kern_thr.c revision 284665
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 284665 2015-06-21 06:28:26Z trasz $"); 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 if (racct_enable) { 285 PROC_LOCK(p); 286 racct_sub(p, RACCT_NTHR, 1); 287 PROC_UNLOCK(p); 288 } 289#endif 290 return (error); 291} 292 293int 294sys_thr_self(struct thread *td, struct thr_self_args *uap) 295 /* long *id */ 296{ 297 int error; 298 299 error = suword_lwpid(uap->id, (unsigned)td->td_tid); 300 if (error == -1) 301 return (EFAULT); 302 return (0); 303} 304 305int 306sys_thr_exit(struct thread *td, struct thr_exit_args *uap) 307 /* long *state */ 308{ 309 struct proc *p; 310 311 p = td->td_proc; 312 313 /* Signal userland that it can free the stack. */ 314 if ((void *)uap->state != NULL) { 315 suword_lwpid(uap->state, 1); 316 kern_umtx_wake(td, uap->state, INT_MAX, 0); 317 } 318 319 rw_wlock(&tidhash_lock); 320 321 PROC_LOCK(p); 322 323 if (p->p_numthreads != 1) { 324 racct_sub(p, RACCT_NTHR, 1); 325 LIST_REMOVE(td, td_hash); 326 rw_wunlock(&tidhash_lock); 327 tdsigcleanup(td); 328 umtx_thread_exit(td); 329 PROC_SLOCK(p); 330 thread_stopped(p); 331 thread_exit(); 332 /* NOTREACHED */ 333 } 334 335 /* 336 * Ignore attempts to shut down last thread in the proc. This 337 * will actually call _exit(2) in the usermode trampoline when 338 * it returns. 339 */ 340 PROC_UNLOCK(p); 341 rw_wunlock(&tidhash_lock); 342 return (0); 343} 344 345int 346sys_thr_kill(struct thread *td, struct thr_kill_args *uap) 347 /* long id, int sig */ 348{ 349 ksiginfo_t ksi; 350 struct thread *ttd; 351 struct proc *p; 352 int error; 353 354 p = td->td_proc; 355 ksiginfo_init(&ksi); 356 ksi.ksi_signo = uap->sig; 357 ksi.ksi_code = SI_LWP; 358 ksi.ksi_pid = p->p_pid; 359 ksi.ksi_uid = td->td_ucred->cr_ruid; 360 if (uap->id == -1) { 361 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 362 error = EINVAL; 363 } else { 364 error = ESRCH; 365 PROC_LOCK(p); 366 FOREACH_THREAD_IN_PROC(p, ttd) { 367 if (ttd != td) { 368 error = 0; 369 if (uap->sig == 0) 370 break; 371 tdksignal(ttd, uap->sig, &ksi); 372 } 373 } 374 PROC_UNLOCK(p); 375 } 376 } else { 377 error = 0; 378 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 379 if (ttd == NULL) 380 return (ESRCH); 381 if (uap->sig == 0) 382 ; 383 else if (!_SIG_VALID(uap->sig)) 384 error = EINVAL; 385 else 386 tdksignal(ttd, uap->sig, &ksi); 387 PROC_UNLOCK(ttd->td_proc); 388 } 389 return (error); 390} 391 392int 393sys_thr_kill2(struct thread *td, struct thr_kill2_args *uap) 394 /* pid_t pid, long id, int sig */ 395{ 396 ksiginfo_t ksi; 397 struct thread *ttd; 398 struct proc *p; 399 int error; 400 401 AUDIT_ARG_SIGNUM(uap->sig); 402 403 ksiginfo_init(&ksi); 404 ksi.ksi_signo = uap->sig; 405 ksi.ksi_code = SI_LWP; 406 ksi.ksi_pid = td->td_proc->p_pid; 407 ksi.ksi_uid = td->td_ucred->cr_ruid; 408 if (uap->id == -1) { 409 if ((p = pfind(uap->pid)) == NULL) 410 return (ESRCH); 411 AUDIT_ARG_PROCESS(p); 412 error = p_cansignal(td, p, uap->sig); 413 if (error) { 414 PROC_UNLOCK(p); 415 return (error); 416 } 417 