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("$FreeBSD$"); 34 35 #include "opt_compat.h" 36 #include "opt_ddb.h" 37 #include "opt_ktrace.h" 38 #include "opt_kstack_pages.h" 39 #include "opt_stack.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/elf.h> 44 #include <sys/exec.h> 45 #include <sys/kernel.h> 46 #include <sys/limits.h> 47 #include <sys/lock.h> 48 #include <sys/loginclass.h> 49 #include <sys/malloc.h> 50 #include <sys/mman.h> 51 #include <sys/mount.h> 52 #include <sys/mutex.h> 53 #include <sys/proc.h> 54 #include <sys/ptrace.h> 55 #include <sys/refcount.h> 56 #include <sys/resourcevar.h> 57 #include <sys/rwlock.h> 58 #include <sys/sbuf.h> 59 #include <sys/sysent.h> 60 #include <sys/sched.h> 61 #include <sys/smp.h> 62 #include <sys/stack.h> 63 #include <sys/stat.h> 64 #include <sys/sysctl.h> 65 #include <sys/filedesc.h> 66 #include <sys/tty.h> 67 #include <sys/signalvar.h> 68 #include <sys/sdt.h> 69 #include <sys/sx.h> 70 #include <sys/user.h> 71 #include <sys/jail.h> 72 #include <sys/vnode.h> 73 #include <sys/eventhandler.h> 74 75 #ifdef DDB 76 #include <ddb/ddb.h> 77 #endif 78 79 #include <vm/vm.h> 80 #include <vm/vm_param.h> 81 #include <vm/vm_extern.h> 82 #include <vm/pmap.h> 83 #include <vm/vm_map.h> 84 #include <vm/vm_object.h> 85 #include <vm/vm_page.h> 86 #include <vm/uma.h> 87 88 #ifdef COMPAT_FREEBSD32 89 #include <compat/freebsd32/freebsd32.h> 90 #include <compat/freebsd32/freebsd32_util.h> 91 #endif 92 93 SDT_PROVIDER_DEFINE(proc); 94 SDT_PROBE_DEFINE4(proc, kernel, ctor, entry, "struct proc *", "int", 95 "void *", "int"); 96 SDT_PROBE_DEFINE4(proc, kernel, ctor, return, "struct proc *", "int", 97 "void *", "int"); 98 SDT_PROBE_DEFINE4(proc, kernel, dtor, entry, "struct proc *", "int", 99 "void *", "struct thread *"); 100 SDT_PROBE_DEFINE3(proc, kernel, dtor, return, "struct proc *", "int", 101 "void *"); 102 SDT_PROBE_DEFINE3(proc, kernel, init, entry, "struct proc *", "int", 103 "int"); 104 SDT_PROBE_DEFINE3(proc, kernel, init, return, "struct proc *", "int", 105 "int"); 106 107 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header"); 108 MALLOC_DEFINE(M_SESSION, "session", "session header"); 109 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); 110 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 111 112 static void doenterpgrp(struct proc *, struct pgrp *); 113 static void orphanpg(struct pgrp *pg); 114 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp); 115 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp); 116 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, 117 int preferthread); 118 static void pgadjustjobc(struct pgrp *pgrp, int entering); 119 static void pgdelete(struct pgrp *); 120 static int proc_ctor(void *mem, int size, void *arg, int flags); 121 static void proc_dtor(void *mem, int size, void *arg); 122 static int proc_init(void *mem, int size, int flags); 123 static void proc_fini(void *mem, int size); 124 static void pargs_free(struct pargs *pa); 125 static struct proc *zpfind_locked(pid_t pid); 126 127 /* 128 * Other process lists 129 */ 130 struct pidhashhead *pidhashtbl; 131 u_long pidhash; 132 struct pgrphashhead *pgrphashtbl; 133 u_long pgrphash; 134 struct proclist allproc; 135 struct proclist zombproc; 136 struct sx allproc_lock; 137 struct sx proctree_lock; 138 struct mtx ppeers_lock; 139 uma_zone_t proc_zone; 140 141 int kstack_pages = KSTACK_PAGES; 142 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, 143 "Kernel stack size in pages"); 144 static int vmmap_skip_res_cnt = 0; 145 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW, 146 &vmmap_skip_res_cnt, 0, 147 "Skip calculation of the pages resident count in kern.proc.vmmap"); 148 149 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); 150 #ifdef COMPAT_FREEBSD32 151 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE); 152 #endif 153 154 /* 155 * Initialize global process hashing structures. 156 */ 157 void 158 procinit() 159 { 160 161 sx_init(&allproc_lock, "allproc"); 162 sx_init(&proctree_lock, "proctree"); 163 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF); 164 LIST_INIT(&allproc); 165 LIST_INIT(&zombproc); 166 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); 167 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); 168 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(), 169 proc_ctor, proc_dtor, proc_init, proc_fini, 170 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 171 uihashinit(); 172 } 173 174 /* 175 * Prepare a proc for use. 176 */ 177 static int 178 proc_ctor(void *mem, int size, void *arg, int flags) 179 { 180 struct proc *p; 181 182 p = (struct proc *)mem; 183 SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0); 184 EVENTHANDLER_INVOKE(process_ctor, p); 185 SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0); 186 return (0); 187 } 188 189 /* 190 * Reclaim a proc after use. 191 */ 192 static void 193 proc_dtor(void *mem, int size, void *arg) 194 { 195 struct proc *p; 196 struct thread *td; 197 198 /* INVARIANTS checks go here */ 199 p = (struct proc *)mem; 200 td = FIRST_THREAD_IN_PROC(p); 201 SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0); 202 if (td != NULL) { 203 #ifdef INVARIANTS 204 KASSERT((p->p_numthreads == 1), 205 ("bad number of threads in exiting process")); 206 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr")); 207 #endif 208 /* Free all OSD associated to this thread. */ 209 osd_thread_exit(td); 210 } 211 EVENTHANDLER_INVOKE(process_dtor, p); 212 if (p->p_ksi != NULL) 213 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue")); 214 SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0); 215 } 216 217 /* 218 * Initialize type-stable parts of a proc (when newly created). 219 */ 220 static int 221 proc_init(void *mem, int size, int flags) 222 { 223 struct proc *p; 224 225 p = (struct proc *)mem; 226 SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0); 227 p->p_sched = (struct p_sched *)&p[1]; 228 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW); 229 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW); 230 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW); 231 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW); 232 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW); 233 cv_init(&p->p_pwait, "ppwait"); 234 cv_init(&p->p_dbgwait, "dbgwait"); 235 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 236 EVENTHANDLER_INVOKE(process_init, p); 237 p->p_stats = pstats_alloc(); 238 SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0); 239 return (0); 240 } 241 242 /* 243 * UMA should ensure that this function is never called. 244 * Freeing a proc structure would violate type stability. 245 */ 246 static void 247 proc_fini(void *mem, int size) 248 { 249 #ifdef notnow 250 struct proc *p; 251 252 p = (struct proc *)mem; 253 EVENTHANDLER_INVOKE(process_fini, p); 254 pstats_free(p->p_stats); 255 thread_free(FIRST_THREAD_IN_PROC(p)); 256 mtx_destroy(&p->p_mtx); 257 if (p->p_ksi != NULL) 258 ksiginfo_free(p->p_ksi); 259 #else 260 panic("proc reclaimed"); 261 #endif 262 } 263 264 /* 265 * Is p an inferior of the current process? 266 */ 267 int 268 inferior(struct proc *p) 269 { 270 271 sx_assert(&proctree_lock, SX_LOCKED); 272 PROC_LOCK_ASSERT(p, MA_OWNED); 273 for (; p != curproc; p = proc_realparent(p)) { 274 if (p->p_pid == 0) 275 return (0); 276 } 277 return (1); 278 } 279 280 struct proc * 281 pfind_locked(pid_t pid) 282 { 283 struct proc *p; 284 285 sx_assert(&allproc_lock, SX_LOCKED); 286 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 287 if (p->p_pid == pid) { 288 PROC_LOCK(p); 289 if (p->p_state == PRS_NEW) { 290 PROC_UNLOCK(p); 291 p = NULL; 292 } 293 break; 294 } 295 } 296 return (p); 297 } 298 299 /* 300 * Locate a process by number; return only "live" processes -- i.e., neither 301 * zombies nor newly born but incompletely initialized processes. By not 302 * returning processes in the PRS_NEW state, we allow callers to avoid 303 * testing for that condition to avoid dereferencing p_ucred, et al. 304 */ 305 struct proc * 306 pfind(pid_t pid) 307 { 308 struct proc *p; 309 310 sx_slock(&allproc_lock); 311 p = pfind_locked(pid); 312 sx_sunlock(&allproc_lock); 313 return (p); 314 } 315 316 static struct proc * 317 pfind_tid_locked(pid_t tid) 318 { 319 struct proc *p; 320 struct thread *td; 321 322 sx_assert(&allproc_lock, SX_LOCKED); 323 FOREACH_PROC_IN_SYSTEM(p) { 324 PROC_LOCK(p); 325 if (p->p_state == PRS_NEW) { 326 PROC_UNLOCK(p); 327 continue; 328 } 329 FOREACH_THREAD_IN_PROC(p, td) { 330 if (td->td_tid == tid) 331 goto found; 332 } 333 PROC_UNLOCK(p); 334 } 335 found: 336 return (p); 337 } 338 339 /* 340 * Locate a process group by number. 341 * The caller must hold proctree_lock. 342 */ 343 struct pgrp * 344 pgfind(pgid) 345 register pid_t pgid; 346 { 347 register struct pgrp *pgrp; 348 349 sx_assert(&proctree_lock, SX_LOCKED); 350 351 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { 352 if (pgrp->pg_id == pgid) { 353 PGRP_LOCK(pgrp); 354 return (pgrp); 355 } 356 } 357 return (NULL); 358 } 359 360 /* 361 * Locate process and do additional manipulations, depending on flags. 362 */ 363 int 364 pget(pid_t pid, int flags, struct proc **pp) 365 { 366 struct proc *p; 367 int error; 368 369 sx_slock(&allproc_lock); 370 if (pid <= PID_MAX) { 371 p = pfind_locked(pid); 372 if (p == NULL && (flags & PGET_NOTWEXIT) == 0) 373 p = zpfind_locked(pid); 374 } else if ((flags & PGET_NOTID) == 0) { 375 p = pfind_tid_locked(pid); 376 } else { 377 p = NULL; 378 } 379 sx_sunlock(&allproc_lock); 380 if (p == NULL) 381 return (ESRCH); 382 if ((flags & PGET_CANSEE) != 0) { 383 error = p_cansee(curthread, p); 384 if (error != 0) 385 goto errout; 386 } 387 if ((flags & PGET_CANDEBUG) != 0) { 388 error = p_candebug(curthread, p); 389 if (error != 0) 390 goto errout; 391 } 392 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) { 393 error = EPERM; 394 goto errout; 395 } 396 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) { 397 error = ESRCH; 398 goto errout; 399 } 400 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) { 401 /* 402 * XXXRW: Not clear ESRCH is the right error during proc 403 * execve(). 