kern_proc.c revision 276272
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 276272 2014-12-27 00:55:14Z kib $"); 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 error = sbuf_bcat(sb, &ki32, sizeof(ki32)); 1207 } else 1208#endif 1209 error = sbuf_bcat(sb, &ki, sizeof(ki)); 1210 } else { 1211 FOREACH_THREAD_IN_PROC(p, td) { 1212 fill_kinfo_thread(td, &ki, 1); 1213#ifdef COMPAT_FREEBSD32 1214 if ((flags & KERN_PROC_MASK32) != 0) { 1215 freebsd32_kinfo_proc_out(&ki, &ki32); 1216 error = sbuf_bcat(sb, &ki32, sizeof(ki32)); 1217 } else 1218#endif 1219 error = sbuf_bcat(sb, &ki, sizeof(ki)); 1220 if (error) 1221 break; 1222 } 1223 } 1224 PROC_UNLOCK(p); 1225 return (error); 1226} 1227 1228static int 1229sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags, 1230 int doingzomb) 1231{ 1232 struct sbuf sb; 1233 struct kinfo_proc ki; 1234 struct proc *np; 1235 int error, error2; 1236 pid_t pid; 1237 1238 pid = p->p_pid; 1239 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); 1240 error = kern_proc_out(p, &sb, flags); 1241 error2 = sbuf_finish(&sb); 1242 sbuf_delete(&sb); 1243 if (error != 0) 1244 return (error); 1245 else if (error2 != 0) 1246 return (error2); 1247 if (doingzomb) 1248 np = zpfind(pid); 1249 else { 1250 if (pid == 0) 1251 return (0); 1252 np = pfind(pid); 1253 } 1254 if (np == NULL) 1255 return (ESRCH); 1256 if (np != p) { 1257 PROC_UNLOCK(np); 1258 return (ESRCH); 1259 } 1260 PROC_UNLOCK(np); 1261 return (0); 1262} 1263 1264static int 1265sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 1266{ 1267 int *name = (int *)arg1; 1268 u_int namelen = arg2; 1269 struct proc *p; 1270 int flags, doingzomb, oid_number; 1271 int error = 0; 1272 1273 oid_number = oidp->oid_number; 1274 if (oid_number != KERN_PROC_ALL && 1275 (oid_number & KERN_PROC_INC_THREAD) == 0) 1276 flags = KERN_PROC_NOTHREADS; 1277 else { 1278 flags = 0; 1279 oid_number &= ~KERN_PROC_INC_THREAD; 1280 } 1281#ifdef COMPAT_FREEBSD32 1282 if (req->flags & SCTL_MASK32) 1283 flags |= KERN_PROC_MASK32; 1284#endif 1285 if (oid_number == KERN_PROC_PID) { 1286 if (namelen != 1) 1287 return (EINVAL); 1288 error = sysctl_wire_old_buffer(req, 0); 1289 if (error) 1290 return (error); 1291 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1292 if (error != 0) 1293 return (error); 1294 error = sysctl_out_proc(p, req, flags, 0); 1295 return (error); 1296 } 1297 1298 switch (oid_number) { 1299 case KERN_PROC_ALL: 1300 if (namelen != 0) 1301 return (EINVAL); 1302 break; 1303 case KERN_PROC_PROC: 1304 if (namelen != 0 && namelen != 1) 1305 return (EINVAL); 1306 break; 1307 default: 1308 if (namelen != 1) 1309 return (EINVAL); 1310 break; 1311 } 1312 1313 if (!req->oldptr) { 1314 /* overestimate by 5 procs */ 1315 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1316 if (error) 1317 return (error); 1318 } 1319 error = sysctl_wire_old_buffer(req, 0); 1320 if (error != 0) 1321 return (error); 1322 sx_slock(&allproc_lock); 1323 for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) { 1324 if (!doingzomb) 1325 p = LIST_FIRST(&allproc); 1326 else 1327 p = LIST_FIRST(&zombproc); 1328 for (; p != 0; p = LIST_NEXT(p, p_list)) { 1329 /* 1330 * Skip embryonic processes. 1331 */ 1332 PROC_LOCK(p); 1333 if (p->p_state == PRS_NEW) { 1334 PROC_UNLOCK(p); 1335 continue; 1336 } 1337 KASSERT(p->p_ucred != NULL, 1338 ("process credential is NULL for non-NEW proc")); 1339 /* 1340 * Show a user only appropriate processes. 1341 */ 1342 if (p_cansee(curthread, p)) { 1343 PROC_UNLOCK(p); 1344 continue; 1345 } 1346 /* 1347 * TODO - make more efficient (see notes below). 1348 * do by session. 1349 */ 1350 switch (oid_number) { 1351 1352 case KERN_PROC_GID: 1353 if (p->p_ucred->cr_gid != (gid_t)name[0]) { 1354 PROC_UNLOCK(p); 1355 continue; 1356 } 1357 break; 1358 1359 case KERN_PROC_PGRP: 1360 /* could do this by traversing pgrp */ 1361 if (p->p_pgrp == NULL || 1362 p->p_pgrp->pg_id != (pid_t)name[0]) { 1363 PROC_UNLOCK(p); 1364 continue; 1365 } 1366 break; 1367 1368 case KERN_PROC_RGID: 1369 if (p->p_ucred->cr_rgid != (gid_t)name[0]) { 1370 PROC_UNLOCK(p); 1371 continue; 1372 } 1373 break; 1374 1375 case KERN_PROC_SESSION: 1376 if (p->p_session == NULL || 1377 p->p_session->s_sid != (pid_t)name[0]) { 1378 PROC_UNLOCK(p); 1379 continue; 1380 } 1381 break; 1382 1383 case KERN_PROC_TTY: 1384 if ((p->p_flag & P_CONTROLT) == 0 || 1385 p->p_session == NULL) { 1386 PROC_UNLOCK(p); 1387 continue; 1388 } 1389 /* XXX proctree_lock */ 1390 SESS_LOCK(p->p_session); 1391 if (p->p_session->s_ttyp == NULL || 1392 tty_udev(p->p_session->s_ttyp) != 1393 (dev_t)name[0]) { 1394 SESS_UNLOCK(p->p_session); 1395 PROC_UNLOCK(p); 1396 continue; 1397 } 1398 SESS_UNLOCK(p->p_session); 1399 break; 1400 1401 case KERN_PROC_UID: 1402 if (p->p_ucred->cr_uid != (uid_t)name[0]) { 1403 PROC_UNLOCK(p); 1404 continue; 1405 } 1406 break; 1407 1408 case KERN_PROC_RUID: 1409 if (p->p_ucred->cr_ruid != (uid_t)name[0]) { 1410 PROC_UNLOCK(p); 1411 continue; 1412 } 1413 break; 1414 1415 case KERN_PROC_PROC: 1416 break; 1417 1418 default: 1419 break; 1420 1421 } 1422 1423 error = sysctl_out_proc(p, req, flags, doingzomb); 1424 if (error) { 1425 sx_sunlock(&allproc_lock); 1426 return (error); 1427 } 1428 } 1429 } 1430 sx_sunlock(&allproc_lock); 1431 return (0); 1432} 1433 1434struct pargs * 1435pargs_alloc(int len) 1436{ 1437 struct pargs *pa; 1438 1439 pa = malloc(sizeof(struct pargs) + len, M_PARGS, 1440 M_WAITOK); 1441 refcount_init(&pa->ar_ref, 1); 1442 pa->ar_length = len; 1443 return (pa); 1444} 1445 1446static void 1447pargs_free(struct pargs *pa) 1448{ 1449 1450 free(pa, M_PARGS); 1451} 1452 1453void 1454pargs_hold(struct pargs *pa) 1455{ 1456 1457 if (pa == NULL) 1458 return; 1459 refcount_acquire(&pa->ar_ref); 1460} 1461 1462void 1463pargs_drop(struct pargs *pa) 1464{ 1465 1466 if (pa == NULL) 1467 return; 1468 if (refcount_release(&pa->ar_ref)) 1469 pargs_free(pa); 1470} 1471 1472static int 1473proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf, 1474 size_t