vfs_vnops.c revision 1.106.4.4
1 /* $NetBSD: vfs_vnops.c,v 1.106.4.4 2006/05/11 23:30:15 elad 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 1.106.4.4 2006/05/11 23:30:15 elad 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 #include <sys/kauth.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 #ifdef VERIFIED_EXEC 76 #include <sys/verified_exec.h> 77 #endif 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 mount *mp; 103 struct lwp *l = ndp->ni_cnd.cn_lwp; 104 kauth_cred_t cred = l->l_proc->p_cred; 105 struct vattr va; 106 int error; 107 #ifdef VERIFIED_EXEC 108 struct veriexec_hash_entry *vhe = NULL; 109 char pathbuf[MAXPATHLEN]; 110 size_t pathlen; 111 int (*copyfun)(const void *, void *, size_t, size_t *) = 112 ndp->ni_segflg == UIO_SYSSPACE ? copystr : copyinstr; 113 #endif /* VERIFIED_EXEC */ 114 115 #ifdef VERIFIED_EXEC 116 error = (*copyfun)(ndp->ni_dirp, pathbuf, sizeof(pathbuf), &pathlen); 117 if (error) { 118 if (veriexec_verbose >= 1) 119 printf("veriexec: Can't copy path. (error=%d)\n", 120 error); 121 122 return (error); 123 } 124 #endif /* VERIFIED_EXEC */ 125 126 restart: 127 if (fmode & O_CREAT) { 128 ndp->ni_cnd.cn_nameiop = CREATE; 129 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 130 if ((fmode & O_EXCL) == 0 && 131 ((fmode & O_NOFOLLOW) == 0)) 132 ndp->ni_cnd.cn_flags |= FOLLOW; 133 if ((error = namei(ndp)) != 0) 134 return (error); 135 if (ndp->ni_vp == NULL) { 136 #ifdef VERIFIED_EXEC 137 /* Lockdown mode: Prevent creation of new files. */ 138 if (veriexec_strict >= 3) { 139 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 140 141 printf("Veriexec: vn_open: Preventing " 142 "new file creation in %s.\n", 143 pathbuf); 144 145 vp = ndp->ni_dvp; 146 error = EPERM; 147 goto bad; 148 } 149 #endif /* VERIFIED_EXEC */ 150 151 VATTR_NULL(&va); 152 va.va_type = VREG; 153 va.va_mode = cmode; 154 if (fmode & O_EXCL) 155 va.va_vaflags |= VA_EXCLUSIVE; 156 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 157 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 158 vput(ndp->ni_dvp); 159 if ((error = vn_start_write(NULL, &mp, 160 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 161 return (error); 162 goto restart; 163 } 164 VOP_LEASE(ndp->ni_dvp, l, cred, LEASE_WRITE); 165 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 166 &ndp->ni_cnd, &va); 167 vn_finished_write(mp, 0); 168 if (error) 169 return (error); 170 fmode &= ~O_TRUNC; 171 vp = ndp->ni_vp; 172 } else { 173 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 174 if (ndp->ni_dvp == ndp->ni_vp) 175 vrele(ndp->ni_dvp); 176 else 177 vput(ndp->ni_dvp); 178 ndp->ni_dvp = NULL; 179 vp = ndp->ni_vp; 180 if (fmode & O_EXCL) { 181 error = EEXIST; 182 goto bad; 183 } 184 fmode &= ~O_CREAT; 185 } 186 } else { 187 ndp->ni_cnd.cn_nameiop = LOOKUP; 188 ndp->ni_cnd.cn_flags = LOCKLEAF; 189 if ((fmode & O_NOFOLLOW) == 0) 190 ndp->ni_cnd.cn_flags |= FOLLOW; 191 if ((error = namei(ndp)) != 0) 192 return (error); 193 vp = ndp->ni_vp; 194 } 195 if (vp->v_type == VSOCK) { 196 error = EOPNOTSUPP; 197 goto bad; 198 } 199 if (ndp->ni_vp->v_type == VLNK) { 200 error = EFTYPE; 201 goto bad; 202 } 203 204 #ifdef VERIFIED_EXEC 205 if ((error = VOP_GETATTR(vp, &va, cred, l)) != 0) 206 goto bad; 207 #endif 208 209 if ((fmode & O_CREAT) == 0) { 210 #ifdef VERIFIED_EXEC 211 if ((error = veriexec_verify(l, vp, &va, pathbuf, 212 VERIEXEC_FILE, &vhe)) != 0) 213 goto bad; 214 #endif 215 216 if (fmode & FREAD) { 217 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 218 goto bad; 219 } 220 221 if (fmode & (FWRITE | O_TRUNC)) { 222 if (vp->v_type == VDIR) { 223 error = EISDIR; 224 goto bad; 225 } 226 if ((error = vn_writechk(vp)) != 0 || 227 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 228 goto bad; 229 #ifdef VERIFIED_EXEC 230 if (vhe != NULL) { 231 veriexec_report("Write access request.", 232 pathbuf, &va, l, 233 REPORT_NOVERBOSE, 234 REPORT_ALARM, 235 REPORT_NOPANIC); 236 237 /* IPS mode: Deny writing to monitored files. */ 238 if (veriexec_strict >= 2) { 239 error = EPERM; 240 goto bad; 241 } else { 242 vhe->status = FINGERPRINT_NOTEVAL; 243 } 244 } 245 #endif 246 } 247 } 248 249 if (fmode & O_TRUNC) { 250 VOP_UNLOCK(vp, 0); /* XXX */ 251 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 252 vrele(vp); 253 return (error); 254 } 255 VOP_LEASE(vp, l, cred, LEASE_WRITE); 256 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 257 VATTR_NULL(&va); 258 va.va_size = 0; 259 error = VOP_SETATTR(vp, &va, cred, l); 260 vn_finished_write(mp, 0); 261 if (error != 0) 262 goto bad; 263 } 264 if ((error = VOP_OPEN(vp, fmode, cred, l)) != 0) 265 goto bad; 266 if (vp->v_type == VREG && 267 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 268 error = EIO; 269 goto bad; 270 } 271 if (fmode & FWRITE) 272 vp->v_writecount++; 273 274 return (0); 275 bad: 276 vput(vp); 277 return (error); 278 } 279 280 /* 281 * Check for write permissions on the specified vnode. 282 * Prototype text segments cannot be written. 283 */ 284 int 285 vn_writechk(struct vnode *vp) 286 { 287 288 /* 289 * If the vnode is in use as a process's text, 290 * we can't allow writing. 291 */ 292 if (vp->v_flag & VTEXT) 293 return (ETXTBSY); 294 return (0); 295 } 296 297 /* 298 * Mark a vnode as having executable mappings. 299 */ 300 void 301 vn_markexec(struct vnode *vp) 302 { 303 if ((vp->v_flag & VEXECMAP) == 0) { 304 uvmexp.filepages -= vp->v_uobj.uo_npages; 305 uvmexp.execpages += vp->v_uobj.uo_npages; 306 } 307 vp->v_flag |= VEXECMAP; 308 } 309 310 /* 311 * Mark a vnode as being the text of a process. 312 * Fail if the vnode is currently writable. 313 */ 314 int 315 vn_marktext(struct vnode *vp) 316 { 317 318 if (vp->v_writecount != 0) { 319 KASSERT((vp->v_flag & VTEXT) == 0); 320 return (ETXTBSY); 321 } 322 vp->v_flag |= VTEXT; 323 vn_markexec(vp); 324 return (0); 325 } 326 327 /* 328 * Vnode close call 329 * 330 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 331 */ 332 int 333 vn_close(struct vnode *vp, int flags, kauth_cred_t cred, struct lwp *l) 334 { 335 int error; 336 337 if (flags & FWRITE) 338 vp->v_writecount--; 339 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 340 error = VOP_CLOSE(vp, flags, cred, l); 341 vput(vp); 342 return (error); 343 } 344 345 /* 346 * Package up an I/O request on a vnode into a uio and do it. 347 */ 348 int 349 vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len, off_t offset, 350 enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid, 351 struct lwp *l) 352 { 353 struct uio auio; 354 struct iovec aiov; 355 struct mount *mp = NULL; 356 int error; 357 358 if ((ioflg & IO_NODELOCKED) == 0) { 359 if (rw == UIO_READ) { 360 vn_lock(vp, LK_SHARED | LK_RETRY); 361 } else /* UIO_WRITE */ { 362 if (vp->v_type != VCHR && 363 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 364 != 0) 365 return (error); 366 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 