1 /* 2 * Copyright (c) 1988 University of Utah. 3 * Copyright (c) 1990, 1993 4 * The Regents of the University of California. All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * the Systems Programming Group of the University of Utah Computer 8 * Science Department. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * from: Utah Hdr: vn.c 1.13 94/04/02 39 * 40 * from: @(#)vn.c 8.6 (Berkeley) 4/1/94 41 * $FreeBSD: src/sys/dev/vn/vn.c,v 1.105.2.4 2001/11/18 07:11:00 dillon Exp $ 42 * $DragonFly: src/sys/dev/disk/vn/vn.c,v 1.11 2004/05/26 01:14:36 dillon Exp $ 43 */ 44 45 /* 46 * Vnode disk driver. 47 * 48 * Block/character interface to a vnode. Allows one to treat a file 49 * as a disk (e.g. build a filesystem in it, mount it, etc.). 50 * 51 * NOTE 1: This uses the VOP_BMAP/VOP_STRATEGY interface to the vnode 52 * instead of a simple VOP_RDWR. We do this to avoid distorting the 53 * local buffer cache. 54 * 55 * NOTE 2: There is a security issue involved with this driver. 56 * Once mounted all access to the contents of the "mapped" file via 57 * the special file is controlled by the permissions on the special 58 * file, the protection of the mapped file is ignored (effectively, 59 * by using root credentials in all transactions). 60 * 61 * NOTE 3: Doesn't interact with leases, should it? 62 */ 63 64 #include <sys/param.h> 65 #include <sys/systm.h> 66 #include <sys/kernel.h> 67 #include <sys/proc.h> 68 #include <sys/namei.h> 69 #include <sys/buf.h> 70 #include <sys/malloc.h> 71 #include <sys/mount.h> 72 #include <sys/vnode.h> 73 #include <sys/fcntl.h> 74 #include <sys/conf.h> 75 #include <sys/disklabel.h> 76 #include <sys/diskslice.h> 77 #include <sys/stat.h> 78 #include <sys/conf.h> 79 #include <sys/module.h> 80 #include <sys/vnioctl.h> 81 82 #include <vm/vm.h> 83 #include <vm/vm_object.h> 84 #include <vm/vm_page.h> 85 #include <vm/vm_pager.h> 86 #include <vm/vm_pageout.h> 87 #include <vm/swap_pager.h> 88 #include <vm/vm_extern.h> 89 #include <vm/vm_zone.h> 90 91 static d_ioctl_t vnioctl; 92 static d_open_t vnopen; 93 static d_close_t vnclose; 94 static d_psize_t vnsize; 95 static d_strategy_t vnstrategy; 96 97 #define CDEV_MAJOR 43 98 99 #define VN_BSIZE_BEST 8192 100 101 /* 102 * cdevsw 103 * D_DISK we want to look like a disk 104 * D_CANFREE We support B_FREEBUF 105 */ 106 107 static struct cdevsw vn_cdevsw = { 108 /* name */ "vn", 109 /* maj */ CDEV_MAJOR, 110 /* flags */ D_DISK|D_CANFREE, 111 /* port */ NULL, 112 /* clone */ NULL, 113 114 /* open */ vnopen, 115 /* close */ vnclose, 116 /* read */ physread, 117 /* write */ physwrite, 118 /* ioctl */ vnioctl, 119 /* poll */ nopoll, 120 /* mmap */ nommap, 121 /* strategy */ vnstrategy, 122 /* dump */ nodump, 123 /* psize */ vnsize 124 }; 125 126 #define getvnbuf() \ 127 ((struct buf *)malloc(sizeof(struct buf), M_DEVBUF, M_WAITOK)) 128 129 #define putvnbuf(bp) \ 