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 1.24 2006/12/17 20:07:29 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 addr, 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 addr = trunc_page(uio->uio_offset); 196 eaddr = round_page(uio->uio_offset + c); 197 198 if (addr < (vm_offset_t)VADDR(PTDPTDI, 0)) 199 return EFAULT; 200 if (eaddr >= (vm_offset_t)VADDR(APTDPTDI, 0)) 201 return EFAULT; 202 for (; addr < eaddr; addr += PAGE_SIZE) 203 if (pmap_extract(kernel_pmap, addr) == 0) 204 return EFAULT; 205 206 if (!kernacc((caddr_t)(int)uio->uio_offset, c, 207 uio->uio_rw == UIO_READ ? 208 VM_PROT_READ : VM_PROT_WRITE)) 209 return (EFAULT); 210 error = uiomove((caddr_t)(int)uio->uio_offset, (int)c, uio); 211 continue; 212 } 213 case 2: 214 /* 215 * minor device 2 is EOF/RATHOLE 216 */ 217 if (uio->uio_rw == UIO_READ) 218 return (0); 219 c = iov->iov_len; 220 break; 221 case 3: 222 /* 223 * minor device 3 (/dev/random) is source of filth 224 * on read, seeder on write 225 */ 226 if (buf == NULL) 227 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK); 228 c = min(iov->iov_len, PAGE_SIZE); 229 if (uio->uio_rw == UIO_WRITE) { 230 error = uiomove(buf, (int)c, uio); 231 if (error == 0) 232 error = add_buffer_randomness(buf, c); 233 } else { 234 poolsize = read_random(buf, c); 235 if (poolsize == 0) { 236 if (buf) 237 kfree(buf, M_TEMP); 238 if ((flags & IO_NDELAY) != 0) 239 return (EWOULDBLOCK); 240 return (0); 241 } 242 c = min(c, poolsize); 243 error = uiomove(buf, (int)c, uio); 244 } 245 continue; 246 case 4: 247 /* 248 * minor device 4 (/dev/urandom) is source of muck 249 * on read, writes are disallowed. 250 */ 251 c = min(iov->iov_len, PAGE_SIZE); 252 if (uio->uio_rw == UIO_WRITE) { 253 error = EPERM; 254 break; 255 } 256 if (CURSIG(curproc) != 0) { 257 /* 258 * Use tsleep() to get the error code right. 259 * It should return immediately. 260 */ 261 error = tsleep(&rand_bolt, PCATCH, "urand", 1); 262 if (error != 0 && error != EWOULDBLOCK) 263 continue; 264 } 265 if (buf == NULL) 266 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK); 267 poolsize = read_random_unlimited(buf, c); 268 c = min(c, poolsize); 269 error = uiomove(buf, (int)c, uio); 270 continue; 271 case 12: 272 /* 273 * minor device 12 (/dev/zero) is source of nulls 274 * on read, write are disallowed. 275 */ 276 if (uio->uio_rw == UIO_WRITE) { 277 c = iov->iov_len; 278 break; 279 } 280 if (zbuf == NULL) { 281 zbuf = (caddr_t) 282 kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK); 283 bzero(zbuf, PAGE_SIZE); 284 } 285 c = min(iov->iov_len, PAGE_SIZE); 286 error = uiomove(zbuf, (int)c, uio); 287 continue; 288 default: 289 return (ENODEV); 290 } 291 if (error) 292 break; 293 iov->iov_base += c; 294 iov->iov_len -= c; 295 uio->uio_offset += c; 296 uio->uio_resid -= c; 297 } 298 if (buf) 299 kfree(buf, M_TEMP); 300 return (error); 301 } 302 303 static int 304 mmread(struct dev_read_args *ap) 305 { 306 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag)); 307 } 308 309 static int 310 mmwrite(struct dev_write_args *ap) 311 { 312 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag)); 313 } 314 315 316 317 318 319 /*******************************************************\ 320 * allow user processes to MMAP some memory sections * 321 * instead of going through read/write * 322 \*******************************************************/ 323 324 static int 325 memmmap(struct dev_mmap_args *ap) 326 { 327 cdev_t dev = ap->a_head.a_dev; 328 329 switch (minor(dev)) { 330 case 0: 331 /* 332 * minor device 0 is physical memory 333 */ 334 ap->a_result = i386_btop(ap->a_offset); 335 return 0; 336 case 1: 337 /* 338 * minor device 1 is kernel memory 339 */ 340 ap->a_result = i386_btop(vtophys(ap->a_offset)); 341 return 0; 342 343 default: 344 return EINVAL; 345 } 346 } 347 348 static int 349 mmioctl(struct dev_ioctl_args *ap) 350 { 351 cdev_t dev = ap->a_head.a_dev; 352 353 switch (minor(dev)) { 354 case 0: 355 return mem_ioctl(dev, ap->a_cmd, ap->a_data, 356 ap->a_fflag, ap->a_cred); 357 case 3: 358 case 4: 359 return random_ioctl(dev, ap->a_cmd, ap->a_data, 360 ap->a_fflag, ap->a_cred); 361 } 362 return (ENODEV); 363 } 364 365 /* 366 * Operations for changing memory attributes. 