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. All advertising materials mentioning features or use of this software
  20  *    must display the following acknowledgement:
  21  *      This product includes software developed by the University of
  22  *      California, Berkeley and its contributors.
  23  * 4. Neither the name of the University nor the names of its contributors
  24  *    may be used to endorse or promote products derived from this software
  25  *    without specific prior written permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  37  * SUCH DAMAGE.
  38  *
  39  *      from: Utah $Hdr: mem.c 1.13 89/10/08$
  40  *      from: @(#)mem.c 7.2 (Berkeley) 5/9/91
  41  * $FreeBSD: src/sys/i386/i386/mem.c,v 1.79.2.9 2003/01/04 22:58:01 njl Exp $
  42  * $DragonFly: src/sys/kern/kern_memio.c,v 1.11 2004/05/19 22:52:57 dillon Exp $
  43  */
  44
  45 /*
  46  * Memory special file
  47  */
  48
  49 #include <sys/param.h>
  50 #include <sys/systm.h>
  51 #include <sys/buf.h>
  52 #include <sys/conf.h>
  53 #include <sys/fcntl.h>
  54 #include <sys/filio.h>
  55 #include <sys/ioccom.h>
  56 #include <sys/kernel.h>
  57 #include <sys/malloc.h>
  58 #include <sys/memrange.h>
  59 #include <sys/proc.h>
  60 #include <sys/random.h>
  61 #include <sys/signalvar.h>
  62 #include <sys/uio.h>
  63 #include <sys/vnode.h>
  64
  65 #include <machine/frame.h>
  66 #include <machine/psl.h>
  67 #include <machine/specialreg.h>
  68 #include <i386/isa/intr_machdep.h>
  69
  70 #include <vm/vm.h>
  71 #include <vm/pmap.h>
  72 #include <vm/vm_extern.h>
  73
  74
  75 static  d_open_t        mmopen;
  76 static  d_close_t       mmclose;
  77 static  d_read_t        mmrw;
  78 static  d_ioctl_t       mmioctl;
  79 static  d_mmap_t        memmmap;
  80 static  d_poll_t        mmpoll;
  81
  82 #define CDEV_MAJOR 2
  83 static struct cdevsw mem_cdevsw = {
  84         /* name */      "mem",
  85         /* maj */       CDEV_MAJOR,
  86         /* flags */     D_MEM,
  87         /* port */      NULL,
  88         /* clone */     NULL,
  89
  90         /* open */      mmopen,
  91         /* close */     mmclose,
  92         /* read */      mmrw,
  93         /* write */     mmrw,
  94         /* ioctl */     mmioctl,
  95         /* poll */      mmpoll,
  96         /* mmap */      memmmap,
  97         /* strategy */  nostrategy,
  98         /* dump */      nodump,
  99         /* psize */     nopsize
 100 };
 101
 102 static int rand_bolt;
 103 static caddr_t  zbuf;
 104
 105 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors");
 106 static int mem_ioctl (dev_t, u_long, caddr_t, int, struct thread *);
 107 static int random_ioctl (dev_t, u_long, caddr_t, int, struct thread *);
 108
 109 struct mem_range_softc mem_range_softc;
 110
 111
 112 static int
 113 mmclose(dev_t dev, int flags, int fmt, struct thread *td)
 114 {
 115         struct proc *p = td->td_proc;
 116
 117         switch (minor(dev)) {
 118         case 14:
 119                 p->p_md.md_regs->tf_eflags &= ~PSL_IOPL;
 120                 break;
 121         default:
 122                 break;
 123         }
 124         return (0);
 125 }
 126
 127 static int
 128 mmopen(dev_t dev, int flags, int fmt, struct thread *td)
 129 {
 130         int error;
 131         struct proc *p = td->td_proc;
 132
 133         switch (minor(dev)) {
 134         case 0:
 135         case 1:
 136                 if ((flags & FWRITE) && securelevel > 0)
 137                         return (EPERM);
 138                 break;
 139         case 14:
 140                 error = suser(td);
 141                 if (error != 0)
 142                         return (error);
 143                 if (securelevel > 0)
 144                         return (EPERM);
 145                 p->p_md.md_regs->tf_eflags |= PSL_IOPL;
 146                 break;
 147         default:
 148                 break;
 149         }
 150         return (0);
 151 }
 152
 153 static int
 154 mmrw(dev, uio, flags)
 155         dev_t dev;
 156         struct uio *uio;
 157         int flags;
 158 {
 159         int o;
 160         u_int c, v;
 161         u_int poolsize;
 162         struct iovec *iov;
 163         int error = 0;
 164         caddr_t buf = NULL;
 165
 166         while (uio->uio_resid > 0 && error == 0) {
 167                 iov = uio->uio_iov;
 168                 if (iov->iov_len == 0) {
