vfs_vnops.c revision 1.135.2.2
1 /* $NetBSD: vfs_vnops.c,v 1.135.2.2 2007/03/21 20:11:54 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_vnops.c 8.14 (Berkeley) 6/15/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.135.2.2 2007/03/21 20:11:54 ad Exp $"); 41 42 #include "fs_union.h" 43 #include "veriexec.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/kernel.h> 48 #include <sys/file.h> 49 #include <sys/stat.h> 50 #include <sys/buf.h> 51 #include <sys/proc.h> 52 #include <sys/malloc.h> 53 #include <sys/mount.h> 54 #include <sys/namei.h> 55 #include <sys/vnode.h> 56 #include <sys/ioctl.h> 57 #include <sys/tty.h> 58 #include <sys/poll.h> 59 #include <sys/kauth.h> 60 #include <sys/syslog.h> 61 62 #include <miscfs/specfs/specdev.h> 63 64 #include <uvm/uvm_extern.h> 65 #include <uvm/uvm_readahead.h> 66 67 #ifdef UNION 68 #include <fs/union/union.h> 69 #endif 70 71 #if defined(LKM) || defined(UNION) 72 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct lwp *); 73 #endif 74 75 #if NVERIEXEC > 0 76 #include <sys/verified_exec.h> 77 #endif /* NVERIEXEC > 0 */ 78 79 static int vn_read(struct file *fp, off_t *offset, struct uio *uio, 80 kauth_cred_t cred, int flags); 81 static int vn_write(struct file *fp, off_t *offset, struct uio *uio, 82 kauth_cred_t cred, int flags); 83 static int vn_closefile(struct file *fp, struct lwp *l); 84 static int vn_poll(struct file *fp, int events, struct lwp *l); 85 static int vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l); 86 static int vn_statfile(struct file *fp, struct stat *sb, struct lwp *l); 87 static int vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l); 88 89 const struct fileops vnops = { 90 vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, 91 vn_statfile, vn_closefile, vn_kqfilter 92 }; 93 94 /* 95 * Common code for vnode open operations. 96 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 97 */ 98 int 99 vn_open(struct nameidata *ndp, int fmode, int cmode) 100 { 101 struct vnode *vp; 102 struct mount *mp = NULL; /* XXX: GCC */ 103 struct lwp *l = ndp->ni_cnd.cn_lwp; 104 kauth_cred_t cred = l->l_cred; 105 struct vattr va; 106 int error; 107 pathname_t pn = NULL; 108 109 restart: 110 if (fmode & O_CREAT) { 111 ndp->ni_cnd.cn_nameiop = CREATE; 112 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 113 if ((fmode & O_EXCL) == 0 && 114 ((fmode & O_NOFOLLOW) == 0)) 115 ndp->ni_cnd.cn_flags |= FOLLOW; 116 } else { 117 ndp->ni_cnd.cn_nameiop = LOOKUP; 118 ndp->ni_cnd.cn_flags = LOCKLEAF; 119 if ((fmode & O_NOFOLLOW) == 0) 120 ndp->ni_cnd.cn_flags |= FOLLOW; 121 } 122 #if NVERIEXEC > 0 123 error = pathname_get(ndp->ni_dirp, ndp->ni_segflg, &pn); 124 if (error) 125 goto bad2; 126 ndp->ni_dirp = pathname_path(pn); 127 ndp->ni_segflg = UIO_SYSSPACE; 128 #endif /* NVERIEXEC > 0 */ 129 error = namei(ndp); 130 if (error) 131 goto bad2; 132 133 vp = ndp->ni_vp; 134 135 #if NVERIEXEC > 0 136 error = veriexec_openchk(l, ndp->ni_vp, ndp->ni_dirp, fmode); 137 if (error) 138 goto bad; 139 #endif /* NVERIEXEC > 0 */ 140 141 if (fmode & O_CREAT) { 142 if (ndp->ni_vp == NULL) { 143 VATTR_NULL(&va); 144 va.va_type = VREG; 145 va.va_mode = cmode; 146 if (fmode & O_EXCL) 147 va.va_vaflags |= VA_EXCLUSIVE; 148 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 149 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 150 vput(ndp->ni_dvp); 