kern_resource.c revision 1.126.2.2
1 /* $NetBSD: kern_resource.c,v 1.126.2.2 2007/12/15 03:16:57 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 1.126.2.2 2007/12/15 03:16:57 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/kauth.h> 53 #include <sys/atomic.h> 54 #include <sys/mount.h> 55 #include <sys/syscallargs.h> 56 57 #include <uvm/uvm_extern.h> 58 59 /* 60 * Maximum process data and stack limits. 61 * They are variables so they are patchable. 62 */ 63 rlim_t maxdmap = MAXDSIZ; 64 rlim_t maxsmap = MAXSSIZ; 65 66 struct uihashhead *uihashtbl; 67 u_long uihash; /* size of hash table - 1 */ 68 kmutex_t uihashtbl_lock; 69 70 static pool_cache_t plimit_cache; 71 static pool_cache_t pstats_cache; 72 73 void 74 resource_init(void) 75 { 76 77 plimit_cache = pool_cache_init(sizeof(struct plimit), 0, 0, 0, 78 "plimitpl", NULL, IPL_NONE, NULL, NULL, NULL); 79 pstats_cache = pool_cache_init(sizeof(struct pstats), 0, 0, 0, 80 "pstatspl", NULL, IPL_NONE, NULL, NULL, NULL); 81 } 82 83 /* 84 * Resource controls and accounting. 85 */ 86 87 int 88 sys_getpriority(struct lwp *l, void *v, register_t *retval) 89 { 90 struct sys_getpriority_args /* { 91 syscallarg(int) which; 92 syscallarg(id_t) who; 93 } */ *uap = v; 94 struct proc *curp = l->l_proc, *p; 95 int low = NZERO + PRIO_MAX + 1; 96 int who = SCARG(uap, who); 97 98 mutex_enter(&proclist_lock); 99 switch (SCARG(uap, which)) { 100 case PRIO_PROCESS: 101 if (who == 0) 102 p = curp; 103 else 104 p = p_find(who, PFIND_LOCKED); 105 if (p != NULL) 106 low = p->p_nice; 107 break; 108 109 case PRIO_PGRP: { 110 struct pgrp *pg; 111 112 if (who == 0) 113 pg = curp->p_pgrp; 114 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 115 break; 116 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 117 if (p->p_nice < low) 118 low = p->p_nice; 119 } 120 break; 121 } 122 123 case PRIO_USER: 124 if (who == 0) 125 who = (int)kauth_cred_geteuid(l->l_cred); 126 PROCLIST_FOREACH(p, &allproc) { 127 mutex_enter(&p->p_mutex); 128 if (kauth_cred_geteuid(p->p_cred) == 129 (uid_t)who && p->p_nice < low) 130 low = p->p_nice; 131 mutex_exit(&p->p_mutex); 132 } 133 break; 134 135 default: 136 mutex_exit(&proclist_lock); 137 return (EINVAL); 138 } 139 mutex_exit(&proclist_lock); 140 141 if (low == NZERO + PRIO_MAX + 1) 142 return (ESRCH); 143 *retval = low - NZERO; 144 return (0); 145 } 146 147 /* ARGSUSED */ 148 int 149 sys_setpriority(struct lwp *l, void *v, register_t *retval) 150 { 151 struct sys_setpriority_args /* { 152 syscallarg(int) which; 153 syscallarg(id_t) who; 154 syscallarg(int) prio; 155 } */ *uap = v; 156 struct proc *curp = l->l_proc, *p; 157 int found = 0, error = 0; 158 int who = SCARG(uap, who); 159 160 mutex_enter(&proclist_lock); 161 switch (SCARG(uap, which)) { 162 case PRIO_PROCESS: 163 if (who == 0) 164 p = curp; 165 else 166 p = p_find(who, PFIND_LOCKED); 167 if (p != 0) { 168 mutex_enter(&p->p_mutex); 169 error = donice(l, p, SCARG(uap, prio)); 170 mutex_exit(&p->p_mutex); 171 } 172 found++; 173 break; 174 175 case PRIO_PGRP: { 176 struct pgrp *pg; 177 178 if (who == 0) 179 pg = curp->p_pgrp; 180 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 181 break; 182 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 183 mutex_enter(&p->p_mutex); 184 error = donice(l, p, SCARG(uap, prio)); 185 mutex_exit(&p->p_mutex); 186 found++; 187 } 188 break; 189 } 190 191 case PRIO_USER: 192 if (who == 0) 193 who = (int)kauth_cred_geteuid(l->l_cred); 194 PROCLIST_FOREACH(p, &allproc) { 195 mutex_enter(&p->p_mutex); 196 if (kauth_cred_geteuid(p->p_cred) == 197 (uid_t)SCARG(uap, who)) { 198 error = donice(l, p, SCARG(uap, prio)); 199 found++; 200 } 201 mutex_exit(&p->p_mutex); 202 } 203 break; 204 205 default: 206 error = EINVAL; 207 break; 208 } 209 mutex_exit(&proclist_lock); 210 if (found == 0) 211 return (ESRCH); 212 return (error); 213 } 214 215 /* 216 * Renice a process. 