1 /* $NetBSD: kern_resource.c,v 1.142 2008/05/31 21:26:01 ad 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 1.142 2008/05/31 21:26:01 ad 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 SLIST_HEAD(uihashhead, uidinfo) *uihashtbl; 70 static u_long uihash; 71 72 #define UIHASH(uid) (&uihashtbl[(uid) & uihash]) 73 74 static pool_cache_t plimit_cache; 75 static pool_cache_t pstats_cache; 76 77 void 78 resource_init(void) 79 { 80 /* 81 * In case of MP system, SLIST_FOREACH would force a cache line 82 * write-back for every modified 'uidinfo', thus we try to keep the 83 * lists short. 84 */ 85 const u_int uihash_sz = (maxproc > 1 ? 1024 : 64); 86 87 plimit_cache = pool_cache_init(sizeof(struct plimit), 0, 0, 0, 88 "plimitpl", NULL, IPL_NONE, NULL, NULL, NULL); 89 pstats_cache = pool_cache_init(sizeof(struct pstats), 0, 0, 0, 90 "pstatspl", NULL, IPL_NONE, NULL, NULL, NULL); 91 uihashtbl = hashinit(uihash_sz, HASH_SLIST, true, &uihash); 92 } 93 94 /* 95 * Resource controls and accounting. 96 */ 97 98 int 99 sys_getpriority(struct lwp *l, const struct sys_getpriority_args *uap, 100 register_t *retval) 101 { 102 /* { 103 syscallarg(int) which; 104 syscallarg(id_t) who; 105 } */ 106 struct proc *curp = l->l_proc, *p; 107 int low = NZERO + PRIO_MAX + 1; 108 int who = SCARG(uap, who); 109 110 mutex_enter(proc_lock); 111 switch (SCARG(uap, which)) { 112 case PRIO_PROCESS: 113 if (who == 0) 114 p = curp; 115 else 116 p = p_find(who, PFIND_LOCKED); 117 if (p != NULL) 118 low = p->p_nice; 119 break; 120 121 case PRIO_PGRP: { 122 struct pgrp *pg; 123 124 if (who == 0) 125 pg = curp->p_pgrp; 126 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 127 break; 128 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 129 if (p->p_nice < low) 130 low = p->p_nice; 131 } 132 break; 133 } 134 135 case PRIO_USER: 136 if (who == 0) 137 who = (int)kauth_cred_geteuid(l->l_cred); 138 PROCLIST_FOREACH(p, &allproc) { 139 if ((p->p_flag & PK_MARKER) != 0) 140 continue; 141 mutex_enter(p->p_lock); 142 if (kauth_cred_geteuid(p->p_cred) == 143 (uid_t)who && p->p_nice < low) 144 low = p->p_nice; 145 mutex_exit(p->p_lock); 146 } 147 break; 148 149 default: 150 mutex_exit(proc_lock); 151 return (EINVAL); 152 } 153 mutex_exit(proc_lock); 154 155 if (low == NZERO + PRIO_MAX + 1) 156 return (ESRCH); 157 *retval = low - NZERO; 158 return (0); 159 } 160 161 /* ARGSUSED */ 162 int 163 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap, 164 register_t *retval) 165 { 166 /* { 167 syscallarg(int) which; 168 syscallarg(id_t) who; 169 syscallarg(int) prio; 170 } */ 171 struct proc *curp = l->l_proc, *p; 172 int found = 0, error = 0; 173 int who = SCARG(uap, who); 174 175 mutex_enter(proc_lock); 176 switch (SCARG(uap, which)) { 177 case PRIO_PROCESS: 178 if (who == 0) 179 p = curp; 180 else 181 p = p_find(who, PFIND_LOCKED); 182 if (p != 0) { 183 mutex_enter(p->p_lock); 184 error = donice(l, p, SCARG(uap, prio)); 185 mutex_exit(p->p_lock); 186 } 187 found++; 188 break; 189 190 case PRIO_PGRP: { 191 struct pgrp *pg; 192 193 if (who == 0) 194 pg = curp->p_pgrp; 195 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 196 break; 197 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 198 mutex_enter(p->p_lock); 199 error = donice(l, p, SCARG(uap, prio)); 200 mutex_exit(p->p_lock); 201 found++; 202 } 203 break; 204 } 205 206 case