kern_resource.c revision 1.103.4.10
1 /* $NetBSD: kern_resource.c,v 1.103.4.10 2007/02/01 08:48:38 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 1.103.4.10 2007/02/01 08:48:38 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 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 struct simplelock uihashtbl_slock = SIMPLELOCK_INITIALIZER; 69 70 71 /* 72 * Resource controls and accounting. 73 */ 74 75 int 76 sys_getpriority(struct lwp *l, void *v, register_t *retval) 77 { 78 struct sys_getpriority_args /* { 79 syscallarg(int) which; 80 syscallarg(id_t) who; 81 } */ *uap = v; 82 struct proc *curp = l->l_proc, *p; 83 int low = NZERO + PRIO_MAX + 1; 84 int who = SCARG(uap, who); 85 86 switch (SCARG(uap, which)) { 87 88 case PRIO_PROCESS: 89 if (who == 0) 90 p = curp; 91 else 92 p = p_find(who, 0); 93 if (p != NULL) 94 low = p->p_nice; 95 if (who != 0) 96 rw_exit(&proclist_lock); 97 break; 98 99 case PRIO_PGRP: { 100 struct pgrp *pg; 101 102 rw_enter(&proclist_lock, RW_READER); 103 if (who == 0) 104 pg = curp->p_pgrp; 105 else if ((pg = pg_find(who, PFIND_LOCKED | PFIND_UNLOCK_FAIL)) 106 == NULL) 107 break; 108 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 109 if (p->p_nice < low) 110 low = p->p_nice; 111 } 112 rw_exit(&proclist_lock); 113 break; 114 } 115 116 case PRIO_USER: 117 if (who == 0) 118 who = (int)kauth_cred_geteuid(l->l_cred); 119 rw_enter(&proclist_lock, RW_READER); 120 PROCLIST_FOREACH(p, &allproc) { 121 mutex_enter(&p->p_mutex); 122 if (kauth_cred_geteuid(p->p_cred) == 123 (uid_t)who && p->p_nice < low) 124 low = p->p_nice; 125 mutex_exit(&p->p_mutex); 126 } 127 rw_exit(&proclist_lock); 128 break; 129 130 default: 131 return (EINVAL); 132 } 133 if (low == NZERO + PRIO_MAX + 1) 134 return (ESRCH); 135 *retval = low - NZERO; 136 return (0); 137 } 138 139 /* ARGSUSED */ 140 int 141 sys_setpriority(struct lwp *l, void *v, register_t *retval) 142 { 143 struct sys_setpriority_args /* { 144 syscallarg(int) which; 145 syscallarg(id_t) who; 146 syscallarg(int) prio; 147 } */ *uap = v; 148 struct proc *curp = l->l_proc, *p; 149 int found = 0, error = 0; 150 int who = SCARG(uap, who); 151 152 switch (SCARG(uap, which)) { 153 154 case PRIO_PROCESS: 155 if (who == 0) 156 p = curp; 157 else 158 p = p_find(who, 0); 159 if (p != 0) { 160 mutex_enter(&p->p_mutex); 161 error = donice(l, p, SCARG(uap, prio)); 162 mutex_exit(&p->p_mutex); 163 } 164 if (who != 0) 165 rw_exit(&proclist_lock); 166 found++; 167 break; 168 169 case PRIO_PGRP: { 170 struct pgrp *pg; 171 172 rw_enter(&proclist_lock, RW_READER); 173 if (who == 0) 174 pg = curp->p_pgrp; 175 else if ((pg = pg_find(who, PFIND_LOCKED | PFIND_UNLOCK_FAIL)) == NULL) 176 break; 177 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 178 mutex_enter(&p->p_mutex); 179 error = donice(l, p, SCARG(uap, prio)); 180 mutex_exit(&p->p_mutex); 181 found++; 182 } 183 rw_exit(&proclist_lock); 184 break; 185 } 186 187 case PRIO_USER: 188 if (who == 0) 189 who = (int)kauth_cred_geteuid(l->l_cred); 190 rw_enter(&proclist_lock, RW_READER); 191 PROCLIST_FOREACH(p, &allproc) { 192 mutex_enter(&p->p_mutex); 193 if (kauth_cred_geteuid(p->p_cred) == 194 (uid_t)SCARG(uap, who)) { 195 error = donice(l, p, SCARG(uap, prio)); 196 found++; 197 } 198 mutex_exit(&p->p_mutex); 199 } 200 rw_exit(&proclist_lock); 201 break; 202 203 default: 204 return (EINVAL); 205 } 206 if (found == 0) 207 return (ESRCH); 208 return (error); 209 } 210 211 /* 212 * Renice a process. 