md.c revision 70595
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: head/sys/dev/md/md.c 70595 2001-01-02 09:42:47Z phk $ 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/kernel.h> 70#include <sys/malloc.h> 71#include <sys/sysctl.h> 72#include <sys/linker.h> 73#include <sys/queue.h> 74#include <sys/mdioctl.h> 75#include <sys/vnode.h> 76#include <sys/namei.h> 77#include <sys/fcntl.h> 78#include <sys/proc.h> 79#include <machine/atomic.h> 80 81#include <vm/vm.h> 82#include <vm/vm_object.h> 83#include <vm/vm_page.h> 84#include <vm/vm_pager.h> 85#include <vm/vm_zone.h> 86#include <vm/swap_pager.h> 87 88#ifndef MD_NSECT 89#define MD_NSECT (10000 * 2) 90#endif 91 92MALLOC_DEFINE(M_MD, "MD disk", "Memory Disk"); 93MALLOC_DEFINE(M_MDSECT, "MD sectors", "Memory Disk Sectors"); 94 95static int md_debug; 96SYSCTL_INT(_debug, OID_AUTO, mddebug, CTLFLAG_RW, &md_debug, 0, ""); 97 98#if defined(MFS_ROOT) && !defined(MD_ROOT) 99#define MD_ROOT MFS_ROOT 100#warning "option MFS_ROOT has been superceeded by MD_ROOT" 101#endif 102 103#if defined(MFS_ROOT_SIZE) && !defined(MD_ROOT_SIZE) 104#define MD_ROOT_SIZE MFS_ROOT_SIZE 105#warning "option MFS_ROOT_SIZE has been superceeded by MD_ROOT_SIZE" 106#endif 107 108#if defined(MD_ROOT) && defined(MD_ROOT_SIZE) 109/* Image gets put here: */ 110static u_char mfs_root[MD_ROOT_SIZE*1024] = "MFS Filesystem goes here"; 111static u_char end_mfs_root[] __unused = "MFS Filesystem had better STOP here"; 112#endif 113 114static int mdrootready; 115static int mdunits; 116 117#define CDEV_MAJOR 95 118 119static d_strategy_t mdstrategy; 120static d_open_t mdopen; 121static d_ioctl_t mdioctl, mdctlioctl; 122 123static struct cdevsw md_cdevsw = { 124 /* open */ mdopen, 125 /* close */ nullclose, 126 /* read */ physread, 127 /* write */ physwrite, 128 /* ioctl */ mdioctl, 129 /* poll */ nopoll, 130 /* mmap */ nommap, 131 /* strategy */ mdstrategy, 132 /* name */ "md", 133 /* maj */ CDEV_MAJOR, 134 /* dump */ nodump, 135 /* psize */ nopsize, 136 /* flags */ D_DISK | D_CANFREE | D_MEMDISK, 137}; 138 139static struct cdevsw mdctl_cdevsw = { 140 /* open */ nullopen, 141 /* close */ nullclose, 142 /* read */ noread, 143 /* write */ nowrite, 144 /* ioctl */ mdctlioctl, 145 /* poll */ nopoll, 146 /* mmap */ nommap, 147 /* strategy */ nostrategy, 148 /* name */ "md", 149 /* maj */ CDEV_MAJOR 150}; 151 152static struct cdevsw mddisk_cdevsw; 153 154static LIST_HEAD(, md_s) md_softc_list = LIST_HEAD_INITIALIZER(&md_softc_list); 155 156struct md_s { 157 int unit; 158 LIST_ENTRY(md_s) list; 159 struct devstat stats; 160 struct bio_queue_head bio_queue; 161 struct disk disk; 162 dev_t dev; 163 int busy; 164 enum md_types type; 165 unsigned nsect; 166 unsigned secsize; 167 unsigned flags; 168 169 /* MD_MALLOC related fields */ 170 unsigned nsecp; 171 u_char **secp; 172 173 /* MD_PRELOAD related fields */ 174 u_char *pl_ptr; 175 unsigned pl_len; 176 177 /* MD_VNODE related