PRIO_USER: 207 if (who == 0) 208 who = (int)kauth_cred_geteuid(l->l_cred); 209 PROCLIST_FOREACH(p, &allproc) { 210 if ((p->p_flag & PK_MARKER) != 0) 211 continue; 212 mutex_enter(p->p_lock); 213 if (kauth_cred_geteuid(p->p_cred) == 214 (uid_t)SCARG(uap, who)) { 215 error = donice(l, p, SCARG(uap, prio)); 216 found++; 217 } 218 mutex_exit(p->p_lock); 219 } 220 break; 221 222 default: 223 error = EINVAL; 224 break; 225 } 226 mutex_exit(proc_lock); 227 if (found == 0) 228 return (ESRCH); 229 return (error); 230 } 231 232 /* 233 * Renice a process. 234 * 235 * Call with the target process' credentials locked. 236 */ 237 int 238 donice(struct lwp *l, struct proc *chgp, int n) 239 { 240 kauth_cred_t cred = l->l_cred; 241 242 KASSERT(mutex_owned(chgp->p_lock)); 243 244 if (n > PRIO_MAX) 245 n = PRIO_MAX; 246 if (n < PRIO_MIN) 247 n = PRIO_MIN; 248 n += NZERO; 249 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 250 KAUTH_ARG(n), NULL, NULL)) 251 return (EACCES); 252 sched_nice(chgp, n); 253 return (0); 254 } 255 256 /* ARGSUSED */ 257 int 258 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap, 259 register_t *retval) 260 { 261 /* { 262 syscallarg(int) which; 263 syscallarg(const struct rlimit *) rlp; 264 } */ 265 int which = SCARG(uap, which); 266 struct rlimit alim; 267 int error; 268 269 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 270 if (error) 271 return (error); 272 return (dosetrlimit(l, l->l_proc, which, &alim)); 273 } 274 275 int 276 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 277 { 278 struct rlimit *alimp; 279 int error; 280 281 if ((u_int)which >= RLIM_NLIMITS) 282 return (EINVAL); 283 284 if (limp->rlim_cur < 0 || limp->rlim_max < 0) 285 return (EINVAL); 286 287 if (limp->rlim_cur > limp->rlim_max) { 288 /* 289 * This is programming error. According to SUSv2, we should 290 * return error in this case. 291 */ 292 return (EINVAL); 293 } 294 295 alimp = &p->p_rlimit[which]; 296 /* if we don't change the value, no need to limcopy() */ 297 if (limp->rlim_cur == alimp->rlim_cur && 298 limp->rlim_max == alimp->rlim_max) 299 return 0; 300 301 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 302 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_SET), limp, KAUTH_ARG(which)); 303 if (error) 304 return (error); 305 306 lim_privatise(p, false); 307 /* p->p_limit is now unchangeable */ 308 alimp = &p->p_rlimit[which]; 309 310 switch (which) { 311 312 case RLIMIT_DATA: 313 if (limp->rlim_cur > maxdmap) 314 limp->rlim_cur = maxdmap; 315 if (limp->rlim_max > maxdmap) 316 limp->rlim_max = maxdmap; 317 break; 318 319 case RLIMIT_STACK: 320 if (limp->rlim_cur > maxsmap) 321 limp->rlim_cur = maxsmap; 322 if (limp->rlim_max > maxsmap) 323 limp->rlim_max = maxsmap; 324 325 /* 326 * Return EINVAL if the new stack size limit is lower than 327 * current usage. Otherwise, the process would get SIGSEGV the 328 * moment it would try to access anything on it's current stack. 329 * This conforms to SUSv2. 330 */ 331 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 332 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 333 return (EINVAL); 334 } 335 336 /* 337 * Stack is allocated to the max at exec time with 338 * only "rlim_cur" bytes accessible (In other words, 339 * allocates stack dividing two contiguous regions at 340 * "rlim_cur" bytes boundary). 