367 } 368 } 369 auio.uio_iov = &aiov; 370 auio.uio_iovcnt = 1; 371 aiov.iov_base = base; 372 aiov.iov_len = len; 373 auio.uio_resid = len; 374 auio.uio_offset = offset; 375 auio.uio_rw = rw; 376 if (segflg == UIO_SYSSPACE) { 377 UIO_SETUP_SYSSPACE(&auio); 378 } else { 379 auio.uio_vmspace = l->l_proc->p_vmspace; 380 } 381 if (rw == UIO_READ) { 382 error = VOP_READ(vp, &auio, ioflg, cred); 383 } else { 384 error = VOP_WRITE(vp, &auio, ioflg, cred); 385 } 386 if (aresid) 387 *aresid = auio.uio_resid; 388 else 389 if (auio.uio_resid && error == 0) 390 error = EIO; 391 if ((ioflg & IO_NODELOCKED) == 0) { 392 if (rw == UIO_WRITE) 393 vn_finished_write(mp, 0); 394 VOP_UNLOCK(vp, 0); 395 } 396 return (error); 397 } 398 399 int 400 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 401 struct lwp *l, off_t **cookies, int *ncookies) 402 { 403 struct vnode *vp = (struct vnode *)fp->f_data; 404 struct iovec aiov; 405 struct uio auio; 406 int error, eofflag; 407 408 unionread: 409 if (vp->v_type != VDIR) 410 return (EINVAL); 411 aiov.iov_base = bf; 412 aiov.iov_len = count; 413 auio.uio_iov = &aiov; 414 auio.uio_iovcnt = 1; 415 auio.uio_rw = UIO_READ; 416 if (segflg == UIO_SYSSPACE) { 417 UIO_SETUP_SYSSPACE(&auio); 418 } else { 419 KASSERT(l == curlwp); 420 auio.uio_vmspace = l->l_proc->p_vmspace; 421 } 422 auio.uio_resid = count; 423 vn_lock(vp, LK_SHARED | LK_RETRY); 424 auio.uio_offset = fp->f_offset; 425 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 426 ncookies); 427 fp->f_offset = auio.uio_offset; 428 VOP_UNLOCK(vp, 0); 429 if (error) 430 return (error); 431 432 #if defined(UNION) || defined(LKM) 433 if (count == auio.uio_resid && vn_union_readdir_hook) { 434 struct vnode *ovp = vp; 435 436 error = (*vn_union_readdir_hook)(&vp, fp, l); 437 if (error) 438 return (error); 439 if (vp != ovp) 440 goto unionread; 441 } 442 #endif /* UNION || LKM */ 443 444 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 445 (vp->v_mount->mnt_flag & MNT_UNION)) { 446 struct vnode *tvp = vp; 447 vp = vp->v_mount->mnt_vnodecovered; 448 VREF(vp); 449 fp->f_data = vp; 450 fp->f_offset = 0; 451 vrele(tvp); 452 goto unionread; 453 } 454 *done = count - auio.uio_resid; 455 return error; 456 } 457 458 /* 459 * File table vnode read routine. 460 */ 461 static int 462 vn_read(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 463 int flags) 464 { 465 struct vnode *vp = (struct vnode *)fp->f_data; 466 int count, error, ioflag; 467 struct lwp *l = curlwp; 468 469 VOP_LEASE(vp, l, cred, LEASE_READ); 470 ioflag = IO_ADV_ENCODE(fp->f_advice); 471 if (fp->f_flag & FNONBLOCK) 472 ioflag |= IO_NDELAY; 473 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 474 ioflag |= IO_SYNC; 475 if (fp->f_flag & FALTIO) 476 ioflag |= IO_ALTSEMANTICS; 477 vn_lock(vp, LK_SHARED | LK_RETRY); 478 uio->uio_offset = *offset; 479 count = uio->uio_resid; 480 error = VOP_READ(vp, uio, ioflag, cred); 481 if (flags & FOF_UPDATE_OFFSET) 482 *offset += count - uio->uio_resid; 483 VOP_UNLOCK(vp, 0); 484 return (error); 485 } 486 487 /* 488 * File table vnode write routine. 