213 * 214 * Call with the target process' credentials locked. 215 */ 216 int 217 donice(struct lwp *l, struct proc *chgp, int n) 218 { 219 kauth_cred_t cred = l->l_cred; 220 int onice; 221 222 LOCK_ASSERT(mutex_owned(&chgp->p_mutex)); 223 224 if (n > PRIO_MAX) 225 n = PRIO_MAX; 226 if (n < PRIO_MIN) 227 n = PRIO_MIN; 228 n += NZERO; 229 onice = chgp->p_nice; 230 onice = chgp->p_nice; 231 232 again: 233 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 234 KAUTH_ARG(n), NULL, NULL)) 235 return (EACCES); 236 mutex_enter(&chgp->p_stmutex); 237 if (onice != chgp->p_nice) { 238 mutex_exit(&chgp->p_stmutex); 239 goto again; 240 } 241 chgp->p_nice = n; 242 (void)resetprocpriority(chgp); 243 mutex_exit(&chgp->p_stmutex); 244 return (0); 245 } 246 247 /* ARGSUSED */ 248 int 249 sys_setrlimit(struct lwp *l, void *v, register_t *retval) 250 { 251 struct sys_setrlimit_args /* { 252 syscallarg(int) which; 253 syscallarg(const struct rlimit *) rlp; 254 } */ *uap = v; 255 int which = SCARG(uap, which); 256 struct rlimit alim; 257 int error; 258 259 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 260 if (error) 261 return (error); 262 return (dosetrlimit(l, l->l_proc, which, &alim)); 263 } 264 265 int 266 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 267 { 268 struct rlimit *alimp; 269 struct plimit *oldplim; 270 int error; 271 272 if ((u_int)which >= RLIM_NLIMITS) 273 return (EINVAL); 274 275 if (limp->rlim_cur < 0 || limp->rlim_max < 0) 276 return (EINVAL); 277 278 alimp = &p->p_rlimit[which]; 279 /* if we don't change the value, no need to limcopy() */ 280 if (limp->rlim_cur == alimp->rlim_cur && 281 limp->rlim_max == alimp->rlim_max) 282 return 0; 283 284 if (limp->rlim_cur > limp->rlim_max) { 285 /* 286 * This is programming error. According to SUSv2, we should 287 * return error in this case. 288 */ 289 return (EINVAL); 290 } 291 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 292 p, limp, KAUTH_ARG(which), NULL); 293 if (error) 294 return (error); 295 296 if (p->p_limit->p_refcnt > 1 && 297 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 298 p->p_limit = limcopy(oldplim = p->p_limit); 299 limfree(oldplim); 300 alimp = &p->p_rlimit[which]; 301 } 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 * Stack is allocated to the max at exec time with 330 * only "rlim_cur" bytes accessible (In other words, 331 * allocates stack dividing two contiguous regions at 332 * "rlim_cur" bytes boundary). 333 * 334 * Since allocation is done in terms of page, roundup 335 * "rlim_cur" (otherwise, contiguous regions 336 * overlap). If stack limit is going up make more 337 * accessible, if going down make inaccessible. 