1 /* 2 * Copyright (c) 1982, 1986, 1989, 1991, 1993 3 * The Regents of the University of California. 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, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95 34 * $FreeBSD$ 35 */ 36 37 #include "opt_ktrace.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/kernel.h> 42 #include <sys/lock.h> 43 #include <sys/malloc.h> 44 #include <sys/mutex.h> 45 #include <sys/proc.h> 46 #include <sys/sysproto.h> 47 #include <sys/kse.h> 48 #include <sys/sysctl.h> 49 #include <sys/filedesc.h> 50 #include <sys/tty.h> 51 #include <sys/signalvar.h> 52 #include <sys/sx.h> 53 #include <sys/user.h> 54 #include <sys/jail.h> 55 #ifdef KTRACE 56 #include <sys/uio.h> 57 #include <sys/ktrace.h> 58 #endif 59 60 #include <vm/vm.h> 61 #include <vm/vm_extern.h> 62 #include <vm/pmap.h> 63 #include <vm/vm_map.h> 64 #include <vm/uma.h> 65 #include <machine/critical.h> 66 67 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header"); 68 MALLOC_DEFINE(M_SESSION, "session", "session header"); 69 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); 70 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 71 72 static struct proc *dopfind(register pid_t); 73 74 static void doenterpgrp(struct proc *, struct pgrp *); 75 76 static void pgdelete(struct pgrp *); 77 78 static void orphanpg(struct pgrp *pg); 79 80 static void proc_ctor(void *mem, int size, void *arg); 81 static void proc_dtor(void *mem, int size, void *arg); 82 static void proc_init(void *mem, int size); 83 static void proc_fini(void *mem, int size); 84 85 /* 86 * Other process lists 87 */ 88 struct pidhashhead *pidhashtbl; 89 u_long pidhash; 90 struct pgrphashhead *pgrphashtbl; 91 u_long pgrphash; 92 struct proclist allproc; 93 struct proclist zombproc; 94 struct sx allproc_lock; 95 struct sx proctree_lock; 96 struct mtx pargs_ref_lock; 97 uma_zone_t proc_zone; 98 uma_zone_t ithread_zone; 99 100 static int active_procs; 101 static int cached_procs; 102 static int allocated_procs; 103 104 #define RANGEOF(type, start, end) (offsetof(type, end) - offsetof(type, start)) 105 106 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); 107 108 /* 109 * Initialize global process hashing structures. 110 */ 111 void 112 procinit() 113 { 114 115 sx_init(&allproc_lock, "allproc"); 116 sx_init(&proctree_lock, "proctree"); 117 mtx_init(&pargs_ref_lock, "struct pargs.ref", NULL, MTX_DEF); 118 LIST_INIT(&allproc); 119 LIST_INIT(&zombproc); 120 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); 121 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); 122 proc_zone = uma_zcreate("PROC", sizeof (struct proc), 123 proc_ctor, proc_dtor, proc_init, proc_fini, 124 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 125 uihashinit(); 126 } 127 128 /* 129 * Prepare a proc for use. 130 */ 131 static void 132 proc_ctor(void *mem, int size, void *arg) 133 { 134 struct proc *p; 135 136 KASSERT((size == sizeof(struct proc)), 137 ("size mismatch: %d != %d\n", size, (int)sizeof(struct proc))); 138 p = (struct proc *)mem; 139 #if 0 140 /* 141 * Maybe move these from process creation, but maybe not. 142 * Moving them here takes them away from their "natural" place 143 * in the fork process. 144 */ 145 bzero(&p->p_startzero, 146 (unsigned) RANGEOF(struct proc, p_startzero, p_endzero)); 147 p->p_state = PRS_NEW; 148 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK); 149 LIST_INIT(&p->p_children); 150 callout_init(&p->p_itcallout, 0); 151 #endif 152 cached_procs--; 153 active_procs++; 154 } 155 156 /* 157 * Reclaim a proc after use. 158 */ 159 static void 160 proc_dtor(void *mem, int size, void *arg) 161 { 162 struct proc *p; 163 164 KASSERT((size == sizeof(struct proc)), 165 ("size mismatch: %d != %d\n", size, (int)sizeof(struct proc))); 166 p = (struct proc *)mem; 167 /* INVARIANTS checks go here */ 168 #if 0 /* See comment in proc_ctor about seperating things */ 169 mtx_destroy(&p->p_mtx); 170 #endif 171 active_procs--; 172 cached_procs++; 173 } 174 175 /* 176 * Initialize type-stable parts of a proc (when newly created). 