1 /* $NetBSD: kern_resource.c,v 1.182 2019/04/05 00:33:21 mlelstv 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 1.182 2019/04/05 00:33:21 mlelstv 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/kmem.h> 48 #include <sys/namei.h> 49 #include <sys/pool.h> 50 #include <sys/proc.h> 51 #include <sys/sysctl.h> 52 #include <sys/timevar.h> 53 #include <sys/kauth.h> 54 #include <sys/atomic.h> 55 #include <sys/mount.h> 56 #include <sys/syscallargs.h> 57 #include <sys/atomic.h> 58 59 #include <uvm/uvm_extern.h> 60 61 /* 62 * Maximum process data and stack limits. 63 * They are variables so they are patchable. 64 */ 65 rlim_t maxdmap = MAXDSIZ; 66 rlim_t maxsmap = MAXSSIZ; 67 68 static pool_cache_t plimit_cache __read_mostly; 69 static pool_cache_t pstats_cache __read_mostly; 70 71 static kauth_listener_t resource_listener; 72 static struct sysctllog *proc_sysctllog; 73 74 static int donice(struct lwp *, struct proc *, int); 75 static void sysctl_proc_setup(void); 76 77 static int 78 resource_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 79 void *arg0, void *arg1, void *arg2, void *arg3) 80 { 81 struct proc *p; 82 int result; 83 84 result = KAUTH_RESULT_DEFER; 85 p = arg0; 86 87 switch (action) { 88 case KAUTH_PROCESS_NICE: 89 if (kauth_cred_geteuid(cred) != kauth_cred_geteuid(p->p_cred) && 90 kauth_cred_getuid(cred) != kauth_cred_geteuid(p->p_cred)) { 91 break; 92 } 93 94 if ((u_long)arg1 >= p->p_nice) 95 result = KAUTH_RESULT_ALLOW; 96 97 break; 98 99 case KAUTH_PROCESS_RLIMIT: { 100 enum kauth_process_req req; 101 102 req = (enum kauth_process_req)(unsigned long)arg1; 103 104 switch (req) { 105 case KAUTH_REQ_PROCESS_RLIMIT_GET: 106 result = KAUTH_RESULT_ALLOW; 107 break; 108 109 case KAUTH_REQ_PROCESS_RLIMIT_SET: { 110 struct rlimit *new_rlimit; 111 u_long which; 112 113 if ((p != curlwp->l_proc) && 114 (proc_uidmatch(cred, p->p_cred) != 0)) 115 break; 116 117 new_rlimit = arg2; 118 which = (u_long)arg3; 119 120 if (new_rlimit->rlim_max <= p->p_rlimit[which].rlim_max) 121 result = KAUTH_RESULT_ALLOW; 122 123 break; 124 } 125 126 default: 127 break; 128 } 129 130 break; 131 } 132 133 default: 134 break; 135 } 136 137 return result; 138 } 139 140 void 141 resource_init(void) 142 { 143 144 plimit_cache = pool_cache_init(sizeof(struct plimit), 0, 0, 0, 145 "plimitpl", NULL, IPL_NONE, NULL, NULL, NULL); 146 pstats_cache = pool_cache_init(sizeof(struct pstats), 0, 0, 0, 147 "pstatspl", NULL, IPL_NONE, NULL, NULL, NULL); 148 149 resource_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 150 resource_listener_cb, NULL); 151 152 sysctl_proc_setup(); 153 } 154 155 /* 156 * Resource controls and accounting. 157 */ 158 159 int 160 sys_getpriority(struct lwp *l, const struct sys_getpriority_args *uap, 161 register_t *retval) 162 { 163 /* { 164 syscallarg(int) which; 165 syscallarg(id_t) who; 166 } */ 167 struct proc *curp = l->l_proc, *p; 168 id_t who = SCARG(uap, who); 169 int low = NZERO + PRIO_MAX + 1; 170 171 mutex_enter(proc_lock); 172 switch (SCARG(uap, which)) { 173 case PRIO_PROCESS: 174 p = who ? proc_find(who) : curp; 175 if (p != NULL) 176 low = p->p_nice; 177 break; 178 179 case PRIO_PGRP: { 180 struct pgrp *pg; 181 182 if (who == 0) 183 pg = curp->p_pgrp; 184 else if ((pg = pgrp_find(who)) == NULL) 185 break; 186 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 187 if (p->p_nice < low) 188 low = p->p_nice; 189 } 190 break; 191 } 192 193 case PRIO_USER: 194 if (who == 0) 195 who = (int)kauth_cred_geteuid(l->l_cred); 196 PROCLIST_FOREACH(p, &allproc) { 197 mutex_enter(p->p_lock); 198 if (kauth_cred_geteuid(p->p_cred) == 199 (uid_t)who && p->p_nice < low) 200 low = p->p_nice; 201 mutex_exit(p->p_lock); 202 } 203 break; 204 205 default: 206 mutex_exit(proc_lock); 207 return EINVAL; 208 } 209 mutex_exit(proc_lock); 