kern_resource.c revision 1.129
1 /* $NetBSD: kern_resource.c,v 1.129 2007/12/22 01:14:54 yamt 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 1.129 2007/12/22 01:14:54 yamt 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 /* 72 * Resource controls and accounting. 73 */ 74 75 int 76 sys_getpriority(struct lwp *l, const struct sys_getpriority_args *uap, register_t *retval) 77 { 78 /* { 79 syscallarg(int) which; 80 syscallarg(id_t) who; 81 } */ 82 struct proc *curp = l->l_proc, *p; 83 int low = NZERO + PRIO_MAX + 1; 84 int who = SCARG(uap, who); 85 86 mutex_enter(&proclist_lock); 87 switch (SCARG(uap, which)) { 88 case PRIO_PROCESS: 89 if (who == 0) 90 p = curp; 91 else 92 p = p_find(who, PFIND_LOCKED); 93 if (p != NULL) 94 low = p->p_nice; 95 break; 96 97 case PRIO_PGRP: { 98 struct pgrp *pg; 99 100 if (who == 0) 101 pg = curp->p_pgrp; 102 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 103 break; 104 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 105 if (p->p_nice < low) 106 low = p->p_nice; 107 } 108 break; 109 } 110 111 case PRIO_USER: 112 if (who == 0) 113 who = (int)kauth_cred_geteuid(l->l_cred); 114 PROCLIST_FOREACH(p, &allproc) { 115 mutex_enter(&p->p_mutex); 116 if (kauth_cred_geteuid(p->p_cred) == 117 (uid_t)who && p->p_nice < low) 118 low = p->p_nice; 119 mutex_exit(&p->p_mutex); 120 } 121 break; 122 123 default: 124 mutex_exit(&proclist_lock); 125 return (EINVAL); 126 } 127 mutex_exit(&proclist_lock); 128 129 if (low == NZERO + PRIO_MAX + 1) 130 return (ESRCH); 131 *retval = low - NZERO; 132 return (0); 133 } 134 135 /* ARGSUSED */ 136 int 137 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap, register_t *retval) 138 { 139 /* { 140 syscallarg(int) which; 141 syscallarg(id_t) who; 142 syscallarg(int) prio; 143 } */ 144 struct proc *curp = l->l_proc, *p; 145 int found = 0, error = 0; 146 int who = SCARG(uap, who); 147 148 mutex_enter(&proclist_lock); 149 switch (SCARG(uap, which)) { 150 case PRIO_PROCESS: 151 if (who == 0) 152 p = curp; 153 else 154 p = p_find(who, PFIND_LOCKED); 155 if (p != 0) { 156 mutex_enter(&p->p_mutex); 157 error = donice(l, p, SCARG(uap, prio)); 158 mutex_exit(&p->p_mutex); 159 } 160 found++; 161 break; 162 163 case PRIO_PGRP: { 164 struct pgrp *pg; 165 166 if (who == 0) 167 pg = curp->p_pgrp; 168 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 169 break; 170 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 171 mutex_enter(&p->p_mutex); 172 error = donice(l, p, SCARG(uap, prio)); 173 mutex_exit(&p->p_mutex); 174 found++; 175 } 176 break; 177 } 178 179 case PRIO_USER: 180 if (who == 0) 181 who = (int)kauth_cred_geteuid(l->l_cred); 182 PROCLIST_FOREACH(p, &allproc) { 183 mutex_enter(&p->p_mutex); 184 if (kauth_cred_geteuid(p->p_cred) == 185 (uid_t)SCARG(uap, who)) { 186 error = donice(l, p, SCARG(uap, prio)); 187 found++; 188 } 189 mutex_exit(&p->p_mutex); 190 } 191 break; 192 193 default: 194 error = EINVAL; 195 break; 196 } 197 mutex_exit(&proclist_lock); 198 if (found == 0) 199 return (ESRCH); 200 return (error); 201 } 202 203 /* 204 * Renice a process. 205 * 206 * Call with the target process' credentials locked. 