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/conf.h> 42 #include <sys/elf.h> 43 #include <sys/eventhandler.h> 44 #include <sys/exec.h> 45 #include <sys/fcntl.h> 46 #include <sys/ipc.h> 47 #include <sys/jail.h> 48 #include <sys/kernel.h> 49 #include <sys/limits.h> 50 #include <sys/lock.h> 51 #include <sys/loginclass.h> 52 #include <sys/malloc.h> 53 #include <sys/mman.h> 54 #include <sys/mount.h> 55 #include <sys/mutex.h> 56 #include <sys/namei.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/shm.h> 66 #include <sys/smp.h> 67 #include <sys/stack.h> 68 #include <sys/stat.h> 69 #include <sys/dtrace_bsd.h> 70 #include <sys/sysctl.h> 71 #include <sys/filedesc.h> 72 #include <sys/tty.h> 73 #include <sys/signalvar.h> 74 #include <sys/sdt.h> 75 #include <sys/sx.h> 76 #include <sys/user.h> 77 #include <sys/vnode.h> 78 #include <sys/wait.h> 79 #ifdef KTRACE 80 #include <sys/ktrace.h> 81 #endif 82 83 #ifdef DDB 84 #include <ddb/ddb.h> 85 #endif 86 87 #include <vm/vm.h> 88 #include <vm/vm_param.h> 89 #include <vm/vm_extern.h> 90 #include <vm/pmap.h> 91 #include <vm/vm_map.h> 92 #include <vm/vm_object.h> 93 #include <vm/vm_page.h> 94 #include <vm/vm_pager.h> 95 #include <vm/vm_radix.h> 96 #include <vm/uma.h> 97 98 #include <fs/devfs/devfs.h> 99 100 #ifdef COMPAT_FREEBSD32 101 #include <compat/freebsd32/freebsd32.h> 102 #include <compat/freebsd32/freebsd32_util.h> 103 #endif 104 105 SDT_PROVIDER_DEFINE(proc); 106 107 MALLOC_DEFINE(M_SESSION, "session", "session header"); 108 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); 109 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 110 111 static void doenterpgrp(struct proc *, struct pgrp *); 112 static void orphanpg(struct pgrp *pg); 113 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp); 114 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp); 115 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, 116 int preferthread); 117 static void pgdelete(struct pgrp *); 118 static int pgrp_init(void *mem, int size, int flags); 119 static int proc_ctor(void *mem, int size, void *arg, int flags); 120 static void proc_dtor(void *mem, int size, void *arg); 121 static int proc_init(void *mem, int size, int flags); 122 static void proc_fini(void *mem, int size); 123 static void pargs_free(struct pargs *pa); 124 125 /* 126 * Other process lists 127 */ 128 struct pidhashhead *pidhashtbl = NULL; 129 struct sx *pidhashtbl_lock; 130 u_long pidhash; 131 u_long pidhashlock; 132 struct pgrphashhead *pgrphashtbl; 133 u_long pgrphash; 134 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc); 135 struct sx __exclusive_cache_line allproc_lock; 136 struct sx __exclusive_cache_line proctree_lock; 137 struct mtx __exclusive_cache_line ppeers_lock; 138 struct mtx __exclusive_cache_line procid_lock; 139 uma_zone_t proc_zone; 140 uma_zone_t pgrp_zone; 141 142 /* 143 * The offset of various fields in struct proc and struct thread. 144 * These are used by kernel debuggers to enumerate kernel threads and 145 * processes. 146 */ 147 const int proc_off_p_pid = offsetof(struct proc, p_pid); 148 const int proc_off_p_comm = offsetof(struct proc, p_comm); 149 const int proc_off_p_list = offsetof(struct proc, p_list); 150 const int proc_off_p_hash = offsetof(struct proc, p_hash); 151 const int proc_off_p_threads = offsetof(struct proc, p_threads); 152 const int thread_off_td_tid = offsetof(struct thread, td_tid); 153 const int thread_off_td_name = offsetof(struct thread, td_name); 154 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu); 155 const int thread_off_td_pcb = offsetof(struct thread, td_pcb); 156 const int thread_off_td_plist = offsetof(struct thread, td_plist); 157 158 EVENTHANDLER_LIST_DEFINE(process_ctor); 159 EVENTHANDLER_LIST_DEFINE(process_dtor); 160 EVENTHANDLER_LIST_DEFINE(process_init); 161 EVENTHANDLER_LIST_DEFINE(process_fini); 162 EVENTHANDLER_LIST_DEFINE(process_exit); 163 EVENTHANDLER_LIST_DEFINE(process_fork); 164 EVENTHANDLER_LIST_DEFINE(process_exec); 165 166 int kstack_pages = KSTACK_PAGES; 167 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, 168 &kstack_pages, 0, 169 "Kernel stack size in pages"); 170 static int vmmap_skip_res_cnt = 0; 171 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW, 172 &vmmap_skip_res_cnt, 0, 173 "Skip calculation of the pages resident count in kern.proc.vmmap"); 174 175 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); 176 #ifdef COMPAT_FREEBSD32 177 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE); 178 #endif 179 180 /* 181 * Initialize global process hashing structures. 182 */ 183 void 184 procinit(void) 185 { 186 u_long i; 187 188 sx_init(&allproc_lock, "allproc"); 189 sx_init(&proctree_lock, "proctree"); 190 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF); 191 mtx_init(&procid_lock, "procid", NULL, MTX_DEF); 192 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); 193 pidhashlock = (pidhash + 1) / 64; 194 if (pidhashlock > 0) 195 pidhashlock--; 196 pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1), 197 M_PROC, M_WAITOK | M_ZERO); 198 for (i = 0; i < pidhashlock + 1; i++) 199 sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK); 200 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); 201 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(), 202 proc_ctor, proc_dtor, proc_init, proc_fini, 203 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 204 pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL, 205 pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 206 uihashinit(); 207 } 208 209 /* 210 * Prepare a proc for use. 211 */ 212 static int 213 proc_ctor(void *mem, int size, void *arg, int flags) 214 { 215 struct proc *p; 216 struct thread *td; 217 218 p = (struct proc *)mem; 219 #ifdef KDTRACE_HOOKS 220 kdtrace_proc_ctor(p); 221 #endif 222 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p); 223 td = FIRST_THREAD_IN_PROC(p); 224 if (td != NULL) { 225 /* Make sure all thread constructors are executed */ 226 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td); 227 } 228 return (0); 229 } 230 231 /* 232 * Reclaim a proc after use. 233 */ 234 static void 235 proc_dtor(void *mem, int size, void *arg) 236 { 237 struct proc *p; 238 struct thread *td; 239 240 p = mem; 241 td = FIRST_THREAD_IN_PROC(p); 242 if (td != NULL) { 243 KASSERT(p->p_numthreads == 1, 244 ("too many threads in exiting process")); 245 246 /* Free all OSD associated to this thread. */ 247 osd_thread_exit(td); 248 ast_kclear(td); 249 250 /* Make sure all thread destructors are executed */ 251 EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td); 252 } 253 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr")); 254 EVENTHANDLER_DIRECT_INVOKE(process_dtor, p); 255 #ifdef KDTRACE_HOOKS 256 kdtrace_proc_dtor(p); 257 #endif 258 KASSERT(p->p_ksi == NULL || !KSI_ONQ(p->p_ksi), ("SIGCHLD queue")); 259 } 260 261 /* 262 * Initialize type-stable parts of a proc (when newly created). 263 */ 264 static int 265 proc_init(void *mem, int size, int flags) 266 { 267 struct proc *p; 268 269 p = (struct proc *)mem; 270 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW); 271 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW); 272 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW); 273 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW); 274 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW); 275 cv_init(&p->p_pwait, "ppwait"); 276 TAILQ_INIT(&p->p_threads); /* all threads in proc */ 277 EVENTHANDLER_DIRECT_INVOKE(process_init, p); 278 p->p_stats = pstats_alloc(); 279 p->p_pgrp = NULL; 280 TAILQ_INIT(&p->p_kqtim_stop); 281 STAILQ_INIT(&p->p_ktr); 282 return (0); 283 } 284 285 /* 286 * UMA should ensure that this function is never called. 287 * Freeing a proc structure would violate type stability. 288 */ 289 static void 290 proc_fini(void *mem, int size) 291 { 292 #ifdef notnow 293 struct proc *p; 294 295 p = (struct proc *)mem; 296 EVENTHANDLER_DIRECT_INVOKE(process_fini, p); 297 pstats_free(p->p_stats); 298 thread_free(FIRST_THREAD_IN_PROC(p)); 299 mtx_destroy(&p->p_mtx); 300 if (p->p_ksi != NULL) 301 ksiginfo_free(p->p_ksi); 302 #else 303 panic("proc reclaimed"); 304 #endif 305 } 306 307 static int 308 pgrp_init(void *mem, int size, int flags) 309 { 310 struct pgrp *pg; 311 312 pg = mem; 313 mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 314 sx_init(&pg->pg_killsx, "killpg racer"); 315 return (0); 316 } 317 318 /* 319 * PID space management. 320 * 321 * These bitmaps are used by fork_findpid. 322 */ 323 bitstr_t bit_decl(proc_id_pidmap, PID_MAX); 324 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX); 325 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX); 326 bitstr_t bit_decl(proc_id_reapmap, PID_MAX); 327 328 static bitstr_t *proc_id_array[] = { 329 proc_id_pidmap, 330 proc_id_grpidmap, 331 proc_id_sessidmap, 332 proc_id_reapmap, 333 }; 334 335 void 336 proc_id_set(int type, pid_t id) 337 { 338 339 KASSERT(type >= 0 && type < nitems(proc_id_array), 340 ("invalid type %d\n", type)); 341 mtx_lock(&procid_lock); 342 KASSERT(bit_test(proc_id_array[type], id) == 0, 343 ("bit %d already set in %d\n", id, type)); 344 bit_set(proc_id_array[type], id); 345 mtx_unlock(&procid_lock); 346 } 347 348 void 349 proc_id_set_cond(int type, pid_t id) 350 { 351 352 KASSERT(type >= 0 && type < nitems(proc_id_array), 353 ("invalid type %d\n", type)); 354 if (bit_test(proc_id_array[type], id)) 355 return; 356 mtx_lock(&procid_lock); 357 bit_set(proc_id_array[type], id); 358 mtx_unlock(&procid_lock); 359 } 360 361 void 362 proc_id_clear(int type, pid_t id) 363 { 364 365 KASSERT(type >= 0 && type < nitems(proc_id_array), 366 ("invalid type %d\n", type)); 367 mtx_lock(&procid_lock); 368 KASSERT(bit_test(proc_id_array[type], id) != 0, 369 ("bit %d not set in %d\n", id, type)); 370 bit_clear(proc_id_array[type], id); 371 mtx_unlock(&procid_lock); 372 } 373 374 /* 375 * Is p an inferior of the current process? 376 */ 377 int 378 inferior(struct proc *p) 379 { 380 381 sx_assert(&proctree_lock, SX_LOCKED); 382 PROC_LOCK_ASSERT(p, MA_OWNED); 383 for (; p != curproc; p = proc_realparent(p)) { 384 if (p->p_pid == 0) 385 return (0); 386 } 387 return (1); 388 } 389 390 /* 391 * Shared lock all the pid hash lists. 392 */ 393 void 394 pidhash_slockall(void) 395 { 396 u_long i; 397 398 for (i = 0; i < pidhashlock + 1; i++) 399 sx_slock(&pidhashtbl_lock[i]); 400 } 401 402 /* 403 * Shared unlock all the pid hash lists. 404 */ 405 void 406 pidhash_sunlockall(void) 407 { 408 u_long i; 409 410 for (i = 0; i < pidhashlock + 1; i++) 411 sx_sunlock(&pidhashtbl_lock[i]); 412 } 413 414 /* 415 * Similar to pfind(), this function locate a process by number. 416 */ 417 struct proc * 418 pfind_any_locked(pid_t pid) 419 { 420 struct proc *p; 421 422 sx_assert(PIDHASHLOCK(pid), SX_LOCKED); 423 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 424 if (p->p_pid == pid) { 425 PROC_LOCK(p); 426 if (p->p_state == PRS_NEW) { 427 PROC_UNLOCK(p); 428 p = NULL; 429 } 430 break; 431 } 432 } 433 return (p); 434 } 435 436 /* 437 * Locate a process by number. 438 * 439 * By not returning processes in the PRS_NEW state, we allow callers to avoid 440 * testing for that condition to avoid dereferencing p_ucred, et al. 441 */ 442 static __always_inline struct proc * 443 _pfind(pid_t pid, bool zombie) 444 { 445 struct proc *p; 446 447 p = curproc; 448 if (p->p_pid == pid) { 449 PROC_LOCK(p); 450 return (p); 451 } 452 sx_slock(PIDHASHLOCK(pid)); 453 LIST_FOREACH(p, PIDHASH(pid), p_hash) { 454 if (p->p_pid == pid) { 455 PROC_LOCK(p); 456 if (p->p_state == PRS_NEW || 457 (!zombie && p->p_state == PRS_ZOMBIE)) { 458 PROC_UNLOCK(p); 459 p = NULL; 460 } 461 break; 462 } 463 } 464 sx_sunlock(PIDHASHLOCK(pid)); 465 return (p); 466 } 467 468 struct proc * 469 pfind(pid_t pid) 470 { 471 472 return (_pfind(pid, false)); 473 } 474 475 /* 476 * Same as pfind but allow zombies. 477 */ 478 struct proc * 479 pfind_any(pid_t pid) 480 { 481 482 return (_pfind(pid, true)); 483 } 484 485 /* 486 * Locate a process group by number. 487 * The caller must hold proctree_lock. 488 */ 489 struct pgrp * 490 pgfind(pid_t pgid) 491 { 492 struct pgrp *pgrp; 493 494 sx_assert(&proctree_lock, SX_LOCKED); 495 496 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { 497 if (pgrp->pg_id == pgid) { 498 PGRP_LOCK(pgrp); 499 return (pgrp); 500 } 501 } 502 return (NULL); 503 } 504 505 /* 506 * Locate process and do additional manipulations, depending on flags. 