kern_proc.c revision 288499
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 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95 30 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: stable/10/sys/kern/kern_proc.c 288499 2015-10-02 14:36:41Z vangyzen $"); 34 35#include "opt_compat.h" 36#include "opt_ddb.h" 37#include "opt_kdtrace.h" 38#include "opt_ktrace.h" 39#include "opt_kstack_pages.h" 40#include "opt_stack.h" 41 42#include <sys/param.h> 43#include <sys/systm.h> 44#include <sys/elf.h> 45#include <sys/exec.h> 46#include <sys/kernel.h> 47#include <sys/limits.h> 48#include <sys/lock.h> 49#include <sys/loginclass.h> 50#include <sys/malloc.h> 51#include <sys/mman.h> 52#include <sys/mount.h> 53#include <sys/mutex.h> 54#include <sys/proc.h> 55#include <sys/ptrace.h> 56#include <sys/refcount.h> 57#include <sys/resourcevar.h> 58#include <sys/rwlock.h> 59#include <sys/sbuf.h> 60#include <sys/sysent.h> 61#include <sys/sched.h> 62#include <sys/smp.h> 63#include <sys/stack.h> 64#include <sys/stat.h> 65#include <sys/sysctl.h> 66#include <sys/filedesc.h> 67#include <sys/tty.h> 68#include <sys/signalvar.h> 69#include <sys/sdt.h> 70#include <sys/sx.h> 71#include <sys/user.h> 72#include <sys/jail.h> 73#include <sys/vnode.h> 74#include <sys/eventhandler.h> 75 76#ifdef DDB 77#include <ddb/ddb.h> 78#endif 79 80#include <vm/vm.h> 81#include <vm/vm_param.h> 82#include <vm/vm_extern.h> 83#include <vm/pmap.h> 84#include <vm/vm_map.h> 85#include <vm/vm_object.h> 86#include <vm/vm_page.h> 87#include <vm/uma.h> 88 89#ifdef COMPAT_FREEBSD32 90#include <compat/freebsd32/freebsd32.h> 91#include <compat/freebsd32/freebsd32_util.h> 92#endif 93 94SDT_PROVIDER_DEFINE(proc); 95SDT_PROBE_DEFINE4(proc, kernel, ctor, entry, "struct proc *", "int", 96 "void *", "int"); 97SDT_PROBE_DEFINE4(proc, kernel, ctor, return, "struct proc *", "int", 98 "void *", "int"); 99SDT_PROBE_DEFINE4(proc, kernel, dtor, entry, "struct proc *", "int", 100 "void *", "struct thread *"); 101SDT_PROBE_DEFINE3(proc, kernel, dtor, return, "struct proc *", "int", 102 "void *"); 103SDT_PROBE_DEFINE3(proc, kernel, init, entry, "struct proc *", "int", 104 "int"); 105SDT_PROBE_DEFINE3(proc, kernel, init, return, "struct proc *", "int", 106 "int"); 107 108MALLOC_DEFINE(M_PGRP, "pgrp", "process group header"); 109MALLOC_DEFINE(M_SESSION, "session", "session header"); 110static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); 111MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 112 113static void doenterpgrp(struct proc *, struct pgrp *); 114static void orphanpg(struct pgrp *pg); 115static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp); 116static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp); 117static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, 118 int preferthread); 119static void pgadjustjobc(struct pgrp *pgrp, int entering); 120static void pgdelete(struct pgrp *); 121static int proc_ctor(void *mem, int size, void *arg, int flags); 122static void proc_dtor(void *mem, int size, void *arg); 123static int proc_init(void *mem, int size, int flags); 124static void proc_fini(void *mem, int size); 125static void pargs_free(struct pargs *pa); 126static struct proc *zpfind_locked(pid_t pid); 127 128/* 129 * Other process lists 130 */ 131struct pidhashhead *pidhashtbl; 132u_long pidhash; 133struct pgrphashhead *pgrphashtbl; 134u_long pgrphash; 135struct proclist allproc; 136struct proclist zombproc; 137struct sx allproc_lock; 138struct sx proctree_lock; 139struct mtx ppeers_lock; 140uma_zone_t proc_zone; 141 142int kstack_pages = KSTACK_PAGES; 143SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, 144 "Kernel stack size in pages"); 145static int vmmap_skip_res_cnt = 0; 146SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW, 147 &vmmap_skip_res_cnt, 0, 148 "Skip calculation of the pages resident count in kern.proc.vmmap"); 149 150CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); 151#ifdef COMPAT_FREEBSD32 152CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE); 153#endif 154 155/* 156 * Initialize global process hashing structures. 157 */ 158void 159procinit() 160{ 161 162 sx_init(&allproc_lock, "allproc"); 163 sx_init(&proctree_lock, "proctree"); 164 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF); 165 LIST_INIT(&allproc); 166 LIST_INIT(&zombproc); 167 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); 168 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); 169 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(), 170 proc_ctor, proc_dtor, proc_init, proc_fini, 171 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 172 uihashinit(); 173} 174 175/* 176 * Prepare a proc for use. 177 */ 178static int 179proc_ctor(void *mem, int size, void *arg, int flags) 180{ 181 struct proc *p; 182 183 p = (struct proc *)mem; 184 SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0); 185 EVENTHANDLER_INVOKE(process_ctor, p); 186 SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0); 187 return (0); 188} 189 190/* 191 * Reclaim a proc after use. 192 */ 193static void 194proc_dtor(void *mem, int size, void *arg) 195{ 196 struct proc *p; 197 struct thread *td; 198 199 /* INVARIANTS checks go here */ 200 p = (struct proc *)mem; 201 td = FIRST_THREAD_IN_PROC(p); 202 SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0); 203 if (td != NULL) { 204#ifdef INVARIANTS 205 KASSERT((p->p_numthreads == 1), 206 ("bad number of threads in exiting process")); 207 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr")); 208#endif 209 /* Free all OSD associated to this thread. */ 210 osd_thread_exit(td); 211 } 212 EVENTHANDLER_INVOKE(process_dtor, p); 213 if (p->p_ksi != NULL) 214 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue")); 215 SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0); 216} 217 218/* 219 * Initialize type-stable parts of a proc (when newly created). 220 */ 221static int 222proc_init(void *mem, int size, int flags) 223{ 224 struct proc *p; 225 226 p = (struct proc *)mem; 227 SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0); 228 p->p_sched = (struct p_sched *)&p[1]; 229 bzero(&p->p_mtx, sizeof(struct mtx)); 230 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK); 231 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE); 232 cv_init(&p->p_pwait, "ppwait"); 233 cv_init(&p->p_dbgwait, "dbgwait"); 234 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 235 EVENTHANDLER_INVOKE(process_init, p); 236 p->p_stats = pstats_alloc(); 237 SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0); 238 return (0); 239} 240 241/* 242 * UMA should ensure that this function is never called. 243 * Freeing a proc structure would violate type stability. 244 */ 245static void 246proc_fini(void *mem, int size) 247{ 248#ifdef notnow 249 struct proc *p; 250 251 p = (struct proc *)mem; 252 EVENTHANDLER_INVOKE(process_fini, p); 253 pstats_free(p->p_stats); 254 thread_free(FIRST_THREAD_IN_PROC(p)); 255 mtx_destroy(&p->p_mtx); 256 if (p->p_ksi != NULL) 257 ksiginfo_free(p->p_ksi); 258#else 259 panic("proc reclaimed"); 260#endif 261} 262 263/* 264 * Is p an inferior of the current process? 265 */ 266int 267inferior(struct proc *p) 268{ 269 270 sx_assert(&proctree_lock, SX_LOCKED); 271 PROC_LOCK_ASSERT(p, MA_OWNED); 272 for (; p != curproc; p = proc_realparent(p)) { 273 if (p->p_pid == 0) 274 return (0); 275 } 276 return (1); 277} 278 279struct proc * 280pfind_locked(pid_t pid) 281{ 282 struct proc *p; 283 284 sx_assert(&allproc_lock, SX_LOCKED); 285 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 286 if (p->p_pid == pid) { 287 PROC_LOCK(p); 288 if (p->p_state == PRS_NEW) { 289 PROC_UNLOCK(p); 290 p = NULL; 291 } 292 break; 293 } 294 } 295 return (p); 296} 297 298/* 299 * Locate a process by number; return only "live" processes -- i.e., neither 300 * zombies nor newly born but incompletely initialized processes. By not 301 * returning processes in the PRS_NEW state, we allow callers to avoid 302 * testing for that condition to avoid dereferencing p_ucred, et al. 303 */ 304struct proc * 305pfind(pid_t pid) 306{ 307 struct proc *p; 308 309 sx_slock(&allproc_lock); 310 p = pfind_locked(pid); 311 sx_sunlock(&allproc_lock); 312 return (p); 313} 314 315static struct proc * 316pfind_tid_locked(pid_t tid) 317{ 318 struct proc *p; 319 struct thread *td; 320 321 sx_assert(&allproc_lock, SX_LOCKED); 322 FOREACH_PROC_IN_SYSTEM(p) { 323 PROC_LOCK(p); 324 if (p->p_state == PRS_NEW) { 325 PROC_UNLOCK(p); 326 continue; 327 } 328 FOREACH_THREAD_IN_PROC(p, td) { 329 if (td->td_tid == tid) 330 goto found; 331 } 332 PROC_UNLOCK(p); 333 } 334found: 335 return (p); 336} 337 338/* 339 * Locate a process group by number. 340 * The caller must hold proctree_lock. 341 */ 342struct pgrp * 343pgfind(pgid) 344 register pid_t pgid; 345{ 346 register struct pgrp *pgrp; 347 348 sx_assert(&proctree_lock, SX_LOCKED); 349 350 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { 351 if (pgrp->pg_id == pgid) { 352 PGRP_LOCK(pgrp); 353 return (pgrp); 354 } 355 } 356 return (NULL); 357} 358 359/* 360 * Locate process and do additional manipulations, depending on flags. 361 */ 362int 363pget(pid_t pid, int flags, struct proc **pp) 364{ 365 struct proc *p; 366 int error; 367 368 sx_slock(&allproc_lock); 369 if (pid <= PID_MAX) { 370 p = pfind_locked(pid); 371 if (p == NULL && (flags & PGET_NOTWEXIT) == 0) 372 p = zpfind_locked(pid); 373 } else if ((flags & PGET_NOTID) == 0) { 374 p = pfind_tid_locked(pid); 375 } else { 376 p = NULL; 377 } 378 sx_sunlock(&allproc_lock); 379 if (p == NULL) 380 return (ESRCH); 381 if ((flags & PGET_CANSEE) != 0) { 382 error = p_cansee(curthread, p); 383 if (error != 0) 384 goto errout; 385 } 386 if ((flags & PGET_CANDEBUG) != 0) { 387 error = p_candebug(curthread, p); 388 if (error != 0) 389 goto errout; 390 } 391 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) { 392 error = EPERM; 393 goto errout; 394 } 395 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) { 396 error = ESRCH; 397 goto errout; 398 } 399 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) { 400 /* 401 * XXXRW: Not clear ESRCH is the right error during proc 402 * execve(). 