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/queue.h>
  57 #include <sys/random.h>
  58 #include <sys/signalvar.h>
  59 #include <sys/uio.h>
  60 #include <sys/vnode.h>
  61 #include <sys/sysctl.h>
  62
  63 #include <sys/signal2.h>
  64 #include <sys/spinlock2.h>
  65
  66 #include <vm/vm.h>
  67 #include <vm/pmap.h>
  68 #include <vm/vm_map.h>
  69 #include <vm/vm_extern.h>
  70
  71
  72 static  d_open_t        mmopen;
  73 static  d_close_t       mmclose;
  74 static  d_read_t        mmread;
  75 static  d_write_t       mmwrite;
  76 static  d_ioctl_t       mmioctl;
  77 #if 0
  78 static  d_mmap_t        memmmap;
  79 #endif
  80 static  d_kqfilter_t    mmkqfilter;
  81 static int memuksmap(vm_map_backing_t ba, int op, cdev_t dev, vm_page_t fake);
  82
  83 #define CDEV_MAJOR 2
  84 static struct dev_ops mem_ops = {
  85         { "mem", 0, D_MPSAFE | D_QUICK },
  86         .d_open =       mmopen,
  87         .d_close =      mmclose,
  88         .d_read =       mmread,
  89         .d_write =      mmwrite,
  90         .d_ioctl =      mmioctl,
  91         .d_kqfilter =   mmkqfilter,
  92 #if 0
  93         .d_mmap =       memmmap,
  94 #endif
  95         .d_uksmap =     memuksmap
  96 };
  97
  98 static struct dev_ops mem_ops_mem = {
  99         { "mem", 0, D_MEM | D_MPSAFE | D_QUICK },
 100         .d_open =       mmopen,
 101         .d_close =      mmclose,
 102         .d_read =       mmread,
 103         .d_write =      mmwrite,
 104         .d_ioctl =      mmioctl,
 105         .d_kqfilter =   mmkqfilter,
 106 #if 0
 107         .d_mmap =       memmmap,
 108 #endif
 109         .d_uksmap =     memuksmap
 110 };
 111
 112 static struct dev_ops mem_ops_noq = {
 113         { "mem", 0, D_MPSAFE },
 114         .d_open =       mmopen,
 115         .d_close =      mmclose,
 116         .d_read =       mmread,
 117         .d_write =      mmwrite,
 118         .d_ioctl =      mmioctl,
 119         .d_kqfilter =   mmkqfilter,
 120 #if 0
 121         .d_mmap =       memmmap,
 122 #endif
 123         .d_uksmap =     memuksmap
 124 };
 125
 126 static int rand_bolt;
 127 static caddr_t  zbuf;
 128 static cdev_t   zerodev = NULL;
 129 static struct lock mem_lock = LOCK_INITIALIZER("memlk", 0, 0);
 130
 131 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors");
 132 static int mem_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
 133 static int random_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
 134
 135 struct mem_range_softc mem_range_softc;
 136
 137 static int seedenable;
 138 SYSCTL_INT(_kern, OID_AUTO, seedenable, CTLFLAG_RW, &seedenable, 0, "");
 139
 140 static int
 141 mmopen(struct dev_open_args *ap)
 142 {
 143         cdev_t dev = ap->a_head.a_dev;
 144         int error;
 145
 146         switch (minor(dev)) {
 147         case 0:
 148         case 1:
 149                 /*
 150                  * /dev/mem and /dev/kmem
 151                  */
 152                 if (ap->a_oflags & FWRITE) {
 153                         if (securelevel > 0 || kernel_mem_readonly)
 154                                 return (EPERM);
 155                 }
 156                 error = 0;
 157                 break;
 158         case 6:
 159                 /*
 160                  * /dev/kpmap can only be opened for reading.