177 */ 178 static void 179 proc_init(void *mem, int size) 180 { 181 struct proc *p; 182 183 KASSERT((size == sizeof(struct proc)), 184 ("size mismatch: %d != %d\n", size, (int)sizeof(struct proc))); 185 p = (struct proc *)mem; 186 vm_proc_new(p); 187 cached_procs++; 188 allocated_procs++; 189 } 190 191 /* 192 * Tear down type-stable parts of a proc (just before being discarded) 193 */ 194 static void 195 proc_fini(void *mem, int size) 196 { 197 struct proc *p; 198 199 KASSERT((size == sizeof(struct proc)), 200 ("size mismatch: %d != %d\n", size, (int)sizeof(struct proc))); 201 p = (struct proc *)mem; 202 vm_proc_dispose(p); 203 cached_procs--; 204 allocated_procs--; 205 } 206 207 /* 208 * KSE is linked onto the idle queue. 209 */ 210 void 211 kse_link(struct kse *ke, struct ksegrp *kg) 212 { 213 struct proc *p = kg->kg_proc; 214 215 TAILQ_INSERT_HEAD(&kg->kg_kseq, ke, ke_kglist); 216 kg->kg_kses++; 217 ke->ke_state = KES_IDLE; 218 TAILQ_INSERT_HEAD(&kg->kg_iq, ke, ke_kgrlist); 219 kg->kg_idle_kses++; 220 ke->ke_proc = p; 221 ke->ke_ksegrp = kg; 222 ke->ke_thread = NULL; 223 ke->ke_oncpu = NOCPU; 224 } 225 226 void 227 ksegrp_link(struct ksegrp *kg, struct proc *p) 228 { 229 230 TAILQ_INIT(&kg->kg_threads); 231 TAILQ_INIT(&kg->kg_runq); /* links with td_runq */ 232 TAILQ_INIT(&kg->kg_slpq); /* links with td_runq */ 233 TAILQ_INIT(&kg->kg_kseq); /* all kses in ksegrp */ 234 TAILQ_INIT(&kg->kg_iq); /* all kses in ksegrp */ 235 kg->kg_proc = p; 236 /* the following counters are in the -zero- section and may not need clearing */ 237 kg->kg_numthreads = 0; 238 kg->kg_runnable = 0; 239 kg->kg_kses = 0; 240 kg->kg_idle_kses = 0; 241 kg->kg_runq_kses = 0; /* XXXKSE change name */ 242 /* link it in now that it's consitant */ 243 p->p_numksegrps++; 244 TAILQ_INSERT_HEAD(&p->p_ksegrps, kg, kg_ksegrp); 245 } 246 247 /* 248 * for a newly created process, 249 * link up a the structure and its initial threads etc. 250 */ 251 void 252 proc_linkup(struct proc *p, struct ksegrp *kg, 253 struct kse *ke, struct thread *td) 254 { 255 256 TAILQ_INIT(&p->p_ksegrps); /* all ksegrps in proc */ 257 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 258 TAILQ_INIT(&p->p_suspended); /* Threads suspended */ 259 260 ksegrp_link(kg, p); 261 kse_link(ke, kg); 262 thread_link(td, kg); 263 } 264 265 int 266 thread_wakeup(struct thread *td, struct thread_wakeup_args *uap) 267 { 268 269 return(ENOSYS); 270 } 271 272 int 273 kse_exit(struct thread *td, struct kse_exit_args *uap) 274 { 275 276 return(ENOSYS); 277 } 278 279 int 280 kse_yield(struct thread *td, struct kse_yield_args *uap) 281 { 282 283 PROC_LOCK(td->td_proc); 284 mtx_lock_spin(&sched_lock); 285 thread_exit(); 286 /* NOTREACHED */ 287 return(0); 288 } 289 290 int kse_wakeup(struct thread *td, struct kse_wakeup_args *uap) 291 { 292 293 return(ENOSYS); 294 } 295 296 /* 297 * No new KSEG: first call: use current KSE, don't schedule an upcall 298 * All other situations, do alloate a new KSE and schedule an upcall on it. 299 */ 300 /* struct kse_new_args { 301 struct kse_mailbox *mbx; 302 int new_grp_flag; 303 }; */ 304 int 305 kse_new(struct thread *td, struct kse_new_args *uap) 306 { 307 struct kse *newkse; 308 struct proc *p; 309 struct kse_mailbox mbx; 310 int err; 311 312 p = td->td_proc; 313 if ((err = copyin(uap->mbx, &mbx, sizeof(mbx)))) 314 return (err); 315 PROC_LOCK(p); 316 /* 317 * If we have no KSE mode set, just set it, and skip KSE and KSEGRP 318 * creation. You cannot request a new group with the first one as 319 * you are effectively getting one. Instead, go directly to saving 320 * the upcall info. 321 */ 322 if ((td->td_proc->p_flag & P_KSES) || (uap->new_grp_flag)) { 323 324 return (EINVAL); /* XXX */ 325 /* 326 * If newgroup then create the new group. 