kern_resource.c revision 1.130
1 /* $NetBSD: kern_resource.c,v 1.130 2007/12/26 16:01:36 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 1.130 2007/12/26 16:01:36 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/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 58 #include <uvm/uvm_extern.h> 59 60 /* 61 * Maximum process data and stack limits. 62 * They are variables so they are patchable. 63 */ 64 rlim_t maxdmap = MAXDSIZ; 65 rlim_t maxsmap = MAXSSIZ; 66 67 struct uihashhead *uihashtbl; 68 u_long uihash; /* size of hash table - 1 */ 69 kmutex_t uihashtbl_lock; 70 71 static pool_cache_t plimit_cache; 72 static pool_cache_t pstats_cache; 73 74 void 75 resource_init(void) 76 { 77 78 plimit_cache = pool_cache_init(sizeof(struct plimit), 0, 0, 0, 79 "plimitpl", NULL, IPL_NONE, NULL, NULL, NULL); 80 pstats_cache = pool_cache_init(sizeof(struct pstats), 0, 0, 0, 81 "pstatspl", NULL, IPL_NONE, NULL, NULL, NULL); 82 } 83 84 /* 85 * Resource controls and accounting. 86 */ 87 88 int 89 sys_getpriority(struct lwp *l, const struct sys_getpriority_args *uap, register_t *retval) 90 { 91 /* { 92 syscallarg(int) which; 93 syscallarg(id_t) who; 94 } */ 95 struct proc *curp = l->l_proc, *p; 96 int low = NZERO + PRIO_MAX + 1; 97 int who = SCARG(uap, who); 98 99 mutex_enter(&proclist_lock); 100 switch (SCARG(uap, which)) { 101 case PRIO_PROCESS: 102 if (who == 0) 103 p = curp; 104 else 105 p = p_find(who, PFIND_LOCKED); 106 if (p != NULL) 107 low = p->p_nice; 108 break; 109 110 case PRIO_PGRP: { 111 struct pgrp *pg; 112 113 if (who == 0) 114 pg = curp->p_pgrp; 115 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 116 break; 117 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 118 if (p->p_nice < low) 119 low = p->p_nice; 120 } 121 break; 122 } 123 124 case PRIO_USER: 125 if (who == 0) 126 who = (int)kauth_cred_geteuid(l->l_cred); 127 PROCLIST_FOREACH(p, &allproc) { 128 mutex_enter(&p->p_mutex); 129 if (kauth_cred_geteuid(p->p_cred) == 130 (uid_t)who && p->p_nice < low) 131 low = p->p_nice; 132 mutex_exit(&p->p_mutex); 133 } 134 break; 135 136 default: 137 mutex_exit(&proclist_lock); 138 return (EINVAL); 139 } 140 mutex_exit(&proclist_lock); 141 142 if (low == NZERO + PRIO_MAX + 1) 143 return (ESRCH); 144 *retval = low - NZERO; 145 return (0); 146 } 147 148 /* ARGSUSED */ 149 int 150 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap, register_t *retval) 151 { 152 /* { 153 syscallarg(int) which; 154 syscallarg(id_t) who; 155 syscallarg(int) prio; 156 } */ 157 struct proc *curp = l->l_proc, *p; 158 int found = 0, error = 0; 159 int who = SCARG(uap, who); 160 161 mutex_enter(&proclist_lock); 162 switch (SCARG(uap, which)) { 163 case PRIO_PROCESS: 164 if (who == 0) 165 p = curp; 166 else 167 p = p_find(who, PFIND_LOCKED); 168 if (p != 0) { 169 mutex_enter(&p->p_mutex); 170 error = donice(l, p, SCARG(uap, prio)); 171 mutex_exit(&p->p_mutex); 172 } 173 found++; 174 break; 175 176 case PRIO_PGRP: { 177 struct pgrp *pg; 178 179 if (who == 0) 180 pg = curp->p_pgrp; 181 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 182 break; 183 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 184 mutex_enter(&p->p_mutex); 185 error = donice(l, p, SCARG(uap, prio)); 186 mutex_exit(&p->p_mutex); 187 found++; 188 } 189 break; 190 } 191 192 case PRIO_USER: 193 if (who == 0) 194 who = (int)kauth_cred_geteuid(l->l_cred); 195 PROCLIST_FOREACH(p, &allproc) { 196 mutex_enter(&p->p_mutex); 197 if (kauth_cred_geteuid(p->p_cred) == 198 (uid_t)SCARG(uap, who)) { 199 error = donice(l, p, SCARG(uap, prio)); 200 found++; 201 } 202 mutex_exit(&p->p_mutex); 203 } 204 break; 205 206 default: 207 error = EINVAL; 208 break; 209 } 210 mutex_exit(&proclist_lock); 211 if (found == 0) 212 return (ESRCH); 213 return (error); 214 } 215 216 /* 217 * Renice a process. 