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 */ 43 44 /* 45 * Vnode disk driver. 46 * 47 * Block/character interface to a vnode. Allows one to treat a file 48 * as a disk (e.g. build a filesystem in it, mount it, etc.). 49 * 50 * NOTE 1: There is a security issue involved with this driver. 51 * Once mounted all access to the contents of the "mapped" file via 52 * the special file is controlled by the permissions on the special 53 * file, the protection of the mapped file is ignored (effectively, 54 * by using root credentials in all transactions). 55 * 56 * NOTE 2: Doesn't interact with leases, should it? 57 */ 58 59 #include "use_vn.h" 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/kernel.h> 63 #include <sys/proc.h> 64 #include <sys/priv.h> 65 #include <sys/nlookup.h> 66 #include <sys/buf.h> 67 #include <sys/malloc.h> 68 #include <sys/mount.h> 69 #include <sys/vnode.h> 70 #include <sys/fcntl.h> 71 #include <sys/conf.h> 72 #include <sys/diskslice.h> 73 #include <sys/disk.h> 74 #include <sys/stat.h> 75 #include <sys/module.h> 76 #include <sys/vnioctl.h> 77 78 #include <vm/vm.h> 79 #include <vm/vm_object.h> 80 #include <vm/vm_page.h> 81 #include <vm/vm_pager.h> 82 #include <vm/vm_pageout.h> 83 #include <vm/swap_pager.h> 84 #include <vm/vm_extern.h> 85 #include <vm/vm_zone.h> 86 #include <sys/devfs.h> 87 88 static d_ioctl_t vnioctl; 89 static d_open_t vnopen; 90 static d_close_t vnclose; 91 static d_psize_t vnsize; 92 static d_strategy_t vnstrategy; 93 static d_clone_t vnclone; 94 95 MALLOC_DEFINE(M_VN, "vn_softc", "vn driver structures"); 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 VN_BSIZE_BEST 8192 105 106 /* 107 * dev_ops 108 * D_DISK we want to look like a disk 109 * D_CANFREE We support BUF_CMD_FREEBLKS 110 */ 111 112 static struct dev_ops vn_ops = { 113 { "vn", 0, D_DISK | D_CANFREE }, 114 .d_open = vnopen, 115 .d_close = vnclose, 116 .d_read = physread, 117 .d_write = physwrite, 118 .d_ioctl = vnioctl, 119 .d_strategy = vnstrategy, 120 .d_psize = vnsize 121 }; 122 123 struct vn_softc { 124 int sc_unit; 125 int sc_flags; /* flags */ 126 u_int64_t sc_size; /* size of vn, sc_secsize scale */ 127 int sc_secsize; /* sector size */ 128 struct disk sc_disk; 129 struct vnode *sc_vp; /* vnode if not NULL */ 130 vm_object_t sc_object; /* backing object if not NULL */ 131 struct ucred *sc_cred; /* credentials */ 132 int sc_maxactive; /* max # of active requests */ 133 struct buf sc_tab; /* transfer queue */ 134 u_long sc_options; /* options */ 135 cdev_t sc_dev; /* devices that refer to this unit */ 136 SLIST_ENTRY(vn_softc) sc_list; 137 }; 138 139 static SLIST_HEAD(, vn_softc) vn_list; 140 141 /* sc_flags */ 142 #define VNF_INITED 0x01 143 #define VNF_READONLY 0x02 144 #define VNF_OPENED 0x10 145 #define VNF_DESTROY 0x20 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, int clone); 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), 1); 167 168 return 0; 169 } 170 171 static int 172 vnclose(struct dev_close_args *ap) 173 { 174 cdev_t dev = ap->a_head.a_dev; 175 struct vn_softc *vn; 