151 if ((error = vn_start_write(NULL, &mp, 152 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 153 goto bad2; 154 goto restart; 155 } 156 VOP_LEASE(ndp->ni_dvp, l, cred, LEASE_WRITE); 157 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 158 &ndp->ni_cnd, &va); 159 vn_finished_write(mp, 0); 160 if (error) 161 goto bad2; 162 fmode &= ~O_TRUNC; 163 vp = ndp->ni_vp; 164 } else { 165 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 166 if (ndp->ni_dvp == ndp->ni_vp) 167 vrele(ndp->ni_dvp); 168 else 169 vput(ndp->ni_dvp); 170 ndp->ni_dvp = NULL; 171 vp = ndp->ni_vp; 172 if (fmode & O_EXCL) { 173 error = EEXIST; 174 goto bad; 175 } 176 fmode &= ~O_CREAT; 177 } 178 } else { 179 vp = ndp->ni_vp; 180 } 181 if (vp->v_type == VSOCK) { 182 error = EOPNOTSUPP; 183 goto bad; 184 } 185 if (ndp->ni_vp->v_type == VLNK) { 186 error = EFTYPE; 187 goto bad; 188 } 189 190 if ((fmode & O_CREAT) == 0) { 191 if (fmode & FREAD) { 192 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 193 goto bad; 194 } 195 196 if (fmode & (FWRITE | O_TRUNC)) { 197 if (vp->v_type == VDIR) { 198 error = EISDIR; 199 goto bad; 200 } 201 if ((error = vn_writechk(vp)) != 0 || 202 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 203 goto bad; 204 } 205 } 206 207 if (fmode & O_TRUNC) { 208 VOP_UNLOCK(vp, 0); /* XXX */ 209 210 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 211 vrele(vp); 212 goto bad2; 213 } 214 VOP_LEASE(vp, l, cred, LEASE_WRITE); 215 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 216 VATTR_NULL(&va); 217 va.va_size = 0; 218 error = VOP_SETATTR(vp, &va, cred, l); 219 vn_finished_write(mp, 0); 220 if (error != 0) 221 goto bad; 222 } 223 if ((error = VOP_OPEN(vp, fmode, cred, l)) != 0) 224 goto bad; 225 if (vp->v_type == VREG && 226 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 227 error = EIO; 228 goto bad; 229 } 230 if (fmode & FWRITE) 231 vp->v_writecount++; 232 233 bad: 234 if (error) 235 vput(vp); 236 237 bad2: 238 pathname_put(pn); 239 240 return (error); 241 } 242 243 /* 244 * Check for write permissions on the specified vnode. 245 * Prototype text segments cannot be written. 246 */ 247 int 248 vn_writechk(struct vnode *vp) 249 { 250 251 /* 252 * If the vnode is in use as a process's text, 253 * we can't allow writing. 254 */ 255 if (vp->v_flag & VTEXT) 256 return (ETXTBSY); 257 return (0); 258 } 259 260 /* 261 * Mark a vnode as having executable mappings. 262 */ 263 void 264 vn_markexec(struct vnode *vp) 265 { 266 if ((vp->v_flag & VEXECMAP) == 0) { 267 uvmexp.filepages -= vp->v_uobj.uo_npages; 268 uvmexp.execpages += vp->v_uobj.uo_npages; 269 } 270 vp->v_flag |= VEXECMAP; 271 } 272 273 /* 274 * Mark a vnode as being the text of a process. 275 * Fail if the vnode is currently writable. 276 */ 277 int 278 vn_marktext(struct vnode *vp) 279 { 280 281 if (vp->v_writecount != 0) { 282 KASSERT((vp->v_flag & VTEXT) == 0); 283 return (ETXTBSY); 284 } 285 vp->v_flag |= VTEXT; 286 vn_markexec(vp); 287 return (0); 288 } 289 290 /* 291 * Vnode close call 292 * 293 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 294 */ 295 int 296 vn_close(struct vnode *vp, int flags, kauth_cred_t cred, struct lwp *l) 297 { 298 int error; 299 300 if (flags & FWRITE) 301 vp->v_writecount--; 302 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 303 error = VOP_CLOSE(vp, flags, cred, l); 304 vput(vp); 305 