217 * 218 * Call with the target process' credentials locked. 219 */ 220 int 221 donice(struct lwp *l, struct proc *chgp, int n) 222 { 223 kauth_cred_t cred = l->l_cred; 224 int onice; 225 226 KASSERT(mutex_owned(&chgp->p_mutex)); 227 228 if (n > PRIO_MAX) 229 n = PRIO_MAX; 230 if (n < PRIO_MIN) 231 n = PRIO_MIN; 232 n += NZERO; 233 onice = chgp->p_nice; 234 onice = chgp->p_nice; 235 236 again: 237 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 238 KAUTH_ARG(n), NULL, NULL)) 239 return (EACCES); 240 mutex_spin_enter(&chgp->p_smutex); 241 if (onice != chgp->p_nice) { 242 mutex_spin_exit(&chgp->p_smutex); 243 goto again; 244 } 245 sched_nice(chgp, n); 246 mutex_spin_exit(&chgp->p_smutex); 247 return (0); 248 } 249 250 /* ARGSUSED */ 251 int 252 sys_setrlimit(struct lwp *l, void *v, register_t *retval) 253 { 254 struct sys_setrlimit_args /* { 255 syscallarg(int) which; 256 syscallarg(const struct rlimit *) rlp; 257 } */ *uap = v; 258 int which = SCARG(uap, which); 259 struct rlimit alim; 260 int error; 261 262 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 263 if (error) 264 return (error); 265 return (dosetrlimit(l, l->l_proc, which, &alim)); 266 } 267 268 int 269 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 270 { 271 struct rlimit *alimp; 272 int error; 273 274 if ((u_int)which >= RLIM_NLIMITS) 275 return (EINVAL); 276 277 if (limp->rlim_cur < 0 || limp->rlim_max < 0) 278 return (EINVAL); 279 280 if (limp->rlim_cur > limp->rlim_max) { 281 /* 282 * This is programming error. According to SUSv2, we should 283 * return error in this case. 284 */ 285 return (EINVAL); 286 } 287 288 alimp = &p->p_rlimit[which]; 289 /* if we don't change the value, no need to limcopy() */ 290 if (limp->rlim_cur == alimp->rlim_cur && 291 limp->rlim_max == alimp->rlim_max) 292 return 0; 293 294 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 295 p, limp, KAUTH_ARG(which), NULL); 296 if (error) 297 return (error); 298 299 lim_privatise(p, false); 300 /* p->p_limit is now unchangeable */ 301 alimp = &p->p_rlimit[which]; 302 303 switch (which) { 304 305 case RLIMIT_DATA: 306 if (limp->rlim_cur > maxdmap) 307 limp->rlim_cur = maxdmap; 308 if (limp->rlim_max > maxdmap) 309 limp->rlim_max = maxdmap; 310 break; 311 312 case RLIMIT_STACK: 313 if (limp->rlim_cur > maxsmap) 314 limp->rlim_cur = maxsmap; 315 if (limp->rlim_max > maxsmap) 316 limp->rlim_max = maxsmap; 317 318 /* 319 * Return EINVAL if the new stack size limit is lower than 320 * current usage. Otherwise, the process would get SIGSEGV the 321 * moment it would try to access anything on it's current stack. 322 * This conforms to SUSv2. 323 */ 324 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 325 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 326 return (EINVAL); 327 } 328 329 /* 330 * Stack is allocated to the max at exec time with 331 * only "rlim_cur" bytes accessible (In other words, 332 * allocates stack dividing two contiguous regions at 333 * "rlim_cur" bytes boundary). 334 * 335 * Since allocation is done in terms of page, roundup 336 * "rlim_cur" (otherwise, contiguous regions 337 * overlap). If stack limit is going up make more 338 * accessible, if going down make inaccessible. 