fields */ 178 struct vnode *vnode; 179 struct ucred *cred; 180 181 /* MD_OBJET related fields */ 182 vm_object_t object; 183}; 184 185static int 186mdopen(dev_t dev, int flag, int fmt, struct proc *p) 187{ 188 struct md_s *sc; 189 struct disklabel *dl; 190 191 if (md_debug) 192 printf("mdopen(%s %x %x %p)\n", 193 devtoname(dev), flag, fmt, p); 194 195 sc = dev->si_drv1; 196 197 dl = &sc->disk.d_label; 198 bzero(dl, sizeof(*dl)); 199 dl->d_secsize = sc->secsize; 200 dl->d_nsectors = sc->nsect > 1024 ? 1024 : sc->nsect; 201 dl->d_ntracks = 1; 202 dl->d_secpercyl = dl->d_nsectors * dl->d_ntracks; 203 dl->d_secperunit = sc->nsect; 204 dl->d_ncylinders = dl->d_secperunit / dl->d_secpercyl; 205 return (0); 206} 207 208static int 209mdioctl(dev_t dev, u_long cmd, caddr_t addr, int flags, struct proc *p) 210{ 211 212 if (md_debug) 213 printf("mdioctl(%s %lx %p %x %p)\n", 214 devtoname(dev), cmd, addr, flags, p); 215 216 return (ENOIOCTL); 217} 218 219static void 220mdstart_malloc(struct md_s *sc) 221{ 222 int i; 223 struct bio *bp; 224 devstat_trans_flags dop; 225 u_char *secp, **secpp, *dst; 226 unsigned secno, nsec, secval, uc; 227 228 for (;;) { 229 /* XXX: LOCK(unique unit numbers) */ 230 bp = bioq_first(&sc->bio_queue); 231 if (bp) 232 bioq_remove(&sc->bio_queue, bp); 233 /* XXX: UNLOCK(unique unit numbers) */ 234 if (!bp) 235 break; 236 237 devstat_start_transaction(&sc->stats); 238 239 if (bp->bio_cmd == BIO_DELETE) 240 dop = DEVSTAT_NO_DATA; 241 else if (bp->bio_cmd == BIO_READ) 242 dop = DEVSTAT_READ; 243 else 244 dop = DEVSTAT_WRITE; 245 246 nsec = bp->bio_bcount / DEV_BSIZE; 247 secno = bp->bio_pblkno; 248 dst = bp->bio_data; 249 while (nsec--) { 250 if (secno < sc->nsecp) { 251 secpp = &sc->secp[secno]; 252 if ((uintptr_t)*secpp > 255) { 253 secp = *secpp; 254 secval = 0; 255 } else { 256 secp = 0; 257 secval = (uintptr_t) *secpp; 258 } 259 } else { 260 secpp = 0; 261 secp = 0; 262 secval = 0; 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 (secpp && !(sc->flags & MD_RESERVE)) { 270 if (secp) 271 FREE(secp, M_MDSECT); 272 *secpp = 0; 273 } 274 } else if (bp->bio_cmd == BIO_READ) { 275 if (secp) { 276 bcopy(secp, dst, DEV_BSIZE); 277 } else if (secval) { 278 for (i = 0; i < DEV_BSIZE; i++) 279 dst[i] = secval; 280 } else { 281 bzero(dst, DEV_BSIZE); 282 } 283 } else { 284 if (sc->flags & MD_COMPRESS) { 285 uc = dst[0]; 286 for (i = 1; i < DEV_BSIZE; i++) 287 if (dst[i] != uc) 288 break; 289 } else { 290 i = 0; 291 uc = 0; 292 } 293 if (i == DEV_BSIZE && !uc) { 294 if (secp) 295 FREE(secp, M_MDSECT); 296 if (secpp) 297 *secpp = (u_char *)(uintptr_t)uc; 298 } else { 299 if (!secpp) { 300 MALLOC(secpp, u_char **, (secno + nsec + 1) * sizeof(u_char *), M_MD, M_WAITOK | M_ZERO); 301 bcopy(sc->secp, secpp, sc->nsecp * sizeof(u_char *)); 302 FREE(sc->secp, M_MD); 303 sc->secp = secpp; 304 sc->nsecp = secno + nsec + 1; 305 secpp = &sc->secp[secno]; 306 } 307 if (i == DEV_BSIZE) { 