vfs_vnops.c revision 1.135.2.6
1 /* $NetBSD: vfs_vnops.c,v 1.135.2.6 2007/06/17 21:31:35 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 1.135.2.6 2007/06/17 21:31:35 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 #include <sys/verified_exec.h> 76 77 static int vn_read(struct file *fp, off_t *offset, struct uio *uio, 78 kauth_cred_t cred, int flags); 79 static int vn_write(struct file *fp, off_t *offset, struct uio *uio, 80 kauth_cred_t cred, int flags); 81 static int vn_closefile(struct file *fp, struct lwp *l); 82 static int vn_poll(struct file *fp, int events, struct lwp *l); 83 static int vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l); 84 static int vn_statfile(struct file *fp, struct stat *sb, struct lwp *l); 85 static int vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l); 86 87 const struct fileops vnops = { 88 vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, 89 vn_statfile, vn_closefile, vn_kqfilter 90 }; 91 92 /* 93 * Common code for vnode open operations. 94 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 95 */ 96 int 97 vn_open(struct nameidata *ndp, int fmode, int cmode) 98 { 99 struct vnode *vp; 100 struct lwp *l = ndp->ni_cnd.cn_lwp; 101 kauth_cred_t cred = l->l_cred; 102 struct vattr va; 103 int error; 104 char *path; 105 106 ndp->ni_cnd.cn_flags &= TRYEMULROOT; 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 } 120 121 VERIEXEC_PATH_GET(ndp->ni_dirp, ndp->ni_segflg, ndp->ni_dirp, path); 122 123 error = namei(ndp); 124 if (error) 125 goto out; 126 127 vp = ndp->ni_vp; 128 129 #if NVERIEXEC > 0 130 error = veriexec_openchk(l, ndp->ni_vp, ndp->ni_dirp, fmode); 131 if (error) 132 goto bad; 133 #endif /* NVERIEXEC > 0 */ 134 135 if (fmode & O_CREAT) { 136 if (ndp->ni_vp == NULL) { 137 VATTR_NULL(&va); 138 va.va_type = VREG; 139 va.va_mode = cmode; 140 if (fmode & O_EXCL) 141 va.va_vaflags |= VA_EXCLUSIVE; 142 VOP_LEASE(ndp->ni_dvp, l, cred, LEASE_WRITE); 143 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 144 &ndp->ni_cnd, &va); 145 if (error) 146 goto out; 147 fmode &= ~O_TRUNC; 148 vp = ndp->ni_vp; 149 } else { 150 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 151 if (ndp->ni_dvp == ndp->ni_vp) 152 vrele(ndp->ni_dvp); 153 else 154 vput(ndp->ni_dvp); 155 ndp->ni_dvp = NULL; 156 vp = ndp->ni_vp; 157 if (fmode & O_EXCL) { 158 error = EEXIST; 159 goto bad; 160 } 161 fmode &= ~O_CREAT; 162 } 163 } else { 164 vp = ndp->ni_vp; 165 } 166 if (vp->v_type == VSOCK) { 167 error = EOPNOTSUPP; 168 goto bad; 169 } 170 if (ndp->ni_vp->v_type == VLNK) { 171 error = EFTYPE; 172 goto bad; 173 } 174 175 if ((fmode & O_CREAT) == 0) { 176 if (fmode & FREAD) { 177 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 178 goto bad; 179 } 180 181 if (fmode & (FWRITE | O_TRUNC)) { 182 if (vp->v_type == VDIR) { 183 error = EISDIR; 184 goto bad; 185 } 186 if ((error = vn_writechk(vp)) != 0 || 187 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 188 goto bad; 189 } 190 } 191 192 if (fmode & O_TRUNC) { 193 VOP_UNLOCK(vp, 0); /* XXX */ 194 195 VOP_LEASE(vp, l, cred, LEASE_WRITE); 196 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 197 VATTR_NULL(&va); 198 va.va_size = 0; 199 error = VOP_SETATTR(vp, &va, cred, l); 200 if (error != 0) 201 goto bad; 202 } 203 if ((error = VOP_OPEN(vp, fmode, cred, l)) != 0) 204 goto bad; 205 if (vp->v_type == VREG && 206 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 207 error = EIO; 208 goto bad; 209 } 210 if (fmode & FWRITE) { 211 mutex_enter(&vp->v_interlock); 212 vp->v_writecount++; 213 mutex_exit(&vp->v_interlock); 214 } 215 216 bad: 217 if (error) 218 vput(vp); 219 out: 220 VERIEXEC_PATH_PUT(path); 221 return (error); 222 } 223 224 /* 225 * Check for write permissions on the specified vnode. 