sys/kern/kern_resource.c

   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. Neither the name of the University nor the names of its contributors
  19  *    may be used to endorse or promote products derived from this software
  20  *    without specific prior written permission.
  21  *
  22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  *      @(#)kern_resource.c     8.5 (Berkeley) 1/21/94
  35  * $FreeBSD: src/sys/kern/kern_resource.c,v 1.55.2.5 2001/11/03 01:41:08 ps Exp $
  36  */
  37
  38 #include "opt_compat.h"
  39
  40 #include <sys/param.h>
  41 #include <sys/systm.h>
  42 #include <sys/sysproto.h>
  43 #include <sys/file.h>
  44 #include <sys/kern_syscall.h>
  45 #include <sys/kernel.h>
  46 #include <sys/resourcevar.h>
  47 #include <sys/malloc.h>
  48 #include <sys/proc.h>
  49 #include <sys/priv.h>
  50 #include <sys/time.h>
  51 #include <sys/lockf.h>
  52
  53 #include <vm/vm.h>
  54 #include <vm/vm_param.h>
  55 #include <sys/lock.h>
  56 #include <vm/pmap.h>
  57 #include <vm/vm_map.h>
  58
  59 #include <sys/thread2.h>
  60 #include <sys/spinlock2.h>
  61
  62 static int donice (struct proc *chgp, int n);
  63 static int doionice (struct proc *chgp, int n);
  64
  65 static MALLOC_DEFINE(M_UIDINFO, "uidinfo", "uidinfo structures");
  66 #define UIHASH(uid)     (&uihashtbl[(uid) & uihash])
  67 static struct spinlock uihash_lock;
  68 static LIST_HEAD(uihashhead, uidinfo) *uihashtbl;
  69 static u_long uihash;           /* size of hash table - 1 */
  70
  71 static struct uidinfo   *uicreate (uid_t uid);
  72 static struct uidinfo   *uilookup (uid_t uid);
  73
  74 /*
  75  * Resource controls and accounting.
  76  */
  77
  78 struct getpriority_info {
  79         int low;
  80         int who;
  81 };
  82
  83 static int getpriority_callback(struct proc *p, void *data);
  84
  85 /*
  86  * MPALMOSTSAFE
  87  */
  88 int
  89 sys_getpriority(struct getpriority_args *uap)
  90 {
  91         struct getpriority_info info;
  92         thread_t curtd = curthread;
  93         struct proc *curp = curproc;
  94         struct proc *p;
  95         struct pgrp *pg;
  96         int low = PRIO_MAX + 1;
  97         int error;
  98
  99         switch (uap->which) {
 100         case PRIO_PROCESS:
 101                 if (uap->who == 0) {
 102                         low = curp->p_nice;
 103                 } else {
 104                         p = pfind(uap->who);
 105                         if (p) {
 106                                 lwkt_gettoken_shared(&p->p_token);
 107                                 if (PRISON_CHECK(curtd->td_ucred, p->p_ucred))
 108                                         low = p->p_nice;
 109                                 lwkt_reltoken(&p->p_token);
 110                                 PRELE(p);
 111                         }
 112                 }
 113                 break;
 114         case PRIO_PGRP:
 115                 if (uap->who == 0) {
 116                         lwkt_gettoken_shared(&curp->p_token);
 117                         pg = curp->p_pgrp;
 118                         pgref(pg);
 119                         lwkt_reltoken(&curp->p_token);
 120                 } else if ((pg = pgfind(uap->who)) == NULL) {
 121                         break;
 122                 } /* else ref held from pgfind */
 123
 124                 lwkt_gettoken_shared(&pg->pg_token);
 125                 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 126                         if (PRISON_CHECK(curtd->td_ucred, p->p_ucred) &&
 127                             p->p_nice < low) {
 128                                 low = p->p_nice;
 129                         }
 130                 }
 131                 lwkt_reltoken(&pg->pg_token);
 132                 pgrel(pg);
 133                 break;
 134         case PRIO_USER:
 135                 if (uap->who == 0)
 136                         uap->who = curtd->td_ucred->cr_uid;
 137                 info.low = low;
 138                 info.who = uap->who;
 139                 allproc_scan(getpriority_callback, &info);
 140                 low = info.low;
 141                 break;
 142
 143         default:
 144                 error = EINVAL;
 145                 goto done;
 146         }
 147         if (low == PRIO_MAX + 1) {
 148                 error = ESRCH;
 149                 goto done;
 150         }
 151         uap->sysmsg_result = low;
 152         error = 0;
 153 done:
 154         return (error);
 155 }
 156
 157 /*
 158  * Figure out the current lowest nice priority for processes owned
 159  * by the specified user.
