md.c revision 76365
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 76365 2001-05-08 09:09:32Z 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/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 96MALLOC_DEFINE(M_MD, "MD disk", "Memory Disk"); 97MALLOC_DEFINE(M_MDSECT, "MD sectors", "Memory Disk Sectors"); 98 99static int md_debug; 100SYSCTL_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: */ 114static u_char mfs_root[MD_ROOT_SIZE*1024] = "MFS Filesystem goes here"; 115static u_char end_mfs_root[] __unused = "MFS Filesystem had better STOP here"; 116#endif 117 118static int mdrootready; 119static int mdunits; 120static dev_t status_dev = 0; 121 122 123#define CDEV_MAJOR 95 124 125static d_strategy_t mdstrategy; 126static d_open_t mdopen; 127static d_ioctl_t mdioctl, mdctlioctl; 128 129static 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 145static 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 158static struct cdevsw mddisk_cdevsw; 159 160static LIST_HEAD(, md_s) md_softc_list = LIST_HEAD_INITIALIZER(&md_softc_list); 161 162struct 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 190static int 191mdopen(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 213static int 214mdioctl(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 224static void 225mdstart_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 biofinish(bp, &sc->stats, 0); 308 } 309 return; 310} 311 312 313static void 314mdstart_preload(struct md_s *sc) 315{ 316 struct bio *bp; 317 devstat_trans_flags dop; 318 319 for (;;) { 320 /* XXX: LOCK(unique unit numbers) */ 321 bp = bioq_first(&sc->bio_queue); 322 if (bp) 323 bioq_remove(&sc->bio_queue, bp); 324 /* XXX: UNLOCK(unique unit numbers) */ 325 if (!bp) 326 break; 327 328 devstat_start_transaction(&sc->stats); 329 330 if (bp->bio_cmd == BIO_DELETE) { 331 dop = DEVSTAT_NO_DATA; 332 } else if (bp->bio_cmd == BIO_READ) { 333 dop = DEVSTAT_READ; 334 bcopy(sc->pl_ptr + (bp->bio_pblkno << DEV_BSHIFT), bp->bio_data, bp->bio_bcount); 335 } else { 336 dop = DEVSTAT_WRITE; 337 bcopy(bp->bio_data, sc->pl_ptr + (bp->bio_pblkno << DEV_BSHIFT), bp->bio_bcount); 338 } 339 bp->bio_resid = 0; 340 biofinish(bp, &sc->stats, 0); 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 biofinish(bp, &sc->stats, error); 401 } 402 return; 403} 404 405static void 406mdstart_swap(struct md_s *sc) 407{ 408 struct bio *bp; 409 410 for (;;) { 411 /* XXX: LOCK(unique unit numbers) */ 412 bp = bioq_first(&sc->bio_queue); 413 if (bp) 414 bioq_remove(&sc->bio_queue, bp); 415 /* XXX: UNLOCK(unique unit numbers) */ 416 if (!bp) 417 break; 418 419#if 0 420 devstat_start_transaction(&sc->stats); 421#endif 422 423 if ((bp->bio_cmd == BIO_DELETE) && (sc->flags & MD_RESERVE)) 424 biodone(bp); 425 else 426 vm_pager_strategy(sc->object, bp); 427 428#if 0 429 devstat_end_transaction_bio(&sc->stats, bp); 430#endif 431 } 432 return; 433} 434 435static void 436mdstrategy(struct bio *bp) 437{ 438 struct md_s *sc; 439 440 if (md_debug > 1) 441 printf("mdstrategy(%p) %s %x, %d, %ld, %p)\n", 442 bp, devtoname(bp->bio_dev), bp->bio_flags, bp->bio_blkno, 443 bp->bio_bcount / DEV_BSIZE, bp->bio_data); 444 445 sc = bp->bio_dev->si_drv1; 446 447 /* XXX: LOCK(sc->lock) */ 448 bioqdisksort(&sc->bio_queue, bp); 449 /* XXX: UNLOCK(sc->lock) */ 450 451 if (atomic_cmpset_int(&sc->busy, 0, 1) == 0) 452 return; 453 454 switch (sc->type) { 455 case MD_MALLOC: 456 mdstart_malloc(sc); 457 break; 458 case MD_PRELOAD: 459 mdstart_preload(sc); 460 break; 461 case MD_VNODE: 462 mdstart_vnode(sc); 463 break; 464 case MD_SWAP: 465 mdstart_swap(sc); 466 break; 467 default: 468 panic("Impossible md(type)"); 469 break; 470 } 471 sc->busy = 0; 472} 473 474static struct md_s * 475mdfind(int unit) 476{ 477 struct md_s *sc; 478 479 /* XXX: LOCK(unique unit numbers) */ 480 LIST_FOREACH(sc, &md_softc_list, list) { 481 if (sc->unit == unit) 482 break; 483 } 484 /* XXX: UNLOCK(unique unit numbers) */ 485 return (sc); 486} 487 488static struct md_s * 489mdnew(int unit) 490{ 491 struct md_s *sc; 492 int max = -1; 493 494 /* XXX: LOCK(unique unit numbers) */ 495 LIST_FOREACH(sc, &md_softc_list, list) { 496 if (sc->unit == unit) { 497 /* XXX: UNLOCK(unique unit numbers) */ 498 return (NULL); 499 } 500 if (sc->unit > max) 501 max = sc->unit; 502 } 503 if (unit == -1) 504 unit = max + 1; 505 if (unit > DKMAXUNIT) 506 return (NULL); 507 MALLOC(sc, struct md_s *,sizeof(*sc), M_MD, M_WAITOK | M_ZERO); 508 sc->unit = unit; 509 LIST_INSERT_HEAD(&md_softc_list, sc, list); 510 /* XXX: UNLOCK(unique unit numbers) */ 511 return (sc); 512} 513 514static void 515mdinit(struct md_s *sc) 516{ 517 518 bioq_init(&sc->bio_queue); 519 devstat_add_entry(&sc->stats, MD_NAME, sc->unit, sc->secsize, 520 DEVSTAT_NO_ORDERED_TAGS, 521 DEVSTAT_TYPE_DIRECT | DEVSTAT_TYPE_IF_OTHER, 522 DEVSTAT_PRIORITY_OTHER); 523 sc->dev = disk_create(sc->unit, &sc->disk, 0, &md_cdevsw, &mddisk_cdevsw); 524 sc->dev->si_drv1 = sc; 525} 526 527/* 528 * XXX: we should check that the range they feed us is mapped. 529 * XXX: we should implement read-only. 530 */ 531 532static int 533mdcreate_preload(struct md_ioctl *mdio) 534{ 535 struct md_s *sc; 536 537 if (mdio->md_size == 0) 538 return(EINVAL); 539 if (mdio->md_options & ~(MD_AUTOUNIT)) 540 return(EINVAL); 541 if (mdio->md_options & MD_AUTOUNIT) { 542 sc = mdnew(-1); 543 if (sc == NULL) 544 return (ENOMEM); 545 mdio->md_unit = sc->unit; 546 } else { 547 sc = mdnew(mdio->md_unit); 548 if (sc == NULL) 549 return (EBUSY); 550 } 551 sc->type = MD_PRELOAD; 552 sc->secsize = DEV_BSIZE; 553 sc->nsect = mdio->md_size; 554 /* Cast to pointer size, then to pointer to avoid warning */ 555 sc->pl_ptr = (u_char *)(uintptr_t)mdio->md_base; 556 sc->pl_len = (mdio->md_size << DEV_BSHIFT); 557 mdinit(sc); 558 return (0); 559} 560 561 562static int 563mdcreate_malloc(struct md_ioctl *mdio) 564{ 565 struct md_s *sc; 566 unsigned u; 567 568 if (mdio->md_size == 0) 569 return(EINVAL); 570 if (mdio->md_options & ~(MD_AUTOUNIT | MD_COMPRESS | MD_RESERVE)) 571 return(EINVAL); 572 /* Compression doesn't make