339 */ 340 limp->rlim_cur = round_page(limp->rlim_cur); 341 if (limp->rlim_cur != alimp->rlim_cur) { 342 vaddr_t addr; 343 vsize_t size; 344 vm_prot_t prot; 345 346 if (limp->rlim_cur > alimp->rlim_cur) { 347 prot = VM_PROT_READ | VM_PROT_WRITE; 348 size = limp->rlim_cur - alimp->rlim_cur; 349 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 350 limp->rlim_cur; 351 } else { 352 prot = VM_PROT_NONE; 353 size = alimp->rlim_cur - limp->rlim_cur; 354 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 355 alimp->rlim_cur; 356 } 357 (void) uvm_map_protect(&p->p_vmspace->vm_map, 358 addr, addr+size, prot, false); 359 } 360 break; 361 362 case RLIMIT_NOFILE: 363 if (limp->rlim_cur > maxfiles) 364 limp->rlim_cur = maxfiles; 365 if (limp->rlim_max > maxfiles) 366 limp->rlim_max = maxfiles; 367 break; 368 369 case RLIMIT_NPROC: 370 if (limp->rlim_cur > maxproc) 371 limp->rlim_cur = maxproc; 372 if (limp->rlim_max > maxproc) 373 limp->rlim_max = maxproc; 374 break; 375 } 376 377 mutex_enter(&p->p_limit->pl_lock); 378 *alimp = *limp; 379 mutex_exit(&p->p_limit->pl_lock); 380 return (0); 381 } 382 383 /* ARGSUSED */ 384 int 385 sys_getrlimit(struct lwp *l, void *v, register_t *retval) 386 { 387 struct sys_getrlimit_args /* { 388 syscallarg(int) which; 389 syscallarg(struct rlimit *) rlp; 390 } */ *uap = v; 391 struct proc *p = l->l_proc; 392 int which = SCARG(uap, which); 393 struct rlimit rl; 394 395 if ((u_int)which >= RLIM_NLIMITS) 396 return (EINVAL); 397 398 mutex_enter(&p->p_mutex); 399 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 400 mutex_exit(&p->p_mutex); 401 402 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 403 } 404 405 /* 406 * Transform the running time and tick information in proc p into user, 407 * system, and interrupt time usage. 408 * 409 * Should be called with p->p_smutex held unless called from exit1(). 410 */ 411 void 412 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 413 struct timeval *ip, struct timeval *rp) 414 { 415 u_quad_t u, st, ut, it, tot; 416 unsigned long sec; 417 long usec; 418 struct timeval tv; 419 struct lwp *l; 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 sec = p->p_rtime.tv_sec; 428 usec = p->p_rtime.tv_usec; 429 430 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 431 lwp_lock(l); 432 sec += l->l_rtime.tv_sec; 433 if ((usec += l->l_rtime.tv_usec) >= 1000000) { 434 sec++; 435 usec -= 1000000; 436 } 437 if ((l->l_flag & LW_RUNNING) != 0) { 438 /* 439 * Adjust for the current time slice. This is 440 * actually fairly important since the error 441 * here is on the order of a time quantum, 442 * which is much greater than the sampling 443 * error. 444 */ 445 microtime(&tv); 446 sec += tv.tv_sec - l->l_stime.tv_sec; 447 usec += tv.tv_usec - l->l_stime.tv_usec; 448 if (usec >= 1000000) { 449 sec++; 450 usec -= 1000000; 451 } 452 } 453 lwp_unlock(l); 454 } 455 456 tot = st + ut + it; 457 u = sec * 1000000ull + 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_sec = sec; 482 rp->tv_usec = usec; 483 } 484 } 485 486 /* ARGSUSED */ 487 int 488 sys_getrusage(struct lwp *l, void *v, register_t *retval) 489 { 490 struct sys_getrusage_args /* { 491 syscallarg(int) who; 492 syscallarg(struct rusage *) rusage; 493 } */ *uap = v; 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_smutex); 500 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 501 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 502 mutex_exit(&p->p_smutex); 503 break; 504 505 case RUSAGE_CHILDREN: 506 mutex_enter(&p->p_smutex); 507 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 508 mutex_exit(&p->p_smutex); 509 break; 510 511 default: 512 return EINVAL; 513 } 514 515 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 516 } 517 518 void 519 ruadd(struct rusage *ru, struct rusage *ru2) 520 { 521 long *ip, *ip2; 522 int i; 523 524 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 525 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 526 if (ru->ru_maxrss < ru2->ru_maxrss) 527 ru->ru_maxrss = ru2->ru_maxrss; 528 ip = &ru->ru_first; ip2 = &ru2->ru_first; 529 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 530 *ip++ += *ip2++; 531 } 532 533 /* 534 * Make a copy of the plimit structure. 