md.c revision 74521
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 74521 2001-03-20 12:31:53Z 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#define MD_MODVER 1 89 90#ifndef MD_NSECT 91#define MD_NSECT (10000 * 2) 92#endif 93 94MALLOC_DEFINE(M_MD, "MD disk", "Memory Disk"); 95MALLOC_DEFINE(M_MDSECT, "MD sectors", "Memory Disk Sectors"); 96 97static int md_debug; 98SYSCTL_INT(_debug, OID_AUTO, mddebug, CTLFLAG_RW, &md_debug, 0, ""); 99 100#if defined(MFS_ROOT) && !defined(MD_ROOT) 101#define MD_ROOT MFS_ROOT 102#warning "option MFS_ROOT has been superceeded by MD_ROOT" 103#endif 104 105#if defined(MFS_ROOT_SIZE) && !defined(MD_ROOT_SIZE) 106#define MD_ROOT_SIZE MFS_ROOT_SIZE 107#warning "option MFS_ROOT_SIZE has been superceeded by MD_ROOT_SIZE" 108#endif 109 110#if defined(MD_ROOT) && defined(MD_ROOT_SIZE) 111/* Image gets put here: */ 112static u_char mfs_root[MD_ROOT_SIZE*1024] = "MFS Filesystem goes here"; 113static u_char end_mfs_root[] __unused = "MFS Filesystem had better STOP here"; 114#endif 115 116static int mdrootready; 117static int mdunits; 118static dev_t status_dev = 0; 119 120 121#define CDEV_MAJOR 95 122 123static d_strategy_t mdstrategy; 124static d_open_t mdopen; 125static d_ioctl_t mdioctl, mdctlioctl; 126 127static struct cdevsw md_cdevsw = { 128 /* open */ mdopen, 129 /* close */ nullclose, 130 /* read */ physread, 131 /* write */ physwrite, 132 /* ioctl */ mdioctl, 133 /* poll */ nopoll, 134 /* mmap */ nommap, 135 /* strategy */ mdstrategy, 136 /* name */ MD_NAME, 137 /* maj */ CDEV_MAJOR, 138 /* dump */ nodump, 139 /* psize */ nopsize, 140 /* flags */ D_DISK | D_CANFREE | D_MEMDISK, 141}; 142 143static struct cdevsw mdctl_cdevsw = { 144 /* open */ nullopen, 145 /* close */ nullclose, 146 /* read */ noread, 147 /* write */ nowrite, 148 /* ioctl */ mdctlioctl, 149 /* poll */ nopoll, 150 /* mmap */ nommap, 151 /* strategy */ nostrategy, 152 /* name */ MD_NAME, 153 /* maj */ CDEV_MAJOR 154}; 155 156static struct cdevsw mddisk_cdevsw; 157 158static LIST_HEAD(, md_s) md_softc_list = LIST_HEAD_INITIALIZER(&md_softc_list); 159 160struct md_s { 161 int unit; 162 LIST_ENTRY(md_s) list; 163 struct devstat stats; 164 struct bio_queue_head bio_queue; 165 struct disk disk; 166 dev_t dev; 167 int busy; 168 enum md_types type; 169 unsigned nsect; 170 unsigned secsize; 171 unsigned flags; 172 173 /* MD_MALLOC related fields */ 174 u_char **secp; 175 176 /* MD_PRELOAD related fields */ 177 u_char *pl_ptr; 178 unsigned pl_len; 179 180 /* MD_VNODE related fields */ 181 struct vnode *vnode; 182 struct ucred *cred; 183 184 /* MD_OBJET related fields */ 185 vm_object_t object; 186}; 187 188static int 189mdopen(dev_t dev, int flag, int fmt, struct proc *p) 190{ 191 struct md_s *sc; 192 struct disklabel *dl; 193 194 if (md_debug) 195 printf("mdopen(%s %x %x %p)\n", 196 devtoname(dev), flag, fmt, p); 197 198 sc = dev->si_drv1; 199 200 dl = &sc->disk.d_label; 201 bzero(dl, sizeof(*dl)); 202 dl->d_secsize = sc->secsize; 203 dl->d_nsectors = sc->nsect > 63 ? 