1 /*- 2 * Copyright (C) 2001 Julian Elischer <julian@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(s), this list of conditions and the following disclaimer as 10 * the first lines of this file unmodified other than the possible 11 * addition of one or more copyright notices. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice(s), this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY 17 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 18 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 19 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY 20 * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 21 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 22 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 23 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH 26 * DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include <sys/param.h> 33 #include <sys/systm.h> 34 #include <sys/kernel.h> 35 #include <sys/lock.h> 36 #include <sys/mutex.h> 37 #include <sys/proc.h> 38 #include <sys/smp.h> 39 #include <sys/sysctl.h> 40 #include <sys/sched.h> 41 #include <sys/sleepqueue.h> 42 #include <sys/turnstile.h> 43 #include <sys/ktr.h> 44 #include <sys/umtx.h> 45 46 #include <security/audit/audit.h> 47 48 #include <vm/vm.h> 49 #include <vm/vm_extern.h> 50 #include <vm/uma.h> 51 52 /* 53 * KSEGRP related storage. 54 */ 55 static uma_zone_t ksegrp_zone; 56 static uma_zone_t thread_zone; 57 58 /* DEBUG ONLY */ 59 SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0, "thread allocation"); 60 static int thread_debug = 0; 61 SYSCTL_INT(_kern_threads, OID_AUTO, debug, CTLFLAG_RW, 62 &thread_debug, 0, "thread debug"); 63 64 int max_threads_per_proc = 1500; 65 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_per_proc, CTLFLAG_RW, 66 &max_threads_per_proc, 0, "Limit on threads per proc"); 67 68 int max_groups_per_proc = 1500; 69 SYSCTL_INT(_kern_threads, OID_AUTO, max_groups_per_proc, CTLFLAG_RW, 70 &max_groups_per_proc, 0, "Limit on thread groups per proc"); 71 72 int max_threads_hits; 73 SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_hits, CTLFLAG_RD, 74 &max_threads_hits, 0, ""); 75 76 int virtual_cpu; 77 78 TAILQ_HEAD(, thread) zombie_threads = TAILQ_HEAD_INITIALIZER(zombie_threads); 79 TAILQ_HEAD(, ksegrp) zombie_ksegrps = TAILQ_HEAD_INITIALIZER(zombie_ksegrps); 80 struct mtx kse_zombie_lock; 81 MTX_SYSINIT(kse_zombie_lock, &kse_zombie_lock, "kse zombie lock", MTX_SPIN); 82 83 static int 84 sysctl_kse_virtual_cpu(SYSCTL_HANDLER_ARGS) 85 { 86 int error, new_val; 87 int def_val; 88 89 def_val = mp_ncpus; 90 if (virtual_cpu == 0) 91 new_val = def_val; 92 else 93 new_val = virtual_cpu; 94 error = sysctl_handle_int(oidp, &new_val, 0, req); 95 if (error != 0 || req->newptr == NULL) 96 return (error); 97 if (new_val < 0) 98 return (EINVAL); 99 virtual_cpu = new_val; 100 return (0); 101 } 102 103 /* DEBUG ONLY */ 104 SYSCTL_PROC(_kern_threads, OID_AUTO, virtual_cpu, CTLTYPE_INT|CTLFLAG_RW, 105 0, sizeof(virtual_cpu), sysctl_kse_virtual_cpu, "I", 106 "debug virtual cpus"); 107 108 struct mtx tid_lock; 109 static struct unrhdr *tid_unrhdr; 110 111 /* 112 * Prepare a thread for use. 113 */ 114 static int 115 thread_ctor(void *mem, int size, void *arg, int flags) 116 { 117 struct thread *td; 118 119 td = (struct thread *)mem; 120 td->td_state = TDS_INACTIVE; 121 td->td_oncpu = NOCPU; 122 123 td->td_tid = alloc_unr(tid_unrhdr); 124 125 /* 126 * Note that td_critnest begins life as 1 because the thread is not 127 * running and is thereby implicitly waiting to be on the receiving 128 * end of a context switch. A context switch must occur inside a 129 * critical section, and in fact, includes hand-off of the sched_lock. 130 * After a context switch to a newly created thread, it will release 131 * sched_lock for the first time, and its td_critnest will hit 0 for 132 * the first time. This happens on the far end of a context switch, 133 * and when it context switches away from itself, it will in fact go 134 * back into a critical section, and hand off the sched lock to the 135 * next thread. 