len) 1475{ 1476 struct iovec iov; 1477 struct uio uio; 1478 1479 iov.iov_base = (caddr_t)buf; 1480 iov.iov_len = len; 1481 uio.uio_iov = &iov; 1482 uio.uio_iovcnt = 1; 1483 uio.uio_offset = offset; 1484 uio.uio_resid = (ssize_t)len; 1485 uio.uio_segflg = UIO_SYSSPACE; 1486 uio.uio_rw = UIO_READ; 1487 uio.uio_td = td; 1488 1489 return (proc_rwmem(p, &uio)); 1490} 1491 1492static int 1493proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, 1494 size_t len) 1495{ 1496 size_t i; 1497 int error; 1498 1499 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len); 1500 /* 1501 * Reading the chunk may validly return EFAULT if the string is shorter 1502 * than the chunk and is aligned at the end of the page, assuming the 1503 * next page is not mapped. So if EFAULT is returned do a fallback to 1504 * one byte read loop. 1505 */ 1506 if (error == EFAULT) { 1507 for (i = 0; i < len; i++, buf++, sptr++) { 1508 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1); 1509 if (error != 0) 1510 return (error); 1511 if (*buf == '\0') 1512 break; 1513 } 1514 error = 0; 1515 } 1516 return (error); 1517} 1518 1519#define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ 1520 1521enum proc_vector_type { 1522 PROC_ARG, 1523 PROC_ENV, 1524 PROC_AUX, 1525}; 1526 1527#ifdef COMPAT_FREEBSD32 1528static int 1529get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, 1530 size_t *vsizep, enum proc_vector_type type) 1531{ 1532 struct freebsd32_ps_strings pss; 1533 Elf32_Auxinfo aux; 1534 vm_offset_t vptr, ptr; 1535 uint32_t *proc_vector32; 1536 char **proc_vector; 1537 size_t vsize, size; 1538 int i, error; 1539 1540 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), 1541 &pss, sizeof(pss)); 1542 if (error != 0) 1543 return (error); 1544 switch (type) { 1545 case PROC_ARG: 1546 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); 1547 vsize = pss.ps_nargvstr; 1548 if (vsize > ARG_MAX) 1549 return (ENOEXEC); 1550 size = vsize * sizeof(int32_t); 1551 break; 1552 case PROC_ENV: 1553 vptr = (vm_offset_t)PTRIN(pss.ps_envstr); 1554 vsize = pss.ps_nenvstr; 1555 if (vsize > ARG_MAX) 1556 return (ENOEXEC); 1557 size = vsize * sizeof(int32_t); 1558 break; 1559 case PROC_AUX: 1560 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + 1561 (pss.ps_nenvstr + 1) * sizeof(int32_t); 1562 if (vptr % 4 != 0) 1563 return (ENOEXEC); 1564 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1565 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); 1566 if (error != 0) 1567 return (error); 1568 if (aux.a_type == AT_NULL) 1569 break; 1570 ptr += sizeof(aux); 1571 } 1572 if (aux.a_type != AT_NULL) 1573 return (ENOEXEC); 1574 vsize = i + 1; 1575 size = vsize * sizeof(aux); 1576 break; 1577 default: 1578 KASSERT(0, ("Wrong proc vector type: %d", type)); 1579 return (EINVAL); 1580 } 1581 proc_vector32 = malloc(size, M_TEMP, M_WAITOK); 1582 error = proc_read_mem(td, p, vptr, proc_vector32, size); 1583 if (error != 0) 1584 goto done; 1585 if (type == PROC_AUX) { 1586 *proc_vectorp = (char **)proc_vector32; 1587 *vsizep = vsize; 1588 return (0); 1589 } 1590 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); 1591 for (i = 0; i < (int)vsize; i++) 1592 proc_vector[i] = PTRIN(proc_vector32[i]); 1593 *proc_vectorp = proc_vector; 1594 *vsizep = vsize; 1595done: 1596 free(proc_vector32, M_TEMP); 1597 return (error); 1598} 1599#endif 1600 1601static int 1602get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, 1603 size_t *vsizep, enum proc_vector_type type) 1604{ 1605 struct ps_strings pss; 1606 Elf_Auxinfo aux; 1607 vm_offset_t vptr, ptr; 1608 char **proc_vector; 1609 size_t vsize, size; 1610 int error, i; 1611 1612#ifdef COMPAT_FREEBSD32 1613 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1614 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); 1615#endif 1616 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), 1617 &pss, sizeof(pss)); 1618 if (error != 0) 1619 return (error); 1620 switch (type) { 1621 case PROC_ARG: 1622 vptr = (vm_offset_t)pss.ps_argvstr; 1623 vsize = pss.ps_nargvstr; 1624 if (vsize > ARG_MAX) 1625 return (ENOEXEC); 1626 size = vsize * sizeof(char *); 1627 break; 1628 case PROC_ENV: 1629 vptr = (vm_offset_t)pss.ps_envstr; 1630 vsize = pss.ps_nenvstr; 1631 if (vsize > ARG_MAX) 1632 return (ENOEXEC); 1633 size = vsize * sizeof(char *); 1634 break; 1635 case PROC_AUX: 1636 /* 1637 * The aux array is just above env array on the stack. Check 1638 * that the address is naturally aligned. 1639 */ 1640 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) 1641 * sizeof(char *); 1642#if __ELF_WORD_SIZE == 64 1643 if (vptr % sizeof(uint64_t) != 0) 1644#else 1645 if (vptr % sizeof(uint32_t) != 0) 1646#endif 1647 return (ENOEXEC); 1648 /* 1649 * We count the array size reading the aux vectors from the 1650 * stack until AT_NULL vector is returned. So (to keep the code 1651 * simple) we read the process stack twice: the first time here 1652 * to find the size and the second time when copying the vectors 1653 * to the allocated proc_vector. 1654 */ 1655 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1656 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); 1657 if (error != 0) 1658 return (error); 1659 if (aux.a_type == AT_NULL) 1660 break; 1661 ptr += sizeof(aux); 1662 } 1663 /* 1664 * If the PROC_AUXV_MAX entries are iterated over, and we have 1665 * not reached AT_NULL, it is most likely we are reading wrong 1666 * data: either the process doesn't have auxv array or data has 1667 * been modified. Return the error in this case. 1668 */ 1669 if (aux.a_type != AT_NULL) 1670 return (ENOEXEC); 1671 vsize = i + 1; 1672 size = vsize * sizeof(aux); 1673 break; 1674 default: 1675 KASSERT(0, ("Wrong proc vector type: %d", type)); 1676 return (EINVAL); /* In case we are built without INVARIANTS. */ 1677 } 1678 proc_vector = malloc(size, M_TEMP, M_WAITOK); 1679 if (proc_vector == NULL) 1680 return (ENOMEM); 1681 error = proc_read_mem(td, p, vptr, proc_vector, size); 1682 if (error != 0) { 1683 free(proc_vector, M_TEMP); 1684 return (error); 1685 } 1686 *proc_vectorp = proc_vector; 1687 *vsizep = vsize; 1688 1689 return (0); 1690} 1691 1692#define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ 1693 1694static int 1695get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, 1696 enum proc_vector_type type) 1697{ 1698 size_t done, len, nchr, vsize; 1699 int error, i; 1700 char **proc_vector, *sptr; 1701 char pss_string[GET_PS_STRINGS_CHUNK_SZ]; 1702 1703 PROC_ASSERT_HELD(p); 1704 1705 /* 1706 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. 