kern_resource.c revision 1.132.2.1
1 /* $NetBSD: kern_resource.c,v 1.132.2.1 2008/03/24 07:16:13 keiichi 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 1.132.2.1 2008/03/24 07:16:13 keiichi 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, M_PROC, M_WAITOK, &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(&proclist_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 mutex_enter(&p->p_mutex); 140 if (kauth_cred_geteuid(p->p_cred) == 141 (uid_t)who && p->p_nice < low) 142 low = p->p_nice; 143 mutex_exit(&p->p_mutex); 144 } 145 break; 146 147 default: 148 mutex_exit(&proclist_lock); 149 return (EINVAL); 150 } 151 mutex_exit(&proclist_lock); 152 153 if (low == NZERO + PRIO_MAX + 1) 154 return (ESRCH); 155 *retval = low - NZERO; 156 return (0); 157 } 158 159 /* ARGSUSED */ 160 int 161 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap, 162 register_t *retval) 163 { 164 /* { 165 syscallarg(int) which; 166 syscallarg(id_t) who; 167 syscallarg(int) prio; 168 } */ 169 struct proc *curp = l->l_proc, *p; 170 int found = 0, error = 0; 171 int who = SCARG(uap, who); 172 173 mutex_enter(&proclist_lock); 174 switch (SCARG(uap, which)) { 175 case PRIO_PROCESS: 176 if (who == 0) 177 p = curp; 178 else 179 p = p_find(who, PFIND_LOCKED); 180 if (p != 0) { 181 mutex_enter(&p->p_mutex); 182 error = donice(l, p, SCARG(uap, prio)); 183 mutex_exit(&p->p_mutex); 184 } 185 found++; 186 break; 187 188 case PRIO_PGRP: { 189 struct pgrp *pg; 190 191 if (who == 0) 192 pg = curp->p_pgrp; 193 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 194 break; 195 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 196 mutex_enter(&p->p_mutex); 197 error = donice(l, p, SCARG(uap, prio)); 198 mutex_exit(&p->p_mutex); 199 found++; 200 } 201 break; 202 } 203 204 case PRIO_USER: 205 if (who == 0) 206 who = (int)kauth_cred_geteuid(l->l_cred); 207 PROCLIST_FOREACH(p, &allproc) { 208 mutex_enter(&p->p_mutex); 209 if (kauth_cred_geteuid(p->p_cred) == 210 (uid_t)SCARG(uap, who)) { 211 error = donice(l, p, SCARG(uap, prio)); 212 found++; 213 } 214 mutex_exit(&p->p_mutex); 215 } 216 break; 217 218 default: 219 error = EINVAL; 220 break; 221 } 222 mutex_exit(&proclist_lock); 223 if (found == 0) 224 return (ESRCH); 225 return (error); 226 } 227 228 /* 229 * Renice a process. 230 * 231 * Call with the target process' credentials locked. 232 */ 233 int 234 donice(struct lwp *l, struct proc *chgp, int n) 235 { 236 kauth_cred_t cred = l->l_cred; 237 int onice; 238 239 KASSERT(mutex_owned(&chgp->p_mutex)); 240 241 if (n > PRIO_MAX) 242 n = PRIO_MAX; 243 if (n < PRIO_MIN) 244 n = PRIO_MIN; 245 n += NZERO; 246 onice = chgp->p_nice; 247 onice = chgp->p_nice; 248 249 again: 250 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 251 KAUTH_ARG(n), NULL, NULL)) 252 return (EACCES); 253 mutex_spin_enter(&chgp->p_smutex); 254 if (onice != chgp->p_nice) { 255 mutex_spin_exit(&chgp->p_smutex); 256 goto again; 257 } 258 sched_nice(chgp, n); 259 mutex_spin_exit(&chgp->p_smutex); 260 return (0); 261 } 262 263 /* ARGSUSED */ 264 int 265 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap, 266 register_t *retval) 267 { 268 /* { 269 syscallarg(int) which; 270 syscallarg(const struct rlimit *) rlp; 271 } */ 272 int which = SCARG(uap, which); 273 struct rlimit alim; 274 int error; 275 276 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 277 if (error) 278 return (error); 279 return (dosetrlimit(l, l->l_proc, which, &alim)); 280 } 281 282 int 283 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 284 { 285 struct rlimit *alimp; 286 int error; 287 288 if ((u_int)which >= RLIM_NLIMITS) 