Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/kern/vfs_syscalls.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
1 /*- 2 * Copyright (c) 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 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94 35 */ 36 37 #include <sys/cdefs.h> 38 __FBSDID("$FreeBSD: releng/9.1/sys/kern/vfs_syscalls.c 235449 2012-05-14 15:46:37Z jh $"); 39 40 #include "opt_capsicum.h" 41 #include "opt_compat.h" 42 #include "opt_kdtrace.h" 43 #include "opt_ktrace.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/bio.h> 48 #include <sys/buf.h> 49 #include <sys/capability.h> 50 #include <sys/disk.h> 51 #include <sys/sysent.h> 52 #include <sys/malloc.h> 53 #include <sys/mount.h> 54 #include <sys/mutex.h> 55 #include <sys/sysproto.h> 56 #include <sys/namei.h> 57 #include <sys/filedesc.h> 58 #include <sys/kernel.h> 59 #include <sys/fcntl.h> 60 #include <sys/file.h> 61 #include <sys/filio.h> 62 #include <sys/limits.h> 63 #include <sys/linker.h> 64 #include <sys/sdt.h> 65 #include <sys/stat.h> 66 #include <sys/sx.h> 67 #include <sys/unistd.h> 68 #include <sys/vnode.h> 69 #include <sys/priv.h> 70 #include <sys/proc.h> 71 #include <sys/dirent.h> 72 #include <sys/jail.h> 73 #include <sys/syscallsubr.h> 74 #include <sys/sysctl.h> 75 #ifdef KTRACE 76 #include <sys/ktrace.h> 77 #endif 78 79 #include <machine/stdarg.h> 80 81 #include <security/audit/audit.h> 82 #include <security/mac/mac_framework.h> 83 84 #include <vm/vm.h> 85 #include <vm/vm_object.h> 86 #include <vm/vm_page.h> 87 #include <vm/uma.h> 88 89 #include <ufs/ufs/quota.h> 90 91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information"); 92 93 SDT_PROVIDER_DEFINE(vfs); 94 SDT_PROBE_DEFINE(vfs, , stat, mode, mode); 95 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *"); 96 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int"); 97 SDT_PROBE_DEFINE(vfs, , stat, reg, reg); 98 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *"); 99 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int"); 100 101 static int chroot_refuse_vdir_fds(struct filedesc *fdp); 102 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *); 103 static int setfflags(struct thread *td, struct vnode *, int); 104 static int setutimes(struct thread *td, struct vnode *, 105 const struct timespec *, int, int); 106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 107 struct thread *td); 108 109 /* 110 * The module initialization routine for POSIX asynchronous I/O will 111 * set this to the version of AIO that it implements. (Zero means 112 * that it is not implemented.) This value is used here by pathconf() 113 * and in kern_descrip.c by fpathconf(). 114 */ 115 int async_io_version; 116 117 #ifdef DEBUG 118 static int syncprt = 0; 119 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, ""); 120 #endif 121 122 /* 123 * Sync each mounted filesystem. 124 */ 125 #ifndef _SYS_SYSPROTO_H_ 126 struct sync_args { 127 int dummy; 128 }; 129 #endif 130 /* ARGSUSED */ 131 int 132 sys_sync(td, uap) 133 struct thread *td; 134 struct sync_args *uap; 135 { 136 struct mount *mp, *nmp; 137 int save, vfslocked; 138 139 mtx_lock(&mountlist_mtx); 140 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 141 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) { 142 nmp = TAILQ_NEXT(mp, mnt_list); 143 continue; 144 } 145 vfslocked = VFS_LOCK_GIANT(mp); 146 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 147 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 148 save = curthread_pflags_set(TDP_SYNCIO); 149 vfs_msync(mp, MNT_NOWAIT); 150 VFS_SYNC(mp, MNT_NOWAIT); 151 curthread_pflags_restore(save); 152 vn_finished_write(mp); 153 } 154 VFS_UNLOCK_GIANT(vfslocked); 155 mtx_lock(&mountlist_mtx); 156 nmp = TAILQ_NEXT(mp, mnt_list); 157 vfs_unbusy(mp); 158 } 159 mtx_unlock(&mountlist_mtx); 160 return (0); 161 } 162 163 /* 164 * Change filesystem quotas. 165 */ 166 #ifndef _SYS_SYSPROTO_H_ 167 struct quotactl_args { 168 char *path; 169 int cmd; 170 int uid; 171 caddr_t arg; 172 }; 173 #endif 174 int 175 sys_quotactl(td, uap) 176 struct thread *td; 177 register struct quotactl_args /* { 178 char *path; 179 int cmd; 180 int uid; 181 caddr_t arg; 182 } */ *uap; 183 { 184 struct mount *mp; 185 int vfslocked; 186 int error; 187 struct nameidata nd; 188 189 AUDIT_ARG_CMD(uap->cmd); 190 AUDIT_ARG_UID(uap->uid); 191 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS)) 192 return (EPERM); 193 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1, 194 UIO_USERSPACE, uap->path, td); 195 if ((error = namei(&nd)) != 0) 196 return (error); 197 vfslocked = NDHASGIANT(&nd); 198 NDFREE(&nd, NDF_ONLY_PNBUF); 199 mp = nd.ni_vp->v_mount; 200 vfs_ref(mp); 201 vput(nd.ni_vp); 202 error = vfs_busy(mp, 0); 203 vfs_rel(mp); 204 if (error) { 205 VFS_UNLOCK_GIANT(vfslocked); 206 return (error); 207 } 208 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg); 209 210 /* 211 * Since quota on operation typically needs to open quota 212 * file, the Q_QUOTAON handler needs to unbusy the mount point 213 * before calling into namei. Otherwise, unmount might be 214 * started between two vfs_busy() invocations (first is our, 215 * second is from mount point cross-walk code in lookup()), 216 * causing deadlock. 217 * 218 * Require that Q_QUOTAON handles the vfs_busy() reference on 219 * its own, always returning with ubusied mount point. 220 */ 221 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON) 222 vfs_unbusy(mp); 223 VFS_UNLOCK_GIANT(vfslocked); 224 return (error); 225 } 226 227 /* 228 * Used by statfs conversion routines to scale the block size up if 229 * necessary so that all of the block counts are <= 'max_size'. Note 230 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero 231 * value of 'n'. 232 */ 233 void 234 statfs_scale_blocks(struct statfs *sf, long max_size) 235 { 236 uint64_t count; 237 int shift; 238 239 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__)); 240 241 /* 242 * Attempt to scale the block counts to give a more accurate 243 * overview to userland of the ratio of free space to used 244 * space. To do this, find the largest block count and compute 245 * a divisor that lets it fit into a signed integer <= max_size. 246 */ 247 if (sf->f_bavail < 0) 248 count = -sf->f_bavail; 249 else 250 count = sf->f_bavail; 251 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count)); 252 if (count <= max_size) 253 return; 254 255 count >>= flsl(max_size); 256 shift = 0; 257 while (count > 0) { 258 shift++; 259 count >>=1; 260 } 261 262 sf->f_bsize <<= shift; 263 sf->f_blocks >>= shift; 264 sf->f_bfree >>= shift; 265 sf->f_bavail >>= shift; 266 } 267 268 /* 269 * Get filesystem statistics. 270 */ 271 #ifndef _SYS_SYSPROTO_H_ 272 struct statfs_args { 273 char *path; 274 struct statfs *buf; 275 }; 276 #endif 277 int 278 sys_statfs(td, uap) 279 struct thread *td; 280 register struct statfs_args /* { 281 char *path; 282 struct statfs *buf; 283 } */ *uap; 284 { 285 struct statfs sf; 286 int error; 287 288 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf); 289 if (error == 0) 290 error = copyout(&sf, uap->buf, sizeof(sf)); 291 return (error); 292 } 293 294 int 295 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg, 296 struct statfs *buf) 297 { 298 struct mount *mp; 299 struct statfs *sp, sb; 300 int vfslocked; 301 int error; 302 struct nameidata nd; 303 304 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE | 305 AUDITVNODE1, pathseg, path, td); 306 error = namei(&nd); 307 if (error) 308 return (error); 309 vfslocked = NDHASGIANT(&nd); 310 mp = nd.ni_vp->v_mount; 311 vfs_ref(mp); 312 NDFREE(&nd, NDF_ONLY_PNBUF); 313 vput(nd.ni_vp); 314 error = vfs_busy(mp, 0); 315 vfs_rel(mp); 316 if (error) { 317 VFS_UNLOCK_GIANT(vfslocked); 318 return (error); 319 } 320 #ifdef MAC 321 error = mac_mount_check_stat(td->td_ucred, mp); 322 if (error) 323 goto out; 324 #endif 325 /* 326 * Set these in case the underlying filesystem fails to do so. 327 */ 328 sp = &mp->mnt_stat; 329 sp->f_version = STATFS_VERSION; 330 sp->f_namemax = NAME_MAX; 331 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 332 error = VFS_STATFS(mp, sp); 333 if (error) 334 goto out; 335 if (priv_check(td, PRIV_VFS_GENERATION)) { 336 bcopy(sp, &sb, sizeof(sb)); 337 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0; 338 prison_enforce_statfs(td->td_ucred, mp, &sb); 339 sp = &sb; 340 } 341 *buf = *sp; 342 out: 343 vfs_unbusy(mp); 344 VFS_UNLOCK_GIANT(vfslocked); 345 return (error); 346 } 347 348 /* 349 * Get filesystem statistics. 350 */ 351 #ifndef _SYS_SYSPROTO_H_ 352 struct fstatfs_args { 353 int fd; 354 struct statfs *buf; 355 }; 356 #endif 357 int 358 sys_fstatfs(td, uap) 359 struct thread *td; 360 register struct fstatfs_args /* { 361 int fd; 362 struct statfs *buf; 363 } */ *uap; 364 { 365 struct statfs sf; 366 int error; 367 368 error = kern_fstatfs(td, uap->fd, &sf); 369 if (error == 0) 370 error = copyout(&sf, uap->buf, sizeof(sf)); 371 return (error); 372 } 373 374 int 375 kern_fstatfs(struct thread *td, int fd, struct statfs *buf) 376 { 377 struct file *fp; 378 struct mount *mp; 379 struct statfs *sp, sb; 380 int vfslocked; 381 struct vnode *vp; 382 int error; 383 384 AUDIT_ARG_FD(fd); 385 error = getvnode(td->td_proc->p_fd, fd, CAP_FSTATFS, &fp); 386 if (error) 387 return (error); 388 vp = fp->f_vnode; 389 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 390 vn_lock(vp, LK_SHARED | LK_RETRY); 391 #ifdef AUDIT 392 AUDIT_ARG_VNODE1(vp); 393 #endif 394 mp = vp->v_mount; 395 if (mp) 396 vfs_ref(mp); 397 VOP_UNLOCK(vp, 0); 398 fdrop(fp, td); 399 if (mp == NULL) { 400 error = EBADF; 401 goto out; 402 } 403 error = vfs_busy(mp, 0); 404 vfs_rel(mp); 405 if (error) { 406 VFS_UNLOCK_GIANT(vfslocked); 407 return (error); 408 } 409 #ifdef MAC 410 error = mac_mount_check_stat(td->td_ucred, mp); 411 if (error) 412 goto out; 413 #endif 414 /* 415 * Set these in case the underlying filesystem fails to do so. 416 */ 417 sp = &mp->mnt_stat; 418 sp->f_version = STATFS_VERSION; 419 sp->f_namemax = NAME_MAX; 420 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 421 error = VFS_STATFS(mp, sp); 422 if (error) 423 goto out; 424 if (priv_check(td, PRIV_VFS_GENERATION)) { 425 bcopy(sp, &sb, sizeof(sb)); 426 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0; 427 prison_enforce_statfs(td->td_ucred, mp, &sb); 428 sp = &sb; 429 } 430 *buf = *sp; 431 out: 432 if (mp) 433 vfs_unbusy(mp); 434 VFS_UNLOCK_GIANT(vfslocked); 435 return (error); 436 } 437 438 /* 439 * Get statistics on all filesystems. 440 */ 441 #ifndef _SYS_SYSPROTO_H_ 442 struct getfsstat_args { 443 struct statfs *buf; 444 long bufsize; 445 int flags; 446 }; 447 #endif 448 int 449 sys_getfsstat(td, uap) 450 struct thread *td; 451 register struct getfsstat_args /* { 452 struct statfs *buf; 453 long bufsize; 454 int flags; 455 } */ *uap; 456 { 457 458 return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE, 459 uap->flags)); 460 } 461 462 /* 463 * If (bufsize > 0 && bufseg == UIO_SYSSPACE) 464 * The caller is responsible for freeing memory which will be allocated 465 * in '*buf'. 466 */ 467 int 468 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize, 469 enum uio_seg bufseg, int flags) 470 { 471 struct mount *mp, *nmp; 472 struct statfs *sfsp, *sp, sb; 473 size_t count, maxcount; 474 int vfslocked; 475 int error; 476 477 maxcount = bufsize / sizeof(struct statfs); 478 if (bufsize == 0) 479 sfsp = NULL; 480 else if (bufseg == UIO_USERSPACE) 481 sfsp = *buf; 482 else /* if (bufseg == UIO_SYSSPACE) */ { 483 count = 0; 484 mtx_lock(&mountlist_mtx); 485 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 486 count++; 487 } 488 mtx_unlock(&mountlist_mtx); 489 if (maxcount > count) 490 maxcount = count; 491 sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP, 492 M_WAITOK); 493 } 494 count = 0; 495 mtx_lock(&mountlist_mtx); 496 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 497 if (prison_canseemount(td->td_ucred, mp) != 0) { 498 nmp = TAILQ_NEXT(mp, mnt_list); 499 continue; 500 } 501 #ifdef MAC 502 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 503 nmp = TAILQ_NEXT(mp, mnt_list); 504 continue; 505 } 506 #endif 507 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) { 508 nmp = TAILQ_NEXT(mp, mnt_list); 509 continue; 510 } 511 vfslocked = VFS_LOCK_GIANT(mp); 512 if (sfsp && count < maxcount) { 513 sp = &mp->mnt_stat; 514 /* 515 * Set these in case the underlying filesystem 516 * fails to do so. 517 */ 518 sp->f_version = STATFS_VERSION; 519 sp->f_namemax = NAME_MAX; 520 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 521 /* 522 * If MNT_NOWAIT or MNT_LAZY is specified, do not 523 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY 524 * overrides MNT_WAIT. 525 */ 526 if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 || 527 (flags & MNT_WAIT)) && 528 (error = VFS_STATFS(mp, sp))) { 529 VFS_UNLOCK_GIANT(vfslocked); 530 mtx_lock(&mountlist_mtx); 531 nmp = TAILQ_NEXT(mp, mnt_list); 532 vfs_unbusy(mp); 533 continue; 534 } 535 if (priv_check(td, PRIV_VFS_GENERATION)) { 536 bcopy(sp, &sb, sizeof(sb)); 537 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0; 538 prison_enforce_statfs(td->td_ucred, mp, &sb); 539 sp = &sb; 540 } 541 if (bufseg == UIO_SYSSPACE) 542 bcopy(sp, sfsp, sizeof(*sp)); 543 else /* if (bufseg == UIO_USERSPACE) */ { 544 error = copyout(sp, sfsp, sizeof(*sp)); 545 if (error) { 546 vfs_unbusy(mp); 547 VFS_UNLOCK_GIANT(vfslocked); 548 return (error); 549 } 550 } 551 sfsp++; 552 } 553 VFS_UNLOCK_GIANT(vfslocked); 554 count++; 555 mtx_lock(&mountlist_mtx); 556 nmp = TAILQ_NEXT(mp, mnt_list); 557 vfs_unbusy(mp); 558 } 559 mtx_unlock(&mountlist_mtx); 560 if (sfsp && count > maxcount) 561 td->td_retval[0] = maxcount; 562 else 563 td->td_retval[0] = count; 564 return (0); 565 } 566 567 #ifdef COMPAT_FREEBSD4 568 /* 569 * Get old format filesystem statistics. 570 */ 571 static void cvtstatfs(struct statfs *, struct ostatfs *); 572 573 #ifndef _SYS_SYSPROTO_H_ 574 struct freebsd4_statfs_args { 575 char *path; 576 struct ostatfs *buf; 577 }; 578 #endif 579 int 580 freebsd4_statfs(td, uap) 581 struct thread *td; 582 struct freebsd4_statfs_args /* { 583 char *path; 584 struct ostatfs *buf; 585 } */ *uap; 586 { 587 struct ostatfs osb; 588 struct statfs sf; 589 int error; 590 591 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf); 592 if (error) 593 return (error); 594 cvtstatfs(&sf, &osb); 595 return (copyout(&osb, uap->buf, sizeof(osb))); 596 } 597 598 /* 599 * Get filesystem statistics. 600 */ 601 #ifndef _SYS_SYSPROTO_H_ 602 struct freebsd4_fstatfs_args { 603 int fd; 604 struct ostatfs *buf; 605 }; 606 #endif 607 int 608 freebsd4_fstatfs(td, uap) 609 struct thread *td; 610 struct freebsd4_fstatfs_args /* { 611 int fd; 612 struct ostatfs *buf; 613 } */ *uap; 614 { 615 struct ostatfs osb; 616 struct statfs sf; 617 int error; 618 619 error = kern_fstatfs(td, uap->fd, &sf); 620 if (error) 621 return (error); 622 cvtstatfs(&sf, &osb); 623 return (copyout(&osb, uap->buf, sizeof(osb))); 624 } 625 626 /* 627 * Get statistics on all filesystems. 