kern_resource.c revision 1.156
1 /* $NetBSD: kern_resource.c,v 1.156 2010/05/26 23:53:21 pooka 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 1.156 2010/05/26 23:53:21 pooka 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 pool_cache_t plimit_cache; 70 static pool_cache_t pstats_cache; 71 72 static kauth_listener_t resource_listener; 73 74 static void sysctl_proc_setup(void); 75 76 static int 77 resource_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 78 void *arg0, void *arg1, void *arg2, void *arg3) 79 { 80 struct proc *p; 81 int result; 82 83 result = KAUTH_RESULT_DEFER; 84 p = arg0; 85 86 switch (action) { 87 case KAUTH_PROCESS_NICE: 88 if (kauth_cred_geteuid(cred) != kauth_cred_geteuid(p->p_cred) && 89 kauth_cred_getuid(cred) != kauth_cred_geteuid(p->p_cred)) { 90 break; 91 } 92 93 if ((u_long)arg1 >= p->p_nice) 94 result = KAUTH_RESULT_ALLOW; 95 96 break; 97 98 case KAUTH_PROCESS_RLIMIT: { 99 enum kauth_process_req req; 100 101 req = (enum kauth_process_req)(unsigned long)arg1; 102 103 switch (req) { 104 case KAUTH_REQ_PROCESS_RLIMIT_GET: 105 result = KAUTH_RESULT_ALLOW; 106 break; 107 108 case KAUTH_REQ_PROCESS_RLIMIT_SET: { 109 struct rlimit *new_rlimit; 110 u_long which; 111 112 if ((p != curlwp->l_proc) && 113 (proc_uidmatch(cred, p->p_cred) != 0)) 114 break; 115 116 new_rlimit = arg2; 117 which = (u_long)arg3; 118 119 if (new_rlimit->rlim_max <= p->p_rlimit[which].rlim_max) 120 result = KAUTH_RESULT_ALLOW; 121 122 break; 123 } 124 125 default: 126 break; 127 } 128 129 break; 130 } 131 132 default: 133 break; 134 } 135 136 return result; 137 } 138 139 void 140 resource_init(void) 141 { 142 143 plimit_cache = pool_cache_init(sizeof(struct plimit), 0, 0, 0, 144 "plimitpl", NULL, IPL_NONE, NULL, NULL, NULL); 145 pstats_cache = pool_cache_init(sizeof(struct pstats), 0, 0, 0, 146 "pstatspl", NULL, IPL_NONE, NULL, NULL, NULL); 147 148 resource_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 149 resource_listener_cb, NULL); 150 151 sysctl_proc_setup(); 152 } 153 154 /* 155 * Resource controls and accounting. 156 */ 157 158 int 159 sys_getpriority(struct lwp *l, const struct sys_getpriority_args *uap, 160 register_t *retval) 161 { 162 /* { 163 syscallarg(int) which; 164 syscallarg(id_t) who; 165 } */ 166 struct proc *curp = l->l_proc, *p; 167 int low = NZERO + PRIO_MAX + 1; 168 int who = SCARG(uap, who); 169 170 mutex_enter(proc_lock); 171 switch (SCARG(uap, which)) { 172 case PRIO_PROCESS: 173 if (who == 0) 174 p = curp; 175 else 176 p = p_find(who, PFIND_LOCKED); 177 if (p != NULL) 178 low = p->p_nice; 179 break; 180 181 case PRIO_PGRP: { 182 struct pgrp *pg; 183 184 if (who == 0) 185 pg = curp->p_pgrp; 186 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 187 break; 188 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 189 if (p->p_nice < low) 190 low = p->p_nice; 191 } 192 break; 193 } 194 195 case PRIO_USER: 196 if (who == 0) 197 who = (int)kauth_cred_geteuid(l->l_cred); 198 PROCLIST_FOREACH(p, &allproc) { 199 mutex_enter(p->p_lock); 200 if (kauth_cred_geteuid(p->p_cred) == 201 (uid_t)who && p->p_nice < low) 202 low = p->p_nice; 203 mutex_exit(p->p_lock); 204 } 205 break; 206 207 default: 208 mutex_exit(proc_lock); 209 return (EINVAL); 210 } 211 mutex_exit(proc_lock); 212 213 if (low == NZERO + PRIO_MAX + 1) 214 return (ESRCH); 215 *retval = low - NZERO; 216 return (0); 217 } 218 219 /* ARGSUSED */ 220 int 221 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap, 222 register_t *retval) 223 { 224 /* { 225 syscallarg(int) which; 226 syscallarg(id_t) who; 227 syscallarg(int) prio; 228 } */ 229 struct proc *curp = l->l_proc, *p; 230 int found = 0, error = 0; 231 int who = SCARG(uap, who); 232 233 mutex_enter(proc_lock); 234 switch (SCARG(uap, which)) { 235 case PRIO_PROCESS: 