1 /* $NetBSD: kern_resource.c,v 1.164 2011/05/14 17:57:05 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 1.164 2011/05/14 17:57:05 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/kmem.h> 48 #include <sys/namei.h> 49 #include <sys/pool.h> 50 #include <sys/proc.h> 51 #include <sys/sysctl.h> 52 #include <sys/timevar.h> 53 #include <sys/kauth.h> 54 #include <sys/atomic.h> 55 #include <sys/mount.h> 56 #include <sys/syscallargs.h> 57 #include <sys/atomic.h> 58 59 #include <uvm/uvm_extern.h> 60 61 /* 62 * Maximum process data and stack limits. 63 * They are variables so they are patchable. 64 */ 65 const rlim_t maxdmap = MAXDSIZ; 66 const rlim_t maxsmap = MAXSSIZ; 67 68 static pool_cache_t plimit_cache __read_mostly; 69 static pool_cache_t pstats_cache __read_mostly; 70 71 static kauth_listener_t resource_listener; 72 static struct sysctllog *proc_sysctllog; 73 74 static int donice(struct lwp *, struct proc *, int); 75 static void sysctl_proc_setup(void); 76 77 static int 78 resource_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 79 void *arg0, void *arg1, void *arg2, void *arg3) 80 { 81 struct proc *p; 82 int result; 83 84 result = KAUTH_RESULT_DEFER; 85 p = arg0; 86 87 switch (action) { 88 case KAUTH_PROCESS_NICE: 89 if (kauth_cred_geteuid(cred) != kauth_cred_geteuid(p->p_cred) && 90 kauth_cred_getuid(cred) != kauth_cred_geteuid(p->p_cred)) { 91 break; 92 } 93 94 if ((u_long)arg1 >= p->p_nice) 95 result = KAUTH_RESULT_ALLOW; 96 97 break; 98 99 case KAUTH_PROCESS_RLIMIT: { 100 enum kauth_process_req req; 101 102 req = (enum kauth_process_req)(unsigned long)arg1; 103 104 switch (req) { 105 case KAUTH_REQ_PROCESS_RLIMIT_GET: 106 result = KAUTH_RESULT_ALLOW; 107 break; 108 109 case KAUTH_REQ_PROCESS_RLIMIT_SET: { 110 struct rlimit *new_rlimit; 111 u_long which; 112 113 if ((p != curlwp->l_proc) && 114 (proc_uidmatch(cred, p->p_cred) != 0)) 115 break; 116 117 new_rlimit = arg2; 118 which = (u_long)arg3; 119 120 if (new_rlimit->rlim_max <= p->p_rlimit[which].rlim_max) 121 result = KAUTH_RESULT_ALLOW; 122 123 break; 124 } 125 126 default: 127 break; 128 } 129 130 break; 131 } 132 133 default: 134 break; 135 } 136 137 return result; 138 } 139 140 void 141 resource_init(void) 142 { 143 144 plimit_cache = pool_cache_init(sizeof(struct plimit), 0, 0, 0, 145 "plimitpl", NULL, IPL_NONE, NULL, NULL, NULL); 146 pstats_cache = pool_cache_init(sizeof(struct pstats), 0, 0, 0, 147 "pstatspl", NULL, IPL_NONE, NULL, NULL, NULL); 148 149 resource_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 150 resource_listener_cb, NULL); 151 152 sysctl_proc_setup(); 153 } 154 155 /* 156 * Resource controls and accounting. 157 */ 158 159 int 160 sys_getpriority(struct lwp *l, const struct sys_getpriority_args *uap, 161 register_t *retval) 162 { 163 /* { 164 syscallarg(int) which; 165 syscallarg(id_t) who; 166 } */ 167 struct proc *curp = l->l_proc, *p; 168 id_t who = SCARG(uap, who); 169 int low = NZERO + PRIO_MAX + 1; 170 171 mutex_enter(proc_lock); 172 switch (SCARG(uap, which)) { 173 case PRIO_PROCESS: 174 p = who ? proc_find(who) : curp;; 175 if (p != NULL) 176 low = p->p_nice; 177 break; 178 179 case PRIO_PGRP: { 180 struct pgrp *pg; 181 182 if (who == 0) 183 pg = curp->p_pgrp; 184 else if ((pg = pgrp_find(who)) == NULL) 185 break; 186 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 187 if (p->p_nice < low) 188 low = p->p_nice; 189 } 190 break; 191 } 192 193 case PRIO_USER: 194 if (who == 0) 195 who = (int)kauth_cred_geteuid(l->l_cred); 196 PROCLIST_FOREACH(p, &allproc) { 197 mutex_enter(p->p_lock); 198 if (kauth_cred_geteuid(p->p_cred) == 199 (uid_t)who && p->p_nice < low) 200 low = p->p_nice; 201 mutex_exit(p->p_lock); 202 } 203 break; 204 205 default: 206 mutex_exit(proc_lock); 207 return EINVAL; 208 } 209 mutex_exit(proc_lock); 210 211 if (low == NZERO + PRIO_MAX + 1) { 212 return ESRCH; 213 } 214 *retval = low - NZERO; 215 return 0; 216 } 217 218 int 219 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap, 220 register_t *retval) 221 { 222 /* { 223 syscallarg(int) which; 224 syscallarg(id_t) who; 225 syscallarg(int) prio; 226 } */ 227 struct proc *curp = l->l_proc, *p; 228 id_t who = SCARG(uap, who); 229 int found = 0, error = 0; 230 231 mutex_enter(proc_lock); 232 switch (SCARG(uap, which)) { 233 case PRIO_PROCESS: 234 p = who ? proc_find(who) : curp; 235 if (p != NULL) { 236 mutex_enter(p->p_lock); 237 found++; 238 error = donice(l, p, SCARG(uap, prio)); 239 mutex_exit(p->p_lock); 240 } 241 break; 242 243 case PRIO_PGRP: { 244 struct pgrp *pg; 245 246 if (who == 0) 247 pg = curp->p_pgrp; 248 else if ((pg = pgrp_find(who)) == NULL) 249 break; 250 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 251 mutex_enter(p->p_lock); 252 found++; 253 error = donice(l, p, SCARG(uap, prio)); 254 mutex_exit(p->p_lock); 255 if (error) 256 break; 257 } 258 break; 259 } 260 261 case PRIO_USER: 262 if (who == 0) 263 who = (int)kauth_cred_geteuid(l->l_cred); 264 PROCLIST_FOREACH(p, &allproc) { 265 mutex_enter(p->p_lock); 266 if (kauth_cred_geteuid(p->p_cred) == 267 (uid_t)SCARG(uap, who)) { 268 found++; 269 error = donice(l, p, SCARG(uap, prio)); 270 } 271 mutex_exit(p->p_lock); 272 if (error) 273 break; 274 } 275 break; 276 277 default: 278 mutex_exit(proc_lock); 279 return EINVAL; 280 } 281 mutex_exit(proc_lock); 282 283 return (found == 0) ? ESRCH : error; 284 } 285 286 /* 287 * Renice a process. 288 * 289 * Call with the target process' credentials locked. 290 */ 291 static int 292 donice(struct lwp *l, struct proc *chgp, int n) 293 { 294 kauth_cred_t cred = l->l_cred; 295 296 KASSERT(mutex_owned(chgp->p_lock)); 297 298 if (kauth_cred_geteuid(cred) && kauth_cred_getuid(cred) && 299 kauth_cred_geteuid(cred) != kauth_cred_geteuid(chgp->p_cred) && 300 kauth_cred_getuid(cred) != kauth_cred_geteuid(chgp->p_cred)) 301 return EPERM; 302 303 if (n > PRIO_MAX) { 304 n = PRIO_MAX; 305 } 306 if (n < PRIO_MIN) { 307 n = PRIO_MIN; 308 } 309 n += NZERO; 310 311 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 312 KAUTH_ARG(n), NULL, NULL)) { 313 return EACCES; 314 } 315 316 sched_nice(chgp, n); 317 return 0; 318 } 319 320 int 321 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap, 322 register_t *retval) 323 { 324 /* { 325 syscallarg(int) which; 326 syscallarg(const struct rlimit *) rlp; 327 } */ 328 int error, which = SCARG(uap, which); 329 struct rlimit alim; 330 331 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 332 if (error) { 333 return error; 334 } 335 return dosetrlimit(l, l->l_proc, which, &alim); 336 } 337 338 int 339 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 340 { 341 struct rlimit *alimp; 342 int error; 343 344 if ((u_int)which >= RLIM_NLIMITS) 345 return EINVAL; 346 347 if (limp->rlim_cur > limp->rlim_max) { 