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 * 3. 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("$FreeBSD$"); 39 40 #include "opt_capsicum.h" 41 #include "opt_compat.h" 42 #include "opt_ktrace.h" 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/bio.h> 47 #include <sys/buf.h> 48 #include <sys/capsicum.h> 49 #include <sys/disk.h> 50 #include <sys/sysent.h> 51 #include <sys/malloc.h> 52 #include <sys/mount.h> 53 #include <sys/mutex.h> 54 #include <sys/sysproto.h> 55 #include <sys/namei.h> 56 #include <sys/filedesc.h> 57 #include <sys/kernel.h> 58 #include <sys/fcntl.h> 59 #include <sys/file.h> 60 #include <sys/filio.h> 61 #include <sys/limits.h> 62 #include <sys/linker.h> 63 #include <sys/rwlock.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_DEFINE2(vfs, , stat, mode, "char *", "int"); 95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int"); 96 97 static int kern_chflagsat(struct thread *td, int fd, const char *path, 98 enum uio_seg pathseg, u_long flags, int atflag); 99 static int setfflags(struct thread *td, struct vnode *, u_long); 100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *); 101 static int getutimens(const struct timespec *, enum uio_seg, 102 struct timespec *, int *); 103 static int setutimes(struct thread *td, struct vnode *, 104 const struct timespec *, int, int); 105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 106 struct thread *td); 107 108 /* 109 * Sync each mounted filesystem. 110 */ 111 #ifndef _SYS_SYSPROTO_H_ 112 struct sync_args { 113 int dummy; 114 }; 115 #endif 116 /* ARGSUSED */ 117 int 118 sys_sync(td, uap) 119 struct thread *td; 120 struct sync_args *uap; 121 { 122 struct mount *mp, *nmp; 123 int save; 124 125 mtx_lock(&mountlist_mtx); 126 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 127 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) { 128 nmp = TAILQ_NEXT(mp, mnt_list); 129 continue; 130 } 131 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 132 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 133 save = curthread_pflags_set(TDP_SYNCIO); 134 vfs_msync(mp, MNT_NOWAIT); 135 VFS_SYNC(mp, MNT_NOWAIT); 136 curthread_pflags_restore(save); 137 vn_finished_write(mp); 138 } 139 mtx_lock(&mountlist_mtx); 140 nmp = TAILQ_NEXT(mp, mnt_list); 141 vfs_unbusy(mp); 142 } 143 mtx_unlock(&mountlist_mtx); 144 return (0); 145 } 146 147 /* 148 * Change filesystem quotas. 149 */ 150 #ifndef _SYS_SYSPROTO_H_ 151 struct quotactl_args { 152 char *path; 153 int cmd; 154 int uid; 155 caddr_t arg; 156 }; 157 #endif 158 int 159 sys_quotactl(td, uap) 160 struct thread *td; 161 register struct quotactl_args /* { 162 char *path; 163 int cmd; 164 int uid; 165 caddr_t arg; 166 } */ *uap; 167 { 168 struct mount *mp; 169 struct nameidata nd; 170 int error; 171 172 AUDIT_ARG_CMD(uap->cmd); 173 AUDIT_ARG_UID(uap->uid); 174 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS)) 175 return (EPERM); 176 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 177 uap->path, td); 178 if ((error = namei(&nd)) != 0) 179 return (error); 180 NDFREE(&nd, NDF_ONLY_PNBUF); 181 mp = nd.ni_vp->v_mount; 182 vfs_ref(mp); 183 vput(nd.ni_vp); 184 error = vfs_busy(mp, 0); 185 vfs_rel(mp); 186 if (error != 0) 187 return (error); 188 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg); 189 190 /* 191 * Since quota on operation typically needs to open quota 192 * file, the Q_QUOTAON handler needs to unbusy the mount point 193 * before calling into namei. Otherwise, unmount might be 194 * started between two vfs_busy() invocations (first is our, 195 * second is from mount point cross-walk code in lookup()), 196 * causing deadlock. 197 * 198 * Require that Q_QUOTAON handles the vfs_busy() reference on 199 * its own, always returning with ubusied mount point. 200 */ 201 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON) 202 vfs_unbusy(mp); 203 return (error); 204 } 205 206 /* 207 * Used by statfs conversion routines to scale the block size up if 208 * necessary so that all of the block counts are <= 'max_size'. Note 209 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero 210 * value of 'n'. 211 */ 212 void 213 statfs_scale_blocks(struct statfs *sf, long max_size) 214 { 215 uint64_t count; 216 int shift; 217 218 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__)); 219 220 /* 221 * Attempt to scale the block counts to give a more accurate 222 * overview to userland of the ratio of free space to used 223 * space. To do this, find the largest block count and compute 224 * a divisor that lets it fit into a signed integer <= max_size. 225 */ 226 if (sf->f_bavail < 0) 227 count = -sf->f_bavail; 228 else 229 count = sf->f_bavail; 230 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count)); 231 if (count <= max_size) 232 return; 233 234 count >>= flsl(max_size); 235 shift = 0; 236 while (count > 0) { 237 shift++; 238 count >>=1; 239 } 240 241 sf->f_bsize <<= shift; 242 sf->f_blocks >>= shift; 243 sf->f_bfree >>= shift; 244 sf->f_bavail >>= shift; 245 } 246 247 static int 248 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf) 249 { 250 struct statfs *sp; 251 int error; 252 253 if (mp == NULL) 254 return (EBADF); 255 error = vfs_busy(mp, 0); 256 vfs_rel(mp); 257 if (error != 0) 258 return (error); 259 #ifdef MAC 260 error = mac_mount_check_stat(td->td_ucred, mp); 261 if (error != 0) 262 goto out; 263 #endif 264 /* 265 * Set these in case the underlying filesystem fails to do so. 266 */ 267 sp = &mp->mnt_stat; 268 sp->f_version = STATFS_VERSION; 269 sp->f_namemax = NAME_MAX; 270 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 271 error = VFS_STATFS(mp, sp); 272 if (error != 0) 273 goto out; 274 *buf = *sp; 275 if (priv_check(td, PRIV_VFS_GENERATION)) { 276 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 277 prison_enforce_statfs(td->td_ucred, mp, buf); 278 } 279 out: 280 vfs_unbusy(mp); 281 return (error); 282 } 283 284 /* 285 * Get filesystem statistics. 286 */ 287 #ifndef _SYS_SYSPROTO_H_ 288 struct statfs_args { 289 char *path; 290 struct statfs *buf; 291 }; 292 #endif 293 int 294 sys_statfs(td, uap) 295 struct thread *td; 296 register struct statfs_args /* { 297 char *path; 298 struct statfs *buf; 299 } */ *uap; 300 { 301 struct statfs *sfp; 302 int error; 303 304 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 305 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 306 if (error == 0) 307 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 308 free(sfp, M_STATFS); 309 return (error); 310 } 311 312 int 313 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg, 314 struct statfs *buf) 315 { 316 struct mount *mp; 317 struct nameidata nd; 318 int error; 319 320 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 321 pathseg, path, td); 322 error = namei(&nd); 323 if (error != 0) 324 return (error); 325 mp = nd.ni_vp->v_mount; 326 vfs_ref(mp); 327 NDFREE(&nd, NDF_ONLY_PNBUF); 328 vput(nd.ni_vp); 329 return (kern_do_statfs(td, mp, buf)); 330 } 331 332 /* 333 * Get filesystem statistics. 334 */ 335 #ifndef _SYS_SYSPROTO_H_ 336 struct fstatfs_args { 337 int fd; 338 struct statfs *buf; 339 }; 340 #endif 341 int 342 sys_fstatfs(td, uap) 343 struct thread *td; 344 register struct fstatfs_args /* { 345 int fd; 346 struct statfs *buf; 347 } */ *uap; 348 { 349 struct statfs *sfp; 350 int error; 351 352 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 353 error = kern_fstatfs(td, uap->fd, sfp); 354 if (error == 0) 355 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 356 free(sfp, M_STATFS); 357 return (error); 358 } 359 360 int 361 kern_fstatfs(struct thread *td, int fd, struct statfs *buf) 362 { 363 struct file *fp; 364 struct mount *mp; 365 struct vnode *vp; 366 cap_rights_t rights; 367 int error; 368 369 AUDIT_ARG_FD(fd); 370 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp); 371 if (error != 0) 372 return (error); 373 vp = fp->f_vnode; 374 vn_lock(vp, LK_SHARED | LK_RETRY); 375 #ifdef AUDIT 376 AUDIT_ARG_VNODE1(vp); 377 #endif 378 mp = vp->v_mount; 379 if (mp != NULL) 380 vfs_ref(mp); 381 VOP_UNLOCK(vp, 0); 382 fdrop(fp, td); 383 return (kern_do_statfs(td, mp, buf)); 384 } 385 386 /* 387 * Get statistics on all filesystems. 388 */ 389 #ifndef _SYS_SYSPROTO_H_ 390 struct getfsstat_args { 391 struct statfs *buf; 392 long bufsize; 393 int mode; 394 }; 395 #endif 396 int 397 sys_getfsstat(td, uap) 398 struct thread *td; 399 register struct getfsstat_args /* { 400 struct statfs *buf; 401 long bufsize; 402 int mode; 403 } */ *uap; 404 { 405 size_t count; 406 int error; 407 408 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX) 409 return (EINVAL); 410 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count, 411 UIO_USERSPACE, uap->mode); 412 if (error == 0) 413 td->td_retval[0] = count; 414 return (error); 415 } 416 417 /* 418 * If (bufsize > 0 && bufseg == UIO_SYSSPACE) 419 * The caller is responsible for freeing memory which will be allocated 420 * in '*buf'. 421 */ 422 int 423 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize, 424 size_t *countp, enum uio_seg bufseg, int mode) 425 { 426 struct mount *mp, *nmp; 427 struct statfs *sfsp, *sp, *sptmp, *tofree; 428 size_t count, maxcount; 429 int error; 430 431 switch (mode) { 432 case MNT_WAIT: 433 case MNT_NOWAIT: 434 break; 435 default: 436 if (bufseg == UIO_SYSSPACE) 437 *buf = NULL; 438 return (EINVAL); 439 } 440 restart: 441 maxcount = bufsize / sizeof(struct statfs); 442 if (bufsize == 0) { 443 sfsp = NULL; 444 tofree = NULL; 445 } else if (bufseg == UIO_USERSPACE) { 446 sfsp = *buf; 447 tofree = NULL; 448 } else /* if (bufseg == UIO_SYSSPACE) */ { 449 count = 0; 450 mtx_lock(&mountlist_mtx); 451 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 452 count++; 453 } 454 mtx_unlock(&mountlist_mtx); 455 if (maxcount > count) 456 maxcount = count; 457 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs), 458 M_STATFS, M_WAITOK); 459 } 460 count = 0; 461 mtx_lock(&mountlist_mtx); 462 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 463 if (prison_canseemount(td->td_ucred, mp) != 0) { 464 nmp = TAILQ_NEXT(mp, mnt_list); 465 continue; 466 } 467 #ifdef MAC 468 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 469 nmp = TAILQ_NEXT(mp, mnt_list); 470 continue; 471 } 472 #endif 473 if (mode == MNT_WAIT) { 474 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) { 475 /* 476 * If vfs_busy() failed, and MBF_NOWAIT 477 * wasn't passed, then the mp is gone. 478 * Furthermore, because of MBF_MNTLSTLOCK, 479 * the mountlist_mtx was dropped. We have 480 * no other choice than to start over. 481 */ 482 mtx_unlock(&mountlist_mtx); 483 free(tofree, M_STATFS); 484 goto restart; 485 } 486 } else { 487 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) { 488 nmp = TAILQ_NEXT(mp, mnt_list); 489 continue; 490 } 491 } 492 if (sfsp != NULL && count < maxcount) { 493 sp = &mp->mnt_stat; 494 /* 495 * Set these in case the underlying filesystem 496 * fails to do so. 497 */ 498 sp->f_version = STATFS_VERSION; 499 sp->f_namemax = NAME_MAX; 500 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 501 /* 502 * If MNT_NOWAIT is specified, do not refresh 503 * the fsstat cache. 504 */ 505 if (mode != MNT_NOWAIT) { 506 error = VFS_STATFS(mp, sp); 507 if (error != 0) { 508 mtx_lock(&mountlist_mtx); 509 nmp = TAILQ_NEXT(mp, mnt_list); 510 vfs_unbusy(mp); 511 continue; 512 } 513 } 514 if (priv_check(td, PRIV_VFS_GENERATION)) { 515 sptmp = malloc(sizeof(struct statfs), M_STATFS, 516 M_WAITOK); 517 *sptmp = *sp; 518 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0; 519 prison_enforce_statfs(td->td_ucred, mp, sptmp); 520 sp = sptmp; 521 } else 522 sptmp = NULL; 523 if (bufseg == UIO_SYSSPACE) { 524 bcopy(sp, sfsp, sizeof(*sp)); 525 free(sptmp, M_STATFS); 526 } else /* if (bufseg == UIO_USERSPACE) */ { 527 error = copyout(sp, sfsp, sizeof(*sp)); 528 free(sptmp, M_STATFS); 529 if (error != 0) { 530 vfs_unbusy(mp); 531 return (error); 532 } 533 } 534 sfsp++; 535 } 536 count++; 537 mtx_lock(&mountlist_mtx); 538 nmp = TAILQ_NEXT(mp, mnt_list); 539 vfs_unbusy(mp); 540 } 541 mtx_unlock(&mountlist_mtx); 542 if (sfsp != NULL && count > maxcount) 543 *countp = maxcount; 544 else 545 *countp = count; 546 return (0); 547 } 548 549 #ifdef COMPAT_FREEBSD4 550 /* 551 * Get old format filesystem statistics. 552 */ 553 static void cvtstatfs(struct statfs *, struct ostatfs *); 554 555 #ifndef _SYS_SYSPROTO_H_ 556 struct freebsd4_statfs_args { 557 char *path; 558 struct ostatfs *buf; 559 }; 560 #endif 561 int 562 freebsd4_statfs(td, uap) 563 struct thread *td; 564 struct freebsd4_statfs_args /* { 565 char *path; 566 struct ostatfs *buf; 567 } */ *uap; 568 { 569 struct ostatfs osb; 570 struct statfs *sfp; 571 int error; 572 573 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 574 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 575 if (error == 0) { 576 cvtstatfs(sfp, &osb); 577 error = copyout(&osb, uap->buf, sizeof(osb)); 578 } 579 free(sfp, M_STATFS); 580 return (error); 581 } 582 583 /* 584 * Get filesystem statistics. 585 */ 586 #ifndef _SYS_SYSPROTO_H_ 587 struct freebsd4_fstatfs_args { 588 int fd; 589 struct ostatfs *buf; 590 }; 591 #endif 592 int 593 freebsd4_fstatfs(td, uap) 594 struct thread *td; 595 struct freebsd4_fstatfs_args /* { 596 int fd; 597 struct ostatfs *buf; 598 } */ *uap; 599 { 600 struct ostatfs osb; 601 struct statfs *sfp; 602 int error; 603 604 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 605 error = kern_fstatfs(td, uap->fd, sfp); 606 if (error == 0) { 607 cvtstatfs(sfp, &osb); 608 error = copyout(&osb, uap->buf, sizeof(osb)); 609 } 610 free(sfp, M_STATFS); 611 return (error); 612 } 613 614 /* 615 * Get statistics on all filesystems. 