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 60 #include <sys/signal2.h> 61 #include <sys/mplock2.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 static d_mmap_t memmmap; 74 static d_kqfilter_t mmkqfilter; 75 76 #define CDEV_MAJOR 2 77 static struct dev_ops mem_ops = { 78 { "mem", 0, D_MPSAFE }, 79 .d_open = mmopen, 80 .d_close = mmclose, 81 .d_read = mmread, 82 .d_write = mmwrite, 83 .d_ioctl = mmioctl, 84 .d_kqfilter = mmkqfilter, 85 .d_mmap = memmmap, 86 }; 87 88 static int rand_bolt; 89 static caddr_t zbuf; 90 static cdev_t zerodev = NULL; 91 92 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors"); 93 static int mem_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *); 94 static int random_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *); 95 96 struct mem_range_softc mem_range_softc; 97 98 99 static int 100 mmopen(struct dev_open_args *ap) 101 { 102 cdev_t dev = ap->a_head.a_dev; 103 int error; 104 105 switch (minor(dev)) { 106 case 0: 107 case 1: 108 if (ap->a_oflags & FWRITE) { 109 if (securelevel > 0 || kernel_mem_readonly) 110 return (EPERM); 111 } 112 error = 0; 113 break; 114 case 14: 115 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0); 116 if (error != 0) 117 break; 118 if (securelevel > 0 || kernel_mem_readonly) { 119 error = EPERM; 120 break; 121 } 122 error = cpu_set_iopl(); 123 break; 124 default: 125 error = 0; 126 break; 127 } 128 return (error); 129 } 130 131 static int 132 mmclose(struct dev_close_args *ap) 133 { 134 cdev_t dev = ap->a_head.a_dev; 135 int error; 136 137 switch (minor(dev)) { 138 case 14: 139 error = cpu_clr_iopl(); 140 break; 141 default: 142 error = 0; 143 break; 144 } 145 return (error); 146 } 147 148 149 static int 150 mmrw(cdev_t dev, struct uio *uio, int flags) 151 { 152 int o; 153 u_int c; 154 u_int poolsize; 155 u_long v; 156 struct iovec *iov; 157 int error = 0; 158 caddr_t buf = NULL; 159 160 while (uio->uio_resid > 0 && error == 0) { 161 iov = uio->uio_iov; 162 if (iov->iov_len == 0) { 163 uio->uio_iov++; 164 uio->uio_iovcnt--; 165 if (uio->uio_iovcnt < 0) 166 panic("mmrw"); 167 continue; 168 } 169 switch (minor(dev)) { 170 case 0: 171 /* 172 * minor device 0 is physical memory, /dev/mem 173 */ 174 v = uio->uio_offset; 175 v &= ~(long)PAGE_MASK; 176 pmap_kenter((vm_offset_t)ptvmmap, v); 177 o = (int)uio->uio_offset & PAGE_MASK; 178 c = (u_int)(PAGE_SIZE - ((uintptr_t)iov->iov_base & PAGE_MASK)); 179 c = min(c, (u_int)(PAGE_SIZE - o)); 180 c = min(c, (u_int)iov->iov_len); 181 error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio); 182 pmap_kremove((vm_offset_t)ptvmmap); 183 continue; 184 185 case 1: { 186 /* 187 * minor device 1 is kernel memory, /dev/kmem 188 */ 189 vm_offset_t saddr, eaddr; 190 int prot; 191 192 c = iov->iov_len; 193 194 /* 195 * Make sure that all of the pages are currently 196 * resident so that we don't create any zero-fill 197 * pages. 198 */ 199 saddr = trunc_page(uio->uio_offset); 200 eaddr = round_page(uio->uio_offset + c); 201 if (saddr > eaddr) 202 return EFAULT; 203 204 /* 205 * Make sure the kernel addresses are mapped. 206 * platform_direct_mapped() can be used to bypass 207 * default mapping via the page table (virtual kernels 208 * contain a lot of out-of-band data). 