1 /* 2 * ---------------------------------------------------------------------------- 3 * "THE BEER-WARE LICENSE" (Revision 42): 4 * <phk@FreeBSD.ORG> wrote this file. As long as you retain this notice you 5 * can do whatever you want with this stuff. If we meet some day, and you think 6 * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp 7 * ---------------------------------------------------------------------------- 8 * 9 * $FreeBSD$ 10 * 11 */ 12 13 /* 14 * The following functions are based in the vn(4) driver: mdstart_swap(), 15 * mdstart_vnode(), mdcreate_swap(), mdcreate_vnode() and mddestroy(), 16 * and as such under the following copyright: 17 * 18 * Copyright (c) 1988 University of Utah. 19 * Copyright (c) 1990, 1993 20 * The Regents of the University of California. All rights reserved. 21 * 22 * This code is derived from software contributed to Berkeley by 23 * the Systems Programming Group of the University of Utah Computer 24 * Science Department. 25 * 26 * Redistribution and use in source and binary forms, with or without 27 * modification, are permitted provided that the following conditions 28 * are met: 29 * 1. Redistributions of source code must retain the above copyright 30 * notice, this list of conditions and the following disclaimer. 31 * 2. Redistributions in binary form must reproduce the above copyright 32 * notice, this list of conditions and the following disclaimer in the 33 * documentation and/or other materials provided with the distribution. 34 * 3. All advertising materials mentioning features or use of this software 35 * must display the following acknowledgement: 36 * This product includes software developed by the University of 37 * California, Berkeley and its contributors. 38 * 4. Neither the name of the University nor the names of its contributors 39 * may be used to endorse or promote products derived from this software 40 * without specific prior written permission. 41 * 42 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 43 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 44 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 45 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 46 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 47 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 48 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 49 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 50 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 51 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 52 * SUCH DAMAGE. 53 * 54 * from: Utah Hdr: vn.c 1.13 94/04/02 55 * 56 * from: @(#)vn.c 8.6 (Berkeley) 4/1/94 57 * From: src/sys/dev/vn/vn.c,v 1.122 2000/12/16 16:06:03 58 */ 59 60 #include "opt_mfs.h" /* We have adopted some tasks from MFS */ 61 #include "opt_md.h" 62 63 #include <sys/param.h> 64 #include <sys/systm.h> 65 #include <sys/bio.h> 66 #include <sys/conf.h> 67 #include <sys/devicestat.h> 68 #include <sys/disk.h> 69 #include <sys/fcntl.h> 70 #include <sys/kernel.h> 71 #include <sys/linker.h> 72 #include <sys/lock.h> 73 #include <sys/malloc.h> 74 #include <sys/mdioctl.h> 75 #include <sys/namei.h> 76 #include <sys/proc.h> 77 #include <sys/queue.h> 78 #include <sys/sysctl.h> 79 #include <sys/vnode.h> 80 81 #include <machine/atomic.h> 82 83 #include <vm/vm.h> 84 #include <vm/vm_object.h> 85 #include <vm/vm_page.h> 86 #include <vm/vm_pager.h> 87 #include <vm/vm_zone.h> 88 #include <vm/swap_pager.h> 89 90 #define MD_MODVER 1 91 92 #ifndef MD_NSECT 93 #define MD_NSECT (10000 * 2) 94 #endif 95 96 MALLOC_DEFINE(M_MD, "MD disk", "Memory Disk"); 97 MALLOC_DEFINE(M_MDSECT, "MD sectors", "Memory Disk Sectors"); 98 99 static int md_debug; 100 SYSCTL_INT(_debug, OID_AUTO, mddebug, CTLFLAG_RW, &md_debug, 0, ""); 101 102 #if defined(MFS_ROOT) && !defined(MD_ROOT) 103 #define MD_ROOT MFS_ROOT 104 #warning "option MFS_ROOT has been superceeded by MD_ROOT" 105 #endif 106 107 #if defined(MFS_ROOT_SIZE) && !defined(MD_ROOT_SIZE) 108 #define MD_ROOT_SIZE MFS_ROOT_SIZE 109 #warning "option MFS_ROOT_SIZE has been superceeded by MD_ROOT_SIZE" 110 #endif 111 112 #if defined(MD_ROOT) && defined(MD_ROOT_SIZE) 113 /* Image gets put here: */ 114 static u_char mfs_root[MD_ROOT_SIZE*1024] = "MFS Filesystem goes here"; 115 static u_char end_mfs_root[] __unused = "MFS Filesystem had better STOP here"; 116 #endif 117 118 static int mdrootready; 119 static int mdunits; 120 static dev_t status_dev = 0; 121 122 123 #define CDEV_MAJOR 95 124 125 static d_strategy_t mdstrategy; 126 static d_open_t mdopen; 127 static d_ioctl_t mdioctl, mdctlioctl; 128 129 static struct cdevsw md_cdevsw = { 130 /* open */ mdopen, 131 /* close */ nullclose, 132 /* read */ physread, 133 /* write */ physwrite, 134 /* ioctl */ mdioctl, 135 /* poll */ nopoll, 136 /* mmap */ nommap, 137 /* strategy */ mdstrategy, 138 /* name */ MD_NAME, 139 /* maj */ CDEV_MAJOR, 140 /* dump */ nodump, 141 /* psize */ nopsize, 142 /* flags */ D_DISK | D_CANFREE | D_MEMDISK, 143 }; 144 145 static struct cdevsw mdctl_cdevsw = { 146 /* open */ nullopen, 147 /* close */ nullclose, 148 /* read */ noread, 149 /* write */ nowrite, 150 /* ioctl */ mdctlioctl, 151 /* poll */ nopoll, 152 /* mmap */ nommap, 153 /* strategy */ nostrategy, 154 /* name */ MD_NAME, 155 /* maj */ CDEV_MAJOR 156 }; 157 158 static struct cdevsw mddisk_cdevsw; 159 160 static LIST_HEAD(, md_s) md_softc_list = LIST_HEAD_INITIALIZER(&md_softc_list); 161 162 struct md_s { 163 int unit; 164 LIST_ENTRY(md_s) list; 165 struct devstat stats; 166 struct bio_queue_head bio_queue; 167 struct disk disk; 168 dev_t dev; 169 int busy; 170 enum md_types type; 171 unsigned nsect; 172 unsigned secsize; 173 unsigned flags; 174 175 /* MD_MALLOC related fields */ 176 u_char **secp; 177 178 /* MD_PRELOAD related fields */ 179 u_char *pl_ptr; 180 unsigned pl_len; 181 182 /* MD_VNODE related fields */ 183 struct vnode *vnode; 184 struct ucred *cred; 185 186 /* MD_OBJET related fields */ 187 vm_object_t object; 188 }; 189 190 static int 191 mdopen(dev_t dev, int flag, int fmt, struct proc *p) 192 { 193 struct md_s *sc; 194 struct disklabel *dl; 195 196 if (md_debug) 197 printf("mdopen(%s %x %x %p)\n", 198 devtoname(dev), flag, fmt, p); 199 200 sc = dev->si_drv1; 201 202 dl = &sc->disk.d_label; 203 bzero(dl, sizeof(*dl)); 204 dl->d_secsize = sc->secsize; 205 dl->d_nsectors = sc->nsect > 63 ? 