vnd.c revision 1.6
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 * @(#)vn.c 8.6 (Berkeley) 4/1/94 41 */ 42 43 /* 44 * Vnode disk driver. 45 * 46 * Block/character interface to a vnode. Allows one to treat a file 47 * as a disk (e.g. build a filesystem in it, mount it, etc.). 48 * 49 * NOTE 1: This uses the VOP_BMAP/VOP_STRATEGY interface to the vnode 50 * instead of a simple VOP_RDWR. We do this to avoid distorting the 51 * local buffer cache. 52 * 53 * NOTE 2: There is a security issue involved with this driver. 54 * Once mounted all access to the contents of the "mapped" file via 55 * the special file is controlled by the permissions on the special 56 * file, the protection of the mapped file is ignored (effectively, 57 * by using root credentials in all transactions). 58 * 59 * NOTE 3: Doesn't interact with leases, should it? 60 */ 61 #include "vn.h" 62 #if NVN > 0 63 64 #include <sys/param.h> 65 #include <sys/systm.h> 66 #include <sys/namei.h> 67 #include <sys/proc.h> 68 #include <sys/errno.h> 69 #include <sys/dkstat.h> 70 #include <sys/buf.h> 71 #include <sys/malloc.h> 72 #include <sys/ioctl.h> 73 #include <sys/mount.h> 74 #include <sys/vnode.h> 75 #include <sys/file.h> 76 #include <sys/uio.h> 77 78 #include <miscfs/specfs/specdev.h> 79 80 #include <dev/vnioctl.h> 81 82 #ifdef DEBUG 83 int dovncluster = 1; 84 int vndebug = 0x00; 85 #define VDB_FOLLOW 0x01 86 #define VDB_INIT 0x02 87 #define VDB_IO 0x04 88 #endif 89 90 #define b_cylin b_resid 91 92 #define vnunit(x) ((minor(x) >> 3) & 0x7) /* for consistency */ 93 94 #define getvnbuf() \ 95 ((struct buf *)malloc(sizeof(struct buf), M_DEVBUF, M_WAITOK)) 96 #define putvnbuf(bp) \ 97 free((caddr_t)(bp), M_DEVBUF) 98 99 struct vn_softc { 100 int sc_flags; /* flags */ 101 size_t sc_size; /* size of vn */ 102 struct vnode *sc_vp; /* vnode */ 103 struct ucred *sc_cred; /* credentials */ 104 int sc_maxactive; /* max # of active requests */ 105 struct buf sc_tab; /* transfer queue */ 106 }; 107 108 /* sc_flags */ 109 #define VNF_ALIVE 0x01 110 #define VNF_INITED 0x02 111 112 #if 0 /* if you need static allocation */ 113 struct vn_softc vn_softc[NVN]; 114 int numvnd = NVN; 115 #else 116 struct vn_softc *vn_softc; 117 int numvnd; 118 #endif 119 120 void 121 vnattach(num) 122 int num; 123 { 124 char *mem; 125 register u_long size; 126 127 if (num <= 0) 128 return; 129 size = num * sizeof(struct vn_softc); 130 mem = malloc(size, M_DEVBUF, M_NOWAIT); 131 if (mem == NULL) { 132 printf("WARNING: no memory for vnode disks\n"); 133 return; 134 } 135 bzero(mem, size); 136 vn_softc = (struct vn_softc *)mem; 137 numvnd = num; 138 } 139 140 int 141 vnopen(dev, flags, mode, p) 142 dev_t dev; 143 int flags, mode; 144 struct proc *p; 145 { 146 int unit = vnunit(dev); 147 148 #ifdef DEBUG 149 if (vndebug & VDB_FOLLOW) 150 printf("vnopen(%x, %x, %x, %x)\n", dev, flags, mode, p); 151 #endif 152 if (unit >= numvnd) 153 return(ENXIO); 154 return(0); 155 } 156 157 int 158 vnclose(dev, flags, mode, p) 159 dev_t dev; 160 int flags, mode; 161 struct proc *p; 162 { 163 #ifdef DEBUG 164 if (vndebug & VDB_FOLLOW) 165 printf("vnclose(%x, %x, %x, %x)\n", dev, flags, mode, p); 166 #endif 167 return 0; 168 } 169 170 /* 171 * Break the request into bsize pieces and submit using VOP_BMAP/VOP_STRATEGY. 172 * Note that this driver can only be used for swapping over NFS on the hp 173 * since nfs_strategy on the vax cannot handle u-areas and page tables. 