327 * Check we have the resources for this. 328 */ 329 /* Copy lots of fields from the current KSEGRP. */ 330 /* Create the new KSE */ 331 /* Copy lots of fields from the current KSE. */ 332 } else { 333 /* 334 * We are switching to KSEs so just 335 * use the preallocated ones for this call. 336 * XXXKSE if we have to initialise any fields for KSE 337 * mode operation, do it here. 338 */ 339 newkse = td->td_kse; 340 } 341 /* 342 * Fill out the KSE-mode specific fields of the new kse. 343 */ 344 PROC_UNLOCK(p); 345 mtx_lock_spin(&sched_lock); 346 mi_switch(); /* Save current registers to PCB. */ 347 mtx_unlock_spin(&sched_lock); 348 newkse->ke_upcall = mbx.kmbx_upcall; 349 newkse->ke_stackbase = mbx.kmbx_stackbase; 350 newkse->ke_stacksize = mbx.kmbx_stacksize; 351 newkse->ke_mailbox = uap->mbx; 352 cpu_save_upcall(td, newkse); 353 /* Note that we are the returning syscall */ 354 td->td_retval[0] = 0; 355 td->td_retval[1] = 0; 356 357 if ((td->td_proc->p_flag & P_KSES) || (uap->new_grp_flag)) { 358 thread_schedule_upcall(td, newkse); 359 } else { 360 /* 361 * Don't set this until we are truely ready, because 362 * things will start acting differently. Return to the 363 * calling code for the first time. Assuming we set up 364 * the mailboxes right, all syscalls after this will be 365 * asynchronous. 366 */ 367 td->td_proc->p_flag |= P_KSES; 368 } 369 return (0); 370 } 371 372 /* 373 * Is p an inferior of the current process? 374 */ 375 int 376 inferior(p) 377 register struct proc *p; 378 { 379 380 sx_assert(&proctree_lock, SX_LOCKED); 381 for (; p != curproc; p = p->p_pptr) 382 if (p->p_pid == 0) 383 return (0); 384 return (1); 385 } 386 387 /* 388 * Locate a process by number 389 */ 390 struct proc * 391 pfind(pid) 392 register pid_t pid; 393 { 394 register struct proc *p; 395 396 sx_slock(&allproc_lock); 397 p = dopfind(pid); 398 sx_sunlock(&allproc_lock); 399 return (p); 400 } 401 402 static struct proc * 403 dopfind(pid) 404 register pid_t pid; 405 { 406 register struct proc *p; 407 408 sx_assert(&allproc_lock, SX_LOCKED); 409 410 LIST_FOREACH(p, PIDHASH(pid), p_hash) 411 if (p->p_pid == pid) { 412 PROC_LOCK(p); 413 break; 414 } 415 return (p); 416 } 417 418 /* 419 * Locate a process group by number. 420 * The caller must hold proctree_lock. 421 */ 422 struct pgrp * 423 pgfind(pgid) 424 register pid_t pgid; 425 { 426 register struct pgrp *pgrp; 427 428 sx_assert(&proctree_lock, SX_LOCKED); 429 430 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { 431 if (pgrp->pg_id == pgid) { 432 PGRP_LOCK(pgrp); 433 return (pgrp); 434 } 435 } 436 return (NULL); 437 } 438 439 /* 440 * Create a new process group. 441 * pgid must be equal to the pid of p. 442 * Begin a new session if required. 443 */ 444 int 445 enterpgrp(p, pgid, pgrp, sess) 446 register struct proc *p; 447 pid_t pgid; 448 struct pgrp *pgrp; 449 struct session *sess; 450 { 451 struct pgrp *pgrp2; 452 453 sx_assert(&proctree_lock, SX_XLOCKED); 454 455 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL")); 456 KASSERT(p->p_pid == pgid, 457 ("enterpgrp: new pgrp and pid != pgid")); 458 459 pgrp2 = pgfind(pgid); 460 461 KASSERT(pgrp2 == NULL, 462 ("enterpgrp: pgrp with pgid exists")); 463 KASSERT(!SESS_LEADER(p), 464 ("enterpgrp: session leader attempted setpgrp")); 465 466 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 467 468 if (sess != NULL) { 469 /* 470 * new session 471 */ 472 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); 473 PROC_LOCK(p); 474 p->p_flag &= ~P_CONTROLT; 475 PROC_UNLOCK(p); 476 PGRP_LOCK(pgrp); 477 sess->s_leader = p; 478 sess->s_sid = p->p_pid; 479 sess->s_count = 1; 480 sess->s_ttyvp = NULL; 481 sess->s_ttyp = NULL; 482 bcopy(p->p_session->s_login, sess->s_login, 483 sizeof(sess->s_login)); 484 pgrp->pg_session = sess; 485 KASSERT(p == curproc, 486 ("enterpgrp: mksession and p != curproc")); 487 } else { 488 pgrp->pg_session = p->p_session; 489 SESS_LOCK(pgrp->pg_session); 490 pgrp->pg_session->s_count++; 491 SESS_UNLOCK(pgrp->pg_session); 492 PGRP_LOCK(pgrp); 493 } 494 pgrp->pg_id = pgid; 495 LIST_INIT(&pgrp->pg_members); 496 497 /* 498 * As we have an exclusive lock of proctree_lock, 499 * this should not deadlock. 