1 /* $NetBSD: kern_resource.c,v 1.157 2010/07/01 02:38:30 rmind Exp $ */ 2 3 /*- 4 * Copyright (c) 1982, 1986, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)kern_resource.c 8.8 (Berkeley) 2/14/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: kern_resource.c,v 1.157 2010/07/01 02:38:30 rmind Exp $"); 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/kernel.h> 45 #include <sys/file.h> 46 #include <sys/resourcevar.h> 47 #include <sys/malloc.h> 48 #include <sys/kmem.h> 49 #include <sys/namei.h> 50 #include <sys/pool.h> 51 #include <sys/proc.h> 52 #include <sys/sysctl.h> 53 #include <sys/timevar.h> 54 #include <sys/kauth.h> 55 #include <sys/atomic.h> 56 #include <sys/mount.h> 57 #include <sys/syscallargs.h> 58 #include <sys/atomic.h> 59 60 #include <uvm/uvm_extern.h> 61 62 /* 63 * Maximum process data and stack limits. 64 * They are variables so they are patchable. 65 */ 66 rlim_t maxdmap = MAXDSIZ; 67 rlim_t maxsmap = MAXSSIZ; 68 69 static pool_cache_t plimit_cache; 70 static pool_cache_t pstats_cache; 71 72 static kauth_listener_t resource_listener; 73 74 static void sysctl_proc_setup(void); 75 76 static int 77 resource_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 78 void *arg0, void *arg1, void *arg2, void *arg3) 79 { 80 struct proc *p; 81 int result; 82 83 result = KAUTH_RESULT_DEFER; 84 p = arg0; 85 86 switch (action) { 87 case KAUTH_PROCESS_NICE: 88 if (kauth_cred_geteuid(cred) != kauth_cred_geteuid(p->p_cred) && 89 kauth_cred_getuid(cred) != kauth_cred_geteuid(p->p_cred)) { 90 break; 91 } 92 93 if ((u_long)arg1 >= p->p_nice) 94 result = KAUTH_RESULT_ALLOW; 95 96 break; 97 98 case KAUTH_PROCESS_RLIMIT: { 99 enum kauth_process_req req; 100 101 req = (enum kauth_process_req)(unsigned long)arg1; 102 103 switch (req) { 104 case KAUTH_REQ_PROCESS_RLIMIT_GET: 105 result = KAUTH_RESULT_ALLOW; 106 break; 107 108 case KAUTH_REQ_PROCESS_RLIMIT_SET: { 109 struct rlimit *new_rlimit; 110 u_long which; 111 112 if ((p != curlwp->l_proc) && 113 (proc_uidmatch(cred, p->p_cred) != 0)) 114 break; 115 116 new_rlimit = arg2; 117 which = (u_long)arg3; 118 119 if (new_rlimit->rlim_max <= p->p_rlimit[which].rlim_max) 120 result = KAUTH_RESULT_ALLOW; 121 122 break; 123 } 124 125 default: 126 break; 127 } 128 129 break; 130 } 131 132 default: 133 break; 134 } 135 136 return result; 137 } 138 139 void 140 resource_init(void) 141 { 142 143 plimit_cache = pool_cache_init(sizeof(struct plimit), 0, 0, 0, 144 "plimitpl", NULL, IPL_NONE, NULL, NULL, NULL); 145 pstats_cache = pool_cache_init(sizeof(struct pstats), 0, 0, 0, 146 "pstatspl", NULL, IPL_NONE, NULL, NULL, NULL); 147 148 resource_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 149 resource_listener_cb, NULL); 150 151 sysctl_proc_setup(); 152 } 153 154 /* 155 * Resource controls and accounting. 156 */ 157 158 int 159 sys_getpriority(struct lwp *l, const struct sys_getpriority_args *uap, 160 register_t *retval) 161 { 162 /* { 163 syscallarg(int) which; 164 syscallarg(id_t) who; 165 } */ 166 struct proc *curp = l->l_proc, *p; 167 int low = NZERO + PRIO_MAX + 1; 168 int who = SCARG(uap, who); 169 170 mutex_enter(proc_lock); 171 switch (SCARG(uap, which)) { 172 case PRIO_PROCESS: 173 p = who ? proc_find(who) : curp;; 174 if (p != NULL) 175 low = p->p_nice; 176 break; 177 178 case PRIO_PGRP: { 179 struct pgrp *pg; 180 181 if (who == 0) 182 pg = curp->p_pgrp; 183 else if ((pg = pgrp_find(who)) == NULL) 184 break; 185 