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/eventhandler.h> 45 #include <sys/exec.h> 46 #include <sys/jail.h> 47 #include <sys/kernel.h> 48 #include <sys/limits.h> 49 #include <sys/lock.h> 50 #include <sys/loginclass.h> 51 #include <sys/malloc.h> 52 #include <sys/mman.h> 53 #include <sys/mount.h> 54 #include <sys/mutex.h> 55 #include <sys/proc.h> 56 #include <sys/ptrace.h> 57 #include <sys/refcount.h> 58 #include <sys/resourcevar.h> 59 #include <sys/rwlock.h> 60 #include <sys/sbuf.h> 61 #include <sys/sysent.h> 62 #include <sys/sched.h> 63 #include <sys/smp.h> 64 #include <sys/stack.h> 65 #include <sys/stat.h> 66 #include <sys/sysctl.h> 67 #include <sys/filedesc.h> 68 #include <sys/tty.h> 69 #include <sys/signalvar.h> 70 #include <sys/sdt.h> 71 #include <sys/sx.h> 72 #include <sys/user.h> 73 #include <sys/vnode.h> 74 #include <sys/wait.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 94 SDT_PROVIDER_DEFINE(proc); 95 SDT_PROBE_DEFINE4(proc, kernel, ctor, entry, "struct proc *", "int", 96 "void *", "int"); 97 SDT_PROBE_DEFINE4(proc, kernel, ctor, return, "struct proc *", "int", 98 "void *", "int"); 99 SDT_PROBE_DEFINE4(proc, kernel, dtor, entry, "struct proc *", "int", 100 "void *", "struct thread *"); 101 SDT_PROBE_DEFINE3(proc, kernel, dtor, return, "struct proc *", "int", 102 "void *"); 103 SDT_PROBE_DEFINE3(proc, kernel, init, entry, "struct proc *", "int", 104 "int"); 105 SDT_PROBE_DEFINE3(proc, kernel, init, return, "struct proc *", "int", 106 "int"); 107 108 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header"); 109 MALLOC_DEFINE(M_SESSION, "session", "session header"); 110 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); 111 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 112 113 static void doenterpgrp(struct proc *, struct pgrp *); 114 static void orphanpg(struct pgrp *pg); 115 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp); 116 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp); 117 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, 118 int preferthread); 119 static void pgadjustjobc(struct pgrp *pgrp, int entering); 120 static void pgdelete(struct pgrp *); 121 static int proc_ctor(void *mem, int size, void *arg, int flags); 122 static void proc_dtor(void *mem, int size, void *arg); 123 static int proc_init(void *mem, int size, int flags); 124 static void proc_fini(void *mem, int size); 125 static void pargs_free(struct pargs *pa); 126 static struct proc *zpfind_locked(pid_t pid); 127 128 /* 129 * Other process lists 130 */ 131 struct pidhashhead *pidhashtbl; 132 u_long pidhash; 133 struct pgrphashhead *pgrphashtbl; 134 u_long pgrphash; 135 struct proclist allproc; 136 struct proclist zombproc; 137 struct sx allproc_lock; 138 struct sx proctree_lock; 139 struct mtx ppeers_lock; 140 uma_zone_t proc_zone; 141 142 /* 143 * The offset of various fields in struct proc and struct thread. 144 * These are used by kernel debuggers to enumerate kernel threads and 145 * processes. 146 */ 147 const int proc_off_p_pid = offsetof(struct proc, p_pid); 148 const int proc_off_p_comm = offsetof(struct proc, p_comm); 149 const int proc_off_p_list = offsetof(struct proc, p_list); 150 const int proc_off_p_threads = offsetof(struct proc, p_threads); 151 const int thread_off_td_tid = offsetof(struct thread, td_tid); 152 const int thread_off_td_name = offsetof(struct thread, td_name); 153 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu); 154 const int thread_off_td_pcb = offsetof(struct thread, td_pcb); 155 const int thread_off_td_plist = offsetof(struct thread, td_plist); 156 157 int kstack_pages = KSTACK_PAGES; 158 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, 159 "Kernel stack size in pages"); 160 static int vmmap_skip_res_cnt = 0; 161 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW, 162 &vmmap_skip_res_cnt, 0, 163 "Skip calculation of the pages resident count in kern.proc.vmmap"); 164 165 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); 166 #ifdef COMPAT_FREEBSD32 167 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE); 168 #endif 169 170 /* 171 * Initialize global process hashing structures. 172 */ 173 void 174 procinit() 175 { 176 177 sx_init(&allproc_lock, "allproc"); 178 sx_init(&proctree_lock, "proctree"); 179 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF); 180 LIST_INIT(&allproc); 181 LIST_INIT(&zombproc); 182 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); 183 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); 184 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(), 185 proc_ctor, proc_dtor, proc_init, proc_fini, 186 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 187 uihashinit(); 188 } 189 190 /* 191 * Prepare a proc for use. 192 */ 193 static int 194 proc_ctor(void *mem, int size, void *arg, int flags) 195 { 196 struct proc *p; 197 198 p = (struct proc *)mem; 199 SDT_PROBE4(proc, kernel, ctor , entry, p, size, arg, flags); 200 EVENTHANDLER_INVOKE(process_ctor, p); 201 SDT_PROBE4(proc, kernel, ctor , return, p, size, arg, flags); 202 return (0); 203 } 204 205 /* 206 * Reclaim a proc after use. 207 */ 208 static void 209 proc_dtor(void *mem, int size, void *arg) 210 { 211 struct proc *p; 212 struct thread *td; 213 214 /* INVARIANTS checks go here */ 215 p = (struct proc *)mem; 216 td = FIRST_THREAD_IN_PROC(p); 217 SDT_PROBE4(proc, kernel, dtor, entry, p, size, arg, td); 218 if (td != NULL) { 219 #ifdef INVARIANTS 220 KASSERT((p->p_numthreads == 1), 221 ("bad number of threads in exiting process")); 222 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr")); 223 #endif 224 /* Free all OSD associated to this thread. */ 225 osd_thread_exit(td); 226 } 227 EVENTHANDLER_INVOKE(process_dtor, p); 228 if (p->p_ksi != NULL) 229 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue")); 230 SDT_PROBE3(proc, kernel, dtor, return, p, size, arg); 231 } 232 233 /* 234 * Initialize type-stable parts of a proc (when newly created). 235 */ 236 static int 237 proc_init(void *mem, int size, int flags) 238 { 239 struct proc *p; 240 241 p = (struct proc *)mem; 242 SDT_PROBE3(proc, kernel, init, entry, p, size, flags); 243 p->p_sched = (struct p_sched *)&p[1]; 244 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW); 245 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW); 246 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW); 247 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW); 248 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW); 249 cv_init(&p->p_pwait, "ppwait"); 250 cv_init(&p->p_dbgwait, "dbgwait"); 251 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 252 EVENTHANDLER_INVOKE(process_init, p); 253 p->p_stats = pstats_alloc(); 254 SDT_PROBE3(proc, kernel, init, return, p, size, flags); 255 return (0); 256 } 257 258 /* 259 * UMA should ensure that this function is never called. 260 * Freeing a proc structure would violate type stability. 261 */ 262 static void 263 proc_fini(void *mem, int size) 264 { 265 #ifdef notnow 266 struct proc *p; 267 268 p = (struct proc *)mem; 269 EVENTHANDLER_INVOKE(process_fini, p); 270 pstats_free(p->p_stats); 271 thread_free(FIRST_THREAD_IN_PROC(p)); 272 mtx_destroy(&p->p_mtx); 273 if (p->p_ksi != NULL) 274 ksiginfo_free(p->p_ksi); 275 #else 276 panic("proc reclaimed"); 277 #endif 278 } 279 280 /* 281 * Is p an inferior of the current process? 282 */ 283 int 284 inferior(struct proc *p) 285 { 286 287 sx_assert(&proctree_lock, SX_LOCKED); 288 PROC_LOCK_ASSERT(p, MA_OWNED); 289 for (; p != curproc; p = proc_realparent(p)) { 290 if (p->p_pid == 0) 291 return (0); 292 } 293 return (1); 294 } 295 296 struct proc * 297 pfind_locked(pid_t pid) 298 { 299 struct proc *p; 300 301 sx_assert(&allproc_lock, SX_LOCKED); 302 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 303 if (p->p_pid == pid) { 304 PROC_LOCK(p); 305 if (p->p_state == PRS_NEW) { 306 PROC_UNLOCK(p); 307 p = NULL; 308 } 309 break; 310 } 311 } 312 return (p); 313 } 314 315 /* 316 * Locate a process by number; return only "live" processes -- i.e., neither 317 * zombies nor newly born but incompletely initialized processes. By not 318 * returning processes in the PRS_NEW state, we allow callers to avoid 319 * testing for that condition to avoid dereferencing p_ucred, et al. 320 */ 321 struct proc * 322 pfind(pid_t pid) 323 { 324 struct proc *p; 325 326 sx_slock(&allproc_lock); 327 p = pfind_locked(pid); 328 sx_sunlock(&allproc_lock); 329 return (p); 330 } 331 332 static struct proc * 333 pfind_tid_locked(pid_t tid) 334 { 335 struct proc *p; 336 struct thread *td; 337 338 sx_assert(&allproc_lock, SX_LOCKED); 339 FOREACH_PROC_IN_SYSTEM(p) { 340 PROC_LOCK(p); 341 if (p->p_state == PRS_NEW) { 342 PROC_UNLOCK(p); 343 continue; 344 } 345 FOREACH_THREAD_IN_PROC(p, td) { 346 if (td->td_tid == tid) 347 goto found; 348 } 349 PROC_UNLOCK(p); 350 } 351 found: 352 return (p); 353 } 354 355 /* 356 * Locate a process group by number. 357 * The caller must hold proctree_lock. 358 */ 359 struct pgrp * 360 pgfind(pgid) 361 register pid_t pgid; 362 { 363 register struct pgrp *pgrp; 364 365 sx_assert(&proctree_lock, SX_LOCKED); 366 367 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { 368 if (pgrp->pg_id == pgid) { 369 PGRP_LOCK(pgrp); 370 return (pgrp); 371 } 372 } 373 return (NULL); 374 } 375 376 /* 377 * Locate process and do additional manipulations, depending on flags. 378 */ 379 int 380 pget(pid_t pid, int flags, struct proc **pp) 381 { 382 struct proc *p; 383 int error; 384 385 sx_slock(&allproc_lock); 386 if (pid <= PID_MAX) { 387 p = pfind_locked(pid); 388 if (p == NULL && (flags & PGET_NOTWEXIT) == 0) 389 p = zpfind_locked(pid); 390 } else if ((flags & PGET_NOTID) == 0) { 391 p = pfind_tid_locked(pid); 392 } else { 393 p = NULL; 394 } 395 sx_sunlock(&allproc_lock); 396 if (p == NULL) 397 return (ESRCH); 398 if ((flags & PGET_CANSEE) != 0) { 399 error = p_cansee(curthread, p); 400 if (error != 0) 401 goto errout; 402 } 403 if ((flags & PGET_CANDEBUG) != 0) { 404 error = p_candebug(curthread, p); 405 if (error != 0) 406 goto errout; 407 } 408 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) { 409 error = EPERM; 410 goto errout; 411 } 412 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) { 413 error = ESRCH; 414 goto errout; 415 } 416 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) { 417 /* 418 * XXXRW: Not clear ESRCH is the right error during proc 419 * execve(). 