vfs_vnops.c revision 1.110
1/* $NetBSD: vfs_vnops.c,v 1.110 2006/05/14 05:30:31 christos 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 1.110 2006/05/14 05:30:31 christos Exp $"); 41 42#include "opt_verified_exec.h" 43 44#include "fs_union.h" 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/kernel.h> 49#include <sys/file.h> 50#include <sys/stat.h> 51#include <sys/buf.h> 52#include <sys/proc.h> 53#include <sys/malloc.h> 54#include <sys/mount.h> 55#include <sys/namei.h> 56#include <sys/vnode.h> 57#include <sys/ioctl.h> 58#include <sys/tty.h> 59#include <sys/poll.h> 60 61#include <miscfs/specfs/specdev.h> 62 63#include <uvm/uvm_extern.h> 64#include <uvm/uvm_readahead.h> 65 66#ifdef UNION 67#include <fs/union/union.h> 68#endif 69 70#if defined(LKM) || defined(UNION) 71int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct lwp *); 72#endif 73 74#ifdef VERIFIED_EXEC 75#include <sys/verified_exec.h> 76#endif 77 78static int vn_read(struct file *fp, off_t *offset, struct uio *uio, 79 struct ucred *cred, int flags); 80static int vn_write(struct file *fp, off_t *offset, struct uio *uio, 81 struct ucred *cred, int flags); 82static int vn_closefile(struct file *fp, struct lwp *l); 83static int vn_poll(struct file *fp, int events, struct lwp *l); 84static int vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l); 85static int vn_statfile(struct file *fp, struct stat *sb, struct lwp *l); 86static int vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l); 87 88const struct fileops vnops = { 89 vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, 90 vn_statfile, vn_closefile, vn_kqfilter 91}; 92 93/* 94 * Common code for vnode open operations. 95 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 96 */ 97int 98vn_open(struct nameidata *ndp, int fmode, int cmode) 99{ 100 struct vnode *vp; 101 struct mount *mp = NULL; /* XXX: GCC */ 102 struct lwp *l = ndp->ni_cnd.cn_lwp; 103 struct ucred *cred = l->l_proc->p_ucred; 104 struct vattr va; 105 int error; 106#ifdef VERIFIED_EXEC 107 struct veriexec_hash_entry *vhe = NULL; 108 char pathbuf[MAXPATHLEN]; 109 size_t pathlen; 110 int (*copyfun)(const void *, void *, size_t, size_t *) = 111 ndp->ni_segflg == UIO_SYSSPACE ? copystr : copyinstr; 112#endif /* VERIFIED_EXEC */ 113 114#ifdef VERIFIED_EXEC 115 error = (*copyfun)(ndp->ni_dirp, pathbuf, sizeof(pathbuf), &pathlen); 116 if (error) { 117 if (veriexec_verbose >= 1) 118 printf("veriexec: Can't copy path. (error=%d)\n", 119 error); 120 121 return (error); 122 } 123#endif /* VERIFIED_EXEC */ 124 125restart: 126 if (fmode & O_CREAT) { 127 ndp->ni_cnd.cn_nameiop = CREATE; 128 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 129 if ((fmode & O_EXCL) == 0 && 130 ((fmode & O_NOFOLLOW) == 0)) 131 ndp->ni_cnd.cn_flags |= FOLLOW; 132 if ((error = namei(ndp)) != 0) 133 return (error); 134 if (ndp->ni_vp == NULL) { 135#ifdef VERIFIED_EXEC 136 /* Lockdown mode: Prevent creation of new files. */ 137 if (veriexec_strict >= 3) { 138 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 139 140 printf("Veriexec: vn_open: Preventing " 141 "new file creation in %s.