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 186 if (p->p_nice < low) 187 low = p->p_nice; 188 } 189 break; 190 } 191 192 case PRIO_USER: 193 if (who == 0) 194 who = (int)kauth_cred_geteuid(l->l_cred); 195 PROCLIST_FOREACH(p, &allproc) { 196 mutex_enter(p->p_lock); 197 if (kauth_cred_geteuid(p->p_cred) == 198 (uid_t)who && p->p_nice < low) 199 low = p->p_nice; 200 mutex_exit(p->p_lock); 201 } 202 break; 203 204 default: 205 mutex_exit(proc_lock); 206 return (EINVAL); 207 } 208 mutex_exit(proc_lock); 209 210 if (low == NZERO + PRIO_MAX + 1) 211 return (ESRCH); 212 *retval = low - NZERO; 213 return (0); 214 } 215 216 /* ARGSUSED */ 217 int 218 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap, 219 register_t *retval) 220 { 221 /* { 222 syscallarg(int) which; 223 syscallarg(id_t) who; 224 syscallarg(int) prio; 225 } */ 226 struct proc *curp = l->l_proc, *p; 227 int found = 0, error = 0; 228 int who = SCARG(uap, who); 229 230 mutex_enter(proc_lock); 231 switch (SCARG(uap, which)) { 232 case PRIO_PROCESS: 233 p = who ? proc_find(who) : curp; 234 if (p != NULL) { 235 mutex_enter(p->p_lock); 236 error = donice(l, p, SCARG(uap, prio)); 237 mutex_exit(p->p_lock); 238 found++; 239 } 240 break; 241 242 case PRIO_PGRP: { 243 struct pgrp *pg; 244 245 if (who == 0) 246 pg = curp->p_pgrp; 247 else if ((pg = pgrp_find(who)) == NULL) 248 break; 249 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 250 mutex_enter(p->p_lock); 251 error = donice(l, p, SCARG(uap, prio)); 252 mutex_exit(p->p_lock); 253 found++; 254 } 255 break; 256 } 257 258 case PRIO_USER: 259 if (who == 0) 260 who = (int)kauth_cred_geteuid(l->l_cred); 261 PROCLIST_FOREACH(p, &allproc) { 262 mutex_enter(p->p_lock); 263 if (kauth_cred_geteuid(p->p_cred) == 264 (uid_t)SCARG(uap, who)) { 265 error = donice(l, p, SCARG(uap, prio)); 266 found++; 267 } 268 mutex_exit(p->p_lock); 269 } 270 break; 271 272 default: 273 mutex_exit(proc_lock); 274 return EINVAL; 275 } 276 mutex_exit(proc_lock); 277 if (found == 0) 278 return (ESRCH); 279 return (error); 280 } 281 282 /* 283 * Renice a process. 284 * 285 * Call with the target process' credentials locked. 286 */ 287 int 288 donice(struct lwp *l, struct proc *chgp, int n) 289 { 290 kauth_cred_t cred = l->l_cred; 291 292 KASSERT(mutex_owned(chgp->p_lock)); 293 294 if (kauth_cred_geteuid(cred) && kauth_cred_getuid(cred) && 295 kauth_cred_geteuid(cred) != kauth_cred_geteuid(chgp->p_cred) && 296 kauth_cred_getuid(cred) != kauth_cred_geteuid(chgp->p_cred)) 297 return (EPERM); 298 299 if (n > PRIO_MAX) 300 n = PRIO_MAX; 301 if (n < PRIO_MIN) 302 n = PRIO_MIN; 303 n += NZERO; 304 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 305 KAUTH_ARG(n), NULL, NULL)) 306 return (EACCES); 307 sched_nice(chgp, n); 308 return (0); 309 } 310 311 /* ARGSUSED */ 312 int 313 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap, 314 register_t *retval) 315 { 316 /* { 317 syscallarg(int) which; 318 syscallarg(const struct rlimit *) rlp; 319 } */ 320 int which = SCARG(uap, which); 321 struct rlimit alim; 322 int error; 323 324 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 325 if (error) 326 return (error); 327 return (dosetrlimit(l, l->l_proc, which, &alim)); 328 } 329 330 int 331 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 332 { 333 struct rlimit *alimp; 334 int error; 335 336 if ((u_int)which >= RLIM_NLIMITS) 337 return (EINVAL); 338 339 if (limp->rlim_cur > limp->rlim_max) { 340 /* 341 * This is programming