1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1989, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_ddb.h" 38 #include "opt_ktrace.h" 39 #include "opt_kstack_pages.h" 40 #include "opt_stack.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/bitstring.h> 45 #include <sys/elf.h> 46 #include <sys/eventhandler.h> 47 #include <sys/exec.h> 48 #include <sys/jail.h> 49 #include <sys/kernel.h> 50 #include <sys/limits.h> 51 #include <sys/lock.h> 52 #include <sys/loginclass.h> 53 #include <sys/malloc.h> 54 #include <sys/mman.h> 55 #include <sys/mount.h> 56 #include <sys/mutex.h> 57 #include <sys/proc.h> 58 #include <sys/ptrace.h> 59 #include <sys/refcount.h> 60 #include <sys/resourcevar.h> 61 #include <sys/rwlock.h> 62 #include <sys/sbuf.h> 63 #include <sys/sysent.h> 64 #include <sys/sched.h> 65 #include <sys/smp.h> 66 #include <sys/stack.h> 67 #include <sys/stat.h> 68 #include <sys/dtrace_bsd.h> 69 #include <sys/sysctl.h> 70 #include <sys/filedesc.h> 71 #include <sys/tty.h> 72 #include <sys/signalvar.h> 73 #include <sys/sdt.h> 74 #include <sys/sx.h> 75 #include <sys/user.h> 76 #include <sys/vnode.h> 77 #include <sys/wait.h> 78 #ifdef KTRACE 79 #include <sys/ktrace.h> 80 #endif 81 82 #ifdef DDB 83 #include <ddb/ddb.h> 84 #endif 85 86 #include <vm/vm.h> 87 #include <vm/vm_param.h> 88 #include <vm/vm_extern.h> 89 #include <vm/pmap.h> 90 #include <vm/vm_map.h> 91 #include <vm/vm_object.h> 92 #include <vm/vm_page.h> 93 #include <vm/uma.h> 94 95 #include <fs/devfs/devfs.h> 96 97 #ifdef COMPAT_FREEBSD32 98 #include <compat/freebsd32/freebsd32.h> 99 #include <compat/freebsd32/freebsd32_util.h> 100 #endif 101 102 SDT_PROVIDER_DEFINE(proc); 103 104 MALLOC_DEFINE(M_SESSION, "session", "session header"); 105 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); 106 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 107 108 static void doenterpgrp(struct proc *, struct pgrp *); 109 static void orphanpg(struct pgrp *pg); 110 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp); 111 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp); 112 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, 113 int preferthread); 114 static void pgdelete(struct pgrp *); 115 static int pgrp_init(void *mem, int size, int flags); 116 static int proc_ctor(void *mem, int size, void *arg, int flags); 117 static void proc_dtor(void *mem, int size, void *arg); 118 static int proc_init(void *mem, int size, int flags); 119 static void proc_fini(void *mem, int size); 120 static void pargs_free(struct pargs *pa); 121 122 /* 123 * Other process lists 124 */ 125 struct pidhashhead *pidhashtbl; 126 struct sx *pidhashtbl_lock; 127 u_long pidhash; 128 u_long pidhashlock; 129 struct pgrphashhead *pgrphashtbl; 130 u_long pgrphash; 131 struct proclist allproc; 132 struct sx __exclusive_cache_line allproc_lock; 133 struct sx __exclusive_cache_line proctree_lock; 134 struct mtx __exclusive_cache_line ppeers_lock; 135 struct mtx __exclusive_cache_line procid_lock; 136 uma_zone_t proc_zone; 137 uma_zone_t pgrp_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_hash = offsetof(struct proc, p_hash); 148 const int proc_off_p_threads = offsetof(struct proc, p_threads); 149 const int thread_off_td_tid = offsetof(struct thread, td_tid); 150 const int thread_off_td_name = offsetof(struct thread, td_name); 151 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu); 152 const int thread_off_td_pcb = offsetof(struct thread, td_pcb); 153 const int thread_off_td_plist = offsetof(struct thread, td_plist); 154 155 EVENTHANDLER_LIST_DEFINE(process_ctor); 156 EVENTHANDLER_LIST_DEFINE(process_dtor); 157 EVENTHANDLER_LIST_DEFINE(process_init); 158 EVENTHANDLER_LIST_DEFINE(process_fini); 159 EVENTHANDLER_LIST_DEFINE(process_exit); 160 EVENTHANDLER_LIST_DEFINE(process_fork); 161 EVENTHANDLER_LIST_DEFINE(process_exec); 162 163 int kstack_pages = KSTACK_PAGES; 164 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, 165 "Kernel stack size in pages"); 166 static int vmmap_skip_res_cnt = 0; 167 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW, 168 &vmmap_skip_res_cnt, 0, 169 "Skip calculation of the pages resident count in kern.proc.vmmap"); 170 171 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); 172 #ifdef COMPAT_FREEBSD32 173 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE); 174 #endif 175 176 /* 177 * Initialize global process hashing structures. 178 */ 179 void 180 procinit(void) 181 { 182 u_long i; 183 184 sx_init(&allproc_lock, "allproc"); 185 sx_init(&proctree_lock, "proctree"); 186 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF); 187 mtx_init(&procid_lock, "procid", NULL, MTX_DEF); 188 LIST_INIT(&allproc); 189 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); 190 pidhashlock = (pidhash + 1) / 64; 191 if (pidhashlock > 0) 192 pidhashlock--; 193 pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1), 194 M_PROC, M_WAITOK | M_ZERO); 195 for (i = 0; i < pidhashlock + 1; i++) 196 sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK); 197 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); 198 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(), 199 proc_ctor, proc_dtor, proc_init, proc_fini, 200 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 201 pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL, 202 pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 203 uihashinit(); 204 } 205 206 /* 207 * Prepare a proc for use. 208 */ 209 static int 210 proc_ctor(void *mem, int size, void *arg, int flags) 211 { 212 struct proc *p; 213 struct thread *td; 214 215 p = (struct proc *)mem; 216 #ifdef KDTRACE_HOOKS 217 kdtrace_proc_ctor(p); 218 #endif 219 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p); 220 td = FIRST_THREAD_IN_PROC(p); 221 if (td != NULL) { 222 /* Make sure all thread constructors are executed */ 223 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td); 224 } 225 return (0); 226 } 227 228 /* 229 * Reclaim a proc after use. 230 */ 231 static void 232 proc_dtor(void *mem, int size, void *arg) 233 { 234 struct proc *p; 235 struct thread *td; 236 237 /* INVARIANTS checks go here */ 238 p = (struct proc *)mem; 239 td = FIRST_THREAD_IN_PROC(p); 240 if (td != NULL) { 241 #ifdef INVARIANTS 242 KASSERT((p->p_numthreads == 1), 243 ("bad number of threads in exiting process")); 244 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr")); 245 #endif 246 /* Free all OSD associated to this thread. */ 247 osd_thread_exit(td); 248 td_softdep_cleanup(td); 249 MPASS(td->td_su == NULL); 250 251 /* Make sure all thread destructors are executed */ 252 EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td); 253 } 254 EVENTHANDLER_DIRECT_INVOKE(process_dtor, p); 255 #ifdef KDTRACE_HOOKS 256 kdtrace_proc_dtor(p); 257 #endif 258 if (p->p_ksi != NULL) 259 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue")); 260 } 261 262 /* 263 * Initialize type-stable parts of a proc (when newly created). 264 */ 265 static int 266 proc_init(void *mem, int size, int flags) 267 { 268 struct proc *p; 269 270 p = (struct proc *)mem; 271 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW); 272 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW); 273 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW); 274 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW); 275 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW); 276 cv_init(&p->p_pwait, "ppwait"); 277 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 278 EVENTHANDLER_DIRECT_INVOKE(process_init, p); 279 p->p_stats = pstats_alloc(); 280 p->p_pgrp = NULL; 281 return (0); 282 } 283 284 /* 285 * UMA should ensure that this function is never called. 286 * Freeing a proc structure would violate type stability. 287 */ 288 static void 289 proc_fini(void *mem, int size) 290 { 291 #ifdef notnow 292 struct proc *p; 293 294 p = (struct proc *)mem; 295 EVENTHANDLER_DIRECT_INVOKE(process_fini, p); 296 pstats_free(p->p_stats); 297 thread_free(FIRST_THREAD_IN_PROC(p)); 298 mtx_destroy(&p->p_mtx); 299 if (p->p_ksi != NULL) 300 ksiginfo_free(p->p_ksi); 301 #else 302 panic("proc reclaimed"); 303 #endif 304 } 305 306 static int 307 pgrp_init(void *mem, int size, int flags) 308 { 309 struct pgrp *pg; 310 311 pg = mem; 312 mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 313 return (0); 314 } 315 316 /* 317 * PID space management. 318 * 319 * These bitmaps are used by fork_findpid. 320 */ 321 bitstr_t bit_decl(proc_id_pidmap, PID_MAX); 322 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX); 323 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX); 324 bitstr_t bit_decl(proc_id_reapmap, PID_MAX); 325 326 static bitstr_t *proc_id_array[] = { 327 proc_id_pidmap, 328 proc_id_grpidmap, 329 proc_id_sessidmap, 330 proc_id_reapmap, 331 }; 332 333 void 334 proc_id_set(int type, pid_t id) 335 { 336 337 KASSERT(type >= 0 && type < nitems(proc_id_array), 338 ("invalid type %d\n", type)); 339 mtx_lock(&procid_lock); 340 KASSERT(bit_test(proc_id_array[type], id) == 0, 341 ("bit %d already set in %d\n", id, type)); 342 bit_set(proc_id_array[type], id); 343 mtx_unlock(&procid_lock); 344 } 345 346 void 347 proc_id_set_cond(int type, pid_t id) 348 { 349 350 KASSERT(type >= 0 && type < nitems(proc_id_array), 351 ("invalid type %d\n", type)); 352 if (bit_test(proc_id_array[type], id)) 353 return; 354 mtx_lock(&procid_lock); 355 bit_set(proc_id_array[type], id); 356 mtx_unlock(&procid_lock); 357 } 358 359 void 360 proc_id_clear(int type, pid_t id) 361 { 362 363 KASSERT(type >= 0 && type < nitems(proc_id_array), 364 ("invalid type %d\n", type)); 365 mtx_lock(&procid_lock); 366 KASSERT(bit_test(proc_id_array[type], id) != 0, 367 ("bit %d not set in %d\n", id, type)); 368 bit_clear(proc_id_array[type], id); 369 mtx_unlock(&procid_lock); 370 } 371 372 /* 373 * Is p an inferior of the current process? 374 */ 375 int 376 inferior(struct proc *p) 377 { 378 379 sx_assert(&proctree_lock, SX_LOCKED); 380 PROC_LOCK_ASSERT(p, MA_OWNED); 381 for (; p != curproc; p = proc_realparent(p)) { 382 if (p->p_pid == 0) 383 return (0); 384 } 385 return (1); 386 } 387 388 /* 389 * Shared lock all the pid hash lists. 390 */ 391 void 392 pidhash_slockall(void) 393 { 394 u_long i; 395 396 for (i = 0; i < pidhashlock + 1; i++) 397 sx_slock(&pidhashtbl_lock[i]); 398 } 399 400 /* 401 * Shared unlock all the pid hash lists. 402 */ 403 void 404 pidhash_sunlockall(void) 405 { 406 u_long i; 407 408 for (i = 0; i < pidhashlock + 1; i++) 409 sx_sunlock(&pidhashtbl_lock[i]); 410 } 411 412 /* 413 * Similar to pfind_any(), this function finds zombies. 414 */ 415 struct proc * 416 pfind_any_locked(pid_t pid) 417 { 418 struct proc *p; 419 420 sx_assert(PIDHASHLOCK(pid), SX_LOCKED); 421 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 422 if (p->p_pid == pid) { 423 PROC_LOCK(p); 424 if (p->p_state == PRS_NEW) { 425 PROC_UNLOCK(p); 426 p = NULL; 427 } 428 break; 429 } 430 } 431 return (p); 432 } 433 434 /* 435 * Locate a process by number. 436 * 437 * By not returning processes in the PRS_NEW state, we allow callers to avoid 438 * testing for that condition to avoid dereferencing p_ucred, et al. 439 */ 440 static __always_inline struct proc * 441 _pfind(pid_t pid, bool zombie) 442 { 443 struct proc *p; 444 445 p = curproc; 446 if (p->p_pid == pid) { 447 PROC_LOCK(p); 448 return (p); 449 } 450 sx_slock(PIDHASHLOCK(pid)); 451 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 452 if (p->p_pid == pid) { 453 PROC_LOCK(p); 454 if (p->p_state == PRS_NEW || 455 (!zombie && p->p_state == PRS_ZOMBIE)) { 456 PROC_UNLOCK(p); 457 p = NULL; 458 } 459 break; 460 } 461 } 462 sx_sunlock(PIDHASHLOCK(pid)); 463 return (p); 464 } 465 466 struct proc * 467 pfind(pid_t pid) 468 { 469 470 return (_pfind(pid, false)); 471 } 472 473 /* 474 * Same as pfind but allow zombies. 