210 211 if (low == NZERO + PRIO_MAX + 1) { 212 return ESRCH; 213 } 214 *retval = low - NZERO; 215 return 0; 216 } 217 218 int 219 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap, 220 register_t *retval) 221 { 222 /* { 223 syscallarg(int) which; 224 syscallarg(id_t) who; 225 syscallarg(int) prio; 226 } */ 227 struct proc *curp = l->l_proc, *p; 228 id_t who = SCARG(uap, who); 229 int found = 0, error = 0; 230 231 mutex_enter(proc_lock); 232 switch (SCARG(uap, which)) { 233 case PRIO_PROCESS: 234 p = who ? proc_find(who) : curp; 235 if (p != NULL) { 236 mutex_enter(p->p_lock); 237 found++; 238 error = donice(l, p, SCARG(uap, prio)); 239 mutex_exit(p->p_lock); 240 } 241 break; 242 243 case PRIO_PGRP: { 244 struct pgrp *pg; 245 246 if (who == 0) 247 pg = curp->p_pgrp; 248 else if ((pg = pgrp_find(who)) == NULL) 249 break; 250 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 251 mutex_enter(p->p_lock); 252 found++; 253 error = donice(l, p, SCARG(uap, prio)); 254 mutex_exit(p->p_lock); 255 if (error) 256 break; 257 } 258 break; 259 } 260 261 case PRIO_USER: 262 if (who == 0) 263 who = (int)kauth_cred_geteuid(l->l_cred); 264 PROCLIST_FOREACH(p, &allproc) { 265 mutex_enter(p->p_lock); 266 if (kauth_cred_geteuid(p->p_cred) == 267 (uid_t)SCARG(uap, who)) { 268 found++; 269 error = donice(l, p, SCARG(uap, prio)); 270 } 271 mutex_exit(p->p_lock); 272 if (error) 273 break; 274 } 275 break; 276 277 default: 278 mutex_exit(proc_lock); 279 return EINVAL; 280 } 281 mutex_exit(proc_lock); 282 283 return (found == 0) ? ESRCH : error; 284 } 285 286 /* 287 * Renice a process. 288 * 289 * Call with the target process' credentials locked. 290 */ 291 static int 292 donice(struct lwp *l, struct proc *chgp, int n) 293 { 294 kauth_cred_t cred = l->l_cred; 295 296 KASSERT(mutex_owned(chgp->p_lock)); 297 298 if (kauth_cred_geteuid(cred) && kauth_cred_getuid(cred) && 299 kauth_cred_geteuid(cred) != kauth_cred_geteuid(chgp->p_cred) && 300 kauth_cred_getuid(cred) != kauth_cred_geteuid(chgp->p_cred)) 301 return EPERM; 302 303 if (n > PRIO_MAX) { 304 n = PRIO_MAX; 305 } 306 if (n < PRIO_MIN) { 307 n = PRIO_MIN; 308 } 309 n += NZERO; 310 311 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp, 312 KAUTH_ARG(n), NULL, NULL)) { 313 return EACCES; 314 } 315 316 sched_nice(chgp, n); 317 return 0; 318 } 319 320 int 321 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap, 322 register_t *retval) 323 { 324 /* { 325 syscallarg(int) which; 326 syscallarg(const struct rlimit *) rlp; 327 } */ 328 int error, which = SCARG(uap, which); 329 struct rlimit alim; 330 331 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit)); 332 if (error) { 333 return error; 334 } 335 return dosetrlimit(l, l->l_proc, which, &alim); 336 } 337 338 int 339 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp) 340 { 341 struct rlimit *alimp; 342 int error; 343 344 if ((u_int)which >= RLIM_NLIMITS) 345 return EINVAL; 346 347 if (limp->rlim_cur > limp->rlim_max) { 348 /* 349 * This is programming error. According to SUSv2, we should 350 * return error in this case. 351 */ 352 return EINVAL; 353 } 354 355 alimp = &p->p_rlimit[which]; 356 /* if we don't change the value, no need to limcopy() */ 357 if (limp->rlim_cur == alimp->rlim_cur && 358 limp->rlim_max == alimp->rlim_max) 359 return 0; 360 361 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 362 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_SET), limp, KAUTH_ARG(which)); 363 if (error) 364 return error; 365 366 lim_privatise(p); 367 /* p->p_limit is now unchangeable */ 368 alimp = &p->p_rlimit[which]; 369 370 switch (which) { 371 372 case RLIMIT_DATA: 373 if (limp->rlim_cur > maxdmap) 374 limp->rlim_cur = maxdmap; 375 if (limp->rlim_max > maxdmap) 376 limp->rlim_max = maxdmap; 377 break; 378 379 case RLIMIT_STACK: 380 if (limp->rlim_cur > maxsmap) 381 limp->rlim_cur = maxsmap; 382 if (limp->rlim_max > maxsmap) 383 limp->rlim_max = maxsmap; 384 385 /* 386 * Return EINVAL if the new stack size limit is lower than 387 * current usage. Otherwise, the process would get SIGSEGV the 388 * moment it would try to access anything on its current stack. 