176 177 vn = dev->si_drv1; 178 KKASSERT(vn != NULL); 179 180 vn->sc_flags &= ~VNF_OPENED; 181 182 /* The disk has been detached and can now be safely destroyed */ 183 if (vn->sc_flags & VNF_DESTROY) { 184 KKASSERT(disk_getopencount(&vn->sc_disk) == 0); 185 disk_destroy(&vn->sc_disk); 186 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(vn), dkunit(dev)); 187 SLIST_REMOVE(&vn_list, vn, vn_softc, sc_list); 188 kfree(vn, M_VN); 189 } 190 return (0); 191 } 192 193 static struct vn_softc * 194 vncreatevn(void) 195 { 196 struct vn_softc *vn; 197 198 vn = kmalloc(sizeof *vn, M_VN, M_WAITOK | M_ZERO); 199 return vn; 200 } 201 202 static void 203 vninitvn(struct vn_softc *vn, cdev_t dev) 204 { 205 int unit; 206 207 KKASSERT(vn != NULL); 208 KKASSERT(dev != NULL); 209 unit = dkunit(dev); 210 211 vn->sc_unit = unit; 212 dev->si_drv1 = vn; 213 vn->sc_dev = dev; 214 215 SLIST_INSERT_HEAD(&vn_list, vn, sc_list); 216 } 217 218 static int 219 vnopen(struct dev_open_args *ap) 220 { 221 cdev_t dev = ap->a_head.a_dev; 222 struct vn_softc *vn; 223 224 /* 225 * Locate preexisting device 226 */ 227 228 vn = dev->si_drv1; 229 KKASSERT(vn != NULL); 230 231 /* 232 * Update si_bsize fields for device. This data will be overriden by 233 * the slice/parition code for vn accesses through partitions, and 234 * used directly if you open the 'whole disk' device. 235 * 236 * si_bsize_best must be reinitialized in case VN has been 237 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency. 238 */ 239 dev->si_bsize_phys = vn->sc_secsize; 240 dev->si_bsize_best = vn->sc_secsize; 241 if (dev->si_bsize_best < VN_BSIZE_BEST) 242 dev->si_bsize_best = VN_BSIZE_BEST; 243 244 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY)) 245 return (EACCES); 246 247 IFOPT(vn, VN_FOLLOW) 248 kprintf("vnopen(%s, 0x%x, 0x%x)\n", 249 devtoname(dev), ap->a_oflags, ap->a_devtype); 250 251 vn->sc_flags |= VNF_OPENED; 252 return(0); 253 } 254 255 /* 256 * vnstrategy: 257 * 258 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls 259 * for vnode-backed vn's, and the swap_pager_strategy() call for 260 * vm_object-backed vn's. 261 */ 262 static int 263 vnstrategy(struct dev_strategy_args *ap) 264 { 265 cdev_t dev = ap->a_head.a_dev; 266 struct bio *bio = ap->a_bio; 267 struct buf *bp; 268 struct bio *nbio; 269 int unit; 270 struct vn_softc *vn; 271 int error; 272 273 unit = dkunit(dev); 274 vn = dev->si_drv1; 275 KKASSERT(vn != NULL); 276 277 bp = bio->bio_buf; 278 279 IFOPT(vn, VN_DEBUG) 280 kprintf("vnstrategy(%p): unit %d\n", bp, unit); 281 282 if ((vn->sc_flags & VNF_INITED) == 0) { 283 bp->b_error = ENXIO; 284 bp->b_flags |= B_ERROR; 285 biodone(bio); 286 return(0); 287 } 288 289 bp->b_resid = bp->b_bcount; 290 291 /* 292 * The vnode device is using disk/slice label support. 293 * 294 * The dscheck() function is called for validating the 295 * slices that exist ON the vnode device itself, and 296 * translate the "slice-relative" block number, again. 