kern_resource.c revision 1.137.2.2
1 /* $NetBSD: kern_resource.c,v 1.137.2.2 2008/04/03 13:05:14 christos 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 1.137.2.2 2008/04/03 13:05:14 christos 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/kmem.h> 49 #include <sys/namei.h> 50 #include <sys/pool.h> 51 #include <sys/proc.h> 52 #include <sys/sysctl.h> 53 #include <sys/timevar.h> 54 #include <sys/kauth.h> 55 #include <sys/atomic.h> 56 #include <sys/mount.h> 57 #include <sys/syscallargs.h> 58 #include <sys/atomic.h> 59 60 #include <uvm/uvm_extern.h> 61 62 /* 63 * Maximum process data and stack limits. 64 * They are variables so they are patchable. 65 */ 66 rlim_t maxdmap = MAXDSIZ; 67 rlim_t maxsmap = MAXSSIZ; 68 69 static SLIST_HEAD(uihashhead, uidinfo) *uihashtbl; 70 static u_long uihash; 71 72 #define UIHASH(uid) (&uihashtbl[(uid) & uihash]) 73 74 static pool_cache_t plimit_cache; 75 static pool_cache_t pstats_cache; 76 77 void 78 resource_init(void) 79 { 80 /* 81 * In case of MP system, SLIST_FOREACH would force a cache line 82 * write-back for every modified 'uidinfo', thus we try to keep the 83 * lists short. 84 */ 85 const u_int uihash_sz = (maxproc > 1 ? 1024 : 64); 86 87 plimit_cache = pool_cache_init(sizeof(struct plimit), 0, 0, 0, 88 "plimitpl", NULL, IPL_NONE, NULL, NULL, NULL); 89 pstats_cache = pool_cache_init(sizeof(struct pstats), 0, 0, 0, 90 "pstatspl", NULL, IPL_NONE, NULL, NULL, NULL); 91 uihashtbl = hashinit(uihash_sz, HASH_SLIST, M_PROC, M_WAITOK, &uihash); 92 } 93 94 /* 95 * Resource controls and accounting. 96 */ 97 98 int 99 sys_getpriority(struct lwp *l, const struct sys_getpriority_args *uap, 100 register_t *retval) 101 { 102 /* { 103 syscallarg(int) which; 104 syscallarg(id_t) who; 105 } */ 106 struct proc *curp = l->l_proc, *p; 107 int low = NZERO + PRIO_MAX + 1; 108 int who = SCARG(uap, who); 109 110 mutex_enter(&proclist_lock); 111 switch (SCARG(uap, which)) { 112 case PRIO_PROCESS: 113 if (who == 0) 114 p = curp; 115 else 116 p = p_find(who, PFIND_LOCKED); 117 if (p != NULL) 118 low = p->p_nice; 119 break; 120 121 case PRIO_PGRP: { 122 struct pgrp *pg; 123 124 if (who == 0) 125 pg = curp->p_pgrp; 126 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 127 break; 128 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 129 if (p->p_nice < low) 130 low = p->p_nice; 131 } 132 break; 133 } 134 135 case PRIO_USER: 136 if (who == 0) 137 who = (int)kauth_cred_geteuid(l->l_cred); 138 PROCLIST_FOREACH(p, &allproc) { 139 mutex_enter(&p->p_mutex); 140 if (kauth_cred_geteuid(p->p_cred) == 141 (uid_t)who && p->p_nice < low) 142 low = p->p_nice; 143 mutex_exit(&p->p_mutex); 144 } 145 break; 146 147 default: 148 mutex_exit(&proclist_lock); 149 return (EINVAL); 150 } 151 mutex_exit(&proclist_lock); 152 153 if (low == NZERO + PRIO_MAX + 1) 154 return (ESRCH); 155 *retval = low - NZERO; 156 return (0); 157 } 158 159 /* ARGSUSED */ 160 int 161 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap, 162 register_t *retval) 163 { 164 /* { 165 syscallarg(int) which; 166 syscallarg(id_t) who; 167 syscallarg(int) prio; 168 } */ 169 struct proc *curp = l->l_proc, *p; 170 int found = 0, error = 0; 171 int who = SCARG(uap, who); 172 173 mutex_enter(&proclist_lock); 174 switch (SCARG(uap, which)) { 175 case PRIO_PROCESS: 176 if (who == 0) 177 p = curp; 178 else 179 p = p_find(who, PFIND_LOCKED); 180 if (p != 0) { 181 mutex_enter(&p->p_mutex); 182 error = donice(l, p, SCARG(uap, prio)); 