403 */ 404 error = ESRCH; 405 goto errout; 406 } 407 if ((flags & PGET_HOLD) != 0) { 408 _PHOLD(p); 409 PROC_UNLOCK(p); 410 } 411 *pp = p; 412 return (0); 413errout: 414 PROC_UNLOCK(p); 415 return (error); 416} 417 418/* 419 * Create a new process group. 420 * pgid must be equal to the pid of p. 421 * Begin a new session if required. 422 */ 423int 424enterpgrp(p, pgid, pgrp, sess) 425 register struct proc *p; 426 pid_t pgid; 427 struct pgrp *pgrp; 428 struct session *sess; 429{ 430 431 sx_assert(&proctree_lock, SX_XLOCKED); 432 433 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL")); 434 KASSERT(p->p_pid == pgid, 435 ("enterpgrp: new pgrp and pid != pgid")); 436 KASSERT(pgfind(pgid) == NULL, 437 ("enterpgrp: pgrp with pgid exists")); 438 KASSERT(!SESS_LEADER(p), 439 ("enterpgrp: session leader attempted setpgrp")); 440 441 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 442 443 if (sess != NULL) { 444 /* 445 * new session 446 */ 447 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); 448 PROC_LOCK(p); 449 p->p_flag &= ~P_CONTROLT; 450 PROC_UNLOCK(p); 451 PGRP_LOCK(pgrp); 452 sess->s_leader = p; 453 sess->s_sid = p->p_pid; 454 refcount_init(&sess->s_count, 1); 455 sess->s_ttyvp = NULL; 456 sess->s_ttydp = NULL; 457 sess->s_ttyp = NULL; 458 bcopy(p->p_session->s_login, sess->s_login, 459 sizeof(sess->s_login)); 460 pgrp->pg_session = sess; 461 KASSERT(p == curproc, 462 ("enterpgrp: mksession and p != curproc")); 463 } else { 464 pgrp->pg_session = p->p_session; 465 sess_hold(pgrp->pg_session); 466 PGRP_LOCK(pgrp); 467 } 468 pgrp->pg_id = pgid; 469 LIST_INIT(&pgrp->pg_members); 470 471 /* 472 * As we have an exclusive lock of proctree_lock, 473 * this should not deadlock. 474 */ 475 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); 476 pgrp->pg_jobc = 0; 477 SLIST_INIT(&pgrp->pg_sigiolst); 478 PGRP_UNLOCK(pgrp); 479 480 doenterpgrp(p, pgrp); 481 482 return (0); 483} 484 485/* 486 * Move p to an existing process group 487 */ 488int 489enterthispgrp(p, pgrp) 490 register struct proc *p; 491 struct pgrp *pgrp; 492{ 493 494 sx_assert(&proctree_lock, SX_XLOCKED); 495 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 496 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 497 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 498 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 499 KASSERT(pgrp->pg_session == p->p_session, 500 ("%s: pgrp's session %p, p->p_session %p.\n", 501 __func__, 502 pgrp->pg_session, 503 p->p_session)); 504 KASSERT(pgrp != p->p_pgrp, 505 ("%s: p belongs to pgrp.", __func__)); 506 507 doenterpgrp(p, pgrp); 508 509 return (0); 510} 511 512/* 513 * Move p to a process group 514 */ 515static void 516doenterpgrp(p, pgrp) 517 struct proc *p; 518 struct pgrp *pgrp; 519{ 520 struct pgrp *savepgrp; 521 522 sx_assert(&proctree_lock, SX_XLOCKED); 523 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 524 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 525 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 526 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 527 528 savepgrp = p->p_pgrp; 529 530 /* 531 * Adjust eligibility of affected pgrps to participate in job control. 532 * Increment eligibility counts before decrementing, otherwise we 533 * could reach 0 spuriously during the first call. 534 */ 535 fixjobc(p, pgrp, 1); 536 fixjobc(p, p->p_pgrp, 0); 537 538 PGRP_LOCK(pgrp); 539 PGRP_LOCK(savepgrp); 540 PROC_LOCK(p); 541 LIST_REMOVE(p, p_pglist); 542 p->p_pgrp = pgrp; 543 PROC_UNLOCK(p); 544 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 545 PGRP_UNLOCK(savepgrp); 546 PGRP_UNLOCK(pgrp); 547 if (LIST_EMPTY(&savepgrp->pg_members)) 548 pgdelete(savepgrp); 549} 550 551/* 552 * remove process from process group 553 */ 554int 555leavepgrp(p) 556 register struct proc *p; 557{ 558 struct pgrp *savepgrp; 559 560 sx_assert(&proctree_lock, SX_XLOCKED); 561 savepgrp = p->p_pgrp; 562 PGRP_LOCK(savepgrp); 563 PROC_LOCK(p); 564 LIST_REMOVE(p, p_pglist); 565 p->p_pgrp = NULL; 566 PROC_UNLOCK(p); 567 PGRP_UNLOCK(savepgrp); 568 if (LIST_EMPTY(&savepgrp->pg_members)) 569 pgdelete(savepgrp); 570 return (0); 571} 572 573/* 574 * delete a process group 575 */ 576static void 577pgdelete(pgrp) 578 register struct pgrp *pgrp; 579{ 580 struct session *savesess; 581 struct tty *tp; 582 583 sx_assert(&proctree_lock, SX_XLOCKED); 584 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 585 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 586 587 /* 588 * Reset any sigio structures pointing to us as a result of 589 * F_SETOWN with our pgid. 590 */ 591 funsetownlst(&pgrp->pg_sigiolst); 592 593 PGRP_LOCK(pgrp); 594 tp = pgrp->pg_session->s_ttyp; 595 LIST_REMOVE(pgrp, pg_hash); 596 savesess = pgrp->pg_session; 597 PGRP_UNLOCK(pgrp); 598 599 /* Remove the reference to the pgrp before deallocating it. */ 600 if (tp != NULL) { 601 tty_lock(tp); 602 tty_rel_pgrp(tp, pgrp); 603 } 604 605 mtx_destroy(&pgrp->pg_mtx); 606 free(pgrp, M_PGRP); 607 sess_release(savesess); 608} 609 610static void 611pgadjustjobc(pgrp, entering) 612 struct pgrp *pgrp; 613 int entering; 614{ 615 616 PGRP_LOCK(pgrp); 617 if (entering) 618 pgrp->pg_jobc++; 619 else { 620 --pgrp->pg_jobc; 621 if (pgrp->pg_jobc == 0) 622 orphanpg(pgrp); 623 } 624 PGRP_UNLOCK(pgrp); 625} 626 627/* 628 * Adjust pgrp jobc counters when specified process changes process group. 629 * We count the number of processes in each process group that "qualify" 630 * the group for terminal job control (those with a parent in a different 631 * process group of the same session). If that count reaches zero, the 632 * process group becomes orphaned. Check both the specified process' 633 * process group and that of its children. 634 * entering == 0 => p is leaving specified group. 635 * entering == 1 => p is entering specified group. 636 */ 637void 638fixjobc(p, pgrp, entering) 639 register struct proc *p; 640 register struct pgrp *pgrp; 641 int entering; 642{ 643 register struct pgrp *hispgrp; 644 register struct session *mysession; 645 646 sx_assert(&proctree_lock, SX_LOCKED); 647 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 648 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 649 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 650 651 /* 652 * Check p's parent to see whether p qualifies its own process 653 * group; if so, adjust count for p's process group. 654 */ 655 mysession = pgrp->pg_session; 656 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp && 657 hispgrp->pg_session == mysession) 658 pgadjustjobc(pgrp, entering); 659 660 /* 661 * Check this process' children to see whether they qualify 662 * their process groups; if so, adjust counts for children's 663 * process groups. 664 */ 665 LIST_FOREACH(p, &p->p_children, p_sibling) { 666 hispgrp = p->p_pgrp; 667 if (hispgrp == pgrp || 668 hispgrp->pg_session != mysession) 669 continue; 670 PROC_LOCK(p); 671 if (p->p_state == PRS_ZOMBIE) { 672 PROC_UNLOCK(p); 673 continue; 674 } 675 PROC_UNLOCK(p); 676 pgadjustjobc(hispgrp, entering); 677 } 678} 679 680/* 681 * A process group has become orphaned; 682 * if there are any stopped processes in the group, 683 * hang-up all process in that group. 684 */ 685static void 686orphanpg(pg) 687 struct pgrp *pg; 688{ 689 register struct proc *p; 690 691 PGRP_LOCK_ASSERT(pg, MA_OWNED); 692 693 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 694 PROC_LOCK(p); 695 if (P_SHOULDSTOP(p)) { 696 PROC_UNLOCK(p); 697 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 698 PROC_LOCK(p); 699 kern_psignal(p, SIGHUP); 700 kern_psignal(p, SIGCONT); 701 PROC_UNLOCK(p); 702 } 703 return; 704 } 705 PROC_UNLOCK(p); 706 } 707} 708 709void 710sess_hold(struct session *s) 711{ 712 713 refcount_acquire(&s->s_count); 714} 715 716void 717sess_release(struct session *s) 718{ 719 720 if (refcount_release(&s->s_count)) { 721 if (s->s_ttyp != NULL) { 722 tty_lock(s->s_ttyp); 723 tty_rel_sess(s->s_ttyp, s); 724 } 725 mtx_destroy(&s->s_mtx); 726 free(s, M_SESSION); 727 } 728} 729 730#ifdef DDB 731 732DB_SHOW_COMMAND(pgrpdump, pgrpdump) 733{ 734 register struct pgrp *pgrp; 735 register struct proc *p; 736 register int i; 737 738 for (i = 0; i <= pgrphash; i++) { 739 if (!LIST_EMPTY(&pgrphashtbl[i])) { 740 printf("\tindx %d\n", i); 741 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { 742 printf( 743 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n", 744 (void *)pgrp, (long)pgrp->pg_id, 745 (void *)pgrp->pg_session, 746 pgrp->pg_session->s_count, 747 (void *)LIST_FIRST(&pgrp->pg_members)); 748 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 749 printf("\t\tpid %ld addr %p pgrp %p\n", 750 (long)p->p_pid, (void *)p, 751 (void *)p->p_pgrp); 752 } 753 } 754 } 755 } 756} 757#endif /* DDB */ 758 759/* 760 * Calculate the kinfo_proc members which contain process-wide 761 * informations. 762 * Must be called with the target process locked. 763 */ 764static void 765fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp) 766{ 767 struct thread *td; 768 769 PROC_LOCK_ASSERT(p, MA_OWNED); 770 771 kp->ki_estcpu = 0; 772 kp->ki_pctcpu = 0; 773 FOREACH_THREAD_IN_PROC(p, td) { 774 thread_lock(td); 775 kp->ki_pctcpu += sched_pctcpu(td); 776 kp->ki_estcpu += td->td_estcpu; 777 thread_unlock(td); 778 } 779} 780 781/* 782 * Clear kinfo_proc and fill in any information that is common 783 * to all threads in the process. 784 * Must be called with the target process locked. 