\n", 142 pathbuf); 143 144 vp = ndp->ni_dvp; 145 error = EPERM; 146 goto bad; 147 } 148#endif /* VERIFIED_EXEC */ 149 150 VATTR_NULL(&va); 151 va.va_type = VREG; 152 va.va_mode = cmode; 153 if (fmode & O_EXCL) 154 va.va_vaflags |= VA_EXCLUSIVE; 155 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 156 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 157 vput(ndp->ni_dvp); 158 if ((error = vn_start_write(NULL, &mp, 159 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 160 return (error); 161 goto restart; 162 } 163 VOP_LEASE(ndp->ni_dvp, l, cred, LEASE_WRITE); 164 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 165 &ndp->ni_cnd, &va); 166 vn_finished_write(mp, 0); 167 if (error) 168 return (error); 169 fmode &= ~O_TRUNC; 170 vp = ndp->ni_vp; 171 } else { 172 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 173 if (ndp->ni_dvp == ndp->ni_vp) 174 vrele(ndp->ni_dvp); 175 else 176 vput(ndp->ni_dvp); 177 ndp->ni_dvp = NULL; 178 vp = ndp->ni_vp; 179 if (fmode & O_EXCL) { 180 error = EEXIST; 181 goto bad; 182 } 183 fmode &= ~O_CREAT; 184 } 185 } else { 186 ndp->ni_cnd.cn_nameiop = LOOKUP; 187 ndp->ni_cnd.cn_flags = LOCKLEAF; 188 if ((fmode & O_NOFOLLOW) == 0) 189 ndp->ni_cnd.cn_flags |= FOLLOW; 190 if ((error = namei(ndp)) != 0) 191 return (error); 192 vp = ndp->ni_vp; 193 } 194 if (vp->v_type == VSOCK) { 195 error = EOPNOTSUPP; 196 goto bad; 197 } 198 if (ndp->ni_vp->v_type == VLNK) { 199 error = EFTYPE; 200 goto bad; 201 } 202 203#ifdef VERIFIED_EXEC 204 if ((error = VOP_GETATTR(vp, &va, cred, l)) != 0) 205 goto bad; 206#endif 207 208 if ((fmode & O_CREAT) == 0) { 209#ifdef VERIFIED_EXEC 210 if ((error = veriexec_verify(l, vp, &va, pathbuf, 211 VERIEXEC_FILE, &vhe)) != 0) 212 goto bad; 213#endif 214 215 if (fmode & FREAD) { 216 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 217 goto bad; 218 } 219 220 if (fmode & (FWRITE | O_TRUNC)) { 221 if (vp->v_type == VDIR) { 222 error = EISDIR; 223 goto bad; 224 } 225 if ((error = vn_writechk(vp)) != 0 || 226 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 227 goto bad; 228#ifdef VERIFIED_EXEC 229 if (vhe != NULL) { 230 veriexec_report("Write access request.", 231 pathbuf, &va, l, 232 REPORT_NOVERBOSE, 233 REPORT_ALARM, 234 REPORT_NOPANIC); 235 236 /* IPS mode: Deny writing to monitored files. */ 237 if (veriexec_strict >= 2) { 238 error = EPERM; 239 goto bad; 240 } else { 241 vhe->status = FINGERPRINT_NOTEVAL; 242 } 243 } 244#endif 245 } 246 } 247 248 if (fmode & O_TRUNC) { 249 VOP_UNLOCK(vp, 0); /* XXX */ 250 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 251 vrele(vp); 252 return (error); 253 } 254 VOP_LEASE(vp, l, cred, LEASE_WRITE); 255 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 256 VATTR_NULL(&va); 257 va.va_size = 0; 258 error = VOP_SETATTR(vp, &va, cred, l); 259 vn_finished_write(mp, 0); 260 if (error != 0) 261 goto bad; 262 } 263 if ((error = VOP_OPEN(vp, fmode, cred, l)) != 0) 264 goto bad; 265 if (vp->v_type == VREG && 266 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 267 error = EIO; 268 goto bad; 269 } 270 if (fmode & FWRITE) 271 vp->v_writecount++; 272 273 return (0); 274bad: 275 vput(vp); 276 return (error); 277} 278 279/* 280 * Check for write permissions on the specified vnode. 