kern_thr.c revision 284199
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: stable/10/sys/kern/kern_thr.c 284199 2015-06-10 02:04:02Z kib $"); 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 61static SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0, 62 "thread allocation"); 63 64static int max_threads_per_proc = 1500; 65SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_per_proc, CTLFLAG_RW, 66 &max_threads_per_proc, 0, "Limit on threads per proc"); 67 68static int max_threads_hits; 69SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_hits, CTLFLAG_RD, 70 &max_threads_hits, 0, "kern.threads.max_threads_per_proc hit count"); 71 72#ifdef COMPAT_FREEBSD32 73 74static inline int 75suword_lwpid(void *addr, lwpid_t lwpid) 76{ 77 int error; 78 79 if (SV_CURPROC_FLAG(SV_LP64)) 80 error = suword(addr, lwpid); 81 else 82 error = suword32(addr, lwpid); 83 return (error); 84} 85 86#else 87#define suword_lwpid suword 88#endif 89 90static int create_thread(struct thread *td, mcontext_t *ctx, 91 void (*start_func)(void *), void *arg, 92 char *stack_base, size_t stack_size, 93 char *tls_base, 94 long *child_tid, long *parent_tid, 95 int flags, struct rtprio *rtp); 96 97/* 98 * System call interface. 99 */ 100int 101sys_thr_create(struct thread *td, struct thr_create_args *uap) 102 /* ucontext_t *ctx, long *id, int flags */ 103{ 104 ucontext_t ctx; 105 int error; 106 107 if ((error = copyin(uap->ctx, &ctx, sizeof(ctx)))) 108 return (error); 109 110 error = create_thread(td, &ctx.uc_mcontext, NULL, NULL, 111 NULL, 0, NULL, uap->id, NULL, uap->flags, NULL); 112 return (error); 113} 114 115int 116sys_thr_new(struct thread *td, struct thr_new_args *uap) 117 /* struct thr_param * */ 118{ 119 struct thr_param param; 120 int error; 121 122 if (uap->param_size < 0 || uap->param_size > sizeof(param)) 123 return (EINVAL); 124 bzero(¶m, sizeof(param)); 125 if ((error = copyin(uap->param, ¶m, uap->param_size))) 126 return (error); 127 return (kern_thr_new(td, ¶m)); 128} 129 130int 131kern_thr_new(struct thread *td, struct thr_param *param) 132{ 133 struct rtprio rtp, *rtpp; 134 int error; 135 136 rtpp = NULL; 137 if (param->rtp != 0) { 138 error = copyin(param->rtp, &rtp, sizeof(struct rtprio)); 139 if (error) 140 return (error); 141 rtpp = &rtp; 142 } 143 error = create_thread(td, NULL, param->start_func, param->arg, 144 param->stack_base, param->stack_size, param->tls_base, 145 param->child_tid, param->parent_tid, param->flags, 146 rtpp); 147 return (error); 148} 149 150static int 151create_thread(struct thread *td, mcontext_t *ctx, 152 void (*start_func)(void *), void *arg, 153 char *stack_base, size_t stack_size, 154 char *tls_base, 155 long *child_tid, long *parent_tid, 156 int flags, struct rtprio *rtp) 157{ 158 stack_t stack; 159 struct thread *newtd; 160 struct proc *p; 161 int error; 162 163 p = td->td_proc; 164 165 /* Have race condition but it is cheap. */ 166 if (p->p_numthreads >= max_threads_per_proc) { 167 ++max_threads_hits; 168 return (EPROCLIM); 169 } 170 171 