1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94 37 */ 38 39 #include <sys/cdefs.h> 40 __FBSDID("$FreeBSD$"); 41 42 #include "opt_capsicum.h" 43 #include "opt_ktrace.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/bio.h> 48 #include <sys/buf.h> 49 #include <sys/capsicum.h> 50 #include <sys/disk.h> 51 #include <sys/sysent.h> 52 #include <sys/malloc.h> 53 #include <sys/mount.h> 54 #include <sys/mutex.h> 55 #include <sys/sysproto.h> 56 #include <sys/namei.h> 57 #include <sys/filedesc.h> 58 #include <sys/kernel.h> 59 #include <sys/fcntl.h> 60 #include <sys/file.h> 61 #include <sys/filio.h> 62 #include <sys/limits.h> 63 #include <sys/linker.h> 64 #include <sys/rwlock.h> 65 #include <sys/sdt.h> 66 #include <sys/stat.h> 67 #include <sys/sx.h> 68 #include <sys/unistd.h> 69 #include <sys/vnode.h> 70 #include <sys/priv.h> 71 #include <sys/proc.h> 72 #include <sys/dirent.h> 73 #include <sys/jail.h> 74 #include <sys/syscallsubr.h> 75 #include <sys/sysctl.h> 76 #ifdef KTRACE 77 #include <sys/ktrace.h> 78 #endif 79 80 #include <machine/stdarg.h> 81 82 #include <security/audit/audit.h> 83 #include <security/mac/mac_framework.h> 84 85 #include <vm/vm.h> 86 #include <vm/vm_object.h> 87 #include <vm/vm_page.h> 88 #include <vm/uma.h> 89 90 #include <fs/devfs/devfs.h> 91 92 #include <ufs/ufs/quota.h> 93 94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information"); 95 96 static int kern_chflagsat(struct thread *td, int fd, const char *path, 97 enum uio_seg pathseg, u_long flags, int atflag); 98 static int setfflags(struct thread *td, struct vnode *, u_long); 99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *); 100 static int getutimens(const struct timespec *, enum uio_seg, 101 struct timespec *, int *); 102 static int setutimes(struct thread *td, struct vnode *, 103 const struct timespec *, int, int); 104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 105 struct thread *td); 106 static int kern_fhlinkat(struct thread *td, int fd, const char *path, 107 enum uio_seg pathseg, fhandle_t *fhp); 108 static int kern_getfhat(struct thread *td, int flags, int fd, 109 const char *path, enum uio_seg pathseg, fhandle_t *fhp); 110 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, 111 size_t count, struct thread *td); 112 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, 113 const char *path, enum uio_seg segflag); 114 115 static uint64_t 116 at2cnpflags(u_int at_flags, u_int mask) 117 { 118 u_int64_t res; 119 120 MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) != 121 (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)); 122 123 res = 0; 124 at_flags &= mask; 125 if ((at_flags & AT_BENEATH) != 0) 126 res |= BENEATH; 127 if ((at_flags & AT_RESOLVE_BENEATH) != 0) 128 res |= RBENEATH; 129 if ((at_flags & AT_SYMLINK_FOLLOW) != 0) 130 res |= FOLLOW; 131 /* NOFOLLOW is pseudo flag */ 132 if ((mask & AT_SYMLINK_NOFOLLOW) != 0) { 133 res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW : 134 FOLLOW; 135 } 136 return (res); 137 } 138 139 int 140 kern_sync(struct thread *td) 141 { 142 struct mount *mp, *nmp; 143 int save; 144 145 mtx_lock(&mountlist_mtx); 146 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 147 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) { 148 nmp = TAILQ_NEXT(mp, mnt_list); 149 continue; 150 } 151 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 152 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 153 save = curthread_pflags_set(TDP_SYNCIO); 154 vfs_periodic(mp, MNT_NOWAIT); 155 VFS_SYNC(mp, MNT_NOWAIT); 156 curthread_pflags_restore(save); 157 vn_finished_write(mp); 158 } 159 mtx_lock(&mountlist_mtx); 160 nmp = TAILQ_NEXT(mp, mnt_list); 161 vfs_unbusy(mp); 162 } 163 mtx_unlock(&mountlist_mtx); 164 return (0); 165 } 166 167 /* 168 * Sync each mounted filesystem. 169 */ 170 #ifndef _SYS_SYSPROTO_H_ 171 struct sync_args { 172 int dummy; 173 }; 174 #endif 175 /* ARGSUSED */ 176 int 177 sys_sync(struct thread *td, struct sync_args *uap) 178 { 179 180 return (kern_sync(td)); 181 } 182 183 /* 184 * Change filesystem quotas. 185 */ 186 #ifndef _SYS_SYSPROTO_H_ 187 struct quotactl_args { 188 char *path; 189 int cmd; 190 int uid; 191 caddr_t arg; 192 }; 193 #endif 194 int 195 sys_quotactl(struct thread *td, struct quotactl_args *uap) 196 { 197 struct mount *mp; 198 struct nameidata nd; 199 int error; 200 201 AUDIT_ARG_CMD(uap->cmd); 202 AUDIT_ARG_UID(uap->uid); 203 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS)) 204 return (EPERM); 205 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 206 uap->path, td); 207 if ((error = namei(&nd)) != 0) 208 return (error); 209 NDFREE(&nd, NDF_ONLY_PNBUF); 210 mp = nd.ni_vp->v_mount; 211 vfs_ref(mp); 212 vput(nd.ni_vp); 213 error = vfs_busy(mp, 0); 214 if (error != 0) { 215 vfs_rel(mp); 216 return (error); 217 } 218 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg); 219 220 /* 221 * Since quota on operation typically needs to open quota 222 * file, the Q_QUOTAON handler needs to unbusy the mount point 223 * before calling into namei. Otherwise, unmount might be 224 * started between two vfs_busy() invocations (first is our, 225 * second is from mount point cross-walk code in lookup()), 226 * causing deadlock. 227 * 228 * Require that Q_QUOTAON handles the vfs_busy() reference on 229 * its own, always returning with ubusied mount point. 230 */ 231 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON && 232 (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF) 233 vfs_unbusy(mp); 234 vfs_rel(mp); 235 return (error); 236 } 237 238 /* 239 * Used by statfs conversion routines to scale the block size up if 240 * necessary so that all of the block counts are <= 'max_size'. Note 241 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero 242 * value of 'n'. 243 */ 244 void 245 statfs_scale_blocks(struct statfs *sf, long max_size) 246 { 247 uint64_t count; 248 int shift; 249 250 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__)); 251 252 /* 253 * Attempt to scale the block counts to give a more accurate 254 * overview to userland of the ratio of free space to used 255 * space. To do this, find the largest block count and compute 256 * a divisor that lets it fit into a signed integer <= max_size. 257 */ 258 if (sf->f_bavail < 0) 259 count = -sf->f_bavail; 260 else 261 count = sf->f_bavail; 262 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count)); 263 if (count <= max_size) 264 return; 265 266 count >>= flsl(max_size); 267 shift = 0; 268 while (count > 0) { 269 shift++; 270 count >>=1; 271 } 272 273 sf->f_bsize <<= shift; 274 sf->f_blocks >>= shift; 275 sf->f_bfree >>= shift; 276 sf->f_bavail >>= shift; 277 } 278 279 static int 280 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf) 281 { 282 int error; 283 284 if (mp == NULL) 285 return (EBADF); 286 error = vfs_busy(mp, 0); 287 vfs_rel(mp); 288 if (error != 0) 289 return (error); 290 #ifdef MAC 291 error = mac_mount_check_stat(td->td_ucred, mp); 292 if (error != 0) 293 goto out; 294 #endif 295 error = VFS_STATFS(mp, buf); 296 if (error != 0) 297 goto out; 298 if (priv_check_cred_vfs_generation(td->td_ucred)) { 299 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 300 prison_enforce_statfs(td->td_ucred, mp, buf); 301 } 302 out: 303 vfs_unbusy(mp); 304 return (error); 305 } 306 307 /* 308 * Get filesystem statistics. 309 */ 310 #ifndef _SYS_SYSPROTO_H_ 311 struct statfs_args { 312 char *path; 313 struct statfs *buf; 314 }; 315 #endif 316 int 317 sys_statfs(struct thread *td, struct statfs_args *uap) 318 { 319 struct statfs *sfp; 320 int error; 321 322 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 323 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 324 if (error == 0) 325 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 326 free(sfp, M_STATFS); 327 return (error); 328 } 329 330 int 331 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg, 332 struct statfs *buf) 333 { 334 struct mount *mp; 335 struct nameidata nd; 336 int error; 337 338 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 339 pathseg, path, td); 340 error = namei(&nd); 341 if (error != 0) 342 return (error); 343 mp = nd.ni_vp->v_mount; 344 vfs_ref(mp); 345 NDFREE(&nd, NDF_ONLY_PNBUF); 346 vput(nd.ni_vp); 347 return (kern_do_statfs(td, mp, buf)); 348 } 349 350 /* 351 * Get filesystem statistics. 352 */ 353 #ifndef _SYS_SYSPROTO_H_ 354 struct fstatfs_args { 355 int fd; 356 struct statfs *buf; 357 }; 358 #endif 359 int 360 sys_fstatfs(struct thread *td, struct fstatfs_args *uap) 361 { 362 struct statfs *sfp; 363 int error; 364 365 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 366 error = kern_fstatfs(td, uap->fd, sfp); 367 if (error == 0) 368 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 369 free(sfp, M_STATFS); 370 return (error); 371 } 372 373 int 374 kern_fstatfs(struct thread *td, int fd, struct statfs *buf) 375 { 376 struct file *fp; 377 struct mount *mp; 378 struct vnode *vp; 379 int error; 380 381 AUDIT_ARG_FD(fd); 382 error = getvnode(td, fd, &cap_fstatfs_rights, &fp); 383 if (error != 0) 384 return (error); 385 vp = fp->f_vnode; 386 vn_lock(vp, LK_SHARED | LK_RETRY); 387 #ifdef AUDIT 388 AUDIT_ARG_VNODE1(vp); 389 #endif 390 mp = vp->v_mount; 391 if (mp != NULL) 392 vfs_ref(mp); 393 VOP_UNLOCK(vp); 394 fdrop(fp, td); 395 return (kern_do_statfs(td, mp, buf)); 396 } 397 398 /* 399 * Get statistics on all filesystems. 400 */ 401 #ifndef _SYS_SYSPROTO_H_ 402 struct getfsstat_args { 403 struct statfs *buf; 404 long bufsize; 405 int mode; 406 }; 407 #endif 408 int 409 sys_getfsstat(struct thread *td, struct getfsstat_args *uap) 410 { 411 size_t count; 412 int error; 413 414 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX) 415 return (EINVAL); 416 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count, 417 UIO_USERSPACE, uap->mode); 418 if (error == 0) 419 td->td_retval[0] = count; 420 return (error); 421 } 422 423 /* 424 * If (bufsize > 0 && bufseg == UIO_SYSSPACE) 425 * The caller is responsible for freeing memory which will be allocated 426 * in '*buf'. 427 */ 428 int 429 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize, 430 size_t *countp, enum uio_seg bufseg, int mode) 431 { 432 struct mount *mp, *nmp; 433 struct statfs *sfsp, *sp, *sptmp, *tofree; 434 size_t count, maxcount; 435 int error; 436 437 switch (mode) { 438 case MNT_WAIT: 439 case MNT_NOWAIT: 440 break; 441 default: 442 if (bufseg == UIO_SYSSPACE) 443 *buf = NULL; 444 return (EINVAL); 445 } 446 restart: 447 maxcount = bufsize / sizeof(struct statfs); 448 if (bufsize == 0) { 449 sfsp = NULL; 450 tofree = NULL; 451 } else if (bufseg == UIO_USERSPACE) { 452 sfsp = *buf; 453 tofree = NULL; 454 } else /* if (bufseg == UIO_SYSSPACE) */ { 455 count = 0; 456 mtx_lock(&mountlist_mtx); 457 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 458 count++; 459 } 460 mtx_unlock(&mountlist_mtx); 461 if (maxcount > count) 462 maxcount = count; 463 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs), 464 M_STATFS, M_WAITOK); 465 } 466 467 count = 0; 468 469 /* 470 * If there is no target buffer they only want the count. 471 * 472 * This could be TAILQ_FOREACH but it is open-coded to match the original 473 * code below. 474 */ 475 if (sfsp == NULL) { 476 mtx_lock(&mountlist_mtx); 477 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 478 if (prison_canseemount(td->td_ucred, mp) != 0) { 479 nmp = TAILQ_NEXT(mp, mnt_list); 480 continue; 481 } 482 #ifdef MAC 483 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 484 nmp = TAILQ_NEXT(mp, mnt_list); 485 continue; 486 } 487 #endif 488 count++; 489 nmp = TAILQ_NEXT(mp, mnt_list); 490 } 491 mtx_unlock(&mountlist_mtx); 492 *countp = count; 493 return (0); 494 } 495 496 /* 497 * They want the entire thing. 498 * 499 * Short-circuit the corner case of no room for anything, avoids 500 * relocking below. 501 */ 502 if (maxcount < 1) { 503 goto out; 504 } 505 506 mtx_lock(&mountlist_mtx); 507 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 508 if (prison_canseemount(td->td_ucred, mp) != 0) { 509 nmp = TAILQ_NEXT(mp, mnt_list); 510 continue; 511 } 512 #ifdef MAC 513 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 514 nmp = TAILQ_NEXT(mp, mnt_list); 515 continue; 516 } 517 #endif 518 if (mode == MNT_WAIT) { 519 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) { 520 /* 521 * If vfs_busy() failed, and MBF_NOWAIT 522 * wasn't passed, then the mp is gone. 523 * Furthermore, because of MBF_MNTLSTLOCK, 524 * the mountlist_mtx was dropped. We have 525 * no other choice than to start over. 526 */ 527 mtx_unlock(&mountlist_mtx); 528 free(tofree, M_STATFS); 529 goto restart; 530 } 531 } else { 532 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) { 533 nmp = TAILQ_NEXT(mp, mnt_list); 534 continue; 535 } 536 } 537 sp = &mp->mnt_stat; 538 /* 539 * If MNT_NOWAIT is specified, do not refresh 540 * the fsstat cache. 541 */ 542 if (mode != MNT_NOWAIT) { 543 error = VFS_STATFS(mp, sp); 544 if (error != 0) { 545 mtx_lock(&mountlist_mtx); 546 nmp = TAILQ_NEXT(mp, mnt_list); 547 vfs_unbusy(mp); 548 continue; 549 } 550 } 551 if (priv_check_cred_vfs_generation(td->td_ucred)) { 552 sptmp = malloc(sizeof(struct statfs), M_STATFS, 553 M_WAITOK); 554 *sptmp = *sp; 555 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0; 556 prison_enforce_statfs(td->td_ucred, mp, sptmp); 557 sp = sptmp; 558 } else 559 sptmp = NULL; 560 if (bufseg == UIO_SYSSPACE) { 561 bcopy(sp, sfsp, sizeof(*sp)); 562 free(sptmp, M_STATFS); 563 } else /* if (bufseg == UIO_USERSPACE) */ { 564 error = copyout(sp, sfsp, sizeof(*sp)); 565 free(sptmp, M_STATFS); 566 if (error != 0) { 567 vfs_unbusy(mp); 568 return (error); 569 } 570 } 571 sfsp++; 572 count++; 573 574 if (count == maxcount) { 575 vfs_unbusy(mp); 576 goto out; 577 } 578 579 mtx_lock(&mountlist_mtx); 580 nmp = TAILQ_NEXT(mp, mnt_list); 581 vfs_unbusy(mp); 582 } 583 mtx_unlock(&mountlist_mtx); 584 out: 585 *countp = count; 586 return (0); 587 } 588 589 #ifdef COMPAT_FREEBSD4 590 /* 591 * Get old format filesystem statistics. 592 */ 593 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *); 594 595 #ifndef _SYS_SYSPROTO_H_ 596 struct freebsd4_statfs_args { 597 char *path; 598 struct ostatfs *buf; 599 }; 600 #endif 601 int 602 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap) 603 { 604 struct ostatfs osb; 605 struct statfs *sfp; 606 int error; 607 608 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 609 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 610 if (error == 0) { 611 freebsd4_cvtstatfs(sfp, &osb); 612 error = copyout(&osb, uap->buf, sizeof(osb)); 613 } 614 free(sfp, M_STATFS); 615 return (error); 616 } 617 618 /* 619 * Get filesystem statistics. 620 */ 621 #ifndef _SYS_SYSPROTO_H_ 622 struct freebsd4_fstatfs_args { 623 int fd; 624 struct ostatfs *buf; 625 }; 626 #endif 627 int 628 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap) 629 { 630 struct ostatfs osb; 631 struct statfs *sfp; 632 int error; 633 634 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 635 error = kern_fstatfs(td, uap->fd, sfp); 636 if (error == 0) { 637 freebsd4_cvtstatfs(sfp, &osb); 638 error = copyout(&osb, uap->buf, sizeof(osb)); 639 } 640 free(sfp, M_STATFS); 641 return (error); 642 } 643 644 /* 645 * Get statistics on all filesystems. 