535 * We share these structures copy-on-write after fork, 536 * and copy when a limit is changed. 537 * 538 * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure 539 * we are copying to change beneath our feet! 540 */ 541 struct plimit * 542 lim_copy(struct plimit *lim) 543 { 544 struct plimit *newlim; 545 char *corename; 546 size_t alen, len; 547 548 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 549 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 550 newlim->pl_flags = 0; 551 newlim->pl_refcnt = 1; 552 newlim->pl_sv_limit = NULL; 553 554 mutex_enter(&lim->pl_lock); 555 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 556 sizeof(struct rlimit) * RLIM_NLIMITS); 557 558 alen = 0; 559 corename = NULL; 560 for (;;) { 561 if (lim->pl_corename == defcorename) { 562 newlim->pl_corename = defcorename; 563 break; 564 } 565 len = strlen(lim->pl_corename) + 1; 566 if (len <= alen) { 567 newlim->pl_corename = corename; 568 memcpy(corename, lim->pl_corename, len); 569 corename = NULL; 570 break; 571 } 572 mutex_exit(&lim->pl_lock); 573 if (corename != NULL) 574 free(corename, M_TEMP); 575 alen = len; 576 corename = malloc(alen, M_TEMP, M_WAITOK); 577 mutex_enter(&lim->pl_lock); 578 } 579 mutex_exit(&lim->pl_lock); 580 if (corename != NULL) 581 free(corename, M_TEMP); 582 return newlim; 583 } 584 585 void 586 lim_addref(struct plimit *lim) 587 { 588 atomic_inc_uint(&lim->pl_refcnt); 589 } 590 591 /* 592 * Give a process it's own private plimit structure. 593 * This will only be shared (in fork) if modifications are to be shared. 594 */ 595 void 596 lim_privatise(struct proc *p, bool set_shared) 597 { 598 struct plimit *lim, *newlim; 599 600 lim = p->p_limit; 601 if (lim->pl_flags & PL_WRITEABLE) { 602 if (set_shared) 603 lim->pl_flags |= PL_SHAREMOD; 604 return; 605 } 606 607 if (set_shared && lim->pl_flags & PL_SHAREMOD) 608 return; 609 610 newlim = lim_copy(lim); 611 612 mutex_enter(&p->p_mutex); 613 if (p->p_limit->pl_flags & PL_WRITEABLE) { 614 /* Someone crept in while we were busy */ 615 mutex_exit(&p->p_mutex); 616 limfree(newlim); 617 if (set_shared) 618 p->p_limit->pl_flags |= PL_SHAREMOD; 619 return; 620 } 621 622 /* 623 * Since most accesses to p->p_limit aren't locked, we must not 624 * delete the old limit structure yet. 625 */ 626 newlim->pl_sv_limit = p->p_limit; 627 newlim->pl_flags |= PL_WRITEABLE; 628 if (set_shared) 629 newlim->pl_flags |= PL_SHAREMOD; 630 p->p_limit = newlim; 631 mutex_exit(&p->p_mutex); 632 } 633 634 void 635 limfree(struct plimit *lim) 636 { 637 struct plimit *sv_lim; 638 639 do { 640 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) 641 return; 642 if (lim->pl_corename != defcorename) 643 free(lim->pl_corename, M_TEMP); 644 sv_lim = lim->pl_sv_limit; 645 mutex_destroy(&lim->pl_lock); 646 pool_cache_put(plimit_cache, lim); 647 } while ((lim = sv_lim) != NULL); 648 } 649 650 struct pstats * 651 pstatscopy(struct pstats *ps) 652 { 653 654 struct pstats *newps; 655 656 newps = pool_cache_get(pstats_cache, PR_WAITOK); 657 658 memset(&newps->pstat_startzero, 0, 659 (unsigned) ((char *)&newps->pstat_endzero - 660 (char *)&newps->pstat_startzero)); 661 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 662 ((char *)&newps->pstat_endcopy - 663 (char *)&newps->pstat_startcopy)); 664 665 return (newps); 666 667 } 668 669 void 670 pstatsfree(struct pstats *ps) 671 { 672 673 pool_cache_put(pstats_cache, ps); 674 } 675 676 /* 677 * sysctl interface in five parts 678 */ 679 680 /* 681 * a routine for sysctl proc subtree helpers that need to pick a valid 682 * process by pid. 683 */ 684 static int 685 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 686 { 687 struct proc *ptmp; 688 int error = 0; 689 690 if (pid == PROC_CURPROC) 691 ptmp = l->l_proc; 692 else if ((ptmp = pfind(pid)) == NULL) 693 error = ESRCH; 694 695 *p2 = ptmp; 696 return (error); 697 } 698 699 /* 700 * sysctl helper routine for setting a process's specific corefile 701 * name. picks the process based on the given pid and checks the 702 * correctness of the new value. 