1707 */ 1708 nchr = 2 * (PATH_MAX + ARG_MAX); 1709 1710 error = get_proc_vector(td, p, &proc_vector, &vsize, type); 1711 if (error != 0) 1712 return (error); 1713 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { 1714 /* 1715 * The program may have scribbled into its argv array, e.g. to 1716 * remove some arguments. If that has happened, break out 1717 * before trying to read from NULL. 1718 */ 1719 if (proc_vector[i] == NULL) 1720 break; 1721 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { 1722 error = proc_read_string(td, p, sptr, pss_string, 1723 sizeof(pss_string)); 1724 if (error != 0) 1725 goto done; 1726 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); 1727 if (done + len >= nchr) 1728 len = nchr - done - 1; 1729 sbuf_bcat(sb, pss_string, len); 1730 if (len != GET_PS_STRINGS_CHUNK_SZ) 1731 break; 1732 done += GET_PS_STRINGS_CHUNK_SZ; 1733 } 1734 sbuf_bcat(sb, "", 1); 1735 done += len + 1; 1736 } 1737done: 1738 free(proc_vector, M_TEMP); 1739 return (error); 1740} 1741 1742int 1743proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) 1744{ 1745 1746 return (get_ps_strings(curthread, p, sb, PROC_ARG)); 1747} 1748 1749int 1750proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) 1751{ 1752 1753 return (get_ps_strings(curthread, p, sb, PROC_ENV)); 1754} 1755 1756int 1757proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) 1758{ 1759 size_t vsize, size; 1760 char **auxv; 1761 int error; 1762 1763 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); 1764 if (error == 0) { 1765#ifdef COMPAT_FREEBSD32 1766 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1767 size = vsize * sizeof(Elf32_Auxinfo); 1768 else 1769#endif 1770 size = vsize * sizeof(Elf_Auxinfo); 1771 error = sbuf_bcat(sb, auxv, size); 1772 free(auxv, M_TEMP); 1773 } 1774 return (error); 1775} 1776 1777/* 1778 * This sysctl allows a process to retrieve the argument list or process 1779 * title for another process without groping around in the address space 1780 * of the other process. It also allow a process to set its own "process 1781 * title to a string of its own choice. 1782 */ 1783static int 1784sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 1785{ 1786 int *name = (int *)arg1; 1787 u_int namelen = arg2; 1788 struct pargs *newpa, *pa; 1789 struct proc *p; 1790 struct sbuf sb; 1791 int flags, error = 0, error2; 1792 1793 if (namelen != 1) 1794 return (EINVAL); 1795 1796 flags = PGET_CANSEE; 1797 if (req->newptr != NULL) 1798 flags |= PGET_ISCURRENT; 1799 error = pget((pid_t)name[0], flags, &p); 1800 if (error) 1801 return (error); 1802 1803 pa = p->p_args; 1804 if (pa != NULL) { 1805 pargs_hold(pa); 1806 PROC_UNLOCK(p); 1807 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 1808 pargs_drop(pa); 1809 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { 1810 _PHOLD(p); 1811 PROC_UNLOCK(p); 1812 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1813 error = proc_getargv(curthread, p, &sb); 1814 error2 = sbuf_finish(&sb); 1815 PRELE(p); 1816 sbuf_delete(&sb); 1817 if (error == 0 && error2 != 0) 1818 error = error2; 1819 } else { 1820 PROC_UNLOCK(p); 1821 } 1822 if (error != 0 || req->newptr == NULL) 1823 return (error); 1824 1825 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit) 1826 return (ENOMEM); 1827 newpa = pargs_alloc(req->newlen); 1828 error = SYSCTL_IN(req, newpa->ar_args, req->newlen); 1829 if (error != 0) { 1830 pargs_free(newpa); 1831 return (error); 1832 } 1833 PROC_LOCK(p); 1834 pa = p->p_args; 1835 p->p_args = newpa; 1836 PROC_UNLOCK(p); 1837 pargs_drop(pa); 1838 return (0); 1839} 1840 1841/* 1842 * This sysctl allows a process to retrieve environment of another process. 1843 */ 1844static int 1845sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) 1846{ 1847 int *name = (int *)arg1; 1848 u_int namelen = arg2; 1849 struct proc *p; 1850 struct sbuf sb; 1851 int error, error2; 1852 1853 if (namelen != 1) 1854 return (EINVAL); 1855 1856 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1857 if (error != 0) 1858 return (error); 1859 if ((p->p_flag & P_SYSTEM) != 0) { 1860 PRELE(p); 1861 return (0); 1862 } 1863 1864 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1865 error = proc_getenvv(curthread, p, &sb); 1866 error2 = sbuf_finish(&sb); 1867 PRELE(p); 1868 sbuf_delete(&sb); 1869 return (error != 0 ? error : error2); 1870} 1871 1872/* 1873 * This sysctl allows a process to retrieve ELF auxiliary vector of 1874 * another process. 1875 */ 1876static int 1877sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) 1878{ 1879 int *name = (int *)arg1; 1880 u_int namelen = arg2; 1881 struct proc *p; 1882 struct sbuf sb; 1883 int error, error2; 1884 1885 if (namelen != 1) 1886 return (EINVAL); 1887 1888 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1889 if (error != 0) 1890 return (error); 1891 if ((p->p_flag & P_SYSTEM) != 0) { 1892 PRELE(p); 1893 return (0); 1894 } 1895 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1896 error = proc_getauxv(curthread, p, &sb); 1897 error2 = sbuf_finish(&sb); 1898 PRELE(p); 1899 sbuf_delete(&sb); 1900 return (error != 0 ? error : error2); 1901} 1902 1903/* 1904 * This sysctl allows a process to retrieve the path of the executable for 1905 * itself or another process. 