 169                         uio->uio_iov++;
 170                         uio->uio_iovcnt--;
 171                         if (uio->uio_iovcnt < 0)
 172                                 panic("mmrw");
 173                         continue;
 174                 }
 175                 switch (minor(dev)) {
 176
 177 /* minor device 0 is physical memory */
 178                 case 0:
 179                         v = uio->uio_offset;
 180                         v &= ~PAGE_MASK;
 181                         pmap_kenter((vm_offset_t)ptvmmap, v);
 182                         o = (int)uio->uio_offset & PAGE_MASK;
 183                         c = (u_int)(PAGE_SIZE - ((int)iov->iov_base & PAGE_MASK));
 184                         c = min(c, (u_int)(PAGE_SIZE - o));
 185                         c = min(c, (u_int)iov->iov_len);
 186                         error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
 187                         pmap_kremove((vm_offset_t)ptvmmap);
 188                         continue;
 189
 190 /* minor device 1 is kernel memory */
 191                 case 1: {
 192                         vm_offset_t addr, eaddr;
 193                         c = iov->iov_len;
 194
 195                         /*
 196                          * Make sure that all of the pages are currently resident so
 197                          * that we don't create any zero-fill pages.
 198                          */
 199                         addr = trunc_page(uio->uio_offset);
 200                         eaddr = round_page(uio->uio_offset + c);
 201
 202                         if (addr < (vm_offset_t)VADDR(PTDPTDI, 0))
 203                                 return EFAULT;
 204                         if (eaddr >= (vm_offset_t)VADDR(APTDPTDI, 0))
 205                                 return EFAULT;
 206                         for (; addr < eaddr; addr += PAGE_SIZE)
 207                                 if (pmap_extract(kernel_pmap, addr) == 0)
 208                                         return EFAULT;
 209
 210                         if (!kernacc((caddr_t)(int)uio->uio_offset, c,
 211                             uio->uio_rw == UIO_READ ?
 212                             VM_PROT_READ : VM_PROT_WRITE))
 213                                 return (EFAULT);
 214                         error = uiomove((caddr_t)(int)uio->uio_offset, (int)c, uio);
 215                         continue;
 216                 }
 217
 218 /* minor device 2 is EOF/RATHOLE */
 219                 case 2:
 220                         if (uio->uio_rw == UIO_READ)
 221                                 return (0);
 222                         c = iov->iov_len;
 223                         break;
 224
 225 /* minor device 3 (/dev/random) is source of filth on read, rathole on write */
 226                 case 3:
 227                         if (uio->uio_rw == UIO_WRITE) {
 228                                 c = iov->iov_len;
 229                                 break;
 230                         }
 231                         if (buf == NULL)
 232                                 buf = (caddr_t)
 233                                     malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
 234                         c = min(iov->iov_len, PAGE_SIZE);
 235                         poolsize = read_random(buf, c);
 236                         if (poolsize == 0) {
 237                                 if (buf)
 238                                         free(buf, M_TEMP);
 239                                 if ((flags & IO_NDELAY) != 0)
 240                                         return (EWOULDBLOCK);
 241                                 return (0);
 242                         }
 243                         c = min(c, poolsize);
 244                         error = uiomove(buf, (int)c, uio);
 245                         continue;
 246
 247 /* minor device 4 (/dev/urandom) is source of muck on read, rathole on write */
 248                 case 4:
 249                         if (uio->uio_rw == UIO_WRITE) {
 250                                 c = iov->iov_len;
 251                                 break;
 252                         }
 253                         if (CURSIG(curproc) != 0) {
 254                                 /*
 255                                  * Use tsleep() to get the error code right.
 256                                  * It should return immediately.