63 : sc->nsect; 204 dl->d_ntracks = 1; 205 dl->d_secpercyl = dl->d_nsectors * dl->d_ntracks; 206 dl->d_secperunit = sc->nsect; 207 dl->d_ncylinders = dl->d_secperunit / dl->d_secpercyl; 208 return (0); 209} 210 211static int 212mdioctl(dev_t dev, u_long cmd, caddr_t addr, int flags, struct proc *p) 213{ 214 215 if (md_debug) 216 printf("mdioctl(%s %lx %p %x %p)\n", 217 devtoname(dev), cmd, addr, flags, p); 218 219 return (ENOIOCTL); 220} 221 222static void 223mdstart_malloc(struct md_s *sc) 224{ 225 int i; 226 struct bio *bp; 227 devstat_trans_flags dop; 228 u_char *secp, **secpp, *dst; 229 unsigned secno, nsec, secval, uc; 230 231 for (;;) { 232 /* XXX: LOCK(unique unit numbers) */ 233 bp = bioq_first(&sc->bio_queue); 234 if (bp) 235 bioq_remove(&sc->bio_queue, bp); 236 /* XXX: UNLOCK(unique unit numbers) */ 237 if (!bp) 238 break; 239 240 devstat_start_transaction(&sc->stats); 241 242 if (bp->bio_cmd == BIO_DELETE) 243 dop = DEVSTAT_NO_DATA; 244 else if (bp->bio_cmd == BIO_READ) 245 dop = DEVSTAT_READ; 246 else 247 dop = DEVSTAT_WRITE; 248 249 nsec = bp->bio_bcount / sc->secsize; 250 secno = bp->bio_pblkno; 251 dst = bp->bio_data; 252 while (nsec--) { 253 secpp = &sc->secp[secno]; 254 if ((uintptr_t)*secpp > 255) { 255 secp = *secpp; 256 secval = 0; 257 } else { 258 secp = NULL; 259 secval = (uintptr_t) *secpp; 260 } 261 262 if (md_debug > 2) 263 printf("%x %p %p %d\n", 264 bp->bio_flags, secpp, secp, secval); 265 266 if (bp->bio_cmd == BIO_DELETE) { 267 if (!(sc->flags & MD_RESERVE) && secp != NULL) { 268 FREE(secp, M_MDSECT); 269 *secpp = 0; 270 } 271 } else if (bp->bio_cmd == BIO_READ) { 272 if (secp != NULL) { 273 bcopy(secp, dst, sc->secsize); 274 } else if (secval) { 275 for (i = 0; i < sc->secsize; i++) 276 dst[i] = secval; 277 } else { 278 bzero(dst, sc->secsize); 279 } 280 } else { 281 if (sc->flags & MD_COMPRESS) { 282 uc = dst[0]; 283 for (i = 1; i < sc->secsize; i++) 284 if (dst[i] != uc) 285 break; 286 } else { 287 i = 0; 288 uc = 0; 289 } 290 if (i == sc->secsize) { 291 if (secp) 292 FREE(secp, M_MDSECT); 293 *secpp = (u_char *)(uintptr_t)uc; 294 } else { 295 if (secp == NULL) 296 MALLOC(secp, u_char *, sc->secsize, M_MDSECT, M_WAITOK); 297 bcopy(dst, secp, sc->secsize); 298 *secpp = secp; 299 } 300 } 301 secno++; 302 dst += sc->secsize; 303 } 304 bp->bio_resid = 0; 305 devstat_end_transaction_bio(&sc->stats, bp); 306 biodone(bp); 307 } 308 return; 309} 310 311 312static void 313mdstart_preload(struct md_s *sc) 314{ 315 struct bio *bp; 316 devstat_trans_flags dop; 317 318 for (;;) { 319 /* XXX: LOCK(unique unit numbers) */ 320 bp = bioq_first(&sc->bio_queue); 321 if (bp) 322 bioq_remove(&sc->bio_queue, bp); 323 /* XXX: UNLOCK(unique unit numbers) */ 324 if (!bp) 325 break; 326 327 devstat_start_transaction(&sc->stats); 328 329 if (bp->bio_cmd == BIO_DELETE) { 330 dop = DEVSTAT_NO_DATA; 331 } else if (bp->bio_cmd == BIO_READ) { 332 dop = DEVSTAT_READ; 333 bcopy(sc->pl_ptr + (bp->bio_pblkno << DEV_BSHIFT), bp->bio_data, bp->bio_bcount); 334 } else { 335 dop = DEVSTAT_WRITE; 336 bcopy(bp->bio_data, sc->pl_ptr + (bp->bio_pblkno << DEV_BSHIFT), bp->bio_bcount); 337 } 338 bp->bio_resid = 0; 339 devstat_end_transaction_bio(&sc->stats, bp); 340 biodone(bp); 341 } 342 return; 343} 344 345static void 346mdstart_vnode(struct md_s *sc) 347{ 348 int error; 349 struct bio *bp; 350 struct uio auio; 351 struct iovec aiov; 352 struct mount *mp; 353 354 /* 355 * VNODE I/O 356 * 357 * If an error occurs, we set BIO_ERROR but we do not set 358 * B_INVAL because (for a write anyway), the buffer is 359 * still valid. 360 */ 361 362 for (;;) { 363 /* XXX: LOCK(unique unit numbers) */ 364 bp = bioq_first(&sc->bio_queue); 365 if (bp) 366 bioq_remove(&sc->bio_queue, bp); 367 /* XXX: UNLOCK(unique unit numbers) */ 368 if (!bp) 369 break; 370 371 devstat_start_transaction(&sc->stats); 372 373 bzero(&auio, sizeof(auio)); 374 375 aiov.iov_base = bp->bio_data; 376 aiov.iov_len = bp->bio_bcount; 377 auio.uio_iov = &aiov; 378 auio.uio_iovcnt = 1; 379 auio.uio_offset = (vm_ooffset_t)bp->bio_pblkno * sc->secsize; 380 auio.uio_segflg = UIO_SYSSPACE; 381 if(bp->bio_cmd == BIO_READ) 382 auio.uio_rw = UIO_READ; 383 else 384 auio.uio_rw = UIO_WRITE; 385 auio.uio_resid = bp->bio_bcount; 386 auio.uio_procp = curproc; 387 if (VOP_ISLOCKED(sc->vnode, NULL)) 388 vprint("unexpected md driver lock", sc->vnode); 389 if (bp->bio_cmd == BIO_READ) { 390 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 391 error = VOP_READ(sc->vnode, &auio, 0, sc->cred); 392 } else { 393 (void) vn_start_write(sc->vnode, &mp, V_WAIT); 394 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 395 error = VOP_WRITE(sc->vnode, &auio, 0, sc->cred); 396 vn_finished_write(mp); 397 } 398 VOP_UNLOCK(sc->vnode, 0, curproc); 399 bp->bio_resid = auio.uio_resid; 400 401 if (error) { 402 bp->bio_error = error; 403 bp->bio_flags |= BIO_ERROR; 404 } 405 devstat_end_transaction_bio(&sc->stats, bp); 406 biodone(bp); 407 } 408 return; 409} 410 411static void 412mdstart_swap(struct md_s *sc) 413{ 414 struct bio *bp; 415 416 for (;;) { 417 /* XXX: LOCK(unique unit numbers) */ 418 bp = bioq_first(&sc->bio_queue); 419 if (bp) 420 bioq_remove(&sc->bio_queue, bp); 421 /* XXX: UNLOCK(unique unit numbers) */ 422 if (!bp) 423 break; 424 425#if 0 426 devstat_start_transaction(&sc->stats); 427#endif 428 429 if ((bp->bio_cmd == BIO_DELETE) && (sc->flags & MD_RESERVE)) 430 biodone(bp); 431 else 432 vm_pager_strategy(sc->object, bp); 433 434#if 0 435 devstat_end_transaction_bio(&sc->stats, bp); 436#endif 437 } 438 return; 439} 440 441static void 442mdstrategy(struct bio *bp) 443{ 444 struct md_s *sc; 445 446 if (md_debug > 1) 447 printf("mdstrategy(%p) %s %x, %d, %ld, %p)\n", 448 bp, devtoname(bp->bio_dev), bp->bio_flags, bp->bio_blkno, 449 bp->bio_bcount / DEV_BSIZE, bp->bio_data); 450 451 sc = bp->bio_dev->si_drv1; 452 453 /* XXX: LOCK(sc->lock) */ 454 bioqdisksort(&sc->bio_queue, bp); 