1906 */ 1907static int 1908sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) 1909{ 1910 pid_t *pidp = (pid_t *)arg1; 1911 unsigned int arglen = arg2; 1912 struct proc *p; 1913 struct vnode *vp; 1914 char *retbuf, *freebuf; 1915 int error; 1916 1917 if (arglen != 1) 1918 return (EINVAL); 1919 if (*pidp == -1) { /* -1 means this process */ 1920 p = req->td->td_proc; 1921 } else { 1922 error = pget(*pidp, PGET_CANSEE, &p); 1923 if (error != 0) 1924 return (error); 1925 } 1926 1927 vp = p->p_textvp; 1928 if (vp == NULL) { 1929 if (*pidp != -1) 1930 PROC_UNLOCK(p); 1931 return (0); 1932 } 1933 vref(vp); 1934 if (*pidp != -1) 1935 PROC_UNLOCK(p); 1936 error = vn_fullpath(req->td, vp, &retbuf, &freebuf); 1937 vrele(vp); 1938 if (error) 1939 return (error); 1940 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); 1941 free(freebuf, M_TEMP); 1942 return (error); 1943} 1944 1945static int 1946sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) 1947{ 1948 struct proc *p; 1949 char *sv_name; 1950 int *name; 1951 int namelen; 1952 int error; 1953 1954 namelen = arg2; 1955 if (namelen != 1) 1956 return (EINVAL); 1957 1958 name = (int *)arg1; 1959 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1960 if (error != 0) 1961 return (error); 1962 sv_name = p->p_sysent->sv_name; 1963 PROC_UNLOCK(p); 1964 return (sysctl_handle_string(oidp, sv_name, 0, req)); 1965} 1966 1967#ifdef KINFO_OVMENTRY_SIZE 1968CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); 1969#endif 1970 1971#ifdef COMPAT_FREEBSD7 1972static int 1973sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) 1974{ 1975 vm_map_entry_t entry, tmp_entry; 1976 unsigned int last_timestamp; 1977 char *fullpath, *freepath; 1978 struct kinfo_ovmentry *kve; 1979 struct vattr va; 1980 struct ucred *cred; 1981 int error, *name; 1982 struct vnode *vp; 1983 struct proc *p; 1984 vm_map_t map; 1985 struct vmspace *vm; 1986 1987 name = (int *)arg1; 1988 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1989 if (error != 0) 1990 return (error); 1991 vm = vmspace_acquire_ref(p); 1992 if (vm == NULL) { 1993 PRELE(p); 1994 return (ESRCH); 1995 } 1996 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 1997 1998 map = &vm->vm_map; 1999 vm_map_lock_read(map); 2000 for (entry = map->header.next; entry != &map->header; 2001 entry = entry->next) { 2002 vm_object_t obj, tobj, lobj; 2003 vm_offset_t addr; 2004 2005 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2006 continue; 2007 2008 bzero(kve, sizeof(*kve)); 2009 kve->kve_structsize = sizeof(*kve); 2010 2011 kve->kve_private_resident = 0; 2012 obj = entry->object.vm_object; 2013 if (obj != NULL) { 2014 VM_OBJECT_RLOCK(obj); 2015 if (obj->shadow_count == 1) 2016 kve->kve_private_resident = 2017 obj->resident_page_count; 2018 } 2019 kve->kve_resident = 0; 2020 addr = entry->start; 2021 while (addr < entry->end) { 2022 if (pmap_extract(map->pmap, addr)) 2023 kve->kve_resident++; 2024 addr += PAGE_SIZE; 2025 } 2026 2027 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { 2028 if (tobj != obj) 2029 VM_OBJECT_RLOCK(tobj); 2030 if (lobj != obj) 2031 VM_OBJECT_RUNLOCK(lobj); 2032 lobj = tobj; 2033 } 2034 2035 kve->kve_start = (void*)entry->start; 2036 kve->kve_end = (void*)entry->end; 2037 kve->kve_offset = (off_t)entry->offset; 2038 2039 if (entry->protection & VM_PROT_READ) 2040 kve->kve_protection |= KVME_PROT_READ; 2041 if (entry->protection & VM_PROT_WRITE) 2042 kve->kve_protection |= KVME_PROT_WRITE; 2043 if (entry->protection & VM_PROT_EXECUTE) 2044 kve->kve_protection |= KVME_PROT_EXEC; 2045 2046 if (entry->eflags & MAP_ENTRY_COW) 2047 kve->kve_flags |= KVME_FLAG_COW; 2048 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2049 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2050 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2051 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2052 2053 last_timestamp = map->timestamp; 2054 vm_map_unlock_read(map); 2055 2056 kve->kve_fileid = 0; 2057 kve->kve_fsid = 0; 2058 freepath = NULL; 2059 fullpath = ""; 2060 if (lobj) { 2061 vp = NULL; 2062 switch (lobj->type) { 2063 case OBJT_DEFAULT: 2064 kve->kve_type = KVME_TYPE_DEFAULT; 2065 break; 2066 case OBJT_VNODE: 2067 kve->kve_type = KVME_TYPE_VNODE; 2068 vp = lobj->handle; 2069 vref(vp); 2070 break; 2071 case OBJT_SWAP: 2072 kve->kve_type = KVME_TYPE_SWAP; 2073 break; 2074 case OBJT_DEVICE: 2075 kve->kve_type = KVME_TYPE_DEVICE; 2076 break; 2077 case OBJT_PHYS: 2078 kve->kve_type = KVME_TYPE_PHYS; 2079 break; 2080 case OBJT_DEAD: 2081 kve->kve_type = KVME_TYPE_DEAD; 2082 break; 2083 case OBJT_SG: 2084 kve->kve_type = KVME_TYPE_SG; 2085 break; 2086 default: 2087 kve->kve_type = KVME_TYPE_UNKNOWN; 2088 break; 2089 } 2090 if (lobj != obj) 2091 VM_OBJECT_RUNLOCK(lobj); 2092 2093 kve->kve_ref_count = obj->ref_count; 2094 kve->kve_shadow_count = obj->shadow_count; 2095 VM_OBJECT_RUNLOCK(obj); 2096 if (vp != NULL) { 2097 vn_fullpath(curthread, vp, &fullpath, 2098 &freepath); 2099 cred = curthread->td_ucred; 2100 vn_lock(vp, LK_SHARED | LK_RETRY); 2101 if (VOP_GETATTR(vp, &va, cred) == 0) { 2102 kve->kve_fileid = va.va_fileid; 2103 kve->kve_fsid = va.va_fsid; 2104 } 2105 vput(vp); 2106 } 2107 } else { 2108 kve->kve_type = KVME_TYPE_NONE; 2109 kve->kve_ref_count = 0; 2110 kve->kve_shadow_count = 0; 2111 } 2112 2113 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2114 if (freepath != NULL) 2115 free(freepath, M_TEMP); 2116 2117 error = SYSCTL_OUT(req, kve, sizeof(*kve)); 2118 vm_map_lock_read(map); 2119 if (error) 2120 break; 2121 if (last_timestamp != map->timestamp) { 2122 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2123 entry = tmp_entry; 2124 } 2125 } 2126 vm_map_unlock_read(map); 2127 vmspace_free(vm); 2128 PRELE(p); 2129 free(kve, M_TEMP); 2130 return (error); 2131} 2132#endif /* COMPAT_FREEBSD7 */ 2133 2134#ifdef KINFO_VMENTRY_SIZE 2135CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); 2136#endif 2137 2138static void 2139kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, 2140 struct kinfo_vmentry *kve) 2141{ 2142 vm_object_t obj, tobj; 2143 vm_page_t m, m_adv; 2144 vm_offset_t addr; 2145 vm_paddr_t locked_pa; 2146 vm_pindex_t pi, pi_adv, pindex; 2147 2148 