646 */ 647 #ifndef _SYS_SYSPROTO_H_ 648 struct freebsd4_getfsstat_args { 649 struct ostatfs *buf; 650 long bufsize; 651 int mode; 652 }; 653 #endif 654 int 655 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap) 656 { 657 struct statfs *buf, *sp; 658 struct ostatfs osb; 659 size_t count, size; 660 int error; 661 662 if (uap->bufsize < 0) 663 return (EINVAL); 664 count = uap->bufsize / sizeof(struct ostatfs); 665 if (count > SIZE_MAX / sizeof(struct statfs)) 666 return (EINVAL); 667 size = count * sizeof(struct statfs); 668 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, 669 uap->mode); 670 if (error == 0) 671 td->td_retval[0] = count; 672 if (size != 0) { 673 sp = buf; 674 while (count != 0 && error == 0) { 675 freebsd4_cvtstatfs(sp, &osb); 676 error = copyout(&osb, uap->buf, sizeof(osb)); 677 sp++; 678 uap->buf++; 679 count--; 680 } 681 free(buf, M_STATFS); 682 } 683 return (error); 684 } 685 686 /* 687 * Implement fstatfs() for (NFS) file handles. 688 */ 689 #ifndef _SYS_SYSPROTO_H_ 690 struct freebsd4_fhstatfs_args { 691 struct fhandle *u_fhp; 692 struct ostatfs *buf; 693 }; 694 #endif 695 int 696 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap) 697 { 698 struct ostatfs osb; 699 struct statfs *sfp; 700 fhandle_t fh; 701 int error; 702 703 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 704 if (error != 0) 705 return (error); 706 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 707 error = kern_fhstatfs(td, fh, sfp); 708 if (error == 0) { 709 freebsd4_cvtstatfs(sfp, &osb); 710 error = copyout(&osb, uap->buf, sizeof(osb)); 711 } 712 free(sfp, M_STATFS); 713 return (error); 714 } 715 716 /* 717 * Convert a new format statfs structure to an old format statfs structure. 718 */ 719 static void 720 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp) 721 { 722 723 statfs_scale_blocks(nsp, LONG_MAX); 724 bzero(osp, sizeof(*osp)); 725 osp->f_bsize = nsp->f_bsize; 726 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX); 727 osp->f_blocks = nsp->f_blocks; 728 osp->f_bfree = nsp->f_bfree; 729 osp->f_bavail = nsp->f_bavail; 730 osp->f_files = MIN(nsp->f_files, LONG_MAX); 731 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX); 732 osp->f_owner = nsp->f_owner; 733 osp->f_type = nsp->f_type; 734 osp->f_flags = nsp->f_flags; 735 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX); 736 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX); 737 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX); 738 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX); 739 strlcpy(osp->f_fstypename, nsp->f_fstypename, 740 MIN(MFSNAMELEN, OMFSNAMELEN)); 741 strlcpy(osp->f_mntonname, nsp->f_mntonname, 742 MIN(MNAMELEN, OMNAMELEN)); 743 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 744 MIN(MNAMELEN, OMNAMELEN)); 745 osp->f_fsid = nsp->f_fsid; 746 } 747 #endif /* COMPAT_FREEBSD4 */ 748 749 #if defined(COMPAT_FREEBSD11) 750 /* 751 * Get old format filesystem statistics. 752 */ 753 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *); 754 755 int 756 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap) 757 { 758 struct freebsd11_statfs osb; 759 struct statfs *sfp; 760 int error; 761 762 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 763 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 764 if (error == 0) { 765 freebsd11_cvtstatfs(sfp, &osb); 766 error = copyout(&osb, uap->buf, sizeof(osb)); 767 } 768 free(sfp, M_STATFS); 769 return (error); 770 } 771 772 /* 773 * Get filesystem statistics. 774 */ 775 int 776 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap) 777 { 778 struct freebsd11_statfs osb; 779 struct statfs *sfp; 780 int error; 781 782 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 783 error = kern_fstatfs(td, uap->fd, sfp); 784 if (error == 0) { 785 freebsd11_cvtstatfs(sfp, &osb); 786 error = copyout(&osb, uap->buf, sizeof(osb)); 787 } 788 free(sfp, M_STATFS); 789 return (error); 790 } 791 792 /* 793 * Get statistics on all filesystems. 794 */ 795 int 796 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap) 797 { 798 struct freebsd11_statfs osb; 799 struct statfs *buf, *sp; 800 size_t count, size; 801 int error; 802 803 count = uap->bufsize / sizeof(struct ostatfs); 804 size = count * sizeof(struct statfs); 805 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, 806 uap->mode); 807 if (error == 0) 808 td->td_retval[0] = count; 809 if (size > 0) { 810 sp = buf; 811 while (count > 0 && error == 0) { 812 freebsd11_cvtstatfs(sp, &osb); 813 error = copyout(&osb, uap->buf, sizeof(osb)); 814 sp++; 815 uap->buf++; 816 count--; 817 } 818 free(buf, M_STATFS); 819 } 820 return (error); 821 } 822 823 /* 824 * Implement fstatfs() for (NFS) file handles. 825 */ 826 int 827 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap) 828 { 829 struct freebsd11_statfs osb; 830 struct statfs *sfp; 831 fhandle_t fh; 832 int error; 833 834 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 835 if (error) 836 return (error); 837 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 838 error = kern_fhstatfs(td, fh, sfp); 839 if (error == 0) { 840 freebsd11_cvtstatfs(sfp, &osb); 841 error = copyout(&osb, uap->buf, sizeof(osb)); 842 } 843 free(sfp, M_STATFS); 844 return (error); 845 } 846 847 /* 848 * Convert a new format statfs structure to an old format statfs structure. 849 */ 850 static void 851 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp) 852 { 853 854 bzero(osp, sizeof(*osp)); 855 osp->f_version = FREEBSD11_STATFS_VERSION; 856 osp->f_type = nsp->f_type; 857 osp->f_flags = nsp->f_flags; 858 osp->f_bsize = nsp->f_bsize; 859 osp->f_iosize = nsp->f_iosize; 860 osp->f_blocks = nsp->f_blocks; 861 osp->f_bfree = nsp->f_bfree; 862 osp->f_bavail = nsp->f_bavail; 863 osp->f_files = nsp->f_files; 864 osp->f_ffree = nsp->f_ffree; 865 osp->f_syncwrites = nsp->f_syncwrites; 866 osp->f_asyncwrites = nsp->f_asyncwrites; 867 osp->f_syncreads = nsp->f_syncreads; 868 osp->f_asyncreads = nsp->f_asyncreads; 869 osp->f_namemax = nsp->f_namemax; 870 osp->f_owner = nsp->f_owner; 871 osp->f_fsid = nsp->f_fsid; 872 strlcpy(osp->f_fstypename, nsp->f_fstypename, 873 MIN(MFSNAMELEN, sizeof(osp->f_fstypename))); 874 strlcpy(osp->f_mntonname, nsp->f_mntonname, 875 MIN(MNAMELEN, sizeof(osp->f_mntonname))); 876 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 877 MIN(MNAMELEN, sizeof(osp->f_mntfromname))); 878 } 879 #endif /* COMPAT_FREEBSD11 */ 880 881 /* 882 * Change current working directory to a given file descriptor. 883 */ 884 #ifndef _SYS_SYSPROTO_H_ 885 struct fchdir_args { 886 int fd; 887 }; 888 #endif 889 int 890 sys_fchdir(struct thread *td, struct fchdir_args *uap) 891 { 892 struct vnode *vp, *tdp; 893 struct mount *mp; 894 struct file *fp; 895 int error; 896 897 AUDIT_ARG_FD(uap->fd); 898 error = getvnode(td, uap->fd, &cap_fchdir_rights, 899 &fp); 900 if (error != 0) 901 return (error); 902 vp = fp->f_vnode; 903 vrefact(vp); 904 fdrop(fp, td); 905 vn_lock(vp, LK_SHARED | LK_RETRY); 906 AUDIT_ARG_VNODE1(vp); 907 error = change_dir(vp, td); 908 while (!error && (mp = vp->v_mountedhere) != NULL) { 909 if (vfs_busy(mp, 0)) 910 continue; 911 error = VFS_ROOT(mp, LK_SHARED, &tdp); 912 vfs_unbusy(mp); 913 if (error != 0) 914 break; 915 vput(vp); 916 vp = tdp; 917 } 918 if (error != 0) { 919 vput(vp); 920 return (error); 921 } 922 VOP_UNLOCK(vp); 923 pwd_chdir(td, vp); 924 return (0); 925 } 926 927 /* 928 * Change current working directory (``.''). 929 */ 930 #ifndef _SYS_SYSPROTO_H_ 931 struct chdir_args { 932 char *path; 933 }; 934 #endif 935 int 936 sys_chdir(struct thread *td, struct chdir_args *uap) 937 { 938 939 return (kern_chdir(td, uap->path, UIO_USERSPACE)); 940 } 941 942 int 943 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg) 944 { 945 struct nameidata nd; 946 int error; 947 948 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 949 pathseg, path, td); 950 if ((error = namei(&nd)) != 0) 951 return (error); 952 if ((error = change_dir(nd.ni_vp, td)) != 0) { 953 vput(nd.ni_vp); 954 NDFREE(&nd, NDF_ONLY_PNBUF); 955 return (error); 956 } 957 VOP_UNLOCK(nd.ni_vp); 958 NDFREE(&nd, NDF_ONLY_PNBUF); 959 pwd_chdir(td, nd.ni_vp); 960 return (0); 961 } 962 963 /* 964 * Change notion of root (``/'') directory. 965 */ 966 #ifndef _SYS_SYSPROTO_H_ 967 struct chroot_args { 968 char *path; 969 }; 970 #endif 971 int 972 sys_chroot(struct thread *td, struct chroot_args *uap) 973 { 974 struct nameidata nd; 975 int error; 976 977 error = priv_check(td, PRIV_VFS_CHROOT); 978 if (error != 0) 979 return (error); 980 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 981 UIO_USERSPACE, uap->path, td); 982 error = namei(&nd); 983 if (error != 0) 984 goto error; 985 error = change_dir(nd.ni_vp, td); 986 if (error != 0) 987 goto e_vunlock; 988 #ifdef MAC 989 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp); 990 if (error != 0) 991 goto e_vunlock; 992 #endif 993 VOP_UNLOCK(nd.ni_vp); 994 error = pwd_chroot(td, nd.ni_vp); 995 vrele(nd.ni_vp); 996 NDFREE(&nd, NDF_ONLY_PNBUF); 997 return (error); 998 e_vunlock: 999 vput(nd.ni_vp); 1000 error: 1001 NDFREE(&nd, NDF_ONLY_PNBUF); 1002 return (error); 1003 } 1004 1005 /* 1006 * Common routine for chroot and chdir. Callers must provide a locked vnode 1007 * instance. 1008 */ 1009 int 1010 change_dir(struct vnode *vp, struct thread *td) 1011 { 1012 #ifdef MAC 1013 int error; 1014 #endif 1015 1016 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked"); 1017 if (vp->v_type != VDIR) 1018 return (ENOTDIR); 1019 #ifdef MAC 1020 error = mac_vnode_check_chdir(td->td_ucred, vp); 1021 if (error != 0) 1022 return (error); 1023 #endif 1024 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td)); 1025 } 1026 1027 static __inline void 1028 flags_to_rights(int flags, cap_rights_t *rightsp) 1029 { 1030 1031 if (flags & O_EXEC) { 1032 cap_rights_set_one(rightsp, CAP_FEXECVE); 1033 } else { 1034 switch ((flags & O_ACCMODE)) { 1035 case O_RDONLY: 1036 cap_rights_set_one(rightsp, CAP_READ); 1037 break; 1038 case O_RDWR: 1039 cap_rights_set_one(rightsp, CAP_READ); 1040 /* FALLTHROUGH */ 1041 case O_WRONLY: 1042 cap_rights_set_one(rightsp, CAP_WRITE); 1043 if (!(flags & (O_APPEND | O_TRUNC))) 1044 cap_rights_set_one(rightsp, CAP_SEEK); 1045 break; 1046 } 1047 } 1048 1049 if (flags & O_CREAT) 1050 cap_rights_set_one(rightsp, CAP_CREATE); 1051 1052 if (flags & O_TRUNC) 1053 cap_rights_set_one(rightsp, CAP_FTRUNCATE); 1054 1055 if (flags & (O_SYNC | O_FSYNC)) 1056 cap_rights_set_one(rightsp, CAP_FSYNC); 1057 1058 if (flags & (O_EXLOCK | O_SHLOCK)) 1059 cap_rights_set_one(rightsp, CAP_FLOCK); 1060 } 1061 1062 /* 1063 * Check permissions, allocate an open file structure, and call the device 1064 * open routine if any. 1065 */ 1066 #ifndef _SYS_SYSPROTO_H_ 1067 struct open_args { 1068 char *path; 1069 int flags; 1070 int mode; 1071 }; 1072 #endif 1073 int 1074 sys_open(struct thread *td, struct open_args *uap) 1075 { 1076 1077 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1078 uap->flags, uap->mode)); 1079 } 1080 1081 #ifndef _SYS_SYSPROTO_H_ 1082 struct openat_args { 1083 int fd; 1084 char *path; 1085 int flag; 1086 int mode; 1087 }; 1088 #endif 1089 int 1090 sys_openat(struct thread *td, struct openat_args *uap) 1091 { 1092 1093 AUDIT_ARG_FD(uap->fd); 1094 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 1095 uap->mode)); 1096 } 1097 1098 int 1099 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg, 1100 int flags, int mode) 1101 { 1102 struct proc *p = td->td_proc; 1103 struct filedesc *fdp = p->p_fd; 1104 struct file *fp; 1105 struct vnode *vp; 1106 struct nameidata nd; 1107 cap_rights_t rights; 1108 int cmode, error, indx; 1109 1110 indx = -1; 1111 1112 AUDIT_ARG_FFLAGS(flags); 1113 AUDIT_ARG_MODE(mode); 1114 cap_rights_init_one(&rights, CAP_LOOKUP); 1115 flags_to_rights(flags, &rights); 1116 /* 1117 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 1118 * may be specified. 1119 */ 1120 if (flags & O_EXEC) { 1121 if (flags & O_ACCMODE) 1122 return (EINVAL); 1123 } else if ((flags & O_ACCMODE) == O_ACCMODE) { 1124 return (EINVAL); 1125 } else { 1126 flags = FFLAGS(flags); 1127 } 1128 1129 /* 1130 * Allocate a file structure. The descriptor to reference it 1131 * is allocated and set by finstall() below. 1132 */ 1133 error = falloc_noinstall(td, &fp); 1134 if (error != 0) 1135 return (error); 1136 /* 1137 * An extra reference on `fp' has been held for us by 1138 * falloc_noinstall(). 1139 */ 1140 /* Set the flags early so the finit in devfs can pick them up. */ 1141 fp->f_flag = flags & FMASK; 1142 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT; 1143 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 1144 &rights, td); 1145 td->td_dupfd = -1; /* XXX check for fdopen */ 1146 error = vn_open(&nd, &flags, cmode, fp); 1147 if (error != 0) { 1148 /* 1149 * If the vn_open replaced the method vector, something 1150 * wonderous happened deep below and we just pass it up 1151 * pretending we know what we do. 1152 */ 1153 if (error == ENXIO && fp->f_ops != &badfileops) 1154 goto success; 1155 1156 /* 1157 * Handle special fdopen() case. bleh. 1158 * 1159 * Don't do this for relative (capability) lookups; we don't 1160 * understand exactly what would happen, and we don't think 1161 * that it ever should. 1162 */ 1163 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 && 1164 (error == ENODEV || error == ENXIO) && 1165 td->td_dupfd >= 0) { 1166 error = dupfdopen(td, fdp, td->td_dupfd, flags, error, 1167 &indx); 1168 if (error == 0) 1169 goto success; 1170 } 1171 1172 goto bad; 1173 } 1174 td->td_dupfd = 0; 1175 NDFREE(&nd, NDF_ONLY_PNBUF); 1176 vp = nd.ni_vp; 1177 1178 /* 1179 * Store the vnode, for any f_type. Typically, the vnode use 1180 * count is decremented by direct call to vn_closefile() for 1181 * files that switched type in the cdevsw fdopen() method. 1182 */ 1183 fp->f_vnode = vp; 1184 /* 1185 * If the file wasn't claimed by devfs bind it to the normal 1186 * vnode operations here. 1187 */ 1188 if (fp->f_ops == &badfileops) { 1189 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo.")); 1190 finit_vnode(fp, flags, NULL, &vnops); 1191 } 1192 1193 VOP_UNLOCK(vp); 1194 if (flags & O_TRUNC) { 1195 error = fo_truncate(fp, 0, td->td_ucred, td); 1196 if (error != 0) 1197 goto bad; 1198 } 1199 success: 1200 /* 1201 * If we haven't already installed the FD (for dupfdopen), do so now. 1202 */ 1203 if (indx == -1) { 1204 struct filecaps *fcaps; 1205 1206 #ifdef CAPABILITIES 1207 if ((nd.ni_resflags & NIRES_STRICTREL) != 0) 1208 fcaps = &nd.ni_filecaps; 1209 else 1210 #endif 1211 fcaps = NULL; 1212 error = finstall(td, fp, &indx, flags, fcaps); 1213 /* On success finstall() consumes fcaps. */ 1214 if (error != 0) { 1215 filecaps_free(&nd.ni_filecaps); 1216 goto bad; 1217 } 1218 } else { 1219 filecaps_free(&nd.ni_filecaps); 1220 } 1221 1222 /* 1223 * Release our private reference, leaving the one associated with 1224 * the descriptor table intact. 