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 * $FreeBSD$ 40 */ 41 42 #include "opt_compat.h" 43 #include "opt_rlimit.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/sysproto.h> 48 #include <sys/file.h> 49 #include <sys/kernel.h> 50 #include <sys/lock.h> 51 #include <sys/malloc.h> 52 #include <sys/mutex.h> 53 #include <sys/proc.h> 54 #include <sys/resourcevar.h> 55 #include <sys/sx.h> 56 #include <sys/time.h> 57 58 #include <vm/vm.h> 59 #include <vm/vm_param.h> 60 #include <vm/pmap.h> 61 #include <vm/vm_map.h> 62 63 static int donice __P((struct proc *curp, struct proc *chgp, int n)); 64 /* dosetrlimit non-static: Needed by SysVR4 emulator */ 65 int dosetrlimit __P((struct proc *p, u_int which, struct rlimit *limp)); 66 67 static MALLOC_DEFINE(M_UIDINFO, "uidinfo", "uidinfo structures"); 68 #define UIHASH(uid) (&uihashtbl[(uid) & uihash]) 69 static struct mtx uihashtbl_mtx; 70 static LIST_HEAD(uihashhead, uidinfo) *uihashtbl; 71 static u_long uihash; /* size of hash table - 1 */ 72 73 static struct uidinfo *uilookup __P((uid_t uid)); 74 75 /* 76 * Resource controls and accounting. 77 */ 78 79 #ifndef _SYS_SYSPROTO_H_ 80 struct getpriority_args { 81 int which; 82 int who; 83 }; 84 #endif 85 int 86 getpriority(curp, uap) 87 struct proc *curp; 88 register struct getpriority_args *uap; 89 { 90 register struct proc *p; 91 register int low = PRIO_MAX + 1; 92 93 switch (uap->which) { 94 95 case PRIO_PROCESS: 96 if (uap->who == 0) 97 low = curp->p_nice; 98 else { 99 p = pfind(uap->who); 100 if (p == NULL) 101 break; 102 if (p_can(curp, p, P_CAN_SEE, NULL) == 0) 103 low = p->p_nice; 104 PROC_UNLOCK(p); 105 } 106 break; 107 108 case PRIO_PGRP: { 109 register struct pgrp *pg; 110 111 if (uap->who == 0) 112 pg = curp->p_pgrp; 113 else if ((pg = pgfind(uap->who)) == NULL) 114 break; 115 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 116 if (!p_can(curp, p, P_CAN_SEE, NULL) && p->p_nice < low) 117 low = p->p_nice; 118 } 119 break; 120 } 121 122 case PRIO_USER: 123 if (uap->who == 0) 124 uap->who = curp->p_ucred->cr_uid; 125 sx_slock(&allproc_lock); 126 LIST_FOREACH(p, &allproc, p_list) 127 if (!p_can(curp, p, P_CAN_SEE, NULL) && 128 p->p_ucred->cr_uid == uap->who && 129 p->p_nice < low) 130 low = p->p_nice; 131 sx_sunlock(&allproc_lock); 132 break; 133 134 default: 135 return (EINVAL); 136 } 137 if (low == PRIO_MAX + 1) 138 return (ESRCH); 139 curp->p_retval[0] = low; 140 return (0); 141 } 142 143 #ifndef _SYS_SYSPROTO_H_ 144 struct setpriority_args { 145 int which; 146 int who; 147 int prio; 148 }; 149 #endif 150 /* ARGSUSED */ 151 int 152 setpriority(curp, uap) 153 struct proc *curp; 154 register struct setpriority_args *uap; 155 { 156 register struct proc *p; 157 int found = 0, error = 0; 158 159 switch (uap->which) { 160 161 case PRIO_PROCESS: 162 if (uap->who == 0) 163 error = donice(curp, curp, uap->prio); 164 else { 165 p = pfind(uap->who); 166 if (p == 0) 167 break; 168 if (p_can(curp, p, P_CAN_SEE, NULL) == 0) 169 error = donice(curp, p, uap->prio); 170 