kern_thr.c revision 289780
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 289780 2015-10-23 01:27:44Z jhb $"); 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 90/* 91 * System call interface. 92 */ 93 94struct thr_create_initthr_args { 95 ucontext_t ctx; 96 long *tid; 97}; 98 99static int 100thr_create_initthr(struct thread *td, void *thunk) 101{ 102 struct thr_create_initthr_args *args; 103 104 /* Copy out the child tid. */ 105 args = thunk; 106 if (args->tid != NULL && suword_lwpid(args->tid, td->td_tid)) 107 return (EFAULT); 108 109 return (set_mcontext(td, &args->ctx.uc_mcontext)); 110} 111 112int 113sys_thr_create(struct thread *td, struct thr_create_args *uap) 114 /* ucontext_t *ctx, long *id, int flags */ 115{ 116 struct thr_create_initthr_args args; 117 int error; 118 119 if ((error = copyin(uap->ctx, &args.ctx, sizeof(args.ctx)))) 120 return (error); 121 args.tid = uap->id; 122 return (thread_create(td, NULL, thr_create_initthr, &args)); 123} 124 125int 126sys_thr_new(struct thread *td, struct thr_new_args *uap) 127 /* struct thr_param * */ 128{ 129 struct thr_param param; 130 int error; 131 132 if (uap->param_size < 0 || uap->param_size > sizeof(param)) 133 return (EINVAL); 134 bzero(¶m, sizeof(param)); 135 if ((error = copyin(uap->param, ¶m, uap->param_size))) 136 return (error); 137 return (kern_thr_new(td, ¶m)); 138} 139 140static int 141thr_new_initthr(struct thread *td, void *thunk) 142{ 143 stack_t stack; 144 struct thr_param *param; 145 146 /* 147 * Here we copy out tid to two places, one for child and one 148 * for parent, because pthread can create a detached thread, 149 * if parent wants to safely access child tid, it has to provide 150 * its storage, because child thread may exit quickly and 151 * memory is freed before parent thread can access it. 152 */ 153 param = thunk; 154 if ((param->child_tid != NULL && 155 suword_lwpid(param->child_tid, td->td_tid)) || 156 (param->parent_tid != NULL && 157 suword_lwpid(param->parent_tid, td->td_tid))) 158 return (EFAULT); 159 160 /* Set up our machine context. */ 161 stack.ss_sp = param->stack_base; 162 stack.ss_size = param->stack_size; 163 /* Set upcall address to user thread entry function. */ 164 cpu_set_upcall_kse(td, param->start_func, param->arg, &stack); 165 /* Setup user TLS address and TLS pointer register. */ 166 return (cpu_set_user_tls(td, param->tls_base)); 167} 168 169int 170kern_thr_new(struct thread *td, struct thr_param *param) 171{ 172 struct rtprio rtp, *rtpp; 173 int error; 174 175 rtpp = NULL; 176 if (param->rtp != 0) { 177 error = copyin(param->rtp, &rtp, sizeof(struct rtprio)); 178 if (error) 179 return (error); 180 rtpp = &rtp; 181 } 182 return (thread_create(td, rtpp, thr_new_initthr, param)); 183} 184 185int 186thread_create(struct thread *td, struct rtprio *rtp, 187 int (*initialize_thread)(struct thread *, void *), void *thunk) 188{ 189 struct thread *newtd; 190 struct proc *p; 191 int error; 192 193 p = td->td_proc; 194 195 /* Have race condition but it is cheap. */ 196 if (p->p_numthreads >= max_threads_per_proc) { 197 ++max_threads_hits; 198 return (EPROCLIM); 199 } 200 201 if (rtp != NULL) { 202 switch(rtp->type) { 203 case RTP_PRIO_REALTIME: 204 case RTP_PRIO_FIFO: 205 /* Only root can set scheduler policy */ 206 if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0) 207 return (EPERM); 208 if (rtp->prio > RTP_PRIO_MAX) 209 return (EINVAL); 210 