226 * Prototype text segments cannot be written. 227 */ 228 int 229 vn_writechk(struct vnode *vp) 230 { 231 232 /* 233 * If the vnode is in use as a process's text, 234 * we can't allow writing. 235 */ 236 if (vp->v_iflag & VI_TEXT) 237 return (ETXTBSY); 238 return (0); 239 } 240 241 /* 242 * Mark a vnode as having executable mappings. 243 */ 244 void 245 vn_markexec(struct vnode *vp) 246 { 247 248 KASSERT(mutex_owned(&vp->v_interlock)); 249 250 if ((vp->v_iflag & VI_EXECMAP) == 0) { 251 /* XXXSMP should be atomic */ 252 uvmexp.filepages -= vp->v_uobj.uo_npages; 253 uvmexp.execpages += vp->v_uobj.uo_npages; 254 } 255 vp->v_iflag |= VI_EXECMAP; 256 } 257 258 /* 259 * Mark a vnode as being the text of a process. 260 * Fail if the vnode is currently writable. 261 */ 262 int 263 vn_marktext(struct vnode *vp) 264 { 265 266 mutex_enter(&vp->v_interlock); 267 if (vp->v_writecount != 0) { 268 KASSERT((vp->v_iflag & VI_TEXT) == 0); 269 mutex_exit(&vp->v_interlock); 270 return (ETXTBSY); 271 } 272 vp->v_iflag |= VI_TEXT; 273 vn_markexec(vp); 274 mutex_exit(&vp->v_interlock); 275 return (0); 276 } 277 278 /* 279 * Vnode close call 280 * 281 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 282 */ 283 int 284 vn_close(struct vnode *vp, int flags, kauth_cred_t cred, struct lwp *l) 285 { 286 int error; 287 288 mutex_enter(&vp->v_interlock); 289 if (flags & FWRITE) 290 vp->v_writecount--; 291 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_INTERLOCK); 292 error = VOP_CLOSE(vp, flags, cred, l); 293 vput(vp); 294 return (error); 295 } 296 297 /* 298 * Package up an I/O request on a vnode into a uio and do it. 299 */ 300 int 301 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset, 302 enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid, 303 struct lwp *l) 304 { 305 struct uio auio; 306 struct iovec aiov; 307 int error; 308 309 if ((ioflg & IO_NODELOCKED) == 0) { 310 if (rw == UIO_READ) { 311 vn_lock(vp, LK_SHARED | LK_RETRY); 312 } else /* UIO_WRITE */ { 313 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 314 } 315 } 316 auio.uio_iov = &aiov; 317 auio.uio_iovcnt = 1; 318 aiov.iov_base = base; 319 aiov.iov_len = len; 320 auio.uio_resid = len; 321 auio.uio_offset = offset; 322 auio.uio_rw = rw; 323 if (segflg == UIO_SYSSPACE) { 324 UIO_SETUP_SYSSPACE(&auio); 325 } else { 326 auio.uio_vmspace = l->l_proc->p_vmspace; 327 } 328 if (rw == UIO_READ) { 329 error = VOP_READ(vp, &auio, ioflg, cred); 330 } else { 331 error = VOP_WRITE(vp, &auio, ioflg, cred); 332 } 333 if (aresid) 334 *aresid = auio.uio_resid; 335 else 336 if (auio.uio_resid && error == 0) 337 error = EIO; 338 if ((ioflg & IO_NODELOCKED) == 0) { 339 VOP_UNLOCK(vp, 0); 340 } 341 return (error); 342 } 343 344 int 345 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 346 struct lwp *l, off_t **cookies, int *ncookies) 347 { 348 struct vnode *vp = (struct vnode *)fp->f_data; 349 struct iovec aiov; 350 struct uio auio; 351 int error, eofflag; 352 353 /* Limit the size on any kernel buffers used by VOP_READDIR */ 