 161                  */
 162                 if (ap->a_oflags & FWRITE)
 163                         return (EPERM);
 164                 error = 0;
 165                 break;
 166         case 14:
 167                 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0);
 168                 if (error != 0)
 169                         break;
 170                 if (securelevel > 0 || kernel_mem_readonly) {
 171                         error = EPERM;
 172                         break;
 173                 }
 174                 error = cpu_set_iopl();
 175                 break;
 176         default:
 177                 error = 0;
 178                 break;
 179         }
 180         return (error);
 181 }
 182
 183 static int
 184 mmclose(struct dev_close_args *ap)
 185 {
 186         cdev_t dev = ap->a_head.a_dev;
 187         int error;
 188
 189         switch (minor(dev)) {
 190         case 14:
 191                 error = cpu_clr_iopl();
 192                 break;
 193         default:
 194                 error = 0;
 195                 break;
 196         }
 197         return (error);
 198 }
 199
 200
 201 static int
 202 mmrw(cdev_t dev, struct uio *uio, int flags)
 203 {
 204         int o;
 205         u_int c;
 206         u_int poolsize;
 207         u_long v;
 208         struct iovec *iov;
 209         int error = 0;
 210         caddr_t buf = NULL;
 211
 212         while (uio->uio_resid > 0 && error == 0) {
 213                 iov = uio->uio_iov;
 214                 if (iov->iov_len == 0) {
 215                         uio->uio_iov++;
 216                         uio->uio_iovcnt--;
 217                         if (uio->uio_iovcnt < 0)
 218                                 panic("mmrw");
 219                         continue;
 220                 }
 221                 switch (minor(dev)) {
 222                 case 0:
 223                         /*
 224                          * minor device 0 is physical memory, /dev/mem
 225                          */
 226                         v = uio->uio_offset;
 227                         v &= ~(long)PAGE_MASK;
 228                         pmap_kenter((vm_offset_t)ptvmmap, v);
 229                         o = (int)uio->uio_offset & PAGE_MASK;
 230                         c = (u_int)(PAGE_SIZE - ((uintptr_t)iov->iov_base & PAGE_MASK));
 231                         c = min(c, (u_int)(PAGE_SIZE - o));
 232                         c = min(c, (u_int)iov->iov_len);
 233                         error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
 234                         pmap_kremove((vm_offset_t)ptvmmap);
 235                         continue;
 236
 237                 case 1: {
 238                         /*
 239                          * minor device 1 is kernel memory, /dev/kmem
 240                          */
 241                         vm_offset_t saddr, eaddr;
 242                         int prot;
 243
 244                         c = iov->iov_len;
 245
 246                         /*
 247                          * Make sure that all of the pages are currently
 248                          * resident so that we don't create any zero-fill
 249                          * pages.
 250                          */
 251                         saddr = trunc_page(uio->uio_offset);
 252                         eaddr = round_page(uio->uio_offset + c);
 253                         if (saddr > eaddr)
 254                                 return EFAULT;
 255
 256                         /*
 257                          * Make sure the kernel addresses are mapped.
 258                          * platform_direct_mapped() can be used to bypass
 259                          * default mapping via the page table (virtual kernels
 260                          * contain a lot of out-of-band data).
 261                          */
 262                         prot = VM_PROT_READ;
 263                         if (uio->uio_rw != UIO_READ)
 264                                 prot |= VM_PROT_WRITE;
 265                         error = kvm_access_check(saddr, eaddr, prot);
 266                         if (error)
 267                                 return (error);
 268                         error = uiomove((caddr_t)(vm_offset_t)uio->uio_offset,
 269                                         (int)c, uio);
 270                         continue;
 271                 }
 272                 case 2:
 273                         /*
 274                          * minor device 2 (/dev/null) is EOF/RATHOLE
 275                          */
 276                         if (uio->uio_rw == UIO_READ)
 277                                 return (0);
 278                         c = iov->iov_len;
 279                         break;
 280                 case 3:
 281                         /*
 282                          * minor device 3 (/dev/random) is source of filth
 283                          * on read, seeder on write
 284                          */
 285                         if (buf == NULL)
 286                                 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
 287                         c = min(iov->iov_len, PAGE_SIZE);
 288                         if (uio->uio_rw == UIO_WRITE) {
 289                                 error = uiomove(buf, (int)c, uio);
 290                                 if (error == 0 &&
 291                                     seedenable &&
 292                                     securelevel <= 0) {
 293                                         error = add_buffer_randomness_src(buf, c, RAND_SRC_SEEDING);
 294                                 } else if (error == 0) {
 295                                         error = EPERM;
 296                                 }
 297                         } else {
 298                                 poolsize = read_random(buf, c, 0);
 299                                 if (poolsize == 0) {
 300                                         if (buf)
 301                                                 kfree(buf, M_TEMP);
 302                                         if ((flags & IO_NDELAY) != 0)
 303                                                 return (EWOULDBLOCK);
 304                                         return (0);
 305                                 }
 306                                 c = min(c, poolsize);
 307                                 error = uiomove(buf, (int)c, uio);
 308                         }
 309                         continue;
 310                 case 4:
 311                         /*
 312                          * minor device 4 (/dev/urandom) is source of muck
 313                          * on read, writes are disallowed.
 314                          */
 315                         c = min(iov->iov_len, PAGE_SIZE);
 316                         if (uio->uio_rw == UIO_WRITE) {
 317                                 error = EPERM;
 318                                 break;
 319                         }
 320                         if (CURSIG(curthread->td_lwp) != 0) {
 321                                 /*
 322                                  * Use tsleep() to get the error code right.
 323                                  * It should return immediately.