183 mutex_exit(&p->p_mutex); 184 } 185 found++; 186 break; 187 188 case PRIO_PGRP: { 189 struct pgrp *pg; 190 191 if (who == 0) 192 pg = curp->p_pgrp; 193 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 194 break; 195 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 196 mutex_enter(&p->p_mutex); 197 error = donice(l, p, SCARG(uap, prio)); 198 mutex_exit(&p->p_mutex); 199 found++; 200 } 201 break; 202 } 203 204 case PRIO_USER: 205 if (who == 0) 206 who = (int)kauth_cred_geteuid(l->l_cred); 207 PROCLIST_FOREACH(p, &allproc) { 208 mutex_enter(&p->p_mutex); 209 if (kauth_cred_geteuid(p->p_cred) == 210 (uid_t)SCARG(uap, who)) { 211 error = donice(l, p, SCARG(uap, prio)); 212 found++; 213 } 214 mutex_exit(&p->p_mutex); 215 } 216 break; 217 218 default: 219 error = EINVAL; 220 break; 221 } 222 mutex_exit(&proclist_lock); 223 if (found == 0) 224 return (ESRCH); 225 return (error); 226 } 227 228 /* 229 * Renice a process. 230 * 231 * Call with the target process' credentials locked. 232 */ 233 int 234 donice(struct lwp *l, struct proc *chgp, int n) 235 { 236 kauth_cred_t cred = l->l_cred; 237 int onice; 238 239 KASSERT(mutex_owned(&chgp->p_mutex)); 240 241 if (n > PRIO_MAX) 242 n = PRIO_MAX; 243 if (n < PRIO_MIN) 244 n = PRIO_MIN; 245 n += NZERO; 246 onice = chgp->p_nice; 247 onice = chgp->p_nice; 248 249 again: 250 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 251 KAUTH_ARG(n), NULL, NULL)) 252 return (EACCES); 253 mutex_spin_enter(&chgp->p_smutex); 254 if (onice != chgp->p_nice) { 255 mutex_spin_exit(&chgp->p_smutex); 256 goto again; 257 } 258 sched_nice(chgp, n); 259 mutex_spin_exit(&chgp->p_smutex); 260 return (0); 261 } 262 263 /* ARGSUSED */ 264 int 265 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap, 266 register_t *retval) 267 { 268 /* { 269 syscallarg(int) which; 270 syscallarg(const struct rlimit *) rlp; 271 } */ 272 int which = SCARG(uap, which); 273 struct rlimit alim; 274 int error; 275 276 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 277 if (error) 278 return (error); 279 return (dosetrlimit(l, l->l_proc, which, &alim)); 280 } 281 282 int 283 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 284 { 285 struct rlimit *alimp; 286 int error; 287 288 if ((u_int)which >= RLIM_NLIMITS) 289 return (EINVAL); 290 291 if (limp->rlim_cur > limp->rlim_max) { 292 /* 293 * This is programming error. According to SUSv2, we should 294 * return error in this case. 295 */ 296 return (EINVAL); 297 } 298 299 alimp = &p->p_rlimit[which]; 300 /* if we don't change the value, no need to limcopy() */ 301 if (limp->rlim_cur == alimp->rlim_cur && 302 limp->rlim_max == alimp->rlim_max) 303 return 0; 304 305 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 306 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_SET), limp, KAUTH_ARG(which)); 307 if (error) 308 return (error); 309 310 lim_privatise(p, false); 311 /* p->p_limit is now unchangeable */ 312 alimp = &p->p_rlimit[which]; 313 314 switch (which) { 315 316 case RLIMIT_DATA: 317 if (limp->rlim_cur > maxdmap) 318 limp->rlim_cur = maxdmap; 319 if (limp->rlim_max > maxdmap) 320 limp->rlim_max = maxdmap; 321 break; 322 323 case RLIMIT_STACK: 324 if (limp->rlim_cur > maxsmap) 325 limp->rlim_cur = maxsmap; 326 if (limp->rlim_max > maxsmap) 327 limp->rlim_max = maxsmap; 328 329 /* 330 * Return EINVAL if the new stack size limit is lower than 331 * current usage. Otherwise, the process would get SIGSEGV the 332 * moment it would try to access anything on it's current stack. 