1225 */ 1226 fdrop(fp, td); 1227 td->td_retval[0] = indx; 1228 return (0); 1229 bad: 1230 KASSERT(indx == -1, ("indx=%d, should be -1", indx)); 1231 fdrop(fp, td); 1232 return (error); 1233 } 1234 1235 #ifdef COMPAT_43 1236 /* 1237 * Create a file. 1238 */ 1239 #ifndef _SYS_SYSPROTO_H_ 1240 struct ocreat_args { 1241 char *path; 1242 int mode; 1243 }; 1244 #endif 1245 int 1246 ocreat(struct thread *td, struct ocreat_args *uap) 1247 { 1248 1249 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1250 O_WRONLY | O_CREAT | O_TRUNC, uap->mode)); 1251 } 1252 #endif /* COMPAT_43 */ 1253 1254 /* 1255 * Create a special file. 1256 */ 1257 #ifndef _SYS_SYSPROTO_H_ 1258 struct mknodat_args { 1259 int fd; 1260 char *path; 1261 mode_t mode; 1262 dev_t dev; 1263 }; 1264 #endif 1265 int 1266 sys_mknodat(struct thread *td, struct mknodat_args *uap) 1267 { 1268 1269 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1270 uap->dev)); 1271 } 1272 1273 #if defined(COMPAT_FREEBSD11) 1274 int 1275 freebsd11_mknod(struct thread *td, 1276 struct freebsd11_mknod_args *uap) 1277 { 1278 1279 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1280 uap->mode, uap->dev)); 1281 } 1282 1283 int 1284 freebsd11_mknodat(struct thread *td, 1285 struct freebsd11_mknodat_args *uap) 1286 { 1287 1288 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1289 uap->dev)); 1290 } 1291 #endif /* COMPAT_FREEBSD11 */ 1292 1293 int 1294 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg, 1295 int mode, dev_t dev) 1296 { 1297 struct vnode *vp; 1298 struct mount *mp; 1299 struct vattr vattr; 1300 struct nameidata nd; 1301 int error, whiteout = 0; 1302 1303 AUDIT_ARG_MODE(mode); 1304 AUDIT_ARG_DEV(dev); 1305 switch (mode & S_IFMT) { 1306 case S_IFCHR: 1307 case S_IFBLK: 1308 error = priv_check(td, PRIV_VFS_MKNOD_DEV); 1309 if (error == 0 && dev == VNOVAL) 1310 error = EINVAL; 1311 break; 1312 case S_IFWHT: 1313 error = priv_check(td, PRIV_VFS_MKNOD_WHT); 1314 break; 1315 case S_IFIFO: 1316 if (dev == 0) 1317 return (kern_mkfifoat(td, fd, path, pathseg, mode)); 1318 /* FALLTHROUGH */ 1319 default: 1320 error = EINVAL; 1321 break; 1322 } 1323 if (error != 0) 1324 return (error); 1325 restart: 1326 bwillwrite(); 1327 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1328 NOCACHE, pathseg, path, fd, &cap_mknodat_rights, 1329 td); 1330 if ((error = namei(&nd)) != 0) 1331 return (error); 1332 vp = nd.ni_vp; 1333 if (vp != NULL) { 1334 NDFREE(&nd, NDF_ONLY_PNBUF); 1335 if (vp == nd.ni_dvp) 1336 vrele(nd.ni_dvp); 1337 else 1338 vput(nd.ni_dvp); 1339 vrele(vp); 1340 return (EEXIST); 1341 } else { 1342 VATTR_NULL(&vattr); 1343 vattr.va_mode = (mode & ALLPERMS) & 1344 ~td->td_proc->p_fd->fd_cmask; 1345 vattr.va_rdev = dev; 1346 whiteout = 0; 1347 1348 switch (mode & S_IFMT) { 1349 case S_IFCHR: 1350 vattr.va_type = VCHR; 1351 break; 1352 case S_IFBLK: 1353 vattr.va_type = VBLK; 1354 break; 1355 case S_IFWHT: 1356 whiteout = 1; 1357 break; 1358 default: 1359 panic("kern_mknod: invalid mode"); 1360 } 1361 } 1362 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1363 NDFREE(&nd, NDF_ONLY_PNBUF); 1364 vput(nd.ni_dvp); 1365 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1366 return (error); 1367 goto restart; 1368 } 1369 #ifdef MAC 1370 if (error == 0 && !whiteout) 1371 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, 1372 &nd.ni_cnd, &vattr); 1373 #endif 1374 if (error == 0) { 1375 if (whiteout) 1376 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1377 else { 1378 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1379 &nd.ni_cnd, &vattr); 1380 if (error == 0) 1381 vput(nd.ni_vp); 1382 } 1383 } 1384 NDFREE(&nd, NDF_ONLY_PNBUF); 1385 vput(nd.ni_dvp); 1386 vn_finished_write(mp); 1387 return (error); 1388 } 1389 1390 /* 1391 * Create a named pipe. 1392 */ 1393 #ifndef _SYS_SYSPROTO_H_ 1394 struct mkfifo_args { 1395 char *path; 1396 int mode; 1397 }; 1398 #endif 1399 int 1400 sys_mkfifo(struct thread *td, struct mkfifo_args *uap) 1401 { 1402 1403 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1404 uap->mode)); 1405 } 1406 1407 #ifndef _SYS_SYSPROTO_H_ 1408 struct mkfifoat_args { 1409 int fd; 1410 char *path; 1411 mode_t mode; 1412 }; 1413 #endif 1414 int 1415 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap) 1416 { 1417 1418 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE, 1419 uap->mode)); 1420 } 1421 1422 int 1423 kern_mkfifoat(struct thread *td, int fd, const char *path, 1424 enum uio_seg pathseg, int mode) 1425 { 1426 struct mount *mp; 1427 struct vattr vattr; 1428 struct nameidata nd; 1429 int error; 1430 1431 AUDIT_ARG_MODE(mode); 1432 restart: 1433 bwillwrite(); 1434 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1435 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights, 1436 td); 1437 if ((error = namei(&nd)) != 0) 1438 return (error); 1439 if (nd.ni_vp != NULL) { 1440 NDFREE(&nd, NDF_ONLY_PNBUF); 1441 if (nd.ni_vp == nd.ni_dvp) 1442 vrele(nd.ni_dvp); 1443 else 1444 vput(nd.ni_dvp); 1445 vrele(nd.ni_vp); 1446 return (EEXIST); 1447 } 1448 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1449 NDFREE(&nd, NDF_ONLY_PNBUF); 1450 vput(nd.ni_dvp); 1451 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1452 return (error); 1453 goto restart; 1454 } 1455 VATTR_NULL(&vattr); 1456 vattr.va_type = VFIFO; 1457 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask; 1458 #ifdef MAC 1459 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1460 &vattr); 1461 if (error != 0) 1462 goto out; 1463 #endif 1464 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1465 if (error == 0) 1466 vput(nd.ni_vp); 1467 #ifdef MAC 1468 out: 1469 #endif 1470 vput(nd.ni_dvp); 1471 vn_finished_write(mp); 1472 NDFREE(&nd, NDF_ONLY_PNBUF); 1473 return (error); 1474 } 1475 1476 /* 1477 * Make a hard file link. 1478 */ 1479 #ifndef _SYS_SYSPROTO_H_ 1480 struct link_args { 1481 char *path; 1482 char *link; 1483 }; 1484 #endif 1485 int 1486 sys_link(struct thread *td, struct link_args *uap) 1487 { 1488 1489 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link, 1490 UIO_USERSPACE, FOLLOW)); 1491 } 1492 1493 #ifndef _SYS_SYSPROTO_H_ 1494 struct linkat_args { 1495 int fd1; 1496 char *path1; 1497 int fd2; 1498 char *path2; 1499 int flag; 1500 }; 1501 #endif 1502 int 1503 sys_linkat(struct thread *td, struct linkat_args *uap) 1504 { 1505 int flag; 1506 1507 flag = uap->flag; 1508 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH | 1509 AT_RESOLVE_BENEATH)) != 0) 1510 return (EINVAL); 1511 1512 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2, 1513 UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW | AT_BENEATH | 1514 AT_RESOLVE_BENEATH))); 1515 } 1516 1517 int hardlink_check_uid = 0; 1518 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW, 1519 &hardlink_check_uid, 0, 1520 "Unprivileged processes cannot create hard links to files owned by other " 1521 "users"); 1522 static int hardlink_check_gid = 0; 1523 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW, 1524 &hardlink_check_gid, 0, 1525 "Unprivileged processes cannot create hard links to files owned by other " 1526 "groups"); 1527 1528 static int 1529 can_hardlink(struct vnode *vp, struct ucred *cred) 1530 { 1531 struct vattr va; 1532 int error; 1533 1534 if (!hardlink_check_uid && !hardlink_check_gid) 1535 return (0); 1536 1537 error = VOP_GETATTR(vp, &va, cred); 1538 if (error != 0) 1539 return (error); 1540 1541 if (hardlink_check_uid && cred->cr_uid != va.va_uid) { 1542 error = priv_check_cred(cred, PRIV_VFS_LINK); 1543 if (error != 0) 1544 return (error); 1545 } 1546 1547 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) { 1548 error = priv_check_cred(cred, PRIV_VFS_LINK); 1549 if (error != 0) 1550 return (error); 1551 } 1552 1553 return (0); 1554 } 1555 1556 int 1557 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1, 1558 const char *path2, enum uio_seg segflag, int follow) 1559 { 1560 struct nameidata nd; 1561 int error; 1562 1563 do { 1564 bwillwrite(); 1565 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag, 1566 path1, fd1, &cap_linkat_source_rights, td); 1567 if ((error = namei(&nd)) != 0) 1568 return (error); 1569 NDFREE(&nd, NDF_ONLY_PNBUF); 1570 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag); 1571 } while (error == EAGAIN); 1572 return (error); 1573 } 1574 1575 static int 1576 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path, 1577 enum uio_seg segflag) 1578 { 1579 struct nameidata nd; 1580 struct mount *mp; 1581 int error; 1582 1583 if (vp->v_type == VDIR) { 1584 vrele(vp); 1585 return (EPERM); /* POSIX */ 1586 } 1587 NDINIT_ATRIGHTS(&nd, CREATE, 1588 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd, 1589 &cap_linkat_target_rights, td); 1590 if ((error = namei(&nd)) == 0) { 1591 if (nd.ni_vp != NULL) { 1592 NDFREE(&nd, NDF_ONLY_PNBUF); 1593 if (nd.ni_dvp == nd.ni_vp) 1594 vrele(nd.ni_dvp); 1595 else 1596 vput(nd.ni_dvp); 1597 vrele(nd.ni_vp); 1598 vrele(vp); 1599 return (EEXIST); 1600 } else if (nd.ni_dvp->v_mount != vp->v_mount) { 1601 /* 1602 * Cross-device link. No need to recheck 1603 * vp->v_type, since it cannot change, except 1604 * to VBAD. 1605 */ 1606 NDFREE(&nd, NDF_ONLY_PNBUF); 1607 vput(nd.ni_dvp); 1608 vrele(vp); 1609 return (EXDEV); 1610 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) { 1611 error = can_hardlink(vp, td->td_ucred); 1612 #ifdef MAC 1613 if (error == 0) 1614 error = mac_vnode_check_link(td->td_ucred, 1615 nd.ni_dvp, vp, &nd.ni_cnd); 1616 #endif 1617 if (error != 0) { 1618 vput(vp); 1619 vput(nd.ni_dvp); 1620 NDFREE(&nd, NDF_ONLY_PNBUF); 1621 return (error); 1622 } 1623 error = vn_start_write(vp, &mp, V_NOWAIT); 1624 if (error != 0) { 1625 vput(vp); 1626 vput(nd.ni_dvp); 1627 NDFREE(&nd, NDF_ONLY_PNBUF); 1628 error = vn_start_write(NULL, &mp, 1629 V_XSLEEP | PCATCH); 1630 if (error != 0) 1631 return (error); 1632 return (EAGAIN); 1633 } 1634 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1635 VOP_UNLOCK(vp); 1636 vput(nd.ni_dvp); 1637 vn_finished_write(mp); 1638 NDFREE(&nd, NDF_ONLY_PNBUF); 1639 } else { 1640 vput(nd.ni_dvp); 1641 NDFREE(&nd, NDF_ONLY_PNBUF); 1642 vrele(vp); 1643 return (EAGAIN); 1644 } 1645 } 1646 vrele(vp); 1647 return (error); 1648 } 1649 1650 /* 1651 * Make a symbolic link. 1652 */ 1653 #ifndef _SYS_SYSPROTO_H_ 1654 struct symlink_args { 1655 char *path; 1656 char *link; 1657 }; 1658 #endif 1659 int 1660 sys_symlink(struct thread *td, struct symlink_args *uap) 1661 { 1662 1663 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link, 1664 UIO_USERSPACE)); 1665 } 1666 1667 #ifndef _SYS_SYSPROTO_H_ 1668 struct symlinkat_args { 1669 char *path; 1670 int fd; 1671 char *path2; 1672 }; 1673 #endif 1674 int 1675 sys_symlinkat(struct thread *td, struct symlinkat_args *uap) 1676 { 1677 1678 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2, 1679 UIO_USERSPACE)); 1680 } 1681 1682 int 1683 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2, 1684 enum uio_seg segflg) 1685 { 1686 struct mount *mp; 1687 struct vattr vattr; 1688 const char *syspath; 1689 char *tmppath; 1690 struct nameidata nd; 1691 int error; 1692 1693 if (segflg == UIO_SYSSPACE) { 1694 syspath = path1; 1695 } else { 1696 tmppath = uma_zalloc(namei_zone, M_WAITOK); 1697 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0) 1698 goto out; 1699 syspath = tmppath; 1700 } 1701 AUDIT_ARG_TEXT(syspath); 1702 restart: 1703 bwillwrite(); 1704 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 1705 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights, 1706 td); 1707 if ((error = namei(&nd)) != 0) 1708 goto out; 1709 if (nd.ni_vp) { 1710 NDFREE(&nd, NDF_ONLY_PNBUF); 1711 if (nd.ni_vp == nd.ni_dvp) 1712 vrele(nd.ni_dvp); 1713 else 1714 vput(nd.ni_dvp); 1715 vrele(nd.ni_vp); 1716 error = EEXIST; 1717 goto out; 1718 } 1719 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1720 NDFREE(&nd, NDF_ONLY_PNBUF); 1721 vput(nd.ni_dvp); 1722 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1723 goto out; 1724 goto restart; 1725 } 1726 VATTR_NULL(&vattr); 1727 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask; 1728 #ifdef MAC 1729 vattr.va_type = VLNK; 1730 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1731 &vattr); 1732 if (error != 0) 1733 goto out2; 1734 #endif 1735 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath); 1736 if (error == 0) 1737 vput(nd.ni_vp); 1738 #ifdef MAC 1739 out2: 1740 #endif 1741 NDFREE(&nd, NDF_ONLY_PNBUF); 1742 vput(nd.ni_dvp); 1743 vn_finished_write(mp); 1744 out: 1745 if (segflg != UIO_SYSSPACE) 1746 uma_zfree(namei_zone, tmppath); 1747 return (error); 1748 } 1749 1750 /* 1751 * Delete a whiteout from the filesystem. 1752 */ 1753 #ifndef _SYS_SYSPROTO_H_ 1754 struct undelete_args { 1755 char *path; 1756 }; 1757 #endif 1758 int 1759 sys_undelete(struct thread *td, struct undelete_args *uap) 1760 { 1761 struct mount *mp; 1762 struct nameidata nd; 1763 int error; 1764 1765 restart: 1766 bwillwrite(); 1767 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1, 1768 UIO_USERSPACE, uap->path, td); 1769 error = namei(&nd); 1770 if (error != 0) 1771 return (error); 1772 1773 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1774 NDFREE(&nd, NDF_ONLY_PNBUF); 1775 if (nd.ni_vp == nd.ni_dvp) 1776 vrele(nd.ni_dvp); 1777 else 1778 vput(nd.ni_dvp); 1779 if (nd.ni_vp) 1780 vrele(nd.ni_vp); 1781 return (EEXIST); 1782 } 1783 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1784 NDFREE(&nd, NDF_ONLY_PNBUF); 1785 vput(nd.ni_dvp); 1786 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 1787 return (error); 1788 goto restart; 1789 } 1790 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE); 1791 NDFREE(&nd, NDF_ONLY_PNBUF); 1792 vput(nd.ni_dvp); 1793 vn_finished_write(mp); 1794 return (error); 1795 } 1796 1797 /* 1798 * Delete a name from the filesystem. 1799 */ 1800 #ifndef _SYS_SYSPROTO_H_ 1801 struct unlink_args { 1802 char *path; 1803 }; 1804 #endif 1805 int 1806 sys_unlink(struct thread *td, struct unlink_args *uap) 1807 { 1808 1809 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE, 1810 0, 0)); 1811 } 1812 1813 static int 1814 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd, 1815 int flag, enum uio_seg pathseg, ino_t oldinum) 1816 { 1817 1818 if ((flag & ~AT_REMOVEDIR) != 0) 1819 return (EINVAL); 1820 1821 if ((flag & AT_REMOVEDIR) != 0) 1822 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0)); 1823 1824 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0)); 1825 } 1826 1827 #ifndef _SYS_SYSPROTO_H_ 1828 struct unlinkat_args { 1829 int fd; 1830 char *path; 1831 int flag; 1832 }; 1833 #endif 1834 int 1835 sys_unlinkat(struct thread *td, struct unlinkat_args *uap) 1836 { 1837 1838 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag, 1839 UIO_USERSPACE, 0)); 1840 } 1841 1842 #ifndef _SYS_SYSPROTO_H_ 1843 struct funlinkat_args { 1844 int dfd; 1845 const char *path; 1846 int fd; 1847 int flag; 1848 }; 1849 #endif 1850 int 1851 sys_funlinkat(struct thread *td, struct funlinkat_args *uap) 1852 { 1853 1854 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag, 1855 UIO_USERSPACE, 0)); 1856 } 1857 1858 int 1859 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd, 1860 enum uio_seg pathseg, int flag, ino_t oldinum) 1861 { 1862 struct mount *mp; 1863 struct file *fp; 1864 struct vnode *vp; 1865 struct nameidata nd; 1866 struct stat sb; 1867 int error; 1868 1869 fp = NULL; 1870 if (fd != FD_NONE) { 1871 error = getvnode(td, fd, &cap_no_rights, &fp); 1872 if (error != 0) 1873 return (error); 1874 } 1875 1876 restart: 1877 bwillwrite(); 1878 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 | 1879 at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH), 1880 pathseg, path, dfd, &cap_unlinkat_rights, td); 1881 if ((error = namei(&nd)) != 0) { 1882 if (error == EINVAL) 1883 error = EPERM; 1884 goto fdout; 1885 } 1886 vp = nd.ni_vp; 1887 if (vp->v_type == VDIR && oldinum == 0) { 1888 error = EPERM; /* POSIX */ 1889 } else if (oldinum != 0 && 1890 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) && 1891 sb.st_ino != oldinum) { 1892 error = EIDRM; /* Identifier removed */ 1893 } else if (fp != NULL && fp->f_vnode != vp) { 1894 if (VN_IS_DOOMED(fp->f_vnode)) 1895 error = EBADF; 1896 else 1897 error = EDEADLK; 1898 } else { 1899 /* 1900 * The root of a mounted filesystem cannot be deleted. 