507 */ 508 int 509 pget(pid_t pid, int flags, struct proc **pp) 510 { 511 struct proc *p; 512 struct thread *td1; 513 int error; 514 515 p = curproc; 516 if (p->p_pid == pid) { 517 PROC_LOCK(p); 518 } else { 519 p = NULL; 520 if (pid <= PID_MAX) { 521 if ((flags & PGET_NOTWEXIT) == 0) 522 p = pfind_any(pid); 523 else 524 p = pfind(pid); 525 } else if ((flags & PGET_NOTID) == 0) { 526 td1 = tdfind(pid, -1); 527 if (td1 != NULL) 528 p = td1->td_proc; 529 } 530 if (p == NULL) 531 return (ESRCH); 532 if ((flags & PGET_CANSEE) != 0) { 533 error = p_cansee(curthread, p); 534 if (error != 0) 535 goto errout; 536 } 537 } 538 if ((flags & PGET_CANDEBUG) != 0) { 539 error = p_candebug(curthread, p); 540 if (error != 0) 541 goto errout; 542 } 543 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) { 544 error = EPERM; 545 goto errout; 546 } 547 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) { 548 error = ESRCH; 549 goto errout; 550 } 551 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) { 552 /* 553 * XXXRW: Not clear ESRCH is the right error during proc 554 * execve(). 555 */ 556 error = ESRCH; 557 goto errout; 558 } 559 if ((flags & PGET_HOLD) != 0) { 560 _PHOLD(p); 561 PROC_UNLOCK(p); 562 } 563 *pp = p; 564 return (0); 565 errout: 566 PROC_UNLOCK(p); 567 return (error); 568 } 569 570 /* 571 * Create a new process group. 572 * pgid must be equal to the pid of p. 573 * Begin a new session if required. 574 */ 575 int 576 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess) 577 { 578 struct pgrp *old_pgrp; 579 580 sx_assert(&proctree_lock, SX_XLOCKED); 581 582 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL")); 583 KASSERT(p->p_pid == pgid, 584 ("enterpgrp: new pgrp and pid != pgid")); 585 KASSERT(pgfind(pgid) == NULL, 586 ("enterpgrp: pgrp with pgid exists")); 587 KASSERT(!SESS_LEADER(p), 588 ("enterpgrp: session leader attempted setpgrp")); 589 590 old_pgrp = p->p_pgrp; 591 if (!sx_try_xlock(&old_pgrp->pg_killsx)) { 592 sx_xunlock(&proctree_lock); 593 sx_xlock(&old_pgrp->pg_killsx); 594 sx_xunlock(&old_pgrp->pg_killsx); 595 return (ERESTART); 596 } 597 MPASS(old_pgrp == p->p_pgrp); 598 599 if (sess != NULL) { 600 /* 601 * new session 602 */ 603 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); 604 PROC_LOCK(p); 605 p->p_flag &= ~P_CONTROLT; 606 PROC_UNLOCK(p); 607 PGRP_LOCK(pgrp); 608 sess->s_leader = p; 609 sess->s_sid = p->p_pid; 610 proc_id_set(PROC_ID_SESSION, p->p_pid); 611 refcount_init(&sess->s_count, 1); 612 sess->s_ttyvp = NULL; 613 sess->s_ttydp = NULL; 614 sess->s_ttyp = NULL; 615 bcopy(p->p_session->s_login, sess->s_login, 616 sizeof(sess->s_login)); 617 pgrp->pg_session = sess; 618 KASSERT(p == curproc, 619 ("enterpgrp: mksession and p != curproc")); 620 } else { 621 pgrp->pg_session = p->p_session; 622 sess_hold(pgrp->pg_session); 623 PGRP_LOCK(pgrp); 624 } 625 pgrp->pg_id = pgid; 626 proc_id_set(PROC_ID_GROUP, p->p_pid); 627 LIST_INIT(&pgrp->pg_members); 628 pgrp->pg_flags = 0; 629 630 /* 631 * As we have an exclusive lock of proctree_lock, 632 * this should not deadlock. 633 */ 634 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); 635 SLIST_INIT(&pgrp->pg_sigiolst); 636 PGRP_UNLOCK(pgrp); 637 638 doenterpgrp(p, pgrp); 639 640 sx_xunlock(&old_pgrp->pg_killsx); 641 return (0); 642 } 643 644 /* 645 * Move p to an existing process group 646 */ 647 int 648 enterthispgrp(struct proc *p, struct pgrp *pgrp) 649 { 650 struct pgrp *old_pgrp; 651 652 sx_assert(&proctree_lock, SX_XLOCKED); 653 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 654 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 655 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 656 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 657 KASSERT(pgrp->pg_session == p->p_session, 658 ("%s: pgrp's session %p, p->p_session %p proc %p\n", 659 __func__, pgrp->pg_session, p->p_session, p)); 660 KASSERT(pgrp != p->p_pgrp, 661 ("%s: p %p belongs to pgrp %p", __func__, p, pgrp)); 662 663 old_pgrp = p->p_pgrp; 664 if (!sx_try_xlock(&old_pgrp->pg_killsx)) { 665 sx_xunlock(&proctree_lock); 666 sx_xlock(&old_pgrp->pg_killsx); 667 sx_xunlock(&old_pgrp->pg_killsx); 668 return (ERESTART); 669 } 670 MPASS(old_pgrp == p->p_pgrp); 671 if (!sx_try_xlock(&pgrp->pg_killsx)) { 672 sx_xunlock(&old_pgrp->pg_killsx); 673 sx_xunlock(&proctree_lock); 674 sx_xlock(&pgrp->pg_killsx); 675 sx_xunlock(&pgrp->pg_killsx); 676 return (ERESTART); 677 } 678 679 doenterpgrp(p, pgrp); 680 681 sx_xunlock(&pgrp->pg_killsx); 682 sx_xunlock(&old_pgrp->pg_killsx); 683 return (0); 684 } 685 686 /* 687 * If true, any child of q which belongs to group pgrp, qualifies the 688 * process group pgrp as not orphaned. 689 */ 690 static bool 691 isjobproc(struct proc *q, struct pgrp *pgrp) 692 { 693 sx_assert(&proctree_lock, SX_LOCKED); 694 695 return (q->p_pgrp != pgrp && 696 q->p_pgrp->pg_session == pgrp->pg_session); 697 } 698 699 static struct proc * 700 jobc_reaper(struct proc *p) 701 { 702 struct proc *pp; 703 704 sx_assert(&proctree_lock, SA_LOCKED); 705 706 for (pp = p;;) { 707 pp = pp->p_reaper; 708 if (pp->p_reaper == pp || 709 (pp->p_treeflag & P_TREE_GRPEXITED) == 0) 710 return (pp); 711 } 712 } 713 714 static struct proc * 715 jobc_parent(struct proc *p, struct proc *p_exiting) 716 { 717 struct proc *pp; 718 719 sx_assert(&proctree_lock, SA_LOCKED); 720 721 pp = proc_realparent(p); 722 if (pp->p_pptr == NULL || pp == p_exiting || 723 (pp->p_treeflag & P_TREE_GRPEXITED) == 0) 724 return (pp); 725 return (jobc_reaper(pp)); 726 } 727 728 int 729 pgrp_calc_jobc(struct pgrp *pgrp) 730 { 731 struct proc *q; 732 int cnt; 733 734 #ifdef INVARIANTS 735 if (!mtx_owned(&pgrp->pg_mtx)) 736 sx_assert(&proctree_lock, SA_LOCKED); 737 #endif 738 739 cnt = 0; 740 LIST_FOREACH(q, &pgrp->pg_members, p_pglist) { 741 if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 || 742 q->p_pptr == NULL) 743 continue; 744 if (isjobproc(jobc_parent(q, NULL), pgrp)) 745 cnt++; 746 } 747 return (cnt); 748 } 749 750 /* 751 * Move p to a process group 752 */ 753 static void 754 doenterpgrp(struct proc *p, struct pgrp *pgrp) 755 { 756 struct pgrp *savepgrp; 757 struct proc *pp; 758 759 sx_assert(&proctree_lock, SX_XLOCKED); 760 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 761 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 762 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); 763 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); 764 765 savepgrp = p->p_pgrp; 766 pp = jobc_parent(p, NULL); 767 768 PGRP_LOCK(pgrp); 769 PGRP_LOCK(savepgrp); 770 if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1) 771 orphanpg(savepgrp); 772 PROC_LOCK(p); 773 LIST_REMOVE(p, p_pglist); 774 p->p_pgrp = pgrp; 775 PROC_UNLOCK(p); 776 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 777 if (isjobproc(pp, pgrp)) 778 pgrp->pg_flags &= ~PGRP_ORPHANED; 779 PGRP_UNLOCK(savepgrp); 780 PGRP_UNLOCK(pgrp); 781 if (LIST_EMPTY(&savepgrp->pg_members)) 782 pgdelete(savepgrp); 783 } 784 785 /* 786 * remove process from process group 787 */ 788 int 789 leavepgrp(struct proc *p) 790 { 791 struct pgrp *savepgrp; 792 793 sx_assert(&proctree_lock, SX_XLOCKED); 794 savepgrp = p->p_pgrp; 795 PGRP_LOCK(savepgrp); 796 PROC_LOCK(p); 797 LIST_REMOVE(p, p_pglist); 798 p->p_pgrp = NULL; 799 PROC_UNLOCK(p); 800 PGRP_UNLOCK(savepgrp); 801 if (LIST_EMPTY(&savepgrp->pg_members)) 802 pgdelete(savepgrp); 803 return (0); 804 } 805 806 /* 807 * delete a process group 808 */ 809 static void 810 pgdelete(struct pgrp *pgrp) 811 { 812 struct session *savesess; 813 struct tty *tp; 814 815 sx_assert(&proctree_lock, SX_XLOCKED); 816 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 817 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 818 819 /* 820 * Reset any sigio structures pointing to us as a result of 821 * F_SETOWN with our pgid. The proctree lock ensures that 822 * new sigio structures will not be added after this point. 823 */ 824 funsetownlst(&pgrp->pg_sigiolst); 825 826 PGRP_LOCK(pgrp); 827 tp = pgrp->pg_session->s_ttyp; 828 LIST_REMOVE(pgrp, pg_hash); 829 savesess = pgrp->pg_session; 830 PGRP_UNLOCK(pgrp); 831 832 /* Remove the reference to the pgrp before deallocating it. */ 833 if (tp != NULL) { 834 tty_lock(tp); 835 tty_rel_pgrp(tp, pgrp); 836 } 837 838 proc_id_clear(PROC_ID_GROUP, pgrp->pg_id); 839 uma_zfree(pgrp_zone, pgrp); 840 sess_release(savesess); 841 } 842 843 844 static void 845 fixjobc_kill(struct proc *p) 846 { 847 struct proc *q; 848 struct pgrp *pgrp; 849 850 sx_assert(&proctree_lock, SX_LOCKED); 851 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 852 pgrp = p->p_pgrp; 853 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); 854 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); 855 856 /* 857 * p no longer affects process group orphanage for children. 858 * It is marked by the flag because p is only physically 859 * removed from its process group on wait(2). 860 */ 861 MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0); 862 p->p_treeflag |= P_TREE_GRPEXITED; 863 864 /* 865 * Check if exiting p orphans its own group. 866 */ 867 pgrp = p->p_pgrp; 868 if (isjobproc(jobc_parent(p, NULL), pgrp)) { 869 PGRP_LOCK(pgrp); 870 if (pgrp_calc_jobc(pgrp) == 0) 871 orphanpg(pgrp); 872 PGRP_UNLOCK(pgrp); 873 } 874 875 /* 876 * Check this process' children to see whether they qualify 877 * their process groups after reparenting to reaper. 878 */ 879 LIST_FOREACH(q, &p->p_children, p_sibling) { 880 pgrp = q->p_pgrp; 881 PGRP_LOCK(pgrp); 882 if (pgrp_calc_jobc(pgrp) == 0) { 883 /* 884 * We want to handle exactly the children that 885 * has p as realparent. Then, when calculating 886 * jobc_parent for children, we should ignore 887 * P_TREE_GRPEXITED flag already set on p. 888 */ 889 if (jobc_parent(q, p) == p && isjobproc(p, pgrp)) 890 orphanpg(pgrp); 891 } else 892 pgrp->pg_flags &= ~PGRP_ORPHANED; 893 PGRP_UNLOCK(pgrp); 894 } 895 LIST_FOREACH(q, &p->p_orphans, p_orphan) { 896 pgrp = q->p_pgrp; 897 PGRP_LOCK(pgrp); 898 if (pgrp_calc_jobc(pgrp) == 0) { 899 if (isjobproc(p, pgrp)) 900 orphanpg(pgrp); 901 } else 902 pgrp->pg_flags &= ~PGRP_ORPHANED; 903 PGRP_UNLOCK(pgrp); 904 } 905 } 906 907 void 908 killjobc(void) 909 { 910 struct session *sp; 911 struct tty *tp; 912 struct proc *p; 913 struct vnode *ttyvp; 914 915 p = curproc; 916 MPASS(p->p_flag & P_WEXIT); 917 sx_assert(&proctree_lock, SX_LOCKED); 918 919 if (SESS_LEADER(p)) { 920 sp = p->p_session; 921 922 /* 923 * s_ttyp is not zero'd; we use this to indicate that 924 * the session once had a controlling terminal. (for 925 * logging and informational purposes) 926 */ 927 SESS_LOCK(sp); 928 ttyvp = sp->s_ttyvp; 929 tp = sp->s_ttyp; 930 sp->s_ttyvp = NULL; 931 sp->s_ttydp = NULL; 932 sp->s_leader = NULL; 933 SESS_UNLOCK(sp); 934 935 /* 936 * Signal foreground pgrp and revoke access to 937 * controlling terminal if it has not been revoked 938 * already. 939 * 940 * Because the TTY may have been revoked in the mean 941 * time and could already have a new session associated 942 * with it, make sure we don't send a SIGHUP to a 943 * foreground process group that does not belong to this 944 * session. 945 */ 946 947 if (tp != NULL) { 948 tty_lock(tp); 949 if (tp->t_session == sp) 950 tty_signal_pgrp(tp, SIGHUP); 951 tty_unlock(tp); 952 } 953 954 if (ttyvp != NULL) { 955 sx_xunlock(&proctree_lock); 956 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) { 957 VOP_REVOKE(ttyvp, REVOKEALL); 958 VOP_UNLOCK(ttyvp); 959 } 960 devfs_ctty_unref(ttyvp); 961 sx_xlock(&proctree_lock); 962 } 963 } 964 fixjobc_kill(p); 965 } 966 967 /* 968 * A process group has become orphaned, mark it as such for signal 969 * delivery code. If there are any stopped processes in the group, 970 * hang-up all process in that group. 