475 */ 476 struct proc * 477 pfind_any(pid_t pid) 478 { 479 480 return (_pfind(pid, true)); 481 } 482 483 /* 484 * Locate a process group by number. 485 * The caller must hold proctree_lock. 486 */ 487 struct pgrp * 488 pgfind(pid_t pgid) 489 { 490 struct pgrp *pgrp; 491 492 sx_assert(&proctree_lock, SX_LOCKED); 493 494 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { 495 if (pgrp->pg_id == pgid) { 496 PGRP_LOCK(pgrp); 497 return (pgrp); 498 } 499 } 500 return (NULL); 501 } 502 503 /* 504 * Locate process and do additional manipulations, depending on flags. 505 */ 506 int 507 pget(pid_t pid, int flags, struct proc **pp) 508 { 509 struct proc *p; 510 struct thread *td1; 511 int error; 512 513 p = curproc; 514 if (p->p_pid == pid) { 515 PROC_LOCK(p); 516 } else { 517 p = NULL; 518 if (pid <= PID_MAX) { 519 if ((flags & PGET_NOTWEXIT) == 0) 520 p = pfind_any(pid); 521 else 522 p = pfind(pid); 523 } else if ((flags & PGET_NOTID) == 0) { 524 td1 = tdfind(pid, -1); 525 if (td1 != NULL) 526 p = td1->td_proc; 527 } 528 if (p == NULL) 529 return (ESRCH); 530 if ((flags & PGET_CANSEE) != 0) { 531 error = p_cansee(curthread, p); 532 if (error != 0) 533 goto errout; 534 } 535 } 536 if ((flags & PGET_CANDEBUG) != 0) { 537 error = p_candebug(curthread, p); 538 if (error != 0) 539 goto errout; 540 } 541 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) { 542 error = EPERM; 543 goto errout; 544 } 545 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) { 546 error = ESRCH; 547 goto errout; 548 } 549 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) { 550 /* 551 * XXXRW: Not clear ESRCH is the right error during proc 552 * execve(). 553 */ 554 error = ESRCH; 555 goto errout; 556 } 557 if ((flags & PGET_HOLD) != 0) { 558 _PHOLD(p); 559 PROC_UNLOCK(p); 560 } 561 *pp = p; 562 return (0); 563 errout: 564 PROC_UNLOCK(p); 565 return (error); 566 } 567 568 /* 569 * Create a new process group. 570 * pgid must be equal to the pid of p. 571 * Begin a new session if required. 572 */ 573 int 574 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess) 575 { 576 577 sx_assert(&proctree_lock, SX_XLOCKED); 578 579 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL")); 580 KASSERT(p->p_pid == pgid, 581 ("enterpgrp: new pgrp and pid != pgid")); 582 KASSERT(pgfind(pgid) == NULL, 583 ("enterpgrp: pgrp with pgid exists")); 584 KASSERT(!SESS_LEADER(p), 585 ("enterpgrp: session leader attempted setpgrp")); 586 587 if (sess != NULL) { 588 /* 589 * new session 590 */ 591 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); 592 PROC_LOCK(p); 593 p->p_flag &= ~P_CONTROLT; 594 PROC_UNLOCK(p); 595 PGRP_LOCK(pgrp); 596 sess->s_leader = p; 597 sess->s_sid = p->p_pid; 598 proc_id_set(PROC_ID_SESSION, p->p_pid); 599 refcount_init(&sess->s_count, 1); 600 sess->s_ttyvp = NULL; 601 sess->s_ttydp = NULL; 602 sess->s_ttyp = NULL; 603 bcopy(p->p_session->s_login, sess->s_login, 604 sizeof(sess->s_login)); 605 pgrp->pg_session = sess; 606 KASSERT(p == curproc, 607 ("enterpgrp: mksession and p != curproc")); 608 } else { 609 pgrp->pg_session = p->p_session; 610 sess_hold(pgrp->pg_session); 611 PGRP_LOCK(pgrp); 612 } 613 pgrp->pg_id = pgid; 614 proc_id_set(PROC_ID_GROUP, p->p_pid); 615 LIST_INIT(&pgrp->pg_members); 616 pgrp->pg_flags = 0; 617 618 /* 619 * As we have an exclusive lock of proctree_lock, 620 * this should not deadlock. 621 */ 622 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); 623 SLIST_INIT(&pgrp->pg_sigiolst); 624 PGRP_UNLOCK(pgrp); 625 626 doenterpgrp(p, pgrp); 627 628 return (0); 629 } 630 631 /* 632 * Move p to an existing process group 633 */ 634 int 635 enterthispgrp(struct proc *p, struct pgrp *pgrp) 636 { 637 638 sx_assert(&proctree_lock, SX_XLOCKED); 639 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 640 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 641 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 642 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 643 KASSERT(pgrp->pg_session == p->p_session, 644 ("%s: pgrp's session %p, p->p_session %p proc %p\n", 645 __func__, pgrp->pg_session, p->p_session, p)); 646 KASSERT(pgrp != p->p_pgrp, 647 ("%s: p %p belongs to pgrp %p", __func__, p, pgrp)); 648 649 doenterpgrp(p, pgrp); 650 651 return (0); 652 } 653 654 /* 655 * If true, any child of q which belongs to group pgrp, qualifies the 656 * process group pgrp as not orphaned. 657 */ 658 static bool 659 isjobproc(struct proc *q, struct pgrp *pgrp) 660 { 661 sx_assert(&proctree_lock, SX_LOCKED); 662 663 return (q->p_pgrp != pgrp && 664 q->p_pgrp->pg_session == pgrp->pg_session); 665 } 666 667 static struct proc * 668 jobc_reaper(struct proc *p) 669 { 670 struct proc *pp; 671 672 sx_assert(&proctree_lock, SA_LOCKED); 673 674 for (pp = p;;) { 675 pp = pp->p_reaper; 676 if (pp->p_reaper == pp || 677 (pp->p_treeflag & P_TREE_GRPEXITED) == 0) 678 return (pp); 679 } 680 } 681 682 static struct proc * 683 jobc_parent(struct proc *p, struct proc *p_exiting) 684 { 685 struct proc *pp; 686 687 sx_assert(&proctree_lock, SA_LOCKED); 688 689 pp = proc_realparent(p); 690 if (pp->p_pptr == NULL || pp == p_exiting || 691 (pp->p_treeflag & P_TREE_GRPEXITED) == 0) 692 return (pp); 693 return (jobc_reaper(pp)); 694 } 695 696 static int 697 pgrp_calc_jobc(struct pgrp *pgrp) 698 { 699 struct proc *q; 700 int cnt; 701 702 #ifdef INVARIANTS 703 if (!mtx_owned(&pgrp->pg_mtx)) 704 sx_assert(&proctree_lock, SA_LOCKED); 705 #endif 706 707 cnt = 0; 708 LIST_FOREACH(q, &pgrp->pg_members, p_pglist) { 709 if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 || 710 q->p_pptr == NULL) 711 continue; 712 if (isjobproc(jobc_parent(q, NULL), pgrp)) 713 cnt++; 714 } 715 return (cnt); 716 } 717 718 /* 719 * Move p to a process group 720 */ 721 static void 722 doenterpgrp(struct proc *p, struct pgrp *pgrp) 723 { 724 struct pgrp *savepgrp; 725 struct proc *pp; 726 727 sx_assert(&proctree_lock, SX_XLOCKED); 728 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 729 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 730 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 731 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 732 733 savepgrp = p->p_pgrp; 734 pp = jobc_parent(p, NULL); 735 736 PGRP_LOCK(pgrp); 737 PGRP_LOCK(savepgrp); 738 if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1) 739 orphanpg(savepgrp); 740 PROC_LOCK(p); 741 LIST_REMOVE(p, p_pglist); 742 p->p_pgrp = pgrp; 743 PROC_UNLOCK(p); 744 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 745 if (isjobproc(pp, pgrp)) 746 pgrp->pg_flags &= ~PGRP_ORPHANED; 747 PGRP_UNLOCK(savepgrp); 748 PGRP_UNLOCK(pgrp); 749 if (LIST_EMPTY(&savepgrp->pg_members)) 750 pgdelete(savepgrp); 751 } 752 753 /* 754 * remove process from process group 755 */ 756 int 757 leavepgrp(struct proc *p) 758 { 759 struct pgrp *savepgrp; 760 761 sx_assert(&proctree_lock, SX_XLOCKED); 762 savepgrp = p->p_pgrp; 763 PGRP_LOCK(savepgrp); 764 PROC_LOCK(p); 765 LIST_REMOVE(p, p_pglist); 766 p->p_pgrp = NULL; 767 PROC_UNLOCK(p); 768 PGRP_UNLOCK(savepgrp); 769 if (LIST_EMPTY(&savepgrp->pg_members)) 770 pgdelete(savepgrp); 771 return (0); 772 } 773 774 /* 775 * delete a process group 776 */ 777 static void 778 pgdelete(struct pgrp *pgrp) 779 { 780 struct session *savesess; 781 struct tty *tp; 782 783 sx_assert(&proctree_lock, SX_XLOCKED); 784 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 785 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 786 787 /* 788 * Reset any sigio structures pointing to us as a result of 789 * F_SETOWN with our pgid. The proctree lock ensures that 790 * new sigio structures will not be added after this point. 791 */ 792 funsetownlst(&pgrp->pg_sigiolst); 793 794 PGRP_LOCK(pgrp); 795 tp = pgrp->pg_session->s_ttyp; 796 LIST_REMOVE(pgrp, pg_hash); 797 savesess = pgrp->pg_session; 798 PGRP_UNLOCK(pgrp); 799 800 /* Remove the reference to the pgrp before deallocating it. */ 801 if (tp != NULL) { 802 tty_lock(tp); 803 tty_rel_pgrp(tp, pgrp); 804 } 805 806 proc_id_clear(PROC_ID_GROUP, pgrp->pg_id); 807 uma_zfree(pgrp_zone, pgrp); 808 sess_release(savesess); 809 } 810 811 812 static void 813 fixjobc_kill(struct proc *p) 814 { 815 struct proc *q; 816 struct pgrp *pgrp; 817 818 sx_assert(&proctree_lock, SX_LOCKED); 819 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 820 pgrp = p->p_pgrp; 821 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 822 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 823 824 /* 825 * p no longer affects process group orphanage for children. 826 * It is marked by the flag because p is only physically 827 * removed from its process group on wait(2). 828 */ 829 MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0); 830 p->p_treeflag |= P_TREE_GRPEXITED; 831 832 /* 833 * Check if exiting p orphans its own group. 834 */ 835 pgrp = p->p_pgrp; 836 if (isjobproc(jobc_parent(p, NULL), pgrp)) { 837 PGRP_LOCK(pgrp); 838 if (pgrp_calc_jobc(pgrp) == 0) 839 orphanpg(pgrp); 840 PGRP_UNLOCK(pgrp); 841 } 842 843 /* 844 * Check this process' children to see whether they qualify 845 * their process groups after reparenting to reaper. 846 */ 847 LIST_FOREACH(q, &p->p_children, p_sibling) { 848 pgrp = q->p_pgrp; 849 PGRP_LOCK(pgrp); 850 if (pgrp_calc_jobc(pgrp) == 0) { 851 /* 852 * We want to handle exactly the children that 853 * has p as realparent. Then, when calculating 854 * jobc_parent for children, we should ignore 855 * P_TREE_GRPEXITED flag already set on p. 856 */ 857 if (jobc_parent(q, p) == p && isjobproc(p, pgrp)) 858 orphanpg(pgrp); 859 } else 860 pgrp->pg_flags &= ~PGRP_ORPHANED; 861 PGRP_UNLOCK(pgrp); 862 } 863 LIST_FOREACH(q, &p->p_orphans, p_orphan) { 864 pgrp = q->p_pgrp; 865 PGRP_LOCK(pgrp); 866 if (pgrp_calc_jobc(pgrp) == 0) { 867 if (isjobproc(p, pgrp)) 868 orphanpg(pgrp); 869 } else 870 pgrp->pg_flags &= ~PGRP_ORPHANED; 871 PGRP_UNLOCK(pgrp); 872 } 873 } 874 875 void 876 killjobc(void) 877 { 878 struct session *sp; 879 struct tty *tp; 880 struct proc *p; 881 struct vnode *ttyvp; 882 883 p = curproc; 884 MPASS(p->p_flag & P_WEXIT); 885 sx_assert(&proctree_lock, SX_LOCKED); 886 887 if (SESS_LEADER(p)) { 888 sp = p->p_session; 889 890 /* 891 * s_ttyp is not zero'd; we use this to indicate that 892 * the session once had a controlling terminal. (for 893 * logging and informational purposes) 894 */ 895 SESS_LOCK(sp); 896 ttyvp = sp->s_ttyvp; 897 tp = sp->s_ttyp; 898 sp->s_ttyvp = NULL; 899 sp->s_ttydp = NULL; 900 sp->s_leader = NULL; 901 SESS_UNLOCK(sp); 902 903 /* 904 * Signal foreground pgrp and revoke access to 905 * controlling terminal if it has not been revoked 906 * already. 907 * 908 * Because the TTY may have been revoked in the mean 909 * time and could already have a new session associated 910 * with it, make sure we don't send a SIGHUP to a 911 * foreground process group that does not belong to this 912 * session. 