1901 * 1902 * XXX: can this only be a VDIR case? 1903 */ 1904 if (vp->v_vflag & VV_ROOT) 1905 error = EBUSY; 1906 } 1907 if (error == 0) { 1908 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1909 NDFREE(&nd, NDF_ONLY_PNBUF); 1910 vput(nd.ni_dvp); 1911 if (vp == nd.ni_dvp) 1912 vrele(vp); 1913 else 1914 vput(vp); 1915 if ((error = vn_start_write(NULL, &mp, 1916 V_XSLEEP | PCATCH)) != 0) { 1917 goto fdout; 1918 } 1919 goto restart; 1920 } 1921 #ifdef MAC 1922 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 1923 &nd.ni_cnd); 1924 if (error != 0) 1925 goto out; 1926 #endif 1927 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 1928 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd); 1929 #ifdef MAC 1930 out: 1931 #endif 1932 vn_finished_write(mp); 1933 } 1934 NDFREE(&nd, NDF_ONLY_PNBUF); 1935 vput(nd.ni_dvp); 1936 if (vp == nd.ni_dvp) 1937 vrele(vp); 1938 else 1939 vput(vp); 1940 fdout: 1941 if (fp != NULL) 1942 fdrop(fp, td); 1943 return (error); 1944 } 1945 1946 /* 1947 * Reposition read/write file offset. 1948 */ 1949 #ifndef _SYS_SYSPROTO_H_ 1950 struct lseek_args { 1951 int fd; 1952 int pad; 1953 off_t offset; 1954 int whence; 1955 }; 1956 #endif 1957 int 1958 sys_lseek(struct thread *td, struct lseek_args *uap) 1959 { 1960 1961 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1962 } 1963 1964 int 1965 kern_lseek(struct thread *td, int fd, off_t offset, int whence) 1966 { 1967 struct file *fp; 1968 int error; 1969 1970 AUDIT_ARG_FD(fd); 1971 error = fget(td, fd, &cap_seek_rights, &fp); 1972 if (error != 0) 1973 return (error); 1974 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 1975 fo_seek(fp, offset, whence, td) : ESPIPE; 1976 fdrop(fp, td); 1977 return (error); 1978 } 1979 1980 #if defined(COMPAT_43) 1981 /* 1982 * Reposition read/write file offset. 1983 */ 1984 #ifndef _SYS_SYSPROTO_H_ 1985 struct olseek_args { 1986 int fd; 1987 long offset; 1988 int whence; 1989 }; 1990 #endif 1991 int 1992 olseek(struct thread *td, struct olseek_args *uap) 1993 { 1994 1995 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 1996 } 1997 #endif /* COMPAT_43 */ 1998 1999 #if defined(COMPAT_FREEBSD6) 2000 /* Version with the 'pad' argument */ 2001 int 2002 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap) 2003 { 2004 2005 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 2006 } 2007 #endif 2008 2009 /* 2010 * Check access permissions using passed credentials. 2011 */ 2012 static int 2013 vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 2014 struct thread *td) 2015 { 2016 accmode_t accmode; 2017 int error; 2018 2019 /* Flags == 0 means only check for existence. */ 2020 if (user_flags == 0) 2021 return (0); 2022 2023 accmode = 0; 2024 if (user_flags & R_OK) 2025 accmode |= VREAD; 2026 if (user_flags & W_OK) 2027 accmode |= VWRITE; 2028 if (user_flags & X_OK) 2029 accmode |= VEXEC; 2030 #ifdef MAC 2031 error = mac_vnode_check_access(cred, vp, accmode); 2032 if (error != 0) 2033 return (error); 2034 #endif 2035 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 2036 error = VOP_ACCESS(vp, accmode, cred, td); 2037 return (error); 2038 } 2039 2040 /* 2041 * Check access permissions using "real" credentials. 2042 */ 2043 #ifndef _SYS_SYSPROTO_H_ 2044 struct access_args { 2045 char *path; 2046 int amode; 2047 }; 2048 #endif 2049 int 2050 sys_access(struct thread *td, struct access_args *uap) 2051 { 2052 2053 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2054 0, uap->amode)); 2055 } 2056 2057 #ifndef _SYS_SYSPROTO_H_ 2058 struct faccessat_args { 2059 int dirfd; 2060 char *path; 2061 int amode; 2062 int flag; 2063 } 2064 #endif 2065 int 2066 sys_faccessat(struct thread *td, struct faccessat_args *uap) 2067 { 2068 2069 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 2070 uap->amode)); 2071 } 2072 2073 int 2074 kern_accessat(struct thread *td, int fd, const char *path, 2075 enum uio_seg pathseg, int flag, int amode) 2076 { 2077 struct ucred *cred, *usecred; 2078 struct vnode *vp; 2079 struct nameidata nd; 2080 int error; 2081 2082 if ((flag & ~(AT_EACCESS | AT_BENEATH | AT_RESOLVE_BENEATH)) != 0) 2083 return (EINVAL); 2084 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0) 2085 return (EINVAL); 2086 2087 /* 2088 * Create and modify a temporary credential instead of one that 2089 * is potentially shared (if we need one). 2090 */ 2091 cred = td->td_ucred; 2092 if ((flag & AT_EACCESS) == 0 && 2093 ((cred->cr_uid != cred->cr_ruid || 2094 cred->cr_rgid != cred->cr_groups[0]))) { 2095 usecred = crdup(cred); 2096 usecred->cr_uid = cred->cr_ruid; 2097 usecred->cr_groups[0] = cred->cr_rgid; 2098 td->td_ucred = usecred; 2099 } else 2100 usecred = cred; 2101 AUDIT_ARG_VALUE(amode); 2102 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | 2103 AUDITVNODE1 | at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH), 2104 pathseg, path, fd, &cap_fstat_rights, td); 2105 if ((error = namei(&nd)) != 0) 2106 goto out; 2107 vp = nd.ni_vp; 2108 2109 error = vn_access(vp, amode, usecred, td); 2110 NDFREE(&nd, NDF_ONLY_PNBUF); 2111 vput(vp); 2112 out: 2113 if (usecred != cred) { 2114 td->td_ucred = cred; 2115 crfree(usecred); 2116 } 2117 return (error); 2118 } 2119 2120 /* 2121 * Check access permissions using "effective" credentials. 2122 */ 2123 #ifndef _SYS_SYSPROTO_H_ 2124 struct eaccess_args { 2125 char *path; 2126 int amode; 2127 }; 2128 #endif 2129 int 2130 sys_eaccess(struct thread *td, struct eaccess_args *uap) 2131 { 2132 2133 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2134 AT_EACCESS, uap->amode)); 2135 } 2136 2137 #if defined(COMPAT_43) 2138 /* 2139 * Get file status; this version follows links. 2140 */ 2141 #ifndef _SYS_SYSPROTO_H_ 2142 struct ostat_args { 2143 char *path; 2144 struct ostat *ub; 2145 }; 2146 #endif 2147 int 2148 ostat(struct thread *td, struct ostat_args *uap) 2149 { 2150 struct stat sb; 2151 struct ostat osb; 2152 int error; 2153 2154 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2155 &sb, NULL); 2156 if (error != 0) 2157 return (error); 2158 cvtstat(&sb, &osb); 2159 return (copyout(&osb, uap->ub, sizeof (osb))); 2160 } 2161 2162 /* 2163 * Get file status; this version does not follow links. 2164 */ 2165 #ifndef _SYS_SYSPROTO_H_ 2166 struct olstat_args { 2167 char *path; 2168 struct ostat *ub; 2169 }; 2170 #endif 2171 int 2172 olstat(struct thread *td, struct olstat_args *uap) 2173 { 2174 struct stat sb; 2175 struct ostat osb; 2176 int error; 2177 2178 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2179 UIO_USERSPACE, &sb, NULL); 2180 if (error != 0) 2181 return (error); 2182 cvtstat(&sb, &osb); 2183 return (copyout(&osb, uap->ub, sizeof (osb))); 2184 } 2185 2186 /* 2187 * Convert from an old to a new stat structure. 2188 * XXX: many values are blindly truncated. 2189 */ 2190 void 2191 cvtstat(struct stat *st, struct ostat *ost) 2192 { 2193 2194 bzero(ost, sizeof(*ost)); 2195 ost->st_dev = st->st_dev; 2196 ost->st_ino = st->st_ino; 2197 ost->st_mode = st->st_mode; 2198 ost->st_nlink = st->st_nlink; 2199 ost->st_uid = st->st_uid; 2200 ost->st_gid = st->st_gid; 2201 ost->st_rdev = st->st_rdev; 2202 ost->st_size = MIN(st->st_size, INT32_MAX); 2203 ost->st_atim = st->st_atim; 2204 ost->st_mtim = st->st_mtim; 2205 ost->st_ctim = st->st_ctim; 2206 ost->st_blksize = st->st_blksize; 2207 ost->st_blocks = st->st_blocks; 2208 ost->st_flags = st->st_flags; 2209 ost->st_gen = st->st_gen; 2210 } 2211 #endif /* COMPAT_43 */ 2212 2213 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 2214 int ino64_trunc_error; 2215 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW, 2216 &ino64_trunc_error, 0, 2217 "Error on truncation of device, file or inode number, or link count"); 2218 2219 int 2220 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost) 2221 { 2222 2223 ost->st_dev = st->st_dev; 2224 if (ost->st_dev != st->st_dev) { 2225 switch (ino64_trunc_error) { 2226 default: 2227 /* 2228 * Since dev_t is almost raw, don't clamp to the 2229 * maximum for case 2, but ignore the error. 2230 */ 2231 break; 2232 case 1: 2233 return (EOVERFLOW); 2234 } 2235 } 2236 ost->st_ino = st->st_ino; 2237 if (ost->st_ino != st->st_ino) { 2238 switch (ino64_trunc_error) { 2239 default: 2240 case 0: 2241 break; 2242 case 1: 2243 return (EOVERFLOW); 2244 case 2: 2245 ost->st_ino = UINT32_MAX; 2246 break; 2247 } 2248 } 2249 ost->st_mode = st->st_mode; 2250 ost->st_nlink = st->st_nlink; 2251 if (ost->st_nlink != st->st_nlink) { 2252 switch (ino64_trunc_error) { 2253 default: 2254 case 0: 2255 break; 2256 case 1: 2257 return (EOVERFLOW); 2258 case 2: 2259 ost->st_nlink = UINT16_MAX; 2260 break; 2261 } 2262 } 2263 ost->st_uid = st->st_uid; 2264 ost->st_gid = st->st_gid; 2265 ost->st_rdev = st->st_rdev; 2266 if (ost->st_rdev != st->st_rdev) { 2267 switch (ino64_trunc_error) { 2268 default: 2269 break; 2270 case 1: 2271 return (EOVERFLOW); 2272 } 2273 } 2274 ost->st_atim = st->st_atim; 2275 ost->st_mtim = st->st_mtim; 2276 ost->st_ctim = st->st_ctim; 2277 ost->st_size = st->st_size; 2278 ost->st_blocks = st->st_blocks; 2279 ost->st_blksize = st->st_blksize; 2280 ost->st_flags = st->st_flags; 2281 ost->st_gen = st->st_gen; 2282 ost->st_lspare = 0; 2283 ost->st_birthtim = st->st_birthtim; 2284 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim), 2285 sizeof(*ost) - offsetof(struct freebsd11_stat, 2286 st_birthtim) - sizeof(ost->st_birthtim)); 2287 return (0); 2288 } 2289 2290 int 2291 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap) 2292 { 2293 struct stat sb; 2294 struct freebsd11_stat osb; 2295 int error; 2296 2297 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2298 &sb, NULL); 2299 if (error != 0) 2300 return (error); 2301 error = freebsd11_cvtstat(&sb, &osb); 2302 if (error == 0) 2303 error = copyout(&osb, uap->ub, sizeof(osb)); 2304 return (error); 2305 } 2306 2307 int 2308 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap) 2309 { 2310 struct stat sb; 2311 struct freebsd11_stat osb; 2312 int error; 2313 2314 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2315 UIO_USERSPACE, &sb, NULL); 2316 if (error != 0) 2317 return (error); 2318 error = freebsd11_cvtstat(&sb, &osb); 2319 if (error == 0) 2320 error = copyout(&osb, uap->ub, sizeof(osb)); 2321 return (error); 2322 } 2323 2324 int 2325 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap) 2326 { 2327 struct fhandle fh; 2328 struct stat sb; 2329 struct freebsd11_stat osb; 2330 int error; 2331 2332 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 2333 if (error != 0) 2334 return (error); 2335 error = kern_fhstat(td, fh, &sb); 2336 if (error != 0) 2337 return (error); 2338 error = freebsd11_cvtstat(&sb, &osb); 2339 if (error == 0) 2340 error = copyout(&osb, uap->sb, sizeof(osb)); 2341 return (error); 2342 } 2343 2344 int 2345 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap) 2346 { 2347 struct stat sb; 2348 struct freebsd11_stat osb; 2349 int error; 2350 2351 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2352 UIO_USERSPACE, &sb, NULL); 2353 if (error != 0) 2354 return (error); 2355 error = freebsd11_cvtstat(&sb, &osb); 2356 if (error == 0) 2357 error = copyout(&osb, uap->buf, sizeof(osb)); 2358 return (error); 2359 } 2360 #endif /* COMPAT_FREEBSD11 */ 2361 2362 /* 2363 * Get file status 2364 */ 2365 #ifndef _SYS_SYSPROTO_H_ 2366 struct fstatat_args { 2367 int fd; 2368 char *path; 2369 struct stat *buf; 2370 int flag; 2371 } 2372 #endif 2373 int 2374 sys_fstatat(struct thread *td, struct fstatat_args *uap) 2375 { 2376 struct stat sb; 2377 int error; 2378 2379 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2380 UIO_USERSPACE, &sb, NULL); 2381 if (error == 0) 2382 error = copyout(&sb, uap->buf, sizeof (sb)); 2383 return (error); 2384 } 2385 2386 int 2387 kern_statat(struct thread *td, int flag, int fd, const char *path, 2388 enum uio_seg pathseg, struct stat *sbp, 2389 void (*hook)(struct vnode *vp, struct stat *sbp)) 2390 { 2391 struct nameidata nd; 2392 int error; 2393 2394 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH | 2395 AT_RESOLVE_BENEATH)) != 0) 2396 return (EINVAL); 2397 2398 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_BENEATH | 2399 AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW) | LOCKSHARED | LOCKLEAF | 2400 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td); 2401 2402 if ((error = namei(&nd)) != 0) 2403 return (error); 2404 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td); 2405 if (error == 0) { 2406 if (__predict_false(hook != NULL)) 2407 hook(nd.ni_vp, sbp); 2408 } 2409 NDFREE(&nd, NDF_ONLY_PNBUF); 2410 vput(nd.ni_vp); 2411 #ifdef __STAT_TIME_T_EXT 2412 sbp->st_atim_ext = 0; 2413 sbp->st_mtim_ext = 0; 2414 sbp->st_ctim_ext = 0; 2415 sbp->st_btim_ext = 0; 2416 #endif 2417 #ifdef KTRACE 2418 if (KTRPOINT(td, KTR_STRUCT)) 2419 ktrstat_error(sbp, error); 2420 #endif 2421 return (error); 2422 } 2423 2424 #if defined(COMPAT_FREEBSD11) 2425 /* 2426 * Implementation of the NetBSD [l]stat() functions. 2427 */ 2428 void 2429 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb) 2430 { 2431 2432 bzero(nsb, sizeof(*nsb)); 2433 nsb->st_dev = sb->st_dev; 2434 nsb->st_ino = sb->st_ino; 2435 nsb->st_mode = sb->st_mode; 2436 nsb->st_nlink = sb->st_nlink; 2437 nsb->st_uid = sb->st_uid; 2438 nsb->st_gid = sb->st_gid; 2439 nsb->st_rdev = sb->st_rdev; 2440 nsb->st_atim = sb->st_atim; 2441 nsb->st_mtim = sb->st_mtim; 2442 nsb->st_ctim = sb->st_ctim; 2443 nsb->st_size = sb->st_size; 2444 nsb->st_blocks = sb->st_blocks; 2445 nsb->st_blksize = sb->st_blksize; 2446 nsb->st_flags = sb->st_flags; 2447 nsb->st_gen = sb->st_gen; 2448 nsb->st_birthtim = sb->st_birthtim; 2449 } 2450 2451 #ifndef _SYS_SYSPROTO_H_ 2452 struct freebsd11_nstat_args { 2453 char *path; 2454 struct nstat *ub; 2455 }; 2456 #endif 2457 int 2458 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap) 2459 { 2460 struct stat sb; 2461 struct nstat nsb; 2462 int error; 2463 2464 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, 2465 &sb, NULL); 2466 if (error != 0) 2467 return (error); 2468 freebsd11_cvtnstat(&sb, &nsb); 2469 return (copyout(&nsb, uap->ub, sizeof (nsb))); 2470 } 2471 2472 /* 2473 * NetBSD lstat. Get file status; this version does not follow links. 2474 */ 2475 #ifndef _SYS_SYSPROTO_H_ 2476 struct freebsd11_nlstat_args { 2477 char *path; 2478 struct nstat *ub; 2479 }; 2480 #endif 2481 int 2482 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap) 2483 { 2484 struct stat sb; 2485 struct nstat nsb; 2486 int error; 2487 2488 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2489 UIO_USERSPACE, &sb, NULL); 2490 if (error != 0) 2491 return (error); 2492 freebsd11_cvtnstat(&sb, &nsb); 2493 return (copyout(&nsb, uap->ub, sizeof (nsb))); 2494 } 2495 #endif /* COMPAT_FREEBSD11 */ 2496 2497 /* 2498 * Get configurable pathname variables. 