locked_pa = 0; 2149 obj = entry->object.vm_object; 2150 addr = entry->start; 2151 m_adv = NULL; 2152 pi = OFF_TO_IDX(entry->offset); 2153 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { 2154 if (m_adv != NULL) { 2155 m = m_adv; 2156 } else { 2157 pi_adv = OFF_TO_IDX(entry->end - addr); 2158 pindex = pi; 2159 for (tobj = obj;; tobj = tobj->backing_object) { 2160 m = vm_page_find_least(tobj, pindex); 2161 if (m != NULL) { 2162 if (m->pindex == pindex) 2163 break; 2164 if (pi_adv > m->pindex - pindex) { 2165 pi_adv = m->pindex - pindex; 2166 m_adv = m; 2167 } 2168 } 2169 if (tobj->backing_object == NULL) 2170 goto next; 2171 pindex += OFF_TO_IDX(tobj-> 2172 backing_object_offset); 2173 } 2174 } 2175 m_adv = NULL; 2176 if (m->psind != 0 && addr + pagesizes[1] <= entry->end && 2177 (addr & (pagesizes[1] - 1)) == 0 && 2178 (pmap_mincore(map->pmap, addr, &locked_pa) & 2179 MINCORE_SUPER) != 0) { 2180 kve->kve_flags |= KVME_FLAG_SUPER; 2181 pi_adv = OFF_TO_IDX(pagesizes[1]); 2182 } else { 2183 /* 2184 * We do not test the found page on validity. 2185 * Either the page is busy and being paged in, 2186 * or it was invalidated. The first case 2187 * should be counted as resident, the second 2188 * is not so clear; we do account both. 2189 */ 2190 pi_adv = 1; 2191 } 2192 kve->kve_resident += pi_adv; 2193next:; 2194 } 2195 PA_UNLOCK_COND(locked_pa); 2196} 2197 2198/* 2199 * Must be called with the process locked and will return unlocked. 2200 */ 2201int 2202kern_proc_vmmap_out(struct proc *p, struct sbuf *sb) 2203{ 2204 vm_map_entry_t entry, tmp_entry; 2205 struct vattr va; 2206 vm_map_t map; 2207 vm_object_t obj, tobj, lobj; 2208 char *fullpath, *freepath; 2209 struct kinfo_vmentry *kve; 2210 struct ucred *cred; 2211 struct vnode *vp; 2212 struct vmspace *vm; 2213 vm_offset_t addr; 2214 unsigned int last_timestamp; 2215 int error; 2216 2217 PROC_LOCK_ASSERT(p, MA_OWNED); 2218 2219 _PHOLD(p); 2220 PROC_UNLOCK(p); 2221 vm = vmspace_acquire_ref(p); 2222 if (vm == NULL) { 2223 PRELE(p); 2224 return (ESRCH); 2225 } 2226 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2227 2228 error = 0; 2229 map = &vm->vm_map; 2230 vm_map_lock_read(map); 2231 for (entry = map->header.next; entry != &map->header; 2232 entry = entry->next) { 2233 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2234 continue; 2235 2236 addr = entry->end; 2237 bzero(kve, sizeof(*kve)); 2238 obj = entry->object.vm_object; 2239 if (obj != NULL) { 2240 for (tobj = obj; tobj != NULL; 2241 tobj = tobj->backing_object) { 2242 VM_OBJECT_RLOCK(tobj); 2243 lobj = tobj; 2244 } 2245 if (obj->backing_object == NULL) 2246 kve->kve_private_resident = 2247 obj->resident_page_count; 2248 if (!vmmap_skip_res_cnt) 2249 kern_proc_vmmap_resident(map, entry, kve); 2250 for (tobj = obj; tobj != NULL; 2251 tobj = tobj->backing_object) { 2252 if (tobj != obj && tobj != lobj) 2253 VM_OBJECT_RUNLOCK(tobj); 2254 } 2255 } else { 2256 lobj = NULL; 2257 } 2258 2259 kve->kve_start = entry->start; 2260 kve->kve_end = entry->end; 2261 kve->kve_offset = entry->offset; 2262 2263 if (entry->protection & VM_PROT_READ) 2264 kve->kve_protection |= KVME_PROT_READ; 2265 if (entry->protection & VM_PROT_WRITE) 2266 kve->kve_protection |= KVME_PROT_WRITE; 2267 if (entry->protection & VM_PROT_EXECUTE) 2268 kve->kve_protection |= KVME_PROT_EXEC; 2269 2270 if (entry->eflags & MAP_ENTRY_COW) 2271 kve->kve_flags |= KVME_FLAG_COW; 2272 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2273 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2274 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2275 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2276 if (entry->eflags & MAP_ENTRY_GROWS_UP) 2277 kve->kve_flags |= KVME_FLAG_GROWS_UP; 2278 if (entry->eflags & MAP_ENTRY_GROWS_DOWN) 2279 kve->kve_flags |= KVME_FLAG_GROWS_DOWN; 2280 2281 last_timestamp = map->timestamp; 2282 vm_map_unlock_read(map); 2283 2284 freepath = NULL; 2285 fullpath = ""; 2286 if (lobj != NULL) { 2287 vp = NULL; 2288 switch (lobj->type) { 2289 case OBJT_DEFAULT: 2290 kve->kve_type = KVME_TYPE_DEFAULT; 2291 break; 2292 case OBJT_VNODE: 2293 kve->kve_type = KVME_TYPE_VNODE; 2294 vp = lobj->handle; 2295 vref(vp); 2296 break; 2297 case OBJT_SWAP: 2298 kve->kve_type = KVME_TYPE_SWAP; 2299 break; 2300 case OBJT_DEVICE: 2301 kve->kve_type = KVME_TYPE_DEVICE; 2302 break; 2303 case OBJT_PHYS: 2304 kve->kve_type = KVME_TYPE_PHYS; 2305 break; 2306 case OBJT_DEAD: 2307 kve->kve_type = KVME_TYPE_DEAD; 2308 break; 2309 case OBJT_SG: 2310 kve->kve_type = KVME_TYPE_SG; 2311 break; 2312 case OBJT_MGTDEVICE: 2313 kve->kve_type = KVME_TYPE_MGTDEVICE; 2314 break; 2315 default: 2316 kve->kve_type = KVME_TYPE_UNKNOWN; 2317 break; 2318 } 2319 if (lobj != obj) 2320 VM_OBJECT_RUNLOCK(lobj); 2321 2322 kve->kve_ref_count = obj->ref_count; 2323 kve->kve_shadow_count = obj->shadow_count; 2324 VM_OBJECT_RUNLOCK(obj); 2325 if (vp != NULL) { 2326 vn_fullpath(curthread, vp, &fullpath, 2327 &freepath); 2328 kve->kve_vn_type = vntype_to_kinfo(vp->v_type); 2329 cred = curthread->td_ucred; 2330 vn_lock(vp, LK_SHARED | LK_RETRY); 2331 if (VOP_GETATTR(vp, &va, cred) == 0) { 2332 kve->kve_vn_fileid = va.va_fileid; 2333 kve->kve_vn_fsid = va.va_fsid; 2334 kve->kve_vn_mode = 2335 MAKEIMODE(va.va_type, va.va_mode); 2336 kve->kve_vn_size = va.va_size; 2337 kve->kve_vn_rdev = va.va_rdev; 2338 kve->kve_status = KF_ATTR_VALID; 2339 } 2340 vput(vp); 2341 } 2342 } else { 2343 kve->kve_type = KVME_TYPE_NONE; 2344 kve->kve_ref_count = 0; 2345 kve->kve_shadow_count = 0; 2346 } 2347 2348 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2349 if (freepath != NULL) 2350 free(freepath, M_TEMP); 2351 2352 /* Pack record size down */ 2353 kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) + 2354 strlen(kve->kve_path) + 1; 2355 kve->kve_structsize = roundup(kve->kve_structsize, 2356 sizeof(uint64_t)); 2357 error = sbuf_bcat(sb, kve, kve->kve_structsize); 2358 vm_map_lock_read(map); 2359 if (error) 2360 break; 2361 if (last_timestamp != map->timestamp) { 