389 * This conforms to SUSv2. 390 */ 391 if (btoc(limp->rlim_cur) < p->p_vmspace->vm_ssize || 392 btoc(limp->rlim_max) < p->p_vmspace->vm_ssize) { 393 return EINVAL; 394 } 395 396 /* 397 * Stack is allocated to the max at exec time with 398 * only "rlim_cur" bytes accessible (In other words, 399 * allocates stack dividing two contiguous regions at 400 * "rlim_cur" bytes boundary). 401 * 402 * Since allocation is done in terms of page, roundup 403 * "rlim_cur" (otherwise, contiguous regions 404 * overlap). If stack limit is going up make more 405 * accessible, if going down make inaccessible. 406 */ 407 limp->rlim_max = round_page(limp->rlim_max); 408 limp->rlim_cur = round_page(limp->rlim_cur); 409 if (limp->rlim_cur != alimp->rlim_cur) { 410 vaddr_t addr; 411 vsize_t size; 412 vm_prot_t prot; 413 char *base, *tmp; 414 415 base = p->p_vmspace->vm_minsaddr; 416 if (limp->rlim_cur > alimp->rlim_cur) { 417 prot = VM_PROT_READ | VM_PROT_WRITE; 418 size = limp->rlim_cur - alimp->rlim_cur; 419 tmp = STACK_GROW(base, alimp->rlim_cur); 420 } else { 421 prot = VM_PROT_NONE; 422 size = alimp->rlim_cur - limp->rlim_cur; 423 tmp = STACK_GROW(base, limp->rlim_cur); 424 } 425 addr = (vaddr_t)STACK_ALLOC(tmp, size); 426 (void) uvm_map_protect(&p->p_vmspace->vm_map, 427 addr, addr + size, prot, false); 428 } 429 break; 430 431 case RLIMIT_NOFILE: 432 if (limp->rlim_cur > maxfiles) 433 limp->rlim_cur = maxfiles; 434 if (limp->rlim_max > maxfiles) 435 limp->rlim_max = maxfiles; 436 break; 437 438 case RLIMIT_NPROC: 439 if (limp->rlim_cur > maxproc) 440 limp->rlim_cur = maxproc; 441 if (limp->rlim_max > maxproc) 442 limp->rlim_max = maxproc; 443 break; 444 445 case RLIMIT_NTHR: 446 if (limp->rlim_cur > maxlwp) 447 limp->rlim_cur = maxlwp; 448 if (limp->rlim_max > maxlwp) 449 limp->rlim_max = maxlwp; 450 break; 451 } 452 453 mutex_enter(&p->p_limit->pl_lock); 454 *alimp = *limp; 455 mutex_exit(&p->p_limit->pl_lock); 456 return 0; 457 } 458 459 int 460 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap, 461 register_t *retval) 462 { 463 /* { 464 syscallarg(int) which; 465 syscallarg(struct rlimit *) rlp; 466 } */ 467 struct proc *p = l->l_proc; 468 int which = SCARG(uap, which); 469 struct rlimit rl; 470 471 if ((u_int)which >= RLIM_NLIMITS) 472 return EINVAL; 473 474 mutex_enter(p->p_lock); 475 memcpy(&rl, &p->p_rlimit[which], sizeof(rl)); 476 mutex_exit(p->p_lock); 477 478 return copyout(&rl, SCARG(uap, rlp), sizeof(rl)); 479 } 480 481 /* 482 * Transform the running time and tick information in proc p into user, 483 * system, and interrupt time usage. 484 * 485 * Should be called with p->p_lock held unless called from exit1(). 486 */ 487 void 488 calcru(struct proc *p, struct timeval *up, struct timeval *sp, 489 struct timeval *ip, struct timeval *rp) 490 { 491 uint64_t u, st, ut, it, tot, dt; 492 struct lwp *l; 493 struct bintime tm; 494 struct timeval tv; 495 496 KASSERT(p->p_stat == SDEAD || mutex_owned(p->p_lock)); 497 498 mutex_spin_enter(&p->p_stmutex); 499 st = p->p_sticks; 500 ut = p->p_uticks; 501 it = p->p_iticks; 502 mutex_spin_exit(&p->p_stmutex); 503 504 tm = p->p_rtime; 505 506 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 507 lwp_lock(l); 508 bintime_add(&tm, &l->l_rtime); 509 if ((l->l_pflag & LP_RUNNING) != 0) { 510 struct bintime diff; 511 /* 512 * Adjust for the current time slice. This is 513 * actually fairly important since the error 514 * here is on the order of a time quantum, 515 * which is much greater than the sampling 516 * error. 