vfs_vnops.c revision 74437
1/* 2 * Copyright (c) 1982, 1986, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)vfs_vnops.c 8.2 (Berkeley) 1/21/94 39 * $FreeBSD: head/sys/kern/vfs_vnops.c 74437 2001-03-19 05:44:15Z rwatson $ 40 */ 41 42#include <sys/param.h> 43#include <sys/systm.h> 44#include <sys/fcntl.h> 45#include <sys/file.h> 46#include <sys/stat.h> 47#include <sys/proc.h> 48#include <sys/mount.h> 49#include <sys/mutex.h> 50#include <sys/namei.h> 51#include <sys/vnode.h> 52#include <sys/bio.h> 53#include <sys/buf.h> 54#include <sys/filio.h> 55#include <sys/ttycom.h> 56#include <sys/conf.h> 57 58static int vn_closefile __P((struct file *fp, struct proc *p)); 59static int vn_ioctl __P((struct file *fp, u_long com, caddr_t data, 60 struct proc *p)); 61static int vn_read __P((struct file *fp, struct uio *uio, 62 struct ucred *cred, int flags, struct proc *p)); 63static int vn_poll __P((struct file *fp, int events, struct ucred *cred, 64 struct proc *p)); 65static int vn_kqfilter __P((struct file *fp, struct knote *kn)); 66static int vn_statfile __P((struct file *fp, struct stat *sb, struct proc *p)); 67static int vn_write __P((struct file *fp, struct uio *uio, 68 struct ucred *cred, int flags, struct proc *p)); 69 70struct fileops vnops = { 71 vn_read, vn_write, vn_ioctl, vn_poll, vn_kqfilter, 72 vn_statfile, vn_closefile 73}; 74 75/* 76 * Common code for vnode open operations. 77 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 78 * 79 * Note that this does NOT free nameidata for the successful case, 80 * due to the NDINIT being done elsewhere. 81 */ 82int 83vn_open(ndp, flagp, cmode) 84 register struct nameidata *ndp; 85 int *flagp, cmode; 86{ 87 struct vnode *vp; 88 struct mount *mp; 89 struct proc *p = ndp->ni_cnd.cn_proc; 90 struct ucred *cred = p->p_ucred; 91 struct vattr vat; 92 struct vattr *vap = &vat; 93 int mode, fmode, error; 94 95restart: 96 fmode = *flagp; 97 if (fmode & O_CREAT) { 98 ndp->ni_cnd.cn_nameiop = CREATE; 99 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 100 if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0) 101 ndp->ni_cnd.cn_flags |= FOLLOW; 102 bwillwrite(); 103 if ((error = namei(ndp)) != 0) 104 return (error); 105 if (ndp->ni_vp == NULL) { 106 VATTR_NULL(vap); 107 vap->va_type = VREG; 108 vap->va_mode = cmode; 109 if (fmode & O_EXCL) 110 vap->va_vaflags |= VA_EXCLUSIVE; 111 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 112 NDFREE(ndp, NDF_ONLY_PNBUF); 113 vput(ndp->ni_dvp); 114 if ((error = vn_start_write(NULL, &mp, 115 V_XSLEEP | PCATCH)) != 0) 116 return (error); 117 goto restart; 118 } 119 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 120 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 121 &ndp->ni_cnd, vap); 122 vput(ndp->ni_dvp); 123 vn_finished_write(mp); 124 if (error) { 125 NDFREE(ndp, NDF_ONLY_PNBUF); 126 return (error); 127 } 128 ASSERT_VOP_UNLOCKED(ndp->ni_dvp, "create"); 129 ASSERT_VOP_LOCKED(ndp->ni_vp, "create"); 130 fmode &= ~O_TRUNC; 131 vp = ndp->ni_vp; 132 } else { 133 if (ndp->ni_dvp == ndp->ni_vp) 134 vrele(ndp->ni_dvp); 135 else 136 vput(ndp->ni_dvp); 137 ndp->ni_dvp = NULL; 138 vp = ndp->ni_vp; 139 if (fmode & O_EXCL) { 140 error = EEXIST; 141 goto bad; 142 } 143 fmode &= ~O_CREAT; 144 } 145 } else { 146 ndp->ni_cnd.cn_nameiop = LOOKUP; 147 ndp->ni_cnd.cn_flags = 148 ((fmode & O_NOFOLLOW) ? NOFOLLOW : FOLLOW) | LOCKLEAF; 149 if ((error = namei(ndp)) != 0) 150 return (error); 151 vp = ndp->ni_vp; 152 } 153 if (vp->v_type == VLNK) { 154 error = EMLINK; 155 goto bad; 156 } 157 if (vp->v_type == VSOCK) { 158 error = EOPNOTSUPP; 159 goto bad; 160 } 161 if ((fmode & O_CREAT) == 0) { 162 mode = 0; 163 if (fmode & (FWRITE | O_TRUNC)) { 164 if (vp->v_type == VDIR) { 165 error = EISDIR; 166 goto bad; 167 } 168 error = vn_writechk(vp); 169 if (error) 170 goto bad; 171 mode |= VWRITE; 172 } 173 if (fmode & FREAD) 174 mode |= VREAD; 175 if (mode) { 176 error = VOP_ACCESS(vp, mode, cred, p); 177 if (error) 178 goto bad; 179 } 180 } 181 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 182 goto bad; 183 /* 184 * Make sure that a VM object is created for VMIO support. 185 */ 186 if (vn_canvmio(vp) == TRUE) { 187 if ((error = vfs_object_create(vp, p, cred)) != 0) 188 goto bad; 189 } 190 191 if (fmode & FWRITE) 192 vp->v_writecount++; 193 *flagp = fmode; 194 return (0); 195bad: 196 NDFREE(ndp, NDF_ONLY_PNBUF); 197 vput(vp); 198 *flagp = fmode; 199 return (error); 200} 201 202/* 203 * Check for write permissions on the specified vnode. 204 * Prototype text segments cannot be written. 205 */ 206int 207vn_writechk(vp) 208 register struct vnode *vp; 209{ 210 211 /* 212 * If there's shared text associated with 213 * the vnode, try to free it up once. If 214 * we fail, we can't allow writing. 215 */ 216 if (vp->v_flag & VTEXT) 217 return (ETXTBSY); 218 return (0); 219} 220 221/* 222 * Vnode close call 223 */ 224int 225vn_close(vp, flags, cred, p) 226 register struct vnode *vp; 227 int flags; 228 struct ucred *cred; 229 struct proc *p; 230{ 231 int error; 232 233 if (flags & FWRITE) 234 vp->v_writecount--; 235 error = VOP_CLOSE(vp, flags, cred, p); 236 vrele(vp); 237 return (error); 238} 239 240static __inline 241int 242sequential_heuristic(struct uio *uio, struct file *fp) 243{ 244 /* 245 * Sequential heuristic - detect sequential operation 246 */ 247 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) || 248 uio->uio_offset == fp->f_nextoff) { 249 /* 250 * XXX we assume that the filesystem block size is 251 * the default. Not true, but still gives us a pretty 252 * good indicator of how sequential the read operations 253 * are. 254 */ 255 fp->f_seqcount += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE; 256 if (fp->f_seqcount >= 127) 257 fp->f_seqcount = 127; 258 return(fp->f_seqcount << 16); 259 } 260 261 /* 262 * Not sequential, quick draw-down of seqcount 263 */ 264 if (fp->f_seqcount > 1) 265 fp->f_seqcount = 1; 266 else 267 fp->f_seqcount = 0; 268 return(0); 269} 270 271/* 272 * Package up an I/O request on a vnode into a uio and do it. 273 */ 274int 275vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p) 276 enum uio_rw rw; 277 struct vnode *vp; 278 caddr_t base; 279 int len; 280 off_t offset; 281 enum uio_seg segflg; 282 int ioflg; 283 struct ucred *cred; 284 int *aresid; 285 struct proc *p; 286{ 287 struct uio auio; 288 struct iovec aiov; 289 struct mount *mp; 290 int error; 291 292 if ((ioflg & IO_NODELOCKED) == 0) { 293 mp = NULL; 294 if (rw == UIO_WRITE && 295 vp->v_type != VCHR && 296 (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 297 return (error); 298 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); 299 } 300 auio.uio_iov = &aiov; 301 auio.uio_iovcnt = 1; 302 aiov.iov_base = base; 303 aiov.iov_len = len; 304 auio.uio_resid = len; 305 auio.uio_offset = offset; 306 auio.uio_segflg = segflg; 307 auio.uio_rw = rw; 308 auio.uio_procp = p; 309 if (rw == UIO_READ) { 310 error = VOP_READ(vp, &auio, ioflg, cred); 311 } else { 312 error = VOP_WRITE(vp, &auio, ioflg, cred); 313 } 314 if (aresid) 315 *aresid = auio.uio_resid; 316 else 317 if (auio.uio_resid && error == 0) 318 error = EIO; 319 if ((ioflg & IO_NODELOCKED) == 0) { 320 vn_finished_write(mp); 321 VOP_UNLOCK(vp, 0, p); 322 } 323 return (error); 324} 325 326/* 327 * File table vnode read routine. 