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