130 free((caddr_t)(bp), M_DEVBUF) 131 132 struct vn_softc { 133 int sc_unit; 134 int sc_flags; /* flags */ 135 int sc_size; /* size of vn, sc_secsize scale */ 136 int sc_secsize; /* sector size */ 137 struct diskslices *sc_slices; 138 struct vnode *sc_vp; /* vnode if not NULL */ 139 vm_object_t sc_object; /* backing object if not NULL */ 140 struct ucred *sc_cred; /* credentials */ 141 int sc_maxactive; /* max # of active requests */ 142 struct buf sc_tab; /* transfer queue */ 143 u_long sc_options; /* options */ 144 dev_t sc_devlist; /* devices that refer to this unit */ 145 SLIST_ENTRY(vn_softc) sc_list; 146 }; 147 148 static SLIST_HEAD(, vn_softc) vn_list; 149 150 /* sc_flags */ 151 #define VNF_INITED 0x01 152 #define VNF_READONLY 0x02 153 154 static u_long vn_options; 155 156 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt)) 157 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt)) 158 159 static int vnsetcred (struct vn_softc *vn, struct ucred *cred); 160 static void vnclear (struct vn_softc *vn); 161 static int vn_modevent (module_t, int, void *); 162 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct thread *p); 163 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct thread *p); 164 165 static int 166 vnclose(dev_t dev, int flags, int mode, struct thread *td) 167 { 168 struct vn_softc *vn = dev->si_drv1; 169 170 IFOPT(vn, VN_LABELS) 171 if (vn->sc_slices != NULL) 172 dsclose(dev, mode, vn->sc_slices); 173 return (0); 174 } 175 176 /* 177 * Called only when si_drv1 is NULL. Locate the associated vn node and 178 * attach the device to it. 179 */ 180 static struct vn_softc * 181 vnfindvn(dev_t dev) 182 { 183 int unit; 184 struct vn_softc *vn; 185 186 unit = dkunit(dev); 187 SLIST_FOREACH(vn, &vn_list, sc_list) { 188 if (vn->sc_unit == unit) { 189 dev->si_drv1 = vn; 190 dev->si_drv2 = vn->sc_devlist; 191 vn->sc_devlist = dev; 192 reference_dev(dev); 193 break; 194 } 195 } 196 if (vn == NULL) { 197 vn = malloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO); 198 vn->sc_unit = unit; 199 dev->si_drv1 = vn; 200 vn->sc_devlist = make_dev(&vn_cdevsw, 0, UID_ROOT, 201 GID_OPERATOR, 0640, "vn%d", unit); 202 if (vn->sc_devlist->si_drv1 == NULL) { 203 reference_dev(vn->sc_devlist); 204 vn->sc_devlist->si_drv1 = vn; 205 vn->sc_devlist->si_drv2 = NULL; 206 } 207 if (vn->sc_devlist != dev) { 208 dev->si_drv1 = vn; 209 dev->si_drv2 = vn->sc_devlist; 210 vn->sc_devlist = dev; 211 reference_dev(dev); 212 } 213 SLIST_INSERT_HEAD(&vn_list, vn, sc_list); 214 } 215 return (vn); 216 } 217 218 static int 219 vnopen(dev_t dev, int flags, int mode, struct thread *td) 220 { 221 struct vn_softc *vn; 222 223 /* 224 * Locate preexisting device 225 */ 226 227 if ((vn = dev->si_drv1) == NULL) 228 vn = vnfindvn(dev); 229 230 /* 231 * Update si_bsize fields for device. This data will be overriden by 232 * the slice/parition code for vn accesses through partitions, and 233 * used directly if you open the 'whole disk' device. 