174 */ 175 void 176 vnstrategy(bp) 177 register struct buf *bp; 178 { 179 int unit = vnunit(bp->b_dev); 180 register struct vn_softc *vn = &vn_softc[unit]; 181 register struct buf *nbp; 182 register int bn, bsize, resid; 183 register caddr_t addr; 184 int sz, flags, error; 185 extern void vniodone(); 186 187 #ifdef DEBUG 188 if (vndebug & VDB_FOLLOW) 189 printf("vnstrategy(%x): unit %d\n", bp, unit); 190 #endif 191 if ((vn->sc_flags & VNF_INITED) == 0) { 192 bp->b_error = ENXIO; 193 bp->b_flags |= B_ERROR; 194 biodone(bp); 195 return; 196 } 197 bn = bp->b_blkno; 198 sz = howmany(bp->b_bcount, DEV_BSIZE); 199 bp->b_resid = bp->b_bcount; 200 if (bn < 0 || bn + sz > vn->sc_size) { 201 if (bn != vn->sc_size) { 202 bp->b_error = EINVAL; 203 bp->b_flags |= B_ERROR; 204 } 205 biodone(bp); 206 return; 207 } 208 bn = dbtob(bn); 209 bsize = vn->sc_vp->v_mount->mnt_stat.f_iosize; 210 addr = bp->b_data; 211 flags = bp->b_flags | B_CALL; 212 for (resid = bp->b_resid; resid; resid -= sz) { 213 struct vnode *vp; 214 daddr_t nbn; 215 int off, s, nra; 216 217 nra = 0; 218 #if (BSD > 199103) 219 error = VOP_BMAP(vn->sc_vp, bn / bsize, &vp, &nbn, &nra); 220 #else 221 error = VOP_BMAP(vn->sc_vp, bn / bsize, &vp, &nbn); 222 #endif 223 if (error == 0 && (long)nbn == -1) 224 error = EIO; 225 #ifdef DEBUG 226 if (!dovncluster) 227 nra = 0; 228 #endif 229 230 if (off = bn % bsize) 231 sz = bsize - off; 232 else 233 sz = (1 + nra) * bsize; 234 if (resid < sz) 235 sz = resid; 236 #ifdef DEBUG 237 if (vndebug & VDB_IO) 238 printf("vnstrategy: vp %x/%x bn %x/%x sz %x\n", 239 vn->sc_vp, vp, bn, nbn, sz); 240 #endif 241 242 nbp = getvnbuf(); 243 nbp->b_flags = flags; 244 nbp->b_bcount = sz; 245 nbp->b_bufsize = bp->b_bufsize; 246 nbp->b_error = 0; 247 if (vp->v_type == VBLK || vp->v_type == VCHR) 248 nbp->b_dev = vp->v_rdev; 249 else 250 nbp->b_dev = NODEV; 251 nbp->b_data = addr; 252 nbp->b_blkno = nbn + btodb(off); 253 nbp->b_proc = bp->b_proc; 254 nbp->b_iodone = vniodone; 255 nbp->b_vp = vp; 256 nbp->b_pfcent = (int) bp; /* XXX */ 257 nbp->b_rcred = vn->sc_cred; /* XXX crdup? */ 258 nbp->b_wcred = vn->sc_cred; /* XXX crdup? */ 259 nbp->b_dirtyoff = bp->b_dirtyoff; 260 nbp->b_dirtyend = bp->b_dirtyend; 261 nbp->b_validoff = bp->b_validoff; 262 nbp->b_validend = bp->b_validend; 263 /* 264 * If there was an error or a hole in the file...punt. 265 * Note that we deal with this after the nbp allocation. 266 * This ensures that we properly clean up any operations 267 * that we have already fired off. 268 * 269 * XXX we could deal with holes here but it would be 270 * a hassle (in the write case). 271 */ 272 if (error) { 273 nbp->b_error = error; 274 nbp->b_flags |= B_ERROR; 275 bp->b_resid -= (resid - sz); 276 biodone(nbp); 277 return; 278 } 279 /* 280 * Just sort by block number 281 */ 282 nbp->b_cylin = nbp->b_blkno; 283 s = splbio(); 284 disksort(&vn->sc_tab, nbp); 285 if (vn->sc_tab.b_active < vn->sc_maxactive) { 286 vn->sc_tab.b_active++; 287 vnstart(vn); 288 } 289 splx(s); 290 bn += sz; 291 addr += sz; 292 } 293 } 294 295 /* 296 * Feed requests sequentially. 297 * We do it this way to keep from flooding NFS servers if we are connected 298 * to an NFS file. This places the burden on the client rather than the 299 * server. 