517 */ 518 binuptime(&diff); 519 bintime_sub(&diff, &l->l_stime); 520 bintime_add(&tm, &diff); 521 } 522 lwp_unlock(l); 523 } 524 525 tot = st + ut + it; 526 bintime2timeval(&tm, &tv); 527 u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec; 528 529 if (tot == 0) { 530 /* No ticks, so can't use to share time out, split 50-50 */ 531 st = ut = u / 2; 532 } else { 533 st = (u * st) / tot; 534 ut = (u * ut) / tot; 535 } 536 537 /* 538 * Try to avoid lying to the users (too much) 539 * 540 * Of course, user/sys time are based on sampling (ie: statistics) 541 * so that would be impossible, but convincing the mark 542 * that we have used less ?time this call than we had 543 * last time, is beyond reasonable... (the con fails!) 544 * 545 * Note that since actual used time cannot decrease, either 546 * utime or stime (or both) must be greater now than last time 547 * (or both the same) - if one seems to have decreased, hold 548 * it constant and steal the necessary bump from the other 549 * which must have increased. 550 */ 551 if (p->p_xutime > ut) { 552 dt = p->p_xutime - ut; 553 st -= uimin(dt, st); 554 ut = p->p_xutime; 555 } else if (p->p_xstime > st) { 556 dt = p->p_xstime - st; 557 ut -= uimin(dt, ut); 558 st = p->p_xstime; 559 } 560 561 if (sp != NULL) { 562 p->p_xstime = st; 563 sp->tv_sec = st / 1000000; 564 sp->tv_usec = st % 1000000; 565 } 566 if (up != NULL) { 567 p->p_xutime = ut; 568 up->tv_sec = ut / 1000000; 569 up->tv_usec = ut % 1000000; 570 } 571 if (ip != NULL) { 572 if (it != 0) /* it != 0 --> tot != 0 */ 573 it = (u * it) / tot; 574 ip->tv_sec = it / 1000000; 575 ip->tv_usec = it % 1000000; 576 } 577 if (rp != NULL) { 578 *rp = tv; 579 } 580 } 581 582 int 583 sys___getrusage50(struct lwp *l, const struct sys___getrusage50_args *uap, 584 register_t *retval) 585 { 586 /* { 587 syscallarg(int) who; 588 syscallarg(struct rusage *) rusage; 589 } */ 590 int error; 591 struct rusage ru; 592 struct proc *p = l->l_proc; 593 594 error = getrusage1(p, SCARG(uap, who), &ru); 595 if (error != 0) 596 return error; 597 598 return copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 599 } 600 601 int 602 getrusage1(struct proc *p, int who, struct rusage *ru) { 603 604 switch (who) { 605 case RUSAGE_SELF: 606 mutex_enter(p->p_lock); 607 ruspace(p); 608 memcpy(ru, &p->p_stats->p_ru, sizeof(*ru)); 609 calcru(p, &ru->ru_utime, &ru->ru_stime, NULL, NULL); 610 rulwps(p, ru); 611 mutex_exit(p->p_lock); 612 break; 613 case RUSAGE_CHILDREN: 614 mutex_enter(p->p_lock); 615 memcpy(ru, &p->p_stats->p_cru, sizeof(*ru)); 616 mutex_exit(p->p_lock); 617 break; 618 default: 619 return EINVAL; 620 } 621 622 return 0; 623 } 624 625 void 626 ruspace(struct proc *p) 627 { 628 struct vmspace *vm = p->p_vmspace; 629 struct rusage *ru = &p->p_stats->p_ru; 630 631 ru->ru_ixrss = vm->vm_tsize << (PAGE_SHIFT - 10); 632 ru->ru_idrss = vm->vm_dsize << (PAGE_SHIFT - 10); 633 ru->ru_isrss = vm->vm_ssize << (PAGE_SHIFT - 10); 634 #ifdef __HAVE_NO_PMAP_STATS 635 /* We don't keep track of the max so we get the current */ 636 ru->ru_maxrss = vm_resident_count(vm) << (PAGE_SHIFT - 10); 637 #else 638 ru->ru_maxrss = vm->vm_rssmax << (PAGE_SHIFT - 10); 639 #endif 640 } 641 642 void 643 ruadd(struct rusage *ru, struct rusage *ru2) 644 { 645 long *ip, *ip2; 646 int i; 647 648 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime); 649 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime); 650 if (ru->ru_maxrss < ru2->ru_maxrss) 651 ru->ru_maxrss = ru2->ru_maxrss; 652 ip = &ru->ru_first; ip2 = &ru2->ru_first; 653 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 654 *ip++ += *ip2++; 655 } 656 657 void 658 rulwps(proc_t *p, struct rusage *ru) 659 { 660 lwp_t *l; 661 662 KASSERT(mutex_owned(p->p_lock)); 663 664 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 665 ruadd(ru, &l->l_ru); 666 ru->ru_nvcsw += (l->l_ncsw - l->l_nivcsw); 667 ru->ru_nivcsw += l->l_nivcsw; 668 } 669 } 670 671 /* 672 * lim_copy: make a copy of the plimit structure. 