sys/kern/kern_plimit.c

   1 /*
   2  * Copyright (c) 2006 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Matthew Dillon <dillon@backplane.com>
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in
  15  *    the documentation and/or other materials provided with the
  16  *    distribution.
  17  * 3. Neither the name of The DragonFly Project nor the names of its
  18  *    contributors may be used to endorse or promote products derived
  19  *    from this software without specific, prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  */
  34 /*
  35  * Copyright (c) 1982, 1986, 1991, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  * (c) UNIX System Laboratories, Inc.
  38  * All or some portions of this file are derived from material licensed
  39  * to the University of California by American Telephone and Telegraph
  40  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
  41  * the permission of UNIX System Laboratories, Inc.
  42  *
  43  * Redistribution and use in source and binary forms, with or without
  44  * modification, are permitted provided that the following conditions
  45  * are met:
  46  * 1. Redistributions of source code must retain the above copyright
  47  *    notice, this list of conditions and the following disclaimer.
  48  * 2. Redistributions in binary form must reproduce the above copyright
  49  *    notice, this list of conditions and the following disclaimer in the
  50  *    documentation and/or other materials provided with the distribution.
  51  * 3. Neither the name of the University nor the names of its contributors
  52  *    may be used to endorse or promote products derived from this software
  53  *    without specific prior written permission.
  54  *
  55  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  56  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  57  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  58  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  59  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  60  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  61  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  62  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  63  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  64  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  65  * SUCH DAMAGE.
  66  *
  67  *      @(#)kern_resource.c     8.5 (Berkeley) 1/21/94
  68  * $FreeBSD: src/sys/kern/kern_resource.c,v 1.55.2.5 2001/11/03 01:41:08 ps Exp $
  69  * $DragonFly: src/sys/kern/kern_plimit.c,v 1.3 2008/05/08 01:26:00 dillon Exp $
  70  */
  71
  72 #include <sys/resource.h>
  73 #include <sys/spinlock.h>
  74 #include <sys/proc.h>
  75 #include <sys/file.h>
  76 #include <sys/lockf.h>
  77 #include <sys/kern_syscall.h>
  78
  79 #include <vm/vm_param.h>
  80 #include <vm/vm.h>
  81 #include <vm/vm_map.h>
  82
  83 #include <machine/pmap.h>
  84
  85 #include <sys/spinlock2.h>
  86
  87 static void plimit_copy(struct plimit *olimit, struct plimit *nlimit);
  88
  89 /*
  90  * Initialize proc0's plimit structure.  All later plimit structures
  91  * are inherited through fork.
  92  */
  93 void
  94 plimit_init0(struct plimit *limit)
  95 {
  96         int i;
  97         rlim_t lim;
  98
  99         for (i = 0; i < RLIM_NLIMITS; ++i) {
 100                 limit->pl_rlimit[i].rlim_cur = RLIM_INFINITY;
 101                 limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
 102         }
 103         limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = maxfiles;
 104         limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
 105         limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = maxproc;
 106         limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
 107         lim = ptoa((rlim_t)vmstats.v_free_count);
 108         limit->pl_rlimit[RLIMIT_RSS].rlim_max = lim;
 109         limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
 110         limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
 111         limit->p_cpulimit = RLIM_INFINITY;
 112         limit->p_refcnt = 1;
 113         spin_init(&limit->p_spin);
 114 }
 115
 116 /*
 117  * Return a plimit for use by a new forked process given the one
 118  * contained in the parent process.
 119  *
 120  * MPSAFE
 121  */
 122 struct plimit *
 123 plimit_fork(struct proc *p1)
 124 {
 125         struct plimit *olimit = p1->p_limit;
 126         struct plimit *nlimit = NULL;
 127         struct plimit *rlimit;
 128
 129         /*
 130          * If we are exclusive (but not threaded-exclusive), but have only
 131          * one reference, we can convert the structure to copy-on-write
 132          * again.
 133          *
 134          * If we were threaded but are no longer threaded we can do the same
 135          * thing.
 136          */
 137         if (olimit->p_exclusive == 1) {
 138                 KKASSERT(olimit->p_refcnt == 1);
 139                 olimit->p_exclusive = 0;
 140         } else if (olimit->p_exclusive == 2 && p1->p_nthreads == 1) {
 141                 KKASSERT(olimit->p_refcnt == 1);
 142                 olimit->p_exclusive = 0;
 143         }
 144
 145         /*
 146          * Take a short-cut that requires limited spin locks.  If we aren't
 147          * exclusive we will not be threaded and we can just bump the ref
 148          * count.  If that is true and we also have only one ref then there
 149          * can be no other accessors.
