1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1989, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_ddb.h" 38 #include "opt_ktrace.h" 39 #include "opt_kstack_pages.h" 40 #include "opt_stack.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/bitstring.h> 45 #include <sys/elf.h> 46 #include <sys/eventhandler.h> 47 #include <sys/exec.h> 48 #include <sys/jail.h> 49 #include <sys/kernel.h> 50 #include <sys/limits.h> 51 #include <sys/lock.h> 52 #include <sys/loginclass.h> 53 #include <sys/malloc.h> 54 #include <sys/mman.h> 55 #include <sys/mount.h> 56 #include <sys/mutex.h> 57 #include <sys/proc.h> 58 #include <sys/ptrace.h> 59 #include <sys/refcount.h> 60 #include <sys/resourcevar.h> 61 #include <sys/rwlock.h> 62 #include <sys/sbuf.h> 63 #include <sys/sysent.h> 64 #include <sys/sched.h> 65 #include <sys/smp.h> 66 #include <sys/stack.h> 67 #include <sys/stat.h> 68 #include <sys/sysctl.h> 69 #include <sys/filedesc.h> 70 #include <sys/tty.h> 71 #include <sys/signalvar.h> 72 #include <sys/sdt.h> 73 #include <sys/sx.h> 74 #include <sys/user.h> 75 #include <sys/vnode.h> 76 #include <sys/wait.h> 77 78 #ifdef DDB 79 #include <ddb/ddb.h> 80 #endif 81 82 #include <vm/vm.h> 83 #include <vm/vm_param.h> 84 #include <vm/vm_extern.h> 85 #include <vm/pmap.h> 86 #include <vm/vm_map.h> 87 #include <vm/vm_object.h> 88 #include <vm/vm_page.h> 89 #include <vm/uma.h> 90 91 #ifdef COMPAT_FREEBSD32 92 #include <compat/freebsd32/freebsd32.h> 93 #include <compat/freebsd32/freebsd32_util.h> 94 #endif 95 96 SDT_PROVIDER_DEFINE(proc); 97 98 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header"); 99 MALLOC_DEFINE(M_SESSION, "session", "session header"); 100 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); 101 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 102 103 static void doenterpgrp(struct proc *, struct pgrp *); 104 static void orphanpg(struct pgrp *pg); 105 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp); 106 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp); 107 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, 108 int preferthread); 109 static void pgadjustjobc(struct pgrp *pgrp, int entering); 110 static void pgdelete(struct pgrp *); 111 static int proc_ctor(void *mem, int size, void *arg, int flags); 112 static void proc_dtor(void *mem, int size, void *arg); 113 static int proc_init(void *mem, int size, int flags); 114 static void proc_fini(void *mem, int size); 115 static void pargs_free(struct pargs *pa); 116 117 /* 118 * Other process lists 119 */ 120 struct pidhashhead *pidhashtbl; 121 struct sx *pidhashtbl_lock; 122 u_long pidhash; 123 u_long pidhashlock; 124 struct pgrphashhead *pgrphashtbl; 125 u_long pgrphash; 126 struct proclist allproc; 127 struct sx __exclusive_cache_line allproc_lock; 128 struct sx __exclusive_cache_line proctree_lock; 129 struct mtx __exclusive_cache_line ppeers_lock; 130 struct mtx __exclusive_cache_line procid_lock; 131 uma_zone_t proc_zone; 132 133 /* 134 * The offset of various fields in struct proc and struct thread. 135 * These are used by kernel debuggers to enumerate kernel threads and 136 * processes. 137 */ 138 const int proc_off_p_pid = offsetof(struct proc, p_pid); 139 const int proc_off_p_comm = offsetof(struct proc, p_comm); 140 const int proc_off_p_list = offsetof(struct proc, p_list); 141 const int proc_off_p_threads = offsetof(struct proc, p_threads); 142 const int thread_off_td_tid = offsetof(struct thread, td_tid); 143 const int thread_off_td_name = offsetof(struct thread, td_name); 144 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu); 145 const int thread_off_td_pcb = offsetof(struct thread, td_pcb); 146 const int thread_off_td_plist = offsetof(struct thread, td_plist); 147 148 EVENTHANDLER_LIST_DEFINE(process_ctor); 149 EVENTHANDLER_LIST_DEFINE(process_dtor); 150 EVENTHANDLER_LIST_DEFINE(process_init); 151 EVENTHANDLER_LIST_DEFINE(process_fini); 152 EVENTHANDLER_LIST_DEFINE(process_exit); 153 EVENTHANDLER_LIST_DEFINE(process_fork); 154 EVENTHANDLER_LIST_DEFINE(process_exec); 155 156 int kstack_pages = KSTACK_PAGES; 157 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, 158 "Kernel stack size in pages"); 159 static int vmmap_skip_res_cnt = 0; 160 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW, 161 &vmmap_skip_res_cnt, 0, 162 "Skip calculation of the pages resident count in kern.proc.vmmap"); 163 164 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); 165 #ifdef COMPAT_FREEBSD32 166 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE); 167 #endif 168 169 /* 170 * Initialize global process hashing structures. 171 */ 172 void 173 procinit(void) 174 { 175 u_long i; 176 177 sx_init(&allproc_lock, "allproc"); 178 sx_init(&proctree_lock, "proctree"); 179 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF); 180 mtx_init(&procid_lock, "procid", NULL, MTX_DEF); 181 LIST_INIT(&allproc); 182 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); 183 pidhashlock = (pidhash + 1) / 64; 184 if (pidhashlock > 0) 185 pidhashlock--; 186 pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1), 187 M_PROC, M_WAITOK | M_ZERO); 188 for (i = 0; i < pidhashlock + 1; i++) 189 sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK); 190 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); 191 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(), 192 proc_ctor, proc_dtor, proc_init, proc_fini, 193 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 194 uihashinit(); 195 } 196 197 /* 198 * Prepare a proc for use. 199 */ 200 static int 201 proc_ctor(void *mem, int size, void *arg, int flags) 202 { 203 struct proc *p; 204 struct thread *td; 205 206 p = (struct proc *)mem; 207 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p); 208 td = FIRST_THREAD_IN_PROC(p); 209 if (td != NULL) { 210 /* Make sure all thread constructors are executed */ 211 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td); 212 } 213 return (0); 214 } 215 216 /* 217 * Reclaim a proc after use. 218 */ 219 static void 220 proc_dtor(void *mem, int size, void *arg) 221 { 222 struct proc *p; 223 struct thread *td; 224 225 /* INVARIANTS checks go here */ 226 p = (struct proc *)mem; 227 td = FIRST_THREAD_IN_PROC(p); 228 if (td != NULL) { 229 #ifdef INVARIANTS 230 KASSERT((p->p_numthreads == 1), 231 ("bad number of threads in exiting process")); 232 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr")); 233 #endif 234 /* Free all OSD associated to this thread. */ 235 osd_thread_exit(td); 236 td_softdep_cleanup(td); 237 MPASS(td->td_su == NULL); 238 239 /* Make sure all thread destructors are executed */ 240 EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td); 241 } 242 EVENTHANDLER_DIRECT_INVOKE(process_dtor, p); 243 if (p->p_ksi != NULL) 244 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue")); 245 } 246 247 /* 248 * Initialize type-stable parts of a proc (when newly created). 249 */ 250 static int 251 proc_init(void *mem, int size, int flags) 252 { 253 struct proc *p; 254 255 p = (struct proc *)mem; 256 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW); 257 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW); 258 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW); 259 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW); 260 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW); 261 cv_init(&p->p_pwait, "ppwait"); 262 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 263 EVENTHANDLER_DIRECT_INVOKE(process_init, p); 264 p->p_stats = pstats_alloc(); 265 p->p_pgrp = NULL; 266 return (0); 267 } 268 269 /* 270 * UMA should ensure that this function is never called. 271 * Freeing a proc structure would violate type stability. 272 */ 273 static void 274 proc_fini(void *mem, int size) 275 { 276 #ifdef notnow 277 struct proc *p; 278 279 p = (struct proc *)mem; 280 EVENTHANDLER_DIRECT_INVOKE(process_fini, p); 281 pstats_free(p->p_stats); 282 thread_free(FIRST_THREAD_IN_PROC(p)); 283 mtx_destroy(&p->p_mtx); 284 if (p->p_ksi != NULL) 285 ksiginfo_free(p->p_ksi); 286 #else 287 panic("proc reclaimed"); 288 #endif 289 } 290 291 /* 292 * PID space management. 293 * 294 * These bitmaps are used by fork_findpid. 295 */ 296 bitstr_t bit_decl(proc_id_pidmap, PID_MAX); 297 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX); 298 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX); 299 bitstr_t bit_decl(proc_id_reapmap, PID_MAX); 300 301 static bitstr_t *proc_id_array[] = { 302 proc_id_pidmap, 303 proc_id_grpidmap, 304 proc_id_sessidmap, 305 proc_id_reapmap, 306 }; 307 308 void 309 proc_id_set(int type, pid_t id) 310 { 311 312 KASSERT(type >= 0 && type < nitems(proc_id_array), 313 ("invalid type %d\n", type)); 314 mtx_lock(&procid_lock); 315 KASSERT(bit_test(proc_id_array[type], id) == 0, 316 ("bit %d already set in %d\n", id, type)); 317 bit_set(proc_id_array[type], id); 318 mtx_unlock(&procid_lock); 319 } 320 321 void 322 proc_id_set_cond(int type, pid_t id) 323 { 324 325 KASSERT(type >= 0 && type < nitems(proc_id_array), 326 ("invalid type %d\n", type)); 327 if (bit_test(proc_id_array[type], id)) 328 return; 329 mtx_lock(&procid_lock); 330 bit_set(proc_id_array[type], id); 331 mtx_unlock(&procid_lock); 332 } 333 334 void 335 proc_id_clear(int type, pid_t id) 336 { 337 338 KASSERT(type >= 0 && type < nitems(proc_id_array), 339 ("invalid type %d\n", type)); 340 mtx_lock(&procid_lock); 341 KASSERT(bit_test(proc_id_array[type], id) != 0, 342 ("bit %d not set in %d\n", id, type)); 343 bit_clear(proc_id_array[type], id); 344 mtx_unlock(&procid_lock); 345 } 346 347 /* 348 * Is p an inferior of the current process? 349 */ 350 int 351 inferior(struct proc *p) 352 { 353 354 sx_assert(&proctree_lock, SX_LOCKED); 355 PROC_LOCK_ASSERT(p, MA_OWNED); 356 for (; p != curproc; p = proc_realparent(p)) { 357 if (p->p_pid == 0) 358 return (0); 359 } 360 return (1); 361 } 362 363 /* 364 * Shared lock all the pid hash lists. 365 */ 366 void 367 pidhash_slockall(void) 368 { 369 u_long i; 370 371 for (i = 0; i < pidhashlock + 1; i++) 372 sx_slock(&pidhashtbl_lock[i]); 373 } 374 375 /* 376 * Shared unlock all the pid hash lists. 377 */ 378 void 379 pidhash_sunlockall(void) 380 { 381 u_long i; 382 383 for (i = 0; i < pidhashlock + 1; i++) 384 sx_sunlock(&pidhashtbl_lock[i]); 385 } 386 387 /* 388 * Similar to pfind_any(), this function finds zombies. 389 */ 390 struct proc * 391 pfind_any_locked(pid_t pid) 392 { 393 struct proc *p; 394 395 sx_assert(PIDHASHLOCK(pid), SX_LOCKED); 396 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 397 if (p->p_pid == pid) { 398 PROC_LOCK(p); 399 if (p->p_state == PRS_NEW) { 400 PROC_UNLOCK(p); 401 p = NULL; 402 } 403 break; 404 } 405 } 406 return (p); 407 } 408 409 /* 410 * Locate a process by number. 