break; 211 case RTP_PRIO_NORMAL: 212 rtp->prio = 0; 213 break; 214 default: 215 return (EINVAL); 216 } 217 } 218 219#ifdef RACCT 220 PROC_LOCK(td->td_proc); 221 error = racct_add(p, RACCT_NTHR, 1); 222 PROC_UNLOCK(td->td_proc); 223 if (error != 0) 224 return (EPROCLIM); 225#endif 226 227 /* Initialize our td */ 228 newtd = thread_alloc(0); 229 if (newtd == NULL) { 230 error = ENOMEM; 231 goto fail; 232 } 233 234 cpu_set_upcall(newtd, td); 235 236 bzero(&newtd->td_startzero, 237 __rangeof(struct thread, td_startzero, td_endzero)); 238 newtd->td_su = NULL; 239 bcopy(&td->td_startcopy, &newtd->td_startcopy, 240 __rangeof(struct thread, td_startcopy, td_endcopy)); 241 newtd->td_proc = td->td_proc; 242 newtd->td_ucred = crhold(td->td_ucred); 243 newtd->td_dbg_sc_code = td->td_dbg_sc_code; 244 newtd->td_dbg_sc_narg = td->td_dbg_sc_narg; 245 246 error = initialize_thread(newtd, thunk); 247 if (error != 0) { 248 thread_free(newtd); 249 crfree(td->td_ucred); 250 goto fail; 251 } 252 253 PROC_LOCK(td->td_proc); 254 td->td_proc->p_flag |= P_HADTHREADS; 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 281fail: 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 292int 293sys_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 304int 305sys_thr_exit(struct thread *td, struct thr_exit_args *uap) 306 /* long *state */ 307{ 308 struct proc *p; 309 310 p = td->td_proc; 311 312 /* Signal userland that it can free the stack. */ 313 if ((void *)uap->state != NULL) { 314 suword_lwpid(uap->state, 1); 315 kern_umtx_wake(td, uap->state, INT_MAX, 0); 316 } 317 318 rw_wlock(&tidhash_lock); 319 320 PROC_LOCK(p); 321 322 if (p->p_numthreads != 1) { 323 racct_sub(p, RACCT_NTHR, 1); 324 LIST_REMOVE(td, td_hash); 325 rw_wunlock(&tidhash_lock); 326 tdsigcleanup(td); 327 umtx_thread_exit(td); 328 PROC_SLOCK(p); 329 thread_stopped(p); 330 thread_exit(); 331 /* NOTREACHED */ 332 } 333 334 /* 335 * Ignore attempts to shut down last thread in the proc. This 336 * will actually call _exit(2) in the usermode trampoline when 337 * it returns. 338 */ 339 PROC_UNLOCK(p); 340 rw_wunlock(&tidhash_lock); 341 return (0); 342} 343 344int 345sys_thr_kill(struct thread *td, struct thr_kill_args *uap) 346 /* long id, int sig */ 347{ 348 ksiginfo_t ksi; 349 struct thread *ttd; 350 struct proc *p; 351 int error; 352 353 p = td->td_proc; 354 ksiginfo_init(&ksi); 355 ksi.ksi_signo = uap->sig; 356 ksi.ksi_code = SI_LWP; 357 ksi.ksi_pid = p->p_pid; 358 ksi.ksi_uid = td->td_ucred->cr_ruid; 359 if (uap->id == -1) { 360 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 361 error = EINVAL; 362 } else { 363 error = ESRCH; 364 PROC_LOCK(p); 365 FOREACH_THREAD_IN_PROC(p, ttd) { 366 if (ttd != td) { 367 error = 0; 368 if (uap->sig == 0) 369 break; 370 tdksignal(ttd, uap->sig, &ksi); 371 } 372 } 373 PROC_UNLOCK(p); 374 } 375 } else { 376 error = 0; 377 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 378 if (ttd == NULL) 379 return (ESRCH); 380 if (uap->sig == 0) 381 ; 382 else if (!_SIG_VALID(uap->sig)) 383 error = EINVAL; 384 else 385 tdksignal(ttd, uap->sig, &ksi); 386 PROC_UNLOCK(ttd->td_proc); 387 } 388 return (error); 389} 390 391int 392sys_thr_kill2(struct thread *td, struct thr_kill2_args *uap) 393 /* pid_t pid, long id, int sig */ 394{ 395 ksiginfo_t ksi; 396 struct thread *ttd; 397 struct proc *p; 398 int error; 399 400 AUDIT_ARG_SIGNUM(uap->sig); 401 402 ksiginfo_init(&ksi); 403 ksi.ksi_signo = uap->sig; 404 ksi.ksi_code = SI_LWP; 405 ksi.ksi_pid = td->td_proc->p_pid; 406 ksi.ksi_uid = td->td_ucred->cr_ruid; 407 if (uap->id == -1) { 408 if ((p = pfind(uap->pid)) == NULL) 409 return (ESRCH); 410 AUDIT_ARG_PROCESS(p); 411 error = p_cansignal(td, p, uap->sig); 412 if (error) { 413 PROC_UNLOCK(p); 414 return (error); 415 } 416 if (uap->sig != 0 && !