308 if (secp) 309 FREE(secp, M_MDSECT); 310 *secpp = (u_char *)(uintptr_t)uc; 311 } else { 312 if (!secp) 313 MALLOC(secp, u_char *, DEV_BSIZE, M_MDSECT, M_WAITOK); 314 bcopy(dst, secp, DEV_BSIZE); 315 *secpp = secp; 316 } 317 } 318 } 319 secno++; 320 dst += DEV_BSIZE; 321 } 322 bp->bio_resid = 0; 323 devstat_end_transaction_bio(&sc->stats, bp); 324 biodone(bp); 325 } 326 return; 327} 328 329 330static void 331mdstart_preload(struct md_s *sc) 332{ 333 struct bio *bp; 334 devstat_trans_flags dop; 335 336 for (;;) { 337 /* XXX: LOCK(unique unit numbers) */ 338 bp = bioq_first(&sc->bio_queue); 339 if (bp) 340 bioq_remove(&sc->bio_queue, bp); 341 /* XXX: UNLOCK(unique unit numbers) */ 342 if (!bp) 343 break; 344 345 devstat_start_transaction(&sc->stats); 346 347 if (bp->bio_cmd == BIO_DELETE) { 348 dop = DEVSTAT_NO_DATA; 349 } else if (bp->bio_cmd == BIO_READ) { 350 dop = DEVSTAT_READ; 351 bcopy(sc->pl_ptr + (bp->bio_pblkno << DEV_BSHIFT), bp->bio_data, bp->bio_bcount); 352 } else { 353 dop = DEVSTAT_WRITE; 354 bcopy(bp->bio_data, sc->pl_ptr + (bp->bio_pblkno << DEV_BSHIFT), bp->bio_bcount); 355 } 356 bp->bio_resid = 0; 357 devstat_end_transaction_bio(&sc->stats, bp); 358 biodone(bp); 359 } 360 return; 361} 362 363static void 364mdstart_vnode(struct md_s *sc) 365{ 366 int error; 367 struct bio *bp; 368 struct uio auio; 369 struct iovec aiov; 370 struct mount *mp; 371 372 /* 373 * VNODE I/O 374 * 375 * If an error occurs, we set BIO_ERROR but we do not set 376 * B_INVAL because (for a write anyway), the buffer is 377 * still valid. 378 */ 379 380 for (;;) { 381 /* XXX: LOCK(unique unit numbers) */ 382 bp = bioq_first(&sc->bio_queue); 383 if (bp) 384 bioq_remove(&sc->bio_queue, bp); 385 /* XXX: UNLOCK(unique unit numbers) */ 386 if (!bp) 387 break; 388 389 devstat_start_transaction(&sc->stats); 390 391 bzero(&auio, sizeof(auio)); 392 393 aiov.iov_base = bp->bio_data; 394 aiov.iov_len = bp->bio_bcount; 395 auio.uio_iov = &aiov; 396 auio.uio_iovcnt = 1; 397 auio.uio_offset = (vm_ooffset_t)bp->bio_pblkno * sc->secsize; 398 auio.uio_segflg = UIO_SYSSPACE; 399 if(bp->bio_cmd == BIO_READ) 400 auio.uio_rw = UIO_READ; 401 else 402 auio.uio_rw = UIO_WRITE; 403 auio.uio_resid = bp->bio_bcount; 404 auio.uio_procp = curproc; 405 if (VOP_ISLOCKED(sc->vnode, NULL)) 406 vprint("unexpected vn driver lock", sc->vnode); 407 if (bp->bio_cmd == BIO_READ) { 408 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 409 error = VOP_READ(sc->vnode, &auio, 0, sc->cred); 410 } else { 411 (void) vn_start_write(sc->vnode, &mp, V_WAIT); 412 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 413 error = VOP_WRITE(sc->vnode, &auio, 0, sc->cred); 414 vn_finished_write(mp); 415 } 416 VOP_UNLOCK(sc->vnode, 0, curproc); 417 bp->bio_resid = auio.uio_resid; 418 419 if (error) { 420 bp->bio_error = error; 421 bp->bio_flags |= BIO_ERROR; 422 } 423 devstat_end_transaction_bio(&sc->stats, bp); 424 biodone(bp); 425 } 426 return; 427} 428 429static void 430mdstart_swap(struct md_s *sc) 431{ 432 struct bio *bp; 433 434 for (;;) { 435 /* XXX: LOCK(unique unit numbers) */ 436 bp = bioq_first(&sc->bio_queue); 437 if (bp) 438 bioq_remove(&sc->bio_queue, bp); 439 /* XXX: UNLOCK(unique unit numbers) */ 440 if (!bp) 441 break; 442 443#if 0 444 devstat_start_transaction(&sc->stats); 445#endif 446 447 if ((bp->bio_cmd == BIO_DELETE) && (sc->flags & MD_RESERVE)) 448 biodone(bp); 449 else 450 vm_pager_strategy(sc->object, bp); 451 452#if 0 453 devstat_end_transaction_bio(&sc->stats, bp); 454#endif 455 } 456 return; 457} 458 459static void 460mdstrategy(struct bio *bp) 461{ 462 struct md_s *sc; 463 464 if (md_debug > 1) 465 printf("mdstrategy(%p) %s %x, %d, %ld, %p)\n", 466 bp, devtoname(bp->bio_dev), bp->bio_flags, bp->bio_blkno, 467 bp->bio_bcount / DEV_BSIZE, bp->bio_data); 468 469 sc = bp->bio_dev->si_drv1; 470 471 /* XXX: LOCK(sc->lock) */ 472 bioqdisksort(&sc->bio_queue, bp); 473 /* XXX: UNLOCK(sc->lock) */ 474 475 if (atomic_cmpset_int(&sc->busy, 0, 1) == 0) 476 return; 477 478 switch (sc->type) { 479 case MD_MALLOC: 480 mdstart_malloc(sc); 481 break; 482 case MD_PRELOAD: 483 mdstart_preload(sc); 484 break; 485 case MD_VNODE: 486 mdstart_vnode(sc); 487 break; 488 case MD_SWAP: 489 mdstart_swap(sc); 490 break; 491 default: 492 panic("Impossible md(type)"); 493 break; 494 } 495 sc->busy = 0; 496} 497 498static struct md_s * 499mdfind(int unit) 500{ 501 struct md_s *sc; 502 503 /* XXX: LOCK(unique unit numbers) */ 504 LIST_FOREACH(sc, &md_softc_list, list) { 505 if (sc->unit == unit) 506 break; 507 } 508 /* XXX: UNLOCK(unique unit numbers) */ 509 return (sc); 510} 511 512static struct md_s * 513mdnew(int unit) 514{ 515 struct md_s *sc; 516 int max = -1; 517 518 /* XXX: LOCK(unique unit numbers) */ 519 LIST_FOREACH(sc, &md_softc_list, list) { 520 if (sc->unit == unit) { 521 /* XXX: UNLOCK(unique unit numbers) */ 522 return (NULL); 523 } 524 if (sc->unit > max) 525 max = sc->unit; 526 } 527 if (unit == -1) 528 unit = max + 1; 529 if (unit > DKMAXUNIT) 530 return (NULL); 531 MALLOC(sc, struct md_s *,sizeof(*sc), M_MD, M_WAITOK | M_ZERO); 532 sc->unit = unit; 533 LIST_INSERT_HEAD(&md_softc_list, sc, list); 534 /* XXX: UNLOCK(unique unit numbers) */ 535 return (sc); 536} 537 538static void 539mddelete(struct md_s *sc) 540{ 541 542 /* XXX: LOCK(unique unit numbers) */ 543 LIST_REMOVE(sc, list); 544 /* XXX: UNLOCK(unique unit numbers) */ 545 FREE(sc, M_MD); 546} 547 548static void 549mdinit(struct md_s *sc) 550{ 551 552 bioq_init(&sc->bio_queue); 553 devstat_add_entry(&sc->stats, "md", sc->unit, DEV_BSIZE, 554 DEVSTAT_NO_ORDERED_TAGS, 555 DEVSTAT_TYPE_DIRECT | DEVSTAT_TYPE_IF_OTHER, 556 DEVSTAT_PRIORITY_OTHER); 557 sc->dev = disk_create(sc->unit, &sc->disk, 0, &md_cdevsw, &mddisk_cdevsw); 558 sc->dev->si_drv1 = sc; 559} 560 561static int 562mdcreate_preload(struct md_ioctl *mdio) 563{ 564 struct md_s *sc; 