971 */ 972 static void 973 orphanpg(struct pgrp *pg) 974 { 975 struct proc *p; 976 977 PGRP_LOCK_ASSERT(pg, MA_OWNED); 978 979 pg->pg_flags |= PGRP_ORPHANED; 980 981 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 982 PROC_LOCK(p); 983 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) { 984 PROC_UNLOCK(p); 985 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 986 PROC_LOCK(p); 987 kern_psignal(p, SIGHUP); 988 kern_psignal(p, SIGCONT); 989 PROC_UNLOCK(p); 990 } 991 return; 992 } 993 PROC_UNLOCK(p); 994 } 995 } 996 997 void 998 sess_hold(struct session *s) 999 { 1000 1001 refcount_acquire(&s->s_count); 1002 } 1003 1004 void 1005 sess_release(struct session *s) 1006 { 1007 1008 if (refcount_release(&s->s_count)) { 1009 if (s->s_ttyp != NULL) { 1010 tty_lock(s->s_ttyp); 1011 tty_rel_sess(s->s_ttyp, s); 1012 } 1013 proc_id_clear(PROC_ID_SESSION, s->s_sid); 1014 mtx_destroy(&s->s_mtx); 1015 free(s, M_SESSION); 1016 } 1017 } 1018 1019 #ifdef DDB 1020 1021 static void 1022 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p) 1023 { 1024 db_printf( 1025 " pid %d at %p pr %d pgrp %p e %d jc %d\n", 1026 p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid, 1027 p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0, 1028 p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp)); 1029 } 1030 1031 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE) 1032 { 1033 struct pgrp *pgrp; 1034 struct proc *p; 1035 int i; 1036 1037 for (i = 0; i <= pgrphash; i++) { 1038 if (!LIST_EMPTY(&pgrphashtbl[i])) { 1039 db_printf("indx %d\n", i); 1040 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { 1041 db_printf( 1042 " pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n", 1043 pgrp, (int)pgrp->pg_id, pgrp->pg_session, 1044 pgrp->pg_session->s_count, 1045 LIST_FIRST(&pgrp->pg_members)); 1046 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) 1047 db_print_pgrp_one(pgrp, p); 1048 } 1049 } 1050 } 1051 } 1052 #endif /* DDB */ 1053 1054 /* 1055 * Calculate the kinfo_proc members which contain process-wide 1056 * informations. 1057 * Must be called with the target process locked. 1058 */ 1059 static void 1060 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp) 1061 { 1062 struct thread *td; 1063 1064 PROC_LOCK_ASSERT(p, MA_OWNED); 1065 1066 kp->ki_estcpu = 0; 1067 kp->ki_pctcpu = 0; 1068 FOREACH_THREAD_IN_PROC(p, td) { 1069 thread_lock(td); 1070 kp->ki_pctcpu += sched_pctcpu(td); 1071 kp->ki_estcpu += sched_estcpu(td); 1072 thread_unlock(td); 1073 } 1074 } 1075 1076 /* 1077 * Fill in any information that is common to all threads in the process. 1078 * Must be called with the target process locked. 1079 */ 1080 static void 1081 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp) 1082 { 1083 struct thread *td0; 1084 struct ucred *cred; 1085 struct sigacts *ps; 1086 struct timeval boottime; 1087 1088 PROC_LOCK_ASSERT(p, MA_OWNED); 1089 1090 kp->ki_structsize = sizeof(*kp); 1091 kp->ki_paddr = p; 1092 kp->ki_addr =/* p->p_addr; */0; /* XXX */ 1093 kp->ki_args = p->p_args; 1094 kp->ki_textvp = p->p_textvp; 1095 #ifdef KTRACE 1096 kp->ki_tracep = ktr_get_tracevp(p, false); 1097 kp->ki_traceflag = p->p_traceflag; 1098 #endif 1099 kp->ki_fd = p->p_fd; 1100 kp->ki_pd = p->p_pd; 1101 kp->ki_vmspace = p->p_vmspace; 1102 kp->ki_flag = p->p_flag; 1103 kp->ki_flag2 = p->p_flag2; 1104 cred = p->p_ucred; 1105 if (cred) { 1106 kp->ki_uid = cred->cr_uid; 1107 kp->ki_ruid = cred->cr_ruid; 1108 kp->ki_svuid = cred->cr_svuid; 1109 kp->ki_cr_flags = 0; 1110 if (cred->cr_flags & CRED_FLAG_CAPMODE) 1111 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE; 1112 /* XXX bde doesn't like KI_NGROUPS */ 1113 if (1 + cred->cr_ngroups > KI_NGROUPS) { 1114 kp->ki_ngroups = KI_NGROUPS; 1115 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW; 1116 } else 1117 kp->ki_ngroups = 1 + cred->cr_ngroups; 1118 kp->ki_groups[0] = cred->cr_gid; 1119 bcopy(cred->cr_groups, kp->ki_groups + 1, 1120 (kp->ki_ngroups - 1) * sizeof(gid_t)); 1121 kp->ki_rgid = cred->cr_rgid; 1122 kp->ki_svgid = cred->cr_svgid; 1123 /* If jailed(cred), emulate the old P_JAILED flag. */ 1124 if (jailed(cred)) { 1125 kp->ki_flag |= P_JAILED; 1126 /* If inside the jail, use 0 as a jail ID. */ 1127 if (cred->cr_prison != curthread->td_ucred->cr_prison) 1128 kp->ki_jid = cred->cr_prison->pr_id; 1129 } 1130 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name, 1131 sizeof(kp->ki_loginclass)); 1132 } 1133 ps = p->p_sigacts; 1134 if (ps) { 1135 mtx_lock(&ps->ps_mtx); 1136 kp->ki_sigignore = ps->ps_sigignore; 1137 kp->ki_sigcatch = ps->ps_sigcatch; 1138 mtx_unlock(&ps->ps_mtx); 1139 } 1140 if (p->p_state != PRS_NEW && 1141 p->p_state != PRS_ZOMBIE && 1142 p->p_vmspace != NULL) { 1143 struct vmspace *vm = p->p_vmspace; 1144 1145 kp->ki_size = vm->vm_map.size; 1146 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ 1147 FOREACH_THREAD_IN_PROC(p, td0) 1148 kp->ki_rssize += td0->td_kstack_pages; 1149 kp->ki_swrss = vm->vm_swrss; 1150 kp->ki_tsize = vm->vm_tsize; 1151 kp->ki_dsize = vm->vm_dsize; 1152 kp->ki_ssize = vm->vm_ssize; 1153 } else if (p->p_state == PRS_ZOMBIE) 1154 kp->ki_stat = SZOMB; 1155 kp->ki_sflag = PS_INMEM; 1156 /* Calculate legacy swtime as seconds since 'swtick'. */ 1157 kp->ki_swtime = (ticks - p->p_swtick) / hz; 1158 kp->ki_pid = p->p_pid; 1159 kp->ki_nice = p->p_nice; 1160 kp->ki_fibnum = p->p_fibnum; 1161 kp->ki_start = p->p_stats->p_start; 1162 getboottime(&boottime); 1163 timevaladd(&kp->ki_start, &boottime); 1164 PROC_STATLOCK(p); 1165 rufetch(p, &kp->ki_rusage); 1166 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime); 1167 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime); 1168 PROC_STATUNLOCK(p); 1169 calccru(p, &kp->ki_childutime, &kp->ki_childstime); 1170 /* Some callers want child times in a single value. */ 1171 kp->ki_childtime = kp->ki_childstime; 1172 timevaladd(&kp->ki_childtime, &kp->ki_childutime); 1173 1174 FOREACH_THREAD_IN_PROC(p, td0) 1175 kp->ki_cow += td0->td_cow; 1176 1177 if (p->p_comm[0] != '\0') 1178 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm)); 1179 if (p->p_sysent && p->p_sysent->sv_name != NULL && 1180 p->p_sysent->sv_name[0] != '\0') 1181 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul)); 1182 kp->ki_siglist = p->p_siglist; 1183 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig); 1184 kp->ki_acflag = p->p_acflag; 1185 kp->ki_lock = p->p_lock; 1186 if (p->p_pptr) { 1187 kp->ki_ppid = p->p_oppid; 1188 if (p->p_flag & P_TRACED) 1189 kp->ki_tracer = p->p_pptr->p_pid; 1190 } 1191 } 1192 1193 /* 1194 * Fill job-related process information. 1195 */ 1196 static void 1197 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp) 1198 { 1199 struct tty *tp; 1200 struct session *sp; 1201 struct pgrp *pgrp; 1202 1203 sx_assert(&proctree_lock, SA_LOCKED); 1204 PROC_LOCK_ASSERT(p, MA_OWNED); 1205 1206 pgrp = p->p_pgrp; 1207 if (pgrp == NULL) 1208 return; 1209 1210 kp->ki_pgid = pgrp->pg_id; 1211 kp->ki_jobc = pgrp_calc_jobc(pgrp); 1212 1213 sp = pgrp->pg_session; 1214 tp = NULL; 1215 1216 if (sp != NULL) { 1217 kp->ki_sid = sp->s_sid; 1218 SESS_LOCK(sp); 1219 strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login)); 1220 if (sp->s_ttyvp) 1221 kp->ki_kiflag |= KI_CTTY; 1222 if (SESS_LEADER(p)) 1223 kp->ki_kiflag |= KI_SLEADER; 1224 tp = sp->s_ttyp; 1225 SESS_UNLOCK(sp); 1226 } 1227 1228 if ((p->p_flag & P_CONTROLT) && tp != NULL) { 1229 kp->ki_tdev = tty_udev(tp); 1230 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 1231 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; 1232 if (tp->t_session) 1233 kp->ki_tsid = tp->t_session->s_sid; 1234 } else { 1235 kp->ki_tdev = NODEV; 1236 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */ 1237 } 1238 } 1239 1240 /* 1241 * Fill in information that is thread specific. Must be called with 1242 * target process locked. If 'preferthread' is set, overwrite certain 1243 * process-related fields that are maintained for both threads and 1244 * processes. 1245 */ 1246 static void 1247 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread) 1248 { 1249 struct proc *p; 1250 1251 p = td->td_proc; 1252 kp->ki_tdaddr = td; 1253 PROC_LOCK_ASSERT(p, MA_OWNED); 1254 1255 if (preferthread) 1256 PROC_STATLOCK(p); 1257 thread_lock(td); 1258 if (td->td_wmesg != NULL) 1259 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg)); 1260 else 1261 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg)); 1262 if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >= 1263 sizeof(kp->ki_tdname)) { 1264 strlcpy(kp->ki_moretdname, 1265 td->td_name + sizeof(kp->ki_tdname) - 1, 1266 sizeof(kp->ki_moretdname)); 1267 } else { 1268 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname)); 1269 } 1270 if (TD_ON_LOCK(td)) { 1271 kp->ki_kiflag |= KI_LOCKBLOCK; 1272 strlcpy(kp->ki_lockname, td->td_lockname, 1273 sizeof(kp->ki_lockname)); 1274 } else { 1275 kp->ki_kiflag &= ~KI_LOCKBLOCK; 1276 bzero(kp->ki_lockname, sizeof(kp->ki_lockname)); 1277 } 1278 1279 if (p->p_state == PRS_NORMAL) { /* approximate. */ 1280 if (TD_ON_RUNQ(td) || 1281 TD_CAN_RUN(td) || 1282 TD_IS_RUNNING(td)) { 1283 kp->ki_stat = SRUN; 1284 } else if (P_SHOULDSTOP(p)) { 1285 kp->ki_stat = SSTOP; 1286 } else if (TD_IS_SLEEPING(td)) { 1287 kp->ki_stat = SSLEEP; 1288 } else if (TD_ON_LOCK(td)) { 1289 kp->ki_stat = SLOCK; 1290 } else { 1291 kp->ki_stat = SWAIT; 1292 } 1293 } else if (p->p_state == PRS_ZOMBIE) { 1294 kp->ki_stat = SZOMB; 1295 } else { 1296 kp->ki_stat = SIDL; 1297 } 1298 1299 /* Things in the thread */ 1300 kp->ki_wchan = td->td_wchan; 1301 kp->ki_pri.pri_level = td->td_priority; 1302 kp->ki_pri.pri_native = td->td_base_pri; 1303 1304 /* 1305 * Note: legacy fields; clamp at the old NOCPU value and/or 1306 * the maximum u_char CPU value. 1307 */ 1308 if (td->td_lastcpu == NOCPU) 1309 kp->ki_lastcpu_old = NOCPU_OLD; 1310 else if (td->td_lastcpu > MAXCPU_OLD) 1311 kp->ki_lastcpu_old = MAXCPU_OLD; 1312 else 1313 kp->ki_lastcpu_old = td->td_lastcpu; 1314 1315 if (td->td_oncpu == NOCPU) 1316 kp->ki_oncpu_old = NOCPU_OLD; 1317 else if (td->td_oncpu > MAXCPU_OLD) 1318 kp->ki_oncpu_old = MAXCPU_OLD; 1319 else 1320 kp->ki_oncpu_old = td->td_oncpu; 1321 1322 kp->ki_lastcpu = td->td_lastcpu; 1323 kp->ki_oncpu = td->td_oncpu; 1324 kp->ki_tdflags = td->td_flags; 1325 kp->ki_tid = td->td_tid; 1326 kp->ki_numthreads = p->p_numthreads; 1327 kp->ki_pcb = td->td_pcb; 1328 kp->ki_kstack = (void *)td->td_kstack; 1329 kp->ki_slptime = (ticks - td->td_slptick) / hz; 1330 kp->ki_pri.pri_class = td->td_pri_class; 1331 kp->ki_pri.pri_user = td->td_user_pri; 1332 1333 if (preferthread) { 1334 rufetchtd(td, &kp->ki_rusage); 1335 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime); 1336 kp->ki_pctcpu = sched_pctcpu(td); 1337 kp->ki_estcpu = sched_estcpu(td); 1338 kp->ki_cow = td->td_cow; 1339 } 1340 1341 /* We can't get this anymore but ps etc never used it anyway. */ 1342 kp->ki_rqindex = 0; 1343 1344 if (preferthread) 1345 kp->ki_siglist = td->td_siglist; 1346 kp->ki_sigmask = td->td_sigmask; 1347 thread_unlock(td); 1348 if (preferthread) 1349 PROC_STATUNLOCK(p); 1350 1351 if ((td->td_pflags & TDP2_UEXTERR) != 0) 1352 kp->ki_uerrmsg = td->td_exterr_ptr; 1353 } 1354 1355 /* 1356 * Fill in a kinfo_proc structure for the specified process. 1357 * Must be called with the target process locked. 1358 */ 1359 void 1360 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp) 1361 { 1362 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1363 1364 bzero(kp, sizeof(*kp)); 1365 1366 fill_kinfo_proc_pgrp(p,kp); 1367 fill_kinfo_proc_only(p, kp); 1368 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0); 1369 fill_kinfo_aggregate(p, kp); 1370 } 1371 1372 struct pstats * 1373 pstats_alloc(void) 1374 { 1375 1376 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK)); 1377 } 1378 1379 /* 1380 * Copy parts of p_stats; zero the rest of p_stats (statistics). 1381 */ 1382 void 1383 pstats_fork(struct pstats *src, struct pstats *dst) 1384 { 1385 1386 bzero(&dst->pstat_startzero, 1387 __rangeof(struct pstats, pstat_startzero, pstat_endzero)); 1388 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy, 1389 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy)); 1390 } 1391 1392 void 1393 pstats_free(struct pstats *ps) 1394 { 1395 1396 free(ps, M_SUBPROC); 1397 } 1398 1399 #ifdef COMPAT_FREEBSD32 1400 1401 /* 1402 * This function is typically used to copy out the kernel address, so 1403 * it can be replaced by assignment of zero. 1404 */ 1405 static inline uint32_t 1406 ptr32_trim(const void *ptr) 1407 { 1408 uintptr_t uptr; 1409 1410 uptr = (uintptr_t)ptr; 1411 return ((uptr > UINT_MAX) ? 