return (error); 306 } 307 308 /* 309 * Package up an I/O request on a vnode into a uio and do it. 310 */ 311 int 312 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset, 313 enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid, 314 struct lwp *l) 315 { 316 struct uio auio; 317 struct iovec aiov; 318 struct mount *mp = NULL; 319 int error; 320 321 if ((ioflg & IO_NODELOCKED) == 0) { 322 if (rw == UIO_READ) { 323 vn_lock(vp, LK_SHARED | LK_RETRY); 324 } else /* UIO_WRITE */ { 325 if (vp->v_type != VCHR && 326 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 327 != 0) 328 return (error); 329 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 330 } 331 } 332 auio.uio_iov = &aiov; 333 auio.uio_iovcnt = 1; 334 aiov.iov_base = base; 335 aiov.iov_len = len; 336 auio.uio_resid = len; 337 auio.uio_offset = offset; 338 auio.uio_rw = rw; 339 if (segflg == UIO_SYSSPACE) { 340 UIO_SETUP_SYSSPACE(&auio); 341 } else { 342 auio.uio_vmspace = l->l_proc->p_vmspace; 343 } 344 if (rw == UIO_READ) { 345 error = VOP_READ(vp, &auio, ioflg, cred); 346 } else { 347 error = VOP_WRITE(vp, &auio, ioflg, cred); 348 } 349 if (aresid) 350 *aresid = auio.uio_resid; 351 else 352 if (auio.uio_resid && error == 0) 353 error = EIO; 354 if ((ioflg & IO_NODELOCKED) == 0) { 355 if (rw == UIO_WRITE) 356 vn_finished_write(mp, 0); 357 VOP_UNLOCK(vp, 0); 358 } 359 return (error); 360 } 361 362 int 363 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 364 struct lwp *l, off_t **cookies, int *ncookies) 365 { 366 struct vnode *vp = (struct vnode *)fp->f_data; 367 struct iovec aiov; 368 struct uio auio; 369 int error, eofflag; 370 371 /* Limit the size on any kernel buffers used by VOP_READDIR */ 372 count = min(MAXBSIZE, count); 373 374 unionread: 375 if (vp->v_type != VDIR) 376 return (EINVAL); 377 aiov.iov_base = bf; 378 aiov.iov_len = count; 379 auio.uio_iov = &aiov; 380 auio.uio_iovcnt = 1; 381 auio.uio_rw = UIO_READ; 382 if (segflg == UIO_SYSSPACE) { 383 UIO_SETUP_SYSSPACE(&auio); 384 } else { 385 KASSERT(l == curlwp); 386 auio.uio_vmspace = l->l_proc->p_vmspace; 387 } 388 auio.uio_resid = count; 389 vn_lock(vp, LK_SHARED | LK_RETRY); 390 auio.uio_offset = fp->f_offset; 391 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 392 ncookies); 393 mutex_enter(&fp->f_lock); 394 fp->f_offset = auio.uio_offset; 395 mutex_exit(&fp->f_lock); 396 VOP_UNLOCK(vp, 0); 397 if (error) 398 return (error); 399 400 #if defined(UNION) || defined(LKM) 401 if (count == auio.uio_resid && vn_union_readdir_hook) { 402 struct vnode *ovp = vp; 403 404 error = (*vn_union_readdir_hook)(&vp, fp, l); 405 if (error) 406 return (error); 407 if (vp != ovp) 408 goto unionread; 409 } 410 #endif /* UNION || LKM */ 411 412 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 413 (vp->v_mount->mnt_flag & MNT_UNION)) { 414 struct vnode *tvp = vp; 415 vp = vp->v_mount->mnt_vnodecovered; 416 VREF(vp); 417 fp->f_data = vp; 418 mutex_enter(&fp->f_lock); 419 fp->f_offset = 0; 420 mutex_exit(&fp->f_lock); 421 vrele(tvp); 422 goto unionread; 423 } 424 *done = count - auio.uio_resid; 425 return error; 426 } 427 428 /* 429 * File table vnode read routine. 