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 1.26 2007/01/02 04:21:13 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 <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) && securelevel > 0) 110 return (EPERM); 111 error = 0; 112 break; 113 case 14: 114 error = suser_cred(ap->a_cred, 0); 115 if (error != 0) 116 break; 117 if (securelevel > 0) { 118 error = EPERM; 119 break; 120 } 121 error = cpu_set_iopl(); 122 break; 123 default: 124 error = 0; 125 break; 126 } 127 return (error); 128 } 129 130 static int 131 mmclose(struct dev_close_args *ap) 132 { 133 cdev_t dev = ap->a_head.a_dev; 134 int error; 135 136 switch (minor(dev)) { 137 case 14: 138 error = cpu_clr_iopl(); 139 break; 140 default: 141 error = 0; 142 break; 143 } 144 return (error); 145 } 146 147 148 static int 149 mmrw(cdev_t dev, struct uio *uio, int flags) 150 { 151 int o; 152 u_int c, v; 153 u_int poolsize; 154 struct iovec *iov; 155 int error = 0; 156 caddr_t buf = NULL; 157 158 while (uio->uio_resid > 0 && error == 0) { 159 iov = uio->uio_iov; 160 if (iov->iov_len == 0) { 161 uio->uio_iov++; 162 uio->uio_iovcnt--; 163 if (uio->uio_iovcnt < 0) 164 panic("mmrw"); 165 continue; 166 } 167 switch (minor(dev)) { 168 case 0: 169 /* 170 * minor device 0 is physical memory, /dev/mem 171 */ 172 v = uio->uio_offset; 173 v &= ~PAGE_MASK; 174 pmap_kenter((vm_offset_t)ptvmmap, v); 175 o = (int)uio->uio_offset & PAGE_MASK; 176 c = (u_int)(PAGE_SIZE - ((int)iov->iov_base & PAGE_MASK)); 177 c = min(c, (u_int)(PAGE_SIZE - o)); 178 c = min(c, (u_int)iov->iov_len); 179 error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio); 180 pmap_kremove((vm_offset_t)ptvmmap); 181 continue; 182 183 case 1: { 184 /* 185 * minor device 1 is kernel memory, /dev/kmem 186 */ 187 vm_offset_t saddr, eaddr; 188 c = iov->iov_len; 189 190 /* 191 * Make sure that all of the pages are currently 192 * resident so that we don't create any zero-fill 193 * pages. 194 */ 195 saddr = trunc_page(uio->uio_offset); 196 eaddr = round_page(uio->uio_offset + c); 197 198 if (saddr < KvaStart) 199 return EFAULT; 200 if (eaddr >= KvaEnd) 201 return EFAULT; 202 for (; saddr < eaddr; saddr += PAGE_SIZE) { 203 if (pmap_extract(&kernel_pmap, saddr) == 0) 204 return EFAULT; 205 } 206 207 if (!kernacc((caddr_t)(int)uio->uio_offset, c, 208 uio->uio_rw == UIO_READ ? 209 VM_PROT_READ : VM_PROT_WRITE)) 210 return (EFAULT); 211 error = uiomove((caddr_t)(vm_offset_t)uio->uio_offset, (int)c, uio); 212 continue; 213 } 214 case 2: 215 /* 216 * minor device 2 is EOF/RATHOLE 217 */ 218 if (uio->uio_rw == UIO_READ) 219 return (0); 220 c = iov->iov_len; 221 break; 222 case 3: 223 /* 224 * minor device 3 (/dev/random) is source of filth 225 * on read, seeder on write 226 */ 227 if (buf == NULL) 228 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK); 229 c = min(iov->iov_len, PAGE_SIZE); 230 if (uio->uio_rw == UIO_WRITE) { 231 error = uiomove(buf, (int)c, uio); 232 if (error == 0) 233 error = add_buffer_randomness(buf, c); 234 } else { 235 poolsize = read_random(buf, c); 236 if (poolsize == 0) { 237 if (buf) 238 kfree(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 } 246 continue; 247 case 4: 248 /* 249 * minor device 4 (/dev/urandom) is source of muck 250 * on read, writes are disallowed. 251 */ 252 c = min(iov->iov_len, PAGE_SIZE); 253 if (uio->uio_rw == UIO_WRITE) { 254 error = EPERM; 255 break; 256 } 257 if (CURSIG(curproc) != 0) { 258 /* 259 * Use tsleep() to get the error code right. 260 * It should return immediately. 