136 */ 137 td->td_critnest = 1; 138 139 #ifdef AUDIT 140 audit_thread_alloc(td); 141 #endif 142 return (0); 143 } 144 145 /* 146 * Reclaim a thread after use. 147 */ 148 static void 149 thread_dtor(void *mem, int size, void *arg) 150 { 151 struct thread *td; 152 153 td = (struct thread *)mem; 154 155 #ifdef INVARIANTS 156 /* Verify that this thread is in a safe state to free. */ 157 switch (td->td_state) { 158 case TDS_INHIBITED: 159 case TDS_RUNNING: 160 case TDS_CAN_RUN: 161 case TDS_RUNQ: 162 /* 163 * We must never unlink a thread that is in one of 164 * these states, because it is currently active. 165 */ 166 panic("bad state for thread unlinking"); 167 /* NOTREACHED */ 168 case TDS_INACTIVE: 169 break; 170 default: 171 panic("bad thread state"); 172 /* NOTREACHED */ 173 } 174 #endif 175 #ifdef AUDIT 176 audit_thread_free(td); 177 #endif 178 free_unr(tid_unrhdr, td->td_tid); 179 sched_newthread(td); 180 } 181 182 /* 183 * Initialize type-stable parts of a thread (when newly created). 184 */ 185 static int 186 thread_init(void *mem, int size, int flags) 187 { 188 struct thread *td; 189 190 td = (struct thread *)mem; 191 192 vm_thread_new(td, 0); 193 cpu_thread_setup(td); 194 td->td_sleepqueue = sleepq_alloc(); 195 td->td_turnstile = turnstile_alloc(); 196 td->td_umtxq = umtxq_alloc(); 197 td->td_sched = (struct td_sched *)&td[1]; 198 sched_newthread(td); 199 return (0); 200 } 201 202 /* 203 * Tear down type-stable parts of a thread (just before being discarded). 204 */ 205 static void 206 thread_fini(void *mem, int size) 207 { 208 struct thread *td; 209 210 td = (struct thread *)mem; 211 turnstile_free(td->td_turnstile); 212 sleepq_free(td->td_sleepqueue); 213 umtxq_free(td->td_umtxq); 214 vm_thread_dispose(td); 215 } 216 217 /* 218 * Initialize type-stable parts of a ksegrp (when newly created). 219 */ 220 static int 221 ksegrp_ctor(void *mem, int size, void *arg, int flags) 222 { 223 struct ksegrp *kg; 224 225 kg = (struct ksegrp *)mem; 226 bzero(mem, size); 227 kg->kg_sched = (struct kg_sched *)&kg[1]; 228 return (0); 229 } 230 231 void 232 ksegrp_link(struct ksegrp *kg, struct proc *p) 233 { 234 235 TAILQ_INIT(&kg->kg_threads); 236 TAILQ_INIT(&kg->kg_runq); /* links with td_runq */ 237 TAILQ_INIT(&kg->kg_upcalls); /* all upcall structure in ksegrp */ 238 kg->kg_proc = p; 239 /* 240 * the following counters are in the -zero- section 241 * and may not need clearing 242 */ 243 kg->kg_numthreads = 0; 244 kg->kg_numupcalls = 0; 245 /* link it in now that it's consistent */ 246 p->p_numksegrps++; 247 TAILQ_INSERT_HEAD(&p->p_ksegrps, kg, kg_ksegrp); 248 } 249 250 /* 251 * Called from: 252 * thread-exit() 253 */ 254 void 255 ksegrp_unlink(struct ksegrp *kg) 256 { 257 struct proc *p; 258 259 mtx_assert(&sched_lock, MA_OWNED); 260 KASSERT((kg->kg_numthreads == 0), ("ksegrp_unlink: residual threads")); 261 KASSERT((kg->kg_numupcalls == 0), ("ksegrp_unlink: residual upcalls")); 262 263 p = kg->kg_proc; 264 TAILQ_REMOVE(&p->p_ksegrps, kg, kg_ksegrp); 265 p->p_numksegrps--; 266 /* 267 * Aggregate stats from the KSE 268 */ 269 if (p->p_procscopegrp == kg) 270 p->p_procscopegrp = NULL; 271 } 272 273 /* 274 * For a newly created process, 275 * link up all the structures and its initial threads etc. 276 * called from: 277 * {arch}/{arch}/machdep.c ia64_init(), init386() etc. 278 * proc_dtor() (should