354 count = min(MAXBSIZE, count); 355 356 unionread: 357 if (vp->v_type != VDIR) 358 return (EINVAL); 359 aiov.iov_base = bf; 360 aiov.iov_len = count; 361 auio.uio_iov = &aiov; 362 auio.uio_iovcnt = 1; 363 auio.uio_rw = UIO_READ; 364 if (segflg == UIO_SYSSPACE) { 365 UIO_SETUP_SYSSPACE(&auio); 366 } else { 367 KASSERT(l == curlwp); 368 auio.uio_vmspace = l->l_proc->p_vmspace; 369 } 370 auio.uio_resid = count; 371 vn_lock(vp, LK_SHARED | LK_RETRY); 372 auio.uio_offset = fp->f_offset; 373 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 374 ncookies); 375 mutex_enter(&fp->f_lock); 376 fp->f_offset = auio.uio_offset; 377 mutex_exit(&fp->f_lock); 378 VOP_UNLOCK(vp, 0); 379 if (error) 380 return (error); 381 382 #if defined(UNION) || defined(LKM) 383 if (count == auio.uio_resid && vn_union_readdir_hook) { 384 struct vnode *ovp = vp; 385 386 error = (*vn_union_readdir_hook)(&vp, fp, l); 387 if (error) 388 return (error); 389 if (vp != ovp) 390 goto unionread; 391 } 392 #endif /* UNION || LKM */ 393 394 if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) && 395 (vp->v_mount->mnt_flag & MNT_UNION)) { 396 struct vnode *tvp = vp; 397 vp = vp->v_mount->mnt_vnodecovered; 398 VREF(vp); 399 mutex_enter(&fp->f_lock); 400 fp->f_data = vp; 401 fp->f_offset = 0; 402 mutex_exit(&fp->f_lock); 403 vrele(tvp); 404 goto unionread; 405 } 406 *done = count - auio.uio_resid; 407 return error; 408 } 409 410 /* 411 * File table vnode read routine. 412 */ 413 static int 414 vn_read(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 415 int flags) 416 { 417 struct vnode *vp = (struct vnode *)fp->f_data; 418 int count, error, ioflag; 419 struct lwp *l = curlwp; 420 421 VOP_LEASE(vp, l, cred, LEASE_READ); 422 mutex_enter(&fp->f_lock); 423 ioflag = IO_ADV_ENCODE(fp->f_advice); 424 if (fp->f_flag & FNONBLOCK) 425 ioflag |= IO_NDELAY; 426 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 427 ioflag |= IO_SYNC; 428 if (fp->f_flag & FALTIO) 429 ioflag |= IO_ALTSEMANTICS; 430 if (fp->f_flag & FDIRECT) 431 ioflag |= IO_DIRECT; 432 mutex_exit(&fp->f_lock); 433 vn_lock(vp, LK_SHARED | LK_RETRY); 434 uio->uio_offset = *offset; 435 count = uio->uio_resid; 436 error = VOP_READ(vp, uio, ioflag, cred); 437 if (flags & FOF_UPDATE_OFFSET) 438 *offset += count - uio->uio_resid; 439 VOP_UNLOCK(vp, 0); 440 return (error); 441 } 442 443 /* 444 * File table vnode write routine. 445 */ 446 static int 447 vn_write(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 448 int flags) 449 { 450 struct vnode *vp = (struct vnode *)fp->f_data; 451 int count, error, ioflag = IO_UNIT; 452 struct lwp *l = curlwp; 453 454 mutex_enter(&fp->f_lock); 455 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 456 ioflag |= IO_APPEND; 457 if (fp->f_flag & FNONBLOCK) 458 ioflag |= IO_NDELAY; 459 if (fp->f_flag & FFSYNC || 460 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 461 ioflag |= IO_SYNC; 462 else if (fp->f_flag & FDSYNC) 463 ioflag |= IO_DSYNC; 464 if (fp->f_flag & FALTIO) 465 ioflag |= IO_ALTSEMANTICS; 466 if (fp->f_flag & FDIRECT) 467 ioflag |= IO_DIRECT; 468 mutex_exit(&fp->f_lock); 469 VOP_LEASE(vp, l, cred, LEASE_WRITE); 470 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 471 uio->uio_offset = *offset; 472 count = uio->uio_resid; 473 error = VOP_WRITE(vp, uio, ioflag, cred); 474 if (flags & FOF_UPDATE_OFFSET) { 475 if (ioflag & IO_APPEND) 476 *offset = uio->uio_offset; 477 else 478 *offset += count - uio->uio_resid; 479 } 480 VOP_UNLOCK(vp, 0); 481 return (error); 482 } 483 484 /* 485 * File table vnode stat routine. 