1 /*- 2 * Copyright (c) 1982, 1986, 1989, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. 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, , ctor, entry, "struct proc *", "int", "void *", 96 "int"); 97 SDT_PROBE_DEFINE4(proc, , ctor, return, "struct proc *", "int", "void *", 98 "int"); 99 SDT_PROBE_DEFINE4(proc, , dtor, entry, "struct proc *", "int", "void *", 100 "struct thread *"); 101 SDT_PROBE_DEFINE3(proc, , dtor, return, "struct proc *", "int", "void *"); 102 SDT_PROBE_DEFINE3(proc, , init, entry, "struct proc *", "int", "int"); 103 SDT_PROBE_DEFINE3(proc, , init, return, "struct proc *", "int", "int"); 104 105 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header"); 106 MALLOC_DEFINE(M_SESSION, "session", "session header"); 107 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); 108 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 109 110 static void doenterpgrp(struct proc *, struct pgrp *); 111 static void orphanpg(struct pgrp *pg); 112 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp); 113 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp); 114 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, 115 int preferthread); 116 static void pgadjustjobc(struct pgrp *pgrp, int entering); 117 static void pgdelete(struct pgrp *); 118 static int proc_ctor(void *mem, int size, void *arg, int flags); 119 static void proc_dtor(void *mem, int size, void *arg); 120 static int proc_init(void *mem, int size, int flags); 121 static void proc_fini(void *mem, int size); 122 static void pargs_free(struct pargs *pa); 123 static struct proc *zpfind_locked(pid_t pid); 124 125 /* 126 * Other process lists 127 */ 128 struct pidhashhead *pidhashtbl; 129 u_long pidhash; 130 struct pgrphashhead *pgrphashtbl; 131 u_long pgrphash; 132 struct proclist allproc; 133 struct proclist zombproc; 134 struct sx allproc_lock; 135 struct sx proctree_lock; 136 struct mtx ppeers_lock; 137 uma_zone_t proc_zone; 138 139 /* 140 * The offset of various fields in struct proc and struct thread. 141 * These are used by kernel debuggers to enumerate kernel threads and 142 * processes. 143 */ 144 const int proc_off_p_pid = offsetof(struct proc, p_pid); 145 const int proc_off_p_comm = offsetof(struct proc, p_comm); 146 const int proc_off_p_list = offsetof(struct proc, p_list); 147 const int proc_off_p_threads = offsetof(struct proc, p_threads); 148 const int thread_off_td_tid = offsetof(struct thread, td_tid); 149 const int thread_off_td_name = offsetof(struct thread, td_name); 150 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu); 151 const int thread_off_td_pcb = offsetof(struct thread, td_pcb); 152 const int thread_off_td_plist = offsetof(struct thread, td_plist); 153 154 int kstack_pages = KSTACK_PAGES; 155 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, 156 "Kernel stack size in pages"); 157 static int vmmap_skip_res_cnt = 0; 158 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW, 159 &vmmap_skip_res_cnt, 0, 160 "Skip calculation of the pages resident count in kern.proc.vmmap"); 161 162 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); 163 #ifdef COMPAT_FREEBSD32 164 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE); 165 #endif 166 167 /* 168 * Initialize global process hashing structures. 169 */ 170 void 171 procinit(void) 172 { 173 174 sx_init(&allproc_lock, "allproc"); 175 sx_init(&proctree_lock, "proctree"); 176 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF); 177 LIST_INIT(&allproc); 178 LIST_INIT(&zombproc); 179 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); 180 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); 181 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(), 182 proc_ctor, proc_dtor, proc_init, proc_fini, 183 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 184 uihashinit(); 185 } 186 187 /* 188 * Prepare a proc for use. 189 */ 190 static int 191 proc_ctor(void *mem, int size, void *arg, int flags) 192 { 193 struct proc *p; 194 struct thread *td; 195 196 p = (struct proc *)mem; 197 SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags); 198 EVENTHANDLER_INVOKE(process_ctor, p); 199 SDT_PROBE4(proc, , ctor , return, p, size, arg, flags); 200 td = FIRST_THREAD_IN_PROC(p); 201 if (td != NULL) { 202 /* Make sure all thread constructors are executed */ 203 EVENTHANDLER_INVOKE(thread_ctor, td); 204 } 205 return (0); 206 } 207 208 /* 209 * Reclaim a proc after use. 210 */ 211 static void 212 proc_dtor(void *mem, int size, void *arg) 213 { 214 struct proc *p; 215 struct thread *td; 216 217 /* INVARIANTS checks go here */ 218 p = (struct proc *)mem; 219 td = FIRST_THREAD_IN_PROC(p); 220 SDT_PROBE4(proc, , dtor, entry, p, size, arg, td); 221 if (td != NULL) { 222 #ifdef INVARIANTS 223 KASSERT((p->p_numthreads == 1), 224 ("bad number of threads in exiting process")); 225 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr")); 226 #endif 227 /* Free all OSD associated to this thread. */ 228 osd_thread_exit(td); 229 td_softdep_cleanup(td); 230 MPASS(td->td_su == NULL); 231 232 /* Make sure all thread destructors are executed */ 233 EVENTHANDLER_INVOKE(thread_dtor, td); 234 } 235 EVENTHANDLER_INVOKE(process_dtor, p); 236 if (p->p_ksi != NULL) 237 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue")); 238 SDT_PROBE3(proc, , dtor, return, p, size, arg); 239 } 240 241 /* 242 * Initialize type-stable parts of a proc (when newly created). 243 */ 244 static int 245 proc_init(void *mem, int size, int flags) 246 { 247 struct proc *p; 248 249 p = (struct proc *)mem; 250 SDT_PROBE3(proc, , init, entry, p, size, flags); 251 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW); 252 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW); 253 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW); 254 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW); 255 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW); 256 cv_init(&p->p_pwait, "ppwait"); 257 cv_init(&p->p_dbgwait, "dbgwait"); 258 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 259 EVENTHANDLER_INVOKE(process_init, p); 260 p->p_stats = pstats_alloc(); 261 p->p_pgrp = NULL; 262 SDT_PROBE3(proc, , init, return, p, size, flags); 263 return (0); 264 } 265 266 /* 267 * UMA should ensure that this function is never called. 268 * Freeing a proc structure would violate type stability. 269 */ 270 static void 271 proc_fini(void *mem, int size) 272 { 273 #ifdef notnow 274 struct proc *p; 275 276 p = (struct proc *)mem; 277 EVENTHANDLER_INVOKE(process_fini, p); 278 pstats_free(p->p_stats); 279 thread_free(FIRST_THREAD_IN_PROC(p)); 280 mtx_destroy(&p->p_mtx); 281 if (p->p_ksi != NULL) 282 ksiginfo_free(p->p_ksi); 283 #else 284 panic("proc reclaimed"); 285 #endif 286 } 287 288 /* 289 * Is p an inferior of the current process? 290 */ 291 int 292 inferior(struct proc *p) 293 { 294 295 sx_assert(&proctree_lock, SX_LOCKED); 296 PROC_LOCK_ASSERT(p, MA_OWNED); 297 for (; p != curproc; p = proc_realparent(p)) { 298 if (p->p_pid == 0) 299 return (0); 300 } 301 return (1); 302 } 303 304 struct proc * 305 pfind_locked(pid_t pid) 306 { 307 struct proc *p; 308 309 sx_assert(&allproc_lock, SX_LOCKED); 310 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 311 if (p->p_pid == pid) { 312 PROC_LOCK(p); 313 if (p->p_state == PRS_NEW) { 314 PROC_UNLOCK(p); 315 p = NULL; 316 } 317 break; 318 } 319 } 320 return (p); 321 } 322 323 /* 324 * Locate a process by number; return only "live" processes -- i.e., neither 325 * zombies nor newly born but incompletely initialized processes. By not 326 * returning processes in the PRS_NEW state, we allow callers to avoid 327 * testing for that condition to avoid dereferencing p_ucred, et al. 328 */ 329 struct proc * 330 pfind(pid_t pid) 331 { 332 struct proc *p; 333 334 sx_slock(&allproc_lock); 335 p = pfind_locked(pid); 336 sx_sunlock(&allproc_lock); 337 return (p); 338 } 339 340 static struct proc * 341 pfind_tid_locked(pid_t tid) 342 { 343 struct proc *p; 344 struct thread *td; 345 346 sx_assert(&allproc_lock, SX_LOCKED); 347 FOREACH_PROC_IN_SYSTEM(p) { 348 PROC_LOCK(p); 349 if (p->p_state == PRS_NEW) { 350 PROC_UNLOCK(p); 351 continue; 352 } 353 FOREACH_THREAD_IN_PROC(p, td) { 354 if (td->td_tid == tid) 355 goto found; 356 } 357 PROC_UNLOCK(p); 358 } 359 found: 360 return (p); 361 } 362 363 /* 364 * Locate a process group by number. 365 * The caller must hold proctree_lock. 366 */ 367 struct pgrp * 368 pgfind(pid_t pgid) 369 { 370 struct pgrp *pgrp; 371 372 sx_assert(&proctree_lock, SX_LOCKED); 373 374 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { 375 if (pgrp->pg_id == pgid) { 376 PGRP_LOCK(pgrp); 377 return (pgrp); 378 } 379 } 380 return (NULL); 381 } 382 383 /* 384 * Locate process and do additional manipulations, depending on flags. 385 */ 386 int 387 pget(pid_t pid, int flags, struct proc **pp) 388 { 389 struct proc *p; 390 int error; 391 392 sx_slock(&allproc_lock); 393 if (pid <= PID_MAX) { 394 p = pfind_locked(pid); 395 if (p == NULL && (flags & PGET_NOTWEXIT) == 0) 396 p = zpfind_locked(pid); 397 } else if ((flags & PGET_NOTID) == 0) { 398 p = pfind_tid_locked(pid); 399 } else { 400 p = NULL; 401 } 402 sx_sunlock(&allproc_lock); 403 if (p == NULL) 404 return (ESRCH); 405 if ((flags & PGET_CANSEE) != 0) { 406 error = p_cansee(curthread, p); 407 if (error != 0) 408 goto errout; 409 } 410 if ((flags & PGET_CANDEBUG) != 0) { 411 error = p_candebug(curthread, p); 412 if (error != 0) 413 goto errout; 414 } 415 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) { 416 error = EPERM; 417 goto errout; 418 } 419 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) { 420 error = ESRCH; 421 goto errout; 422 } 423 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) { 424 /* 425 * XXXRW: Not clear ESRCH is the right error during proc 426 * execve(). 427 */ 428 error = ESRCH; 429 goto errout; 430 } 431 if ((flags & PGET_HOLD) != 0) { 432 _PHOLD(p); 433 PROC_UNLOCK(p); 434 } 435 *pp = p; 436 return (0); 437 errout: 438 PROC_UNLOCK(p); 439 return (error); 440 } 441 442 /* 443 * Create a new process group. 444 * pgid must be equal to the pid of p. 445 * Begin a new session if required. 