 160  */
 161 static
 162 int
 163 getpriority_callback(struct proc *p, void *data)
 164 {
 165         struct getpriority_info *info = data;
 166
 167         lwkt_gettoken_shared(&p->p_token);
 168         if (PRISON_CHECK(curthread->td_ucred, p->p_ucred) &&
 169             p->p_ucred->cr_uid == info->who &&
 170             p->p_nice < info->low) {
 171                 info->low = p->p_nice;
 172         }
 173         lwkt_reltoken(&p->p_token);
 174         return(0);
 175 }
 176
 177 struct setpriority_info {
 178         int prio;
 179         int who;
 180         int error;
 181         int found;
 182 };
 183
 184 static int setpriority_callback(struct proc *p, void *data);
 185
 186 /*
 187  * MPALMOSTSAFE
 188  */
 189 int
 190 sys_setpriority(struct setpriority_args *uap)
 191 {
 192         struct setpriority_info info;
 193         thread_t curtd = curthread;
 194         struct proc *curp = curproc;
 195         struct proc *p;
 196         struct pgrp *pg;
 197         int found = 0, error = 0;
 198
 199         switch (uap->which) {
 200         case PRIO_PROCESS:
 201                 if (uap->who == 0) {
 202                         lwkt_gettoken(&curp->p_token);
 203                         error = donice(curp, uap->prio);
 204                         found++;
 205                         lwkt_reltoken(&curp->p_token);
 206                 } else {
 207                         p = pfind(uap->who);
 208                         if (p) {
 209                                 lwkt_gettoken(&p->p_token);
 210                                 if (PRISON_CHECK(curtd->td_ucred, p->p_ucred)) {
 211                                         error = donice(p, uap->prio);
 212                                         found++;
 213                                 }
 214                                 lwkt_reltoken(&p->p_token);
 215                                 PRELE(p);
 216                         }
 217                 }
 218                 break;
 219         case PRIO_PGRP:
 220                 if (uap->who == 0) {
 221                         lwkt_gettoken_shared(&curp->p_token);
 222                         pg = curp->p_pgrp;
 223                         pgref(pg);
 224                         lwkt_reltoken(&curp->p_token);
 225                 } else if ((pg = pgfind(uap->who)) == NULL) {
 226                         break;
 227                 } /* else ref held from pgfind */
 228
 229                 lwkt_gettoken(&pg->pg_token);
 230 restart:
 231                 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 232                         PHOLD(p);
 233                         lwkt_gettoken(&p->p_token);
 234                         if (p->p_pgrp == pg &&
 235                             PRISON_CHECK(curtd->td_ucred, p->p_ucred)) {
 236                                 error = donice(p, uap->prio);
 237                                 found++;
 238                         }
 239                         lwkt_reltoken(&p->p_token);
 240                         if (p->p_pgrp != pg) {
 241                                 PRELE(p);
 242                                 goto restart;
 243                         }
 244                         PRELE(p);
 245                 }
 246                 lwkt_reltoken(&pg->pg_token);
 247                 pgrel(pg);
 248                 break;
 249         case PRIO_USER:
 250                 if (uap->who == 0)
 251                         uap->who = curtd->td_ucred->cr_uid;
 252                 info.prio = uap->prio;
 253                 info.who = uap->who;
 254                 info.error = 0;
 255                 info.found = 0;
 256                 allproc_scan(setpriority_callback, &info);
 257                 error = info.error;
 258                 found = info.found;
 259                 break;
 260         default:
 261                 error = EINVAL;
 262                 found = 1;
 263                 break;
 264         }
 265
 266         if (found == 0)
 267                 error = ESRCH;
 268         return (error);
 269 }
 270
 271 static
 272 int
 273 setpriority_callback(struct proc *p, void *data)
 274 {
 275         struct setpriority_info *info = data;
 276         int error;
 277
 278         lwkt_gettoken(&p->p_token);
 279         if (p->p_ucred->cr_uid == info->who &&
 280             PRISON_CHECK(curthread->td_ucred, p->p_ucred)) {
 281                 error = donice(p, info->prio);
 282                 if (error)
 283                         info->error = error;
 284                 ++info->found;
 285         }
 286         lwkt_reltoken(&p->p_token);
 287         return(0);
 288 }
 289
 290 /*
 291  * Caller must hold chgp->p_token
 292  */
 293 static int
 294 donice(struct proc *chgp, int n)
 295 {
 296         struct ucred *cr = curthread->td_ucred;
 297         struct lwp *lp;
 298
 299         if (cr->cr_uid && cr->cr_ruid &&
 300             cr->cr_uid != chgp->p_ucred->cr_uid &&
 301             cr->cr_ruid != chgp->p_ucred->cr_uid)
 302                 return (EPERM);
 303         if (n > PRIO_MAX)
 304                 n = PRIO_MAX;