if (rtp != NULL) { 172 switch(rtp->type) { 173 case RTP_PRIO_REALTIME: 174 case RTP_PRIO_FIFO: 175 /* Only root can set scheduler policy */ 176 if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0) 177 return (EPERM); 178 if (rtp->prio > RTP_PRIO_MAX) 179 return (EINVAL); 180 break; 181 case RTP_PRIO_NORMAL: 182 rtp->prio = 0; 183 break; 184 default: 185 return (EINVAL); 186 } 187 } 188 189#ifdef RACCT 190 PROC_LOCK(td->td_proc); 191 error = racct_add(p, RACCT_NTHR, 1); 192 PROC_UNLOCK(td->td_proc); 193 if (error != 0) 194 return (EPROCLIM); 195#endif 196 197 /* Initialize our td */ 198 newtd = thread_alloc(0); 199 if (newtd == NULL) { 200 error = ENOMEM; 201 goto fail; 202 } 203 204 cpu_set_upcall(newtd, td); 205 206 /* 207 * Try the copyout as soon as we allocate the td so we don't 208 * have to tear things down in a failure case below. 209 * Here we copy out tid to two places, one for child and one 210 * for parent, because pthread can create a detached thread, 211 * if parent wants to safely access child tid, it has to provide 212 * its storage, because child thread may exit quickly and 213 * memory is freed before parent thread can access it. 214 */ 215 if ((child_tid != NULL && 216 suword_lwpid(child_tid, newtd->td_tid)) || 217 (parent_tid != NULL && 218 suword_lwpid(parent_tid, newtd->td_tid))) { 219 thread_free(newtd); 220 error = EFAULT; 221 goto fail; 222 } 223 224 bzero(&newtd->td_startzero, 225 __rangeof(struct thread, td_startzero, td_endzero)); 226 newtd->td_su = NULL; 227 bcopy(&td->td_startcopy, &newtd->td_startcopy, 228 __rangeof(struct thread, td_startcopy, td_endcopy)); 229 newtd->td_proc = td->td_proc; 230 newtd->td_ucred = crhold(td->td_ucred); 231 232 if (ctx != NULL) { /* old way to set user context */ 233 error = set_mcontext(newtd, ctx); 234 if (error != 0) { 235 thread_free(newtd); 236 crfree(td->td_ucred); 237 goto fail; 238 } 239 } else { 240 /* Set up our machine context. */ 241 stack.ss_sp = stack_base; 242 stack.ss_size = stack_size; 243 /* Set upcall address to user thread entry function. */ 244 cpu_set_upcall_kse(newtd, start_func, arg, &stack); 245 /* Setup user TLS address and TLS pointer register. */ 246 error = cpu_set_user_tls(newtd, tls_base); 247 if (error != 0) { 248 thread_free(newtd); 249 crfree(td->td_ucred); 250 goto fail; 251 } 252 } 253 254 PROC_LOCK(td->td_proc); 255 td->td_proc->p_flag |= P_HADTHREADS; 256 thread_link(newtd, p); 257 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); 258 thread_lock(td); 259 /* let the scheduler know about these things. */ 260 sched_fork_thread(td, newtd); 261 thread_unlock(td); 262 if (P_SHOULDSTOP(p)) 263 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; 264 PROC_UNLOCK(p); 265 266 tidhash_add(newtd); 267 268 thread_lock(newtd); 269 if (rtp != NULL) { 270 if (!