236 if (who == 0) 237 p = curp; 238 else 239 p = p_find(who, PFIND_LOCKED); 240 if (p != 0) { 241 mutex_enter(p->p_lock); 242 error = donice(l, p, SCARG(uap, prio)); 243 mutex_exit(p->p_lock); 244 found++; 245 } 246 break; 247 248 case PRIO_PGRP: { 249 struct pgrp *pg; 250 251 if (who == 0) 252 pg = curp->p_pgrp; 253 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL) 254 break; 255 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 256 mutex_enter(p->p_lock); 257 error = donice(l, p, SCARG(uap, prio)); 258 mutex_exit(p->p_lock); 259 found++; 260 } 261 break; 262 } 263 264 case PRIO_USER: 265 if (who == 0) 266 who = (int)kauth_cred_geteuid(l->l_cred); 267 PROCLIST_FOREACH(p, &allproc) { 268 mutex_enter(p->p_lock); 269 if (kauth_cred_geteuid(p->p_cred) == 270 (uid_t)SCARG(uap, who)) { 271 error = donice(l, p, SCARG(uap, prio)); 272 found++; 273 } 274 mutex_exit(p->p_lock); 275 } 276 break; 277 278 default: 279 mutex_exit(proc_lock); 280 return EINVAL; 281 } 282 mutex_exit(proc_lock); 283 if (found == 0) 284 return (ESRCH); 285 return (error); 286 } 287 288 /* 289 * Renice a process. 290 * 291 * Call with the target process' credentials locked. 292 */ 293 int 294 donice(struct lwp *l, struct proc *chgp, int n) 295 { 296 kauth_cred_t cred = l->l_cred; 297 298 KASSERT(mutex_owned(chgp->p_lock)); 299 300 if (kauth_cred_geteuid(cred) && kauth_cred_getuid(cred) && 301 kauth_cred_geteuid(cred) != kauth_cred_geteuid(chgp->p_cred) && 302 kauth_cred_getuid(cred) != kauth_cred_geteuid(chgp->p_cred)) 303 return (EPERM); 304 305 if (n > PRIO_MAX) 306 n = PRIO_MAX; 307 if (n < PRIO_MIN) 308 n = PRIO_MIN; 309 n += NZERO; 310 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 311 KAUTH_ARG(n), NULL, NULL)) 312 return (EACCES); 313 sched_nice(chgp, n); 314 return (0); 315 } 316 317 /* ARGSUSED */ 318 int 319 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap, 320 register_t *retval) 321 { 322 /* { 323 syscallarg(int) which; 324 syscallarg(const struct rlimit *) rlp; 325 } */ 326 int which = SCARG(uap, which); 327 struct rlimit alim; 328 int error; 329 330 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 331 if (error) 332 return (error); 333 return (dosetrlimit(l, l->l_proc, which, &alim)); 334 } 335 336 int 337 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 338 { 339 struct rlimit *alimp; 340 int error; 341 342 if ((u_int)which >= RLIM_NLIMITS) 343 return (EINVAL); 344 345 if (limp->rlim_cur > limp->rlim_max) { 346 /* 347 * This is programming error. According to SUSv2, we should 348 * return error in this case. 349 */ 350 return (EINVAL); 351 } 352 353 alimp = &p->p_rlimit[which]; 354 /* if we don't change the value, no need to limcopy() */ 355 if (limp->rlim_cur == alimp->rlim_cur && 356 limp->rlim_max == alimp->rlim_max) 357 return 0; 358 359 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 360 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_SET), limp, KAUTH_ARG(which)); 361 if (error) 362 return (error); 363 364 lim_privatise(p, false); 365 /* p->p_limit is now unchangeable */ 366 alimp = &p->p_rlimit[which]; 367 368 switch (which) { 369 370 case RLIMIT_DATA: 371 if (limp->rlim_cur > maxdmap) 372 limp->rlim_cur = maxdmap; 373 if (limp->rlim_max > maxdmap) 374 limp->rlim_max = maxdmap; 375 break; 376 377 case RLIMIT_STACK: 378 if (limp->rlim_cur > maxsmap) 379 limp->rlim_cur = maxsmap; 380 if (limp->rlim_max > maxsmap) 381 limp->rlim_max = maxsmap; 382 383 /* 384 * Return EINVAL if the new stack size limit is lower than 385 * current usage. Otherwise, the process would get SIGSEGV the 386 * moment it would try to access anything on it's current stack. 387 * This conforms to SUSv2. 