1 /* $NetBSD: vnd.c,v 1.69 2000/11/27 08:39:41 chs Exp $ */ 2 3 /*- 4 * Copyright (c) 1996, 1997, 1998 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe. 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 NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39 /* 40 * Copyright (c) 1988 University of Utah. 41 * Copyright (c) 1990, 1993 42 * The Regents of the University of California. All rights reserved. 43 * 44 * This code is derived from software contributed to Berkeley by 45 * the Systems Programming Group of the University of Utah Computer 46 * Science Department. 47 * 48 * Redistribution and use in source and binary forms, with or without 49 * modification, are permitted provided that the following conditions 50 * are met: 51 * 1. Redistributions of source code must retain the above copyright 52 * notice, this list of conditions and the following disclaimer. 53 * 2. Redistributions in binary form must reproduce the above copyright 54 * notice, this list of conditions and the following disclaimer in the 55 * documentation and/or other materials provided with the distribution. 56 * 3. All advertising materials mentioning features or use of this software 57 * must display the following acknowledgement: 58 * This product includes software developed by the University of 59 * California, Berkeley and its contributors. 60 * 4. Neither the name of the University nor the names of its contributors 61 * may be used to endorse or promote products derived from this software 62 * without specific prior written permission. 63 * 64 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 65 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 66 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 67 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 68 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 69 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 70 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 71 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 72 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 73 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 74 * SUCH DAMAGE. 75 * 76 * from: Utah $Hdr: vn.c 1.13 94/04/02$ 77 * 78 * @(#)vn.c 8.9 (Berkeley) 5/14/95 79 */ 80 81 /* 82 * Vnode disk driver. 83 * 84 * Block/character interface to a vnode. Allows one to treat a file 85 * as a disk (e.g. build a filesystem in it, mount it, etc.). 86 * 87 * NOTE 1: This uses the VOP_BMAP/VOP_STRATEGY interface to the vnode 88 * instead of a simple VOP_RDWR. We do this to avoid distorting the 89 * local buffer cache. 90 * 91 * NOTE 2: There is a security issue involved with this driver. 92 * Once mounted all access to the contents of the "mapped" file via 93 * the special file is controlled by the permissions on the special 94 * file, the protection of the mapped file is ignored (effectively, 95 * by using root credentials in all transactions). 96 * 97 * NOTE 3: Doesn't interact with leases, should it? 98 */ 99 100 #include "fs_nfs.h" 101 102 #include <sys/param.h> 103 #include <sys/systm.h> 104 #include <sys/namei.h> 105 #include <sys/proc.h> 106 #include <sys/errno.h> 107 #include <sys/buf.h> 108 #include <sys/malloc.h> 109 #include <sys/ioctl.h> 110 #include <sys/disklabel.h> 111 #include <sys/device.h> 112 #include <sys/disk.h> 113 #include <sys/stat.h> 114 #include <sys/mount.h> 115 #include <sys/vnode.h> 116 #include <sys/file.h> 117 #include <sys/uio.h> 118 #include <sys/conf.h> 119 120 #include <miscfs/specfs/specdev.h> 121 122 #include <dev/vndvar.h> 123 124 #if defined(VNDDEBUG) && !defined(DEBUG) 125 #define DEBUG 126 #endif 127 128 #ifdef DEBUG 129 int dovndcluster = 1; 130 #define VDB_FOLLOW 0x01 131 #define VDB_INIT 0x02 132 #define VDB_IO 0x04 133 #define