486 */ 487 static int 488 vn_statfile(struct file *fp, struct stat *sb, struct lwp *l) 489 { 490 struct vnode *vp = (struct vnode *)fp->f_data; 491 492 return vn_stat(vp, sb, l); 493 } 494 495 int 496 vn_stat(struct vnode *vp, struct stat *sb, struct lwp *l) 497 { 498 struct vattr va; 499 int error; 500 mode_t mode; 501 502 error = VOP_GETATTR(vp, &va, l->l_cred, l); 503 if (error) 504 return (error); 505 /* 506 * Copy from vattr table 507 */ 508 sb->st_dev = va.va_fsid; 509 sb->st_ino = va.va_fileid; 510 mode = va.va_mode; 511 switch (vp->v_type) { 512 case VREG: 513 mode |= S_IFREG; 514 break; 515 case VDIR: 516 mode |= S_IFDIR; 517 break; 518 case VBLK: 519 mode |= S_IFBLK; 520 break; 521 case VCHR: 522 mode |= S_IFCHR; 523 break; 524 case VLNK: 525 mode |= S_IFLNK; 526 break; 527 case VSOCK: 528 mode |= S_IFSOCK; 529 break; 530 case VFIFO: 531 mode |= S_IFIFO; 532 break; 533 default: 534 return (EBADF); 535 }; 536 sb->st_mode = mode; 537 sb->st_nlink = va.va_nlink; 538 sb->st_uid = va.va_uid; 539 sb->st_gid = va.va_gid; 540 sb->st_rdev = va.va_rdev; 541 sb->st_size = va.va_size; 542 sb->st_atimespec = va.va_atime; 543 sb->st_mtimespec = va.va_mtime; 544 sb->st_ctimespec = va.va_ctime; 545 sb->st_birthtimespec = va.va_birthtime; 546 sb->st_blksize = va.va_blocksize; 547 sb->st_flags = va.va_flags; 548 sb->st_gen = 0; 549 sb->st_blocks = va.va_bytes / S_BLKSIZE; 550 return (0); 551 } 552 553 /* 554 * File table vnode fcntl routine. 555 */ 556 static int 557 vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l) 558 { 559 struct vnode *vp = ((struct vnode *)fp->f_data); 560 int error; 561 562 error = VOP_FCNTL(vp, com, data, fp->f_flag, l->l_cred, l); 563 return (error); 564 } 565 566 /* 567 * File table vnode ioctl routine. 568 */ 569 static int 570 vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l) 571 { 572 struct vnode *vp = ((struct vnode *)fp->f_data), *ovp; 573 struct proc *p = l->l_proc; 574 struct vattr vattr; 575 int error; 576 577 switch (vp->v_type) { 578 579 case VREG: 580 case VDIR: 581 if (com == FIONREAD) { 582 error = VOP_GETATTR(vp, &vattr, l->l_cred, l); 583 if (error) 584 return (error); 585 *(int *)data = vattr.va_size - fp->f_offset; 586 return (0); 587 } 588 if ((com == FIONWRITE) || (com == FIONSPACE)) { 589 /* 590 * Files don't have send queues, so there never 591 * are any bytes in them, nor is there any 592 * open space in them. 593 */ 594 *(int *)data = 0; 595 return (0); 596 } 597 if (com == FIOGETBMAP) { 598 daddr_t *block; 599 600 if (*(daddr_t *)data < 0) 601 return (EINVAL); 602 block = (daddr_t *)data; 603 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 604 } 605 if (com == OFIOGETBMAP) { 606 daddr_t ibn, obn; 607 608 if (*(int32_t *)data < 0) 609 return (EINVAL); 610 ibn = (daddr_t)*(int32_t *)data; 611 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 612 *(int32_t *)data = (int32_t)obn; 613 return error; 614 } 615 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 616 return (0); /* XXX */ 617 /* fall into ... */ 618 case VFIFO: 619 case VCHR: 620 case VBLK: 621 error = VOP_IOCTL(vp, com, data, fp->f_flag, 622 l->l_cred, l); 623 if (error == 0 && com == TIOCSCTTY) { 624 VREF(vp); 625 mutex_enter(&proclist_lock); 626 ovp = p->p_session->s_ttyvp; 627 p->p_session->s_ttyvp = vp; 628 mutex_exit(&proclist_lock); 629 if (ovp != NULL) 630 vrele(ovp); 631 } 632 return (error); 633 634 default: 635 return (EPASSTHROUGH); 636 } 637 } 638 639 /* 640 * File table vnode poll routine. 