281 * Prototype text segments cannot be written. 282 */ 283int 284vn_writechk(struct vnode *vp) 285{ 286 287 /* 288 * If the vnode is in use as a process's text, 289 * we can't allow writing. 290 */ 291 if (vp->v_flag & VTEXT) 292 return (ETXTBSY); 293 return (0); 294} 295 296/* 297 * Mark a vnode as having executable mappings. 298 */ 299void 300vn_markexec(struct vnode *vp) 301{ 302 if ((vp->v_flag & VEXECMAP) == 0) { 303 uvmexp.filepages -= vp->v_uobj.uo_npages; 304 uvmexp.execpages += vp->v_uobj.uo_npages; 305 } 306 vp->v_flag |= VEXECMAP; 307} 308 309/* 310 * Mark a vnode as being the text of a process. 311 * Fail if the vnode is currently writable. 312 */ 313int 314vn_marktext(struct vnode *vp) 315{ 316 317 if (vp->v_writecount != 0) { 318 KASSERT((vp->v_flag & VTEXT) == 0); 319 return (ETXTBSY); 320 } 321 vp->v_flag |= VTEXT; 322 vn_markexec(vp); 323 return (0); 324} 325 326/* 327 * Vnode close call 328 * 329 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 330 */ 331int 332vn_close(struct vnode *vp, int flags, struct ucred *cred, struct lwp *l) 333{ 334 int error; 335 336 if (flags & FWRITE) 337 vp->v_writecount--; 338 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 339 error = VOP_CLOSE(vp, flags, cred, l); 340 vput(vp); 341 return (error); 342} 343 344/* 345 * Package up an I/O request on a vnode into a uio and do it. 346 */ 347int 348vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len, off_t offset, 349 enum uio_seg segflg, int ioflg, struct ucred *cred, size_t *aresid, 350 struct lwp *l) 351{ 352 struct uio auio; 353 struct iovec aiov; 354 struct mount *mp = NULL; 355 int error; 356 357 if ((ioflg & IO_NODELOCKED) == 0) { 358 if (rw == UIO_READ) { 359 vn_lock(vp, LK_SHARED | LK_RETRY); 360 } else /* UIO_WRITE */ { 361 if (vp->v_type != VCHR && 362 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 363 != 0) 364 return (error); 365 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 366 } 367 } 368 auio.uio_iov = &aiov; 369 auio.uio_iovcnt = 1; 370 aiov.iov_base = base; 371 aiov.iov_len = len; 372 auio.uio_resid = len; 373 auio.uio_offset = offset; 374 auio.uio_rw = rw; 375 if (segflg == UIO_SYSSPACE) { 376 UIO_SETUP_SYSSPACE(&auio); 377 } else { 378 auio.uio_vmspace = l->l_proc->p_vmspace; 379 } 380 if (rw == UIO_READ) { 381 error = VOP_READ(vp, &auio, ioflg, cred); 382 } else { 383 error = VOP_WRITE(vp, &auio, ioflg, cred); 384 } 385 if (aresid) 386 *aresid = auio.uio_resid; 387 else 388 if (auio.uio_resid && error == 0) 389 error = EIO; 390 if ((ioflg & IO_NODELOCKED) == 0) { 391 if (rw == UIO_WRITE) 392 vn_finished_write(mp, 0); 393 VOP_UNLOCK(vp, 0); 394 } 395 return (error); 396} 397 398int 399vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 400 struct lwp *l, off_t **cookies, int *ncookies) 401{ 402 struct vnode *vp = (struct vnode *)fp->f_data; 403 struct iovec aiov; 404 struct uio auio; 405 int error, eofflag; 406 407unionread: 408 if (vp->v_type != VDIR) 409 return (EINVAL); 410 aiov.iov_base = bf; 411 aiov.iov_len = count; 412 auio.uio_iov = &aiov; 413 auio.uio_iovcnt = 1; 414 auio.uio_rw = UIO_READ; 415 if (segflg == UIO_SYSSPACE) { 416 