489 */ 490 static int 491 vn_write(struct file *fp, off_t *offset, struct uio *uio, kauth_cred_t cred, 492 int flags) 493 { 494 struct vnode *vp = (struct vnode *)fp->f_data; 495 struct mount *mp; 496 int count, error, ioflag = IO_UNIT; 497 struct lwp *l = curlwp; 498 499 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 500 ioflag |= IO_APPEND; 501 if (fp->f_flag & FNONBLOCK) 502 ioflag |= IO_NDELAY; 503 if (fp->f_flag & FFSYNC || 504 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 505 ioflag |= IO_SYNC; 506 else if (fp->f_flag & FDSYNC) 507 ioflag |= IO_DSYNC; 508 if (fp->f_flag & FALTIO) 509 ioflag |= IO_ALTSEMANTICS; 510 mp = NULL; 511 if (vp->v_type != VCHR && 512 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 513 return (error); 514 VOP_LEASE(vp, l, cred, LEASE_WRITE); 515 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 516 uio->uio_offset = *offset; 517 count = uio->uio_resid; 518 error = VOP_WRITE(vp, uio, ioflag, cred); 519 if (flags & FOF_UPDATE_OFFSET) { 520 if (ioflag & IO_APPEND) 521 *offset = uio->uio_offset; 522 else 523 *offset += count - uio->uio_resid; 524 } 525 VOP_UNLOCK(vp, 0); 526 vn_finished_write(mp, 0); 527 return (error); 528 } 529 530 /* 531 * File table vnode stat routine. 532 */ 533 static int 534 vn_statfile(struct file *fp, struct stat *sb, struct lwp *l) 535 { 536 struct vnode *vp = (struct vnode *)fp->f_data; 537 538 return vn_stat(vp, sb, l); 539 } 540 541 int 542 vn_stat(struct vnode *vp, struct stat *sb, struct lwp *l) 543 { 544 struct vattr va; 545 int error; 546 mode_t mode; 547 548 error = VOP_GETATTR(vp, &va, l->l_proc->p_cred, l); 549 if (error) 550 return (error); 551 /* 552 * Copy from vattr table 553 */ 554 sb->st_dev = va.va_fsid; 555 sb->st_ino = va.va_fileid; 556 mode = va.va_mode; 557 switch (vp->v_type) { 558 case VREG: 559 mode |= S_IFREG; 560 break; 561 case VDIR: 562 mode |= S_IFDIR; 563 break; 564 case VBLK: 565 mode |= S_IFBLK; 566 break; 567 case VCHR: 568 mode |= S_IFCHR; 569 break; 570 case VLNK: 571 mode |= S_IFLNK; 572 break; 573 case VSOCK: 574 mode |= S_IFSOCK; 575 break; 576 case VFIFO: 577 mode |= S_IFIFO; 578 break; 579 default: 580 return (EBADF); 581 }; 582 sb->st_mode = mode; 583 sb->st_nlink = va.va_nlink; 584 sb->st_uid = va.va_uid; 585 sb->st_gid = va.va_gid; 586 sb->st_rdev = va.va_rdev; 587 sb->st_size = va.va_size; 588 sb->st_atimespec = va.va_atime; 589 sb->st_mtimespec = va.va_mtime; 590 sb->st_ctimespec = va.va_ctime; 591 sb->st_birthtimespec = va.va_birthtime; 592 sb->st_blksize = va.va_blocksize; 593 sb->st_flags = va.va_flags; 594 sb->st_gen = 0; 595 sb->st_blocks = va.va_bytes / S_BLKSIZE; 596 return (0); 597 } 598 599 /* 600 * File table vnode fcntl routine. 601 */ 602 static int 603 vn_fcntl(struct file *fp, u_int com, void *data, struct lwp *l) 604 { 605 struct vnode *vp = ((struct vnode *)fp->f_data); 606 int error; 607 608 error = VOP_FCNTL(vp, com, data, fp->f_flag, l->l_proc->p_cred, l); 609 return (error); 610 } 611 612 /* 613 * File table vnode ioctl routine. 614 */ 615 static int 616 vn_ioctl(struct file *fp, u_long com, void *data, struct lwp *l) 617 { 618 struct vnode *vp = ((struct vnode *)fp->f_data); 619 struct proc *p = l->l_proc; 620 struct vattr vattr; 621 int error; 622 623 switch (vp->v_type) { 624 625 case VREG: 626 case VDIR: 627 if (com == FIONREAD) { 628 error = VOP_GETATTR(vp, &vattr, l->l_proc->p_cred, l); 629 if (error) 630 return (error); 631 *(int *)data = vattr.va_size - fp->f_offset; 632 return (0); 633 } 634 if ((com == FIONWRITE) || (com == FIONSPACE)) { 635 /* 636 * Files don't have send queues, so there never 637 * are any bytes in them, nor is there any 638 * open space in them. 639 */ 640 *(int *)data = 0; 641 return (0); 642 } 643 if (com == FIOGETBMAP) { 644 daddr_t *block; 