261 */ 262 error = tsleep(&rand_bolt, PCATCH, "urand", 1); 263 if (error != 0 && error != EWOULDBLOCK) 264 continue; 265 } 266 if (buf == NULL) 267 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK); 268 poolsize = read_random_unlimited(buf, c); 269 c = min(c, poolsize); 270 error = uiomove(buf, (int)c, uio); 271 continue; 272 case 12: 273 /* 274 * minor device 12 (/dev/zero) is source of nulls 275 * on read, write are disallowed. 276 */ 277 if (uio->uio_rw == UIO_WRITE) { 278 c = iov->iov_len; 279 break; 280 } 281 if (zbuf == NULL) { 282 zbuf = (caddr_t) 283 kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK); 284 bzero(zbuf, PAGE_SIZE); 285 } 286 c = min(iov->iov_len, PAGE_SIZE); 287 error = uiomove(zbuf, (int)c, uio); 288 continue; 289 default: 290 return (ENODEV); 291 } 292 if (error) 293 break; 294 iov->iov_base += c; 295 iov->iov_len -= c; 296 uio->uio_offset += c; 297 uio->uio_resid -= c; 298 } 299 if (buf) 300 kfree(buf, M_TEMP); 301 return (error); 302 } 303 304 static int 305 mmread(struct dev_read_args *ap) 306 { 307 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag)); 308 } 309 310 static int 311 mmwrite(struct dev_write_args *ap) 312 { 313 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag)); 314 } 315 316 317 318 319 320 /*******************************************************\ 321 * allow user processes to MMAP some memory sections * 322 * instead of going through read/write * 323 \*******************************************************/ 324 325 static int 326 memmmap(struct dev_mmap_args *ap) 327 { 328 cdev_t dev = ap->a_head.a_dev; 329 330 switch (minor(dev)) { 331 case 0: 332 /* 333 * minor device 0 is physical memory 334 */ 335 ap->a_result = i386_btop(ap->a_offset); 336 return 0; 337 case 1: 338 /* 339 * minor device 1 is kernel memory 340 */ 341 ap->a_result = i386_btop(vtophys(ap->a_offset)); 342 return 0; 343 344 default: 345 return EINVAL; 346 } 347 } 348 349 static int 350 mmioctl(struct dev_ioctl_args *ap) 351 { 352 cdev_t dev = ap->a_head.a_dev; 353 354 switch (minor(dev)) { 355 case 0: 356 return mem_ioctl(dev, ap->a_cmd, ap->a_data, 357 ap->a_fflag, ap->a_cred); 358 case 3: 359 case 4: 360 return random_ioctl(dev, ap->a_cmd, ap->a_data, 361 ap->a_fflag, ap->a_cred); 362 } 363 return (ENODEV); 364 } 365 366 /* 367 * Operations for changing memory attributes. 368 * 369 * This is basically just an ioctl shim for mem_range_attr_get 370 * and mem_range_attr_set. 371 */ 372 static int 373 mem_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred) 374 { 375 int nd, error = 0; 376 struct mem_range_op *mo = (struct mem_range_op *)data; 377 struct mem_range_desc *md; 378 379 /* is this for us? */ 380 if ((cmd != MEMRANGE_GET) && 381 (cmd != MEMRANGE_SET)) 382 return (ENOTTY); 383 384 /* any chance we can handle this? */ 385 if (mem_range_softc.mr_op == NULL) 386 return (EOPNOTSUPP); 387 388 /* do we have any descriptors? */ 389 if (mem_range_softc.mr_ndesc == 0) 390 return (ENXIO); 391 392 switch (cmd) { 393 case MEMRANGE_GET: 394 nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc); 395 if (nd > 0) { 396 md = (struct mem_range_desc *) 397 kmalloc(nd * sizeof(struct mem_range_desc), 398 M_MEMDESC, M_WAITOK); 399 error = mem_range_attr_get(md, &nd); 400 if (!error) 401 error = copyout(md, mo->mo_desc, 402 nd * sizeof(struct mem_range_desc)); 403 kfree(md, M_MEMDESC); 404 } else { 405 nd = mem_range_softc.mr_ndesc; 406 } 407 mo->mo_arg[0] = nd; 408 break; 409 410 case MEMRANGE_SET: 411 md = (struct mem_range_desc *)kmalloc(sizeof(struct mem_range_desc), 412 M_MEMDESC, M_WAITOK); 413 error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc)); 414 /* clamp description string */ 415 md->mr_owner[sizeof(md->mr_owner) - 1] = 0; 416 if (error == 0) 417 error = mem_range_attr_set(md, &mo->mo_arg[0]); 418 kfree(md, M_MEMDESC); 419 break; 420 } 421 return (error); 422 } 423 424 /* 425 * Implementation-neutral, kernel-callable functions for manipulating 426 * memory range attributes. 