616 */ 617 #ifndef _SYS_SYSPROTO_H_ 618 struct freebsd4_getfsstat_args { 619 struct ostatfs *buf; 620 long bufsize; 621 int mode; 622 }; 623 #endif 624 int 625 freebsd4_getfsstat(td, uap) 626 struct thread *td; 627 register struct freebsd4_getfsstat_args /* { 628 struct ostatfs *buf; 629 long bufsize; 630 int mode; 631 } */ *uap; 632 { 633 struct statfs *buf, *sp; 634 struct ostatfs osb; 635 size_t count, size; 636 int error; 637 638 if (uap->bufsize < 0) 639 return (EINVAL); 640 count = uap->bufsize / sizeof(struct ostatfs); 641 if (count > SIZE_MAX / sizeof(struct statfs)) 642 return (EINVAL); 643 size = count * sizeof(struct statfs); 644 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, 645 uap->mode); 646 td->td_retval[0] = count; 647 if (size != 0) { 648 sp = buf; 649 while (count != 0 && error == 0) { 650 cvtstatfs(sp, &osb); 651 error = copyout(&osb, uap->buf, sizeof(osb)); 652 sp++; 653 uap->buf++; 654 count--; 655 } 656 free(buf, M_STATFS); 657 } 658 return (error); 659 } 660 661 /* 662 * Implement fstatfs() for (NFS) file handles. 663 */ 664 #ifndef _SYS_SYSPROTO_H_ 665 struct freebsd4_fhstatfs_args { 666 struct fhandle *u_fhp; 667 struct ostatfs *buf; 668 }; 669 #endif 670 int 671 freebsd4_fhstatfs(td, uap) 672 struct thread *td; 673 struct freebsd4_fhstatfs_args /* { 674 struct fhandle *u_fhp; 675 struct ostatfs *buf; 676 } */ *uap; 677 { 678 struct ostatfs osb; 679 struct statfs *sfp; 680 fhandle_t fh; 681 int error; 682 683 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 684 if (error != 0) 685 return (error); 686 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 687 error = kern_fhstatfs(td, fh, sfp); 688 if (error == 0) { 689 cvtstatfs(sfp, &osb); 690 error = copyout(&osb, uap->buf, sizeof(osb)); 691 } 692 free(sfp, M_STATFS); 693 return (error); 694 } 695 696 /* 697 * Convert a new format statfs structure to an old format statfs structure. 698 */ 699 static void 700 cvtstatfs(nsp, osp) 701 struct statfs *nsp; 702 struct ostatfs *osp; 703 { 704 705 statfs_scale_blocks(nsp, LONG_MAX); 706 bzero(osp, sizeof(*osp)); 707 osp->f_bsize = nsp->f_bsize; 708 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX); 709 osp->f_blocks = nsp->f_blocks; 710 osp->f_bfree = nsp->f_bfree; 711 osp->f_bavail = nsp->f_bavail; 712 osp->f_files = MIN(nsp->f_files, LONG_MAX); 713 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX); 714 osp->f_owner = nsp->f_owner; 715 osp->f_type = nsp->f_type; 716 osp->f_flags = nsp->f_flags; 717 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX); 718 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX); 719 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX); 720 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX); 721 strlcpy(osp->f_fstypename, nsp->f_fstypename, 722 MIN(MFSNAMELEN, OMFSNAMELEN)); 723 strlcpy(osp->f_mntonname, nsp->f_mntonname, 724 MIN(MNAMELEN, OMNAMELEN)); 725 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 726 MIN(MNAMELEN, OMNAMELEN)); 727 osp->f_fsid = nsp->f_fsid; 728 } 729 #endif /* COMPAT_FREEBSD4 */ 730 731 /* 732 * Change current working directory to a given file descriptor. 733 */ 734 #ifndef _SYS_SYSPROTO_H_ 735 struct fchdir_args { 736 int fd; 737 }; 738 #endif 739 int 740 sys_fchdir(td, uap) 741 struct thread *td; 742 struct fchdir_args /* { 743 int fd; 744 } */ *uap; 745 { 746 struct vnode *vp, *tdp; 747 struct mount *mp; 748 struct file *fp; 749 cap_rights_t rights; 750 int error; 751 752 AUDIT_ARG_FD(uap->fd); 753 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR), 754 &fp); 755 if (error != 0) 756 return (error); 757 vp = fp->f_vnode; 758 vrefact(vp); 759 fdrop(fp, td); 760 vn_lock(vp, LK_SHARED | LK_RETRY); 761 AUDIT_ARG_VNODE1(vp); 762 error = change_dir(vp, td); 763 while (!error && (mp = vp->v_mountedhere) != NULL) { 764 if (vfs_busy(mp, 0)) 765 continue; 766 error = VFS_ROOT(mp, LK_SHARED, &tdp); 767 vfs_unbusy(mp); 768 if (error != 0) 769 break; 770 vput(vp); 771 vp = tdp; 772 } 773 if (error != 0) { 774 vput(vp); 775 return (error); 776 } 777 VOP_UNLOCK(vp, 0); 778 pwd_chdir(td, vp); 779 return (0); 780 } 781 782 /* 783 * Change current working directory (``.''). 784 */ 785 #ifndef _SYS_SYSPROTO_H_ 786 struct chdir_args { 787 char *path; 788 }; 789 #endif 790 int 791 sys_chdir(td, uap) 792 struct thread *td; 793 struct chdir_args /* { 794 char *path; 795 } */ *uap; 796 { 797 798 return (kern_chdir(td, uap->path, UIO_USERSPACE)); 799 } 800 801 int 802 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg) 803 { 804 struct nameidata nd; 805 int error; 806 807 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 808 pathseg, path, td); 809 if ((error = namei(&nd)) != 0) 810 return (error); 811 if ((error = change_dir(nd.ni_vp, td)) != 0) { 812 vput(nd.ni_vp); 813 NDFREE(&nd, NDF_ONLY_PNBUF); 814 return (error); 815 } 816 VOP_UNLOCK(nd.ni_vp, 0); 817 NDFREE(&nd, NDF_ONLY_PNBUF); 818 pwd_chdir(td, nd.ni_vp); 819 return (0); 820 } 821 822 /* 823 * Change notion of root (``/'') directory. 824 */ 825 #ifndef _SYS_SYSPROTO_H_ 826 struct chroot_args { 827 char *path; 828 }; 829 #endif 830 int 831 sys_chroot(td, uap) 832 struct thread *td; 833 struct chroot_args /* { 834 char *path; 835 } */ *uap; 836 { 837 struct nameidata nd; 838 int error; 839 840 error = priv_check(td, PRIV_VFS_CHROOT); 841 if (error != 0) 842 return (error); 843 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 844 UIO_USERSPACE, uap->path, td); 845 error = namei(&nd); 846 if (error != 0) 847 goto error; 848 error = change_dir(nd.ni_vp, td); 849 if (error != 0) 850 goto e_vunlock; 851 #ifdef MAC 852 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp); 853 if (error != 0) 854 goto e_vunlock; 855 #endif 856 VOP_UNLOCK(nd.ni_vp, 0); 857 error = pwd_chroot(td, nd.ni_vp); 858 vrele(nd.ni_vp); 859 NDFREE(&nd, NDF_ONLY_PNBUF); 860 return (error); 861 e_vunlock: 862 vput(nd.ni_vp); 863 error: 864 NDFREE(&nd, NDF_ONLY_PNBUF); 865 return (error); 866 } 867 868 /* 869 * Common routine for chroot and chdir. Callers must provide a locked vnode 870 * instance. 871 */ 872 int 873 change_dir(vp, td) 874 struct vnode *vp; 875 struct thread *td; 876 { 877 #ifdef MAC 878 int error; 879 #endif 880 881 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked"); 882 if (vp->v_type != VDIR) 883 return (ENOTDIR); 884 #ifdef MAC 885 error = mac_vnode_check_chdir(td->td_ucred, vp); 886 if (error != 0) 887 return (error); 888 #endif 889 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td)); 890 } 891 892 static __inline void 893 flags_to_rights(int flags, cap_rights_t *rightsp) 894 { 895 896 if (flags & O_EXEC) { 897 cap_rights_set(rightsp, CAP_FEXECVE); 898 } else { 899 switch ((flags & O_ACCMODE)) { 900 case O_RDONLY: 901 cap_rights_set(rightsp, CAP_READ); 902 break; 903 case O_RDWR: 904 cap_rights_set(rightsp, CAP_READ); 905 /* FALLTHROUGH */ 906 case O_WRONLY: 907 cap_rights_set(rightsp, CAP_WRITE); 908 if (!(flags & (O_APPEND | O_TRUNC))) 909 cap_rights_set(rightsp, CAP_SEEK); 910 break; 911 } 912 } 913 914 if (flags & O_CREAT) 915 cap_rights_set(rightsp, CAP_CREATE); 916 917 if (flags & O_TRUNC) 918 cap_rights_set(rightsp, CAP_FTRUNCATE); 919 920 if (flags & (O_SYNC | O_FSYNC)) 921 cap_rights_set(rightsp, CAP_FSYNC); 922 923 if (flags & (O_EXLOCK | O_SHLOCK)) 924 cap_rights_set(rightsp, CAP_FLOCK); 925 } 926 927 /* 928 * Check permissions, allocate an open file structure, and call the device 929 * open routine if any. 930 */ 931 #ifndef _SYS_SYSPROTO_H_ 932 struct open_args { 933 char *path; 934 int flags; 935 int mode; 936 }; 937 #endif 938 int 939 sys_open(td, uap) 940 struct thread *td; 941 register struct open_args /* { 942 char *path; 943 int flags; 944 int mode; 945 } */ *uap; 946 { 947 948 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 949 uap->flags, uap->mode)); 950 } 951 952 #ifndef _SYS_SYSPROTO_H_ 953 struct openat_args { 954 int fd; 955 char *path; 956 int flag; 957 int mode; 958 }; 959 #endif 960 int 961 sys_openat(struct thread *td, struct openat_args *uap) 962 { 963 964 AUDIT_ARG_FD(uap->fd); 965 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 966 uap->mode)); 967 } 968 969 int 970 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 971 int flags, int mode) 972 { 973 struct proc *p = td->td_proc; 974 struct filedesc *fdp = p->p_fd; 975 struct file *fp; 976 struct vnode *vp; 977 struct nameidata nd; 978 cap_rights_t rights; 979 int cmode, error, indx; 980 981 indx = -1; 982 983 AUDIT_ARG_FFLAGS(flags); 984 AUDIT_ARG_MODE(mode); 985 cap_rights_init(&rights, CAP_LOOKUP); 986 flags_to_rights(flags, &rights); 987 /* 988 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 989 * may be specified. 990 */ 991 if (flags & O_EXEC) { 992 if (flags & O_ACCMODE) 993 return (EINVAL); 994 } else if ((flags & O_ACCMODE) == O_ACCMODE) { 995 return (EINVAL); 996 } else { 997 flags = FFLAGS(flags); 998 } 999 1000 /* 1001 * Allocate a file structure. The descriptor to reference it 1002 * is allocated and set by finstall() below. 1003 */ 1004 error = falloc_noinstall(td, &fp); 1005 if (error != 0) 1006 return (error); 1007 /* 1008 * An extra reference on `fp' has been held for us by 1009 * falloc_noinstall(). 1010 */ 1011 /* Set the flags early so the finit in devfs can pick them up. */ 1012 fp->f_flag = flags & FMASK; 1013 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT; 1014 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 1015 &rights, td); 1016 td->td_dupfd = -1; /* XXX check for fdopen */ 1017 error = vn_open(&nd, &flags, cmode, fp); 1018 if (error != 0) { 1019 /* 1020 * If the vn_open replaced the method vector, something 1021 * wonderous happened deep below and we just pass it up 1022 * pretending we know what we do. 1023 */ 1024 if (error == ENXIO && fp->f_ops != &badfileops) 1025 goto success; 1026 1027 /* 1028 * Handle special fdopen() case. bleh. 1029 * 1030 * Don't do this for relative (capability) lookups; we don't 1031 * understand exactly what would happen, and we don't think 1032 * that it ever should. 1033 */ 1034 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 && 1035 (error == ENODEV || error == ENXIO) && 1036 td->td_dupfd >= 0) { 1037 error = dupfdopen(td, fdp, td->td_dupfd, flags, error, 1038 &indx); 1039 if (error == 0) 1040 goto success; 1041 } 1042 1043 goto bad; 1044 } 1045 td->td_dupfd = 0; 1046 NDFREE(&nd, NDF_ONLY_PNBUF); 1047 vp = nd.ni_vp; 1048 1049 /* 1050 * Store the vnode, for any f_type. Typically, the vnode use 1051 * count is decremented by direct call to vn_closefile() for 1052 * files that switched type in the cdevsw fdopen() method. 1053 */ 1054 fp->f_vnode = vp; 1055 /* 1056 * If the file wasn't claimed by devfs bind it to the normal 1057 * vnode operations here. 1058 */ 1059 if (fp->f_ops == &badfileops) { 1060 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo.")); 1061 fp->f_seqcount = 1; 1062 finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK), 1063 DTYPE_VNODE, vp, &vnops); 1064 } 1065 1066 VOP_UNLOCK(vp, 0); 1067 if (flags & O_TRUNC) { 1068 error = fo_truncate(fp, 0, td->td_ucred, td); 1069 if (error != 0) 1070 goto bad; 1071 } 1072 success: 1073 /* 1074 * If we haven't already installed the FD (for dupfdopen), do so now. 1075 */ 1076 if (indx == -1) { 1077 struct filecaps *fcaps; 1078 1079 #ifdef CAPABILITIES 1080 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0) 1081 fcaps = &nd.ni_filecaps; 1082 else 1083 #endif 1084 fcaps = NULL; 1085 error = finstall(td, fp, &indx, flags, fcaps); 1086 /* On success finstall() consumes fcaps. */ 1087 if (error != 0) { 1088 filecaps_free(&nd.ni_filecaps); 1089 goto bad; 1090 } 1091 } else { 1092 filecaps_free(&nd.ni_filecaps); 1093 } 1094 1095 /* 1096 * Release our private reference, leaving the one associated with 1097 * the descriptor table intact. 1098 */ 1099 fdrop(fp, td); 1100 td->td_retval[0] = indx; 1101 return (0); 1102 bad: 1103 KASSERT(indx == -1, ("indx=%d, should be -1", indx)); 1104 fdrop(fp, td); 1105 return (error); 1106 } 1107 1108 #ifdef COMPAT_43 1109 /* 1110 * Create a file. 1111 */ 1112 #ifndef _SYS_SYSPROTO_H_ 1113 struct ocreat_args { 1114 char *path; 1115 int mode; 1116 }; 1117 #endif 1118 int 1119 ocreat(td, uap) 1120 struct thread *td; 1121 register struct ocreat_args /* { 1122 char *path; 1123 int mode; 1124 } */ *uap; 1125 { 1126 1127 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1128 O_WRONLY | O_CREAT | O_TRUNC, uap->mode)); 1129 } 1130 #endif /* COMPAT_43 */ 1131 1132 /* 1133 * Create a special file. 1134 */ 1135 #ifndef _SYS_SYSPROTO_H_ 1136 struct mknod_args { 1137 char *path; 1138 int mode; 1139 int dev; 1140 }; 1141 #endif 1142 int 1143 sys_mknod(td, uap) 1144 struct thread *td; 1145 register struct mknod_args /* { 1146 char *path; 1147 int mode; 1148 int dev; 1149 } */ *uap; 1150 { 1151 1152 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1153 uap->mode, uap->dev)); 1154 } 1155 1156 #ifndef _SYS_SYSPROTO_H_ 1157 struct mknodat_args { 1158 int fd; 1159 char *path; 1160 mode_t mode; 1161 dev_t dev; 1162 }; 1163 #endif 1164 int 1165 sys_mknodat(struct thread *td, struct mknodat_args *uap) 1166 { 1167 1168 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1169 uap->dev)); 1170 } 1171 1172 int 1173 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1174 int mode, int dev) 1175 { 1176 struct vnode *vp; 1177 struct mount *mp; 1178 struct vattr vattr; 1179 struct nameidata nd; 1180 cap_rights_t rights; 1181 int error, whiteout = 0; 1182 1183 AUDIT_ARG_MODE(mode); 1184 AUDIT_ARG_DEV(dev); 1185 switch (mode & S_IFMT) { 1186 case S_IFCHR: 1187 case S_IFBLK: 1188 error = priv_check(td, PRIV_VFS_MKNOD_DEV); 1189 if (error == 0 && dev == VNOVAL) 1190 error = EINVAL; 1191 break; 1192 case S_IFMT: 1193 error = priv_check(td, PRIV_VFS_MKNOD_BAD); 1194 break; 1195 case S_IFWHT: 1196 error = priv_check(td, PRIV_VFS_MKNOD_WHT); 1197 break; 1198 case S_IFIFO: 1199 if (dev == 0) 1200 return (kern_mkfifoat(td, fd, path, pathseg, mode)); 1201 /* FALLTHROUGH */ 1202 default: 1203 error = EINVAL; 1204 break; 1205 } 1206 if (error != 0) 1207 return (error); 1208 restart: 1209 bwillwrite(); 1210 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1211 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT), 1212 td); 1213 if ((error = namei(&nd)) != 0) 1214 return (error); 1215 vp = nd.ni_vp; 1216 if (vp != NULL) { 1217 NDFREE(&nd, NDF_ONLY_PNBUF); 1218 if (vp == nd.ni_dvp) 1219 vrele(nd.ni_dvp); 1220 else 1221 vput(nd.ni_dvp); 1222 vrele(vp); 1223 return (EEXIST); 1224 } else { 1225 VATTR_NULL(&vattr); 1226 vattr.va_mode = (mode & ALLPERMS) & 1227 ~td->td_proc->p_fd->fd_cmask; 1228 vattr.va_rdev = dev; 1229 whiteout = 0; 1230 1231 switch (mode & S_IFMT) { 1232 case S_IFMT: /* used by badsect to flag bad sectors */ 1233 vattr.va_type = VBAD; 1234 break; 1235 case S_IFCHR: 1236 vattr.va_type = VCHR; 1237 break; 1238 case S_IFBLK: 1239 vattr.va_type = VBLK; 1240 break; 1241 case S_IFWHT: 1242 whiteout = 1; 1243 break; 1244 default: 1245 panic("kern_mknod: invalid mode"); 1246 } 1247 } 1248 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1249 NDFREE(&nd, NDF_ONLY_PNBUF); 1250 vput(nd.ni_dvp); 1251 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1252 return (error); 1253 goto restart; 1254 } 1255 #ifdef MAC 1256 if (error == 0 && !whiteout) 1257 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, 1258 &nd.ni_cnd, &vattr); 1259 #endif 1260 if (error == 0) { 1261 if (whiteout) 1262 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1263 else { 1264 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1265 &nd.ni_cnd, &vattr); 1266 if (error == 0) 1267 vput(nd.ni_vp); 1268 } 1269 } 1270 NDFREE(&nd, NDF_ONLY_PNBUF); 1271 vput(nd.ni_dvp); 1272 vn_finished_write(mp); 1273 return (error); 1274 } 1275 1276 /* 1277 * Create a named pipe. 1278 */ 1279 #ifndef _SYS_SYSPROTO_H_ 1280 struct mkfifo_args { 1281 char *path; 1282 int mode; 1283 }; 1284 #endif 1285 int 1286 sys_mkfifo(td, uap) 1287 struct thread *td; 1288 register struct mkfifo_args /* { 1289 char *path; 1290 int mode; 1291 } */ *uap; 1292 { 1293 1294 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1295 uap->mode)); 1296 } 1297 1298 #ifndef _SYS_SYSPROTO_H_ 1299 struct mkfifoat_args { 1300 int fd; 1301 char *path; 1302 mode_t mode; 1303 }; 1304 #endif 1305 int 1306 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap) 1307 { 1308 1309 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE, 1310 uap->mode)); 1311 } 1312 1313 int 1314 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1315 int mode) 1316 { 1317 struct mount *mp; 1318 struct vattr vattr; 1319 struct nameidata nd; 1320 cap_rights_t rights; 1321 int error; 1322 1323 AUDIT_ARG_MODE(mode); 1324 restart: 1325 bwillwrite(); 1326 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1327 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT), 1328 td); 1329 if ((error = namei(&nd)) != 0) 1330 return (error); 1331 if (nd.ni_vp != NULL) { 1332 NDFREE(&nd, NDF_ONLY_PNBUF); 1333 if (nd.ni_vp == nd.ni_dvp) 1334 vrele(nd.ni_dvp); 1335 else 1336 vput(nd.ni_dvp); 1337 vrele(nd.ni_vp); 1338 return (EEXIST); 1339 } 1340 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1341 NDFREE(&nd, NDF_ONLY_PNBUF); 1342 vput(nd.ni_dvp); 1343 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1344 return (error); 1345 goto restart; 1346 } 1347 VATTR_NULL(&vattr); 1348 vattr.va_type = VFIFO; 1349 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask; 1350 #ifdef MAC 1351 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1352 &vattr); 1353 if (error != 0) 1354 goto out; 1355 #endif 1356 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1357 if (error == 0) 1358 vput(nd.ni_vp); 1359 #ifdef MAC 1360 out: 1361 #endif 1362 vput(nd.ni_dvp); 1363 vn_finished_write(mp); 1364 NDFREE(&nd, NDF_ONLY_PNBUF); 1365 return (error); 1366 } 1367 1368 /* 1369 * Make a hard file link. 1370 */ 1371 #ifndef _SYS_SYSPROTO_H_ 1372 struct link_args { 1373 char *path; 1374 char *link; 1375 }; 1376 #endif 1377 int 1378 sys_link(td, uap) 1379 struct thread *td; 1380 register struct link_args /* { 1381 char *path; 1382 char *link; 1383 } */ *uap; 1384 { 1385 1386 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link, 1387 UIO_USERSPACE, FOLLOW)); 1388 } 1389 1390 #ifndef _SYS_SYSPROTO_H_ 1391 struct linkat_args { 1392 int fd1; 1393 char *path1; 1394 int fd2; 1395 char *path2; 1396 int flag; 1397 }; 1398 #endif 1399 int 1400 sys_linkat(struct thread *td, struct linkat_args *uap) 1401 { 1402 int flag; 1403 1404 flag = uap->flag; 1405 if (flag & ~AT_SYMLINK_FOLLOW) 1406 return (EINVAL); 1407 1408 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2, 1409 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW)); 1410 } 1411 1412 int hardlink_check_uid = 0; 1413 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW, 1414 &hardlink_check_uid, 0, 1415 "Unprivileged processes cannot create hard links to files owned by other " 1416 "users"); 1417 static int hardlink_check_gid = 0; 1418 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW, 1419 &hardlink_check_gid, 0, 1420 "Unprivileged processes cannot create hard links to files owned by other " 1421 "groups"); 1422 1423 static int 1424 can_hardlink(struct vnode *vp, struct ucred *cred) 1425 { 1426 struct vattr va; 1427 int error; 1428 1429 if (!hardlink_check_uid && !hardlink_check_gid) 1430 return (0); 1431 1432 error = VOP_GETATTR(vp, &va, cred); 1433 if (error != 0) 1434 return (error); 1435 1436 if (hardlink_check_uid && cred->cr_uid != va.va_uid) { 1437 error = priv_check_cred(cred, PRIV_VFS_LINK, 0); 1438 if (error != 0) 1439 return (error); 1440 } 1441 1442 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) { 1443 error = priv_check_cred(cred, PRIV_VFS_LINK, 0); 1444 if (error != 0) 1445 return (error); 1446 } 1447 1448 return (0); 1449 } 1450 1451 int 1452 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2, 1453 enum uio_seg segflg, int follow) 1454 { 1455 struct vnode *vp; 1456 struct mount *mp; 1457 struct nameidata nd; 1458 cap_rights_t rights; 1459 int error; 1460 1461 again: 1462 bwillwrite(); 1463 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, 1464 cap_rights_init(&rights, CAP_LINKAT_SOURCE), td); 1465 1466 if ((error = namei(&nd)) != 0) 1467 return (error); 1468 NDFREE(&nd, NDF_ONLY_PNBUF); 1469 vp = nd.ni_vp; 1470 if (vp->v_type == VDIR) { 1471 vrele(vp); 1472 return (EPERM); /* POSIX */ 1473 } 1474 NDINIT_ATRIGHTS(&nd, CREATE, 1475 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2, 1476 cap_rights_init(&rights, CAP_LINKAT_TARGET), td); 1477 if ((error = namei(&nd)) == 0) { 1478 if (nd.ni_vp != NULL) { 1479 NDFREE(&nd, NDF_ONLY_PNBUF); 1480 if (nd.ni_dvp == nd.ni_vp) 1481 vrele(nd.ni_dvp); 1482 else 1483 vput(nd.ni_dvp); 1484 vrele(nd.ni_vp); 1485 vrele(vp); 1486 return (EEXIST); 1487 } else if (nd.ni_dvp->v_mount != vp->v_mount) { 1488 /* 1489 * Cross-device link. No need to recheck 1490 * vp->v_type, since it cannot change, except 1491 * to VBAD. 1492 */ 1493 NDFREE(&nd, NDF_ONLY_PNBUF); 1494 vput(nd.ni_dvp); 1495 vrele(vp); 1496 return (EXDEV); 1497 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) { 1498 error = can_hardlink(vp, td->td_ucred); 1499 #ifdef MAC 1500 if (error == 0) 1501 error = mac_vnode_check_link(td->td_ucred, 1502 nd.ni_dvp, vp, &nd.ni_cnd); 1503 #endif 1504 if (error != 0) { 1505 vput(vp); 1506 vput(nd.ni_dvp); 1507 NDFREE(&nd, NDF_ONLY_PNBUF); 1508 return (error); 1509 } 1510 error = vn_start_write(vp, &mp, V_NOWAIT); 1511 if (error != 0) { 1512 vput(vp); 1513 vput(nd.ni_dvp); 1514 NDFREE(&nd, NDF_ONLY_PNBUF); 1515 error = vn_start_write(NULL, &mp, 1516 V_XSLEEP | PCATCH); 1517 if (error != 0) 1518 return (error); 1519 goto again; 1520 } 1521 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1522 VOP_UNLOCK(vp, 0); 1523 vput(nd.ni_dvp); 1524 vn_finished_write(mp); 1525 NDFREE(&nd, NDF_ONLY_PNBUF); 1526 } else { 1527 vput(nd.ni_dvp); 1528 NDFREE(&nd, NDF_ONLY_PNBUF); 1529 vrele(vp); 1530 goto again; 1531 } 1532 } 1533 vrele(vp); 1534 return (error); 1535 } 1536 1537 /* 1538 * Make a symbolic link. 1539 */ 1540 #ifndef _SYS_SYSPROTO_H_ 1541 struct symlink_args { 1542 char *path; 1543 char *link; 1544 }; 1545 #endif 1546 int 1547 sys_symlink(td, uap) 1548 struct thread *td; 1549 register struct symlink_args /* { 1550 char *path; 1551 char *link; 1552 } */ *uap; 1553 { 1554 1555 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link, 1556 UIO_USERSPACE)); 1557 } 1558 1559 #ifndef _SYS_SYSPROTO_H_ 1560 struct symlinkat_args { 1561 char *path; 1562 int fd; 1563 char *path2; 1564 }; 1565 #endif 1566 int 1567 sys_symlinkat(struct thread *td, struct symlinkat_args *uap) 1568 { 1569 1570 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2, 1571 UIO_USERSPACE)); 1572 } 1573 1574 int 1575 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2, 1576 enum uio_seg segflg) 1577 { 1578 struct mount *mp; 1579 struct vattr vattr; 1580 char *syspath; 1581 struct nameidata nd; 1582 int error; 1583 cap_rights_t rights; 1584 1585 if (segflg == UIO_SYSSPACE) { 1586 syspath = path1; 1587 } else { 1588 syspath = uma_zalloc(namei_zone, M_WAITOK); 1589 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0) 1590 goto out; 1591 } 1592 AUDIT_ARG_TEXT(syspath); 1593 restart: 1594 bwillwrite(); 1595 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1596 NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT), 1597 td); 1598 if ((error = namei(&nd)) != 0) 1599 goto out; 1600 if (nd.ni_vp) { 1601 NDFREE(&nd, NDF_ONLY_PNBUF); 1602 if (nd.ni_vp == nd.ni_dvp) 1603 vrele(nd.ni_dvp); 1604 else 1605 vput(nd.ni_dvp); 1606 vrele(nd.ni_vp); 1607 error = EEXIST; 1608 goto out; 1609 } 1610 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1611 NDFREE(&nd, NDF_ONLY_PNBUF); 1612 vput(nd.ni_dvp); 1613 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1614 goto out; 1615 goto restart; 1616 } 1617 VATTR_NULL(&vattr); 1618 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask; 1619 #ifdef MAC 1620 vattr.va_type = VLNK; 1621 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1622 &vattr); 1623 if (error != 0) 1624 goto out2; 1625 #endif 1626 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath); 1627 if (error == 0) 1628 vput(nd.ni_vp); 1629 #ifdef MAC 1630 out2: 1631 #endif 1632 NDFREE(&nd, NDF_ONLY_PNBUF); 1633 vput(nd.ni_dvp); 1634 vn_finished_write(mp); 1635 out: 1636 if (segflg != UIO_SYSSPACE) 1637 uma_zfree(namei_zone, syspath); 1638 return (error); 1639 } 1640 1641 /* 1642 * Delete a whiteout from the filesystem. 1643 */ 1644 int 1645 sys_undelete(td, uap) 1646 struct thread *td; 1647 register struct undelete_args /* { 1648 char *path; 1649 } */ *uap; 1650 { 1651 struct mount *mp; 1652 struct nameidata nd; 1653 int error; 1654 1655 restart: 1656 bwillwrite(); 1657 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1, 1658 UIO_USERSPACE, uap->path, td); 1659 error = namei(&nd); 1660 if (error != 0) 1661 return (error); 1662 1663 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1664 NDFREE(&nd, NDF_ONLY_PNBUF); 1665 if (nd.ni_vp == nd.ni_dvp) 1666 vrele(nd.ni_dvp); 1667 else 1668 vput(nd.ni_dvp); 1669 if (nd.ni_vp) 1670 vrele(nd.ni_vp); 1671 return (EEXIST); 1672 } 1673 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1674 NDFREE(&nd, NDF_ONLY_PNBUF); 1675 vput(nd.ni_dvp); 1676 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1677 return (error); 1678 goto restart; 1679 } 1680 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE); 1681 NDFREE(&nd, NDF_ONLY_PNBUF); 1682 vput(nd.ni_dvp); 1683 vn_finished_write(mp); 1684 return (error); 1685 } 1686 1687 /* 1688 * Delete a name from the filesystem. 1689 */ 1690 #ifndef _SYS_SYSPROTO_H_ 1691 struct unlink_args { 1692 char *path; 1693 }; 1694 #endif 1695 int 1696 sys_unlink(td, uap) 1697 struct thread *td; 1698 struct unlink_args /* { 1699 char *path; 1700 } */ *uap; 1701 { 1702 1703 return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0)); 1704 } 1705 1706 #ifndef _SYS_SYSPROTO_H_ 1707 struct unlinkat_args { 1708 int fd; 1709 char *path; 1710 int flag; 1711 }; 1712 #endif 1713 int 1714 sys_unlinkat(struct thread *td, struct unlinkat_args *uap) 1715 { 1716 int flag = uap->flag; 1717 int fd = uap->fd; 1718 char *path = uap->path; 1719 1720 if (flag & ~AT_REMOVEDIR) 1721 return (EINVAL); 1722 1723 if (flag & AT_REMOVEDIR) 1724 return (kern_rmdirat(td, fd, path, UIO_USERSPACE)); 1725 else 1726 return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0)); 1727 } 1728 1729 int 1730 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1731 ino_t oldinum) 1732 { 1733 struct mount *mp; 1734 struct vnode *vp; 1735 struct nameidata nd; 1736 struct stat sb; 1737 cap_rights_t rights; 1738 int error; 1739 1740 restart: 1741 bwillwrite(); 1742 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1, 1743 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td); 1744 if ((error = namei(&nd)) != 0) 1745 return (error == EINVAL ? EPERM : error); 1746 vp = nd.ni_vp; 1747 if (vp->v_type == VDIR && oldinum == 0) { 1748 error = EPERM; /* POSIX */ 1749 } else if (oldinum != 0 && 1750 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) && 1751 sb.st_ino != oldinum) { 1752 error = EIDRM; /* Identifier removed */ 1753 } else { 1754 /* 1755 * The root of a mounted filesystem cannot be deleted. 1756 * 1757 * XXX: can this only be a VDIR case? 1758 */ 1759 if (vp->v_vflag & VV_ROOT) 1760 error = EBUSY; 1761 } 1762 if (error == 0) { 1763 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1764 NDFREE(&nd, NDF_ONLY_PNBUF); 1765 vput(nd.ni_dvp); 1766 if (vp == nd.ni_dvp) 1767 vrele(vp); 1768 else 1769 vput(vp); 1770 if ((error = vn_start_write(NULL, &mp, 1771 V_XSLEEP | PCATCH)) != 0) 1772 return (error); 1773 goto restart; 1774 } 1775 #ifdef MAC 1776 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 1777 &nd.ni_cnd); 1778 if (error != 0) 1779 goto out; 1780 #endif 1781 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 1782 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd); 1783 #ifdef MAC 1784 out: 1785 #endif 1786 vn_finished_write(mp); 1787 } 1788 NDFREE(&nd, NDF_ONLY_PNBUF); 1789 vput(nd.ni_dvp); 1790 if (vp == nd.ni_dvp) 1791 vrele(vp); 1792 else 1793 vput(vp); 1794 return (error); 1795 } 1796 1797 /* 1798 * Reposition read/write file offset. 