kern_resource.c revision 96886
1/*- 2 * Copyright (c) 1982, 1986, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)kern_resource.c 8.5 (Berkeley) 1/21/94 39 * $FreeBSD: head/sys/kern/kern_resource.c 96886 2002-05-19 00:14:50Z jhb $ 40 */ 41 42#include "opt_compat.h" 43 44#include <sys/param.h> 45#include <sys/systm.h> 46#include <sys/sysproto.h> 47#include <sys/file.h> 48#include <sys/kernel.h> 49#include <sys/lock.h> 50#include <sys/malloc.h> 51#include <sys/mutex.h> 52#include <sys/proc.h> 53#include <sys/resourcevar.h> 54#include <sys/sx.h> 55#include <sys/time.h> 56 57#include <vm/vm.h> 58#include <vm/vm_param.h> 59#include <vm/pmap.h> 60#include <vm/vm_map.h> 61 62static int donice(struct thread *td, struct proc *chgp, int n); 63 64static MALLOC_DEFINE(M_UIDINFO, "uidinfo", "uidinfo structures"); 65#define UIHASH(uid) (&uihashtbl[(uid) & uihash]) 66static struct mtx uihashtbl_mtx; 67static LIST_HEAD(uihashhead, uidinfo) *uihashtbl; 68static u_long uihash; /* size of hash table - 1 */ 69 70static struct uidinfo *uilookup(uid_t uid); 71 72/* 73 * Resource controls and accounting. 74 */ 75 76#ifndef _SYS_SYSPROTO_H_ 77struct getpriority_args { 78 int which; 79 int who; 80}; 81#endif 82/* 83 * MPSAFE 84 */ 85int 86getpriority(td, uap) 87 struct thread *td; 88 register struct getpriority_args *uap; 89{ 90 register struct proc *p; 91 register int low = PRIO_MAX + 1; 92 int error = 0; 93 94 mtx_lock(&Giant); 95 96 switch (uap->which) { 97 case PRIO_PROCESS: 98 if (uap->who == 0) 99 low = td->td_ksegrp->kg_nice; 100 else { 101 p = pfind(uap->who); 102 if (p == NULL) 103 break; 104 if (p_cansee(td, p) == 0) 105 low = p->p_ksegrp.kg_nice /* XXXKSE */ ; 106 PROC_UNLOCK(p); 107 } 108 break; 109 110 case PRIO_PGRP: { 111 register struct pgrp *pg; 112 113 sx_slock(&proctree_lock); 114 if (uap->who == 0) { 115 pg = td->td_proc->p_pgrp; 116 PGRP_LOCK(pg); 117 } else { 118 pg = pgfind(uap->who); 119 if (pg == NULL) { 120 sx_sunlock(&proctree_lock); 121 break; 122 } 123 } 124 sx_sunlock(&proctree_lock); 125 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 126 PROC_LOCK(p); 127 if (!p_cansee(td, p) && p->p_ksegrp.kg_nice /* XXXKSE */ < low) 128 low = p->p_ksegrp.kg_nice /* XXXKSE */ ; 129 PROC_UNLOCK(p); 130 } 131 PGRP_UNLOCK(pg); 132 break; 133 } 134 135 case PRIO_USER: 136 if (uap->who == 0) 137 uap->who = td->td_ucred->cr_uid; 138 sx_slock(&allproc_lock); 139 LIST_FOREACH(p, &allproc, p_list) { 140 PROC_LOCK(p); 141 if (!p_cansee(td, p) && 142 p->p_ucred->cr_uid == uap->who && 143 p->p_ksegrp.kg_nice /* XXXKSE */ < low) 144 low = p->p_ksegrp.kg_nice /* XXXKSE */ ; 145 PROC_UNLOCK(p); 146 } 147 sx_sunlock(&allproc_lock); 148 break; 149 150 default: 151 error = EINVAL; 152 break; 153 } 154 if (low == PRIO_MAX + 1 && error == 0) 155 error = ESRCH; 156 td->td_retval[0] = low; 157 mtx_unlock(&Giant); 158 return (error); 159} 160 161#ifndef _SYS_SYSPROTO_H_ 162struct setpriority_args { 163 int which; 164 int who; 165 int prio; 166}; 167#endif 168/* 169 * MPSAFE 170 */ 171/* ARGSUSED */ 172int 173setpriority(td, uap) 174 struct thread *td; 175 register struct setpriority_args *uap; 176{ 177 struct proc *curp = td->td_proc; 178 register struct proc *p; 179 int found = 0, error = 0; 180 181 mtx_lock(&Giant); 182 183 switch (uap->which) { 184 case PRIO_PROCESS: 185 if (uap->who == 0) { 186 PROC_LOCK(curp); 187 error = donice(td, curp, uap->prio); 188 PROC_UNLOCK(curp); 189 } else { 190 p = pfind(uap->who); 191 if (p == 0) 192 break; 193 if (p_cansee(td, p) == 0) 194 error = donice(td, p, uap->prio); 195 PROC_UNLOCK(p); 196 } 197 found++; 198 break; 199 200 case PRIO_PGRP: { 201 register struct pgrp *pg; 202 203 sx_slock(&proctree_lock); 204 if (uap->who == 0) { 205 pg = curp->p_pgrp; 206 PGRP_LOCK(pg); 207 } else { 208 pg = pgfind(uap->who); 209 if (pg == NULL) { 210 sx_sunlock(&proctree_lock); 211 break; 212 } 213 } 214 sx_sunlock(&proctree_lock); 215 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 216 PROC_LOCK(p); 217 if (!p_cansee(td, p)) { 218 error = donice(td, p, uap->prio); 219 found++; 220 } 221 PROC_UNLOCK(p); 222 } 223 PGRP_UNLOCK(pg); 224 break; 225 } 226 227 case PRIO_USER: 228 if (uap->who == 0) 229 uap->who = td->td_ucred->cr_uid; 230 sx_slock(&allproc_lock); 231 FOREACH_PROC_IN_SYSTEM(p) { 232 PROC_LOCK(p); 233 if (p->p_ucred->cr_uid == uap->who && 234 !p_cansee(td, p)) { 235 error = donice(td, p, uap->prio); 236 found++; 237 } 238 PROC_UNLOCK(p); 239 } 240 sx_sunlock(&allproc_lock); 241 break; 242 243 default: 244 error = EINVAL; 245 break; 246 } 247 if (found == 0 && error == 0) 248 error = ESRCH; 249 mtx_unlock(&Giant); 250 return (error); 251} 252 253static int 254donice(td, chgp, n) 255 struct thread *td; 256 register struct proc *chgp; 257 register int n; 258{ 259 int error; 260 261 PROC_LOCK_ASSERT(chgp, MA_OWNED); 262 if ((error = p_cansched(td, chgp))) 263 return (error); 264 if (n > PRIO_MAX) 265 n = PRIO_MAX; 266 if (n < PRIO_MIN) 267 n = PRIO_MIN; 268 if (n < chgp->p_ksegrp.kg_nice /* XXXKSE */ && suser(td)) 269 return (EACCES); 270 chgp->p_ksegrp.kg_nice /* XXXKSE */ = n; 271 (void)resetpriority(&chgp->p_ksegrp); /* XXXKSE */ 272 return (0); 273} 274 275/* rtprio system call */ 276#ifndef _SYS_SYSPROTO_H_ 277struct rtprio_args { 278 int function; 279 pid_t pid; 280 struct rtprio *rtp; 281}; 282#endif 283 284/* 285 * Set realtime priority 286 */ 287 288/* 289 * MPSAFE 290 */ 291/* ARGSUSED */ 292int 293rtprio(td, uap) 294 struct thread *td; 295 register struct rtprio_args *uap; 296{ 297 struct proc *curp = td->td_proc; 298 register struct proc *p; 299 struct rtprio rtp; 300 int error, cierror = 0; 301 302 /* Perform copyin before acquiring locks if needed. */ 303 if (uap->function == RTP_SET) 304 cierror = copyin(uap->rtp, &rtp, sizeof(struct rtprio)); 305 306 if (uap->pid == 0) { 307 p = curp; 308 PROC_LOCK(p); 309 } else { 310 p = pfind(uap->pid); 311 if (p == NULL) 312 return (ESRCH); 313 } 314 315 switch (uap->function) { 316 case RTP_LOOKUP: 317 if ((error = p_cansee(td, p))) 318 break; 319 mtx_lock_spin(&sched_lock); 320 pri_to_rtp(&p->p_ksegrp /* XXXKSE */ , &rtp); 321 mtx_unlock_spin(&sched_lock); 322 PROC_UNLOCK(p); 323 return (copyout(&rtp, uap->rtp, sizeof(struct rtprio))); 324 case RTP_SET: 325 if ((error = p_cansched(td, p)) || (error = cierror)) 326 break; 327 /* disallow setting rtprio in most cases if not superuser */ 328 if (suser(td) != 0) { 329 /* can't set someone else's */ 330 if (uap->pid) { 331 error = EPERM; 332 break; 333 } 334 /* can't set realtime priority */ 335/* 336 * Realtime priority has to be restricted for reasons which should be 337 * obvious. However, for idle priority, there is a potential for 338 * system deadlock if an idleprio process gains a lock on a resource 339 * that other processes need (and the idleprio process can't run 340 * due to a CPU-bound normal process). Fix me! XXX 341 */ 342#if 0 343 if (RTP_PRIO_IS_REALTIME(rtp.type)) 344#endif 345 if (rtp.type != RTP_PRIO_NORMAL) { 346 error = EPERM; 347 break; 348 } 349 } 350 mtx_lock_spin(&sched_lock); 351 error = rtp_to_pri(&rtp, &p->p_ksegrp); 352 mtx_unlock_spin(&sched_lock); 353 break; 354 default: 355 error = EINVAL; 356 break; 357 } 358 PROC_UNLOCK(p); 359 return (error); 360} 361 362int 363rtp_to_pri(struct rtprio *rtp, struct ksegrp *kg) 364{ 365 366 if (rtp->prio > RTP_PRIO_MAX) 367 return (EINVAL); 368 switch (RTP_PRIO_BASE(rtp->type)) { 369 case RTP_PRIO_REALTIME: 370 kg->kg_user_pri = PRI_MIN_REALTIME + rtp->prio; 371 break; 372 case RTP_PRIO_NORMAL: 373 kg->kg_user_pri = PRI_MIN_TIMESHARE + rtp->prio; 374 break; 375 case RTP_PRIO_IDLE: 376 kg->kg_user_pri = PRI_MIN_IDLE + rtp->prio; 377 break; 378 default: 379 return (EINVAL); 380 } 381 kg->kg_pri_class = rtp->type; 382 if (curthread->td_ksegrp == kg) { 383 curthread->td_base_pri = kg->kg_user_pri; 384 curthread->td_priority = kg->kg_user_pri; /* XXX dubious */ 385 } 386 return (0); 387} 388 389void 390pri_to_rtp(struct ksegrp *kg, struct rtprio *rtp) 391{ 392 393 switch (PRI_BASE(kg->kg_pri_class)) { 394 case PRI_REALTIME: 395 rtp->prio = kg->kg_user_pri - PRI_MIN_REALTIME; 396 break; 397 case PRI_TIMESHARE: 398 rtp->prio = kg->kg_user_pri - PRI_MIN_TIMESHARE; 399 break; 400 case PRI_IDLE: 401 rtp->prio = kg->kg_user_pri - PRI_MIN_IDLE; 402 break; 403 default: 404 break; 405 } 406 rtp->type = kg->kg_pri_class; 407} 408 409#if defined(COMPAT_43) || defined(COMPAT_SUNOS) 410#ifndef _SYS_SYSPROTO_H_ 411struct osetrlimit_args { 412 u_int which; 413 struct orlimit *rlp; 414}; 415#endif 416/* 417 * MPSAFE 418 */ 419/* ARGSUSED */ 420int 421osetrlimit(td, uap) 422 struct thread *td; 423 register struct osetrlimit_args *uap; 424{ 425 struct orlimit olim; 426 struct rlimit lim; 427 int error; 428 429 if ((error = 430 