420 */ 421 error = ESRCH; 422 goto errout; 423 } 424 if ((flags & PGET_HOLD) != 0) { 425 _PHOLD(p); 426 PROC_UNLOCK(p); 427 } 428 *pp = p; 429 return (0); 430 errout: 431 PROC_UNLOCK(p); 432 return (error); 433 } 434 435 /* 436 * Create a new process group. 437 * pgid must be equal to the pid of p. 438 * Begin a new session if required. 439 */ 440 int 441 enterpgrp(p, pgid, pgrp, sess) 442 register struct proc *p; 443 pid_t pgid; 444 struct pgrp *pgrp; 445 struct session *sess; 446 { 447 448 sx_assert(&proctree_lock, SX_XLOCKED); 449 450 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL")); 451 KASSERT(p->p_pid == pgid, 452 ("enterpgrp: new pgrp and pid != pgid")); 453 KASSERT(pgfind(pgid) == NULL, 454 ("enterpgrp: pgrp with pgid exists")); 455 KASSERT(!SESS_LEADER(p), 456 ("enterpgrp: session leader attempted setpgrp")); 457 458 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 459 460 if (sess != NULL) { 461 /* 462 * new session 463 */ 464 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); 465 PROC_LOCK(p); 466 p->p_flag &= ~P_CONTROLT; 467 PROC_UNLOCK(p); 468 PGRP_LOCK(pgrp); 469 sess->s_leader = p; 470 sess->s_sid = p->p_pid; 471 refcount_init(&sess->s_count, 1); 472 sess->s_ttyvp = NULL; 473 sess->s_ttydp = NULL; 474 sess->s_ttyp = NULL; 475 bcopy(p->p_session->s_login, sess->s_login, 476 sizeof(sess->s_login)); 477 pgrp->pg_session = sess; 478 KASSERT(p == curproc, 479 ("enterpgrp: mksession and p != curproc")); 480 } else { 481 pgrp->pg_session = p->p_session; 482 sess_hold(pgrp->pg_session); 483 PGRP_LOCK(pgrp); 484 } 485 pgrp->pg_id = pgid; 486 LIST_INIT(&pgrp->pg_members); 487 488 /* 489 * As we have an exclusive lock of proctree_lock, 490 * this should not deadlock. 491 */ 492 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); 493 pgrp->pg_jobc = 0; 494 SLIST_INIT(&pgrp->pg_sigiolst); 495 PGRP_UNLOCK(pgrp); 496 497 doenterpgrp(p, pgrp); 498 499 return (0); 500 } 501 502 /* 503 * Move p to an existing process group 504 */ 505 int 506 enterthispgrp(p, pgrp) 507 register struct proc *p; 508 struct pgrp *pgrp; 509 { 510 511 sx_assert(&proctree_lock, SX_XLOCKED); 512 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 513 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 514 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 515 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 516 KASSERT(pgrp->pg_session == p->p_session, 517 ("%s: pgrp's session %p, p->p_session %p.\n", 518 __func__, 519 pgrp->pg_session, 520 p->p_session)); 521 KASSERT(pgrp != p->p_pgrp, 522 ("%s: p belongs to pgrp.", __func__)); 523 524 doenterpgrp(p, pgrp); 525 526 return (0); 527 } 528 529 /* 530 * Move p to a process group 531 */ 532 static void 533 doenterpgrp(p, pgrp) 534 struct proc *p; 535 struct pgrp *pgrp; 536 { 537 struct pgrp *savepgrp; 538 539 sx_assert(&proctree_lock, SX_XLOCKED); 540 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 541 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 542 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 543 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 544 545 savepgrp = p->p_pgrp; 546 547 /* 548 * Adjust eligibility of affected pgrps to participate in job control. 549 * Increment eligibility counts before decrementing, otherwise we 550 * could reach 0 spuriously during the first call. 551 */ 552 fixjobc(p, pgrp, 1); 553 fixjobc(p, p->p_pgrp, 0); 554 555 PGRP_LOCK(pgrp); 556 PGRP_LOCK(savepgrp); 557 PROC_LOCK(p); 558 LIST_REMOVE(p, p_pglist); 559 p->p_pgrp = pgrp; 560 PROC_UNLOCK(p); 561 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 562 PGRP_UNLOCK(savepgrp); 563 PGRP_UNLOCK(pgrp); 564 if (LIST_EMPTY(&savepgrp->pg_members)) 565 pgdelete(savepgrp); 566 } 567 568 /* 569 * remove process from process group 570 */ 571 int 572 leavepgrp(p) 573 register struct proc *p; 574 { 575 struct pgrp *savepgrp; 576 577 sx_assert(&proctree_lock, SX_XLOCKED); 578 savepgrp = p->p_pgrp; 579 PGRP_LOCK(savepgrp); 580 PROC_LOCK(p); 581 LIST_REMOVE(p, p_pglist); 582 p->p_pgrp = NULL; 583 PROC_UNLOCK(p); 584 PGRP_UNLOCK(savepgrp); 585 if (LIST_EMPTY(&savepgrp->pg_members)) 586 pgdelete(savepgrp); 587 return (0); 588 } 589 590 /* 591 * delete a process group 592 */ 593 static void 594 pgdelete(pgrp) 595 register struct pgrp *pgrp; 596 { 597 struct session *savesess; 598 struct tty *tp; 599 600 sx_assert(&proctree_lock, SX_XLOCKED); 601 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 602 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 603 604 /* 605 * Reset any sigio structures pointing to us as a result of 606 * F_SETOWN with our pgid. 607 */ 608 funsetownlst(&pgrp->pg_sigiolst); 609 610 PGRP_LOCK(pgrp); 611 tp = pgrp->pg_session->s_ttyp; 612 LIST_REMOVE(pgrp, pg_hash); 613 savesess = pgrp->pg_session; 614 PGRP_UNLOCK(pgrp); 615 616 /* Remove the reference to the pgrp before deallocating it. */ 617 if (tp != NULL) { 618 tty_lock(tp); 619 tty_rel_pgrp(tp, pgrp); 620 } 621 622 mtx_destroy(&pgrp->pg_mtx); 623 free(pgrp, M_PGRP); 624 sess_release(savesess); 625 } 626 627 static void 628 pgadjustjobc(pgrp, entering) 629 struct pgrp *pgrp; 630 int entering; 631 { 632 633 PGRP_LOCK(pgrp); 634 if (entering) 635 pgrp->pg_jobc++; 636 else { 637 --pgrp->pg_jobc; 638 if (pgrp->pg_jobc == 0) 639 orphanpg(pgrp); 640 } 641 PGRP_UNLOCK(pgrp); 642 } 643 644 /* 645 * Adjust pgrp jobc counters when specified process changes process group. 646 * We count the number of processes in each process group that "qualify" 647 * the group for terminal job control (those with a parent in a different 648 * process group of the same session). If that count reaches zero, the 649 * process group becomes orphaned. Check both the specified process' 650 * process group and that of its children. 651 * entering == 0 => p is leaving specified group. 652 * entering == 1 => p is entering specified group. 653 */ 654 void 655 fixjobc(p, pgrp, entering) 656 register struct proc *p; 657 register struct pgrp *pgrp; 658 int entering; 659 { 660 register struct pgrp *hispgrp; 661 register struct session *mysession; 662 663 sx_assert(&proctree_lock, SX_LOCKED); 664 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 665 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 666 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 667 668 /* 669 * Check p's parent to see whether p qualifies its own process 670 * group; if so, adjust count for p's process group. 671 */ 672 mysession = pgrp->pg_session; 673 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp && 674 hispgrp->pg_session == mysession) 675 pgadjustjobc(pgrp, entering); 676 677 /* 678 * Check this process' children to see whether they qualify 679 * their process groups; if so, adjust counts for children's 680 * process groups. 681 */ 682 LIST_FOREACH(p, &p->p_children, p_sibling) { 683 hispgrp = p->p_pgrp; 684 if (hispgrp == pgrp || 685 hispgrp->pg_session != mysession) 686 continue; 687 PROC_LOCK(p); 688 if (p->p_state == PRS_ZOMBIE) { 689 PROC_UNLOCK(p); 690 continue; 691 } 692 PROC_UNLOCK(p); 693 pgadjustjobc(hispgrp, entering); 694 } 695 } 696 697 /* 698 * A process group has become orphaned; 699 * if there are any stopped processes in the group, 700 * hang-up all process in that group. 701 */ 702 static void 703 orphanpg(pg) 704 struct pgrp *pg; 705 { 706 register struct proc *p; 707 708 PGRP_LOCK_ASSERT(pg, MA_OWNED); 709 710 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 711 PROC_LOCK(p); 712 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) { 713 PROC_UNLOCK(p); 714 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 715 PROC_LOCK(p); 716 kern_psignal(p, SIGHUP); 717 kern_psignal(p, SIGCONT); 718 PROC_UNLOCK(p); 719 } 720 return; 721 } 722 PROC_UNLOCK(p); 723 } 724 } 725 726 void 727 sess_hold(struct session *s) 728 { 729 730 refcount_acquire(&s->s_count); 731 } 732 733 void 734 sess_release(struct session *s) 735 { 736 737 if (refcount_release(&s->s_count)) { 738 if (s->s_ttyp != NULL) { 739 tty_lock(s->s_ttyp); 740 tty_rel_sess(s->s_ttyp, s); 741 } 742 mtx_destroy(&s->s_mtx); 743 free(s, M_SESSION); 744 } 745 } 746 747 #ifdef DDB 748 749 DB_SHOW_COMMAND(pgrpdump, pgrpdump) 750 { 751 register struct pgrp *pgrp; 752 register struct proc *p; 753 register int i; 754 755 for (i = 0; i <= pgrphash; i++) { 756 if (!LIST_EMPTY(&pgrphashtbl[i])) { 757 printf("\tindx %d\n", i); 758 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { 759 printf( 760 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n", 761 (void *)pgrp, (long)pgrp->pg_id, 762 (void *)pgrp->pg_session, 763 pgrp->pg_session->s_count, 764 (void *)LIST_FIRST(&pgrp->pg_members)); 765 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 766 printf("\t\tpid %ld addr %p pgrp %p\n", 767 (long)p->p_pid, (void *)p, 768 (void *)p->p_pgrp); 769 } 770 } 771 } 772 } 773 } 774 #endif /* DDB */ 775 776 /* 777 * Calculate the kinfo_proc members which contain process-wide 778 * informations. 779 * Must be called with the target process locked. 780 */ 781 static void 782 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp) 783 { 784 struct thread *td; 785 786 PROC_LOCK_ASSERT(p, MA_OWNED); 787 788 kp->ki_estcpu = 0; 789 kp->ki_pctcpu = 0; 790 FOREACH_THREAD_IN_PROC(p, td) { 791 thread_lock(td); 792 kp->ki_pctcpu += sched_pctcpu(td); 793 kp->ki_estcpu += td->td_estcpu; 794 thread_unlock(td); 795 } 796 } 797 798 /* 799 * Clear kinfo_proc and fill in any information that is common 800 * to all threads in the process. 801 * Must be called with the target process locked. 