913 */ 914 915 if (tp != NULL) { 916 tty_lock(tp); 917 if (tp->t_session == sp) 918 tty_signal_pgrp(tp, SIGHUP); 919 tty_unlock(tp); 920 } 921 922 if (ttyvp != NULL) { 923 sx_xunlock(&proctree_lock); 924 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) { 925 VOP_REVOKE(ttyvp, REVOKEALL); 926 VOP_UNLOCK(ttyvp); 927 } 928 devfs_ctty_unref(ttyvp); 929 sx_xlock(&proctree_lock); 930 } 931 } 932 fixjobc_kill(p); 933 } 934 935 /* 936 * A process group has become orphaned, mark it as such for signal 937 * delivery code. If there are any stopped processes in the group, 938 * hang-up all process in that group. 939 */ 940 static void 941 orphanpg(struct pgrp *pg) 942 { 943 struct proc *p; 944 945 PGRP_LOCK_ASSERT(pg, MA_OWNED); 946 947 pg->pg_flags |= PGRP_ORPHANED; 948 949 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 950 PROC_LOCK(p); 951 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) { 952 PROC_UNLOCK(p); 953 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 954 PROC_LOCK(p); 955 kern_psignal(p, SIGHUP); 956 kern_psignal(p, SIGCONT); 957 PROC_UNLOCK(p); 958 } 959 return; 960 } 961 PROC_UNLOCK(p); 962 } 963 } 964 965 void 966 sess_hold(struct session *s) 967 { 968 969 refcount_acquire(&s->s_count); 970 } 971 972 void 973 sess_release(struct session *s) 974 { 975 976 if (refcount_release(&s->s_count)) { 977 if (s->s_ttyp != NULL) { 978 tty_lock(s->s_ttyp); 979 tty_rel_sess(s->s_ttyp, s); 980 } 981 proc_id_clear(PROC_ID_SESSION, s->s_sid); 982 mtx_destroy(&s->s_mtx); 983 free(s, M_SESSION); 984 } 985 } 986 987 #ifdef DDB 988 989 static void 990 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p) 991 { 992 db_printf( 993 " pid %d at %p pr %d pgrp %p e %d jc %d\n", 994 p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid, 995 p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0, 996 p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp)); 997 } 998 999 DB_SHOW_COMMAND(pgrpdump, pgrpdump) 1000 { 1001 struct pgrp *pgrp; 1002 struct proc *p; 1003 int i; 1004 1005 for (i = 0; i <= pgrphash; i++) { 1006 if (!LIST_EMPTY(&pgrphashtbl[i])) { 1007 db_printf("indx %d\n", i); 1008 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { 1009 db_printf( 1010 " pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n", 1011 pgrp, (int)pgrp->pg_id, pgrp->pg_session, 1012 pgrp->pg_session->s_count, 1013 LIST_FIRST(&pgrp->pg_members)); 1014 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) 1015 db_print_pgrp_one(pgrp, p); 1016 } 1017 } 1018 } 1019 } 1020 #endif /* DDB */ 1021 1022 /* 1023 * Calculate the kinfo_proc members which contain process-wide 1024 * informations. 1025 * Must be called with the target process locked. 1026 */ 1027 static void 1028 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp) 1029 { 1030 struct thread *td; 1031 1032 PROC_LOCK_ASSERT(p, MA_OWNED); 1033 1034 kp->ki_estcpu = 0; 1035 kp->ki_pctcpu = 0; 1036 FOREACH_THREAD_IN_PROC(p, td) { 1037 thread_lock(td); 1038 kp->ki_pctcpu += sched_pctcpu(td); 1039 kp->ki_estcpu += sched_estcpu(td); 1040 thread_unlock(td); 1041 } 1042 } 1043 1044 /* 1045 * Fill in any information that is common to all threads in the process. 1046 * Must be called with the target process locked. 1047 */ 1048 static void 1049 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp) 1050 { 1051 struct thread *td0; 1052 struct ucred *cred; 1053 struct sigacts *ps; 1054 struct timeval boottime; 1055 1056 PROC_LOCK_ASSERT(p, MA_OWNED); 1057 1058 kp->ki_structsize = sizeof(*kp); 1059 kp->ki_paddr = p; 1060 kp->ki_addr =/* p->p_addr; */0; /* XXX */ 1061 kp->ki_args = p->p_args; 1062 kp->ki_textvp = p->p_textvp; 1063 #ifdef KTRACE 1064 kp->ki_tracep = ktr_get_tracevp(p, false); 1065 kp->ki_traceflag = p->p_traceflag; 1066 #endif 1067 kp->ki_fd = p->p_fd; 1068 kp->ki_pd = p->p_pd; 1069 kp->ki_vmspace = p->p_vmspace; 1070 kp->ki_flag = p->p_flag; 1071 kp->ki_flag2 = p->p_flag2; 1072 cred = p->p_ucred; 1073 if (cred) { 1074 kp->ki_uid = cred->cr_uid; 1075 kp->ki_ruid = cred->cr_ruid; 1076 kp->ki_svuid = cred->cr_svuid; 1077 kp->ki_cr_flags = 0; 1078 if (cred->cr_flags & CRED_FLAG_CAPMODE) 1079 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE; 1080 /* XXX bde doesn't like KI_NGROUPS */ 1081 if (cred->cr_ngroups > KI_NGROUPS) { 1082 kp->ki_ngroups = KI_NGROUPS; 1083 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW; 1084 } else 1085 kp->ki_ngroups = cred->cr_ngroups; 1086 bcopy(cred->cr_groups, kp->ki_groups, 1087 kp->ki_ngroups * sizeof(gid_t)); 1088 kp->ki_rgid = cred->cr_rgid; 1089 kp->ki_svgid = cred->cr_svgid; 1090 /* If jailed(cred), emulate the old P_JAILED flag. */ 1091 if (jailed(cred)) { 1092 kp->ki_flag |= P_JAILED; 1093 /* If inside the jail, use 0 as a jail ID. */ 1094 if (cred->cr_prison != curthread->td_ucred->cr_prison) 1095 kp->ki_jid = cred->cr_prison->pr_id; 1096 } 1097 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name, 1098 sizeof(kp->ki_loginclass)); 1099 } 1100 ps = p->p_sigacts; 1101 if (ps) { 1102 mtx_lock(&ps->ps_mtx); 1103 kp->ki_sigignore = ps->ps_sigignore; 1104 kp->ki_sigcatch = ps->ps_sigcatch; 1105 mtx_unlock(&ps->ps_mtx); 1106 } 1107 if (p->p_state != PRS_NEW && 1108 p->p_state != PRS_ZOMBIE && 1109 p->p_vmspace != NULL) { 1110 struct vmspace *vm = p->p_vmspace; 1111 1112 kp->ki_size = vm->vm_map.size; 1113 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ 1114 FOREACH_THREAD_IN_PROC(p, td0) { 1115 if (!TD_IS_SWAPPED(td0)) 1116 kp->ki_rssize += td0->td_kstack_pages; 1117 } 1118 kp->ki_swrss = vm->vm_swrss; 1119 kp->ki_tsize = vm->vm_tsize; 1120 kp->ki_dsize = vm->vm_dsize; 1121 kp->ki_ssize = vm->vm_ssize; 1122 } else if (p->p_state == PRS_ZOMBIE) 1123 kp->ki_stat = SZOMB; 1124 if (kp->ki_flag & P_INMEM) 1125 kp->ki_sflag = PS_INMEM; 1126 else 1127 kp->ki_sflag = 0; 1128 /* Calculate legacy swtime as seconds since 'swtick'. */ 1129 kp->ki_swtime = (ticks - p->p_swtick) / hz; 1130 kp->ki_pid = p->p_pid; 1131 kp->ki_nice = p->p_nice; 1132 kp->ki_fibnum = p->p_fibnum; 1133 kp->ki_start = p->p_stats->p_start; 1134 getboottime(&boottime); 1135 timevaladd(&kp->ki_start, &boottime); 1136 PROC_STATLOCK(p); 1137 rufetch(p, &kp->ki_rusage); 1138 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime); 1139 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime); 1140 PROC_STATUNLOCK(p); 1141 calccru(p, &kp->ki_childutime, &kp->ki_childstime); 1142 /* Some callers want child times in a single value. */ 1143 kp->ki_childtime = kp->ki_childstime; 1144 timevaladd(&kp->ki_childtime, &kp->ki_childutime); 1145 1146 FOREACH_THREAD_IN_PROC(p, td0) 1147 kp->ki_cow += td0->td_cow; 1148 1149 if (p->p_comm[0] != '\0') 1150 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm)); 1151 if (p->p_sysent && p->p_sysent->sv_name != NULL && 1152 p->p_sysent->sv_name[0] != '\0') 1153 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul)); 1154 kp->ki_siglist = p->p_siglist; 1155 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig); 1156 kp->ki_acflag = p->p_acflag; 1157 kp->ki_lock = p->p_lock; 1158 if (p->p_pptr) { 1159 kp->ki_ppid = p->p_oppid; 1160 if (p->p_flag & P_TRACED) 1161 kp->ki_tracer = p->p_pptr->p_pid; 1162 } 1163 } 1164 1165 /* 1166 * Fill job-related process information. 1167 */ 1168 static void 1169 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp) 1170 { 1171 struct tty *tp; 1172 struct session *sp; 1173 struct pgrp *pgrp; 1174 1175 sx_assert(&proctree_lock, SA_LOCKED); 1176 PROC_LOCK_ASSERT(p, MA_OWNED); 1177 1178 pgrp = p->p_pgrp; 1179 if (pgrp == NULL) 1180 return; 1181 1182 kp->ki_pgid = pgrp->pg_id; 1183 kp->ki_jobc = pgrp_calc_jobc(pgrp); 1184 1185 sp = pgrp->pg_session; 1186 tp = NULL; 1187 1188 if (sp != NULL) { 1189 kp->ki_sid = sp->s_sid; 1190 SESS_LOCK(sp); 1191 strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login)); 1192 if (sp->s_ttyvp) 1193 kp->ki_kiflag |= KI_CTTY; 1194 if (SESS_LEADER(p)) 1195 kp->ki_kiflag |= KI_SLEADER; 1196 tp = sp->s_ttyp; 1197 SESS_UNLOCK(sp); 1198 } 1199 1200 if ((p->p_flag & P_CONTROLT) && tp != NULL) { 1201 kp->ki_tdev = tty_udev(tp); 1202 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 1203 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; 1204 if (tp->t_session) 1205 kp->ki_tsid = tp->t_session->s_sid; 1206 } else { 1207 kp->ki_tdev = NODEV; 1208 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 1209 } 1210 } 1211 1212 /* 1213 * Fill in information that is thread specific. Must be called with 1214 * target process locked. If 'preferthread' is set, overwrite certain 1215 * process-related fields that are maintained for both threads and 1216 * processes. 1217 */ 1218 static void 1219 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread) 1220 { 1221 struct proc *p; 1222 1223 p = td->td_proc; 1224 kp->ki_tdaddr = td; 1225 PROC_LOCK_ASSERT(p, MA_OWNED); 1226 1227 if (preferthread) 1228 PROC_STATLOCK(p); 1229 thread_lock(td); 1230 if (td->td_wmesg != NULL) 1231 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg)); 1232 else 1233 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg)); 1234 if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >= 1235 sizeof(kp->ki_tdname)) { 1236 strlcpy(kp->ki_moretdname, 1237 td->td_name + sizeof(kp->ki_tdname) - 1, 1238 sizeof(kp->ki_moretdname)); 1239 } else { 1240 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname)); 1241 } 1242 if (TD_ON_LOCK(td)) { 1243 kp->ki_kiflag |= KI_LOCKBLOCK; 1244 strlcpy(kp->ki_lockname, td->td_lockname, 1245 sizeof(kp->ki_lockname)); 1246 } else { 1247 kp->ki_kiflag &= ~KI_LOCKBLOCK; 1248 bzero(kp->ki_lockname, sizeof(kp->ki_lockname)); 1249 } 1250 1251 if (p->p_state == PRS_NORMAL) { /* approximate. */ 1252 if (TD_ON_RUNQ(td) || 1253 TD_CAN_RUN(td) || 1254 TD_IS_RUNNING(td)) { 1255 kp->ki_stat = SRUN; 1256 } else if (P_SHOULDSTOP(p)) { 1257 kp->ki_stat = SSTOP; 1258 } else if (TD_IS_SLEEPING(td)) { 1259 kp->ki_stat = SSLEEP; 1260 } else if (TD_ON_LOCK(td)) { 1261 kp->ki_stat = SLOCK; 1262 } else { 1263 kp->ki_stat = SWAIT; 1264 } 1265 } else if (p->p_state == PRS_ZOMBIE) { 1266 kp->ki_stat = SZOMB; 1267 } else { 1268 kp->ki_stat = SIDL; 1269 } 1270 1271 /* Things in the thread */ 1272 kp->ki_wchan = td->td_wchan; 1273 kp->ki_pri.pri_level = td->td_priority; 1274 kp->ki_pri.pri_native = td->td_base_pri; 1275 1276 /* 1277 * Note: legacy fields; clamp at the old NOCPU value and/or 1278 * the maximum u_char CPU value. 1279 */ 1280 if (td->td_lastcpu == NOCPU) 1281 kp->ki_lastcpu_old = NOCPU_OLD; 1282 else if (td->td_lastcpu > MAXCPU_OLD) 1283 kp->ki_lastcpu_old = MAXCPU_OLD; 1284 else 1285 kp->ki_lastcpu_old = td->td_lastcpu; 1286 1287 if (td->td_oncpu == NOCPU) 1288 kp->ki_oncpu_old = NOCPU_OLD; 1289 else if (td->td_oncpu > MAXCPU_OLD) 1290 kp->ki_oncpu_old = MAXCPU_OLD; 1291 else 1292 kp->ki_oncpu_old = td->td_oncpu; 1293 1294 kp->ki_lastcpu = td->td_lastcpu; 1295 kp->ki_oncpu = td->td_oncpu; 1296 kp->ki_tdflags = td->td_flags; 1297 kp->ki_tid = td->td_tid; 1298 kp->ki_numthreads = p->p_numthreads; 1299 kp->ki_pcb = td->td_pcb; 1300 kp->ki_kstack = (void *)td->td_kstack; 1301 kp->ki_slptime = (ticks - td->td_slptick) / hz; 1302 kp->ki_pri.pri_class = td->td_pri_class; 1303 kp->ki_pri.pri_user = td->td_user_pri; 1304 1305 if (preferthread) { 1306 rufetchtd(td, &kp->ki_rusage); 1307 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime); 1308 kp->ki_pctcpu = sched_pctcpu(td); 1309 kp->ki_estcpu = sched_estcpu(td); 1310 kp->ki_cow = td->td_cow; 1311 } 1312 1313 /* We can't get this anymore but ps etc never used it anyway. */ 1314 kp->ki_rqindex = 0; 1315 1316 if (preferthread) 1317 kp->ki_siglist = td->td_siglist; 1318 kp->ki_sigmask = td->td_sigmask; 1319 thread_unlock(td); 1320 if (preferthread) 1321 PROC_STATUNLOCK(p); 1322 } 1323 1324 /* 1325 * Fill in a kinfo_proc structure for the specified process. 