446 */ 447 int 448 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess) 449 { 450 451 sx_assert(&proctree_lock, SX_XLOCKED); 452 453 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL")); 454 KASSERT(p->p_pid == pgid, 455 ("enterpgrp: new pgrp and pid != pgid")); 456 KASSERT(pgfind(pgid) == NULL, 457 ("enterpgrp: pgrp with pgid exists")); 458 KASSERT(!SESS_LEADER(p), 459 ("enterpgrp: session leader attempted setpgrp")); 460 461 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 462 463 if (sess != NULL) { 464 /* 465 * new session 466 */ 467 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); 468 PROC_LOCK(p); 469 p->p_flag &= ~P_CONTROLT; 470 PROC_UNLOCK(p); 471 PGRP_LOCK(pgrp); 472 sess->s_leader = p; 473 sess->s_sid = p->p_pid; 474 refcount_init(&sess->s_count, 1); 475 sess->s_ttyvp = NULL; 476 sess->s_ttydp = NULL; 477 sess->s_ttyp = NULL; 478 bcopy(p->p_session->s_login, sess->s_login, 479 sizeof(sess->s_login)); 480 pgrp->pg_session = sess; 481 KASSERT(p == curproc, 482 ("enterpgrp: mksession and p != curproc")); 483 } else { 484 pgrp->pg_session = p->p_session; 485 sess_hold(pgrp->pg_session); 486 PGRP_LOCK(pgrp); 487 } 488 pgrp->pg_id = pgid; 489 LIST_INIT(&pgrp->pg_members); 490 491 /* 492 * As we have an exclusive lock of proctree_lock, 493 * this should not deadlock. 494 */ 495 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); 496 pgrp->pg_jobc = 0; 497 SLIST_INIT(&pgrp->pg_sigiolst); 498 PGRP_UNLOCK(pgrp); 499 500 doenterpgrp(p, pgrp); 501 502 return (0); 503 } 504 505 /* 506 * Move p to an existing process group 507 */ 508 int 509 enterthispgrp(struct proc *p, struct pgrp *pgrp) 510 { 511 512 sx_assert(&proctree_lock, SX_XLOCKED); 513 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 514 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 515 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 516 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 517 KASSERT(pgrp->pg_session == p->p_session, 518 ("%s: pgrp's session %p, p->p_session %p.\n", 519 __func__, 520 pgrp->pg_session, 521 p->p_session)); 522 KASSERT(pgrp != p->p_pgrp, 523 ("%s: p belongs to pgrp.", __func__)); 524 525 doenterpgrp(p, pgrp); 526 527 return (0); 528 } 529 530 /* 531 * Move p to a process group 532 */ 533 static void 534 doenterpgrp(struct proc *p, struct pgrp *pgrp) 535 { 536 struct pgrp *savepgrp; 537 538 sx_assert(&proctree_lock, SX_XLOCKED); 539 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 540 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 541 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 542 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 543 544 savepgrp = p->p_pgrp; 545 546 /* 547 * Adjust eligibility of affected pgrps to participate in job control. 548 * Increment eligibility counts before decrementing, otherwise we 549 * could reach 0 spuriously during the first call. 550 */ 551 fixjobc(p, pgrp, 1); 552 fixjobc(p, p->p_pgrp, 0); 553 554 PGRP_LOCK(pgrp); 555 PGRP_LOCK(savepgrp); 556 PROC_LOCK(p); 557 LIST_REMOVE(p, p_pglist); 558 p->p_pgrp = pgrp; 559 PROC_UNLOCK(p); 560 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 561 PGRP_UNLOCK(savepgrp); 562 PGRP_UNLOCK(pgrp); 563 if (LIST_EMPTY(&savepgrp->pg_members)) 564 pgdelete(savepgrp); 565 } 566 567 /* 568 * remove process from process group 569 */ 570 int 571 leavepgrp(struct proc *p) 572 { 573 struct pgrp *savepgrp; 574 575 sx_assert(&proctree_lock, SX_XLOCKED); 576 savepgrp = p->p_pgrp; 577 PGRP_LOCK(savepgrp); 578 PROC_LOCK(p); 579 LIST_REMOVE(p, p_pglist); 580 p->p_pgrp = NULL; 581 PROC_UNLOCK(p); 582 PGRP_UNLOCK(savepgrp); 583 if (LIST_EMPTY(&savepgrp->pg_members)) 584 pgdelete(savepgrp); 585 return (0); 586 } 587 588 /* 589 * delete a process group 590 */ 591 static void 592 pgdelete(struct pgrp *pgrp) 593 { 594 struct session *savesess; 595 struct tty *tp; 596 597 sx_assert(&proctree_lock, SX_XLOCKED); 598 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 599 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 600 601 /* 602 * Reset any sigio structures pointing to us as a result of 603 * F_SETOWN with our pgid. 604 */ 605 funsetownlst(&pgrp->pg_sigiolst); 606 607 PGRP_LOCK(pgrp); 608 tp = pgrp->pg_session->s_ttyp; 609 LIST_REMOVE(pgrp, pg_hash); 610 savesess = pgrp->pg_session; 611 PGRP_UNLOCK(pgrp); 612 613 /* Remove the reference to the pgrp before deallocating it. */ 614 if (tp != NULL) { 615 tty_lock(tp); 616 tty_rel_pgrp(tp, pgrp); 617 } 618 619 mtx_destroy(&pgrp->pg_mtx); 620 free(pgrp, M_PGRP); 621 sess_release(savesess); 622 } 623 624 static void 625 pgadjustjobc(struct pgrp *pgrp, int entering) 626 { 627 628 PGRP_LOCK(pgrp); 629 if (entering) 630 pgrp->pg_jobc++; 631 else { 632 --pgrp->pg_jobc; 633 if (pgrp->pg_jobc == 0) 634 orphanpg(pgrp); 635 } 636 PGRP_UNLOCK(pgrp); 637 } 638 639 /* 640 * Adjust pgrp jobc counters when specified process changes process group. 641 * We count the number of processes in each process group that "qualify" 642 * the group for terminal job control (those with a parent in a different 643 * process group of the same session). If that count reaches zero, the 644 * process group becomes orphaned. Check both the specified process' 645 * process group and that of its children. 646 * entering == 0 => p is leaving specified group. 647 * entering == 1 => p is entering specified group. 648 */ 649 void 650 fixjobc(struct proc *p, struct pgrp *pgrp, int entering) 651 { 652 struct pgrp *hispgrp; 653 struct session *mysession; 654 struct proc *q; 655 656 sx_assert(&proctree_lock, SX_LOCKED); 657 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 658 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 659 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 660 661 /* 662 * Check p's parent to see whether p qualifies its own process 663 * group; if so, adjust count for p's process group. 664 */ 665 mysession = pgrp->pg_session; 666 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp && 667 hispgrp->pg_session == mysession) 668 pgadjustjobc(pgrp, entering); 669 670 /* 671 * Check this process' children to see whether they qualify 672 * their process groups; if so, adjust counts for children's 673 * process groups. 674 */ 675 LIST_FOREACH(q, &p->p_children, p_sibling) { 676 hispgrp = q->p_pgrp; 677 if (hispgrp == pgrp || 678 hispgrp->pg_session != mysession) 679 continue; 680 if (q->p_state == PRS_ZOMBIE) 681 continue; 682 pgadjustjobc(hispgrp, entering); 683 } 684 } 685 686 void 687 killjobc(void) 688 { 689 struct session *sp; 690 struct tty *tp; 691 struct proc *p; 692 struct vnode *ttyvp; 693 694 p = curproc; 695 MPASS(p->p_flag & P_WEXIT); 696 /* 697 * Do a quick check to see if there is anything to do with the 698 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc(). 699 */ 700 PROC_LOCK(p); 701 if (!SESS_LEADER(p) && 702 (p->p_pgrp == p->p_pptr->p_pgrp) && 703 LIST_EMPTY(&p->p_children)) { 704 PROC_UNLOCK(p); 705 return; 706 } 707 PROC_UNLOCK(p); 708 709 sx_xlock(&proctree_lock); 710 if (SESS_LEADER(p)) { 711 sp = p->p_session; 712 713 /* 714 * s_ttyp is not zero'd; we use this to indicate that 715 * the session once had a controlling terminal. (for 716 * logging and informational purposes) 717 */ 718 SESS_LOCK(sp); 719 ttyvp = sp->s_ttyvp; 720 tp = sp->s_ttyp; 721 sp->s_ttyvp = NULL; 722 sp->s_ttydp = NULL; 723 sp->s_leader = NULL; 724 SESS_UNLOCK(sp); 725 726 /* 727 * Signal foreground pgrp and revoke access to 728 * controlling terminal if it has not been revoked 729 * already. 730 * 731 * Because the TTY may have been revoked in the mean 732 * time and could already have a new session associated 733 * with it, make sure we don't send a SIGHUP to a 734 * foreground process group that does not belong to this 735 * session. 736 */ 737 738 if (tp != NULL) { 739 tty_lock(tp); 740 if (tp->t_session == sp) 741 tty_signal_pgrp(tp, SIGHUP); 742 tty_unlock(tp); 743 } 744 745 if (ttyvp != NULL) { 746 sx_xunlock(&proctree_lock); 747 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) { 748 VOP_REVOKE(ttyvp, REVOKEALL); 749 VOP_UNLOCK(ttyvp, 0); 750 } 751 vrele(ttyvp); 752 sx_xlock(&proctree_lock); 753 } 754 } 755 fixjobc(p, p->p_pgrp, 0); 756 sx_xunlock(&proctree_lock); 757 } 758 759 /* 760 * A process group has become orphaned; 761 * if there are any stopped processes in the group, 762 * hang-up all process in that group. 763 */ 764 static void 765 orphanpg(struct pgrp *pg) 766 { 767 struct proc *p; 768 769 PGRP_LOCK_ASSERT(pg, MA_OWNED); 770 771 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 772 PROC_LOCK(p); 773 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) { 774 PROC_UNLOCK(p); 775 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 776 PROC_LOCK(p); 777 kern_psignal(p, SIGHUP); 778 kern_psignal(p, SIGCONT); 779 PROC_UNLOCK(p); 780 } 781 return; 782 } 783 PROC_UNLOCK(p); 784 } 785 } 786 787 void 788 sess_hold(struct session *s) 789 { 790 791 refcount_acquire(&s->s_count); 792 } 793 794 void 795 sess_release(struct session *s) 796 { 797 798 if (refcount_release(&s->s_count)) { 799 if (s->s_ttyp != NULL) { 800 tty_lock(s->s_ttyp); 801 tty_rel_sess(s->s_ttyp, s); 802 } 803 mtx_destroy(&s->s_mtx); 804 free(s, M_SESSION); 805 } 806 } 807 808 #ifdef DDB 809 810 DB_SHOW_COMMAND(pgrpdump, pgrpdump) 811 { 812 struct pgrp *pgrp; 813 struct proc *p; 814 int i; 815 816 for (i = 0; i <= pgrphash; i++) { 817 if (!LIST_EMPTY(&pgrphashtbl[i])) { 818 printf("\tindx %d\n", i); 819 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { 820 printf( 821 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n", 822 (void *)pgrp, (long)pgrp->pg_id, 823 (void *)pgrp->pg_session, 824 pgrp->pg_session->s_count, 825 (void *)LIST_FIRST(&pgrp->pg_members)); 826 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 827 printf("\t\tpid %ld addr %p pgrp %p\n", 828 (long)p->p_pid, (void *)p, 829 (void *)p->p_pgrp); 830 } 831 } 832 } 833 } 834 } 835 #endif /* DDB */ 836 837 /* 838 * Calculate the kinfo_proc members which contain process-wide 839 * informations. 840 * Must be called with the target process locked. 841 */ 842 static void 843 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp) 844 { 845 struct thread *td; 846 847 PROC_LOCK_ASSERT(p, MA_OWNED); 848 849 kp->ki_estcpu = 0; 850 kp->ki_pctcpu = 0; 851 FOREACH_THREAD_IN_PROC(p, td) { 852 thread_lock(td); 853 kp->ki_pctcpu += sched_pctcpu(td); 854 kp->ki_estcpu += sched_estcpu(td); 855 thread_unlock(td); 856 } 857 } 858 859 /* 860 * Clear kinfo_proc and fill in any information that is common 861 * to all threads in the process. 862 * Must be called with the target process locked. 