645 646 if (*(daddr_t *)data < 0) 647 return (EINVAL); 648 block = (daddr_t *)data; 649 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 650 } 651 if (com == OFIOGETBMAP) { 652 daddr_t ibn, obn; 653 654 if (*(int32_t *)data < 0) 655 return (EINVAL); 656 ibn = (daddr_t)*(int32_t *)data; 657 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 658 *(int32_t *)data = (int32_t)obn; 659 return error; 660 } 661 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 662 return (0); /* XXX */ 663 /* fall into ... */ 664 case VFIFO: 665 case VCHR: 666 case VBLK: 667 error = VOP_IOCTL(vp, com, data, fp->f_flag, 668 l->l_proc->p_cred, l); 669 if (error == 0 && com == TIOCSCTTY) { 670 if (p->p_session->s_ttyvp) 671 vrele(p->p_session->s_ttyvp); 672 p->p_session->s_ttyvp = vp; 673 VREF(vp); 674 } 675 return (error); 676 677 default: 678 return (EPASSTHROUGH); 679 } 680 } 681 682 /* 683 * File table vnode poll routine. 684 */ 685 static int 686 vn_poll(struct file *fp, int events, struct lwp *l) 687 { 688 689 return (VOP_POLL(((struct vnode *)fp->f_data), events, l)); 690 } 691 692 /* 693 * File table vnode kqfilter routine. 694 */ 695 int 696 vn_kqfilter(struct file *fp, struct knote *kn) 697 { 698 699 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 700 } 701 702 /* 703 * Check that the vnode is still valid, and if so 704 * acquire requested lock. 705 */ 706 int 707 vn_lock(struct vnode *vp, int flags) 708 { 709 int error; 710 711 #if 0 712 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 713 || (vp->v_flag & VONWORKLST) != 0); 714 #endif 715 KASSERT((flags & 716 ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_DRAIN|LK_NOWAIT|LK_RETRY| 717 LK_SETRECURSE|LK_CANRECURSE)) 718 == 0); 719 720 do { 721 if ((flags & LK_INTERLOCK) == 0) 722 simple_lock(&vp->v_interlock); 723 if (vp->v_flag & VXLOCK) { 724 if (flags & LK_NOWAIT) { 725 simple_unlock(&vp->v_interlock); 726 return EBUSY; 727 } 728 vp->v_flag |= VXWANT; 729 ltsleep(vp, PINOD | PNORELOCK, 730 "vn_lock", 0, &vp->v_interlock); 731 error = ENOENT; 732 } else { 733 error = VOP_LOCK(vp, 734 (flags & ~LK_RETRY) | LK_INTERLOCK); 735 if (error == 0 || error == EDEADLK || error == EBUSY) 736 return (error); 737 } 738 flags &= ~LK_INTERLOCK; 739 } while (flags & LK_RETRY); 740 return (error); 741 } 742 743 /* 744 * File table vnode close routine. 745 */ 746 static int 747 vn_closefile(struct file *fp, struct lwp *l) 748 { 749 750 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 751 fp->f_cred, l)); 752 } 753 754 /* 755 * Enable LK_CANRECURSE on lock. Return prior status. 756 */ 757 u_int 758 vn_setrecurse(struct vnode *vp) 759 { 760 struct lock *lkp = &vp->v_lock; 761 u_int retval = lkp->lk_flags & LK_CANRECURSE; 762 763 lkp->lk_flags |= LK_CANRECURSE; 764 return retval; 765 } 766 767 /* 768 * Called when done with locksetrecurse. 769 */ 770 void 771 vn_restorerecurse(struct vnode *vp, u_int flags) 772 { 773 struct lock *lkp = &vp->v_lock; 774 775 lkp->lk_flags &= ~LK_CANRECURSE; 776 lkp->lk_flags |= flags; 777 } 778 779 int 780 vn_cow_establish(struct vnode *vp, 781 int (*func)(void *, struct buf *), void *cookie) 782 { 783 int s; 784 struct spec_cow_entry *e; 785 786 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 787 M_DEVBUF, M_WAITOK); 788 e->ce_func = func; 789 e->ce_cookie = cookie; 790 791 SPEC_COW_LOCK(vp->v_specinfo, s); 792 vp->v_spec_cow_req++; 793 while (vp->v_spec_cow_count > 0) 794 