cr_ruid; 318 PROC_LOCK(p); 319 if (uap->id == -1) { 320 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 321 error = EINVAL; 322 } else { 323 error = ESRCH; 324 FOREACH_THREAD_IN_PROC(p, ttd) { 325 if (ttd != td) { 326 error = 0; 327 if (uap->sig == 0) 328 break; 329 tdksignal(ttd, uap->sig, &ksi); 330 } 331 } 332 } 333 } else { 334 if (uap->id != td->td_tid) 335 ttd = thread_find(p, uap->id); 336 else 337 ttd = td; 338 if (ttd == NULL) 339 error = ESRCH; 340 else if (uap->sig == 0) 341 ; 342 else if (!_SIG_VALID(uap->sig)) 343 error = EINVAL; 344 else 345 tdksignal(ttd, uap->sig, &ksi); 346 } 347 PROC_UNLOCK(p); 348 return (error); 349 } 350 351 int 352 thr_kill2(struct thread *td, struct thr_kill2_args *uap) 353 /* pid_t pid, long id, int sig */ 354 { 355 ksiginfo_t ksi; 356 struct thread *ttd; 357 struct proc *p; 358 int error; 359 360 AUDIT_ARG_SIGNUM(uap->sig); 361 362 if (uap->pid == td->td_proc->p_pid) { 363 p = td->td_proc; 364 PROC_LOCK(p); 365 } else if ((p = pfind(uap->pid)) == NULL) { 366 return (ESRCH); 367 } 368 AUDIT_ARG_PROCESS(p); 369 370 error = p_cansignal(td, p, uap->sig); 371 if (error == 0) { 372 ksiginfo_init(&ksi); 373 ksi.ksi_signo = uap->sig; 374 ksi.ksi_code = SI_USER; 375 ksi.ksi_pid = td->td_proc->p_pid; 376 ksi.ksi_uid = td->td_ucred->cr_ruid; 377 if (uap->id == -1) { 378 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 379 error = EINVAL; 380 } else { 381 error = ESRCH; 382 FOREACH_THREAD_IN_PROC(p, ttd) { 383 if (ttd != td) { 384 error = 0; 385 if (uap->sig == 0) 386 break; 387 tdksignal(ttd, uap->sig, &ksi); 388 } 389 } 390 } 391 } else { 392 if (uap->id != td->td_tid) 393 ttd = thread_find(p, uap->id); 394 else 395 ttd = td; 396 if (ttd == NULL) 397 error = ESRCH; 398 else if (uap->sig == 0) 399 ; 400 else if (!_SIG_VALID(uap->sig)) 401 error = EINVAL; 402 else 403 tdksignal(ttd, uap->sig, &ksi); 404 } 405 } 406 PROC_UNLOCK(p); 407 return (error); 408 } 409 410 int 411 thr_suspend(struct thread *td, struct thr_suspend_args *uap) 412 /* const struct timespec *timeout */ 413 { 414 struct timespec ts, *tsp; 415 int error; 416 417 tsp = NULL; 418 if (uap->timeout != NULL) { 419 error = copyin((const void *)uap->timeout, (void *)&ts, 420 sizeof(struct timespec)); 421 if (error != 0) 422 return (error); 423 tsp = &ts; 424 } 425 426 return (kern_thr_suspend(td, tsp)); 427 } 428 429 int 430 kern_thr_suspend(struct thread *td, struct timespec *tsp) 431 { 432 struct timeval tv; 433 int error = 0, hz = 0; 434 435 if (tsp != NULL) { 436 if (tsp->tv_nsec < 0 || tsp->tv_nsec > 1000000000) 437 return (EINVAL); 438 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) 439 return (ETIMEDOUT); 440 TIMESPEC_TO_TIMEVAL(&tv, tsp); 441 hz = tvtohz(&tv); 442 } 443 444 if (td->td_pflags & TDP_WAKEUP) { 445 td->td_pflags &= ~TDP_WAKEUP; 446 return (0); 447 } 448 449 PROC_LOCK(td->td_proc); 450 if ((td->td_flags & TDF_THRWAKEUP) == 0) 451 error = msleep((void *)td, &td->td_proc->p_mtx, PCATCH, "lthr", 452 hz); 453 if (td->td_flags & TDF_THRWAKEUP) { 454 thread_lock(td); 455 td->td_flags &= ~TDF_THRWAKEUP; 456 thread_unlock(td); 457 PROC_UNLOCK(td->td_proc); 458 return (0); 459 } 460 PROC_UNLOCK(td->td_proc); 461 if (error == EWOULDBLOCK) 462 error = ETIMEDOUT; 463 else if (error == ERESTART) { 464 if (hz != 0) 465 error = EINTR; 466 } 467 return (error); 468 } 469 470 int 471 thr_wake(struct thread *td, struct thr_wake_args *uap) 472 /* long id */ 473 { 474 struct proc *p; 475 struct thread *ttd; 476 477 if (uap->id == td->td_tid) { 478 td->td_pflags |= TDP_WAKEUP; 479 return (0); 480 } 481 482 p = td->td_proc; 483 PROC_LOCK(p); 484 ttd = thread_find(p, uap->id); 485 if (ttd == NULL) { 486 PROC_UNLOCK(p); 487 return (ESRCH); 488 } 489 thread_lock(ttd); 490 ttd->td_flags |= TDF_THRWAKEUP; 491 thread_unlock(ttd); 492 wakeup((void *)ttd); 493 PROC_UNLOCK(p); 494 return (0); 495 } 496 497 int 498 thr_set_name(struct thread *td, struct thr_set_name_args *uap) 499 { 500 struct proc *p = td->td_proc; 501 char name[MAXCOMLEN + 1]; 502 struct thread *ttd; 503 int error; 504 505 error = 0; 506 name[0] = '\0'; 507 if (uap->name != NULL) { 508 error = copyinstr(uap->name, name, sizeof(name), 509 NULL); 510 if (error) 511 return (error); 512 } 513 PROC_LOCK(p); 514 if (uap->id == td->td_tid) 515 ttd = td; 516 else 517 ttd = thread_find(p, uap->id); 518 if (ttd != NULL) 519 strcpy(ttd->td_name, name); 520 else 521 error = ESRCH; 522 PROC_UNLOCK(p); 523 return (error); 524 } 525