207 */ 208 int 209 donice(struct lwp *l, struct proc *chgp, int n) 210 { 211 kauth_cred_t cred = l->l_cred; 212 int onice; 213 214 KASSERT(mutex_owned(&chgp->p_mutex)); 215 216 if (n > PRIO_MAX) 217 n = PRIO_MAX; 218 if (n < PRIO_MIN) 219 n = PRIO_MIN; 220 n += NZERO; 221 onice = chgp->p_nice; 222 onice = chgp->p_nice; 223 224 again: 225 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 226 KAUTH_ARG(n), NULL, NULL)) 227 return (EACCES); 228 mutex_spin_enter(&chgp->p_smutex); 229 if (onice != chgp->p_nice) { 230 mutex_spin_exit(&chgp->p_smutex); 231 goto again; 232 } 233 sched_nice(chgp, n); 234 mutex_spin_exit(&chgp->p_smutex); 235 return (0); 236 } 237 238 /* ARGSUSED */ 239 int 240 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap, register_t *retval) 241 { 242 /* { 243 syscallarg(int) which; 244 syscallarg(const struct rlimit *) rlp; 245 } */ 246 int which = SCARG(uap, which); 247 struct rlimit alim; 248 int error; 249 250 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 251 if (error) 252 return (error); 253 return (dosetrlimit(l, l->l_proc, which, &alim)); 254 } 255 256 int 257 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 258 { 259 struct rlimit *alimp; 260 int error; 261 262 if ((u_int)which >= RLIM_NLIMITS) 263 return (EINVAL); 264 265 if (limp->rlim_cur < 0 || limp->rlim_max < 0) 266 return (EINVAL); 267 268 if (limp->rlim_cur > limp->rlim_max) { 269 /* 270 * This is programming error. According to SUSv2, we should 271 * return error in this case. 272 */ 273 return (EINVAL); 274 } 275 276 alimp = &p->p_rlimit[which]; 277 /* if we don't change the value, no need to limcopy() */ 278 if (limp->rlim_cur == alimp->rlim_cur && 279 limp->rlim_max == alimp->rlim_max) 280 return 0; 281 282 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 283 p, limp, KAUTH_ARG(which), NULL); 284 if (error) 285 return (error); 286 287 lim_privatise(p, false); 288 /* p->p_limit is now unchangeable */ 289 alimp = &p->p_rlimit[which]; 290 291 switch (which) { 292 293 case RLIMIT_DATA: 294 if (limp->rlim_cur > maxdmap) 295 limp->rlim_cur = maxdmap; 296 if (limp->rlim_max > maxdmap) 297 limp->rlim_max = maxdmap; 298 break; 299 300 case RLIMIT_STACK: 301 if (limp->rlim_cur > maxsmap) 302 limp->rlim_cur = maxsmap; 303 if (limp->rlim_max > maxsmap) 304 limp->rlim_max = maxsmap; 305 306 /* 307 * Return EINVAL if the new stack size limit is lower than 308 * current usage. Otherwise, the process would get SIGSEGV the 309 * moment it would try to access anything on it's current stack. 310 * This conforms to SUSv2. 311 */ 312 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 313 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 314 return (EINVAL); 315 } 316 317 /* 318 * Stack is allocated to the max at exec time with 319 * only "rlim_cur" bytes accessible (In other words, 320 * allocates stack dividing two contiguous regions at 321 * "rlim_cur" bytes boundary). 322 * 323 * Since allocation is done in terms of page, roundup 324 * "rlim_cur" (otherwise, contiguous regions 325 * overlap). If stack limit is going up make more 326 * accessible, if going down make inaccessible. 327 */ 328 limp->rlim_cur = round_page(limp->rlim_cur); 329 if (limp->rlim_cur != alimp->rlim_cur) { 330 vaddr_t addr; 331 vsize_t size; 332 vm_prot_t prot; 333 334 if (limp->rlim_cur > alimp->rlim_cur) { 335 prot = VM_PROT_READ | VM_PROT_WRITE; 336 size = limp->rlim_cur - alimp->rlim_cur; 337 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 338 limp->rlim_cur; 339 } else { 340 prot = VM_PROT_NONE; 341 size = alimp->rlim_cur - limp->rlim_cur; 342 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 343 alimp->rlim_cur; 344 } 345 (void) uvm_map_protect(&p->p_vmspace->vm_map, 346 addr, addr+size, prot, false); 347 } 348 break; 349 350 case RLIMIT_NOFILE: 351 if (limp->rlim_cur > maxfiles) 352 limp->rlim_cur = maxfiles; 353 if (limp->rlim_max > maxfiles) 354 limp->rlim_max = maxfiles; 355 break; 356 357 case RLIMIT_NPROC: 358 if (limp->rlim_cur > maxproc) 359 limp->rlim_cur = maxproc; 360 if (limp->rlim_max > maxproc) 361 limp->rlim_max = maxproc; 362 break; 363 } 364 365 mutex_enter(&p->p_limit->pl_lock); 366 *alimp = *limp; 367 mutex_exit(&p->p_limit->pl_lock); 368 return (0); 369 } 370 371 /* ARGSUSED */ 372 int 373 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, register_t *retval) 374 { 375 /* { 376 syscallarg(int) which; 377 syscallarg(struct rlimit *) rlp; 378 } */ 379 struct proc *p = l->l_proc; 380 int which = SCARG(uap, which); 381 struct rlimit rl; 382 383 if ((u_int)which >= RLIM_NLIMITS) 384 return (EINVAL); 385 386 mutex_enter(&p->p_mutex); 387 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 388 mutex_exit(&p->p_mutex); 389 390 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 391 } 392 393 /* 394 * Transform the running time and tick information in proc p into user, 395 * system, and interrupt time usage. 396 * 397 * Should be called with p->p_smutex held unless called from exit1(). 398 */ 399 void 400 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 401 struct timeval *ip, struct timeval *rp) 402 { 403 uint64_t u, st, ut, it, tot; 404 struct lwp *l; 405 struct bintime tm; 406 struct timeval tv; 407 408 mutex_spin_enter(&p->p_stmutex); 409 st = p->p_sticks; 410 ut = p->p_uticks; 411 it = p->p_iticks; 412 mutex_spin_exit(&p->p_stmutex); 413 414 tm = p->p_rtime; 415 416 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 417 lwp_lock(l); 418 bintime_add(&tm, &l->l_rtime); 419 if ((l->l_flag & LW_RUNNING) != 0) { 420 struct bintime diff; 421 /* 422 * Adjust for the current time slice. This is 423 * actually fairly important since the error 424 * here is on the order of a time quantum, 425 * which is much greater than the sampling 426 * error. 427 */ 428 binuptime(&diff); 429 bintime_sub(&diff, &l->l_stime); 430 bintime_add(&tm, &diff); 431 } 432 lwp_unlock(l); 433 } 434 435 tot = st + ut + it; 436 bintime2timeval(&tm, &tv); 437 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 438 439 if (tot == 0) { 440 /* No ticks, so can't use to share time out, split 50-50 */ 441 st = ut = u / 2; 442 } else { 443 st = (u * st) / tot; 444 ut = (u * ut) / tot; 445 } 446 if (sp != NULL) { 447 sp->tv_sec = st / 1000000; 448 sp->tv_usec = st % 1000000; 449 } 450 if (up != NULL) { 451 up->tv_sec = ut / 1000000; 452 up->tv_usec = ut % 1000000; 453 } 454 if (ip != NULL) { 455 if (it != 0) 456 it = (u * it) / tot; 457 ip->tv_sec = it / 1000000; 458 ip->tv_usec = it % 1000000; 459 } 460 if (rp != NULL) { 461 *rp = tv; 462 } 463 } 464 465 /* ARGSUSED */ 466 int 467 sys_getrusage(struct lwp *l, const struct sys_getrusage_args *uap, register_t *retval) 468 { 469 /* { 470 syscallarg(int) who; 471 syscallarg(struct rusage *) rusage; 472 } */ 473 struct rusage ru; 474 struct proc *p = l->l_proc; 475 476 switch (SCARG(uap, who)) { 477 case RUSAGE_SELF: 478 mutex_enter(&p->p_smutex); 479 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 480 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 481 mutex_exit(&p->p_smutex); 482 break; 483 484 case RUSAGE_CHILDREN: 485 mutex_enter(&p->p_smutex); 486 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 487 mutex_exit(&p->p_smutex); 488 break; 489 490 default: 491 return EINVAL; 492 } 493 494 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 495 } 496 497 void 498 ruadd(struct rusage *ru, struct rusage *ru2) 499 { 500 long *ip, *ip2; 501 int i; 502 503 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 504 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 505 if (ru->ru_maxrss < ru2->ru_maxrss) 506 ru->ru_maxrss = ru2->ru_maxrss; 507 ip = &ru->ru_first; ip2 = &ru2->ru_first; 508 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 509 *ip++ += *ip2++; 510 } 511 512 /* 513 * Make a copy of the plimit structure. 