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 1.38 2008/07/01 02:02:53 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: There is a security issue involved with this driver. 52 * Once mounted all access to the contents of the "mapped" file via 53 * the special file is controlled by the permissions on the special 54 * file, the protection of the mapped file is ignored (effectively, 55 * by using root credentials in all transactions). 56 * 57 * NOTE 2: Doesn't interact with leases, should it? 58 */ 59 60 #include "use_vn.h" 61 #include <sys/param.h> 62 #include <sys/systm.h> 63 #include <sys/kernel.h> 64 #include <sys/proc.h> 65 #include <sys/priv.h> 66 #include <sys/nlookup.h> 67 #include <sys/buf.h> 68 #include <sys/malloc.h> 69 #include <sys/mount.h> 70 #include <sys/vnode.h> 71 #include <sys/fcntl.h> 72 #include <sys/conf.h> 73 #include <sys/diskslice.h> 74 #include <sys/disk.h> 75 #include <sys/stat.h> 76 #include <sys/module.h> 77 #include <sys/vnioctl.h> 78 79 #include <vm/vm.h> 80 #include <vm/vm_object.h> 81 #include <vm/vm_page.h> 82 #include <vm/vm_pager.h> 83 #include <vm/vm_pageout.h> 84 #include <vm/swap_pager.h> 85 #include <vm/vm_extern.h> 86 #include <vm/vm_zone.h> 87 #include <sys/devfs.h> 88 89 static d_ioctl_t vnioctl; 90 static d_open_t vnopen; 91 static d_close_t vnclose; 92 static d_psize_t vnsize; 93 static d_strategy_t vnstrategy; 94 static d_clone_t vnclone; 95 96 DEVFS_DECLARE_CLONE_BITMAP(vn); 97 98 #if NVN <= 1 99 #define VN_PREALLOCATED_UNITS 4 100 #else 101 #define VN_PREALLOCATED_UNITS NVN 102 #endif 103 104 #define CDEV_MAJOR 43 105 106 #define VN_BSIZE_BEST 8192 107 108 /* 109 * dev_ops 110 * D_DISK we want to look like a disk 111 * D_CANFREE We support BUF_CMD_FREEBLKS 112 */ 113 114 static struct dev_ops vn_ops = { 115 { "vn", CDEV_MAJOR, D_DISK | D_CANFREE }, 116 .d_open = vnopen, 117 .d_close = vnclose, 118 .d_read = physread, 119 .d_write = physwrite, 120 .d_ioctl = vnioctl, 121 .d_strategy = vnstrategy, 122 .d_psize = vnsize 123 }; 124 125 struct vn_softc { 126 int sc_unit; 127 int sc_flags; /* flags */ 128 u_int64_t sc_size; /* size of vn, sc_secsize scale */ 129 int sc_secsize; /* sector size */ 130 struct disk sc_disk; 131 struct vnode *sc_vp; /* vnode if not NULL */ 132 vm_object_t sc_object; /* backing object if not NULL */ 133 struct ucred *sc_cred; /* credentials */ 134 int sc_maxactive; /* max # of active requests */ 135 struct buf sc_tab; /* transfer queue */ 136 u_long sc_options; /* options */ 137 cdev_t sc_devlist; /* devices that refer to this unit */ 138 SLIST_ENTRY(vn_softc) sc_list; 139 }; 140 141 static SLIST_HEAD(, vn_softc) vn_list; 142 143 /* sc_flags */ 144 #define VNF_INITED 0x01 145 #define VNF_READONLY 0x02 146 147 static u_long vn_options; 148 149 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt)) 150 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt)) 151 152 static int vnsetcred (struct vn_softc *vn, struct ucred *cred); 153 static void vnclear (struct vn_softc *vn); 154 static int vnget (cdev_t dev, struct vn_softc *vn , struct vn_user *vnu); 155 static int vn_modevent (module_t, int, void *); 156 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred); 157 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred); 158 static cdev_t vn_create(int unit, struct devfs_bitmap *bitmap); 159 160 static int 161 vnclone(struct dev_clone_args *ap) 162 { 163 int unit; 164 165 unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(vn), 0); 166 ap->a_dev = vn_create(unit, &DEVFS_CLONE_BITMAP(vn)); 167 168 return 0; 169 } 170 171 static int 172 vnclose(struct dev_close_args *ap) 173 { 174 return (0); 175 } 176 177 static struct vn_softc * 178 vncreatevn(void) 179 { 180 struct vn_softc *vn; 181 182 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO); 183 return vn; 184 } 185 186 static void 187 vninitvn(struct vn_softc *vn, cdev_t dev) 188 { 189 int unit; 190 191 KKASSERT(vn != NULL); 192 KKASSERT(dev != NULL); 193 unit = dkunit(dev); 194 195 vn->sc_unit = unit; 196 dev->si_drv1 = vn; 197 vn->sc_devlist = dev; 198 if (vn->sc_devlist->si_drv1 == NULL) { 199 reference_dev(vn->sc_devlist); 200 vn->sc_devlist->si_drv1 = vn; 201 vn->sc_devlist->si_drv2 = NULL; 202 } 203 if (vn->sc_devlist != dev) { 204 dev->si_drv1 = vn; 205 dev->si_drv2 = vn->sc_devlist; 206 vn->sc_devlist = dev; 207 reference_dev(dev); 208 } 209 SLIST_INSERT_HEAD(&vn_list, vn, sc_list); 210 } 211 212 /* 213 * Called only when si_drv1 is NULL. Locate the associated vn node and 214 * attach the device to it. 215 */ 216 static struct vn_softc * 217 vnfindvn(cdev_t dev) 218 { 219 int unit; 220 struct vn_softc *vn; 221 222 unit = dkunit(dev); 223 SLIST_FOREACH(vn, &vn_list, sc_list) { 224 if (vn->sc_unit == unit) { 225 dev->si_drv1 = vn; 226 dev->si_drv2 = vn->sc_devlist; 227 vn->sc_devlist = dev; 228 reference_dev(dev); 229 break; 230 } 231 } 232 233 KKASSERT(vn != NULL); 234 235 return vn; 236 } 237 238 static int 239 vnopen(struct dev_open_args *ap) 240 { 241 cdev_t dev = ap->a_head.a_dev; 242 struct vn_softc *vn; 243 244 /* 245 * Locate preexisting device 246 */ 247 248 if ((vn = dev->si_drv1) == NULL) 249 vn = vnfindvn(dev); 250 251 /* 252 * Update si_bsize fields for device. This data will be overriden by 253 * the slice/parition code for vn accesses through partitions, and 254 * used directly if you open the 'whole disk' device. 255 * 256 * si_bsize_best must be reinitialized in case VN has been 257 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency. 258 */ 259 dev->si_bsize_phys = vn->sc_secsize; 260 dev->si_bsize_best = vn->sc_secsize; 261 if (dev->si_bsize_best < VN_BSIZE_BEST) 262 dev->si_bsize_best = VN_BSIZE_BEST; 263 264 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY)) 265 return (EACCES); 266 267 IFOPT(vn, VN_FOLLOW) 268 kprintf("vnopen(%s, 0x%x, 0x%x)\n", 269 devtoname(dev), ap->a_oflags, ap->a_devtype); 270 271 return(0); 272 } 273 274 /* 275 * vnstrategy: 276 * 277 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls 278 * for vnode-backed vn's, and the new vm_pager_strategy() call for 279 * vm_object-backed vn's. 280 */ 281 static int 282 