2362 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2363 entry = tmp_entry; 2364 } 2365 } 2366 vm_map_unlock_read(map); 2367 vmspace_free(vm); 2368 PRELE(p); 2369 free(kve, M_TEMP); 2370 return (error); 2371} 2372 2373static int 2374sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) 2375{ 2376 struct proc *p; 2377 struct sbuf sb; 2378 int error, error2, *name; 2379 2380 name = (int *)arg1; 2381 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); 2382 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); 2383 if (error != 0) { 2384 sbuf_delete(&sb); 2385 return (error); 2386 } 2387 error = kern_proc_vmmap_out(p, &sb); 2388 error2 = sbuf_finish(&sb); 2389 sbuf_delete(&sb); 2390 return (error != 0 ? error : error2); 2391} 2392 2393#if defined(STACK) || defined(DDB) 2394static int 2395sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) 2396{ 2397 struct kinfo_kstack *kkstp; 2398 int error, i, *name, numthreads; 2399 lwpid_t *lwpidarray; 2400 struct thread *td; 2401 struct stack *st; 2402 struct sbuf sb; 2403 struct proc *p; 2404 2405 name = (int *)arg1; 2406 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); 2407 if (error != 0) 2408 return (error); 2409 2410 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); 2411 st = stack_create(); 2412 2413 lwpidarray = NULL; 2414 numthreads = 0; 2415 PROC_LOCK(p); 2416repeat: 2417 if (numthreads < p->p_numthreads) { 2418 if (lwpidarray != NULL) { 2419 free(lwpidarray, M_TEMP); 2420 lwpidarray = NULL; 2421 } 2422 numthreads = p->p_numthreads; 2423 PROC_UNLOCK(p); 2424 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, 2425 M_WAITOK | M_ZERO); 2426 PROC_LOCK(p); 2427 goto repeat; 2428 } 2429 i = 0; 2430 2431 /* 2432 * XXXRW: During the below loop, execve(2) and countless other sorts 2433 * of changes could have taken place. Should we check to see if the 2434 * vmspace has been replaced, or the like, in order to prevent 2435 * giving a snapshot that spans, say, execve(2), with some threads 2436 * before and some after? Among other things, the credentials could 2437 * have changed, in which case the right to extract debug info might 2438 * no longer be assured. 2439 */ 2440 FOREACH_THREAD_IN_PROC(p, td) { 2441 KASSERT(i < numthreads, 2442 ("sysctl_kern_proc_kstack: numthreads")); 2443 lwpidarray[i] = td->td_tid; 2444 i++; 2445 } 2446 numthreads = i; 2447 for (i = 0; i < numthreads; i++) { 2448 td = thread_find(p, lwpidarray[i]); 2449 if (td == NULL) { 2450 continue; 2451 } 2452 bzero(kkstp, sizeof(*kkstp)); 2453 (void)sbuf_new(&sb, kkstp->kkst_trace, 2454 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); 2455 thread_lock(td); 2456 kkstp->kkst_tid = td->td_tid; 2457 if (TD_IS_SWAPPED(td)) 2458 kkstp->kkst_state = KKST_STATE_SWAPPED; 2459 else if (TD_IS_RUNNING(td)) 2460 kkstp->kkst_state = KKST_STATE_RUNNING; 2461 else { 2462 kkstp->kkst_state = KKST_STATE_STACKOK; 2463 stack_save_td(st, td); 2464 } 2465 thread_unlock(td); 2466 PROC_UNLOCK(p); 2467 stack_sbuf_print(&sb, st); 2468 sbuf_finish(&sb); 2469 sbuf_delete(&sb); 2470 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); 2471 PROC_LOCK(p); 2472 if (error) 2473 break; 2474 } 2475 _PRELE(p); 2476 PROC_UNLOCK(p); 2477 if (lwpidarray != NULL) 2478 free(lwpidarray, M_TEMP); 2479 stack_destroy(st); 2480 free(kkstp, M_TEMP); 2481 return (error); 2482} 2483#endif 2484 2485/* 2486 * This sysctl allows a process to retrieve the full list of groups from 2487 * itself or another process. 2488 */ 2489static int 2490sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) 2491{ 2492 pid_t *pidp = (pid_t *)arg1; 2493 unsigned int arglen = arg2; 2494 struct proc *p; 2495 struct ucred *cred; 2496 int error; 2497 2498 if (arglen != 1) 2499 return (EINVAL); 2500 if (*pidp == -1) { /* -1 means this process */ 2501 p = req->td->td_proc; 2502 } else { 2503 error = pget(*pidp, PGET_CANSEE, &p); 2504 if (error != 0) 2505 return (error); 2506 } 2507 2508 cred = crhold(p->p_ucred); 2509 if (*pidp != -1) 2510 PROC_UNLOCK(p); 2511 2512 error = SYSCTL_OUT(req, cred->cr_groups, 2513 cred->cr_ngroups * sizeof(gid_t)); 2514 crfree(cred); 2515 return (error); 2516} 2517 2518/* 2519 * This sysctl allows a process to retrieve or/and set the resource limit for 2520 * another process. 2521 */ 2522static int 2523sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) 2524{ 2525 int *name = (int *)arg1; 2526 u_int namelen = arg2; 2527 struct rlimit rlim; 2528 struct proc *p; 2529 u_int which; 2530 int flags, error; 2531 2532 if (namelen != 2) 2533 return (EINVAL); 2534 2535 which = (u_int)name[1]; 2536 if (which >= RLIM_NLIMITS) 2537 return (EINVAL); 2538 2539 if (req->newptr != NULL && req->newlen != sizeof(rlim)) 2540 return (EINVAL); 2541 2542 flags = PGET_HOLD | PGET_NOTWEXIT; 2543 if (req->newptr != NULL) 2544 flags |= PGET_CANDEBUG; 2545 else 2546 flags |= PGET_CANSEE; 2547 error = pget((pid_t)name[0], flags, &p); 2548 if (error != 0) 2549 return (error); 2550 2551 /* 2552 * Retrieve limit. 2553 */ 2554 if (req->oldptr != NULL) { 2555 PROC_LOCK(p); 2556 lim_rlimit(p, which, &rlim); 2557 PROC_UNLOCK(p); 2558 } 2559 error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); 2560 if (error != 0) 2561 goto errout; 2562 2563 /* 2564 * Set limit. 2565 */ 2566 if (req->newptr != NULL) { 2567 error = SYSCTL_IN(req, &rlim, sizeof(rlim)); 2568 if (error == 0) 2569 error = kern_proc_setrlimit(curthread, p, which, &rlim); 2570 } 2571 2572errout: 2573 PRELE(p); 2574 return (error); 2575} 2576 2577/* 2578 * This sysctl allows a process to retrieve ps_strings structure location of 2579 * another process. 2580 */ 2581static int 2582sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) 2583{ 2584 int *name = (int *)arg1; 2585 u_int namelen = arg2; 2586 struct proc *p; 2587 vm_offset_t ps_strings; 2588 int error; 2589#ifdef COMPAT_FREEBSD32 2590 uint32_t ps_strings32; 2591#endif 2592 2593 if (namelen != 1) 2594 return (EINVAL); 2595 2596 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2597 if (error != 0) 2598 return (error); 2599#ifdef COMPAT_FREEBSD32 2600 if ((req->flags & SCTL_MASK32) != 0) { 2601 /* 2602 * We return 0 if the 32 bit emulation request is for a 64 bit 2603 * process. 