1799 */ 1800 #ifndef _SYS_SYSPROTO_H_ 1801 struct lseek_args { 1802 int fd; 1803 int pad; 1804 off_t offset; 1805 int whence; 1806 }; 1807 #endif 1808 int 1809 sys_lseek(struct thread *td, struct lseek_args *uap) 1810 { 1811 1812 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1813 } 1814 1815 int 1816 kern_lseek(struct thread *td, int fd, off_t offset, int whence) 1817 { 1818 struct file *fp; 1819 cap_rights_t rights; 1820 int error; 1821 1822 AUDIT_ARG_FD(fd); 1823 error = fget(td, fd, cap_rights_init(&rights, CAP_SEEK), &fp); 1824 if (error != 0) 1825 return (error); 1826 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 1827 fo_seek(fp, offset, whence, td) : ESPIPE; 1828 fdrop(fp, td); 1829 return (error); 1830 } 1831 1832 #if defined(COMPAT_43) 1833 /* 1834 * Reposition read/write file offset. 1835 */ 1836 #ifndef _SYS_SYSPROTO_H_ 1837 struct olseek_args { 1838 int fd; 1839 long offset; 1840 int whence; 1841 }; 1842 #endif 1843 int 1844 olseek(struct thread *td, struct olseek_args *uap) 1845 { 1846 1847 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1848 } 1849 #endif /* COMPAT_43 */ 1850 1851 #if defined(COMPAT_FREEBSD6) 1852 /* Version with the 'pad' argument */ 1853 int 1854 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap) 1855 { 1856 1857 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1858 } 1859 #endif 1860 1861 /* 1862 * Check access permissions using passed credentials. 1863 */ 1864 static int 1865 vn_access(vp, user_flags, cred, td) 1866 struct vnode *vp; 1867 int user_flags; 1868 struct ucred *cred; 1869 struct thread *td; 1870 { 1871 accmode_t accmode; 1872 int error; 1873 1874 /* Flags == 0 means only check for existence. */ 1875 if (user_flags == 0) 1876 return (0); 1877 1878 accmode = 0; 1879 if (user_flags & R_OK) 1880 accmode |= VREAD; 1881 if (user_flags & W_OK) 1882 accmode |= VWRITE; 1883 if (user_flags & X_OK) 1884 accmode |= VEXEC; 1885 #ifdef MAC 1886 error = mac_vnode_check_access(cred, vp, accmode); 1887 if (error != 0) 1888 return (error); 1889 #endif 1890 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 1891 error = VOP_ACCESS(vp, accmode, cred, td); 1892 return (error); 1893 } 1894 1895 /* 1896 * Check access permissions using "real" credentials. 1897 */ 1898 #ifndef _SYS_SYSPROTO_H_ 1899 struct access_args { 1900 char *path; 1901 int amode; 1902 }; 1903 #endif 1904 int 1905 sys_access(td, uap) 1906 struct thread *td; 1907 register struct access_args /* { 1908 char *path; 1909 int amode; 1910 } */ *uap; 1911 { 1912 1913 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1914 0, uap->amode)); 1915 } 1916 1917 #ifndef _SYS_SYSPROTO_H_ 1918 struct faccessat_args { 1919 int dirfd; 1920 char *path; 1921 int amode; 1922 int flag; 1923 } 1924 #endif 1925 int 1926 sys_faccessat(struct thread *td, struct faccessat_args *uap) 1927 { 1928 1929 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 1930 uap->amode)); 1931 } 1932 1933 int 1934 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 1935 int flag, int amode) 1936 { 1937 struct ucred *cred, *usecred; 1938 struct vnode *vp; 1939 struct nameidata nd; 1940 cap_rights_t rights; 1941 int error; 1942 1943 if (flag & ~AT_EACCESS) 1944 return (EINVAL); 1945 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0) 1946 return (EINVAL); 1947 1948 /* 1949 * Create and modify a temporary credential instead of one that 1950 * is potentially shared (if we need one). 1951 */ 1952 cred = td->td_ucred; 1953 if ((flag & AT_EACCESS) == 0 && 1954 ((cred->cr_uid != cred->cr_ruid || 1955 cred->cr_rgid != cred->cr_groups[0]))) { 1956 usecred = crdup(cred); 1957 usecred->cr_uid = cred->cr_ruid; 1958 usecred->cr_groups[0] = cred->cr_rgid; 1959 td->td_ucred = usecred; 1960 } else 1961 usecred = cred; 1962 AUDIT_ARG_VALUE(amode); 1963 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | 1964 AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT), 1965 td); 1966 if ((error = namei(&nd)) != 0) 1967 goto out; 1968 vp = nd.ni_vp; 1969 1970 error = vn_access(vp, amode, usecred, td); 1971 NDFREE(&nd, NDF_ONLY_PNBUF); 1972 vput(vp); 1973 out: 1974 if (usecred != cred) { 1975 td->td_ucred = cred; 1976 crfree(usecred); 1977 } 1978 return (error); 1979 } 1980 1981 /* 1982 * Check access permissions using "effective" credentials. 1983 */ 1984 #ifndef _SYS_SYSPROTO_H_ 1985 struct eaccess_args { 1986 char *path; 1987 int amode; 1988 }; 1989 #endif 1990 int 1991 sys_eaccess(td, uap) 1992 struct thread *td; 1993 register struct eaccess_args /* { 1994 char *path; 1995 int amode; 1996 } */ *uap; 1997 { 1998 1999 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2000 AT_EACCESS, uap->amode)); 2001 } 2002 2003 #if defined(COMPAT_43) 2004 /* 2005 * Get file status; this version follows links. 2006 */ 2007 #ifndef _SYS_SYSPROTO_H_ 2008 struct ostat_args { 2009 char *path; 2010 struct ostat *ub; 2011 }; 2012 #endif 2013 int 2014 ostat(td, uap) 2015 struct thread *td; 2016 register struct ostat_args /* { 2017 char *path; 2018 struct ostat *ub; 2019 } */ *uap; 2020 { 2021 struct stat sb; 2022 struct ostat osb; 2023 int error; 2024 2025 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2026 &sb, NULL); 2027 if (error != 0) 2028 return (error); 2029 cvtstat(&sb, &osb); 2030 return (copyout(&osb, uap->ub, sizeof (osb))); 2031 } 2032 2033 /* 2034 * Get file status; this version does not follow links. 2035 */ 2036 #ifndef _SYS_SYSPROTO_H_ 2037 struct olstat_args { 2038 char *path; 2039 struct ostat *ub; 2040 }; 2041 #endif 2042 int 2043 olstat(td, uap) 2044 struct thread *td; 2045 register struct olstat_args /* { 2046 char *path; 2047 struct ostat *ub; 2048 } */ *uap; 2049 { 2050 struct stat sb; 2051 struct ostat osb; 2052 int error; 2053 2054 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2055 UIO_USERSPACE, &sb, NULL); 2056 if (error != 0) 2057 return (error); 2058 cvtstat(&sb, &osb); 2059 return (copyout(&osb, uap->ub, sizeof (osb))); 2060 } 2061 2062 /* 2063 * Convert from an old to a new stat structure. 2064 */ 2065 void 2066 cvtstat(st, ost) 2067 struct stat *st; 2068 struct ostat *ost; 2069 { 2070 2071 bzero(ost, sizeof(*ost)); 2072 ost->st_dev = st->st_dev; 2073 ost->st_ino = st->st_ino; 2074 ost->st_mode = st->st_mode; 2075 ost->st_nlink = st->st_nlink; 2076 ost->st_uid = st->st_uid; 2077 ost->st_gid = st->st_gid; 2078 ost->st_rdev = st->st_rdev; 2079 if (st->st_size < (quad_t)1 << 32) 2080 ost->st_size = st->st_size; 2081 else 2082 ost->st_size = -2; 2083 ost->st_atim = st->st_atim; 2084 ost->st_mtim = st->st_mtim; 2085 ost->st_ctim = st->st_ctim; 2086 ost->st_blksize = st->st_blksize; 2087 ost->st_blocks = st->st_blocks; 2088 ost->st_flags = st->st_flags; 2089 ost->st_gen = st->st_gen; 2090 } 2091 #endif /* COMPAT_43 */ 2092 2093 /* 2094 * Get file status; this version follows links. 2095 */ 2096 #ifndef _SYS_SYSPROTO_H_ 2097 struct stat_args { 2098 char *path; 2099 struct stat *ub; 2100 }; 2101 #endif 2102 int 2103 sys_stat(td, uap) 2104 struct thread *td; 2105 register struct stat_args /* { 2106 char *path; 2107 struct stat *ub; 2108 } */ *uap; 2109 { 2110 struct stat sb; 2111 int error; 2112 2113 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2114 &sb, NULL); 2115 if (error == 0) 2116 error = copyout(&sb, uap->ub, sizeof (sb)); 2117 return (error); 2118 } 2119 2120 #ifndef _SYS_SYSPROTO_H_ 2121 struct fstatat_args { 2122 int fd; 2123 char *path; 2124 struct stat *buf; 2125 int flag; 2126 } 2127 #endif 2128 int 2129 sys_fstatat(struct thread *td, struct fstatat_args *uap) 2130 { 2131 struct stat sb; 2132 int error; 2133 2134 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2135 UIO_USERSPACE, &sb, NULL); 2136 if (error == 0) 2137 error = copyout(&sb, uap->buf, sizeof (sb)); 2138 return (error); 2139 } 2140 2141 int 2142 kern_statat(struct thread *td, int flag, int fd, char *path, 2143 enum uio_seg pathseg, struct stat *sbp, 2144 void (*hook)(struct vnode *vp, struct stat *sbp)) 2145 { 2146 struct nameidata nd; 2147 struct stat sb; 2148 cap_rights_t rights; 2149 int error; 2150 2151 if (flag & ~AT_SYMLINK_NOFOLLOW) 2152 return (EINVAL); 2153 2154 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : 2155 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd, 2156 cap_rights_init(&rights, CAP_FSTAT), td); 2157 2158 if ((error = namei(&nd)) != 0) 2159 return (error); 2160 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td); 2161 if (error == 0) { 2162 SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode); 2163 if (S_ISREG(sb.st_mode)) 2164 SDT_PROBE2(vfs, , stat, reg, path, pathseg); 2165 if (__predict_false(hook != NULL)) 2166 hook(nd.ni_vp, &sb); 2167 } 2168 NDFREE(&nd, NDF_ONLY_PNBUF); 2169 vput(nd.ni_vp); 2170 if (error != 0) 2171 return (error); 2172 *sbp = sb; 2173 #ifdef KTRACE 2174 if (KTRPOINT(td, KTR_STRUCT)) 2175 ktrstat(&sb); 2176 #endif 2177 return (0); 2178 } 2179 2180 /* 2181 * Get file status; this version does not follow links. 2182 */ 2183 #ifndef _SYS_SYSPROTO_H_ 2184 struct lstat_args { 2185 char *path; 2186 struct stat *ub; 2187 }; 2188 #endif 2189 int 2190 sys_lstat(td, uap) 2191 struct thread *td; 2192 register struct lstat_args /* { 2193 char *path; 2194 struct stat *ub; 2195 } */ *uap; 2196 { 2197 struct stat sb; 2198 int error; 2199 2200 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2201 UIO_USERSPACE, &sb, NULL); 2202 if (error == 0) 2203 error = copyout(&sb, uap->ub, sizeof (sb)); 2204 return (error); 2205 } 2206 2207 /* 2208 * Implementation of the NetBSD [l]stat() functions. 2209 */ 2210 void 2211 cvtnstat(sb, nsb) 2212 struct stat *sb; 2213 struct nstat *nsb; 2214 { 2215 2216 bzero(nsb, sizeof *nsb); 2217 nsb->st_dev = sb->st_dev; 2218 nsb->st_ino = sb->st_ino; 2219 nsb->st_mode = sb->st_mode; 2220 nsb->st_nlink = sb->st_nlink; 2221 nsb->st_uid = sb->st_uid; 2222 nsb->st_gid = sb->st_gid; 2223 nsb->st_rdev = sb->st_rdev; 2224 nsb->st_atim = sb->st_atim; 2225 nsb->st_mtim = sb->st_mtim; 2226 nsb->st_ctim = sb->st_ctim; 2227 nsb->st_size = sb->st_size; 2228 nsb->st_blocks = sb->st_blocks; 2229 nsb->st_blksize = sb->st_blksize; 2230 nsb->st_flags = sb->st_flags; 2231 nsb->st_gen = sb->st_gen; 2232 nsb->st_birthtim = sb->st_birthtim; 2233 } 2234 2235 #ifndef _SYS_SYSPROTO_H_ 2236 struct nstat_args { 2237 char *path; 2238 struct nstat *ub; 2239 }; 2240 #endif 2241 int 2242 sys_nstat(td, uap) 2243 struct thread *td; 2244 register struct nstat_args /* { 2245 char *path; 2246 struct nstat *ub; 2247 } */ *uap; 2248 { 2249 struct stat sb; 2250 struct nstat nsb; 2251 int error; 2252 2253 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2254 &sb, NULL); 2255 if (error != 0) 2256 return (error); 2257 cvtnstat(&sb, &nsb); 2258 return (copyout(&nsb, uap->ub, sizeof (nsb))); 2259 } 2260 2261 /* 2262 * NetBSD lstat. Get file status; this version does not follow links. 2263 */ 2264 #ifndef _SYS_SYSPROTO_H_ 2265 struct lstat_args { 2266 char *path; 2267 struct stat *ub; 2268 }; 2269 #endif 2270 int 2271 sys_nlstat(td, uap) 2272 struct thread *td; 2273 register struct nlstat_args /* { 2274 char *path; 2275 struct nstat *ub; 2276 } */ *uap; 2277 { 2278 struct stat sb; 2279 struct nstat nsb; 2280 int error; 2281 2282 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2283 UIO_USERSPACE, &sb, NULL); 2284 if (error != 0) 2285 return (error); 2286 cvtnstat(&sb, &nsb); 2287 return (copyout(&nsb, uap->ub, sizeof (nsb))); 2288 } 2289 2290 /* 2291 * Get configurable pathname variables. 2292 */ 2293 #ifndef _SYS_SYSPROTO_H_ 2294 struct pathconf_args { 2295 char *path; 2296 int name; 2297 }; 2298 #endif 2299 int 2300 sys_pathconf(td, uap) 2301 struct thread *td; 2302 register struct pathconf_args /* { 2303 char *path; 2304 int name; 2305 } */ *uap; 2306 { 2307 2308 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW)); 2309 } 2310 2311 #ifndef _SYS_SYSPROTO_H_ 2312 struct lpathconf_args { 2313 char *path; 2314 int name; 2315 }; 2316 #endif 2317 int 2318 sys_lpathconf(td, uap) 2319 struct thread *td; 2320 register struct lpathconf_args /* { 2321 char *path; 2322 int name; 2323 } */ *uap; 2324 { 2325 2326 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, 2327 NOFOLLOW)); 2328 } 2329 2330 int 2331 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name, 2332 u_long flags) 2333 { 2334 struct nameidata nd; 2335 int error; 2336 2337 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags, 2338 pathseg, path, td); 2339 if ((error = namei(&nd)) != 0) 2340 return (error); 2341 NDFREE(&nd, NDF_ONLY_PNBUF); 2342 2343 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval); 2344 vput(nd.ni_vp); 2345 return (error); 2346 } 2347 2348 /* 2349 * Return target name of a symbolic link. 2350 */ 2351 #ifndef _SYS_SYSPROTO_H_ 2352 struct readlink_args { 2353 char *path; 2354 char *buf; 2355 size_t count; 2356 }; 2357 #endif 2358 int 2359 sys_readlink(td, uap) 2360 struct thread *td; 2361 register struct readlink_args /* { 2362 char *path; 2363 char *buf; 2364 size_t count; 2365 } */ *uap; 2366 { 2367 2368 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2369 uap->buf, UIO_USERSPACE, uap->count)); 2370 } 2371 #ifndef _SYS_SYSPROTO_H_ 2372 struct readlinkat_args { 2373 int fd; 2374 char *path; 2375 char *buf; 2376 size_t bufsize; 2377 }; 2378 #endif 2379 int 2380 sys_readlinkat(struct thread *td, struct readlinkat_args *uap) 2381 { 2382 2383 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE, 2384 uap->buf, UIO_USERSPACE, uap->bufsize)); 2385 } 2386 2387 int 2388 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 2389 char *buf, enum uio_seg bufseg, size_t count) 2390 { 2391 struct vnode *vp; 2392 struct iovec aiov; 2393 struct uio auio; 2394 struct nameidata nd; 2395 int error; 2396 2397 if (count > IOSIZE_MAX) 2398 return (EINVAL); 2399 2400 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 2401 pathseg, path, fd, td); 2402 2403 if ((error = namei(&nd)) != 0) 2404 return (error); 2405 NDFREE(&nd, NDF_ONLY_PNBUF); 2406 vp = nd.ni_vp; 2407 #ifdef MAC 2408 error = mac_vnode_check_readlink(td->td_ucred, vp); 2409 if (error != 0) { 2410 vput(vp); 2411 return (error); 2412 } 2413 #endif 2414 if (vp->v_type != VLNK) 2415 error = EINVAL; 2416 else { 2417 aiov.iov_base = buf; 2418 aiov.iov_len = count; 2419 auio.uio_iov = &aiov; 2420 auio.uio_iovcnt = 1; 2421 auio.uio_offset = 0; 2422 auio.uio_rw = UIO_READ; 2423 auio.uio_segflg = bufseg; 2424 auio.uio_td = td; 2425 auio.uio_resid = count; 2426 error = VOP_READLINK(vp, &auio, td->td_ucred); 2427 td->td_retval[0] = count - auio.uio_resid; 2428 } 2429 vput(vp); 2430 return (error); 2431 } 2432 2433 /* 2434 * Common implementation code for chflags() and fchflags(). 