430 */ 431 static int 432 vn_read(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 433 int flags) 434 { 435 struct vnode *vp = (struct vnode *)fp->f_data; 436 int count, error, ioflag; 437 struct lwp *l = curlwp; 438 439 VOP_LEASE(vp, l, cred, LEASE_READ); 440 ioflag = IO_ADV_ENCODE(fp->f_advice); 441 if (fp->f_flag & FNONBLOCK) 442 ioflag |= IO_NDELAY; 443 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 444 ioflag |= IO_SYNC; 445 if (fp->f_flag & FALTIO) 446 ioflag |= IO_ALTSEMANTICS; 447 if (fp->f_flag & FDIRECT) 448 ioflag |= IO_DIRECT; 449 vn_lock(vp, LK_SHARED | LK_RETRY); 450 uio->uio_offset = *offset; 451 count = uio->uio_resid; 452 error = VOP_READ(vp, uio, ioflag, cred); 453 if (flags & FOF_UPDATE_OFFSET) 454 *offset += count - uio->uio_resid; 455 VOP_UNLOCK(vp, 0); 456 return (error); 457 } 458 459 /* 460 * File table vnode write routine. 461 */ 462 static int 463 vn_write(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 464 int flags) 465 { 466 struct vnode *vp = (struct vnode *)fp->f_data; 467 struct mount *mp; 468 int count, error, ioflag = IO_UNIT; 469 struct lwp *l = curlwp; 470 471 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 472 ioflag |= IO_APPEND; 473 if (fp->f_flag & FNONBLOCK) 474 ioflag |= IO_NDELAY; 475 if (fp->f_flag & FFSYNC || 476 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 477 ioflag |= IO_SYNC; 478 else if (fp->f_flag & FDSYNC) 479 ioflag |= IO_DSYNC; 480 if (fp->f_flag & FALTIO) 481 ioflag |= IO_ALTSEMANTICS; 482 if (fp->f_flag & FDIRECT) 483 ioflag |= IO_DIRECT; 484 mp = NULL; 485 if (vp->v_type != VCHR && 486 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 487 return (error); 488 VOP_LEASE(vp, l, cred, LEASE_WRITE); 489 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 490 uio->uio_offset = *offset; 491 count = uio->uio_resid; 492 error = VOP_WRITE(vp, uio, ioflag, cred); 493 if (flags & FOF_UPDATE_OFFSET) { 494 if (ioflag & IO_APPEND) 495 *offset = uio->uio_offset; 496 else 497 *offset += count - uio->uio_resid; 498 } 499 VOP_UNLOCK(vp, 0); 500 vn_finished_write(mp, 0); 501 return (error); 502 } 503 504 /* 505 * File table vnode stat routine. 506 */ 507 static int 508 vn_statfile(struct file *fp, struct stat *sb, struct lwp *l) 509 { 510 struct vnode *vp = (struct vnode *)fp->f_data; 511 512 return vn_stat(vp, sb, l); 513 } 514 515 int 516 vn_stat(struct vnode *vp, struct stat *sb, struct lwp *l) 517 { 518 struct vattr va; 519 int error; 520 mode_t mode; 521 522 error = VOP_GETATTR(vp, &va, l->l_cred, l); 523 if (error) 524 return (error); 525 /* 526 * Copy from vattr table 527 */ 528 sb->st_dev = va.va_fsid; 529 sb->st_ino = va.va_fileid; 530 mode = va.va_mode; 531 switch (vp->v_type) { 532 case VREG: 533 mode |= S_IFREG; 534 break; 535 case VDIR: 536 mode |= S_IFDIR; 537 break; 538 case VBLK: 539 mode |= S_IFBLK; 540 break; 541 case VCHR: 542 mode |= S_IFCHR; 543 break; 544 case VLNK: 545 mode |= S_IFLNK; 546 break; 547 case VSOCK: 548 mode |= S_IFSOCK; 549 break; 550 case VFIFO: 551 mode |= S_IFIFO; 552 break; 553 default: 554 return (EBADF); 555 }; 556 sb->st_mode = mode; 557 sb->st_nlink = va.va_nlink; 558 sb->st_uid = va.va_uid; 559 sb->st_gid = va.va_gid; 560 sb->st_rdev = va.va_rdev; 561 sb->st_size = va.va_size; 562 sb->st_atimespec = va.va_atime; 563 sb->st_mtimespec = va.va_mtime; 564 sb->st_ctimespec = va.va_ctime; 565 sb->st_birthtimespec = va.va_birthtime; 566 sb->st_blksize = va.va_blocksize; 567 sb->st_flags = va.va_flags; 568 sb->st_gen = 0; 569 sb->st_blocks = va.va_bytes / S_BLKSIZE; 570 return (0); 571 } 572 573 /* 574 * File table vnode fcntl routine. 