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 32 #include <sys/cdefs.h> 33 #include "opt_ddb.h" 34 #include "opt_ktrace.h" 35 #include "opt_kstack_pages.h" 36 #include "opt_stack.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/bitstring.h> 41 #include <sys/elf.h> 42 #include <sys/eventhandler.h> 43 #include <sys/exec.h> 44 #include <sys/fcntl.h> 45 #include <sys/jail.h> 46 #include <sys/kernel.h> 47 #include <sys/limits.h> 48 #include <sys/lock.h> 49 #include <sys/loginclass.h> 50 #include <sys/malloc.h> 51 #include <sys/mman.h> 52 #include <sys/mount.h> 53 #include <sys/mutex.h> 54 #include <sys/namei.h> 55 #include <sys/proc.h> 56 #include <sys/ptrace.h> 57 #include <sys/refcount.h> 58 #include <sys/resourcevar.h> 59 #include <sys/rwlock.h> 60 #include <sys/sbuf.h> 61 #include <sys/sysent.h> 62 #include <sys/sched.h> 63 #include <sys/smp.h> 64 #include <sys/stack.h> 65 #include <sys/stat.h> 66 #include <sys/dtrace_bsd.h> 67 #include <sys/sysctl.h> 68 #include <sys/filedesc.h> 69 #include <sys/tty.h> 70 #include <sys/signalvar.h> 71 #include <sys/sdt.h> 72 #include <sys/sx.h> 73 #include <sys/user.h> 74 #include <sys/vnode.h> 75 #include <sys/wait.h> 76 #ifdef KTRACE 77 #include <sys/ktrace.h> 78 #endif 79 80 #ifdef DDB 81 #include <ddb/ddb.h> 82 #endif 83 84 #include <vm/vm.h> 85 #include <vm/vm_param.h> 86 #include <vm/vm_extern.h> 87 #include <vm/pmap.h> 88 #include <vm/vm_map.h> 89 #include <vm/vm_object.h> 90 #include <vm/vm_page.h> 91 #include <vm/uma.h> 92 93 #include <fs/devfs/devfs.h> 94 95 #ifdef COMPAT_FREEBSD32 96 #include <compat/freebsd32/freebsd32.h> 97 #include <compat/freebsd32/freebsd32_util.h> 98 #endif 99 100 SDT_PROVIDER_DEFINE(proc); 101 102 MALLOC_DEFINE(M_SESSION, "session", "session header"); 103 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); 104 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 105 106 static void doenterpgrp(struct proc *, struct pgrp *); 107 static void orphanpg(struct pgrp *pg); 108 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp); 109 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp); 110 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, 111 int preferthread); 112 static void pgdelete(struct pgrp *); 113 static int pgrp_init(void *mem, int size, int flags); 114 static int proc_ctor(void *mem, int size, void *arg, int flags); 115 static void proc_dtor(void *mem, int size, void *arg); 116 static int proc_init(void *mem, int size, int flags); 117 static void proc_fini(void *mem, int size); 118 static void pargs_free(struct pargs *pa); 119 120 /* 121 * Other process lists 122 */ 123 struct pidhashhead *pidhashtbl = NULL; 124 struct sx *pidhashtbl_lock; 125 u_long pidhash; 126 u_long pidhashlock; 127 struct pgrphashhead *pgrphashtbl; 128 u_long pgrphash; 129 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc); 130 struct sx __exclusive_cache_line allproc_lock; 131 struct sx __exclusive_cache_line proctree_lock; 132 struct mtx __exclusive_cache_line ppeers_lock; 133 struct mtx __exclusive_cache_line procid_lock; 134 uma_zone_t proc_zone; 135 uma_zone_t pgrp_zone; 136 137 /* 138 * The offset of various fields in struct proc and struct thread. 139 * These are used by kernel debuggers to enumerate kernel threads and 140 * processes. 141 */ 142 const int proc_off_p_pid = offsetof(struct proc, p_pid); 143 const int proc_off_p_comm = offsetof(struct proc, p_comm); 144 const int proc_off_p_list = offsetof(struct proc, p_list); 145 const int proc_off_p_hash = offsetof(struct proc, p_hash); 146 const int proc_off_p_threads = offsetof(struct proc, p_threads); 147 const int thread_off_td_tid = offsetof(struct thread, td_tid); 148 const int thread_off_td_name = offsetof(struct thread, td_name); 149 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu); 150 const int thread_off_td_pcb = offsetof(struct thread, td_pcb); 151 const int thread_off_td_plist = offsetof(struct thread, td_plist); 152 153 EVENTHANDLER_LIST_DEFINE(process_ctor); 154 EVENTHANDLER_LIST_DEFINE(process_dtor); 155 EVENTHANDLER_LIST_DEFINE(process_init); 156 EVENTHANDLER_LIST_DEFINE(process_fini); 157 EVENTHANDLER_LIST_DEFINE(process_exit); 158 EVENTHANDLER_LIST_DEFINE(process_fork); 159 EVENTHANDLER_LIST_DEFINE(process_exec); 160 161 int kstack_pages = KSTACK_PAGES; 162 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, 163 &kstack_pages, 0, 164 "Kernel stack size in pages"); 165 static int vmmap_skip_res_cnt = 0; 166 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW, 167 &vmmap_skip_res_cnt, 0, 168 "Skip calculation of the pages resident count in kern.proc.vmmap"); 169 170 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); 171 #ifdef COMPAT_FREEBSD32 172 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE); 173 #endif 174 175 /* 176 * Initialize global process hashing structures. 177 */ 178 void 179 procinit(void) 180 { 181 u_long i; 182 183 sx_init(&allproc_lock, "allproc"); 184 sx_init(&proctree_lock, "proctree"); 185 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF); 186 mtx_init(&procid_lock, "procid", NULL, MTX_DEF); 187 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); 188 pidhashlock = (pidhash + 1) / 64; 189 if (pidhashlock > 0) 190 pidhashlock--; 191 pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1), 192 M_PROC, M_WAITOK | M_ZERO); 193 for (i = 0; i < pidhashlock + 1; i++) 194 sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK); 195 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); 196 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(), 197 proc_ctor, proc_dtor, proc_init, proc_fini, 198 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 199 pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL, 200 pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 201 uihashinit(); 202 } 203 204 /* 205 * Prepare a proc for use. 206 */ 207 static int 208 proc_ctor(void *mem, int size, void *arg, int flags) 209 { 210 struct proc *p; 211 struct thread *td; 212 213 p = (struct proc *)mem; 214 #ifdef KDTRACE_HOOKS 215 kdtrace_proc_ctor(p); 216 #endif 217 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p); 218 td = FIRST_THREAD_IN_PROC(p); 219 if (td != NULL) { 220 /* Make sure all thread constructors are executed */ 221 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td); 222 } 223 return (0); 224 } 225 226 /* 227 * Reclaim a proc after use. 228 */ 229 static void 230 proc_dtor(void *mem, int size, void *arg) 231 { 232 struct proc *p; 233 struct thread *td; 234 235 /* INVARIANTS checks go here */ 236 p = (struct proc *)mem; 237 td = FIRST_THREAD_IN_PROC(p); 238 if (td != NULL) { 239 #ifdef INVARIANTS 240 KASSERT((p->p_numthreads == 1), 241 ("bad number of threads in exiting process")); 242 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr")); 243 #endif 244 /* Free all OSD associated to this thread. */ 245 osd_thread_exit(td); 246 ast_kclear(td); 247 248 /* Make sure all thread destructors are executed */ 249 EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td); 250 } 251 EVENTHANDLER_DIRECT_INVOKE(process_dtor, p); 252 #ifdef KDTRACE_HOOKS 253 kdtrace_proc_dtor(p); 254 #endif 255 if (p->p_ksi != NULL) 256 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue")); 257 } 258 259 /* 260 * Initialize type-stable parts of a proc (when newly created). 261 */ 262 static int 263 proc_init(void *mem, int size, int flags) 264 { 265 struct proc *p; 266 267 p = (struct proc *)mem; 268 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW); 269 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW); 270 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW); 271 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW); 272 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW); 273 cv_init(&p->p_pwait, "ppwait"); 274 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 275 EVENTHANDLER_DIRECT_INVOKE(process_init, p); 276 p->p_stats = pstats_alloc(); 277 p->p_pgrp = NULL; 278 TAILQ_INIT(&p->p_kqtim_stop); 279 return (0); 280 } 281 282 /* 283 * UMA should ensure that this function is never called. 284 * Freeing a proc structure would violate type stability. 285 */ 286 static void 287 proc_fini(void *mem, int size) 288 { 289 #ifdef notnow 290 struct proc *p; 291 292 p = (struct proc *)mem; 293 EVENTHANDLER_DIRECT_INVOKE(process_fini, p); 294 pstats_free(p->p_stats); 295 thread_free(FIRST_THREAD_IN_PROC(p)); 296 mtx_destroy(&p->p_mtx); 297 if (p->p_ksi != NULL) 298 ksiginfo_free(p->p_ksi); 299 #else 300 panic("proc reclaimed"); 301 #endif 302 } 303 304 static int 305 pgrp_init(void *mem, int size, int flags) 306 { 307 struct pgrp *pg; 308 309 pg = mem; 310 mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 311 sx_init(&pg->pg_killsx, "killpg racer"); 312 return (0); 313 } 314 315 /* 316 * PID space management. 317 * 318 * These bitmaps are used by fork_findpid. 319 */ 320 bitstr_t bit_decl(proc_id_pidmap, PID_MAX); 321 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX); 322 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX); 323 bitstr_t bit_decl(proc_id_reapmap, PID_MAX); 324 325 static bitstr_t *proc_id_array[] = { 326 proc_id_pidmap, 327 proc_id_grpidmap, 328 proc_id_sessidmap, 329 proc_id_reapmap, 330 }; 331 332 void 333 proc_id_set(int type, pid_t id) 334 { 335 336 KASSERT(type >= 0 && type < nitems(proc_id_array), 337 ("invalid type %d\n", type)); 338 mtx_lock(&procid_lock); 339 KASSERT(bit_test(proc_id_array[type], id) == 0, 340 ("bit %d already set in %d\n", id, type)); 341 bit_set(proc_id_array[type], id); 342 mtx_unlock(&procid_lock); 343 } 344 345 void 346 proc_id_set_cond(int type, pid_t id) 347 { 348 349 KASSERT(type >= 0 && type < nitems(proc_id_array), 350 ("invalid type %d\n", type)); 351 if (bit_test(proc_id_array[type], id)) 352 return; 353 mtx_lock(&procid_lock); 354 bit_set(proc_id_array[type], id); 355 mtx_unlock(&procid_lock); 356 } 357 358 void 359 proc_id_clear(int type, pid_t id) 360 { 361 362 KASSERT(type >= 0 && type < nitems(proc_id_array), 363 ("invalid type %d\n", type)); 364 mtx_lock(&procid_lock); 365 KASSERT(bit_test(proc_id_array[type], id) != 0, 366 ("bit %d not set in %d\n", id, type)); 367 bit_clear(proc_id_array[type], id); 368 mtx_unlock(&procid_lock); 369 } 370 371 /* 372 * Is p an inferior of the current process? 373 */ 374 int 375 inferior(struct proc *p) 376 { 377 378 sx_assert(&proctree_lock, SX_LOCKED); 379 PROC_LOCK_ASSERT(p, MA_OWNED); 380 for (; p != curproc; p = proc_realparent(p)) { 381 if (p->p_pid == 0) 382 return (0); 383 } 384 return (1); 385 } 386 387 /* 388 * Shared lock all the pid hash lists. 389 */ 390 void 391 pidhash_slockall(void) 392 { 393 u_long i; 394 395 for (i = 0; i < pidhashlock + 1; i++) 396 sx_slock(&pidhashtbl_lock[i]); 397 } 398 399 /* 400 * Shared unlock all the pid hash lists. 401 */ 402 void 403 pidhash_sunlockall(void) 404 { 405 u_long i; 406 407 for (i = 0; i < pidhashlock + 1; i++) 408 sx_sunlock(&pidhashtbl_lock[i]); 409 } 410 411 /* 412 * Similar to pfind_any(), this function finds zombies. 413 */ 414 struct proc * 415 pfind_any_locked(pid_t pid) 416 { 417 struct proc *p; 418 419 sx_assert(PIDHASHLOCK(pid), SX_LOCKED); 420 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 421 if (p->p_pid == pid) { 422 PROC_LOCK(p); 423 if (p->p_state == PRS_NEW) { 424 PROC_UNLOCK(p); 425 p = NULL; 426 } 427 break; 428 } 429 } 430 return (p); 431 } 432 433 /* 434 * Locate a process by number. 435 * 436 * By not returning processes in the PRS_NEW state, we allow callers to avoid 437 * testing for that condition to avoid dereferencing p_ucred, et al. 438 */ 439 static __always_inline struct proc * 440 _pfind(pid_t pid, bool zombie) 441 { 442 struct proc *p; 443 444 p = curproc; 445 if (p->p_pid == pid) { 446 PROC_LOCK(p); 447 return (p); 448 } 449 sx_slock(PIDHASHLOCK(pid)); 450 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 451 if (p->p_pid == pid) { 452 PROC_LOCK(p); 453 if (p->p_state == PRS_NEW || 454 (!zombie && p->p_state == PRS_ZOMBIE)) { 455 PROC_UNLOCK(p); 456 p = NULL; 457 } 458 break; 459 } 460 } 461 sx_sunlock(PIDHASHLOCK(pid)); 462 return (p); 463 } 464 465 struct proc * 466 pfind(pid_t pid) 467 { 468 469 return (_pfind(pid, false)); 470 } 471 472 /* 473 * Same as pfind but allow zombies. 