sense if we have reserved space */ 573 if (mdio->md_options & MD_RESERVE) 574 mdio->md_options &= ~MD_COMPRESS; 575 if (mdio->md_options & MD_AUTOUNIT) { 576 sc = mdnew(-1); 577 if (sc == NULL) 578 return (ENOMEM); 579 mdio->md_unit = sc->unit; 580 } else { 581 sc = mdnew(mdio->md_unit); 582 if (sc == NULL) 583 return (EBUSY); 584 } 585 sc->type = MD_MALLOC; 586 sc->secsize = DEV_BSIZE; 587 sc->nsect = mdio->md_size; 588 sc->flags = mdio->md_options & MD_COMPRESS; 589 MALLOC(sc->secp, u_char **, sc->nsect * sizeof(u_char *), M_MD, M_WAITOK | M_ZERO); 590 if (mdio->md_options & MD_RESERVE) { 591 for (u = 0; u < sc->nsect; u++) 592 MALLOC(sc->secp[u], u_char *, DEV_BSIZE, M_MDSECT, M_WAITOK | M_ZERO); 593 } 594 printf("%s%d: Malloc disk\n", MD_NAME, sc->unit); 595 mdinit(sc); 596 return (0); 597} 598 599 600static int 601mdsetcred(struct md_s *sc, struct ucred *cred) 602{ 603 char *tmpbuf; 604 int error = 0; 605 606 /* 607 * Set credits in our softc 608 */ 609 610 if (sc->cred) 611 crfree(sc->cred); 612 sc->cred = crdup(cred); 613 614 /* 615 * Horrible kludge to establish credentials for NFS XXX. 616 */ 617 618 if (sc->vnode) { 619 struct uio auio; 620 struct iovec aiov; 621 622 tmpbuf = malloc(sc->secsize, M_TEMP, M_WAITOK); 623 bzero(&auio, sizeof(auio)); 624 625 aiov.iov_base = tmpbuf; 626 aiov.iov_len = sc->secsize; 627 auio.uio_iov = &aiov; 628 auio.uio_iovcnt = 1; 629 auio.uio_offset = 0; 630 auio.uio_rw = UIO_READ; 631 auio.uio_segflg = UIO_SYSSPACE; 632 auio.uio_resid = aiov.iov_len; 633 vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curproc); 634 error = VOP_READ(sc->vnode, &auio, 0, sc->cred); 635 VOP_UNLOCK(sc->vnode, 0, curproc); 636 free(tmpbuf, M_TEMP); 637 } 638 return (error); 639} 640 641static int 642mdcreate_vnode(struct md_ioctl *mdio, struct proc *p) 643{ 644 struct md_s *sc; 645 struct vattr vattr; 646 struct nameidata nd; 647 int error, flags; 648 649 if (mdio->md_options & MD_AUTOUNIT) { 650 sc = mdnew(-1); 651 mdio->md_unit = sc->unit; 652 } else { 653 sc = mdnew(mdio->md_unit); 654 } 655 if (sc == NULL) 656 return (EBUSY); 657 658 sc->type = MD_VNODE; 659 660 flags = FREAD|FWRITE; 661 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, mdio->md_file, p); 662 error = vn_open(&nd, &flags, 0); 663 if (error) { 664 if (error != EACCES && error != EPERM && error != EROFS) 665 return (error); 666 flags &= ~FWRITE; 667 sc->flags |= MD_READONLY; 668 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, mdio->md_file, p); 669 error = vn_open(&nd, &flags, 0); 670 if (error) 671 return (error); 672 } 673 NDFREE(&nd, NDF_ONLY_PNBUF); 674 if (nd.ni_vp->v_type != VREG || 675 (error = VOP_GETATTR(nd.ni_vp, &vattr, p->p_ucred, p))) { 676 VOP_UNLOCK(nd.ni_vp, 0, p); 677 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p); 678 return (error ? error : EINVAL); 679 } 680 VOP_UNLOCK(nd.ni_vp, 0, p); 681 sc->secsize = DEV_BSIZE; 682 sc->vnode = nd.ni_vp; 683 684 /* 685 * If the size is specified, override the file attributes. 