kern_resource.c revision 1.122
1 /* $NetBSD: kern_resource.c,v 1.122 2007/09/29 12:22:31 dsl 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 1.122 2007/09/29 12:22:31 dsl 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 int error; 260 261 if ((u_int)which >= RLIM_NLIMITS) 262 return (EINVAL); 263 264 if (limp->rlim_cur < 0 || limp->rlim_max < 0) 265 return (EINVAL); 266 267 if (limp->rlim_cur > limp->rlim_max) { 268 /* 269 * This is programming error. According to SUSv2, we should 270 * return error in this case. 271 */ 272 return (EINVAL); 273 } 274 275 alimp = &p->p_rlimit[which]; 276 /* if we don't change the value, no need to limcopy() */ 277 if (limp->rlim_cur == alimp->rlim_cur && 278 limp->rlim_max == alimp->rlim_max) 279 return 0; 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 lim_privatise(p, false); 287 /* p->p_limit is now unchangeable */ 288 alimp = &p->p_rlimit[which]; 289 290 switch (which) { 291 292 case RLIMIT_DATA: 293 if (limp->rlim_cur > maxdmap) 294 limp->rlim_cur = maxdmap; 295 if (limp->rlim_max > maxdmap) 296 limp->rlim_max = maxdmap; 297 break; 298 299 case RLIMIT_STACK: 300 if (limp->rlim_cur > maxsmap) 301 limp->rlim_cur = maxsmap; 302 if (limp->rlim_max > maxsmap) 303 limp->rlim_max = maxsmap; 304 305 /* 306 * Return EINVAL if the new stack size limit is lower than 307 * current usage. Otherwise, the process would get SIGSEGV the 308 * moment it would try to access anything on it's current stack. 309 * This conforms to SUSv2. 310 */ 311 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 312 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 313 return (EINVAL); 314 } 315 316 /* 317 * Stack is allocated to the max at exec time with 318 * only "rlim_cur" bytes accessible (In other words, 319 * allocates stack dividing two contiguous regions at 320 * "rlim_cur" bytes boundary). 321 * 322 * Since allocation is done in terms of page, roundup 323 * "rlim_cur" (otherwise, contiguous regions 324 * overlap). If stack limit is going up make more 325 * accessible, if going down make inaccessible. 326 */ 327 limp->rlim_cur = round_page(limp->rlim_cur); 328 if (limp->rlim_cur != alimp->rlim_cur) { 329 vaddr_t addr; 330 vsize_t size; 331 vm_prot_t prot; 332 333 if (limp->rlim_cur > alimp->rlim_cur) { 334 prot = VM_PROT_READ | VM_PROT_WRITE; 335 size = limp->rlim_cur - alimp->rlim_cur; 336 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 337 limp->rlim_cur; 338 } else { 339 prot = VM_PROT_NONE; 340 size = alimp->rlim_cur - limp->rlim_cur; 341 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 342 alimp->rlim_cur; 343 } 344 (void) uvm_map_protect(&p->p_vmspace->vm_map, 345 addr, addr+size, prot, false); 346 } 347 break; 348 349 case RLIMIT_NOFILE: 350 if (limp->rlim_cur > maxfiles) 351 limp->rlim_cur = maxfiles; 352 if (limp->rlim_max > maxfiles) 353 limp->rlim_max = maxfiles; 354 break; 355 356 case RLIMIT_NPROC: 357 if (limp->rlim_cur > maxproc) 358 limp->rlim_cur = maxproc; 359 if (limp->rlim_max > maxproc) 360 limp->rlim_max = maxproc; 361 break; 362 } 363 364 mutex_enter(&p->p_limit->pl_lock); 365 *alimp = *limp; 366 mutex_exit(&p->p_limit->pl_lock); 367 return (0); 368 } 369 370 /* ARGSUSED */ 371 int 372 sys_getrlimit(struct lwp *l, void *v, register_t *retval) 373 { 374 struct sys_getrlimit_args /* { 375 syscallarg(int) which; 376 syscallarg(struct rlimit *) rlp; 377 } */ *uap = v; 378 struct proc *p = l->l_proc; 379 int which = SCARG(uap, which); 380 struct rlimit rl; 381 382 if ((u_int)which >= RLIM_NLIMITS) 383 return (EINVAL); 384 385 mutex_enter(&p->p_mutex); 386 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 387 mutex_exit(&p->p_mutex); 388 389 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 390 } 391 392 /* 393 * Transform the running time and tick information in proc p into user, 394 * system, and interrupt time usage. 