 257                                  */
 258                                 error = tsleep(&rand_bolt, PCATCH, "urand", 1);
 259                                 if (error != 0 && error != EWOULDBLOCK)
 260                                         continue;
 261                         }
 262                         if (buf == NULL)
 263                                 buf = (caddr_t)
 264                                     malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
 265                         c = min(iov->iov_len, PAGE_SIZE);
 266                         poolsize = read_random_unlimited(buf, c);
 267                         c = min(c, poolsize);
 268                         error = uiomove(buf, (int)c, uio);
 269                         continue;
 270
 271 /* minor device 12 (/dev/zero) is source of nulls on read, rathole on write */
 272                 case 12:
 273                         if (uio->uio_rw == UIO_WRITE) {
 274                                 c = iov->iov_len;
 275                                 break;
 276                         }
 277                         if (zbuf == NULL) {
 278                                 zbuf = (caddr_t)
 279                                     malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
 280                                 bzero(zbuf, PAGE_SIZE);
 281                         }
 282                         c = min(iov->iov_len, PAGE_SIZE);
 283                         error = uiomove(zbuf, (int)c, uio);
 284                         continue;
 285
 286                 default:
 287                         return (ENODEV);
 288                 }
 289                 if (error)
 290                         break;
 291                 iov->iov_base += c;
 292                 iov->iov_len -= c;
 293                 uio->uio_offset += c;
 294                 uio->uio_resid -= c;
 295         }
 296         if (buf)
 297                 free(buf, M_TEMP);
 298         return (error);
 299 }
 300
 301
 302
 303
 304 /*******************************************************\
 305 * allow user processes to MMAP some memory sections     *
 306 * instead of going through read/write                   *
 307 \*******************************************************/
 308 static int
 309 memmmap(dev_t dev, vm_offset_t offset, int nprot)
 310 {
 311         switch (minor(dev))
 312         {
 313
 314 /* minor device 0 is physical memory */
 315         case 0:
 316                 return i386_btop(offset);
 317
 318 /* minor device 1 is kernel memory */
 319         case 1:
 320                 return i386_btop(vtophys(offset));
 321
 322         default:
 323                 return -1;
 324         }
 325 }
 326
 327 static int
 328 mmioctl(dev_t dev, u_long cmd, caddr_t data, int flags, struct thread *td)
 329 {
 330
 331         switch (minor(dev)) {
 332         case 0:
 333                 return mem_ioctl(dev, cmd, data, flags, td);
 334         case 3:
 335         case 4:
 336                 return random_ioctl(dev, cmd, data, flags, td);
 337         }
 338         return (ENODEV);
 339 }
 340
 341 /*
 342  * Operations for changing memory attributes.
 343  *
 344  * This is basically just an ioctl shim for mem_range_attr_get
 345  * and mem_range_attr_set.
 346  */
 347 static int
 348 mem_ioctl(dev_t dev, u_long cmd, caddr_t data, int flags, struct thread *td)
 349 {
 350         int nd, error = 0;
 351         struct mem_range_op *mo = (struct mem_range_op *)data;
 352         struct mem_range_desc *md;
 353
 354         /* is this for us? */
 355         if ((cmd != MEMRANGE_GET) &&
 356             (cmd != MEMRANGE_SET))
 357                 return (ENOTTY);
 358
 359         /* any chance we can handle this? */
 360         if (mem_range_softc.mr_op == NULL)
 361                 return (EOPNOTSUPP);
 362
 363         /* do we have any descriptors? */
 364         if (mem_range_softc.mr_ndesc == 0)
 365                 return (ENXIO);
 366
 367         switch (cmd) {
 368         case MEMRANGE_GET:
 369                 nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
 370                 if (nd > 0) {
 371                         md = (struct mem_range_desc *)
 372                                 malloc(nd * sizeof(struct mem_range_desc),
 373                                        M_MEMDESC, M_WAITOK);
 374                         error = mem_range_attr_get(md, &nd);
 375                         if (!error)
 376                                 error = copyout(md, mo->mo_desc,
 377                                         nd * sizeof(struct mem_range_desc));
 378                         free(md, M_MEMDESC);
 379                 } else {
 380                         nd = mem_range_softc.mr_ndesc;
 381                 }
 382                 mo->mo_arg[0] = nd;
 383                 break;
 384
 385         case MEMRANGE_SET:
 386                 md = (struct mem_range_desc *)malloc(sizeof(struct mem_range_desc),
 387                                                     M_MEMDESC, M_WAITOK);
 388                 error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
 389                 /* clamp description string */
 390                 md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
 391                 if (error == 0)
 392                         error = mem_range_attr_set(md, &mo->mo_arg[0]);
 393                 free(md, M_MEMDESC);
 394                 break;
 395         }
 396         return (error);
 397 }
 398
 399 /*
 400  * Implementation-neutral, kernel-callable functions for manipulating
 401  * memory range attributes.
