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 1.2 2006/09/05 00:55:45 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 struct plimit *plimit_copy(struct plimit *olimit);
  88 static void plimit_exclusive(struct plimit **limitp);
  89
  90 /*
  91  * Initialize proc0's plimit structure.  All later plimit structures
  92  * are inherited through fork.
  93  */
  94 void
  95 plimit_init0(struct plimit *limit)
  96 {
  97         int i;
  98         rlim_t lim;
  99
 100         for (i = 0; i < RLIM_NLIMITS; ++i) {
 101                 limit->pl_rlimit[i].rlim_cur = RLIM_INFINITY;
 102                 limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
 103         }
 104         limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = maxfiles;
 105         limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
 106         limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = maxproc;
 107         limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
 108         lim = ptoa((rlim_t)vmstats.v_free_count);
 109         limit->pl_rlimit[RLIMIT_RSS].rlim_max = lim;
 110         limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
 111         limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
 112         limit->p_cpulimit = RLIM_INFINITY;
 113         limit->p_refcnt = 1;
 114         spin_init(&limit->p_spin);
 115 }
 116
 117 /*
 118  * Return a plimit for use by a new forked process given the one
 119  * contained in the parent process.  p_exclusive will be set if
 120  * the parent process has more then one thread, requiring a copy.
 121  * Otherwise we can just inherit a reference.
 122  *
 123  * MPSAFE
 124  */
 125 struct plimit *
 126 plimit_fork(struct plimit *limit)
 127 {
 128         if (limit->p_exclusive) {
 129                 limit = plimit_copy(limit);
 130         } else {
 131                 spin_lock_wr(&limit->p_spin);
 132                 ++limit->p_refcnt;
 133                 spin_unlock_wr(&limit->p_spin);
 134         }
 135         return(limit);
 136 }
 137
 138 /*
 139  * This routine is called when the second LWP is created for a process.
 140  * The process's limit structure must be made exclusive and a copy is
 141  * made if necessary.
 142  *
 143  * If the refcnt is 1 only the LWPs associated with the caller's process
 144  * have access to the structure, and since all we do is set the exclusive
 145  * but we don't need a spinlock.
 146  *
 147  * MPSAFE
 148  */
 149 static
 150 void
 151 plimit_exclusive(struct plimit **limitp)
 152 {
 153         struct plimit *olimit = *limitp;
 154         struct plimit *nlimit;
 155
 156         if (olimit->p_refcnt == 1) {
 157                 olimit->p_exclusive = 1;
 158         } else {
 159                 nlimit = plimit_copy(olimit);
 160                 nlimit->p_exclusive = 1;
 161                 *limitp = nlimit;
 162                 spin_lock_wr(&olimit->p_spin);
 163                 if (--olimit->p_refcnt == 0) {
 164                         spin_unlock_wr(&olimit->p_spin);
 165                         kfree(olimit, M_SUBPROC);
 166                 } else {
 167                         spin_unlock_wr(&olimit->p_spin);
 168                 }
 169         }
 170 }
 171
 172 /*
 173  * Return a copy of the passed plimit.  The returned copy will have a refcnt
 174  * of 1 and p_exclusive will be cleared.
 175  *
 176  * MPSAFE
 177  */
 178 static
 179 struct plimit *
 180 plimit_copy(struct plimit *olimit)
 181 {
 182         struct plimit *nlimit;
 183
 184         nlimit = kmalloc(sizeof(struct plimit), M_SUBPROC, M_WAITOK);
 185
 186         spin_lock_rd(&olimit->p_spin);
 187         *nlimit = *olimit;
 188         spin_unlock_rd(&olimit->p_spin);
 189
 190         spin_init(&nlimit->p_spin);
 191         nlimit->p_refcnt = 1;
 192         nlimit->p_exclusive = 0;
 193         return (nlimit);
 194 }
 195
 196 /*
 197  * This routine is called to fixup a proces's p_limit structure prior
 198  * to it being modified.  If index >= 0 the specified modified is also
 199  * made.
 200  *
 201  * A limit structure is potentially shared only if the process has exactly
 202  * one LWP, otherwise it is guarenteed to be exclusive and no copy needs
 203  * to be made.  This means that we can safely replace *limitp in the copy
 204  * case.