455 /* XXX: UNLOCK(sc->lock) */ 456 457 if (atomic_cmpset_int(&sc->busy, 0, 1) == 0) 458 return; 459 460 switch (sc->type) { 461 case MD_MALLOC: 462 mdstart_malloc(sc); 463 break; 464 case MD_PRELOAD: 465 mdstart_preload(sc); 466 break; 467 case MD_VNODE: 468 mdstart_vnode(sc); 469 break; 470 case MD_SWAP: 471 mdstart_swap(sc); 472 break; 473 default: 474 panic("Impossible md(type)"); 475 break; 476 } 477 sc->busy = 0; 478} 479 480static struct md_s * 481mdfind(int unit) 482{ 483 struct md_s *sc; 484 485 /* XXX: LOCK(unique unit numbers) */ 486 LIST_FOREACH(sc, &md_softc_list, list) { 487 if (sc->unit == unit) 488 break; 489 } 490 /* XXX: UNLOCK(unique unit numbers) */ 491 return (sc); 492} 493 494static struct md_s * 495mdnew(int unit) 496{ 497 struct md_s *sc; 498 int max = -1; 499 500 /* XXX: LOCK(unique unit numbers) */ 501 LIST_FOREACH(sc, &md_softc_list, list) { 502 if (sc->unit == unit) { 503 /* XXX: UNLOCK(unique unit numbers) */ 504 return (NULL); 505 } 506 if (sc->unit > max) 507 max = sc->unit; 508 } 509 if (unit == -1) 510 unit = max + 1; 511 if (unit > DKMAXUNIT) 512 return (NULL); 513 MALLOC(sc, struct md_s *,sizeof(*sc), M_MD, M_WAITOK | M_ZERO); 514 sc->unit = unit; 515 LIST_INSERT_HEAD(&md_softc_list, sc, list); 516 /* XXX: UNLOCK(unique unit numbers) */ 517 return (sc); 518} 519 520static void 521mddelete(struct md_s *sc) 522{ 523 524 devstat_remove_entry(&sc->stats); 525 /* XXX: LOCK(unique unit numbers) */ 526 LIST_REMOVE(sc, list); 527 /* XXX: UNLOCK(unique unit numbers) */ 528 FREE(sc, M_MD); 529} 530 531static void 532mdinit(struct md_s *sc) 533{ 534 535 bioq_init(&sc->bio_queue); 536 devstat_add_entry(&sc->stats, MD_NAME, sc->unit, sc->secsize, 537 DEVSTAT_NO_ORDERED_TAGS, 538 DEVSTAT_TYPE_DIRECT | DEVSTAT_TYPE_IF_OTHER, 539 DEVSTAT_PRIORITY_OTHER); 540 sc->dev = disk_create(sc->unit, &sc->disk, 0, &md_cdevsw, &mddisk_cdevsw); 541 sc->dev->si_drv1 = sc; 542} 543 544/* 545 * XXX: we should check that the range they feed us is mapped. 546 * XXX: we should implement read-only. 547 */ 548 549static int 550mdcreate_preload(struct md_ioctl *mdio) 551{ 552 struct md_s *sc; 553 554 if (mdio->md_size == 0) 555 return(EINVAL); 556 if (mdio->md_options & ~(MD_AUTOUNIT)) 557 return(EINVAL); 558 if (mdio->md_options & MD_AUTOUNIT) { 559 sc = mdnew(-1); 560 if (sc == NULL) 561 return (ENOMEM); 562 mdio->md_unit = sc->unit; 563 } else { 564 sc = mdnew(mdio->md_unit); 565 if (sc == NULL) 566 return (EBUSY); 567 } 568 sc->type = MD_PRELOAD; 569 sc->secsize = DEV_BSIZE; 570 sc->nsect = mdio->md_size; 571 /* Cast to pointer size, then to pointer to avoid warning */ 572 sc->pl_ptr = (u_char *)(uintptr_t)mdio->md_base; 573 sc->pl_len = (mdio->md_size << DEV_BSHIFT); 574 mdinit(sc); 575 return (0); 576} 577 578 579static int 580mdcreate_malloc(struct md_ioctl *mdio) 581{ 582 struct md_s *sc; 583 unsigned u; 584 585 if (mdio->md_size == 0) 586 return(EINVAL); 587 if (mdio->md_options & ~(MD_AUTOUNIT | MD_COMPRESS | MD_RESERVE)) 588 return(EINVAL); 589 /* Compression doesn't make sense if we have reserved space */ 590 if (mdio->md_options & MD_RESERVE) 591 mdio->md_options &= ~MD_COMPRESS; 592 if (mdio->md_options & MD_AUTOUNIT) { 593 sc = mdnew(-1); 594 if (sc == NULL) 595 return (ENOMEM); 596 mdio->md_unit = sc->unit; 597 } else { 598 sc = mdnew(mdio->md_unit); 599 if (sc == NULL) 600 return (EBUSY); 601 } 602 sc->type = MD_MALLOC; 603 sc->secsize = DEV_BSIZE; 604 sc->nsect = mdio->md_size; 605 sc->flags = mdio->md_options & MD_COMPRESS; 606 MALLOC(sc->secp, u_char **, sc->nsect * sizeof(u_char *), M_MD, M_WAITOK | M_ZERO); 607 if (mdio->md_options & MD_RESERVE) { 608 for (u = 0; u < sc->nsect; u++) 609 MALLOC(sc->secp[u], u_char *, DEV_BSIZE, M_MDSECT, M_WAITOK | M_ZERO); 610 } 611 printf("%s%d: Malloc disk\n", MD_NAME, sc->unit); 612 mdinit(sc); 613 return (0); 614} 615 616 617static int 618mdsetcred(struct md_s *sc, struct ucred *cred) 619{ 620 char *tmpbuf; 621 int error = 0; 622 623 /* 624 * Set credits in our softc 625 */ 626 627 if (sc->cred) 628 crfree(sc->cred); 629 sc->cred = crdup(cred); 630 631 /* 632 * Horrible kludge to establish credentials for NFS XXX. 633 */ 634 635 if (sc->vnode) { 636 struct uio auio; 637 struct iovec aiov; 638 639 tmpbuf = malloc(sc->secsize, M_TEMP, M_WAITOK); 640 bzero(&auio, sizeof(auio)); 641 642 aiov.iov_base = tmpbuf; 643 aiov.iov_len = sc->secsize; 644 auio.uio_iov = &aiov; 645 auio.uio_iovcnt = 1; 646 auio.uio_offset = 0; 647 auio.uio_rw = UIO_READ; 648 auio.uio_segflg = UIO_SYSSPACE; 649 auio.uio_resid = aiov.iov_len; 650 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 651 error = VOP_READ(sc->vnode, &auio, 0, sc->cred); 652 VOP_UNLOCK(sc->vnode, 0, curproc); 653 free(tmpbuf, M_TEMP); 654 } 655 return (error); 656} 657 658static int 659mdcreate_vnode(struct md_ioctl *mdio, struct proc *p) 660{ 661 struct md_s *sc; 662 struct vattr vattr; 663 struct nameidata nd; 664 int error, flags; 665 666 if (mdio->md_options & MD_AUTOUNIT) { 667 sc = mdnew(-1); 668 mdio->md_unit = sc->unit; 669 } else { 670 sc = mdnew(mdio->md_unit); 671 } 672 if (sc == NULL) 673 return (EBUSY); 674 675 sc->type = MD_VNODE; 676 677 flags = FREAD|FWRITE; 678 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, mdio->md_file, p); 679 error = vn_open(&nd, &flags, 0); 680 if (error) { 681 if (error != EACCES && error != EPERM && error != EROFS) 682 return (error); 683 flags &= ~FWRITE; 684 sc->flags |= MD_READONLY; 685 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, mdio->md_file, p); 686 error = vn_open(&nd, &flags, 0); 687 if (error) 688 return (error); 689 } 690 NDFREE(&nd, NDF_ONLY_PNBUF); 691 if (nd.ni_vp->v_type != VREG || 692 (error = VOP_GETATTR(nd.ni_vp, &vattr, p->p_ucred, p))) { 693 VOP_UNLOCK(nd.ni_vp, 0, p); 694 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p); 695 return (error ? error : EINVAL); 696 } 697 VOP_UNLOCK(nd.ni_vp, 0, p); 698 sc->secsize = DEV_BSIZE; 699 sc->vnode = nd.ni_vp; 700 701 /* 702 * If the size is specified, override the file attributes. 