802 */ 803 static void 804 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp) 805 { 806 struct thread *td0; 807 struct tty *tp; 808 struct session *sp; 809 struct ucred *cred; 810 struct sigacts *ps; 811 812 /* For proc_realparent. */ 813 sx_assert(&proctree_lock, SX_LOCKED); 814 PROC_LOCK_ASSERT(p, MA_OWNED); 815 bzero(kp, sizeof(*kp)); 816 817 kp->ki_structsize = sizeof(*kp); 818 kp->ki_paddr = p; 819 kp->ki_addr =/* p->p_addr; */0; /* XXX */ 820 kp->ki_args = p->p_args; 821 kp->ki_textvp = p->p_textvp; 822 #ifdef KTRACE 823 kp->ki_tracep = p->p_tracevp; 824 kp->ki_traceflag = p->p_traceflag; 825 #endif 826 kp->ki_fd = p->p_fd; 827 kp->ki_vmspace = p->p_vmspace; 828 kp->ki_flag = p->p_flag; 829 kp->ki_flag2 = p->p_flag2; 830 cred = p->p_ucred; 831 if (cred) { 832 kp->ki_uid = cred->cr_uid; 833 kp->ki_ruid = cred->cr_ruid; 834 kp->ki_svuid = cred->cr_svuid; 835 kp->ki_cr_flags = 0; 836 if (cred->cr_flags & CRED_FLAG_CAPMODE) 837 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE; 838 /* XXX bde doesn't like KI_NGROUPS */ 839 if (cred->cr_ngroups > KI_NGROUPS) { 840 kp->ki_ngroups = KI_NGROUPS; 841 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW; 842 } else 843 kp->ki_ngroups = cred->cr_ngroups; 844 bcopy(cred->cr_groups, kp->ki_groups, 845 kp->ki_ngroups * sizeof(gid_t)); 846 kp->ki_rgid = cred->cr_rgid; 847 kp->ki_svgid = cred->cr_svgid; 848 /* If jailed(cred), emulate the old P_JAILED flag. */ 849 if (jailed(cred)) { 850 kp->ki_flag |= P_JAILED; 851 /* If inside the jail, use 0 as a jail ID. */ 852 if (cred->cr_prison != curthread->td_ucred->cr_prison) 853 kp->ki_jid = cred->cr_prison->pr_id; 854 } 855 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name, 856 sizeof(kp->ki_loginclass)); 857 } 858 ps = p->p_sigacts; 859 if (ps) { 860 mtx_lock(&ps->ps_mtx); 861 kp->ki_sigignore = ps->ps_sigignore; 862 kp->ki_sigcatch = ps->ps_sigcatch; 863 mtx_unlock(&ps->ps_mtx); 864 } 865 if (p->p_state != PRS_NEW && 866 p->p_state != PRS_ZOMBIE && 867 p->p_vmspace != NULL) { 868 struct vmspace *vm = p->p_vmspace; 869 870 kp->ki_size = vm->vm_map.size; 871 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ 872 FOREACH_THREAD_IN_PROC(p, td0) { 873 if (!TD_IS_SWAPPED(td0)) 874 kp->ki_rssize += td0->td_kstack_pages; 875 } 876 kp->ki_swrss = vm->vm_swrss; 877 kp->ki_tsize = vm->vm_tsize; 878 kp->ki_dsize = vm->vm_dsize; 879 kp->ki_ssize = vm->vm_ssize; 880 } else if (p->p_state == PRS_ZOMBIE) 881 kp->ki_stat = SZOMB; 882 if (kp->ki_flag & P_INMEM) 883 kp->ki_sflag = PS_INMEM; 884 else 885 kp->ki_sflag = 0; 886 /* Calculate legacy swtime as seconds since 'swtick'. */ 887 kp->ki_swtime = (ticks - p->p_swtick) / hz; 888 kp->ki_pid = p->p_pid; 889 kp->ki_nice = p->p_nice; 890 kp->ki_fibnum = p->p_fibnum; 891 kp->ki_start = p->p_stats->p_start; 892 timevaladd(&kp->ki_start, &boottime); 893 PROC_STATLOCK(p); 894 rufetch(p, &kp->ki_rusage); 895 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime); 896 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime); 897 PROC_STATUNLOCK(p); 898 calccru(p, &kp->ki_childutime, &kp->ki_childstime); 899 /* Some callers want child times in a single value. */ 900 kp->ki_childtime = kp->ki_childstime; 901 timevaladd(&kp->ki_childtime, &kp->ki_childutime); 902 903 FOREACH_THREAD_IN_PROC(p, td0) 904 kp->ki_cow += td0->td_cow; 905 906 tp = NULL; 907 if (p->p_pgrp) { 908 kp->ki_pgid = p->p_pgrp->pg_id; 909 kp->ki_jobc = p->p_pgrp->pg_jobc; 910 sp = p->p_pgrp->pg_session; 911 912 if (sp != NULL) { 913 kp->ki_sid = sp->s_sid; 914 SESS_LOCK(sp); 915 strlcpy(kp->ki_login, sp->s_login, 916 sizeof(kp->ki_login)); 917 if (sp->s_ttyvp) 918 kp->ki_kiflag |= KI_CTTY; 919 if (SESS_LEADER(p)) 920 kp->ki_kiflag |= KI_SLEADER; 921 /* XXX proctree_lock */ 922 tp = sp->s_ttyp; 923 SESS_UNLOCK(sp); 924 } 925 } 926 if ((p->p_flag & P_CONTROLT) && tp != NULL) { 927 kp->ki_tdev = tty_udev(tp); 928 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; 929 if (tp->t_session) 930 kp->ki_tsid = tp->t_session->s_sid; 931 } else 932 kp->ki_tdev = NODEV; 933 if (p->p_comm[0] != '\0') 934 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm)); 935 if (p->p_sysent && p->p_sysent->sv_name != NULL && 936 p->p_sysent->sv_name[0] != '\0') 937 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul)); 938 kp->ki_siglist = p->p_siglist; 939 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig); 940 kp->ki_acflag = p->p_acflag; 941 kp->ki_lock = p->p_lock; 942 if (p->p_pptr) { 943 kp->ki_ppid = proc_realparent(p)->p_pid; 944 if (p->p_flag & P_TRACED) 945 kp->ki_tracer = p->p_pptr->p_pid; 946 } 947 } 948 949 /* 950 * Fill in information that is thread specific. Must be called with 951 * target process locked. If 'preferthread' is set, overwrite certain 952 * process-related fields that are maintained for both threads and 953 * processes. 954 */ 955 static void 956 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread) 957 { 958 struct proc *p; 959 960 p = td->td_proc; 961 kp->ki_tdaddr = td; 962 PROC_LOCK_ASSERT(p, MA_OWNED); 963 964 if (preferthread) 965 PROC_STATLOCK(p); 966 thread_lock(td); 967 if (td->td_wmesg != NULL) 968 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg)); 969 else 970 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg)); 971 strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)); 972 if (TD_ON_LOCK(td)) { 973 kp->ki_kiflag |= KI_LOCKBLOCK; 974 strlcpy(kp->ki_lockname, td->td_lockname, 975 sizeof(kp->ki_lockname)); 976 } else { 977 kp->ki_kiflag &= ~KI_LOCKBLOCK; 978 bzero(kp->ki_lockname, sizeof(kp->ki_lockname)); 979 } 980 981 if (p->p_state == PRS_NORMAL) { /* approximate. */ 982 if (TD_ON_RUNQ(td) || 983 TD_CAN_RUN(td) || 984 TD_IS_RUNNING(td)) { 985 kp->ki_stat = SRUN; 986 } else if (P_SHOULDSTOP(p)) { 987 kp->ki_stat = SSTOP; 988 } else if (TD_IS_SLEEPING(td)) { 989 kp->ki_stat = SSLEEP; 990 } else if (TD_ON_LOCK(td)) { 991 kp->ki_stat = SLOCK; 992 } else { 993 kp->ki_stat = SWAIT; 994 } 995 } else if (p->p_state == PRS_ZOMBIE) { 996 kp->ki_stat = SZOMB; 997 } else { 998 kp->ki_stat = SIDL; 999 } 1000 1001 /* Things in the thread */ 1002 kp->ki_wchan = td->td_wchan; 1003 kp->ki_pri.pri_level = td->td_priority; 1004 kp->ki_pri.pri_native = td->td_base_pri; 1005 1006 /* 1007 * Note: legacy fields; clamp at the old NOCPU value and/or 1008 * the maximum u_char CPU value. 1009 */ 1010 if (td->td_lastcpu == NOCPU) 1011 kp->ki_lastcpu_old = NOCPU_OLD; 1012 else if (td->td_lastcpu > MAXCPU_OLD) 1013 kp->ki_lastcpu_old = MAXCPU_OLD; 1014 else 1015 kp->ki_lastcpu_old = td->td_lastcpu; 1016 1017 if (td->td_oncpu == NOCPU) 1018 kp->ki_oncpu_old = NOCPU_OLD; 1019 else if (td->td_oncpu > MAXCPU_OLD) 1020 kp->ki_oncpu_old = MAXCPU_OLD; 1021 else 1022 kp->ki_oncpu_old = td->td_oncpu; 1023 1024 kp->ki_lastcpu = td->td_lastcpu; 1025 kp->ki_oncpu = td->td_oncpu; 1026 kp->ki_tdflags = td->td_flags; 1027 kp->ki_tid = td->td_tid; 1028 kp->ki_numthreads = p->p_numthreads; 1029 kp->ki_pcb = td->td_pcb; 1030 kp->ki_kstack = (void *)td->td_kstack; 1031 kp->ki_slptime = (ticks - td->td_slptick) / hz; 1032 kp->ki_pri.pri_class = td->td_pri_class; 1033 kp->ki_pri.pri_user = td->td_user_pri; 1034 1035 if (preferthread) { 1036 rufetchtd(td, &kp->ki_rusage); 1037 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime); 1038 kp->ki_pctcpu = sched_pctcpu(td); 1039 kp->ki_estcpu = td->td_estcpu; 1040 kp->ki_cow = td->td_cow; 1041 } 1042 1043 /* We can't get this anymore but ps etc never used it anyway. */ 1044 kp->ki_rqindex = 0; 1045 1046 if (preferthread) 1047 kp->ki_siglist = td->td_siglist; 1048 kp->ki_sigmask = td->td_sigmask; 1049 thread_unlock(td); 1050 if (preferthread) 1051 PROC_STATUNLOCK(p); 1052 } 1053 1054 /* 1055 * Fill in a kinfo_proc structure for the specified process. 1056 * Must be called with the target process locked. 1057 */ 1058 void 1059 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp) 1060 { 1061 1062 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1063 1064 fill_kinfo_proc_only(p, kp); 1065 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0); 1066 fill_kinfo_aggregate(p, kp); 1067 } 1068 1069 struct pstats * 1070 pstats_alloc(void) 1071 { 1072 1073 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK)); 1074 } 1075 1076 /* 1077 * Copy parts of p_stats; zero the rest of p_stats (statistics). 1078 */ 1079 void 1080 pstats_fork(struct pstats *src, struct pstats *dst) 1081 { 1082 1083 bzero(&dst->pstat_startzero, 1084 __rangeof(struct pstats, pstat_startzero, pstat_endzero)); 1085 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy, 1086 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy)); 1087 } 1088 1089 void 1090 pstats_free(struct pstats *ps) 1091 { 1092 1093 free(ps, M_SUBPROC); 1094 } 1095 1096 static struct proc * 1097 zpfind_locked(pid_t pid) 1098 { 1099 struct proc *p; 1100 1101 sx_assert(&allproc_lock, SX_LOCKED); 1102 LIST_FOREACH(p, &zombproc, p_list) { 1103 if (p->p_pid == pid) { 1104 PROC_LOCK(p); 1105 break; 1106 } 1107 } 1108 return (p); 1109 } 1110 1111 /* 1112 * Locate a zombie process by number 1113 */ 1114 struct proc * 1115 zpfind(pid_t pid) 1116 { 1117 struct proc *p; 1118 1119 sx_slock(&allproc_lock); 1120 p = zpfind_locked(pid); 1121 sx_sunlock(&allproc_lock); 1122 return (p); 1123 } 1124 1125 #ifdef COMPAT_FREEBSD32 1126 1127 /* 1128 * This function is typically used to copy out the kernel address, so 1129 * it can be replaced by assignment of zero. 1130 */ 1131 static inline uint32_t 1132 ptr32_trim(void *ptr) 1133 { 1134 uintptr_t uptr; 1135 1136 uptr = (uintptr_t)ptr; 1137 return ((uptr > UINT_MAX) ? 