VDB_LABEL 0x08 134 int vnddebug = 0x00; 135 #endif 136 137 #define vndunit(x) DISKUNIT(x) 138 139 struct vndxfer { 140 struct buf *vx_bp; /* Pointer to parent buffer */ 141 int vx_error; 142 int vx_pending; /* # of pending aux buffers */ 143 int vx_flags; 144 #define VX_BUSY 1 145 }; 146 147 struct vndbuf { 148 struct buf vb_buf; 149 struct vndxfer *vb_xfer; 150 }; 151 152 #define VND_GETXFER(vnd) pool_get(&(vnd)->sc_vxpool, PR_NOWAIT) 153 #define VND_PUTXFER(vnd, vx) pool_put(&(vnd)->sc_vxpool, (vx)) 154 155 #define VND_GETBUF(vnd) pool_get(&(vnd)->sc_vbpool, PR_NOWAIT) 156 #define VND_PUTBUF(vnd, vb) pool_put(&(vnd)->sc_vbpool, (vb)) 157 158 struct vnd_softc *vnd_softc; 159 int numvnd = 0; 160 161 #define VNDLABELDEV(dev) \ 162 (MAKEDISKDEV(major((dev)), vndunit((dev)), RAW_PART)) 163 164 /* called by main() at boot time */ 165 void vndattach __P((int)); 166 167 void vndclear __P((struct vnd_softc *)); 168 void vndstart __P((struct vnd_softc *)); 169 int vndsetcred __P((struct vnd_softc *, struct ucred *)); 170 void vndthrottle __P((struct vnd_softc *, struct vnode *)); 171 void vndiodone __P((struct buf *)); 172 void vndshutdown __P((void)); 173 174 void vndgetdefaultlabel __P((struct vnd_softc *, struct disklabel *)); 175 void vndgetdisklabel __P((dev_t)); 176 177 static int vndlock __P((struct vnd_softc *)); 178 static void vndunlock __P((struct vnd_softc *)); 179 180 void 181 vndattach(num) 182 int num; 183 { 184 int i; 185 char *mem; 186 187 if (num <= 0) 188 return; 189 i = num * sizeof(struct vnd_softc); 190 mem = malloc(i, M_DEVBUF, M_NOWAIT); 191 if (mem == NULL) { 192 printf("WARNING: no memory for vnode disks\n"); 193 return; 194 } 195 bzero(mem, i); 196 vnd_softc = (struct vnd_softc *)mem; 197 numvnd = num; 198 199 for (i = 0; i < numvnd; i++) 200 BUFQ_INIT(&vnd_softc[i].sc_tab); 201 } 202 203 int 204 vndopen(dev, flags, mode, p) 205 dev_t dev; 206 int flags, mode; 207 struct proc *p; 208 { 209 int unit = vndunit(dev); 210 struct vnd_softc *sc; 211 int error = 0, part, pmask; 212 struct disklabel *lp; 213 214 #ifdef DEBUG 215 if (vnddebug & VDB_FOLLOW) 216 printf("vndopen(0x%x, 0x%x, 0x%x, %p)\n", dev, flags, mode, p); 217 #endif 218 if (unit >= numvnd) 219 return (ENXIO); 220 sc = &vnd_softc[unit]; 221 222 if ((error = vndlock(sc)) != 0) 223 return (error); 224 225 lp = sc->sc_dkdev.dk_label; 226 227 part = DISKPART(dev); 228 pmask = (1 << part); 229 230 /* 231 * If we're initialized, check to see if there are any other 232 * open partitions. If not, then it's safe to update the 233 * in-core disklabel. 234 */ 235 if ((sc->sc_flags & VNF_INITED) && (sc->sc_dkdev.dk_openmask == 0)) 236 vndgetdisklabel(dev); 237 238 /* Check that the partitions exists. */ 239 if (part != RAW_PART) { 240 if (((sc->sc_flags & VNF_INITED) == 0) || 241 ((part >= lp->d_npartitions) || 242 (lp->d_partitions[part].p_fstype == FS_UNUSED))) { 243 error = ENXIO; 244 goto done; 245 } 246 } 247 248 /* Prevent our unit from being unconfigured while open. */ 249 switch (mode) { 250 case S_IFCHR: 251 sc->sc_dkdev.dk_copenmask |= pmask; 252 break; 253 254 case S_IFBLK: 255 sc->sc_dkdev.dk_bopenmask |= pmask; 256 break; 257 } 258 sc->sc_dkdev.dk_openmask = 259 sc->sc_dkdev.dk_copenmask | sc->sc_dkdev.dk_bopenmask; 260 261 done: 262 vndunlock(sc); 263 return (error); 264 } 265 266 int 267 vndclose(dev, flags, mode, p) 268 dev_t dev; 269 int flags, mode; 270 struct proc *p; 271 { 272 int unit = vndunit(dev); 273 struct vnd_softc *sc; 274 int error = 0, part; 275 276 #ifdef DEBUG 277 if (vnddebug & VDB_FOLLOW) 278 printf("vndclose(0x%x, 0x%x, 0x%x, %p)\n", dev, flags, mode, p); 279 #endif 280 281 if (unit >= numvnd) 282 return (ENXIO); 283 sc = &vnd_softc[unit]; 284 285 if ((error = vndlock(sc)) != 0) 286 return (error); 287 288 part = DISKPART(dev); 289 290 /* ...that much closer to allowing unconfiguration... */ 291 switch (mode) { 292 case S_IFCHR: 293 sc->sc_dkdev.dk_copenmask &= ~(1 << part); 294 break; 295 296 case S_IFBLK: 297 sc->sc_dkdev.dk_bopenmask &= ~(1 << part); 298 break; 299 } 300 sc->sc_dkdev.dk_openmask = 301 sc->sc_dkdev.dk_copenmask | sc->sc_dkdev.dk_bopenmask; 302 303 vndunlock(sc); 304 return (0); 305 } 306 307 /* 308 * Break the request into bsize pieces and submit using VOP_BMAP/VOP_STRATEGY. 