404 */ 405 error = ESRCH; 406 goto errout; 407 } 408 if ((flags & PGET_HOLD) != 0) { 409 _PHOLD(p); 410 PROC_UNLOCK(p); 411 } 412 *pp = p; 413 return (0); 414 errout: 415 PROC_UNLOCK(p); 416 return (error); 417 } 418 419 /* 420 * Create a new process group. 421 * pgid must be equal to the pid of p. 422 * Begin a new session if required. 423 */ 424 int 425 enterpgrp(p, pgid, pgrp, sess) 426 register struct proc *p; 427 pid_t pgid; 428 struct pgrp *pgrp; 429 struct session *sess; 430 { 431 432 sx_assert(&proctree_lock, SX_XLOCKED); 433 434 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL")); 435 KASSERT(p->p_pid == pgid, 436 ("enterpgrp: new pgrp and pid != pgid")); 437 KASSERT(pgfind(pgid) == NULL, 438 ("enterpgrp: pgrp with pgid exists")); 439 KASSERT(!SESS_LEADER(p), 440 ("enterpgrp: session leader attempted setpgrp")); 441 442 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 443 444 if (sess != NULL) { 445 /* 446 * new session 447 */ 448 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); 449 PROC_LOCK(p); 450 p->p_flag &= ~P_CONTROLT; 451 PROC_UNLOCK(p); 452 PGRP_LOCK(pgrp); 453 sess->s_leader = p; 454 sess->s_sid = p->p_pid; 455 refcount_init(&sess->s_count, 1); 456 sess->s_ttyvp = NULL; 457 sess->s_ttydp = NULL; 458 sess->s_ttyp = NULL; 459 bcopy(p->p_session->s_login, sess->s_login, 460 sizeof(sess->s_login)); 461 pgrp->pg_session = sess; 462 KASSERT(p == curproc, 463 ("enterpgrp: mksession and p != curproc")); 464 } else { 465 pgrp->pg_session = p->p_session; 466 sess_hold(pgrp->pg_session); 467 PGRP_LOCK(pgrp); 468 } 469 pgrp->pg_id = pgid; 470 LIST_INIT(&pgrp->pg_members); 471 472 /* 473 * As we have an exclusive lock of proctree_lock, 474 * this should not deadlock. 475 */ 476 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); 477 pgrp->pg_jobc = 0; 478 SLIST_INIT(&pgrp->pg_sigiolst); 479 PGRP_UNLOCK(pgrp); 480 481 doenterpgrp(p, pgrp); 482 483 return (0); 484 } 485 486 /* 487 * Move p to an existing process group 488 */ 489 int 490 enterthispgrp(p, pgrp) 491 register struct proc *p; 492 struct pgrp *pgrp; 493 { 494 495 sx_assert(&proctree_lock, SX_XLOCKED); 496 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 497 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 498 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 499 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 500 KASSERT(pgrp->pg_session == p->p_session, 501 ("%s: pgrp's session %p, p->p_session %p.\n", 502 __func__, 503 pgrp->pg_session, 504 p->p_session)); 505 KASSERT(pgrp != p->p_pgrp, 506 ("%s: p belongs to pgrp.", __func__)); 507 508 doenterpgrp(p, pgrp); 509 510 return (0); 511 } 512 513 /* 514 * Move p to a process group 515 */ 516 static void 517 doenterpgrp(p, pgrp) 518 struct proc *p; 519 struct pgrp *pgrp; 520 { 521 struct pgrp *savepgrp; 522 523 sx_assert(&proctree_lock, SX_XLOCKED); 524 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 525 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 526 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 527 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 528 529 savepgrp = p->p_pgrp; 530 531 /* 532 * Adjust eligibility of affected pgrps to participate in job control. 533 * Increment eligibility counts before decrementing, otherwise we 534 * could reach 0 spuriously during the first call. 535 */ 536 fixjobc(p, pgrp, 1); 537 fixjobc(p, p->p_pgrp, 0); 538 539 PGRP_LOCK(pgrp); 540 PGRP_LOCK(savepgrp); 541 PROC_LOCK(p); 542 LIST_REMOVE(p, p_pglist); 543 p->p_pgrp = pgrp; 544 PROC_UNLOCK(p); 545 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 546 PGRP_UNLOCK(savepgrp); 547 PGRP_UNLOCK(pgrp); 548 if (LIST_EMPTY(&savepgrp->pg_members)) 549 pgdelete(savepgrp); 550 } 551 552 /* 553 * remove process from process group 554 */ 555 int 556 leavepgrp(p) 557 register struct proc *p; 558 { 559 struct pgrp *savepgrp; 560 561 sx_assert(&proctree_lock, SX_XLOCKED); 562 savepgrp = p->p_pgrp; 563 PGRP_LOCK(savepgrp); 564 PROC_LOCK(p); 565 LIST_REMOVE(p, p_pglist); 566 p->p_pgrp = NULL; 567 PROC_UNLOCK(p); 568 PGRP_UNLOCK(savepgrp); 569 if (LIST_EMPTY(&savepgrp->pg_members)) 570 pgdelete(savepgrp); 571 return (0); 572 } 573 574 /* 575 * delete a process group 576 */ 577 static void 578 pgdelete(pgrp) 579 register struct pgrp *pgrp; 580 { 581 struct session *savesess; 582 struct tty *tp; 583 584 sx_assert(&proctree_lock, SX_XLOCKED); 585 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 586 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 587 588 /* 589 * Reset any sigio structures pointing to us as a result of 590 * F_SETOWN with our pgid. 591 */ 592 funsetownlst(&pgrp->pg_sigiolst); 593 594 PGRP_LOCK(pgrp); 595 tp = pgrp->pg_session->s_ttyp; 596 LIST_REMOVE(pgrp, pg_hash); 597 savesess = pgrp->pg_session; 598 PGRP_UNLOCK(pgrp); 599 600 /* Remove the reference to the pgrp before deallocating it. */ 601 if (tp != NULL) { 602 tty_lock(tp); 603 tty_rel_pgrp(tp, pgrp); 604 } 605 606 mtx_destroy(&pgrp->pg_mtx); 607 free(pgrp, M_PGRP); 608 sess_release(savesess); 609 } 610 611 static void 612 pgadjustjobc(pgrp, entering) 613 struct pgrp *pgrp; 614 int entering; 615 { 616 617 PGRP_LOCK(pgrp); 618 if (entering) 619 pgrp->pg_jobc++; 620 else { 621 --pgrp->pg_jobc; 622 if (pgrp->pg_jobc == 0) 623 orphanpg(pgrp); 624 } 625 PGRP_UNLOCK(pgrp); 626 } 627 628 /* 629 * Adjust pgrp jobc counters when specified process changes process group. 630 * We count the number of processes in each process group that "qualify" 631 * the group for terminal job control (those with a parent in a different 632 * process group of the same session). If that count reaches zero, the 633 * process group becomes orphaned. Check both the specified process' 634 * process group and that of its children. 635 * entering == 0 => p is leaving specified group. 636 * entering == 1 => p is entering specified group. 637 */ 638 void 639 fixjobc(p, pgrp, entering) 640 register struct proc *p; 641 register struct pgrp *pgrp; 642 int entering; 643 { 644 register struct pgrp *hispgrp; 645 register struct session *mysession; 646 647 sx_assert(&proctree_lock, SX_LOCKED); 648 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 649 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 650 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 651 652 /* 653 * Check p's parent to see whether p qualifies its own process 654 * group; if so, adjust count for p's process group. 655 */ 656 mysession = pgrp->pg_session; 657 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp && 658 hispgrp->pg_session == mysession) 659 pgadjustjobc(pgrp, entering); 660 661 /* 662 * Check this process' children to see whether they qualify 663 * their process groups; if so, adjust counts for children's 664 * process groups. 665 */ 666 LIST_FOREACH(p, &p->p_children, p_sibling) { 667 hispgrp = p->p_pgrp; 668 if (hispgrp == pgrp || 669 hispgrp->pg_session != mysession) 670 continue; 671 PROC_LOCK(p); 672 if (p->p_state == PRS_ZOMBIE) { 673 PROC_UNLOCK(p); 674 continue; 675 } 676 PROC_UNLOCK(p); 677 pgadjustjobc(hispgrp, entering); 678 } 679 } 680 681 /* 682 * A process group has become orphaned; 683 * if there are any stopped processes in the group, 684 * hang-up all process in that group. 685 */ 686 static void 687 orphanpg(pg) 688 struct pgrp *pg; 689 { 690 register struct proc *p; 691 692 PGRP_LOCK_ASSERT(pg, MA_OWNED); 693 694 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 695 PROC_LOCK(p); 696 if (P_SHOULDSTOP(p)) { 697 PROC_UNLOCK(p); 698 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 699 PROC_LOCK(p); 700 kern_psignal(p, SIGHUP); 701 kern_psignal(p, SIGCONT); 702 PROC_UNLOCK(p); 703 } 704 return; 705 } 706 PROC_UNLOCK(p); 707 } 708 } 709 710 void 711 sess_hold(struct session *s) 712 { 713 714 refcount_acquire(&s->s_count); 715 } 716 717 void 718 sess_release(struct session *s) 719 { 720 721 if (refcount_release(&s->s_count)) { 722 if (s->s_ttyp != NULL) { 723 tty_lock(s->s_ttyp); 724 tty_rel_sess(s->s_ttyp, s); 725 } 726 mtx_destroy(&s->s_mtx); 727 free(s, M_SESSION); 728 } 729 } 730 731 #ifdef DDB 732 733 DB_SHOW_COMMAND(pgrpdump, pgrpdump) 734 { 735 register struct pgrp *pgrp; 736 register struct proc *p; 737 register int i; 738 739 for (i = 0; i <= pgrphash; i++) { 740 if (!LIST_EMPTY(&pgrphashtbl[i])) { 741 printf("\tindx %d\n", i); 742 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { 743 printf( 744 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n", 745 (void *)pgrp, (long)pgrp->pg_id, 746 (void *)pgrp->pg_session, 747 pgrp->pg_session->s_count, 748 (void *)LIST_FIRST(&pgrp->pg_members)); 749 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 750 printf("\t\tpid %ld addr %p pgrp %p\n", 751 (long)p->p_pid, (void *)p, 752 (void *)p->p_pgrp); 753 } 754 } 755 } 756 } 757 } 758 #endif /* DDB */ 759 760 /* 761 * Calculate the kinfo_proc members which contain process-wide 762 * informations. 763 * Must be called with the target process locked. 764 */ 765 static void 766 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp) 767 { 768 struct thread *td; 769 770 PROC_LOCK_ASSERT(p, MA_OWNED); 771 772 kp->ki_estcpu = 0; 773 kp->ki_pctcpu = 0; 774 FOREACH_THREAD_IN_PROC(p, td) { 775 thread_lock(td); 776 kp->ki_pctcpu += sched_pctcpu(td); 777 kp->ki_estcpu += td->td_estcpu; 778 thread_unlock(td); 779 } 780 } 781 782 /* 783 * Clear kinfo_proc and fill in any information that is common 784 * to all threads in the process. 785 * Must be called with the target process locked. 