565 566 if (mdio->md_size == 0) 567 return(EINVAL); 568 if (mdio->md_options & ~(MD_AUTOUNIT)) 569 return(EINVAL); 570 if (mdio->md_options & MD_AUTOUNIT) { 571 sc = mdnew(-1); 572 if (sc == NULL) 573 return (ENOMEM); 574 mdio->md_unit = sc->unit; 575 } else { 576 sc = mdnew(mdio->md_unit); 577 if (sc == NULL) 578 return (EBUSY); 579 } 580 sc->type = MD_PRELOAD; 581 sc->secsize = DEV_BSIZE; 582 sc->nsect = mdio->md_size; 583 sc->pl_ptr = (u_char *)mdio->md_base; 584 sc->pl_len = (mdio->md_size << DEV_BSHIFT); 585 mdinit(sc); 586 return (0); 587} 588 589 590static int 591mdcreate_malloc(struct md_ioctl *mdio) 592{ 593 struct md_s *sc; 594 unsigned u; 595 596 if (mdio->md_size == 0) 597 return(EINVAL); 598 if (mdio->md_options & ~(MD_AUTOUNIT | MD_COMPRESS | MD_RESERVE)) 599 return(EINVAL); 600 /* Compression doesn't make sense if we have reserved space */ 601 if (mdio->md_options & MD_RESERVE) 602 mdio->md_options &= ~MD_COMPRESS; 603 if (mdio->md_options & MD_AUTOUNIT) { 604 sc = mdnew(-1); 605 if (sc == NULL) 606 return (ENOMEM); 607 mdio->md_unit = sc->unit; 608 } else { 609 sc = mdnew(mdio->md_unit); 610 if (sc == NULL) 611 return (EBUSY); 612 } 613 sc->type = MD_MALLOC; 614 sc->secsize = DEV_BSIZE; 615 sc->nsect = mdio->md_size; 616 sc->flags = mdio->md_options & MD_COMPRESS; 617 if (!(mdio->md_options & MD_RESERVE)) { 618 MALLOC(sc->secp, u_char **, sizeof(u_char *), M_MD, M_WAITOK | M_ZERO); 619 sc->nsecp = 1; 620 } else { 621 MALLOC(sc->secp, u_char **, sc->nsect * sizeof(u_char *), M_MD, M_WAITOK | M_ZERO); 622 sc->nsecp = sc->nsect; 623 for (u = 0; u < sc->nsect; u++) 624 MALLOC(sc->secp[u], u_char *, DEV_BSIZE, M_MDSECT, M_WAITOK | M_ZERO); 625 } 626 printf("md%d: Malloc disk\n", sc->unit); 627 mdinit(sc); 628 return (0); 629} 630 631 632static int 633mdsetcred(struct md_s *sc, struct ucred *cred) 634{ 635 char *tmpbuf; 636 int error = 0; 637 638 /* 639 * Set credits in our softc 640 */ 641 642 if (sc->cred) 643 crfree(sc->cred); 644 sc->cred = crdup(cred); 645 646 /* 647 * Horrible kludge to establish credentials for NFS XXX. 648 */ 649 650 if (sc->vnode) { 651 struct uio auio; 652 struct iovec aiov; 653 654 tmpbuf = malloc(sc->secsize, M_TEMP, M_WAITOK); 655 bzero(&auio, sizeof(auio)); 656 657 aiov.iov_base = tmpbuf; 658 aiov.iov_len = sc->secsize; 659 auio.uio_iov = &aiov; 660 auio.uio_iovcnt = 1; 661 auio.uio_offset = 0; 662 auio.uio_rw = UIO_READ; 663 auio.uio_segflg = UIO_SYSSPACE; 664 auio.uio_resid = aiov.iov_len; 665 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 666 error = VOP_READ(sc->vnode, &auio, 0, sc->cred); 667 VOP_UNLOCK(sc->vnode, 0, curproc); 668 free(tmpbuf, M_TEMP); 669 } 670 return (error); 671} 672 673static int 674mdcreate_vnode(struct md_ioctl *mdio, struct proc *p) 675{ 676 struct md_s *sc; 677 struct vattr vattr; 678 struct nameidata nd; 679 int error, flags; 680 681 if (mdio->md_options & MD_AUTOUNIT) { 682 sc = mdnew(-1); 683 