703 */ 704 if (mdio->md_size) 705 sc->nsect = mdio->md_size; 706 else 707 sc->nsect = vattr.va_size / sc->secsize; /* XXX: round up ? */ 708 error = mdsetcred(sc, p->p_ucred); 709 if (error) { 710 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p); 711 return(error); 712 } 713 mdinit(sc); 714 return (0); 715} 716 717static int 718mddestroy(struct md_s *sc, struct md_ioctl *mdio, struct proc *p) 719{ 720 unsigned u; 721 722 if (sc->dev != NULL) 723 disk_destroy(sc->dev); 724 if (sc->vnode != NULL) 725 (void)vn_close(sc->vnode, sc->flags & MD_READONLY ? FREAD : (FREAD|FWRITE), sc->cred, p); 726 if (sc->cred != NULL) 727 crfree(sc->cred); 728 if (sc->object != NULL) 729 vm_pager_deallocate(sc->object); 730 if (sc->secp != NULL) { 731 for (u = 0; u < sc->nsect; u++) 732 if ((uintptr_t)sc->secp[u] > 255) 733 FREE(sc->secp[u], M_MDSECT); 734 FREE(sc->secp, M_MD); 735 } 736 mddelete(sc); 737 return (0); 738} 739 740static int 741mdcreate_swap(struct md_ioctl *mdio, struct proc *p) 742{ 743 int error; 744 struct md_s *sc; 745 746 if (mdio->md_options & MD_AUTOUNIT) { 747 sc = mdnew(-1); 748 mdio->md_unit = sc->unit; 749 } else { 750 sc = mdnew(mdio->md_unit); 751 } 752 if (sc == NULL) 753 return (EBUSY); 754 755 sc->type = MD_SWAP; 756 757 /* 758 * Range check. Disallow negative sizes or any size less then the 759 * size of a page. Then round to a page. 760 */ 761 762 if (mdio->md_size == 0) 763 return(EDOM); 764 765 /* 766 * Allocate an OBJT_SWAP object. 767 * 768 * sc_secsize is PAGE_SIZE'd 769 * 770 * mdio->size is in DEV_BSIZE'd chunks. 771 * Note the truncation. 772 */ 773 774 sc->secsize = PAGE_SIZE; 775 sc->nsect = mdio->md_size / (PAGE_SIZE / DEV_BSIZE); 776 sc->object = vm_pager_allocate(OBJT_SWAP, NULL, sc->secsize * (vm_offset_t)sc->nsect, VM_PROT_DEFAULT, 0); 777 if (mdio->md_options & MD_RESERVE) { 778 if (swap_pager_reserve(sc->object, 0, sc->nsect) < 0) { 779 vm_pager_deallocate(sc->object); 780 sc->object = NULL; 781 return(EDOM); 782 } 783 } 784 error = mdsetcred(sc, p->p_ucred); 785 if (error) 786 mddestroy(sc, mdio, p); 787 else 788 mdinit(sc); 789 return(error); 790} 791 792static int 793mdctlioctl(dev_t dev, u_long cmd, caddr_t addr, int flags, struct proc *p) 794{ 795 struct md_ioctl *mdio; 796 struct md_s *sc; 797 798 if (md_debug) 799 printf("mdctlioctl(%s %lx %p %x %p)\n", 800 devtoname(dev), cmd, addr, flags, p); 801 802 mdio = (struct md_ioctl *)addr; 803 switch (cmd) { 804 case MDIOCATTACH: 805 switch (mdio->md_type) { 806 case MD_MALLOC: 807 return(mdcreate_malloc(mdio)); 808 case MD_PRELOAD: 809 return(mdcreate_preload(mdio)); 810 case MD_VNODE: 811 return(mdcreate_vnode(mdio, p)); 812 case MD_SWAP: 813 return(mdcreate_swap(mdio, p)); 814 default: 815 return (EINVAL); 816 } 817 case MDIOCDETACH: 818 if (mdio->md_file != NULL) 819 return(EINVAL); 820 if (mdio->md_size != 0) 821 return(EINVAL); 822 if (mdio->md_options != 0) 823 