2435 */ 2436 static int 2437 setfflags(td, vp, flags) 2438 struct thread *td; 2439 struct vnode *vp; 2440 u_long flags; 2441 { 2442 struct mount *mp; 2443 struct vattr vattr; 2444 int error; 2445 2446 /* We can't support the value matching VNOVAL. */ 2447 if (flags == VNOVAL) 2448 return (EOPNOTSUPP); 2449 2450 /* 2451 * Prevent non-root users from setting flags on devices. When 2452 * a device is reused, users can retain ownership of the device 2453 * if they are allowed to set flags and programs assume that 2454 * chown can't fail when done as root. 2455 */ 2456 if (vp->v_type == VCHR || vp->v_type == VBLK) { 2457 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV); 2458 if (error != 0) 2459 return (error); 2460 } 2461 2462 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2463 return (error); 2464 VATTR_NULL(&vattr); 2465 vattr.va_flags = flags; 2466 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2467 #ifdef MAC 2468 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags); 2469 if (error == 0) 2470 #endif 2471 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 2472 VOP_UNLOCK(vp, 0); 2473 vn_finished_write(mp); 2474 return (error); 2475 } 2476 2477 /* 2478 * Change flags of a file given a path name. 2479 */ 2480 #ifndef _SYS_SYSPROTO_H_ 2481 struct chflags_args { 2482 const char *path; 2483 u_long flags; 2484 }; 2485 #endif 2486 int 2487 sys_chflags(td, uap) 2488 struct thread *td; 2489 register struct chflags_args /* { 2490 const char *path; 2491 u_long flags; 2492 } */ *uap; 2493 { 2494 2495 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2496 uap->flags, 0)); 2497 } 2498 2499 #ifndef _SYS_SYSPROTO_H_ 2500 struct chflagsat_args { 2501 int fd; 2502 const char *path; 2503 u_long flags; 2504 int atflag; 2505 } 2506 #endif 2507 int 2508 sys_chflagsat(struct thread *td, struct chflagsat_args *uap) 2509 { 2510 int fd = uap->fd; 2511 const char *path = uap->path; 2512 u_long flags = uap->flags; 2513 int atflag = uap->atflag; 2514 2515 if (atflag & ~AT_SYMLINK_NOFOLLOW) 2516 return (EINVAL); 2517 2518 return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag)); 2519 } 2520 2521 /* 2522 * Same as chflags() but doesn't follow symlinks. 2523 */ 2524 int 2525 sys_lchflags(td, uap) 2526 struct thread *td; 2527 register struct lchflags_args /* { 2528 const char *path; 2529 u_long flags; 2530 } */ *uap; 2531 { 2532 2533 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2534 uap->flags, AT_SYMLINK_NOFOLLOW)); 2535 } 2536 2537 static int 2538 kern_chflagsat(struct thread *td, int fd, const char *path, 2539 enum uio_seg pathseg, u_long flags, int atflag) 2540 { 2541 struct nameidata nd; 2542 cap_rights_t rights; 2543 int error, follow; 2544 2545 AUDIT_ARG_FFLAGS(flags); 2546 follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 2547 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd, 2548 cap_rights_init(&rights, CAP_FCHFLAGS), td); 2549 if ((error = namei(&nd)) != 0) 2550 return (error); 2551 NDFREE(&nd, NDF_ONLY_PNBUF); 2552 error = setfflags(td, nd.ni_vp, flags); 2553 vrele(nd.ni_vp); 2554 return (error); 2555 } 2556 2557 /* 2558 * Change flags of a file given a file descriptor. 2559 */ 2560 #ifndef _SYS_SYSPROTO_H_ 2561 struct fchflags_args { 2562 int fd; 2563 u_long flags; 2564 }; 2565 #endif 2566 int 2567 sys_fchflags(td, uap) 2568 struct thread *td; 2569 register struct fchflags_args /* { 2570 int fd; 2571 u_long flags; 2572 } */ *uap; 2573 { 2574 struct file *fp; 2575 cap_rights_t rights; 2576 int error; 2577 2578 AUDIT_ARG_FD(uap->fd); 2579 AUDIT_ARG_FFLAGS(uap->flags); 2580 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS), 2581 &fp); 2582 if (error != 0) 2583 return (error); 2584 #ifdef AUDIT 2585 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 2586 AUDIT_ARG_VNODE1(fp->f_vnode); 2587 VOP_UNLOCK(fp->f_vnode, 0); 2588 #endif 2589 error = setfflags(td, fp->f_vnode, uap->flags); 2590 fdrop(fp, td); 2591 return (error); 2592 } 2593 2594 /* 2595 * Common implementation code for chmod(), lchmod() and fchmod(). 2596 */ 2597 int 2598 setfmode(td, cred, vp, mode) 2599 struct thread *td; 2600 struct ucred *cred; 2601 struct vnode *vp; 2602 int mode; 2603 { 2604 struct mount *mp; 2605 struct vattr vattr; 2606 int error; 2607 2608 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2609 return (error); 2610 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2611 VATTR_NULL(&vattr); 2612 vattr.va_mode = mode & ALLPERMS; 2613 #ifdef MAC 2614 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode); 2615 if (error == 0) 2616 #endif 2617 error = VOP_SETATTR(vp, &vattr, cred); 2618 VOP_UNLOCK(vp, 0); 2619 vn_finished_write(mp); 2620 return (error); 2621 } 2622 2623 /* 2624 * Change mode of a file given path name. 2625 */ 2626 #ifndef _SYS_SYSPROTO_H_ 2627 struct chmod_args { 2628 char *path; 2629 int mode; 2630 }; 2631 #endif 2632 int 2633 sys_chmod(td, uap) 2634 struct thread *td; 2635 register struct chmod_args /* { 2636 char *path; 2637 int mode; 2638 } */ *uap; 2639 { 2640 2641 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2642 uap->mode, 0)); 2643 } 2644 2645 #ifndef _SYS_SYSPROTO_H_ 2646 struct fchmodat_args { 2647 int dirfd; 2648 char *path; 2649 mode_t mode; 2650 int flag; 2651 } 2652 #endif 2653 int 2654 sys_fchmodat(struct thread *td, struct fchmodat_args *uap) 2655 { 2656 int flag = uap->flag; 2657 int fd = uap->fd; 2658 char *path = uap->path; 2659 mode_t mode = uap->mode; 2660 2661 if (flag & ~AT_SYMLINK_NOFOLLOW) 2662 return (EINVAL); 2663 2664 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag)); 2665 } 2666 2667 /* 2668 * Change mode of a file given path name (don't follow links.) 2669 */ 2670 #ifndef _SYS_SYSPROTO_H_ 2671 struct lchmod_args { 2672 char *path; 2673 int mode; 2674 }; 2675 #endif 2676 int 2677 sys_lchmod(td, uap) 2678 struct thread *td; 2679 register struct lchmod_args /* { 2680 char *path; 2681 int mode; 2682 } */ *uap; 2683 { 2684 2685 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2686 uap->mode, AT_SYMLINK_NOFOLLOW)); 2687 } 2688 2689 int 2690 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 2691 mode_t mode, int flag) 2692 { 2693 struct nameidata nd; 2694 cap_rights_t rights; 2695 int error, follow; 2696 2697 AUDIT_ARG_MODE(mode); 2698 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 2699 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd, 2700 cap_rights_init(&rights, CAP_FCHMOD), td); 2701 if ((error = namei(&nd)) != 0) 2702 return (error); 2703 NDFREE(&nd, NDF_ONLY_PNBUF); 2704 error = setfmode(td, td->td_ucred, nd.ni_vp, mode); 2705 vrele(nd.ni_vp); 2706 return (error); 2707 } 2708 2709 /* 2710 * Change mode of a file given a file descriptor. 2711 */ 2712 #ifndef _SYS_SYSPROTO_H_ 2713 struct fchmod_args { 2714 int fd; 2715 int mode; 2716 }; 2717 #endif 2718 int 2719 sys_fchmod(struct thread *td, struct fchmod_args *uap) 2720 { 2721 struct file *fp; 2722 cap_rights_t rights; 2723 int error; 2724 2725 AUDIT_ARG_FD(uap->fd); 2726 AUDIT_ARG_MODE(uap->mode); 2727 2728 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp); 2729 if (error != 0) 2730 return (error); 2731 error = fo_chmod(fp, uap->mode, td->td_ucred, td); 2732 fdrop(fp, td); 2733 return (error); 2734 } 2735 2736 /* 2737 * Common implementation for chown(), lchown(), and fchown() 2738 */ 2739 int 2740 setfown(td, cred, vp, uid, gid) 2741 struct thread *td; 2742 struct ucred *cred; 2743 struct vnode *vp; 2744 uid_t uid; 2745 gid_t gid; 2746 { 2747 struct mount *mp; 2748 struct vattr vattr; 2749 int error; 2750 2751 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2752 return (error); 2753 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2754 VATTR_NULL(&vattr); 2755 vattr.va_uid = uid; 2756 vattr.va_gid = gid; 2757 #ifdef MAC 2758 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid, 2759 vattr.va_gid); 2760 if (error == 0) 2761 #endif 2762 error = VOP_SETATTR(vp, &vattr, cred); 2763 VOP_UNLOCK(vp, 0); 2764 vn_finished_write(mp); 2765 return (error); 2766 } 2767 2768 /* 2769 * Set ownership given a path name. 2770 */ 2771 #ifndef _SYS_SYSPROTO_H_ 2772 struct chown_args { 2773 char *path; 2774 int uid; 2775 int gid; 2776 }; 2777 #endif 2778 int 2779 sys_chown(td, uap) 2780 struct thread *td; 2781 register struct chown_args /* { 2782 char *path; 2783 int uid; 2784 int gid; 2785 } */ *uap; 2786 { 2787 2788 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid, 2789 uap->gid, 0)); 2790 } 2791 2792 #ifndef _SYS_SYSPROTO_H_ 2793 struct fchownat_args { 2794 int fd; 2795 const char * path; 2796 uid_t uid; 2797 gid_t gid; 2798 int flag; 2799 }; 2800 #endif 2801 int 2802 sys_fchownat(struct thread *td, struct fchownat_args *uap) 2803 { 2804 int flag; 2805 2806 flag = uap->flag; 2807 if (flag & ~AT_SYMLINK_NOFOLLOW) 2808 return (EINVAL); 2809 2810 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid, 2811 uap->gid, uap->flag)); 2812 } 2813 2814 int 2815 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 2816 int uid, int gid, int flag) 2817 { 2818 struct nameidata nd; 2819 cap_rights_t rights; 2820 int error, follow; 2821 2822 AUDIT_ARG_OWNER(uid, gid); 2823 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 2824 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd, 2825 cap_rights_init(&rights, CAP_FCHOWN), td); 2826 2827 if ((error = namei(&nd)) != 0) 2828 return (error); 2829 NDFREE(&nd, NDF_ONLY_PNBUF); 2830 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid); 2831 vrele(nd.ni_vp); 2832 return (error); 2833 } 2834 2835 /* 2836 * Set ownership given a path name, do not cross symlinks. 2837 */ 2838 #ifndef _SYS_SYSPROTO_H_ 2839 struct lchown_args { 2840 char *path; 2841 int uid; 2842 int gid; 2843 }; 2844 #endif 2845 int 2846 sys_lchown(td, uap) 2847 struct thread *td; 2848 register struct lchown_args /* { 2849 char *path; 2850 int uid; 2851 int gid; 2852 } */ *uap; 2853 { 2854 2855 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2856 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW)); 2857 } 2858 2859 /* 2860 * Set ownership given a file descriptor. 2861 */ 2862 #ifndef _SYS_SYSPROTO_H_ 2863 struct fchown_args { 2864 int fd; 2865 int uid; 2866 int gid; 2867 }; 2868 #endif 2869 int 2870 sys_fchown(td, uap) 2871 struct thread *td; 2872 register struct fchown_args /* { 2873 int fd; 2874 int uid; 2875 int gid; 2876 } */ *uap; 2877 { 2878 struct file *fp; 2879 cap_rights_t rights; 2880 int error; 2881 2882 AUDIT_ARG_FD(uap->fd); 2883 AUDIT_ARG_OWNER(uap->uid, uap->gid); 2884 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp); 2885 if (error != 0) 2886 return (error); 2887 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td); 2888 fdrop(fp, td); 2889 return (error); 2890 } 2891 2892 /* 2893 * Common implementation code for utimes(), lutimes(), and futimes(). 2894 */ 2895 static int 2896 getutimes(usrtvp, tvpseg, tsp) 2897 const struct timeval *usrtvp; 2898 enum uio_seg tvpseg; 2899 struct timespec *tsp; 2900 { 2901 struct timeval tv[2]; 2902 const struct timeval *tvp; 2903 int error; 2904 2905 if (usrtvp == NULL) { 2906 vfs_timestamp(&tsp[0]); 2907 tsp[1] = tsp[0]; 2908 } else { 2909 if (tvpseg == UIO_SYSSPACE) { 2910 tvp = usrtvp; 2911 } else { 2912 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0) 2913 return (error); 2914 tvp = tv; 2915 } 2916 2917 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 || 2918 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000) 2919 return (EINVAL); 2920 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]); 2921 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]); 2922 } 2923 return (0); 2924 } 2925 2926 /* 2927 * Common implementation code for futimens(), utimensat(). 2928 */ 2929 #define UTIMENS_NULL 0x1 2930 #define UTIMENS_EXIT 0x2 2931 static int 2932 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg, 2933 struct timespec *tsp, int *retflags) 2934 { 2935 struct timespec tsnow; 2936 int error; 2937 2938 vfs_timestamp(&tsnow); 2939 *retflags = 0; 2940 if (usrtsp == NULL) { 2941 tsp[0] = tsnow; 2942 tsp[1] = tsnow; 2943 *retflags |= UTIMENS_NULL; 2944 return (0); 2945 } 2946 if (tspseg == UIO_SYSSPACE) { 2947 tsp[0] = usrtsp[0]; 2948 tsp[1] = usrtsp[1]; 2949 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0) 2950 return (error); 2951 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT) 2952 *retflags |= UTIMENS_EXIT; 2953 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW) 2954 *retflags |= UTIMENS_NULL; 2955 if (tsp[0].tv_nsec == UTIME_OMIT) 2956 tsp[0].tv_sec = VNOVAL; 2957 else if (tsp[0].tv_nsec == UTIME_NOW) 2958 tsp[0] = tsnow; 2959 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L) 2960 return (EINVAL); 2961 if (tsp[1].tv_nsec == UTIME_OMIT) 2962 tsp[1].tv_sec = VNOVAL; 2963 else if (tsp[1].tv_nsec == UTIME_NOW) 2964 tsp[1] = tsnow; 2965 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L) 2966 return (EINVAL); 2967 2968 return (0); 2969 } 2970 2971 /* 2972 * Common implementation code for utimes(), lutimes(), futimes(), futimens(), 2973 * and utimensat(). 