289 return (EINVAL); 290 291 if (limp->rlim_cur < 0 || limp->rlim_max < 0) 292 return (EINVAL); 293 294 if (limp->rlim_cur > limp->rlim_max) { 295 /* 296 * This is programming error. According to SUSv2, we should 297 * return error in this case. 298 */ 299 return (EINVAL); 300 } 301 302 alimp = &p->p_rlimit[which]; 303 /* if we don't change the value, no need to limcopy() */ 304 if (limp->rlim_cur == alimp->rlim_cur && 305 limp->rlim_max == alimp->rlim_max) 306 return 0; 307 308 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 309 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_SET), limp, KAUTH_ARG(which)); 310 if (error) 311 return (error); 312 313 lim_privatise(p, false); 314 /* p->p_limit is now unchangeable */ 315 alimp = &p->p_rlimit[which]; 316 317 switch (which) { 318 319 case RLIMIT_DATA: 320 if (limp->rlim_cur > maxdmap) 321 limp->rlim_cur = maxdmap; 322 if (limp->rlim_max > maxdmap) 323 limp->rlim_max = maxdmap; 324 break; 325 326 case RLIMIT_STACK: 327 if (limp->rlim_cur > maxsmap) 328 limp->rlim_cur = maxsmap; 329 if (limp->rlim_max > maxsmap) 330 limp->rlim_max = maxsmap; 331 332 /* 333 * Return EINVAL if the new stack size limit is lower than 334 * current usage. Otherwise, the process would get SIGSEGV the 335 * moment it would try to access anything on it's current stack. 336 * This conforms to SUSv2. 337 */ 338 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 339 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 340 return (EINVAL); 341 } 342 343 /* 344 * Stack is allocated to the max at exec time with 345 * only "rlim_cur" bytes accessible (In other words, 346 * allocates stack dividing two contiguous regions at 347 * "rlim_cur" bytes boundary). 348 * 349 * Since allocation is done in terms of page, roundup 350 * "rlim_cur" (otherwise, contiguous regions 351 * overlap). If stack limit is going up make more 352 * accessible, if going down make inaccessible. 353 */ 354 limp->rlim_cur = round_page(limp->rlim_cur); 355 if (limp->rlim_cur != alimp->rlim_cur) { 356 vaddr_t addr; 357 vsize_t size; 358 vm_prot_t prot; 359 360 if (limp->rlim_cur > alimp->rlim_cur) { 361 prot = VM_PROT_READ | VM_PROT_WRITE; 362 size = limp->rlim_cur - alimp->rlim_cur; 363 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 364 limp->rlim_cur; 365 } else { 366 prot = VM_PROT_NONE; 367 size = alimp->rlim_cur - limp->rlim_cur; 368 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 369 alimp->rlim_cur; 370 } 371 (void) uvm_map_protect(&p->p_vmspace->vm_map, 372 addr, addr+size, prot, false); 373 } 374 break; 375 376 case RLIMIT_NOFILE: 377 if (limp->rlim_cur > maxfiles) 378 limp->rlim_cur = maxfiles; 379 if (limp->rlim_max > maxfiles) 380 limp->rlim_max = maxfiles; 381 break; 382 383 case RLIMIT_NPROC: 384 if (limp->rlim_cur > maxproc) 385 limp->rlim_cur = maxproc; 386 if (limp->rlim_max > maxproc) 387 limp->rlim_max = maxproc; 388 break; 389 } 390 391 mutex_enter(&p->p_limit->pl_lock); 392 *alimp = *limp; 393 mutex_exit(&p->p_limit->pl_lock); 394 return (0); 395 } 396 397 /* ARGSUSED */ 398 int 399 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, 400 register_t *retval) 401 { 402 /* { 403 syscallarg(int) which; 404 syscallarg(struct rlimit *) rlp; 405 } */ 406 struct proc *p = l->l_proc; 407 int which = SCARG(uap, which); 408 struct rlimit rl; 409 410 if ((u_int)which >= RLIM_NLIMITS) 411 return (EINVAL); 412 413 mutex_enter(&p->p_mutex); 414 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 415 mutex_exit(&p->p_mutex); 416 417 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 418 } 419 420 /* 421 * Transform the running time and tick information in proc p into user, 422 * system, and interrupt time usage. 