go away) 279 * proc_init() 280 */ 281 void 282 proc_linkup(struct proc *p, struct ksegrp *kg, struct thread *td) 283 { 284 285 TAILQ_INIT(&p->p_ksegrps); /* all ksegrps in proc */ 286 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 287 TAILQ_INIT(&p->p_suspended); /* Threads suspended */ 288 sigqueue_init(&p->p_sigqueue, p); 289 p->p_ksi = ksiginfo_alloc(1); 290 if (p->p_ksi != NULL) { 291 /* XXX p_ksi may be null if ksiginfo zone is not ready */ 292 p->p_ksi->ksi_flags = KSI_EXT | KSI_INS; 293 } 294 LIST_INIT(&p->p_mqnotifier); 295 p->p_numksegrps = 0; 296 p->p_numthreads = 0; 297 298 ksegrp_link(kg, p); 299 thread_link(td, kg); 300 } 301 302 /* 303 * Initialize global thread allocation resources. 304 */ 305 void 306 threadinit(void) 307 { 308 309 mtx_init(&tid_lock, "TID lock", NULL, MTX_DEF); 310 tid_unrhdr = new_unrhdr(PID_MAX + 1, INT_MAX, &tid_lock); 311 312 thread_zone = uma_zcreate("THREAD", sched_sizeof_thread(), 313 thread_ctor, thread_dtor, thread_init, thread_fini, 314 UMA_ALIGN_CACHE, 0); 315 ksegrp_zone = uma_zcreate("KSEGRP", sched_sizeof_ksegrp(), 316 ksegrp_ctor, NULL, NULL, NULL, 317 UMA_ALIGN_CACHE, 0); 318 kseinit(); /* set up kse specific stuff e.g. upcall zone*/ 319 } 320 321 /* 322 * Stash an embarasingly extra thread into the zombie thread queue. 323 */ 324 void 325 thread_stash(struct thread *td) 326 { 327 mtx_lock_spin(&kse_zombie_lock); 328 TAILQ_INSERT_HEAD(&zombie_threads, td, td_runq); 329 mtx_unlock_spin(&kse_zombie_lock); 330 } 331 332 /* 333 * Stash an embarasingly extra ksegrp into the zombie ksegrp queue. 334 */ 335 void 336 ksegrp_stash(struct ksegrp *kg) 337 { 338 mtx_lock_spin(&kse_zombie_lock); 339 TAILQ_INSERT_HEAD(&zombie_ksegrps, kg, kg_ksegrp); 340 mtx_unlock_spin(&kse_zombie_lock); 341 } 342 343 /* 344 * Reap zombie kse resource. 345 */ 346 void 347 thread_reap(void) 348 { 349 struct thread *td_first, *td_next; 350 struct ksegrp *kg_first, * kg_next; 351 352 /* 353 * Don't even bother to lock if none at this instant, 354 * we really don't care about the next instant.. 355 */ 356 if ((!TAILQ_EMPTY(&zombie_threads)) 357 || (!TAILQ_EMPTY(&zombie_ksegrps))) { 358 mtx_lock_spin(&kse_zombie_lock); 359 td_first = TAILQ_FIRST(&zombie_threads); 360 kg_first = TAILQ_FIRST(&zombie_ksegrps); 361 if (td_first) 362 TAILQ_INIT(&zombie_threads); 363 if (kg_first) 364 TAILQ_INIT(&zombie_ksegrps); 365 mtx_unlock_spin(&kse_zombie_lock); 366 while (td_first) { 367 td_next = TAILQ_NEXT(td_first, td_runq); 368 if (td_first->td_ucred) 369 crfree(td_first->td_ucred); 370 thread_free(td_first); 371 td_first = td_next; 372 } 373 while (kg_first) { 374 kg_next = TAILQ_NEXT(kg_first, kg_ksegrp); 375 ksegrp_free(kg_first); 376 kg_first = kg_next; 377 } 378 /* 379 * there will always be a thread on the list if one of these 380 * is there. 381 */ 382 kse_GC(); 383 } 384 } 385 386 /* 387 * Allocate a ksegrp. 388 */ 389 struct ksegrp * 390 ksegrp_alloc(void) 391 { 392 return (uma_zalloc(ksegrp_zone, M_WAITOK)); 393 } 394 395 /* 396 * Allocate a thread. 397 */ 398 struct thread * 399 thread_alloc(void) 400 { 401 thread_reap(); /* check if any zombies to get */ 402 return (uma_zalloc(thread_zone, M_WAITOK)); 403 } 404 405 /* 406 * Deallocate a ksegrp. 407 */ 408 void 409 ksegrp_free(struct ksegrp *td) 410 { 411 uma_zfree(ksegrp_zone, td); 412 } 413 414 /* 415 * Deallocate a thread. 416 */ 417 void 418 thread_free(struct thread *td) 419 { 420 421 cpu_thread_clean(td); 422 uma_zfree(thread_zone, td); 423 } 424 425 /* 426 * Discard the current thread and exit from its context. 427 * Always called with scheduler locked. 