785 */ 786static void 787fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp) 788{ 789 struct thread *td0; 790 struct tty *tp; 791 struct session *sp; 792 struct ucred *cred; 793 struct sigacts *ps; 794 795 PROC_LOCK_ASSERT(p, MA_OWNED); 796 bzero(kp, sizeof(*kp)); 797 798 kp->ki_structsize = sizeof(*kp); 799 kp->ki_paddr = p; 800 kp->ki_addr =/* p->p_addr; */0; /* XXX */ 801 kp->ki_args = p->p_args; 802 kp->ki_textvp = p->p_textvp; 803#ifdef KTRACE 804 kp->ki_tracep = p->p_tracevp; 805 kp->ki_traceflag = p->p_traceflag; 806#endif 807 kp->ki_fd = p->p_fd; 808 kp->ki_vmspace = p->p_vmspace; 809 kp->ki_flag = p->p_flag; 810 kp->ki_flag2 = p->p_flag2; 811 cred = p->p_ucred; 812 if (cred) { 813 kp->ki_uid = cred->cr_uid; 814 kp->ki_ruid = cred->cr_ruid; 815 kp->ki_svuid = cred->cr_svuid; 816 kp->ki_cr_flags = 0; 817 if (cred->cr_flags & CRED_FLAG_CAPMODE) 818 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE; 819 /* XXX bde doesn't like KI_NGROUPS */ 820 if (cred->cr_ngroups > KI_NGROUPS) { 821 kp->ki_ngroups = KI_NGROUPS; 822 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW; 823 } else 824 kp->ki_ngroups = cred->cr_ngroups; 825 bcopy(cred->cr_groups, kp->ki_groups, 826 kp->ki_ngroups * sizeof(gid_t)); 827 kp->ki_rgid = cred->cr_rgid; 828 kp->ki_svgid = cred->cr_svgid; 829 /* If jailed(cred), emulate the old P_JAILED flag. */ 830 if (jailed(cred)) { 831 kp->ki_flag |= P_JAILED; 832 /* If inside the jail, use 0 as a jail ID. */ 833 if (cred->cr_prison != curthread->td_ucred->cr_prison) 834 kp->ki_jid = cred->cr_prison->pr_id; 835 } 836 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name, 837 sizeof(kp->ki_loginclass)); 838 } 839 ps = p->p_sigacts; 840 if (ps) { 841 mtx_lock(&ps->ps_mtx); 842 kp->ki_sigignore = ps->ps_sigignore; 843 kp->ki_sigcatch = ps->ps_sigcatch; 844 mtx_unlock(&ps->ps_mtx); 845 } 846 if (p->p_state != PRS_NEW && 847 p->p_state != PRS_ZOMBIE && 848 p->p_vmspace != NULL) { 849 struct vmspace *vm = p->p_vmspace; 850 851 kp->ki_size = vm->vm_map.size; 852 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ 853 FOREACH_THREAD_IN_PROC(p, td0) { 854 if (!TD_IS_SWAPPED(td0)) 855 kp->ki_rssize += td0->td_kstack_pages; 856 } 857 kp->ki_swrss = vm->vm_swrss; 858 kp->ki_tsize = vm->vm_tsize; 859 kp->ki_dsize = vm->vm_dsize; 860 kp->ki_ssize = vm->vm_ssize; 861 } else if (p->p_state == PRS_ZOMBIE) 862 kp->ki_stat = SZOMB; 863 if (kp->ki_flag & P_INMEM) 864 kp->ki_sflag = PS_INMEM; 865 else 866 kp->ki_sflag = 0; 867 /* Calculate legacy swtime as seconds since 'swtick'. */ 868 kp->ki_swtime = (ticks - p->p_swtick) / hz; 869 kp->ki_pid = p->p_pid; 870 kp->ki_nice = p->p_nice; 871 kp->ki_fibnum = p->p_fibnum; 872 kp->ki_start = p->p_stats->p_start; 873 timevaladd(&kp->ki_start, &boottime); 874 PROC_SLOCK(p); 875 rufetch(p, &kp->ki_rusage); 876 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime); 877 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime); 878 PROC_SUNLOCK(p); 879 calccru(p, &kp->ki_childutime, &kp->ki_childstime); 880 /* Some callers want child times in a single value. */ 881 kp->ki_childtime = kp->ki_childstime; 882 timevaladd(&kp->ki_childtime, &kp->ki_childutime); 883 884 FOREACH_THREAD_IN_PROC(p, td0) 885 kp->ki_cow += td0->td_cow; 886 887 tp = NULL; 888 if (p->p_pgrp) { 889 kp->ki_pgid = p->p_pgrp->pg_id; 890 kp->ki_jobc = p->p_pgrp->pg_jobc; 891 sp = p->p_pgrp->pg_session; 892 893 if (sp != NULL) { 894 kp->ki_sid = sp->s_sid; 895 SESS_LOCK(sp); 896 strlcpy(kp->ki_login, sp->s_login, 897 sizeof(kp->ki_login)); 898 if (sp->s_ttyvp) 899 kp->ki_kiflag |= KI_CTTY; 900 if (SESS_LEADER(p)) 901 kp->ki_kiflag |= KI_SLEADER; 902 /* XXX proctree_lock */ 903 tp = sp->s_ttyp; 904 SESS_UNLOCK(sp); 905 } 906 } 907 if ((p->p_flag & P_CONTROLT) && tp != NULL) { 908 kp->ki_tdev = tty_udev(tp); 909 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; 910 if (tp->t_session) 911 kp->ki_tsid = tp->t_session->s_sid; 912 } else 913 kp->ki_tdev = NODEV; 914 if (p->p_comm[0] != '\0') 915 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm)); 916 if (p->p_sysent && p->p_sysent->sv_name != NULL && 917 p->p_sysent->sv_name[0] != '\0') 918 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul)); 919 kp->ki_siglist = p->p_siglist; 920 kp->ki_xstat = p->p_xstat; 921 kp->ki_acflag = p->p_acflag; 922 kp->ki_lock = p->p_lock; 923 if (p->p_pptr) 924 kp->ki_ppid = p->p_pptr->p_pid; 925} 926 927/* 928 * Fill in information that is thread specific. Must be called with 929 * target process locked. If 'preferthread' is set, overwrite certain 930 * process-related fields that are maintained for both threads and 931 * processes. 932 */ 933static void 934fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread) 935{ 936 struct proc *p; 937 938 p = td->td_proc; 939 kp->ki_tdaddr = td; 940 PROC_LOCK_ASSERT(p, MA_OWNED); 941 942 if (preferthread) 943 PROC_SLOCK(p); 944 thread_lock(td); 945 if (td->td_wmesg != NULL) 946 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg)); 947 else 948 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg)); 949 strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)); 950 if (TD_ON_LOCK(td)) { 951 kp->ki_kiflag |= KI_LOCKBLOCK; 952 strlcpy(kp->ki_lockname, td->td_lockname, 953 sizeof(kp->ki_lockname)); 954 } else { 955 kp->ki_kiflag &= ~KI_LOCKBLOCK; 956 bzero(kp->ki_lockname, sizeof(kp->ki_lockname)); 957 } 958 959 if (p->p_state == PRS_NORMAL) { /* approximate. */ 960 if (TD_ON_RUNQ(td) || 961 TD_CAN_RUN(td) || 962 TD_IS_RUNNING(td)) { 963 kp->ki_stat = SRUN; 964 } else if (P_SHOULDSTOP(p)) { 965 kp->ki_stat = SSTOP; 966 } else if (TD_IS_SLEEPING(td)) { 967 kp->ki_stat = SSLEEP; 968 } else if (TD_ON_LOCK(td)) { 969 kp->ki_stat = SLOCK; 970 } else { 971 kp->ki_stat = SWAIT; 972 } 973 } else if (p->p_state == PRS_ZOMBIE) { 974 kp->ki_stat = SZOMB; 975 } else { 976 kp->ki_stat = SIDL; 977 } 978 979 /* Things in the thread */ 980 kp->ki_wchan = td->td_wchan; 981 kp->ki_pri.pri_level = td->td_priority; 982 kp->ki_pri.pri_native = td->td_base_pri; 983 kp->ki_lastcpu = td->td_lastcpu; 984 kp->ki_oncpu = td->td_oncpu; 985 kp->ki_tdflags = td->td_flags; 986 kp->ki_tid = td->td_tid; 987 kp->ki_numthreads = p->p_numthreads; 988 kp->ki_pcb = td->td_pcb; 989 kp->ki_kstack = (void *)td->td_kstack; 990 kp->ki_slptime = (ticks - td->td_slptick) / hz; 991 kp->ki_pri.pri_class = td->td_pri_class; 992 kp->ki_pri.pri_user = td->td_user_pri; 993 994 if (preferthread) { 995 rufetchtd(td, &kp->ki_rusage); 996 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime); 997 kp->ki_pctcpu = sched_pctcpu(td); 998 kp->ki_estcpu = td->td_estcpu; 999 kp->ki_cow = td->td_cow; 1000 } 1001 1002 /* We can't get this anymore but ps etc never used it anyway. */ 1003 kp->ki_rqindex = 0; 1004 1005 if (preferthread) 1006 kp->ki_siglist = td->td_siglist; 1007 kp->ki_sigmask = td->td_sigmask; 1008 thread_unlock(td); 1009 if (preferthread) 1010 PROC_SUNLOCK(p); 1011} 1012 1013/* 1014 * Fill in a kinfo_proc structure for the specified process. 1015 * Must be called with the target process locked. 1016 */ 1017void 1018fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp) 1019{ 1020 1021 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1022 1023 fill_kinfo_proc_only(p, kp); 1024 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0); 1025 fill_kinfo_aggregate(p, kp); 1026} 1027 1028struct pstats * 1029pstats_alloc(void) 1030{ 1031 1032 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK)); 1033} 1034 1035/* 1036 * Copy parts of p_stats; zero the rest of p_stats (statistics). 1037 */ 1038void 1039pstats_fork(struct pstats *src, struct pstats *dst) 1040{ 1041 1042 bzero(&dst->pstat_startzero, 1043 __rangeof(struct pstats, pstat_startzero, pstat_endzero)); 1044 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy, 1045 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy)); 1046} 1047 1048void 1049pstats_free(struct pstats *ps) 1050{ 1051 1052 free(ps, M_SUBPROC); 1053} 1054 1055static struct proc * 1056zpfind_locked(pid_t pid) 1057{ 1058 struct proc *p; 1059 1060 sx_assert(&allproc_lock, SX_LOCKED); 1061 LIST_FOREACH(p, &zombproc, p_list) { 1062 if (p->p_pid == pid) { 1063 PROC_LOCK(p); 1064 break; 1065 } 1066 } 1067 return (p); 1068} 1069 1070/* 1071 * Locate a zombie process by number 1072 */ 1073struct proc * 1074zpfind(pid_t pid) 1075{ 1076 struct proc *p; 1077 1078 sx_slock(&allproc_lock); 1079 p = zpfind_locked(pid); 1080 sx_sunlock(&allproc_lock); 1081 return (p); 1082} 1083 1084#ifdef COMPAT_FREEBSD32 1085 1086/* 1087 * This function is typically used to copy out the kernel address, so 1088 * it can be replaced by assignment of zero. 1089 */ 1090static inline uint32_t 1091ptr32_trim(void *ptr) 1092{ 1093 uintptr_t uptr; 1094 1095 uptr = (uintptr_t)ptr; 1096 return ((uptr > UINT_MAX) ? 0 : uptr); 1097} 1098 1099#define PTRTRIM_CP(src,dst,fld) \ 1100 do { (dst).fld = ptr32_trim((src).fld); } while (0) 1101 1102static void 1103freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32) 1104{ 1105 int i; 1106 1107 bzero(ki32, sizeof(struct kinfo_proc32)); 1108 ki32->ki_structsize = sizeof(struct kinfo_proc32); 1109 CP(*ki, *ki32, ki_layout); 1110 PTRTRIM_CP(*ki, *ki32, ki_args); 1111 PTRTRIM_CP(*ki, *ki32, ki_paddr); 1112 PTRTRIM_CP(*ki, *ki32, ki_addr); 1113 PTRTRIM_CP(*ki, *ki32, ki_tracep); 1114 PTRTRIM_CP(*ki, *ki32, ki_textvp); 1115 PTRTRIM_CP(*ki, *ki32, ki_fd); 1116 PTRTRIM_CP(*ki, *ki32, ki_vmspace); 1117 PTRTRIM_CP(*ki, *ki32, ki_wchan); 1118 CP(*ki, *ki32, ki_pid); 1119 CP(*ki, *ki32, ki_ppid); 1120 CP(*ki, *ki32, ki_pgid); 1121 CP(*ki, *ki32, ki_tpgid); 1122 CP(*ki, *ki32, ki_sid); 1123 CP(*ki, *ki32, ki_tsid); 1124 CP(*ki, *ki32, ki_jobc); 1125 CP(*ki, *ki32, ki_tdev); 1126 CP(*ki, *ki32, ki_siglist); 1127 CP(*ki, *ki32, ki_sigmask); 1128 CP(*ki, *ki32, ki_sigignore); 1129 CP(*ki, *ki32, ki_sigcatch); 1130 CP(*ki, *ki32, ki_uid); 1131 CP(*ki, *ki32, ki_ruid); 1132 CP(*ki, *ki32, ki_svuid); 1133 CP(*ki, *ki32, ki_rgid); 1134 CP(*ki, *ki32, ki_svgid); 1135 CP(*ki, *ki32, ki_ngroups); 1136 for (i = 0; i < KI_NGROUPS; i++) 1137 CP(*ki, *ki32, ki_groups[i]); 1138 CP(*ki, *ki32, ki_size); 1139 CP(*ki, *ki32, ki_rssize); 1140 CP(*ki, *ki32, ki_swrss); 1141 CP(*ki, *ki32, ki_tsize); 1142 CP(*ki, *ki32, ki_dsize); 1143 CP(*ki, *ki32, ki_ssize); 1144 CP(*ki, *ki32, ki_xstat); 1145 CP(*ki, *ki32, ki_acflag); 1146 CP(*ki, *ki32, ki_pctcpu); 1147 CP(*ki, *ki32, ki_estcpu); 1148 CP(*ki, *ki32, ki_slptime); 1149 CP(*ki, *ki32, ki_swtime); 1150 CP(*ki, *ki32, ki_cow); 1151 CP(*ki, *ki32, ki_runtime); 1152 TV_CP(*ki, *ki32, ki_start); 1153 TV_CP(*ki, *ki32, ki_childtime); 1154 