388 */ 389 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE 390 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) { 391 return (EINVAL); 392 } 393 394 /* 395 * Stack is allocated to the max at exec time with 396 * only "rlim_cur" bytes accessible (In other words, 397 * allocates stack dividing two contiguous regions at 398 * "rlim_cur" bytes boundary). 399 * 400 * Since allocation is done in terms of page, roundup 401 * "rlim_cur" (otherwise, contiguous regions 402 * overlap). If stack limit is going up make more 403 * accessible, if going down make inaccessible. 404 */ 405 limp->rlim_cur = round_page(limp->rlim_cur); 406 if (limp->rlim_cur != alimp->rlim_cur) { 407 vaddr_t addr; 408 vsize_t size; 409 vm_prot_t prot; 410 411 if (limp->rlim_cur > alimp->rlim_cur) { 412 prot = VM_PROT_READ | VM_PROT_WRITE; 413 size = limp->rlim_cur - alimp->rlim_cur; 414 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 415 limp->rlim_cur; 416 } else { 417 prot = VM_PROT_NONE; 418 size = alimp->rlim_cur - limp->rlim_cur; 419 addr = (vaddr_t)p->p_vmspace->vm_minsaddr - 420 alimp->rlim_cur; 421 } 422 (void) uvm_map_protect(&p->p_vmspace->vm_map, 423 addr, addr+size, prot, false); 424 } 425 break; 426 427 case RLIMIT_NOFILE: 428 if (limp->rlim_cur > maxfiles) 429 limp->rlim_cur = maxfiles; 430 if (limp->rlim_max > maxfiles) 431 limp->rlim_max = maxfiles; 432 break; 433 434 case RLIMIT_NPROC: 435 if (limp->rlim_cur > maxproc) 436 limp->rlim_cur = maxproc; 437 if (limp->rlim_max > maxproc) 438 limp->rlim_max = maxproc; 439 break; 440 } 441 442 mutex_enter(&p->p_limit->pl_lock); 443 *alimp = *limp; 444 mutex_exit(&p->p_limit->pl_lock); 445 return (0); 446 } 447 448 /* ARGSUSED */ 449 int 450 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, 451 register_t *retval) 452 { 453 /* { 454 syscallarg(int) which; 455 syscallarg(struct rlimit *) rlp; 456 } */ 457 struct proc *p = l->l_proc; 458 int which = SCARG(uap, which); 459 struct rlimit rl; 460 461 if ((u_int)which >= RLIM_NLIMITS) 462 return (EINVAL); 463 464 mutex_enter(p->p_lock); 465 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 466 mutex_exit(p->p_lock); 467 468 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 469 } 470 471 /* 472 * Transform the running time and tick information in proc p into user, 473 * system, and interrupt time usage. 474 * 475 * Should be called with p->p_lock held unless called from exit1(). 476 */ 477 void 478 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 479 struct timeval *ip, struct timeval *rp) 480 { 481 uint64_t u, st, ut, it, tot; 482 struct lwp *l; 483 struct bintime tm; 484 struct timeval tv; 485 486 mutex_spin_enter(&p->p_stmutex); 487 st = p->p_sticks; 488 ut = p->p_uticks; 489 it = p->p_iticks; 490 mutex_spin_exit(&p->p_stmutex); 491 492 tm = p->p_rtime; 493 494 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 495 lwp_lock(l); 496 bintime_add(&tm, &l->l_rtime); 497 if ((l->l_pflag & LP_RUNNING) != 0) { 498 struct bintime diff; 499 /* 500 * Adjust for the current time slice. This is 501 * actually fairly important since the error 502 * here is on the order of a time quantum, 503 * which is much greater than the sampling 504 * error. 505 */ 506 binuptime(&diff); 507 bintime_sub(&diff, &l->l_stime); 508 bintime_add(&tm, &diff); 509 } 510 lwp_unlock(l); 511 } 512 513 tot = st + ut + it; 514 bintime2timeval(&tm, &tv); 515 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 516 517 if (tot == 0) { 518 /* No ticks, so can't use to share time out, split 50-50 */ 519 st = ut = u / 2; 520 } else { 521 st = (u * st) / tot; 522 ut = (u * ut) / tot; 523 } 524 if (sp != NULL) { 525 sp->tv_sec = st / 1000000; 526 sp->tv_usec = st % 1000000; 527 } 528 if (up != NULL) { 529 up->tv_sec = ut / 1000000; 530 up->tv_usec = ut % 1000000; 531 } 532 if (ip != NULL) { 533 if (it != 0) 534 it = (u * it) / tot; 535 ip->tv_sec = it / 1000000; 536 ip->tv_usec = it % 1000000; 537 } 538 if (rp != NULL) { 539 *rp = tv; 540 } 541 } 542 543 /* ARGSUSED */ 544 int 545 sys___getrusage50(struct lwp *l, const struct sys___getrusage50_args *uap, 546 register_t *retval) 547 { 548 /* { 549 syscallarg(int) who; 550 syscallarg(struct rusage *) rusage; 551 } */ 552 struct rusage ru; 553 struct proc *p = l->l_proc; 554 555 switch (SCARG(uap, who)) { 556 case RUSAGE_SELF: 557 mutex_enter(p->p_lock); 558 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 559 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 560 rulwps(p, &ru); 561 mutex_exit(p->p_lock); 562 break; 563 564 case RUSAGE_CHILDREN: 565 mutex_enter(p->p_lock); 566 memcpy(&ru, &p->p_stats->p_cru, sizeof(ru)); 567 mutex_exit(p->p_lock); 568 break; 569 570 default: 571 return EINVAL; 572 } 573 574 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 575 } 576 577 void 578 ruadd(struct rusage *ru, struct rusage *ru2) 579 { 580 long *ip, *ip2; 581 int i; 582 583 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 584 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 585 if (ru->ru_maxrss < ru2->ru_maxrss) 586 ru->ru_maxrss = ru2->ru_maxrss; 587 ip = &ru->ru_first; ip2 = &ru2->ru_first; 588 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 589 *ip++ += *ip2++; 590 } 591 592 void 593 rulwps(proc_t *p, struct rusage *ru) 594 { 595 lwp_t *l; 596 597 KASSERT(mutex_owned(p->p_lock)); 598 599 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 600 ruadd(ru, &l->l_ru); 601 ru->ru_nvcsw += (l->l_ncsw - l->l_nivcsw); 602 ru->ru_nivcsw += l->l_nivcsw; 603 } 604 } 605 606 /* 607 * Make a copy of the plimit structure. 608 * We share these structures copy-on-write after fork, 609 * and copy when a limit is changed. 610 * 611 * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure 612 * we are copying to change beneath our feet! 613 */ 614 struct plimit * 615 lim_copy(struct plimit *lim) 616 { 617 struct plimit *newlim; 618 char *corename; 619 size_t alen, len; 620 621 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 622 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 623 newlim->pl_flags = 0; 624 newlim->pl_refcnt = 1; 625 newlim->pl_sv_limit = NULL; 626 627 mutex_enter(&lim->pl_lock); 628 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 629 sizeof(struct rlimit) * RLIM_NLIMITS); 630 631 alen = 0; 632 corename = NULL; 633 for (;;) { 634 if (lim->pl_corename == defcorename) { 635 newlim->pl_corename = defcorename; 636 break; 637 } 638 len = strlen(lim->pl_corename) + 1; 639 if (len <= alen) { 640 newlim->pl_corename = corename; 641 memcpy(corename, lim->pl_corename, len); 642 corename = NULL; 643 break; 644 } 645 mutex_exit(&lim->pl_lock); 646 if (corename != NULL) 647 free(corename, M_TEMP); 648 alen = len; 649 corename = malloc(alen, M_TEMP, M_WAITOK); 650 mutex_enter(&lim->pl_lock); 651 } 652 mutex_exit(&lim->pl_lock); 653 if (corename != NULL) 654 free(corename, M_TEMP); 655 return newlim; 656 } 657 658 void 659 lim_addref(struct plimit *lim) 660 { 661 atomic_inc_uint(&lim->pl_refcnt); 662 } 663 664 /* 665 * Give a process it's own private plimit structure. 666 * This will only be shared (in fork) if modifications are to be shared. 