411 * 412 * By not returning processes in the PRS_NEW state, we allow callers to avoid 413 * testing for that condition to avoid dereferencing p_ucred, et al. 414 */ 415 static __always_inline struct proc * 416 _pfind(pid_t pid, bool zombie) 417 { 418 struct proc *p; 419 420 p = curproc; 421 if (p->p_pid == pid) { 422 PROC_LOCK(p); 423 return (p); 424 } 425 sx_slock(PIDHASHLOCK(pid)); 426 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 427 if (p->p_pid == pid) { 428 PROC_LOCK(p); 429 if (p->p_state == PRS_NEW || 430 (!zombie && p->p_state == PRS_ZOMBIE)) { 431 PROC_UNLOCK(p); 432 p = NULL; 433 } 434 break; 435 } 436 } 437 sx_sunlock(PIDHASHLOCK(pid)); 438 return (p); 439 } 440 441 struct proc * 442 pfind(pid_t pid) 443 { 444 445 return (_pfind(pid, false)); 446 } 447 448 /* 449 * Same as pfind but allow zombies. 450 */ 451 struct proc * 452 pfind_any(pid_t pid) 453 { 454 455 return (_pfind(pid, true)); 456 } 457 458 /* 459 * Locate a process group by number. 460 * The caller must hold proctree_lock. 461 */ 462 struct pgrp * 463 pgfind(pid_t pgid) 464 { 465 struct pgrp *pgrp; 466 467 sx_assert(&proctree_lock, SX_LOCKED); 468 469 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { 470 if (pgrp->pg_id == pgid) { 471 PGRP_LOCK(pgrp); 472 return (pgrp); 473 } 474 } 475 return (NULL); 476 } 477 478 /* 479 * Locate process and do additional manipulations, depending on flags. 480 */ 481 int 482 pget(pid_t pid, int flags, struct proc **pp) 483 { 484 struct proc *p; 485 struct thread *td1; 486 int error; 487 488 p = curproc; 489 if (p->p_pid == pid) { 490 PROC_LOCK(p); 491 } else { 492 p = NULL; 493 if (pid <= PID_MAX) { 494 if ((flags & PGET_NOTWEXIT) == 0) 495 p = pfind_any(pid); 496 else 497 p = pfind(pid); 498 } else if ((flags & PGET_NOTID) == 0) { 499 td1 = tdfind(pid, -1); 500 if (td1 != NULL) 501 p = td1->td_proc; 502 } 503 if (p == NULL) 504 return (ESRCH); 505 if ((flags & PGET_CANSEE) != 0) { 506 error = p_cansee(curthread, p); 507 if (error != 0) 508 goto errout; 509 } 510 } 511 if ((flags & PGET_CANDEBUG) != 0) { 512 error = p_candebug(curthread, p); 513 if (error != 0) 514 goto errout; 515 } 516 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) { 517 error = EPERM; 518 goto errout; 519 } 520 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) { 521 error = ESRCH; 522 goto errout; 523 } 524 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) { 525 /* 526 * XXXRW: Not clear ESRCH is the right error during proc 527 * execve(). 528 */ 529 error = ESRCH; 530 goto errout; 531 } 532 if ((flags & PGET_HOLD) != 0) { 533 _PHOLD(p); 534 PROC_UNLOCK(p); 535 } 536 *pp = p; 537 return (0); 538 errout: 539 PROC_UNLOCK(p); 540 return (error); 541 } 542 543 /* 544 * Create a new process group. 545 * pgid must be equal to the pid of p. 546 * Begin a new session if required. 547 */ 548 int 549 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess) 550 { 551 552 sx_assert(&proctree_lock, SX_XLOCKED); 553 554 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL")); 555 KASSERT(p->p_pid == pgid, 556 ("enterpgrp: new pgrp and pid != pgid")); 557 KASSERT(pgfind(pgid) == NULL, 558 ("enterpgrp: pgrp with pgid exists")); 559 KASSERT(!SESS_LEADER(p), 560 ("enterpgrp: session leader attempted setpgrp")); 561 562 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 563 564 if (sess != NULL) { 565 /* 566 * new session 567 */ 568 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); 569 PROC_LOCK(p); 570 p->p_flag &= ~P_CONTROLT; 571 PROC_UNLOCK(p); 572 PGRP_LOCK(pgrp); 573 sess->s_leader = p; 574 sess->s_sid = p->p_pid; 575 proc_id_set(PROC_ID_SESSION, p->p_pid); 576 refcount_init(&sess->s_count, 1); 577 sess->s_ttyvp = NULL; 578 sess->s_ttydp = NULL; 579 sess->s_ttyp = NULL; 580 bcopy(p->p_session->s_login, sess->s_login, 581 sizeof(sess->s_login)); 582 pgrp->pg_session = sess; 583 KASSERT(p == curproc, 584 ("enterpgrp: mksession and p != curproc")); 585 } else { 586 pgrp->pg_session = p->p_session; 587 sess_hold(pgrp->pg_session); 588 PGRP_LOCK(pgrp); 589 } 590 pgrp->pg_id = pgid; 591 proc_id_set(PROC_ID_GROUP, p->p_pid); 592 LIST_INIT(&pgrp->pg_members); 593 594 /* 595 * As we have an exclusive lock of proctree_lock, 596 * this should not deadlock. 597 */ 598 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); 599 pgrp->pg_jobc = 0; 600 SLIST_INIT(&pgrp->pg_sigiolst); 601 PGRP_UNLOCK(pgrp); 602 603 doenterpgrp(p, pgrp); 604 605 return (0); 606 } 607 608 /* 609 * Move p to an existing process group 610 */ 611 int 612 enterthispgrp(struct proc *p, struct pgrp *pgrp) 613 { 614 615 sx_assert(&proctree_lock, SX_XLOCKED); 616 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 617 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 618 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 619 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 620 KASSERT(pgrp->pg_session == p->p_session, 621 ("%s: pgrp's session %p, p->p_session %p.\n", 622 __func__, 623 pgrp->pg_session, 624 p->p_session)); 625 KASSERT(pgrp != p->p_pgrp, 626 ("%s: p belongs to pgrp.", __func__)); 627 628 doenterpgrp(p, pgrp); 629 630 return (0); 631 } 632 633 /* 634 * Move p to a process group 635 */ 636 static void 637 doenterpgrp(struct proc *p, struct pgrp *pgrp) 638 { 639 struct pgrp *savepgrp; 640 641 sx_assert(&proctree_lock, SX_XLOCKED); 642 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 643 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 644 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 645 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 646 647 savepgrp = p->p_pgrp; 648 649 /* 650 * Adjust eligibility of affected pgrps to participate in job control. 651 * Increment eligibility counts before decrementing, otherwise we 652 * could reach 0 spuriously during the first call. 653 */ 654 fixjobc(p, pgrp, 1); 655 fixjobc(p, p->p_pgrp, 0); 656 657 PGRP_LOCK(pgrp); 658 PGRP_LOCK(savepgrp); 659 PROC_LOCK(p); 660 LIST_REMOVE(p, p_pglist); 661 p->p_pgrp = pgrp; 662 PROC_UNLOCK(p); 663 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 664 PGRP_UNLOCK(savepgrp); 665 PGRP_UNLOCK(pgrp); 666 if (LIST_EMPTY(&savepgrp->pg_members)) 667 pgdelete(savepgrp); 668 } 669 670 /* 671 * remove process from process group 672 */ 673 int 674 leavepgrp(struct proc *p) 675 { 676 struct pgrp *savepgrp; 677 678 sx_assert(&proctree_lock, SX_XLOCKED); 679 savepgrp = p->p_pgrp; 680 PGRP_LOCK(savepgrp); 681 PROC_LOCK(p); 682 LIST_REMOVE(p, p_pglist); 683 p->p_pgrp = NULL; 684 PROC_UNLOCK(p); 685 PGRP_UNLOCK(savepgrp); 686 if (LIST_EMPTY(&savepgrp->pg_members)) 687 pgdelete(savepgrp); 688 return (0); 689 } 690 691 /* 692 * delete a process group 693 */ 694 static void 695 pgdelete(struct pgrp *pgrp) 696 { 697 struct session *savesess; 698 struct tty *tp; 699 700 sx_assert(&proctree_lock, SX_XLOCKED); 701 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 702 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 703 704 /* 705 * Reset any sigio structures pointing to us as a result of 706 * F_SETOWN with our pgid. 707 */ 708 funsetownlst(&pgrp->pg_sigiolst); 709 710 PGRP_LOCK(pgrp); 711 tp = pgrp->pg_session->s_ttyp; 712 LIST_REMOVE(pgrp, pg_hash); 713 savesess = pgrp->pg_session; 714 PGRP_UNLOCK(pgrp); 715 716 /* Remove the reference to the pgrp before deallocating it. */ 717 if (tp != NULL) { 718 tty_lock(tp); 719 tty_rel_pgrp(tp, pgrp); 720 } 721 722 proc_id_clear(PROC_ID_GROUP, pgrp->pg_id); 723 mtx_destroy(&pgrp->pg_mtx); 724 free(pgrp, M_PGRP); 725 sess_release(savesess); 726 } 727 728 static void 729 pgadjustjobc(struct pgrp *pgrp, int entering) 730 { 731 732 PGRP_LOCK(pgrp); 733 if (entering) { 734 #ifdef notyet 735 MPASS(pgrp->pg_jobc >= 0); 736 #endif 737 pgrp->pg_jobc++; 738 } else { 739 #ifdef notyet 740 MPASS(pgrp->pg_jobc > 0); 741 #endif 742 --pgrp->pg_jobc; 743 if (pgrp->pg_jobc == 0) 744 orphanpg(pgrp); 745 } 746 PGRP_UNLOCK(pgrp); 747 } 748 749 /* 750 * Adjust pgrp jobc counters when specified process changes process group. 751 * We count the number of processes in each process group that "qualify" 752 * the group for terminal job control (those with a parent in a different 753 * process group of the same session). If that count reaches zero, the 754 * process group becomes orphaned. Check both the specified process' 755 * process group and that of its children. 756 * entering == 0 => p is leaving specified group. 757 * entering == 1 => p is entering specified group. 758 */ 759 void 760 fixjobc(struct proc *p, struct pgrp *pgrp, int entering) 761 { 762 struct pgrp *hispgrp; 763 struct session *mysession; 764 struct proc *q; 765 766 sx_assert(&proctree_lock, SX_LOCKED); 767 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 768 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 769 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 770 771 /* 772 * Check p's parent to see whether p qualifies its own process 773 * group; if so, adjust count for p's process group. 774 */ 775 mysession = pgrp->pg_session; 776 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp && 777 hispgrp->pg_session == mysession) 778 pgadjustjobc(pgrp, entering); 779 780 /* 781 * Check this process' children to see whether they qualify 782 * their process groups; if so, adjust counts for children's 783 * process groups. 784 */ 785 LIST_FOREACH(q, &p->p_children, p_sibling) { 786 hispgrp = q->p_pgrp; 787 if (hispgrp == pgrp || 788 hispgrp->pg_session != mysession) 789 continue; 790 if (q->p_state == PRS_ZOMBIE) 791 continue; 792 pgadjustjobc(hispgrp, entering); 793 } 794 } 795 796 void 797 killjobc(void) 798 { 799 struct session *sp; 800 struct tty *tp; 801 struct proc *p; 802 struct vnode *ttyvp; 803 804 p = curproc; 805 MPASS(p->p_flag & P_WEXIT); 806 /* 807 * Do a quick check to see if there is anything to do with the 808 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc(). 809 */ 810 PROC_LOCK(p); 811 if (!SESS_LEADER(p) && 812 (p->p_pgrp == p->p_pptr->p_pgrp) && 813 LIST_EMPTY(&p->p_children)) { 814 PROC_UNLOCK(p); 815 return; 816 } 817 PROC_UNLOCK(p); 818 819 sx_xlock(&proctree_lock); 820 if (SESS_LEADER(p)) { 821 sp = p->p_session; 822 823 /* 824 * s_ttyp is not zero'd; we use this to indicate that 825 * the session once had a controlling terminal. (for 826 * logging and informational purposes) 827 */ 828 SESS_LOCK(sp); 829 ttyvp = sp->s_ttyvp; 830 tp = sp->s_ttyp; 831 sp->s_ttyvp = NULL; 832 sp->s_ttydp = NULL; 833 sp->s_leader = NULL; 834 SESS_UNLOCK(sp); 835 836 /* 837 * Signal foreground pgrp and revoke access to 838 * controlling terminal if it has not been revoked 839 * already. 840 * 841 * Because the TTY may have been revoked in the mean 842 * time and could already have a new session associated 843 * with it, make sure we don't send a SIGHUP to a 844 * foreground process group that does not belong to this 845 * session. 846 */ 847 848 if (tp != NULL) { 849 tty_lock(tp); 850 if (tp->t_session == sp) 851 tty_signal_pgrp(tp, SIGHUP); 852 tty_unlock(tp); 853 } 854 855 if (ttyvp != NULL) { 856 sx_xunlock(&proctree_lock); 857 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) { 858 VOP_REVOKE(ttyvp, REVOKEALL); 859 VOP_UNLOCK(ttyvp); 860 } 861 vrele(ttyvp); 862 sx_xlock(&proctree_lock); 863 } 864 } 865 fixjobc(p, p->p_pgrp, 0); 866 sx_xunlock(&proctree_lock); 867 } 868 869 /* 870 * A process group has become orphaned; 871 * if there are any stopped processes in the group, 872 * hang-up all process in that group. 