628 */ 629 #ifndef _SYS_SYSPROTO_H_ 630 struct freebsd4_getfsstat_args { 631 struct ostatfs *buf; 632 long bufsize; 633 int flags; 634 }; 635 #endif 636 int 637 freebsd4_getfsstat(td, uap) 638 struct thread *td; 639 register struct freebsd4_getfsstat_args /* { 640 struct ostatfs *buf; 641 long bufsize; 642 int flags; 643 } */ *uap; 644 { 645 struct statfs *buf, *sp; 646 struct ostatfs osb; 647 size_t count, size; 648 int error; 649 650 count = uap->bufsize / sizeof(struct ostatfs); 651 size = count * sizeof(struct statfs); 652 error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags); 653 if (size > 0) { 654 count = td->td_retval[0]; 655 sp = buf; 656 while (count > 0 && error == 0) { 657 cvtstatfs(sp, &osb); 658 error = copyout(&osb, uap->buf, sizeof(osb)); 659 sp++; 660 uap->buf++; 661 count--; 662 } 663 free(buf, M_TEMP); 664 } 665 return (error); 666 } 667 668 /* 669 * Implement fstatfs() for (NFS) file handles. 670 */ 671 #ifndef _SYS_SYSPROTO_H_ 672 struct freebsd4_fhstatfs_args { 673 struct fhandle *u_fhp; 674 struct ostatfs *buf; 675 }; 676 #endif 677 int 678 freebsd4_fhstatfs(td, uap) 679 struct thread *td; 680 struct freebsd4_fhstatfs_args /* { 681 struct fhandle *u_fhp; 682 struct ostatfs *buf; 683 } */ *uap; 684 { 685 struct ostatfs osb; 686 struct statfs sf; 687 fhandle_t fh; 688 int error; 689 690 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 691 if (error) 692 return (error); 693 error = kern_fhstatfs(td, fh, &sf); 694 if (error) 695 return (error); 696 cvtstatfs(&sf, &osb); 697 return (copyout(&osb, uap->buf, sizeof(osb))); 698 } 699 700 /* 701 * Convert a new format statfs structure to an old format statfs structure. 702 */ 703 static void 704 cvtstatfs(nsp, osp) 705 struct statfs *nsp; 706 struct ostatfs *osp; 707 { 708 709 statfs_scale_blocks(nsp, LONG_MAX); 710 bzero(osp, sizeof(*osp)); 711 osp->f_bsize = nsp->f_bsize; 712 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX); 713 osp->f_blocks = nsp->f_blocks; 714 osp->f_bfree = nsp->f_bfree; 715 osp->f_bavail = nsp->f_bavail; 716 osp->f_files = MIN(nsp->f_files, LONG_MAX); 717 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX); 718 osp->f_owner = nsp->f_owner; 719 osp->f_type = nsp->f_type; 720 osp->f_flags = nsp->f_flags; 721 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX); 722 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX); 723 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX); 724 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX); 725 strlcpy(osp->f_fstypename, nsp->f_fstypename, 726 MIN(MFSNAMELEN, OMFSNAMELEN)); 727 strlcpy(osp->f_mntonname, nsp->f_mntonname, 728 MIN(MNAMELEN, OMNAMELEN)); 729 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 730 MIN(MNAMELEN, OMNAMELEN)); 731 osp->f_fsid = nsp->f_fsid; 732 } 733 #endif /* COMPAT_FREEBSD4 */ 734 735 /* 736 * Change current working directory to a given file descriptor. 737 */ 738 #ifndef _SYS_SYSPROTO_H_ 739 struct fchdir_args { 740 int fd; 741 }; 742 #endif 743 int 744 sys_fchdir(td, uap) 745 struct thread *td; 746 struct fchdir_args /* { 747 int fd; 748 } */ *uap; 749 { 750 register struct filedesc *fdp = td->td_proc->p_fd; 751 struct vnode *vp, *tdp, *vpold; 752 struct mount *mp; 753 struct file *fp; 754 int vfslocked; 755 int error; 756 757 AUDIT_ARG_FD(uap->fd); 758 if ((error = getvnode(fdp, uap->fd, CAP_FCHDIR, &fp)) != 0) 759 return (error); 760 vp = fp->f_vnode; 761 VREF(vp); 762 fdrop(fp, td); 763 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 764 vn_lock(vp, LK_SHARED | LK_RETRY); 765 AUDIT_ARG_VNODE1(vp); 766 error = change_dir(vp, td); 767 while (!error && (mp = vp->v_mountedhere) != NULL) { 768 int tvfslocked; 769 if (vfs_busy(mp, 0)) 770 continue; 771 tvfslocked = VFS_LOCK_GIANT(mp); 772 error = VFS_ROOT(mp, LK_SHARED, &tdp); 773 vfs_unbusy(mp); 774 if (error) { 775 VFS_UNLOCK_GIANT(tvfslocked); 776 break; 777 } 778 vput(vp); 779 VFS_UNLOCK_GIANT(vfslocked); 780 vp = tdp; 781 vfslocked = tvfslocked; 782 } 783 if (error) { 784 vput(vp); 785 VFS_UNLOCK_GIANT(vfslocked); 786 return (error); 787 } 788 VOP_UNLOCK(vp, 0); 789 VFS_UNLOCK_GIANT(vfslocked); 790 FILEDESC_XLOCK(fdp); 791 vpold = fdp->fd_cdir; 792 fdp->fd_cdir = vp; 793 FILEDESC_XUNLOCK(fdp); 794 vfslocked = VFS_LOCK_GIANT(vpold->v_mount); 795 vrele(vpold); 796 VFS_UNLOCK_GIANT(vfslocked); 797 return (0); 798 } 799 800 /* 801 * Change current working directory (``.''). 802 */ 803 #ifndef _SYS_SYSPROTO_H_ 804 struct chdir_args { 805 char *path; 806 }; 807 #endif 808 int 809 sys_chdir(td, uap) 810 struct thread *td; 811 struct chdir_args /* { 812 char *path; 813 } */ *uap; 814 { 815 816 return (kern_chdir(td, uap->path, UIO_USERSPACE)); 817 } 818 819 int 820 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg) 821 { 822 register struct filedesc *fdp = td->td_proc->p_fd; 823 int error; 824 struct nameidata nd; 825 struct vnode *vp; 826 int vfslocked; 827 828 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | 829 MPSAFE, pathseg, path, td); 830 if ((error = namei(&nd)) != 0) 831 return (error); 832 vfslocked = NDHASGIANT(&nd); 833 if ((error = change_dir(nd.ni_vp, td)) != 0) { 834 vput(nd.ni_vp); 835 VFS_UNLOCK_GIANT(vfslocked); 836 NDFREE(&nd, NDF_ONLY_PNBUF); 837 return (error); 838 } 839 VOP_UNLOCK(nd.ni_vp, 0); 840 VFS_UNLOCK_GIANT(vfslocked); 841 NDFREE(&nd, NDF_ONLY_PNBUF); 842 FILEDESC_XLOCK(fdp); 843 vp = fdp->fd_cdir; 844 fdp->fd_cdir = nd.ni_vp; 845 FILEDESC_XUNLOCK(fdp); 846 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 847 vrele(vp); 848 VFS_UNLOCK_GIANT(vfslocked); 849 return (0); 850 } 851 852 /* 853 * Helper function for raised chroot(2) security function: Refuse if 854 * any filedescriptors are open directories. 855 */ 856 static int 857 chroot_refuse_vdir_fds(fdp) 858 struct filedesc *fdp; 859 { 860 struct vnode *vp; 861 struct file *fp; 862 int fd; 863 864 FILEDESC_LOCK_ASSERT(fdp); 865 866 for (fd = 0; fd < fdp->fd_nfiles ; fd++) { 867 fp = fget_locked(fdp, fd); 868 if (fp == NULL) 869 continue; 870 if (fp->f_type == DTYPE_VNODE) { 871 vp = fp->f_vnode; 872 if (vp->v_type == VDIR) 873 return (EPERM); 874 } 875 } 876 return (0); 877 } 878 879 /* 880 * This sysctl determines if we will allow a process to chroot(2) if it 881 * has a directory open: 882 * 0: disallowed for all processes. 883 * 1: allowed for processes that were not already chroot(2)'ed. 884 * 2: allowed for all processes. 885 */ 886 887 static int chroot_allow_open_directories = 1; 888 889 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW, 890 &chroot_allow_open_directories, 0, ""); 891 892 /* 893 * Change notion of root (``/'') directory. 894 */ 895 #ifndef _SYS_SYSPROTO_H_ 896 struct chroot_args { 897 char *path; 898 }; 899 #endif 900 int 901 sys_chroot(td, uap) 902 struct thread *td; 903 struct chroot_args /* { 904 char *path; 905 } */ *uap; 906 { 907 int error; 908 struct nameidata nd; 909 int vfslocked; 910 911 error = priv_check(td, PRIV_VFS_CHROOT); 912 if (error) 913 return (error); 914 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE | 915 AUDITVNODE1, UIO_USERSPACE, uap->path, td); 916 error = namei(&nd); 917 if (error) 918 goto error; 919 vfslocked = NDHASGIANT(&nd); 920 if ((error = change_dir(nd.ni_vp, td)) != 0) 921 goto e_vunlock; 922 #ifdef MAC 923 if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp))) 924 goto e_vunlock; 925 #endif 926 VOP_UNLOCK(nd.ni_vp, 0); 927 error = change_root(nd.ni_vp, td); 928 vrele(nd.ni_vp); 929 VFS_UNLOCK_GIANT(vfslocked); 930 NDFREE(&nd, NDF_ONLY_PNBUF); 931 return (error); 932 e_vunlock: 933 vput(nd.ni_vp); 934 VFS_UNLOCK_GIANT(vfslocked); 935 error: 936 NDFREE(&nd, NDF_ONLY_PNBUF); 937 return (error); 938 } 939 940 /* 941 * Common routine for chroot and chdir. Callers must provide a locked vnode 942 * instance. 943 */ 944 int 945 change_dir(vp, td) 946 struct vnode *vp; 947 struct thread *td; 948 { 949 int error; 950 951 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked"); 952 if (vp->v_type != VDIR) 953 return (ENOTDIR); 954 #ifdef MAC 955 error = mac_vnode_check_chdir(td->td_ucred, vp); 956 if (error) 957 return (error); 958 #endif 959 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 960 return (error); 961 } 962 963 /* 964 * Common routine for kern_chroot() and jail_attach(). The caller is 965 * responsible for invoking priv_check() and mac_vnode_check_chroot() to 966 * authorize this operation. 967 */ 968 int 969 change_root(vp, td) 970 struct vnode *vp; 971 struct thread *td; 972 { 973 struct filedesc *fdp; 974 struct vnode *oldvp; 975 int vfslocked; 976 int error; 977 978 VFS_ASSERT_GIANT(vp->v_mount); 979 fdp = td->td_proc->p_fd; 980 FILEDESC_XLOCK(fdp); 981 if (chroot_allow_open_directories == 0 || 982 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) { 983 error = chroot_refuse_vdir_fds(fdp); 984 if (error) { 985 FILEDESC_XUNLOCK(fdp); 986 return (error); 987 } 988 } 989 oldvp = fdp->fd_rdir; 990 fdp->fd_rdir = vp; 991 VREF(fdp->fd_rdir); 992 if (!fdp->fd_jdir) { 993 fdp->fd_jdir = vp; 994 VREF(fdp->fd_jdir); 995 } 996 FILEDESC_XUNLOCK(fdp); 997 vfslocked = VFS_LOCK_GIANT(oldvp->v_mount); 998 vrele(oldvp); 999 VFS_UNLOCK_GIANT(vfslocked); 1000 return (0); 1001 } 1002 1003 static __inline cap_rights_t 1004 flags_to_rights(int flags) 1005 { 1006 cap_rights_t rights = 0; 1007 1008 switch ((flags & O_ACCMODE)) { 1009 case O_RDONLY: 1010 rights |= CAP_READ; 1011 break; 1012 1013 case O_RDWR: 1014 rights |= CAP_READ; 1015 /* fall through */ 1016 1017 case O_WRONLY: 1018 rights |= CAP_WRITE; 1019 break; 1020 1021 case O_EXEC: 1022 rights |= CAP_FEXECVE; 1023 break; 1024 } 1025 1026 if (flags & O_CREAT) 1027 rights |= CAP_CREATE; 1028 1029 if (flags & O_TRUNC) 1030 rights |= CAP_FTRUNCATE; 1031 1032 if ((flags & O_EXLOCK) || (flags & O_SHLOCK)) 1033 rights |= CAP_FLOCK; 1034 1035 return (rights); 1036 } 1037 1038 /* 1039 * Check permissions, allocate an open file structure, and call the device 1040 * open routine if any. 1041 */ 1042 #ifndef _SYS_SYSPROTO_H_ 1043 struct open_args { 1044 char *path; 1045 int flags; 1046 int mode; 1047 }; 1048 #endif 1049 int 1050 sys_open(td, uap) 1051 struct thread *td; 1052 register struct open_args /* { 1053 char *path; 1054 int flags; 1055 int mode; 1056 } */ *uap; 1057 { 1058 1059 return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode)); 1060 } 1061 1062 #ifndef _SYS_SYSPROTO_H_ 1063 struct openat_args { 1064 int fd; 1065 char *path; 1066 int flag; 1067 int mode; 1068 }; 1069 #endif 1070 int 1071 sys_openat(struct thread *td, struct openat_args *uap) 1072 { 1073 1074 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 1075 uap->mode)); 1076 } 1077 1078 int 1079 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags, 1080 int mode) 1081 { 1082 1083 return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode)); 1084 } 1085 1086 int 1087 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1088 int flags, int mode) 1089 { 1090 struct proc *p = td->td_proc; 1091 struct filedesc *fdp = p->p_fd; 1092 struct file *fp; 1093 struct vnode *vp; 1094 int cmode; 1095 struct file *nfp; 1096 int type, indx = -1, error, error_open; 1097 struct flock lf; 1098 struct nameidata nd; 1099 int vfslocked; 1100 cap_rights_t rights_needed = CAP_LOOKUP; 1101 1102 AUDIT_ARG_FFLAGS(flags); 1103 AUDIT_ARG_MODE(mode); 1104 /* XXX: audit dirfd */ 1105 rights_needed |= flags_to_rights(flags); 1106 /* 1107 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 1108 * may be specified. 1109 */ 1110 if (flags & O_EXEC) { 1111 if (flags & O_ACCMODE) 1112 return (EINVAL); 1113 } else if ((flags & O_ACCMODE) == O_ACCMODE) 1114 return (EINVAL); 1115 else 1116 flags = FFLAGS(flags); 1117 1118 /* 1119 * allocate the file descriptor, but don't install a descriptor yet 1120 */ 1121 error = falloc_noinstall(td, &nfp); 1122 if (error) 1123 return (error); 1124 /* An extra reference on `nfp' has been held for us by falloc_noinstall(). */ 1125 fp = nfp; 1126 /* Set the flags early so the finit in devfs can pick them up. */ 1127 fp->f_flag = flags & FMASK; 1128 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT; 1129 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg, 1130 path, fd, rights_needed, td); 1131 td->td_dupfd = -1; /* XXX check for fdopen */ 1132 error = vn_open(&nd, &flags, cmode, fp); 1133 if (error) { 1134 /* 1135 * If the vn_open replaced the method vector, something 1136 * wonderous happened deep below and we just pass it up 1137 * pretending we know what we do. 1138 */ 1139 if (error == ENXIO && fp->f_ops != &badfileops) 1140 goto success; 1141 1142 /* 1143 * handle special fdopen() case. bleh. dupfdopen() is 1144 * responsible for dropping the old contents of ofiles[indx] 1145 * if it succeeds. 1146 * 1147 * Don't do this for relative (capability) lookups; we don't 1148 * understand exactly what would happen, and we don't think 1149 * that it ever should. 1150 */ 1151 if ((nd.ni_strictrelative == 0) && 1152 (error == ENODEV || error == ENXIO) && 1153 (td->td_dupfd >= 0)) { 1154 /* XXX from fdopen */ 1155 error_open = error; 1156 if ((error = finstall(td, fp, &indx, flags)) != 0) 1157 goto bad_unlocked; 1158 if ((error = dupfdopen(td, fdp, indx, td->td_dupfd, 1159 flags, error_open)) == 0) 1160 goto success; 1161 } 1162 /* 1163 * Clean up the descriptor, but only if another thread hadn't 1164 * replaced or closed it. 1165 */ 1166 if (indx != -1) 1167 fdclose(fdp, fp, indx, td); 1168 fdrop(fp, td); 1169 1170 if (error == ERESTART) 1171 error = EINTR; 1172 return (error); 1173 } 1174 td->td_dupfd = 0; 1175 vfslocked = NDHASGIANT(&nd); 1176 NDFREE(&nd, NDF_ONLY_PNBUF); 1177 vp = nd.ni_vp; 1178 1179 /* 1180 * Store the vnode, for any f_type. Typically, the vnode use 1181 * count is decremented by direct call to vn_closefile() for 1182 * files that switched type in the cdevsw fdopen() method. 1183 */ 1184 fp->f_vnode = vp; 1185 /* 1186 * If the file wasn't claimed by devfs bind it to the normal 1187 * vnode operations here. 1188 */ 1189 if (fp->f_ops == &badfileops) { 1190 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo.")); 1191 fp->f_seqcount = 1; 1192 finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops); 1193 } 1194 1195 VOP_UNLOCK(vp, 0); 1196 if (fp->f_type == DTYPE_VNODE && (flags & (O_EXLOCK | O_SHLOCK)) != 0) { 1197 lf.l_whence = SEEK_SET; 1198 lf.l_start = 0; 1199 lf.l_len = 0; 1200 if (flags & O_EXLOCK) 1201 lf.l_type = F_WRLCK; 1202 else 1203 lf.l_type = F_RDLCK; 1204 type = F_FLOCK; 1205 if ((flags & FNONBLOCK) == 0) 1206 type |= F_WAIT; 1207 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, 1208 type)) != 0) 1209 goto bad; 1210 atomic_set_int(&fp->f_flag, FHASLOCK); 1211 } 1212 if (flags & O_TRUNC) { 1213 error = fo_truncate(fp, 0, td->td_ucred, td); 1214 if (error) 1215 goto bad; 1216 } 1217 VFS_UNLOCK_GIANT(vfslocked); 1218 success: 1219 /* 1220 * If we haven't already installed the FD (for dupfdopen), do so now. 1221 */ 1222 if (indx == -1) { 1223 #ifdef CAPABILITIES 1224 if (nd.ni_strictrelative == 1) { 1225 /* 1226 * We are doing a strict relative lookup; wrap the 1227 * result in a capability. 1228 */ 1229 if ((error = kern_capwrap(td, fp, nd.ni_baserights, 1230 &indx)) != 0) 1231 goto bad_unlocked; 1232 } else 1233 #endif 1234 if ((error = finstall(td, fp, &indx, flags)) != 0) 1235 goto bad_unlocked; 1236 1237 } 1238 1239 /* 1240 * Release our private reference, leaving the one associated with 1241 * the descriptor table intact. 