703 */ 704 static int 705 sysctl_proc_corename(SYSCTLFN_ARGS) 706 { 707 struct proc *ptmp; 708 struct plimit *lim; 709 int error = 0, len; 710 char *cname; 711 char *ocore; 712 char *tmp; 713 struct sysctlnode node; 714 715 /* 716 * is this all correct? 717 */ 718 if (namelen != 0) 719 return (EINVAL); 720 if (name[-1] != PROC_PID_CORENAME) 721 return (EINVAL); 722 723 /* 724 * whom are we tweaking? 725 */ 726 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 727 if (error) 728 return (error); 729 730 /* XXX this should be in p_find() */ 731 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 732 ptmp, NULL, NULL, NULL); 733 if (error) 734 return (error); 735 736 /* 737 * let them modify a temporary copy of the core name 738 */ 739 cname = PNBUF_GET(); 740 lim = ptmp->p_limit; 741 mutex_enter(&lim->pl_lock); 742 strlcpy(cname, lim->pl_corename, MAXPATHLEN); 743 mutex_exit(&lim->pl_lock); 744 745 node = *rnode; 746 node.sysctl_data = cname; 747 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 748 749 /* 750 * if that failed, or they have nothing new to say, or we've 751 * heard it before... 752 */ 753 if (error || newp == NULL) 754 goto done; 755 lim = ptmp->p_limit; 756 mutex_enter(&lim->pl_lock); 757 error = strcmp(cname, lim->pl_corename); 758 mutex_exit(&lim->pl_lock); 759 if (error == 0) 760 /* Unchanged */ 761 goto done; 762 763 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 764 ptmp, cname, NULL, NULL); 765 if (error) 766 return (error); 767 768 /* 769 * no error yet and cname now has the new core name in it. 770 * let's see if it looks acceptable. it must be either "core" 771 * or end in ".core" or "/core". 772 */ 773 len = strlen(cname); 774 if (len < 4) { 775 error = EINVAL; 776 } else if (strcmp(cname + len - 4, "core") != 0) { 777 error = EINVAL; 778 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 779 error = EINVAL; 780 } 781 if (error != 0) { 782 goto done; 783 } 784 785 /* 786 * hmm...looks good. now...where do we put it? 787 */ 788 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 789 if (tmp == NULL) { 790 error = ENOMEM; 791 goto done; 792 } 793 memcpy(tmp, cname, len + 1); 794 795 lim_privatise(ptmp, false); 796 lim = ptmp->p_limit; 797 mutex_enter(&lim->pl_lock); 798 ocore = lim->pl_corename; 799 lim->pl_corename = tmp; 800 mutex_exit(&lim->pl_lock); 801 if (ocore != defcorename) 802 free(ocore, M_TEMP); 803 804 done: 805 PNBUF_PUT(cname); 806 return error; 807 } 808 809 /* 810 * sysctl helper routine for checking/setting a process's stop flags, 811 * one for fork and one for exec. 812 */ 813 static int 814 sysctl_proc_stop(SYSCTLFN_ARGS) 815 { 816 struct proc *ptmp; 817 int i, f, error = 0; 818 struct sysctlnode node; 819 820 if (namelen != 0) 821 return (EINVAL); 822 823 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 824 if (error) 825 return (error); 826 827 /* XXX this should be in p_find() */ 828 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 829 ptmp, NULL, NULL, NULL); 830 if (error) 831 return (error); 832 833 switch (rnode->sysctl_num) { 834 case PROC_PID_STOPFORK: 835 f = PS_STOPFORK; 836 break; 837 case PROC_PID_STOPEXEC: 838 f = PS_STOPEXEC; 839 break; 840 case PROC_PID_STOPEXIT: 841 f = PS_STOPEXIT; 842 break; 843 default: 844 return (EINVAL); 845 } 846 847 i = (ptmp->p_flag & f) ? 