333 * This conforms to SUSv2. 334 */ 335 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 336 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 337 return (EINVAL); 338 } 339 340 /* 341 * Stack is allocated to the max at exec time with 342 * only "rlim_cur" bytes accessible (In other words, 343 * allocates stack dividing two contiguous regions at 344 * "rlim_cur" bytes boundary). 345 * 346 * Since allocation is done in terms of page, roundup 347 * "rlim_cur" (otherwise, contiguous regions 348 * overlap). If stack limit is going up make more 349 * accessible, if going down make inaccessible. 350 */ 351 limp->rlim_cur = round_page(limp->rlim_cur); 352 if (limp->rlim_cur != alimp->rlim_cur) { 353 vaddr_t addr; 354 vsize_t size; 355 vm_prot_t prot; 356 357 if (limp->rlim_cur > alimp->rlim_cur) { 358 prot = VM_PROT_READ | VM_PROT_WRITE; 359 size = limp->rlim_cur - alimp->rlim_cur; 360 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 361 limp->rlim_cur; 362 } else { 363 prot = VM_PROT_NONE; 364 size = alimp->rlim_cur - limp->rlim_cur; 365 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 366 alimp->rlim_cur; 367 } 368 (void) uvm_map_protect(&p->p_vmspace->vm_map, 369 addr, addr+size, prot, false); 370 } 371 break; 372 373 case RLIMIT_NOFILE: 374 if (limp->rlim_cur > maxfiles) 375 limp->rlim_cur = maxfiles; 376 if (limp->rlim_max > maxfiles) 377 limp->rlim_max = maxfiles; 378 break; 379 380 case RLIMIT_NPROC: 381 if (limp->rlim_cur > maxproc) 382 limp->rlim_cur = maxproc; 383 if (limp->rlim_max > maxproc) 384 limp->rlim_max = maxproc; 385 break; 386 } 387 388 mutex_enter(&p->p_limit->pl_lock); 389 *alimp = *limp; 390 mutex_exit(&p->p_limit->pl_lock); 391 return (0); 392 } 393 394 /* ARGSUSED */ 395 int 396 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, 397 register_t *retval) 398 { 399 /* { 400 syscallarg(int) which; 401 syscallarg(struct rlimit *) rlp; 402 } */ 403 struct proc *p = l->l_proc; 404 int which = SCARG(uap, which); 405 struct rlimit rl; 406 407 if ((u_int)which >= RLIM_NLIMITS) 408 return (EINVAL); 409 410 mutex_enter(&p->p_mutex); 411 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 412 mutex_exit(&p->p_mutex); 413 414 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 415 } 416 417 /* 418 * Transform the running time and tick information in proc p into user, 419 * system, and interrupt time usage. 420 * 421 * Should be called with p->p_smutex held unless called from exit1(). 422 */ 423 void 424 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 425 struct timeval *ip, struct timeval *rp) 426 { 427 uint64_t u, st, ut, it, tot; 428 struct lwp *l; 429 struct bintime tm; 430 struct timeval tv; 431 432 mutex_spin_enter(&p->p_stmutex); 433 st = p->p_sticks; 434 ut = p->p_uticks; 435 it = p->p_iticks; 436 mutex_spin_exit(&p->p_stmutex); 437 438 tm = p->p_rtime; 439 440 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 441 lwp_lock(l); 442 bintime_add(&tm, &l->l_rtime); 443 if ((l->l_flag & LW_RUNNING) != 0) { 444 struct bintime diff; 445 /* 446 * Adjust for the current time slice. This is 447 * actually fairly important since the error 448 * here is on the order of a time quantum, 449 * which is much greater than the sampling 450 * error. 