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 #ifdef COMPAT_FREEBSD11 48 #include <sys/abi_compat.h> 49 #endif 50 #include <sys/bio.h> 51 #include <sys/buf.h> 52 #include <sys/capsicum.h> 53 #include <sys/disk.h> 54 #include <sys/malloc.h> 55 #include <sys/mount.h> 56 #include <sys/mutex.h> 57 #include <sys/sysproto.h> 58 #include <sys/namei.h> 59 #include <sys/filedesc.h> 60 #include <sys/kernel.h> 61 #include <sys/fcntl.h> 62 #include <sys/file.h> 63 #include <sys/filio.h> 64 #include <sys/limits.h> 65 #include <sys/linker.h> 66 #include <sys/rwlock.h> 67 #include <sys/sdt.h> 68 #include <sys/stat.h> 69 #include <sys/sx.h> 70 #include <sys/unistd.h> 71 #include <sys/vnode.h> 72 #include <sys/priv.h> 73 #include <sys/proc.h> 74 #include <sys/dirent.h> 75 #include <sys/jail.h> 76 #include <sys/syscallsubr.h> 77 #include <sys/sysctl.h> 78 #ifdef KTRACE 79 #include <sys/ktrace.h> 80 #endif 81 82 #include <machine/stdarg.h> 83 84 #include <security/audit/audit.h> 85 #include <security/mac/mac_framework.h> 86 87 #include <vm/vm.h> 88 #include <vm/vm_object.h> 89 #include <vm/vm_page.h> 90 #include <vm/uma.h> 91 92 #include <fs/devfs/devfs.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_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, 109 size_t count, struct thread *td); 110 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, 111 const char *path, enum uio_seg segflag); 112 113 uint64_t 114 at2cnpflags(u_int at_flags, u_int mask) 115 { 116 uint64_t res; 117 118 MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) != 119 (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)); 120 121 res = 0; 122 at_flags &= mask; 123 if ((at_flags & AT_RESOLVE_BENEATH) != 0) 124 res |= RBENEATH; 125 if ((at_flags & AT_SYMLINK_FOLLOW) != 0) 126 res |= FOLLOW; 127 /* NOFOLLOW is pseudo flag */ 128 if ((mask & AT_SYMLINK_NOFOLLOW) != 0) { 129 res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW : 130 FOLLOW; 131 } 132 if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0) 133 res |= EMPTYPATH; 134 return (res); 135 } 136 137 int 138 kern_sync(struct thread *td) 139 { 140 struct mount *mp, *nmp; 141 int save; 142 143 mtx_lock(&mountlist_mtx); 144 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 145 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) { 146 nmp = TAILQ_NEXT(mp, mnt_list); 147 continue; 148 } 149 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 150 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 151 save = curthread_pflags_set(TDP_SYNCIO); 152 vfs_periodic(mp, MNT_NOWAIT); 153 VFS_SYNC(mp, MNT_NOWAIT); 154 curthread_pflags_restore(save); 155 vn_finished_write(mp); 156 } 157 mtx_lock(&mountlist_mtx); 158 nmp = TAILQ_NEXT(mp, mnt_list); 159 vfs_unbusy(mp); 160 } 161 mtx_unlock(&mountlist_mtx); 162 return (0); 163 } 164 165 /* 166 * Sync each mounted filesystem. 167 */ 168 #ifndef _SYS_SYSPROTO_H_ 169 struct sync_args { 170 int dummy; 171 }; 172 #endif 173 /* ARGSUSED */ 174 int 175 sys_sync(struct thread *td, struct sync_args *uap) 176 { 177 178 return (kern_sync(td)); 179 } 180 181 /* 182 * Change filesystem quotas. 183 */ 184 #ifndef _SYS_SYSPROTO_H_ 185 struct quotactl_args { 186 char *path; 187 int cmd; 188 int uid; 189 caddr_t arg; 190 }; 191 #endif 192 int 193 sys_quotactl(struct thread *td, struct quotactl_args *uap) 194 { 195 struct mount *mp; 196 struct nameidata nd; 197 int error; 198 bool mp_busy; 199 200 AUDIT_ARG_CMD(uap->cmd); 201 AUDIT_ARG_UID(uap->uid); 202 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS)) 203 return (EPERM); 204 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 205 uap->path); 206 if ((error = namei(&nd)) != 0) 207 return (error); 208 NDFREE_PNBUF(&nd); 209 mp = nd.ni_vp->v_mount; 210 vfs_ref(mp); 211 vput(nd.ni_vp); 212 error = vfs_busy(mp, 0); 213 if (error != 0) { 214 vfs_rel(mp); 215 return (error); 216 } 217 mp_busy = true; 218 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy); 219 220 /* 221 * Since quota on/off operations typically need to open quota 222 * files, the implementation may need to unbusy the mount point 223 * before calling into namei. Otherwise, unmount might be 224 * started between two vfs_busy() invocations (first is ours, 225 * second is from mount point cross-walk code in lookup()), 226 * causing deadlock. 227 * 228 * Avoid unbusying mp if the implementation indicates it has 229 * already done so. 230 */ 231 if (mp_busy) 232 vfs_unbusy(mp); 233 vfs_rel(mp); 234 return (error); 235 } 236 237 /* 238 * Used by statfs conversion routines to scale the block size up if 239 * necessary so that all of the block counts are <= 'max_size'. Note 240 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero 241 * value of 'n'. 242 */ 243 void 244 statfs_scale_blocks(struct statfs *sf, long max_size) 245 { 246 uint64_t count; 247 int shift; 248 249 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__)); 250 251 /* 252 * Attempt to scale the block counts to give a more accurate 253 * overview to userland of the ratio of free space to used 254 * space. To do this, find the largest block count and compute 255 * a divisor that lets it fit into a signed integer <= max_size. 256 */ 257 if (sf->f_bavail < 0) 258 count = -sf->f_bavail; 259 else 260 count = sf->f_bavail; 261 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count)); 262 if (count <= max_size) 263 return; 264 265 count >>= flsl(max_size); 266 shift = 0; 267 while (count > 0) { 268 shift++; 269 count >>=1; 270 } 271 272 sf->f_bsize <<= shift; 273 sf->f_blocks >>= shift; 274 sf->f_bfree >>= shift; 275 sf->f_bavail >>= shift; 276 } 277 278 static int 279 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf) 280 { 281 int error; 282 283 if (mp == NULL) 284 return (EBADF); 285 error = vfs_busy(mp, 0); 286 vfs_rel(mp); 287 if (error != 0) 288 return (error); 289 #ifdef MAC 290 error = mac_mount_check_stat(td->td_ucred, mp); 291 if (error != 0) 292 goto out; 293 #endif 294 error = VFS_STATFS(mp, buf); 295 if (error != 0) 296 goto out; 297 if (priv_check_cred_vfs_generation(td->td_ucred)) { 298 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 299 prison_enforce_statfs(td->td_ucred, mp, buf); 300 } 301 out: 302 vfs_unbusy(mp); 303 return (error); 304 } 305 306 /* 307 * Get filesystem statistics. 308 */ 309 #ifndef _SYS_SYSPROTO_H_ 310 struct statfs_args { 311 char *path; 312 struct statfs *buf; 313 }; 314 #endif 315 int 316 sys_statfs(struct thread *td, struct statfs_args *uap) 317 { 318 struct statfs *sfp; 319 int error; 320 321 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 322 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 323 if (error == 0) 324 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 325 free(sfp, M_STATFS); 326 return (error); 327 } 328 329 int 330 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg, 331 struct statfs *buf) 332 { 333 struct mount *mp; 334 struct nameidata nd; 335 int error; 336 337 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path); 338 error = namei(&nd); 339 if (error != 0) 340 return (error); 341 NDFREE_PNBUF(&nd); 342 mp = vfs_ref_from_vp(nd.ni_vp); 343 vrele(nd.ni_vp); 344 return (kern_do_statfs(td, mp, buf)); 345 } 346 347 /* 348 * Get filesystem statistics. 349 */ 350 #ifndef _SYS_SYSPROTO_H_ 351 struct fstatfs_args { 352 int fd; 353 struct statfs *buf; 354 }; 355 #endif 356 int 357 sys_fstatfs(struct thread *td, struct fstatfs_args *uap) 358 { 359 struct statfs *sfp; 360 int error; 361 362 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 363 error = kern_fstatfs(td, uap->fd, sfp); 364 if (error == 0) 365 error = copyout(sfp, uap->buf, sizeof(struct statfs)); 366 free(sfp, M_STATFS); 367 return (error); 368 } 369 370 int 371 kern_fstatfs(struct thread *td, int fd, struct statfs *buf) 372 { 373 struct file *fp; 374 struct mount *mp; 375 struct vnode *vp; 376 int error; 377 378 AUDIT_ARG_FD(fd); 379 error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp); 380 if (error != 0) 381 return (error); 382 vp = fp->f_vnode; 383 #ifdef AUDIT 384 if (AUDITING_TD(td)) { 385 vn_lock(vp, LK_SHARED | LK_RETRY); 386 AUDIT_ARG_VNODE1(vp); 387 VOP_UNLOCK(vp); 388 } 389 #endif 390 mp = vfs_ref_from_vp(vp); 391 fdrop(fp, td); 392 return (kern_do_statfs(td, mp, buf)); 393 } 394 395 /* 396 * Get statistics on all filesystems. 397 */ 398 #ifndef _SYS_SYSPROTO_H_ 399 struct getfsstat_args { 400 struct statfs *buf; 401 long bufsize; 402 int mode; 403 }; 404 #endif 405 int 406 sys_getfsstat(struct thread *td, struct getfsstat_args *uap) 407 { 408 size_t count; 409 int error; 410 411 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX) 412 return (EINVAL); 413 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count, 414 UIO_USERSPACE, uap->mode); 415 if (error == 0) 416 td->td_retval[0] = count; 417 return (error); 418 } 419 420 /* 421 * If (bufsize > 0 && bufseg == UIO_SYSSPACE) 422 * The caller is responsible for freeing memory which will be allocated 423 * in '*buf'. 424 */ 425 int 426 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize, 427 size_t *countp, enum uio_seg bufseg, int mode) 428 { 429 struct mount *mp, *nmp; 430 struct statfs *sfsp, *sp, *sptmp, *tofree; 431 size_t count, maxcount; 432 int error; 433 434 switch (mode) { 435 case MNT_WAIT: 436 case MNT_NOWAIT: 437 break; 438 default: 439 if (bufseg == UIO_SYSSPACE) 440 *buf = NULL; 441 return (EINVAL); 442 } 443 restart: 444 maxcount = bufsize / sizeof(struct statfs); 445 if (bufsize == 0) { 446 sfsp = NULL; 447 tofree = NULL; 448 } else if (bufseg == UIO_USERSPACE) { 449 sfsp = *buf; 450 tofree = NULL; 451 } else /* if (bufseg == UIO_SYSSPACE) */ { 452 count = 0; 453 mtx_lock(&mountlist_mtx); 454 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 455 count++; 456 } 457 mtx_unlock(&mountlist_mtx); 458 if (maxcount > count) 459 maxcount = count; 460 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs), 461 M_STATFS, M_WAITOK); 462 } 463 464 count = 0; 465 466 /* 467 * If there is no target buffer they only want the count. 468 * 469 * This could be TAILQ_FOREACH but it is open-coded to match the original 470 * code below. 471 */ 472 if (sfsp == NULL) { 473 mtx_lock(&mountlist_mtx); 474 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 475 if (prison_canseemount(td->td_ucred, mp) != 0) { 476 nmp = TAILQ_NEXT(mp, mnt_list); 477 continue; 478 } 479 #ifdef MAC 480 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 481 nmp = TAILQ_NEXT(mp, mnt_list); 482 continue; 483 } 484 #endif 485 count++; 486 nmp = TAILQ_NEXT(mp, mnt_list); 487 } 488 mtx_unlock(&mountlist_mtx); 489 *countp = count; 490 return (0); 491 } 492 493 /* 494 * They want the entire thing. 495 * 496 * Short-circuit the corner case of no room for anything, avoids 497 * relocking below. 498 */ 499 if (maxcount < 1) { 500 goto out; 501 } 502 503 mtx_lock(&mountlist_mtx); 504 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 505 if (prison_canseemount(td->td_ucred, mp) != 0) { 506 nmp = TAILQ_NEXT(mp, mnt_list); 507 continue; 508 } 509 #ifdef MAC 510 if (mac_mount_check_stat(td->td_ucred, mp) != 0) { 511 nmp = TAILQ_NEXT(mp, mnt_list); 512 continue; 513 } 514 #endif 515 if (mode == MNT_WAIT) { 516 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) { 517 /* 518 * If vfs_busy() failed, and MBF_NOWAIT 519 * wasn't passed, then the mp is gone. 520 * Furthermore, because of MBF_MNTLSTLOCK, 521 * the mountlist_mtx was dropped. We have 522 * no other choice than to start over. 523 */ 524 mtx_unlock(&mountlist_mtx); 525 free(tofree, M_STATFS); 526 goto restart; 527 } 528 } else { 529 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) { 530 nmp = TAILQ_NEXT(mp, mnt_list); 531 continue; 532 } 533 } 534 sp = &mp->mnt_stat; 535 /* 536 * If MNT_NOWAIT is specified, do not refresh 537 * the fsstat cache. 538 */ 539 if (mode != MNT_NOWAIT) { 540 error = VFS_STATFS(mp, sp); 541 if (error != 0) { 542 mtx_lock(&mountlist_mtx); 543 nmp = TAILQ_NEXT(mp, mnt_list); 544 vfs_unbusy(mp); 545 continue; 546 } 547 } 548 if (priv_check_cred_vfs_generation(td->td_ucred)) { 549 sptmp = malloc(sizeof(struct statfs), M_STATFS, 550 M_WAITOK); 551 *sptmp = *sp; 552 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0; 553 prison_enforce_statfs(td->td_ucred, mp, sptmp); 554 sp = sptmp; 555 } else 556 sptmp = NULL; 557 if (bufseg == UIO_SYSSPACE) { 558 bcopy(sp, sfsp, sizeof(*sp)); 559 free(sptmp, M_STATFS); 560 } else /* if (bufseg == UIO_USERSPACE) */ { 561 error = copyout(sp, sfsp, sizeof(*sp)); 562 free(sptmp, M_STATFS); 563 if (error != 0) { 564 vfs_unbusy(mp); 565 return (error); 566 } 567 } 568 sfsp++; 569 count++; 570 571 if (count == maxcount) { 572 vfs_unbusy(mp); 573 goto out; 574 } 575 576 mtx_lock(&mountlist_mtx); 577 nmp = TAILQ_NEXT(mp, mnt_list); 578 vfs_unbusy(mp); 579 } 580 mtx_unlock(&mountlist_mtx); 581 out: 582 *countp = count; 583 return (0); 584 } 585 586 #ifdef COMPAT_FREEBSD4 587 /* 588 * Get old format filesystem statistics. 589 */ 590 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *); 591 592 #ifndef _SYS_SYSPROTO_H_ 593 struct freebsd4_statfs_args { 594 char *path; 595 struct ostatfs *buf; 596 }; 597 #endif 598 int 599 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap) 600 { 601 struct ostatfs osb; 602 struct statfs *sfp; 603 int error; 604 605 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 606 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 607 if (error == 0) { 608 freebsd4_cvtstatfs(sfp, &osb); 609 error = copyout(&osb, uap->buf, sizeof(osb)); 610 } 611 free(sfp, M_STATFS); 612 return (error); 613 } 614 615 /* 616 * Get filesystem statistics. 617 */ 618 #ifndef _SYS_SYSPROTO_H_ 619 struct freebsd4_fstatfs_args { 620 int fd; 621 struct ostatfs *buf; 622 }; 623 #endif 624 int 625 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap) 626 { 627 struct ostatfs osb; 628 struct statfs *sfp; 629 int error; 630 631 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 632 error = kern_fstatfs(td, uap->fd, sfp); 633 if (error == 0) { 634 freebsd4_cvtstatfs(sfp, &osb); 635 error = copyout(&osb, uap->buf, sizeof(osb)); 636 } 637 free(sfp, M_STATFS); 638 return (error); 639 } 640 641 /* 642 * Get statistics on all filesystems. 643 */ 644 #ifndef _SYS_SYSPROTO_H_ 645 struct freebsd4_getfsstat_args { 646 struct ostatfs *buf; 647 long bufsize; 648 int mode; 649 }; 650 #endif 651 int 652 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap) 653 { 654 struct statfs *buf, *sp; 655 struct ostatfs osb; 656 size_t count, size; 657 int error; 658 659 if (uap->bufsize < 0) 660 return (EINVAL); 661 count = uap->bufsize / sizeof(struct ostatfs); 662 if (count > SIZE_MAX / sizeof(struct statfs)) 663 return (EINVAL); 664 size = count * sizeof(struct statfs); 665 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, 666 uap->mode); 667 if (error == 0) 668 td->td_retval[0] = count; 669 if (size != 0) { 670 sp = buf; 671 while (count != 0 && error == 0) { 672 freebsd4_cvtstatfs(sp, &osb); 673 error = copyout(&osb, uap->buf, sizeof(osb)); 674 sp++; 675 uap->buf++; 676 count--; 677 } 678 free(buf, M_STATFS); 679 } 680 return (error); 681 } 682 683 /* 684 * Implement fstatfs() for (NFS) file handles. 