vfs_vnops.c revision 1.135.2.4
1 /* $NetBSD: vfs_vnops.c,v 1.135.2.4 2007/04/13 15:49:49 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 1.135.2.4 2007/04/13 15:49:49 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 mutex_enter(&vp->v_interlock); 216 vp->v_writecount++; 217 mutex_exit(&vp->v_interlock); 218 } 219 220 bad: 221 if (error) 222 vput(vp); 223 224 bad2: 225 pathname_put(pn); 226 227 return (error); 228 } 229 230 /* 231 * Check for write permissions on the specified vnode. 232 * Prototype text segments cannot be written. 233 */ 234 int 235 vn_writechk(struct vnode *vp) 236 { 237 238 /* 239 * If the vnode is in use as a process's text, 240 * we can't allow writing. 241 */ 242 if (vp->v_flag & VTEXT) 243 return (ETXTBSY); 244 return (0); 245 } 246 247 /* 248 * Mark a vnode as having executable mappings. 249 */ 250 void 251 vn_markexec(struct vnode *vp) 252 { 253 254 KASSERT(mutex_owned(&vp->v_interlock)); 255 256 if ((vp->v_flag & VEXECMAP) == 0) { 257 /* XXXSMP should be atomic */ 258 uvmexp.filepages -= vp->v_uobj.uo_npages; 259 uvmexp.execpages += vp->v_uobj.uo_npages; 260 } 261 vp->v_flag |= VEXECMAP; 262 } 263 264 /* 265 * Mark a vnode as being the text of a process. 266 * Fail if the vnode is currently writable. 267 */ 268 int 269 vn_marktext(struct vnode *vp) 270 { 271 272 mutex_enter(&vp->v_interlock); 273 if (vp->v_writecount != 0) { 274 KASSERT((vp->v_flag & VTEXT) == 0); 275 mutex_exit(&vp->v_interlock); 276 return (ETXTBSY); 277 } 278 vp->v_flag |= VTEXT; 279 vn_markexec(vp); 280 mutex_exit(&vp->v_interlock); 281 return (0); 282 } 283 284 /* 285 * Vnode close call 286 * 287 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 288 */ 289 int 290 vn_close(struct vnode *vp, int flags, kauth_cred_t cred, struct lwp *l) 291 { 292 int error; 293 294 mutex_enter(&vp->v_interlock); 295 if (flags & FWRITE) 296 vp->v_writecount--; 297 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_INTERLOCK); 298 error = VOP_CLOSE(vp, flags, cred, l); 299 vput(vp); 300 return (error); 301 } 302 303 /* 304 * Package up an I/O request on a vnode into a uio and do it. 305 */ 306 int 307 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset, 308 enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid, 309 struct lwp *l) 310 { 311 struct uio auio; 312 struct iovec aiov; 313 int error; 314 315 if ((ioflg & IO_NODELOCKED) == 0) { 316 if (rw == UIO_READ) { 317 vn_lock(vp, LK_SHARED | LK_RETRY); 318 } else /* UIO_WRITE */ { 319 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 320 } 321 } 322 auio.uio_iov = &aiov; 323 auio.uio_iovcnt = 1; 324 aiov.iov_base = base; 325 aiov.iov_len = len; 326 auio.uio_resid = len; 327 auio.uio_offset = offset; 328 auio.uio_rw = rw; 329 if (segflg == UIO_SYSSPACE) { 330 UIO_SETUP_SYSSPACE(&auio); 331 } else { 332 auio.uio_vmspace = l->l_proc->p_vmspace; 333 } 334 if (rw == UIO_READ) { 335 error = VOP_READ(vp, &auio, ioflg, cred); 336 } else { 337 error = VOP_WRITE(vp, &auio, ioflg, cred); 338 } 339 if (aresid) 340 *aresid = auio.uio_resid; 341 else 342 if (auio.uio_resid && error == 0) 343 error = EIO; 344 if ((ioflg & IO_NODELOCKED) == 0) { 345 VOP_UNLOCK(vp, 0); 346 } 347 return (error); 348 } 349 350 int 351 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 352 struct lwp *l, off_t **cookies, int *ncookies) 353 { 354 struct vnode *vp = (struct vnode *)fp->f_data; 355 struct iovec aiov; 356 struct uio