sys/kern/kern_memio.c

   1 /*-
   2  * Copyright (c) 1988 University of Utah.
   3  * Copyright (c) 1982, 1986, 1990 The Regents of the University of California.
   4  * All rights reserved.
   5  *
   6  * This code is derived from software contributed to Berkeley by
   7  * the Systems Programming Group of the University of Utah Computer
   8  * Science Department, and code derived from software contributed to
   9  * Berkeley by William Jolitz.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 3. Neither the name of the University nor the names of its contributors
  20  *    may be used to endorse or promote products derived from this software
  21  *    without specific prior written permission.
  22  *
  23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  33  * SUCH DAMAGE.
  34  *
  35  *      from: Utah $Hdr: mem.c 1.13 89/10/08$
  36  *      from: @(#)mem.c 7.2 (Berkeley) 5/9/91
  37  * $FreeBSD: src/sys/i386/i386/mem.c,v 1.79.2.9 2003/01/04 22:58:01 njl Exp $
  38  */
  39
  40 /*
  41  * Memory special file
  42  */
  43
  44 #include <sys/param.h>
  45 #include <sys/systm.h>
  46 #include <sys/buf.h>
  47 #include <sys/conf.h>
  48 #include <sys/fcntl.h>
  49 #include <sys/filio.h>
  50 #include <sys/kernel.h>
  51 #include <sys/malloc.h>
  52 #include <sys/memrange.h>
  53 #include <sys/proc.h>
  54 #include <sys/priv.h>
  55 #include <sys/random.h>
  56 #include <sys/signalvar.h>
  57 #include <sys/uio.h>
  58 #include <sys/vnode.h>
  59 #include <sys/sysctl.h>
  60
  61 #include <sys/signal2.h>
  62
  63 #include <vm/vm.h>
  64 #include <vm/pmap.h>
  65 #include <vm/vm_extern.h>
  66
  67
  68 static  d_open_t        mmopen;
  69 static  d_close_t       mmclose;
  70 static  d_read_t        mmread;
  71 static  d_write_t       mmwrite;
  72 static  d_ioctl_t       mmioctl;
  73 #if 0
  74 static  d_mmap_t        memmmap;
  75 #endif
  76 static  d_kqfilter_t    mmkqfilter;
  77 static int memuksmap(cdev_t dev, vm_page_t fake);
  78
  79 #define CDEV_MAJOR 2
  80 static struct dev_ops mem_ops = {
  81         { "mem", 0, D_MPSAFE | D_QUICK },
  82         .d_open =       mmopen,
  83         .d_close =      mmclose,
  84         .d_read =       mmread,
  85         .d_write =      mmwrite,
  86         .d_ioctl =      mmioctl,
  87         .d_kqfilter =   mmkqfilter,
  88 #if 0
  89         .d_mmap =       memmmap,
  90 #endif
  91         .d_uksmap =     memuksmap
  92 };
  93
  94 static struct dev_ops mem_ops_mem = {
  95         { "mem", 0, D_MEM | D_MPSAFE | D_QUICK },
  96         .d_open =       mmopen,
  97         .d_close =      mmclose,
  98         .d_read =       mmread,
  99         .d_write =      mmwrite,
 100         .d_ioctl =      mmioctl,
 101         .d_kqfilter =   mmkqfilter,
 102 #if 0
 103         .d_mmap =       memmmap,
 104 #endif
 105         .d_uksmap =     memuksmap
 106 };
 107
 108 static struct dev_ops mem_ops_noq = {
 109         { "mem", 0, D_MPSAFE },
 110         .d_open =       mmopen,
 111         .d_close =      mmclose,
 112         .d_read =       mmread,
 113         .d_write =      mmwrite,
 114         .d_ioctl =      mmioctl,
 115         .d_kqfilter =   mmkqfilter,
 116 #if 0
 117         .d_mmap =       memmmap,
 118 #endif
 119         .d_uksmap =     memuksmap
 120 };
 121
 122 static int rand_bolt;
 123 static caddr_t  zbuf;
 124 static cdev_t   zerodev = NULL;
 125 static struct lock mem_lock = LOCK_INITIALIZER("memlk", 0, 0);
 126
 127 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors");
 128 static int mem_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
 129 static int random_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
 130
 131 struct mem_range_softc mem_range_softc;
 132
 133 static int seedenable;
 134 SYSCTL_INT(_kern, OID_AUTO, seedenable, CTLFLAG_RW, &seedenable, 0, "");
 135
 136 static int
 137 mmopen(struct dev_open_args *ap)
 138 {
 139         cdev_t dev = ap->a_head.a_dev;
 140         int error;
 141
 142         switch (minor(dev)) {
 143         case 0:
 144         case 1:
 145                 /*
 146                  * /dev/mem and /dev/kmem
 147                  */
 148                 if (ap->a_oflags & FWRITE) {
 149                         if (securelevel > 0 || kernel_mem_readonly)
 150                                 return (EPERM);
 151                 }
 152                 error = 0;
 153                 break;
 154         case 6:
 155                 /*
 156                  * /dev/kpmap can only be opened for reading.
