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/interrupt.h>
  51 #include <sys/kernel.h>
  52 #include <sys/malloc.h>
  53 #include <sys/memrange.h>
  54 #include <sys/proc.h>
  55 #include <sys/priv.h>
  56 #include <sys/random.h>
  57 #include <sys/signalvar.h>
  58 #include <sys/uio.h>
  59 #include <sys/vnode.h>
  60 #include <sys/sysctl.h>
  61
  62 #include <sys/signal2.h>
  63
  64 #include <vm/vm.h>
  65 #include <vm/pmap.h>
  66 #include <vm/vm_extern.h>
  67
  68
  69 static  d_open_t        mmopen;
  70 static  d_close_t       mmclose;
  71 static  d_read_t        mmread;
  72 static  d_write_t       mmwrite;
  73 static  d_ioctl_t       mmioctl;
  74 #if 0
  75 static  d_mmap_t        memmmap;
  76 #endif
  77 static  d_kqfilter_t    mmkqfilter;
  78 static int memuksmap(cdev_t dev, vm_page_t fake);
  79
  80 #define CDEV_MAJOR 2
  81 static struct dev_ops mem_ops = {
  82         { "mem", 0, D_MPSAFE | D_QUICK },
  83         .d_open =       mmopen,
  84         .d_close =      mmclose,
  85         .d_read =       mmread,
  86         .d_write =      mmwrite,
  87         .d_ioctl =      mmioctl,
  88         .d_kqfilter =   mmkqfilter,
  89 #if 0
  90         .d_mmap =       memmmap,
  91 #endif
  92         .d_uksmap =     memuksmap
  93 };
  94
  95 static struct dev_ops mem_ops_mem = {
  96         { "mem", 0, D_MEM | D_MPSAFE | D_QUICK },
  97         .d_open =       mmopen,
  98         .d_close =      mmclose,
  99         .d_read =       mmread,
 100         .d_write =      mmwrite,
 101         .d_ioctl =      mmioctl,
 102         .d_kqfilter =   mmkqfilter,
 103 #if 0
 104         .d_mmap =       memmmap,
 105 #endif
 106         .d_uksmap =     memuksmap
 107 };
 108
 109 static struct dev_ops mem_ops_noq = {
 110         { "mem", 0, D_MPSAFE },
 111         .d_open =       mmopen,
 112         .d_close =      mmclose,
 113         .d_read =       mmread,
 114         .d_write =      mmwrite,
 115         .d_ioctl =      mmioctl,
 116         .d_kqfilter =   mmkqfilter,
 117 #if 0
 118         .d_mmap =       memmmap,
 119 #endif
 120         .d_uksmap =     memuksmap
 121 };
 122
 123 static int rand_bolt;
 124 static caddr_t  zbuf;
 125 static cdev_t   zerodev = NULL;
 126 static struct lock mem_lock = LOCK_INITIALIZER("memlk", 0, 0);
 127
 128 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors");
 129 static int mem_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
 130 static int random_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
 131
 132 struct mem_range_softc mem_range_softc;
 133
 134 static int seedenable;
 135 SYSCTL_INT(_kern, OID_AUTO, seedenable, CTLFLAG_RW, &seedenable, 0, "");
 136
 137 static int
 138 mmopen(struct dev_open_args *ap)
 139 {
 140         cdev_t dev = ap->a_head.a_dev;
 141         int error;
 142
 143         switch (minor(dev)) {
 144         case 0:
 145         case 1:
 146                 /*
 147                  * /dev/mem and /dev/kmem
 148                  */
 149                 if (ap->a_oflags & FWRITE) {
 150                         if (securelevel > 0 || kernel_mem_readonly)
 151                                 return (EPERM);
 152                 }
 153                 error = 0;
 154                 break;
 155         case 6:
 156                 /*
 157                  * /dev/kpmap can only be opened for reading.