685 */ 686 #ifndef _SYS_SYSPROTO_H_ 687 struct freebsd4_fhstatfs_args { 688 struct fhandle *u_fhp; 689 struct ostatfs *buf; 690 }; 691 #endif 692 int 693 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap) 694 { 695 struct ostatfs osb; 696 struct statfs *sfp; 697 fhandle_t fh; 698 int error; 699 700 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 701 if (error != 0) 702 return (error); 703 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 704 error = kern_fhstatfs(td, fh, sfp); 705 if (error == 0) { 706 freebsd4_cvtstatfs(sfp, &osb); 707 error = copyout(&osb, uap->buf, sizeof(osb)); 708 } 709 free(sfp, M_STATFS); 710 return (error); 711 } 712 713 /* 714 * Convert a new format statfs structure to an old format statfs structure. 715 */ 716 static void 717 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp) 718 { 719 720 statfs_scale_blocks(nsp, LONG_MAX); 721 bzero(osp, sizeof(*osp)); 722 osp->f_bsize = nsp->f_bsize; 723 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX); 724 osp->f_blocks = nsp->f_blocks; 725 osp->f_bfree = nsp->f_bfree; 726 osp->f_bavail = nsp->f_bavail; 727 osp->f_files = MIN(nsp->f_files, LONG_MAX); 728 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX); 729 osp->f_owner = nsp->f_owner; 730 osp->f_type = nsp->f_type; 731 osp->f_flags = nsp->f_flags; 732 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX); 733 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX); 734 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX); 735 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX); 736 strlcpy(osp->f_fstypename, nsp->f_fstypename, 737 MIN(MFSNAMELEN, OMFSNAMELEN)); 738 strlcpy(osp->f_mntonname, nsp->f_mntonname, 739 MIN(MNAMELEN, OMNAMELEN)); 740 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 741 MIN(MNAMELEN, OMNAMELEN)); 742 osp->f_fsid = nsp->f_fsid; 743 } 744 #endif /* COMPAT_FREEBSD4 */ 745 746 #if defined(COMPAT_FREEBSD11) 747 /* 748 * Get old format filesystem statistics. 749 */ 750 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *); 751 752 int 753 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap) 754 { 755 struct freebsd11_statfs osb; 756 struct statfs *sfp; 757 int error; 758 759 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 760 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp); 761 if (error == 0) { 762 freebsd11_cvtstatfs(sfp, &osb); 763 error = copyout(&osb, uap->buf, sizeof(osb)); 764 } 765 free(sfp, M_STATFS); 766 return (error); 767 } 768 769 /* 770 * Get filesystem statistics. 771 */ 772 int 773 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap) 774 { 775 struct freebsd11_statfs osb; 776 struct statfs *sfp; 777 int error; 778 779 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 780 error = kern_fstatfs(td, uap->fd, sfp); 781 if (error == 0) { 782 freebsd11_cvtstatfs(sfp, &osb); 783 error = copyout(&osb, uap->buf, sizeof(osb)); 784 } 785 free(sfp, M_STATFS); 786 return (error); 787 } 788 789 /* 790 * Get statistics on all filesystems. 791 */ 792 int 793 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap) 794 { 795 return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode)); 796 } 797 798 int 799 kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf, 800 long bufsize, int mode) 801 { 802 struct freebsd11_statfs osb; 803 struct statfs *buf, *sp; 804 size_t count, size; 805 int error; 806 807 if (bufsize < 0) 808 return (EINVAL); 809 810 count = bufsize / sizeof(struct ostatfs); 811 size = count * sizeof(struct statfs); 812 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode); 813 if (error == 0) 814 td->td_retval[0] = count; 815 if (size > 0) { 816 sp = buf; 817 while (count > 0 && error == 0) { 818 freebsd11_cvtstatfs(sp, &osb); 819 error = copyout(&osb, ubuf, sizeof(osb)); 820 sp++; 821 ubuf++; 822 count--; 823 } 824 free(buf, M_STATFS); 825 } 826 return (error); 827 } 828 829 /* 830 * Implement fstatfs() for (NFS) file handles. 831 */ 832 int 833 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap) 834 { 835 struct freebsd11_statfs osb; 836 struct statfs *sfp; 837 fhandle_t fh; 838 int error; 839 840 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 841 if (error) 842 return (error); 843 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 844 error = kern_fhstatfs(td, fh, sfp); 845 if (error == 0) { 846 freebsd11_cvtstatfs(sfp, &osb); 847 error = copyout(&osb, uap->buf, sizeof(osb)); 848 } 849 free(sfp, M_STATFS); 850 return (error); 851 } 852 853 /* 854 * Convert a new format statfs structure to an old format statfs structure. 855 */ 856 static void 857 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp) 858 { 859 860 bzero(osp, sizeof(*osp)); 861 osp->f_version = FREEBSD11_STATFS_VERSION; 862 osp->f_type = nsp->f_type; 863 osp->f_flags = nsp->f_flags; 864 osp->f_bsize = nsp->f_bsize; 865 osp->f_iosize = nsp->f_iosize; 866 osp->f_blocks = nsp->f_blocks; 867 osp->f_bfree = nsp->f_bfree; 868 osp->f_bavail = nsp->f_bavail; 869 osp->f_files = nsp->f_files; 870 osp->f_ffree = nsp->f_ffree; 871 osp->f_syncwrites = nsp->f_syncwrites; 872 osp->f_asyncwrites = nsp->f_asyncwrites; 873 osp->f_syncreads = nsp->f_syncreads; 874 osp->f_asyncreads = nsp->f_asyncreads; 875 osp->f_namemax = nsp->f_namemax; 876 osp->f_owner = nsp->f_owner; 877 osp->f_fsid = nsp->f_fsid; 878 strlcpy(osp->f_fstypename, nsp->f_fstypename, 879 MIN(MFSNAMELEN, sizeof(osp->f_fstypename))); 880 strlcpy(osp->f_mntonname, nsp->f_mntonname, 881 MIN(MNAMELEN, sizeof(osp->f_mntonname))); 882 strlcpy(osp->f_mntfromname, nsp->f_mntfromname, 883 MIN(MNAMELEN, sizeof(osp->f_mntfromname))); 884 } 885 #endif /* COMPAT_FREEBSD11 */ 886 887 /* 888 * Change current working directory to a given file descriptor. 889 */ 890 #ifndef _SYS_SYSPROTO_H_ 891 struct fchdir_args { 892 int fd; 893 }; 894 #endif 895 int 896 sys_fchdir(struct thread *td, struct fchdir_args *uap) 897 { 898 struct vnode *vp, *tdp; 899 struct mount *mp; 900 struct file *fp; 901 int error; 902 903 AUDIT_ARG_FD(uap->fd); 904 error = getvnode_path(td, uap->fd, &cap_fchdir_rights, 905 &fp); 906 if (error != 0) 907 return (error); 908 vp = fp->f_vnode; 909 vrefact(vp); 910 fdrop(fp, td); 911 vn_lock(vp, LK_SHARED | LK_RETRY); 912 AUDIT_ARG_VNODE1(vp); 913 error = change_dir(vp, td); 914 while (!error && (mp = vp->v_mountedhere) != NULL) { 915 if (vfs_busy(mp, 0)) 916 continue; 917 error = VFS_ROOT(mp, LK_SHARED, &tdp); 918 vfs_unbusy(mp); 919 if (error != 0) 920 break; 921 vput(vp); 922 vp = tdp; 923 } 924 if (error != 0) { 925 vput(vp); 926 return (error); 927 } 928 VOP_UNLOCK(vp); 929 pwd_chdir(td, vp); 930 return (0); 931 } 932 933 /* 934 * Change current working directory (``.''). 935 */ 936 #ifndef _SYS_SYSPROTO_H_ 937 struct chdir_args { 938 char *path; 939 }; 940 #endif 941 int 942 sys_chdir(struct thread *td, struct chdir_args *uap) 943 { 944 945 return (kern_chdir(td, uap->path, UIO_USERSPACE)); 946 } 947 948 int 949 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg) 950 { 951 struct nameidata nd; 952 int error; 953 954 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 955 pathseg, path); 956 if ((error = namei(&nd)) != 0) 957 return (error); 958 if ((error = change_dir(nd.ni_vp, td)) != 0) { 959 vput(nd.ni_vp); 960 NDFREE_PNBUF(&nd); 961 return (error); 962 } 963 VOP_UNLOCK(nd.ni_vp); 964 NDFREE_PNBUF(&nd); 965 pwd_chdir(td, nd.ni_vp); 966 return (0); 967 } 968 969 static int unprivileged_chroot = 0; 970 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW, 971 &unprivileged_chroot, 0, 972 "Unprivileged processes can use chroot(2)"); 973 /* 974 * Change notion of root (``/'') directory. 975 */ 976 #ifndef _SYS_SYSPROTO_H_ 977 struct chroot_args { 978 char *path; 979 }; 980 #endif 981 int 982 sys_chroot(struct thread *td, struct chroot_args *uap) 983 { 984 struct nameidata nd; 985 struct proc *p; 986 int error; 987 988 error = priv_check(td, PRIV_VFS_CHROOT); 989 if (error != 0) { 990 p = td->td_proc; 991 PROC_LOCK(p); 992 if (unprivileged_chroot == 0 || 993 (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) { 994 PROC_UNLOCK(p); 995 return (error); 996 } 997 PROC_UNLOCK(p); 998 } 999 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1, 1000 UIO_USERSPACE, uap->path); 1001 error = namei(&nd); 1002 if (error != 0) 1003 return (error); 1004 NDFREE_PNBUF(&nd); 1005 error = change_dir(nd.ni_vp, td); 1006 if (error != 0) 1007 goto e_vunlock; 1008 #ifdef MAC 1009 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp); 1010 if (error != 0) 1011 goto e_vunlock; 1012 #endif 1013 VOP_UNLOCK(nd.ni_vp); 1014 error = pwd_chroot(td, nd.ni_vp); 1015 vrele(nd.ni_vp); 1016 return (error); 1017 e_vunlock: 1018 vput(nd.ni_vp); 1019 return (error); 1020 } 1021 1022 /* 1023 * Common routine for chroot and chdir. Callers must provide a locked vnode 1024 * instance. 1025 */ 1026 int 1027 change_dir(struct vnode *vp, struct thread *td) 1028 { 1029 #ifdef MAC 1030 int error; 1031 #endif 1032 1033 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked"); 1034 if (vp->v_type != VDIR) 1035 return (ENOTDIR); 1036 #ifdef MAC 1037 error = mac_vnode_check_chdir(td->td_ucred, vp); 1038 if (error != 0) 1039 return (error); 1040 #endif 1041 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td)); 1042 } 1043 1044 static __inline void 1045 flags_to_rights(int flags, cap_rights_t *rightsp) 1046 { 1047 if (flags & O_EXEC) { 1048 cap_rights_set_one(rightsp, CAP_FEXECVE); 1049 if (flags & O_PATH) 1050 return; 1051 } else { 1052 switch ((flags & O_ACCMODE)) { 1053 case O_RDONLY: 1054 cap_rights_set_one(rightsp, CAP_READ); 1055 break; 1056 case O_RDWR: 1057 cap_rights_set_one(rightsp, CAP_READ); 1058 /* FALLTHROUGH */ 1059 case O_WRONLY: 1060 cap_rights_set_one(rightsp, CAP_WRITE); 1061 if (!(flags & (O_APPEND | O_TRUNC))) 1062 cap_rights_set_one(rightsp, CAP_SEEK); 1063 break; 1064 } 1065 } 1066 1067 if (flags & O_CREAT) 1068 cap_rights_set_one(rightsp, CAP_CREATE); 1069 1070 if (flags & O_TRUNC) 1071 cap_rights_set_one(rightsp, CAP_FTRUNCATE); 1072 1073 if (flags & (O_SYNC | O_FSYNC)) 1074 cap_rights_set_one(rightsp, CAP_FSYNC); 1075 1076 if (flags & (O_EXLOCK | O_SHLOCK)) 1077 cap_rights_set_one(rightsp, CAP_FLOCK); 1078 } 1079 1080 /* 1081 * Check permissions, allocate an open file structure, and call the device 1082 * open routine if any. 1083 */ 1084 #ifndef _SYS_SYSPROTO_H_ 1085 struct open_args { 1086 char *path; 1087 int flags; 1088 int mode; 1089 }; 1090 #endif 1091 int 1092 sys_open(struct thread *td, struct open_args *uap) 1093 { 1094 1095 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1096 uap->flags, uap->mode)); 1097 } 1098 1099 #ifndef _SYS_SYSPROTO_H_ 1100 struct openat_args { 1101 int fd; 1102 char *path; 1103 int flag; 1104 int mode; 1105 }; 1106 #endif 1107 int 1108 sys_openat(struct thread *td, struct openat_args *uap) 1109 { 1110 1111 AUDIT_ARG_FD(uap->fd); 1112 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 1113 uap->mode)); 1114 } 1115 1116 int 1117 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg, 1118 int flags, int mode) 1119 { 1120 struct proc *p = td->td_proc; 1121 struct filedesc *fdp; 1122 struct pwddesc *pdp; 1123 struct file *fp; 1124 struct vnode *vp; 1125 struct nameidata nd; 1126 cap_rights_t rights; 1127 int cmode, error, indx; 1128 1129 indx = -1; 1130 fdp = p->p_fd; 1131 pdp = p->p_pd; 1132 1133 AUDIT_ARG_FFLAGS(flags); 1134 AUDIT_ARG_MODE(mode); 1135 cap_rights_init_one(&rights, CAP_LOOKUP); 1136 flags_to_rights(flags, &rights); 1137 1138 /* 1139 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 1140 * may be specified. On the other hand, for O_PATH any mode 1141 * except O_EXEC is ignored. 1142 */ 1143 if ((flags & O_PATH) != 0) { 1144 flags &= ~(O_CREAT | O_ACCMODE); 1145 } else if ((flags & O_EXEC) != 0) { 1146 if (flags & O_ACCMODE) 1147 return (EINVAL); 1148 } else if ((flags & O_ACCMODE) == O_ACCMODE) { 1149 return (EINVAL); 1150 } else { 1151 flags = FFLAGS(flags); 1152 } 1153 1154 /* 1155 * Allocate a file structure. The descriptor to reference it 1156 * is allocated and used by finstall_refed() below. 1157 */ 1158 error = falloc_noinstall(td, &fp); 1159 if (error != 0) 1160 return (error); 1161 /* Set the flags early so the finit in devfs can pick them up. */ 1162 fp->f_flag = flags & FMASK; 1163 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT; 1164 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | WANTIOCTLCAPS, 1165 pathseg, path, fd, &rights); 1166 td->td_dupfd = -1; /* XXX check for fdopen */ 1167 error = vn_open_cred(&nd, &flags, cmode, VN_OPEN_WANTIOCTLCAPS, 1168 td->td_ucred, fp); 1169 if (error != 0) { 1170 /* 1171 * If the vn_open replaced the method vector, something 1172 * wonderous happened deep below and we just pass it up 1173 * pretending we know what we do. 1174 */ 1175 if (error == ENXIO && fp->f_ops != &badfileops) { 1176 MPASS((flags & O_PATH) == 0); 1177 goto success; 1178 } 1179 1180 /* 1181 * Handle special fdopen() case. bleh. 1182 * 1183 * Don't do this for relative (capability) lookups; we don't 1184 * understand exactly what would happen, and we don't think 1185 * that it ever should. 1186 */ 1187 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 && 1188 (error == ENODEV || error == ENXIO) && 1189 td->td_dupfd >= 0) { 1190 error = dupfdopen(td, fdp, td->td_dupfd, flags, error, 1191 &indx); 1192 if (error == 0) 1193 goto success; 1194 } 1195 1196 goto bad; 1197 } 1198 td->td_dupfd = 0; 1199 NDFREE_PNBUF(&nd); 1200 vp = nd.ni_vp; 1201 1202 /* 1203 * Store the vnode, for any f_type. Typically, the vnode use 1204 * count is decremented by direct call to vn_closefile() for 1205 * files that switched type in the cdevsw fdopen() method. 