PROC_UNLOCK(p); 171 } 172 found++; 173 break; 174 175 case PRIO_PGRP: { 176 register struct pgrp *pg; 177 178 if (uap->who == 0) 179 pg = curp->p_pgrp; 180 else if ((pg = pgfind(uap->who)) == NULL) 181 break; 182 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 183 if (!p_can(curp, p, P_CAN_SEE, NULL)) { 184 error = donice(curp, p, uap->prio); 185 found++; 186 } 187 } 188 break; 189 } 190 191 case PRIO_USER: 192 if (uap->who == 0) 193 uap->who = curp->p_ucred->cr_uid; 194 sx_slock(&allproc_lock); 195 LIST_FOREACH(p, &allproc, p_list) 196 if (p->p_ucred->cr_uid == uap->who && 197 !p_can(curp, p, P_CAN_SEE, NULL)) { 198 error = donice(curp, p, uap->prio); 199 found++; 200 } 201 sx_sunlock(&allproc_lock); 202 break; 203 204 default: 205 return (EINVAL); 206 } 207 if (found == 0) 208 return (ESRCH); 209 return (error); 210 } 211 212 static int 213 donice(curp, chgp, n) 214 register struct proc *curp, *chgp; 215 register int n; 216 { 217 int error; 218 219 if ((error = p_can(curp, chgp, P_CAN_SCHED, NULL))) 220 return (error); 221 if (n > PRIO_MAX) 222 n = PRIO_MAX; 223 if (n < PRIO_MIN) 224 n = PRIO_MIN; 225 if (n < chgp->p_nice && suser(curp)) 226 return (EACCES); 227 chgp->p_nice = n; 228 (void)resetpriority(chgp); 229 return (0); 230 } 231 232 /* rtprio system call */ 233 #ifndef _SYS_SYSPROTO_H_ 234 struct rtprio_args { 235 int function; 236 pid_t pid; 237 struct rtprio *rtp; 238 }; 239 #endif 240 241 /* 242 * Set realtime priority 243 */ 244 245 /* ARGSUSED */ 246 int 247 rtprio(curp, uap) 248 struct proc *curp; 249 register struct rtprio_args *uap; 250 { 251 register struct proc *p; 252 struct rtprio rtp; 253 int error; 254 255 if (uap->pid == 0) { 256 p = curp; 257 PROC_LOCK(p); 258 } else 259 p = pfind(uap->pid); 260 261 if (p == NULL) 262 return (ESRCH); 263 264 switch (uap->function) { 265 case RTP_LOOKUP: 266 if ((error = p_can(curp, p, P_CAN_SEE, NULL))) 267 break; 268 pri_to_rtp(&p->p_pri, &rtp); 269 error = copyout(&rtp, uap->rtp, sizeof(struct rtprio)); 270 break; 271 case RTP_SET: 272 if ((error = p_can(curp, p, P_CAN_SCHED, NULL)) || 273 (error = copyin(uap->rtp, &rtp, sizeof(struct rtprio)))) 274 break; 275 /* disallow setting rtprio in most cases if not superuser */ 276 if (suser(curp) != 0) { 277 /* can't set someone else's */ 278 if (uap->pid) { 279 error = EPERM; 280 break; 281 } 282 /* can't set realtime priority */ 283 /* 284 * Realtime priority has to be restricted for reasons which should be 285 * obvious. However, for idle priority, there is a potential for 286 * system deadlock if an idleprio process gains a lock on a resource 287 * that other processes need (and the idleprio process can't run 288 * due to a CPU-bound normal process). Fix me! XXX 289 */ 290 #if 0 291 if (RTP_PRIO_IS_REALTIME(rtp.type)) 292 #endif 293 if (rtp.type != RTP_PRIO_NORMAL) { 294 error = EPERM; 295 break; 296 } 297 } 298 error = rtp_to_pri(&rtp, &p->p_pri); 299 break; 300 default: 301 error = EINVAL; 302 break; 303 } 304 PROC_UNLOCK(p); 305 return (error); 306 } 307 308 int 309 