297 * dscheck() will call biodone() and return NULL if 298 * we are at EOF or beyond the device size. 299 */ 300 301 nbio = bio; 302 303 /* 304 * Use the translated nbio from this point on 305 */ 306 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) { 307 /* 308 * Freeblks is not handled for vnode-backed elements yet. 309 */ 310 bp->b_resid = 0; 311 /* operation complete */ 312 } else if (vn->sc_vp) { 313 /* 314 * VNODE I/O 315 * 316 * If an error occurs, we set B_ERROR but we do not set 317 * B_INVAL because (for a write anyway), the buffer is 318 * still valid. 319 */ 320 struct uio auio; 321 struct iovec aiov; 322 323 bzero(&auio, sizeof(auio)); 324 325 aiov.iov_base = bp->b_data; 326 aiov.iov_len = bp->b_bcount; 327 auio.uio_iov = &aiov; 328 auio.uio_iovcnt = 1; 329 auio.uio_offset = nbio->bio_offset; 330 auio.uio_segflg = UIO_SYSSPACE; 331 if (bp->b_cmd == BUF_CMD_READ) 332 auio.uio_rw = UIO_READ; 333 else 334 auio.uio_rw = UIO_WRITE; 335 auio.uio_resid = bp->b_bcount; 336 auio.uio_td = curthread; 337 338 /* 339 * Don't use IO_DIRECT here, it really gets in the way 340 * due to typical blocksize differences between the 341 * fs backing the VN device and whatever is running on 342 * the VN device. 343 */ 344 switch (bp->b_cmd) { 345 case (BUF_CMD_READ): 346 vn_lock(vn->sc_vp, LK_SHARED | LK_RETRY); 347 error = VOP_READ(vn->sc_vp, &auio, IO_RECURSE, 348 vn->sc_cred); 349 break; 350 351 case (BUF_CMD_WRITE): 352 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY); 353 error = VOP_WRITE(vn->sc_vp, &auio, IO_RECURSE, 354 vn->sc_cred); 355 break; 356 357 case (BUF_CMD_FLUSH): 358 auio.uio_resid = 0; 359 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY); 360 error = VOP_FSYNC(vn->sc_vp, MNT_WAIT, 0); 361 break; 362 default: 363 auio.uio_resid = 0; 364 error = 0; 365 break; 366 } 367 vn_unlock(vn->sc_vp); 368 bp->b_resid = auio.uio_resid; 369 if (error) { 370 bp->b_error = error; 371 bp->b_flags |= B_ERROR; 372 } 373 /* operation complete */ 374 } else if (vn->sc_object) { 375 /* 376 * OBJT_SWAP I/O (handles read, write, freebuf) 377 * 378 * We have nothing to do if freeing blocks on a reserved 379 * swap area, othrewise execute the op. 380 */ 381 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) { 382 bp->b_resid = 0; 383 /* operation complete */ 384 } else { 385 swap_pager_strategy(vn->sc_object, nbio); 386 return(0); 387 /* NOT REACHED */ 388 } 389 } else { 390 bp->b_resid = bp->b_bcount; 391 bp->b_flags |= B_ERROR | B_INVAL; 392 bp->b_error = EINVAL; 393 /* operation complete */ 394 } 395 biodone(nbio); 396 return(0); 397 } 398 399 /* ARGSUSED */ 400 static int 401 vnioctl(struct dev_ioctl_args *ap) 402 { 403 cdev_t dev = ap->a_head.a_dev; 404 struct vn_softc *vn; 405 struct vn_ioctl *vio; 406 int error; 407 u_long *f; 408 409 vn = dev->si_drv1; 410 IFOPT(vn,VN_FOLLOW) { 411 kprintf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n", 412 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag, 413 dkunit(dev)); 414 } 415 416 switch (ap->a_cmd) { 417 case VNIOCATTACH: 418 case VNIOCDETACH: 419 case VNIOCGSET: 420 case VNIOCGCLEAR: 421 case VNIOCGET: 422 case VNIOCUSET: 423 case VNIOCUCLEAR: 