328 */ 329static int 330vn_read(fp, uio, cred, flags, p) 331 struct file *fp; 332 struct uio *uio; 333 struct ucred *cred; 334 struct proc *p; 335 int flags; 336{ 337 struct vnode *vp; 338 int error, ioflag; 339 340 KASSERT(uio->uio_procp == p, ("uio_procp %p is not p %p", 341 uio->uio_procp, p)); 342 vp = (struct vnode *)fp->f_data; 343 ioflag = 0; 344 if (fp->f_flag & FNONBLOCK) 345 ioflag |= IO_NDELAY; 346 VOP_LEASE(vp, p, cred, LEASE_READ); 347 vn_lock(vp, LK_SHARED | LK_NOPAUSE | LK_RETRY, p); 348 if ((flags & FOF_OFFSET) == 0) 349 uio->uio_offset = fp->f_offset; 350 351 ioflag |= sequential_heuristic(uio, fp); 352 353 error = VOP_READ(vp, uio, ioflag, cred); 354 if ((flags & FOF_OFFSET) == 0) 355 fp->f_offset = uio->uio_offset; 356 fp->f_nextoff = uio->uio_offset; 357 VOP_UNLOCK(vp, 0, p); 358 return (error); 359} 360 361/* 362 * File table vnode write routine. 363 */ 364static int 365vn_write(fp, uio, cred, flags, p) 366 struct file *fp; 367 struct uio *uio; 368 struct ucred *cred; 369 struct proc *p; 370 int flags; 371{ 372 struct vnode *vp; 373 struct mount *mp; 374 int error, ioflag; 375 376 KASSERT(uio->uio_procp == p, ("uio_procp %p is not p %p", 377 uio->uio_procp, p)); 378 vp = (struct vnode *)fp->f_data; 379 if (vp->v_type == VREG) 380 bwillwrite(); 381 vp = (struct vnode *)fp->f_data; /* XXX needed? */ 382 ioflag = IO_UNIT; 383 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 384 ioflag |= IO_APPEND; 385 if (fp->f_flag & FNONBLOCK) 386 ioflag |= IO_NDELAY; 387 if ((fp->f_flag & O_FSYNC) || 388 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 389 ioflag |= IO_SYNC; 390 mp = NULL; 391 if (vp->v_type != VCHR && 392 (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 393 return (error); 394 VOP_LEASE(vp, p, cred, LEASE_WRITE); 395 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); 396 if ((flags & FOF_OFFSET) == 0) 397 uio->uio_offset = fp->f_offset; 398 ioflag |= sequential_heuristic(uio, fp); 399 error = VOP_WRITE(vp, uio, ioflag, cred); 400 if ((flags & FOF_OFFSET) == 0) 401 fp->f_offset = uio->uio_offset; 402 fp->f_nextoff = uio->uio_offset; 403 VOP_UNLOCK(vp, 0, p); 404 vn_finished_write(mp); 405 return (error); 406} 407 408/* 409 * File table vnode stat routine. 410 */ 411static int 412vn_statfile(fp, sb, p) 413 struct file *fp; 414 struct stat *sb; 415 struct proc *p; 416{ 417 struct vnode *vp = (struct vnode *)fp->f_data; 418 419 return vn_stat(vp, sb, p); 420} 421 422int 423vn_stat(vp, sb, p) 424 struct vnode *vp; 425 register struct stat *sb; 426 struct proc *p; 427{ 428 struct vattr vattr; 429 register struct vattr *vap; 430 int error; 431 u_short mode; 432 433 vap = &vattr; 434 error = VOP_GETATTR(vp, vap, p->p_ucred, p); 435 if (error) 436 return (error); 437 438 /* 439 * Zero the spare stat fields 440 */ 441 sb->st_lspare = 0; 442 sb->st_qspare[0] = 0; 443 sb->st_qspare[1] = 0; 444 445 /* 446 * Copy from vattr table 447 */ 448 if (vap->va_fsid != VNOVAL) 449 sb->st_dev = vap->va_fsid; 450 else 451 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0]; 452 sb->st_ino = vap->va_fileid; 453 mode = vap->va_mode; 454 switch (vap->va_type) { 455 case VREG: 456 mode |= S_IFREG; 457 break; 458 case VDIR: 459 mode |= S_IFDIR; 460 break; 461 case VBLK: 462 mode |= S_IFBLK; 463 break; 464 case VCHR: 465 mode |= S_IFCHR; 466 break; 467 case VLNK: 468 mode |= S_IFLNK; 469 /* This is a cosmetic change, symlinks do not have