63 : sc->nsect; 206 dl->d_ntracks = 1; 207 dl->d_secpercyl = dl->d_nsectors * dl->d_ntracks; 208 dl->d_secperunit = sc->nsect; 209 dl->d_ncylinders = dl->d_secperunit / dl->d_secpercyl; 210 return (0); 211 } 212 213 static int 214 mdioctl(dev_t dev, u_long cmd, caddr_t addr, int flags, struct proc *p) 215 { 216 217 if (md_debug) 218 printf("mdioctl(%s %lx %p %x %p)\n", 219 devtoname(dev), cmd, addr, flags, p); 220 221 return (ENOIOCTL); 222 } 223 224 static void 225 mdstart_malloc(struct md_s *sc) 226 { 227 int i; 228 struct bio *bp; 229 devstat_trans_flags dop; 230 u_char *secp, **secpp, *dst; 231 unsigned secno, nsec, secval, uc; 232 233 for (;;) { 234 /* XXX: LOCK(unique unit numbers) */ 235 bp = bioq_first(&sc->bio_queue); 236 if (bp) 237 bioq_remove(&sc->bio_queue, bp); 238 /* XXX: UNLOCK(unique unit numbers) */ 239 if (!bp) 240 break; 241 242 devstat_start_transaction(&sc->stats); 243 244 if (bp->bio_cmd == BIO_DELETE) 245 dop = DEVSTAT_NO_DATA; 246 else if (bp->bio_cmd == BIO_READ) 247 dop = DEVSTAT_READ; 248 else 249 dop = DEVSTAT_WRITE; 250 251 nsec = bp->bio_bcount / sc->secsize; 252 secno = bp->bio_pblkno; 253 dst = bp->bio_data; 254 while (nsec--) { 255 secpp = &sc->secp[secno]; 256 if ((uintptr_t)*secpp > 255) { 257 secp = *secpp; 258 secval = 0; 259 } else { 260 secp = NULL; 261 secval = (uintptr_t) *secpp; 262 } 263 264 if (md_debug > 2) 265 printf("%x %p %p %d\n", 266 bp->bio_flags, secpp, secp, secval); 267 268 if (bp->bio_cmd == BIO_DELETE) { 269 if (!(sc->flags & MD_RESERVE) && secp != NULL) { 270 FREE(secp, M_MDSECT); 271 *secpp = 0; 272 } 273 } else if (bp->bio_cmd == BIO_READ) { 274 if (secp != NULL) { 275 bcopy(secp, dst, sc->secsize); 276 } else if (secval) { 277 for (i = 0; i < sc->secsize; i++) 278 dst[i] = secval; 279 } else { 280 bzero(dst, sc->secsize); 281 } 282 } else { 283 if (sc->flags & MD_COMPRESS) { 284 uc = dst[0]; 285 for (i = 1; i < sc->secsize; i++) 286 if (dst[i] != uc) 287 break; 288 } else { 289 i = 0; 290 uc = 0; 291 } 292 if (i == sc->secsize) { 293 if (secp) 294 FREE(secp, M_MDSECT); 295 *secpp = (u_char *)(uintptr_t)uc; 296 } else { 297 if (secp == NULL) 298 MALLOC(secp, u_char *, sc->secsize, M_MDSECT, M_WAITOK); 299 bcopy(dst, secp, sc->secsize); 300 *secpp = secp; 301 } 302 } 303 secno++; 304 dst += sc->secsize; 305 } 306 bp->bio_resid = 0; 307 devstat_end_transaction_bio(&sc->stats, bp); 308 biodone(bp); 309 } 310 return; 311 } 312 313 314 static void 315 mdstart_preload(struct md_s *sc) 316 { 317 struct bio *bp; 318 devstat_trans_flags dop; 319 320 for (;;) { 321 /* XXX: LOCK(unique unit numbers) */ 322 bp = bioq_first(&sc->bio_queue); 323 if (bp) 324 bioq_remove(&sc->bio_queue, bp); 325 /* XXX: UNLOCK(unique unit numbers) */ 326 if (!bp) 327 break; 328 