424 goto vn_specific; 425 } 426 427 #if 0 428 if (dkslice(dev) != WHOLE_DISK_SLICE || 429 dkpart(dev) != WHOLE_SLICE_PART) 430 return (ENOTTY); 431 #endif 432 433 vn_specific: 434 435 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0); 436 if (error) 437 return (error); 438 439 vio = (struct vn_ioctl *)ap->a_data; 440 f = (u_long*)ap->a_data; 441 442 switch (ap->a_cmd) { 443 case VNIOCATTACH: 444 if (vn->sc_flags & VNF_INITED) 445 return(EBUSY); 446 447 if (vn->sc_flags & VNF_DESTROY) 448 return(ENXIO); 449 450 if (vio->vn_file == NULL) 451 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred); 452 else 453 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred); 454 break; 455 456 case VNIOCDETACH: 457 if ((vn->sc_flags & VNF_INITED) == 0) 458 return(ENXIO); 459 /* 460 * XXX handle i/o in progress. Return EBUSY, or wait, or 461 * flush the i/o. 462 * XXX handle multiple opens of the device. Return EBUSY, 463 * or revoke the fd's. 464 * How are these problems handled for removable and failing 465 * hardware devices? (Hint: They are not) 466 */ 467 if ((disk_getopencount(&vn->sc_disk)) > 1) 468 return (EBUSY); 469 470 vnclear(vn); 471 IFOPT(vn, VN_FOLLOW) 472 kprintf("vnioctl: CLRed\n"); 473 474 if (dkunit(dev) >= VN_PREALLOCATED_UNITS) { 475 vn->sc_flags |= VNF_DESTROY; 476 } 477 478 break; 479 480 case VNIOCGET: 481 error = vnget(dev, vn, (struct vn_user *) ap->a_data); 482 break; 483 484 case VNIOCGSET: 485 vn_options |= *f; 486 *f = vn_options; 487 break; 488 489 case VNIOCGCLEAR: 490 vn_options &= ~(*f); 491 *f = vn_options; 492 break; 493 494 case VNIOCUSET: 495 vn->sc_options |= *f; 496 *f = vn->sc_options; 497 break; 498 499 case VNIOCUCLEAR: 500 vn->sc_options &= ~(*f); 501 *f = vn->sc_options; 502 break; 503 504 default: 505 error = ENOTTY; 506 break; 507 } 508 return(error); 509 } 510 511 /* 512 * vniocattach_file: 513 * 514 * Attach a file to a VN partition. Return the size in the vn_size 515 * field. 516 */ 517 518 static int 519 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev, 520 int flag, struct ucred *cred) 521 { 522 struct vattr vattr; 523 struct nlookupdata nd; 524 int error, flags; 525 struct vnode *vp; 526 struct disk_info info; 527 528 flags = FREAD|FWRITE; 529 error = nlookup_init(&nd, vio->vn_file, 530 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP); 531 if (error) 532 return (error); 533 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) { 534 if (error != EACCES && error != EPERM && error != EROFS) 535 goto done; 536 flags &= ~FWRITE; 537 nlookup_done(&nd); 538 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP); 539 if (error) 540 return (error); 541 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) 542 goto done; 543 } 544 vp = nd.nl_open_vp; 545 if (vp->v_type != VREG || 546 (error = VOP_GETATTR(vp, &vattr))) { 547 if (error == 0) 548 error = EINVAL; 549 goto done; 550 } 551 vn_unlock(vp); 552 vn->sc_secsize = DEV_BSIZE; 553 vn->sc_vp = vp; 554 nd.nl_open_vp = NULL; 555 556 /* 557 * If the size is specified, override the file attributes. Note that 558 * the vn_size argument is in PAGE_SIZE sized blocks. 