338 */ 339 limp->rlim_cur = round_page(limp->rlim_cur); 340 if (limp->rlim_cur != alimp->rlim_cur) { 341 vaddr_t addr; 342 vsize_t size; 343 vm_prot_t prot; 344 345 if (limp->rlim_cur > alimp->rlim_cur) { 346 prot = VM_PROT_READ | VM_PROT_WRITE; 347 size = limp->rlim_cur - alimp->rlim_cur; 348 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 349 limp->rlim_cur; 350 } else { 351 prot = VM_PROT_NONE; 352 size = alimp->rlim_cur - limp->rlim_cur; 353 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 354 alimp->rlim_cur; 355 } 356 (void) uvm_map_protect(&p->p_vmspace->vm_map, 357 addr, addr+size, prot, FALSE); 358 } 359 break; 360 361 case RLIMIT_NOFILE: 362 if (limp->rlim_cur > maxfiles) 363 limp->rlim_cur = maxfiles; 364 if (limp->rlim_max > maxfiles) 365 limp->rlim_max = maxfiles; 366 break; 367 368 case RLIMIT_NPROC: 369 if (limp->rlim_cur > maxproc) 370 limp->rlim_cur = maxproc; 371 if (limp->rlim_max > maxproc) 372 limp->rlim_max = maxproc; 373 break; 374 } 375 *alimp = *limp; 376 return (0); 377 } 378 379 /* ARGSUSED */ 380 int 381 sys_getrlimit(struct lwp *l, void *v, register_t *retval) 382 { 383 struct sys_getrlimit_args /* { 384 syscallarg(int) which; 385 syscallarg(struct rlimit *) rlp; 386 } */ *uap = v; 387 struct proc *p = l->l_proc; 388 int which = SCARG(uap, which); 389 390 if ((u_int)which >= RLIM_NLIMITS) 391 return (EINVAL); 392 return (copyout(&p->p_rlimit[which], SCARG(uap, rlp), 393 sizeof(struct rlimit))); 394 } 395 396 /* 397 * Transform the running time and tick information in proc p into user, 398 * system, and interrupt time usage. 399 * 400 * Should be called with p->p_smutex held unless called from exit1(). 401 */ 402 void 403 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 404 struct timeval *ip, struct timeval *rp) 405 { 406 u_quad_t u, st, ut, it, tot; 407 unsigned long sec; 408 long usec; 409 struct timeval tv; 410 struct lwp *l; 411 412 mutex_enter(&p->p_stmutex); 413 st = p->p_sticks; 414 ut = p->p_uticks; 415 it = p->p_iticks; 416 mutex_exit(&p->p_stmutex); 417 418 sec = p->p_rtime.tv_sec; 419 usec = p->p_rtime.tv_usec; 420 421 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 422 lwp_lock(l); 423 sec += l->l_rtime.tv_sec; 424 if ((usec += l->l_rtime.tv_usec) >= 1000000) { 425 sec++; 426 usec -= 1000000; 427 } 428 if (l->l_cpu == curcpu()) { 429 struct schedstate_percpu *spc; 430 431 KDASSERT(l->l_cpu != NULL); 432 spc = &l->l_cpu->ci_schedstate; 433 434 /* 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 microtime(&tv); 442 sec += tv.tv_sec - spc->spc_runtime.tv_sec; 443 usec += tv.tv_usec - spc->spc_runtime.tv_usec; 444 if (usec >= 1000000) { 445 sec++; 446 usec -= 1000000; 447 } 448 } 449 lwp_unlock(l); 450 } 451 452 tot = st + ut + it; 453 u = sec * 1000000ull + usec; 454 455 if (tot == 0) { 456 /* No ticks, so can't use to share time out, split 50-50 */ 457 st = ut = u / 2; 458 } else { 459 st = (u * st) / tot; 460 ut = (u * ut) / tot; 461 } 462 if (sp != NULL) { 463 sp->tv_sec = st / 1000000; 464 sp->tv_usec = st % 1000000; 465 } 466 if (up != NULL) { 467 up->tv_sec = ut / 1000000; 468 up->tv_usec = ut % 1000000; 469 } 470 if (ip != NULL) { 471 if (it != 0) 472 it = (u * it) / tot; 473 ip->tv_sec = it / 1000000; 474 ip->tv_usec = it % 1000000; 475 } 476 if (rp != NULL) { 477 rp->tv_sec = sec; 478 rp->tv_usec = usec; 479 } 480 } 481 482 /* ARGSUSED */ 483 int 484 sys_getrusage(struct lwp *l, void *v, register_t *retval) 485 { 486 struct sys_getrusage_args /* { 487 syscallarg(int) who; 488 syscallarg(struct rusage *) rusage; 489 } */ *uap = v; 490 struct rusage *rup; 491 struct proc *p = l->l_proc; 492 493 switch (SCARG(uap, who)) { 494 495 case