2974 */ 2975 static int 2976 setutimes(td, vp, ts, numtimes, nullflag) 2977 struct thread *td; 2978 struct vnode *vp; 2979 const struct timespec *ts; 2980 int numtimes; 2981 int nullflag; 2982 { 2983 struct mount *mp; 2984 struct vattr vattr; 2985 int error, setbirthtime; 2986 2987 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2988 return (error); 2989 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2990 setbirthtime = 0; 2991 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) && 2992 timespeccmp(&ts[1], &vattr.va_birthtime, < )) 2993 setbirthtime = 1; 2994 VATTR_NULL(&vattr); 2995 vattr.va_atime = ts[0]; 2996 vattr.va_mtime = ts[1]; 2997 if (setbirthtime) 2998 vattr.va_birthtime = ts[1]; 2999 if (numtimes > 2) 3000 vattr.va_birthtime = ts[2]; 3001 if (nullflag) 3002 vattr.va_vaflags |= VA_UTIMES_NULL; 3003 #ifdef MAC 3004 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime, 3005 vattr.va_mtime); 3006 #endif 3007 if (error == 0) 3008 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3009 VOP_UNLOCK(vp, 0); 3010 vn_finished_write(mp); 3011 return (error); 3012 } 3013 3014 /* 3015 * Set the access and modification times of a file. 3016 */ 3017 #ifndef _SYS_SYSPROTO_H_ 3018 struct utimes_args { 3019 char *path; 3020 struct timeval *tptr; 3021 }; 3022 #endif 3023 int 3024 sys_utimes(td, uap) 3025 struct thread *td; 3026 register struct utimes_args /* { 3027 char *path; 3028 struct timeval *tptr; 3029 } */ *uap; 3030 { 3031 3032 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3033 uap->tptr, UIO_USERSPACE)); 3034 } 3035 3036 #ifndef _SYS_SYSPROTO_H_ 3037 struct futimesat_args { 3038 int fd; 3039 const char * path; 3040 const struct timeval * times; 3041 }; 3042 #endif 3043 int 3044 sys_futimesat(struct thread *td, struct futimesat_args *uap) 3045 { 3046 3047 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE, 3048 uap->times, UIO_USERSPACE)); 3049 } 3050 3051 int 3052 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 3053 struct timeval *tptr, enum uio_seg tptrseg) 3054 { 3055 struct nameidata nd; 3056 struct timespec ts[2]; 3057 cap_rights_t rights; 3058 int error; 3059 3060 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3061 return (error); 3062 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 3063 cap_rights_init(&rights, CAP_FUTIMES), td); 3064 3065 if ((error = namei(&nd)) != 0) 3066 return (error); 3067 NDFREE(&nd, NDF_ONLY_PNBUF); 3068 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3069 vrele(nd.ni_vp); 3070 return (error); 3071 } 3072 3073 /* 3074 * Set the access and modification times of a file. 3075 */ 3076 #ifndef _SYS_SYSPROTO_H_ 3077 struct lutimes_args { 3078 char *path; 3079 struct timeval *tptr; 3080 }; 3081 #endif 3082 int 3083 sys_lutimes(td, uap) 3084 struct thread *td; 3085 register struct lutimes_args /* { 3086 char *path; 3087 struct timeval *tptr; 3088 } */ *uap; 3089 { 3090 3091 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr, 3092 UIO_USERSPACE)); 3093 } 3094 3095 int 3096 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg, 3097 struct timeval *tptr, enum uio_seg tptrseg) 3098 { 3099 struct timespec ts[2]; 3100 struct nameidata nd; 3101 int error; 3102 3103 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3104 return (error); 3105 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td); 3106 if ((error = namei(&nd)) != 0) 3107 return (error); 3108 NDFREE(&nd, NDF_ONLY_PNBUF); 3109 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3110 vrele(nd.ni_vp); 3111 return (error); 3112 } 3113 3114 /* 3115 * Set the access and modification times of a file. 3116 */ 3117 #ifndef _SYS_SYSPROTO_H_ 3118 struct futimes_args { 3119 int fd; 3120 struct timeval *tptr; 3121 }; 3122 #endif 3123 int 3124 sys_futimes(td, uap) 3125 struct thread *td; 3126 register struct futimes_args /* { 3127 int fd; 3128 struct timeval *tptr; 3129 } */ *uap; 3130 { 3131 3132 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE)); 3133 } 3134 3135 int 3136 kern_futimes(struct thread *td, int fd, struct timeval *tptr, 3137 enum uio_seg tptrseg) 3138 { 3139 struct timespec ts[2]; 3140 struct file *fp; 3141 cap_rights_t rights; 3142 int error; 3143 3144 AUDIT_ARG_FD(fd); 3145 error = getutimes(tptr, tptrseg, ts); 3146 if (error != 0) 3147 return (error); 3148 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp); 3149 if (error != 0) 3150 return (error); 3151 #ifdef AUDIT 3152 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3153 AUDIT_ARG_VNODE1(fp->f_vnode); 3154 VOP_UNLOCK(fp->f_vnode, 0); 3155 #endif 3156 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL); 3157 fdrop(fp, td); 3158 return (error); 3159 } 3160 3161 int 3162 sys_futimens(struct thread *td, struct futimens_args *uap) 3163 { 3164 3165 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE)); 3166 } 3167 3168 int 3169 kern_futimens(struct thread *td, int fd, struct timespec *tptr, 3170 enum uio_seg tptrseg) 3171 { 3172 struct timespec ts[2]; 3173 struct file *fp; 3174 cap_rights_t rights; 3175 int error, flags; 3176 3177 AUDIT_ARG_FD(fd); 3178 error = getutimens(tptr, tptrseg, ts, &flags); 3179 if (error != 0) 3180 return (error); 3181 if (flags & UTIMENS_EXIT) 3182 return (0); 3183 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp); 3184 if (error != 0) 3185 return (error); 3186 #ifdef AUDIT 3187 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3188 AUDIT_ARG_VNODE1(fp->f_vnode); 3189 VOP_UNLOCK(fp->f_vnode, 0); 3190 #endif 3191 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL); 3192 fdrop(fp, td); 3193 return (error); 3194 } 3195 3196 int 3197 sys_utimensat(struct thread *td, struct utimensat_args *uap) 3198 { 3199 3200 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE, 3201 uap->times, UIO_USERSPACE, uap->flag)); 3202 } 3203 3204 int 3205 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg, 3206 struct timespec *tptr, enum uio_seg tptrseg, int flag) 3207 { 3208 struct nameidata nd; 3209 struct timespec ts[2]; 3210 cap_rights_t rights; 3211 int error, flags; 3212 3213 if (flag & ~AT_SYMLINK_NOFOLLOW) 3214 return (EINVAL); 3215 3216 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0) 3217 return (error); 3218 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : 3219 FOLLOW) | AUDITVNODE1, pathseg, path, fd, 3220 cap_rights_init(&rights, CAP_FUTIMES), td); 3221 if ((error = namei(&nd)) != 0) 3222 return (error); 3223 /* 3224 * We are allowed to call namei() regardless of 2xUTIME_OMIT. 3225 * POSIX states: 3226 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected." 3227 * "Search permission is denied by a component of the path prefix." 3228 */ 3229 NDFREE(&nd, NDF_ONLY_PNBUF); 3230 if ((flags & UTIMENS_EXIT) == 0) 3231 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL); 3232 vrele(nd.ni_vp); 3233 return (error); 3234 } 3235 3236 /* 3237 * Truncate a file given its path name. 3238 */ 3239 #ifndef _SYS_SYSPROTO_H_ 3240 struct truncate_args { 3241 char *path; 3242 int pad; 3243 off_t length; 3244 }; 3245 #endif 3246 int 3247 sys_truncate(td, uap) 3248 struct thread *td; 3249 register struct truncate_args /* { 3250 char *path; 3251 int pad; 3252 off_t length; 3253 } */ *uap; 3254 { 3255 3256 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3257 } 3258 3259 int 3260 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length) 3261 { 3262 struct mount *mp; 3263 struct vnode *vp; 3264 void *rl_cookie; 3265 struct vattr vattr; 3266 struct nameidata nd; 3267 int error; 3268 3269 if (length < 0) 3270 return(EINVAL); 3271 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td); 3272 if ((error = namei(&nd)) != 0) 3273 return (error); 3274 vp = nd.ni_vp; 3275 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX); 3276 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) { 3277 vn_rangelock_unlock(vp, rl_cookie); 3278 vrele(vp); 3279 return (error); 3280 } 3281 NDFREE(&nd, NDF_ONLY_PNBUF); 3282 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3283 if (vp->v_type == VDIR) 3284 error = EISDIR; 3285 #ifdef MAC 3286 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) { 3287 } 3288 #endif 3289 else if ((error = vn_writechk(vp)) == 0 && 3290 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) { 3291 VATTR_NULL(&vattr); 3292 vattr.va_size = length; 3293 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3294 } 3295 VOP_UNLOCK(vp, 0); 3296 vn_finished_write(mp); 3297 vn_rangelock_unlock(vp, rl_cookie); 3298 vrele(vp); 3299 return (error); 3300 } 3301 3302 #if defined(COMPAT_43) 3303 /* 3304 * Truncate a file given its path name. 3305 */ 3306 #ifndef _SYS_SYSPROTO_H_ 3307 struct otruncate_args { 3308 char *path; 3309 long length; 3310 }; 3311 #endif 3312 int 3313 otruncate(struct thread *td, struct otruncate_args *uap) 3314 { 3315 3316 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3317 } 3318 #endif /* COMPAT_43 */ 3319 3320 #if defined(COMPAT_FREEBSD6) 3321 /* Versions with the pad argument */ 3322 int 3323 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap) 3324 { 3325 3326 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3327 } 3328 3329 int 3330 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap) 3331 { 3332 3333 return (kern_ftruncate(td, uap->fd, uap->length)); 3334 } 3335 #endif 3336 3337 int 3338 kern_fsync(struct thread *td, int fd, bool fullsync) 3339 { 3340 struct vnode *vp; 3341 struct mount *mp; 3342 struct file *fp; 3343 cap_rights_t rights; 3344 int error, lock_flags; 3345 3346 AUDIT_ARG_FD(fd); 3347 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp); 3348 if (error != 0) 3349 return (error); 3350 vp = fp->f_vnode; 3351 #if 0 3352 if (!fullsync) 3353 /* XXXKIB: compete outstanding aio writes */; 3354 #endif 3355 error = vn_start_write(vp, &mp, V_WAIT | PCATCH); 3356 if (error != 0) 3357 goto drop; 3358 if (MNT_SHARED_WRITES(mp) || 3359 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) { 3360 lock_flags = LK_SHARED; 3361 } else { 3362 lock_flags = LK_EXCLUSIVE; 3363 } 3364 vn_lock(vp, lock_flags | LK_RETRY); 3365 AUDIT_ARG_VNODE1(vp); 3366 if (vp->v_object != NULL) { 3367 VM_OBJECT_WLOCK(vp->v_object); 3368 vm_object_page_clean(vp->v_object, 0, 0, 0); 3369 VM_OBJECT_WUNLOCK(vp->v_object); 3370 } 3371 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td); 3372 VOP_UNLOCK(vp, 0); 3373 vn_finished_write(mp); 3374 drop: 3375 fdrop(fp, td); 3376 return (error); 3377 } 3378 3379 /* 3380 * Sync an open file. 3381 */ 3382 #ifndef _SYS_SYSPROTO_H_ 3383 struct fsync_args { 3384 int fd; 3385 }; 3386 #endif 3387 int 3388 sys_fsync(struct thread *td, struct fsync_args *uap) 3389 { 3390 3391 return (kern_fsync(td, uap->fd, true)); 3392 } 3393 3394 int 3395 sys_fdatasync(struct thread *td, struct fdatasync_args *uap) 3396 { 3397 3398 return (kern_fsync(td, uap->fd, false)); 3399 } 3400 3401 /* 3402 * Rename files. Source and destination must either both be directories, or 3403 * both not be directories. If target is a directory, it must be empty. 3404 */ 3405 #ifndef _SYS_SYSPROTO_H_ 3406 struct rename_args { 3407 char *from; 3408 char *to; 3409 }; 3410 #endif 3411 int 3412 sys_rename(td, uap) 3413 struct thread *td; 3414 register struct rename_args /* { 3415 char *from; 3416 char *to; 3417 } */ *uap; 3418 { 3419 3420 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD, 3421 uap->to, UIO_USERSPACE)); 3422 } 3423 3424 #ifndef _SYS_SYSPROTO_H_ 3425 struct renameat_args { 3426 int oldfd; 3427 char *old; 3428 int newfd; 3429 char *new; 3430 }; 3431 #endif 3432 int 3433 sys_renameat(struct thread *td, struct renameat_args *uap) 3434 { 3435 3436 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new, 3437 UIO_USERSPACE)); 3438 } 3439 3440 int 3441 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new, 3442 enum uio_seg pathseg) 3443 { 3444 struct mount *mp = NULL; 3445 struct vnode *tvp, *fvp, *tdvp; 3446 struct nameidata fromnd, tond; 3447 cap_rights_t rights; 3448 int error; 3449 3450 again: 3451 bwillwrite(); 3452 #ifdef MAC 3453 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | 3454 AUDITVNODE1, pathseg, old, oldfd, 3455 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td); 3456 #else 3457 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1, 3458 pathseg, old, oldfd, 3459 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td); 3460 #endif 3461 3462 if ((error = namei(&fromnd)) != 0) 3463 return (error); 3464 #ifdef MAC 3465 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp, 3466 fromnd.ni_vp, &fromnd.ni_cnd); 3467 VOP_UNLOCK(fromnd.ni_dvp, 0); 3468 if (fromnd.ni_dvp != fromnd.ni_vp) 3469 VOP_UNLOCK(fromnd.ni_vp, 0); 3470 #endif 3471 fvp = fromnd.ni_vp; 3472 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | 3473 SAVESTART | AUDITVNODE2, pathseg, new, newfd, 3474 cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td); 3475 if (fromnd.ni_vp->v_type == VDIR) 3476 tond.ni_cnd.cn_flags |= WILLBEDIR; 3477 if ((error = namei(&tond)) != 0) { 3478 /* Translate error code for rename("dir1", "dir2/."). */ 3479 if (error == EISDIR && fvp->v_type == VDIR) 3480 error = EINVAL; 3481 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3482 vrele(fromnd.ni_dvp); 3483 vrele(fvp); 3484 goto out1; 3485 } 3486 tdvp = tond.ni_dvp; 3487 tvp = tond.ni_vp; 3488 error = vn_start_write(fvp, &mp, V_NOWAIT); 3489 if (error != 0) { 3490 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3491 NDFREE(&tond, NDF_ONLY_PNBUF); 3492 if (tvp != NULL) 3493 vput(tvp); 3494 if (tdvp == tvp) 3495 vrele(tdvp); 3496 else 3497 vput(tdvp); 3498 vrele(fromnd.ni_dvp); 3499 vrele(fvp); 3500 vrele(tond.ni_startdir); 3501 if (fromnd.ni_startdir != NULL) 3502 vrele(fromnd.ni_startdir); 3503 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH); 3504 if (error != 0) 3505 return (error); 3506 goto again; 3507 } 3508 if (tvp != NULL) { 3509 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3510 error = ENOTDIR; 3511 goto out; 3512 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3513 error = EISDIR; 3514 goto out; 3515 } 3516 #ifdef CAPABILITIES 3517 if (newfd != AT_FDCWD) { 3518 /* 3519 * If the target already exists we require CAP_UNLINKAT 3520 * from 'newfd'. 