786 */ 787 static void 788 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp) 789 { 790 struct thread *td0; 791 struct tty *tp; 792 struct session *sp; 793 struct ucred *cred; 794 struct sigacts *ps; 795 796 /* For proc_realparent. */ 797 sx_assert(&proctree_lock, SX_LOCKED); 798 PROC_LOCK_ASSERT(p, MA_OWNED); 799 bzero(kp, sizeof(*kp)); 800 801 kp->ki_structsize = sizeof(*kp); 802 kp->ki_paddr = p; 803 kp->ki_addr =/* p->p_addr; */0; /* XXX */ 804 kp->ki_args = p->p_args; 805 kp->ki_textvp = p->p_textvp; 806 #ifdef KTRACE 807 kp->ki_tracep = p->p_tracevp; 808 kp->ki_traceflag = p->p_traceflag; 809 #endif 810 kp->ki_fd = p->p_fd; 811 kp->ki_vmspace = p->p_vmspace; 812 kp->ki_flag = p->p_flag; 813 kp->ki_flag2 = p->p_flag2; 814 cred = p->p_ucred; 815 if (cred) { 816 kp->ki_uid = cred->cr_uid; 817 kp->ki_ruid = cred->cr_ruid; 818 kp->ki_svuid = cred->cr_svuid; 819 kp->ki_cr_flags = 0; 820 if (cred->cr_flags & CRED_FLAG_CAPMODE) 821 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE; 822 /* XXX bde doesn't like KI_NGROUPS */ 823 if (cred->cr_ngroups > KI_NGROUPS) { 824 kp->ki_ngroups = KI_NGROUPS; 825 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW; 826 } else 827 kp->ki_ngroups = cred->cr_ngroups; 828 bcopy(cred->cr_groups, kp->ki_groups, 829 kp->ki_ngroups * sizeof(gid_t)); 830 kp->ki_rgid = cred->cr_rgid; 831 kp->ki_svgid = cred->cr_svgid; 832 /* If jailed(cred), emulate the old P_JAILED flag. */ 833 if (jailed(cred)) { 834 kp->ki_flag |= P_JAILED; 835 /* If inside the jail, use 0 as a jail ID. */ 836 if (cred->cr_prison != curthread->td_ucred->cr_prison) 837 kp->ki_jid = cred->cr_prison->pr_id; 838 } 839 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name, 840 sizeof(kp->ki_loginclass)); 841 } 842 ps = p->p_sigacts; 843 if (ps) { 844 mtx_lock(&ps->ps_mtx); 845 kp->ki_sigignore = ps->ps_sigignore; 846 kp->ki_sigcatch = ps->ps_sigcatch; 847 mtx_unlock(&ps->ps_mtx); 848 } 849 if (p->p_state != PRS_NEW && 850 p->p_state != PRS_ZOMBIE && 851 p->p_vmspace != NULL) { 852 struct vmspace *vm = p->p_vmspace; 853 854 kp->ki_size = vm->vm_map.size; 855 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ 856 FOREACH_THREAD_IN_PROC(p, td0) { 857 if (!TD_IS_SWAPPED(td0)) 858 kp->ki_rssize += td0->td_kstack_pages; 859 } 860 kp->ki_swrss = vm->vm_swrss; 861 kp->ki_tsize = vm->vm_tsize; 862 kp->ki_dsize = vm->vm_dsize; 863 kp->ki_ssize = vm->vm_ssize; 864 } else if (p->p_state == PRS_ZOMBIE) 865 kp->ki_stat = SZOMB; 866 if (kp->ki_flag & P_INMEM) 867 kp->ki_sflag = PS_INMEM; 868 else 869 kp->ki_sflag = 0; 870 /* Calculate legacy swtime as seconds since 'swtick'. */ 871 kp->ki_swtime = (ticks - p->p_swtick) / hz; 872 kp->ki_pid = p->p_pid; 873 kp->ki_nice = p->p_nice; 874 kp->ki_fibnum = p->p_fibnum; 875 kp->ki_start = p->p_stats->p_start; 876 timevaladd(&kp->ki_start, &boottime); 877 PROC_STATLOCK(p); 878 rufetch(p, &kp->ki_rusage); 879 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime); 880 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime); 881 PROC_STATUNLOCK(p); 882 calccru(p, &kp->ki_childutime, &kp->ki_childstime); 883 /* Some callers want child times in a single value. */ 884 kp->ki_childtime = kp->ki_childstime; 885 timevaladd(&kp->ki_childtime, &kp->ki_childutime); 886 887 FOREACH_THREAD_IN_PROC(p, td0) 888 kp->ki_cow += td0->td_cow; 889 890 tp = NULL; 891 if (p->p_pgrp) { 892 kp->ki_pgid = p->p_pgrp->pg_id; 893 kp->ki_jobc = p->p_pgrp->pg_jobc; 894 sp = p->p_pgrp->pg_session; 895 896 if (sp != NULL) { 897 kp->ki_sid = sp->s_sid; 898 SESS_LOCK(sp); 899 strlcpy(kp->ki_login, sp->s_login, 900 sizeof(kp->ki_login)); 901 if (sp->s_ttyvp) 902 kp->ki_kiflag |= KI_CTTY; 903 if (SESS_LEADER(p)) 904 kp->ki_kiflag |= KI_SLEADER; 905 /* XXX proctree_lock */ 906 tp = sp->s_ttyp; 907 SESS_UNLOCK(sp); 908 } 909 } 910 if ((p->p_flag & P_CONTROLT) && tp != NULL) { 911 kp->ki_tdev = tty_udev(tp); 912 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; 913 if (tp->t_session) 914 kp->ki_tsid = tp->t_session->s_sid; 915 } else 916 kp->ki_tdev = NODEV; 917 if (p->p_comm[0] != '\0') 918 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm)); 919 if (p->p_sysent && p->p_sysent->sv_name != NULL && 920 p->p_sysent->sv_name[0] != '\0') 921 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul)); 922 kp->ki_siglist = p->p_siglist; 923 kp->ki_xstat = p->p_xstat; 924 kp->ki_acflag = p->p_acflag; 925 kp->ki_lock = p->p_lock; 926 if (p->p_pptr) { 927 kp->ki_ppid = proc_realparent(p)->p_pid; 928 if (p->p_flag & P_TRACED) 929 kp->ki_tracer = p->p_pptr->p_pid; 930 } 931 } 932 933 /* 934 * Fill in information that is thread specific. Must be called with 935 * target process locked. If 'preferthread' is set, overwrite certain 936 * process-related fields that are maintained for both threads and 937 * processes. 938 */ 939 static void 940 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread) 941 { 942 struct proc *p; 943 944 p = td->td_proc; 945 kp->ki_tdaddr = td; 946 PROC_LOCK_ASSERT(p, MA_OWNED); 947 948 if (preferthread) 949 PROC_STATLOCK(p); 950 thread_lock(td); 951 if (td->td_wmesg != NULL) 952 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg)); 953 else 954 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg)); 955 strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)); 956 if (TD_ON_LOCK(td)) { 957 kp->ki_kiflag |= KI_LOCKBLOCK; 958 strlcpy(kp->ki_lockname, td->td_lockname, 959 sizeof(kp->ki_lockname)); 960 } else { 961 kp->ki_kiflag &= ~KI_LOCKBLOCK; 962 bzero(kp->ki_lockname, sizeof(kp->ki_lockname)); 963 } 964 965 if (p->p_state == PRS_NORMAL) { /* approximate. */ 966 if (TD_ON_RUNQ(td) || 967 TD_CAN_RUN(td) || 968 TD_IS_RUNNING(td)) { 969 kp->ki_stat = SRUN; 970 } else if (P_SHOULDSTOP(p)) { 971 kp->ki_stat = SSTOP; 972 } else if (TD_IS_SLEEPING(td)) { 973 kp->ki_stat = SSLEEP; 974 } else if (TD_ON_LOCK(td)) { 975 kp->ki_stat = SLOCK; 976 } else { 977 kp->ki_stat = SWAIT; 978 } 979 } else if (p->p_state == PRS_ZOMBIE) { 980 kp->ki_stat = SZOMB; 981 } else { 982 kp->ki_stat = SIDL; 983 } 984 985 /* Things in the thread */ 986 kp->ki_wchan = td->td_wchan; 987 kp->ki_pri.pri_level = td->td_priority; 988 kp->ki_pri.pri_native = td->td_base_pri; 989 990 /* 991 * Note: legacy fields; clamp at the old NOCPU value and/or 992 * the maximum u_char CPU value. 993 */ 994 if (td->td_lastcpu == NOCPU) 995 kp->ki_lastcpu_old = NOCPU_OLD; 996 else if (td->td_lastcpu > MAXCPU_OLD) 997 kp->ki_lastcpu_old = MAXCPU_OLD; 998 else 999 kp->ki_lastcpu_old = td->td_lastcpu; 1000 1001 if (td->td_oncpu == NOCPU) 1002 kp->ki_oncpu_old = NOCPU_OLD; 1003 else if (td->td_oncpu > MAXCPU_OLD) 1004 kp->ki_oncpu_old = MAXCPU_OLD; 1005 else 1006 kp->ki_oncpu_old = td->td_oncpu; 1007 1008 kp->ki_lastcpu = td->td_lastcpu; 1009 kp->ki_oncpu = td->td_oncpu; 1010 kp->ki_tdflags = td->td_flags; 1011 kp->ki_tid = td->td_tid; 1012 kp->ki_numthreads = p->p_numthreads; 1013 kp->ki_pcb = td->td_pcb; 1014 kp->ki_kstack = (void *)td->td_kstack; 1015 kp->ki_slptime = (ticks - td->td_slptick) / hz; 1016 kp->ki_pri.pri_class = td->td_pri_class; 1017 kp->ki_pri.pri_user = td->td_user_pri; 1018 1019 if (preferthread) { 1020 rufetchtd(td, &kp->ki_rusage); 1021 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime); 1022 kp->ki_pctcpu = sched_pctcpu(td); 1023 kp->ki_estcpu = td->td_estcpu; 1024 kp->ki_cow = td->td_cow; 1025 } 1026 1027 /* We can't get this anymore but ps etc never used it anyway. */ 1028 kp->ki_rqindex = 0; 1029 1030 if (preferthread) 1031 kp->ki_siglist = td->td_siglist; 1032 kp->ki_sigmask = td->td_sigmask; 1033 thread_unlock(td); 1034 if (preferthread) 1035 PROC_STATUNLOCK(p); 1036 } 1037 1038 /* 1039 * Fill in a kinfo_proc structure for the specified process. 1040 * Must be called with the target process locked. 1041 */ 1042 void 1043 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp) 1044 { 1045 1046 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1047 1048 fill_kinfo_proc_only(p, kp); 1049 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0); 1050 fill_kinfo_aggregate(p, kp); 1051 } 1052 1053 struct pstats * 1054 pstats_alloc(void) 1055 { 1056 1057 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK)); 1058 } 1059 1060 /* 1061 * Copy parts of p_stats; zero the rest of p_stats (statistics). 1062 */ 1063 void 1064 pstats_fork(struct pstats *src, struct pstats *dst) 1065 { 1066 1067 bzero(&dst->pstat_startzero, 1068 __rangeof(struct pstats, pstat_startzero, pstat_endzero)); 1069 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy, 1070 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy)); 1071 } 1072 1073 void 1074 pstats_free(struct pstats *ps) 1075 { 1076 1077 free(ps, M_SUBPROC); 1078 } 1079 1080 static struct proc * 1081 zpfind_locked(pid_t pid) 1082 { 1083 struct proc *p; 1084 1085 sx_assert(&allproc_lock, SX_LOCKED); 1086 LIST_FOREACH(p, &zombproc, p_list) { 1087 if (p->p_pid == pid) { 1088 PROC_LOCK(p); 1089 break; 1090 } 1091 } 1092 return (p); 1093 } 1094 1095 /* 1096 * Locate a zombie process by number 1097 */ 1098 struct proc * 1099 zpfind(pid_t pid) 1100 { 1101 struct proc *p; 1102 1103 sx_slock(&allproc_lock); 1104 p = zpfind_locked(pid); 1105 sx_sunlock(&allproc_lock); 1106 return (p); 1107 } 1108 1109 #ifdef COMPAT_FREEBSD32 1110 1111 /* 1112 * This function is typically used to copy out the kernel address, so 1113 * it can be replaced by assignment of zero. 1114 */ 1115 static inline uint32_t 1116 ptr32_trim(void *ptr) 1117 { 1118 uintptr_t uptr; 1119 1120 uptr = (uintptr_t)ptr; 1121 return ((uptr > UINT_MAX) ? 