RUSAGE_SELF: 496 rup = &p->p_stats->p_ru; 497 mutex_enter(&p->p_smutex); 498 calcru(p, &rup->ru_utime, &rup->ru_stime, NULL, NULL); 499 mutex_exit(&p->p_smutex); 500 break; 501 502 case RUSAGE_CHILDREN: 503 rup = &p->p_stats->p_cru; 504 break; 505 506 default: 507 return (EINVAL); 508 } 509 return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage))); 510 } 511 512 void 513 ruadd(struct rusage *ru, struct rusage *ru2) 514 { 515 long *ip, *ip2; 516 int i; 517 518 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 519 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 520 if (ru->ru_maxrss < ru2->ru_maxrss) 521 ru->ru_maxrss = ru2->ru_maxrss; 522 ip = &ru->ru_first; ip2 = &ru2->ru_first; 523 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 524 *ip++ += *ip2++; 525 } 526 527 /* 528 * Make a copy of the plimit structure. 529 * We share these structures copy-on-write after fork, 530 * and copy when a limit is changed. 531 */ 532 struct plimit * 533 limcopy(struct plimit *lim) 534 { 535 struct plimit *newlim; 536 size_t l = 0; 537 538 simple_lock(&lim->p_slock); 539 if (lim->pl_corename != defcorename) 540 l = strlen(lim->pl_corename) + 1; 541 simple_unlock(&lim->p_slock); 542 543 newlim = pool_get(&plimit_pool, PR_WAITOK); 544 simple_lock_init(&newlim->p_slock); 545 newlim->p_lflags = 0; 546 newlim->p_refcnt = 1; 547 newlim->pl_corename = (l != 0) 548 ? malloc(l, M_TEMP, M_WAITOK) 549 : defcorename; 550 551 simple_lock(&lim->p_slock); 552 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 553 sizeof(struct rlimit) * RLIM_NLIMITS); 554 555 if (l != 0) 556 strlcpy(newlim->pl_corename, lim->pl_corename, l); 557 simple_unlock(&lim->p_slock); 558 559 return (newlim); 560 } 561 562 void 563 limfree(struct plimit *lim) 564 { 565 int n; 566 567 simple_lock(&lim->p_slock); 568 n = --lim->p_refcnt; 569 simple_unlock(&lim->p_slock); 570 if (n > 0) 571 return; 572 #ifdef DIAGNOSTIC 573 if (n < 0) 574 panic("limfree"); 575 #endif 576 if (lim->pl_corename != defcorename) 577 free(lim->pl_corename, M_TEMP); 578 pool_put(&plimit_pool, lim); 579 } 580 581 struct pstats * 582 pstatscopy(struct pstats *ps) 583 { 584 585 struct pstats *newps; 586 587 newps = pool_get(&pstats_pool, PR_WAITOK); 588 589 memset(&newps->pstat_startzero, 0, 590 (unsigned) ((caddr_t)&newps->pstat_endzero - 591 (caddr_t)&newps->pstat_startzero)); 592 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 593 ((caddr_t)&newps->pstat_endcopy - 594 (caddr_t)&newps->pstat_startcopy)); 595 596 return (newps); 597 598 } 599 600 void 601 pstatsfree(struct pstats *ps) 602 { 603 604 pool_put(&pstats_pool, ps); 605 } 606 607 /* 608 * sysctl interface in five parts 609 */ 610 611 /* 612 * a routine for sysctl proc subtree helpers that need to pick a valid 613 * process by pid. 614 */ 615 static int 616 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 617 { 618 struct proc *ptmp; 619 int error = 0; 620 621 if (pid == PROC_CURPROC) 622 ptmp = l->l_proc; 623 else if ((ptmp = pfind(pid)) == NULL) 624 error = ESRCH; 625 626 *p2 = ptmp; 627 return (error); 628 } 629 630 /* 631 * sysctl helper routine for setting a process's specific corefile 632 * name. picks the process based on the given pid and checks the 633 * correctness of the new value. 