 402  */
 403 int
 404 mem_range_attr_get(mrd, arg)
 405         struct mem_range_desc *mrd;
 406         int *arg;
 407 {
 408         /* can we handle this? */
 409         if (mem_range_softc.mr_op == NULL)
 410                 return (EOPNOTSUPP);
 411
 412         if (*arg == 0) {
 413                 *arg = mem_range_softc.mr_ndesc;
 414         } else {
 415                 bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc));
 416         }
 417         return (0);
 418 }
 419
 420 int
 421 mem_range_attr_set(mrd, arg)
 422         struct mem_range_desc *mrd;
 423         int *arg;
 424 {
 425         /* can we handle this? */
 426         if (mem_range_softc.mr_op == NULL)
 427                 return (EOPNOTSUPP);
 428
 429         return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
 430 }
 431
 432 #ifdef SMP
 433 void
 434 mem_range_AP_init(void)
 435 {
 436         if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
 437                 return (mem_range_softc.mr_op->initAP(&mem_range_softc));
 438 }
 439 #endif
 440
 441 static int
 442 random_ioctl(dev_t dev, u_long cmd, caddr_t data, int flags, struct thread *td)
 443 {
 444         static intrmask_t interrupt_allowed;
 445         intrmask_t interrupt_mask;
 446         int error, intr;
 447
 448         /*
 449          * We're the random or urandom device.  The only ioctls are for
 450          * selecting and inspecting which interrupts are used in the muck
 451          * gathering business and the fcntl() stuff.
 452          */
 453         if (cmd != MEM_SETIRQ && cmd != MEM_CLEARIRQ && cmd != MEM_RETURNIRQ
 454                 && cmd != FIONBIO && cmd != FIOASYNC)
 455                 return (ENOTTY);
 456
 457         /*
 458          * XXX the data is 16-bit due to a historical botch, so we use
 459          * magic 16's instead of ICU_LEN and can't support 24 interrupts
 460          * under SMP.
 461          * Even inspecting the state is privileged, since it gives a hint
 462          * about how easily the randomness might be guessed.
 463          */
 464         intr = *(int16_t *)data;
 465         interrupt_mask = 1 << intr;
 466         switch (cmd) {
 467         /* Really handled in upper layer */
 468         case FIOASYNC:
 469         case FIONBIO:
 470                 break;
 471         case MEM_SETIRQ:
 472                 error = suser(td);
 473                 if (error != 0)
 474                         return (error);
 475                 if (intr < 0 || intr >= 16)
 476                         return (EINVAL);
 477                 if (interrupt_allowed & interrupt_mask)
 478                         break;
 479                 interrupt_allowed |= interrupt_mask;
 480                 register_randintr(intr);
 481                 break;
 482         case MEM_CLEARIRQ:
 483                 error = suser(td);
 484                 if (error != 0)
 485                         return (error);
 486                 if (intr < 0 || intr >= 16)
 487                         return (EINVAL);
 488                 if (!(interrupt_allowed & interrupt_mask))
 489                         break;
 490                 interrupt_allowed &= ~interrupt_mask;
 491                 unregister_randintr(intr);
 492                 break;
 493         case MEM_RETURNIRQ:
 494                 error = suser(td);
 495                 if (error != 0)
 496                         return (error);
 497                 *(u_int16_t *)data = interrupt_allowed;
 498                 break;
 499         }
 500         return (0);
 501 }
 502
 503 int
 504 mmpoll(dev_t dev, int events, struct thread *td)
 505 {
 506         switch (minor(dev)) {
 507         case 3:         /* /dev/random */
 508                 return random_poll(dev, events, td);
 509         case 4:         /* /dev/urandom */
 510         default:
 511                 return seltrue(dev, events, td);
 512         }
 513 }
 514
 515 int
 516 iszerodev(dev)
 517         dev_t dev;
 518 {
 519         return ((major(dev) == mem_cdevsw.d_maj)
 520           && minor(dev) == 12);
 521 }
 522
 523 static void
 524 mem_drvinit(void *unused)
 525 {
 526
 527         /* Initialise memory range handling */
 528         if (mem_range_softc.mr_op != NULL)
 529                 mem_range_softc.mr_op->init(&mem_range_softc);
 530
 531         cdevsw_add(&mem_cdevsw, 0xf0, 0);
 532         make_dev(&mem_cdevsw, 0, UID_ROOT, GID_KMEM, 0640, "mem");
 533         make_dev(&mem_cdevsw, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
 534         make_dev(&mem_cdevsw, 2, UID_ROOT, GID_WHEEL, 0666, "null");
 535         make_dev(&mem_cdevsw, 3, UID_ROOT, GID_WHEEL, 0644, "random");
 536         make_dev(&mem_cdevsw, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
 537         make_dev(&mem_cdevsw, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
 538         make_dev(&mem_cdevsw, 14, UID_ROOT, GID_WHEEL, 0600, "io");
 539 }
 540
 541 SYSINIT(memdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,mem_drvinit,NULL)
 542