218 * 219 * Call with the target process' credentials locked. 220 */ 221 int 222 donice(struct lwp *l, struct proc *chgp, int n) 223 { 224 kauth_cred_t cred = l->l_cred; 225 int onice; 226 227 KASSERT(mutex_owned(&chgp->p_mutex)); 228 229 if (n > PRIO_MAX) 230 n = PRIO_MAX; 231 if (n < PRIO_MIN) 232 n = PRIO_MIN; 233 n += NZERO; 234 onice = chgp->p_nice; 235 onice = chgp->p_nice; 236 237 again: 238 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 239 KAUTH_ARG(n), NULL, NULL)) 240 return (EACCES); 241 mutex_spin_enter(&chgp->p_smutex); 242 if (onice != chgp->p_nice) { 243 mutex_spin_exit(&chgp->p_smutex); 244 goto again; 245 } 246 sched_nice(chgp, n); 247 mutex_spin_exit(&chgp->p_smutex); 248 return (0); 249 } 250 251 /* ARGSUSED */ 252 int 253 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap, register_t *retval) 254 { 255 /* { 256 syscallarg(int) which; 257 syscallarg(const struct rlimit *) rlp; 258 } */ 259 int which = SCARG(uap, which); 260 struct rlimit alim; 261 int error; 262 263 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 264 if (error) 265 return (error); 266 return (dosetrlimit(l, l->l_proc, which, &alim)); 267 } 268 269 int 270 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 271 { 272 struct rlimit *alimp; 273 int error; 274 275 if ((u_int)which >= RLIM_NLIMITS) 276 return (EINVAL); 277 278 if (limp->rlim_cur < 0 || limp->rlim_max < 0) 279 return (EINVAL); 280 281 if (limp->rlim_cur > limp->rlim_max) { 282 /* 283 * This is programming error. According to SUSv2, we should 284 * return error in this case. 285 */ 286 return (EINVAL); 287 } 288 289 alimp = &p->p_rlimit[which]; 290 /* if we don't change the value, no need to limcopy() */ 291 if (limp->rlim_cur == alimp->rlim_cur && 292 limp->rlim_max == alimp->rlim_max) 293 return 0; 294 295 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 296 p, limp, KAUTH_ARG(which), NULL); 297 if (error) 298 return (error); 299 300 lim_privatise(p, false); 301 /* p->p_limit is now unchangeable */ 302 alimp = &p->p_rlimit[which]; 303 304 switch (which) { 305 306 case RLIMIT_DATA: 307 if (limp->rlim_cur > maxdmap) 308 limp->rlim_cur = maxdmap; 309 if (limp->rlim_max > maxdmap) 310 limp->rlim_max = maxdmap; 311 break; 312 313 case RLIMIT_STACK: 314 if (limp->rlim_cur > maxsmap) 315 limp->rlim_cur = maxsmap; 316 if (limp->rlim_max > maxsmap) 317 limp->rlim_max = maxsmap; 318 319 /* 320 * Return EINVAL if the new stack size limit is lower than 321 * current usage. Otherwise, the process would get SIGSEGV the 322 * moment it would try to access anything on it's current stack. 323 * This conforms to SUSv2. 324 */ 325 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 326 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 327 return (EINVAL); 328 } 329 330 /* 331 * Stack is allocated to the max at exec time with 332 * only "rlim_cur" bytes accessible (In other words, 333 * allocates stack dividing two contiguous regions at 334 * "rlim_cur" bytes boundary). 335 * 336 * Since allocation is done in terms of page, roundup 337 * "rlim_cur" (otherwise, contiguous regions 338 * overlap). If stack limit is going up make more 339 * accessible, if going down make inaccessible. 