234 * 235 * si_bsize_best must be reinitialized in case VN has been 236 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency. 237 */ 238 dev->si_bsize_phys = vn->sc_secsize; 239 dev->si_bsize_best = vn->sc_secsize; 240 if (dev->si_bsize_best < VN_BSIZE_BEST) 241 dev->si_bsize_best = VN_BSIZE_BEST; 242 243 if ((flags & FWRITE) && (vn->sc_flags & VNF_READONLY)) 244 return (EACCES); 245 246 IFOPT(vn, VN_FOLLOW) 247 printf("vnopen(%s, 0x%x, 0x%x, %p)\n", 248 devtoname(dev), flags, mode, (void *)td); 249 250 /* 251 * Initialize label 252 */ 253 254 IFOPT(vn, VN_LABELS) { 255 if (vn->sc_flags & VNF_INITED) { 256 struct disklabel label; 257 258 /* Build label for whole disk. */ 259 bzero(&label, sizeof label); 260 label.d_secsize = vn->sc_secsize; 261 label.d_nsectors = 32; 262 label.d_ntracks = 64 / (vn->sc_secsize / DEV_BSIZE); 263 label.d_secpercyl = label.d_nsectors * label.d_ntracks; 264 label.d_ncylinders = vn->sc_size / label.d_secpercyl; 265 label.d_secperunit = vn->sc_size; 266 label.d_partitions[RAW_PART].p_size = vn->sc_size; 267 268 return (dsopen(dev, mode, 0, &vn->sc_slices, &label)); 269 } 270 if (dkslice(dev) != WHOLE_DISK_SLICE || 271 dkpart(dev) != RAW_PART || 272 mode != S_IFCHR) { 273 return (ENXIO); 274 } 275 } 276 return(0); 277 } 278 279 /* 280 * vnstrategy: 281 * 282 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls 283 * for vnode-backed vn's, and the new vm_pager_strategy() call for 284 * vm_object-backed vn's. 285 * 286 * Currently B_ASYNC is only partially handled - for OBJT_SWAP I/O only. 287 * 288 * NOTE: bp->b_blkno is DEV_BSIZE'd. We must generate bp->b_pblkno for 289 * our uio or vn_pager_strategy() call that is vn->sc_secsize'd 290 */ 291 292 static void 293 vnstrategy(struct buf *bp) 294 { 295 int unit; 296 struct vn_softc *vn; 297 int error; 298 299 unit = dkunit(bp->b_dev); 300 if ((vn = bp->b_dev->si_drv1) == NULL) 301 vn = vnfindvn(bp->b_dev); 302 303 IFOPT(vn, VN_DEBUG) 304 printf("vnstrategy(%p): unit %d\n", bp, unit); 305 306 if ((vn->sc_flags & VNF_INITED) == 0) { 307 bp->b_error = ENXIO; 308 bp->b_flags |= B_ERROR; 309 biodone(bp); 310 return; 311 } 312 313 bp->b_resid = bp->b_bcount; 314 315 IFOPT(vn, VN_LABELS) { 316 if (vn->sc_slices != NULL && dscheck(bp, vn->sc_slices) <= 0) { 317 bp->b_flags |= B_INVAL; 318 biodone(bp); 319 return; 320 } 321 } else { 322 int pbn; /* in sc_secsize chunks */ 323 long sz; /* in sc_secsize chunks */ 324 325 /* 326 * Check for required alignment. Transfers must be a valid 327 * multiple of the sector size. 328 */ 329 if (bp->b_bcount % vn->sc_secsize != 0 || 330 bp->b_blkno % (vn->sc_secsize / DEV_BSIZE) != 0) { 331 bp->b_error = EINVAL; 332 bp->b_flags |= B_ERROR | B_INVAL; 333 biodone(bp); 334 return; 335 } 336 337 pbn = bp->b_blkno / (vn->sc_secsize / DEV_BSIZE); 338 sz = howmany(bp->b_bcount, vn->sc_secsize); 339 340 /* 341 * If out of bounds return an error. If at the EOF point, 342 * simply read or write less. 343 */ 344 if (pbn < 0 || pbn >= vn->sc_size) { 345 if (pbn != vn->sc_size) { 346 bp->b_error = EINVAL; 347 bp->b_flags |= B_ERROR | B_INVAL; 348 } 349 biodone(bp); 350 return; 351 } 352 353 /* 354 * If the request crosses EOF, truncate the request. 355 */ 356 if (pbn + sz > vn->sc_size) { 357 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize; 358 bp->b_resid = bp->b_bcount; 359 } 360 bp->b_pblkno = pbn; 361 } 362 363 if (vn->sc_vp && (bp->b_flags & B_FREEBUF)) { 364 /* 365 * Not handled for vnode-backed element yet. 366 */ 367 biodone(bp); 368 } else if (vn->sc_vp) { 369 /* 370 * VNODE I/O 371 * 372 * If an error occurs, we set B_ERROR but we do not set 373 * B_INVAL because (for a write anyway), the buffer is 374 * still valid. 375 */ 376 struct uio auio; 377 struct iovec aiov; 378 379 bzero(&auio, sizeof(auio)); 380 381 aiov.iov_base = bp->b_data; 382 aiov.iov_len = bp->b_bcount; 383 auio.uio_iov = &aiov; 384 auio.uio_iovcnt = 1; 385 auio.uio_offset = (vm_ooffset_t)bp->b_pblkno * vn->sc_secsize; 386 auio.uio_segflg = UIO_SYSSPACE; 387 if( bp->b_flags & B_READ) 388 auio.uio_rw = UIO_READ; 389 else 390 auio.uio_rw = UIO_WRITE; 391 auio.uio_resid = bp->b_bcount; 392 auio.uio_td = curthread; 393 vn_lock(vn->sc_vp, NULL, LK_EXCLUSIVE | LK_RETRY, curthread); 394 if (bp->b_flags & B_READ) 395 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT, vn->sc_cred); 396 else 397 error = VOP_WRITE(vn->sc_vp, &auio, IO_NOWDRAIN, vn->sc_cred); 398 VOP_UNLOCK(vn->sc_vp, NULL, 0, curthread); 399 bp->b_resid = auio.uio_resid; 400 401 if (error) { 402 bp->b_error = error; 403 bp->b_flags |= B_ERROR; 404 } 405 biodone(bp); 406 } else if (vn->sc_object) { 407 /* 408 * OBJT_SWAP I/O 409 * 410 * ( handles read, write, freebuf ) 411 * 412 * Note: if we pre-reserved swap, B_FREEBUF is disabled 413 */ 414 KASSERT((bp->b_bufsize & (vn->sc_secsize - 1)) == 0, 415 ("vnstrategy: buffer %p too small for physio", bp)); 416 417 if ((bp->b_flags & B_FREEBUF) && TESTOPT(vn, VN_RESERVE)) { 418 biodone(bp); 419 } else { 420 vm_pager_strategy(vn->sc_object, bp); 421 } 422 } else { 423 bp->b_flags |= B_ERROR; 424 bp->b_error = EINVAL; 425 biodone(bp); 426 } 427 } 428 429 /* ARGSUSED */ 430 static int 431 vnioctl(dev_t dev, u_long cmd, caddr_t data, int flag, struct thread *td) 432 { 433 struct vn_softc *vn; 434 struct vn_ioctl *vio; 435 int error; 436 u_long *f; 437 438 vn = dev->si_drv1; 439 IFOPT(vn,VN_FOLLOW) 440 printf("vnioctl(%s, 0x%lx, %p, 0x%x, %p): unit %d\n", 441 devtoname(dev), cmd, (void *)data, flag, (void *)td, 442 dkunit(dev)); 443 444 switch (cmd) { 445 case VNIOCATTACH: 446 case VNIOCDETACH: 447 case VNIOCGSET: 448 case VNIOCGCLEAR: 