641 */ 642 static int 643 vn_poll(struct file *fp, int events, struct lwp *l) 644 { 645 646 return (VOP_POLL(((struct vnode *)fp->f_data), events, l)); 647 } 648 649 /* 650 * File table vnode kqfilter routine. 651 */ 652 int 653 vn_kqfilter(struct file *fp, struct knote *kn) 654 { 655 656 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 657 } 658 659 /* 660 * Check that the vnode is still valid, and if so 661 * acquire requested lock. 662 */ 663 int 664 vn_lock(struct vnode *vp, int flags) 665 { 666 int error; 667 668 #if 0 669 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 670 || (vp->v_iflag & VI_ONWORKLST) != 0); 671 #endif 672 KASSERT((flags & 673 ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_DRAIN|LK_NOWAIT|LK_RETRY| 674 LK_SETRECURSE|LK_CANRECURSE)) 675 == 0); 676 677 do { 678 if ((flags & LK_INTERLOCK) == 0) 679 mutex_enter(&vp->v_interlock); 680 if (vp->v_iflag & VI_XLOCK) { 681 if (flags & LK_NOWAIT) { 682 mutex_exit(&vp->v_interlock); 683 return EBUSY; 684 } 685 vwait(vp, VI_XLOCK); 686 mutex_exit(&vp->v_interlock); 687 error = ENOENT; 688 } else { 689 error = VOP_LOCK(vp, 690 (flags & ~LK_RETRY) | LK_INTERLOCK); 691 if (error == 0 || error == EDEADLK || error == EBUSY) 692 return (error); 693 } 694 flags &= ~LK_INTERLOCK; 695 } while (flags & LK_RETRY); 696 return (error); 697 } 698 699 /* 700 * File table vnode close routine. 701 */ 702 static int 703 vn_closefile(struct file *fp, struct lwp *l) 704 { 705 706 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 707 fp->f_cred, l)); 708 } 709 710 /* 711 * Enable LK_CANRECURSE on lock. Return prior status. 712 */ 713 u_int 714 vn_setrecurse(struct vnode *vp) 715 { 716 struct lock *lkp = &vp->v_lock; 717 u_int retval; 718 719 mutex_enter(&lkp->lk_interlock); 720 retval = lkp->lk_flags & LK_CANRECURSE; 721 lkp->lk_flags |= LK_CANRECURSE; 722 mutex_exit(&lkp->lk_interlock); 723 724 return retval; 725 } 726 727 /* 728 * Called when done with locksetrecurse. 729 */ 730 void 731 vn_restorerecurse(struct vnode *vp, u_int flags) 732 { 733 struct lock *lkp = &vp->v_lock; 734 735 mutex_enter(&lkp->lk_interlock); 736 lkp->lk_flags &= ~LK_CANRECURSE; 737 lkp->lk_flags |= flags; 738 mutex_exit(&lkp->lk_interlock); 739 } 740 741 int 742 vn_cow_establish(struct vnode *vp, 743 int (*func)(void *, struct buf *), void *cookie) 744 { 745 struct spec_cow_entry *e; 746 747 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 748 M_DEVBUF, M_WAITOK); 749 e->ce_func = func; 750 e->ce_cookie = cookie; 751 752 mutex_enter(&vp->v_spec_cow_lock); 753 vp->v_spec_cow_req++; 754 while (vp->v_spec_cow_count > 0) 755 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 756 &vp->v_spec_cow_lock); 757 758 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 759 760 vp->v_spec_cow_req--; 761 if (vp->v_spec_cow_req == 0) 762 wakeup(&vp->v_spec_cow_req); 763 mutex_exit(&vp->v_spec_cow_lock); 764 765 return 0; 766 } 767 768 int 769 vn_cow_disestablish(struct vnode *vp, 770 int (*func)(void *, struct buf *), void *cookie) 771 { 772 struct spec_cow_entry *e; 773 774 mutex_enter(&vp->v_spec_cow_lock); 775 vp->v_spec_cow_req++; 776 while (vp->v_spec_cow_count > 0) 777 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 778 &vp->v_spec_cow_lock); 779 780 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 781 if (e->ce_func == func && e->ce_cookie == cookie) { 782 SLIST_REMOVE(&vp->v_spec_cow_head, e, 783 spec_cow_entry, ce_list); 784 FREE(e, M_DEVBUF); 785 break; 786 } 787 788 vp->v_spec_cow_req--; 789 if (vp->v_spec_cow_req == 0) 790 wakeup(&vp->v_spec_cow_req); 791 mutex_exit(&vp->v_spec_cow_lock); 792 793 return e ? 