209 */ 210 prot = VM_PROT_READ; 211 if (uio->uio_rw != UIO_READ) 212 prot |= VM_PROT_WRITE; 213 error = kvm_access_check(saddr, eaddr, prot); 214 if (error) 215 return (error); 216 error = uiomove((caddr_t)(vm_offset_t)uio->uio_offset, 217 (int)c, uio); 218 continue; 219 } 220 case 2: 221 /* 222 * minor device 2 (/dev/null) is EOF/RATHOLE 223 */ 224 if (uio->uio_rw == UIO_READ) 225 return (0); 226 c = iov->iov_len; 227 break; 228 case 3: 229 /* 230 * minor device 3 (/dev/random) is source of filth 231 * on read, seeder on write 232 */ 233 if (buf == NULL) 234 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK); 235 c = min(iov->iov_len, PAGE_SIZE); 236 if (uio->uio_rw == UIO_WRITE) { 237 error = uiomove(buf, (int)c, uio); 238 if (error == 0) 239 error = add_buffer_randomness(buf, c); 240 } else { 241 poolsize = read_random(buf, c); 242 if (poolsize == 0) { 243 if (buf) 244 kfree(buf, M_TEMP); 245 if ((flags & IO_NDELAY) != 0) 246 return (EWOULDBLOCK); 247 return (0); 248 } 249 c = min(c, poolsize); 250 error = uiomove(buf, (int)c, uio); 251 } 252 continue; 253 case 4: 254 /* 255 * minor device 4 (/dev/urandom) is source of muck 256 * on read, writes are disallowed. 257 */ 258 c = min(iov->iov_len, PAGE_SIZE); 259 if (uio->uio_rw == UIO_WRITE) { 260 error = EPERM; 261 break; 262 } 263 if (CURSIG(curthread->td_lwp) != 0) { 264 /* 265 * Use tsleep() to get the error code right. 266 * It should return immediately. 267 */ 268 error = tsleep(&rand_bolt, PCATCH, "urand", 1); 269 if (error != 0 && error != EWOULDBLOCK) 270 continue; 271 } 272 if (buf == NULL) 273 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK); 274 poolsize = read_random_unlimited(buf, c); 275 c = min(c, poolsize); 276 error = uiomove(buf, (int)c, uio); 277 continue; 278 case 12: 279 /* 280 * minor device 12 (/dev/zero) is source of nulls 281 * on read, write are disallowed. 282 */ 283 if (uio->uio_rw == UIO_WRITE) { 284 c = iov->iov_len; 285 break; 286 } 287 if (zbuf == NULL) { 288 zbuf = (caddr_t)kmalloc(PAGE_SIZE, M_TEMP, 289 M_WAITOK | M_ZERO); 290 } 291 c = min(iov->iov_len, PAGE_SIZE); 292 error = uiomove(zbuf, (int)c, uio); 293 continue; 294 default: 295 return (ENODEV); 296 } 297 if (error) 298 break; 299 iov->iov_base = (char *)iov->iov_base + c; 300 iov->iov_len -= c; 301 uio->uio_offset += c; 302 uio->uio_resid -= c; 303 } 304 if (buf) 305 kfree(buf, M_TEMP); 306 return (error); 307 } 308 309 static int 310 mmread(struct dev_read_args *ap) 311 { 312 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag)); 313 } 314 315 static int 316 mmwrite(struct dev_write_args *ap) 317 { 318 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag)); 319 } 320 321 322 323 324 325 /*******************************************************\ 326 * allow user processes to MMAP some memory sections * 327 * instead of going through read/write * 328 \*******************************************************/ 329 330 static int 331 memmmap(struct dev_mmap_args *ap) 332 { 333 cdev_t dev = ap->a_head.a_dev; 334 335 switch (minor(dev)) { 336 case 0: 337 /* 338 * minor device 0 is physical memory 339 */ 340 #if defined(__i386__) 341 ap->a_result = i386_btop(ap->a_offset); 342 #elif defined(__x86_64__) 343 ap->a_result = x86_64_btop(ap->a_offset); 344 #endif 345 return 0; 346 case 1: 347 /* 348 * minor device 1 is kernel memory 349 */ 350 #if defined(__i386__) 351 ap->a_result = i386_btop(vtophys(ap->a_offset)); 352 #elif defined(__x86_64__) 353 ap->a_result = x86_64_btop(vtophys(ap->a_offset)); 354 #endif 355 return 0; 356 357 default: 358 return EINVAL; 359 } 360 } 361 362 static int 363 mmioctl(struct dev_ioctl_args *ap) 364 { 365 cdev_t dev = ap->a_head.a_dev; 366 int error; 367 368 get_mplock(); 369 370 switch (minor(dev)) { 371 case 0: 372 error = mem_ioctl(dev, ap->a_cmd, ap->a_data, 373 ap->a_fflag, ap->a_cred); 374 break; 375 case 3: 376 case 4: 377 error = random_ioctl(dev, ap->a_cmd, ap->a_data, 378 ap->a_fflag, ap->a_cred); 379 break; 380 default: 381 error = ENODEV; 382 break; 383 } 384 385 rel_mplock(); 386 return (error); 387 } 388 389 /* 390 * Operations for changing memory attributes. 391 * 392 * This is basically just an ioctl shim for mem_range_attr_get 393 * and mem_range_attr_set. 394 */ 395 static int 396 mem_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred) 397 { 398 int nd, error = 0; 399 struct mem_range_op *mo = (struct mem_range_op *)data; 400 struct mem_range_desc *md; 401 402 /* is this for us? */ 403 if ((cmd != MEMRANGE_GET) && 404 (cmd != MEMRANGE_SET)) 405 return (ENOTTY); 406 407 /* any chance we can handle this? */ 408 if (mem_range_softc.mr_op == NULL) 409 return (EOPNOTSUPP); 410 411 /* do we have any descriptors? */ 412 if (mem_range_softc.mr_ndesc == 0) 413 return (ENXIO); 414 415 switch (cmd) { 416 case MEMRANGE_GET: 417 nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc); 418 if (nd > 0) { 419 md = (struct mem_range_desc *) 420 kmalloc(nd * sizeof(struct mem_range_desc), 421 M_MEMDESC, M_WAITOK); 422 error = mem_range_attr_get(md, &nd); 423 if (!error) 424 error = copyout(md, mo->mo_desc, 425 nd * sizeof(struct mem_range_desc)); 426 kfree(md, M_MEMDESC); 427 } else { 428 nd = mem_range_softc.mr_ndesc; 429 } 430 mo->mo_arg[0] = nd; 431 break; 432 433 case MEMRANGE_SET: 434 md = (struct mem_range_desc *)kmalloc(sizeof(struct mem_range_desc), 435 M_MEMDESC, M_WAITOK); 436 error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc)); 437 /* clamp description string */ 438 md->mr_owner[sizeof(md->mr_owner) - 1] = 0; 439 if (error == 0) 440 error = mem_range_attr_set(md, &mo->mo_arg[0]); 441 kfree(md, M_MEMDESC); 442 break; 443 } 444 return (error); 445 } 446 447 /* 448 * Implementation-neutral, kernel-callable functions for manipulating 449 * memory range attributes. 450 */ 451 int 452 mem_range_attr_get(struct mem_range_desc *mrd, int *arg) 453 { 454 /* can we handle this? */ 455 if (mem_range_softc.mr_op == NULL) 456 return (EOPNOTSUPP); 457 458 if (*arg == 0) { 459 *arg = mem_range_softc.mr_ndesc; 460 } else { 461 bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc)); 462 } 463 return (0); 464 } 465 466 int 467 mem_range_attr_set(struct mem_range_desc *mrd, int *arg) 468 { 469 /* can we handle this? */ 470 if (mem_range_softc.mr_op == NULL) 471 return (EOPNOTSUPP); 472 473 return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg)); 474 } 475 476 void 477 mem_range_AP_init(void) 478 { 479 if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP) 480 mem_range_softc.mr_op->initAP(&mem_range_softc); 481 } 482 483 static int 484 random_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred) 485 { 486 int error; 487 int intr; 488 489 /* 490 * Even inspecting the state is privileged, since it gives a hint 491 * about how easily the randomness might be guessed. 