686 */ 687 if (mdio->md_size) 688 sc->nsect = mdio->md_size; 689 else 690 sc->nsect = vattr.va_size / sc->secsize; /* XXX: round up ? */ 691 error = mdsetcred(sc, p->p_ucred); 692 if (error) { 693 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p); 694 return(error); 695 } 696 mdinit(sc); 697 return (0); 698} 699 700static int 701mddestroy(struct md_s *sc, struct md_ioctl *mdio, struct proc *p) 702{ 703 unsigned u; 704 705 if (sc->dev != NULL) { 706 devstat_remove_entry(&sc->stats); 707 disk_destroy(sc->dev); 708 } 709 if (sc->vnode != NULL) 710 (void)vn_close(sc->vnode, sc->flags & MD_READONLY ? FREAD : (FREAD|FWRITE), sc->cred, p); 711 if (sc->cred != NULL) 712 crfree(sc->cred); 713 if (sc->object != NULL) 714 vm_pager_deallocate(sc->object); 715 if (sc->secp != NULL) { 716 for (u = 0; u < sc->nsect; u++) 717 if ((uintptr_t)sc->secp[u] > 255) 718 FREE(sc->secp[u], M_MDSECT); 719 FREE(sc->secp, M_MD); 720 } 721 722 /* XXX: LOCK(unique unit numbers) */ 723 LIST_REMOVE(sc, list); 724 /* XXX: UNLOCK(unique unit numbers) */ 725 FREE(sc, M_MD); 726 return (0); 727} 728 729static int 730mdcreate_swap(struct md_ioctl *mdio, struct proc *p) 731{ 732 int error; 733 struct md_s *sc; 734 735 if (mdio->md_options & MD_AUTOUNIT) { 736 sc = mdnew(-1); 737 mdio->md_unit = sc->unit; 738 } else { 739 sc = mdnew(mdio->md_unit); 740 } 741 if (sc == NULL) 742 return (EBUSY); 743 744 sc->type = MD_SWAP; 745 746 /* 747 * Range check. Disallow negative sizes or any size less then the 748 * size of a page. Then round to a page. 749 */ 750 751 if (mdio->md_size == 0) { 752 mddestroy(sc, mdio, p); 753 return(EDOM); 754 } 755 756 /* 757 * Allocate an OBJT_SWAP object. 758 * 759 * sc_secsize is PAGE_SIZE'd 760 * 761 * mdio->size is in DEV_BSIZE'd chunks. 762 * Note the truncation. 763 */ 764 765 sc->secsize = PAGE_SIZE; 766 sc->nsect = mdio->md_size / (PAGE_SIZE / DEV_BSIZE); 767 sc->object = vm_pager_allocate(OBJT_SWAP, NULL, sc->secsize * (vm_offset_t)sc->nsect, VM_PROT_DEFAULT, 0); 768 if (mdio->md_options & MD_RESERVE) { 769 if (swap_pager_reserve(sc->object, 0, sc->nsect) < 0) { 770 vm_pager_deallocate(sc->object); 771 sc->object = NULL; 772 mddestroy(sc, mdio, p); 773 return(EDOM); 774 } 775 } 776 error = mdsetcred(sc, p->p_ucred); 777 if (error) 778 mddestroy(sc, mdio, p); 779 else 780 mdinit(sc); 781 return(error); 782} 783 784static int 785mdctlioctl(dev_t dev, u_long cmd, caddr_t addr, int flags, struct proc *p) 786{ 787 struct md_ioctl *mdio; 788 struct md_s *sc; 789 790 if (md_debug) 791 printf("mdctlioctl(%s %lx %p %x %p)\n", 792 devtoname(dev), cmd, addr, flags, p); 793 794 mdio = (struct md_ioctl *)addr; 795 switch (cmd) { 796 case MDIOCATTACH: 797 switch (mdio->md_type) { 798 case MD_MALLOC: 799 return(mdcreate_malloc(mdio)); 800 case MD_PRELOAD: 801 return(mdcreate_preload(mdio)); 802 case MD_VNODE: 803 return(mdcreate_vnode(mdio, p)); 804 case MD_SWAP: 805 return(mdcreate_swap(mdio, p)); 806 default: 807 return (EINVAL); 808 } 809 case MDIOCDETACH: 810 if (mdio->md_file != NULL) 811 return(EINVAL); 812 if (mdio->md_size != 