1326 * Must be called with the target process locked. 1327 */ 1328 void 1329 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp) 1330 { 1331 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1332 1333 bzero(kp, sizeof(*kp)); 1334 1335 fill_kinfo_proc_pgrp(p,kp); 1336 fill_kinfo_proc_only(p, kp); 1337 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0); 1338 fill_kinfo_aggregate(p, kp); 1339 } 1340 1341 struct pstats * 1342 pstats_alloc(void) 1343 { 1344 1345 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK)); 1346 } 1347 1348 /* 1349 * Copy parts of p_stats; zero the rest of p_stats (statistics). 1350 */ 1351 void 1352 pstats_fork(struct pstats *src, struct pstats *dst) 1353 { 1354 1355 bzero(&dst->pstat_startzero, 1356 __rangeof(struct pstats, pstat_startzero, pstat_endzero)); 1357 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy, 1358 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy)); 1359 } 1360 1361 void 1362 pstats_free(struct pstats *ps) 1363 { 1364 1365 free(ps, M_SUBPROC); 1366 } 1367 1368 #ifdef COMPAT_FREEBSD32 1369 1370 /* 1371 * This function is typically used to copy out the kernel address, so 1372 * it can be replaced by assignment of zero. 1373 */ 1374 static inline uint32_t 1375 ptr32_trim(const void *ptr) 1376 { 1377 uintptr_t uptr; 1378 1379 uptr = (uintptr_t)ptr; 1380 return ((uptr > UINT_MAX) ? 0 : uptr); 1381 } 1382 1383 #define PTRTRIM_CP(src,dst,fld) \ 1384 do { (dst).fld = ptr32_trim((src).fld); } while (0) 1385 1386 static void 1387 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32) 1388 { 1389 int i; 1390 1391 bzero(ki32, sizeof(struct kinfo_proc32)); 1392 ki32->ki_structsize = sizeof(struct kinfo_proc32); 1393 CP(*ki, *ki32, ki_layout); 1394 PTRTRIM_CP(*ki, *ki32, ki_args); 1395 PTRTRIM_CP(*ki, *ki32, ki_paddr); 1396 PTRTRIM_CP(*ki, *ki32, ki_addr); 1397 PTRTRIM_CP(*ki, *ki32, ki_tracep); 1398 PTRTRIM_CP(*ki, *ki32, ki_textvp); 1399 PTRTRIM_CP(*ki, *ki32, ki_fd); 1400 PTRTRIM_CP(*ki, *ki32, ki_vmspace); 1401 PTRTRIM_CP(*ki, *ki32, ki_wchan); 1402 CP(*ki, *ki32, ki_pid); 1403 CP(*ki, *ki32, ki_ppid); 1404 CP(*ki, *ki32, ki_pgid); 1405 CP(*ki, *ki32, ki_tpgid); 1406 CP(*ki, *ki32, ki_sid); 1407 CP(*ki, *ki32, ki_tsid); 1408 CP(*ki, *ki32, ki_jobc); 1409 CP(*ki, *ki32, ki_tdev); 1410 CP(*ki, *ki32, ki_tdev_freebsd11); 1411 CP(*ki, *ki32, ki_siglist); 1412 CP(*ki, *ki32, ki_sigmask); 1413 CP(*ki, *ki32, ki_sigignore); 1414 CP(*ki, *ki32, ki_sigcatch); 1415 CP(*ki, *ki32, ki_uid); 1416 CP(*ki, *ki32, ki_ruid); 1417 CP(*ki, *ki32, ki_svuid); 1418 CP(*ki, *ki32, ki_rgid); 1419 CP(*ki, *ki32, ki_svgid); 1420 CP(*ki, *ki32, ki_ngroups); 1421 for (i = 0; i < KI_NGROUPS; i++) 1422 CP(*ki, *ki32, ki_groups[i]); 1423 CP(*ki, *ki32, ki_size); 1424 CP(*ki, *ki32, ki_rssize); 1425 CP(*ki, *ki32, ki_swrss); 1426 CP(*ki, *ki32, ki_tsize); 1427 CP(*ki, *ki32, ki_dsize); 1428 CP(*ki, *ki32, ki_ssize); 1429 CP(*ki, *ki32, ki_xstat); 1430 CP(*ki, *ki32, ki_acflag); 1431 CP(*ki, *ki32, ki_pctcpu); 1432 CP(*ki, *ki32, ki_estcpu); 1433 CP(*ki, *ki32, ki_slptime); 1434 CP(*ki, *ki32, ki_swtime); 1435 CP(*ki, *ki32, ki_cow); 1436 CP(*ki, *ki32, ki_runtime); 1437 TV_CP(*ki, *ki32, ki_start); 1438 TV_CP(*ki, *ki32, ki_childtime); 1439 CP(*ki, *ki32, ki_flag); 1440 CP(*ki, *ki32, ki_kiflag); 1441 CP(*ki, *ki32, ki_traceflag); 1442 CP(*ki, *ki32, ki_stat); 1443 CP(*ki, *ki32, ki_nice); 1444 CP(*ki, *ki32, ki_lock); 1445 CP(*ki, *ki32, ki_rqindex); 1446 CP(*ki, *ki32, ki_oncpu); 1447 CP(*ki, *ki32, ki_lastcpu); 1448 1449 /* XXX TODO: wrap cpu value as appropriate */ 1450 CP(*ki, *ki32, ki_oncpu_old); 1451 CP(*ki, *ki32, ki_lastcpu_old); 1452 1453 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1); 1454 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1); 1455 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1); 1456 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1); 1457 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1); 1458 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1); 1459 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1); 1460 bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1); 1461 CP(*ki, *ki32, ki_tracer); 1462 CP(*ki, *ki32, ki_flag2); 1463 CP(*ki, *ki32, ki_fibnum); 1464 CP(*ki, *ki32, ki_cr_flags); 1465 CP(*ki, *ki32, ki_jid); 1466 CP(*ki, *ki32, ki_numthreads); 1467 CP(*ki, *ki32, ki_tid); 1468 CP(*ki, *ki32, ki_pri); 1469 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage); 1470 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch); 1471 PTRTRIM_CP(*ki, *ki32, ki_pcb); 1472 PTRTRIM_CP(*ki, *ki32, ki_kstack); 1473 PTRTRIM_CP(*ki, *ki32, ki_udata); 1474 PTRTRIM_CP(*ki, *ki32, ki_tdaddr); 1475 CP(*ki, *ki32, ki_sflag); 1476 CP(*ki, *ki32, ki_tdflags); 1477 } 1478 #endif 1479 1480 static ssize_t 1481 kern_proc_out_size(struct proc *p, int flags) 1482 { 1483 ssize_t size = 0; 1484 1485 PROC_LOCK_ASSERT(p, MA_OWNED); 1486 1487 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1488 #ifdef COMPAT_FREEBSD32 1489 if ((flags & KERN_PROC_MASK32) != 0) { 1490 size += sizeof(struct kinfo_proc32); 1491 } else 1492 #endif 1493 size += sizeof(struct kinfo_proc); 1494 } else { 1495 #ifdef COMPAT_FREEBSD32 1496 if ((flags & KERN_PROC_MASK32) != 0) 1497 size += sizeof(struct kinfo_proc32) * p->p_numthreads; 1498 else 1499 #endif 1500 size += sizeof(struct kinfo_proc) * p->p_numthreads; 1501 } 1502 PROC_UNLOCK(p); 1503 return (size); 1504 } 1505 1506 int 1507 kern_proc_out(struct proc *p, struct sbuf *sb, int flags) 1508 { 1509 struct thread *td; 1510 struct kinfo_proc ki; 1511 #ifdef COMPAT_FREEBSD32 1512 struct kinfo_proc32 ki32; 1513 #endif 1514 int error; 1515 1516 PROC_LOCK_ASSERT(p, MA_OWNED); 1517 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1518 1519 error = 0; 1520 fill_kinfo_proc(p, &ki); 1521 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1522 #ifdef COMPAT_FREEBSD32 1523 if ((flags & KERN_PROC_MASK32) != 0) { 1524 freebsd32_kinfo_proc_out(&ki, &ki32); 1525 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1526 error = ENOMEM; 1527 } else 1528 #endif 1529 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1530 error = ENOMEM; 1531 } else { 1532 FOREACH_THREAD_IN_PROC(p, td) { 1533 fill_kinfo_thread(td, &ki, 1); 1534 #ifdef COMPAT_FREEBSD32 1535 if ((flags & KERN_PROC_MASK32) != 0) { 1536 freebsd32_kinfo_proc_out(&ki, &ki32); 1537 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1538 error = ENOMEM; 1539 } else 1540 #endif 1541 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1542 error = ENOMEM; 1543 if (error != 0) 1544 break; 1545 } 1546 } 1547 PROC_UNLOCK(p); 1548 return (error); 1549 } 1550 1551 static int 1552 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags) 1553 { 1554 struct sbuf sb; 1555 struct kinfo_proc ki; 1556 int error, error2; 1557 1558 if (req->oldptr == NULL) 1559 return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags))); 1560 1561 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); 1562 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1563 error = kern_proc_out(p, &sb, flags); 1564 error2 = sbuf_finish(&sb); 1565 sbuf_delete(&sb); 1566 if (error != 0) 1567 return (error); 1568 else if (error2 != 0) 1569 return (error2); 1570 return (0); 1571 } 1572 1573 int 1574 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg) 1575 { 1576 struct proc *p; 1577 int error, i, j; 1578 1579 for (i = 0; i < pidhashlock + 1; i++) { 1580 sx_slock(&proctree_lock); 1581 sx_slock(&pidhashtbl_lock[i]); 1582 for (j = i; j <= pidhash; j += pidhashlock + 1) { 1583 LIST_FOREACH(p, &pidhashtbl[j], p_hash) { 1584 if (p->p_state == PRS_NEW) 1585 continue; 1586 error = cb(p, cbarg); 1587 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 1588 if (error != 0) { 1589 sx_sunlock(&pidhashtbl_lock[i]); 1590 sx_sunlock(&proctree_lock); 1591 return (error); 1592 } 1593 } 1594 } 1595 sx_sunlock(&pidhashtbl_lock[i]); 1596 sx_sunlock(&proctree_lock); 1597 } 1598 return (0); 1599 } 1600 1601 struct kern_proc_out_args { 1602 struct sysctl_req *req; 1603 int flags; 1604 int oid_number; 1605 int *name; 1606 }; 1607 1608 static int 1609 sysctl_kern_proc_iterate(struct proc *p, void *origarg) 1610 { 1611 struct kern_proc_out_args *arg = origarg; 1612 int *name = arg->name; 1613 int oid_number = arg->oid_number; 1614 int flags = arg->flags; 1615 struct sysctl_req *req = arg->req; 1616 int error = 0; 1617 1618 PROC_LOCK(p); 1619 1620 KASSERT(p->p_ucred != NULL, 1621 ("process credential is NULL for non-NEW proc")); 1622 /* 1623 * Show a user only appropriate processes. 1624 */ 1625 if (p_cansee(curthread, p)) 1626 goto skip; 1627 /* 1628 * TODO - make more efficient (see notes below). 1629 * do by session. 1630 */ 1631 switch (oid_number) { 1632 case KERN_PROC_GID: 1633 if (p->p_ucred->cr_gid != (gid_t)name[0]) 1634 goto skip; 1635 break; 1636 1637 case KERN_PROC_PGRP: 1638 /* could do this by traversing pgrp */ 1639 if (p->p_pgrp == NULL || 1640 p->p_pgrp->pg_id != (pid_t)name[0]) 1641 goto skip; 1642 break; 1643 1644 case KERN_PROC_RGID: 1645 if (p->p_ucred->cr_rgid != (gid_t)name[0]) 1646 goto skip; 1647 break; 1648 1649 case KERN_PROC_SESSION: 1650 if (p->p_session == NULL || 1651 p->p_session->s_sid != (pid_t)name[0]) 1652 goto skip; 1653 break; 1654 1655 case KERN_PROC_TTY: 1656 if ((p->p_flag & P_CONTROLT) == 0 || 1657 p->p_session == NULL) 1658 goto skip; 1659 /* XXX proctree_lock */ 1660 SESS_LOCK(p->p_session); 1661 if (p->p_session->s_ttyp == NULL || 1662 tty_udev(p->p_session->s_ttyp) != 1663 (dev_t)name[0]) { 1664 SESS_UNLOCK(p->p_session); 1665 goto skip; 1666 } 1667 SESS_UNLOCK(p->p_session); 1668 break; 1669 1670 case KERN_PROC_UID: 1671 if (p->p_ucred->cr_uid != (uid_t)name[0]) 1672 goto skip; 1673 break; 1674 1675 case KERN_PROC_RUID: 1676 if (p->p_ucred->cr_ruid != (uid_t)name[0]) 1677 goto skip; 1678 break; 1679 1680 case KERN_PROC_PROC: 1681 break; 1682 1683 default: 1684 break; 1685 } 1686 error = sysctl_out_proc(p, req, flags); 1687 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 1688 return (error); 1689 skip: 1690 PROC_UNLOCK(p); 1691 return (0); 1692 } 1693 1694 static int 1695 sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 1696 { 1697 struct kern_proc_out_args iterarg; 1698 int *name = (int *)arg1; 1699 u_int namelen = arg2; 1700 struct proc *p; 1701 int flags, oid_number; 1702 int error = 0; 1703 1704 oid_number = oidp->oid_number; 1705 if (oid_number != KERN_PROC_ALL && 1706 (oid_number & KERN_PROC_INC_THREAD) == 0) 1707 flags = KERN_PROC_NOTHREADS; 1708 else { 1709 flags = 0; 1710 oid_number &= ~KERN_PROC_INC_THREAD; 1711 } 1712 #ifdef COMPAT_FREEBSD32 1713 if (req->flags & SCTL_MASK32) 1714 flags |= KERN_PROC_MASK32; 1715 #endif 1716 if (oid_number == KERN_PROC_PID) { 1717 if (namelen != 1) 1718 return (EINVAL); 1719 error = sysctl_wire_old_buffer(req, 0); 1720 if (error) 1721 return (error); 1722 sx_slock(&proctree_lock); 1723 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1724 if (error == 0) 1725 error = sysctl_out_proc(p, req, flags); 1726 sx_sunlock(&proctree_lock); 1727 return (error); 1728 } 1729 1730 switch (oid_number) { 1731 case KERN_PROC_ALL: 1732 if (namelen != 0) 1733 return (EINVAL); 1734 break; 1735 case KERN_PROC_PROC: 1736 if (namelen != 0 && namelen != 1) 1737 return (EINVAL); 1738 break; 1739 default: 1740 if (namelen != 1) 1741 return (EINVAL); 1742 break; 1743 } 1744 1745 if (req->oldptr == NULL) { 1746 /* overestimate by 5 procs */ 1747 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1748 if (error) 1749 return (error); 1750 } else { 1751 error = sysctl_wire_old_buffer(req, 0); 1752 if (error != 0) 1753 return (error); 1754 } 1755 iterarg.flags = flags; 1756 iterarg.oid_number = oid_number; 1757 iterarg.req = req; 1758 iterarg.name = name; 1759 error = proc_iterate(sysctl_kern_proc_iterate, &iterarg); 1760 return (error); 1761 } 1762 1763 struct pargs * 1764 pargs_alloc(int len) 1765 { 1766 struct pargs *pa; 1767 1768 pa = malloc(sizeof(struct pargs) + len, M_PARGS, 1769 M_WAITOK); 1770 refcount_init(&pa->ar_ref, 1); 1771 pa->ar_length = len; 1772 return (pa); 1773 } 1774 1775 static void 1776 pargs_free(struct pargs *pa) 1777 { 1778 1779 free(pa, M_PARGS); 1780 } 1781 1782 void 1783 pargs_hold(struct pargs *pa) 1784 { 1785 1786 if (pa == NULL) 1787 return; 1788 refcount_acquire(&pa->ar_ref); 1789 } 1790 1791 void 1792 pargs_drop(struct pargs *pa) 1793 { 1794 1795 if (pa == NULL) 1796 return; 1797 if (refcount_release(&pa->ar_ref)) 1798 pargs_free(pa); 1799 } 1800 1801 static int 1802 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, 1803 size_t len) 1804 { 1805 ssize_t n; 1806 1807 /* 1808 * This may return a short read if the string is shorter than the chunk 1809 * and is aligned at the end of the page, and the following page is not 1810 * mapped. 