1 : 0; 848 node = *rnode; 849 node.sysctl_data = &i; 850 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 851 if (error || newp == NULL) 852 return (error); 853 854 mutex_enter(&ptmp->p_smutex); 855 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 856 ptmp, KAUTH_ARG(f), NULL, NULL); 857 if (error) 858 return (error); 859 if (i) 860 ptmp->p_sflag |= f; 861 else 862 ptmp->p_sflag &= ~f; 863 mutex_exit(&ptmp->p_smutex); 864 865 return (0); 866 } 867 868 /* 869 * sysctl helper routine for a process's rlimits as exposed by sysctl. 870 */ 871 static int 872 sysctl_proc_plimit(SYSCTLFN_ARGS) 873 { 874 struct proc *ptmp; 875 u_int limitno; 876 int which, error = 0; 877 struct rlimit alim; 878 struct sysctlnode node; 879 880 if (namelen != 0) 881 return (EINVAL); 882 883 which = name[-1]; 884 if (which != PROC_PID_LIMIT_TYPE_SOFT && 885 which != PROC_PID_LIMIT_TYPE_HARD) 886 return (EINVAL); 887 888 limitno = name[-2] - 1; 889 if (limitno >= RLIM_NLIMITS) 890 return (EINVAL); 891 892 if (name[-3] != PROC_PID_LIMIT) 893 return (EINVAL); 894 895 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 896 if (error) 897 return (error); 898 899 /* XXX this should be in p_find() */ 900 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 901 ptmp, NULL, NULL, NULL); 902 if (error) 903 return (error); 904 905 node = *rnode; 906 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 907 if (which == PROC_PID_LIMIT_TYPE_HARD) 908 node.sysctl_data = &alim.rlim_max; 909 else 910 node.sysctl_data = &alim.rlim_cur; 911 912 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 913 if (error || newp == NULL) 914 return (error); 915 916 return (dosetrlimit(l, ptmp, limitno, &alim)); 917 } 918 919 /* 920 * and finally, the actually glue that sticks it to the tree 921 */ 922 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup") 923 { 924 925 sysctl_createv(clog, 0, NULL, NULL, 926 CTLFLAG_PERMANENT, 927 CTLTYPE_NODE, "proc", NULL, 928 NULL, 0, NULL, 0, 929 CTL_PROC, CTL_EOL); 930 sysctl_createv(clog, 0, NULL, NULL, 931 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 932 CTLTYPE_NODE, "curproc", 933 SYSCTL_DESCR("Per-process settings"), 934 NULL, 0, NULL, 0, 935 CTL_PROC, PROC_CURPROC, CTL_EOL); 936 937 sysctl_createv(clog, 0, NULL, NULL, 938 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 939 CTLTYPE_STRING, "corename", 940 SYSCTL_DESCR("Core file name"), 941 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 942 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 943 sysctl_createv(clog, 0, NULL, NULL, 944 CTLFLAG_PERMANENT, 945 CTLTYPE_NODE, "rlimit", 946 SYSCTL_DESCR("Process limits"), 947 NULL, 0, NULL, 0, 948 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 949 950 #define create_proc_plimit(s, n) do { \ 951 sysctl_createv(clog, 0, NULL, NULL, \ 952 CTLFLAG_PERMANENT, \ 953 CTLTYPE_NODE, s, \ 954 SYSCTL_DESCR("Process " s " limits"), \ 955 NULL, 0, NULL, 0, \ 956 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 957 CTL_EOL); \ 958 sysctl_createv(clog, 0, NULL, NULL, \ 959 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 960 CTLTYPE_QUAD, "soft", \ 961 SYSCTL_DESCR("Process soft " s " limit"), \ 962 sysctl_proc_plimit, 0, NULL, 0, \ 963 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 964 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 965 sysctl_createv(clog, 0, NULL, NULL, \ 966 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 967 CTLTYPE_QUAD, "hard", \ 968 SYSCTL_DESCR("Process hard " s " limit"), \ 969 sysctl_proc_plimit, 0, NULL, 0, \ 970 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 971 