873 */ 874 static void 875 orphanpg(struct pgrp *pg) 876 { 877 struct proc *p; 878 879 PGRP_LOCK_ASSERT(pg, MA_OWNED); 880 881 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 882 PROC_LOCK(p); 883 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) { 884 PROC_UNLOCK(p); 885 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 886 PROC_LOCK(p); 887 kern_psignal(p, SIGHUP); 888 kern_psignal(p, SIGCONT); 889 PROC_UNLOCK(p); 890 } 891 return; 892 } 893 PROC_UNLOCK(p); 894 } 895 } 896 897 void 898 sess_hold(struct session *s) 899 { 900 901 refcount_acquire(&s->s_count); 902 } 903 904 void 905 sess_release(struct session *s) 906 { 907 908 if (refcount_release(&s->s_count)) { 909 if (s->s_ttyp != NULL) { 910 tty_lock(s->s_ttyp); 911 tty_rel_sess(s->s_ttyp, s); 912 } 913 proc_id_clear(PROC_ID_SESSION, s->s_sid); 914 mtx_destroy(&s->s_mtx); 915 free(s, M_SESSION); 916 } 917 } 918 919 #ifdef DDB 920 921 DB_SHOW_COMMAND(pgrpdump, pgrpdump) 922 { 923 struct pgrp *pgrp; 924 struct proc *p; 925 int i; 926 927 for (i = 0; i <= pgrphash; i++) { 928 if (!LIST_EMPTY(&pgrphashtbl[i])) { 929 printf("\tindx %d\n", i); 930 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { 931 printf( 932 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n", 933 (void *)pgrp, (long)pgrp->pg_id, 934 (void *)pgrp->pg_session, 935 pgrp->pg_session->s_count, 936 (void *)LIST_FIRST(&pgrp->pg_members)); 937 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 938 printf("\t\tpid %ld addr %p pgrp %p\n", 939 (long)p->p_pid, (void *)p, 940 (void *)p->p_pgrp); 941 } 942 } 943 } 944 } 945 } 946 #endif /* DDB */ 947 948 /* 949 * Calculate the kinfo_proc members which contain process-wide 950 * informations. 951 * Must be called with the target process locked. 952 */ 953 static void 954 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp) 955 { 956 struct thread *td; 957 958 PROC_LOCK_ASSERT(p, MA_OWNED); 959 960 kp->ki_estcpu = 0; 961 kp->ki_pctcpu = 0; 962 FOREACH_THREAD_IN_PROC(p, td) { 963 thread_lock(td); 964 kp->ki_pctcpu += sched_pctcpu(td); 965 kp->ki_estcpu += sched_estcpu(td); 966 thread_unlock(td); 967 } 968 } 969 970 /* 971 * Clear kinfo_proc and fill in any information that is common 972 * to all threads in the process. 973 * Must be called with the target process locked. 974 */ 975 static void 976 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp) 977 { 978 struct thread *td0; 979 struct tty *tp; 980 struct session *sp; 981 struct ucred *cred; 982 struct sigacts *ps; 983 struct timeval boottime; 984 985 PROC_LOCK_ASSERT(p, MA_OWNED); 986 bzero(kp, sizeof(*kp)); 987 988 kp->ki_structsize = sizeof(*kp); 989 kp->ki_paddr = p; 990 kp->ki_addr =/* p->p_addr; */0; /* XXX */ 991 kp->ki_args = p->p_args; 992 kp->ki_textvp = p->p_textvp; 993 #ifdef KTRACE 994 kp->ki_tracep = p->p_tracevp; 995 kp->ki_traceflag = p->p_traceflag; 996 #endif 997 kp->ki_fd = p->p_fd; 998 kp->ki_vmspace = p->p_vmspace; 999 kp->ki_flag = p->p_flag; 1000 kp->ki_flag2 = p->p_flag2; 1001 cred = p->p_ucred; 1002 if (cred) { 1003 kp->ki_uid = cred->cr_uid; 1004 kp->ki_ruid = cred->cr_ruid; 1005 kp->ki_svuid = cred->cr_svuid; 1006 kp->ki_cr_flags = 0; 1007 if (cred->cr_flags & CRED_FLAG_CAPMODE) 1008 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE; 1009 /* XXX bde doesn't like KI_NGROUPS */ 1010 if (cred->cr_ngroups > KI_NGROUPS) { 1011 kp->ki_ngroups = KI_NGROUPS; 1012 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW; 1013 } else 1014 kp->ki_ngroups = cred->cr_ngroups; 1015 bcopy(cred->cr_groups, kp->ki_groups, 1016 kp->ki_ngroups * sizeof(gid_t)); 1017 kp->ki_rgid = cred->cr_rgid; 1018 kp->ki_svgid = cred->cr_svgid; 1019 /* If jailed(cred), emulate the old P_JAILED flag. */ 1020 if (jailed(cred)) { 1021 kp->ki_flag |= P_JAILED; 1022 /* If inside the jail, use 0 as a jail ID. */ 1023 if (cred->cr_prison != curthread->td_ucred->cr_prison) 1024 kp->ki_jid = cred->cr_prison->pr_id; 1025 } 1026 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name, 1027 sizeof(kp->ki_loginclass)); 1028 } 1029 ps = p->p_sigacts; 1030 if (ps) { 1031 mtx_lock(&ps->ps_mtx); 1032 kp->ki_sigignore = ps->ps_sigignore; 1033 kp->ki_sigcatch = ps->ps_sigcatch; 1034 mtx_unlock(&ps->ps_mtx); 1035 } 1036 if (p->p_state != PRS_NEW && 1037 p->p_state != PRS_ZOMBIE && 1038 p->p_vmspace != NULL) { 1039 struct vmspace *vm = p->p_vmspace; 1040 1041 kp->ki_size = vm->vm_map.size; 1042 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ 1043 FOREACH_THREAD_IN_PROC(p, td0) { 1044 if (!TD_IS_SWAPPED(td0)) 1045 kp->ki_rssize += td0->td_kstack_pages; 1046 } 1047 kp->ki_swrss = vm->vm_swrss; 1048 kp->ki_tsize = vm->vm_tsize; 1049 kp->ki_dsize = vm->vm_dsize; 1050 kp->ki_ssize = vm->vm_ssize; 1051 } else if (p->p_state == PRS_ZOMBIE) 1052 kp->ki_stat = SZOMB; 1053 if (kp->ki_flag & P_INMEM) 1054 kp->ki_sflag = PS_INMEM; 1055 else 1056 kp->ki_sflag = 0; 1057 /* Calculate legacy swtime as seconds since 'swtick'. */ 1058 kp->ki_swtime = (ticks - p->p_swtick) / hz; 1059 kp->ki_pid = p->p_pid; 1060 kp->ki_nice = p->p_nice; 1061 kp->ki_fibnum = p->p_fibnum; 1062 kp->ki_start = p->p_stats->p_start; 1063 getboottime(&boottime); 1064 timevaladd(&kp->ki_start, &boottime); 1065 PROC_STATLOCK(p); 1066 rufetch(p, &kp->ki_rusage); 1067 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime); 1068 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime); 1069 PROC_STATUNLOCK(p); 1070 calccru(p, &kp->ki_childutime, &kp->ki_childstime); 1071 /* Some callers want child times in a single value. */ 1072 kp->ki_childtime = kp->ki_childstime; 1073 timevaladd(&kp->ki_childtime, &kp->ki_childutime); 1074 1075 FOREACH_THREAD_IN_PROC(p, td0) 1076 kp->ki_cow += td0->td_cow; 1077 1078 tp = NULL; 1079 if (p->p_pgrp) { 1080 kp->ki_pgid = p->p_pgrp->pg_id; 1081 kp->ki_jobc = p->p_pgrp->pg_jobc; 1082 sp = p->p_pgrp->pg_session; 1083 1084 if (sp != NULL) { 1085 kp->ki_sid = sp->s_sid; 1086 SESS_LOCK(sp); 1087 strlcpy(kp->ki_login, sp->s_login, 1088 sizeof(kp->ki_login)); 1089 if (sp->s_ttyvp) 1090 kp->ki_kiflag |= KI_CTTY; 1091 if (SESS_LEADER(p)) 1092 kp->ki_kiflag |= KI_SLEADER; 1093 /* XXX proctree_lock */ 1094 tp = sp->s_ttyp; 1095 SESS_UNLOCK(sp); 1096 } 1097 } 1098 if ((p->p_flag & P_CONTROLT) && tp != NULL) { 1099 kp->ki_tdev = tty_udev(tp); 1100 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 1101 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; 1102 if (tp->t_session) 1103 kp->ki_tsid = tp->t_session->s_sid; 1104 } else { 1105 kp->ki_tdev = NODEV; 1106 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 1107 } 1108 if (p->p_comm[0] != '\0') 1109 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm)); 1110 if (p->p_sysent && p->p_sysent->sv_name != NULL && 1111 p->p_sysent->sv_name[0] != '\0') 1112 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul)); 1113 kp->ki_siglist = p->p_siglist; 1114 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig); 1115 kp->ki_acflag = p->p_acflag; 1116 kp->ki_lock = p->p_lock; 1117 if (p->p_pptr) { 1118 kp->ki_ppid = p->p_oppid; 1119 if (p->p_flag & P_TRACED) 1120 kp->ki_tracer = p->p_pptr->p_pid; 1121 } 1122 } 1123 1124 /* 1125 * Fill in information that is thread specific. Must be called with 1126 * target process locked. If 'preferthread' is set, overwrite certain 1127 * process-related fields that are maintained for both threads and 1128 * processes. 1129 */ 1130 static void 1131 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread) 1132 { 1133 struct proc *p; 1134 1135 p = td->td_proc; 1136 kp->ki_tdaddr = td; 1137 PROC_LOCK_ASSERT(p, MA_OWNED); 1138 1139 if (preferthread) 1140 PROC_STATLOCK(p); 1141 thread_lock(td); 1142 if (td->td_wmesg != NULL) 1143 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg)); 1144 else 1145 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg)); 1146 if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >= 1147 sizeof(kp->ki_tdname)) { 1148 strlcpy(kp->ki_moretdname, 1149 td->td_name + sizeof(kp->ki_tdname) - 1, 1150 sizeof(kp->ki_moretdname)); 1151 } else { 1152 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname)); 1153 } 1154 if (TD_ON_LOCK(td)) { 1155 kp->ki_kiflag |= KI_LOCKBLOCK; 1156 strlcpy(kp->ki_lockname, td->td_lockname, 1157 sizeof(kp->ki_lockname)); 1158 } else { 1159 kp->ki_kiflag &= ~KI_LOCKBLOCK; 1160 bzero(kp->ki_lockname, sizeof(kp->ki_lockname)); 1161 } 1162 1163 if (p->p_state == PRS_NORMAL) { /* approximate. */ 1164 if (TD_ON_RUNQ(td) || 1165 TD_CAN_RUN(td) || 1166 TD_IS_RUNNING(td)) { 1167 kp->ki_stat = SRUN; 1168 } else if (P_SHOULDSTOP(p)) { 1169 kp->ki_stat = SSTOP; 1170 } else if (TD_IS_SLEEPING(td)) { 1171 kp->ki_stat = SSLEEP; 1172 } else if (TD_ON_LOCK(td)) { 1173 kp->ki_stat = SLOCK; 1174 } else { 1175 kp->ki_stat = SWAIT; 1176 } 1177 } else if (p->p_state == PRS_ZOMBIE) { 1178 kp->ki_stat = SZOMB; 1179 } else { 1180 kp->ki_stat = SIDL; 1181 } 1182 1183 /* Things in the thread */ 1184 kp->ki_wchan = td->td_wchan; 1185 kp->ki_pri.pri_level = td->td_priority; 1186 kp->ki_pri.pri_native = td->td_base_pri; 1187 1188 /* 1189 * Note: legacy fields; clamp at the old NOCPU value and/or 1190 * the maximum u_char CPU value. 1191 */ 1192 if (td->td_lastcpu == NOCPU) 1193 kp->ki_lastcpu_old = NOCPU_OLD; 1194 else if (td->td_lastcpu > MAXCPU_OLD) 1195 kp->ki_lastcpu_old = MAXCPU_OLD; 1196 else 1197 kp->ki_lastcpu_old = td->td_lastcpu; 1198 1199 if (td->td_oncpu == NOCPU) 1200 kp->ki_oncpu_old = NOCPU_OLD; 1201 else if (td->td_oncpu > MAXCPU_OLD) 1202 kp->ki_oncpu_old = MAXCPU_OLD; 1203 else 1204 kp->ki_oncpu_old = td->td_oncpu; 1205 1206 kp->ki_lastcpu = td->td_lastcpu; 1207 kp->ki_oncpu = td->td_oncpu; 1208 kp->ki_tdflags = td->td_flags; 1209 kp->ki_tid = td->td_tid; 1210 kp->ki_numthreads = p->p_numthreads; 1211 kp->ki_pcb = td->td_pcb; 1212 kp->ki_kstack = (void *)td->td_kstack; 1213 kp->ki_slptime = (ticks - td->td_slptick) / hz; 1214 kp->ki_pri.pri_class = td->td_pri_class; 1215 kp->ki_pri.pri_user = td->td_user_pri; 1216 1217 if (preferthread) { 1218 rufetchtd(td, &kp->ki_rusage); 1219 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime); 1220 kp->ki_pctcpu = sched_pctcpu(td); 1221 kp->ki_estcpu = sched_estcpu(td); 1222 kp->ki_cow = td->td_cow; 1223 } 1224 1225 /* We can't get this anymore but ps etc never used it anyway. */ 1226 kp->ki_rqindex = 0; 1227 1228 if (preferthread) 1229 kp->ki_siglist = td->td_siglist; 1230 kp->ki_sigmask = td->td_sigmask; 1231 thread_unlock(td); 1232 if (preferthread) 1233 PROC_STATUNLOCK(p); 1234 } 1235 1236 /* 1237 * Fill in a kinfo_proc structure for the specified process. 1238 * Must be called with the target process locked. 1239 */ 1240 void 1241 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp) 1242 { 1243 1244 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1245 1246 fill_kinfo_proc_only(p, kp); 1247 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0); 1248 fill_kinfo_aggregate(p, kp); 1249 } 1250 1251 struct pstats * 1252 pstats_alloc(void) 1253 { 1254 1255 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK)); 1256 } 1257 1258 /* 1259 * Copy parts of p_stats; zero the rest of p_stats (statistics). 