0 : uptr); 1412 } 1413 1414 #define PTRTRIM_CP(src,dst,fld) \ 1415 do { (dst).fld = ptr32_trim((src).fld); } while (0) 1416 1417 static void 1418 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32) 1419 { 1420 int i; 1421 1422 bzero(ki32, sizeof(struct kinfo_proc32)); 1423 ki32->ki_structsize = sizeof(struct kinfo_proc32); 1424 CP(*ki, *ki32, ki_layout); 1425 PTRTRIM_CP(*ki, *ki32, ki_args); 1426 PTRTRIM_CP(*ki, *ki32, ki_paddr); 1427 PTRTRIM_CP(*ki, *ki32, ki_addr); 1428 PTRTRIM_CP(*ki, *ki32, ki_tracep); 1429 PTRTRIM_CP(*ki, *ki32, ki_textvp); 1430 PTRTRIM_CP(*ki, *ki32, ki_fd); 1431 PTRTRIM_CP(*ki, *ki32, ki_vmspace); 1432 PTRTRIM_CP(*ki, *ki32, ki_wchan); 1433 CP(*ki, *ki32, ki_pid); 1434 CP(*ki, *ki32, ki_ppid); 1435 CP(*ki, *ki32, ki_pgid); 1436 CP(*ki, *ki32, ki_tpgid); 1437 CP(*ki, *ki32, ki_sid); 1438 CP(*ki, *ki32, ki_tsid); 1439 CP(*ki, *ki32, ki_jobc); 1440 CP(*ki, *ki32, ki_tdev); 1441 CP(*ki, *ki32, ki_tdev_freebsd11); 1442 CP(*ki, *ki32, ki_siglist); 1443 CP(*ki, *ki32, ki_sigmask); 1444 CP(*ki, *ki32, ki_sigignore); 1445 CP(*ki, *ki32, ki_sigcatch); 1446 CP(*ki, *ki32, ki_uid); 1447 CP(*ki, *ki32, ki_ruid); 1448 CP(*ki, *ki32, ki_svuid); 1449 CP(*ki, *ki32, ki_rgid); 1450 CP(*ki, *ki32, ki_svgid); 1451 CP(*ki, *ki32, ki_ngroups); 1452 for (i = 0; i < KI_NGROUPS; i++) 1453 CP(*ki, *ki32, ki_groups[i]); 1454 CP(*ki, *ki32, ki_size); 1455 CP(*ki, *ki32, ki_rssize); 1456 CP(*ki, *ki32, ki_swrss); 1457 CP(*ki, *ki32, ki_tsize); 1458 CP(*ki, *ki32, ki_dsize); 1459 CP(*ki, *ki32, ki_ssize); 1460 CP(*ki, *ki32, ki_xstat); 1461 CP(*ki, *ki32, ki_acflag); 1462 CP(*ki, *ki32, ki_pctcpu); 1463 CP(*ki, *ki32, ki_estcpu); 1464 CP(*ki, *ki32, ki_slptime); 1465 CP(*ki, *ki32, ki_swtime); 1466 CP(*ki, *ki32, ki_cow); 1467 CP(*ki, *ki32, ki_runtime); 1468 TV_CP(*ki, *ki32, ki_start); 1469 TV_CP(*ki, *ki32, ki_childtime); 1470 CP(*ki, *ki32, ki_flag); 1471 CP(*ki, *ki32, ki_kiflag); 1472 CP(*ki, *ki32, ki_traceflag); 1473 CP(*ki, *ki32, ki_stat); 1474 CP(*ki, *ki32, ki_nice); 1475 CP(*ki, *ki32, ki_lock); 1476 CP(*ki, *ki32, ki_rqindex); 1477 CP(*ki, *ki32, ki_oncpu); 1478 CP(*ki, *ki32, ki_lastcpu); 1479 1480 /* XXX TODO: wrap cpu value as appropriate */ 1481 CP(*ki, *ki32, ki_oncpu_old); 1482 CP(*ki, *ki32, ki_lastcpu_old); 1483 1484 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1); 1485 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1); 1486 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1); 1487 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1); 1488 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1); 1489 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1); 1490 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1); 1491 bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1); 1492 CP(*ki, *ki32, ki_tracer); 1493 CP(*ki, *ki32, ki_flag2); 1494 CP(*ki, *ki32, ki_fibnum); 1495 CP(*ki, *ki32, ki_cr_flags); 1496 CP(*ki, *ki32, ki_jid); 1497 CP(*ki, *ki32, ki_numthreads); 1498 CP(*ki, *ki32, ki_tid); 1499 CP(*ki, *ki32, ki_pri); 1500 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage); 1501 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch); 1502 PTRTRIM_CP(*ki, *ki32, ki_pcb); 1503 PTRTRIM_CP(*ki, *ki32, ki_kstack); 1504 PTRTRIM_CP(*ki, *ki32, ki_udata); 1505 PTRTRIM_CP(*ki, *ki32, ki_tdaddr); 1506 CP(*ki, *ki32, ki_sflag); 1507 CP(*ki, *ki32, ki_tdflags); 1508 PTRTRIM_CP(*ki, *ki32, ki_uerrmsg); 1509 } 1510 #endif 1511 1512 static ssize_t 1513 kern_proc_out_size(struct proc *p, int flags) 1514 { 1515 ssize_t size = 0; 1516 1517 PROC_LOCK_ASSERT(p, MA_OWNED); 1518 1519 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1520 #ifdef COMPAT_FREEBSD32 1521 if ((flags & KERN_PROC_MASK32) != 0) { 1522 size += sizeof(struct kinfo_proc32); 1523 } else 1524 #endif 1525 size += sizeof(struct kinfo_proc); 1526 } else { 1527 #ifdef COMPAT_FREEBSD32 1528 if ((flags & KERN_PROC_MASK32) != 0) 1529 size += sizeof(struct kinfo_proc32) * p->p_numthreads; 1530 else 1531 #endif 1532 size += sizeof(struct kinfo_proc) * p->p_numthreads; 1533 } 1534 PROC_UNLOCK(p); 1535 return (size); 1536 } 1537 1538 int 1539 kern_proc_out(struct proc *p, struct sbuf *sb, int flags) 1540 { 1541 struct thread *td; 1542 struct kinfo_proc ki; 1543 #ifdef COMPAT_FREEBSD32 1544 struct kinfo_proc32 ki32; 1545 #endif 1546 int error; 1547 1548 PROC_LOCK_ASSERT(p, MA_OWNED); 1549 MPASS(FIRST_THREAD_IN_PROC(p) != NULL); 1550 1551 error = 0; 1552 fill_kinfo_proc(p, &ki); 1553 if ((flags & KERN_PROC_NOTHREADS) != 0) { 1554 #ifdef COMPAT_FREEBSD32 1555 if ((flags & KERN_PROC_MASK32) != 0) { 1556 freebsd32_kinfo_proc_out(&ki, &ki32); 1557 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1558 error = ENOMEM; 1559 } else 1560 #endif 1561 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1562 error = ENOMEM; 1563 } else { 1564 FOREACH_THREAD_IN_PROC(p, td) { 1565 fill_kinfo_thread(td, &ki, 1); 1566 #ifdef COMPAT_FREEBSD32 1567 if ((flags & KERN_PROC_MASK32) != 0) { 1568 freebsd32_kinfo_proc_out(&ki, &ki32); 1569 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) 1570 error = ENOMEM; 1571 } else 1572 #endif 1573 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) 1574 error = ENOMEM; 1575 if (error != 0) 1576 break; 1577 } 1578 } 1579 PROC_UNLOCK(p); 1580 return (error); 1581 } 1582 1583 static int 1584 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags) 1585 { 1586 struct sbuf sb; 1587 struct kinfo_proc ki; 1588 int error, error2; 1589 1590 if (req->oldptr == NULL) 1591 return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags))); 1592 1593 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); 1594 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 1595 error = kern_proc_out(p, &sb, flags); 1596 error2 = sbuf_finish(&sb); 1597 sbuf_delete(&sb); 1598 if (error != 0) 1599 return (error); 1600 else if (error2 != 0) 1601 return (error2); 1602 return (0); 1603 } 1604 1605 int 1606 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg) 1607 { 1608 struct proc *p; 1609 int error, i, j; 1610 1611 for (i = 0; i < pidhashlock + 1; i++) { 1612 sx_slock(&proctree_lock); 1613 sx_slock(&pidhashtbl_lock[i]); 1614 for (j = i; j <= pidhash; j += pidhashlock + 1) { 1615 LIST_FOREACH(p, &pidhashtbl[j], p_hash) { 1616 if (p->p_state == PRS_NEW) 1617 continue; 1618 error = cb(p, cbarg); 1619 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 1620 if (error != 0) { 1621 sx_sunlock(&pidhashtbl_lock[i]); 1622 sx_sunlock(&proctree_lock); 1623 return (error); 1624 } 1625 } 1626 } 1627 sx_sunlock(&pidhashtbl_lock[i]); 1628 sx_sunlock(&proctree_lock); 1629 } 1630 return (0); 1631 } 1632 1633 struct kern_proc_out_args { 1634 struct sysctl_req *req; 1635 int flags; 1636 int oid_number; 1637 int *name; 1638 }; 1639 1640 static int 1641 sysctl_kern_proc_iterate(struct proc *p, void *origarg) 1642 { 1643 struct kern_proc_out_args *arg = origarg; 1644 int *name = arg->name; 1645 int oid_number = arg->oid_number; 1646 int flags = arg->flags; 1647 struct sysctl_req *req = arg->req; 1648 int error = 0; 1649 1650 PROC_LOCK(p); 1651 1652 KASSERT(p->p_ucred != NULL, 1653 ("process credential is NULL for non-NEW proc")); 1654 /* 1655 * Show a user only appropriate processes. 1656 */ 1657 if (p_cansee(curthread, p)) 1658 goto skip; 1659 /* 1660 * TODO - make more efficient (see notes below). 1661 * do by session. 1662 */ 1663 switch (oid_number) { 1664 case KERN_PROC_GID: 1665 if (p->p_ucred->cr_gid != (gid_t)name[0]) 1666 goto skip; 1667 break; 1668 1669 case KERN_PROC_PGRP: 1670 /* could do this by traversing pgrp */ 1671 if (p->p_pgrp == NULL || 1672 p->p_pgrp->pg_id != (pid_t)name[0]) 1673 goto skip; 1674 break; 1675 1676 case KERN_PROC_RGID: 1677 if (p->p_ucred->cr_rgid != (gid_t)name[0]) 1678 goto skip; 1679 break; 1680 1681 case KERN_PROC_SESSION: 1682 if (p->p_session == NULL || 1683 p->p_session->s_sid != (pid_t)name[0]) 1684 goto skip; 1685 break; 1686 1687 case KERN_PROC_TTY: 1688 if ((p->p_flag & P_CONTROLT) == 0 || 1689 p->p_session == NULL) 1690 goto skip; 1691 /* XXX proctree_lock */ 1692 SESS_LOCK(p->p_session); 1693 if (p->p_session->s_ttyp == NULL || 1694 tty_udev(p->p_session->s_ttyp) != 1695 (dev_t)name[0]) { 1696 SESS_UNLOCK(p->p_session); 1697 goto skip; 1698 } 1699 SESS_UNLOCK(p->p_session); 1700 break; 1701 1702 case KERN_PROC_UID: 1703 if (p->p_ucred->cr_uid != (uid_t)name[0]) 1704 goto skip; 1705 break; 1706 1707 case KERN_PROC_RUID: 1708 if (p->p_ucred->cr_ruid != (uid_t)name[0]) 1709 goto skip; 1710 break; 1711 1712 case KERN_PROC_PROC: 1713 break; 1714 1715 default: 1716 break; 1717 } 1718 error = sysctl_out_proc(p, req, flags); 1719 PROC_LOCK_ASSERT(p, MA_NOTOWNED); 1720 return (error); 1721 skip: 1722 PROC_UNLOCK(p); 1723 return (0); 1724 } 1725 1726 static int 1727 sysctl_kern_proc(SYSCTL_HANDLER_ARGS) 1728 { 1729 struct kern_proc_out_args iterarg; 1730 int *name = (int *)arg1; 1731 u_int namelen = arg2; 1732 struct proc *p; 1733 int flags, oid_number; 1734 int error = 0; 1735 1736 oid_number = oidp->oid_number; 1737 if (oid_number != KERN_PROC_ALL && 1738 (oid_number & KERN_PROC_INC_THREAD) == 0) 1739 flags = KERN_PROC_NOTHREADS; 1740 else { 1741 flags = 0; 1742 oid_number &= ~KERN_PROC_INC_THREAD; 1743 } 1744 #ifdef COMPAT_FREEBSD32 1745 if (req->flags & SCTL_MASK32) 1746 flags |= KERN_PROC_MASK32; 1747 #endif 1748 if (oid_number == KERN_PROC_PID) { 1749 if (namelen != 1) 1750 return (EINVAL); 1751 error = sysctl_wire_old_buffer(req, 0); 1752 if (error) 1753 return (error); 1754 sx_slock(&proctree_lock); 1755 error = pget((pid_t)name[0], PGET_CANSEE, &p); 1756 if (error == 0) 1757 error = sysctl_out_proc(p, req, flags); 1758 sx_sunlock(&proctree_lock); 1759 return (error); 1760 } 1761 1762 switch (oid_number) { 1763 case KERN_PROC_ALL: 1764 if (namelen != 0) 1765 return (EINVAL); 1766 break; 1767 case KERN_PROC_PROC: 1768 if (namelen != 0 && namelen != 1) 1769 return (EINVAL); 1770 break; 1771 default: 1772 if (namelen != 1) 1773 return (EINVAL); 1774 break; 1775 } 1776 1777 if (req->oldptr == NULL) { 1778 /* overestimate by 5 procs */ 1779 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); 1780 if (error) 1781 return (error); 1782 } else { 1783 error = sysctl_wire_old_buffer(req, 0); 1784 if (error != 0) 1785 return (error); 1786 } 1787 iterarg.flags = flags; 1788 iterarg.oid_number = oid_number; 1789 iterarg.req = req; 1790 iterarg.name = name; 1791 error = proc_iterate(sysctl_kern_proc_iterate, &iterarg); 1792 return (error); 1793 } 1794 1795 struct pargs * 1796 pargs_alloc(int len) 1797 { 1798 struct pargs *pa; 1799 1800 pa = malloc(sizeof(struct pargs) + len, M_PARGS, 1801 M_WAITOK); 1802 refcount_init(&pa->ar_ref, 1); 1803 pa->ar_length = len; 1804 return (pa); 1805 } 1806 1807 static void 1808 pargs_free(struct pargs *pa) 1809 { 1810 1811 free(pa, M_PARGS); 1812 } 1813 1814 void 1815 pargs_hold(struct pargs *pa) 1816 { 1817 1818 if (pa == NULL) 1819 return; 1820 refcount_acquire(&pa->ar_ref); 1821 } 1822 1823 void 1824 pargs_drop(struct pargs *pa) 1825 { 1826 1827 if (pa == NULL) 1828 return; 1829 if (refcount_release(&pa->ar_ref)) 1830 pargs_free(pa); 1831 } 1832 1833 static int 1834 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, 1835 size_t len) 1836 { 1837 ssize_t n; 1838 1839 /* 1840 * This may return a short read if the string is shorter than the chunk 1841 * and is aligned at the end of the page, and the following page is not 1842 * mapped. 