1206 */ 1207 fp->f_vnode = vp; 1208 1209 /* 1210 * If the file wasn't claimed by devfs bind it to the normal 1211 * vnode operations here. 1212 */ 1213 if (fp->f_ops == &badfileops) { 1214 KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0, 1215 ("Unexpected fifo fp %p vp %p", fp, vp)); 1216 if ((flags & O_PATH) != 0) { 1217 finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED), 1218 DTYPE_VNODE, NULL, &path_fileops); 1219 } else { 1220 finit_vnode(fp, flags, NULL, &vnops); 1221 } 1222 } 1223 1224 VOP_UNLOCK(vp); 1225 if (flags & O_TRUNC) { 1226 error = fo_truncate(fp, 0, td->td_ucred, td); 1227 if (error != 0) 1228 goto bad; 1229 } 1230 success: 1231 /* 1232 * If we haven't already installed the FD (for dupfdopen), do so now. 1233 */ 1234 if (indx == -1) { 1235 struct filecaps *fcaps; 1236 1237 #ifdef CAPABILITIES 1238 if ((nd.ni_resflags & NIRES_STRICTREL) != 0) 1239 fcaps = &nd.ni_filecaps; 1240 else 1241 #endif 1242 fcaps = NULL; 1243 error = finstall_refed(td, fp, &indx, flags, fcaps); 1244 /* On success finstall_refed() consumes fcaps. */ 1245 if (error != 0) { 1246 goto bad; 1247 } 1248 } else { 1249 NDFREE_IOCTLCAPS(&nd); 1250 falloc_abort(td, fp); 1251 } 1252 1253 td->td_retval[0] = indx; 1254 return (0); 1255 bad: 1256 KASSERT(indx == -1, ("indx=%d, should be -1", indx)); 1257 NDFREE_IOCTLCAPS(&nd); 1258 falloc_abort(td, fp); 1259 return (error); 1260 } 1261 1262 #ifdef COMPAT_43 1263 /* 1264 * Create a file. 1265 */ 1266 #ifndef _SYS_SYSPROTO_H_ 1267 struct ocreat_args { 1268 char *path; 1269 int mode; 1270 }; 1271 #endif 1272 int 1273 ocreat(struct thread *td, struct ocreat_args *uap) 1274 { 1275 1276 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1277 O_WRONLY | O_CREAT | O_TRUNC, uap->mode)); 1278 } 1279 #endif /* COMPAT_43 */ 1280 1281 /* 1282 * Create a special file. 1283 */ 1284 #ifndef _SYS_SYSPROTO_H_ 1285 struct mknodat_args { 1286 int fd; 1287 char *path; 1288 mode_t mode; 1289 dev_t dev; 1290 }; 1291 #endif 1292 int 1293 sys_mknodat(struct thread *td, struct mknodat_args *uap) 1294 { 1295 1296 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1297 uap->dev)); 1298 } 1299 1300 #if defined(COMPAT_FREEBSD11) 1301 int 1302 freebsd11_mknod(struct thread *td, 1303 struct freebsd11_mknod_args *uap) 1304 { 1305 1306 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1307 uap->mode, uap->dev)); 1308 } 1309 1310 int 1311 freebsd11_mknodat(struct thread *td, 1312 struct freebsd11_mknodat_args *uap) 1313 { 1314 1315 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode, 1316 uap->dev)); 1317 } 1318 #endif /* COMPAT_FREEBSD11 */ 1319 1320 int 1321 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg, 1322 int mode, dev_t dev) 1323 { 1324 struct vnode *vp; 1325 struct mount *mp; 1326 struct vattr vattr; 1327 struct nameidata nd; 1328 int error, whiteout = 0; 1329 1330 AUDIT_ARG_MODE(mode); 1331 AUDIT_ARG_DEV(dev); 1332 switch (mode & S_IFMT) { 1333 case S_IFCHR: 1334 case S_IFBLK: 1335 error = priv_check(td, PRIV_VFS_MKNOD_DEV); 1336 if (error == 0 && dev == VNOVAL) 1337 error = EINVAL; 1338 break; 1339 case S_IFWHT: 1340 error = priv_check(td, PRIV_VFS_MKNOD_WHT); 1341 break; 1342 case S_IFIFO: 1343 if (dev == 0) 1344 return (kern_mkfifoat(td, fd, path, pathseg, mode)); 1345 /* FALLTHROUGH */ 1346 default: 1347 error = EINVAL; 1348 break; 1349 } 1350 if (error != 0) 1351 return (error); 1352 NDPREINIT(&nd); 1353 restart: 1354 bwillwrite(); 1355 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, 1356 pathseg, path, fd, &cap_mknodat_rights); 1357 if ((error = namei(&nd)) != 0) 1358 return (error); 1359 vp = nd.ni_vp; 1360 if (vp != NULL) { 1361 NDFREE_PNBUF(&nd); 1362 if (vp == nd.ni_dvp) 1363 vrele(nd.ni_dvp); 1364 else 1365 vput(nd.ni_dvp); 1366 vrele(vp); 1367 return (EEXIST); 1368 } else { 1369 VATTR_NULL(&vattr); 1370 vattr.va_mode = (mode & ALLPERMS) & 1371 ~td->td_proc->p_pd->pd_cmask; 1372 vattr.va_rdev = dev; 1373 whiteout = 0; 1374 1375 switch (mode & S_IFMT) { 1376 case S_IFCHR: 1377 vattr.va_type = VCHR; 1378 break; 1379 case S_IFBLK: 1380 vattr.va_type = VBLK; 1381 break; 1382 case S_IFWHT: 1383 whiteout = 1; 1384 break; 1385 default: 1386 panic("kern_mknod: invalid mode"); 1387 } 1388 } 1389 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1390 NDFREE_PNBUF(&nd); 1391 vput(nd.ni_dvp); 1392 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 1393 return (error); 1394 goto restart; 1395 } 1396 #ifdef MAC 1397 if (error == 0 && !whiteout) 1398 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, 1399 &nd.ni_cnd, &vattr); 1400 #endif 1401 if (error == 0) { 1402 if (whiteout) 1403 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1404 else { 1405 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1406 &nd.ni_cnd, &vattr); 1407 } 1408 } 1409 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL, 1410 true); 1411 vn_finished_write(mp); 1412 NDFREE_PNBUF(&nd); 1413 if (error == ERELOOKUP) 1414 goto restart; 1415 return (error); 1416 } 1417 1418 /* 1419 * Create a named pipe. 1420 */ 1421 #ifndef _SYS_SYSPROTO_H_ 1422 struct mkfifo_args { 1423 char *path; 1424 int mode; 1425 }; 1426 #endif 1427 int 1428 sys_mkfifo(struct thread *td, struct mkfifo_args *uap) 1429 { 1430 1431 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 1432 uap->mode)); 1433 } 1434 1435 #ifndef _SYS_SYSPROTO_H_ 1436 struct mkfifoat_args { 1437 int fd; 1438 char *path; 1439 mode_t mode; 1440 }; 1441 #endif 1442 int 1443 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap) 1444 { 1445 1446 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE, 1447 uap->mode)); 1448 } 1449 1450 int 1451 kern_mkfifoat(struct thread *td, int fd, const char *path, 1452 enum uio_seg pathseg, int mode) 1453 { 1454 struct mount *mp; 1455 struct vattr vattr; 1456 struct nameidata nd; 1457 int error; 1458 1459 AUDIT_ARG_MODE(mode); 1460 NDPREINIT(&nd); 1461 restart: 1462 bwillwrite(); 1463 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, 1464 pathseg, path, fd, &cap_mkfifoat_rights); 1465 if ((error = namei(&nd)) != 0) 1466 return (error); 1467 if (nd.ni_vp != NULL) { 1468 NDFREE_PNBUF(&nd); 1469 if (nd.ni_vp == nd.ni_dvp) 1470 vrele(nd.ni_dvp); 1471 else 1472 vput(nd.ni_dvp); 1473 vrele(nd.ni_vp); 1474 return (EEXIST); 1475 } 1476 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1477 NDFREE_PNBUF(&nd); 1478 vput(nd.ni_dvp); 1479 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 1480 return (error); 1481 goto restart; 1482 } 1483 VATTR_NULL(&vattr); 1484 vattr.va_type = VFIFO; 1485 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask; 1486 #ifdef MAC 1487 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1488 &vattr); 1489 if (error != 0) 1490 goto out; 1491 #endif 1492 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1493 #ifdef MAC 1494 out: 1495 #endif 1496 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true); 1497 vn_finished_write(mp); 1498 NDFREE_PNBUF(&nd); 1499 if (error == ERELOOKUP) 1500 goto restart; 1501 return (error); 1502 } 1503 1504 /* 1505 * Make a hard file link. 1506 */ 1507 #ifndef _SYS_SYSPROTO_H_ 1508 struct link_args { 1509 char *path; 1510 char *link; 1511 }; 1512 #endif 1513 int 1514 sys_link(struct thread *td, struct link_args *uap) 1515 { 1516 1517 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link, 1518 UIO_USERSPACE, AT_SYMLINK_FOLLOW)); 1519 } 1520 1521 #ifndef _SYS_SYSPROTO_H_ 1522 struct linkat_args { 1523 int fd1; 1524 char *path1; 1525 int fd2; 1526 char *path2; 1527 int flag; 1528 }; 1529 #endif 1530 int 1531 sys_linkat(struct thread *td, struct linkat_args *uap) 1532 { 1533 1534 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2, 1535 UIO_USERSPACE, uap->flag)); 1536 } 1537 1538 int hardlink_check_uid = 0; 1539 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW, 1540 &hardlink_check_uid, 0, 1541 "Unprivileged processes cannot create hard links to files owned by other " 1542 "users"); 1543 static int hardlink_check_gid = 0; 1544 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW, 1545 &hardlink_check_gid, 0, 1546 "Unprivileged processes cannot create hard links to files owned by other " 1547 "groups"); 1548 1549 static int 1550 can_hardlink(struct vnode *vp, struct ucred *cred) 1551 { 1552 struct vattr va; 1553 int error; 1554 1555 if (!hardlink_check_uid && !hardlink_check_gid) 1556 return (0); 1557 1558 error = VOP_GETATTR(vp, &va, cred); 1559 if (error != 0) 1560 return (error); 1561 1562 if (hardlink_check_uid && cred->cr_uid != va.va_uid) { 1563 error = priv_check_cred(cred, PRIV_VFS_LINK); 1564 if (error != 0) 1565 return (error); 1566 } 1567 1568 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) { 1569 error = priv_check_cred(cred, PRIV_VFS_LINK); 1570 if (error != 0) 1571 return (error); 1572 } 1573 1574 return (0); 1575 } 1576 1577 int 1578 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1, 1579 const char *path2, enum uio_seg segflag, int flag) 1580 { 1581 struct nameidata nd; 1582 int error; 1583 1584 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | 1585 AT_EMPTY_PATH)) != 0) 1586 return (EINVAL); 1587 1588 NDPREINIT(&nd); 1589 do { 1590 bwillwrite(); 1591 NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag, 1592 AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH), 1593 segflag, path1, fd1, &cap_linkat_source_rights); 1594 if ((error = namei(&nd)) != 0) 1595 return (error); 1596 NDFREE_PNBUF(&nd); 1597 if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) { 1598 error = priv_check(td, PRIV_VFS_FHOPEN); 1599 if (error != 0) { 1600 vrele(nd.ni_vp); 1601 return (error); 1602 } 1603 } 1604 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag); 1605 } while (error == EAGAIN || error == ERELOOKUP); 1606 return (error); 1607 } 1608 1609 static int 1610 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path, 1611 enum uio_seg segflag) 1612 { 1613 struct nameidata nd; 1614 struct mount *mp; 1615 int error; 1616 1617 if (vp->v_type == VDIR) { 1618 vrele(vp); 1619 return (EPERM); /* POSIX */ 1620 } 1621 NDINIT_ATRIGHTS(&nd, CREATE, 1622 LOCKPARENT | AUDITVNODE2 | NOCACHE, segflag, path, fd, 1623 &cap_linkat_target_rights); 1624 if ((error = namei(&nd)) == 0) { 1625 if (nd.ni_vp != NULL) { 1626 NDFREE_PNBUF(&nd); 1627 if (nd.ni_dvp == nd.ni_vp) 1628 vrele(nd.ni_dvp); 1629 else 1630 vput(nd.ni_dvp); 1631 vrele(nd.ni_vp); 1632 vrele(vp); 1633 return (EEXIST); 1634 } else if (nd.ni_dvp->v_mount != vp->v_mount) { 1635 /* 1636 * Cross-device link. No need to recheck 1637 * vp->v_type, since it cannot change, except 1638 * to VBAD. 1639 */ 1640 NDFREE_PNBUF(&nd); 1641 vput(nd.ni_dvp); 1642 vrele(vp); 1643 return (EXDEV); 1644 } else if (vn_lock(vp, LK_EXCLUSIVE) == 0) { 1645 error = can_hardlink(vp, td->td_ucred); 1646 #ifdef MAC 1647 if (error == 0) 1648 error = mac_vnode_check_link(td->td_ucred, 1649 nd.ni_dvp, vp, &nd.ni_cnd); 1650 #endif 1651 if (error != 0) { 1652 vput(vp); 1653 vput(nd.ni_dvp); 1654 NDFREE_PNBUF(&nd); 1655 return (error); 1656 } 1657 error = vn_start_write(vp, &mp, V_NOWAIT); 1658 if (error != 0) { 1659 vput(vp); 1660 vput(nd.ni_dvp); 1661 NDFREE_PNBUF(&nd); 1662 error = vn_start_write(NULL, &mp, 1663 V_XSLEEP | V_PCATCH); 1664 if (error != 0) 1665 return (error); 1666 return (EAGAIN); 1667 } 1668 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1669 VOP_VPUT_PAIR(nd.ni_dvp, &vp, true); 1670 vn_finished_write(mp); 1671 NDFREE_PNBUF(&nd); 1672 vp = NULL; 1673 } else { 1674 vput(nd.ni_dvp); 1675 NDFREE_PNBUF(&nd); 1676 vrele(vp); 1677 return (EAGAIN); 1678 } 1679 } 1680 if (vp != NULL) 1681 vrele(vp); 1682 return (error); 1683 } 1684 1685 /* 1686 * Make a symbolic link. 1687 */ 1688 #ifndef _SYS_SYSPROTO_H_ 1689 struct symlink_args { 1690 char *path; 1691 char *link; 1692 }; 1693 #endif 1694 int 1695 sys_symlink(struct thread *td, struct symlink_args *uap) 1696 { 1697 1698 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link, 1699 UIO_USERSPACE)); 1700 } 1701 1702 #ifndef _SYS_SYSPROTO_H_ 1703 struct symlinkat_args { 1704 char *path; 1705 int fd; 1706 char *path2; 1707 }; 1708 #endif 1709 int 1710 sys_symlinkat(struct thread *td, struct symlinkat_args *uap) 1711 { 1712 1713 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2, 1714 UIO_USERSPACE)); 1715 } 1716 1717 int 1718 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2, 1719 enum uio_seg segflg) 1720 { 1721 struct mount *mp; 1722 struct vattr vattr; 1723 const char *syspath; 1724 char *tmppath; 1725 struct nameidata nd; 1726 int error; 1727 1728 if (segflg == UIO_SYSSPACE) { 1729 syspath = path1; 1730 } else { 1731 tmppath = uma_zalloc(namei_zone, M_WAITOK); 1732 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0) 1733 goto out; 1734 syspath = tmppath; 1735 } 1736 AUDIT_ARG_TEXT(syspath); 1737 NDPREINIT(&nd); 1738 restart: 1739 bwillwrite(); 1740 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, segflg, 1741 path2, fd, &cap_symlinkat_rights); 1742 if ((error = namei(&nd)) != 0) 1743 goto out; 1744 if (nd.ni_vp) { 1745 NDFREE_PNBUF(&nd); 1746 if (nd.ni_vp == nd.ni_dvp) 1747 vrele(nd.ni_dvp); 1748 else 1749 vput(nd.ni_dvp); 1750 vrele(nd.ni_vp); 1751 nd.ni_vp = NULL; 1752 error = EEXIST; 1753 goto out; 1754 } 1755 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1756 NDFREE_PNBUF(&nd); 1757 vput(nd.ni_dvp); 1758 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 1759 goto out; 1760 goto restart; 1761 } 1762 VATTR_NULL(&vattr); 1763 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask; 1764 #ifdef MAC 1765 vattr.va_type = VLNK; 1766 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 1767 &vattr); 1768 if (error != 0) 1769 goto out2; 1770 #endif 1771 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath); 1772 #ifdef MAC 1773 out2: 1774 #endif 1775 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true); 1776 vn_finished_write(mp); 1777 NDFREE_PNBUF(&nd); 1778 if (error == ERELOOKUP) 1779 goto restart; 1780 out: 1781 if (segflg != UIO_SYSSPACE) 1782 uma_zfree(namei_zone, tmppath); 1783 return (error); 1784 } 1785 1786 /* 1787 * Delete a whiteout from the filesystem. 1788 */ 1789 #ifndef _SYS_SYSPROTO_H_ 1790 struct undelete_args { 1791 char *path; 1792 }; 1793 #endif 1794 int 1795 sys_undelete(struct thread *td, struct undelete_args *uap) 1796 { 1797 struct mount *mp; 1798 struct nameidata nd; 1799 int error; 1800 1801 NDPREINIT(&nd); 1802 restart: 1803 bwillwrite(); 1804 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1, 1805 UIO_USERSPACE, uap->path); 1806 error = namei(&nd); 1807 if (error != 0) 1808 return (error); 1809 1810 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1811 NDFREE_PNBUF(&nd); 1812 if (nd.ni_vp == nd.ni_dvp) 1813 vrele(nd.ni_dvp); 1814 else 1815 vput(nd.ni_dvp); 1816 if (nd.ni_vp) 1817 vrele(nd.ni_vp); 1818 return (EEXIST); 1819 } 1820 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1821 NDFREE_PNBUF(&nd); 1822 vput(nd.ni_dvp); 1823 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 1824 return (error); 1825 goto restart; 1826 } 1827 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE); 1828 NDFREE_PNBUF(&nd); 1829 vput(nd.ni_dvp); 1830 vn_finished_write(mp); 1831 if (error == ERELOOKUP) 1832 goto restart; 1833 return (error); 1834 } 1835 1836 /* 1837 * Delete a name from the filesystem. 1838 */ 1839 #ifndef _SYS_SYSPROTO_H_ 1840 struct unlink_args { 1841 char *path; 1842 }; 1843 #endif 1844 int 1845 sys_unlink(struct thread *td, struct unlink_args *uap) 1846 { 1847 1848 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE, 1849 0, 0)); 1850 } 1851 1852 static int 1853 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd, 1854 int flag, enum uio_seg pathseg, ino_t oldinum) 1855 { 1856 1857 if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0) 1858 return (EINVAL); 1859 1860 if ((flag & AT_REMOVEDIR) != 0) 1861 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0)); 1862 1863 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0)); 1864 } 1865 1866 #ifndef _SYS_SYSPROTO_H_ 1867 struct unlinkat_args { 1868 int fd; 1869 char *path; 1870 int flag; 1871 }; 1872 #endif 1873 int 1874 sys_unlinkat(struct thread *td, struct unlinkat_args *uap) 1875 { 1876 1877 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag, 1878 UIO_USERSPACE, 0)); 1879 } 1880 1881 #ifndef _SYS_SYSPROTO_H_ 1882 struct funlinkat_args { 1883 int dfd; 1884 const char *path; 1885 int fd; 1886 int flag; 1887 }; 1888 #endif 1889 int 1890 sys_funlinkat(struct thread *td, struct funlinkat_args *uap) 1891 { 1892 1893 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag, 1894 UIO_USERSPACE, 0)); 1895 } 1896 1897 int 1898 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd, 1899 enum uio_seg pathseg, int flag, ino_t oldinum) 1900 { 1901 struct mount *mp; 1902 struct file *fp; 1903 struct vnode *vp; 1904 struct nameidata nd; 1905 struct stat sb; 1906 int error; 1907 1908 fp = NULL; 1909 if (fd != FD_NONE) { 1910 error = getvnode_path(td, fd, &cap_no_rights, &fp); 1911 if (error != 0) 1912 return (error); 1913 } 1914 1915 NDPREINIT(&nd); 1916 restart: 1917 bwillwrite(); 1918 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 | 1919 at2cnpflags(flag, AT_RESOLVE_BENEATH), 1920 pathseg, path, dfd, &cap_unlinkat_rights); 1921 if ((error = namei(&nd)) != 0) { 1922 if (error == EINVAL) 1923 error = EPERM; 1924 goto fdout; 1925 } 1926 vp = nd.ni_vp; 1927 if (vp->v_type == VDIR && oldinum == 0) { 1928 error = EPERM; /* POSIX */ 1929 } else if (oldinum != 0 && 1930 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) && 1931 sb.st_ino != oldinum) { 1932 error = EIDRM; /* Identifier removed */ 1933 } else if (fp != NULL && fp->f_vnode != vp) { 1934 if (VN_IS_DOOMED(fp->f_vnode)) 1935 error = EBADF; 1936 else 1937 error = EDEADLK; 1938 } else { 1939 /* 1940 * The root of a mounted filesystem cannot be deleted. 