0 : uptr); 1138 } 1139 1140 #define PTRTRIM_CP(src,dst,fld) \ 1141 do { (dst).fld = ptr32_trim((src).fld); } while (0) 1142 1143 static void 1144 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32) 1145 { 1146 int i; 1147 1148 bzero(ki32, sizeof(struct kinfo_proc32)); 1149 ki32->ki_structsize = sizeof(struct kinfo_proc32); 1150 CP(*ki, *ki32, ki_layout); 1151 PTRTRIM_CP(*ki, *ki32, ki_args); 1152 PTRTRIM_CP(*ki, *ki32, ki_paddr); 1153 PTRTRIM_CP(*ki, *ki32, ki_addr); 1154 PTRTRIM_CP(*ki, *ki32, ki_tracep); 1155 PTRTRIM_CP(*ki, *ki32, ki_textvp); 1156 PTRTRIM_CP(*ki, *ki32, ki_fd); 1157 PTRTRIM_CP(*ki, *ki32, ki_vmspace); 1158 PTRTRIM_CP(*ki, *ki32, ki_wchan); 1159 CP(*ki, *ki32, ki_pid); 1160 CP(*ki, *ki32, ki_ppid); 1161 CP(*ki, *ki32, ki_pgid); 1162 CP(*ki, *ki32, ki_tpgid); 1163 CP(*ki, *ki32, ki_sid); 1164 CP(*ki, *ki32, ki_tsid); 1165 CP(*ki, *ki32, ki_jobc); 1166 CP(*ki, *ki32, ki_tdev); 1167 CP(*ki, *ki32, ki_siglist); 1168 CP(*ki, *ki32, ki_sigmask); 1169 CP(*ki, *ki32, ki_sigignore); 1170 CP(*ki, *ki32, ki_sigcatch); 1171 CP(*ki, *ki32, ki_uid); 1172 CP(*ki, *ki32, ki_ruid); 1173 CP(*ki, *ki32, ki_svuid); 1174 CP(*ki, *ki32, ki_rgid); 1175 CP(*ki, *ki32, ki_svgid); 1176 CP(*ki, *ki32, ki_ngroups); 1177 for (i = 0; i < KI_NGROUPS; i++) 1178 CP(*ki, *ki32, ki_groups[i]); 1179 CP(*ki, *ki32, ki_size); 1180 CP(*ki, *ki32, ki_rssize); 1181 CP(*ki, *ki32, ki_swrss); 1182 CP(*ki, *ki32, ki_tsize); 1183 CP(*ki, *ki32, ki_dsize); 1184 CP(*ki, *ki32, ki_ssize); 1185 CP(*ki, *ki32, ki_xstat); 1186 CP(*ki, *ki32, ki_acflag); 1187 CP(*ki, *ki32, ki_pctcpu); 1188 CP(*ki, *ki32, ki_estcpu); 1189 CP(*ki, *ki32, ki_slptime); 1190 CP(*ki, *ki32, ki_swtime); 1191 CP(*ki, *ki32, ki_cow); 1192 CP(*ki, *ki32, ki_runtime); 1193 TV_CP(*ki, *ki32, ki_start); 1194 TV_CP(*ki, *ki32, ki_childtime); 1195 CP(*ki, *ki32, ki_flag); 1196 CP(*ki, *ki32, ki_kiflag); 1197 CP(*ki, *ki32, ki_traceflag); 1198 CP(*ki, *ki32, ki_stat); 1199 CP(*ki, *ki32, ki_nice); 1200 CP(*ki, *ki32, ki_lock); 1201 CP(*ki, *ki32, ki_rqindex); 1202 CP(*ki, *ki32, ki_oncpu); 1203 CP(*ki, *ki32, ki_lastcpu); 1204 1205 /* XXX TODO: wrap cpu value as appropriate */ 1206 CP(*ki, *ki32, ki_oncpu_old); 1207 CP(*ki, *ki32, ki_lastcpu_old); 1208 1209 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1); 1210 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1); 1211 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1); 1212 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1); 1213 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1); 1214 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1); 1215 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1); 1216 CP(*ki, *ki32, ki_tracer); 1217 CP(*ki, *ki32, ki_flag2); 1218 CP(*ki, *ki32, ki_fibnum); 1219 CP(*ki, *ki32, ki_cr_flags); 1220 CP(*ki, *ki32, ki_jid); 1221 CP(*ki, *ki32, ki_numthreads); 1222 CP(*ki, *ki32, ki_tid); 1223 CP(*ki, *ki32, ki_pri); 1224 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage); 1225 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch); 1226 PTRTRIM_CP(*ki, *ki32, ki_pcb); 1227 PTRTRIM_CP(*ki, *ki32, ki_kstack); 1228 PTRTRIM_CP(*ki, *ki32, ki_udata); 1229 CP(*ki, *ki32, ki_sflag); 1230 CP(*ki, *ki32, ki_tdflags); 1231 } 1232 #endif 1233 1234 int 1235 kern_proc_out(struct proc *p, struct sbuf *sb, int flags) 1236 { 1237 struct thread *td; 1238 struct kinfo_proc ki; 1239 #ifdef COMPAT_FREEBSD32 1240 struct kinfo_proc32 ki32; 1241 #endif 1242 int error; 1243 1244 PROC_LOCK_ASSERT(p, MA_OWNED); 1245 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1246 1247 error = 0; 1248 fill_kinfo_proc(p, &ki); 1249 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1250 #ifdef COMPAT_FREEBSD32 1251 if ((flags & KERN_PROC_MASK32) != 0) { 1252 freebsd32_kinfo_proc_out(&ki, &ki32); 1253 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1254 error = ENOMEM; 1255 } else 1256 #endif 1257 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1258 error = ENOMEM; 1259 } else { 1260 FOREACH_THREAD_IN_PROC(p, td) { 1261 fill_kinfo_thread(td, &ki, 1); 1262 #ifdef COMPAT_FREEBSD32 1263 if ((flags & KERN_PROC_MASK32) != 0) { 1264 freebsd32_kinfo_proc_out(&ki, &ki32); 1265 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1266 error = ENOMEM; 1267 } else 1268 #endif 1269 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1270 error = ENOMEM; 1271 if (error != 0) 1272 break; 1273 } 1274 } 1275 PROC_UNLOCK(p); 1276 return (error); 1277 } 1278 1279 static int 1280 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags, 1281 int doingzomb) 1282 { 1283 struct sbuf sb; 1284 struct kinfo_proc ki; 1285 struct proc *np; 1286 int error, error2; 1287 pid_t pid; 1288 1289 pid = p->p_pid; 1290 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); 1291 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1292 error = kern_proc_out(p, &sb, flags); 1293 error2 = sbuf_finish(&sb); 1294 sbuf_delete(&sb); 1295 if (error != 0) 1296 return (error); 1297 else if (error2 != 0) 1298 return (error2); 1299 if (doingzomb) 1300 np = zpfind(pid); 1301 else { 1302 if (pid == 0) 1303 return (0); 1304 np = pfind(pid); 1305 } 1306 if (np == NULL) 1307 return (ESRCH); 1308 if (np != p) { 1309 PROC_UNLOCK(np); 1310 return (ESRCH); 1311 } 1312 PROC_UNLOCK(np); 1313 return (0); 1314 } 1315 1316 static int 1317 sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 1318 { 1319 int *name = (int *)arg1; 1320 u_int namelen = arg2; 1321 struct proc *p; 1322 int flags, doingzomb, oid_number; 1323 int error = 0; 1324 1325 oid_number = oidp->oid_number; 1326 if (oid_number != KERN_PROC_ALL && 1327 (oid_number & KERN_PROC_INC_THREAD) == 0) 1328 flags = KERN_PROC_NOTHREADS; 1329 else { 1330 flags = 0; 1331 oid_number &= ~KERN_PROC_INC_THREAD; 1332 } 1333 #ifdef COMPAT_FREEBSD32 1334 if (req->flags & SCTL_MASK32) 1335 flags |= KERN_PROC_MASK32; 1336 #endif 1337 if (oid_number == KERN_PROC_PID) { 1338 if (namelen != 1) 1339 return (EINVAL); 1340 error = sysctl_wire_old_buffer(req, 0); 1341 if (error) 1342 return (error); 1343 sx_slock(&proctree_lock); 1344 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1345 if (error == 0) 1346 error = sysctl_out_proc(p, req, flags, 0); 1347 sx_sunlock(&proctree_lock); 1348 return (error); 1349 } 1350 1351 switch (oid_number) { 1352 case KERN_PROC_ALL: 1353 if (namelen != 0) 1354 return (EINVAL); 1355 break; 1356 case KERN_PROC_PROC: 1357 if (namelen != 0 && namelen != 1) 1358 return (EINVAL); 1359 break; 1360 default: 1361 if (namelen != 1) 1362 return (EINVAL); 1363 break; 1364 } 1365 1366 if (!req->oldptr) { 1367 /* overestimate by 5 procs */ 1368 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1369 if (error) 1370 return (error); 1371 } 1372 error = sysctl_wire_old_buffer(req, 0); 1373 if (error != 0) 1374 return (error); 1375 sx_slock(&proctree_lock); 1376 sx_slock(&allproc_lock); 1377 for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) { 1378 if (!doingzomb) 1379 p = LIST_FIRST(&allproc); 1380 else 1381 p = LIST_FIRST(&zombproc); 1382 for (; p != 0; p = LIST_NEXT(p, p_list)) { 1383 /* 1384 * Skip embryonic processes. 1385 */ 1386 PROC_LOCK(p); 1387 if (p->p_state == PRS_NEW) { 1388 PROC_UNLOCK(p); 1389 continue; 1390 } 1391 KASSERT(p->p_ucred != NULL, 1392 ("process credential is NULL for non-NEW proc")); 1393 /* 1394 * Show a user only appropriate processes. 1395 */ 1396 if (p_cansee(curthread, p)) { 1397 PROC_UNLOCK(p); 1398 continue; 1399 } 1400 /* 1401 * TODO - make more efficient (see notes below). 1402 * do by session. 1403 */ 1404 switch (oid_number) { 1405 1406 case KERN_PROC_GID: 1407 if (p->p_ucred->cr_gid != (gid_t)name[0]) { 1408 PROC_UNLOCK(p); 1409 continue; 1410 } 1411 break; 1412 1413 case KERN_PROC_PGRP: 1414 /* could do this by traversing pgrp */ 1415 if (p->p_pgrp == NULL || 1416 p->p_pgrp->pg_id != (pid_t)name[0]) { 1417 PROC_UNLOCK(p); 1418 continue; 1419 } 1420 break; 1421 1422 case KERN_PROC_RGID: 1423 if (p->p_ucred->cr_rgid != (gid_t)name[0]) { 1424 PROC_UNLOCK(p); 1425 continue; 1426 } 1427 break; 1428 1429 case KERN_PROC_SESSION: 1430 if (p->p_session == NULL || 1431 p->p_session->s_sid != (pid_t)name[0]) { 1432 PROC_UNLOCK(p); 1433 continue; 1434 } 1435 break; 1436 1437 case KERN_PROC_TTY: 1438 if ((p->p_flag & P_CONTROLT) == 0 || 1439 p->p_session == NULL) { 1440 PROC_UNLOCK(p); 1441 continue; 1442 } 1443 /* XXX proctree_lock */ 1444 SESS_LOCK(p->p_session); 1445 if (p->p_session->s_ttyp == NULL || 1446 tty_udev(p->p_session->s_ttyp) != 1447 (dev_t)name[0]) { 1448 SESS_UNLOCK(p->p_session); 1449 PROC_UNLOCK(p); 1450 continue; 1451 } 1452 SESS_UNLOCK(p->p_session); 1453 break; 1454 1455 case KERN_PROC_UID: 1456 if (p->p_ucred->cr_uid != (uid_t)name[0]) { 1457 PROC_UNLOCK(p); 1458 continue; 1459 } 1460 break; 1461 1462 case KERN_PROC_RUID: 1463 if (p->p_ucred->cr_ruid != (uid_t)name[0]) { 1464 PROC_UNLOCK(p); 1465 continue; 1466 } 1467 break; 1468 1469 case KERN_PROC_PROC: 1470 break; 1471 1472 default: 1473 break; 1474 1475 } 1476 1477 error = sysctl_out_proc(p, req, flags, doingzomb); 1478 if (error) { 1479 sx_sunlock(&allproc_lock); 1480 sx_sunlock(&proctree_lock); 1481 return (error); 1482 } 1483 } 1484 } 1485 sx_sunlock(&allproc_lock); 1486 sx_sunlock(&proctree_lock); 1487 return (0); 1488 } 1489 1490 struct pargs * 1491 pargs_alloc(int len) 1492 { 1493 struct pargs *pa; 1494 1495 pa = malloc(sizeof(struct pargs) + len, M_PARGS, 1496 M_WAITOK); 1497 refcount_init(&pa->ar_ref, 1); 1498 pa->ar_length = len; 1499 return (pa); 1500 } 1501 1502 static void 1503 pargs_free(struct pargs *pa) 1504 { 1505 1506 free(pa, M_PARGS); 1507 } 1508 1509 void 1510 pargs_hold(struct pargs *pa) 1511 { 1512 1513 if (pa == NULL) 1514 return; 1515 refcount_acquire(&pa->ar_ref); 1516 } 1517 1518 void 1519 pargs_drop(struct pargs *pa) 1520 { 1521 1522 if (pa == NULL) 1523 return; 1524 if (refcount_release(&pa->ar_ref)) 1525 pargs_free(pa); 1526 } 1527 1528 static int 1529 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf, 1530 size_t len) 1531 { 1532 struct iovec iov; 1533 struct uio uio; 1534 1535 iov.iov_base = (caddr_t)buf; 1536 iov.iov_len = len; 1537 uio.uio_iov = &iov; 1538 uio.uio_iovcnt = 1; 1539 uio.uio_offset = offset; 1540 uio.uio_resid = (ssize_t)len; 1541 uio.uio_segflg = UIO_SYSSPACE; 1542 uio.uio_rw = UIO_READ; 1543 uio.uio_td = td; 1544 1545 return (proc_rwmem(p, &uio)); 1546 } 1547 1548 static int 1549 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, 1550 size_t len) 1551 { 1552 size_t i; 1553 int error; 1554 1555 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len); 1556 /* 1557 * Reading the chunk may validly return EFAULT if the string is shorter 1558 * than the chunk and is aligned at the end of the page, assuming the 1559 * next page is not mapped. So if EFAULT is returned do a fallback to 1560 * one byte read loop. 1561 */ 1562 if (error == EFAULT) { 1563 for (i = 0; i < len; i++, buf++, sptr++) { 1564 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1); 1565 if (error != 0) 1566 return (error); 1567 if (*buf == '\0') 1568 break; 1569 } 1570 error = 0; 1571 } 1572 return (error); 1573 } 1574 1575 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ 1576 1577 enum proc_vector_type { 1578 PROC_ARG, 1579 PROC_ENV, 1580 PROC_AUX, 1581 }; 1582 1583 #ifdef COMPAT_FREEBSD32 1584 static int 1585 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, 1586 size_t *vsizep, enum proc_vector_type type) 1587 { 1588 struct freebsd32_ps_strings pss; 1589 Elf32_Auxinfo aux; 1590 vm_offset_t vptr, ptr; 1591 uint32_t *proc_vector32; 1592 char **proc_vector; 1593 size_t vsize, size; 1594 int i, error; 1595 1596 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), 1597 &pss, sizeof(pss)); 1598 if (error != 0) 1599 return (error); 1600 switch (type) { 1601 case PROC_ARG: 1602 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); 1603 vsize = pss.ps_nargvstr; 