474 */ 475 struct proc * 476 pfind_any(pid_t pid) 477 { 478 479 return (_pfind(pid, true)); 480 } 481 482 /* 483 * Locate a process group by number. 484 * The caller must hold proctree_lock. 485 */ 486 struct pgrp * 487 pgfind(pid_t pgid) 488 { 489 struct pgrp *pgrp; 490 491 sx_assert(&proctree_lock, SX_LOCKED); 492 493 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { 494 if (pgrp->pg_id == pgid) { 495 PGRP_LOCK(pgrp); 496 return (pgrp); 497 } 498 } 499 return (NULL); 500 } 501 502 /* 503 * Locate process and do additional manipulations, depending on flags. 504 */ 505 int 506 pget(pid_t pid, int flags, struct proc **pp) 507 { 508 struct proc *p; 509 struct thread *td1; 510 int error; 511 512 p = curproc; 513 if (p->p_pid == pid) { 514 PROC_LOCK(p); 515 } else { 516 p = NULL; 517 if (pid <= PID_MAX) { 518 if ((flags & PGET_NOTWEXIT) == 0) 519 p = pfind_any(pid); 520 else 521 p = pfind(pid); 522 } else if ((flags & PGET_NOTID) == 0) { 523 td1 = tdfind(pid, -1); 524 if (td1 != NULL) 525 p = td1->td_proc; 526 } 527 if (p == NULL) 528 return (ESRCH); 529 if ((flags & PGET_CANSEE) != 0) { 530 error = p_cansee(curthread, p); 531 if (error != 0) 532 goto errout; 533 } 534 } 535 if ((flags & PGET_CANDEBUG) != 0) { 536 error = p_candebug(curthread, p); 537 if (error != 0) 538 goto errout; 539 } 540 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) { 541 error = EPERM; 542 goto errout; 543 } 544 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) { 545 error = ESRCH; 546 goto errout; 547 } 548 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) { 549 /* 550 * XXXRW: Not clear ESRCH is the right error during proc 551 * execve(). 552 */ 553 error = ESRCH; 554 goto errout; 555 } 556 if ((flags & PGET_HOLD) != 0) { 557 _PHOLD(p); 558 PROC_UNLOCK(p); 559 } 560 *pp = p; 561 return (0); 562 errout: 563 PROC_UNLOCK(p); 564 return (error); 565 } 566 567 /* 568 * Create a new process group. 569 * pgid must be equal to the pid of p. 570 * Begin a new session if required. 571 */ 572 int 573 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess) 574 { 575 struct pgrp *old_pgrp; 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 old_pgrp = p->p_pgrp; 588 if (!sx_try_xlock(&old_pgrp->pg_killsx)) { 589 sx_xunlock(&proctree_lock); 590 sx_xlock(&old_pgrp->pg_killsx); 591 sx_xunlock(&old_pgrp->pg_killsx); 592 return (ERESTART); 593 } 594 MPASS(old_pgrp == p->p_pgrp); 595 596 if (sess != NULL) { 597 /* 598 * new session 599 */ 600 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); 601 PROC_LOCK(p); 602 p->p_flag &= ~P_CONTROLT; 603 PROC_UNLOCK(p); 604 PGRP_LOCK(pgrp); 605 sess->s_leader = p; 606 sess->s_sid = p->p_pid; 607 proc_id_set(PROC_ID_SESSION, p->p_pid); 608 refcount_init(&sess->s_count, 1); 609 sess->s_ttyvp = NULL; 610 sess->s_ttydp = NULL; 611 sess->s_ttyp = NULL; 612 bcopy(p->p_session->s_login, sess->s_login, 613 sizeof(sess->s_login)); 614 pgrp->pg_session = sess; 615 KASSERT(p == curproc, 616 ("enterpgrp: mksession and p != curproc")); 617 } else { 618 pgrp->pg_session = p->p_session; 619 sess_hold(pgrp->pg_session); 620 PGRP_LOCK(pgrp); 621 } 622 pgrp->pg_id = pgid; 623 proc_id_set(PROC_ID_GROUP, p->p_pid); 624 LIST_INIT(&pgrp->pg_members); 625 pgrp->pg_flags = 0; 626 627 /* 628 * As we have an exclusive lock of proctree_lock, 629 * this should not deadlock. 630 */ 631 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); 632 SLIST_INIT(&pgrp->pg_sigiolst); 633 PGRP_UNLOCK(pgrp); 634 635 doenterpgrp(p, pgrp); 636 637 sx_xunlock(&old_pgrp->pg_killsx); 638 return (0); 639 } 640 641 /* 642 * Move p to an existing process group 643 */ 644 int 645 enterthispgrp(struct proc *p, struct pgrp *pgrp) 646 { 647 struct pgrp *old_pgrp; 648 649 sx_assert(&proctree_lock, SX_XLOCKED); 650 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 651 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 652 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 653 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 654 KASSERT(pgrp->pg_session == p->p_session, 655 ("%s: pgrp's session %p, p->p_session %p proc %p\n", 656 __func__, pgrp->pg_session, p->p_session, p)); 657 KASSERT(pgrp != p->p_pgrp, 658 ("%s: p %p belongs to pgrp %p", __func__, p, pgrp)); 659 660 old_pgrp = p->p_pgrp; 661 if (!sx_try_xlock(&old_pgrp->pg_killsx)) { 662 sx_xunlock(&proctree_lock); 663 sx_xlock(&old_pgrp->pg_killsx); 664 sx_xunlock(&old_pgrp->pg_killsx); 665 return (ERESTART); 666 } 667 MPASS(old_pgrp == p->p_pgrp); 668 if (!sx_try_xlock(&pgrp->pg_killsx)) { 669 sx_xunlock(&old_pgrp->pg_killsx); 670 sx_xunlock(&proctree_lock); 671 sx_xlock(&pgrp->pg_killsx); 672 sx_xunlock(&pgrp->pg_killsx); 673 return (ERESTART); 674 } 675 676 doenterpgrp(p, pgrp); 677 678 sx_xunlock(&pgrp->pg_killsx); 679 sx_xunlock(&old_pgrp->pg_killsx); 680 return (0); 681 } 682 683 /* 684 * If true, any child of q which belongs to group pgrp, qualifies the 685 * process group pgrp as not orphaned. 686 */ 687 static bool 688 isjobproc(struct proc *q, struct pgrp *pgrp) 689 { 690 sx_assert(&proctree_lock, SX_LOCKED); 691 692 return (q->p_pgrp != pgrp && 693 q->p_pgrp->pg_session == pgrp->pg_session); 694 } 695 696 static struct proc * 697 jobc_reaper(struct proc *p) 698 { 699 struct proc *pp; 700 701 sx_assert(&proctree_lock, SA_LOCKED); 702 703 for (pp = p;;) { 704 pp = pp->p_reaper; 705 if (pp->p_reaper == pp || 706 (pp->p_treeflag & P_TREE_GRPEXITED) == 0) 707 return (pp); 708 } 709 } 710 711 static struct proc * 712 jobc_parent(struct proc *p, struct proc *p_exiting) 713 { 714 struct proc *pp; 715 716 sx_assert(&proctree_lock, SA_LOCKED); 717 718 pp = proc_realparent(p); 719 if (pp->p_pptr == NULL || pp == p_exiting || 720 (pp->p_treeflag & P_TREE_GRPEXITED) == 0) 721 return (pp); 722 return (jobc_reaper(pp)); 723 } 724 725 static int 726 pgrp_calc_jobc(struct pgrp *pgrp) 727 { 728 struct proc *q; 729 int cnt; 730 731 #ifdef INVARIANTS 732 if (!mtx_owned(&pgrp->pg_mtx)) 733 sx_assert(&proctree_lock, SA_LOCKED); 734 #endif 735 736 cnt = 0; 737 LIST_FOREACH(q, &pgrp->pg_members, p_pglist) { 738 if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 || 739 q->p_pptr == NULL) 740 continue; 741 if (isjobproc(jobc_parent(q, NULL), pgrp)) 742 cnt++; 743 } 744 return (cnt); 745 } 746 747 /* 748 * Move p to a process group 749 */ 750 static void 751 doenterpgrp(struct proc *p, struct pgrp *pgrp) 752 { 753 struct pgrp *savepgrp; 754 struct proc *pp; 755 756 sx_assert(&proctree_lock, SX_XLOCKED); 757 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 758 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 759 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 760 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 761 762 savepgrp = p->p_pgrp; 763 pp = jobc_parent(p, NULL); 764 765 PGRP_LOCK(pgrp); 766 PGRP_LOCK(savepgrp); 767 if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1) 768 orphanpg(savepgrp); 769 PROC_LOCK(p); 770 LIST_REMOVE(p, p_pglist); 771 p->p_pgrp = pgrp; 772 PROC_UNLOCK(p); 773 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 774 if (isjobproc(pp, pgrp)) 775 pgrp->pg_flags &= ~PGRP_ORPHANED; 776 PGRP_UNLOCK(savepgrp); 777 PGRP_UNLOCK(pgrp); 778 if (LIST_EMPTY(&savepgrp->pg_members)) 779 pgdelete(savepgrp); 780 } 781 782 /* 783 * remove process from process group 784 */ 785 int 786 leavepgrp(struct proc *p) 787 { 788 struct pgrp *savepgrp; 789 790 sx_assert(&proctree_lock, SX_XLOCKED); 791 savepgrp = p->p_pgrp; 792 PGRP_LOCK(savepgrp); 793 PROC_LOCK(p); 794 LIST_REMOVE(p, p_pglist); 795 p->p_pgrp = NULL; 796 PROC_UNLOCK(p); 797 PGRP_UNLOCK(savepgrp); 798 if (LIST_EMPTY(&savepgrp->pg_members)) 799 pgdelete(savepgrp); 800 return (0); 801 } 802 803 /* 804 * delete a process group 805 */ 806 static void 807 pgdelete(struct pgrp *pgrp) 808 { 809 struct session *savesess; 810 struct tty *tp; 811 812 sx_assert(&proctree_lock, SX_XLOCKED); 813 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 814 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 815 816 /* 817 * Reset any sigio structures pointing to us as a result of 818 * F_SETOWN with our pgid. The proctree lock ensures that 819 * new sigio structures will not be added after this point. 820 */ 821 funsetownlst(&pgrp->pg_sigiolst); 822 823 PGRP_LOCK(pgrp); 824 tp = pgrp->pg_session->s_ttyp; 825 LIST_REMOVE(pgrp, pg_hash); 826 savesess = pgrp->pg_session; 827 PGRP_UNLOCK(pgrp); 828 829 /* Remove the reference to the pgrp before deallocating it. */ 830 if (tp != NULL) { 831 tty_lock(tp); 832 tty_rel_pgrp(tp, pgrp); 833 } 834 835 proc_id_clear(PROC_ID_GROUP, pgrp->pg_id); 836 uma_zfree(pgrp_zone, pgrp); 837 sess_release(savesess); 838 } 839 840 841 static void 842 fixjobc_kill(struct proc *p) 843 { 844 struct proc *q; 845 struct pgrp *pgrp; 846 847 sx_assert(&proctree_lock, SX_LOCKED); 848 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 849 pgrp = p->p_pgrp; 850 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 851 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 852 853 /* 854 * p no longer affects process group orphanage for children. 855 * It is marked by the flag because p is only physically 856 * removed from its process group on wait(2). 857 */ 858 MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0); 859 p->p_treeflag |= P_TREE_GRPEXITED; 860 861 /* 862 * Check if exiting p orphans its own group. 863 */ 864 pgrp = p->p_pgrp; 865 if (isjobproc(jobc_parent(p, NULL), pgrp)) { 866 PGRP_LOCK(pgrp); 867 if (pgrp_calc_jobc(pgrp) == 0) 868 orphanpg(pgrp); 869 PGRP_UNLOCK(pgrp); 870 } 871 872 /* 873 * Check this process' children to see whether they qualify 874 * their process groups after reparenting to reaper. 875 */ 876 LIST_FOREACH(q, &p->p_children, p_sibling) { 877 pgrp = q->p_pgrp; 878 PGRP_LOCK(pgrp); 879 if (pgrp_calc_jobc(pgrp) == 0) { 880 /* 881 * We want to handle exactly the children that 882 * has p as realparent. Then, when calculating 883 * jobc_parent for children, we should ignore 884 * P_TREE_GRPEXITED flag already set on p. 885 */ 886 if (jobc_parent(q, p) == p && isjobproc(p, pgrp)) 887 orphanpg(pgrp); 888 } else 889 pgrp->pg_flags &= ~PGRP_ORPHANED; 890 PGRP_UNLOCK(pgrp); 891 } 892 LIST_FOREACH(q, &p->p_orphans, p_orphan) { 893 pgrp = q->p_pgrp; 894 PGRP_LOCK(pgrp); 895 if (pgrp_calc_jobc(pgrp) == 0) { 896 if (isjobproc(p, pgrp)) 897 orphanpg(pgrp); 898 } else 899 pgrp->pg_flags &= ~PGRP_ORPHANED; 900 PGRP_UNLOCK(pgrp); 901 } 902 } 903 904 void 905 killjobc(void) 906 { 907 struct session *sp; 908 struct tty *tp; 909 struct proc *p; 910 struct vnode *ttyvp; 911 912 p = curproc; 913 MPASS(p->p_flag & P_WEXIT); 914 sx_assert(&proctree_lock, SX_LOCKED); 915 916 if (SESS_LEADER(p)) { 917 sp = p->p_session; 918 919 /* 920 * s_ttyp is not zero'd; we use this to indicate that 921 * the session once had a controlling terminal. (for 922 * logging and informational purposes) 923 */ 924 SESS_LOCK(sp); 925 ttyvp = sp->s_ttyvp; 926 tp = sp->s_ttyp; 927 sp->s_ttyvp = NULL; 928 sp->s_ttydp = NULL; 929 sp->s_leader = NULL; 930 SESS_UNLOCK(sp); 931 932 /* 933 * Signal foreground pgrp and revoke access to 934 * controlling terminal if it has not been revoked 935 * already. 936 * 937 * Because the TTY may have been revoked in the mean 938 * time and could already have a new session associated 939 * with it, make sure we don't send a SIGHUP to a 940 * foreground process group that does not belong to this 941 * session. 942 */ 943 944 if (tp != NULL) { 945 tty_lock(tp); 946 if (tp->t_session == sp) 947 tty_signal_pgrp(tp, SIGHUP); 948 tty_unlock(tp); 949 } 950 951 if (ttyvp != NULL) { 952 sx_xunlock(&proctree_lock); 953 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) { 954 VOP_REVOKE(ttyvp, REVOKEALL); 955 VOP_UNLOCK(ttyvp); 956 } 957 devfs_ctty_unref(ttyvp); 958 sx_xlock(&proctree_lock); 959 } 960 } 961 fixjobc_kill(p); 962 } 963 964 /* 965 * A process group has become orphaned, mark it as such for signal 966 * delivery code. If there are any stopped processes in the group, 967 * hang-up all process in that group. 