1843 */ 1844 n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len); 1845 if (n <= 0) 1846 return (ENOMEM); 1847 return (0); 1848 } 1849 1850 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ 1851 1852 enum proc_vector_type { 1853 PROC_ARG, 1854 PROC_ENV, 1855 PROC_AUX, 1856 }; 1857 1858 #ifdef COMPAT_FREEBSD32 1859 static int 1860 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, 1861 size_t *vsizep, enum proc_vector_type type) 1862 { 1863 struct freebsd32_ps_strings pss; 1864 Elf32_Auxinfo aux; 1865 vm_offset_t vptr, ptr; 1866 uint32_t *proc_vector32; 1867 char **proc_vector; 1868 size_t vsize, size; 1869 int i, error; 1870 1871 error = 0; 1872 if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) != 1873 sizeof(pss)) 1874 return (ENOMEM); 1875 switch (type) { 1876 case PROC_ARG: 1877 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); 1878 vsize = pss.ps_nargvstr; 1879 if (vsize > ARG_MAX) 1880 return (ENOEXEC); 1881 size = vsize * sizeof(int32_t); 1882 break; 1883 case PROC_ENV: 1884 vptr = (vm_offset_t)PTRIN(pss.ps_envstr); 1885 vsize = pss.ps_nenvstr; 1886 if (vsize > ARG_MAX) 1887 return (ENOEXEC); 1888 size = vsize * sizeof(int32_t); 1889 break; 1890 case PROC_AUX: 1891 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + 1892 (pss.ps_nenvstr + 1) * sizeof(int32_t); 1893 if (vptr % 4 != 0) 1894 return (ENOEXEC); 1895 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1896 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1897 sizeof(aux)) 1898 return (ENOMEM); 1899 if (aux.a_type == AT_NULL) 1900 break; 1901 ptr += sizeof(aux); 1902 } 1903 if (aux.a_type != AT_NULL) 1904 return (ENOEXEC); 1905 vsize = i + 1; 1906 size = vsize * sizeof(aux); 1907 break; 1908 default: 1909 KASSERT(0, ("Wrong proc vector type: %d", type)); 1910 return (EINVAL); 1911 } 1912 proc_vector32 = malloc(size, M_TEMP, M_WAITOK); 1913 if (proc_readmem(td, p, vptr, proc_vector32, size) != size) { 1914 error = ENOMEM; 1915 goto done; 1916 } 1917 if (type == PROC_AUX) { 1918 *proc_vectorp = (char **)proc_vector32; 1919 *vsizep = vsize; 1920 return (0); 1921 } 1922 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); 1923 for (i = 0; i < (int)vsize; i++) 1924 proc_vector[i] = PTRIN(proc_vector32[i]); 1925 *proc_vectorp = proc_vector; 1926 *vsizep = vsize; 1927 done: 1928 free(proc_vector32, M_TEMP); 1929 return (error); 1930 } 1931 #endif 1932 1933 static int 1934 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, 1935 size_t *vsizep, enum proc_vector_type type) 1936 { 1937 struct ps_strings pss; 1938 Elf_Auxinfo aux; 1939 vm_offset_t vptr, ptr; 1940 char **proc_vector; 1941 size_t vsize, size; 1942 int i; 1943 1944 #ifdef COMPAT_FREEBSD32 1945 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 1946 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); 1947 #endif 1948 if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) != 1949 sizeof(pss)) 1950 return (ENOMEM); 1951 switch (type) { 1952 case PROC_ARG: 1953 vptr = (vm_offset_t)pss.ps_argvstr; 1954 vsize = pss.ps_nargvstr; 1955 if (vsize > ARG_MAX) 1956 return (ENOEXEC); 1957 size = vsize * sizeof(char *); 1958 break; 1959 case PROC_ENV: 1960 vptr = (vm_offset_t)pss.ps_envstr; 1961 vsize = pss.ps_nenvstr; 1962 if (vsize > ARG_MAX) 1963 return (ENOEXEC); 1964 size = vsize * sizeof(char *); 1965 break; 1966 case PROC_AUX: 1967 /* 1968 * The aux array is just above env array on the stack. Check 1969 * that the address is naturally aligned. 1970 */ 1971 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) 1972 * sizeof(char *); 1973 #if __ELF_WORD_SIZE == 64 1974 if (vptr % sizeof(uint64_t) != 0) 1975 #else 1976 if (vptr % sizeof(uint32_t) != 0) 1977 #endif 1978 return (ENOEXEC); 1979 /* 1980 * We count the array size reading the aux vectors from the 1981 * stack until AT_NULL vector is returned. So (to keep the code 1982 * simple) we read the process stack twice: the first time here 1983 * to find the size and the second time when copying the vectors 1984 * to the allocated proc_vector. 1985 */ 1986 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { 1987 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) != 1988 sizeof(aux)) 1989 return (ENOMEM); 1990 if (aux.a_type == AT_NULL) 1991 break; 1992 ptr += sizeof(aux); 1993 } 1994 /* 1995 * If the PROC_AUXV_MAX entries are iterated over, and we have 1996 * not reached AT_NULL, it is most likely we are reading wrong 1997 * data: either the process doesn't have auxv array or data has 1998 * been modified. Return the error in this case. 1999 */ 2000 if (aux.a_type != AT_NULL) 2001 return (ENOEXEC); 2002 vsize = i + 1; 2003 size = vsize * sizeof(aux); 2004 break; 2005 default: 2006 KASSERT(0, ("Wrong proc vector type: %d", type)); 2007 return (EINVAL); /* In case we are built without INVARIANTS. */ 2008 } 2009 proc_vector = malloc(size, M_TEMP, M_WAITOK); 2010 if (proc_readmem(td, p, vptr, proc_vector, size) != size) { 2011 free(proc_vector, M_TEMP); 2012 return (ENOMEM); 2013 } 2014 *proc_vectorp = proc_vector; 2015 *vsizep = vsize; 2016 2017 return (0); 2018 } 2019 2020 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ 2021 2022 static int 2023 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, 2024 enum proc_vector_type type) 2025 { 2026 size_t done, len, nchr, vsize; 2027 int error, i; 2028 char **proc_vector, *sptr; 2029 char pss_string[GET_PS_STRINGS_CHUNK_SZ]; 2030 2031 PROC_ASSERT_HELD(p); 2032 2033 /* 2034 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. 2035 */ 2036 nchr = 2 * (PATH_MAX + ARG_MAX); 2037 2038 error = get_proc_vector(td, p, &proc_vector, &vsize, type); 2039 if (error != 0) 2040 return (error); 2041 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { 2042 /* 2043 * The program may have scribbled into its argv array, e.g. to 2044 * remove some arguments. If that has happened, break out 2045 * before trying to read from NULL. 2046 */ 2047 if (proc_vector[i] == NULL) 2048 break; 2049 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { 2050 error = proc_read_string(td, p, sptr, pss_string, 2051 sizeof(pss_string)); 2052 if (error != 0) 2053 goto done; 2054 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); 2055 if (done + len >= nchr) 2056 len = nchr - done - 1; 2057 sbuf_bcat(sb, pss_string, len); 2058 if (len != GET_PS_STRINGS_CHUNK_SZ) 2059 break; 2060 done += GET_PS_STRINGS_CHUNK_SZ; 2061 } 2062 sbuf_bcat(sb, "", 1); 2063 done += len + 1; 2064 } 2065 done: 2066 free(proc_vector, M_TEMP); 2067 return (error); 2068 } 2069 2070 int 2071 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) 2072 { 2073 2074 return (get_ps_strings(curthread, p, sb, PROC_ARG)); 2075 } 2076 2077 int 2078 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) 2079 { 2080 2081 return (get_ps_strings(curthread, p, sb, PROC_ENV)); 2082 } 2083 2084 int 2085 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) 2086 { 2087 size_t vsize, size; 2088 char **auxv; 2089 int error; 2090 2091 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); 2092 if (error == 0) { 2093 #ifdef COMPAT_FREEBSD32 2094 if (SV_PROC_FLAG(p, SV_ILP32) != 0) 2095 size = vsize * sizeof(Elf32_Auxinfo); 2096 else 2097 #endif 2098 size = vsize * sizeof(Elf_Auxinfo); 2099 if (sbuf_bcat(sb, auxv, size) != 0) 2100 error = ENOMEM; 2101 free(auxv, M_TEMP); 2102 } 2103 return (error); 2104 } 2105 2106 /* 2107 * This sysctl allows a process to retrieve the argument list or process 2108 * title for another process without groping around in the address space 2109 * of the other process. It also allow a process to set its own "process 2110 * title to a string of its own choice. 2111 */ 2112 static int 2113 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) 2114 { 2115 int *name = (int *)arg1; 2116 u_int namelen = arg2; 2117 struct pargs *newpa, *pa; 2118 struct proc *p; 2119 struct sbuf sb; 2120 int flags, error = 0, error2; 2121 pid_t pid; 2122 2123 if (namelen != 1) 2124 return (EINVAL); 2125 2126 p = curproc; 2127 pid = (pid_t)name[0]; 2128 if (pid == -1) { 2129 pid = p->p_pid; 2130 } 2131 2132 /* 2133 * If the query is for this process and it is single-threaded, there 2134 * is nobody to modify pargs, thus we can just read. 2135 */ 2136 if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL && 2137 (pa = p->p_args) != NULL) 2138 return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length)); 2139 2140 flags = PGET_CANSEE; 2141 if (req->newptr != NULL) 2142 flags |= PGET_ISCURRENT; 2143 error = pget(pid, flags, &p); 2144 if (error) 2145 return (error); 2146 2147 pa = p->p_args; 2148 if (pa != NULL) { 2149 pargs_hold(pa); 2150 PROC_UNLOCK(p); 2151 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); 2152 pargs_drop(pa); 2153 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { 2154 _PHOLD(p); 2155 PROC_UNLOCK(p); 2156 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2157 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2158 error = proc_getargv(curthread, p, &sb); 2159 error2 = sbuf_finish(&sb); 2160 PRELE(p); 2161 sbuf_delete(&sb); 2162 if (error == 0 && error2 != 0) 2163 error = error2; 2164 } else { 2165 PROC_UNLOCK(p); 2166 } 2167 if (error != 0 || req->newptr == NULL) 2168 return (error); 2169 2170 if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs)) 2171 return (ENOMEM); 2172 2173 if (req->newlen == 0) { 2174 /* 2175 * Clear the argument pointer, so that we'll fetch arguments 2176 * with proc_getargv() until further notice. 2177 */ 2178 newpa = NULL; 2179 } else { 2180 newpa = pargs_alloc(req->newlen); 2181 error = SYSCTL_IN(req, newpa->ar_args, req->newlen); 2182 if (error != 0) { 2183 pargs_free(newpa); 2184 return (error); 2185 } 2186 } 2187 PROC_LOCK(p); 2188 pa = p->p_args; 2189 p->p_args = newpa; 2190 PROC_UNLOCK(p); 2191 pargs_drop(pa); 2192 return (0); 2193 } 2194 2195 /* 2196 * This sysctl allows a process to retrieve environment of another process. 2197 */ 2198 static int 2199 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) 2200 { 2201 int *name = (int *)arg1; 2202 u_int namelen = arg2; 2203 struct proc *p; 2204 struct sbuf sb; 2205 int error, error2; 2206 2207 if (namelen != 1) 2208 return (EINVAL); 2209 2210 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2211 if (error != 0) 2212 return (error); 2213 if ((p->p_flag & P_SYSTEM) != 0) { 2214 PRELE(p); 2215 return (0); 2216 } 2217 2218 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2219 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2220 error = proc_getenvv(curthread, p, &sb); 2221 error2 = sbuf_finish(&sb); 2222 PRELE(p); 2223 sbuf_delete(&sb); 2224 return (error != 0 ? error : error2); 2225 } 2226 2227 /* 2228 * This sysctl allows a process to retrieve ELF auxiliary vector of 2229 * another process. 2230 */ 2231 static int 2232 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) 2233 { 2234 int *name = (int *)arg1; 2235 u_int namelen = arg2; 2236 struct proc *p; 2237 struct sbuf sb; 2238 int error, error2; 2239 2240 if (namelen != 1) 2241 return (EINVAL); 2242 2243 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2244 if (error != 0) 2245 return (error); 2246 if ((p->p_flag & P_SYSTEM) != 0) { 2247 PRELE(p); 2248 return (0); 2249 } 2250 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); 2251 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2252 error = proc_getauxv(curthread, p, &sb); 2253 error2 = sbuf_finish(&sb); 2254 PRELE(p); 2255 sbuf_delete(&sb); 2256 return (error != 0 ? error : error2); 2257 } 2258 2259 /* 2260 * Look up the canonical executable path running in the specified process. 2261 * It tries to return the same hardlink name as was used for execve(2). 2262 * This allows the programs that modify their behavior based on their progname, 2263 * to operate correctly. 2264 * 2265 * Result is returned in retbuf, it must not be freed, similar to vn_fullpath() 2266 * calling conventions. 2267 * binname is a pointer to temporary string buffer of length MAXPATHLEN, 2268 * allocated and freed by caller. 2269 * freebuf should be freed by caller, from the M_TEMP malloc type. 2270 */ 2271 int 2272 proc_get_binpath(struct proc *p, char *binname, char **retbuf, 2273 char **freebuf) 2274 { 2275 struct nameidata nd; 2276 struct vnode *vp, *dvp; 2277 size_t freepath_size; 2278 int error; 2279 bool do_fullpath; 2280 2281 PROC_LOCK_ASSERT(p, MA_OWNED); 2282 2283 vp = p->p_textvp; 2284 if (vp == NULL) { 2285 PROC_UNLOCK(p); 2286 *retbuf = ""; 2287 *freebuf = NULL; 2288 return (0); 2289 } 2290 vref(vp); 2291 dvp = p->p_textdvp; 2292 if (dvp != NULL) 2293 vref(dvp); 2294 if (p->p_binname != NULL) 2295 strlcpy(binname, p->p_binname, MAXPATHLEN); 2296 PROC_UNLOCK(p); 2297 2298 do_fullpath = true; 2299 *freebuf = NULL; 2300 if (dvp != NULL && binname[0] != '\0') { 2301 freepath_size = MAXPATHLEN; 2302 if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname), 2303 retbuf, freebuf, &freepath_size) == 0) { 2304 /* 2305 * Recheck the looked up path. The binary 2306 * might have been renamed or replaced, in 2307 * which case we should not report old name. 2308 */ 2309 NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf); 2310 error = namei(&nd); 2311 if (error == 0) { 2312 if (nd.ni_vp == vp) 2313 do_fullpath = false; 2314 vrele(nd.ni_vp); 2315 NDFREE_PNBUF(&nd); 2316 } 2317 } 2318 } 2319 if (do_fullpath) { 2320 free(*freebuf, M_TEMP); 2321 *freebuf = NULL; 2322 error = vn_fullpath(vp, retbuf, freebuf); 2323 } 2324 vrele(vp); 2325 if (dvp != NULL) 2326 vrele(dvp); 2327 return (error); 2328 } 2329 2330 /* 2331 * This sysctl allows a process to retrieve the path of the executable for 2332 * itself or another process. 