rtp_to_pri(struct rtprio *rtp, struct priority *pri) 310 { 311 312 if (rtp->prio > RTP_PRIO_MAX) 313 return (EINVAL); 314 switch (RTP_PRIO_BASE(rtp->type)) { 315 case RTP_PRIO_REALTIME: 316 pri->pri_level = PRI_MIN_REALTIME + rtp->prio; 317 break; 318 case RTP_PRIO_NORMAL: 319 pri->pri_level = PRI_MIN_TIMESHARE + rtp->prio; 320 break; 321 case RTP_PRIO_IDLE: 322 pri->pri_level = PRI_MIN_IDLE + rtp->prio; 323 break; 324 default: 325 return (EINVAL); 326 } 327 pri->pri_class = rtp->type; 328 pri->pri_native = pri->pri_level; 329 pri->pri_user = pri->pri_level; 330 return (0); 331 } 332 333 void 334 pri_to_rtp(struct priority *pri, struct rtprio *rtp) 335 { 336 337 switch (PRI_BASE(pri->pri_class)) { 338 case PRI_REALTIME: 339 rtp->prio = pri->pri_level - PRI_MIN_REALTIME; 340 break; 341 case PRI_TIMESHARE: 342 rtp->prio = pri->pri_level - PRI_MIN_TIMESHARE; 343 break; 344 case PRI_IDLE: 345 rtp->prio = pri->pri_level - PRI_MIN_IDLE; 346 break; 347 default: 348 break; 349 } 350 rtp->type = pri->pri_class; 351 } 352 353 #if defined(COMPAT_43) || defined(COMPAT_SUNOS) 354 #ifndef _SYS_SYSPROTO_H_ 355 struct osetrlimit_args { 356 u_int which; 357 struct orlimit *rlp; 358 }; 359 #endif 360 /* ARGSUSED */ 361 int 362 osetrlimit(p, uap) 363 struct proc *p; 364 register struct osetrlimit_args *uap; 365 { 366 struct orlimit olim; 367 struct rlimit lim; 368 int error; 369 370 if ((error = 371 copyin((caddr_t)uap->rlp, (caddr_t)&olim, sizeof(struct orlimit)))) 372 return (error); 373 lim.rlim_cur = olim.rlim_cur; 374 lim.rlim_max = olim.rlim_max; 375 return (dosetrlimit(p, uap->which, &lim)); 376 } 377 378 #ifndef _SYS_SYSPROTO_H_ 379 struct ogetrlimit_args { 380 u_int which; 381 struct orlimit *rlp; 382 }; 383 #endif 384 /* ARGSUSED */ 385 int 386 ogetrlimit(p, uap) 387 struct proc *p; 388 register struct ogetrlimit_args *uap; 389 { 390 struct orlimit olim; 391 392 if (uap->which >= RLIM_NLIMITS) 393 return (EINVAL); 394 olim.rlim_cur = p->p_rlimit[uap->which].rlim_cur; 395 if (olim.rlim_cur == -1) 396 olim.rlim_cur = 0x7fffffff; 397 olim.rlim_max = p->p_rlimit[uap->which].rlim_max; 398 if (olim.rlim_max == -1) 399 olim.rlim_max = 0x7fffffff; 400 return (copyout((caddr_t)&olim, (caddr_t)uap->rlp, sizeof(olim))); 401 } 402 #endif /* COMPAT_43 || COMPAT_SUNOS */ 403 404 #ifndef _SYS_SYSPROTO_H_ 405 struct __setrlimit_args { 406 u_int which; 407 struct rlimit *rlp; 408 }; 409 #endif 410 /* ARGSUSED */ 411 int 412 setrlimit(p, uap) 413 struct proc *p; 414 register struct __setrlimit_args *uap; 415 { 416 struct rlimit alim; 417 int error; 418 419 if ((error = 420 copyin((caddr_t)uap->rlp, (caddr_t)&alim, sizeof (struct rlimit)))) 421 return (error); 422 return (dosetrlimit(p, uap->which, &alim)); 423 } 424 425 int 426 dosetrlimit(p, which, limp) 427 struct proc *p; 428 u_int which; 429 struct rlimit *limp; 430 { 431 register struct rlimit *alimp; 432 int error; 433 434 if (which >= RLIM_NLIMITS) 435 return (EINVAL); 436 alimp = &p->p_rlimit[which]; 437 438 /* 439 * Preserve historical bugs by treating negative limits as unsigned. 