863 */ 864 static void 865 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp) 866 { 867 struct thread *td0; 868 struct tty *tp; 869 struct session *sp; 870 struct ucred *cred; 871 struct sigacts *ps; 872 struct timeval boottime; 873 874 /* For proc_realparent. */ 875 sx_assert(&proctree_lock, SX_LOCKED); 876 PROC_LOCK_ASSERT(p, MA_OWNED); 877 bzero(kp, sizeof(*kp)); 878 879 kp->ki_structsize = sizeof(*kp); 880 kp->ki_paddr = p; 881 kp->ki_addr =/* p->p_addr; */0; /* XXX */ 882 kp->ki_args = p->p_args; 883 kp->ki_textvp = p->p_textvp; 884 #ifdef KTRACE 885 kp->ki_tracep = p->p_tracevp; 886 kp->ki_traceflag = p->p_traceflag; 887 #endif 888 kp->ki_fd = p->p_fd; 889 kp->ki_vmspace = p->p_vmspace; 890 kp->ki_flag = p->p_flag; 891 kp->ki_flag2 = p->p_flag2; 892 cred = p->p_ucred; 893 if (cred) { 894 kp->ki_uid = cred->cr_uid; 895 kp->ki_ruid = cred->cr_ruid; 896 kp->ki_svuid = cred->cr_svuid; 897 kp->ki_cr_flags = 0; 898 if (cred->cr_flags & CRED_FLAG_CAPMODE) 899 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE; 900 /* XXX bde doesn't like KI_NGROUPS */ 901 if (cred->cr_ngroups > KI_NGROUPS) { 902 kp->ki_ngroups = KI_NGROUPS; 903 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW; 904 } else 905 kp->ki_ngroups = cred->cr_ngroups; 906 bcopy(cred->cr_groups, kp->ki_groups, 907 kp->ki_ngroups * sizeof(gid_t)); 908 kp->ki_rgid = cred->cr_rgid; 909 kp->ki_svgid = cred->cr_svgid; 910 /* If jailed(cred), emulate the old P_JAILED flag. */ 911 if (jailed(cred)) { 912 kp->ki_flag |= P_JAILED; 913 /* If inside the jail, use 0 as a jail ID. */ 914 if (cred->cr_prison != curthread->td_ucred->cr_prison) 915 kp->ki_jid = cred->cr_prison->pr_id; 916 } 917 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name, 918 sizeof(kp->ki_loginclass)); 919 } 920 ps = p->p_sigacts; 921 if (ps) { 922 mtx_lock(&ps->ps_mtx); 923 kp->ki_sigignore = ps->ps_sigignore; 924 kp->ki_sigcatch = ps->ps_sigcatch; 925 mtx_unlock(&ps->ps_mtx); 926 } 927 if (p->p_state != PRS_NEW && 928 p->p_state != PRS_ZOMBIE && 929 p->p_vmspace != NULL) { 930 struct vmspace *vm = p->p_vmspace; 931 932 kp->ki_size = vm->vm_map.size; 933 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ 934 FOREACH_THREAD_IN_PROC(p, td0) { 935 if (!TD_IS_SWAPPED(td0)) 936 kp->ki_rssize += td0->td_kstack_pages; 937 } 938 kp->ki_swrss = vm->vm_swrss; 939 kp->ki_tsize = vm->vm_tsize; 940 kp->ki_dsize = vm->vm_dsize; 941 kp->ki_ssize = vm->vm_ssize; 942 } else if (p->p_state == PRS_ZOMBIE) 943 kp->ki_stat = SZOMB; 944 if (kp->ki_flag & P_INMEM) 945 kp->ki_sflag = PS_INMEM; 946 else 947 kp->ki_sflag = 0; 948 /* Calculate legacy swtime as seconds since 'swtick'. */ 949 kp->ki_swtime = (ticks - p->p_swtick) / hz; 950 kp->ki_pid = p->p_pid; 951 kp->ki_nice = p->p_nice; 952 kp->ki_fibnum = p->p_fibnum; 953 kp->ki_start = p->p_stats->p_start; 954 getboottime(&boottime); 955 timevaladd(&kp->ki_start, &boottime); 956 PROC_STATLOCK(p); 957 rufetch(p, &kp->ki_rusage); 958 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime); 959 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime); 960 PROC_STATUNLOCK(p); 961 calccru(p, &kp->ki_childutime, &kp->ki_childstime); 962 /* Some callers want child times in a single value. */ 963 kp->ki_childtime = kp->ki_childstime; 964 timevaladd(&kp->ki_childtime, &kp->ki_childutime); 965 966 FOREACH_THREAD_IN_PROC(p, td0) 967 kp->ki_cow += td0->td_cow; 968 969 tp = NULL; 970 if (p->p_pgrp) { 971 kp->ki_pgid = p->p_pgrp->pg_id; 972 kp->ki_jobc = p->p_pgrp->pg_jobc; 973 sp = p->p_pgrp->pg_session; 974 975 if (sp != NULL) { 976 kp->ki_sid = sp->s_sid; 977 SESS_LOCK(sp); 978 strlcpy(kp->ki_login, sp->s_login, 979 sizeof(kp->ki_login)); 980 if (sp->s_ttyvp) 981 kp->ki_kiflag |= KI_CTTY; 982 if (SESS_LEADER(p)) 983 kp->ki_kiflag |= KI_SLEADER; 984 /* XXX proctree_lock */ 985 tp = sp->s_ttyp; 986 SESS_UNLOCK(sp); 987 } 988 } 989 if ((p->p_flag & P_CONTROLT) && tp != NULL) { 990 kp->ki_tdev = tty_udev(tp); 991 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 992 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; 993 if (tp->t_session) 994 kp->ki_tsid = tp->t_session->s_sid; 995 } else { 996 kp->ki_tdev = NODEV; 997 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 998 } 999 if (p->p_comm[0] != '\0') 1000 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm)); 1001 if (p->p_sysent && p->p_sysent->sv_name != NULL && 1002 p->p_sysent->sv_name[0] != '\0') 1003 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul)); 1004 kp->ki_siglist = p->p_siglist; 1005 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig); 1006 kp->ki_acflag = p->p_acflag; 1007 kp->ki_lock = p->p_lock; 1008 if (p->p_pptr) { 1009 kp->ki_ppid = proc_realparent(p)->p_pid; 1010 if (p->p_flag & P_TRACED) 1011 kp->ki_tracer = p->p_pptr->p_pid; 1012 } 1013 } 1014 1015 /* 1016 * Fill in information that is thread specific. Must be called with 1017 * target process locked. If 'preferthread' is set, overwrite certain 1018 * process-related fields that are maintained for both threads and 1019 * processes. 1020 */ 1021 static void 1022 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread) 1023 { 1024 struct proc *p; 1025 1026 p = td->td_proc; 1027 kp->ki_tdaddr = td; 1028 PROC_LOCK_ASSERT(p, MA_OWNED); 1029 1030 if (preferthread) 1031 PROC_STATLOCK(p); 1032 thread_lock(td); 1033 if (td->td_wmesg != NULL) 1034 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg)); 1035 else 1036 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg)); 1037 if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >= 1038 sizeof(kp->ki_tdname)) { 1039 strlcpy(kp->ki_moretdname, 1040 td->td_name + sizeof(kp->ki_tdname) - 1, 1041 sizeof(kp->ki_moretdname)); 1042 } else { 1043 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname)); 1044 } 1045 if (TD_ON_LOCK(td)) { 1046 kp->ki_kiflag |= KI_LOCKBLOCK; 1047 strlcpy(kp->ki_lockname, td->td_lockname, 1048 sizeof(kp->ki_lockname)); 1049 } else { 1050 kp->ki_kiflag &= ~KI_LOCKBLOCK; 1051 bzero(kp->ki_lockname, sizeof(kp->ki_lockname)); 1052 } 1053 1054 if (p->p_state == PRS_NORMAL) { /* approximate. */ 1055 if (TD_ON_RUNQ(td) || 1056 TD_CAN_RUN(td) || 1057 TD_IS_RUNNING(td)) { 1058 kp->ki_stat = SRUN; 1059 } else if (P_SHOULDSTOP(p)) { 1060 kp->ki_stat = SSTOP; 1061 } else if (TD_IS_SLEEPING(td)) { 1062 kp->ki_stat = SSLEEP; 1063 } else if (TD_ON_LOCK(td)) { 1064 kp->ki_stat = SLOCK; 1065 } else { 1066 kp->ki_stat = SWAIT; 1067 } 1068 } else if (p->p_state == PRS_ZOMBIE) { 1069 kp->ki_stat = SZOMB; 1070 } else { 1071 kp->ki_stat = SIDL; 1072 } 1073 1074 /* Things in the thread */ 1075 kp->ki_wchan = td->td_wchan; 1076 kp->ki_pri.pri_level = td->td_priority; 1077 kp->ki_pri.pri_native = td->td_base_pri; 1078 1079 /* 1080 * Note: legacy fields; clamp at the old NOCPU value and/or 1081 * the maximum u_char CPU value. 1082 */ 1083 if (td->td_lastcpu == NOCPU) 1084 kp->ki_lastcpu_old = NOCPU_OLD; 1085 else if (td->td_lastcpu > MAXCPU_OLD) 1086 kp->ki_lastcpu_old = MAXCPU_OLD; 1087 else 1088 kp->ki_lastcpu_old = td->td_lastcpu; 1089 1090 if (td->td_oncpu == NOCPU) 1091 kp->ki_oncpu_old = NOCPU_OLD; 1092 else if (td->td_oncpu > MAXCPU_OLD) 1093 kp->ki_oncpu_old = MAXCPU_OLD; 1094 else 1095 kp->ki_oncpu_old = td->td_oncpu; 1096 1097 kp->ki_lastcpu = td->td_lastcpu; 1098 kp->ki_oncpu = td->td_oncpu; 1099 kp->ki_tdflags = td->td_flags; 1100 kp->ki_tid = td->td_tid; 1101 kp->ki_numthreads = p->p_numthreads; 1102 kp->ki_pcb = td->td_pcb; 1103 kp->ki_kstack = (void *)td->td_kstack; 1104 kp->ki_slptime = (ticks - td->td_slptick) / hz; 1105 kp->ki_pri.pri_class = td->td_pri_class; 1106 kp->ki_pri.pri_user = td->td_user_pri; 1107 1108 if (preferthread) { 1109 rufetchtd(td, &kp->ki_rusage); 1110 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime); 1111 kp->ki_pctcpu = sched_pctcpu(td); 1112 kp->ki_estcpu = sched_estcpu(td); 1113 kp->ki_cow = td->td_cow; 1114 } 1115 1116 /* We can't get this anymore but ps etc never used it anyway. */ 1117 kp->ki_rqindex = 0; 1118 1119 if (preferthread) 1120 kp->ki_siglist = td->td_siglist; 1121 kp->ki_sigmask = td->td_sigmask; 1122 thread_unlock(td); 1123 if (preferthread) 1124 PROC_STATUNLOCK(p); 1125 } 1126 1127 /* 1128 * Fill in a kinfo_proc structure for the specified process. 1129 * Must be called with the target process locked. 1130 */ 1131 void 1132 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp) 1133 { 1134 1135 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1136 1137 fill_kinfo_proc_only(p, kp); 1138 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0); 1139 fill_kinfo_aggregate(p, kp); 1140 } 1141 1142 struct pstats * 1143 pstats_alloc(void) 1144 { 1145 1146 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK)); 1147 } 1148 1149 /* 1150 * Copy parts of p_stats; zero the rest of p_stats (statistics). 1151 */ 1152 void 1153 pstats_fork(struct pstats *src, struct pstats *dst) 1154 { 1155 1156 bzero(&dst->pstat_startzero, 1157 __rangeof(struct pstats, pstat_startzero, pstat_endzero)); 1158 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy, 1159 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy)); 1160 } 1161 1162 void 1163 pstats_free(struct pstats *ps) 1164 { 1165 1166 free(ps, M_SUBPROC); 1167 } 1168 1169 static struct proc * 1170 zpfind_locked(pid_t pid) 1171 { 1172 struct proc *p; 1173 1174 sx_assert(&allproc_lock, SX_LOCKED); 1175 LIST_FOREACH(p, &zombproc, p_list) { 1176 if (p->p_pid == pid) { 1177 PROC_LOCK(p); 1178 break; 1179 } 1180 } 1181 return (p); 1182 } 1183 1184 /* 1185 * Locate a zombie process by number 1186 */ 1187 struct proc * 1188 zpfind(pid_t pid) 1189 { 1190 struct proc *p; 1191 1192 sx_slock(&allproc_lock); 1193 p = zpfind_locked(pid); 1194 sx_sunlock(&allproc_lock); 1195 return (p); 1196 } 1197 1198 #ifdef COMPAT_FREEBSD32 1199 1200 /* 1201 * This function is typically used to copy out the kernel address, so 1202 * it can be replaced by assignment of zero. 1203 */ 1204 static inline uint32_t 1205 ptr32_trim(void *ptr) 1206 { 1207 uintptr_t uptr; 1208 1209 uptr = (uintptr_t)ptr; 1210 return ((uptr > UINT_MAX) ? 