 305         if (n < PRIO_MIN)
 306                 n = PRIO_MIN;
 307         if (n < chgp->p_nice && priv_check_cred(cr, PRIV_SCHED_SETPRIORITY, 0))
 308                 return (EACCES);
 309         chgp->p_nice = n;
 310         FOREACH_LWP_IN_PROC(lp, chgp) {
 311                 LWPHOLD(lp);
 312                 chgp->p_usched->resetpriority(lp);
 313                 LWPRELE(lp);
 314         }
 315         return (0);
 316 }
 317
 318
 319 struct ioprio_get_info {
 320         int high;
 321         int who;
 322 };
 323
 324 static int ioprio_get_callback(struct proc *p, void *data);
 325
 326 /*
 327  * MPALMOSTSAFE
 328  */
 329 int
 330 sys_ioprio_get(struct ioprio_get_args *uap)
 331 {
 332         struct ioprio_get_info info;
 333         thread_t curtd = curthread;
 334         struct proc *curp = curproc;
 335         struct proc *p;
 336         struct pgrp *pg;
 337         int high = IOPRIO_MIN-2;
 338         int error;
 339
 340         switch (uap->which) {
 341         case PRIO_PROCESS:
 342                 if (uap->who == 0) {
 343                         high = curp->p_ionice;
 344                 } else {
 345                         p = pfind(uap->who);
 346                         if (p) {
 347                                 lwkt_gettoken_shared(&p->p_token);
 348                                 if (PRISON_CHECK(curtd->td_ucred, p->p_ucred))
 349                                         high = p->p_ionice;
 350                                 lwkt_reltoken(&p->p_token);
 351                                 PRELE(p);
 352                         }
 353                 }
 354                 break;
 355         case PRIO_PGRP:
 356                 if (uap->who == 0) {
 357                         lwkt_gettoken_shared(&curp->p_token);
 358                         pg = curp->p_pgrp;
 359                         pgref(pg);
 360                         lwkt_reltoken(&curp->p_token);
 361                 } else if ((pg = pgfind(uap->who)) == NULL) {
 362                         break;
 363                 } /* else ref held from pgfind */
 364
 365                 lwkt_gettoken_shared(&pg->pg_token);
 366                 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 367                         if (PRISON_CHECK(curtd->td_ucred, p->p_ucred) &&
 368                             p->p_nice > high)
 369                                 high = p->p_ionice;
 370                 }
 371                 lwkt_reltoken(&pg->pg_token);
 372                 pgrel(pg);
 373                 break;
 374         case PRIO_USER:
 375                 if (uap->who == 0)
 376                         uap->who = curtd->td_ucred->cr_uid;
 377                 info.high = high;
 378                 info.who = uap->who;
 379                 allproc_scan(ioprio_get_callback, &info);
 380                 high = info.high;
 381                 break;
 382         default:
 383                 error = EINVAL;
 384                 goto done;
 385         }
 386         if (high == IOPRIO_MIN-2) {
 387                 error = ESRCH;
 388                 goto done;
 389         }
 390         uap->sysmsg_result = high;
 391         error = 0;
 392 done:
 393         return (error);
 394 }
 395
 396 /*
 397  * Figure out the current lowest nice priority for processes owned
 398  * by the specified user.
 399  */
 400 static
 401 int
 402 ioprio_get_callback(struct proc *p, void *data)
 403 {
 404         struct ioprio_get_info *info = data;
 405
 406         lwkt_gettoken_shared(&p->p_token);
 407         if (PRISON_CHECK(curthread->td_ucred, p->p_ucred) &&
 408             p->p_ucred->cr_uid == info->who &&
 409             p->p_ionice > info->high) {
 410                 info->high = p->p_ionice;
 411         }
 412         lwkt_reltoken(&p->p_token);
 413         return(0);
 414 }
 415
 416
 417 struct ioprio_set_info {
 418         int prio;
 419         int who;
 420         int error;
 421         int found;
 422 };
 423
 424 static int ioprio_set_callback(struct proc *p, void *data);
 425
 426 /*
 427  * MPALMOSTSAFE
 428  */
 429 int
 430 sys_ioprio_set(struct ioprio_set_args *uap)
 431 {
 432         struct ioprio_set_info info;
 433         thread_t curtd = curthread;
 434         struct proc *curp = curproc;
 435         struct proc *p;
 436         struct pgrp *pg;
 437         int found = 0, error = 0;
 438
 439         switch (uap->which) {
 440         case PRIO_PROCESS:
 441                 if (uap->who == 0) {
 442                         lwkt_gettoken(&curp->p_token);
 443                         error = doionice(curp, uap->prio);
 444                         lwkt_reltoken(&curp->p_token);
 445                         found++;
 446                 } else {
 447                         p = pfind(uap->who);
 448                         if (p) {
 449                                 lwkt_gettoken(&p->p_token);
 450                                 if (PRISON_CHECK(curtd->td_ucred, p->p_ucred)) {
 451                                         error = doionice(p, uap->prio);
 452                                         found++;