440 */ 441 if (limp->rlim_cur < 0) 442 limp->rlim_cur = RLIM_INFINITY; 443 if (limp->rlim_max < 0) 444 limp->rlim_max = RLIM_INFINITY; 445 446 if (limp->rlim_cur > alimp->rlim_max || 447 limp->rlim_max > alimp->rlim_max) 448 if ((error = suser_xxx(0, p, PRISON_ROOT))) 449 return (error); 450 if (limp->rlim_cur > limp->rlim_max) 451 limp->rlim_cur = limp->rlim_max; 452 if (p->p_limit->p_refcnt > 1 && 453 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 454 p->p_limit->p_refcnt--; 455 p->p_limit = limcopy(p->p_limit); 456 alimp = &p->p_rlimit[which]; 457 } 458 459 switch (which) { 460 461 case RLIMIT_CPU: 462 if (limp->rlim_cur > RLIM_INFINITY / (rlim_t)1000000) 463 p->p_limit->p_cpulimit = RLIM_INFINITY; 464 else 465 p->p_limit->p_cpulimit = 466 (rlim_t)1000000 * limp->rlim_cur; 467 break; 468 case RLIMIT_DATA: 469 if (limp->rlim_cur > MAXDSIZ) 470 limp->rlim_cur = MAXDSIZ; 471 if (limp->rlim_max > MAXDSIZ) 472 limp->rlim_max = MAXDSIZ; 473 break; 474 475 case RLIMIT_STACK: 476 if (limp->rlim_cur > MAXSSIZ) 477 limp->rlim_cur = MAXSSIZ; 478 if (limp->rlim_max > MAXSSIZ) 479 limp->rlim_max = MAXSSIZ; 480 /* 481 * Stack is allocated to the max at exec time with only 482 * "rlim_cur" bytes accessible. If stack limit is going 483 * up make more accessible, if going down make inaccessible. 484 */ 485 if (limp->rlim_cur != alimp->rlim_cur) { 486 vm_offset_t addr; 487 vm_size_t size; 488 vm_prot_t prot; 489 490 if (limp->rlim_cur > alimp->rlim_cur) { 491 prot = VM_PROT_ALL; 492 size = limp->rlim_cur - alimp->rlim_cur; 493 addr = USRSTACK - limp->rlim_cur; 494 } else { 495 prot = VM_PROT_NONE; 496 size = alimp->rlim_cur - limp->rlim_cur; 497 addr = USRSTACK - alimp->rlim_cur; 498 } 499 addr = trunc_page(addr); 500 size = round_page(size); 501 (void) vm_map_protect(&p->p_vmspace->vm_map, 502 addr, addr+size, prot, FALSE); 503 } 504 break; 505 506 case RLIMIT_NOFILE: 507 if (limp->rlim_cur > maxfilesperproc) 508 limp->rlim_cur = maxfilesperproc; 509 if (limp->rlim_max > maxfilesperproc) 510 limp->rlim_max = maxfilesperproc; 511 break; 512 513 case RLIMIT_NPROC: 514 if (limp->rlim_cur > maxprocperuid) 515 limp->rlim_cur = maxprocperuid; 516 if (limp->rlim_max > maxprocperuid) 517 limp->rlim_max = maxprocperuid; 518 if (limp->rlim_cur < 1) 519 limp->rlim_cur = 1; 520 if (limp->rlim_max < 1) 521 limp->rlim_max = 1; 522 break; 523 } 524 *alimp = *limp; 525 return (0); 526 } 527 528 #ifndef _SYS_SYSPROTO_H_ 529 struct __getrlimit_args { 530 u_int which; 531 struct rlimit *rlp; 532 }; 533 #endif 534 /* ARGSUSED */ 535 int 536 getrlimit(p, uap) 537 struct proc *p; 538 register struct __getrlimit_args *uap; 539 { 540 541 if (uap->which >= RLIM_NLIMITS) 542 return (EINVAL); 543 return (copyout((caddr_t)&p->p_rlimit[uap->which], (caddr_t)uap->rlp, 544 sizeof (struct rlimit))); 545 } 546 547 /* 548 * Transform the running time and tick information in proc p into user, 549 * system, and interrupt time usage. 