mdio->md_unit = sc->unit; 684 } else { 685 sc = mdnew(mdio->md_unit); 686 } 687 if (sc == NULL) 688 return (EBUSY); 689 690 sc->type = MD_VNODE; 691 692 flags = FREAD|FWRITE; 693 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, mdio->md_file, p); 694 error = vn_open(&nd, &flags, 0); 695 if (error) { 696 if (error != EACCES && error != EPERM && error != EROFS) 697 return (error); 698 flags &= ~FWRITE; 699 sc->flags |= MD_READONLY; 700 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, mdio->md_file, p); 701 error = vn_open(&nd, &flags, 0); 702 if (error) 703 return (error); 704 } 705 NDFREE(&nd, NDF_ONLY_PNBUF); 706 if (nd.ni_vp->v_type != VREG || 707 (error = VOP_GETATTR(nd.ni_vp, &vattr, p->p_ucred, p))) { 708 VOP_UNLOCK(nd.ni_vp, 0, p); 709 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p); 710 return (error ? error : EINVAL); 711 } 712 VOP_UNLOCK(nd.ni_vp, 0, p); 713 sc->secsize = DEV_BSIZE; 714 sc->vnode = nd.ni_vp; 715 716 /* 717 * If the size is specified, override the file attributes. 718 */ 719 if (mdio->md_size) 720 sc->nsect = mdio->md_size; 721 else 722 sc->nsect = vattr.va_size / sc->secsize; /* XXX: round up ? */ 723 error = mdsetcred(sc, p->p_ucred); 724 if (error) { 725 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p); 726 return(error); 727 } 728 mdinit(sc); 729 return (0); 730} 731 732static int 733mddestroy(struct md_s *sc, struct md_ioctl *mdio, struct proc *p) 734{ 735 unsigned u; 736 737 if (sc->dev != NULL) 738 disk_destroy(sc->dev); 739 if (sc->vnode != NULL) 740 (void)vn_close(sc->vnode, sc->flags & MD_READONLY ? FREAD : (FREAD|FWRITE), sc->cred, p); 741 if (sc->cred != NULL) 742 crfree(sc->cred); 743 if (sc->object != NULL) 744 vm_pager_deallocate(sc->object); 745 if (sc->secp != NULL) { 746 for (u = 0; u < sc->nsecp; u++) 747 if ((uintptr_t)sc->secp[u] > 255) 748 FREE(sc->secp[u], M_MDSECT); 749 FREE(sc->secp, M_MD); 750 } 751 mddelete(sc); 752 return (0); 753} 754 755static int 756mdcreate_swap(struct md_ioctl *mdio, struct proc *p) 757{ 758 int error; 759 struct md_s *sc; 760 761 if (mdio->md_options & MD_AUTOUNIT) { 762 sc = mdnew(-1); 763 mdio->md_unit = sc->unit; 764 } else { 765 sc = mdnew(mdio->md_unit); 766 } 767 if (sc == NULL) 768 return (EBUSY); 769 770 sc->type = MD_SWAP; 771 772 /* 773 * Range check. Disallow negative sizes or any size less then the 774 * size of a page. Then round to a page. 775 */ 776 777 if (mdio->md_size == 0) 778 return(EDOM); 779 780 /* 781 * Allocate an OBJT_SWAP object. 782 * 783 * sc_secsize is PAGE_SIZE'd 784 * 785 * mdio->size is in DEV_BSIZE'd chunks. 786 * Note the truncation. 