3521 */ 3522 error = cap_check(&tond.ni_filecaps.fc_rights, 3523 cap_rights_init(&rights, CAP_UNLINKAT)); 3524 if (error != 0) 3525 goto out; 3526 } 3527 #endif 3528 } 3529 if (fvp == tdvp) { 3530 error = EINVAL; 3531 goto out; 3532 } 3533 /* 3534 * If the source is the same as the destination (that is, if they 3535 * are links to the same vnode), then there is nothing to do. 3536 */ 3537 if (fvp == tvp) 3538 error = -1; 3539 #ifdef MAC 3540 else 3541 error = mac_vnode_check_rename_to(td->td_ucred, tdvp, 3542 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd); 3543 #endif 3544 out: 3545 if (error == 0) { 3546 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3547 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3548 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3549 NDFREE(&tond, NDF_ONLY_PNBUF); 3550 } else { 3551 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3552 NDFREE(&tond, NDF_ONLY_PNBUF); 3553 if (tvp != NULL) 3554 vput(tvp); 3555 if (tdvp == tvp) 3556 vrele(tdvp); 3557 else 3558 vput(tdvp); 3559 vrele(fromnd.ni_dvp); 3560 vrele(fvp); 3561 } 3562 vrele(tond.ni_startdir); 3563 vn_finished_write(mp); 3564 out1: 3565 if (fromnd.ni_startdir) 3566 vrele(fromnd.ni_startdir); 3567 if (error == -1) 3568 return (0); 3569 return (error); 3570 } 3571 3572 /* 3573 * Make a directory file. 3574 */ 3575 #ifndef _SYS_SYSPROTO_H_ 3576 struct mkdir_args { 3577 char *path; 3578 int mode; 3579 }; 3580 #endif 3581 int 3582 sys_mkdir(td, uap) 3583 struct thread *td; 3584 register struct mkdir_args /* { 3585 char *path; 3586 int mode; 3587 } */ *uap; 3588 { 3589 3590 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3591 uap->mode)); 3592 } 3593 3594 #ifndef _SYS_SYSPROTO_H_ 3595 struct mkdirat_args { 3596 int fd; 3597 char *path; 3598 mode_t mode; 3599 }; 3600 #endif 3601 int 3602 sys_mkdirat(struct thread *td, struct mkdirat_args *uap) 3603 { 3604 3605 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode)); 3606 } 3607 3608 int 3609 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg, 3610 int mode) 3611 { 3612 struct mount *mp; 3613 struct vnode *vp; 3614 struct vattr vattr; 3615 struct nameidata nd; 3616 cap_rights_t rights; 3617 int error; 3618 3619 AUDIT_ARG_MODE(mode); 3620 restart: 3621 bwillwrite(); 3622 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 3623 NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT), 3624 td); 3625 nd.ni_cnd.cn_flags |= WILLBEDIR; 3626 if ((error = namei(&nd)) != 0) 3627 return (error); 3628 vp = nd.ni_vp; 3629 if (vp != NULL) { 3630 NDFREE(&nd, NDF_ONLY_PNBUF); 3631 /* 3632 * XXX namei called with LOCKPARENT but not LOCKLEAF has 3633 * the strange behaviour of leaving the vnode unlocked 3634 * if the target is the same vnode as the parent. 3635 */ 3636 if (vp == nd.ni_dvp) 3637 vrele(nd.ni_dvp); 3638 else 3639 vput(nd.ni_dvp); 3640 vrele(vp); 3641 return (EEXIST); 3642 } 3643 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3644 NDFREE(&nd, NDF_ONLY_PNBUF); 3645 vput(nd.ni_dvp); 3646 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3647 return (error); 3648 goto restart; 3649 } 3650 VATTR_NULL(&vattr); 3651 vattr.va_type = VDIR; 3652 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask; 3653 #ifdef MAC 3654 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 3655 &vattr); 3656 if (error != 0) 3657 goto out; 3658 #endif 3659 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3660 #ifdef MAC 3661 out: 3662 #endif 3663 NDFREE(&nd, NDF_ONLY_PNBUF); 3664 vput(nd.ni_dvp); 3665 if (error == 0) 3666 vput(nd.ni_vp); 3667 vn_finished_write(mp); 3668 return (error); 3669 } 3670 3671 /* 3672 * Remove a directory file. 3673 */ 3674 #ifndef _SYS_SYSPROTO_H_ 3675 struct rmdir_args { 3676 char *path; 3677 }; 3678 #endif 3679 int 3680 sys_rmdir(td, uap) 3681 struct thread *td; 3682 struct rmdir_args /* { 3683 char *path; 3684 } */ *uap; 3685 { 3686 3687 return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE)); 3688 } 3689 3690 int 3691 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg) 3692 { 3693 struct mount *mp; 3694 struct vnode *vp; 3695 struct nameidata nd; 3696 cap_rights_t rights; 3697 int error; 3698 3699 restart: 3700 bwillwrite(); 3701 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1, 3702 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td); 3703 if ((error = namei(&nd)) != 0) 3704 return (error); 3705 vp = nd.ni_vp; 3706 if (vp->v_type != VDIR) { 3707 error = ENOTDIR; 3708 goto out; 3709 } 3710 /* 3711 * No rmdir "." please. 3712 */ 3713 if (nd.ni_dvp == vp) { 3714 error = EINVAL; 3715 goto out; 3716 } 3717 /* 3718 * The root of a mounted filesystem cannot be deleted. 3719 */ 3720 if (vp->v_vflag & VV_ROOT) { 3721 error = EBUSY; 3722 goto out; 3723 } 3724 #ifdef MAC 3725 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 3726 &nd.ni_cnd); 3727 if (error != 0) 3728 goto out; 3729 #endif 3730 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3731 NDFREE(&nd, NDF_ONLY_PNBUF); 3732 vput(vp); 3733 if (nd.ni_dvp == vp) 3734 vrele(nd.ni_dvp); 3735 else 3736 vput(nd.ni_dvp); 3737 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3738 return (error); 3739 goto restart; 3740 } 3741 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 3742 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3743 vn_finished_write(mp); 3744 out: 3745 NDFREE(&nd, NDF_ONLY_PNBUF); 3746 vput(vp); 3747 if (nd.ni_dvp == vp) 3748 vrele(nd.ni_dvp); 3749 else 3750 vput(nd.ni_dvp); 3751 return (error); 3752 } 3753 3754 #ifdef COMPAT_43 3755 /* 3756 * Read a block of directory entries in a filesystem independent format. 3757 */ 3758 #ifndef _SYS_SYSPROTO_H_ 3759 struct ogetdirentries_args { 3760 int fd; 3761 char *buf; 3762 u_int count; 3763 long *basep; 3764 }; 3765 #endif 3766 int 3767 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap) 3768 { 3769 long loff; 3770 int error; 3771 3772 error = kern_ogetdirentries(td, uap, &loff); 3773 if (error == 0) 3774 error = copyout(&loff, uap->basep, sizeof(long)); 3775 return (error); 3776 } 3777 3778 int 3779 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap, 3780 long *ploff) 3781 { 3782 struct vnode *vp; 3783 struct file *fp; 3784 struct uio auio, kuio; 3785 struct iovec aiov, kiov; 3786 struct dirent *dp, *edp; 3787 cap_rights_t rights; 3788 caddr_t dirbuf; 3789 int error, eofflag, readcnt; 3790 long loff; 3791 off_t foffset; 3792 3793 /* XXX arbitrary sanity limit on `count'. */ 3794 if (uap->count > 64 * 1024) 3795 return (EINVAL); 3796 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp); 3797 if (error != 0) 3798 return (error); 3799 if ((fp->f_flag & FREAD) == 0) { 3800 fdrop(fp, td); 3801 return (EBADF); 3802 } 3803 vp = fp->f_vnode; 3804 foffset = foffset_lock(fp, 0); 3805 unionread: 3806 if (vp->v_type != VDIR) { 3807 foffset_unlock(fp, foffset, 0); 3808 fdrop(fp, td); 3809 return (EINVAL); 3810 } 3811 aiov.iov_base = uap->buf; 3812 aiov.iov_len = uap->count; 3813 auio.uio_iov = &aiov; 3814 auio.uio_iovcnt = 1; 3815 auio.uio_rw = UIO_READ; 3816 auio.uio_segflg = UIO_USERSPACE; 3817 auio.uio_td = td; 3818 auio.uio_resid = uap->count; 3819 vn_lock(vp, LK_SHARED | LK_RETRY); 3820 loff = auio.uio_offset = foffset; 3821 #ifdef MAC 3822 error = mac_vnode_check_readdir(td->td_ucred, vp); 3823 if (error != 0) { 3824 VOP_UNLOCK(vp, 0); 3825 foffset_unlock(fp, foffset, FOF_NOUPDATE); 3826 fdrop(fp, td); 3827 return (error); 3828 } 3829 #endif 3830 # if (BYTE_ORDER != LITTLE_ENDIAN) 3831 if (vp->v_mount->mnt_maxsymlinklen <= 0) { 3832 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, 3833 NULL, NULL); 3834 foffset = auio.uio_offset; 3835 } else 3836 # endif 3837 { 3838 kuio = auio; 3839 kuio.uio_iov = &kiov; 3840 kuio.uio_segflg = UIO_SYSSPACE; 3841 kiov.iov_len = uap->count; 3842 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK); 3843 kiov.iov_base = dirbuf; 3844 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag, 3845 NULL, NULL); 3846 foffset = kuio.uio_offset; 3847 if (error == 0) { 3848 readcnt = uap->count - kuio.uio_resid; 3849 edp = (struct dirent *)&dirbuf[readcnt]; 3850 for (dp = (struct dirent *)dirbuf; dp < edp; ) { 3851 # if (BYTE_ORDER == LITTLE_ENDIAN) 3852 /* 3853 * The expected low byte of 3854 * dp->d_namlen is our dp->d_type. 3855 * The high MBZ byte of dp->d_namlen 3856 * is our dp->d_namlen. 3857 */ 3858 dp->d_type = dp->d_namlen; 3859 dp->d_namlen = 0; 3860 # else 3861 /* 3862 * The dp->d_type is the high byte 3863 * of the expected dp->d_namlen, 3864 * so must be zero'ed. 3865 */ 3866 dp->d_type = 0; 3867 # endif 3868 if (dp->d_reclen > 0) { 3869 dp = (struct dirent *) 3870 ((char *)dp + dp->d_reclen); 3871 } else { 3872 error = EIO; 3873 break; 3874 } 3875 } 3876 if (dp >= edp) 3877 error = uiomove(dirbuf, readcnt, &auio); 3878 } 3879 free(dirbuf, M_TEMP); 3880 } 3881 if (error != 0) { 3882 VOP_UNLOCK(vp, 0); 3883 foffset_unlock(fp, foffset, 0); 3884 fdrop(fp, td); 3885 return (error); 3886 } 3887 if (uap->count == auio.uio_resid && 3888 (vp->v_vflag & VV_ROOT) && 3889 (vp->v_mount->mnt_flag & MNT_UNION)) { 3890 struct vnode *tvp = vp; 3891 vp = vp->v_mount->mnt_vnodecovered; 3892 VREF(vp); 3893 fp->f_vnode = vp; 3894 fp->f_data = vp; 3895 foffset = 0; 3896 vput(tvp); 3897 goto unionread; 3898 } 3899 VOP_UNLOCK(vp, 0); 3900 foffset_unlock(fp, foffset, 0); 3901 fdrop(fp, td); 3902 td->td_retval[0] = uap->count - auio.uio_resid; 3903 if (error == 0) 3904 *ploff = loff; 3905 return (error); 3906 } 3907 #endif /* COMPAT_43 */ 3908 3909 /* 3910 * Read a block of directory entries in a filesystem independent format. 3911 */ 3912 #ifndef _SYS_SYSPROTO_H_ 3913 struct getdirentries_args { 3914 int fd; 3915 char *buf; 3916 u_int count; 3917 long *basep; 3918 }; 3919 #endif 3920 int 3921 sys_getdirentries(td, uap) 3922 struct thread *td; 3923 register struct getdirentries_args /* { 3924 int fd; 3925 char *buf; 3926 u_int count; 3927 long *basep; 3928 } */ *uap; 3929 { 3930 long base; 3931 int error; 3932 3933 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base, 3934 NULL, UIO_USERSPACE); 3935 if (error != 0) 3936 return (error); 3937 if (uap->basep != NULL) 3938 error = copyout(&base, uap->basep, sizeof(long)); 3939 return (error); 3940 } 3941 3942 int 3943 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count, 3944 long *basep, ssize_t *residp, enum uio_seg bufseg) 3945 { 3946 struct vnode *vp; 3947 struct file *fp; 3948 struct uio auio; 3949 struct iovec aiov; 3950 cap_rights_t rights; 3951 long loff; 3952 int error, eofflag; 3953 off_t foffset; 3954 3955 AUDIT_ARG_FD(fd); 3956 if (count > IOSIZE_MAX) 3957 return (EINVAL); 3958 auio.uio_resid = count; 3959 error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp); 3960 if (error != 0) 3961 return (error); 3962 if ((fp->f_flag & FREAD) == 0) { 3963 fdrop(fp, td); 3964 return (EBADF); 3965 } 3966 vp = fp->f_vnode; 3967 foffset = foffset_lock(fp, 0); 3968 unionread: 3969 if (vp->v_type != VDIR) { 3970 error = EINVAL; 3971 goto fail; 3972 } 3973 aiov.iov_base = buf; 3974 aiov.iov_len = count; 3975 auio.uio_iov = &aiov; 3976 auio.uio_iovcnt = 1; 3977 auio.uio_rw = UIO_READ; 3978 auio.uio_segflg = bufseg; 3979 auio.uio_td = td; 3980 vn_lock(vp, LK_SHARED | LK_RETRY); 3981 AUDIT_ARG_VNODE1(vp); 3982 loff = auio.uio_offset = foffset; 3983 #ifdef MAC 3984 error = mac_vnode_check_readdir(td->td_ucred, vp); 3985 if (error == 0) 3986 #endif 3987 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, 3988 NULL); 3989 foffset = auio.uio_offset; 3990 if (error != 0) { 3991 VOP_UNLOCK(vp, 0); 3992 goto fail; 3993 } 3994 if (count == auio.uio_resid && 3995 (vp->v_vflag & VV_ROOT) && 3996 (vp->v_mount->mnt_flag & MNT_UNION)) { 3997 struct vnode *tvp = vp; 3998 3999 vp = vp->v_mount->mnt_vnodecovered; 4000 VREF(vp); 4001 fp->f_vnode = vp; 4002 fp->f_data = vp; 4003 foffset = 0; 4004 vput(tvp); 4005 goto unionread; 4006 } 4007 VOP_UNLOCK(vp, 0); 4008 *basep = loff; 4009 if (residp != NULL) 4010 *residp = auio.uio_resid; 4011 td->td_retval[0] = count - auio.uio_resid; 4012 fail: 4013 foffset_unlock(fp, foffset, 0); 4014 fdrop(fp, td); 4015 return (error); 4016 } 4017 4018 #ifndef _SYS_SYSPROTO_H_ 4019 struct getdents_args { 4020 int fd; 4021 char *buf; 4022 size_t count; 4023 }; 4024 #endif 4025 int 4026 sys_getdents(td, uap) 4027 struct thread *td; 4028 register struct getdents_args /* { 4029 int fd; 4030 char *buf; 4031 u_int count; 4032 } */ *uap; 4033 { 4034 struct getdirentries_args ap; 4035 4036 ap.fd = uap->fd; 4037 ap.buf = uap->buf; 4038 ap.count = uap->count; 4039 ap.basep = NULL; 4040 return (sys_getdirentries(td, &ap)); 4041 } 4042 4043 /* 4044 * Set the mode mask for creation of filesystem nodes. 4045 */ 4046 #ifndef _SYS_SYSPROTO_H_ 4047 struct umask_args { 4048 int newmask; 4049 }; 4050 #endif 4051 int 4052 sys_umask(td, uap) 4053 struct thread *td; 4054 struct umask_args /* { 4055 int newmask; 4056 } */ *uap; 4057 { 4058 struct filedesc *fdp; 4059 4060 fdp = td->td_proc->p_fd; 4061 FILEDESC_XLOCK(fdp); 4062 td->td_retval[0] = fdp->fd_cmask; 4063 fdp->fd_cmask = uap->newmask & ALLPERMS; 4064 FILEDESC_XUNLOCK(fdp); 4065 return (0); 4066 } 4067 4068 /* 4069 * Void all references to file by ripping underlying filesystem away from 4070 * vnode. 4071 */ 4072 #ifndef _SYS_SYSPROTO_H_ 4073 struct revoke_args { 4074 char *path; 4075 }; 4076 #endif 4077 int 4078 sys_revoke(td, uap) 4079 struct thread *td; 4080 register struct revoke_args /* { 4081 char *path; 4082 } */ *uap; 4083 { 4084 struct vnode *vp; 4085 struct vattr vattr; 4086 struct nameidata nd; 4087 int error; 4088 4089 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4090 uap->path, td); 4091 if ((error = namei(&nd)) != 0) 4092 return (error); 4093 vp = nd.ni_vp; 4094 NDFREE(&nd, NDF_ONLY_PNBUF); 4095 if (vp->v_type != VCHR || vp->v_rdev == NULL) { 4096 error = EINVAL; 4097 goto out; 4098 } 4099 #ifdef MAC 4100 error = mac_vnode_check_revoke(td->td_ucred, vp); 4101 if (error != 0) 4102 goto out; 4103 #endif 4104 error = VOP_GETATTR(vp, &vattr, td->td_ucred); 4105 if (error != 0) 4106 goto out; 4107 if (td->td_ucred->cr_uid != vattr.va_uid) { 4108 error = priv_check(td, PRIV_VFS_ADMIN); 4109 if (error != 0) 4110 goto out; 4111 } 4112 if (vcount(vp) > 1) 4113 VOP_REVOKE(vp, REVOKEALL); 4114 out: 4115 vput(vp); 4116 return (error); 4117 } 4118 4119 /* 4120 * Convert a user file descriptor to a kernel file entry and check that, if it 4121 * is a capability, the correct rights are present. A reference on the file 4122 * entry is held upon returning. 