395 * 396 * Should be called with p->p_smutex held unless called from exit1(). 397 */ 398 void 399 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 400 struct timeval *ip, struct timeval *rp) 401 { 402 u_quad_t u, st, ut, it, tot; 403 unsigned long sec; 404 long usec; 405 struct timeval tv; 406 struct lwp *l; 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 sec = p->p_rtime.tv_sec; 415 usec = p->p_rtime.tv_usec; 416 417 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 418 lwp_lock(l); 419 sec += l->l_rtime.tv_sec; 420 if ((usec += l->l_rtime.tv_usec) >= 1000000) { 421 sec++; 422 usec -= 1000000; 423 } 424 if (l->l_cpu == curcpu()) { 425 struct schedstate_percpu *spc; 426 427 KDASSERT(l->l_cpu != NULL); 428 spc = &l->l_cpu->ci_schedstate; 429 430 /* 431 * Adjust for the current time slice. This is 432 * actually fairly important since the error 433 * here is on the order of a time quantum, 434 * which is much greater than the sampling 435 * error. 436 */ 437 microtime(&tv); 438 sec += tv.tv_sec - spc->spc_runtime.tv_sec; 439 usec += tv.tv_usec - spc->spc_runtime.tv_usec; 440 if (usec >= 1000000) { 441 sec++; 442 usec -= 1000000; 443 } 444 } 445 lwp_unlock(l); 446 } 447 448 tot = st + ut + it; 449 u = sec * 1000000ull + usec; 450 451 if (tot == 0) { 452 /* No ticks, so can't use to share time out, split 50-50 */ 453 st = ut = u / 2; 454 } else { 455 st = (u * st) / tot; 456 ut = (u * ut) / tot; 457 } 458 if (sp != NULL) { 459 sp->tv_sec = st / 1000000; 460 sp->tv_usec = st % 1000000; 461 } 462 if (up != NULL) { 463 up->tv_sec = ut / 1000000; 464 up->tv_usec = ut % 1000000; 465 } 466 if (ip != NULL) { 467 if (it != 0) 468 it = (u * it) / tot; 469 ip->tv_sec = it / 1000000; 470 ip->tv_usec = it % 1000000; 471 } 472 if (rp != NULL) { 473 rp->tv_sec = sec; 474 rp->tv_usec = usec; 475 } 476 } 477 478 /* ARGSUSED */ 479 int 480 sys_getrusage(struct lwp *l, void *v, register_t *retval) 481 { 482 struct sys_getrusage_args /* { 483 syscallarg(int) who; 484 syscallarg(struct rusage *) rusage; 485 } */ *uap = v; 486 struct rusage ru; 487 struct proc *p = l->l_proc; 488 489 switch (SCARG(uap, who)) { 490 case RUSAGE_SELF: 491 mutex_enter(&p->p_smutex); 492 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 493 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 494 mutex_exit(&p->p_smutex); 495 break; 496 497 case RUSAGE_CHILDREN: 498 mutex_enter(&p->p_smutex); 499 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 500 mutex_exit(&p->p_smutex); 501 break; 502 503 default: 504 return EINVAL; 505 } 506 507 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 508 } 509 510 void 511 ruadd(struct rusage *ru, struct rusage *ru2) 512 { 513 long *ip, *ip2; 514 int i; 515 516 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 517 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 518 if (ru->ru_maxrss < ru2->ru_maxrss) 519 ru->ru_maxrss = ru2->ru_maxrss; 520 ip = &ru->ru_first; ip2 = &ru2->ru_first; 521 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 522 *ip++ += *ip2++; 523 } 524 525 /* 526 * Make a copy of the plimit structure. 527 * We share these structures copy-on-write after fork, 528 * and copy when a limit is changed. 