 158                  */
 159                 if (ap->a_oflags & FWRITE)
 160                         return (EPERM);
 161                 error = 0;
 162                 break;
 163         case 14:
 164                 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0);
 165                 if (error != 0)
 166                         break;
 167                 if (securelevel > 0 || kernel_mem_readonly) {
 168                         error = EPERM;
 169                         break;
 170                 }
 171                 error = cpu_set_iopl();
 172                 break;
 173         default:
 174                 error = 0;
 175                 break;
 176         }
 177         return (error);
 178 }
 179
 180 static int
 181 mmclose(struct dev_close_args *ap)
 182 {
 183         cdev_t dev = ap->a_head.a_dev;
 184         int error;
 185
 186         switch (minor(dev)) {
 187         case 14:
 188                 error = cpu_clr_iopl();
 189                 break;
 190         default:
 191                 error = 0;
 192                 break;
 193         }
 194         return (error);
 195 }
 196
 197
 198 static int
 199 mmrw(cdev_t dev, struct uio *uio, int flags)
 200 {
 201         int o;
 202         u_int c;
 203         u_int poolsize;
 204         u_long v;
 205         struct iovec *iov;
 206         int error = 0;
 207         caddr_t buf = NULL;
 208
 209         while (uio->uio_resid > 0 && error == 0) {
 210                 iov = uio->uio_iov;
 211                 if (iov->iov_len == 0) {
 212                         uio->uio_iov++;
 213                         uio->uio_iovcnt--;
 214                         if (uio->uio_iovcnt < 0)
 215                                 panic("mmrw");
 216                         continue;
 217                 }
 218                 switch (minor(dev)) {
 219                 case 0:
 220                         /*
 221                          * minor device 0 is physical memory, /dev/mem
 222                          */
 223                         v = uio->uio_offset;
 224                         v &= ~(long)PAGE_MASK;
 225                         pmap_kenter((vm_offset_t)ptvmmap, v);
 226                         o = (int)uio->uio_offset & PAGE_MASK;
 227                         c = (u_int)(PAGE_SIZE - ((uintptr_t)iov->iov_base & PAGE_MASK));
 228                         c = min(c, (u_int)(PAGE_SIZE - o));
 229                         c = min(c, (u_int)iov->iov_len);
 230                         error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
 231                         pmap_kremove((vm_offset_t)ptvmmap);
 232                         continue;
 233
 234                 case 1: {
 235                         /*
 236                          * minor device 1 is kernel memory, /dev/kmem
 237                          */
 238                         vm_offset_t saddr, eaddr;
 239                         int prot;
 240
 241                         c = iov->iov_len;
 242
 243                         /*
 244                          * Make sure that all of the pages are currently
 245                          * resident so that we don't create any zero-fill
 246                          * pages.
 247                          */
 248                         saddr = trunc_page(uio->uio_offset);
 249                         eaddr = round_page(uio->uio_offset + c);
 250                         if (saddr > eaddr)
 251                                 return EFAULT;
 252
 253                         /*
 254                          * Make sure the kernel addresses are mapped.
 255                          * platform_direct_mapped() can be used to bypass
 256                          * default mapping via the page table (virtual kernels
 257                          * contain a lot of out-of-band data).
 258                          */
 259                         prot = VM_PROT_READ;
 260                         if (uio->uio_rw != UIO_READ)
 261                                 prot |= VM_PROT_WRITE;
 262                         error = kvm_access_check(saddr, eaddr, prot);
 263                         if (error)
 264                                 return (error);
 265                         error = uiomove((caddr_t)(vm_offset_t)uio->uio_offset,
 266                                         (int)c, uio);
 267                         continue;
 268                 }
 269                 case 2:
 270                         /*
 271                          * minor device 2 (/dev/null) is EOF/RATHOLE
 272                          */
 273                         if (uio->uio_rw == UIO_READ)
 274                                 return (0);
 275                         c = iov->iov_len;
 276                         break;
 277                 case 3:
 278                         /*
 279                          * minor device 3 (/dev/random) is source of filth
 280                          * on read, seeder on write
 281                          */
 282                         if (buf == NULL)
 283                                 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
 284                         c = min(iov->iov_len, PAGE_SIZE);
 285                         if (uio->uio_rw == UIO_WRITE) {
 286                                 error = uiomove(buf, (int)c, uio);
 287                                 if (error == 0 &&
 288                                     seedenable &&
 289                                     securelevel <= 0) {
 290                                         error = add_buffer_randomness_src(buf, c, RAND_SRC_SEEDING);
 291                                 } else if (error == 0) {
 292                                         error = EPERM;
 293                                 }
 294                         } else {
 295                                 poolsize = read_random(buf, c);
 296                                 if (poolsize == 0) {
 297                                         if (buf)
 298                                                 kfree(buf, M_TEMP);
 299                                         if ((flags & IO_NDELAY) != 0)
 300                                                 return (EWOULDBLOCK);
 301                                         return (0);
 302                                 }
 303                                 c = min(c, poolsize);
 304                                 error = uiomove(buf, (int)c, uio);
 305                         }
 306                         continue;
 307                 case 4:
 308                         /*
 309                          * minor device 4 (/dev/urandom) is source of muck
 310                          * on read, writes are disallowed.