2499 */ 2500 #ifndef _SYS_SYSPROTO_H_ 2501 struct pathconf_args { 2502 char *path; 2503 int name; 2504 }; 2505 #endif 2506 int 2507 sys_pathconf(struct thread *td, struct pathconf_args *uap) 2508 { 2509 long value; 2510 int error; 2511 2512 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW, 2513 &value); 2514 if (error == 0) 2515 td->td_retval[0] = value; 2516 return (error); 2517 } 2518 2519 #ifndef _SYS_SYSPROTO_H_ 2520 struct lpathconf_args { 2521 char *path; 2522 int name; 2523 }; 2524 #endif 2525 int 2526 sys_lpathconf(struct thread *td, struct lpathconf_args *uap) 2527 { 2528 long value; 2529 int error; 2530 2531 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, 2532 NOFOLLOW, &value); 2533 if (error == 0) 2534 td->td_retval[0] = value; 2535 return (error); 2536 } 2537 2538 int 2539 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg, 2540 int name, u_long flags, long *valuep) 2541 { 2542 struct nameidata nd; 2543 int error; 2544 2545 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags, 2546 pathseg, path, td); 2547 if ((error = namei(&nd)) != 0) 2548 return (error); 2549 NDFREE(&nd, NDF_ONLY_PNBUF); 2550 2551 error = VOP_PATHCONF(nd.ni_vp, name, valuep); 2552 vput(nd.ni_vp); 2553 return (error); 2554 } 2555 2556 /* 2557 * Return target name of a symbolic link. 2558 */ 2559 #ifndef _SYS_SYSPROTO_H_ 2560 struct readlink_args { 2561 char *path; 2562 char *buf; 2563 size_t count; 2564 }; 2565 #endif 2566 int 2567 sys_readlink(struct thread *td, struct readlink_args *uap) 2568 { 2569 2570 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2571 uap->buf, UIO_USERSPACE, uap->count)); 2572 } 2573 #ifndef _SYS_SYSPROTO_H_ 2574 struct readlinkat_args { 2575 int fd; 2576 char *path; 2577 char *buf; 2578 size_t bufsize; 2579 }; 2580 #endif 2581 int 2582 sys_readlinkat(struct thread *td, struct readlinkat_args *uap) 2583 { 2584 2585 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE, 2586 uap->buf, UIO_USERSPACE, uap->bufsize)); 2587 } 2588 2589 int 2590 kern_readlinkat(struct thread *td, int fd, const char *path, 2591 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count) 2592 { 2593 struct vnode *vp; 2594 struct nameidata nd; 2595 int error; 2596 2597 if (count > IOSIZE_MAX) 2598 return (EINVAL); 2599 2600 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 2601 pathseg, path, fd, td); 2602 2603 if ((error = namei(&nd)) != 0) 2604 return (error); 2605 NDFREE(&nd, NDF_ONLY_PNBUF); 2606 vp = nd.ni_vp; 2607 2608 error = kern_readlink_vp(vp, buf, bufseg, count, td); 2609 vput(vp); 2610 2611 return (error); 2612 } 2613 2614 /* 2615 * Helper function to readlink from a vnode 2616 */ 2617 static int 2618 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count, 2619 struct thread *td) 2620 { 2621 struct iovec aiov; 2622 struct uio auio; 2623 int error; 2624 2625 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked"); 2626 #ifdef MAC 2627 error = mac_vnode_check_readlink(td->td_ucred, vp); 2628 if (error != 0) 2629 return (error); 2630 #endif 2631 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0) 2632 return (EINVAL); 2633 2634 aiov.iov_base = buf; 2635 aiov.iov_len = count; 2636 auio.uio_iov = &aiov; 2637 auio.uio_iovcnt = 1; 2638 auio.uio_offset = 0; 2639 auio.uio_rw = UIO_READ; 2640 auio.uio_segflg = bufseg; 2641 auio.uio_td = td; 2642 auio.uio_resid = count; 2643 error = VOP_READLINK(vp, &auio, td->td_ucred); 2644 td->td_retval[0] = count - auio.uio_resid; 2645 return (error); 2646 } 2647 2648 /* 2649 * Common implementation code for chflags() and fchflags(). 2650 */ 2651 static int 2652 setfflags(struct thread *td, struct vnode *vp, u_long flags) 2653 { 2654 struct mount *mp; 2655 struct vattr vattr; 2656 int error; 2657 2658 /* We can't support the value matching VNOVAL. */ 2659 if (flags == VNOVAL) 2660 return (EOPNOTSUPP); 2661 2662 /* 2663 * Prevent non-root users from setting flags on devices. When 2664 * a device is reused, users can retain ownership of the device 2665 * if they are allowed to set flags and programs assume that 2666 * chown can't fail when done as root. 2667 */ 2668 if (vp->v_type == VCHR || vp->v_type == VBLK) { 2669 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV); 2670 if (error != 0) 2671 return (error); 2672 } 2673 2674 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2675 return (error); 2676 VATTR_NULL(&vattr); 2677 vattr.va_flags = flags; 2678 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2679 #ifdef MAC 2680 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags); 2681 if (error == 0) 2682 #endif 2683 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 2684 VOP_UNLOCK(vp); 2685 vn_finished_write(mp); 2686 return (error); 2687 } 2688 2689 /* 2690 * Change flags of a file given a path name. 2691 */ 2692 #ifndef _SYS_SYSPROTO_H_ 2693 struct chflags_args { 2694 const char *path; 2695 u_long flags; 2696 }; 2697 #endif 2698 int 2699 sys_chflags(struct thread *td, struct chflags_args *uap) 2700 { 2701 2702 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2703 uap->flags, 0)); 2704 } 2705 2706 #ifndef _SYS_SYSPROTO_H_ 2707 struct chflagsat_args { 2708 int fd; 2709 const char *path; 2710 u_long flags; 2711 int atflag; 2712 } 2713 #endif 2714 int 2715 sys_chflagsat(struct thread *td, struct chflagsat_args *uap) 2716 { 2717 2718 if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH | 2719 AT_RESOLVE_BENEATH)) != 0) 2720 return (EINVAL); 2721 2722 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE, 2723 uap->flags, uap->atflag)); 2724 } 2725 2726 /* 2727 * Same as chflags() but doesn't follow symlinks. 2728 */ 2729 #ifndef _SYS_SYSPROTO_H_ 2730 struct lchflags_args { 2731 const char *path; 2732 u_long flags; 2733 }; 2734 #endif 2735 int 2736 sys_lchflags(struct thread *td, struct lchflags_args *uap) 2737 { 2738 2739 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2740 uap->flags, AT_SYMLINK_NOFOLLOW)); 2741 } 2742 2743 static int 2744 kern_chflagsat(struct thread *td, int fd, const char *path, 2745 enum uio_seg pathseg, u_long flags, int atflag) 2746 { 2747 struct nameidata nd; 2748 int error; 2749 2750 AUDIT_ARG_FFLAGS(flags); 2751 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW | 2752 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd, 2753 &cap_fchflags_rights, td); 2754 if ((error = namei(&nd)) != 0) 2755 return (error); 2756 NDFREE(&nd, NDF_ONLY_PNBUF); 2757 error = setfflags(td, nd.ni_vp, flags); 2758 vrele(nd.ni_vp); 2759 return (error); 2760 } 2761 2762 /* 2763 * Change flags of a file given a file descriptor. 2764 */ 2765 #ifndef _SYS_SYSPROTO_H_ 2766 struct fchflags_args { 2767 int fd; 2768 u_long flags; 2769 }; 2770 #endif 2771 int 2772 sys_fchflags(struct thread *td, struct fchflags_args *uap) 2773 { 2774 struct file *fp; 2775 int error; 2776 2777 AUDIT_ARG_FD(uap->fd); 2778 AUDIT_ARG_FFLAGS(uap->flags); 2779 error = getvnode(td, uap->fd, &cap_fchflags_rights, 2780 &fp); 2781 if (error != 0) 2782 return (error); 2783 #ifdef AUDIT 2784 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 2785 AUDIT_ARG_VNODE1(fp->f_vnode); 2786 VOP_UNLOCK(fp->f_vnode); 2787 #endif 2788 error = setfflags(td, fp->f_vnode, uap->flags); 2789 fdrop(fp, td); 2790 return (error); 2791 } 2792 2793 /* 2794 * Common implementation code for chmod(), lchmod() and fchmod(). 2795 */ 2796 int 2797 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode) 2798 { 2799 struct mount *mp; 2800 struct vattr vattr; 2801 int error; 2802 2803 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2804 return (error); 2805 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2806 VATTR_NULL(&vattr); 2807 vattr.va_mode = mode & ALLPERMS; 2808 #ifdef MAC 2809 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode); 2810 if (error == 0) 2811 #endif 2812 error = VOP_SETATTR(vp, &vattr, cred); 2813 VOP_UNLOCK(vp); 2814 vn_finished_write(mp); 2815 return (error); 2816 } 2817 2818 /* 2819 * Change mode of a file given path name. 2820 */ 2821 #ifndef _SYS_SYSPROTO_H_ 2822 struct chmod_args { 2823 char *path; 2824 int mode; 2825 }; 2826 #endif 2827 int 2828 sys_chmod(struct thread *td, struct chmod_args *uap) 2829 { 2830 2831 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2832 uap->mode, 0)); 2833 } 2834 2835 #ifndef _SYS_SYSPROTO_H_ 2836 struct fchmodat_args { 2837 int dirfd; 2838 char *path; 2839 mode_t mode; 2840 int flag; 2841 } 2842 #endif 2843 int 2844 sys_fchmodat(struct thread *td, struct fchmodat_args *uap) 2845 { 2846 2847 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH | 2848 AT_RESOLVE_BENEATH)) != 0) 2849 return (EINVAL); 2850 2851 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE, 2852 uap->mode, uap->flag)); 2853 } 2854 2855 /* 2856 * Change mode of a file given path name (don't follow links.) 2857 */ 2858 #ifndef _SYS_SYSPROTO_H_ 2859 struct lchmod_args { 2860 char *path; 2861 int mode; 2862 }; 2863 #endif 2864 int 2865 sys_lchmod(struct thread *td, struct lchmod_args *uap) 2866 { 2867 2868 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2869 uap->mode, AT_SYMLINK_NOFOLLOW)); 2870 } 2871 2872 int 2873 kern_fchmodat(struct thread *td, int fd, const char *path, 2874 enum uio_seg pathseg, mode_t mode, int flag) 2875 { 2876 struct nameidata nd; 2877 int error; 2878 2879 AUDIT_ARG_MODE(mode); 2880 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW | 2881 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd, 2882 &cap_fchmod_rights, td); 2883 if ((error = namei(&nd)) != 0) 2884 return (error); 2885 NDFREE(&nd, NDF_ONLY_PNBUF); 2886 error = setfmode(td, td->td_ucred, nd.ni_vp, mode); 2887 vrele(nd.ni_vp); 2888 return (error); 2889 } 2890 2891 /* 2892 * Change mode of a file given a file descriptor. 2893 */ 2894 #ifndef _SYS_SYSPROTO_H_ 2895 struct fchmod_args { 2896 int fd; 2897 int mode; 2898 }; 2899 #endif 2900 int 2901 sys_fchmod(struct thread *td, struct fchmod_args *uap) 2902 { 2903 struct file *fp; 2904 int error; 2905 2906 AUDIT_ARG_FD(uap->fd); 2907 AUDIT_ARG_MODE(uap->mode); 2908 2909 error = fget(td, uap->fd, &cap_fchmod_rights, &fp); 2910 if (error != 0) 2911 return (error); 2912 error = fo_chmod(fp, uap->mode, td->td_ucred, td); 2913 fdrop(fp, td); 2914 return (error); 2915 } 2916 2917 /* 2918 * Common implementation for chown(), lchown(), and fchown() 2919 */ 2920 int 2921 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid, 2922 gid_t gid) 2923 { 2924 struct mount *mp; 2925 struct vattr vattr; 2926 int error; 2927 2928 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 2929 return (error); 2930 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2931 VATTR_NULL(&vattr); 2932 vattr.va_uid = uid; 2933 vattr.va_gid = gid; 2934 #ifdef MAC 2935 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid, 2936 vattr.va_gid); 2937 if (error == 0) 2938 #endif 2939 error = VOP_SETATTR(vp, &vattr, cred); 2940 VOP_UNLOCK(vp); 2941 vn_finished_write(mp); 2942 return (error); 2943 } 2944 2945 /* 2946 * Set ownership given a path name. 2947 */ 2948 #ifndef _SYS_SYSPROTO_H_ 2949 struct chown_args { 2950 char *path; 2951 int uid; 2952 int gid; 2953 }; 2954 #endif 2955 int 2956 sys_chown(struct thread *td, struct chown_args *uap) 2957 { 2958 2959 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid, 2960 uap->gid, 0)); 2961 } 2962 2963 #ifndef _SYS_SYSPROTO_H_ 2964 struct fchownat_args { 2965 int fd; 2966 const char * path; 2967 uid_t uid; 2968 gid_t gid; 2969 int flag; 2970 }; 2971 #endif 2972 int 2973 sys_fchownat(struct thread *td, struct fchownat_args *uap) 2974 { 2975 2976 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH | 2977 AT_RESOLVE_BENEATH)) != 0) 2978 return (EINVAL); 2979 2980 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid, 2981 uap->gid, uap->flag)); 2982 } 2983 2984 int 2985 kern_fchownat(struct thread *td, int fd, const char *path, 2986 enum uio_seg pathseg, int uid, int gid, int flag) 2987 { 2988 struct nameidata nd; 2989 int error; 2990 2991 AUDIT_ARG_OWNER(uid, gid); 2992 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW | 2993 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd, 2994 &cap_fchown_rights, td); 2995 2996 if ((error = namei(&nd)) != 0) 2997 return (error); 2998 NDFREE(&nd, NDF_ONLY_PNBUF); 2999 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid); 3000 vrele(nd.ni_vp); 3001 return (error); 3002 } 3003 3004 /* 3005 * Set ownership given a path name, do not cross symlinks. 3006 */ 3007 #ifndef _SYS_SYSPROTO_H_ 3008 struct lchown_args { 3009 char *path; 3010 int uid; 3011 int gid; 3012 }; 3013 #endif 3014 int 3015 sys_lchown(struct thread *td, struct lchown_args *uap) 3016 { 3017 3018 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3019 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW)); 3020 } 3021 3022 /* 3023 * Set ownership given a file descriptor. 3024 */ 3025 #ifndef _SYS_SYSPROTO_H_ 3026 struct fchown_args { 3027 int fd; 3028 int uid; 3029 int gid; 3030 }; 3031 #endif 3032 int 3033 sys_fchown(struct thread *td, struct fchown_args *uap) 3034 { 3035 struct file *fp; 3036 int error; 3037 3038 AUDIT_ARG_FD(uap->fd); 3039 AUDIT_ARG_OWNER(uap->uid, uap->gid); 3040 error = fget(td, uap->fd, &cap_fchown_rights, &fp); 3041 if (error != 0) 3042 return (error); 3043 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td); 3044 fdrop(fp, td); 3045 return (error); 3046 } 3047 3048 /* 3049 * Common implementation code for utimes(), lutimes(), and futimes(). 3050 */ 3051 static int 3052 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg, 3053 struct timespec *tsp) 3054 { 3055 struct timeval tv[2]; 3056 const struct timeval *tvp; 3057 int error; 3058 3059 if (usrtvp == NULL) { 3060 vfs_timestamp(&tsp[0]); 3061 tsp[1] = tsp[0]; 3062 } else { 3063 if (tvpseg == UIO_SYSSPACE) { 3064 tvp = usrtvp; 3065 } else { 3066 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0) 3067 return (error); 3068 tvp = tv; 3069 } 3070 3071 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 || 3072 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000) 3073 return (EINVAL); 3074 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]); 3075 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]); 3076 } 3077 return (0); 3078 } 3079 3080 /* 3081 * Common implementation code for futimens(), utimensat(). 3082 */ 3083 #define UTIMENS_NULL 0x1 3084 #define UTIMENS_EXIT 0x2 3085 static int 3086 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg, 3087 struct timespec *tsp, int *retflags) 3088 { 3089 struct timespec tsnow; 3090 int error; 3091 3092 vfs_timestamp(&tsnow); 3093 *retflags = 0; 3094 if (usrtsp == NULL) { 3095 tsp[0] = tsnow; 3096 tsp[1] = tsnow; 3097 *retflags |= UTIMENS_NULL; 3098 return (0); 3099 } 3100 if (tspseg == UIO_SYSSPACE) { 3101 tsp[0] = usrtsp[0]; 3102 tsp[1] = usrtsp[1]; 3103 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0) 3104 return (error); 3105 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT) 3106 *retflags |= UTIMENS_EXIT; 3107 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW) 3108 *retflags |= UTIMENS_NULL; 3109 if (tsp[0].tv_nsec == UTIME_OMIT) 3110 tsp[0].tv_sec = VNOVAL; 3111 else if (tsp[0].tv_nsec == UTIME_NOW) 3112 tsp[0] = tsnow; 3113 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L) 3114 return (EINVAL); 3115 if (tsp[1].tv_nsec == UTIME_OMIT) 3116 tsp[1].tv_sec = VNOVAL; 3117 else if (tsp[1].tv_nsec == UTIME_NOW) 3118 tsp[1] = tsnow; 3119 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L) 3120 return (EINVAL); 3121 3122 return (0); 3123 } 3124 3125 /* 3126 * Common implementation code for utimes(), lutimes(), futimes(), futimens(), 3127 * and utimensat(). 