 205  *
 206  * We call plimit_exclusive() to do all the hard work, but the result does
 207  * not actually have to be exclusive since the original was not, so just
 208  * clear p_exclusive afterwords.
 209  *
 210  * MPSAFE
 211  */
 212 void
 213 plimit_modify(struct plimit **limitp, int index, struct rlimit *rlim)
 214 {
 215         struct plimit *limit = *limitp;
 216
 217         if (limit->p_exclusive == 0 && limit->p_refcnt > 1) {
 218                 plimit_exclusive(limitp);
 219                 limit = *limitp;
 220                 limit->p_exclusive = 0;
 221         }
 222         if (index >= 0) {
 223                 spin_lock_wr(&limit->p_spin);
 224                 limit->pl_rlimit[index] = *rlim;
 225                 spin_unlock_wr(&limit->p_spin);
 226         }
 227 }
 228
 229 /*
 230  * Destroy a process's plimit structure.
 231  *
 232  * MPSAFE
 233  */
 234 void
 235 plimit_free(struct plimit **limitp)
 236 {
 237         struct plimit *limit;
 238
 239         if ((limit = *limitp) != NULL) {
 240                 *limitp = NULL;
 241
 242                 spin_lock_wr(&limit->p_spin);
 243                 if (--limit->p_refcnt == 0) {
 244                         spin_unlock_wr(&limit->p_spin);
 245                         kfree(limit, M_SUBPROC);
 246                 } else {
 247                         spin_unlock_wr(&limit->p_spin);
 248                 }
 249         }
 250 }
 251
 252 /*
 253  * Modify a resource limit (from system call)
 254  *
 255  * MPSAFE
 256  */
 257 int
 258 kern_setrlimit(u_int which, struct rlimit *limp)
 259 {
 260         struct proc *p = curproc;
 261         struct plimit *limit;
 262         struct rlimit *alimp;
 263         int error;
 264
 265         if (which >= RLIM_NLIMITS)
 266                 return (EINVAL);
 267
 268         /*
 269          * We will be modifying a resource, make a copy if necessary.
 270          */
 271         plimit_modify(&p->p_limit, -1, NULL);
 272         limit = p->p_limit;
 273         alimp = &limit->pl_rlimit[which];
 274
 275         /*
 276          * Preserve historical bugs by treating negative limits as unsigned.
 277          */
 278         if (limp->rlim_cur < 0)
 279                 limp->rlim_cur = RLIM_INFINITY;
 280         if (limp->rlim_max < 0)
 281                 limp->rlim_max = RLIM_INFINITY;
 282
 283         spin_lock_rd(&limit->p_spin);
 284         if (limp->rlim_cur > alimp->rlim_max ||
 285             limp->rlim_max > alimp->rlim_max) {
 286                 spin_unlock_rd(&limit->p_spin);
 287                 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)))
 288                         return (error);
 289         } else {
 290                 spin_unlock_rd(&limit->p_spin);
 291         }
 292         if (limp->rlim_cur > limp->rlim_max)
 293                 limp->rlim_cur = limp->rlim_max;
 294
 295         switch (which) {
 296         case RLIMIT_CPU:
 297                 spin_lock_wr(&limit->p_spin);
 298                 if (limp->rlim_cur > RLIM_INFINITY / (rlim_t)1000000)
 299                         limit->p_cpulimit = RLIM_INFINITY;
 300                 else
 301                         limit->p_cpulimit = (rlim_t)1000000 * limp->rlim_cur;
 302                 spin_unlock_wr(&limit->p_spin);
 303                 break;
 304         case RLIMIT_DATA:
 305                 if (limp->rlim_cur > maxdsiz)
 306                         limp->rlim_cur = maxdsiz;
 307                 if (limp->rlim_max > maxdsiz)
 308                         limp->rlim_max = maxdsiz;
 309                 break;
 310
 311         case RLIMIT_STACK:
 312                 if (limp->rlim_cur > maxssiz)
 313                         limp->rlim_cur = maxssiz;
 314                 if (limp->rlim_max > maxssiz)
 315                         limp->rlim_max = maxssiz;
 316                 /*
 317                  * Stack is allocated to the max at exec time with only
 318                  * "rlim_cur" bytes accessible.  If stack limit is going
 319                  * up make more accessible, if going down make inaccessible.