a mode. */ 470 if (vp->v_mount->mnt_flag & MNT_NOSYMFOLLOW) 471 sb->st_mode &= ~ACCESSPERMS; /* 0000 */ 472 else 473 sb->st_mode |= ACCESSPERMS; /* 0777 */ 474 break; 475 case VSOCK: 476 mode |= S_IFSOCK; 477 break; 478 case VFIFO: 479 mode |= S_IFIFO; 480 break; 481 default: 482 return (EBADF); 483 }; 484 sb->st_mode = mode; 485 sb->st_nlink = vap->va_nlink; 486 sb->st_uid = vap->va_uid; 487 sb->st_gid = vap->va_gid; 488 sb->st_rdev = vap->va_rdev; 489 sb->st_size = vap->va_size; 490 sb->st_atimespec = vap->va_atime; 491 sb->st_mtimespec = vap->va_mtime; 492 sb->st_ctimespec = vap->va_ctime; 493 494 /* 495 * According to www.opengroup.org, the meaning of st_blksize is 496 * "a filesystem-specific preferred I/O block size for this 497 * object. In some filesystem types, this may vary from file 498 * to file" 499 * Default to zero to catch bogus uses of this field. 500 */ 501 502 if (vap->va_type == VREG) { 503 sb->st_blksize = vap->va_blocksize; 504 } else if (vn_isdisk(vp, NULL)) { 505 sb->st_blksize = vp->v_rdev->si_bsize_best; 506 if (sb->st_blksize < vp->v_rdev->si_bsize_phys) 507 sb->st_blksize = vp->v_rdev->si_bsize_phys; 508 if (sb->st_blksize < BLKDEV_IOSIZE) 509 sb->st_blksize = BLKDEV_IOSIZE; 510 } else { 511 sb->st_blksize = 0; 512 } 513 514 sb->st_flags = vap->va_flags; 515 if (suser_xxx(p->p_ucred, 0, 0)) 516 sb->st_gen = 0; 517 else 518 sb->st_gen = vap->va_gen; 519 520#if (S_BLKSIZE == 512) 521 /* Optimize this case */ 522 sb->st_blocks = vap->va_bytes >> 9; 523#else 524 sb->st_blocks = vap->va_bytes / S_BLKSIZE; 525#endif 526 return (0); 527} 528 529/* 530 * File table vnode ioctl routine. 531 */ 532static int 533vn_ioctl(fp, com, data, p) 534 struct file *fp; 535 u_long com; 536 caddr_t data; 537 struct proc *p; 538{ 539 register struct vnode *vp = ((struct vnode *)fp->f_data); 540 struct vattr vattr; 541 int error; 542 543 switch (vp->v_type) { 544 545 case VREG: 546 case VDIR: 547 if (com == FIONREAD) { 548 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 549 if (error) 550 return (error); 551 *(int *)data = vattr.va_size - fp->f_offset; 552 return (0); 553 } 554 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 555 return (0); /* XXX */ 556 /* fall into ... */ 557 558 default: 559#if 0 560 return (ENOTTY); 561#endif 562 case VFIFO: 563 case VCHR: 564 case VBLK: 565 if (com == FIODTYPE) { 566 if (vp->v_type != VCHR && vp->v_type != VBLK) 567 return (ENOTTY); 568 *(int *)data = devsw(vp->v_rdev)->d_flags & D_TYPEMASK; 569 return (0); 570 } 571 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 572 if (error == 0 && com == TIOCSCTTY) { 573 574 /* Do nothing if reassigning same control tty */ 575 if (p->p_session->s_ttyvp == vp) 576 return (0); 577 578 /* Get rid of reference to old control tty */ 579 if (p->p_session->s_ttyvp) 580 vrele(p->p_session->s_ttyvp); 581 582 p->p_session->s_ttyvp = vp; 583 VREF(vp); 584 } 585 return (error); 586 } 587} 588 589/* 590 * File table vnode poll routine. 591 */ 592static int 593vn_poll(fp, events, cred, p) 594 struct file *fp; 595 int events; 596 struct ucred *cred; 597 struct proc *p; 598{ 599 600 return (VOP_POLL(((struct vnode *)fp->f_data), events, cred, p)); 601} 602 603/* 604 * Check that the vnode is still valid, and if so 605 * acquire requested lock. 606 */ 607int 608#ifndef DEBUG_LOCKS 609vn_lock(vp, flags, p) 610#else 611debug_vn_lock(vp, flags, p, filename, line) 612#endif 613 struct vnode *vp; 614 int flags; 615 struct proc *p; 616#ifdef DEBUG_LOCKS 617 const char *filename; 618 int line; 619#endif 620{ 621 int error; 622 623 