451 */ 452 binuptime(&diff); 453 bintime_sub(&diff, &l->l_stime); 454 bintime_add(&tm, &diff); 455 } 456 lwp_unlock(l); 457 } 458 459 tot = st + ut + it; 460 bintime2timeval(&tm, &tv); 461 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 462 463 if (tot == 0) { 464 /* No ticks, so can't use to share time out, split 50-50 */ 465 st = ut = u / 2; 466 } else { 467 st = (u * st) / tot; 468 ut = (u * ut) / tot; 469 } 470 if (sp != NULL) { 471 sp->tv_sec = st / 1000000; 472 sp->tv_usec = st % 1000000; 473 } 474 if (up != NULL) { 475 up->tv_sec = ut / 1000000; 476 up->tv_usec = ut % 1000000; 477 } 478 if (ip != NULL) { 479 if (it != 0) 480 it = (u * it) / tot; 481 ip->tv_sec = it / 1000000; 482 ip->tv_usec = it % 1000000; 483 } 484 if (rp != NULL) { 485 *rp = tv; 486 } 487 } 488 489 /* ARGSUSED */ 490 int 491 sys___getrusage50(struct lwp *l, const struct sys___getrusage50_args *uap, 492 register_t *retval) 493 { 494 /* { 495 syscallarg(int) who; 496 syscallarg(struct rusage *) rusage; 497 } */ 498 struct rusage ru; 499 struct proc *p = l->l_proc; 500 501 switch (SCARG(uap, who)) { 502 case RUSAGE_SELF: 503 mutex_enter(&p->p_smutex); 504 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 505 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 506 rulwps(p, &ru); 507 mutex_exit(&p->p_smutex); 508 break; 509 510 case RUSAGE_CHILDREN: 511 mutex_enter(&p->p_smutex); 512 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 513 mutex_exit(&p->p_smutex); 514 break; 515 516 default: 517 return EINVAL; 518 } 519 520 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 521 } 522 523 void 524 ruadd(struct rusage *ru, struct rusage *ru2) 525 { 526 long *ip, *ip2; 527 int i; 528 529 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 530 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 531 if (ru->ru_maxrss < ru2->ru_maxrss) 532 ru->ru_maxrss = ru2->ru_maxrss; 533 ip = &ru->ru_first; ip2 = &ru2->ru_first; 534 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 535 *ip++ += *ip2++; 536 } 537 538 void 539 rulwps(proc_t *p, struct rusage *ru) 540 { 541 lwp_t *l; 542 543 KASSERT(mutex_owned(&p->p_smutex)); 544 545 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 546 ruadd(ru, &l->l_ru); 547 ru->ru_nvcsw += (l->l_ncsw - l->l_nivcsw); 548 ru->ru_nivcsw += l->l_nivcsw; 549 } 550 } 551 552 /* 553 * Make a copy of the plimit structure. 554 * We share these structures copy-on-write after fork, 555 * and copy when a limit is changed. 556 * 557 * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure 558 * we are copying to change beneath our feet! 559 */ 560 struct plimit * 561 lim_copy(struct plimit *lim) 562 { 563 struct plimit *newlim; 564 char *corename; 565 size_t alen, len; 566 567 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 568 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 569 newlim->pl_flags = 0; 570 newlim->pl_refcnt = 1; 571 newlim->pl_sv_limit = NULL; 572 573 mutex_enter(&lim->pl_lock); 574 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 575 sizeof(struct rlimit) * RLIM_NLIMITS); 576 577 alen = 0; 578 corename = NULL; 579 for (;;) { 580 if (lim->pl_corename == defcorename) { 581 newlim->pl_corename = defcorename; 582 break; 583 } 584 len = strlen(lim->pl_corename) + 1; 585 if (len <= alen) { 586 newlim->pl_corename = corename; 587 memcpy(corename, lim->pl_corename, len); 588 corename = NULL; 589 break; 590 } 591 mutex_exit(&lim->pl_lock); 592 if (corename != NULL) 593 free(corename, M_TEMP); 594 alen = len; 595 corename = malloc(alen, M_TEMP, M_WAITOK); 596 mutex_enter(&lim->pl_lock); 597 } 598 mutex_exit(&lim->pl_lock); 599 if (corename != NULL) 600 free(corename, M_TEMP); 601 return newlim; 602 } 603 604 void 605 lim_addref(struct plimit *lim) 606 { 607 atomic_inc_uint(&lim->pl_refcnt); 608 } 609 610 /* 611 * Give a process it's own private plimit structure. 