449 case VNIOCUSET: 450 case VNIOCUCLEAR: 451 goto vn_specific; 452 } 453 454 IFOPT(vn,VN_LABELS) { 455 if (vn->sc_slices != NULL) { 456 error = dsioctl(dev, cmd, data, flag, &vn->sc_slices); 457 if (error != ENOIOCTL) 458 return (error); 459 } 460 if (dkslice(dev) != WHOLE_DISK_SLICE || 461 dkpart(dev) != RAW_PART) 462 return (ENOTTY); 463 } 464 465 vn_specific: 466 467 error = suser(td); 468 if (error) 469 return (error); 470 471 vio = (struct vn_ioctl *)data; 472 f = (u_long*)data; 473 switch (cmd) { 474 475 case VNIOCATTACH: 476 if (vn->sc_flags & VNF_INITED) 477 return(EBUSY); 478 479 if (vio->vn_file == NULL) 480 error = vniocattach_swap(vn, vio, dev, flag, td); 481 else 482 error = vniocattach_file(vn, vio, dev, flag, td); 483 break; 484 485 case VNIOCDETACH: 486 if ((vn->sc_flags & VNF_INITED) == 0) 487 return(ENXIO); 488 /* 489 * XXX handle i/o in progress. Return EBUSY, or wait, or 490 * flush the i/o. 491 * XXX handle multiple opens of the device. Return EBUSY, 492 * or revoke the fd's. 493 * How are these problems handled for removable and failing 494 * hardware devices? (Hint: They are not) 495 */ 496 vnclear(vn); 497 IFOPT(vn, VN_FOLLOW) 498 printf("vnioctl: CLRed\n"); 499 break; 500 501 case VNIOCGSET: 502 vn_options |= *f; 503 *f = vn_options; 504 break; 505 506 case VNIOCGCLEAR: 507 vn_options &= ~(*f); 508 *f = vn_options; 509 break; 510 511 case VNIOCUSET: 512 vn->sc_options |= *f; 513 *f = vn->sc_options; 514 break; 515 516 case VNIOCUCLEAR: 517 vn->sc_options &= ~(*f); 518 *f = vn->sc_options; 519 break; 520 521 default: 522 error = ENOTTY; 523 break; 524 } 525 return(error); 526 } 527 528 /* 529 * vniocattach_file: 530 * 531 * Attach a file to a VN partition. Return the size in the vn_size 532 * field. 533 */ 534 535 static int 536 vniocattach_file(vn, vio, dev, flag, td) 537 struct vn_softc *vn; 538 struct vn_ioctl *vio; 539 dev_t dev; 540 int flag; 541 struct thread *td; 542 { 543 struct vattr vattr; 544 struct nameidata nd; 545 int error, flags; 546 struct proc *p = td->td_proc; 547 548 KKASSERT(p != NULL); 549 550 flags = FREAD|FWRITE; 551 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, vio->vn_file, td); 552 error = vn_open(&nd, flags, 0); 553 if (error) { 554 if (error != EACCES && error != EPERM && error != EROFS) 555 return (error); 556 flags &= ~FWRITE; 557 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, 558 UIO_USERSPACE, vio->vn_file, td); 559 error = vn_open(&nd, flags, 0); 560 if (error) 561 return (error); 562 } 563 NDFREE(&nd, NDF_ONLY_PNBUF); 564 if (nd.ni_vp->v_type != VREG || 565 (error = VOP_GETATTR(nd.ni_vp, &vattr, td))) { 566 VOP_UNLOCK(nd.ni_vp, NULL, 0, td); 567 (void) vn_close(nd.ni_vp, flags, td); 568 return (error ? error : EINVAL); 569 } 570 VOP_UNLOCK(nd.ni_vp, NULL, 0, td); 571 vn->sc_secsize = DEV_BSIZE; 572 vn->sc_vp = nd.ni_vp; 573 574 /* 575 * If the size is specified, override the file attributes. Note that 576 * the vn_size argument is in PAGE_SIZE sized blocks. 