634 */ 635 static int 636 sysctl_proc_corename(SYSCTLFN_ARGS) 637 { 638 struct proc *ptmp; 639 struct plimit *lim; 640 int error = 0, len; 641 char *cname; 642 char *tmp; 643 struct sysctlnode node; 644 645 /* 646 * is this all correct? 647 */ 648 if (namelen != 0) 649 return (EINVAL); 650 if (name[-1] != PROC_PID_CORENAME) 651 return (EINVAL); 652 653 /* 654 * whom are we tweaking? 655 */ 656 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 657 if (error) 658 return (error); 659 660 /* XXX this should be in p_find() */ 661 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 662 ptmp, NULL, NULL, NULL); 663 if (error) 664 return (error); 665 666 cname = PNBUF_GET(); 667 /* 668 * let them modify a temporary copy of the core name 669 */ 670 node = *rnode; 671 strlcpy(cname, ptmp->p_limit->pl_corename, MAXPATHLEN); 672 node.sysctl_data = cname; 673 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 674 675 /* 676 * if that failed, or they have nothing new to say, or we've 677 * heard it before... 678 */ 679 if (error || newp == NULL || 680 strcmp(cname, ptmp->p_limit->pl_corename) == 0) { 681 goto done; 682 } 683 684 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 685 ptmp, cname, NULL, NULL); 686 if (error) 687 return (error); 688 689 /* 690 * no error yet and cname now has the new core name in it. 691 * let's see if it looks acceptable. it must be either "core" 692 * or end in ".core" or "/core". 693 */ 694 len = strlen(cname); 695 if (len < 4) { 696 error = EINVAL; 697 } else if (strcmp(cname + len - 4, "core") != 0) { 698 error = EINVAL; 699 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 700 error = EINVAL; 701 } 702 if (error != 0) { 703 goto done; 704 } 705 706 /* 707 * hmm...looks good. now...where do we put it? 708 */ 709 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 710 if (tmp == NULL) { 711 error = ENOMEM; 712 goto done; 713 } 714 strlcpy(tmp, cname, len + 1); 715 716 lim = ptmp->p_limit; 717 if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) { 718 ptmp->p_limit = limcopy(lim); 719 limfree(lim); 720 lim = ptmp->p_limit; 721 } 722 if (lim->pl_corename != defcorename) 723 free(lim->pl_corename, M_TEMP); 724 lim->pl_corename = tmp; 725 done: 726 PNBUF_PUT(cname); 727 return error; 728 } 729 730 /* 731 * sysctl helper routine for checking/setting a process's stop flags, 732 * one for fork and one for exec. 733 */ 734 static int 735 sysctl_proc_stop(SYSCTLFN_ARGS) 736 { 737 struct proc *ptmp; 738 int i, f, error = 0; 739 struct sysctlnode node; 740 741 if (namelen != 0) 742 return (EINVAL); 743 744 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 745 if (error) 746 return (error); 747 748 /* XXX this should be in p_find() */ 749 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 750 ptmp, NULL, NULL, NULL); 751 if (error) 752 return (error); 753 754 switch (rnode->sysctl_num) { 755 case PROC_PID_STOPFORK: 756 f = PS_STOPFORK; 757 break; 758 case PROC_PID_STOPEXEC: 759 f = PS_STOPEXEC; 760 break; 761 case PROC_PID_STOPEXIT: 762 f = PS_STOPEXIT; 763 break; 764 default: 765 return (EINVAL); 766 } 767 768 i = (ptmp->p_flag & f) ? 