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 795 &vp->v_spec_cow_slock); 796 797 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 798 799 vp->v_spec_cow_req--; 800 if (vp->v_spec_cow_req == 0) 801 wakeup(&vp->v_spec_cow_req); 802 SPEC_COW_UNLOCK(vp->v_specinfo, s); 803 804 return 0; 805 } 806 807 int 808 vn_cow_disestablish(struct vnode *vp, 809 int (*func)(void *, struct buf *), void *cookie) 810 { 811 int s; 812 struct spec_cow_entry *e; 813 814 SPEC_COW_LOCK(vp->v_specinfo, s); 815 vp->v_spec_cow_req++; 816 while (vp->v_spec_cow_count > 0) 817 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 818 &vp->v_spec_cow_slock); 819 820 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 821 if (e->ce_func == func && e->ce_cookie == cookie) { 822 SLIST_REMOVE(&vp->v_spec_cow_head, e, 823 spec_cow_entry, ce_list); 824 FREE(e, M_DEVBUF); 825 break; 826 } 827 828 vp->v_spec_cow_req--; 829 if (vp->v_spec_cow_req == 0) 830 wakeup(&vp->v_spec_cow_req); 831 SPEC_COW_UNLOCK(vp->v_specinfo, s); 832 833 return e ? 0 : EINVAL; 834 } 835 836 /* 837 * Simplified in-kernel wrapper calls for extended attribute access. 838 * Both calls pass in a NULL credential, authorizing a "kernel" access. 839 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 840 */ 841 int 842 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 843 const char *attrname, size_t *buflen, void *bf, struct lwp *l) 844 { 845 struct uio auio; 846 struct iovec aiov; 847 int error; 848 849 aiov.iov_len = *buflen; 850 aiov.iov_base = bf; 851 852 auio.uio_iov = &aiov; 853 auio.uio_iovcnt = 1; 854 auio.uio_rw = UIO_READ; 855 auio.uio_offset = 0; 856 auio.uio_resid = *buflen; 857 UIO_SETUP_SYSSPACE(&auio); 858 859 if ((ioflg & IO_NODELOCKED) == 0) 860 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 861 862 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 863 l); 864 865 if ((ioflg & IO_NODELOCKED) == 0) 866 VOP_UNLOCK(vp, 0); 867 868 if (error == 0) 869 *buflen = *buflen - auio.uio_resid; 870 871 return (error); 872 } 873 874 /* 875 * XXX Failure mode if partially written? 876 */ 877 int 878 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 879 const char *attrname, size_t buflen, const void *bf, struct lwp *l) 880 { 881 struct uio auio; 882 struct iovec aiov; 883 struct mount *mp; 884 int error; 885 886 aiov.iov_len = buflen; 887 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 888 889 auio.uio_iov = &aiov; 890 auio.uio_iovcnt = 1; 891 auio.uio_rw = UIO_WRITE; 892 auio.uio_offset = 0; 893 auio.uio_resid = buflen; 894 UIO_SETUP_SYSSPACE(&auio); 895 896 if ((ioflg & IO_NODELOCKED) == 0) { 897 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 898 return (error); 899 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 900 } 901 902 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, l); 903 904 if ((ioflg & IO_NODELOCKED) == 0) { 905 vn_finished_write(mp, 0); 906 VOP_UNLOCK(vp, 0); 907 } 908 909 return (error); 910 } 911 912 int 913 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 914 const char *attrname, struct lwp *l) 915 { 916 struct mount *mp; 917 int error; 918 919 if ((ioflg & IO_NODELOCKED) == 0) { 920 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 921 return (error); 922 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 923 } 924 925 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, l); 926 if (error == EOPNOTSUPP) 927 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 928 NULL, l); 929 930 if ((ioflg & IO_NODELOCKED) == 0) { 931 vn_finished_write(mp, 0); 932 VOP_UNLOCK(vp, 0); 933 } 934 935 return (error); 936 } 937 938 /* 939 * Preparing to start a filesystem write operation. If the operation is 940 * permitted, then we bump the count of operations in progress and 941 * proceed. If a suspend request is in progress, we wait until the 942 * suspension is over, and then proceed. 