return(EINVAL); 824 sc = mdfind(mdio->md_unit); 825 if (sc == NULL) 826 return (ENOENT); 827 switch(sc->type) { 828 case MD_VNODE: 829 case MD_SWAP: 830 case MD_MALLOC: 831 case MD_PRELOAD: 832 return(mddestroy(sc, mdio, p)); 833 default: 834 return (EOPNOTSUPP); 835 } 836 case MDIOCQUERY: 837 sc = mdfind(mdio->md_unit); 838 if (sc == NULL) 839 return (ENOENT); 840 mdio->md_type = sc->type; 841 mdio->md_options = sc->flags; 842 switch (sc->type) { 843 case MD_MALLOC: 844 mdio->md_size = sc->nsect; 845 break; 846 case MD_PRELOAD: 847 mdio->md_size = sc->nsect; 848 (u_char *)(uintptr_t)mdio->md_base = sc->pl_ptr; 849 break; 850 case MD_SWAP: 851 mdio->md_size = sc->nsect * (PAGE_SIZE / DEV_BSIZE); 852 break; 853 case MD_VNODE: 854 mdio->md_size = sc->nsect; 855 /* XXX fill this in */ 856 mdio->md_file = NULL; 857 break; 858 } 859 return (0); 860 default: 861 return (ENOIOCTL); 862 }; 863 return (ENOIOCTL); 864} 865 866static void 867md_preloaded(u_char *image, unsigned length) 868{ 869 struct md_s *sc; 870 871 sc = mdnew(-1); 872 if (sc == NULL) 873 return; 874 sc->type = MD_PRELOAD; 875 sc->secsize = DEV_BSIZE; 876 sc->nsect = length / DEV_BSIZE; 877 sc->pl_ptr = image; 878 sc->pl_len = length; 879 if (sc->unit == 0) 880 mdrootready = 1; 881 mdinit(sc); 882} 883 884static void 885md_drvinit(void *unused) 886{ 887 888 caddr_t mod; 889 caddr_t c; 890 u_char *ptr, *name, *type; 891 unsigned len; 892 893#ifdef MD_ROOT_SIZE 894 md_preloaded(mfs_root, MD_ROOT_SIZE*1024); 895#endif 896 mod = NULL; 897 while ((mod = preload_search_next_name(mod)) != NULL) { 898 name = (char *)preload_search_info(mod, MODINFO_NAME); 899 type = (char *)preload_search_info(mod, MODINFO_TYPE); 900 if (name == NULL) 901 continue; 902 if (type == NULL) 903 continue; 904 if (strcmp(type, "md_image") && strcmp(type, "mfs_root")) 905 continue; 906 c = preload_search_info(mod, MODINFO_ADDR); 907 ptr = *(u_char **)c; 908 c = preload_search_info(mod, MODINFO_SIZE); 909 len = *(unsigned *)c; 910 printf("md%d: Preloaded image <%s> %d bytes at %p\n", 911 mdunits, name, len, ptr); 912 md_preloaded(ptr, len); 913 } 914 status_dev = make_dev(&mdctl_cdevsw, 0xffff00ff, UID_ROOT, GID_WHEEL, 0600, "mdctl"); 915} 916 917static int 918md_modevent(module_t mod, int type, void *data) 919{ 920 switch (type) { 921 case MOD_LOAD: 922 md_drvinit(NULL); 923 break; 924 case MOD_UNLOAD: 925 if (!LIST_EMPTY(&md_softc_list)) 926 return EBUSY; 927 if (status_dev) 928 destroy_dev(status_dev); 929 status_dev = 0; 930 break; 931 default: 932 break; 933 } 934 return 0; 935} 936 937static moduledata_t md_mod = { 938 "md", 939 md_modevent, 940 NULL 941}; 942DECLARE_MODULE(md, md_mod, SI_SUB_DRIVERS, SI_ORDER_MIDDLE+CDEV_MAJOR); 943MODULE_VERSION(md, MD_MODVER); 944 945 946#ifdef MD_ROOT 947static void 948md_takeroot(void *junk) 949{ 950 if (mdrootready) 951 rootdevnames[0] = "ufs:/dev/md0c"; 952} 953 954SYSINIT(md_root, SI_SUB_MOUNT_ROOT, SI_ORDER_FIRST, md_takeroot, NULL); 955#endif 956 957