559 */ 560 if (vio->vn_size) 561 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize; 562 else 563 vn->sc_size = vattr.va_size / vn->sc_secsize; 564 error = vnsetcred(vn, cred); 565 if (error) { 566 vn->sc_vp = NULL; 567 vn_close(vp, flags, NULL); 568 goto done; 569 } 570 vn->sc_flags |= VNF_INITED; 571 if (flags == FREAD) 572 vn->sc_flags |= VNF_READONLY; 573 574 /* 575 * Set the disk info so that probing is triggered 576 */ 577 bzero(&info, sizeof(struct disk_info)); 578 info.d_media_blksize = vn->sc_secsize; 579 info.d_media_blocks = vn->sc_size; 580 /* 581 * reserve mbr sector for backwards compatibility 582 * when no slices exist. 583 */ 584 info.d_dsflags = DSO_COMPATMBR | DSO_RAWPSIZE; 585 info.d_secpertrack = 32; 586 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE); 587 info.d_secpercyl = info.d_secpertrack * info.d_nheads; 588 info.d_ncylinders = vn->sc_size / info.d_secpercyl; 589 disk_setdiskinfo_sync(&vn->sc_disk, &info); 590 591 error = dev_dopen(dev, flag, S_IFCHR, cred, NULL); 592 if (error) 593 vnclear(vn); 594 595 IFOPT(vn, VN_FOLLOW) 596 kprintf("vnioctl: SET vp %p size %llx blks\n", 597 vn->sc_vp, (long long)vn->sc_size); 598 done: 599 nlookup_done(&nd); 600 return(error); 601 } 602 603 /* 604 * vniocattach_swap: 605 * 606 * Attach swap backing store to a VN partition of the size specified 607 * in vn_size. 608 */ 609 610 static int 611 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev, 612 int flag, struct ucred *cred) 613 { 614 int error; 615 struct disk_info info; 616 617 /* 618 * Range check. Disallow negative sizes or any size less then the 619 * size of a page. Then round to a page. 620 */ 621 622 if (vio->vn_size <= 0) 623 return(EDOM); 624 625 /* 626 * Allocate an OBJT_SWAP object. 627 * 628 * sc_secsize is PAGE_SIZE'd 629 * 630 * vio->vn_size is in PAGE_SIZE'd chunks. 631 * sc_size must be in PAGE_SIZE'd chunks. 632 * Note the truncation. 633 */ 634 635 vn->sc_secsize = PAGE_SIZE; 636 vn->sc_size = vio->vn_size; 637 vn->sc_object = swap_pager_alloc(NULL, 638 vn->sc_secsize * (off_t)vio->vn_size, 639 VM_PROT_DEFAULT, 0); 640 IFOPT(vn, VN_RESERVE) { 641 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) { 642 vm_pager_deallocate(vn->sc_object); 643 vn->sc_object = NULL; 644 return(EDOM); 645 } 646 } 647 vn->sc_flags |= VNF_INITED; 648 649 error = vnsetcred(vn, cred); 650 if (error == 0) { 651 /* 652 * Set the disk info so that probing is triggered 653 */ 654 bzero(&info, sizeof(struct disk_info)); 655 info.d_media_blksize = vn->sc_secsize; 656 info.d_media_blocks = vn->sc_size; 657 /* 658 * reserve mbr sector for backwards compatibility 659 * when no slices exist. 