423 * 424 * Should be called with p->p_smutex held unless called from exit1(). 425 */ 426 void 427 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 428 struct timeval *ip, struct timeval *rp) 429 { 430 uint64_t u, st, ut, it, tot; 431 struct lwp *l; 432 struct bintime tm; 433 struct timeval tv; 434 435 mutex_spin_enter(&p->p_stmutex); 436 st = p->p_sticks; 437 ut = p->p_uticks; 438 it = p->p_iticks; 439 mutex_spin_exit(&p->p_stmutex); 440 441 tm = p->p_rtime; 442 443 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 444 lwp_lock(l); 445 bintime_add(&tm, &l->l_rtime); 446 if ((l->l_flag & LW_RUNNING) != 0) { 447 struct bintime diff; 448 /* 449 * Adjust for the current time slice. This is 450 * actually fairly important since the error 451 * here is on the order of a time quantum, 452 * which is much greater than the sampling 453 * error. 454 */ 455 binuptime(&diff); 456 bintime_sub(&diff, &l->l_stime); 457 bintime_add(&tm, &diff); 458 } 459 lwp_unlock(l); 460 } 461 462 tot = st + ut + it; 463 bintime2timeval(&tm, &tv); 464 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 465 466 if (tot == 0) { 467 /* No ticks, so can't use to share time out, split 50-50 */ 468 st = ut = u / 2; 469 } else { 470 st = (u * st) / tot; 471 ut = (u * ut) / tot; 472 } 473 if (sp != NULL) { 474 sp->tv_sec = st / 1000000; 475 sp->tv_usec = st % 1000000; 476 } 477 if (up != NULL) { 478 up->tv_sec = ut / 1000000; 479 up->tv_usec = ut % 1000000; 480 } 481 if (ip != NULL) { 482 if (it != 0) 483 it = (u * it) / tot; 484 ip->tv_sec = it / 1000000; 485 ip->tv_usec = it % 1000000; 486 } 487 if (rp != NULL) { 488 *rp = tv; 489 } 490 } 491 492 /* ARGSUSED */ 493 int 494 sys_getrusage(struct lwp *l, const struct sys_getrusage_args *uap, 495 register_t *retval) 496 { 497 /* { 498 syscallarg(int) who; 499 syscallarg(struct rusage *) rusage; 500 } */ 501 struct rusage ru; 502 struct proc *p = l->l_proc; 503 504 switch (SCARG(uap, who)) { 505 case RUSAGE_SELF: 506 mutex_enter(&p->p_smutex); 507 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 508 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 509 mutex_exit(&p->p_smutex); 510 break; 511 512 case RUSAGE_CHILDREN: 513 mutex_enter(&p->p_smutex); 514 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 515 mutex_exit(&p->p_smutex); 516 break; 517 518 default: 519 return EINVAL; 520 } 521 522 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 523 } 524 525 void 526 ruadd(struct rusage *ru, struct rusage *ru2) 527 { 528 long *ip, *ip2; 529 int i; 530 531 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 532 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 533 if (ru->ru_maxrss < ru2->ru_maxrss) 534 ru->ru_maxrss = ru2->ru_maxrss; 535 ip = &ru->ru_first; ip2 = &ru2->ru_first; 536 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 537 *ip++ += *ip2++; 538 } 539 540 /* 541 * Make a copy of the plimit structure. 542 * We share these structures copy-on-write after fork, 543 * and copy when a limit is changed. 