550 */ 551 void 552 calcru(p, up, sp, ip) 553 struct proc *p; 554 struct timeval *up; 555 struct timeval *sp; 556 struct timeval *ip; 557 { 558 /* {user, system, interrupt, total} {ticks, usec}; previous tu: */ 559 u_int64_t ut, uu, st, su, it, iu, tt, tu, ptu; 560 int s; 561 struct timeval tv; 562 563 mtx_assert(&sched_lock, MA_OWNED); 564 /* XXX: why spl-protect ? worst case is an off-by-one report */ 565 s = splstatclock(); 566 ut = p->p_uticks; 567 st = p->p_sticks; 568 it = p->p_iticks; 569 splx(s); 570 571 tt = ut + st + it; 572 if (tt == 0) { 573 st = 1; 574 tt = 1; 575 } 576 577 tu = p->p_runtime; 578 if (p == curproc) { 579 /* 580 * Adjust for the current time slice. This is actually fairly 581 * important since the error here is on the order of a time 582 * quantum, which is much greater than the sampling error. 583 */ 584 microuptime(&tv); 585 if (timevalcmp(&tv, PCPU_PTR(switchtime), <)) 586 printf("microuptime() went backwards (%ld.%06ld -> %ld.%06ld)\n", 587 PCPU_GET(switchtime.tv_sec), PCPU_GET(switchtime.tv_usec), 588 tv.tv_sec, tv.tv_usec); 589 else 590 tu += (tv.tv_usec - PCPU_GET(switchtime.tv_usec)) + 591 (tv.tv_sec - PCPU_GET(switchtime.tv_sec)) * 592 (int64_t)1000000; 593 } 594 ptu = p->p_uu + p->p_su + p->p_iu; 595 if (tu < ptu || (int64_t)tu < 0) { 596 /* XXX no %qd in kernel. Truncate. */ 597 printf("calcru: negative time of %ld usec for pid %d (%s)\n", 598 (long)tu, p->p_pid, p->p_comm); 599 tu = ptu; 600 } 601 602 /* Subdivide tu. */ 603 uu = (tu * ut) / tt; 604 su = (tu * st) / tt; 605 iu = tu - uu - su; 606 607 /* Enforce monotonicity. */ 608 if (uu < p->p_uu || su < p->p_su || iu < p->p_iu) { 609 if (uu < p->p_uu) 610 uu = p->p_uu; 611 else if (uu + p->p_su + p->p_iu > tu) 612 uu = tu - p->p_su - p->p_iu; 613 if (st == 0) 614 su = p->p_su; 615 else { 616 su = ((tu - uu) * st) / (st + it); 617 if (su < p->p_su) 618 su = p->p_su; 619 else if (uu + su + p->p_iu > tu) 620 su = tu - uu - p->p_iu; 621 } 622 KASSERT(uu + su + p->p_iu <= tu, 623 ("calcru: monotonisation botch 1")); 624 iu = tu - uu - su; 625 KASSERT(iu >= p->p_iu, 626 ("calcru: monotonisation botch 2")); 627 } 628 p->p_uu = uu; 629 p->p_su = su; 630 p->p_iu = iu; 631 632 up->tv_sec = uu / 1000000; 633 up->tv_usec = uu % 1000000; 634 sp->tv_sec = su / 1000000; 635 sp->tv_usec = su % 1000000; 636 if (ip != NULL) { 637 ip->tv_sec = iu / 1000000; 638 ip->tv_usec = iu % 1000000; 639 } 640 } 641 642 #ifndef _SYS_SYSPROTO_H_ 643 struct getrusage_args { 644 int who; 645 struct rusage *rusage; 646 }; 647 #endif 648 /* ARGSUSED */ 649 int 650 getrusage(p, uap) 651 register struct proc *p; 652 register struct getrusage_args *uap; 653 { 654 register struct rusage *rup; 655 656 switch (uap->who) { 657 658 case RUSAGE_SELF: 659 rup = &p->p_stats->p_ru; 660 mtx_lock_spin(&sched_lock); 661 calcru(p, &rup->ru_utime, &rup->ru_stime, NULL); 662 mtx_unlock_spin(&sched_lock); 663 break; 664 665 case RUSAGE_CHILDREN: 666 rup = &p->p_stats->p_cru; 667 break; 668 669 default: 670 return (EINVAL); 671 } 672 return (copyout((caddr_t)rup, (caddr_t)uap->rusage, 673 