575 */ 576 static int 577 vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l) 578 { 579 struct vnode *vp = ((struct vnode *)fp->f_data); 580 int error; 581 582 error = VOP_FCNTL(vp, com, data, fp->f_flag, l->l_cred, l); 583 return (error); 584 } 585 586 /* 587 * File table vnode ioctl routine. 588 */ 589 static int 590 vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l) 591 { 592 struct vnode *vp = ((struct vnode *)fp->f_data), *ovp; 593 struct proc *p = l->l_proc; 594 struct vattr vattr; 595 int error; 596 597 switch (vp->v_type) { 598 599 case VREG: 600 case VDIR: 601 if (com == FIONREAD) { 602 error = VOP_GETATTR(vp, &vattr, l->l_cred, l); 603 if (error) 604 return (error); 605 *(int *)data = vattr.va_size - fp->f_offset; 606 return (0); 607 } 608 if ((com == FIONWRITE) || (com == FIONSPACE)) { 609 /* 610 * Files don't have send queues, so there never 611 * are any bytes in them, nor is there any 612 * open space in them. 613 */ 614 *(int *)data = 0; 615 return (0); 616 } 617 if (com == FIOGETBMAP) { 618 daddr_t *block; 619 620 if (*(daddr_t *)data < 0) 621 return (EINVAL); 622 block = (daddr_t *)data; 623 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 624 } 625 if (com == OFIOGETBMAP) { 626 daddr_t ibn, obn; 627 628 if (*(int32_t *)data < 0) 629 return (EINVAL); 630 ibn = (daddr_t)*(int32_t *)data; 631 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 632 *(int32_t *)data = (int32_t)obn; 633 return error; 634 } 635 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 636 return (0); /* XXX */ 637 /* fall into ... */ 638 case VFIFO: 639 case VCHR: 640 case VBLK: 641 error = VOP_IOCTL(vp, com, data, fp->f_flag, 642 l->l_cred, l); 643 if (error == 0 && com == TIOCSCTTY) { 644 VREF(vp); 645 mutex_enter(&proclist_lock); 646 ovp = p->p_session->s_ttyvp; 647 p->p_session->s_ttyvp = vp; 648 mutex_exit(&proclist_lock); 649 if (ovp != NULL) 650 vrele(ovp); 651 } 652 return (error); 653 654 default: 655 return (EPASSTHROUGH); 656 } 657 } 658 659 /* 660 * File table vnode poll routine. 661 */ 662 static int 663 vn_poll(struct file *fp, int events, struct lwp *l) 664 { 665 666 return (VOP_POLL(((struct vnode *)fp->f_data), events, l)); 667 } 668 669 /* 670 * File table vnode kqfilter routine. 671 */ 672 int 673 vn_kqfilter(struct file *fp, struct knote *kn) 674 { 675 676 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 677 } 678 679 /* 680 * Check that the vnode is still valid, and if so 681 * acquire requested lock. 682 */ 683 int 684 vn_lock(struct vnode *vp, int flags) 685 { 686 int error; 687 688 #if 0 689 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 690 || (vp->v_flag & VONWORKLST) != 0); 691 #endif 692 KASSERT((flags & 693 ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_DRAIN|LK_NOWAIT|LK_RETRY| 694 LK_SETRECURSE|LK_CANRECURSE)) 695 == 0); 696 697 do { 698 if ((flags & LK_INTERLOCK) == 0) 699 mutex_enter(&vp->v_interlock); 700 if (vp->v_flag & VXLOCK) { 701 if (flags & LK_NOWAIT) { 702 mutex_exit(&vp->v_interlock); 703 return EBUSY; 704 } 705 vp->v_flag |= VXWANT; 706 cv_wait(&vp->v_cv, &vp->v_interlock); 707 mutex_exit(&vp->v_interlock); 708 error = ENOENT; 709 } else { 710 error = VOP_LOCK(vp, 711 (flags & ~LK_RETRY) | LK_INTERLOCK); 712 if (error == 0 || error == EDEADLK || error == EBUSY) 713 return (error); 714 } 715 flags &= ~LK_INTERLOCK; 716 } while (flags & LK_RETRY); 717 return (error); 718 } 719 720 /* 721 * File table vnode close routine. 