428 * 429 * Because we can't free a thread while we're operating under its context, 430 * push the current thread into our CPU's deadthread holder. This means 431 * we needn't worry about someone else grabbing our context before we 432 * do a cpu_throw(). This may not be needed now as we are under schedlock. 433 * Maybe we can just do a thread_stash() as thr_exit1 does. 434 */ 435 /* XXX 436 * libthr expects its thread exit to return for the last 437 * thread, meaning that the program is back to non-threaded 438 * mode I guess. Because we do this (cpu_throw) unconditionally 439 * here, they have their own version of it. (thr_exit1()) 440 * that doesn't do it all if this was the last thread. 441 * It is also called from thread_suspend_check(). 442 * Of course in the end, they end up coming here through exit1 443 * anyhow.. After fixing 'thr' to play by the rules we should be able 444 * to merge these two functions together. 445 * 446 * called from: 447 * exit1() 448 * kse_exit() 449 * thr_exit() 450 * thread_user_enter() 451 * thread_userret() 452 * thread_suspend_check() 453 */ 454 void 455 thread_exit(void) 456 { 457 struct thread *td; 458 struct proc *p; 459 struct ksegrp *kg; 460 461 td = curthread; 462 kg = td->td_ksegrp; 463 p = td->td_proc; 464 465 mtx_assert(&sched_lock, MA_OWNED); 466 mtx_assert(&Giant, MA_NOTOWNED); 467 PROC_LOCK_ASSERT(p, MA_OWNED); 468 KASSERT(p != NULL, ("thread exiting without a process")); 469 KASSERT(kg != NULL, ("thread exiting without a kse group")); 470 CTR3(KTR_PROC, "thread_exit: thread %p (pid %ld, %s)", td, 471 (long)p->p_pid, p->p_comm); 472 KASSERT(TAILQ_EMPTY(&td->td_sigqueue.sq_list), ("signal pending")); 473 474 #ifdef AUDIT 475 AUDIT_SYSCALL_EXIT(0, td); 476 #endif 477 478 if (td->td_standin != NULL) { 479 /* 480 * Note that we don't need to free the cred here as it 481 * is done in thread_reap(). 482 */ 483 thread_stash(td->td_standin); 484 td->td_standin = NULL; 485 } 486 487 /* 488 * drop FPU & debug register state storage, or any other 489 * architecture specific resources that 490 * would not be on a new untouched process. 491 */ 492 cpu_thread_exit(td); /* XXXSMP */ 493 494 /* 495 * The thread is exiting. scheduler can release its stuff 496 * and collect stats etc. 497 */ 498 sched_thread_exit(td); 499 500 /* 501 * The last thread is left attached to the process 502 * So that the whole bundle gets recycled. Skip 503 * all this stuff if we never had threads. 504 * EXIT clears all sign of other threads when 505 * it goes to single threading, so the last thread always 506 * takes the short path. 507 */ 508 if (p->p_flag & P_HADTHREADS) { 509 if (p->p_numthreads > 1) { 510 thread_unlink(td); 511 512 /* XXX first arg not used in 4BSD or ULE */ 513 sched_exit_thread(FIRST_THREAD_IN_PROC(p), td); 514 515 /* 516 * The test below is NOT true if we are the 517 * sole exiting thread. P_STOPPED_SNGL is unset 518 * in exit1() after it is the only survivor. 519 */ 520 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 521 if (p->p_numthreads == p->p_suspcount) { 522 thread_unsuspend_one(p->p_singlethread); 523 } 524 } 525 526 /* 527 * Because each upcall structure has an owner thread, 528 * owner thread exits only when process is in exiting 529 * state, so upcall to userland is no longer needed, 530 * deleting upcall structure is safe here. 531 * So when all threads in a group is exited, all upcalls 532 * in the group should be automatically freed. 533 * XXXKSE This is a KSE thing and should be exported 534 * there somehow. 535 */ 536 upcall_remove(td); 537 538 /* 539 * If the thread we unlinked above was the last one, 540 * then this ksegrp should go away too. 541 */ 542 if (kg->kg_numthreads == 0) { 543 /* 544 * let the scheduler know about this in case 545 * it needs to recover stats or resources. 