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 972 } while (0/*CONSTCOND*/) 973 974 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 975 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 976 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 977 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 978 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 979 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 980 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 981 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 982 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 983 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 984 985 #undef create_proc_plimit 986 987 sysctl_createv(clog, 0, NULL, NULL, 988 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 989 CTLTYPE_INT, "stopfork", 990 SYSCTL_DESCR("Stop process at fork(2)"), 991 sysctl_proc_stop, 0, NULL, 0, 992 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 993 sysctl_createv(clog, 0, NULL, NULL, 994 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 995 CTLTYPE_INT, "stopexec", 996 SYSCTL_DESCR("Stop process at execve(2)"), 997 sysctl_proc_stop, 0, NULL, 0, 998 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 999 sysctl_createv(clog, 0, NULL, NULL, 1000 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1001 CTLTYPE_INT, "stopexit", 1002 SYSCTL_DESCR("Stop process before completing exit"), 1003 sysctl_proc_stop, 0, NULL, 0, 1004 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1005 } 1006 1007 void 1008 uid_init(void) 1009 { 1010 1011 /* 1012 * XXXSMP This could be at IPL_SOFTNET, but for now we want 1013 * to to be deadlock free, so it must be at IPL_VM. 1014 */ 1015 mutex_init(&uihashtbl_lock, MUTEX_DEFAULT, IPL_VM); 1016 1017 /* 1018 * Ensure that uid 0 is always in the user hash table, as 1019 * sbreserve() expects it available from interrupt context. 1020 */ 1021 (void)uid_find(0); 1022 } 1023 1024 struct uidinfo * 1025 uid_find(uid_t uid) 1026 { 1027 struct uidinfo *uip; 1028 struct uidinfo *newuip = NULL; 1029 struct uihashhead *uipp; 1030 1031 uipp = UIHASH(uid); 1032 1033 again: 1034 mutex_enter(&uihashtbl_lock); 1035 LIST_FOREACH(uip, uipp, ui_hash) 1036 if (uip->ui_uid == uid) { 1037 mutex_exit(&uihashtbl_lock); 1038 if (newuip) { 1039 mutex_destroy(&newuip->ui_lock); 1040 free(newuip, M_PROC); 1041 } 1042 return uip; 1043 } 1044 if (newuip == NULL) { 1045 mutex_exit(&uihashtbl_lock); 1046 /* Must not be called from interrupt context. */ 1047 newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO); 1048 /* XXX this could be IPL_SOFTNET */ 1049 mutex_init(&newuip->ui_lock, MUTEX_DEFAULT, IPL_VM); 1050 goto again; 1051 } 1052 uip = newuip; 1053 1054 LIST_INSERT_HEAD(uipp, uip, ui_hash); 1055 uip->ui_uid = uid; 1056 mutex_exit(&uihashtbl_lock); 1057 1058 return uip; 1059 } 1060 1061 /* 1062 * Change the count associated with number of processes 1063 * a given user is using. 1064 */ 1065 int 1066 chgproccnt(uid_t uid, int diff) 1067 { 1068 struct uidinfo *uip; 1069 1070 if (diff == 0) 1071 return 0; 1072 1073 uip = uid_find(uid); 1074 mutex_enter(&uip->ui_lock); 1075 uip->ui_proccnt += diff; 1076 KASSERT(uip->ui_proccnt >= 0); 1077 mutex_exit(&uip->ui_lock); 1078 return uip->ui_proccnt; 1079 } 1080 1081 int 1082 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax) 1083 { 1084 rlim_t nsb; 1085 1086 mutex_enter(&uip->ui_lock); 1087 nsb = uip->ui_sbsize + to - *hiwat; 1088 if (to > *hiwat && nsb > xmax) { 1089 mutex_exit(&uip->ui_lock); 1090 return 0; 1091 } 1092 *hiwat = to; 1093 uip->ui_sbsize = nsb; 1094 KASSERT(uip->ui_sbsize >= 0); 1095 mutex_exit(&uip->ui_lock); 1096 return 1; 1097 } 1098
Indexes created Sun Sep 21 20:09:37 GMT 2025