968 */ 969 static void 970 orphanpg(struct pgrp *pg) 971 { 972 struct proc *p; 973 974 PGRP_LOCK_ASSERT(pg, MA_OWNED); 975 976 pg->pg_flags |= PGRP_ORPHANED; 977 978 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 979 PROC_LOCK(p); 980 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) { 981 PROC_UNLOCK(p); 982 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 983 PROC_LOCK(p); 984 kern_psignal(p, SIGHUP); 985 kern_psignal(p, SIGCONT); 986 PROC_UNLOCK(p); 987 } 988 return; 989 } 990 PROC_UNLOCK(p); 991 } 992 } 993 994 void 995 sess_hold(struct session *s) 996 { 997 998 refcount_acquire(&s->s_count); 999 } 1000 1001 void 1002 sess_release(struct session *s) 1003 { 1004 1005 if (refcount_release(&s->s_count)) { 1006 if (s->s_ttyp != NULL) { 1007 tty_lock(s->s_ttyp); 1008 tty_rel_sess(s->s_ttyp, s); 1009 } 1010 proc_id_clear(PROC_ID_SESSION, s->s_sid); 1011 mtx_destroy(&s->s_mtx); 1012 free(s, M_SESSION); 1013 } 1014 } 1015 1016 #ifdef DDB 1017 1018 static void 1019 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p) 1020 { 1021 db_printf( 1022 " pid %d at %p pr %d pgrp %p e %d jc %d\n", 1023 p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid, 1024 p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0, 1025 p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp)); 1026 } 1027 1028 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE) 1029 { 1030 struct pgrp *pgrp; 1031 struct proc *p; 1032 int i; 1033 1034 for (i = 0; i <= pgrphash; i++) { 1035 if (!LIST_EMPTY(&pgrphashtbl[i])) { 1036 db_printf("indx %d\n", i); 1037 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { 1038 db_printf( 1039 " pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n", 1040 pgrp, (int)pgrp->pg_id, pgrp->pg_session, 1041 pgrp->pg_session->s_count, 1042 LIST_FIRST(&pgrp->pg_members)); 1043 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) 1044 db_print_pgrp_one(pgrp, p); 1045 } 1046 } 1047 } 1048 } 1049 #endif /* DDB */ 1050 1051 /* 1052 * Calculate the kinfo_proc members which contain process-wide 1053 * informations. 1054 * Must be called with the target process locked. 1055 */ 1056 static void 1057 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp) 1058 { 1059 struct thread *td; 1060 1061 PROC_LOCK_ASSERT(p, MA_OWNED); 1062 1063 kp->ki_estcpu = 0; 1064 kp->ki_pctcpu = 0; 1065 FOREACH_THREAD_IN_PROC(p, td) { 1066 thread_lock(td); 1067 kp->ki_pctcpu += sched_pctcpu(td); 1068 kp->ki_estcpu += sched_estcpu(td); 1069 thread_unlock(td); 1070 } 1071 } 1072 1073 /* 1074 * Fill in any information that is common to all threads in the process. 1075 * Must be called with the target process locked. 1076 */ 1077 static void 1078 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp) 1079 { 1080 struct thread *td0; 1081 struct ucred *cred; 1082 struct sigacts *ps; 1083 struct timeval boottime; 1084 1085 PROC_LOCK_ASSERT(p, MA_OWNED); 1086 1087 kp->ki_structsize = sizeof(*kp); 1088 kp->ki_paddr = p; 1089 kp->ki_addr =/* p->p_addr; */0; /* XXX */ 1090 kp->ki_args = p->p_args; 1091 kp->ki_textvp = p->p_textvp; 1092 #ifdef KTRACE 1093 kp->ki_tracep = ktr_get_tracevp(p, false); 1094 kp->ki_traceflag = p->p_traceflag; 1095 #endif 1096 kp->ki_fd = p->p_fd; 1097 kp->ki_pd = p->p_pd; 1098 kp->ki_vmspace = p->p_vmspace; 1099 kp->ki_flag = p->p_flag; 1100 kp->ki_flag2 = p->p_flag2; 1101 cred = p->p_ucred; 1102 if (cred) { 1103 kp->ki_uid = cred->cr_uid; 1104 kp->ki_ruid = cred->cr_ruid; 1105 kp->ki_svuid = cred->cr_svuid; 1106 kp->ki_cr_flags = 0; 1107 if (cred->cr_flags & CRED_FLAG_CAPMODE) 1108 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE; 1109 /* XXX bde doesn't like KI_NGROUPS */ 1110 if (cred->cr_ngroups > KI_NGROUPS) { 1111 kp->ki_ngroups = KI_NGROUPS; 1112 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW; 1113 } else 1114 kp->ki_ngroups = cred->cr_ngroups; 1115 bcopy(cred->cr_groups, kp->ki_groups, 1116 kp->ki_ngroups * sizeof(gid_t)); 1117 kp->ki_rgid = cred->cr_rgid; 1118 kp->ki_svgid = cred->cr_svgid; 1119 /* If jailed(cred), emulate the old P_JAILED flag. */ 1120 if (jailed(cred)) { 1121 kp->ki_flag |= P_JAILED; 1122 /* If inside the jail, use 0 as a jail ID. */ 1123 if (cred->cr_prison != curthread->td_ucred->cr_prison) 1124 kp->ki_jid = cred->cr_prison->pr_id; 1125 } 1126 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name, 1127 sizeof(kp->ki_loginclass)); 1128 } 1129 ps = p->p_sigacts; 1130 if (ps) { 1131 mtx_lock(&ps->ps_mtx); 1132 kp->ki_sigignore = ps->ps_sigignore; 1133 kp->ki_sigcatch = ps->ps_sigcatch; 1134 mtx_unlock(&ps->ps_mtx); 1135 } 1136 if (p->p_state != PRS_NEW && 1137 p->p_state != PRS_ZOMBIE && 1138 p->p_vmspace != NULL) { 1139 struct vmspace *vm = p->p_vmspace; 1140 1141 kp->ki_size = vm->vm_map.size; 1142 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ 1143 FOREACH_THREAD_IN_PROC(p, td0) 1144 kp->ki_rssize += td0->td_kstack_pages; 1145 kp->ki_swrss = vm->vm_swrss; 1146 kp->ki_tsize = vm->vm_tsize; 1147 kp->ki_dsize = vm->vm_dsize; 1148 kp->ki_ssize = vm->vm_ssize; 1149 } else if (p->p_state == PRS_ZOMBIE) 1150 kp->ki_stat = SZOMB; 1151 if (kp->ki_flag & P_INMEM) 1152 kp->ki_sflag = PS_INMEM; 1153 else 1154 kp->ki_sflag = 0; 1155 /* Calculate legacy swtime as seconds since 'swtick'. */ 1156 kp->ki_swtime = (ticks - p->p_swtick) / hz; 1157 kp->ki_pid = p->p_pid; 1158 kp->ki_nice = p->p_nice; 1159 kp->ki_fibnum = p->p_fibnum; 1160 kp->ki_start = p->p_stats->p_start; 1161 getboottime(&boottime); 1162 timevaladd(&kp->ki_start, &boottime); 1163 PROC_STATLOCK(p); 1164 rufetch(p, &kp->ki_rusage); 1165 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime); 1166 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime); 1167 PROC_STATUNLOCK(p); 1168 calccru(p, &kp->ki_childutime, &kp->ki_childstime); 1169 /* Some callers want child times in a single value. */ 1170 kp->ki_childtime = kp->ki_childstime; 1171 timevaladd(&kp->ki_childtime, &kp->ki_childutime); 1172 1173 FOREACH_THREAD_IN_PROC(p, td0) 1174 kp->ki_cow += td0->td_cow; 1175 1176 if (p->p_comm[0] != '\0') 1177 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm)); 1178 if (p->p_sysent && p->p_sysent->sv_name != NULL && 1179 p->p_sysent->sv_name[0] != '\0') 1180 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul)); 1181 kp->ki_siglist = p->p_siglist; 1182 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig); 1183 kp->ki_acflag = p->p_acflag; 1184 kp->ki_lock = p->p_lock; 1185 if (p->p_pptr) { 1186 kp->ki_ppid = p->p_oppid; 1187 if (p->p_flag & P_TRACED) 1188 kp->ki_tracer = p->p_pptr->p_pid; 1189 } 1190 } 1191 1192 /* 1193 * Fill job-related process information. 1194 */ 1195 static void 1196 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp) 1197 { 1198 struct tty *tp; 1199 struct session *sp; 1200 struct pgrp *pgrp; 1201 1202 sx_assert(&proctree_lock, SA_LOCKED); 1203 PROC_LOCK_ASSERT(p, MA_OWNED); 1204 1205 pgrp = p->p_pgrp; 1206 if (pgrp == NULL) 1207 return; 1208 1209 kp->ki_pgid = pgrp->pg_id; 1210 kp->ki_jobc = pgrp_calc_jobc(pgrp); 1211 1212 sp = pgrp->pg_session; 1213 tp = NULL; 1214 1215 if (sp != NULL) { 1216 kp->ki_sid = sp->s_sid; 1217 SESS_LOCK(sp); 1218 strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login)); 1219 if (sp->s_ttyvp) 1220 kp->ki_kiflag |= KI_CTTY; 1221 if (SESS_LEADER(p)) 1222 kp->ki_kiflag |= KI_SLEADER; 1223 tp = sp->s_ttyp; 1224 SESS_UNLOCK(sp); 1225 } 1226 1227 if ((p->p_flag & P_CONTROLT) && tp != NULL) { 1228 kp->ki_tdev = tty_udev(tp); 1229 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 1230 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; 1231 if (tp->t_session) 1232 kp->ki_tsid = tp->t_session->s_sid; 1233 } else { 1234 kp->ki_tdev = NODEV; 1235 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 1236 } 1237 } 1238 1239 /* 1240 * Fill in information that is thread specific. Must be called with 1241 * target process locked. If 'preferthread' is set, overwrite certain 1242 * process-related fields that are maintained for both threads and 1243 * processes. 1244 */ 1245 static void 1246 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread) 1247 { 1248 struct proc *p; 1249 1250 p = td->td_proc; 1251 kp->ki_tdaddr = td; 1252 PROC_LOCK_ASSERT(p, MA_OWNED); 1253 1254 if (preferthread) 1255 PROC_STATLOCK(p); 1256 thread_lock(td); 1257 if (td->td_wmesg != NULL) 1258 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg)); 1259 else 1260 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg)); 1261 if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >= 1262 sizeof(kp->ki_tdname)) { 1263 strlcpy(kp->ki_moretdname, 1264 td->td_name + sizeof(kp->ki_tdname) - 1, 1265 sizeof(kp->ki_moretdname)); 1266 } else { 1267 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname)); 1268 } 1269 if (TD_ON_LOCK(td)) { 1270 kp->ki_kiflag |= KI_LOCKBLOCK; 1271 strlcpy(kp->ki_lockname, td->td_lockname, 1272 sizeof(kp->ki_lockname)); 1273 } else { 1274 kp->ki_kiflag &= ~KI_LOCKBLOCK; 1275 bzero(kp->ki_lockname, sizeof(kp->ki_lockname)); 1276 } 1277 1278 if (p->p_state == PRS_NORMAL) { /* approximate. */ 1279 if (TD_ON_RUNQ(td) || 1280 TD_CAN_RUN(td) || 1281 TD_IS_RUNNING(td)) { 1282 kp->ki_stat = SRUN; 1283 } else if (P_SHOULDSTOP(p)) { 1284 kp->ki_stat = SSTOP; 1285 } else if (TD_IS_SLEEPING(td)) { 1286 kp->ki_stat = SSLEEP; 1287 } else if (TD_ON_LOCK(td)) { 1288 kp->ki_stat = SLOCK; 1289 } else { 1290 kp->ki_stat = SWAIT; 1291 } 1292 } else if (p->p_state == PRS_ZOMBIE) { 1293 kp->ki_stat = SZOMB; 1294 } else { 1295 kp->ki_stat = SIDL; 1296 } 1297 1298 /* Things in the thread */ 1299 kp->ki_wchan = td->td_wchan; 1300 kp->ki_pri.pri_level = td->td_priority; 1301 kp->ki_pri.pri_native = td->td_base_pri; 1302 1303 /* 1304 * Note: legacy fields; clamp at the old NOCPU value and/or 1305 * the maximum u_char CPU value. 1306 */ 1307 if (td->td_lastcpu == NOCPU) 1308 kp->ki_lastcpu_old = NOCPU_OLD; 1309 else if (td->td_lastcpu > MAXCPU_OLD) 1310 kp->ki_lastcpu_old = MAXCPU_OLD; 1311 else 1312 kp->ki_lastcpu_old = td->td_lastcpu; 1313 1314 if (td->td_oncpu == NOCPU) 1315 kp->ki_oncpu_old = NOCPU_OLD; 1316 else if (td->td_oncpu > MAXCPU_OLD) 1317 kp->ki_oncpu_old = MAXCPU_OLD; 1318 else 1319 kp->ki_oncpu_old = td->td_oncpu; 1320 1321 kp->ki_lastcpu = td->td_lastcpu; 1322 kp->ki_oncpu = td->td_oncpu; 1323 kp->ki_tdflags = td->td_flags; 1324 kp->ki_tid = td->td_tid; 1325 kp->ki_numthreads = p->p_numthreads; 1326 kp->ki_pcb = td->td_pcb; 1327 kp->ki_kstack = (void *)td->td_kstack; 1328 kp->ki_slptime = (ticks - td->td_slptick) / hz; 1329 kp->ki_pri.pri_class = td->td_pri_class; 1330 kp->ki_pri.pri_user = td->td_user_pri; 1331 1332 if (preferthread) { 1333 rufetchtd(td, &kp->ki_rusage); 1334 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime); 1335 kp->ki_pctcpu = sched_pctcpu(td); 1336 kp->ki_estcpu = sched_estcpu(td); 1337 kp->ki_cow = td->td_cow; 1338 } 1339 1340 /* We can't get this anymore but ps etc never used it anyway. */ 1341 kp->ki_rqindex = 0; 1342 1343 if (preferthread) 1344 kp->ki_siglist = td->td_siglist; 1345 kp->ki_sigmask = td->td_sigmask; 1346 thread_unlock(td); 1347 if (preferthread) 1348 PROC_STATUNLOCK(p); 1349 } 1350 1351 /* 1352 * Fill in a kinfo_proc structure for the specified process. 1353 * Must be called with the target process locked. 1354 */ 1355 void 1356 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp) 1357 { 1358 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1359 1360 bzero(kp, sizeof(*kp)); 1361 1362 fill_kinfo_proc_pgrp(p,kp); 1363 fill_kinfo_proc_only(p, kp); 1364 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0); 1365 fill_kinfo_aggregate(p, kp); 1366 } 1367 1368 struct pstats * 1369 pstats_alloc(void) 1370 { 1371 1372 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK)); 1373 } 1374 1375 /* 1376 * Copy parts of p_stats; zero the rest of p_stats (statistics). 1377 */ 1378 void 1379 pstats_fork(struct pstats *src, struct pstats *dst) 1380 { 1381 1382 bzero(&dst->pstat_startzero, 1383 __rangeof(struct pstats, pstat_startzero, pstat_endzero)); 1384 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy, 1385 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy)); 1386 } 1387 1388 void 1389 pstats_free(struct pstats *ps) 1390 { 1391 1392 free(ps, M_SUBPROC); 1393 } 1394 1395 #ifdef COMPAT_FREEBSD32 1396 1397 /* 1398 * This function is typically used to copy out the kernel address, so 1399 * it can be replaced by assignment of zero. 1400 */ 1401 static inline uint32_t 1402 ptr32_trim(const void *ptr) 1403 { 1404 uintptr_t uptr; 1405 1406 uptr = (uintptr_t)ptr; 1407 return ((uptr > UINT_MAX) ? 