0 : uptr); 1122 } 1123 1124 #define PTRTRIM_CP(src,dst,fld) \ 1125 do { (dst).fld = ptr32_trim((src).fld); } while (0) 1126 1127 static void 1128 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32) 1129 { 1130 int i; 1131 1132 bzero(ki32, sizeof(struct kinfo_proc32)); 1133 ki32->ki_structsize = sizeof(struct kinfo_proc32); 1134 CP(*ki, *ki32, ki_layout); 1135 PTRTRIM_CP(*ki, *ki32, ki_args); 1136 PTRTRIM_CP(*ki, *ki32, ki_paddr); 1137 PTRTRIM_CP(*ki, *ki32, ki_addr); 1138 PTRTRIM_CP(*ki, *ki32, ki_tracep); 1139 PTRTRIM_CP(*ki, *ki32, ki_textvp); 1140 PTRTRIM_CP(*ki, *ki32, ki_fd); 1141 PTRTRIM_CP(*ki, *ki32, ki_vmspace); 1142 PTRTRIM_CP(*ki, *ki32, ki_wchan); 1143 CP(*ki, *ki32, ki_pid); 1144 CP(*ki, *ki32, ki_ppid); 1145 CP(*ki, *ki32, ki_pgid); 1146 CP(*ki, *ki32, ki_tpgid); 1147 CP(*ki, *ki32, ki_sid); 1148 CP(*ki, *ki32, ki_tsid); 1149 CP(*ki, *ki32, ki_jobc); 1150 CP(*ki, *ki32, ki_tdev); 1151 CP(*ki, *ki32, ki_siglist); 1152 CP(*ki, *ki32, ki_sigmask); 1153 CP(*ki, *ki32, ki_sigignore); 1154 CP(*ki, *ki32, ki_sigcatch); 1155 CP(*ki, *ki32, ki_uid); 1156 CP(*ki, *ki32, ki_ruid); 1157 CP(*ki, *ki32, ki_svuid); 1158 CP(*ki, *ki32, ki_rgid); 1159 CP(*ki, *ki32, ki_svgid); 1160 CP(*ki, *ki32, ki_ngroups); 1161 for (i = 0; i < KI_NGROUPS; i++) 1162 CP(*ki, *ki32, ki_groups[i]); 1163 CP(*ki, *ki32, ki_size); 1164 CP(*ki, *ki32, ki_rssize); 1165 CP(*ki, *ki32, ki_swrss); 1166 CP(*ki, *ki32, ki_tsize); 1167 CP(*ki, *ki32, ki_dsize); 1168 CP(*ki, *ki32, ki_ssize); 1169 CP(*ki, *ki32, ki_xstat); 1170 CP(*ki, *ki32, ki_acflag); 1171 CP(*ki, *ki32, ki_pctcpu); 1172 CP(*ki, *ki32, ki_estcpu); 1173 CP(*ki, *ki32, ki_slptime); 1174 CP(*ki, *ki32, ki_swtime); 1175 CP(*ki, *ki32, ki_cow); 1176 CP(*ki, *ki32, ki_runtime); 1177 TV_CP(*ki, *ki32, ki_start); 1178 TV_CP(*ki, *ki32, ki_childtime); 1179 CP(*ki, *ki32, ki_flag); 1180 CP(*ki, *ki32, ki_kiflag); 1181 CP(*ki, *ki32, ki_traceflag); 1182 CP(*ki, *ki32, ki_stat); 1183 CP(*ki, *ki32, ki_nice); 1184 CP(*ki, *ki32, ki_lock); 1185 CP(*ki, *ki32, ki_rqindex); 1186 CP(*ki, *ki32, ki_oncpu); 1187 CP(*ki, *ki32, ki_lastcpu); 1188 1189 /* XXX TODO: wrap cpu value as appropriate */ 1190 CP(*ki, *ki32, ki_oncpu_old); 1191 CP(*ki, *ki32, ki_lastcpu_old); 1192 1193 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1); 1194 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1); 1195 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1); 1196 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1); 1197 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1); 1198 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1); 1199 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1); 1200 CP(*ki, *ki32, ki_tracer); 1201 CP(*ki, *ki32, ki_flag2); 1202 CP(*ki, *ki32, ki_fibnum); 1203 CP(*ki, *ki32, ki_cr_flags); 1204 CP(*ki, *ki32, ki_jid); 1205 CP(*ki, *ki32, ki_numthreads); 1206 CP(*ki, *ki32, ki_tid); 1207 CP(*ki, *ki32, ki_pri); 1208 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage); 1209 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch); 1210 PTRTRIM_CP(*ki, *ki32, ki_pcb); 1211 PTRTRIM_CP(*ki, *ki32, ki_kstack); 1212 PTRTRIM_CP(*ki, *ki32, ki_udata); 1213 CP(*ki, *ki32, ki_sflag); 1214 CP(*ki, *ki32, ki_tdflags); 1215 } 1216 #endif 1217 1218 int 1219 kern_proc_out(struct proc *p, struct sbuf *sb, int flags) 1220 { 1221 struct thread *td; 1222 struct kinfo_proc ki; 1223 #ifdef COMPAT_FREEBSD32 1224 struct kinfo_proc32 ki32; 1225 #endif 1226 int error; 1227 1228 PROC_LOCK_ASSERT(p, MA_OWNED); 1229 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1230 1231 error = 0; 1232 fill_kinfo_proc(p, &ki); 1233 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1234 #ifdef COMPAT_FREEBSD32 1235 if ((flags & KERN_PROC_MASK32) != 0) { 1236 freebsd32_kinfo_proc_out(&ki, &ki32); 1237 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1238 error = ENOMEM; 1239 } else 1240 #endif 1241 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1242 error = ENOMEM; 1243 } else { 1244 FOREACH_THREAD_IN_PROC(p, td) { 1245 fill_kinfo_thread(td, &ki, 1); 1246 #ifdef COMPAT_FREEBSD32 1247 if ((flags & KERN_PROC_MASK32) != 0) { 1248 freebsd32_kinfo_proc_out(&ki, &ki32); 1249 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1250 error = ENOMEM; 1251 } else 1252 #endif 1253 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1254 error = ENOMEM; 1255 if (error != 0) 1256 break; 1257 } 1258 } 1259 PROC_UNLOCK(p); 1260 return (error); 1261 } 1262 1263 static int 1264 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags, 1265 int doingzomb) 1266 { 1267 struct sbuf sb; 1268 struct kinfo_proc ki; 1269 struct proc *np; 1270 int error, error2; 1271 pid_t pid; 1272 1273 pid = p->p_pid; 1274 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); 1275 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1276 error = kern_proc_out(p, &sb, flags); 1277 error2 = sbuf_finish(&sb); 1278 sbuf_delete(&sb); 1279 if (error != 0) 1280 return (error); 1281 else if (error2 != 0) 1282 return (error2); 1283 if (doingzomb) 1284 np = zpfind(pid); 1285 else { 1286 if (pid == 0) 1287 return (0); 1288 np = pfind(pid); 1289 } 1290 if (np == NULL) 1291 return (ESRCH); 1292 if (np != p) { 1293 PROC_UNLOCK(np); 1294 return (ESRCH); 1295 } 1296 PROC_UNLOCK(np); 1297 return (0); 1298 } 1299 1300 static int 1301 sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 1302 { 1303 int *name = (int *)arg1; 1304 u_int namelen = arg2; 1305 struct proc *p; 1306 int flags, doingzomb, oid_number; 1307 int error = 0; 1308 1309 oid_number = oidp->oid_number; 1310 if (oid_number != KERN_PROC_ALL && 1311 (oid_number & KERN_PROC_INC_THREAD) == 0) 1312 flags = KERN_PROC_NOTHREADS; 1313 else { 1314 flags = 0; 1315 oid_number &= ~KERN_PROC_INC_THREAD; 1316 } 1317 #ifdef COMPAT_FREEBSD32 1318 if (req->flags & SCTL_MASK32) 1319 flags |= KERN_PROC_MASK32; 1320 #endif 1321 if (oid_number == KERN_PROC_PID) { 1322 if (namelen != 1) 1323 return (EINVAL); 1324 error = sysctl_wire_old_buffer(req, 0); 1325 if (error) 1326 return (error); 1327 sx_slock(&proctree_lock); 1328 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1329 if (error == 0) 1330 error = sysctl_out_proc(p, req, flags, 0); 1331 sx_sunlock(&proctree_lock); 1332 return (error); 1333 } 1334 1335 switch (oid_number) { 1336 case KERN_PROC_ALL: 1337 if (namelen != 0) 1338 return (EINVAL); 1339 break; 1340 case KERN_PROC_PROC: 1341 if (namelen != 0 && namelen != 1) 1342 return (EINVAL); 1343 break; 1344 default: 1345 if (namelen != 1) 1346 return (EINVAL); 1347 break; 1348 } 1349 1350 if (!req->oldptr) { 1351 /* overestimate by 5 procs */ 1352 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1353 if (error) 1354 return (error); 1355 } 1356 error = sysctl_wire_old_buffer(req, 0); 1357 if (error != 0) 1358 return (error); 1359 sx_slock(&proctree_lock); 1360 sx_slock(&allproc_lock); 1361 for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) { 1362 if (!doingzomb) 1363 p = LIST_FIRST(&allproc); 1364 else 1365 p = LIST_FIRST(&zombproc); 1366 for (; p != 0; p = LIST_NEXT(p, p_list)) { 1367 /* 1368 * Skip embryonic processes. 1369 */ 1370 PROC_LOCK(p); 1371 if (p->p_state == PRS_NEW) { 1372 PROC_UNLOCK(p); 1373 continue; 1374 } 1375 KASSERT(p->p_ucred != NULL, 1376 ("process credential is NULL for non-NEW proc")); 1377 /* 1378 * Show a user only appropriate processes. 1379 */ 1380 if (p_cansee(curthread, p)) { 1381 PROC_UNLOCK(p); 1382 continue; 1383 } 1384 /* 1385 * TODO - make more efficient (see notes below). 1386 * do by session. 1387 */ 1388 switch (oid_number) { 1389 1390 case KERN_PROC_GID: 1391 if (p->p_ucred->cr_gid != (gid_t)name[0]) { 1392 PROC_UNLOCK(p); 1393 continue; 1394 } 1395 break; 1396 1397 case KERN_PROC_PGRP: 1398 /* could do this by traversing pgrp */ 1399 if (p->p_pgrp == NULL || 1400 p->p_pgrp->pg_id != (pid_t)name[0]) { 1401 PROC_UNLOCK(p); 1402 continue; 1403 } 1404 break; 1405 1406 case KERN_PROC_RGID: 1407 if (p->p_ucred->cr_rgid != (gid_t)name[0]) { 1408 PROC_UNLOCK(p); 1409 continue; 1410 } 1411 break; 1412 1413 case KERN_PROC_SESSION: 1414 if (p->p_session == NULL || 1415 p->p_session->s_sid != (pid_t)name[0]) { 1416 PROC_UNLOCK(p); 1417 continue; 1418 } 1419 break; 1420 1421 case KERN_PROC_TTY: 1422 if ((p->p_flag & P_CONTROLT) == 0 || 1423 p->p_session == NULL) { 1424 PROC_UNLOCK(p); 1425 continue; 1426 } 1427 /* XXX proctree_lock */ 1428 SESS_LOCK(p->p_session); 1429 if (p->p_session->s_ttyp == NULL || 1430 tty_udev(p->p_session->s_ttyp) != 1431 (dev_t)name[0]) { 1432 SESS_UNLOCK(p->p_session); 1433 PROC_UNLOCK(p); 1434 continue; 1435 } 1436 SESS_UNLOCK(p->p_session); 1437 break; 1438 1439 case KERN_PROC_UID: 1440 if (p->p_ucred->cr_uid != (uid_t)name[0]) { 1441 PROC_UNLOCK(p); 1442 continue; 1443 } 1444 break; 1445 1446 case KERN_PROC_RUID: 1447 if (p->p_ucred->cr_ruid != (uid_t)name[0]) { 1448 PROC_UNLOCK(p); 1449 continue; 1450 } 1451 break; 1452 1453 case KERN_PROC_PROC: 1454 break; 1455 1456 default: 1457 break; 1458 1459 } 1460 1461 error = sysctl_out_proc(p, req, flags, doingzomb); 1462 if (error) { 1463 sx_sunlock(&allproc_lock); 1464 sx_sunlock(&proctree_lock); 1465 return (error); 1466 } 1467 } 1468 } 1469 sx_sunlock(&allproc_lock); 1470 sx_sunlock(&proctree_lock); 1471 return (0); 1472 } 1473 1474 struct pargs * 1475 pargs_alloc(int len) 1476 { 1477 struct pargs *pa; 1478 1479 pa = malloc(sizeof(struct pargs) + len, M_PARGS, 1480 M_WAITOK); 1481 refcount_init(&pa->ar_ref, 1); 1482 pa->ar_length = len; 1483 return (pa); 1484 } 1485 1486 static void 1487 pargs_free(struct pargs *pa) 1488 { 1489 1490 free(pa, M_PARGS); 1491 } 1492 1493 void 1494 pargs_hold(struct pargs *pa) 1495 { 1496 1497 if (pa == NULL) 1498 return; 1499 refcount_acquire(&pa->ar_ref); 1500 } 1501 1502 void 1503 pargs_drop(struct pargs *pa) 1504 { 1505 1506 if (pa == NULL) 1507 return; 1508 if (refcount_release(&pa->ar_ref)) 1509 pargs_free(pa); 1510 } 1511 1512 static int 1513 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf, 1514 size_t len) 1515 { 1516 struct iovec iov; 1517 struct uio uio; 1518 1519 iov.iov_base = (caddr_t)buf; 1520 iov.iov_len = len; 1521 uio.uio_iov = &iov; 1522 uio.uio_iovcnt = 1; 1523 uio.uio_offset = offset; 1524 uio.uio_resid = (ssize_t)len; 1525 uio.uio_segflg = UIO_SYSSPACE; 1526 uio.uio_rw = UIO_READ; 1527 uio.uio_td = td; 1528 1529 return (proc_rwmem(p, &uio)); 1530 } 1531 1532 static int 1533 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, 1534 size_t len) 1535 { 1536 size_t i; 1537 int error; 1538 1539 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len); 1540 /* 1541 * Reading the chunk may validly return EFAULT if the string is shorter 1542 * than the chunk and is aligned at the end of the page, assuming the 1543 * next page is not mapped. So if EFAULT is returned do a fallback to 1544 * one byte read loop. 1545 */ 1546 if (error == EFAULT) { 1547 for (i = 0; i < len; i++, buf++, sptr++) { 1548 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1); 1549 if (error != 0) 1550 return (error); 1551 if (*buf == '\0') 1552 break; 1553 } 1554 error = 0; 1555 } 1556 return (error); 1557 } 1558 1559 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ 1560 1561 enum proc_vector_type { 1562 PROC_ARG, 1563 PROC_ENV, 1564 PROC_AUX, 1565 }; 1566 1567 #ifdef COMPAT_FREEBSD32 1568 static int 1569 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, 1570 size_t *vsizep, enum proc_vector_type type) 1571 { 1572 struct freebsd32_ps_strings pss; 1573 Elf32_Auxinfo aux; 1574 vm_offset_t vptr, ptr; 1575 uint32_t *proc_vector32; 1576 