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