514 * We share these structures copy-on-write after fork, 515 * and copy when a limit is changed. 516 * 517 * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure 518 * we are copying to change beneath our feet! 519 */ 520 struct plimit * 521 lim_copy(struct plimit *lim) 522 { 523 struct plimit *newlim; 524 char *corename; 525 size_t alen, len; 526 527 newlim = pool_get(&plimit_pool, PR_WAITOK); 528 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 529 newlim->pl_flags = 0; 530 newlim->pl_refcnt = 1; 531 newlim->pl_sv_limit = NULL; 532 533 mutex_enter(&lim->pl_lock); 534 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 535 sizeof(struct rlimit) * RLIM_NLIMITS); 536 537 alen = 0; 538 corename = NULL; 539 for (;;) { 540 if (lim->pl_corename == defcorename) { 541 newlim->pl_corename = defcorename; 542 break; 543 } 544 len = strlen(lim->pl_corename) + 1; 545 if (len <= alen) { 546 newlim->pl_corename = corename; 547 memcpy(corename, lim->pl_corename, len); 548 corename = NULL; 549 break; 550 } 551 mutex_exit(&lim->pl_lock); 552 if (corename != NULL) 553 free(corename, M_TEMP); 554 alen = len; 555 corename = malloc(alen, M_TEMP, M_WAITOK); 556 mutex_enter(&lim->pl_lock); 557 } 558 mutex_exit(&lim->pl_lock); 559 if (corename != NULL) 560 free(corename, M_TEMP); 561 return newlim; 562 } 563 564 void 565 lim_addref(struct plimit *lim) 566 { 567 atomic_inc_uint(&lim->pl_refcnt); 568 } 569 570 /* 571 * Give a process it's own private plimit structure. 572 * This will only be shared (in fork) if modifications are to be shared. 573 */ 574 void 575 lim_privatise(struct proc *p, bool set_shared) 576 { 577 struct plimit *lim, *newlim; 578 579 lim = p->p_limit; 580 if (lim->pl_flags & PL_WRITEABLE) { 581 if (set_shared) 582 lim->pl_flags |= PL_SHAREMOD; 583 return; 584 } 585 586 if (set_shared && lim->pl_flags & PL_SHAREMOD) 587 return; 588 589 newlim = lim_copy(lim); 590 591 mutex_enter(&p->p_mutex); 592 if (p->p_limit->pl_flags & PL_WRITEABLE) { 593 /* Someone crept in while we were busy */ 594 mutex_exit(&p->p_mutex); 595 limfree(newlim); 596 if (set_shared) 597 p->p_limit->pl_flags |= PL_SHAREMOD; 598 return; 599 } 600 601 /* 602 * Since most accesses to p->p_limit aren't locked, we must not 603 * delete the old limit structure yet. 604 */ 605 newlim->pl_sv_limit = p->p_limit; 606 newlim->pl_flags |= PL_WRITEABLE; 607 if (set_shared) 608 newlim->pl_flags |= PL_SHAREMOD; 609 p->p_limit = newlim; 610 mutex_exit(&p->p_mutex); 611 } 612 613 void 614 limfree(struct plimit *lim) 615 { 616 struct plimit *sv_lim; 617 618 do { 619 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) 620 return; 621 if (lim->pl_corename != defcorename) 622 free(lim->pl_corename, M_TEMP); 623 sv_lim = lim->pl_sv_limit; 624 mutex_destroy(&lim->pl_lock); 625 pool_put(&plimit_pool, lim); 626 } while ((lim = sv_lim) != NULL); 627 } 628 629 struct pstats * 630 pstatscopy(struct pstats *ps) 631 { 632 633 struct pstats *newps; 634 635 newps = pool_get(&pstats_pool, PR_WAITOK); 636 637 memset(&newps->pstat_startzero, 0, 638 (unsigned) ((char *)&newps->pstat_endzero - 639 (char *)&newps->pstat_startzero)); 640 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 641 ((char *)&newps->pstat_endcopy - 642 (char *)&newps->pstat_startcopy)); 643 644 return (newps); 645 646 } 647 648 void 649 pstatsfree(struct pstats *ps) 650 { 651 652 pool_put(&pstats_pool, ps); 653 } 654 655 /* 656 * sysctl interface in five parts 657 */ 658 659 /* 660 * a routine for sysctl proc subtree helpers that need to pick a valid 661 * process by pid. 662 */ 663 static int 664 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 665 { 666 struct proc *ptmp; 667 int error = 0; 668 669 if (pid == PROC_CURPROC) 670 ptmp = l->l_proc; 671 else if ((ptmp = pfind(pid)) == NULL) 672 error = ESRCH; 673 674 *p2 = ptmp; 675 return (error); 676 } 677 678 /* 679 * sysctl helper routine for setting a process's specific corefile 680 * name. picks the process based on the given pid and checks the 681 * correctness of the new value. 