667 */ 668 void 669 lim_privatise(struct proc *p, bool set_shared) 670 { 671 struct plimit *lim, *newlim; 672 673 lim = p->p_limit; 674 if (lim->pl_flags & PL_WRITEABLE) { 675 if (set_shared) 676 lim->pl_flags |= PL_SHAREMOD; 677 return; 678 } 679 680 if (set_shared && lim->pl_flags & PL_SHAREMOD) 681 return; 682 683 newlim = lim_copy(lim); 684 685 mutex_enter(p->p_lock); 686 if (p->p_limit->pl_flags & PL_WRITEABLE) { 687 /* Someone crept in while we were busy */ 688 mutex_exit(p->p_lock); 689 limfree(newlim); 690 if (set_shared) 691 p->p_limit->pl_flags |= PL_SHAREMOD; 692 return; 693 } 694 695 /* 696 * Since most accesses to p->p_limit aren't locked, we must not 697 * delete the old limit structure yet. 698 */ 699 newlim->pl_sv_limit = p->p_limit; 700 newlim->pl_flags |= PL_WRITEABLE; 701 if (set_shared) 702 newlim->pl_flags |= PL_SHAREMOD; 703 p->p_limit = newlim; 704 mutex_exit(p->p_lock); 705 } 706 707 void 708 limfree(struct plimit *lim) 709 { 710 struct plimit *sv_lim; 711 712 do { 713 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) 714 return; 715 if (lim->pl_corename != defcorename) 716 free(lim->pl_corename, M_TEMP); 717 sv_lim = lim->pl_sv_limit; 718 mutex_destroy(&lim->pl_lock); 719 pool_cache_put(plimit_cache, lim); 720 } while ((lim = sv_lim) != NULL); 721 } 722 723 struct pstats * 724 pstatscopy(struct pstats *ps) 725 { 726 727 struct pstats *newps; 728 729 newps = pool_cache_get(pstats_cache, PR_WAITOK); 730 731 memset(&newps->pstat_startzero, 0, 732 (unsigned) ((char *)&newps->pstat_endzero - 733 (char *)&newps->pstat_startzero)); 734 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy, 735 ((char *)&newps->pstat_endcopy - 736 (char *)&newps->pstat_startcopy)); 737 738 return (newps); 739 740 } 741 742 void 743 pstatsfree(struct pstats *ps) 744 { 745 746 pool_cache_put(pstats_cache, ps); 747 } 748 749 /* 750 * sysctl interface in five parts 751 */ 752 753 /* 754 * a routine for sysctl proc subtree helpers that need to pick a valid 755 * process by pid. 756 */ 757 static int 758 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid) 759 { 760 struct proc *ptmp; 761 int error = 0; 762 763 if (pid == PROC_CURPROC) 764 ptmp = l->l_proc; 765 else if ((ptmp = pfind(pid)) == NULL) 766 error = ESRCH; 767 768 *p2 = ptmp; 769 return (error); 770 } 771 772 /* 773 * sysctl helper routine for setting a process's specific corefile 774 * name. picks the process based on the given pid and checks the 775 * correctness of the new value. 776 */ 777 static int 778 sysctl_proc_corename(SYSCTLFN_ARGS) 779 { 780 struct proc *ptmp; 781 struct plimit *lim; 782 int error = 0, len; 783 char *cname; 784 char *ocore; 785 char *tmp; 786 struct sysctlnode node; 787 788 /* 789 * is this all correct? 790 */ 791 if (namelen != 0) 792 return (EINVAL); 793 if (name[-1] != PROC_PID_CORENAME) 794 return (EINVAL); 795 796 /* 797 * whom are we tweaking? 798 */ 799 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 800 if (error) 801 return (error); 802 803 /* XXX-elad */ 804 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 805 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 806 if (error) 807 return (error); 808 809 if (newp == NULL) { 810 error = kauth_authorize_process(l->l_cred, 811 KAUTH_PROCESS_CORENAME, ptmp, 812 KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_GET), NULL, NULL); 813 if (error) 814 return (error); 815 } 816 817 /* 818 * let them modify a temporary copy of the core name 819 */ 820 cname = PNBUF_GET(); 821 lim = ptmp->p_limit; 822 mutex_enter(&lim->pl_lock); 823 strlcpy(cname, lim->pl_corename, MAXPATHLEN); 824 mutex_exit(&lim->pl_lock); 825 826 node = *rnode; 827 node.sysctl_data = cname; 828 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 829 830 /* 831 * if that failed, or they have nothing new to say, or we've 832 * heard it before... 833 */ 834 if (error || newp == NULL) 835 goto done; 836 lim = ptmp->p_limit; 837 mutex_enter(&lim->pl_lock); 838 error = strcmp(cname, lim->pl_corename); 839 mutex_exit(&lim->pl_lock); 840 if (error == 0) 841 /* Unchanged */ 842 goto done; 843 844 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 845 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_SET), cname, NULL); 846 if (error) 847 return (error); 848 849 /* 850 * no error yet and cname now has the new core name in it. 851 * let's see if it looks acceptable. it must be either "core" 852 * or end in ".core" or "/core". 