309 */ 310 void 311 vndstrategy(bp) 312 struct buf *bp; 313 { 314 int unit = vndunit(bp->b_dev); 315 struct vnd_softc *vnd = &vnd_softc[unit]; 316 struct vndxfer *vnx; 317 int s, bsize, resid; 318 off_t bn; 319 caddr_t addr; 320 int sz, flags, error, wlabel; 321 struct disklabel *lp; 322 struct partition *pp; 323 324 #ifdef DEBUG 325 if (vnddebug & VDB_FOLLOW) 326 printf("vndstrategy(%p): unit %d\n", bp, unit); 327 #endif 328 if ((vnd->sc_flags & VNF_INITED) == 0) { 329 bp->b_error = ENXIO; 330 bp->b_flags |= B_ERROR; 331 goto done; 332 } 333 334 /* If it's a nil transfer, wake up the top half now. */ 335 if (bp->b_bcount == 0) 336 goto done; 337 338 lp = vnd->sc_dkdev.dk_label; 339 340 /* 341 * The transfer must be a whole number of blocks. 342 */ 343 if ((bp->b_bcount % lp->d_secsize) != 0) { 344 bp->b_error = EINVAL; 345 bp->b_flags |= B_ERROR; 346 goto done; 347 } 348 349 /* 350 * Do bounds checking and adjust transfer. If there's an error, 351 * the bounds check will flag that for us. 352 */ 353 wlabel = vnd->sc_flags & (VNF_WLABEL|VNF_LABELLING); 354 if (DISKPART(bp->b_dev) != RAW_PART) 355 if (bounds_check_with_label(bp, lp, wlabel) <= 0) 356 goto done; 357 358 bp->b_resid = bp->b_bcount; 359 360 /* 361 * Put the block number in terms of the logical blocksize 362 * of the "device". 363 */ 364 bn = bp->b_blkno / (lp->d_secsize / DEV_BSIZE); 365 366 /* 367 * Translate the partition-relative block number to an absolute. 368 */ 369 if (DISKPART(bp->b_dev) != RAW_PART) { 370 pp = &vnd->sc_dkdev.dk_label->d_partitions[DISKPART(bp->b_dev)]; 371 bn += pp->p_offset; 372 } 373 374 /* ...and convert to a byte offset within the file. */ 375 bn *= lp->d_secsize; 376 377 bsize = vnd->sc_vp->v_mount->mnt_stat.f_iosize; 378 addr = bp->b_data; 379 flags = (bp->b_flags & (B_READ|B_ASYNC)) | B_CALL; 380 381 /* Allocate a header for this transfer and link it to the buffer */ 382 s = splbio(); 383 vnx = VND_GETXFER(vnd); 384 splx(s); 385 vnx->vx_flags = VX_BUSY; 386 vnx->vx_error = 0; 387 vnx->vx_pending = 0; 388 vnx->vx_bp = bp; 389 390 for (resid = bp->b_resid; resid; resid -= sz) { 391 struct vndbuf *nbp; 392 struct vnode *vp; 393 daddr_t nbn; 394 int off, nra; 395 396 nra = 0; 397 vn_lock(vnd->sc_vp, LK_EXCLUSIVE | LK_RETRY | LK_CANRECURSE); 398 error = VOP_BMAP(vnd->sc_vp, bn / bsize, &vp, &nbn, &nra); 399 VOP_UNLOCK(vnd->sc_vp, 0); 400 401 if (error == 0 && (long)nbn == -1) 402 error = EIO; 403 404 /* 405 * If there was an error or a hole in the file...punt. 406 * Note that we may have to wait for any operations 407 * that we have already fired off before releasing 408 * the buffer. 409 * 410 * XXX we could deal with holes here but it would be 411 * a hassle (in the write case). 