0 : uptr); 1408 } 1409 1410 #define PTRTRIM_CP(src,dst,fld) \ 1411 do { (dst).fld = ptr32_trim((src).fld); } while (0) 1412 1413 static void 1414 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32) 1415 { 1416 int i; 1417 1418 bzero(ki32, sizeof(struct kinfo_proc32)); 1419 ki32->ki_structsize = sizeof(struct kinfo_proc32); 1420 CP(*ki, *ki32, ki_layout); 1421 PTRTRIM_CP(*ki, *ki32, ki_args); 1422 PTRTRIM_CP(*ki, *ki32, ki_paddr); 1423 PTRTRIM_CP(*ki, *ki32, ki_addr); 1424 PTRTRIM_CP(*ki, *ki32, ki_tracep); 1425 PTRTRIM_CP(*ki, *ki32, ki_textvp); 1426 PTRTRIM_CP(*ki, *ki32, ki_fd); 1427 PTRTRIM_CP(*ki, *ki32, ki_vmspace); 1428 PTRTRIM_CP(*ki, *ki32, ki_wchan); 1429 CP(*ki, *ki32, ki_pid); 1430 CP(*ki, *ki32, ki_ppid); 1431 CP(*ki, *ki32, ki_pgid); 1432 CP(*ki, *ki32, ki_tpgid); 1433 CP(*ki, *ki32, ki_sid); 1434 CP(*ki, *ki32, ki_tsid); 1435 CP(*ki, *ki32, ki_jobc); 1436 CP(*ki, *ki32, ki_tdev); 1437 CP(*ki, *ki32, ki_tdev_freebsd11); 1438 CP(*ki, *ki32, ki_siglist); 1439 CP(*ki, *ki32, ki_sigmask); 1440 CP(*ki, *ki32, ki_sigignore); 1441 CP(*ki, *ki32, ki_sigcatch); 1442 CP(*ki, *ki32, ki_uid); 1443 CP(*ki, *ki32, ki_ruid); 1444 CP(*ki, *ki32, ki_svuid); 1445 CP(*ki, *ki32, ki_rgid); 1446 CP(*ki, *ki32, ki_svgid); 1447 CP(*ki, *ki32, ki_ngroups); 1448 for (i = 0; i < KI_NGROUPS; i++) 1449 CP(*ki, *ki32, ki_groups[i]); 1450 CP(*ki, *ki32, ki_size); 1451 CP(*ki, *ki32, ki_rssize); 1452 CP(*ki, *ki32, ki_swrss); 1453 CP(*ki, *ki32, ki_tsize); 1454 CP(*ki, *ki32, ki_dsize); 1455 CP(*ki, *ki32, ki_ssize); 1456 CP(*ki, *ki32, ki_xstat); 1457 CP(*ki, *ki32, ki_acflag); 1458 CP(*ki, *ki32, ki_pctcpu); 1459 CP(*ki, *ki32, ki_estcpu); 1460 CP(*ki, *ki32, ki_slptime); 1461 CP(*ki, *ki32, ki_swtime); 1462 CP(*ki, *ki32, ki_cow); 1463 CP(*ki, *ki32, ki_runtime); 1464 TV_CP(*ki, *ki32, ki_start); 1465 TV_CP(*ki, *ki32, ki_childtime); 1466 CP(*ki, *ki32, ki_flag); 1467 CP(*ki, *ki32, ki_kiflag); 1468 CP(*ki, *ki32, ki_traceflag); 1469 CP(*ki, *ki32, ki_stat); 1470 CP(*ki, *ki32, ki_nice); 1471 CP(*ki, *ki32, ki_lock); 1472 CP(*ki, *ki32, ki_rqindex); 1473 CP(*ki, *ki32, ki_oncpu); 1474 CP(*ki, *ki32, ki_lastcpu); 1475 1476 /* XXX TODO: wrap cpu value as appropriate */ 1477 CP(*ki, *ki32, ki_oncpu_old); 1478 CP(*ki, *ki32, ki_lastcpu_old); 1479 1480 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1); 1481 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1); 1482 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1); 1483 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1); 1484 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1); 1485 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1); 1486 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1); 1487 bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1); 1488 CP(*ki, *ki32, ki_tracer); 1489 CP(*ki, *ki32, ki_flag2); 1490 CP(*ki, *ki32, ki_fibnum); 1491 CP(*ki, *ki32, ki_cr_flags); 1492 CP(*ki, *ki32, ki_jid); 1493 CP(*ki, *ki32, ki_numthreads); 1494 CP(*ki, *ki32, ki_tid); 1495 CP(*ki, *ki32, ki_pri); 1496 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage); 1497 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch); 1498 PTRTRIM_CP(*ki, *ki32, ki_pcb); 1499 PTRTRIM_CP(*ki, *ki32, ki_kstack); 1500 PTRTRIM_CP(*ki, *ki32, ki_udata); 1501 PTRTRIM_CP(*ki, *ki32, ki_tdaddr); 1502 CP(*ki, *ki32, ki_sflag); 1503 CP(*ki, *ki32, ki_tdflags); 1504 } 1505 #endif 1506 1507 static ssize_t 1508 kern_proc_out_size(struct proc *p, int flags) 1509 { 1510 ssize_t size = 0; 1511 1512 PROC_LOCK_ASSERT(p, MA_OWNED); 1513 1514 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1515 #ifdef COMPAT_FREEBSD32 1516 if ((flags & KERN_PROC_MASK32) != 0) { 1517 size += sizeof(struct kinfo_proc32); 1518 } else 1519 #endif 1520 size += sizeof(struct kinfo_proc); 1521 } else { 1522 #ifdef COMPAT_FREEBSD32 1523 if ((flags & KERN_PROC_MASK32) != 0) 1524 size += sizeof(struct kinfo_proc32) * p->p_numthreads; 1525 else 1526 #endif 1527 size += sizeof(struct kinfo_proc) * p->p_numthreads; 1528 } 1529 PROC_UNLOCK(p); 1530 return (size); 1531 } 1532 1533 int 1534 kern_proc_out(struct proc *p, struct sbuf *sb, int flags) 1535 { 1536 struct thread *td; 1537 struct kinfo_proc ki; 1538 #ifdef COMPAT_FREEBSD32 1539 struct kinfo_proc32 ki32; 1540 #endif 1541 int error; 1542 1543 PROC_LOCK_ASSERT(p, MA_OWNED); 1544 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1545 1546 error = 0; 1547 fill_kinfo_proc(p, &ki); 1548 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1549 #ifdef COMPAT_FREEBSD32 1550 if ((flags & KERN_PROC_MASK32) != 0) { 1551 freebsd32_kinfo_proc_out(&ki, &ki32); 1552 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1553 error = ENOMEM; 1554 } else 1555 #endif 1556 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1557 error = ENOMEM; 1558 } else { 1559 FOREACH_THREAD_IN_PROC(p, td) { 1560 fill_kinfo_thread(td, &ki, 1); 1561 #ifdef COMPAT_FREEBSD32 1562 if ((flags & KERN_PROC_MASK32) != 0) { 1563 freebsd32_kinfo_proc_out(&ki, &ki32); 1564 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1565 error = ENOMEM; 1566 } else 1567 #endif 1568 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1569 error = ENOMEM; 1570 if (error != 0) 1571 break; 1572 } 1573 } 1574 PROC_UNLOCK(p); 1575 return (error); 1576 } 1577 1578 static int 1579 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags) 1580 { 1581 struct sbuf sb; 1582 struct kinfo_proc ki; 1583 int error, error2; 1584 1585 if (req->oldptr == NULL) 1586 return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags))); 1587 1588 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); 1589 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1590 error = kern_proc_out(p, &sb, flags); 1591 error2 = sbuf_finish(&sb); 1592 sbuf_delete(&sb); 1593 if (error != 0) 1594 return (error); 1595 else if (error2 != 0) 1596 return (error2); 1597 return (0); 1598 } 1599 1600 int 1601 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg) 1602 { 1603 struct proc *p; 1604 int error, i, j; 1605 1606 for (i = 0; i < pidhashlock + 1; i++) { 1607 sx_slock(&proctree_lock); 1608 sx_slock(&pidhashtbl_lock[i]); 1609 for (j = i; j <= pidhash; j += pidhashlock + 1) { 1610 LIST_FOREACH(p, &pidhashtbl[j], p_hash) { 1611 if (p->p_state == PRS_NEW) 1612 continue; 1613 error = cb(p, cbarg); 1614 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 1615 if (error != 0) { 1616 sx_sunlock(&pidhashtbl_lock[i]); 1617 sx_sunlock(&proctree_lock); 1618 return (error); 1619 } 1620 } 1621 } 1622 sx_sunlock(&pidhashtbl_lock[i]); 1623 sx_sunlock(&proctree_lock); 1624 } 1625 return (0); 1626 } 1627 1628 struct kern_proc_out_args { 1629 struct sysctl_req *req; 1630 int flags; 1631 int oid_number; 1632 int *name; 1633 }; 1634 1635 static int 1636 sysctl_kern_proc_iterate(struct proc *p, void *origarg) 1637 { 1638 struct kern_proc_out_args *arg = origarg; 1639 int *name = arg->name; 1640 int oid_number = arg->oid_number; 1641 int flags = arg->flags; 1642 struct sysctl_req *req = arg->req; 1643 int error = 0; 1644 1645 PROC_LOCK(p); 1646 1647 KASSERT(p->p_ucred != NULL, 1648 ("process credential is NULL for non-NEW proc")); 1649 /* 1650 * Show a user only appropriate processes. 1651 */ 1652 if (p_cansee(curthread, p)) 1653 goto skip; 1654 /* 1655 * TODO - make more efficient (see notes below). 1656 * do by session. 1657 */ 1658 switch (oid_number) { 1659 case KERN_PROC_GID: 1660 if (p->p_ucred->cr_gid != (gid_t)name[0]) 1661 goto skip; 1662 break; 1663 1664 case KERN_PROC_PGRP: 1665 /* could do this by traversing pgrp */ 1666 if (p->p_pgrp == NULL || 1667 p->p_pgrp->pg_id != (pid_t)name[0]) 1668 goto skip; 1669 break; 1670 1671 case KERN_PROC_RGID: 1672 if (p->p_ucred->cr_rgid != (gid_t)name[0]) 1673 goto skip; 1674 break; 1675 1676 case KERN_PROC_SESSION: 1677 if (p->p_session == NULL || 1678 p->p_session->s_sid != (pid_t)name[0]) 1679 goto skip; 1680 break; 1681 1682 case KERN_PROC_TTY: 1683 if ((p->p_flag & P_CONTROLT) == 0 || 1684 p->p_session == NULL) 1685 goto skip; 1686 /* XXX proctree_lock */ 1687 SESS_LOCK(p->p_session); 1688 if (p->p_session->s_ttyp == NULL || 1689 tty_udev(p->p_session->s_ttyp) != 1690 (dev_t)name[0]) { 1691 SESS_UNLOCK(p->p_session); 1692 goto skip; 1693 } 1694 SESS_UNLOCK(p->p_session); 1695 break; 1696 1697 case KERN_PROC_UID: 1698 if (p->p_ucred->cr_uid != (uid_t)name[0]) 1699 goto skip; 1700 break; 1701 1702 case KERN_PROC_RUID: 1703 if (p->p_ucred->cr_ruid != (uid_t)name[0]) 1704 goto skip; 1705 break; 1706 1707 case KERN_PROC_PROC: 1708 break; 1709 1710 default: 1711 break; 1712 } 1713 error = sysctl_out_proc(p, req, flags); 1714 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 1715 return (error); 1716 skip: 1717 PROC_UNLOCK(p); 1718 return (0); 1719 } 1720 1721 static int 1722 sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 1723 { 1724 struct kern_proc_out_args iterarg; 1725 int *name = (int *)arg1; 1726 u_int namelen = arg2; 1727 struct proc *p; 1728 int flags, oid_number; 1729 int error = 0; 1730 1731 oid_number = oidp->oid_number; 1732 if (oid_number != KERN_PROC_ALL && 1733 (oid_number & KERN_PROC_INC_THREAD) == 0) 1734 flags = KERN_PROC_NOTHREADS; 1735 else { 1736 flags = 0; 1737 oid_number &= ~KERN_PROC_INC_THREAD; 1738 } 1739 #ifdef COMPAT_FREEBSD32 1740 if (req->flags & SCTL_MASK32) 1741 flags |= KERN_PROC_MASK32; 1742 #endif 1743 if (oid_number == KERN_PROC_PID) { 1744 if (namelen != 1) 1745 return (EINVAL); 1746 error = sysctl_wire_old_buffer(req, 0); 1747 if (error) 1748 return (error); 1749 sx_slock(&proctree_lock); 1750 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1751 if (error == 0) 1752 error = sysctl_out_proc(p, req, flags); 1753 sx_sunlock(&proctree_lock); 1754 return (error); 1755 } 1756 1757 switch (oid_number) { 1758 case KERN_PROC_ALL: 1759 if (namelen != 0) 1760 return (EINVAL); 1761 break; 1762 case KERN_PROC_PROC: 1763 if (namelen != 0 && namelen != 1) 1764 return (EINVAL); 1765 break; 1766 default: 1767 if (namelen != 1) 1768 return (EINVAL); 1769 break; 1770 } 1771 1772 if (req->oldptr == NULL) { 1773 /* overestimate by 5 procs */ 1774 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1775 if (error) 1776 return (error); 1777 } else { 1778 error = sysctl_wire_old_buffer(req, 0); 1779 if (error != 0) 1780 return (error); 1781 } 1782 iterarg.flags = flags; 1783 iterarg.oid_number = oid_number; 1784 iterarg.req = req; 1785 iterarg.name = name; 1786 error = proc_iterate(sysctl_kern_proc_iterate, &iterarg); 1787 return (error); 1788 } 1789 1790 struct pargs * 1791 pargs_alloc(int len) 1792 { 1793 struct pargs *pa; 1794 1795 pa = malloc(sizeof(struct pargs) + len, M_PARGS, 1796 M_WAITOK); 1797 refcount_init(&pa->ar_ref, 1); 1798 pa->ar_length = len; 1799 return (pa); 1800 } 1801 1802 static void 1803 pargs_free(struct pargs *pa) 1804 { 1805 1806 free(pa, M_PARGS); 1807 } 1808 1809 void 1810 pargs_hold(struct pargs *pa) 1811 { 1812 1813 if (pa == NULL) 1814 return; 1815 refcount_acquire(&pa->ar_ref); 1816 } 1817 1818 void 1819 pargs_drop(struct pargs *pa) 1820 { 1821 1822 if (pa == NULL) 1823 return; 1824 if (refcount_release(&pa->ar_ref)) 1825 pargs_free(pa); 1826 } 1827 1828 static int 1829 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, 1830 size_t len) 1831 { 1832 ssize_t n; 1833 1834 /* 1835 * This may return a short read if the string is shorter than the chunk 1836 * and is aligned at the end of the page, and the following page is not 1837 * mapped. 