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 418 error = EINVAL; 419 } else { 420 error = ESRCH; 421 FOREACH_THREAD_IN_PROC(p, ttd) { 422 if (ttd != td) { 423 error = 0; 424 if (uap->sig == 0) 425 break; 426 tdksignal(ttd, uap->sig, &ksi); 427 } 428 } 429 } 430 PROC_UNLOCK(p); 431 } else { 432 ttd = tdfind((lwpid_t)uap->id, uap->pid); 433 if (ttd == NULL) 434 return (ESRCH); 435 p = ttd->td_proc; 436 AUDIT_ARG_PROCESS(p); 437 error = p_cansignal(td, p, uap->sig); 438 if (uap->sig == 0) 439 ; 440 else if (!_SIG_VALID(uap->sig)) 441 error = EINVAL; 442 else 443 tdksignal(ttd, uap->sig, &ksi); 444 PROC_UNLOCK(p); 445 } 446 return (error); 447} 448 449int 450sys_thr_suspend(struct thread *td, struct thr_suspend_args *uap) 451 /* const struct timespec *timeout */ 452{ 453 struct timespec ts, *tsp; 454 int error; 455 456 tsp = NULL; 457 if (uap->timeout != NULL) { 458 error = umtx_copyin_timeout(uap->timeout, &ts); 459 if (error != 0) 460 return (error); 461 tsp = &ts; 462 } 463 464 return (kern_thr_suspend(td, tsp)); 465} 466 467int 468kern_thr_suspend(struct thread *td, struct timespec *tsp) 469{ 470 struct proc *p = td->td_proc; 471 struct timeval tv; 472 int error = 0; 473 int timo = 0; 474 475 if (td->td_pflags & TDP_WAKEUP) { 476 td->td_pflags &= ~TDP_WAKEUP; 477 return (0); 478 } 479 480 if (tsp != NULL) { 481 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) 482 error = EWOULDBLOCK; 483 else { 484 TIMESPEC_TO_TIMEVAL(&tv, tsp); 485 timo = tvtohz(&tv); 486 } 487 } 488 489 PROC_LOCK(p); 490 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0) 491 error = msleep((void *)td, &p->p_mtx, 492 PCATCH, "lthr", timo); 493 494 if (td->td_flags & TDF_THRWAKEUP) { 495 thread_lock(td); 496 td->td_flags &= ~TDF_THRWAKEUP; 497 thread_unlock(td); 498 PROC_UNLOCK(p); 499 return (0); 500 } 501 PROC_UNLOCK(p); 502 if (error == EWOULDBLOCK) 503 error = ETIMEDOUT; 504 else if (error == ERESTART) { 505 if (timo != 0) 506 error = EINTR; 507 } 508 return (error); 509} 510 511int 512sys_thr_wake(struct thread *td, struct thr_wake_args *uap) 513 /* long id */ 514{ 515 struct proc *p; 516 struct thread *ttd; 517 518 if (uap->id == td->td_tid) { 519 td->td_pflags |= TDP_WAKEUP; 520 return (0); 521 } 522 523 p = td->td_proc; 524 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 525 if (ttd == NULL) 526 return (ESRCH); 527 thread_lock(ttd); 528 ttd->td_flags |= TDF_THRWAKEUP; 529 thread_unlock(ttd); 530 wakeup((void *)ttd); 531 PROC_UNLOCK(p); 532 return (0); 533} 534 535int 536sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap) 537{ 538 struct proc *p; 539 char name[MAXCOMLEN + 1]; 540 struct thread *ttd; 541 int error; 542 543 error = 0; 544 name[0] = '\0'; 545 if (uap->name != NULL) { 546 error = copyinstr(uap->name, name, sizeof(name), 547 NULL); 548 if (error) 549 return (error); 550 } 551 p = td->td_proc; 552 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 553 if (ttd == NULL) 554 return (ESRCH); 555 strcpy(ttd->td_name, name); 556#ifdef KTR 557 sched_clear_tdname(ttd); 558#endif 559 PROC_UNLOCK(p); 560 return (error); 561} 562