340 */ 341 limp->rlim_cur = round_page(limp->rlim_cur); 342 if (limp->rlim_cur != alimp->rlim_cur) { 343 vaddr_t addr; 344 vsize_t size; 345 vm_prot_t prot; 346 347 if (limp->rlim_cur > alimp->rlim_cur) { 348 prot = VM_PROT_READ | VM_PROT_WRITE; 349 size = limp->rlim_cur - alimp->rlim_cur; 350 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 351 limp->rlim_cur; 352 } else { 353 prot = VM_PROT_NONE; 354 size = alimp->rlim_cur - limp->rlim_cur; 355 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 356 alimp->rlim_cur; 357 } 358 (void) uvm_map_protect(&p->p_vmspace->vm_map, 359 addr, addr+size, prot, false); 360 } 361 break; 362 363 case RLIMIT_NOFILE: 364 if (limp->rlim_cur > maxfiles) 365 limp->rlim_cur = maxfiles; 366 if (limp->rlim_max > maxfiles) 367 limp->rlim_max = maxfiles; 368 break; 369 370 case RLIMIT_NPROC: 371 if (limp->rlim_cur > maxproc) 372 limp->rlim_cur = maxproc; 373 if (limp->rlim_max > maxproc) 374 limp->rlim_max = maxproc; 375 break; 376 } 377 378 mutex_enter(&p->p_limit->pl_lock); 379 *alimp = *limp; 380 mutex_exit(&p->p_limit->pl_lock); 381 return (0); 382 } 383 384 /* ARGSUSED */ 385 int 386 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, register_t *retval) 387 { 388 /* { 389 syscallarg(int) which; 390 syscallarg(struct rlimit *) rlp; 391 } */ 392 struct proc *p = l->l_proc; 393 int which = SCARG(uap, which); 394 struct rlimit rl; 395 396 if ((u_int)which >= RLIM_NLIMITS) 397 return (EINVAL); 398 399 mutex_enter(&p->p_mutex); 400 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 401 mutex_exit(&p->p_mutex); 402 403 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 404 } 405 406 /* 407 * Transform the running time and tick information in proc p into user, 408 * system, and interrupt time usage. 409 * 410 * Should be called with p->p_smutex held unless called from exit1(). 411 */ 412 void 413 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 414 struct timeval *ip, struct timeval *rp) 415 { 416 uint64_t u, st, ut, it, tot; 417 struct lwp *l; 418 struct bintime tm; 419 struct timeval tv; 420 421 mutex_spin_enter(&p->p_stmutex); 422 st = p->p_sticks; 423 ut = p->p_uticks; 424 it = p->p_iticks; 425 mutex_spin_exit(&p->p_stmutex); 426 427 tm = p->p_rtime; 428 429 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 430 lwp_lock(l); 431 bintime_add(&tm, &l->l_rtime); 432 if ((l->l_flag & LW_RUNNING) != 0) { 433 struct bintime diff; 434 /* 435 * Adjust for the current time slice. This is 436 * actually fairly important since the error 437 * here is on the order of a time quantum, 438 * which is much greater than the sampling 439 * error. 440 */ 441 binuptime(&diff); 442 bintime_sub(&diff, &l->l_stime); 443 bintime_add(&tm, &diff); 444 } 445 lwp_unlock(l); 446 } 447 448 tot = st + ut + it; 449 bintime2timeval(&tm, &tv); 450 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 451 452 if (tot == 0) { 453 /* No ticks, so can't use to share time out, split 50-50 */ 454 st = ut = u / 2; 455 } else { 456 st = (u * st) / tot; 457 ut = (u * ut) / tot; 458 } 459 if (sp != NULL) { 460 sp->tv_sec = st / 1000000; 461 sp->tv_usec = st % 1000000; 462 } 463 if (up != NULL) { 464 up->tv_sec = ut / 1000000; 465 up->tv_usec = ut % 1000000; 466 } 467 if (ip != NULL) { 468 if (it != 0) 469 it = (u * it) / tot; 470 ip->tv_sec = it / 1000000; 471 ip->tv_usec = it % 1000000; 472 } 473 if (rp != NULL) { 474 *rp = tv; 475 } 476 } 477 478 /* ARGSUSED */ 479 int 480 sys_getrusage(struct lwp *l, const struct sys_getrusage_args *uap, register_t *retval) 481 { 482 /* { 483 syscallarg(int) who; 484 syscallarg(struct rusage *) rusage; 485 } */ 486 struct rusage ru; 487 struct proc *p = l->l_proc; 488 489 switch (SCARG(uap, who)) { 490 case RUSAGE_SELF: 491 mutex_enter(&p->p_smutex); 492 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 493 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 494 mutex_exit(&p->p_smutex); 495 break; 496 497 case RUSAGE_CHILDREN: 498 mutex_enter(&p->p_smutex); 499 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 500 mutex_exit(&p->p_smutex); 501 break; 502 503 default: 504 return EINVAL; 505 } 506 507 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 508 } 509 510 void 511 ruadd(struct rusage *ru, struct rusage *ru2) 512 { 513 long *ip, *ip2; 514 int i; 515 516 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 517 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 518 if (ru->ru_maxrss < ru2->ru_maxrss) 519 ru->ru_maxrss = ru2->ru_maxrss; 520 ip = &ru->ru_first; ip2 = &ru2->ru_first; 521 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 522 *ip++ += *ip2++; 523 } 524 525 /* 526 * Make a copy of the plimit structure. 