_SIG_VALID(uap->sig)) { 417 error = EINVAL; 418 } else { 419 error = ESRCH; 420 FOREACH_THREAD_IN_PROC(p, ttd) { 421 if (ttd != td) { 422 error = 0; 423 if (uap->sig == 0) 424 break; 425 tdksignal(ttd, uap->sig, &ksi); 426 } 427 } 428 } 429 PROC_UNLOCK(p); 430 } else { 431 ttd = tdfind((lwpid_t)uap->id, uap->pid); 432 if (ttd == NULL) 433 return (ESRCH); 434 p = ttd->td_proc; 435 AUDIT_ARG_PROCESS(p); 436 error = p_cansignal(td, p, uap->sig); 437 if (uap->sig == 0) 438 ; 439 else if (!_SIG_VALID(uap->sig)) 440 error = EINVAL; 441 else 442 tdksignal(ttd, uap->sig, &ksi); 443 PROC_UNLOCK(p); 444 } 445 return (error); 446} 447 448int 449sys_thr_suspend(struct thread *td, struct thr_suspend_args *uap) 450 /* const struct timespec *timeout */ 451{ 452 struct timespec ts, *tsp; 453 int error; 454 455 tsp = NULL; 456 if (uap->timeout != NULL) { 457 error = umtx_copyin_timeout(uap->timeout, &ts); 458 if (error != 0) 459 return (error); 460 tsp = &ts; 461 } 462 463 return (kern_thr_suspend(td, tsp)); 464} 465 466int 467kern_thr_suspend(struct thread *td, struct timespec *tsp) 468{ 469 struct proc *p = td->td_proc; 470 struct timeval tv; 471 int error = 0; 472 int timo = 0; 473 474 if (td->td_pflags & TDP_WAKEUP) { 475 td->td_pflags &= ~TDP_WAKEUP; 476 return (0); 477 } 478 479 if (tsp != NULL) { 480 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) 481 error = EWOULDBLOCK; 482 else { 483 TIMESPEC_TO_TIMEVAL(&tv, tsp); 484 timo = tvtohz(&tv); 485 } 486 } 487 488 PROC_LOCK(p); 489 if (error == 0 && (td->td_flags & TDF_THRWAKEUP) == 0) 490 error = msleep((void *)td, &p->p_mtx, 491 PCATCH, "lthr", timo); 492 493 if (td->td_flags & TDF_THRWAKEUP) { 494 thread_lock(td); 495 td->td_flags &= ~TDF_THRWAKEUP; 496 thread_unlock(td); 497 PROC_UNLOCK(p); 498 return (0); 499 } 500 PROC_UNLOCK(p); 501 if (error == EWOULDBLOCK) 502 error = ETIMEDOUT; 503 else if (error == ERESTART) { 504 if (timo != 0) 505 error = EINTR; 506 } 507 return (error); 508} 509 510int 511sys_thr_wake(struct thread *td, struct thr_wake_args *uap) 512 /* long id */ 513{ 514 struct proc *p; 515 struct thread *ttd; 516 517 if (uap->id == td->td_tid) { 518 td->td_pflags |= TDP_WAKEUP; 519 return (0); 520 } 521 522 p = td->td_proc; 523 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 524 if (ttd == NULL) 525 return (ESRCH); 526 thread_lock(ttd); 527 ttd->td_flags |= TDF_THRWAKEUP; 528 thread_unlock(ttd); 529 wakeup((void *)ttd); 530 PROC_UNLOCK(p); 531 return (0); 532} 533 534int 535sys_thr_set_name(struct thread *td, struct thr_set_name_args *uap) 536{ 537 struct proc *p; 538 char name[MAXCOMLEN + 1]; 539 struct thread *ttd; 540 int error; 541 542 error = 0; 543 name[0] = '\0'; 544 if (uap->name != NULL) { 545 error = copyinstr(uap->name, name, sizeof(name), 546 NULL); 547 if (error) 548 return (error); 549 } 550 p = td->td_proc; 551 ttd = tdfind((lwpid_t)uap->id, p->p_pid); 552 if (ttd == NULL) 553 return (ESRCH); 554 strcpy(ttd->td_name, name); 555#ifdef KTR 556 sched_clear_tdname(ttd); 557#endif 558 PROC_UNLOCK(p); 559 return (error); 560} 561