vfs_vnops.c revision 1.133.2.3
1 /* $NetBSD: vfs_vnops.c,v 1.133.2.3 2007/05/07 10:55:51 yamt 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 1.133.2.3 2007/05/07 10:55:51 yamt 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 lwp *l = ndp->ni_cnd.cn_lwp; 103 kauth_cred_t cred = l->l_cred; 104 struct vattr va; 105 int error; 106 pathname_t pn = NULL; 107 108 ndp->ni_cnd.cn_flags &= TRYEMULROOT; 109 110 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 VOP_LEASE(ndp->ni_dvp, l, cred, LEASE_WRITE); 149 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 150 &ndp->ni_cnd, &va); 151 if (error) 152 goto bad2; 153 fmode &= ~O_TRUNC; 154 vp = ndp->ni_vp; 155 } else { 156 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 157 if (ndp->ni_dvp == ndp->ni_vp) 158 vrele(ndp->ni_dvp); 159 else 160 vput(ndp->ni_dvp); 161 ndp->ni_dvp = NULL; 162 vp = ndp->ni_vp; 163 if (fmode & O_EXCL) { 164 error = EEXIST; 165 goto bad; 166 } 167 fmode &= ~O_CREAT; 168 } 169 } else { 170 vp = ndp->ni_vp; 171 } 172 if (vp->v_type == VSOCK) { 173 error = EOPNOTSUPP; 174 goto bad; 175 } 176 if (ndp->ni_vp->v_type == VLNK) { 177 error = EFTYPE; 178 goto bad; 179 } 180 181 if ((fmode & O_CREAT) == 0) { 182 if (fmode & FREAD) { 183 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 184 goto bad; 185 } 186 187 if (fmode & (FWRITE | O_TRUNC)) { 188 if (vp->v_type == VDIR) { 189 error = EISDIR; 190 goto bad; 191 } 192 if ((error = vn_writechk(vp)) != 0 || 193 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 194 goto bad; 195 } 196 } 197 198 if (fmode & O_TRUNC) { 199 VOP_UNLOCK(vp, 0); /* XXX */ 200 201 VOP_LEASE(vp, l, cred, LEASE_WRITE); 202 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 203 VATTR_NULL(&va); 204 va.va_size = 0; 205 error = VOP_SETATTR(vp, &va, cred, l); 206 if (error != 0) 207 goto bad; 208 } 209 if ((error = VOP_OPEN(vp, fmode, cred, l)) != 0) 210 goto bad; 211 if (vp->v_type == VREG && 212 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 213 error = EIO; 214 goto bad; 215 } 216 if (fmode & FWRITE) 217 vp->v_writecount++; 218 219 bad: 220 if (error) 221 vput(vp); 222 223 bad2: 224 pathname_put(pn); 225 226 return (error); 227 } 228 229 /* 230 * Check for write permissions on the specified vnode. 231 * Prototype text segments cannot be written. 232 */ 233 int 234 vn_writechk(struct vnode *vp) 235 { 236 237 /* 238 * If the vnode is in use as a process's text, 239 * we can't allow writing. 240 */ 241 if (vp->v_flag & VTEXT) 242 return (ETXTBSY); 243 return (0); 244 } 245 246 /* 247 * Mark a vnode as having executable mappings. 248 */ 249 void 250 vn_markexec(struct vnode *vp) 251 { 252 if ((vp->v_flag & VEXECMAP) == 0) { 253 uvmexp.filepages -= vp->v_uobj.uo_npages; 254 uvmexp.execpages += vp->v_uobj.uo_npages; 255 } 256 vp->v_flag |= VEXECMAP; 257 } 258 259 /* 260 * Mark a vnode as being the text of a process. 261 * Fail if the vnode is currently writable. 262 */ 263 int 264 vn_marktext(struct vnode *vp) 265 { 266 267 if (vp->v_writecount != 0) { 268 KASSERT((vp->v_flag & VTEXT) == 0); 269 return (ETXTBSY); 270 } 271 vp->v_flag |= VTEXT; 272 vn_markexec(vp); 273 return (0); 274 } 275 276 /* 277 * Vnode close call 278 * 279 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 280 */ 281 int 282 vn_close(struct vnode *vp, int flags, kauth_cred_t cred, struct lwp *l) 283 { 284 int error; 285 286 if (flags & FWRITE) 287 vp->v_writecount--; 288 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 289 error = VOP_CLOSE(vp, flags, cred, l); 290 vput(vp); 291 return (error); 292 } 293 294 /* 295 * Package up an I/O request on a vnode into a uio and do it. 296 */ 297 int 298 