vnstrategy(struct dev_strategy_args *ap) 283 { 284 cdev_t dev = ap->a_head.a_dev; 285 struct bio *bio = ap->a_bio; 286 struct buf *bp; 287 struct bio *nbio; 288 int unit; 289 struct vn_softc *vn; 290 int error; 291 292 unit = dkunit(dev); 293 if ((vn = dev->si_drv1) == NULL) 294 vn = vnfindvn(dev); 295 296 bp = bio->bio_buf; 297 298 IFOPT(vn, VN_DEBUG) 299 kprintf("vnstrategy(%p): unit %d\n", bp, unit); 300 301 if ((vn->sc_flags & VNF_INITED) == 0) { 302 bp->b_error = ENXIO; 303 bp->b_flags |= B_ERROR; 304 biodone(bio); 305 return(0); 306 } 307 308 bp->b_resid = bp->b_bcount; 309 310 /* 311 * The vnode device is using disk/slice label support. 312 * 313 * The dscheck() function is called for validating the 314 * slices that exist ON the vnode device itself, and 315 * translate the "slice-relative" block number, again. 316 * dscheck() will call biodone() and return NULL if 317 * we are at EOF or beyond the device size. 318 */ 319 320 nbio = bio; 321 322 /* 323 * Use the translated nbio from this point on 324 */ 325 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) { 326 /* 327 * Freeblks is not handled for vnode-backed elements yet. 328 */ 329 bp->b_resid = 0; 330 /* operation complete */ 331 } else if (vn->sc_vp) { 332 /* 333 * VNODE I/O 334 * 335 * If an error occurs, we set B_ERROR but we do not set 336 * B_INVAL because (for a write anyway), the buffer is 337 * still valid. 338 */ 339 struct uio auio; 340 struct iovec aiov; 341 342 bzero(&auio, sizeof(auio)); 343 344 aiov.iov_base = bp->b_data; 345 aiov.iov_len = bp->b_bcount; 346 auio.uio_iov = &aiov; 347 auio.uio_iovcnt = 1; 348 auio.uio_offset = nbio->bio_offset; 349 auio.uio_segflg = UIO_SYSSPACE; 350 if (bp->b_cmd == BUF_CMD_READ) 351 auio.uio_rw = UIO_READ; 352 else 353 auio.uio_rw = UIO_WRITE; 354 auio.uio_resid = bp->b_bcount; 355 auio.uio_td = curthread; 356 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY); 357 if (bp->b_cmd == BUF_CMD_READ) 358 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT | IO_RECURSE, vn->sc_cred); 359 else 360 error = VOP_WRITE(vn->sc_vp, &auio, IO_DIRECT | IO_RECURSE, vn->sc_cred); 361 vn_unlock(vn->sc_vp); 362 bp->b_resid = auio.uio_resid; 363 if (error) { 364 bp->b_error = error; 365 bp->b_flags |= B_ERROR; 366 } 367 /* operation complete */ 368 } else if (vn->sc_object) { 369 /* 370 * OBJT_SWAP I/O (handles read, write, freebuf) 371 * 372 * We have nothing to do if freeing blocks on a reserved 373 * swap area, othrewise execute the op. 374 */ 375 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) { 376 bp->b_resid = 0; 377 /* operation complete */ 378 } else { 379 vm_pager_strategy(vn->sc_object, nbio); 380 return(0); 381 /* NOT REACHED */ 382 } 383 } else { 384 bp->b_resid = bp->b_bcount; 385 bp->b_flags |= B_ERROR | B_INVAL; 386 bp->b_error = EINVAL; 387 /* operation complete */ 388 } 389 biodone(nbio); 390 return(0); 391 } 392 393 /* ARGSUSED */ 394 static int 395 vnioctl(struct dev_ioctl_args *ap) 396 { 397 cdev_t dev = ap->a_head.a_dev; 398 struct vn_softc *vn; 399 struct vn_ioctl *vio; 400 int error; 