329 devstat_start_transaction(&sc->stats); 330 331 if (bp->bio_cmd == BIO_DELETE) { 332 dop = DEVSTAT_NO_DATA; 333 } else if (bp->bio_cmd == BIO_READ) { 334 dop = DEVSTAT_READ; 335 bcopy(sc->pl_ptr + (bp->bio_pblkno << DEV_BSHIFT), bp->bio_data, bp->bio_bcount); 336 } else { 337 dop = DEVSTAT_WRITE; 338 bcopy(bp->bio_data, sc->pl_ptr + (bp->bio_pblkno << DEV_BSHIFT), bp->bio_bcount); 339 } 340 bp->bio_resid = 0; 341 devstat_end_transaction_bio(&sc->stats, bp); 342 biodone(bp); 343 } 344 return; 345 } 346 347 static void 348 mdstart_vnode(struct md_s *sc) 349 { 350 int error; 351 struct bio *bp; 352 struct uio auio; 353 struct iovec aiov; 354 struct mount *mp; 355 356 /* 357 * VNODE I/O 358 * 359 * If an error occurs, we set BIO_ERROR but we do not set 360 * B_INVAL because (for a write anyway), the buffer is 361 * still valid. 362 */ 363 364 for (;;) { 365 /* XXX: LOCK(unique unit numbers) */ 366 bp = bioq_first(&sc->bio_queue); 367 if (bp) 368 bioq_remove(&sc->bio_queue, bp); 369 /* XXX: UNLOCK(unique unit numbers) */ 370 if (!bp) 371 break; 372 373 devstat_start_transaction(&sc->stats); 374 375 bzero(&auio, sizeof(auio)); 376 377 aiov.iov_base = bp->bio_data; 378 aiov.iov_len = bp->bio_bcount; 379 auio.uio_iov = &aiov; 380 auio.uio_iovcnt = 1; 381 auio.uio_offset = (vm_ooffset_t)bp->bio_pblkno * sc->secsize; 382 auio.uio_segflg = UIO_SYSSPACE; 383 if(bp->bio_cmd == BIO_READ) 384 auio.uio_rw = UIO_READ; 385 else 386 auio.uio_rw = UIO_WRITE; 387 auio.uio_resid = bp->bio_bcount; 388 auio.uio_procp = curproc; 389 if (VOP_ISLOCKED(sc->vnode, NULL)) 390 vprint("unexpected md driver lock", sc->vnode); 391 if (bp->bio_cmd == BIO_READ) { 392 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 393 error = VOP_READ(sc->vnode, &auio, 0, sc->cred); 394 } else { 395 (void) vn_start_write(sc->vnode, &mp, V_WAIT); 396 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 397 error = VOP_WRITE(sc->vnode, &auio, 0, sc->cred); 398 vn_finished_write(mp); 399 } 400 VOP_UNLOCK(sc->vnode, 0, curproc); 401 bp->bio_resid = auio.uio_resid; 402 403 if (error) { 404 bp->bio_error = error; 405 bp->bio_flags |= BIO_ERROR; 406 } 407 devstat_end_transaction_bio(&sc->stats, bp); 408 biodone(bp); 409 } 410 return; 411 } 412 413 static void 414 mdstart_swap(struct md_s *sc) 415 { 416 struct bio *bp; 417 418 for (;;) { 419 /* XXX: LOCK(unique unit numbers) */ 420 bp = bioq_first(&sc->bio_queue); 421 if (bp) 422 bioq_remove(&sc->bio_queue, bp); 423 /* XXX: UNLOCK(unique unit numbers) */ 424 if (!bp) 425 break; 426 427 #if 0 428 devstat_start_transaction(&sc->stats); 429 #endif 430 431 if ((bp->bio_cmd == BIO_DELETE) && (sc->flags & MD_RESERVE)) 432 biodone(bp); 433 else 434 vm_pager_strategy(sc->object, bp); 435 436 #if 0 437 devstat_end_transaction_bio(&sc->stats, bp); 438 #endif 439 } 440 return; 441 } 442 443 static void 444 mdstrategy(struct bio *bp) 445 { 446 struct md_s *sc; 447 448 if (md_debug > 1) 449 printf("mdstrategy(%p) %s %x, %d, %ld, %p)\n", 450 bp, devtoname(bp->bio_dev), bp->bio_flags, bp->bio_blkno, 