auio; 357 int error, eofflag; 358 359 /* Limit the size on any kernel buffers used by VOP_READDIR */ 360 count = min(MAXBSIZE, count); 361 362 unionread: 363 if (vp->v_type != VDIR) 364 return (EINVAL); 365 aiov.iov_base = bf; 366 aiov.iov_len = count; 367 auio.uio_iov = &aiov; 368 auio.uio_iovcnt = 1; 369 auio.uio_rw = UIO_READ; 370 if (segflg == UIO_SYSSPACE) { 371 UIO_SETUP_SYSSPACE(&auio); 372 } else { 373 KASSERT(l == curlwp); 374 auio.uio_vmspace = l->l_proc->p_vmspace; 375 } 376 auio.uio_resid = count; 377 vn_lock(vp, LK_SHARED | LK_RETRY); 378 auio.uio_offset = fp->f_offset; 379 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 380 ncookies); 381 mutex_enter(&fp->f_lock); 382 fp->f_offset = auio.uio_offset; 383 mutex_exit(&fp->f_lock); 384 VOP_UNLOCK(vp, 0); 385 if (error) 386 return (error); 387 388 #if defined(UNION) || defined(LKM) 389 if (count == auio.uio_resid && vn_union_readdir_hook) { 390 struct vnode *ovp = vp; 391 392 error = (*vn_union_readdir_hook)(&vp, fp, l); 393 if (error) 394 return (error); 395 if (vp != ovp) 396 goto unionread; 397 } 398 #endif /* UNION || LKM */ 399 400 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 401 (vp->v_mount->mnt_flag & MNT_UNION)) { 402 struct vnode *tvp = vp; 403 vp = vp->v_mount->mnt_vnodecovered; 404 VREF(vp); 405 mutex_enter(&fp->f_lock); 406 fp->f_data = vp; 407 fp->f_offset = 0; 408 mutex_exit(&fp->f_lock); 409 vrele(tvp); 410 goto unionread; 411 } 412 *done = count - auio.uio_resid; 413 return error; 414 } 415 416 /* 417 * File table vnode read routine. 418 */ 419 static int 420 vn_read(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 421 int flags) 422 { 423 struct vnode *vp = (struct vnode *)fp->f_data; 424 int count, error, ioflag; 425 struct lwp *l = curlwp; 426 427 VOP_LEASE(vp, l, cred, LEASE_READ); 428 mutex_enter(&fp->f_lock); 429 ioflag = IO_ADV_ENCODE(fp->f_advice); 430 if (fp->f_flag & FNONBLOCK) 431 ioflag |= IO_NDELAY; 432 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 433 ioflag |= IO_SYNC; 434 if (fp->f_flag & FALTIO) 435 ioflag |= IO_ALTSEMANTICS; 436 if (fp->f_flag & FDIRECT) 437 ioflag |= IO_DIRECT; 438 mutex_exit(&fp->f_lock); 439 vn_lock(vp, LK_SHARED | LK_RETRY); 440 uio->uio_offset = *offset; 441 count = uio->uio_resid; 442 error = VOP_READ(vp, uio, ioflag, cred); 443 if (flags & FOF_UPDATE_OFFSET) 444 *offset += count - uio->uio_resid; 445 VOP_UNLOCK(vp, 0); 446 return (error); 447 } 448 449 /* 450 * File table vnode write routine. 451 */ 452 static int 453 vn_write(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 454 int flags) 455 { 456 struct vnode *vp = (struct vnode *)fp->f_data; 457 int count, error, ioflag = IO_UNIT; 458 struct lwp *l = curlwp; 459 460 mutex_enter(&fp->f_lock); 461 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 462 ioflag |= IO_APPEND; 463 if (fp->f_flag & FNONBLOCK) 464 ioflag |= IO_NDELAY; 465 if (fp->f_flag & FFSYNC || 466 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 467 ioflag |= IO_SYNC; 468 else if (fp->f_flag & FDSYNC) 469 ioflag |= IO_DSYNC; 470 if (fp->f_flag & FALTIO) 471 ioflag |= IO_ALTSEMANTICS; 472 if (fp->f_flag & FDIRECT) 473 ioflag |= IO_DIRECT; 474 mutex_exit(&fp->f_lock); 475 VOP_LEASE(vp, l, cred, LEASE_WRITE); 476 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 477 uio->uio_offset = *offset; 478 count = uio->uio_resid; 479 error = VOP_WRITE(vp, uio, ioflag, cred); 480 if (flags & FOF_UPDATE_OFFSET) { 481 if (ioflag & IO_APPEND) 482 *offset = uio->uio_offset; 483 else 484 *offset += count - uio->uio_resid; 485 } 486 VOP_UNLOCK(vp, 0); 487 return (error); 488 } 489 490 /* 491 * File table vnode stat routine. 