 453                                 }
 454                                 lwkt_reltoken(&p->p_token);
 455                                 PRELE(p);
 456                         }
 457                 }
 458                 break;
 459         case PRIO_PGRP:
 460                 if (uap->who == 0) {
 461                         lwkt_gettoken_shared(&curp->p_token);
 462                         pg = curp->p_pgrp;
 463                         pgref(pg);
 464                         lwkt_reltoken(&curp->p_token);
 465                 } else if ((pg = pgfind(uap->who)) == NULL) {
 466                         break;
 467                 } /* else ref held from pgfind */
 468
 469                 lwkt_gettoken(&pg->pg_token);
 470 restart:
 471                 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 472                         PHOLD(p);
 473                         lwkt_gettoken(&p->p_token);
 474                         if (p->p_pgrp == pg &&
 475                             PRISON_CHECK(curtd->td_ucred, p->p_ucred)) {
 476                                 error = doionice(p, uap->prio);
 477                                 found++;
 478                         }
 479                         lwkt_reltoken(&p->p_token);
 480                         if (p->p_pgrp != pg) {
 481                                 PRELE(p);
 482                                 goto restart;
 483                         }
 484                         PRELE(p);
 485                 }
 486                 lwkt_reltoken(&pg->pg_token);
 487                 pgrel(pg);
 488                 break;
 489         case PRIO_USER:
 490                 if (uap->who == 0)
 491                         uap->who = curtd->td_ucred->cr_uid;
 492                 info.prio = uap->prio;
 493                 info.who = uap->who;
 494                 info.error = 0;
 495                 info.found = 0;
 496                 allproc_scan(ioprio_set_callback, &info);
 497                 error = info.error;
 498                 found = info.found;
 499                 break;
 500         default:
 501                 error = EINVAL;
 502                 found = 1;
 503                 break;
 504         }
 505
 506         if (found == 0)
 507                 error = ESRCH;
 508         return (error);
 509 }
 510
 511 static
 512 int
 513 ioprio_set_callback(struct proc *p, void *data)
 514 {
 515         struct ioprio_set_info *info = data;
 516         int error;
 517
 518         lwkt_gettoken(&p->p_token);
 519         if (p->p_ucred->cr_uid == info->who &&
 520             PRISON_CHECK(curthread->td_ucred, p->p_ucred)) {
 521                 error = doionice(p, info->prio);
 522                 if (error)
 523                         info->error = error;
 524                 ++info->found;
 525         }
 526         lwkt_reltoken(&p->p_token);
 527         return(0);
 528 }
 529
 530 static int
 531 doionice(struct proc *chgp, int n)
 532 {
 533         struct ucred *cr = curthread->td_ucred;
 534
 535         if (cr->cr_uid && cr->cr_ruid &&
 536             cr->cr_uid != chgp->p_ucred->cr_uid &&
 537             cr->cr_ruid != chgp->p_ucred->cr_uid)
 538                 return (EPERM);
 539         if (n > IOPRIO_MAX)
 540                 n = IOPRIO_MAX;
 541         if (n < IOPRIO_MIN)
 542                 n = IOPRIO_MIN;
 543         if (n < chgp->p_ionice &&
 544             priv_check_cred(cr, PRIV_SCHED_SETPRIORITY, 0))
 545                 return (EACCES);
 546         chgp->p_ionice = n;
 547
 548         return (0);
 549
 550 }
 551
 552 /*
 553  * MPALMOSTSAFE
 554  */
 555 int
 556 sys_lwp_rtprio(struct lwp_rtprio_args *uap)
 557 {
 558         struct ucred *cr = curthread->td_ucred;
 559         struct proc *p;
 560         struct lwp *lp;
 561         struct rtprio rtp;
 562         int error;
 563
 564         error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
 565         if (error)
 566                 return error;
 567         if (uap->pid < 0)
 568                 return EINVAL;
 569
 570         if (uap->pid == 0) {
 571                 p = curproc;
 572                 PHOLD(p);
 573         } else {
 574                 p = pfind(uap->pid);
 575         }
 576         if (p == NULL) {
 577                 error = ESRCH;
 578                 goto done;
 579         }
 580         lwkt_gettoken(&p->p_token);
 581
 582         if (uap->tid < -1) {
 583                 error = EINVAL;
 584                 goto done;
 585         }
 586         if (uap->tid == -1) {
 587                 /*
 588                  * sadly, tid can be 0 so we can't use 0 here
 589                  * like sys_rtprio()
 590                  */
 591                 lp = curthread->td_lwp;
 592         } else {
 593                 lp = lwp_rb_tree_RB_LOOKUP(&p->p_lwp_tree, uap->tid);
 594                 if (lp == NULL) {
 595                         error = ESRCH;
 596                         goto done;
 597                 }
 598         }
 599
 600         /*
 601          * Make sure that this lwp is not ripped if any of the following
 602          * code blocks, e.g. copyout.