341 * 342 * Since allocation is done in terms of page, roundup 343 * "rlim_cur" (otherwise, contiguous regions 344 * overlap). If stack limit is going up make more 345 * accessible, if going down make inaccessible. 346 */ 347 limp->rlim_cur = round_page(limp->rlim_cur); 348 if (limp->rlim_cur != alimp->rlim_cur) { 349 vaddr_t addr; 350 vsize_t size; 351 vm_prot_t prot; 352 353 if (limp->rlim_cur > alimp->rlim_cur) { 354 prot = VM_PROT_READ | VM_PROT_WRITE; 355 size = limp->rlim_cur - alimp->rlim_cur; 356 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 357 limp->rlim_cur; 358 } else { 359 prot = VM_PROT_NONE; 360 size = alimp->rlim_cur - limp->rlim_cur; 361 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 362 alimp->rlim_cur; 363 } 364 (void) uvm_map_protect(&p->p_vmspace->vm_map, 365 addr, addr+size, prot, false); 366 } 367 break; 368 369 case RLIMIT_NOFILE: 370 if (limp->rlim_cur > maxfiles) 371 limp->rlim_cur = maxfiles; 372 if (limp->rlim_max > maxfiles) 373 limp->rlim_max = maxfiles; 374 break; 375 376 case RLIMIT_NPROC: 377 if (limp->rlim_cur > maxproc) 378 limp->rlim_cur = maxproc; 379 if (limp->rlim_max > maxproc) 380 limp->rlim_max = maxproc; 381 break; 382 } 383 384 mutex_enter(&p->p_limit->pl_lock); 385 *alimp = *limp; 386 mutex_exit(&p->p_limit->pl_lock); 387 return (0); 388 } 389 390 /* ARGSUSED */ 391 int 392 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, 393 register_t *retval) 394 { 395 /* { 396 syscallarg(int) which; 397 syscallarg(struct rlimit *) rlp; 398 } */ 399 struct proc *p = l->l_proc; 400 int which = SCARG(uap, which); 401 struct rlimit rl; 402 403 if ((u_int)which >= RLIM_NLIMITS) 404 return (EINVAL); 405 406 mutex_enter(p->p_lock); 407 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 408 mutex_exit(p->p_lock); 409 410 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 411 } 412 413 /* 414 * Transform the running time and tick information in proc p into user, 415 * system, and interrupt time usage. 416 * 417 * Should be called with p->p_lock held unless called from exit1(). 418 */ 419 void 420 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 421 struct timeval *ip, struct timeval *rp) 422 { 423 uint64_t u, st, ut, it, tot; 424 struct lwp *l; 425 struct bintime tm; 426 struct timeval tv; 427 428 mutex_spin_enter(&p->p_stmutex); 429 st = p->p_sticks; 430 ut = p->p_uticks; 431 it = p->p_iticks; 432 mutex_spin_exit(&p->p_stmutex); 433 434 tm = p->p_rtime; 435 436 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 437 lwp_lock(l); 438 bintime_add(&tm, &l->l_rtime); 439 if ((l->l_pflag & LP_RUNNING) != 0) { 440 struct bintime diff; 441 /* 442 * Adjust for the current time slice. This is 443 * actually fairly important since the error 444 * here is on the order of a time quantum, 445 * which is much greater than the sampling 446 * error. 447 */ 448 binuptime(&diff); 449 bintime_sub(&diff, &l->l_stime); 450 bintime_add(&tm, &diff); 451 } 452 lwp_unlock(l); 453 } 454 455 tot = st + ut + it; 456 bintime2timeval(&tm, &tv); 457 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 458 459 if (tot == 0) { 460 /* No ticks, so can't use to share time out, split 50-50 */ 461 st = ut = u / 2; 462 } else { 463 st = (u * st) / tot; 464 ut = (u * ut) / tot; 465 } 466 if (sp != NULL) { 467 sp->tv_sec = st / 1000000; 468 sp->tv_usec = st % 1000000; 469 } 470 if (up != NULL) { 471 up->tv_sec = ut / 1000000; 472 up->tv_usec = ut % 1000000; 473 } 474 if (ip != NULL) { 475 if (it != 0) 476 it = (u * it) / tot; 477 ip->tv_sec = it / 1000000; 478 ip->tv_usec = it % 1000000; 479 } 480 if (rp != NULL) { 481 *rp = tv; 482 } 483 } 484 485 /* ARGSUSED */ 486 int 487 sys_getrusage(struct lwp *l, const struct sys_getrusage_args *uap, 488 register_t *retval) 489 { 490 /* { 491 syscallarg(int) who; 492 syscallarg(struct rusage *) rusage; 493 } */ 494 struct rusage ru; 495 struct proc *p = l->l_proc; 496 497 switch (SCARG(uap, who)) { 498 case RUSAGE_SELF: 499 mutex_enter(p->p_lock); 500 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 501 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 502 rulwps(p, &ru); 503 mutex_exit(p->p_lock); 504 break; 505 506 case RUSAGE_CHILDREN: 507 mutex_enter(p->p_lock); 508 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 509 mutex_exit(p->p_lock); 510 break; 511 512 default: 513 return EINVAL; 514 } 515 516 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 517 } 518 519 void 520 ruadd(struct rusage *ru, struct rusage *ru2) 521 { 522 long *ip, *ip2; 523 int i; 524 525 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 526 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 527 if (ru->ru_maxrss < ru2->ru_maxrss) 528 ru->ru_maxrss = ru2->ru_maxrss; 529 ip = &ru->ru_first; ip2 = &ru2->ru_first; 530 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 531 *ip++ += *ip2++; 532 } 533 534 void 535 rulwps(proc_t *p, struct rusage *ru) 536 { 537 lwp_t *l; 538 539 KASSERT(mutex_owned(p->p_lock)); 540 541 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 542 ruadd(ru, &l->l_ru); 543 ru->ru_nvcsw += (l->l_ncsw - l->l_nivcsw); 544 ru->ru_nivcsw += l->l_nivcsw; 545 } 546 } 547 548 /* 549 * Make a copy of the plimit structure. 550 * We share these structures copy-on-write after fork, 551 * and copy when a limit is changed. 552 * 553 * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure 554 * we are copying to change beneath our feet! 555 */ 556 struct plimit * 557 lim_copy(struct plimit *lim) 558 { 559 struct plimit *newlim; 560 char *corename; 561 size_t alen, len; 562 563 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 564 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 565 newlim->pl_flags = 0; 566 newlim->pl_refcnt = 1; 567 newlim->pl_sv_limit = NULL; 568 569 mutex_enter(&lim->pl_lock); 570 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 571 sizeof(struct rlimit) * RLIM_NLIMITS); 572 573 alen = 0; 574 corename = NULL; 575 for (;;) { 576 if (lim->pl_corename == defcorename) { 577 newlim->pl_corename = defcorename; 578 break; 579 } 580 len = strlen(lim->pl_corename) + 1; 581 if (len <= alen) { 582 newlim->pl_corename = corename; 583 memcpy(corename, lim->pl_corename, len); 584 corename = NULL; 585 break; 586 } 587 mutex_exit(&lim->pl_lock); 588 if (corename != NULL) 589 free(corename, M_TEMP); 590 alen = len; 591 corename = malloc(alen, M_TEMP, M_WAITOK); 592 mutex_enter(&lim->pl_lock); 593 } 594 mutex_exit(&lim->pl_lock); 595 if (corename != NULL) 596 free(corename, M_TEMP); 597 return newlim; 598 } 599 600 void 601 lim_addref(struct plimit *lim) 602 { 603 atomic_inc_uint(&lim->pl_refcnt); 604 } 605 606 /* 607 * Give a process it's own private plimit structure. 608 * This will only be shared (in fork) if modifications are to be shared. 