1242 */ 1243 fdrop(fp, td); 1244 td->td_retval[0] = indx; 1245 return (0); 1246 bad: 1247 VFS_UNLOCK_GIANT(vfslocked); 1248 bad_unlocked: 1249 if (indx != -1) 1250 fdclose(fdp, fp, indx, td); 1251 fdrop(fp, td); 1252 td->td_retval[0] = -1; 1253 return (error); 1254 } 1255 1256 #ifdef COMPAT_43 1257 /* 1258 * Create a file. 1259 */ 1260 #ifndef _SYS_SYSPROTO_H_ 1261 struct ocreat_args { 1262 char *path; 1263 int mode; 1264 }; 1265 #endif 1266 int 1267 ocreat(td, uap) 1268 struct thread *td; 1269 register struct ocreat_args /* { 1270 char *path; 1271 int mode; 1272 } */ *uap; 1273 { 1274 1275 return (kern_open(td, uap->path, UIO_USERSPACE, 1276 O_WRONLY | O_CREAT | O_TRUNC, uap->mode)); 1277 } 1278 #endif /* COMPAT_43 */ 1279 1280 /* 1281 * Create a special file. 1282 */ 1283 #ifndef _SYS_SYSPROTO_H_ 1284 struct mknod_args { 1285 char *path; 1286 int mode; 1287 int dev; 1288 }; 1289 #endif 1290 int 1291 sys_mknod(td, uap) 1292 struct thread *td; 1293 register struct mknod_args /* { 1294 char *path; 1295 int mode; 1296 int dev; 1297 } */ *uap; 1298 { 1299 1300 return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev)); 1301 } 1302 1303 #ifndef _SYS_SYSPROTO_H_ 1304 struct mknodat_args { 1305 int fd; 1306 char *path; 1307 mode_t mode; 1308 dev_t dev; 1309 }; 1310 #endif 1311 int 1312 sys_mknodat(struct thread *td, struct mknodat_args *uap) 1313 { 1314 1315 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1316 uap->dev)); 1317 } 1318 1319 int 1320 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode, 1321 int dev) 1322 { 1323 1324 return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev)); 1325 } 1326 1327 int 1328 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1329 int mode, int dev) 1330 { 1331 struct vnode *vp; 1332 struct mount *mp; 1333 struct vattr vattr; 1334 int error; 1335 int whiteout = 0; 1336 struct nameidata nd; 1337 int vfslocked; 1338 1339 AUDIT_ARG_MODE(mode); 1340 AUDIT_ARG_DEV(dev); 1341 switch (mode & S_IFMT) { 1342 case S_IFCHR: 1343 case S_IFBLK: 1344 error = priv_check(td, PRIV_VFS_MKNOD_DEV); 1345 break; 1346 case S_IFMT: 1347 error = priv_check(td, PRIV_VFS_MKNOD_BAD); 1348 break; 1349 case S_IFWHT: 1350 error = priv_check(td, PRIV_VFS_MKNOD_WHT); 1351 break; 1352 case S_IFIFO: 1353 if (dev == 0) 1354 return (kern_mkfifoat(td, fd, path, pathseg, mode)); 1355 /* FALLTHROUGH */ 1356 default: 1357 error = EINVAL; 1358 break; 1359 } 1360 if (error) 1361 return (error); 1362 restart: 1363 bwillwrite(); 1364 NDINIT_ATRIGHTS(&nd, CREATE, 1365 LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1, pathseg, path, fd, 1366 CAP_MKFIFO, td); 1367 if ((error = namei(&nd)) != 0) 1368 return (error); 1369 vfslocked = NDHASGIANT(&nd); 1370 vp = nd.ni_vp; 1371 if (vp != NULL) { 1372 NDFREE(&nd, NDF_ONLY_PNBUF); 1373 if (vp == nd.ni_dvp) 1374 vrele(nd.ni_dvp); 1375 else 1376 vput(nd.ni_dvp); 1377 vrele(vp); 1378 VFS_UNLOCK_GIANT(vfslocked); 1379 return (EEXIST); 1380 } else { 1381 VATTR_NULL(&vattr); 1382 vattr.va_mode = (mode & ALLPERMS) & 1383 ~td->td_proc->p_fd->fd_cmask; 1384 vattr.va_rdev = dev; 1385 whiteout = 0; 1386 1387 switch (mode & S_IFMT) { 1388 case S_IFMT: /* used by badsect to flag bad sectors */ 1389 vattr.va_type = VBAD; 1390 break; 1391 case S_IFCHR: 1392 vattr.va_type = VCHR; 1393 break; 1394 case S_IFBLK: 1395 vattr.va_type = VBLK; 1396 break; 1397 case S_IFWHT: 1398 whiteout = 1; 1399 break; 1400 default: 1401 panic("kern_mknod: invalid mode"); 1402 } 1403 } 1404 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1405 NDFREE(&nd, NDF_ONLY_PNBUF); 1406 vput(nd.ni_dvp); 1407 VFS_UNLOCK_GIANT(vfslocked); 1408 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1409 return (error); 1410 goto restart; 1411 } 1412 #ifdef MAC 1413 if (error == 0 && !whiteout) 1414 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, 1415 &nd.ni_cnd, &vattr); 1416 #endif 1417 if (!error) { 1418 if (whiteout) 1419 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1420 else { 1421 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1422 &nd.ni_cnd, &vattr); 1423 if (error == 0) 1424 vput(nd.ni_vp); 1425 } 1426 } 1427 NDFREE(&nd, NDF_ONLY_PNBUF); 1428 vput(nd.ni_dvp); 1429 vn_finished_write(mp); 1430 VFS_UNLOCK_GIANT(vfslocked); 1431 return (error); 1432 } 1433 1434 /* 1435 * Create a named pipe. 1436 */ 1437 #ifndef _SYS_SYSPROTO_H_ 1438 struct mkfifo_args { 1439 char *path; 1440 int mode; 1441 }; 1442 #endif 1443 int 1444 sys_mkfifo(td, uap) 1445 struct thread *td; 1446 register struct mkfifo_args /* { 1447 char *path; 1448 int mode; 1449 } */ *uap; 1450 { 1451 1452 return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode)); 1453 } 1454 1455 #ifndef _SYS_SYSPROTO_H_ 1456 struct mkfifoat_args { 1457 int fd; 1458 char *path; 1459 mode_t mode; 1460 }; 1461 #endif 1462 int 1463 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap) 1464 { 1465 1466 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE, 1467 uap->mode)); 1468 } 1469 1470 int 1471 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode) 1472 { 1473 1474 return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode)); 1475 } 1476 1477 int 1478 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1479 int mode) 1480 { 1481 struct mount *mp; 1482 struct vattr vattr; 1483 int error; 1484 struct nameidata nd; 1485 int vfslocked; 1486 1487 AUDIT_ARG_MODE(mode); 1488 restart: 1489 bwillwrite(); 1490 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1, 1491 pathseg, path, fd, td); 1492 if ((error = namei(&nd)) != 0) 1493 return (error); 1494 vfslocked = NDHASGIANT(&nd); 1495 if (nd.ni_vp != NULL) { 1496 NDFREE(&nd, NDF_ONLY_PNBUF); 1497 if (nd.ni_vp == nd.ni_dvp) 1498 vrele(nd.ni_dvp); 1499 else 1500 vput(nd.ni_dvp); 1501 vrele(nd.ni_vp); 1502 VFS_UNLOCK_GIANT(vfslocked); 1503 return (EEXIST); 1504 } 1505 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1506 NDFREE(&nd, NDF_ONLY_PNBUF); 1507 vput(nd.ni_dvp); 1508 VFS_UNLOCK_GIANT(vfslocked); 1509 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1510 return (error); 1511 goto restart; 1512 } 1513 VATTR_NULL(&vattr); 1514 vattr.va_type = VFIFO; 1515 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask; 1516 #ifdef MAC 1517 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1518 &vattr); 1519 if (error) 1520 goto out; 1521 #endif 1522 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1523 if (error == 0) 1524 vput(nd.ni_vp); 1525 #ifdef MAC 1526 out: 1527 #endif 1528 vput(nd.ni_dvp); 1529 vn_finished_write(mp); 1530 VFS_UNLOCK_GIANT(vfslocked); 1531 NDFREE(&nd, NDF_ONLY_PNBUF); 1532 return (error); 1533 } 1534 1535 /* 1536 * Make a hard file link. 1537 */ 1538 #ifndef _SYS_SYSPROTO_H_ 1539 struct link_args { 1540 char *path; 1541 char *link; 1542 }; 1543 #endif 1544 int 1545 sys_link(td, uap) 1546 struct thread *td; 1547 register struct link_args /* { 1548 char *path; 1549 char *link; 1550 } */ *uap; 1551 { 1552 1553 return (kern_link(td, uap->path, uap->link, UIO_USERSPACE)); 1554 } 1555 1556 #ifndef _SYS_SYSPROTO_H_ 1557 struct linkat_args { 1558 int fd1; 1559 char *path1; 1560 int fd2; 1561 char *path2; 1562 int flag; 1563 }; 1564 #endif 1565 int 1566 sys_linkat(struct thread *td, struct linkat_args *uap) 1567 { 1568 int flag; 1569 1570 flag = uap->flag; 1571 if (flag & ~AT_SYMLINK_FOLLOW) 1572 return (EINVAL); 1573 1574 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2, 1575 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW)); 1576 } 1577 1578 int hardlink_check_uid = 0; 1579 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW, 1580 &hardlink_check_uid, 0, 1581 "Unprivileged processes cannot create hard links to files owned by other " 1582 "users"); 1583 static int hardlink_check_gid = 0; 1584 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW, 1585 &hardlink_check_gid, 0, 1586 "Unprivileged processes cannot create hard links to files owned by other " 1587 "groups"); 1588 1589 static int 1590 can_hardlink(struct vnode *vp, struct ucred *cred) 1591 { 1592 struct vattr va; 1593 int error; 1594 1595 if (!hardlink_check_uid && !hardlink_check_gid) 1596 return (0); 1597 1598 error = VOP_GETATTR(vp, &va, cred); 1599 if (error != 0) 1600 return (error); 1601 1602 if (hardlink_check_uid && cred->cr_uid != va.va_uid) { 1603 error = priv_check_cred(cred, PRIV_VFS_LINK, 0); 1604 if (error) 1605 return (error); 1606 } 1607 1608 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) { 1609 error = priv_check_cred(cred, PRIV_VFS_LINK, 0); 1610 if (error) 1611 return (error); 1612 } 1613 1614 return (0); 1615 } 1616 1617 int 1618 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg) 1619 { 1620 1621 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW)); 1622 } 1623 1624 int 1625 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2, 1626 enum uio_seg segflg, int follow) 1627 { 1628 struct vnode *vp; 1629 struct mount *mp; 1630 struct nameidata nd; 1631 int vfslocked; 1632 int lvfslocked; 1633 int error; 1634 1635 bwillwrite(); 1636 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1, 1637 fd1, td); 1638 1639 if ((error = namei(&nd)) != 0) 1640 return (error); 1641 vfslocked = NDHASGIANT(&nd); 1642 NDFREE(&nd, NDF_ONLY_PNBUF); 1643 vp = nd.ni_vp; 1644 if (vp->v_type == VDIR) { 1645 vrele(vp); 1646 VFS_UNLOCK_GIANT(vfslocked); 1647 return (EPERM); /* POSIX */ 1648 } 1649 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) { 1650 vrele(vp); 1651 VFS_UNLOCK_GIANT(vfslocked); 1652 return (error); 1653 } 1654 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2, 1655 segflg, path2, fd2, td); 1656 if ((error = namei(&nd)) == 0) { 1657 lvfslocked = NDHASGIANT(&nd); 1658 if (nd.ni_vp != NULL) { 1659 if (nd.ni_dvp == nd.ni_vp) 1660 vrele(nd.ni_dvp); 1661 else 1662 vput(nd.ni_dvp); 1663 vrele(nd.ni_vp); 1664 error = EEXIST; 1665 } else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) 1666 == 0) { 1667 error = can_hardlink(vp, td->td_ucred); 1668 if (error == 0) 1669 #ifdef MAC 1670 error = mac_vnode_check_link(td->td_ucred, 1671 nd.ni_dvp, vp, &nd.ni_cnd); 1672 if (error == 0) 1673 #endif 1674 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1675 VOP_UNLOCK(vp, 0); 1676 vput(nd.ni_dvp); 1677 } 1678 NDFREE(&nd, NDF_ONLY_PNBUF); 1679 VFS_UNLOCK_GIANT(lvfslocked); 1680 } 1681 vrele(vp); 1682 vn_finished_write(mp); 1683 VFS_UNLOCK_GIANT(vfslocked); 1684 return (error); 1685 } 1686 1687 /* 1688 * Make a symbolic link. 1689 */ 1690 #ifndef _SYS_SYSPROTO_H_ 1691 struct symlink_args { 1692 char *path; 1693 char *link; 1694 }; 1695 #endif 1696 int 1697 sys_symlink(td, uap) 1698 struct thread *td; 1699 register struct symlink_args /* { 1700 char *path; 1701 char *link; 1702 } */ *uap; 1703 { 1704 1705 return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE)); 1706 } 1707 1708 #ifndef _SYS_SYSPROTO_H_ 1709 struct symlinkat_args { 1710 char *path; 1711 int fd; 1712 char *path2; 1713 }; 1714 #endif 1715 int 1716 sys_symlinkat(struct thread *td, struct symlinkat_args *uap) 1717 { 1718 1719 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2, 1720 UIO_USERSPACE)); 1721 } 1722 1723 int 1724 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg) 1725 { 1726 1727 return (kern_symlinkat(td, path, AT_FDCWD, link, segflg)); 1728 } 1729 1730 int 1731 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2, 1732 enum uio_seg segflg) 1733 { 1734 struct mount *mp; 1735 struct vattr vattr; 1736 char *syspath; 1737 int error; 1738 struct nameidata nd; 1739 int vfslocked; 1740 1741 if (segflg == UIO_SYSSPACE) { 1742 syspath = path1; 1743 } else { 1744 syspath = uma_zalloc(namei_zone, M_WAITOK); 1745 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0) 1746 goto out; 1747 } 1748 AUDIT_ARG_TEXT(syspath); 1749 restart: 1750 bwillwrite(); 1751 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1, 1752 segflg, path2, fd, td); 1753 if ((error = namei(&nd)) != 0) 1754 goto out; 1755 vfslocked = NDHASGIANT(&nd); 1756 if (nd.ni_vp) { 1757 NDFREE(&nd, NDF_ONLY_PNBUF); 1758 if (nd.ni_vp == nd.ni_dvp) 1759 vrele(nd.ni_dvp); 1760 else 1761 vput(nd.ni_dvp); 1762 vrele(nd.ni_vp); 1763 VFS_UNLOCK_GIANT(vfslocked); 1764 error = EEXIST; 1765 goto out; 1766 } 1767 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1768 NDFREE(&nd, NDF_ONLY_PNBUF); 1769 vput(nd.ni_dvp); 1770 VFS_UNLOCK_GIANT(vfslocked); 1771 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1772 goto out; 1773 goto restart; 1774 } 1775 VATTR_NULL(&vattr); 1776 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask; 1777 #ifdef MAC 1778 vattr.va_type = VLNK; 1779 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1780 &vattr); 1781 if (error) 1782 goto out2; 1783 #endif 1784 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath); 1785 if (error == 0) 1786 vput(nd.ni_vp); 1787 #ifdef MAC 1788 out2: 1789 #endif 1790 NDFREE(&nd, NDF_ONLY_PNBUF); 1791 vput(nd.ni_dvp); 1792 vn_finished_write(mp); 1793 VFS_UNLOCK_GIANT(vfslocked); 1794 out: 1795 if (segflg != UIO_SYSSPACE) 1796 uma_zfree(namei_zone, syspath); 1797 return (error); 1798 } 1799 1800 /* 1801 * Delete a whiteout from the filesystem. 1802 */ 1803 int 1804 sys_undelete(td, uap) 1805 struct thread *td; 1806 register struct undelete_args /* { 1807 char *path; 1808 } */ *uap; 1809 { 1810 int error; 1811 struct mount *mp; 1812 struct nameidata nd; 1813 int vfslocked; 1814 1815 restart: 1816 bwillwrite(); 1817 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1, 1818 UIO_USERSPACE, uap->path, td); 1819 error = namei(&nd); 1820 if (error) 1821 return (error); 1822 vfslocked = NDHASGIANT(&nd); 1823 1824 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1825 NDFREE(&nd, NDF_ONLY_PNBUF); 1826 if (nd.ni_vp == nd.ni_dvp) 1827 vrele(nd.ni_dvp); 1828 else 1829 vput(nd.ni_dvp); 1830 if (nd.ni_vp) 1831 vrele(nd.ni_vp); 1832 VFS_UNLOCK_GIANT(vfslocked); 1833 return (EEXIST); 1834 } 1835 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1836 NDFREE(&nd, NDF_ONLY_PNBUF); 1837 vput(nd.ni_dvp); 1838 VFS_UNLOCK_GIANT(vfslocked); 1839 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1840 return (error); 1841 goto restart; 1842 } 1843 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE); 1844 NDFREE(&nd, NDF_ONLY_PNBUF); 1845 vput(nd.ni_dvp); 1846 vn_finished_write(mp); 1847 VFS_UNLOCK_GIANT(vfslocked); 1848 return (error); 1849 } 1850 1851 /* 1852 * Delete a name from the filesystem. 