2604 */ 2605 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? 2606 PTROUT(p->p_sysent->sv_psstrings) : 0; 2607 PROC_UNLOCK(p); 2608 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); 2609 return (error); 2610 } 2611#endif 2612 ps_strings = p->p_sysent->sv_psstrings; 2613 PROC_UNLOCK(p); 2614 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); 2615 return (error); 2616} 2617 2618/* 2619 * This sysctl allows a process to retrieve umask of another process. 2620 */ 2621static int 2622sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) 2623{ 2624 int *name = (int *)arg1; 2625 u_int namelen = arg2; 2626 struct proc *p; 2627 int error; 2628 u_short fd_cmask; 2629 2630 if (namelen != 1) 2631 return (EINVAL); 2632 2633 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2634 if (error != 0) 2635 return (error); 2636 2637 FILEDESC_SLOCK(p->p_fd); 2638 fd_cmask = p->p_fd->fd_cmask; 2639 FILEDESC_SUNLOCK(p->p_fd); 2640 PRELE(p); 2641 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask)); 2642 return (error); 2643} 2644 2645/* 2646 * This sysctl allows a process to set and retrieve binary osreldate of 2647 * another process. 2648 */ 2649static int 2650sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) 2651{ 2652 int *name = (int *)arg1; 2653 u_int namelen = arg2; 2654 struct proc *p; 2655 int flags, error, osrel; 2656 2657 if (namelen != 1) 2658 return (EINVAL); 2659 2660 if (req->newptr != NULL && req->newlen != sizeof(osrel)) 2661 return (EINVAL); 2662 2663 flags = PGET_HOLD | PGET_NOTWEXIT; 2664 if (req->newptr != NULL) 2665 flags |= PGET_CANDEBUG; 2666 else 2667 flags |= PGET_CANSEE; 2668 error = pget((pid_t)name[0], flags, &p); 2669 if (error != 0) 2670 return (error); 2671 2672 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); 2673 if (error != 0) 2674 goto errout; 2675 2676 if (req->newptr != NULL) { 2677 error = SYSCTL_IN(req, &osrel, sizeof(osrel)); 2678 if (error != 0) 2679 goto errout; 2680 if (osrel < 0) { 2681 error = EINVAL; 2682 goto errout; 2683 } 2684 p->p_osrel = osrel; 2685 } 2686errout: 2687 PRELE(p); 2688 return (error); 2689} 2690 2691static int 2692sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) 2693{ 2694 int *name = (int *)arg1; 2695 u_int namelen = arg2; 2696 struct proc *p; 2697 struct kinfo_sigtramp kst; 2698 const struct sysentvec *sv; 2699 int error; 2700#ifdef COMPAT_FREEBSD32 2701 struct kinfo_sigtramp32 kst32; 2702#endif 2703 2704 if (namelen != 1) 2705 return (EINVAL); 2706 2707 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2708 if (error != 0) 2709 return (error); 2710 sv = p->p_sysent; 2711#ifdef COMPAT_FREEBSD32 2712 if ((req->flags & SCTL_MASK32) != 0) { 2713 bzero(&kst32, sizeof(kst32)); 2714 if (SV_PROC_FLAG(p, SV_ILP32)) { 2715 if (sv->sv_sigcode_base != 0) { 2716 kst32.ksigtramp_start = sv->sv_sigcode_base; 2717 kst32.ksigtramp_end = sv->sv_sigcode_base + 2718 *sv->sv_szsigcode; 2719 } else { 2720 kst32.ksigtramp_start = sv->sv_psstrings - 2721 *sv->sv_szsigcode; 2722 kst32.ksigtramp_end = sv->sv_psstrings; 2723 } 2724 } 2725 PROC_UNLOCK(p); 2726 error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); 2727 return (error); 2728 } 2729#endif 2730 bzero(&kst, sizeof(kst)); 2731 if (sv->sv_sigcode_base != 0) { 2732 kst.ksigtramp_start = (char *)sv->sv_sigcode_base; 2733 kst.ksigtramp_end = (char *)sv->sv_sigcode_base + 2734 *sv->sv_szsigcode; 2735 } else { 2736 kst.ksigtramp_start = (char *)sv->sv_psstrings - 2737 *sv->sv_szsigcode; 2738 kst.ksigtramp_end = (char *)sv->sv_psstrings; 2739 } 2740 PROC_UNLOCK(p); 2741 error = SYSCTL_OUT(req, &kst, sizeof(kst)); 2742 return (error); 2743} 2744 2745SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table"); 2746 2747SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| 2748 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", 2749 "Return entire process table"); 2750 2751static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2752 sysctl_kern_proc, "Process table"); 2753 2754static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, 2755 sysctl_kern_proc, "Process table"); 2756 2757static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2758 sysctl_kern_proc, "Process table"); 2759 2760static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | 2761 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2762 2763static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 2764 sysctl_kern_proc, "Process table"); 2765 2766static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2767 sysctl_kern_proc, "Process table"); 2768 2769static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2770 sysctl_kern_proc, "Process table"); 2771 2772static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2773 sysctl_kern_proc, "Process table"); 2774 2775static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 2776 sysctl_kern_proc, "Return process table, no threads"); 2777 2778static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, 2779 CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, 2780 sysctl_kern_proc_args, "Process argument list"); 2781 2782static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, 2783 sysctl_kern_proc_env, "Process environment"); 2784 2785static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | 2786 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); 2787 2788static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | 2789 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); 2790 2791static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | 2792 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, 2793 "Process syscall vector name (ABI type)"); 2794 2795static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, 2796 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2797 2798static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, 