1260 */ 1261 void 1262 pstats_fork(struct pstats *src, struct pstats *dst) 1263 { 1264 1265 bzero(&dst->pstat_startzero, 1266 __rangeof(struct pstats, pstat_startzero, pstat_endzero)); 1267 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy, 1268 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy)); 1269 } 1270 1271 void 1272 pstats_free(struct pstats *ps) 1273 { 1274 1275 free(ps, M_SUBPROC); 1276 } 1277 1278 #ifdef COMPAT_FREEBSD32 1279 1280 /* 1281 * This function is typically used to copy out the kernel address, so 1282 * it can be replaced by assignment of zero. 1283 */ 1284 static inline uint32_t 1285 ptr32_trim(const void *ptr) 1286 { 1287 uintptr_t uptr; 1288 1289 uptr = (uintptr_t)ptr; 1290 return ((uptr > UINT_MAX) ? 0 : uptr); 1291 } 1292 1293 #define PTRTRIM_CP(src,dst,fld) \ 1294 do { (dst).fld = ptr32_trim((src).fld); } while (0) 1295 1296 static void 1297 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32) 1298 { 1299 int i; 1300 1301 bzero(ki32, sizeof(struct kinfo_proc32)); 1302 ki32->ki_structsize = sizeof(struct kinfo_proc32); 1303 CP(*ki, *ki32, ki_layout); 1304 PTRTRIM_CP(*ki, *ki32, ki_args); 1305 PTRTRIM_CP(*ki, *ki32, ki_paddr); 1306 PTRTRIM_CP(*ki, *ki32, ki_addr); 1307 PTRTRIM_CP(*ki, *ki32, ki_tracep); 1308 PTRTRIM_CP(*ki, *ki32, ki_textvp); 1309 PTRTRIM_CP(*ki, *ki32, ki_fd); 1310 PTRTRIM_CP(*ki, *ki32, ki_vmspace); 1311 PTRTRIM_CP(*ki, *ki32, ki_wchan); 1312 CP(*ki, *ki32, ki_pid); 1313 CP(*ki, *ki32, ki_ppid); 1314 CP(*ki, *ki32, ki_pgid); 1315 CP(*ki, *ki32, ki_tpgid); 1316 CP(*ki, *ki32, ki_sid); 1317 CP(*ki, *ki32, ki_tsid); 1318 CP(*ki, *ki32, ki_jobc); 1319 CP(*ki, *ki32, ki_tdev); 1320 CP(*ki, *ki32, ki_tdev_freebsd11); 1321 CP(*ki, *ki32, ki_siglist); 1322 CP(*ki, *ki32, ki_sigmask); 1323 CP(*ki, *ki32, ki_sigignore); 1324 CP(*ki, *ki32, ki_sigcatch); 1325 CP(*ki, *ki32, ki_uid); 1326 CP(*ki, *ki32, ki_ruid); 1327 CP(*ki, *ki32, ki_svuid); 1328 CP(*ki, *ki32, ki_rgid); 1329 CP(*ki, *ki32, ki_svgid); 1330 CP(*ki, *ki32, ki_ngroups); 1331 for (i = 0; i < KI_NGROUPS; i++) 1332 CP(*ki, *ki32, ki_groups[i]); 1333 CP(*ki, *ki32, ki_size); 1334 CP(*ki, *ki32, ki_rssize); 1335 CP(*ki, *ki32, ki_swrss); 1336 CP(*ki, *ki32, ki_tsize); 1337 CP(*ki, *ki32, ki_dsize); 1338 CP(*ki, *ki32, ki_ssize); 1339 CP(*ki, *ki32, ki_xstat); 1340 CP(*ki, *ki32, ki_acflag); 1341 CP(*ki, *ki32, ki_pctcpu); 1342 CP(*ki, *ki32, ki_estcpu); 1343 CP(*ki, *ki32, ki_slptime); 1344 CP(*ki, *ki32, ki_swtime); 1345 CP(*ki, *ki32, ki_cow); 1346 CP(*ki, *ki32, ki_runtime); 1347 TV_CP(*ki, *ki32, ki_start); 1348 TV_CP(*ki, *ki32, ki_childtime); 1349 CP(*ki, *ki32, ki_flag); 1350 CP(*ki, *ki32, ki_kiflag); 1351 CP(*ki, *ki32, ki_traceflag); 1352 CP(*ki, *ki32, ki_stat); 1353 CP(*ki, *ki32, ki_nice); 1354 CP(*ki, *ki32, ki_lock); 1355 CP(*ki, *ki32, ki_rqindex); 1356 CP(*ki, *ki32, ki_oncpu); 1357 CP(*ki, *ki32, ki_lastcpu); 1358 1359 /* XXX TODO: wrap cpu value as appropriate */ 1360 CP(*ki, *ki32, ki_oncpu_old); 1361 CP(*ki, *ki32, ki_lastcpu_old); 1362 1363 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1); 1364 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1); 1365 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1); 1366 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1); 1367 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1); 1368 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1); 1369 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1); 1370 bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1); 1371 CP(*ki, *ki32, ki_tracer); 1372 CP(*ki, *ki32, ki_flag2); 1373 CP(*ki, *ki32, ki_fibnum); 1374 CP(*ki, *ki32, ki_cr_flags); 1375 CP(*ki, *ki32, ki_jid); 1376 CP(*ki, *ki32, ki_numthreads); 1377 CP(*ki, *ki32, ki_tid); 1378 CP(*ki, *ki32, ki_pri); 1379 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage); 1380 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch); 1381 PTRTRIM_CP(*ki, *ki32, ki_pcb); 1382 PTRTRIM_CP(*ki, *ki32, ki_kstack); 1383 PTRTRIM_CP(*ki, *ki32, ki_udata); 1384 PTRTRIM_CP(*ki, *ki32, ki_tdaddr); 1385 CP(*ki, *ki32, ki_sflag); 1386 CP(*ki, *ki32, ki_tdflags); 1387 } 1388 #endif 1389 1390 static ssize_t 1391 kern_proc_out_size(struct proc *p, int flags) 1392 { 1393 ssize_t size = 0; 1394 1395 PROC_LOCK_ASSERT(p, MA_OWNED); 1396 1397 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1398 #ifdef COMPAT_FREEBSD32 1399 if ((flags & KERN_PROC_MASK32) != 0) { 1400 size += sizeof(struct kinfo_proc32); 1401 } else 1402 #endif 1403 size += sizeof(struct kinfo_proc); 1404 } else { 1405 #ifdef COMPAT_FREEBSD32 1406 if ((flags & KERN_PROC_MASK32) != 0) 1407 size += sizeof(struct kinfo_proc32) * p->p_numthreads; 1408 else 1409 #endif 1410 size += sizeof(struct kinfo_proc) * p->p_numthreads; 1411 } 1412 PROC_UNLOCK(p); 1413 return (size); 1414 } 1415 1416 int 1417 kern_proc_out(struct proc *p, struct sbuf *sb, int flags) 1418 { 1419 struct thread *td; 1420 struct kinfo_proc ki; 1421 #ifdef COMPAT_FREEBSD32 1422 struct kinfo_proc32 ki32; 1423 #endif 1424 int error; 1425 1426 PROC_LOCK_ASSERT(p, MA_OWNED); 1427 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1428 1429 error = 0; 1430 fill_kinfo_proc(p, &ki); 1431 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1432 #ifdef COMPAT_FREEBSD32 1433 if ((flags & KERN_PROC_MASK32) != 0) { 1434 freebsd32_kinfo_proc_out(&ki, &ki32); 1435 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1436 error = ENOMEM; 1437 } else 1438 #endif 1439 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1440 error = ENOMEM; 1441 } else { 1442 FOREACH_THREAD_IN_PROC(p, td) { 1443 fill_kinfo_thread(td, &ki, 1); 1444 #ifdef COMPAT_FREEBSD32 1445 if ((flags & KERN_PROC_MASK32) != 0) { 1446 freebsd32_kinfo_proc_out(&ki, &ki32); 1447 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1448 error = ENOMEM; 1449 } else 1450 #endif 1451 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1452 error = ENOMEM; 1453 if (error != 0) 1454 break; 1455 } 1456 } 1457 PROC_UNLOCK(p); 1458 return (error); 1459 } 1460 1461 static int 1462 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags) 1463 { 1464 struct sbuf sb; 1465 struct kinfo_proc ki; 1466 int error, error2; 1467 1468 if (req->oldptr == NULL) 1469 return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags))); 1470 1471 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); 1472 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1473 error = kern_proc_out(p, &sb, flags); 1474 error2 = sbuf_finish(&sb); 1475 sbuf_delete(&sb); 1476 if (error != 0) 1477 return (error); 1478 else if (error2 != 0) 1479 return (error2); 1480 return (0); 1481 } 1482 1483 int 1484 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg) 1485 { 1486 struct proc *p; 1487 int error, i, j; 1488 1489 for (i = 0; i < pidhashlock + 1; i++) { 1490 sx_slock(&pidhashtbl_lock[i]); 1491 for (j = i; j <= pidhash; j += pidhashlock + 1) { 1492 LIST_FOREACH(p, &pidhashtbl[j], p_hash) { 1493 if (p->p_state == PRS_NEW) 1494 continue; 1495 error = cb(p, cbarg); 1496 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 1497 if (error != 0) { 1498 sx_sunlock(&pidhashtbl_lock[i]); 1499 return (error); 1500 } 1501 } 1502 } 1503 sx_sunlock(&pidhashtbl_lock[i]); 1504 } 1505 return (0); 1506 } 1507 1508 struct kern_proc_out_args { 1509 struct sysctl_req *req; 1510 int flags; 1511 int oid_number; 1512 int *name; 1513 }; 1514 1515 static int 1516 sysctl_kern_proc_iterate(struct proc *p, void *origarg) 1517 { 1518 struct kern_proc_out_args *arg = origarg; 1519 int *name = arg->name; 1520 int oid_number = arg->oid_number; 1521 int flags = arg->flags; 1522 struct sysctl_req *req = arg->req; 1523 int error = 0; 1524 1525 PROC_LOCK(p); 1526 1527 KASSERT(p->p_ucred != NULL, 1528 ("process credential is NULL for non-NEW proc")); 1529 /* 1530 * Show a user only appropriate processes. 1531 */ 1532 if (p_cansee(curthread, p)) 1533 goto skip; 1534 /* 1535 * TODO - make more efficient (see notes below). 1536 * do by session. 1537 */ 1538 switch (oid_number) { 1539 1540 case KERN_PROC_GID: 1541 if (p->p_ucred->cr_gid != (gid_t)name[0]) 1542 goto skip; 1543 break; 1544 1545 case KERN_PROC_PGRP: 1546 /* could do this by traversing pgrp */ 1547 if (p->p_pgrp == NULL || 1548 p->p_pgrp->pg_id != (pid_t)name[0]) 1549 goto skip; 1550 break; 1551 1552 case KERN_PROC_RGID: 1553 if (p->p_ucred->cr_rgid != (gid_t)name[0]) 1554 goto skip; 1555 break; 1556 1557 case KERN_PROC_SESSION: 1558 if (p->p_session == NULL || 1559 p->p_session->s_sid != (pid_t)name[0]) 1560 goto skip; 1561 break; 1562 1563 case KERN_PROC_TTY: 1564 if ((p->p_flag & P_CONTROLT) == 0 || 1565 p->p_session == NULL) 1566 goto skip; 1567 /* XXX proctree_lock */ 1568 SESS_LOCK(p->p_session); 1569 if (p->p_session->s_ttyp == NULL || 1570 tty_udev(p->p_session->s_ttyp) != 1571 (dev_t)name[0]) { 1572 SESS_UNLOCK(p->p_session); 1573 goto skip; 1574 } 1575 SESS_UNLOCK(p->p_session); 1576 break; 1577 1578 case KERN_PROC_UID: 1579 if (p->p_ucred->cr_uid != (uid_t)name[0]) 1580 goto skip; 1581 break; 1582 1583 case KERN_PROC_RUID: 1584 if (p->p_ucred->cr_ruid != (uid_t)name[0]) 1585 goto skip; 1586 break; 1587 1588 case KERN_PROC_PROC: 1589 break; 1590 1591 default: 1592 break; 1593 1594 } 1595 error = sysctl_out_proc(p, req, flags); 1596 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 1597 return (error); 1598 skip: 1599 PROC_UNLOCK(p); 1600 return (0); 1601 } 1602 1603 static int 1604 sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 1605 { 1606 struct kern_proc_out_args iterarg; 1607 int *name = (int *)arg1; 1608 u_int namelen = arg2; 1609 struct proc *p; 1610 int flags, oid_number; 1611 int error = 0; 1612 1613 oid_number = oidp->oid_number; 1614 if (oid_number != KERN_PROC_ALL && 1615 (oid_number & KERN_PROC_INC_THREAD) == 0) 1616 flags = KERN_PROC_NOTHREADS; 1617 else { 1618 flags = 0; 1619 oid_number &= ~KERN_PROC_INC_THREAD; 1620 } 1621 #ifdef COMPAT_FREEBSD32 1622 if (req->flags & SCTL_MASK32) 1623 flags |= KERN_PROC_MASK32; 1624 #endif 1625 if (oid_number == KERN_PROC_PID) { 1626 if (namelen != 1) 1627 return (EINVAL); 1628 error = sysctl_wire_old_buffer(req, 0); 1629 if (error) 1630 return (error); 1631 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1632 if (error == 0) 1633 error = sysctl_out_proc(p, req, flags); 1634 return (error); 1635 } 1636 1637 switch (oid_number) { 1638 case KERN_PROC_ALL: 1639 if (namelen != 0) 1640 return (EINVAL); 1641 break; 1642 case KERN_PROC_PROC: 1643 if (namelen != 0 && namelen != 1) 1644 return (EINVAL); 1645 break; 1646 default: 1647 if (namelen != 1) 1648 return (EINVAL); 1649 break; 1650 } 1651 1652 if (req->oldptr == NULL) { 1653 /* overestimate by 5 procs */ 1654 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1655 if (error) 1656 return (error); 1657 } else { 1658 error = sysctl_wire_old_buffer(req, 0); 1659 if (error != 0) 1660 return (error); 1661 } 1662 iterarg.flags = flags; 1663 iterarg.oid_number = oid_number; 1664 iterarg.req = req; 1665 iterarg.name = name; 1666 error = proc_iterate(sysctl_kern_proc_iterate, &iterarg); 1667 return (error); 1668 } 1669 1670 struct pargs * 1671 pargs_alloc(int len) 1672 { 1673 struct pargs *pa; 1674 1675 pa = malloc(sizeof(struct pargs) + len, M_PARGS, 1676 M_WAITOK); 1677 refcount_init(&pa->ar_ref, 1); 1678 pa->ar_length = len; 1679 return (pa); 1680 } 1681 1682 static void 1683 pargs_free(struct pargs *pa) 1684 { 1685 1686 free(pa, M_PARGS); 1687 } 1688 1689 void 1690 pargs_hold(struct pargs *pa) 1691 { 1692 1693 if (pa == NULL) 1694 return; 1695 refcount_acquire(&pa->ar_ref); 1696 } 1697 1698 void 1699 pargs_drop(struct pargs *pa) 1700 { 1701 1702 if (pa == NULL) 1703 return; 1704 if (refcount_release(&pa->ar_ref)) 1705 pargs_free(pa); 1706 } 1707 1708 static int 1709 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, 1710 size_t len) 1711 { 1712 ssize_t n; 1713 1714 /* 1715 * This may return a short read if the string is shorter than the chunk 1716 * and is aligned at the end of the page, and the following page is not 1717 * mapped. 