348 /* 349 * This is programming error. According to SUSv2, we should 350 * return error in this case. 351 */ 352 return EINVAL; 353 } 354 355 alimp = &p->p_rlimit[which]; 356 /* if we don't change the value, no need to limcopy() */ 357 if (limp->rlim_cur == alimp->rlim_cur && 358 limp->rlim_max == alimp->rlim_max) 359 return 0; 360 361 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 362 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_SET), limp, KAUTH_ARG(which)); 363 if (error) 364 return error; 365 366 lim_privatise(p); 367 /* p->p_limit is now unchangeable */ 368 alimp = &p->p_rlimit[which]; 369 370 switch (which) { 371 372 case RLIMIT_DATA: 373 if (limp->rlim_cur > maxdmap) 374 limp->rlim_cur = maxdmap; 375 if (limp->rlim_max > maxdmap) 376 limp->rlim_max = maxdmap; 377 break; 378 379 case RLIMIT_STACK: 380 if (limp->rlim_cur > maxsmap) 381 limp->rlim_cur = maxsmap; 382 if (limp->rlim_max > maxsmap) 383 limp->rlim_max = maxsmap; 384 385 /* 386 * Return EINVAL if the new stack size limit is lower than 387 * current usage. Otherwise, the process would get SIGSEGV the 388 * moment it would try to access anything on it's current stack. 389 * This conforms to SUSv2. 390 */ 391 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE || 392 limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 393 return EINVAL; 394 } 395 396 /* 397 * Stack is allocated to the max at exec time with 398 * only "rlim_cur" bytes accessible (In other words, 399 * allocates stack dividing two contiguous regions at 400 * "rlim_cur" bytes boundary). 401 * 402 * Since allocation is done in terms of page, roundup 403 * "rlim_cur" (otherwise, contiguous regions 404 * overlap). If stack limit is going up make more 405 * accessible, if going down make inaccessible. 406 */ 407 limp->rlim_cur = round_page(limp->rlim_cur); 408 if (limp->rlim_cur != alimp->rlim_cur) { 409 vaddr_t addr; 410 vsize_t size; 411 vm_prot_t prot; 412 413 if (limp->rlim_cur > alimp->rlim_cur) { 414 prot = VM_PROT_READ | VM_PROT_WRITE; 415 size = limp->rlim_cur - alimp->rlim_cur; 416 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 417 limp->rlim_cur; 418 } else { 419 prot = VM_PROT_NONE; 420 size = alimp->rlim_cur - limp->rlim_cur; 421 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 422 alimp->rlim_cur; 423 } 424 (void) uvm_map_protect(&p->p_vmspace->vm_map, 425 addr, addr+size, prot, false); 426 } 427 break; 428 429 case RLIMIT_NOFILE: 430 if (limp->rlim_cur > maxfiles) 431 limp->rlim_cur = maxfiles; 432 if (limp->rlim_max > maxfiles) 433 limp->rlim_max = maxfiles; 434 break; 435 436 case RLIMIT_NPROC: 437 if (limp->rlim_cur > maxproc) 438 limp->rlim_cur = maxproc; 439 if (limp->rlim_max > maxproc) 440 limp->rlim_max = maxproc; 441 break; 442 } 443 444 mutex_enter(&p->p_limit->pl_lock); 445 *alimp = *limp; 446 mutex_exit(&p->p_limit->pl_lock); 447 return 0; 448 } 449 450 int 451 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, 452 register_t *retval) 453 { 454 /* { 455 syscallarg(int) which; 456 syscallarg(struct rlimit *) rlp; 457 } */ 458 struct proc *p = l->l_proc; 459 int which = SCARG(uap, which); 460 struct rlimit rl; 461 462 if ((u_int)which >= RLIM_NLIMITS) 463 return EINVAL; 464 465 mutex_enter(p->p_lock); 466 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 467 mutex_exit(p->p_lock); 468 469 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 470 } 471 472 /* 473 * Transform the running time and tick information in proc p into user, 474 * system, and interrupt time usage. 