529 * 530 * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure 531 * we are copying to change beneath our feet! 532 */ 533 struct plimit * 534 lim_copy(struct plimit *lim) 535 { 536 struct plimit *newlim; 537 char *corename; 538 size_t alen, len; 539 540 newlim = pool_get(&plimit_pool, PR_WAITOK); 541 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 542 newlim->pl_flags = 0; 543 newlim->pl_refcnt = 1; 544 newlim->pl_sv_limit = NULL; 545 546 mutex_enter(&lim->pl_lock); 547 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 548 sizeof(struct rlimit) * RLIM_NLIMITS); 549 550 alen = 0; 551 corename = NULL; 552 for (;;) { 553 if (lim->pl_corename == defcorename) { 554 newlim->pl_corename = defcorename; 555 break; 556 } 557 len = strlen(lim->pl_corename) + 1; 558 if (len <= alen) { 559 newlim->pl_corename = corename; 560 memcpy(corename, lim->pl_corename, len); 561 corename = NULL; 562 break; 563 } 564 mutex_exit(&lim->pl_lock); 565 if (corename != NULL) 566 free(corename, M_TEMP); 567 alen = len; 568 corename = malloc(alen, M_TEMP, M_WAITOK); 569 mutex_enter(&lim->pl_lock); 570 } 571 mutex_exit(&lim->pl_lock); 572 if (corename != NULL) 573 free(corename, M_TEMP); 574 return newlim; 575 } 576 577 void 578 lim_addref(struct plimit *lim) 579 { 580 mutex_enter(&lim->pl_lock); 581 lim->pl_refcnt++; 582 mutex_exit(&lim->pl_lock); 583 } 584 585 /* 586 * Give a process it's own private plimit structure. 587 * This will only be shared (in fork) if modifications are to be shared. 588 */ 589 void 590 lim_privatise(struct proc *p, bool set_shared) 591 { 592 struct plimit *lim, *newlim; 593 594 lim = p->p_limit; 595 if (lim->pl_flags & PL_WRITEABLE) { 596 if (set_shared) 597 lim->pl_flags |= PL_SHAREMOD; 598 return; 599 } 600 601 if (set_shared && lim->pl_flags & PL_SHAREMOD) 602 return; 603 604 newlim = lim_copy(lim); 605 606 mutex_enter(&p->p_mutex); 607 if (p->p_limit->pl_flags & PL_WRITEABLE) { 608 /* Someone crept in while we were busy */ 609 mutex_exit(&p->p_mutex); 610 limfree(newlim); 611 if (set_shared) 612 p->p_limit->pl_flags |= PL_SHAREMOD; 613 return; 614 } 615 616 /* 617 * Since most accesses to p->p_limit aren't locked, we must not 618 * delete the old limit structure yet. 619 */ 620 newlim->pl_sv_limit = p->p_limit; 621 newlim->pl_flags |= PL_WRITEABLE; 622 if (set_shared) 623 newlim->pl_flags |= PL_SHAREMOD; 624 p->p_limit = newlim; 625 mutex_exit(&p->p_mutex); 626 } 627 628 void 629 limfree(struct plimit *lim) 630 { 631 struct plimit *sv_lim; 632 int n; 633 634 do { 635 mutex_enter(&lim->pl_lock); 636 n = --lim->pl_refcnt; 637 mutex_exit(&lim->pl_lock); 638 if (n > 0) 639 return; 640 #ifdef DIAGNOSTIC 641 if (n < 0) 642 panic("limfree"); 643 #endif 644 if (lim->pl_corename != defcorename) 645 free(lim->pl_corename, M_TEMP); 646 sv_lim = lim->pl_sv_limit; 647 mutex_destroy(&lim->pl_lock); 648 pool_put(&plimit_pool, lim); 649 } while ((lim = sv_lim) != NULL); 650 } 651 652 struct pstats * 653 pstatscopy(struct pstats *ps) 654 { 655 656 struct pstats *newps; 657 658 newps = pool_get(&pstats_pool, PR_WAITOK); 659 660 memset(&newps->pstat_startzero, 0, 661 (unsigned) ((char *)&newps->pstat_endzero - 662 (char *)&newps->pstat_startzero)); 663 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 664 ((char *)&newps->pstat_endcopy - 665 (char *)&newps->pstat_startcopy)); 666 667 return (newps); 668 669 } 670 671 void 672 pstatsfree(struct pstats *ps) 673 { 674 675 pool_put(&pstats_pool, ps); 676 } 677 678 /* 679 * sysctl interface in five parts 680 */ 681 682 /* 683 * a routine for sysctl proc subtree helpers that need to pick a valid 684 * process by pid. 685 */ 686 static int 687 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 688 { 689 struct proc *ptmp; 690 int error = 0; 691 692 if (pid == PROC_CURPROC) 693 ptmp = l->l_proc; 694 else if ((ptmp = pfind(pid)) == NULL) 695 error = ESRCH; 696 697 *p2 = ptmp; 698 return (error); 699 } 700 701 /* 702 * sysctl helper routine for setting a process's specific corefile 703 * name. picks the process based on the given pid and checks the 704 * correctness of the new value. 