1838 */ 1839 n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len); 1840 if (n <= 0) 1841 return (ENOMEM); 1842 return (0); 1843 } 1844 1845 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ 1846 1847 enum proc_vector_type { 1848 PROC_ARG, 1849 PROC_ENV, 1850 PROC_AUX, 1851 }; 1852 1853 #ifdef COMPAT_FREEBSD32 1854 static int 1855 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, 1856 size_t *vsizep, enum proc_vector_type type) 1857 { 1858 struct freebsd32_ps_strings pss; 1859 Elf32_Auxinfo aux; 1860 vm_offset_t vptr, ptr; 1861 uint32_t *proc_vector32; 1862 char **proc_vector; 1863 size_t vsize, size; 1864 int i, error; 1865 1866 error = 0; 1867 if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) != 1868 sizeof(pss)) 1869 return (ENOMEM); 1870 switch (type) { 1871 case PROC_ARG: 1872 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); 1873 vsize = pss.ps_nargvstr; 1874 if (vsize > ARG_MAX) 1875 return (ENOEXEC); 1876 size = vsize * sizeof(int32_t); 1877 break; 1878 case PROC_ENV: 1879 vptr = (vm_offset_t)PTRIN(pss.ps_envstr); 1880 vsize = pss.ps_nenvstr; 1881 if (vsize > ARG_MAX) 1882 return (ENOEXEC); 1883 size = vsize * sizeof(int32_t); 1884 break; 1885 case PROC_AUX: 1886 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + 1887 (pss.ps_nenvstr + 1) * sizeof(int32_t); 1888 if (vptr % 4 != 0) 1889 return (ENOEXEC); 1890 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1891 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1892 sizeof(aux)) 1893 return (ENOMEM); 1894 if (aux.a_type == AT_NULL) 1895 break; 1896 ptr += sizeof(aux); 1897 } 1898 if (aux.a_type != AT_NULL) 1899 return (ENOEXEC); 1900 vsize = i + 1; 1901 size = vsize * sizeof(aux); 1902 break; 1903 default: 1904 KASSERT(0, ("Wrong proc vector type: %d", type)); 1905 return (EINVAL); 1906 } 1907 proc_vector32 = malloc(size, M_TEMP, M_WAITOK); 1908 if (proc_readmem(td, p, vptr, proc_vector32, size) != size) { 1909 error = ENOMEM; 1910 goto done; 1911 } 1912 if (type == PROC_AUX) { 1913 *proc_vectorp = (char **)proc_vector32; 1914 *vsizep = vsize; 1915 return (0); 1916 } 1917 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); 1918 for (i = 0; i < (int)vsize; i++) 1919 proc_vector[i] = PTRIN(proc_vector32[i]); 1920 *proc_vectorp = proc_vector; 1921 *vsizep = vsize; 1922 done: 1923 free(proc_vector32, M_TEMP); 1924 return (error); 1925 } 1926 #endif 1927 1928 static int 1929 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, 1930 size_t *vsizep, enum proc_vector_type type) 1931 { 1932 struct ps_strings pss; 1933 Elf_Auxinfo aux; 1934 vm_offset_t vptr, ptr; 1935 char **proc_vector; 1936 size_t vsize, size; 1937 int i; 1938 1939 #ifdef COMPAT_FREEBSD32 1940 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1941 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); 1942 #endif 1943 if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) != 1944 sizeof(pss)) 1945 return (ENOMEM); 1946 switch (type) { 1947 case PROC_ARG: 1948 vptr = (vm_offset_t)pss.ps_argvstr; 1949 vsize = pss.ps_nargvstr; 1950 if (vsize > ARG_MAX) 1951 return (ENOEXEC); 1952 size = vsize * sizeof(char *); 1953 break; 1954 case PROC_ENV: 1955 vptr = (vm_offset_t)pss.ps_envstr; 1956 vsize = pss.ps_nenvstr; 1957 if (vsize > ARG_MAX) 1958 return (ENOEXEC); 1959 size = vsize * sizeof(char *); 1960 break; 1961 case PROC_AUX: 1962 /* 1963 * The aux array is just above env array on the stack. Check 1964 * that the address is naturally aligned. 1965 */ 1966 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) 1967 * sizeof(char *); 1968 #if __ELF_WORD_SIZE == 64 1969 if (vptr % sizeof(uint64_t) != 0) 1970 #else 1971 if (vptr % sizeof(uint32_t) != 0) 1972 #endif 1973 return (ENOEXEC); 1974 /* 1975 * We count the array size reading the aux vectors from the 1976 * stack until AT_NULL vector is returned. So (to keep the code 1977 * simple) we read the process stack twice: the first time here 1978 * to find the size and the second time when copying the vectors 1979 * to the allocated proc_vector. 1980 */ 1981 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1982 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1983 sizeof(aux)) 1984 return (ENOMEM); 1985 if (aux.a_type == AT_NULL) 1986 break; 1987 ptr += sizeof(aux); 1988 } 1989 /* 1990 * If the PROC_AUXV_MAX entries are iterated over, and we have 1991 * not reached AT_NULL, it is most likely we are reading wrong 1992 * data: either the process doesn't have auxv array or data has 1993 * been modified. Return the error in this case. 1994 */ 1995 if (aux.a_type != AT_NULL) 1996 return (ENOEXEC); 1997 vsize = i + 1; 1998 size = vsize * sizeof(aux); 1999 break; 2000 default: 2001 KASSERT(0, ("Wrong proc vector type: %d", type)); 2002 return (EINVAL); /* In case we are built without INVARIANTS. */ 2003 } 2004 proc_vector = malloc(size, M_TEMP, M_WAITOK); 2005 if (proc_readmem(td, p, vptr, proc_vector, size) != size) { 2006 free(proc_vector, M_TEMP); 2007 return (ENOMEM); 2008 } 2009 *proc_vectorp = proc_vector; 2010 *vsizep = vsize; 2011 2012 return (0); 2013 } 2014 2015 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ 2016 2017 static int 2018 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, 2019 enum proc_vector_type type) 2020 { 2021 size_t done, len, nchr, vsize; 2022 int error, i; 2023 char **proc_vector, *sptr; 2024 char pss_string[GET_PS_STRINGS_CHUNK_SZ]; 2025 2026 PROC_ASSERT_HELD(p); 2027 2028 /* 2029 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. 2030 */ 2031 nchr = 2 * (PATH_MAX + ARG_MAX); 2032 2033 error = get_proc_vector(td, p, &proc_vector, &vsize, type); 2034 if (error != 0) 2035 return (error); 2036 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { 2037 /* 2038 * The program may have scribbled into its argv array, e.g. to 2039 * remove some arguments. If that has happened, break out 2040 * before trying to read from NULL. 2041 */ 2042 if (proc_vector[i] == NULL) 2043 break; 2044 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { 2045 error = proc_read_string(td, p, sptr, pss_string, 2046 sizeof(pss_string)); 2047 if (error != 0) 2048 goto done; 2049 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); 2050 if (done + len >= nchr) 2051 len = nchr - done - 1; 2052 sbuf_bcat(sb, pss_string, len); 2053 if (len != GET_PS_STRINGS_CHUNK_SZ) 2054 break; 2055 done += GET_PS_STRINGS_CHUNK_SZ; 2056 } 2057 sbuf_bcat(sb, "", 1); 2058 done += len + 1; 2059 } 2060 done: 2061 free(proc_vector, M_TEMP); 2062 return (error); 2063 } 2064 2065 int 2066 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) 2067 { 2068 2069 return (get_ps_strings(curthread, p, sb, PROC_ARG)); 2070 } 2071 2072 int 2073 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) 2074 { 2075 2076 return (get_ps_strings(curthread, p, sb, PROC_ENV)); 2077 } 2078 2079 int 2080 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) 2081 { 2082 size_t vsize, size; 2083 char **auxv; 2084 int error; 2085 2086 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); 2087 if (error == 0) { 2088 #ifdef COMPAT_FREEBSD32 2089 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 2090 size = vsize * sizeof(Elf32_Auxinfo); 2091 else 2092 #endif 2093 size = vsize * sizeof(Elf_Auxinfo); 2094 if (sbuf_bcat(sb, auxv, size) != 0) 2095 error = ENOMEM; 2096 free(auxv, M_TEMP); 2097 } 2098 return (error); 2099 } 2100 2101 /* 2102 * This sysctl allows a process to retrieve the argument list or process 2103 * title for another process without groping around in the address space 2104 * of the other process. It also allow a process to set its own "process 2105 * title to a string of its own choice. 2106 */ 2107 static int 2108 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 2109 { 2110 int *name = (int *)arg1; 2111 u_int namelen = arg2; 2112 struct pargs *newpa, *pa; 2113 struct proc *p; 2114 struct sbuf sb; 2115 int flags, error = 0, error2; 2116 pid_t pid; 2117 2118 if (namelen != 1) 2119 return (EINVAL); 2120 2121 p = curproc; 2122 pid = (pid_t)name[0]; 2123 if (pid == -1) { 2124 pid = p->p_pid; 2125 } 2126 2127 /* 2128 * If the query is for this process and it is single-threaded, there 2129 * is nobody to modify pargs, thus we can just read. 2130 */ 2131 if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL && 2132 (pa = p->p_args) != NULL) 2133 return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length)); 2134 2135 flags = PGET_CANSEE; 2136 if (req->newptr != NULL) 2137 flags |= PGET_ISCURRENT; 2138 error = pget(pid, flags, &p); 2139 if (error) 2140 return (error); 2141 2142 pa = p->p_args; 2143 if (pa != NULL) { 2144 pargs_hold(pa); 2145 PROC_UNLOCK(p); 2146 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 2147 pargs_drop(pa); 2148 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { 2149 _PHOLD(p); 2150 PROC_UNLOCK(p); 2151 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2152 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2153 error = proc_getargv(curthread, p, &sb); 2154 error2 = sbuf_finish(&sb); 2155 PRELE(p); 2156 sbuf_delete(&sb); 2157 if (error == 0 && error2 != 0) 2158 error = error2; 2159 } else { 2160 PROC_UNLOCK(p); 2161 } 2162 if (error != 0 || req->newptr == NULL) 2163 return (error); 2164 2165 if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs)) 2166 return (ENOMEM); 2167 2168 if (req->newlen == 0) { 2169 /* 2170 * Clear the argument pointer, so that we'll fetch arguments 2171 * with proc_getargv() until further notice. 2172 */ 2173 newpa = NULL; 2174 } else { 2175 newpa = pargs_alloc(req->newlen); 2176 error = SYSCTL_IN(req, newpa->ar_args, req->newlen); 2177 if (error != 0) { 2178 pargs_free(newpa); 2179 return (error); 2180 } 2181 } 2182 PROC_LOCK(p); 2183 pa = p->p_args; 2184 p->p_args = newpa; 2185 PROC_UNLOCK(p); 2186 pargs_drop(pa); 2187 return (0); 2188 } 2189 2190 /* 2191 * This sysctl allows a process to retrieve environment of another process. 2192 */ 2193 static int 2194 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) 2195 { 2196 int *name = (int *)arg1; 2197 u_int namelen = arg2; 2198 struct proc *p; 2199 struct sbuf sb; 2200 int error, error2; 2201 2202 if (namelen != 1) 2203 return (EINVAL); 2204 2205 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2206 if (error != 0) 2207 return (error); 2208 if ((p->p_flag & P_SYSTEM) != 0) { 2209 PRELE(p); 2210 return (0); 2211 } 2212 2213 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2214 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2215 error = proc_getenvv(curthread, p, &sb); 2216 error2 = sbuf_finish(&sb); 2217 PRELE(p); 2218 sbuf_delete(&sb); 2219 return (error != 0 ? error : error2); 2220 } 2221 2222 /* 2223 * This sysctl allows a process to retrieve ELF auxiliary vector of 2224 * another process. 2225 */ 2226 static int 2227 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) 2228 { 2229 int *name = (int *)arg1; 2230 u_int namelen = arg2; 2231 struct proc *p; 2232 struct sbuf sb; 2233 int error, error2; 2234 2235 if (namelen != 1) 2236 return (EINVAL); 2237 2238 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2239 if (error != 0) 2240 return (error); 2241 if ((p->p_flag & P_SYSTEM) != 0) { 2242 PRELE(p); 2243 return (0); 2244 } 2245 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2246 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2247 error = proc_getauxv(curthread, p, &sb); 2248 error2 = sbuf_finish(&sb); 2249 PRELE(p); 2250 sbuf_delete(&sb); 2251 return (error != 0 ? error : error2); 2252 } 2253 2254 /* 2255 * Look up the canonical executable path running in the specified process. 2256 * It tries to return the same hardlink name as was used for execve(2). 2257 * This allows the programs that modify their behavior based on their progname, 2258 * to operate correctly. 2259 * 2260 * Result is returned in retbuf, it must not be freed, similar to vn_fullpath() 2261 * calling conventions. 2262 * binname is a pointer to temporary string buffer of length MAXPATHLEN, 2263 * allocated and freed by caller. 2264 * freebuf should be freed by caller, from the M_TEMP malloc type. 2265 */ 2266 int 2267 proc_get_binpath(struct proc *p, char *binname, char **retbuf, 2268 char **freebuf) 2269 { 2270 struct nameidata nd; 2271 struct vnode *vp, *dvp; 2272 size_t freepath_size; 2273 int error; 2274 bool do_fullpath; 2275 2276 PROC_LOCK_ASSERT(p, MA_OWNED); 2277 2278 vp = p->p_textvp; 2279 if (vp == NULL) { 2280 PROC_UNLOCK(p); 2281 *retbuf = ""; 2282 *freebuf = NULL; 2283 return (0); 2284 } 2285 vref(vp); 2286 dvp = p->p_textdvp; 2287 if (dvp != NULL) 2288 vref(dvp); 2289 if (p->p_binname != NULL) 2290 strlcpy(binname, p->p_binname, MAXPATHLEN); 2291 PROC_UNLOCK(p); 2292 2293 do_fullpath = true; 2294 *freebuf = NULL; 2295 if (dvp != NULL && binname[0] != '\0') { 2296 freepath_size = MAXPATHLEN; 2297 if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname), 2298 retbuf, freebuf, &freepath_size) == 0) { 2299 /* 2300 * Recheck the looked up path. The binary 2301 * might have been renamed or replaced, in 2302 * which case we should not report old name. 2303 */ 2304 NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf); 2305 error = namei(&nd); 2306 if (error == 0) { 2307 if (nd.ni_vp == vp) 2308 do_fullpath = false; 2309 vrele(nd.ni_vp); 2310 NDFREE_PNBUF(&nd); 2311 } 2312 } 2313 } 2314 if (do_fullpath) { 2315 free(*freebuf, M_TEMP); 2316 *freebuf = NULL; 2317 error = vn_fullpath(vp, retbuf, freebuf); 2318 } 2319 vrele(vp); 2320 if (dvp != NULL) 2321 vrele(dvp); 2322 return (error); 2323 } 2324 2325 /* 2326 * This sysctl allows a process to retrieve the path of the executable for 2327 * itself or another process. 