CP(*ki, *ki32, ki_flag); 1155 CP(*ki, *ki32, ki_kiflag); 1156 CP(*ki, *ki32, ki_traceflag); 1157 CP(*ki, *ki32, ki_stat); 1158 CP(*ki, *ki32, ki_nice); 1159 CP(*ki, *ki32, ki_lock); 1160 CP(*ki, *ki32, ki_rqindex); 1161 CP(*ki, *ki32, ki_oncpu); 1162 CP(*ki, *ki32, ki_lastcpu); 1163 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1); 1164 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1); 1165 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1); 1166 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1); 1167 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1); 1168 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1); 1169 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1); 1170 CP(*ki, *ki32, ki_flag2); 1171 CP(*ki, *ki32, ki_fibnum); 1172 CP(*ki, *ki32, ki_cr_flags); 1173 CP(*ki, *ki32, ki_jid); 1174 CP(*ki, *ki32, ki_numthreads); 1175 CP(*ki, *ki32, ki_tid); 1176 CP(*ki, *ki32, ki_pri); 1177 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage); 1178 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch); 1179 PTRTRIM_CP(*ki, *ki32, ki_pcb); 1180 PTRTRIM_CP(*ki, *ki32, ki_kstack); 1181 PTRTRIM_CP(*ki, *ki32, ki_udata); 1182 CP(*ki, *ki32, ki_sflag); 1183 CP(*ki, *ki32, ki_tdflags); 1184} 1185#endif 1186 1187int 1188kern_proc_out(struct proc *p, struct sbuf *sb, int flags) 1189{ 1190 struct thread *td; 1191 struct kinfo_proc ki; 1192#ifdef COMPAT_FREEBSD32 1193 struct kinfo_proc32 ki32; 1194#endif 1195 int error; 1196 1197 PROC_LOCK_ASSERT(p, MA_OWNED); 1198 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1199 1200 error = 0; 1201 fill_kinfo_proc(p, &ki); 1202 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1203#ifdef COMPAT_FREEBSD32 1204 if ((flags & KERN_PROC_MASK32) != 0) { 1205 freebsd32_kinfo_proc_out(&ki, &ki32); 1206 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1207 error = ENOMEM; 1208 } else 1209#endif 1210 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1211 error = ENOMEM; 1212 } else { 1213 FOREACH_THREAD_IN_PROC(p, td) { 1214 fill_kinfo_thread(td, &ki, 1); 1215#ifdef COMPAT_FREEBSD32 1216 if ((flags & KERN_PROC_MASK32) != 0) { 1217 freebsd32_kinfo_proc_out(&ki, &ki32); 1218 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1219 error = ENOMEM; 1220 } else 1221#endif 1222 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1223 error = ENOMEM; 1224 if (error != 0) 1225 break; 1226 } 1227 } 1228 PROC_UNLOCK(p); 1229 return (error); 1230} 1231 1232static int 1233sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags, 1234 int doingzomb) 1235{ 1236 struct sbuf sb; 1237 struct kinfo_proc ki; 1238 struct proc *np; 1239 int error, error2; 1240 pid_t pid; 1241 1242 pid = p->p_pid; 1243 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); 1244 error = kern_proc_out(p, &sb, flags); 1245 error2 = sbuf_finish(&sb); 1246 sbuf_delete(&sb); 1247 if (error != 0) 1248 return (error); 1249 else if (error2 != 0) 1250 return (error2); 1251 if (doingzomb) 1252 np = zpfind(pid); 1253 else { 1254 if (pid == 0) 1255 return (0); 1256 np = pfind(pid); 1257 } 1258 if (np == NULL) 1259 return (ESRCH); 1260 if (np != p) { 1261 PROC_UNLOCK(np); 1262 return (ESRCH); 1263 } 1264 PROC_UNLOCK(np); 1265 return (0); 1266} 1267 1268static int 1269sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 1270{ 1271 int *name = (int *)arg1; 1272 u_int namelen = arg2; 1273 struct proc *p; 1274 int flags, doingzomb, oid_number; 1275 int error = 0; 1276 1277 oid_number = oidp->oid_number; 1278 if (oid_number != KERN_PROC_ALL && 1279 (oid_number & KERN_PROC_INC_THREAD) == 0) 1280 flags = KERN_PROC_NOTHREADS; 1281 else { 1282 flags = 0; 1283 oid_number &= ~KERN_PROC_INC_THREAD; 1284 } 1285#ifdef COMPAT_FREEBSD32 1286 if (req->flags & SCTL_MASK32) 1287 flags |= KERN_PROC_MASK32; 1288#endif 1289 if (oid_number == KERN_PROC_PID) { 1290 if (namelen != 1) 1291 return (EINVAL); 1292 error = sysctl_wire_old_buffer(req, 0); 1293 if (error) 1294 return (error); 1295 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1296 if (error != 0) 1297 return (error); 1298 error = sysctl_out_proc(p, req, flags, 0); 1299 return (error); 1300 } 1301 1302 switch (oid_number) { 1303 case KERN_PROC_ALL: 1304 if (namelen != 0) 1305 return (EINVAL); 1306 break; 1307 case KERN_PROC_PROC: 1308 if (namelen != 0 && namelen != 1) 1309 return (EINVAL); 1310 break; 1311 default: 1312 if (namelen != 1) 1313 return (EINVAL); 1314 break; 1315 } 1316 1317 if (!req->oldptr) { 1318 /* overestimate by 5 procs */ 1319 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1320 if (error) 1321 return (error); 1322 } 1323 error = sysctl_wire_old_buffer(req, 0); 1324 if (error != 0) 1325 return (error); 1326 sx_slock(&allproc_lock); 1327 for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) { 1328 if (!doingzomb) 1329 p = LIST_FIRST(&allproc); 1330 else 1331 p = LIST_FIRST(&zombproc); 1332 for (; p != 0; p = LIST_NEXT(p, p_list)) { 1333 /* 1334 * Skip embryonic processes. 1335 */ 1336 PROC_LOCK(p); 1337 if (p->p_state == PRS_NEW) { 1338 PROC_UNLOCK(p); 1339 continue; 1340 } 1341 KASSERT(p->p_ucred != NULL, 1342 ("process credential is NULL for non-NEW proc")); 1343 /* 1344 * Show a user only appropriate processes. 1345 */ 1346 if (p_cansee(curthread, p)) { 1347 PROC_UNLOCK(p); 1348 continue; 1349 } 1350 /* 1351 * TODO - make more efficient (see notes below). 1352 * do by session. 1353 */ 1354 switch (oid_number) { 1355 1356 case KERN_PROC_GID: 1357 if (p->p_ucred->cr_gid != (gid_t)name[0]) { 1358 PROC_UNLOCK(p); 1359 continue; 1360 } 1361 break; 1362 1363 case KERN_PROC_PGRP: 1364 /* could do this by traversing pgrp */ 1365 if (p->p_pgrp == NULL || 1366 p->p_pgrp->pg_id != (pid_t)name[0]) { 1367 PROC_UNLOCK(p); 1368 continue; 1369 } 1370 break; 1371 1372 case KERN_PROC_RGID: 1373 if (p->p_ucred->cr_rgid != (gid_t)name[0]) { 1374 PROC_UNLOCK(p); 1375 continue; 1376 } 1377 break; 1378 1379 case KERN_PROC_SESSION: 1380 if (p->p_session == NULL || 1381 p->p_session->s_sid != (pid_t)name[0]) { 1382 PROC_UNLOCK(p); 1383 continue; 1384 } 1385 break; 1386 1387 case KERN_PROC_TTY: 1388 if ((p->p_flag & P_CONTROLT) == 0 || 1389 p->p_session == NULL) { 1390 PROC_UNLOCK(p); 1391 continue; 1392 } 1393 /* XXX proctree_lock */ 1394 SESS_LOCK(p->p_session); 1395 if (p->p_session->s_ttyp == NULL || 1396 tty_udev(p->p_session->s_ttyp) != 1397 (dev_t)name[0]) { 1398 SESS_UNLOCK(p->p_session); 1399 PROC_UNLOCK(p); 1400 continue; 1401 } 1402 SESS_UNLOCK(p->p_session); 1403 break; 1404 1405 case KERN_PROC_UID: 1406 if (p->p_ucred->cr_uid != (uid_t)name[0]) { 1407 PROC_UNLOCK(p); 1408 continue; 1409 } 1410 break; 1411 1412 case KERN_PROC_RUID: 1413 if (p->p_ucred->cr_ruid != (uid_t)name[0]) { 1414 PROC_UNLOCK(p); 1415 continue; 1416 } 1417 break; 1418 1419 case KERN_PROC_PROC: 1420 break; 1421 1422 default: 1423 break; 1424 1425 } 1426 1427 error = sysctl_out_proc(p, req, flags, doingzomb); 1428 if (error) { 1429 sx_sunlock(&allproc_lock); 1430 return (error); 1431 } 1432 } 1433 } 1434 sx_sunlock(&allproc_lock); 1435 return (0); 1436} 1437 1438struct pargs * 1439pargs_alloc(int len) 1440{ 1441 struct pargs *pa; 1442 1443 pa = malloc(sizeof(struct pargs) + len, M_PARGS, 1444 M_WAITOK); 1445 refcount_init(&pa->ar_ref, 1); 1446 pa->ar_length = len; 1447 return (pa); 1448} 1449 1450static void 1451pargs_free(struct pargs *pa) 1452{ 1453 1454 free(pa, M_PARGS); 1455} 1456 1457void 1458pargs_hold(struct pargs *pa) 1459{ 1460 1461 if (pa == NULL) 1462 return; 1463 refcount_acquire(&pa->ar_ref); 1464} 1465 1466void 1467pargs_drop(struct pargs *pa) 1468{ 1469 1470 if (pa == NULL) 1471 return; 1472 if (refcount_release(&pa->ar_ref)) 1473 pargs_free(pa); 1474} 1475 1476static int 1477proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf, 1478 size_t len) 1479{ 1480 struct iovec iov; 1481 struct uio uio; 1482 1483 iov.iov_base = (caddr_t)buf; 1484 iov.iov_len = len; 1485 uio.uio_iov = &iov; 1486 uio.uio_iovcnt = 1; 1487 uio.uio_offset = offset; 1488 uio.uio_resid = (ssize_t)len; 1489 uio.uio_segflg = UIO_SYSSPACE; 1490 uio.uio_rw = UIO_READ; 1491 uio.uio_td = td; 1492 1493 return (proc_rwmem(p, &uio)); 1494} 1495 1496static int 1497proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, 1498 size_t len) 1499{ 1500 size_t i; 1501 int error; 1502 1503 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len); 1504 /* 1505 * Reading the chunk may validly return EFAULT if the string is shorter 1506 * than the chunk and is aligned at the end of the page, assuming the 1507 * next page is not mapped. So if EFAULT is returned do a fallback to 1508 * one byte read loop. 1509 */ 1510 if (error == EFAULT) { 1511 for (i = 0; i < len; i++, buf++, sptr++) { 1512 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1); 1513 if (error != 0) 1514 return (error); 1515 if (*buf == '\0') 1516 break; 1517 } 1518 error = 0; 1519 } 1520 return (error); 1521} 1522 1523#define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ 1524 1525enum proc_vector_type { 1526 PROC_ARG, 1527 PROC_ENV, 1528 PROC_AUX, 1529}; 1530 1531#ifdef COMPAT_FREEBSD32 1532static int 1533get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, 1534 size_t *vsizep, enum proc_vector_type type) 1535{ 1536 struct freebsd32_ps_strings pss; 1537 Elf32_Auxinfo aux; 1538 vm_offset_t vptr, ptr; 1539 uint32_t *proc_vector32; 1540 char **proc_vector; 1541 size_t vsize, size; 1542 int i, error; 1543 1544 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), 1545 &pss, sizeof(pss)); 1546 if (error != 0) 1547 return (error); 1548 switch (type) { 1549 case PROC_ARG: 1550 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); 1551 vsize = pss.ps_nargvstr; 1552 if (vsize > ARG_MAX) 1553 return (ENOEXEC); 1554 size = vsize * sizeof(int32_t); 1555 break; 1556 case PROC_ENV: 1557 vptr = (vm_offset_t)PTRIN(pss.ps_envstr); 1558 vsize = pss.ps_nenvstr; 1559 if (vsize > ARG_MAX) 1560 return (ENOEXEC); 1561 size = vsize * sizeof(int32_t); 1562 break; 1563 case PROC_AUX: 1564 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + 1565 (pss.ps_nenvstr + 1) * sizeof(int32_t); 1566 