705 */ 706 static int 707 sysctl_proc_corename(SYSCTLFN_ARGS) 708 { 709 struct proc *ptmp; 710 struct plimit *lim; 711 int error = 0, len; 712 char *cname; 713 char *ocore; 714 char *tmp; 715 struct sysctlnode node; 716 717 /* 718 * is this all correct? 719 */ 720 if (namelen != 0) 721 return (EINVAL); 722 if (name[-1] != PROC_PID_CORENAME) 723 return (EINVAL); 724 725 /* 726 * whom are we tweaking? 727 */ 728 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 729 if (error) 730 return (error); 731 732 /* XXX this should be in p_find() */ 733 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 734 ptmp, NULL, NULL, NULL); 735 if (error) 736 return (error); 737 738 /* 739 * let them modify a temporary copy of the core name 740 */ 741 cname = PNBUF_GET(); 742 lim = ptmp->p_limit; 743 mutex_enter(&lim->pl_lock); 744 strlcpy(cname, lim->pl_corename, MAXPATHLEN); 745 mutex_exit(&lim->pl_lock); 746 747 node = *rnode; 748 node.sysctl_data = cname; 749 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 750 751 /* 752 * if that failed, or they have nothing new to say, or we've 753 * heard it before... 754 */ 755 if (error || newp == NULL) 756 goto done; 757 lim = ptmp->p_limit; 758 mutex_enter(&lim->pl_lock); 759 error = strcmp(cname, lim->pl_corename); 760 mutex_exit(&lim->pl_lock); 761 if (error == 0) 762 /* Unchanged */ 763 goto done; 764 765 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 766 ptmp, cname, NULL, NULL); 767 if (error) 768 return (error); 769 770 /* 771 * no error yet and cname now has the new core name in it. 772 * let's see if it looks acceptable. it must be either "core" 773 * or end in ".core" or "/core". 774 */ 775 len = strlen(cname); 776 if (len < 4) { 777 error = EINVAL; 778 } else if (strcmp(cname + len - 4, "core") != 0) { 779 error = EINVAL; 780 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 781 error = EINVAL; 782 } 783 if (error != 0) { 784 goto done; 785 } 786 787 /* 788 * hmm...looks good. now...where do we put it? 789 */ 790 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 791 if (tmp == NULL) { 792 error = ENOMEM; 793 goto done; 794 } 795 memcpy(tmp, cname, len + 1); 796 797 lim_privatise(ptmp, false); 798 lim = ptmp->p_limit; 799 mutex_enter(&lim->pl_lock); 800 ocore = lim->pl_corename; 801 lim->pl_corename = tmp; 802 mutex_exit(&lim->pl_lock); 803 if (ocore != defcorename) 804 free(ocore, M_TEMP); 805 806 done: 807 PNBUF_PUT(cname); 808 return error; 809 } 810 811 /* 812 * sysctl helper routine for checking/setting a process's stop flags, 813 * one for fork and one for exec. 814 */ 815 static int 816 sysctl_proc_stop(SYSCTLFN_ARGS) 817 { 818 struct proc *ptmp; 819 int i, f, error = 0; 820 struct sysctlnode node; 821 822 if (namelen != 0) 823 return (EINVAL); 824 825 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 826 if (error) 827 return (error); 828 829 /* XXX this should be in p_find() */ 830 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 831 ptmp, NULL, NULL, NULL); 832 if (error) 833 return (error); 834 835 switch (rnode->sysctl_num) { 836 case PROC_PID_STOPFORK: 837 f = PS_STOPFORK; 838 break; 839 case PROC_PID_STOPEXEC: 840 f = PS_STOPEXEC; 841 break; 842 case PROC_PID_STOPEXIT: 843 f = PS_STOPEXIT; 844 break; 845 default: 846 return (EINVAL); 847 } 848 849 i = (ptmp->p_flag & f) ? 