943 * V_PCATCH adds PCATCH to the tsleep flags. 944 * V_WAIT waits until suspension is over. Otherwise returns EWOULDBLOCK. 945 * V_SLEEPONLY wait, but do not bump the operations count. 946 * V_LOWER this is a lower level operation. No further vnodes should be 947 * locked. Otherwise it is a upper level operation. No vnodes 948 * should be locked. 949 */ 950 int 951 vn_start_write(struct vnode *vp, struct mount **mpp, int flags) 952 { 953 struct mount *mp; 954 int error, mask, prio; 955 956 /* 957 * If a vnode is provided, get and return the mount point that 958 * to which it will write. 959 */ 960 if (vp != NULL) { 961 *mpp = vp->v_mount; 962 } 963 if ((mp = *mpp) == NULL) 964 return (0); 965 mp = mp->mnt_leaf; 966 /* 967 * Check on status of suspension. 968 */ 969 prio = PUSER - 1; 970 if (flags & V_PCATCH) 971 prio |= PCATCH; 972 973 if ((flags & V_LOWER) == 0) 974 mask = IMNT_SUSPEND; 975 else 976 mask = IMNT_SUSPENDLOW; 977 978 while ((mp->mnt_iflag & mask) != 0) { 979 if ((flags & V_WAIT) == 0) 980 return (EWOULDBLOCK); 981 error = tsleep(&mp->mnt_flag, prio, "suspfs", 0); 982 if (error) 983 return (error); 984 } 985 if (flags & V_SLEEPONLY) 986 return (0); 987 simple_lock(&mp->mnt_slock); 988 if ((flags & V_LOWER) == 0) 989 mp->mnt_writeopcountupper++; 990 else 991 mp->mnt_writeopcountlower++; 992 simple_unlock(&mp->mnt_slock); 993 return (0); 994 } 995 996 /* 997 * Filesystem write operation has completed. If we are suspending and this 998 * operation is the last one, notify the suspender that the suspension is 999 * now in effect. 1000 */ 1001 void 1002 vn_finished_write(struct mount *mp, int flags) 1003 { 1004 if (mp == NULL) 1005 return; 1006 mp = mp->mnt_leaf; 1007 simple_lock(&mp->mnt_slock); 1008 if ((flags & V_LOWER) == 0) { 1009 mp->mnt_writeopcountupper--; 1010 if (mp->mnt_writeopcountupper < 0) 1011 printf("vn_finished_write: neg cnt upper=%d\n", 1012 mp->mnt_writeopcountupper); 1013 if ((mp->mnt_iflag & IMNT_SUSPEND) != 0 && 1014 mp->mnt_writeopcountupper <= 0) 1015 wakeup(&mp->mnt_writeopcountupper); 1016 } else { 1017 mp->mnt_writeopcountlower--; 1018 if (mp->mnt_writeopcountlower < 0) 1019 printf("vn_finished_write: neg cnt lower=%d\n", 1020 mp->mnt_writeopcountlower); 1021 if ((mp->mnt_iflag & IMNT_SUSPENDLOW) != 0 && 1022 mp->mnt_writeopcountupper <= 0) 1023 wakeup(&mp->mnt_writeopcountlower); 1024 } 1025 simple_unlock(&mp->mnt_slock); 1026 } 1027 1028 void 1029 vn_ra_allocctx(struct vnode *vp) 1030 { 1031 struct uvm_ractx *ra = NULL; 1032 1033 if (vp->v_type != VREG) { 1034 return; 1035 } 1036 if (vp->v_ractx != NULL) { 1037 return; 1038 } 1039 simple_lock(&vp->v_interlock); 1040 if (vp->v_ractx == NULL) { 1041 simple_unlock(&vp->v_interlock); 1042 ra = uvm_ra_allocctx(); 1043 simple_lock(&vp->v_interlock); 1044 if (ra != NULL && vp->v_ractx == NULL) { 1045 vp->v_ractx = ra; 1046 ra = NULL; 1047 } 1048 } 1049 simple_unlock(&vp->v_interlock); 1050 if (ra != NULL) { 1051 uvm_ra_freectx(ra); 1052 } 1053 } 1054
Indexes created Sat Sep 20 22:09:52 GMT 2025