 603          */
 604         LWPHOLD(lp);
 605         switch (uap->function) {
 606         case RTP_LOOKUP:
 607                 error = copyout(&lp->lwp_rtprio, uap->rtp,
 608                                 sizeof(struct rtprio));
 609                 break;
 610         case RTP_SET:
 611                 if (cr->cr_uid && cr->cr_ruid &&
 612                     cr->cr_uid != p->p_ucred->cr_uid &&
 613                     cr->cr_ruid != p->p_ucred->cr_uid) {
 614                         error = EPERM;
 615                         break;
 616                 }
 617                 /* disallow setting rtprio in most cases if not superuser */
 618                 if (priv_check_cred(cr, PRIV_SCHED_RTPRIO, 0)) {
 619                         /* can't set someone else's */
 620                         if (uap->pid) { /* XXX */
 621                                 error = EPERM;
 622                                 break;
 623                         }
 624                         /* can't set realtime priority */
 625 /*
 626  * Realtime priority has to be restricted for reasons which should be
 627  * obvious. However, for idle priority, there is a potential for
 628  * system deadlock if an idleprio process gains a lock on a resource
 629  * that other processes need (and the idleprio process can't run
 630  * due to a CPU-bound normal process). Fix me! XXX
 631  */
 632                         if (RTP_PRIO_IS_REALTIME(rtp.type)) {
 633                                 error = EPERM;
 634                                 break;
 635                         }
 636                 }
 637                 switch (rtp.type) {
 638 #ifdef RTP_PRIO_FIFO
 639                 case RTP_PRIO_FIFO:
 640 #endif
 641                 case RTP_PRIO_REALTIME:
 642                 case RTP_PRIO_NORMAL:
 643                 case RTP_PRIO_IDLE:
 644                         if (rtp.prio > RTP_PRIO_MAX) {
 645                                 error = EINVAL;
 646                         } else {
 647                                 lp->lwp_rtprio = rtp;
 648                                 error = 0;
 649                         }
 650                         break;
 651                 default:
 652                         error = EINVAL;
 653                         break;
 654                 }
 655                 break;
 656         default:
 657                 error = EINVAL;
 658                 break;
 659         }
 660         LWPRELE(lp);
 661
 662 done:
 663         if (p) {
 664                 lwkt_reltoken(&p->p_token);
 665                 PRELE(p);
 666         }
 667         return (error);
 668 }
 669
 670 /*
 671  * Set realtime priority
 672  *
 673  * MPALMOSTSAFE
 674  */
 675 int
 676 sys_rtprio(struct rtprio_args *uap)
 677 {
 678         struct ucred *cr = curthread->td_ucred;
 679         struct proc *p;
 680         struct lwp *lp;
 681         struct rtprio rtp;
 682         int error;
 683
 684         error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
 685         if (error)
 686                 return (error);
 687
 688         if (uap->pid == 0) {
 689                 p = curproc;
 690                 PHOLD(p);
 691         } else {
 692                 p = pfind(uap->pid);
 693         }
 694
 695         if (p == NULL) {
 696                 error = ESRCH;
 697                 goto done;
 698         }
 699         lwkt_gettoken(&p->p_token);
 700
 701         /* XXX lwp */
 702         lp = FIRST_LWP_IN_PROC(p);
 703         switch (uap->function) {
 704         case RTP_LOOKUP:
 705                 error = copyout(&lp->lwp_rtprio, uap->rtp,
 706                                 sizeof(struct rtprio));
 707                 break;
 708         case RTP_SET:
 709                 if (cr->cr_uid && cr->cr_ruid &&
 710                     cr->cr_uid != p->p_ucred->cr_uid &&
 711                     cr->cr_ruid != p->p_ucred->cr_uid) {
 712                         error = EPERM;
 713                         break;
 714                 }
 715                 /* disallow setting rtprio in most cases if not superuser */
 716                 if (priv_check_cred(cr, PRIV_SCHED_RTPRIO, 0)) {
 717                         /* can't set someone else's */
 718                         if (uap->pid) {
 719                                 error = EPERM;