500 */ 501 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); 502 pgrp->pg_jobc = 0; 503 SLIST_INIT(&pgrp->pg_sigiolst); 504 PGRP_UNLOCK(pgrp); 505 506 doenterpgrp(p, pgrp); 507 508 return (0); 509 } 510 511 /* 512 * Move p to an existing process group 513 */ 514 int 515 enterthispgrp(p, pgrp) 516 register struct proc *p; 517 struct pgrp *pgrp; 518 { 519 520 sx_assert(&proctree_lock, SX_XLOCKED); 521 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 522 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 523 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 524 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 525 KASSERT(pgrp->pg_session == p->p_session, 526 ("%s: pgrp's session %p, p->p_session %p.\n", 527 __func__, 528 pgrp->pg_session, 529 p->p_session)); 530 KASSERT(pgrp != p->p_pgrp, 531 ("%s: p belongs to pgrp.", __func__)); 532 533 doenterpgrp(p, pgrp); 534 535 return (0); 536 } 537 538 /* 539 * Move p to a process group 540 */ 541 static void 542 doenterpgrp(p, pgrp) 543 struct proc *p; 544 struct pgrp *pgrp; 545 { 546 struct pgrp *savepgrp; 547 548 sx_assert(&proctree_lock, SX_XLOCKED); 549 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 550 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 551 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 552 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 553 554 savepgrp = p->p_pgrp; 555 556 /* 557 * Adjust eligibility of affected pgrps to participate in job control. 558 * Increment eligibility counts before decrementing, otherwise we 559 * could reach 0 spuriously during the first call. 560 */ 561 fixjobc(p, pgrp, 1); 562 fixjobc(p, p->p_pgrp, 0); 563 564 PGRP_LOCK(pgrp); 565 PGRP_LOCK(savepgrp); 566 PROC_LOCK(p); 567 LIST_REMOVE(p, p_pglist); 568 p->p_pgrp = pgrp; 569 PROC_UNLOCK(p); 570 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 571 PGRP_UNLOCK(savepgrp); 572 PGRP_UNLOCK(pgrp); 573 if (LIST_EMPTY(&savepgrp->pg_members)) 574 pgdelete(savepgrp); 575 } 576 577 /* 578 * remove process from process group 579 */ 580 int 581 leavepgrp(p) 582 register struct proc *p; 583 { 584 struct pgrp *savepgrp; 585 586 sx_assert(&proctree_lock, SX_XLOCKED); 587 savepgrp = p->p_pgrp; 588 PGRP_LOCK(savepgrp); 589 PROC_LOCK(p); 590 LIST_REMOVE(p, p_pglist); 591 p->p_pgrp = NULL; 592 PROC_UNLOCK(p); 593 PGRP_UNLOCK(savepgrp); 594 if (LIST_EMPTY(&savepgrp->pg_members)) 595 pgdelete(savepgrp); 596 return (0); 597 } 598 599 /* 600 * delete a process group 601 */ 602 static void 603 pgdelete(pgrp) 604 register struct pgrp *pgrp; 605 { 606 struct session *savesess; 607 608 sx_assert(&proctree_lock, SX_XLOCKED); 609 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 610 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 611 612 /* 613 * Reset any sigio structures pointing to us as a result of 614 * F_SETOWN with our pgid. 615 */ 616 funsetownlst(&pgrp->pg_sigiolst); 617 618 PGRP_LOCK(pgrp); 619 if (pgrp->pg_session->s_ttyp != NULL && 620 pgrp->pg_session->s_ttyp->t_pgrp == pgrp) 621 pgrp->pg_session->s_ttyp->t_pgrp = NULL; 622 LIST_REMOVE(pgrp, pg_hash); 623 savesess = pgrp->pg_session; 624 SESS_LOCK(savesess); 625 savesess->s_count--; 626 SESS_UNLOCK(savesess); 627 PGRP_UNLOCK(pgrp); 628 if (savesess->s_count == 0) { 629 mtx_destroy(&savesess->s_mtx); 630 FREE(pgrp->pg_session, M_SESSION); 631 } 632 mtx_destroy(&pgrp->pg_mtx); 633 FREE(pgrp, M_PGRP); 634 } 635 636 /* 637 * Adjust pgrp jobc counters when specified process changes process group. 638 * We count the number of processes in each process group that "qualify" 639 * the group for terminal job control (those with a parent in a different 640 * process group of the same session). If that count reaches zero, the 641 * process group becomes orphaned. Check both the specified process' 642 * process group and that of its children. 