527 * We share these structures copy-on-write after fork, 528 * and copy when a limit is changed. 529 * 530 * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure 531 * we are copying to change beneath our feet! 532 */ 533 struct plimit * 534 lim_copy(struct plimit *lim) 535 { 536 struct plimit *newlim; 537 char *corename; 538 size_t alen, len; 539 540 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 541 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 542 newlim->pl_flags = 0; 543 newlim->pl_refcnt = 1; 544 newlim->pl_sv_limit = NULL; 545 546 mutex_enter(&lim->pl_lock); 547 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 548 sizeof(struct rlimit) * RLIM_NLIMITS); 549 550 alen = 0; 551 corename = NULL; 552 for (;;) { 553 if (lim->pl_corename == defcorename) { 554 newlim->pl_corename = defcorename; 555 break; 556 } 557 len = strlen(lim->pl_corename) + 1; 558 if (len <= alen) { 559 newlim->pl_corename = corename; 560 memcpy(corename, lim->pl_corename, len); 561 corename = NULL; 562 break; 563 } 564 mutex_exit(&lim->pl_lock); 565 if (corename != NULL) 566 free(corename, M_TEMP); 567 alen = len; 568 corename = malloc(alen, M_TEMP, M_WAITOK); 569 mutex_enter(&lim->pl_lock); 570 } 571 mutex_exit(&lim->pl_lock); 572 if (corename != NULL) 573 free(corename, M_TEMP); 574 return newlim; 575 } 576 577 void 578 lim_addref(struct plimit *lim) 579 { 580 atomic_inc_uint(&lim->pl_refcnt); 581 } 582 583 /* 584 * Give a process it's own private plimit structure. 585 * This will only be shared (in fork) if modifications are to be shared. 586 */ 587 void 588 lim_privatise(struct proc *p, bool set_shared) 589 { 590 struct plimit *lim, *newlim; 591 592 lim = p->p_limit; 593 if (lim->pl_flags & PL_WRITEABLE) { 594 if (set_shared) 595 lim->pl_flags |= PL_SHAREMOD; 596 return; 597 } 598 599 if (set_shared && lim->pl_flags & PL_SHAREMOD) 600 return; 601 602 newlim = lim_copy(lim); 603 604 mutex_enter(&p->p_mutex); 605 if (p->p_limit->pl_flags & PL_WRITEABLE) { 606 /* Someone crept in while we were busy */ 607 mutex_exit(&p->p_mutex); 608 limfree(newlim); 609 if (set_shared) 610 p->p_limit->pl_flags |= PL_SHAREMOD; 611 return; 612 } 613 614 /* 615 * Since most accesses to p->p_limit aren't locked, we must not 616 * delete the old limit structure yet. 617 */ 618 newlim->pl_sv_limit = p->p_limit; 619 newlim->pl_flags |= PL_WRITEABLE; 620 if (set_shared) 621 newlim->pl_flags |= PL_SHAREMOD; 622 p->p_limit = newlim; 623 mutex_exit(&p->p_mutex); 624 } 625 626 void 627 limfree(struct plimit *lim) 628 { 629 struct plimit *sv_lim; 630 631 do { 632 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) 633 return; 634 if (lim->pl_corename != defcorename) 635 free(lim->pl_corename, M_TEMP); 636 sv_lim = lim->pl_sv_limit; 637 mutex_destroy(&lim->pl_lock); 638 pool_cache_put(plimit_cache, lim); 639 } while ((lim = sv_lim) != NULL); 640 } 641 642 struct pstats * 643 pstatscopy(struct pstats *ps) 644 { 645 646 struct pstats *newps; 647 648 newps = pool_cache_get(pstats_cache, PR_WAITOK); 649 650 memset(&newps->pstat_startzero, 0, 651 (unsigned) ((char *)&newps->pstat_endzero - 652 (char *)&newps->pstat_startzero)); 653 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 654 ((char *)&newps->pstat_endcopy - 655 (char *)&newps->pstat_startcopy)); 656 657 return (newps); 658 659 } 660 661 void 662 pstatsfree(struct pstats *ps) 663 { 664 665 