612 * This will only be shared (in fork) if modifications are to be shared. 613 */ 614 void 615 lim_privatise(struct proc *p, bool set_shared) 616 { 617 struct plimit *lim, *newlim; 618 619 lim = p->p_limit; 620 if (lim->pl_flags & PL_WRITEABLE) { 621 if (set_shared) 622 lim->pl_flags |= PL_SHAREMOD; 623 return; 624 } 625 626 if (set_shared && lim->pl_flags & PL_SHAREMOD) 627 return; 628 629 newlim = lim_copy(lim); 630 631 mutex_enter(&p->p_mutex); 632 if (p->p_limit->pl_flags & PL_WRITEABLE) { 633 /* Someone crept in while we were busy */ 634 mutex_exit(&p->p_mutex); 635 limfree(newlim); 636 if (set_shared) 637 p->p_limit->pl_flags |= PL_SHAREMOD; 638 return; 639 } 640 641 /* 642 * Since most accesses to p->p_limit aren't locked, we must not 643 * delete the old limit structure yet. 644 */ 645 newlim->pl_sv_limit = p->p_limit; 646 newlim->pl_flags |= PL_WRITEABLE; 647 if (set_shared) 648 newlim->pl_flags |= PL_SHAREMOD; 649 p->p_limit = newlim; 650 mutex_exit(&p->p_mutex); 651 } 652 653 void 654 limfree(struct plimit *lim) 655 { 656 struct plimit *sv_lim; 657 658 do { 659 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) 660 return; 661 if (lim->pl_corename != defcorename) 662 free(lim->pl_corename, M_TEMP); 663 sv_lim = lim->pl_sv_limit; 664 mutex_destroy(&lim->pl_lock); 665 pool_cache_put(plimit_cache, lim); 666 } while ((lim = sv_lim) != NULL); 667 } 668 669 struct pstats * 670 pstatscopy(struct pstats *ps) 671 { 672 673 struct pstats *newps; 674 675 newps = pool_cache_get(pstats_cache, PR_WAITOK); 676 677 memset(&newps->pstat_startzero, 0, 678 (unsigned) ((char *)&newps->pstat_endzero - 679 (char *)&newps->pstat_startzero)); 680 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 681 ((char *)&newps->pstat_endcopy - 682 (char *)&newps->pstat_startcopy)); 683 684 return (newps); 685 686 } 687 688 void 689 pstatsfree(struct pstats *ps) 690 { 691 692 pool_cache_put(pstats_cache, ps); 693 } 694 695 /* 696 * sysctl interface in five parts 697 */ 698 699 /* 700 * a routine for sysctl proc subtree helpers that need to pick a valid 701 * process by pid. 702 */ 703 static int 704 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 705 { 706 struct proc *ptmp; 707 int error = 0; 708 709 if (pid == PROC_CURPROC) 710 ptmp = l->l_proc; 711 else if ((ptmp = pfind(pid)) == NULL) 712 error = ESRCH; 713 714 *p2 = ptmp; 715 return (error); 716 } 717 718 /* 719 * sysctl helper routine for setting a process's specific corefile 720 * name. picks the process based on the given pid and checks the 721 * correctness of the new value. 722 */ 723 static int 724 sysctl_proc_corename(SYSCTLFN_ARGS) 725 { 726 struct proc *ptmp; 727 struct plimit *lim; 728 int error = 0, len; 729 char *cname; 730 char *ocore; 731 char *tmp; 732 struct sysctlnode node; 733 734 /* 735 * is this all correct? 736 */ 737 if (namelen != 0) 738 return (EINVAL); 739 if (name[-1] != PROC_PID_CORENAME) 740 return (EINVAL); 741 742 /* 743 * whom are we tweaking? 