643 * entering == 0 => p is leaving specified group. 644 * entering == 1 => p is entering specified group. 645 */ 646 void 647 fixjobc(p, pgrp, entering) 648 register struct proc *p; 649 register struct pgrp *pgrp; 650 int entering; 651 { 652 register struct pgrp *hispgrp; 653 register struct session *mysession; 654 655 sx_assert(&proctree_lock, SX_LOCKED); 656 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 657 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 658 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 659 660 /* 661 * Check p's parent to see whether p qualifies its own process 662 * group; if so, adjust count for p's process group. 663 */ 664 mysession = pgrp->pg_session; 665 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp && 666 hispgrp->pg_session == mysession) { 667 PGRP_LOCK(pgrp); 668 if (entering) 669 pgrp->pg_jobc++; 670 else { 671 --pgrp->pg_jobc; 672 if (pgrp->pg_jobc == 0) 673 orphanpg(pgrp); 674 } 675 PGRP_UNLOCK(pgrp); 676 } 677 678 /* 679 * Check this process' children to see whether they qualify 680 * their process groups; if so, adjust counts for children's 681 * process groups. 682 */ 683 LIST_FOREACH(p, &p->p_children, p_sibling) { 684 if ((hispgrp = p->p_pgrp) != pgrp && 685 hispgrp->pg_session == mysession && 686 p->p_state != PRS_ZOMBIE) { 687 PGRP_LOCK(hispgrp); 688 if (entering) 689 hispgrp->pg_jobc++; 690 else { 691 --hispgrp->pg_jobc; 692 if (hispgrp->pg_jobc == 0) 693 orphanpg(hispgrp); 694 } 695 PGRP_UNLOCK(hispgrp); 696 } 697 } 698 } 699 700 /* 701 * A process group has become orphaned; 702 * if there are any stopped processes in the group, 703 * hang-up all process in that group. 704 */ 705 static void 706 orphanpg(pg) 707 struct pgrp *pg; 708 { 709 register struct proc *p; 710 711 PGRP_LOCK_ASSERT(pg, MA_OWNED); 712 713 mtx_lock_spin(&sched_lock); 714 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 715 if (P_SHOULDSTOP(p)) { 716 mtx_unlock_spin(&sched_lock); 717 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 718 PROC_LOCK(p); 719 psignal(p, SIGHUP); 720 psignal(p, SIGCONT); 721 PROC_UNLOCK(p); 722 } 723 return; 724 } 725 } 726 mtx_unlock_spin(&sched_lock); 727 } 728 729 #include "opt_ddb.h" 730 #ifdef DDB 731 #include <ddb/ddb.h> 732 733 DB_SHOW_COMMAND(pgrpdump, pgrpdump) 734 { 735 register struct pgrp *pgrp; 736 register struct proc *p; 737 register int i; 738 739 for (i = 0; i <= pgrphash; i++) { 740 if (!LIST_EMPTY(&pgrphashtbl[i])) { 741 printf("\tindx %d\n", i); 742 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { 743 printf( 744 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n", 745 (void *)pgrp, (long)pgrp->pg_id, 746 (void *)pgrp->pg_session, 747 pgrp->pg_session->s_count, 748 (void *)LIST_FIRST(&pgrp->pg_members)); 749 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 750 printf("\t\tpid %ld addr %p pgrp %p\n", 751 (long)p->p_pid, (void *)p, 752 (void *)p->p_pgrp); 753 } 754 } 755 } 756 } 757 } 758 #endif /* DDB */ 759 760 /* 761 * Fill in an kinfo_proc structure for the specified process. 762 * Must be called with the target process locked. 763 */ 764 void 765 fill_kinfo_proc(p, kp) 766 struct proc *p; 767 struct kinfo_proc *kp; 768 { 769 struct thread *td; 770 struct tty *tp; 771 struct session *sp; 772 struct timeval tv; 773 774 bzero(kp, sizeof(*kp)); 775 776 kp->ki_structsize = sizeof(*kp); 777 kp->ki_paddr = p; 778 PROC_LOCK_ASSERT(p, MA_OWNED); 779 kp->ki_addr =/* p->p_addr; */0; /* XXXKSE */ 780 kp->ki_args = p->p_args; 781 kp->ki_textvp = p->p_textvp; 782 #ifdef KTRACE 783 kp->ki_tracep = p->p_tracep; 784 mtx_lock(&ktrace_mtx); 785 kp->ki_traceflag = p->p_traceflag; 786 mtx_unlock(&ktrace_mtx); 787 #endif 788 kp->ki_fd = p->p_fd; 789 kp->ki_vmspace = p->p_vmspace; 790 if (p->p_ucred) { 791 