if (vptr % 4 != 0) 1567 return (ENOEXEC); 1568 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1569 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); 1570 if (error != 0) 1571 return (error); 1572 if (aux.a_type == AT_NULL) 1573 break; 1574 ptr += sizeof(aux); 1575 } 1576 if (aux.a_type != AT_NULL) 1577 return (ENOEXEC); 1578 vsize = i + 1; 1579 size = vsize * sizeof(aux); 1580 break; 1581 default: 1582 KASSERT(0, ("Wrong proc vector type: %d", type)); 1583 return (EINVAL); 1584 } 1585 proc_vector32 = malloc(size, M_TEMP, M_WAITOK); 1586 error = proc_read_mem(td, p, vptr, proc_vector32, size); 1587 if (error != 0) 1588 goto done; 1589 if (type == PROC_AUX) { 1590 *proc_vectorp = (char **)proc_vector32; 1591 *vsizep = vsize; 1592 return (0); 1593 } 1594 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); 1595 for (i = 0; i < (int)vsize; i++) 1596 proc_vector[i] = PTRIN(proc_vector32[i]); 1597 *proc_vectorp = proc_vector; 1598 *vsizep = vsize; 1599done: 1600 free(proc_vector32, M_TEMP); 1601 return (error); 1602} 1603#endif 1604 1605static int 1606get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, 1607 size_t *vsizep, enum proc_vector_type type) 1608{ 1609 struct ps_strings pss; 1610 Elf_Auxinfo aux; 1611 vm_offset_t vptr, ptr; 1612 char **proc_vector; 1613 size_t vsize, size; 1614 int error, i; 1615 1616#ifdef COMPAT_FREEBSD32 1617 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1618 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); 1619#endif 1620 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), 1621 &pss, sizeof(pss)); 1622 if (error != 0) 1623 return (error); 1624 switch (type) { 1625 case PROC_ARG: 1626 vptr = (vm_offset_t)pss.ps_argvstr; 1627 vsize = pss.ps_nargvstr; 1628 if (vsize > ARG_MAX) 1629 return (ENOEXEC); 1630 size = vsize * sizeof(char *); 1631 break; 1632 case PROC_ENV: 1633 vptr = (vm_offset_t)pss.ps_envstr; 1634 vsize = pss.ps_nenvstr; 1635 if (vsize > ARG_MAX) 1636 return (ENOEXEC); 1637 size = vsize * sizeof(char *); 1638 break; 1639 case PROC_AUX: 1640 /* 1641 * The aux array is just above env array on the stack. Check 1642 * that the address is naturally aligned. 1643 */ 1644 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) 1645 * sizeof(char *); 1646#if __ELF_WORD_SIZE == 64 1647 if (vptr % sizeof(uint64_t) != 0) 1648#else 1649 if (vptr % sizeof(uint32_t) != 0) 1650#endif 1651 return (ENOEXEC); 1652 /* 1653 * We count the array size reading the aux vectors from the 1654 * stack until AT_NULL vector is returned. So (to keep the code 1655 * simple) we read the process stack twice: the first time here 1656 * to find the size and the second time when copying the vectors 1657 * to the allocated proc_vector. 1658 */ 1659 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1660 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); 1661 if (error != 0) 1662 return (error); 1663 if (aux.a_type == AT_NULL) 1664 break; 1665 ptr += sizeof(aux); 1666 } 1667 /* 1668 * If the PROC_AUXV_MAX entries are iterated over, and we have 1669 * not reached AT_NULL, it is most likely we are reading wrong 1670 * data: either the process doesn't have auxv array or data has 1671 * been modified. Return the error in this case. 1672 */ 1673 if (aux.a_type != AT_NULL) 1674 return (ENOEXEC); 1675 vsize = i + 1; 1676 size = vsize * sizeof(aux); 1677 break; 1678 default: 1679 KASSERT(0, ("Wrong proc vector type: %d", type)); 1680 return (EINVAL); /* In case we are built without INVARIANTS. */ 1681 } 1682 proc_vector = malloc(size, M_TEMP, M_WAITOK); 1683 if (proc_vector == NULL) 1684 return (ENOMEM); 1685 error = proc_read_mem(td, p, vptr, proc_vector, size); 1686 if (error != 0) { 1687 free(proc_vector, M_TEMP); 1688 return (error); 1689 } 1690 *proc_vectorp = proc_vector; 1691 *vsizep = vsize; 1692 1693 return (0); 1694} 1695 1696#define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ 1697 1698static int 1699get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, 1700 enum proc_vector_type type) 1701{ 1702 size_t done, len, nchr, vsize; 1703 int error, i; 1704 char **proc_vector, *sptr; 1705 char pss_string[GET_PS_STRINGS_CHUNK_SZ]; 1706 1707 PROC_ASSERT_HELD(p); 1708 1709 /* 1710 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. 1711 */ 1712 nchr = 2 * (PATH_MAX + ARG_MAX); 1713 1714 error = get_proc_vector(td, p, &proc_vector, &vsize, type); 1715 if (error != 0) 1716 return (error); 1717 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { 1718 /* 1719 * The program may have scribbled into its argv array, e.g. to 1720 * remove some arguments. If that has happened, break out 1721 * before trying to read from NULL. 1722 */ 1723 if (proc_vector[i] == NULL) 1724 break; 1725 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { 1726 error = proc_read_string(td, p, sptr, pss_string, 1727 sizeof(pss_string)); 1728 if (error != 0) 1729 goto done; 1730 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); 1731 if (done + len >= nchr) 1732 len = nchr - done - 1; 1733 sbuf_bcat(sb, pss_string, len); 1734 if (len != GET_PS_STRINGS_CHUNK_SZ) 1735 break; 1736 done += GET_PS_STRINGS_CHUNK_SZ; 1737 } 1738 sbuf_bcat(sb, "", 1); 1739 done += len + 1; 1740 } 1741done: 1742 free(proc_vector, M_TEMP); 1743 return (error); 1744} 1745 1746int 1747proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) 1748{ 1749 1750 return (get_ps_strings(curthread, p, sb, PROC_ARG)); 1751} 1752 1753int 1754proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) 1755{ 1756 1757 return (get_ps_strings(curthread, p, sb, PROC_ENV)); 1758} 1759 1760int 1761proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) 1762{ 1763 size_t vsize, size; 1764 char **auxv; 1765 int error; 1766 1767 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); 1768 if (error == 0) { 1769#ifdef COMPAT_FREEBSD32 1770 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1771 size = vsize * sizeof(Elf32_Auxinfo); 1772 else 1773#endif 1774 size = vsize * sizeof(Elf_Auxinfo); 1775 if (sbuf_bcat(sb, auxv, size) != 0) 1776 error = ENOMEM; 1777 free(auxv, M_TEMP); 1778 } 1779 return (error); 1780} 1781 1782/* 1783 * This sysctl allows a process to retrieve the argument list or process 1784 * title for another process without groping around in the address space 1785 * of the other process. It also allow a process to set its own "process 1786 * title to a string of its own choice. 1787 */ 1788static int 1789sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 1790{ 1791 int *name = (int *)arg1; 1792 u_int namelen = arg2; 1793 struct pargs *newpa, *pa; 1794 struct proc *p; 1795 struct sbuf sb; 1796 int flags, error = 0, error2; 1797 1798 if (namelen != 1) 1799 return (EINVAL); 1800 1801 flags = PGET_CANSEE; 1802 if (req->newptr != NULL) 1803 flags |= PGET_ISCURRENT; 1804 error = pget((pid_t)name[0], flags, &p); 1805 if (error) 1806 return (error); 1807 1808 pa = p->p_args; 1809 if (pa != NULL) { 1810 pargs_hold(pa); 1811 PROC_UNLOCK(p); 1812 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 1813 pargs_drop(pa); 1814 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { 1815 _PHOLD(p); 1816 PROC_UNLOCK(p); 1817 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1818 error = proc_getargv(curthread, p, &sb); 1819 error2 = sbuf_finish(&sb); 1820 PRELE(p); 1821 sbuf_delete(&sb); 1822 if (error == 0 && error2 != 0) 1823 error = error2; 1824 } else { 1825 PROC_UNLOCK(p); 1826 } 1827 if (error != 0 || req->newptr == NULL) 1828 return (error); 1829 1830 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit) 1831 return (ENOMEM); 1832 newpa = pargs_alloc(req->newlen); 1833 error = SYSCTL_IN(req, newpa->ar_args, req->newlen); 1834 if (error != 0) { 1835 pargs_free(newpa); 1836 return (error); 1837 } 1838 PROC_LOCK(p); 1839 pa = p->p_args; 1840 p->p_args = newpa; 1841 PROC_UNLOCK(p); 1842 pargs_drop(pa); 1843 return (0); 1844} 1845 1846/* 1847 * This sysctl allows a process to retrieve environment of another process. 1848 */ 1849static int 1850sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) 1851{ 1852 int *name = (int *)arg1; 1853 u_int namelen = arg2; 1854 struct proc *p; 1855 struct sbuf sb; 1856 int error, error2; 1857 1858 if (namelen != 1) 1859 return (EINVAL); 1860 1861 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1862 if (error != 0) 1863 return (error); 1864 if ((p->p_flag & P_SYSTEM) != 0) { 1865 PRELE(p); 1866 return (0); 1867 } 1868 1869 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1870 error = proc_getenvv(curthread, p, &sb); 1871 error2 = sbuf_finish(&sb); 1872 PRELE(p); 1873 sbuf_delete(&sb); 1874 return (error != 0 ? error : error2); 1875} 1876 1877/* 1878 * This sysctl allows a process to retrieve ELF auxiliary vector of 1879 * another process. 1880 */ 1881static int 1882sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) 1883{ 1884 int *name = (int *)arg1; 1885 u_int namelen = arg2; 1886 struct proc *p; 1887 struct sbuf sb; 1888 int error, error2; 1889 1890 if (namelen != 1) 1891 return (EINVAL); 1892 1893 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1894 if (error != 0) 1895 return (error); 1896 if ((p->p_flag & P_SYSTEM) != 0) { 1897 PRELE(p); 1898 return (0); 1899 } 1900 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1901 error = proc_getauxv(curthread, p, &sb); 1902 error2 = sbuf_finish(&sb); 1903 PRELE(p); 1904 sbuf_delete(&sb); 1905 return (error != 0 ? error : error2); 1906} 1907 1908/* 1909 * This sysctl allows a process to retrieve the path of the executable for 1910 * itself or another process. 