1 : 0; 850 node = *rnode; 851 node.sysctl_data = &i; 852 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 853 if (error || newp == NULL) 854 return (error); 855 856 mutex_enter(&ptmp->p_smutex); 857 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 858 ptmp, KAUTH_ARG(f), NULL, NULL); 859 if (error) 860 return (error); 861 if (i) 862 ptmp->p_sflag |= f; 863 else 864 ptmp->p_sflag &= ~f; 865 mutex_exit(&ptmp->p_smutex); 866 867 return (0); 868 } 869 870 /* 871 * sysctl helper routine for a process's rlimits as exposed by sysctl. 872 */ 873 static int 874 sysctl_proc_plimit(SYSCTLFN_ARGS) 875 { 876 struct proc *ptmp; 877 u_int limitno; 878 int which, error = 0; 879 struct rlimit alim; 880 struct sysctlnode node; 881 882 if (namelen != 0) 883 return (EINVAL); 884 885 which = name[-1]; 886 if (which != PROC_PID_LIMIT_TYPE_SOFT && 887 which != PROC_PID_LIMIT_TYPE_HARD) 888 return (EINVAL); 889 890 limitno = name[-2] - 1; 891 if (limitno >= RLIM_NLIMITS) 892 return (EINVAL); 893 894 if (name[-3] != PROC_PID_LIMIT) 895 return (EINVAL); 896 897 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 898 if (error) 899 return (error); 900 901 /* XXX this should be in p_find() */ 902 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 903 ptmp, NULL, NULL, NULL); 904 if (error) 905 return (error); 906 907 node = *rnode; 908 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 909 if (which == PROC_PID_LIMIT_TYPE_HARD) 910 node.sysctl_data = &alim.rlim_max; 911 else 912 node.sysctl_data = &alim.rlim_cur; 913 914 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 915 if (error || newp == NULL) 916 return (error); 917 918 return (dosetrlimit(l, ptmp, limitno, &alim)); 919 } 920 921 /* 922 * and finally, the actually glue that sticks it to the tree 923 */ 924 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup") 925 { 926 927 sysctl_createv(clog, 0, NULL, NULL, 928 CTLFLAG_PERMANENT, 929 CTLTYPE_NODE, "proc", NULL, 930 NULL, 0, NULL, 0, 931 CTL_PROC, CTL_EOL); 932 sysctl_createv(clog, 0, NULL, NULL, 933 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 934 CTLTYPE_NODE, "curproc", 935 SYSCTL_DESCR("Per-process settings"), 936 NULL, 0, NULL, 0, 937 CTL_PROC, PROC_CURPROC, CTL_EOL); 938 939 sysctl_createv(clog, 0, NULL, NULL, 940 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 941 CTLTYPE_STRING, "corename", 942 SYSCTL_DESCR("Core file name"), 943 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 944 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 945 sysctl_createv(clog, 0, NULL, NULL, 946 CTLFLAG_PERMANENT, 947 CTLTYPE_NODE, "rlimit", 948 SYSCTL_DESCR("Process limits"), 949 NULL, 0, NULL, 0, 950 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 951 952 #define create_proc_plimit(s, n) do { \ 953 sysctl_createv(clog, 0, NULL, NULL, \ 954 CTLFLAG_PERMANENT, \ 955 CTLTYPE_NODE, s, \ 956 SYSCTL_DESCR("Process " s " limits"), \ 957 NULL, 0, NULL, 0, \ 958 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 959 CTL_EOL); \ 960 sysctl_createv(clog, 0, NULL, NULL, \ 961 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 962 CTLTYPE_QUAD, "soft", \ 963 SYSCTL_DESCR("Process soft " s " limit"), \ 964 sysctl_proc_plimit, 0, NULL, 0, \ 965 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 966 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 967 sysctl_createv(clog, 0, NULL, NULL, \ 968 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 969 CTLTYPE_QUAD, "hard", \ 970 SYSCTL_DESCR("Process hard " s " limit"), \ 971 sysctl_proc_plimit, 0, NULL, 0, \ 972 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 973 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 974 } while (0/*CONSTCOND*/) 975 976 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 977 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 978 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 979 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 980 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 981 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 982 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 983 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 984 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 985 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 986 987 #undef create_proc_plimit 988 989 sysctl_createv(clog, 0, NULL, NULL, 990 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 991 CTLTYPE_INT, "stopfork", 992 SYSCTL_DESCR("Stop process at fork(2)"), 993 sysctl_proc_stop, 0, NULL, 0, 994 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 995 sysctl_createv(clog, 0, NULL, NULL, 996 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 997 CTLTYPE_INT, "stopexec", 998 SYSCTL_DESCR("Stop process at execve(2)"), 999 sysctl_proc_stop, 0, NULL, 0, 1000 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 1001 sysctl_createv(clog, 0, NULL, NULL, 1002 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1003 CTLTYPE_INT, "stopexit", 1004 SYSCTL_DESCR("Stop process before completing exit"), 1005 sysctl_proc_stop, 0, NULL, 0, 1006 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1007 } 1008 1009 void 1010 uid_init(void) 1011 { 1012 1013 /* 1014 * XXXSMP This could be at IPL_SOFTNET, but for now we want 1015 * to to be deadlock free, so it must be at IPL_VM. 1016 */ 1017 mutex_init(&uihashtbl_lock, MUTEX_DRIVER, IPL_VM); 1018 1019 /* 1020 * Ensure that uid 0 is always in the user hash table, as 1021 * sbreserve() expects it available from interrupt context. 1022 */ 1023 (void)uid_find(0); 1024 } 1025 1026 struct uidinfo * 1027 uid_find(uid_t uid) 1028 { 1029 struct uidinfo *uip; 1030 struct uidinfo *newuip = NULL; 1031 struct uihashhead *uipp; 1032 1033 uipp = UIHASH(uid); 1034 1035 again: 1036 mutex_enter(&uihashtbl_lock); 1037 LIST_FOREACH(uip, uipp, ui_hash) 1038 if (uip->ui_uid == uid) { 1039 mutex_exit(&uihashtbl_lock); 1040 if (newuip) { 1041 mutex_destroy(&newuip->ui_lock); 1042 free(newuip, M_PROC); 1043 } 1044 return uip; 1045 } 1046 if (newuip == NULL) { 1047 mutex_exit(&uihashtbl_lock); 1048 /* Must not be called from interrupt context. */ 1049 newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO); 1050 mutex_init(&newuip->ui_lock, MUTEX_DRIVER, IPL_SOFTNET); 1051 goto again; 1052 } 1053 uip = newuip; 1054 1055 LIST_INSERT_HEAD(uipp, uip, ui_hash); 1056 uip->ui_uid = uid; 1057 mutex_exit(&uihashtbl_lock); 1058 1059 return uip; 1060 } 1061 1062 /* 1063 * Change the count associated with number of processes 1064 * a given user is using. 1065 */ 1066 int 1067 chgproccnt(uid_t uid, int diff) 1068 { 1069 struct uidinfo *uip; 1070 1071 if (diff == 0) 1072 return 0; 1073 1074 uip = uid_find(uid); 1075 mutex_enter(&uip->ui_lock); 1076 uip->ui_proccnt += diff; 1077 KASSERT(uip->ui_proccnt >= 0); 1078 mutex_exit(&uip->ui_lock); 1079 return uip->ui_proccnt; 1080 } 1081 1082 int 1083 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax) 1084 { 1085 rlim_t nsb; 1086 1087 mutex_enter(&uip->ui_lock); 1088 nsb = uip->ui_sbsize + to - *hiwat; 1089 if (to > *hiwat && nsb > xmax) { 1090 mutex_exit(&uip->ui_lock); 1091 return 0; 1092 } 1093 *hiwat = to; 1094 uip->ui_sbsize = nsb; 1095 KASSERT(uip->ui_sbsize >= 0); 1096 mutex_exit(&uip->ui_lock); 1097 return 1; 1098 } 1099
Indexes created Sun Sep 21 20:09:37 GMT 2025