 157                  */
 158                 if (ap->a_oflags & FWRITE)
 159                         return (EPERM);
 160                 error = 0;
 161                 break;
 162         case 14:
 163                 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0);
 164                 if (error != 0)
 165                         break;
 166                 if (securelevel > 0 || kernel_mem_readonly) {
 167                         error = EPERM;
 168                         break;
 169                 }
 170                 error = cpu_set_iopl();
 171                 break;
 172         default:
 173                 error = 0;
 174                 break;
 175         }
 176         return (error);
 177 }
 178
 179 static int
 180 mmclose(struct dev_close_args *ap)
 181 {
 182         cdev_t dev = ap->a_head.a_dev;
 183         int error;
 184
 185         switch (minor(dev)) {
 186         case 14:
 187                 error = cpu_clr_iopl();
 188                 break;
 189         default:
 190                 error = 0;
 191                 break;
 192         }
 193         return (error);
 194 }
 195
 196
 197 static int
 198 mmrw(cdev_t dev, struct uio *uio, int flags)
 199 {
 200         int o;
 201         u_int c;
 202         u_int poolsize;
 203         u_long v;
 204         struct iovec *iov;
 205         int error = 0;
 206         caddr_t buf = NULL;
 207
 208         while (uio->uio_resid > 0 && error == 0) {
 209                 iov = uio->uio_iov;
 210                 if (iov->iov_len == 0) {
 211                         uio->uio_iov++;
 212                         uio->uio_iovcnt--;
 213                         if (uio->uio_iovcnt < 0)
 214                                 panic("mmrw");
 215                         continue;
 216                 }
 217                 switch (minor(dev)) {
 218                 case 0:
 219                         /*
 220                          * minor device 0 is physical memory, /dev/mem
 221                          */
 222                         v = uio->uio_offset;
 223                         v &= ~(long)PAGE_MASK;
 224                         pmap_kenter((vm_offset_t)ptvmmap, v);
 225                         o = (int)uio->uio_offset & PAGE_MASK;
 226                         c = (u_int)(PAGE_SIZE - ((uintptr_t)iov->iov_base & PAGE_MASK));
 227                         c = min(c, (u_int)(PAGE_SIZE - o));
 228                         c = min(c, (u_int)iov->iov_len);
 229                         error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
 230                         pmap_kremove((vm_offset_t)ptvmmap);
 231                         continue;
 232
 233                 case 1: {
 234                         /*
 235                          * minor device 1 is kernel memory, /dev/kmem
 236                          */
 237                         vm_offset_t saddr, eaddr;
 238                         int prot;
 239
 240                         c = iov->iov_len;
 241
 242                         /*
 243                          * Make sure that all of the pages are currently
 244                          * resident so that we don't create any zero-fill
 245                          * pages.
 246                          */
 247                         saddr = trunc_page(uio->uio_offset);
 248                         eaddr = round_page(uio->uio_offset + c);
 249                         if (saddr > eaddr)
 250                                 return EFAULT;
 251
 252                         /*
 253                          * Make sure the kernel addresses are mapped.
 254                          * platform_direct_mapped() can be used to bypass
 255                          * default mapping via the page table (virtual kernels
 256                          * contain a lot of out-of-band data).