492 */ 493 static int 494 vn_statfile(struct file *fp, struct stat *sb, struct lwp *l) 495 { 496 struct vnode *vp = (struct vnode *)fp->f_data; 497 498 return vn_stat(vp, sb, l); 499 } 500 501 int 502 vn_stat(struct vnode *vp, struct stat *sb, struct lwp *l) 503 { 504 struct vattr va; 505 int error; 506 mode_t mode; 507 508 error = VOP_GETATTR(vp, &va, l->l_cred, l); 509 if (error) 510 return (error); 511 /* 512 * Copy from vattr table 513 */ 514 sb->st_dev = va.va_fsid; 515 sb->st_ino = va.va_fileid; 516 mode = va.va_mode; 517 switch (vp->v_type) { 518 case VREG: 519 mode |= S_IFREG; 520 break; 521 case VDIR: 522 mode |= S_IFDIR; 523 break; 524 case VBLK: 525 mode |= S_IFBLK; 526 break; 527 case VCHR: 528 mode |= S_IFCHR; 529 break; 530 case VLNK: 531 mode |= S_IFLNK; 532 break; 533 case VSOCK: 534 mode |= S_IFSOCK; 535 break; 536 case VFIFO: 537 mode |= S_IFIFO; 538 break; 539 default: 540 return (EBADF); 541 }; 542 sb->st_mode = mode; 543 sb->st_nlink = va.va_nlink; 544 sb->st_uid = va.va_uid; 545 sb->st_gid = va.va_gid; 546 sb->st_rdev = va.va_rdev; 547 sb->st_size = va.va_size; 548 sb->st_atimespec = va.va_atime; 549 sb->st_mtimespec = va.va_mtime; 550 sb->st_ctimespec = va.va_ctime; 551 sb->st_birthtimespec = va.va_birthtime; 552 sb->st_blksize = va.va_blocksize; 553 sb->st_flags = va.va_flags; 554 sb->st_gen = 0; 555 sb->st_blocks = va.va_bytes / S_BLKSIZE; 556 return (0); 557 } 558 559 /* 560 * File table vnode fcntl routine. 561 */ 562 static int 563 vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l) 564 { 565 struct vnode *vp = ((struct vnode *)fp->f_data); 566 int error; 567 568 error = VOP_FCNTL(vp, com, data, fp->f_flag, l->l_cred, l); 569 return (error); 570 } 571 572 /* 573 * File table vnode ioctl routine. 574 */ 575 static int 576 vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l) 577 { 578 struct vnode *vp = ((struct vnode *)fp->f_data), *ovp; 579 struct proc *p = l->l_proc; 580 struct vattr vattr; 581 int error; 582 583 switch (vp->v_type) { 584 585 case VREG: 586 case VDIR: 587 if (com == FIONREAD) { 588 error = VOP_GETATTR(vp, &vattr, l->l_cred, l); 589 if (error) 590 return (error); 591 *(int *)data = vattr.va_size - fp->f_offset; 592 return (0); 593 } 594 if ((com == FIONWRITE) || (com == FIONSPACE)) { 595 /* 596 * Files don't have send queues, so there never 597 * are any bytes in them, nor is there any 598 * open space in them. 599 */ 600 *(int *)data = 0; 601 return (0); 602 } 603 if (com == FIOGETBMAP) { 604 daddr_t *block; 605 606 if (*(daddr_t *)data < 0) 607 return (EINVAL); 608 block = (daddr_t *)data; 609 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 610 } 611 if (com == OFIOGETBMAP) { 612 daddr_t ibn, obn; 613 614 if (*(int32_t *)data < 0) 615 return (EINVAL); 616 ibn = (daddr_t)*(int32_t *)data; 617 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 618 *(int32_t *)data = (int32_t)obn; 619 return error; 620 } 621 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 622 return (0); /* XXX */ 623 /* fall into ... */ 624 case VFIFO: 625 case VCHR: 626 case VBLK: 627 error = VOP_IOCTL(vp, com, data, fp->f_flag, 628 l->l_cred, l); 629 if (error == 0 && com == TIOCSCTTY) { 630 VREF(vp); 631 mutex_enter(&proclist_lock); 632 ovp = p->p_session->s_ttyvp; 633 p->p_session->s_ttyvp = vp; 634 mutex_exit(&proclist_lock); 635 if (ovp != NULL) 636 vrele(ovp); 637 } 638 return (error); 639 640 default: 641 return (EPASSTHROUGH); 642 } 643 } 644 645 /* 646 * File table vnode poll routine. 