546 * Theoretically we could let 547 * sched_exit_ksegrp() do the equivalent of 548 * setting the concurrency to 0 549 * but don't do it yet to avoid changing 550 * the existing scheduler code until we 551 * are ready. 552 * We supply a random other ksegrp 553 * as the recipient of any built up 554 * cpu usage etc. (If the scheduler wants it). 555 * XXXKSE 556 * This is probably not fair so think of 557 * a better answer. 558 */ 559 sched_exit_ksegrp(FIRST_KSEGRP_IN_PROC(p), td); 560 sched_set_concurrency(kg, 0); /* XXX TEMP */ 561 ksegrp_unlink(kg); 562 ksegrp_stash(kg); 563 } 564 PROC_UNLOCK(p); 565 td->td_ksegrp = NULL; 566 PCPU_SET(deadthread, td); 567 } else { 568 /* 569 * The last thread is exiting.. but not through exit() 570 * what should we do? 571 * Theoretically this can't happen 572 * exit1() - clears threading flags before coming here 573 * kse_exit() - treats last thread specially 574 * thr_exit() - treats last thread specially 575 * thread_user_enter() - only if more exist 576 * thread_userret() - only if more exist 577 * thread_suspend_check() - only if more exist 578 */ 579 panic ("thread_exit: Last thread exiting on its own"); 580 } 581 } else { 582 /* 583 * non threaded process comes here. 584 * This includes an EX threaded process that is coming 585 * here via exit1(). (exit1 dethreads the proc first). 586 */ 587 PROC_UNLOCK(p); 588 } 589 td->td_state = TDS_INACTIVE; 590 CTR1(KTR_PROC, "thread_exit: cpu_throw() thread %p", td); 591 cpu_throw(td, choosethread()); 592 panic("I'm a teapot!"); 593 /* NOTREACHED */ 594 } 595 596 /* 597 * Do any thread specific cleanups that may be needed in wait() 598 * called with Giant, proc and schedlock not held. 599 */ 600 void 601 thread_wait(struct proc *p) 602 { 603 struct thread *td; 604 605 mtx_assert(&Giant, MA_NOTOWNED); 606 KASSERT((p->p_numthreads == 1), ("Multiple threads in wait1()")); 607 KASSERT((p->p_numksegrps == 1), ("Multiple ksegrps in wait1()")); 608 FOREACH_THREAD_IN_PROC(p, td) { 609 if (td->td_standin != NULL) { 610 if (td->td_standin->td_ucred != NULL) { 611 crfree(td->td_standin->td_ucred); 612 td->td_standin->td_ucred = NULL; 613 } 614 thread_free(td->td_standin); 615 td->td_standin = NULL; 616 } 617 cpu_thread_clean(td); 618 crfree(td->td_ucred); 619 } 620 thread_reap(); /* check for zombie threads etc. */ 621 } 622 623 /* 624 * Link a thread to a process. 625 * set up anything that needs to be initialized for it to 626 * be used by the process. 627 * 628 * Note that we do not link to the proc's ucred here. 629 * The thread is linked as if running but no KSE assigned. 630 * Called from: 631 * proc_linkup() 632 * thread_schedule_upcall() 633 * thr_create() 634 */ 635 void 636 thread_link(struct thread *td, struct ksegrp *kg) 637 { 638 struct proc *p; 639 640 p = kg->kg_proc; 641 td->td_state = TDS_INACTIVE; 642 td->td_proc = p; 643 td->td_ksegrp = kg; 644 td->td_flags = 0; 645 td->td_kflags = 0; 646 647 LIST_INIT(&td->td_contested); 648 sigqueue_init(&td->td_sigqueue, p); 649 callout_init(&td->td_slpcallout, CALLOUT_MPSAFE); 650 TAILQ_INSERT_HEAD(&p->p_threads, td, td_plist); 651 TAILQ_INSERT_HEAD(&kg->kg_threads, td, td_kglist); 652 p->p_numthreads++; 653 kg->kg_numthreads++; 654 } 655 656 /* 657 * Convert a process with one thread to an unthreaded process. 