char **proc_vector; 1577 size_t vsize, size; 1578 int i, error; 1579 1580 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), 1581 &pss, sizeof(pss)); 1582 if (error != 0) 1583 return (error); 1584 switch (type) { 1585 case PROC_ARG: 1586 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); 1587 vsize = pss.ps_nargvstr; 1588 if (vsize > ARG_MAX) 1589 return (ENOEXEC); 1590 size = vsize * sizeof(int32_t); 1591 break; 1592 case PROC_ENV: 1593 vptr = (vm_offset_t)PTRIN(pss.ps_envstr); 1594 vsize = pss.ps_nenvstr; 1595 if (vsize > ARG_MAX) 1596 return (ENOEXEC); 1597 size = vsize * sizeof(int32_t); 1598 break; 1599 case PROC_AUX: 1600 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + 1601 (pss.ps_nenvstr + 1) * sizeof(int32_t); 1602 if (vptr % 4 != 0) 1603 return (ENOEXEC); 1604 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1605 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); 1606 if (error != 0) 1607 return (error); 1608 if (aux.a_type == AT_NULL) 1609 break; 1610 ptr += sizeof(aux); 1611 } 1612 if (aux.a_type != AT_NULL) 1613 return (ENOEXEC); 1614 vsize = i + 1; 1615 size = vsize * sizeof(aux); 1616 break; 1617 default: 1618 KASSERT(0, ("Wrong proc vector type: %d", type)); 1619 return (EINVAL); 1620 } 1621 proc_vector32 = malloc(size, M_TEMP, M_WAITOK); 1622 error = proc_read_mem(td, p, vptr, proc_vector32, size); 1623 if (error != 0) 1624 goto done; 1625 if (type == PROC_AUX) { 1626 *proc_vectorp = (char **)proc_vector32; 1627 *vsizep = vsize; 1628 return (0); 1629 } 1630 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); 1631 for (i = 0; i < (int)vsize; i++) 1632 proc_vector[i] = PTRIN(proc_vector32[i]); 1633 *proc_vectorp = proc_vector; 1634 *vsizep = vsize; 1635 done: 1636 free(proc_vector32, M_TEMP); 1637 return (error); 1638 } 1639 #endif 1640 1641 static int 1642 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, 1643 size_t *vsizep, enum proc_vector_type type) 1644 { 1645 struct ps_strings pss; 1646 Elf_Auxinfo aux; 1647 vm_offset_t vptr, ptr; 1648 char **proc_vector; 1649 size_t vsize, size; 1650 int error, i; 1651 1652 #ifdef COMPAT_FREEBSD32 1653 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1654 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); 1655 #endif 1656 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), 1657 &pss, sizeof(pss)); 1658 if (error != 0) 1659 return (error); 1660 switch (type) { 1661 case PROC_ARG: 1662 vptr = (vm_offset_t)pss.ps_argvstr; 1663 vsize = pss.ps_nargvstr; 1664 if (vsize > ARG_MAX) 1665 return (ENOEXEC); 1666 size = vsize * sizeof(char *); 1667 break; 1668 case PROC_ENV: 1669 vptr = (vm_offset_t)pss.ps_envstr; 1670 vsize = pss.ps_nenvstr; 1671 if (vsize > ARG_MAX) 1672 return (ENOEXEC); 1673 size = vsize * sizeof(char *); 1674 break; 1675 case PROC_AUX: 1676 /* 1677 * The aux array is just above env array on the stack. Check 1678 * that the address is naturally aligned. 1679 */ 1680 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) 1681 * sizeof(char *); 1682 #if __ELF_WORD_SIZE == 64 1683 if (vptr % sizeof(uint64_t) != 0) 1684 #else 1685 if (vptr % sizeof(uint32_t) != 0) 1686 #endif 1687 return (ENOEXEC); 1688 /* 1689 * We count the array size reading the aux vectors from the 1690 * stack until AT_NULL vector is returned. So (to keep the code 1691 * simple) we read the process stack twice: the first time here 1692 * to find the size and the second time when copying the vectors 1693 * to the allocated proc_vector. 1694 */ 1695 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1696 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); 1697 if (error != 0) 1698 return (error); 1699 if (aux.a_type == AT_NULL) 1700 break; 1701 ptr += sizeof(aux); 1702 } 1703 /* 1704 * If the PROC_AUXV_MAX entries are iterated over, and we have 1705 * not reached AT_NULL, it is most likely we are reading wrong 1706 * data: either the process doesn't have auxv array or data has 1707 * been modified. Return the error in this case. 1708 */ 1709 if (aux.a_type != AT_NULL) 1710 return (ENOEXEC); 1711 vsize = i + 1; 1712 size = vsize * sizeof(aux); 1713 break; 1714 default: 1715 KASSERT(0, ("Wrong proc vector type: %d", type)); 1716 return (EINVAL); /* In case we are built without INVARIANTS. */ 1717 } 1718 proc_vector = malloc(size, M_TEMP, M_WAITOK); 1719 if (proc_vector == NULL) 1720 return (ENOMEM); 1721 error = proc_read_mem(td, p, vptr, proc_vector, size); 1722 if (error != 0) { 1723 free(proc_vector, M_TEMP); 1724 return (error); 1725 } 1726 *proc_vectorp = proc_vector; 1727 *vsizep = vsize; 1728 1729 return (0); 1730 } 1731 1732 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ 1733 1734 static int 1735 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, 1736 enum proc_vector_type type) 1737 { 1738 size_t done, len, nchr, vsize; 1739 int error, i; 1740 char **proc_vector, *sptr; 1741 char pss_string[GET_PS_STRINGS_CHUNK_SZ]; 1742 1743 PROC_ASSERT_HELD(p); 1744 1745 /* 1746 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. 1747 */ 1748 nchr = 2 * (PATH_MAX + ARG_MAX); 1749 1750 error = get_proc_vector(td, p, &proc_vector, &vsize, type); 1751 if (error != 0) 1752 return (error); 1753 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { 1754 /* 1755 * The program may have scribbled into its argv array, e.g. to 1756 * remove some arguments. If that has happened, break out 1757 * before trying to read from NULL. 1758 */ 1759 if (proc_vector[i] == NULL) 1760 break; 1761 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { 1762 error = proc_read_string(td, p, sptr, pss_string, 1763 sizeof(pss_string)); 1764 if (error != 0) 1765 goto done; 1766 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); 1767 if (done + len >= nchr) 1768 len = nchr - done - 1; 1769 sbuf_bcat(sb, pss_string, len); 1770 if (len != GET_PS_STRINGS_CHUNK_SZ) 1771 break; 1772 done += GET_PS_STRINGS_CHUNK_SZ; 1773 } 1774 sbuf_bcat(sb, "", 1); 1775 done += len + 1; 1776 } 1777 done: 1778 free(proc_vector, M_TEMP); 1779 return (error); 1780 } 1781 1782 int 1783 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) 1784 { 1785 1786 return (get_ps_strings(curthread, p, sb, PROC_ARG)); 1787 } 1788 1789 int 1790 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) 1791 { 1792 1793 return (get_ps_strings(curthread, p, sb, PROC_ENV)); 1794 } 1795 1796 int 1797 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) 1798 { 1799 size_t vsize, size; 1800 char **auxv; 1801 int error; 1802 1803 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); 1804 if (error == 0) { 1805 #ifdef COMPAT_FREEBSD32 1806 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1807 size = vsize * sizeof(Elf32_Auxinfo); 1808 else 1809 #endif 1810 size = vsize * sizeof(Elf_Auxinfo); 1811 if (sbuf_bcat(sb, auxv, size) != 0) 1812 error = ENOMEM; 1813 free(auxv, M_TEMP); 1814 } 1815 return (error); 1816 } 1817 1818 /* 1819 * This sysctl allows a process to retrieve the argument list or process 1820 * title for another process without groping around in the address space 1821 * of the other process. It also allow a process to set its own "process 1822 * title to a string of its own choice. 1823 */ 1824 static int 1825 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 1826 { 1827 int *name = (int *)arg1; 1828 u_int namelen = arg2; 1829 struct pargs *newpa, *pa; 1830 struct proc *p; 1831 struct sbuf sb; 1832 int flags, error = 0, error2; 1833 1834 if (namelen != 1) 1835 return (EINVAL); 1836 1837 flags = PGET_CANSEE; 1838 if (req->newptr != NULL) 1839 flags |= PGET_ISCURRENT; 1840 error = pget((pid_t)name[0], flags, &p); 1841 if (error) 1842 return (error); 1843 1844 pa = p->p_args; 1845 if (pa != NULL) { 1846 pargs_hold(pa); 1847 PROC_UNLOCK(p); 1848 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 1849 pargs_drop(pa); 1850 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { 1851 _PHOLD(p); 1852 PROC_UNLOCK(p); 1853 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1854 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1855 error = proc_getargv(curthread, p, &sb); 1856 error2 = sbuf_finish(&sb); 1857 PRELE(p); 1858 sbuf_delete(&sb); 1859 if (error == 0 && error2 != 0) 1860 error = error2; 1861 } else { 1862 PROC_UNLOCK(p); 1863 } 1864 if (error != 0 || req->newptr == NULL) 1865 return (error); 1866 1867 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit) 1868 return (ENOMEM); 1869 newpa = pargs_alloc(req->newlen); 1870 error = SYSCTL_IN(req, newpa->ar_args, req->newlen); 1871 if (error != 0) { 1872 pargs_free(newpa); 1873 return (error); 1874 } 1875 PROC_LOCK(p); 1876 pa = p->p_args; 1877 p->p_args = newpa; 1878 PROC_UNLOCK(p); 1879 pargs_drop(pa); 1880 return (0); 1881 } 1882 1883 /* 1884 * This sysctl allows a process to retrieve environment of another process. 1885 */ 1886 static int 1887 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) 1888 { 1889 int *name = (int *)arg1; 1890 u_int namelen = arg2; 1891 struct proc *p; 1892 struct sbuf sb; 1893 int error, error2; 1894 1895 if (namelen != 1) 1896 return (EINVAL); 1897 1898 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1899 if (error != 0) 1900 return (error); 1901 if ((p->p_flag & P_SYSTEM) != 0) { 1902 PRELE(p); 1903 return (0); 1904 } 1905 1906 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1907 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1908 error = proc_getenvv(curthread, p, &sb); 1909 error2 = sbuf_finish(&sb); 1910 PRELE(p); 1911 sbuf_delete(&sb); 1912 return (error != 0 ? error : error2); 1913 } 1914 1915 /* 1916 * This sysctl allows a process to retrieve ELF auxiliary vector of 1917 * another process. 1918 */ 1919 static int 1920 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) 1921 { 1922 int *name = (int *)arg1; 1923 u_int namelen = arg2; 1924 struct proc *p; 1925 struct sbuf sb; 1926 int error, error2; 1927 1928 if (namelen != 1) 1929 return (EINVAL); 1930 1931 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1932 if (error != 0) 1933 return (error); 1934 if ((p->p_flag & P_SYSTEM) != 0) { 1935 PRELE(p); 1936 return (0); 1937 } 1938 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1939 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1940 error = proc_getauxv(curthread, p, &sb); 1941 error2 = sbuf_finish(&sb); 1942 PRELE(p); 1943 sbuf_delete(&sb); 1944 return (error != 0 ? error : error2); 1945 } 1946 1947 /* 1948 * This sysctl allows a process to retrieve the path of the executable for 1949 * itself or another process. 