647 */ 648 static int 649 vn_poll(struct file *fp, int events, struct lwp *l) 650 { 651 652 return (VOP_POLL(((struct vnode *)fp->f_data), events, l)); 653 } 654 655 /* 656 * File table vnode kqfilter routine. 657 */ 658 int 659 vn_kqfilter(struct file *fp, struct knote *kn) 660 { 661 662 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 663 } 664 665 /* 666 * Check that the vnode is still valid, and if so 667 * acquire requested lock. 668 */ 669 int 670 vn_lock(struct vnode *vp, int flags) 671 { 672 int error; 673 674 #if 0 675 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 676 || (vp->v_flag & VONWORKLST) != 0); 677 #endif 678 KASSERT((flags & 679 ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_DRAIN|LK_NOWAIT|LK_RETRY| 680 LK_SETRECURSE|LK_CANRECURSE)) 681 == 0); 682 683 do { 684 if ((flags & LK_INTERLOCK) == 0) 685 mutex_enter(&vp->v_interlock); 686 if (vp->v_flag & VXLOCK) { 687 if (flags & LK_NOWAIT) { 688 mutex_exit(&vp->v_interlock); 689 return EBUSY; 690 } 691 vwait(vp, VXLOCK); 692 mutex_exit(&vp->v_interlock); 693 error = ENOENT; 694 } else { 695 error = VOP_LOCK(vp, 696 (flags & ~LK_RETRY) | LK_INTERLOCK); 697 if (error == 0 || error == EDEADLK || error == EBUSY) 698 return (error); 699 } 700 flags &= ~LK_INTERLOCK; 701 } while (flags & LK_RETRY); 702 return (error); 703 } 704 705 /* 706 * File table vnode close routine. 707 */ 708 static int 709 vn_closefile(struct file *fp, struct lwp *l) 710 { 711 712 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 713 fp->f_cred, l)); 714 } 715 716 /* 717 * Enable LK_CANRECURSE on lock. Return prior status. 718 */ 719 u_int 720 vn_setrecurse(struct vnode *vp) 721 { 722 struct lock *lkp = &vp->v_lock; 723 u_int retval; 724 725 mutex_enter(&lkp->lk_interlock); 726 retval = lkp->lk_flags & LK_CANRECURSE; 727 lkp->lk_flags |= LK_CANRECURSE; 728 mutex_exit(&lkp->lk_interlock); 729 730 return retval; 731 } 732 733 /* 734 * Called when done with locksetrecurse. 735 */ 736 void 737 vn_restorerecurse(struct vnode *vp, u_int flags) 738 { 739 struct lock *lkp = &vp->v_lock; 740 741 mutex_enter(&lkp->lk_interlock); 742 lkp->lk_flags &= ~LK_CANRECURSE; 743 lkp->lk_flags |= flags; 744 mutex_exit(&lkp->lk_interlock); 745 } 746 747 int 748 vn_cow_establish(struct vnode *vp, 749 int (*func)(void *, struct buf *), void *cookie) 750 { 751 struct spec_cow_entry *e; 752 753 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 754 M_DEVBUF, M_WAITOK); 755 e->ce_func = func; 756 e->ce_cookie = cookie; 757 758 mutex_enter(&vp->v_spec_cow_lock); 759 vp->v_spec_cow_req++; 760 while (vp->v_spec_cow_count > 0) 761 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 762 &vp->v_spec_cow_lock); 763 764 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 765 766 vp->v_spec_cow_req--; 767 if (vp->v_spec_cow_req == 0) 768 wakeup(&vp->v_spec_cow_req); 769 mutex_exit(&vp->v_spec_cow_lock); 770 771 return 0; 772 } 773 774 int 775 vn_cow_disestablish(struct vnode *vp, 776 int (*func)(void *, struct buf *), void *cookie) 777 { 778 struct spec_cow_entry *e; 779 780 mutex_enter(&vp->v_spec_cow_lock); 781 