2333 */ 2334 static int 2335 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) 2336 { 2337 pid_t *pidp = (pid_t *)arg1; 2338 unsigned int arglen = arg2; 2339 struct proc *p; 2340 char *retbuf, *freebuf, *binname; 2341 int error; 2342 2343 if (arglen != 1) 2344 return (EINVAL); 2345 binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 2346 binname[0] = '\0'; 2347 if (*pidp == -1) { /* -1 means this process */ 2348 error = 0; 2349 p = req->td->td_proc; 2350 PROC_LOCK(p); 2351 } else { 2352 error = pget(*pidp, PGET_CANSEE, &p); 2353 } 2354 2355 if (error == 0) 2356 error = proc_get_binpath(p, binname, &retbuf, &freebuf); 2357 free(binname, M_TEMP); 2358 if (error != 0) 2359 return (error); 2360 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); 2361 free(freebuf, M_TEMP); 2362 return (error); 2363 } 2364 2365 static int 2366 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) 2367 { 2368 struct proc *p; 2369 char *sv_name; 2370 int *name; 2371 int namelen; 2372 int error; 2373 2374 namelen = arg2; 2375 if (namelen != 1) 2376 return (EINVAL); 2377 2378 name = (int *)arg1; 2379 error = pget((pid_t)name[0], PGET_CANSEE, &p); 2380 if (error != 0) 2381 return (error); 2382 sv_name = p->p_sysent->sv_name; 2383 PROC_UNLOCK(p); 2384 return (sysctl_handle_string(oidp, sv_name, 0, req)); 2385 } 2386 2387 #ifdef KINFO_OVMENTRY_SIZE 2388 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); 2389 #endif 2390 2391 #ifdef COMPAT_FREEBSD7 2392 static int 2393 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) 2394 { 2395 vm_map_entry_t entry, tmp_entry; 2396 unsigned int last_timestamp, namelen; 2397 char *fullpath, *freepath; 2398 struct kinfo_ovmentry *kve; 2399 struct vattr va; 2400 struct ucred *cred; 2401 int error, *name; 2402 struct vnode *vp; 2403 struct proc *p; 2404 vm_map_t map; 2405 struct vmspace *vm; 2406 2407 namelen = arg2; 2408 if (namelen != 1) 2409 return (EINVAL); 2410 2411 name = (int *)arg1; 2412 error = pget((pid_t)name[0], PGET_WANTREAD, &p); 2413 if (error != 0) 2414 return (error); 2415 vm = vmspace_acquire_ref(p); 2416 if (vm == NULL) { 2417 PRELE(p); 2418 return (ESRCH); 2419 } 2420 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); 2421 2422 map = &vm->vm_map; 2423 vm_map_lock_read(map); 2424 VM_MAP_ENTRY_FOREACH(entry, map) { 2425 vm_object_t obj, tobj, lobj; 2426 vm_offset_t addr; 2427 2428 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2429 continue; 2430 2431 bzero(kve, sizeof(*kve)); 2432 kve->kve_structsize = sizeof(*kve); 2433 2434 kve->kve_private_resident = 0; 2435 obj = entry->object.vm_object; 2436 if (obj != NULL) { 2437 VM_OBJECT_RLOCK(obj); 2438 if (obj->shadow_count == 1) 2439 kve->kve_private_resident = 2440 obj->resident_page_count; 2441 } 2442 kve->kve_resident = 0; 2443 addr = entry->start; 2444 while (addr < entry->end) { 2445 if (pmap_extract(map->pmap, addr)) 2446 kve->kve_resident++; 2447 addr += PAGE_SIZE; 2448 } 2449 2450 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { 2451 if (tobj != obj) { 2452 VM_OBJECT_RLOCK(tobj); 2453 kve->kve_offset += tobj->backing_object_offset; 2454 } 2455 if (lobj != obj) 2456 VM_OBJECT_RUNLOCK(lobj); 2457 lobj = tobj; 2458 } 2459 2460 kve->kve_start = (void*)entry->start; 2461 kve->kve_end = (void*)entry->end; 2462 kve->kve_offset += (off_t)entry->offset; 2463 2464 if (entry->protection & VM_PROT_READ) 2465 kve->kve_protection |= KVME_PROT_READ; 2466 if (entry->protection & VM_PROT_WRITE) 2467 kve->kve_protection |= KVME_PROT_WRITE; 2468 if (entry->protection & VM_PROT_EXECUTE) 2469 kve->kve_protection |= KVME_PROT_EXEC; 2470 2471 if (entry->eflags & MAP_ENTRY_COW) 2472 kve->kve_flags |= KVME_FLAG_COW; 2473 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2474 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2475 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2476 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2477 2478 last_timestamp = map->timestamp; 2479 vm_map_unlock_read(map); 2480 2481 kve->kve_fileid = 0; 2482 kve->kve_fsid = 0; 2483 freepath = NULL; 2484 fullpath = ""; 2485 if (lobj) { 2486 kve->kve_type = vm_object_kvme_type(lobj, &vp); 2487 if (kve->kve_type == KVME_TYPE_MGTDEVICE) 2488 kve->kve_type = KVME_TYPE_UNKNOWN; 2489 if (vp != NULL) 2490 vref(vp); 2491 if (lobj != obj) 2492 VM_OBJECT_RUNLOCK(lobj); 2493 2494 kve->kve_ref_count = obj->ref_count; 2495 kve->kve_shadow_count = obj->shadow_count; 2496 VM_OBJECT_RUNLOCK(obj); 2497 if (vp != NULL) { 2498 vn_fullpath(vp, &fullpath, &freepath); 2499 cred = curthread->td_ucred; 2500 vn_lock(vp, LK_SHARED | LK_RETRY); 2501 if (VOP_GETATTR(vp, &va, cred) == 0) { 2502 kve->kve_fileid = va.va_fileid; 2503 /* truncate */ 2504 kve->kve_fsid = va.va_fsid; 2505 } 2506 vput(vp); 2507 } 2508 } else { 2509 kve->kve_type = KVME_TYPE_NONE; 2510 kve->kve_ref_count = 0; 2511 kve->kve_shadow_count = 0; 2512 } 2513 2514 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); 2515 if (freepath != NULL) 2516 free(freepath, M_TEMP); 2517 2518 error = SYSCTL_OUT(req, kve, sizeof(*kve)); 2519 vm_map_lock_read(map); 2520 if (error) 2521 break; 2522 if (last_timestamp != map->timestamp) { 2523 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2524 entry = tmp_entry; 2525 } 2526 } 2527 vm_map_unlock_read(map); 2528 vmspace_free(vm); 2529 PRELE(p); 2530 free(kve, M_TEMP); 2531 return (error); 2532 } 2533 #endif /* COMPAT_FREEBSD7 */ 2534 2535 #ifdef KINFO_VMENTRY_SIZE 2536 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); 2537 #endif 2538 2539 void 2540 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, 2541 int *resident_count, bool *super) 2542 { 2543 vm_object_t obj, tobj; 2544 vm_page_t m, m_adv; 2545 vm_offset_t addr; 2546 vm_paddr_t pa; 2547 vm_pindex_t pi, pi_adv, pindex; 2548 int incore; 2549 2550 *super = false; 2551 *resident_count = 0; 2552 if (vmmap_skip_res_cnt) 2553 return; 2554 2555 pa = 0; 2556 obj = entry->object.vm_object; 2557 addr = entry->start; 2558 m_adv = NULL; 2559 pi = OFF_TO_IDX(entry->offset); 2560 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { 2561 if (m_adv != NULL) { 2562 m = m_adv; 2563 } else { 2564 pi_adv = atop(entry->end - addr); 2565 pindex = pi; 2566 for (tobj = obj;; tobj = tobj->backing_object) { 2567 m = vm_radix_lookup_ge(&tobj->rtree, pindex); 2568 if (m != NULL) { 2569 if (m->pindex == pindex) 2570 break; 2571 if (pi_adv > m->pindex - pindex) { 2572 pi_adv = m->pindex - pindex; 2573 m_adv = m; 2574 } 2575 } 2576 if (tobj->backing_object == NULL) 2577 goto next; 2578 pindex += OFF_TO_IDX(tobj-> 2579 backing_object_offset); 2580 } 2581 } 2582 m_adv = NULL; 2583 if (m->psind != 0 && addr + pagesizes[1] <= entry->end && 2584 (addr & (pagesizes[1] - 1)) == 0 && (incore = 2585 pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) { 2586 *super = true; 2587 /* 2588 * The virtual page might be smaller than the physical 2589 * page, so we use the page size reported by the pmap 2590 * rather than m->psind. 2591 */ 2592 pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]); 2593 } else { 2594 /* 2595 * We do not test the found page on validity. 2596 * Either the page is busy and being paged in, 2597 * or it was invalidated. The first case 2598 * should be counted as resident, the second 2599 * is not so clear; we do account both. 2600 */ 2601 pi_adv = 1; 2602 } 2603 *resident_count += pi_adv; 2604 next:; 2605 } 2606 } 2607 2608 /* 2609 * Must be called with the process locked and will return unlocked. 2610 */ 2611 int 2612 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags) 2613 { 2614 vm_map_entry_t entry, tmp_entry; 2615 struct vattr va; 2616 vm_map_t map; 2617 vm_object_t lobj, nobj, obj, tobj; 2618 char *fullpath, *freepath; 2619 struct kinfo_vmentry *kve; 2620 struct ucred *cred; 2621 struct vnode *vp; 2622 struct vmspace *vm; 2623 vm_offset_t addr; 2624 unsigned int last_timestamp; 2625 int error; 2626 key_t key; 2627 unsigned short seq; 2628 bool guard, super; 2629 2630 PROC_LOCK_ASSERT(p, MA_OWNED); 2631 2632 _PHOLD(p); 2633 PROC_UNLOCK(p); 2634 vm = vmspace_acquire_ref(p); 2635 if (vm == NULL) { 2636 PRELE(p); 2637 return (ESRCH); 2638 } 2639 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO); 2640 2641 error = 0; 2642 map = &vm->vm_map; 2643 vm_map_lock_read(map); 2644 VM_MAP_ENTRY_FOREACH(entry, map) { 2645 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) 2646 continue; 2647 2648 addr = entry->end; 2649 bzero(kve, sizeof(*kve)); 2650 obj = entry->object.vm_object; 2651 if (obj != NULL) { 2652 if ((obj->flags & OBJ_ANON) != 0) 2653 kve->kve_obj = (uintptr_t)obj; 2654 2655 for (tobj = obj; tobj != NULL; 2656 tobj = tobj->backing_object) { 2657 VM_OBJECT_RLOCK(tobj); 2658 kve->kve_offset += tobj->backing_object_offset; 2659 lobj = tobj; 2660 } 2661 if (obj->backing_object == NULL) 2662 kve->kve_private_resident = 2663 obj->resident_page_count; 2664 kern_proc_vmmap_resident(map, entry, 2665 &kve->kve_resident, &super); 2666 if (super) 2667 kve->kve_flags |= KVME_FLAG_SUPER; 2668 for (tobj = obj; tobj != NULL; tobj = nobj) { 2669 nobj = tobj->backing_object; 2670 if (tobj != obj && tobj != lobj) 2671 VM_OBJECT_RUNLOCK(tobj); 2672 } 2673 } else { 2674 lobj = NULL; 2675 } 2676 2677 kve->kve_start = entry->start; 2678 kve->kve_end = entry->end; 2679 kve->kve_offset += entry->offset; 2680 2681 if (entry->protection & VM_PROT_READ) 2682 kve->kve_protection |= KVME_PROT_READ; 2683 if (entry->protection & VM_PROT_WRITE) 2684 kve->kve_protection |= KVME_PROT_WRITE; 2685 if (entry->protection & VM_PROT_EXECUTE) 2686 kve->kve_protection |= KVME_PROT_EXEC; 2687 if (entry->max_protection & VM_PROT_READ) 2688 kve->kve_protection |= KVME_MAX_PROT_READ; 2689 if (entry->max_protection & VM_PROT_WRITE) 2690 kve->kve_protection |= KVME_MAX_PROT_WRITE; 2691 if (entry->max_protection & VM_PROT_EXECUTE) 2692 kve->kve_protection |= KVME_MAX_PROT_EXEC; 2693 2694 if (entry->eflags & MAP_ENTRY_COW) 2695 kve->kve_flags |= KVME_FLAG_COW; 2696 if (entry->eflags & MAP_ENTRY_NEEDS_COPY) 2697 kve->kve_flags |= KVME_FLAG_NEEDS_COPY; 2698 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 2699 kve->kve_flags |= KVME_FLAG_NOCOREDUMP; 2700 if (entry->eflags & MAP_ENTRY_GROWS_DOWN) 2701 kve->kve_flags |= KVME_FLAG_GROWS_DOWN; 2702 if (entry->eflags & MAP_ENTRY_USER_WIRED) 2703 kve->kve_flags |= KVME_FLAG_USER_WIRED; 2704 2705 guard = (entry->eflags & MAP_ENTRY_GUARD) != 0; 2706 2707 last_timestamp = map->timestamp; 2708 vm_map_unlock_read(map); 2709 2710 freepath = NULL; 2711 fullpath = ""; 2712 if (lobj != NULL) { 2713 kve->kve_type = vm_object_kvme_type(lobj, &vp); 2714 if (vp != NULL) 2715 vref(vp); 2716 if (lobj != obj) 2717 VM_OBJECT_RUNLOCK(lobj); 2718 2719 kve->kve_ref_count = obj->ref_count; 2720 kve->kve_shadow_count = obj->shadow_count; 2721 if (obj->type == OBJT_DEVICE || 2722 obj->type == OBJT_MGTDEVICE) { 2723 cdev_pager_get_path(obj, kve->kve_path, 2724 sizeof(kve->kve_path)); 2725 } 2726 VM_OBJECT_RUNLOCK(obj); 2727 if ((lobj->flags & OBJ_SYSVSHM) != 0) { 2728 kve->kve_flags |= KVME_FLAG_SYSVSHM; 2729 shmobjinfo(lobj, &key, &seq); 2730 kve->kve_vn_fileid = key; 2731 kve->kve_vn_fsid_freebsd11 = seq; 2732 } 2733 if ((lobj->flags & OBJ_POSIXSHM) != 0) { 2734 kve->kve_flags |= KVME_FLAG_POSIXSHM; 2735 shm_get_path(lobj, kve->kve_path, 2736 sizeof(kve->kve_path)); 2737 } 2738 if (vp != NULL) { 2739 vn_fullpath(vp, &fullpath, &freepath); 2740 kve->kve_vn_type = vntype_to_kinfo(vp->v_type); 2741 cred = curthread->td_ucred; 2742 vn_lock(vp, LK_SHARED | LK_RETRY); 2743 if (VOP_GETATTR(vp, &va, cred) == 0) { 2744 kve->kve_vn_fileid = va.va_fileid; 2745 kve->kve_vn_fsid = va.va_fsid; 2746 kve->kve_vn_fsid_freebsd11 = 2747 kve->kve_vn_fsid; /* truncate */ 2748 kve->kve_vn_mode = 2749 MAKEIMODE(va.va_type, va.va_mode); 2750 kve->kve_vn_size = va.va_size; 2751 kve->kve_vn_rdev = va.va_rdev; 2752 kve->kve_vn_rdev_freebsd11 = 2753 kve->kve_vn_rdev; /* truncate */ 2754 kve->kve_status = KF_ATTR_VALID; 2755 } 2756 vput(vp); 2757 strlcpy(kve->kve_path, fullpath, sizeof( 2758 kve->kve_path)); 2759 free(freepath, M_TEMP); 2760 } 2761 } else { 2762 kve->kve_type = guard ? KVME_TYPE_GUARD : 2763 KVME_TYPE_NONE; 2764 kve->kve_ref_count = 0; 2765 kve->kve_shadow_count = 0; 2766 } 2767 2768 /* Pack record size down */ 2769 if ((flags & KERN_VMMAP_PACK_KINFO) != 0) 2770 kve->kve_structsize = 2771 offsetof(struct kinfo_vmentry, kve_path) + 2772 strlen(kve->kve_path) + 