1941 * 1942 * XXX: can this only be a VDIR case? 1943 */ 1944 if (vp->v_vflag & VV_ROOT) 1945 error = EBUSY; 1946 } 1947 if (error == 0) { 1948 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1949 NDFREE_PNBUF(&nd); 1950 vput(nd.ni_dvp); 1951 if (vp == nd.ni_dvp) 1952 vrele(vp); 1953 else 1954 vput(vp); 1955 if ((error = vn_start_write(NULL, &mp, 1956 V_XSLEEP | V_PCATCH)) != 0) { 1957 goto fdout; 1958 } 1959 goto restart; 1960 } 1961 #ifdef MAC 1962 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 1963 &nd.ni_cnd); 1964 if (error != 0) 1965 goto out; 1966 #endif 1967 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 1968 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd); 1969 #ifdef MAC 1970 out: 1971 #endif 1972 vn_finished_write(mp); 1973 } 1974 NDFREE_PNBUF(&nd); 1975 vput(nd.ni_dvp); 1976 if (vp == nd.ni_dvp) 1977 vrele(vp); 1978 else 1979 vput(vp); 1980 if (error == ERELOOKUP) 1981 goto restart; 1982 fdout: 1983 if (fp != NULL) 1984 fdrop(fp, td); 1985 return (error); 1986 } 1987 1988 /* 1989 * Reposition read/write file offset. 1990 */ 1991 #ifndef _SYS_SYSPROTO_H_ 1992 struct lseek_args { 1993 int fd; 1994 int pad; 1995 off_t offset; 1996 int whence; 1997 }; 1998 #endif 1999 int 2000 sys_lseek(struct thread *td, struct lseek_args *uap) 2001 { 2002 2003 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 2004 } 2005 2006 int 2007 kern_lseek(struct thread *td, int fd, off_t offset, int whence) 2008 { 2009 struct file *fp; 2010 int error; 2011 2012 AUDIT_ARG_FD(fd); 2013 error = fget(td, fd, &cap_seek_rights, &fp); 2014 if (error != 0) 2015 return (error); 2016 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ? 2017 fo_seek(fp, offset, whence, td) : ESPIPE; 2018 fdrop(fp, td); 2019 return (error); 2020 } 2021 2022 #if defined(COMPAT_43) 2023 /* 2024 * Reposition read/write file offset. 2025 */ 2026 #ifndef _SYS_SYSPROTO_H_ 2027 struct olseek_args { 2028 int fd; 2029 long offset; 2030 int whence; 2031 }; 2032 #endif 2033 int 2034 olseek(struct thread *td, struct olseek_args *uap) 2035 { 2036 2037 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 2038 } 2039 #endif /* COMPAT_43 */ 2040 2041 #if defined(COMPAT_FREEBSD6) 2042 /* Version with the 'pad' argument */ 2043 int 2044 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap) 2045 { 2046 2047 return (kern_lseek(td, uap->fd, uap->offset, uap->whence)); 2048 } 2049 #endif 2050 2051 /* 2052 * Check access permissions using passed credentials. 2053 */ 2054 static int 2055 vn_access(struct vnode *vp, int user_flags, struct ucred *cred, 2056 struct thread *td) 2057 { 2058 accmode_t accmode; 2059 int error; 2060 2061 /* Flags == 0 means only check for existence. */ 2062 if (user_flags == 0) 2063 return (0); 2064 2065 accmode = 0; 2066 if (user_flags & R_OK) 2067 accmode |= VREAD; 2068 if (user_flags & W_OK) 2069 accmode |= VWRITE; 2070 if (user_flags & X_OK) 2071 accmode |= VEXEC; 2072 #ifdef MAC 2073 error = mac_vnode_check_access(cred, vp, accmode); 2074 if (error != 0) 2075 return (error); 2076 #endif 2077 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 2078 error = VOP_ACCESS(vp, accmode, cred, td); 2079 return (error); 2080 } 2081 2082 /* 2083 * Check access permissions using "real" credentials. 2084 */ 2085 #ifndef _SYS_SYSPROTO_H_ 2086 struct access_args { 2087 char *path; 2088 int amode; 2089 }; 2090 #endif 2091 int 2092 sys_access(struct thread *td, struct access_args *uap) 2093 { 2094 2095 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2096 0, uap->amode)); 2097 } 2098 2099 #ifndef _SYS_SYSPROTO_H_ 2100 struct faccessat_args { 2101 int dirfd; 2102 char *path; 2103 int amode; 2104 int flag; 2105 } 2106 #endif 2107 int 2108 sys_faccessat(struct thread *td, struct faccessat_args *uap) 2109 { 2110 2111 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag, 2112 uap->amode)); 2113 } 2114 2115 int 2116 kern_accessat(struct thread *td, int fd, const char *path, 2117 enum uio_seg pathseg, int flag, int amode) 2118 { 2119 struct ucred *cred, *usecred; 2120 struct vnode *vp; 2121 struct nameidata nd; 2122 int error; 2123 2124 if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0) 2125 return (EINVAL); 2126 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0) 2127 return (EINVAL); 2128 2129 /* 2130 * Create and modify a temporary credential instead of one that 2131 * is potentially shared (if we need one). 2132 */ 2133 cred = td->td_ucred; 2134 if ((flag & AT_EACCESS) == 0 && 2135 ((cred->cr_uid != cred->cr_ruid || 2136 cred->cr_rgid != cred->cr_groups[0]))) { 2137 usecred = crdup(cred); 2138 usecred->cr_uid = cred->cr_ruid; 2139 usecred->cr_groups[0] = cred->cr_rgid; 2140 td->td_ucred = usecred; 2141 } else 2142 usecred = cred; 2143 AUDIT_ARG_VALUE(amode); 2144 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | 2145 AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH | 2146 AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights); 2147 if ((error = namei(&nd)) != 0) 2148 goto out; 2149 vp = nd.ni_vp; 2150 2151 error = vn_access(vp, amode, usecred, td); 2152 NDFREE_PNBUF(&nd); 2153 vput(vp); 2154 out: 2155 if (usecred != cred) { 2156 td->td_ucred = cred; 2157 crfree(usecred); 2158 } 2159 return (error); 2160 } 2161 2162 /* 2163 * Check access permissions using "effective" credentials. 2164 */ 2165 #ifndef _SYS_SYSPROTO_H_ 2166 struct eaccess_args { 2167 char *path; 2168 int amode; 2169 }; 2170 #endif 2171 int 2172 sys_eaccess(struct thread *td, struct eaccess_args *uap) 2173 { 2174 2175 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2176 AT_EACCESS, uap->amode)); 2177 } 2178 2179 #if defined(COMPAT_43) 2180 /* 2181 * Get file status; this version follows links. 2182 */ 2183 #ifndef _SYS_SYSPROTO_H_ 2184 struct ostat_args { 2185 char *path; 2186 struct ostat *ub; 2187 }; 2188 #endif 2189 int 2190 ostat(struct thread *td, struct ostat_args *uap) 2191 { 2192 struct stat sb; 2193 struct ostat osb; 2194 int error; 2195 2196 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb); 2197 if (error != 0) 2198 return (error); 2199 cvtstat(&sb, &osb); 2200 return (copyout(&osb, uap->ub, sizeof (osb))); 2201 } 2202 2203 /* 2204 * Get file status; this version does not follow links. 2205 */ 2206 #ifndef _SYS_SYSPROTO_H_ 2207 struct olstat_args { 2208 char *path; 2209 struct ostat *ub; 2210 }; 2211 #endif 2212 int 2213 olstat(struct thread *td, struct olstat_args *uap) 2214 { 2215 struct stat sb; 2216 struct ostat osb; 2217 int error; 2218 2219 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2220 UIO_USERSPACE, &sb); 2221 if (error != 0) 2222 return (error); 2223 cvtstat(&sb, &osb); 2224 return (copyout(&osb, uap->ub, sizeof (osb))); 2225 } 2226 2227 /* 2228 * Convert from an old to a new stat structure. 2229 * XXX: many values are blindly truncated. 2230 */ 2231 void 2232 cvtstat(struct stat *st, struct ostat *ost) 2233 { 2234 2235 bzero(ost, sizeof(*ost)); 2236 ost->st_dev = st->st_dev; 2237 ost->st_ino = st->st_ino; 2238 ost->st_mode = st->st_mode; 2239 ost->st_nlink = st->st_nlink; 2240 ost->st_uid = st->st_uid; 2241 ost->st_gid = st->st_gid; 2242 ost->st_rdev = st->st_rdev; 2243 ost->st_size = MIN(st->st_size, INT32_MAX); 2244 ost->st_atim = st->st_atim; 2245 ost->st_mtim = st->st_mtim; 2246 ost->st_ctim = st->st_ctim; 2247 ost->st_blksize = st->st_blksize; 2248 ost->st_blocks = st->st_blocks; 2249 ost->st_flags = st->st_flags; 2250 ost->st_gen = st->st_gen; 2251 } 2252 #endif /* COMPAT_43 */ 2253 2254 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 2255 int ino64_trunc_error; 2256 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW, 2257 &ino64_trunc_error, 0, 2258 "Error on truncation of device, file or inode number, or link count"); 2259 2260 int 2261 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost) 2262 { 2263 2264 ost->st_dev = st->st_dev; 2265 if (ost->st_dev != st->st_dev) { 2266 switch (ino64_trunc_error) { 2267 default: 2268 /* 2269 * Since dev_t is almost raw, don't clamp to the 2270 * maximum for case 2, but ignore the error. 2271 */ 2272 break; 2273 case 1: 2274 return (EOVERFLOW); 2275 } 2276 } 2277 ost->st_ino = st->st_ino; 2278 if (ost->st_ino != st->st_ino) { 2279 switch (ino64_trunc_error) { 2280 default: 2281 case 0: 2282 break; 2283 case 1: 2284 return (EOVERFLOW); 2285 case 2: 2286 ost->st_ino = UINT32_MAX; 2287 break; 2288 } 2289 } 2290 ost->st_mode = st->st_mode; 2291 ost->st_nlink = st->st_nlink; 2292 if (ost->st_nlink != st->st_nlink) { 2293 switch (ino64_trunc_error) { 2294 default: 2295 case 0: 2296 break; 2297 case 1: 2298 return (EOVERFLOW); 2299 case 2: 2300 ost->st_nlink = UINT16_MAX; 2301 break; 2302 } 2303 } 2304 ost->st_uid = st->st_uid; 2305 ost->st_gid = st->st_gid; 2306 ost->st_rdev = st->st_rdev; 2307 if (ost->st_rdev != st->st_rdev) { 2308 switch (ino64_trunc_error) { 2309 default: 2310 break; 2311 case 1: 2312 return (EOVERFLOW); 2313 } 2314 } 2315 ost->st_atim = st->st_atim; 2316 ost->st_mtim = st->st_mtim; 2317 ost->st_ctim = st->st_ctim; 2318 ost->st_size = st->st_size; 2319 ost->st_blocks = st->st_blocks; 2320 ost->st_blksize = st->st_blksize; 2321 ost->st_flags = st->st_flags; 2322 ost->st_gen = st->st_gen; 2323 ost->st_lspare = 0; 2324 ost->st_birthtim = st->st_birthtim; 2325 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim), 2326 sizeof(*ost) - offsetof(struct freebsd11_stat, 2327 st_birthtim) - sizeof(ost->st_birthtim)); 2328 return (0); 2329 } 2330 2331 int 2332 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap) 2333 { 2334 struct stat sb; 2335 struct freebsd11_stat osb; 2336 int error; 2337 2338 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb); 2339 if (error != 0) 2340 return (error); 2341 error = freebsd11_cvtstat(&sb, &osb); 2342 if (error == 0) 2343 error = copyout(&osb, uap->ub, sizeof(osb)); 2344 return (error); 2345 } 2346 2347 int 2348 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap) 2349 { 2350 struct stat sb; 2351 struct freebsd11_stat osb; 2352 int error; 2353 2354 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2355 UIO_USERSPACE, &sb); 2356 if (error != 0) 2357 return (error); 2358 error = freebsd11_cvtstat(&sb, &osb); 2359 if (error == 0) 2360 error = copyout(&osb, uap->ub, sizeof(osb)); 2361 return (error); 2362 } 2363 2364 int 2365 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap) 2366 { 2367 struct fhandle fh; 2368 struct stat sb; 2369 struct freebsd11_stat osb; 2370 int error; 2371 2372 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 2373 if (error != 0) 2374 return (error); 2375 error = kern_fhstat(td, fh, &sb); 2376 if (error != 0) 2377 return (error); 2378 error = freebsd11_cvtstat(&sb, &osb); 2379 if (error == 0) 2380 error = copyout(&osb, uap->sb, sizeof(osb)); 2381 return (error); 2382 } 2383 2384 int 2385 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap) 2386 { 2387 struct stat sb; 2388 struct freebsd11_stat osb; 2389 int error; 2390 2391 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2392 UIO_USERSPACE, &sb); 2393 if (error != 0) 2394 return (error); 2395 error = freebsd11_cvtstat(&sb, &osb); 2396 if (error == 0) 2397 error = copyout(&osb, uap->buf, sizeof(osb)); 2398 return (error); 2399 } 2400 #endif /* COMPAT_FREEBSD11 */ 2401 2402 /* 2403 * Get file status 2404 */ 2405 #ifndef _SYS_SYSPROTO_H_ 2406 struct fstatat_args { 2407 int fd; 2408 char *path; 2409 struct stat *buf; 2410 int flag; 2411 } 2412 #endif 2413 int 2414 sys_fstatat(struct thread *td, struct fstatat_args *uap) 2415 { 2416 struct stat sb; 2417 int error; 2418 2419 error = kern_statat(td, uap->flag, uap->fd, uap->path, 2420 UIO_USERSPACE, &sb); 2421 if (error == 0) 2422 error = copyout(&sb, uap->buf, sizeof (sb)); 2423 return (error); 2424 } 2425 2426 int 2427 kern_statat(struct thread *td, int flag, int fd, const char *path, 2428 enum uio_seg pathseg, struct stat *sbp) 2429 { 2430 struct nameidata nd; 2431 int error; 2432 2433 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 2434 AT_EMPTY_PATH)) != 0) 2435 return (EINVAL); 2436 2437 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH | 2438 AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF | 2439 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights); 2440 2441 if ((error = namei(&nd)) != 0) 2442 return (error); 2443 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED); 2444 NDFREE_PNBUF(&nd); 2445 vput(nd.ni_vp); 2446 #ifdef __STAT_TIME_T_EXT 2447 sbp->st_atim_ext = 0; 2448 sbp->st_mtim_ext = 0; 2449 sbp->st_ctim_ext = 0; 2450 sbp->st_btim_ext = 0; 2451 #endif 2452 #ifdef KTRACE 2453 if (KTRPOINT(td, KTR_STRUCT)) 2454 ktrstat_error(sbp, error); 2455 #endif 2456 return (error); 2457 } 2458 2459 #if defined(COMPAT_FREEBSD11) 2460 /* 2461 * Implementation of the NetBSD [l]stat() functions. 2462 */ 2463 int 2464 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb) 2465 { 2466 struct freebsd11_stat sb11; 2467 int error; 2468 2469 error = freebsd11_cvtstat(sb, &sb11); 2470 if (error != 0) 2471 return (error); 2472 2473 bzero(nsb, sizeof(*nsb)); 2474 CP(sb11, *nsb, st_dev); 2475 CP(sb11, *nsb, st_ino); 2476 CP(sb11, *nsb, st_mode); 2477 CP(sb11, *nsb, st_nlink); 2478 CP(sb11, *nsb, st_uid); 2479 CP(sb11, *nsb, st_gid); 2480 CP(sb11, *nsb, st_rdev); 2481 CP(sb11, *nsb, st_atim); 2482 CP(sb11, *nsb, st_mtim); 2483 CP(sb11, *nsb, st_ctim); 2484 CP(sb11, *nsb, st_size); 2485 CP(sb11, *nsb, st_blocks); 2486 CP(sb11, *nsb, st_blksize); 2487 CP(sb11, *nsb, st_flags); 2488 CP(sb11, *nsb, st_gen); 2489 CP(sb11, *nsb, st_birthtim); 2490 return (0); 2491 } 2492 2493 #ifndef _SYS_SYSPROTO_H_ 2494 struct freebsd11_nstat_args { 2495 char *path; 2496 struct nstat *ub; 2497 }; 2498 #endif 2499 int 2500 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap) 2501 { 2502 struct stat sb; 2503 struct nstat nsb; 2504 int error; 2505 2506 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb); 2507 if (error != 0) 2508 return (error); 2509 error = freebsd11_cvtnstat(&sb, &nsb); 2510 if (error == 0) 2511 error = copyout(&nsb, uap->ub, sizeof (nsb)); 2512 return (error); 2513 } 2514 2515 /* 2516 * NetBSD lstat. Get file status; this version does not follow links. 2517 */ 2518 #ifndef _SYS_SYSPROTO_H_ 2519 struct freebsd11_nlstat_args { 2520 char *path; 2521 struct nstat *ub; 2522 }; 2523 #endif 2524 int 2525 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap) 2526 { 2527 struct stat sb; 2528 struct nstat nsb; 2529 int error; 2530 2531 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path, 2532 UIO_USERSPACE, &sb); 2533 if (error != 0) 2534 return (error); 2535 error = freebsd11_cvtnstat(&sb, &nsb); 2536 if (error == 0) 2537 error = copyout(&nsb, uap->ub, sizeof (nsb)); 2538 return (error); 2539 } 2540 #endif /* COMPAT_FREEBSD11 */ 2541 2542 /* 2543 * Get configurable pathname variables. 