609 */ 610 void 611 lim_privatise(struct proc *p, bool set_shared) 612 { 613 struct plimit *lim, *newlim; 614 615 lim = p->p_limit; 616 if (lim->pl_flags & PL_WRITEABLE) { 617 if (set_shared) 618 lim->pl_flags |= PL_SHAREMOD; 619 return; 620 } 621 622 if (set_shared && lim->pl_flags & PL_SHAREMOD) 623 return; 624 625 newlim = lim_copy(lim); 626 627 mutex_enter(p->p_lock); 628 if (p->p_limit->pl_flags & PL_WRITEABLE) { 629 /* Someone crept in while we were busy */ 630 mutex_exit(p->p_lock); 631 limfree(newlim); 632 if (set_shared) 633 p->p_limit->pl_flags |= PL_SHAREMOD; 634 return; 635 } 636 637 /* 638 * Since most accesses to p->p_limit aren't locked, we must not 639 * delete the old limit structure yet. 640 */ 641 newlim->pl_sv_limit = p->p_limit; 642 newlim->pl_flags |= PL_WRITEABLE; 643 if (set_shared) 644 newlim->pl_flags |= PL_SHAREMOD; 645 p->p_limit = newlim; 646 mutex_exit(p->p_lock); 647 } 648 649 void 650 limfree(struct plimit *lim) 651 { 652 struct plimit *sv_lim; 653 654 do { 655 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) 656 return; 657 if (lim->pl_corename != defcorename) 658 free(lim->pl_corename, M_TEMP); 659 sv_lim = lim->pl_sv_limit; 660 mutex_destroy(&lim->pl_lock); 661 pool_cache_put(plimit_cache, lim); 662 } while ((lim = sv_lim) != NULL); 663 } 664 665 struct pstats * 666 pstatscopy(struct pstats *ps) 667 { 668 669 struct pstats *newps; 670 671 newps = pool_cache_get(pstats_cache, PR_WAITOK); 672 673 memset(&newps->pstat_startzero, 0, 674 (unsigned) ((char *)&newps->pstat_endzero - 675 (char *)&newps->pstat_startzero)); 676 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 677 ((char *)&newps->pstat_endcopy - 678 (char *)&newps->pstat_startcopy)); 679 680 return (newps); 681 682 } 683 684 void 685 pstatsfree(struct pstats *ps) 686 { 687 688 pool_cache_put(pstats_cache, ps); 689 } 690 691 /* 692 * sysctl interface in five parts 693 */ 694 695 /* 696 * a routine for sysctl proc subtree helpers that need to pick a valid 697 * process by pid. 698 */ 699 static int 700 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 701 { 702 struct proc *ptmp; 703 int error = 0; 704 705 if (pid == PROC_CURPROC) 706 ptmp = l->l_proc; 707 else if ((ptmp = pfind(pid)) == NULL) 708 error = ESRCH; 709 710 *p2 = ptmp; 711 return (error); 712 } 713 714 /* 715 * sysctl helper routine for setting a process's specific corefile 716 * name. picks the process based on the given pid and checks the 717 * correctness of the new value. 718 */ 719 static int 720 sysctl_proc_corename(SYSCTLFN_ARGS) 721 { 722 struct proc *ptmp; 723 struct plimit *lim; 724 int error = 0, len; 725 char *cname; 726 char *ocore; 727 char *tmp; 728 struct sysctlnode node; 729 730 /* 731 * is this all correct? 732 */ 733 if (namelen != 0) 734 return (EINVAL); 735 if (name[-1] != PROC_PID_CORENAME) 736 return (EINVAL); 737 738 /* 739 * whom are we tweaking? 740 */ 741 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 742 if (error) 743 return (error); 744 745 /* XXX-elad */ 746 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 747 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 748 if (error) 749 return (error); 750 751 if (newp == NULL) { 752 error = kauth_authorize_process(l->l_cred, 753 KAUTH_PROCESS_CORENAME, ptmp, 754 KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_GET), NULL, NULL); 755 if (error) 756 return (error); 757 } 758 759 /* 760 * let them modify a temporary copy of the core name 761 */ 762 cname = PNBUF_GET(); 763 lim = ptmp->p_limit; 764 mutex_enter(&lim->pl_lock); 765 strlcpy(cname, lim->pl_corename, MAXPATHLEN); 766 mutex_exit(&lim->pl_lock); 767 768 node = *rnode; 769 node.sysctl_data = cname; 770 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 771 772 /* 773 * if that failed, or they have nothing new to say, or we've 774 * heard it before... 