pool_cache_put(pstats_cache, ps); 666 } 667 668 /* 669 * sysctl interface in five parts 670 */ 671 672 /* 673 * a routine for sysctl proc subtree helpers that need to pick a valid 674 * process by pid. 675 */ 676 static int 677 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 678 { 679 struct proc *ptmp; 680 int error = 0; 681 682 if (pid == PROC_CURPROC) 683 ptmp = l->l_proc; 684 else if ((ptmp = pfind(pid)) == NULL) 685 error = ESRCH; 686 687 *p2 = ptmp; 688 return (error); 689 } 690 691 /* 692 * sysctl helper routine for setting a process's specific corefile 693 * name. picks the process based on the given pid and checks the 694 * correctness of the new value. 695 */ 696 static int 697 sysctl_proc_corename(SYSCTLFN_ARGS) 698 { 699 struct proc *ptmp; 700 struct plimit *lim; 701 int error = 0, len; 702 char *cname; 703 char *ocore; 704 char *tmp; 705 struct sysctlnode node; 706 707 /* 708 * is this all correct? 709 */ 710 if (namelen != 0) 711 return (EINVAL); 712 if (name[-1] != PROC_PID_CORENAME) 713 return (EINVAL); 714 715 /* 716 * whom are we tweaking? 717 */ 718 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 719 if (error) 720 return (error); 721 722 /* XXX this should be in p_find() */ 723 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 724 ptmp, NULL, NULL, NULL); 725 if (error) 726 return (error); 727 728 /* 729 * let them modify a temporary copy of the core name 730 */ 731 cname = PNBUF_GET(); 732 lim = ptmp->p_limit; 733 mutex_enter(&lim->pl_lock); 734 strlcpy(cname, lim->pl_corename, MAXPATHLEN); 735 mutex_exit(&lim->pl_lock); 736 737 node = *rnode; 738 node.sysctl_data = cname; 739 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 740 741 /* 742 * if that failed, or they have nothing new to say, or we've 743 * heard it before... 744 */ 745 if (error || newp == NULL) 746 goto done; 747 lim = ptmp->p_limit; 748 mutex_enter(&lim->pl_lock); 749 error = strcmp(cname, lim->pl_corename); 750 mutex_exit(&lim->pl_lock); 751 if (error == 0) 752 /* Unchanged */ 753 goto done; 754 755 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 756 ptmp, cname, NULL, NULL); 757 if (error) 758 return (error); 759 760 /* 761 * no error yet and cname now has the new core name in it. 762 * let's see if it looks acceptable. it must be either "core" 763 * or end in ".core" or "/core". 764 */ 765 len = strlen(cname); 766 if (len < 4) { 767 error = EINVAL; 768 } else if (strcmp(cname + len - 4, "core") != 0) { 769 error = EINVAL; 770 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 771 error = EINVAL; 772 } 773 if (error != 0) { 774 goto done; 775 } 776 777 /* 778 * hmm...looks good. now...where do we put it? 779 */ 780 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 781 if (tmp == NULL) { 782 error = ENOMEM; 783 goto done; 784 } 785 memcpy(tmp, cname, len + 1); 786 787 lim_privatise(ptmp, false); 788 lim = ptmp->p_limit; 789 mutex_enter(&lim->pl_lock); 790 ocore = lim->pl_corename; 791 lim->pl_corename = tmp; 792 mutex_exit(&lim->pl_lock); 793 if (ocore != defcorename) 794 free(ocore, M_TEMP); 795 796 done: 797 PNBUF_PUT(cname); 798 return error; 799 } 800 801 /* 802 * sysctl helper routine for checking/setting a process's stop flags, 803 * one for fork and one for exec. 804 */ 805 static int 806 sysctl_proc_stop(SYSCTLFN_ARGS) 807 { 808 struct proc *ptmp; 809 int i, f, error = 0; 810 struct sysctlnode node; 811 812 if (namelen != 0) 813 return (EINVAL); 814 815 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 816 if (error) 817 return (error); 818 819 /* XXX this should be in p_find() */ 820 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 821 ptmp, NULL, NULL, NULL); 822 if (error) 823 return (error); 824 825 switch (rnode->sysctl_num) { 826 case PROC_PID_STOPFORK: 827 f = PS_STOPFORK; 828 break; 829 case PROC_PID_STOPEXEC: 830 f = PS_STOPEXEC; 831 break; 832 case PROC_PID_STOPEXIT: 833 f = PS_STOPEXIT; 834 break; 835 default: 836 return (EINVAL); 837 } 838 839 i = (ptmp->p_flag & f) ? 