475 * 476 * Should be called with p->p_lock held unless called from exit1(). 477 */ 478 void 479 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 480 struct timeval *ip, struct timeval *rp) 481 { 482 uint64_t u, st, ut, it, tot; 483 struct lwp *l; 484 struct bintime tm; 485 struct timeval tv; 486 487 mutex_spin_enter(&p->p_stmutex); 488 st = p->p_sticks; 489 ut = p->p_uticks; 490 it = p->p_iticks; 491 mutex_spin_exit(&p->p_stmutex); 492 493 tm = p->p_rtime; 494 495 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 496 lwp_lock(l); 497 bintime_add(&tm, &l->l_rtime); 498 if ((l->l_pflag & LP_RUNNING) != 0) { 499 struct bintime diff; 500 /* 501 * Adjust for the current time slice. This is 502 * actually fairly important since the error 503 * here is on the order of a time quantum, 504 * which is much greater than the sampling 505 * error. 506 */ 507 binuptime(&diff); 508 bintime_sub(&diff, &l->l_stime); 509 bintime_add(&tm, &diff); 510 } 511 lwp_unlock(l); 512 } 513 514 tot = st + ut + it; 515 bintime2timeval(&tm, &tv); 516 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 517 518 if (tot == 0) { 519 /* No ticks, so can't use to share time out, split 50-50 */ 520 st = ut = u / 2; 521 } else { 522 st = (u * st) / tot; 523 ut = (u * ut) / tot; 524 } 525 if (sp != NULL) { 526 sp->tv_sec = st / 1000000; 527 sp->tv_usec = st % 1000000; 528 } 529 if (up != NULL) { 530 up->tv_sec = ut / 1000000; 531 up->tv_usec = ut % 1000000; 532 } 533 if (ip != NULL) { 534 if (it != 0) 535 it = (u * it) / tot; 536 ip->tv_sec = it / 1000000; 537 ip->tv_usec = it % 1000000; 538 } 539 if (rp != NULL) { 540 *rp = tv; 541 } 542 } 543 544 int 545 sys___getrusage50(struct lwp *l, const struct sys___getrusage50_args *uap, 546 register_t *retval) 547 { 548 /* { 549 syscallarg(int) who; 550 syscallarg(struct rusage *) rusage; 551 } */ 552 struct rusage ru; 553 struct proc *p = l->l_proc; 554 555 switch (SCARG(uap, who)) { 556 case RUSAGE_SELF: 557 mutex_enter(p->p_lock); 558 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 559 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 560 rulwps(p, &ru); 561 mutex_exit(p->p_lock); 562 break; 563 564 case RUSAGE_CHILDREN: 565 mutex_enter(p->p_lock); 566 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 567 mutex_exit(p->p_lock); 568 break; 569 570 default: 571 return EINVAL; 572 } 573 574 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 575 } 576 577 void 578 ruadd(struct rusage *ru, struct rusage *ru2) 579 { 580 long *ip, *ip2; 581 int i; 582 583 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 584 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 585 if (ru->ru_maxrss < ru2->ru_maxrss) 586 ru->ru_maxrss = ru2->ru_maxrss; 587 ip = &ru->ru_first; ip2 = &ru2->ru_first; 588 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 589 *ip++ += *ip2++; 590 } 591 592 void 593 rulwps(proc_t *p, struct rusage *ru) 594 { 595 lwp_t *l; 596 597 KASSERT(mutex_owned(p->p_lock)); 598 599 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 600 ruadd(ru, &l->l_ru); 601 ru->ru_nvcsw += (l->l_ncsw - l->l_nivcsw); 602 ru->ru_nivcsw += l->l_nivcsw; 603 } 604 } 605 606 /* 607 * lim_copy: make a copy of the plimit structure. 