copyin((caddr_t)uap->rlp, (caddr_t)&olim, sizeof(struct orlimit)))) 431 return (error); 432 lim.rlim_cur = olim.rlim_cur; 433 lim.rlim_max = olim.rlim_max; 434 mtx_lock(&Giant); 435 error = dosetrlimit(td, uap->which, &lim); 436 mtx_unlock(&Giant); 437 return (error); 438} 439 440#ifndef _SYS_SYSPROTO_H_ 441struct ogetrlimit_args { 442 u_int which; 443 struct orlimit *rlp; 444}; 445#endif 446/* 447 * MPSAFE 448 */ 449/* ARGSUSED */ 450int 451ogetrlimit(td, uap) 452 struct thread *td; 453 register struct ogetrlimit_args *uap; 454{ 455 struct proc *p = td->td_proc; 456 struct orlimit olim; 457 int error; 458 459 if (uap->which >= RLIM_NLIMITS) 460 return (EINVAL); 461 mtx_lock(&Giant); 462 olim.rlim_cur = p->p_rlimit[uap->which].rlim_cur; 463 if (olim.rlim_cur == -1) 464 olim.rlim_cur = 0x7fffffff; 465 olim.rlim_max = p->p_rlimit[uap->which].rlim_max; 466 if (olim.rlim_max == -1) 467 olim.rlim_max = 0x7fffffff; 468 error = copyout((caddr_t)&olim, (caddr_t)uap->rlp, sizeof(olim)); 469 mtx_unlock(&Giant); 470 return (error); 471} 472#endif /* COMPAT_43 || COMPAT_SUNOS */ 473 474#ifndef _SYS_SYSPROTO_H_ 475struct __setrlimit_args { 476 u_int which; 477 struct rlimit *rlp; 478}; 479#endif 480/* 481 * MPSAFE 482 */ 483/* ARGSUSED */ 484int 485setrlimit(td, uap) 486 struct thread *td; 487 register struct __setrlimit_args *uap; 488{ 489 struct rlimit alim; 490 int error; 491 492 if ((error = 493 copyin((caddr_t)uap->rlp, (caddr_t)&alim, sizeof (struct rlimit)))) 494 return (error); 495 mtx_lock(&Giant); 496 error = dosetrlimit(td, uap->which, &alim); 497 mtx_unlock(&Giant); 498 return (error); 499} 500 501int 502dosetrlimit(td, which, limp) 503 struct thread *td; 504 u_int which; 505 struct rlimit *limp; 506{ 507 struct proc *p = td->td_proc; 508 register struct rlimit *alimp; 509 int error; 510 511 GIANT_REQUIRED; 512 513 if (which >= RLIM_NLIMITS) 514 return (EINVAL); 515 alimp = &p->p_rlimit[which]; 516 517 /* 518 * Preserve historical bugs by treating negative limits as unsigned. 519 */ 520 if (limp->rlim_cur < 0) 521 limp->rlim_cur = RLIM_INFINITY; 522 if (limp->rlim_max < 0) 523 limp->rlim_max = RLIM_INFINITY; 524 525 if (limp->rlim_cur > alimp->rlim_max || 526 limp->rlim_max > alimp->rlim_max) 527 if ((error = suser_cred(td->td_ucred, PRISON_ROOT))) 528 return (error); 529 if (limp->rlim_cur > limp->rlim_max) 530 limp->rlim_cur = limp->rlim_max; 531 if (p->p_limit->p_refcnt > 1 && 532 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 533 p->p_limit->p_refcnt--; 534 p->p_limit = limcopy(p->p_limit); 535 alimp = &p->p_rlimit[which]; 536 } 537 538 switch (which) { 539 540 case RLIMIT_CPU: 541 if (limp->rlim_cur > RLIM_INFINITY / (rlim_t)1000000) 542 p->p_limit->p_cpulimit = RLIM_INFINITY; 543 else 544 p->p_limit->p_cpulimit = 545 (rlim_t)1000000 * limp->rlim_cur; 546 break; 547 case RLIMIT_DATA: 548 if (limp->rlim_cur > maxdsiz) 549 limp->rlim_cur = maxdsiz; 550 if (limp->rlim_max > maxdsiz) 551 limp->rlim_max = maxdsiz; 552 break; 553 554 case RLIMIT_STACK: 555 if (limp->rlim_cur > maxssiz) 556 limp->rlim_cur = maxssiz; 557 if (limp->rlim_max > maxssiz) 558 limp->rlim_max = maxssiz; 559 /* 560 * Stack is allocated to the max at exec time with only 561 * "rlim_cur" bytes accessible. If stack limit is going 562 * up make more accessible, if going down make inaccessible. 