660 */ 661 info.d_dsflags = DSO_COMPATMBR | DSO_RAWPSIZE; 662 info.d_secpertrack = 32; 663 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE); 664 info.d_secpercyl = info.d_secpertrack * info.d_nheads; 665 info.d_ncylinders = vn->sc_size / info.d_secpercyl; 666 disk_setdiskinfo_sync(&vn->sc_disk, &info); 667 668 error = dev_dopen(dev, flag, S_IFCHR, cred, NULL); 669 } 670 if (error == 0) { 671 IFOPT(vn, VN_FOLLOW) { 672 kprintf("vnioctl: SET vp %p size %llx\n", 673 vn->sc_vp, (long long)vn->sc_size); 674 } 675 } 676 if (error) 677 vnclear(vn); 678 return(error); 679 } 680 681 /* 682 * Duplicate the current processes' credentials. Since we are called only 683 * as the result of a SET ioctl and only root can do that, any future access 684 * to this "disk" is essentially as root. Note that credentials may change 685 * if some other uid can write directly to the mapped file (NFS). 686 */ 687 int 688 vnsetcred(struct vn_softc *vn, struct ucred *cred) 689 { 690 char *tmpbuf; 691 int error = 0; 692 693 /* 694 * Set credits in our softc 695 */ 696 697 if (vn->sc_cred) 698 crfree(vn->sc_cred); 699 vn->sc_cred = crdup(cred); 700 701 /* 702 * Horrible kludge to establish credentials for NFS XXX. 703 */ 704 705 if (vn->sc_vp) { 706 struct uio auio; 707 struct iovec aiov; 708 709 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK); 710 bzero(&auio, sizeof(auio)); 711 712 aiov.iov_base = tmpbuf; 713 aiov.iov_len = vn->sc_secsize; 714 auio.uio_iov = &aiov; 715 auio.uio_iovcnt = 1; 716 auio.uio_offset = 0; 717 auio.uio_rw = UIO_READ; 718 auio.uio_segflg = UIO_SYSSPACE; 719 auio.uio_resid = aiov.iov_len; 720 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY); 721 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred); 722 vn_unlock(vn->sc_vp); 723 kfree(tmpbuf, M_TEMP); 724 } 725 return (error); 726 } 727 728 void 729 vnclear(struct vn_softc *vn) 730 { 731 IFOPT(vn, VN_FOLLOW) 732 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp); 733 vn->sc_flags &= ~VNF_INITED; 734 if (vn->sc_vp != NULL) { 735 vn_close(vn->sc_vp, 736 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE), 737 NULL); 738 vn->sc_vp = NULL; 739 } 740 vn->sc_flags &= ~VNF_READONLY; 741 if (vn->sc_cred) { 742 crfree(vn->sc_cred); 743 vn->sc_cred = NULL; 744 } 745 if (vn->sc_object != NULL) { 746 vm_pager_deallocate(vn->sc_object); 747 vn->sc_object = NULL; 748 } 749 750 disk_unprobe(&vn->sc_disk); 751 752 vn->sc_size = 0; 753 } 754 755 /* 756 * vnget: 757 * 758 * populate a struct vn_user for the VNIOCGET ioctl. 759 * interface conventions defined in sys/sys/vnioctl.h. 760 */ 761 762 static int 763 vnget(cdev_t dev, struct vn_softc *vn, struct vn_user *vnu) 764 { 765 int error, found = 0; 766 char *freepath, *fullpath; 767 struct vattr vattr; 768 769 if (vnu->vnu_unit == -1) { 770 vnu->vnu_unit = dkunit(dev); 771 } 772 else if (vnu->vnu_unit < 0) 773 return (EINVAL); 774 775 SLIST_FOREACH(vn, &vn_list, sc_list) { 776 777 if(vn->sc_unit != vnu->vnu_unit) 778 continue; 779 780 found = 1; 781 782 if (vn->sc_flags & VNF_INITED && vn->sc_vp != NULL) { 783 784 /* note: u_cred checked in vnioctl above */ 785 error = VOP_GETATTR(vn->sc_vp, &vattr); 786 if (error) { 787 kprintf("vnget: VOP_GETATTR for %p failed\n", 788 vn->sc_vp); 789 return (error); 790 } 791 792 error = vn_fullpath(curproc, vn->sc_vp, 793 &fullpath, &freepath, 0); 794 795 if (error) { 796 kprintf("vnget: unable to