2328 */ 2329 static int 2330 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) 2331 { 2332 pid_t *pidp = (pid_t *)arg1; 2333 unsigned int arglen = arg2; 2334 struct proc *p; 2335 char *retbuf, *freebuf, *binname; 2336 int error; 2337 2338 if (arglen != 1) 2339 return (EINVAL); 2340 binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 2341 binname[0] = '\0'; 2342 if (*pidp == -1) { /* -1 means this process */ 2343 error = 0; 2344 p = req->td->td_proc; 2345 PROC_LOCK(p); 2346 } else { 2347 error = pget(*pidp, PGET_CANSEE, &p); 2348 } 2349 2350 if (error == 0) 2351 error = proc_get_binpath(p, binname, &retbuf, &freebuf); 2352 free(binname, M_TEMP); 2353 if (error != 0) 2354 return (error); 2355 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); 2356 free(freebuf, M_TEMP); 2357 return (error); 2358 } 2359 2360 static int 2361 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) 2362 { 2363 struct proc *p; 2364 char *sv_name; 2365 int *name; 2366 int namelen; 2367 int error; 2368 2369 namelen = arg2; 2370 if (namelen != 1) 2371 return (EINVAL); 2372 2373 name = (int *)arg1; 2374 error = pget((pid_t)name[0], PGET_CANSEE, &p); 2375 if (error != 0) 2376 return (error); 2377 sv_name = p->p_sysent->sv_name; 2378 PROC_UNLOCK(p); 2379 return (sysctl_handle_string(oidp, sv_name, 0, req)); 2380 } 2381 2382 #ifdef KINFO_OVMENTRY_SIZE 2383 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); 2384 #endif 2385 2386 #ifdef COMPAT_FREEBSD7 2387 static int 2388 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) 2389 { 2390 vm_map_entry_t entry, tmp_entry; 2391 unsigned int last_timestamp, namelen; 2392 char *fullpath, *freepath; 2393 struct kinfo_ovmentry *kve; 2394 struct vattr va; 2395 struct ucred *cred; 2396 int error, *name; 2397 struct vnode *vp; 2398 struct proc *p; 2399 vm_map_t map; 2400 struct vmspace *vm; 2401 2402 namelen = arg2; 2403 if (namelen != 1) 2404 return (EINVAL); 2405 2406 name = (int *)arg1; 2407 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2408 if (error != 0) 2409 return (error); 2410 vm = vmspace_acquire_ref(p); 2411 if (vm == NULL) { 2412 PRELE(p); 2413 return (ESRCH); 2414 } 2415 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2416 2417 map = &vm->vm_map; 2418 vm_map_lock_read(map); 2419 VM_MAP_ENTRY_FOREACH(entry, map) { 2420 vm_object_t obj, tobj, lobj; 2421 vm_offset_t addr; 2422 2423 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2424 continue; 2425 2426 bzero(kve, sizeof(*kve)); 2427 kve->kve_structsize = sizeof(*kve); 2428 2429 kve->kve_private_resident = 0; 2430 obj = entry->object.vm_object; 2431 if (obj != NULL) { 2432 VM_OBJECT_RLOCK(obj); 2433 if (obj->shadow_count == 1) 2434 kve->kve_private_resident = 2435 obj->resident_page_count; 2436 } 2437 kve->kve_resident = 0; 2438 addr = entry->start; 2439 while (addr < entry->end) { 2440 if (pmap_extract(map->pmap, addr)) 2441 kve->kve_resident++; 2442 addr += PAGE_SIZE; 2443 } 2444 2445 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { 2446 if (tobj != obj) { 2447 VM_OBJECT_RLOCK(tobj); 2448 kve->kve_offset += tobj->backing_object_offset; 2449 } 2450 if (lobj != obj) 2451 VM_OBJECT_RUNLOCK(lobj); 2452 lobj = tobj; 2453 } 2454 2455 kve->kve_start = (void*)entry->start; 2456 kve->kve_end = (void*)entry->end; 2457 kve->kve_offset += (off_t)entry->offset; 2458 2459 if (entry->protection & VM_PROT_READ) 2460 kve->kve_protection |= KVME_PROT_READ; 2461 if (entry->protection & VM_PROT_WRITE) 2462 kve->kve_protection |= KVME_PROT_WRITE; 2463 if (entry->protection & VM_PROT_EXECUTE) 2464 kve->kve_protection |= KVME_PROT_EXEC; 2465 2466 if (entry->eflags & MAP_ENTRY_COW) 2467 kve->kve_flags |= KVME_FLAG_COW; 2468 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2469 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2470 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2471 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2472 2473 last_timestamp = map->timestamp; 2474 vm_map_unlock_read(map); 2475 2476 kve->kve_fileid = 0; 2477 kve->kve_fsid = 0; 2478 freepath = NULL; 2479 fullpath = ""; 2480 if (lobj) { 2481 kve->kve_type = vm_object_kvme_type(lobj, &vp); 2482 if (kve->kve_type == KVME_TYPE_MGTDEVICE) 2483 kve->kve_type = KVME_TYPE_UNKNOWN; 2484 if (vp != NULL) 2485 vref(vp); 2486 if (lobj != obj) 2487 VM_OBJECT_RUNLOCK(lobj); 2488 2489 kve->kve_ref_count = obj->ref_count; 2490 kve->kve_shadow_count = obj->shadow_count; 2491 VM_OBJECT_RUNLOCK(obj); 2492 if (vp != NULL) { 2493 vn_fullpath(vp, &fullpath, &freepath); 2494 cred = curthread->td_ucred; 2495 vn_lock(vp, LK_SHARED | LK_RETRY); 2496 if (VOP_GETATTR(vp, &va, cred) == 0) { 2497 kve->kve_fileid = va.va_fileid; 2498 /* truncate */ 2499 kve->kve_fsid = va.va_fsid; 2500 } 2501 vput(vp); 2502 } 2503 } else { 2504 kve->kve_type = KVME_TYPE_NONE; 2505 kve->kve_ref_count = 0; 2506 kve->kve_shadow_count = 0; 2507 } 2508 2509 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2510 if (freepath != NULL) 2511 free(freepath, M_TEMP); 2512 2513 error = SYSCTL_OUT(req, kve, sizeof(*kve)); 2514 vm_map_lock_read(map); 2515 if (error) 2516 break; 2517 if (last_timestamp != map->timestamp) { 2518 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2519 entry = tmp_entry; 2520 } 2521 } 2522 vm_map_unlock_read(map); 2523 vmspace_free(vm); 2524 PRELE(p); 2525 free(kve, M_TEMP); 2526 return (error); 2527 } 2528 #endif /* COMPAT_FREEBSD7 */ 2529 2530 #ifdef KINFO_VMENTRY_SIZE 2531 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); 2532 #endif 2533 2534 void 2535 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, 2536 int *resident_count, bool *super) 2537 { 2538 vm_object_t obj, tobj; 2539 vm_page_t m, m_adv; 2540 vm_offset_t addr; 2541 vm_paddr_t pa; 2542 vm_pindex_t pi, pi_adv, pindex; 2543 int incore; 2544 2545 *super = false; 2546 *resident_count = 0; 2547 if (vmmap_skip_res_cnt) 2548 return; 2549 2550 pa = 0; 2551 obj = entry->object.vm_object; 2552 addr = entry->start; 2553 m_adv = NULL; 2554 pi = OFF_TO_IDX(entry->offset); 2555 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { 2556 if (m_adv != NULL) { 2557 m = m_adv; 2558 } else { 2559 pi_adv = atop(entry->end - addr); 2560 pindex = pi; 2561 for (tobj = obj;; tobj = tobj->backing_object) { 2562 m = vm_page_find_least(tobj, pindex); 2563 if (m != NULL) { 2564 if (m->pindex == pindex) 2565 break; 2566 if (pi_adv > m->pindex - pindex) { 2567 pi_adv = m->pindex - pindex; 2568 m_adv = m; 2569 } 2570 } 2571 if (tobj->backing_object == NULL) 2572 goto next; 2573 pindex += OFF_TO_IDX(tobj-> 2574 backing_object_offset); 2575 } 2576 } 2577 m_adv = NULL; 2578 if (m->psind != 0 && addr + pagesizes[1] <= entry->end && 2579 (addr & (pagesizes[1] - 1)) == 0 && (incore = 2580 pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) { 2581 *super = true; 2582 /* 2583 * The virtual page might be smaller than the physical 2584 * page, so we use the page size reported by the pmap 2585 * rather than m->psind. 2586 */ 2587 pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]); 2588 } else { 2589 /* 2590 * We do not test the found page on validity. 2591 * Either the page is busy and being paged in, 2592 * or it was invalidated. The first case 2593 * should be counted as resident, the second 2594 * is not so clear; we do account both. 2595 */ 2596 pi_adv = 1; 2597 } 2598 *resident_count += pi_adv; 2599 next:; 2600 } 2601 } 2602 2603 /* 2604 * Must be called with the process locked and will return unlocked. 2605 */ 2606 int 2607 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags) 2608 { 2609 vm_map_entry_t entry, tmp_entry; 2610 struct vattr va; 2611 vm_map_t map; 2612 vm_object_t lobj, nobj, obj, tobj; 2613 char *fullpath, *freepath; 2614 struct kinfo_vmentry *kve; 2615 struct ucred *cred; 2616 struct vnode *vp; 2617 struct vmspace *vm; 2618 vm_offset_t addr; 2619 unsigned int last_timestamp; 2620 int error; 2621 bool guard, super; 2622 2623 PROC_LOCK_ASSERT(p, MA_OWNED); 2624 2625 _PHOLD(p); 2626 PROC_UNLOCK(p); 2627 vm = vmspace_acquire_ref(p); 2628 if (vm == NULL) { 2629 PRELE(p); 2630 return (ESRCH); 2631 } 2632 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO); 2633 2634 error = 0; 2635 map = &vm->vm_map; 2636 vm_map_lock_read(map); 2637 VM_MAP_ENTRY_FOREACH(entry, map) { 2638 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2639 continue; 2640 2641 addr = entry->end; 2642 bzero(kve, sizeof(*kve)); 2643 obj = entry->object.vm_object; 2644 if (obj != NULL) { 2645 if ((obj->flags & OBJ_ANON) != 0) 2646 kve->kve_obj = (uintptr_t)obj; 2647 2648 for (tobj = obj; tobj != NULL; 2649 tobj = tobj->backing_object) { 2650 VM_OBJECT_RLOCK(tobj); 2651 kve->kve_offset += tobj->backing_object_offset; 2652 lobj = tobj; 2653 } 2654 if (obj->backing_object == NULL) 2655 kve->kve_private_resident = 2656 obj->resident_page_count; 2657 kern_proc_vmmap_resident(map, entry, 2658 &kve->kve_resident, &super); 2659 if (super) 2660 kve->kve_flags |= KVME_FLAG_SUPER; 2661 for (tobj = obj; tobj != NULL; tobj = nobj) { 2662 nobj = tobj->backing_object; 2663 if (tobj != obj && tobj != lobj) 2664 VM_OBJECT_RUNLOCK(tobj); 2665 } 2666 } else { 2667 lobj = NULL; 2668 } 2669 2670 kve->kve_start = entry->start; 2671 kve->kve_end = entry->end; 2672 kve->kve_offset += entry->offset; 2673 2674 if (entry->protection & VM_PROT_READ) 2675 kve->kve_protection |= KVME_PROT_READ; 2676 if (entry->protection & VM_PROT_WRITE) 2677 kve->kve_protection |= KVME_PROT_WRITE; 2678 if (entry->protection & VM_PROT_EXECUTE) 2679 kve->kve_protection |= KVME_PROT_EXEC; 2680 2681 if (entry->eflags & MAP_ENTRY_COW) 2682 kve->kve_flags |= KVME_FLAG_COW; 2683 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2684 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2685 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2686 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2687 if (entry->eflags & MAP_ENTRY_GROWS_UP) 2688 kve->kve_flags |= KVME_FLAG_GROWS_UP; 2689 if (entry->eflags & MAP_ENTRY_GROWS_DOWN) 2690 kve->kve_flags |= KVME_FLAG_GROWS_DOWN; 2691 if (entry->eflags & MAP_ENTRY_USER_WIRED) 2692 kve->kve_flags |= KVME_FLAG_USER_WIRED; 2693 2694 guard = (entry->eflags & MAP_ENTRY_GUARD) != 0; 2695 2696 last_timestamp = map->timestamp; 2697 vm_map_unlock_read(map); 2698 2699 freepath = NULL; 2700 fullpath = ""; 2701 if (lobj != NULL) { 2702 kve->kve_type = vm_object_kvme_type(lobj, &vp); 2703 if (vp != NULL) 2704 vref(vp); 2705 if (lobj != obj) 2706 VM_OBJECT_RUNLOCK(lobj); 2707 2708 kve->kve_ref_count = obj->ref_count; 2709 kve->kve_shadow_count = obj->shadow_count; 2710 VM_OBJECT_RUNLOCK(obj); 2711 if (vp != NULL) { 2712 vn_fullpath(vp, &fullpath, &freepath); 2713 kve->kve_vn_type = vntype_to_kinfo(vp->v_type); 2714 cred = curthread->td_ucred; 2715 vn_lock(vp, LK_SHARED | LK_RETRY); 2716 if (VOP_GETATTR(vp, &va, cred) == 0) { 2717 kve->kve_vn_fileid = va.va_fileid; 2718 kve->kve_vn_fsid = va.va_fsid; 2719 kve->kve_vn_fsid_freebsd11 = 2720 kve->kve_vn_fsid; /* truncate */ 2721 kve->kve_vn_mode = 2722 MAKEIMODE(va.va_type, va.va_mode); 2723 kve->kve_vn_size = va.va_size; 2724 kve->kve_vn_rdev = va.va_rdev; 2725 kve->kve_vn_rdev_freebsd11 = 2726 kve->kve_vn_rdev; /* truncate */ 2727 kve->kve_status = KF_ATTR_VALID; 2728 } 2729 vput(vp); 2730 } 2731 } else { 2732 kve->kve_type = guard ? KVME_TYPE_GUARD : 2733 KVME_TYPE_NONE; 2734 kve->kve_ref_count = 0; 2735 kve->kve_shadow_count = 0; 2736 } 2737 2738 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2739 if (freepath != NULL) 2740 free(freepath, M_TEMP); 2741 2742 /* Pack record size down */ 2743 if ((flags & KERN_VMMAP_PACK_KINFO) != 0) 2744 kve->kve_structsize = 2745 offsetof(struct kinfo_vmentry, kve_path) + 2746 strlen(kve->kve_path) + 1; 2747 else 2748 kve->kve_structsize = sizeof(*kve); 2749 kve->kve_structsize = roundup(kve->kve_structsize, 2750 