492 */ 493 error = 0; 494 495 switch (cmd) { 496 /* Really handled in upper layer */ 497 case FIOASYNC: 498 break; 499 case MEM_SETIRQ: 500 intr = *(int16_t *)data; 501 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0) 502 break; 503 if (intr < 0 || intr >= MAX_INTS) 504 return (EINVAL); 505 register_randintr(intr); 506 break; 507 case MEM_CLEARIRQ: 508 intr = *(int16_t *)data; 509 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0) 510 break; 511 if (intr < 0 || intr >= MAX_INTS) 512 return (EINVAL); 513 unregister_randintr(intr); 514 break; 515 case MEM_RETURNIRQ: 516 error = ENOTSUP; 517 break; 518 case MEM_FINDIRQ: 519 intr = *(int16_t *)data; 520 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0) 521 break; 522 if (intr < 0 || intr >= MAX_INTS) 523 return (EINVAL); 524 intr = next_registered_randintr(intr); 525 if (intr == MAX_INTS) 526 return (ENOENT); 527 *(u_int16_t *)data = intr; 528 break; 529 default: 530 error = ENOTSUP; 531 break; 532 } 533 return (error); 534 } 535 536 static int 537 mm_filter_read(struct knote *kn, long hint) 538 { 539 return (1); 540 } 541 542 static int 543 mm_filter_write(struct knote *kn, long hint) 544 { 545 return (1); 546 } 547 548 static void 549 dummy_filter_detach(struct knote *kn) {} 550 551 /* Implemented in kern_nrandom.c */ 552 static struct filterops random_read_filtops = 553 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, random_filter_read }; 554 555 static struct filterops mm_read_filtops = 556 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_read }; 557 558 static struct filterops mm_write_filtops = 559 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_write }; 560 561 int 562 mmkqfilter(struct dev_kqfilter_args *ap) 563 { 564 struct knote *kn = ap->a_kn; 565 cdev_t dev = ap->a_head.a_dev; 566 567 ap->a_result = 0; 568 switch (kn->kn_filter) { 569 case EVFILT_READ: 570 switch (minor(dev)) { 571 case 3: 572 kn->kn_fop = &random_read_filtops; 573 break; 574 default: 575 kn->kn_fop = &mm_read_filtops; 576 break; 577 } 578 break; 579 case EVFILT_WRITE: 580 kn->kn_fop = &mm_write_filtops; 581 break; 582 default: 583 ap->a_result = EOPNOTSUPP; 584 return (0); 585 } 586 587 return (0); 588 } 589 590 int 591 iszerodev(cdev_t dev) 592 { 593 return (zerodev == dev); 594 } 595 596 static void 597 mem_drvinit(void *unused) 598 { 599 600 /* Initialise memory range handling */ 601 if (mem_range_softc.mr_op != NULL) 602 mem_range_softc.mr_op->init(&mem_range_softc); 603 604 make_dev(&mem_ops, 0, UID_ROOT, GID_KMEM, 0640, "mem"); 605 make_dev(&mem_ops, 1, UID_ROOT, GID_KMEM, 0640, "kmem"); 606 make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null"); 607 make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random"); 608 make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom"); 609 zerodev = make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero"); 610 make_dev(&mem_ops, 14, UID_ROOT, GID_WHEEL, 0600, "io"); 611 } 612 613 SYSINIT(memdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,mem_drvinit,NULL) 614 615