401 u_long *f; 402 403 vn = dev->si_drv1; 404 IFOPT(vn,VN_FOLLOW) { 405 kprintf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n", 406 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag, 407 dkunit(dev)); 408 } 409 410 switch (ap->a_cmd) { 411 case VNIOCATTACH: 412 case VNIOCDETACH: 413 case VNIOCGSET: 414 case VNIOCGCLEAR: 415 case VNIOCGET: 416 case VNIOCUSET: 417 case VNIOCUCLEAR: 418 goto vn_specific; 419 } 420 421 #if 0 422 if (dkslice(dev) != WHOLE_DISK_SLICE || 423 dkpart(dev) != WHOLE_SLICE_PART) 424 return (ENOTTY); 425 #endif 426 427 vn_specific: 428 429 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0); 430 if (error) 431 return (error); 432 433 vio = (struct vn_ioctl *)ap->a_data; 434 f = (u_long*)ap->a_data; 435 436 switch (ap->a_cmd) { 437 case VNIOCATTACH: 438 if (vn->sc_flags & VNF_INITED) 439 return(EBUSY); 440 441 if (vio->vn_file == NULL) 442 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred); 443 else 444 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred); 445 break; 446 447 case VNIOCDETACH: 448 if ((vn->sc_flags & VNF_INITED) == 0) 449 return(ENXIO); 450 /* 451 * XXX handle i/o in progress. Return EBUSY, or wait, or 452 * flush the i/o. 453 * XXX handle multiple opens of the device. Return EBUSY, 454 * or revoke the fd's. 455 * How are these problems handled for removable and failing 456 * hardware devices? (Hint: They are not) 457 */ 458 if (count_dev(vn->sc_devlist) > 1) 459 return (EBUSY); 460 461 vnclear(vn); 462 IFOPT(vn, VN_FOLLOW) 463 kprintf("vnioctl: CLRed\n"); 464 465 if (dkunit(dev) >= VN_PREALLOCATED_UNITS) { 466 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(vn), dkunit(dev)); 467 disk_destroy(&vn->sc_disk); 468 } 469 470 break; 471 472 case VNIOCGET: 473 error = vnget(dev, vn, (struct vn_user *) ap->a_data); 474 break; 475 476 case VNIOCGSET: 477 vn_options |= *f; 478 *f = vn_options; 479 break; 480 481 case VNIOCGCLEAR: 482 vn_options &= ~(*f); 483 *f = vn_options; 484 break; 485 486 case VNIOCUSET: 487 vn->sc_options |= *f; 488 *f = vn->sc_options; 489 break; 490 491 case VNIOCUCLEAR: 492 vn->sc_options &= ~(*f); 493 *f = vn->sc_options; 494 break; 495 496 default: 497 error = ENOTTY; 498 break; 499 } 500 return(error); 501 } 502 503 /* 504 * vniocattach_file: 505 * 506 * Attach a file to a VN partition. Return the size in the vn_size 507 * field. 508 */ 509 510 static int 511 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev, 512 int flag, struct ucred *cred) 513 { 514 struct vattr vattr; 515 struct nlookupdata nd; 516 int error, flags; 517 struct vnode *vp; 518 struct disk_info info; 519 520 flags = FREAD|FWRITE; 521 error = nlookup_init(&nd, vio->vn_file, 522 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP); 523 if (error) 524 return (error); 525 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) { 526 if (error != EACCES && error != EPERM && error != EROFS) 527 goto done; 528 flags &= ~FWRITE; 529 nlookup_done(&nd); 530 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP); 531 if (error) 532 return (error); 533 