1853 */ 1854 #ifndef _SYS_SYSPROTO_H_ 1855 struct unlink_args { 1856 char *path; 1857 }; 1858 #endif 1859 int 1860 sys_unlink(td, uap) 1861 struct thread *td; 1862 struct unlink_args /* { 1863 char *path; 1864 } */ *uap; 1865 { 1866 1867 return (kern_unlink(td, uap->path, UIO_USERSPACE)); 1868 } 1869 1870 #ifndef _SYS_SYSPROTO_H_ 1871 struct unlinkat_args { 1872 int fd; 1873 char *path; 1874 int flag; 1875 }; 1876 #endif 1877 int 1878 sys_unlinkat(struct thread *td, struct unlinkat_args *uap) 1879 { 1880 int flag = uap->flag; 1881 int fd = uap->fd; 1882 char *path = uap->path; 1883 1884 if (flag & ~AT_REMOVEDIR) 1885 return (EINVAL); 1886 1887 if (flag & AT_REMOVEDIR) 1888 return (kern_rmdirat(td, fd, path, UIO_USERSPACE)); 1889 else 1890 return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0)); 1891 } 1892 1893 int 1894 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg) 1895 { 1896 1897 return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0)); 1898 } 1899 1900 int 1901 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1902 ino_t oldinum) 1903 { 1904 struct mount *mp; 1905 struct vnode *vp; 1906 int error; 1907 struct nameidata nd; 1908 struct stat sb; 1909 int vfslocked; 1910 1911 restart: 1912 bwillwrite(); 1913 NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1, 1914 pathseg, path, fd, td); 1915 if ((error = namei(&nd)) != 0) 1916 return (error == EINVAL ? EPERM : error); 1917 vfslocked = NDHASGIANT(&nd); 1918 vp = nd.ni_vp; 1919 if (vp->v_type == VDIR && oldinum == 0) { 1920 error = EPERM; /* POSIX */ 1921 } else if (oldinum != 0 && 1922 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) && 1923 sb.st_ino != oldinum) { 1924 error = EIDRM; /* Identifier removed */ 1925 } else { 1926 /* 1927 * The root of a mounted filesystem cannot be deleted. 1928 * 1929 * XXX: can this only be a VDIR case? 1930 */ 1931 if (vp->v_vflag & VV_ROOT) 1932 error = EBUSY; 1933 } 1934 if (error == 0) { 1935 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1936 NDFREE(&nd, NDF_ONLY_PNBUF); 1937 vput(nd.ni_dvp); 1938 if (vp == nd.ni_dvp) 1939 vrele(vp); 1940 else 1941 vput(vp); 1942 VFS_UNLOCK_GIANT(vfslocked); 1943 if ((error = vn_start_write(NULL, &mp, 1944 V_XSLEEP | PCATCH)) != 0) 1945 return (error); 1946 goto restart; 1947 } 1948 #ifdef MAC 1949 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 1950 &nd.ni_cnd); 1951 if (error) 1952 goto out; 1953 #endif 1954 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd); 1955 #ifdef MAC 1956 out: 1957 #endif 1958 vn_finished_write(mp); 1959 } 1960 NDFREE(&nd, NDF_ONLY_PNBUF); 1961 vput(nd.ni_dvp); 1962 if (vp == nd.ni_dvp) 1963 vrele(vp); 1964 else 1965 vput(vp); 1966 VFS_UNLOCK_GIANT(vfslocked); 1967 return (error); 1968 } 1969 1970 /* 1971 * Reposition read/write file offset. 1972 */ 1973 #ifndef _SYS_SYSPROTO_H_ 1974 struct lseek_args { 1975 int fd; 1976 int pad; 1977 off_t offset; 1978 int whence; 1979 }; 1980 #endif 1981 int 1982 sys_lseek(td, uap) 1983 struct thread *td; 1984 register struct lseek_args /* { 1985 int fd; 1986 int pad; 1987 off_t offset; 1988 int whence; 1989 } */ *uap; 1990 { 1991 struct ucred *cred = td->td_ucred; 1992 struct file *fp; 1993 struct vnode *vp; 1994 struct vattr vattr; 1995 off_t offset, size; 1996 int error, noneg; 1997 int vfslocked; 1998 1999 AUDIT_ARG_FD(uap->fd); 2000 if ((error = fget(td, uap->fd, CAP_SEEK, &fp)) != 0) 2001 return (error); 2002 if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) { 2003 fdrop(fp, td); 2004 return (ESPIPE); 2005 } 2006 vp = fp->f_vnode; 2007 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 2008 noneg = (vp->v_type != VCHR); 2009 offset = uap->offset; 2010 switch (uap->whence) { 2011 case L_INCR: 2012 if (noneg && 2013 (fp->f_offset < 0 || 2014 (offset > 0 && fp->f_offset > OFF_MAX - offset))) { 2015 error = EOVERFLOW; 2016 break; 2017 } 2018 offset += fp->f_offset; 2019 break; 2020 case L_XTND: 2021 vn_lock(vp, LK_SHARED | LK_RETRY); 2022 error = VOP_GETATTR(vp, &vattr, cred); 2023 VOP_UNLOCK(vp, 0); 2024 if (error) 2025 break; 2026 2027 /* 2028 * If the file references a disk device, then fetch 2029 * the media size and use that to determine the ending 2030 * offset. 2031 */ 2032 if (vattr.va_size == 0 && vp->v_type == VCHR && 2033 fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0) 2034 vattr.va_size = size; 2035 if (noneg && 2036 (vattr.va_size > OFF_MAX || 2037 (offset > 0 && vattr.va_size > OFF_MAX - offset))) { 2038 error = EOVERFLOW; 2039 break; 2040 } 2041 offset += vattr.va_size; 2042 break; 2043 case L_SET: 2044 break; 2045 case SEEK_DATA: 2046 error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td); 2047 break; 2048 case SEEK_HOLE: 2049 error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td); 2050 break; 2051 default: 2052 error = EINVAL; 2053 } 2054 if (error == 0 && noneg && offset < 0) 2055 error = EINVAL; 2056 if (error != 0) 2057 goto drop; 2058 fp->f_offset = offset; 2059 VFS_KNOTE_UNLOCKED(vp, 0); 2060 *(off_t *)(td->td_retval) = fp->f_offset; 2061 drop: 2062 fdrop(fp, td); 2063 VFS_UNLOCK_GIANT(vfslocked); 2064 return (error); 2065 } 2066 2067 #if defined(COMPAT_43) 2068 /* 2069 * Reposition read/write file offset. 2070 */ 2071 #ifndef _SYS_SYSPROTO_H_ 2072 struct olseek_args { 2073 int fd; 2074 long offset; 2075 int whence; 2076 }; 2077 #endif 2078 int 2079 olseek(td, uap) 2080 struct thread *td; 2081 register struct olseek_args /* { 2082 int fd; 2083 long offset; 2084 int whence; 2085 } */ *uap; 2086 { 2087 struct lseek_args /* { 2088 int fd; 2089 int pad; 2090 off_t offset; 2091 int whence; 2092 } */ nuap; 2093 2094 nuap.fd = uap->fd; 2095 nuap.offset = uap->offset; 2096 nuap.whence = uap->whence; 2097 return (sys_lseek(td, &nuap)); 2098 } 2099 #endif /* COMPAT_43 */ 2100 2101 /* Version with the 'pad' argument */ 2102 int 2103 freebsd6_lseek(td, uap) 2104 struct thread *td; 2105 register struct freebsd6_lseek_args *uap; 2106 { 2107 struct lseek_args ouap; 2108 2109 ouap.fd = uap->fd; 2110 ouap.offset = uap->offset; 2111 ouap.whence = uap->whence; 2112 return (sys_lseek(td, &ouap)); 2113 } 2114 2115 /* 2116 * Check access permissions using passed credentials. 2117 */ 2118 static int 2119 vn_access(vp, user_flags, cred, td) 2120 struct vnode *vp; 2121 int user_flags; 2122 struct ucred *cred; 2123 struct thread *td; 2124 { 2125 int error; 2126 accmode_t accmode; 2127 2128 /* Flags == 0 means only check for existence. */ 2129 error = 0; 2130 if (user_flags) { 2131 accmode = 0; 2132 if (user_flags & R_OK) 2133 accmode |= VREAD; 2134 if (user_flags & W_OK) 2135 accmode |= VWRITE; 2136 if (user_flags & X_OK) 2137 accmode |= VEXEC; 2138 #ifdef MAC 2139 error = mac_vnode_check_access(cred, vp, accmode); 2140 if (error) 2141 return (error); 2142 #endif 2143 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 2144 error = VOP_ACCESS(vp, accmode, cred, td); 2145 } 2146 return (error); 2147 } 2148 2149 /* 2150 * Check access permissions using "real" credentials. 2151 */ 2152 #ifndef _SYS_SYSPROTO_H_ 2153 struct access_args { 2154 char *path; 2155 int flags; 2156 }; 2157 #endif 2158 int 2159 sys_access(td, uap) 2160 struct thread *td; 2161 register struct access_args /* { 2162 char *path; 2163 int flags; 2164 } */ *uap; 2165 { 2166 2167 return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags)); 2168 } 2169 2170 #ifndef _SYS_SYSPROTO_H_ 2171 struct faccessat_args { 2172 int dirfd; 2173 char *path; 2174 int mode; 2175 int flag; 2176 } 2177 #endif 2178 int 2179 sys_faccessat(struct thread *td, struct faccessat_args *uap) 2180 { 2181 2182 if (uap->flag & ~AT_EACCESS) 2183 return (EINVAL); 2184 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 2185 uap->mode)); 2186 } 2187 2188 int 2189 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode) 2190 { 2191 2192 return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode)); 2193 } 2194 2195 int 2196 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 2197 int flags, int mode) 2198 { 2199 struct ucred *cred, *tmpcred; 2200 struct vnode *vp; 2201 struct nameidata nd; 2202 int vfslocked; 2203 int error; 2204 2205 /* 2206 * Create and modify a temporary credential instead of one that 2207 * is potentially shared. 2208 */ 2209 if (!(flags & AT_EACCESS)) { 2210 cred = td->td_ucred; 2211 tmpcred = crdup(cred); 2212 tmpcred->cr_uid = cred->cr_ruid; 2213 tmpcred->cr_groups[0] = cred->cr_rgid; 2214 td->td_ucred = tmpcred; 2215 } else 2216 cred = tmpcred = td->td_ucred; 2217 AUDIT_ARG_VALUE(mode); 2218 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE | 2219 AUDITVNODE1, pathseg, path, fd, CAP_FSTAT, td); 2220 if ((error = namei(&nd)) != 0) 2221 goto out1; 2222 vfslocked = NDHASGIANT(&nd); 2223 vp = nd.ni_vp; 2224 2225 error = vn_access(vp, mode, tmpcred, td); 2226 NDFREE(&nd, NDF_ONLY_PNBUF); 2227 vput(vp); 2228 VFS_UNLOCK_GIANT(vfslocked); 2229 out1: 2230 if (!(flags & AT_EACCESS)) { 2231 td->td_ucred = cred; 2232 crfree(tmpcred); 2233 } 2234 return (error); 2235 } 2236 2237 /* 2238 * Check access permissions using "effective" credentials. 2239 */ 2240 #ifndef _SYS_SYSPROTO_H_ 2241 struct eaccess_args { 2242 char *path; 2243 int flags; 2244 }; 2245 #endif 2246 int 2247 sys_eaccess(td, uap) 2248 struct thread *td; 2249 register struct eaccess_args /* { 2250 char *path; 2251 int flags; 2252 } */ *uap; 2253 { 2254 2255 return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags)); 2256 } 2257 2258 int 2259 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags) 2260 { 2261 2262 return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags)); 2263 } 2264 2265 #if defined(COMPAT_43) 2266 /* 2267 * Get file status; this version follows links. 2268 */ 2269 #ifndef _SYS_SYSPROTO_H_ 2270 struct ostat_args { 2271 char *path; 2272 struct ostat *ub; 2273 }; 2274 #endif 2275 int 2276 ostat(td, uap) 2277 struct thread *td; 2278 register struct ostat_args /* { 2279 char *path; 2280 struct ostat *ub; 2281 } */ *uap; 2282 { 2283 struct stat sb; 2284 struct ostat osb; 2285 int error; 2286 2287 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb); 2288 if (error) 2289 return (error); 2290 cvtstat(&sb, &osb); 2291 error = copyout(&osb, uap->ub, sizeof (osb)); 2292 return (error); 2293 } 2294 2295 /* 2296 * Get file status; this version does not follow links. 2297 */ 2298 #ifndef _SYS_SYSPROTO_H_ 2299 struct olstat_args { 2300 char *path; 2301 struct ostat *ub; 2302 }; 2303 #endif 2304 int 2305 olstat(td, uap) 2306 struct thread *td; 2307 register struct olstat_args /* { 2308 char *path; 2309 struct ostat *ub; 2310 } */ *uap; 2311 { 2312 struct stat sb; 2313 struct ostat osb; 2314 int error; 2315 2316 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb); 2317 if (error) 2318 return (error); 2319 cvtstat(&sb, &osb); 2320 error = copyout(&osb, uap->ub, sizeof (osb)); 2321 return (error); 2322 } 2323 2324 /* 2325 * Convert from an old to a new stat structure. 2326 */ 2327 void 2328 cvtstat(st, ost) 2329 struct stat *st; 2330 struct ostat *ost; 2331 { 2332 2333 ost->st_dev = st->st_dev; 2334 ost->st_ino = st->st_ino; 2335 ost->st_mode = st->st_mode; 2336 ost->st_nlink = st->st_nlink; 2337 ost->st_uid = st->st_uid; 2338 ost->st_gid = st->st_gid; 2339 ost->st_rdev = st->st_rdev; 2340 if (st->st_size < (quad_t)1 << 32) 2341 ost->st_size = st->st_size; 2342 else 2343 ost->st_size = -2; 2344 ost->st_atim = st->st_atim; 2345 ost->st_mtim = st->st_mtim; 2346 ost->st_ctim = st->st_ctim; 2347 ost->st_blksize = st->st_blksize; 2348 ost->st_blocks = st->st_blocks; 2349 ost->st_flags = st->st_flags; 2350 ost->st_gen = st->st_gen; 2351 } 2352 #endif /* COMPAT_43 */ 2353 2354 /* 2355 * Get file status; this version follows links. 2356 */ 2357 #ifndef _SYS_SYSPROTO_H_ 2358 struct stat_args { 2359 char *path; 2360 struct stat *ub; 2361 }; 2362 #endif 2363 int 2364 sys_stat(td, uap) 2365 struct thread *td; 2366 register struct stat_args /* { 2367 char *path; 2368 struct stat *ub; 2369 } */ *uap; 2370 { 2371 struct stat sb; 2372 int error; 2373 2374 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb); 2375 if (error == 0) 2376 error = copyout(&sb, uap->ub, sizeof (sb)); 2377 return (error); 2378 } 2379 2380 #ifndef _SYS_SYSPROTO_H_ 2381 struct fstatat_args { 2382 int fd; 2383 char *path; 2384 struct stat *buf; 2385 int flag; 2386 } 2387 #endif 2388 int 2389 sys_fstatat(struct thread *td, struct fstatat_args *uap) 2390 { 2391 struct stat sb; 2392 int error; 2393 2394 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2395 UIO_USERSPACE, &sb); 2396 if (error == 0) 2397 error = copyout(&sb, uap->buf, sizeof (sb)); 2398 return (error); 2399 } 2400 2401 int 2402 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp) 2403 { 2404 2405 return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp)); 2406 } 2407 2408 int 2409 kern_statat(struct thread *td, int flag, int fd, char *path, 2410 enum uio_seg pathseg, struct stat *sbp) 2411 { 2412 2413 return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL)); 2414 } 2415 2416 int 2417 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path, 2418 enum uio_seg pathseg, struct stat *sbp, 2419 void (*hook)(struct vnode *vp, struct stat *sbp)) 2420 { 2421 struct nameidata nd; 2422 struct stat sb; 2423 int error, vfslocked; 2424 2425 if (flag & ~AT_SYMLINK_NOFOLLOW) 2426 return (EINVAL); 2427 2428 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : 2429 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg, 2430 path, fd, CAP_FSTAT, td); 2431 2432 if ((error = namei(&nd)) != 0) 2433 return (error); 2434 vfslocked = NDHASGIANT(&nd); 2435 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td); 2436 if (!error) { 2437 SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0); 2438 if (S_ISREG(sb.st_mode)) 2439 SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0); 2440 if (__predict_false(hook != NULL)) 2441 hook(nd.ni_vp, &sb); 2442 } 2443 NDFREE(&nd, NDF_ONLY_PNBUF); 2444 vput(nd.ni_vp); 2445 VFS_UNLOCK_GIANT(vfslocked); 2446 if (error) 2447 return (error); 2448 *sbp = sb; 2449 #ifdef KTRACE 2450 if (KTRPOINT(td, KTR_STRUCT)) 2451 ktrstat(&sb); 2452 #endif 2453 return (0); 2454 } 2455 2456 /* 2457 * Get file status; this version does not follow links. 2458 */ 2459 #ifndef _SYS_SYSPROTO_H_ 2460 struct lstat_args { 2461 char *path; 2462 struct stat *ub; 2463 }; 2464 #endif 2465 int 2466 sys_lstat(td, uap) 2467 struct thread *td; 2468 register struct lstat_args /* { 2469 char *path; 2470 struct stat *ub; 2471 } */ *uap; 2472 { 2473 struct stat sb; 2474 int error; 2475 2476 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb); 2477 if (error == 0) 2478 error = copyout(&sb, uap->ub, sizeof (sb)); 2479 return (error); 2480 } 2481 2482 int 2483 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp) 2484 { 2485 2486 return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg, 2487 sbp)); 2488 } 2489 2490 /* 2491 * Implementation of the NetBSD [l]stat() functions. 2492 */ 2493 void 2494 cvtnstat(sb, nsb) 2495 struct stat *sb; 2496 struct nstat *nsb; 2497 { 2498 bzero(nsb, sizeof *nsb); 2499 nsb->st_dev = sb->st_dev; 2500 nsb->st_ino = sb->st_ino; 2501 nsb->st_mode = sb->st_mode; 2502 nsb->st_nlink = sb->st_nlink; 2503 nsb->st_uid = sb->st_uid; 2504 nsb->st_gid = sb->st_gid; 2505 nsb->st_rdev = sb->st_rdev; 2506 nsb->st_atim = sb->st_atim; 2507 nsb->st_mtim = sb->st_mtim; 2508 nsb->st_ctim = sb->st_ctim; 2509 nsb->st_size = sb->st_size; 2510 nsb->st_blocks = sb->st_blocks; 2511 nsb->st_blksize = sb->st_blksize; 2512 nsb->st_flags = sb->st_flags; 2513 nsb->st_gen = sb->st_gen; 2514 nsb->st_birthtim = sb->st_birthtim; 2515 } 2516 2517 #ifndef _SYS_SYSPROTO_H_ 2518 struct nstat_args { 2519 char *path; 2520 struct nstat *ub; 2521 }; 2522 #endif 2523 int 2524 sys_nstat(td, uap) 2525 struct thread *td; 2526 register struct nstat_args /* { 2527 char *path; 2528 struct nstat *ub; 2529 } */ *uap; 2530 { 2531 struct stat sb; 2532 struct nstat nsb; 2533 int error; 2534 2535 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb); 2536 if (error) 2537 return (error); 2538 cvtnstat(&sb, &nsb); 2539 error = copyout(&nsb, uap->ub, sizeof (nsb)); 2540 return (error); 2541 } 2542 2543 /* 2544 * NetBSD lstat. Get file status; this version does not follow links. 