853 */ 854 len = strlen(cname); 855 if (len < 4) { 856 error = EINVAL; 857 } else if (strcmp(cname + len - 4, "core") != 0) { 858 error = EINVAL; 859 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') { 860 error = EINVAL; 861 } 862 if (error != 0) { 863 goto done; 864 } 865 866 /* 867 * hmm...looks good. now...where do we put it? 868 */ 869 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL); 870 if (tmp == NULL) { 871 error = ENOMEM; 872 goto done; 873 } 874 memcpy(tmp, cname, len + 1); 875 876 lim_privatise(ptmp, false); 877 lim = ptmp->p_limit; 878 mutex_enter(&lim->pl_lock); 879 ocore = lim->pl_corename; 880 lim->pl_corename = tmp; 881 mutex_exit(&lim->pl_lock); 882 if (ocore != defcorename) 883 free(ocore, M_TEMP); 884 885 done: 886 PNBUF_PUT(cname); 887 return error; 888 } 889 890 /* 891 * sysctl helper routine for checking/setting a process's stop flags, 892 * one for fork and one for exec. 893 */ 894 static int 895 sysctl_proc_stop(SYSCTLFN_ARGS) 896 { 897 struct proc *ptmp; 898 int i, f, error = 0; 899 struct sysctlnode node; 900 901 if (namelen != 0) 902 return (EINVAL); 903 904 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]); 905 if (error) 906 return (error); 907 908 /* XXX-elad */ 909 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 910 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 911 if (error) 912 return (error); 913 914 switch (rnode->sysctl_num) { 915 case PROC_PID_STOPFORK: 916 f = PS_STOPFORK; 917 break; 918 case PROC_PID_STOPEXEC: 919 f = PS_STOPEXEC; 920 break; 921 case PROC_PID_STOPEXIT: 922 f = PS_STOPEXIT; 923 break; 924 default: 925 return (EINVAL); 926 } 927 928 i = (ptmp->p_flag & f) ? 1 : 0; 929 node = *rnode; 930 node.sysctl_data = &i; 931 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 932 if (error || newp == NULL) 933 return (error); 934 935 mutex_enter(ptmp->p_lock); 936 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 937 ptmp, KAUTH_ARG(f), NULL, NULL); 938 if (!error) { 939 if (i) { 940 ptmp->p_sflag |= f; 941 } else { 942 ptmp->p_sflag &= ~f; 943 } 944 } 945 mutex_exit(ptmp->p_lock); 946 947 return error; 948 } 949 950 /* 951 * sysctl helper routine for a process's rlimits as exposed by sysctl. 952 */ 953 static int 954 sysctl_proc_plimit(SYSCTLFN_ARGS) 955 { 956 struct proc *ptmp; 957 u_int limitno; 958 int which, error = 0; 959 struct rlimit alim; 960 struct sysctlnode node; 961 962 if (namelen != 0) 963 return (EINVAL); 964 965 which = name[-1]; 966 if (which != PROC_PID_LIMIT_TYPE_SOFT && 967 which != PROC_PID_LIMIT_TYPE_HARD) 968 return (EINVAL); 969 970 limitno = name[-2] - 1; 971 if (limitno >= RLIM_NLIMITS) 972 return (EINVAL); 973 974 if (name[-3] != PROC_PID_LIMIT) 975 return (EINVAL); 976 977 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]); 978 if (error) 979 return (error); 980 981 /* XXX-elad */ 982 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp, 983 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 984 if (error) 985 return (error); 986 987 /* Check if we can view limits. */ 988 if (newp == NULL) { 989 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 990 ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_GET), &alim, 991 KAUTH_ARG(which)); 992 if (error) 993 return (error); 994 } 995 996 node = *rnode; 997 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim)); 998 if (which == PROC_PID_LIMIT_TYPE_HARD) 999 node.sysctl_data = &alim.rlim_max; 1000 else 1001 node.sysctl_data = &alim.rlim_cur; 1002 1003 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1004 if (error || newp == NULL) 1005 return (error); 1006 1007 return (dosetrlimit(l, ptmp, limitno, &alim)); 1008 } 1009 1010 static