error. According to SUSv2, we should 342 * return error in this case. 343 */ 344 return (EINVAL); 345 } 346 347 alimp = &p->p_rlimit[which]; 348 /* if we don't change the value, no need to limcopy() */ 349 if (limp->rlim_cur == alimp->rlim_cur && 350 limp->rlim_max == alimp->rlim_max) 351 return 0; 352 353 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 354 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_SET), limp, KAUTH_ARG(which)); 355 if (error) 356 return (error); 357 358 lim_privatise(p, false); 359 /* p->p_limit is now unchangeable */ 360 alimp = &p->p_rlimit[which]; 361 362 switch (which) { 363 364 case RLIMIT_DATA: 365 if (limp->rlim_cur > maxdmap) 366 limp->rlim_cur = maxdmap; 367 if (limp->rlim_max > maxdmap) 368 limp->rlim_max = maxdmap; 369 break; 370 371 case RLIMIT_STACK: 372 if (limp->rlim_cur > maxsmap) 373 limp->rlim_cur = maxsmap; 374 if (limp->rlim_max > maxsmap) 375 limp->rlim_max = maxsmap; 376 377 /* 378 * Return EINVAL if the new stack size limit is lower than 379 * current usage. Otherwise, the process would get SIGSEGV the 380 * moment it would try to access anything on it's current stack. 381 * This conforms to SUSv2. 382 */ 383 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 384 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 385 return (EINVAL); 386 } 387 388 /* 389 * Stack is allocated to the max at exec time with 390 * only "rlim_cur" bytes accessible (In other words, 391 * allocates stack dividing two contiguous regions at 392 * "rlim_cur" bytes boundary). 393 * 394 * Since allocation is done in terms of page, roundup 395 * "rlim_cur" (otherwise, contiguous regions 396 * overlap). If stack limit is going up make more 397 * accessible, if going down make inaccessible. 398 */ 399 limp->rlim_cur = round_page(limp->rlim_cur); 400 if (limp->rlim_cur != alimp->rlim_cur) { 401 vaddr_t addr; 402 vsize_t size; 403 vm_prot_t prot; 404 405 if (limp->rlim_cur > alimp->rlim_cur) { 406 prot = VM_PROT_READ | VM_PROT_WRITE; 407 size = limp->rlim_cur - alimp->rlim_cur; 408 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 409 limp->rlim_cur; 410 } else { 411 prot = VM_PROT_NONE; 412 size = alimp->rlim_cur - limp->rlim_cur; 413 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 414 alimp->rlim_cur; 415 } 416 (void) uvm_map_protect(&p->p_vmspace->vm_map, 417 addr, addr+size, prot, false); 418 } 419 break; 420 421 case RLIMIT_NOFILE: 422 if (limp->rlim_cur > maxfiles) 423 limp->rlim_cur = maxfiles; 424 if (limp->rlim_max > maxfiles) 425 limp->rlim_max = maxfiles; 426 break; 427 428 case RLIMIT_NPROC: 429 if (limp->rlim_cur > maxproc) 430 limp->rlim_cur = maxproc; 431 if (limp->rlim_max > maxproc) 432 limp->rlim_max = maxproc; 433 break; 434 } 435 436 mutex_enter(&p->p_limit->pl_lock); 437 *alimp = *limp; 438 mutex_exit(&p->p_limit->pl_lock); 439 return (0); 440 } 441 442 /* ARGSUSED */ 443 int 444 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, 445 register_t *retval) 446 { 447 /* { 448 syscallarg(int) which; 449 syscallarg(struct rlimit *) rlp; 450 } */ 451 struct proc *p = l->l_proc; 452 int which = SCARG(uap, which); 453 struct rlimit rl; 454 455 if ((u_int)which >= RLIM_NLIMITS) 456 return (EINVAL); 457 458 mutex_enter(p->p_lock); 459 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 460 mutex_exit(p->p_lock); 461 462 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 463 } 464 465 /* 466 * Transform the running time and tick information in proc p into user, 467 * system, and interrupt time usage. 