2545 */ 2546 #ifndef _SYS_SYSPROTO_H_ 2547 struct lstat_args { 2548 char *path; 2549 struct stat *ub; 2550 }; 2551 #endif 2552 int 2553 sys_nlstat(td, uap) 2554 struct thread *td; 2555 register struct nlstat_args /* { 2556 char *path; 2557 struct nstat *ub; 2558 } */ *uap; 2559 { 2560 struct stat sb; 2561 struct nstat nsb; 2562 int error; 2563 2564 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb); 2565 if (error) 2566 return (error); 2567 cvtnstat(&sb, &nsb); 2568 error = copyout(&nsb, uap->ub, sizeof (nsb)); 2569 return (error); 2570 } 2571 2572 /* 2573 * Get configurable pathname variables. 2574 */ 2575 #ifndef _SYS_SYSPROTO_H_ 2576 struct pathconf_args { 2577 char *path; 2578 int name; 2579 }; 2580 #endif 2581 int 2582 sys_pathconf(td, uap) 2583 struct thread *td; 2584 register struct pathconf_args /* { 2585 char *path; 2586 int name; 2587 } */ *uap; 2588 { 2589 2590 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW)); 2591 } 2592 2593 #ifndef _SYS_SYSPROTO_H_ 2594 struct lpathconf_args { 2595 char *path; 2596 int name; 2597 }; 2598 #endif 2599 int 2600 sys_lpathconf(td, uap) 2601 struct thread *td; 2602 register struct lpathconf_args /* { 2603 char *path; 2604 int name; 2605 } */ *uap; 2606 { 2607 2608 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW)); 2609 } 2610 2611 int 2612 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name, 2613 u_long flags) 2614 { 2615 struct nameidata nd; 2616 int error, vfslocked; 2617 2618 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1 | 2619 flags, pathseg, path, td); 2620 if ((error = namei(&nd)) != 0) 2621 return (error); 2622 vfslocked = NDHASGIANT(&nd); 2623 NDFREE(&nd, NDF_ONLY_PNBUF); 2624 2625 /* If asynchronous I/O is available, it works for all files. */ 2626 if (name == _PC_ASYNC_IO) 2627 td->td_retval[0] = async_io_version; 2628 else 2629 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval); 2630 vput(nd.ni_vp); 2631 VFS_UNLOCK_GIANT(vfslocked); 2632 return (error); 2633 } 2634 2635 /* 2636 * Return target name of a symbolic link. 2637 */ 2638 #ifndef _SYS_SYSPROTO_H_ 2639 struct readlink_args { 2640 char *path; 2641 char *buf; 2642 size_t count; 2643 }; 2644 #endif 2645 int 2646 sys_readlink(td, uap) 2647 struct thread *td; 2648 register struct readlink_args /* { 2649 char *path; 2650 char *buf; 2651 size_t count; 2652 } */ *uap; 2653 { 2654 2655 return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf, 2656 UIO_USERSPACE, uap->count)); 2657 } 2658 #ifndef _SYS_SYSPROTO_H_ 2659 struct readlinkat_args { 2660 int fd; 2661 char *path; 2662 char *buf; 2663 size_t bufsize; 2664 }; 2665 #endif 2666 int 2667 sys_readlinkat(struct thread *td, struct readlinkat_args *uap) 2668 { 2669 2670 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE, 2671 uap->buf, UIO_USERSPACE, uap->bufsize)); 2672 } 2673 2674 int 2675 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf, 2676 enum uio_seg bufseg, size_t count) 2677 { 2678 2679 return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg, 2680 count)); 2681 } 2682 2683 int 2684 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 2685 char *buf, enum uio_seg bufseg, size_t count) 2686 { 2687 struct vnode *vp; 2688 struct iovec aiov; 2689 struct uio auio; 2690 int error; 2691 struct nameidata nd; 2692 int vfslocked; 2693 2694 if (count > IOSIZE_MAX) 2695 return (EINVAL); 2696 2697 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE | 2698 AUDITVNODE1, pathseg, path, fd, td); 2699 2700 if ((error = namei(&nd)) != 0) 2701 return (error); 2702 NDFREE(&nd, NDF_ONLY_PNBUF); 2703 vfslocked = NDHASGIANT(&nd); 2704 vp = nd.ni_vp; 2705 #ifdef MAC 2706 error = mac_vnode_check_readlink(td->td_ucred, vp); 2707 if (error) { 2708 vput(vp); 2709 VFS_UNLOCK_GIANT(vfslocked); 2710 return (error); 2711 } 2712 #endif 2713 if (vp->v_type != VLNK) 2714 error = EINVAL; 2715 else { 2716 aiov.iov_base = buf; 2717 aiov.iov_len = count; 2718 auio.uio_iov = &aiov; 2719 auio.uio_iovcnt = 1; 2720 auio.uio_offset = 0; 2721 auio.uio_rw = UIO_READ; 2722 auio.uio_segflg = bufseg; 2723 auio.uio_td = td; 2724 auio.uio_resid = count; 2725 error = VOP_READLINK(vp, &auio, td->td_ucred); 2726 } 2727 vput(vp); 2728 VFS_UNLOCK_GIANT(vfslocked); 2729 td->td_retval[0] = count - auio.uio_resid; 2730 return (error); 2731 } 2732 2733 /* 2734 * Common implementation code for chflags() and fchflags(). 2735 */ 2736 static int 2737 setfflags(td, vp, flags) 2738 struct thread *td; 2739 struct vnode *vp; 2740 int flags; 2741 { 2742 int error; 2743 struct mount *mp; 2744 struct vattr vattr; 2745 2746 /* We can't support the value matching VNOVAL. */ 2747 if (flags == VNOVAL) 2748 return (EOPNOTSUPP); 2749 2750 /* 2751 * Prevent non-root users from setting flags on devices. When 2752 * a device is reused, users can retain ownership of the device 2753 * if they are allowed to set flags and programs assume that 2754 * chown can't fail when done as root. 2755 */ 2756 if (vp->v_type == VCHR || vp->v_type == VBLK) { 2757 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV); 2758 if (error) 2759 return (error); 2760 } 2761 2762 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2763 return (error); 2764 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2765 VATTR_NULL(&vattr); 2766 vattr.va_flags = flags; 2767 #ifdef MAC 2768 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags); 2769 if (error == 0) 2770 #endif 2771 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 2772 VOP_UNLOCK(vp, 0); 2773 vn_finished_write(mp); 2774 return (error); 2775 } 2776 2777 /* 2778 * Change flags of a file given a path name. 2779 */ 2780 #ifndef _SYS_SYSPROTO_H_ 2781 struct chflags_args { 2782 char *path; 2783 int flags; 2784 }; 2785 #endif 2786 int 2787 sys_chflags(td, uap) 2788 struct thread *td; 2789 register struct chflags_args /* { 2790 char *path; 2791 int flags; 2792 } */ *uap; 2793 { 2794 int error; 2795 struct nameidata nd; 2796 int vfslocked; 2797 2798 AUDIT_ARG_FFLAGS(uap->flags); 2799 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE, 2800 uap->path, td); 2801 if ((error = namei(&nd)) != 0) 2802 return (error); 2803 NDFREE(&nd, NDF_ONLY_PNBUF); 2804 vfslocked = NDHASGIANT(&nd); 2805 error = setfflags(td, nd.ni_vp, uap->flags); 2806 vrele(nd.ni_vp); 2807 VFS_UNLOCK_GIANT(vfslocked); 2808 return (error); 2809 } 2810 2811 /* 2812 * Same as chflags() but doesn't follow symlinks. 2813 */ 2814 int 2815 sys_lchflags(td, uap) 2816 struct thread *td; 2817 register struct lchflags_args /* { 2818 char *path; 2819 int flags; 2820 } */ *uap; 2821 { 2822 int error; 2823 struct nameidata nd; 2824 int vfslocked; 2825 2826 AUDIT_ARG_FFLAGS(uap->flags); 2827 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE, 2828 uap->path, td); 2829 if ((error = namei(&nd)) != 0) 2830 return (error); 2831 vfslocked = NDHASGIANT(&nd); 2832 NDFREE(&nd, NDF_ONLY_PNBUF); 2833 error = setfflags(td, nd.ni_vp, uap->flags); 2834 vrele(nd.ni_vp); 2835 VFS_UNLOCK_GIANT(vfslocked); 2836 return (error); 2837 } 2838 2839 /* 2840 * Change flags of a file given a file descriptor. 2841 */ 2842 #ifndef _SYS_SYSPROTO_H_ 2843 struct fchflags_args { 2844 int fd; 2845 int flags; 2846 }; 2847 #endif 2848 int 2849 sys_fchflags(td, uap) 2850 struct thread *td; 2851 register struct fchflags_args /* { 2852 int fd; 2853 int flags; 2854 } */ *uap; 2855 { 2856 struct file *fp; 2857 int vfslocked; 2858 int error; 2859 2860 AUDIT_ARG_FD(uap->fd); 2861 AUDIT_ARG_FFLAGS(uap->flags); 2862 if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FCHFLAGS, 2863 &fp)) != 0) 2864 return (error); 2865 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount); 2866 #ifdef AUDIT 2867 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 2868 AUDIT_ARG_VNODE1(fp->f_vnode); 2869 VOP_UNLOCK(fp->f_vnode, 0); 2870 #endif 2871 error = setfflags(td, fp->f_vnode, uap->flags); 2872 VFS_UNLOCK_GIANT(vfslocked); 2873 fdrop(fp, td); 2874 return (error); 2875 } 2876 2877 /* 2878 * Common implementation code for chmod(), lchmod() and fchmod(). 2879 */ 2880 int 2881 setfmode(td, cred, vp, mode) 2882 struct thread *td; 2883 struct ucred *cred; 2884 struct vnode *vp; 2885 int mode; 2886 { 2887 int error; 2888 struct mount *mp; 2889 struct vattr vattr; 2890 2891 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2892 return (error); 2893 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2894 VATTR_NULL(&vattr); 2895 vattr.va_mode = mode & ALLPERMS; 2896 #ifdef MAC 2897 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode); 2898 if (error == 0) 2899 #endif 2900 error = VOP_SETATTR(vp, &vattr, cred); 2901 VOP_UNLOCK(vp, 0); 2902 vn_finished_write(mp); 2903 return (error); 2904 } 2905 2906 /* 2907 * Change mode of a file given path name. 2908 */ 2909 #ifndef _SYS_SYSPROTO_H_ 2910 struct chmod_args { 2911 char *path; 2912 int mode; 2913 }; 2914 #endif 2915 int 2916 sys_chmod(td, uap) 2917 struct thread *td; 2918 register struct chmod_args /* { 2919 char *path; 2920 int mode; 2921 } */ *uap; 2922 { 2923 2924 return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode)); 2925 } 2926 2927 #ifndef _SYS_SYSPROTO_H_ 2928 struct fchmodat_args { 2929 int dirfd; 2930 char *path; 2931 mode_t mode; 2932 int flag; 2933 } 2934 #endif 2935 int 2936 sys_fchmodat(struct thread *td, struct fchmodat_args *uap) 2937 { 2938 int flag = uap->flag; 2939 int fd = uap->fd; 2940 char *path = uap->path; 2941 mode_t mode = uap->mode; 2942 2943 if (flag & ~AT_SYMLINK_NOFOLLOW) 2944 return (EINVAL); 2945 2946 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag)); 2947 } 2948 2949 int 2950 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode) 2951 { 2952 2953 return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0)); 2954 } 2955 2956 /* 2957 * Change mode of a file given path name (don't follow links.) 2958 */ 2959 #ifndef _SYS_SYSPROTO_H_ 2960 struct lchmod_args { 2961 char *path; 2962 int mode; 2963 }; 2964 #endif 2965 int 2966 sys_lchmod(td, uap) 2967 struct thread *td; 2968 register struct lchmod_args /* { 2969 char *path; 2970 int mode; 2971 } */ *uap; 2972 { 2973 2974 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2975 uap->mode, AT_SYMLINK_NOFOLLOW)); 2976 } 2977 2978 2979 int 2980 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 2981 mode_t mode, int flag) 2982 { 2983 int error; 2984 struct nameidata nd; 2985 int vfslocked; 2986 int follow; 2987 2988 AUDIT_ARG_MODE(mode); 2989 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 2990 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, 2991 path, fd, CAP_FCHMOD, td); 2992 if ((error = namei(&nd)) != 0) 2993 return (error); 2994 vfslocked = NDHASGIANT(&nd); 2995 NDFREE(&nd, NDF_ONLY_PNBUF); 2996 error = setfmode(td, td->td_ucred, nd.ni_vp, mode); 2997 vrele(nd.ni_vp); 2998 VFS_UNLOCK_GIANT(vfslocked); 2999 return (error); 3000 } 3001 3002 /* 3003 * Change mode of a file given a file descriptor. 3004 */ 3005 #ifndef _SYS_SYSPROTO_H_ 3006 struct fchmod_args { 3007 int fd; 3008 int mode; 3009 }; 3010 #endif 3011 int 3012 sys_fchmod(struct thread *td, struct fchmod_args *uap) 3013 { 3014 struct file *fp; 3015 int error; 3016 3017 AUDIT_ARG_FD(uap->fd); 3018 AUDIT_ARG_MODE(uap->mode); 3019 3020 error = fget(td, uap->fd, CAP_FCHMOD, &fp); 3021 if (error != 0) 3022 return (error); 3023 error = fo_chmod(fp, uap->mode, td->td_ucred, td); 3024 fdrop(fp, td); 3025 return (error); 3026 } 3027 3028 /* 3029 * Common implementation for chown(), lchown(), and fchown() 3030 */ 3031 int 3032 setfown(td, cred, vp, uid, gid) 3033 struct thread *td; 3034 struct ucred *cred; 3035 struct vnode *vp; 3036 uid_t uid; 3037 gid_t gid; 3038 { 3039 int error; 3040 struct mount *mp; 3041 struct vattr vattr; 3042 3043 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 3044 return (error); 3045 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3046 VATTR_NULL(&vattr); 3047 vattr.va_uid = uid; 3048 vattr.va_gid = gid; 3049 #ifdef MAC 3050 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid, 3051 vattr.va_gid); 3052 if (error == 0) 3053 #endif 3054 error = VOP_SETATTR(vp, &vattr, cred); 3055 VOP_UNLOCK(vp, 0); 3056 vn_finished_write(mp); 3057 return (error); 3058 } 3059 3060 /* 3061 * Set ownership given a path name. 3062 */ 3063 #ifndef _SYS_SYSPROTO_H_ 3064 struct chown_args { 3065 char *path; 3066 int uid; 3067 int gid; 3068 }; 3069 #endif 3070 int 3071 sys_chown(td, uap) 3072 struct thread *td; 3073 register struct chown_args /* { 3074 char *path; 3075 int uid; 3076 int gid; 3077 } */ *uap; 3078 { 3079 3080 return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid)); 3081 } 3082 3083 #ifndef _SYS_SYSPROTO_H_ 3084 struct fchownat_args { 3085 int fd; 3086 const char * path; 3087 uid_t uid; 3088 gid_t gid; 3089 int flag; 3090 }; 3091 #endif 3092 int 3093 sys_fchownat(struct thread *td, struct fchownat_args *uap) 3094 { 3095 int flag; 3096 3097 flag = uap->flag; 3098 if (flag & ~AT_SYMLINK_NOFOLLOW) 3099 return (EINVAL); 3100 3101 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid, 3102 uap->gid, uap->flag)); 3103 } 3104 3105 int 3106 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid, 3107 int gid) 3108 { 3109 3110 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0)); 3111 } 3112 3113 int 3114 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 3115 int uid, int gid, int flag) 3116 { 3117 struct nameidata nd; 3118 int error, vfslocked, follow; 3119 3120 AUDIT_ARG_OWNER(uid, gid); 3121 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3122 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, 3123 path, fd, CAP_FCHOWN, td); 3124 3125 if ((error = namei(&nd)) != 0) 3126 return (error); 3127 vfslocked = NDHASGIANT(&nd); 3128 NDFREE(&nd, NDF_ONLY_PNBUF); 3129 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid); 3130 vrele(nd.ni_vp); 3131 VFS_UNLOCK_GIANT(vfslocked); 3132 return (error); 3133 } 3134 3135 /* 3136 * Set ownership given a path name, do not cross symlinks. 3137 */ 3138 #ifndef _SYS_SYSPROTO_H_ 3139 struct lchown_args { 3140 char *path; 3141 int uid; 3142 int gid; 3143 }; 3144 #endif 3145 int 3146 sys_lchown(td, uap) 3147 struct thread *td; 3148 register struct lchown_args /* { 3149 char *path; 3150 int uid; 3151 int gid; 3152 } */ *uap; 3153 { 3154 3155 return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid)); 3156 } 3157 3158 int 3159 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid, 3160 int gid) 3161 { 3162 3163 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 3164 AT_SYMLINK_NOFOLLOW)); 3165 } 3166 3167 /* 3168 * Set ownership given a file descriptor. 