vfs_vnops.c revision 1.135.2.3
1 /* $NetBSD: vfs_vnops.c,v 1.135.2.3 2007/04/10 13:26:43 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 1.135.2.3 2007/04/10 13:26:43 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 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 if (fmode & O_CREAT) { 109 ndp->ni_cnd.cn_nameiop = CREATE; 110 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 111 if ((fmode & O_EXCL) == 0 && 112 ((fmode & O_NOFOLLOW) == 0)) 113 ndp->ni_cnd.cn_flags |= FOLLOW; 114 } else { 115 ndp->ni_cnd.cn_nameiop = LOOKUP; 116 ndp->ni_cnd.cn_flags = LOCKLEAF; 117 if ((fmode & O_NOFOLLOW) == 0) 118 ndp->ni_cnd.cn_flags |= FOLLOW; 119 } 120 #if NVERIEXEC > 0 121 error = pathname_get(ndp->ni_dirp, ndp->ni_segflg, &pn); 122 if (error) 123 goto bad2; 124 ndp->ni_dirp = pathname_path(pn); 125 ndp->ni_segflg = UIO_SYSSPACE; 126 #endif /* NVERIEXEC > 0 */ 127 error = namei(ndp); 128 if (error) 129 goto bad2; 130 131 vp = ndp->ni_vp; 132 133 #if NVERIEXEC > 0 134 error = veriexec_openchk(l, ndp->ni_vp, ndp->ni_dirp, fmode); 135 if (error) 136 goto bad; 137 #endif /* NVERIEXEC > 0 */ 138 139 if (fmode & O_CREAT) { 140 if (ndp->ni_vp == NULL) { 141 VATTR_NULL(&va); 142 va.va_type = VREG; 143 va.va_mode = cmode; 144 if (fmode & O_EXCL) 145 va.va_vaflags |= VA_EXCLUSIVE; 146 VOP_LEASE(ndp->ni_dvp, l, cred, LEASE_WRITE); 147 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 148 &ndp->ni_cnd, &va); 149 if (error) 150 goto bad2; 151 fmode &= ~O_TRUNC; 152 vp = ndp->ni_vp; 153 } else { 154 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 155 if (ndp->ni_dvp == ndp->ni_vp) 156 vrele(ndp->ni_dvp); 157 else 158 vput(ndp->ni_dvp); 159 ndp->ni_dvp = NULL; 160 vp = ndp->ni_vp; 161 if (fmode & O_EXCL) { 162 error = EEXIST; 163 goto bad; 164 } 165 fmode &= ~O_CREAT; 166 } 167 } else { 168 vp = ndp->ni_vp; 169 } 170 if (vp->v_type == VSOCK) { 171 error = EOPNOTSUPP; 172 goto bad; 173 } 174 if (ndp->ni_vp->v_type == VLNK) { 175 error = EFTYPE; 176 goto bad; 177 } 178 179 if ((fmode & O_CREAT) == 0) { 180 if (fmode & FREAD) { 181 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 182 goto bad; 183 } 184 185 if (fmode & (FWRITE | O_TRUNC)) { 186 if (vp->v_type == VDIR) { 187 error = EISDIR; 188 goto bad; 189 } 190 if ((error = vn_writechk(vp)) != 0 || 191 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 192 goto bad; 193 } 194 } 195 196 if (fmode & O_TRUNC) { 197 VOP_UNLOCK(vp, 0); /* XXX */ 198 199 VOP_LEASE(vp, l, cred, LEASE_WRITE); 200 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 201 VATTR_NULL(&va); 202 va.va_size = 0; 203 error = VOP_SETATTR(vp, &va, cred, l); 204 if (error != 0) 205 goto bad; 206 } 207 if ((error = VOP_OPEN(vp, fmode, cred, l)) != 0) 208 goto bad; 209 if (vp->v_type == VREG && 210 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 211 error = EIO; 212 goto bad; 213 } 214 if (fmode & FWRITE) 215 vp->v_writecount++; 216 217 bad: 218 if (error) 219 vput(vp); 220 221 bad2: 222 pathname_put(pn); 223 224 return (error); 225 } 226 227 /* 228 * Check for write permissions on the specified vnode. 229 * Prototype text segments cannot be written. 