kern_resource.c revision 1.116.2.5
1 /* $NetBSD: kern_resource.c,v 1.116.2.5 2007/08/20 21:27:32 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 1.116.2.5 2007/08/20 21:27:32 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 kmutex_t uihashtbl_lock; 69 70 /* 71 * Resource controls and accounting. 72 */ 73 74 int 75 sys_getpriority(struct lwp *l, void *v, register_t *retval) 76 { 77 struct sys_getpriority_args /* { 78 syscallarg(int) which; 79 syscallarg(id_t) who; 80 } */ *uap = v; 81 struct proc *curp = l->l_proc, *p; 82 int low = NZERO + PRIO_MAX + 1; 83 int who = SCARG(uap, who); 84 85 mutex_enter(&proclist_lock); 86 switch (SCARG(uap, which)) { 87 case PRIO_PROCESS: 88 if (who == 0) 89 p = curp; 90 else 91 p = p_find(who, PFIND_LOCKED); 92 if (p != NULL) 93 low = p->p_nice; 94 break; 95 96 case PRIO_PGRP: { 97 struct pgrp *pg; 98 99 if (who == 0) 100 pg = curp->p_pgrp; 101 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 102 break; 103 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 104 if (p->p_nice < low) 105 low = p->p_nice; 106 } 107 break; 108 } 109 110 case PRIO_USER: 111 if (who == 0) 112 who = (int)kauth_cred_geteuid(l->l_cred); 113 PROCLIST_FOREACH(p, &allproc) { 114 mutex_enter(&p->p_mutex); 115 if (kauth_cred_geteuid(p->p_cred) == 116 (uid_t)who && p->p_nice < low) 117 low = p->p_nice; 118 mutex_exit(&p->p_mutex); 119 } 120 break; 121 122 default: 123 mutex_exit(&proclist_lock); 124 return (EINVAL); 125 } 126 mutex_exit(&proclist_lock); 127 128 if (low == NZERO + PRIO_MAX + 1) 129 return (ESRCH); 130 *retval = low - NZERO; 131 return (0); 132 } 133 134 /* ARGSUSED */ 135 int 136 sys_setpriority(struct lwp *l, void *v, register_t *retval) 137 { 138 struct sys_setpriority_args /* { 139 syscallarg(int) which; 140 syscallarg(id_t) who; 141 syscallarg(int) prio; 142 } */ *uap = v; 143 struct proc *curp = l->l_proc, *p; 144 int found = 0, error = 0; 145 int who = SCARG(uap, who); 146 147 mutex_enter(&proclist_lock); 148 switch (SCARG(uap, which)) { 149 case PRIO_PROCESS: 150 if (who == 0) 151 p = curp; 152 else 153 p = p_find(who, PFIND_LOCKED); 154 if (p != 0) { 155 mutex_enter(&p->p_mutex); 156 error = donice(l, p, SCARG(uap, prio)); 157 mutex_exit(&p->p_mutex); 158 } 159 found++; 160 break; 161 162 case PRIO_PGRP: { 163 struct pgrp *pg; 164 165 if (who == 0) 166 pg = curp->p_pgrp; 167 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 168 break; 169 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 170 mutex_enter(&p->p_mutex); 171 error = donice(l, p, SCARG(uap, prio)); 172 mutex_exit(&p->p_mutex); 173 found++; 174 } 175 break; 176 } 177 178 case PRIO_USER: 179 if (who == 0) 180 who = (int)kauth_cred_geteuid(l->l_cred); 181 PROCLIST_FOREACH(p, &allproc) { 182 mutex_enter(&p->p_mutex); 183 if (kauth_cred_geteuid(p->p_cred) == 184 (uid_t)SCARG(uap, who)) { 185 error = donice(l, p, SCARG(uap, prio)); 186 found++; 187 } 188 mutex_exit(&p->p_mutex); 189 } 190 break; 191 192 default: 193 error = EINVAL; 194 break; 195 } 196 mutex_exit(&proclist_lock); 197 if (found == 0) 198 return (ESRCH); 199 return (error); 200 } 201 202 /* 203 * Renice a process. 204 * 205 * Call with the target process' credentials locked. 