682 */ 683 static int 684 sysctl_proc_corename(SYSCTLFN_ARGS) 685 { 686 struct proc *ptmp; 687 struct plimit *lim; 688 int error = 0, len; 689 char *cname; 690 char *ocore; 691 char *tmp; 692 struct sysctlnode node; 693 694 /* 695 * is this all correct? 696 */ 697 if (namelen != 0) 698 return (EINVAL); 699 if (name[-1] != PROC_PID_CORENAME) 700 return (EINVAL); 701 702 /* 703 * whom are we tweaking? 704 */ 705 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 706 if (error) 707 return (error); 708 709 /* XXX this should be in p_find() */ 710 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 711 ptmp, NULL, NULL, NULL); 712 if (error) 713 return (error); 714 715 /* 716 * let them modify a temporary copy of the core name 717 */ 718 cname = PNBUF_GET(); 719 lim = ptmp->p_limit; 720 mutex_enter(&lim->pl_lock); 721 strlcpy(cname, lim->pl_corename, MAXPATHLEN); 722 mutex_exit(&lim->pl_lock); 723 724 node = *rnode; 725 node.sysctl_data = cname; 726 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 727 728 /* 729 * if that failed, or they have nothing new to say, or we've 730 * heard it before... 731 */ 732 if (error || newp == NULL) 733 goto done; 734 lim = ptmp->p_limit; 735 mutex_enter(&lim->pl_lock); 736 error = strcmp(cname, lim->pl_corename); 737 mutex_exit(&lim->pl_lock); 738 if (error == 0) 739 /* Unchanged */ 740 goto done; 741 742 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 743 ptmp, cname, NULL, NULL); 744 if (error) 745 return (error); 746 747 /* 748 * no error yet and cname now has the new core name in it. 749 * let's see if it looks acceptable. it must be either "core" 750 * or end in ".core" or "/core". 751 */ 752 len = strlen(cname); 753 if (len < 4) { 754 error = EINVAL; 755 } else if (strcmp(cname + len - 4, "core") != 0) { 756 error = EINVAL; 757 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 758 error = EINVAL; 759 } 760 if (error != 0) { 761 goto done; 762 } 763 764 /* 765 * hmm...looks good. now...where do we put it? 766 */ 767 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 768 if (tmp == NULL) { 769 error = ENOMEM; 770 goto done; 771 } 772 memcpy(tmp, cname, len + 1); 773 774 lim_privatise(ptmp, false); 775 lim = ptmp->p_limit; 776 mutex_enter(&lim->pl_lock); 777 ocore = lim->pl_corename; 778 lim->pl_corename = tmp; 779 mutex_exit(&lim->pl_lock); 780 if (ocore != defcorename) 781 free(ocore, M_TEMP); 782 783 done: 784 PNBUF_PUT(cname); 785 return error; 786 } 787 788 /* 789 * sysctl helper routine for checking/setting a process's stop flags, 790 * one for fork and one for exec. 791 */ 792 static int 793 sysctl_proc_stop(SYSCTLFN_ARGS) 794 { 795 struct proc *ptmp; 796 int i, f, error = 0; 797 struct sysctlnode node; 798 799 if (namelen != 0) 800 return (EINVAL); 801 802 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 803 if (error) 804 return (error); 805 806 /* XXX this should be in p_find() */ 807 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 808 ptmp, NULL, NULL, NULL); 809 if (error) 810 return (error); 811 812 switch (rnode->sysctl_num) { 813 case PROC_PID_STOPFORK: 814 f = PS_STOPFORK; 815 break; 816 case PROC_PID_STOPEXEC: 817 f = PS_STOPEXEC; 818 break; 819 case PROC_PID_STOPEXIT: 820 f = PS_STOPEXIT; 821 break; 822 default: 823 return (EINVAL); 824 } 825 826 i = (ptmp->p_flag & f) ? 