230 */ 231 int 232 vn_writechk(struct vnode *vp) 233 { 234 235 /* 236 * If the vnode is in use as a process's text, 237 * we can't allow writing. 238 */ 239 if (vp->v_flag & VTEXT) 240 return (ETXTBSY); 241 return (0); 242 } 243 244 /* 245 * Mark a vnode as having executable mappings. 246 */ 247 void 248 vn_markexec(struct vnode *vp) 249 { 250 if ((vp->v_flag & VEXECMAP) == 0) { 251 uvmexp.filepages -= vp->v_uobj.uo_npages; 252 uvmexp.execpages += vp->v_uobj.uo_npages; 253 } 254 vp->v_flag |= VEXECMAP; 255 } 256 257 /* 258 * Mark a vnode as being the text of a process. 259 * Fail if the vnode is currently writable. 260 */ 261 int 262 vn_marktext(struct vnode *vp) 263 { 264 265 if (vp->v_writecount != 0) { 266 KASSERT((vp->v_flag & VTEXT) == 0); 267 return (ETXTBSY); 268 } 269 vp->v_flag |= VTEXT; 270 vn_markexec(vp); 271 return (0); 272 } 273 274 /* 275 * Vnode close call 276 * 277 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 278 */ 279 int 280 vn_close(struct vnode *vp, int flags, kauth_cred_t cred, struct lwp *l) 281 { 282 int error; 283 284 if (flags & FWRITE) 285 vp->v_writecount--; 286 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 287 error = VOP_CLOSE(vp, flags, cred, l); 288 vput(vp); 289 return (error); 290 } 291 292 /* 293 * Package up an I/O request on a vnode into a uio and do it. 294 */ 295 int 296 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset, 297 enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid, 298 struct lwp *l) 299 { 300 struct uio auio; 301 struct iovec aiov; 302 int error; 303 304 if ((ioflg & IO_NODELOCKED) == 0) { 305 if (rw == UIO_READ) { 306 vn_lock(vp, LK_SHARED | LK_RETRY); 307 } else /* UIO_WRITE */ { 308 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 309 } 310 } 311 auio.uio_iov = &aiov; 312 auio.uio_iovcnt = 1; 313 aiov.iov_base = base; 314 aiov.iov_len = len; 315 auio.uio_resid = len; 316 auio.uio_offset = offset; 317 auio.uio_rw = rw; 318 if (segflg == UIO_SYSSPACE) { 319 UIO_SETUP_SYSSPACE(&auio); 320 } else { 321 auio.uio_vmspace = l->l_proc->p_vmspace; 322 } 323 if (rw == UIO_READ) { 324 error = VOP_READ(vp, &auio, ioflg, cred); 325 } else { 326 error = VOP_WRITE(vp, &auio, ioflg, cred); 327 } 328 if (aresid) 329 *aresid = auio.uio_resid; 330 else 331 if (auio.uio_resid && error == 0) 332 error = EIO; 333 if ((ioflg & IO_NODELOCKED) == 0) { 334 VOP_UNLOCK(vp, 0); 335 } 336 return (error); 337 } 338 339 int 340 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 341 struct lwp *l, off_t **cookies, int *ncookies) 342 { 343 struct vnode *vp = (struct vnode *)fp->f_data; 344 struct iovec aiov; 345 struct uio auio; 346 int error, eofflag; 347 348 /* Limit the size on any kernel buffers used by VOP_READDIR */ 349 count = min(MAXBSIZE, count); 350 351 unionread: 352 if (vp->v_type != VDIR) 353 return (EINVAL); 354 aiov.iov_base = bf; 355 aiov.iov_len = count; 356 auio.uio_iov = &aiov; 357 auio.uio_iovcnt = 1; 358 auio.uio_rw = UIO_READ; 359 if (segflg == UIO_SYSSPACE) { 360 UIO_SETUP_SYSSPACE(&auio); 361 } else { 362 KASSERT(l == curlwp); 363 auio.uio_vmspace = l->l_proc->p_vmspace; 364 } 365 auio.uio_resid = count; 366 vn_lock(vp, LK_SHARED | LK_RETRY); 367 auio.uio_offset = fp->f_offset; 368 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 369 ncookies); 370 mutex_enter(&fp->f_lock); 371 fp->f_offset = auio.uio_offset; 372 mutex_exit(&fp->f_lock); 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 mutex_enter(&fp->f_lock); 396 fp->f_offset = 0; 397 mutex_exit(&fp->f_lock); 398 vrele(tvp); 399 goto unionread; 400 } 401 *done = count - auio.uio_resid; 402 return error; 403 } 404 405 /* 406 * File table vnode read routine. 