1; 2773 else 2774 kve->kve_structsize = sizeof(*kve); 2775 kve->kve_structsize = roundup(kve->kve_structsize, 2776 sizeof(uint64_t)); 2777 2778 /* Halt filling and truncate rather than exceeding maxlen */ 2779 if (maxlen != -1 && maxlen < kve->kve_structsize) { 2780 error = 0; 2781 vm_map_lock_read(map); 2782 break; 2783 } else if (maxlen != -1) 2784 maxlen -= kve->kve_structsize; 2785 2786 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0) 2787 error = ENOMEM; 2788 vm_map_lock_read(map); 2789 if (error != 0) 2790 break; 2791 if (last_timestamp != map->timestamp) { 2792 vm_map_lookup_entry(map, addr - 1, &tmp_entry); 2793 entry = tmp_entry; 2794 } 2795 } 2796 vm_map_unlock_read(map); 2797 vmspace_free(vm); 2798 PRELE(p); 2799 free(kve, M_TEMP); 2800 return (error); 2801 } 2802 2803 static int 2804 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) 2805 { 2806 struct proc *p; 2807 struct sbuf sb; 2808 u_int namelen; 2809 int error, error2, *name; 2810 2811 namelen = arg2; 2812 if (namelen != 1) 2813 return (EINVAL); 2814 2815 name = (int *)arg1; 2816 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); 2817 sbuf_clear_flags(&sb, SBUF_INCLUDENUL); 2818 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); 2819 if (error != 0) { 2820 sbuf_delete(&sb); 2821 return (error); 2822 } 2823 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO); 2824 error2 = sbuf_finish(&sb); 2825 sbuf_delete(&sb); 2826 return (error != 0 ? error : error2); 2827 } 2828 2829 #if defined(STACK) || defined(DDB) 2830 static int 2831 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) 2832 { 2833 struct kinfo_kstack *kkstp; 2834 int error, i, *name, numthreads; 2835 lwpid_t *lwpidarray; 2836 struct thread *td; 2837 struct stack *st; 2838 struct sbuf sb; 2839 struct proc *p; 2840 u_int namelen; 2841 2842 namelen = arg2; 2843 if (namelen != 1) 2844 return (EINVAL); 2845 2846 name = (int *)arg1; 2847 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); 2848 if (error != 0) 2849 return (error); 2850 2851 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); 2852 st = stack_create(M_WAITOK); 2853 2854 lwpidarray = NULL; 2855 PROC_LOCK(p); 2856 do { 2857 if (lwpidarray != NULL) { 2858 free(lwpidarray, M_TEMP); 2859 lwpidarray = NULL; 2860 } 2861 numthreads = p->p_numthreads; 2862 PROC_UNLOCK(p); 2863 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, 2864 M_WAITOK | M_ZERO); 2865 PROC_LOCK(p); 2866 } while (numthreads < p->p_numthreads); 2867 2868 /* 2869 * XXXRW: During the below loop, execve(2) and countless other sorts 2870 * of changes could have taken place. Should we check to see if the 2871 * vmspace has been replaced, or the like, in order to prevent 2872 * giving a snapshot that spans, say, execve(2), with some threads 2873 * before and some after? Among other things, the credentials could 2874 * have changed, in which case the right to extract debug info might 2875 * no longer be assured. 2876 */ 2877 i = 0; 2878 FOREACH_THREAD_IN_PROC(p, td) { 2879 KASSERT(i < numthreads, 2880 ("sysctl_kern_proc_kstack: numthreads")); 2881 lwpidarray[i] = td->td_tid; 2882 i++; 2883 } 2884 PROC_UNLOCK(p); 2885 numthreads = i; 2886 for (i = 0; i < numthreads; i++) { 2887 td = tdfind(lwpidarray[i], p->p_pid); 2888 if (td == NULL) { 2889 continue; 2890 } 2891 bzero(kkstp, sizeof(*kkstp)); 2892 (void)sbuf_new(&sb, kkstp->kkst_trace, 2893 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); 2894 thread_lock(td); 2895 kkstp->kkst_tid = td->td_tid; 2896 if (stack_save_td(st, td) == 0) 2897 kkstp->kkst_state = KKST_STATE_STACKOK; 2898 else 2899 kkstp->kkst_state = KKST_STATE_RUNNING; 2900 thread_unlock(td); 2901 PROC_UNLOCK(p); 2902 stack_sbuf_print(&sb, st); 2903 sbuf_finish(&sb); 2904 sbuf_delete(&sb); 2905 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); 2906 if (error) 2907 break; 2908 } 2909 PRELE(p); 2910 if (lwpidarray != NULL) 2911 free(lwpidarray, M_TEMP); 2912 stack_destroy(st); 2913 free(kkstp, M_TEMP); 2914 return (error); 2915 } 2916 #endif 2917 2918 /* 2919 * This sysctl allows a process to retrieve the full list of groups from 2920 * itself or another process. 2921 */ 2922 static int 2923 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) 2924 { 2925 pid_t *pidp = (pid_t *)arg1; 2926 unsigned int arglen = arg2; 2927 struct proc *p; 2928 struct ucred *cred; 2929 int error; 2930 2931 if (arglen != 1) 2932 return (EINVAL); 2933 if (*pidp == -1) { /* -1 means this process */ 2934 p = req->td->td_proc; 2935 PROC_LOCK(p); 2936 } else { 2937 error = pget(*pidp, PGET_CANSEE, &p); 2938 if (error != 0) 2939 return (error); 2940 } 2941 2942 cred = crhold(p->p_ucred); 2943 PROC_UNLOCK(p); 2944 2945 error = SYSCTL_OUT(req, &cred->cr_gid, sizeof(gid_t)); 2946 if (error == 0) 2947 error = SYSCTL_OUT(req, cred->cr_groups, 2948 cred->cr_ngroups * sizeof(gid_t)); 2949 2950 crfree(cred); 2951 return (error); 2952 } 2953 2954 /* 2955 * This sysctl allows a process to retrieve or/and set the resource limit for 2956 * another process. 2957 */ 2958 static int 2959 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) 2960 { 2961 int *name = (int *)arg1; 2962 u_int namelen = arg2; 2963 struct rlimit rlim; 2964 struct proc *p; 2965 u_int which; 2966 int flags, error; 2967 2968 if (namelen != 2) 2969 return (EINVAL); 2970 2971 which = (u_int)name[1]; 2972 if (which >= RLIM_NLIMITS) 2973 return (EINVAL); 2974 2975 if (req->newptr != NULL && req->newlen != sizeof(rlim)) 2976 return (EINVAL); 2977 2978 flags = PGET_HOLD | PGET_NOTWEXIT; 2979 if (req->newptr != NULL) 2980 flags |= PGET_CANDEBUG; 2981 else 2982 flags |= PGET_CANSEE; 2983 error = pget((pid_t)name[0], flags, &p); 2984 if (error != 0) 2985 return (error); 2986 2987 /* 2988 * Retrieve limit. 2989 */ 2990 if (req->oldptr != NULL) { 2991 PROC_LOCK(p); 2992 lim_rlimit_proc(p, which, &rlim); 2993 PROC_UNLOCK(p); 2994 } 2995 error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); 2996 if (error != 0) 2997 goto errout; 2998 2999 /* 3000 * Set limit. 3001 */ 3002 if (req->newptr != NULL) { 3003 error = SYSCTL_IN(req, &rlim, sizeof(rlim)); 3004 if (error == 0) 3005 error = kern_proc_setrlimit(curthread, p, which, &rlim); 3006 } 3007 3008 errout: 3009 PRELE(p); 3010 return (error); 3011 } 3012 3013 /* 3014 * This sysctl allows a process to retrieve ps_strings structure location of 3015 * another process. 3016 */ 3017 static int 3018 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) 3019 { 3020 int *name = (int *)arg1; 3021 u_int namelen = arg2; 3022 struct proc *p; 3023 vm_offset_t ps_strings; 3024 int error; 3025 #ifdef COMPAT_FREEBSD32 3026 uint32_t ps_strings32; 3027 #endif 3028 3029 if (namelen != 1) 3030 return (EINVAL); 3031 3032 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 3033 if (error != 0) 3034 return (error); 3035 #ifdef COMPAT_FREEBSD32 3036 if ((req->flags & SCTL_MASK32) != 0) { 3037 /* 3038 * We return 0 if the 32 bit emulation request is for a 64 bit 3039 * process. 3040 */ 3041 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? 3042 PTROUT(PROC_PS_STRINGS(p)) : 0; 3043 PROC_UNLOCK(p); 3044 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); 3045 return (error); 3046 } 3047 #endif 3048 ps_strings = PROC_PS_STRINGS(p); 3049 PROC_UNLOCK(p); 3050 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); 3051 return (error); 3052 } 3053 3054 /* 3055 * This sysctl allows a process to retrieve umask of another process. 3056 */ 3057 static int 3058 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) 3059 { 3060 int *name = (int *)arg1; 3061 u_int namelen = arg2; 3062 struct proc *p; 3063 int error; 3064 u_short cmask; 3065 pid_t pid; 3066 3067 if (namelen != 1) 3068 return (EINVAL); 3069 3070 pid = (pid_t)name[0]; 3071 p = curproc; 3072 if (pid == p->p_pid || pid == 0) { 3073 cmask = p->p_pd->pd_cmask; 3074 goto out; 3075 } 3076 3077 error = pget(pid, PGET_WANTREAD, &p); 3078 if (error != 0) 3079 return (error); 3080 3081 cmask = p->p_pd->pd_cmask; 3082 PRELE(p); 3083 out: 3084 error = SYSCTL_OUT(req, &cmask, sizeof(cmask)); 3085 return (error); 3086 } 3087 3088 /* 3089 * This sysctl allows a process to set and retrieve binary osreldate of 3090 * another process. 3091 */ 3092 static int 3093 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) 3094 { 3095 int *name = (int *)arg1; 3096 u_int namelen = arg2; 3097 struct proc *p; 3098 int flags, error, osrel; 3099 3100 if (namelen != 1) 3101 return (EINVAL); 3102 3103 if (req->newptr != NULL && req->newlen != sizeof(osrel)) 3104 return (EINVAL); 3105 3106 flags = PGET_HOLD | PGET_NOTWEXIT; 3107 if (req->newptr != NULL) 3108 flags |= PGET_CANDEBUG; 3109 else 3110 flags |= PGET_CANSEE; 3111 error = pget((pid_t)name[0], flags, &p); 3112 if (error != 0) 3113 return (error); 3114 3115 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); 3116 if (error != 0) 3117 goto errout; 3118 3119 if (req->newptr != NULL) { 3120 error = SYSCTL_IN(req, &osrel, sizeof(osrel)); 3121 if (error != 0) 3122 goto errout; 3123 if (osrel < 0) { 3124 error = EINVAL; 3125 goto errout; 3126 } 3127 p->p_osrel = osrel; 3128 } 3129 errout: 3130 PRELE(p); 3131 return (error); 3132 } 3133 3134 static int 3135 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) 3136 { 3137 int *name = (int *)arg1; 3138 u_int namelen = arg2; 3139 struct proc *p; 3140 struct kinfo_sigtramp kst; 3141 const struct sysentvec *sv; 3142 int error; 3143 #ifdef COMPAT_FREEBSD32 3144 struct kinfo_sigtramp32 kst32; 3145 #endif 3146 3147 if (namelen != 1) 3148 return (EINVAL); 3149 3150 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 3151 if (error != 0) 3152 return (error); 3153 sv = p->p_sysent; 3154 #ifdef COMPAT_FREEBSD32 3155 if ((req->flags & SCTL_MASK32) != 0) { 3156 bzero(&kst32, sizeof(kst32)); 3157 if (SV_PROC_FLAG(p, SV_ILP32)) { 3158 if (PROC_HAS_SHP(p)) { 3159 kst32.ksigtramp_start = PROC_SIGCODE(p); 3160 kst32.ksigtramp_end = kst32.ksigtramp_start + 3161 ((sv->sv_flags & SV_DSO_SIG) == 0 ? 3162 *sv->sv_szsigcode : 3163 (uintptr_t)sv->sv_szsigcode); 3164 } else { 3165 kst32.ksigtramp_start = PROC_PS_STRINGS(p) - 3166 *sv->sv_szsigcode; 3167 kst32.ksigtramp_end = PROC_PS_STRINGS(p); 3168 } 3169 } 3170 PROC_UNLOCK(p); 3171 error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); 3172 return (error); 3173 } 3174 #endif 3175 bzero(&kst, sizeof(kst)); 3176 if (PROC_HAS_SHP(p)) { 3177 kst.ksigtramp_start = (char *)PROC_SIGCODE(p); 3178 kst.ksigtramp_end = (char *)kst.ksigtramp_start + 3179 ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode : 3180 (uintptr_t)sv->sv_szsigcode); 3181 } else { 3182 kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) - 3183 *sv->sv_szsigcode; 3184 kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p); 3185 } 3186 PROC_UNLOCK(p); 3187 error = SYSCTL_OUT(req, &kst, sizeof(kst)); 3188 return (error); 3189 } 3190 3191 static int 3192 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS) 3193 { 3194 int *name = (int *)arg1; 3195 u_int namelen = arg2; 3196 pid_t pid; 3197 struct proc *p; 3198 struct thread *td1; 3199 uintptr_t addr; 3200 #ifdef COMPAT_FREEBSD32 3201 uint32_t addr32; 3202 #endif 3203 int error; 3204 3205 if (namelen != 1 || req->newptr != NULL) 3206 return (EINVAL); 3207 3208 pid = (pid_t)name[0]; 3209 error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p); 3210 if (error != 0) 3211 return (error); 3212 3213 PROC_LOCK(p); 3214 #ifdef COMPAT_FREEBSD32 3215 if (SV_CURPROC_FLAG(SV_ILP32)) { 3216 if (!SV_PROC_FLAG(p, SV_ILP32)) { 3217 error = EINVAL; 3218 goto errlocked; 3219 } 3220 } 3221 #endif 3222 if (pid <= PID_MAX) { 3223 td1 = FIRST_THREAD_IN_PROC(p); 3224 } else { 3225 FOREACH_THREAD_IN_PROC(p, td1) { 3226 if (td1->td_tid == pid) 3227 break; 3228 } 3229 } 3230 if (td1 == NULL) { 3231 error = ESRCH; 3232 goto errlocked; 3233 } 3234 /* 3235 * The access to the private thread flags. It is fine as far 3236 * as no out-of-thin-air values are read from td_pflags, and 3237 * usermode read of the td_sigblock_ptr is racy inherently, 3238 * since target process might have already changed it 3239 * meantime. 