1718 */ 1719 n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len); 1720 if (n <= 0) 1721 return (ENOMEM); 1722 return (0); 1723 } 1724 1725 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ 1726 1727 enum proc_vector_type { 1728 PROC_ARG, 1729 PROC_ENV, 1730 PROC_AUX, 1731 }; 1732 1733 #ifdef COMPAT_FREEBSD32 1734 static int 1735 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, 1736 size_t *vsizep, enum proc_vector_type type) 1737 { 1738 struct freebsd32_ps_strings pss; 1739 Elf32_Auxinfo aux; 1740 vm_offset_t vptr, ptr; 1741 uint32_t *proc_vector32; 1742 char **proc_vector; 1743 size_t vsize, size; 1744 int i, error; 1745 1746 error = 0; 1747 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss, 1748 sizeof(pss)) != sizeof(pss)) 1749 return (ENOMEM); 1750 switch (type) { 1751 case PROC_ARG: 1752 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); 1753 vsize = pss.ps_nargvstr; 1754 if (vsize > ARG_MAX) 1755 return (ENOEXEC); 1756 size = vsize * sizeof(int32_t); 1757 break; 1758 case PROC_ENV: 1759 vptr = (vm_offset_t)PTRIN(pss.ps_envstr); 1760 vsize = pss.ps_nenvstr; 1761 if (vsize > ARG_MAX) 1762 return (ENOEXEC); 1763 size = vsize * sizeof(int32_t); 1764 break; 1765 case PROC_AUX: 1766 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + 1767 (pss.ps_nenvstr + 1) * sizeof(int32_t); 1768 if (vptr % 4 != 0) 1769 return (ENOEXEC); 1770 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1771 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1772 sizeof(aux)) 1773 return (ENOMEM); 1774 if (aux.a_type == AT_NULL) 1775 break; 1776 ptr += sizeof(aux); 1777 } 1778 if (aux.a_type != AT_NULL) 1779 return (ENOEXEC); 1780 vsize = i + 1; 1781 size = vsize * sizeof(aux); 1782 break; 1783 default: 1784 KASSERT(0, ("Wrong proc vector type: %d", type)); 1785 return (EINVAL); 1786 } 1787 proc_vector32 = malloc(size, M_TEMP, M_WAITOK); 1788 if (proc_readmem(td, p, vptr, proc_vector32, size) != size) { 1789 error = ENOMEM; 1790 goto done; 1791 } 1792 if (type == PROC_AUX) { 1793 *proc_vectorp = (char **)proc_vector32; 1794 *vsizep = vsize; 1795 return (0); 1796 } 1797 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); 1798 for (i = 0; i < (int)vsize; i++) 1799 proc_vector[i] = PTRIN(proc_vector32[i]); 1800 *proc_vectorp = proc_vector; 1801 *vsizep = vsize; 1802 done: 1803 free(proc_vector32, M_TEMP); 1804 return (error); 1805 } 1806 #endif 1807 1808 static int 1809 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, 1810 size_t *vsizep, enum proc_vector_type type) 1811 { 1812 struct ps_strings pss; 1813 Elf_Auxinfo aux; 1814 vm_offset_t vptr, ptr; 1815 char **proc_vector; 1816 size_t vsize, size; 1817 int i; 1818 1819 #ifdef COMPAT_FREEBSD32 1820 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1821 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); 1822 #endif 1823 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss, 1824 sizeof(pss)) != sizeof(pss)) 1825 return (ENOMEM); 1826 switch (type) { 1827 case PROC_ARG: 1828 vptr = (vm_offset_t)pss.ps_argvstr; 1829 vsize = pss.ps_nargvstr; 1830 if (vsize > ARG_MAX) 1831 return (ENOEXEC); 1832 size = vsize * sizeof(char *); 1833 break; 1834 case PROC_ENV: 1835 vptr = (vm_offset_t)pss.ps_envstr; 1836 vsize = pss.ps_nenvstr; 1837 if (vsize > ARG_MAX) 1838 return (ENOEXEC); 1839 size = vsize * sizeof(char *); 1840 break; 1841 case PROC_AUX: 1842 /* 1843 * The aux array is just above env array on the stack. Check 1844 * that the address is naturally aligned. 1845 */ 1846 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) 1847 * sizeof(char *); 1848 #if __ELF_WORD_SIZE == 64 1849 if (vptr % sizeof(uint64_t) != 0) 1850 #else 1851 if (vptr % sizeof(uint32_t) != 0) 1852 #endif 1853 return (ENOEXEC); 1854 /* 1855 * We count the array size reading the aux vectors from the 1856 * stack until AT_NULL vector is returned. So (to keep the code 1857 * simple) we read the process stack twice: the first time here 1858 * to find the size and the second time when copying the vectors 1859 * to the allocated proc_vector. 1860 */ 1861 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1862 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1863 sizeof(aux)) 1864 return (ENOMEM); 1865 if (aux.a_type == AT_NULL) 1866 break; 1867 ptr += sizeof(aux); 1868 } 1869 /* 1870 * If the PROC_AUXV_MAX entries are iterated over, and we have 1871 * not reached AT_NULL, it is most likely we are reading wrong 1872 * data: either the process doesn't have auxv array or data has 1873 * been modified. Return the error in this case. 1874 */ 1875 if (aux.a_type != AT_NULL) 1876 return (ENOEXEC); 1877 vsize = i + 1; 1878 size = vsize * sizeof(aux); 1879 break; 1880 default: 1881 KASSERT(0, ("Wrong proc vector type: %d", type)); 1882 return (EINVAL); /* In case we are built without INVARIANTS. */ 1883 } 1884 proc_vector = malloc(size, M_TEMP, M_WAITOK); 1885 if (proc_readmem(td, p, vptr, proc_vector, size) != size) { 1886 free(proc_vector, M_TEMP); 1887 return (ENOMEM); 1888 } 1889 *proc_vectorp = proc_vector; 1890 *vsizep = vsize; 1891 1892 return (0); 1893 } 1894 1895 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ 1896 1897 static int 1898 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, 1899 enum proc_vector_type type) 1900 { 1901 size_t done, len, nchr, vsize; 1902 int error, i; 1903 char **proc_vector, *sptr; 1904 char pss_string[GET_PS_STRINGS_CHUNK_SZ]; 1905 1906 PROC_ASSERT_HELD(p); 1907 1908 /* 1909 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. 1910 */ 1911 nchr = 2 * (PATH_MAX + ARG_MAX); 1912 1913 error = get_proc_vector(td, p, &proc_vector, &vsize, type); 1914 if (error != 0) 1915 return (error); 1916 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { 1917 /* 1918 * The program may have scribbled into its argv array, e.g. to 1919 * remove some arguments. If that has happened, break out 1920 * before trying to read from NULL. 1921 */ 1922 if (proc_vector[i] == NULL) 1923 break; 1924 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { 1925 error = proc_read_string(td, p, sptr, pss_string, 1926 sizeof(pss_string)); 1927 if (error != 0) 1928 goto done; 1929 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); 1930 if (done + len >= nchr) 1931 len = nchr - done - 1; 1932 sbuf_bcat(sb, pss_string, len); 1933 if (len != GET_PS_STRINGS_CHUNK_SZ) 1934 break; 1935 done += GET_PS_STRINGS_CHUNK_SZ; 1936 } 1937 sbuf_bcat(sb, "", 1); 1938 done += len + 1; 1939 } 1940 done: 1941 free(proc_vector, M_TEMP); 1942 return (error); 1943 } 1944 1945 int 1946 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) 1947 { 1948 1949 return (get_ps_strings(curthread, p, sb, PROC_ARG)); 1950 } 1951 1952 int 1953 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) 1954 { 1955 1956 return (get_ps_strings(curthread, p, sb, PROC_ENV)); 1957 } 1958 1959 int 1960 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) 1961 { 1962 size_t vsize, size; 1963 char **auxv; 1964 int error; 1965 1966 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); 1967 if (error == 0) { 1968 #ifdef COMPAT_FREEBSD32 1969 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1970 size = vsize * sizeof(Elf32_Auxinfo); 1971 else 1972 #endif 1973 size = vsize * sizeof(Elf_Auxinfo); 1974 if (sbuf_bcat(sb, auxv, size) != 0) 1975 error = ENOMEM; 1976 free(auxv, M_TEMP); 1977 } 1978 return (error); 1979 } 1980 1981 /* 1982 * This sysctl allows a process to retrieve the argument list or process 1983 * title for another process without groping around in the address space 1984 * of the other process. It also allow a process to set its own "process 1985 * title to a string of its own choice. 1986 */ 1987 static int 1988 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 1989 { 1990 int *name = (int *)arg1; 1991 u_int namelen = arg2; 1992 struct pargs *newpa, *pa; 1993 struct proc *p; 1994 struct sbuf sb; 1995 int flags, error = 0, error2; 1996 pid_t pid; 1997 1998 if (namelen != 1) 1999 return (EINVAL); 2000 2001 pid = (pid_t)name[0]; 2002 /* 2003 * If the query is for this process and it is single-threaded, there 2004 * is nobody to modify pargs, thus we can just read. 2005 */ 2006 p = curproc; 2007 if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL && 2008 (pa = p->p_args) != NULL) 2009 return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length)); 2010 2011 flags = PGET_CANSEE; 2012 if (req->newptr != NULL) 2013 flags |= PGET_ISCURRENT; 2014 error = pget(pid, flags, &p); 2015 if (error) 2016 return (error); 2017 2018 pa = p->p_args; 2019 if (pa != NULL) { 2020 pargs_hold(pa); 2021 PROC_UNLOCK(p); 2022 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 2023 pargs_drop(pa); 2024 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { 2025 _PHOLD(p); 2026 PROC_UNLOCK(p); 2027 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2028 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2029 error = proc_getargv(curthread, p, &sb); 2030 error2 = sbuf_finish(&sb); 2031 PRELE(p); 2032 sbuf_delete(&sb); 2033 if (error == 0 && error2 != 0) 2034 error = error2; 2035 } else { 2036 PROC_UNLOCK(p); 2037 } 2038 if (error != 0 || req->newptr == NULL) 2039 return (error); 2040 2041 if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs)) 2042 return (ENOMEM); 2043 2044 if (req->newlen == 0) { 2045 /* 2046 * Clear the argument pointer, so that we'll fetch arguments 2047 * with proc_getargv() until further notice. 2048 */ 2049 newpa = NULL; 2050 } else { 2051 newpa = pargs_alloc(req->newlen); 2052 error = SYSCTL_IN(req, newpa->ar_args, req->newlen); 2053 if (error != 0) { 2054 pargs_free(newpa); 2055 return (error); 2056 } 2057 } 2058 PROC_LOCK(p); 2059 pa = p->p_args; 2060 p->p_args = newpa; 2061 PROC_UNLOCK(p); 2062 pargs_drop(pa); 2063 return (0); 2064 } 2065 2066 /* 2067 * This sysctl allows a process to retrieve environment of another process. 