1 : 0; 827 node = *rnode; 828 node.sysctl_data = &i; 829 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 830 if (error || newp == NULL) 831 return (error); 832 833 mutex_enter(&ptmp->p_smutex); 834 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 835 ptmp, KAUTH_ARG(f), NULL, NULL); 836 if (error) 837 return (error); 838 if (i) 839 ptmp->p_sflag |= f; 840 else 841 ptmp->p_sflag &= ~f; 842 mutex_exit(&ptmp->p_smutex); 843 844 return (0); 845 } 846 847 /* 848 * sysctl helper routine for a process's rlimits as exposed by sysctl. 849 */ 850 static int 851 sysctl_proc_plimit(SYSCTLFN_ARGS) 852 { 853 struct proc *ptmp; 854 u_int limitno; 855 int which, error = 0; 856 struct rlimit alim; 857 struct sysctlnode node; 858 859 if (namelen != 0) 860 return (EINVAL); 861 862 which = name[-1]; 863 if (which != PROC_PID_LIMIT_TYPE_SOFT && 864 which != PROC_PID_LIMIT_TYPE_HARD) 865 return (EINVAL); 866 867 limitno = name[-2] - 1; 868 if (limitno >= RLIM_NLIMITS) 869 return (EINVAL); 870 871 if (name[-3] != PROC_PID_LIMIT) 872 return (EINVAL); 873 874 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 875 if (error) 876 return (error); 877 878 /* XXX this should be in p_find() */ 879 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 880 ptmp, NULL, NULL, NULL); 881 if (error) 882 return (error); 883 884 node = *rnode; 885 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 886 if (which == PROC_PID_LIMIT_TYPE_HARD) 887 node.sysctl_data = &alim.rlim_max; 888 else 889 node.sysctl_data = &alim.rlim_cur; 890 891 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 892 if (error || newp == NULL) 893 return (error); 894 895 return (dosetrlimit(l, ptmp, limitno, &alim)); 896 } 897 898 /* 899 * and finally, the actually glue that sticks it to the tree 900 */ 901 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup") 902 { 903 904 sysctl_createv(clog, 0, NULL, NULL, 905 CTLFLAG_PERMANENT, 906 CTLTYPE_NODE, "proc", NULL, 907 NULL, 0, NULL, 0, 908 CTL_PROC, CTL_EOL); 909 sysctl_createv(clog, 0, NULL, NULL, 910 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 911 CTLTYPE_NODE, "curproc", 912 SYSCTL_DESCR("Per-process settings"), 913 NULL, 0, NULL, 0, 914 CTL_PROC, PROC_CURPROC, CTL_EOL); 915 916 sysctl_createv(clog, 0, NULL, NULL, 917 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 918 CTLTYPE_STRING, "corename", 919 SYSCTL_DESCR("Core file name"), 920 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 921 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 922 sysctl_createv(clog, 0, NULL, NULL, 923 CTLFLAG_PERMANENT, 924 CTLTYPE_NODE, "rlimit", 925 SYSCTL_DESCR("Process limits"), 926 NULL, 0, NULL, 0, 927 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 928 929 #define create_proc_plimit(s, n) do { \ 930 sysctl_createv(clog, 0, NULL, NULL, \ 931 CTLFLAG_PERMANENT, \ 932 CTLTYPE_NODE, s, \ 933 SYSCTL_DESCR("Process " s " limits"), \ 934 NULL, 0, NULL, 0, \ 935 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 936 CTL_EOL); \ 937 sysctl_createv(clog, 0, NULL, NULL, \ 938 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 939 CTLTYPE_QUAD, "soft", \ 940 SYSCTL_DESCR("Process soft " s " limit"), \ 941 sysctl_proc_plimit, 0, NULL, 0, \ 942 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 943 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 944 sysctl_createv(clog, 0, NULL, NULL, \ 945 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 946 CTLTYPE_QUAD, "hard", \ 947 SYSCTL_DESCR("Process hard " s " limit"), \ 948 sysctl_proc_plimit, 0, NULL, 0, \ 949 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 950 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 