2544 */ 2545 #ifndef _SYS_SYSPROTO_H_ 2546 struct pathconf_args { 2547 char *path; 2548 int name; 2549 }; 2550 #endif 2551 int 2552 sys_pathconf(struct thread *td, struct pathconf_args *uap) 2553 { 2554 long value; 2555 int error; 2556 2557 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW, 2558 &value); 2559 if (error == 0) 2560 td->td_retval[0] = value; 2561 return (error); 2562 } 2563 2564 #ifndef _SYS_SYSPROTO_H_ 2565 struct lpathconf_args { 2566 char *path; 2567 int name; 2568 }; 2569 #endif 2570 int 2571 sys_lpathconf(struct thread *td, struct lpathconf_args *uap) 2572 { 2573 long value; 2574 int error; 2575 2576 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, 2577 NOFOLLOW, &value); 2578 if (error == 0) 2579 td->td_retval[0] = value; 2580 return (error); 2581 } 2582 2583 int 2584 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg, 2585 int name, u_long flags, long *valuep) 2586 { 2587 struct nameidata nd; 2588 int error; 2589 2590 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags, 2591 pathseg, path); 2592 if ((error = namei(&nd)) != 0) 2593 return (error); 2594 NDFREE_PNBUF(&nd); 2595 2596 error = VOP_PATHCONF(nd.ni_vp, name, valuep); 2597 vput(nd.ni_vp); 2598 return (error); 2599 } 2600 2601 /* 2602 * Return target name of a symbolic link. 2603 */ 2604 #ifndef _SYS_SYSPROTO_H_ 2605 struct readlink_args { 2606 char *path; 2607 char *buf; 2608 size_t count; 2609 }; 2610 #endif 2611 int 2612 sys_readlink(struct thread *td, struct readlink_args *uap) 2613 { 2614 2615 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2616 uap->buf, UIO_USERSPACE, uap->count)); 2617 } 2618 #ifndef _SYS_SYSPROTO_H_ 2619 struct readlinkat_args { 2620 int fd; 2621 char *path; 2622 char *buf; 2623 size_t bufsize; 2624 }; 2625 #endif 2626 int 2627 sys_readlinkat(struct thread *td, struct readlinkat_args *uap) 2628 { 2629 2630 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE, 2631 uap->buf, UIO_USERSPACE, uap->bufsize)); 2632 } 2633 2634 int 2635 kern_readlinkat(struct thread *td, int fd, const char *path, 2636 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count) 2637 { 2638 struct vnode *vp; 2639 struct nameidata nd; 2640 int error; 2641 2642 if (count > IOSIZE_MAX) 2643 return (EINVAL); 2644 2645 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | 2646 EMPTYPATH, pathseg, path, fd); 2647 2648 if ((error = namei(&nd)) != 0) 2649 return (error); 2650 NDFREE_PNBUF(&nd); 2651 vp = nd.ni_vp; 2652 2653 error = kern_readlink_vp(vp, buf, bufseg, count, td); 2654 vput(vp); 2655 2656 return (error); 2657 } 2658 2659 /* 2660 * Helper function to readlink from a vnode 2661 */ 2662 static int 2663 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count, 2664 struct thread *td) 2665 { 2666 struct iovec aiov; 2667 struct uio auio; 2668 int error; 2669 2670 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked"); 2671 #ifdef MAC 2672 error = mac_vnode_check_readlink(td->td_ucred, vp); 2673 if (error != 0) 2674 return (error); 2675 #endif 2676 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0) 2677 return (EINVAL); 2678 2679 aiov.iov_base = buf; 2680 aiov.iov_len = count; 2681 auio.uio_iov = &aiov; 2682 auio.uio_iovcnt = 1; 2683 auio.uio_offset = 0; 2684 auio.uio_rw = UIO_READ; 2685 auio.uio_segflg = bufseg; 2686 auio.uio_td = td; 2687 auio.uio_resid = count; 2688 error = VOP_READLINK(vp, &auio, td->td_ucred); 2689 td->td_retval[0] = count - auio.uio_resid; 2690 return (error); 2691 } 2692 2693 /* 2694 * Common implementation code for chflags() and fchflags(). 2695 */ 2696 static int 2697 setfflags(struct thread *td, struct vnode *vp, u_long flags) 2698 { 2699 struct mount *mp; 2700 struct vattr vattr; 2701 int error; 2702 2703 /* We can't support the value matching VNOVAL. */ 2704 if (flags == VNOVAL) 2705 return (EOPNOTSUPP); 2706 2707 /* 2708 * Prevent non-root users from setting flags on devices. When 2709 * a device is reused, users can retain ownership of the device 2710 * if they are allowed to set flags and programs assume that 2711 * chown can't fail when done as root. 2712 */ 2713 if (vp->v_type == VCHR || vp->v_type == VBLK) { 2714 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV); 2715 if (error != 0) 2716 return (error); 2717 } 2718 2719 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2720 return (error); 2721 VATTR_NULL(&vattr); 2722 vattr.va_flags = flags; 2723 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2724 #ifdef MAC 2725 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags); 2726 if (error == 0) 2727 #endif 2728 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 2729 VOP_UNLOCK(vp); 2730 vn_finished_write(mp); 2731 return (error); 2732 } 2733 2734 /* 2735 * Change flags of a file given a path name. 2736 */ 2737 #ifndef _SYS_SYSPROTO_H_ 2738 struct chflags_args { 2739 const char *path; 2740 u_long flags; 2741 }; 2742 #endif 2743 int 2744 sys_chflags(struct thread *td, struct chflags_args *uap) 2745 { 2746 2747 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2748 uap->flags, 0)); 2749 } 2750 2751 #ifndef _SYS_SYSPROTO_H_ 2752 struct chflagsat_args { 2753 int fd; 2754 const char *path; 2755 u_long flags; 2756 int atflag; 2757 } 2758 #endif 2759 int 2760 sys_chflagsat(struct thread *td, struct chflagsat_args *uap) 2761 { 2762 2763 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE, 2764 uap->flags, uap->atflag)); 2765 } 2766 2767 /* 2768 * Same as chflags() but doesn't follow symlinks. 2769 */ 2770 #ifndef _SYS_SYSPROTO_H_ 2771 struct lchflags_args { 2772 const char *path; 2773 u_long flags; 2774 }; 2775 #endif 2776 int 2777 sys_lchflags(struct thread *td, struct lchflags_args *uap) 2778 { 2779 2780 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2781 uap->flags, AT_SYMLINK_NOFOLLOW)); 2782 } 2783 2784 static int 2785 kern_chflagsat(struct thread *td, int fd, const char *path, 2786 enum uio_seg pathseg, u_long flags, int atflag) 2787 { 2788 struct nameidata nd; 2789 int error; 2790 2791 if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 2792 AT_EMPTY_PATH)) != 0) 2793 return (EINVAL); 2794 2795 AUDIT_ARG_FFLAGS(flags); 2796 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW | 2797 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path, 2798 fd, &cap_fchflags_rights); 2799 if ((error = namei(&nd)) != 0) 2800 return (error); 2801 NDFREE_PNBUF(&nd); 2802 error = setfflags(td, nd.ni_vp, flags); 2803 vrele(nd.ni_vp); 2804 return (error); 2805 } 2806 2807 /* 2808 * Change flags of a file given a file descriptor. 2809 */ 2810 #ifndef _SYS_SYSPROTO_H_ 2811 struct fchflags_args { 2812 int fd; 2813 u_long flags; 2814 }; 2815 #endif 2816 int 2817 sys_fchflags(struct thread *td, struct fchflags_args *uap) 2818 { 2819 struct file *fp; 2820 int error; 2821 2822 AUDIT_ARG_FD(uap->fd); 2823 AUDIT_ARG_FFLAGS(uap->flags); 2824 error = getvnode(td, uap->fd, &cap_fchflags_rights, 2825 &fp); 2826 if (error != 0) 2827 return (error); 2828 #ifdef AUDIT 2829 if (AUDITING_TD(td)) { 2830 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 2831 AUDIT_ARG_VNODE1(fp->f_vnode); 2832 VOP_UNLOCK(fp->f_vnode); 2833 } 2834 #endif 2835 error = setfflags(td, fp->f_vnode, uap->flags); 2836 fdrop(fp, td); 2837 return (error); 2838 } 2839 2840 /* 2841 * Common implementation code for chmod(), lchmod() and fchmod(). 2842 */ 2843 int 2844 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode) 2845 { 2846 struct mount *mp; 2847 struct vattr vattr; 2848 int error; 2849 2850 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2851 return (error); 2852 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2853 VATTR_NULL(&vattr); 2854 vattr.va_mode = mode & ALLPERMS; 2855 #ifdef MAC 2856 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode); 2857 if (error == 0) 2858 #endif 2859 error = VOP_SETATTR(vp, &vattr, cred); 2860 VOP_UNLOCK(vp); 2861 vn_finished_write(mp); 2862 return (error); 2863 } 2864 2865 /* 2866 * Change mode of a file given path name. 2867 */ 2868 #ifndef _SYS_SYSPROTO_H_ 2869 struct chmod_args { 2870 char *path; 2871 int mode; 2872 }; 2873 #endif 2874 int 2875 sys_chmod(struct thread *td, struct chmod_args *uap) 2876 { 2877 2878 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2879 uap->mode, 0)); 2880 } 2881 2882 #ifndef _SYS_SYSPROTO_H_ 2883 struct fchmodat_args { 2884 int dirfd; 2885 char *path; 2886 mode_t mode; 2887 int flag; 2888 } 2889 #endif 2890 int 2891 sys_fchmodat(struct thread *td, struct fchmodat_args *uap) 2892 { 2893 2894 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE, 2895 uap->mode, uap->flag)); 2896 } 2897 2898 /* 2899 * Change mode of a file given path name (don't follow links.) 2900 */ 2901 #ifndef _SYS_SYSPROTO_H_ 2902 struct lchmod_args { 2903 char *path; 2904 int mode; 2905 }; 2906 #endif 2907 int 2908 sys_lchmod(struct thread *td, struct lchmod_args *uap) 2909 { 2910 2911 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 2912 uap->mode, AT_SYMLINK_NOFOLLOW)); 2913 } 2914 2915 int 2916 kern_fchmodat(struct thread *td, int fd, const char *path, 2917 enum uio_seg pathseg, mode_t mode, int flag) 2918 { 2919 struct nameidata nd; 2920 int error; 2921 2922 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 2923 AT_EMPTY_PATH)) != 0) 2924 return (EINVAL); 2925 2926 AUDIT_ARG_MODE(mode); 2927 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW | 2928 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path, 2929 fd, &cap_fchmod_rights); 2930 if ((error = namei(&nd)) != 0) 2931 return (error); 2932 NDFREE_PNBUF(&nd); 2933 error = setfmode(td, td->td_ucred, nd.ni_vp, mode); 2934 vrele(nd.ni_vp); 2935 return (error); 2936 } 2937 2938 /* 2939 * Change mode of a file given a file descriptor. 2940 */ 2941 #ifndef _SYS_SYSPROTO_H_ 2942 struct fchmod_args { 2943 int fd; 2944 int mode; 2945 }; 2946 #endif 2947 int 2948 sys_fchmod(struct thread *td, struct fchmod_args *uap) 2949 { 2950 struct file *fp; 2951 int error; 2952 2953 AUDIT_ARG_FD(uap->fd); 2954 AUDIT_ARG_MODE(uap->mode); 2955 2956 error = fget(td, uap->fd, &cap_fchmod_rights, &fp); 2957 if (error != 0) 2958 return (error); 2959 error = fo_chmod(fp, uap->mode, td->td_ucred, td); 2960 fdrop(fp, td); 2961 return (error); 2962 } 2963 2964 /* 2965 * Common implementation for chown(), lchown(), and fchown() 2966 */ 2967 int 2968 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid, 2969 gid_t gid) 2970 { 2971 struct mount *mp; 2972 struct vattr vattr; 2973 int error; 2974 2975 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2976 return (error); 2977 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2978 VATTR_NULL(&vattr); 2979 vattr.va_uid = uid; 2980 vattr.va_gid = gid; 2981 #ifdef MAC 2982 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid, 2983 vattr.va_gid); 2984 if (error == 0) 2985 #endif 2986 error = VOP_SETATTR(vp, &vattr, cred); 2987 VOP_UNLOCK(vp); 2988 vn_finished_write(mp); 2989 return (error); 2990 } 2991 2992 /* 2993 * Set ownership given a path name. 2994 */ 2995 #ifndef _SYS_SYSPROTO_H_ 2996 struct chown_args { 2997 char *path; 2998 int uid; 2999 int gid; 3000 }; 3001 #endif 3002 int 3003 sys_chown(struct thread *td, struct chown_args *uap) 3004 { 3005 3006 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid, 3007 uap->gid, 0)); 3008 } 3009 3010 #ifndef _SYS_SYSPROTO_H_ 3011 struct fchownat_args { 3012 int fd; 3013 const char * path; 3014 uid_t uid; 3015 gid_t gid; 3016 int flag; 3017 }; 3018 #endif 3019 int 3020 sys_fchownat(struct thread *td, struct fchownat_args *uap) 3021 { 3022 3023 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid, 3024 uap->gid, uap->flag)); 3025 } 3026 3027 int 3028 kern_fchownat(struct thread *td, int fd, const char *path, 3029 enum uio_seg pathseg, int uid, int gid, int flag) 3030 { 3031 struct nameidata nd; 3032 int error; 3033 3034 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 3035 AT_EMPTY_PATH)) != 0) 3036 return (EINVAL); 3037 3038 AUDIT_ARG_OWNER(uid, gid); 3039 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW | 3040 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path, 3041 fd, &cap_fchown_rights); 3042 3043 if ((error = namei(&nd)) != 0) 3044 return (error); 3045 NDFREE_PNBUF(&nd); 3046 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid); 3047 vrele(nd.ni_vp); 3048 return (error); 3049 } 3050 3051 /* 3052 * Set ownership given a path name, do not cross symlinks. 3053 */ 3054 #ifndef _SYS_SYSPROTO_H_ 3055 struct lchown_args { 3056 char *path; 3057 int uid; 3058 int gid; 3059 }; 3060 #endif 3061 int 3062 sys_lchown(struct thread *td, struct lchown_args *uap) 3063 { 3064 3065 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3066 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW)); 3067 } 3068 3069 /* 3070 * Set ownership given a file descriptor. 3071 */ 3072 #ifndef _SYS_SYSPROTO_H_ 3073 struct fchown_args { 3074 int fd; 3075 int uid; 3076 int gid; 3077 }; 3078 #endif 3079 int 3080 sys_fchown(struct thread *td, struct fchown_args *uap) 3081 { 3082 struct file *fp; 3083 int error; 3084 3085 AUDIT_ARG_FD(uap->fd); 3086 AUDIT_ARG_OWNER(uap->uid, uap->gid); 3087 error = fget(td, uap->fd, &cap_fchown_rights, &fp); 3088 if (error != 0) 3089 return (error); 3090 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td); 3091 fdrop(fp, td); 3092 return (error); 3093 } 3094 3095 /* 3096 * Common implementation code for utimes(), lutimes(), and futimes(). 3097 */ 3098 static int 3099 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg, 3100 struct timespec *tsp) 3101 { 3102 struct timeval tv[2]; 3103 const struct timeval *tvp; 3104 int error; 3105 3106 if (usrtvp == NULL) { 3107 vfs_timestamp(&tsp[0]); 3108 tsp[1] = tsp[0]; 3109 } else { 3110 if (tvpseg == UIO_SYSSPACE) { 3111 tvp = usrtvp; 3112 } else { 3113 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0) 3114 return (error); 3115 tvp = tv; 3116 } 3117 3118 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 || 3119 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000) 3120 return (EINVAL); 3121 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]); 3122 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]); 3123 } 3124 return (0); 3125 } 3126 3127 /* 3128 * Common implementation code for futimens(), utimensat(). 3129 */ 3130 #define UTIMENS_NULL 0x1 3131 #define UTIMENS_EXIT 0x2 3132 static int 3133 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg, 3134 struct timespec *tsp, int *retflags) 3135 { 3136 struct timespec tsnow; 3137 int error; 3138 3139 vfs_timestamp(&tsnow); 3140 *retflags = 0; 3141 if (usrtsp == NULL) { 3142 tsp[0] = tsnow; 3143 tsp[1] = tsnow; 3144 *retflags |= UTIMENS_NULL; 3145 return (0); 3146 } 3147 if (tspseg == UIO_SYSSPACE) { 3148 tsp[0] = usrtsp[0]; 3149 tsp[1] = usrtsp[1]; 3150 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0) 3151 return (error); 3152 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT) 3153 *retflags |= UTIMENS_EXIT; 3154 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW) 3155 *retflags |= UTIMENS_NULL; 3156 if (tsp[0].tv_nsec == UTIME_OMIT) 3157 tsp[0].tv_sec = VNOVAL; 3158 else if (tsp[0].tv_nsec == UTIME_NOW) 3159 tsp[0] = tsnow; 3160 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L) 3161 return (EINVAL); 3162 if (tsp[1].tv_nsec == UTIME_OMIT) 3163 tsp[1].tv_sec = VNOVAL; 3164 else if (tsp[1].tv_nsec == UTIME_NOW) 3165 tsp[1] = tsnow; 3166 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L) 3167 return (EINVAL); 3168 3169 return (0); 3170 } 3171 3172 /* 3173 * Common implementation code for utimes(), lutimes(), futimes(), futimens(), 3174 * and utimensat(). 