 257                          */
 258                         prot = VM_PROT_READ;
 259                         if (uio->uio_rw != UIO_READ)
 260                                 prot |= VM_PROT_WRITE;
 261                         error = kvm_access_check(saddr, eaddr, prot);
 262                         if (error)
 263                                 return (error);
 264                         error = uiomove((caddr_t)(vm_offset_t)uio->uio_offset,
 265                                         (int)c, uio);
 266                         continue;
 267                 }
 268                 case 2:
 269                         /*
 270                          * minor device 2 (/dev/null) is EOF/RATHOLE
 271                          */
 272                         if (uio->uio_rw == UIO_READ)
 273                                 return (0);
 274                         c = iov->iov_len;
 275                         break;
 276                 case 3:
 277                         /*
 278                          * minor device 3 (/dev/random) is source of filth
 279                          * on read, seeder on write
 280                          */
 281                         if (buf == NULL)
 282                                 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
 283                         c = min(iov->iov_len, PAGE_SIZE);
 284                         if (uio->uio_rw == UIO_WRITE) {
 285                                 error = uiomove(buf, (int)c, uio);
 286                                 if (error == 0 &&
 287                                     seedenable &&
 288                                     securelevel <= 0) {
 289                                         error = add_buffer_randomness_src(buf, c, RAND_SRC_SEEDING);
 290                                 } else if (error == 0) {
 291                                         error = EPERM;
 292                                 }
 293                         } else {
 294                                 poolsize = read_random(buf, c);
 295                                 if (poolsize == 0) {
 296                                         if (buf)
 297                                                 kfree(buf, M_TEMP);
 298                                         if ((flags & IO_NDELAY) != 0)
 299                                                 return (EWOULDBLOCK);
 300                                         return (0);
 301                                 }
 302                                 c = min(c, poolsize);
 303                                 error = uiomove(buf, (int)c, uio);
 304                         }
 305                         continue;
 306                 case 4:
 307                         /*
 308                          * minor device 4 (/dev/urandom) is source of muck
 309                          * on read, writes are disallowed.
 310                          */
 311                         c = min(iov->iov_len, PAGE_SIZE);
 312                         if (uio->uio_rw == UIO_WRITE) {
 313                                 error = EPERM;
 314                                 break;
 315                         }
 316                         if (CURSIG(curthread->td_lwp) != 0) {
 317                                 /*
 318                                  * Use tsleep() to get the error code right.
 319                                  * It should return immediately.
 320                                  */
 321                                 error = tsleep(&rand_bolt, PCATCH, "urand", 1);
 322                                 if (error != 0 && error != EWOULDBLOCK)
 323                                         continue;
 324                         }
 325                         if (buf == NULL)
 326                                 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
 327                         poolsize = read_random_unlimited(buf, c);
 328                         c = min(c, poolsize);
 329                         error = uiomove(buf, (int)c, uio);
 330                         continue;
 331                 /* case 5: read/write not supported, mmap only */
 332                 /* case 6: read/write not supported, mmap only */
 333                 case 12:
 334                         /*
 335                          * minor device 12 (/dev/zero) is source of nulls
 336                          * on read, write are disallowed.
 337                          */
 338                         if (uio->uio_rw == UIO_WRITE) {
 339                                 c = iov->iov_len;
 340                                 break;
 341                         }
 342                         if (zbuf == NULL) {
 343                                 zbuf = (caddr_t)kmalloc(PAGE_SIZE, M_TEMP,
 344                                     M_WAITOK | M_ZERO);
 345                         }
