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