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset, 299 enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid, 300 struct lwp *l) 301 { 302 struct uio auio; 303 struct iovec aiov; 304 int error; 305 306 if ((ioflg & IO_NODELOCKED) == 0) { 307 if (rw == UIO_READ) { 308 vn_lock(vp, LK_SHARED | LK_RETRY); 309 } else /* UIO_WRITE */ { 310 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 311 } 312 } 313 auio.uio_iov = &aiov; 314 auio.uio_iovcnt = 1; 315 aiov.iov_base = base; 316 aiov.iov_len = len; 317 auio.uio_resid = len; 318 auio.uio_offset = offset; 319 auio.uio_rw = rw; 320 if (segflg == UIO_SYSSPACE) { 321 UIO_SETUP_SYSSPACE(&auio); 322 } else { 323 auio.uio_vmspace = l->l_proc->p_vmspace; 324 } 325 if (rw == UIO_READ) { 326 error = VOP_READ(vp, &auio, ioflg, cred); 327 } else { 328 error = VOP_WRITE(vp, &auio, ioflg, cred); 329 } 330 if (aresid) 331 *aresid = auio.uio_resid; 332 else 333 if (auio.uio_resid && error == 0) 334 error = EIO; 335 if ((ioflg & IO_NODELOCKED) == 0) { 336 VOP_UNLOCK(vp, 0); 337 } 338 return (error); 339 } 340 341 int 342 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 343 struct lwp *l, off_t **cookies, int *ncookies) 344 { 345 struct vnode *vp = (struct vnode *)fp->f_data; 346 struct iovec aiov; 347 struct uio auio; 348 int error, eofflag; 349 350 /* Limit the size on any kernel buffers used by VOP_READDIR */ 351 count = min(MAXBSIZE, count); 352 353 unionread: 354 if (vp->v_type != VDIR) 355 return (EINVAL); 356 aiov.iov_base = bf; 357 aiov.iov_len = count; 358 auio.uio_iov = &aiov; 359 auio.uio_iovcnt = 1; 360 auio.uio_rw = UIO_READ; 361 if (segflg == UIO_SYSSPACE) { 362 UIO_SETUP_SYSSPACE(&auio); 363 } else { 364 KASSERT(l == curlwp); 365 auio.uio_vmspace = l->l_proc->p_vmspace; 366 } 367 auio.uio_resid = count; 368 vn_lock(vp, LK_SHARED | LK_RETRY); 369 auio.uio_offset = fp->f_offset; 370 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 371 ncookies); 372 fp->f_offset = auio.uio_offset; 373 VOP_UNLOCK(vp, 0); 374 if (error) 375 return (error); 376 377 #if defined(UNION) || defined(LKM) 378 if (count == auio.uio_resid && vn_union_readdir_hook) { 379 struct vnode *ovp = vp; 380 381 error = (*vn_union_readdir_hook)(&vp, fp, l); 382 if (error) 383 return (error); 384 if (vp != ovp) 385 goto unionread; 386 } 387 #endif /* UNION || LKM */ 388 389 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 390 (vp->v_mount->mnt_flag & MNT_UNION)) { 391 struct vnode *tvp = vp; 392 vp = vp->v_mount->mnt_vnodecovered; 393 VREF(vp); 394 fp->f_data = vp; 395 fp->f_offset = 0; 396 vrele(tvp); 397 goto unionread; 398 } 399 *done = count - auio.uio_resid; 400 return error; 401 } 402 403 /* 404 * File table vnode read routine. 405 */ 406 static int 407 vn_read(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 408 int flags) 409 { 410 struct vnode *vp = (struct vnode *)fp->f_data; 411 int count, error, ioflag; 412 struct lwp *l = curlwp; 413 414 VOP_LEASE(vp, l, cred, LEASE_READ); 415 ioflag = IO_ADV_ENCODE(fp->f_advice); 416 if (fp->f_flag & FNONBLOCK) 417 ioflag |= IO_NDELAY; 418 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 419 ioflag |= IO_SYNC; 420 if (fp->f_flag & FALTIO) 421 ioflag |= IO_ALTSEMANTICS; 422 if (fp->f_flag & FDIRECT) 423 ioflag |= IO_DIRECT; 424 vn_lock(vp, LK_SHARED | LK_RETRY); 425 uio->uio_offset = *offset; 426 count = uio->uio_resid; 427 error = VOP_READ(vp, uio, ioflag, cred); 428 if (flags & FOF_UPDATE_OFFSET) 429 *offset += count - uio->uio_resid; 430 VOP_UNLOCK(vp, 0); 431 return (error); 432 } 433 434 /* 435 * File table vnode write routine. 