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 1.29 2007/02/03 17:05:57 corecode 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 <vm/vm.h> 66 #include <vm/pmap.h> 67 #include <vm/vm_extern.h> 68 69 70 static d_open_t mmopen; 71 static d_close_t mmclose; 72 static d_read_t mmread; 73 static d_write_t mmwrite; 74 static d_ioctl_t mmioctl; 75 static d_mmap_t memmmap; 76 static d_poll_t mmpoll; 77 78 #define CDEV_MAJOR 2 79 static struct dev_ops mem_ops = { 80 { "mem", CDEV_MAJOR, D_MEM }, 81 .d_open = mmopen, 82 .d_close = mmclose, 83 .d_read = mmread, 84 .d_write = mmwrite, 85 .d_ioctl = mmioctl, 86 .d_poll = mmpoll, 87 .d_mmap = memmmap, 88 }; 89 90 static int rand_bolt; 91 static caddr_t zbuf; 92 93 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors"); 94 static int mem_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *); 95 static int random_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *); 96 97 struct mem_range_softc mem_range_softc; 98 99 100 static int 101 mmopen(struct dev_open_args *ap) 102 { 103 cdev_t dev = ap->a_head.a_dev; 104 int error; 105 106 switch (minor(dev)) { 107 case 0: 108 case 1: 109 if (ap->a_oflags & FWRITE) { 110 if (securelevel > 0 || kernel_mem_readonly) 111 return (EPERM); 112 } 113 error = 0; 114 break; 115 case 14: 116 error = suser_cred(ap->a_cred, 0); 117 if (error != 0) 118 break; 119 if (securelevel > 0 || kernel_mem_readonly) { 120 error = EPERM; 121 break; 122 } 123 error = cpu_set_iopl(); 124 break; 125 default: 126 error = 0; 127 break; 128 } 129 return (error); 130 } 131 132 static int 133 mmclose(struct dev_close_args *ap) 134 { 135 cdev_t dev = ap->a_head.a_dev; 136 int error; 137 138 switch (minor(dev)) { 139 case 14: 140 error = cpu_clr_iopl(); 141 break; 142 default: 143 error = 0; 144 break; 145 } 146 return (error); 147 } 148 149 150 static int 151 mmrw(cdev_t dev, struct uio *uio, int flags) 152 { 153 int o; 154 u_int c, v; 155 u_int poolsize; 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 &= ~PAGE_MASK; 176 pmap_kenter((vm_offset_t)ptvmmap, v); 177 o = (int)uio->uio_offset & PAGE_MASK; 178 c = (u_int)(PAGE_SIZE - ((int)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 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) 289 kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK); 290 bzero(zbuf, PAGE_SIZE); 291 } 292 c = min(iov->iov_len, PAGE_SIZE); 293 error = uiomove(zbuf, (int)c, uio); 294 continue; 295 default: 296 return (ENODEV); 297 } 298 if (error) 299 break; 300 iov->iov_base += c; 301 iov->iov_len -= c; 302 uio->uio_offset += c; 303 uio->uio_resid -= c; 304 } 305 if (buf) 306 kfree(buf, M_TEMP); 307 return (error); 308 } 309 310 static int 311 mmread(struct dev_read_args *ap) 312 { 313 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag)); 314 } 315 316 static int 317 mmwrite(struct dev_write_args *ap) 318 { 319 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag)); 320 } 321 322 323 324 325 326 /*******************************************************\ 327 * allow user processes to MMAP some memory sections * 328 * instead of going through read/write * 329 \*******************************************************/ 330 331 static int 332 memmmap(struct dev_mmap_args *ap) 333 { 334 cdev_t dev = ap->a_head.a_dev; 335 336 switch (minor(dev)) { 337 case 0: 338 /* 339 * minor device 0 is physical memory 340 */ 341 ap->a_result = i386_btop(ap->a_offset); 342 return 0; 343 case 1: 344 /* 345 * minor device 1 is kernel memory 346 */ 347 ap->a_result = i386_btop(vtophys(ap->a_offset)); 348 return 0; 349 350 default: 351 return EINVAL; 352 } 353 } 354 355 static int 356 mmioctl(struct dev_ioctl_args *ap) 357 { 358 cdev_t dev = ap->a_head.a_dev; 359 360 switch (minor(dev)) { 361 case 0: 362 return mem_ioctl(dev, ap->a_cmd, ap->a_data, 363 ap->a_fflag, ap->a_cred); 364 case 3: 365 case 4: 366 return random_ioctl(dev, ap->a_cmd, ap->a_data, 367 ap->a_fflag, ap->a_cred); 368 } 369 return (ENODEV); 370 } 371 372 /* 373 * Operations for changing memory attributes. 