608 * 609 * We use copy-on-write after fork, and copy when a limit is changed. 610 */ 611 struct plimit * 612 lim_copy(struct plimit *lim) 613 { 614 struct plimit *newlim; 615 char *corename; 616 size_t alen, len; 617 618 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 619 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 620 newlim->pl_writeable = false; 621 newlim->pl_refcnt = 1; 622 newlim->pl_sv_limit = NULL; 623 624 mutex_enter(&lim->pl_lock); 625 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 626 sizeof(struct rlimit) * RLIM_NLIMITS); 627 628 /* 629 * Note: the common case is a use of default core name. 630 */ 631 alen = 0; 632 corename = NULL; 633 for (;;) { 634 if (lim->pl_corename == defcorename) { 635 newlim->pl_corename = defcorename; 636 newlim->pl_cnlen = 0; 637 break; 638 } 639 len = lim->pl_cnlen; 640 if (len == alen) { 641 newlim->pl_corename = corename; 642 newlim->pl_cnlen = len; 643 memcpy(corename, lim->pl_corename, len); 644 corename = NULL; 645 break; 646 } 647 mutex_exit(&lim->pl_lock); 648 if (corename) { 649 kmem_free(corename, alen); 650 } 651 alen = len; 652 corename = kmem_alloc(alen, KM_SLEEP); 653 mutex_enter(&lim->pl_lock); 654 } 655 mutex_exit(&lim->pl_lock); 656 657 if (corename) { 658 kmem_free(corename, alen); 659 } 660 return newlim; 661 } 662 663 void 664 lim_addref(struct plimit *lim) 665 { 666 atomic_inc_uint(&lim->pl_refcnt); 667 } 668 669 /* 670 * lim_privatise: give a process its own private plimit structure. 671 */ 672 void 673 lim_privatise(proc_t *p) 674 { 675 struct plimit *lim = p->p_limit, *newlim; 676 677 if (lim->pl_writeable) { 678 return; 679 } 680 681 newlim = lim_copy(lim); 682 683 mutex_enter(p->p_lock); 684 if (p->p_limit->pl_writeable) { 685 /* Other thread won the race. */ 686 mutex_exit(p->p_lock); 687 lim_free(newlim); 688 return; 689 } 690 691 /* 692 * Since p->p_limit can be accessed without locked held, 693 * old limit structure must not be deleted yet. 694 */ 695 newlim->pl_sv_limit = p->p_limit; 696 newlim->pl_writeable = true; 697 p->p_limit = newlim; 698 mutex_exit(p->p_lock); 699 } 700 701 void 702 lim_setcorename(proc_t *p, char *name, size_t len) 703 { 704 struct plimit *lim; 705 char *oname; 706 size_t olen; 707 708 lim_privatise(p); 709 lim = p->p_limit; 710 711 mutex_enter(&lim->pl_lock); 712 oname = lim->pl_corename; 713 olen = lim->pl_cnlen; 714 lim->pl_corename = name; 715 lim->pl_cnlen = len; 716 mutex_exit(&lim->pl_lock); 717 718 if (oname != defcorename) { 719 kmem_free(oname, olen); 720 } 721 } 722 723 void 724 lim_free(struct plimit *lim) 725 { 726 struct plimit *sv_lim; 727 728 do { 729 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) { 730 return; 731 } 732 if (lim->pl_corename != defcorename) { 733 kmem_free(lim->pl_corename, lim->pl_cnlen); 734 } 735 sv_lim = lim->pl_sv_limit; 736 mutex_destroy(&lim->pl_lock); 737 pool_cache_put(plimit_cache, lim); 738 } while ((lim = sv_lim) != NULL); 739 } 740 741 struct pstats * 742 pstatscopy(struct pstats *ps) 743 { 744 struct pstats *nps; 745 size_t len; 746 747 nps = pool_cache_get(pstats_cache, PR_WAITOK); 748 749 len = (char *)&nps->pstat_endzero - (char *)&nps->pstat_startzero; 750 memset(&nps->pstat_startzero, 0, len); 751 752 len = (char *)&nps->pstat_endcopy - (char *)&nps->pstat_startcopy; 753 memcpy(&nps->pstat_startcopy, &ps->pstat_startcopy, len); 754 755 return nps; 756 } 757 758 void 759 pstatsfree(struct pstats *ps) 760 { 761 762 pool_cache_put(pstats_cache, ps); 763 } 764 765 /* 766 * sysctl_proc_findproc: a routine for sysctl proc subtree helpers that 767 * need to pick a valid process by PID. 768 * 769 * => Hold a reference on the process, on success. 