1 : 0; 840 node = *rnode; 841 node.sysctl_data = &i; 842 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 843 if (error || newp == NULL) 844 return (error); 845 846 mutex_enter(&ptmp->p_smutex); 847 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 848 ptmp, KAUTH_ARG(f), NULL, NULL); 849 if (error) 850 return (error); 851 if (i) 852 ptmp->p_sflag |= f; 853 else 854 ptmp->p_sflag &= ~f; 855 mutex_exit(&ptmp->p_smutex); 856 857 return (0); 858 } 859 860 /* 861 * sysctl helper routine for a process's rlimits as exposed by sysctl. 862 */ 863 static int 864 sysctl_proc_plimit(SYSCTLFN_ARGS) 865 { 866 struct proc *ptmp; 867 u_int limitno; 868 int which, error = 0; 869 struct rlimit alim; 870 struct sysctlnode node; 871 872 if (namelen != 0) 873 return (EINVAL); 874 875 which = name[-1]; 876 if (which != PROC_PID_LIMIT_TYPE_SOFT && 877 which != PROC_PID_LIMIT_TYPE_HARD) 878 return (EINVAL); 879 880 limitno = name[-2] - 1; 881 if (limitno >= RLIM_NLIMITS) 882 return (EINVAL); 883 884 if (name[-3] != PROC_PID_LIMIT) 885 return (EINVAL); 886 887 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 888 if (error) 889 return (error); 890 891 /* XXX this should be in p_find() */ 892 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 893 ptmp, NULL, NULL, NULL); 894 if (error) 895 return (error); 896 897 node = *rnode; 898 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 899 if (which == PROC_PID_LIMIT_TYPE_HARD) 900 node.sysctl_data = &alim.rlim_max; 901 else 902 node.sysctl_data = &alim.rlim_cur; 903 904 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 905 if (error || newp == NULL) 906 return (error); 907 908 return (dosetrlimit(l, ptmp, limitno, &alim)); 909 } 910 911 /* 912 * and finally, the actually glue that sticks it to the tree 913 */ 914 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup") 915 { 916 917 sysctl_createv(clog, 0, NULL, NULL, 918 CTLFLAG_PERMANENT, 919 CTLTYPE_NODE, "proc", NULL, 920 NULL, 0, NULL, 0, 921 CTL_PROC, CTL_EOL); 922 sysctl_createv(clog, 0, NULL, NULL, 923 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 924 CTLTYPE_NODE, "curproc", 925 SYSCTL_DESCR("Per-process settings"), 926 NULL, 0, NULL, 0, 927 CTL_PROC, PROC_CURPROC, CTL_EOL); 928 929 sysctl_createv(clog, 0, NULL, NULL, 930 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 931 CTLTYPE_STRING, "corename", 932 SYSCTL_DESCR("Core file name"), 933 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 934 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 935 sysctl_createv(clog, 0, NULL, NULL, 936 CTLFLAG_PERMANENT, 937 CTLTYPE_NODE, "rlimit", 938 SYSCTL_DESCR("Process limits"), 939 NULL, 0, NULL, 0, 940 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 941 942 #define create_proc_plimit(s, n) do { \ 943 sysctl_createv(clog, 0, NULL, NULL, \ 944 CTLFLAG_PERMANENT, \ 945 CTLTYPE_NODE, s, \ 946 SYSCTL_DESCR("Process " s " limits"), \ 947 NULL, 0, NULL, 0, \ 948 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 949 CTL_EOL); \ 950 sysctl_createv(clog, 0, NULL, NULL, \ 951 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 952 CTLTYPE_QUAD, "soft", \ 953 SYSCTL_DESCR("Process soft " s " limit"), \ 954 sysctl_proc_plimit, 0, NULL, 0, \ 955 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 956 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 957 sysctl_createv(clog, 0, NULL, NULL, \ 958 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 959 CTLTYPE_QUAD, "hard", \ 960 SYSCTL_DESCR("Process hard " s " limit"), \ 961 sysctl_proc_plimit, 0, NULL, 0, \ 962 