468 * 469 * Should be called with p->p_lock held unless called from exit1(). 470 */ 471 void 472 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 473 struct timeval *ip, struct timeval *rp) 474 { 475 uint64_t u, st, ut, it, tot; 476 struct lwp *l; 477 struct bintime tm; 478 struct timeval tv; 479 480 mutex_spin_enter(&p->p_stmutex); 481 st = p->p_sticks; 482 ut = p->p_uticks; 483 it = p->p_iticks; 484 mutex_spin_exit(&p->p_stmutex); 485 486 tm = p->p_rtime; 487 488 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 489 lwp_lock(l); 490 bintime_add(&tm, &l->l_rtime); 491 if ((l->l_pflag & LP_RUNNING) != 0) { 492 struct bintime diff; 493 /* 494 * Adjust for the current time slice. This is 495 * actually fairly important since the error 496 * here is on the order of a time quantum, 497 * which is much greater than the sampling 498 * error. 499 */ 500 binuptime(&diff); 501 bintime_sub(&diff, &l->l_stime); 502 bintime_add(&tm, &diff); 503 } 504 lwp_unlock(l); 505 } 506 507 tot = st + ut + it; 508 bintime2timeval(&tm, &tv); 509 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 510 511 if (tot == 0) { 512 /* No ticks, so can't use to share time out, split 50-50 */ 513 st = ut = u / 2; 514 } else { 515 st = (u * st) / tot; 516 ut = (u * ut) / tot; 517 } 518 if (sp != NULL) { 519 sp->tv_sec = st / 1000000; 520 sp->tv_usec = st % 1000000; 521 } 522 if (up != NULL) { 523 up->tv_sec = ut / 1000000; 524 up->tv_usec = ut % 1000000; 525 } 526 if (ip != NULL) { 527 if (it != 0) 528 it = (u * it) / tot; 529 ip->tv_sec = it / 1000000; 530 ip->tv_usec = it % 1000000; 531 } 532 if (rp != NULL) { 533 *rp = tv; 534 } 535 } 536 537 /* ARGSUSED */ 538 int 539 sys___getrusage50(struct lwp *l, const struct sys___getrusage50_args *uap, 540 register_t *retval) 541 { 542 /* { 543 syscallarg(int) who; 544 syscallarg(struct rusage *) rusage; 545 } */ 546 struct rusage ru; 547 struct proc *p = l->l_proc; 548 549 switch (SCARG(uap, who)) { 550 case RUSAGE_SELF: 551 mutex_enter(p->p_lock); 552 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 553 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 554 rulwps(p, &ru); 555 mutex_exit(p->p_lock); 556 break; 557 558 case RUSAGE_CHILDREN: 559 mutex_enter(p->p_lock); 560 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 561 mutex_exit(p->p_lock); 562 break; 563 564 default: 565 return EINVAL; 566 } 567 568 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 569 } 570 571 void 572 ruadd(struct rusage *ru, struct rusage *ru2) 573 { 574 long *ip, *ip2; 575 int i; 576 577 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 578 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 579 if (ru->ru_maxrss < ru2->ru_maxrss) 580 ru->ru_maxrss = ru2->ru_maxrss; 581 ip = &ru->ru_first; ip2 = &ru2->ru_first; 582 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 583 *ip++ += *ip2++; 584 } 585 586 void 587 rulwps(proc_t *p, struct rusage *ru) 588 { 589 lwp_t *l; 590 591 KASSERT(mutex_owned(p->p_lock)); 592 593 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 594 ruadd(ru, &l->l_ru); 595 ru->ru_nvcsw += (l->l_ncsw - l->l_nivcsw); 596 ru->ru_nivcsw += l->l_nivcsw; 597 } 598 } 599 600 /* 601 * Make a copy of the plimit structure. 602 * We share these structures copy-on-write after fork, 603 * and copy when a limit is changed. 