resolve vp %p\n", 797 vn->sc_vp); 798 return(error); 799 } 800 801 strlcpy(vnu->vnu_file, fullpath, 802 sizeof(vnu->vnu_file)); 803 kfree(freepath, M_TEMP); 804 vnu->vnu_dev = vattr.va_fsid; 805 vnu->vnu_ino = vattr.va_fileid; 806 807 } 808 else if (vn->sc_flags & VNF_INITED && vn->sc_object != NULL){ 809 810 strlcpy(vnu->vnu_file, _VN_USER_SWAP, 811 sizeof(vnu->vnu_file)); 812 vnu->vnu_size = vn->sc_size; 813 vnu->vnu_secsize = vn->sc_secsize; 814 815 } else { 816 817 bzero(vnu->vnu_file, sizeof(vnu->vnu_file)); 818 vnu->vnu_dev = 0; 819 vnu->vnu_ino = 0; 820 821 } 822 break; 823 } 824 825 if (!found) 826 return(ENXIO); 827 828 return(0); 829 } 830 831 static int 832 vnsize(struct dev_psize_args *ap) 833 { 834 cdev_t dev = ap->a_head.a_dev; 835 struct vn_softc *vn; 836 837 vn = dev->si_drv1; 838 if (!vn) 839 return(ENXIO); 840 if ((vn->sc_flags & VNF_INITED) == 0) 841 return(ENXIO); 842 ap->a_result = (int64_t)vn->sc_size; 843 return(0); 844 } 845 846 static cdev_t 847 vn_create(int unit, struct devfs_bitmap *bitmap, int clone) 848 { 849 struct vn_softc *vn; 850 struct disk_info info; 851 cdev_t dev, ret_dev; 852 853 vn = vncreatevn(); 854 if (clone) { 855 /* 856 * For clone devices we need to return the top-level cdev, 857 * not the raw dev we'd normally work with. 858 */ 859 dev = disk_create_clone(unit, &vn->sc_disk, &vn_ops); 860 ret_dev = vn->sc_disk.d_cdev; 861 } else { 862 ret_dev = dev = disk_create(unit, &vn->sc_disk, &vn_ops); 863 } 864 vninitvn(vn, dev); 865 866 bzero(&info, sizeof(struct disk_info)); 867 info.d_media_blksize = 512; 868 info.d_media_blocks = 0; 869 info.d_dsflags = DSO_MBRQUIET | DSO_RAWPSIZE; 870 info.d_secpertrack = 32; 871 info.d_nheads = 64; 872 info.d_secpercyl = info.d_secpertrack * info.d_nheads; 873 info.d_ncylinders = 0; 874 disk_setdiskinfo_sync(&vn->sc_disk, &info); 875 876 if (bitmap != NULL) 877 devfs_clone_bitmap_set(bitmap, unit); 878 879 return ret_dev; 880 } 881 882 static int 883 vn_modevent(module_t mod, int type, void *data) 884 { 885 struct vn_softc *vn; 886 static cdev_t dev = NULL; 887 int i; 888 889 switch (type) { 890 case MOD_LOAD: 891 dev = make_autoclone_dev(&vn_ops, &DEVFS_CLONE_BITMAP(vn), vnclone, UID_ROOT, 892 GID_OPERATOR, 0640, "vn"); 893 894 for (i = 0; i < VN_PREALLOCATED_UNITS; i++) { 895 vn_create(i, &DEVFS_CLONE_BITMAP(vn), 0); 896 } 897 break; 898 899 case MOD_UNLOAD: 900 case MOD_SHUTDOWN: 901 while ((vn = SLIST_FIRST(&vn_list)) != NULL) { 902 /* 903 * XXX: no idea if we can return EBUSY even in the 904 * shutdown case, so err on the side of caution 905 * and just rip stuff out on shutdown. 906 */ 907 if (type != MOD_SHUTDOWN) { 908 if (vn->sc_flags & VNF_OPENED) 909 return (EBUSY); 910 } 911 912 disk_destroy(&vn->sc_disk); 913 914 SLIST_REMOVE_HEAD(&vn_list, sc_list); 915 916 if (vn->sc_flags & VNF_INITED) 917 vnclear(vn); 918 919 kfree(vn, M_VN); 920 } 921 destroy_autoclone_dev(dev, &DEVFS_CLONE_BITMAP(vn)); 922 dev_ops_remove_all(&vn_ops); 923 break; 924 default: 925 break; 926 } 927 return 0; 928 } 929 930 DEV_MODULE(vn, vn_modevent, 0); 931