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) 534 goto done; 535 } 536 vp = nd.nl_open_vp; 537 if (vp->v_type != VREG || 538 (error = VOP_GETATTR(vp, &vattr))) { 539 if (error == 0) 540 error = EINVAL; 541 goto done; 542 } 543 vn_unlock(vp); 544 vn->sc_secsize = DEV_BSIZE; 545 vn->sc_vp = vp; 546 nd.nl_open_vp = NULL; 547 548 /* 549 * If the size is specified, override the file attributes. Note that 550 * the vn_size argument is in PAGE_SIZE sized blocks. 551 */ 552 if (vio->vn_size) 553 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize; 554 else 555 vn->sc_size = vattr.va_size / vn->sc_secsize; 556 error = vnsetcred(vn, cred); 557 if (error) { 558 vn->sc_vp = NULL; 559 vn_close(vp, flags); 560 goto done; 561 } 562 vn->sc_flags |= VNF_INITED; 563 if (flags == FREAD) 564 vn->sc_flags |= VNF_READONLY; 565 566 /* 567 * Set the disk info so that probing is triggered 568 */ 569 bzero(&info, sizeof(struct disk_info)); 570 info.d_media_blksize = vn->sc_secsize; 571 info.d_media_blocks = vn->sc_size; 572 /* 573 * reserve mbr sector for backwards compatibility 574 * when no slices exist. 575 */ 576 info.d_dsflags = DSO_COMPATMBR; 577 info.d_secpertrack = 32; 578 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE); 579 info.d_secpercyl = info.d_secpertrack * info.d_nheads; 580 info.d_ncylinders = vn->sc_size / info.d_secpercyl; 581 disk_setdiskinfo_sync(&vn->sc_disk, &info); 582 583 error = dev_dopen(dev, flag, S_IFCHR, cred); 584 if (error) 585 vnclear(vn); 586 587 IFOPT(vn, VN_FOLLOW) 588 kprintf("vnioctl: SET vp %p size %llx blks\n", 589 vn->sc_vp, (long long)vn->sc_size); 590 done: 591 nlookup_done(&nd); 592 return(error); 593 } 594 595 /* 596 * vniocattach_swap: 597 * 598 * Attach swap backing store to a VN partition of the size specified 599 * in vn_size. 600 */ 601 602 static int 603 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev, 604 int flag, struct ucred *cred) 605 { 606 int error; 607 struct disk_info info; 608 609 /* 610 * Range check. Disallow negative sizes or any size less then the 611 * size of a page. Then round to a page. 612 */ 613 614 if (vio->vn_size <= 0) 615 return(EDOM); 616 617 /* 618 * Allocate an OBJT_SWAP object. 619 * 620 * sc_secsize is PAGE_SIZE'd 621 * 622 * vio->vn_size is in PAGE_SIZE'd chunks. 623 * sc_size must be in PAGE_SIZE'd chunks. 624 * Note the truncation. 625 */ 626 627 vn->sc_secsize = PAGE_SIZE; 628 vn->sc_size = vio->vn_size; 629 vn->sc_object = vm_pager_allocate(OBJT_SWAP, NULL, 630 vn->sc_secsize * (off_t)vio->vn_size, 631 VM_PROT_DEFAULT, 0); 632 IFOPT(vn, VN_RESERVE) { 633 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) { 634 vm_pager_deallocate(vn->sc_object); 635 vn->sc_object = NULL; 636 return(EDOM); 637 } 638 } 639 vn->sc_flags |= VNF_INITED; 640 641 error = vnsetcred(vn, cred); 642 if (error == 0) { 643 /* 644 * Set the disk info so that probing is triggered 645 */ 646 bzero(&info, sizeof(struct disk_info)); 647 info.d_media_blksize = vn->sc_secsize; 648 info.d_media_blocks = vn->sc_size; 649 /* 650 * reserve mbr sector for backwards compatibility 651 * when no slices exist. 