436 */ 437 static int 438 vn_write(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 439 int flags) 440 { 441 struct vnode *vp = (struct vnode *)fp->f_data; 442 int count, error, ioflag = IO_UNIT; 443 struct lwp *l = curlwp; 444 445 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 446 ioflag |= IO_APPEND; 447 if (fp->f_flag & FNONBLOCK) 448 ioflag |= IO_NDELAY; 449 if (fp->f_flag & FFSYNC || 450 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 451 ioflag |= IO_SYNC; 452 else if (fp->f_flag & FDSYNC) 453 ioflag |= IO_DSYNC; 454 if (fp->f_flag & FALTIO) 455 ioflag |= IO_ALTSEMANTICS; 456 if (fp->f_flag & FDIRECT) 457 ioflag |= IO_DIRECT; 458 VOP_LEASE(vp, l, cred, LEASE_WRITE); 459 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 460 uio->uio_offset = *offset; 461 count = uio->uio_resid; 462 error = VOP_WRITE(vp, uio, ioflag, cred); 463 if (flags & FOF_UPDATE_OFFSET) { 464 if (ioflag & IO_APPEND) 465 *offset = uio->uio_offset; 466 else 467 *offset += count - uio->uio_resid; 468 } 469 VOP_UNLOCK(vp, 0); 470 return (error); 471 } 472 473 /* 474 * File table vnode stat routine. 475 */ 476 static int 477 vn_statfile(struct file *fp, struct stat *sb, struct lwp *l) 478 { 479 struct vnode *vp = (struct vnode *)fp->f_data; 480 481 return vn_stat(vp, sb, l); 482 } 483 484 int 485 vn_stat(struct vnode *vp, struct stat *sb, struct lwp *l) 486 { 487 struct vattr va; 488 int error; 489 mode_t mode; 490 491 error = VOP_GETATTR(vp, &va, l->l_cred, l); 492 if (error) 493 return (error); 494 /* 495 * Copy from vattr table 496 */ 497 sb->st_dev = va.va_fsid; 498 sb->st_ino = va.va_fileid; 499 mode = va.va_mode; 500 switch (vp->v_type) { 501 case VREG: 502 mode |= S_IFREG; 503 break; 504 case VDIR: 505 mode |= S_IFDIR; 506 break; 507 case VBLK: 508 mode |= S_IFBLK; 509 break; 510 case VCHR: 511 mode |= S_IFCHR; 512 break; 513 case VLNK: 514 mode |= S_IFLNK; 515 break; 516 case VSOCK: 517 mode |= S_IFSOCK; 518 break; 519 case VFIFO: 520 mode |= S_IFIFO; 521 break; 522 default: 523 return (EBADF); 524 }; 525 sb->st_mode = mode; 526 sb->st_nlink = va.va_nlink; 527 sb->st_uid = va.va_uid; 528 sb->st_gid = va.va_gid; 529 sb->st_rdev = va.va_rdev; 530 sb->st_size = va.va_size; 531 sb->st_atimespec = va.va_atime; 532 sb->st_mtimespec = va.va_mtime; 533 sb->st_ctimespec = va.va_ctime; 534 sb->st_birthtimespec = va.va_birthtime; 535 sb->st_blksize = va.va_blocksize; 536 sb->st_flags = va.va_flags; 537 sb->st_gen = 0; 538 sb->st_blocks = va.va_bytes / S_BLKSIZE; 539 return (0); 540 } 541 542 /* 543 * File table vnode fcntl routine. 544 */ 545 static int 546 vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l) 547 { 548 struct vnode *vp = ((struct vnode *)fp->f_data); 549 int error; 550 551 error = VOP_FCNTL(vp, com, data, fp->f_flag, l->l_cred, l); 552 return (error); 553 } 554 555 /* 556 * File table vnode ioctl routine. 557 */ 558 static int 559 vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l) 560 { 561 struct vnode *vp = ((struct vnode *)fp->f_data), *ovp; 562 struct proc *p = l->l_proc; 563 struct vattr vattr; 564 int error; 565 566 switch (vp->v_type) { 567 568 case VREG: 569 case VDIR: 570 if (com == FIONREAD) { 571 error = VOP_GETATTR(vp, &vattr, l->l_cred, l); 572 if (error) 573 return (error); 574 *(int *)data = vattr.va_size - fp->f_offset; 575 return (0); 576 } 577 if ((com == FIONWRITE) || (com == FIONSPACE)) { 578 /* 579 * Files don't have send queues, so there never 580 * are any bytes in them, nor is there any 581 * open space in them. 582 */ 583 *(int *)data = 0; 584 return (0); 585 } 586 if (com == FIOGETBMAP) { 587 daddr_t *block; 588 589 if (*(daddr_t *)data < 0) 590 return (EINVAL); 591 block = (daddr_t *)data; 592 