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