1811 */ 1812 n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len); 1813 if (n <= 0) 1814 return (ENOMEM); 1815 return (0); 1816 } 1817 1818 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ 1819 1820 enum proc_vector_type { 1821 PROC_ARG, 1822 PROC_ENV, 1823 PROC_AUX, 1824 }; 1825 1826 #ifdef COMPAT_FREEBSD32 1827 static int 1828 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, 1829 size_t *vsizep, enum proc_vector_type type) 1830 { 1831 struct freebsd32_ps_strings pss; 1832 Elf32_Auxinfo aux; 1833 vm_offset_t vptr, ptr; 1834 uint32_t *proc_vector32; 1835 char **proc_vector; 1836 size_t vsize, size; 1837 int i, error; 1838 1839 error = 0; 1840 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss, 1841 sizeof(pss)) != sizeof(pss)) 1842 return (ENOMEM); 1843 switch (type) { 1844 case PROC_ARG: 1845 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); 1846 vsize = pss.ps_nargvstr; 1847 if (vsize > ARG_MAX) 1848 return (ENOEXEC); 1849 size = vsize * sizeof(int32_t); 1850 break; 1851 case PROC_ENV: 1852 vptr = (vm_offset_t)PTRIN(pss.ps_envstr); 1853 vsize = pss.ps_nenvstr; 1854 if (vsize > ARG_MAX) 1855 return (ENOEXEC); 1856 size = vsize * sizeof(int32_t); 1857 break; 1858 case PROC_AUX: 1859 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + 1860 (pss.ps_nenvstr + 1) * sizeof(int32_t); 1861 if (vptr % 4 != 0) 1862 return (ENOEXEC); 1863 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1864 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1865 sizeof(aux)) 1866 return (ENOMEM); 1867 if (aux.a_type == AT_NULL) 1868 break; 1869 ptr += sizeof(aux); 1870 } 1871 if (aux.a_type != AT_NULL) 1872 return (ENOEXEC); 1873 vsize = i + 1; 1874 size = vsize * sizeof(aux); 1875 break; 1876 default: 1877 KASSERT(0, ("Wrong proc vector type: %d", type)); 1878 return (EINVAL); 1879 } 1880 proc_vector32 = malloc(size, M_TEMP, M_WAITOK); 1881 if (proc_readmem(td, p, vptr, proc_vector32, size) != size) { 1882 error = ENOMEM; 1883 goto done; 1884 } 1885 if (type == PROC_AUX) { 1886 *proc_vectorp = (char **)proc_vector32; 1887 *vsizep = vsize; 1888 return (0); 1889 } 1890 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); 1891 for (i = 0; i < (int)vsize; i++) 1892 proc_vector[i] = PTRIN(proc_vector32[i]); 1893 *proc_vectorp = proc_vector; 1894 *vsizep = vsize; 1895 done: 1896 free(proc_vector32, M_TEMP); 1897 return (error); 1898 } 1899 #endif 1900 1901 static int 1902 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, 1903 size_t *vsizep, enum proc_vector_type type) 1904 { 1905 struct ps_strings pss; 1906 Elf_Auxinfo aux; 1907 vm_offset_t vptr, ptr; 1908 char **proc_vector; 1909 size_t vsize, size; 1910 int i; 1911 1912 #ifdef COMPAT_FREEBSD32 1913 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1914 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); 1915 #endif 1916 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss, 1917 sizeof(pss)) != sizeof(pss)) 1918 return (ENOMEM); 1919 switch (type) { 1920 case PROC_ARG: 1921 vptr = (vm_offset_t)pss.ps_argvstr; 1922 vsize = pss.ps_nargvstr; 1923 if (vsize > ARG_MAX) 1924 return (ENOEXEC); 1925 size = vsize * sizeof(char *); 1926 break; 1927 case PROC_ENV: 1928 vptr = (vm_offset_t)pss.ps_envstr; 1929 vsize = pss.ps_nenvstr; 1930 if (vsize > ARG_MAX) 1931 return (ENOEXEC); 1932 size = vsize * sizeof(char *); 1933 break; 1934 case PROC_AUX: 1935 /* 1936 * The aux array is just above env array on the stack. Check 1937 * that the address is naturally aligned. 1938 */ 1939 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) 1940 * sizeof(char *); 1941 #if __ELF_WORD_SIZE == 64 1942 if (vptr % sizeof(uint64_t) != 0) 1943 #else 1944 if (vptr % sizeof(uint32_t) != 0) 1945 #endif 1946 return (ENOEXEC); 1947 /* 1948 * We count the array size reading the aux vectors from the 1949 * stack until AT_NULL vector is returned. So (to keep the code 1950 * simple) we read the process stack twice: the first time here 1951 * to find the size and the second time when copying the vectors 1952 * to the allocated proc_vector. 1953 */ 1954 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1955 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1956 sizeof(aux)) 1957 return (ENOMEM); 1958 if (aux.a_type == AT_NULL) 1959 break; 1960 ptr += sizeof(aux); 1961 } 1962 /* 1963 * If the PROC_AUXV_MAX entries are iterated over, and we have 1964 * not reached AT_NULL, it is most likely we are reading wrong 1965 * data: either the process doesn't have auxv array or data has 1966 * been modified. Return the error in this case. 1967 */ 1968 if (aux.a_type != AT_NULL) 1969 return (ENOEXEC); 1970 vsize = i + 1; 1971 size = vsize * sizeof(aux); 1972 break; 1973 default: 1974 KASSERT(0, ("Wrong proc vector type: %d", type)); 1975 return (EINVAL); /* In case we are built without INVARIANTS. */ 1976 } 1977 proc_vector = malloc(size, M_TEMP, M_WAITOK); 1978 if (proc_readmem(td, p, vptr, proc_vector, size) != size) { 1979 free(proc_vector, M_TEMP); 1980 return (ENOMEM); 1981 } 1982 *proc_vectorp = proc_vector; 1983 *vsizep = vsize; 1984 1985 return (0); 1986 } 1987 1988 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ 1989 1990 static int 1991 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, 1992 enum proc_vector_type type) 1993 { 1994 size_t done, len, nchr, vsize; 1995 int error, i; 1996 char **proc_vector, *sptr; 1997 char pss_string[GET_PS_STRINGS_CHUNK_SZ]; 1998 1999 PROC_ASSERT_HELD(p); 2000 2001 /* 2002 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. 2003 */ 2004 nchr = 2 * (PATH_MAX + ARG_MAX); 2005 2006 error = get_proc_vector(td, p, &proc_vector, &vsize, type); 2007 if (error != 0) 2008 return (error); 2009 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { 2010 /* 2011 * The program may have scribbled into its argv array, e.g. to 2012 * remove some arguments. If that has happened, break out 2013 * before trying to read from NULL. 2014 */ 2015 if (proc_vector[i] == NULL) 2016 break; 2017 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { 2018 error = proc_read_string(td, p, sptr, pss_string, 2019 sizeof(pss_string)); 2020 if (error != 0) 2021 goto done; 2022 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); 2023 if (done + len >= nchr) 2024 len = nchr - done - 1; 2025 sbuf_bcat(sb, pss_string, len); 2026 if (len != GET_PS_STRINGS_CHUNK_SZ) 2027 break; 2028 done += GET_PS_STRINGS_CHUNK_SZ; 2029 } 2030 sbuf_bcat(sb, "", 1); 2031 done += len + 1; 2032 } 2033 done: 2034 free(proc_vector, M_TEMP); 2035 return (error); 2036 } 2037 2038 int 2039 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) 2040 { 2041 2042 return (get_ps_strings(curthread, p, sb, PROC_ARG)); 2043 } 2044 2045 int 2046 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) 2047 { 2048 2049 return (get_ps_strings(curthread, p, sb, PROC_ENV)); 2050 } 2051 2052 int 2053 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) 2054 { 2055 size_t vsize, size; 2056 char **auxv; 2057 int error; 2058 2059 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); 2060 if (error == 0) { 2061 #ifdef COMPAT_FREEBSD32 2062 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 2063 size = vsize * sizeof(Elf32_Auxinfo); 2064 else 2065 #endif 2066 size = vsize * sizeof(Elf_Auxinfo); 2067 if (sbuf_bcat(sb, auxv, size) != 0) 2068 error = ENOMEM; 2069 free(auxv, M_TEMP); 2070 } 2071 return (error); 2072 } 2073 2074 /* 2075 * This sysctl allows a process to retrieve the argument list or process 2076 * title for another process without groping around in the address space 2077 * of the other process. It also allow a process to set its own "process 2078 * title to a string of its own choice. 2079 */ 2080 static int 2081 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 2082 { 2083 int *name = (int *)arg1; 2084 u_int namelen = arg2; 2085 struct pargs *newpa, *pa; 2086 struct proc *p; 2087 struct sbuf sb; 2088 int flags, error = 0, error2; 2089 pid_t pid; 2090 2091 if (namelen != 1) 2092 return (EINVAL); 2093 2094 p = curproc; 2095 pid = (pid_t)name[0]; 2096 if (pid == -1) { 2097 pid = p->p_pid; 2098 } 2099 2100 /* 2101 * If the query is for this process and it is single-threaded, there 2102 * is nobody to modify pargs, thus we can just read. 2103 */ 2104 if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL && 2105 (pa = p->p_args) != NULL) 2106 return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length)); 2107 2108 flags = PGET_CANSEE; 2109 if (req->newptr != NULL) 2110 flags |= PGET_ISCURRENT; 2111 error = pget(pid, flags, &p); 2112 if (error) 2113 return (error); 2114 2115 pa = p->p_args; 2116 if (pa != NULL) { 2117 pargs_hold(pa); 2118 PROC_UNLOCK(p); 2119 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 2120 pargs_drop(pa); 2121 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { 2122 _PHOLD(p); 2123 PROC_UNLOCK(p); 2124 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2125 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2126 error = proc_getargv(curthread, p, &sb); 2127 error2 = sbuf_finish(&sb); 2128 PRELE(p); 2129 sbuf_delete(&sb); 2130 if (error == 0 && error2 != 0) 2131 error = error2; 2132 } else { 2133 PROC_UNLOCK(p); 2134 } 2135 if (error != 0 || req->newptr == NULL) 2136 return (error); 2137 2138 if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs)) 2139 return (ENOMEM); 2140 2141 if (req->newlen == 0) { 2142 /* 2143 * Clear the argument pointer, so that we'll fetch arguments 2144 * with proc_getargv() until further notice. 