3169 */ 3170 #ifndef _SYS_SYSPROTO_H_ 3171 struct fchown_args { 3172 int fd; 3173 int uid; 3174 int gid; 3175 }; 3176 #endif 3177 int 3178 sys_fchown(td, uap) 3179 struct thread *td; 3180 register struct fchown_args /* { 3181 int fd; 3182 int uid; 3183 int gid; 3184 } */ *uap; 3185 { 3186 struct file *fp; 3187 int error; 3188 3189 AUDIT_ARG_FD(uap->fd); 3190 AUDIT_ARG_OWNER(uap->uid, uap->gid); 3191 error = fget(td, uap->fd, CAP_FCHOWN, &fp); 3192 if (error != 0) 3193 return (error); 3194 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td); 3195 fdrop(fp, td); 3196 return (error); 3197 } 3198 3199 /* 3200 * Common implementation code for utimes(), lutimes(), and futimes(). 3201 */ 3202 static int 3203 getutimes(usrtvp, tvpseg, tsp) 3204 const struct timeval *usrtvp; 3205 enum uio_seg tvpseg; 3206 struct timespec *tsp; 3207 { 3208 struct timeval tv[2]; 3209 const struct timeval *tvp; 3210 int error; 3211 3212 if (usrtvp == NULL) { 3213 vfs_timestamp(&tsp[0]); 3214 tsp[1] = tsp[0]; 3215 } else { 3216 if (tvpseg == UIO_SYSSPACE) { 3217 tvp = usrtvp; 3218 } else { 3219 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0) 3220 return (error); 3221 tvp = tv; 3222 } 3223 3224 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 || 3225 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000) 3226 return (EINVAL); 3227 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]); 3228 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]); 3229 } 3230 return (0); 3231 } 3232 3233 /* 3234 * Common implementation code for utimes(), lutimes(), and futimes(). 3235 */ 3236 static int 3237 setutimes(td, vp, ts, numtimes, nullflag) 3238 struct thread *td; 3239 struct vnode *vp; 3240 const struct timespec *ts; 3241 int numtimes; 3242 int nullflag; 3243 { 3244 int error, setbirthtime; 3245 struct mount *mp; 3246 struct vattr vattr; 3247 3248 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 3249 return (error); 3250 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3251 setbirthtime = 0; 3252 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) && 3253 timespeccmp(&ts[1], &vattr.va_birthtime, < )) 3254 setbirthtime = 1; 3255 VATTR_NULL(&vattr); 3256 vattr.va_atime = ts[0]; 3257 vattr.va_mtime = ts[1]; 3258 if (setbirthtime) 3259 vattr.va_birthtime = ts[1]; 3260 if (numtimes > 2) 3261 vattr.va_birthtime = ts[2]; 3262 if (nullflag) 3263 vattr.va_vaflags |= VA_UTIMES_NULL; 3264 #ifdef MAC 3265 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime, 3266 vattr.va_mtime); 3267 #endif 3268 if (error == 0) 3269 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3270 VOP_UNLOCK(vp, 0); 3271 vn_finished_write(mp); 3272 return (error); 3273 } 3274 3275 /* 3276 * Set the access and modification times of a file. 3277 */ 3278 #ifndef _SYS_SYSPROTO_H_ 3279 struct utimes_args { 3280 char *path; 3281 struct timeval *tptr; 3282 }; 3283 #endif 3284 int 3285 sys_utimes(td, uap) 3286 struct thread *td; 3287 register struct utimes_args /* { 3288 char *path; 3289 struct timeval *tptr; 3290 } */ *uap; 3291 { 3292 3293 return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr, 3294 UIO_USERSPACE)); 3295 } 3296 3297 #ifndef _SYS_SYSPROTO_H_ 3298 struct futimesat_args { 3299 int fd; 3300 const char * path; 3301 const struct timeval * times; 3302 }; 3303 #endif 3304 int 3305 sys_futimesat(struct thread *td, struct futimesat_args *uap) 3306 { 3307 3308 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE, 3309 uap->times, UIO_USERSPACE)); 3310 } 3311 3312 int 3313 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg, 3314 struct timeval *tptr, enum uio_seg tptrseg) 3315 { 3316 3317 return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg)); 3318 } 3319 3320 int 3321 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 3322 struct timeval *tptr, enum uio_seg tptrseg) 3323 { 3324 struct nameidata nd; 3325 struct timespec ts[2]; 3326 int error, vfslocked; 3327 3328 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3329 return (error); 3330 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, 3331 path, fd, CAP_FUTIMES, td); 3332 3333 if ((error = namei(&nd)) != 0) 3334 return (error); 3335 vfslocked = NDHASGIANT(&nd); 3336 NDFREE(&nd, NDF_ONLY_PNBUF); 3337 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3338 vrele(nd.ni_vp); 3339 VFS_UNLOCK_GIANT(vfslocked); 3340 return (error); 3341 } 3342 3343 /* 3344 * Set the access and modification times of a file. 3345 */ 3346 #ifndef _SYS_SYSPROTO_H_ 3347 struct lutimes_args { 3348 char *path; 3349 struct timeval *tptr; 3350 }; 3351 #endif 3352 int 3353 sys_lutimes(td, uap) 3354 struct thread *td; 3355 register struct lutimes_args /* { 3356 char *path; 3357 struct timeval *tptr; 3358 } */ *uap; 3359 { 3360 3361 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr, 3362 UIO_USERSPACE)); 3363 } 3364 3365 int 3366 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg, 3367 struct timeval *tptr, enum uio_seg tptrseg) 3368 { 3369 struct timespec ts[2]; 3370 int error; 3371 struct nameidata nd; 3372 int vfslocked; 3373 3374 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3375 return (error); 3376 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td); 3377 if ((error = namei(&nd)) != 0) 3378 return (error); 3379 vfslocked = NDHASGIANT(&nd); 3380 NDFREE(&nd, NDF_ONLY_PNBUF); 3381 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3382 vrele(nd.ni_vp); 3383 VFS_UNLOCK_GIANT(vfslocked); 3384 return (error); 3385 } 3386 3387 /* 3388 * Set the access and modification times of a file. 3389 */ 3390 #ifndef _SYS_SYSPROTO_H_ 3391 struct futimes_args { 3392 int fd; 3393 struct timeval *tptr; 3394 }; 3395 #endif 3396 int 3397 sys_futimes(td, uap) 3398 struct thread *td; 3399 register struct futimes_args /* { 3400 int fd; 3401 struct timeval *tptr; 3402 } */ *uap; 3403 { 3404 3405 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE)); 3406 } 3407 3408 int 3409 kern_futimes(struct thread *td, int fd, struct timeval *tptr, 3410 enum uio_seg tptrseg) 3411 { 3412 struct timespec ts[2]; 3413 struct file *fp; 3414 int vfslocked; 3415 int error; 3416 3417 AUDIT_ARG_FD(fd); 3418 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3419 return (error); 3420 if ((error = getvnode(td->td_proc->p_fd, fd, CAP_FUTIMES, &fp)) 3421 != 0) 3422 return (error); 3423 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount); 3424 #ifdef AUDIT 3425 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3426 AUDIT_ARG_VNODE1(fp->f_vnode); 3427 VOP_UNLOCK(fp->f_vnode, 0); 3428 #endif 3429 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL); 3430 VFS_UNLOCK_GIANT(vfslocked); 3431 fdrop(fp, td); 3432 return (error); 3433 } 3434 3435 /* 3436 * Truncate a file given its path name. 3437 */ 3438 #ifndef _SYS_SYSPROTO_H_ 3439 struct truncate_args { 3440 char *path; 3441 int pad; 3442 off_t length; 3443 }; 3444 #endif 3445 int 3446 sys_truncate(td, uap) 3447 struct thread *td; 3448 register struct truncate_args /* { 3449 char *path; 3450 int pad; 3451 off_t length; 3452 } */ *uap; 3453 { 3454 3455 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3456 } 3457 3458 int 3459 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length) 3460 { 3461 struct mount *mp; 3462 struct vnode *vp; 3463 struct vattr vattr; 3464 int error; 3465 struct nameidata nd; 3466 int vfslocked; 3467 3468 if (length < 0) 3469 return(EINVAL); 3470 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td); 3471 if ((error = namei(&nd)) != 0) 3472 return (error); 3473 vfslocked = NDHASGIANT(&nd); 3474 vp = nd.ni_vp; 3475 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) { 3476 vrele(vp); 3477 VFS_UNLOCK_GIANT(vfslocked); 3478 return (error); 3479 } 3480 NDFREE(&nd, NDF_ONLY_PNBUF); 3481 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3482 if (vp->v_type == VDIR) 3483 error = EISDIR; 3484 #ifdef MAC 3485 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) { 3486 } 3487 #endif 3488 else if ((error = vn_writechk(vp)) == 0 && 3489 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) { 3490 VATTR_NULL(&vattr); 3491 vattr.va_size = length; 3492 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3493 } 3494 vput(vp); 3495 vn_finished_write(mp); 3496 VFS_UNLOCK_GIANT(vfslocked); 3497 return (error); 3498 } 3499 3500 #if defined(COMPAT_43) 3501 /* 3502 * Truncate a file given its path name. 3503 */ 3504 #ifndef _SYS_SYSPROTO_H_ 3505 struct otruncate_args { 3506 char *path; 3507 long length; 3508 }; 3509 #endif 3510 int 3511 otruncate(td, uap) 3512 struct thread *td; 3513 register struct otruncate_args /* { 3514 char *path; 3515 long length; 3516 } */ *uap; 3517 { 3518 struct truncate_args /* { 3519 char *path; 3520 int pad; 3521 off_t length; 3522 } */ nuap; 3523 3524 nuap.path = uap->path; 3525 nuap.length = uap->length; 3526 return (sys_truncate(td, &nuap)); 3527 } 3528 #endif /* COMPAT_43 */ 3529 3530 /* Versions with the pad argument */ 3531 int 3532 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap) 3533 { 3534 struct truncate_args ouap; 3535 3536 ouap.path = uap->path; 3537 ouap.length = uap->length; 3538 return (sys_truncate(td, &ouap)); 3539 } 3540 3541 int 3542 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap) 3543 { 3544 struct ftruncate_args ouap; 3545 3546 ouap.fd = uap->fd; 3547 ouap.length = uap->length; 3548 return (sys_ftruncate(td, &ouap)); 3549 } 3550 3551 /* 3552 * Sync an open file. 3553 */ 3554 #ifndef _SYS_SYSPROTO_H_ 3555 struct fsync_args { 3556 int fd; 3557 }; 3558 #endif 3559 int 3560 sys_fsync(td, uap) 3561 struct thread *td; 3562 struct fsync_args /* { 3563 int fd; 3564 } */ *uap; 3565 { 3566 struct vnode *vp; 3567 struct mount *mp; 3568 struct file *fp; 3569 int vfslocked; 3570 int error, lock_flags; 3571 3572 AUDIT_ARG_FD(uap->fd); 3573 if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FSYNC, 3574 &fp)) != 0) 3575 return (error); 3576 vp = fp->f_vnode; 3577 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 3578 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 3579 goto drop; 3580 if (MNT_SHARED_WRITES(mp) || 3581 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) { 3582 lock_flags = LK_SHARED; 3583 } else { 3584 lock_flags = LK_EXCLUSIVE; 3585 } 3586 vn_lock(vp, lock_flags | LK_RETRY); 3587 AUDIT_ARG_VNODE1(vp); 3588 if (vp->v_object != NULL) { 3589 VM_OBJECT_LOCK(vp->v_object); 3590 vm_object_page_clean(vp->v_object, 0, 0, 0); 3591 VM_OBJECT_UNLOCK(vp->v_object); 3592 } 3593 error = VOP_FSYNC(vp, MNT_WAIT, td); 3594 3595 VOP_UNLOCK(vp, 0); 3596 vn_finished_write(mp); 3597 drop: 3598 VFS_UNLOCK_GIANT(vfslocked); 3599 fdrop(fp, td); 3600 return (error); 3601 } 3602 3603 /* 3604 * Rename files. Source and destination must either both be directories, or 3605 * both not be directories. If target is a directory, it must be empty. 3606 */ 3607 #ifndef _SYS_SYSPROTO_H_ 3608 struct rename_args { 3609 char *from; 3610 char *to; 3611 }; 3612 #endif 3613 int 3614 sys_rename(td, uap) 3615 struct thread *td; 3616 register struct rename_args /* { 3617 char *from; 3618 char *to; 3619 } */ *uap; 3620 { 3621 3622 return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE)); 3623 } 3624 3625 #ifndef _SYS_SYSPROTO_H_ 3626 struct renameat_args { 3627 int oldfd; 3628 char *old; 3629 int newfd; 3630 char *new; 3631 }; 3632 #endif 3633 int 3634 sys_renameat(struct thread *td, struct renameat_args *uap) 3635 { 3636 3637 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new, 3638 UIO_USERSPACE)); 3639 } 3640 3641 int 3642 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg) 3643 { 3644 3645 return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg)); 3646 } 3647 3648 int 3649 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new, 3650 enum uio_seg pathseg) 3651 { 3652 struct mount *mp = NULL; 3653 struct vnode *tvp, *fvp, *tdvp; 3654 struct nameidata fromnd, tond; 3655 int tvfslocked; 3656 int fvfslocked; 3657 int error; 3658 3659 bwillwrite(); 3660 #ifdef MAC 3661 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | 3662 MPSAFE | AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td); 3663 #else 3664 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE | 3665 AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td); 3666 #endif 3667 3668 if ((error = namei(&fromnd)) != 0) 3669 return (error); 3670 fvfslocked = NDHASGIANT(&fromnd); 3671 tvfslocked = 0; 3672 #ifdef MAC 3673 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp, 3674 fromnd.ni_vp, &fromnd.ni_cnd); 3675 VOP_UNLOCK(fromnd.ni_dvp, 0); 3676 if (fromnd.ni_dvp != fromnd.ni_vp) 3677 VOP_UNLOCK(fromnd.ni_vp, 0); 3678 #endif 3679 fvp = fromnd.ni_vp; 3680 if (error == 0) 3681 error = vn_start_write(fvp, &mp, V_WAIT | PCATCH); 3682 if (error != 0) { 3683 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3684 vrele(fromnd.ni_dvp); 3685 vrele(fvp); 3686 goto out1; 3687 } 3688 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | 3689 SAVESTART | MPSAFE | AUDITVNODE2, pathseg, new, newfd, CAP_CREATE, 3690 td); 3691 if (fromnd.ni_vp->v_type == VDIR) 3692 tond.ni_cnd.cn_flags |= WILLBEDIR; 3693 if ((error = namei(&tond)) != 0) { 3694 /* Translate error code for rename("dir1", "dir2/."). */ 3695 if (error == EISDIR && fvp->v_type == VDIR) 3696 error = EINVAL; 3697 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3698 vrele(fromnd.ni_dvp); 3699 vrele(fvp); 3700 vn_finished_write(mp); 3701 goto out1; 3702 } 3703 tvfslocked = NDHASGIANT(&tond); 3704 tdvp = tond.ni_dvp; 3705 tvp = tond.ni_vp; 3706 if (tvp != NULL) { 3707 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3708 error = ENOTDIR; 3709 goto out; 3710 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3711 error = EISDIR; 3712 goto out; 3713 } 3714 } 3715 if (fvp == tdvp) { 3716 error = EINVAL; 3717 goto out; 3718 } 3719 /* 3720 * If the source is the same as the destination (that is, if they 3721 * are links to the same vnode), then there is nothing to do. 3722 */ 3723 if (fvp == tvp) 3724 error = -1; 3725 #ifdef MAC 3726 else 3727 error = mac_vnode_check_rename_to(td->td_ucred, tdvp, 3728 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd); 3729 #endif 3730 out: 3731 if (!error) { 3732 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3733 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3734 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3735 NDFREE(&tond, NDF_ONLY_PNBUF); 3736 } else { 3737 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3738 NDFREE(&tond, NDF_ONLY_PNBUF); 3739 if (tvp) 3740 vput(tvp); 3741 if (tdvp == tvp) 3742 vrele(tdvp); 3743 else 3744 vput(tdvp); 3745 vrele(fromnd.ni_dvp); 3746 vrele(fvp); 3747 } 3748 vrele(tond.ni_startdir); 3749 vn_finished_write(mp); 3750 out1: 3751 if (fromnd.ni_startdir) 3752 vrele(fromnd.ni_startdir); 3753 VFS_UNLOCK_GIANT(fvfslocked); 3754 VFS_UNLOCK_GIANT(tvfslocked); 3755 if (error == -1) 3756 return (0); 3757 return (error); 3758 } 3759 3760 /* 3761 * Make a directory file. 3762 */ 3763 #ifndef _SYS_SYSPROTO_H_ 3764 struct mkdir_args { 3765 char *path; 3766 int mode; 3767 }; 3768 #endif 3769 int 3770 sys_mkdir(td, uap) 3771 struct thread *td; 3772 register struct mkdir_args /* { 3773 char *path; 3774 int mode; 3775 } */ *uap; 3776 { 3777 3778 return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode)); 3779 } 3780 3781 #ifndef _SYS_SYSPROTO_H_ 3782 struct mkdirat_args { 3783 int fd; 3784 char *path; 3785 mode_t mode; 3786 }; 3787 #endif 3788 int 3789 sys_mkdirat(struct thread *td, struct mkdirat_args *uap) 3790 { 3791 3792 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode)); 3793 } 3794 3795 int 3796 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode) 3797 { 3798 3799 return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode)); 3800 } 3801 3802 int 3803 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg, 3804 int mode) 3805 { 3806 struct mount *mp; 3807 struct vnode *vp; 3808 struct vattr vattr; 3809 int error; 3810 struct nameidata nd; 3811 int vfslocked; 3812 3813 AUDIT_ARG_MODE(mode); 3814 restart: 3815 bwillwrite(); 3816 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | 3817 AUDITVNODE1, segflg, path, fd, CAP_MKDIR, td); 3818 nd.ni_cnd.cn_flags |= WILLBEDIR; 3819 if ((error = namei(&nd)) != 0) 3820 return (error); 3821 vfslocked = NDHASGIANT(&nd); 3822 vp = nd.ni_vp; 3823 if (vp != NULL) { 3824 NDFREE(&nd, NDF_ONLY_PNBUF); 3825 /* 3826 * XXX namei called with LOCKPARENT but not LOCKLEAF has 3827 * the strange behaviour of leaving the vnode unlocked 3828 * if the target is the same vnode as the parent. 