407 */ 408 static int 409 vn_read(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 410 int flags) 411 { 412 struct vnode *vp = (struct vnode *)fp->f_data; 413 int count, error, ioflag; 414 struct lwp *l = curlwp; 415 416 VOP_LEASE(vp, l, cred, LEASE_READ); 417 ioflag = IO_ADV_ENCODE(fp->f_advice); 418 if (fp->f_flag & FNONBLOCK) 419 ioflag |= IO_NDELAY; 420 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 421 ioflag |= IO_SYNC; 422 if (fp->f_flag & FALTIO) 423 ioflag |= IO_ALTSEMANTICS; 424 if (fp->f_flag & FDIRECT) 425 ioflag |= IO_DIRECT; 426 vn_lock(vp, LK_SHARED | LK_RETRY); 427 uio->uio_offset = *offset; 428 count = uio->uio_resid; 429 error = VOP_READ(vp, uio, ioflag, cred); 430 if (flags & FOF_UPDATE_OFFSET) 431 *offset += count - uio->uio_resid; 432 VOP_UNLOCK(vp, 0); 433 return (error); 434 } 435 436 /* 437 * File table vnode write routine. 438 */ 439 static int 440 vn_write(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 441 int flags) 442 { 443 struct vnode *vp = (struct vnode *)fp->f_data; 444 int count, error, ioflag = IO_UNIT; 445 struct lwp *l = curlwp; 446 447 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 448 ioflag |= IO_APPEND; 449 if (fp->f_flag & FNONBLOCK) 450 ioflag |= IO_NDELAY; 451 if (fp->f_flag & FFSYNC || 452 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 453 ioflag |= IO_SYNC; 454 else if (fp->f_flag & FDSYNC) 455 ioflag |= IO_DSYNC; 456 if (fp->f_flag & FALTIO) 457 ioflag |= IO_ALTSEMANTICS; 458 if (fp->f_flag & FDIRECT) 459 ioflag |= IO_DIRECT; 460 VOP_LEASE(vp, l, cred, LEASE_WRITE); 461 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 462 uio->uio_offset = *offset; 463 count = uio->uio_resid; 464 error = VOP_WRITE(vp, uio, ioflag, cred); 465 if (flags & FOF_UPDATE_OFFSET) { 466 if (ioflag & IO_APPEND) 467 *offset = uio->uio_offset; 468 else 469 *offset += count - uio->uio_resid; 470 } 471 VOP_UNLOCK(vp, 0); 472 return (error); 473 } 474 475 /* 476 * File table vnode stat routine. 477 */ 478 static int 479 vn_statfile(struct file *fp, struct stat *sb, struct lwp *l) 480 { 481 struct vnode *vp = (struct vnode *)fp->f_data; 482 483 return vn_stat(vp, sb, l); 484 } 485 486 int 487 vn_stat(struct vnode *vp, struct stat *sb, struct lwp *l) 488 { 489 struct vattr va; 490 int error; 491 mode_t mode; 492 493 error = VOP_GETATTR(vp, &va, l->l_cred, l); 494 if (error) 495 return (error); 496 /* 497 * Copy from vattr table 498 */ 499 sb->st_dev = va.va_fsid; 500 sb->st_ino = va.va_fileid; 501 mode = va.va_mode; 502 switch (vp->v_type) { 503 case VREG: 504 mode |= S_IFREG; 505 break; 506 case VDIR: 507 mode |= S_IFDIR; 508 break; 509 case VBLK: 510 mode |= S_IFBLK; 511 break; 512 case VCHR: 513 mode |= S_IFCHR; 