(td->td_pri_class == PRI_TIMESHARE && 271 rtp->type == RTP_PRIO_NORMAL)) { 272 rtp_to_pri(rtp, newtd); 273 sched_prio(newtd, newtd->td_user_pri); 274 } /* ignore timesharing class */ 275 } 276 TD_SET_CAN_RUN(newtd); 277 sched_add(newtd, SRQ_BORING); 278 thread_unlock(newtd); 279 280 return (0); 281 282fail: 283#ifdef RACCT 284 PROC_LOCK(p); 285 racct_sub(p, RACCT_NTHR, 1); 286 PROC_UNLOCK(p); 287#endif 288 return (error); 289} 290 291int 292sys_thr_self(struct thread *td, struct thr_self_args *uap) 293 /* long *id */ 294{ 295 int error; 296 297 error = suword_lwpid(uap->id, (unsigned)td->td_tid); 298 if (error == -1) 299 return (EFAULT); 300 return (0); 301} 302 303int 304sys_thr_exit(struct thread *td, struct thr_exit_args *uap) 305 /* long *state */ 306{ 307 struct proc *p; 308 309 p = td->td_proc; 310 311 /* Signal userland that it can free the stack. */ 312 if ((void *)uap->state != NULL) { 313 suword_lwpid(uap->state, 1); 314 kern_umtx_wake(td, uap->state, INT_MAX, 0); 315 } 316 317 rw_wlock(&tidhash_lock); 318 319 PROC_LOCK(p); 320 321 if (p->p_numthreads != 1) { 322 racct_sub(p, RACCT_NTHR, 1); 323 LIST_REMOVE(td, td_hash); 324 rw_wunlock(&tidhash_lock); 325 tdsigcleanup(td); 326 umtx_thread_exit(td); 327 PROC_SLOCK(p); 328 thread_stopped(p); 329 thread_exit(); 330 /* NOTREACHED */ 331 } 332 333 /* 334 * Ignore attempts to shut down last thread in the proc. This 335 * will actually call _exit(2) in the usermode trampoline when 336 * it returns. 337 */ 338 PROC_UNLOCK(p); 339 rw_wunlock(&tidhash_lock); 340 return (0); 341} 342 343int 344sys_thr_kill(struct thread *td, struct thr_kill_args *uap) 345 /* long id, int sig */ 346{ 347 ksiginfo_t ksi; 348 struct thread *ttd; 349 struct proc *p; 350 int error; 351 352 p = td->td_proc; 353 ksiginfo_init(&ksi); 354 ksi.ksi_signo = uap->sig; 355 ksi.ksi_code = SI_LWP; 356 ksi.ksi_pid = p->p_pid; 357 ksi.ksi_uid = td->td_ucred->cr_ruid; 358 if (uap->id == -1) { 359 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 360 error = EINVAL; 361 } else { 362 error = ESRCH; 363 PROC_LOCK(p); 364 FOREACH_THREAD_IN_PROC(p, ttd) { 365 if (ttd != td) { 366 error = 0; 367 if (uap->sig == 0) 368 break; 369 tdksignal(ttd, uap->sig, &ksi); 370 } 371 } 372 PROC_UNLOCK(p); 373 } 374 } else { 375 error = 0; 376 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 377 if (ttd == NULL) 378 return (ESRCH); 379 if (uap->sig == 0) 380 ; 381 else if (!_SIG_VALID(uap->sig)) 382 error = EINVAL; 383 else 384 tdksignal(ttd, uap->sig, &ksi); 385 PROC_UNLOCK(ttd->td_proc); 386 } 387 return (error); 388} 389 390int 391sys_thr_kill2(struct thread *td, struct thr_kill2_args *uap) 392 /* pid_t pid, long id, int sig */ 393{ 394 ksiginfo_t ksi; 395 struct thread *ttd; 396 struct proc *p; 397 int error; 398 399 AUDIT_ARG_SIGNUM(uap->sig); 400 401 ksiginfo_init(&ksi); 402 ksi.ksi_signo = uap->sig; 403 ksi.ksi_code = SI_LWP; 404 ksi.ksi_pid = td->td_proc->p_pid; 405 ksi.ksi_uid = td->td_ucred->cr_ruid; 406 if (uap->id == -1) { 407 if ((p = pfind(uap->pid)) == NULL) 408 return (ESRCH); 409 AUDIT_ARG_PROCESS(p); 410 error = p_cansignal(td, p, uap->sig); 411 if (error) { 412 PROC_UNLOCK(p); 413 return (error); 414 } 415 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 416 error = EINVAL; 417 } else { 418 error = ESRCH; 419 FOREACH_THREAD_IN_PROC(p, ttd) { 420 if (ttd != td) { 421 error = 0; 422 if (uap->sig == 0) 423 break; 424 tdksignal(ttd, uap->sig, &ksi); 425 } 426 } 427 } 428 PROC_UNLOCK(p); 429 } else { 430 ttd = tdfind((lwpid_t)uap->id, uap->pid); 431 if (ttd == NULL) 432 return (ESRCH); 433 p = ttd->td_proc; 434 AUDIT_ARG_PROCESS(p); 435 error = p_cansignal(td, p, uap->sig); 436 if (uap->sig == 0) 437 ; 438 else if (!_SIG_VALID(uap->sig)) 439 error = EINVAL; 440 else 441 tdksignal(ttd, uap->sig, &ksi); 442 PROC_UNLOCK(p); 443 } 444 return (error); 445} 446 447int 448sys_thr_suspend(struct thread *td, struct thr_suspend_args *uap) 449 /* const struct timespec *timeout */ 450{ 451 struct timespec ts, *tsp; 452 int error; 453 454 tsp = NULL; 455 if (uap->timeout != NULL) { 456 error = umtx_copyin_timeout(uap->timeout, &ts); 457 if (error != 0) 458 return (error); 459 tsp = &ts; 460 } 461 462 return (kern_thr_suspend(td, tsp)); 463} 464 465int 466kern_thr_suspend(struct thread *td, struct timespec *tsp) 467{ 468 struct proc *p = td->td_proc; 469 struct timeval tv; 470 int error = 0; 471 int timo = 0; 472 473 if (td->td_pflags & TDP_WAKEUP) { 474 td->td_pflags &= ~TDP_WAKEUP; 475 return (0); 476 } 477 478 if (tsp != NULL) { 479 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) 480 error = EWOULDBLOCK; 481 else { 482 TIMESPEC_TO_TIMEVAL(&tv, tsp); 483 timo = tvtohz(&tv); 484 } 485 } 486 487 PROC_LOCK(p); 488 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0) 489 error = msleep((void *)td, &p->p_mtx, 490 PCATCH, "lthr", timo); 491 492 if (td->td_flags & TDF_THRWAKEUP) { 493 thread_lock(td); 494 td->td_flags &= ~TDF_THRWAKEUP; 495 thread_unlock(td); 496 PROC_UNLOCK(p); 497 return (0); 498 } 499 PROC_UNLOCK(p); 500 if (error == EWOULDBLOCK) 501 error = ETIMEDOUT; 502 else if (error == ERESTART) { 503 if (timo != 0) 504 error = EINTR; 505 } 506 return (error); 507} 508 509int 510sys_thr_wake(struct thread *td, struct thr_wake_args *uap) 511 /* long id */ 512{ 513 struct proc *p; 514 struct thread *ttd; 515 516 if (uap->id == td->td_tid) { 517 td->td_pflags |= TDP_WAKEUP; 518 return (0); 519 } 520 521 p = td->td_proc; 522 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 523 if (ttd == NULL) 524 return (ESRCH); 525 thread_lock(ttd); 526 ttd->td_flags |= TDF_THRWAKEUP; 527 thread_unlock(ttd); 528 wakeup((void *)ttd); 529 PROC_UNLOCK(p); 530 return (0); 531} 532 533int 534sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap) 535{ 536 struct proc *p; 537 char name[MAXCOMLEN + 1]; 538 struct thread *ttd; 539 int error; 540 541 error = 0; 542 name[0] = '\0'; 543 if (uap->name != NULL) { 544 error = copyinstr(uap->name, name, sizeof(name), 545 NULL); 546 if (error) 547 return (error); 548 } 549 p = td->td_proc; 550 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 551 if (ttd == NULL) 552 return (ESRCH); 553 strcpy(ttd->td_name, name); 554#ifdef KTR 555 sched_clear_tdname(ttd); 556#endif 557 PROC_UNLOCK(p); 558 return (error); 559} 560