vp->v_spec_cow_req++; 782 while (vp->v_spec_cow_count > 0) 783 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 784 &vp->v_spec_cow_lock); 785 786 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 787 if (e->ce_func == func && e->ce_cookie == cookie) { 788 SLIST_REMOVE(&vp->v_spec_cow_head, e, 789 spec_cow_entry, ce_list); 790 FREE(e, M_DEVBUF); 791 break; 792 } 793 794 vp->v_spec_cow_req--; 795 if (vp->v_spec_cow_req == 0) 796 wakeup(&vp->v_spec_cow_req); 797 mutex_exit(&vp->v_spec_cow_lock); 798 799 return e ? 0 : EINVAL; 800 } 801 802 /* 803 * Simplified in-kernel wrapper calls for extended attribute access. 804 * Both calls pass in a NULL credential, authorizing a "kernel" access. 805 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 806 */ 807 int 808 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 809 const char *attrname, size_t *buflen, void *bf, struct lwp *l) 810 { 811 struct uio auio; 812 struct iovec aiov; 813 int error; 814 815 aiov.iov_len = *buflen; 816 aiov.iov_base = bf; 817 818 auio.uio_iov = &aiov; 819 auio.uio_iovcnt = 1; 820 auio.uio_rw = UIO_READ; 821 auio.uio_offset = 0; 822 auio.uio_resid = *buflen; 823 UIO_SETUP_SYSSPACE(&auio); 824 825 if ((ioflg & IO_NODELOCKED) == 0) 826 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 827 828 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 829 l); 830 831 if ((ioflg & IO_NODELOCKED) == 0) 832 VOP_UNLOCK(vp, 0); 833 834 if (error == 0) 835 *buflen = *buflen - auio.uio_resid; 836 837 return (error); 838 } 839 840 /* 841 * XXX Failure mode if partially written? 842 */ 843 int 844 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 845 const char *attrname, size_t buflen, const void *bf, struct lwp *l) 846 { 847 struct uio auio; 848 struct iovec aiov; 849 int error; 850 851 aiov.iov_len = buflen; 852 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 853 854 auio.uio_iov = &aiov; 855 auio.uio_iovcnt = 1; 856 auio.uio_rw = UIO_WRITE; 857 auio.uio_offset = 0; 858 auio.uio_resid = buflen; 859 UIO_SETUP_SYSSPACE(&auio); 860 861 if ((ioflg & IO_NODELOCKED) == 0) { 862 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 863 } 864 865 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, l); 866 867 if ((ioflg & IO_NODELOCKED) == 0) { 868 VOP_UNLOCK(vp, 0); 869 } 870 871 return (error); 872 } 873 874 int 875 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 876 const char *attrname, struct lwp *l) 877 { 878 int error; 879 880 if ((ioflg & IO_NODELOCKED) == 0) { 881 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 882 } 883 884 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, l); 885 if (error == EOPNOTSUPP) 886 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 887 NULL, l); 888 889 if ((ioflg & IO_NODELOCKED) == 0) { 890 VOP_UNLOCK(vp, 0); 891 } 892 893 return (error); 894 } 895 896 void 897 vn_ra_allocctx(struct vnode *vp) 898 { 899 struct uvm_ractx *ra = NULL; 900 901 if (vp->v_type != VREG) { 902 return; 903 } 904 if (vp->v_ractx != NULL) { 905 return; 906 } 907 mutex_enter(&vp->v_interlock); 908 if (vp->v_ractx == NULL) { 909 mutex_exit(&vp->v_interlock); 910 ra = uvm_ra_allocctx(); 911 mutex_enter(&vp->v_interlock); 912 if (ra != NULL && vp->v_ractx == NULL) { 913 vp->v_ractx = ra; 914 ra = NULL; 915 } 916 } 917 mutex_exit(&vp->v_interlock); 918 if (ra != NULL) { 919 uvm_ra_freectx(ra); 920 } 921 } 922
Indexes created Sat Sep 20 22:09:52 GMT 2025