206 */ 207 int 208 donice(struct lwp *l, struct proc *chgp, int n) 209 { 210 kauth_cred_t cred = l->l_cred; 211 int onice; 212 213 KASSERT(mutex_owned(&chgp->p_mutex)); 214 215 if (n > PRIO_MAX) 216 n = PRIO_MAX; 217 if (n < PRIO_MIN) 218 n = PRIO_MIN; 219 n += NZERO; 220 onice = chgp->p_nice; 221 onice = chgp->p_nice; 222 223 again: 224 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 225 KAUTH_ARG(n), NULL, NULL)) 226 return (EACCES); 227 mutex_spin_enter(&chgp->p_stmutex); 228 if (onice != chgp->p_nice) { 229 mutex_spin_exit(&chgp->p_stmutex); 230 goto again; 231 } 232 sched_nice(chgp, n); 233 mutex_spin_exit(&chgp->p_stmutex); 234 return (0); 235 } 236 237 /* ARGSUSED */ 238 int 239 sys_setrlimit(struct lwp *l, void *v, register_t *retval) 240 { 241 struct sys_setrlimit_args /* { 242 syscallarg(int) which; 243 syscallarg(const struct rlimit *) rlp; 244 } */ *uap = v; 245 int which = SCARG(uap, which); 246 struct rlimit alim; 247 int error; 248 249 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 250 if (error) 251 return (error); 252 return (dosetrlimit(l, l->l_proc, which, &alim)); 253 } 254 255 int 256 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 257 { 258 struct rlimit *alimp; 259 struct plimit *oldplim; 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 alimp = &p->p_rlimit[which]; 269 /* if we don't change the value, no need to limcopy() */ 270 if (limp->rlim_cur == alimp->rlim_cur && 271 limp->rlim_max == alimp->rlim_max) 272 return 0; 273 274 if (limp->rlim_cur > limp->rlim_max) { 275 /* 276 * This is programming error. According to SUSv2, we should 277 * return error in this case. 278 */ 279 return (EINVAL); 280 } 281 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 282 p, limp, KAUTH_ARG(which), NULL); 283 if (error) 284 return (error); 285 286 mutex_enter(&p->p_mutex); 287 if (p->p_limit->p_refcnt > 1 && 288 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 289 oldplim = p->p_limit; 290 p->p_limit = limcopy(p); 291 limfree(oldplim); 292 alimp = &p->p_rlimit[which]; 293 } 294 295 switch (which) { 296 297 case RLIMIT_DATA: 298 if (limp->rlim_cur > maxdmap) 299 limp->rlim_cur = maxdmap; 300 if (limp->rlim_max > maxdmap) 301 limp->rlim_max = maxdmap; 302 break; 303 304 case RLIMIT_STACK: 305 if (limp->rlim_cur > maxsmap) 306 limp->rlim_cur = maxsmap; 307 if (limp->rlim_max > maxsmap) 308 limp->rlim_max = maxsmap; 309 310 /* 311 * Return EINVAL if the new stack size limit is lower than 312 * current usage. Otherwise, the process would get SIGSEGV the 313 * moment it would try to access anything on it's current stack. 314 * This conforms to SUSv2. 315 */ 316 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 317 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 318 mutex_exit(&p->p_mutex); 319 return (EINVAL); 320 } 321 322 /* 323 * Stack is allocated to the max at exec time with 324 * only "rlim_cur" bytes accessible (In other words, 325 * allocates stack dividing two contiguous regions at 326 * "rlim_cur" bytes boundary). 327 * 328 * Since allocation is done in terms of page, roundup 329 * "rlim_cur" (otherwise, contiguous regions 330 * overlap). If stack limit is going up make more 331 * accessible, if going down make inaccessible. 