3829 */ 3830 if (vp == nd.ni_dvp) 3831 vrele(nd.ni_dvp); 3832 else 3833 vput(nd.ni_dvp); 3834 vrele(vp); 3835 VFS_UNLOCK_GIANT(vfslocked); 3836 return (EEXIST); 3837 } 3838 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3839 NDFREE(&nd, NDF_ONLY_PNBUF); 3840 vput(nd.ni_dvp); 3841 VFS_UNLOCK_GIANT(vfslocked); 3842 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3843 return (error); 3844 goto restart; 3845 } 3846 VATTR_NULL(&vattr); 3847 vattr.va_type = VDIR; 3848 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask; 3849 #ifdef MAC 3850 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 3851 &vattr); 3852 if (error) 3853 goto out; 3854 #endif 3855 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3856 #ifdef MAC 3857 out: 3858 #endif 3859 NDFREE(&nd, NDF_ONLY_PNBUF); 3860 vput(nd.ni_dvp); 3861 if (!error) 3862 vput(nd.ni_vp); 3863 vn_finished_write(mp); 3864 VFS_UNLOCK_GIANT(vfslocked); 3865 return (error); 3866 } 3867 3868 /* 3869 * Remove a directory file. 3870 */ 3871 #ifndef _SYS_SYSPROTO_H_ 3872 struct rmdir_args { 3873 char *path; 3874 }; 3875 #endif 3876 int 3877 sys_rmdir(td, uap) 3878 struct thread *td; 3879 struct rmdir_args /* { 3880 char *path; 3881 } */ *uap; 3882 { 3883 3884 return (kern_rmdir(td, uap->path, UIO_USERSPACE)); 3885 } 3886 3887 int 3888 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg) 3889 { 3890 3891 return (kern_rmdirat(td, AT_FDCWD, path, pathseg)); 3892 } 3893 3894 int 3895 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg) 3896 { 3897 struct mount *mp; 3898 struct vnode *vp; 3899 int error; 3900 struct nameidata nd; 3901 int vfslocked; 3902 3903 restart: 3904 bwillwrite(); 3905 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | 3906 AUDITVNODE1, pathseg, path, fd, CAP_RMDIR, td); 3907 if ((error = namei(&nd)) != 0) 3908 return (error); 3909 vfslocked = NDHASGIANT(&nd); 3910 vp = nd.ni_vp; 3911 if (vp->v_type != VDIR) { 3912 error = ENOTDIR; 3913 goto out; 3914 } 3915 /* 3916 * No rmdir "." please. 3917 */ 3918 if (nd.ni_dvp == vp) { 3919 error = EINVAL; 3920 goto out; 3921 } 3922 /* 3923 * The root of a mounted filesystem cannot be deleted. 3924 */ 3925 if (vp->v_vflag & VV_ROOT) { 3926 error = EBUSY; 3927 goto out; 3928 } 3929 #ifdef MAC 3930 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 3931 &nd.ni_cnd); 3932 if (error) 3933 goto out; 3934 #endif 3935 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3936 NDFREE(&nd, NDF_ONLY_PNBUF); 3937 vput(vp); 3938 if (nd.ni_dvp == vp) 3939 vrele(nd.ni_dvp); 3940 else 3941 vput(nd.ni_dvp); 3942 VFS_UNLOCK_GIANT(vfslocked); 3943 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3944 return (error); 3945 goto restart; 3946 } 3947 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3948 vn_finished_write(mp); 3949 out: 3950 NDFREE(&nd, NDF_ONLY_PNBUF); 3951 vput(vp); 3952 if (nd.ni_dvp == vp) 3953 vrele(nd.ni_dvp); 3954 else 3955 vput(nd.ni_dvp); 3956 VFS_UNLOCK_GIANT(vfslocked); 3957 return (error); 3958 } 3959 3960 #ifdef COMPAT_43 3961 /* 3962 * Read a block of directory entries in a filesystem independent format. 3963 */ 3964 #ifndef _SYS_SYSPROTO_H_ 3965 struct ogetdirentries_args { 3966 int fd; 3967 char *buf; 3968 u_int count; 3969 long *basep; 3970 }; 3971 #endif 3972 int 3973 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap) 3974 { 3975 long loff; 3976 int error; 3977 3978 error = kern_ogetdirentries(td, uap, &loff); 3979 if (error == 0) 3980 error = copyout(&loff, uap->basep, sizeof(long)); 3981 return (error); 3982 } 3983 3984 int 3985 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap, 3986 long *ploff) 3987 { 3988 struct vnode *vp; 3989 struct file *fp; 3990 struct uio auio, kuio; 3991 struct iovec aiov, kiov; 3992 struct dirent *dp, *edp; 3993 caddr_t dirbuf; 3994 int error, eofflag, readcnt, vfslocked; 3995 long loff; 3996 3997 /* XXX arbitrary sanity limit on `count'. */ 3998 if (uap->count > 64 * 1024) 3999 return (EINVAL); 4000 if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_READ, 4001 &fp)) != 0) 4002 return (error); 4003 if ((fp->f_flag & FREAD) == 0) { 4004 fdrop(fp, td); 4005 return (EBADF); 4006 } 4007 vp = fp->f_vnode; 4008 unionread: 4009 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 4010 if (vp->v_type != VDIR) { 4011 VFS_UNLOCK_GIANT(vfslocked); 4012 fdrop(fp, td); 4013 return (EINVAL); 4014 } 4015 aiov.iov_base = uap->buf; 4016 aiov.iov_len = uap->count; 4017 auio.uio_iov = &aiov; 4018 auio.uio_iovcnt = 1; 4019 auio.uio_rw = UIO_READ; 4020 auio.uio_segflg = UIO_USERSPACE; 4021 auio.uio_td = td; 4022 auio.uio_resid = uap->count; 4023 vn_lock(vp, LK_SHARED | LK_RETRY); 4024 loff = auio.uio_offset = fp->f_offset; 4025 #ifdef MAC 4026 error = mac_vnode_check_readdir(td->td_ucred, vp); 4027 if (error) { 4028 VOP_UNLOCK(vp, 0); 4029 VFS_UNLOCK_GIANT(vfslocked); 4030 fdrop(fp, td); 4031 return (error); 4032 } 4033 #endif 4034 # if (BYTE_ORDER != LITTLE_ENDIAN) 4035 if (vp->v_mount->mnt_maxsymlinklen <= 0) { 4036 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, 4037 NULL, NULL); 4038 fp->f_offset = auio.uio_offset; 4039 } else 4040 # endif 4041 { 4042 kuio = auio; 4043 kuio.uio_iov = &kiov; 4044 kuio.uio_segflg = UIO_SYSSPACE; 4045 kiov.iov_len = uap->count; 4046 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK); 4047 kiov.iov_base = dirbuf; 4048 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag, 4049 NULL, NULL); 4050 fp->f_offset = kuio.uio_offset; 4051 if (error == 0) { 4052 readcnt = uap->count - kuio.uio_resid; 4053 edp = (struct dirent *)&dirbuf[readcnt]; 4054 for (dp = (struct dirent *)dirbuf; dp < edp; ) { 4055 # if (BYTE_ORDER == LITTLE_ENDIAN) 4056 /* 4057 * The expected low byte of 4058 * dp->d_namlen is our dp->d_type. 4059 * The high MBZ byte of dp->d_namlen 4060 * is our dp->d_namlen. 4061 */ 4062 dp->d_type = dp->d_namlen; 4063 dp->d_namlen = 0; 4064 # else 4065 /* 4066 * The dp->d_type is the high byte 4067 * of the expected dp->d_namlen, 4068 * so must be zero'ed. 4069 */ 4070 dp->d_type = 0; 4071 # endif 4072 if (dp->d_reclen > 0) { 4073 dp = (struct dirent *) 4074 ((char *)dp + dp->d_reclen); 4075 } else { 4076 error = EIO; 4077 break; 4078 } 4079 } 4080 if (dp >= edp) 4081 error = uiomove(dirbuf, readcnt, &auio); 4082 } 4083 free(dirbuf, M_TEMP); 4084 } 4085 if (error) { 4086 VOP_UNLOCK(vp, 0); 4087 VFS_UNLOCK_GIANT(vfslocked); 4088 fdrop(fp, td); 4089 return (error); 4090 } 4091 if (uap->count == auio.uio_resid && 4092 (vp->v_vflag & VV_ROOT) && 4093 (vp->v_mount->mnt_flag & MNT_UNION)) { 4094 struct vnode *tvp = vp; 4095 vp = vp->v_mount->mnt_vnodecovered; 4096 VREF(vp); 4097 fp->f_vnode = vp; 4098 fp->f_data = vp; 4099 fp->f_offset = 0; 4100 vput(tvp); 4101 VFS_UNLOCK_GIANT(vfslocked); 4102 goto unionread; 4103 } 4104 VOP_UNLOCK(vp, 0); 4105 VFS_UNLOCK_GIANT(vfslocked); 4106 fdrop(fp, td); 4107 td->td_retval[0] = uap->count - auio.uio_resid; 4108 if (error == 0) 4109 *ploff = loff; 4110 return (error); 4111 } 4112 #endif /* COMPAT_43 */ 4113 4114 /* 4115 * Read a block of directory entries in a filesystem independent format. 4116 */ 4117 #ifndef _SYS_SYSPROTO_H_ 4118 struct getdirentries_args { 4119 int fd; 4120 char *buf; 4121 u_int count; 4122 long *basep; 4123 }; 4124 #endif 4125 int 4126 sys_getdirentries(td, uap) 4127 struct thread *td; 4128 register struct getdirentries_args /* { 4129 int fd; 4130 char *buf; 4131 u_int count; 4132 long *basep; 4133 } */ *uap; 4134 { 4135 long base; 4136 int error; 4137 4138 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base); 4139 if (error) 4140 return (error); 4141 if (uap->basep != NULL) 4142 error = copyout(&base, uap->basep, sizeof(long)); 4143 return (error); 4144 } 4145 4146 int 4147 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count, 4148 long *basep) 4149 { 4150 struct vnode *vp; 4151 struct file *fp; 4152 struct uio auio; 4153 struct iovec aiov; 4154 int vfslocked; 4155 long loff; 4156 int error, eofflag; 4157 4158 AUDIT_ARG_FD(fd); 4159 auio.uio_resid = count; 4160 if (auio.uio_resid > IOSIZE_MAX) 4161 return (EINVAL); 4162 if ((error = getvnode(td->td_proc->p_fd, fd, CAP_READ | CAP_SEEK, 4163 &fp)) != 0) 4164 return (error); 4165 if ((fp->f_flag & FREAD) == 0) { 4166 fdrop(fp, td); 4167 return (EBADF); 4168 } 4169 vp = fp->f_vnode; 4170 unionread: 4171 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 4172 if (vp->v_type != VDIR) { 4173 VFS_UNLOCK_GIANT(vfslocked); 4174 error = EINVAL; 4175 goto fail; 4176 } 4177 aiov.iov_base = buf; 4178 aiov.iov_len = count; 4179 auio.uio_iov = &aiov; 4180 auio.uio_iovcnt = 1; 4181 auio.uio_rw = UIO_READ; 4182 auio.uio_segflg = UIO_USERSPACE; 4183 auio.uio_td = td; 4184 vn_lock(vp, LK_SHARED | LK_RETRY); 4185 AUDIT_ARG_VNODE1(vp); 4186 loff = auio.uio_offset = fp->f_offset; 4187 #ifdef MAC 4188 error = mac_vnode_check_readdir(td->td_ucred, vp); 4189 if (error == 0) 4190 #endif 4191 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, 4192 NULL); 4193 fp->f_offset = auio.uio_offset; 4194 if (error) { 4195 VOP_UNLOCK(vp, 0); 4196 VFS_UNLOCK_GIANT(vfslocked); 4197 goto fail; 4198 } 4199 if (count == auio.uio_resid && 4200 (vp->v_vflag & VV_ROOT) && 4201 (vp->v_mount->mnt_flag & MNT_UNION)) { 4202 struct vnode *tvp = vp; 4203 vp = vp->v_mount->mnt_vnodecovered; 4204 VREF(vp); 4205 fp->f_vnode = vp; 4206 fp->f_data = vp; 4207 fp->f_offset = 0; 4208 vput(tvp); 4209 VFS_UNLOCK_GIANT(vfslocked); 4210 goto unionread; 4211 } 4212 VOP_UNLOCK(vp, 0); 4213 VFS_UNLOCK_GIANT(vfslocked); 4214 *basep = loff; 4215 td->td_retval[0] = count - auio.uio_resid; 4216 fail: 4217 fdrop(fp, td); 4218 return (error); 4219 } 4220 4221 #ifndef _SYS_SYSPROTO_H_ 4222 struct getdents_args { 4223 int fd; 4224 char *buf; 4225 size_t count; 4226 }; 4227 #endif 4228 int 4229 sys_getdents(td, uap) 4230 struct thread *td; 4231 register struct getdents_args /* { 4232 int fd; 4233 char *buf; 4234 u_int count; 4235 } */ *uap; 4236 { 4237 struct getdirentries_args ap; 4238 ap.fd = uap->fd; 4239 ap.buf = uap->buf; 4240 ap.count = uap->count; 4241 ap.basep = NULL; 4242 return (sys_getdirentries(td, &ap)); 4243 } 4244 4245 /* 4246 * Set the mode mask for creation of filesystem nodes. 4247 */ 4248 #ifndef _SYS_SYSPROTO_H_ 4249 struct umask_args { 4250 int newmask; 4251 }; 4252 #endif 4253 int 4254 sys_umask(td, uap) 4255 struct thread *td; 4256 struct umask_args /* { 4257 int newmask; 4258 } */ *uap; 4259 { 4260 register struct filedesc *fdp; 4261 4262 FILEDESC_XLOCK(td->td_proc->p_fd); 4263 fdp = td->td_proc->p_fd; 4264 td->td_retval[0] = fdp->fd_cmask; 4265 fdp->fd_cmask = uap->newmask & ALLPERMS; 4266 FILEDESC_XUNLOCK(td->td_proc->p_fd); 4267 return (0); 4268 } 4269 4270 /* 4271 * Void all references to file by ripping underlying filesystem away from 4272 * vnode. 4273 */ 4274 #ifndef _SYS_SYSPROTO_H_ 4275 struct revoke_args { 4276 char *path; 4277 }; 4278 #endif 4279 int 4280 sys_revoke(td, uap) 4281 struct thread *td; 4282 register struct revoke_args /* { 4283 char *path; 4284 } */ *uap; 4285 { 4286 struct vnode *vp; 4287 struct vattr vattr; 4288 int error; 4289 struct nameidata nd; 4290 int vfslocked; 4291 4292 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1, 4293 UIO_USERSPACE, uap->path, td); 4294 if ((error = namei(&nd)) != 0) 4295 return (error); 4296 vfslocked = NDHASGIANT(&nd); 4297 vp = nd.ni_vp; 4298 NDFREE(&nd, NDF_ONLY_PNBUF); 4299 if (vp->v_type != VCHR || vp->v_rdev == NULL) { 4300 error = EINVAL; 4301 goto out; 4302 } 4303 #ifdef MAC 4304 error = mac_vnode_check_revoke(td->td_ucred, vp); 4305 if (error) 4306 goto out; 4307 #endif 4308 error = VOP_GETATTR(vp, &vattr, td->td_ucred); 4309 if (error) 4310 goto out; 4311 if (td->td_ucred->cr_uid != vattr.va_uid) { 4312 error = priv_check(td, PRIV_VFS_ADMIN); 4313 if (error) 4314 goto out; 4315 } 4316 if (vcount(vp) > 1) 4317 VOP_REVOKE(vp, REVOKEALL); 4318 out: 4319 vput(vp); 4320 VFS_UNLOCK_GIANT(vfslocked); 4321 return (error); 4322 } 4323 4324 /* 4325 * Convert a user file descriptor to a kernel file entry and check that, if it 4326 * is a capability, the correct rights are present. A reference on the file 4327 * entry is held upon returning. 4328 */ 4329 int 4330 getvnode(struct filedesc *fdp, int fd, cap_rights_t rights, 4331 struct file **fpp) 4332 { 4333 struct file *fp; 4334 #ifdef CAPABILITIES 4335 struct file *fp_fromcap; 4336 #endif 4337 int error; 4338 4339 error = 0; 4340 fp = NULL; 4341 if ((fdp == NULL) || (fp = fget_unlocked(fdp, fd)) == NULL) 4342 return (EBADF); 4343 #ifdef CAPABILITIES 4344 /* 4345 * If the file descriptor is for a capability, test rights and use the 4346 * file descriptor referenced by the capability. 4347 */ 4348 error = cap_funwrap(fp, rights, &fp_fromcap); 4349 if (error) { 4350 fdrop(fp, curthread); 4351 return (error); 4352 } 4353 if (fp != fp_fromcap) { 4354 fhold(fp_fromcap); 4355 fdrop(fp, curthread); 4356 fp = fp_fromcap; 4357 } 4358 #endif /* CAPABILITIES */ 4359 4360 /* 4361 * The file could be not of the vnode type, or it may be not 4362 * yet fully initialized, in which case the f_vnode pointer 4363 * may be set, but f_ops is still badfileops. E.g., 4364 * devfs_open() transiently create such situation to 4365 * facilitate csw d_fdopen(). 4366 * 4367 * Dupfdopen() handling in kern_openat() installs the 4368 * half-baked file into the process descriptor table, allowing 4369 * other thread to dereference it. Guard against the race by 4370 * checking f_ops. 4371 */ 4372 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) { 4373 fdrop(fp, curthread); 4374 return (EINVAL); 4375 } 4376 *fpp = fp; 4377 return (0); 4378 } 4379 4380 4381 /* 4382 * Get an (NFS) file handle. 