770 */ 771 static int 772 sysctl_proc_findproc(lwp_t *l, pid_t pid, proc_t **p2) 773 { 774 proc_t *p; 775 int error; 776 777 if (pid == PROC_CURPROC) { 778 p = l->l_proc; 779 } else { 780 mutex_enter(proc_lock); 781 p = proc_find(pid); 782 if (p == NULL) { 783 mutex_exit(proc_lock); 784 return ESRCH; 785 } 786 } 787 error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY; 788 if (pid != PROC_CURPROC) { 789 mutex_exit(proc_lock); 790 } 791 *p2 = p; 792 return error; 793 } 794 795 /* 796 * sysctl_proc_corename: helper routine to get or set the core file name 797 * for a process specified by PID. 798 */ 799 static int 800 sysctl_proc_corename(SYSCTLFN_ARGS) 801 { 802 struct proc *p; 803 struct plimit *lim; 804 char *cnbuf, *cname; 805 struct sysctlnode node; 806 size_t len; 807 int error; 808 809 /* First, validate the request. */ 810 if (namelen != 0 || name[-1] != PROC_PID_CORENAME) 811 return EINVAL; 812 813 /* Find the process. Hold a reference (p_reflock), if found. */ 814 error = sysctl_proc_findproc(l, (pid_t)name[-2], &p); 815 if (error) 816 return error; 817 818 /* XXX-elad */ 819 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 820 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 821 if (error) { 822 rw_exit(&p->p_reflock); 823 return error; 824 } 825 826 cnbuf = PNBUF_GET(); 827 828 if (newp == NULL) { 829 /* Get case: copy the core name into the buffer. */ 830 error = kauth_authorize_process(l->l_cred, 831 KAUTH_PROCESS_CORENAME, p, 832 KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_GET), NULL, NULL); 833 if (error) { 834 goto done; 835 } 836 lim = p->p_limit; 837 mutex_enter(&lim->pl_lock); 838 strlcpy(cnbuf, lim->pl_corename, MAXPATHLEN); 839 mutex_exit(&lim->pl_lock); 840 } else { 841 /* Set case: just use the temporary buffer. */ 842 error = kauth_authorize_process(l->l_cred, 843 KAUTH_PROCESS_CORENAME, p, 844 KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_SET), cnbuf, NULL); 845 if (error) { 846 goto done; 847 } 848 } 849 850 node = *rnode; 851 node.sysctl_data = cnbuf; 852 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 853 854 /* Return if error, or if caller is only getting the core name. */ 855 if (error || newp == NULL) { 856 goto done; 857 } 858 859 /* 860 * Validate new core name. It must be either "core", "/core", 861 * or end in ".core". 862 */ 863 len = strlen(cnbuf); 864 if ((len < 4 || strcmp(cnbuf + len - 4, "core") != 0) || 865 (len > 4 && cnbuf[len - 5] != '/' && cnbuf[len - 5] != '.')) { 866 error = EINVAL; 867 goto done; 868 } 869 870 /* Allocate, copy and set the new core name for plimit structure. */ 871 cname = kmem_alloc(++len, KM_NOSLEEP); 872 if (cname == NULL) { 873 error = ENOMEM; 874 goto done; 875 } 876 memcpy(cname, cnbuf, len); 877 lim_setcorename(p, cname, len); 878 done: 879 rw_exit(&p->p_reflock); 880 PNBUF_PUT(cnbuf); 881 return error; 882 } 883 884 /* 885 * sysctl_proc_stop: helper routine for checking/setting the stop flags. 