412 */ 413 if (error) { 414 s = splbio(); 415 vnx->vx_error = error; 416 goto out; 417 } 418 419 #ifdef DEBUG 420 if (!dovndcluster) 421 nra = 0; 422 #endif 423 424 if ((off = bn % bsize) != 0) 425 sz = bsize - off; 426 else 427 sz = (1 + nra) * bsize; 428 if (resid < sz) 429 sz = resid; 430 #ifdef DEBUG 431 if (vnddebug & VDB_IO) 432 printf("vndstrategy: vp %p/%p bn 0x%qx/0x%x sz 0x%x\n", 433 vnd->sc_vp, vp, (long long)bn, nbn, sz); 434 #endif 435 436 s = splbio(); 437 nbp = VND_GETBUF(vnd); 438 splx(s); 439 nbp->vb_buf.b_flags = flags; 440 nbp->vb_buf.b_bcount = sz; 441 nbp->vb_buf.b_bufsize = bp->b_bufsize; 442 nbp->vb_buf.b_error = 0; 443 nbp->vb_buf.b_data = addr; 444 nbp->vb_buf.b_blkno = nbp->vb_buf.b_rawblkno = nbn + btodb(off); 445 nbp->vb_buf.b_proc = bp->b_proc; 446 nbp->vb_buf.b_iodone = vndiodone; 447 nbp->vb_buf.b_vp = NULLVP; 448 LIST_INIT(&nbp->vb_buf.b_dep); 449 450 nbp->vb_xfer = vnx; 451 452 /* 453 * Just sort by block number 454 */ 455 s = splbio(); 456 if (vnx->vx_error != 0) { 457 VND_PUTBUF(vnd, nbp); 458 goto out; 459 } 460 vnx->vx_pending++; 461 bgetvp(vp, &nbp->vb_buf); 462 disksort_blkno(&vnd->sc_tab, &nbp->vb_buf); 463 vndstart(vnd); 464 splx(s); 465 bn += sz; 466 addr += sz; 467 } 468 469 s = splbio(); 470 471 out: /* Arrive here at splbio */ 472 vnx->vx_flags &= ~VX_BUSY; 473 if (vnx->vx_pending == 0) { 474 if (vnx->vx_error != 0) { 475 bp->b_error = vnx->vx_error; 476 bp->b_flags |= B_ERROR; 477 } 478 VND_PUTXFER(vnd, vnx); 479 biodone(bp); 480 } 481 splx(s); 482 return; 483 484 done: 485 biodone(bp); 486 } 487 488 /* 489 * Feed requests sequentially. 490 * We do it this way to keep from flooding NFS servers if we are connected 491 * to an NFS file. This places the burden on the client rather than the 492 * server. 493 */ 494 void 495 vndstart(vnd) 496 struct vnd_softc *vnd; 497 { 498 struct buf *bp; 499 500 /* 501 * Dequeue now since lower level strategy routine might 502 * queue using same links 503 */ 504 505 if ((vnd->sc_flags & VNF_BUSY) != 0) 506 return; 507 508 vnd->sc_flags |= VNF_BUSY; 509 510 while (vnd->sc_active < vnd->sc_maxactive) { 511 bp = BUFQ_FIRST(&vnd->sc_tab); 512 if (bp == NULL) 513 break; 514 BUFQ_REMOVE(&vnd->sc_tab, bp); 515 vnd->sc_active++; 516 #ifdef DEBUG 517 if (vnddebug & VDB_IO) 518 printf("vndstart(%ld): bp %p vp %p blkno 0x%x" 519 " flags %lx addr %p cnt 0x%lx\n", 520 (long) (vnd-vnd_softc), bp, bp->b_vp, bp->b_blkno, 521 bp->b_flags, bp->b_data, bp->b_bcount); 522 #endif 523 524 /* Instrumentation. */ 525 disk_busy(&vnd->sc_dkdev); 526 527 if ((bp->b_flags & B_READ) == 0) 528 bp->b_vp->v_numoutput++; 529 VOP_STRATEGY(bp); 530 } 531 vnd->sc_flags &= ~VNF_BUSY; 532 } 533 534 void 535 vndiodone(bp) 536 struct buf *bp; 537 { 538 struct vndbuf *vbp = (struct vndbuf *) bp; 539 struct vndxfer *vnx = (struct vndxfer *)vbp->vb_xfer; 540 struct buf *pbp = vnx->vx_bp; 541 struct vnd_softc *vnd = &vnd_softc[vndunit(pbp->b_dev)]; 542 int s, resid; 543 544 s = splbio(); 545 #ifdef DEBUG 546 if (vnddebug & VDB_IO) 547 printf("vndiodone(%ld): vbp %p vp %p blkno 0x%x addr %p cnt 0x%lx\n", 548 (long) (vnd-vnd_softc), vbp, vbp->vb_buf.b_vp, 549 vbp->vb_buf.b_blkno, vbp->vb_buf.b_data, 550 vbp->vb_buf.b_bcount); 551 #endif 552 553 resid = vbp->vb_buf.b_bcount - vbp->vb_buf.b_resid; 554 pbp->b_resid -= resid; 555 disk_unbusy(&vnd->sc_dkdev, resid); 556 vnx->vx_pending--; 557 558 if (vbp->vb_buf.b_error) { 559 #ifdef DEBUG 560 if (vnddebug & VDB_IO) 561 printf("vndiodone: vbp %p error %d\n", vbp, 562 vbp->vb_buf.b_error); 563 #endif 564 vnx->vx_error = vbp->vb_buf.b_error; 565 } 566 567 if (vbp->vb_buf.b_vp != NULLVP) 568 brelvp(&vbp->vb_buf); 569 570 VND_PUTBUF(vnd, vbp); 571 572 /* 573 * Wrap up this transaction if it has run to completion or, in 574 * case of an error, when all auxiliary buffers have returned. 