 311                          */
 312                         c = min(iov->iov_len, PAGE_SIZE);
 313                         if (uio->uio_rw == UIO_WRITE) {
 314                                 error = EPERM;
 315                                 break;
 316                         }
 317                         if (CURSIG(curthread->td_lwp) != 0) {
 318                                 /*
 319                                  * Use tsleep() to get the error code right.
 320                                  * It should return immediately.
 321                                  */
 322                                 error = tsleep(&rand_bolt, PCATCH, "urand", 1);
 323                                 if (error != 0 && error != EWOULDBLOCK)
 324                                         continue;
 325                         }
 326                         if (buf == NULL)
 327                                 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
 328                         poolsize = read_random_unlimited(buf, c);
 329                         c = min(c, poolsize);
 330                         error = uiomove(buf, (int)c, uio);
 331                         continue;
 332                 /* case 5: read/write not supported, mmap only */
 333                 /* case 6: read/write not supported, mmap only */
 334                 case 12:
 335                         /*
 336                          * minor device 12 (/dev/zero) is source of nulls
 337                          * on read, write are disallowed.
 338                          */
 339                         if (uio->uio_rw == UIO_WRITE) {
 340                                 c = iov->iov_len;
 341                                 break;
 342                         }
 343                         if (zbuf == NULL) {
 344                                 zbuf = (caddr_t)kmalloc(PAGE_SIZE, M_TEMP,
 345                                     M_WAITOK | M_ZERO);
 346                         }
 347                         c = min(iov->iov_len, PAGE_SIZE);
 348                         error = uiomove(zbuf, (int)c, uio);
 349                         continue;
 350                 default:
 351                         return (ENODEV);
 352                 }
 353                 if (error)
 354                         break;
 355                 iov->iov_base = (char *)iov->iov_base + c;
 356                 iov->iov_len -= c;
 357                 uio->uio_offset += c;
 358                 uio->uio_resid -= c;
 359         }
 360         if (buf)
 361                 kfree(buf, M_TEMP);
 362         return (error);
 363 }
 364
 365 static int
 366 mmread(struct dev_read_args *ap)
 367 {
 368         return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
 369 }
 370
 371 static int
 372 mmwrite(struct dev_write_args *ap)
 373 {
 374         return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
 375 }
 376
 377 /*******************************************************\
 378 * allow user processes to MMAP some memory sections     *
 379 * instead of going through read/write                   *
 380 \*******************************************************/
 381
 382 static int user_kernel_mapping(int num, vm_ooffset_t offset,
 383                                 vm_ooffset_t *resultp);
 384
 385 #if 0
 386
 387 static int
 388 memmmap(struct dev_mmap_args *ap)
 389 {
 390         cdev_t dev = ap->a_head.a_dev;
 391         vm_ooffset_t result;
 392         int error;
 393
 394         switch (minor(dev)) {
 395         case 0:
 396                 /*
 397                  * minor device 0 is physical memory
 398                  */
 399                 ap->a_result = atop(ap->a_offset);
 400                 error = 0;
 401                 break;
 402         case 1:
 403                 /*
 404                  * minor device 1 is kernel memory
 405                  */
 406                 ap->a_result = atop(vtophys(ap->a_offset));
 407                 error = 0;
 408                 break;
 409         case 5:
 410         case 6:
 411                 /*
 412                  * minor device 5 is /dev/upmap (see sys/upmap.h)
 413                  * minor device 6 is /dev/kpmap (see sys/upmap.h)
 414                  */
 415                 result = 0;
 416                 error = user_kernel_mapping(minor(dev), ap->a_offset, &result);
 417                 ap->a_result = atop(result);
 418                 break;
 419         default:
 420                 error = EINVAL;
 421                 break;
 422         }
 423         return error;
 424 }
 425