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 593 } 594 if (com == OFIOGETBMAP) { 595 daddr_t ibn, obn; 596 597 if (*(int32_t *)data < 0) 598 return (EINVAL); 599 ibn = (daddr_t)*(int32_t *)data; 600 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 601 *(int32_t *)data = (int32_t)obn; 602 return error; 603 } 604 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 605 return (0); /* XXX */ 606 /* fall into ... */ 607 case VFIFO: 608 case VCHR: 609 case VBLK: 610 error = VOP_IOCTL(vp, com, data, fp->f_flag, 611 l->l_cred, l); 612 if (error == 0 && com == TIOCSCTTY) { 613 VREF(vp); 614 mutex_enter(&proclist_lock); 615 ovp = p->p_session->s_ttyvp; 616 p->p_session->s_ttyvp = vp; 617 mutex_exit(&proclist_lock); 618 if (ovp != NULL) 619 vrele(ovp); 620 } 621 return (error); 622 623 default: 624 return (EPASSTHROUGH); 625 } 626 } 627 628 /* 629 * File table vnode poll routine. 630 */ 631 static int 632 vn_poll(struct file *fp, int events, struct lwp *l) 633 { 634 635 return (VOP_POLL(((struct vnode *)fp->f_data), events, l)); 636 } 637 638 /* 639 * File table vnode kqfilter routine. 640 */ 641 int 642 vn_kqfilter(struct file *fp, struct knote *kn) 643 { 644 645 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 646 } 647 648 /* 649 * Check that the vnode is still valid, and if so 650 * acquire requested lock. 651 */ 652 int 653 vn_lock(struct vnode *vp, int flags) 654 { 655 int error; 656 657 #if 0 658 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 659 || (vp->v_flag & VONWORKLST) != 0); 660 #endif 661 KASSERT((flags & 662 ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_DRAIN|LK_NOWAIT|LK_RETRY| 663 LK_SETRECURSE|LK_CANRECURSE)) 664 == 0); 665 666 do { 667 if ((flags & LK_INTERLOCK) == 0) 668 simple_lock(&vp->v_interlock); 669 if (vp->v_flag & VXLOCK) { 670 if (flags & LK_NOWAIT) { 671 simple_unlock(&vp->v_interlock); 672 return EBUSY; 673 } 674 vp->v_flag |= VXWANT; 675 ltsleep(vp, PINOD | PNORELOCK, 676 "vn_lock", 0, &vp->v_interlock); 677 error = ENOENT; 678 } else { 679 error = VOP_LOCK(vp, 680 (flags & ~LK_RETRY) | LK_INTERLOCK); 681 if (error == 0 || error == EDEADLK || error == EBUSY) 682 return (error); 683 } 684 flags &= ~LK_INTERLOCK; 685 } while (flags & LK_RETRY); 686 return (error); 687 } 688 689 /* 690 * File table vnode close routine. 691 */ 692 static int 693 vn_closefile(struct file *fp, struct lwp *l) 694 { 695 696 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 697 fp->f_cred, l)); 698 } 699 700 /* 701 * Enable LK_CANRECURSE on lock. Return prior status. 702 */ 703 u_int 704 vn_setrecurse(struct vnode *vp) 705 { 706 struct lock *lkp = &vp->v_lock; 707 u_int retval = lkp->lk_flags & LK_CANRECURSE; 708 709 lkp->lk_flags |= LK_CANRECURSE; 710 return retval; 711 } 712 713 /* 714 * Called when done with locksetrecurse. 715 */ 716 void 717 vn_restorerecurse(struct vnode *vp, u_int flags) 718 { 719 struct lock *lkp = &vp->v_lock; 720 721 lkp->lk_flags &= ~LK_CANRECURSE; 722 lkp->lk_flags |= flags; 723 } 724 725 int 726 vn_cow_establish(struct vnode *vp, 727 int (*func)(void *, struct buf *), void *cookie) 728 { 729 int s; 730 struct spec_cow_entry *e; 731 732 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 733 M_DEVBUF, M_WAITOK); 734 e->ce_func = func; 735 e->ce_cookie = cookie; 736 737 SPEC_COW_LOCK(vp->v_specinfo, s); 738 vp->v_spec_cow_req++; 739 while (vp->v_spec_cow_count > 0) 740 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 741 &vp->v_spec_cow_slock); 742 743 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 744 745 vp->v_spec_cow_req--; 746 if (vp->v_spec_cow_req == 0) 747 wakeup(&vp->v_spec_cow_req); 748 SPEC_COW_UNLOCK(vp->v_specinfo, s); 