1 /* $NetBSD: kern_resource.c,v 1.174 2014/10/18 08:33:29 snj 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 1.174 2014/10/18 08:33:29 snj 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 rlim_t maxdmap = MAXDSIZ; 66 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 its 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 char *base, *tmp; 413 414 base = p->p_vmspace->vm_minsaddr; 415 if (limp->rlim_cur > alimp->rlim_cur) { 416 prot = VM_PROT_READ | VM_PROT_WRITE; 417 size = limp->rlim_cur - alimp->rlim_cur; 418 tmp = STACK_GROW(base, alimp->rlim_cur); 419 } else { 420 prot = VM_PROT_NONE; 421 size = alimp->rlim_cur - limp->rlim_cur; 422 tmp = STACK_GROW(base, limp->rlim_cur); 423 } 424 addr = (vaddr_t)STACK_ALLOC(tmp, size); 425 (void) uvm_map_protect(&p->p_vmspace->vm_map, 426 addr, addr + size, prot, false); 427 } 428 break; 429 430 case RLIMIT_NOFILE: 431 if (limp->rlim_cur > maxfiles) 432 limp->rlim_cur = maxfiles; 433 if (limp->rlim_max > maxfiles) 434 limp->rlim_max = maxfiles; 435 break; 436 437 case RLIMIT_NPROC: 438 if (limp->rlim_cur > maxproc) 439 limp->rlim_cur = maxproc; 440 if (limp->rlim_max > maxproc) 441 limp->rlim_max = maxproc; 442 break; 443 444 case RLIMIT_NTHR: 445 if (limp->rlim_cur > maxlwp) 446 limp->rlim_cur = maxlwp; 447 if (limp->rlim_max > maxlwp) 448 limp->rlim_max = maxlwp; 449 break; 450 } 451 452 mutex_enter(&p->p_limit->pl_lock); 453 *alimp = *limp; 454 mutex_exit(&p->p_limit->pl_lock); 455 return 0; 456 } 457 458 int 459 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, 460 register_t *retval) 461 { 462 /* { 463 syscallarg(int) which; 464 syscallarg(struct rlimit *) rlp; 465 } */ 466 struct proc *p = l->l_proc; 467 int which = SCARG(uap, which); 468 struct rlimit rl; 469 470 if ((u_int)which >= RLIM_NLIMITS) 471 return EINVAL; 472 473 mutex_enter(p->p_lock); 474 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 475 mutex_exit(p->p_lock); 476 477 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 478 } 479 480 /* 481 * Transform the running time and tick information in proc p into user, 482 * system, and interrupt time usage. 483 * 484 * Should be called with p->p_lock held unless called from exit1(). 485 */ 486 void 487 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 488 struct timeval *ip, struct timeval *rp) 489 { 490 uint64_t u, st, ut, it, tot; 491 struct lwp *l; 492 struct bintime tm; 493 struct timeval tv; 494 495 KASSERT(p->p_stat == SDEAD || mutex_owned(p->p_lock)); 496 497 mutex_spin_enter(&p->p_stmutex); 498 st = p->p_sticks; 499 ut = p->p_uticks; 500 it = p->p_iticks; 501 mutex_spin_exit(&p->p_stmutex); 502 503 tm = p->p_rtime; 504 505 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 506 lwp_lock(l); 507 bintime_add(&tm, &l->l_rtime); 508 if ((l->l_pflag & LP_RUNNING) != 0) { 509 struct bintime diff; 510 /* 511 * Adjust for the current time slice. This is 512 * actually fairly important since the error 513 * here is on the order of a time quantum, 514 * which is much greater than the sampling 515 * error. 516 */ 517 binuptime(&diff); 518 bintime_sub(&diff, &l->l_stime); 519 bintime_add(&tm, &diff); 520 } 521 lwp_unlock(l); 522 } 523 524 tot = st + ut + it; 525 bintime2timeval(&tm, &tv); 526 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 527 528 if (tot == 0) { 529 /* No ticks, so can't use to share time out, split 50-50 */ 530 st = ut = u / 2; 531 } else { 532 st = (u * st) / tot; 533 ut = (u * ut) / tot; 534 } 535 if (sp != NULL) { 536 sp->tv_sec = st / 1000000; 537 sp->tv_usec = st % 1000000; 538 } 539 if (up != NULL) { 540 up->tv_sec = ut / 1000000; 541 up->tv_usec = ut % 1000000; 542 } 543 if (ip != NULL) { 544 if (it != 0) 545 it = (u * it) / tot; 546 ip->tv_sec = it / 1000000; 547 ip->tv_usec = it % 1000000; 548 } 549 if (rp != NULL) { 550 *rp = tv; 551 } 552 } 553 554 int 555 sys___getrusage50(struct lwp *l, const struct sys___getrusage50_args *uap, 556 register_t *retval) 557 { 558 /* { 559 syscallarg(int) who; 560 syscallarg(struct rusage *) rusage; 561 } */ 562 int error; 563 struct rusage ru; 564 struct proc *p = l->l_proc; 565 566 error = getrusage1(p, SCARG(uap, who), &ru); 567 if (error != 0) 568 return error; 569 570 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 571 } 572 573 int 574 getrusage1(struct proc *p, int who, struct rusage *ru) { 575 576 switch (who) { 577 case RUSAGE_SELF: 578 mutex_enter(p->p_lock); 579 memcpy(ru, &p->p_stats->p_ru, sizeof(*ru)); 580 calcru(p, &ru->ru_utime, &ru->ru_stime, NULL, NULL); 581 rulwps(p, ru); 582 mutex_exit(p->p_lock); 583 break; 584 case RUSAGE_CHILDREN: 585 mutex_enter(p->p_lock); 586 memcpy(ru, &p->p_stats->p_cru, sizeof(*ru)); 587 mutex_exit(p->p_lock); 588 break; 589 default: 590 return EINVAL; 591 } 592 593 return 0; 594 } 595 596 void 597 ruadd(struct rusage *ru, struct rusage *ru2) 598 { 599 long *ip, *ip2; 600 int i; 601 602 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 603 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 604 if (ru->ru_maxrss < ru2->ru_maxrss) 605 ru->ru_maxrss = ru2->ru_maxrss; 606 ip = &ru->ru_first; ip2 = &ru2->ru_first; 607 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 608 *ip++ += *ip2++; 609 } 610 611 void 612 rulwps(proc_t *p, struct rusage *ru) 613 { 614 lwp_t *l; 615 616 KASSERT(mutex_owned(p->p_lock)); 617 618 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 619 ruadd(ru, &l->l_ru); 620 ru->ru_nvcsw += (l->l_ncsw - l->l_nivcsw); 621 ru->ru_nivcsw += l->l_nivcsw; 622 } 623 } 624 625 /* 626 * lim_copy: make a copy of the plimit structure. 627 * 628 * We use copy-on-write after fork, and copy when a limit is changed. 629 */ 630 struct plimit * 631 lim_copy(struct plimit *lim) 632 { 633 struct plimit *newlim; 634 char *corename; 635 size_t alen, len; 636 637 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 638 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 639 newlim->pl_writeable = false; 640 newlim->pl_refcnt = 1; 641 newlim->pl_sv_limit = NULL; 642 643 mutex_enter(&lim->pl_lock); 644 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 645 sizeof(struct rlimit) * RLIM_NLIMITS); 646 647 /* 648 * Note: the common case is a use of default core name. 649 */ 650 alen = 0; 651 corename = NULL; 652 for (;;) { 653 if (lim->pl_corename == defcorename) { 654 newlim->pl_corename = defcorename; 655 newlim->pl_cnlen = 0; 656 break; 657 } 658 len = lim->pl_cnlen; 659 if (len == alen) { 660 newlim->pl_corename = corename; 661 newlim->pl_cnlen = len; 662 memcpy(corename, lim->pl_corename, len); 663 corename = NULL; 664 break; 665 } 666 mutex_exit(&lim->pl_lock); 667 if (corename) { 668 kmem_free(corename, alen); 669 } 670 alen = len; 671 corename = kmem_alloc(alen, KM_SLEEP); 672 mutex_enter(&lim->pl_lock); 673 } 674 mutex_exit(&lim->pl_lock); 675 676 if (corename) { 677 kmem_free(corename, alen); 678 } 679 return newlim; 680 } 681 682 void 683 lim_addref(struct plimit *lim) 684 { 685 atomic_inc_uint(&lim->pl_refcnt); 686 } 687 688 /* 689 * lim_privatise: give a process its own private plimit structure. 