vfs_vnops.c revision 1.135.2.7
1 /* $NetBSD: vfs_vnops.c,v 1.135.2.7 2007/08/20 21:27:45 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 1.135.2.7 2007/08/20 21:27:45 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 (fmode & FWRITE) { 206 mutex_enter(&vp->v_interlock); 207 vp->v_writecount++; 208 mutex_exit(&vp->v_interlock); 209 } 210 211 bad: 212 if (error) 213 vput(vp); 214 out: 215 VERIEXEC_PATH_PUT(path); 216 return (error); 217 } 218 219 /* 220 * Check for write permissions on the specified vnode. 221 * Prototype text segments cannot be written. 222 */ 223 int 224 vn_writechk(struct vnode *vp) 225 { 226 227 /* 228 * If the vnode is in use as a process's text, 229 * we can't allow writing. 230 */ 231 if (vp->v_iflag & VI_TEXT) 232 return (ETXTBSY); 233 return (0); 234 } 235 236 /* 237 * Mark a vnode as having executable mappings. 238 */ 239 void 240 vn_markexec(struct vnode *vp) 241 { 242 243 KASSERT(mutex_owned(&vp->v_interlock)); 244 245 if ((vp->v_iflag & VI_EXECMAP) == 0) { 246 /* XXXSMP should be atomic */ 247 uvmexp.filepages -= vp->v_uobj.uo_npages; 248 uvmexp.execpages += vp->v_uobj.uo_npages; 249 } 250 vp->v_iflag |= VI_EXECMAP; 251 } 252 253 /* 254 * Mark a vnode as being the text of a process. 255 * Fail if the vnode is currently writable. 256 */ 257 int 258 vn_marktext(struct vnode *vp) 259 { 260 261 mutex_enter(&vp->v_interlock); 262 if (vp->v_writecount != 0) { 263 KASSERT((vp->v_iflag & VI_TEXT) == 0); 264 mutex_exit(&vp->v_interlock); 265 return (ETXTBSY); 266 } 267 vp->v_iflag |= VI_TEXT; 268 vn_markexec(vp); 269 mutex_exit(&vp->v_interlock); 270 return (0); 271 } 272 273 /* 274 * Vnode close call 275 * 276 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 277 */ 278 int 279 vn_close(struct vnode *vp, int flags, kauth_cred_t cred, struct lwp *l) 280 { 281 int error; 282 283 mutex_enter(&vp->v_interlock); 284 if (flags & FWRITE) 285 vp->v_writecount--; 286 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_INTERLOCK); 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_vflag & VV_ROOT) && 390 (vp->v_mount->mnt_flag & MNT_UNION)) { 391 struct vnode *tvp = vp; 392 vp = vp->v_mount->mnt_vnodecovered; 393 VREF(vp); 394 mutex_enter(&fp->f_lock); 395 fp->f_data = vp; 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 mutex_enter(&fp->f_lock); 418 ioflag = IO_ADV_ENCODE(fp->f_advice); 419 if (fp->f_flag & FNONBLOCK) 420 ioflag |= IO_NDELAY; 421 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 422 ioflag |= IO_SYNC; 423 if (fp->f_flag & FALTIO) 424 ioflag |= IO_ALTSEMANTICS; 425 if (fp->f_flag & FDIRECT) 426 ioflag |= IO_DIRECT; 427 mutex_exit(&fp->f_lock); 428 vn_lock(vp, LK_SHARED | LK_RETRY); 429 uio->uio_offset = *offset; 430 count = uio->uio_resid; 431 error = VOP_READ(vp, uio, ioflag, cred); 432 if (flags & FOF_UPDATE_OFFSET) 433 *offset += count - uio->uio_resid; 434 VOP_UNLOCK(vp, 0); 435 return (error); 436 } 437 438 /* 439 * File table vnode write routine. 440 */ 441 static int 442 vn_write(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 443 int flags) 444 { 445 struct vnode *vp = (struct vnode *)fp->f_data; 446 int count, error, ioflag = IO_UNIT; 447 struct lwp *l = curlwp; 448 449 mutex_enter(&fp->f_lock); 450 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 451 ioflag |= IO_APPEND; 452 if (fp->f_flag & FNONBLOCK) 453 ioflag |= IO_NDELAY; 454 if (fp->f_flag & FFSYNC || 455 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 456 ioflag |= IO_SYNC; 457 else if (fp->f_flag & FDSYNC) 458 ioflag |= IO_DSYNC; 459 