951 } while (0/*CONSTCOND*/) 952 953 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 954 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 955 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 956 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 957 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 958 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 959 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 960 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 961 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 962 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 963 964 #undef create_proc_plimit 965 966 sysctl_createv(clog, 0, NULL, NULL, 967 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 968 CTLTYPE_INT, "stopfork", 969 SYSCTL_DESCR("Stop process at fork(2)"), 970 sysctl_proc_stop, 0, NULL, 0, 971 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 972 sysctl_createv(clog, 0, NULL, NULL, 973 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 974 CTLTYPE_INT, "stopexec", 975 SYSCTL_DESCR("Stop process at execve(2)"), 976 sysctl_proc_stop, 0, NULL, 0, 977 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 978 sysctl_createv(clog, 0, NULL, NULL, 979 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 980 CTLTYPE_INT, "stopexit", 981 SYSCTL_DESCR("Stop process before completing exit"), 982 sysctl_proc_stop, 0, NULL, 0, 983 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 984 } 985 986 void 987 uid_init(void) 988 { 989 990 /* 991 * XXXSMP This could be at IPL_SOFTNET, but for now we want 992 * to to be deadlock free, so it must be at IPL_VM. 993 */ 994 mutex_init(&uihashtbl_lock, MUTEX_DEFAULT, IPL_VM); 995 996 /* 997 * Ensure that uid 0 is always in the user hash table, as 998 * sbreserve() expects it available from interrupt context. 999 */ 1000 (void)uid_find(0); 1001 } 1002 1003 struct uidinfo * 1004 uid_find(uid_t uid) 1005 { 1006 struct uidinfo *uip; 1007 struct uidinfo *newuip = NULL; 1008 struct uihashhead *uipp; 1009 1010 uipp = UIHASH(uid); 1011 1012 again: 1013 mutex_enter(&uihashtbl_lock); 1014 LIST_FOREACH(uip, uipp, ui_hash) 1015 if (uip->ui_uid == uid) { 1016 mutex_exit(&uihashtbl_lock); 1017 if (newuip) { 1018 mutex_destroy(&newuip->ui_lock); 1019 free(newuip, M_PROC); 1020 } 1021 return uip; 1022 } 1023 if (newuip == NULL) { 1024 mutex_exit(&uihashtbl_lock); 1025 /* Must not be called from interrupt context. */ 1026 newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO); 1027 /* XXX this could be IPL_SOFTNET */ 1028 mutex_init(&newuip->ui_lock, MUTEX_DEFAULT, IPL_VM); 1029 goto again; 1030 } 1031 uip = newuip; 1032 1033 LIST_INSERT_HEAD(uipp, uip, ui_hash); 1034 uip->ui_uid = uid; 1035 mutex_exit(&uihashtbl_lock); 1036 1037 return uip; 1038 } 1039 1040 /* 1041 * Change the count associated with number of processes 1042 * a given user is using. 1043 */ 1044 int 1045 chgproccnt(uid_t uid, int diff) 1046 { 1047 struct uidinfo *uip; 1048 1049 if (diff == 0) 1050 return 0; 1051 1052 uip = uid_find(uid); 1053 mutex_enter(&uip->ui_lock); 1054 uip->ui_proccnt += diff; 1055 KASSERT(uip->ui_proccnt >= 0); 1056 mutex_exit(&uip->ui_lock); 1057 return uip->ui_proccnt; 1058 } 1059 1060 int 1061 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax) 1062 { 1063 rlim_t nsb; 1064 1065 mutex_enter(&uip->ui_lock); 1066 nsb = uip->ui_sbsize + to - *hiwat; 1067 if (to > *hiwat && nsb > xmax) { 1068 mutex_exit(&uip->ui_lock); 1069 return 0; 1070 } 1071 *hiwat = to; 1072 uip->ui_sbsize = nsb; 1073 KASSERT(uip->ui_sbsize >= 0); 1074 mutex_exit(&uip->ui_lock); 1075 return 1; 1076 } 1077
Indexes created Sun Sep 21 20:09:37 GMT 2025