1604 if (vsize > ARG_MAX) 1605 return (ENOEXEC); 1606 size = vsize * sizeof(int32_t); 1607 break; 1608 case PROC_ENV: 1609 vptr = (vm_offset_t)PTRIN(pss.ps_envstr); 1610 vsize = pss.ps_nenvstr; 1611 if (vsize > ARG_MAX) 1612 return (ENOEXEC); 1613 size = vsize * sizeof(int32_t); 1614 break; 1615 case PROC_AUX: 1616 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + 1617 (pss.ps_nenvstr + 1) * sizeof(int32_t); 1618 if (vptr % 4 != 0) 1619 return (ENOEXEC); 1620 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1621 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); 1622 if (error != 0) 1623 return (error); 1624 if (aux.a_type == AT_NULL) 1625 break; 1626 ptr += sizeof(aux); 1627 } 1628 if (aux.a_type != AT_NULL) 1629 return (ENOEXEC); 1630 vsize = i + 1; 1631 size = vsize * sizeof(aux); 1632 break; 1633 default: 1634 KASSERT(0, ("Wrong proc vector type: %d", type)); 1635 return (EINVAL); 1636 } 1637 proc_vector32 = malloc(size, M_TEMP, M_WAITOK); 1638 error = proc_read_mem(td, p, vptr, proc_vector32, size); 1639 if (error != 0) 1640 goto done; 1641 if (type == PROC_AUX) { 1642 *proc_vectorp = (char **)proc_vector32; 1643 *vsizep = vsize; 1644 return (0); 1645 } 1646 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); 1647 for (i = 0; i < (int)vsize; i++) 1648 proc_vector[i] = PTRIN(proc_vector32[i]); 1649 *proc_vectorp = proc_vector; 1650 *vsizep = vsize; 1651 done: 1652 free(proc_vector32, M_TEMP); 1653 return (error); 1654 } 1655 #endif 1656 1657 static int 1658 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, 1659 size_t *vsizep, enum proc_vector_type type) 1660 { 1661 struct ps_strings pss; 1662 Elf_Auxinfo aux; 1663 vm_offset_t vptr, ptr; 1664 char **proc_vector; 1665 size_t vsize, size; 1666 int error, i; 1667 1668 #ifdef COMPAT_FREEBSD32 1669 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1670 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); 1671 #endif 1672 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), 1673 &pss, sizeof(pss)); 1674 if (error != 0) 1675 return (error); 1676 switch (type) { 1677 case PROC_ARG: 1678 vptr = (vm_offset_t)pss.ps_argvstr; 1679 vsize = pss.ps_nargvstr; 1680 if (vsize > ARG_MAX) 1681 return (ENOEXEC); 1682 size = vsize * sizeof(char *); 1683 break; 1684 case PROC_ENV: 1685 vptr = (vm_offset_t)pss.ps_envstr; 1686 vsize = pss.ps_nenvstr; 1687 if (vsize > ARG_MAX) 1688 return (ENOEXEC); 1689 size = vsize * sizeof(char *); 1690 break; 1691 case PROC_AUX: 1692 /* 1693 * The aux array is just above env array on the stack. Check 1694 * that the address is naturally aligned. 1695 */ 1696 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) 1697 * sizeof(char *); 1698 #if __ELF_WORD_SIZE == 64 1699 if (vptr % sizeof(uint64_t) != 0) 1700 #else 1701 if (vptr % sizeof(uint32_t) != 0) 1702 #endif 1703 return (ENOEXEC); 1704 /* 1705 * We count the array size reading the aux vectors from the 1706 * stack until AT_NULL vector is returned. So (to keep the code 1707 * simple) we read the process stack twice: the first time here 1708 * to find the size and the second time when copying the vectors 1709 * to the allocated proc_vector. 1710 */ 1711 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1712 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); 1713 if (error != 0) 1714 return (error); 1715 if (aux.a_type == AT_NULL) 1716 break; 1717 ptr += sizeof(aux); 1718 } 1719 /* 1720 * If the PROC_AUXV_MAX entries are iterated over, and we have 1721 * not reached AT_NULL, it is most likely we are reading wrong 1722 * data: either the process doesn't have auxv array or data has 1723 * been modified. Return the error in this case. 1724 */ 1725 if (aux.a_type != AT_NULL) 1726 return (ENOEXEC); 1727 vsize = i + 1; 1728 size = vsize * sizeof(aux); 1729 break; 1730 default: 1731 KASSERT(0, ("Wrong proc vector type: %d", type)); 1732 return (EINVAL); /* In case we are built without INVARIANTS. */ 1733 } 1734 proc_vector = malloc(size, M_TEMP, M_WAITOK); 1735 if (proc_vector == NULL) 1736 return (ENOMEM); 1737 error = proc_read_mem(td, p, vptr, proc_vector, size); 1738 if (error != 0) { 1739 free(proc_vector, M_TEMP); 1740 return (error); 1741 } 1742 *proc_vectorp = proc_vector; 1743 *vsizep = vsize; 1744 1745 return (0); 1746 } 1747 1748 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ 1749 1750 static int 1751 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, 1752 enum proc_vector_type type) 1753 { 1754 size_t done, len, nchr, vsize; 1755 int error, i; 1756 char **proc_vector, *sptr; 1757 char pss_string[GET_PS_STRINGS_CHUNK_SZ]; 1758 1759 PROC_ASSERT_HELD(p); 1760 1761 /* 1762 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. 1763 */ 1764 nchr = 2 * (PATH_MAX + ARG_MAX); 1765 1766 error = get_proc_vector(td, p, &proc_vector, &vsize, type); 1767 if (error != 0) 1768 return (error); 1769 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { 1770 /* 1771 * The program may have scribbled into its argv array, e.g. to 1772 * remove some arguments. If that has happened, break out 1773 * before trying to read from NULL. 1774 */ 1775 if (proc_vector[i] == NULL) 1776 break; 1777 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { 1778 error = proc_read_string(td, p, sptr, pss_string, 1779 sizeof(pss_string)); 1780 if (error != 0) 1781 goto done; 1782 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); 1783 if (done + len >= nchr) 1784 len = nchr - done - 1; 1785 sbuf_bcat(sb, pss_string, len); 1786 if (len != GET_PS_STRINGS_CHUNK_SZ) 1787 break; 1788 done += GET_PS_STRINGS_CHUNK_SZ; 1789 } 1790 sbuf_bcat(sb, "", 1); 1791 done += len + 1; 1792 } 1793 done: 1794 free(proc_vector, M_TEMP); 1795 return (error); 1796 } 1797 1798 int 1799 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) 1800 { 1801 1802 return (get_ps_strings(curthread, p, sb, PROC_ARG)); 1803 } 1804 1805 int 1806 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) 1807 { 1808 1809 return (get_ps_strings(curthread, p, sb, PROC_ENV)); 1810 } 1811 1812 int 1813 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) 1814 { 1815 size_t vsize, size; 1816 char **auxv; 1817 int error; 1818 1819 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); 1820 if (error == 0) { 1821 #ifdef COMPAT_FREEBSD32 1822 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1823 size = vsize * sizeof(Elf32_Auxinfo); 1824 else 1825 #endif 1826 size = vsize * sizeof(Elf_Auxinfo); 1827 if (sbuf_bcat(sb, auxv, size) != 0) 1828 error = ENOMEM; 1829 free(auxv, M_TEMP); 1830 } 1831 return (error); 1832 } 1833 1834 /* 1835 * This sysctl allows a process to retrieve the argument list or process 1836 * title for another process without groping around in the address space 1837 * of the other process. It also allow a process to set its own "process 1838 * title to a string of its own choice. 1839 */ 1840 static int 1841 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 1842 { 1843 int *name = (int *)arg1; 1844 u_int namelen = arg2; 1845 struct pargs *newpa, *pa; 1846 struct proc *p; 1847 struct sbuf sb; 1848 int flags, error = 0, error2; 1849 1850 if (namelen != 1) 1851 return (EINVAL); 1852 1853 flags = PGET_CANSEE; 1854 if (req->newptr != NULL) 1855 flags |= PGET_ISCURRENT; 1856 error = pget((pid_t)name[0], flags, &p); 1857 if (error) 1858 return (error); 1859 1860 pa = p->p_args; 1861 if (pa != NULL) { 1862 pargs_hold(pa); 1863 PROC_UNLOCK(p); 1864 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 1865 pargs_drop(pa); 1866 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { 1867 _PHOLD(p); 1868 PROC_UNLOCK(p); 1869 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1870 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1871 error = proc_getargv(curthread, p, &sb); 1872 error2 = sbuf_finish(&sb); 1873 PRELE(p); 1874 sbuf_delete(&sb); 1875 if (error == 0 && error2 != 0) 1876 error = error2; 1877 } else { 1878 PROC_UNLOCK(p); 1879 } 1880 if (error != 0 || req->newptr == NULL) 1881 return (error); 1882 1883 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit) 1884 return (ENOMEM); 1885 newpa = pargs_alloc(req->newlen); 1886 error = SYSCTL_IN(req, newpa->ar_args, req->newlen); 1887 if (error != 0) { 1888 pargs_free(newpa); 1889 return (error); 1890 } 1891 PROC_LOCK(p); 1892 pa = p->p_args; 1893 p->p_args = newpa; 1894 PROC_UNLOCK(p); 1895 pargs_drop(pa); 1896 return (0); 1897 } 1898 1899 /* 1900 * This sysctl allows a process to retrieve environment of another process. 1901 */ 1902 static int 1903 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) 1904 { 1905 int *name = (int *)arg1; 1906 u_int namelen = arg2; 1907 struct proc *p; 1908 struct sbuf sb; 1909 int error, error2; 1910 1911 if (namelen != 1) 1912 return (EINVAL); 1913 1914 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1915 if (error != 0) 1916 return (error); 1917 if ((p->p_flag & P_SYSTEM) != 0) { 1918 PRELE(p); 1919 return (0); 1920 } 1921 1922 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1923 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1924 error = proc_getenvv(curthread, p, &sb); 1925 error2 = sbuf_finish(&sb); 1926 PRELE(p); 1927 sbuf_delete(&sb); 1928 return (error != 0 ? error : error2); 1929 } 1930 1931 /* 1932 * This sysctl allows a process to retrieve ELF auxiliary vector of 1933 * another process. 1934 */ 1935 static int 1936 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) 1937 { 1938 int *name = (int *)arg1; 1939 u_int namelen = arg2; 1940 struct proc *p; 1941 struct sbuf sb; 1942 int error, error2; 1943 1944 if (namelen != 1) 1945 return (EINVAL); 1946 1947 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1948 if (error != 0) 1949 return (error); 1950 if ((p->p_flag & P_SYSTEM) != 0) { 1951 PRELE(p); 1952 return (0); 1953 } 1954 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1955 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1956 error = proc_getauxv(curthread, p, &sb); 1957 error2 = sbuf_finish(&sb); 1958 PRELE(p); 1959 sbuf_delete(&sb); 1960 return (error != 0 ? error : error2); 1961 } 1962 1963 /* 1964 * This sysctl allows a process to retrieve the path of the executable for 1965 * itself or another process. 