544 * 545 * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure 546 * we are copying to change beneath our feet! 547 */ 548 struct plimit * 549 lim_copy(struct plimit *lim) 550 { 551 struct plimit *newlim; 552 char *corename; 553 size_t alen, len; 554 555 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 556 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 557 newlim->pl_flags = 0; 558 newlim->pl_refcnt = 1; 559 newlim->pl_sv_limit = NULL; 560 561 mutex_enter(&lim->pl_lock); 562 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 563 sizeof(struct rlimit) * RLIM_NLIMITS); 564 565 alen = 0; 566 corename = NULL; 567 for (;;) { 568 if (lim->pl_corename == defcorename) { 569 newlim->pl_corename = defcorename; 570 break; 571 } 572 len = strlen(lim->pl_corename) + 1; 573 if (len <= alen) { 574 newlim->pl_corename = corename; 575 memcpy(corename, lim->pl_corename, len); 576 corename = NULL; 577 break; 578 } 579 mutex_exit(&lim->pl_lock); 580 if (corename != NULL) 581 free(corename, M_TEMP); 582 alen = len; 583 corename = malloc(alen, M_TEMP, M_WAITOK); 584 mutex_enter(&lim->pl_lock); 585 } 586 mutex_exit(&lim->pl_lock); 587 if (corename != NULL) 588 free(corename, M_TEMP); 589 return newlim; 590 } 591 592 void 593 lim_addref(struct plimit *lim) 594 { 595 atomic_inc_uint(&lim->pl_refcnt); 596 } 597 598 /* 599 * Give a process it's own private plimit structure. 600 * This will only be shared (in fork) if modifications are to be shared. 601 */ 602 void 603 lim_privatise(struct proc *p, bool set_shared) 604 { 605 struct plimit *lim, *newlim; 606 607 lim = p->p_limit; 608 if (lim->pl_flags & PL_WRITEABLE) { 609 if (set_shared) 610 lim->pl_flags |= PL_SHAREMOD; 611 return; 612 } 613 614 if (set_shared && lim->pl_flags & PL_SHAREMOD) 615 return; 616 617 newlim = lim_copy(lim); 618 619 mutex_enter(&p->p_mutex); 620 if (p->p_limit->pl_flags & PL_WRITEABLE) { 621 /* Someone crept in while we were busy */ 622 mutex_exit(&p->p_mutex); 623 limfree(newlim); 624 if (set_shared) 625 p->p_limit->pl_flags |= PL_SHAREMOD; 626 return; 627 } 628 629 /* 630 * Since most accesses to p->p_limit aren't locked, we must not 631 * delete the old limit structure yet. 632 */ 633 newlim->pl_sv_limit = p->p_limit; 634 newlim->pl_flags |= PL_WRITEABLE; 635 if (set_shared) 636 newlim->pl_flags |= PL_SHAREMOD; 637 p->p_limit = newlim; 638 mutex_exit(&p->p_mutex); 639 } 640 641 void 642 limfree(struct plimit *lim) 643 { 644 struct plimit *sv_lim; 645 646 do { 647 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) 648 return; 649 if (lim->pl_corename != defcorename) 650 free(lim->pl_corename, M_TEMP); 651 sv_lim = lim->pl_sv_limit; 652 mutex_destroy(&lim->pl_lock); 653 pool_cache_put(plimit_cache, lim); 654 } while ((lim = sv_lim) != NULL); 655 } 656 657 struct pstats * 658 pstatscopy(struct pstats *ps) 659 { 660 661 struct pstats *newps; 662 663 newps = pool_cache_get(pstats_cache, PR_WAITOK); 664 665 memset(&newps->pstat_startzero, 0, 666 (unsigned) ((char *)&newps->pstat_endzero - 667 (char *)&newps->pstat_startzero)); 668 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 669 ((char *)&newps->pstat_endcopy - 670 (char *)&newps->pstat_startcopy)); 671 672 return (newps); 673 674 } 675 676 void 677 pstatsfree(struct pstats *ps) 678 { 679 680 pool_cache_put(pstats_cache, ps); 681 } 682 683 /* 684 * sysctl interface in five parts 685 */ 686 687 /* 688 * a routine for sysctl proc subtree helpers that need to pick a valid 689 * process by pid. 690 */ 691 static int 692 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 693 { 694 struct proc *ptmp; 695 int error = 0; 696 697 if (pid == PROC_CURPROC) 698 ptmp = l->l_proc; 699 else if ((ptmp = pfind(pid)) == NULL) 700 error = ESRCH; 701 702 *p2 = ptmp; 703 return (error); 704 } 705 706 /* 707 * sysctl helper routine for setting a process's specific corefile 708 * name. picks the process based on the given pid and checks the 709 * correctness of the new value. 710 */ 711 static int 712 sysctl_proc_corename(SYSCTLFN_ARGS) 713 { 714 struct proc *ptmp; 715 struct plimit *lim; 716 int error = 0, len; 717 char *cname; 718 char *ocore; 719 char *tmp; 720 struct sysctlnode node; 721 722 /* 723 * is this all correct? 