575 */ 576 if (vnx->vx_error != 0) { 577 pbp->b_flags |= B_ERROR; 578 pbp->b_error = vnx->vx_error; 579 if ((vnx->vx_flags & VX_BUSY) == 0 && vnx->vx_pending == 0) { 580 581 #ifdef DEBUG 582 if (vnddebug & VDB_IO) 583 printf("vndiodone: pbp %p iodone: error %d\n", 584 pbp, vnx->vx_error); 585 #endif 586 VND_PUTXFER(vnd, vnx); 587 biodone(pbp); 588 } 589 } else if (pbp->b_resid == 0) { 590 591 #ifdef DIAGNOSTIC 592 if (vnx->vx_pending != 0) 593 panic("vndiodone: vnx pending: %d", vnx->vx_pending); 594 #endif 595 596 if ((vnx->vx_flags & VX_BUSY) == 0) { 597 #ifdef DEBUG 598 if (vnddebug & VDB_IO) 599 printf("vndiodone: pbp %p iodone\n", pbp); 600 #endif 601 VND_PUTXFER(vnd, vnx); 602 biodone(pbp); 603 } 604 } 605 606 vnd->sc_active--; 607 vndstart(vnd); 608 splx(s); 609 } 610 611 /* ARGSUSED */ 612 int 613 vndread(dev, uio, flags) 614 dev_t dev; 615 struct uio *uio; 616 int flags; 617 { 618 int unit = vndunit(dev); 619 struct vnd_softc *sc; 620 621 #ifdef DEBUG 622 if (vnddebug & VDB_FOLLOW) 623 printf("vndread(0x%x, %p)\n", dev, uio); 624 #endif 625 626 if (unit >= numvnd) 627 return (ENXIO); 628 sc = &vnd_softc[unit]; 629 630 if ((sc->sc_flags & VNF_INITED) == 0) 631 return (ENXIO); 632 633 return (physio(vndstrategy, NULL, dev, B_READ, minphys, uio)); 634 } 635 636 /* ARGSUSED */ 637 int 638 vndwrite(dev, uio, flags) 639 dev_t dev; 640 struct uio *uio; 641 int flags; 642 { 643 int unit = vndunit(dev); 644 struct vnd_softc *sc; 645 646 #ifdef DEBUG 647 if (vnddebug & VDB_FOLLOW) 648 printf("vndwrite(0x%x, %p)\n", dev, uio); 649 #endif 650 651 if (unit >= numvnd) 652 return (ENXIO); 653 sc = &vnd_softc[unit]; 654 655 if ((sc->sc_flags & VNF_INITED) == 0) 656 return (ENXIO); 657 658 return (physio(vndstrategy, NULL, dev, B_WRITE, minphys, uio)); 659 } 660 661 /* ARGSUSED */ 662 int 663 vndioctl(dev, cmd, data, flag, p) 664 dev_t dev; 665 u_long cmd; 666 caddr_t data; 667 int flag; 668 struct proc *p; 669 { 670 int unit = vndunit(dev); 671 struct vnd_softc *vnd; 672 struct vnd_ioctl *vio; 673 struct vattr vattr; 674 struct nameidata nd; 675 int error, part, pmask; 676 size_t geomsize; 677 678 #ifdef DEBUG 679 if (vnddebug & VDB_FOLLOW) 680 printf("vndioctl(0x%x, 0x%lx, %p, 0x%x, %p): unit %d\n", 681 dev, cmd, data, flag, p, unit); 682 #endif 683 error = suser(p->p_ucred, &p->p_acflag); 684 if (error) 685 return (error); 686 if (unit >= numvnd) 687 return (ENXIO); 688 689 vnd = &vnd_softc[unit]; 690 vio = (struct vnd_ioctl *)data; 691 692 /* Must be open for writes for these commands... */ 693 switch (cmd) { 694 case VNDIOCSET: 695 case VNDIOCCLR: 696 case DIOCSDINFO: 697 case DIOCWDINFO: 698 case DIOCWLABEL: 699 if ((flag & FWRITE) == 0) 700 return (EBADF); 701 } 702 703 /* Must be initialized for these... */ 704 switch (cmd) { 705 case VNDIOCCLR: 706 case DIOCGDINFO: 707 case DIOCSDINFO: 708 case DIOCWDINFO: 709 case DIOCGPART: 710 case DIOCWLABEL: 711 case DIOCGDEFLABEL: 712 if ((vnd->sc_flags & VNF_INITED) == 0) 713 return (ENXIO); 714 } 715 716 switch (cmd) { 717 case VNDIOCSET: 718 if (vnd->sc_flags & VNF_INITED) 719 return (EBUSY); 720 721 if ((error = vndlock(vnd)) != 0) 722 return (error); 723 724 /* 725 * Always open for read and write. 726 * This is probably bogus, but it lets vn_open() 727 * weed out directories, sockets, etc. so we don't 728 * have to worry about them. 729 */ 730 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, vio->vnd_file, p); 731 if ((error = vn_open(&nd, FREAD|FWRITE, 0)) != 0) { 732 vndunlock(vnd); 733 return(error); 734 } 735 error = VOP_GETATTR(nd.ni_vp, &vattr, p->p_ucred, p); 736 if (error) { 737 VOP_UNLOCK(nd.ni_vp, 0); 738 (void) vn_close(nd.ni_vp, FREAD|FWRITE, p->p_ucred, p); 739 vndunlock(vnd); 740 return(error); 741 } 742 VOP_UNLOCK(nd.ni_vp, 0); 743 vnd->sc_vp = nd.ni_vp; 744 vnd->sc_size = btodb(vattr.va_size); /* note truncation */ 745 746 /* 747 * Use pseudo-geometry specified. If none was provided, 748 * use "standard" Adaptec fictitious geometry. 