744 */ 745 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 746 if (error) 747 return (error); 748 749 /* XXX-elad */ 750 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 751 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 752 if (error) 753 return (error); 754 755 if (newp == NULL) { 756 error = kauth_authorize_process(l->l_cred, 757 KAUTH_PROCESS_CORENAME, ptmp, 758 KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_GET), NULL, NULL); 759 if (error) 760 return (error); 761 } 762 763 /* 764 * let them modify a temporary copy of the core name 765 */ 766 cname = PNBUF_GET(); 767 lim = ptmp->p_limit; 768 mutex_enter(&lim->pl_lock); 769 strlcpy(cname, lim->pl_corename, MAXPATHLEN); 770 mutex_exit(&lim->pl_lock); 771 772 node = *rnode; 773 node.sysctl_data = cname; 774 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 775 776 /* 777 * if that failed, or they have nothing new to say, or we've 778 * heard it before... 779 */ 780 if (error || newp == NULL) 781 goto done; 782 lim = ptmp->p_limit; 783 mutex_enter(&lim->pl_lock); 784 error = strcmp(cname, lim->pl_corename); 785 mutex_exit(&lim->pl_lock); 786 if (error == 0) 787 /* Unchanged */ 788 goto done; 789 790 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 791 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_SET), cname, NULL); 792 if (error) 793 return (error); 794 795 /* 796 * no error yet and cname now has the new core name in it. 797 * let's see if it looks acceptable. it must be either "core" 798 * or end in ".core" or "/core". 799 */ 800 len = strlen(cname); 801 if (len < 4) { 802 error = EINVAL; 803 } else if (strcmp(cname + len - 4, "core") != 0) { 804 error = EINVAL; 805 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 806 error = EINVAL; 807 } 808 if (error != 0) { 809 goto done; 810 } 811 812 /* 813 * hmm...looks good. now...where do we put it? 814 */ 815 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 816 if (tmp == NULL) { 817 error = ENOMEM; 818 goto done; 819 } 820 memcpy(tmp, cname, len + 1); 821 822 lim_privatise(ptmp, false); 823 lim = ptmp->p_limit; 824 mutex_enter(&lim->pl_lock); 825 ocore = lim->pl_corename; 826 lim->pl_corename = tmp; 827 mutex_exit(&lim->pl_lock); 828 if (ocore != defcorename) 829 free(ocore, M_TEMP); 830 831 done: 832 PNBUF_PUT(cname); 833 return error; 834 } 835 836 /* 837 * sysctl helper routine for checking/setting a process's stop flags, 838 * one for fork and one for exec. 839 */ 840 static int 841 sysctl_proc_stop(SYSCTLFN_ARGS) 842 { 843 struct proc *ptmp; 844 int i, f, error = 0; 845 struct sysctlnode node; 846 847 if (namelen != 0) 848 return (EINVAL); 849 850 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 851 if (error) 852 return (error); 853 854 /* XXX-elad */ 855 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 856 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 857 if (error) 858 return (error); 859 860 switch (rnode->sysctl_num) { 861 case PROC_PID_STOPFORK: 862 f = PS_STOPFORK; 863 break; 864 case PROC_PID_STOPEXEC: 865 f = PS_STOPEXEC; 866 break; 867 case PROC_PID_STOPEXIT: 868 f = PS_STOPEXIT; 869 break; 870 default: 871 return (EINVAL); 872 } 873 874 i = (ptmp->p_flag & f) ? 1 : 0; 875 node = *rnode; 876 node.sysctl_data = &i; 877 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 878 if (error || newp == NULL) 879 return (error); 880 881 mutex_enter(&ptmp->p_smutex); 882 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 883 ptmp, KAUTH_ARG(f), NULL, NULL); 884 if (error) 885 return (error); 886 if (i) 887 ptmp->p_sflag |= f; 888 else 889 ptmp->p_sflag &= ~f; 890 mutex_exit(&ptmp->p_smutex); 891 892 return (0); 893 } 894 895 /* 896 * sysctl helper routine for a process's rlimits as exposed by sysctl. 