 346                         c = min(iov->iov_len, PAGE_SIZE);
 347                         error = uiomove(zbuf, (int)c, uio);
 348                         continue;
 349                 default:
 350                         return (ENODEV);
 351                 }
 352                 if (error)
 353                         break;
 354                 iov->iov_base = (char *)iov->iov_base + c;
 355                 iov->iov_len -= c;
 356                 uio->uio_offset += c;
 357                 uio->uio_resid -= c;
 358         }
 359         if (buf)
 360                 kfree(buf, M_TEMP);
 361         return (error);
 362 }
 363
 364 static int
 365 mmread(struct dev_read_args *ap)
 366 {
 367         return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
 368 }
 369
 370 static int
 371 mmwrite(struct dev_write_args *ap)
 372 {
 373         return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
 374 }
 375
 376 /*******************************************************\
 377 * allow user processes to MMAP some memory sections     *
 378 * instead of going through read/write                   *
 379 \*******************************************************/
 380
 381 static int user_kernel_mapping(int num, vm_ooffset_t offset,
 382                                 vm_ooffset_t *resultp);
 383
 384 #if 0
 385
 386 static int
 387 memmmap(struct dev_mmap_args *ap)
 388 {
 389         cdev_t dev = ap->a_head.a_dev;
 390         vm_ooffset_t result;
 391         int error;
 392
 393         switch (minor(dev)) {
 394         case 0:
 395                 /*
 396                  * minor device 0 is physical memory
 397                  */
 398                 ap->a_result = atop(ap->a_offset);
 399                 error = 0;
 400                 break;
 401         case 1:
 402                 /*
 403                  * minor device 1 is kernel memory
 404                  */
 405                 ap->a_result = atop(vtophys(ap->a_offset));
 406                 error = 0;
 407                 break;
 408         case 5:
 409         case 6:
 410                 /*
 411                  * minor device 5 is /dev/upmap (see sys/upmap.h)
 412                  * minor device 6 is /dev/kpmap (see sys/upmap.h)
 413                  */
 414                 result = 0;
 415                 error = user_kernel_mapping(minor(dev), ap->a_offset, &result);
 416                 ap->a_result = atop(result);
 417                 break;
 418         default:
 419                 error = EINVAL;
 420                 break;
 421         }
 422         return error;
 423 }
 424
 425 #endif
 426
 427 static int
 428 memuksmap(cdev_t dev, vm_page_t fake)
 429 {
 430         vm_ooffset_t result;
 431         int error;
 432
 433         switch (minor(dev)) {
 434         case 0:
 435                 /*
 436                  * minor device 0 is physical memory
 437                  */
 438                 fake->phys_addr = ptoa(fake->pindex);
 439                 error = 0;
 440                 break;
 441         case 1:
 442                 /*
 443                  * minor device 1 is kernel memory
 444                  */
 445                 fake->phys_addr = vtophys(ptoa(fake->pindex));
 446                 error = 0;
 447                 break;
 448         case 5:
 449         case 6:
 450                 /*
 451                  * minor device 5 is /dev/upmap (see sys/upmap.h)
 452                  * minor device 6 is /dev/kpmap (see sys/upmap.h)
 453                  */
 454                 result = 0;
 455                 error = user_kernel_mapping(minor(dev),
 456                                             ptoa(fake->pindex), &result);
 457                 fake->phys_addr = result;
 458                 break;
 459         default:
 460                 error = EINVAL;
 461                 break;
 462         }
 463         return error;
 464 }
 465
 466 static int
 467 mmioctl(struct dev_ioctl_args *ap)
 468 {
 469         cdev_t dev = ap->a_head.a_dev;
 470         int error;
 471
 472         lockmgr(&mem_lock, LK_EXCLUSIVE);
 473
 474         switch (minor(dev)) {
 475         case 0:
 476                 error = mem_ioctl(dev, ap->a_cmd, ap->a_data,
 477                                   ap->a_fflag, ap->a_cred);
 478                 break;
 479         case 3:
 480         case 4:
 481                 error = random_ioctl(dev, ap->a_cmd, ap->a_data,
 482                                      ap->a_fflag, ap->a_cred);
 483                 break;
 484         default:
 485                 error = ENODEV;
 486                 break;
 487         }
 488
 489         lockmgr(&mem_lock, LK_RELEASE);
 490
 491         return (error);
 492 }
 493
 494 /*
 495  * Operations for changing memory attributes.
 496  *
 497  * This is basically just an ioctl shim for mem_range_attr_get
 498  * and mem_range_attr_set.