749 */ 750 if (vio->vnd_flags & VNDIOF_HASGEOM) { 751 752 bcopy(&vio->vnd_geom, &vnd->sc_geom, 753 sizeof(vio->vnd_geom)); 754 755 /* 756 * Sanity-check the sector size. 757 * XXX Don't allow secsize < DEV_BSIZE. Should 758 * XXX we? 759 */ 760 if (vnd->sc_geom.vng_secsize < DEV_BSIZE || 761 (vnd->sc_geom.vng_secsize % DEV_BSIZE) != 0) { 762 (void) vn_close(nd.ni_vp, FREAD|FWRITE, 763 p->p_ucred, p); 764 vndunlock(vnd); 765 return (EINVAL); 766 } 767 768 /* 769 * Compute the size (in DEV_BSIZE blocks) specified 770 * by the geometry. 771 */ 772 geomsize = (vnd->sc_geom.vng_nsectors * 773 vnd->sc_geom.vng_ntracks * 774 vnd->sc_geom.vng_ncylinders) * 775 (vnd->sc_geom.vng_secsize / DEV_BSIZE); 776 777 /* 778 * Sanity-check the size against the specified 779 * geometry. 780 */ 781 if (vnd->sc_size < geomsize) { 782 (void) vn_close(nd.ni_vp, FREAD|FWRITE, 783 p->p_ucred, p); 784 vndunlock(vnd); 785 return (EINVAL); 786 } 787 } else { 788 /* 789 * Size must be at least 2048 DEV_BSIZE blocks 790 * (1M) in order to use this geometry. 791 */ 792 if (vnd->sc_size < (32 * 64)) { 793 vndunlock(vnd); 794 return (EINVAL); 795 } 796 797 vnd->sc_geom.vng_secsize = DEV_BSIZE; 798 vnd->sc_geom.vng_nsectors = 32; 799 vnd->sc_geom.vng_ntracks = 64; 800 vnd->sc_geom.vng_ncylinders = vnd->sc_size / (64 * 32); 801 802 /* 803 * Compute the actual size allowed by this geometry. 804 */ 805 geomsize = 32 * 64 * vnd->sc_geom.vng_ncylinders; 806 } 807 808 /* 809 * Truncate the size to that specified by 810 * the geometry. 811 * XXX Should we even bother with this? 812 */ 813 vnd->sc_size = geomsize; 814 815 if ((error = vndsetcred(vnd, p->p_ucred)) != 0) { 816 (void) vn_close(nd.ni_vp, FREAD|FWRITE, p->p_ucred, p); 817 vndunlock(vnd); 818 return(error); 819 } 820 vndthrottle(vnd, vnd->sc_vp); 821 vio->vnd_size = dbtob(vnd->sc_size); 822 vnd->sc_flags |= VNF_INITED; 823 #ifdef DEBUG 824 if (vnddebug & VDB_INIT) 825 printf("vndioctl: SET vp %p size 0x%lx %d/%d/%d/%d\n", 826 vnd->sc_vp, (unsigned long) vnd->sc_size, 827 vnd->sc_geom.vng_secsize, 828 vnd->sc_geom.vng_nsectors, 829 vnd->sc_geom.vng_ntracks, 830 vnd->sc_geom.vng_ncylinders); 831 #endif 832 833 /* Attach the disk. */ 834 bzero(vnd->sc_xname, sizeof(vnd->sc_xname)); /* XXX */ 835 sprintf(vnd->sc_xname, "vnd%d", unit); /* XXX */ 836 vnd->sc_dkdev.dk_name = vnd->sc_xname; 837 disk_attach(&vnd->sc_dkdev); 838 839 /* Initialize the xfer and buffer pools. */ 840 pool_init(&vnd->sc_vxpool, sizeof(struct vndxfer), 0, 841 0, 0, "vndxpl", 0, NULL, NULL, M_DEVBUF); 842 pool_init(&vnd->sc_vbpool, sizeof(struct vndbuf), 0, 843 0, 0, "vndbpl", 0, NULL, NULL, M_DEVBUF); 844 845 /* Try and read the disklabel. */ 846 vndgetdisklabel(dev); 847 848 vndunlock(vnd); 849 850 break; 851 852 case VNDIOCCLR: 853 if ((error = vndlock(vnd)) != 0) 854 return (error); 855 856 /* 857 * Don't unconfigure if any other partitions are open 858 * or if both the character and block flavors of this 859 * partition are open. 860 */ 861 part = DISKPART(dev); 862 pmask = (1 << part); 863 if ((vnd->sc_dkdev.dk_openmask & ~pmask) || 864 ((vnd->sc_dkdev.dk_bopenmask & pmask) && 865 (vnd->sc_dkdev.dk_copenmask & pmask))) { 866 vndunlock(vnd); 867 return (EBUSY); 868 } 869 870 vndclear(vnd); 871 #ifdef DEBUG 872 if (vnddebug & VDB_INIT) 873 printf("vndioctl: CLRed\n"); 874 #endif 875 876 /* Destroy the xfer and buffer pools. */ 877 pool_destroy(&vnd->sc_vxpool); 878 pool_destroy(&vnd->sc_vbpool); 879 880 /* Detatch the disk. */ 881 disk_detach(&vnd->sc_dkdev); 882 883 vndunlock(vnd); 884 885 break; 886 887 case DIOCGDINFO: 888 *(struct disklabel *)data = *(vnd->sc_dkdev.dk_label); 889 break; 890 891 case DIOCGPART: 892 ((struct partinfo *)data)->disklab = vnd->sc_dkdev.dk_label; 893 ((struct partinfo *)data)->part = 894 &vnd->sc_dkdev.dk_label->d_partitions[DISKPART(dev)]; 895 break; 896 897 case DIOCWDINFO: 898 case DIOCSDINFO: 899 if ((error = vndlock(vnd)) != 0) 900 return (error); 901 902 vnd->sc_flags |= VNF_LABELLING; 903 904 error = setdisklabel(vnd->sc_dkdev.dk_label, 905 (struct disklabel *)data, 0, vnd->sc_dkdev.dk_cpulabel); 906 if (error == 0) { 907 if (cmd == DIOCWDINFO) 908 error = writedisklabel(VNDLABELDEV(dev), 909 vndstrategy, vnd->sc_dkdev.dk_label, 910 vnd->sc_dkdev.dk_cpulabel); 911 } 912 913 vnd->sc_flags &= ~VNF_LABELLING; 914 915 vndunlock(vnd); 916 917 if (error) 918 return (error); 919 break; 920 921 case DIOCWLABEL: 922 if (*(int *)data != 0) 923 vnd->sc_flags |= VNF_WLABEL; 924 else 925 vnd->sc_flags &= ~VNF_WLABEL; 926 break; 927 928 case DIOCGDEFLABEL: 929 vndgetdefaultlabel(vnd, (struct disklabel *)data); 930 break; 931 932 default: 933 return (ENOTTY); 934 } 935 936 return (0); 937 } 938 939 /* 940 * Duplicate the current processes' credentials. Since we are called only 941 * as the result of a SET ioctl and only root can do that, any future access 942 * to this "disk" is essentially as root. Note that credentials may change 943 * if some other uid can write directly to the mapped file (NFS). 944 */ 945 int 946 vndsetcred(vnd, cred) 947 struct vnd_softc *vnd; 948 struct ucred *cred; 949 { 950 struct uio auio; 951 struct iovec aiov; 952 char *tmpbuf; 953 int error; 954 955 vnd->sc_cred = crdup(cred); 956 tmpbuf = malloc(DEV_BSIZE, M_TEMP, M_WAITOK); 957 958 /* XXX: Horrible kludge to establish credentials for NFS */ 959 aiov.iov_base = tmpbuf; 960 aiov.iov_len = min(DEV_BSIZE, dbtob(vnd->sc_size)); 961 auio.uio_iov = &aiov; 962 auio.uio_iovcnt = 1; 963 auio.uio_offset = 0; 964 auio.uio_rw = UIO_READ; 965 auio.uio_segflg = UIO_SYSSPACE; 966 auio.uio_resid = aiov.iov_len; 967 vn_lock(vnd->sc_vp, LK_EXCLUSIVE | LK_RETRY); 968 error = VOP_READ(vnd->sc_vp, &auio, 0, vnd->sc_cred); 969 if (error == 0) { 970 /* 971 * Because vnd does all IO directly through the vnode 972 * we need to flush (at least) the buffer from the above 973 * VOP_READ from the buffer cache to prevent cache 974 * incoherencies. Also, be careful to write dirty 975 * buffers back to stable storage. 976 */ 977 error = vinvalbuf(vnd->sc_vp, V_SAVE, vnd->sc_cred, 978 curproc, 0, 0); 979 } 980 VOP_UNLOCK(vnd->sc_vp, 0); 981 982 free(tmpbuf, M_TEMP); 983 return (error); 984 } 985 986 /* 987 * Set maxactive based on FS type 988 */ 989 void 990 vndthrottle(vnd, vp) 991 struct vnd_softc *vnd; 992 struct vnode *vp; 993 { 994 #ifdef NFS 995 extern int (**nfsv2_vnodeop_p) __P((void *)); 996 997 if (vp->v_op == nfsv2_vnodeop_p) 998 vnd->sc_maxactive = 2; 999 else 1000 #endif 1001 vnd->sc_maxactive = 8; 1002 1003 if (vnd->sc_maxactive < 1) 1004 vnd->sc_maxactive = 1; 1005 } 1006 1007 void 1008 vndshutdown() 1009 { 1010 struct vnd_softc *vnd; 1011 1012 for (vnd = &vnd_softc[0]; vnd < &vnd_softc[numvnd]; vnd++) 1013 if (vnd->sc_flags & VNF_INITED) 1014 vndclear(vnd); 1015 } 1016 1017 void 1018 vndclear(vnd) 1019 struct vnd_softc *vnd; 1020 { 1021 struct vnode *vp = vnd->sc_vp; 1022 struct proc *p = curproc; /* XXX */ 1023 1024 #ifdef DEBUG 1025 if (vnddebug & VDB_FOLLOW) 1026 printf("vndclear(%p): vp %p\n", vnd, vp); 1027 #endif 1028 vnd->sc_flags &= ~VNF_INITED; 1029 if (vp == (struct vnode *)0) 1030 panic("vndioctl: null vp"); 1031 (void) vn_close(vp, FREAD|FWRITE, vnd->sc_cred, p); 1032 crfree(vnd->sc_cred); 1033 vnd->sc_vp = (struct vnode *)0; 1034 vnd->sc_cred = (struct ucred *)0; 1035 vnd->sc_size = 0; 1036 } 1037 1038 int 1039 vndsize(dev) 1040 dev_t dev; 1041 { 1042 struct vnd_softc *sc; 1043 struct disklabel *lp; 1044 int part, unit, omask; 1045 int size; 1046 1047 unit = vndunit(dev); 1048 if (unit >= numvnd) 1049 return (-1); 1050 sc = &vnd_softc[unit]; 1051 1052 if ((sc->sc_flags & VNF_INITED) == 0) 1053 return (-1); 1054 1055 part = DISKPART(dev); 1056 omask = sc->sc_dkdev.dk_openmask & (1 << part); 1057 lp = sc->sc_dkdev.dk_label; 1058 1059 if (omask == 0 && vndopen(dev, 0, S_IFBLK, curproc)) 1060 return (-1); 1061 1062 if (lp->d_partitions[part].p_fstype != FS_SWAP) 1063 size = -1; 1064 else 1065 size = lp->d_partitions[part].p_size * 1066 (lp->d_secsize / DEV_BSIZE); 1067 1068 if (omask == 0 && vndclose(dev, 0, S_IFBLK, curproc)) 1069 return (-1); 1070 1071 return (size); 1072 } 1073 1074 int 1075 vnddump(dev, blkno, va, size) 1076 dev_t dev; 1077 daddr_t blkno; 1078 caddr_t va; 1079 size_t size; 1080 { 1081 1082 /* Not implemented. */ 1083 return ENXIO; 1084 } 1085 1086 void 1087 vndgetdefaultlabel(sc, lp) 1088 struct vnd_softc *sc; 1089 struct disklabel *lp; 1090 { 1091 struct vndgeom *vng = &sc->sc_geom; 1092 struct partition *pp; 1093 1094 bzero(lp, sizeof(*lp)); 1095 1096 lp->d_secperunit = sc->sc_size / (vng->vng_secsize / DEV_BSIZE); 1097 lp->d_secsize = vng->vng_secsize; 1098 lp->d_nsectors = vng->vng_nsectors; 1099 lp->d_ntracks = vng->vng_ntracks; 1100 lp->d_ncylinders = vng->vng_ncylinders; 1101 lp->d_secpercyl = lp->d_ntracks * lp->d_nsectors; 1102 1103 strncpy(lp->d_typename, "vnd", sizeof(lp->d_typename)); 1104 lp->d_type = DTYPE_VND; 1105 strncpy(lp->d_packname, "fictitious", sizeof(lp->d_packname)); 1106 lp->d_rpm = 3600; 1107 lp->d_interleave = 1; 1108 lp->d_flags = 0; 1109 1110 pp = &lp->d_partitions[RAW_PART]; 1111 pp->p_offset = 0; 1112 pp->p_size = lp->d_secperunit; 1113 pp->p_fstype = FS_UNUSED; 1114 lp->d_npartitions = RAW_PART + 1; 1115 1116 lp->d_magic = DISKMAGIC; 1117 lp->d_magic2 = DISKMAGIC; 1118 lp->d_checksum = dkcksum(lp); 1119 } 1120 1121 /* 1122 * Read the disklabel from a vnd. If one is not present, create a fake one. 1123 */ 1124 void 1125 vndgetdisklabel(dev) 1126 dev_t dev; 1127 { 1128 struct vnd_softc *sc = &vnd_softc[vndunit(dev)]; 1129 char *errstring; 1130 struct disklabel *lp = sc->sc_dkdev.dk_label; 1131 struct cpu_disklabel *clp = sc->sc_dkdev.dk_cpulabel; 1132 int i; 1133 1134 bzero(clp, sizeof(*clp)); 1135 1136 vndgetdefaultlabel(sc, lp); 1137 1138 /* 1139 * Call the generic disklabel extraction routine. 1140 */ 1141 errstring = readdisklabel(VNDLABELDEV(dev), vndstrategy, lp, clp); 1142 if (errstring) { 1143 /* 1144 * Lack of disklabel is common, but we print the warning 1145 * anyway, since it might contain other useful information. 1146 */ 1147 printf("%s: %s\n", sc->sc_xname, errstring); 1148 1149 /* 1150 * For historical reasons, if there's no disklabel 1151 * present, all partitions must be FS_BSDFFS and 1152 * occupy the entire disk. 1153 */ 1154 for (i = 0; i < MAXPARTITIONS; i++) { 1155 /* 1156 * Don't wipe out port specific hack (such as 1157 * dos partition hack of i386 port). 1158 */ 1159 if (lp->d_partitions[i].p_fstype != FS_UNUSED) 1160 continue; 1161 1162 lp->d_partitions[i].p_size = lp->d_secperunit; 1163 lp->d_partitions[i].p_offset = 0; 1164 lp->d_partitions[i].p_fstype = FS_BSDFFS; 1165 } 1166 1167 strncpy(lp->d_packname, "default label", 1168 sizeof(lp->d_packname)); 1169 1170 lp->d_checksum = dkcksum(lp); 1171 } 1172 } 1173 1174 /* 1175 * Wait interruptibly for an exclusive lock. 1176 * 1177 * XXX 1178 * Several drivers do this; it should be abstracted and made MP-safe. 1179 */ 1180 static int 1181 vndlock(sc) 1182 struct vnd_softc *sc; 1183 { 1184 int error; 1185 1186 while ((sc->sc_flags & VNF_LOCKED) != 0) { 1187 sc->sc_flags |= VNF_WANTED; 1188 if ((error = tsleep(sc, PRIBIO | PCATCH, "vndlck", 0)) != 0) 1189 return (error); 1190 } 1191 sc->sc_flags |= VNF_LOCKED; 1192 return (0); 1193 } 1194 1195 /* 1196 * Unlock and wake up any waiters. 1197 */ 1198 static void 1199 vndunlock(sc) 1200 struct vnd_softc *sc; 1201 { 1202 1203 sc->sc_flags &= ~VNF_LOCKED; 1204 if ((sc->sc_flags & VNF_WANTED) != 0) { 1205 sc->sc_flags &= ~VNF_WANTED; 1206 wakeup(sc); 1207 } 1208 } 1209