 324                                  */
 325                                 error = tsleep(&rand_bolt, PCATCH, "urand", 1);
 326                                 if (error != 0 && error != EWOULDBLOCK)
 327                                         continue;
 328                         }
 329                         if (buf == NULL)
 330                                 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
 331                         poolsize = read_random(buf, c, 1);
 332                         c = min(c, poolsize);
 333                         error = uiomove(buf, (int)c, uio);
 334                         continue;
 335                 /* case 5: read/write not supported, mmap only */
 336                 /* case 6: read/write not supported, mmap only */
 337                 case 12:
 338                         /*
 339                          * minor device 12 (/dev/zero) is source of nulls
 340                          * on read, write are disallowed.
 341                          */
 342                         if (uio->uio_rw == UIO_WRITE) {
 343                                 c = iov->iov_len;
 344                                 break;
 345                         }
 346                         if (zbuf == NULL) {
 347                                 zbuf = (caddr_t)kmalloc(PAGE_SIZE, M_TEMP,
 348                                     M_WAITOK | M_ZERO);
 349                         }
 350                         c = min(iov->iov_len, PAGE_SIZE);
 351                         error = uiomove(zbuf, (int)c, uio);
 352                         continue;
 353                 default:
 354                         return (ENODEV);
 355                 }
 356                 if (error)
 357                         break;
 358                 iov->iov_base = (char *)iov->iov_base + c;
 359                 iov->iov_len -= c;
 360                 uio->uio_offset += c;
 361                 uio->uio_resid -= c;
 362         }
 363         if (buf)
 364                 kfree(buf, M_TEMP);
 365         return (error);
 366 }
 367
 368 static int
 369 mmread(struct dev_read_args *ap)
 370 {
 371         return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
 372 }
 373
 374 static int
 375 mmwrite(struct dev_write_args *ap)
 376 {
 377         return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
 378 }
 379
 380 /*******************************************************\
 381 * allow user processes to MMAP some memory sections     *
 382 * instead of going through read/write                   *
 383 \*******************************************************/
 384
 385 static int user_kernel_mapping(vm_map_backing_t ba, int num,
 386                         vm_ooffset_t offset, vm_ooffset_t *resultp);
 387
 388 static int
 389 memuksmap(vm_map_backing_t ba, int op, cdev_t dev, vm_page_t fake)
 390 {
 391         vm_ooffset_t result;
 392         int error;
 393         struct lwp *lp;
 394
 395         error = 0;
 396
 397         switch(op) {
 398         case UKSMAPOP_ADD:
 399                 /*
 400                  * We only need to track mappings for /dev/lpmap, all process
 401                  * mappings will be deleted when the process exits and we
 402                  * do not need to track kernel mappings.
 403                  */
 404                 if (minor(dev) == 7) {
 405                         lp = ba->aux_info;
 406                         spin_lock(&lp->lwp_spin);
 407                         TAILQ_INSERT_TAIL(&lp->lwp_lpmap_backing_list,
 408                                           ba, entry);
 409                         spin_unlock(&lp->lwp_spin);
 410                 }
 411                 break;
 412         case UKSMAPOP_REM:
 413                 /*
 414                  * We only need to track mappings for /dev/lpmap, all process
 415                  * mappings will be deleted when the process exits and we
 416                  * do not need to track kernel mappings.
 417                  */
 418                 if (minor(dev) == 7) {
 419                         lp = ba->aux_info;
 420                         spin_lock(&lp->lwp_spin);
 421                         TAILQ_REMOVE(&lp->lwp_lpmap_backing_list, ba, entry);
 422                         spin_unlock(&lp->lwp_spin);
 423                 }
 424                 break;
 425         case UKSMAPOP_FAULT:
 426                 switch (minor(dev)) {
 427                 case 0:
 428                         /*
 429                          * minor device 0 is physical memory
 430                          */
 431                         fake->phys_addr = ptoa(fake->pindex);
 432                         break;
 433                 case 1:
 434                         /*
 435                          * minor device 1 is kernel memory
 436                          */
 437                         fake->phys_addr = vtophys(ptoa(fake->pindex));
 438                         break;
 439                 case 5:
 440                 case 6:
 441                 case 7:
 442                         /*
 443                          * minor device 5 is /dev/upmap (see sys/upmap.h)
 444                          * minor device 6 is /dev/kpmap (see sys/upmap.h)
 445                          * minor device 7 is /dev/lpmap (see sys/upmap.h)
 446                          */
 447                         result = 0;
 448                         error = user_kernel_mapping(ba,
 449                                                     minor(dev),
 450                                                     ptoa(fake->pindex),
 451                                                     &result);
 452                         fake->phys_addr = result;
 453                         break;
 454                 default:
 455                         error = EINVAL;
 456                         break;
 457                 }
 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/lpmap, /dev/upmap, /dev/kpmap.