332 */ 333 limp->rlim_cur = round_page(limp->rlim_cur); 334 if (limp->rlim_cur != alimp->rlim_cur) { 335 vaddr_t addr; 336 vsize_t size; 337 vm_prot_t prot; 338 339 if (limp->rlim_cur > alimp->rlim_cur) { 340 prot = VM_PROT_READ | VM_PROT_WRITE; 341 size = limp->rlim_cur - alimp->rlim_cur; 342 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 343 limp->rlim_cur; 344 } else { 345 prot = VM_PROT_NONE; 346 size = alimp->rlim_cur - limp->rlim_cur; 347 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 348 alimp->rlim_cur; 349 } 350 (void) uvm_map_protect(&p->p_vmspace->vm_map, 351 addr, addr+size, prot, false); 352 } 353 break; 354 355 case RLIMIT_NOFILE: 356 if (limp->rlim_cur > maxfiles) 357 limp->rlim_cur = maxfiles; 358 if (limp->rlim_max > maxfiles) 359 limp->rlim_max = maxfiles; 360 break; 361 362 case RLIMIT_NPROC: 363 if (limp->rlim_cur > maxproc) 364 limp->rlim_cur = maxproc; 365 if (limp->rlim_max > maxproc) 366 limp->rlim_max = maxproc; 367 break; 368 } 369 *alimp = *limp; 370 mutex_exit(&p->p_mutex); 371 return (0); 372 } 373 374 /* ARGSUSED */ 375 int 376 sys_getrlimit(struct lwp *l, void *v, register_t *retval) 377 { 378 struct sys_getrlimit_args /* { 379 syscallarg(int) which; 380 syscallarg(struct rlimit *) rlp; 381 } */ *uap = v; 382 struct proc *p = l->l_proc; 383 int which = SCARG(uap, which); 384 struct rlimit rl; 385 386 if ((u_int)which >= RLIM_NLIMITS) 387 return (EINVAL); 388 389 mutex_enter(&p->p_mutex); 390 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 391 mutex_exit(&p->p_mutex); 392 393 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 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_spin_enter(&p->p_stmutex); 413 st = p->p_sticks; 414 ut = p->p_uticks; 415 it = p->p_iticks; 416 mutex_spin_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_flag & LW_RUNNING) != 0) { 429 /* 430 * Adjust for the current time slice. This is 431 * actually fairly important since the error 432 * here is on the order of a time quantum, 433 * which is much greater than the sampling 434 * error. 435 */ 436 microtime(&tv); 437 sec += tv.tv_sec - l->l_stime.tv_sec; 438 usec += tv.tv_usec - l->l_stime.tv_usec; 439 if (usec >= 1000000) { 440 sec++; 441 usec -= 1000000; 442 } 443 } 444 lwp_unlock(l); 445 } 446 447 tot = st + ut + it; 448 u = sec * 1000000ull + usec; 449 450 if (tot == 0) { 451 /* No ticks, so can't use to share time out, split 50-50 */ 452 st = ut = u / 2; 453 } else { 454 st = (u * st) / tot; 455 ut = (u * ut) / tot; 456 } 457 if (sp != NULL) { 458 sp->tv_sec = st / 1000000; 459 sp->tv_usec = st % 1000000; 460 } 461 if (up != NULL) { 462 up->tv_sec = ut / 1000000; 463 up->tv_usec = ut % 1000000; 464 } 465 if (ip != NULL) { 466 if (it != 0) 467 it = (u * it) / tot; 468 ip->tv_sec = it / 1000000; 469 ip->tv_usec = it % 1000000; 470 } 471 if (rp != NULL) { 472 rp->tv_sec = sec; 473 rp->tv_usec = usec; 474 } 475 } 476 477 /* ARGSUSED */ 478 int 479 sys_getrusage(struct lwp *l, void *v, register_t *retval) 480 { 481 struct sys_getrusage_args /* { 482 syscallarg(int) who; 483 syscallarg(struct rusage *) rusage; 484 } */ *uap = v; 485 struct rusage ru; 486 struct proc *p = l->l_proc; 487 488 switch (SCARG(uap, who)) { 489 case RUSAGE_SELF: 490 mutex_enter(&p->p_smutex); 491 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 492 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 493 mutex_exit(&p->p_smutex); 494 break; 495 496 case RUSAGE_CHILDREN: 497 mutex_enter(&p->p_smutex); 498 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 499 mutex_exit(&p->p_smutex); 500 break; 501 502 default: 503 return EINVAL; 504 } 505 506 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 507 } 508 509 void 510 ruadd(struct rusage *ru, struct rusage *ru2) 511 { 512 long *ip, *ip2; 513 int i; 514 515 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 516 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 517 if (ru->ru_maxrss < ru2->ru_maxrss) 518 ru->ru_maxrss = ru2->ru_maxrss; 519 ip = &ru->ru_first; ip2 = &ru2->ru_first; 520 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 521 *ip++ += *ip2++; 522 } 523 524 /* 525 * Make a copy of the plimit structure. 