sizeof(uint64_t)); 2751 2752 /* Halt filling and truncate rather than exceeding maxlen */ 2753 if (maxlen != -1 && maxlen < kve->kve_structsize) { 2754 error = 0; 2755 vm_map_lock_read(map); 2756 break; 2757 } else if (maxlen != -1) 2758 maxlen -= kve->kve_structsize; 2759 2760 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0) 2761 error = ENOMEM; 2762 vm_map_lock_read(map); 2763 if (error != 0) 2764 break; 2765 if (last_timestamp != map->timestamp) { 2766 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2767 entry = tmp_entry; 2768 } 2769 } 2770 vm_map_unlock_read(map); 2771 vmspace_free(vm); 2772 PRELE(p); 2773 free(kve, M_TEMP); 2774 return (error); 2775 } 2776 2777 static int 2778 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) 2779 { 2780 struct proc *p; 2781 struct sbuf sb; 2782 u_int namelen; 2783 int error, error2, *name; 2784 2785 namelen = arg2; 2786 if (namelen != 1) 2787 return (EINVAL); 2788 2789 name = (int *)arg1; 2790 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); 2791 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2792 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); 2793 if (error != 0) { 2794 sbuf_delete(&sb); 2795 return (error); 2796 } 2797 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO); 2798 error2 = sbuf_finish(&sb); 2799 sbuf_delete(&sb); 2800 return (error != 0 ? error : error2); 2801 } 2802 2803 #if defined(STACK) || defined(DDB) 2804 static int 2805 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) 2806 { 2807 struct kinfo_kstack *kkstp; 2808 int error, i, *name, numthreads; 2809 lwpid_t *lwpidarray; 2810 struct thread *td; 2811 struct stack *st; 2812 struct sbuf sb; 2813 struct proc *p; 2814 u_int namelen; 2815 2816 namelen = arg2; 2817 if (namelen != 1) 2818 return (EINVAL); 2819 2820 name = (int *)arg1; 2821 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); 2822 if (error != 0) 2823 return (error); 2824 2825 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); 2826 st = stack_create(M_WAITOK); 2827 2828 lwpidarray = NULL; 2829 PROC_LOCK(p); 2830 do { 2831 if (lwpidarray != NULL) { 2832 free(lwpidarray, M_TEMP); 2833 lwpidarray = NULL; 2834 } 2835 numthreads = p->p_numthreads; 2836 PROC_UNLOCK(p); 2837 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, 2838 M_WAITOK | M_ZERO); 2839 PROC_LOCK(p); 2840 } while (numthreads < p->p_numthreads); 2841 2842 /* 2843 * XXXRW: During the below loop, execve(2) and countless other sorts 2844 * of changes could have taken place. Should we check to see if the 2845 * vmspace has been replaced, or the like, in order to prevent 2846 * giving a snapshot that spans, say, execve(2), with some threads 2847 * before and some after? Among other things, the credentials could 2848 * have changed, in which case the right to extract debug info might 2849 * no longer be assured. 2850 */ 2851 i = 0; 2852 FOREACH_THREAD_IN_PROC(p, td) { 2853 KASSERT(i < numthreads, 2854 ("sysctl_kern_proc_kstack: numthreads")); 2855 lwpidarray[i] = td->td_tid; 2856 i++; 2857 } 2858 PROC_UNLOCK(p); 2859 numthreads = i; 2860 for (i = 0; i < numthreads; i++) { 2861 td = tdfind(lwpidarray[i], p->p_pid); 2862 if (td == NULL) { 2863 continue; 2864 } 2865 bzero(kkstp, sizeof(*kkstp)); 2866 (void)sbuf_new(&sb, kkstp->kkst_trace, 2867 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); 2868 thread_lock(td); 2869 kkstp->kkst_tid = td->td_tid; 2870 if (stack_save_td(st, td) == 0) 2871 kkstp->kkst_state = KKST_STATE_STACKOK; 2872 else 2873 kkstp->kkst_state = KKST_STATE_RUNNING; 2874 thread_unlock(td); 2875 PROC_UNLOCK(p); 2876 stack_sbuf_print(&sb, st); 2877 sbuf_finish(&sb); 2878 sbuf_delete(&sb); 2879 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); 2880 if (error) 2881 break; 2882 } 2883 PRELE(p); 2884 if (lwpidarray != NULL) 2885 free(lwpidarray, M_TEMP); 2886 stack_destroy(st); 2887 free(kkstp, M_TEMP); 2888 return (error); 2889 } 2890 #endif 2891 2892 /* 2893 * This sysctl allows a process to retrieve the full list of groups from 2894 * itself or another process. 2895 */ 2896 static int 2897 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) 2898 { 2899 pid_t *pidp = (pid_t *)arg1; 2900 unsigned int arglen = arg2; 2901 struct proc *p; 2902 struct ucred *cred; 2903 int error; 2904 2905 if (arglen != 1) 2906 return (EINVAL); 2907 if (*pidp == -1) { /* -1 means this process */ 2908 p = req->td->td_proc; 2909 PROC_LOCK(p); 2910 } else { 2911 error = pget(*pidp, PGET_CANSEE, &p); 2912 if (error != 0) 2913 return (error); 2914 } 2915 2916 cred = crhold(p->p_ucred); 2917 PROC_UNLOCK(p); 2918 2919 error = SYSCTL_OUT(req, cred->cr_groups, 2920 cred->cr_ngroups * sizeof(gid_t)); 2921 crfree(cred); 2922 return (error); 2923 } 2924 2925 /* 2926 * This sysctl allows a process to retrieve or/and set the resource limit for 2927 * another process. 2928 */ 2929 static int 2930 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) 2931 { 2932 int *name = (int *)arg1; 2933 u_int namelen = arg2; 2934 struct rlimit rlim; 2935 struct proc *p; 2936 u_int which; 2937 int flags, error; 2938 2939 if (namelen != 2) 2940 return (EINVAL); 2941 2942 which = (u_int)name[1]; 2943 if (which >= RLIM_NLIMITS) 2944 return (EINVAL); 2945 2946 if (req->newptr != NULL && req->newlen != sizeof(rlim)) 2947 return (EINVAL); 2948 2949 flags = PGET_HOLD | PGET_NOTWEXIT; 2950 if (req->newptr != NULL) 2951 flags |= PGET_CANDEBUG; 2952 else 2953 flags |= PGET_CANSEE; 2954 error = pget((pid_t)name[0], flags, &p); 2955 if (error != 0) 2956 return (error); 2957 2958 /* 2959 * Retrieve limit. 2960 */ 2961 if (req->oldptr != NULL) { 2962 PROC_LOCK(p); 2963 lim_rlimit_proc(p, which, &rlim); 2964 PROC_UNLOCK(p); 2965 } 2966 error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); 2967 if (error != 0) 2968 goto errout; 2969 2970 /* 2971 * Set limit. 2972 */ 2973 if (req->newptr != NULL) { 2974 error = SYSCTL_IN(req, &rlim, sizeof(rlim)); 2975 if (error == 0) 2976 error = kern_proc_setrlimit(curthread, p, which, &rlim); 2977 } 2978 2979 errout: 2980 PRELE(p); 2981 return (error); 2982 } 2983 2984 /* 2985 * This sysctl allows a process to retrieve ps_strings structure location of 2986 * another process. 2987 */ 2988 static int 2989 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) 2990 { 2991 int *name = (int *)arg1; 2992 u_int namelen = arg2; 2993 struct proc *p; 2994 vm_offset_t ps_strings; 2995 int error; 2996 #ifdef COMPAT_FREEBSD32 2997 uint32_t ps_strings32; 2998 #endif 2999 3000 if (namelen != 1) 3001 return (EINVAL); 3002 3003 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 3004 if (error != 0) 3005 return (error); 3006 #ifdef COMPAT_FREEBSD32 3007 if ((req->flags & SCTL_MASK32) != 0) { 3008 /* 3009 * We return 0 if the 32 bit emulation request is for a 64 bit 3010 * process. 3011 */ 3012 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? 3013 PTROUT(PROC_PS_STRINGS(p)) : 0; 3014 PROC_UNLOCK(p); 3015 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); 3016 return (error); 3017 } 3018 #endif 3019 ps_strings = PROC_PS_STRINGS(p); 3020 PROC_UNLOCK(p); 3021 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); 3022 return (error); 3023 } 3024 3025 /* 3026 * This sysctl allows a process to retrieve umask of another process. 3027 */ 3028 static int 3029 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) 3030 { 3031 int *name = (int *)arg1; 3032 u_int namelen = arg2; 3033 struct proc *p; 3034 int error; 3035 u_short cmask; 3036 pid_t pid; 3037 3038 if (namelen != 1) 3039 return (EINVAL); 3040 3041 pid = (pid_t)name[0]; 3042 p = curproc; 3043 if (pid == p->p_pid || pid == 0) { 3044 cmask = p->p_pd->pd_cmask; 3045 goto out; 3046 } 3047 3048 error = pget(pid, PGET_WANTREAD, &p); 3049 if (error != 0) 3050 return (error); 3051 3052 cmask = p->p_pd->pd_cmask; 3053 PRELE(p); 3054 out: 3055 error = SYSCTL_OUT(req, &cmask, sizeof(cmask)); 3056 return (error); 3057 } 3058 3059 /* 3060 * This sysctl allows a process to set and retrieve binary osreldate of 3061 * another process. 3062 */ 3063 static int 3064 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) 3065 { 3066 int *name = (int *)arg1; 3067 u_int namelen = arg2; 3068 struct proc *p; 3069 int flags, error, osrel; 3070 3071 if (namelen != 1) 3072 return (EINVAL); 3073 3074 if (req->newptr != NULL && req->newlen != sizeof(osrel)) 3075 return (EINVAL); 3076 3077 flags = PGET_HOLD | PGET_NOTWEXIT; 3078 if (req->newptr != NULL) 3079 flags |= PGET_CANDEBUG; 3080 else 3081 flags |= PGET_CANSEE; 3082 error = pget((pid_t)name[0], flags, &p); 3083 if (error != 0) 3084 return (error); 3085 3086 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); 3087 if (error != 0) 3088 goto errout; 3089 3090 if (req->newptr != NULL) { 3091 error = SYSCTL_IN(req, &osrel, sizeof(osrel)); 3092 if (error != 0) 3093 goto errout; 3094 if (osrel < 0) { 3095 error = EINVAL; 3096 goto errout; 3097 } 3098 p->p_osrel = osrel; 3099 } 3100 errout: 3101 PRELE(p); 3102 return (error); 3103 } 3104 3105 static int 3106 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) 3107 { 3108 int *name = (int *)arg1; 3109 u_int namelen = arg2; 3110 struct proc *p; 3111 struct kinfo_sigtramp kst; 3112 const struct sysentvec *sv; 3113 int error; 3114 #ifdef COMPAT_FREEBSD32 3115 struct kinfo_sigtramp32 kst32; 3116 #endif 3117 3118 if (namelen != 1) 3119 return (EINVAL); 3120 3121 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 3122 if (error != 0) 3123 return (error); 3124 sv = p->p_sysent; 3125 #ifdef COMPAT_FREEBSD32 3126 if ((req->flags & SCTL_MASK32) != 0) { 3127 bzero(&kst32, sizeof(kst32)); 3128 if (SV_PROC_FLAG(p, SV_ILP32)) { 3129 if (PROC_HAS_SHP(p)) { 3130 kst32.ksigtramp_start = PROC_SIGCODE(p); 3131 kst32.ksigtramp_end = kst32.ksigtramp_start + 3132 ((sv->sv_flags & SV_DSO_SIG) == 0 ? 3133 *sv->sv_szsigcode : 3134 (uintptr_t)sv->sv_szsigcode); 3135 } else { 3136 kst32.ksigtramp_start = PROC_PS_STRINGS(p) - 3137 *sv->sv_szsigcode; 3138 kst32.ksigtramp_end = PROC_PS_STRINGS(p); 3139 } 3140 } 3141 PROC_UNLOCK(p); 3142 error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); 3143 return (error); 3144 } 3145 #endif 3146 bzero(&kst, sizeof(kst)); 3147 if (PROC_HAS_SHP(p)) { 3148 kst.ksigtramp_start = (char *)PROC_SIGCODE(p); 3149 kst.ksigtramp_end = (char *)kst.ksigtramp_start + 3150 ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode : 3151 (uintptr_t)sv->sv_szsigcode); 3152 } else { 3153 kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) - 3154 *sv->sv_szsigcode; 3155 kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p); 3156 } 3157 PROC_UNLOCK(p); 3158 error = SYSCTL_OUT(req, &kst, sizeof(kst)); 3159 return (error); 3160 } 3161 3162 static int 3163 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS) 3164 { 3165 int *name = (int *)arg1; 3166 u_int namelen = arg2; 3167 pid_t pid; 3168 struct proc *p; 3169 struct thread *td1; 3170 uintptr_t addr; 3171 #ifdef COMPAT_FREEBSD32 3172 uint32_t addr32; 3173 #endif 3174 int error; 3175 3176 if (namelen != 1 || req->newptr != NULL) 3177 return (EINVAL); 3178 3179 pid = (pid_t)name[0]; 3180 error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p); 3181 if (error != 0) 3182 return (error); 3183 3184 PROC_LOCK(p); 3185 #ifdef COMPAT_FREEBSD32 3186 if (SV_CURPROC_FLAG(SV_ILP32)) { 3187 if (!SV_PROC_FLAG(p, SV_ILP32)) { 3188 error = EINVAL; 3189 goto errlocked; 3190 } 3191 } 3192 #endif 3193 if (pid <= PID_MAX) { 3194 td1 = FIRST_THREAD_IN_PROC(p); 3195 } else { 3196 FOREACH_THREAD_IN_PROC(p, td1) { 3197 if (td1->td_tid == pid) 3198 break; 3199 } 3200 } 3201 if (td1 == NULL) { 3202 error = ESRCH; 3203 goto errlocked; 3204 } 3205 /* 3206 * The access to the private thread flags. It is fine as far 3207 * as no out-of-thin-air values are read from td_pflags, and 3208 * usermode read of the td_sigblock_ptr is racy inherently, 3209 * since target process might have already changed it 3210 * meantime. 