673 * 674 * We use copy-on-write after fork, and copy when a limit is changed. 675 */ 676 struct plimit * 677 lim_copy(struct plimit *lim) 678 { 679 struct plimit *newlim; 680 char *corename; 681 size_t alen, len; 682 683 newlim = pool_cache_get(plimit_cache, PR_WAITOK); 684 mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE); 685 newlim->pl_writeable = false; 686 newlim->pl_refcnt = 1; 687 newlim->pl_sv_limit = NULL; 688 689 mutex_enter(&lim->pl_lock); 690 memcpy(newlim->pl_rlimit, lim->pl_rlimit, 691 sizeof(struct rlimit) * RLIM_NLIMITS); 692 693 /* 694 * Note: the common case is a use of default core name. 695 */ 696 alen = 0; 697 corename = NULL; 698 for (;;) { 699 if (lim->pl_corename == defcorename) { 700 newlim->pl_corename = defcorename; 701 newlim->pl_cnlen = 0; 702 break; 703 } 704 len = lim->pl_cnlen; 705 if (len == alen) { 706 newlim->pl_corename = corename; 707 newlim->pl_cnlen = len; 708 memcpy(corename, lim->pl_corename, len); 709 corename = NULL; 710 break; 711 } 712 mutex_exit(&lim->pl_lock); 713 if (corename) { 714 kmem_free(corename, alen); 715 } 716 alen = len; 717 corename = kmem_alloc(alen, KM_SLEEP); 718 mutex_enter(&lim->pl_lock); 719 } 720 mutex_exit(&lim->pl_lock); 721 722 if (corename) { 723 kmem_free(corename, alen); 724 } 725 return newlim; 726 } 727 728 void 729 lim_addref(struct plimit *lim) 730 { 731 atomic_inc_uint(&lim->pl_refcnt); 732 } 733 734 /* 735 * lim_privatise: give a process its own private plimit structure. 736 */ 737 void 738 lim_privatise(proc_t *p) 739 { 740 struct plimit *lim = p->p_limit, *newlim; 741 742 if (lim->pl_writeable) { 743 return; 744 } 745 746 newlim = lim_copy(lim); 747 748 mutex_enter(p->p_lock); 749 if (p->p_limit->pl_writeable) { 750 /* Other thread won the race. */ 751 mutex_exit(p->p_lock); 752 lim_free(newlim); 753 return; 754 } 755 756 /* 757 * Since p->p_limit can be accessed without locked held, 758 * old limit structure must not be deleted yet. 759 */ 760 newlim->pl_sv_limit = p->p_limit; 761 newlim->pl_writeable = true; 762 p->p_limit = newlim; 763 mutex_exit(p->p_lock); 764 } 765 766 void 767 lim_setcorename(proc_t *p, char *name, size_t len) 768 { 769 struct plimit *lim; 770 char *oname; 771 size_t olen; 772 773 lim_privatise(p); 774 lim = p->p_limit; 775 776 mutex_enter(&lim->pl_lock); 777 oname = lim->pl_corename; 778 olen = lim->pl_cnlen; 779 lim->pl_corename = name; 780 lim->pl_cnlen = len; 781 mutex_exit(&lim->pl_lock); 782 783 if (oname != defcorename) { 784 kmem_free(oname, olen); 785 } 786 } 787 788 void 789 lim_free(struct plimit *lim) 790 { 791 struct plimit *sv_lim; 792 793 do { 794 if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0) { 795 return; 796 } 797 if (lim->pl_corename != defcorename) { 798 kmem_free(lim->pl_corename, lim->pl_cnlen); 799 } 800 sv_lim = lim->pl_sv_limit; 801 mutex_destroy(&lim->pl_lock); 802 pool_cache_put(plimit_cache, lim); 803 } while ((lim = sv_lim) != NULL); 804 } 805 806 struct pstats * 807 pstatscopy(struct pstats *ps) 808 { 809 struct pstats *nps; 810 size_t len; 811 812 nps = pool_cache_get(pstats_cache, PR_WAITOK); 813 814 len = (char *)&nps->pstat_endzero - (char *)&nps->pstat_startzero; 815 memset(&nps->pstat_startzero, 0, len); 816 817 len = (char *)&nps->pstat_endcopy - (char *)&nps->pstat_startcopy; 818 memcpy(&nps->pstat_startcopy, &ps->pstat_startcopy, len); 819 820 return nps; 821 } 822 823 void 824 pstatsfree(struct pstats *ps) 825 { 826 827 pool_cache_put(pstats_cache, ps); 828 } 829 830 /* 831 * sysctl_proc_findproc: a routine for sysctl proc subtree helpers that 832 * need to pick a valid process by PID. 833 * 834 * => Hold a reference on the process, on success. 