367 * 368 * This is basically just an ioctl shim for mem_range_attr_get 369 * and mem_range_attr_set. 370 */ 371 static int 372 mem_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred) 373 { 374 int nd, error = 0; 375 struct mem_range_op *mo = (struct mem_range_op *)data; 376 struct mem_range_desc *md; 377 378 /* is this for us? */ 379 if ((cmd != MEMRANGE_GET) && 380 (cmd != MEMRANGE_SET)) 381 return (ENOTTY); 382 383 /* any chance we can handle this? */ 384 if (mem_range_softc.mr_op == NULL) 385 return (EOPNOTSUPP); 386 387 /* do we have any descriptors? */ 388 if (mem_range_softc.mr_ndesc == 0) 389 return (ENXIO); 390 391 switch (cmd) { 392 case MEMRANGE_GET: 393 nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc); 394 if (nd > 0) { 395 md = (struct mem_range_desc *) 396 kmalloc(nd * sizeof(struct mem_range_desc), 397 M_MEMDESC, M_WAITOK); 398 error = mem_range_attr_get(md, &nd); 399 if (!error) 400 error = copyout(md, mo->mo_desc, 401 nd * sizeof(struct mem_range_desc)); 402 kfree(md, M_MEMDESC); 403 } else { 404 nd = mem_range_softc.mr_ndesc; 405 } 406 mo->mo_arg[0] = nd; 407 break; 408 409 case MEMRANGE_SET: 410 md = (struct mem_range_desc *)kmalloc(sizeof(struct mem_range_desc), 411 M_MEMDESC, M_WAITOK); 412 error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc)); 413 /* clamp description string */ 414 md->mr_owner[sizeof(md->mr_owner) - 1] = 0; 415 if (error == 0) 416 error = mem_range_attr_set(md, &mo->mo_arg[0]); 417 kfree(md, M_MEMDESC); 418 break; 419 } 420 return (error); 421 } 422 423 /* 424 * Implementation-neutral, kernel-callable functions for manipulating 425 * memory range attributes. 426 */ 427 int 428 mem_range_attr_get(struct mem_range_desc *mrd, int *arg) 429 { 430 /* can we handle this? */ 431 if (mem_range_softc.mr_op == NULL) 432 return (EOPNOTSUPP); 433 434 if (*arg == 0) { 435 *arg = mem_range_softc.mr_ndesc; 436 } else { 437 bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc)); 438 } 439 return (0); 440 } 441 442 int 443 mem_range_attr_set(struct mem_range_desc *mrd, int *arg) 444 { 445 /* can we handle this? */ 446 if (mem_range_softc.mr_op == NULL) 447 return (EOPNOTSUPP); 448 449 return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg)); 450 } 451 452 #ifdef SMP 453 void 454 mem_range_AP_init(void) 455 { 456 if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP) 457 return (mem_range_softc.mr_op->initAP(&mem_range_softc)); 458 } 459 #endif 460 461 static int 462 random_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred) 463 { 464 int error; 465 int intr; 466 467 /* 468 * Even inspecting the state is privileged, since it gives a hint 469 * about how easily the randomness might be guessed. 470 */ 471 error = 0; 472 473 switch (cmd) { 474 /* Really handled in upper layer */ 475 case FIOASYNC: 476 break; 477 case MEM_SETIRQ: 478 intr = *(int16_t *)data; 479 if ((error = suser_cred(cred, 0)) != 0) 480 break; 481 if (intr < 0 || intr >= MAX_INTS) 482 return (EINVAL); 483 register_randintr(intr); 484 break; 485 case MEM_CLEARIRQ: 486 intr = *(int16_t *)data; 487 if ((error = suser_cred(cred, 0)) != 0) 488 break; 489 if (intr < 0 || intr >= MAX_INTS) 490 return (EINVAL); 491 unregister_randintr(intr); 492 break; 493 case MEM_RETURNIRQ: 494 error = ENOTSUP; 495 break; 496 case MEM_FINDIRQ: 497 intr = *(int16_t *)data; 498 if ((error = suser_cred(cred, 0)) != 0) 499 break; 500 if (intr < 0 || intr >= MAX_INTS) 501 return (EINVAL); 502 intr = next_registered_randintr(intr); 503 if (intr == MAX_INTS) 504 return (ENOENT); 505 *(u_int16_t *)data = intr; 506 break; 507 default: 508 error = ENOTSUP; 509 break; 510 } 511 return (error); 512 } 513 514 int 515 mmpoll(struct dev_poll_args *ap) 516 { 517 cdev_t dev = ap->a_head.a_dev; 518 int revents; 519 520 switch (minor(dev)) { 521 case 3: /* /dev/random */ 522 revents = random_poll(dev, ap->a_events); 523 break; 524 case 4: /* /dev/urandom */ 525 default: 526 revents = seltrue(dev, ap->a_events); 527 break; 528 } 529 ap->a_events = revents; 530 return (0); 531 } 532 533 int 534 iszerodev(cdev_t dev) 535 { 536 return ((major(dev) == mem_ops.head.maj) 537 && minor(dev) == 12); 538 } 539 540 static void 541 mem_drvinit(void *unused) 542 { 543 544 /* Initialise memory range handling */ 545 if (mem_range_softc.mr_op != NULL) 546 mem_range_softc.mr_op->init(&mem_range_softc); 547 548 dev_ops_add(&mem_ops, 0xf0, 0); 549 make_dev(&mem_ops, 0, UID_ROOT, GID_KMEM, 0640, "mem"); 550 make_dev(&mem_ops, 1, UID_ROOT, GID_KMEM, 0640, "kmem"); 551 make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null"); 552 make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random"); 553 make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom"); 554 make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero"); 555 make_dev(&mem_ops, 14, UID_ROOT, GID_WHEEL, 0600, "io"); 556 } 557 558 SYSINIT(memdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,mem_drvinit,NULL) 559 560