604 * 605 * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure 606 * we are copying to change beneath our feet! 607 */ 608 struct plimit * 609 lim_copy(struct plimit *lim) 610 { 611 struct plimit *newlim; 612 char *corename; 613 size_t alen, len; 614 615 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 616 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 617 newlim->pl_flags = 0; 618 newlim->pl_refcnt = 1; 619 newlim->pl_sv_limit = NULL; 620 621 mutex_enter(&lim->pl_lock); 622 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 623 sizeof(struct rlimit) * RLIM_NLIMITS); 624 625 alen = 0; 626 corename = NULL; 627 for (;;) { 628 if (lim->pl_corename == defcorename) { 629 newlim->pl_corename = defcorename; 630 break; 631 } 632 len = strlen(lim->pl_corename) + 1; 633 if (len <= alen) { 634 newlim->pl_corename = corename; 635 memcpy(corename, lim->pl_corename, len); 636 corename = NULL; 637 break; 638 } 639 mutex_exit(&lim->pl_lock); 640 if (corename != NULL) 641 free(corename, M_TEMP); 642 alen = len; 643 corename = malloc(alen, M_TEMP, M_WAITOK); 644 mutex_enter(&lim->pl_lock); 645 } 646 mutex_exit(&lim->pl_lock); 647 if (corename != NULL) 648 free(corename, M_TEMP); 649 return newlim; 650 } 651 652 void 653 lim_addref(struct plimit *lim) 654 { 655 atomic_inc_uint(&lim->pl_refcnt); 656 } 657 658 /* 659 * Give a process it's own private plimit structure. 660 * This will only be shared (in fork) if modifications are to be shared. 661 */ 662 void 663 lim_privatise(struct proc *p, bool set_shared) 664 { 665 struct plimit *lim, *newlim; 666 667 lim = p->p_limit; 668 if (lim->pl_flags & PL_WRITEABLE) { 669 if (set_shared) 670 lim->pl_flags |= PL_SHAREMOD; 671 return; 672 } 673 674 if (set_shared && lim->pl_flags & PL_SHAREMOD) 675 return; 676 677 newlim = lim_copy(lim); 678 679 mutex_enter(p->p_lock); 680 if (p->p_limit->pl_flags & PL_WRITEABLE) { 681 /* Someone crept in while we were busy */ 682 mutex_exit(p->p_lock); 683 limfree(newlim); 684 if (set_shared) 685 p->p_limit->pl_flags |= PL_SHAREMOD; 686 return; 687 } 688 689 /* 690 * Since most accesses to p->p_limit aren't locked, we must not 691 * delete the old limit structure yet. 692 */ 693 newlim->pl_sv_limit = p->p_limit; 694 newlim->pl_flags |= PL_WRITEABLE; 695 if (set_shared) 696 newlim->pl_flags |= PL_SHAREMOD; 697 p->p_limit = newlim; 698 mutex_exit(p->p_lock); 699 } 700 701 void 702 limfree(struct plimit *lim) 703 { 704 struct plimit *sv_lim; 705 706 do { 707 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) 708 return; 709 if (lim->pl_corename != defcorename) 710 free(lim->pl_corename, M_TEMP); 711 sv_lim = lim->pl_sv_limit; 712 mutex_destroy(&lim->pl_lock); 713 pool_cache_put(plimit_cache, lim); 714 } while ((lim = sv_lim) != NULL); 715 } 716 717 struct pstats * 718 pstatscopy(struct pstats *ps) 719 { 720 721 struct pstats *newps; 722 723 newps = pool_cache_get(pstats_cache, PR_WAITOK); 724 725 memset(&newps->pstat_startzero, 0, 726 (unsigned) ((char *)&newps->pstat_endzero - 727 (char *)&newps->pstat_startzero)); 728 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 729 ((char *)&newps->pstat_endcopy - 730 (char *)&newps->pstat_startcopy)); 731 732 return (newps); 733 734 } 735 736 void 737 pstatsfree(struct pstats *ps) 738 { 739 740 pool_cache_put(pstats_cache, ps); 741 } 742 743 /* 744 * sysctl interface in five parts 745 */ 746 747 /* 748 * sysctl_proc_findproc: a routine for sysctl proc subtree helpers that 749 * need to pick a valid process by PID. 750 * 751 * => Hold a reference on the process, on success. 