UIO_SETUP_SYSSPACE(&auio); 417 } else { 418 KASSERT(l == curlwp); 419 auio.uio_vmspace = l->l_proc->p_vmspace; 420 } 421 auio.uio_resid = count; 422 vn_lock(vp, LK_SHARED | LK_RETRY); 423 auio.uio_offset = fp->f_offset; 424 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 425 ncookies); 426 fp->f_offset = auio.uio_offset; 427 VOP_UNLOCK(vp, 0); 428 if (error) 429 return (error); 430 431#if defined(UNION) || defined(LKM) 432 if (count == auio.uio_resid && vn_union_readdir_hook) { 433 struct vnode *ovp = vp; 434 435 error = (*vn_union_readdir_hook)(&vp, fp, l); 436 if (error) 437 return (error); 438 if (vp != ovp) 439 goto unionread; 440 } 441#endif /* UNION || LKM */ 442 443 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 444 (vp->v_mount->mnt_flag & MNT_UNION)) { 445 struct vnode *tvp = vp; 446 vp = vp->v_mount->mnt_vnodecovered; 447 VREF(vp); 448 fp->f_data = vp; 449 fp->f_offset = 0; 450 vrele(tvp); 451 goto unionread; 452 } 453 *done = count - auio.uio_resid; 454 return error; 455} 456 457/* 458 * File table vnode read routine. 459 */ 460static int 461vn_read(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 462 int flags) 463{ 464 struct vnode *vp = (struct vnode *)fp->f_data; 465 int count, error, ioflag; 466 struct lwp *l = curlwp; 467 468 VOP_LEASE(vp, l, cred, LEASE_READ); 469 ioflag = IO_ADV_ENCODE(fp->f_advice); 470 if (fp->f_flag & FNONBLOCK) 471 ioflag |= IO_NDELAY; 472 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 473 ioflag |= IO_SYNC; 474 if (fp->f_flag & FALTIO) 475 ioflag |= IO_ALTSEMANTICS; 476 vn_lock(vp, LK_SHARED | LK_RETRY); 477 uio->uio_offset = *offset; 478 count = uio->uio_resid; 479 error = VOP_READ(vp, uio, ioflag, cred); 480 if (flags & FOF_UPDATE_OFFSET) 481 *offset += count - uio->uio_resid; 482 VOP_UNLOCK(vp, 0); 483 return (error); 484} 485 486/* 487 * File table vnode write routine. 488 */ 489static int 490vn_write(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 491 int flags) 492{ 493 struct vnode *vp = (struct vnode *)fp->f_data; 494 struct mount *mp; 495 int count, error, ioflag = IO_UNIT; 496 struct lwp *l = curlwp; 497 498 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 499 ioflag |= IO_APPEND; 500 if (fp->f_flag & FNONBLOCK) 501 ioflag |= IO_NDELAY; 502 if (fp->f_flag & FFSYNC || 503 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 504 ioflag |= IO_SYNC; 505 else if (fp->f_flag & FDSYNC) 506 ioflag |= IO_DSYNC; 507 if (fp->f_flag & FALTIO) 508 ioflag |= IO_ALTSEMANTICS; 509 mp = NULL; 510 if (vp->v_type != VCHR && 511 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 512 return (error); 513 VOP_LEASE(vp, l, cred, LEASE_WRITE); 514 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 515 uio->uio_offset = *offset; 516 count = uio->uio_resid; 517 error = VOP_WRITE(vp, uio, ioflag, cred); 518 if (flags & FOF_UPDATE_OFFSET) { 519 if (ioflag & IO_APPEND) 520 *offset = uio->uio_offset; 521 else 522 *offset += count - uio->uio_resid; 523 } 524 VOP_UNLOCK(vp, 0); 525 vn_finished_write(mp, 0); 526 return (error); 527} 528 529/* 530 * File table vnode stat routine. 