526 * We share these structures copy-on-write after fork, 527 * and copy when a limit is changed. 528 * 529 * XXXSMP This is atrocious, need to simplify. 530 */ 531 struct plimit * 532 limcopy(struct proc *p) 533 { 534 struct plimit *lim, *newlim; 535 char *corename; 536 size_t l; 537 538 KASSERT(mutex_owned(&p->p_mutex)); 539 540 mutex_exit(&p->p_mutex); 541 newlim = pool_get(&plimit_pool, PR_WAITOK); 542 mutex_init(&newlim->p_lock, MUTEX_DEFAULT, IPL_NONE); 543 newlim->p_lflags = 0; 544 newlim->p_refcnt = 1; 545 mutex_enter(&p->p_mutex); 546 547 for (;;) { 548 lim = p->p_limit; 549 mutex_enter(&lim->p_lock); 550 if (lim->pl_corename != defcorename) { 551 l = strlen(lim->pl_corename) + 1; 552 553 mutex_exit(&lim->p_lock); 554 mutex_exit(&p->p_mutex); 555 corename = malloc(l, M_TEMP, M_WAITOK); 556 mutex_enter(&p->p_mutex); 557 mutex_enter(&lim->p_lock); 558 559 if (l != strlen(lim->pl_corename) + 1) { 560 mutex_exit(&lim->p_lock); 561 mutex_exit(&p->p_mutex); 562 free(corename, M_TEMP); 563 mutex_enter(&p->p_mutex); 564 continue; 565 } 566 } else 567 l = 0; 568 569 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 570 sizeof(struct rlimit) * RLIM_NLIMITS); 571 if (l != 0) 572 strlcpy(newlim->pl_corename, lim->pl_corename, l); 573 else 574 newlim->pl_corename = defcorename; 575 mutex_exit(&lim->p_lock); 576 break; 577 } 578 579 return (newlim); 580 } 581 582 void 583 limfree(struct plimit *lim) 584 { 585 int n; 586 587 mutex_enter(&lim->p_lock); 588 n = --lim->p_refcnt; 589 mutex_exit(&lim->p_lock); 590 if (n > 0) 591 return; 592 #ifdef DIAGNOSTIC 593 if (n < 0) 594 panic("limfree"); 595 #endif 596 if (lim->pl_corename != defcorename) 597 free(lim->pl_corename, M_TEMP); 598 mutex_destroy(&lim->p_lock); 599 pool_put(&plimit_pool, lim); 600 } 601 602 struct pstats * 603 pstatscopy(struct pstats *ps) 604 { 605 606 struct pstats *newps; 607 608 newps = pool_get(&pstats_pool, PR_WAITOK); 609 610 memset(&newps->pstat_startzero, 0, 611 (unsigned) ((char *)&newps->pstat_endzero - 612 (char *)&newps->pstat_startzero)); 613 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 614 ((char *)&newps->pstat_endcopy - 615 (char *)&newps->pstat_startcopy)); 616 617 return (newps); 618 619 } 620 621 void 622 pstatsfree(struct pstats *ps) 623 { 624 625 pool_put(&pstats_pool, ps); 626 } 627 628 /* 629 * sysctl interface in five parts 630 */ 631 632 /* 633 * a routine for sysctl proc subtree helpers that need to pick a valid 634 * process by pid. 635 */ 636 static int 637 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 638 { 639 struct proc *ptmp; 640 int error = 0; 641 642 if (pid == PROC_CURPROC) 643 ptmp = l->l_proc; 644 else if ((ptmp = pfind(pid)) == NULL) 645 error = ESRCH; 646 647 *p2 = ptmp; 648 return (error); 649 } 650 651 /* 652 * sysctl helper routine for setting a process's specific corefile 653 * name. picks the process based on the given pid and checks the 654 * correctness of the new value. 655 */ 656 static int 657 sysctl_proc_corename(SYSCTLFN_ARGS) 658 { 659 struct proc *ptmp; 660 struct plimit *lim; 661 int error = 0, len; 662 char *cname; 663 char *tmp; 664 struct sysctlnode node; 665 666 /* 667 * is this all correct? 