897 */ 898 static int 899 sysctl_proc_plimit(SYSCTLFN_ARGS) 900 { 901 struct proc *ptmp; 902 u_int limitno; 903 int which, error = 0; 904 struct rlimit alim; 905 struct sysctlnode node; 906 907 if (namelen != 0) 908 return (EINVAL); 909 910 which = name[-1]; 911 if (which != PROC_PID_LIMIT_TYPE_SOFT && 912 which != PROC_PID_LIMIT_TYPE_HARD) 913 return (EINVAL); 914 915 limitno = name[-2] - 1; 916 if (limitno >= RLIM_NLIMITS) 917 return (EINVAL); 918 919 if (name[-3] != PROC_PID_LIMIT) 920 return (EINVAL); 921 922 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 923 if (error) 924 return (error); 925 926 /* XXX-elad */ 927 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 928 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 929 if (error) 930 return (error); 931 932 /* Check if we can view limits. */ 933 if (newp == NULL) { 934 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 935 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_GET), &alim, 936 KAUTH_ARG(which)); 937 if (error) 938 return (error); 939 } 940 941 node = *rnode; 942 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 943 if (which == PROC_PID_LIMIT_TYPE_HARD) 944 node.sysctl_data = &alim.rlim_max; 945 else 946 node.sysctl_data = &alim.rlim_cur; 947 948 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 949 if (error || newp == NULL) 950 return (error); 951 952 return (dosetrlimit(l, ptmp, limitno, &alim)); 953 } 954 955 /* 956 * and finally, the actually glue that sticks it to the tree 957 */ 958 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup") 959 { 960 961 sysctl_createv(clog, 0, NULL, NULL, 962 CTLFLAG_PERMANENT, 963 CTLTYPE_NODE, "proc", NULL, 964 NULL, 0, NULL, 0, 965 CTL_PROC, CTL_EOL); 966 sysctl_createv(clog, 0, NULL, NULL, 967 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 968 CTLTYPE_NODE, "curproc", 969 SYSCTL_DESCR("Per-process settings"), 970 NULL, 0, NULL, 0, 971 CTL_PROC, PROC_CURPROC, CTL_EOL); 972 973 sysctl_createv(clog, 0, NULL, NULL, 974 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 975 CTLTYPE_STRING, "corename", 976 SYSCTL_DESCR("Core file name"), 977 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 978 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 979 sysctl_createv(clog, 0, NULL, NULL, 980 CTLFLAG_PERMANENT, 981 CTLTYPE_NODE, "rlimit", 982 SYSCTL_DESCR("Process limits"), 983 NULL, 0, NULL, 0, 984 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 985 986 #define create_proc_plimit(s, n) do { \ 987 sysctl_createv(clog, 0, NULL, NULL, \ 988 CTLFLAG_PERMANENT, \ 989 CTLTYPE_NODE, s, \ 990 SYSCTL_DESCR("Process " s " limits"), \ 991 NULL, 0, NULL, 0, \ 992 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 993 CTL_EOL); \ 994 sysctl_createv(clog, 0, NULL, NULL, \ 995 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 996 CTLTYPE_QUAD, "soft", \ 997 SYSCTL_DESCR("Process soft " s " limit"), \ 998 sysctl_proc_plimit, 0, NULL, 0, \ 999 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1000 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 1001 sysctl_createv(clog, 0, NULL, NULL, \ 1002 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1003 CTLTYPE_QUAD, "hard", \ 1004 SYSCTL_DESCR("Process hard " s " limit"), \ 1005 sysctl_proc_plimit, 0, NULL, 0, \ 1006 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1007 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 1008 } while (0/*CONSTCOND*/) 