577 */ 578 if (vio->vn_size) 579 vn->sc_size = (quad_t)vio->vn_size * PAGE_SIZE / vn->sc_secsize; 580 else 581 vn->sc_size = vattr.va_size / vn->sc_secsize; 582 error = vnsetcred(vn, p->p_ucred); 583 if (error) { 584 (void) vn_close(nd.ni_vp, flags, td); 585 return(error); 586 } 587 vn->sc_flags |= VNF_INITED; 588 if (flags == FREAD) 589 vn->sc_flags |= VNF_READONLY; 590 IFOPT(vn, VN_LABELS) { 591 /* 592 * Reopen so that `ds' knows which devices are open. 593 * If this is the first VNIOCSET, then we've 594 * guaranteed that the device is the cdev and that 595 * no other slices or labels are open. Otherwise, 596 * we rely on VNIOCCLR not being abused. 597 */ 598 error = vnopen(dev, flag, S_IFCHR, td); 599 if (error) 600 vnclear(vn); 601 } 602 IFOPT(vn, VN_FOLLOW) 603 printf("vnioctl: SET vp %p size %x blks\n", 604 vn->sc_vp, vn->sc_size); 605 return(0); 606 } 607 608 /* 609 * vniocattach_swap: 610 * 611 * Attach swap backing store to a VN partition of the size specified 612 * in vn_size. 613 */ 614 615 static int 616 vniocattach_swap(vn, vio, dev, flag, td) 617 struct vn_softc *vn; 618 struct vn_ioctl *vio; 619 dev_t dev; 620 int flag; 621 struct thread *td; 622 { 623 int error; 624 struct proc *p = td->td_proc; 625 626 KKASSERT(p != NULL); 627 /* 628 * Range check. Disallow negative sizes or any size less then the 629 * size of a page. Then round to a page. 630 */ 631 632 if (vio->vn_size <= 0) 633 return(EDOM); 634 635 /* 636 * Allocate an OBJT_SWAP object. 637 * 638 * sc_secsize is PAGE_SIZE'd 639 * 640 * vio->vn_size is in PAGE_SIZE'd chunks. 641 * sc_size must be in PAGE_SIZE'd chunks. 642 * Note the truncation. 643 */ 644 645 vn->sc_secsize = PAGE_SIZE; 646 vn->sc_size = vio->vn_size; 647 vn->sc_object = 648 vm_pager_allocate(OBJT_SWAP, NULL, vn->sc_secsize * (vm_ooffset_t)vio->vn_size, VM_PROT_DEFAULT, 0); 649 IFOPT(vn, VN_RESERVE) { 650 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) { 651 vm_pager_deallocate(vn->sc_object); 652 vn->sc_object = NULL; 653 return(EDOM); 654 } 655 } 656 vn->sc_flags |= VNF_INITED; 657 658 error = vnsetcred(vn, p->p_ucred); 659 if (error == 0) { 660 IFOPT(vn, VN_LABELS) { 661 /* 662 * Reopen so that `ds' knows which devices are open. 663 * If this is the first VNIOCSET, then we've 664 * guaranteed that the device is the cdev and that 665 * no other slices or labels are open. Otherwise, 666 * we rely on VNIOCCLR not being abused. 667 */ 668 error = vnopen(dev, flag, S_IFCHR, td); 669 } 670 } 671 if (error == 0) { 672 IFOPT(vn, VN_FOLLOW) { 673 printf("vnioctl: SET vp %p size %x\n", 674 vn->sc_vp, vn->sc_size); 675 } 676 } 677 if (error) 678 vnclear(vn); 679 return(error); 680 } 681 682 /* 683 * Duplicate the current processes' credentials. Since we are called only 684 * as the result of a SET ioctl and only root can do that, any future access 685 * to this "disk" is essentially as root. Note that credentials may change 686 * if some other uid can write directly to the mapped file (NFS). 