724 */ 725 if (namelen != 0) 726 return (EINVAL); 727 if (name[-1] != PROC_PID_CORENAME) 728 return (EINVAL); 729 730 /* 731 * whom are we tweaking? 732 */ 733 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 734 if (error) 735 return (error); 736 737 /* XXX-elad */ 738 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 739 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 740 if (error) 741 return (error); 742 743 if (newp == NULL) { 744 error = kauth_authorize_process(l->l_cred, 745 KAUTH_PROCESS_CORENAME, ptmp, 746 KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_GET), NULL, NULL); 747 if (error) 748 return (error); 749 } 750 751 /* 752 * let them modify a temporary copy of the core name 753 */ 754 cname = PNBUF_GET(); 755 lim = ptmp->p_limit; 756 mutex_enter(&lim->pl_lock); 757 strlcpy(cname, lim->pl_corename, MAXPATHLEN); 758 mutex_exit(&lim->pl_lock); 759 760 node = *rnode; 761 node.sysctl_data = cname; 762 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 763 764 /* 765 * if that failed, or they have nothing new to say, or we've 766 * heard it before... 767 */ 768 if (error || newp == NULL) 769 goto done; 770 lim = ptmp->p_limit; 771 mutex_enter(&lim->pl_lock); 772 error = strcmp(cname, lim->pl_corename); 773 mutex_exit(&lim->pl_lock); 774 if (error == 0) 775 /* Unchanged */ 776 goto done; 777 778 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 779 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_SET), cname, NULL); 780 if (error) 781 return (error); 782 783 /* 784 * no error yet and cname now has the new core name in it. 785 * let's see if it looks acceptable. it must be either "core" 786 * or end in ".core" or "/core". 787 */ 788 len = strlen(cname); 789 if (len < 4) { 790 error = EINVAL; 791 } else if (strcmp(cname + len - 4, "core") != 0) { 792 error = EINVAL; 793 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 794 error = EINVAL; 795 } 796 if (error != 0) { 797 goto done; 798 } 799 800 /* 801 * hmm...looks good. now...where do we put it? 802 */ 803 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 804 if (tmp == NULL) { 805 error = ENOMEM; 806 goto done; 807 } 808 memcpy(tmp, cname, len + 1); 809 810 lim_privatise(ptmp, false); 811 lim = ptmp->p_limit; 812 mutex_enter(&lim->pl_lock); 813 ocore = lim->pl_corename; 814 lim->pl_corename = tmp; 815 mutex_exit(&lim->pl_lock); 816 if (ocore != defcorename) 817 free(ocore, M_TEMP); 818 819 done: 820 PNBUF_PUT(cname); 821 return error; 822 } 823 824 /* 825 * sysctl helper routine for checking/setting a process's stop flags, 826 * one for fork and one for exec. 827 */ 828 static int 829 sysctl_proc_stop(SYSCTLFN_ARGS) 830 { 831 struct proc *ptmp; 832 int i, f, error = 0; 833 struct sysctlnode node; 834 835 if (namelen != 0) 836 return (EINVAL); 837 838 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 839 if (error) 840 return (error); 841 842 /* XXX-elad */ 843 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 844 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 845 if (error) 846 return (error); 847 848 switch (rnode->sysctl_num) { 849 case PROC_PID_STOPFORK: 850 f = PS_STOPFORK; 851 break; 852 case PROC_PID_STOPEXEC: 853 f = PS_STOPEXEC; 854 break; 855 case PROC_PID_STOPEXIT: 856 f = PS_STOPEXIT; 857 break; 858 default: 859 return (EINVAL); 860 } 861 862 i = (ptmp->p_flag & f) ? 