 426 #endif
 427
 428 static int
 429 memuksmap(cdev_t dev, vm_page_t fake)
 430 {
 431         vm_ooffset_t result;
 432         int error;
 433
 434         switch (minor(dev)) {
 435         case 0:
 436                 /*
 437                  * minor device 0 is physical memory
 438                  */
 439                 fake->phys_addr = ptoa(fake->pindex);
 440                 error = 0;
 441                 break;
 442         case 1:
 443                 /*
 444                  * minor device 1 is kernel memory
 445                  */
 446                 fake->phys_addr = vtophys(ptoa(fake->pindex));
 447                 error = 0;
 448                 break;
 449         case 5:
 450         case 6:
 451                 /*
 452                  * minor device 5 is /dev/upmap (see sys/upmap.h)
 453                  * minor device 6 is /dev/kpmap (see sys/upmap.h)
 454                  */
 455                 result = 0;
 456                 error = user_kernel_mapping(minor(dev),
 457                                             ptoa(fake->pindex), &result);
 458                 fake->phys_addr = result;
 459                 break;
 460         default:
 461                 error = EINVAL;
 462                 break;
 463         }
 464         return error;
 465 }
 466
 467 static int
 468 mmioctl(struct dev_ioctl_args *ap)
 469 {
 470         cdev_t dev = ap->a_head.a_dev;
 471         int error;
 472
 473         lockmgr(&mem_lock, LK_EXCLUSIVE);
 474
 475         switch (minor(dev)) {
 476         case 0:
 477                 error = mem_ioctl(dev, ap->a_cmd, ap->a_data,
 478                                   ap->a_fflag, ap->a_cred);
 479                 break;
 480         case 3:
 481         case 4:
 482                 error = random_ioctl(dev, ap->a_cmd, ap->a_data,
 483                                      ap->a_fflag, ap->a_cred);
 484                 break;
 485         default:
 486                 error = ENODEV;
 487                 break;
 488         }
 489
 490         lockmgr(&mem_lock, LK_RELEASE);
 491
 492         return (error);
 493 }
 494
 495 /*
 496  * Operations for changing memory attributes.
 497  *
 498  * This is basically just an ioctl shim for mem_range_attr_get
 499  * and mem_range_attr_set.
 500  */
 501 static int
 502 mem_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
 503 {
 504         int nd, error = 0;
 505         struct mem_range_op *mo = (struct mem_range_op *)data;
 506         struct mem_range_desc *md;
 507
 508         /* is this for us? */
 509         if ((cmd != MEMRANGE_GET) &&
 510             (cmd != MEMRANGE_SET))
 511                 return (ENOTTY);
 512
 513         /* any chance we can handle this? */
 514         if (mem_range_softc.mr_op == NULL)
 515                 return (EOPNOTSUPP);
 516
 517         /* do we have any descriptors? */
 518         if (mem_range_softc.mr_ndesc == 0)
 519                 return (ENXIO);
 520
 521         switch (cmd) {
 522         case MEMRANGE_GET:
 523                 nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
 524                 if (nd > 0) {
 525                         md = (struct mem_range_desc *)
 526                                 kmalloc(nd * sizeof(struct mem_range_desc),
 527                                        M_MEMDESC, M_WAITOK);
 528                         error = mem_range_attr_get(md, &nd);
 529                         if (!error)
 530                                 error = copyout(md, mo->mo_desc,
 531                                         nd * sizeof(struct mem_range_desc));
 532                         kfree(md, M_MEMDESC);
 533                 } else {
 534                         nd = mem_range_softc.mr_ndesc;
 535                 }
 536                 mo->mo_arg[0] = nd;
 537                 break;
 538
 539         case MEMRANGE_SET:
 540                 md = (struct mem_range_desc *)kmalloc(sizeof(struct mem_range_desc),
 541                                                     M_MEMDESC, M_WAITOK);
 542                 error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
 543                 /* clamp description string */
 544                 md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
 545                 if (error == 0)
 546                         error = mem_range_attr_set(md, &mo->mo_arg[0]);
 547                 kfree(md, M_MEMDESC);
 548                 break;
 549         }
 550         return (error);
 551 }
 552
 553 /*
 554  * Implementation-neutral, kernel-callable functions for manipulating
 555  * memory range attributes.