0 : uptr); 1211 } 1212 1213 #define PTRTRIM_CP(src,dst,fld) \ 1214 do { (dst).fld = ptr32_trim((src).fld); } while (0) 1215 1216 static void 1217 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32) 1218 { 1219 int i; 1220 1221 bzero(ki32, sizeof(struct kinfo_proc32)); 1222 ki32->ki_structsize = sizeof(struct kinfo_proc32); 1223 CP(*ki, *ki32, ki_layout); 1224 PTRTRIM_CP(*ki, *ki32, ki_args); 1225 PTRTRIM_CP(*ki, *ki32, ki_paddr); 1226 PTRTRIM_CP(*ki, *ki32, ki_addr); 1227 PTRTRIM_CP(*ki, *ki32, ki_tracep); 1228 PTRTRIM_CP(*ki, *ki32, ki_textvp); 1229 PTRTRIM_CP(*ki, *ki32, ki_fd); 1230 PTRTRIM_CP(*ki, *ki32, ki_vmspace); 1231 PTRTRIM_CP(*ki, *ki32, ki_wchan); 1232 CP(*ki, *ki32, ki_pid); 1233 CP(*ki, *ki32, ki_ppid); 1234 CP(*ki, *ki32, ki_pgid); 1235 CP(*ki, *ki32, ki_tpgid); 1236 CP(*ki, *ki32, ki_sid); 1237 CP(*ki, *ki32, ki_tsid); 1238 CP(*ki, *ki32, ki_jobc); 1239 CP(*ki, *ki32, ki_tdev); 1240 CP(*ki, *ki32, ki_tdev_freebsd11); 1241 CP(*ki, *ki32, ki_siglist); 1242 CP(*ki, *ki32, ki_sigmask); 1243 CP(*ki, *ki32, ki_sigignore); 1244 CP(*ki, *ki32, ki_sigcatch); 1245 CP(*ki, *ki32, ki_uid); 1246 CP(*ki, *ki32, ki_ruid); 1247 CP(*ki, *ki32, ki_svuid); 1248 CP(*ki, *ki32, ki_rgid); 1249 CP(*ki, *ki32, ki_svgid); 1250 CP(*ki, *ki32, ki_ngroups); 1251 for (i = 0; i < KI_NGROUPS; i++) 1252 CP(*ki, *ki32, ki_groups[i]); 1253 CP(*ki, *ki32, ki_size); 1254 CP(*ki, *ki32, ki_rssize); 1255 CP(*ki, *ki32, ki_swrss); 1256 CP(*ki, *ki32, ki_tsize); 1257 CP(*ki, *ki32, ki_dsize); 1258 CP(*ki, *ki32, ki_ssize); 1259 CP(*ki, *ki32, ki_xstat); 1260 CP(*ki, *ki32, ki_acflag); 1261 CP(*ki, *ki32, ki_pctcpu); 1262 CP(*ki, *ki32, ki_estcpu); 1263 CP(*ki, *ki32, ki_slptime); 1264 CP(*ki, *ki32, ki_swtime); 1265 CP(*ki, *ki32, ki_cow); 1266 CP(*ki, *ki32, ki_runtime); 1267 TV_CP(*ki, *ki32, ki_start); 1268 TV_CP(*ki, *ki32, ki_childtime); 1269 CP(*ki, *ki32, ki_flag); 1270 CP(*ki, *ki32, ki_kiflag); 1271 CP(*ki, *ki32, ki_traceflag); 1272 CP(*ki, *ki32, ki_stat); 1273 CP(*ki, *ki32, ki_nice); 1274 CP(*ki, *ki32, ki_lock); 1275 CP(*ki, *ki32, ki_rqindex); 1276 CP(*ki, *ki32, ki_oncpu); 1277 CP(*ki, *ki32, ki_lastcpu); 1278 1279 /* XXX TODO: wrap cpu value as appropriate */ 1280 CP(*ki, *ki32, ki_oncpu_old); 1281 CP(*ki, *ki32, ki_lastcpu_old); 1282 1283 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1); 1284 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1); 1285 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1); 1286 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1); 1287 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1); 1288 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1); 1289 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1); 1290 bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1); 1291 CP(*ki, *ki32, ki_tracer); 1292 CP(*ki, *ki32, ki_flag2); 1293 CP(*ki, *ki32, ki_fibnum); 1294 CP(*ki, *ki32, ki_cr_flags); 1295 CP(*ki, *ki32, ki_jid); 1296 CP(*ki, *ki32, ki_numthreads); 1297 CP(*ki, *ki32, ki_tid); 1298 CP(*ki, *ki32, ki_pri); 1299 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage); 1300 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch); 1301 PTRTRIM_CP(*ki, *ki32, ki_pcb); 1302 PTRTRIM_CP(*ki, *ki32, ki_kstack); 1303 PTRTRIM_CP(*ki, *ki32, ki_udata); 1304 PTRTRIM_CP(*ki, *ki32, ki_tdaddr); 1305 CP(*ki, *ki32, ki_sflag); 1306 CP(*ki, *ki32, ki_tdflags); 1307 } 1308 #endif 1309 1310 int 1311 kern_proc_out(struct proc *p, struct sbuf *sb, int flags) 1312 { 1313 struct thread *td; 1314 struct kinfo_proc ki; 1315 #ifdef COMPAT_FREEBSD32 1316 struct kinfo_proc32 ki32; 1317 #endif 1318 int error; 1319 1320 PROC_LOCK_ASSERT(p, MA_OWNED); 1321 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1322 1323 error = 0; 1324 fill_kinfo_proc(p, &ki); 1325 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1326 #ifdef COMPAT_FREEBSD32 1327 if ((flags & KERN_PROC_MASK32) != 0) { 1328 freebsd32_kinfo_proc_out(&ki, &ki32); 1329 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1330 error = ENOMEM; 1331 } else 1332 #endif 1333 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1334 error = ENOMEM; 1335 } else { 1336 FOREACH_THREAD_IN_PROC(p, td) { 1337 fill_kinfo_thread(td, &ki, 1); 1338 #ifdef COMPAT_FREEBSD32 1339 if ((flags & KERN_PROC_MASK32) != 0) { 1340 freebsd32_kinfo_proc_out(&ki, &ki32); 1341 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1342 error = ENOMEM; 1343 } else 1344 #endif 1345 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1346 error = ENOMEM; 1347 if (error != 0) 1348 break; 1349 } 1350 } 1351 PROC_UNLOCK(p); 1352 return (error); 1353 } 1354 1355 static int 1356 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags, 1357 int doingzomb) 1358 { 1359 struct sbuf sb; 1360 struct kinfo_proc ki; 1361 struct proc *np; 1362 int error, error2; 1363 pid_t pid; 1364 1365 pid = p->p_pid; 1366 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); 1367 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1368 error = kern_proc_out(p, &sb, flags); 1369 error2 = sbuf_finish(&sb); 1370 sbuf_delete(&sb); 1371 if (error != 0) 1372 return (error); 1373 else if (error2 != 0) 1374 return (error2); 1375 if (doingzomb) 1376 np = zpfind(pid); 1377 else { 1378 if (pid == 0) 1379 return (0); 1380 np = pfind(pid); 1381 } 1382 if (np == NULL) 1383 return (ESRCH); 1384 if (np != p) { 1385 PROC_UNLOCK(np); 1386 return (ESRCH); 1387 } 1388 PROC_UNLOCK(np); 1389 return (0); 1390 } 1391 1392 static int 1393 sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 1394 { 1395 int *name = (int *)arg1; 1396 u_int namelen = arg2; 1397 struct proc *p; 1398 int flags, doingzomb, oid_number; 1399 int error = 0; 1400 1401 oid_number = oidp->oid_number; 1402 if (oid_number != KERN_PROC_ALL && 1403 (oid_number & KERN_PROC_INC_THREAD) == 0) 1404 flags = KERN_PROC_NOTHREADS; 1405 else { 1406 flags = 0; 1407 oid_number &= ~KERN_PROC_INC_THREAD; 1408 } 1409 #ifdef COMPAT_FREEBSD32 1410 if (req->flags & SCTL_MASK32) 1411 flags |= KERN_PROC_MASK32; 1412 #endif 1413 if (oid_number == KERN_PROC_PID) { 1414 if (namelen != 1) 1415 return (EINVAL); 1416 error = sysctl_wire_old_buffer(req, 0); 1417 if (error) 1418 return (error); 1419 sx_slock(&proctree_lock); 1420 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1421 if (error == 0) 1422 error = sysctl_out_proc(p, req, flags, 0); 1423 sx_sunlock(&proctree_lock); 1424 return (error); 1425 } 1426 1427 switch (oid_number) { 1428 case KERN_PROC_ALL: 1429 if (namelen != 0) 1430 return (EINVAL); 1431 break; 1432 case KERN_PROC_PROC: 1433 if (namelen != 0 && namelen != 1) 1434 return (EINVAL); 1435 break; 1436 default: 1437 if (namelen != 1) 1438 return (EINVAL); 1439 break; 1440 } 1441 1442 if (!req->oldptr) { 1443 /* overestimate by 5 procs */ 1444 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1445 if (error) 1446 return (error); 1447 } 1448 error = sysctl_wire_old_buffer(req, 0); 1449 if (error != 0) 1450 return (error); 1451 sx_slock(&proctree_lock); 1452 sx_slock(&allproc_lock); 1453 for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) { 1454 if (!doingzomb) 1455 p = LIST_FIRST(&allproc); 1456 else 1457 p = LIST_FIRST(&zombproc); 1458 for (; p != NULL; p = LIST_NEXT(p, p_list)) { 1459 /* 1460 * Skip embryonic processes. 1461 */ 1462 PROC_LOCK(p); 1463 if (p->p_state == PRS_NEW) { 1464 PROC_UNLOCK(p); 1465 continue; 1466 } 1467 KASSERT(p->p_ucred != NULL, 1468 ("process credential is NULL for non-NEW proc")); 1469 /* 1470 * Show a user only appropriate processes. 1471 */ 1472 if (p_cansee(curthread, p)) { 1473 PROC_UNLOCK(p); 1474 continue; 1475 } 1476 /* 1477 * TODO - make more efficient (see notes below). 1478 * do by session. 1479 */ 1480 switch (oid_number) { 1481 1482 case KERN_PROC_GID: 1483 if (p->p_ucred->cr_gid != (gid_t)name[0]) { 1484 PROC_UNLOCK(p); 1485 continue; 1486 } 1487 break; 1488 1489 case KERN_PROC_PGRP: 1490 /* could do this by traversing pgrp */ 1491 if (p->p_pgrp == NULL || 1492 p->p_pgrp->pg_id != (pid_t)name[0]) { 1493 PROC_UNLOCK(p); 1494 continue; 1495 } 1496 break; 1497 1498 case KERN_PROC_RGID: 1499 if (p->p_ucred->cr_rgid != (gid_t)name[0]) { 1500 PROC_UNLOCK(p); 1501 continue; 1502 } 1503 break; 1504 1505 case KERN_PROC_SESSION: 1506 if (p->p_session == NULL || 1507 p->p_session->s_sid != (pid_t)name[0]) { 1508 PROC_UNLOCK(p); 1509 continue; 1510 } 1511 break; 1512 1513 case KERN_PROC_TTY: 1514 if ((p->p_flag & P_CONTROLT) == 0 || 1515 p->p_session == NULL) { 1516 PROC_UNLOCK(p); 1517 continue; 1518 } 1519 /* XXX proctree_lock */ 1520 SESS_LOCK(p->p_session); 1521 if (p->p_session->s_ttyp == NULL || 1522 tty_udev(p->p_session->s_ttyp) != 1523 (dev_t)name[0]) { 1524 SESS_UNLOCK(p->p_session); 1525 PROC_UNLOCK(p); 1526 continue; 1527 } 1528 SESS_UNLOCK(p->p_session); 1529 break; 1530 1531 case KERN_PROC_UID: 1532 if (p->p_ucred->cr_uid != (uid_t)name[0]) { 1533 PROC_UNLOCK(p); 1534 continue; 1535 } 1536 break; 1537 1538 case KERN_PROC_RUID: 1539 if (p->p_ucred->cr_ruid != (uid_t)name[0]) { 1540 PROC_UNLOCK(p); 1541 continue; 1542 } 1543 break; 1544 1545 case KERN_PROC_PROC: 1546 break; 1547 1548 default: 1549 break; 1550 1551 } 1552 1553 error = sysctl_out_proc(p, req, flags, doingzomb); 1554 if (error) { 1555 sx_sunlock(&allproc_lock); 1556 sx_sunlock(&proctree_lock); 1557 return (error); 1558 } 1559 } 1560 } 1561 sx_sunlock(&allproc_lock); 1562 sx_sunlock(&proctree_lock); 1563 return (0); 1564 } 1565 1566 struct pargs * 1567 pargs_alloc(int len) 1568 { 1569 struct pargs *pa; 1570 1571 pa = malloc(sizeof(struct pargs) + len, M_PARGS, 1572 M_WAITOK); 1573 refcount_init(&pa->ar_ref, 1); 1574 pa->ar_length = len; 1575 return (pa); 1576 } 1577 1578 static void 1579 pargs_free(struct pargs *pa) 1580 { 1581 1582 free(pa, M_PARGS); 1583 } 1584 1585 void 1586 pargs_hold(struct pargs *pa) 1587 { 1588 1589 if (pa == NULL) 1590 return; 1591 refcount_acquire(&pa->ar_ref); 1592 } 1593 1594 void 1595 pargs_drop(struct pargs *pa) 1596 { 1597 1598 if (pa == NULL) 1599 return; 1600 if (refcount_release(&pa->ar_ref)) 1601 pargs_free(pa); 1602 } 1603 1604 static int 1605 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, 1606 size_t len) 1607 { 1608 ssize_t n; 1609 1610 /* 1611 * This may return a short read if the string is shorter than the chunk 1612 * and is aligned at the end of the page, and the following page is not 1613 * mapped. 