 320                  */
 321                 spin_lock_rd(&limit->p_spin);
 322                 if (limp->rlim_cur != alimp->rlim_cur) {
 323                         vm_offset_t addr;
 324                         vm_size_t size;
 325                         vm_prot_t prot;
 326
 327                         if (limp->rlim_cur > alimp->rlim_cur) {
 328                                 prot = VM_PROT_ALL;
 329                                 size = limp->rlim_cur - alimp->rlim_cur;
 330                                 addr = USRSTACK - limp->rlim_cur;
 331                         } else {
 332                                 prot = VM_PROT_NONE;
 333                                 size = alimp->rlim_cur - limp->rlim_cur;
 334                                 addr = USRSTACK - alimp->rlim_cur;
 335                         }
 336                         spin_unlock_rd(&limit->p_spin);
 337                         addr = trunc_page(addr);
 338                         size = round_page(size);
 339                         (void) vm_map_protect(&p->p_vmspace->vm_map,
 340                                               addr, addr+size, prot, FALSE);
 341                 } else {
 342                         spin_unlock_rd(&limit->p_spin);
 343                 }
 344                 break;
 345
 346         case RLIMIT_NOFILE:
 347                 if (limp->rlim_cur > maxfilesperproc)
 348                         limp->rlim_cur = maxfilesperproc;
 349                 if (limp->rlim_max > maxfilesperproc)
 350                         limp->rlim_max = maxfilesperproc;
 351                 break;
 352
 353         case RLIMIT_NPROC:
 354                 if (limp->rlim_cur > maxprocperuid)
 355                         limp->rlim_cur = maxprocperuid;
 356                 if (limp->rlim_max > maxprocperuid)
 357                         limp->rlim_max = maxprocperuid;
 358                 if (limp->rlim_cur < 1)
 359                         limp->rlim_cur = 1;
 360                 if (limp->rlim_max < 1)
 361                         limp->rlim_max = 1;
 362                 break;
 363         case RLIMIT_POSIXLOCKS:
 364                 if (limp->rlim_cur > maxposixlocksperuid)
 365                         limp->rlim_cur = maxposixlocksperuid;
 366                 if (limp->rlim_max > maxposixlocksperuid)
 367                         limp->rlim_max = maxposixlocksperuid;
 368                 break;
 369         }
 370         spin_lock_wr(&limit->p_spin);
 371         *alimp = *limp;
 372         spin_unlock_wr(&limit->p_spin);
 373         return (0);
 374 }
 375
 376 /*
 377  * The rlimit indexed by which is returned in the second argument.
 378  *
 379  * MPSAFE
 380  */
 381 int
 382 kern_getrlimit(u_int which, struct rlimit *limp)
 383 {
 384         struct proc *p = curproc;
 385         struct plimit *limit;
 386
 387         if (which >= RLIM_NLIMITS)
 388                 return (EINVAL);
 389
 390         limit = p->p_limit;
 391         spin_lock_rd(&limit->p_spin);
 392         *limp = p->p_rlimit[which];
 393         spin_unlock_rd(&limit->p_spin);
 394         return (0);
 395 }
 396
 397 /*
 398  * Determine if the cpu limit has been reached and return an operations
 399  * code for the caller to perform.
 400  *
 401  * MPSAFE
 402  */
 403 int
 404 plimit_testcpulimit(struct plimit *limit, u_int64_t ttime)
 405 {
 406         struct rlimit *rlim;
 407         int mode;
 408
 409         mode = PLIMIT_TESTCPU_OK;
 410         if (limit->p_cpulimit != RLIM_INFINITY) {
 411                 spin_lock_rd(&limit->p_spin);
 412                 if (ttime > limit->p_cpulimit) {
 413                         rlim = &limit->pl_rlimit[RLIMIT_CPU];
 414                         if (ttime / (rlim_t)1000000 >= rlim->rlim_max + 5) {
 415                                 mode = PLIMIT_TESTCPU_KILL;
 416                         } else {
 417                                 mode = PLIMIT_TESTCPU_XCPU;
 418                         }
 419                 }
 420                 spin_unlock_rd(&limit->p_spin);
 421         }
 422         return(mode);
 423 }
 424