 499  */
 500 static int
 501 mem_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
 502 {
 503         int nd, error = 0;
 504         struct mem_range_op *mo = (struct mem_range_op *)data;
 505         struct mem_range_desc *md;
 506
 507         /* is this for us? */
 508         if ((cmd != MEMRANGE_GET) &&
 509             (cmd != MEMRANGE_SET))
 510                 return (ENOTTY);
 511
 512         /* any chance we can handle this? */
 513         if (mem_range_softc.mr_op == NULL)
 514                 return (EOPNOTSUPP);
 515
 516         /* do we have any descriptors? */
 517         if (mem_range_softc.mr_ndesc == 0)
 518                 return (ENXIO);
 519
 520         switch (cmd) {
 521         case MEMRANGE_GET:
 522                 nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
 523                 if (nd > 0) {
 524                         md = (struct mem_range_desc *)
 525                                 kmalloc(nd * sizeof(struct mem_range_desc),
 526                                        M_MEMDESC, M_WAITOK);
 527                         error = mem_range_attr_get(md, &nd);
 528                         if (!error)
 529                                 error = copyout(md, mo->mo_desc,
 530                                         nd * sizeof(struct mem_range_desc));
 531                         kfree(md, M_MEMDESC);
 532                 } else {
 533                         nd = mem_range_softc.mr_ndesc;
 534                 }
 535                 mo->mo_arg[0] = nd;
 536                 break;
 537
 538         case MEMRANGE_SET:
 539                 md = (struct mem_range_desc *)kmalloc(sizeof(struct mem_range_desc),
 540                                                     M_MEMDESC, M_WAITOK);
 541                 error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
 542                 /* clamp description string */
 543                 md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
 544                 if (error == 0)
 545                         error = mem_range_attr_set(md, &mo->mo_arg[0]);
 546                 kfree(md, M_MEMDESC);
 547                 break;
 548         }
 549         return (error);
 550 }
 551
 552 /*
 553  * Implementation-neutral, kernel-callable functions for manipulating
 554  * memory range attributes.
 555  */
 556 int
 557 mem_range_attr_get(struct mem_range_desc *mrd, int *arg)
 558 {
 559         /* can we handle this? */
 560         if (mem_range_softc.mr_op == NULL)
 561                 return (EOPNOTSUPP);
 562
 563         if (*arg == 0) {
 564                 *arg = mem_range_softc.mr_ndesc;
 565         } else {
 566                 bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc));
 567         }
 568         return (0);
 569 }
 570
 571 int
 572 mem_range_attr_set(struct mem_range_desc *mrd, int *arg)
 573 {
 574         /* can we handle this? */
 575         if (mem_range_softc.mr_op == NULL)
 576                 return (EOPNOTSUPP);
 577
 578         return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
 579 }
 580
 581 void
 582 mem_range_AP_init(void)
 583 {
 584         if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
 585                 mem_range_softc.mr_op->initAP(&mem_range_softc);
 586 }
 587
 588 static int
 589 random_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
 590 {
 591         int error;
 592         int intr;
 593
 594         /*
 595          * Even inspecting the state is privileged, since it gives a hint
 596          * about how easily the randomness might be guessed.
 597          */
 598         error = 0;
 599
 600         switch (cmd) {
 601         /* Really handled in upper layer */
 602         case FIOASYNC:
 603                 break;
 604         case MEM_SETIRQ:
 605                 intr = *(int16_t *)data;
 606                 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
 607                         break;
 608                 if (intr < 0 || intr >= MAX_INTS)
 609                         return (EINVAL);
 610                 register_randintr(intr);
 611                 break;
 612         case MEM_CLEARIRQ:
 613                 intr = *(int16_t *)data;
 614                 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
 615                         break;
 616                 if (intr < 0 || intr >= MAX_INTS)
 617                         return (EINVAL);
 618                 unregister_randintr(intr);
 619                 break;
 620         case MEM_RETURNIRQ:
 621                 error = ENOTSUP;
 622                 break;
 623         case MEM_FINDIRQ:
 624                 intr = *(int16_t *)data;
 625                 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
 626                         break;
 627                 if (intr < 0 || intr >= MAX_INTS)
 628                         return (EINVAL);
 629                 intr = next_registered_randintr(intr);
 630                 if (intr == MAX_INTS)
 631                         return (ENOENT);
 632                 *(u_int16_t *)data = intr;
 633                 break;
 634         default:
 635                 error = ENOTSUP;
 636                 break;
 637         }
 638         return (error);
 639 }
 640
 641 static int
 642 mm_filter_read(struct knote *kn, long hint)
 643 {
 644         return (1);
 645 }
 646
 647 static int
 648 mm_filter_write(struct knote *kn, long hint)
 649 {
 650         return (1);
 651 }
 652
 653 static void
 654 dummy_filter_detach(struct knote *kn) {}
 655
 656 /* Implemented in kern_nrandom.c */
 657 static struct filterops random_read_filtops =
 658         { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, random_filter_read };
 659
 660 static struct filterops mm_read_filtops =
 661         { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_read };
 662
 663 static struct filterops mm_write_filtops =
 664         { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_write };
 665
 666 static int
 667 mmkqfilter(struct dev_kqfilter_args *ap)
 668 {
 669         struct knote *kn = ap->a_kn;
 670         cdev_t dev = ap->a_head.a_dev;
 671
 672         ap->a_result = 0;
 673         switch (kn->kn_filter) {
 674         case EVFILT_READ:
 675                 switch (minor(dev)) {
 676                 case 3:
 677                         kn->kn_fop = &random_read_filtops;
 678                         break;
 679                 default:
 680                         kn->kn_fop = &mm_read_filtops;
 681                         break;
 682                 }
 683                 break;
 684         case EVFILT_WRITE:
 685                 kn->kn_fop = &mm_write_filtops;
 686                 break;
 687         default:
 688                 ap->a_result = EOPNOTSUPP;
 689                 return (0);
 690         }
 691
 692         return (0);
 693 }
 694
 695 int
 696 iszerodev(cdev_t dev)
 697 {
 698         return (zerodev == dev);
 699 }
 700
 701 /*
 702  * /dev/upmap and /dev/kpmap.