752 */ 753 static int 754 sysctl_proc_findproc(lwp_t *l, pid_t pid, proc_t **p2) 755 { 756 proc_t *p; 757 int error; 758 759 if (pid == PROC_CURPROC) { 760 p = l->l_proc; 761 } else { 762 mutex_enter(proc_lock); 763 p = proc_find(pid); 764 if (p == NULL) { 765 mutex_exit(proc_lock); 766 return ESRCH; 767 } 768 } 769 error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY; 770 if (pid != PROC_CURPROC) { 771 mutex_exit(proc_lock); 772 } 773 *p2 = p; 774 return error; 775 } 776 777 /* 778 * sysctl helper routine for setting a process's specific corefile 779 * name. picks the process based on the given pid and checks the 780 * correctness of the new value. 781 */ 782 static int 783 sysctl_proc_corename(SYSCTLFN_ARGS) 784 { 785 struct proc *ptmp; 786 struct plimit *lim; 787 char *cname, *ocore, *tmp; 788 struct sysctlnode node; 789 int error = 0, len; 790 791 /* 792 * is this all correct? 793 */ 794 if (namelen != 0) 795 return (EINVAL); 796 if (name[-1] != PROC_PID_CORENAME) 797 return (EINVAL); 798 799 /* Find the process. Hold a reference (p_reflock), if found. */ 800 error = sysctl_proc_findproc(l, (pid_t)name[-2], &ptmp); 801 if (error) 802 return error; 803 804 /* XXX-elad */ 805 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 806 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 807 if (error) { 808 rw_exit(&ptmp->p_reflock); 809 return error; 810 } 811 812 if (newp == NULL) { 813 error = kauth_authorize_process(l->l_cred, 814 KAUTH_PROCESS_CORENAME, ptmp, 815 KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_GET), NULL, NULL); 816 if (error) { 817 rw_exit(&ptmp->p_reflock); 818 return error; 819 } 820 } 821 822 /* 823 * let them modify a temporary copy of the core name 824 */ 825 cname = PNBUF_GET(); 826 lim = ptmp->p_limit; 827 mutex_enter(&lim->pl_lock); 828 strlcpy(cname, lim->pl_corename, MAXPATHLEN); 829 mutex_exit(&lim->pl_lock); 830 831 node = *rnode; 832 node.sysctl_data = cname; 833 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 834 835 /* 836 * if that failed, or they have nothing new to say, or we've 837 * heard it before... 838 */ 839 if (error || newp == NULL) 840 goto done; 841 lim = ptmp->p_limit; 842 mutex_enter(&lim->pl_lock); 843 error = strcmp(cname, lim->pl_corename); 844 mutex_exit(&lim->pl_lock); 845 if (error == 0) { 846 /* Unchanged */ 847 goto done; 848 } 849 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 850 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_SET), cname, NULL); 851 if (error) 852 goto done; 853 854 /* 855 * no error yet and cname now has the new core name in it. 856 * let's see if it looks acceptable. it must be either "core" 857 * or end in ".core" or "/core". 858 */ 859 len = strlen(cname); 860 if (len < 4) { 861 error = EINVAL; 862 } else if (strcmp(cname + len - 4, "core") != 0) { 863 error = EINVAL; 864 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 865 error = EINVAL; 866 } 867 if (error != 0) { 868 goto done; 869 } 870 871 /* 872 * hmm...looks good. now...where do we put it? 873 */ 874 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 875 if (tmp == NULL) { 876 error = ENOMEM; 877 goto done; 878 } 879 memcpy(tmp, cname, len + 1); 880 881 lim_privatise(ptmp, false); 882 lim = ptmp->p_limit; 883 mutex_enter(&lim->pl_lock); 884 ocore = lim->pl_corename; 885 lim->pl_corename = tmp; 886 mutex_exit(&lim->pl_lock); 887 if (ocore != defcorename) 888 free(ocore, M_TEMP); 889 890 