1911 */ 1912static int 1913sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) 1914{ 1915 pid_t *pidp = (pid_t *)arg1; 1916 unsigned int arglen = arg2; 1917 struct proc *p; 1918 struct vnode *vp; 1919 char *retbuf, *freebuf; 1920 int error; 1921 1922 if (arglen != 1) 1923 return (EINVAL); 1924 if (*pidp == -1) { /* -1 means this process */ 1925 p = req->td->td_proc; 1926 } else { 1927 error = pget(*pidp, PGET_CANSEE, &p); 1928 if (error != 0) 1929 return (error); 1930 } 1931 1932 vp = p->p_textvp; 1933 if (vp == NULL) { 1934 if (*pidp != -1) 1935 PROC_UNLOCK(p); 1936 return (0); 1937 } 1938 vref(vp); 1939 if (*pidp != -1) 1940 PROC_UNLOCK(p); 1941 error = vn_fullpath(req->td, vp, &retbuf, &freebuf); 1942 vrele(vp); 1943 if (error) 1944 return (error); 1945 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); 1946 free(freebuf, M_TEMP); 1947 return (error); 1948} 1949 1950static int 1951sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) 1952{ 1953 struct proc *p; 1954 char *sv_name; 1955 int *name; 1956 int namelen; 1957 int error; 1958 1959 namelen = arg2; 1960 if (namelen != 1) 1961 return (EINVAL); 1962 1963 name = (int *)arg1; 1964 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1965 if (error != 0) 1966 return (error); 1967 sv_name = p->p_sysent->sv_name; 1968 PROC_UNLOCK(p); 1969 return (sysctl_handle_string(oidp, sv_name, 0, req)); 1970} 1971 1972#ifdef KINFO_OVMENTRY_SIZE 1973CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); 1974#endif 1975 1976#ifdef COMPAT_FREEBSD7 1977static int 1978sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) 1979{ 1980 vm_map_entry_t entry, tmp_entry; 1981 unsigned int last_timestamp; 1982 char *fullpath, *freepath; 1983 struct kinfo_ovmentry *kve; 1984 struct vattr va; 1985 struct ucred *cred; 1986 int error, *name; 1987 struct vnode *vp; 1988 struct proc *p; 1989 vm_map_t map; 1990 struct vmspace *vm; 1991 1992 name = (int *)arg1; 1993 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1994 if (error != 0) 1995 return (error); 1996 vm = vmspace_acquire_ref(p); 1997 if (vm == NULL) { 1998 PRELE(p); 1999 return (ESRCH); 2000 } 2001 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2002 2003 map = &vm->vm_map; 2004 vm_map_lock_read(map); 2005 for (entry = map->header.next; entry != &map->header; 2006 entry = entry->next) { 2007 vm_object_t obj, tobj, lobj; 2008 vm_offset_t addr; 2009 2010 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2011 continue; 2012 2013 bzero(kve, sizeof(*kve)); 2014 kve->kve_structsize = sizeof(*kve); 2015 2016 kve->kve_private_resident = 0; 2017 obj = entry->object.vm_object; 2018 if (obj != NULL) { 2019 VM_OBJECT_RLOCK(obj); 2020 if (obj->shadow_count == 1) 2021 kve->kve_private_resident = 2022 obj->resident_page_count; 2023 } 2024 kve->kve_resident = 0; 2025 addr = entry->start; 2026 while (addr < entry->end) { 2027 if (pmap_extract(map->pmap, addr)) 2028 kve->kve_resident++; 2029 addr += PAGE_SIZE; 2030 } 2031 2032 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { 2033 if (tobj != obj) 2034 VM_OBJECT_RLOCK(tobj); 2035 if (lobj != obj) 2036 VM_OBJECT_RUNLOCK(lobj); 2037 lobj = tobj; 2038 } 2039 2040 kve->kve_start = (void*)entry->start; 2041 kve->kve_end = (void*)entry->end; 2042 kve->kve_offset = (off_t)entry->offset; 2043 2044 if (entry->protection & VM_PROT_READ) 2045 kve->kve_protection |= KVME_PROT_READ; 2046 if (entry->protection & VM_PROT_WRITE) 2047 kve->kve_protection |= KVME_PROT_WRITE; 2048 if (entry->protection & VM_PROT_EXECUTE) 2049 kve->kve_protection |= KVME_PROT_EXEC; 2050 2051 if (entry->eflags & MAP_ENTRY_COW) 2052 kve->kve_flags |= KVME_FLAG_COW; 2053 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2054 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2055 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2056 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2057 2058 last_timestamp = map->timestamp; 2059 vm_map_unlock_read(map); 2060 2061 kve->kve_fileid = 0; 2062 kve->kve_fsid = 0; 2063 freepath = NULL; 2064 fullpath = ""; 2065 if (lobj) { 2066 vp = NULL; 2067 switch (lobj->type) { 2068 case OBJT_DEFAULT: 2069 kve->kve_type = KVME_TYPE_DEFAULT; 2070 break; 2071 case OBJT_VNODE: 2072 kve->kve_type = KVME_TYPE_VNODE; 2073 vp = lobj->handle; 2074 vref(vp); 2075 break; 2076 case OBJT_SWAP: 2077 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) { 2078 kve->kve_type = KVME_TYPE_VNODE; 2079 if ((lobj->flags & OBJ_TMPFS) != 0) { 2080 vp = lobj->un_pager.swp.swp_tmpfs; 2081 vref(vp); 2082 } 2083 } else { 2084 kve->kve_type = KVME_TYPE_SWAP; 2085 } 2086 break; 2087 case OBJT_DEVICE: 2088 kve->kve_type = KVME_TYPE_DEVICE; 2089 break; 2090 case OBJT_PHYS: 2091 kve->kve_type = KVME_TYPE_PHYS; 2092 break; 2093 case OBJT_DEAD: 2094 kve->kve_type = KVME_TYPE_DEAD; 2095 break; 2096 case OBJT_SG: 2097 kve->kve_type = KVME_TYPE_SG; 2098 break; 2099 default: 2100 kve->kve_type = KVME_TYPE_UNKNOWN; 2101 break; 2102 } 2103 if (lobj != obj) 2104 VM_OBJECT_RUNLOCK(lobj); 2105 2106 kve->kve_ref_count = obj->ref_count; 2107 kve->kve_shadow_count = obj->shadow_count; 2108 VM_OBJECT_RUNLOCK(obj); 2109 if (vp != NULL) { 2110 vn_fullpath(curthread, vp, &fullpath, 2111 &freepath); 2112 cred = curthread->td_ucred; 2113 vn_lock(vp, LK_SHARED | LK_RETRY); 2114 if (VOP_GETATTR(vp, &va, cred) == 0) { 2115 kve->kve_fileid = va.va_fileid; 2116 kve->kve_fsid = va.va_fsid; 2117 } 2118 vput(vp); 2119 } 2120 } else { 2121 kve->kve_type = KVME_TYPE_NONE; 2122 kve->kve_ref_count = 0; 2123 kve->kve_shadow_count = 0; 2124 } 2125 2126 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2127 if (freepath != NULL) 2128 free(freepath, M_TEMP); 2129 2130 error = SYSCTL_OUT(req, kve, sizeof(*kve)); 2131 vm_map_lock_read(map); 2132 if (error) 2133 break; 2134 if (last_timestamp != map->timestamp) { 2135 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2136 entry = tmp_entry; 2137 } 2138 } 2139 vm_map_unlock_read(map); 2140 vmspace_free(vm); 2141 PRELE(p); 2142 free(kve, M_TEMP); 2143 return (error); 2144} 2145#endif /* COMPAT_FREEBSD7 */ 2146 2147#ifdef KINFO_VMENTRY_SIZE 2148CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); 2149#endif 2150 2151static void 2152kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, 2153 struct kinfo_vmentry *kve) 2154{ 2155 vm_object_t obj, tobj; 2156 vm_page_t m, m_adv; 2157 vm_offset_t addr; 2158 vm_paddr_t locked_pa; 2159 vm_pindex_t pi, pi_adv, pindex; 2160 2161 locked_pa = 0; 2162 obj = entry->object.vm_object; 2163 addr = entry->start; 2164 m_adv = NULL; 2165 pi = OFF_TO_IDX(entry->offset); 2166 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { 2167 if (m_adv != NULL) { 2168 m = m_adv; 2169 } else { 2170 pi_adv = OFF_TO_IDX(entry->end - addr); 2171 pindex = pi; 2172 for (tobj = obj;; tobj = tobj->backing_object) { 2173 m = vm_page_find_least(tobj, pindex); 2174 if (m != NULL) { 2175 if (m->pindex == pindex) 2176 break; 2177 if (pi_adv > m->pindex - pindex) { 2178 pi_adv = m->pindex - pindex; 2179 m_adv = m; 2180 } 2181 } 2182 if (tobj->backing_object == NULL) 2183 goto next; 2184 pindex += OFF_TO_IDX(tobj-> 2185 backing_object_offset); 2186 } 2187 } 2188 m_adv = NULL; 2189 if (m->psind != 0 && addr + pagesizes[1] <= entry->end && 2190 (addr & (pagesizes[1] - 1)) == 0 && 2191 (pmap_mincore(map->pmap, addr, &locked_pa) & 2192 MINCORE_SUPER) != 0) { 2193 kve->kve_flags |= KVME_FLAG_SUPER; 2194 pi_adv = OFF_TO_IDX(pagesizes[1]); 2195 } else { 2196 /* 2197 * We do not test the found page on validity. 2198 * Either the page is busy and being paged in, 2199 * or it was invalidated. The first case 2200 * should be counted as resident, the second 2201 * is not so clear; we do account both. 2202 */ 2203 pi_adv = 1; 2204 } 2205 kve->kve_resident += pi_adv; 2206next:; 2207 } 2208 PA_UNLOCK_COND(locked_pa); 2209} 2210 2211/* 2212 * Must be called with the process locked and will return unlocked. 2213 */ 2214int 2215kern_proc_vmmap_out(struct proc *p, struct sbuf *sb) 2216{ 2217 vm_map_entry_t entry, tmp_entry; 2218 struct vattr va; 2219 vm_map_t map; 2220 vm_object_t obj, tobj, lobj; 2221 char *fullpath, *freepath; 2222 struct kinfo_vmentry *kve; 2223 struct ucred *cred; 2224 struct vnode *vp; 2225 struct vmspace *vm; 2226 vm_offset_t addr; 2227 unsigned int last_timestamp; 2228 int error; 2229 2230 PROC_LOCK_ASSERT(p, MA_OWNED); 2231 2232 _PHOLD(p); 2233 PROC_UNLOCK(p); 2234 vm = vmspace_acquire_ref(p); 2235 if (vm == NULL) { 2236 PRELE(p); 2237 return (ESRCH); 2238 } 2239 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2240 2241 error = 0; 2242 map = &vm->vm_map; 2243 vm_map_lock_read(map); 2244 for (entry = map->header.next; entry != &map->header; 2245 entry = entry->next) { 2246 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2247 continue; 2248 2249 addr = entry->end; 2250 bzero(kve, sizeof(*kve)); 2251 obj = entry->object.vm_object; 2252 if (obj != NULL) { 2253 for (tobj = obj; tobj != NULL; 2254 tobj = tobj->backing_object) { 2255 VM_OBJECT_RLOCK(tobj); 2256 lobj = tobj; 2257 } 2258 if (obj->backing_object == NULL) 2259 kve->kve_private_resident = 2260 obj->resident_page_count; 2261 if (!vmmap_skip_res_cnt) 2262 kern_proc_vmmap_resident(map, entry, kve); 2263 for (tobj = obj; tobj != NULL; 2264 tobj = tobj->backing_object) { 2265 if (tobj != obj && tobj != lobj) 2266 VM_OBJECT_RUNLOCK(tobj); 2267 } 2268 } else { 2269 lobj = NULL; 2270 } 2271 2272 kve->kve_start = entry->start; 2273 kve->kve_end = entry->end; 2274 kve->kve_offset = entry->offset; 2275 2276 if (entry->protection & VM_PROT_READ) 2277 kve->kve_protection |= KVME_PROT_READ; 2278 if (entry->protection & VM_PROT_WRITE) 2279 kve->kve_protection |= KVME_PROT_WRITE; 2280 if (entry->protection & VM_PROT_EXECUTE) 2281 kve->kve_protection |= KVME_PROT_EXEC; 2282 2283 if (entry->eflags & MAP_ENTRY_COW) 2284 kve->kve_flags |= KVME_FLAG_COW; 2285 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2286 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2287 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2288 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2289 if (entry->eflags & MAP_ENTRY_GROWS_UP) 2290 kve->kve_flags |= KVME_FLAG_GROWS_UP; 2291 if (entry->eflags & MAP_ENTRY_GROWS_DOWN) 2292 kve->kve_flags |= KVME_FLAG_GROWS_DOWN; 2293 2294 last_timestamp = map->timestamp; 2295 vm_map_unlock_read(map); 2296 2297 freepath = NULL; 2298 fullpath = ""; 2299 if (lobj != NULL) { 2300 vp = NULL; 2301 switch (lobj->type) { 2302 case OBJT_DEFAULT: 2303 kve->kve_type = KVME_TYPE_DEFAULT; 2304 break; 2305 case OBJT_VNODE: 2306 kve->kve_type = KVME_TYPE_VNODE; 2307 vp = lobj->handle; 2308 vref(vp); 2309 break; 2310 case OBJT_SWAP: 2311 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) { 2312 kve->kve_type = KVME_TYPE_VNODE; 2313 if ((lobj->flags & OBJ_TMPFS) != 0) { 2314 vp = lobj->un_pager.swp.swp_tmpfs; 2315 vref(vp); 2316 } 2317 } else { 2318 kve->kve_type = KVME_TYPE_SWAP; 2319 } 2320 break; 2321 case OBJT_DEVICE: 2322 kve->kve_type = KVME_TYPE_DEVICE; 2323 break; 2324 case OBJT_PHYS: 2325 kve->kve_type = KVME_TYPE_PHYS; 2326 break; 2327 case OBJT_DEAD: 2328 kve->kve_type = KVME_TYPE_DEAD; 2329 break; 2330 case OBJT_SG: 2331 kve->kve_type = KVME_TYPE_SG; 2332 break; 2333 case OBJT_MGTDEVICE: 2334 kve->kve_type = KVME_TYPE_MGTDEVICE; 2335 break; 2336 default: 2337 kve->kve_type = KVME_TYPE_UNKNOWN; 2338 break; 2339 } 2340 if (lobj != obj) 2341 VM_OBJECT_RUNLOCK(lobj); 2342 2343 kve->kve_ref_count = obj->ref_count; 2344 kve->kve_shadow_count = obj->shadow_count; 2345 VM_OBJECT_RUNLOCK(obj); 2346 if (vp != NULL) { 2347 vn_fullpath(curthread, vp, &fullpath, 2348 &freepath); 2349 kve->kve_vn_type = vntype_to_kinfo(vp->v_type); 2350 cred = curthread->td_ucred; 2351 vn_lock(vp, LK_SHARED | LK_RETRY); 2352 if (VOP_GETATTR(vp, &va, cred) == 0) { 2353 kve->kve_vn_fileid = va.va_fileid; 2354 kve->kve_vn_fsid = va.va_fsid; 2355 kve->kve_vn_mode = 2356 MAKEIMODE(va.va_type, va.va_mode); 2357 kve->kve_vn_size = va.va_size; 2358 kve->kve_vn_rdev = va.va_rdev; 2359 kve->kve_status = KF_ATTR_VALID; 2360 } 2361 vput(vp); 2362 } 2363 } else { 2364 kve->kve_type = KVME_TYPE_NONE; 2365 kve->kve_ref_count = 0; 2366 kve->kve_shadow_count = 0; 2367 } 2368 2369 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2370 if (freepath != NULL) 2371 free(freepath, M_TEMP); 2372 2373 /* Pack record size down */ 2374 kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) + 2375 strlen(kve->kve_path) + 1; 2376 kve->kve_structsize = roundup(kve->kve_structsize, 2377 sizeof(uint64_t)); 2378 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0) 2379 error = ENOMEM; 2380 vm_map_lock_read(map); 2381 if (error != 0) 2382 break; 2383 if (last_timestamp != map->timestamp) { 2384 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2385 entry = tmp_entry; 2386 } 2387 } 2388 vm_map_unlock_read(map); 2389 vmspace_free(vm); 2390 PRELE(p); 2391 free(kve, M_TEMP); 2392 return (error); 2393} 2394 2395static int 2396sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) 2397{ 2398 struct proc *p; 2399 struct sbuf sb; 2400 int error, error2, *name; 2401 2402 name = (int *)arg1; 2403 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); 2404 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); 2405 if (error != 0) { 2406 sbuf_delete(&sb); 2407 return (error); 2408 } 2409 error = kern_proc_vmmap_out(p, &sb); 2410 error2 = sbuf_finish(&sb); 2411 sbuf_delete(&sb); 2412 return (error != 0 ? error : error2); 2413} 2414 2415#if defined(STACK) || defined(DDB) 2416static int 2417sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) 2418{ 2419 struct kinfo_kstack *kkstp; 2420 int error, i, *name, numthreads; 2421 lwpid_t *lwpidarray; 2422 struct thread *td; 2423 struct stack *st; 2424 struct sbuf sb; 2425 struct proc *p; 2426 2427 name = (int *)arg1; 2428 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); 2429 if (error != 0) 2430 return (error); 2431 2432 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); 2433 st = stack_create(); 2434 2435 lwpidarray = NULL; 2436 numthreads = 0; 2437 PROC_LOCK(p); 2438repeat: 2439 if (numthreads < p->p_numthreads) { 2440 if (lwpidarray != NULL) { 2441 free(lwpidarray, M_TEMP); 2442 lwpidarray = NULL; 2443 } 2444 numthreads = p->p_numthreads; 2445 PROC_UNLOCK(p); 2446 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, 2447 M_WAITOK | M_ZERO); 2448 PROC_LOCK(p); 2449 goto repeat; 2450 } 2451 i = 0; 2452 2453 /* 2454 * XXXRW: During the below loop, execve(2) and countless other sorts 2455 * of changes could have taken place. Should we check to see if the 2456 * vmspace has been replaced, or the like, in order to prevent 2457 * giving a snapshot that spans, say, execve(2), with some threads 2458 * before and some after? Among other things, the credentials could 2459 * have changed, in which case the right to extract debug info might 2460 * no longer be assured. 2461 */ 2462 FOREACH_THREAD_IN_PROC(p, td) { 2463 KASSERT(i < numthreads, 2464 ("sysctl_kern_proc_kstack: numthreads")); 2465 lwpidarray[i] = td->td_tid; 2466 i++; 2467 } 2468 numthreads = i; 2469 for (i = 0; i < numthreads; i++) { 2470 td = thread_find(p, lwpidarray[i]); 2471 if (td == NULL) { 2472 continue; 2473 } 2474 bzero(kkstp, sizeof(*kkstp)); 2475 (void)sbuf_new(&sb, kkstp->kkst_trace, 2476 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); 2477 thread_lock(td); 2478 kkstp->kkst_tid = td->td_tid; 2479 if (TD_IS_SWAPPED(td)) 2480 kkstp->kkst_state = KKST_STATE_SWAPPED; 2481 else if (TD_IS_RUNNING(td)) 2482 kkstp->kkst_state = KKST_STATE_RUNNING; 2483 else { 2484 kkstp->kkst_state = KKST_STATE_STACKOK; 2485 stack_save_td(st, td); 2486 } 2487 thread_unlock(td); 2488 PROC_UNLOCK(p); 2489 stack_sbuf_print(&sb, st); 2490 sbuf_finish(&sb); 2491 sbuf_delete(&sb); 2492 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); 2493 PROC_LOCK(p); 2494 if (error) 2495 break; 2496 } 2497 _PRELE(p); 2498 PROC_UNLOCK(p); 2499 if (lwpidarray != NULL) 2500 free(lwpidarray, M_TEMP); 2501 stack_destroy(st); 2502 free(kkstp, M_TEMP); 2503 return (error); 2504} 2505#endif 2506 2507/* 2508 * This sysctl allows a process to retrieve the full list of groups from 2509 * itself or another process. 2510 */ 2511static int 2512sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) 2513{ 2514 pid_t *pidp = (pid_t *)arg1; 2515 unsigned int arglen = arg2; 2516 struct proc *p; 2517 struct ucred *cred; 2518 int error; 2519 2520 if (arglen != 1) 2521 return (EINVAL); 2522 if (*pidp == -1) { /* -1 means this process */ 2523 p = req->td->td_proc; 2524 } else { 2525 error = pget(*pidp, PGET_CANSEE, &p); 2526 if (error != 0) 2527 return (error); 2528 } 2529 2530 cred = crhold(p->p_ucred); 2531 if (*pidp != -1) 2532 PROC_UNLOCK(p); 2533 2534 error = SYSCTL_OUT(req, cred->cr_groups, 2535 cred->cr_ngroups * sizeof(gid_t)); 2536 crfree(cred); 2537 return (error); 2538} 2539 2540/* 2541 * This sysctl allows a process to retrieve or/and set the resource limit for 2542 * another process. 2543 */ 2544static int 2545sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) 2546{ 2547 int *name = (int *)arg1; 2548 u_int namelen = arg2; 2549 struct rlimit rlim; 2550 struct proc *p; 2551 u_int which; 2552 int flags, error; 2553 2554 if (namelen != 2) 2555 return (EINVAL); 2556 2557 which = (u_int)name[1]; 2558 if (which >= RLIM_NLIMITS) 2559 return (EINVAL); 2560 2561 if (req->newptr != NULL && req->newlen != sizeof(rlim)) 2562 return (EINVAL); 2563 2564 flags = PGET_HOLD | PGET_NOTWEXIT; 2565 if (req->newptr != NULL) 2566 flags |= PGET_CANDEBUG; 2567 else 2568 flags |= PGET_CANSEE; 2569 error = pget((pid_t)name[0], flags, &p); 2570 if (error != 0) 2571 return (error); 2572 2573 /* 2574 * Retrieve limit. 2575 */ 2576 if (req->oldptr != NULL) { 2577 PROC_LOCK(p); 2578 lim_rlimit(p, which, &rlim); 2579 PROC_UNLOCK(p); 2580 } 2581 error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); 2582 if (error != 0) 2583 goto errout; 2584 2585 /* 2586 * Set limit. 2587 */ 2588 if (req->newptr != NULL) { 2589 error = SYSCTL_IN(req, &rlim, sizeof(rlim)); 2590 if (error == 0) 2591 error = kern_proc_setrlimit(curthread, p, which, &rlim); 2592 } 2593 2594errout: 2595 PRELE(p); 2596 return (error); 2597} 2598 2599/* 2600 * This sysctl allows a process to retrieve ps_strings structure location of 2601 * another process. 2602 */ 2603static int 2604sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) 2605{ 2606 int *name = (int *)arg1; 2607 u_int namelen = arg2; 2608 struct proc *p; 2609 vm_offset_t ps_strings; 2610 int error; 2611#ifdef COMPAT_FREEBSD32 2612 uint32_t ps_strings32; 2613#endif 2614 2615 if (namelen != 1) 2616 return (EINVAL); 2617 2618 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2619 if (error != 0) 2620 return (error); 2621#ifdef COMPAT_FREEBSD32 2622 if ((req->flags & SCTL_MASK32) != 0) { 2623 /* 2624 * We return 0 if the 32 bit emulation request is for a 64 bit 2625 * process. 2626 */ 2627 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? 2628 PTROUT(p->p_sysent->sv_psstrings) : 0; 2629 PROC_UNLOCK(p); 2630 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); 2631 return (error); 2632 } 2633#endif 2634 ps_strings = p->p_sysent->sv_psstrings; 2635 PROC_UNLOCK(p); 2636 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); 2637 return (error); 2638} 2639 2640/* 2641 * This sysctl allows a process to retrieve umask of another process. 2642 */ 2643static int 2644sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) 2645{ 2646 int *name = (int *)arg1; 2647 u_int namelen = arg2; 2648 struct proc *p; 2649 int error; 2650 u_short fd_cmask; 2651 2652 if (namelen != 1) 2653 return (EINVAL); 2654 2655 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2656 if (error != 0) 2657 return (error); 2658 2659 FILEDESC_SLOCK(p->p_fd); 2660 fd_cmask = p->p_fd->fd_cmask; 2661 FILEDESC_SUNLOCK(p->p_fd); 2662 PRELE(p); 2663 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask)); 2664 return (error); 2665} 2666 2667/* 2668 * This sysctl allows a process to set and retrieve binary osreldate of 2669 * another process. 