775 */ 776 if (error || newp == NULL) 777 goto done; 778 lim = ptmp->p_limit; 779 mutex_enter(&lim->pl_lock); 780 error = strcmp(cname, lim->pl_corename); 781 mutex_exit(&lim->pl_lock); 782 if (error == 0) 783 /* Unchanged */ 784 goto done; 785 786 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 787 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_SET), cname, NULL); 788 if (error) 789 return (error); 790 791 /* 792 * no error yet and cname now has the new core name in it. 793 * let's see if it looks acceptable. it must be either "core" 794 * or end in ".core" or "/core". 795 */ 796 len = strlen(cname); 797 if (len < 4) { 798 error = EINVAL; 799 } else if (strcmp(cname + len - 4, "core") != 0) { 800 error = EINVAL; 801 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 802 error = EINVAL; 803 } 804 if (error != 0) { 805 goto done; 806 } 807 808 /* 809 * hmm...looks good. now...where do we put it? 810 */ 811 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 812 if (tmp == NULL) { 813 error = ENOMEM; 814 goto done; 815 } 816 memcpy(tmp, cname, len + 1); 817 818 lim_privatise(ptmp, false); 819 lim = ptmp->p_limit; 820 mutex_enter(&lim->pl_lock); 821 ocore = lim->pl_corename; 822 lim->pl_corename = tmp; 823 mutex_exit(&lim->pl_lock); 824 if (ocore != defcorename) 825 free(ocore, M_TEMP); 826 827 done: 828 PNBUF_PUT(cname); 829 return error; 830 } 831 832 /* 833 * sysctl helper routine for checking/setting a process's stop flags, 834 * one for fork and one for exec. 835 */ 836 static int 837 sysctl_proc_stop(SYSCTLFN_ARGS) 838 { 839 struct proc *ptmp; 840 int i, f, error = 0; 841 struct sysctlnode node; 842 843 if (namelen != 0) 844 return (EINVAL); 845 846 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 847 if (error) 848 return (error); 849 850 /* XXX-elad */ 851 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 852 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 853 if (error) 854 return (error); 855 856 switch (rnode->sysctl_num) { 857 case PROC_PID_STOPFORK: 858 f = PS_STOPFORK; 859 break; 860 case PROC_PID_STOPEXEC: 861 f = PS_STOPEXEC; 862 break; 863 case PROC_PID_STOPEXIT: 864 f = PS_STOPEXIT; 865 break; 866 default: 867 return (EINVAL); 868 } 869 870 i = (ptmp->p_flag & f) ? 1 : 0; 871 node = *rnode; 872 node.sysctl_data = &i; 873 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 874 if (error || newp == NULL) 875 return (error); 876 877 mutex_enter(ptmp->p_lock); 878 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 879 ptmp, KAUTH_ARG(f), NULL, NULL); 880 if (error) 881 return (error); 882 if (i) 883 ptmp->p_sflag |= f; 884 else 885 ptmp->p_sflag &= ~f; 886 mutex_exit(ptmp->p_lock); 887 888 return (0); 889 } 890 891 /* 892 * sysctl helper routine for a process's rlimits as exposed by sysctl. 