2799 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2800 2801static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, 2802 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2803 2804static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), 2805 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2806 2807static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, 2808 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2809 2810static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, 2811 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2812 2813static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, 2814 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2815 2816static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, 2817 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2818 2819static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, 2820 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, 2821 "Return process table, no threads"); 2822 2823#ifdef COMPAT_FREEBSD7 2824static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | 2825 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); 2826#endif 2827 2828static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | 2829 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); 2830 2831#if defined(STACK) || defined(DDB) 2832static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | 2833 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); 2834#endif 2835 2836static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | 2837 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); 2838 2839static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | 2840 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, 2841 "Process resource limits"); 2842 2843static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | 2844 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, 2845 "Process ps_strings location"); 2846 2847static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | 2848 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); 2849 2850static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | 2851 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, 2852 "Process binary osreldate"); 2853 2854static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | 2855 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, 2856 "Process signal trampoline location"); 2857 2858int allproc_gen; 2859 2860void 2861stop_all_proc(void) 2862{ 2863 struct proc *cp, *p; 2864 int r, gen; 2865 bool restart, seen_stopped, seen_exiting, stopped_some; 2866 2867 cp = curproc; 2868 /* 2869 * stop_all_proc() assumes that all process which have 2870 * usermode must be stopped, except current process, for 2871 * obvious reasons. Since other threads in the process 2872 * establishing global stop could unstop something, disable 2873 * calls from multithreaded processes as precaution. The 2874 * service must not be user-callable anyway. 2875 */ 2876 KASSERT((cp->p_flag & P_HADTHREADS) == 0 || 2877 (cp->p_flag & P_KTHREAD) != 0, ("mt stop_all_proc")); 2878 2879allproc_loop: 2880 sx_xlock(&allproc_lock); 2881 gen = allproc_gen; 2882 seen_exiting = seen_stopped = stopped_some = restart = false; 2883 LIST_REMOVE(cp, p_list); 2884 LIST_INSERT_HEAD(&allproc, cp, p_list); 2885 for (;;) { 2886 p = LIST_NEXT(cp, p_list); 2887 if (p == NULL) 2888 break; 2889 LIST_REMOVE(cp, p_list); 2890 LIST_INSERT_AFTER(p, cp, p_list); 2891 PROC_LOCK(p); 2892 if ((p->p_flag & (P_KTHREAD | P_SYSTEM | 2893 P_TOTAL_STOP)) != 0) { 2894 PROC_UNLOCK(p); 2895 continue; 2896 } 2897 if ((p->p_flag & P_WEXIT) != 0) { 2898 seen_exiting = true; 2899 PROC_UNLOCK(p); 2900 continue; 2901 } 2902 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 2903 /* 2904 * Stopped processes are tolerated when there 2905 * are no other processes which might continue 2906 * them. P_STOPPED_SINGLE but not 2907 * P_TOTAL_STOP process still has at least one 2908 * thread running. 2909 */ 2910 seen_stopped = true; 2911 PROC_UNLOCK(p); 2912 continue; 2913 } 2914 _PHOLD(p); 2915 sx_xunlock(&allproc_lock); 2916 r = thread_single(p, SINGLE_ALLPROC); 2917 if (r != 0) 2918 restart = true; 2919 else 2920 stopped_some = true; 2921 _PRELE(p); 2922 PROC_UNLOCK(p); 2923 sx_xlock(&allproc_lock); 2924 } 2925 /* Catch forked children we did not see in iteration. */ 2926 if (gen != allproc_gen) 2927 restart = true; 2928 sx_xunlock(&allproc_lock); 2929 if (restart || stopped_some || seen_exiting || seen_stopped) { 2930 kern_yield(PRI_USER); 2931 goto allproc_loop; 2932 } 2933} 2934 2935void 2936resume_all_proc(void) 2937{ 2938 struct proc *cp, *p; 2939 2940 cp = curproc; 2941 sx_xlock(&allproc_lock); 2942 LIST_REMOVE(cp, p_list); 2943 LIST_INSERT_HEAD(&allproc, cp, p_list); 2944 for (;;) { 2945 p = LIST_NEXT(cp, p_list); 2946 if (p == NULL) 2947 break; 2948 LIST_REMOVE(cp, p_list); 2949 LIST_INSERT_AFTER(p, cp, p_list); 2950 PROC_LOCK(p); 2951 if ((p->p_flag & P_TOTAL_STOP) != 0) { 2952 sx_xunlock(&allproc_lock); 2953 _PHOLD(p); 2954 thread_single_end(p, SINGLE_ALLPROC); 2955 _PRELE(p); 2956 PROC_UNLOCK(p); 2957 sx_xlock(&allproc_lock); 2958 } else { 2959 PROC_UNLOCK(p); 2960 } 2961 } 2962 sx_xunlock(&allproc_lock); 2963} 2964 2965#define TOTAL_STOP_DEBUG 1 2966#ifdef TOTAL_STOP_DEBUG 2967volatile static int ap_resume; 2968#include <sys/mount.h> 2969 2970static int 2971sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) 2972{ 2973 int error, val; 2974 2975 val = 0; 2976 ap_resume = 0; 2977 error = sysctl_handle_int(oidp, &val, 0, req); 2978 if (error != 0 || req->newptr == NULL) 2979 return (error); 2980 if (val != 0) { 2981 stop_all_proc(); 2982 syncer_suspend(); 2983 while (ap_resume == 0) 2984 ; 2985 syncer_resume(); 2986 resume_all_proc(); 2987 } 2988 return (0); 2989} 2990 2991SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | 2992 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, 2993 sysctl_debug_stop_all_proc, "I", 2994 ""); 2995#endif 2996