687 */ 688 int 689 vnsetcred(struct vn_softc *vn, struct ucred *cred) 690 { 691 char *tmpbuf; 692 int error = 0; 693 694 /* 695 * Set credits in our softc 696 */ 697 698 if (vn->sc_cred) 699 crfree(vn->sc_cred); 700 vn->sc_cred = crdup(cred); 701 702 /* 703 * Horrible kludge to establish credentials for NFS XXX. 704 */ 705 706 if (vn->sc_vp) { 707 struct uio auio; 708 struct iovec aiov; 709 710 tmpbuf = malloc(vn->sc_secsize, M_TEMP, M_WAITOK); 711 bzero(&auio, sizeof(auio)); 712 713 aiov.iov_base = tmpbuf; 714 aiov.iov_len = vn->sc_secsize; 715 auio.uio_iov = &aiov; 716 auio.uio_iovcnt = 1; 717 auio.uio_offset = 0; 718 auio.uio_rw = UIO_READ; 719 auio.uio_segflg = UIO_SYSSPACE; 720 auio.uio_resid = aiov.iov_len; 721 vn_lock(vn->sc_vp, NULL, LK_EXCLUSIVE | LK_RETRY, curthread); 722 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred); 723 VOP_UNLOCK(vn->sc_vp, NULL, 0, curthread); 724 free(tmpbuf, M_TEMP); 725 } 726 return (error); 727 } 728 729 void 730 vnclear(struct vn_softc *vn) 731 { 732 struct thread *td = curthread; /* XXX */ 733 734 IFOPT(vn, VN_FOLLOW) 735 printf("vnclear(%p): vp=%p\n", vn, vn->sc_vp); 736 if (vn->sc_slices != NULL) 737 dsgone(&vn->sc_slices); 738 vn->sc_flags &= ~VNF_INITED; 739 if (vn->sc_vp != NULL) { 740 (void)vn_close(vn->sc_vp, vn->sc_flags & VNF_READONLY ? 741 FREAD : (FREAD|FWRITE), td); 742 vn->sc_vp = NULL; 743 } 744 vn->sc_flags &= ~VNF_READONLY; 745 if (vn->sc_cred) { 746 crfree(vn->sc_cred); 747 vn->sc_cred = NULL; 748 } 749 if (vn->sc_object != NULL) { 750 vm_pager_deallocate(vn->sc_object); 751 vn->sc_object = NULL; 752 } 753 vn->sc_size = 0; 754 } 755 756 static int 757 vnsize(dev_t dev) 758 { 759 struct vn_softc *vn; 760 761 vn = dev->si_drv1; 762 if (!vn) 763 return(-1); 764 if ((vn->sc_flags & VNF_INITED) == 0) 765 return(-1); 766 767 return(vn->sc_size); 768 } 769 770 static int 771 vn_modevent(module_t mod, int type, void *data) 772 { 773 struct vn_softc *vn; 774 dev_t dev; 775 776 switch (type) { 777 case MOD_LOAD: 778 cdevsw_add(&vn_cdevsw, 0, 0); 779 break; 780 case MOD_UNLOAD: 781 /* fall through */ 782 case MOD_SHUTDOWN: 783 while ((vn = SLIST_FIRST(&vn_list)) != NULL) { 784 SLIST_REMOVE_HEAD(&vn_list, sc_list); 785 if (vn->sc_flags & VNF_INITED) 786 vnclear(vn); 787 /* Cleanup all dev_t's that refer to this unit */ 788 while ((dev = vn->sc_devlist) != NULL) { 789 vn->sc_devlist = dev->si_drv2; 790 dev->si_drv1 = dev->si_drv2 = NULL; 791 destroy_dev(dev); 792 } 793 free(vn, M_DEVBUF); 794 } 795 cdevsw_remove(&vn_cdevsw, 0, 0); 796 break; 797 default: 798 break; 799 } 800 return 0; 801 } 802 803 DEV_MODULE(vn, vn_modevent, 0); 804