722 */ 723 static int 724 vn_closefile(struct file *fp, struct lwp *l) 725 { 726 727 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 728 fp->f_cred, l)); 729 } 730 731 /* 732 * Enable LK_CANRECURSE on lock. Return prior status. 733 */ 734 u_int 735 vn_setrecurse(struct vnode *vp) 736 { 737 struct lock *lkp = &vp->v_lock; 738 u_int retval = lkp->lk_flags & LK_CANRECURSE; 739 740 lkp->lk_flags |= LK_CANRECURSE; 741 return retval; 742 } 743 744 /* 745 * Called when done with locksetrecurse. 746 */ 747 void 748 vn_restorerecurse(struct vnode *vp, u_int flags) 749 { 750 struct lock *lkp = &vp->v_lock; 751 752 lkp->lk_flags &= ~LK_CANRECURSE; 753 lkp->lk_flags |= flags; 754 } 755 756 int 757 vn_cow_establish(struct vnode *vp, 758 int (*func)(void *, struct buf *), void *cookie) 759 { 760 struct spec_cow_entry *e; 761 762 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 763 M_DEVBUF, M_WAITOK); 764 e->ce_func = func; 765 e->ce_cookie = cookie; 766 767 mutex_enter(&vp->v_spec_cow_lock); 768 vp->v_spec_cow_req++; 769 while (vp->v_spec_cow_count > 0) 770 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 771 &vp->v_spec_cow_lock); 772 773 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 774 775 vp->v_spec_cow_req--; 776 if (vp->v_spec_cow_req == 0) 777 wakeup(&vp->v_spec_cow_req); 778 mutex_exit(&vp->v_spec_cow_lock); 779 780 return 0; 781 } 782 783 int 784 vn_cow_disestablish(struct vnode *vp, 785 int (*func)(void *, struct buf *), void *cookie) 786 { 787 struct spec_cow_entry *e; 788 789 mutex_enter(&vp->v_spec_cow_lock); 790 vp->v_spec_cow_req++; 791 while (vp->v_spec_cow_count > 0) 792 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 793 &vp->v_spec_cow_lock); 794 795 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 796 if (e->ce_func == func && e->ce_cookie == cookie) { 797 SLIST_REMOVE(&vp->v_spec_cow_head, e, 798 spec_cow_entry, ce_list); 799 FREE(e, M_DEVBUF); 800 break; 801 } 802 803 vp->v_spec_cow_req--; 804 if (vp->v_spec_cow_req == 0) 805 wakeup(&vp->v_spec_cow_req); 806 mutex_exit(&vp->v_spec_cow_lock); 807 808 return e ? 0 : EINVAL; 809 } 810 811 /* 812 * Simplified in-kernel wrapper calls for extended attribute access. 813 * Both calls pass in a NULL credential, authorizing a "kernel" access. 814 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 815 */ 816 int 817 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 818 const char *attrname, size_t *buflen, void *bf, struct lwp *l) 819 { 820 struct uio auio; 821 struct iovec aiov; 822 int error; 823 824 aiov.iov_len = *buflen; 825 aiov.iov_base = bf; 826 827 auio.uio_iov = &aiov; 828 auio.uio_iovcnt = 1; 829 auio.uio_rw = UIO_READ; 830 auio.uio_offset = 0; 831 auio.uio_resid = *buflen; 832 UIO_SETUP_SYSSPACE(&auio); 833 834 if ((ioflg & IO_NODELOCKED) == 0) 835 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 836 837 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 838 l); 839 840 if ((ioflg & IO_NODELOCKED) == 0) 841 VOP_UNLOCK(vp, 0); 842 843 if (error == 0) 844 *buflen = *buflen - auio.uio_resid; 845 846 return (error); 847 } 848 849 /* 850 * XXX Failure mode if partially written? 