4383 */ 4384 #ifndef _SYS_SYSPROTO_H_ 4385 struct lgetfh_args { 4386 char *fname; 4387 fhandle_t *fhp; 4388 }; 4389 #endif 4390 int 4391 sys_lgetfh(td, uap) 4392 struct thread *td; 4393 register struct lgetfh_args *uap; 4394 { 4395 struct nameidata nd; 4396 fhandle_t fh; 4397 register struct vnode *vp; 4398 int vfslocked; 4399 int error; 4400 4401 error = priv_check(td, PRIV_VFS_GETFH); 4402 if (error) 4403 return (error); 4404 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1, 4405 UIO_USERSPACE, uap->fname, td); 4406 error = namei(&nd); 4407 if (error) 4408 return (error); 4409 vfslocked = NDHASGIANT(&nd); 4410 NDFREE(&nd, NDF_ONLY_PNBUF); 4411 vp = nd.ni_vp; 4412 bzero(&fh, sizeof(fh)); 4413 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4414 error = VOP_VPTOFH(vp, &fh.fh_fid); 4415 vput(vp); 4416 VFS_UNLOCK_GIANT(vfslocked); 4417 if (error) 4418 return (error); 4419 error = copyout(&fh, uap->fhp, sizeof (fh)); 4420 return (error); 4421 } 4422 4423 #ifndef _SYS_SYSPROTO_H_ 4424 struct getfh_args { 4425 char *fname; 4426 fhandle_t *fhp; 4427 }; 4428 #endif 4429 int 4430 sys_getfh(td, uap) 4431 struct thread *td; 4432 register struct getfh_args *uap; 4433 { 4434 struct nameidata nd; 4435 fhandle_t fh; 4436 register struct vnode *vp; 4437 int vfslocked; 4438 int error; 4439 4440 error = priv_check(td, PRIV_VFS_GETFH); 4441 if (error) 4442 return (error); 4443 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1, 4444 UIO_USERSPACE, uap->fname, td); 4445 error = namei(&nd); 4446 if (error) 4447 return (error); 4448 vfslocked = NDHASGIANT(&nd); 4449 NDFREE(&nd, NDF_ONLY_PNBUF); 4450 vp = nd.ni_vp; 4451 bzero(&fh, sizeof(fh)); 4452 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4453 error = VOP_VPTOFH(vp, &fh.fh_fid); 4454 vput(vp); 4455 VFS_UNLOCK_GIANT(vfslocked); 4456 if (error) 4457 return (error); 4458 error = copyout(&fh, uap->fhp, sizeof (fh)); 4459 return (error); 4460 } 4461 4462 /* 4463 * syscall for the rpc.lockd to use to translate a NFS file handle into an 4464 * open descriptor. 4465 * 4466 * warning: do not remove the priv_check() call or this becomes one giant 4467 * security hole. 4468 */ 4469 #ifndef _SYS_SYSPROTO_H_ 4470 struct fhopen_args { 4471 const struct fhandle *u_fhp; 4472 int flags; 4473 }; 4474 #endif 4475 int 4476 sys_fhopen(td, uap) 4477 struct thread *td; 4478 struct fhopen_args /* { 4479 const struct fhandle *u_fhp; 4480 int flags; 4481 } */ *uap; 4482 { 4483 struct proc *p = td->td_proc; 4484 struct mount *mp; 4485 struct vnode *vp; 4486 struct fhandle fhp; 4487 struct vattr vat; 4488 struct vattr *vap = &vat; 4489 struct flock lf; 4490 struct file *fp; 4491 register struct filedesc *fdp = p->p_fd; 4492 int fmode, error, type; 4493 accmode_t accmode; 4494 struct file *nfp; 4495 int vfslocked; 4496 int indx; 4497 4498 error = priv_check(td, PRIV_VFS_FHOPEN); 4499 if (error) 4500 return (error); 4501 fmode = FFLAGS(uap->flags); 4502 /* why not allow a non-read/write open for our lockd? */ 4503 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT)) 4504 return (EINVAL); 4505 error = copyin(uap->u_fhp, &fhp, sizeof(fhp)); 4506 if (error) 4507 return(error); 4508 /* find the mount point */ 4509 mp = vfs_busyfs(&fhp.fh_fsid); 4510 if (mp == NULL) 4511 return (ESTALE); 4512 vfslocked = VFS_LOCK_GIANT(mp); 4513 /* now give me my vnode, it gets returned to me locked */ 4514 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp); 4515 vfs_unbusy(mp); 4516 if (error) 4517 goto out; 4518 /* 4519 * from now on we have to make sure not 4520 * to forget about the vnode 4521 * any error that causes an abort must vput(vp) 4522 * just set error = err and 'goto bad;'. 4523 */ 4524 4525 /* 4526 * from vn_open 4527 */ 4528 if (vp->v_type == VLNK) { 4529 error = EMLINK; 4530 goto bad; 4531 } 4532 if (vp->v_type == VSOCK) { 4533 error = EOPNOTSUPP; 4534 goto bad; 4535 } 4536 if (vp->v_type != VDIR && fmode & O_DIRECTORY) { 4537 error = ENOTDIR; 4538 goto bad; 4539 } 4540 accmode = 0; 4541 if (fmode & (FWRITE | O_TRUNC)) { 4542 if (vp->v_type == VDIR) { 4543 error = EISDIR; 4544 goto bad; 4545 } 4546 error = vn_writechk(vp); 4547 if (error) 4548 goto bad; 4549 accmode |= VWRITE; 4550 } 4551 if (fmode & FREAD) 4552 accmode |= VREAD; 4553 if ((fmode & O_APPEND) && (fmode & FWRITE)) 4554 accmode |= VAPPEND; 4555 #ifdef MAC 4556 error = mac_vnode_check_open(td->td_ucred, vp, accmode); 4557 if (error) 4558 goto bad; 4559 #endif 4560 if (accmode) { 4561 error = VOP_ACCESS(vp, accmode, td->td_ucred, td); 4562 if (error) 4563 goto bad; 4564 } 4565 if (fmode & O_TRUNC) { 4566 vfs_ref(mp); 4567 VOP_UNLOCK(vp, 0); /* XXX */ 4568 if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) { 4569 vrele(vp); 4570 vfs_rel(mp); 4571 goto out; 4572 } 4573 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 4574 vfs_rel(mp); 4575 #ifdef MAC 4576 /* 4577 * We don't yet have fp->f_cred, so use td->td_ucred, which 4578 * should be right. 4579 */ 4580 error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp); 4581 if (error == 0) { 4582 #endif 4583 VATTR_NULL(vap); 4584 vap->va_size = 0; 4585 error = VOP_SETATTR(vp, vap, td->td_ucred); 4586 #ifdef MAC 4587 } 4588 #endif 4589 vn_finished_write(mp); 4590 if (error) 4591 goto bad; 4592 } 4593 error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL); 4594 if (error) 4595 goto bad; 4596 4597 if (fmode & FWRITE) 4598 vp->v_writecount++; 4599 4600 /* 4601 * end of vn_open code 4602 */ 4603 4604 if ((error = falloc(td, &nfp, &indx, fmode)) != 0) { 4605 if (fmode & FWRITE) 4606 vp->v_writecount--; 4607 goto bad; 4608 } 4609 /* An extra reference on `nfp' has been held for us by falloc(). */ 4610 fp = nfp; 4611 nfp->f_vnode = vp; 4612 finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops); 4613 if (fmode & (O_EXLOCK | O_SHLOCK)) { 4614 lf.l_whence = SEEK_SET; 4615 lf.l_start = 0; 4616 lf.l_len = 0; 4617 if (fmode & O_EXLOCK) 4618 lf.l_type = F_WRLCK; 4619 else 4620 lf.l_type = F_RDLCK; 4621 type = F_FLOCK; 4622 if ((fmode & FNONBLOCK) == 0) 4623 type |= F_WAIT; 4624 VOP_UNLOCK(vp, 0); 4625 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, 4626 type)) != 0) { 4627 /* 4628 * The lock request failed. Normally close the 4629 * descriptor but handle the case where someone might 4630 * have dup()d or close()d it when we weren't looking. 4631 */ 4632 fdclose(fdp, fp, indx, td); 4633 4634 /* 4635 * release our private reference 4636 */ 4637 fdrop(fp, td); 4638 goto out; 4639 } 4640 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4641 atomic_set_int(&fp->f_flag, FHASLOCK); 4642 } 4643 4644 VOP_UNLOCK(vp, 0); 4645 fdrop(fp, td); 4646 VFS_UNLOCK_GIANT(vfslocked); 4647 td->td_retval[0] = indx; 4648 return (0); 4649 4650 bad: 4651 vput(vp); 4652 out: 4653 VFS_UNLOCK_GIANT(vfslocked); 4654 return (error); 4655 } 4656 4657 /* 4658 * Stat an (NFS) file handle. 4659 */ 4660 #ifndef _SYS_SYSPROTO_H_ 4661 struct fhstat_args { 4662 struct fhandle *u_fhp; 4663 struct stat *sb; 4664 }; 4665 #endif 4666 int 4667 sys_fhstat(td, uap) 4668 struct thread *td; 4669 register struct fhstat_args /* { 4670 struct fhandle *u_fhp; 4671 struct stat *sb; 4672 } */ *uap; 4673 { 4674 struct stat sb; 4675 fhandle_t fh; 4676 struct mount *mp; 4677 struct vnode *vp; 4678 int vfslocked; 4679 int error; 4680 4681 error = priv_check(td, PRIV_VFS_FHSTAT); 4682 if (error) 4683 return (error); 4684 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 4685 if (error) 4686 return (error); 4687 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4688 return (ESTALE); 4689 vfslocked = VFS_LOCK_GIANT(mp); 4690 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4691 vfs_unbusy(mp); 4692 if (error) { 4693 VFS_UNLOCK_GIANT(vfslocked); 4694 return (error); 4695 } 4696 error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td); 4697 vput(vp); 4698 VFS_UNLOCK_GIANT(vfslocked); 4699 if (error) 4700 return (error); 4701 error = copyout(&sb, uap->sb, sizeof(sb)); 4702 return (error); 4703 } 4704 4705 /* 4706 * Implement fstatfs() for (NFS) file handles. 4707 */ 4708 #ifndef _SYS_SYSPROTO_H_ 4709 struct fhstatfs_args { 4710 struct fhandle *u_fhp; 4711 struct statfs *buf; 4712 }; 4713 #endif 4714 int 4715 sys_fhstatfs(td, uap) 4716 struct thread *td; 4717 struct fhstatfs_args /* { 4718 struct fhandle *u_fhp; 4719 struct statfs *buf; 4720 } */ *uap; 4721 { 4722 struct statfs sf; 4723 fhandle_t fh; 4724 int error; 4725 4726 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 4727 if (error) 4728 return (error); 4729 error = kern_fhstatfs(td, fh, &sf); 4730 if (error) 4731 return (error); 4732 return (copyout(&sf, uap->buf, sizeof(sf))); 4733 } 4734 4735 int 4736 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf) 4737 { 4738 struct statfs *sp; 4739 struct mount *mp; 4740 struct vnode *vp; 4741 int vfslocked; 4742 int error; 4743 4744 error = priv_check(td, PRIV_VFS_FHSTATFS); 4745 if (error) 4746 return (error); 4747 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4748 return (ESTALE); 4749 vfslocked = VFS_LOCK_GIANT(mp); 4750 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4751 if (error) { 4752 vfs_unbusy(mp); 4753 VFS_UNLOCK_GIANT(vfslocked); 4754 return (error); 4755 } 4756 vput(vp); 4757 error = prison_canseemount(td->td_ucred, mp); 4758 if (error) 4759 goto out; 4760 #ifdef MAC 4761 error = mac_mount_check_stat(td->td_ucred, mp); 4762 if (error) 4763 goto out; 4764 #endif 4765 /* 4766 * Set these in case the underlying filesystem fails to do so. 4767 */ 4768 sp = &mp->mnt_stat; 4769 sp->f_version = STATFS_VERSION; 4770 sp->f_namemax = NAME_MAX; 4771 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 4772 error = VFS_STATFS(mp, sp); 4773 if (error == 0) 4774 *buf = *sp; 4775 out: 4776 vfs_unbusy(mp); 4777 VFS_UNLOCK_GIANT(vfslocked); 4778 return (error); 4779 } 4780 4781 int 4782 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len) 4783 { 4784 struct file *fp; 4785 struct mount *mp; 4786 struct vnode *vp; 4787 off_t olen, ooffset; 4788 int error, vfslocked; 4789 4790 fp = NULL; 4791 vfslocked = 0; 4792 error = fget(td, fd, CAP_WRITE, &fp); 4793 if (error != 0) 4794 goto out; 4795 4796 switch (fp->f_type) { 4797 case DTYPE_VNODE: 4798 break; 4799 case DTYPE_PIPE: 4800 case DTYPE_FIFO: 4801 error = ESPIPE; 4802 goto out; 4803 default: 4804 error = ENODEV; 4805 goto out; 4806 } 4807 if ((fp->f_flag & FWRITE) == 0) { 4808 error = EBADF; 4809 goto out; 4810 } 4811 vp = fp->f_vnode; 4812 if (vp->v_type != VREG) { 4813 error = ENODEV; 4814 goto out; 4815 } 4816 if (offset < 0 || len <= 0) { 4817 error = EINVAL; 4818 goto out; 4819 } 4820 /* Check for wrap. */ 4821 if (offset > OFF_MAX - len) { 4822 error = EFBIG; 4823 goto out; 4824 } 4825 4826 /* Allocating blocks may take a long time, so iterate. */ 4827 for (;;) { 4828 olen = len; 4829 ooffset = offset; 4830 4831 bwillwrite(); 4832 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 4833 mp = NULL; 4834 error = vn_start_write(vp, &mp, V_WAIT | PCATCH); 4835 if (error != 0) { 4836 VFS_UNLOCK_GIANT(vfslocked); 4837 break; 4838 } 4839 error = vn_lock(vp, LK_EXCLUSIVE); 4840 if (error != 0) { 4841 vn_finished_write(mp); 4842 VFS_UNLOCK_GIANT(vfslocked); 4843 break; 4844 } 4845 #ifdef MAC 4846 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp); 4847 if (error == 0) 4848 #endif 4849 error = VOP_ALLOCATE(vp, &offset, &len); 4850 VOP_UNLOCK(vp, 0); 4851 vn_finished_write(mp); 4852 VFS_UNLOCK_GIANT(vfslocked); 4853 4854 if (olen + ooffset != offset + len) { 4855 panic("offset + len changed from %jx/%jx to %jx/%jx", 4856 ooffset, olen, offset, len); 4857 } 4858 if (error != 0 || len == 0) 4859 break; 4860 KASSERT(olen > len, ("Iteration did not make progress?")); 4861 maybe_yield(); 4862 } 4863 out: 4864 if (fp != NULL) 4865 fdrop(fp, td); 4866 return (error); 4867 } 4868 4869 int 4870 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap) 4871 { 4872 4873 return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len)); 4874 } 4875 4876 /* 4877 * Unlike madvise(2), we do not make a best effort to remember every 4878 * possible caching hint. Instead, we remember the last setting with 4879 * the exception that we will allow POSIX_FADV_NORMAL to adjust the 4880 * region of any current setting. 4881 */ 4882 int 4883 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len, 4884 int advice) 4885 { 4886 struct fadvise_info *fa, *new; 4887 struct file *fp; 4888 struct vnode *vp; 4889 off_t end; 4890 int error; 4891 4892 if (offset < 0 || len < 0 || offset > OFF_MAX - len) 4893 return (EINVAL); 4894 switch (advice) { 4895 case POSIX_FADV_SEQUENTIAL: 4896 case POSIX_FADV_RANDOM: 4897 case POSIX_FADV_NOREUSE: 4898 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK); 4899 break; 4900 case POSIX_FADV_NORMAL: 4901 case POSIX_FADV_WILLNEED: 4902 case POSIX_FADV_DONTNEED: 4903 new = NULL; 4904 break; 4905 default: 4906 return (EINVAL); 4907 } 4908 /* XXX: CAP_POSIX_FADVISE? */ 4909 error = fget(td, fd, 0, &fp); 4910 if (error != 0) 4911 goto out; 4912 4913 switch (fp->f_type) { 4914 case DTYPE_VNODE: 4915 break; 4916 case DTYPE_PIPE: 4917 case DTYPE_FIFO: 4918 error = ESPIPE; 4919 goto out; 4920 default: 4921 error = ENODEV; 4922 goto out; 4923 } 4924 vp = fp->f_vnode; 4925 if (vp->v_type != VREG) { 4926 error = ENODEV; 4927 goto out; 4928 } 4929 if (len == 0) 4930 end = OFF_MAX; 4931 else 4932 end = offset + len - 1; 4933 switch (advice) { 4934 case POSIX_FADV_SEQUENTIAL: 4935 case POSIX_FADV_RANDOM: 4936 case POSIX_FADV_NOREUSE: 4937 /* 4938 * Try to merge any existing non-standard region with 4939 * this new region if possible, otherwise create a new 4940 * non-standard region for this request. 4941 */ 4942 mtx_pool_lock(mtxpool_sleep, fp); 4943 fa = fp->f_advice; 4944 if (fa != NULL && fa->fa_advice == advice && 4945 ((fa->fa_start <= end && fa->fa_end >= offset) || 4946 (end != OFF_MAX && fa->fa_start == end + 1) || 4947 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) { 4948 if (offset < fa->fa_start) 4949 fa->fa_start = offset; 4950 if (end > fa->fa_end) 4951 fa->fa_end = end; 4952 } else { 4953 new->fa_advice = advice; 4954 new->fa_start = offset; 4955 new->fa_end = end; 4956 fp->f_advice = new; 4957 new = fa; 4958 } 4959 mtx_pool_unlock(mtxpool_sleep, fp); 4960 break; 4961 case POSIX_FADV_NORMAL: 4962 /* 4963 * If a the "normal" region overlaps with an existing 4964 * non-standard region, trim or remove the 4965 * non-standard region. 4966 */ 4967 mtx_pool_lock(mtxpool_sleep, fp); 4968 fa = fp->f_advice; 4969 if (fa != NULL) { 4970 if (offset <= fa->fa_start && end >= fa->fa_end) { 4971 new = fa; 4972 fp->f_advice = NULL; 4973 } else if (offset <= fa->fa_start && 4974 end >= fa->fa_start) 4975 fa->fa_start = end + 1; 4976 else if (offset <= fa->fa_end && end >= fa->fa_end) 4977 fa->fa_end = offset - 1; 4978 else if (offset >= fa->fa_start && end <= fa->fa_end) { 4979 /* 4980 * If the "normal" region is a middle 4981 * portion of the existing 4982 * non-standard region, just remove 4983 * the whole thing rather than picking 4984 * one side or the other to 4985 * preserve. 4986 */ 4987 new = fa; 4988 fp->f_advice = NULL; 4989 } 4990 } 4991 mtx_pool_unlock(mtxpool_sleep, fp); 4992 break; 4993 case POSIX_FADV_WILLNEED: 4994 case POSIX_FADV_DONTNEED: 4995 error = VOP_ADVISE(vp, offset, end, advice); 4996 break; 4997 } 4998 out: 4999 if (fp != NULL) 5000 fdrop(fp, td); 5001 free(new, M_FADVISE); 5002 return (error); 5003 } 5004 5005 int 5006 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap) 5007 { 5008 5009 return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len, 5010 uap->advice)); 5011 }
Cache object: 8a69b8c370fdcba6b86e119ec55e9bf6
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]