 150          */
 151         if (olimit->p_exclusive == 0) {
 152                 if (olimit->p_refcnt == 1) {
 153                         ++olimit->p_refcnt;
 154                 } else {
 155                         spin_lock_wr(&olimit->p_spin);
 156                         ++olimit->p_refcnt;
 157                         spin_unlock_wr(&olimit->p_spin);
 158                 }
 159                 return(olimit);
 160         }
 161
 162         /*
 163          * Full-blown code-up.
 164          */
 165         nlimit = NULL;
 166         spin_lock_wr(&olimit->p_spin);
 167
 168         for (;;) {
 169                 if (olimit->p_exclusive == 0) {
 170                         ++olimit->p_refcnt;
 171                         rlimit = olimit;
 172                         break;
 173                 }
 174                 if (nlimit) {
 175                         plimit_copy(olimit, nlimit);
 176                         rlimit = nlimit;
 177                         nlimit = NULL;
 178                         break;
 179                 }
 180                 spin_unlock_wr(&olimit->p_spin);
 181                 nlimit = kmalloc(sizeof(*nlimit), M_SUBPROC, M_WAITOK);
 182                 spin_lock_wr(&olimit->p_spin);
 183         }
 184         spin_unlock_wr(&olimit->p_spin);
 185         if (nlimit)
 186                 kfree(nlimit, M_SUBPROC);
 187         return(rlimit);
 188 }
 189
 190 /*
 191  * This routine is called when a new LWP is created for a process.  We
 192  * must force exclusivity (=2) so p->p_limit remains stable.
 193  *
 194  * LWPs share the same process structure so this does not bump refcnt.
 195  */
 196 void
 197 plimit_lwp_fork(struct proc *p)
 198 {
 199         struct plimit *olimit;
 200
 201         for (;;) {
 202                 olimit = p->p_limit;
 203                 if (olimit->p_exclusive == 2) {
 204                         KKASSERT(olimit->p_refcnt == 1);
 205                         break;
 206                 }
 207                 if (olimit->p_refcnt == 1) {
 208                         olimit->p_exclusive = 2;
 209                         break;
 210                 }
 211                 plimit_modify(p, -1, NULL);
 212         }
 213 }
 214
 215 /*
 216  * This routine is called to fixup a proces's p_limit structure prior
 217  * to it being modified.  If index >= 0 the specified modification is also
 218  * made.
 219  *
 220  * This routine must make the limit structure exclusive.  A later fork
 221  * will convert it back to copy-on-write if possible.
 222  *
 223  * We can count on p->p_limit being stable since if we had created any
 224  * threads it will have already been made exclusive (=2).
 225  *
 226  * MPSAFE
 227  */
 228 void
 229 plimit_modify(struct proc *p, int index, struct rlimit *rlim)
 230 {
 231         struct plimit *olimit;
 232         struct plimit *nlimit;
 233         struct plimit *rlimit;
 234
 235         /*
 236          * Shortcut.  If we are not threaded we may be able to trivially
 237          * set the structure to exclusive access without needing to acquire
 238          * any spinlocks.   The p_limit structure will be stable.
 239          */
 240         olimit = p->p_limit;
 241         if (p->p_nthreads == 1) {
 242                 if (olimit->p_exclusive == 0 && olimit->p_refcnt == 1)
 243                         olimit->p_exclusive = 1;
 244                 if (olimit->p_exclusive) {
 245                         if (index >= 0)
 246                                 p->p_limit->pl_rlimit[index] = *rlim;
 247                         return;
 248                 }
 249         }
 250
 251         /*
 252          * Full-blown code-up.  Make a copy if we aren't exclusive.  If
 253          * we have only one ref we can safely convert the structure to
 254          * exclusive without copying.