2145 */ 2146 newpa = NULL; 2147 } else { 2148 newpa = pargs_alloc(req->newlen); 2149 error = SYSCTL_IN(req, newpa->ar_args, req->newlen); 2150 if (error != 0) { 2151 pargs_free(newpa); 2152 return (error); 2153 } 2154 } 2155 PROC_LOCK(p); 2156 pa = p->p_args; 2157 p->p_args = newpa; 2158 PROC_UNLOCK(p); 2159 pargs_drop(pa); 2160 return (0); 2161 } 2162 2163 /* 2164 * This sysctl allows a process to retrieve environment of another process. 2165 */ 2166 static int 2167 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) 2168 { 2169 int *name = (int *)arg1; 2170 u_int namelen = arg2; 2171 struct proc *p; 2172 struct sbuf sb; 2173 int error, error2; 2174 2175 if (namelen != 1) 2176 return (EINVAL); 2177 2178 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2179 if (error != 0) 2180 return (error); 2181 if ((p->p_flag & P_SYSTEM) != 0) { 2182 PRELE(p); 2183 return (0); 2184 } 2185 2186 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2187 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2188 error = proc_getenvv(curthread, p, &sb); 2189 error2 = sbuf_finish(&sb); 2190 PRELE(p); 2191 sbuf_delete(&sb); 2192 return (error != 0 ? error : error2); 2193 } 2194 2195 /* 2196 * This sysctl allows a process to retrieve ELF auxiliary vector of 2197 * another process. 2198 */ 2199 static int 2200 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) 2201 { 2202 int *name = (int *)arg1; 2203 u_int namelen = arg2; 2204 struct proc *p; 2205 struct sbuf sb; 2206 int error, error2; 2207 2208 if (namelen != 1) 2209 return (EINVAL); 2210 2211 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2212 if (error != 0) 2213 return (error); 2214 if ((p->p_flag & P_SYSTEM) != 0) { 2215 PRELE(p); 2216 return (0); 2217 } 2218 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2219 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2220 error = proc_getauxv(curthread, p, &sb); 2221 error2 = sbuf_finish(&sb); 2222 PRELE(p); 2223 sbuf_delete(&sb); 2224 return (error != 0 ? error : error2); 2225 } 2226 2227 /* 2228 * This sysctl allows a process to retrieve the path of the executable for 2229 * itself or another process. 2230 */ 2231 static int 2232 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) 2233 { 2234 pid_t *pidp = (pid_t *)arg1; 2235 unsigned int arglen = arg2; 2236 struct proc *p; 2237 struct vnode *vp; 2238 char *retbuf, *freebuf; 2239 int error; 2240 2241 if (arglen != 1) 2242 return (EINVAL); 2243 if (*pidp == -1) { /* -1 means this process */ 2244 p = req->td->td_proc; 2245 } else { 2246 error = pget(*pidp, PGET_CANSEE, &p); 2247 if (error != 0) 2248 return (error); 2249 } 2250 2251 vp = p->p_textvp; 2252 if (vp == NULL) { 2253 if (*pidp != -1) 2254 PROC_UNLOCK(p); 2255 return (0); 2256 } 2257 vref(vp); 2258 if (*pidp != -1) 2259 PROC_UNLOCK(p); 2260 error = vn_fullpath(vp, &retbuf, &freebuf); 2261 vrele(vp); 2262 if (error) 2263 return (error); 2264 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); 2265 free(freebuf, M_TEMP); 2266 return (error); 2267 } 2268 2269 static int 2270 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) 2271 { 2272 struct proc *p; 2273 char *sv_name; 2274 int *name; 2275 int namelen; 2276 int error; 2277 2278 namelen = arg2; 2279 if (namelen != 1) 2280 return (EINVAL); 2281 2282 name = (int *)arg1; 2283 error = pget((pid_t)name[0], PGET_CANSEE, &p); 2284 if (error != 0) 2285 return (error); 2286 sv_name = p->p_sysent->sv_name; 2287 PROC_UNLOCK(p); 2288 return (sysctl_handle_string(oidp, sv_name, 0, req)); 2289 } 2290 2291 #ifdef KINFO_OVMENTRY_SIZE 2292 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); 2293 #endif 2294 2295 #ifdef COMPAT_FREEBSD7 2296 static int 2297 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) 2298 { 2299 vm_map_entry_t entry, tmp_entry; 2300 unsigned int last_timestamp, namelen; 2301 char *fullpath, *freepath; 2302 struct kinfo_ovmentry *kve; 2303 struct vattr va; 2304 struct ucred *cred; 2305 int error, *name; 2306 struct vnode *vp; 2307 struct proc *p; 2308 vm_map_t map; 2309 struct vmspace *vm; 2310 2311 namelen = arg2; 2312 if (namelen != 1) 2313 return (EINVAL); 2314 2315 name = (int *)arg1; 2316 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2317 if (error != 0) 2318 return (error); 2319 vm = vmspace_acquire_ref(p); 2320 if (vm == NULL) { 2321 PRELE(p); 2322 return (ESRCH); 2323 } 2324 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2325 2326 map = &vm->vm_map; 2327 vm_map_lock_read(map); 2328 VM_MAP_ENTRY_FOREACH(entry, map) { 2329 vm_object_t obj, tobj, lobj; 2330 vm_offset_t addr; 2331 2332 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2333 continue; 2334 2335 bzero(kve, sizeof(*kve)); 2336 kve->kve_structsize = sizeof(*kve); 2337 2338 kve->kve_private_resident = 0; 2339 obj = entry->object.vm_object; 2340 if (obj != NULL) { 2341 VM_OBJECT_RLOCK(obj); 2342 if (obj->shadow_count == 1) 2343 kve->kve_private_resident = 2344 obj->resident_page_count; 2345 } 2346 kve->kve_resident = 0; 2347 addr = entry->start; 2348 while (addr < entry->end) { 2349 if (pmap_extract(map->pmap, addr)) 2350 kve->kve_resident++; 2351 addr += PAGE_SIZE; 2352 } 2353 2354 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { 2355 if (tobj != obj) { 2356 VM_OBJECT_RLOCK(tobj); 2357 kve->kve_offset += tobj->backing_object_offset; 2358 } 2359 if (lobj != obj) 2360 VM_OBJECT_RUNLOCK(lobj); 2361 lobj = tobj; 2362 } 2363 2364 kve->kve_start = (void*)entry->start; 2365 kve->kve_end = (void*)entry->end; 2366 kve->kve_offset += (off_t)entry->offset; 2367 2368 if (entry->protection & VM_PROT_READ) 2369 kve->kve_protection |= KVME_PROT_READ; 2370 if (entry->protection & VM_PROT_WRITE) 2371 kve->kve_protection |= KVME_PROT_WRITE; 2372 if (entry->protection & VM_PROT_EXECUTE) 2373 kve->kve_protection |= KVME_PROT_EXEC; 2374 2375 if (entry->eflags & MAP_ENTRY_COW) 2376 kve->kve_flags |= KVME_FLAG_COW; 2377 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2378 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2379 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2380 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2381 2382 last_timestamp = map->timestamp; 2383 vm_map_unlock_read(map); 2384 2385 kve->kve_fileid = 0; 2386 kve->kve_fsid = 0; 2387 freepath = NULL; 2388 fullpath = ""; 2389 if (lobj) { 2390 kve->kve_type = vm_object_kvme_type(lobj, &vp); 2391 if (kve->kve_type == KVME_TYPE_MGTDEVICE) 2392 kve->kve_type = KVME_TYPE_UNKNOWN; 2393 if (vp != NULL) 2394 vref(vp); 2395 if (lobj != obj) 2396 VM_OBJECT_RUNLOCK(lobj); 2397 2398 kve->kve_ref_count = obj->ref_count; 2399 kve->kve_shadow_count = obj->shadow_count; 2400 VM_OBJECT_RUNLOCK(obj); 2401 if (vp != NULL) { 2402 vn_fullpath(vp, &fullpath, &freepath); 2403 cred = curthread->td_ucred; 2404 vn_lock(vp, LK_SHARED | LK_RETRY); 2405 if (VOP_GETATTR(vp, &va, cred) == 0) { 2406 kve->kve_fileid = va.va_fileid; 2407 /* truncate */ 2408 kve->kve_fsid = va.va_fsid; 2409 } 2410 vput(vp); 2411 } 2412 } else { 2413 kve->kve_type = KVME_TYPE_NONE; 2414 kve->kve_ref_count = 0; 2415 kve->kve_shadow_count = 0; 2416 } 2417 2418 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2419 if (freepath != NULL) 2420 free(freepath, M_TEMP); 2421 2422 error = SYSCTL_OUT(req, kve, sizeof(*kve)); 2423 vm_map_lock_read(map); 2424 if (error) 2425 break; 2426 if (last_timestamp != map->timestamp) { 2427 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2428 entry = tmp_entry; 2429 } 2430 } 2431 vm_map_unlock_read(map); 2432 vmspace_free(vm); 2433 PRELE(p); 2434 free(kve, M_TEMP); 2435 return (error); 2436 } 2437 #endif /* COMPAT_FREEBSD7 */ 2438 2439 #ifdef KINFO_VMENTRY_SIZE 2440 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); 2441 #endif 2442 2443 void 2444 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, 2445 int *resident_count, bool *super) 2446 { 2447 vm_object_t obj, tobj; 2448 vm_page_t m, m_adv; 2449 vm_offset_t addr; 2450 vm_paddr_t pa; 2451 vm_pindex_t pi, pi_adv, pindex; 2452 2453 *super = false; 2454 *resident_count = 0; 2455 if (vmmap_skip_res_cnt) 2456 return; 2457 2458 pa = 0; 2459 obj = entry->object.vm_object; 2460 addr = entry->start; 2461 m_adv = NULL; 2462 pi = OFF_TO_IDX(entry->offset); 2463 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { 2464 if (m_adv != NULL) { 2465 m = m_adv; 2466 } else { 2467 pi_adv = atop(entry->end - addr); 2468 pindex = pi; 2469 for (tobj = obj;; tobj = tobj->backing_object) { 2470 m = vm_page_find_least(tobj, pindex); 2471 if (m != NULL) { 2472 if (m->pindex == pindex) 2473 break; 2474 if (pi_adv > m->pindex - pindex) { 2475 pi_adv = m->pindex - pindex; 2476 m_adv = m; 2477 } 2478 } 2479 if (tobj->backing_object == NULL) 2480 goto next; 2481 pindex += OFF_TO_IDX(tobj-> 2482 backing_object_offset); 2483 } 2484 } 2485 m_adv = NULL; 2486 if (m->psind != 0 && addr + pagesizes[1] <= entry->end && 2487 (addr & (pagesizes[1] - 1)) == 0 && 2488 (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) { 2489 *super = true; 2490 pi_adv = atop(pagesizes[1]); 2491 } else { 2492 /* 2493 * We do not test the found page on validity. 2494 * Either the page is busy and being paged in, 2495 * or it was invalidated. The first case 2496 * should be counted as resident, the second 2497 * is not so clear; we do account both. 2498 */ 2499 pi_adv = 1; 2500 } 2501 *resident_count += pi_adv; 2502 next:; 2503 } 2504 } 2505 2506 /* 2507 * Must be called with the process locked and will return unlocked. 2508 */ 2509 int 2510 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags) 2511 { 2512 vm_map_entry_t entry, tmp_entry; 2513 struct vattr va; 2514 vm_map_t map; 2515 vm_object_t lobj, nobj, obj, tobj; 2516 char *fullpath, *freepath; 2517 struct kinfo_vmentry *kve; 2518 struct ucred *cred; 2519 struct vnode *vp; 2520 struct vmspace *vm; 2521 vm_offset_t addr; 2522 unsigned int last_timestamp; 2523 int error; 2524 bool guard, super; 2525 2526 PROC_LOCK_ASSERT(p, MA_OWNED); 2527 2528 _PHOLD(p); 2529 PROC_UNLOCK(p); 2530 vm = vmspace_acquire_ref(p); 2531 if (vm == NULL) { 2532 PRELE(p); 2533 return (ESRCH); 2534 } 2535 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO); 2536 2537 error = 0; 2538 map = &vm->vm_map; 2539 vm_map_lock_read(map); 2540 VM_MAP_ENTRY_FOREACH(entry, map) { 2541 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2542 continue; 2543 2544 addr = entry->end; 2545 bzero(kve, sizeof(*kve)); 2546 obj = entry->object.vm_object; 2547 if (obj != NULL) { 2548 if ((obj->flags & OBJ_ANON) != 0) 2549 kve->kve_obj = (uintptr_t)obj; 2550 2551 for (tobj = obj; tobj != NULL; 2552 tobj = tobj->backing_object) { 2553 VM_OBJECT_RLOCK(tobj); 2554 kve->kve_offset += tobj->backing_object_offset; 2555 lobj = tobj; 2556 } 2557 if (obj->backing_object == NULL) 2558 kve->kve_private_resident = 2559 obj->resident_page_count; 2560 kern_proc_vmmap_resident(map, entry, 2561 &kve->kve_resident, &super); 2562 if (super) 2563 kve->kve_flags |= KVME_FLAG_SUPER; 2564 for (tobj = obj; tobj != NULL; tobj = nobj) { 2565 nobj = tobj->backing_object; 2566 if (tobj != obj && tobj != lobj) 2567 VM_OBJECT_RUNLOCK(tobj); 2568 } 2569 } else { 2570 lobj = NULL; 2571 } 2572 2573 kve->kve_start = entry->start; 2574 kve->kve_end = entry->end; 2575 kve->kve_offset += entry->offset; 2576 2577 if (entry->protection & VM_PROT_READ) 2578 kve->kve_protection |= KVME_PROT_READ; 2579 if (entry->protection & VM_PROT_WRITE) 2580 kve->kve_protection |= KVME_PROT_WRITE; 2581 if (entry->protection & VM_PROT_EXECUTE) 2582 kve->kve_protection |= KVME_PROT_EXEC; 2583 2584 if (entry->eflags & MAP_ENTRY_COW) 2585 kve->kve_flags |= KVME_FLAG_COW; 2586 