531 */ 532static int 533vn_statfile(struct file *fp, struct stat *sb, struct lwp *l) 534{ 535 struct vnode *vp = (struct vnode *)fp->f_data; 536 537 return vn_stat(vp, sb, l); 538} 539 540int 541vn_stat(struct vnode *vp, struct stat *sb, struct lwp *l) 542{ 543 struct vattr va; 544 int error; 545 mode_t mode; 546 547 error = VOP_GETATTR(vp, &va, l->l_proc->p_ucred, l); 548 if (error) 549 return (error); 550 /* 551 * Copy from vattr table 552 */ 553 sb->st_dev = va.va_fsid; 554 sb->st_ino = va.va_fileid; 555 mode = va.va_mode; 556 switch (vp->v_type) { 557 case VREG: 558 mode |= S_IFREG; 559 break; 560 case VDIR: 561 mode |= S_IFDIR; 562 break; 563 case VBLK: 564 mode |= S_IFBLK; 565 break; 566 case VCHR: 567 mode |= S_IFCHR; 568 break; 569 case VLNK: 570 mode |= S_IFLNK; 571 break; 572 case VSOCK: 573 mode |= S_IFSOCK; 574 break; 575 case VFIFO: 576 mode |= S_IFIFO; 577 break; 578 default: 579 return (EBADF); 580 }; 581 sb->st_mode = mode; 582 sb->st_nlink = va.va_nlink; 583 sb->st_uid = va.va_uid; 584 sb->st_gid = va.va_gid; 585 sb->st_rdev = va.va_rdev; 586 sb->st_size = va.va_size; 587 sb->st_atimespec = va.va_atime; 588 sb->st_mtimespec = va.va_mtime; 589 sb->st_ctimespec = va.va_ctime; 590 sb->st_birthtimespec = va.va_birthtime; 591 sb->st_blksize = va.va_blocksize; 592 sb->st_flags = va.va_flags; 593 sb->st_gen = 0; 594 sb->st_blocks = va.va_bytes / S_BLKSIZE; 595 return (0); 596} 597 598/* 599 * File table vnode fcntl routine. 600 */ 601static int 602vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l) 603{ 604 struct vnode *vp = ((struct vnode *)fp->f_data); 605 int error; 606 607 error = VOP_FCNTL(vp, com, data, fp->f_flag, l->l_proc->p_ucred, l); 608 return (error); 609} 610 611/* 612 * File table vnode ioctl routine. 613 */ 614static int 615vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l) 616{ 617 struct vnode *vp = ((struct vnode *)fp->f_data); 618 struct proc *p = l->l_proc; 619 struct vattr vattr; 620 int error; 621 622 switch (vp->v_type) { 623 624 case VREG: 625 case VDIR: 626 if (com == FIONREAD) { 627 error = VOP_GETATTR(vp, &vattr, l->l_proc->p_ucred, l); 628 if (error) 629 return (error); 630 *(int *)data = vattr.va_size - fp->f_offset; 631 return (0); 632 } 633 if ((com == FIONWRITE) || (com == FIONSPACE)) { 634 /* 635 * Files don't have send queues, so there never 636 * are any bytes in them, nor is there any 637 * open space in them. 638 */ 639 *(int *)data = 0; 640 return (0); 641 } 642 if (com == FIOGETBMAP) { 643 daddr_t *block; 644 645 if (*(daddr_t *)data < 0) 646 return (EINVAL); 647 block = (daddr_t *)data; 648 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 649 } 650 if (com == OFIOGETBMAP) { 651 daddr_t ibn, obn; 652 653 if (*(int32_t *)data < 0) 654 return (EINVAL); 655 ibn = (daddr_t)*(int32_t *)data; 656 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 657 *(int32_t *)data = (int32_t)obn; 658 return error; 659 } 660 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 661 return (0); /* XXX */ 662 /* fall into ... */ 663 case VFIFO: 664 case VCHR: 665 case VBLK: 666 error = VOP_IOCTL(vp, com, data, fp->f_flag, 667 l->l_proc->p_ucred, l); 668 if (error == 0 && com == TIOCSCTTY) { 669 if (p->p_session->s_ttyvp) 670 vrele(p->p_session->s_ttyvp); 671 p->p_session->s_ttyvp = vp; 672 VREF(vp); 673 } 674 return (error); 675 676 default: 677 return (EPASSTHROUGH); 678 } 679} 680 681/* 682 * File table vnode poll routine. 