3175 */ 3176 static int 3177 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts, 3178 int numtimes, int nullflag) 3179 { 3180 struct mount *mp; 3181 struct vattr vattr; 3182 int error; 3183 bool setbirthtime; 3184 3185 setbirthtime = false; 3186 vattr.va_birthtime.tv_sec = VNOVAL; 3187 vattr.va_birthtime.tv_nsec = 0; 3188 3189 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3190 return (error); 3191 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3192 if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 && 3193 timespeccmp(&ts[1], &vattr.va_birthtime, < )) 3194 setbirthtime = true; 3195 VATTR_NULL(&vattr); 3196 vattr.va_atime = ts[0]; 3197 vattr.va_mtime = ts[1]; 3198 if (setbirthtime) 3199 vattr.va_birthtime = ts[1]; 3200 if (numtimes > 2) 3201 vattr.va_birthtime = ts[2]; 3202 if (nullflag) 3203 vattr.va_vaflags |= VA_UTIMES_NULL; 3204 #ifdef MAC 3205 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime, 3206 vattr.va_mtime); 3207 #endif 3208 if (error == 0) 3209 error = VOP_SETATTR(vp, &vattr, td->td_ucred); 3210 VOP_UNLOCK(vp); 3211 vn_finished_write(mp); 3212 return (error); 3213 } 3214 3215 /* 3216 * Set the access and modification times of a file. 3217 */ 3218 #ifndef _SYS_SYSPROTO_H_ 3219 struct utimes_args { 3220 char *path; 3221 struct timeval *tptr; 3222 }; 3223 #endif 3224 int 3225 sys_utimes(struct thread *td, struct utimes_args *uap) 3226 { 3227 3228 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3229 uap->tptr, UIO_USERSPACE)); 3230 } 3231 3232 #ifndef _SYS_SYSPROTO_H_ 3233 struct futimesat_args { 3234 int fd; 3235 const char * path; 3236 const struct timeval * times; 3237 }; 3238 #endif 3239 int 3240 sys_futimesat(struct thread *td, struct futimesat_args *uap) 3241 { 3242 3243 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE, 3244 uap->times, UIO_USERSPACE)); 3245 } 3246 3247 int 3248 kern_utimesat(struct thread *td, int fd, const char *path, 3249 enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg) 3250 { 3251 struct nameidata nd; 3252 struct timespec ts[2]; 3253 int error; 3254 3255 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3256 return (error); 3257 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd, 3258 &cap_futimes_rights); 3259 3260 if ((error = namei(&nd)) != 0) 3261 return (error); 3262 NDFREE_PNBUF(&nd); 3263 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3264 vrele(nd.ni_vp); 3265 return (error); 3266 } 3267 3268 /* 3269 * Set the access and modification times of a file. 3270 */ 3271 #ifndef _SYS_SYSPROTO_H_ 3272 struct lutimes_args { 3273 char *path; 3274 struct timeval *tptr; 3275 }; 3276 #endif 3277 int 3278 sys_lutimes(struct thread *td, struct lutimes_args *uap) 3279 { 3280 3281 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr, 3282 UIO_USERSPACE)); 3283 } 3284 3285 int 3286 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg, 3287 const struct timeval *tptr, enum uio_seg tptrseg) 3288 { 3289 struct timespec ts[2]; 3290 struct nameidata nd; 3291 int error; 3292 3293 if ((error = getutimes(tptr, tptrseg, ts)) != 0) 3294 return (error); 3295 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path); 3296 if ((error = namei(&nd)) != 0) 3297 return (error); 3298 NDFREE_PNBUF(&nd); 3299 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL); 3300 vrele(nd.ni_vp); 3301 return (error); 3302 } 3303 3304 /* 3305 * Set the access and modification times of a file. 3306 */ 3307 #ifndef _SYS_SYSPROTO_H_ 3308 struct futimes_args { 3309 int fd; 3310 struct timeval *tptr; 3311 }; 3312 #endif 3313 int 3314 sys_futimes(struct thread *td, struct futimes_args *uap) 3315 { 3316 3317 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE)); 3318 } 3319 3320 int 3321 kern_futimes(struct thread *td, int fd, const struct timeval *tptr, 3322 enum uio_seg tptrseg) 3323 { 3324 struct timespec ts[2]; 3325 struct file *fp; 3326 int error; 3327 3328 AUDIT_ARG_FD(fd); 3329 error = getutimes(tptr, tptrseg, ts); 3330 if (error != 0) 3331 return (error); 3332 error = getvnode(td, fd, &cap_futimes_rights, &fp); 3333 if (error != 0) 3334 return (error); 3335 #ifdef AUDIT 3336 if (AUDITING_TD(td)) { 3337 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3338 AUDIT_ARG_VNODE1(fp->f_vnode); 3339 VOP_UNLOCK(fp->f_vnode); 3340 } 3341 #endif 3342 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL); 3343 fdrop(fp, td); 3344 return (error); 3345 } 3346 3347 int 3348 sys_futimens(struct thread *td, struct futimens_args *uap) 3349 { 3350 3351 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE)); 3352 } 3353 3354 int 3355 kern_futimens(struct thread *td, int fd, const struct timespec *tptr, 3356 enum uio_seg tptrseg) 3357 { 3358 struct timespec ts[2]; 3359 struct file *fp; 3360 int error, flags; 3361 3362 AUDIT_ARG_FD(fd); 3363 error = getutimens(tptr, tptrseg, ts, &flags); 3364 if (error != 0) 3365 return (error); 3366 if (flags & UTIMENS_EXIT) 3367 return (0); 3368 error = getvnode(td, fd, &cap_futimes_rights, &fp); 3369 if (error != 0) 3370 return (error); 3371 #ifdef AUDIT 3372 if (AUDITING_TD(td)) { 3373 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY); 3374 AUDIT_ARG_VNODE1(fp->f_vnode); 3375 VOP_UNLOCK(fp->f_vnode); 3376 } 3377 #endif 3378 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL); 3379 fdrop(fp, td); 3380 return (error); 3381 } 3382 3383 int 3384 sys_utimensat(struct thread *td, struct utimensat_args *uap) 3385 { 3386 3387 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE, 3388 uap->times, UIO_USERSPACE, uap->flag)); 3389 } 3390 3391 int 3392 kern_utimensat(struct thread *td, int fd, const char *path, 3393 enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg, 3394 int flag) 3395 { 3396 struct nameidata nd; 3397 struct timespec ts[2]; 3398 int error, flags; 3399 3400 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH | 3401 AT_EMPTY_PATH)) != 0) 3402 return (EINVAL); 3403 3404 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0) 3405 return (error); 3406 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW | 3407 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, 3408 pathseg, path, fd, &cap_futimes_rights); 3409 if ((error = namei(&nd)) != 0) 3410 return (error); 3411 /* 3412 * We are allowed to call namei() regardless of 2xUTIME_OMIT. 3413 * POSIX states: 3414 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected." 3415 * "Search permission is denied by a component of the path prefix." 3416 */ 3417 NDFREE_PNBUF(&nd); 3418 if ((flags & UTIMENS_EXIT) == 0) 3419 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL); 3420 vrele(nd.ni_vp); 3421 return (error); 3422 } 3423 3424 /* 3425 * Truncate a file given its path name. 3426 */ 3427 #ifndef _SYS_SYSPROTO_H_ 3428 struct truncate_args { 3429 char *path; 3430 int pad; 3431 off_t length; 3432 }; 3433 #endif 3434 int 3435 sys_truncate(struct thread *td, struct truncate_args *uap) 3436 { 3437 3438 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3439 } 3440 3441 int 3442 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg, 3443 off_t length) 3444 { 3445 struct mount *mp; 3446 struct vnode *vp; 3447 void *rl_cookie; 3448 struct nameidata nd; 3449 int error; 3450 3451 if (length < 0) 3452 return (EINVAL); 3453 NDPREINIT(&nd); 3454 retry: 3455 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path); 3456 if ((error = namei(&nd)) != 0) 3457 return (error); 3458 vp = nd.ni_vp; 3459 NDFREE_PNBUF(&nd); 3460 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX); 3461 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3462 vn_rangelock_unlock(vp, rl_cookie); 3463 vrele(vp); 3464 return (error); 3465 } 3466 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3467 if (vp->v_type == VDIR) { 3468 error = EISDIR; 3469 goto out; 3470 } 3471 #ifdef MAC 3472 error = mac_vnode_check_write(td->td_ucred, NOCRED, vp); 3473 if (error != 0) 3474 goto out; 3475 #endif 3476 error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td); 3477 if (error != 0) 3478 goto out; 3479 3480 error = vn_truncate_locked(vp, length, false, td->td_ucred); 3481 out: 3482 VOP_UNLOCK(vp); 3483 vn_finished_write(mp); 3484 vn_rangelock_unlock(vp, rl_cookie); 3485 vrele(vp); 3486 if (error == ERELOOKUP) 3487 goto retry; 3488 return (error); 3489 } 3490 3491 #if defined(COMPAT_43) 3492 /* 3493 * Truncate a file given its path name. 3494 */ 3495 #ifndef _SYS_SYSPROTO_H_ 3496 struct otruncate_args { 3497 char *path; 3498 long length; 3499 }; 3500 #endif 3501 int 3502 otruncate(struct thread *td, struct otruncate_args *uap) 3503 { 3504 3505 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3506 } 3507 #endif /* COMPAT_43 */ 3508 3509 #if defined(COMPAT_FREEBSD6) 3510 /* Versions with the pad argument */ 3511 int 3512 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap) 3513 { 3514 3515 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length)); 3516 } 3517 3518 int 3519 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap) 3520 { 3521 3522 return (kern_ftruncate(td, uap->fd, uap->length)); 3523 } 3524 #endif 3525 3526 int 3527 kern_fsync(struct thread *td, int fd, bool fullsync) 3528 { 3529 struct vnode *vp; 3530 struct mount *mp; 3531 struct file *fp; 3532 int error; 3533 3534 AUDIT_ARG_FD(fd); 3535 error = getvnode(td, fd, &cap_fsync_rights, &fp); 3536 if (error != 0) 3537 return (error); 3538 vp = fp->f_vnode; 3539 #if 0 3540 if (!fullsync) 3541 /* XXXKIB: compete outstanding aio writes */; 3542 #endif 3543 retry: 3544 error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH); 3545 if (error != 0) 3546 goto drop; 3547 vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY); 3548 AUDIT_ARG_VNODE1(vp); 3549 if (vp->v_object != NULL) { 3550 VM_OBJECT_WLOCK(vp->v_object); 3551 vm_object_page_clean(vp->v_object, 0, 0, 0); 3552 VM_OBJECT_WUNLOCK(vp->v_object); 3553 } 3554 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td); 3555 VOP_UNLOCK(vp); 3556 vn_finished_write(mp); 3557 if (error == ERELOOKUP) 3558 goto retry; 3559 drop: 3560 fdrop(fp, td); 3561 return (error); 3562 } 3563 3564 /* 3565 * Sync an open file. 3566 */ 3567 #ifndef _SYS_SYSPROTO_H_ 3568 struct fsync_args { 3569 int fd; 3570 }; 3571 #endif 3572 int 3573 sys_fsync(struct thread *td, struct fsync_args *uap) 3574 { 3575 3576 return (kern_fsync(td, uap->fd, true)); 3577 } 3578 3579 int 3580 sys_fdatasync(struct thread *td, struct fdatasync_args *uap) 3581 { 3582 3583 return (kern_fsync(td, uap->fd, false)); 3584 } 3585 3586 /* 3587 * Rename files. Source and destination must either both be directories, or 3588 * both not be directories. If target is a directory, it must be empty. 3589 */ 3590 #ifndef _SYS_SYSPROTO_H_ 3591 struct rename_args { 3592 char *from; 3593 char *to; 3594 }; 3595 #endif 3596 int 3597 sys_rename(struct thread *td, struct rename_args *uap) 3598 { 3599 3600 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD, 3601 uap->to, UIO_USERSPACE)); 3602 } 3603 3604 #ifndef _SYS_SYSPROTO_H_ 3605 struct renameat_args { 3606 int oldfd; 3607 char *old; 3608 int newfd; 3609 char *new; 3610 }; 3611 #endif 3612 int 3613 sys_renameat(struct thread *td, struct renameat_args *uap) 3614 { 3615 3616 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new, 3617 UIO_USERSPACE)); 3618 } 3619 3620 #ifdef MAC 3621 static int 3622 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd, 3623 const char *new, enum uio_seg pathseg, struct nameidata *fromnd) 3624 { 3625 int error; 3626 3627 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1, 3628 pathseg, old, oldfd, &cap_renameat_source_rights); 3629 if ((error = namei(fromnd)) != 0) 3630 return (error); 3631 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp, 3632 fromnd->ni_vp, &fromnd->ni_cnd); 3633 VOP_UNLOCK(fromnd->ni_dvp); 3634 if (fromnd->ni_dvp != fromnd->ni_vp) 3635 VOP_UNLOCK(fromnd->ni_vp); 3636 if (error != 0) { 3637 NDFREE_PNBUF(fromnd); 3638 vrele(fromnd->ni_dvp); 3639 vrele(fromnd->ni_vp); 3640 } 3641 return (error); 3642 } 3643 #endif 3644 3645 int 3646 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd, 3647 const char *new, enum uio_seg pathseg) 3648 { 3649 struct mount *mp = NULL; 3650 struct vnode *tvp, *fvp, *tdvp; 3651 struct nameidata fromnd, tond; 3652 uint64_t tondflags; 3653 int error; 3654 3655 again: 3656 bwillwrite(); 3657 #ifdef MAC 3658 if (mac_vnode_check_rename_from_enabled()) { 3659 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg, 3660 &fromnd); 3661 if (error != 0) 3662 return (error); 3663 } else { 3664 #endif 3665 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | AUDITVNODE1, 3666 pathseg, old, oldfd, &cap_renameat_source_rights); 3667 if ((error = namei(&fromnd)) != 0) 3668 return (error); 3669 #ifdef MAC 3670 } 3671 #endif 3672 fvp = fromnd.ni_vp; 3673 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | AUDITVNODE2; 3674 if (fromnd.ni_vp->v_type == VDIR) 3675 tondflags |= WILLBEDIR; 3676 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd, 3677 &cap_renameat_target_rights); 3678 if ((error = namei(&tond)) != 0) { 3679 /* Translate error code for rename("dir1", "dir2/."). */ 3680 if (error == EISDIR && fvp->v_type == VDIR) 3681 error = EINVAL; 3682 NDFREE_PNBUF(&fromnd); 3683 vrele(fromnd.ni_dvp); 3684 vrele(fvp); 3685 goto out1; 3686 } 3687 tdvp = tond.ni_dvp; 3688 tvp = tond.ni_vp; 3689 error = vn_start_write(fvp, &mp, V_NOWAIT); 3690 if (error != 0) { 3691 NDFREE_PNBUF(&fromnd); 3692 NDFREE_PNBUF(&tond); 3693 if (tvp != NULL) 3694 vput(tvp); 3695 if (tdvp == tvp) 3696 vrele(tdvp); 3697 else 3698 vput(tdvp); 3699 vrele(fromnd.ni_dvp); 3700 vrele(fvp); 3701 error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH); 3702 if (error != 0) 3703 return (error); 3704 goto again; 3705 } 3706 if (tvp != NULL) { 3707 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3708 error = ENOTDIR; 3709 goto out; 3710 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3711 error = EISDIR; 3712 goto out; 3713 } 3714 #ifdef CAPABILITIES 3715 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) { 3716 /* 3717 * If the target already exists we require CAP_UNLINKAT 3718 * from 'newfd', when newfd was used for the lookup. 