658 * Called from: 659 * thread_single(exit) (called from execve and exit) 660 * kse_exit() XXX may need cleaning up wrt KSE stuff 661 */ 662 void 663 thread_unthread(struct thread *td) 664 { 665 struct proc *p = td->td_proc; 666 667 KASSERT((p->p_numthreads == 1), ("Unthreading with >1 threads")); 668 upcall_remove(td); 669 p->p_flag &= ~(P_SA|P_HADTHREADS); 670 td->td_mailbox = NULL; 671 td->td_pflags &= ~(TDP_SA | TDP_CAN_UNBIND); 672 if (td->td_standin != NULL) { 673 thread_stash(td->td_standin); 674 td->td_standin = NULL; 675 } 676 sched_set_concurrency(td->td_ksegrp, 1); 677 } 678 679 /* 680 * Called from: 681 * thread_exit() 682 */ 683 void 684 thread_unlink(struct thread *td) 685 { 686 struct proc *p = td->td_proc; 687 struct ksegrp *kg = td->td_ksegrp; 688 689 mtx_assert(&sched_lock, MA_OWNED); 690 TAILQ_REMOVE(&p->p_threads, td, td_plist); 691 p->p_numthreads--; 692 TAILQ_REMOVE(&kg->kg_threads, td, td_kglist); 693 kg->kg_numthreads--; 694 /* could clear a few other things here */ 695 /* Must NOT clear links to proc and ksegrp! */ 696 } 697 698 /* 699 * Enforce single-threading. 700 * 701 * Returns 1 if the caller must abort (another thread is waiting to 702 * exit the process or similar). Process is locked! 703 * Returns 0 when you are successfully the only thread running. 704 * A process has successfully single threaded in the suspend mode when 705 * There are no threads in user mode. Threads in the kernel must be 706 * allowed to continue until they get to the user boundary. They may even 707 * copy out their return values and data before suspending. They may however be 708 * accellerated in reaching the user boundary as we will wake up 709 * any sleeping threads that are interruptable. (PCATCH). 710 */ 711 int 712 thread_single(int mode) 713 { 714 struct thread *td; 715 struct thread *td2; 716 struct proc *p; 717 int remaining; 718 719 td = curthread; 720 p = td->td_proc; 721 mtx_assert(&Giant, MA_NOTOWNED); 722 PROC_LOCK_ASSERT(p, MA_OWNED); 723 KASSERT((td != NULL), ("curthread is NULL")); 724 725 if ((p->p_flag & P_HADTHREADS) == 0) 726 return (0); 727 728 /* Is someone already single threading? */ 729 if (p->p_singlethread != NULL && p->p_singlethread != td) 730 return (1); 731 732 if (mode == SINGLE_EXIT) { 733 p->p_flag |= P_SINGLE_EXIT; 734 p->p_flag &= ~P_SINGLE_BOUNDARY; 735 } else { 736 p->p_flag &= ~P_SINGLE_EXIT; 737 if (mode == SINGLE_BOUNDARY) 738 p->p_flag |= P_SINGLE_BOUNDARY; 739 else 740 p->p_flag &= ~P_SINGLE_BOUNDARY; 741 } 742 p->p_flag |= P_STOPPED_SINGLE; 743 mtx_lock_spin(&sched_lock); 744 p->p_singlethread = td; 745 if (mode == SINGLE_EXIT) 746 remaining = p->p_numthreads; 747 else if (mode == SINGLE_BOUNDARY) 748 remaining = p->p_numthreads - p->p_boundary_count; 749 else 750 remaining = p->p_numthreads - p->p_suspcount; 751 while (remaining != 1) { 752 FOREACH_THREAD_IN_PROC(p, td2) { 753 if (td2 == td) 754 continue; 755 td2->td_flags |= TDF_ASTPENDING; 756 if (TD_IS_INHIBITED(td2)) { 757 switch (mode) { 758 case SINGLE_EXIT: 759 if (td->td_flags & TDF_DBSUSPEND) 760 td->td_flags &= ~TDF_DBSUSPEND; 761 if (TD_IS_SUSPENDED(td2)) 762 thread_unsuspend_one(td2); 763 if (TD_ON_SLEEPQ(td2) && 764 (td2->td_flags & TDF_SINTR)) 765 sleepq_abort(td2); 766 break; 767 case SINGLE_BOUNDARY: 768 if (TD_IS_SUSPENDED(td2) && 769 !(td2->td_flags & TDF_BOUNDARY)) 770 thread_unsuspend_one(td2); 771 if (TD_ON_SLEEPQ(td2) && 772 (td2->td_flags & TDF_SINTR)) 773 sleepq_abort(td2); 774 break; 775 default: 776 if (TD_IS_SUSPENDED(td2)) 777 continue; 778 /* 779 * maybe other inhibitted states too? 780 */ 781 if ((td2->td_flags & TDF_SINTR) && 782 (td2->td_inhibitors & 783 (TDI_SLEEPING | TDI_SWAPPED))) 784 thread_suspend_one(td2); 785 break; 786 } 787 } 788 } 789 if (mode == SINGLE_EXIT) 790 remaining = p->p_numthreads; 791 else if (mode == SINGLE_BOUNDARY) 792 remaining = p->p_numthreads - p->p_boundary_count; 793 else 794 remaining = p->p_numthreads - p->p_suspcount; 795 796 /* 797 * Maybe we suspended some threads.. was it enough? 798 */ 799 if (remaining == 1) 800 break; 801 802 /* 803 * Wake us up when everyone else has suspended. 