1 : 0; 769 node = *rnode; 770 node.sysctl_data = &i; 771 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 772 if (error || newp == NULL) 773 return (error); 774 775 mutex_enter(&ptmp->p_smutex); 776 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 777 ptmp, KAUTH_ARG(f), NULL, NULL); 778 if (error) 779 return (error); 780 if (i) 781 ptmp->p_sflag |= f; 782 else 783 ptmp->p_sflag &= ~f; 784 mutex_exit(&ptmp->p_smutex); 785 786 return (0); 787 } 788 789 /* 790 * sysctl helper routine for a process's rlimits as exposed by sysctl. 791 */ 792 static int 793 sysctl_proc_plimit(SYSCTLFN_ARGS) 794 { 795 struct proc *ptmp; 796 u_int limitno; 797 int which, error = 0; 798 struct rlimit alim; 799 struct sysctlnode node; 800 801 if (namelen != 0) 802 return (EINVAL); 803 804 which = name[-1]; 805 if (which != PROC_PID_LIMIT_TYPE_SOFT && 806 which != PROC_PID_LIMIT_TYPE_HARD) 807 return (EINVAL); 808 809 limitno = name[-2] - 1; 810 if (limitno >= RLIM_NLIMITS) 811 return (EINVAL); 812 813 if (name[-3] != PROC_PID_LIMIT) 814 return (EINVAL); 815 816 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 817 if (error) 818 return (error); 819 820 /* XXX this should be in p_find() */ 821 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 822 ptmp, NULL, NULL, NULL); 823 if (error) 824 return (error); 825 826 node = *rnode; 827 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 828 if (which == PROC_PID_LIMIT_TYPE_HARD) 829 node.sysctl_data = &alim.rlim_max; 830 else 831 node.sysctl_data = &alim.rlim_cur; 832 833 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 834 if (error || newp == NULL) 835 return (error); 836 837 return (dosetrlimit(l, ptmp, limitno, &alim)); 838 } 839 840 /* 841 * and finally, the actually glue that sticks it to the tree 842 */ 843 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup") 844 { 845 846 sysctl_createv(clog, 0, NULL, NULL, 847 CTLFLAG_PERMANENT, 848 CTLTYPE_NODE, "proc", NULL, 849 NULL, 0, NULL, 0, 850 CTL_PROC, CTL_EOL); 851 sysctl_createv(clog, 0, NULL, NULL, 852 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 853 CTLTYPE_NODE, "curproc", 854 SYSCTL_DESCR("Per-process settings"), 855 NULL, 0, NULL, 0, 856 CTL_PROC, PROC_CURPROC, CTL_EOL); 857 858 sysctl_createv(clog, 0, NULL, NULL, 859 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 860 CTLTYPE_STRING, "corename", 861 SYSCTL_DESCR("Core file name"), 862 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 863 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 864 sysctl_createv(clog, 0, NULL, NULL, 865 CTLFLAG_PERMANENT, 866 CTLTYPE_NODE, "rlimit", 867 SYSCTL_DESCR("Process limits"), 868 NULL, 0, NULL, 0, 869 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 870 871 #define create_proc_plimit(s, n) do { \ 872 sysctl_createv(clog, 0, NULL, NULL, \ 873 CTLFLAG_PERMANENT, \ 874 CTLTYPE_NODE, s, \ 875 SYSCTL_DESCR("Process " s " limits"), \ 876 NULL, 0, NULL, 0, \ 877 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 878 CTL_EOL); \ 879 sysctl_createv(clog, 0, NULL, NULL, \ 880 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 881 CTLTYPE_QUAD, "soft", \ 882 SYSCTL_DESCR("Process soft " s " limit"), \ 883 sysctl_proc_plimit, 0, NULL, 0, \ 884 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 885 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 886 sysctl_createv(clog, 0, NULL, NULL, \ 887 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 