451 bp->bio_bcount / DEV_BSIZE, bp->bio_data); 452 453 sc = bp->bio_dev->si_drv1; 454 455 /* XXX: LOCK(sc->lock) */ 456 bioqdisksort(&sc->bio_queue, bp); 457 /* XXX: UNLOCK(sc->lock) */ 458 459 if (atomic_cmpset_int(&sc->busy, 0, 1) == 0) 460 return; 461 462 switch (sc->type) { 463 case MD_MALLOC: 464 mdstart_malloc(sc); 465 break; 466 case MD_PRELOAD: 467 mdstart_preload(sc); 468 break; 469 case MD_VNODE: 470 mdstart_vnode(sc); 471 break; 472 case MD_SWAP: 473 mdstart_swap(sc); 474 break; 475 default: 476 panic("Impossible md(type)"); 477 break; 478 } 479 sc->busy = 0; 480 } 481 482 static struct md_s * 483 mdfind(int unit) 484 { 485 struct md_s *sc; 486 487 /* XXX: LOCK(unique unit numbers) */ 488 LIST_FOREACH(sc, &md_softc_list, list) { 489 if (sc->unit == unit) 490 break; 491 } 492 /* XXX: UNLOCK(unique unit numbers) */ 493 return (sc); 494 } 495 496 static struct md_s * 497 mdnew(int unit) 498 { 499 struct md_s *sc; 500 int max = -1; 501 502 /* XXX: LOCK(unique unit numbers) */ 503 LIST_FOREACH(sc, &md_softc_list, list) { 504 if (sc->unit == unit) { 505 /* XXX: UNLOCK(unique unit numbers) */ 506 return (NULL); 507 } 508 if (sc->unit > max) 509 max = sc->unit; 510 } 511 if (unit == -1) 512 unit = max + 1; 513 if (unit > DKMAXUNIT) 514 return (NULL); 515 MALLOC(sc, struct md_s *,sizeof(*sc), M_MD, M_WAITOK | M_ZERO); 516 sc->unit = unit; 517 LIST_INSERT_HEAD(&md_softc_list, sc, list); 518 /* XXX: UNLOCK(unique unit numbers) */ 519 return (sc); 520 } 521 522 static void 523 mddelete(struct md_s *sc) 524 { 525 526 devstat_remove_entry(&sc->stats); 527 /* XXX: LOCK(unique unit numbers) */ 528 LIST_REMOVE(sc, list); 529 /* XXX: UNLOCK(unique unit numbers) */ 530 FREE(sc, M_MD); 531 } 532 533 static void 534 mdinit(struct md_s *sc) 535 { 536 537 bioq_init(&sc->bio_queue); 538 devstat_add_entry(&sc->stats, MD_NAME, sc->unit, sc->secsize, 539 DEVSTAT_NO_ORDERED_TAGS, 540 DEVSTAT_TYPE_DIRECT | DEVSTAT_TYPE_IF_OTHER, 541 DEVSTAT_PRIORITY_OTHER); 542 sc->dev = disk_create(sc->unit, &sc->disk, 0, &md_cdevsw, &mddisk_cdevsw); 543 sc->dev->si_drv1 = sc; 544 } 545 546 /* 547 * XXX: we should check that the range they feed us is mapped. 548 * XXX: we should implement read-only. 549 */ 550 551 static int 552 mdcreate_preload(struct md_ioctl *mdio) 553 { 554 struct md_s *sc; 555 556 if (mdio->md_size == 0) 557 return(EINVAL); 558 if (mdio->md_options & ~(MD_AUTOUNIT)) 559 return(EINVAL); 560 if (mdio->md_options & MD_AUTOUNIT) { 561 sc = mdnew(-1); 562 if (sc == NULL) 563 return (ENOMEM); 564 mdio->md_unit = sc->unit; 565 } else { 566 sc = mdnew(mdio->md_unit); 567 if (sc == NULL) 568 return (EBUSY); 569 } 570 sc->type = MD_PRELOAD; 571 sc->secsize = DEV_BSIZE; 572 sc->nsect = mdio->md_size; 573 /* Cast to pointer size, then to pointer to avoid warning */ 574 sc->pl_ptr = (u_char *)(uintptr_t)mdio->md_base; 575 sc->pl_len = (mdio->md_size << DEV_BSHIFT); 576 mdinit(sc); 577 return (0); 578 } 579 580 581 static int 582 mdcreate_malloc(struct md_ioctl *mdio) 583 { 584 struct md_s *sc; 585 