1966 */ 1967 static int 1968 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) 1969 { 1970 pid_t *pidp = (pid_t *)arg1; 1971 unsigned int arglen = arg2; 1972 struct proc *p; 1973 struct vnode *vp; 1974 char *retbuf, *freebuf; 1975 int error; 1976 1977 if (arglen != 1) 1978 return (EINVAL); 1979 if (*pidp == -1) { /* -1 means this process */ 1980 p = req->td->td_proc; 1981 } else { 1982 error = pget(*pidp, PGET_CANSEE, &p); 1983 if (error != 0) 1984 return (error); 1985 } 1986 1987 vp = p->p_textvp; 1988 if (vp == NULL) { 1989 if (*pidp != -1) 1990 PROC_UNLOCK(p); 1991 return (0); 1992 } 1993 vref(vp); 1994 if (*pidp != -1) 1995 PROC_UNLOCK(p); 1996 error = vn_fullpath(req->td, vp, &retbuf, &freebuf); 1997 vrele(vp); 1998 if (error) 1999 return (error); 2000 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); 2001 free(freebuf, M_TEMP); 2002 return (error); 2003 } 2004 2005 static int 2006 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) 2007 { 2008 struct proc *p; 2009 char *sv_name; 2010 int *name; 2011 int namelen; 2012 int error; 2013 2014 namelen = arg2; 2015 if (namelen != 1) 2016 return (EINVAL); 2017 2018 name = (int *)arg1; 2019 error = pget((pid_t)name[0], PGET_CANSEE, &p); 2020 if (error != 0) 2021 return (error); 2022 sv_name = p->p_sysent->sv_name; 2023 PROC_UNLOCK(p); 2024 return (sysctl_handle_string(oidp, sv_name, 0, req)); 2025 } 2026 2027 #ifdef KINFO_OVMENTRY_SIZE 2028 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); 2029 #endif 2030 2031 #ifdef COMPAT_FREEBSD7 2032 static int 2033 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) 2034 { 2035 vm_map_entry_t entry, tmp_entry; 2036 unsigned int last_timestamp; 2037 char *fullpath, *freepath; 2038 struct kinfo_ovmentry *kve; 2039 struct vattr va; 2040 struct ucred *cred; 2041 int error, *name; 2042 struct vnode *vp; 2043 struct proc *p; 2044 vm_map_t map; 2045 struct vmspace *vm; 2046 2047 name = (int *)arg1; 2048 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2049 if (error != 0) 2050 return (error); 2051 vm = vmspace_acquire_ref(p); 2052 if (vm == NULL) { 2053 PRELE(p); 2054 return (ESRCH); 2055 } 2056 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2057 2058 map = &vm->vm_map; 2059 vm_map_lock_read(map); 2060 for (entry = map->header.next; entry != &map->header; 2061 entry = entry->next) { 2062 vm_object_t obj, tobj, lobj; 2063 vm_offset_t addr; 2064 2065 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2066 continue; 2067 2068 bzero(kve, sizeof(*kve)); 2069 kve->kve_structsize = sizeof(*kve); 2070 2071 kve->kve_private_resident = 0; 2072 obj = entry->object.vm_object; 2073 if (obj != NULL) { 2074 VM_OBJECT_RLOCK(obj); 2075 if (obj->shadow_count == 1) 2076 kve->kve_private_resident = 2077 obj->resident_page_count; 2078 } 2079 kve->kve_resident = 0; 2080 addr = entry->start; 2081 while (addr < entry->end) { 2082 if (pmap_extract(map->pmap, addr)) 2083 kve->kve_resident++; 2084 addr += PAGE_SIZE; 2085 } 2086 2087 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { 2088 if (tobj != obj) 2089 VM_OBJECT_RLOCK(tobj); 2090 if (lobj != obj) 2091 VM_OBJECT_RUNLOCK(lobj); 2092 lobj = tobj; 2093 } 2094 2095 kve->kve_start = (void*)entry->start; 2096 kve->kve_end = (void*)entry->end; 2097 kve->kve_offset = (off_t)entry->offset; 2098 2099 if (entry->protection & VM_PROT_READ) 2100 kve->kve_protection |= KVME_PROT_READ; 2101 if (entry->protection & VM_PROT_WRITE) 2102 kve->kve_protection |= KVME_PROT_WRITE; 2103 if (entry->protection & VM_PROT_EXECUTE) 2104 kve->kve_protection |= KVME_PROT_EXEC; 2105 2106 if (entry->eflags & MAP_ENTRY_COW) 2107 kve->kve_flags |= KVME_FLAG_COW; 2108 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2109 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2110 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2111 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2112 2113 last_timestamp = map->timestamp; 2114 vm_map_unlock_read(map); 2115 2116 kve->kve_fileid = 0; 2117 kve->kve_fsid = 0; 2118 freepath = NULL; 2119 fullpath = ""; 2120 if (lobj) { 2121 vp = NULL; 2122 switch (lobj->type) { 2123 case OBJT_DEFAULT: 2124 kve->kve_type = KVME_TYPE_DEFAULT; 2125 break; 2126 case OBJT_VNODE: 2127 kve->kve_type = KVME_TYPE_VNODE; 2128 vp = lobj->handle; 2129 vref(vp); 2130 break; 2131 case OBJT_SWAP: 2132 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) { 2133 kve->kve_type = KVME_TYPE_VNODE; 2134 if ((lobj->flags & OBJ_TMPFS) != 0) { 2135 vp = lobj->un_pager.swp.swp_tmpfs; 2136 vref(vp); 2137 } 2138 } else { 2139 kve->kve_type = KVME_TYPE_SWAP; 2140 } 2141 break; 2142 case OBJT_DEVICE: 2143 kve->kve_type = KVME_TYPE_DEVICE; 2144 break; 2145 case OBJT_PHYS: 2146 kve->kve_type = KVME_TYPE_PHYS; 2147 break; 2148 case OBJT_DEAD: 2149 kve->kve_type = KVME_TYPE_DEAD; 2150 break; 2151 case OBJT_SG: 2152 kve->kve_type = KVME_TYPE_SG; 2153 break; 2154 default: 2155 kve->kve_type = KVME_TYPE_UNKNOWN; 2156 break; 2157 } 2158 if (lobj != obj) 2159 VM_OBJECT_RUNLOCK(lobj); 2160 2161 kve->kve_ref_count = obj->ref_count; 2162 kve->kve_shadow_count = obj->shadow_count; 2163 VM_OBJECT_RUNLOCK(obj); 2164 if (vp != NULL) { 2165 vn_fullpath(curthread, vp, &fullpath, 2166 &freepath); 2167 cred = curthread->td_ucred; 2168 vn_lock(vp, LK_SHARED | LK_RETRY); 2169 if (VOP_GETATTR(vp, &va, cred) == 0) { 2170 kve->kve_fileid = va.va_fileid; 2171 kve->kve_fsid = va.va_fsid; 2172 } 2173 vput(vp); 2174 } 2175 } else { 2176 kve->kve_type = KVME_TYPE_NONE; 2177 kve->kve_ref_count = 0; 2178 kve->kve_shadow_count = 0; 2179 } 2180 2181 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2182 if (freepath != NULL) 2183 free(freepath, M_TEMP); 2184 2185 error = SYSCTL_OUT(req, kve, sizeof(*kve)); 2186 vm_map_lock_read(map); 2187 if (error) 2188 break; 2189 if (last_timestamp != map->timestamp) { 2190 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2191 entry = tmp_entry; 2192 } 2193 } 2194 vm_map_unlock_read(map); 2195 vmspace_free(vm); 2196 PRELE(p); 2197 free(kve, M_TEMP); 2198 return (error); 2199 } 2200 #endif /* COMPAT_FREEBSD7 */ 2201 2202 #ifdef KINFO_VMENTRY_SIZE 2203 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); 2204 #endif 2205 2206 static void 2207 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, 2208 struct kinfo_vmentry *kve) 2209 { 2210 vm_object_t obj, tobj; 2211 vm_page_t m, m_adv; 2212 vm_offset_t addr; 2213 vm_paddr_t locked_pa; 2214 vm_pindex_t pi, pi_adv, pindex; 2215 2216 locked_pa = 0; 2217 obj = entry->object.vm_object; 2218 addr = entry->start; 2219 m_adv = NULL; 2220 pi = OFF_TO_IDX(entry->offset); 2221 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { 2222 if (m_adv != NULL) { 2223 m = m_adv; 2224 } else { 2225 pi_adv = OFF_TO_IDX(entry->end - addr); 2226 pindex = pi; 2227 for (tobj = obj;; tobj = tobj->backing_object) { 2228 m = vm_page_find_least(tobj, pindex); 2229 if (m != NULL) { 2230 if (m->pindex == pindex) 2231 break; 2232 if (pi_adv > m->pindex - pindex) { 2233 pi_adv = m->pindex - pindex; 2234 m_adv = m; 2235 } 2236 } 2237 if (tobj->backing_object == NULL) 2238 goto next; 2239 pindex += OFF_TO_IDX(tobj-> 2240 backing_object_offset); 2241 } 2242 } 2243 m_adv = NULL; 2244 if (m->psind != 0 && addr + pagesizes[1] <= entry->end && 2245 (addr & (pagesizes[1] - 1)) == 0 && 2246 (pmap_mincore(map->pmap, addr, &locked_pa) & 2247 MINCORE_SUPER) != 0) { 2248 kve->kve_flags |= KVME_FLAG_SUPER; 2249 pi_adv = OFF_TO_IDX(pagesizes[1]); 2250 } else { 2251 /* 2252 * We do not test the found page on validity. 2253 * Either the page is busy and being paged in, 2254 * or it was invalidated. The first case 2255 * should be counted as resident, the second 2256 * is not so clear; we do account both. 2257 */ 2258 pi_adv = 1; 2259 } 2260 kve->kve_resident += pi_adv; 2261 next:; 2262 } 2263 PA_UNLOCK_COND(locked_pa); 2264 } 2265 2266 /* 2267 * Must be called with the process locked and will return unlocked. 2268 */ 2269 int 2270 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags) 2271 { 2272 vm_map_entry_t entry, tmp_entry; 2273 struct vattr va; 2274 vm_map_t map; 2275 vm_object_t obj, tobj, lobj; 2276 char *fullpath, *freepath; 2277 struct kinfo_vmentry *kve; 2278 struct ucred *cred; 2279 struct vnode *vp; 2280 struct vmspace *vm; 2281 vm_offset_t addr; 2282 unsigned int last_timestamp; 2283 int error; 2284 2285 PROC_LOCK_ASSERT(p, MA_OWNED); 2286 2287 _PHOLD(p); 2288 PROC_UNLOCK(p); 2289 vm = vmspace_acquire_ref(p); 2290 if (vm == NULL) { 2291 PRELE(p); 2292 return (ESRCH); 2293 } 2294 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO); 2295 2296 error = 0; 2297 map = &vm->vm_map; 2298 vm_map_lock_read(map); 2299 for (entry = map->header.next; entry != &map->header; 2300 entry = entry->next) { 2301 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2302 continue; 2303 2304 addr = entry->end; 2305 bzero(kve, sizeof(*kve)); 2306 obj = entry->object.vm_object; 2307 if (obj != NULL) { 2308 for (tobj = obj; tobj != NULL; 2309 tobj = tobj->backing_object) { 2310 VM_OBJECT_RLOCK(tobj); 2311 lobj = tobj; 2312 } 2313 if (obj->backing_object == NULL) 2314 kve->kve_private_resident = 2315 obj->resident_page_count; 2316 if (!vmmap_skip_res_cnt) 2317 kern_proc_vmmap_resident(map, entry, kve); 2318 for (tobj = obj; tobj != NULL; 2319 tobj = tobj->backing_object) { 2320 if (tobj != obj && tobj != lobj) 2321 VM_OBJECT_RUNLOCK(tobj); 2322 } 2323 } else { 2324 lobj = NULL; 2325 } 2326 2327 kve->kve_start = entry->start; 2328 kve->kve_end = entry->end; 2329 kve->kve_offset = entry->offset; 2330 2331 if (entry->protection & VM_PROT_READ) 2332 kve->kve_protection |= KVME_PROT_READ; 2333 if (entry->protection & VM_PROT_WRITE) 2334 kve->kve_protection |= KVME_PROT_WRITE; 2335 if (entry->protection & VM_PROT_EXECUTE) 2336 kve->kve_protection |= KVME_PROT_EXEC; 2337 2338 if (entry->eflags & MAP_ENTRY_COW) 2339 kve->kve_flags |= KVME_FLAG_COW; 2340 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2341 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2342 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2343 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2344 if (entry->eflags & MAP_ENTRY_GROWS_UP) 2345 kve->kve_flags |= KVME_FLAG_GROWS_UP; 2346 if (entry->eflags & MAP_ENTRY_GROWS_DOWN) 2347 kve->kve_flags |= KVME_FLAG_GROWS_DOWN; 2348 2349 last_timestamp = map->timestamp; 2350 vm_map_unlock_read(map); 2351 2352 freepath = NULL; 2353 fullpath = ""; 2354 if (lobj != NULL) { 2355 vp = NULL; 2356 switch (lobj->type) { 2357 case OBJT_DEFAULT: 2358 kve->kve_type = KVME_TYPE_DEFAULT; 