3128 */ 3129 static int 3130 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts, 3131 int numtimes, int nullflag) 3132 { 3133 struct mount *mp; 3134 struct vattr vattr; 3135 int error, setbirthtime; 3136 3137 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) 3138 return (error); 3139 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3140 setbirthtime = 0; 3141 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) && 3142 timespeccmp(&ts[1], &vattr.va_birthtime, < )) 3143 setbirthtime = 1; 3144 VATTR_NULL(&vattr); 3145 vattr.va_atime = ts[0]; 3146 vattr.va_mtime = ts[1]; 3147 if (setbirthtime) 3148 vattr.va_birthtime = ts[1]; 3149 if (numtimes > 2) 3150 vattr.va_birthtime = ts[2]; 3151 if (nullflag) 3152 vattr.va_vaflags |= VA_UTIMES_NULL; 3153 #ifdef MAC 3154 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime, 3155 vattr.va_mtime); 3156 #endif 3157 if (error == 0) 3158 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3159 VOP_UNLOCK(vp); 3160 vn_finished_write(mp); 3161 return (error); 3162 } 3163 3164 /* 3165 * Set the access and modification times of a file. 3166 */ 3167 #ifndef _SYS_SYSPROTO_H_ 3168 struct utimes_args { 3169 char *path; 3170 struct timeval *tptr; 3171 }; 3172 #endif 3173 int 3174 sys_utimes(struct thread *td, struct utimes_args *uap) 3175 { 3176 3177 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3178 uap->tptr, UIO_USERSPACE)); 3179 } 3180 3181 #ifndef _SYS_SYSPROTO_H_ 3182 struct futimesat_args { 3183 int fd; 3184 const char * path; 3185 const struct timeval * times; 3186 }; 3187 #endif 3188 int 3189 sys_futimesat(struct thread *td, struct futimesat_args *uap) 3190 { 3191 3192 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE, 3193 uap->times, UIO_USERSPACE)); 3194 } 3195 3196 int 3197 kern_utimesat(struct thread *td, int fd, const char *path, 3198 enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg) 3199 { 3200 struct nameidata nd; 3201 struct timespec ts[2]; 3202 int error; 3203 3204 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3205 return (error); 3206 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 3207 &cap_futimes_rights, td); 3208 3209 if ((error = namei(&nd)) != 0) 3210 return (error); 3211 NDFREE(&nd, NDF_ONLY_PNBUF); 3212 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3213 vrele(nd.ni_vp); 3214 return (error); 3215 } 3216 3217 /* 3218 * Set the access and modification times of a file. 3219 */ 3220 #ifndef _SYS_SYSPROTO_H_ 3221 struct lutimes_args { 3222 char *path; 3223 struct timeval *tptr; 3224 }; 3225 #endif 3226 int 3227 sys_lutimes(struct thread *td, struct lutimes_args *uap) 3228 { 3229 3230 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr, 3231 UIO_USERSPACE)); 3232 } 3233 3234 int 3235 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg, 3236 struct timeval *tptr, enum uio_seg tptrseg) 3237 { 3238 struct timespec ts[2]; 3239 struct nameidata nd; 3240 int error; 3241 3242 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3243 return (error); 3244 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td); 3245 if ((error = namei(&nd)) != 0) 3246 return (error); 3247 NDFREE(&nd, NDF_ONLY_PNBUF); 3248 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3249 vrele(nd.ni_vp); 3250 return (error); 3251 } 3252 3253 /* 3254 * Set the access and modification times of a file. 3255 */ 3256 #ifndef _SYS_SYSPROTO_H_ 3257 struct futimes_args { 3258 int fd; 3259 struct timeval *tptr; 3260 }; 3261 #endif 3262 int 3263 sys_futimes(struct thread *td, struct futimes_args *uap) 3264 { 3265 3266 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE)); 3267 } 3268 3269 int 3270 kern_futimes(struct thread *td, int fd, struct timeval *tptr, 3271 enum uio_seg tptrseg) 3272 { 3273 struct timespec ts[2]; 3274 struct file *fp; 3275 int error; 3276 3277 AUDIT_ARG_FD(fd); 3278 error = getutimes(tptr, tptrseg, ts); 3279 if (error != 0) 3280 return (error); 3281 error = getvnode(td, fd, &cap_futimes_rights, &fp); 3282 if (error != 0) 3283 return (error); 3284 #ifdef AUDIT 3285 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3286 AUDIT_ARG_VNODE1(fp->f_vnode); 3287 VOP_UNLOCK(fp->f_vnode); 3288 #endif 3289 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL); 3290 fdrop(fp, td); 3291 return (error); 3292 } 3293 3294 int 3295 sys_futimens(struct thread *td, struct futimens_args *uap) 3296 { 3297 3298 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE)); 3299 } 3300 3301 int 3302 kern_futimens(struct thread *td, int fd, struct timespec *tptr, 3303 enum uio_seg tptrseg) 3304 { 3305 struct timespec ts[2]; 3306 struct file *fp; 3307 int error, flags; 3308 3309 AUDIT_ARG_FD(fd); 3310 error = getutimens(tptr, tptrseg, ts, &flags); 3311 if (error != 0) 3312 return (error); 3313 if (flags & UTIMENS_EXIT) 3314 return (0); 3315 error = getvnode(td, fd, &cap_futimes_rights, &fp); 3316 if (error != 0) 3317 return (error); 3318 #ifdef AUDIT 3319 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3320 AUDIT_ARG_VNODE1(fp->f_vnode); 3321 VOP_UNLOCK(fp->f_vnode); 3322 #endif 3323 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL); 3324 fdrop(fp, td); 3325 return (error); 3326 } 3327 3328 int 3329 sys_utimensat(struct thread *td, struct utimensat_args *uap) 3330 { 3331 3332 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE, 3333 uap->times, UIO_USERSPACE, uap->flag)); 3334 } 3335 3336 int 3337 kern_utimensat(struct thread *td, int fd, const char *path, 3338 enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg, 3339 int flag) 3340 { 3341 struct nameidata nd; 3342 struct timespec ts[2]; 3343 int error, flags; 3344 3345 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH | 3346 AT_RESOLVE_BENEATH)) != 0) 3347 return (EINVAL); 3348 3349 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0) 3350 return (error); 3351 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW | 3352 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, 3353 pathseg, path, fd, &cap_futimes_rights, td); 3354 if ((error = namei(&nd)) != 0) 3355 return (error); 3356 /* 3357 * We are allowed to call namei() regardless of 2xUTIME_OMIT. 3358 * POSIX states: 3359 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected." 3360 * "Search permission is denied by a component of the path prefix." 3361 */ 3362 NDFREE(&nd, NDF_ONLY_PNBUF); 3363 if ((flags & UTIMENS_EXIT) == 0) 3364 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL); 3365 vrele(nd.ni_vp); 3366 return (error); 3367 } 3368 3369 /* 3370 * Truncate a file given its path name. 3371 */ 3372 #ifndef _SYS_SYSPROTO_H_ 3373 struct truncate_args { 3374 char *path; 3375 int pad; 3376 off_t length; 3377 }; 3378 #endif 3379 int 3380 sys_truncate(struct thread *td, struct truncate_args *uap) 3381 { 3382 3383 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3384 } 3385 3386 int 3387 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg, 3388 off_t length) 3389 { 3390 struct mount *mp; 3391 struct vnode *vp; 3392 void *rl_cookie; 3393 struct vattr vattr; 3394 struct nameidata nd; 3395 int error; 3396 3397 if (length < 0) 3398 return(EINVAL); 3399 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td); 3400 if ((error = namei(&nd)) != 0) 3401 return (error); 3402 vp = nd.ni_vp; 3403 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX); 3404 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) { 3405 vn_rangelock_unlock(vp, rl_cookie); 3406 vrele(vp); 3407 return (error); 3408 } 3409 NDFREE(&nd, NDF_ONLY_PNBUF); 3410 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3411 if (vp->v_type == VDIR) 3412 error = EISDIR; 3413 #ifdef MAC 3414 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) { 3415 } 3416 #endif 3417 else if ((error = vn_writechk(vp)) == 0 && 3418 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) { 3419 VATTR_NULL(&vattr); 3420 vattr.va_size = length; 3421 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3422 } 3423 VOP_UNLOCK(vp); 3424 vn_finished_write(mp); 3425 vn_rangelock_unlock(vp, rl_cookie); 3426 vrele(vp); 3427 return (error); 3428 } 3429 3430 #if defined(COMPAT_43) 3431 /* 3432 * Truncate a file given its path name. 3433 */ 3434 #ifndef _SYS_SYSPROTO_H_ 3435 struct otruncate_args { 3436 char *path; 3437 long length; 3438 }; 3439 #endif 3440 int 3441 otruncate(struct thread *td, struct otruncate_args *uap) 3442 { 3443 3444 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3445 } 3446 #endif /* COMPAT_43 */ 3447 3448 #if defined(COMPAT_FREEBSD6) 3449 /* Versions with the pad argument */ 3450 int 3451 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap) 3452 { 3453 3454 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3455 } 3456 3457 int 3458 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap) 3459 { 3460 3461 return (kern_ftruncate(td, uap->fd, uap->length)); 3462 } 3463 #endif 3464 3465 int 3466 kern_fsync(struct thread *td, int fd, bool fullsync) 3467 { 3468 struct vnode *vp; 3469 struct mount *mp; 3470 struct file *fp; 3471 int error, lock_flags; 3472 3473 AUDIT_ARG_FD(fd); 3474 error = getvnode(td, fd, &cap_fsync_rights, &fp); 3475 if (error != 0) 3476 return (error); 3477 vp = fp->f_vnode; 3478 #if 0 3479 if (!fullsync) 3480 /* XXXKIB: compete outstanding aio writes */; 3481 #endif 3482 error = vn_start_write(vp, &mp, V_WAIT | PCATCH); 3483 if (error != 0) 3484 goto drop; 3485 if (MNT_SHARED_WRITES(mp) || 3486 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) { 3487 lock_flags = LK_SHARED; 3488 } else { 3489 lock_flags = LK_EXCLUSIVE; 3490 } 3491 vn_lock(vp, lock_flags | LK_RETRY); 3492 AUDIT_ARG_VNODE1(vp); 3493 if (vp->v_object != NULL) { 3494 VM_OBJECT_WLOCK(vp->v_object); 3495 vm_object_page_clean(vp->v_object, 0, 0, 0); 3496 VM_OBJECT_WUNLOCK(vp->v_object); 3497 } 3498 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td); 3499 VOP_UNLOCK(vp); 3500 vn_finished_write(mp); 3501 drop: 3502 fdrop(fp, td); 3503 return (error); 3504 } 3505 3506 /* 3507 * Sync an open file. 3508 */ 3509 #ifndef _SYS_SYSPROTO_H_ 3510 struct fsync_args { 3511 int fd; 3512 }; 3513 #endif 3514 int 3515 sys_fsync(struct thread *td, struct fsync_args *uap) 3516 { 3517 3518 return (kern_fsync(td, uap->fd, true)); 3519 } 3520 3521 int 3522 sys_fdatasync(struct thread *td, struct fdatasync_args *uap) 3523 { 3524 3525 return (kern_fsync(td, uap->fd, false)); 3526 } 3527 3528 /* 3529 * Rename files. Source and destination must either both be directories, or 3530 * both not be directories. If target is a directory, it must be empty. 3531 */ 3532 #ifndef _SYS_SYSPROTO_H_ 3533 struct rename_args { 3534 char *from; 3535 char *to; 3536 }; 3537 #endif 3538 int 3539 sys_rename(struct thread *td, struct rename_args *uap) 3540 { 3541 3542 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD, 3543 uap->to, UIO_USERSPACE)); 3544 } 3545 3546 #ifndef _SYS_SYSPROTO_H_ 3547 struct renameat_args { 3548 int oldfd; 3549 char *old; 3550 int newfd; 3551 char *new; 3552 }; 3553 #endif 3554 int 3555 sys_renameat(struct thread *td, struct renameat_args *uap) 3556 { 3557 3558 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new, 3559 UIO_USERSPACE)); 3560 } 3561 3562 #ifdef MAC 3563 static int 3564 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd, 3565 const char *new, enum uio_seg pathseg, struct nameidata *fromnd) 3566 { 3567 int error; 3568 3569 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | 3570 AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td); 3571 if ((error = namei(fromnd)) != 0) 3572 return (error); 3573 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp, 3574 fromnd->ni_vp, &fromnd->ni_cnd); 3575 VOP_UNLOCK(fromnd->ni_dvp); 3576 if (fromnd->ni_dvp != fromnd->ni_vp) 3577 VOP_UNLOCK(fromnd->ni_vp); 3578 if (error != 0) { 3579 NDFREE(fromnd, NDF_ONLY_PNBUF); 3580 vrele(fromnd->ni_dvp); 3581 vrele(fromnd->ni_vp); 3582 if (fromnd->ni_startdir) 3583 vrele(fromnd->ni_startdir); 3584 } 3585 return (error); 3586 } 3587 #endif 3588 3589 int 3590 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd, 3591 const char *new, enum uio_seg pathseg) 3592 { 3593 struct mount *mp = NULL; 3594 struct vnode *tvp, *fvp, *tdvp; 3595 struct nameidata fromnd, tond; 3596 int error; 3597 3598 again: 3599 bwillwrite(); 3600 #ifdef MAC 3601 if (mac_vnode_check_rename_from_enabled()) { 3602 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg, 3603 &fromnd); 3604 if (error != 0) 3605 return (error); 3606 } else { 3607 #endif 3608 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1, 3609 pathseg, old, oldfd, &cap_renameat_source_rights, td); 3610 if ((error = namei(&fromnd)) != 0) 3611 return (error); 3612 #ifdef MAC 3613 } 3614 #endif 3615 fvp = fromnd.ni_vp; 3616 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | 3617 SAVESTART | AUDITVNODE2, pathseg, new, newfd, 3618 &cap_renameat_target_rights, td); 3619 if (fromnd.ni_vp->v_type == VDIR) 3620 tond.ni_cnd.cn_flags |= WILLBEDIR; 3621 if ((error = namei(&tond)) != 0) { 3622 /* Translate error code for rename("dir1", "dir2/."). */ 3623 if (error == EISDIR && fvp->v_type == VDIR) 3624 error = EINVAL; 3625 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3626 vrele(fromnd.ni_dvp); 3627 vrele(fvp); 3628 goto out1; 3629 } 3630 tdvp = tond.ni_dvp; 3631 tvp = tond.ni_vp; 3632 error = vn_start_write(fvp, &mp, V_NOWAIT); 3633 if (error != 0) { 3634 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3635 NDFREE(&tond, NDF_ONLY_PNBUF); 3636 if (tvp != NULL) 3637 vput(tvp); 3638 if (tdvp == tvp) 3639 vrele(tdvp); 3640 else 3641 vput(tdvp); 3642 vrele(fromnd.ni_dvp); 3643 vrele(fvp); 3644 vrele(tond.ni_startdir); 3645 if (fromnd.ni_startdir != NULL) 3646 vrele(fromnd.ni_startdir); 3647 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH); 3648 if (error != 0) 3649 return (error); 3650 goto again; 3651 } 3652 if (tvp != NULL) { 3653 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3654 error = ENOTDIR; 3655 goto out; 3656 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3657 error = EISDIR; 3658 goto out; 3659 } 3660 #ifdef CAPABILITIES 3661 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) { 3662 /* 3663 * If the target already exists we require CAP_UNLINKAT 3664 * from 'newfd', when newfd was used for the lookup. 3665 */ 3666 error = cap_check(&tond.ni_filecaps.fc_rights, 3667 &cap_unlinkat_rights); 3668 if (error != 0) 3669 goto out; 3670 } 3671 #endif 3672 } 3673 if (fvp == tdvp) { 3674 error = EINVAL; 3675 goto out; 3676 } 3677 /* 3678 * If the source is the same as the destination (that is, if they 3679 * are links to the same vnode), then there is nothing to do. 3680 */ 3681 if (fvp == tvp) 3682 error = -1; 3683 #ifdef MAC 3684 else 3685 error = mac_vnode_check_rename_to(td->td_ucred, tdvp, 3686 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd); 3687 #endif 3688 out: 3689 if (error == 0) { 3690 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3691 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3692 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3693 NDFREE(&tond, NDF_ONLY_PNBUF); 3694 } else { 3695 NDFREE(&fromnd, NDF_ONLY_PNBUF); 3696 NDFREE(&tond, NDF_ONLY_PNBUF); 3697 if (tvp != NULL) 3698 vput(tvp); 3699 if (tdvp == tvp) 3700 vrele(tdvp); 3701 else 3702 vput(tdvp); 3703 vrele(fromnd.ni_dvp); 3704 vrele(fvp); 3705 } 3706 vrele(tond.ni_startdir); 3707 vn_finished_write(mp); 3708 out1: 3709 if (fromnd.ni_startdir) 3710 vrele(fromnd.ni_startdir); 3711 if (error == -1) 3712 return (0); 3713 return (error); 3714 } 3715 3716 /* 3717 * Make a directory file. 3718 */ 3719 #ifndef _SYS_SYSPROTO_H_ 3720 struct mkdir_args { 3721 char *path; 3722 int mode; 3723 }; 3724 #endif 3725 int 3726 sys_mkdir(struct thread *td, struct mkdir_args *uap) 3727 { 3728 3729 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3730 uap->mode)); 3731 } 3732 3733 #ifndef _SYS_SYSPROTO_H_ 3734 struct mkdirat_args { 3735 int fd; 3736 char *path; 3737 mode_t mode; 3738 }; 3739 #endif 3740 int 3741 sys_mkdirat(struct thread *td, struct mkdirat_args *uap) 3742 { 3743 3744 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode)); 3745 } 3746 3747 int 3748 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg, 3749 int mode) 3750 { 3751 struct mount *mp; 3752 struct vnode *vp; 3753 struct vattr vattr; 3754 struct nameidata nd; 3755 int error; 3756 3757 AUDIT_ARG_MODE(mode); 3758 restart: 3759 bwillwrite(); 3760 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 | 3761 NOCACHE, segflg, path, fd, &cap_mkdirat_rights, 3762 td); 3763 nd.ni_cnd.cn_flags |= WILLBEDIR; 3764 if ((error = namei(&nd)) != 0) 3765 return (error); 3766 vp = nd.ni_vp; 3767 if (vp != NULL) { 3768 NDFREE(&nd, NDF_ONLY_PNBUF); 3769 /* 3770 * XXX namei called with LOCKPARENT but not LOCKLEAF has 3771 * the strange behaviour of leaving the vnode unlocked 3772 * if the target is the same vnode as the parent. 