1 : 0; 863 node = *rnode; 864 node.sysctl_data = &i; 865 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 866 if (error || newp == NULL) 867 return (error); 868 869 mutex_enter(&ptmp->p_smutex); 870 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 871 ptmp, KAUTH_ARG(f), NULL, NULL); 872 if (error) 873 return (error); 874 if (i) 875 ptmp->p_sflag |= f; 876 else 877 ptmp->p_sflag &= ~f; 878 mutex_exit(&ptmp->p_smutex); 879 880 return (0); 881 } 882 883 /* 884 * sysctl helper routine for a process's rlimits as exposed by sysctl. 885 */ 886 static int 887 sysctl_proc_plimit(SYSCTLFN_ARGS) 888 { 889 struct proc *ptmp; 890 u_int limitno; 891 int which, error = 0; 892 struct rlimit alim; 893 struct sysctlnode node; 894 895 if (namelen != 0) 896 return (EINVAL); 897 898 which = name[-1]; 899 if (which != PROC_PID_LIMIT_TYPE_SOFT && 900 which != PROC_PID_LIMIT_TYPE_HARD) 901 return (EINVAL); 902 903 limitno = name[-2] - 1; 904 if (limitno >= RLIM_NLIMITS) 905 return (EINVAL); 906 907 if (name[-3] != PROC_PID_LIMIT) 908 return (EINVAL); 909 910 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 911 if (error) 912 return (error); 913 914 /* XXX-elad */ 915 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 916 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 917 if (error) 918 return (error); 919 920 /* Check if we can view limits. */ 921 if (newp == NULL) { 922 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 923 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_GET), &alim, 924 KAUTH_ARG(which)); 925 if (error) 926 return (error); 927 } 928 929 node = *rnode; 930 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 931 if (which == PROC_PID_LIMIT_TYPE_HARD) 932 node.sysctl_data = &alim.rlim_max; 933 else 934 node.sysctl_data = &alim.rlim_cur; 935 936 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 937 if (error || newp == NULL) 938 return (error); 939 940 return (dosetrlimit(l, ptmp, limitno, &alim)); 941 } 942 943 /* 944 * and finally, the actually glue that sticks it to the tree 945 */ 946 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup") 947 { 948 949 sysctl_createv(clog, 0, NULL, NULL, 950 CTLFLAG_PERMANENT, 951 CTLTYPE_NODE, "proc", NULL, 952 NULL, 0, NULL, 0, 953 CTL_PROC, CTL_EOL); 954 sysctl_createv(clog, 0, NULL, NULL, 955 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 956 CTLTYPE_NODE, "curproc", 957 SYSCTL_DESCR("Per-process settings"), 958 NULL, 0, NULL, 0, 959 CTL_PROC, PROC_CURPROC, CTL_EOL); 960 961 sysctl_createv(clog, 0, NULL, NULL, 962 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 963 CTLTYPE_STRING, "corename", 964 SYSCTL_DESCR("Core file name"), 965 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 966 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 967 sysctl_createv(clog, 0, NULL, NULL, 968 CTLFLAG_PERMANENT, 969 CTLTYPE_NODE, "rlimit", 970 SYSCTL_DESCR("Process limits"), 971 NULL, 0, NULL, 0, 972 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 973 974 #define create_proc_plimit(s, n) do { \ 975 sysctl_createv(clog, 0, NULL, NULL, \ 976 CTLFLAG_PERMANENT, \ 977 CTLTYPE_NODE, s, \ 978 SYSCTL_DESCR("Process " s " limits"), \ 979 NULL, 0, NULL, 0, \ 980 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 981 CTL_EOL); \ 982 sysctl_createv(clog, 0, NULL, NULL, \ 983 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 984 CTLTYPE_QUAD, "soft", \ 985 SYSCTL_DESCR("Process soft " s " limit"), \ 986 sysctl_proc_plimit, 0, NULL, 0, \ 987 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 988 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 989 sysctl_createv(clog, 0, NULL, NULL, \ 990 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 991 CTLTYPE_QUAD, "hard", \ 992 SYSCTL_DESCR("Process hard " s " limit"), \ 993 sysctl_proc_plimit, 