2359 break; 2360 case OBJT_VNODE: 2361 kve->kve_type = KVME_TYPE_VNODE; 2362 vp = lobj->handle; 2363 vref(vp); 2364 break; 2365 case OBJT_SWAP: 2366 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) { 2367 kve->kve_type = KVME_TYPE_VNODE; 2368 if ((lobj->flags & OBJ_TMPFS) != 0) { 2369 vp = lobj->un_pager.swp.swp_tmpfs; 2370 vref(vp); 2371 } 2372 } else { 2373 kve->kve_type = KVME_TYPE_SWAP; 2374 } 2375 break; 2376 case OBJT_DEVICE: 2377 kve->kve_type = KVME_TYPE_DEVICE; 2378 break; 2379 case OBJT_PHYS: 2380 kve->kve_type = KVME_TYPE_PHYS; 2381 break; 2382 case OBJT_DEAD: 2383 kve->kve_type = KVME_TYPE_DEAD; 2384 break; 2385 case OBJT_SG: 2386 kve->kve_type = KVME_TYPE_SG; 2387 break; 2388 case OBJT_MGTDEVICE: 2389 kve->kve_type = KVME_TYPE_MGTDEVICE; 2390 break; 2391 default: 2392 kve->kve_type = KVME_TYPE_UNKNOWN; 2393 break; 2394 } 2395 if (lobj != obj) 2396 VM_OBJECT_RUNLOCK(lobj); 2397 2398 kve->kve_ref_count = obj->ref_count; 2399 kve->kve_shadow_count = obj->shadow_count; 2400 VM_OBJECT_RUNLOCK(obj); 2401 if (vp != NULL) { 2402 vn_fullpath(curthread, vp, &fullpath, 2403 &freepath); 2404 kve->kve_vn_type = vntype_to_kinfo(vp->v_type); 2405 cred = curthread->td_ucred; 2406 vn_lock(vp, LK_SHARED | LK_RETRY); 2407 if (VOP_GETATTR(vp, &va, cred) == 0) { 2408 kve->kve_vn_fileid = va.va_fileid; 2409 kve->kve_vn_fsid = va.va_fsid; 2410 kve->kve_vn_mode = 2411 MAKEIMODE(va.va_type, va.va_mode); 2412 kve->kve_vn_size = va.va_size; 2413 kve->kve_vn_rdev = va.va_rdev; 2414 kve->kve_status = KF_ATTR_VALID; 2415 } 2416 vput(vp); 2417 } 2418 } else { 2419 kve->kve_type = KVME_TYPE_NONE; 2420 kve->kve_ref_count = 0; 2421 kve->kve_shadow_count = 0; 2422 } 2423 2424 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2425 if (freepath != NULL) 2426 free(freepath, M_TEMP); 2427 2428 /* Pack record size down */ 2429 if ((flags & KERN_VMMAP_PACK_KINFO) != 0) 2430 kve->kve_structsize = 2431 offsetof(struct kinfo_vmentry, kve_path) + 2432 strlen(kve->kve_path) + 1; 2433 else 2434 kve->kve_structsize = sizeof(*kve); 2435 kve->kve_structsize = roundup(kve->kve_structsize, 2436 sizeof(uint64_t)); 2437 2438 /* Halt filling and truncate rather than exceeding maxlen */ 2439 if (maxlen != -1 && maxlen < kve->kve_structsize) { 2440 error = 0; 2441 vm_map_lock_read(map); 2442 break; 2443 } else if (maxlen != -1) 2444 maxlen -= kve->kve_structsize; 2445 2446 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0) 2447 error = ENOMEM; 2448 vm_map_lock_read(map); 2449 if (error != 0) 2450 break; 2451 if (last_timestamp != map->timestamp) { 2452 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2453 entry = tmp_entry; 2454 } 2455 } 2456 vm_map_unlock_read(map); 2457 vmspace_free(vm); 2458 PRELE(p); 2459 free(kve, M_TEMP); 2460 return (error); 2461 } 2462 2463 static int 2464 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) 2465 { 2466 struct proc *p; 2467 struct sbuf sb; 2468 int error, error2, *name; 2469 2470 name = (int *)arg1; 2471 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); 2472 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2473 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); 2474 if (error != 0) { 2475 sbuf_delete(&sb); 2476 return (error); 2477 } 2478 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO); 2479 error2 = sbuf_finish(&sb); 2480 sbuf_delete(&sb); 2481 return (error != 0 ? error : error2); 2482 } 2483 2484 #if defined(STACK) || defined(DDB) 2485 static int 2486 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) 2487 { 2488 struct kinfo_kstack *kkstp; 2489 int error, i, *name, numthreads; 2490 lwpid_t *lwpidarray; 2491 struct thread *td; 2492 struct stack *st; 2493 struct sbuf sb; 2494 struct proc *p; 2495 2496 name = (int *)arg1; 2497 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); 2498 if (error != 0) 2499 return (error); 2500 2501 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); 2502 st = stack_create(); 2503 2504 lwpidarray = NULL; 2505 numthreads = 0; 2506 PROC_LOCK(p); 2507 repeat: 2508 if (numthreads < p->p_numthreads) { 2509 if (lwpidarray != NULL) { 2510 free(lwpidarray, M_TEMP); 2511 lwpidarray = NULL; 2512 } 2513 numthreads = p->p_numthreads; 2514 PROC_UNLOCK(p); 2515 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, 2516 M_WAITOK | M_ZERO); 2517 PROC_LOCK(p); 2518 goto repeat; 2519 } 2520 i = 0; 2521 2522 /* 2523 * XXXRW: During the below loop, execve(2) and countless other sorts 2524 * of changes could have taken place. Should we check to see if the 2525 * vmspace has been replaced, or the like, in order to prevent 2526 * giving a snapshot that spans, say, execve(2), with some threads 2527 * before and some after? Among other things, the credentials could 2528 * have changed, in which case the right to extract debug info might 2529 * no longer be assured. 2530 */ 2531 FOREACH_THREAD_IN_PROC(p, td) { 2532 KASSERT(i < numthreads, 2533 ("sysctl_kern_proc_kstack: numthreads")); 2534 lwpidarray[i] = td->td_tid; 2535 i++; 2536 } 2537 numthreads = i; 2538 for (i = 0; i < numthreads; i++) { 2539 td = thread_find(p, lwpidarray[i]); 2540 if (td == NULL) { 2541 continue; 2542 } 2543 bzero(kkstp, sizeof(*kkstp)); 2544 (void)sbuf_new(&sb, kkstp->kkst_trace, 2545 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); 2546 thread_lock(td); 2547 kkstp->kkst_tid = td->td_tid; 2548 if (TD_IS_SWAPPED(td)) { 2549 kkstp->kkst_state = KKST_STATE_SWAPPED; 2550 } else if (TD_IS_RUNNING(td)) { 2551 if (stack_save_td_running(st, td) == 0) 2552 kkstp->kkst_state = KKST_STATE_STACKOK; 2553 else 2554 kkstp->kkst_state = KKST_STATE_RUNNING; 2555 } else { 2556 kkstp->kkst_state = KKST_STATE_STACKOK; 2557 stack_save_td(st, td); 2558 } 2559 thread_unlock(td); 2560 PROC_UNLOCK(p); 2561 stack_sbuf_print(&sb, st); 2562 sbuf_finish(&sb); 2563 sbuf_delete(&sb); 2564 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); 2565 PROC_LOCK(p); 2566 if (error) 2567 break; 2568 } 2569 _PRELE(p); 2570 PROC_UNLOCK(p); 2571 if (lwpidarray != NULL) 2572 free(lwpidarray, M_TEMP); 2573 stack_destroy(st); 2574 free(kkstp, M_TEMP); 2575 return (error); 2576 } 2577 #endif 2578 2579 /* 2580 * This sysctl allows a process to retrieve the full list of groups from 2581 * itself or another process. 2582 */ 2583 static int 2584 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) 2585 { 2586 pid_t *pidp = (pid_t *)arg1; 2587 unsigned int arglen = arg2; 2588 struct proc *p; 2589 struct ucred *cred; 2590 int error; 2591 2592 if (arglen != 1) 2593 return (EINVAL); 2594 if (*pidp == -1) { /* -1 means this process */ 2595 p = req->td->td_proc; 2596 PROC_LOCK(p); 2597 } else { 2598 error = pget(*pidp, PGET_CANSEE, &p); 2599 if (error != 0) 2600 return (error); 2601 } 2602 2603 cred = crhold(p->p_ucred); 2604 PROC_UNLOCK(p); 2605 2606 error = SYSCTL_OUT(req, cred->cr_groups, 2607 cred->cr_ngroups * sizeof(gid_t)); 2608 crfree(cred); 2609 return (error); 2610 } 2611 2612 /* 2613 * This sysctl allows a process to retrieve or/and set the resource limit for 2614 * another process. 2615 */ 2616 static int 2617 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) 2618 { 2619 int *name = (int *)arg1; 2620 u_int namelen = arg2; 2621 struct rlimit rlim; 2622 struct proc *p; 2623 u_int which; 2624 int flags, error; 2625 2626 if (namelen != 2) 2627 return (EINVAL); 2628 2629 which = (u_int)name[1]; 2630 if (which >= RLIM_NLIMITS) 2631 return (EINVAL); 2632 2633 if (req->newptr != NULL && req->newlen != sizeof(rlim)) 2634 return (EINVAL); 2635 2636 flags = PGET_HOLD | PGET_NOTWEXIT; 2637 if (req->newptr != NULL) 2638 flags |= PGET_CANDEBUG; 2639 else 2640 flags |= PGET_CANSEE; 2641 error = pget((pid_t)name[0], flags, &p); 2642 if (error != 0) 2643 return (error); 2644 2645 /* 2646 * Retrieve limit. 2647 */ 2648 if (req->oldptr != NULL) { 2649 PROC_LOCK(p); 2650 lim_rlimit_proc(p, which, &rlim); 2651 PROC_UNLOCK(p); 2652 } 2653 error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); 2654 if (error != 0) 2655 goto errout; 2656 2657 /* 2658 * Set limit. 2659 */ 2660 if (req->newptr != NULL) { 2661 error = SYSCTL_IN(req, &rlim, sizeof(rlim)); 2662 if (error == 0) 2663 error = kern_proc_setrlimit(curthread, p, which, &rlim); 2664 } 2665 2666 errout: 2667 PRELE(p); 2668 return (error); 2669 } 2670 2671 /* 2672 * This sysctl allows a process to retrieve ps_strings structure location of 2673 * another process. 2674 */ 2675 static int 2676 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) 2677 { 2678 int *name = (int *)arg1; 2679 u_int namelen = arg2; 2680 struct proc *p; 2681 vm_offset_t ps_strings; 2682 int error; 2683 #ifdef COMPAT_FREEBSD32 2684 uint32_t ps_strings32; 2685 #endif 2686 2687 if (namelen != 1) 2688 return (EINVAL); 2689 2690 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2691 if (error != 0) 2692 return (error); 2693 #ifdef COMPAT_FREEBSD32 2694 if ((req->flags & SCTL_MASK32) != 0) { 2695 /* 2696 * We return 0 if the 32 bit emulation request is for a 64 bit 2697 * process. 2698 */ 2699 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? 2700 PTROUT(p->p_sysent->sv_psstrings) : 0; 2701 PROC_UNLOCK(p); 2702 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); 2703 return (error); 2704 } 2705 #endif 2706 ps_strings = p->p_sysent->sv_psstrings; 2707 PROC_UNLOCK(p); 2708 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); 2709 return (error); 2710 } 2711 2712 /* 2713 * This sysctl allows a process to retrieve umask of another process. 2714 */ 2715 static int 2716 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) 2717 { 2718 int *name = (int *)arg1; 2719 u_int namelen = arg2; 2720 struct proc *p; 2721 int error; 2722 u_short fd_cmask; 2723 2724 if (namelen != 1) 2725 return (EINVAL); 2726 2727 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2728 if (error != 0) 2729 return (error); 2730 2731 FILEDESC_SLOCK(p->p_fd); 2732 fd_cmask = p->p_fd->fd_cmask; 2733 FILEDESC_SUNLOCK(p->p_fd); 2734 PRELE(p); 2735 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask)); 2736 return (error); 2737 } 2738 2739 /* 2740 * This sysctl allows a process to set and retrieve binary osreldate of 2741 * another process. 