563 */ 564 if (limp->rlim_cur != alimp->rlim_cur) { 565 vm_offset_t addr; 566 vm_size_t size; 567 vm_prot_t prot; 568 569 if (limp->rlim_cur > alimp->rlim_cur) { 570 prot = VM_PROT_ALL; 571 size = limp->rlim_cur - alimp->rlim_cur; 572 addr = USRSTACK - limp->rlim_cur; 573 } else { 574 prot = VM_PROT_NONE; 575 size = alimp->rlim_cur - limp->rlim_cur; 576 addr = USRSTACK - alimp->rlim_cur; 577 } 578 addr = trunc_page(addr); 579 size = round_page(size); 580 (void) vm_map_protect(&p->p_vmspace->vm_map, 581 addr, addr+size, prot, FALSE); 582 } 583 break; 584 585 case RLIMIT_NOFILE: 586 if (limp->rlim_cur > maxfilesperproc) 587 limp->rlim_cur = maxfilesperproc; 588 if (limp->rlim_max > maxfilesperproc) 589 limp->rlim_max = maxfilesperproc; 590 break; 591 592 case RLIMIT_NPROC: 593 if (limp->rlim_cur > maxprocperuid) 594 limp->rlim_cur = maxprocperuid; 595 if (limp->rlim_max > maxprocperuid) 596 limp->rlim_max = maxprocperuid; 597 if (limp->rlim_cur < 1) 598 limp->rlim_cur = 1; 599 if (limp->rlim_max < 1) 600 limp->rlim_max = 1; 601 break; 602 } 603 *alimp = *limp; 604 return (0); 605} 606 607#ifndef _SYS_SYSPROTO_H_ 608struct __getrlimit_args { 609 u_int which; 610 struct rlimit *rlp; 611}; 612#endif 613/* 614 * MPSAFE 615 */ 616/* ARGSUSED */ 617int 618getrlimit(td, uap) 619 struct thread *td; 620 register struct __getrlimit_args *uap; 621{ 622 int error; 623 struct proc *p = td->td_proc; 624 625 if (uap->which >= RLIM_NLIMITS) 626 return (EINVAL); 627 mtx_lock(&Giant); 628 error = copyout((caddr_t)&p->p_rlimit[uap->which], (caddr_t)uap->rlp, 629 sizeof (struct rlimit)); 630 mtx_unlock(&Giant); 631 return(error); 632} 633 634/* 635 * Transform the running time and tick information in proc p into user, 636 * system, and interrupt time usage. 637 */ 638void 639calcru(p, up, sp, ip) 640 struct proc *p; 641 struct timeval *up; 642 struct timeval *sp; 643 struct timeval *ip; 644{ 645 /* {user, system, interrupt, total} {ticks, usec}; previous tu: */ 646 u_int64_t ut, uu, st, su, it, iu, tt, tu, ptu; 647 u_int64_t uut = 0, sut = 0, iut = 0; 648 int s; 649 struct timeval tv; 650 struct bintime bt; 651 struct kse *ke; 652 struct ksegrp *kg; 653 654 mtx_assert(&sched_lock, MA_OWNED); 655 /* XXX: why spl-protect ? worst case is an off-by-one report */ 656 657 FOREACH_KSEGRP_IN_PROC(p, kg) { 658 /* we could accumulate per ksegrp and per process here*/ 659 FOREACH_KSE_IN_GROUP(kg, ke) { 660 s = splstatclock(); 661 ut = ke->ke_uticks; 662 st = ke->ke_sticks; 663 it = ke->ke_iticks; 664 splx(s); 665 666 tt = ut + st + it; 667 if (tt == 0) { 668 st = 1; 669 tt = 1; 670 } 671 672 if (ke == curthread->td_kse) { 673 /* 674 * Adjust for the current time slice. This is actually fairly 675 * important since the error here is on the order of a time 676 * quantum, which is much greater than the sampling error. 