804 * In the mean time we suspend as well. 805 */ 806 thread_suspend_one(td); 807 PROC_UNLOCK(p); 808 mi_switch(SW_VOL, NULL); 809 mtx_unlock_spin(&sched_lock); 810 PROC_LOCK(p); 811 mtx_lock_spin(&sched_lock); 812 if (mode == SINGLE_EXIT) 813 remaining = p->p_numthreads; 814 else if (mode == SINGLE_BOUNDARY) 815 remaining = p->p_numthreads - p->p_boundary_count; 816 else 817 remaining = p->p_numthreads - p->p_suspcount; 818 } 819 if (mode == SINGLE_EXIT) { 820 /* 821 * We have gotten rid of all the other threads and we 822 * are about to either exit or exec. In either case, 823 * we try our utmost to revert to being a non-threaded 824 * process. 825 */ 826 p->p_singlethread = NULL; 827 p->p_flag &= ~(P_STOPPED_SINGLE | P_SINGLE_EXIT); 828 thread_unthread(td); 829 } 830 mtx_unlock_spin(&sched_lock); 831 return (0); 832 } 833 834 /* 835 * Called in from locations that can safely check to see 836 * whether we have to suspend or at least throttle for a 837 * single-thread event (e.g. fork). 838 * 839 * Such locations include userret(). 840 * If the "return_instead" argument is non zero, the thread must be able to 841 * accept 0 (caller may continue), or 1 (caller must abort) as a result. 842 * 843 * The 'return_instead' argument tells the function if it may do a 844 * thread_exit() or suspend, or whether the caller must abort and back 845 * out instead. 846 * 847 * If the thread that set the single_threading request has set the 848 * P_SINGLE_EXIT bit in the process flags then this call will never return 849 * if 'return_instead' is false, but will exit. 850 * 851 * P_SINGLE_EXIT | return_instead == 0| return_instead != 0 852 *---------------+--------------------+--------------------- 853 * 0 | returns 0 | returns 0 or 1 854 * | when ST ends | immediatly 855 *---------------+--------------------+--------------------- 856 * 1 | thread exits | returns 1 857 * | | immediatly 858 * 0 = thread_exit() or suspension ok, 859 * other = return error instead of stopping the thread. 860 * 861 * While a full suspension is under effect, even a single threading 862 * thread would be suspended if it made this call (but it shouldn't). 863 * This call should only be made from places where 864 * thread_exit() would be safe as that may be the outcome unless 865 * return_instead is set. 866 */ 867 int 868 thread_suspend_check(int return_instead) 869 { 870 struct thread *td; 871 struct proc *p; 872 873 td = curthread; 874 p = td->td_proc; 875 mtx_assert(&Giant, MA_NOTOWNED); 876 PROC_LOCK_ASSERT(p, MA_OWNED); 877 while (P_SHOULDSTOP(p) || 878 ((p->p_flag & P_TRACED) && (td->td_flags & TDF_DBSUSPEND))) { 879 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 880 KASSERT(p->p_singlethread != NULL, 881 ("singlethread not set")); 882 /* 883 * The only suspension in action is a 884 * single-threading. Single threader need not stop. 885 * XXX Should be safe to access unlocked 886 * as it can only be set to be true by us. 887 */ 888 if (p->p_singlethread == td) 889 return (0); /* Exempt from stopping. */ 890 } 891 if ((p->p_flag & P_SINGLE_EXIT) && return_instead) 892 return (1); 893 894 /* Should we goto user boundary if we didn't come from there? */ 895 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE && 896 (p->p_flag & P_SINGLE_BOUNDARY) && return_instead) 897 return (1); 898 899 /* If thread will exit, flush its pending signals */ 900 if ((p->p_flag & P_SINGLE_EXIT) && (p->p_singlethread != td)) 901 sigqueue_flush(&td->td_sigqueue); 902 903 mtx_lock_spin(&sched_lock); 904 thread_stopped(p); 905 /* 906 * If the process is waiting for us to exit, 907 * this thread should just suicide. 908 * Assumes that P_SINGLE_EXIT implies P_STOPPED_SINGLE. 