unsigned u; 586 587 if (mdio->md_size == 0) 588 return(EINVAL); 589 if (mdio->md_options & ~(MD_AUTOUNIT | MD_COMPRESS | MD_RESERVE)) 590 return(EINVAL); 591 /* Compression doesn't make sense if we have reserved space */ 592 if (mdio->md_options & MD_RESERVE) 593 mdio->md_options &= ~MD_COMPRESS; 594 if (mdio->md_options & MD_AUTOUNIT) { 595 sc = mdnew(-1); 596 if (sc == NULL) 597 return (ENOMEM); 598 mdio->md_unit = sc->unit; 599 } else { 600 sc = mdnew(mdio->md_unit); 601 if (sc == NULL) 602 return (EBUSY); 603 } 604 sc->type = MD_MALLOC; 605 sc->secsize = DEV_BSIZE; 606 sc->nsect = mdio->md_size; 607 sc->flags = mdio->md_options & MD_COMPRESS; 608 MALLOC(sc->secp, u_char **, sc->nsect * sizeof(u_char *), M_MD, M_WAITOK | M_ZERO); 609 if (mdio->md_options & MD_RESERVE) { 610 for (u = 0; u < sc->nsect; u++) 611 MALLOC(sc->secp[u], u_char *, DEV_BSIZE, M_MDSECT, M_WAITOK | M_ZERO); 612 } 613 printf("%s%d: Malloc disk\n", MD_NAME, sc->unit); 614 mdinit(sc); 615 return (0); 616 } 617 618 619 static int 620 mdsetcred(struct md_s *sc, struct ucred *cred) 621 { 622 char *tmpbuf; 623 int error = 0; 624 625 /* 626 * Set credits in our softc 627 */ 628 629 if (sc->cred) 630 crfree(sc->cred); 631 sc->cred = crdup(cred); 632 633 /* 634 * Horrible kludge to establish credentials for NFS XXX. 635 */ 636 637 if (sc->vnode) { 638 struct uio auio; 639 struct iovec aiov; 640 641 tmpbuf = malloc(sc->secsize, M_TEMP, M_WAITOK); 642 bzero(&auio, sizeof(auio)); 643 644 aiov.iov_base = tmpbuf; 645 aiov.iov_len = sc->secsize; 646 auio.uio_iov = &aiov; 647 auio.uio_iovcnt = 1; 648 auio.uio_offset = 0; 649 auio.uio_rw = UIO_READ; 650 auio.uio_segflg = UIO_SYSSPACE; 651 auio.uio_resid = aiov.iov_len; 652 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 653 error = VOP_READ(sc->vnode, &auio, 0, sc->cred); 654 VOP_UNLOCK(sc->vnode, 0, curproc); 655 free(tmpbuf, M_TEMP); 656 } 657 return (error); 658 } 659 660 static int 661 mdcreate_vnode(struct md_ioctl *mdio, struct proc *p) 662 { 663 struct md_s *sc; 664 struct vattr vattr; 665 struct nameidata nd; 666 int error, flags; 667 668 if (mdio->md_options & MD_AUTOUNIT) { 669 sc = mdnew(-1); 670 mdio->md_unit = sc->unit; 671 } else { 672 sc = mdnew(mdio->md_unit); 673 } 674 if (sc == NULL) 675 return (EBUSY); 676 677 sc->type = MD_VNODE; 678 679 flags = FREAD|FWRITE; 680 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, mdio->md_file, p); 681 error = vn_open(&nd, &flags, 0); 682 if (error) { 683 if (error != EACCES && error != EPERM && error != EROFS) 684 return (error); 685 flags &= ~FWRITE; 686 sc->flags |= MD_READONLY; 687 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, mdio->md_file, p); 688 error = vn_open(&nd, &flags, 0); 689 if (error) 690 return (error); 691 } 692 NDFREE(&nd, NDF_ONLY_PNBUF); 693 if (nd.ni_vp->v_type != VREG || 694 (error = VOP_GETATTR(nd.ni_vp, &vattr, p->p_ucred, p))) { 695 VOP_UNLOCK(nd.ni_vp, 0, p); 696 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p); 697 return (error ? error : EINVAL); 698 } 699 VOP_UNLOCK(nd.ni_vp, 0, p); 700 sc->secsize = DEV_BSIZE; 701 sc->vnode = nd.ni_vp; 702 703 /* 704 * If the size is specified, override the file attributes. 