300 */ 301 vnstart(vn) 302 register struct vn_softc *vn; 303 { 304 register struct buf *bp; 305 306 /* 307 * Dequeue now since lower level strategy routine might 308 * queue using same links 309 */ 310 bp = vn->sc_tab.b_actf; 311 vn->sc_tab.b_actf = bp->b_actf; 312 #ifdef DEBUG 313 if (vndebug & VDB_IO) 314 printf("vnstart(%d): bp %x vp %x blkno %x addr %x cnt %x\n", 315 vn-vn_softc, bp, bp->b_vp, bp->b_blkno, bp->b_data, 316 bp->b_bcount); 317 #endif 318 if ((bp->b_flags & B_READ) == 0) 319 bp->b_vp->v_numoutput++; 320 VOP_STRATEGY(bp); 321 } 322 323 void 324 vniodone(bp) 325 register struct buf *bp; 326 { 327 register struct buf *pbp = (struct buf *)bp->b_pfcent; /* XXX */ 328 register struct vn_softc *vn = &vn_softc[vnunit(pbp->b_dev)]; 329 int s; 330 331 s = splbio(); 332 #ifdef DEBUG 333 if (vndebug & VDB_IO) 334 printf("vniodone(%d): bp %x vp %x blkno %x addr %x cnt %x\n", 335 vn-vn_softc, bp, bp->b_vp, bp->b_blkno, bp->b_data, 336 bp->b_bcount); 337 #endif 338 if (bp->b_error) { 339 #ifdef DEBUG 340 if (vndebug & VDB_IO) 341 printf("vniodone: bp %x error %d\n", bp, bp->b_error); 342 #endif 343 pbp->b_flags |= B_ERROR; 344 pbp->b_error = biowait(bp); 345 } 346 pbp->b_resid -= bp->b_bcount; 347 putvnbuf(bp); 348 if (pbp->b_resid == 0) { 349 #ifdef DEBUG 350 if (vndebug & VDB_IO) 351 printf("vniodone: pbp %x iodone\n", pbp); 352 #endif 353 biodone(pbp); 354 } 355 if (vn->sc_tab.b_actf) 356 vnstart(vn); 357 else 358 vn->sc_tab.b_active--; 359 splx(s); 360 } 361 362 vnread(dev, uio, flags, p) 363 dev_t dev; 364 struct uio *uio; 365 int flags; 366 struct proc *p; 367 { 368 369 #ifdef DEBUG 370 if (vndebug & VDB_FOLLOW) 371 printf("vnread(%x, %x, %x, %x)\n", dev, uio, flags, p); 372 #endif 373 return(physio(vnstrategy, NULL, dev, B_READ, minphys, uio)); 374 } 375 376 vnwrite(dev, uio, flags, p) 377 dev_t dev; 378 struct uio *uio; 379 int flags; 380 struct proc *p; 381 { 382 383 #ifdef DEBUG 384 if (vndebug & VDB_FOLLOW) 385 printf("vnwrite(%x, %x, %x, %x)\n", dev, uio, flags, p); 386 #endif 387 return(physio(vnstrategy, NULL, dev, B_WRITE, minphys, uio)); 388 } 389 390 /* ARGSUSED */ 391 vnioctl(dev, cmd, data, flag, p) 392 dev_t dev; 393 u_long cmd; 394 caddr_t data; 395 int flag; 396 struct proc *p; 397 { 398 int unit = vnunit(dev); 399 register struct vn_softc *vn; 400 struct vn_ioctl *vio; 401 struct vattr vattr; 402 struct nameidata nd; 403 int error; 404 405 #ifdef DEBUG 406 if (vndebug & VDB_FOLLOW) 407 printf("vnioctl(%x, %x, %x, %x, %x): unit %d\n", 408 dev, cmd, data, flag, p, unit); 409 #endif 410 error = suser(p->p_ucred, &p->p_acflag); 411 if (error) 412 return (error); 413 if (unit >= numvnd) 414 return (ENXIO); 415 416 vn = &vn_softc[unit]; 417 vio = (struct vn_ioctl *)data; 418 switch (cmd) { 419 420 case VNIOCSET: 421 if (vn->sc_flags & VNF_INITED) 422 return(EBUSY); 423 /* 424 * Always open for read and write. 425 * This is probably bogus, but it lets vn_open() 426 * weed out directories, sockets, etc. so we don't 427 * have to worry about them. 428 */ 429 #if (BSD > 199103) 430 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, vio->vn_file, p); 431 if (error = vn_open(&nd, FREAD|FWRITE, 0)) 432 return(error); 433 #else 434 nd.ni_nameiop = LOOKUP | FOLLOW; 435 nd.ni_segflg = UIO_USERSPACE; 436 nd.ni_dirp = vio->vn_file; 437 if (error = vn_open(&nd, p, FREAD|FWRITE, 0)) 438 return(error); 439 #endif 440 if (error = VOP_GETATTR(nd.ni_vp, &vattr, p->p_ucred, p)) { 441 VOP_UNLOCK(nd.ni_vp); 442 (void) vn_close(nd.ni_vp, FREAD|FWRITE, p->p_ucred, p); 443 return(error); 444 } 445 VOP_UNLOCK(nd.ni_vp); 446 vn->sc_vp = nd.ni_vp; 447 vn->sc_size = btodb(vattr.va_size); /* note truncation */ 448 if (error = vnsetcred(vn, p->p_ucred)) { 449 (void) vn_close(nd.ni_vp, FREAD|FWRITE, p->p_ucred, p); 450 return(error); 451 } 452 vnthrottle(vn, vn->sc_vp); 453 vio->vn_size = dbtob(vn->sc_size); 454 vn->sc_flags |= VNF_INITED; 455 #ifdef DEBUG 456 if (vndebug & VDB_INIT) 457 printf("vnioctl: SET vp %x size %x\n", 458 vn->sc_vp, vn->sc_size); 459 #endif 460 break; 461 462 case VNIOCCLR: 463 if ((vn->sc_flags & VNF_INITED) == 0) 464 return(ENXIO); 465 vnclear(vn); 466 #ifdef DEBUG 467 if (vndebug & VDB_INIT) 468 printf("vnioctl: CLRed\n"); 469 #endif 470 break; 471 472 default: 473 return(ENXIO); 474 } 475 return(0); 476 } 477 478 /* 479 * Duplicate the current processes' credentials. Since we are called only 480 * as the result of a SET ioctl and only root can do that, any future access 481 * to this "disk" is essentially as root. Note that credentials may change 482 * if some other uid can write directly to the mapped file (NFS). 483 */ 484 vnsetcred(vn, cred) 485 register struct vn_softc *vn; 486 struct ucred *cred; 487 { 488 struct uio auio; 489 struct iovec aiov; 490 char *tmpbuf; 491 int error; 492 493 vn->sc_cred = crdup(cred); 494 tmpbuf = malloc(DEV_BSIZE, M_TEMP, M_WAITOK); 495 496 /* XXX: Horrible kludge to establish credentials for NFS */ 497 aiov.iov_base = tmpbuf; 498 aiov.iov_len = min(DEV_BSIZE, dbtob(vn->sc_size)); 499 auio.uio_iov = &aiov; 500 auio.uio_iovcnt = 1; 501 auio.uio_offset = 0; 502 auio.uio_rw = UIO_READ; 503 auio.uio_segflg = UIO_SYSSPACE; 504 auio.uio_resid = aiov.iov_len; 505 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred); 506 507 free(tmpbuf, M_TEMP); 508 return (error); 509 } 510 511 /* 512 * Set maxactive based on FS type 513 */ 514 vnthrottle(vn, vp) 515 register struct vn_softc *vn; 516 struct vnode *vp; 517 { 518 #ifdef NFSCLIENT 519 #if (BSD > 199103) 520 extern int (**nfsv2_vnodeop_p)(); 521 522 if (vp->v_op == nfsv2_vnodeop_p) 523 #else 524 extern struct vnodeops nfsv2_vnodeops; 525 526 if (vp->v_op == &nfsv2_vnodeops) 527 #endif 528 vn->sc_maxactive = 2; 529 else 530 #endif 531 vn->sc_maxactive = 8; 532 533 if (vn->sc_maxactive < 1) 534 vn->sc_maxactive = 1; 535 } 536 537 vnshutdown() 538 { 539 register struct vn_softc *vn; 540 541 for (vn = &vn_softc[0]; vn < &vn_softc[numvnd]; vn++) 542 if (vn->sc_flags & VNF_INITED) 543 vnclear(vn); 544 } 545 546 vnclear(vn) 547 register struct vn_softc *vn; 548 { 549 register struct vnode *vp = vn->sc_vp; 550 struct proc *p = curproc; /* XXX */ 551 552 #ifdef DEBUG 553 if (vndebug & VDB_FOLLOW) 554 printf("vnclear(%x): vp %x\n", vp); 555 #endif 556 vn->sc_flags &= ~VNF_INITED; 557 if (vp == (struct vnode *)0) 558 panic("vnioctl: null vp"); 559 (void) vn_close(vp, FREAD|FWRITE, vn->sc_cred, p); 560 crfree(vn->sc_cred); 561 vn->sc_vp = (struct vnode *)0; 562 vn->sc_cred = (struct ucred *)0; 563 vn->sc_size = 0; 564 } 565 566 vnsize(dev) 567 dev_t dev; 568 { 569 int unit = vnunit(dev); 570 register struct vn_softc *vn = &vn_softc[unit]; 571 572 if (unit >= numvnd || (vn->sc_flags & VNF_INITED) == 0) 573 return(-1); 574 return(vn->sc_size); 575 } 576 577 vndump(dev) 578 { 579 return(ENXIO); 580 } 581 #endif 582
Indexes created Tue Sep 30 11:09:46 GMT 2025