668 */ 669 if (namelen != 0) 670 return (EINVAL); 671 if (name[-1] != PROC_PID_CORENAME) 672 return (EINVAL); 673 674 /* 675 * whom are we tweaking? 676 */ 677 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 678 if (error) 679 return (error); 680 681 /* XXX this should be in p_find() */ 682 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 683 ptmp, NULL, NULL, NULL); 684 if (error) 685 return (error); 686 687 cname = PNBUF_GET(); 688 /* 689 * let them modify a temporary copy of the core name 690 */ 691 node = *rnode; 692 strlcpy(cname, ptmp->p_limit->pl_corename, MAXPATHLEN); 693 node.sysctl_data = cname; 694 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 695 696 /* 697 * if that failed, or they have nothing new to say, or we've 698 * heard it before... 699 */ 700 if (error || newp == NULL || 701 strcmp(cname, ptmp->p_limit->pl_corename) == 0) { 702 goto done; 703 } 704 705 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 706 ptmp, cname, NULL, NULL); 707 if (error) 708 return (error); 709 710 /* 711 * no error yet and cname now has the new core name in it. 712 * let's see if it looks acceptable. it must be either "core" 713 * or end in ".core" or "/core". 714 */ 715 len = strlen(cname); 716 if (len < 4) { 717 error = EINVAL; 718 } else if (strcmp(cname + len - 4, "core") != 0) { 719 error = EINVAL; 720 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 721 error = EINVAL; 722 } 723 if (error != 0) { 724 goto done; 725 } 726 727 /* 728 * hmm...looks good. now...where do we put it? 729 */ 730 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 731 if (tmp == NULL) { 732 error = ENOMEM; 733 goto done; 734 } 735 strlcpy(tmp, cname, len + 1); 736 737 mutex_enter(&ptmp->p_mutex); 738 lim = ptmp->p_limit; 739 if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) { 740 ptmp->p_limit = limcopy(ptmp); 741 limfree(lim); 742 lim = ptmp->p_limit; 743 } 744 if (lim->pl_corename != defcorename) 745 free(lim->pl_corename, M_TEMP); 746 lim->pl_corename = tmp; 747 mutex_exit(&ptmp->p_mutex); 748 done: 749 PNBUF_PUT(cname); 750 return error; 751 } 752 753 /* 754 * sysctl helper routine for checking/setting a process's stop flags, 755 * one for fork and one for exec. 756 */ 757 static int 758 sysctl_proc_stop(SYSCTLFN_ARGS) 759 { 760 struct proc *ptmp; 761 int i, f, error = 0; 762 struct sysctlnode node; 763 764 if (namelen != 0) 765 return (EINVAL); 766 767 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 768 if (error) 769 return (error); 770 771 /* XXX this should be in p_find() */ 772 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 773 ptmp, NULL, NULL, NULL); 774 if (error) 775 return (error); 776 777 switch (rnode->sysctl_num) { 778 case PROC_PID_STOPFORK: 779 f = PS_STOPFORK; 780 break; 781 case PROC_PID_STOPEXEC: 782 f = PS_STOPEXEC; 783 break; 784 case PROC_PID_STOPEXIT: 785 f = PS_STOPEXIT; 786 break; 787 default: 788 return (EINVAL); 789 } 790 791 i = (ptmp->p_flag & f) ? 1 : 0; 792 node = *rnode; 793 node.sysctl_data = &i; 794 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 795 if (error || newp == NULL) 796 return (error); 797 798 mutex_enter(&ptmp->p_smutex); 799 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 800 ptmp, KAUTH_ARG(f), NULL, NULL); 801 if (error) 802 return (error); 803 if (i) 804 ptmp->p_sflag |= f; 805 else 806 ptmp->p_sflag &= ~f; 807 mutex_exit(&ptmp->p_smutex); 808 809 return (0); 810 } 811 812 /* 813 * sysctl helper routine for a process's rlimits as exposed by sysctl. 