sizeof (struct rusage))); 674 } 675 676 void 677 ruadd(ru, ru2) 678 register struct rusage *ru, *ru2; 679 { 680 register long *ip, *ip2; 681 register int i; 682 683 timevaladd(&ru->ru_utime, &ru2->ru_utime); 684 timevaladd(&ru->ru_stime, &ru2->ru_stime); 685 if (ru->ru_maxrss < ru2->ru_maxrss) 686 ru->ru_maxrss = ru2->ru_maxrss; 687 ip = &ru->ru_first; ip2 = &ru2->ru_first; 688 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) 689 *ip++ += *ip2++; 690 } 691 692 /* 693 * Make a copy of the plimit structure. 694 * We share these structures copy-on-write after fork, 695 * and copy when a limit is changed. 696 */ 697 struct plimit * 698 limcopy(lim) 699 struct plimit *lim; 700 { 701 register struct plimit *copy; 702 703 MALLOC(copy, struct plimit *, sizeof(struct plimit), 704 M_SUBPROC, M_WAITOK); 705 bcopy(lim->pl_rlimit, copy->pl_rlimit, sizeof(struct plimit)); 706 copy->p_lflags = 0; 707 copy->p_refcnt = 1; 708 return (copy); 709 } 710 711 /* 712 * Find the uidinfo structure for a uid. This structure is used to 713 * track the total resource consumption (process count, socket buffer 714 * size, etc.) for the uid and impose limits. 715 */ 716 void 717 uihashinit() 718 { 719 720 uihashtbl = hashinit(maxproc / 16, M_UIDINFO, &uihash); 721 mtx_init(&uihashtbl_mtx, "uidinfo hash", MTX_DEF); 722 } 723 724 /* 725 * lookup a uidinfo struct for the parameter uid. 726 * uihashtbl_mtx must be locked. 727 */ 728 static struct uidinfo * 729 uilookup(uid) 730 uid_t uid; 731 { 732 struct uihashhead *uipp; 733 struct uidinfo *uip; 734 735 mtx_assert(&uihashtbl_mtx, MA_OWNED); 736 uipp = UIHASH(uid); 737 LIST_FOREACH(uip, uipp, ui_hash) 738 if (uip->ui_uid == uid) 739 break; 740 741 return (uip); 742 } 743 744 /* 745 * Find or allocate a struct uidinfo for a particular uid. 746 * Increase refcount on uidinfo struct returned. 747 * uifree() should be called on a struct uidinfo when released. 748 */ 749 struct uidinfo * 750 uifind(uid) 751 uid_t uid; 752 { 753 struct uidinfo *uip; 754 755 mtx_lock(&uihashtbl_mtx); 756 uip = uilookup(uid); 757 if (uip == NULL) { 758 struct uidinfo *old_uip; 759 760 mtx_unlock(&uihashtbl_mtx); 761 uip = malloc(sizeof(*uip), M_UIDINFO, M_WAITOK | M_ZERO); 762 mtx_lock(&uihashtbl_mtx); 763 /* 764 * There's a chance someone created our uidinfo while we 765 * were in malloc and not holding the lock, so we have to 766 * make sure we don't insert a duplicate uidinfo 767 */ 768 if ((old_uip = uilookup(uid)) != NULL) { 769 /* someone else beat us to it */ 770 free(uip, M_UIDINFO); 771 uip = old_uip; 772 } else { 773 mtx_init(&uip->ui_mtx, "uidinfo struct", MTX_DEF); 774 uip->ui_uid = uid; 775 LIST_INSERT_HEAD(UIHASH(uid), uip, ui_hash); 776 } 777 } 778 uihold(uip); 779 mtx_unlock(&uihashtbl_mtx); 780 return (uip); 781 } 782 783 /* 784 * Place another refcount on a uidinfo struct. 785 */ 786 void 787 uihold(uip) 788 struct uidinfo *uip; 789 { 790 791 mtx_lock(&uip->ui_mtx); 792 uip->ui_ref++; 793 mtx_unlock(&uip->ui_mtx); 794 } 795 796 /*- 797 * Since uidinfo structs have a long lifetime, we use an 798 * opportunistic refcounting scheme to avoid locking the lookup hash 799 * for each release. 