kp->ki_uid = p->p_ucred->cr_uid; 792 kp->ki_ruid = p->p_ucred->cr_ruid; 793 kp->ki_svuid = p->p_ucred->cr_svuid; 794 /* XXX bde doesn't like KI_NGROUPS */ 795 kp->ki_ngroups = min(p->p_ucred->cr_ngroups, KI_NGROUPS); 796 bcopy(p->p_ucred->cr_groups, kp->ki_groups, 797 kp->ki_ngroups * sizeof(gid_t)); 798 kp->ki_rgid = p->p_ucred->cr_rgid; 799 kp->ki_svgid = p->p_ucred->cr_svgid; 800 } 801 if (p->p_procsig) { 802 kp->ki_sigignore = p->p_procsig->ps_sigignore; 803 kp->ki_sigcatch = p->p_procsig->ps_sigcatch; 804 } 805 mtx_lock_spin(&sched_lock); 806 if (p->p_state != PRS_NEW && 807 p->p_state != PRS_ZOMBIE && 808 p->p_vmspace != NULL) { 809 struct vmspace *vm = p->p_vmspace; 810 811 kp->ki_size = vm->vm_map.size; 812 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ 813 if (p->p_sflag & PS_INMEM) 814 kp->ki_rssize += UAREA_PAGES; 815 FOREACH_THREAD_IN_PROC(p, td) /* XXXKSE: thread swapout check */ 816 kp->ki_rssize += KSTACK_PAGES; 817 kp->ki_swrss = vm->vm_swrss; 818 kp->ki_tsize = vm->vm_tsize; 819 kp->ki_dsize = vm->vm_dsize; 820 kp->ki_ssize = vm->vm_ssize; 821 } 822 if ((p->p_sflag & PS_INMEM) && p->p_stats) { 823 kp->ki_start = p->p_stats->p_start; 824 kp->ki_rusage = p->p_stats->p_ru; 825 kp->ki_childtime.tv_sec = p->p_stats->p_cru.ru_utime.tv_sec + 826 p->p_stats->p_cru.ru_stime.tv_sec; 827 kp->ki_childtime.tv_usec = p->p_stats->p_cru.ru_utime.tv_usec + 828 p->p_stats->p_cru.ru_stime.tv_usec; 829 } 830 if (p->p_state != PRS_ZOMBIE) { 831 td = FIRST_THREAD_IN_PROC(p); 832 if (!(p->p_flag & P_KSES)) { 833 if (td->td_wmesg != NULL) { 834 strncpy(kp->ki_wmesg, td->td_wmesg, 835 sizeof(kp->ki_wmesg) - 1); 836 } 837 if (td->td_state == TDS_MTX) { 838 kp->ki_kiflag |= KI_MTXBLOCK; 839 strncpy(kp->ki_mtxname, td->td_mtxname, 840 sizeof(kp->ki_mtxname) - 1); 841 } 842 } 843 844 if (p->p_state == PRS_NORMAL) { /* XXXKSE very aproximate */ 845 if ((td->td_state == TDS_RUNQ) || 846 (td->td_state == TDS_RUNNING)) { 847 kp->ki_stat = SRUN; 848 } else if (td->td_state == TDS_SLP) { 849 kp->ki_stat = SSLEEP; 850 } else if (P_SHOULDSTOP(p)) { 851 kp->ki_stat = SSTOP; 852 } else if (td->td_state == TDS_MTX) { 853 kp->ki_stat = SMTX; 854 } else { 855 kp->ki_stat = SWAIT; 856 } 857 } else { 858 kp->ki_stat = SIDL; 859 } 860 861 kp->ki_sflag = p->p_sflag; 862 kp->ki_swtime = p->p_swtime; 863 kp->ki_pid = p->p_pid; 864 /* vvv XXXKSE */ 865 if (!(p->p_flag & P_KSES)) { 866 bintime2timeval(&p->p_runtime, &tv); 867 kp->ki_runtime = tv.tv_sec * (u_int64_t)1000000 + tv.tv_usec; 868 kp->ki_pctcpu = p->p_kse.ke_pctcpu; 869 kp->ki_estcpu = p->p_ksegrp.kg_estcpu; 870 kp->ki_slptime = p->p_ksegrp.kg_slptime; 871 kp->ki_wchan = td->td_wchan; 872 kp->ki_pri.pri_level = td->td_priority; 873 kp->ki_pri.pri_user = p->p_ksegrp.kg_user_pri; 874 kp->ki_pri.pri_class = p->p_ksegrp.kg_pri_class; 875 kp->ki_pri.pri_native = td->td_base_pri; 876 kp->ki_nice = p->p_ksegrp.kg_nice; 877 kp->ki_rqindex = p->p_kse.ke_rqindex; 878 kp->ki_oncpu = p->p_kse.ke_oncpu; 879 kp->ki_lastcpu = td->td_lastcpu; 880 kp->ki_tdflags = td->td_flags; 881 kp->ki_pcb = td->td_pcb; 882 kp->ki_kstack = (void *)td->td_kstack; 883 } else { 884 kp->ki_oncpu = -1; 885 kp->ki_lastcpu = -1; 886 kp->ki_tdflags = -1; 887 /* All the reast are 0 */ 888 } 889 } else { 890 kp->ki_stat = SZOMB; 891 } 892 /* ^^^ XXXKSE */ 893 mtx_unlock_spin(&sched_lock); 894 sp = NULL; 895 tp = NULL; 896 if (p->p_pgrp) { 897 kp->ki_pgid = p->p_pgrp->pg_id; 898 kp->ki_jobc = p->p_pgrp->pg_jobc; 899 sp = p->p_pgrp->pg_session; 900 901 if (sp != NULL) { 902 kp->ki_sid = sp->s_sid; 903 SESS_LOCK(sp); 904 strncpy(kp->ki_login, sp->s_login, 905 sizeof(kp->ki_login) - 1); 906 if (sp->s_ttyvp) 907 kp->ki_kiflag |= KI_CTTY; 908 if (SESS_LEADER(p)) 909 kp->ki_kiflag |= KI_SLEADER; 910 tp = sp->s_ttyp; 911 SESS_UNLOCK(sp); 912 } 913 } 914 if ((p->p_flag & P_CONTROLT) && tp != NULL) { 915 kp->ki_tdev = dev2udev(tp->t_dev); 916 