427 */ 428 int 429 mem_range_attr_get(struct mem_range_desc *mrd, int *arg) 430 { 431 /* can we handle this? */ 432 if (mem_range_softc.mr_op == NULL) 433 return (EOPNOTSUPP); 434 435 if (*arg == 0) { 436 *arg = mem_range_softc.mr_ndesc; 437 } else { 438 bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc)); 439 } 440 return (0); 441 } 442 443 int 444 mem_range_attr_set(struct mem_range_desc *mrd, int *arg) 445 { 446 /* can we handle this? */ 447 if (mem_range_softc.mr_op == NULL) 448 return (EOPNOTSUPP); 449 450 return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg)); 451 } 452 453 #ifdef SMP 454 void 455 mem_range_AP_init(void) 456 { 457 if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP) 458 return (mem_range_softc.mr_op->initAP(&mem_range_softc)); 459 } 460 #endif 461 462 static int 463 random_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred) 464 { 465 int error; 466 int intr; 467 468 /* 469 * Even inspecting the state is privileged, since it gives a hint 470 * about how easily the randomness might be guessed. 471 */ 472 error = 0; 473 474 switch (cmd) { 475 /* Really handled in upper layer */ 476 case FIOASYNC: 477 break; 478 case MEM_SETIRQ: 479 intr = *(int16_t *)data; 480 if ((error = suser_cred(cred, 0)) != 0) 481 break; 482 if (intr < 0 || intr >= MAX_INTS) 483 return (EINVAL); 484 register_randintr(intr); 485 break; 486 case MEM_CLEARIRQ: 487 intr = *(int16_t *)data; 488 if ((error = suser_cred(cred, 0)) != 0) 489 break; 490 if (intr < 0 || intr >= MAX_INTS) 491 return (EINVAL); 492 unregister_randintr(intr); 493 break; 494 case MEM_RETURNIRQ: 495 error = ENOTSUP; 496 break; 497 case MEM_FINDIRQ: 498 intr = *(int16_t *)data; 499 if ((error = suser_cred(cred, 0)) != 0) 500 break; 501 if (intr < 0 || intr >= MAX_INTS) 502 return (EINVAL); 503 intr = next_registered_randintr(intr); 504 if (intr == MAX_INTS) 505 return (ENOENT); 506 *(u_int16_t *)data = intr; 507 break; 508 default: 509 error = ENOTSUP; 510 break; 511 } 512 return (error); 513 } 514 515 int 516 mmpoll(struct dev_poll_args *ap) 517 { 518 cdev_t dev = ap->a_head.a_dev; 519 int revents; 520 521 switch (minor(dev)) { 522 case 3: /* /dev/random */ 523 revents = random_poll(dev, ap->a_events); 524 break; 525 case 4: /* /dev/urandom */ 526 default: 527 revents = seltrue(dev, ap->a_events); 528 break; 529 } 530 ap->a_events = revents; 531 return (0); 532 } 533 534 int 535 iszerodev(cdev_t dev) 536 { 537 return ((major(dev) == mem_ops.head.maj) 538 && minor(dev) == 12); 539 } 540 541 static void 542 mem_drvinit(void *unused) 543 { 544 545 /* Initialise memory range handling */ 546 if (mem_range_softc.mr_op != NULL) 547 mem_range_softc.mr_op->init(&mem_range_softc); 548 549 dev_ops_add(&mem_ops, 0xf0, 0); 550 make_dev(&mem_ops, 0, UID_ROOT, GID_KMEM, 0640, "mem"); 551 make_dev(&mem_ops, 1, UID_ROOT, GID_KMEM, 0640, "kmem"); 552 make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null"); 553 make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random"); 554 make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom"); 555 make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero"); 556 make_dev(&mem_ops, 14, UID_ROOT, GID_WHEEL, 0600, "io"); 557 } 558 559 SYSINIT(memdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,mem_drvinit,NULL) 560 561