888 CTLTYPE_QUAD, "hard", \ 889 SYSCTL_DESCR("Process hard " s " limit"), \ 890 sysctl_proc_plimit, 0, NULL, 0, \ 891 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 892 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 893 } while (0/*CONSTCOND*/) 894 895 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 896 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 897 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 898 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 899 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 900 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 901 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 902 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 903 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 904 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 905 906 #undef create_proc_plimit 907 908 sysctl_createv(clog, 0, NULL, NULL, 909 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 910 CTLTYPE_INT, "stopfork", 911 SYSCTL_DESCR("Stop process at fork(2)"), 912 sysctl_proc_stop, 0, NULL, 0, 913 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 914 sysctl_createv(clog, 0, NULL, NULL, 915 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 916 CTLTYPE_INT, "stopexec", 917 SYSCTL_DESCR("Stop process at execve(2)"), 918 sysctl_proc_stop, 0, NULL, 0, 919 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 920 sysctl_createv(clog, 0, NULL, NULL, 921 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 922 CTLTYPE_INT, "stopexit", 923 SYSCTL_DESCR("Stop process before completing exit"), 924 sysctl_proc_stop, 0, NULL, 0, 925 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 926 } 927 928 struct uidinfo * 929 uid_find(uid_t uid) 930 { 931 struct uidinfo *uip; 932 struct uidinfo *newuip = NULL; 933 struct uihashhead *uipp; 934 935 uipp = UIHASH(uid); 936 937 again: 938 simple_lock(&uihashtbl_slock); 939 LIST_FOREACH(uip, uipp, ui_hash) 940 if (uip->ui_uid == uid) { 941 simple_unlock(&uihashtbl_slock); 942 if (newuip) 943 free(newuip, M_PROC); 944 return uip; 945 } 946 947 if (newuip == NULL) { 948 simple_unlock(&uihashtbl_slock); 949 newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO); 950 goto again; 951 } 952 uip = newuip; 953 954 LIST_INSERT_HEAD(uipp, uip, ui_hash); 955 uip->ui_uid = uid; 956 simple_lock_init(&uip->ui_slock); 957 simple_unlock(&uihashtbl_slock); 958 959 return uip; 960 } 961 962 /* 963 * Change the count associated with number of processes 964 * a given user is using. 965 */ 966 int 967 chgproccnt(uid_t uid, int diff) 968 { 969 struct uidinfo *uip; 970 int s; 971 972 if (diff == 0) 973 return 0; 974 975 uip = uid_find(uid); 976 UILOCK(uip, s); 977 uip->ui_proccnt += diff; 978 KASSERT(uip->ui_proccnt >= 0); 979 UIUNLOCK(uip, s); 980 return uip->ui_proccnt; 981 } 982 983 int 984 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax) 985 { 986 rlim_t nsb; 987 int s; 988 989 UILOCK(uip, s); 990 nsb = uip->ui_sbsize + to - *hiwat; 991 if (to > *hiwat && nsb > xmax) { 992 UIUNLOCK(uip, s); 993 splx(s); 994 return 0; 995 } 996 *hiwat = to; 997 uip->ui_sbsize = nsb; 998 KASSERT(uip->ui_sbsize >= 0); 999 UIUNLOCK(uip, s); 1000 return 1; 1001 } 1002
Indexes created Sun Sep 21 14:09:52 GMT 2025