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 963 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 964 } while (0/*CONSTCOND*/) 965 966 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 967 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 968 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 969 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 970 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 971 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 972 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 973 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 974 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 975 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 976 977 #undef create_proc_plimit 978 979 sysctl_createv(clog, 0, NULL, NULL, 980 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 981 CTLTYPE_INT, "stopfork", 982 SYSCTL_DESCR("Stop process at fork(2)"), 983 sysctl_proc_stop, 0, NULL, 0, 984 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 985 sysctl_createv(clog, 0, NULL, NULL, 986 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 987 CTLTYPE_INT, "stopexec", 988 SYSCTL_DESCR("Stop process at execve(2)"), 989 sysctl_proc_stop, 0, NULL, 0, 990 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 991 sysctl_createv(clog, 0, NULL, NULL, 992 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 993 CTLTYPE_INT, "stopexit", 994 SYSCTL_DESCR("Stop process before completing exit"), 995 sysctl_proc_stop, 0, NULL, 0, 996 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 997 } 998 999 void 1000 uid_init(void) 1001 { 1002 1003 /* 1004 * XXXSMP This could be at IPL_SOFTNET, but for now we want 1005 * to to be deadlock free, so it must be at IPL_VM. 1006 */ 1007 mutex_init(&uihashtbl_lock, MUTEX_DEFAULT, IPL_VM); 1008 1009 /* 1010 * Ensure that uid 0 is always in the user hash table, as 1011 * sbreserve() expects it available from interrupt context. 1012 */ 1013 (void)uid_find(0); 1014 } 1015 1016 struct uidinfo * 1017 uid_find(uid_t uid) 1018 { 1019 struct uidinfo *uip; 1020 struct uidinfo *newuip = NULL; 1021 struct uihashhead *uipp; 1022 1023 uipp = UIHASH(uid); 1024 1025 again: 1026 mutex_enter(&uihashtbl_lock); 1027 LIST_FOREACH(uip, uipp, ui_hash) 1028 if (uip->ui_uid == uid) { 1029 mutex_exit(&uihashtbl_lock); 1030 if (newuip) { 1031 mutex_destroy(&newuip->ui_lock); 1032 free(newuip, M_PROC); 1033 } 1034 return uip; 1035 } 1036 if (newuip == NULL) { 1037 mutex_exit(&uihashtbl_lock); 1038 /* Must not be called from interrupt context. */ 1039 newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO); 1040 /* XXX this could be IPL_SOFTNET */ 1041 mutex_init(&newuip->ui_lock, MUTEX_DEFAULT, IPL_VM); 1042 goto again; 1043 } 1044 uip = newuip; 1045 1046 LIST_INSERT_HEAD(uipp, uip, ui_hash); 1047 uip->ui_uid = uid; 1048 mutex_exit(&uihashtbl_lock); 1049 1050 return uip; 1051 } 1052 1053 /* 1054 * Change the count associated with number of processes 1055 * a given user is using. 1056 */ 1057 int 1058 chgproccnt(uid_t uid, int diff) 1059 { 1060 struct uidinfo *uip; 1061 1062 if (diff == 0) 1063 return 0; 1064 1065 uip = uid_find(uid); 1066 mutex_enter(&uip->ui_lock); 1067 uip->ui_proccnt += diff; 1068 KASSERT(uip->ui_proccnt >= 0); 1069 mutex_exit(&uip->ui_lock); 1070 return uip->ui_proccnt; 1071 } 1072 1073 int 1074 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax) 1075 { 1076 rlim_t nsb; 1077 1078 mutex_enter(&uip->ui_lock); 1079 nsb = uip->ui_sbsize + to - *hiwat; 1080 if (to > *hiwat && nsb > xmax) { 1081 mutex_exit(&uip->ui_lock); 1082 return 0; 1083 } 1084 *hiwat = to; 1085 uip->ui_sbsize = nsb; 1086 KASSERT(uip->ui_sbsize >= 0); 1087 mutex_exit(&uip->ui_lock); 1088 return 1; 1089 } 1090
Indexes created Sun Sep 21 19:09:51 GMT 2025