749 750 return 0; 751 } 752 753 int 754 vn_cow_disestablish(struct vnode *vp, 755 int (*func)(void *, struct buf *), void *cookie) 756 { 757 int s; 758 struct spec_cow_entry *e; 759 760 SPEC_COW_LOCK(vp->v_specinfo, s); 761 vp->v_spec_cow_req++; 762 while (vp->v_spec_cow_count > 0) 763 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 764 &vp->v_spec_cow_slock); 765 766 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 767 if (e->ce_func == func && e->ce_cookie == cookie) { 768 SLIST_REMOVE(&vp->v_spec_cow_head, e, 769 spec_cow_entry, ce_list); 770 FREE(e, M_DEVBUF); 771 break; 772 } 773 774 vp->v_spec_cow_req--; 775 if (vp->v_spec_cow_req == 0) 776 wakeup(&vp->v_spec_cow_req); 777 SPEC_COW_UNLOCK(vp->v_specinfo, s); 778 779 return e ? 0 : EINVAL; 780 } 781 782 /* 783 * Simplified in-kernel wrapper calls for extended attribute access. 784 * Both calls pass in a NULL credential, authorizing a "kernel" access. 785 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 786 */ 787 int 788 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 789 const char *attrname, size_t *buflen, void *bf, struct lwp *l) 790 { 791 struct uio auio; 792 struct iovec aiov; 793 int error; 794 795 aiov.iov_len = *buflen; 796 aiov.iov_base = bf; 797 798 auio.uio_iov = &aiov; 799 auio.uio_iovcnt = 1; 800 auio.uio_rw = UIO_READ; 801 auio.uio_offset = 0; 802 auio.uio_resid = *buflen; 803 UIO_SETUP_SYSSPACE(&auio); 804 805 if ((ioflg & IO_NODELOCKED) == 0) 806 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 807 808 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 809 l); 810 811 if ((ioflg & IO_NODELOCKED) == 0) 812 VOP_UNLOCK(vp, 0); 813 814 if (error == 0) 815 *buflen = *buflen - auio.uio_resid; 816 817 return (error); 818 } 819 820 /* 821 * XXX Failure mode if partially written? 822 */ 823 int 824 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 825 const char *attrname, size_t buflen, const void *bf, struct lwp *l) 826 { 827 struct uio auio; 828 struct iovec aiov; 829 int error; 830 831 aiov.iov_len = buflen; 832 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 833 834 auio.uio_iov = &aiov; 835 auio.uio_iovcnt = 1; 836 auio.uio_rw = UIO_WRITE; 837 auio.uio_offset = 0; 838 auio.uio_resid = buflen; 839 UIO_SETUP_SYSSPACE(&auio); 840 841 if ((ioflg & IO_NODELOCKED) == 0) { 842 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 843 } 844 845 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, l); 846 847 if ((ioflg & IO_NODELOCKED) == 0) { 848 VOP_UNLOCK(vp, 0); 849 } 850 851 return (error); 852 } 853 854 int 855 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 856 const char *attrname, struct lwp *l) 857 { 858 int error; 859 860 if ((ioflg & IO_NODELOCKED) == 0) { 861 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 862 } 863 864 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, l); 865 if (error == EOPNOTSUPP) 866 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 867 NULL, l); 868 869 if ((ioflg & IO_NODELOCKED) == 0) { 870 VOP_UNLOCK(vp, 0); 871 } 872 873 return (error); 874 } 875 876 void 877 vn_ra_allocctx(struct vnode *vp) 878 { 879 struct uvm_ractx *ra = NULL; 880 881 if (vp->v_type != VREG) { 882 return; 883 } 884 if (vp->v_ractx != NULL) { 885 return; 886 } 887 simple_lock(&vp->v_interlock); 888 if (vp->v_ractx == NULL) { 889 simple_unlock(&vp->v_interlock); 890 ra = uvm_ra_allocctx(); 891 simple_lock(&vp->v_interlock); 892 if (ra != NULL && vp->v_ractx == NULL) { 893 vp->v_ractx = ra; 894 ra = NULL; 895 } 896 } 897 simple_unlock(&vp->v_interlock); 898 if (ra != NULL) { 899 uvm_ra_freectx(ra); 900 } 901 } 902
Indexes created Sat Sep 20 22:09:52 GMT 2025