 720                                 break;
 721                         }
 722                         /* can't set realtime priority */
 723 /*
 724  * Realtime priority has to be restricted for reasons which should be
 725  * obvious. However, for idle priority, there is a potential for
 726  * system deadlock if an idleprio process gains a lock on a resource
 727  * that other processes need (and the idleprio process can't run
 728  * due to a CPU-bound normal process). Fix me! XXX
 729  */
 730                         if (RTP_PRIO_IS_REALTIME(rtp.type)) {
 731                                 error = EPERM;
 732                                 break;
 733                         }
 734                 }
 735                 switch (rtp.type) {
 736 #ifdef RTP_PRIO_FIFO
 737                 case RTP_PRIO_FIFO:
 738 #endif
 739                 case RTP_PRIO_REALTIME:
 740                 case RTP_PRIO_NORMAL:
 741                 case RTP_PRIO_IDLE:
 742                         if (rtp.prio > RTP_PRIO_MAX) {
 743                                 error = EINVAL;
 744                                 break;
 745                         }
 746                         lp->lwp_rtprio = rtp;
 747                         error = 0;
 748                         break;
 749                 default:
 750                         error = EINVAL;
 751                         break;
 752                 }
 753                 break;
 754         default:
 755                 error = EINVAL;
 756                 break;
 757         }
 758 done:
 759         if (p) {
 760                 lwkt_reltoken(&p->p_token);
 761                 PRELE(p);
 762         }
 763
 764         return (error);
 765 }
 766
 767 /*
 768  * MPSAFE
 769  */
 770 int
 771 sys_setrlimit(struct __setrlimit_args *uap)
 772 {
 773         struct rlimit alim;
 774         int error;
 775
 776         error = copyin(uap->rlp, &alim, sizeof(alim));
 777         if (error)
 778                 return (error);
 779
 780         error = kern_setrlimit(uap->which, &alim);
 781
 782         return (error);
 783 }
 784
 785 /*
 786  * MPSAFE
 787  */
 788 int
 789 sys_getrlimit(struct __getrlimit_args *uap)
 790 {
 791         struct rlimit lim;
 792         int error;
 793
 794         error = kern_getrlimit(uap->which, &lim);
 795
 796         if (error == 0)
 797                 error = copyout(&lim, uap->rlp, sizeof(*uap->rlp));
 798         return error;
 799 }
 800
 801 /*
 802  * Transform the running time and tick information in lwp lp's thread into user,
 803  * system, and interrupt time usage.
 804  *
 805  * Since we are limited to statclock tick granularity this is a statisical
 806  * calculation which will be correct over the long haul, but should not be
 807  * expected to measure fine grained deltas.
 808  *
 809  * It is possible to catch a lwp in the midst of being created, so
 810  * check whether lwp_thread is NULL or not.
 811  */
 812 void
 813 calcru(struct lwp *lp, struct timeval *up, struct timeval *sp)
 814 {
 815         struct thread *td;
 816
 817         /*
 818          * Calculate at the statclock level.  YYY if the thread is owned by
 819          * another cpu we need to forward the request to the other cpu, or
 820          * have a token to interlock the information in order to avoid racing
 821          * thread destruction.
 822          */
 823         if ((td = lp->lwp_thread) != NULL) {
 824                 crit_enter();
 825                 up->tv_sec = td->td_uticks / 1000000;
 826                 up->tv_usec = td->td_uticks % 1000000;
 827                 sp->tv_sec = td->td_sticks / 1000000;
 828                 sp->tv_usec = td->td_sticks % 1000000;
 829                 crit_exit();
 830         }
 831 }
 832
 833 /*
 834  * Aggregate resource statistics of all lwps of a process.
 835  *
 836  * proc.p_ru keeps track of all statistics directly related to a proc.  This
 837  * consists of RSS usage and nswap information and aggregate numbers for all
 838  * former lwps of this proc.
 839  *
 840  * proc.p_cru is the sum of all stats of reaped children.
 841  *
 842  * lwp.lwp_ru contains the stats directly related to one specific lwp, meaning
 843  * packet, scheduler switch or page fault counts, etc.  This information gets
 844  * added to lwp.lwp_proc.p_ru when the lwp exits.