do { 624 if ((flags & LK_INTERLOCK) == 0) 625 mtx_lock(&vp->v_interlock); 626 if ((vp->v_flag & VXLOCK) && vp->v_vxproc != curproc) { 627 vp->v_flag |= VXWANT; 628 mtx_unlock(&vp->v_interlock); 629 tsleep((caddr_t)vp, PINOD, "vn_lock", 0); 630 error = ENOENT; 631 } else { 632 if (vp->v_vxproc != NULL) 633 printf("VXLOCK interlock avoided in vn_lock\n"); 634#ifdef DEBUG_LOCKS 635 vp->filename = filename; 636 vp->line = line; 637#endif 638 error = VOP_LOCK(vp, 639 flags | LK_NOPAUSE | LK_INTERLOCK, p); 640 if (error == 0) 641 return (error); 642 } 643 flags &= ~LK_INTERLOCK; 644 } while (flags & LK_RETRY); 645 return (error); 646} 647 648/* 649 * File table vnode close routine. 650 */ 651static int 652vn_closefile(fp, p) 653 struct file *fp; 654 struct proc *p; 655{ 656 657 fp->f_ops = &badfileops; 658 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 659 fp->f_cred, p)); 660} 661 662/* 663 * Preparing to start a filesystem write operation. If the operation is 664 * permitted, then we bump the count of operations in progress and 665 * proceed. If a suspend request is in progress, we wait until the 666 * suspension is over, and then proceed. 667 */ 668int 669vn_start_write(vp, mpp, flags) 670 struct vnode *vp; 671 struct mount **mpp; 672 int flags; 673{ 674 struct mount *mp; 675 int error; 676 677 /* 678 * If a vnode is provided, get and return the mount point that 679 * to which it will write. 680 */ 681 if (vp != NULL) { 682 if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) { 683 *mpp = NULL; 684 if (error != EOPNOTSUPP) 685 return (error); 686 return (0); 687 } 688 } 689 if ((mp = *mpp) == NULL) 690 return (0); 691 /* 692 * Check on status of suspension. 693 */ 694 while ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) { 695 if (flags & V_NOWAIT) 696 return (EWOULDBLOCK); 697 error = tsleep(&mp->mnt_flag, (PUSER - 1) | (flags & PCATCH), 698 "suspfs", 0); 699 if (error) 700 return (error); 701 } 702 if (flags & V_XSLEEP) 703 return (0); 704 mp->mnt_writeopcount++; 705 return (0); 706} 707 708/* 709 * Secondary suspension. Used by operations such as vop_inactive 710 * routines that are needed by the higher level functions. These 711 * are allowed to proceed until all the higher level functions have 712 * completed (indicated by mnt_writeopcount dropping to zero). At that 713 * time, these operations are halted until the suspension is over. 714 */ 715int 716vn_write_suspend_wait(vp, mp, flags) 717 struct vnode *vp; 718 struct mount *mp; 719 int flags; 720{ 721 int error; 722 723 if (vp != NULL) { 724 if ((error = VOP_GETWRITEMOUNT(vp, &mp)) != 0) { 725 if (error != EOPNOTSUPP) 726 return (error); 727 return (0); 728 } 729 } 730 /* 731 * If we are not suspended or have not yet reached suspended 732 * mode, then let the operation proceed. 733 */ 734 if (mp == NULL || (mp->mnt_kern_flag & MNTK_SUSPENDED) == 0) 735 return (0); 736 if (flags & V_NOWAIT) 737 return (EWOULDBLOCK); 738 /* 739 * Wait for the suspension to finish. 740 */ 741 return (tsleep(&mp->mnt_flag, (PUSER - 1) | (flags & PCATCH), 742 "suspfs", 0)); 743} 744 745/* 746 * Filesystem write operation has completed. If we are suspending and this 747 * operation is the last one, notify the suspender that the suspension is 748 * now in effect. 749 */ 750void 751vn_finished_write(mp) 752 struct mount *mp; 753{ 754 755 if (mp == NULL) 756 return; 757 mp->mnt_writeopcount--; 758 if (mp->mnt_writeopcount < 0) 759 panic("vn_finished_write: neg cnt"); 760 if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 && 761 mp->mnt_writeopcount <= 0) 762 wakeup(&mp->mnt_writeopcount); 763} 764 765/* 766 * Request a filesystem to suspend write operations. 