514 break; 515 case VLNK: 516 mode |= S_IFLNK; 517 break; 518 case VSOCK: 519 mode |= S_IFSOCK; 520 break; 521 case VFIFO: 522 mode |= S_IFIFO; 523 break; 524 default: 525 return (EBADF); 526 }; 527 sb->st_mode = mode; 528 sb->st_nlink = va.va_nlink; 529 sb->st_uid = va.va_uid; 530 sb->st_gid = va.va_gid; 531 sb->st_rdev = va.va_rdev; 532 sb->st_size = va.va_size; 533 sb->st_atimespec = va.va_atime; 534 sb->st_mtimespec = va.va_mtime; 535 sb->st_ctimespec = va.va_ctime; 536 sb->st_birthtimespec = va.va_birthtime; 537 sb->st_blksize = va.va_blocksize; 538 sb->st_flags = va.va_flags; 539 sb->st_gen = 0; 540 sb->st_blocks = va.va_bytes / S_BLKSIZE; 541 return (0); 542 } 543 544 /* 545 * File table vnode fcntl routine. 546 */ 547 static int 548 vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l) 549 { 550 struct vnode *vp = ((struct vnode *)fp->f_data); 551 int error; 552 553 error = VOP_FCNTL(vp, com, data, fp->f_flag, l->l_cred, l); 554 return (error); 555 } 556 557 /* 558 * File table vnode ioctl routine. 559 */ 560 static int 561 vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l) 562 { 563 struct vnode *vp = ((struct vnode *)fp->f_data), *ovp; 564 struct proc *p = l->l_proc; 565 struct vattr vattr; 566 int error; 567 568 switch (vp->v_type) { 569 570 case VREG: 571 case VDIR: 572 if (com == FIONREAD) { 573 error = VOP_GETATTR(vp, &vattr, l->l_cred, l); 574 if (error) 575 return (error); 576 *(int *)data = vattr.va_size - fp->f_offset; 577 return (0); 578 } 579 if ((com == FIONWRITE) || (com == FIONSPACE)) { 580 /* 581 * Files don't have send queues, so there never 582 * are any bytes in them, nor is there any 583 * open space in them. 584 */ 585 *(int *)data = 0; 586 return (0); 587 } 588 if (com == FIOGETBMAP) { 589 daddr_t *block; 590 591 if (*(daddr_t *)data < 0) 592 return (EINVAL); 593 block = (daddr_t *)data; 594 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 595 } 596 if (com == OFIOGETBMAP) { 597 daddr_t ibn, obn; 598 599 if (*(int32_t *)data < 0) 600 return (EINVAL); 601 ibn = (daddr_t)*(int32_t *)data; 602 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 603 *(int32_t *)data = (int32_t)obn; 604 return error; 605 } 606 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 607 return (0); /* XXX */ 608 /* fall into ... */ 609 case VFIFO: 610 case VCHR: 611 case VBLK: 612 error = VOP_IOCTL(vp, com, data, fp->f_flag, 613 l->l_cred, l); 614 if (error == 0 && com == TIOCSCTTY) { 615 VREF(vp); 616 mutex_enter(&proclist_lock); 617 ovp = p->p_session->s_ttyvp; 618 p->p_session->s_ttyvp = vp; 619 mutex_exit(&proclist_lock); 620 if (ovp != NULL) 621 vrele(ovp); 622 } 623 return (error); 624 625 default: 626 return (EPASSTHROUGH); 627 } 628 } 629 630 /* 631 * File table vnode poll routine. 632 */ 633 static int 634 vn_poll(struct file *fp, int events, struct lwp *l) 635 { 636 637 return (VOP_POLL(((struct vnode *)fp->f_data), events, l)); 638 } 639 640 /* 641 * File table vnode kqfilter routine. 