705 */ 706 if (mdio->md_size) 707 sc->nsect = mdio->md_size; 708 else 709 sc->nsect = vattr.va_size / sc->secsize; /* XXX: round up ? */ 710 error = mdsetcred(sc, p->p_ucred); 711 if (error) { 712 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p); 713 return(error); 714 } 715 mdinit(sc); 716 return (0); 717 } 718 719 static int 720 mddestroy(struct md_s *sc, struct md_ioctl *mdio, struct proc *p) 721 { 722 unsigned u; 723 724 if (sc->dev != NULL) 725 disk_destroy(sc->dev); 726 if (sc->vnode != NULL) 727 (void)vn_close(sc->vnode, sc->flags & MD_READONLY ? FREAD : (FREAD|FWRITE), sc->cred, p); 728 if (sc->cred != NULL) 729 crfree(sc->cred); 730 if (sc->object != NULL) 731 vm_pager_deallocate(sc->object); 732 if (sc->secp != NULL) { 733 for (u = 0; u < sc->nsect; u++) 734 if ((uintptr_t)sc->secp[u] > 255) 735 FREE(sc->secp[u], M_MDSECT); 736 FREE(sc->secp, M_MD); 737 } 738 mddelete(sc); 739 return (0); 740 } 741 742 static int 743 mdcreate_swap(struct md_ioctl *mdio, struct proc *p) 744 { 745 int error; 746 struct md_s *sc; 747 748 if (mdio->md_options & MD_AUTOUNIT) { 749 sc = mdnew(-1); 750 mdio->md_unit = sc->unit; 751 } else { 752 sc = mdnew(mdio->md_unit); 753 } 754 if (sc == NULL) 755 return (EBUSY); 756 757 sc->type = MD_SWAP; 758 759 /* 760 * Range check. Disallow negative sizes or any size less then the 761 * size of a page. Then round to a page. 762 */ 763 764 if (mdio->md_size == 0) 765 return(EDOM); 766 767 /* 768 * Allocate an OBJT_SWAP object. 769 * 770 * sc_secsize is PAGE_SIZE'd 771 * 772 * mdio->size is in DEV_BSIZE'd chunks. 773 * Note the truncation. 774 */ 775 776 sc->secsize = PAGE_SIZE; 777 sc->nsect = mdio->md_size / (PAGE_SIZE / DEV_BSIZE); 778 sc->object = vm_pager_allocate(OBJT_SWAP, NULL, sc->secsize * (vm_offset_t)sc->nsect, VM_PROT_DEFAULT, 0); 779 if (mdio->md_options & MD_RESERVE) { 780 if (swap_pager_reserve(sc->object, 0, sc->nsect) < 0) { 781 vm_pager_deallocate(sc->object); 782 sc->object = NULL; 783 return(EDOM); 784 } 785 } 786 error = mdsetcred(sc, p->p_ucred); 787 if (error) 788 mddestroy(sc, mdio, p); 789 else 790 mdinit(sc); 791 return(error); 792 } 793 794 static int 795 mdctlioctl(dev_t dev, u_long cmd, caddr_t addr, int flags, struct proc *p) 796 { 797 struct md_ioctl *mdio; 798 struct md_s *sc; 799 800 if (md_debug) 801 printf("mdctlioctl(%s %lx %p %x %p)\n", 802 devtoname(dev), cmd, addr, flags, p); 803 804 mdio = (struct md_ioctl *)addr; 805 switch (cmd) { 806 case MDIOCATTACH: 807 switch (mdio->md_type) { 808 case MD_MALLOC: 809 return(mdcreate_malloc(mdio)); 810 case MD_PRELOAD: 811 return(mdcreate_preload(mdio)); 812 case MD_VNODE: 813 return(mdcreate_vnode(mdio, p)); 814 case MD_SWAP: 815 return(mdcreate_swap(mdio, p)); 816 default: 817 return (EINVAL); 818 } 819 case MDIOCDETACH: 820 if (mdio->md_file != NULL) 821 return(EINVAL); 822 if (mdio->md_size != 0) 823 return(EINVAL); 824 if (mdio->md_options != 0) 825 return(EINVAL); 826 sc = mdfind(mdio->md_unit); 827 if (sc == NULL) 828 return (ENOENT); 829 switch(sc->type) { 830 case MD_VNODE: 831 case MD_SWAP: 832 case MD_MALLOC: 833 case MD_PRELOAD: 834 return(mddestroy(sc, mdio, p)); 835 default: 836 return (EOPNOTSUPP); 837 } 838 case MDIOCQUERY: 839 sc = mdfind(mdio->md_unit); 840 if (sc == NULL) 841 return (ENOENT); 842 mdio->md_type = sc->type; 843 mdio->md_options = sc->flags; 844 switch (sc->type) { 845 case MD_MALLOC: 846 mdio->md_size = sc->nsect; 847 break; 848 case MD_PRELOAD: 849 mdio->md_size = sc->nsect; 850 (u_char *)(uintptr_t)mdio->md_base = sc->pl_ptr; 851 break; 852 case MD_SWAP: 853 mdio->md_size = sc->nsect * (PAGE_SIZE / DEV_BSIZE); 854 break; 855 case MD_VNODE: 856 mdio->md_size = sc->nsect; 857 /* XXX fill this in */ 858 mdio->md_file = NULL; 859 break; 860 } 861 return (0); 862 default: 863 return (ENOIOCTL); 864 }; 865 return (ENOIOCTL); 866 } 867 868 static void 869 md_preloaded(u_char *image, unsigned length) 870 { 871 struct md_s *sc; 872 873 sc = mdnew(-1); 874 if (sc == NULL) 875 return; 876 sc->type = MD_PRELOAD; 877 sc->secsize = DEV_BSIZE; 878 sc->nsect = length / DEV_BSIZE; 879 sc->pl_ptr = image; 880 sc->pl_len = length; 881 if (sc->unit == 0) 882 mdrootready = 1; 883 mdinit(sc); 884 } 885 886 static void 887 md_drvinit(void *unused) 888 { 889 890 caddr_t mod; 891 caddr_t c; 892 u_char *ptr, *name, *type; 893 unsigned len; 894 895 #ifdef MD_ROOT_SIZE 896 md_preloaded(mfs_root, MD_ROOT_SIZE*1024); 897 #endif 898 mod = NULL; 899 while ((mod = preload_search_next_name(mod)) != NULL) { 900 name = (char *)preload_search_info(mod, MODINFO_NAME); 901 type = (char *)preload_search_info(mod, MODINFO_TYPE); 902 if (name == NULL) 903 continue; 904 if (type == NULL) 905 continue; 906 if (strcmp(type, "md_image") && strcmp(type, "mfs_root")) 907 continue; 908 c = preload_search_info(mod, MODINFO_ADDR); 909 ptr = *(u_char **)c; 910 c = preload_search_info(mod, MODINFO_SIZE); 911 len = *(unsigned *)c; 912 printf("md%d: Preloaded image <%s> %d bytes at %p\n", 913 mdunits, name, len, ptr); 914 md_preloaded(ptr, len); 915 } 916 status_dev = make_dev(&mdctl_cdevsw, 0xffff00ff, UID_ROOT, GID_WHEEL, 0600, "mdctl"); 917 } 918 919 static int 920 md_modevent(module_t mod, int type, void *data) 921 { 922 switch (type) { 923 case MOD_LOAD: 924 md_drvinit(NULL); 925 break; 926 case MOD_UNLOAD: 927 if (!LIST_EMPTY(&md_softc_list)) 928 return EBUSY; 929 if (status_dev) 930 destroy_dev(status_dev); 931 status_dev = 0; 932 break; 933 default: 934 break; 935 } 936 return 0; 937 } 938 939 static moduledata_t md_mod = { 940 "md", 941 md_modevent, 942 NULL 943 }; 944 DECLARE_MODULE(md, md_mod, SI_SUB_DRIVERS, SI_ORDER_MIDDLE+CDEV_MAJOR); 945 MODULE_VERSION(md, MD_MODVER); 946 947 948 #ifdef MD_ROOT 949 static void 950 md_takeroot(void *junk) 951 { 952 if (mdrootready) 953 rootdevnames[0] = "ufs:/dev/md0c"; 954 } 955 956 SYSINIT(md_root, SI_SUB_MOUNT_ROOT, SI_ORDER_FIRST, md_takeroot, NULL); 957 #endif 958 959