1614 */ 1615 n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len); 1616 if (n <= 0) 1617 return (ENOMEM); 1618 return (0); 1619 } 1620 1621 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ 1622 1623 enum proc_vector_type { 1624 PROC_ARG, 1625 PROC_ENV, 1626 PROC_AUX, 1627 }; 1628 1629 #ifdef COMPAT_FREEBSD32 1630 static int 1631 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, 1632 size_t *vsizep, enum proc_vector_type type) 1633 { 1634 struct freebsd32_ps_strings pss; 1635 Elf32_Auxinfo aux; 1636 vm_offset_t vptr, ptr; 1637 uint32_t *proc_vector32; 1638 char **proc_vector; 1639 size_t vsize, size; 1640 int i, error; 1641 1642 error = 0; 1643 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss, 1644 sizeof(pss)) != sizeof(pss)) 1645 return (ENOMEM); 1646 switch (type) { 1647 case PROC_ARG: 1648 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); 1649 vsize = pss.ps_nargvstr; 1650 if (vsize > ARG_MAX) 1651 return (ENOEXEC); 1652 size = vsize * sizeof(int32_t); 1653 break; 1654 case PROC_ENV: 1655 vptr = (vm_offset_t)PTRIN(pss.ps_envstr); 1656 vsize = pss.ps_nenvstr; 1657 if (vsize > ARG_MAX) 1658 return (ENOEXEC); 1659 size = vsize * sizeof(int32_t); 1660 break; 1661 case PROC_AUX: 1662 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + 1663 (pss.ps_nenvstr + 1) * sizeof(int32_t); 1664 if (vptr % 4 != 0) 1665 return (ENOEXEC); 1666 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1667 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1668 sizeof(aux)) 1669 return (ENOMEM); 1670 if (aux.a_type == AT_NULL) 1671 break; 1672 ptr += sizeof(aux); 1673 } 1674 if (aux.a_type != AT_NULL) 1675 return (ENOEXEC); 1676 vsize = i + 1; 1677 size = vsize * sizeof(aux); 1678 break; 1679 default: 1680 KASSERT(0, ("Wrong proc vector type: %d", type)); 1681 return (EINVAL); 1682 } 1683 proc_vector32 = malloc(size, M_TEMP, M_WAITOK); 1684 if (proc_readmem(td, p, vptr, proc_vector32, size) != size) { 1685 error = ENOMEM; 1686 goto done; 1687 } 1688 if (type == PROC_AUX) { 1689 *proc_vectorp = (char **)proc_vector32; 1690 *vsizep = vsize; 1691 return (0); 1692 } 1693 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); 1694 for (i = 0; i < (int)vsize; i++) 1695 proc_vector[i] = PTRIN(proc_vector32[i]); 1696 *proc_vectorp = proc_vector; 1697 *vsizep = vsize; 1698 done: 1699 free(proc_vector32, M_TEMP); 1700 return (error); 1701 } 1702 #endif 1703 1704 static int 1705 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, 1706 size_t *vsizep, enum proc_vector_type type) 1707 { 1708 struct ps_strings pss; 1709 Elf_Auxinfo aux; 1710 vm_offset_t vptr, ptr; 1711 char **proc_vector; 1712 size_t vsize, size; 1713 int i; 1714 1715 #ifdef COMPAT_FREEBSD32 1716 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1717 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); 1718 #endif 1719 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss, 1720 sizeof(pss)) != sizeof(pss)) 1721 return (ENOMEM); 1722 switch (type) { 1723 case PROC_ARG: 1724 vptr = (vm_offset_t)pss.ps_argvstr; 1725 vsize = pss.ps_nargvstr; 1726 if (vsize > ARG_MAX) 1727 return (ENOEXEC); 1728 size = vsize * sizeof(char *); 1729 break; 1730 case PROC_ENV: 1731 vptr = (vm_offset_t)pss.ps_envstr; 1732 vsize = pss.ps_nenvstr; 1733 if (vsize > ARG_MAX) 1734 return (ENOEXEC); 1735 size = vsize * sizeof(char *); 1736 break; 1737 case PROC_AUX: 1738 /* 1739 * The aux array is just above env array on the stack. Check 1740 * that the address is naturally aligned. 1741 */ 1742 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) 1743 * sizeof(char *); 1744 #if __ELF_WORD_SIZE == 64 1745 if (vptr % sizeof(uint64_t) != 0) 1746 #else 1747 if (vptr % sizeof(uint32_t) != 0) 1748 #endif 1749 return (ENOEXEC); 1750 /* 1751 * We count the array size reading the aux vectors from the 1752 * stack until AT_NULL vector is returned. So (to keep the code 1753 * simple) we read the process stack twice: the first time here 1754 * to find the size and the second time when copying the vectors 1755 * to the allocated proc_vector. 1756 */ 1757 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1758 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1759 sizeof(aux)) 1760 return (ENOMEM); 1761 if (aux.a_type == AT_NULL) 1762 break; 1763 ptr += sizeof(aux); 1764 } 1765 /* 1766 * If the PROC_AUXV_MAX entries are iterated over, and we have 1767 * not reached AT_NULL, it is most likely we are reading wrong 1768 * data: either the process doesn't have auxv array or data has 1769 * been modified. Return the error in this case. 1770 */ 1771 if (aux.a_type != AT_NULL) 1772 return (ENOEXEC); 1773 vsize = i + 1; 1774 size = vsize * sizeof(aux); 1775 break; 1776 default: 1777 KASSERT(0, ("Wrong proc vector type: %d", type)); 1778 return (EINVAL); /* In case we are built without INVARIANTS. */ 1779 } 1780 proc_vector = malloc(size, M_TEMP, M_WAITOK); 1781 if (proc_readmem(td, p, vptr, proc_vector, size) != size) { 1782 free(proc_vector, M_TEMP); 1783 return (ENOMEM); 1784 } 1785 *proc_vectorp = proc_vector; 1786 *vsizep = vsize; 1787 1788 return (0); 1789 } 1790 1791 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ 1792 1793 static int 1794 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, 1795 enum proc_vector_type type) 1796 { 1797 size_t done, len, nchr, vsize; 1798 int error, i; 1799 char **proc_vector, *sptr; 1800 char pss_string[GET_PS_STRINGS_CHUNK_SZ]; 1801 1802 PROC_ASSERT_HELD(p); 1803 1804 /* 1805 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. 1806 */ 1807 nchr = 2 * (PATH_MAX + ARG_MAX); 1808 1809 error = get_proc_vector(td, p, &proc_vector, &vsize, type); 1810 if (error != 0) 1811 return (error); 1812 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { 1813 /* 1814 * The program may have scribbled into its argv array, e.g. to 1815 * remove some arguments. If that has happened, break out 1816 * before trying to read from NULL. 1817 */ 1818 if (proc_vector[i] == NULL) 1819 break; 1820 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { 1821 error = proc_read_string(td, p, sptr, pss_string, 1822 sizeof(pss_string)); 1823 if (error != 0) 1824 goto done; 1825 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); 1826 if (done + len >= nchr) 1827 len = nchr - done - 1; 1828 sbuf_bcat(sb, pss_string, len); 1829 if (len != GET_PS_STRINGS_CHUNK_SZ) 1830 break; 1831 done += GET_PS_STRINGS_CHUNK_SZ; 1832 } 1833 sbuf_bcat(sb, "", 1); 1834 done += len + 1; 1835 } 1836 done: 1837 free(proc_vector, M_TEMP); 1838 return (error); 1839 } 1840 1841 int 1842 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) 1843 { 1844 1845 return (get_ps_strings(curthread, p, sb, PROC_ARG)); 1846 } 1847 1848 int 1849 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) 1850 { 1851 1852 return (get_ps_strings(curthread, p, sb, PROC_ENV)); 1853 } 1854 1855 int 1856 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) 1857 { 1858 size_t vsize, size; 1859 char **auxv; 1860 int error; 1861 1862 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); 1863 if (error == 0) { 1864 #ifdef COMPAT_FREEBSD32 1865 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1866 size = vsize * sizeof(Elf32_Auxinfo); 1867 else 1868 #endif 1869 size = vsize * sizeof(Elf_Auxinfo); 1870 if (sbuf_bcat(sb, auxv, size) != 0) 1871 error = ENOMEM; 1872 free(auxv, M_TEMP); 1873 } 1874 return (error); 1875 } 1876 1877 /* 1878 * This sysctl allows a process to retrieve the argument list or process 1879 * title for another process without groping around in the address space 1880 * of the other process. It also allow a process to set its own "process 1881 * title to a string of its own choice. 1882 */ 1883 static int 1884 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 1885 { 1886 int *name = (int *)arg1; 1887 u_int namelen = arg2; 1888 struct pargs *newpa, *pa; 1889 struct proc *p; 1890 struct sbuf sb; 1891 int flags, error = 0, error2; 1892 1893 if (namelen != 1) 1894 return (EINVAL); 1895 1896 flags = PGET_CANSEE; 1897 if (req->newptr != NULL) 1898 flags |= PGET_ISCURRENT; 1899 error = pget((pid_t)name[0], flags, &p); 1900 if (error) 1901 return (error); 1902 1903 pa = p->p_args; 1904 if (pa != NULL) { 1905 pargs_hold(pa); 1906 PROC_UNLOCK(p); 1907 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 1908 pargs_drop(pa); 1909 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { 1910 _PHOLD(p); 1911 PROC_UNLOCK(p); 1912 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1913 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1914 error = proc_getargv(curthread, p, &sb); 1915 error2 = sbuf_finish(&sb); 1916 PRELE(p); 1917 sbuf_delete(&sb); 1918 if (error == 0 && error2 != 0) 1919 error = error2; 1920 } else { 1921 PROC_UNLOCK(p); 1922 } 1923 if (error != 0 || req->newptr == NULL) 1924 return (error); 1925 1926 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit) 1927 return (ENOMEM); 1928 newpa = pargs_alloc(req->newlen); 1929 error = SYSCTL_IN(req, newpa->ar_args, req->newlen); 1930 if (error != 0) { 1931 pargs_free(newpa); 1932 return (error); 1933 } 1934 PROC_LOCK(p); 1935 pa = p->p_args; 1936 p->p_args = newpa; 1937 PROC_UNLOCK(p); 1938 pargs_drop(pa); 1939 return (0); 1940 } 1941 1942 /* 1943 * This sysctl allows a process to retrieve environment of another process. 1944 */ 1945 static int 1946 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) 1947 { 1948 int *name = (int *)arg1; 1949 u_int namelen = arg2; 1950 struct proc *p; 1951 struct sbuf sb; 1952 int error, error2; 1953 1954 if (namelen != 1) 1955 return (EINVAL); 1956 1957 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1958 if (error != 0) 1959 return (error); 1960 if ((p->p_flag & P_SYSTEM) != 0) { 1961 PRELE(p); 1962 return (0); 1963 } 1964 1965 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1966 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1967 error = proc_getenvv(curthread, p, &sb); 1968 error2 = sbuf_finish(&sb); 1969 PRELE(p); 1970 sbuf_delete(&sb); 1971 return (error != 0 ? error : error2); 1972 } 1973 1974 /* 1975 * This sysctl allows a process to retrieve ELF auxiliary vector of 1976 * another process. 1977 */ 1978 static int 1979 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) 1980 { 1981 int *name = (int *)arg1; 1982 u_int namelen = arg2; 1983 struct proc *p; 1984 struct sbuf sb; 1985 int error, error2; 1986 1987 if (namelen != 1) 1988 return (EINVAL); 1989 1990 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 1991 if (error != 0) 1992 return (error); 1993 if ((p->p_flag & P_SYSTEM) != 0) { 1994 PRELE(p); 1995 return (0); 1996 } 1997 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 1998 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1999 error = proc_getauxv(curthread, p, &sb); 2000 error2 = sbuf_finish(&sb); 2001 PRELE(p); 2002 sbuf_delete(&sb); 2003 return (error != 0 ? error : error2); 2004 } 2005 2006 /* 2007 * This sysctl allows a process to retrieve the path of the executable for 2008 * itself or another process. 