done: 891 rw_exit(&ptmp->p_reflock); 892 PNBUF_PUT(cname); 893 return error; 894 } 895 896 /* 897 * sysctl helper routine for checking/setting a process's stop flags, 898 * one for fork and one for exec. 899 */ 900 static int 901 sysctl_proc_stop(SYSCTLFN_ARGS) 902 { 903 struct proc *ptmp; 904 int i, f, error = 0; 905 struct sysctlnode node; 906 907 if (namelen != 0) 908 return (EINVAL); 909 910 /* Find the process. Hold a reference (p_reflock), if found. */ 911 error = sysctl_proc_findproc(l, (pid_t)name[-2], &ptmp); 912 if (error) 913 return error; 914 915 /* XXX-elad */ 916 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 917 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 918 if (error) 919 goto out; 920 921 switch (rnode->sysctl_num) { 922 case PROC_PID_STOPFORK: 923 f = PS_STOPFORK; 924 break; 925 case PROC_PID_STOPEXEC: 926 f = PS_STOPEXEC; 927 break; 928 case PROC_PID_STOPEXIT: 929 f = PS_STOPEXIT; 930 break; 931 default: 932 error = EINVAL; 933 goto out; 934 } 935 936 i = (ptmp->p_flag & f) ? 1 : 0; 937 node = *rnode; 938 node.sysctl_data = &i; 939 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 940 if (error || newp == NULL) 941 goto out; 942 943 mutex_enter(ptmp->p_lock); 944 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 945 ptmp, KAUTH_ARG(f), NULL, NULL); 946 if (!error) { 947 if (i) { 948 ptmp->p_sflag |= f; 949 } else { 950 ptmp->p_sflag &= ~f; 951 } 952 } 953 mutex_exit(ptmp->p_lock); 954 out: 955 rw_exit(&ptmp->p_reflock); 956 return error; 957 } 958 959 /* 960 * sysctl helper routine for a process's rlimits as exposed by sysctl. 961 */ 962 static int 963 sysctl_proc_plimit(SYSCTLFN_ARGS) 964 { 965 struct proc *ptmp; 966 u_int limitno; 967 int which, error = 0; 968 struct rlimit alim; 969 struct sysctlnode node; 970 971 if (namelen != 0) 972 return (EINVAL); 973 974 which = name[-1]; 975 if (which != PROC_PID_LIMIT_TYPE_SOFT && 976 which != PROC_PID_LIMIT_TYPE_HARD) 977 return (EINVAL); 978 979 limitno = name[-2] - 1; 980 if (limitno >= RLIM_NLIMITS) 981 return (EINVAL); 982 983 if (name[-3] != PROC_PID_LIMIT) 984 return (EINVAL); 985 986 /* Find the process. Hold a reference (p_reflock), if found. */ 987 error = sysctl_proc_findproc(l, (pid_t)name[-4], &ptmp); 988 if (error) 989 return error; 990 991 /* XXX-elad */ 992 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 993 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 994 if (error) 995 goto out; 996 997 /* Check if we can view limits. */ 998 if (newp == NULL) { 999 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 1000 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_GET), &alim, 1001 KAUTH_ARG(which)); 1002 if (error) 1003 goto out; 1004 } 1005 1006 node = *rnode; 1007 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 1008 if (which == PROC_PID_LIMIT_TYPE_HARD) 1009 node.sysctl_data = &alim.rlim_max; 1010 else 1011 node.sysctl_data = &alim.rlim_cur; 1012 1013 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1014 if (error || newp == NULL) { 1015 goto out; 1016 } 1017 error = dosetrlimit(l, ptmp, limitno, &alim); 1018 out: 1019 rw_exit(&ptmp->p_reflock); 1020 return error; 1021 } 1022 1023 static struct sysctllog *proc_sysctllog; 1024 1025 /* 1026 * and finally, the actually glue that sticks it to the tree 1027 */ 1028 static void 1029 sysctl_proc_setup() 1030 { 1031 1032 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1033 CTLFLAG_PERMANENT, 1034 CTLTYPE_NODE, "proc", NULL, 