374 * 375 * This is basically just an ioctl shim for mem_range_attr_get 376 * and mem_range_attr_set. 377 */ 378 static int 379 mem_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred) 380 { 381 int nd, error = 0; 382 struct mem_range_op *mo = (struct mem_range_op *)data; 383 struct mem_range_desc *md; 384 385 /* is this for us? */ 386 if ((cmd != MEMRANGE_GET) && 387 (cmd != MEMRANGE_SET)) 388 return (ENOTTY); 389 390 /* any chance we can handle this? */ 391 if (mem_range_softc.mr_op == NULL) 392 return (EOPNOTSUPP); 393 394 /* do we have any descriptors? */ 395 if (mem_range_softc.mr_ndesc == 0) 396 return (ENXIO); 397 398 switch (cmd) { 399 case MEMRANGE_GET: 400 nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc); 401 if (nd > 0) { 402 md = (struct mem_range_desc *) 403 kmalloc(nd * sizeof(struct mem_range_desc), 404 M_MEMDESC, M_WAITOK); 405 error = mem_range_attr_get(md, &nd); 406 if (!error) 407 error = copyout(md, mo->mo_desc, 408 nd * sizeof(struct mem_range_desc)); 409 kfree(md, M_MEMDESC); 410 } else { 411 nd = mem_range_softc.mr_ndesc; 412 } 413 mo->mo_arg[0] = nd; 414 break; 415 416 case MEMRANGE_SET: 417 md = (struct mem_range_desc *)kmalloc(sizeof(struct mem_range_desc), 418 M_MEMDESC, M_WAITOK); 419 error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc)); 420 /* clamp description string */ 421 md->mr_owner[sizeof(md->mr_owner) - 1] = 0; 422 if (error == 0) 423 error = mem_range_attr_set(md, &mo->mo_arg[0]); 424 kfree(md, M_MEMDESC); 425 break; 426 } 427 return (error); 428 } 429 430 /* 431 * Implementation-neutral, kernel-callable functions for manipulating 432 * memory range attributes. 433 */ 434 int 435 mem_range_attr_get(struct mem_range_desc *mrd, int *arg) 436 { 437 /* can we handle this? */ 438 if (mem_range_softc.mr_op == NULL) 439 return (EOPNOTSUPP); 440 441 if (*arg == 0) { 442 *arg = mem_range_softc.mr_ndesc; 443 } else { 444 bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc)); 445 } 446 return (0); 447 } 448 449 int 450 mem_range_attr_set(struct mem_range_desc *mrd, int *arg) 451 { 452 /* can we handle this? */ 453 if (mem_range_softc.mr_op == NULL) 454 return (EOPNOTSUPP); 455 456 return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg)); 457 } 458 459 #ifdef SMP 460 void 461 mem_range_AP_init(void) 462 { 463 if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP) 464 return (mem_range_softc.mr_op->initAP(&mem_range_softc)); 465 } 466 #endif 467 468 static int 469 random_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred) 470 { 471 int error; 472 int intr; 473 474 /* 475 * Even inspecting the state is privileged, since it gives a hint 476 * about how easily the randomness might be guessed. 477 */ 478 error = 0; 479 480 switch (cmd) { 481 /* Really handled in upper layer */ 482 case FIOASYNC: 483 break; 484 case MEM_SETIRQ: 485 intr = *(int16_t *)data; 486 if ((error = suser_cred(cred, 0)) != 0) 487 break; 488 if (intr < 0 || intr >= MAX_INTS) 489 return (EINVAL); 490 register_randintr(intr); 491 break; 492 case MEM_CLEARIRQ: 493 intr = *(int16_t *)data; 494 if ((error = suser_cred(cred, 0)) != 0) 495 break; 496 if (intr < 0 || intr >= MAX_INTS) 497 return (EINVAL); 498 unregister_randintr(intr); 499 break; 500 case MEM_RETURNIRQ: 501 error = ENOTSUP; 502 break; 503 case MEM_FINDIRQ: 504 intr = *(int16_t *)data; 505 if ((error = suser_cred(cred, 0)) != 0) 506 break; 507 if (intr < 0 || intr >= MAX_INTS) 508 return (EINVAL); 509 intr = next_registered_randintr(intr); 510 if (intr == MAX_INTS) 511 return (ENOENT); 512 *(u_int16_t *)data = intr; 513 break; 514 default: 515 error = ENOTSUP; 516 break; 517 } 518 return (error); 519 } 520 521 int 522 mmpoll(struct dev_poll_args *ap) 523 { 524 cdev_t dev = ap->a_head.a_dev; 525 int revents; 526 527 switch (minor(dev)) { 528 case 3: /* /dev/random */ 529 revents = random_poll(dev, ap->a_events); 530 break; 531 case 4: /* /dev/urandom */ 532 default: 533 revents = seltrue(dev, ap->a_events); 534 break; 535 } 536 ap->a_events = revents; 537 return (0); 538 } 539 540 int 541 iszerodev(cdev_t dev) 542 { 543 return ((major(dev) == mem_ops.head.maj) 544 && minor(dev) == 12); 545 } 546 547 static void 548 mem_drvinit(void *unused) 549 { 550 551 /* Initialise memory range handling */ 552 if (mem_range_softc.mr_op != NULL) 553 mem_range_softc.mr_op->init(&mem_range_softc); 554 555 dev_ops_add(&mem_ops, 0xf0, 0); 556 make_dev(&mem_ops, 0, UID_ROOT, GID_KMEM, 0640, "mem"); 557 make_dev(&mem_ops, 1, UID_ROOT, GID_KMEM, 0640, "kmem"); 558 make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null"); 559 make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random"); 560 make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom"); 561 make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero"); 562 make_dev(&mem_ops, 14, UID_ROOT, GID_WHEEL, 0600, "io"); 563 } 564 565 SYSINIT(memdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,mem_drvinit,NULL) 566 567