690 */ 691 void 692 lim_privatise(proc_t *p) 693 { 694 struct plimit *lim = p->p_limit, *newlim; 695 696 if (lim->pl_writeable) { 697 return; 698 } 699 700 newlim = lim_copy(lim); 701 702 mutex_enter(p->p_lock); 703 if (p->p_limit->pl_writeable) { 704 /* Other thread won the race. */ 705 mutex_exit(p->p_lock); 706 lim_free(newlim); 707 return; 708 } 709 710 /* 711 * Since p->p_limit can be accessed without locked held, 712 * old limit structure must not be deleted yet. 713 */ 714 newlim->pl_sv_limit = p->p_limit; 715 newlim->pl_writeable = true; 716 p->p_limit = newlim; 717 mutex_exit(p->p_lock); 718 } 719 720 void 721 lim_setcorename(proc_t *p, char *name, size_t len) 722 { 723 struct plimit *lim; 724 char *oname; 725 size_t olen; 726 727 lim_privatise(p); 728 lim = p->p_limit; 729 730 mutex_enter(&lim->pl_lock); 731 oname = lim->pl_corename; 732 olen = lim->pl_cnlen; 733 lim->pl_corename = name; 734 lim->pl_cnlen = len; 735 mutex_exit(&lim->pl_lock); 736 737 if (oname != defcorename) { 738 kmem_free(oname, olen); 739 } 740 } 741 742 void 743 lim_free(struct plimit *lim) 744 { 745 struct plimit *sv_lim; 746 747 do { 748 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) { 749 return; 750 } 751 if (lim->pl_corename != defcorename) { 752 kmem_free(lim->pl_corename, lim->pl_cnlen); 753 } 754 sv_lim = lim->pl_sv_limit; 755 mutex_destroy(&lim->pl_lock); 756 pool_cache_put(plimit_cache, lim); 757 } while ((lim = sv_lim) != NULL); 758 } 759 760 struct pstats * 761 pstatscopy(struct pstats *ps) 762 { 763 struct pstats *nps; 764 size_t len; 765 766 nps = pool_cache_get(pstats_cache, PR_WAITOK); 767 768 len = (char *)&nps->pstat_endzero - (char *)&nps->pstat_startzero; 769 memset(&nps->pstat_startzero, 0, len); 770 771 len = (char *)&nps->pstat_endcopy - (char *)&nps->pstat_startcopy; 772 memcpy(&nps->pstat_startcopy, &ps->pstat_startcopy, len); 773 774 return nps; 775 } 776 777 void 778 pstatsfree(struct pstats *ps) 779 { 780 781 pool_cache_put(pstats_cache, ps); 782 } 783 784 /* 785 * sysctl_proc_findproc: a routine for sysctl proc subtree helpers that 786 * need to pick a valid process by PID. 787 * 788 * => Hold a reference on the process, on success. 789 */ 790 static int 791 sysctl_proc_findproc(lwp_t *l, pid_t pid, proc_t **p2) 792 { 793 proc_t *p; 794 int error; 795 796 if (pid == PROC_CURPROC) { 797 p = l->l_proc; 798 } else { 799 mutex_enter(proc_lock); 800 p = proc_find(pid); 801 if (p == NULL) { 802 mutex_exit(proc_lock); 803 return ESRCH; 804 } 805 } 806 error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY; 807 if (pid != PROC_CURPROC) { 808 mutex_exit(proc_lock); 809 } 810 *p2 = p; 811 return error; 812 } 813 814 /* 815 * sysctl_proc_corename: helper routine to get or set the core file name 816 * for a process specified by PID. 817 */ 818 static int 819 sysctl_proc_corename(SYSCTLFN_ARGS) 820 { 821 struct proc *p; 822 struct plimit *lim; 823 char *cnbuf, *cname; 824 struct sysctlnode node; 825 size_t len; 826 int error; 827 828 /* First, validate the request. */ 829 if (namelen != 0 || name[-1] != PROC_PID_CORENAME) 830 return EINVAL; 831 832 /* Find the process. Hold a reference (p_reflock), if found. */ 833 error = sysctl_proc_findproc(l, (pid_t)name[-2], &p); 834 if (error) 835 return error; 836 837 /* XXX-elad */ 838 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 839 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 840 if (error) { 841 rw_exit(&p->p_reflock); 842 return error; 843 } 844 845 cnbuf = PNBUF_GET(); 846 847 if (oldp) { 848 /* Get case: copy the core name into the buffer. */ 849 error = kauth_authorize_process(l->l_cred, 850 KAUTH_PROCESS_CORENAME, p, 851 KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_GET), NULL, NULL); 852 if (error) { 853 goto done; 854 } 855 lim = p->p_limit; 856 mutex_enter(&lim->pl_lock); 857 strlcpy(cnbuf, lim->pl_corename, MAXPATHLEN); 858 mutex_exit(&lim->pl_lock); 859 } 860 861 node = *rnode; 862 node.sysctl_data = cnbuf; 863 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 864 865 /* Return if error, or if caller is only getting the core name. */ 866 if (error || newp == NULL) { 867 goto done; 868 } 869 870 /* 871 * Set case. Check permission and then validate new core name. 872 * It must be either "core", "/core", or end in ".core". 