642 */ 643 int 644 vn_kqfilter(struct file *fp, struct knote *kn) 645 { 646 647 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 648 } 649 650 /* 651 * Check that the vnode is still valid, and if so 652 * acquire requested lock. 653 */ 654 int 655 vn_lock(struct vnode *vp, int flags) 656 { 657 int error; 658 659 #if 0 660 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 661 || (vp->v_flag & VONWORKLST) != 0); 662 #endif 663 KASSERT((flags & 664 ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_DRAIN|LK_NOWAIT|LK_RETRY| 665 LK_SETRECURSE|LK_CANRECURSE)) 666 == 0); 667 668 do { 669 if ((flags & LK_INTERLOCK) == 0) 670 mutex_enter(&vp->v_interlock); 671 if (vp->v_flag & VXLOCK) { 672 if (flags & LK_NOWAIT) { 673 mutex_exit(&vp->v_interlock); 674 return EBUSY; 675 } 676 vp->v_flag |= VXWANT; 677 cv_wait(&vp->v_cv, &vp->v_interlock); 678 mutex_exit(&vp->v_interlock); 679 error = ENOENT; 680 } else { 681 error = VOP_LOCK(vp, 682 (flags & ~LK_RETRY) | LK_INTERLOCK); 683 if (error == 0 || error == EDEADLK || error == EBUSY) 684 return (error); 685 } 686 flags &= ~LK_INTERLOCK; 687 } while (flags & LK_RETRY); 688 return (error); 689 } 690 691 /* 692 * File table vnode close routine. 693 */ 694 static int 695 vn_closefile(struct file *fp, struct lwp *l) 696 { 697 698 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 699 fp->f_cred, l)); 700 } 701 702 /* 703 * Enable LK_CANRECURSE on lock. Return prior status. 704 */ 705 u_int 706 vn_setrecurse(struct vnode *vp) 707 { 708 struct lock *lkp = &vp->v_lock; 709 u_int retval = lkp->lk_flags & LK_CANRECURSE; 710 711 lkp->lk_flags |= LK_CANRECURSE; 712 return retval; 713 } 714 715 /* 716 * Called when done with locksetrecurse. 717 */ 718 void 719 vn_restorerecurse(struct vnode *vp, u_int flags) 720 { 721 struct lock *lkp = &vp->v_lock; 722 723 lkp->lk_flags &= ~LK_CANRECURSE; 724 lkp->lk_flags |= flags; 725 } 726 727 int 728 vn_cow_establish(struct vnode *vp, 729 int (*func)(void *, struct buf *), void *cookie) 730 { 731 struct spec_cow_entry *e; 732 733 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 734 M_DEVBUF, M_WAITOK); 735 e->ce_func = func; 736 e->ce_cookie = cookie; 737 738 mutex_enter(&vp->v_spec_cow_lock); 739 vp->v_spec_cow_req++; 740 while (vp->v_spec_cow_count > 0) 741 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 742 &vp->v_spec_cow_lock); 743 744 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 745 746 vp->v_spec_cow_req--; 747 if (vp->v_spec_cow_req == 0) 748 wakeup(&vp->v_spec_cow_req); 749 mutex_exit(&vp->v_spec_cow_lock); 750 751 return 0; 752 } 753 754 int 755 vn_cow_disestablish(struct vnode *vp, 756 int (*func)(void *, struct buf *), void *cookie) 757 { 758 struct spec_cow_entry *e; 759 760 mutex_enter(&vp->v_spec_cow_lock); 761 vp->v_spec_cow_req++; 762 while (vp->v_spec_cow_count > 0) 763 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 764 &vp->v_spec_cow_lock); 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 mutex_exit(&vp->v_spec_cow_lock); 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 mutex_enter(&vp->v_interlock); 888 if (vp->v_ractx == NULL) { 889 mutex_exit(&vp->v_interlock); 890 ra = uvm_ra_allocctx(); 891 mutex_enter(&vp->v_interlock); 892 if (ra != NULL && vp->v_ractx == NULL) { 893 vp->v_ractx = ra; 894 ra = NULL; 895 } 896 } 897 mutex_exit(&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