677 * XXXKSE use a different test due to threads on other 678 * processors also being 'current'. 679 */ 680 681 binuptime(&bt); 682 bintime_sub(&bt, PCPU_PTR(switchtime)); 683 bintime_add(&bt, &p->p_runtime); 684 } else { 685 bt = p->p_runtime; 686 } 687 bintime2timeval(&bt, &tv); 688 tu = (u_int64_t)tv.tv_sec * 1000000 + tv.tv_usec; 689 ptu = ke->ke_uu + ke->ke_su + ke->ke_iu; 690 if (tu < ptu || (int64_t)tu < 0) { 691 /* XXX no %qd in kernel. Truncate. */ 692 printf("calcru: negative time of %ld usec for pid %d (%s)\n", 693 (long)tu, p->p_pid, p->p_comm); 694 tu = ptu; 695 } 696 697 /* Subdivide tu. */ 698 uu = (tu * ut) / tt; 699 su = (tu * st) / tt; 700 iu = tu - uu - su; 701 702 /* Enforce monotonicity. */ 703 if (uu < ke->ke_uu || su < ke->ke_su || iu < ke->ke_iu) { 704 if (uu < ke->ke_uu) 705 uu = ke->ke_uu; 706 else if (uu + ke->ke_su + ke->ke_iu > tu) 707 uu = tu - ke->ke_su - ke->ke_iu; 708 if (st == 0) 709 su = ke->ke_su; 710 else { 711 su = ((tu - uu) * st) / (st + it); 712 if (su < ke->ke_su) 713 su = ke->ke_su; 714 else if (uu + su + ke->ke_iu > tu) 715 su = tu - uu - ke->ke_iu; 716 } 717 KASSERT(uu + su + ke->ke_iu <= tu, 718 ("calcru: monotonisation botch 1")); 719 iu = tu - uu - su; 720 KASSERT(iu >= ke->ke_iu, 721 ("calcru: monotonisation botch 2")); 722 } 723 ke->ke_uu = uu; 724 ke->ke_su = su; 725 ke->ke_iu = iu; 726 uut += uu; 727 sut += su; 728 iut += iu; 729 730 } /* end kse loop */ 731 } /* end kseg loop */ 732 up->tv_sec = uut / 1000000; 733 up->tv_usec = uut % 1000000; 734 sp->tv_sec = sut / 1000000; 735 sp->tv_usec = sut % 1000000; 736 if (ip != NULL) { 737 ip->tv_sec = iut / 1000000; 738 ip->tv_usec = iut % 1000000; 739 } 740} 741 742#ifndef _SYS_SYSPROTO_H_ 743struct getrusage_args { 744 int who; 745 struct rusage *rusage; 746}; 747#endif 748/* 749 * MPSAFE 750 */ 751/* ARGSUSED */ 752int 753getrusage(td, uap) 754 register struct thread *td; 755 register struct getrusage_args *uap; 756{ 757 struct proc *p = td->td_proc; 758 register struct rusage *rup; 759 int error = 0; 760 761 mtx_lock(&Giant); 762 763 switch (uap->who) { 764 case RUSAGE_SELF: 765 rup = &p->p_stats->p_ru; 766 mtx_lock_spin(&sched_lock); 767 calcru(p, &rup->ru_utime, &rup->ru_stime, NULL); 768 mtx_unlock_spin(&sched_lock); 769 break; 770 771 case RUSAGE_CHILDREN: 772 rup = &p->p_stats->p_cru; 773 break; 774 775 default: 776 rup = NULL; 777 error = EINVAL; 778 break; 779 } 780 mtx_unlock(&Giant); 781 if (error == 0) { 782 error = copyout((caddr_t)rup, (caddr_t)uap->rusage, 783 sizeof (struct rusage)); 784 } 785 return(error); 786} 787 788void 789ruadd(ru, ru2) 790 register struct rusage *ru, *ru2; 791{ 792 register long *ip, *ip2; 793 register int i; 794 795 timevaladd(&ru->ru_utime, &ru2->ru_utime); 796 timevaladd(&ru->ru_stime, &ru2->ru_stime); 797 if (ru->ru_maxrss < ru2->ru_maxrss) 798 ru->ru_maxrss = ru2->ru_maxrss; 799 ip = &ru->ru_first; ip2 = &ru2->ru_first; 800 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 801 *ip++ += *ip2++; 802} 803 804/* 805 * Make a copy of the plimit structure. 