652 */ 653 info.d_dsflags = DSO_COMPATMBR; 654 info.d_secpertrack = 32; 655 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE); 656 info.d_secpercyl = info.d_secpertrack * info.d_nheads; 657 info.d_ncylinders = vn->sc_size / info.d_secpercyl; 658 disk_setdiskinfo_sync(&vn->sc_disk, &info); 659 660 error = dev_dopen(dev, flag, S_IFCHR, cred); 661 } 662 if (error == 0) { 663 IFOPT(vn, VN_FOLLOW) { 664 kprintf("vnioctl: SET vp %p size %llx\n", 665 vn->sc_vp, (long long)vn->sc_size); 666 } 667 } 668 if (error) 669 vnclear(vn); 670 return(error); 671 } 672 673 /* 674 * Duplicate the current processes' credentials. Since we are called only 675 * as the result of a SET ioctl and only root can do that, any future access 676 * to this "disk" is essentially as root. Note that credentials may change 677 * if some other uid can write directly to the mapped file (NFS). 678 */ 679 int 680 vnsetcred(struct vn_softc *vn, struct ucred *cred) 681 { 682 char *tmpbuf; 683 int error = 0; 684 685 /* 686 * Set credits in our softc 687 */ 688 689 if (vn->sc_cred) 690 crfree(vn->sc_cred); 691 vn->sc_cred = crdup(cred); 692 693 /* 694 * Horrible kludge to establish credentials for NFS XXX. 695 */ 696 697 if (vn->sc_vp) { 698 struct uio auio; 699 struct iovec aiov; 700 701 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK); 702 bzero(&auio, sizeof(auio)); 703 704 aiov.iov_base = tmpbuf; 705 aiov.iov_len = vn->sc_secsize; 706 auio.uio_iov = &aiov; 707 auio.uio_iovcnt = 1; 708 auio.uio_offset = 0; 709 auio.uio_rw = UIO_READ; 710 auio.uio_segflg = UIO_SYSSPACE; 711 auio.uio_resid = aiov.iov_len; 712 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY); 713 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred); 714 vn_unlock(vn->sc_vp); 715 kfree(tmpbuf, M_TEMP); 716 } 717 return (error); 718 } 719 720 void 721 vnclear(struct vn_softc *vn) 722 { 723 IFOPT(vn, VN_FOLLOW) 724 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp); 725 vn->sc_flags &= ~VNF_INITED; 726 if (vn->sc_vp != NULL) { 727 vn_close(vn->sc_vp, 728 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE)); 729 vn->sc_vp = NULL; 730 } 731 vn->sc_flags &= ~VNF_READONLY; 732 if (vn->sc_cred) { 733 crfree(vn->sc_cred); 734 vn->sc_cred = NULL; 735 } 736 if (vn->sc_object != NULL) { 737 vm_pager_deallocate(vn->sc_object); 738 vn->sc_object = NULL; 739 } 740 741 disk_unprobe(&vn->sc_disk); 742 743 vn->sc_size = 0; 744 } 745 746 /* 747 * vnget: 748 * 749 * populate a struct vn_user for the VNIOCGET ioctl. 750 * interface conventions defined in sys/sys/vnioctl.h. 751 */ 752 753 static int 754 vnget(cdev_t dev, struct vn_softc *vn, struct vn_user *vnu) 755 { 756 int error, found = 0; 757 char *freepath, *fullpath; 758 struct vattr vattr; 759 760 if (vnu->vnu_unit == -1) { 761 vnu->vnu_unit = dkunit(dev); 762 } 763 else if (vnu->vnu_unit < 0) 764 return (EINVAL); 765 766 SLIST_FOREACH(vn, &vn_list, sc_list) { 767 768 if(vn->sc_unit != vnu->vnu_unit) 769 continue; 770 771 found = 1; 772 773 if (vn->sc_flags & VNF_INITED && vn->sc_vp != NULL) { 774 775 /* note: u_cred checked in vnioctl above */ 776 error = VOP_GETATTR(vn->sc_vp, &vattr); 777 if (error) { 778 kprintf("vnget: VOP_GETATTR for %p failed\n", 779 vn->sc_vp); 780 return (error); 781 } 782 783 error = vn_fullpath(curproc, vn->sc_vp, 784 &fullpath, &freepath); 785 786 if (error) { 787 kprintf("vnget: unable to resolve vp %p\n", 788 vn->sc_vp); 789 return(error); 790 } 791 792 strlcpy(vnu->vnu_file, fullpath, 793 sizeof(vnu->vnu_file)); 794 kfree(freepath, M_TEMP); 795 vnu->vnu_dev = vattr.va_fsid; 796 vnu->vnu_ino = vattr.va_fileid; 797 798 } 799 else if (vn->sc_flags & VNF_INITED && vn->sc_object != NULL){ 800 801 strlcpy(vnu->vnu_file, _VN_USER_SWAP, 802 sizeof(vnu->vnu_file)); 803 vnu->vnu_size = vn->sc_size; 804 vnu->vnu_secsize = vn->sc_secsize; 805 806 } else { 807 808 bzero(vnu->vnu_file, sizeof(vnu->vnu_file)); 809 vnu->vnu_dev = 0; 810 vnu->vnu_ino = 0; 811 812 } 813 break; 814 } 815 816 if (!found) 817 return(ENXIO); 818 819 return(0); 820 } 821 822 static int 823 vnsize(struct dev_psize_args *ap) 824 { 825 cdev_t dev = ap->a_head.a_dev; 826 struct vn_softc *vn; 827 828 vn = dev->si_drv1; 829 if (!vn) 830 return(ENXIO); 831 if ((vn->sc_flags & VNF_INITED) == 0) 832 return(ENXIO); 833 ap->a_result = (int64_t)vn->sc_size; 834 return(0); 835 } 836 837 static cdev_t 838 vn_create(int unit, struct devfs_bitmap *bitmap) 839 { 840 struct vn_softc *vn; 841 struct disk_info info; 842 cdev_t dev; 843 844 vn = vncreatevn(); 845 dev = disk_create(unit, &vn->sc_disk, &vn_ops); 846 vninitvn(vn, dev); 847 848 bzero(&info, sizeof(struct disk_info)); 849 info.d_media_blksize = 512; 850 info.d_media_blocks = 0; 851 info.d_dsflags = DSO_MBRQUIET; 852 info.d_secpertrack = 32; 853 info.d_nheads = 64; 854 info.d_secpercyl = info.d_secpertrack * info.d_nheads; 855 info.d_ncylinders = 0; 856 disk_setdiskinfo_sync(&vn->sc_disk, &info); 857 858 if (bitmap != NULL) 859 devfs_clone_bitmap_set(bitmap, unit); 860 861 return dev; 862 } 863 864 static int 865 vn_modevent(module_t mod, int type, void *data) 866 { 867 struct vn_softc *vn; 868 struct disk_info info; 869 static cdev_t dev = NULL; 870 int i; 871 872 switch (type) { 873 case MOD_LOAD: 874 dev = make_autoclone_dev(&vn_ops, &DEVFS_CLONE_BITMAP(vn), vnclone, UID_ROOT, 875 GID_OPERATOR, 0640, "vn"); 876 877 for (i = 0; i < VN_PREALLOCATED_UNITS; i++) { 878 vn_create(i, &DEVFS_CLONE_BITMAP(vn)); 879 } 880 break; 881 case MOD_UNLOAD: 882 destroy_autoclone_dev(dev, &DEVFS_CLONE_BITMAP(vn)); 883 /* fall through */ 884 case MOD_SHUTDOWN: 885 while ((vn = SLIST_FIRST(&vn_list)) != NULL) { 886 SLIST_REMOVE_HEAD(&vn_list, sc_list); 887 if (vn->sc_flags & VNF_INITED) 888 vnclear(vn); 889 /* Cleanup all cdev_t's that refer to this unit */ 890 while ((dev = vn->sc_devlist) != NULL) { 891 vn->sc_devlist = dev->si_drv2; 892 dev->si_drv1 = dev->si_drv2 = NULL; 893 destroy_dev(dev); 894 } 895 kfree(vn, M_DEVBUF); 896 } 897 dev_ops_remove_all(&vn_ops); 898 break; 899 default: 900 break; 901 } 902 return 0; 903 } 904 905 DEV_MODULE(vn, vn_modevent, 0); 906