1950 */ 1951 static int 1952 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) 1953 { 1954 pid_t *pidp = (pid_t *)arg1; 1955 unsigned int arglen = arg2; 1956 struct proc *p; 1957 struct vnode *vp; 1958 char *retbuf, *freebuf; 1959 int error; 1960 1961 if (arglen != 1) 1962 return (EINVAL); 1963 if (*pidp == -1) { /* -1 means this process */ 1964 p = req->td->td_proc; 1965 } else { 1966 error = pget(*pidp, PGET_CANSEE, &p); 1967 if (error != 0) 1968 return (error); 1969 } 1970 1971 vp = p->p_textvp; 1972 if (vp == NULL) { 1973 if (*pidp != -1) 1974 PROC_UNLOCK(p); 1975 return (0); 1976 } 1977 vref(vp); 1978 if (*pidp != -1) 1979 PROC_UNLOCK(p); 1980 error = vn_fullpath(req->td, vp, &retbuf, &freebuf); 1981 vrele(vp); 1982 if (error) 1983 return (error); 1984 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); 1985 free(freebuf, M_TEMP); 1986 return (error); 1987 } 1988 1989 static int 1990 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) 1991 { 1992 struct proc *p; 1993 char *sv_name; 1994 int *name; 1995 int namelen; 1996 int error; 1997 1998 namelen = arg2; 1999 if (namelen != 1) 2000 return (EINVAL); 2001 2002 name = (int *)arg1; 2003 error = pget((pid_t)name[0], PGET_CANSEE, &p); 2004 if (error != 0) 2005 return (error); 2006 sv_name = p->p_sysent->sv_name; 2007 PROC_UNLOCK(p); 2008 return (sysctl_handle_string(oidp, sv_name, 0, req)); 2009 } 2010 2011 #ifdef KINFO_OVMENTRY_SIZE 2012 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); 2013 #endif 2014 2015 #ifdef COMPAT_FREEBSD7 2016 static int 2017 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) 2018 { 2019 vm_map_entry_t entry, tmp_entry; 2020 unsigned int last_timestamp; 2021 char *fullpath, *freepath; 2022 struct kinfo_ovmentry *kve; 2023 struct vattr va; 2024 struct ucred *cred; 2025 int error, *name; 2026 struct vnode *vp; 2027 struct proc *p; 2028 vm_map_t map; 2029 struct vmspace *vm; 2030 2031 name = (int *)arg1; 2032 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2033 if (error != 0) 2034 return (error); 2035 vm = vmspace_acquire_ref(p); 2036 if (vm == NULL) { 2037 PRELE(p); 2038 return (ESRCH); 2039 } 2040 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2041 2042 map = &vm->vm_map; 2043 vm_map_lock_read(map); 2044 for (entry = map->header.next; entry != &map->header; 2045 entry = entry->next) { 2046 vm_object_t obj, tobj, lobj; 2047 vm_offset_t addr; 2048 2049 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2050 continue; 2051 2052 bzero(kve, sizeof(*kve)); 2053 kve->kve_structsize = sizeof(*kve); 2054 2055 kve->kve_private_resident = 0; 2056 obj = entry->object.vm_object; 2057 if (obj != NULL) { 2058 VM_OBJECT_RLOCK(obj); 2059 if (obj->shadow_count == 1) 2060 kve->kve_private_resident = 2061 obj->resident_page_count; 2062 } 2063 kve->kve_resident = 0; 2064 addr = entry->start; 2065 while (addr < entry->end) { 2066 if (pmap_extract(map->pmap, addr)) 2067 kve->kve_resident++; 2068 addr += PAGE_SIZE; 2069 } 2070 2071 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { 2072 if (tobj != obj) 2073 VM_OBJECT_RLOCK(tobj); 2074 if (lobj != obj) 2075 VM_OBJECT_RUNLOCK(lobj); 2076 lobj = tobj; 2077 } 2078 2079 kve->kve_start = (void*)entry->start; 2080 kve->kve_end = (void*)entry->end; 2081 kve->kve_offset = (off_t)entry->offset; 2082 2083 if (entry->protection & VM_PROT_READ) 2084 kve->kve_protection |= KVME_PROT_READ; 2085 if (entry->protection & VM_PROT_WRITE) 2086 kve->kve_protection |= KVME_PROT_WRITE; 2087 if (entry->protection & VM_PROT_EXECUTE) 2088 kve->kve_protection |= KVME_PROT_EXEC; 2089 2090 if (entry->eflags & MAP_ENTRY_COW) 2091 kve->kve_flags |= KVME_FLAG_COW; 2092 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2093 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2094 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2095 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2096 2097 last_timestamp = map->timestamp; 2098 vm_map_unlock_read(map); 2099 2100 kve->kve_fileid = 0; 2101 kve->kve_fsid = 0; 2102 freepath = NULL; 2103 fullpath = ""; 2104 if (lobj) { 2105 vp = NULL; 2106 switch (lobj->type) { 2107 case OBJT_DEFAULT: 2108 kve->kve_type = KVME_TYPE_DEFAULT; 2109 break; 2110 case OBJT_VNODE: 2111 kve->kve_type = KVME_TYPE_VNODE; 2112 vp = lobj->handle; 2113 vref(vp); 2114 break; 2115 case OBJT_SWAP: 2116 kve->kve_type = KVME_TYPE_SWAP; 2117 break; 2118 case OBJT_DEVICE: 2119 kve->kve_type = KVME_TYPE_DEVICE; 2120 break; 2121 case OBJT_PHYS: 2122 kve->kve_type = KVME_TYPE_PHYS; 2123 break; 2124 case OBJT_DEAD: 2125 kve->kve_type = KVME_TYPE_DEAD; 2126 break; 2127 case OBJT_SG: 2128 kve->kve_type = KVME_TYPE_SG; 2129 break; 2130 default: 2131 kve->kve_type = KVME_TYPE_UNKNOWN; 2132 break; 2133 } 2134 if (lobj != obj) 2135 VM_OBJECT_RUNLOCK(lobj); 2136 2137 kve->kve_ref_count = obj->ref_count; 2138 kve->kve_shadow_count = obj->shadow_count; 2139 VM_OBJECT_RUNLOCK(obj); 2140 if (vp != NULL) { 2141 vn_fullpath(curthread, vp, &fullpath, 2142 &freepath); 2143 cred = curthread->td_ucred; 2144 vn_lock(vp, LK_SHARED | LK_RETRY); 2145 if (VOP_GETATTR(vp, &va, cred) == 0) { 2146 kve->kve_fileid = va.va_fileid; 2147 kve->kve_fsid = va.va_fsid; 2148 } 2149 vput(vp); 2150 } 2151 } else { 2152 kve->kve_type = KVME_TYPE_NONE; 2153 kve->kve_ref_count = 0; 2154 kve->kve_shadow_count = 0; 2155 } 2156 2157 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2158 if (freepath != NULL) 2159 free(freepath, M_TEMP); 2160 2161 error = SYSCTL_OUT(req, kve, sizeof(*kve)); 2162 vm_map_lock_read(map); 2163 if (error) 2164 break; 2165 if (last_timestamp != map->timestamp) { 2166 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2167 entry = tmp_entry; 2168 } 2169 } 2170 vm_map_unlock_read(map); 2171 vmspace_free(vm); 2172 PRELE(p); 2173 free(kve, M_TEMP); 2174 return (error); 2175 } 2176 #endif /* COMPAT_FREEBSD7 */ 2177 2178 #ifdef KINFO_VMENTRY_SIZE 2179 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); 2180 #endif 2181 2182 static void 2183 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, 2184 struct kinfo_vmentry *kve) 2185 { 2186 vm_object_t obj, tobj; 2187 vm_page_t m, m_adv; 2188 vm_offset_t addr; 2189 vm_paddr_t locked_pa; 2190 vm_pindex_t pi, pi_adv, pindex; 2191 2192 locked_pa = 0; 2193 obj = entry->object.vm_object; 2194 addr = entry->start; 2195 m_adv = NULL; 2196 pi = OFF_TO_IDX(entry->offset); 2197 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { 2198 if (m_adv != NULL) { 2199 m = m_adv; 2200 } else { 2201 pi_adv = OFF_TO_IDX(entry->end - addr); 2202 pindex = pi; 2203 for (tobj = obj;; tobj = tobj->backing_object) { 2204 m = vm_page_find_least(tobj, pindex); 2205 if (m != NULL) { 2206 if (m->pindex == pindex) 2207 break; 2208 if (pi_adv > m->pindex - pindex) { 2209 pi_adv = m->pindex - pindex; 2210 m_adv = m; 2211 } 2212 } 2213 if (tobj->backing_object == NULL) 2214 goto next; 2215 pindex += OFF_TO_IDX(tobj-> 2216 backing_object_offset); 2217 } 2218 } 2219 m_adv = NULL; 2220 if (m->psind != 0 && addr + pagesizes[1] <= entry->end && 2221 (addr & (pagesizes[1] - 1)) == 0 && 2222 (pmap_mincore(map->pmap, addr, &locked_pa) & 2223 MINCORE_SUPER) != 0) { 2224 kve->kve_flags |= KVME_FLAG_SUPER; 2225 pi_adv = OFF_TO_IDX(pagesizes[1]); 2226 } else { 2227 /* 2228 * We do not test the found page on validity. 2229 * Either the page is busy and being paged in, 2230 * or it was invalidated. The first case 2231 * should be counted as resident, the second 2232 * is not so clear; we do account both. 2233 */ 2234 pi_adv = 1; 2235 } 2236 kve->kve_resident += pi_adv; 2237 next:; 2238 } 2239 PA_UNLOCK_COND(locked_pa); 2240 } 2241 2242 /* 2243 * Must be called with the process locked and will return unlocked. 2244 */ 2245 int 2246 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb) 2247 { 2248 vm_map_entry_t entry, tmp_entry; 2249 struct vattr va; 2250 vm_map_t map; 2251 vm_object_t obj, tobj, lobj; 2252 char *fullpath, *freepath; 2253 struct kinfo_vmentry *kve; 2254 struct ucred *cred; 2255 struct vnode *vp; 2256 struct vmspace *vm; 2257 vm_offset_t addr; 2258 unsigned int last_timestamp; 2259 int error; 2260 2261 PROC_LOCK_ASSERT(p, MA_OWNED); 2262 2263 _PHOLD(p); 2264 PROC_UNLOCK(p); 2265 vm = vmspace_acquire_ref(p); 2266 if (vm == NULL) { 2267 PRELE(p); 2268 return (ESRCH); 2269 } 2270 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2271 2272 error = 0; 2273 map = &vm->vm_map; 2274 vm_map_lock_read(map); 2275 for (entry = map->header.next; entry != &map->header; 2276 entry = entry->next) { 2277 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2278 continue; 2279 2280 addr = entry->end; 2281 bzero(kve, sizeof(*kve)); 2282 obj = entry->object.vm_object; 2283 if (obj != NULL) { 2284 for (tobj = obj; tobj != NULL; 2285 tobj = tobj->backing_object) { 2286 VM_OBJECT_RLOCK(tobj); 2287 lobj = tobj; 2288 } 2289 if (obj->backing_object == NULL) 2290 kve->kve_private_resident = 2291 obj->resident_page_count; 2292 if (!vmmap_skip_res_cnt) 2293 kern_proc_vmmap_resident(map, entry, kve); 2294 for (tobj = obj; tobj != NULL; 2295 tobj = tobj->backing_object) { 2296 if (tobj != obj && tobj != lobj) 2297 VM_OBJECT_RUNLOCK(tobj); 2298 } 2299 } else { 2300 lobj = NULL; 2301 } 2302 2303 kve->kve_start = entry->start; 2304 kve->kve_end = entry->end; 2305 kve->kve_offset = entry->offset; 2306 2307 if (entry->protection & VM_PROT_READ) 2308 kve->kve_protection |= KVME_PROT_READ; 2309 if (entry->protection & VM_PROT_WRITE) 2310 kve->kve_protection |= KVME_PROT_WRITE; 2311 if (entry->protection & VM_PROT_EXECUTE) 2312 kve->kve_protection |= KVME_PROT_EXEC; 2313 2314 if (entry->eflags & MAP_ENTRY_COW) 2315 kve->kve_flags |= KVME_FLAG_COW; 2316 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2317 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2318 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2319 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2320 if (entry->eflags & MAP_ENTRY_GROWS_UP) 2321 kve->kve_flags |= KVME_FLAG_GROWS_UP; 2322 if (entry->eflags & MAP_ENTRY_GROWS_DOWN) 2323 kve->kve_flags |= KVME_FLAG_GROWS_DOWN; 2324 2325 last_timestamp = map->timestamp; 2326 vm_map_unlock_read(map); 2327 2328 freepath = NULL; 2329 fullpath = ""; 2330 if (lobj != NULL) { 2331 vp = NULL; 2332 switch (lobj->type) { 2333 case OBJT_DEFAULT: 2334 kve->kve_type = KVME_TYPE_DEFAULT; 2335 break; 2336 case OBJT_VNODE: 2337 kve->kve_type = KVME_TYPE_VNODE; 2338 vp = lobj->handle; 2339 vref(vp); 2340 break; 2341 case OBJT_SWAP: 2342 kve->kve_type = KVME_TYPE_SWAP; 2343 break; 2344 case OBJT_DEVICE: 2345 kve->kve_type = KVME_TYPE_DEVICE; 2346 break; 2347 case OBJT_PHYS: 2348 kve->kve_type = KVME_TYPE_PHYS; 2349 break; 2350 case OBJT_DEAD: 2351 kve->kve_type = KVME_TYPE_DEAD; 2352 break; 2353 case OBJT_SG: 2354 kve->kve_type = KVME_TYPE_SG; 2355 break; 2356 case OBJT_MGTDEVICE: 2357 kve->kve_type = KVME_TYPE_MGTDEVICE; 2358 break; 2359 default: 2360 kve->kve_type = KVME_TYPE_UNKNOWN; 2361 break; 2362 } 2363 if (lobj != obj) 2364 VM_OBJECT_RUNLOCK(lobj); 2365 2366 kve->kve_ref_count = obj->ref_count; 2367 kve->kve_shadow_count = obj->shadow_count; 2368 VM_OBJECT_RUNLOCK(obj); 2369 if (vp != NULL) { 2370 vn_fullpath(curthread, vp, &fullpath, 2371 &freepath); 2372 kve->kve_vn_type = vntype_to_kinfo(vp->v_type); 2373 cred = curthread->td_ucred; 2374 vn_lock(vp, LK_SHARED | LK_RETRY); 2375 if (VOP_GETATTR(vp, &va, cred) == 0) { 2376 kve->kve_vn_fileid = va.va_fileid; 2377 kve->kve_vn_fsid = va.va_fsid; 2378 kve->kve_vn_mode = 2379 MAKEIMODE(va.va_type, va.va_mode); 2380 kve->kve_vn_size = va.va_size; 2381 kve->kve_vn_rdev = va.va_rdev; 2382 kve->kve_status = KF_ATTR_VALID; 2383 } 2384 vput(vp); 2385 } 2386 } else { 2387 kve->kve_type = KVME_TYPE_NONE; 2388 kve->kve_ref_count = 0; 2389 kve->kve_shadow_count = 0; 2390 } 2391 2392 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2393 if (freepath != NULL) 2394 free(freepath, M_TEMP); 2395 2396 /* Pack record size down */ 2397 kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) + 2398 strlen(kve->kve_path) + 1; 2399 kve->kve_structsize = roundup(kve->kve_structsize, 2400 sizeof(uint64_t)); 2401 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0) 2402 error = ENOMEM; 2403 vm_map_lock_read(map); 2404 if (error != 0) 2405 break; 2406 if (last_timestamp != map->timestamp) { 2407 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2408 entry = tmp_entry; 2409 } 2410 } 2411 vm_map_unlock_read(map); 2412 vmspace_free(vm); 2413 PRELE(p); 2414 free(kve, M_TEMP); 2415 return (error); 2416 } 2417 2418 static int 2419 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) 2420 { 2421 struct proc *p; 2422 struct sbuf sb; 2423 int error, error2, *name; 2424 2425 name = (int *)arg1; 2426 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); 2427 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2428 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); 2429 if (error != 0) { 2430 sbuf_delete(&sb); 2431 return (error); 2432 } 2433 error = kern_proc_vmmap_out(p, &sb); 2434 error2 = sbuf_finish(&sb); 2435 sbuf_delete(&sb); 2436 return (error != 0 ? error : error2); 2437 } 2438 2439 #if defined(STACK) || defined(DDB) 2440 static int 2441 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) 2442 { 2443 struct kinfo_kstack *kkstp; 2444 int error, i, *name, numthreads; 2445 lwpid_t *lwpidarray; 2446 struct thread *td; 2447 struct stack *st; 2448 struct sbuf sb; 2449 struct proc *p; 2450 2451 name = (int *)arg1; 2452 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); 2453 if (error != 0) 2454 return (error); 2455 2456 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); 2457 st = stack_create(); 2458 2459 lwpidarray = NULL; 2460 numthreads = 0; 2461 PROC_LOCK(p); 2462 repeat: 2463 if (numthreads < p->p_numthreads) { 2464 if (lwpidarray != NULL) { 2465 free(lwpidarray, M_TEMP); 2466 lwpidarray = NULL; 2467 } 2468 numthreads = p->p_numthreads; 2469 PROC_UNLOCK(p); 2470 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, 2471 M_WAITOK | M_ZERO); 2472 PROC_LOCK(p); 2473 goto repeat; 2474 } 2475 i = 0; 2476 2477 /* 2478 * XXXRW: During the below loop, execve(2) and countless other sorts 2479 * of changes could have taken place. Should we check to see if the 2480 * vmspace has been replaced, or the like, in order to prevent 2481 * giving a snapshot that spans, say, execve(2), with some threads 2482 * before and some after? Among other things, the credentials could 2483 * have changed, in which case the right to extract debug info might 2484 * no longer be assured. 2485 */ 2486 FOREACH_THREAD_IN_PROC(p, td) { 2487 KASSERT(i < numthreads, 2488 ("sysctl_kern_proc_kstack: numthreads")); 2489 lwpidarray[i] = td->td_tid; 2490 i++; 2491 } 2492 numthreads = i; 2493 for (i = 0; i < numthreads; i++) { 2494 td = thread_find(p, lwpidarray[i]); 2495 if (td == NULL) { 2496 continue; 2497 } 2498 bzero(kkstp, sizeof(*kkstp)); 2499 (void)sbuf_new(&sb, kkstp->kkst_trace, 2500 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); 2501 thread_lock(td); 2502 kkstp->kkst_tid = td->td_tid; 2503 if (TD_IS_SWAPPED(td)) 2504 kkstp->kkst_state = KKST_STATE_SWAPPED; 2505 else if (TD_IS_RUNNING(td)) 2506 kkstp->kkst_state = KKST_STATE_RUNNING; 2507 else { 2508 kkstp->kkst_state = KKST_STATE_STACKOK; 2509 stack_save_td(st, td); 2510 } 2511 thread_unlock(td); 2512 PROC_UNLOCK(p); 2513 stack_sbuf_print(&sb, st); 2514 sbuf_finish(&sb); 2515 sbuf_delete(&sb); 2516 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); 2517 PROC_LOCK(p); 2518 if (error) 2519 break; 2520 } 2521 _PRELE(p); 2522 PROC_UNLOCK(p); 2523 if (lwpidarray != NULL) 2524 free(lwpidarray, M_TEMP); 2525 stack_destroy(st); 2526 free(kkstp, M_TEMP); 2527 return (error); 2528 } 2529 #endif 2530 2531 /* 2532 * This sysctl allows a process to retrieve the full list of groups from 2533 * itself or another process. 2534 */ 2535 static int 2536 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) 2537 { 2538 pid_t *pidp = (pid_t *)arg1; 2539 unsigned int arglen = arg2; 2540 struct proc *p; 2541 struct ucred *cred; 2542 int error; 2543 2544 if (arglen != 1) 2545 return (EINVAL); 2546 if (*pidp == -1) { /* -1 means this process */ 2547 p = req->td->td_proc; 2548 PROC_LOCK(p); 2549 } else { 2550 error = pget(*pidp, PGET_CANSEE, &p); 2551 if (error != 0) 2552 return (error); 2553 } 2554 2555 cred = crhold(p->p_ucred); 2556 PROC_UNLOCK(p); 2557 2558 error = SYSCTL_OUT(req, cred->cr_groups, 2559 cred->cr_ngroups * sizeof(gid_t)); 2560 crfree(cred); 2561 return (error); 2562 } 2563 2564 /* 2565 * This sysctl allows a process to retrieve or/and set the resource limit for 2566 * another process. 2567 */ 2568 static int 2569 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) 2570 { 2571 int *name = (int *)arg1; 2572 u_int namelen = arg2; 2573 struct rlimit rlim; 2574 struct proc *p; 2575 u_int which; 2576 int flags, error; 2577 2578 if (namelen != 2) 2579 return (EINVAL); 2580 2581 which = (u_int)name[1]; 2582 if (which >= RLIM_NLIMITS) 2583 return (EINVAL); 2584 2585 if (req->newptr != NULL && req->newlen != sizeof(rlim)) 2586 return (EINVAL); 2587 2588 flags = PGET_HOLD | PGET_NOTWEXIT; 2589 if (req->newptr != NULL) 2590 flags |= PGET_CANDEBUG; 2591 else 2592 flags |= PGET_CANSEE; 2593 error = pget((pid_t)name[0], flags, &p); 2594 if (error != 0) 2595 return (error); 2596 2597 /* 2598 * Retrieve limit. 2599 */ 2600 if (req->oldptr != NULL) { 2601 PROC_LOCK(p); 2602 lim_rlimit(p, which, &rlim); 2603 PROC_UNLOCK(p); 2604 } 2605 error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); 2606 if (error != 0) 2607 goto errout; 2608 2609 /* 2610 * Set limit. 2611 */ 2612 if (req->newptr != NULL) { 2613 error = SYSCTL_IN(req, &rlim, sizeof(rlim)); 2614 if (error == 0) 2615 error = kern_proc_setrlimit(curthread, p, which, &rlim); 2616 } 2617 2618 errout: 2619 PRELE(p); 2620 return (error); 2621 } 2622 2623 /* 2624 * This sysctl allows a process to retrieve ps_strings structure location of 2625 * another process. 2626 */ 2627 static int 2628 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) 2629 { 2630 int *name = (int *)arg1; 2631 u_int namelen = arg2; 2632 struct proc *p; 2633 vm_offset_t ps_strings; 2634 int error; 2635 #ifdef COMPAT_FREEBSD32 2636 uint32_t ps_strings32; 2637 #endif 2638 2639 if (namelen != 1) 2640 return (EINVAL); 2641 2642 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2643 if (error != 0) 2644 return (error); 2645 #ifdef COMPAT_FREEBSD32 2646 if ((req->flags & SCTL_MASK32) != 0) { 2647 /* 2648 * We return 0 if the 32 bit emulation request is for a 64 bit 2649 * process. 2650 */ 2651 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? 2652 PTROUT(p->p_sysent->sv_psstrings) : 0; 2653 PROC_UNLOCK(p); 2654 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); 2655 return (error); 2656 } 2657 #endif 2658 ps_strings = p->p_sysent->sv_psstrings; 2659 PROC_UNLOCK(p); 2660 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); 2661 return (error); 2662 } 2663 2664 /* 2665 * This sysctl allows a process to retrieve umask of another process. 2666 */ 2667 static int 2668 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) 2669 { 2670 int *name = (int *)arg1; 2671 u_int namelen = arg2; 2672 struct proc *p; 2673 int error; 2674 u_short fd_cmask; 2675 2676 if (namelen != 1) 2677 return (EINVAL); 2678 2679 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2680 if (error != 0) 2681 return (error); 2682 2683 FILEDESC_SLOCK(p->p_fd); 2684 fd_cmask = p->p_fd->fd_cmask; 2685 FILEDESC_SUNLOCK(p->p_fd); 2686 PRELE(p); 2687 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask)); 2688 return (error); 2689 } 2690 2691 /* 2692 * This sysctl allows a process to set and retrieve binary osreldate of 2693 * another process. 