806 * We share these structures copy-on-write after fork, 807 * and copy when a limit is changed. 808 */ 809struct plimit * 810limcopy(lim) 811 struct plimit *lim; 812{ 813 register struct plimit *copy; 814 815 MALLOC(copy, struct plimit *, sizeof(struct plimit), 816 M_SUBPROC, M_WAITOK); 817 bcopy(lim->pl_rlimit, copy->pl_rlimit, sizeof(struct plimit)); 818 copy->p_lflags = 0; 819 copy->p_refcnt = 1; 820 return (copy); 821} 822 823/* 824 * Find the uidinfo structure for a uid. This structure is used to 825 * track the total resource consumption (process count, socket buffer 826 * size, etc.) for the uid and impose limits. 827 */ 828void 829uihashinit() 830{ 831 832 uihashtbl = hashinit(maxproc / 16, M_UIDINFO, &uihash); 833 mtx_init(&uihashtbl_mtx, "uidinfo hash", NULL, MTX_DEF); 834} 835 836/* 837 * lookup a uidinfo struct for the parameter uid. 838 * uihashtbl_mtx must be locked. 839 */ 840static struct uidinfo * 841uilookup(uid) 842 uid_t uid; 843{ 844 struct uihashhead *uipp; 845 struct uidinfo *uip; 846 847 mtx_assert(&uihashtbl_mtx, MA_OWNED); 848 uipp = UIHASH(uid); 849 LIST_FOREACH(uip, uipp, ui_hash) 850 if (uip->ui_uid == uid) 851 break; 852 853 return (uip); 854} 855 856/* 857 * Find or allocate a struct uidinfo for a particular uid. 858 * Increase refcount on uidinfo struct returned. 859 * uifree() should be called on a struct uidinfo when released. 860 */ 861struct uidinfo * 862uifind(uid) 863 uid_t uid; 864{ 865 struct uidinfo *uip; 866 867 mtx_lock(&uihashtbl_mtx); 868 uip = uilookup(uid); 869 if (uip == NULL) { 870 struct uidinfo *old_uip; 871 872 mtx_unlock(&uihashtbl_mtx); 873 uip = malloc(sizeof(*uip), M_UIDINFO, M_WAITOK | M_ZERO); 874 mtx_lock(&uihashtbl_mtx); 875 /* 876 * There's a chance someone created our uidinfo while we 877 * were in malloc and not holding the lock, so we have to 878 * make sure we don't insert a duplicate uidinfo 879 */ 880 if ((old_uip = uilookup(uid)) != NULL) { 881 /* someone else beat us to it */ 882 free(uip, M_UIDINFO); 883 uip = old_uip; 884 } else { 885 uip->ui_mtxp = mtx_pool_alloc(); 886 uip->ui_uid = uid; 887 LIST_INSERT_HEAD(UIHASH(uid), uip, ui_hash); 888 } 889 } 890 uihold(uip); 891 mtx_unlock(&uihashtbl_mtx); 892 return (uip); 893} 894 895/* 896 * Place another refcount on a uidinfo struct. 897 */ 898void 899uihold(uip) 900 struct uidinfo *uip; 901{ 902 903 UIDINFO_LOCK(uip); 904 uip->ui_ref++; 905 UIDINFO_UNLOCK(uip); 906} 907 908/*- 909 * Since uidinfo structs have a long lifetime, we use an 910 * opportunistic refcounting scheme to avoid locking the lookup hash 911 * for each release. 