835 */ 836 static int 837 sysctl_proc_findproc(lwp_t *l, pid_t pid, proc_t **p2) 838 { 839 proc_t *p; 840 int error; 841 842 if (pid == PROC_CURPROC) { 843 p = l->l_proc; 844 } else { 845 mutex_enter(proc_lock); 846 p = proc_find(pid); 847 if (p == NULL) { 848 mutex_exit(proc_lock); 849 return ESRCH; 850 } 851 } 852 error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY; 853 if (pid != PROC_CURPROC) { 854 mutex_exit(proc_lock); 855 } 856 *p2 = p; 857 return error; 858 } 859 860 /* 861 * sysctl_proc_paxflags: helper routine to get process's paxctl flags 862 */ 863 static int 864 sysctl_proc_paxflags(SYSCTLFN_ARGS) 865 { 866 struct proc *p; 867 struct sysctlnode node; 868 int paxflags; 869 int error; 870 871 /* First, validate the request. */ 872 if (namelen != 0 || name[-1] != PROC_PID_PAXFLAGS) 873 return EINVAL; 874 875 /* Find the process. Hold a reference (p_reflock), if found. */ 876 error = sysctl_proc_findproc(l, (pid_t)name[-2], &p); 877 if (error) 878 return error; 879 880 /* XXX-elad */ 881 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 882 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 883 if (error) { 884 rw_exit(&p->p_reflock); 885 return error; 886 } 887 888 /* Retrieve the limits. */ 889 node = *rnode; 890 paxflags = p->p_pax; 891 node.sysctl_data = &paxflags; 892 893 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 894 895 /* If attempting to write new value, it's an error */ 896 if (error == 0 && newp != NULL) 897 error = EACCES; 898 899 rw_exit(&p->p_reflock); 900 return error; 901 } 902 903 /* 904 * sysctl_proc_corename: helper routine to get or set the core file name 905 * for a process specified by PID. 906 */ 907 static int 908 sysctl_proc_corename(SYSCTLFN_ARGS) 909 { 910 struct proc *p; 911 struct plimit *lim; 912 char *cnbuf, *cname; 913 struct sysctlnode node; 914 size_t len; 915 int error; 916 917 /* First, validate the request. */ 918 if (namelen != 0 || name[-1] != PROC_PID_CORENAME) 919 return EINVAL; 920 921 /* Find the process. Hold a reference (p_reflock), if found. */ 922 error = sysctl_proc_findproc(l, (pid_t)name[-2], &p); 923 if (error) 924 return error; 925 926 /* XXX-elad */ 927 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 928 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 929 if (error) { 930 rw_exit(&p->p_reflock); 931 return error; 932 } 933 934 cnbuf = PNBUF_GET(); 935 936 if (oldp) { 937 /* Get case: copy the core name into the buffer. */ 938 error = kauth_authorize_process(l->l_cred, 939 KAUTH_PROCESS_CORENAME, p, 940 KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_GET), NULL, NULL); 941 if (error) { 942 goto done; 943 } 944 lim = p->p_limit; 945 mutex_enter(&lim->pl_lock); 946 strlcpy(cnbuf, lim->pl_corename, MAXPATHLEN); 947 mutex_exit(&lim->pl_lock); 948 } 949 950 node = *rnode; 951 node.sysctl_data = cnbuf; 952 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 953 954 /* Return if error, or if caller is only getting the core name. */ 955 if (error || newp == NULL) { 956 goto done; 957 } 958 959 /* 960 * Set case. Check permission and then validate new core name. 961 * It must be either "core", "/core", or end in ".core". 962 */ 963 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME, 964 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_SET), cnbuf, NULL); 965 if (error) { 966 goto done; 967 } 968 len = strlen(cnbuf); 969 if ((len < 4 || strcmp(cnbuf + len - 4, "core") != 0) || 970 (len > 4 && cnbuf[len - 5] != '/' && cnbuf[len - 5] != '.')) { 971 error = EINVAL; 972 goto done; 973 } 974 975 /* Allocate, copy and set the new core name for plimit structure. */ 976 cname = kmem_alloc(++len, KM_NOSLEEP); 977 if (cname == NULL) { 978 error = ENOMEM; 979 goto done; 980 } 981 memcpy(cname, cnbuf, len); 982 lim_setcorename(p, cname, len); 983 done: 984 rw_exit(&p->p_reflock); 985 PNBUF_PUT(cnbuf); 986 return error; 987 } 988 989 /* 990 * sysctl_proc_stop: helper routine for checking/setting the stop flags. 