3211 */ 3212 if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0) 3213 addr = (uintptr_t)td1->td_sigblock_ptr; 3214 else 3215 error = ENOTTY; 3216 3217 errlocked: 3218 _PRELE(p); 3219 PROC_UNLOCK(p); 3220 if (error != 0) 3221 return (error); 3222 3223 #ifdef COMPAT_FREEBSD32 3224 if (SV_CURPROC_FLAG(SV_ILP32)) { 3225 addr32 = addr; 3226 error = SYSCTL_OUT(req, &addr32, sizeof(addr32)); 3227 } else 3228 #endif 3229 error = SYSCTL_OUT(req, &addr, sizeof(addr)); 3230 return (error); 3231 } 3232 3233 static int 3234 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS) 3235 { 3236 struct kinfo_vm_layout kvm; 3237 struct proc *p; 3238 struct vmspace *vmspace; 3239 int error, *name; 3240 3241 name = (int *)arg1; 3242 if ((u_int)arg2 != 1) 3243 return (EINVAL); 3244 3245 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 3246 if (error != 0) 3247 return (error); 3248 #ifdef COMPAT_FREEBSD32 3249 if (SV_CURPROC_FLAG(SV_ILP32)) { 3250 if (!SV_PROC_FLAG(p, SV_ILP32)) { 3251 PROC_UNLOCK(p); 3252 return (EINVAL); 3253 } 3254 } 3255 #endif 3256 vmspace = vmspace_acquire_ref(p); 3257 PROC_UNLOCK(p); 3258 3259 memset(&kvm, 0, sizeof(kvm)); 3260 kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map); 3261 kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map); 3262 kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr; 3263 kvm.kvm_text_size = vmspace->vm_tsize; 3264 kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr; 3265 kvm.kvm_data_size = vmspace->vm_dsize; 3266 kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr; 3267 kvm.kvm_stack_size = vmspace->vm_ssize; 3268 kvm.kvm_shp_addr = vmspace->vm_shp_base; 3269 kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len; 3270 if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0) 3271 kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE; 3272 if ((vmspace->vm_map.flags & MAP_ASLR) != 0) 3273 kvm.kvm_map_flags |= KMAP_FLAG_ASLR; 3274 if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0) 3275 kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART; 3276 if ((vmspace->vm_map.flags & MAP_WXORX) != 0) 3277 kvm.kvm_map_flags |= KMAP_FLAG_WXORX; 3278 if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0) 3279 kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK; 3280 if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base && 3281 PROC_HAS_SHP(p)) 3282 kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE; 3283 3284 #ifdef COMPAT_FREEBSD32 3285 if (SV_CURPROC_FLAG(SV_ILP32)) { 3286 struct kinfo_vm_layout32 kvm32; 3287 3288 memset(&kvm32, 0, sizeof(kvm32)); 3289 kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr; 3290 kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr; 3291 kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr; 3292 kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size; 3293 kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr; 3294 kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size; 3295 kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr; 3296 kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size; 3297 kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr; 3298 kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size; 3299 kvm32.kvm_map_flags = kvm.kvm_map_flags; 3300 error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32)); 3301 goto out; 3302 } 3303 #endif 3304 3305 error = SYSCTL_OUT(req, &kvm, sizeof(kvm)); 3306 #ifdef COMPAT_FREEBSD32 3307 out: 3308 #endif 3309 vmspace_free(vmspace); 3310 return (error); 3311 } 3312 3313 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 3314 "Process table"); 3315 3316 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| 3317 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", 3318 "Return entire process table"); 3319 3320 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3321 sysctl_kern_proc, "Process table"); 3322 3323 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, 3324 sysctl_kern_proc, "Process table"); 3325 3326 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3327 sysctl_kern_proc, "Process table"); 3328 3329 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | 3330 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3331 3332 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 3333 sysctl_kern_proc, "Process table"); 3334 3335 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3336 sysctl_kern_proc, "Process table"); 3337 3338 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3339 sysctl_kern_proc, "Process table"); 3340 3341 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3342 sysctl_kern_proc, "Process table"); 3343 3344 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 3345 sysctl_kern_proc, "Return process table, no threads"); 3346 3347 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, 3348 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, 3349 sysctl_kern_proc_args, "Process argument list"); 3350 3351 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, 3352 sysctl_kern_proc_env, "Process environment"); 3353 3354 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | 3355 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); 3356 3357 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | 3358 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); 3359 3360 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | 3361 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, 3362 "Process syscall vector name (ABI type)"); 3363 3364 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, 3365 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3366 3367 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, 3368 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3369 3370 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, 3371 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3372 3373 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), 3374 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3375 3376 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, 3377 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3378 3379 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, 3380 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3381 3382 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, 3383 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3384 3385 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, 3386 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3387 3388 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, 3389 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, 3390 "Return process table, including threads"); 3391 3392 #ifdef COMPAT_FREEBSD7 3393 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | 3394 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); 3395 #endif 3396 3397 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | 3398 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); 3399 3400 #if defined(STACK) || defined(DDB) 3401 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | 3402 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); 3403 #endif 3404 3405 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | 3406 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); 3407 3408 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | 3409 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, 3410 "Process resource limits"); 3411 3412 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | 3413 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, 3414 "Process ps_strings location"); 3415 3416 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | 3417 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); 3418 3419 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | 3420 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, 3421 "Process binary osreldate"); 3422 3423 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | 3424 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, 3425 "Process signal trampoline location"); 3426 3427 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD | 3428 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk, 3429 "Thread sigfastblock address"); 3430 3431 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD | 3432 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout, 3433 "Process virtual address space layout info"); 3434 3435 static struct sx stop_all_proc_blocker; 3436 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk"); 3437 3438 bool 3439 stop_all_proc_block(void) 3440 { 3441 return (sx_xlock_sig(&stop_all_proc_blocker) == 0); 3442 } 3443 3444 void 3445 stop_all_proc_unblock(void) 3446 { 3447 sx_xunlock(&stop_all_proc_blocker); 3448 } 3449 3450 int allproc_gen; 3451 3452 /* 3453 * stop_all_proc() purpose is to stop all process which have usermode, 3454 * except current process for obvious reasons. This makes it somewhat 3455 * unreliable when invoked from multithreaded process. The service 3456 * must not be user-callable anyway. 3457 */ 3458 void 3459 stop_all_proc(void) 3460 { 3461 struct proc *cp, *p; 3462 int r, gen; 3463 bool restart, seen_stopped, seen_exiting, stopped_some; 3464 3465 if (!stop_all_proc_block()) 3466 return; 3467 3468 cp = curproc; 3469 allproc_loop: 3470 sx_xlock(&allproc_lock); 3471 gen = allproc_gen; 3472 seen_exiting = seen_stopped = stopped_some = restart = false; 3473 LIST_REMOVE(cp, p_list); 3474 LIST_INSERT_HEAD(&allproc, cp, p_list); 3475 for (;;) { 3476 p = LIST_NEXT(cp, p_list); 3477 if (p == NULL) 3478 break; 3479 LIST_REMOVE(cp, p_list); 3480 LIST_INSERT_AFTER(p, cp, p_list); 3481 PROC_LOCK(p); 3482 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP | 3483 P_STOPPED_SIG)) != 0) { 3484 PROC_UNLOCK(p); 3485 continue; 3486 } 3487 if ((p->p_flag2 & P2_WEXIT) != 0) { 3488 seen_exiting = true; 3489 PROC_UNLOCK(p); 3490 continue; 3491 } 3492 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 3493 /* 3494 * Stopped processes are tolerated when there 3495 * are no other processes which might continue 3496 * them. P_STOPPED_SINGLE but not 3497 * P_TOTAL_STOP process still has at least one 3498 * thread running. 3499 */ 3500 seen_stopped = true; 3501 PROC_UNLOCK(p); 3502 continue; 3503 } 3504 if ((p->p_flag & P_TRACED) != 0) { 3505 /* 3506 * thread_single() below cannot stop traced p, 3507 * so skip it. OTOH, we cannot require 3508 * restart because debugger might be either 3509 * already stopped or traced as well. 3510 */ 3511 PROC_UNLOCK(p); 3512 continue; 3513 } 3514 sx_xunlock(&allproc_lock); 3515 _PHOLD(p); 3516 r = thread_single(p, SINGLE_ALLPROC); 3517 if (r != 0) 3518 restart = true; 3519 else 3520 stopped_some = true; 3521 _PRELE(p); 3522 PROC_UNLOCK(p); 3523 sx_xlock(&allproc_lock); 3524 } 3525 /* Catch forked children we did not see in iteration. */ 3526 if (gen != allproc_gen) 3527 restart = true; 3528 sx_xunlock(&allproc_lock); 3529 if (restart || stopped_some || seen_exiting || seen_stopped) { 3530 kern_yield(PRI_USER); 3531 goto allproc_loop; 3532 } 3533 } 3534 3535 void 3536 resume_all_proc(void) 3537 { 3538 struct proc *cp, *p; 3539 3540 cp = curproc; 3541 sx_xlock(&allproc_lock); 3542 again: 3543 LIST_REMOVE(cp, p_list); 3544 LIST_INSERT_HEAD(&allproc, cp, p_list); 3545 for (;;) { 3546 p = LIST_NEXT(cp, p_list); 3547 if (p == NULL) 3548 break; 3549 LIST_REMOVE(cp, p_list); 3550 LIST_INSERT_AFTER(p, cp, p_list); 3551 PROC_LOCK(p); 3552 if ((p->p_flag & P_TOTAL_STOP) != 0) { 3553 sx_xunlock(&allproc_lock); 3554 _PHOLD(p); 3555 thread_single_end(p, SINGLE_ALLPROC); 3556 _PRELE(p); 3557 PROC_UNLOCK(p); 3558 sx_xlock(&allproc_lock); 3559 } else { 3560 PROC_UNLOCK(p); 3561 } 3562 } 3563 /* Did the loop above missed any stopped process ? */ 3564 FOREACH_PROC_IN_SYSTEM(p) { 3565 /* No need for proc lock. */ 3566 if ((p->p_flag & P_TOTAL_STOP) != 0) 3567 goto again; 3568 } 3569 sx_xunlock(&allproc_lock); 3570 3571 stop_all_proc_unblock(); 3572 } 3573 3574 /* #define TOTAL_STOP_DEBUG 1 */ 3575 #ifdef TOTAL_STOP_DEBUG 3576 volatile static int ap_resume; 3577 #include <sys/mount.h> 3578 3579 static int 3580 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) 3581 { 3582 int error, val; 3583 3584 val = 0; 3585 ap_resume = 0; 3586 error = sysctl_handle_int(oidp, &val, 0, req); 3587 if (error != 0 || req->newptr == NULL) 3588 return (error); 3589 if (val != 0) { 3590 stop_all_proc(); 3591 syncer_suspend(); 3592 while (ap_resume == 0) 3593 ; 3594 syncer_resume(); 3595 resume_all_proc(); 3596 } 3597 return (0); 3598 } 3599 3600 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | 3601 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, 3602 sysctl_debug_stop_all_proc, "I", 3603 ""); 3604 #endif 3605