if (fp->f_flag & FALTIO) 460 ioflag |= IO_ALTSEMANTICS; 461 if (fp->f_flag & FDIRECT) 462 ioflag |= IO_DIRECT; 463 mutex_exit(&fp->f_lock); 464 VOP_LEASE(vp, l, cred, LEASE_WRITE); 465 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 466 uio->uio_offset = *offset; 467 count = uio->uio_resid; 468 error = VOP_WRITE(vp, uio, ioflag, cred); 469 if (flags & FOF_UPDATE_OFFSET) { 470 if (ioflag & IO_APPEND) 471 *offset = uio->uio_offset; 472 else 473 *offset += count - uio->uio_resid; 474 } 475 VOP_UNLOCK(vp, 0); 476 return (error); 477 } 478 479 /* 480 * File table vnode stat routine. 481 */ 482 static int 483 vn_statfile(struct file *fp, struct stat *sb, struct lwp *l) 484 { 485 struct vnode *vp = (struct vnode *)fp->f_data; 486 487 return vn_stat(vp, sb, l); 488 } 489 490 int 491 vn_stat(struct vnode *vp, struct stat *sb, struct lwp *l) 492 { 493 struct vattr va; 494 int error; 495 mode_t mode; 496 497 error = VOP_GETATTR(vp, &va, l->l_cred, l); 498 if (error) 499 return (error); 500 /* 501 * Copy from vattr table 502 */ 503 sb->st_dev = va.va_fsid; 504 sb->st_ino = va.va_fileid; 505 mode = va.va_mode; 506 switch (vp->v_type) { 507 case VREG: 508 mode |= S_IFREG; 509 break; 510 case VDIR: 511 mode |= S_IFDIR; 512 break; 513 case VBLK: 514 mode |= S_IFBLK; 515 break; 516 case VCHR: 517 mode |= S_IFCHR; 518 break; 519 case VLNK: 520 mode |= S_IFLNK; 521 break; 522 case VSOCK: 523 mode |= S_IFSOCK; 524 break; 525 case VFIFO: 526 mode |= S_IFIFO; 527 break; 528 default: 529 return (EBADF); 530 }; 531 sb->st_mode = mode; 532 sb->st_nlink = va.va_nlink; 533 sb->st_uid = va.va_uid; 534 sb->st_gid = va.va_gid; 535 sb->st_rdev = va.va_rdev; 536 sb->st_size = va.va_size; 537 sb->st_atimespec = va.va_atime; 538 sb->st_mtimespec = va.va_mtime; 539 sb->st_ctimespec = va.va_ctime; 540 sb->st_birthtimespec = va.va_birthtime; 541 sb->st_blksize = va.va_blocksize; 542 sb->st_flags = va.va_flags; 543 sb->st_gen = 0; 544 sb->st_blocks = va.va_bytes / S_BLKSIZE; 545 return (0); 546 } 547 548 /* 549 * File table vnode fcntl routine. 550 */ 551 static int 552 vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l) 553 { 554 struct vnode *vp = ((struct vnode *)fp->f_data); 555 int error; 556 557 error = VOP_FCNTL(vp, com, data, fp->f_flag, l->l_cred, l); 558 return (error); 559 } 560 561 /* 562 * File table vnode ioctl routine. 563 */ 564 static int 565 vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l) 566 { 567 struct vnode *vp = ((struct vnode *)fp->f_data), *ovp; 568 struct proc *p = l->l_proc; 569 struct vattr vattr; 570 int error; 571 572 switch (vp->v_type) { 573 574 case VREG: 575 case VDIR: 576 if (com == FIONREAD) { 577 error = VOP_GETATTR(vp, &vattr, l->l_cred, l); 578 if (error) 579 return (error); 580 *(int *)data = vattr.va_size - fp->f_offset; 581 return (0); 582 } 583 if ((com == FIONWRITE) || (com == FIONSPACE)) { 584 /* 585 * Files don't have send queues, so there never 586 * are any bytes in them, nor is there any 587 * open space in them. 588 */ 589 *(int *)data = 0; 590 return (0); 591 } 592 if (com == FIOGETBMAP) { 593 daddr_t *block; 594 595 if (*(daddr_t *)data < 0) 596 return (EINVAL); 597 block = (daddr_t *)data; 598 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 599 } 600 if (com == OFIOGETBMAP) { 601 daddr_t ibn, obn; 602 603 if (*(int32_t *)data < 0) 604 return (EINVAL); 605 ibn = (daddr_t)*(int32_t *)data; 606 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 607 *(int32_t *)data = (int32_t)obn; 608 return error; 609 } 610 