 556  */
 557 int
 558 mem_range_attr_get(struct mem_range_desc *mrd, int *arg)
 559 {
 560         /* can we handle this? */
 561         if (mem_range_softc.mr_op == NULL)
 562                 return (EOPNOTSUPP);
 563
 564         if (*arg == 0) {
 565                 *arg = mem_range_softc.mr_ndesc;
 566         } else {
 567                 bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc));
 568         }
 569         return (0);
 570 }
 571
 572 int
 573 mem_range_attr_set(struct mem_range_desc *mrd, int *arg)
 574 {
 575         /* can we handle this? */
 576         if (mem_range_softc.mr_op == NULL)
 577                 return (EOPNOTSUPP);
 578
 579         return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
 580 }
 581
 582 void
 583 mem_range_AP_init(void)
 584 {
 585         if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
 586                 mem_range_softc.mr_op->initAP(&mem_range_softc);
 587 }
 588
 589 static int
 590 random_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
 591 {
 592         int error;
 593         int intr;
 594
 595         /*
 596          * Even inspecting the state is privileged, since it gives a hint
 597          * about how easily the randomness might be guessed.
 598          */
 599         error = 0;
 600
 601         switch (cmd) {
 602         /* Really handled in upper layer */
 603         case FIOASYNC:
 604                 break;
 605         case MEM_SETIRQ:
 606                 intr = *(int16_t *)data;
 607                 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
 608                         break;
 609                 if (intr < 0 || intr >= MAX_INTS)
 610                         return (EINVAL);
 611                 register_randintr(intr);
 612                 break;
 613         case MEM_CLEARIRQ:
 614                 intr = *(int16_t *)data;
 615                 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
 616                         break;
 617                 if (intr < 0 || intr >= MAX_INTS)
 618                         return (EINVAL);
 619                 unregister_randintr(intr);
 620                 break;
 621         case MEM_RETURNIRQ:
 622                 error = ENOTSUP;
 623                 break;
 624         case MEM_FINDIRQ:
 625                 intr = *(int16_t *)data;
 626                 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
 627                         break;
 628                 if (intr < 0 || intr >= MAX_INTS)
 629                         return (EINVAL);
 630                 intr = next_registered_randintr(intr);
 631                 if (intr == MAX_INTS)
 632                         return (ENOENT);
 633                 *(u_int16_t *)data = intr;
 634                 break;
 635         default:
 636                 error = ENOTSUP;
 637                 break;
 638         }
 639         return (error);
 640 }
 641
 642 static int
 643 mm_filter_read(struct knote *kn, long hint)
 644 {
 645         return (1);
 646 }
 647
 648 static int
 649 mm_filter_write(struct knote *kn, long hint)
 650 {
 651         return (1);
 652 }
 653
 654 static void
 655 dummy_filter_detach(struct knote *kn) {}
 656
 657 /* Implemented in kern_nrandom.c */
 658 static struct filterops random_read_filtops =
 659         { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, random_filter_read };
 660
 661 static struct filterops mm_read_filtops =
 662         { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_read };
 663
 664 static struct filterops mm_write_filtops =
 665         { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_write };
 666
 667 static int
 668 mmkqfilter(struct dev_kqfilter_args *ap)
 669 {
 670         struct knote *kn = ap->a_kn;
 671         cdev_t dev = ap->a_head.a_dev;
 672
 673         ap->a_result = 0;
 674         switch (kn->kn_filter) {
 675         case EVFILT_READ:
 676                 switch (minor(dev)) {
 677                 case 3:
 678                         kn->kn_fop = &random_read_filtops;
 679                         break;
 680                 default:
 681                         kn->kn_fop = &mm_read_filtops;
 682                         break;
 683                 }
 684                 break;
 685         case EVFILT_WRITE:
 686                 kn->kn_fop = &mm_write_filtops;
 687                 break;
 688         default:
 689                 ap->a_result = EOPNOTSUPP;
 690                 return (0);
 691         }
 692
 693         return (0);
 694 }
 695
 696 int
 697 iszerodev(cdev_t dev)
 698 {
 699         return (zerodev == dev);
 700 }
 701
 702 /*
 703  * /dev/upmap and /dev/kpmap.