909 */ 910 if ((p->p_flag & P_SINGLE_EXIT) && (p->p_singlethread != td)) 911 thread_exit(); 912 913 /* 914 * When a thread suspends, it just 915 * moves to the processes's suspend queue 916 * and stays there. 917 */ 918 thread_suspend_one(td); 919 if (return_instead == 0) { 920 p->p_boundary_count++; 921 td->td_flags |= TDF_BOUNDARY; 922 } 923 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 924 if (p->p_numthreads == p->p_suspcount) 925 thread_unsuspend_one(p->p_singlethread); 926 } 927 PROC_UNLOCK(p); 928 mi_switch(SW_INVOL, NULL); 929 if (return_instead == 0) { 930 p->p_boundary_count--; 931 td->td_flags &= ~TDF_BOUNDARY; 932 } 933 mtx_unlock_spin(&sched_lock); 934 PROC_LOCK(p); 935 } 936 return (0); 937 } 938 939 void 940 thread_suspend_one(struct thread *td) 941 { 942 struct proc *p = td->td_proc; 943 944 mtx_assert(&sched_lock, MA_OWNED); 945 PROC_LOCK_ASSERT(p, MA_OWNED); 946 KASSERT(!TD_IS_SUSPENDED(td), ("already suspended")); 947 p->p_suspcount++; 948 TD_SET_SUSPENDED(td); 949 TAILQ_INSERT_TAIL(&p->p_suspended, td, td_runq); 950 } 951 952 void 953 thread_unsuspend_one(struct thread *td) 954 { 955 struct proc *p = td->td_proc; 956 957 mtx_assert(&sched_lock, MA_OWNED); 958 PROC_LOCK_ASSERT(p, MA_OWNED); 959 TAILQ_REMOVE(&p->p_suspended, td, td_runq); 960 TD_CLR_SUSPENDED(td); 961 p->p_suspcount--; 962 setrunnable(td); 963 } 964 965 /* 966 * Allow all threads blocked by single threading to continue running. 967 */ 968 void 969 thread_unsuspend(struct proc *p) 970 { 971 struct thread *td; 972 973 mtx_assert(&sched_lock, MA_OWNED); 974 PROC_LOCK_ASSERT(p, MA_OWNED); 975 if (!P_SHOULDSTOP(p)) { 976 while ((td = TAILQ_FIRST(&p->p_suspended))) { 977 thread_unsuspend_one(td); 978 } 979 } else if ((P_SHOULDSTOP(p) == P_STOPPED_SINGLE) && 980 (p->p_numthreads == p->p_suspcount)) { 981 /* 982 * Stopping everything also did the job for the single 983 * threading request. Now we've downgraded to single-threaded, 984 * let it continue. 985 */ 986 thread_unsuspend_one(p->p_singlethread); 987 } 988 } 989 990 /* 991 * End the single threading mode.. 992 */ 993 void 994 thread_single_end(void) 995 { 996 struct thread *td; 997 struct proc *p; 998 999 td = curthread; 1000 p = td->td_proc; 1001 PROC_LOCK_ASSERT(p, MA_OWNED); 1002 p->p_flag &= ~(P_STOPPED_SINGLE | P_SINGLE_EXIT | P_SINGLE_BOUNDARY); 1003 mtx_lock_spin(&sched_lock); 1004 p->p_singlethread = NULL; 1005 p->p_procscopegrp = NULL; 1006 /* 1007 * If there are other threads they mey now run, 1008 * unless of course there is a blanket 'stop order' 1009 * on the process. The single threader must be allowed 1010 * to continue however as this is a bad place to stop. 1011 */ 1012 if ((p->p_numthreads != 1) && (!P_SHOULDSTOP(p))) { 1013 while ((td = TAILQ_FIRST(&p->p_suspended))) { 1014 thread_unsuspend_one(td); 1015 } 1016 } 1017 mtx_unlock_spin(&sched_lock); 1018 } 1019 1020 /* 1021 * Called before going into an interruptible sleep to see if we have been 1022 * interrupted or requested to exit. 1023 */ 1024 int 1025 thread_sleep_check(struct thread *td) 1026 { 1027 struct proc *p; 1028 1029 p = td->td_proc; 1030 mtx_assert(&sched_lock, MA_OWNED); 1031 if (p->p_flag & P_HADTHREADS) { 1032 if (p->p_singlethread != td) { 1033 if (p->p_flag & P_SINGLE_EXIT) 1034 return (EINTR); 1035 if (p->p_flag & P_SINGLE_BOUNDARY) 1036 return (ERESTART); 1037 } 1038 if (td->td_flags & TDF_INTERRUPT) 1039 return (td->td_intrval); 1040 } 1041 return (0); 1042 } 1043 1044 struct thread * 1045 thread_find(struct proc *p, lwpid_t tid) 1046 { 1047 struct thread *td; 1048 1049 PROC_LOCK_ASSERT(p, MA_OWNED); 1050 mtx_lock_spin(&sched_lock); 1051 FOREACH_THREAD_IN_PROC(p, td) { 1052 if (td->td_tid == tid) 1053 break; 1054 } 1055 mtx_unlock_spin(&sched_lock); 1056 return (td); 1057 } 1058