767 */ 768void 769vfs_write_suspend(mp) 770 struct mount *mp; 771{ 772 struct proc *p = curproc; 773 774 if (mp->mnt_kern_flag & MNTK_SUSPEND) 775 return; 776 mp->mnt_kern_flag |= MNTK_SUSPEND; 777 if (mp->mnt_writeopcount > 0) 778 (void) tsleep(&mp->mnt_writeopcount, PUSER - 1, "suspwt", 0); 779 VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p); 780 mp->mnt_kern_flag |= MNTK_SUSPENDED; 781} 782 783/* 784 * Request a filesystem to resume write operations. 785 */ 786void 787vfs_write_resume(mp) 788 struct mount *mp; 789{ 790 791 if ((mp->mnt_kern_flag & MNTK_SUSPEND) == 0) 792 return; 793 mp->mnt_kern_flag &= ~(MNTK_SUSPEND | MNTK_SUSPENDED); 794 wakeup(&mp->mnt_writeopcount); 795 wakeup(&mp->mnt_flag); 796} 797 798static int 799vn_kqfilter(struct file *fp, struct knote *kn) 800{ 801 802 return (VOP_KQFILTER(((struct vnode *)fp->f_data), kn)); 803} 804 805/* 806 * Simplified in-kernel wrapper calls for extended attribute access. 807 * Both calls pass in a NULL credential, authorizing as "kernel" access. 808 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 809 */ 810int 811vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 812 const char *attrname, int *buflen, char *buf, struct proc *p) 813{ 814 struct uio auio; 815 struct iovec iov; 816 int error; 817 818 iov.iov_len = *buflen; 819 iov.iov_base = buf; 820 821 auio.uio_iov = &iov; 822 auio.uio_iovcnt = 1; 823 auio.uio_rw = UIO_READ; 824 auio.uio_segflg = UIO_SYSSPACE; 825 auio.uio_procp = p; 826 auio.uio_offset = 0; 827 auio.uio_resid = *buflen; 828 829 if ((ioflg & IO_NODELOCKED) == 0) 830 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); 831 832 /* authorize attribute retrieval as kernel */ 833 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 834 835 if ((ioflg & IO_NODELOCKED) == 0) 836 VOP_UNLOCK(vp, 0, p); 837 838 if (error == 0) { 839 *buflen = *buflen - auio.uio_resid; 840 } 841 842 return (error); 843} 844 845/* 846 * XXX failure mode if partially written? 847 */ 848int 849vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 850 const char *attrname, int buflen, char *buf, struct proc *p) 851{ 852 struct uio auio; 853 struct iovec iov; 854 struct mount *mp; 855 int error; 856 857 iov.iov_len = buflen; 858 iov.iov_base = buf; 859 860 auio.uio_iov = &iov; 861 auio.uio_iovcnt = 1; 862 auio.uio_rw = UIO_WRITE; 863 auio.uio_segflg = UIO_SYSSPACE; 864 auio.uio_procp = p; 865 auio.uio_offset = 0; 866 auio.uio_resid = buflen; 867 868 if ((ioflg & IO_NODELOCKED) == 0) { 869 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 870 return (error); 871 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); 872 } 873 874 /* authorize attribute setting as kernel */ 875 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 876 877 if ((ioflg & IO_NODELOCKED) == 0) { 878 vn_finished_write(mp); 879 VOP_UNLOCK(vp, 0, p); 880 } 881 882 return (error); 883} 884 885int 886vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 887 const char *attrname, struct proc *p) 888{ 889 struct mount *mp; 890 int error; 891 892 if ((ioflg & IO_NODELOCKED) == 0) { 893 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 894 return (error); 895 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); 896 } 897 898 /* authorize attribute removal as kernel */ 899 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL, p); 900 901 if ((ioflg & IO_NODELOCKED) == 0) { 902 vn_finished_write(mp); 903 VOP_UNLOCK(vp, 0, p); 904 } 905 906 return (error); 907} 908