3773 */ 3774 if (vp == nd.ni_dvp) 3775 vrele(nd.ni_dvp); 3776 else 3777 vput(nd.ni_dvp); 3778 vrele(vp); 3779 return (EEXIST); 3780 } 3781 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3782 NDFREE(&nd, NDF_ONLY_PNBUF); 3783 vput(nd.ni_dvp); 3784 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3785 return (error); 3786 goto restart; 3787 } 3788 VATTR_NULL(&vattr); 3789 vattr.va_type = VDIR; 3790 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask; 3791 #ifdef MAC 3792 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 3793 &vattr); 3794 if (error != 0) 3795 goto out; 3796 #endif 3797 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3798 #ifdef MAC 3799 out: 3800 #endif 3801 NDFREE(&nd, NDF_ONLY_PNBUF); 3802 vput(nd.ni_dvp); 3803 if (error == 0) 3804 vput(nd.ni_vp); 3805 vn_finished_write(mp); 3806 return (error); 3807 } 3808 3809 /* 3810 * Remove a directory file. 3811 */ 3812 #ifndef _SYS_SYSPROTO_H_ 3813 struct rmdir_args { 3814 char *path; 3815 }; 3816 #endif 3817 int 3818 sys_rmdir(struct thread *td, struct rmdir_args *uap) 3819 { 3820 3821 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE, 3822 0)); 3823 } 3824 3825 int 3826 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd, 3827 enum uio_seg pathseg, int flag) 3828 { 3829 struct mount *mp; 3830 struct vnode *vp; 3831 struct file *fp; 3832 struct nameidata nd; 3833 cap_rights_t rights; 3834 int error; 3835 3836 fp = NULL; 3837 if (fd != FD_NONE) { 3838 error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP), 3839 &fp); 3840 if (error != 0) 3841 return (error); 3842 } 3843 3844 restart: 3845 bwillwrite(); 3846 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 | 3847 at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH), 3848 pathseg, path, dfd, &cap_unlinkat_rights, td); 3849 if ((error = namei(&nd)) != 0) 3850 goto fdout; 3851 vp = nd.ni_vp; 3852 if (vp->v_type != VDIR) { 3853 error = ENOTDIR; 3854 goto out; 3855 } 3856 /* 3857 * No rmdir "." please. 3858 */ 3859 if (nd.ni_dvp == vp) { 3860 error = EINVAL; 3861 goto out; 3862 } 3863 /* 3864 * The root of a mounted filesystem cannot be deleted. 3865 */ 3866 if (vp->v_vflag & VV_ROOT) { 3867 error = EBUSY; 3868 goto out; 3869 } 3870 3871 if (fp != NULL && fp->f_vnode != vp) { 3872 if (VN_IS_DOOMED(fp->f_vnode)) 3873 error = EBADF; 3874 else 3875 error = EDEADLK; 3876 goto out; 3877 } 3878 3879 #ifdef MAC 3880 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 3881 &nd.ni_cnd); 3882 if (error != 0) 3883 goto out; 3884 #endif 3885 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3886 NDFREE(&nd, NDF_ONLY_PNBUF); 3887 vput(vp); 3888 if (nd.ni_dvp == vp) 3889 vrele(nd.ni_dvp); 3890 else 3891 vput(nd.ni_dvp); 3892 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) 3893 goto fdout; 3894 goto restart; 3895 } 3896 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 3897 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3898 vn_finished_write(mp); 3899 out: 3900 NDFREE(&nd, NDF_ONLY_PNBUF); 3901 vput(vp); 3902 if (nd.ni_dvp == vp) 3903 vrele(nd.ni_dvp); 3904 else 3905 vput(nd.ni_dvp); 3906 fdout: 3907 if (fp != NULL) 3908 fdrop(fp, td); 3909 return (error); 3910 } 3911 3912 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 3913 int 3914 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count, 3915 long *basep, void (*func)(struct freebsd11_dirent *)) 3916 { 3917 struct freebsd11_dirent dstdp; 3918 struct dirent *dp, *edp; 3919 char *dirbuf; 3920 off_t base; 3921 ssize_t resid, ucount; 3922 int error; 3923 3924 /* XXX arbitrary sanity limit on `count'. */ 3925 count = min(count, 64 * 1024); 3926 3927 dirbuf = malloc(count, M_TEMP, M_WAITOK); 3928 3929 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid, 3930 UIO_SYSSPACE); 3931 if (error != 0) 3932 goto done; 3933 if (basep != NULL) 3934 *basep = base; 3935 3936 ucount = 0; 3937 for (dp = (struct dirent *)dirbuf, 3938 edp = (struct dirent *)&dirbuf[count - resid]; 3939 ucount < count && dp < edp; ) { 3940 if (dp->d_reclen == 0) 3941 break; 3942 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0)); 3943 if (dp->d_namlen >= sizeof(dstdp.d_name)) 3944 continue; 3945 dstdp.d_type = dp->d_type; 3946 dstdp.d_namlen = dp->d_namlen; 3947 dstdp.d_fileno = dp->d_fileno; /* truncate */ 3948 if (dstdp.d_fileno != dp->d_fileno) { 3949 switch (ino64_trunc_error) { 3950 default: 3951 case 0: 3952 break; 3953 case 1: 3954 error = EOVERFLOW; 3955 goto done; 3956 case 2: 3957 dstdp.d_fileno = UINT32_MAX; 3958 break; 3959 } 3960 } 3961 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) + 3962 ((dp->d_namlen + 1 + 3) &~ 3); 3963 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen); 3964 bzero(dstdp.d_name + dstdp.d_namlen, 3965 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) - 3966 dstdp.d_namlen); 3967 MPASS(dstdp.d_reclen <= dp->d_reclen); 3968 MPASS(ucount + dstdp.d_reclen <= count); 3969 if (func != NULL) 3970 func(&dstdp); 3971 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen); 3972 if (error != 0) 3973 break; 3974 dp = (struct dirent *)((char *)dp + dp->d_reclen); 3975 ucount += dstdp.d_reclen; 3976 } 3977 3978 done: 3979 free(dirbuf, M_TEMP); 3980 if (error == 0) 3981 td->td_retval[0] = ucount; 3982 return (error); 3983 } 3984 #endif /* COMPAT */ 3985 3986 #ifdef COMPAT_43 3987 static void 3988 ogetdirentries_cvt(struct freebsd11_dirent *dp) 3989 { 3990 #if (BYTE_ORDER == LITTLE_ENDIAN) 3991 /* 3992 * The expected low byte of dp->d_namlen is our dp->d_type. 3993 * The high MBZ byte of dp->d_namlen is our dp->d_namlen. 3994 */ 3995 dp->d_type = dp->d_namlen; 3996 dp->d_namlen = 0; 3997 #else 3998 /* 3999 * The dp->d_type is the high byte of the expected dp->d_namlen, 4000 * so must be zero'ed. 4001 */ 4002 dp->d_type = 0; 4003 #endif 4004 } 4005 4006 /* 4007 * Read a block of directory entries in a filesystem independent format. 4008 */ 4009 #ifndef _SYS_SYSPROTO_H_ 4010 struct ogetdirentries_args { 4011 int fd; 4012 char *buf; 4013 u_int count; 4014 long *basep; 4015 }; 4016 #endif 4017 int 4018 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap) 4019 { 4020 long loff; 4021 int error; 4022 4023 error = kern_ogetdirentries(td, uap, &loff); 4024 if (error == 0) 4025 error = copyout(&loff, uap->basep, sizeof(long)); 4026 return (error); 4027 } 4028 4029 int 4030 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap, 4031 long *ploff) 4032 { 4033 long base; 4034 int error; 4035 4036 /* XXX arbitrary sanity limit on `count'. */ 4037 if (uap->count > 64 * 1024) 4038 return (EINVAL); 4039 4040 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 4041 &base, ogetdirentries_cvt); 4042 4043 if (error == 0 && uap->basep != NULL) 4044 error = copyout(&base, uap->basep, sizeof(long)); 4045 4046 return (error); 4047 } 4048 #endif /* COMPAT_43 */ 4049 4050 #if defined(COMPAT_FREEBSD11) 4051 #ifndef _SYS_SYSPROTO_H_ 4052 struct freebsd11_getdirentries_args { 4053 int fd; 4054 char *buf; 4055 u_int count; 4056 long *basep; 4057 }; 4058 #endif 4059 int 4060 freebsd11_getdirentries(struct thread *td, 4061 struct freebsd11_getdirentries_args *uap) 4062 { 4063 long base; 4064 int error; 4065 4066 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 4067 &base, NULL); 4068 4069 if (error == 0 && uap->basep != NULL) 4070 error = copyout(&base, uap->basep, sizeof(long)); 4071 return (error); 4072 } 4073 4074 int 4075 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap) 4076 { 4077 struct freebsd11_getdirentries_args ap; 4078 4079 ap.fd = uap->fd; 4080 ap.buf = uap->buf; 4081 ap.count = uap->count; 4082 ap.basep = NULL; 4083 return (freebsd11_getdirentries(td, &ap)); 4084 } 4085 #endif /* COMPAT_FREEBSD11 */ 4086 4087 /* 4088 * Read a block of directory entries in a filesystem independent format. 4089 */ 4090 int 4091 sys_getdirentries(struct thread *td, struct getdirentries_args *uap) 4092 { 4093 off_t base; 4094 int error; 4095 4096 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base, 4097 NULL, UIO_USERSPACE); 4098 if (error != 0) 4099 return (error); 4100 if (uap->basep != NULL) 4101 error = copyout(&base, uap->basep, sizeof(off_t)); 4102 return (error); 4103 } 4104 4105 int 4106 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count, 4107 off_t *basep, ssize_t *residp, enum uio_seg bufseg) 4108 { 4109 struct vnode *vp; 4110 struct file *fp; 4111 struct uio auio; 4112 struct iovec aiov; 4113 off_t loff; 4114 int error, eofflag; 4115 off_t foffset; 4116 4117 AUDIT_ARG_FD(fd); 4118 if (count > IOSIZE_MAX) 4119 return (EINVAL); 4120 auio.uio_resid = count; 4121 error = getvnode(td, fd, &cap_read_rights, &fp); 4122 if (error != 0) 4123 return (error); 4124 if ((fp->f_flag & FREAD) == 0) { 4125 fdrop(fp, td); 4126 return (EBADF); 4127 } 4128 vp = fp->f_vnode; 4129 foffset = foffset_lock(fp, 0); 4130 unionread: 4131 if (vp->v_type != VDIR) { 4132 error = EINVAL; 4133 goto fail; 4134 } 4135 aiov.iov_base = buf; 4136 aiov.iov_len = count; 4137 auio.uio_iov = &aiov; 4138 auio.uio_iovcnt = 1; 4139 auio.uio_rw = UIO_READ; 4140 auio.uio_segflg = bufseg; 4141 auio.uio_td = td; 4142 vn_lock(vp, LK_SHARED | LK_RETRY); 4143 AUDIT_ARG_VNODE1(vp); 4144 loff = auio.uio_offset = foffset; 4145 #ifdef MAC 4146 error = mac_vnode_check_readdir(td->td_ucred, vp); 4147 if (error == 0) 4148 #endif 4149 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, 4150 NULL); 4151 foffset = auio.uio_offset; 4152 if (error != 0) { 4153 VOP_UNLOCK(vp); 4154 goto fail; 4155 } 4156 if (count == auio.uio_resid && 4157 (vp->v_vflag & VV_ROOT) && 4158 (vp->v_mount->mnt_flag & MNT_UNION)) { 4159 struct vnode *tvp = vp; 4160 4161 vp = vp->v_mount->mnt_vnodecovered; 4162 VREF(vp); 4163 fp->f_vnode = vp; 4164 foffset = 0; 4165 vput(tvp); 4166 goto unionread; 4167 } 4168 VOP_UNLOCK(vp); 4169 *basep = loff; 4170 if (residp != NULL) 4171 *residp = auio.uio_resid; 4172 td->td_retval[0] = count - auio.uio_resid; 4173 fail: 4174 foffset_unlock(fp, foffset, 0); 4175 fdrop(fp, td); 4176 return (error); 4177 } 4178 4179 /* 4180 * Set the mode mask for creation of filesystem nodes. 4181 */ 4182 #ifndef _SYS_SYSPROTO_H_ 4183 struct umask_args { 4184 int newmask; 4185 }; 4186 #endif 4187 int 4188 sys_umask(struct thread *td, struct umask_args *uap) 4189 { 4190 struct filedesc *fdp; 4191 4192 fdp = td->td_proc->p_fd; 4193 FILEDESC_XLOCK(fdp); 4194 td->td_retval[0] = fdp->fd_cmask; 4195 fdp->fd_cmask = uap->newmask & ALLPERMS; 4196 FILEDESC_XUNLOCK(fdp); 4197 return (0); 4198 } 4199 4200 /* 4201 * Void all references to file by ripping underlying filesystem away from 4202 * vnode. 4203 */ 4204 #ifndef _SYS_SYSPROTO_H_ 4205 struct revoke_args { 4206 char *path; 4207 }; 4208 #endif 4209 int 4210 sys_revoke(struct thread *td, struct revoke_args *uap) 4211 { 4212 struct vnode *vp; 4213 struct vattr vattr; 4214 struct nameidata nd; 4215 int error; 4216 4217 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4218 uap->path, td); 4219 if ((error = namei(&nd)) != 0) 4220 return (error); 4221 vp = nd.ni_vp; 4222 NDFREE(&nd, NDF_ONLY_PNBUF); 4223 if (vp->v_type != VCHR || vp->v_rdev == NULL) { 4224 error = EINVAL; 4225 goto out; 4226 } 4227 #ifdef MAC 4228 error = mac_vnode_check_revoke(td->td_ucred, vp); 4229 if (error != 0) 4230 goto out; 4231 #endif 4232 error = VOP_GETATTR(vp, &vattr, td->td_ucred); 4233 if (error != 0) 4234 goto out; 4235 if (td->td_ucred->cr_uid != vattr.va_uid) { 4236 error = priv_check(td, PRIV_VFS_ADMIN); 4237 if (error != 0) 4238 goto out; 4239 } 4240 if (devfs_usecount(vp) > 0) 4241 VOP_REVOKE(vp, REVOKEALL); 4242 out: 4243 vput(vp); 4244 return (error); 4245 } 4246 4247 /* 4248 * Convert a user file descriptor to a kernel file entry and check that, if it 4249 * is a capability, the correct rights are present. A reference on the file 4250 * entry is held upon returning. 4251 */ 4252 int 4253 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp) 4254 { 4255 struct file *fp; 4256 int error; 4257 4258 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp); 4259 if (error != 0) 4260 return (error); 4261 4262 /* 4263 * The file could be not of the vnode type, or it may be not 4264 * yet fully initialized, in which case the f_vnode pointer 4265 * may be set, but f_ops is still badfileops. E.g., 4266 * devfs_open() transiently create such situation to 4267 * facilitate csw d_fdopen(). 4268 * 4269 * Dupfdopen() handling in kern_openat() installs the 4270 * half-baked file into the process descriptor table, allowing 4271 * other thread to dereference it. Guard against the race by 4272 * checking f_ops. 4273 */ 4274 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) { 4275 fdrop(fp, td); 4276 return (EINVAL); 4277 } 4278 *fpp = fp; 4279 return (0); 4280 } 4281 4282 /* 4283 * Get an (NFS) file handle. 4284 */ 4285 #ifndef _SYS_SYSPROTO_H_ 4286 struct lgetfh_args { 4287 char *fname; 4288 fhandle_t *fhp; 4289 }; 4290 #endif 4291 int 4292 sys_lgetfh(struct thread *td, struct lgetfh_args *uap) 4293 { 4294 4295 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname, 4296 UIO_USERSPACE, uap->fhp)); 4297 } 4298 4299 #ifndef _SYS_SYSPROTO_H_ 4300 struct getfh_args { 4301 char *fname; 4302 fhandle_t *fhp; 4303 }; 4304 #endif 4305 int 4306 sys_getfh(struct thread *td, struct getfh_args *uap) 4307 { 4308 4309 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE, 4310 uap->fhp)); 4311 } 4312 4313 /* 4314 * syscall for the rpc.lockd to use to translate an open descriptor into 4315 * a NFS file handle. 