 703  */
 704 static int
 705 user_kernel_mapping(int num, vm_ooffset_t offset, vm_ooffset_t *resultp)
 706 {
 707         struct proc *p;
 708         int error;
 709         int invfork;
 710
 711         if ((p = curproc) == NULL)
 712                 return (EINVAL);
 713
 714         /*
 715          * If this is a child currently in vfork the pmap is shared with
 716          * the parent!  We need to actually set-up the parent's p_upmap,
 717          * not the child's, and we need to set the invfork flag.  Userland
 718          * will probably adjust its static state so it must be consistent
 719          * with the parent or userland will be really badly confused.
 720          *
 721          * (this situation can happen when user code in vfork() calls
 722          *  libc's getpid() or some other function which then decides
 723          *  it wants the upmap).
 724          */
 725         if (p->p_flags & P_PPWAIT) {
 726                 p = p->p_pptr;
 727                 if (p == NULL)
 728                         return (EINVAL);
 729                 invfork = 1;
 730         } else {
 731                 invfork = 0;
 732         }
 733
 734         error = EINVAL;
 735
 736         switch(num) {
 737         case 5:
 738                 /*
 739                  * /dev/upmap - maps RW per-process shared user-kernel area.
 740                  */
 741                 if (p->p_upmap == NULL)
 742                         proc_usermap(p, invfork);
 743                 else if (invfork)
 744                         p->p_upmap->invfork = invfork;
 745
 746                 if (p->p_upmap &&
 747                     offset < roundup2(sizeof(*p->p_upmap), PAGE_SIZE)) {
 748                         /* only good for current process */
 749                         *resultp = pmap_kextract((vm_offset_t)p->p_upmap +
 750                                                  offset);
 751                         error = 0;
 752                 }
 753                 break;
 754         case 6:
 755                 /*
 756                  * /dev/kpmap - maps RO shared kernel global page
 757                  */
 758                 if (kpmap &&
 759                     offset < roundup2(sizeof(*kpmap), PAGE_SIZE)) {
 760                         *resultp = pmap_kextract((vm_offset_t)kpmap +
 761                                                  offset);
 762                         error = 0;
 763                 }
 764                 break;
 765         default:
 766                 break;
 767         }
 768         return error;
 769 }
 770
 771 static void
 772 mem_drvinit(void *unused)
 773 {
 774
 775         /* Initialise memory range handling */
 776         if (mem_range_softc.mr_op != NULL)
 777                 mem_range_softc.mr_op->init(&mem_range_softc);
 778
 779         make_dev(&mem_ops_mem, 0, UID_ROOT, GID_KMEM, 0640, "mem");
 780         make_dev(&mem_ops_mem, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
 781         make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null");
 782         make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random");
 783         make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
 784         make_dev(&mem_ops, 5, UID_ROOT, GID_WHEEL, 0666, "upmap");
 785         make_dev(&mem_ops, 6, UID_ROOT, GID_WHEEL, 0444, "kpmap");
 786         zerodev = make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
 787         make_dev(&mem_ops_noq, 14, UID_ROOT, GID_WHEEL, 0600, "io");
 788 }
 789
 790 SYSINIT(memdev, SI_SUB_DRIVERS, SI_ORDER_MIDDLE + CDEV_MAJOR, mem_drvinit,
 791     NULL);
 792