1009 1010 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 1011 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 1012 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 1013 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 1014 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 1015 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 1016 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 1017 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 1018 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 1019 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 1020 1021 #undef create_proc_plimit 1022 1023 sysctl_createv(clog, 0, NULL, NULL, 1024 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1025 CTLTYPE_INT, "stopfork", 1026 SYSCTL_DESCR("Stop process at fork(2)"), 1027 sysctl_proc_stop, 0, NULL, 0, 1028 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 1029 sysctl_createv(clog, 0, NULL, NULL, 1030 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1031 CTLTYPE_INT, "stopexec", 1032 SYSCTL_DESCR("Stop process at execve(2)"), 1033 sysctl_proc_stop, 0, NULL, 0, 1034 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 1035 sysctl_createv(clog, 0, NULL, NULL, 1036 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1037 CTLTYPE_INT, "stopexit", 1038 SYSCTL_DESCR("Stop process before completing exit"), 1039 sysctl_proc_stop, 0, NULL, 0, 1040 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1041 } 1042 1043 void 1044 uid_init(void) 1045 { 1046 1047 /* 1048 * Ensure that uid 0 is always in the user hash table, as 1049 * sbreserve() expects it available from interrupt context. 1050 */ 1051 (void)uid_find(0); 1052 } 1053 1054 struct uidinfo * 1055 uid_find(uid_t uid) 1056 { 1057 struct uidinfo *uip, *uip_first, *newuip; 1058 struct uihashhead *uipp; 1059 1060 uipp = UIHASH(uid); 1061 newuip = NULL; 1062 1063 /* 1064 * To make insertion atomic, abstraction of SLIST will be violated. 1065 */ 1066 uip_first = uipp->slh_first; 1067 again: 1068 SLIST_FOREACH(uip, uipp, ui_hash) { 1069 if (uip->ui_uid != uid) 1070 continue; 1071 if (newuip != NULL) 1072 kmem_free(newuip, sizeof(*newuip)); 1073 return uip; 1074 } 1075 if (newuip == NULL) 1076 newuip = kmem_zalloc(sizeof(*newuip), KM_SLEEP); 1077 newuip->ui_uid = uid; 1078 1079 /* 1080 * If atomic insert is unsuccessful, another thread might be 1081 * allocated this 'uid', thus full re-check is needed. 1082 */ 1083 newuip->ui_hash.sle_next = uip_first; 1084 membar_producer(); 1085 uip = atomic_cas_ptr(&uipp->slh_first, uip_first, newuip); 1086 if (uip != uip_first) { 1087 uip_first = uip; 1088 goto again; 1089 } 1090 1091 return newuip; 1092 } 1093 1094 /* 1095 * Change the count associated with number of processes 1096 * a given user is using. 1097 */ 1098 int 1099 chgproccnt(uid_t uid, int diff) 1100 { 1101 struct uidinfo *uip; 1102 long proccnt; 1103 1104 uip = uid_find(uid); 1105 proccnt = atomic_add_long_nv(&uip->ui_proccnt, diff); 1106 KASSERT(proccnt >= 0); 1107 return proccnt; 1108 } 1109 1110 int 1111 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax) 1112 { 1113 rlim_t nsb; 1114 const long diff = to - *hiwat; 1115 1116 nsb = atomic_add_long_nv((long *)&uip->ui_sbsize, diff); 1117 if (diff > 0 && nsb > xmax) { 1118 atomic_add_long((long *)&uip->ui_sbsize, -diff); 1119 return 0; 1120 } 1121 *hiwat = to; 1122 KASSERT(nsb >= 0); 1123 return 1; 1124 } 1125
Indexes created Sun Sep 21 19:09:51 GMT 2025