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2587 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2588 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2589 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2590 if (entry->eflags & MAP_ENTRY_GROWS_UP) 2591 kve->kve_flags |= KVME_FLAG_GROWS_UP; 2592 if (entry->eflags & MAP_ENTRY_GROWS_DOWN) 2593 kve->kve_flags |= KVME_FLAG_GROWS_DOWN; 2594 if (entry->eflags & MAP_ENTRY_USER_WIRED) 2595 kve->kve_flags |= KVME_FLAG_USER_WIRED; 2596 2597 guard = (entry->eflags & MAP_ENTRY_GUARD) != 0; 2598 2599 last_timestamp = map->timestamp; 2600 vm_map_unlock_read(map); 2601 2602 freepath = NULL; 2603 fullpath = ""; 2604 if (lobj != NULL) { 2605 kve->kve_type = vm_object_kvme_type(lobj, &vp); 2606 if (vp != NULL) 2607 vref(vp); 2608 if (lobj != obj) 2609 VM_OBJECT_RUNLOCK(lobj); 2610 2611 kve->kve_ref_count = obj->ref_count; 2612 kve->kve_shadow_count = obj->shadow_count; 2613 VM_OBJECT_RUNLOCK(obj); 2614 if (vp != NULL) { 2615 vn_fullpath(vp, &fullpath, &freepath); 2616 kve->kve_vn_type = vntype_to_kinfo(vp->v_type); 2617 cred = curthread->td_ucred; 2618 vn_lock(vp, LK_SHARED | LK_RETRY); 2619 if (VOP_GETATTR(vp, &va, cred) == 0) { 2620 kve->kve_vn_fileid = va.va_fileid; 2621 kve->kve_vn_fsid = va.va_fsid; 2622 kve->kve_vn_fsid_freebsd11 = 2623 kve->kve_vn_fsid; /* truncate */ 2624 kve->kve_vn_mode = 2625 MAKEIMODE(va.va_type, va.va_mode); 2626 kve->kve_vn_size = va.va_size; 2627 kve->kve_vn_rdev = va.va_rdev; 2628 kve->kve_vn_rdev_freebsd11 = 2629 kve->kve_vn_rdev; /* truncate */ 2630 kve->kve_status = KF_ATTR_VALID; 2631 } 2632 vput(vp); 2633 } 2634 } else { 2635 kve->kve_type = guard ? KVME_TYPE_GUARD : 2636 KVME_TYPE_NONE; 2637 kve->kve_ref_count = 0; 2638 kve->kve_shadow_count = 0; 2639 } 2640 2641 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2642 if (freepath != NULL) 2643 free(freepath, M_TEMP); 2644 2645 /* Pack record size down */ 2646 if ((flags & KERN_VMMAP_PACK_KINFO) != 0) 2647 kve->kve_structsize = 2648 offsetof(struct kinfo_vmentry, kve_path) + 2649 strlen(kve->kve_path) + 1; 2650 else 2651 kve->kve_structsize = sizeof(*kve); 2652 kve->kve_structsize = roundup(kve->kve_structsize, 2653 sizeof(uint64_t)); 2654 2655 /* Halt filling and truncate rather than exceeding maxlen */ 2656 if (maxlen != -1 && maxlen < kve->kve_structsize) { 2657 error = 0; 2658 vm_map_lock_read(map); 2659 break; 2660 } else if (maxlen != -1) 2661 maxlen -= kve->kve_structsize; 2662 2663 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0) 2664 error = ENOMEM; 2665 vm_map_lock_read(map); 2666 if (error != 0) 2667 break; 2668 if (last_timestamp != map->timestamp) { 2669 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2670 entry = tmp_entry; 2671 } 2672 } 2673 vm_map_unlock_read(map); 2674 vmspace_free(vm); 2675 PRELE(p); 2676 free(kve, M_TEMP); 2677 return (error); 2678 } 2679 2680 static int 2681 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) 2682 { 2683 struct proc *p; 2684 struct sbuf sb; 2685 u_int namelen; 2686 int error, error2, *name; 2687 2688 namelen = arg2; 2689 if (namelen != 1) 2690 return (EINVAL); 2691 2692 name = (int *)arg1; 2693 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); 2694 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2695 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); 2696 if (error != 0) { 2697 sbuf_delete(&sb); 2698 return (error); 2699 } 2700 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO); 2701 error2 = sbuf_finish(&sb); 2702 sbuf_delete(&sb); 2703 return (error != 0 ? error : error2); 2704 } 2705 2706 #if defined(STACK) || defined(DDB) 2707 static int 2708 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) 2709 { 2710 struct kinfo_kstack *kkstp; 2711 int error, i, *name, numthreads; 2712 lwpid_t *lwpidarray; 2713 struct thread *td; 2714 struct stack *st; 2715 struct sbuf sb; 2716 struct proc *p; 2717 u_int namelen; 2718 2719 namelen = arg2; 2720 if (namelen != 1) 2721 return (EINVAL); 2722 2723 name = (int *)arg1; 2724 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); 2725 if (error != 0) 2726 return (error); 2727 2728 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); 2729 st = stack_create(M_WAITOK); 2730 2731 lwpidarray = NULL; 2732 PROC_LOCK(p); 2733 do { 2734 if (lwpidarray != NULL) { 2735 free(lwpidarray, M_TEMP); 2736 lwpidarray = NULL; 2737 } 2738 numthreads = p->p_numthreads; 2739 PROC_UNLOCK(p); 2740 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, 2741 M_WAITOK | M_ZERO); 2742 PROC_LOCK(p); 2743 } while (numthreads < p->p_numthreads); 2744 2745 /* 2746 * XXXRW: During the below loop, execve(2) and countless other sorts 2747 * of changes could have taken place. Should we check to see if the 2748 * vmspace has been replaced, or the like, in order to prevent 2749 * giving a snapshot that spans, say, execve(2), with some threads 2750 * before and some after? Among other things, the credentials could 2751 * have changed, in which case the right to extract debug info might 2752 * no longer be assured. 2753 */ 2754 i = 0; 2755 FOREACH_THREAD_IN_PROC(p, td) { 2756 KASSERT(i < numthreads, 2757 ("sysctl_kern_proc_kstack: numthreads")); 2758 lwpidarray[i] = td->td_tid; 2759 i++; 2760 } 2761 PROC_UNLOCK(p); 2762 numthreads = i; 2763 for (i = 0; i < numthreads; i++) { 2764 td = tdfind(lwpidarray[i], p->p_pid); 2765 if (td == NULL) { 2766 continue; 2767 } 2768 bzero(kkstp, sizeof(*kkstp)); 2769 (void)sbuf_new(&sb, kkstp->kkst_trace, 2770 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); 2771 thread_lock(td); 2772 kkstp->kkst_tid = td->td_tid; 2773 if (TD_IS_SWAPPED(td)) 2774 kkstp->kkst_state = KKST_STATE_SWAPPED; 2775 else if (stack_save_td(st, td) == 0) 2776 kkstp->kkst_state = KKST_STATE_STACKOK; 2777 else 2778 kkstp->kkst_state = KKST_STATE_RUNNING; 2779 thread_unlock(td); 2780 PROC_UNLOCK(p); 2781 stack_sbuf_print(&sb, st); 2782 sbuf_finish(&sb); 2783 sbuf_delete(&sb); 2784 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); 2785 if (error) 2786 break; 2787 } 2788 PRELE(p); 2789 if (lwpidarray != NULL) 2790 free(lwpidarray, M_TEMP); 2791 stack_destroy(st); 2792 free(kkstp, M_TEMP); 2793 return (error); 2794 } 2795 #endif 2796 2797 /* 2798 * This sysctl allows a process to retrieve the full list of groups from 2799 * itself or another process. 2800 */ 2801 static int 2802 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) 2803 { 2804 pid_t *pidp = (pid_t *)arg1; 2805 unsigned int arglen = arg2; 2806 struct proc *p; 2807 struct ucred *cred; 2808 int error; 2809 2810 if (arglen != 1) 2811 return (EINVAL); 2812 if (*pidp == -1) { /* -1 means this process */ 2813 p = req->td->td_proc; 2814 PROC_LOCK(p); 2815 } else { 2816 error = pget(*pidp, PGET_CANSEE, &p); 2817 if (error != 0) 2818 return (error); 2819 } 2820 2821 cred = crhold(p->p_ucred); 2822 PROC_UNLOCK(p); 2823 2824 error = SYSCTL_OUT(req, cred->cr_groups, 2825 cred->cr_ngroups * sizeof(gid_t)); 2826 crfree(cred); 2827 return (error); 2828 } 2829 2830 /* 2831 * This sysctl allows a process to retrieve or/and set the resource limit for 2832 * another process. 2833 */ 2834 static int 2835 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) 2836 { 2837 int *name = (int *)arg1; 2838 u_int namelen = arg2; 2839 struct rlimit rlim; 2840 struct proc *p; 2841 u_int which; 2842 int flags, error; 2843 2844 if (namelen != 2) 2845 return (EINVAL); 2846 2847 which = (u_int)name[1]; 2848 if (which >= RLIM_NLIMITS) 2849 return (EINVAL); 2850 2851 if (req->newptr != NULL && req->newlen != sizeof(rlim)) 2852 return (EINVAL); 2853 2854 flags = PGET_HOLD | PGET_NOTWEXIT; 2855 if (req->newptr != NULL) 2856 flags |= PGET_CANDEBUG; 2857 else 2858 flags |= PGET_CANSEE; 2859 error = pget((pid_t)name[0], flags, &p); 2860 if (error != 0) 2861 return (error); 2862 2863 /* 2864 * Retrieve limit. 2865 */ 2866 if (req->oldptr != NULL) { 2867 PROC_LOCK(p); 2868 lim_rlimit_proc(p, which, &rlim); 2869 PROC_UNLOCK(p); 2870 } 2871 error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); 2872 if (error != 0) 2873 goto errout; 2874 2875 /* 2876 * Set limit. 2877 */ 2878 if (req->newptr != NULL) { 2879 error = SYSCTL_IN(req, &rlim, sizeof(rlim)); 2880 if (error == 0) 2881 error = kern_proc_setrlimit(curthread, p, which, &rlim); 2882 } 2883 2884 errout: 2885 PRELE(p); 2886 return (error); 2887 } 2888 2889 /* 2890 * This sysctl allows a process to retrieve ps_strings structure location of 2891 * another process. 2892 */ 2893 static int 2894 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) 2895 { 2896 int *name = (int *)arg1; 2897 u_int namelen = arg2; 2898 struct proc *p; 2899 vm_offset_t ps_strings; 2900 int error; 2901 #ifdef COMPAT_FREEBSD32 2902 uint32_t ps_strings32; 2903 #endif 2904 2905 if (namelen != 1) 2906 return (EINVAL); 2907 2908 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2909 if (error != 0) 2910 return (error); 2911 #ifdef COMPAT_FREEBSD32 2912 if ((req->flags & SCTL_MASK32) != 0) { 2913 /* 2914 * We return 0 if the 32 bit emulation request is for a 64 bit 2915 * process. 2916 */ 2917 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? 2918 PTROUT(p->p_sysent->sv_psstrings) : 0; 2919 PROC_UNLOCK(p); 2920 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); 2921 return (error); 2922 } 2923 #endif 2924 ps_strings = p->p_sysent->sv_psstrings; 2925 PROC_UNLOCK(p); 2926 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); 2927 return (error); 2928 } 2929 2930 /* 2931 * This sysctl allows a process to retrieve umask of another process. 2932 */ 2933 static int 2934 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) 2935 { 2936 int *name = (int *)arg1; 2937 u_int namelen = arg2; 2938 struct proc *p; 2939 int error; 2940 u_short cmask; 2941 pid_t pid; 2942 2943 if (namelen != 1) 2944 return (EINVAL); 2945 2946 pid = (pid_t)name[0]; 2947 p = curproc; 2948 if (pid == p->p_pid || pid == 0) { 2949 cmask = p->p_pd->pd_cmask; 2950 goto out; 2951 } 2952 2953 error = pget(pid, PGET_WANTREAD, &p); 2954 if (error != 0) 2955 return (error); 2956 2957 cmask = p->p_pd->pd_cmask; 2958 PRELE(p); 2959 out: 2960 error = SYSCTL_OUT(req, &cmask, sizeof(cmask)); 2961 return (error); 2962 } 2963 2964 /* 2965 * This sysctl allows a process to set and retrieve binary osreldate of 2966 * another process. 