2742 */ 2743 static int 2744 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) 2745 { 2746 int *name = (int *)arg1; 2747 u_int namelen = arg2; 2748 struct proc *p; 2749 int flags, error, osrel; 2750 2751 if (namelen != 1) 2752 return (EINVAL); 2753 2754 if (req->newptr != NULL && req->newlen != sizeof(osrel)) 2755 return (EINVAL); 2756 2757 flags = PGET_HOLD | PGET_NOTWEXIT; 2758 if (req->newptr != NULL) 2759 flags |= PGET_CANDEBUG; 2760 else 2761 flags |= PGET_CANSEE; 2762 error = pget((pid_t)name[0], flags, &p); 2763 if (error != 0) 2764 return (error); 2765 2766 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); 2767 if (error != 0) 2768 goto errout; 2769 2770 if (req->newptr != NULL) { 2771 error = SYSCTL_IN(req, &osrel, sizeof(osrel)); 2772 if (error != 0) 2773 goto errout; 2774 if (osrel < 0) { 2775 error = EINVAL; 2776 goto errout; 2777 } 2778 p->p_osrel = osrel; 2779 } 2780 errout: 2781 PRELE(p); 2782 return (error); 2783 } 2784 2785 static int 2786 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) 2787 { 2788 int *name = (int *)arg1; 2789 u_int namelen = arg2; 2790 struct proc *p; 2791 struct kinfo_sigtramp kst; 2792 const struct sysentvec *sv; 2793 int error; 2794 #ifdef COMPAT_FREEBSD32 2795 struct kinfo_sigtramp32 kst32; 2796 #endif 2797 2798 if (namelen != 1) 2799 return (EINVAL); 2800 2801 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2802 if (error != 0) 2803 return (error); 2804 sv = p->p_sysent; 2805 #ifdef COMPAT_FREEBSD32 2806 if ((req->flags & SCTL_MASK32) != 0) { 2807 bzero(&kst32, sizeof(kst32)); 2808 if (SV_PROC_FLAG(p, SV_ILP32)) { 2809 if (sv->sv_sigcode_base != 0) { 2810 kst32.ksigtramp_start = sv->sv_sigcode_base; 2811 kst32.ksigtramp_end = sv->sv_sigcode_base + 2812 *sv->sv_szsigcode; 2813 } else { 2814 kst32.ksigtramp_start = sv->sv_psstrings - 2815 *sv->sv_szsigcode; 2816 kst32.ksigtramp_end = sv->sv_psstrings; 2817 } 2818 } 2819 PROC_UNLOCK(p); 2820 error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); 2821 return (error); 2822 } 2823 #endif 2824 bzero(&kst, sizeof(kst)); 2825 if (sv->sv_sigcode_base != 0) { 2826 kst.ksigtramp_start = (char *)sv->sv_sigcode_base; 2827 kst.ksigtramp_end = (char *)sv->sv_sigcode_base + 2828 *sv->sv_szsigcode; 2829 } else { 2830 kst.ksigtramp_start = (char *)sv->sv_psstrings - 2831 *sv->sv_szsigcode; 2832 kst.ksigtramp_end = (char *)sv->sv_psstrings; 2833 } 2834 PROC_UNLOCK(p); 2835 error = SYSCTL_OUT(req, &kst, sizeof(kst)); 2836 return (error); 2837 } 2838 2839 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table"); 2840 2841 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| 2842 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", 2843 "Return entire process table"); 2844 2845 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2846 sysctl_kern_proc, "Process table"); 2847 2848 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, 2849 sysctl_kern_proc, "Process table"); 2850 2851 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2852 sysctl_kern_proc, "Process table"); 2853 2854 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | 2855 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2856 2857 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 2858 sysctl_kern_proc, "Process table"); 2859 2860 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2861 sysctl_kern_proc, "Process table"); 2862 2863 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2864 sysctl_kern_proc, "Process table"); 2865 2866 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2867 sysctl_kern_proc, "Process table"); 2868 2869 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 2870 sysctl_kern_proc, "Return process table, no threads"); 2871 2872 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, 2873 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, 2874 sysctl_kern_proc_args, "Process argument list"); 2875 2876 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, 2877 sysctl_kern_proc_env, "Process environment"); 2878 2879 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | 2880 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); 2881 2882 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | 2883 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); 2884 2885 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | 2886 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, 2887 "Process syscall vector name (ABI type)"); 2888 2889 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, 2890 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2891 2892 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, 2893 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2894 2895 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, 2896 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2897 2898 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), 2899 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2900 2901 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, 2902 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2903 2904 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, 2905 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2906 2907 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, 2908 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2909 2910 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, 2911 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2912 2913 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, 2914 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, 2915 "Return process table, no threads"); 2916 2917 #ifdef COMPAT_FREEBSD7 2918 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | 2919 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); 2920 #endif 2921 2922 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | 2923 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); 2924 2925 #if defined(STACK) || defined(DDB) 2926 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | 2927 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); 2928 #endif 2929 2930 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | 2931 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); 2932 2933 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | 2934 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, 2935 "Process resource limits"); 2936 2937 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | 2938 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, 2939 "Process ps_strings location"); 2940 2941 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | 2942 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); 2943 2944 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | 2945 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, 2946 "Process binary osreldate"); 2947 2948 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | 2949 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, 2950 "Process signal trampoline location"); 2951 2952 int allproc_gen; 2953 2954 void 2955 stop_all_proc(void) 2956 { 2957 struct proc *cp, *p; 2958 int r, gen; 2959 bool restart, seen_stopped, seen_exiting, stopped_some; 2960 2961 cp = curproc; 2962 /* 2963 * stop_all_proc() assumes that all process which have 2964 * usermode must be stopped, except current process, for 2965 * obvious reasons. Since other threads in the process 2966 * establishing global stop could unstop something, disable 2967 * calls from multithreaded processes as precaution. The 2968 * service must not be user-callable anyway. 2969 */ 2970 KASSERT((cp->p_flag & P_HADTHREADS) == 0 || 2971 (cp->p_flag & P_KTHREAD) != 0, ("mt stop_all_proc")); 2972 2973 allproc_loop: 2974 sx_xlock(&allproc_lock); 2975 gen = allproc_gen; 2976 seen_exiting = seen_stopped = stopped_some = restart = false; 2977 LIST_REMOVE(cp, p_list); 2978 LIST_INSERT_HEAD(&allproc, cp, p_list); 2979 for (;;) { 2980 p = LIST_NEXT(cp, p_list); 2981 if (p == NULL) 2982 break; 2983 LIST_REMOVE(cp, p_list); 2984 LIST_INSERT_AFTER(p, cp, p_list); 2985 PROC_LOCK(p); 2986 if ((p->p_flag & (P_KTHREAD | P_SYSTEM | 2987 P_TOTAL_STOP)) != 0) { 2988 PROC_UNLOCK(p); 2989 continue; 2990 } 2991 if ((p->p_flag & P_WEXIT) != 0) { 2992 seen_exiting = true; 2993 PROC_UNLOCK(p); 2994 continue; 2995 } 2996 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 2997 /* 2998 * Stopped processes are tolerated when there 2999 * are no other processes which might continue 3000 * them. P_STOPPED_SINGLE but not 3001 * P_TOTAL_STOP process still has at least one 3002 * thread running. 3003 */ 3004 seen_stopped = true; 3005 PROC_UNLOCK(p); 3006 continue; 3007 } 3008 _PHOLD(p); 3009 sx_xunlock(&allproc_lock); 3010 r = thread_single(p, SINGLE_ALLPROC); 3011 if (r != 0) 3012 restart = true; 3013 else 3014 stopped_some = true; 3015 _PRELE(p); 3016 PROC_UNLOCK(p); 3017 sx_xlock(&allproc_lock); 3018 } 3019 /* Catch forked children we did not see in iteration. */ 3020 if (gen != allproc_gen) 3021 restart = true; 3022 sx_xunlock(&allproc_lock); 3023 if (restart || stopped_some || seen_exiting || seen_stopped) { 3024 kern_yield(PRI_USER); 3025 goto allproc_loop; 3026 } 3027 } 3028 3029 void 3030 resume_all_proc(void) 3031 { 3032 struct proc *cp, *p; 3033 3034 cp = curproc; 3035 sx_xlock(&allproc_lock); 3036 LIST_REMOVE(cp, p_list); 3037 LIST_INSERT_HEAD(&allproc, cp, p_list); 3038 for (;;) { 3039 p = LIST_NEXT(cp, p_list); 3040 if (p == NULL) 3041 break; 3042 LIST_REMOVE(cp, p_list); 3043 LIST_INSERT_AFTER(p, cp, p_list); 3044 PROC_LOCK(p); 3045 if ((p->p_flag & P_TOTAL_STOP) != 0) { 3046 sx_xunlock(&allproc_lock); 3047 _PHOLD(p); 3048 thread_single_end(p, SINGLE_ALLPROC); 3049 _PRELE(p); 3050 PROC_UNLOCK(p); 3051 sx_xlock(&allproc_lock); 3052 } else { 3053 PROC_UNLOCK(p); 3054 } 3055 } 3056 sx_xunlock(&allproc_lock); 3057 } 3058 3059 #define TOTAL_STOP_DEBUG 1 3060 #ifdef TOTAL_STOP_DEBUG 3061 volatile static int ap_resume; 3062 #include <sys/mount.h> 3063 3064 static int 3065 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) 3066 { 3067 int error, val; 3068 3069 val = 0; 3070 ap_resume = 0; 3071 error = sysctl_handle_int(oidp, &val, 0, req); 3072 if (error != 0 || req->newptr == NULL) 3073 return (error); 3074 if (val != 0) { 3075 stop_all_proc(); 3076 syncer_suspend(); 3077 while (ap_resume == 0) 3078 ; 3079 syncer_resume(); 3080 resume_all_proc(); 3081 } 3082 return (0); 3083 } 3084 3085 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | 3086 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, 3087 sysctl_debug_stop_all_proc, "I", 3088 ""); 3089 #endif 3090