886 */ 887 static int 888 sysctl_proc_stop(SYSCTLFN_ARGS) 889 { 890 struct proc *p; 891 int isset, flag, error = 0; 892 struct sysctlnode node; 893 894 if (namelen != 0) 895 return EINVAL; 896 897 /* Find the process. Hold a reference (p_reflock), if found. */ 898 error = sysctl_proc_findproc(l, (pid_t)name[-2], &p); 899 if (error) 900 return error; 901 902 /* XXX-elad */ 903 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 904 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 905 if (error) { 906 goto out; 907 } 908 909 /* Determine the flag. */ 910 switch (rnode->sysctl_num) { 911 case PROC_PID_STOPFORK: 912 flag = PS_STOPFORK; 913 break; 914 case PROC_PID_STOPEXEC: 915 flag = PS_STOPEXEC; 916 break; 917 case PROC_PID_STOPEXIT: 918 flag = PS_STOPEXIT; 919 break; 920 default: 921 error = EINVAL; 922 goto out; 923 } 924 isset = (p->p_flag & flag) ? 1 : 0; 925 node = *rnode; 926 node.sysctl_data = &isset; 927 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 928 929 /* Return if error, or if callers is only getting the flag. */ 930 if (error || newp == NULL) { 931 goto out; 932 } 933 934 /* Check if caller can set the flags. */ 935 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 936 p, KAUTH_ARG(flag), NULL, NULL); 937 if (error) { 938 goto out; 939 } 940 mutex_enter(p->p_lock); 941 if (isset) { 942 p->p_sflag |= flag; 943 } else { 944 p->p_sflag &= ~flag; 945 } 946 mutex_exit(p->p_lock); 947 out: 948 rw_exit(&p->p_reflock); 949 return error; 950 } 951 952 /* 953 * sysctl_proc_plimit: helper routine to get/set rlimits of a process. 954 */ 955 static int 956 sysctl_proc_plimit(SYSCTLFN_ARGS) 957 { 958 struct proc *p; 959 u_int limitno; 960 int which, error = 0; 961 struct rlimit alim; 962 struct sysctlnode node; 963 964 if (namelen != 0) 965 return EINVAL; 966 967 which = name[-1]; 968 if (which != PROC_PID_LIMIT_TYPE_SOFT && 969 which != PROC_PID_LIMIT_TYPE_HARD) 970 return EINVAL; 971 972 limitno = name[-2] - 1; 973 if (limitno >= RLIM_NLIMITS) 974 return EINVAL; 975 976 if (name[-3] != PROC_PID_LIMIT) 977 return EINVAL; 978 979 /* Find the process. Hold a reference (p_reflock), if found. */ 980 error = sysctl_proc_findproc(l, (pid_t)name[-4], &p); 981 if (error) 982 return error; 983 984 /* XXX-elad */ 985 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 986 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 987 if (error) 988 goto out; 989 990 /* Check if caller can retrieve the limits. */ 991 if (newp == NULL) { 992 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 993 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_GET), &alim, 994 KAUTH_ARG(which)); 995 if (error) 996 goto out; 997 } 998 999 /* Retrieve the limits. */ 1000 node = *rnode; 1001 memcpy(&alim, &p->p_rlimit[limitno], sizeof(alim)); 1002 if (which == PROC_PID_LIMIT_TYPE_HARD) { 1003 node.sysctl_data = &alim.rlim_max; 1004 } else { 1005 node.sysctl_data = &alim.rlim_cur; 1006 } 1007 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1008 1009 /* Return if error, or if we are only retrieving the limits. */ 1010 if (error || newp == NULL) { 1011 goto out; 1012 } 1013 error = dosetrlimit(l, p, limitno, &alim); 1014 out: 1015 rw_exit(&p->p_reflock); 1016 return error; 1017 } 1018 1019 /* 1020 * Setup sysctl nodes. 