 845  */
 846 void
 847 calcru_proc(struct proc *p, struct rusage *ru)
 848 {
 849         struct timeval upt, spt;
 850         long *rip1, *rip2;
 851         struct lwp *lp;
 852
 853         *ru = p->p_ru;
 854
 855         FOREACH_LWP_IN_PROC(lp, p) {
 856                 calcru(lp, &upt, &spt);
 857                 timevaladd(&ru->ru_utime, &upt);
 858                 timevaladd(&ru->ru_stime, &spt);
 859                 for (rip1 = &ru->ru_first, rip2 = &lp->lwp_ru.ru_first;
 860                      rip1 <= &ru->ru_last;
 861                      rip1++, rip2++)
 862                         *rip1 += *rip2;
 863         }
 864 }
 865
 866
 867 /*
 868  * MPALMOSTSAFE
 869  */
 870 int
 871 sys_getrusage(struct getrusage_args *uap)
 872 {
 873         struct proc *p = curproc;
 874         struct rusage ru;
 875         struct rusage *rup;
 876         int error;
 877
 878         lwkt_gettoken(&p->p_token);
 879
 880         switch (uap->who) {
 881         case RUSAGE_SELF:
 882                 rup = &ru;
 883                 calcru_proc(p, rup);
 884                 error = 0;
 885                 break;
 886         case RUSAGE_CHILDREN:
 887                 rup = &p->p_cru;
 888                 error = 0;
 889                 break;
 890         default:
 891                 error = EINVAL;
 892                 break;
 893         }
 894         lwkt_reltoken(&p->p_token);
 895
 896         if (error == 0)
 897                 error = copyout(rup, uap->rusage, sizeof(struct rusage));
 898         return (error);
 899 }
 900
 901 void
 902 ruadd(struct rusage *ru, struct rusage *ru2)
 903 {
 904         long *ip, *ip2;
 905         int i;
 906
 907         timevaladd(&ru->ru_utime, &ru2->ru_utime);
 908         timevaladd(&ru->ru_stime, &ru2->ru_stime);
 909         if (ru->ru_maxrss < ru2->ru_maxrss)
 910                 ru->ru_maxrss = ru2->ru_maxrss;
 911         ip = &ru->ru_first; ip2 = &ru2->ru_first;
 912         for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
 913                 *ip++ += *ip2++;
 914 }
 915
 916 /*
 917  * Find the uidinfo structure for a uid.  This structure is used to
 918  * track the total resource consumption (process count, socket buffer
 919  * size, etc.) for the uid and impose limits.
 920  */
 921 void
 922 uihashinit(void)
 923 {
 924         spin_init(&uihash_lock, "uihashinit");
 925         uihashtbl = hashinit(maxproc / 16, M_UIDINFO, &uihash);
 926 }
 927
 928 /*
 929  * NOTE: Must be called with uihash_lock held
 930  *
 931  * MPSAFE
 932  */
 933 static struct uidinfo *
 934 uilookup(uid_t uid)
 935 {
 936         struct  uihashhead *uipp;
 937         struct  uidinfo *uip;
 938
 939         uipp = UIHASH(uid);
 940         LIST_FOREACH(uip, uipp, ui_hash) {
 941                 if (uip->ui_uid == uid)
 942                         break;
 943         }
 944         return (uip);
 945 }
 946
 947 /*
 948  * Helper function to creat ea uid that could not be found.
 949  * This function will properly deal with races.
 950  *
 951  * MPSAFE
 952  */
 953 static struct uidinfo *
 954 uicreate(uid_t uid)
 955 {
 956         struct  uidinfo *uip, *tmp;
 957
 958         /*
 959          * Allocate space and check for a race
 960          */
 961         uip = kmalloc(sizeof(*uip), M_UIDINFO, M_WAITOK|M_ZERO);
 962
 963         /*
 964          * Initialize structure and enter it into the hash table
 965          */
 966         spin_init(&uip->ui_lock, "uicreate");
 967         uip->ui_uid = uid;
 968         uip->ui_ref = 1;        /* we're returning a ref */
 969         varsymset_init(&uip->ui_varsymset, NULL);
 970
 971         /*
 972          * Somebody may have already created the uidinfo for this
 973          * uid. If so, return that instead.
 974          */
 975         spin_lock(&uihash_lock);
 976         tmp = uilookup(uid);
 977         if (tmp != NULL) {
 978                 uihold(tmp);
 979                 spin_unlock(&uihash_lock);
 980
 981                 spin_uninit(&uip->ui_lock);
 982                 varsymset_clean(&uip->ui_varsymset);
 983                 kfree(uip, M_UIDINFO);
 984                 uip = tmp;
 985         } else {
 986                 LIST_INSERT_HEAD(UIHASH(uid), uip, ui_hash);
 987                 spin_unlock(&uihash_lock);
 988         }
 989         return (uip);
 990 }
 991
 992 /*
 993  *
 994  *
 995  * MPSAFE
 996  */
 997 struct uidinfo *
 998 uifind(uid_t uid)
 999 {
1000         struct  uidinfo *uip;
1001
1002         spin_lock(&uihash_lock);
1003         uip = uilookup(uid);
1004         if (uip == NULL) {
1005                 spin_unlock(&uihash_lock);
1006                 uip = uicreate(uid);
1007         } else {
1008                 uihold(uip);
1009                 spin_unlock(&uihash_lock);
1010         }
1011         return (uip);
1012 }
1013
1014 /*
1015  * Helper funtion to remove a uidinfo whos reference count is
1016  * transitioning from 1->0.  The reference count is 1 on call.
1017  *
1018  * Zero is returned on success, otherwise non-zero and the
1019  * uiphas not been removed.