1035 NULL, 0, NULL, 0, 1036 CTL_PROC, CTL_EOL); 1037 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1038 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 1039 CTLTYPE_NODE, "curproc", 1040 SYSCTL_DESCR("Per-process settings"), 1041 NULL, 0, NULL, 0, 1042 CTL_PROC, PROC_CURPROC, CTL_EOL); 1043 1044 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1045 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1046 CTLTYPE_STRING, "corename", 1047 SYSCTL_DESCR("Core file name"), 1048 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 1049 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 1050 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1051 CTLFLAG_PERMANENT, 1052 CTLTYPE_NODE, "rlimit", 1053 SYSCTL_DESCR("Process limits"), 1054 NULL, 0, NULL, 0, 1055 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 1056 1057 #define create_proc_plimit(s, n) do { \ 1058 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1059 CTLFLAG_PERMANENT, \ 1060 CTLTYPE_NODE, s, \ 1061 SYSCTL_DESCR("Process " s " limits"), \ 1062 NULL, 0, NULL, 0, \ 1063 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1064 CTL_EOL); \ 1065 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1066 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1067 CTLTYPE_QUAD, "soft", \ 1068 SYSCTL_DESCR("Process soft " s " limit"), \ 1069 sysctl_proc_plimit, 0, NULL, 0, \ 1070 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1071 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 1072 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1073 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1074 CTLTYPE_QUAD, "hard", \ 1075 SYSCTL_DESCR("Process hard " s " limit"), \ 1076 sysctl_proc_plimit, 0, NULL, 0, \ 1077 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1078 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 1079 } while (0/*CONSTCOND*/) 1080 1081 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 1082 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 1083 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 1084 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 1085 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 1086 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 1087 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 1088 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 1089 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 1090 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 1091 create_proc_plimit("vmemoryuse", PROC_PID_LIMIT_AS); 1092 1093 #undef create_proc_plimit 1094 1095 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1096 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1097 CTLTYPE_INT, "stopfork", 1098 SYSCTL_DESCR("Stop process at fork(2)"), 1099 sysctl_proc_stop, 0, NULL, 0, 1100 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 1101 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1102 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1103 CTLTYPE_INT, "stopexec", 1104 SYSCTL_DESCR("Stop process at execve(2)"), 1105 sysctl_proc_stop, 0, NULL, 0, 1106 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 1107 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1108 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1109 CTLTYPE_INT, "stopexit", 1110 SYSCTL_DESCR("Stop process before completing exit"), 1111 sysctl_proc_stop, 0, NULL, 0, 1112 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1113 } 1114