 255          */
 256         nlimit = NULL;
 257         spin_lock_wr(&olimit->p_spin);
 258
 259         for (;;) {
 260                 if (olimit->p_refcnt == 1) {
 261                         if (olimit->p_exclusive == 0)
 262                                 olimit->p_exclusive = 1;
 263                         rlimit = olimit;
 264                         break;
 265                 }
 266                 KKASSERT(olimit->p_exclusive == 0);
 267                 if (nlimit) {
 268                         plimit_copy(olimit, nlimit);
 269                         nlimit->p_exclusive = 1;
 270                         p->p_limit = nlimit;
 271                         rlimit = nlimit;
 272                         nlimit = NULL;
 273                         break;
 274                 }
 275                 spin_unlock_wr(&olimit->p_spin);
 276                 nlimit = kmalloc(sizeof(*nlimit), M_SUBPROC, M_WAITOK);
 277                 spin_lock_wr(&olimit->p_spin);
 278         }
 279         if (index >= 0)
 280                 rlimit->pl_rlimit[index] = *rlim;
 281         spin_unlock_wr(&olimit->p_spin);
 282         if (nlimit)
 283                 kfree(nlimit, M_SUBPROC);
 284 }
 285
 286 /*
 287  * Destroy a process's plimit structure.
 288  *
 289  * MPSAFE
 290  */
 291 void
 292 plimit_free(struct proc *p)
 293 {
 294         struct plimit *limit;
 295
 296         if ((limit = p->p_limit) != NULL) {
 297                 p->p_limit = NULL;
 298
 299                 if (limit->p_refcnt == 1) {
 300                         limit->p_refcnt = -999;
 301                         kfree(limit, M_SUBPROC);
 302                 } else {
 303                         spin_lock_wr(&limit->p_spin);
 304                         if (--limit->p_refcnt == 0) {
 305                                 spin_unlock_wr(&limit->p_spin);
 306                                 kfree(limit, M_SUBPROC);
 307                         } else {
 308                                 spin_unlock_wr(&limit->p_spin);
 309                         }
 310                 }
 311         }
 312 }
 313
 314 /*
 315  * Modify a resource limit (from system call)
 316  *
 317  * MPSAFE
 318  */
 319 int
 320 kern_setrlimit(u_int which, struct rlimit *limp)
 321 {
 322         struct proc *p = curproc;
 323         struct plimit *limit;
 324         struct rlimit *alimp;
 325         int error;
 326
 327         if (which >= RLIM_NLIMITS)
 328                 return (EINVAL);
 329
 330         /*
 331          * We will be modifying a resource, make a copy if necessary.
 332          */
 333         plimit_modify(p, -1, NULL);
 334         limit = p->p_limit;
 335         alimp = &limit->pl_rlimit[which];
 336
 337         /*
 338          * Preserve historical bugs by treating negative limits as unsigned.
 339          */
 340         if (limp->rlim_cur < 0)
 341                 limp->rlim_cur = RLIM_INFINITY;
 342         if (limp->rlim_max < 0)
 343                 limp->rlim_max = RLIM_INFINITY;
 344
 345         spin_lock_rd(&limit->p_spin);
 346         if (limp->rlim_cur > alimp->rlim_max ||
 347             limp->rlim_max > alimp->rlim_max) {
 348                 spin_unlock_rd(&limit->p_spin);
 349                 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)))
 350                         return (error);
 351         } else {
 352                 spin_unlock_rd(&limit->p_spin);
 353         }
 354         if (limp->rlim_cur > limp->rlim_max)
 355                 limp->rlim_cur = limp->rlim_max;
 356
 357         switch (which) {
 358         case RLIMIT_CPU:
 359                 spin_lock_wr(&limit->p_spin);
 360                 if (limp->rlim_cur > RLIM_INFINITY / (rlim_t)1000000)
 361                         limit->p_cpulimit = RLIM_INFINITY;
 362                 else
 363                         limit->p_cpulimit = (rlim_t)1000000 * limp->rlim_cur;
 364                 spin_unlock_wr(&limit->p_spin);
 365                 break;
 366         case RLIMIT_DATA:
 367                 if (limp->rlim_cur > maxdsiz)
 368                         limp->rlim_cur = maxdsiz;
 369                 if (limp->rlim_max > maxdsiz)
 370                         limp->rlim_max = maxdsiz;
 371                 break;
 372
 373         case RLIMIT_STACK:
 374                 if (limp->rlim_cur > maxssiz)
 375                         limp->rlim_cur = maxssiz;
 376                 if (limp->rlim_max > maxssiz)
 377                         limp->rlim_max = maxssiz;
 378                 /*
 379                  * Stack is allocated to the max at exec time with only
 380                  * "rlim_cur" bytes accessible.  If stack limit is going
 381                  * up make more accessible, if going down make inaccessible.