3719 */ 3720 error = cap_check(&tond.ni_filecaps.fc_rights, 3721 &cap_unlinkat_rights); 3722 if (error != 0) 3723 goto out; 3724 } 3725 #endif 3726 } 3727 if (fvp == tdvp) { 3728 error = EINVAL; 3729 goto out; 3730 } 3731 /* 3732 * If the source is the same as the destination (that is, if they 3733 * are links to the same vnode), then there is nothing to do. 3734 */ 3735 if (fvp == tvp) 3736 error = ERESTART; 3737 #ifdef MAC 3738 else 3739 error = mac_vnode_check_rename_to(td->td_ucred, tdvp, 3740 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd); 3741 #endif 3742 out: 3743 if (error == 0) { 3744 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3745 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3746 NDFREE_PNBUF(&fromnd); 3747 NDFREE_PNBUF(&tond); 3748 } else { 3749 NDFREE_PNBUF(&fromnd); 3750 NDFREE_PNBUF(&tond); 3751 if (tvp != NULL) 3752 vput(tvp); 3753 if (tdvp == tvp) 3754 vrele(tdvp); 3755 else 3756 vput(tdvp); 3757 vrele(fromnd.ni_dvp); 3758 vrele(fvp); 3759 } 3760 vn_finished_write(mp); 3761 out1: 3762 if (error == ERESTART) 3763 return (0); 3764 if (error == ERELOOKUP) 3765 goto again; 3766 return (error); 3767 } 3768 3769 /* 3770 * Make a directory file. 3771 */ 3772 #ifndef _SYS_SYSPROTO_H_ 3773 struct mkdir_args { 3774 char *path; 3775 int mode; 3776 }; 3777 #endif 3778 int 3779 sys_mkdir(struct thread *td, struct mkdir_args *uap) 3780 { 3781 3782 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 3783 uap->mode)); 3784 } 3785 3786 #ifndef _SYS_SYSPROTO_H_ 3787 struct mkdirat_args { 3788 int fd; 3789 char *path; 3790 mode_t mode; 3791 }; 3792 #endif 3793 int 3794 sys_mkdirat(struct thread *td, struct mkdirat_args *uap) 3795 { 3796 3797 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode)); 3798 } 3799 3800 int 3801 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg, 3802 int mode) 3803 { 3804 struct mount *mp; 3805 struct vattr vattr; 3806 struct nameidata nd; 3807 int error; 3808 3809 AUDIT_ARG_MODE(mode); 3810 NDPREINIT(&nd); 3811 restart: 3812 bwillwrite(); 3813 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | 3814 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR, 3815 segflg, path, fd, &cap_mkdirat_rights); 3816 if ((error = namei(&nd)) != 0) 3817 return (error); 3818 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3819 NDFREE_PNBUF(&nd); 3820 vput(nd.ni_dvp); 3821 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 3822 return (error); 3823 goto restart; 3824 } 3825 VATTR_NULL(&vattr); 3826 vattr.va_type = VDIR; 3827 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask; 3828 #ifdef MAC 3829 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, 3830 &vattr); 3831 if (error != 0) 3832 goto out; 3833 #endif 3834 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3835 #ifdef MAC 3836 out: 3837 #endif 3838 NDFREE_PNBUF(&nd); 3839 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true); 3840 vn_finished_write(mp); 3841 if (error == ERELOOKUP) 3842 goto restart; 3843 return (error); 3844 } 3845 3846 /* 3847 * Remove a directory file. 3848 */ 3849 #ifndef _SYS_SYSPROTO_H_ 3850 struct rmdir_args { 3851 char *path; 3852 }; 3853 #endif 3854 int 3855 sys_rmdir(struct thread *td, struct rmdir_args *uap) 3856 { 3857 3858 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE, 3859 0)); 3860 } 3861 3862 int 3863 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd, 3864 enum uio_seg pathseg, int flag) 3865 { 3866 struct mount *mp; 3867 struct vnode *vp; 3868 struct file *fp; 3869 struct nameidata nd; 3870 cap_rights_t rights; 3871 int error; 3872 3873 fp = NULL; 3874 if (fd != FD_NONE) { 3875 error = getvnode(td, fd, cap_rights_init_one(&rights, 3876 CAP_LOOKUP), &fp); 3877 if (error != 0) 3878 return (error); 3879 } 3880 3881 NDPREINIT(&nd); 3882 restart: 3883 bwillwrite(); 3884 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 | 3885 at2cnpflags(flag, AT_RESOLVE_BENEATH), 3886 pathseg, path, dfd, &cap_unlinkat_rights); 3887 if ((error = namei(&nd)) != 0) 3888 goto fdout; 3889 vp = nd.ni_vp; 3890 if (vp->v_type != VDIR) { 3891 error = ENOTDIR; 3892 goto out; 3893 } 3894 /* 3895 * No rmdir "." please. 3896 */ 3897 if (nd.ni_dvp == vp) { 3898 error = EINVAL; 3899 goto out; 3900 } 3901 /* 3902 * The root of a mounted filesystem cannot be deleted. 3903 */ 3904 if (vp->v_vflag & VV_ROOT) { 3905 error = EBUSY; 3906 goto out; 3907 } 3908 3909 if (fp != NULL && fp->f_vnode != vp) { 3910 if (VN_IS_DOOMED(fp->f_vnode)) 3911 error = EBADF; 3912 else 3913 error = EDEADLK; 3914 goto out; 3915 } 3916 3917 #ifdef MAC 3918 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, 3919 &nd.ni_cnd); 3920 if (error != 0) 3921 goto out; 3922 #endif 3923 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3924 NDFREE_PNBUF(&nd); 3925 vput(vp); 3926 if (nd.ni_dvp == vp) 3927 vrele(nd.ni_dvp); 3928 else 3929 vput(nd.ni_dvp); 3930 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0) 3931 goto fdout; 3932 goto restart; 3933 } 3934 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK); 3935 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3936 vn_finished_write(mp); 3937 out: 3938 NDFREE_PNBUF(&nd); 3939 vput(vp); 3940 if (nd.ni_dvp == vp) 3941 vrele(nd.ni_dvp); 3942 else 3943 vput(nd.ni_dvp); 3944 if (error == ERELOOKUP) 3945 goto restart; 3946 fdout: 3947 if (fp != NULL) 3948 fdrop(fp, td); 3949 return (error); 3950 } 3951 3952 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11) 3953 int 3954 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count, 3955 long *basep, void (*func)(struct freebsd11_dirent *)) 3956 { 3957 struct freebsd11_dirent dstdp; 3958 struct dirent *dp, *edp; 3959 char *dirbuf; 3960 off_t base; 3961 ssize_t resid, ucount; 3962 int error; 3963 3964 /* XXX arbitrary sanity limit on `count'. */ 3965 count = min(count, 64 * 1024); 3966 3967 dirbuf = malloc(count, M_TEMP, M_WAITOK); 3968 3969 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid, 3970 UIO_SYSSPACE); 3971 if (error != 0) 3972 goto done; 3973 if (basep != NULL) 3974 *basep = base; 3975 3976 ucount = 0; 3977 for (dp = (struct dirent *)dirbuf, 3978 edp = (struct dirent *)&dirbuf[count - resid]; 3979 ucount < count && dp < edp; ) { 3980 if (dp->d_reclen == 0) 3981 break; 3982 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0)); 3983 if (dp->d_namlen >= sizeof(dstdp.d_name)) 3984 continue; 3985 dstdp.d_type = dp->d_type; 3986 dstdp.d_namlen = dp->d_namlen; 3987 dstdp.d_fileno = dp->d_fileno; /* truncate */ 3988 if (dstdp.d_fileno != dp->d_fileno) { 3989 switch (ino64_trunc_error) { 3990 default: 3991 case 0: 3992 break; 3993 case 1: 3994 error = EOVERFLOW; 3995 goto done; 3996 case 2: 3997 dstdp.d_fileno = UINT32_MAX; 3998 break; 3999 } 4000 } 4001 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) + 4002 ((dp->d_namlen + 1 + 3) &~ 3); 4003 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen); 4004 bzero(dstdp.d_name + dstdp.d_namlen, 4005 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) - 4006 dstdp.d_namlen); 4007 MPASS(dstdp.d_reclen <= dp->d_reclen); 4008 MPASS(ucount + dstdp.d_reclen <= count); 4009 if (func != NULL) 4010 func(&dstdp); 4011 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen); 4012 if (error != 0) 4013 break; 4014 dp = (struct dirent *)((char *)dp + dp->d_reclen); 4015 ucount += dstdp.d_reclen; 4016 } 4017 4018 done: 4019 free(dirbuf, M_TEMP); 4020 if (error == 0) 4021 td->td_retval[0] = ucount; 4022 return (error); 4023 } 4024 #endif /* COMPAT */ 4025 4026 #ifdef COMPAT_43 4027 static void 4028 ogetdirentries_cvt(struct freebsd11_dirent *dp) 4029 { 4030 #if (BYTE_ORDER == LITTLE_ENDIAN) 4031 /* 4032 * The expected low byte of dp->d_namlen is our dp->d_type. 4033 * The high MBZ byte of dp->d_namlen is our dp->d_namlen. 4034 */ 4035 dp->d_type = dp->d_namlen; 4036 dp->d_namlen = 0; 4037 #else 4038 /* 4039 * The dp->d_type is the high byte of the expected dp->d_namlen, 4040 * so must be zero'ed. 4041 */ 4042 dp->d_type = 0; 4043 #endif 4044 } 4045 4046 /* 4047 * Read a block of directory entries in a filesystem independent format. 4048 */ 4049 #ifndef _SYS_SYSPROTO_H_ 4050 struct ogetdirentries_args { 4051 int fd; 4052 char *buf; 4053 u_int count; 4054 long *basep; 4055 }; 4056 #endif 4057 int 4058 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap) 4059 { 4060 long loff; 4061 int error; 4062 4063 error = kern_ogetdirentries(td, uap, &loff); 4064 if (error == 0) 4065 error = copyout(&loff, uap->basep, sizeof(long)); 4066 return (error); 4067 } 4068 4069 int 4070 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap, 4071 long *ploff) 4072 { 4073 long base; 4074 int error; 4075 4076 /* XXX arbitrary sanity limit on `count'. */ 4077 if (uap->count > 64 * 1024) 4078 return (EINVAL); 4079 4080 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 4081 &base, ogetdirentries_cvt); 4082 4083 if (error == 0 && uap->basep != NULL) 4084 error = copyout(&base, uap->basep, sizeof(long)); 4085 4086 return (error); 4087 } 4088 #endif /* COMPAT_43 */ 4089 4090 #if defined(COMPAT_FREEBSD11) 4091 #ifndef _SYS_SYSPROTO_H_ 4092 struct freebsd11_getdirentries_args { 4093 int fd; 4094 char *buf; 4095 u_int count; 4096 long *basep; 4097 }; 4098 #endif 4099 int 4100 freebsd11_getdirentries(struct thread *td, 4101 struct freebsd11_getdirentries_args *uap) 4102 { 4103 long base; 4104 int error; 4105 4106 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count, 4107 &base, NULL); 4108 4109 if (error == 0 && uap->basep != NULL) 4110 error = copyout(&base, uap->basep, sizeof(long)); 4111 return (error); 4112 } 4113 4114 int 4115 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap) 4116 { 4117 struct freebsd11_getdirentries_args ap; 4118 4119 ap.fd = uap->fd; 4120 ap.buf = uap->buf; 4121 ap.count = uap->count; 4122 ap.basep = NULL; 4123 return (freebsd11_getdirentries(td, &ap)); 4124 } 4125 #endif /* COMPAT_FREEBSD11 */ 4126 4127 /* 4128 * Read a block of directory entries in a filesystem independent format. 4129 */ 4130 int 4131 sys_getdirentries(struct thread *td, struct getdirentries_args *uap) 4132 { 4133 off_t base; 4134 int error; 4135 4136 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base, 4137 NULL, UIO_USERSPACE); 4138 if (error != 0) 4139 return (error); 4140 if (uap->basep != NULL) 4141 error = copyout(&base, uap->basep, sizeof(off_t)); 4142 return (error); 4143 } 4144 4145 int 4146 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count, 4147 off_t *basep, ssize_t *residp, enum uio_seg bufseg) 4148 { 4149 struct vnode *vp; 4150 struct file *fp; 4151 struct uio auio; 4152 struct iovec aiov; 4153 off_t loff; 4154 int error, eofflag; 4155 off_t foffset; 4156 4157 AUDIT_ARG_FD(fd); 4158 if (count > IOSIZE_MAX) 4159 return (EINVAL); 4160 auio.uio_resid = count; 4161 error = getvnode(td, fd, &cap_read_rights, &fp); 4162 if (error != 0) 4163 return (error); 4164 if ((fp->f_flag & FREAD) == 0) { 4165 fdrop(fp, td); 4166 return (EBADF); 4167 } 4168 vp = fp->f_vnode; 4169 foffset = foffset_lock(fp, 0); 4170 unionread: 4171 if (vp->v_type != VDIR) { 4172 error = EINVAL; 4173 goto fail; 4174 } 4175 if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) { 4176 error = ENOENT; 4177 goto fail; 4178 } 4179 aiov.iov_base = buf; 4180 aiov.iov_len = count; 4181 auio.uio_iov = &aiov; 4182 auio.uio_iovcnt = 1; 4183 auio.uio_rw = UIO_READ; 4184 auio.uio_segflg = bufseg; 4185 auio.uio_td = td; 4186 vn_lock(vp, LK_SHARED | LK_RETRY); 4187 AUDIT_ARG_VNODE1(vp); 4188 loff = auio.uio_offset = foffset; 4189 #ifdef MAC 4190 error = mac_vnode_check_readdir(td->td_ucred, vp); 4191 if (error == 0) 4192 #endif 4193 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, 4194 NULL); 4195 foffset = auio.uio_offset; 4196 if (error != 0) { 4197 VOP_UNLOCK(vp); 4198 goto fail; 4199 } 4200 if (count == auio.uio_resid && 4201 (vp->v_vflag & VV_ROOT) && 4202 (vp->v_mount->mnt_flag & MNT_UNION)) { 4203 struct vnode *tvp = vp; 4204 4205 vp = vp->v_mount->mnt_vnodecovered; 4206 VREF(vp); 4207 fp->f_vnode = vp; 4208 foffset = 0; 4209 vput(tvp); 4210 goto unionread; 4211 } 4212 VOP_UNLOCK(vp); 4213 *basep = loff; 4214 if (residp != NULL) 4215 *residp = auio.uio_resid; 4216 td->td_retval[0] = count - auio.uio_resid; 4217 fail: 4218 foffset_unlock(fp, foffset, 0); 4219 fdrop(fp, td); 4220 return (error); 4221 } 4222 4223 /* 4224 * Set the mode mask for creation of filesystem nodes. 4225 */ 4226 #ifndef _SYS_SYSPROTO_H_ 4227 struct umask_args { 4228 int newmask; 4229 }; 4230 #endif 4231 int 4232 sys_umask(struct thread *td, struct umask_args *uap) 4233 { 4234 struct pwddesc *pdp; 4235 4236 pdp = td->td_proc->p_pd; 4237 PWDDESC_XLOCK(pdp); 4238 td->td_retval[0] = pdp->pd_cmask; 4239 pdp->pd_cmask = uap->newmask & ALLPERMS; 4240 PWDDESC_XUNLOCK(pdp); 4241 return (0); 4242 } 4243 4244 /* 4245 * Void all references to file by ripping underlying filesystem away from 4246 * vnode. 4247 */ 4248 #ifndef _SYS_SYSPROTO_H_ 4249 struct revoke_args { 4250 char *path; 4251 }; 4252 #endif 4253 int 4254 sys_revoke(struct thread *td, struct revoke_args *uap) 4255 { 4256 struct vnode *vp; 4257 struct vattr vattr; 4258 struct nameidata nd; 4259 int error; 4260 4261 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE, 4262 uap->path); 4263 if ((error = namei(&nd)) != 0) 4264 return (error); 4265 vp = nd.ni_vp; 4266 NDFREE_PNBUF(&nd); 4267 if (vp->v_type != VCHR || vp->v_rdev == NULL) { 4268 error = EINVAL; 4269 goto out; 4270 } 4271 #ifdef MAC 4272 error = mac_vnode_check_revoke(td->td_ucred, vp); 4273 if (error != 0) 4274 goto out; 4275 #endif 4276 error = VOP_GETATTR(vp, &vattr, td->td_ucred); 4277 if (error != 0) 4278 goto out; 4279 if (td->td_ucred->cr_uid != vattr.va_uid) { 4280 error = priv_check(td, PRIV_VFS_ADMIN); 4281 if (error != 0) 4282 goto out; 4283 } 4284 if (devfs_usecount(vp) > 0) 4285 VOP_REVOKE(vp, REVOKEALL); 4286 out: 4287 vput(vp); 4288 return (error); 4289 } 4290 4291 /* 4292 * This variant of getvnode() allows O_PATH files. Caller should 4293 * ensure that returned file and vnode are only used for compatible 4294 * semantics. 4295 */ 4296 int 4297 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp, 4298 struct file **fpp) 4299 { 4300 struct file *fp; 4301 int error; 4302 4303 error = fget_unlocked(td, fd, rightsp, &fp); 4304 if (error != 0) 4305 return (error); 4306 4307 /* 4308 * The file could be not of the vnode type, or it may be not 4309 * yet fully initialized, in which case the f_vnode pointer 4310 * may be set, but f_ops is still badfileops. E.g., 4311 * devfs_open() transiently create such situation to 4312 * facilitate csw d_fdopen(). 4313 * 4314 * Dupfdopen() handling in kern_openat() installs the 4315 * half-baked file into the process descriptor table, allowing 4316 * other thread to dereference it. Guard against the race by 4317 * checking f_ops. 4318 */ 4319 if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) { 4320 fdrop(fp, td); 4321 *fpp = NULL; 4322 return (EINVAL); 4323 } 4324 4325 *fpp = fp; 4326 return (0); 4327 } 4328 4329 /* 4330 * Convert a user file descriptor to a kernel file entry and check 4331 * that, if it is a capability, the correct rights are present. 4332 * A reference on the file entry is held upon returning. 4333 */ 4334 int 4335 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp) 4336 { 4337 int error; 4338 4339 error = getvnode_path(td, fd, rightsp, fpp); 4340 if (__predict_false(error != 0)) 4341 return (error); 4342 4343 /* 4344 * Filter out O_PATH file descriptors, most getvnode() callers 4345 * do not call fo_ methods. 4346 */ 4347 if (__predict_false((*fpp)->f_ops == &path_fileops)) { 4348 fdrop(*fpp, td); 4349 *fpp = NULL; 4350 error = EBADF; 4351 } 4352 4353 return (error); 4354 } 4355 4356 /* 4357 * Get an (NFS) file handle. 4358 */ 4359 #ifndef _SYS_SYSPROTO_H_ 4360 struct lgetfh_args { 4361 char *fname; 4362 fhandle_t *fhp; 4363 }; 4364 #endif 4365 int 4366 sys_lgetfh(struct thread *td, struct lgetfh_args *uap) 4367 { 4368 4369 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname, 4370 UIO_USERSPACE, uap->fhp, UIO_USERSPACE)); 4371 } 4372 4373 #ifndef _SYS_SYSPROTO_H_ 4374 struct getfh_args { 4375 char *fname; 4376 fhandle_t *fhp; 4377 }; 4378 #endif 4379 int 4380 sys_getfh(struct thread *td, struct getfh_args *uap) 4381 { 4382 4383 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE, 4384 uap->fhp, UIO_USERSPACE)); 4385 } 4386 4387 /* 4388 * syscall for the rpc.lockd to use to translate an open descriptor into 4389 * a NFS file handle. 