0) 813 return(EINVAL); 814 if (mdio->md_options != 0) 815 return(EINVAL); 816 sc = mdfind(mdio->md_unit); 817 if (sc == NULL) 818 return (ENOENT); 819 switch(sc->type) { 820 case MD_VNODE: 821 case MD_SWAP: 822 case MD_MALLOC: 823 case MD_PRELOAD: 824 return(mddestroy(sc, mdio, p)); 825 default: 826 return (EOPNOTSUPP); 827 } 828 case MDIOCQUERY: 829 sc = mdfind(mdio->md_unit); 830 if (sc == NULL) 831 return (ENOENT); 832 mdio->md_type = sc->type; 833 mdio->md_options = sc->flags; 834 switch (sc->type) { 835 case MD_MALLOC: 836 mdio->md_size = sc->nsect; 837 break; 838 case MD_PRELOAD: 839 mdio->md_size = sc->nsect; 840 (u_char *)(uintptr_t)mdio->md_base = sc->pl_ptr; 841 break; 842 case MD_SWAP: 843 mdio->md_size = sc->nsect * (PAGE_SIZE / DEV_BSIZE); 844 break; 845 case MD_VNODE: 846 mdio->md_size = sc->nsect; 847 /* XXX fill this in */ 848 mdio->md_file = NULL; 849 break; 850 } 851 return (0); 852 default: 853 return (ENOIOCTL); 854 }; 855 return (ENOIOCTL); 856} 857 858static void 859md_preloaded(u_char *image, unsigned length) 860{ 861 struct md_s *sc; 862 863 sc = mdnew(-1); 864 if (sc == NULL) 865 return; 866 sc->type = MD_PRELOAD; 867 sc->secsize = DEV_BSIZE; 868 sc->nsect = length / DEV_BSIZE; 869 sc->pl_ptr = image; 870 sc->pl_len = length; 871 if (sc->unit == 0) 872 mdrootready = 1; 873 mdinit(sc); 874} 875 876static void 877md_drvinit(void *unused) 878{ 879 880 caddr_t mod; 881 caddr_t c; 882 u_char *ptr, *name, *type; 883 unsigned len; 884 885#ifdef MD_ROOT_SIZE 886 md_preloaded(mfs_root, MD_ROOT_SIZE*1024); 887#endif 888 mod = NULL; 889 while ((mod = preload_search_next_name(mod)) != NULL) { 890 name = (char *)preload_search_info(mod, MODINFO_NAME); 891 type = (char *)preload_search_info(mod, MODINFO_TYPE); 892 if (name == NULL) 893 continue; 894 if (type == NULL) 895 continue; 896 if (strcmp(type, "md_image") && strcmp(type, "mfs_root")) 897 continue; 898 c = preload_search_info(mod, MODINFO_ADDR); 899 ptr = *(u_char **)c; 900 c = preload_search_info(mod, MODINFO_SIZE); 901 len = *(unsigned *)c; 902 printf("md%d: Preloaded image <%s> %d bytes at %p\n", 903 mdunits, name, len, ptr); 904 md_preloaded(ptr, len); 905 } 906 status_dev = make_dev(&mdctl_cdevsw, 0xffff00ff, UID_ROOT, GID_WHEEL, 0600, "mdctl"); 907} 908 909static int 910md_modevent(module_t mod, int type, void *data) 911{ 912 switch (type) { 913 case MOD_LOAD: 914 md_drvinit(NULL); 915 break; 916 case MOD_UNLOAD: 917 if (!LIST_EMPTY(&md_softc_list)) 918 return EBUSY; 919 if (status_dev) 920 destroy_dev(status_dev); 921 status_dev = 0; 922 break; 923 default: 924 break; 925 } 926 return 0; 927} 928 929static moduledata_t md_mod = { 930 "md", 931 md_modevent, 932 NULL 933}; 934DECLARE_MODULE(md, md_mod, SI_SUB_DRIVERS, SI_ORDER_MIDDLE+CDEV_MAJOR); 935MODULE_VERSION(md, MD_MODVER); 936 937 938#ifdef MD_ROOT 939static void 940md_takeroot(void *junk) 941{ 942 if (mdrootready) 943 rootdevnames[0] = "ufs:/dev/md0c"; 944} 945 946SYSINIT(md_root, SI_SUB_MOUNT_ROOT, SI_ORDER_FIRST, md_takeroot, NULL); 947#endif 948 949