2068 */ 2069 static int 2070 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) 2071 { 2072 int *name = (int *)arg1; 2073 u_int namelen = arg2; 2074 struct proc *p; 2075 struct sbuf sb; 2076 int error, error2; 2077 2078 if (namelen != 1) 2079 return (EINVAL); 2080 2081 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2082 if (error != 0) 2083 return (error); 2084 if ((p->p_flag & P_SYSTEM) != 0) { 2085 PRELE(p); 2086 return (0); 2087 } 2088 2089 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2090 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2091 error = proc_getenvv(curthread, p, &sb); 2092 error2 = sbuf_finish(&sb); 2093 PRELE(p); 2094 sbuf_delete(&sb); 2095 return (error != 0 ? error : error2); 2096 } 2097 2098 /* 2099 * This sysctl allows a process to retrieve ELF auxiliary vector of 2100 * another process. 2101 */ 2102 static int 2103 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) 2104 { 2105 int *name = (int *)arg1; 2106 u_int namelen = arg2; 2107 struct proc *p; 2108 struct sbuf sb; 2109 int error, error2; 2110 2111 if (namelen != 1) 2112 return (EINVAL); 2113 2114 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2115 if (error != 0) 2116 return (error); 2117 if ((p->p_flag & P_SYSTEM) != 0) { 2118 PRELE(p); 2119 return (0); 2120 } 2121 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2122 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2123 error = proc_getauxv(curthread, p, &sb); 2124 error2 = sbuf_finish(&sb); 2125 PRELE(p); 2126 sbuf_delete(&sb); 2127 return (error != 0 ? error : error2); 2128 } 2129 2130 /* 2131 * This sysctl allows a process to retrieve the path of the executable for 2132 * itself or another process. 2133 */ 2134 static int 2135 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) 2136 { 2137 pid_t *pidp = (pid_t *)arg1; 2138 unsigned int arglen = arg2; 2139 struct proc *p; 2140 struct vnode *vp; 2141 char *retbuf, *freebuf; 2142 int error; 2143 2144 if (arglen != 1) 2145 return (EINVAL); 2146 if (*pidp == -1) { /* -1 means this process */ 2147 p = req->td->td_proc; 2148 } else { 2149 error = pget(*pidp, PGET_CANSEE, &p); 2150 if (error != 0) 2151 return (error); 2152 } 2153 2154 vp = p->p_textvp; 2155 if (vp == NULL) { 2156 if (*pidp != -1) 2157 PROC_UNLOCK(p); 2158 return (0); 2159 } 2160 vref(vp); 2161 if (*pidp != -1) 2162 PROC_UNLOCK(p); 2163 error = vn_fullpath(req->td, vp, &retbuf, &freebuf); 2164 vrele(vp); 2165 if (error) 2166 return (error); 2167 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); 2168 free(freebuf, M_TEMP); 2169 return (error); 2170 } 2171 2172 static int 2173 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) 2174 { 2175 struct proc *p; 2176 char *sv_name; 2177 int *name; 2178 int namelen; 2179 int error; 2180 2181 namelen = arg2; 2182 if (namelen != 1) 2183 return (EINVAL); 2184 2185 name = (int *)arg1; 2186 error = pget((pid_t)name[0], PGET_CANSEE, &p); 2187 if (error != 0) 2188 return (error); 2189 sv_name = p->p_sysent->sv_name; 2190 PROC_UNLOCK(p); 2191 return (sysctl_handle_string(oidp, sv_name, 0, req)); 2192 } 2193 2194 #ifdef KINFO_OVMENTRY_SIZE 2195 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); 2196 #endif 2197 2198 #ifdef COMPAT_FREEBSD7 2199 static int 2200 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) 2201 { 2202 vm_map_entry_t entry, tmp_entry; 2203 unsigned int last_timestamp; 2204 char *fullpath, *freepath; 2205 struct kinfo_ovmentry *kve; 2206 struct vattr va; 2207 struct ucred *cred; 2208 int error, *name; 2209 struct vnode *vp; 2210 struct proc *p; 2211 vm_map_t map; 2212 struct vmspace *vm; 2213 2214 name = (int *)arg1; 2215 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2216 if (error != 0) 2217 return (error); 2218 vm = vmspace_acquire_ref(p); 2219 if (vm == NULL) { 2220 PRELE(p); 2221 return (ESRCH); 2222 } 2223 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2224 2225 map = &vm->vm_map; 2226 vm_map_lock_read(map); 2227 VM_MAP_ENTRY_FOREACH(entry, map) { 2228 vm_object_t obj, tobj, lobj; 2229 vm_offset_t addr; 2230 2231 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2232 continue; 2233 2234 bzero(kve, sizeof(*kve)); 2235 kve->kve_structsize = sizeof(*kve); 2236 2237 kve->kve_private_resident = 0; 2238 obj = entry->object.vm_object; 2239 if (obj != NULL) { 2240 VM_OBJECT_RLOCK(obj); 2241 if (obj->shadow_count == 1) 2242 kve->kve_private_resident = 2243 obj->resident_page_count; 2244 } 2245 kve->kve_resident = 0; 2246 addr = entry->start; 2247 while (addr < entry->end) { 2248 if (pmap_extract(map->pmap, addr)) 2249 kve->kve_resident++; 2250 addr += PAGE_SIZE; 2251 } 2252 2253 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { 2254 if (tobj != obj) { 2255 VM_OBJECT_RLOCK(tobj); 2256 kve->kve_offset += tobj->backing_object_offset; 2257 } 2258 if (lobj != obj) 2259 VM_OBJECT_RUNLOCK(lobj); 2260 lobj = tobj; 2261 } 2262 2263 kve->kve_start = (void*)entry->start; 2264 kve->kve_end = (void*)entry->end; 2265 kve->kve_offset += (off_t)entry->offset; 2266 2267 if (entry->protection & VM_PROT_READ) 2268 kve->kve_protection |= KVME_PROT_READ; 2269 if (entry->protection & VM_PROT_WRITE) 2270 kve->kve_protection |= KVME_PROT_WRITE; 2271 if (entry->protection & VM_PROT_EXECUTE) 2272 kve->kve_protection |= KVME_PROT_EXEC; 2273 2274 if (entry->eflags & MAP_ENTRY_COW) 2275 kve->kve_flags |= KVME_FLAG_COW; 2276 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2277 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2278 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2279 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2280 2281 last_timestamp = map->timestamp; 2282 vm_map_unlock_read(map); 2283 2284 kve->kve_fileid = 0; 2285 kve->kve_fsid = 0; 2286 freepath = NULL; 2287 fullpath = ""; 2288 if (lobj) { 2289 kve->kve_type = vm_object_kvme_type(lobj, &vp); 2290 if (kve->kve_type == KVME_TYPE_MGTDEVICE) 2291 kve->kve_type = KVME_TYPE_UNKNOWN; 2292 if (vp != NULL) 2293 vref(vp); 2294 if (lobj != obj) 2295 VM_OBJECT_RUNLOCK(lobj); 2296 2297 kve->kve_ref_count = obj->ref_count; 2298 kve->kve_shadow_count = obj->shadow_count; 2299 VM_OBJECT_RUNLOCK(obj); 2300 if (vp != NULL) { 2301 vn_fullpath(curthread, vp, &fullpath, 2302 &freepath); 2303 cred = curthread->td_ucred; 2304 vn_lock(vp, LK_SHARED | LK_RETRY); 2305 if (VOP_GETATTR(vp, &va, cred) == 0) { 2306 kve->kve_fileid = va.va_fileid; 2307 /* truncate */ 2308 kve->kve_fsid = va.va_fsid; 2309 } 2310 vput(vp); 2311 } 2312 } else { 2313 kve->kve_type = KVME_TYPE_NONE; 2314 kve->kve_ref_count = 0; 2315 kve->kve_shadow_count = 0; 2316 } 2317 2318 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2319 if (freepath != NULL) 2320 free(freepath, M_TEMP); 2321 2322 error = SYSCTL_OUT(req, kve, sizeof(*kve)); 2323 vm_map_lock_read(map); 2324 if (error) 2325 break; 2326 if (last_timestamp != map->timestamp) { 2327 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2328 entry = tmp_entry; 2329 } 2330 } 2331 vm_map_unlock_read(map); 2332 vmspace_free(vm); 2333 PRELE(p); 2334 free(kve, M_TEMP); 2335 return (error); 2336 } 2337 #endif /* COMPAT_FREEBSD7 */ 2338 2339 #ifdef KINFO_VMENTRY_SIZE 2340 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); 2341 #endif 2342 2343 void 2344 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, 2345 int *resident_count, bool *super) 2346 { 2347 vm_object_t obj, tobj; 2348 vm_page_t m, m_adv; 2349 vm_offset_t addr; 2350 vm_paddr_t pa; 2351 vm_pindex_t pi, pi_adv, pindex; 2352 2353 *super = false; 2354 *resident_count = 0; 2355 if (vmmap_skip_res_cnt) 2356 return; 2357 2358 pa = 0; 2359 obj = entry->object.vm_object; 2360 addr = entry->start; 2361 m_adv = NULL; 2362 pi = OFF_TO_IDX(entry->offset); 2363 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { 2364 if (m_adv != NULL) { 2365 m = m_adv; 2366 } else { 2367 pi_adv = atop(entry->end - addr); 2368 pindex = pi; 2369 for (tobj = obj;; tobj = tobj->backing_object) { 2370 m = vm_page_find_least(tobj, pindex); 2371 if (m != NULL) { 2372 if (m->pindex == pindex) 2373 break; 2374 if (pi_adv > m->pindex - pindex) { 2375 pi_adv = m->pindex - pindex; 2376 m_adv = m; 2377 } 2378 } 2379 if (tobj->backing_object == NULL) 2380 goto next; 2381 pindex += OFF_TO_IDX(tobj-> 2382 backing_object_offset); 2383 } 2384 } 2385 m_adv = NULL; 2386 if (m->psind != 0 && addr + pagesizes[1] <= entry->end && 2387 (addr & (pagesizes[1] - 1)) == 0 && 2388 (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) { 2389 *super = true; 2390 pi_adv = atop(pagesizes[1]); 2391 } else { 2392 /* 2393 * We do not test the found page on validity. 2394 * Either the page is busy and being paged in, 2395 * or it was invalidated. The first case 2396 * should be counted as resident, the second 2397 * is not so clear; we do account both. 2398 */ 2399 pi_adv = 1; 2400 } 2401 *resident_count += pi_adv; 2402 next:; 2403 } 2404 } 2405 2406 /* 2407 * Must be called with the process locked and will return unlocked. 2408 */ 2409 int 2410 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags) 2411 { 2412 vm_map_entry_t entry, tmp_entry; 2413 struct vattr va; 2414 vm_map_t map; 2415 vm_object_t obj, tobj, lobj; 2416 char *fullpath, *freepath; 2417 struct kinfo_vmentry *kve; 2418 struct ucred *cred; 2419 struct vnode *vp; 2420 struct vmspace *vm; 2421 vm_offset_t addr; 2422 unsigned int last_timestamp; 2423 int error; 2424 bool super; 2425 2426 PROC_LOCK_ASSERT(p, MA_OWNED); 2427 2428 _PHOLD(p); 2429 PROC_UNLOCK(p); 2430 vm = vmspace_acquire_ref(p); 2431 if (vm == NULL) { 2432 PRELE(p); 2433 return (ESRCH); 2434 } 2435 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO); 2436 2437 error = 0; 2438 map = &vm->vm_map; 2439 vm_map_lock_read(map); 2440 VM_MAP_ENTRY_FOREACH(entry, map) { 2441 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2442 continue; 2443 2444 addr = entry->end; 2445 bzero(kve, sizeof(*kve)); 2446 obj = entry->object.vm_object; 2447 if (obj != NULL) { 2448 for (tobj = obj; tobj != NULL; 2449 tobj = tobj->backing_object) { 2450 VM_OBJECT_RLOCK(tobj); 2451 kve->kve_offset += tobj->backing_object_offset; 2452 lobj = tobj; 2453 } 2454 if (obj->backing_object == NULL) 2455 kve->kve_private_resident = 2456 obj->resident_page_count; 2457 kern_proc_vmmap_resident(map, entry, 2458 &kve->kve_resident, &super); 2459 if (super) 2460 kve->kve_flags |= KVME_FLAG_SUPER; 2461 for (tobj = obj; tobj != NULL; 2462 tobj = tobj->backing_object) { 2463 if (tobj != obj && tobj != lobj) 2464 VM_OBJECT_RUNLOCK(tobj); 2465 } 2466 } else { 2467 lobj = NULL; 2468 } 2469 2470 kve->kve_start = entry->start; 2471 kve->kve_end = entry->end; 2472 kve->kve_offset += entry->offset; 2473 2474 if (entry->protection & VM_PROT_READ) 2475 kve->kve_protection |= KVME_PROT_READ; 2476 if (entry->protection & VM_PROT_WRITE) 2477 kve->kve_protection |= KVME_PROT_WRITE; 2478 if (entry->protection & VM_PROT_EXECUTE) 2479 kve->kve_protection |= KVME_PROT_EXEC; 2480 2481 if (entry->eflags & MAP_ENTRY_COW) 2482 kve->kve_flags |= KVME_FLAG_COW; 2483 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2484 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2485 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2486 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2487 if (entry->eflags & MAP_ENTRY_GROWS_UP) 2488 kve->kve_flags |= KVME_FLAG_GROWS_UP; 2489 if (entry->eflags & MAP_ENTRY_GROWS_DOWN) 2490 