912 * 913 * If the refcount hits 0, we need to free the structure, 914 * which means we need to lock the hash. 915 * Optimal case: 916 * After locking the struct and lowering the refcount, if we find 917 * that we don't need to free, simply unlock and return. 918 * Suboptimal case: 919 * If refcount lowering results in need to free, bump the count 920 * back up, loose the lock and aquire the locks in the proper 921 * order to try again. 922 */ 923void 924uifree(uip) 925 struct uidinfo *uip; 926{ 927 928 /* Prepare for optimal case. */ 929 UIDINFO_LOCK(uip); 930 931 if (--uip->ui_ref != 0) { 932 UIDINFO_UNLOCK(uip); 933 return; 934 } 935 936 /* Prepare for suboptimal case. */ 937 uip->ui_ref++; 938 UIDINFO_UNLOCK(uip); 939 mtx_lock(&uihashtbl_mtx); 940 UIDINFO_LOCK(uip); 941 942 /* 943 * We must subtract one from the count again because we backed out 944 * our initial subtraction before dropping the lock. 945 * Since another thread may have added a reference after we dropped the 946 * initial lock we have to test for zero again. 947 */ 948 if (--uip->ui_ref == 0) { 949 LIST_REMOVE(uip, ui_hash); 950 mtx_unlock(&uihashtbl_mtx); 951 if (uip->ui_sbsize != 0) 952 /* XXX no %qd in kernel. Truncate. */ 953 printf("freeing uidinfo: uid = %d, sbsize = %ld\n", 954 uip->ui_uid, (long)uip->ui_sbsize); 955 if (uip->ui_proccnt != 0) 956 printf("freeing uidinfo: uid = %d, proccnt = %ld\n", 957 uip->ui_uid, uip->ui_proccnt); 958 UIDINFO_UNLOCK(uip); 959 FREE(uip, M_UIDINFO); 960 return; 961 } 962 963 mtx_unlock(&uihashtbl_mtx); 964 UIDINFO_UNLOCK(uip); 965} 966 967/* 968 * Change the count associated with number of processes 969 * a given user is using. When 'max' is 0, don't enforce a limit 970 */ 971int 972chgproccnt(uip, diff, max) 973 struct uidinfo *uip; 974 int diff; 975 int max; 976{ 977 978 UIDINFO_LOCK(uip); 979 /* don't allow them to exceed max, but allow subtraction */ 980 if (diff > 0 && uip->ui_proccnt + diff > max && max != 0) { 981 UIDINFO_UNLOCK(uip); 982 return (0); 983 } 984 uip->ui_proccnt += diff; 985 if (uip->ui_proccnt < 0) 986 printf("negative proccnt for uid = %d\n", uip->ui_uid); 987 UIDINFO_UNLOCK(uip); 988 return (1); 989} 990 991/* 992 * Change the total socket buffer size a user has used. 993 */ 994int 995chgsbsize(uip, hiwat, to, max) 996 struct uidinfo *uip; 997 u_long *hiwat; 998 u_long to; 999 rlim_t max; 1000{ 1001 rlim_t new; 1002 int s; 1003 1004 s = splnet(); 1005 UIDINFO_LOCK(uip); 1006 new = uip->ui_sbsize + to - *hiwat; 1007 /* don't allow them to exceed max, but allow subtraction */ 1008 if (to > *hiwat && new > max) { 1009 splx(s); 1010 UIDINFO_UNLOCK(uip); 1011 return (0); 1012 } 1013 uip->ui_sbsize = new; 1014 *hiwat = to; 1015 if (uip->ui_sbsize < 0) 1016 printf("negative sbsize for uid = %d\n", uip->ui_uid); 1017 splx(s); 1018 UIDINFO_UNLOCK(uip); 1019 return (1); 1020} 1021