1021 */ 1022 static void 1023 sysctl_proc_setup(void) 1024 { 1025 1026 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1027 CTLFLAG_PERMANENT, 1028 CTLTYPE_NODE, "proc", NULL, 1029 NULL, 0, NULL, 0, 1030 CTL_PROC, CTL_EOL); 1031 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1032 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 1033 CTLTYPE_NODE, "curproc", 1034 SYSCTL_DESCR("Per-process settings"), 1035 NULL, 0, NULL, 0, 1036 CTL_PROC, PROC_CURPROC, CTL_EOL); 1037 1038 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1039 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1040 CTLTYPE_STRING, "corename", 1041 SYSCTL_DESCR("Core file name"), 1042 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 1043 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 1044 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1045 CTLFLAG_PERMANENT, 1046 CTLTYPE_NODE, "rlimit", 1047 SYSCTL_DESCR("Process limits"), 1048 NULL, 0, NULL, 0, 1049 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 1050 1051 #define create_proc_plimit(s, n) do { \ 1052 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1053 CTLFLAG_PERMANENT, \ 1054 CTLTYPE_NODE, s, \ 1055 SYSCTL_DESCR("Process " s " limits"), \ 1056 NULL, 0, NULL, 0, \ 1057 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1058 CTL_EOL); \ 1059 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1060 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1061 CTLTYPE_QUAD, "soft", \ 1062 SYSCTL_DESCR("Process soft " s " limit"), \ 1063 sysctl_proc_plimit, 0, NULL, 0, \ 1064 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1065 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 1066 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1067 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1068 CTLTYPE_QUAD, "hard", \ 1069 SYSCTL_DESCR("Process hard " s " limit"), \ 1070 sysctl_proc_plimit, 0, NULL, 0, \ 1071 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1072 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 1073 } while (0/*CONSTCOND*/) 1074 1075 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 1076 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 1077 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 1078 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 1079 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 1080 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 1081 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 1082 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 1083 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 1084 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 1085 create_proc_plimit("vmemoryuse", PROC_PID_LIMIT_AS); 1086 1087 #undef create_proc_plimit 1088 1089 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1090 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1091 CTLTYPE_INT, "stopfork", 1092 SYSCTL_DESCR("Stop process at fork(2)"), 1093 sysctl_proc_stop, 0, NULL, 0, 1094 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 1095 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1096 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1097 CTLTYPE_INT, "stopexec", 1098 SYSCTL_DESCR("Stop process at execve(2)"), 1099 sysctl_proc_stop, 0, NULL, 0, 1100 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 1101 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1102 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1103 CTLTYPE_INT, "stopexit", 1104 SYSCTL_DESCR("Stop process before completing exit"), 1105 sysctl_proc_stop, 0, NULL, 0, 1106 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1107 } 1108