893 */ 894 static int 895 sysctl_proc_plimit(SYSCTLFN_ARGS) 896 { 897 struct proc *ptmp; 898 u_int limitno; 899 int which, error = 0; 900 struct rlimit alim; 901 struct sysctlnode node; 902 903 if (namelen != 0) 904 return (EINVAL); 905 906 which = name[-1]; 907 if (which != PROC_PID_LIMIT_TYPE_SOFT && 908 which != PROC_PID_LIMIT_TYPE_HARD) 909 return (EINVAL); 910 911 limitno = name[-2] - 1; 912 if (limitno >= RLIM_NLIMITS) 913 return (EINVAL); 914 915 if (name[-3] != PROC_PID_LIMIT) 916 return (EINVAL); 917 918 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 919 if (error) 920 return (error); 921 922 /* XXX-elad */ 923 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 924 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 925 if (error) 926 return (error); 927 928 /* Check if we can view limits. */ 929 if (newp == NULL) { 930 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 931 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_GET), &alim, 932 KAUTH_ARG(which)); 933 if (error) 934 return (error); 935 } 936 937 node = *rnode; 938 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 939 if (which == PROC_PID_LIMIT_TYPE_HARD) 940 node.sysctl_data = &alim.rlim_max; 941 else 942 node.sysctl_data = &alim.rlim_cur; 943 944 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 945 if (error || newp == NULL) 946 return (error); 947 948 return (dosetrlimit(l, ptmp, limitno, &alim)); 949 } 950 951 /* 952 * and finally, the actually glue that sticks it to the tree 953 */ 954 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup") 955 { 956 957 sysctl_createv(clog, 0, NULL, NULL, 958 CTLFLAG_PERMANENT, 959 CTLTYPE_NODE, "proc", NULL, 960 NULL, 0, NULL, 0, 961 CTL_PROC, CTL_EOL); 962 sysctl_createv(clog, 0, NULL, NULL, 963 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 964 CTLTYPE_NODE, "curproc", 965 SYSCTL_DESCR("Per-process settings"), 966 NULL, 0, NULL, 0, 967 CTL_PROC, PROC_CURPROC, CTL_EOL); 968 969 sysctl_createv(clog, 0, NULL, NULL, 970 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 971 CTLTYPE_STRING, "corename", 972 SYSCTL_DESCR("Core file name"), 973 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 974 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 975 sysctl_createv(clog, 0, NULL, NULL, 976 CTLFLAG_PERMANENT, 977 CTLTYPE_NODE, "rlimit", 978 SYSCTL_DESCR("Process limits"), 979 NULL, 0, NULL, 0, 980 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 981 982 #define create_proc_plimit(s, n) do { \ 983 sysctl_createv(clog, 0, NULL, NULL, \ 984 CTLFLAG_PERMANENT, \ 985 CTLTYPE_NODE, s, \ 986 SYSCTL_DESCR("Process " s " limits"), \ 987 NULL, 0, NULL, 0, \ 988 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 989 CTL_EOL); \ 990 sysctl_createv(clog, 0, NULL, NULL, \ 991 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 992 CTLTYPE_QUAD, "soft", \ 993 SYSCTL_DESCR("Process soft " s " limit"), \ 994 sysctl_proc_plimit, 0, NULL, 0, \ 995 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 996 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 997 sysctl_createv(clog, 0, NULL, NULL, \ 998 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 999 CTLTYPE_QUAD, "hard", \ 1000 SYSCTL_DESCR("Process hard " s " limit"), \ 1001 sysctl_proc_plimit, 0, NULL, 0, \ 1002 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1003 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 1004 } while (0/*CONSTCOND*/) 1005 1006 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 