1020  *
1021  * MPSAFE
1022  */
1023 static __inline int
1024 uifree(struct uidinfo *uip)
1025 {
1026         /*
1027          * If we are still the only holder after acquiring the uihash_lock
1028          * we can safely unlink the uip and destroy it.  Otherwise we lost
1029          * a race and must fail.
1030          */
1031         spin_lock(&uihash_lock);
1032         if (uip->ui_ref != 1) {
1033                 spin_unlock(&uihash_lock);
1034                 return(-1);
1035         }
1036         LIST_REMOVE(uip, ui_hash);
1037         spin_unlock(&uihash_lock);
1038
1039         /*
1040          * The uip is now orphaned and we can destroy it at our
1041          * leisure.
1042          */
1043         if (uip->ui_sbsize != 0)
1044                 kprintf("freeing uidinfo: uid = %d, sbsize = %jd\n",
1045                     uip->ui_uid, (intmax_t)uip->ui_sbsize);
1046         if (uip->ui_proccnt != 0)
1047                 kprintf("freeing uidinfo: uid = %d, proccnt = %ld\n",
1048                     uip->ui_uid, uip->ui_proccnt);
1049
1050         varsymset_clean(&uip->ui_varsymset);
1051         lockuninit(&uip->ui_varsymset.vx_lock);
1052         spin_uninit(&uip->ui_lock);
1053         kfree(uip, M_UIDINFO);
1054         return(0);
1055 }
1056
1057 /*
1058  * MPSAFE
1059  */
1060 void
1061 uihold(struct uidinfo *uip)
1062 {
1063         atomic_add_int(&uip->ui_ref, 1);
1064         KKASSERT(uip->ui_ref >= 0);
1065 }
1066
1067 /*
1068  * NOTE: It is important for us to not drop the ref count to 0
1069  *       because this can cause a 2->0/2->0 race with another
1070  *       concurrent dropper.  Losing the race in that situation
1071  *       can cause uip to become stale for one of the other
1072  *       threads.
1073  *
1074  * MPSAFE
1075  */
1076 void
1077 uidrop(struct uidinfo *uip)
1078 {
1079         int ref;
1080
1081         KKASSERT(uip->ui_ref > 0);
1082
1083         for (;;) {
1084                 ref = uip->ui_ref;
1085                 cpu_ccfence();
1086                 if (ref == 1) {
1087                         if (uifree(uip) == 0)
1088                                 break;
1089                 } else if (atomic_cmpset_int(&uip->ui_ref, ref, ref - 1)) {
1090                         break;
1091                 }
1092                 /* else retry */
1093         }
1094 }
1095
1096 void
1097 uireplace(struct uidinfo **puip, struct uidinfo *nuip)
1098 {
1099         uidrop(*puip);
1100         *puip = nuip;
1101 }
1102
1103 /*
1104  * Change the count associated with number of processes
1105  * a given user is using.  When 'max' is 0, don't enforce a limit
1106  */
1107 int
1108 chgproccnt(struct uidinfo *uip, int diff, int max)
1109 {
1110         int ret;
1111         spin_lock(&uip->ui_lock);
1112         /* don't allow them to exceed max, but allow subtraction */
1113         if (diff > 0 && uip->ui_proccnt + diff > max && max != 0) {
1114                 ret = 0;
1115         } else {
1116                 uip->ui_proccnt += diff;
1117                 if (uip->ui_proccnt < 0)
1118                         kprintf("negative proccnt for uid = %d\n", uip->ui_uid);
1119                 ret = 1;
1120         }
1121         spin_unlock(&uip->ui_lock);
1122         return ret;
1123 }
1124
1125 /*
1126  * Change the total socket buffer size a user has used.
1127  */
1128 int
1129 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t max)
1130 {
1131         rlim_t new;
1132
1133 #ifdef __x86_64__
1134         rlim_t sbsize;
1135
1136         sbsize = atomic_fetchadd_long(&uip->ui_sbsize, to - *hiwat);
1137         new = sbsize + to - *hiwat;
1138 #else
1139         spin_lock(&uip->ui_lock);
1140         new = uip->ui_sbsize + to - *hiwat;
1141         uip->ui_sbsize = new;
1142         spin_unlock(&uip->ui_lock);
1143 #endif
1144         KKASSERT(new >= 0);
1145
1146         /*
1147          * If we are trying to increase the socket buffer size
1148          * Scale down the hi water mark when we exceed the user's
1149          * allowed socket buffer space.
1150          *
1151          * We can't scale down too much or we will blow up atomic packet
1152          * operations.
1153          */
1154         if (to > *hiwat && to > MCLBYTES && new > max) {
1155                 to = to * max / new;
1156                 if (to < MCLBYTES)
1157                         to = MCLBYTES;
1158         }
1159         *hiwat = to;
1160         return (1);
1161 }