683 */ 684static int 685vn_poll(struct file *fp, int events, struct lwp *l) 686{ 687 688 return (VOP_POLL(((struct vnode *)fp->f_data), events, l)); 689} 690 691/* 692 * File table vnode kqfilter routine. 693 */ 694int 695vn_kqfilter(struct file *fp, struct knote *kn) 696{ 697 698 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 699} 700 701/* 702 * Check that the vnode is still valid, and if so 703 * acquire requested lock. 704 */ 705int 706vn_lock(struct vnode *vp, int flags) 707{ 708 int error; 709 710#if 0 711 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 712 || (vp->v_flag & VONWORKLST) != 0); 713#endif 714 KASSERT((flags & 715 ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_DRAIN|LK_NOWAIT|LK_RETRY| 716 LK_SETRECURSE|LK_CANRECURSE)) 717 == 0); 718 719 do { 720 if ((flags & LK_INTERLOCK) == 0) 721 simple_lock(&vp->v_interlock); 722 if (vp->v_flag & VXLOCK) { 723 if (flags & LK_NOWAIT) { 724 simple_unlock(&vp->v_interlock); 725 return EBUSY; 726 } 727 vp->v_flag |= VXWANT; 728 ltsleep(vp, PINOD | PNORELOCK, 729 "vn_lock", 0, &vp->v_interlock); 730 error = ENOENT; 731 } else { 732 error = VOP_LOCK(vp, 733 (flags & ~LK_RETRY) | LK_INTERLOCK); 734 if (error == 0 || error == EDEADLK || error == EBUSY) 735 return (error); 736 } 737 flags &= ~LK_INTERLOCK; 738 } while (flags & LK_RETRY); 739 return (error); 740} 741 742/* 743 * File table vnode close routine. 744 */ 745static int 746vn_closefile(struct file *fp, struct lwp *l) 747{ 748 749 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 750 fp->f_cred, l)); 751} 752 753/* 754 * Enable LK_CANRECURSE on lock. Return prior status. 755 */ 756u_int 757vn_setrecurse(struct vnode *vp) 758{ 759 struct lock *lkp = &vp->v_lock; 760 u_int retval = lkp->lk_flags & LK_CANRECURSE; 761 762 lkp->lk_flags |= LK_CANRECURSE; 763 return retval; 764} 765 766/* 767 * Called when done with locksetrecurse. 768 */ 769void 770vn_restorerecurse(struct vnode *vp, u_int flags) 771{ 772 struct lock *lkp = &vp->v_lock; 773 774 lkp->lk_flags &= ~LK_CANRECURSE; 775 lkp->lk_flags |= flags; 776} 777 778int 779vn_cow_establish(struct vnode *vp, 780 int (*func)(void *, struct buf *), void *cookie) 781{ 782 int s; 783 struct spec_cow_entry *e; 784 785 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 786 M_DEVBUF, M_WAITOK); 787 e->ce_func = func; 788 e->ce_cookie = cookie; 789 790 SPEC_COW_LOCK(vp->v_specinfo, s); 791 vp->v_spec_cow_req++; 792 while (vp->v_spec_cow_count > 0) 793 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 794 &vp->v_spec_cow_slock); 795 796 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 797 798 vp->v_spec_cow_req--; 799 if (vp->v_spec_cow_req == 0) 800 wakeup(&vp->v_spec_cow_req); 801 SPEC_COW_UNLOCK(vp->v_specinfo, s); 802 803 return 0; 804} 805 806int 807vn_cow_disestablish(struct vnode *vp, 808 int (*func)(void *, struct buf *), void *cookie) 809{ 810 int s; 811 struct spec_cow_entry *e; 812 813 SPEC_COW_LOCK(vp->v_specinfo, s); 814 vp->v_spec_cow_req++; 815 