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 611 return (0); /* XXX */ 612 /* fall into ... */ 613 case VFIFO: 614 case VCHR: 615 case VBLK: 616 error = VOP_IOCTL(vp, com, data, fp->f_flag, 617 l->l_cred, l); 618 if (error == 0 && com == TIOCSCTTY) { 619 VREF(vp); 620 mutex_enter(&proclist_lock); 621 ovp = p->p_session->s_ttyvp; 622 p->p_session->s_ttyvp = vp; 623 mutex_exit(&proclist_lock); 624 if (ovp != NULL) 625 vrele(ovp); 626 } 627 return (error); 628 629 default: 630 return (EPASSTHROUGH); 631 } 632 } 633 634 /* 635 * File table vnode poll routine. 636 */ 637 static int 638 vn_poll(struct file *fp, int events, struct lwp *l) 639 { 640 641 return (VOP_POLL(((struct vnode *)fp->f_data), events, l)); 642 } 643 644 /* 645 * File table vnode kqfilter routine. 646 */ 647 int 648 vn_kqfilter(struct file *fp, struct knote *kn) 649 { 650 651 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 652 } 653 654 /* 655 * Check that the vnode is still valid, and if so 656 * acquire requested lock. 657 */ 658 int 659 vn_lock(struct vnode *vp, int flags) 660 { 661 int error; 662 663 #if 0 664 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 665 || (vp->v_iflag & VI_ONWORKLST) != 0); 666 #endif 667 KASSERT((flags & 668 ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_DRAIN|LK_NOWAIT|LK_RETRY| 669 LK_SETRECURSE|LK_CANRECURSE)) 670 == 0); 671 672 do { 673 if ((flags & LK_INTERLOCK) == 0) 674 mutex_enter(&vp->v_interlock); 675 if (vp->v_iflag & VI_XLOCK) { 676 if (flags & LK_NOWAIT) { 677 mutex_exit(&vp->v_interlock); 678 return EBUSY; 679 } 680 vwait(vp, VI_XLOCK); 681 mutex_exit(&vp->v_interlock); 682 error = ENOENT; 683 } else { 684 error = VOP_LOCK(vp, 685 (flags & ~LK_RETRY) | LK_INTERLOCK); 686 if (error == 0 || error == EDEADLK || error == EBUSY) 687 return (error); 688 } 689 flags &= ~LK_INTERLOCK; 690 } while (flags & LK_RETRY); 691 return (error); 692 } 693 694 /* 695 * File table vnode close routine. 696 */ 697 static int 698 vn_closefile(struct file *fp, struct lwp *l) 699 { 700 701 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 702 fp->f_cred, l)); 703 } 704 705 /* 706 * Enable LK_CANRECURSE on lock. Return prior status. 707 */ 708 u_int 709 vn_setrecurse(struct vnode *vp) 710 { 711 struct lock *lkp = &vp->v_lock; 712 u_int retval; 713 714 mutex_enter(&lkp->lk_interlock); 715 retval = lkp->lk_flags & LK_CANRECURSE; 716 lkp->lk_flags |= LK_CANRECURSE; 717 mutex_exit(&lkp->lk_interlock); 718 719 return retval; 720 } 721 722 /* 723 * Called when done with locksetrecurse. 724 */ 725 void 726 vn_restorerecurse(struct vnode *vp, u_int flags) 727 { 728 struct lock *lkp = &vp->v_lock; 729 730 mutex_enter(&lkp->lk_interlock); 731 lkp->lk_flags &= ~LK_CANRECURSE; 732 lkp->lk_flags |= flags; 733 mutex_exit(&lkp->lk_interlock); 734 } 735 736 int 737 vn_cow_establish(struct vnode *vp, 738 int (*func)(void *, struct buf *), void *cookie) 739 { 740 struct spec_cow_entry *e; 741 742 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 743 M_DEVBUF, M_WAITOK); 744 e->ce_func = func; 745 e->ce_cookie = cookie; 746 747 mutex_enter(&vp->v_spec_cow_lock); 748 vp->v_spec_cow_req++; 749 while (vp->v_spec_cow_count > 0) 750 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 751 &vp->v_spec_cow_lock); 752 753 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 754 755 vp->v_spec_cow_req--; 756 if (vp->v_spec_cow_req == 0) 757 wakeup(&vp->v_spec_cow_req); 758 mutex_exit(&vp->v_spec_cow_lock); 759 760 return 0; 761 } 762 763 int 764 vn_cow_disestablish(struct vnode *vp, 765 int (*func)(void *, struct buf *), void *cookie) 766 { 767 struct spec_cow_entry *e; 768 769 mutex_enter(&vp->v_spec_cow_lock); 770 vp->v_spec_cow_req++; 771 while (vp->v_spec_cow_count > 0) 772 mtsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 773 &vp->v_spec_cow_lock); 774 775 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 776 if (e->ce_func == func && e->ce_cookie == cookie) { 777 SLIST_REMOVE(&vp->v_spec_cow_head, e, 778 spec_cow_entry, ce_list); 779 FREE(e, M_DEVBUF); 780 break; 781 } 782 783 vp->v_spec_cow_req--; 784 if (vp->v_spec_cow_req == 0) 785 wakeup(&vp->v_spec_cow_req); 786 mutex_exit(&vp->v_spec_cow_lock); 787 788 return e ? 0 : EINVAL; 789 } 790 791 /* 792 * Simplified in-kernel wrapper calls for extended attribute access. 793 * Both calls pass in a NULL credential, authorizing a "kernel" access. 794 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 795 */ 796 int 797 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 798 const char *attrname, size_t *buflen, void *bf, struct lwp *l) 799 { 800 struct uio auio; 801 struct iovec aiov; 802 int error; 803 804 aiov.iov_len = *buflen; 805 aiov.iov_base = bf; 806 807 auio.uio_iov = &aiov; 808 auio.uio_iovcnt = 1; 809 auio.uio_rw = UIO_READ; 810 auio.uio_offset = 0; 811 auio.uio_resid = *buflen; 812 UIO_SETUP_SYSSPACE(&auio); 813 814 if ((ioflg & IO_NODELOCKED) == 0) 815 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 816 817 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 818 l); 819 820 if ((ioflg & IO_NODELOCKED) == 0) 821 VOP_UNLOCK(vp, 0); 822 823 if (error == 0) 824 *buflen = *buflen - auio.uio_resid; 825 826 return (error); 827 } 828 829 /* 830 * XXX Failure mode if partially written? 831 */ 832 int 833 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 834 const char *attrname, size_t buflen, const void *bf, struct lwp *l) 835 { 836 struct uio auio; 837 struct iovec aiov; 838 int error; 839 840 aiov.iov_len = buflen; 841 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 842 843 auio.uio_iov = &aiov; 844 auio.uio_iovcnt = 1; 845 auio.uio_rw = UIO_WRITE; 846 auio.uio_offset = 0; 847 auio.uio_resid = buflen; 848 UIO_SETUP_SYSSPACE(&auio); 849 850 if ((ioflg & IO_NODELOCKED) == 0) { 851 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 852 } 853 854 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, l); 855 856 if ((ioflg & IO_NODELOCKED) == 0) { 857 VOP_UNLOCK(vp, 0); 858 } 859 860 return (error); 861 } 862 863 int 864 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 865 const char *attrname, struct lwp *l) 866 { 867 int error; 868 869 if ((ioflg & IO_NODELOCKED) == 0) { 870 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 871 } 872 873 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, l); 874 if (error == EOPNOTSUPP) 875 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 876 NULL, l); 877 878 if ((ioflg & IO_NODELOCKED) == 0) { 879 VOP_UNLOCK(vp, 0); 880 } 881 882 return (error); 883 } 884 885 void 886 vn_ra_allocctx(struct vnode *vp) 887 { 888 struct uvm_ractx *ra = NULL; 889 890 if (vp->v_type != VREG) { 891 return; 892 } 893 if (vp->v_ractx != NULL) { 894 return; 895 } 896 mutex_enter(&vp->v_interlock); 897 if (vp->v_ractx == NULL) { 898 mutex_exit(&vp->v_interlock); 899 ra = uvm_ra_allocctx(); 900 mutex_enter(&vp->v_interlock); 901 if (ra != NULL && vp->v_ractx == NULL) { 902 vp->v_ractx = ra; 903 ra = NULL; 904 } 905 } 906 mutex_exit(&vp->v_interlock); 907 if (ra != NULL) { 908 uvm_ra_freectx(ra); 909 } 910 } 911
Indexes created Sat Sep 20 22:09:52 GMT 2025