kve->kve_flags |= KVME_FLAG_GROWS_DOWN; 2491 if (entry->eflags & MAP_ENTRY_USER_WIRED) 2492 kve->kve_flags |= KVME_FLAG_USER_WIRED; 2493 2494 last_timestamp = map->timestamp; 2495 vm_map_unlock_read(map); 2496 2497 freepath = NULL; 2498 fullpath = ""; 2499 if (lobj != NULL) { 2500 kve->kve_type = vm_object_kvme_type(lobj, &vp); 2501 if (vp != NULL) 2502 vref(vp); 2503 if (lobj != obj) 2504 VM_OBJECT_RUNLOCK(lobj); 2505 2506 kve->kve_ref_count = obj->ref_count; 2507 kve->kve_shadow_count = obj->shadow_count; 2508 VM_OBJECT_RUNLOCK(obj); 2509 if (vp != NULL) { 2510 vn_fullpath(curthread, vp, &fullpath, 2511 &freepath); 2512 kve->kve_vn_type = vntype_to_kinfo(vp->v_type); 2513 cred = curthread->td_ucred; 2514 vn_lock(vp, LK_SHARED | LK_RETRY); 2515 if (VOP_GETATTR(vp, &va, cred) == 0) { 2516 kve->kve_vn_fileid = va.va_fileid; 2517 kve->kve_vn_fsid = va.va_fsid; 2518 kve->kve_vn_fsid_freebsd11 = 2519 kve->kve_vn_fsid; /* truncate */ 2520 kve->kve_vn_mode = 2521 MAKEIMODE(va.va_type, va.va_mode); 2522 kve->kve_vn_size = va.va_size; 2523 kve->kve_vn_rdev = va.va_rdev; 2524 kve->kve_vn_rdev_freebsd11 = 2525 kve->kve_vn_rdev; /* truncate */ 2526 kve->kve_status = KF_ATTR_VALID; 2527 } 2528 vput(vp); 2529 } 2530 } else { 2531 kve->kve_type = KVME_TYPE_NONE; 2532 kve->kve_ref_count = 0; 2533 kve->kve_shadow_count = 0; 2534 } 2535 2536 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2537 if (freepath != NULL) 2538 free(freepath, M_TEMP); 2539 2540 /* Pack record size down */ 2541 if ((flags & KERN_VMMAP_PACK_KINFO) != 0) 2542 kve->kve_structsize = 2543 offsetof(struct kinfo_vmentry, kve_path) + 2544 strlen(kve->kve_path) + 1; 2545 else 2546 kve->kve_structsize = sizeof(*kve); 2547 kve->kve_structsize = roundup(kve->kve_structsize, 2548 sizeof(uint64_t)); 2549 2550 /* Halt filling and truncate rather than exceeding maxlen */ 2551 if (maxlen != -1 && maxlen < kve->kve_structsize) { 2552 error = 0; 2553 vm_map_lock_read(map); 2554 break; 2555 } else if (maxlen != -1) 2556 maxlen -= kve->kve_structsize; 2557 2558 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0) 2559 error = ENOMEM; 2560 vm_map_lock_read(map); 2561 if (error != 0) 2562 break; 2563 if (last_timestamp != map->timestamp) { 2564 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2565 entry = tmp_entry; 2566 } 2567 } 2568 vm_map_unlock_read(map); 2569 vmspace_free(vm); 2570 PRELE(p); 2571 free(kve, M_TEMP); 2572 return (error); 2573 } 2574 2575 static int 2576 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) 2577 { 2578 struct proc *p; 2579 struct sbuf sb; 2580 int error, error2, *name; 2581 2582 name = (int *)arg1; 2583 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); 2584 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2585 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); 2586 if (error != 0) { 2587 sbuf_delete(&sb); 2588 return (error); 2589 } 2590 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO); 2591 error2 = sbuf_finish(&sb); 2592 sbuf_delete(&sb); 2593 return (error != 0 ? error : error2); 2594 } 2595 2596 #if defined(STACK) || defined(DDB) 2597 static int 2598 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) 2599 { 2600 struct kinfo_kstack *kkstp; 2601 int error, i, *name, numthreads; 2602 lwpid_t *lwpidarray; 2603 struct thread *td; 2604 struct stack *st; 2605 struct sbuf sb; 2606 struct proc *p; 2607 2608 name = (int *)arg1; 2609 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); 2610 if (error != 0) 2611 return (error); 2612 2613 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); 2614 st = stack_create(M_WAITOK); 2615 2616 lwpidarray = NULL; 2617 PROC_LOCK(p); 2618 do { 2619 if (lwpidarray != NULL) { 2620 free(lwpidarray, M_TEMP); 2621 lwpidarray = NULL; 2622 } 2623 numthreads = p->p_numthreads; 2624 PROC_UNLOCK(p); 2625 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, 2626 M_WAITOK | M_ZERO); 2627 PROC_LOCK(p); 2628 } while (numthreads < p->p_numthreads); 2629 2630 /* 2631 * XXXRW: During the below loop, execve(2) and countless other sorts 2632 * of changes could have taken place. Should we check to see if the 2633 * vmspace has been replaced, or the like, in order to prevent 2634 * giving a snapshot that spans, say, execve(2), with some threads 2635 * before and some after? Among other things, the credentials could 2636 * have changed, in which case the right to extract debug info might 2637 * no longer be assured. 2638 */ 2639 i = 0; 2640 FOREACH_THREAD_IN_PROC(p, td) { 2641 KASSERT(i < numthreads, 2642 ("sysctl_kern_proc_kstack: numthreads")); 2643 lwpidarray[i] = td->td_tid; 2644 i++; 2645 } 2646 numthreads = i; 2647 for (i = 0; i < numthreads; i++) { 2648 td = thread_find(p, lwpidarray[i]); 2649 if (td == NULL) { 2650 continue; 2651 } 2652 bzero(kkstp, sizeof(*kkstp)); 2653 (void)sbuf_new(&sb, kkstp->kkst_trace, 2654 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); 2655 thread_lock(td); 2656 kkstp->kkst_tid = td->td_tid; 2657 if (TD_IS_SWAPPED(td)) 2658 kkstp->kkst_state = KKST_STATE_SWAPPED; 2659 else if (stack_save_td(st, td) == 0) 2660 kkstp->kkst_state = KKST_STATE_STACKOK; 2661 else 2662 kkstp->kkst_state = KKST_STATE_RUNNING; 2663 thread_unlock(td); 2664 PROC_UNLOCK(p); 2665 stack_sbuf_print(&sb, st); 2666 sbuf_finish(&sb); 2667 sbuf_delete(&sb); 2668 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); 2669 PROC_LOCK(p); 2670 if (error) 2671 break; 2672 } 2673 _PRELE(p); 2674 PROC_UNLOCK(p); 2675 if (lwpidarray != NULL) 2676 free(lwpidarray, M_TEMP); 2677 stack_destroy(st); 2678 free(kkstp, M_TEMP); 2679 return (error); 2680 } 2681 #endif 2682 2683 /* 2684 * This sysctl allows a process to retrieve the full list of groups from 2685 * itself or another process. 2686 */ 2687 static int 2688 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) 2689 { 2690 pid_t *pidp = (pid_t *)arg1; 2691 unsigned int arglen = arg2; 2692 struct proc *p; 2693 struct ucred *cred; 2694 int error; 2695 2696 if (arglen != 1) 2697 return (EINVAL); 2698 if (*pidp == -1) { /* -1 means this process */ 2699 p = req->td->td_proc; 2700 PROC_LOCK(p); 2701 } else { 2702 error = pget(*pidp, PGET_CANSEE, &p); 2703 if (error != 0) 2704 return (error); 2705 } 2706 2707 cred = crhold(p->p_ucred); 2708 PROC_UNLOCK(p); 2709 2710 error = SYSCTL_OUT(req, cred->cr_groups, 2711 cred->cr_ngroups * sizeof(gid_t)); 2712 crfree(cred); 2713 return (error); 2714 } 2715 2716 /* 2717 * This sysctl allows a process to retrieve or/and set the resource limit for 2718 * another process. 2719 */ 2720 static int 2721 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) 2722 { 2723 int *name = (int *)arg1; 2724 u_int namelen = arg2; 2725 struct rlimit rlim; 2726 struct proc *p; 2727 u_int which; 2728 int flags, error; 2729 2730 if (namelen != 2) 2731 return (EINVAL); 2732 2733 which = (u_int)name[1]; 2734 if (which >= RLIM_NLIMITS) 2735 return (EINVAL); 2736 2737 if (req->newptr != NULL && req->newlen != sizeof(rlim)) 2738 return (EINVAL); 2739 2740 flags = PGET_HOLD | PGET_NOTWEXIT; 2741 if (req->newptr != NULL) 2742 flags |= PGET_CANDEBUG; 2743 else 2744 flags |= PGET_CANSEE; 2745 error = pget((pid_t)name[0], flags, &p); 2746 if (error != 0) 2747 return (error); 2748 2749 /* 2750 * Retrieve limit. 2751 */ 2752 if (req->oldptr != NULL) { 2753 PROC_LOCK(p); 2754 lim_rlimit_proc(p, which, &rlim); 2755 PROC_UNLOCK(p); 2756 } 2757 error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); 2758 if (error != 0) 2759 goto errout; 2760 2761 /* 2762 * Set limit. 2763 */ 2764 if (req->newptr != NULL) { 2765 error = SYSCTL_IN(req, &rlim, sizeof(rlim)); 2766 if (error == 0) 2767 error = kern_proc_setrlimit(curthread, p, which, &rlim); 2768 } 2769 2770 errout: 2771 PRELE(p); 2772 return (error); 2773 } 2774 2775 /* 2776 * This sysctl allows a process to retrieve ps_strings structure location of 2777 * another process. 2778 */ 2779 static int 2780 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) 2781 { 2782 int *name = (int *)arg1; 2783 u_int namelen = arg2; 2784 struct proc *p; 2785 vm_offset_t ps_strings; 2786 int error; 2787 #ifdef COMPAT_FREEBSD32 2788 uint32_t ps_strings32; 2789 #endif 2790 2791 if (namelen != 1) 2792 return (EINVAL); 2793 2794 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2795 if (error != 0) 2796 return (error); 2797 #ifdef COMPAT_FREEBSD32 2798 if ((req->flags & SCTL_MASK32) != 0) { 2799 /* 2800 * We return 0 if the 32 bit emulation request is for a 64 bit 2801 * process. 2802 */ 2803 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? 2804 PTROUT(p->p_sysent->sv_psstrings) : 0; 2805 PROC_UNLOCK(p); 2806 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); 2807 return (error); 2808 } 2809 #endif 2810 ps_strings = p->p_sysent->sv_psstrings; 2811 PROC_UNLOCK(p); 2812 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); 2813 return (error); 2814 } 2815 2816 /* 2817 * This sysctl allows a process to retrieve umask of another process. 2818 */ 2819 static int 2820 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) 2821 { 2822 int *name = (int *)arg1; 2823 u_int namelen = arg2; 2824 struct proc *p; 2825 int error; 2826 u_short fd_cmask; 2827 pid_t pid; 2828 2829 if (namelen != 1) 2830 return (EINVAL); 2831 2832 pid = (pid_t)name[0]; 2833 p = curproc; 2834 if (pid == p->p_pid || pid == 0) { 2835 fd_cmask = p->p_fd->fd_cmask; 2836 goto out; 2837 } 2838 2839 error = pget(pid, PGET_WANTREAD, &p); 2840 if (error != 0) 2841 return (error); 2842 2843 fd_cmask = p->p_fd->fd_cmask; 2844 PRELE(p); 2845 out: 2846 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask)); 2847 return (error); 2848 } 2849 2850 /* 2851 * This sysctl allows a process to set and retrieve binary osreldate of 2852 * another process. 