2009 */ 2010 static int 2011 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) 2012 { 2013 pid_t *pidp = (pid_t *)arg1; 2014 unsigned int arglen = arg2; 2015 struct proc *p; 2016 struct vnode *vp; 2017 char *retbuf, *freebuf; 2018 int error; 2019 2020 if (arglen != 1) 2021 return (EINVAL); 2022 if (*pidp == -1) { /* -1 means this process */ 2023 p = req->td->td_proc; 2024 } else { 2025 error = pget(*pidp, PGET_CANSEE, &p); 2026 if (error != 0) 2027 return (error); 2028 } 2029 2030 vp = p->p_textvp; 2031 if (vp == NULL) { 2032 if (*pidp != -1) 2033 PROC_UNLOCK(p); 2034 return (0); 2035 } 2036 vref(vp); 2037 if (*pidp != -1) 2038 PROC_UNLOCK(p); 2039 error = vn_fullpath(req->td, vp, &retbuf, &freebuf); 2040 vrele(vp); 2041 if (error) 2042 return (error); 2043 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); 2044 free(freebuf, M_TEMP); 2045 return (error); 2046 } 2047 2048 static int 2049 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) 2050 { 2051 struct proc *p; 2052 char *sv_name; 2053 int *name; 2054 int namelen; 2055 int error; 2056 2057 namelen = arg2; 2058 if (namelen != 1) 2059 return (EINVAL); 2060 2061 name = (int *)arg1; 2062 error = pget((pid_t)name[0], PGET_CANSEE, &p); 2063 if (error != 0) 2064 return (error); 2065 sv_name = p->p_sysent->sv_name; 2066 PROC_UNLOCK(p); 2067 return (sysctl_handle_string(oidp, sv_name, 0, req)); 2068 } 2069 2070 #ifdef KINFO_OVMENTRY_SIZE 2071 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); 2072 #endif 2073 2074 #ifdef COMPAT_FREEBSD7 2075 static int 2076 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) 2077 { 2078 vm_map_entry_t entry, tmp_entry; 2079 unsigned int last_timestamp; 2080 char *fullpath, *freepath; 2081 struct kinfo_ovmentry *kve; 2082 struct vattr va; 2083 struct ucred *cred; 2084 int error, *name; 2085 struct vnode *vp; 2086 struct proc *p; 2087 vm_map_t map; 2088 struct vmspace *vm; 2089 2090 name = (int *)arg1; 2091 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2092 if (error != 0) 2093 return (error); 2094 vm = vmspace_acquire_ref(p); 2095 if (vm == NULL) { 2096 PRELE(p); 2097 return (ESRCH); 2098 } 2099 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2100 2101 map = &vm->vm_map; 2102 vm_map_lock_read(map); 2103 for (entry = map->header.next; entry != &map->header; 2104 entry = entry->next) { 2105 vm_object_t obj, tobj, lobj; 2106 vm_offset_t addr; 2107 2108 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2109 continue; 2110 2111 bzero(kve, sizeof(*kve)); 2112 kve->kve_structsize = sizeof(*kve); 2113 2114 kve->kve_private_resident = 0; 2115 obj = entry->object.vm_object; 2116 if (obj != NULL) { 2117 VM_OBJECT_RLOCK(obj); 2118 if (obj->shadow_count == 1) 2119 kve->kve_private_resident = 2120 obj->resident_page_count; 2121 } 2122 kve->kve_resident = 0; 2123 addr = entry->start; 2124 while (addr < entry->end) { 2125 if (pmap_extract(map->pmap, addr)) 2126 kve->kve_resident++; 2127 addr += PAGE_SIZE; 2128 } 2129 2130 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { 2131 if (tobj != obj) 2132 VM_OBJECT_RLOCK(tobj); 2133 if (lobj != obj) 2134 VM_OBJECT_RUNLOCK(lobj); 2135 lobj = tobj; 2136 } 2137 2138 kve->kve_start = (void*)entry->start; 2139 kve->kve_end = (void*)entry->end; 2140 kve->kve_offset = (off_t)entry->offset; 2141 2142 if (entry->protection & VM_PROT_READ) 2143 kve->kve_protection |= KVME_PROT_READ; 2144 if (entry->protection & VM_PROT_WRITE) 2145 kve->kve_protection |= KVME_PROT_WRITE; 2146 if (entry->protection & VM_PROT_EXECUTE) 2147 kve->kve_protection |= KVME_PROT_EXEC; 2148 2149 if (entry->eflags & MAP_ENTRY_COW) 2150 kve->kve_flags |= KVME_FLAG_COW; 2151 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2152 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2153 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2154 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2155 2156 last_timestamp = map->timestamp; 2157 vm_map_unlock_read(map); 2158 2159 kve->kve_fileid = 0; 2160 kve->kve_fsid = 0; 2161 freepath = NULL; 2162 fullpath = ""; 2163 if (lobj) { 2164 vp = NULL; 2165 switch (lobj->type) { 2166 case OBJT_DEFAULT: 2167 kve->kve_type = KVME_TYPE_DEFAULT; 2168 break; 2169 case OBJT_VNODE: 2170 kve->kve_type = KVME_TYPE_VNODE; 2171 vp = lobj->handle; 2172 vref(vp); 2173 break; 2174 case OBJT_SWAP: 2175 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) { 2176 kve->kve_type = KVME_TYPE_VNODE; 2177 if ((lobj->flags & OBJ_TMPFS) != 0) { 2178 vp = lobj->un_pager.swp.swp_tmpfs; 2179 vref(vp); 2180 } 2181 } else { 2182 kve->kve_type = KVME_TYPE_SWAP; 2183 } 2184 break; 2185 case OBJT_DEVICE: 2186 kve->kve_type = KVME_TYPE_DEVICE; 2187 break; 2188 case OBJT_PHYS: 2189 kve->kve_type = KVME_TYPE_PHYS; 2190 break; 2191 case OBJT_DEAD: 2192 kve->kve_type = KVME_TYPE_DEAD; 2193 break; 2194 case OBJT_SG: 2195 kve->kve_type = KVME_TYPE_SG; 2196 break; 2197 default: 2198 kve->kve_type = KVME_TYPE_UNKNOWN; 2199 break; 2200 } 2201 if (lobj != obj) 2202 VM_OBJECT_RUNLOCK(lobj); 2203 2204 kve->kve_ref_count = obj->ref_count; 2205 kve->kve_shadow_count = obj->shadow_count; 2206 VM_OBJECT_RUNLOCK(obj); 2207 if (vp != NULL) { 2208 vn_fullpath(curthread, vp, &fullpath, 2209 &freepath); 2210 cred = curthread->td_ucred; 2211 vn_lock(vp, LK_SHARED | LK_RETRY); 2212 if (VOP_GETATTR(vp, &va, cred) == 0) { 2213 kve->kve_fileid = va.va_fileid; 2214 /* truncate */ 2215 kve->kve_fsid = va.va_fsid; 2216 } 2217 vput(vp); 2218 } 2219 } else { 2220 kve->kve_type = KVME_TYPE_NONE; 2221 kve->kve_ref_count = 0; 2222 kve->kve_shadow_count = 0; 2223 } 2224 2225 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2226 if (freepath != NULL) 2227 free(freepath, M_TEMP); 2228 2229 error = SYSCTL_OUT(req, kve, sizeof(*kve)); 2230 vm_map_lock_read(map); 2231 if (error) 2232 break; 2233 if (last_timestamp != map->timestamp) { 2234 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2235 entry = tmp_entry; 2236 } 2237 } 2238 vm_map_unlock_read(map); 2239 vmspace_free(vm); 2240 PRELE(p); 2241 free(kve, M_TEMP); 2242 return (error); 2243 } 2244 #endif /* COMPAT_FREEBSD7 */ 2245 2246 #ifdef KINFO_VMENTRY_SIZE 2247 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); 2248 #endif 2249 2250 static void 2251 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, 2252 struct kinfo_vmentry *kve) 2253 { 2254 vm_object_t obj, tobj; 2255 vm_page_t m, m_adv; 2256 vm_offset_t addr; 2257 vm_paddr_t locked_pa; 2258 vm_pindex_t pi, pi_adv, pindex; 2259 2260 locked_pa = 0; 2261 obj = entry->object.vm_object; 2262 addr = entry->start; 2263 m_adv = NULL; 2264 pi = OFF_TO_IDX(entry->offset); 2265 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { 2266 if (m_adv != NULL) { 2267 m = m_adv; 2268 } else { 2269 pi_adv = atop(entry->end - addr); 2270 pindex = pi; 2271 for (tobj = obj;; tobj = tobj->backing_object) { 2272 m = vm_page_find_least(tobj, pindex); 2273 if (m != NULL) { 2274 if (m->pindex == pindex) 2275 break; 2276 if (pi_adv > m->pindex - pindex) { 2277 pi_adv = m->pindex - pindex; 2278 m_adv = m; 2279 } 2280 } 2281 if (tobj->backing_object == NULL) 2282 goto next; 2283 pindex += OFF_TO_IDX(tobj-> 2284 backing_object_offset); 2285 } 2286 } 2287 m_adv = NULL; 2288 if (m->psind != 0 && addr + pagesizes[1] <= entry->end && 2289 (addr & (pagesizes[1] - 1)) == 0 && 2290 (pmap_mincore(map->pmap, addr, &locked_pa) & 2291 MINCORE_SUPER) != 0) { 2292 kve->kve_flags |= KVME_FLAG_SUPER; 2293 pi_adv = atop(pagesizes[1]); 2294 } else { 2295 /* 2296 * We do not test the found page on validity. 2297 * Either the page is busy and being paged in, 2298 * or it was invalidated. The first case 2299 * should be counted as resident, the second 2300 * is not so clear; we do account both. 2301 */ 2302 pi_adv = 1; 2303 } 2304 kve->kve_resident += pi_adv; 2305 next:; 2306 } 2307 PA_UNLOCK_COND(locked_pa); 2308 } 2309 2310 /* 2311 * Must be called with the process locked and will return unlocked. 2312 */ 2313 int 2314 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags) 2315 { 2316 vm_map_entry_t entry, tmp_entry; 2317 struct vattr va; 2318 vm_map_t map; 2319 vm_object_t obj, tobj, lobj; 2320 char *fullpath, *freepath; 2321 struct kinfo_vmentry *kve; 2322 struct ucred *cred; 2323 struct vnode *vp; 2324 struct vmspace *vm; 2325 vm_offset_t addr; 2326 unsigned int last_timestamp; 2327 int error; 2328 2329 PROC_LOCK_ASSERT(p, MA_OWNED); 2330 2331 _PHOLD(p); 2332 PROC_UNLOCK(p); 2333 vm = vmspace_acquire_ref(p); 2334 if (vm == NULL) { 2335 PRELE(p); 2336 return (ESRCH); 2337 } 2338 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO); 2339 2340 error = 0; 2341 map = &vm->vm_map; 2342 vm_map_lock_read(map); 2343 for (entry = map->header.next; entry != &map->header; 2344 entry = entry->next) { 2345 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2346 continue; 2347 2348 addr = entry->end; 2349 bzero(kve, sizeof(*kve)); 2350 obj = entry->object.vm_object; 2351 if (obj != NULL) { 2352 for (tobj = obj; tobj != NULL; 2353 tobj = tobj->backing_object) { 2354 VM_OBJECT_RLOCK(tobj); 2355 lobj = tobj; 2356 } 2357 if (obj->backing_object == NULL) 2358 kve->kve_private_resident = 2359 obj->resident_page_count; 2360 if (!vmmap_skip_res_cnt) 2361 kern_proc_vmmap_resident(map, entry, kve); 2362 for (tobj = obj; tobj != NULL; 2363 tobj = tobj->backing_object) { 2364 if (tobj != obj && tobj != lobj) 2365 VM_OBJECT_RUNLOCK(tobj); 2366 } 2367 } else { 2368 lobj = NULL; 2369 } 2370 2371 kve->kve_start = entry->start; 2372 kve->kve_end = entry->end; 2373 kve->kve_offset = entry->offset; 2374 2375 if (entry->protection & VM_PROT_READ) 2376 kve->kve_protection |= KVME_PROT_READ; 2377 if (entry->protection & VM_PROT_WRITE) 2378 kve->kve_protection |= KVME_PROT_WRITE; 2379 if (entry->protection & VM_PROT_EXECUTE) 2380 kve->kve_protection |= KVME_PROT_EXEC; 2381 2382 if (entry->eflags & MAP_ENTRY_COW) 2383 kve->kve_flags |= KVME_FLAG_COW; 2384 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2385 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2386 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2387 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2388 if (entry->eflags & MAP_ENTRY_GROWS_UP) 2389 kve->kve_flags |= KVME_FLAG_GROWS_UP; 2390 if (entry->eflags & MAP_ENTRY_GROWS_DOWN) 2391 kve->kve_flags |= KVME_FLAG_GROWS_DOWN; 2392 2393 