4316 * 4317 * warning: do not remove the priv_check() call or this becomes one giant 4318 * security hole. 4319 */ 4320 #ifndef _SYS_SYSPROTO_H_ 4321 struct getfhat_args { 4322 int fd; 4323 char *path; 4324 fhandle_t *fhp; 4325 int flags; 4326 }; 4327 #endif 4328 int 4329 sys_getfhat(struct thread *td, struct getfhat_args *uap) 4330 { 4331 4332 if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH | 4333 AT_RESOLVE_BENEATH)) != 0) 4334 return (EINVAL); 4335 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE, 4336 uap->fhp)); 4337 } 4338 4339 static int 4340 kern_getfhat(struct thread *td, int flags, int fd, const char *path, 4341 enum uio_seg pathseg, fhandle_t *fhp) 4342 { 4343 struct nameidata nd; 4344 fhandle_t fh; 4345 struct vnode *vp; 4346 int error; 4347 4348 error = priv_check(td, PRIV_VFS_GETFH); 4349 if (error != 0) 4350 return (error); 4351 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW | 4352 AT_BENEATH | AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, 4353 pathseg, path, fd, td); 4354 error = namei(&nd); 4355 if (error != 0) 4356 return (error); 4357 NDFREE(&nd, NDF_ONLY_PNBUF); 4358 vp = nd.ni_vp; 4359 bzero(&fh, sizeof(fh)); 4360 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4361 error = VOP_VPTOFH(vp, &fh.fh_fid); 4362 vput(vp); 4363 if (error == 0) 4364 error = copyout(&fh, fhp, sizeof (fh)); 4365 return (error); 4366 } 4367 4368 #ifndef _SYS_SYSPROTO_H_ 4369 struct fhlink_args { 4370 fhandle_t *fhp; 4371 const char *to; 4372 }; 4373 #endif 4374 int 4375 sys_fhlink(struct thread *td, struct fhlink_args *uap) 4376 { 4377 4378 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp)); 4379 } 4380 4381 #ifndef _SYS_SYSPROTO_H_ 4382 struct fhlinkat_args { 4383 fhandle_t *fhp; 4384 int tofd; 4385 const char *to; 4386 }; 4387 #endif 4388 int 4389 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap) 4390 { 4391 4392 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp)); 4393 } 4394 4395 static int 4396 kern_fhlinkat(struct thread *td, int fd, const char *path, 4397 enum uio_seg pathseg, fhandle_t *fhp) 4398 { 4399 fhandle_t fh; 4400 struct mount *mp; 4401 struct vnode *vp; 4402 int error; 4403 4404 error = priv_check(td, PRIV_VFS_GETFH); 4405 if (error != 0) 4406 return (error); 4407 error = copyin(fhp, &fh, sizeof(fh)); 4408 if (error != 0) 4409 return (error); 4410 do { 4411 bwillwrite(); 4412 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4413 return (ESTALE); 4414 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp); 4415 vfs_unbusy(mp); 4416 if (error != 0) 4417 return (error); 4418 VOP_UNLOCK(vp); 4419 } while ((error = kern_linkat_vp(td, vp, fd, path, pathseg)) == EAGAIN); 4420 return (error); 4421 } 4422 4423 #ifndef _SYS_SYSPROTO_H_ 4424 struct fhreadlink_args { 4425 fhandle_t *fhp; 4426 char *buf; 4427 size_t bufsize; 4428 }; 4429 #endif 4430 int 4431 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap) 4432 { 4433 fhandle_t fh; 4434 struct mount *mp; 4435 struct vnode *vp; 4436 int error; 4437 4438 error = priv_check(td, PRIV_VFS_GETFH); 4439 if (error != 0) 4440 return (error); 4441 if (uap->bufsize > IOSIZE_MAX) 4442 return (EINVAL); 4443 error = copyin(uap->fhp, &fh, sizeof(fh)); 4444 if (error != 0) 4445 return (error); 4446 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4447 return (ESTALE); 4448 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp); 4449 vfs_unbusy(mp); 4450 if (error != 0) 4451 return (error); 4452 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td); 4453 vput(vp); 4454 return (error); 4455 } 4456 4457 /* 4458 * syscall for the rpc.lockd to use to translate a NFS file handle into an 4459 * open descriptor. 4460 * 4461 * warning: do not remove the priv_check() call or this becomes one giant 4462 * security hole. 4463 */ 4464 #ifndef _SYS_SYSPROTO_H_ 4465 struct fhopen_args { 4466 const struct fhandle *u_fhp; 4467 int flags; 4468 }; 4469 #endif 4470 int 4471 sys_fhopen(struct thread *td, struct fhopen_args *uap) 4472 { 4473 struct mount *mp; 4474 struct vnode *vp; 4475 struct fhandle fhp; 4476 struct file *fp; 4477 int fmode, error; 4478 int indx; 4479 4480 error = priv_check(td, PRIV_VFS_FHOPEN); 4481 if (error != 0) 4482 return (error); 4483 indx = -1; 4484 fmode = FFLAGS(uap->flags); 4485 /* why not allow a non-read/write open for our lockd? */ 4486 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT)) 4487 return (EINVAL); 4488 error = copyin(uap->u_fhp, &fhp, sizeof(fhp)); 4489 if (error != 0) 4490 return(error); 4491 /* find the mount point */ 4492 mp = vfs_busyfs(&fhp.fh_fsid); 4493 if (mp == NULL) 4494 return (ESTALE); 4495 /* now give me my vnode, it gets returned to me locked */ 4496 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp); 4497 vfs_unbusy(mp); 4498 if (error != 0) 4499 return (error); 4500 4501 error = falloc_noinstall(td, &fp); 4502 if (error != 0) { 4503 vput(vp); 4504 return (error); 4505 } 4506 /* 4507 * An extra reference on `fp' has been held for us by 4508 * falloc_noinstall(). 4509 */ 4510 4511 #ifdef INVARIANTS 4512 td->td_dupfd = -1; 4513 #endif 4514 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp); 4515 if (error != 0) { 4516 KASSERT(fp->f_ops == &badfileops, 4517 ("VOP_OPEN in fhopen() set f_ops")); 4518 KASSERT(td->td_dupfd < 0, 4519 ("fhopen() encountered fdopen()")); 4520 4521 vput(vp); 4522 goto bad; 4523 } 4524 #ifdef INVARIANTS 4525 td->td_dupfd = 0; 4526 #endif 4527 fp->f_vnode = vp; 4528 finit_vnode(fp, fmode, NULL, &vnops); 4529 VOP_UNLOCK(vp); 4530 if ((fmode & O_TRUNC) != 0) { 4531 error = fo_truncate(fp, 0, td->td_ucred, td); 4532 if (error != 0) 4533 goto bad; 4534 } 4535 4536 error = finstall(td, fp, &indx, fmode, NULL); 4537 bad: 4538 fdrop(fp, td); 4539 td->td_retval[0] = indx; 4540 return (error); 4541 } 4542 4543 /* 4544 * Stat an (NFS) file handle. 4545 */ 4546 #ifndef _SYS_SYSPROTO_H_ 4547 struct fhstat_args { 4548 struct fhandle *u_fhp; 4549 struct stat *sb; 4550 }; 4551 #endif 4552 int 4553 sys_fhstat(struct thread *td, struct fhstat_args *uap) 4554 { 4555 struct stat sb; 4556 struct fhandle fh; 4557 int error; 4558 4559 error = copyin(uap->u_fhp, &fh, sizeof(fh)); 4560 if (error != 0) 4561 return (error); 4562 error = kern_fhstat(td, fh, &sb); 4563 if (error == 0) 4564 error = copyout(&sb, uap->sb, sizeof(sb)); 4565 return (error); 4566 } 4567 4568 int 4569 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb) 4570 { 4571 struct mount *mp; 4572 struct vnode *vp; 4573 int error; 4574 4575 error = priv_check(td, PRIV_VFS_FHSTAT); 4576 if (error != 0) 4577 return (error); 4578 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4579 return (ESTALE); 4580 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4581 vfs_unbusy(mp); 4582 if (error != 0) 4583 return (error); 4584 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td); 4585 vput(vp); 4586 return (error); 4587 } 4588 4589 /* 4590 * Implement fstatfs() for (NFS) file handles. 4591 */ 4592 #ifndef _SYS_SYSPROTO_H_ 4593 struct fhstatfs_args { 4594 struct fhandle *u_fhp; 4595 struct statfs *buf; 4596 }; 4597 #endif 4598 int 4599 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap) 4600 { 4601 struct statfs *sfp; 4602 fhandle_t fh; 4603 int error; 4604 4605 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 4606 if (error != 0) 4607 return (error); 4608 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 4609 error = kern_fhstatfs(td, fh, sfp); 4610 if (error == 0) 4611 error = copyout(sfp, uap->buf, sizeof(*sfp)); 4612 free(sfp, M_STATFS); 4613 return (error); 4614 } 4615 4616 int 4617 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf) 4618 { 4619 struct mount *mp; 4620 struct vnode *vp; 4621 int error; 4622 4623 error = priv_check(td, PRIV_VFS_FHSTATFS); 4624 if (error != 0) 4625 return (error); 4626 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4627 return (ESTALE); 4628 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4629 if (error != 0) { 4630 vfs_unbusy(mp); 4631 return (error); 4632 } 4633 vput(vp); 4634 error = prison_canseemount(td->td_ucred, mp); 4635 if (error != 0) 4636 goto out; 4637 #ifdef MAC 4638 error = mac_mount_check_stat(td->td_ucred, mp); 4639 if (error != 0) 4640 goto out; 4641 #endif 4642 error = VFS_STATFS(mp, buf); 4643 out: 4644 vfs_unbusy(mp); 4645 return (error); 4646 } 4647 4648 /* 4649 * Unlike madvise(2), we do not make a best effort to remember every 4650 * possible caching hint. Instead, we remember the last setting with 4651 * the exception that we will allow POSIX_FADV_NORMAL to adjust the 4652 * region of any current setting. 4653 */ 4654 int 4655 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len, 4656 int advice) 4657 { 4658 struct fadvise_info *fa, *new; 4659 struct file *fp; 4660 struct vnode *vp; 4661 off_t end; 4662 int error; 4663 4664 if (offset < 0 || len < 0 || offset > OFF_MAX - len) 4665 return (EINVAL); 4666 AUDIT_ARG_VALUE(advice); 4667 switch (advice) { 4668 case POSIX_FADV_SEQUENTIAL: 4669 case POSIX_FADV_RANDOM: 4670 case POSIX_FADV_NOREUSE: 4671 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK); 4672 break; 4673 case POSIX_FADV_NORMAL: 4674 case POSIX_FADV_WILLNEED: 4675 case POSIX_FADV_DONTNEED: 4676 new = NULL; 4677 break; 4678 default: 4679 return (EINVAL); 4680 } 4681 /* XXX: CAP_POSIX_FADVISE? */ 4682 AUDIT_ARG_FD(fd); 4683 error = fget(td, fd, &cap_no_rights, &fp); 4684 if (error != 0) 4685 goto out; 4686 AUDIT_ARG_FILE(td->td_proc, fp); 4687 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) { 4688 error = ESPIPE; 4689 goto out; 4690 } 4691 if (fp->f_type != DTYPE_VNODE) { 4692 error = ENODEV; 4693 goto out; 4694 } 4695 vp = fp->f_vnode; 4696 if (vp->v_type != VREG) { 4697 error = ENODEV; 4698 goto out; 4699 } 4700 if (len == 0) 4701 end = OFF_MAX; 4702 else 4703 end = offset + len - 1; 4704 switch (advice) { 4705 case POSIX_FADV_SEQUENTIAL: 4706 case POSIX_FADV_RANDOM: 4707 case POSIX_FADV_NOREUSE: 4708 /* 4709 * Try to merge any existing non-standard region with 4710 * this new region if possible, otherwise create a new 4711 * non-standard region for this request. 4712 */ 4713 mtx_pool_lock(mtxpool_sleep, fp); 4714 fa = fp->f_advice; 4715 if (fa != NULL && fa->fa_advice == advice && 4716 ((fa->fa_start <= end && fa->fa_end >= offset) || 4717 (end != OFF_MAX && fa->fa_start == end + 1) || 4718 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) { 4719 if (offset < fa->fa_start) 4720 fa->fa_start = offset; 4721 if (end > fa->fa_end) 4722 fa->fa_end = end; 4723 } else { 4724 new->fa_advice = advice; 4725 new->fa_start = offset; 4726 new->fa_end = end; 4727 fp->f_advice = new; 4728 new = fa; 4729 } 4730 mtx_pool_unlock(mtxpool_sleep, fp); 4731 break; 4732 case POSIX_FADV_NORMAL: 4733 /* 4734 * If a the "normal" region overlaps with an existing 4735 * non-standard region, trim or remove the 4736 * non-standard region. 4737 */ 4738 mtx_pool_lock(mtxpool_sleep, fp); 4739 fa = fp->f_advice; 4740 if (fa != NULL) { 4741 if (offset <= fa->fa_start && end >= fa->fa_end) { 4742 new = fa; 4743 fp->f_advice = NULL; 4744 } else if (offset <= fa->fa_start && 4745 end >= fa->fa_start) 4746 fa->fa_start = end + 1; 4747 else if (offset <= fa->fa_end && end >= fa->fa_end) 4748 fa->fa_end = offset - 1; 4749 else if (offset >= fa->fa_start && end <= fa->fa_end) { 4750 /* 4751 * If the "normal" region is a middle 4752 * portion of the existing 4753 * non-standard region, just remove 4754 * the whole thing rather than picking 4755 * one side or the other to 4756 * preserve. 4757 */ 4758 new = fa; 4759 fp->f_advice = NULL; 4760 } 4761 } 4762 mtx_pool_unlock(mtxpool_sleep, fp); 4763 break; 4764 case POSIX_FADV_WILLNEED: 4765 case POSIX_FADV_DONTNEED: 4766 error = VOP_ADVISE(vp, offset, end, advice); 4767 break; 4768 } 4769 out: 4770 if (fp != NULL) 4771 fdrop(fp, td); 4772 free(new, M_FADVISE); 4773 return (error); 4774 } 4775 4776 int 4777 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap) 4778 { 4779 int error; 4780 4781 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len, 4782 uap->advice); 4783 return (kern_posix_error(td, error)); 4784 } 4785 4786 int 4787 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd, 4788 off_t *outoffp, size_t len, unsigned int flags) 4789 { 4790 struct file *infp, *outfp; 4791 struct vnode *invp, *outvp; 4792 int error; 4793 size_t retlen; 4794 void *rl_rcookie, *rl_wcookie; 4795 off_t savinoff, savoutoff; 4796 4797 infp = outfp = NULL; 4798 rl_rcookie = rl_wcookie = NULL; 4799 savinoff = -1; 4800 error = 0; 4801 retlen = 0; 4802 4803 if (flags != 0) { 4804 error = EINVAL; 4805 goto out; 4806 } 4807 if (len > SSIZE_MAX) 4808 /* 4809 * Although the len argument is size_t, the return argument 4810 * is ssize_t (which is signed). Therefore a size that won't 4811 * fit in ssize_t can't be returned. 4812 */ 4813 len = SSIZE_MAX; 4814 4815 /* Get the file structures for the file descriptors. */ 4816 error = fget_read(td, infd, &cap_read_rights, &infp); 4817 if (error != 0) 4818 goto out; 4819 if (infp->f_ops == &badfileops) { 4820 error = EBADF; 4821 goto out; 4822 } 4823 if (infp->f_vnode == NULL) { 4824 error = EINVAL; 4825 goto out; 4826 } 4827 error = fget_write(td, outfd, &cap_write_rights, &outfp); 4828 if (error != 0) 4829 goto out; 4830 if (outfp->f_ops == &badfileops) { 4831 error = EBADF; 4832 goto out; 4833 } 4834 if (outfp->f_vnode == NULL) { 4835 error = EINVAL; 4836 goto out; 4837 } 4838 4839 /* Set the offset pointers to the correct place. */ 4840 if (inoffp == NULL) 4841 inoffp = &infp->f_offset; 4842 if (outoffp == NULL) 4843 outoffp = &outfp->f_offset; 4844 savinoff = *inoffp; 4845 savoutoff = *outoffp; 4846 4847 invp = infp->f_vnode; 4848 outvp = outfp->f_vnode; 4849 /* Sanity check the f_flag bits. */ 4850 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE || 4851 (infp->f_flag & FREAD) == 0) { 4852 error = EBADF; 4853 goto out; 4854 } 4855 4856 /* If len == 0, just return 0. */ 4857 if (len == 0) 4858 goto out; 4859 4860 /* 4861 * If infp and outfp refer to the same file, the byte ranges cannot 4862 * overlap. 4863 */ 4864 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len > 4865 savoutoff) || (savinoff > savoutoff && savoutoff + len > 4866 savinoff))) { 4867 error = EINVAL; 4868 goto out; 4869 } 4870 4871 /* Range lock the byte ranges for both invp and outvp. */ 4872 for (;;) { 4873 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp + 4874 len); 4875 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp + 4876 len); 4877 if (rl_rcookie != NULL) 4878 break; 4879 vn_rangelock_unlock(outvp, rl_wcookie); 4880 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len); 4881 vn_rangelock_unlock(invp, rl_rcookie); 4882 } 4883 4884 retlen = len; 4885 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen, 4886 flags, infp->f_cred, outfp->f_cred, td); 4887 out: 4888 if (rl_rcookie != NULL) 4889 vn_rangelock_unlock(invp, rl_rcookie); 4890 if (rl_wcookie != NULL) 4891 vn_rangelock_unlock(outvp, rl_wcookie); 4892 if (savinoff != -1 && (error == EINTR || error == ERESTART)) { 4893 *inoffp = savinoff; 4894 *outoffp = savoutoff; 4895 } 4896 if (outfp != NULL) 4897 fdrop(outfp, td); 4898 if (infp != NULL) 4899 fdrop(infp, td); 4900 td->td_retval[0] = retlen; 4901 return (error); 4902 } 4903 4904 int 4905 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap) 4906 { 4907 off_t inoff, outoff, *inoffp, *outoffp; 4908 int error; 4909 4910 inoffp = outoffp = NULL; 4911 if (uap->inoffp != NULL) { 4912 error = copyin(uap->inoffp, &inoff, sizeof(off_t)); 4913 if (error != 0) 4914 return (error); 4915 inoffp = &inoff; 4916 } 4917 if (uap->outoffp != NULL) { 4918 error = copyin(uap->outoffp, &outoff, sizeof(off_t)); 4919 if (error != 0) 4920 return (error); 4921 outoffp = &outoff; 4922 } 4923 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd, 4924 outoffp, uap->len, uap->flags); 4925 if (error == 0 && uap->inoffp != NULL) 4926 error = copyout(inoffp, uap->inoffp, sizeof(off_t)); 4927 if (error == 0 && uap->outoffp != NULL) 4928 error = copyout(outoffp, uap->outoffp, sizeof(off_t)); 4929 return (error); 4930 } 4931