814 */ 815 static int 816 sysctl_proc_plimit(SYSCTLFN_ARGS) 817 { 818 struct proc *ptmp; 819 u_int limitno; 820 int which, error = 0; 821 struct rlimit alim; 822 struct sysctlnode node; 823 824 if (namelen != 0) 825 return (EINVAL); 826 827 which = name[-1]; 828 if (which != PROC_PID_LIMIT_TYPE_SOFT && 829 which != PROC_PID_LIMIT_TYPE_HARD) 830 return (EINVAL); 831 832 limitno = name[-2] - 1; 833 if (limitno >= RLIM_NLIMITS) 834 return (EINVAL); 835 836 if (name[-3] != PROC_PID_LIMIT) 837 return (EINVAL); 838 839 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 840 if (error) 841 return (error); 842 843 /* XXX this should be in p_find() */ 844 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 845 ptmp, NULL, NULL, NULL); 846 if (error) 847 return (error); 848 849 node = *rnode; 850 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 851 if (which == PROC_PID_LIMIT_TYPE_HARD) 852 node.sysctl_data = &alim.rlim_max; 853 else 854 node.sysctl_data = &alim.rlim_cur; 855 856 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 857 if (error || newp == NULL) 858 return (error); 859 860 return (dosetrlimit(l, ptmp, limitno, &alim)); 861 } 862 863 /* 864 * and finally, the actually glue that sticks it to the tree 865 */ 866 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup") 867 { 868 869 sysctl_createv(clog, 0, NULL, NULL, 870 CTLFLAG_PERMANENT, 871 CTLTYPE_NODE, "proc", NULL, 872 NULL, 0, NULL, 0, 873 CTL_PROC, CTL_EOL); 874 sysctl_createv(clog, 0, NULL, NULL, 875 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 876 CTLTYPE_NODE, "curproc", 877 SYSCTL_DESCR("Per-process settings"), 878 NULL, 0, NULL, 0, 879 CTL_PROC, PROC_CURPROC, CTL_EOL); 880 881 sysctl_createv(clog, 0, NULL, NULL, 882 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 883 CTLTYPE_STRING, "corename", 884 SYSCTL_DESCR("Core file name"), 885 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 886 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 887 sysctl_createv(clog, 0, NULL, NULL, 888 CTLFLAG_PERMANENT, 889 CTLTYPE_NODE, "rlimit", 890 SYSCTL_DESCR("Process limits"), 891 NULL, 0, NULL, 0, 892 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 893 894 #define create_proc_plimit(s, n) do { \ 895 sysctl_createv(clog, 0, NULL, NULL, \ 896 CTLFLAG_PERMANENT, \ 897 CTLTYPE_NODE, s, \ 898 SYSCTL_DESCR("Process " s " limits"), \ 899 NULL, 0, NULL, 0, \ 900 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 901 CTL_EOL); \ 902 sysctl_createv(clog, 0, NULL, NULL, \ 903 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 904 CTLTYPE_QUAD, "soft", \ 905 SYSCTL_DESCR("Process soft " s " limit"), \ 906 sysctl_proc_plimit, 0, NULL, 0, \ 907 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 908 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 909 sysctl_createv(clog, 0, NULL, NULL, \ 910 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 911 CTLTYPE_QUAD, "hard", \ 912 SYSCTL_DESCR("Process hard " s " limit"), \ 913 sysctl_proc_plimit, 0, NULL, 0, \ 914 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 915 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 916 } while (0/*CONSTCOND*/) 917 918 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 919 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 920 