991 */ 992 static int 993 sysctl_proc_stop(SYSCTLFN_ARGS) 994 { 995 struct proc *p; 996 int isset, flag, error = 0; 997 struct sysctlnode node; 998 999 if (namelen != 0) 1000 return EINVAL; 1001 1002 /* Find the process. Hold a reference (p_reflock), if found. */ 1003 error = sysctl_proc_findproc(l, (pid_t)name[-2], &p); 1004 if (error) 1005 return error; 1006 1007 /* XXX-elad */ 1008 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 1009 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 1010 if (error) { 1011 goto out; 1012 } 1013 1014 /* Determine the flag. */ 1015 switch (rnode->sysctl_num) { 1016 case PROC_PID_STOPFORK: 1017 flag = PS_STOPFORK; 1018 break; 1019 case PROC_PID_STOPEXEC: 1020 flag = PS_STOPEXEC; 1021 break; 1022 case PROC_PID_STOPEXIT: 1023 flag = PS_STOPEXIT; 1024 break; 1025 default: 1026 error = EINVAL; 1027 goto out; 1028 } 1029 isset = (p->p_flag & flag) ? 1 : 0; 1030 node = *rnode; 1031 node.sysctl_data = &isset; 1032 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1033 1034 /* Return if error, or if callers is only getting the flag. */ 1035 if (error || newp == NULL) { 1036 goto out; 1037 } 1038 1039 /* Check if caller can set the flags. */ 1040 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG, 1041 p, KAUTH_ARG(flag), NULL, NULL); 1042 if (error) { 1043 goto out; 1044 } 1045 mutex_enter(p->p_lock); 1046 if (isset) { 1047 p->p_sflag |= flag; 1048 } else { 1049 p->p_sflag &= ~flag; 1050 } 1051 mutex_exit(p->p_lock); 1052 out: 1053 rw_exit(&p->p_reflock); 1054 return error; 1055 } 1056 1057 /* 1058 * sysctl_proc_plimit: helper routine to get/set rlimits of a process. 1059 */ 1060 static int 1061 sysctl_proc_plimit(SYSCTLFN_ARGS) 1062 { 1063 struct proc *p; 1064 u_int limitno; 1065 int which, error = 0; 1066 struct rlimit alim; 1067 struct sysctlnode node; 1068 1069 if (namelen != 0) 1070 return EINVAL; 1071 1072 which = name[-1]; 1073 if (which != PROC_PID_LIMIT_TYPE_SOFT && 1074 which != PROC_PID_LIMIT_TYPE_HARD) 1075 return EINVAL; 1076 1077 limitno = name[-2] - 1; 1078 if (limitno >= RLIM_NLIMITS) 1079 return EINVAL; 1080 1081 if (name[-3] != PROC_PID_LIMIT) 1082 return EINVAL; 1083 1084 /* Find the process. Hold a reference (p_reflock), if found. */ 1085 error = sysctl_proc_findproc(l, (pid_t)name[-4], &p); 1086 if (error) 1087 return error; 1088 1089 /* XXX-elad */ 1090 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, p, 1091 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 1092 if (error) 1093 goto out; 1094 1095 /* Check if caller can retrieve the limits. */ 1096 if (newp == NULL) { 1097 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 1098 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_GET), &alim, 1099 KAUTH_ARG(which)); 1100 if (error) 1101 goto out; 1102 } 1103 1104 /* Retrieve the limits. */ 1105 node = *rnode; 1106 memcpy(&alim, &p->p_rlimit[limitno], sizeof(alim)); 1107 if (which == PROC_PID_LIMIT_TYPE_HARD) { 1108 node.sysctl_data = &alim.rlim_max; 1109 } else { 1110 node.sysctl_data = &alim.rlim_cur; 1111 } 1112 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1113 1114 /* Return if error, or if we are only retrieving the limits. */ 1115 if (error || newp == NULL) { 1116 goto out; 1117 } 1118 error = dosetrlimit(l, p, limitno, &alim); 1119 out: 1120 rw_exit(&p->p_reflock); 1121 return error; 1122 } 1123 1124 /* 1125 * Setup sysctl nodes. 