struct sysctllog *proc_sysctllog; 1011 1012 /* 1013 * and finally, the actually glue that sticks it to the tree 1014 */ 1015 static void 1016 sysctl_proc_setup() 1017 { 1018 1019 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1020 CTLFLAG_PERMANENT, 1021 CTLTYPE_NODE, "proc", NULL, 1022 NULL, 0, NULL, 0, 1023 CTL_PROC, CTL_EOL); 1024 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1025 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 1026 CTLTYPE_NODE, "curproc", 1027 SYSCTL_DESCR("Per-process settings"), 1028 NULL, 0, NULL, 0, 1029 CTL_PROC, PROC_CURPROC, CTL_EOL); 1030 1031 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1032 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1033 CTLTYPE_STRING, "corename", 1034 SYSCTL_DESCR("Core file name"), 1035 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 1036 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 1037 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1038 CTLFLAG_PERMANENT, 1039 CTLTYPE_NODE, "rlimit", 1040 SYSCTL_DESCR("Process limits"), 1041 NULL, 0, NULL, 0, 1042 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 1043 1044 #define create_proc_plimit(s, n) do { \ 1045 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1046 CTLFLAG_PERMANENT, \ 1047 CTLTYPE_NODE, s, \ 1048 SYSCTL_DESCR("Process " s " limits"), \ 1049 NULL, 0, NULL, 0, \ 1050 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1051 CTL_EOL); \ 1052 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1053 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1054 CTLTYPE_QUAD, "soft", \ 1055 SYSCTL_DESCR("Process soft " s " limit"), \ 1056 sysctl_proc_plimit, 0, NULL, 0, \ 1057 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1058 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 1059 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1060 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1061 CTLTYPE_QUAD, "hard", \ 1062 SYSCTL_DESCR("Process hard " s " limit"), \ 1063 sysctl_proc_plimit, 0, NULL, 0, \ 1064 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1065 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 1066 } while (0/*CONSTCOND*/) 1067 1068 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 1069 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 1070 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 1071 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 1072 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 1073 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 1074 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 1075 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 1076 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 1077 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 1078 create_proc_plimit("vmemoryuse", PROC_PID_LIMIT_AS); 1079 1080 #undef create_proc_plimit 1081 1082 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1083 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1084 CTLTYPE_INT, "stopfork", 1085 SYSCTL_DESCR("Stop process at fork(2)"), 1086 sysctl_proc_stop, 0, NULL, 0, 1087 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 1088 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1089 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1090 CTLTYPE_INT, "stopexec", 1091 SYSCTL_DESCR("Stop process at execve(2)"), 1092 sysctl_proc_stop, 0, NULL, 0, 1093 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 1094 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1095 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1096 CTLTYPE_INT, "stopexit", 1097 SYSCTL_DESCR("Stop process before completing exit"), 1098 sysctl_proc_stop, 0, NULL, 0, 1099 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1100 } 1101
Indexes created Sun Sep 21 14:09:52 GMT 2025