2853 */ 2854 static int 2855 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) 2856 { 2857 int *name = (int *)arg1; 2858 u_int namelen = arg2; 2859 struct proc *p; 2860 int flags, error, osrel; 2861 2862 if (namelen != 1) 2863 return (EINVAL); 2864 2865 if (req->newptr != NULL && req->newlen != sizeof(osrel)) 2866 return (EINVAL); 2867 2868 flags = PGET_HOLD | PGET_NOTWEXIT; 2869 if (req->newptr != NULL) 2870 flags |= PGET_CANDEBUG; 2871 else 2872 flags |= PGET_CANSEE; 2873 error = pget((pid_t)name[0], flags, &p); 2874 if (error != 0) 2875 return (error); 2876 2877 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); 2878 if (error != 0) 2879 goto errout; 2880 2881 if (req->newptr != NULL) { 2882 error = SYSCTL_IN(req, &osrel, sizeof(osrel)); 2883 if (error != 0) 2884 goto errout; 2885 if (osrel < 0) { 2886 error = EINVAL; 2887 goto errout; 2888 } 2889 p->p_osrel = osrel; 2890 } 2891 errout: 2892 PRELE(p); 2893 return (error); 2894 } 2895 2896 static int 2897 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) 2898 { 2899 int *name = (int *)arg1; 2900 u_int namelen = arg2; 2901 struct proc *p; 2902 struct kinfo_sigtramp kst; 2903 const struct sysentvec *sv; 2904 int error; 2905 #ifdef COMPAT_FREEBSD32 2906 struct kinfo_sigtramp32 kst32; 2907 #endif 2908 2909 if (namelen != 1) 2910 return (EINVAL); 2911 2912 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 2913 if (error != 0) 2914 return (error); 2915 sv = p->p_sysent; 2916 #ifdef COMPAT_FREEBSD32 2917 if ((req->flags & SCTL_MASK32) != 0) { 2918 bzero(&kst32, sizeof(kst32)); 2919 if (SV_PROC_FLAG(p, SV_ILP32)) { 2920 if (sv->sv_sigcode_base != 0) { 2921 kst32.ksigtramp_start = sv->sv_sigcode_base; 2922 kst32.ksigtramp_end = sv->sv_sigcode_base + 2923 *sv->sv_szsigcode; 2924 } else { 2925 kst32.ksigtramp_start = sv->sv_psstrings - 2926 *sv->sv_szsigcode; 2927 kst32.ksigtramp_end = sv->sv_psstrings; 2928 } 2929 } 2930 PROC_UNLOCK(p); 2931 error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); 2932 return (error); 2933 } 2934 #endif 2935 bzero(&kst, sizeof(kst)); 2936 if (sv->sv_sigcode_base != 0) { 2937 kst.ksigtramp_start = (char *)sv->sv_sigcode_base; 2938 kst.ksigtramp_end = (char *)sv->sv_sigcode_base + 2939 *sv->sv_szsigcode; 2940 } else { 2941 kst.ksigtramp_start = (char *)sv->sv_psstrings - 2942 *sv->sv_szsigcode; 2943 kst.ksigtramp_end = (char *)sv->sv_psstrings; 2944 } 2945 PROC_UNLOCK(p); 2946 error = SYSCTL_OUT(req, &kst, sizeof(kst)); 2947 return (error); 2948 } 2949 2950 static int 2951 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS) 2952 { 2953 int *name = (int *)arg1; 2954 u_int namelen = arg2; 2955 pid_t pid; 2956 struct proc *p; 2957 struct thread *td1; 2958 uintptr_t addr; 2959 #ifdef COMPAT_FREEBSD32 2960 uint32_t addr32; 2961 #endif 2962 int error; 2963 2964 if (namelen != 1 || req->newptr != NULL) 2965 return (EINVAL); 2966 2967 pid = (pid_t)name[0]; 2968 error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p); 2969 if (error != 0) 2970 return (error); 2971 2972 PROC_LOCK(p); 2973 #ifdef COMPAT_FREEBSD32 2974 if (SV_CURPROC_FLAG(SV_ILP32)) { 2975 if (!SV_PROC_FLAG(p, SV_ILP32)) { 2976 error = EINVAL; 2977 goto errlocked; 2978 } 2979 } 2980 #endif 2981 if (pid <= PID_MAX) { 2982 td1 = FIRST_THREAD_IN_PROC(p); 2983 } else { 2984 FOREACH_THREAD_IN_PROC(p, td1) { 2985 if (td1->td_tid == pid) 2986 break; 2987 } 2988 } 2989 if (td1 == NULL) { 2990 error = ESRCH; 2991 goto errlocked; 2992 } 2993 /* 2994 * The access to the private thread flags. It is fine as far 2995 * as no out-of-thin-air values are read from td_pflags, and 2996 * usermode read of the td_sigblock_ptr is racy inherently, 2997 * since target process might have already changed it 2998 * meantime. 2999 */ 3000 if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0) 3001 addr = (uintptr_t)td1->td_sigblock_ptr; 3002 else 3003 error = ENOTTY; 3004 3005 errlocked: 3006 _PRELE(p); 3007 PROC_UNLOCK(p); 3008 if (error != 0) 3009 return (error); 3010 3011 #ifdef COMPAT_FREEBSD32 3012 if (SV_CURPROC_FLAG(SV_ILP32)) { 3013 addr32 = addr; 3014 error = SYSCTL_OUT(req, &addr32, sizeof(addr32)); 3015 } else 3016 #endif 3017 error = SYSCTL_OUT(req, &addr, sizeof(addr)); 3018 return (error); 3019 } 3020 3021 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 3022 "Process table"); 3023 3024 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| 3025 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", 3026 "Return entire process table"); 3027 3028 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3029 sysctl_kern_proc, "Process table"); 3030 3031 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, 3032 sysctl_kern_proc, "Process table"); 3033 3034 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3035 sysctl_kern_proc, "Process table"); 3036 3037 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | 3038 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3039 3040 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 3041 sysctl_kern_proc, "Process table"); 3042 3043 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3044 sysctl_kern_proc, "Process table"); 3045 3046 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3047 sysctl_kern_proc, "Process table"); 3048 3049 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3050 sysctl_kern_proc, "Process table"); 3051 3052 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 3053 sysctl_kern_proc, "Return process table, no threads"); 3054 3055 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, 3056 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, 3057 sysctl_kern_proc_args, "Process argument list"); 3058 3059 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, 3060 sysctl_kern_proc_env, "Process environment"); 3061 3062 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | 3063 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); 3064 3065 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | 3066 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); 3067 3068 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | 3069 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, 3070 "Process syscall vector name (ABI type)"); 3071 3072 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, 3073 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3074 3075 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, 3076 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3077 3078 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, 3079 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3080 3081 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), 3082 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3083 3084 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, 3085 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3086 3087 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, 3088 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3089 3090 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, 3091 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3092 3093 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, 3094 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3095 3096 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, 3097 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, 3098 "Return process table, no threads"); 3099 3100 #ifdef COMPAT_FREEBSD7 3101 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | 3102 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); 3103 #endif 3104 3105 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | 3106 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); 3107 3108 #if defined(STACK) || defined(DDB) 3109 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | 3110 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); 3111 #endif 3112 3113 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | 3114 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); 3115 3116 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | 3117 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, 3118 "Process resource limits"); 3119 3120 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | 3121 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, 3122 "Process ps_strings location"); 3123 3124 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | 3125 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); 3126 3127 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | 3128 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, 3129 "Process binary osreldate"); 3130 3131 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | 3132 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, 3133 "Process signal trampoline location"); 3134 3135 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD | 3136 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk, 3137 "Thread sigfastblock address"); 3138 3139 int allproc_gen; 3140 3141 /* 3142 * stop_all_proc() purpose is to stop all process which have usermode, 3143 * except current process for obvious reasons. This makes it somewhat 3144 * unreliable when invoked from multithreaded process. The service 3145 * must not be user-callable anyway. 3146 */ 3147 void 3148 stop_all_proc(void) 3149 { 3150 struct proc *cp, *p; 3151 int r, gen; 3152 bool restart, seen_stopped, seen_exiting, stopped_some; 3153 3154 cp = curproc; 3155 allproc_loop: 3156 sx_xlock(&allproc_lock); 3157 gen = allproc_gen; 3158 seen_exiting = seen_stopped = stopped_some = restart = false; 3159 LIST_REMOVE(cp, p_list); 3160 LIST_INSERT_HEAD(&allproc, cp, p_list); 3161 for (;;) { 3162 p = LIST_NEXT(cp, p_list); 3163 if (p == NULL) 3164 break; 3165 LIST_REMOVE(cp, p_list); 3166 LIST_INSERT_AFTER(p, cp, p_list); 3167 PROC_LOCK(p); 3168 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) { 3169 PROC_UNLOCK(p); 3170 continue; 3171 } 3172 if ((p->p_flag & P_WEXIT) != 0) { 3173 seen_exiting = true; 3174 PROC_UNLOCK(p); 3175 continue; 3176 } 3177 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 3178 /* 3179 * Stopped processes are tolerated when there 3180 * are no other processes which might continue 3181 * them. P_STOPPED_SINGLE but not 3182 * P_TOTAL_STOP process still has at least one 3183 * thread running. 3184 */ 3185 seen_stopped = true; 3186 PROC_UNLOCK(p); 3187 continue; 3188 } 3189 sx_xunlock(&allproc_lock); 3190 _PHOLD(p); 3191 r = thread_single(p, SINGLE_ALLPROC); 3192 if (r != 0) 3193 restart = true; 3194 else 3195 stopped_some = true; 3196 _PRELE(p); 3197 PROC_UNLOCK(p); 3198 sx_xlock(&allproc_lock); 3199 } 3200 /* Catch forked children we did not see in iteration. */ 3201 if (gen != allproc_gen) 3202 restart = true; 3203 sx_xunlock(&allproc_lock); 3204 if (restart || stopped_some || seen_exiting || seen_stopped) { 3205 kern_yield(PRI_USER); 3206 goto allproc_loop; 3207 } 3208 } 3209 3210 void 3211 resume_all_proc(void) 3212 { 3213 struct proc *cp, *p; 3214 3215 cp = curproc; 3216 sx_xlock(&allproc_lock); 3217 again: 3218 LIST_REMOVE(cp, p_list); 3219 LIST_INSERT_HEAD(&allproc, cp, p_list); 3220 for (;;) { 3221 p = LIST_NEXT(cp, p_list); 3222 if (p == NULL) 3223 break; 3224 LIST_REMOVE(cp, p_list); 3225 LIST_INSERT_AFTER(p, cp, p_list); 3226 PROC_LOCK(p); 3227 if ((p->p_flag & P_TOTAL_STOP) != 0) { 3228 sx_xunlock(&allproc_lock); 3229 _PHOLD(p); 3230 thread_single_end(p, SINGLE_ALLPROC); 3231 _PRELE(p); 3232 PROC_UNLOCK(p); 3233 sx_xlock(&allproc_lock); 3234 } else { 3235 PROC_UNLOCK(p); 3236 } 3237 } 3238 /* Did the loop above missed any stopped process ? */ 3239 FOREACH_PROC_IN_SYSTEM(p) { 3240 /* No need for proc lock. */ 3241 if ((p->p_flag & P_TOTAL_STOP) != 0) 3242 goto again; 3243 } 3244 sx_xunlock(&allproc_lock); 3245 } 3246 3247 /* #define TOTAL_STOP_DEBUG 1 */ 3248 #ifdef TOTAL_STOP_DEBUG 3249 volatile static int ap_resume; 3250 #include <sys/mount.h> 3251 3252 static int 3253 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) 3254 { 3255 int error, val; 3256 3257 val = 0; 3258 ap_resume = 0; 3259 error = sysctl_handle_int(oidp, &val, 0, req); 3260 if (error != 0 || req->newptr == NULL) 3261 return (error); 3262 if (val != 0) { 3263 stop_all_proc(); 3264 syncer_suspend(); 3265 while (ap_resume == 0) 3266 ; 3267 syncer_resume(); 3268 resume_all_proc(); 3269 } 3270 return (0); 3271 } 3272 3273 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | 3274 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, 3275 sysctl_debug_stop_all_proc, "I", 3276 ""); 3277 #endif 3278