2694 */ 2695 static int 2696 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) 2697 { 2698 int *name = (int *)arg1; 2699 u_int namelen = arg2; 2700 struct proc *p; 2701 int flags, error, osrel; 2702 2703 if (namelen != 1) 2704 return (EINVAL); 2705 2706 if (req->newptr != NULL && req->newlen != sizeof(osrel)) 2707 return (EINVAL); 2708 2709 flags = PGET_HOLD | PGET_NOTWEXIT; 2710 if (req->newptr != NULL) 2711 flags |= PGET_CANDEBUG; 2712 else 2713 flags |= PGET_CANSEE; 2714 error = pget((pid_t)name[0], flags, &p); 2715 if (error != 0) 2716 return (error); 2717 2718 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); 2719 if (error != 0) 2720 goto errout; 2721 2722 if (req->newptr != NULL) { 2723 error = SYSCTL_IN(req, &osrel, sizeof(osrel)); 2724 if (error != 0) 2725 goto errout; 2726 if (osrel < 0) { 2727 error = EINVAL; 2728 goto errout; 2729 } 2730 p->p_osrel = osrel; 2731 } 2732 errout: 2733 PRELE(p); 2734 return (error); 2735 } 2736 2737 static int 2738 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) 2739 { 2740 int *name = (int *)arg1; 2741 u_int namelen = arg2; 2742 struct proc *p; 2743 struct kinfo_sigtramp kst; 2744 const struct sysentvec *sv; 2745 int error; 2746 #ifdef COMPAT_FREEBSD32 2747 struct kinfo_sigtramp32 kst32; 2748 #endif 2749 2750 if (namelen != 1) 2751 return (EINVAL); 2752 2753 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2754 if (error != 0) 2755 return (error); 2756 sv = p->p_sysent; 2757 #ifdef COMPAT_FREEBSD32 2758 if ((req->flags & SCTL_MASK32) != 0) { 2759 bzero(&kst32, sizeof(kst32)); 2760 if (SV_PROC_FLAG(p, SV_ILP32)) { 2761 if (sv->sv_sigcode_base != 0) { 2762 kst32.ksigtramp_start = sv->sv_sigcode_base; 2763 kst32.ksigtramp_end = sv->sv_sigcode_base + 2764 *sv->sv_szsigcode; 2765 } else { 2766 kst32.ksigtramp_start = sv->sv_psstrings - 2767 *sv->sv_szsigcode; 2768 kst32.ksigtramp_end = sv->sv_psstrings; 2769 } 2770 } 2771 PROC_UNLOCK(p); 2772 error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); 2773 return (error); 2774 } 2775 #endif 2776 bzero(&kst, sizeof(kst)); 2777 if (sv->sv_sigcode_base != 0) { 2778 kst.ksigtramp_start = (char *)sv->sv_sigcode_base; 2779 kst.ksigtramp_end = (char *)sv->sv_sigcode_base + 2780 *sv->sv_szsigcode; 2781 } else { 2782 kst.ksigtramp_start = (char *)sv->sv_psstrings - 2783 *sv->sv_szsigcode; 2784 kst.ksigtramp_end = (char *)sv->sv_psstrings; 2785 } 2786 PROC_UNLOCK(p); 2787 error = SYSCTL_OUT(req, &kst, sizeof(kst)); 2788 return (error); 2789 } 2790 2791 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table"); 2792 2793 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| 2794 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", 2795 "Return entire process table"); 2796 2797 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2798 sysctl_kern_proc, "Process table"); 2799 2800 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, 2801 sysctl_kern_proc, "Process table"); 2802 2803 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2804 sysctl_kern_proc, "Process table"); 2805 2806 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | 2807 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2808 2809 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 2810 sysctl_kern_proc, "Process table"); 2811 2812 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2813 sysctl_kern_proc, "Process table"); 2814 2815 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2816 sysctl_kern_proc, "Process table"); 2817 2818 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2819 sysctl_kern_proc, "Process table"); 2820 2821 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 2822 sysctl_kern_proc, "Return process table, no threads"); 2823 2824 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, 2825 CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, 2826 sysctl_kern_proc_args, "Process argument list"); 2827 2828 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, 2829 sysctl_kern_proc_env, "Process environment"); 2830 2831 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | 2832 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); 2833 2834 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | 2835 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); 2836 2837 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | 2838 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, 2839 "Process syscall vector name (ABI type)"); 2840 2841 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, 2842 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2843 2844 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, 2845 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2846 2847 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, 2848 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2849 2850 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), 2851 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2852 2853 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, 2854 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2855 2856 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, 2857 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2858 2859 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, 2860 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2861 2862 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, 2863 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2864 2865 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, 2866 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, 2867 "Return process table, no threads"); 2868 2869 #ifdef COMPAT_FREEBSD7 2870 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | 2871 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); 2872 #endif 2873 2874 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | 2875 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); 2876 2877 #if defined(STACK) || defined(DDB) 2878 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | 2879 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); 2880 #endif 2881 2882 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | 2883 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); 2884 2885 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | 2886 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, 2887 "Process resource limits"); 2888 2889 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | 2890 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, 2891 "Process ps_strings location"); 2892 2893 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | 2894 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); 2895 2896 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | 2897 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, 2898 "Process binary osreldate"); 2899 2900 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | 2901 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, 2902 "Process signal trampoline location"); 2903 2904 int allproc_gen; 2905 2906 void 2907 stop_all_proc(void) 2908 { 2909 struct proc *cp, *p; 2910 int r, gen; 2911 bool restart, seen_stopped, seen_exiting, stopped_some; 2912 2913 cp = curproc; 2914 /* 2915 * stop_all_proc() assumes that all process which have 2916 * usermode must be stopped, except current process, for 2917 * obvious reasons. Since other threads in the process 2918 * establishing global stop could unstop something, disable 2919 * calls from multithreaded processes as precaution. The 2920 * service must not be user-callable anyway. 2921 */ 2922 KASSERT((cp->p_flag & P_HADTHREADS) == 0 || 2923 (cp->p_flag & P_KTHREAD) != 0, ("mt stop_all_proc")); 2924 2925 allproc_loop: 2926 sx_xlock(&allproc_lock); 2927 gen = allproc_gen; 2928 seen_exiting = seen_stopped = stopped_some = restart = false; 2929 LIST_REMOVE(cp, p_list); 2930 LIST_INSERT_HEAD(&allproc, cp, p_list); 2931 for (;;) { 2932 p = LIST_NEXT(cp, p_list); 2933 if (p == NULL) 2934 break; 2935 LIST_REMOVE(cp, p_list); 2936 LIST_INSERT_AFTER(p, cp, p_list); 2937 PROC_LOCK(p); 2938 if ((p->p_flag & (P_KTHREAD | P_SYSTEM | 2939 P_TOTAL_STOP)) != 0) { 2940 PROC_UNLOCK(p); 2941 continue; 2942 } 2943 if ((p->p_flag & P_WEXIT) != 0) { 2944 seen_exiting = true; 2945 PROC_UNLOCK(p); 2946 continue; 2947 } 2948 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 2949 /* 2950 * Stopped processes are tolerated when there 2951 * are no other processes which might continue 2952 * them. P_STOPPED_SINGLE but not 2953 * P_TOTAL_STOP process still has at least one 2954 * thread running. 2955 */ 2956 seen_stopped = true; 2957 PROC_UNLOCK(p); 2958 continue; 2959 } 2960 _PHOLD(p); 2961 sx_xunlock(&allproc_lock); 2962 r = thread_single(p, SINGLE_ALLPROC); 2963 if (r != 0) 2964 restart = true; 2965 else 2966 stopped_some = true; 2967 _PRELE(p); 2968 PROC_UNLOCK(p); 2969 sx_xlock(&allproc_lock); 2970 } 2971 /* Catch forked children we did not see in iteration. */ 2972 if (gen != allproc_gen) 2973 restart = true; 2974 sx_xunlock(&allproc_lock); 2975 if (restart || stopped_some || seen_exiting || seen_stopped) { 2976 kern_yield(PRI_USER); 2977 goto allproc_loop; 2978 } 2979 } 2980 2981 void 2982 resume_all_proc(void) 2983 { 2984 struct proc *cp, *p; 2985 2986 cp = curproc; 2987 sx_xlock(&allproc_lock); 2988 LIST_REMOVE(cp, p_list); 2989 LIST_INSERT_HEAD(&allproc, cp, p_list); 2990 for (;;) { 2991 p = LIST_NEXT(cp, p_list); 2992 if (p == NULL) 2993 break; 2994 LIST_REMOVE(cp, p_list); 2995 LIST_INSERT_AFTER(p, cp, p_list); 2996 PROC_LOCK(p); 2997 if ((p->p_flag & P_TOTAL_STOP) != 0) { 2998 sx_xunlock(&allproc_lock); 2999 _PHOLD(p); 3000 thread_single_end(p, SINGLE_ALLPROC); 3001 _PRELE(p); 3002 PROC_UNLOCK(p); 3003 sx_xlock(&allproc_lock); 3004 } else { 3005 PROC_UNLOCK(p); 3006 } 3007 } 3008 sx_xunlock(&allproc_lock); 3009 } 3010 3011 #define TOTAL_STOP_DEBUG 1 3012 #ifdef TOTAL_STOP_DEBUG 3013 volatile static int ap_resume; 3014 #include <sys/mount.h> 3015 3016 static int 3017 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) 3018 { 3019 int error, val; 3020 3021 val = 0; 3022 ap_resume = 0; 3023 error = sysctl_handle_int(oidp, &val, 0, req); 3024 if (error != 0 || req->newptr == NULL) 3025 return (error); 3026 if (val != 0) { 3027 stop_all_proc(); 3028 syncer_suspend(); 3029 while (ap_resume == 0) 3030 ; 3031 syncer_resume(); 3032 resume_all_proc(); 3033 } 3034 return (0); 3035 } 3036 3037 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | 3038 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, 3039 sysctl_debug_stop_all_proc, "I", 3040 ""); 3041 #endif 3042