0, NULL, 0, \ 994 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 995 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 996 } while (0/*CONSTCOND*/) 997 998 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 999 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 1000 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 1001 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 1002 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 1003 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 1004 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 1005 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 1006 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 1007 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 1008 1009 #undef create_proc_plimit 1010 1011 sysctl_createv(clog, 0, NULL, NULL, 1012 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1013 CTLTYPE_INT, "stopfork", 1014 SYSCTL_DESCR("Stop process at fork(2)"), 1015 sysctl_proc_stop, 0, NULL, 0, 1016 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 1017 sysctl_createv(clog, 0, NULL, NULL, 1018 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1019 CTLTYPE_INT, "stopexec", 1020 SYSCTL_DESCR("Stop process at execve(2)"), 1021 sysctl_proc_stop, 0, NULL, 0, 1022 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 1023 sysctl_createv(clog, 0, NULL, NULL, 1024 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1025 CTLTYPE_INT, "stopexit", 1026 SYSCTL_DESCR("Stop process before completing exit"), 1027 sysctl_proc_stop, 0, NULL, 0, 1028 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1029 } 1030 1031 void 1032 uid_init(void) 1033 { 1034 1035 /* 1036 * Ensure that uid 0 is always in the user hash table, as 1037 * sbreserve() expects it available from interrupt context. 1038 */ 1039 (void)uid_find(0); 1040 } 1041 1042 struct uidinfo * 1043 uid_find(uid_t uid) 1044 { 1045 struct uidinfo *uip, *uip_first, *newuip; 1046 struct uihashhead *uipp; 1047 1048 uipp = UIHASH(uid); 1049 newuip = NULL; 1050 1051 /* 1052 * To make insertion atomic, abstraction of SLIST will be violated. 1053 */ 1054 uip_first = uipp->slh_first; 1055 again: 1056 SLIST_FOREACH(uip, uipp, ui_hash) { 1057 if (uip->ui_uid != uid) 1058 continue; 1059 if (newuip != NULL) 1060 kmem_free(newuip, sizeof(*newuip)); 1061 return uip; 1062 } 1063 if (newuip == NULL) 1064 newuip = kmem_zalloc(sizeof(*newuip), KM_SLEEP); 1065 newuip->ui_uid = uid; 1066 1067 /* 1068 * If atomic insert is unsuccessful, another thread might be 1069 * allocated this 'uid', thus full re-check is needed. 1070 */ 1071 newuip->ui_hash.sle_next = uip_first; 1072 membar_producer(); 1073 uip = atomic_cas_ptr(&uipp->slh_first, uip_first, newuip); 1074 if (uip != uip_first) { 1075 uip_first = uip; 1076 goto again; 1077 } 1078 1079 return newuip; 1080 } 1081 1082 /* 1083 * Change the count associated with number of processes 1084 * a given user is using. 1085 */ 1086 int 1087 chgproccnt(uid_t uid, int diff) 1088 { 1089 struct uidinfo *uip; 1090 long proccnt; 1091 1092 uip = uid_find(uid); 1093 proccnt = atomic_add_long_nv(&uip->ui_proccnt, diff); 1094 KASSERT(proccnt >= 0); 1095 return proccnt; 1096 } 1097 1098 int 1099 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax) 1100 { 1101 rlim_t nsb; 1102 const long diff = to - *hiwat; 1103 1104 nsb = atomic_add_long_nv((long *)&uip->ui_sbsize, diff); 1105 if (diff > 0 && nsb > xmax) { 1106 atomic_add_long((long *)&uip->ui_sbsize, -diff); 1107 return 0; 1108 } 1109 *hiwat = to; 1110 KASSERT(nsb >= 0); 1111 return 1; 1112 } 1113
Indexes created Sun Sep 21 19:09:51 GMT 2025