800 * 801 * If the refcount hits 0, we need to free the structure, 802 * which means we need to lock the hash. 803 * Optimal case: 804 * After locking the struct and lowering the refcount, if we find 805 * that we don't need to free, simply unlock and return. 806 * Suboptimal case: 807 * If refcount lowering results in need to free, bump the count 808 * back up, loose the lock and aquire the locks in the proper 809 * order to try again. 810 */ 811 void 812 uifree(uip) 813 struct uidinfo *uip; 814 { 815 816 /* Prepare for optimal case. */ 817 mtx_lock(&uip->ui_mtx); 818 819 if (--uip->ui_ref != 0) { 820 mtx_unlock(&uip->ui_mtx); 821 return; 822 } 823 824 /* Prepare for suboptimal case. */ 825 uip->ui_ref++; 826 mtx_unlock(&uip->ui_mtx); 827 mtx_lock(&uihashtbl_mtx); 828 mtx_lock(&uip->ui_mtx); 829 830 /* 831 * We must subtract one from the count again because we backed out 832 * our initial subtraction before dropping the lock. 833 * Since another thread may have added a reference after we dropped the 834 * initial lock we have to test for zero again. 835 */ 836 if (--uip->ui_ref == 0) { 837 LIST_REMOVE(uip, ui_hash); 838 mtx_unlock(&uihashtbl_mtx); 839 if (uip->ui_sbsize != 0) 840 /* XXX no %qd in kernel. Truncate. */ 841 printf("freeing uidinfo: uid = %d, sbsize = %ld\n", 842 uip->ui_uid, (long)uip->ui_sbsize); 843 if (uip->ui_proccnt != 0) 844 printf("freeing uidinfo: uid = %d, proccnt = %ld\n", 845 uip->ui_uid, uip->ui_proccnt); 846 mtx_destroy(&uip->ui_mtx); 847 FREE(uip, M_UIDINFO); 848 return; 849 } 850 851 mtx_unlock(&uihashtbl_mtx); 852 mtx_unlock(&uip->ui_mtx); 853 } 854 855 /* 856 * Change the count associated with number of processes 857 * a given user is using. When 'max' is 0, don't enforce a limit 858 */ 859 int 860 chgproccnt(uip, diff, max) 861 struct uidinfo *uip; 862 int diff; 863 int max; 864 { 865 866 mtx_lock(&uip->ui_mtx); 867 /* don't allow them to exceed max, but allow subtraction */ 868 if (diff > 0 && uip->ui_proccnt + diff > max && max != 0) { 869 mtx_unlock(&uip->ui_mtx); 870 return (0); 871 } 872 uip->ui_proccnt += diff; 873 if (uip->ui_proccnt < 0) 874 printf("negative proccnt for uid = %d\n", uip->ui_uid); 875 mtx_unlock(&uip->ui_mtx); 876 return (1); 877 } 878 879 /* 880 * Change the total socket buffer size a user has used. 881 */ 882 int 883 chgsbsize(uip, hiwat, to, max) 884 struct uidinfo *uip; 885 u_long *hiwat; 886 u_long to; 887 rlim_t max; 888 { 889 rlim_t new; 890 int s; 891 892 s = splnet(); 893 mtx_lock(&uip->ui_mtx); 894 new = uip->ui_sbsize + to - *hiwat; 895 /* don't allow them to exceed max, but allow subtraction */ 896 if (to > *hiwat && new > max) { 897 splx(s); 898 mtx_unlock(&uip->ui_mtx); 899 return (0); 900 } 901 uip->ui_sbsize = new; 902 *hiwat = to; 903 if (uip->ui_sbsize < 0) 904 printf("negative sbsize for uid = %d\n", uip->ui_uid); 905 splx(s); 906 mtx_unlock(&uip->ui_mtx); 907 return (1); 908 } 909