0 : EINVAL; 794 } 795 796 /* 797 * Simplified in-kernel wrapper calls for extended attribute access. 798 * Both calls pass in a NULL credential, authorizing a "kernel" access. 799 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 800 */ 801 int 802 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 803 const char *attrname, size_t *buflen, void *bf, struct lwp *l) 804 { 805 struct uio auio; 806 struct iovec aiov; 807 int error; 808 809 aiov.iov_len = *buflen; 810 aiov.iov_base = bf; 811 812 auio.uio_iov = &aiov; 813 auio.uio_iovcnt = 1; 814 auio.uio_rw = UIO_READ; 815 auio.uio_offset = 0; 816 auio.uio_resid = *buflen; 817 UIO_SETUP_SYSSPACE(&auio); 818 819 if ((ioflg & IO_NODELOCKED) == 0) 820 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 821 822 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 823 l); 824 825 if ((ioflg & IO_NODELOCKED) == 0) 826 VOP_UNLOCK(vp, 0); 827 828 if (error == 0) 829 *buflen = *buflen - auio.uio_resid; 830 831 return (error); 832 } 833 834 /* 835 * XXX Failure mode if partially written? 836 */ 837 int 838 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 839 const char *attrname, size_t buflen, const void *bf, struct lwp *l) 840 { 841 struct uio auio; 842 struct iovec aiov; 843 int error; 844 845 aiov.iov_len = buflen; 846 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 847 848 auio.uio_iov = &aiov; 849 auio.uio_iovcnt = 1; 850 auio.uio_rw = UIO_WRITE; 851 auio.uio_offset = 0; 852 auio.uio_resid = buflen; 853 UIO_SETUP_SYSSPACE(&auio); 854 855 if ((ioflg & IO_NODELOCKED) == 0) { 856 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 857 } 858 859 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, l); 860 861 if ((ioflg & IO_NODELOCKED) == 0) { 862 VOP_UNLOCK(vp, 0); 863 } 864 865 return (error); 866 } 867 868 int 869 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 870 const char *attrname, struct lwp *l) 871 { 872 int error; 873 874 if ((ioflg & IO_NODELOCKED) == 0) { 875 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 876 } 877 878 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, l); 879 if (error == EOPNOTSUPP) 880 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 881 NULL, l); 882 883 if ((ioflg & IO_NODELOCKED) == 0) { 884 VOP_UNLOCK(vp, 0); 885 } 886 887 return (error); 888 } 889 890 void 891 vn_ra_allocctx(struct vnode *vp) 892 { 893 struct uvm_ractx *ra = NULL; 894 895 if (vp->v_type != VREG) { 896 return; 897 } 898 if (vp->v_ractx != NULL) { 899 return; 900 } 901 mutex_enter(&vp->v_interlock); 902 if (vp->v_ractx == NULL) { 903 mutex_exit(&vp->v_interlock); 904 ra = uvm_ra_allocctx(); 905 mutex_enter(&vp->v_interlock); 906 if (ra != NULL && vp->v_ractx == NULL) { 907 vp->v_ractx = ra; 908 ra = NULL; 909 } 910 } 911 mutex_exit(&vp->v_interlock); 912 if (ra != NULL) { 913 uvm_ra_freectx(ra); 914 } 915 } 916
Indexes created Sat Sep 20 22:09:52 GMT 2025