last_timestamp = map->timestamp; 2394 vm_map_unlock_read(map); 2395 2396 freepath = NULL; 2397 fullpath = ""; 2398 if (lobj != NULL) { 2399 vp = NULL; 2400 switch (lobj->type) { 2401 case OBJT_DEFAULT: 2402 kve->kve_type = KVME_TYPE_DEFAULT; 2403 break; 2404 case OBJT_VNODE: 2405 kve->kve_type = KVME_TYPE_VNODE; 2406 vp = lobj->handle; 2407 vref(vp); 2408 break; 2409 case OBJT_SWAP: 2410 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) { 2411 kve->kve_type = KVME_TYPE_VNODE; 2412 if ((lobj->flags & OBJ_TMPFS) != 0) { 2413 vp = lobj->un_pager.swp.swp_tmpfs; 2414 vref(vp); 2415 } 2416 } else { 2417 kve->kve_type = KVME_TYPE_SWAP; 2418 } 2419 break; 2420 case OBJT_DEVICE: 2421 kve->kve_type = KVME_TYPE_DEVICE; 2422 break; 2423 case OBJT_PHYS: 2424 kve->kve_type = KVME_TYPE_PHYS; 2425 break; 2426 case OBJT_DEAD: 2427 kve->kve_type = KVME_TYPE_DEAD; 2428 break; 2429 case OBJT_SG: 2430 kve->kve_type = KVME_TYPE_SG; 2431 break; 2432 case OBJT_MGTDEVICE: 2433 kve->kve_type = KVME_TYPE_MGTDEVICE; 2434 break; 2435 default: 2436 kve->kve_type = KVME_TYPE_UNKNOWN; 2437 break; 2438 } 2439 if (lobj != obj) 2440 VM_OBJECT_RUNLOCK(lobj); 2441 2442 kve->kve_ref_count = obj->ref_count; 2443 kve->kve_shadow_count = obj->shadow_count; 2444 VM_OBJECT_RUNLOCK(obj); 2445 if (vp != NULL) { 2446 vn_fullpath(curthread, vp, &fullpath, 2447 &freepath); 2448 kve->kve_vn_type = vntype_to_kinfo(vp->v_type); 2449 cred = curthread->td_ucred; 2450 vn_lock(vp, LK_SHARED | LK_RETRY); 2451 if (VOP_GETATTR(vp, &va, cred) == 0) { 2452 kve->kve_vn_fileid = va.va_fileid; 2453 kve->kve_vn_fsid = va.va_fsid; 2454 kve->kve_vn_fsid_freebsd11 = 2455 kve->kve_vn_fsid; /* truncate */ 2456 kve->kve_vn_mode = 2457 MAKEIMODE(va.va_type, va.va_mode); 2458 kve->kve_vn_size = va.va_size; 2459 kve->kve_vn_rdev = va.va_rdev; 2460 kve->kve_vn_rdev_freebsd11 = 2461 kve->kve_vn_rdev; /* truncate */ 2462 kve->kve_status = KF_ATTR_VALID; 2463 } 2464 vput(vp); 2465 } 2466 } else { 2467 kve->kve_type = KVME_TYPE_NONE; 2468 kve->kve_ref_count = 0; 2469 kve->kve_shadow_count = 0; 2470 } 2471 2472 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2473 if (freepath != NULL) 2474 free(freepath, M_TEMP); 2475 2476 /* Pack record size down */ 2477 if ((flags & KERN_VMMAP_PACK_KINFO) != 0) 2478 kve->kve_structsize = 2479 offsetof(struct kinfo_vmentry, kve_path) + 2480 strlen(kve->kve_path) + 1; 2481 else 2482 kve->kve_structsize = sizeof(*kve); 2483 kve->kve_structsize = roundup(kve->kve_structsize, 2484 sizeof(uint64_t)); 2485 2486 /* Halt filling and truncate rather than exceeding maxlen */ 2487 if (maxlen != -1 && maxlen < kve->kve_structsize) { 2488 error = 0; 2489 vm_map_lock_read(map); 2490 break; 2491 } else if (maxlen != -1) 2492 maxlen -= kve->kve_structsize; 2493 2494 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0) 2495 error = ENOMEM; 2496 vm_map_lock_read(map); 2497 if (error != 0) 2498 break; 2499 if (last_timestamp != map->timestamp) { 2500 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2501 entry = tmp_entry; 2502 } 2503 } 2504 vm_map_unlock_read(map); 2505 vmspace_free(vm); 2506 PRELE(p); 2507 free(kve, M_TEMP); 2508 return (error); 2509 } 2510 2511 static int 2512 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) 2513 { 2514 struct proc *p; 2515 struct sbuf sb; 2516 int error, error2, *name; 2517 2518 name = (int *)arg1; 2519 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); 2520 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2521 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); 2522 if (error != 0) { 2523 sbuf_delete(&sb); 2524 return (error); 2525 } 2526 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO); 2527 error2 = sbuf_finish(&sb); 2528 sbuf_delete(&sb); 2529 return (error != 0 ? error : error2); 2530 } 2531 2532 #if defined(STACK) || defined(DDB) 2533 static int 2534 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) 2535 { 2536 struct kinfo_kstack *kkstp; 2537 int error, i, *name, numthreads; 2538 lwpid_t *lwpidarray; 2539 struct thread *td; 2540 struct stack *st; 2541 struct sbuf sb; 2542 struct proc *p; 2543 2544 name = (int *)arg1; 2545 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); 2546 if (error != 0) 2547 return (error); 2548 2549 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); 2550 st = stack_create(); 2551 2552 lwpidarray = NULL; 2553 PROC_LOCK(p); 2554 do { 2555 if (lwpidarray != NULL) { 2556 free(lwpidarray, M_TEMP); 2557 lwpidarray = NULL; 2558 } 2559 numthreads = p->p_numthreads; 2560 PROC_UNLOCK(p); 2561 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, 2562 M_WAITOK | M_ZERO); 2563 PROC_LOCK(p); 2564 } while (numthreads < p->p_numthreads); 2565 2566 /* 2567 * XXXRW: During the below loop, execve(2) and countless other sorts 2568 * of changes could have taken place. Should we check to see if the 2569 * vmspace has been replaced, or the like, in order to prevent 2570 * giving a snapshot that spans, say, execve(2), with some threads 2571 * before and some after? Among other things, the credentials could 2572 * have changed, in which case the right to extract debug info might 2573 * no longer be assured. 2574 */ 2575 i = 0; 2576 FOREACH_THREAD_IN_PROC(p, td) { 2577 KASSERT(i < numthreads, 2578 ("sysctl_kern_proc_kstack: numthreads")); 2579 lwpidarray[i] = td->td_tid; 2580 i++; 2581 } 2582 numthreads = i; 2583 for (i = 0; i < numthreads; i++) { 2584 td = thread_find(p, lwpidarray[i]); 2585 if (td == NULL) { 2586 continue; 2587 } 2588 bzero(kkstp, sizeof(*kkstp)); 2589 (void)sbuf_new(&sb, kkstp->kkst_trace, 2590 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); 2591 thread_lock(td); 2592 kkstp->kkst_tid = td->td_tid; 2593 if (TD_IS_SWAPPED(td)) { 2594 kkstp->kkst_state = KKST_STATE_SWAPPED; 2595 } else if (TD_IS_RUNNING(td)) { 2596 if (stack_save_td_running(st, td) == 0) 2597 kkstp->kkst_state = KKST_STATE_STACKOK; 2598 else 2599 kkstp->kkst_state = KKST_STATE_RUNNING; 2600 } else { 2601 kkstp->kkst_state = KKST_STATE_STACKOK; 2602 stack_save_td(st, td); 2603 } 2604 thread_unlock(td); 2605 PROC_UNLOCK(p); 2606 stack_sbuf_print(&sb, st); 2607 sbuf_finish(&sb); 2608 sbuf_delete(&sb); 2609 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); 2610 PROC_LOCK(p); 2611 if (error) 2612 break; 2613 } 2614 _PRELE(p); 2615 PROC_UNLOCK(p); 2616 if (lwpidarray != NULL) 2617 free(lwpidarray, M_TEMP); 2618 stack_destroy(st); 2619 free(kkstp, M_TEMP); 2620 return (error); 2621 } 2622 #endif 2623 2624 /* 2625 * This sysctl allows a process to retrieve the full list of groups from 2626 * itself or another process. 2627 */ 2628 static int 2629 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) 2630 { 2631 pid_t *pidp = (pid_t *)arg1; 2632 unsigned int arglen = arg2; 2633 struct proc *p; 2634 struct ucred *cred; 2635 int error; 2636 2637 if (arglen != 1) 2638 return (EINVAL); 2639 if (*pidp == -1) { /* -1 means this process */ 2640 p = req->td->td_proc; 2641 PROC_LOCK(p); 2642 } else { 2643 error = pget(*pidp, PGET_CANSEE, &p); 2644 if (error != 0) 2645 return (error); 2646 } 2647 2648 cred = crhold(p->p_ucred); 2649 PROC_UNLOCK(p); 2650 2651 error = SYSCTL_OUT(req, cred->cr_groups, 2652 cred->cr_ngroups * sizeof(gid_t)); 2653 crfree(cred); 2654 return (error); 2655 } 2656 2657 /* 2658 * This sysctl allows a process to retrieve or/and set the resource limit for 2659 * another process. 2660 */ 2661 static int 2662 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) 2663 { 2664 int *name = (int *)arg1; 2665 u_int namelen = arg2; 2666 struct rlimit rlim; 2667 struct proc *p; 2668 u_int which; 2669 int flags, error; 2670 2671 if (namelen != 2) 2672 return (EINVAL); 2673 2674 which = (u_int)name[1]; 2675 if (which >= RLIM_NLIMITS) 2676 return (EINVAL); 2677 2678 if (req->newptr != NULL && req->newlen != sizeof(rlim)) 2679 return (EINVAL); 2680 2681 flags = PGET_HOLD | PGET_NOTWEXIT; 2682 if (req->newptr != NULL) 2683 flags |= PGET_CANDEBUG; 2684 else 2685 flags |= PGET_CANSEE; 2686 error = pget((pid_t)name[0], flags, &p); 2687 if (error != 0) 2688 return (error); 2689 2690 /* 2691 * Retrieve limit. 2692 */ 2693 if (req->oldptr != NULL) { 2694 PROC_LOCK(p); 2695 lim_rlimit_proc(p, which, &rlim); 2696 PROC_UNLOCK(p); 2697 } 2698 error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); 2699 if (error != 0) 2700 goto errout; 2701 2702 /* 2703 * Set limit. 2704 */ 2705 if (req->newptr != NULL) { 2706 error = SYSCTL_IN(req, &rlim, sizeof(rlim)); 2707 if (error == 0) 2708 error = kern_proc_setrlimit(curthread, p, which, &rlim); 2709 } 2710 2711 errout: 2712 PRELE(p); 2713 return (error); 2714 } 2715 2716 /* 2717 * This sysctl allows a process to retrieve ps_strings structure location of 2718 * another process. 2719 */ 2720 static int 2721 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) 2722 { 2723 int *name = (int *)arg1; 2724 u_int namelen = arg2; 2725 struct proc *p; 2726 vm_offset_t ps_strings; 2727 int error; 2728 #ifdef COMPAT_FREEBSD32 2729 uint32_t ps_strings32; 2730 #endif 2731 2732 if (namelen != 1) 2733 return (EINVAL); 2734 2735 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2736 if (error != 0) 2737 return (error); 2738 #ifdef COMPAT_FREEBSD32 2739 if ((req->flags & SCTL_MASK32) != 0) { 2740 /* 2741 * We return 0 if the 32 bit emulation request is for a 64 bit 2742 * process. 2743 */ 2744 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? 2745 PTROUT(p->p_sysent->sv_psstrings) : 0; 2746 PROC_UNLOCK(p); 2747 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); 2748 return (error); 2749 } 2750 #endif 2751 ps_strings = p->p_sysent->sv_psstrings; 2752 PROC_UNLOCK(p); 2753 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); 2754 return (error); 2755 } 2756 2757 /* 2758 * This sysctl allows a process to retrieve umask of another process. 2759 */ 2760 static int 2761 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) 2762 { 2763 int *name = (int *)arg1; 2764 u_int namelen = arg2; 2765 struct proc *p; 2766 int error; 2767 u_short fd_cmask; 2768 2769 if (namelen != 1) 2770 return (EINVAL); 2771 2772 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2773 if (error != 0) 2774 return (error); 2775 2776 FILEDESC_SLOCK(p->p_fd); 2777 fd_cmask = p->p_fd->fd_cmask; 2778 FILEDESC_SUNLOCK(p->p_fd); 2779 PRELE(p); 2780 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask)); 2781 return (error); 2782 } 2783 2784 /* 2785 * This sysctl allows a process to set and retrieve binary osreldate of 2786 * another process. 