while (vp->v_spec_cow_count > 0) 816 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 817 &vp->v_spec_cow_slock); 818 819 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 820 if (e->ce_func == func && e->ce_cookie == cookie) { 821 SLIST_REMOVE(&vp->v_spec_cow_head, e, 822 spec_cow_entry, ce_list); 823 FREE(e, M_DEVBUF); 824 break; 825 } 826 827 vp->v_spec_cow_req--; 828 if (vp->v_spec_cow_req == 0) 829 wakeup(&vp->v_spec_cow_req); 830 SPEC_COW_UNLOCK(vp->v_specinfo, s); 831 832 return e ? 0 : EINVAL; 833} 834 835/* 836 * Simplified in-kernel wrapper calls for extended attribute access. 837 * Both calls pass in a NULL credential, authorizing a "kernel" access. 838 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 839 */ 840int 841vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 842 const char *attrname, size_t *buflen, void *bf, struct lwp *l) 843{ 844 struct uio auio; 845 struct iovec aiov; 846 int error; 847 848 aiov.iov_len = *buflen; 849 aiov.iov_base = bf; 850 851 auio.uio_iov = &aiov; 852 auio.uio_iovcnt = 1; 853 auio.uio_rw = UIO_READ; 854 auio.uio_offset = 0; 855 auio.uio_resid = *buflen; 856 UIO_SETUP_SYSSPACE(&auio); 857 858 if ((ioflg & IO_NODELOCKED) == 0) 859 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 860 861 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 862 l); 863 864 if ((ioflg & IO_NODELOCKED) == 0) 865 VOP_UNLOCK(vp, 0); 866 867 if (error == 0) 868 *buflen = *buflen - auio.uio_resid; 869 870 return (error); 871} 872 873/* 874 * XXX Failure mode if partially written? 875 */ 876int 877vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 878 const char *attrname, size_t buflen, const void *bf, struct lwp *l) 879{ 880 struct uio auio; 881 struct iovec aiov; 882 struct mount *mp = NULL; /* XXX: GCC */ 883 int error; 884 885 aiov.iov_len = buflen; 886 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 887 888 auio.uio_iov = &aiov; 889 auio.uio_iovcnt = 1; 890 auio.uio_rw = UIO_WRITE; 891 auio.uio_offset = 0; 892 auio.uio_resid = buflen; 893 UIO_SETUP_SYSSPACE(&auio); 894 895 if ((ioflg & IO_NODELOCKED) == 0) { 896 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 897 return (error); 898 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 899 } 900 901 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, l); 902 903 if ((ioflg & IO_NODELOCKED) == 0) { 904 vn_finished_write(mp, 0); 905 VOP_UNLOCK(vp, 0); 906 } 907 908 return (error); 909} 910 911int 912vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 913 const char *attrname, struct lwp *l) 914{ 915 struct mount *mp = NULL; /* XXX: GCC */ 916 int error; 917 918 if ((ioflg & IO_NODELOCKED) == 0) { 919 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 920 return (error); 921 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 922 } 923 924 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, l); 925 if (error == EOPNOTSUPP) 926 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 927 NULL, l); 928 929 if ((ioflg & IO_NODELOCKED) == 0) { 930 vn_finished_write(mp, 0); 931 VOP_UNLOCK(vp, 0); 932 } 933 934 return (error); 935} 936 937/* 938 * Preparing to start a filesystem write operation. If the operation is 939 * permitted, then we bump the count of operations in progress and 940 * proceed. If a suspend request is in progress, we wait until the 941 * suspension is over, and then proceed. 