 704  */
 705 static int
 706 user_kernel_mapping(int num, vm_ooffset_t offset, vm_ooffset_t *resultp)
 707 {
 708         struct proc *p;
 709         int error;
 710         int invfork;
 711
 712         if ((p = curproc) == NULL)
 713                 return (EINVAL);
 714
 715         /*
 716          * If this is a child currently in vfork the pmap is shared with
 717          * the parent!  We need to actually set-up the parent's p_upmap,
 718          * not the child's, and we need to set the invfork flag.  Userland
 719          * will probably adjust its static state so it must be consistent
 720          * with the parent or userland will be really badly confused.
 721          *
 722          * (this situation can happen when user code in vfork() calls
 723          *  libc's getpid() or some other function which then decides
 724          *  it wants the upmap).
 725          */
 726         if (p->p_flags & P_PPWAIT) {
 727                 p = p->p_pptr;
 728                 if (p == NULL)
 729                         return (EINVAL);
 730                 invfork = 1;
 731         } else {
 732                 invfork = 0;
 733         }
 734
 735         error = EINVAL;
 736
 737         switch(num) {
 738         case 5:
 739                 /*
 740                  * /dev/upmap - maps RW per-process shared user-kernel area.
 741                  */
 742                 if (p->p_upmap == NULL)
 743                         proc_usermap(p, invfork);
 744                 else if (invfork)
 745                         p->p_upmap->invfork = invfork;
 746
 747                 if (p->p_upmap &&
 748                     offset < roundup2(sizeof(*p->p_upmap), PAGE_SIZE)) {
 749                         /* only good for current process */
 750                         *resultp = pmap_kextract((vm_offset_t)p->p_upmap +
 751                                                  offset);
 752                         error = 0;
 753                 }
 754                 break;
 755         case 6:
 756                 /*
 757                  * /dev/kpmap - maps RO shared kernel global page
 758                  */
 759                 if (kpmap &&
 760                     offset < roundup2(sizeof(*kpmap), PAGE_SIZE)) {
 761                         *resultp = pmap_kextract((vm_offset_t)kpmap +
 762                                                  offset);
 763                         error = 0;
 764                 }
 765                 break;
 766         default:
 767                 break;
 768         }
 769         return error;
 770 }
 771
 772 static void
 773 mem_drvinit(void *unused)
 774 {
 775
 776         /* Initialise memory range handling */
 777         if (mem_range_softc.mr_op != NULL)
 778                 mem_range_softc.mr_op->init(&mem_range_softc);
 779
 780         make_dev(&mem_ops_mem, 0, UID_ROOT, GID_KMEM, 0640, "mem");
 781         make_dev(&mem_ops_mem, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
 782         make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null");
 783         make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random");
 784         make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
 785         make_dev(&mem_ops, 5, UID_ROOT, GID_WHEEL, 0666, "upmap");
 786         make_dev(&mem_ops, 6, UID_ROOT, GID_WHEEL, 0444, "kpmap");
 787         zerodev = make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
 788         make_dev(&mem_ops_noq, 14, UID_ROOT, GID_WHEEL, 0600, "io");
 789 }
 790
 791 SYSINIT(memdev, SI_SUB_DRIVERS, SI_ORDER_MIDDLE + CDEV_MAJOR, mem_drvinit,
 792     NULL);
 793