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; 917 if (tp->t_session) 918 kp->ki_tsid = tp->t_session->s_sid; 919 } else 920 kp->ki_tdev = NOUDEV; 921 if (p->p_comm[0] != '\0') { 922 strncpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm) - 1); 923 strncpy(kp->ki_ocomm, p->p_comm, sizeof(kp->ki_ocomm) - 1); 924 } 925 kp->ki_siglist = p->p_siglist; 926 kp->ki_sigmask = p->p_sigmask; 927 kp->ki_xstat = p->p_xstat; 928 kp->ki_acflag = p->p_acflag; 929 kp->ki_flag = p->p_flag; 930 /* If jailed(p->p_ucred), emulate the old P_JAILED flag. */ 931 if (jailed(p->p_ucred)) 932 kp->ki_flag |= P_JAILED; 933 kp->ki_lock = p->p_lock; 934 if (p->p_pptr) 935 kp->ki_ppid = p->p_pptr->p_pid; 936 } 937 938 /* 939 * Locate a zombie process by number 940 */ 941 struct proc * 942 zpfind(pid_t pid) 943 { 944 struct proc *p; 945 946 sx_slock(&allproc_lock); 947 LIST_FOREACH(p, &zombproc, p_list) 948 if (p->p_pid == pid) { 949 PROC_LOCK(p); 950 break; 951 } 952 sx_sunlock(&allproc_lock); 953 return (p); 954 } 955 956 957 /* 958 * Must be called with the process locked and will return with it unlocked. 959 */ 960 static int 961 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int doingzomb) 962 { 963 struct kinfo_proc kinfo_proc; 964 int error; 965 struct proc *np; 966 pid_t pid = p->p_pid; 967 968 PROC_LOCK_ASSERT(p, MA_OWNED); 969 fill_kinfo_proc(p, &kinfo_proc); 970 PROC_UNLOCK(p); 971 error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc, sizeof(kinfo_proc)); 972 if (error) 973 return (error); 974 if (doingzomb) 975 np = zpfind(pid); 976 else { 977 if (pid == 0) 978 return (0); 979 np = pfind(pid); 980 } 981 if (np == NULL) 982 return EAGAIN; 983 if (np != p) { 984 PROC_UNLOCK(np); 985 return EAGAIN; 986 } 987 PROC_UNLOCK(np); 988 return (0); 989 } 990 991 static int 992 sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 993 { 994 int *name = (int*) arg1; 995 u_int namelen = arg2; 996 struct proc *p; 997 int doingzomb; 998 int error = 0; 999 1000 if (oidp->oid_number == KERN_PROC_PID) { 1001 if (namelen != 1) 1002 return (EINVAL); 1003 p = pfind((pid_t)name[0]); 1004 if (!p) 1005 return (0); 1006 if (p_cansee(curthread, p)) { 1007 PROC_UNLOCK(p); 1008 return (0); 1009 } 1010 error = sysctl_out_proc(p, req, 0); 1011 return (error); 1012 } 1013 if (oidp->oid_number == KERN_PROC_ALL && !namelen) 1014 ; 1015 else if (oidp->oid_number != KERN_PROC_ALL && namelen == 1) 1016 ; 1017 else 1018 return (EINVAL); 1019 1020 if (!req->oldptr) { 1021 /* overestimate by 5 procs */ 1022 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1023 if (error) 1024 return (error); 1025 } 1026 sx_slock(&allproc_lock); 1027 for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) { 1028 if (!doingzomb) 1029 p = LIST_FIRST(&allproc); 1030 else 1031 p = LIST_FIRST(&zombproc); 1032 for (; p != 0; p = LIST_NEXT(p, p_list)) { 1033 PROC_LOCK(p); 1034 /* 1035 * Show a user only appropriate processes. 1036 */ 1037 if (p_cansee(curthread, p)) { 1038 PROC_UNLOCK(p); 1039 continue; 1040 } 1041 /* 1042 * Skip embryonic processes. 1043 */ 1044 if (p->p_state == PRS_NEW) { 1045 PROC_UNLOCK(p); 1046 continue; 1047 } 1048 /* 1049 * TODO - make more efficient (see notes below). 1050 * do by session. 1051 */ 1052 switch (oidp->oid_number) { 1053 1054 case KERN_PROC_PGRP: 1055 /* could do this by traversing pgrp */ 1056 if (p->p_pgrp == NULL || 1057 p->p_pgrp->pg_id != (pid_t)name[0]) { 1058 PROC_UNLOCK(p); 1059 continue; 1060 } 1061 break; 1062 1063 case KERN_PROC_TTY: 1064 if ((p->p_flag & P_CONTROLT) == 0 || 1065 p->p_session == NULL) { 1066 PROC_UNLOCK(p); 1067 continue; 1068 } 1069 SESS_LOCK(p->p_session); 1070 if (p->p_session->s_ttyp == NULL || 1071 dev2udev(p->p_session->s_ttyp->t_dev) != 1072 (udev_t)name[0]) { 1073 SESS_UNLOCK(p->p_session); 1074 PROC_UNLOCK(p); 1075 continue; 1076 } 1077 SESS_UNLOCK(p->p_session); 1078 break; 1079 1080 case KERN_PROC_UID: 1081 if (p->p_ucred == NULL || 1082 p->p_ucred->cr_uid != (uid_t)name[0]) { 1083 PROC_UNLOCK(p); 1084 continue; 1085 } 1086 break; 1087 1088 case KERN_PROC_RUID: 1089 if (p->p_ucred == NULL || 1090 p->p_ucred->cr_ruid != (uid_t)name[0]) { 1091 PROC_UNLOCK(p); 1092 continue; 1093 } 1094 break; 1095 } 1096 1097 error = sysctl_out_proc(p, req, doingzomb); 1098 if (error) { 1099 sx_sunlock(&allproc_lock); 1100 return (error); 1101 } 1102 } 1103 } 1104 sx_sunlock(&allproc_lock); 1105 return (0); 1106 } 1107 1108 struct pargs * 1109 pargs_alloc(int len) 1110 { 1111 struct pargs *pa; 1112 1113 MALLOC(pa, struct pargs *, sizeof(struct pargs) + len, M_PARGS, 1114 M_WAITOK); 1115 pa->ar_ref = 1; 1116 pa->ar_length = len; 1117 return (pa); 1118 } 1119 1120 void 1121 pargs_free(struct pargs *pa) 1122 { 1123 1124 FREE(pa, M_PARGS); 1125 } 1126 1127 void 1128 pargs_hold(struct pargs *pa) 1129 { 1130 1131 if (pa == NULL) 1132 return; 1133 PARGS_LOCK(pa); 1134 pa->ar_ref++; 1135 PARGS_UNLOCK(pa); 1136 } 1137 1138 void 1139 pargs_drop(struct pargs *pa) 1140 { 1141 1142 if (pa == NULL) 1143 return; 1144 PARGS_LOCK(pa); 1145 if (--pa->ar_ref == 0) { 1146 PARGS_UNLOCK(pa); 1147 pargs_free(pa); 1148 } else 1149 PARGS_UNLOCK(pa); 1150 } 1151 1152 /* 1153 * This sysctl allows a process to retrieve the argument list or process 1154 * title for another process without groping around in the address space 1155 * of the other process. It also allow a process to set its own "process 1156 * title to a string of its own choice. 1157 */ 1158 static int 1159 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 1160 { 1161 int *name = (int*) arg1; 1162 u_int namelen = arg2; 1163 struct proc *p; 1164 struct pargs *pa; 1165 int error = 0; 1166 1167 if (namelen != 1) 1168 return (EINVAL); 1169 1170 p = pfind((pid_t)name[0]); 1171 if (!p) 1172 return (0); 1173 1174 if ((!ps_argsopen) && p_cansee(curthread, p)) { 1175 PROC_UNLOCK(p); 1176 return (0); 1177 } 1178 PROC_UNLOCK(p); 1179 1180 if (req->newptr && curproc != p) 1181 return (EPERM); 1182 1183 PROC_LOCK(p); 1184 pa = p->p_args; 1185 pargs_hold(pa); 1186 PROC_UNLOCK(p); 1187 if (req->oldptr && pa != NULL) { 1188 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 1189 } 1190 pargs_drop(pa); 1191 if (req->newptr == NULL) 1192 return (error); 1193 1194 PROC_LOCK(p); 1195 pa = p->p_args; 1196 p->p_args = NULL; 1197 PROC_UNLOCK(p); 1198 pargs_drop(pa); 1199 1200 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit) 1201 return (error); 1202 1203 pa = pargs_alloc(req->newlen); 1204 error = SYSCTL_IN(req, pa->ar_args, req->newlen); 1205 if (!error) { 1206 PROC_LOCK(p); 1207 p->p_args = pa; 1208 PROC_UNLOCK(p); 1209 } else 1210 pargs_free(pa); 1211 return (error); 1212 } 1213 1214 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table"); 1215 1216 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT, 1217 0, 0, sysctl_kern_proc, "S,proc", "Return entire process table"); 1218 1219 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD, 1220 sysctl_kern_proc, "Process table"); 1221 1222 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD, 1223 sysctl_kern_proc, "Process table"); 1224 1225 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD, 1226 sysctl_kern_proc, "Process table"); 1227 1228 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD, 1229 sysctl_kern_proc, "Process table"); 1230 1231 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD, 1232 sysctl_kern_proc, "Process table"); 1233 1234 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY, 1235 sysctl_kern_proc_args, "Process argument list"); 1236 1237 SYSCTL_INT(_kern_proc, OID_AUTO, active, CTLFLAG_RD, 1238 &active_procs, 0, "Number of active procs in system."); 1239 1240 SYSCTL_INT(_kern_proc, OID_AUTO, cached, CTLFLAG_RD, 1241 &cached_procs, 0, "Number of procs in proc cache."); 1242 1243 SYSCTL_INT(_kern_proc, OID_AUTO, allocated, CTLFLAG_RD, 1244 &allocated_procs, 0, "Number of procs in zone."); 1245