873 */ 874 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 875 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_SET), cnbuf, NULL); 876 if (error) { 877 goto done; 878 } 879 len = strlen(cnbuf); 880 if ((len < 4 || strcmp(cnbuf + len - 4, "core") != 0) || 881 (len > 4 && cnbuf[len - 5] != '/' && cnbuf[len - 5] != '.')) { 882 error = EINVAL; 883 goto done; 884 } 885 886 /* Allocate, copy and set the new core name for plimit structure. */ 887 cname = kmem_alloc(++len, KM_NOSLEEP); 888 if (cname == NULL) { 889 error = ENOMEM; 890 goto done; 891 } 892 memcpy(cname, cnbuf, len); 893 lim_setcorename(p, cname, len); 894 done: 895 rw_exit(&p->p_reflock); 896 PNBUF_PUT(cnbuf); 897 return error; 898 } 899 900 /* 901 * sysctl_proc_stop: helper routine for checking/setting the stop flags. 902 */ 903 static int 904 sysctl_proc_stop(SYSCTLFN_ARGS) 905 { 906 struct proc *p; 907 int isset, flag, error = 0; 908 struct sysctlnode node; 909 910 if (namelen != 0) 911 return EINVAL; 912 913 /* Find the process. Hold a reference (p_reflock), if found. */ 914 error = sysctl_proc_findproc(l, (pid_t)name[-2], &p); 915 if (error) 916 return error; 917 918 /* XXX-elad */ 919 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 920 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 921 if (error) { 922 goto out; 923 } 924 925 /* Determine the flag. */ 926 switch (rnode->sysctl_num) { 927 case PROC_PID_STOPFORK: 928 flag = PS_STOPFORK; 929 break; 930 case PROC_PID_STOPEXEC: 931 flag = PS_STOPEXEC; 932 break; 933 case PROC_PID_STOPEXIT: 934 flag = PS_STOPEXIT; 935 break; 936 default: 937 error = EINVAL; 938 goto out; 939 } 940 isset = (p->p_flag & flag) ? 1 : 0; 941 node = *rnode; 942 node.sysctl_data = &isset; 943 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 944 945 /* Return if error, or if callers is only getting the flag. */ 946 if (error || newp == NULL) { 947 goto out; 948 } 949 950 /* Check if caller can set the flags. */ 951 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 952 p, KAUTH_ARG(flag), NULL, NULL); 953 if (error) { 954 goto out; 955 } 956 mutex_enter(p->p_lock); 957 if (isset) { 958 p->p_sflag |= flag; 959 } else { 960 p->p_sflag &= ~flag; 961 } 962 mutex_exit(p->p_lock); 963 out: 964 rw_exit(&p->p_reflock); 965 return error; 966 } 967 968 /* 969 * sysctl_proc_plimit: helper routine to get/set rlimits of a process. 970 */ 971 static int 972 sysctl_proc_plimit(SYSCTLFN_ARGS) 973 { 974 struct proc *p; 975 u_int limitno; 976 int which, error = 0; 977 struct rlimit alim; 978 struct sysctlnode node; 979 980 if (namelen != 0) 981 return EINVAL; 982 983 which = name[-1]; 984 if (which != PROC_PID_LIMIT_TYPE_SOFT && 985 which != PROC_PID_LIMIT_TYPE_HARD) 986 return EINVAL; 987 988 limitno = name[-2] - 1; 989 if (limitno >= RLIM_NLIMITS) 990 return EINVAL; 991 992 if (name[-3] != PROC_PID_LIMIT) 993 return EINVAL; 994 995 /* Find the process. Hold a reference (p_reflock), if found. */ 996 error = sysctl_proc_findproc(l, (pid_t)name[-4], &p); 997 if (error) 998 return error; 999 1000 /* XXX-elad */ 1001 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 1002 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 1003 if (error) 1004 goto out; 1005 1006 /* Check if caller can retrieve the limits. */ 1007 if (newp == NULL) { 1008 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 1009 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_GET), &alim, 1010 KAUTH_ARG(which)); 1011 if (error) 1012 goto out; 1013 } 1014 1015 /* Retrieve the limits. */ 1016 node = *rnode; 1017 memcpy(&alim, &p->p_rlimit[limitno], sizeof(alim)); 1018 if (which == PROC_PID_LIMIT_TYPE_HARD) { 1019 node.sysctl_data = &alim.rlim_max; 1020 } else { 1021 node.sysctl_data = &alim.rlim_cur; 1022 } 1023 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1024 1025 /* Return if error, or if we are only retrieving the limits. */ 1026 if (error || newp == NULL) { 1027 goto out; 1028 } 1029 error = dosetrlimit(l, p, limitno, &alim); 1030 out: 1031 rw_exit(&p->p_reflock); 1032 return error; 1033 } 1034 1035 /* 1036 * Setup sysctl nodes. 