 703  */
 704 static int
 705 user_kernel_mapping(vm_map_backing_t ba, int num, vm_ooffset_t offset,
 706                     vm_ooffset_t *resultp)
 707 {
 708         struct proc *p;
 709         struct lwp *lp;
 710         int error;
 711         int invfork;
 712
 713         if (offset < 0)
 714                 return (EINVAL);
 715
 716         error = EINVAL;
 717
 718         switch(num) {
 719         case 5:
 720                 /*
 721                  * /dev/upmap - maps RW per-process shared user-kernel area.
 722                  */
 723
 724                 /*
 725                  * If this is a child currently in vfork the pmap is shared
 726                  * with the parent!  We need to actually set-up the parent's
 727                  * p_upmap, not the child's, and we need to set the invfork
 728                  * flag.  Userland will probably adjust its static state so
 729                  * it must be consistent with the parent or userland will be
 730                  * really badly confused.
 731                  *
 732                  * (this situation can happen when user code in vfork() calls
 733                  *  libc's getpid() or some other function which then decides
 734                  *  it wants the upmap).
 735                  */
 736                 p = ba->aux_info;
 737                 if (p == NULL)
 738                         break;
 739                 if (p->p_flags & P_PPWAIT) {
 740                         p = p->p_pptr;
 741                         if (p == NULL)
 742                                 return (EINVAL);
 743                         invfork = 1;
 744                 } else {
 745                         invfork = 0;
 746                 }
 747
 748                 /*
 749                  * Create the kernel structure as required, set the invfork
 750                  * flag if we are faulting in on a vfork().
 751                  */
 752                 if (p->p_upmap == NULL)
 753                         proc_usermap(p, invfork);
 754                 if (p->p_upmap && invfork)
 755                         p->p_upmap->invfork = invfork;
 756
 757                 /*
 758                  * Extract address for pmap
 759                  */
 760                 if (p->p_upmap &&
 761                     offset < roundup2(sizeof(*p->p_upmap), PAGE_SIZE)) {
 762                         /* only good for current process */
 763                         *resultp = pmap_kextract((vm_offset_t)p->p_upmap +
 764                                                  offset);
 765                         error = 0;
 766                 }
 767                 break;
 768         case 6:
 769                 /*
 770                  * /dev/kpmap - maps RO shared kernel global page
 771                  *
 772                  * Extract address for pmap
 773                  */
 774                 if (kpmap &&
 775                     offset < roundup2(sizeof(*kpmap), PAGE_SIZE)) {
 776                         *resultp = pmap_kextract((vm_offset_t)kpmap + offset);
 777                         error = 0;
 778                 }
 779                 break;
 780         case 7:
 781                 /*
 782                  * /dev/lpmap - maps RW per-thread shared user-kernel area.
 783                  */
 784                 lp = ba->aux_info;
 785                 if (lp == NULL)
 786                         break;
 787
 788                 /*
 789                  * Create the kernel structure as required
 790                  */
 791                 if (lp->lwp_lpmap == NULL)
 792                         lwp_usermap(lp, -1);    /* second arg not yet XXX */
 793
 794                 /*
 795                  * Extract address for pmap
 796                  */
 797                 if (lp->lwp_lpmap &&
 798                     offset < roundup2(sizeof(*lp->lwp_lpmap), PAGE_SIZE)) {
 799                         /* only good for current process */
 800                         *resultp = pmap_kextract((vm_offset_t)lp->lwp_lpmap +
 801                                                  offset);
 802                         error = 0;
 803                 }
 804                 break;
 805         default:
 806                 break;
 807         }
 808         return error;
 809 }
 810
 811 static void
 812 mem_drvinit(void *unused)
 813 {
 814
 815         /* Initialise memory range handling */
 816         if (mem_range_softc.mr_op != NULL)
 817                 mem_range_softc.mr_op->init(&mem_range_softc);
 818
 819         make_dev(&mem_ops_mem, 0, UID_ROOT, GID_KMEM, 0640, "mem");
 820         make_dev(&mem_ops_mem, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
 821         make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null");
 822         make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random");
 823         make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
 824         make_dev(&mem_ops, 5, UID_ROOT, GID_WHEEL, 0666, "upmap");
 825         make_dev(&mem_ops, 6, UID_ROOT, GID_WHEEL, 0444, "kpmap");
 826         make_dev(&mem_ops, 7, UID_ROOT, GID_WHEEL, 0666, "lpmap");
 827         zerodev = make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
 828         make_dev(&mem_ops_noq, 14, UID_ROOT, GID_WHEEL, 0600, "io");
 829 }
 830
 831 SYSINIT(memdev, SI_SUB_DRIVERS, SI_ORDER_MIDDLE + CDEV_MAJOR, mem_drvinit,
 832     NULL);
 833