1126 */ 1127 static void 1128 sysctl_proc_setup(void) 1129 { 1130 1131 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1132 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER, 1133 CTLTYPE_NODE, "curproc", 1134 SYSCTL_DESCR("Per-process settings"), 1135 NULL, 0, NULL, 0, 1136 CTL_PROC, PROC_CURPROC, CTL_EOL); 1137 1138 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1139 CTLFLAG_PERMANENT|CTLFLAG_READONLY, 1140 CTLTYPE_INT, "paxflags", 1141 SYSCTL_DESCR("Process PAX control flags"), 1142 sysctl_proc_paxflags, 0, NULL, 0, 1143 CTL_PROC, PROC_CURPROC, PROC_PID_PAXFLAGS, CTL_EOL); 1144 1145 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1146 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1147 CTLTYPE_STRING, "corename", 1148 SYSCTL_DESCR("Core file name"), 1149 sysctl_proc_corename, 0, NULL, MAXPATHLEN, 1150 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL); 1151 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1152 CTLFLAG_PERMANENT, 1153 CTLTYPE_NODE, "rlimit", 1154 SYSCTL_DESCR("Process limits"), 1155 NULL, 0, NULL, 0, 1156 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL); 1157 1158 #define create_proc_plimit(s, n) do { \ 1159 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1160 CTLFLAG_PERMANENT, \ 1161 CTLTYPE_NODE, s, \ 1162 SYSCTL_DESCR("Process " s " limits"), \ 1163 NULL, 0, NULL, 0, \ 1164 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1165 CTL_EOL); \ 1166 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1167 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1168 CTLTYPE_QUAD, "soft", \ 1169 SYSCTL_DESCR("Process soft " s " limit"), \ 1170 sysctl_proc_plimit, 0, NULL, 0, \ 1171 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1172 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \ 1173 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, \ 1174 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \ 1175 CTLTYPE_QUAD, "hard", \ 1176 SYSCTL_DESCR("Process hard " s " limit"), \ 1177 sysctl_proc_plimit, 0, NULL, 0, \ 1178 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \ 1179 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \ 1180 } while (0/*CONSTCOND*/) 1181 1182 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU); 1183 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE); 1184 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA); 1185 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK); 1186 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE); 1187 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS); 1188 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK); 1189 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC); 1190 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE); 1191 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE); 1192 create_proc_plimit("vmemoryuse", PROC_PID_LIMIT_AS); 1193 create_proc_plimit("maxlwp", PROC_PID_LIMIT_NTHR); 1194 1195 #undef create_proc_plimit 1196 1197 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1198 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1199 CTLTYPE_INT, "stopfork", 1200 SYSCTL_DESCR("Stop process at fork(2)"), 1201 sysctl_proc_stop, 0, NULL, 0, 1202 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL); 1203 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1204 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1205 CTLTYPE_INT, "stopexec", 1206 SYSCTL_DESCR("Stop process at execve(2)"), 1207 sysctl_proc_stop, 0, NULL, 0, 1208 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL); 1209 sysctl_createv(&proc_sysctllog, 0, NULL, NULL, 1210 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, 1211 CTLTYPE_INT, "stopexit", 1212 SYSCTL_DESCR("Stop process before completing exit"), 1213 sysctl_proc_stop, 0, NULL, 0, 1214 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL); 1215 } 1216