2787 */ 2788 static int 2789 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) 2790 { 2791 int *name = (int *)arg1; 2792 u_int namelen = arg2; 2793 struct proc *p; 2794 int flags, error, osrel; 2795 2796 if (namelen != 1) 2797 return (EINVAL); 2798 2799 if (req->newptr != NULL && req->newlen != sizeof(osrel)) 2800 return (EINVAL); 2801 2802 flags = PGET_HOLD | PGET_NOTWEXIT; 2803 if (req->newptr != NULL) 2804 flags |= PGET_CANDEBUG; 2805 else 2806 flags |= PGET_CANSEE; 2807 error = pget((pid_t)name[0], flags, &p); 2808 if (error != 0) 2809 return (error); 2810 2811 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); 2812 if (error != 0) 2813 goto errout; 2814 2815 if (req->newptr != NULL) { 2816 error = SYSCTL_IN(req, &osrel, sizeof(osrel)); 2817 if (error != 0) 2818 goto errout; 2819 if (osrel < 0) { 2820 error = EINVAL; 2821 goto errout; 2822 } 2823 p->p_osrel = osrel; 2824 } 2825 errout: 2826 PRELE(p); 2827 return (error); 2828 } 2829 2830 static int 2831 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) 2832 { 2833 int *name = (int *)arg1; 2834 u_int namelen = arg2; 2835 struct proc *p; 2836 struct kinfo_sigtramp kst; 2837 const struct sysentvec *sv; 2838 int error; 2839 #ifdef COMPAT_FREEBSD32 2840 struct kinfo_sigtramp32 kst32; 2841 #endif 2842 2843 if (namelen != 1) 2844 return (EINVAL); 2845 2846 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2847 if (error != 0) 2848 return (error); 2849 sv = p->p_sysent; 2850 #ifdef COMPAT_FREEBSD32 2851 if ((req->flags & SCTL_MASK32) != 0) { 2852 bzero(&kst32, sizeof(kst32)); 2853 if (SV_PROC_FLAG(p, SV_ILP32)) { 2854 if (sv->sv_sigcode_base != 0) { 2855 kst32.ksigtramp_start = sv->sv_sigcode_base; 2856 kst32.ksigtramp_end = sv->sv_sigcode_base + 2857 *sv->sv_szsigcode; 2858 } else { 2859 kst32.ksigtramp_start = sv->sv_psstrings - 2860 *sv->sv_szsigcode; 2861 kst32.ksigtramp_end = sv->sv_psstrings; 2862 } 2863 } 2864 PROC_UNLOCK(p); 2865 error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); 2866 return (error); 2867 } 2868 #endif 2869 bzero(&kst, sizeof(kst)); 2870 if (sv->sv_sigcode_base != 0) { 2871 kst.ksigtramp_start = (char *)sv->sv_sigcode_base; 2872 kst.ksigtramp_end = (char *)sv->sv_sigcode_base + 2873 *sv->sv_szsigcode; 2874 } else { 2875 kst.ksigtramp_start = (char *)sv->sv_psstrings - 2876 *sv->sv_szsigcode; 2877 kst.ksigtramp_end = (char *)sv->sv_psstrings; 2878 } 2879 PROC_UNLOCK(p); 2880 error = SYSCTL_OUT(req, &kst, sizeof(kst)); 2881 return (error); 2882 } 2883 2884 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table"); 2885 2886 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| 2887 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", 2888 "Return entire process table"); 2889 2890 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2891 sysctl_kern_proc, "Process table"); 2892 2893 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, 2894 sysctl_kern_proc, "Process table"); 2895 2896 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2897 sysctl_kern_proc, "Process table"); 2898 2899 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | 2900 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2901 2902 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 2903 sysctl_kern_proc, "Process table"); 2904 2905 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2906 sysctl_kern_proc, "Process table"); 2907 2908 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2909 sysctl_kern_proc, "Process table"); 2910 2911 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, 2912 sysctl_kern_proc, "Process table"); 2913 2914 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 2915 sysctl_kern_proc, "Return process table, no threads"); 2916 2917 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, 2918 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, 2919 sysctl_kern_proc_args, "Process argument list"); 2920 2921 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, 2922 sysctl_kern_proc_env, "Process environment"); 2923 2924 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | 2925 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); 2926 2927 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | 2928 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); 2929 2930 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | 2931 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, 2932 "Process syscall vector name (ABI type)"); 2933 2934 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, 2935 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2936 2937 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, 2938 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2939 2940 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, 2941 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2942 2943 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), 2944 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2945 2946 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, 2947 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2948 2949 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, 2950 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2951 2952 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, 2953 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2954 2955 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, 2956 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 2957 2958 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, 2959 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, 2960 "Return process table, no threads"); 2961 2962 #ifdef COMPAT_FREEBSD7 2963 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | 2964 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); 2965 #endif 2966 2967 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | 2968 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); 2969 2970 #if defined(STACK) || defined(DDB) 2971 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | 2972 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); 2973 #endif 2974 2975 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | 2976 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); 2977 2978 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | 2979 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, 2980 "Process resource limits"); 2981 2982 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | 2983 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, 2984 "Process ps_strings location"); 2985 2986 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | 2987 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); 2988 2989 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | 2990 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, 2991 "Process binary osreldate"); 2992 2993 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | 2994 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, 2995 "Process signal trampoline location"); 2996 2997 int allproc_gen; 2998 2999 /* 3000 * stop_all_proc() purpose is to stop all process which have usermode, 3001 * except current process for obvious reasons. This makes it somewhat 3002 * unreliable when invoked from multithreaded process. The service 3003 * must not be user-callable anyway. 3004 */ 3005 void 3006 stop_all_proc(void) 3007 { 3008 struct proc *cp, *p; 3009 int r, gen; 3010 bool restart, seen_stopped, seen_exiting, stopped_some; 3011 3012 cp = curproc; 3013 allproc_loop: 3014 sx_xlock(&allproc_lock); 3015 gen = allproc_gen; 3016 seen_exiting = seen_stopped = stopped_some = restart = false; 3017 LIST_REMOVE(cp, p_list); 3018 LIST_INSERT_HEAD(&allproc, cp, p_list); 3019 for (;;) { 3020 p = LIST_NEXT(cp, p_list); 3021 if (p == NULL) 3022 break; 3023 LIST_REMOVE(cp, p_list); 3024 LIST_INSERT_AFTER(p, cp, p_list); 3025 PROC_LOCK(p); 3026 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) { 3027 PROC_UNLOCK(p); 3028 continue; 3029 } 3030 if ((p->p_flag & P_WEXIT) != 0) { 3031 seen_exiting = true; 3032 PROC_UNLOCK(p); 3033 continue; 3034 } 3035 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 3036 /* 3037 * Stopped processes are tolerated when there 3038 * are no other processes which might continue 3039 * them. P_STOPPED_SINGLE but not 3040 * P_TOTAL_STOP process still has at least one 3041 * thread running. 3042 */ 3043 seen_stopped = true; 3044 PROC_UNLOCK(p); 3045 continue; 3046 } 3047 _PHOLD(p); 3048 sx_xunlock(&allproc_lock); 3049 r = thread_single(p, SINGLE_ALLPROC); 3050 if (r != 0) 3051 restart = true; 3052 else 3053 stopped_some = true; 3054 _PRELE(p); 3055 PROC_UNLOCK(p); 3056 sx_xlock(&allproc_lock); 3057 } 3058 /* Catch forked children we did not see in iteration. */ 3059 if (gen != allproc_gen) 3060 restart = true; 3061 sx_xunlock(&allproc_lock); 3062 if (restart || stopped_some || seen_exiting || seen_stopped) { 3063 kern_yield(PRI_USER); 3064 goto allproc_loop; 3065 } 3066 } 3067 3068 void 3069 resume_all_proc(void) 3070 { 3071 struct proc *cp, *p; 3072 3073 cp = curproc; 3074 sx_xlock(&allproc_lock); 3075 LIST_REMOVE(cp, p_list); 3076 LIST_INSERT_HEAD(&allproc, cp, p_list); 3077 for (;;) { 3078 p = LIST_NEXT(cp, p_list); 3079 if (p == NULL) 3080 break; 3081 LIST_REMOVE(cp, p_list); 3082 LIST_INSERT_AFTER(p, cp, p_list); 3083 PROC_LOCK(p); 3084 if ((p->p_flag & P_TOTAL_STOP) != 0) { 3085 sx_xunlock(&allproc_lock); 3086 _PHOLD(p); 3087 thread_single_end(p, SINGLE_ALLPROC); 3088 _PRELE(p); 3089 PROC_UNLOCK(p); 3090 sx_xlock(&allproc_lock); 3091 } else { 3092 PROC_UNLOCK(p); 3093 } 3094 } 3095 sx_xunlock(&allproc_lock); 3096 } 3097 3098 /* #define TOTAL_STOP_DEBUG 1 */ 3099 #ifdef TOTAL_STOP_DEBUG 3100 volatile static int ap_resume; 3101 #include <sys/mount.h> 3102 3103 static int 3104 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) 3105 { 3106 int error, val; 3107 3108 val = 0; 3109 ap_resume = 0; 3110 error = sysctl_handle_int(oidp, &val, 0, req); 3111 if (error != 0 || req->newptr == NULL) 3112 return (error); 3113 if (val != 0) { 3114 stop_all_proc(); 3115 syncer_suspend(); 3116 while (ap_resume == 0) 3117 ; 3118 syncer_resume(); 3119 resume_all_proc(); 3120 } 3121 return (0); 3122 } 3123 3124 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | 3125 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, 3126 sysctl_debug_stop_all_proc, "I", 3127 ""); 3128 #endif 3129