851 */ 852 int 853 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 854 const char *attrname, size_t buflen, const void *bf, struct lwp *l) 855 { 856 struct uio auio; 857 struct iovec aiov; 858 struct mount *mp = NULL; /* XXX: GCC */ 859 int error; 860 861 aiov.iov_len = buflen; 862 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 863 864 auio.uio_iov = &aiov; 865 auio.uio_iovcnt = 1; 866 auio.uio_rw = UIO_WRITE; 867 auio.uio_offset = 0; 868 auio.uio_resid = buflen; 869 UIO_SETUP_SYSSPACE(&auio); 870 871 if ((ioflg & IO_NODELOCKED) == 0) { 872 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 873 return (error); 874 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 875 } 876 877 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, l); 878 879 if ((ioflg & IO_NODELOCKED) == 0) { 880 vn_finished_write(mp, 0); 881 VOP_UNLOCK(vp, 0); 882 } 883 884 return (error); 885 } 886 887 int 888 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 889 const char *attrname, struct lwp *l) 890 { 891 struct mount *mp = NULL; /* XXX: GCC */ 892 int error; 893 894 if ((ioflg & IO_NODELOCKED) == 0) { 895 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 896 return (error); 897 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 898 } 899 900 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, l); 901 if (error == EOPNOTSUPP) 902 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 903 NULL, l); 904 905 if ((ioflg & IO_NODELOCKED) == 0) { 906 vn_finished_write(mp, 0); 907 VOP_UNLOCK(vp, 0); 908 } 909 910 return (error); 911 } 912 913 /* 914 * OBSOLETE -- this function will be removed in the near future! 915 * 916 * Preparing to start a filesystem write operation. If the operation is 917 * permitted, then we bump the count of operations in progress and 918 * proceed. If a suspend request is in progress, we wait until the 919 * suspension is over, and then proceed. 920 * V_PCATCH adds PCATCH to the tsleep flags. 921 * V_WAIT waits until suspension is over. Otherwise returns EWOULDBLOCK. 922 * V_SLEEPONLY wait, but do not bump the operations count. 923 * V_LOWER this is a lower level operation. No further vnodes should be 924 * locked. Otherwise it is a upper level operation. No vnodes 925 * should be locked. 926 */ 927 int 928 vn_start_write(struct vnode *vp, struct mount **mpp, int flags) 929 { 930 931 /* 932 * If a vnode is provided, get and return the mount point that 933 * to which it will write. 934 */ 935 if (vp != NULL) { 936 *mpp = vp->v_mount; 937 } 938 return 0; 939 } 940 941 /* 942 * OBSOLETE -- this function will be removed in the near future! 943 * 944 * Filesystem write operation has completed. If we are suspending and this 945 * operation is the last one, notify the suspender that the suspension is 946 * now in effect. 947 */ 948 void 949 vn_finished_write(struct mount *mp, int flags) 950 { 951 } 952 953 void 954 vn_ra_allocctx(struct vnode *vp) 955 { 956 struct uvm_ractx *ra = NULL; 957 958 if (vp->v_type != VREG) { 959 return; 960 } 961 if (vp->v_ractx != NULL) { 962 return; 963 } 964 mutex_enter(&vp->v_interlock); 965 if (vp->v_ractx == NULL) { 966 mutex_exit(&vp->v_interlock); 967 ra = uvm_ra_allocctx(); 968 mutex_enter(&vp->v_interlock); 969 if (ra != NULL && vp->v_ractx == NULL) { 970 vp->v_ractx = ra; 971 ra = NULL; 972 } 973 } 974 mutex_exit(&vp->v_interlock); 975 if (ra != NULL) { 976 uvm_ra_freectx(ra); 977 } 978 } 979
Indexes created Sat Sep 20 22:09:52 GMT 2025