1037 */ 1038 static void 1039 sysctl_proc_setup(void) 1040 { 1041 1042 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1043 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 1044 CTLTYPE_NODE, "curproc", 1045 SYSCTL_DESCR("Per-process settings"), 1046 NULL, 0, NULL, 0, 1047 CTL_PROC, PROC_CURPROC, CTL_EOL); 1048 1049 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1050 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1051 CTLTYPE_STRING, "corename", 1052 SYSCTL_DESCR("Core file name"), 1053 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 1054 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 1055 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1056 CTLFLAG_PERMANENT, 1057 CTLTYPE_NODE, "rlimit", 1058 SYSCTL_DESCR("Process limits"), 1059 NULL, 0, NULL, 0, 1060 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 1061 1062 #define create_proc_plimit(s, n) do { \ 1063 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1064 CTLFLAG_PERMANENT, \ 1065 CTLTYPE_NODE, s, \ 1066 SYSCTL_DESCR("Process " s " limits"), \ 1067 NULL, 0, NULL, 0, \ 1068 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1069 CTL_EOL); \ 1070 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1071 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1072 CTLTYPE_QUAD, "soft", \ 1073 SYSCTL_DESCR("Process soft " s " limit"), \ 1074 sysctl_proc_plimit, 0, NULL, 0, \ 1075 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1076 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 1077 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1078 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1079 CTLTYPE_QUAD, "hard", \ 1080 SYSCTL_DESCR("Process hard " s " limit"), \ 1081 sysctl_proc_plimit, 0, NULL, 0, \ 1082 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1083 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 1084 } while (0/*CONSTCOND*/) 1085 1086 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 1087 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 1088 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 1089 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 1090 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 1091 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 1092 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 1093 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 1094 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 1095 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 1096 create_proc_plimit("vmemoryuse", PROC_PID_LIMIT_AS); 1097 create_proc_plimit("maxlwp", PROC_PID_LIMIT_NTHR); 1098 1099 #undef create_proc_plimit 1100 1101 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1102 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1103 CTLTYPE_INT, "stopfork", 1104 SYSCTL_DESCR("Stop process at fork(2)"), 1105 sysctl_proc_stop, 0, NULL, 0, 1106 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 1107 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1108 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1109 CTLTYPE_INT, "stopexec", 1110 SYSCTL_DESCR("Stop process at execve(2)"), 1111 sysctl_proc_stop, 0, NULL, 0, 1112 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 1113 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1114 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1115 CTLTYPE_INT, "stopexit", 1116 SYSCTL_DESCR("Stop process before completing exit"), 1117 sysctl_proc_stop, 0, NULL, 0, 1118 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1119 } 1120