787 */ 788 789 sc->secsize = PAGE_SIZE; 790 sc->nsect = mdio->md_size / (PAGE_SIZE / DEV_BSIZE); 791 sc->object = vm_pager_allocate(OBJT_SWAP, NULL, sc->secsize * (vm_offset_t)sc->nsect, VM_PROT_DEFAULT, 0); 792 if (mdio->md_options & MD_RESERVE) { 793 if (swap_pager_reserve(sc->object, 0, sc->nsect) < 0) { 794 vm_pager_deallocate(sc->object); 795 sc->object = NULL; 796 return(EDOM); 797 } 798 } 799 error = mdsetcred(sc, p->p_ucred); 800 if (error) 801 mddestroy(sc, mdio, p); 802 else 803 mdinit(sc); 804 return(error); 805} 806 807static int 808mdctlioctl(dev_t dev, u_long cmd, caddr_t addr, int flags, struct proc *p) 809{ 810 struct md_ioctl *mdio; 811 struct md_s *sc; 812 813 if (md_debug) 814 printf("mdctlioctl(%s %lx %p %x %p)\n", 815 devtoname(dev), cmd, addr, flags, p); 816 817 mdio = (struct md_ioctl *)addr; 818 switch (cmd) { 819 case MDIOCATTACH: 820 switch (mdio->md_type) { 821 case MD_MALLOC: 822 return(mdcreate_malloc(mdio)); 823 case MD_PRELOAD: 824 return(mdcreate_preload(mdio)); 825 case MD_VNODE: 826 return(mdcreate_vnode(mdio, p)); 827 case MD_SWAP: 828 return(mdcreate_swap(mdio, p)); 829 default: 830 return (EINVAL); 831 } 832 case MDIOCDETACH: 833 if (mdio->md_file != NULL) 834 return(EINVAL); 835 if (mdio->md_size != 0) 836 return(EINVAL); 837 if (mdio->md_options != 0) 838 return(EINVAL); 839 sc = mdfind(mdio->md_unit); 840 if (sc == NULL) 841 return (ENOENT); 842 switch(sc->type) { 843 case MD_VNODE: 844 case MD_SWAP: 845 case MD_MALLOC: 846 case MD_PRELOAD: 847 return(mddestroy(sc, mdio, p)); 848 default: 849 return (EOPNOTSUPP); 850 } 851 default: 852 return (ENOIOCTL); 853 }; 854 return (ENOIOCTL); 855} 856 857static void 858md_preloaded(u_char *image, unsigned length) 859{ 860 struct md_s *sc; 861 862 sc = mdnew(-1); 863 if (sc == NULL) 864 return; 865 sc->type = MD_PRELOAD; 866 sc->secsize = DEV_BSIZE; 867 sc->nsect = length / DEV_BSIZE; 868 sc->pl_ptr = image; 869 sc->pl_len = length; 870 if (sc->unit == 0) 871 mdrootready = 1; 872 mdinit(sc); 873} 874 875static void 876md_drvinit(void *unused) 877{ 878 879 caddr_t mod; 880 caddr_t c; 881 u_char *ptr, *name, *type; 882 unsigned len; 883 884#ifdef MD_ROOT_SIZE 885 md_preloaded(mfs_root, MD_ROOT_SIZE*1024); 886#endif 887 mod = NULL; 888 while ((mod = preload_search_next_name(mod)) != NULL) { 889 name = (char *)preload_search_info(mod, MODINFO_NAME); 890 type = (char *)preload_search_info(mod, MODINFO_TYPE); 891 if (name == NULL) 892 continue; 893 if (type == NULL) 894 continue; 895 if (strcmp(type, "md_image") && strcmp(type, "mfs_root")) 896 continue; 897 c = preload_search_info(mod, MODINFO_ADDR); 898 ptr = *(u_char **)c; 899 c = preload_search_info(mod, MODINFO_SIZE); 900 len = *(unsigned *)c; 901 printf("md%d: Preloaded image <%s> %d bytes at %p\n", 902 mdunits, name, len, ptr); 903 md_preloaded(ptr, len); 904 } 905 make_dev(&mdctl_cdevsw, 0xffff00ff, UID_ROOT, GID_WHEEL, 0600, "mdctl"); 906} 907 908SYSINIT(mddev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR, md_drvinit,NULL) 909 910#ifdef MD_ROOT 911static void 912md_takeroot(void *junk) 913{ 914 if (mdrootready) 915 rootdevnames[0] = "ufs:/dev/md0c"; 916} 917 918SYSINIT(md_root, SI_SUB_MOUNT_ROOT, SI_ORDER_FIRST, md_takeroot, NULL); 919#endif 920 921