 382                  */
 383                 spin_lock_rd(&limit->p_spin);
 384                 if (limp->rlim_cur != alimp->rlim_cur) {
 385                         vm_offset_t addr;
 386                         vm_size_t size;
 387                         vm_prot_t prot;
 388
 389                         if (limp->rlim_cur > alimp->rlim_cur) {
 390                                 prot = VM_PROT_ALL;
 391                                 size = limp->rlim_cur - alimp->rlim_cur;
 392                                 addr = USRSTACK - limp->rlim_cur;
 393                         } else {
 394                                 prot = VM_PROT_NONE;
 395                                 size = alimp->rlim_cur - limp->rlim_cur;
 396                                 addr = USRSTACK - alimp->rlim_cur;
 397                         }
 398                         spin_unlock_rd(&limit->p_spin);
 399                         addr = trunc_page(addr);
 400                         size = round_page(size);
 401                         (void) vm_map_protect(&p->p_vmspace->vm_map,
 402                                               addr, addr+size, prot, FALSE);
 403                 } else {
 404                         spin_unlock_rd(&limit->p_spin);
 405                 }
 406                 break;
 407
 408         case RLIMIT_NOFILE:
 409                 if (limp->rlim_cur > maxfilesperproc)
 410                         limp->rlim_cur = maxfilesperproc;
 411                 if (limp->rlim_max > maxfilesperproc)
 412                         limp->rlim_max = maxfilesperproc;
 413                 break;
 414
 415         case RLIMIT_NPROC:
 416                 if (limp->rlim_cur > maxprocperuid)
 417                         limp->rlim_cur = maxprocperuid;
 418                 if (limp->rlim_max > maxprocperuid)
 419                         limp->rlim_max = maxprocperuid;
 420                 if (limp->rlim_cur < 1)
 421                         limp->rlim_cur = 1;
 422                 if (limp->rlim_max < 1)
 423                         limp->rlim_max = 1;
 424                 break;
 425         case RLIMIT_POSIXLOCKS:
 426                 if (limp->rlim_cur > maxposixlocksperuid)
 427                         limp->rlim_cur = maxposixlocksperuid;
 428                 if (limp->rlim_max > maxposixlocksperuid)
 429                         limp->rlim_max = maxposixlocksperuid;
 430                 break;
 431         }
 432         spin_lock_wr(&limit->p_spin);
 433         *alimp = *limp;
 434         spin_unlock_wr(&limit->p_spin);
 435         return (0);
 436 }
 437
 438 /*
 439  * The rlimit indexed by which is returned in the second argument.
 440  *
 441  * MPSAFE
 442  */
 443 int
 444 kern_getrlimit(u_int which, struct rlimit *limp)
 445 {
 446         struct proc *p = curproc;
 447         struct plimit *limit;
 448
 449         if (which >= RLIM_NLIMITS)
 450                 return (EINVAL);
 451
 452         limit = p->p_limit;
 453         spin_lock_rd(&limit->p_spin);
 454         *limp = p->p_rlimit[which];
 455         spin_unlock_rd(&limit->p_spin);
 456         return (0);
 457 }
 458
 459 /*
 460  * Determine if the cpu limit has been reached and return an operations
 461  * code for the caller to perform.
 462  *
 463  * MPSAFE
 464  */
 465 int
 466 plimit_testcpulimit(struct plimit *limit, u_int64_t ttime)
 467 {
 468         struct rlimit *rlim;
 469         int mode;
 470
 471         mode = PLIMIT_TESTCPU_OK;
 472         if (limit->p_cpulimit != RLIM_INFINITY) {
 473                 spin_lock_rd(&limit->p_spin);
 474                 if (ttime > limit->p_cpulimit) {
 475                         rlim = &limit->pl_rlimit[RLIMIT_CPU];
 476                         if (ttime / (rlim_t)1000000 >= rlim->rlim_max + 5) {
 477                                 mode = PLIMIT_TESTCPU_KILL;
 478                         } else {
 479                                 mode = PLIMIT_TESTCPU_XCPU;
 480                         }
 481                 }
 482                 spin_unlock_rd(&limit->p_spin);
 483         }
 484         return(mode);
 485 }
 486
 487 /*
 488  * Helper routine to copy olimit to nlimit and initialize nlimit for
 489  * use.  nlimit's reference count will be set to 1 and its exclusive bit
 490  * will be cleared.
 491  *
 492  * MPSAFE
 493  */
 494 static
 495 void
 496 plimit_copy(struct plimit *olimit, struct plimit *nlimit)
 497 {
 498         *nlimit = *olimit;
 499
 500         spin_init(&nlimit->p_spin);
 501         nlimit->p_refcnt = 1;
 502         nlimit->p_exclusive = 0;
 503 }
 504