4390 * 4391 * warning: do not remove the priv_check() call or this becomes one giant 4392 * security hole. 4393 */ 4394 #ifndef _SYS_SYSPROTO_H_ 4395 struct getfhat_args { 4396 int fd; 4397 char *path; 4398 fhandle_t *fhp; 4399 int flags; 4400 }; 4401 #endif 4402 int 4403 sys_getfhat(struct thread *td, struct getfhat_args *uap) 4404 { 4405 4406 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE, 4407 uap->fhp, UIO_USERSPACE)); 4408 } 4409 4410 int 4411 kern_getfhat(struct thread *td, int flags, int fd, const char *path, 4412 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg) 4413 { 4414 struct nameidata nd; 4415 fhandle_t fh; 4416 struct vnode *vp; 4417 int error; 4418 4419 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0) 4420 return (EINVAL); 4421 error = priv_check(td, PRIV_VFS_GETFH); 4422 if (error != 0) 4423 return (error); 4424 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW | 4425 AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path, 4426 fd); 4427 error = namei(&nd); 4428 if (error != 0) 4429 return (error); 4430 NDFREE_PNBUF(&nd); 4431 vp = nd.ni_vp; 4432 bzero(&fh, sizeof(fh)); 4433 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 4434 error = VOP_VPTOFH(vp, &fh.fh_fid); 4435 vput(vp); 4436 if (error == 0) { 4437 if (fhseg == UIO_USERSPACE) 4438 error = copyout(&fh, fhp, sizeof (fh)); 4439 else 4440 memcpy(fhp, &fh, sizeof(fh)); 4441 } 4442 return (error); 4443 } 4444 4445 #ifndef _SYS_SYSPROTO_H_ 4446 struct fhlink_args { 4447 fhandle_t *fhp; 4448 const char *to; 4449 }; 4450 #endif 4451 int 4452 sys_fhlink(struct thread *td, struct fhlink_args *uap) 4453 { 4454 4455 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp)); 4456 } 4457 4458 #ifndef _SYS_SYSPROTO_H_ 4459 struct fhlinkat_args { 4460 fhandle_t *fhp; 4461 int tofd; 4462 const char *to; 4463 }; 4464 #endif 4465 int 4466 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap) 4467 { 4468 4469 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp)); 4470 } 4471 4472 static int 4473 kern_fhlinkat(struct thread *td, int fd, const char *path, 4474 enum uio_seg pathseg, fhandle_t *fhp) 4475 { 4476 fhandle_t fh; 4477 struct mount *mp; 4478 struct vnode *vp; 4479 int error; 4480 4481 error = priv_check(td, PRIV_VFS_GETFH); 4482 if (error != 0) 4483 return (error); 4484 error = copyin(fhp, &fh, sizeof(fh)); 4485 if (error != 0) 4486 return (error); 4487 do { 4488 bwillwrite(); 4489 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4490 return (ESTALE); 4491 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp); 4492 vfs_unbusy(mp); 4493 if (error != 0) 4494 return (error); 4495 VOP_UNLOCK(vp); 4496 error = kern_linkat_vp(td, vp, fd, path, pathseg); 4497 } while (error == EAGAIN || error == ERELOOKUP); 4498 return (error); 4499 } 4500 4501 #ifndef _SYS_SYSPROTO_H_ 4502 struct fhreadlink_args { 4503 fhandle_t *fhp; 4504 char *buf; 4505 size_t bufsize; 4506 }; 4507 #endif 4508 int 4509 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap) 4510 { 4511 fhandle_t fh; 4512 struct mount *mp; 4513 struct vnode *vp; 4514 int error; 4515 4516 error = priv_check(td, PRIV_VFS_GETFH); 4517 if (error != 0) 4518 return (error); 4519 if (uap->bufsize > IOSIZE_MAX) 4520 return (EINVAL); 4521 error = copyin(uap->fhp, &fh, sizeof(fh)); 4522 if (error != 0) 4523 return (error); 4524 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4525 return (ESTALE); 4526 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp); 4527 vfs_unbusy(mp); 4528 if (error != 0) 4529 return (error); 4530 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td); 4531 vput(vp); 4532 return (error); 4533 } 4534 4535 /* 4536 * syscall for the rpc.lockd to use to translate a NFS file handle into an 4537 * open descriptor. 4538 * 4539 * warning: do not remove the priv_check() call or this becomes one giant 4540 * security hole. 4541 */ 4542 #ifndef _SYS_SYSPROTO_H_ 4543 struct fhopen_args { 4544 const struct fhandle *u_fhp; 4545 int flags; 4546 }; 4547 #endif 4548 int 4549 sys_fhopen(struct thread *td, struct fhopen_args *uap) 4550 { 4551 return (kern_fhopen(td, uap->u_fhp, uap->flags)); 4552 } 4553 4554 int 4555 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags) 4556 { 4557 struct mount *mp; 4558 struct vnode *vp; 4559 struct fhandle fhp; 4560 struct file *fp; 4561 int fmode, error; 4562 int indx; 4563 4564 error = priv_check(td, PRIV_VFS_FHOPEN); 4565 if (error != 0) 4566 return (error); 4567 indx = -1; 4568 fmode = FFLAGS(flags); 4569 /* why not allow a non-read/write open for our lockd? */ 4570 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT)) 4571 return (EINVAL); 4572 error = copyin(u_fhp, &fhp, sizeof(fhp)); 4573 if (error != 0) 4574 return(error); 4575 /* find the mount point */ 4576 mp = vfs_busyfs(&fhp.fh_fsid); 4577 if (mp == NULL) 4578 return (ESTALE); 4579 /* now give me my vnode, it gets returned to me locked */ 4580 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp); 4581 vfs_unbusy(mp); 4582 if (error != 0) 4583 return (error); 4584 4585 error = falloc_noinstall(td, &fp); 4586 if (error != 0) { 4587 vput(vp); 4588 return (error); 4589 } 4590 /* 4591 * An extra reference on `fp' has been held for us by 4592 * falloc_noinstall(). 4593 */ 4594 4595 #ifdef INVARIANTS 4596 td->td_dupfd = -1; 4597 #endif 4598 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp); 4599 if (error != 0) { 4600 KASSERT(fp->f_ops == &badfileops, 4601 ("VOP_OPEN in fhopen() set f_ops")); 4602 KASSERT(td->td_dupfd < 0, 4603 ("fhopen() encountered fdopen()")); 4604 4605 vput(vp); 4606 goto bad; 4607 } 4608 #ifdef INVARIANTS 4609 td->td_dupfd = 0; 4610 #endif 4611 fp->f_vnode = vp; 4612 finit_vnode(fp, fmode, NULL, &vnops); 4613 VOP_UNLOCK(vp); 4614 if ((fmode & O_TRUNC) != 0) { 4615 error = fo_truncate(fp, 0, td->td_ucred, td); 4616 if (error != 0) 4617 goto bad; 4618 } 4619 4620 error = finstall(td, fp, &indx, fmode, NULL); 4621 bad: 4622 fdrop(fp, td); 4623 td->td_retval[0] = indx; 4624 return (error); 4625 } 4626 4627 /* 4628 * Stat an (NFS) file handle. 4629 */ 4630 #ifndef _SYS_SYSPROTO_H_ 4631 struct fhstat_args { 4632 struct fhandle *u_fhp; 4633 struct stat *sb; 4634 }; 4635 #endif 4636 int 4637 sys_fhstat(struct thread *td, struct fhstat_args *uap) 4638 { 4639 struct stat sb; 4640 struct fhandle fh; 4641 int error; 4642 4643 error = copyin(uap->u_fhp, &fh, sizeof(fh)); 4644 if (error != 0) 4645 return (error); 4646 error = kern_fhstat(td, fh, &sb); 4647 if (error == 0) 4648 error = copyout(&sb, uap->sb, sizeof(sb)); 4649 return (error); 4650 } 4651 4652 int 4653 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb) 4654 { 4655 struct mount *mp; 4656 struct vnode *vp; 4657 int error; 4658 4659 error = priv_check(td, PRIV_VFS_FHSTAT); 4660 if (error != 0) 4661 return (error); 4662 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4663 return (ESTALE); 4664 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4665 vfs_unbusy(mp); 4666 if (error != 0) 4667 return (error); 4668 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED); 4669 vput(vp); 4670 return (error); 4671 } 4672 4673 /* 4674 * Implement fstatfs() for (NFS) file handles. 4675 */ 4676 #ifndef _SYS_SYSPROTO_H_ 4677 struct fhstatfs_args { 4678 struct fhandle *u_fhp; 4679 struct statfs *buf; 4680 }; 4681 #endif 4682 int 4683 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap) 4684 { 4685 struct statfs *sfp; 4686 fhandle_t fh; 4687 int error; 4688 4689 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t)); 4690 if (error != 0) 4691 return (error); 4692 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 4693 error = kern_fhstatfs(td, fh, sfp); 4694 if (error == 0) 4695 error = copyout(sfp, uap->buf, sizeof(*sfp)); 4696 free(sfp, M_STATFS); 4697 return (error); 4698 } 4699 4700 int 4701 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf) 4702 { 4703 struct mount *mp; 4704 struct vnode *vp; 4705 int error; 4706 4707 error = priv_check(td, PRIV_VFS_FHSTATFS); 4708 if (error != 0) 4709 return (error); 4710 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL) 4711 return (ESTALE); 4712 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp); 4713 if (error != 0) { 4714 vfs_unbusy(mp); 4715 return (error); 4716 } 4717 vput(vp); 4718 error = prison_canseemount(td->td_ucred, mp); 4719 if (error != 0) 4720 goto out; 4721 #ifdef MAC 4722 error = mac_mount_check_stat(td->td_ucred, mp); 4723 if (error != 0) 4724 goto out; 4725 #endif 4726 error = VFS_STATFS(mp, buf); 4727 out: 4728 vfs_unbusy(mp); 4729 return (error); 4730 } 4731 4732 /* 4733 * Unlike madvise(2), we do not make a best effort to remember every 4734 * possible caching hint. Instead, we remember the last setting with 4735 * the exception that we will allow POSIX_FADV_NORMAL to adjust the 4736 * region of any current setting. 4737 */ 4738 int 4739 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len, 4740 int advice) 4741 { 4742 struct fadvise_info *fa, *new; 4743 struct file *fp; 4744 struct vnode *vp; 4745 off_t end; 4746 int error; 4747 4748 if (offset < 0 || len < 0 || offset > OFF_MAX - len) 4749 return (EINVAL); 4750 AUDIT_ARG_VALUE(advice); 4751 switch (advice) { 4752 case POSIX_FADV_SEQUENTIAL: 4753 case POSIX_FADV_RANDOM: 4754 case POSIX_FADV_NOREUSE: 4755 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK); 4756 break; 4757 case POSIX_FADV_NORMAL: 4758 case POSIX_FADV_WILLNEED: 4759 case POSIX_FADV_DONTNEED: 4760 new = NULL; 4761 break; 4762 default: 4763 return (EINVAL); 4764 } 4765 /* XXX: CAP_POSIX_FADVISE? */ 4766 AUDIT_ARG_FD(fd); 4767 error = fget(td, fd, &cap_no_rights, &fp); 4768 if (error != 0) 4769 goto out; 4770 AUDIT_ARG_FILE(td->td_proc, fp); 4771 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) { 4772 error = ESPIPE; 4773 goto out; 4774 } 4775 if (fp->f_type != DTYPE_VNODE) { 4776 error = ENODEV; 4777 goto out; 4778 } 4779 vp = fp->f_vnode; 4780 if (vp->v_type != VREG) { 4781 error = ENODEV; 4782 goto out; 4783 } 4784 if (len == 0) 4785 end = OFF_MAX; 4786 else 4787 end = offset + len - 1; 4788 switch (advice) { 4789 case POSIX_FADV_SEQUENTIAL: 4790 case POSIX_FADV_RANDOM: 4791 case POSIX_FADV_NOREUSE: 4792 /* 4793 * Try to merge any existing non-standard region with 4794 * this new region if possible, otherwise create a new 4795 * non-standard region for this request. 4796 */ 4797 mtx_pool_lock(mtxpool_sleep, fp); 4798 fa = fp->f_advice; 4799 if (fa != NULL && fa->fa_advice == advice && 4800 ((fa->fa_start <= end && fa->fa_end >= offset) || 4801 (end != OFF_MAX && fa->fa_start == end + 1) || 4802 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) { 4803 if (offset < fa->fa_start) 4804 fa->fa_start = offset; 4805 if (end > fa->fa_end) 4806 fa->fa_end = end; 4807 } else { 4808 new->fa_advice = advice; 4809 new->fa_start = offset; 4810 new->fa_end = end; 4811 fp->f_advice = new; 4812 new = fa; 4813 } 4814 mtx_pool_unlock(mtxpool_sleep, fp); 4815 break; 4816 case POSIX_FADV_NORMAL: 4817 /* 4818 * If a the "normal" region overlaps with an existing 4819 * non-standard region, trim or remove the 4820 * non-standard region. 4821 */ 4822 mtx_pool_lock(mtxpool_sleep, fp); 4823 fa = fp->f_advice; 4824 if (fa != NULL) { 4825 if (offset <= fa->fa_start && end >= fa->fa_end) { 4826 new = fa; 4827 fp->f_advice = NULL; 4828 } else if (offset <= fa->fa_start && 4829 end >= fa->fa_start) 4830 fa->fa_start = end + 1; 4831 else if (offset <= fa->fa_end && end >= fa->fa_end) 4832 fa->fa_end = offset - 1; 4833 else if (offset >= fa->fa_start && end <= fa->fa_end) { 4834 /* 4835 * If the "normal" region is a middle 4836 * portion of the existing 4837 * non-standard region, just remove 4838 * the whole thing rather than picking 4839 * one side or the other to 4840 * preserve. 4841 */ 4842 new = fa; 4843 fp->f_advice = NULL; 4844 } 4845 } 4846 mtx_pool_unlock(mtxpool_sleep, fp); 4847 break; 4848 case POSIX_FADV_WILLNEED: 4849 case POSIX_FADV_DONTNEED: 4850 error = VOP_ADVISE(vp, offset, end, advice); 4851 break; 4852 } 4853 out: 4854 if (fp != NULL) 4855 fdrop(fp, td); 4856 free(new, M_FADVISE); 4857 return (error); 4858 } 4859 4860 int 4861 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap) 4862 { 4863 int error; 4864 4865 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len, 4866 uap->advice); 4867 return (kern_posix_error(td, error)); 4868 } 4869 4870 int 4871 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd, 4872 off_t *outoffp, size_t len, unsigned int flags) 4873 { 4874 struct file *infp, *outfp; 4875 struct vnode *invp, *outvp; 4876 int error; 4877 size_t retlen; 4878 void *rl_rcookie, *rl_wcookie; 4879 off_t savinoff, savoutoff; 4880 4881 infp = outfp = NULL; 4882 rl_rcookie = rl_wcookie = NULL; 4883 savinoff = -1; 4884 error = 0; 4885 retlen = 0; 4886 4887 if (flags != 0) { 4888 error = EINVAL; 4889 goto out; 4890 } 4891 if (len > SSIZE_MAX) 4892 /* 4893 * Although the len argument is size_t, the return argument 4894 * is ssize_t (which is signed). Therefore a size that won't 4895 * fit in ssize_t can't be returned. 4896 */ 4897 len = SSIZE_MAX; 4898 4899 /* Get the file structures for the file descriptors. */ 4900 error = fget_read(td, infd, &cap_read_rights, &infp); 4901 if (error != 0) 4902 goto out; 4903 if (infp->f_ops == &badfileops) { 4904 error = EBADF; 4905 goto out; 4906 } 4907 if (infp->f_vnode == NULL) { 4908 error = EINVAL; 4909 goto out; 4910 } 4911 error = fget_write(td, outfd, &cap_write_rights, &outfp); 4912 if (error != 0) 4913 goto out; 4914 if (outfp->f_ops == &badfileops) { 4915 error = EBADF; 4916 goto out; 4917 } 4918 if (outfp->f_vnode == NULL) { 4919 error = EINVAL; 4920 goto out; 4921 } 4922 4923 /* Set the offset pointers to the correct place. */ 4924 if (inoffp == NULL) 4925 inoffp = &infp->f_offset; 4926 if (outoffp == NULL) 4927 outoffp = &outfp->f_offset; 4928 savinoff = *inoffp; 4929 savoutoff = *outoffp; 4930 4931 invp = infp->f_vnode; 4932 outvp = outfp->f_vnode; 4933 /* Sanity check the f_flag bits. */ 4934 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE || 4935 (infp->f_flag & FREAD) == 0) { 4936 error = EBADF; 4937 goto out; 4938 } 4939 4940 /* If len == 0, just return 0. */ 4941 if (len == 0) 4942 goto out; 4943 4944 /* 4945 * If infp and outfp refer to the same file, the byte ranges cannot 4946 * overlap. 4947 */ 4948 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len > 4949 savoutoff) || (savinoff > savoutoff && savoutoff + len > 4950 savinoff))) { 4951 error = EINVAL; 4952 goto out; 4953 } 4954 4955 /* Range lock the byte ranges for both invp and outvp. */ 4956 for (;;) { 4957 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp + 4958 len); 4959 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp + 4960 len); 4961 if (rl_rcookie != NULL) 4962 break; 4963 vn_rangelock_unlock(outvp, rl_wcookie); 4964 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len); 4965 vn_rangelock_unlock(invp, rl_rcookie); 4966 } 4967 4968 retlen = len; 4969 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen, 4970 flags, infp->f_cred, outfp->f_cred, td); 4971 out: 4972 if (rl_rcookie != NULL) 4973 vn_rangelock_unlock(invp, rl_rcookie); 4974 if (rl_wcookie != NULL) 4975 vn_rangelock_unlock(outvp, rl_wcookie); 4976 if (savinoff != -1 && (error == EINTR || error == ERESTART)) { 4977 *inoffp = savinoff; 4978 *outoffp = savoutoff; 4979 } 4980 if (outfp != NULL) 4981 fdrop(outfp, td); 4982 if (infp != NULL) 4983 fdrop(infp, td); 4984 td->td_retval[0] = retlen; 4985 return (error); 4986 } 4987 4988 int 4989 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap) 4990 { 4991 off_t inoff, outoff, *inoffp, *outoffp; 4992 int error; 4993 4994 inoffp = outoffp = NULL; 4995 if (uap->inoffp != NULL) { 4996 error = copyin(uap->inoffp, &inoff, sizeof(off_t)); 4997 if (error != 0) 4998 return (error); 4999 inoffp = &inoff; 5000 } 5001 if (uap->outoffp != NULL) { 5002 error = copyin(uap->outoffp, &outoff, sizeof(off_t)); 5003 if (error != 0) 5004 return (error); 5005 outoffp = &outoff; 5006 } 5007 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd, 5008 outoffp, uap->len, uap->flags); 5009 if (error == 0 && uap->inoffp != NULL) 5010 error = copyout(inoffp, uap->inoffp, sizeof(off_t)); 5011 if (error == 0 && uap->outoffp != NULL) 5012 error = copyout(outoffp, uap->outoffp, sizeof(off_t)); 5013 return (error); 5014 } 5015