2670 */ 2671static int 2672sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) 2673{ 2674 int *name = (int *)arg1; 2675 u_int namelen = arg2; 2676 struct proc *p; 2677 int flags, error, osrel; 2678 2679 if (namelen != 1) 2680 return (EINVAL); 2681 2682 if (req->newptr != NULL && req->newlen != sizeof(osrel)) 2683 return (EINVAL); 2684 2685 flags = PGET_HOLD | PGET_NOTWEXIT; 2686 if (req->newptr != NULL) 2687 flags |= PGET_CANDEBUG; 2688 else 2689 flags |= PGET_CANSEE; 2690 error = pget((pid_t)name[0], flags, &p); 2691 if (error != 0) 2692 return (error); 2693 2694 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); 2695 if (error != 0) 2696 goto errout; 2697 2698 if (req->newptr != NULL) { 2699 error = SYSCTL_IN(req, &osrel, sizeof(osrel)); 2700 if (error != 0) 2701 goto errout; 2702 if (osrel < 0) { 2703 error = EINVAL; 2704 goto errout; 2705 } 2706 p->p_osrel = osrel; 2707 } 2708errout: 2709 PRELE(p); 2710 return (error); 2711} 2712 2713static int 2714sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) 2715{ 2716 int *name = (int *)arg1; 2717 u_int namelen = arg2; 2718 struct proc *p; 2719 struct kinfo_sigtramp kst; 2720 const struct sysentvec *sv; 2721 int error; 2722#ifdef COMPAT_FREEBSD32 2723 struct kinfo_sigtramp32 kst32; 2724#endif 2725 2726 if (namelen != 1) 2727 return (EINVAL); 2728 2729 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2730 if (error != 0) 2731 return (error); 2732 sv = p->p_sysent; 2733#ifdef COMPAT_FREEBSD32 2734 if ((req->flags & SCTL_MASK32) != 0) { 2735 bzero(&kst32, sizeof(kst32)); 2736 if (SV_PROC_FLAG(p, SV_ILP32)) { 2737 if (sv->sv_sigcode_base != 0) { 2738 kst32.ksigtramp_start = sv->sv_sigcode_base; 2739 kst32.ksigtramp_end = sv->sv_sigcode_base + 2740 *sv->sv_szsigcode; 2741 } else { 2742 kst32.ksigtramp_start = sv->sv_psstrings - 2743 *sv->sv_szsigcode; 2744 kst32.ksigtramp_end = sv->sv_psstrings; 2745 } 2746 } 2747 PROC_UNLOCK(p); 2748 error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); 2749 return (error); 2750 } 2751#endif 2752 bzero(&kst, sizeof(kst)); 2753 if (sv->sv_sigcode_base != 0) { 2754 kst.ksigtramp_start = (char *)sv->sv_sigcode_base; 2755 kst.ksigtramp_end = (char *)sv->sv_sigcode_base + 2756 *sv->sv_szsigcode; 2757 } else { 2758 kst.ksigtramp_start = (char *)sv->sv_psstrings - 2759 *sv->sv_szsigcode; 2760 kst.ksigtramp_end = (char *)sv->sv_psstrings; 2761 } 2762 PROC_UNLOCK(p); 2763 error = SYSCTL_OUT(req, &kst, sizeof(kst)); 2764 return (error); 2765} 2766 2767SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table"); 2768 2769SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| 2770 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", 2771 "Return entire process table"); 2772 2773static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2774 sysctl_kern_proc, "Process table"); 2775 2776static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, 2777 sysctl_kern_proc, "Process table"); 2778 2779static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2780 sysctl_kern_proc, "Process table"); 2781 2782static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | 2783 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2784 2785static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 2786 sysctl_kern_proc, "Process table"); 2787 2788static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2789 sysctl_kern_proc, "Process table"); 2790 2791static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2792 sysctl_kern_proc, "Process table"); 2793 2794static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2795 sysctl_kern_proc, "Process table"); 2796 2797static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 2798 sysctl_kern_proc, "Return process table, no threads"); 2799 2800static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, 2801 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, 2802 sysctl_kern_proc_args, "Process argument list"); 2803 2804static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, 2805 sysctl_kern_proc_env, "Process environment"); 2806 2807static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | 2808 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); 2809 2810static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | 2811 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); 2812 2813static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | 2814 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, 2815 "Process syscall vector name (ABI type)"); 2816 2817static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, 2818 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2819 2820static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, 2821 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2822 2823static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, 2824 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2825 2826static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), 2827 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2828 2829static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, 2830 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2831 2832static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, 2833 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2834 2835static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, 2836 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2837 2838static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, 2839 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2840 2841static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, 2842 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, 2843 "Return process table, no threads"); 2844 2845#ifdef COMPAT_FREEBSD7 2846static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | 2847 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); 2848#endif 2849 2850static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | 2851 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); 2852 2853#if defined(STACK) || defined(DDB) 2854static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | 2855 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); 2856#endif 2857 2858static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | 2859 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); 2860 2861static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | 2862 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, 2863 "Process resource limits"); 2864 2865static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | 2866 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, 2867 "Process ps_strings location"); 2868 2869static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | 2870 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); 2871 2872static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | 2873 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, 2874 "Process binary osreldate"); 2875 2876static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | 2877 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, 2878 "Process signal trampoline location"); 2879 2880int allproc_gen; 2881 2882void 2883stop_all_proc(void) 2884{ 2885 struct proc *cp, *p; 2886 int r, gen; 2887 bool restart, seen_stopped, seen_exiting, stopped_some; 2888 2889 cp = curproc; 2890 /* 2891 * stop_all_proc() assumes that all process which have 2892 * usermode must be stopped, except current process, for 2893 * obvious reasons. Since other threads in the process 2894 * establishing global stop could unstop something, disable 2895 * calls from multithreaded processes as precaution. The 2896 * service must not be user-callable anyway. 2897 */ 2898 KASSERT((cp->p_flag & P_HADTHREADS) == 0 || 2899 (cp->p_flag & P_KTHREAD) != 0, ("mt stop_all_proc")); 2900 2901allproc_loop: 2902 sx_xlock(&allproc_lock); 2903 gen = allproc_gen; 2904 seen_exiting = seen_stopped = stopped_some = restart = false; 2905 LIST_REMOVE(cp, p_list); 2906 LIST_INSERT_HEAD(&allproc, cp, p_list); 2907 for (;;) { 2908 p = LIST_NEXT(cp, p_list); 2909 if (p == NULL) 2910 break; 2911 LIST_REMOVE(cp, p_list); 2912 LIST_INSERT_AFTER(p, cp, p_list); 2913 PROC_LOCK(p); 2914 if ((p->p_flag & (P_KTHREAD | P_SYSTEM | 2915 P_TOTAL_STOP)) != 0) { 2916 PROC_UNLOCK(p); 2917 continue; 2918 } 2919 if ((p->p_flag & P_WEXIT) != 0) { 2920 seen_exiting = true; 2921 PROC_UNLOCK(p); 2922 continue; 2923 } 2924 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 2925 /* 2926 * Stopped processes are tolerated when there 2927 * are no other processes which might continue 2928 * them. P_STOPPED_SINGLE but not 2929 * P_TOTAL_STOP process still has at least one 2930 * thread running. 2931 */ 2932 seen_stopped = true; 2933 PROC_UNLOCK(p); 2934 continue; 2935 } 2936 _PHOLD(p); 2937 sx_xunlock(&allproc_lock); 2938 r = thread_single(p, SINGLE_ALLPROC); 2939 if (r != 0) 2940 restart = true; 2941 else 2942 stopped_some = true; 2943 _PRELE(p); 2944 PROC_UNLOCK(p); 2945 sx_xlock(&allproc_lock); 2946 } 2947 /* Catch forked children we did not see in iteration. */ 2948 if (gen != allproc_gen) 2949 restart = true; 2950 sx_xunlock(&allproc_lock); 2951 if (restart || stopped_some || seen_exiting || seen_stopped) { 2952 kern_yield(PRI_USER); 2953 goto allproc_loop; 2954 } 2955} 2956 2957void 2958resume_all_proc(void) 2959{ 2960 struct proc *cp, *p; 2961 2962 cp = curproc; 2963 sx_xlock(&allproc_lock); 2964 LIST_REMOVE(cp, p_list); 2965 LIST_INSERT_HEAD(&allproc, cp, p_list); 2966 for (;;) { 2967 p = LIST_NEXT(cp, p_list); 2968 if (p == NULL) 2969 break; 2970 LIST_REMOVE(cp, p_list); 2971 LIST_INSERT_AFTER(p, cp, p_list); 2972 PROC_LOCK(p); 2973 if ((p->p_flag & P_TOTAL_STOP) != 0) { 2974 sx_xunlock(&allproc_lock); 2975 _PHOLD(p); 2976 thread_single_end(p, SINGLE_ALLPROC); 2977 _PRELE(p); 2978 PROC_UNLOCK(p); 2979 sx_xlock(&allproc_lock); 2980 } else { 2981 PROC_UNLOCK(p); 2982 } 2983 } 2984 sx_xunlock(&allproc_lock); 2985} 2986 2987#define TOTAL_STOP_DEBUG 1 2988#ifdef TOTAL_STOP_DEBUG 2989volatile static int ap_resume; 2990#include <sys/mount.h> 2991 2992static int 2993sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) 2994{ 2995 int error, val; 2996 2997 val = 0; 2998 ap_resume = 0; 2999 error = sysctl_handle_int(oidp, &val, 0, req); 3000 if (error != 0 || req->newptr == NULL) 3001 return (error); 3002 if (val != 0) { 3003 stop_all_proc(); 3004 syncer_suspend(); 3005 while (ap_resume == 0) 3006 ; 3007 syncer_resume(); 3008 resume_all_proc(); 3009 } 3010 return (0); 3011} 3012 3013SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | 3014 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, 3015 sysctl_debug_stop_all_proc, "I", 3016 ""); 3017#endif 3018