2967 */ 2968 static int 2969 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) 2970 { 2971 int *name = (int *)arg1; 2972 u_int namelen = arg2; 2973 struct proc *p; 2974 int flags, error, osrel; 2975 2976 if (namelen != 1) 2977 return (EINVAL); 2978 2979 if (req->newptr != NULL && req->newlen != sizeof(osrel)) 2980 return (EINVAL); 2981 2982 flags = PGET_HOLD | PGET_NOTWEXIT; 2983 if (req->newptr != NULL) 2984 flags |= PGET_CANDEBUG; 2985 else 2986 flags |= PGET_CANSEE; 2987 error = pget((pid_t)name[0], flags, &p); 2988 if (error != 0) 2989 return (error); 2990 2991 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); 2992 if (error != 0) 2993 goto errout; 2994 2995 if (req->newptr != NULL) { 2996 error = SYSCTL_IN(req, &osrel, sizeof(osrel)); 2997 if (error != 0) 2998 goto errout; 2999 if (osrel < 0) { 3000 error = EINVAL; 3001 goto errout; 3002 } 3003 p->p_osrel = osrel; 3004 } 3005 errout: 3006 PRELE(p); 3007 return (error); 3008 } 3009 3010 static int 3011 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) 3012 { 3013 int *name = (int *)arg1; 3014 u_int namelen = arg2; 3015 struct proc *p; 3016 struct kinfo_sigtramp kst; 3017 const struct sysentvec *sv; 3018 int error; 3019 #ifdef COMPAT_FREEBSD32 3020 struct kinfo_sigtramp32 kst32; 3021 #endif 3022 3023 if (namelen != 1) 3024 return (EINVAL); 3025 3026 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 3027 if (error != 0) 3028 return (error); 3029 sv = p->p_sysent; 3030 #ifdef COMPAT_FREEBSD32 3031 if ((req->flags & SCTL_MASK32) != 0) { 3032 bzero(&kst32, sizeof(kst32)); 3033 if (SV_PROC_FLAG(p, SV_ILP32)) { 3034 if (sv->sv_sigcode_base != 0) { 3035 kst32.ksigtramp_start = sv->sv_sigcode_base; 3036 kst32.ksigtramp_end = sv->sv_sigcode_base + 3037 *sv->sv_szsigcode; 3038 } else { 3039 kst32.ksigtramp_start = sv->sv_psstrings - 3040 *sv->sv_szsigcode; 3041 kst32.ksigtramp_end = sv->sv_psstrings; 3042 } 3043 } 3044 PROC_UNLOCK(p); 3045 error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); 3046 return (error); 3047 } 3048 #endif 3049 bzero(&kst, sizeof(kst)); 3050 if (sv->sv_sigcode_base != 0) { 3051 kst.ksigtramp_start = (char *)sv->sv_sigcode_base; 3052 kst.ksigtramp_end = (char *)sv->sv_sigcode_base + 3053 *sv->sv_szsigcode; 3054 } else { 3055 kst.ksigtramp_start = (char *)sv->sv_psstrings - 3056 *sv->sv_szsigcode; 3057 kst.ksigtramp_end = (char *)sv->sv_psstrings; 3058 } 3059 PROC_UNLOCK(p); 3060 error = SYSCTL_OUT(req, &kst, sizeof(kst)); 3061 return (error); 3062 } 3063 3064 static int 3065 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS) 3066 { 3067 int *name = (int *)arg1; 3068 u_int namelen = arg2; 3069 pid_t pid; 3070 struct proc *p; 3071 struct thread *td1; 3072 uintptr_t addr; 3073 #ifdef COMPAT_FREEBSD32 3074 uint32_t addr32; 3075 #endif 3076 int error; 3077 3078 if (namelen != 1 || req->newptr != NULL) 3079 return (EINVAL); 3080 3081 pid = (pid_t)name[0]; 3082 error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p); 3083 if (error != 0) 3084 return (error); 3085 3086 PROC_LOCK(p); 3087 #ifdef COMPAT_FREEBSD32 3088 if (SV_CURPROC_FLAG(SV_ILP32)) { 3089 if (!SV_PROC_FLAG(p, SV_ILP32)) { 3090 error = EINVAL; 3091 goto errlocked; 3092 } 3093 } 3094 #endif 3095 if (pid <= PID_MAX) { 3096 td1 = FIRST_THREAD_IN_PROC(p); 3097 } else { 3098 FOREACH_THREAD_IN_PROC(p, td1) { 3099 if (td1->td_tid == pid) 3100 break; 3101 } 3102 } 3103 if (td1 == NULL) { 3104 error = ESRCH; 3105 goto errlocked; 3106 } 3107 /* 3108 * The access to the private thread flags. It is fine as far 3109 * as no out-of-thin-air values are read from td_pflags, and 3110 * usermode read of the td_sigblock_ptr is racy inherently, 3111 * since target process might have already changed it 3112 * meantime. 3113 */ 3114 if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0) 3115 addr = (uintptr_t)td1->td_sigblock_ptr; 3116 else 3117 error = ENOTTY; 3118 3119 errlocked: 3120 _PRELE(p); 3121 PROC_UNLOCK(p); 3122 if (error != 0) 3123 return (error); 3124 3125 #ifdef COMPAT_FREEBSD32 3126 if (SV_CURPROC_FLAG(SV_ILP32)) { 3127 addr32 = addr; 3128 error = SYSCTL_OUT(req, &addr32, sizeof(addr32)); 3129 } else 3130 #endif 3131 error = SYSCTL_OUT(req, &addr, sizeof(addr)); 3132 return (error); 3133 } 3134 3135 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 3136 "Process table"); 3137 3138 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| 3139 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", 3140 "Return entire process table"); 3141 3142 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3143 sysctl_kern_proc, "Process table"); 3144 3145 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, 3146 sysctl_kern_proc, "Process table"); 3147 3148 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3149 sysctl_kern_proc, "Process table"); 3150 3151 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | 3152 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3153 3154 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 3155 sysctl_kern_proc, "Process table"); 3156 3157 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3158 sysctl_kern_proc, "Process table"); 3159 3160 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3161 sysctl_kern_proc, "Process table"); 3162 3163 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3164 sysctl_kern_proc, "Process table"); 3165 3166 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 3167 sysctl_kern_proc, "Return process table, no threads"); 3168 3169 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, 3170 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, 3171 sysctl_kern_proc_args, "Process argument list"); 3172 3173 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, 3174 sysctl_kern_proc_env, "Process environment"); 3175 3176 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | 3177 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); 3178 3179 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | 3180 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); 3181 3182 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | 3183 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, 3184 "Process syscall vector name (ABI type)"); 3185 3186 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, 3187 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3188 3189 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, 3190 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3191 3192 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, 3193 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3194 3195 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), 3196 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3197 3198 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, 3199 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3200 3201 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, 3202 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3203 3204 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, 3205 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3206 3207 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, 3208 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3209 3210 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, 3211 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, 3212 "Return process table, including threads"); 3213 3214 #ifdef COMPAT_FREEBSD7 3215 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | 3216 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); 3217 #endif 3218 3219 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | 3220 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); 3221 3222 #if defined(STACK) || defined(DDB) 3223 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | 3224 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); 3225 #endif 3226 3227 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | 3228 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); 3229 3230 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | 3231 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, 3232 "Process resource limits"); 3233 3234 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | 3235 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, 3236 "Process ps_strings location"); 3237 3238 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | 3239 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); 3240 3241 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | 3242 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, 3243 "Process binary osreldate"); 3244 3245 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | 3246 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, 3247 "Process signal trampoline location"); 3248 3249 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD | 3250 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk, 3251 "Thread sigfastblock address"); 3252 3253 int allproc_gen; 3254 3255 /* 3256 * stop_all_proc() purpose is to stop all process which have usermode, 3257 * except current process for obvious reasons. This makes it somewhat 3258 * unreliable when invoked from multithreaded process. The service 3259 * must not be user-callable anyway. 3260 */ 3261 void 3262 stop_all_proc(void) 3263 { 3264 struct proc *cp, *p; 3265 int r, gen; 3266 bool restart, seen_stopped, seen_exiting, stopped_some; 3267 3268 cp = curproc; 3269 allproc_loop: 3270 sx_xlock(&allproc_lock); 3271 gen = allproc_gen; 3272 seen_exiting = seen_stopped = stopped_some = restart = false; 3273 LIST_REMOVE(cp, p_list); 3274 LIST_INSERT_HEAD(&allproc, cp, p_list); 3275 for (;;) { 3276 p = LIST_NEXT(cp, p_list); 3277 if (p == NULL) 3278 break; 3279 LIST_REMOVE(cp, p_list); 3280 LIST_INSERT_AFTER(p, cp, p_list); 3281 PROC_LOCK(p); 3282 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) { 3283 PROC_UNLOCK(p); 3284 continue; 3285 } 3286 if ((p->p_flag & P_WEXIT) != 0) { 3287 seen_exiting = true; 3288 PROC_UNLOCK(p); 3289 continue; 3290 } 3291 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 3292 /* 3293 * Stopped processes are tolerated when there 3294 * are no other processes which might continue 3295 * them. P_STOPPED_SINGLE but not 3296 * P_TOTAL_STOP process still has at least one 3297 * thread running. 3298 */ 3299 seen_stopped = true; 3300 PROC_UNLOCK(p); 3301 continue; 3302 } 3303 sx_xunlock(&allproc_lock); 3304 _PHOLD(p); 3305 r = thread_single(p, SINGLE_ALLPROC); 3306 if (r != 0) 3307 restart = true; 3308 else 3309 stopped_some = true; 3310 _PRELE(p); 3311 PROC_UNLOCK(p); 3312 sx_xlock(&allproc_lock); 3313 } 3314 /* Catch forked children we did not see in iteration. */ 3315 if (gen != allproc_gen) 3316 restart = true; 3317 sx_xunlock(&allproc_lock); 3318 if (restart || stopped_some || seen_exiting || seen_stopped) { 3319 kern_yield(PRI_USER); 3320 goto allproc_loop; 3321 } 3322 } 3323 3324 void 3325 resume_all_proc(void) 3326 { 3327 struct proc *cp, *p; 3328 3329 cp = curproc; 3330 sx_xlock(&allproc_lock); 3331 again: 3332 LIST_REMOVE(cp, p_list); 3333 LIST_INSERT_HEAD(&allproc, cp, p_list); 3334 for (;;) { 3335 p = LIST_NEXT(cp, p_list); 3336 if (p == NULL) 3337 break; 3338 LIST_REMOVE(cp, p_list); 3339 LIST_INSERT_AFTER(p, cp, p_list); 3340 PROC_LOCK(p); 3341 if ((p->p_flag & P_TOTAL_STOP) != 0) { 3342 sx_xunlock(&allproc_lock); 3343 _PHOLD(p); 3344 thread_single_end(p, SINGLE_ALLPROC); 3345 _PRELE(p); 3346 PROC_UNLOCK(p); 3347 sx_xlock(&allproc_lock); 3348 } else { 3349 PROC_UNLOCK(p); 3350 } 3351 } 3352 /* Did the loop above missed any stopped process ? */ 3353 FOREACH_PROC_IN_SYSTEM(p) { 3354 /* No need for proc lock. */ 3355 if ((p->p_flag & P_TOTAL_STOP) != 0) 3356 goto again; 3357 } 3358 sx_xunlock(&allproc_lock); 3359 } 3360 3361 /* #define TOTAL_STOP_DEBUG 1 */ 3362 #ifdef TOTAL_STOP_DEBUG 3363 volatile static int ap_resume; 3364 #include <sys/mount.h> 3365 3366 static int 3367 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) 3368 { 3369 int error, val; 3370 3371 val = 0; 3372 ap_resume = 0; 3373 error = sysctl_handle_int(oidp, &val, 0, req); 3374 if (error != 0 || req->newptr == NULL) 3375 return (error); 3376 if (val != 0) { 3377 stop_all_proc(); 3378 syncer_suspend(); 3379 while (ap_resume == 0) 3380 ; 3381 syncer_resume(); 3382 resume_all_proc(); 3383 } 3384 return (0); 3385 } 3386 3387 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | 3388 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, 3389 sysctl_debug_stop_all_proc, "I", 3390 ""); 3391 #endif 3392