942 * V_PCATCH adds PCATCH to the tsleep flags. 943 * V_WAIT waits until suspension is over. Otherwise returns EWOULDBLOCK. 944 * V_SLEEPONLY wait, but do not bump the operations count. 945 * V_LOWER this is a lower level operation. No further vnodes should be 946 * locked. Otherwise it is a upper level operation. No vnodes 947 * should be locked. 948 */ 949int 950vn_start_write(struct vnode *vp, struct mount **mpp, int flags) 951{ 952 struct mount *mp; 953 int error, mask, prio; 954 955 /* 956 * If a vnode is provided, get and return the mount point that 957 * to which it will write. 958 */ 959 if (vp != NULL) { 960 *mpp = vp->v_mount; 961 } 962 if ((mp = *mpp) == NULL) 963 return (0); 964 mp = mp->mnt_leaf; 965 /* 966 * Check on status of suspension. 967 */ 968 prio = PUSER - 1; 969 if (flags & V_PCATCH) 970 prio |= PCATCH; 971 972 if ((flags & V_LOWER) == 0) 973 mask = IMNT_SUSPEND; 974 else 975 mask = IMNT_SUSPENDLOW; 976 977 while ((mp->mnt_iflag & mask) != 0) { 978 if ((flags & V_WAIT) == 0) 979 return (EWOULDBLOCK); 980 error = tsleep(&mp->mnt_flag, prio, "suspfs", 0); 981 if (error) 982 return (error); 983 } 984 if (flags & V_SLEEPONLY) 985 return (0); 986 simple_lock(&mp->mnt_slock); 987 if ((flags & V_LOWER) == 0) 988 mp->mnt_writeopcountupper++; 989 else 990 mp->mnt_writeopcountlower++; 991 simple_unlock(&mp->mnt_slock); 992 return (0); 993} 994 995/* 996 * Filesystem write operation has completed. If we are suspending and this 997 * operation is the last one, notify the suspender that the suspension is 998 * now in effect. 999 */ 1000void 1001vn_finished_write(struct mount *mp, int flags) 1002{ 1003 if (mp == NULL) 1004 return; 1005 mp = mp->mnt_leaf; 1006 simple_lock(&mp->mnt_slock); 1007 if ((flags & V_LOWER) == 0) { 1008 mp->mnt_writeopcountupper--; 1009 if (mp->mnt_writeopcountupper < 0) 1010 printf("vn_finished_write: neg cnt upper=%d\n", 1011 mp->mnt_writeopcountupper); 1012 if ((mp->mnt_iflag & IMNT_SUSPEND) != 0 && 1013 mp->mnt_writeopcountupper <= 0) 1014 wakeup(&mp->mnt_writeopcountupper); 1015 } else { 1016 mp->mnt_writeopcountlower--; 1017 if (mp->mnt_writeopcountlower < 0) 1018 printf("vn_finished_write: neg cnt lower=%d\n", 1019 mp->mnt_writeopcountlower); 1020 if ((mp->mnt_iflag & IMNT_SUSPENDLOW) != 0 && 1021 mp->mnt_writeopcountupper <= 0) 1022 wakeup(&mp->mnt_writeopcountlower); 1023 } 1024 simple_unlock(&mp->mnt_slock); 1025} 1026 1027void 1028vn_ra_allocctx(struct vnode *vp) 1029{ 1030 struct uvm_ractx *ra = NULL; 1031 1032 if (vp->v_type != VREG) { 1033 return; 1034 } 1035 if (vp->v_ractx != NULL) { 1036 return; 1037 } 1038 simple_lock(&vp->v_interlock); 1039 if (vp->v_ractx == NULL) { 1040 simple_unlock(&vp->v_interlock); 1041 ra = uvm_ra_allocctx(); 1042 simple_lock(&vp->v_interlock); 1043 if (ra != NULL && vp->v_ractx == NULL) { 1044 vp->v_ractx = ra; 1045 ra = NULL; 1046 } 1047 } 1048 simple_unlock(&vp->v_interlock); 1049 if (ra != NULL) { 1050 uvm_ra_freectx(ra); 1051 } 1052} 1053