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 921 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 922 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 923 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 924 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 925 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 926 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 927 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 928 929 #undef create_proc_plimit 930 931 sysctl_createv(clog, 0, NULL, NULL, 932 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 933 CTLTYPE_INT, "stopfork", 934 SYSCTL_DESCR("Stop process at fork(2)"), 935 sysctl_proc_stop, 0, NULL, 0, 936 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 937 sysctl_createv(clog, 0, NULL, NULL, 938 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 939 CTLTYPE_INT, "stopexec", 940 SYSCTL_DESCR("Stop process at execve(2)"), 941 sysctl_proc_stop, 0, NULL, 0, 942 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 943 sysctl_createv(clog, 0, NULL, NULL, 944 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 945 CTLTYPE_INT, "stopexit", 946 SYSCTL_DESCR("Stop process before completing exit"), 947 sysctl_proc_stop, 0, NULL, 0, 948 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 949 } 950 951 void 952 uid_init(void) 953 { 954 955 /* 956 * XXXSMP This could be at IPL_SOFTNET, but for now we want 957 * to to be deadlock free, so it must be at IPL_VM. 958 */ 959 mutex_init(&uihashtbl_lock, MUTEX_DRIVER, IPL_VM); 960 961 /* 962 * Ensure that uid 0 is always in the user hash table, as 963 * sbreserve() expects it available from interrupt context. 964 */ 965 (void)uid_find(0); 966 } 967 968 struct uidinfo * 969 uid_find(uid_t uid) 970 { 971 struct uidinfo *uip; 972 struct uidinfo *newuip = NULL; 973 struct uihashhead *uipp; 974 975 uipp = UIHASH(uid); 976 977 again: 978 mutex_enter(&uihashtbl_lock); 979 LIST_FOREACH(uip, uipp, ui_hash) 980 if (uip->ui_uid == uid) { 981 mutex_exit(&uihashtbl_lock); 982 if (newuip) { 983 free(newuip, M_PROC); 984 mutex_destroy(&newuip->ui_lock); 985 } 986 return uip; 987 } 988 if (newuip == NULL) { 989 mutex_exit(&uihashtbl_lock); 990 /* Must not be called from interrupt context. */ 991 newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO); 992 mutex_init(&newuip->ui_lock, MUTEX_DRIVER, IPL_SOFTNET); 993 goto again; 994 } 995 uip = newuip; 996 997 LIST_INSERT_HEAD(uipp, uip, ui_hash); 998 uip->ui_uid = uid; 999 mutex_exit(&uihashtbl_lock); 1000 1001 return uip; 1002 } 1003 1004 /* 1005 * Change the count associated with number of processes 1006 * a given user is using. 1007 */ 1008 int 1009 chgproccnt(uid_t uid, int diff) 1010 { 1011 struct uidinfo *uip; 1012 1013 if (diff == 0) 1014 return 0; 1015 1016 uip = uid_find(uid); 1017 mutex_enter(&uip->ui_lock); 1018 uip->ui_proccnt += diff; 1019 KASSERT(uip->ui_proccnt >= 0); 1020 mutex_exit(&uip->ui_lock); 1021 return uip->ui_proccnt; 1022 } 1023 1024 int 1025 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax) 1026 { 1027 rlim_t nsb; 1028 1029 mutex_enter(&uip->ui_lock); 1030 nsb = uip->ui_sbsize + to - *hiwat; 1031 if (to > *hiwat && nsb > xmax) { 1032 mutex_exit(&uip->ui_lock); 1033 return 0; 1034 } 1035 *hiwat = to; 1036 uip->ui_sbsize = nsb; 1037 KASSERT(uip->ui_sbsize >= 0); 1038 mutex_exit(&uip->ui_lock); 1039 return 1; 1040 } 1041
Indexes created Sun Sep 21 19:09:51 GMT 2025