4123 */ 4124 int 4125 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp) 4126 { 4127 struct file *fp; 4128 int error; 4129 4130 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL); 4131 if (error != 0) 4132 return (error); 4133 4134 /* 4135 * The file could be not of the vnode type, or it may be not 4136 * yet fully initialized, in which case the f_vnode pointer 4137 * may be set, but f_ops is still badfileops. E.g., 4138 * devfs_open() transiently create such situation to 4139 * facilitate csw d_fdopen(). 4140 * 4141 * Dupfdopen() handling in kern_openat() installs the 4142 * half-baked file into the process descriptor table, allowing 4143 * other thread to dereference it. Guard against the race by 4144 * checking f_ops. 4145 */ 4146 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) { 4147 fdrop(fp, td); 4148 return (EINVAL); 4149 } 4150 *fpp = fp; 4151 return (0); 4152 } 4153 4154 4155 /* 4156 * Get an (NFS) file handle. 4157 */ 4158 #ifndef _SYS_SYSPROTO_H_ 4159 struct lgetfh_args { 4160 char *fname; 4161 fhandle_t *fhp; 4162 }; 4163 #endif 4164 int 4165 sys_lgetfh(td, uap) 4166 struct thread *td; 4167 register struct lgetfh_args *uap; 4168 { 4169 struct nameidata nd; 4170 fhandle_t fh; 4171 register struct vnode *vp; 4172 int error; 4173 4174 error = priv_check(td, PRIV_VFS_GETFH); 4175 if (error != 0) 4176 return (error); 4177 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4178 uap->fname, td); 4179 error = namei(&nd); 4180 if (error != 0) 4181 return (error); 4182 NDFREE(&nd, NDF_ONLY_PNBUF); 4183 vp = nd.ni_vp; 4184 bzero(&fh, sizeof(fh)); 4185 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4186 error = VOP_VPTOFH(vp, &fh.fh_fid); 4187 vput(vp); 4188 if (error == 0) 4189 error = copyout(&fh, uap->fhp, sizeof (fh)); 4190 return (error); 4191 } 4192 4193 #ifndef _SYS_SYSPROTO_H_ 4194 struct getfh_args { 4195 char *fname; 4196 fhandle_t *fhp; 4197 }; 4198 #endif 4199 int 4200 sys_getfh(td, uap) 4201 struct thread *td; 4202 register struct getfh_args *uap; 4203 { 4204 struct nameidata nd; 4205 fhandle_t fh; 4206 register struct vnode *vp; 4207 int error; 4208 4209 error = priv_check(td, PRIV_VFS_GETFH); 4210 if (error != 0) 4211 return (error); 4212 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4213 uap->fname, td); 4214 error = namei(&nd); 4215 if (error != 0) 4216 return (error); 4217 NDFREE(&nd, NDF_ONLY_PNBUF); 4218 vp = nd.ni_vp; 4219 bzero(&fh, sizeof(fh)); 4220 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4221 error = VOP_VPTOFH(vp, &fh.fh_fid); 4222 vput(vp); 4223 if (error == 0) 4224 error = copyout(&fh, uap->fhp, sizeof (fh)); 4225 return (error); 4226 } 4227 4228 /* 4229 * syscall for the rpc.lockd to use to translate a NFS file handle into an 4230 * open descriptor. 4231 * 4232 * warning: do not remove the priv_check() call or this becomes one giant 4233 * security hole. 4234 */ 4235 #ifndef _SYS_SYSPROTO_H_ 4236 struct fhopen_args { 4237 const struct fhandle *u_fhp; 4238 int flags; 4239 }; 4240 #endif 4241 int 4242 sys_fhopen(td, uap) 4243 struct thread *td; 4244 struct fhopen_args /* { 4245 const struct fhandle *u_fhp; 4246 int flags; 4247 } */ *uap; 4248 { 4249 struct mount *mp; 4250 struct vnode *vp; 4251 struct fhandle fhp; 4252 struct file *fp; 4253 int fmode, error; 4254 int indx; 4255 4256 error = priv_check(td, PRIV_VFS_FHOPEN); 4257 if (error != 0) 4258 return (error); 4259 indx = -1; 4260 fmode = FFLAGS(uap->flags); 4261 /* why not allow a non-read/write open for our lockd? */ 4262 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT)) 4263 return (EINVAL); 4264 error = copyin(uap->u_fhp, &fhp, sizeof(fhp)); 4265 if (error != 0) 4266 return(error); 4267 /* find the mount point */ 4268 mp = vfs_busyfs(&fhp.fh_fsid); 4269 if (mp == NULL) 4270 return (ESTALE); 4271 /* now give me my vnode, it gets returned to me locked */ 4272 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp); 4273 vfs_unbusy(mp); 4274 if (error != 0) 4275 return (error); 4276 4277 error = falloc_noinstall(td, &fp); 4278 if (error != 0) { 4279 vput(vp); 4280 return (error); 4281 } 4282 /* 4283 * An extra reference on `fp' has been held for us by 4284 * falloc_noinstall(). 4285 */ 4286 4287 #ifdef INVARIANTS 4288 td->td_dupfd = -1; 4289 #endif 4290 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp); 4291 if (error != 0) { 4292 KASSERT(fp->f_ops == &badfileops, 4293 ("VOP_OPEN in fhopen() set f_ops")); 4294 KASSERT(td->td_dupfd < 0, 4295 ("fhopen() encountered fdopen()")); 4296 4297 vput(vp); 4298 goto bad; 4299 } 4300 #ifdef INVARIANTS 4301 td->td_dupfd = 0; 4302 #endif 4303 fp->f_vnode = vp; 4304 fp->f_seqcount = 1; 4305 finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp, 4306 &vnops); 4307 VOP_UNLOCK(vp, 0); 4308 if ((fmode & O_TRUNC) != 0) { 4309 error = fo_truncate(fp, 0, td->td_ucred, td); 4310 if (error != 0) 4311 goto bad; 4312 } 4313 4314 error = finstall(td, fp, &indx, fmode, NULL); 4315 bad: 4316 fdrop(fp, td); 4317 td->td_retval[0] = indx; 4318 return (error); 4319 } 4320 4321 /* 4322 * Stat an (NFS) file handle. 4323 */ 4324 #ifndef _SYS_SYSPROTO_H_ 4325 struct fhstat_args { 4326 struct fhandle *u_fhp; 4327 struct stat *sb; 4328 }; 4329 #endif 4330 int 4331 sys_fhstat(td, uap) 4332 struct thread *td; 4333 register struct fhstat_args /* { 4334 struct fhandle *u_fhp; 4335 struct stat *sb; 4336 } */ *uap; 4337 { 4338 struct stat sb; 4339 struct fhandle fh; 4340 int error; 4341 4342 error = copyin(uap->u_fhp, &fh, sizeof(fh)); 4343 if (error != 0) 4344 return (error); 4345 error = kern_fhstat(td, fh, &sb); 4346 if (error == 0) 4347 error = copyout(&sb, uap->sb, sizeof(sb)); 4348 return (error); 4349 } 4350 4351 int 4352 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb) 4353 { 4354 struct mount *mp; 4355 struct vnode *vp; 4356 int error; 4357 4358 error = priv_check(td, PRIV_VFS_FHSTAT); 4359 if (error != 0) 4360 return (error); 4361 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4362 return (ESTALE); 4363 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4364 vfs_unbusy(mp); 4365 if (error != 0) 4366 return (error); 4367 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td); 4368 vput(vp); 4369 return (error); 4370 } 4371 4372 /* 4373 * Implement fstatfs() for (NFS) file handles. 4374 */ 4375 #ifndef _SYS_SYSPROTO_H_ 4376 struct fhstatfs_args { 4377 struct fhandle *u_fhp; 4378 struct statfs *buf; 4379 }; 4380 #endif 4381 int 4382 sys_fhstatfs(td, uap) 4383 struct thread *td; 4384 struct fhstatfs_args /* { 4385 struct fhandle *u_fhp; 4386 struct statfs *buf; 4387 } */ *uap; 4388 { 4389 struct statfs *sfp; 4390 fhandle_t fh; 4391 int error; 4392 4393 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 4394 if (error != 0) 4395 return (error); 4396 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 4397 error = kern_fhstatfs(td, fh, sfp); 4398 if (error == 0) 4399 error = copyout(sfp, uap->buf, sizeof(*sfp)); 4400 free(sfp, M_STATFS); 4401 return (error); 4402 } 4403 4404 int 4405 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf) 4406 { 4407 struct statfs *sp; 4408 struct mount *mp; 4409 struct vnode *vp; 4410 int error; 4411 4412 error = priv_check(td, PRIV_VFS_FHSTATFS); 4413 if (error != 0) 4414 return (error); 4415 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4416 return (ESTALE); 4417 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4418 if (error != 0) { 4419 vfs_unbusy(mp); 4420 return (error); 4421 } 4422 vput(vp); 4423 error = prison_canseemount(td->td_ucred, mp); 4424 if (error != 0) 4425 goto out; 4426 #ifdef MAC 4427 error = mac_mount_check_stat(td->td_ucred, mp); 4428 if (error != 0) 4429 goto out; 4430 #endif 4431 /* 4432 * Set these in case the underlying filesystem fails to do so. 4433 */ 4434 sp = &mp->mnt_stat; 4435 sp->f_version = STATFS_VERSION; 4436 sp->f_namemax = NAME_MAX; 4437 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 4438 error = VFS_STATFS(mp, sp); 4439 if (error == 0) 4440 *buf = *sp; 4441 out: 4442 vfs_unbusy(mp); 4443 return (error); 4444 } 4445 4446 int 4447 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len) 4448 { 4449 struct file *fp; 4450 struct mount *mp; 4451 struct vnode *vp; 4452 cap_rights_t rights; 4453 off_t olen, ooffset; 4454 int error; 4455 4456 if (offset < 0 || len <= 0) 4457 return (EINVAL); 4458 /* Check for wrap. */ 4459 if (offset > OFF_MAX - len) 4460 return (EFBIG); 4461 error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp); 4462 if (error != 0) 4463 return (error); 4464 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) { 4465 error = ESPIPE; 4466 goto out; 4467 } 4468 if ((fp->f_flag & FWRITE) == 0) { 4469 error = EBADF; 4470 goto out; 4471 } 4472 if (fp->f_type != DTYPE_VNODE) { 4473 error = ENODEV; 4474 goto out; 4475 } 4476 vp = fp->f_vnode; 4477 if (vp->v_type != VREG) { 4478 error = ENODEV; 4479 goto out; 4480 } 4481 4482 /* Allocating blocks may take a long time, so iterate. */ 4483 for (;;) { 4484 olen = len; 4485 ooffset = offset; 4486 4487 bwillwrite(); 4488 mp = NULL; 4489 error = vn_start_write(vp, &mp, V_WAIT | PCATCH); 4490 if (error != 0) 4491 break; 4492 error = vn_lock(vp, LK_EXCLUSIVE); 4493 if (error != 0) { 4494 vn_finished_write(mp); 4495 break; 4496 } 4497 #ifdef MAC 4498 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp); 4499 if (error == 0) 4500 #endif 4501 error = VOP_ALLOCATE(vp, &offset, &len); 4502 VOP_UNLOCK(vp, 0); 4503 vn_finished_write(mp); 4504 4505 if (olen + ooffset != offset + len) { 4506 panic("offset + len changed from %jx/%jx to %jx/%jx", 4507 ooffset, olen, offset, len); 4508 } 4509 if (error != 0 || len == 0) 4510 break; 4511 KASSERT(olen > len, ("Iteration did not make progress?")); 4512 maybe_yield(); 4513 } 4514 out: 4515 fdrop(fp, td); 4516 return (error); 4517 } 4518 4519 int 4520 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap) 4521 { 4522 int error; 4523 4524 error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len); 4525 return (kern_posix_error(td, error)); 4526 } 4527 4528 /* 4529 * Unlike madvise(2), we do not make a best effort to remember every 4530 * possible caching hint. Instead, we remember the last setting with 4531 * the exception that we will allow POSIX_FADV_NORMAL to adjust the 4532 * region of any current setting. 4533 */ 4534 int 4535 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len, 4536 int advice) 4537 { 4538 struct fadvise_info *fa, *new; 4539 struct file *fp; 4540 struct vnode *vp; 4541 cap_rights_t rights; 4542 off_t end; 4543 int error; 4544 4545 if (offset < 0 || len < 0 || offset > OFF_MAX - len) 4546 return (EINVAL); 4547 switch (advice) { 4548 case POSIX_FADV_SEQUENTIAL: 4549 case POSIX_FADV_RANDOM: 4550 case POSIX_FADV_NOREUSE: 4551 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK); 4552 break; 4553 case POSIX_FADV_NORMAL: 4554 case POSIX_FADV_WILLNEED: 4555 case POSIX_FADV_DONTNEED: 4556 new = NULL; 4557 break; 4558 default: 4559 return (EINVAL); 4560 } 4561 /* XXX: CAP_POSIX_FADVISE? */ 4562 error = fget(td, fd, cap_rights_init(&rights), &fp); 4563 if (error != 0) 4564 goto out; 4565 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) { 4566 error = ESPIPE; 4567 goto out; 4568 } 4569 if (fp->f_type != DTYPE_VNODE) { 4570 error = ENODEV; 4571 goto out; 4572 } 4573 vp = fp->f_vnode; 4574 if (vp->v_type != VREG) { 4575 error = ENODEV; 4576 goto out; 4577 } 4578 if (len == 0) 4579 end = OFF_MAX; 4580 else 4581 end = offset + len - 1; 4582 switch (advice) { 4583 case POSIX_FADV_SEQUENTIAL: 4584 case POSIX_FADV_RANDOM: 4585 case POSIX_FADV_NOREUSE: 4586 /* 4587 * Try to merge any existing non-standard region with 4588 * this new region if possible, otherwise create a new 4589 * non-standard region for this request. 4590 */ 4591 mtx_pool_lock(mtxpool_sleep, fp); 4592 fa = fp->f_advice; 4593 if (fa != NULL && fa->fa_advice == advice && 4594 ((fa->fa_start <= end && fa->fa_end >= offset) || 4595 (end != OFF_MAX && fa->fa_start == end + 1) || 4596 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) { 4597 if (offset < fa->fa_start) 4598 fa->fa_start = offset; 4599 if (end > fa->fa_end) 4600 fa->fa_end = end; 4601 } else { 4602 new->fa_advice = advice; 4603 new->fa_start = offset; 4604 new->fa_end = end; 4605 fp->f_advice = new; 4606 new = fa; 4607 } 4608 mtx_pool_unlock(mtxpool_sleep, fp); 4609 break; 4610 case POSIX_FADV_NORMAL: 4611 /* 4612 * If a the "normal" region overlaps with an existing 4613 * non-standard region, trim or remove the 4614 * non-standard region. 4615 */ 4616 mtx_pool_lock(mtxpool_sleep, fp); 4617 fa = fp->f_advice; 4618 if (fa != NULL) { 4619 if (offset <= fa->fa_start && end >= fa->fa_end) { 4620 new = fa; 4621 fp->f_advice = NULL; 4622 } else if (offset <= fa->fa_start && 4623 end >= fa->fa_start) 4624 fa->fa_start = end + 1; 4625 else if (offset <= fa->fa_end && end >= fa->fa_end) 4626 fa->fa_end = offset - 1; 4627 else if (offset >= fa->fa_start && end <= fa->fa_end) { 4628 /* 4629 * If the "normal" region is a middle 4630 * portion of the existing 4631 * non-standard region, just remove 4632 * the whole thing rather than picking 4633 * one side or the other to 4634 * preserve. 4635 */ 4636 new = fa; 4637 fp->f_advice = NULL; 4638 } 4639 } 4640 mtx_pool_unlock(mtxpool_sleep, fp); 4641 break; 4642 case POSIX_FADV_WILLNEED: 4643 case POSIX_FADV_DONTNEED: 4644 error = VOP_ADVISE(vp, offset, end, advice); 4645 break; 4646 } 4647 out: 4648 if (fp != NULL) 4649 fdrop(fp, td); 4650 free(new, M_FADVISE); 4651 return (error); 4652 } 4653 4654 int 4655 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap) 4656 { 4657 int error; 4658 4659 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len, 4660 uap->advice); 4661 return (kern_posix_error(td, error)); 4662 } 4663