3240 */ 3241 if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0) 3242 addr = (uintptr_t)td1->td_sigblock_ptr; 3243 else 3244 error = ENOTTY; 3245 3246 errlocked: 3247 _PRELE(p); 3248 PROC_UNLOCK(p); 3249 if (error != 0) 3250 return (error); 3251 3252 #ifdef COMPAT_FREEBSD32 3253 if (SV_CURPROC_FLAG(SV_ILP32)) { 3254 addr32 = addr; 3255 error = SYSCTL_OUT(req, &addr32, sizeof(addr32)); 3256 } else 3257 #endif 3258 error = SYSCTL_OUT(req, &addr, sizeof(addr)); 3259 return (error); 3260 } 3261 3262 static int 3263 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS) 3264 { 3265 struct kinfo_vm_layout kvm; 3266 struct proc *p; 3267 struct vmspace *vmspace; 3268 int error, *name; 3269 3270 name = (int *)arg1; 3271 if ((u_int)arg2 != 1) 3272 return (EINVAL); 3273 3274 error = pget((pid_t)name[0], PGET_CANDEBUG, &p); 3275 if (error != 0) 3276 return (error); 3277 #ifdef COMPAT_FREEBSD32 3278 if (SV_CURPROC_FLAG(SV_ILP32)) { 3279 if (!SV_PROC_FLAG(p, SV_ILP32)) { 3280 PROC_UNLOCK(p); 3281 return (EINVAL); 3282 } 3283 } 3284 #endif 3285 vmspace = vmspace_acquire_ref(p); 3286 PROC_UNLOCK(p); 3287 3288 memset(&kvm, 0, sizeof(kvm)); 3289 kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map); 3290 kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map); 3291 kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr; 3292 kvm.kvm_text_size = vmspace->vm_tsize; 3293 kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr; 3294 kvm.kvm_data_size = vmspace->vm_dsize; 3295 kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr; 3296 kvm.kvm_stack_size = vmspace->vm_ssize; 3297 kvm.kvm_shp_addr = vmspace->vm_shp_base; 3298 kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len; 3299 if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0) 3300 kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE; 3301 if ((vmspace->vm_map.flags & MAP_ASLR) != 0) 3302 kvm.kvm_map_flags |= KMAP_FLAG_ASLR; 3303 if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0) 3304 kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART; 3305 if ((vmspace->vm_map.flags & MAP_WXORX) != 0) 3306 kvm.kvm_map_flags |= KMAP_FLAG_WXORX; 3307 if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0) 3308 kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK; 3309 if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base && 3310 PROC_HAS_SHP(p)) 3311 kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE; 3312 3313 #ifdef COMPAT_FREEBSD32 3314 if (SV_CURPROC_FLAG(SV_ILP32)) { 3315 struct kinfo_vm_layout32 kvm32; 3316 3317 memset(&kvm32, 0, sizeof(kvm32)); 3318 kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr; 3319 kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr; 3320 kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr; 3321 kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size; 3322 kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr; 3323 kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size; 3324 kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr; 3325 kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size; 3326 kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr; 3327 kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size; 3328 kvm32.kvm_map_flags = kvm.kvm_map_flags; 3329 error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32)); 3330 goto out; 3331 } 3332 #endif 3333 3334 error = SYSCTL_OUT(req, &kvm, sizeof(kvm)); 3335 #ifdef COMPAT_FREEBSD32 3336 out: 3337 #endif 3338 vmspace_free(vmspace); 3339 return (error); 3340 } 3341 3342 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 3343 "Process table"); 3344 3345 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| 3346 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", 3347 "Return entire process table"); 3348 3349 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3350 sysctl_kern_proc, "Process table"); 3351 3352 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, 3353 sysctl_kern_proc, "Process table"); 3354 3355 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3356 sysctl_kern_proc, "Process table"); 3357 3358 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | 3359 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3360 3361 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 3362 sysctl_kern_proc, "Process table"); 3363 3364 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3365 sysctl_kern_proc, "Process table"); 3366 3367 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3368 sysctl_kern_proc, "Process table"); 3369 3370 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, 3371 sysctl_kern_proc, "Process table"); 3372 3373 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 3374 sysctl_kern_proc, "Return process table, no threads"); 3375 3376 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, 3377 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, 3378 sysctl_kern_proc_args, "Process argument list"); 3379 3380 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, 3381 sysctl_kern_proc_env, "Process environment"); 3382 3383 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | 3384 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); 3385 3386 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | 3387 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); 3388 3389 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | 3390 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, 3391 "Process syscall vector name (ABI type)"); 3392 3393 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, 3394 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3395 3396 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, 3397 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3398 3399 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, 3400 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3401 3402 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), 3403 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3404 3405 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, 3406 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3407 3408 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, 3409 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3410 3411 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, 3412 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3413 3414 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, 3415 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); 3416 3417 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, 3418 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, 3419 "Return process table, including threads"); 3420 3421 #ifdef COMPAT_FREEBSD7 3422 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | 3423 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); 3424 #endif 3425 3426 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | 3427 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); 3428 3429 #if defined(STACK) || defined(DDB) 3430 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | 3431 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); 3432 #endif 3433 3434 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | 3435 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); 3436 3437 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | 3438 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, 3439 "Process resource limits"); 3440 3441 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | 3442 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, 3443 "Process ps_strings location"); 3444 3445 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | 3446 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); 3447 3448 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | 3449 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, 3450 "Process binary osreldate"); 3451 3452 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | 3453 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, 3454 "Process signal trampoline location"); 3455 3456 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD | 3457 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk, 3458 "Thread sigfastblock address"); 3459 3460 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD | 3461 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout, 3462 "Process virtual address space layout info"); 3463 3464 static struct sx stop_all_proc_blocker; 3465 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk"); 3466 3467 bool 3468 stop_all_proc_block(void) 3469 { 3470 return (sx_xlock_sig(&stop_all_proc_blocker) == 0); 3471 } 3472 3473 void 3474 stop_all_proc_unblock(void) 3475 { 3476 sx_xunlock(&stop_all_proc_blocker); 3477 } 3478 3479 int allproc_gen; 3480 3481 /* 3482 * stop_all_proc() purpose is to stop all process which have usermode, 3483 * except current process for obvious reasons. This makes it somewhat 3484 * unreliable when invoked from multithreaded process. The service 3485 * must not be user-callable anyway. 3486 */ 3487 void 3488 stop_all_proc(void) 3489 { 3490 struct proc *cp, *p; 3491 int r, gen; 3492 bool restart, seen_stopped, seen_exiting, stopped_some; 3493 3494 if (!stop_all_proc_block()) 3495 return; 3496 3497 cp = curproc; 3498 allproc_loop: 3499 sx_xlock(&allproc_lock); 3500 gen = allproc_gen; 3501 seen_exiting = seen_stopped = stopped_some = restart = false; 3502 LIST_REMOVE(cp, p_list); 3503 LIST_INSERT_HEAD(&allproc, cp, p_list); 3504 for (;;) { 3505 p = LIST_NEXT(cp, p_list); 3506 if (p == NULL) 3507 break; 3508 LIST_REMOVE(cp, p_list); 3509 LIST_INSERT_AFTER(p, cp, p_list); 3510 PROC_LOCK(p); 3511 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP | 3512 P_STOPPED_SIG)) != 0) { 3513 PROC_UNLOCK(p); 3514 continue; 3515 } 3516 if ((p->p_flag2 & P2_WEXIT) != 0) { 3517 seen_exiting = true; 3518 PROC_UNLOCK(p); 3519 continue; 3520 } 3521 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { 3522 /* 3523 * Stopped processes are tolerated when there 3524 * are no other processes which might continue 3525 * them. P_STOPPED_SINGLE but not 3526 * P_TOTAL_STOP process still has at least one 3527 * thread running. 3528 */ 3529 seen_stopped = true; 3530 PROC_UNLOCK(p); 3531 continue; 3532 } 3533 if ((p->p_flag & P_TRACED) != 0) { 3534 /* 3535 * thread_single() below cannot stop traced p, 3536 * so skip it. OTOH, we cannot require 3537 * restart because debugger might be either 3538 * already stopped or traced as well. 3539 */ 3540 PROC_UNLOCK(p); 3541 continue; 3542 } 3543 sx_xunlock(&allproc_lock); 3544 _PHOLD(p); 3545 r = thread_single(p, SINGLE_ALLPROC); 3546 if (r != 0) 3547 restart = true; 3548 else 3549 stopped_some = true; 3550 _PRELE(p); 3551 PROC_UNLOCK(p); 3552 sx_xlock(&allproc_lock); 3553 } 3554 /* Catch forked children we did not see in iteration. */ 3555 if (gen != allproc_gen) 3556 restart = true; 3557 sx_xunlock(&allproc_lock); 3558 if (restart || stopped_some || seen_exiting || seen_stopped) { 3559 kern_yield(PRI_USER); 3560 goto allproc_loop; 3561 } 3562 } 3563 3564 void 3565 resume_all_proc(void) 3566 { 3567 struct proc *cp, *p; 3568 3569 cp = curproc; 3570 sx_xlock(&allproc_lock); 3571 again: 3572 LIST_REMOVE(cp, p_list); 3573 LIST_INSERT_HEAD(&allproc, cp, p_list); 3574 for (;;) { 3575 p = LIST_NEXT(cp, p_list); 3576 if (p == NULL) 3577 break; 3578 LIST_REMOVE(cp, p_list); 3579 LIST_INSERT_AFTER(p, cp, p_list); 3580 PROC_LOCK(p); 3581 if ((p->p_flag & P_TOTAL_STOP) != 0) { 3582 sx_xunlock(&allproc_lock); 3583 _PHOLD(p); 3584 thread_single_end(p, SINGLE_ALLPROC); 3585 _PRELE(p); 3586 PROC_UNLOCK(p); 3587 sx_xlock(&allproc_lock); 3588 } else { 3589 PROC_UNLOCK(p); 3590 } 3591 } 3592 /* Did the loop above missed any stopped process ? */ 3593 FOREACH_PROC_IN_SYSTEM(p) { 3594 /* No need for proc lock. */ 3595 if ((p->p_flag & P_TOTAL_STOP) != 0) 3596 goto again; 3597 } 3598 sx_xunlock(&allproc_lock); 3599 3600 stop_all_proc_unblock(); 3601 } 3602 3603 /* #define TOTAL_STOP_DEBUG 1 */ 3604 #ifdef TOTAL_STOP_DEBUG 3605 volatile static int ap_resume; 3606 #include <sys/mount.h> 3607 3608 static int 3609 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) 3610 { 3611 int error, val; 3612 3613 val = 0; 3614 ap_resume = 0; 3615 error = sysctl_handle_int(oidp, &val, 0, req); 3616 if (error != 0 || req->newptr == NULL) 3617 return (error); 3618 if (val != 0) { 3619 stop_all_proc(); 3620 syncer_suspend(); 3621 while (ap_resume == 0) 3622 ; 3623 syncer_resume(); 3624 resume_all_proc(); 3625 } 3626 return (0); 3627 } 3628 3629 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | 3630 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, 3631 sysctl_debug_stop_all_proc, "I", 3632 ""); 3633 #endif 3634