1007 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 1008 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 1009 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 1010 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 1011 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 1012 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 1013 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 1014 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 1015 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 1016 1017 #undef create_proc_plimit 1018 1019 sysctl_createv(clog, 0, NULL, NULL, 1020 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1021 CTLTYPE_INT, "stopfork", 1022 SYSCTL_DESCR("Stop process at fork(2)"), 1023 sysctl_proc_stop, 0, NULL, 0, 1024 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 1025 sysctl_createv(clog, 0, NULL, NULL, 1026 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1027 CTLTYPE_INT, "stopexec", 1028 SYSCTL_DESCR("Stop process at execve(2)"), 1029 sysctl_proc_stop, 0, NULL, 0, 1030 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 1031 sysctl_createv(clog, 0, NULL, NULL, 1032 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1033 CTLTYPE_INT, "stopexit", 1034 SYSCTL_DESCR("Stop process before completing exit"), 1035 sysctl_proc_stop, 0, NULL, 0, 1036 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1037 } 1038 1039 void 1040 uid_init(void) 1041 { 1042 1043 /* 1044 * Ensure that uid 0 is always in the user hash table, as 1045 * sbreserve() expects it available from interrupt context. 1046 */ 1047 (void)uid_find(0); 1048 } 1049 1050 struct uidinfo * 1051 uid_find(uid_t uid) 1052 { 1053 struct uidinfo *uip, *uip_first, *newuip; 1054 struct uihashhead *uipp; 1055 1056 uipp = UIHASH(uid); 1057 newuip = NULL; 1058 1059 /* 1060 * To make insertion atomic, abstraction of SLIST will be violated. 1061 */ 1062 uip_first = uipp->slh_first; 1063 again: 1064 SLIST_FOREACH(uip, uipp, ui_hash) { 1065 if (uip->ui_uid != uid) 1066 continue; 1067 if (newuip != NULL) 1068 kmem_free(newuip, sizeof(*newuip)); 1069 return uip; 1070 } 1071 if (newuip == NULL) 1072 newuip = kmem_zalloc(sizeof(*newuip), KM_SLEEP); 1073 newuip->ui_uid = uid; 1074 1075 /* 1076 * If atomic insert is unsuccessful, another thread might be 1077 * allocated this 'uid', thus full re-check is needed. 1078 */ 1079 newuip->ui_hash.sle_next = uip_first; 1080 membar_producer(); 1081 uip = atomic_cas_ptr(&uipp->slh_first, uip_first, newuip); 1082 if (uip != uip_first) { 1083 uip_first = uip; 1084 goto again; 1085 } 1086 1087 return newuip; 1088 } 1089 1090 /* 1091 * Change the count associated with number of processes 1092 * a given user is using. 1093 */ 1094 int 1095 chgproccnt(uid_t uid, int diff) 1096 { 1097 struct uidinfo *uip; 1098 long proccnt; 1099 1100 uip = uid_find(uid); 1101 proccnt = atomic_add_long_nv(&uip->ui_proccnt, diff); 1102 KASSERT(proccnt >= 0); 1103 return proccnt; 1104 } 1105 1106 int 1107 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax) 1108 { 1109 rlim_t nsb; 1110 const long diff = to - *hiwat; 1111 1112 nsb = atomic_add_long_nv((long *)&uip->ui_sbsize, diff); 1113 if (diff > 0 && nsb > xmax) { 1114 atomic_add_long((long *)&uip->ui_sbsize, -diff); 1115 return 0; 1116 } 1117 *hiwat = to; 1118 KASSERT(nsb >= 0); 1119 return 1; 1120 } 1121